Fix old comment about HyperClockCache and SecondaryCache (#11517)

Summary:
Support was added in 8.1.0

Pull Request resolved: https://github.com/facebook/rocksdb/pull/11517

Test Plan: comments only

Reviewed By: anand1976

Differential Revision: D46489929

Pulled By: pdillinger

fbshipit-source-id: 4fd30078389065c9ec225bf55b6773f1641f0646

.circleci/config.yml

@@ -1,7 +1,7 @@
 version: 2.1
 
 orbs:
-  win: circleci/windows@2.4.0
+  win: circleci/windows@5.0.0
 
 commands:
   install-cmake-on-macos:
@@ -45,6 +45,38 @@ commands:
             echo "export LZ4_DOWNLOAD_BASE=https://rocksdb-deps.s3.us-west-2.amazonaws.com/pkgs/lz4" >> $BASH_ENV
             echo "export ZSTD_DOWNLOAD_BASE=https://rocksdb-deps.s3.us-west-2.amazonaws.com/pkgs/zstd" >> $BASH_ENV
 
+  windows-build-steps:
+    steps:
+      - checkout
+      - run:
+          name: "Install thirdparty dependencies"
+          command: |
+            echo "Installing CMake..."
+            choco install cmake --installargs 'ADD_CMAKE_TO_PATH=System' -y
+            mkdir $Env:THIRDPARTY_HOME
+            cd $Env:THIRDPARTY_HOME
+            echo "Building Snappy dependency..."
+            curl https://github.com/google/snappy/archive/refs/tags/1.1.8.zip -O snappy-1.1.8.zip
+            unzip -q snappy-1.1.8.zip
+            cd snappy-1.1.8
+            mkdir build
+            cd build
+            & $Env:CMAKE_BIN -G "$Env:CMAKE_GENERATOR" ..
+            msbuild.exe Snappy.sln -maxCpuCount -property:Configuration=Debug -property:Platform=x64
+      - run:
+          name: "Build RocksDB"
+          command: |
+            mkdir build
+            cd build
+            & $Env:CMAKE_BIN -G "$Env:CMAKE_GENERATOR" -DCMAKE_BUILD_TYPE=Debug -DOPTDBG=1 -DPORTABLE=1 -DSNAPPY=1 -DJNI=1 ..
+            cd ..
+            echo "Building with VS version: $Env:CMAKE_GENERATOR"
+            msbuild.exe build/rocksdb.sln -maxCpuCount -property:Configuration=Debug -property:Platform=x64
+      - run:
+          name: "Test RocksDB"
+          shell: powershell.exe
+          command: |
+            build_tools\run_ci_db_test.ps1 -SuiteRun arena_test,db_basic_test,db_test,db_test2,db_merge_operand_test,bloom_test,c_test,coding_test,crc32c_test,dynamic_bloom_test,env_basic_test,env_test,hash_test,random_test -Concurrency 16
   pre-steps-macos:
       steps:
         - pre-steps
@@ -120,15 +152,15 @@ commands:
     steps:
       - run:
           name: "Test low-variance benchmarks"
-          command: ./tools/benchmark_ci.py --db_dir /tmp/rocksdb-benchmark-datadir --output_dir /tmp/benchmark-results --num_keys 10000000
+          command: ./tools/benchmark_ci.py --db_dir /tmp/rocksdb-benchmark-datadir --output_dir /tmp/benchmark-results --num_keys 20000000
           environment:
             LD_LIBRARY_PATH: /usr/local/lib
             # How long to run parts of the test(s)
-            DURATION_RO: 400
-            DURATION_RW: 700
+            DURATION_RO: 300
+            DURATION_RW: 500
             # Keep threads within physical capacity of server (much lower than default)
             NUM_THREADS: 1
-            MAX_BACKGROUND_JOBS: 3
+            MAX_BACKGROUND_JOBS: 4
             # Don't run a couple of "optional" initial tests
             CI_TESTS_ONLY: "true"
             # Reduce configured size of levels to ensure more levels in the leveled compaction LSM tree
@@ -138,7 +170,11 @@ commands:
             # The benchmark host has 32GB memory
             # The following values are tailored to work with that
             # Note, tests may not exercise the targeted issues if the memory is increased on new test hosts.
-
+            COMPRESSION_TYPE: "none"
+            CACHE_INDEX_AND_FILTER_BLOCKS: 1
+            MIN_LEVEL_TO_COMPRESS: 3
+            CACHE_SIZE_MB: 10240
+            MB_WRITE_PER_SEC: 2
 
   post-benchmarks:
     steps:
@@ -154,11 +190,6 @@ commands:
             true
 
 executors:
-  windows-2xlarge:
-    machine:
-      image: 'windows-server-2019-vs2019:stable'
-      resource_class: windows.2xlarge
-      shell: bash.exe
   linux-docker:
     docker:
       # The image configuration is build_tools/ubuntu20_image/Dockerfile
@@ -242,12 +273,12 @@ jobs:
           ./sst_dump --help | grep -E -q 'Supported compression types: kNoCompression$' # Verify no compiled in compression
       - post-steps
 
-  build-linux-shared_lib-alt_namespace-status_checked:
+  build-linux-static_lib-alt_namespace-status_checked:
     executor: linux-docker
     resource_class: 2xlarge
     steps:
       - pre-steps
-      - run: ASSERT_STATUS_CHECKED=1 TEST_UINT128_COMPAT=1 ROCKSDB_MODIFY_NPHASH=1 LIB_MODE=shared OPT="-DROCKSDB_NAMESPACE=alternative_rocksdb_ns" make V=1 -j32 check
+      - run: ASSERT_STATUS_CHECKED=1 TEST_UINT128_COMPAT=1 ROCKSDB_MODIFY_NPHASH=1 LIB_MODE=static OPT="-DROCKSDB_NAMESPACE=alternative_rocksdb_ns" make V=1 -j24 check
       - post-steps
 
   build-linux-release:
@@ -255,11 +286,21 @@ jobs:
     resource_class: 2xlarge
     steps:
       - checkout # check out the code in the project directory
+      - run: make V=1 -j32 LIB_MODE=shared release
+      - run: ls librocksdb.so # ensure shared lib built
+      - run: ./db_stress --version # ensure with gflags
+      - run: make clean
       - run: make V=1 -j32 release
+      - run: ls librocksdb.a # ensure static lib built
       - run: ./db_stress --version # ensure with gflags
       - run: make clean
       - run: apt-get remove -y libgflags-dev
+      - run: make V=1 -j32 LIB_MODE=shared release
+      - run: ls librocksdb.so # ensure shared lib built
+      - run: if ./db_stress --version; then false; else true; fi # ensure without gflags
+      - run: make clean
       - run: make V=1 -j32 release
+      - run: ls librocksdb.a # ensure static lib built
       - run: if ./db_stress --version; then false; else true; fi # ensure without gflags
       - post-steps
 
@@ -275,27 +316,6 @@ jobs:
       - run: USE_RTTI=1 DEBUG_LEVEL=0 make V=1 -j16 static_lib tools db_bench
       - run: if ./db_stress --version; then false; else true; fi # ensure without gflags
 
-  build-linux-lite:
-    executor: linux-docker
-    resource_class: large
-    steps:
-      - pre-steps
-      - run: LITE=1 make V=1 J=8 -j8 check
-      - post-steps
-
-  build-linux-lite-release:
-    executor: linux-docker
-    resource_class: large
-    steps:
-      - checkout # check out the code in the project directory
-      - run: LITE=1 make V=1 -j8 release
-      - run: ./db_stress --version # ensure with gflags
-      - run: make clean
-      - run: apt-get remove -y libgflags-dev
-      - run: LITE=1 make V=1 -j8 release
-      - run: if ./db_stress --version; then false; else true; fi # ensure without gflags
-      - post-steps
-
   build-linux-clang-no_test_run:
     executor: linux-docker
     resource_class: xlarge
@@ -400,7 +420,10 @@ jobs:
     steps:
       - checkout # check out the code in the project directory
       - run: apt-get update -y && apt-get install -y libgflags-dev
-      - run: make V=1 -j8 unity_test
+      - run:
+          name: "Unity build"
+          command: make V=1 -j8 unity_test
+          no_output_timeout: 20m
       - run: make V=1 -j8 -k check-headers # could be moved to a different build
       - post-steps
 
@@ -411,7 +434,7 @@ jobs:
       - pre-steps
       - setup-folly
       - build-folly
-      - run: USE_FOLLY=1 CC=gcc-7 CXX=g++-7 V=1 make -j32 check
+      - run: USE_FOLLY=1 LIB_MODE=static CC=gcc-7 CXX=g++-7 V=1 make -j32 check # TODO: LIB_MODE only to work around unresolved linker failures
       - post-steps
 
   build-linux-gcc-7-with-folly-lite-no-test:
@@ -457,7 +480,7 @@ jobs:
     resource_class: 2xlarge
     steps:
       - pre-steps
-      - run: CC=gcc-11 CXX=g++-11 V=1 make -j32 all microbench
+      - run: LIB_MODE=static CC=gcc-11 CXX=g++-11 V=1 make -j32 all microbench # TODO: LIB_MODE only to work around unresolved linker failures
       - post-steps
 
   build-linux-clang-13-no_test_run:
@@ -476,7 +499,7 @@ jobs:
       - pre-steps
       - setup-folly
       - build-folly
-      - run: CC=clang-13 CXX=clang++-13 USE_CLANG=1 USE_FOLLY=1 COMPILE_WITH_UBSAN=1 COMPILE_WITH_ASAN=1 make -j32 check
+      - run: CC=clang-13 CXX=clang++-13 LIB_MODE=static USE_CLANG=1 USE_FOLLY=1 COMPILE_WITH_UBSAN=1 COMPILE_WITH_ASAN=1 make -j32 check # TODO: LIB_MODE only to work around unresolved linker failures
       - post-steps
 
   # This job is only to make sure the microbench tests are able to run, the benchmark result is not meaningful as the CI host is changing.
@@ -493,7 +516,7 @@ jobs:
     resource_class: large
     steps:
       - pre-steps
-      - run: ulimit -S -n `ulimit -H -n` && make V=1 -j8 CRASH_TEST_EXT_ARGS='--duration=960 --max_key=2500000' blackbox_crash_test_with_atomic_flush
+      - run: ulimit -S -n `ulimit -H -n` && make V=1 -j8 CRASH_TEST_EXT_ARGS='--duration=960 --max_key=2500000 --use_io_uring=0' blackbox_crash_test_with_atomic_flush
       - post-steps
 
   build-linux-crashtest-tiered-storage-bb:
@@ -503,7 +526,7 @@ jobs:
       - pre-steps
       - run:
           name: "run crashtest"
-          command: ulimit -S -n `ulimit -H -n` && make V=1 -j32 CRASH_TEST_EXT_ARGS=--duration=10800 blackbox_crash_test_with_tiered_storage
+          command: ulimit -S -n `ulimit -H -n` && make V=1 -j32 CRASH_TEST_EXT_ARGS='--duration=10800 --use_io_uring=0' blackbox_crash_test_with_tiered_storage
           no_output_timeout: 100m
       - post-steps
 
@@ -514,77 +537,39 @@ jobs:
       - pre-steps
       - run:
           name: "run crashtest"
-          command: ulimit -S -n `ulimit -H -n` && make V=1 -j32 CRASH_TEST_EXT_ARGS=--duration=10800 whitebox_crash_test_with_tiered_storage
+          command: ulimit -S -n `ulimit -H -n` && make V=1 -j32 CRASH_TEST_EXT_ARGS='--duration=10800 --use_io_uring=0' whitebox_crash_test_with_tiered_storage
           no_output_timeout: 100m
       - post-steps
 
-  build-windows:
-    executor: windows-2xlarge
-    parameters:
-      extra_cmake_opt:
-        default: ""
-        type: string
-      vs_year:
-        default: "2019"
-        type: string
-      cmake_generator:
-        default: "Visual Studio 16 2019"
-        type: string
+  build-windows-vs2022:
+    executor:
+      name: win/server-2022
+      size: 2xlarge
     environment:
       THIRDPARTY_HOME: C:/Users/circleci/thirdparty
-      CMAKE_HOME: C:/Users/circleci/thirdparty/cmake-3.16.4-win64-x64
-      CMAKE_BIN: C:/Users/circleci/thirdparty/cmake-3.16.4-win64-x64/bin/cmake.exe
-      SNAPPY_HOME: C:/Users/circleci/thirdparty/snappy-1.1.7
-      SNAPPY_INCLUDE: C:/Users/circleci/thirdparty/snappy-1.1.7;C:/Users/circleci/thirdparty/snappy-1.1.7/build
-      SNAPPY_LIB_DEBUG: C:/Users/circleci/thirdparty/snappy-1.1.7/build/Debug/snappy.lib
-      VS_YEAR: <<parameters.vs_year>>
-      CMAKE_GENERATOR: <<parameters.cmake_generator>>
+      CMAKE_HOME: C:/Program Files/CMake
+      CMAKE_BIN: C:/Program Files/CMake/bin/cmake.exe
+      SNAPPY_HOME: C:/Users/circleci/thirdparty/snappy-1.1.8
+      SNAPPY_INCLUDE: C:/Users/circleci/thirdparty/snappy-1.1.8;C:/Users/circleci/thirdparty/snappy-1.1.8/build
+      SNAPPY_LIB_DEBUG: C:/Users/circleci/thirdparty/snappy-1.1.8/build/Debug/snappy.lib
+      CMAKE_GENERATOR: Visual Studio 17 2022
     steps:
-      - checkout
-      - run:
-          name: "Setup VS"
-          command: |
-            if [[ "${VS_YEAR}" == "2019" ]]; then
-              echo "VS2019 already present."
-            elif [[ "${VS_YEAR}" == "2017" ]]; then
-              echo "Installing VS2017..."
-              powershell .circleci/vs2017_install.ps1
-            elif [[ "${VS_YEAR}" == "2015" ]]; then
-              echo "Installing VS2015..."
-              powershell .circleci/vs2015_install.ps1
-            fi
-      - store_artifacts:
-          path: \Users\circleci\AppData\Local\Temp\vslogs.zip
-      - run:
-          name: "Install thirdparty dependencies"
-          command: |
-            mkdir ${THIRDPARTY_HOME}
-            cd ${THIRDPARTY_HOME}
-            echo "Installing CMake..."
-            curl --fail --silent --show-error --output cmake-3.16.4-win64-x64.zip --location https://github.com/Kitware/CMake/releases/download/v3.16.4/cmake-3.16.4-win64-x64.zip
-            unzip -q cmake-3.16.4-win64-x64.zip
-            echo "Building Snappy dependency..."
-            curl --fail --silent --show-error --output snappy-1.1.7.zip --location https://github.com/google/snappy/archive/1.1.7.zip
-            unzip -q snappy-1.1.7.zip
-            cd snappy-1.1.7
-            mkdir build
-            cd build
-            ${CMAKE_BIN} -G "${CMAKE_GENERATOR}" ..
-            msbuild.exe Snappy.sln -maxCpuCount -property:Configuration=Debug -property:Platform=x64
-      - run:
-          name: "Build RocksDB"
-          command: |
-            mkdir build
-            cd build
-            ${CMAKE_BIN} -G "${CMAKE_GENERATOR}" -DCMAKE_BUILD_TYPE=Debug -DOPTDBG=1 -DPORTABLE=1 -DSNAPPY=1 -DJNI=1 << parameters.extra_cmake_opt >> ..
-            cd ..
-            echo "Building with VS version: ${CMAKE_GENERATOR}"
-            msbuild.exe build/rocksdb.sln -maxCpuCount -property:Configuration=Debug -property:Platform=x64
-      - run:
-          name: "Test RocksDB"
-          shell: powershell.exe
-          command: |
-            build_tools\run_ci_db_test.ps1 -SuiteRun db_basic_test,db_test,db_test2,db_merge_operand_test,bloom_test,c_test,coding_test,crc32c_test,dynamic_bloom_test,env_basic_test,env_test,hash_test,random_test -Concurrency 16
+      - windows-build-steps
+
+  build-windows-vs2019:
+    executor:
+      name: win/server-2019
+      size: 2xlarge
+    environment:
+      THIRDPARTY_HOME: C:/Users/circleci/thirdparty
+      CMAKE_HOME: C:/Program Files/CMake
+      CMAKE_BIN: C:/Program Files/CMake/bin/cmake.exe
+      SNAPPY_HOME: C:/Users/circleci/thirdparty/snappy-1.1.8
+      SNAPPY_INCLUDE: C:/Users/circleci/thirdparty/snappy-1.1.8;C:/Users/circleci/thirdparty/snappy-1.1.8/build
+      SNAPPY_LIB_DEBUG: C:/Users/circleci/thirdparty/snappy-1.1.8/build/Debug/snappy.lib
+      CMAKE_GENERATOR: Visual Studio 16 2019
+    steps:
+      - windows-build-steps
 
   build-linux-java:
     executor: linux-docker
@@ -832,18 +817,16 @@ workflows:
       - build-linux-cmake-with-folly-coroutines
       - build-linux-cmake-with-benchmark
       - build-linux-encrypted_env-no_compression
-      - build-linux-lite
   jobs-linux-run-tests-san:
     jobs:
       - build-linux-clang10-asan
       - build-linux-clang10-ubsan
       - build-linux-clang10-mini-tsan
-      - build-linux-shared_lib-alt_namespace-status_checked
+      - build-linux-static_lib-alt_namespace-status_checked
   jobs-linux-no-test-run:
     jobs:
       - build-linux-release
       - build-linux-release-rtti
-      - build-linux-lite-release
       - build-examples
       - build-fuzzers
       - build-linux-clang-no_test_run
@@ -859,15 +842,8 @@ workflows:
       - build-linux-mini-crashtest
   jobs-windows:
     jobs:
-      - build-windows:
-          name: "build-windows-vs2019"
-      - build-windows:
-          name: "build-windows-vs2019-cxx20"
-          extra_cmake_opt: -DCMAKE_CXX_STANDARD=20
-      - build-windows:
-          name: "build-windows-vs2017"
-          vs_year: "2017"
-          cmake_generator: "Visual Studio 15 Win64"
+      - build-windows-vs2022
+      - build-windows-vs2019
       - build-cmake-mingw
   jobs-java:
     jobs:

.circleci/vs2015_install.ps1

@@ -1,24 +0,0 @@
-$VS_DOWNLOAD_LINK = "https://go.microsoft.com/fwlink/?LinkId=691126"
-$COLLECT_DOWNLOAD_LINK = "https://aka.ms/vscollect.exe"
-curl.exe --retry 3 -kL $VS_DOWNLOAD_LINK --output vs_installer.exe
-if ($LASTEXITCODE -ne 0) {
-    echo "Download of the VS 2015 installer failed"
-    exit 1
-}
-$VS_INSTALL_ARGS = @("/Quiet", "/NoRestart")
-$process = Start-Process "${PWD}\vs_installer.exe" -ArgumentList $VS_INSTALL_ARGS -NoNewWindow -Wait -PassThru
-Remove-Item -Path vs_installer.exe -Force
-$exitCode = $process.ExitCode
-if (($exitCode -ne 0) -and ($exitCode -ne 3010)) {
-    echo "VS 2015 installer exited with code $exitCode, which should be one of [0, 3010]."
-    curl.exe --retry 3 -kL $COLLECT_DOWNLOAD_LINK --output Collect.exe
-    if ($LASTEXITCODE -ne 0) {
-        echo "Download of the VS Collect tool failed."
-        exit 1
-    }
-    Start-Process "${PWD}\Collect.exe" -NoNewWindow -Wait -PassThru
-    New-Item -Path "C:\w\build-results" -ItemType "directory" -Force
-    Copy-Item -Path "C:\Users\circleci\AppData\Local\Temp\vslogs.zip" -Destination "C:\w\build-results\"
-    exit 1
-}
-echo "VS 2015 installed."

.circleci/vs2017_install.ps1

@@ -1,35 +0,0 @@
-$VS_DOWNLOAD_LINK = "https://aka.ms/vs/15/release/vs_buildtools.exe"
-$COLLECT_DOWNLOAD_LINK = "https://aka.ms/vscollect.exe"
-$VS_INSTALL_ARGS = @("--nocache","--quiet","--wait", "--add Microsoft.VisualStudio.Workload.VCTools",
-                                                     "--add Microsoft.VisualStudio.Component.VC.Tools.14.13",
-                                                     "--add Microsoft.Component.MSBuild",
-                                                     "--add Microsoft.VisualStudio.Component.Roslyn.Compiler",
-                                                     "--add Microsoft.VisualStudio.Component.TextTemplating",
-                                                     "--add Microsoft.VisualStudio.Component.VC.CoreIde",
-                                                     "--add Microsoft.VisualStudio.Component.VC.Redist.14.Latest",
-                                                     "--add Microsoft.VisualStudio.ComponentGroup.NativeDesktop.Core",
-                                                     "--add Microsoft.VisualStudio.Component.VC.Tools.x86.x64",
-                                                     "--add Microsoft.VisualStudio.ComponentGroup.NativeDesktop.Win81")
-
-curl.exe --retry 3 -kL $VS_DOWNLOAD_LINK --output vs_installer.exe
-if ($LASTEXITCODE -ne 0) {
-    echo "Download of the VS 2017 installer failed"
-    exit 1
-}
-
-$process = Start-Process "${PWD}\vs_installer.exe" -ArgumentList $VS_INSTALL_ARGS -NoNewWindow -Wait -PassThru
-Remove-Item -Path vs_installer.exe -Force
-$exitCode = $process.ExitCode
-if (($exitCode -ne 0) -and ($exitCode -ne 3010)) {
-    echo "VS 2017 installer exited with code $exitCode, which should be one of [0, 3010]."
-    curl.exe --retry 3 -kL $COLLECT_DOWNLOAD_LINK --output Collect.exe
-    if ($LASTEXITCODE -ne 0) {
-        echo "Download of the VS Collect tool failed."
-        exit 1
-    }
-    Start-Process "${PWD}\Collect.exe" -NoNewWindow -Wait -PassThru
-    New-Item -Path "C:\w\build-results" -ItemType "directory" -Force
-    Copy-Item -Path "C:\Users\circleci\AppData\Local\Temp\vslogs.zip" -Destination "C:\w\build-results\"
-    exit 1
-}
-echo "VS 2017 installed."

CMakeLists.txt

@@ -245,11 +245,21 @@ if(CMAKE_SYSTEM_PROCESSOR MATCHES "s390x")
   endif(HAS_S390X_MARCH_NATIVE)
 endif(CMAKE_SYSTEM_PROCESSOR MATCHES "s390x")
 
+if(CMAKE_SYSTEM_PROCESSOR MATCHES "loongarch64")
+  CHECK_C_COMPILER_FLAG("-march=loongarch64" HAS_LOONGARCH64)
+  if(HAS_LOONGARCH64)
+    set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mcpu=loongarch64 -mtune=loongarch64")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -mcpu=loongarch64 -mtune=loongarch64")
+  endif(HAS_LOONGARCH64)
+endif(CMAKE_SYSTEM_PROCESSOR MATCHES "loongarch64")
+
 option(PORTABLE "build a portable binary" OFF)
 option(FORCE_SSE42 "force building with SSE4.2, even when PORTABLE=ON" OFF)
 option(FORCE_AVX "force building with AVX, even when PORTABLE=ON" OFF)
 option(FORCE_AVX2 "force building with AVX2, even when PORTABLE=ON" OFF)
 if(PORTABLE)
+  add_definitions(-DROCKSDB_PORTABLE)
+
   # MSVC does not need a separate compiler flag to enable SSE4.2; if nmmintrin.h
   # is available, it is available by default.
   if(FORCE_SSE42 AND NOT MSVC)
@@ -273,6 +283,9 @@ if(PORTABLE)
     if(CMAKE_SYSTEM_PROCESSOR MATCHES "^s390x")
       set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -march=z196")
     endif()
+    if(CMAKE_SYSTEM_PROCESSOR MATCHES "^loongarch64")
+      set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -march=loongarch64")
+    endif()
   endif()
 else()
   if(MSVC)
@@ -292,8 +305,7 @@ if(NOT MSVC)
   set(CMAKE_REQUIRED_FLAGS "-msse4.2 -mpclmul")
 endif()
 
-if (NOT PORTABLE OR FORCE_SSE42)
-  CHECK_CXX_SOURCE_COMPILES("
+CHECK_CXX_SOURCE_COMPILES("
 #include <cstdint>
 #include <nmmintrin.h>
 #include <wmmintrin.h>
@@ -305,12 +317,11 @@ int main() {
   auto d = _mm_cvtsi128_si64(c);
 }
 " HAVE_SSE42)
-  if(HAVE_SSE42)
-    add_definitions(-DHAVE_SSE42)
-    add_definitions(-DHAVE_PCLMUL)
-  elseif(FORCE_SSE42)
-    message(FATAL_ERROR "FORCE_SSE42=ON but unable to compile with SSE4.2 enabled")
-  endif()
+if(HAVE_SSE42)
+  add_definitions(-DHAVE_SSE42)
+  add_definitions(-DHAVE_PCLMUL)
+elseif(FORCE_SSE42)
+  message(FATAL_ERROR "FORCE_SSE42=ON but unable to compile with SSE4.2 enabled")
 endif()
 
 # Check if -latomic is required or not
@@ -485,12 +496,6 @@ if(CMAKE_COMPILER_IS_GNUCXX)
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fno-builtin-memcmp")
 endif()
 
-option(ROCKSDB_LITE "Build RocksDBLite version" OFF)
-if(ROCKSDB_LITE)
-  add_definitions(-DROCKSDB_LITE)
-  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fno-exceptions -Os")
-endif()
-
 if(CMAKE_SYSTEM_NAME MATCHES "Cygwin")
   add_definitions(-fno-builtin-memcmp -DCYGWIN)
 elseif(CMAKE_SYSTEM_NAME MATCHES "Darwin")
@@ -649,12 +654,14 @@ set(SOURCES
         cache/cache.cc
         cache/cache_entry_roles.cc
         cache/cache_key.cc
+        cache/cache_helpers.cc
         cache/cache_reservation_manager.cc
         cache/charged_cache.cc
         cache/clock_cache.cc
         cache/compressed_secondary_cache.cc
-        cache/fast_lru_cache.cc
         cache/lru_cache.cc
+        cache/secondary_cache.cc
+        cache/secondary_cache_adapter.cc
         cache/sharded_cache.cc
         db/arena_wrapped_db_iter.cc
         db/blob/blob_contents.cc
@@ -741,6 +748,7 @@ set(SOURCES
         db/write_batch.cc
         db/write_batch_base.cc
         db/write_controller.cc
+        db/write_stall_stats.cc
         db/write_thread.cc
         env/composite_env.cc
         env/env.cc
@@ -796,6 +804,7 @@ set(SOURCES
         options/options.cc
         options/options_helper.cc
         options/options_parser.cc
+        port/mmap.cc
         port/stack_trace.cc
         table/adaptive/adaptive_table_factory.cc
         table/block_based/binary_search_index_reader.cc
@@ -805,6 +814,7 @@ set(SOURCES
         table/block_based/block_based_table_iterator.cc
         table/block_based/block_based_table_reader.cc
         table/block_based/block_builder.cc
+        table/block_based/block_cache.cc
         table/block_based/block_prefetcher.cc
         table/block_based/block_prefix_index.cc
         table/block_based/data_block_hash_index.cc
@@ -830,6 +840,7 @@ set(SOURCES
         table/get_context.cc
         table/iterator.cc
         table/merging_iterator.cc
+        table/compaction_merging_iterator.cc
         table/meta_blocks.cc
         table/persistent_cache_helper.cc
         table/plain/plain_table_bloom.cc
@@ -871,6 +882,7 @@ set(SOURCES
         util/compression_context_cache.cc
         util/concurrent_task_limiter_impl.cc
         util/crc32c.cc
+        util/data_structure.cc
         util/dynamic_bloom.cc
         util/hash.cc
         util/murmurhash.cc
@@ -978,6 +990,12 @@ if ( ROCKSDB_PLUGINS )
         plugin/${plugin}/${src}
         PROPERTIES COMPILE_FLAGS "${${plugin}_COMPILE_FLAGS}")
     endforeach()
+    foreach (test ${${plugin}_TESTS})
+      list(APPEND PLUGIN_TESTS plugin/${plugin}/${test})
+      set_source_files_properties(
+        plugin/${plugin}/${test}
+        PROPERTIES COMPILE_FLAGS "${${plugin}_COMPILE_FLAGS}")
+    endforeach()
     foreach (path ${${plugin}_INCLUDE_PATHS})
       include_directories(${path})
     endforeach()
@@ -1263,6 +1281,7 @@ if(WITH_TESTS OR WITH_BENCHMARK_TOOLS)
   add_subdirectory(third-party/gtest-1.8.1/fused-src/gtest)
   add_library(testharness STATIC
   test_util/mock_time_env.cc
+  test_util/secondary_cache_test_util.cc
   test_util/testharness.cc)
   target_link_libraries(testharness gtest)
 endif()
@@ -1469,6 +1488,7 @@ if(WITH_TESTS)
         utilities/ttl/ttl_test.cc
         utilities/util_merge_operators_test.cc
         utilities/write_batch_with_index/write_batch_with_index_test.cc
+	${PLUGIN_TESTS}
     )
   endif()
 
@@ -1572,9 +1592,9 @@ endif()
 
 option(WITH_TRACE_TOOLS "build with trace tools" ON)
 if(WITH_TRACE_TOOLS)
-  add_executable(block_cache_trace_analyzer_tool${ARTIFACT_SUFFIX}
+  add_executable(block_cache_trace_analyzer${ARTIFACT_SUFFIX}
     tools/block_cache_analyzer/block_cache_trace_analyzer_tool.cc)
-  target_link_libraries(block_cache_trace_analyzer_tool${ARTIFACT_SUFFIX}
+  target_link_libraries(block_cache_trace_analyzer${ARTIFACT_SUFFIX}
     ${ROCKSDB_LIB} ${GFLAGS_LIB} ${FOLLY_LIBS})
 
   add_executable(trace_analyzer${ARTIFACT_SUFFIX}

HISTORY.md

@@ -1,20 +1,141 @@
 # Rocksdb Change Log
-## 7.8.3 (11/29/2022)
-* Revert an internal change in 7.8.0 associated with some memory usage churn.
+## 8.2.1 (05/04/2023)
+* Public API comments and a Meta-internal fix to the buck build
 
-## 7.8.2 (11/27/2022)
+## 8.2.0 (04/24/2023)
+### Public API Changes
+* `SstFileWriter::DeleteRange()` now returns `Status::InvalidArgument` if the range's end key comes before its start key according to the user comparator. Previously the behavior was undefined.
+* Add `multi_get_for_update` to C API.
+* Remove unnecessary constructor for CompressionOptions.
+
+### Behavior changes
+* Changed default block cache size from an 8MB to 32MB LRUCache, which increases the default number of cache shards from 16 to 64. This change is intended to minimize cache mutex contention under stress conditions. See https://github.com/facebook/rocksdb/wiki/Block-Cache for more information.
+* For level compaction with `level_compaction_dynamic_level_bytes=true`, RocksDB now trivially moves levels down to fill LSM starting from bottommost level during DB open. See more in comments for option `level_compaction_dynamic_level_bytes` (#11321).
+* User-provided `ReadOptions` take effect for more reads of non-`CacheEntryRole::kDataBlock` blocks.
+* For level compaction with `level_compaction_dynamic_level_bytes=true`, RocksDB now drains unnecessary levels through background compaction automatically (#11340). This together with #11321 makes it automatic to migrate other compaction settings to level compaction with `level_compaction_dynamic_level_bytes=true`. In addition, a live DB that becomes smaller will now have unnecessary levels drained which can help to reduce read and space amp.
+* If `CompactRange()` is called with `CompactRangeOptions::bottommost_level_compaction=kForce*` to compact from L0 to L1, RocksDB now will try to do trivial move from L0 to L1 and then do an intra L1 compaction, instead of a L0 to L1 compaction with trivial move disabled (#11375)).
+
+### Bug Fixes
+* In the DB::VerifyFileChecksums API, ensure that file system reads of SST files are equal to the readahead_size in ReadOptions, if specified. Previously, each read was 2x the readahead_size.
+* In block cache tracing, fixed some cases of bad hit/miss information (and more) with MultiGet.
+
+### New Features
+* Add experimental `PerfContext` counters `iter_{next|prev|seek}_count` for db iterator, each counting the times of corresponding API being called.
+* Allow runtime changes to whether `WriteBufferManager` allows stall or not by calling `SetAllowStall()`
+* Added statistics tickers BYTES_COMPRESSED_FROM, BYTES_COMPRESSED_TO, BYTES_COMPRESSION_BYPASSED, BYTES_COMPRESSION_REJECTED, NUMBER_BLOCK_COMPRESSION_BYPASSED, and NUMBER_BLOCK_COMPRESSION_REJECTED. Disabled/deprecated histograms BYTES_COMPRESSED and BYTES_DECOMPRESSED, and ticker NUMBER_BLOCK_NOT_COMPRESSED. The new tickers offer more inight into compression ratios, rejected vs. disabled compression, etc. (#11388)
+* New statistics `rocksdb.file.read.{flush|compaction}.micros` that measure read time of block-based SST tables or blob files during flush or compaction.
+
+## 8.1.0 (03/18/2023)
+### Behavior changes
+* Compaction output file cutting logic now considers range tombstone start keys. For example, SST partitioner now may receive ParitionRequest for range tombstone start keys.
+* If the async_io ReadOption is specified for MultiGet or NewIterator on a platform that doesn't support IO uring, the option is ignored and synchronous IO is used.
+
+### Bug Fixes
+* Fixed an issue for backward iteration when user defined timestamp is enabled in combination with BlobDB.
+* Fixed a couple of cases where a Merge operand encountered during iteration wasn't reflected in the `internal_merge_count` PerfContext counter.
+* Fixed a bug in CreateColumnFamilyWithImport()/ExportColumnFamily() which did not support range tombstones (#11252).
+* Fixed a bug where an excluded column family from an atomic flush contains unflushed data that should've been included in this atomic flush (i.e, data of seqno less than the max seqno of this atomic flush), leading to potential data loss in this excluded column family when `WriteOptions::disableWAL == true` (#11148).
+
+### New Features
+* Add statistics rocksdb.secondary.cache.filter.hits, rocksdb.secondary.cache.index.hits, and rocksdb.secondary.cache.filter.hits
+* Added a new PerfContext counter `internal_merge_point_lookup_count` which tracks the number of Merge operands applied while serving point lookup queries.
+* Add new statistics rocksdb.table.open.prefetch.tail.read.bytes, rocksdb.table.open.prefetch.tail.{miss|hit}
+* Add support for SecondaryCache with HyperClockCache (`HyperClockCacheOptions` inherits `secondary_cache` option from `ShardedCacheOptions`)
+* Add new db properties `rocksdb.cf-write-stall-stats`, `rocksdb.db-write-stall-stats`and APIs to examine them in a structured way. In particular, users of `GetMapProperty()` with property `kCFWriteStallStats`/`kDBWriteStallStats` can now use the functions in `WriteStallStatsMapKeys` to find stats in the map.
+
+### Public API Changes
+* Changed various functions and features in `Cache` that are mostly relevant to custom implementations or wrappers. Especially, asychronous lookup functionality is moved from `Lookup()` to a new `StartAsyncLookup()` function.
+
+## 8.0.0 (02/19/2023)
+### Behavior changes
+* `ReadOptions::verify_checksums=false` disables checksum verification for more reads of non-`CacheEntryRole::kDataBlock` blocks.
+* In case of scan with async_io enabled, if posix doesn't support IOUring, Status::NotSupported error will be returned to the users. Initially that error was swallowed and reads were switched to synchronous reads.
+
+### Bug Fixes
+* Fixed a data race on `ColumnFamilyData::flush_reason` caused by concurrent flushes.
+* Fixed an issue in `Get` and `MultiGet` when user-defined timestamps is enabled in combination with BlobDB.
+* Fixed some atypical behaviors for `LockWAL()` such as allowing concurrent/recursive use and not expecting `UnlockWAL()` after non-OK result. See API comments.
+* Fixed a feature interaction bug where for blobs `GetEntity` would expose the blob reference instead of the blob value.
+* Fixed `DisableManualCompaction()` and `CompactRangeOptions::canceled` to cancel compactions even when they are waiting on conflicting compactions to finish
+* Fixed a bug in which a successful `GetMergeOperands()` could transiently return `Status::MergeInProgress()`
+* Return the correct error (Status::NotSupported()) to MultiGet caller when ReadOptions::async_io flag is true and IO uring is not enabled. Previously, Status::Corruption() was being returned when the actual failure was lack of async IO support.
+* Fixed a bug in DB open/recovery from a compressed WAL that was caused due to incorrect handling of certain record fragments with the same offset within a WAL block.
+
+### Feature Removal
+* Remove RocksDB Lite.
+* The feature block_cache_compressed is removed. Statistics related to it are removed too.
+* Remove deprecated Env::LoadEnv(). Use Env::CreateFromString() instead.
+* Remove deprecated FileSystem::Load(). Use FileSystem::CreateFromString() instead.
+* Removed the deprecated version of these utility functions and the corresponding Java bindings: `LoadOptionsFromFile`, `LoadLatestOptions`, `CheckOptionsCompatibility`.
+* Remove the FactoryFunc from the LoadObject method from the Customizable helper methods.
+
+### Public API Changes
+* Moved rarely-needed Cache class definition to new advanced_cache.h, and added a CacheWrapper class to advanced_cache.h. Minor changes to SimCache API definitions.
+* Completely removed the following deprecated/obsolete statistics: the tickers `BLOCK_CACHE_INDEX_BYTES_EVICT`, `BLOCK_CACHE_FILTER_BYTES_EVICT`, `BLOOM_FILTER_MICROS`, `NO_FILE_CLOSES`, `STALL_L0_SLOWDOWN_MICROS`, `STALL_MEMTABLE_COMPACTION_MICROS`, `STALL_L0_NUM_FILES_MICROS`, `RATE_LIMIT_DELAY_MILLIS`, `NO_ITERATORS`, `NUMBER_FILTERED_DELETES`, `WRITE_TIMEDOUT`, `BLOB_DB_GC_NUM_KEYS_OVERWRITTEN`, `BLOB_DB_GC_NUM_KEYS_EXPIRED`, `BLOB_DB_GC_BYTES_OVERWRITTEN`, `BLOB_DB_GC_BYTES_EXPIRED`, `BLOCK_CACHE_COMPRESSION_DICT_BYTES_EVICT` as well as the histograms `STALL_L0_SLOWDOWN_COUNT`, `STALL_MEMTABLE_COMPACTION_COUNT`, `STALL_L0_NUM_FILES_COUNT`, `HARD_RATE_LIMIT_DELAY_COUNT`, `SOFT_RATE_LIMIT_DELAY_COUNT`, `BLOB_DB_GC_MICROS`, and `NUM_DATA_BLOCKS_READ_PER_LEVEL`. Note that as a result, the C++ enum values of the still supported statistics have changed. Developers are advised to not rely on the actual numeric values.
+* Deprecated IngestExternalFileOptions::write_global_seqno and change default to false. This option only needs to be set to true to generate a DB compatible with RocksDB versions before 5.16.0.
+* Remove deprecated APIs `GetColumnFamilyOptionsFrom{Map|String}(const ColumnFamilyOptions&, ..)`, `GetDBOptionsFrom{Map|String}(const DBOptions&, ..)`, `GetBlockBasedTableOptionsFrom{Map|String}(const BlockBasedTableOptions& table_options, ..)` and ` GetPlainTableOptionsFrom{Map|String}(const PlainTableOptions& table_options,..)`.
+* Added a subcode of `Status::Corruption`, `Status::SubCode::kMergeOperatorFailed`, for users to identify corruption failures originating in the merge operator, as opposed to RocksDB's internally identified data corruptions
+
+### Build Changes
+* The `make` build now builds a shared library by default instead of a static library. Use `LIB_MODE=static` to override.
+
+### New Features
+* Compaction filters are now supported for wide-column entities by means of the `FilterV3` API. See the comment of the API for more details.
+* Added `do_not_compress_roles` to `CompressedSecondaryCacheOptions` to disable compression on certain kinds of block. Filter blocks are now not compressed by CompressedSecondaryCache by default.
+* Added a new `MultiGetEntity` API that enables batched wide-column point lookups. See the API comments for more details.
+
+## 7.10.0 (01/23/2023)
 ### Behavior changes
 * Make best-efforts recovery verify SST unique ID before Version construction (#10962)
-* Fix failed memtable flush retry bug that could cause wrongly ordered updates, which would surface to writers as `Status::Corruption` in case of `force_consistency_checks=true` (default). It affects use cases that enable both parallel flush (`max_background_flushes > 1` or `max_background_jobs >= 8`) and non-default memtable count (`max_write_buffer_number > 2`).
-* Tiered Storage: fixed excessive keys written to penultimate level in non-debug builds.
+* Introduce `epoch_number` and sort L0 files by `epoch_number` instead of `largest_seqno`. `epoch_number` represents the order of a file being flushed or ingested/imported. Compaction output file will be assigned with the minimum `epoch_number` among input files'. For L0, larger `epoch_number` indicates newer L0 file.
 
 ### Bug Fixes
-* Fixed a regression in scan for async_io. During seek, valid buffers were getting cleared causing a regression.
-* Fixed a performance regression in iterator where range tombstones after `iterate_upper_bound` is processed.
+* Fixed a regression in iterator where range tombstones after `iterate_upper_bound` is processed.
+* Fixed a memory leak in MultiGet with async_io read option, caused by IO errors during table file open
+* Fixed a bug that multi-level FIFO compaction deletes one file in non-L0 even when `CompactionOptionsFIFO::max_table_files_size` is no exceeded since #10348 or 7.8.0.
+* Fixed a bug caused by `DB::SyncWAL()` affecting `track_and_verify_wals_in_manifest`. Without the fix, application may see "open error: Corruption: Missing WAL with log number" while trying to open the db. The corruption is a false alarm but prevents DB open (#10892).
+* Fixed a BackupEngine bug in which RestoreDBFromLatestBackup would fail if the latest backup was deleted and there is another valid backup available.
+* Fix L0 file misorder corruption caused by ingesting files of overlapping seqnos with memtable entries' through introducing `epoch_number`. Before the fix, `force_consistency_checks=true` may catch the corruption before it's exposed to readers, in which case writes returning `Status::Corruption` would be expected. Also replace the previous incomplete fix (#5958) to the same corruption with this new and more complete fix.
+* Fixed a bug in LockWAL() leading to re-locking mutex (#11020).
+* Fixed a heap use after free bug in async scan prefetching when the scan thread and another thread try to read and load the same seek block into cache.
+* Fixed a heap use after free in async scan prefetching if dictionary compression is enabled, in which case sync read of the compression dictionary gets mixed with async prefetching
+* Fixed a data race bug of `CompactRange()` under `change_level=true` acts on overlapping range with an ongoing file ingestion for level compaction. This will either result in overlapping file ranges corruption at a certain level caught by `force_consistency_checks=true` or protentially two same keys both with seqno 0 in two different levels (i.e, new data ends up in lower/older level). The latter will be caught by assertion in debug build but go silently and result in read returning wrong result in release build. This fix is general so it also replaced previous fixes to a similar problem for `CompactFiles()` (#4665), general `CompactRange()` and auto compaction (commit 5c64fb6 and 87dfc1d).
+* Fixed a bug in compaction output cutting where small output files were produced due to TTL file cutting states were not being updated (#11075).
+
+### New Features
+* When an SstPartitionerFactory is configured, CompactRange() now automatically selects for compaction any files overlapping a partition boundary that is in the compaction range, even if no actual entries are in the requested compaction range. With this feature, manual compaction can be used to (re-)establish SST partition points when SstPartitioner changes, without a full compaction.
+* Add BackupEngine feature to exclude files from backup that are known to be backed up elsewhere, using `CreateBackupOptions::exclude_files_callback`. To restore the DB, the excluded files must be provided in alternative backup directories using `RestoreOptions::alternate_dirs`.
+
+### Public API Changes
+* Substantial changes have been made to the Cache class to support internal development goals. Direct use of Cache class members is discouraged and further breaking modifications are expected in the future. SecondaryCache has some related changes and implementations will need to be updated. (Unlike Cache, SecondaryCache is still intended to support user implementations, and disruptive changes will be avoided.) (#10975)
+* Add `MergeOperationOutput::op_failure_scope` for merge operator users to control the blast radius of merge operator failures. Existing merge operator users do not need to make any change to preserve the old behavior
+
+### Performance Improvements
+* Updated xxHash source code, which should improve kXXH3 checksum speed, at least on ARM (#11098).
+* Improved CPU efficiency of DB reads, from block cache access improvements (#10975).
+
+## 7.9.0 (11/21/2022)
+### Performance Improvements
+* Fixed an iterator performance regression for delete range users when scanning through a consecutive sequence of range tombstones (#10877).
 
-## 7.8.1 (11/2/2022)
 ### Bug Fixes
 * Fix memory corruption error in scans if async_io is enabled. Memory corruption happened if there is IOError while reading the data leading to empty buffer and other buffer already in progress of async read goes again for reading.
+* Fix failed memtable flush retry bug that could cause wrongly ordered updates, which would surface to writers as `Status::Corruption` in case of `force_consistency_checks=true` (default). It affects use cases that enable both parallel flush (`max_background_flushes > 1` or `max_background_jobs >= 8`) and non-default memtable count (`max_write_buffer_number > 2`).
+* Fixed an issue where the `READ_NUM_MERGE_OPERANDS` ticker was not updated when the base key-value or tombstone was read from an SST file.
+* Fixed a memory safety bug when using a SecondaryCache with `block_cache_compressed`. `block_cache_compressed` no longer attempts to use SecondaryCache features.
+* Fixed a regression in scan for async_io. During seek, valid buffers were getting cleared causing a regression.
+* Tiered Storage: fixed excessive keys written to penultimate level in non-debug builds.
+
+### New Features
+* Add basic support for user-defined timestamp to Merge (#10819).
+* Add stats for ReadAsync time spent and async read errors.
+* Basic support for the wide-column data model is now available. Wide-column entities can be stored using the `PutEntity` API, and retrieved using `GetEntity` and the new `columns` API of iterator. For compatibility, the classic APIs `Get` and `MultiGet`, as well as iterator's `value` API return the value of the anonymous default column of wide-column entities; also, `GetEntity` and iterator's `columns` return any plain key-values in the form of an entity which only has the anonymous default column. `Merge` (and `GetMergeOperands`) currently also apply to the default column; any other columns of entities are unaffected by `Merge` operations. Note that some features like compaction filters, transactions, user-defined timestamps, and the SST file writer do not yet support wide-column entities; also, there is currently no `MultiGet`-like API to retrieve multiple entities at once. We plan to gradually close the above gaps and also implement new features like column-level operations (e.g. updating or querying only certain columns of an entity).
+* Marked HyperClockCache as a production-ready alternative to LRUCache for the block cache. HyperClockCache greatly improves hot-path CPU efficiency under high parallel load or high contention, with some documented caveats and limitations. As much as 4.5x higher ops/sec vs. LRUCache has been seen in db_bench under high parallel load.
+* Add periodic diagnostics to info_log (LOG file) for HyperClockCache block cache if performance is degraded by bad `estimated_entry_charge` option.
+
+### Public API Changes
+* Marked `block_cache_compressed` as a deprecated feature. Use SecondaryCache instead.
+* Added a `SecondaryCache::InsertSaved()` API, with default implementation depending on `Insert()`. Some implementations might need to add a custom implementation of `InsertSaved()`. (Details in API comments.)
 
 ## 7.8.0 (10/22/2022)
 ### New Features
@@ -45,10 +166,11 @@
 * Try to align the compaction output file boundaries to the next level ones, which can reduce more than 10% compaction load for the default level compaction. The feature is enabled by default, to disable, set `AdvancedColumnFamilyOptions.level_compaction_dynamic_file_size` to false. As a side effect, it can create SSTs larger than the target_file_size (capped at 2x target_file_size) or smaller files.
 * Improve RoundRobin TTL compaction, which is going to be the same as normal RoundRobin compaction to move the compaction cursor.
 * Fix a small CPU regression caused by a change that UserComparatorWrapper was made Customizable, because Customizable itself has small CPU overhead for initialization.
-* Fixed an iterator performance regression for delete range users when scanning through a consecutive sequence of range tombstones (#10877).
 
 ### Behavior Changes
 * Sanitize min_write_buffer_number_to_merge to 1 if atomic flush is enabled to prevent unexpected data loss when WAL is disabled in a multi-column-family setting (#10773).
+* With periodic stat dumper waits up every options.stats_dump_period_sec seconds, it won't dump stats for a CF if it has no change in the period, unless 7 periods have been skipped.
+* Only periodic stats dumper triggered by options.stats_dump_period_sec will update stats interval. Ones triggered by DB::GetProperty() will not update stats interval and will report based on an interval since the last time stats dump period.
 
 ### Public API changes
 * Make kXXH3 checksum the new default, because it is faster on common hardware, especially with kCRC32c affected by a performance bug in some versions of clang (https://github.com/facebook/rocksdb/issues/9891). DBs written with this new setting can be read by RocksDB 6.27 and newer.

INSTALL.md

@@ -48,6 +48,11 @@ to build a portable binary, add `PORTABLE=1` before your make commands, like thi
 * If you wish to build the RocksJava static target, then cmake is required for building Snappy.
 
 * If you wish to run microbench (e.g, `make microbench`, `make ribbon_bench` or `cmake -DWITH_BENCHMARK=1`), Google benchmark >= 1.6.0 is needed.
+* You can do the following to install Google benchmark. These commands are copied from `./build_tools/ubuntu20_image/Dockerfile`:
+
+`$ git clone --depth 1 --branch v1.7.0 https://github.com/google/benchmark.git ~/benchmark`
+
+`$ cd ~/benchmark && mkdir build && cd build && cmake .. -GNinja -DCMAKE_BUILD_TYPE=Release -DBENCHMARK_ENABLE_GTEST_TESTS=0 && ninja && ninja install`
 
 ## Supported platforms
 
@@ -178,7 +183,7 @@ to build a portable binary, add `PORTABLE=1` before your make commands, like thi
         gmake rocksdbjava
 
 * **iOS**:
-  * Run: `TARGET_OS=IOS make static_lib`. When building the project which uses rocksdb iOS library, make sure to define two important pre-processing macros: `ROCKSDB_LITE` and `IOS_CROSS_COMPILE`.
+  * Run: `TARGET_OS=IOS make static_lib`. When building the project which uses rocksdb iOS library, make sure to define an important pre-processing macros: `IOS_CROSS_COMPILE`.
 
 * **Windows** (Visual Studio 2017 to up):
   * Read and follow the instructions at CMakeLists.txt

Makefile

@@ -44,13 +44,6 @@ quoted_perl_command = $(subst ','\'',$(perl_command))
 # Set the default DEBUG_LEVEL to 1
 DEBUG_LEVEL?=1
 
-# LIB_MODE says whether or not to use/build "shared" or "static" libraries.
-# Mode "static" means to link against static libraries (.a)
-# Mode "shared" means to link against shared libraries (.so, .sl, .dylib, etc)
-#
-# Set the default LIB_MODE to static
-LIB_MODE?=static
-
 # OBJ_DIR is where the object files reside.  Default to the current directory
 OBJ_DIR?=.
 
@@ -81,29 +74,23 @@ else ifneq ($(filter jtest rocksdbjava%, $(MAKECMDGOALS)),)
 	endif
 endif
 
-$(info $$DEBUG_LEVEL is ${DEBUG_LEVEL})
+# LIB_MODE says whether or not to use/build "shared" or "static" libraries.
+# Mode "static" means to link against static libraries (.a)
+# Mode "shared" means to link against shared libraries (.so, .sl, .dylib, etc)
+#
+ifeq ($(DEBUG_LEVEL), 0)
+# For optimized, set the default LIB_MODE to static for code size/efficiency
+	LIB_MODE?=static
+else
+# For debug, set the default LIB_MODE to shared for efficient `make check` etc.
+	LIB_MODE?=shared
+endif
 
-# Lite build flag.
-LITE ?= 0
-ifeq ($(LITE), 0)
-ifneq ($(filter -DROCKSDB_LITE,$(OPT)),)
-  # Be backward compatible and support older format where OPT=-DROCKSDB_LITE is
-  # specified instead of LITE=1 on the command line.
-  LITE=1
-endif
-else ifeq ($(LITE), 1)
-ifeq ($(filter -DROCKSDB_LITE,$(OPT)),)
-	OPT += -DROCKSDB_LITE
-endif
-endif
+$(info $$DEBUG_LEVEL is $(DEBUG_LEVEL), $$LIB_MODE is $(LIB_MODE))
 
 # Figure out optimize level.
 ifneq ($(DEBUG_LEVEL), 2)
-ifeq ($(LITE), 0)
 	OPTIMIZE_LEVEL ?= -O2
-else
-	OPTIMIZE_LEVEL ?= -Os
-endif
 endif
 # `OPTIMIZE_LEVEL` is empty when the user does not set it and `DEBUG_LEVEL=2`.
 # In that case, the compiler default (`-O0` for gcc and clang) will be used.
@@ -266,6 +253,7 @@ ROCKSDB_PLUGIN_EXTERNS = $(foreach p, $(ROCKSDB_PLUGIN_W_FUNCS), int $($(p)_FUNC
 ROCKSDB_PLUGIN_BUILTINS = $(foreach p, $(ROCKSDB_PLUGIN_W_FUNCS), {\"$(p)\"\, $($(p)_FUNC)}\,)
 ROCKSDB_PLUGIN_LDFLAGS = $(foreach plugin, $(ROCKSDB_PLUGINS), $($(plugin)_LDFLAGS))
 ROCKSDB_PLUGIN_PKGCONFIG_REQUIRES = $(foreach plugin, $(ROCKSDB_PLUGINS), $($(plugin)_PKGCONFIG_REQUIRES))
+ROCKSDB_PLUGIN_TESTS = $(foreach p, $(ROCKSDB_PLUGINS), $(foreach test, $($(p)_TESTS), plugin/$(p)/$(test)))
 
 CXXFLAGS += $(foreach plugin, $(ROCKSDB_PLUGINS), $($(plugin)_CXXFLAGS))
 PLATFORM_LDFLAGS += $(ROCKSDB_PLUGIN_LDFLAGS)
@@ -336,13 +324,6 @@ endif
 ifeq ($(PLATFORM), OS_SOLARIS)
 	PLATFORM_CXXFLAGS += -D _GLIBCXX_USE_C99
 endif
-ifneq ($(filter -DROCKSDB_LITE,$(OPT)),)
-	# found
-	CFLAGS += -fno-exceptions
-	CXXFLAGS += -fno-exceptions
-	# LUA is not supported under ROCKSDB_LITE
-	LUA_PATH =
-endif
 
 ifeq ($(LIB_MODE),shared)
 # So that binaries are executable from build location, in addition to install location
@@ -356,7 +337,7 @@ ifneq ($(MACHINE), arm64)
 # linking with jemalloc (as it won't be arm64-compatible) and remove some other options
 # set during platform detection
 DISABLE_JEMALLOC=1
-PLATFORM_CFLAGS := $(filter-out -march=native -DHAVE_SSE42 -DHAVE_AVX2, $(PLATFORM_CFLAGS))
+PLATFORM_CCFLAGS := $(filter-out -march=native -DHAVE_SSE42 -DHAVE_AVX2, $(PLATFORM_CCFLAGS))
 PLATFORM_CXXFLAGS := $(filter-out -march=native -DHAVE_SSE42 -DHAVE_AVX2, $(PLATFORM_CXXFLAGS))
 endif
 endif
@@ -558,7 +539,7 @@ endif
 
 ifdef USE_CLANG
 	# Used by some teams in Facebook
-	WARNING_FLAGS += -Wshift-sign-overflow
+	WARNING_FLAGS += -Wshift-sign-overflow -Wambiguous-reversed-operator
 endif
 
 ifeq ($(PLATFORM), OS_OPENBSD)
@@ -647,10 +628,12 @@ STRESS_OBJECTS =  $(patsubst %.cc, $(OBJ_DIR)/%.o, $(STRESS_LIB_SOURCES))
 ALL_SOURCES  = $(filter-out util/build_version.cc, $(LIB_SOURCES)) $(TEST_LIB_SOURCES) $(MOCK_LIB_SOURCES) $(GTEST_DIR)/gtest/gtest-all.cc
 ALL_SOURCES += $(TOOL_LIB_SOURCES) $(BENCH_LIB_SOURCES) $(CACHE_BENCH_LIB_SOURCES) $(ANALYZER_LIB_SOURCES) $(STRESS_LIB_SOURCES)
 ALL_SOURCES += $(TEST_MAIN_SOURCES) $(TOOL_MAIN_SOURCES) $(BENCH_MAIN_SOURCES)
-ALL_SOURCES += $(ROCKSDB_PLUGIN_SOURCES)
+ALL_SOURCES += $(ROCKSDB_PLUGIN_SOURCES) $(ROCKSDB_PLUGIN_TESTS)
 
+PLUGIN_TESTS = $(patsubst %.cc, %, $(notdir $(ROCKSDB_PLUGIN_TESTS)))
 TESTS = $(patsubst %.cc, %, $(notdir $(TEST_MAIN_SOURCES)))
 TESTS += $(patsubst %.c, %, $(notdir $(TEST_MAIN_SOURCES_C)))
+TESTS += $(PLUGIN_TESTS)
 
 # `make check-headers` to very that each header file includes its own
 # dependencies
@@ -702,6 +685,7 @@ NON_PARALLEL_TEST = \
 	env_test \
 	deletefile_test \
 	db_bloom_filter_test \
+	$(PLUGIN_TESTS) \
 
 PARALLEL_TEST = $(filter-out $(NON_PARALLEL_TEST), $(TESTS))
 
@@ -1078,13 +1062,11 @@ check: all
 	rm -rf $(TEST_TMPDIR)
 ifneq ($(PLATFORM), OS_AIX)
 	$(PYTHON) tools/check_all_python.py
-ifeq ($(filter -DROCKSDB_LITE,$(OPT)),)
 ifndef ASSERT_STATUS_CHECKED # not yet working with these tests
 	$(PYTHON) tools/ldb_test.py
 	sh tools/rocksdb_dump_test.sh
 endif
 endif
-endif
 ifndef SKIP_FORMAT_BUCK_CHECKS
 	$(MAKE) check-format
 	$(MAKE) check-buck-targets
@@ -1240,9 +1222,9 @@ clean: clean-ext-libraries-all clean-rocks clean-rocksjava
 clean-not-downloaded: clean-ext-libraries-bin clean-rocks clean-not-downloaded-rocksjava
 
 clean-rocks:
-	echo shared=$(ALL_SHARED_LIBS)
-	echo static=$(ALL_STATIC_LIBS)
-	rm -f $(BENCHMARKS) $(TOOLS) $(TESTS) $(PARALLEL_TEST) $(ALL_STATIC_LIBS) $(ALL_SHARED_LIBS) $(MICROBENCHS)
+# Not practical to exactly match all versions/variants in naming (e.g. debug or not)
+	rm -f ${LIBNAME}*.so* ${LIBNAME}*.a
+	rm -f $(BENCHMARKS) $(TOOLS) $(TESTS) $(PARALLEL_TEST) $(MICROBENCHS)
 	rm -rf $(CLEAN_FILES) ios-x86 ios-arm scan_build_report
 	$(FIND) . -name "*.[oda]" -exec rm -f {} \;
 	$(FIND) . -type f \( -name "*.gcda" -o -name "*.gcno" \) -exec rm -f {} \;
@@ -1355,6 +1337,14 @@ db_sanity_test: $(OBJ_DIR)/tools/db_sanity_test.o $(LIBRARY)
 db_repl_stress: $(OBJ_DIR)/tools/db_repl_stress.o $(LIBRARY)
 	$(AM_LINK)
 
+define MakeTestRule
+$(notdir $(1:%.cc=%)): $(1:%.cc=$$(OBJ_DIR)/%.o) $$(TEST_LIBRARY) $$(LIBRARY)
+	$$(AM_LINK)
+endef
+
+# For each PLUGIN test, create a rule to generate the test executable
+$(foreach test, $(ROCKSDB_PLUGIN_TESTS), $(eval $(call MakeTestRule, $(test))))
+
 arena_test: $(OBJ_DIR)/memory/arena_test.o $(TEST_LIBRARY) $(LIBRARY)
 	$(AM_LINK)
 
@@ -2058,7 +2048,7 @@ JAVA_INCLUDE = -I$(JAVA_HOME)/include/ -I$(JAVA_HOME)/include/linux
 ifeq ($(PLATFORM), OS_SOLARIS)
 	ARCH := $(shell isainfo -b)
 else ifeq ($(PLATFORM), OS_OPENBSD)
-	ifneq (,$(filter amd64 ppc64 ppc64le s390x arm64 aarch64 sparc64, $(MACHINE)))
+	ifneq (,$(filter amd64 ppc64 ppc64le s390x arm64 aarch64 sparc64 loongarch64, $(MACHINE)))
 		ARCH := 64
 	else
 		ARCH := 32
@@ -2079,7 +2069,7 @@ ifneq ($(origin JNI_LIBC), undefined)
 endif
 
 ifeq (,$(ROCKSDBJNILIB))
-ifneq (,$(filter ppc% s390x arm64 aarch64 sparc64, $(MACHINE)))
+ifneq (,$(filter ppc% s390x arm64 aarch64 sparc64 loongarch64, $(MACHINE)))
 	ROCKSDBJNILIB = librocksdbjni-linux-$(MACHINE)$(JNI_LIBC_POSTFIX).so
 else
 	ROCKSDBJNILIB = librocksdbjni-linux$(ARCH)$(JNI_LIBC_POSTFIX).so
@@ -2092,8 +2082,8 @@ ROCKSDB_JAVADOCS_JAR = rocksdbjni-$(ROCKSDB_JAVA_VERSION)-javadoc.jar
 ROCKSDB_SOURCES_JAR = rocksdbjni-$(ROCKSDB_JAVA_VERSION)-sources.jar
 SHA256_CMD = sha256sum
 
-ZLIB_VER ?= 1.2.12
-ZLIB_SHA256 ?= 91844808532e5ce316b3c010929493c0244f3d37593afd6de04f71821d5136d9
+ZLIB_VER ?= 1.2.13
+ZLIB_SHA256 ?= b3a24de97a8fdbc835b9833169501030b8977031bcb54b3b3ac13740f846ab30
 ZLIB_DOWNLOAD_BASE ?= http://zlib.net
 BZIP2_VER ?= 1.0.8
 BZIP2_SHA256 ?= ab5a03176ee106d3f0fa90e381da478ddae405918153cca248e682cd0c4a2269
@@ -2477,18 +2467,6 @@ build_size:
 	$(REPORT_BUILD_STATISTIC) rocksdb.build_size.shared_lib $$(stat --printf="%s" `readlink -f librocksdb.so`)
 	strip `readlink -f librocksdb.so`
 	$(REPORT_BUILD_STATISTIC) rocksdb.build_size.shared_lib_stripped $$(stat --printf="%s" `readlink -f librocksdb.so`)
-	# === lite build, static ===
-	$(MAKE) clean
-	$(MAKE) LITE=1 static_lib
-	$(REPORT_BUILD_STATISTIC) rocksdb.build_size.static_lib_lite $$(stat --printf="%s" librocksdb.a)
-	strip librocksdb.a
-	$(REPORT_BUILD_STATISTIC) rocksdb.build_size.static_lib_lite_stripped $$(stat --printf="%s" librocksdb.a)
-	# === lite build, shared ===
-	$(MAKE) clean
-	$(MAKE) LITE=1 shared_lib
-	$(REPORT_BUILD_STATISTIC) rocksdb.build_size.shared_lib_lite $$(stat --printf="%s" `readlink -f librocksdb.so`)
-	strip `readlink -f librocksdb.so`
-	$(REPORT_BUILD_STATISTIC) rocksdb.build_size.shared_lib_lite_stripped $$(stat --printf="%s" `readlink -f librocksdb.so`)
 
 # ---------------------------------------------------------------------------
 #  	Platform-specific compilation

README.md

@@ -1,8 +1,6 @@
 ## RocksDB: A Persistent Key-Value Store for Flash and RAM Storage
 
 [![CircleCI Status](https://circleci.com/gh/facebook/rocksdb.svg?style=svg)](https://circleci.com/gh/facebook/rocksdb)
-[![Appveyor Build status](https://ci.appveyor.com/api/projects/status/fbgfu0so3afcno78/branch/main?svg=true)](https://ci.appveyor.com/project/Facebook/rocksdb/branch/main)
-[![PPC64le Build Status](http://140-211-168-68-openstack.osuosl.org:8080/buildStatus/icon?job=rocksdb&style=plastic)](http://140-211-168-68-openstack.osuosl.org:8080/job/rocksdb)
 
 RocksDB is developed and maintained by Facebook Database Engineering Team.
 It is built on earlier work on [LevelDB](https://github.com/google/leveldb) by Sanjay Ghemawat (sanjay@google.com)

ROCKSDB_LITE.md

@@ -1,21 +0,0 @@
-# RocksDBLite
-
-RocksDBLite is a project focused on mobile use cases, which don't need a lot of fancy things we've built for server workloads and they are very sensitive to binary size. For that reason, we added a compile flag ROCKSDB_LITE that comments out a lot of the nonessential code and keeps the binary lean.
-
-Some examples of the features disabled by ROCKSDB_LITE:
-* compiled-in support for LDB tool
-* No backup engine
-* No support for replication (which we provide in form of TransactionalIterator)
-* No advanced monitoring tools
-* No special-purpose memtables that are highly optimized for specific use cases
-* No Transactions
-
-When adding a new big feature to RocksDB, please add ROCKSDB_LITE compile guard if:
-* Nobody from mobile really needs your feature,
-* Your feature is adding a lot of weight to the binary.
-
-Don't add ROCKSDB_LITE compile guard if:
-* It would introduce a lot of code complexity. Compile guards make code harder to read. It's a trade-off.
-* Your feature is not adding a lot of weight.
-
-If unsure, ask. :)

TARGETS

@@ -11,13 +11,15 @@ load("//rocks/buckifier:defs.bzl", "cpp_library_wrapper","rocks_cpp_library_wrap
 cpp_library_wrapper(name="rocksdb_lib", srcs=[
         "cache/cache.cc",
         "cache/cache_entry_roles.cc",
+        "cache/cache_helpers.cc",
         "cache/cache_key.cc",
         "cache/cache_reservation_manager.cc",
         "cache/charged_cache.cc",
         "cache/clock_cache.cc",
         "cache/compressed_secondary_cache.cc",
-        "cache/fast_lru_cache.cc",
         "cache/lru_cache.cc",
+        "cache/secondary_cache.cc",
+        "cache/secondary_cache_adapter.cc",
         "cache/sharded_cache.cc",
         "db/arena_wrapped_db_iter.cc",
         "db/blob/blob_contents.cc",
@@ -104,6 +106,7 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
         "db/write_batch.cc",
         "db/write_batch_base.cc",
         "db/write_controller.cc",
+        "db/write_stall_stats.cc",
         "db/write_thread.cc",
         "env/composite_env.cc",
         "env/env.cc",
@@ -163,6 +166,7 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
         "options/options.cc",
         "options/options_helper.cc",
         "options/options_parser.cc",
+        "port/mmap.cc",
         "port/port_posix.cc",
         "port/stack_trace.cc",
         "port/win/env_default.cc",
@@ -179,6 +183,7 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
         "table/block_based/block_based_table_iterator.cc",
         "table/block_based/block_based_table_reader.cc",
         "table/block_based/block_builder.cc",
+        "table/block_based/block_cache.cc",
         "table/block_based/block_prefetcher.cc",
         "table/block_based/block_prefix_index.cc",
         "table/block_based/data_block_footer.cc",
@@ -197,6 +202,7 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
         "table/block_based/reader_common.cc",
         "table/block_based/uncompression_dict_reader.cc",
         "table/block_fetcher.cc",
+        "table/compaction_merging_iterator.cc",
         "table/cuckoo/cuckoo_table_builder.cc",
         "table/cuckoo/cuckoo_table_factory.cc",
         "table/cuckoo/cuckoo_table_reader.cc",
@@ -244,6 +250,7 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
         "util/concurrent_task_limiter_impl.cc",
         "util/crc32c.cc",
         "util/crc32c_arm64.cc",
+        "util/data_structure.cc",
         "util/dynamic_bloom.cc",
         "util/file_checksum_helper.cc",
         "util/hash.cc",
@@ -347,350 +354,14 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
         "//folly/synchronization:distributed_mutex",
     ], headers=None, link_whole=False, extra_test_libs=False)
 
-cpp_library_wrapper(name="rocksdb_whole_archive_lib", srcs=[
-        "cache/cache.cc",
-        "cache/cache_entry_roles.cc",
-        "cache/cache_key.cc",
-        "cache/cache_reservation_manager.cc",
-        "cache/charged_cache.cc",
-        "cache/clock_cache.cc",
-        "cache/compressed_secondary_cache.cc",
-        "cache/fast_lru_cache.cc",
-        "cache/lru_cache.cc",
-        "cache/sharded_cache.cc",
-        "db/arena_wrapped_db_iter.cc",
-        "db/blob/blob_contents.cc",
-        "db/blob/blob_fetcher.cc",
-        "db/blob/blob_file_addition.cc",
-        "db/blob/blob_file_builder.cc",
-        "db/blob/blob_file_cache.cc",
-        "db/blob/blob_file_garbage.cc",
-        "db/blob/blob_file_meta.cc",
-        "db/blob/blob_file_reader.cc",
-        "db/blob/blob_garbage_meter.cc",
-        "db/blob/blob_log_format.cc",
-        "db/blob/blob_log_sequential_reader.cc",
-        "db/blob/blob_log_writer.cc",
-        "db/blob/blob_source.cc",
-        "db/blob/prefetch_buffer_collection.cc",
-        "db/builder.cc",
-        "db/c.cc",
-        "db/column_family.cc",
-        "db/compaction/compaction.cc",
-        "db/compaction/compaction_iterator.cc",
-        "db/compaction/compaction_job.cc",
-        "db/compaction/compaction_outputs.cc",
-        "db/compaction/compaction_picker.cc",
-        "db/compaction/compaction_picker_fifo.cc",
-        "db/compaction/compaction_picker_level.cc",
-        "db/compaction/compaction_picker_universal.cc",
-        "db/compaction/compaction_service_job.cc",
-        "db/compaction/compaction_state.cc",
-        "db/compaction/sst_partitioner.cc",
-        "db/compaction/subcompaction_state.cc",
-        "db/convenience.cc",
-        "db/db_filesnapshot.cc",
-        "db/db_impl/compacted_db_impl.cc",
-        "db/db_impl/db_impl.cc",
-        "db/db_impl/db_impl_compaction_flush.cc",
-        "db/db_impl/db_impl_debug.cc",
-        "db/db_impl/db_impl_experimental.cc",
-        "db/db_impl/db_impl_files.cc",
-        "db/db_impl/db_impl_open.cc",
-        "db/db_impl/db_impl_readonly.cc",
-        "db/db_impl/db_impl_secondary.cc",
-        "db/db_impl/db_impl_write.cc",
-        "db/db_info_dumper.cc",
-        "db/db_iter.cc",
-        "db/dbformat.cc",
-        "db/error_handler.cc",
-        "db/event_helpers.cc",
-        "db/experimental.cc",
-        "db/external_sst_file_ingestion_job.cc",
-        "db/file_indexer.cc",
-        "db/flush_job.cc",
-        "db/flush_scheduler.cc",
-        "db/forward_iterator.cc",
-        "db/import_column_family_job.cc",
-        "db/internal_stats.cc",
-        "db/log_reader.cc",
-        "db/log_writer.cc",
-        "db/logs_with_prep_tracker.cc",
-        "db/malloc_stats.cc",
-        "db/memtable.cc",
-        "db/memtable_list.cc",
-        "db/merge_helper.cc",
-        "db/merge_operator.cc",
-        "db/output_validator.cc",
-        "db/periodic_task_scheduler.cc",
-        "db/range_del_aggregator.cc",
-        "db/range_tombstone_fragmenter.cc",
-        "db/repair.cc",
-        "db/seqno_to_time_mapping.cc",
-        "db/snapshot_impl.cc",
-        "db/table_cache.cc",
-        "db/table_properties_collector.cc",
-        "db/transaction_log_impl.cc",
-        "db/trim_history_scheduler.cc",
-        "db/version_builder.cc",
-        "db/version_edit.cc",
-        "db/version_edit_handler.cc",
-        "db/version_set.cc",
-        "db/wal_edit.cc",
-        "db/wal_manager.cc",
-        "db/wide/wide_column_serialization.cc",
-        "db/wide/wide_columns.cc",
-        "db/write_batch.cc",
-        "db/write_batch_base.cc",
-        "db/write_controller.cc",
-        "db/write_thread.cc",
-        "env/composite_env.cc",
-        "env/env.cc",
-        "env/env_chroot.cc",
-        "env/env_encryption.cc",
-        "env/env_posix.cc",
-        "env/file_system.cc",
-        "env/file_system_tracer.cc",
-        "env/fs_posix.cc",
-        "env/fs_remap.cc",
-        "env/io_posix.cc",
-        "env/mock_env.cc",
-        "env/unique_id_gen.cc",
-        "file/delete_scheduler.cc",
-        "file/file_prefetch_buffer.cc",
-        "file/file_util.cc",
-        "file/filename.cc",
-        "file/line_file_reader.cc",
-        "file/random_access_file_reader.cc",
-        "file/read_write_util.cc",
-        "file/readahead_raf.cc",
-        "file/sequence_file_reader.cc",
-        "file/sst_file_manager_impl.cc",
-        "file/writable_file_writer.cc",
-        "logging/auto_roll_logger.cc",
-        "logging/event_logger.cc",
-        "logging/log_buffer.cc",
-        "memory/arena.cc",
-        "memory/concurrent_arena.cc",
-        "memory/jemalloc_nodump_allocator.cc",
-        "memory/memkind_kmem_allocator.cc",
-        "memory/memory_allocator.cc",
-        "memtable/alloc_tracker.cc",
-        "memtable/hash_linklist_rep.cc",
-        "memtable/hash_skiplist_rep.cc",
-        "memtable/skiplistrep.cc",
-        "memtable/vectorrep.cc",
-        "memtable/write_buffer_manager.cc",
-        "monitoring/histogram.cc",
-        "monitoring/histogram_windowing.cc",
-        "monitoring/in_memory_stats_history.cc",
-        "monitoring/instrumented_mutex.cc",
-        "monitoring/iostats_context.cc",
-        "monitoring/perf_context.cc",
-        "monitoring/perf_level.cc",
-        "monitoring/persistent_stats_history.cc",
-        "monitoring/statistics.cc",
-        "monitoring/thread_status_impl.cc",
-        "monitoring/thread_status_updater.cc",
-        "monitoring/thread_status_updater_debug.cc",
-        "monitoring/thread_status_util.cc",
-        "monitoring/thread_status_util_debug.cc",
-        "options/cf_options.cc",
-        "options/configurable.cc",
-        "options/customizable.cc",
-        "options/db_options.cc",
-        "options/options.cc",
-        "options/options_helper.cc",
-        "options/options_parser.cc",
-        "port/port_posix.cc",
-        "port/stack_trace.cc",
-        "port/win/env_default.cc",
-        "port/win/env_win.cc",
-        "port/win/io_win.cc",
-        "port/win/port_win.cc",
-        "port/win/win_logger.cc",
-        "port/win/win_thread.cc",
-        "table/adaptive/adaptive_table_factory.cc",
-        "table/block_based/binary_search_index_reader.cc",
-        "table/block_based/block.cc",
-        "table/block_based/block_based_table_builder.cc",
-        "table/block_based/block_based_table_factory.cc",
-        "table/block_based/block_based_table_iterator.cc",
-        "table/block_based/block_based_table_reader.cc",
-        "table/block_based/block_builder.cc",
-        "table/block_based/block_prefetcher.cc",
-        "table/block_based/block_prefix_index.cc",
-        "table/block_based/data_block_footer.cc",
-        "table/block_based/data_block_hash_index.cc",
-        "table/block_based/filter_block_reader_common.cc",
-        "table/block_based/filter_policy.cc",
-        "table/block_based/flush_block_policy.cc",
-        "table/block_based/full_filter_block.cc",
-        "table/block_based/hash_index_reader.cc",
-        "table/block_based/index_builder.cc",
-        "table/block_based/index_reader_common.cc",
-        "table/block_based/parsed_full_filter_block.cc",
-        "table/block_based/partitioned_filter_block.cc",
-        "table/block_based/partitioned_index_iterator.cc",
-        "table/block_based/partitioned_index_reader.cc",
-        "table/block_based/reader_common.cc",
-        "table/block_based/uncompression_dict_reader.cc",
-        "table/block_fetcher.cc",
-        "table/cuckoo/cuckoo_table_builder.cc",
-        "table/cuckoo/cuckoo_table_factory.cc",
-        "table/cuckoo/cuckoo_table_reader.cc",
-        "table/format.cc",
-        "table/get_context.cc",
-        "table/iterator.cc",
-        "table/merging_iterator.cc",
-        "table/meta_blocks.cc",
-        "table/persistent_cache_helper.cc",
-        "table/plain/plain_table_bloom.cc",
-        "table/plain/plain_table_builder.cc",
-        "table/plain/plain_table_factory.cc",
-        "table/plain/plain_table_index.cc",
-        "table/plain/plain_table_key_coding.cc",
-        "table/plain/plain_table_reader.cc",
-        "table/sst_file_dumper.cc",
-        "table/sst_file_reader.cc",
-        "table/sst_file_writer.cc",
-        "table/table_factory.cc",
-        "table/table_properties.cc",
-        "table/two_level_iterator.cc",
-        "table/unique_id.cc",
-        "test_util/sync_point.cc",
-        "test_util/sync_point_impl.cc",
-        "test_util/transaction_test_util.cc",
-        "tools/dump/db_dump_tool.cc",
-        "tools/io_tracer_parser_tool.cc",
-        "tools/ldb_cmd.cc",
-        "tools/ldb_tool.cc",
-        "tools/sst_dump_tool.cc",
-        "trace_replay/block_cache_tracer.cc",
-        "trace_replay/io_tracer.cc",
-        "trace_replay/trace_record.cc",
-        "trace_replay/trace_record_handler.cc",
-        "trace_replay/trace_record_result.cc",
-        "trace_replay/trace_replay.cc",
-        "util/async_file_reader.cc",
-        "util/build_version.cc",
-        "util/cleanable.cc",
-        "util/coding.cc",
-        "util/compaction_job_stats_impl.cc",
-        "util/comparator.cc",
-        "util/compression.cc",
-        "util/compression_context_cache.cc",
-        "util/concurrent_task_limiter_impl.cc",
-        "util/crc32c.cc",
-        "util/crc32c_arm64.cc",
-        "util/dynamic_bloom.cc",
-        "util/file_checksum_helper.cc",
-        "util/hash.cc",
-        "util/murmurhash.cc",
-        "util/random.cc",
-        "util/rate_limiter.cc",
-        "util/ribbon_config.cc",
-        "util/slice.cc",
-        "util/status.cc",
-        "util/stderr_logger.cc",
-        "util/string_util.cc",
-        "util/thread_local.cc",
-        "util/threadpool_imp.cc",
-        "util/xxhash.cc",
-        "utilities/agg_merge/agg_merge.cc",
-        "utilities/backup/backup_engine.cc",
-        "utilities/blob_db/blob_compaction_filter.cc",
-        "utilities/blob_db/blob_db.cc",
-        "utilities/blob_db/blob_db_impl.cc",
-        "utilities/blob_db/blob_db_impl_filesnapshot.cc",
-        "utilities/blob_db/blob_dump_tool.cc",
-        "utilities/blob_db/blob_file.cc",
-        "utilities/cache_dump_load.cc",
-        "utilities/cache_dump_load_impl.cc",
-        "utilities/cassandra/cassandra_compaction_filter.cc",
-        "utilities/cassandra/format.cc",
-        "utilities/cassandra/merge_operator.cc",
-        "utilities/checkpoint/checkpoint_impl.cc",
-        "utilities/compaction_filters.cc",
-        "utilities/compaction_filters/remove_emptyvalue_compactionfilter.cc",
-        "utilities/convenience/info_log_finder.cc",
-        "utilities/counted_fs.cc",
-        "utilities/debug.cc",
-        "utilities/env_mirror.cc",
-        "utilities/env_timed.cc",
-        "utilities/fault_injection_env.cc",
-        "utilities/fault_injection_fs.cc",
-        "utilities/fault_injection_secondary_cache.cc",
-        "utilities/leveldb_options/leveldb_options.cc",
-        "utilities/memory/memory_util.cc",
-        "utilities/merge_operators.cc",
-        "utilities/merge_operators/bytesxor.cc",
-        "utilities/merge_operators/max.cc",
-        "utilities/merge_operators/put.cc",
-        "utilities/merge_operators/sortlist.cc",
-        "utilities/merge_operators/string_append/stringappend.cc",
-        "utilities/merge_operators/string_append/stringappend2.cc",
-        "utilities/merge_operators/uint64add.cc",
-        "utilities/object_registry.cc",
-        "utilities/option_change_migration/option_change_migration.cc",
-        "utilities/options/options_util.cc",
-        "utilities/persistent_cache/block_cache_tier.cc",
-        "utilities/persistent_cache/block_cache_tier_file.cc",
-        "utilities/persistent_cache/block_cache_tier_metadata.cc",
-        "utilities/persistent_cache/persistent_cache_tier.cc",
-        "utilities/persistent_cache/volatile_tier_impl.cc",
-        "utilities/simulator_cache/cache_simulator.cc",
-        "utilities/simulator_cache/sim_cache.cc",
-        "utilities/table_properties_collectors/compact_on_deletion_collector.cc",
-        "utilities/trace/file_trace_reader_writer.cc",
-        "utilities/trace/replayer_impl.cc",
-        "utilities/transactions/lock/lock_manager.cc",
-        "utilities/transactions/lock/point/point_lock_manager.cc",
-        "utilities/transactions/lock/point/point_lock_tracker.cc",
-        "utilities/transactions/lock/range/range_tree/lib/locktree/concurrent_tree.cc",
-        "utilities/transactions/lock/range/range_tree/lib/locktree/keyrange.cc",
-        "utilities/transactions/lock/range/range_tree/lib/locktree/lock_request.cc",
-        "utilities/transactions/lock/range/range_tree/lib/locktree/locktree.cc",
-        "utilities/transactions/lock/range/range_tree/lib/locktree/manager.cc",
-        "utilities/transactions/lock/range/range_tree/lib/locktree/range_buffer.cc",
-        "utilities/transactions/lock/range/range_tree/lib/locktree/treenode.cc",
-        "utilities/transactions/lock/range/range_tree/lib/locktree/txnid_set.cc",
-        "utilities/transactions/lock/range/range_tree/lib/locktree/wfg.cc",
-        "utilities/transactions/lock/range/range_tree/lib/standalone_port.cc",
-        "utilities/transactions/lock/range/range_tree/lib/util/dbt.cc",
-        "utilities/transactions/lock/range/range_tree/lib/util/memarena.cc",
-        "utilities/transactions/lock/range/range_tree/range_tree_lock_manager.cc",
-        "utilities/transactions/lock/range/range_tree/range_tree_lock_tracker.cc",
-        "utilities/transactions/optimistic_transaction.cc",
-        "utilities/transactions/optimistic_transaction_db_impl.cc",
-        "utilities/transactions/pessimistic_transaction.cc",
-        "utilities/transactions/pessimistic_transaction_db.cc",
-        "utilities/transactions/snapshot_checker.cc",
-        "utilities/transactions/transaction_base.cc",
-        "utilities/transactions/transaction_db_mutex_impl.cc",
-        "utilities/transactions/transaction_util.cc",
-        "utilities/transactions/write_prepared_txn.cc",
-        "utilities/transactions/write_prepared_txn_db.cc",
-        "utilities/transactions/write_unprepared_txn.cc",
-        "utilities/transactions/write_unprepared_txn_db.cc",
-        "utilities/ttl/db_ttl_impl.cc",
-        "utilities/wal_filter.cc",
-        "utilities/write_batch_with_index/write_batch_with_index.cc",
-        "utilities/write_batch_with_index/write_batch_with_index_internal.cc",
-    ], deps=[
-        "//folly/container:f14_hash",
-        "//folly/experimental/coro:blocking_wait",
-        "//folly/experimental/coro:collect",
-        "//folly/experimental/coro:coroutine",
-        "//folly/experimental/coro:task",
-        "//folly/synchronization:distributed_mutex",
-    ], headers=None, link_whole=True, extra_test_libs=False)
+cpp_library_wrapper(name="rocksdb_whole_archive_lib", srcs=[], deps=[":rocksdb_lib"], headers=None, link_whole=True, extra_test_libs=False)
 
 cpp_library_wrapper(name="rocksdb_test_lib", srcs=[
         "db/db_test_util.cc",
         "db/db_with_timestamp_test_util.cc",
         "table/mock_table.cc",
         "test_util/mock_time_env.cc",
+        "test_util/secondary_cache_test_util.cc",
         "test_util/testharness.cc",
         "test_util/testutil.cc",
         "tools/block_cache_analyzer/block_cache_trace_analyzer.cc",

USERS.md

@@ -15,6 +15,9 @@ At Facebook, we use RocksDB as storage engines in multiple data management servi
 
 [2] https://code.facebook.com/posts/357056558062811/logdevice-a-distributed-data-store-for-logs/
 
+## Microsoft
+The Bing search engine from Microsoft uses RocksDB as the storage engine for its web data platform: https://blogs.bing.com/Engineering-Blog/october-2021/RocksDB-in-Microsoft-Bing
+
 ## LinkedIn
 Two different use cases at Linkedin are using RocksDB as a storage engine:
 
@@ -26,6 +29,9 @@ Learn more about those use cases in a Tech Talk by Ankit Gupta and Naveen Somasu
 ## Yahoo
 Yahoo is using RocksDB as a storage engine for their biggest distributed data store Sherpa. Learn more about it here: http://yahooeng.tumblr.com/post/120730204806/sherpa-scales-new-heights
 
+## Tencent
+[PaxosStore](https://github.com/Tencent/paxosstore) is a distributed database supporting WeChat. It uses RocksDB as its storage engine.
+
 ## Baidu
 [Apache Doris](http://doris.apache.org/master/en/) is a MPP analytical database engine released by Baidu. It [uses RocksDB](http://doris.apache.org/master/en/administrator-guide/operation/tablet-meta-tool.html) to manage its tablet's metadata.
 
@@ -79,6 +85,9 @@ quasardb uses a heavily tuned RocksDB as its persistence layer.
 ## TiKV
 [TiKV](https://github.com/pingcap/tikv) is a GEO-replicated, high-performance, distributed, transactional key-value database. TiKV is powered by Rust and Raft. TiKV uses RocksDB as its persistence layer.
 
+## Apache Spark
+[Spark Structured Streaming](https://docs.databricks.com/structured-streaming/rocksdb-state-store.html) uses RocksDB as the local state store.
+
 ## Apache Flink
 [Apache Flink](https://flink.apache.org/news/2016/03/08/release-1.0.0.html) uses RocksDB to store state locally on a machine.
 

buckifier/buckify_rocksdb.py

@@ -26,7 +26,7 @@ from util import ColorString
 # $python3 buckifier/buckify_rocksdb.py \
 #        '{"fake": {
 #                      "extra_deps": [":test_dep", "//fakes/module:mock1"],
-#                      "extra_compiler_flags": ["-DROCKSDB_LITE", "-Os"]
+#                      "extra_compiler_flags": ["-DFOO_BAR", "-Os"]
 #                  }
 #         }'
 # (Generated TARGETS file has test_dep and mock1 as dependencies for RocksDB
@@ -154,16 +154,9 @@ def generate_targets(repo_path, deps_map):
     # rocksdb_whole_archive_lib
     TARGETS.add_library(
         "rocksdb_whole_archive_lib",
-        src_mk["LIB_SOURCES"] +
-        # always add range_tree, it's only excluded on ppc64, which we don't use internally
-        src_mk["RANGE_TREE_SOURCES"] + src_mk["TOOL_LIB_SOURCES"],
+        [],
         deps=[
-            "//folly/container:f14_hash",
-            "//folly/experimental/coro:blocking_wait",
-            "//folly/experimental/coro:collect",
-            "//folly/experimental/coro:coroutine",
-            "//folly/experimental/coro:task",
-            "//folly/synchronization:distributed_mutex",
+            ":rocksdb_lib",
         ],
         headers=None,
         extra_external_deps="",

build_tools/build_detect_platform

@@ -63,13 +63,7 @@ if [ -z "$ROCKSDB_NO_FBCODE" -a -d /mnt/gvfs/third-party ]; then
     if [ "$LIB_MODE" == "shared" ]; then
       PIC_BUILD=1
     fi
-    if [ -n "$ROCKSDB_FBCODE_BUILD_WITH_PLATFORM010" ]; then
-      source "$PWD/build_tools/fbcode_config_platform010.sh"
-    elif [ -n "$ROCKSDB_FBCODE_BUILD_WITH_PLATFORM009" ]; then
-      source "$PWD/build_tools/fbcode_config_platform009.sh"
-    else
-      source "$PWD/build_tools/fbcode_config_platform009.sh"
-    fi
+    source "$PWD/build_tools/fbcode_config_platform010.sh"
 fi
 
 # Delete existing output, if it exists
@@ -154,7 +148,7 @@ case "$TARGET_OS" in
         ;;
     IOS)
         PLATFORM=IOS
-        COMMON_FLAGS="$COMMON_FLAGS -DOS_MACOSX -DIOS_CROSS_COMPILE -DROCKSDB_LITE"
+        COMMON_FLAGS="$COMMON_FLAGS -DOS_MACOSX -DIOS_CROSS_COMPILE "
         PLATFORM_SHARED_EXT=dylib
         PLATFORM_SHARED_LDFLAGS="-dynamiclib -install_name "
         CROSS_COMPILE=true

build_tools/dependencies_platform009.sh

@@ -1,22 +0,0 @@
-# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
-GCC_BASE=/mnt/gvfs/third-party2/gcc/1795efe5f06778c15a92c8f9a2aba5dc496d9d4d/9.x/centos7-native/3bed279
-CLANG_BASE=/mnt/gvfs/third-party2/llvm-fb/7318eaac22659b6ff2fe43918e4b69fd0772a8a7/9.0.0/platform009/651ee30
-LIBGCC_BASE=/mnt/gvfs/third-party2/libgcc/4959b39cfbe5965a37c861c4c327fa7c5c759b87/9.x/platform009/9202ce7
-GLIBC_BASE=/mnt/gvfs/third-party2/glibc/45ce3375cdc77ecb2520bbf8f0ecddd3f98efd7a/2.30/platform009/f259413
-SNAPPY_BASE=/mnt/gvfs/third-party2/snappy/be4de3205e029101b18aa8103daa696c2bef3b19/1.1.3/platform009/7f3b187
-ZLIB_BASE=/mnt/gvfs/third-party2/zlib/3c160ac5c67e257501e24c6c1d00ad5e01d73db6/1.2.8/platform009/7f3b187
-BZIP2_BASE=/mnt/gvfs/third-party2/bzip2/73a237ac5bc0a5f5d67b39b8d253cfebaab88684/1.0.6/platform009/7f3b187
-LZ4_BASE=/mnt/gvfs/third-party2/lz4/6ca38d3c390be2774d61a300f151464bbd632d62/1.9.1/platform009/7f3b187
-ZSTD_BASE=/mnt/gvfs/third-party2/zstd/64c58a207d2495e83abc57a500a956df09b79a7c/1.4.x/platform009/ba86d1f
-GFLAGS_BASE=/mnt/gvfs/third-party2/gflags/824d0a8a5abb5b121afd1b35fc3896407ea50092/2.2.0/platform009/7f3b187
-JEMALLOC_BASE=/mnt/gvfs/third-party2/jemalloc/b62912d333ef33f9760efa6219dbe3fe6abb3b0e/master/platform009/c305944
-NUMA_BASE=/mnt/gvfs/third-party2/numa/0af65f71e23a67bf65dc91b11f95caa39325c432/2.0.11/platform009/7f3b187
-LIBUNWIND_BASE=/mnt/gvfs/third-party2/libunwind/02486dac347645d31dce116f44e1de3177315be2/1.4/platform009/5191652
-TBB_BASE=/mnt/gvfs/third-party2/tbb/2e0ec671e550bfca347300bf3f789d9c0fff24ad/2018_U5/platform009/7f3b187
-LIBURING_BASE=/mnt/gvfs/third-party2/liburing/70dbd9cfee63a25611417d09433a86d7711b3990/20200729/platform009/7f3b187
-KERNEL_HEADERS_BASE=/mnt/gvfs/third-party2/kernel-headers/32b8a2407b634df3f8f948ba373fc4acc6a18296/fb/platform009/da39a3e
-BINUTILS_BASE=/mnt/gvfs/third-party2/binutils/08634589372fa5f237bfd374e8c644a8364e78c1/2.32/platform009/ba86d1f/
-VALGRIND_BASE=/mnt/gvfs/third-party2/valgrind/6ae525939ad02e5e676855082fbbc7828dbafeac/3.15.0/platform009/7f3b187
-LUA_BASE=/mnt/gvfs/third-party2/lua/162efd9561a3d21f6869f4814011e9cf1b3ff4dc/5.3.4/platform009/a6271c4
-BENCHMARK_BASE=/mnt/gvfs/third-party2/benchmark/30bf49ad6414325e17f3425b0edcb64239427ae3/1.6.1/platform009/7f3b187
-GLOG_BASE=/mnt/gvfs/third-party2/glog/32d751bd5673375b438158717ab6a57c1cc57e3d/0.3.2_fb/platform009/10a364d

build_tools/fbcode_config_platform009.sh

@@ -1,170 +0,0 @@
-#!/bin/sh
-# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
-#
-# Set environment variables so that we can compile rocksdb using
-# fbcode settings.  It uses the latest g++ and clang compilers and also
-# uses jemalloc
-# Environment variables that change the behavior of this script:
-# PIC_BUILD -- if true, it will only take pic versions of libraries from fbcode. libraries that don't have pic variant will not be included
-
-
-BASEDIR=`dirname $BASH_SOURCE`
-source "$BASEDIR/dependencies_platform009.sh"
-
-CFLAGS=""
-
-# libgcc
-LIBGCC_INCLUDE="$LIBGCC_BASE/include/c++/9.3.0 -I $LIBGCC_BASE/include/c++/9.3.0/backward"
-LIBGCC_LIBS=" -L $LIBGCC_BASE/lib"
-
-# glibc
-GLIBC_INCLUDE="$GLIBC_BASE/include"
-GLIBC_LIBS=" -L $GLIBC_BASE/lib"
-
-if test -z $PIC_BUILD; then
-  MAYBE_PIC=
-else
-  MAYBE_PIC=_pic
-fi
-
-if ! test $ROCKSDB_DISABLE_SNAPPY; then
-  # snappy
-  SNAPPY_INCLUDE=" -I $SNAPPY_BASE/include/"
-  SNAPPY_LIBS=" $SNAPPY_BASE/lib/libsnappy${MAYBE_PIC}.a"
-  CFLAGS+=" -DSNAPPY"
-fi
-
-if ! test $ROCKSDB_DISABLE_ZLIB; then
-  # location of zlib headers and libraries
-  ZLIB_INCLUDE=" -I $ZLIB_BASE/include/"
-  ZLIB_LIBS=" $ZLIB_BASE/lib/libz${MAYBE_PIC}.a"
-  CFLAGS+=" -DZLIB"
-fi
-
-if ! test $ROCKSDB_DISABLE_BZIP; then
-  # location of bzip headers and libraries
-  BZIP_INCLUDE=" -I $BZIP2_BASE/include/"
-  BZIP_LIBS=" $BZIP2_BASE/lib/libbz2${MAYBE_PIC}.a"
-  CFLAGS+=" -DBZIP2"
-fi
-
-if ! test $ROCKSDB_DISABLE_LZ4; then
-  LZ4_INCLUDE=" -I $LZ4_BASE/include/"
-  LZ4_LIBS=" $LZ4_BASE/lib/liblz4${MAYBE_PIC}.a"
-  CFLAGS+=" -DLZ4"
-fi
-
-if ! test $ROCKSDB_DISABLE_ZSTD; then
-  ZSTD_INCLUDE=" -I $ZSTD_BASE/include/"
-  ZSTD_LIBS=" $ZSTD_BASE/lib/libzstd${MAYBE_PIC}.a"
-  CFLAGS+=" -DZSTD"
-fi
-
-# location of gflags headers and libraries
-GFLAGS_INCLUDE=" -I $GFLAGS_BASE/include/"
-GFLAGS_LIBS=" $GFLAGS_BASE/lib/libgflags${MAYBE_PIC}.a"
-CFLAGS+=" -DGFLAGS=gflags"
-
-BENCHMARK_INCLUDE=" -I $BENCHMARK_BASE/include/"
-BENCHMARK_LIBS=" $BENCHMARK_BASE/lib/libbenchmark${MAYBE_PIC}.a"
-
-GLOG_INCLUDE=" -I $GLOG_BASE/include/"
-GLOG_LIBS=" $GLOG_BASE/lib/libglog${MAYBE_PIC}.a"
-
-# location of jemalloc
-JEMALLOC_INCLUDE=" -I $JEMALLOC_BASE/include/"
-JEMALLOC_LIB=" $JEMALLOC_BASE/lib/libjemalloc${MAYBE_PIC}.a"
-
-# location of numa
-NUMA_INCLUDE=" -I $NUMA_BASE/include/"
-NUMA_LIB=" $NUMA_BASE/lib/libnuma${MAYBE_PIC}.a"
-CFLAGS+=" -DNUMA"
-
-# location of libunwind
-LIBUNWIND="$LIBUNWIND_BASE/lib/libunwind${MAYBE_PIC}.a"
-
-# location of TBB
-TBB_INCLUDE=" -isystem $TBB_BASE/include/"
-TBB_LIBS="$TBB_BASE/lib/libtbb${MAYBE_PIC}.a"
-CFLAGS+=" -DTBB"
-
-# location of LIBURING
-LIBURING_INCLUDE=" -isystem $LIBURING_BASE/include/"
-LIBURING_LIBS="$LIBURING_BASE/lib/liburing${MAYBE_PIC}.a"
-CFLAGS+=" -DLIBURING"
-
-test "$USE_SSE" || USE_SSE=1
-export USE_SSE
-test "$PORTABLE" || PORTABLE=1
-export PORTABLE
-
-BINUTILS="$BINUTILS_BASE/bin"
-AR="$BINUTILS/ar"
-AS="$BINUTILS/as"
-
-DEPS_INCLUDE="$SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP_INCLUDE $LZ4_INCLUDE $ZSTD_INCLUDE $GFLAGS_INCLUDE $NUMA_INCLUDE $TBB_INCLUDE $LIBURING_INCLUDE $BENCHMARK_INCLUDE $GLOG_INCLUDE"
-
-STDLIBS="-L $GCC_BASE/lib64"
-
-CLANG_BIN="$CLANG_BASE/bin"
-CLANG_LIB="$CLANG_BASE/lib"
-CLANG_SRC="$CLANG_BASE/../../src"
-
-CLANG_ANALYZER="$CLANG_BIN/clang++"
-CLANG_SCAN_BUILD="$CLANG_SRC/llvm/clang/tools/scan-build/bin/scan-build"
-
-if [ -z "$USE_CLANG" ]; then
-  # gcc
-  CC="$GCC_BASE/bin/gcc"
-  CXX="$GCC_BASE/bin/g++"
-  AR="$GCC_BASE/bin/gcc-ar"
-
-  CFLAGS+=" -B$BINUTILS"
-  CFLAGS+=" -isystem $LIBGCC_INCLUDE"
-  CFLAGS+=" -isystem $GLIBC_INCLUDE"
-  JEMALLOC=1
-else
-  # clang
-  CLANG_INCLUDE="$CLANG_LIB/clang/stable/include"
-  CC="$CLANG_BIN/clang"
-  CXX="$CLANG_BIN/clang++"
-  AR="$CLANG_BIN/llvm-ar"
-
-  KERNEL_HEADERS_INCLUDE="$KERNEL_HEADERS_BASE/include"
-
-  CFLAGS+=" -B$BINUTILS -nostdinc -nostdlib"
-  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/9.x "
-  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/9.x/x86_64-facebook-linux "
-  CFLAGS+=" -isystem $GLIBC_INCLUDE"
-  CFLAGS+=" -isystem $LIBGCC_INCLUDE"
-  CFLAGS+=" -isystem $CLANG_INCLUDE"
-  CFLAGS+=" -isystem $KERNEL_HEADERS_INCLUDE/linux "
-  CFLAGS+=" -isystem $KERNEL_HEADERS_INCLUDE "
-  CFLAGS+=" -Wno-expansion-to-defined "
-  CXXFLAGS="-nostdinc++"
-fi
-
-CFLAGS+=" $DEPS_INCLUDE"
-CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE -DROCKSDB_RANGESYNC_PRESENT -DROCKSDB_SCHED_GETCPU_PRESENT -DHAVE_SSE42 -DROCKSDB_IOURING_PRESENT"
-CXXFLAGS+=" $CFLAGS"
-
-EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB $TBB_LIBS $LIBURING_LIBS $BENCHMARK_LIBS"
-EXEC_LDFLAGS+=" -Wl,--dynamic-linker,/usr/local/fbcode/platform009/lib/ld.so"
-EXEC_LDFLAGS+=" $LIBUNWIND"
-EXEC_LDFLAGS+=" -Wl,-rpath=/usr/local/fbcode/platform009/lib"
-EXEC_LDFLAGS+=" -Wl,-rpath=$GCC_BASE/lib64"
-# required by libtbb
-EXEC_LDFLAGS+=" -ldl"
-
-PLATFORM_LDFLAGS="$LIBGCC_LIBS $GLIBC_LIBS $STDLIBS -lgcc -lstdc++"
-PLATFORM_LDFLAGS+=" -B$BINUTILS"
-
-EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $TBB_LIBS $LIBURING_LIBS $BENCHMARK_LIBS"
-
-VALGRIND_VER="$VALGRIND_BASE/bin/"
-
-# lua not supported because it's on track for deprecation, I think
-LUA_PATH=
-LUA_LIB=
-
-export CC CXX AR AS CFLAGS CXXFLAGS EXEC_LDFLAGS EXEC_LDFLAGS_SHARED VALGRIND_VER JEMALLOC_LIB JEMALLOC_INCLUDE CLANG_ANALYZER CLANG_SCAN_BUILD LUA_PATH LUA_LIB

build_tools/regression_build_test.sh

@@ -360,7 +360,7 @@ function send_to_ods {
     echo >&2 "ERROR: Key $key doesn't have a value."
     return
   fi
-  curl --silent "https://www.intern.facebook.com/intern/agent/ods_set.php?entity=rocksdb_build&key=$key&value=$value" \
+  curl --silent "https://www.facebook.com/intern/agent/ods_set.php?entity=rocksdb_build&key=$key&value=$value" \
     --connect-timeout 60
 }
 

build_tools/update_dependencies.sh

@@ -104,46 +104,3 @@ get_lib_base valgrind   LATEST platform010
 get_lib_base lua        5.3.4  platform010
 
 git diff $OUTPUT
-
-
-###########################################################
-#                platform009 dependencies                 #
-###########################################################
-
-OUTPUT="$BASEDIR/dependencies_platform009.sh"
-
-rm -f "$OUTPUT"
-touch "$OUTPUT"
-
-echo "Writing dependencies to $OUTPUT"
-
-# Compilers locations
-GCC_BASE=`readlink -f $TP2_LATEST/gcc/9.x/centos7-native/*/`
-CLANG_BASE=`readlink -f $TP2_LATEST/llvm-fb/9.0.0/platform009/*/`
-
-log_header
-log_variable GCC_BASE
-log_variable CLANG_BASE
-
-# Libraries locations
-get_lib_base libgcc     9.x     platform009
-get_lib_base glibc      2.30    platform009
-get_lib_base snappy     LATEST  platform009
-get_lib_base zlib       LATEST  platform009
-get_lib_base bzip2      LATEST  platform009
-get_lib_base lz4        LATEST  platform009
-get_lib_base zstd       LATEST  platform009
-get_lib_base gflags     LATEST  platform009
-get_lib_base jemalloc   LATEST  platform009
-get_lib_base numa       LATEST  platform009
-get_lib_base libunwind  LATEST  platform009
-get_lib_base tbb        2018_U5 platform009
-get_lib_base liburing   LATEST  platform009
-get_lib_base benchmark  LATEST  platform009
-
-get_lib_base kernel-headers fb platform009
-get_lib_base binutils   LATEST centos7-native
-get_lib_base valgrind   LATEST platform009
-get_lib_base lua        5.3.4  platform009
-
-git diff $OUTPUT

cache/cache.cc

@@ -16,7 +16,8 @@
 #include "util/string_util.h"
 
 namespace ROCKSDB_NAMESPACE {
-#ifndef ROCKSDB_LITE
+const Cache::CacheItemHelper kNoopCacheItemHelper{};
+
 static std::unordered_map<std::string, OptionTypeInfo>
     lru_cache_options_type_info = {
         {"capacity",
@@ -64,7 +65,6 @@ static std::unordered_map<std::string, OptionTypeInfo>
           OptionType::kBoolean, OptionVerificationType::kNormal,
           OptionTypeFlags::kMutable}},
 };
-#endif  // ROCKSDB_LITE
 
 Status SecondaryCache::CreateFromString(
     const ConfigOptions& config_options, const std::string& value,
@@ -75,7 +75,6 @@ Status SecondaryCache::CreateFromString(
     Status status;
     std::shared_ptr<SecondaryCache> sec_cache;
 
-#ifndef ROCKSDB_LITE
     CompressedSecondaryCacheOptions sec_cache_opts;
     status = OptionTypeInfo::ParseStruct(config_options, "",
                                          &comp_sec_cache_options_type_info, "",
@@ -84,19 +83,13 @@ Status SecondaryCache::CreateFromString(
       sec_cache = NewCompressedSecondaryCache(sec_cache_opts);
     }
 
-#else
-    (void)config_options;
-    status = Status::NotSupported(
-        "Cannot load compressed secondary cache in LITE mode ", args);
-#endif  //! ROCKSDB_LITE
 
     if (status.ok()) {
       result->swap(sec_cache);
     }
     return status;
   } else {
-    return LoadSharedObject<SecondaryCache>(config_options, value, nullptr,
-                                            result);
+    return LoadSharedObject<SecondaryCache>(config_options, value, result);
   }
 }
 
@@ -108,7 +101,6 @@ Status Cache::CreateFromString(const ConfigOptions& config_options,
   if (value.find('=') == std::string::npos) {
     cache = NewLRUCache(ParseSizeT(value));
   } else {
-#ifndef ROCKSDB_LITE
     LRUCacheOptions cache_opts;
     status = OptionTypeInfo::ParseStruct(config_options, "",
                                          &lru_cache_options_type_info, "",
@@ -116,14 +108,51 @@ Status Cache::CreateFromString(const ConfigOptions& config_options,
     if (status.ok()) {
       cache = NewLRUCache(cache_opts);
     }
-#else
-    (void)config_options;
-    status = Status::NotSupported("Cannot load cache in LITE mode ", value);
-#endif  //! ROCKSDB_LITE
   }
   if (status.ok()) {
     result->swap(cache);
   }
   return status;
 }
+
+bool Cache::AsyncLookupHandle::IsReady() {
+  return pending_handle == nullptr || pending_handle->IsReady();
+}
+
+bool Cache::AsyncLookupHandle::IsPending() { return pending_handle != nullptr; }
+
+Cache::Handle* Cache::AsyncLookupHandle::Result() {
+  assert(!IsPending());
+  return result_handle;
+}
+
+void Cache::StartAsyncLookup(AsyncLookupHandle& async_handle) {
+  async_handle.found_dummy_entry = false;  // in case re-used
+  assert(!async_handle.IsPending());
+  async_handle.result_handle =
+      Lookup(async_handle.key, async_handle.helper, async_handle.create_context,
+             async_handle.priority, async_handle.stats);
+}
+
+Cache::Handle* Cache::Wait(AsyncLookupHandle& async_handle) {
+  WaitAll(&async_handle, 1);
+  return async_handle.Result();
+}
+
+void Cache::WaitAll(AsyncLookupHandle* async_handles, size_t count) {
+  for (size_t i = 0; i < count; ++i) {
+    if (async_handles[i].IsPending()) {
+      // If a pending handle gets here, it should be marked at "to be handled
+      // by a caller" by that caller erasing the pending_cache on it.
+      assert(async_handles[i].pending_cache == nullptr);
+    }
+  }
+}
+
+void Cache::SetEvictionCallback(EvictionCallback&& fn) {
+  // Overwriting non-empty with non-empty could indicate a bug
+  assert(!eviction_callback_ || !fn);
+  eviction_callback_ = std::move(fn);
+}
+
 }  // namespace ROCKSDB_NAMESPACE

cache/cache_bench_tool.cc

@@ -13,11 +13,10 @@
 #include <set>
 #include <sstream>
 
-#include "cache/fast_lru_cache.h"
 #include "db/db_impl/db_impl.h"
 #include "monitoring/histogram.h"
 #include "port/port.h"
-#include "rocksdb/cache.h"
+#include "rocksdb/advanced_cache.h"
 #include "rocksdb/convenience.h"
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
@@ -78,11 +77,9 @@ DEFINE_bool(lean, false,
             "If true, no additional computation is performed besides cache "
             "operations.");
 
-#ifndef ROCKSDB_LITE
 DEFINE_string(secondary_cache_uri, "",
               "Full URI for creating a custom secondary cache object");
 static class std::shared_ptr<ROCKSDB_NAMESPACE::SecondaryCache> secondary_cache;
-#endif  // ROCKSDB_LITE
 
 DEFINE_string(cache_type, "lru_cache", "Type of block cache.");
 
@@ -227,7 +224,7 @@ struct KeyGen {
   }
 };
 
-char* createValue(Random64& rnd) {
+Cache::ObjectPtr createValue(Random64& rnd) {
   char* rv = new char[FLAGS_value_bytes];
   // Fill with some filler data, and take some CPU time
   for (uint32_t i = 0; i < FLAGS_value_bytes; i += 8) {
@@ -237,28 +234,36 @@ char* createValue(Random64& rnd) {
 }
 
 // Callbacks for secondary cache
-size_t SizeFn(void* /*obj*/) { return FLAGS_value_bytes; }
+size_t SizeFn(Cache::ObjectPtr /*obj*/) { return FLAGS_value_bytes; }
 
-Status SaveToFn(void* obj, size_t /*offset*/, size_t size, void* out) {
-  memcpy(out, obj, size);
+Status SaveToFn(Cache::ObjectPtr from_obj, size_t /*from_offset*/,
+                size_t length, char* out) {
+  memcpy(out, from_obj, length);
   return Status::OK();
 }
 
-// Different deleters to simulate using deleter to gather
-// stats on the code origin and kind of cache entries.
-void deleter1(const Slice& /*key*/, void* value) {
-  delete[] static_cast<char*>(value);
-}
-void deleter2(const Slice& /*key*/, void* value) {
-  delete[] static_cast<char*>(value);
-}
-void deleter3(const Slice& /*key*/, void* value) {
+Status CreateFn(const Slice& data, Cache::CreateContext* /*context*/,
+                MemoryAllocator* /*allocator*/, Cache::ObjectPtr* out_obj,
+                size_t* out_charge) {
+  *out_obj = new char[data.size()];
+  memcpy(*out_obj, data.data(), data.size());
+  *out_charge = data.size();
+  return Status::OK();
+};
+
+void DeleteFn(Cache::ObjectPtr value, MemoryAllocator* /*alloc*/) {
   delete[] static_cast<char*>(value);
 }
 
-Cache::CacheItemHelper helper1(SizeFn, SaveToFn, deleter1);
-Cache::CacheItemHelper helper2(SizeFn, SaveToFn, deleter2);
-Cache::CacheItemHelper helper3(SizeFn, SaveToFn, deleter3);
+Cache::CacheItemHelper helper1_wos(CacheEntryRole::kDataBlock, DeleteFn);
+Cache::CacheItemHelper helper1(CacheEntryRole::kDataBlock, DeleteFn, SizeFn,
+                               SaveToFn, CreateFn, &helper1_wos);
+Cache::CacheItemHelper helper2_wos(CacheEntryRole::kIndexBlock, DeleteFn);
+Cache::CacheItemHelper helper2(CacheEntryRole::kIndexBlock, DeleteFn, SizeFn,
+                               SaveToFn, CreateFn, &helper2_wos);
+Cache::CacheItemHelper helper3_wos(CacheEntryRole::kFilterBlock, DeleteFn);
+Cache::CacheItemHelper helper3(CacheEntryRole::kFilterBlock, DeleteFn, SizeFn,
+                               SaveToFn, CreateFn, &helper3_wos);
 }  // namespace
 
 class CacheBench {
@@ -297,15 +302,10 @@ class CacheBench {
       cache_ = HyperClockCacheOptions(FLAGS_cache_size, FLAGS_value_bytes,
                                       FLAGS_num_shard_bits)
                    .MakeSharedCache();
-    } else if (FLAGS_cache_type == "fast_lru_cache") {
-      cache_ = NewFastLRUCache(
-          FLAGS_cache_size, FLAGS_value_bytes, FLAGS_num_shard_bits,
-          false /*strict_capacity_limit*/, kDefaultCacheMetadataChargePolicy);
     } else if (FLAGS_cache_type == "lru_cache") {
       LRUCacheOptions opts(FLAGS_cache_size, FLAGS_num_shard_bits,
                            false /* strict_capacity_limit */,
                            0.5 /* high_pri_pool_ratio */);
-#ifndef ROCKSDB_LITE
       if (!FLAGS_secondary_cache_uri.empty()) {
         Status s = SecondaryCache::CreateFromString(
             ConfigOptions(), FLAGS_secondary_cache_uri, &secondary_cache);
@@ -318,7 +318,6 @@ class CacheBench {
         }
         opts.secondary_cache = secondary_cache;
       }
-#endif  // ROCKSDB_LITE
 
       cache_ = NewLRUCache(opts);
     } else {
@@ -441,7 +440,7 @@ class CacheBench {
     uint64_t total_entry_count = 0;
     uint64_t table_occupancy = 0;
     uint64_t table_size = 0;
-    std::set<Cache::DeleterFn> deleters;
+    std::set<const Cache::CacheItemHelper*> helpers;
     StopWatchNano timer(clock);
 
     for (;;) {
@@ -466,7 +465,7 @@ class CacheBench {
                  << BytesToHumanString(static_cast<uint64_t>(
                         1.0 * total_charge / total_entry_count))
                  << "\n"
-                 << "Unique deleters: " << deleters.size() << "\n";
+                 << "Unique helpers: " << helpers.size() << "\n";
             *stats_report = ostr.str();
             return;
           }
@@ -482,14 +481,14 @@ class CacheBench {
       total_key_size = 0;
       total_charge = 0;
       total_entry_count = 0;
-      deleters.clear();
-      auto fn = [&](const Slice& key, void* /*value*/, size_t charge,
-                    Cache::DeleterFn deleter) {
+      helpers.clear();
+      auto fn = [&](const Slice& key, Cache::ObjectPtr /*value*/, size_t charge,
+                    const Cache::CacheItemHelper* helper) {
         total_key_size += key.size();
         total_charge += charge;
         ++total_entry_count;
-        // Something slightly more expensive as in (future) stats by category
-        deleters.insert(deleter);
+        // Something slightly more expensive as in stats by category
+        helpers.insert(helper);
       };
       timer.Start();
       Cache::ApplyToAllEntriesOptions opts;
@@ -538,14 +537,6 @@ class CacheBench {
     for (uint64_t i = 0; i < FLAGS_ops_per_thread; i++) {
       Slice key = gen.GetRand(thread->rnd, max_key_, max_log_);
       uint64_t random_op = thread->rnd.Next();
-      Cache::CreateCallback create_cb = [](const void* buf, size_t size,
-                                           void** out_obj,
-                                           size_t* charge) -> Status {
-        *out_obj = reinterpret_cast<void*>(new char[size]);
-        memcpy(*out_obj, buf, size);
-        *charge = size;
-        return Status::OK();
-      };
 
       timer.Start();
 
@@ -555,8 +546,8 @@ class CacheBench {
           handle = nullptr;
         }
         // do lookup
-        handle = cache_->Lookup(key, &helper2, create_cb, Cache::Priority::LOW,
-                                true);
+        handle = cache_->Lookup(key, &helper2, /*context*/ nullptr,
+                                Cache::Priority::LOW);
         if (handle) {
           if (!FLAGS_lean) {
             // do something with the data
@@ -584,8 +575,8 @@ class CacheBench {
           handle = nullptr;
         }
         // do lookup
-        handle = cache_->Lookup(key, &helper2, create_cb, Cache::Priority::LOW,
-                                true);
+        handle = cache_->Lookup(key, &helper2, /*context*/ nullptr,
+                                Cache::Priority::LOW);
         if (handle) {
           if (!FLAGS_lean) {
             // do something with the data

cache/cache_entry_roles.cc

@@ -101,34 +101,4 @@ std::string BlockCacheEntryStatsMapKeys::UsedPercent(CacheEntryRole role) {
   return GetPrefixedCacheEntryRoleName(kPrefix, role);
 }
 
-namespace {
-
-struct Registry {
-  std::mutex mutex;
-  UnorderedMap<Cache::DeleterFn, CacheEntryRole> role_map;
-  void Register(Cache::DeleterFn fn, CacheEntryRole role) {
-    std::lock_guard<std::mutex> lock(mutex);
-    role_map[fn] = role;
-  }
-  UnorderedMap<Cache::DeleterFn, CacheEntryRole> Copy() {
-    std::lock_guard<std::mutex> lock(mutex);
-    return role_map;
-  }
-};
-
-Registry& GetRegistry() {
-  STATIC_AVOID_DESTRUCTION(Registry, registry);
-  return registry;
-}
-
-}  // namespace
-
-void RegisterCacheDeleterRole(Cache::DeleterFn fn, CacheEntryRole role) {
-  GetRegistry().Register(fn, role);
-}
-
-UnorderedMap<Cache::DeleterFn, CacheEntryRole> CopyCacheDeleterRoleMap() {
-  return GetRegistry().Copy();
-}
-
 }  // namespace ROCKSDB_NAMESPACE

cache/cache_entry_roles.h

@@ -7,11 +7,8 @@
 
 #include <array>
 #include <cstdint>
-#include <memory>
-#include <type_traits>
 
 #include "rocksdb/cache.h"
-#include "util/hash_containers.h"
 
 namespace ROCKSDB_NAMESPACE {
 
@@ -20,84 +17,4 @@ extern std::array<std::string, kNumCacheEntryRoles>
 extern std::array<std::string, kNumCacheEntryRoles>
     kCacheEntryRoleToHyphenString;
 
-// To associate cache entries with their role, we use a hack on the
-// existing Cache interface. Because the deleter of an entry can authenticate
-// the code origin of an entry, we can elaborate the choice of deleter to
-// also encode role information, without inferring false role information
-// from entries not choosing to encode a role.
-//
-// The rest of this file is for handling mappings between deleters and
-// roles.
-
-// To infer a role from a deleter, the deleter must be registered. This
-// can be done "manually" with this function. This function is thread-safe,
-// and the registration mappings go into private but static storage. (Note
-// that DeleterFn is a function pointer, not std::function. Registrations
-// should not be too many.)
-void RegisterCacheDeleterRole(Cache::DeleterFn fn, CacheEntryRole role);
-
-// Gets a copy of the registered deleter -> role mappings. This is the only
-// function for reading the mappings made with RegisterCacheDeleterRole.
-// Why only this interface for reading?
-// * This function has to be thread safe, which could incur substantial
-// overhead. We should not pay this overhead for every deleter look-up.
-// * This is suitable for preparing for batch operations, like with
-// CacheEntryStatsCollector.
-// * The number of mappings should be sufficiently small (dozens).
-UnorderedMap<Cache::DeleterFn, CacheEntryRole> CopyCacheDeleterRoleMap();
-
-// ************************************************************** //
-// An automatic registration infrastructure. This enables code
-// to simply ask for a deleter associated with a particular type
-// and role, and registration is automatic. In a sense, this is
-// a small dependency injection infrastructure, because linking
-// in new deleter instantiations is essentially sufficient for
-// making stats collection (using CopyCacheDeleterRoleMap) aware
-// of them.
-
-namespace cache_entry_roles_detail {
-
-template <typename T, CacheEntryRole R>
-struct RegisteredDeleter {
-  RegisteredDeleter() { RegisterCacheDeleterRole(Delete, R); }
-
-  // These have global linkage to help ensure compiler optimizations do not
-  // break uniqueness for each <T,R>
-  static void Delete(const Slice& /* key */, void* value) {
-    // Supports T == Something[], unlike delete operator
-    std::default_delete<T>()(
-        static_cast<typename std::remove_extent<T>::type*>(value));
-  }
-};
-
-template <CacheEntryRole R>
-struct RegisteredNoopDeleter {
-  RegisteredNoopDeleter() { RegisterCacheDeleterRole(Delete, R); }
-
-  static void Delete(const Slice& /* key */, void* /* value */) {
-    // Here was `assert(value == nullptr);` but we can also put pointers
-    // to static data in Cache, for testing at least.
-  }
-};
-
-}  // namespace cache_entry_roles_detail
-
-// Get an automatically registered deleter for value type T and role R.
-// Based on C++ semantics, registration is invoked exactly once in a
-// thread-safe way on first call to this function, for each <T, R>.
-template <typename T, CacheEntryRole R>
-Cache::DeleterFn GetCacheEntryDeleterForRole() {
-  static cache_entry_roles_detail::RegisteredDeleter<T, R> reg;
-  return reg.Delete;
-}
-
-// Get an automatically registered no-op deleter (value should be nullptr)
-// and associated with role R. This is used for Cache "reservation" entries
-// such as for WriteBufferManager.
-template <CacheEntryRole R>
-Cache::DeleterFn GetNoopDeleterForRole() {
-  static cache_entry_roles_detail::RegisteredNoopDeleter<R> reg;
-  return reg.Delete;
-}
-
 }  // namespace ROCKSDB_NAMESPACE

cache/cache_entry_stats.h

@@ -10,8 +10,8 @@
 #include <memory>
 #include <mutex>
 
-#include "cache/cache_helpers.h"
 #include "cache/cache_key.h"
+#include "cache/typed_cache.h"
 #include "port/lang.h"
 #include "rocksdb/cache.h"
 #include "rocksdb/status.h"
@@ -111,11 +111,14 @@ class CacheEntryStatsCollector {
   // Gets or creates a shared instance of CacheEntryStatsCollector in the
   // cache itself, and saves into `ptr`. This shared_ptr will hold the
   // entry in cache until all refs are destroyed.
-  static Status GetShared(Cache *cache, SystemClock *clock,
+  static Status GetShared(Cache *raw_cache, SystemClock *clock,
                           std::shared_ptr<CacheEntryStatsCollector> *ptr) {
-    const Slice &cache_key = GetCacheKey();
+    assert(raw_cache);
+    BasicTypedCacheInterface<CacheEntryStatsCollector, CacheEntryRole::kMisc>
+        cache{raw_cache};
 
-    Cache::Handle *h = cache->Lookup(cache_key);
+    const Slice &cache_key = GetCacheKey();
+    auto h = cache.Lookup(cache_key);
     if (h == nullptr) {
       // Not yet in cache, but Cache doesn't provide a built-in way to
       // avoid racing insert. So we double-check under a shared mutex,
@@ -123,15 +126,15 @@ class CacheEntryStatsCollector {
       STATIC_AVOID_DESTRUCTION(std::mutex, static_mutex);
       std::lock_guard<std::mutex> lock(static_mutex);
 
-      h = cache->Lookup(cache_key);
+      h = cache.Lookup(cache_key);
       if (h == nullptr) {
-        auto new_ptr = new CacheEntryStatsCollector(cache, clock);
+        auto new_ptr = new CacheEntryStatsCollector(cache.get(), clock);
         // TODO: non-zero charge causes some tests that count block cache
         // usage to go flaky. Fix the problem somehow so we can use an
         // accurate charge.
         size_t charge = 0;
-        Status s = cache->Insert(cache_key, new_ptr, charge, Deleter, &h,
-                                 Cache::Priority::HIGH);
+        Status s =
+            cache.Insert(cache_key, new_ptr, charge, &h, Cache::Priority::HIGH);
         if (!s.ok()) {
           assert(h == nullptr);
           delete new_ptr;
@@ -140,11 +143,11 @@ class CacheEntryStatsCollector {
       }
     }
     // If we reach here, shared entry is in cache with handle `h`.
-    assert(cache->GetDeleter(h) == Deleter);
+    assert(cache.get()->GetCacheItemHelper(h) == cache.GetBasicHelper());
 
     // Build an aliasing shared_ptr that keeps `ptr` in cache while there
     // are references.
-    *ptr = MakeSharedCacheHandleGuard<CacheEntryStatsCollector>(cache, h);
+    *ptr = cache.SharedGuard(h);
     return Status::OK();
   }
 
@@ -157,10 +160,6 @@ class CacheEntryStatsCollector {
         cache_(cache),
         clock_(clock) {}
 
-  static void Deleter(const Slice &, void *value) {
-    delete static_cast<CacheEntryStatsCollector *>(value);
-  }
-
   static const Slice &GetCacheKey() {
     // For each template instantiation
     static CacheKey ckey = CacheKey::CreateUniqueForProcessLifetime();

cache/cache_helpers.cc

@@ -0,0 +1,40 @@
+//  Copyright (c) Meta Platforms, Inc. and affiliates.
+//  This source code is licensed under both the GPLv2 (found in the
+//  COPYING file in the root directory) and Apache 2.0 License
+//  (found in the LICENSE.Apache file in the root directory).
+
+#include "cache/cache_helpers.h"
+
+namespace ROCKSDB_NAMESPACE {
+
+void ReleaseCacheHandleCleanup(void* arg1, void* arg2) {
+  Cache* const cache = static_cast<Cache*>(arg1);
+  assert(cache);
+
+  Cache::Handle* const cache_handle = static_cast<Cache::Handle*>(arg2);
+  assert(cache_handle);
+
+  cache->Release(cache_handle);
+}
+
+Status WarmInCache(Cache* cache, const Slice& key, const Slice& saved,
+                   Cache::CreateContext* create_context,
+                   const Cache::CacheItemHelper* helper,
+                   Cache::Priority priority, size_t* out_charge) {
+  assert(helper);
+  assert(helper->create_cb);
+  Cache::ObjectPtr value;
+  size_t charge;
+  Status st = helper->create_cb(saved, create_context,
+                                cache->memory_allocator(), &value, &charge);
+  if (st.ok()) {
+    st =
+        cache->Insert(key, value, helper, charge, /*handle*/ nullptr, priority);
+    if (out_charge) {
+      *out_charge = charge;
+    }
+  }
+  return st;
+}
+
+}  // namespace ROCKSDB_NAMESPACE

cache/cache_helpers.h

@@ -7,7 +7,7 @@
 
 #include <cassert>
 
-#include "rocksdb/cache.h"
+#include "rocksdb/advanced_cache.h"
 #include "rocksdb/rocksdb_namespace.h"
 
 namespace ROCKSDB_NAMESPACE {
@@ -17,22 +17,17 @@ template <typename T>
 T* GetFromCacheHandle(Cache* cache, Cache::Handle* handle) {
   assert(cache);
   assert(handle);
-
   return static_cast<T*>(cache->Value(handle));
 }
 
-// Simple generic deleter for Cache (to be used with Cache::Insert).
-template <typename T>
-void DeleteCacheEntry(const Slice& /* key */, void* value) {
-  delete static_cast<T*>(value);
-}
-
 // Turns a T* into a Slice so it can be used as a key with Cache.
 template <typename T>
-Slice GetSlice(const T* t) {
+Slice GetSliceForKey(const T* t) {
   return Slice(reinterpret_cast<const char*>(t), sizeof(T));
 }
 
+void ReleaseCacheHandleCleanup(void* arg1, void* arg2);
+
 // Generic resource management object for cache handles that releases the handle
 // when destroyed. Has unique ownership of the handle, so copying it is not
 // allowed, while moving it transfers ownership.
@@ -88,7 +83,7 @@ class CacheHandleGuard {
     if (cleanable) {
       if (handle_ != nullptr) {
         assert(cache_);
-        cleanable->RegisterCleanup(&ReleaseCacheHandle, cache_, handle_);
+        cleanable->RegisterCleanup(&ReleaseCacheHandleCleanup, cache_, handle_);
       }
     }
     ResetFields();
@@ -115,16 +110,6 @@ class CacheHandleGuard {
     value_ = nullptr;
   }
 
-  static void ReleaseCacheHandle(void* arg1, void* arg2) {
-    Cache* const cache = static_cast<Cache*>(arg1);
-    assert(cache);
-
-    Cache::Handle* const cache_handle = static_cast<Cache::Handle*>(arg2);
-    assert(cache_handle);
-
-    cache->Release(cache_handle);
-  }
-
  private:
   Cache* cache_ = nullptr;
   Cache::Handle* handle_ = nullptr;
@@ -139,7 +124,16 @@ template <typename T>
 std::shared_ptr<T> MakeSharedCacheHandleGuard(Cache* cache,
                                               Cache::Handle* handle) {
   auto wrapper = std::make_shared<CacheHandleGuard<T>>(cache, handle);
-  return std::shared_ptr<T>(wrapper, static_cast<T*>(cache->Value(handle)));
+  return std::shared_ptr<T>(wrapper, GetFromCacheHandle<T>(cache, handle));
 }
 
+// Given the persistable data (saved) for a block cache entry, parse that
+// into a cache entry object and insert it into the given cache. The charge
+// of the new entry can be returned to the caller through `out_charge`.
+Status WarmInCache(Cache* cache, const Slice& key, const Slice& saved,
+                   Cache::CreateContext* create_context,
+                   const Cache::CacheItemHelper* helper,
+                   Cache::Priority priority = Cache::Priority::LOW,
+                   size_t* out_charge = nullptr);
+
 }  // namespace ROCKSDB_NAMESPACE

cache/cache_key.cc

@@ -8,7 +8,7 @@
 #include <algorithm>
 #include <atomic>
 
-#include "rocksdb/cache.h"
+#include "rocksdb/advanced_cache.h"
 #include "table/unique_id_impl.h"
 #include "util/hash.h"
 #include "util/math.h"

cache/cache_reservation_manager.cc

@@ -13,7 +13,6 @@
 #include <cstring>
 #include <memory>
 
-#include "cache/cache_entry_roles.h"
 #include "rocksdb/cache.h"
 #include "rocksdb/slice.h"
 #include "rocksdb/status.h"
@@ -41,17 +40,17 @@ CacheReservationManagerImpl<
 template <CacheEntryRole R>
 CacheReservationManagerImpl<R>::CacheReservationManagerImpl(
     std::shared_ptr<Cache> cache, bool delayed_decrease)
-    : delayed_decrease_(delayed_decrease),
+    : cache_(cache),
+      delayed_decrease_(delayed_decrease),
       cache_allocated_size_(0),
       memory_used_(0) {
   assert(cache != nullptr);
-  cache_ = cache;
 }
 
 template <CacheEntryRole R>
 CacheReservationManagerImpl<R>::~CacheReservationManagerImpl() {
   for (auto* handle : dummy_handles_) {
-    cache_->Release(handle, true);
+    cache_.ReleaseAndEraseIfLastRef(handle);
   }
 }
 
@@ -115,8 +114,7 @@ Status CacheReservationManagerImpl<R>::IncreaseCacheReservation(
   Status return_status = Status::OK();
   while (new_mem_used > cache_allocated_size_.load(std::memory_order_relaxed)) {
     Cache::Handle* handle = nullptr;
-    return_status = cache_->Insert(GetNextCacheKey(), nullptr, kSizeDummyEntry,
-                                   GetNoopDeleterForRole<R>(), &handle);
+    return_status = cache_.Insert(GetNextCacheKey(), kSizeDummyEntry, &handle);
 
     if (return_status != Status::OK()) {
       return return_status;
@@ -141,7 +139,7 @@ Status CacheReservationManagerImpl<R>::DecreaseCacheReservation(
          cache_allocated_size_.load(std::memory_order_relaxed)) {
     assert(!dummy_handles_.empty());
     auto* handle = dummy_handles_.back();
-    cache_->Release(handle, true);
+    cache_.ReleaseAndEraseIfLastRef(handle);
     dummy_handles_.pop_back();
     cache_allocated_size_ -= kSizeDummyEntry;
   }
@@ -169,8 +167,9 @@ Slice CacheReservationManagerImpl<R>::GetNextCacheKey() {
 }
 
 template <CacheEntryRole R>
-Cache::DeleterFn CacheReservationManagerImpl<R>::TEST_GetNoopDeleterForRole() {
-  return GetNoopDeleterForRole<R>();
+const Cache::CacheItemHelper*
+CacheReservationManagerImpl<R>::TEST_GetCacheItemHelperForRole() {
+  return CacheInterface::GetHelper();
 }
 
 template class CacheReservationManagerImpl<

cache/cache_reservation_manager.h

@@ -18,7 +18,7 @@
 
 #include "cache/cache_entry_roles.h"
 #include "cache/cache_key.h"
-#include "rocksdb/cache.h"
+#include "cache/typed_cache.h"
 #include "rocksdb/slice.h"
 #include "rocksdb/status.h"
 #include "util/coding.h"
@@ -197,10 +197,10 @@ class CacheReservationManagerImpl
 
   static constexpr std::size_t GetDummyEntrySize() { return kSizeDummyEntry; }
 
-  // For testing only - it is to help ensure the NoopDeleterForRole<R>
+  // For testing only - it is to help ensure the CacheItemHelperForRole<R>
   // accessed from CacheReservationManagerImpl and the one accessed from the
   // test are from the same translation units
-  static Cache::DeleterFn TEST_GetNoopDeleterForRole();
+  static const Cache::CacheItemHelper *TEST_GetCacheItemHelperForRole();
 
  private:
   static constexpr std::size_t kSizeDummyEntry = 256 * 1024;
@@ -211,7 +211,8 @@ class CacheReservationManagerImpl
   Status IncreaseCacheReservation(std::size_t new_mem_used);
   Status DecreaseCacheReservation(std::size_t new_mem_used);
 
-  std::shared_ptr<Cache> cache_;
+  using CacheInterface = PlaceholderSharedCacheInterface<R>;
+  CacheInterface cache_;
   bool delayed_decrease_;
   std::atomic<std::size_t> cache_allocated_size_;
   std::size_t memory_used_;

cache/cache_test.cc

@@ -15,15 +15,16 @@
 #include <string>
 #include <vector>
 
-#include "cache/fast_lru_cache.h"
 #include "cache/lru_cache.h"
+#include "cache/typed_cache.h"
 #include "port/stack_trace.h"
+#include "test_util/secondary_cache_test_util.h"
 #include "test_util/testharness.h"
 #include "util/coding.h"
 #include "util/string_util.h"
 
-// FastLRUCache and HyperClockCache only support 16-byte keys, so some of
-// the tests originally wrote for LRUCache do not work on the other caches.
+// HyperClockCache only supports 16-byte keys, so some of the tests
+// originally written for LRUCache do not work on the other caches.
 // Those tests were adapted to use 16-byte keys. We kept the original ones.
 // TODO: Remove the original tests if they ever become unused.
 
@@ -56,43 +57,43 @@ int DecodeKey32Bits(const Slice& k) {
   return DecodeFixed32(k.data());
 }
 
-void* EncodeValue(uintptr_t v) { return reinterpret_cast<void*>(v); }
+Cache::ObjectPtr EncodeValue(uintptr_t v) {
+  return reinterpret_cast<Cache::ObjectPtr>(v);
+}
 
 int DecodeValue(void* v) {
   return static_cast<int>(reinterpret_cast<uintptr_t>(v));
 }
 
-void DumbDeleter(const Slice& /*key*/, void* /*value*/) {}
+const Cache::CacheItemHelper kDumbHelper{
+    CacheEntryRole::kMisc,
+    [](Cache::ObjectPtr /*value*/, MemoryAllocator* /*alloc*/) {}};
 
-void EraseDeleter1(const Slice& /*key*/, void* value) {
-  Cache* cache = reinterpret_cast<Cache*>(value);
-  cache->Erase("foo");
-}
-
-void EraseDeleter2(const Slice& /*key*/, void* value) {
-  Cache* cache = reinterpret_cast<Cache*>(value);
-  cache->Erase(EncodeKey16Bytes(1234));
-}
-
-const std::string kLRU = "lru";
-const std::string kHyperClock = "hyper_clock";
-const std::string kFast = "fast";
+const Cache::CacheItemHelper kEraseOnDeleteHelper1{
+    CacheEntryRole::kMisc,
+    [](Cache::ObjectPtr value, MemoryAllocator* /*alloc*/) {
+      Cache* cache = static_cast<Cache*>(value);
+      cache->Erase("foo");
+    }};
 
+const Cache::CacheItemHelper kEraseOnDeleteHelper2{
+    CacheEntryRole::kMisc,
+    [](Cache::ObjectPtr value, MemoryAllocator* /*alloc*/) {
+      Cache* cache = static_cast<Cache*>(value);
+      cache->Erase(EncodeKey16Bytes(1234));
+    }};
 }  // anonymous namespace
 
-class CacheTest : public testing::TestWithParam<std::string> {
+class CacheTest : public testing::Test,
+                  public secondary_cache_test_util::WithCacheTypeParam {
  public:
   static CacheTest* current_;
   static std::string type_;
 
-  static void Deleter(const Slice& key, void* v) {
-    if (type_ == kFast || type_ == kHyperClock) {
-      current_->deleted_keys_.push_back(DecodeKey16Bytes(key));
-    } else {
-      current_->deleted_keys_.push_back(DecodeKey32Bits(key));
-    }
+  static void Deleter(Cache::ObjectPtr v, MemoryAllocator*) {
     current_->deleted_values_.push_back(DecodeValue(v));
   }
+  static const Cache::CacheItemHelper kHelper;
 
   static const int kCacheSize = 1000;
   static const int kNumShardBits = 4;
@@ -100,13 +101,10 @@ class CacheTest : public testing::TestWithParam<std::string> {
   static const int kCacheSize2 = 100;
   static const int kNumShardBits2 = 2;
 
-  std::vector<int> deleted_keys_;
   std::vector<int> deleted_values_;
   std::shared_ptr<Cache> cache_;
   std::shared_ptr<Cache> cache2_;
 
-  size_t estimated_value_size_ = 1;
-
   CacheTest()
       : cache_(NewCache(kCacheSize, kNumShardBits, false)),
         cache2_(NewCache(kCacheSize2, kNumShardBits2, false)) {
@@ -116,58 +114,13 @@ class CacheTest : public testing::TestWithParam<std::string> {
 
   ~CacheTest() override {}
 
-  std::shared_ptr<Cache> NewCache(size_t capacity) {
-    auto type = GetParam();
-    if (type == kLRU) {
-      return NewLRUCache(capacity);
-    }
-    if (type == kHyperClock) {
-      return HyperClockCacheOptions(
-                 capacity, estimated_value_size_ /*estimated_value_size*/)
-          .MakeSharedCache();
-    }
-    if (type == kFast) {
-      return NewFastLRUCache(
-          capacity, estimated_value_size_, -1 /*num_shard_bits*/,
-          false /*strict_capacity_limit*/, kDefaultCacheMetadataChargePolicy);
-    }
-    return nullptr;
-  }
-
-  std::shared_ptr<Cache> NewCache(
-      size_t capacity, int num_shard_bits, bool strict_capacity_limit,
-      CacheMetadataChargePolicy charge_policy = kDontChargeCacheMetadata) {
-    auto type = GetParam();
-    if (type == kLRU) {
-      LRUCacheOptions co;
-      co.capacity = capacity;
-      co.num_shard_bits = num_shard_bits;
-      co.strict_capacity_limit = strict_capacity_limit;
-      co.high_pri_pool_ratio = 0;
-      co.metadata_charge_policy = charge_policy;
-      return NewLRUCache(co);
-    }
-    if (type == kHyperClock) {
-      return HyperClockCacheOptions(capacity, 1 /*estimated_value_size*/,
-                                    num_shard_bits, strict_capacity_limit,
-                                    nullptr /*allocator*/, charge_policy)
-          .MakeSharedCache();
-    }
-    if (type == kFast) {
-      return NewFastLRUCache(capacity, 1 /*estimated_value_size*/,
-                             num_shard_bits, strict_capacity_limit,
-                             charge_policy);
-    }
-    return nullptr;
-  }
-
   // These functions encode/decode keys in tests cases that use
   // int keys.
   // Currently, HyperClockCache requires keys to be 16B long, whereas
   // LRUCache doesn't, so the encoding depends on the cache type.
   std::string EncodeKey(int k) {
     auto type = GetParam();
-    if (type == kFast || type == kHyperClock) {
+    if (type == kHyperClock) {
       return EncodeKey16Bytes(k);
     } else {
       return EncodeKey32Bits(k);
@@ -176,7 +129,7 @@ class CacheTest : public testing::TestWithParam<std::string> {
 
   int DecodeKey(const Slice& k) {
     auto type = GetParam();
-    if (type == kFast || type == kHyperClock) {
+    if (type == kHyperClock) {
       return DecodeKey16Bytes(k);
     } else {
       return DecodeKey32Bits(k);
@@ -194,39 +147,34 @@ class CacheTest : public testing::TestWithParam<std::string> {
 
   void Insert(std::shared_ptr<Cache> cache, int key, int value,
               int charge = 1) {
-    EXPECT_OK(cache->Insert(EncodeKey(key), EncodeValue(value), charge,
-                            &CacheTest::Deleter));
+    EXPECT_OK(cache->Insert(EncodeKey(key), EncodeValue(value), &kHelper,
+                            charge, /*handle*/ nullptr, Cache::Priority::HIGH));
   }
 
   void Erase(std::shared_ptr<Cache> cache, int key) {
     cache->Erase(EncodeKey(key));
   }
 
-  int Lookup(int key) {
-    return Lookup(cache_, key);
-  }
+  int Lookup(int key) { return Lookup(cache_, key); }
 
   void Insert(int key, int value, int charge = 1) {
     Insert(cache_, key, value, charge);
   }
 
-  void Erase(int key) {
-    Erase(cache_, key);
-  }
+  void Erase(int key) { Erase(cache_, key); }
 
-  int Lookup2(int key) {
-    return Lookup(cache2_, key);
-  }
+  int Lookup2(int key) { return Lookup(cache2_, key); }
 
   void Insert2(int key, int value, int charge = 1) {
     Insert(cache2_, key, value, charge);
   }
 
-  void Erase2(int key) {
-    Erase(cache2_, key);
-  }
+  void Erase2(int key) { Erase(cache2_, key); }
 };
 
+const Cache::CacheItemHelper CacheTest::kHelper{CacheEntryRole::kMisc,
+                                                &CacheTest::Deleter};
+
 CacheTest* CacheTest::current_;
 std::string CacheTest::type_;
 
@@ -256,10 +204,8 @@ TEST_P(CacheTest, UsageTest) {
       key = EncodeKey(i);
     }
     auto kv_size = key.size() + 5;
-    ASSERT_OK(cache->Insert(key, reinterpret_cast<void*>(value), kv_size,
-                            DumbDeleter));
-    ASSERT_OK(precise_cache->Insert(key, reinterpret_cast<void*>(value),
-                                    kv_size, DumbDeleter));
+    ASSERT_OK(cache->Insert(key, value, &kDumbHelper, kv_size));
+    ASSERT_OK(precise_cache->Insert(key, value, &kDumbHelper, kv_size));
     usage += kv_size;
     ASSERT_EQ(usage, cache->GetUsage());
     if (type == kHyperClock) {
@@ -282,10 +228,8 @@ TEST_P(CacheTest, UsageTest) {
     } else {
       key = EncodeKey(static_cast<int>(1000 + i));
     }
-    ASSERT_OK(cache->Insert(key, reinterpret_cast<void*>(value), key.size() + 5,
-                            DumbDeleter));
-    ASSERT_OK(precise_cache->Insert(key, reinterpret_cast<void*>(value),
-                                    key.size() + 5, DumbDeleter));
+    ASSERT_OK(cache->Insert(key, value, &kDumbHelper, key.size() + 5));
+    ASSERT_OK(precise_cache->Insert(key, value, &kDumbHelper, key.size() + 5));
   }
 
   // the usage should be close to the capacity
@@ -340,11 +284,9 @@ TEST_P(CacheTest, PinnedUsageTest) {
     auto kv_size = key.size() + 5;
     Cache::Handle* handle;
     Cache::Handle* handle_in_precise_cache;
-    ASSERT_OK(cache->Insert(key, reinterpret_cast<void*>(value), kv_size,
-                            DumbDeleter, &handle));
+    ASSERT_OK(cache->Insert(key, value, &kDumbHelper, kv_size, &handle));
     assert(handle);
-    ASSERT_OK(precise_cache->Insert(key, reinterpret_cast<void*>(value),
-                                    kv_size, DumbDeleter,
+    ASSERT_OK(precise_cache->Insert(key, value, &kDumbHelper, kv_size,
                                     &handle_in_precise_cache));
     assert(handle_in_precise_cache);
     pinned_usage += kv_size;
@@ -385,10 +327,8 @@ TEST_P(CacheTest, PinnedUsageTest) {
     } else {
       key = EncodeKey(static_cast<int>(1000 + i));
     }
-    ASSERT_OK(cache->Insert(key, reinterpret_cast<void*>(value), key.size() + 5,
-                            DumbDeleter));
-    ASSERT_OK(precise_cache->Insert(key, reinterpret_cast<void*>(value),
-                                    key.size() + 5, DumbDeleter));
+    ASSERT_OK(cache->Insert(key, value, &kDumbHelper, key.size() + 5));
+    ASSERT_OK(precise_cache->Insert(key, value, &kDumbHelper, key.size() + 5));
   }
   ASSERT_EQ(pinned_usage, cache->GetPinnedUsage());
   ASSERT_EQ(precise_cache_pinned_usage, precise_cache->GetPinnedUsage());
@@ -418,13 +358,13 @@ TEST_P(CacheTest, HitAndMiss) {
 
   Insert(100, 101);
   ASSERT_EQ(101, Lookup(100));
-  ASSERT_EQ(-1,  Lookup(200));
-  ASSERT_EQ(-1,  Lookup(300));
+  ASSERT_EQ(-1, Lookup(200));
+  ASSERT_EQ(-1, Lookup(300));
 
   Insert(200, 201);
   ASSERT_EQ(101, Lookup(100));
   ASSERT_EQ(201, Lookup(200));
-  ASSERT_EQ(-1,  Lookup(300));
+  ASSERT_EQ(-1, Lookup(300));
 
   Insert(100, 102);
   if (GetParam() == kHyperClock) {
@@ -434,10 +374,9 @@ TEST_P(CacheTest, HitAndMiss) {
     ASSERT_EQ(102, Lookup(100));
   }
   ASSERT_EQ(201, Lookup(200));
-  ASSERT_EQ(-1,  Lookup(300));
+  ASSERT_EQ(-1, Lookup(300));
 
-  ASSERT_EQ(1U, deleted_keys_.size());
-  ASSERT_EQ(100, deleted_keys_[0]);
+  ASSERT_EQ(1U, deleted_values_.size());
   if (GetParam() == kHyperClock) {
     ASSERT_EQ(102, deleted_values_[0]);
   } else {
@@ -458,21 +397,20 @@ TEST_P(CacheTest, InsertSameKey) {
 
 TEST_P(CacheTest, Erase) {
   Erase(200);
-  ASSERT_EQ(0U, deleted_keys_.size());
+  ASSERT_EQ(0U, deleted_values_.size());
 
   Insert(100, 101);
   Insert(200, 201);
   Erase(100);
-  ASSERT_EQ(-1,  Lookup(100));
+  ASSERT_EQ(-1, Lookup(100));
   ASSERT_EQ(201, Lookup(200));
-  ASSERT_EQ(1U, deleted_keys_.size());
-  ASSERT_EQ(100, deleted_keys_[0]);
+  ASSERT_EQ(1U, deleted_values_.size());
   ASSERT_EQ(101, deleted_values_[0]);
 
   Erase(100);
-  ASSERT_EQ(-1,  Lookup(100));
+  ASSERT_EQ(-1, Lookup(100));
   ASSERT_EQ(201, Lookup(200));
-  ASSERT_EQ(1U, deleted_keys_.size());
+  ASSERT_EQ(1U, deleted_values_.size());
 }
 
 TEST_P(CacheTest, EntriesArePinned) {
@@ -489,23 +427,21 @@ TEST_P(CacheTest, EntriesArePinned) {
   Insert(100, 102);
   Cache::Handle* h2 = cache_->Lookup(EncodeKey(100));
   ASSERT_EQ(102, DecodeValue(cache_->Value(h2)));
-  ASSERT_EQ(0U, deleted_keys_.size());
+  ASSERT_EQ(0U, deleted_values_.size());
   ASSERT_EQ(2U, cache_->GetUsage());
 
   cache_->Release(h1);
-  ASSERT_EQ(1U, deleted_keys_.size());
-  ASSERT_EQ(100, deleted_keys_[0]);
+  ASSERT_EQ(1U, deleted_values_.size());
   ASSERT_EQ(101, deleted_values_[0]);
   ASSERT_EQ(1U, cache_->GetUsage());
 
   Erase(100);
   ASSERT_EQ(-1, Lookup(100));
-  ASSERT_EQ(1U, deleted_keys_.size());
+  ASSERT_EQ(1U, deleted_values_.size());
   ASSERT_EQ(1U, cache_->GetUsage());
 
   cache_->Release(h2);
-  ASSERT_EQ(2U, deleted_keys_.size());
-  ASSERT_EQ(100, deleted_keys_[1]);
+  ASSERT_EQ(2U, deleted_values_.size());
   ASSERT_EQ(102, deleted_values_[1]);
   ASSERT_EQ(0U, cache_->GetUsage());
 }
@@ -515,7 +451,7 @@ TEST_P(CacheTest, EvictionPolicy) {
   Insert(200, 201);
   // Frequently used entry must be kept around
   for (int i = 0; i < 2 * kCacheSize; i++) {
-    Insert(1000+i, 2000+i);
+    Insert(1000 + i, 2000 + i);
     ASSERT_EQ(101, Lookup(100));
   }
   ASSERT_EQ(101, Lookup(100));
@@ -608,9 +544,9 @@ TEST_P(CacheTest, EvictEmptyCache) {
   // Insert item large than capacity to trigger eviction on empty cache.
   auto cache = NewCache(1, 0, false);
   if (type == kLRU) {
-    ASSERT_OK(cache->Insert("foo", nullptr, 10, DumbDeleter));
+    ASSERT_OK(cache->Insert("foo", nullptr, &kDumbHelper, 10));
   } else {
-    ASSERT_OK(cache->Insert(EncodeKey(1000), nullptr, 10, DumbDeleter));
+    ASSERT_OK(cache->Insert(EncodeKey(1000), nullptr, &kDumbHelper, 10));
   }
 }
 
@@ -621,19 +557,19 @@ TEST_P(CacheTest, EraseFromDeleter) {
   // the cache at that point.
   std::shared_ptr<Cache> cache = NewCache(10, 0, false);
   std::string foo, bar;
-  Cache::DeleterFn erase_deleter;
+  const Cache::CacheItemHelper* erase_helper;
   if (type == kLRU) {
     foo = "foo";
     bar = "bar";
-    erase_deleter = EraseDeleter1;
+    erase_helper = &kEraseOnDeleteHelper1;
   } else {
     foo = EncodeKey(1234);
     bar = EncodeKey(5678);
-    erase_deleter = EraseDeleter2;
+    erase_helper = &kEraseOnDeleteHelper2;
   }
 
-  ASSERT_OK(cache->Insert(foo, nullptr, 1, DumbDeleter));
-  ASSERT_OK(cache->Insert(bar, cache.get(), 1, erase_deleter));
+  ASSERT_OK(cache->Insert(foo, nullptr, &kDumbHelper, 1));
+  ASSERT_OK(cache->Insert(bar, cache.get(), erase_helper, 1));
 
   cache->Erase(bar);
   ASSERT_EQ(nullptr, cache->Lookup(foo));
@@ -686,7 +622,7 @@ TEST_P(CacheTest, HeavyEntries) {
       ASSERT_EQ(1000 + i, r);
     }
   }
-  ASSERT_LE(cached_weight, kCacheSize + kCacheSize/10);
+  ASSERT_LE(cached_weight, kCacheSize + kCacheSize / 10);
 }
 
 TEST_P(CacheTest, NewId) {
@@ -695,53 +631,54 @@ TEST_P(CacheTest, NewId) {
   ASSERT_NE(a, b);
 }
 
-class Value {
- public:
-  explicit Value(int v) : v_(v) {}
-
-  int v_;
-};
-
-namespace {
-void deleter(const Slice& /*key*/, void* value) {
-  delete static_cast<Value *>(value);
-}
-}  // namespace
-
 TEST_P(CacheTest, ReleaseAndErase) {
   std::shared_ptr<Cache> cache = NewCache(5, 0, false);
   Cache::Handle* handle;
-  Status s = cache->Insert(EncodeKey(100), EncodeValue(100), 1,
-                           &CacheTest::Deleter, &handle);
+  Status s =
+      cache->Insert(EncodeKey(100), EncodeValue(100), &kHelper, 1, &handle);
   ASSERT_TRUE(s.ok());
   ASSERT_EQ(5U, cache->GetCapacity());
   ASSERT_EQ(1U, cache->GetUsage());
-  ASSERT_EQ(0U, deleted_keys_.size());
+  ASSERT_EQ(0U, deleted_values_.size());
   auto erased = cache->Release(handle, true);
   ASSERT_TRUE(erased);
   // This tests that deleter has been called
-  ASSERT_EQ(1U, deleted_keys_.size());
+  ASSERT_EQ(1U, deleted_values_.size());
 }
 
 TEST_P(CacheTest, ReleaseWithoutErase) {
   std::shared_ptr<Cache> cache = NewCache(5, 0, false);
   Cache::Handle* handle;
-  Status s = cache->Insert(EncodeKey(100), EncodeValue(100), 1,
-                           &CacheTest::Deleter, &handle);
+  Status s =
+      cache->Insert(EncodeKey(100), EncodeValue(100), &kHelper, 1, &handle);
   ASSERT_TRUE(s.ok());
   ASSERT_EQ(5U, cache->GetCapacity());
   ASSERT_EQ(1U, cache->GetUsage());
-  ASSERT_EQ(0U, deleted_keys_.size());
+  ASSERT_EQ(0U, deleted_values_.size());
   auto erased = cache->Release(handle);
   ASSERT_FALSE(erased);
   // This tests that deleter is not called. When cache has free capacity it is
   // not expected to immediately erase the released items.
-  ASSERT_EQ(0U, deleted_keys_.size());
+  ASSERT_EQ(0U, deleted_values_.size());
 }
 
+namespace {
+class Value {
+ public:
+  explicit Value(int v) : v_(v) {}
+
+  int v_;
+
+  static constexpr auto kCacheEntryRole = CacheEntryRole::kMisc;
+};
+
+using SharedCache = BasicTypedSharedCacheInterface<Value>;
+using TypedHandle = SharedCache::TypedHandle;
+}  // namespace
+
 TEST_P(CacheTest, SetCapacity) {
   auto type = GetParam();
-  if (type == kFast || type == kHyperClock) {
+  if (type == kHyperClock) {
     ROCKSDB_GTEST_BYPASS(
         "FastLRUCache and HyperClockCache don't support arbitrary capacity "
         "adjustments.");
@@ -751,19 +688,19 @@ TEST_P(CacheTest, SetCapacity) {
   // lets create a cache with capacity 5,
   // then, insert 5 elements, then increase capacity
   // to 10, returned capacity should be 10, usage=5
-  std::shared_ptr<Cache> cache = NewCache(5, 0, false);
-  std::vector<Cache::Handle*> handles(10);
+  SharedCache cache{NewCache(5, 0, false)};
+  std::vector<TypedHandle*> handles(10);
   // Insert 5 entries, but not releasing.
   for (int i = 0; i < 5; i++) {
     std::string key = EncodeKey(i + 1);
-    Status s = cache->Insert(key, new Value(i + 1), 1, &deleter, &handles[i]);
+    Status s = cache.Insert(key, new Value(i + 1), 1, &handles[i]);
     ASSERT_TRUE(s.ok());
   }
-  ASSERT_EQ(5U, cache->GetCapacity());
-  ASSERT_EQ(5U, cache->GetUsage());
-  cache->SetCapacity(10);
-  ASSERT_EQ(10U, cache->GetCapacity());
-  ASSERT_EQ(5U, cache->GetUsage());
+  ASSERT_EQ(5U, cache.get()->GetCapacity());
+  ASSERT_EQ(5U, cache.get()->GetUsage());
+  cache.get()->SetCapacity(10);
+  ASSERT_EQ(10U, cache.get()->GetCapacity());
+  ASSERT_EQ(5U, cache.get()->GetUsage());
 
   // test2: decrease capacity
   // insert 5 more elements to cache, then release 5,
@@ -771,85 +708,77 @@ TEST_P(CacheTest, SetCapacity) {
   // and usage should be 7
   for (int i = 5; i < 10; i++) {
     std::string key = EncodeKey(i + 1);
-    Status s = cache->Insert(key, new Value(i + 1), 1, &deleter, &handles[i]);
+    Status s = cache.Insert(key, new Value(i + 1), 1, &handles[i]);
     ASSERT_TRUE(s.ok());
   }
-  ASSERT_EQ(10U, cache->GetCapacity());
-  ASSERT_EQ(10U, cache->GetUsage());
+  ASSERT_EQ(10U, cache.get()->GetCapacity());
+  ASSERT_EQ(10U, cache.get()->GetUsage());
   for (int i = 0; i < 5; i++) {
-    cache->Release(handles[i]);
+    cache.Release(handles[i]);
   }
-  ASSERT_EQ(10U, cache->GetCapacity());
-  ASSERT_EQ(10U, cache->GetUsage());
-  cache->SetCapacity(7);
-  ASSERT_EQ(7, cache->GetCapacity());
-  ASSERT_EQ(7, cache->GetUsage());
+  ASSERT_EQ(10U, cache.get()->GetCapacity());
+  ASSERT_EQ(10U, cache.get()->GetUsage());
+  cache.get()->SetCapacity(7);
+  ASSERT_EQ(7, cache.get()->GetCapacity());
+  ASSERT_EQ(7, cache.get()->GetUsage());
 
   // release remaining 5 to keep valgrind happy
   for (int i = 5; i < 10; i++) {
-    cache->Release(handles[i]);
+    cache.Release(handles[i]);
   }
 
   // Make sure this doesn't crash or upset ASAN/valgrind
-  cache->DisownData();
+  cache.get()->DisownData();
 }
 
 TEST_P(LRUCacheTest, SetStrictCapacityLimit) {
-  auto type = GetParam();
-  if (type == kFast) {
-    ROCKSDB_GTEST_BYPASS(
-        "FastLRUCache only supports a limited number of "
-        "inserts beyond "
-        "capacity.");
-    return;
-  }
   // test1: set the flag to false. Insert more keys than capacity. See if they
   // all go through.
-  std::shared_ptr<Cache> cache = NewCache(5, 0, false);
-  std::vector<Cache::Handle*> handles(10);
+  SharedCache cache{NewCache(5, 0, false)};
+  std::vector<TypedHandle*> handles(10);
   Status s;
   for (int i = 0; i < 10; i++) {
     std::string key = EncodeKey(i + 1);
-    s = cache->Insert(key, new Value(i + 1), 1, &deleter, &handles[i]);
+    s = cache.Insert(key, new Value(i + 1), 1, &handles[i]);
     ASSERT_OK(s);
     ASSERT_NE(nullptr, handles[i]);
   }
-  ASSERT_EQ(10, cache->GetUsage());
+  ASSERT_EQ(10, cache.get()->GetUsage());
 
   // test2: set the flag to true. Insert and check if it fails.
   std::string extra_key = EncodeKey(100);
   Value* extra_value = new Value(0);
-  cache->SetStrictCapacityLimit(true);
-  Cache::Handle* handle;
-  s = cache->Insert(extra_key, extra_value, 1, &deleter, &handle);
+  cache.get()->SetStrictCapacityLimit(true);
+  TypedHandle* handle;
+  s = cache.Insert(extra_key, extra_value, 1, &handle);
   ASSERT_TRUE(s.IsMemoryLimit());
   ASSERT_EQ(nullptr, handle);
-  ASSERT_EQ(10, cache->GetUsage());
+  ASSERT_EQ(10, cache.get()->GetUsage());
 
   for (int i = 0; i < 10; i++) {
-    cache->Release(handles[i]);
+    cache.Release(handles[i]);
   }
 
   // test3: init with flag being true.
-  std::shared_ptr<Cache> cache2 = NewCache(5, 0, true);
+  SharedCache cache2{NewCache(5, 0, true)};
   for (int i = 0; i < 5; i++) {
     std::string key = EncodeKey(i + 1);
-    s = cache2->Insert(key, new Value(i + 1), 1, &deleter, &handles[i]);
+    s = cache2.Insert(key, new Value(i + 1), 1, &handles[i]);
     ASSERT_OK(s);
     ASSERT_NE(nullptr, handles[i]);
   }
-  s = cache2->Insert(extra_key, extra_value, 1, &deleter, &handle);
+  s = cache2.Insert(extra_key, extra_value, 1, &handle);
   ASSERT_TRUE(s.IsMemoryLimit());
   ASSERT_EQ(nullptr, handle);
   // test insert without handle
-  s = cache2->Insert(extra_key, extra_value, 1, &deleter);
+  s = cache2.Insert(extra_key, extra_value, 1);
   // AS if the key have been inserted into cache but get evicted immediately.
   ASSERT_OK(s);
-  ASSERT_EQ(5, cache2->GetUsage());
-  ASSERT_EQ(nullptr, cache2->Lookup(extra_key));
+  ASSERT_EQ(5, cache2.get()->GetUsage());
+  ASSERT_EQ(nullptr, cache2.Lookup(extra_key));
 
   for (int i = 0; i < 5; i++) {
-    cache2->Release(handles[i]);
+    cache2.Release(handles[i]);
   }
 }
 
@@ -857,55 +786,54 @@ TEST_P(CacheTest, OverCapacity) {
   size_t n = 10;
 
   // a LRUCache with n entries and one shard only
-  std::shared_ptr<Cache> cache = NewCache(n, 0, false);
-
-  std::vector<Cache::Handle*> handles(n+1);
+  SharedCache cache{NewCache(n, 0, false)};
+  std::vector<TypedHandle*> handles(n + 1);
 
   // Insert n+1 entries, but not releasing.
   for (int i = 0; i < static_cast<int>(n + 1); i++) {
     std::string key = EncodeKey(i + 1);
-    Status s = cache->Insert(key, new Value(i + 1), 1, &deleter, &handles[i]);
+    Status s = cache.Insert(key, new Value(i + 1), 1, &handles[i]);
     ASSERT_TRUE(s.ok());
   }
 
   // Guess what's in the cache now?
   for (int i = 0; i < static_cast<int>(n + 1); i++) {
     std::string key = EncodeKey(i + 1);
-    auto h = cache->Lookup(key);
+    auto h = cache.Lookup(key);
     ASSERT_TRUE(h != nullptr);
-    if (h) cache->Release(h);
+    if (h) cache.Release(h);
   }
 
   // the cache is over capacity since nothing could be evicted
-  ASSERT_EQ(n + 1U, cache->GetUsage());
+  ASSERT_EQ(n + 1U, cache.get()->GetUsage());
   for (int i = 0; i < static_cast<int>(n + 1); i++) {
-    cache->Release(handles[i]);
+    cache.Release(handles[i]);
   }
 
   if (GetParam() == kHyperClock) {
     // Make sure eviction is triggered.
-    ASSERT_OK(cache->Insert(EncodeKey(-1), nullptr, 1, &deleter, &handles[0]));
+    ASSERT_OK(cache.Insert(EncodeKey(-1), nullptr, 1, &handles[0]));
 
     // cache is under capacity now since elements were released
-    ASSERT_GE(n, cache->GetUsage());
+    ASSERT_GE(n, cache.get()->GetUsage());
 
     // clean up
-    cache->Release(handles[0]);
+    cache.Release(handles[0]);
   } else {
     // LRUCache checks for over-capacity in Release.
 
     // cache is exactly at capacity now with minimal eviction
-    ASSERT_EQ(n, cache->GetUsage());
+    ASSERT_EQ(n, cache.get()->GetUsage());
 
     // element 0 is evicted and the rest is there
     // This is consistent with the LRU policy since the element 0
     // was released first
     for (int i = 0; i < static_cast<int>(n + 1); i++) {
       std::string key = EncodeKey(i + 1);
-      auto h = cache->Lookup(key);
+      auto h = cache.Lookup(key);
       if (h) {
         ASSERT_NE(static_cast<size_t>(i), 0U);
-        cache->Release(h);
+        cache.Release(h);
       } else {
         ASSERT_EQ(static_cast<size_t>(i), 0U);
       }
@@ -913,40 +841,15 @@ TEST_P(CacheTest, OverCapacity) {
   }
 }
 
-namespace {
-std::vector<std::pair<int, int>> legacy_callback_state;
-void legacy_callback(void* value, size_t charge) {
-  legacy_callback_state.push_back(
-      {DecodeValue(value), static_cast<int>(charge)});
-}
-};
-
-TEST_P(CacheTest, ApplyToAllCacheEntriesTest) {
-  std::vector<std::pair<int, int>> inserted;
-  legacy_callback_state.clear();
-
-  for (int i = 0; i < 10; ++i) {
-    Insert(i, i * 2, i + 1);
-    inserted.push_back({i * 2, i + 1});
-  }
-  cache_->ApplyToAllCacheEntries(legacy_callback, true);
-
-  std::sort(inserted.begin(), inserted.end());
-  std::sort(legacy_callback_state.begin(), legacy_callback_state.end());
-  ASSERT_EQ(inserted.size(), legacy_callback_state.size());
-  for (int i = 0; i < static_cast<int>(inserted.size()); ++i) {
-    EXPECT_EQ(inserted[i], legacy_callback_state[i]);
-  }
-}
-
 TEST_P(CacheTest, ApplyToAllEntriesTest) {
   std::vector<std::string> callback_state;
-  const auto callback = [&](const Slice& key, void* value, size_t charge,
-                            Cache::DeleterFn deleter) {
+  const auto callback = [&](const Slice& key, Cache::ObjectPtr value,
+                            size_t charge,
+                            const Cache::CacheItemHelper* helper) {
     callback_state.push_back(std::to_string(DecodeKey(key)) + "," +
                              std::to_string(DecodeValue(value)) + "," +
                              std::to_string(charge));
-    assert(deleter == &CacheTest::Deleter);
+    assert(helper == &CacheTest::kHelper);
   };
 
   std::vector<std::string> inserted;
@@ -985,8 +888,8 @@ TEST_P(CacheTest, ApplyToAllEntriesDuringResize) {
 
   // For callback
   int special_count = 0;
-  const auto callback = [&](const Slice&, void*, size_t charge,
-                            Cache::DeleterFn) {
+  const auto callback = [&](const Slice&, Cache::ObjectPtr, size_t charge,
+                            const Cache::CacheItemHelper*) {
     if (charge == static_cast<size_t>(kSpecialCharge)) {
       ++special_count;
     }
@@ -1048,14 +951,14 @@ TEST_P(CacheTest, GetChargeAndDeleter) {
   Cache::Handle* h1 = cache_->Lookup(EncodeKey(1));
   ASSERT_EQ(2, DecodeValue(cache_->Value(h1)));
   ASSERT_EQ(1, cache_->GetCharge(h1));
-  ASSERT_EQ(&CacheTest::Deleter, cache_->GetDeleter(h1));
+  ASSERT_EQ(&CacheTest::kHelper, cache_->GetCacheItemHelper(h1));
   cache_->Release(h1);
 }
 
 INSTANTIATE_TEST_CASE_P(CacheTestInstance, CacheTest,
-                        testing::Values(kLRU, kHyperClock, kFast));
+                        secondary_cache_test_util::GetTestingCacheTypes());
 INSTANTIATE_TEST_CASE_P(CacheTestInstance, LRUCacheTest,
-                        testing::Values(kLRU, kFast));
+                        testing::Values(secondary_cache_test_util::kLRU));
 
 }  // namespace ROCKSDB_NAMESPACE
 

cache/charged_cache.cc

@@ -11,65 +11,57 @@ namespace ROCKSDB_NAMESPACE {
 
 ChargedCache::ChargedCache(std::shared_ptr<Cache> cache,
                            std::shared_ptr<Cache> block_cache)
-    : cache_(cache),
+    : CacheWrapper(cache),
       cache_res_mgr_(std::make_shared<ConcurrentCacheReservationManager>(
           std::make_shared<
               CacheReservationManagerImpl<CacheEntryRole::kBlobCache>>(
               block_cache))) {}
 
-Status ChargedCache::Insert(const Slice& key, void* value, size_t charge,
-                            DeleterFn deleter, Handle** handle,
-                            Priority priority) {
-  Status s = cache_->Insert(key, value, charge, deleter, handle, priority);
-  if (s.ok()) {
-    // Insert may cause the cache entry eviction if the cache is full. So we
-    // directly call the reservation manager to update the total memory used
-    // in the cache.
-    assert(cache_res_mgr_);
-    cache_res_mgr_->UpdateCacheReservation(cache_->GetUsage())
-        .PermitUncheckedError();
-  }
-  return s;
-}
-
-Status ChargedCache::Insert(const Slice& key, void* value,
+Status ChargedCache::Insert(const Slice& key, ObjectPtr obj,
                             const CacheItemHelper* helper, size_t charge,
                             Handle** handle, Priority priority) {
-  Status s = cache_->Insert(key, value, helper, charge, handle, priority);
+  Status s = target_->Insert(key, obj, helper, charge, handle, priority);
   if (s.ok()) {
     // Insert may cause the cache entry eviction if the cache is full. So we
     // directly call the reservation manager to update the total memory used
     // in the cache.
     assert(cache_res_mgr_);
-    cache_res_mgr_->UpdateCacheReservation(cache_->GetUsage())
+    cache_res_mgr_->UpdateCacheReservation(target_->GetUsage())
         .PermitUncheckedError();
   }
   return s;
 }
 
-Cache::Handle* ChargedCache::Lookup(const Slice& key, Statistics* stats) {
-  return cache_->Lookup(key, stats);
-}
-
 Cache::Handle* ChargedCache::Lookup(const Slice& key,
                                     const CacheItemHelper* helper,
-                                    const CreateCallback& create_cb,
-                                    Priority priority, bool wait,
-                                    Statistics* stats) {
-  auto handle = cache_->Lookup(key, helper, create_cb, priority, wait, stats);
+                                    CreateContext* create_context,
+                                    Priority priority, Statistics* stats) {
+  auto handle = target_->Lookup(key, helper, create_context, priority, stats);
   // Lookup may promote the KV pair from the secondary cache to the primary
   // cache. So we directly call the reservation manager to update the total
   // memory used in the cache.
-  assert(cache_res_mgr_);
-  cache_res_mgr_->UpdateCacheReservation(cache_->GetUsage())
-      .PermitUncheckedError();
+  if (helper && helper->create_cb) {
+    assert(cache_res_mgr_);
+    cache_res_mgr_->UpdateCacheReservation(target_->GetUsage())
+        .PermitUncheckedError();
+  }
   return handle;
 }
 
+void ChargedCache::WaitAll(AsyncLookupHandle* async_handles, size_t count) {
+  target_->WaitAll(async_handles, count);
+  // In case of any promotions. Although some could finish by return of
+  // StartAsyncLookup, Wait/WaitAll will generally be used, so simpler to
+  // update here.
+  assert(cache_res_mgr_);
+  cache_res_mgr_->UpdateCacheReservation(target_->GetUsage())
+      .PermitUncheckedError();
+}
+
 bool ChargedCache::Release(Cache::Handle* handle, bool useful,
                            bool erase_if_last_ref) {
-  size_t memory_used_delta = cache_->GetUsage(handle);
-  bool erased = cache_->Release(handle, useful, erase_if_last_ref);
+  size_t memory_used_delta = target_->GetUsage(handle);
+  bool erased = target_->Release(handle, useful, erase_if_last_ref);
   if (erased) {
     assert(cache_res_mgr_);
     cache_res_mgr_
@@ -80,8 +72,8 @@ bool ChargedCache::Release(Cache::Handle* handle, bool useful,
 }
 
 bool ChargedCache::Release(Cache::Handle* handle, bool erase_if_last_ref) {
-  size_t memory_used_delta = cache_->GetUsage(handle);
-  bool erased = cache_->Release(handle, erase_if_last_ref);
+  size_t memory_used_delta = target_->GetUsage(handle);
+  bool erased = target_->Release(handle, erase_if_last_ref);
   if (erased) {
     assert(cache_res_mgr_);
     cache_res_mgr_
@@ -92,25 +84,25 @@ bool ChargedCache::Release(Cache::Handle* handle, bool erase_if_last_ref) {
 }
 
 void ChargedCache::Erase(const Slice& key) {
-  cache_->Erase(key);
+  target_->Erase(key);
   assert(cache_res_mgr_);
-  cache_res_mgr_->UpdateCacheReservation(cache_->GetUsage())
+  cache_res_mgr_->UpdateCacheReservation(target_->GetUsage())
       .PermitUncheckedError();
 }
 
 void ChargedCache::EraseUnRefEntries() {
-  cache_->EraseUnRefEntries();
+  target_->EraseUnRefEntries();
   assert(cache_res_mgr_);
-  cache_res_mgr_->UpdateCacheReservation(cache_->GetUsage())
+  cache_res_mgr_->UpdateCacheReservation(target_->GetUsage())
       .PermitUncheckedError();
 }
 
 void ChargedCache::SetCapacity(size_t capacity) {
-  cache_->SetCapacity(capacity);
+  target_->SetCapacity(capacity);
   // SetCapacity can result in evictions when the cache capacity is decreased,
   // so we would want to update the cache reservation here as well.
   assert(cache_res_mgr_);
-  cache_res_mgr_->UpdateCacheReservation(cache_->GetUsage())
+  cache_res_mgr_->UpdateCacheReservation(target_->GetUsage())
       .PermitUncheckedError();
 }
 

cache/charged_cache.h

@@ -8,7 +8,7 @@
 #include <string>
 
 #include "port/port.h"
-#include "rocksdb/cache.h"
+#include "rocksdb/advanced_cache.h"
 
 namespace ROCKSDB_NAMESPACE {
 
@@ -17,22 +17,21 @@ class ConcurrentCacheReservationManager;
 // A cache interface which wraps around another cache and takes care of
 // reserving space in block cache towards a single global memory limit, and
 // forwards all the calls to the underlying cache.
-class ChargedCache : public Cache {
+class ChargedCache : public CacheWrapper {
  public:
   ChargedCache(std::shared_ptr<Cache> cache,
                std::shared_ptr<Cache> block_cache);
-  ~ChargedCache() override = default;
 
-  Status Insert(const Slice& key, void* value, size_t charge, DeleterFn deleter,
-                Handle** handle, Priority priority) override;
-  Status Insert(const Slice& key, void* value, const CacheItemHelper* helper,
+  Status Insert(const Slice& key, ObjectPtr obj, const CacheItemHelper* helper,
                 size_t charge, Handle** handle = nullptr,
                 Priority priority = Priority::LOW) override;
 
-  Cache::Handle* Lookup(const Slice& key, Statistics* stats) override;
   Cache::Handle* Lookup(const Slice& key, const CacheItemHelper* helper,
-                        const CreateCallback& create_cb, Priority priority,
-                        bool wait, Statistics* stats = nullptr) override;
+                        CreateContext* create_context,
+                        Priority priority = Priority::LOW,
+                        Statistics* stats = nullptr) override;
+
+  void WaitAll(AsyncLookupHandle* async_handles, size_t count) override;
 
   bool Release(Cache::Handle* handle, bool useful,
                bool erase_if_last_ref = false) override;
@@ -44,69 +43,9 @@ class ChargedCache : public Cache {
   static const char* kClassName() { return "ChargedCache"; }
   const char* Name() const override { return kClassName(); }
 
-  uint64_t NewId() override { return cache_->NewId(); }
-
   void SetCapacity(size_t capacity) override;
 
-  void SetStrictCapacityLimit(bool strict_capacity_limit) override {
-    cache_->SetStrictCapacityLimit(strict_capacity_limit);
-  }
-
-  bool HasStrictCapacityLimit() const override {
-    return cache_->HasStrictCapacityLimit();
-  }
-
-  void* Value(Cache::Handle* handle) override { return cache_->Value(handle); }
-
-  bool IsReady(Cache::Handle* handle) override {
-    return cache_->IsReady(handle);
-  }
-
-  void Wait(Cache::Handle* handle) override { cache_->Wait(handle); }
-
-  void WaitAll(std::vector<Handle*>& handles) override {
-    cache_->WaitAll(handles);
-  }
-
-  bool Ref(Cache::Handle* handle) override { return cache_->Ref(handle); }
-
-  size_t GetCapacity() const override { return cache_->GetCapacity(); }
-
-  size_t GetUsage() const override { return cache_->GetUsage(); }
-
-  size_t GetUsage(Cache::Handle* handle) const override {
-    return cache_->GetUsage(handle);
-  }
-
-  size_t GetPinnedUsage() const override { return cache_->GetPinnedUsage(); }
-
-  size_t GetCharge(Cache::Handle* handle) const override {
-    return cache_->GetCharge(handle);
-  }
-
-  Cache::DeleterFn GetDeleter(Cache::Handle* handle) const override {
-    return cache_->GetDeleter(handle);
-  }
-
-  void ApplyToAllEntries(
-      const std::function<void(const Slice& key, void* value, size_t charge,
-                               Cache::DeleterFn deleter)>& callback,
-      const Cache::ApplyToAllEntriesOptions& opts) override {
-    cache_->ApplyToAllEntries(callback, opts);
-  }
-
-  void ApplyToAllCacheEntries(void (*callback)(void* value, size_t charge),
-                              bool thread_safe) override {
-    cache_->ApplyToAllCacheEntries(callback, thread_safe);
-  }
-
-  std::string GetPrintableOptions() const override {
-    return cache_->GetPrintableOptions();
-  }
-
-  void DisownData() override { return cache_->DisownData(); }
-
-  inline Cache* GetCache() const { return cache_.get(); }
+  inline Cache* GetCache() const { return target_.get(); }
 
   inline ConcurrentCacheReservationManager* TEST_GetCacheReservationManager()
       const {
@@ -114,7 +53,6 @@ class ChargedCache : public Cache {
   }
 
  private:
-  std::shared_ptr<Cache> cache_;
   std::shared_ptr<ConcurrentCacheReservationManager> cache_res_mgr_;
 };
 

cache/clock_cache.h

@@ -27,12 +27,13 @@
 
 namespace ROCKSDB_NAMESPACE {
 
-namespace hyper_clock_cache {
+namespace clock_cache {
 
 // Forward declaration of friend class.
 class ClockCacheTest;
 
-// HyperClockCache is an experimental alternative to LRUCache.
+// HyperClockCache is an alternative to LRUCache specifically tailored for
+// use as BlockBasedTableOptions::block_cache
 //
 // Benefits
 // --------
@@ -144,7 +145,7 @@ class ClockCacheTest;
 //     (erased by user) but can be read by existing references, and ref count
 //     changed by Ref and Release.
 //
-// A special case is "detached" entries, which are heap-allocated handles
+// A special case is "standalone" entries, which are heap-allocated handles
 // not in the table. They are always Invisible and freed on zero refs.
 //
 // State transitions:
@@ -199,8 +200,8 @@ class ClockCacheTest;
 // table occupancy limit has been reached. If strict_capacity_limit=false,
 // we must never fail Insert, and if a Handle* is provided, we have to return
 // a usable Cache handle on success. The solution to this (typically rare)
-// problem is "detached" handles, which are usable by the caller but not
-// actually available for Lookup in the Cache. Detached handles are allocated
+// problem is "standalone" handles, which are usable by the caller but not
+// actually available for Lookup in the Cache. Standalone handles are allocated
 // independently on the heap and specially marked so that they are freed on
 // the heap when their last reference is released.
 //
@@ -304,23 +305,23 @@ constexpr double kLoadFactor = 0.7;
 constexpr double kStrictLoadFactor = 0.84;
 
 struct ClockHandleBasicData {
-  void* value = nullptr;
-  Cache::DeleterFn deleter = nullptr;
+  Cache::ObjectPtr value = nullptr;
+  const Cache::CacheItemHelper* helper = nullptr;
   // A lossless, reversible hash of the fixed-size (16 byte) cache key. This
   // eliminates the need to store a hash separately.
   UniqueId64x2 hashed_key = kNullUniqueId64x2;
   size_t total_charge = 0;
 
+  inline size_t GetTotalCharge() const { return total_charge; }
+
   // Calls deleter (if non-null) on cache key and value
-  void FreeData() const;
+  void FreeData(MemoryAllocator* allocator) const;
 
   // Required by concept HandleImpl
   const UniqueId64x2& GetHash() const { return hashed_key; }
 };
 
-// Target size to be exactly a common cache line size (see static_assert in
-// clock_cache.cc)
-struct ALIGN_AS(64U) ClockHandle : public ClockHandleBasicData {
+struct ClockHandle : public ClockHandleBasicData {
   // Constants for handling the atomic `meta` word, which tracks most of the
   // state of the handle. The meta word looks like this:
   // low bits                                                     high bits
@@ -372,32 +373,60 @@ struct ALIGN_AS(64U) ClockHandle : public ClockHandleBasicData {
 
   // See above
   std::atomic<uint64_t> meta{};
-  // The number of elements that hash to this slot or a lower one, but wind
-  // up in this slot or a higher one.
-  std::atomic<uint32_t> displacements{};
 
-  // True iff the handle is allocated separately from hash table.
-  bool detached = false;
+  // Anticipating use for SecondaryCache support
+  void* reserved_for_future_use = nullptr;
 };  // struct ClockHandle
 
-class ClockHandleTable {
+class HyperClockTable {
  public:
-  explicit ClockHandleTable(int hash_bits, bool initial_charge_metadata);
-  ~ClockHandleTable();
+  // Target size to be exactly a common cache line size (see static_assert in
+  // clock_cache.cc)
+  struct ALIGN_AS(64U) HandleImpl : public ClockHandle {
+    // The number of elements that hash to this slot or a lower one, but wind
+    // up in this slot or a higher one.
+    std::atomic<uint32_t> displacements{};
 
-  Status Insert(const ClockHandleBasicData& proto, ClockHandle** handle,
+    // Whether this is a "deteched" handle that is independently allocated
+    // with `new` (so must be deleted with `delete`).
+    // TODO: ideally this would be packed into some other data field, such
+    // as upper bits of total_charge, but that incurs a measurable performance
+    // regression.
+    bool standalone = false;
+
+    inline bool IsStandalone() const { return standalone; }
+
+    inline void SetStandalone() { standalone = true; }
+  };  // struct HandleImpl
+
+  struct Opts {
+    size_t estimated_value_size;
+  };
+
+  HyperClockTable(size_t capacity, bool strict_capacity_limit,
+                  CacheMetadataChargePolicy metadata_charge_policy,
+                  MemoryAllocator* allocator,
+                  const Cache::EvictionCallback* eviction_callback,
+                  const Opts& opts);
+  ~HyperClockTable();
+
+  Status Insert(const ClockHandleBasicData& proto, HandleImpl** handle,
                 Cache::Priority priority, size_t capacity,
                 bool strict_capacity_limit);
 
-  ClockHandle* Lookup(const UniqueId64x2& hashed_key);
+  HandleImpl* CreateStandalone(ClockHandleBasicData& proto, size_t capacity,
+                               bool strict_capacity_limit,
+                               bool allow_uncharged);
 
-  bool Release(ClockHandle* handle, bool useful, bool erase_if_last_ref);
+  HandleImpl* Lookup(const UniqueId64x2& hashed_key);
 
-  void Ref(ClockHandle& handle);
+  bool Release(HandleImpl* handle, bool useful, bool erase_if_last_ref);
+
+  void Ref(HandleImpl& handle);
 
   void Erase(const UniqueId64x2& hashed_key);
 
-  void ConstApplyToEntriesRange(std::function<void(const ClockHandle&)> func,
+  void ConstApplyToEntriesRange(std::function<void(const HandleImpl&)> func,
                                 size_t index_begin, size_t index_end,
                                 bool apply_if_will_be_deleted) const;
 
@@ -407,21 +436,21 @@ class ClockHandleTable {
 
   int GetLengthBits() const { return length_bits_; }
 
-  size_t GetOccupancyLimit() const { return occupancy_limit_; }
-
   size_t GetOccupancy() const {
     return occupancy_.load(std::memory_order_relaxed);
   }
 
+  size_t GetOccupancyLimit() const { return occupancy_limit_; }
+
   size_t GetUsage() const { return usage_.load(std::memory_order_relaxed); }
 
-  size_t GetDetachedUsage() const {
-    return detached_usage_.load(std::memory_order_relaxed);
+  size_t GetStandaloneUsage() const {
+    return standalone_usage_.load(std::memory_order_relaxed);
   }
 
   // Acquire/release N references
-  void TEST_RefN(ClockHandle& handle, size_t n);
-  void TEST_ReleaseN(ClockHandle* handle, size_t n);
+  void TEST_RefN(HandleImpl& handle, size_t n);
+  void TEST_ReleaseN(HandleImpl* handle, size_t n);
 
  private:  // functions
   // Returns x mod 2^{length_bits_}.
@@ -432,8 +461,8 @@ class ClockHandleTable {
   // Runs the clock eviction algorithm trying to reclaim at least
   // requested_charge. Returns how much is evicted, which could be less
   // if it appears impossible to evict the requested amount without blocking.
-  void Evict(size_t requested_charge, size_t* freed_charge,
-             size_t* freed_count);
+  inline void Evict(size_t requested_charge, size_t* freed_charge,
+                    size_t* freed_count);
 
   // Returns the first slot in the probe sequence, starting from the given
   // probe number, with a handle e such that match(e) is true. At every
@@ -446,15 +475,56 @@ class ClockHandleTable {
   // value of probe is one more than the last non-aborting probe during the
   // call. This is so that that the variable can be used to keep track of
   // progress across consecutive calls to FindSlot.
-  inline ClockHandle* FindSlot(const UniqueId64x2& hashed_key,
-                               std::function<bool(ClockHandle*)> match,
-                               std::function<bool(ClockHandle*)> stop,
-                               std::function<void(ClockHandle*)> update,
-                               size_t& probe);
+  inline HandleImpl* FindSlot(const UniqueId64x2& hashed_key,
+                              std::function<bool(HandleImpl*)> match,
+                              std::function<bool(HandleImpl*)> stop,
+                              std::function<void(HandleImpl*)> update,
+                              size_t& probe);
 
   // Re-decrement all displacements in probe path starting from beginning
   // until (not including) the given handle
-  void Rollback(const UniqueId64x2& hashed_key, const ClockHandle* h);
+  inline void Rollback(const UniqueId64x2& hashed_key, const HandleImpl* h);
+
+  // Subtracts `total_charge` from `usage_` and 1 from `occupancy_`.
+  // Ideally this comes after releasing the entry itself so that we
+  // actually have the available occupancy/usage that is claimed.
+  // However, that means total_charge has to be saved from the handle
+  // before releasing it so that it can be provided to this function.
+  inline void ReclaimEntryUsage(size_t total_charge);
+
+  // Helper for updating `usage_` for new entry with given `total_charge`
+  // and evicting if needed under strict_capacity_limit=true rules. This
+  // means the operation might fail with Status::MemoryLimit. If
+  // `need_evict_for_occupancy`, then eviction of at least one entry is
+  // required, and the operation should fail if not possible.
+  // NOTE: Otherwise, occupancy_ is not managed in this function
+  inline Status ChargeUsageMaybeEvictStrict(size_t total_charge,
+                                            size_t capacity,
+                                            bool need_evict_for_occupancy);
+
+  // Helper for updating `usage_` for new entry with given `total_charge`
+  // and evicting if needed under strict_capacity_limit=false rules. This
+  // means that updating `usage_` always succeeds even if forced to exceed
+  // capacity. If `need_evict_for_occupancy`, then eviction of at least one
+  // entry is required, and the operation should return false if such eviction
+  // is not possible. `usage_` is not updated in that case. Otherwise, returns
+  // true, indicating success.
+  // NOTE: occupancy_ is not managed in this function
+  inline bool ChargeUsageMaybeEvictNonStrict(size_t total_charge,
+                                             size_t capacity,
+                                             bool need_evict_for_occupancy);
+
+  // Creates a "standalone" handle for returning from an Insert operation that
+  // cannot be completed by actually inserting into the table.
+  // Updates `standalone_usage_` but not `usage_` nor `occupancy_`.
+  inline HandleImpl* StandaloneInsert(const ClockHandleBasicData& proto);
+
+  MemoryAllocator* GetAllocator() const { return allocator_; }
+
+  // Returns the number of bits used to hash an element in the hash
+  // table.
+  static int CalcHashBits(size_t capacity, size_t estimated_value_size,
+                          CacheMetadataChargePolicy metadata_charge_policy);
 
  private:  // data
   // Number of hash bits used for table index.
@@ -468,7 +538,13 @@ class ClockHandleTable {
   const size_t occupancy_limit_;
 
   // Array of slots comprising the hash table.
-  const std::unique_ptr<ClockHandle[]> array_;
+  const std::unique_ptr<HandleImpl[]> array_;
+
+  // From Cache, for deleter
+  MemoryAllocator* const allocator_;
+
+  // A reference to Cache::eviction_callback_
+  const Cache::EvictionCallback& eviction_callback_;
 
   // We partition the following members into different cache lines
   // to avoid false sharing among Lookup, Release, Erase and Insert
@@ -482,22 +558,25 @@ class ClockHandleTable {
   // Number of elements in the table.
   std::atomic<size_t> occupancy_{};
 
-  // Memory usage by entries tracked by the cache (including detached)
+  // Memory usage by entries tracked by the cache (including standalone)
   std::atomic<size_t> usage_{};
 
-  // Part of usage by detached entries (not in table)
-  std::atomic<size_t> detached_usage_{};
-};  // class ClockHandleTable
+  // Part of usage by standalone entries (not in table)
+  std::atomic<size_t> standalone_usage_{};
+};  // class HyperClockTable
 
 // A single shard of sharded cache.
+template <class Table>
 class ALIGN_AS(CACHE_LINE_SIZE) ClockCacheShard final : public CacheShardBase {
  public:
-  ClockCacheShard(size_t capacity, size_t estimated_value_size,
-                  bool strict_capacity_limit,
-                  CacheMetadataChargePolicy metadata_charge_policy);
+  ClockCacheShard(size_t capacity, bool strict_capacity_limit,
+                  CacheMetadataChargePolicy metadata_charge_policy,
+                  MemoryAllocator* allocator,
+                  const Cache::EvictionCallback* eviction_callback,
+                  const typename Table::Opts& opts);
 
   // For CacheShard concept
-  using HandleImpl = ClockHandle;
+  using HandleImpl = typename Table::HandleImpl;
   // Hash is lossless hash of 128-bit key
   using HashVal = UniqueId64x2;
   using HashCref = const HashVal&;
@@ -531,74 +610,62 @@ class ALIGN_AS(CACHE_LINE_SIZE) ClockCacheShard final : public CacheShardBase {
 
   void SetStrictCapacityLimit(bool strict_capacity_limit);
 
-  Status Insert(const Slice& key, const UniqueId64x2& hashed_key, void* value,
-                size_t charge, Cache::DeleterFn deleter, ClockHandle** handle,
-                Cache::Priority priority);
+  Status Insert(const Slice& key, const UniqueId64x2& hashed_key,
+                Cache::ObjectPtr value, const Cache::CacheItemHelper* helper,
+                size_t charge, HandleImpl** handle, Cache::Priority priority);
 
-  ClockHandle* Lookup(const Slice& key, const UniqueId64x2& hashed_key);
+  HandleImpl* CreateStandalone(const Slice& key, const UniqueId64x2& hashed_key,
+                               Cache::ObjectPtr obj,
+                               const Cache::CacheItemHelper* helper,
+                               size_t charge, bool allow_uncharged);
 
-  bool Release(ClockHandle* handle, bool useful, bool erase_if_last_ref);
+  HandleImpl* Lookup(const Slice& key, const UniqueId64x2& hashed_key);
 
-  bool Release(ClockHandle* handle, bool erase_if_last_ref = false);
+  bool Release(HandleImpl* handle, bool useful, bool erase_if_last_ref);
 
-  bool Ref(ClockHandle* handle);
+  bool Release(HandleImpl* handle, bool erase_if_last_ref = false);
+
+  bool Ref(HandleImpl* handle);
 
   void Erase(const Slice& key, const UniqueId64x2& hashed_key);
 
+  size_t GetCapacity() const;
+
   size_t GetUsage() const;
 
+  size_t GetStandaloneUsage() const;
+
   size_t GetPinnedUsage() const;
 
   size_t GetOccupancyCount() const;
 
+  size_t GetOccupancyLimit() const;
+
   size_t GetTableAddressCount() const;
 
   void ApplyToSomeEntries(
-      const std::function<void(const Slice& key, void* value, size_t charge,
-                               DeleterFn deleter)>& callback,
+      const std::function<void(const Slice& key, Cache::ObjectPtr obj,
+                               size_t charge,
+                               const Cache::CacheItemHelper* helper)>& callback,
       size_t average_entries_per_lock, size_t* state);
 
   void EraseUnRefEntries();
 
   std::string GetPrintableOptions() const { return std::string{}; }
 
-  // SecondaryCache not yet supported
-  Status Insert(const Slice& key, const UniqueId64x2& hashed_key, void* value,
-                const Cache::CacheItemHelper* helper, size_t charge,
-                ClockHandle** handle, Cache::Priority priority) {
-    return Insert(key, hashed_key, value, charge, helper->del_cb, handle,
-                  priority);
-  }
-
-  ClockHandle* Lookup(const Slice& key, const UniqueId64x2& hashed_key,
-                      const Cache::CacheItemHelper* /*helper*/,
-                      const Cache::CreateCallback& /*create_cb*/,
-                      Cache::Priority /*priority*/, bool /*wait*/,
-                      Statistics* /*stats*/) {
+  HandleImpl* Lookup(const Slice& key, const UniqueId64x2& hashed_key,
+                     const Cache::CacheItemHelper* /*helper*/,
+                     Cache::CreateContext* /*create_context*/,
+                     Cache::Priority /*priority*/, Statistics* /*stats*/) {
     return Lookup(key, hashed_key);
   }
 
-  bool IsReady(ClockHandle* /*handle*/) { return true; }
-
-  void Wait(ClockHandle* /*handle*/) {}
-
   // Acquire/release N references
-  void TEST_RefN(ClockHandle* handle, size_t n);
-  void TEST_ReleaseN(ClockHandle* handle, size_t n);
-
- private:  // functions
-  friend class ClockCache;
-  friend class ClockCacheTest;
-
-  ClockHandle* DetachedInsert(const ClockHandleBasicData& h);
-
-  // Returns the number of bits used to hash an element in the hash
-  // table.
-  static int CalcHashBits(size_t capacity, size_t estimated_value_size,
-                          CacheMetadataChargePolicy metadata_charge_policy);
+  void TEST_RefN(HandleImpl* handle, size_t n);
+  void TEST_ReleaseN(HandleImpl* handle, size_t n);
 
  private:  // data
-  ClockHandleTable table_;
+  Table table_;
 
   // Maximum total charge of all elements stored in the table.
   std::atomic<size_t> capacity_;
@@ -611,8 +678,10 @@ class HyperClockCache
 #ifdef NDEBUG
     final
 #endif
-    : public ShardedCache<ClockCacheShard> {
+    : public ShardedCache<ClockCacheShard<HyperClockTable>> {
  public:
+  using Shard = ClockCacheShard<HyperClockTable>;
+
   HyperClockCache(size_t capacity, size_t estimated_value_size,
                   int num_shard_bits, bool strict_capacity_limit,
                   CacheMetadataChargePolicy metadata_charge_policy,
@@ -620,13 +689,16 @@ class HyperClockCache
 
   const char* Name() const override { return "HyperClockCache"; }
 
-  void* Value(Handle* handle) override;
+  Cache::ObjectPtr Value(Handle* handle) override;
 
   size_t GetCharge(Handle* handle) const override;
 
-  DeleterFn GetDeleter(Handle* handle) const override;
+  const CacheItemHelper* GetCacheItemHelper(Handle* handle) const override;
+
+  void ReportProblems(
+      const std::shared_ptr<Logger>& /*info_log*/) const override;
 };  // class HyperClockCache
 
-}  // namespace hyper_clock_cache
+}  // namespace clock_cache
 
 }  // namespace ROCKSDB_NAMESPACE

cache/compressed_secondary_cache.cc

@@ -22,12 +22,13 @@ CompressedSecondaryCache::CompressedSecondaryCache(
     std::shared_ptr<MemoryAllocator> memory_allocator, bool use_adaptive_mutex,
     CacheMetadataChargePolicy metadata_charge_policy,
     CompressionType compression_type, uint32_t compress_format_version,
-    bool enable_custom_split_merge)
+    bool enable_custom_split_merge,
+    const CacheEntryRoleSet& do_not_compress_roles)
     : cache_options_(capacity, num_shard_bits, strict_capacity_limit,
                      high_pri_pool_ratio, low_pri_pool_ratio, memory_allocator,
                      use_adaptive_mutex, metadata_charge_policy,
                      compression_type, compress_format_version,
-                     enable_custom_split_merge) {
+                     enable_custom_split_merge, do_not_compress_roles) {
   cache_ =
       NewLRUCache(capacity, num_shard_bits, strict_capacity_limit,
                   high_pri_pool_ratio, memory_allocator, use_adaptive_mutex,
@@ -37,10 +38,12 @@ CompressedSecondaryCache::CompressedSecondaryCache(
 CompressedSecondaryCache::~CompressedSecondaryCache() { cache_.reset(); }
 
 std::unique_ptr<SecondaryCacheResultHandle> CompressedSecondaryCache::Lookup(
-    const Slice& key, const Cache::CreateCallback& create_cb, bool /*wait*/,
-    bool advise_erase, bool& is_in_sec_cache) {
+    const Slice& key, const Cache::CacheItemHelper* helper,
+    Cache::CreateContext* create_context, bool /*wait*/, bool advise_erase,
+    bool& kept_in_sec_cache) {
+  assert(helper);
   std::unique_ptr<SecondaryCacheResultHandle> handle;
-  is_in_sec_cache = false;
+  kept_in_sec_cache = false;
   Cache::Handle* lru_handle = cache_->Lookup(key);
   if (lru_handle == nullptr) {
     return nullptr;
@@ -64,12 +67,15 @@ std::unique_ptr<SecondaryCacheResultHandle> CompressedSecondaryCache::Lookup(
     ptr = reinterpret_cast<CacheAllocationPtr*>(handle_value);
     handle_value_charge = cache_->GetCharge(lru_handle);
   }
+  MemoryAllocator* allocator = cache_options_.memory_allocator.get();
 
   Status s;
-  void* value{nullptr};
+  Cache::ObjectPtr value{nullptr};
   size_t charge{0};
-  if (cache_options_.compression_type == kNoCompression) {
-    s = create_cb(ptr->get(), handle_value_charge, &value, &charge);
+  if (cache_options_.compression_type == kNoCompression ||
+      cache_options_.do_not_compress_roles.Contains(helper->role)) {
+    s = helper->create_cb(Slice(ptr->get(), handle_value_charge),
+                          create_context, allocator, &value, &charge);
   } else {
     UncompressionContext uncompression_context(cache_options_.compression_type);
     UncompressionInfo uncompression_info(uncompression_context,
@@ -79,14 +85,14 @@ std::unique_ptr<SecondaryCacheResultHandle> CompressedSecondaryCache::Lookup(
     size_t uncompressed_size{0};
     CacheAllocationPtr uncompressed = UncompressData(
         uncompression_info, (char*)ptr->get(), handle_value_charge,
-        &uncompressed_size, cache_options_.compress_format_version,
-        cache_options_.memory_allocator.get());
+        &uncompressed_size, cache_options_.compress_format_version, allocator);
 
     if (!uncompressed) {
       cache_->Release(lru_handle, /*erase_if_last_ref=*/true);
       return nullptr;
     }
-    s = create_cb(uncompressed.get(), uncompressed_size, &value, &charge);
+    s = helper->create_cb(Slice(uncompressed.get(), uncompressed_size),
+                          create_context, allocator, &value, &charge);
   }
 
   if (!s.ok()) {
@@ -98,30 +104,32 @@ std::unique_ptr<SecondaryCacheResultHandle> CompressedSecondaryCache::Lookup(
     cache_->Release(lru_handle, /*erase_if_last_ref=*/true);
     // Insert a dummy handle.
     cache_
-        ->Insert(key, /*value=*/nullptr, /*charge=*/0,
-                 GetDeletionCallback(cache_options_.enable_custom_split_merge))
+        ->Insert(key, /*obj=*/nullptr,
+                 GetHelper(cache_options_.enable_custom_split_merge),
+                 /*charge=*/0)
         .PermitUncheckedError();
   } else {
-    is_in_sec_cache = true;
+    kept_in_sec_cache = true;
     cache_->Release(lru_handle, /*erase_if_last_ref=*/false);
   }
   handle.reset(new CompressedSecondaryCacheResultHandle(value, charge));
   return handle;
 }
 
-Status CompressedSecondaryCache::Insert(const Slice& key, void* value,
+Status CompressedSecondaryCache::Insert(const Slice& key,
+                                        Cache::ObjectPtr value,
                                         const Cache::CacheItemHelper* helper) {
   if (value == nullptr) {
     return Status::InvalidArgument();
   }
 
   Cache::Handle* lru_handle = cache_->Lookup(key);
-  Cache::DeleterFn del_cb =
-      GetDeletionCallback(cache_options_.enable_custom_split_merge);
+  auto internal_helper = GetHelper(cache_options_.enable_custom_split_merge);
   if (lru_handle == nullptr) {
     PERF_COUNTER_ADD(compressed_sec_cache_insert_dummy_count, 1);
     // Insert a dummy handle if the handle is evicted for the first time.
-    return cache_->Insert(key, /*value=*/nullptr, /*charge=*/0, del_cb);
+    return cache_->Insert(key, /*obj=*/nullptr, internal_helper,
+                          /*charge=*/0);
   } else {
     cache_->Release(lru_handle, /*erase_if_last_ref=*/false);
   }
@@ -137,7 +145,8 @@ Status CompressedSecondaryCache::Insert(const Slice& key, void* value,
   Slice val(ptr.get(), size);
 
   std::string compressed_val;
-  if (cache_options_.compression_type != kNoCompression) {
+  if (cache_options_.compression_type != kNoCompression &&
+      !cache_options_.do_not_compress_roles.Contains(helper->role)) {
     PERF_COUNTER_ADD(compressed_sec_cache_uncompressed_bytes, size);
     CompressionOptions compression_opts;
     CompressionContext compression_context(cache_options_.compression_type);
@@ -169,10 +178,10 @@ Status CompressedSecondaryCache::Insert(const Slice& key, void* value,
     size_t charge{0};
     CacheValueChunk* value_chunks_head =
         SplitValueIntoChunks(val, cache_options_.compression_type, charge);
-    return cache_->Insert(key, value_chunks_head, charge, del_cb);
+    return cache_->Insert(key, value_chunks_head, internal_helper, charge);
   } else {
     CacheAllocationPtr* buf = new CacheAllocationPtr(std::move(ptr));
-    return cache_->Insert(key, buf, size, del_cb);
+    return cache_->Insert(key, buf, internal_helper, size);
   }
 }
 
@@ -276,23 +285,29 @@ CacheAllocationPtr CompressedSecondaryCache::MergeChunksIntoValue(
   return ptr;
 }
 
-Cache::DeleterFn CompressedSecondaryCache::GetDeletionCallback(
-    bool enable_custom_split_merge) {
+const Cache::CacheItemHelper* CompressedSecondaryCache::GetHelper(
+    bool enable_custom_split_merge) const {
   if (enable_custom_split_merge) {
-    return [](const Slice& /*key*/, void* obj) {
-      CacheValueChunk* chunks_head = reinterpret_cast<CacheValueChunk*>(obj);
-      while (chunks_head != nullptr) {
-        CacheValueChunk* tmp_chunk = chunks_head;
-        chunks_head = chunks_head->next;
-        tmp_chunk->Free();
-        obj = nullptr;
-      };
-    };
+    static const Cache::CacheItemHelper kHelper{
+        CacheEntryRole::kMisc,
+        [](Cache::ObjectPtr obj, MemoryAllocator* /*alloc*/) {
+          CacheValueChunk* chunks_head = static_cast<CacheValueChunk*>(obj);
+          while (chunks_head != nullptr) {
+            CacheValueChunk* tmp_chunk = chunks_head;
+            chunks_head = chunks_head->next;
+            tmp_chunk->Free();
+            obj = nullptr;
+          };
+        }};
+    return &kHelper;
   } else {
-    return [](const Slice& /*key*/, void* obj) {
-      delete reinterpret_cast<CacheAllocationPtr*>(obj);
-      obj = nullptr;
-    };
+    static const Cache::CacheItemHelper kHelper{
+        CacheEntryRole::kMisc,
+        [](Cache::ObjectPtr obj, MemoryAllocator* /*alloc*/) {
+          delete static_cast<CacheAllocationPtr*>(obj);
+          obj = nullptr;
+        }};
+    return &kHelper;
   }
 }
 
@@ -302,12 +317,13 @@ std::shared_ptr<SecondaryCache> NewCompressedSecondaryCache(
     std::shared_ptr<MemoryAllocator> memory_allocator, bool use_adaptive_mutex,
     CacheMetadataChargePolicy metadata_charge_policy,
     CompressionType compression_type, uint32_t compress_format_version,
-    bool enable_custom_split_merge) {
+    bool enable_custom_split_merge,
+    const CacheEntryRoleSet& do_not_compress_roles) {
   return std::make_shared<CompressedSecondaryCache>(
       capacity, num_shard_bits, strict_capacity_limit, high_pri_pool_ratio,
       low_pri_pool_ratio, memory_allocator, use_adaptive_mutex,
       metadata_charge_policy, compression_type, compress_format_version,
-      enable_custom_split_merge);
+      enable_custom_split_merge, do_not_compress_roles);
 }
 
 std::shared_ptr<SecondaryCache> NewCompressedSecondaryCache(
@@ -319,7 +335,7 @@ std::shared_ptr<SecondaryCache> NewCompressedSecondaryCache(
       opts.high_pri_pool_ratio, opts.low_pri_pool_ratio, opts.memory_allocator,
       opts.use_adaptive_mutex, opts.metadata_charge_policy,
       opts.compression_type, opts.compress_format_version,
-      opts.enable_custom_split_merge);
+      opts.enable_custom_split_merge, opts.do_not_compress_roles);
 }
 
 }  // namespace ROCKSDB_NAMESPACE

cache/compressed_secondary_cache.h

@@ -21,7 +21,7 @@ namespace ROCKSDB_NAMESPACE {
 
 class CompressedSecondaryCacheResultHandle : public SecondaryCacheResultHandle {
  public:
-  CompressedSecondaryCacheResultHandle(void* value, size_t size)
+  CompressedSecondaryCacheResultHandle(Cache::ObjectPtr value, size_t size)
       : value_(value), size_(size) {}
   ~CompressedSecondaryCacheResultHandle() override = default;
 
@@ -34,12 +34,12 @@ class CompressedSecondaryCacheResultHandle : public SecondaryCacheResultHandle {
 
   void Wait() override {}
 
-  void* Value() override { return value_; }
+  Cache::ObjectPtr Value() override { return value_; }
 
   size_t Size() override { return size_; }
 
  private:
-  void* value_;
+  Cache::ObjectPtr value_;
   size_t size_;
 };
 
@@ -78,17 +78,20 @@ class CompressedSecondaryCache : public SecondaryCache {
           kDefaultCacheMetadataChargePolicy,
       CompressionType compression_type = CompressionType::kLZ4Compression,
       uint32_t compress_format_version = 2,
-      bool enable_custom_split_merge = false);
+      bool enable_custom_split_merge = false,
+      const CacheEntryRoleSet& do_not_compress_roles = {
+          CacheEntryRole::kFilterBlock});
   ~CompressedSecondaryCache() override;
 
   const char* Name() const override { return "CompressedSecondaryCache"; }
 
-  Status Insert(const Slice& key, void* value,
+  Status Insert(const Slice& key, Cache::ObjectPtr value,
                 const Cache::CacheItemHelper* helper) override;
 
   std::unique_ptr<SecondaryCacheResultHandle> Lookup(
-      const Slice& key, const Cache::CreateCallback& create_cb, bool /*wait*/,
-      bool advise_erase, bool& is_in_sec_cache) override;
+      const Slice& key, const Cache::CacheItemHelper* helper,
+      Cache::CreateContext* create_context, bool /*wait*/, bool advise_erase,
+      bool& kept_in_sec_cache) override;
 
   bool SupportForceErase() const override { return true; }
 
@@ -103,7 +106,7 @@ class CompressedSecondaryCache : public SecondaryCache {
   std::string GetPrintableOptions() const override;
 
  private:
-  friend class CompressedSecondaryCacheTest;
+  friend class CompressedSecondaryCacheTestBase;
   static constexpr std::array<uint16_t, 8> malloc_bin_sizes_{
       128, 256, 512, 1024, 2048, 4096, 8192, 16384};
 
@@ -129,8 +132,8 @@ class CompressedSecondaryCache : public SecondaryCache {
   CacheAllocationPtr MergeChunksIntoValue(const void* chunks_head,
                                           size_t& charge);
 
-  // An implementation of Cache::DeleterFn.
-  static Cache::DeleterFn GetDeletionCallback(bool enable_custom_split_merge);
+  // TODO: clean up to use cleaner interfaces in typed_cache.h
+  const Cache::CacheItemHelper* GetHelper(bool enable_custom_split_merge) const;
   std::shared_ptr<Cache> cache_;
   CompressedSecondaryCacheOptions cache_options_;
   mutable port::Mutex capacity_mutex_;

cache/compressed_secondary_cache_test.cc

@@ -5,86 +5,43 @@
 
 #include "cache/compressed_secondary_cache.h"
 
+#include <array>
 #include <iterator>
 #include <memory>
 #include <tuple>
 
 #include "memory/jemalloc_nodump_allocator.h"
 #include "rocksdb/convenience.h"
+#include "test_util/secondary_cache_test_util.h"
 #include "test_util/testharness.h"
 #include "test_util/testutil.h"
 
 namespace ROCKSDB_NAMESPACE {
 
-class CompressedSecondaryCacheTest : public testing::Test {
+using secondary_cache_test_util::GetTestingCacheTypes;
+using secondary_cache_test_util::WithCacheType;
+
+// 16 bytes for HCC compatibility
+const std::string key0 = "____    ____key0";
+const std::string key1 = "____    ____key1";
+const std::string key2 = "____    ____key2";
+const std::string key3 = "____    ____key3";
+
+class CompressedSecondaryCacheTestBase : public testing::Test,
+                                         public WithCacheType {
  public:
-  CompressedSecondaryCacheTest() : fail_create_(false) {}
-  ~CompressedSecondaryCacheTest() override = default;
+  CompressedSecondaryCacheTestBase() {}
+  ~CompressedSecondaryCacheTestBase() override = default;
 
  protected:
-  class TestItem {
-   public:
-    TestItem(const char* buf, size_t size) : buf_(new char[size]), size_(size) {
-      memcpy(buf_.get(), buf, size);
-    }
-    ~TestItem() = default;
-
-    char* Buf() { return buf_.get(); }
-    [[nodiscard]] size_t Size() const { return size_; }
-
-   private:
-    std::unique_ptr<char[]> buf_;
-    size_t size_;
-  };
-
-  static size_t SizeCallback(void* obj) {
-    return reinterpret_cast<TestItem*>(obj)->Size();
-  }
-
-  static Status SaveToCallback(void* from_obj, size_t from_offset,
-                               size_t length, void* out) {
-    auto item = reinterpret_cast<TestItem*>(from_obj);
-    const char* buf = item->Buf();
-    EXPECT_EQ(length, item->Size());
-    EXPECT_EQ(from_offset, 0);
-    memcpy(out, buf, length);
-    return Status::OK();
-  }
-
-  static void DeletionCallback(const Slice& /*key*/, void* obj) {
-    delete reinterpret_cast<TestItem*>(obj);
-    obj = nullptr;
-  }
-
-  static Cache::CacheItemHelper helper_;
-
-  static Status SaveToCallbackFail(void* /*obj*/, size_t /*offset*/,
-                                   size_t /*size*/, void* /*out*/) {
-    return Status::NotSupported();
-  }
-
-  static Cache::CacheItemHelper helper_fail_;
-
-  Cache::CreateCallback test_item_creator = [&](const void* buf, size_t size,
-                                                void** out_obj,
-                                                size_t* charge) -> Status {
-    if (fail_create_) {
-      return Status::NotSupported();
-    }
-    *out_obj = reinterpret_cast<void*>(new TestItem((char*)buf, size));
-    *charge = size;
-    return Status::OK();
-  };
-
-  void SetFailCreate(bool fail) { fail_create_ = fail; }
-
   void BasicTestHelper(std::shared_ptr<SecondaryCache> sec_cache,
                        bool sec_cache_is_compressed) {
     get_perf_context()->Reset();
-    bool is_in_sec_cache{true};
+    bool kept_in_sec_cache{true};
     // Lookup an non-existent key.
-    std::unique_ptr<SecondaryCacheResultHandle> handle0 = sec_cache->Lookup(
-        "k0", test_item_creator, true, /*advise_erase=*/true, is_in_sec_cache);
+    std::unique_ptr<SecondaryCacheResultHandle> handle0 =
+        sec_cache->Lookup(key0, GetHelper(), this, true, /*advise_erase=*/true,
+                          kept_in_sec_cache);
     ASSERT_EQ(handle0, nullptr);
 
     Random rnd(301);
@@ -92,25 +49,25 @@ class CompressedSecondaryCacheTest : public testing::Test {
     std::string str1(rnd.RandomString(1000));
     TestItem item1(str1.data(), str1.length());
     // A dummy handle is inserted if the item is inserted for the first time.
-    ASSERT_OK(sec_cache->Insert("k1", &item1,
-                                &CompressedSecondaryCacheTest::helper_));
+    ASSERT_OK(sec_cache->Insert(key1, &item1, GetHelper()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_dummy_count, 1);
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes, 0);
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_compressed_bytes, 0);
 
-    std::unique_ptr<SecondaryCacheResultHandle> handle1_1 = sec_cache->Lookup(
-        "k1", test_item_creator, true, /*advise_erase=*/false, is_in_sec_cache);
+    std::unique_ptr<SecondaryCacheResultHandle> handle1_1 =
+        sec_cache->Lookup(key1, GetHelper(), this, true, /*advise_erase=*/false,
+                          kept_in_sec_cache);
     ASSERT_EQ(handle1_1, nullptr);
 
     // Insert and Lookup the item k1 for the second time and advise erasing it.
-    ASSERT_OK(sec_cache->Insert("k1", &item1,
-                                &CompressedSecondaryCacheTest::helper_));
+    ASSERT_OK(sec_cache->Insert(key1, &item1, GetHelper()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_real_count, 1);
 
-    std::unique_ptr<SecondaryCacheResultHandle> handle1_2 = sec_cache->Lookup(
-        "k1", test_item_creator, true, /*advise_erase=*/true, is_in_sec_cache);
+    std::unique_ptr<SecondaryCacheResultHandle> handle1_2 =
+        sec_cache->Lookup(key1, GetHelper(), this, true, /*advise_erase=*/true,
+                          kept_in_sec_cache);
     ASSERT_NE(handle1_2, nullptr);
-    ASSERT_FALSE(is_in_sec_cache);
+    ASSERT_FALSE(kept_in_sec_cache);
     if (sec_cache_is_compressed) {
       ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes,
                 1000);
@@ -127,22 +84,22 @@ class CompressedSecondaryCacheTest : public testing::Test {
     ASSERT_EQ(memcmp(val1->Buf(), item1.Buf(), item1.Size()), 0);
 
     // Lookup the item k1 again.
-    std::unique_ptr<SecondaryCacheResultHandle> handle1_3 = sec_cache->Lookup(
-        "k1", test_item_creator, true, /*advise_erase=*/true, is_in_sec_cache);
+    std::unique_ptr<SecondaryCacheResultHandle> handle1_3 =
+        sec_cache->Lookup(key1, GetHelper(), this, true, /*advise_erase=*/true,
+                          kept_in_sec_cache);
     ASSERT_EQ(handle1_3, nullptr);
 
     // Insert and Lookup the item k2.
     std::string str2(rnd.RandomString(1000));
     TestItem item2(str2.data(), str2.length());
-    ASSERT_OK(sec_cache->Insert("k2", &item2,
-                                &CompressedSecondaryCacheTest::helper_));
+    ASSERT_OK(sec_cache->Insert(key2, &item2, GetHelper()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_dummy_count, 2);
-    std::unique_ptr<SecondaryCacheResultHandle> handle2_1 = sec_cache->Lookup(
-        "k2", test_item_creator, true, /*advise_erase=*/false, is_in_sec_cache);
+    std::unique_ptr<SecondaryCacheResultHandle> handle2_1 =
+        sec_cache->Lookup(key2, GetHelper(), this, true, /*advise_erase=*/false,
+                          kept_in_sec_cache);
     ASSERT_EQ(handle2_1, nullptr);
 
-    ASSERT_OK(sec_cache->Insert("k2", &item2,
-                                &CompressedSecondaryCacheTest::helper_));
+    ASSERT_OK(sec_cache->Insert(key2, &item2, GetHelper()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_real_count, 2);
     if (sec_cache_is_compressed) {
       ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes,
@@ -153,8 +110,9 @@ class CompressedSecondaryCacheTest : public testing::Test {
       ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes, 0);
       ASSERT_EQ(get_perf_context()->compressed_sec_cache_compressed_bytes, 0);
     }
-    std::unique_ptr<SecondaryCacheResultHandle> handle2_2 = sec_cache->Lookup(
-        "k2", test_item_creator, true, /*advise_erase=*/false, is_in_sec_cache);
+    std::unique_ptr<SecondaryCacheResultHandle> handle2_2 =
+        sec_cache->Lookup(key2, GetHelper(), this, true, /*advise_erase=*/false,
+                          kept_in_sec_cache);
     ASSERT_NE(handle2_2, nullptr);
     std::unique_ptr<TestItem> val2 =
         std::unique_ptr<TestItem>(static_cast<TestItem*>(handle2_2->Value()));
@@ -223,28 +181,26 @@ class CompressedSecondaryCacheTest : public testing::Test {
     std::string str1(rnd.RandomString(1000));
     TestItem item1(str1.data(), str1.length());
     // Insert a dummy handle.
-    ASSERT_OK(sec_cache->Insert("k1", &item1,
-                                &CompressedSecondaryCacheTest::helper_));
+    ASSERT_OK(sec_cache->Insert(key1, &item1, GetHelper()));
     // Insert k1.
-    ASSERT_OK(sec_cache->Insert("k1", &item1,
-                                &CompressedSecondaryCacheTest::helper_));
+    ASSERT_OK(sec_cache->Insert(key1, &item1, GetHelper()));
 
     // Insert and Lookup the second item.
     std::string str2(rnd.RandomString(200));
     TestItem item2(str2.data(), str2.length());
     // Insert a dummy handle, k1 is not evicted.
-    ASSERT_OK(sec_cache->Insert("k2", &item2,
-                                &CompressedSecondaryCacheTest::helper_));
-    bool is_in_sec_cache{false};
-    std::unique_ptr<SecondaryCacheResultHandle> handle1 = sec_cache->Lookup(
-        "k1", test_item_creator, true, /*advise_erase=*/false, is_in_sec_cache);
+    ASSERT_OK(sec_cache->Insert(key2, &item2, GetHelper()));
+    bool kept_in_sec_cache{false};
+    std::unique_ptr<SecondaryCacheResultHandle> handle1 =
+        sec_cache->Lookup(key1, GetHelper(), this, true, /*advise_erase=*/false,
+                          kept_in_sec_cache);
     ASSERT_EQ(handle1, nullptr);
 
     // Insert k2 and k1 is evicted.
-    ASSERT_OK(sec_cache->Insert("k2", &item2,
-                                &CompressedSecondaryCacheTest::helper_));
-    std::unique_ptr<SecondaryCacheResultHandle> handle2 = sec_cache->Lookup(
-        "k2", test_item_creator, true, /*advise_erase=*/false, is_in_sec_cache);
+    ASSERT_OK(sec_cache->Insert(key2, &item2, GetHelper()));
+    std::unique_ptr<SecondaryCacheResultHandle> handle2 =
+        sec_cache->Lookup(key2, GetHelper(), this, true, /*advise_erase=*/false,
+                          kept_in_sec_cache);
     ASSERT_NE(handle2, nullptr);
     std::unique_ptr<TestItem> val2 =
         std::unique_ptr<TestItem>(static_cast<TestItem*>(handle2->Value()));
@@ -252,27 +208,26 @@ class CompressedSecondaryCacheTest : public testing::Test {
     ASSERT_EQ(memcmp(val2->Buf(), item2.Buf(), item2.Size()), 0);
 
     // Insert k1 again and a dummy handle is inserted.
-    ASSERT_OK(sec_cache->Insert("k1", &item1,
-                                &CompressedSecondaryCacheTest::helper_));
+    ASSERT_OK(sec_cache->Insert(key1, &item1, GetHelper()));
 
-    std::unique_ptr<SecondaryCacheResultHandle> handle1_1 = sec_cache->Lookup(
-        "k1", test_item_creator, true, /*advise_erase=*/false, is_in_sec_cache);
+    std::unique_ptr<SecondaryCacheResultHandle> handle1_1 =
+        sec_cache->Lookup(key1, GetHelper(), this, true, /*advise_erase=*/false,
+                          kept_in_sec_cache);
     ASSERT_EQ(handle1_1, nullptr);
 
     // Create Fails.
     SetFailCreate(true);
-    std::unique_ptr<SecondaryCacheResultHandle> handle2_1 = sec_cache->Lookup(
-        "k2", test_item_creator, true, /*advise_erase=*/true, is_in_sec_cache);
+    std::unique_ptr<SecondaryCacheResultHandle> handle2_1 =
+        sec_cache->Lookup(key2, GetHelper(), this, true, /*advise_erase=*/true,
+                          kept_in_sec_cache);
     ASSERT_EQ(handle2_1, nullptr);
 
     // Save Fails.
     std::string str3 = rnd.RandomString(10);
     TestItem item3(str3.data(), str3.length());
     // The Status is OK because a dummy handle is inserted.
-    ASSERT_OK(sec_cache->Insert("k3", &item3,
-                                &CompressedSecondaryCacheTest::helper_fail_));
-    ASSERT_NOK(sec_cache->Insert("k3", &item3,
-                                 &CompressedSecondaryCacheTest::helper_fail_));
+    ASSERT_OK(sec_cache->Insert(key3, &item3, GetHelperFail()));
+    ASSERT_NOK(sec_cache->Insert(key3, &item3, GetHelperFail()));
 
     sec_cache.reset();
   }
@@ -296,28 +251,22 @@ class CompressedSecondaryCacheTest : public testing::Test {
     secondary_cache_opts.enable_custom_split_merge = enable_custom_split_merge;
     std::shared_ptr<SecondaryCache> secondary_cache =
         NewCompressedSecondaryCache(secondary_cache_opts);
-    LRUCacheOptions lru_cache_opts(
+    std::shared_ptr<Cache> cache = NewCache(
         /*_capacity =*/1300, /*_num_shard_bits =*/0,
-        /*_strict_capacity_limit =*/false, /*_high_pri_pool_ratio =*/0.5,
-        /*_memory_allocator =*/nullptr, kDefaultToAdaptiveMutex,
-        kDefaultCacheMetadataChargePolicy, /*_low_pri_pool_ratio =*/0.0);
-    lru_cache_opts.secondary_cache = secondary_cache;
-    std::shared_ptr<Cache> cache = NewLRUCache(lru_cache_opts);
+        /*_strict_capacity_limit =*/true, secondary_cache);
     std::shared_ptr<Statistics> stats = CreateDBStatistics();
 
     get_perf_context()->Reset();
     Random rnd(301);
     std::string str1 = rnd.RandomString(1001);
     auto item1_1 = new TestItem(str1.data(), str1.length());
-    ASSERT_OK(cache->Insert(
-        "k1", item1_1, &CompressedSecondaryCacheTest::helper_, str1.length()));
+    ASSERT_OK(cache->Insert(key1, item1_1, GetHelper(), str1.length()));
 
     std::string str2 = rnd.RandomString(1012);
     auto item2_1 = new TestItem(str2.data(), str2.length());
     // After this Insert, primary cache contains k2 and secondary cache contains
     // k1's dummy item.
-    ASSERT_OK(cache->Insert(
-        "k2", item2_1, &CompressedSecondaryCacheTest::helper_, str2.length()));
+    ASSERT_OK(cache->Insert(key2, item2_1, GetHelper(), str2.length()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_dummy_count, 1);
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes, 0);
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_compressed_bytes, 0);
@@ -326,22 +275,19 @@ class CompressedSecondaryCacheTest : public testing::Test {
     auto item3_1 = new TestItem(str3.data(), str3.length());
     // After this Insert, primary cache contains k3 and secondary cache contains
     // k1's dummy item and k2's dummy item.
-    ASSERT_OK(cache->Insert(
-        "k3", item3_1, &CompressedSecondaryCacheTest::helper_, str3.length()));
+    ASSERT_OK(cache->Insert(key3, item3_1, GetHelper(), str3.length()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_dummy_count, 2);
 
     // After this Insert, primary cache contains k1 and secondary cache contains
     // k1's dummy item, k2's dummy item, and k3's dummy item.
     auto item1_2 = new TestItem(str1.data(), str1.length());
-    ASSERT_OK(cache->Insert(
-        "k1", item1_2, &CompressedSecondaryCacheTest::helper_, str1.length()));
+    ASSERT_OK(cache->Insert(key1, item1_2, GetHelper(), str1.length()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_dummy_count, 3);
 
     // After this Insert, primary cache contains k2 and secondary cache contains
     // k1's item, k2's dummy item, and k3's dummy item.
     auto item2_2 = new TestItem(str2.data(), str2.length());
-    ASSERT_OK(cache->Insert(
-        "k2", item2_2, &CompressedSecondaryCacheTest::helper_, str2.length()));
+    ASSERT_OK(cache->Insert(key2, item2_2, GetHelper(), str2.length()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_real_count, 1);
     if (sec_cache_is_compressed) {
       ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes,
@@ -356,8 +302,7 @@ class CompressedSecondaryCacheTest : public testing::Test {
     // After this Insert, primary cache contains k3 and secondary cache contains
     // k1's item and k2's item.
     auto item3_2 = new TestItem(str3.data(), str3.length());
-    ASSERT_OK(cache->Insert(
-        "k3", item3_2, &CompressedSecondaryCacheTest::helper_, str3.length()));
+    ASSERT_OK(cache->Insert(key3, item3_2, GetHelper(), str3.length()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_real_count, 2);
     if (sec_cache_is_compressed) {
       ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes,
@@ -370,8 +315,7 @@ class CompressedSecondaryCacheTest : public testing::Test {
     }
 
     Cache::Handle* handle;
-    handle = cache->Lookup("k3", &CompressedSecondaryCacheTest::helper_,
-                           test_item_creator, Cache::Priority::LOW, true,
+    handle = cache->Lookup(key3, GetHelper(), this, Cache::Priority::LOW,
                            stats.get());
     ASSERT_NE(handle, nullptr);
     auto val3 = static_cast<TestItem*>(cache->Value(handle));
@@ -380,15 +324,13 @@ class CompressedSecondaryCacheTest : public testing::Test {
     cache->Release(handle);
 
     // Lookup an non-existent key.
-    handle = cache->Lookup("k0", &CompressedSecondaryCacheTest::helper_,
-                           test_item_creator, Cache::Priority::LOW, true,
+    handle = cache->Lookup(key0, GetHelper(), this, Cache::Priority::LOW,
                            stats.get());
     ASSERT_EQ(handle, nullptr);
 
     // This Lookup should just insert a dummy handle in the primary cache
     // and the k1 is still in the secondary cache.
-    handle = cache->Lookup("k1", &CompressedSecondaryCacheTest::helper_,
-                           test_item_creator, Cache::Priority::LOW, true,
+    handle = cache->Lookup(key1, GetHelper(), this, Cache::Priority::LOW,
                            stats.get());
     ASSERT_NE(handle, nullptr);
     ASSERT_EQ(get_perf_context()->block_cache_standalone_handle_count, 1);
@@ -400,8 +342,7 @@ class CompressedSecondaryCacheTest : public testing::Test {
     // This Lookup should erase k1 from the secondary cache and insert
     // it into primary cache; then k3 is demoted.
     // k2 and k3 are in secondary cache.
-    handle = cache->Lookup("k1", &CompressedSecondaryCacheTest::helper_,
-                           test_item_creator, Cache::Priority::LOW, true,
+    handle = cache->Lookup(key1, GetHelper(), this, Cache::Priority::LOW,
                            stats.get());
     ASSERT_NE(handle, nullptr);
     ASSERT_EQ(get_perf_context()->block_cache_standalone_handle_count, 1);
@@ -409,8 +350,7 @@ class CompressedSecondaryCacheTest : public testing::Test {
     cache->Release(handle);
 
     // k2 is still in secondary cache.
-    handle = cache->Lookup("k2", &CompressedSecondaryCacheTest::helper_,
-                           test_item_creator, Cache::Priority::LOW, true,
+    handle = cache->Lookup(key2, GetHelper(), this, Cache::Priority::LOW,
                            stats.get());
     ASSERT_NE(handle, nullptr);
     ASSERT_EQ(get_perf_context()->block_cache_standalone_handle_count, 2);
@@ -418,8 +358,7 @@ class CompressedSecondaryCacheTest : public testing::Test {
 
     // Testing SetCapacity().
     ASSERT_OK(secondary_cache->SetCapacity(0));
-    handle = cache->Lookup("k3", &CompressedSecondaryCacheTest::helper_,
-                           test_item_creator, Cache::Priority::LOW, true,
+    handle = cache->Lookup(key3, GetHelper(), this, Cache::Priority::LOW,
                            stats.get());
     ASSERT_EQ(handle, nullptr);
 
@@ -429,35 +368,30 @@ class CompressedSecondaryCacheTest : public testing::Test {
     ASSERT_EQ(capacity, 7000);
     auto item1_3 = new TestItem(str1.data(), str1.length());
     // After this Insert, primary cache contains k1.
-    ASSERT_OK(cache->Insert(
-        "k1", item1_3, &CompressedSecondaryCacheTest::helper_, str2.length()));
+    ASSERT_OK(cache->Insert(key1, item1_3, GetHelper(), str2.length()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_dummy_count, 3);
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_real_count, 4);
 
     auto item2_3 = new TestItem(str2.data(), str2.length());
     // After this Insert, primary cache contains k2 and secondary cache contains
     // k1's dummy item.
-    ASSERT_OK(cache->Insert(
-        "k2", item2_3, &CompressedSecondaryCacheTest::helper_, str1.length()));
+    ASSERT_OK(cache->Insert(key2, item2_3, GetHelper(), str1.length()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_dummy_count, 4);
 
     auto item1_4 = new TestItem(str1.data(), str1.length());
     // After this Insert, primary cache contains k1 and secondary cache contains
     // k1's dummy item and k2's dummy item.
-    ASSERT_OK(cache->Insert(
-        "k1", item1_4, &CompressedSecondaryCacheTest::helper_, str2.length()));
+    ASSERT_OK(cache->Insert(key1, item1_4, GetHelper(), str2.length()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_dummy_count, 5);
 
     auto item2_4 = new TestItem(str2.data(), str2.length());
     // After this Insert, primary cache contains k2 and secondary cache contains
     // k1's real item and k2's dummy item.
-    ASSERT_OK(cache->Insert(
-        "k2", item2_4, &CompressedSecondaryCacheTest::helper_, str2.length()));
+    ASSERT_OK(cache->Insert(key2, item2_4, GetHelper(), str2.length()));
     ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_real_count, 5);
     // This Lookup should just insert a dummy handle in the primary cache
     // and the k1 is still in the secondary cache.
-    handle = cache->Lookup("k1", &CompressedSecondaryCacheTest::helper_,
-                           test_item_creator, Cache::Priority::LOW, true,
+    handle = cache->Lookup(key1, GetHelper(), this, Cache::Priority::LOW,
                            stats.get());
 
     ASSERT_NE(handle, nullptr);
@@ -485,31 +419,31 @@ class CompressedSecondaryCacheTest : public testing::Test {
     std::shared_ptr<SecondaryCache> secondary_cache =
         NewCompressedSecondaryCache(secondary_cache_opts);
 
-    LRUCacheOptions opts(
+    std::shared_ptr<Cache> cache = NewCache(
         /*_capacity=*/1300, /*_num_shard_bits=*/0,
-        /*_strict_capacity_limit=*/false, /*_high_pri_pool_ratio=*/0.5,
-        /*_memory_allocator=*/nullptr, kDefaultToAdaptiveMutex,
-        kDefaultCacheMetadataChargePolicy, /*_low_pri_pool_ratio=*/0.0);
-    opts.secondary_cache = secondary_cache;
-    std::shared_ptr<Cache> cache = NewLRUCache(opts);
+        /*_strict_capacity_limit=*/false, secondary_cache);
 
     Random rnd(301);
     std::string str1 = rnd.RandomString(1001);
     auto item1 = std::make_unique<TestItem>(str1.data(), str1.length());
-    ASSERT_NOK(cache->Insert("k1", item1.get(), nullptr, str1.length()));
-    ASSERT_OK(cache->Insert("k1", item1.get(),
-                            &CompressedSecondaryCacheTest::helper_,
-                            str1.length()));
+    ASSERT_OK(cache->Insert(key1, item1.get(), GetHelper(), str1.length()));
     item1.release();  // Appease clang-analyze "potential memory leak"
 
     Cache::Handle* handle;
-    handle = cache->Lookup("k2", nullptr, test_item_creator,
-                           Cache::Priority::LOW, true);
+    handle = cache->Lookup(key2, nullptr, this, Cache::Priority::LOW);
     ASSERT_EQ(handle, nullptr);
-    handle = cache->Lookup("k2", &CompressedSecondaryCacheTest::helper_,
-                           test_item_creator, Cache::Priority::LOW, false);
+    handle = cache->Lookup(key2, GetHelper(), this, Cache::Priority::LOW);
     ASSERT_EQ(handle, nullptr);
 
+    Cache::AsyncLookupHandle ah;
+    ah.key = key2;
+    ah.helper = GetHelper();
+    ah.create_context = this;
+    ah.priority = Cache::Priority::LOW;
+    cache->StartAsyncLookup(ah);
+    cache->Wait(ah);
+    ASSERT_EQ(ah.Result(), nullptr);
+
     cache.reset();
     secondary_cache.reset();
   }
@@ -532,40 +466,29 @@ class CompressedSecondaryCacheTest : public testing::Test {
     std::shared_ptr<SecondaryCache> secondary_cache =
         NewCompressedSecondaryCache(secondary_cache_opts);
 
-    LRUCacheOptions opts(
+    std::shared_ptr<Cache> cache = NewCache(
         /*_capacity=*/1300, /*_num_shard_bits=*/0,
-        /*_strict_capacity_limit=*/false, /*_high_pri_pool_ratio=*/0.5,
-        /*_memory_allocator=*/nullptr, kDefaultToAdaptiveMutex,
-        kDefaultCacheMetadataChargePolicy, /*_low_pri_pool_ratio=*/0.0);
-    opts.secondary_cache = secondary_cache;
-    std::shared_ptr<Cache> cache = NewLRUCache(opts);
+        /*_strict_capacity_limit=*/true, secondary_cache);
 
     Random rnd(301);
     std::string str1 = rnd.RandomString(1001);
     auto item1 = new TestItem(str1.data(), str1.length());
-    ASSERT_OK(cache->Insert("k1", item1,
-                            &CompressedSecondaryCacheTest::helper_fail_,
-                            str1.length()));
+    ASSERT_OK(cache->Insert(key1, item1, GetHelperFail(), str1.length()));
 
     std::string str2 = rnd.RandomString(1002);
     auto item2 = new TestItem(str2.data(), str2.length());
     // k1 should be demoted to the secondary cache.
-    ASSERT_OK(cache->Insert("k2", item2,
-                            &CompressedSecondaryCacheTest::helper_fail_,
-                            str2.length()));
+    ASSERT_OK(cache->Insert(key2, item2, GetHelperFail(), str2.length()));
 
     Cache::Handle* handle;
-    handle = cache->Lookup("k2", &CompressedSecondaryCacheTest::helper_fail_,
-                           test_item_creator, Cache::Priority::LOW, true);
+    handle = cache->Lookup(key2, GetHelperFail(), this, Cache::Priority::LOW);
     ASSERT_NE(handle, nullptr);
     cache->Release(handle);
     // This lookup should fail, since k1 demotion would have failed.
-    handle = cache->Lookup("k1", &CompressedSecondaryCacheTest::helper_fail_,
-                           test_item_creator, Cache::Priority::LOW, true);
+    handle = cache->Lookup(key1, GetHelperFail(), this, Cache::Priority::LOW);
     ASSERT_EQ(handle, nullptr);
     // Since k1 was not promoted, k2 should still be in cache.
-    handle = cache->Lookup("k2", &CompressedSecondaryCacheTest::helper_fail_,
-                           test_item_creator, Cache::Priority::LOW, true);
+    handle = cache->Lookup(key2, GetHelperFail(), this, Cache::Priority::LOW);
     ASSERT_NE(handle, nullptr);
     cache->Release(handle);
 
@@ -591,39 +514,30 @@ class CompressedSecondaryCacheTest : public testing::Test {
     std::shared_ptr<SecondaryCache> secondary_cache =
         NewCompressedSecondaryCache(secondary_cache_opts);
 
-    LRUCacheOptions opts(
+    std::shared_ptr<Cache> cache = NewCache(
         /*_capacity=*/1300, /*_num_shard_bits=*/0,
-        /*_strict_capacity_limit=*/false, /*_high_pri_pool_ratio=*/0.5,
-        /*_memory_allocator=*/nullptr, kDefaultToAdaptiveMutex,
-        kDefaultCacheMetadataChargePolicy, /*_low_pri_pool_ratio=*/0.0);
-    opts.secondary_cache = secondary_cache;
-    std::shared_ptr<Cache> cache = NewLRUCache(opts);
+        /*_strict_capacity_limit=*/true, secondary_cache);
 
     Random rnd(301);
     std::string str1 = rnd.RandomString(1001);
     auto item1 = new TestItem(str1.data(), str1.length());
-    ASSERT_OK(cache->Insert("k1", item1, &CompressedSecondaryCacheTest::helper_,
-                            str1.length()));
+    ASSERT_OK(cache->Insert(key1, item1, GetHelper(), str1.length()));
 
     std::string str2 = rnd.RandomString(1002);
     auto item2 = new TestItem(str2.data(), str2.length());
     // k1 should be demoted to the secondary cache.
-    ASSERT_OK(cache->Insert("k2", item2, &CompressedSecondaryCacheTest::helper_,
-                            str2.length()));
+    ASSERT_OK(cache->Insert(key2, item2, GetHelper(), str2.length()));
 
     Cache::Handle* handle;
     SetFailCreate(true);
-    handle = cache->Lookup("k2", &CompressedSecondaryCacheTest::helper_,
-                           test_item_creator, Cache::Priority::LOW, true);
+    handle = cache->Lookup(key2, GetHelper(), this, Cache::Priority::LOW);
     ASSERT_NE(handle, nullptr);
     cache->Release(handle);
     // This lookup should fail, since k1 creation would have failed
-    handle = cache->Lookup("k1", &CompressedSecondaryCacheTest::helper_,
-                           test_item_creator, Cache::Priority::LOW, true);
+    handle = cache->Lookup(key1, GetHelper(), this, Cache::Priority::LOW);
     ASSERT_EQ(handle, nullptr);
     // Since k1 didn't get promoted, k2 should still be in cache
-    handle = cache->Lookup("k2", &CompressedSecondaryCacheTest::helper_,
-                           test_item_creator, Cache::Priority::LOW, true);
+    handle = cache->Lookup(key2, GetHelper(), this, Cache::Priority::LOW);
     ASSERT_NE(handle, nullptr);
     cache->Release(handle);
 
@@ -649,43 +563,34 @@ class CompressedSecondaryCacheTest : public testing::Test {
     std::shared_ptr<SecondaryCache> secondary_cache =
         NewCompressedSecondaryCache(secondary_cache_opts);
 
-    LRUCacheOptions opts(
+    std::shared_ptr<Cache> cache = NewCache(
         /*_capacity=*/1300, /*_num_shard_bits=*/0,
-        /*_strict_capacity_limit=*/false, /*_high_pri_pool_ratio=*/0.5,
-        /*_memory_allocator=*/nullptr, kDefaultToAdaptiveMutex,
-        kDefaultCacheMetadataChargePolicy, /*_low_pri_pool_ratio=*/0.0);
-    opts.secondary_cache = secondary_cache;
-    std::shared_ptr<Cache> cache = NewLRUCache(opts);
+        /*_strict_capacity_limit=*/false, secondary_cache);
 
     Random rnd(301);
     std::string str1 = rnd.RandomString(1001);
     auto item1_1 = new TestItem(str1.data(), str1.length());
-    ASSERT_OK(cache->Insert(
-        "k1", item1_1, &CompressedSecondaryCacheTest::helper_, str1.length()));
+    ASSERT_OK(cache->Insert(key1, item1_1, GetHelper(), str1.length()));
 
     std::string str2 = rnd.RandomString(1002);
     std::string str2_clone{str2};
     auto item2 = new TestItem(str2.data(), str2.length());
     // After this Insert, primary cache contains k2 and secondary cache contains
     // k1's dummy item.
-    ASSERT_OK(cache->Insert("k2", item2, &CompressedSecondaryCacheTest::helper_,
-                            str2.length()));
+    ASSERT_OK(cache->Insert(key2, item2, GetHelper(), str2.length()));
 
     // After this Insert, primary cache contains k1 and secondary cache contains
     // k1's dummy item and k2's dummy item.
     auto item1_2 = new TestItem(str1.data(), str1.length());
-    ASSERT_OK(cache->Insert(
-        "k1", item1_2, &CompressedSecondaryCacheTest::helper_, str1.length()));
+    ASSERT_OK(cache->Insert(key1, item1_2, GetHelper(), str1.length()));
 
     auto item2_2 = new TestItem(str2.data(), str2.length());
     // After this Insert, primary cache contains k2 and secondary cache contains
     // k1's item and k2's dummy item.
-    ASSERT_OK(cache->Insert(
-        "k2", item2_2, &CompressedSecondaryCacheTest::helper_, str2.length()));
+    ASSERT_OK(cache->Insert(key2, item2_2, GetHelper(), str2.length()));
 
     Cache::Handle* handle2;
-    handle2 = cache->Lookup("k2", &CompressedSecondaryCacheTest::helper_,
-                            test_item_creator, Cache::Priority::LOW, true);
+    handle2 = cache->Lookup(key2, GetHelper(), this, Cache::Priority::LOW);
     ASSERT_NE(handle2, nullptr);
     cache->Release(handle2);
 
@@ -693,14 +598,12 @@ class CompressedSecondaryCacheTest : public testing::Test {
     // strict_capacity_limit is true, but the lookup should still succeed.
     // A k1's dummy item is inserted into primary cache.
     Cache::Handle* handle1;
-    handle1 = cache->Lookup("k1", &CompressedSecondaryCacheTest::helper_,
-                            test_item_creator, Cache::Priority::LOW, true);
+    handle1 = cache->Lookup(key1, GetHelper(), this, Cache::Priority::LOW);
     ASSERT_NE(handle1, nullptr);
     cache->Release(handle1);
 
     // Since k1 didn't get inserted, k2 should still be in cache
-    handle2 = cache->Lookup("k2", &CompressedSecondaryCacheTest::helper_,
-                            test_item_creator, Cache::Priority::LOW, true);
+    handle2 = cache->Lookup(key2, GetHelper(), this, Cache::Priority::LOW);
     ASSERT_NE(handle2, nullptr);
     cache->Release(handle2);
 
@@ -741,7 +644,7 @@ class CompressedSecondaryCacheTest : public testing::Test {
     current_chunk = current_chunk->next;
     ASSERT_EQ(current_chunk->size, 98);
 
-    sec_cache->GetDeletionCallback(true)("dummy", chunks_head);
+    sec_cache->GetHelper(true)->del_cb(chunks_head, /*alloc*/ nullptr);
   }
 
   void MergeChunksIntoValueTest() {
@@ -822,31 +725,29 @@ class CompressedSecondaryCacheTest : public testing::Test {
     std::string value_str{value.get(), charge};
     ASSERT_EQ(strcmp(value_str.data(), str.data()), 0);
 
-    sec_cache->GetDeletionCallback(true)("dummy", chunks_head);
+    sec_cache->GetHelper(true)->del_cb(chunks_head, /*alloc*/ nullptr);
   }
-
- private:
-  bool fail_create_;
 };
 
-Cache::CacheItemHelper CompressedSecondaryCacheTest::helper_(
-    CompressedSecondaryCacheTest::SizeCallback,
-    CompressedSecondaryCacheTest::SaveToCallback,
-    CompressedSecondaryCacheTest::DeletionCallback);
+class CompressedSecondaryCacheTest
+    : public CompressedSecondaryCacheTestBase,
+      public testing::WithParamInterface<std::string> {
+  const std::string& Type() override { return GetParam(); }
+};
 
-Cache::CacheItemHelper CompressedSecondaryCacheTest::helper_fail_(
-    CompressedSecondaryCacheTest::SizeCallback,
-    CompressedSecondaryCacheTest::SaveToCallbackFail,
-    CompressedSecondaryCacheTest::DeletionCallback);
+INSTANTIATE_TEST_CASE_P(CompressedSecondaryCacheTest,
+                        CompressedSecondaryCacheTest, GetTestingCacheTypes());
 
 class CompressedSecCacheTestWithCompressAndAllocatorParam
-    : public CompressedSecondaryCacheTest,
-      public ::testing::WithParamInterface<std::tuple<bool, bool>> {
+    : public CompressedSecondaryCacheTestBase,
+      public ::testing::WithParamInterface<
+          std::tuple<bool, bool, std::string>> {
  public:
   CompressedSecCacheTestWithCompressAndAllocatorParam() {
     sec_cache_is_compressed_ = std::get<0>(GetParam());
     use_jemalloc_ = std::get<1>(GetParam());
   }
+  const std::string& Type() override { return std::get<2>(GetParam()); }
   bool sec_cache_is_compressed_;
   bool use_jemalloc_;
 };
@@ -857,20 +758,20 @@ TEST_P(CompressedSecCacheTestWithCompressAndAllocatorParam, BasicTes) {
 
 INSTANTIATE_TEST_CASE_P(CompressedSecCacheTests,
                         CompressedSecCacheTestWithCompressAndAllocatorParam,
-                        ::testing::Combine(testing::Bool(), testing::Bool()));
+                        ::testing::Combine(testing::Bool(), testing::Bool(),
+                                           GetTestingCacheTypes()));
 
 class CompressedSecondaryCacheTestWithCompressionParam
-    : public CompressedSecondaryCacheTest,
-      public ::testing::WithParamInterface<bool> {
+    : public CompressedSecondaryCacheTestBase,
+      public ::testing::WithParamInterface<std::tuple<bool, std::string>> {
  public:
   CompressedSecondaryCacheTestWithCompressionParam() {
-    sec_cache_is_compressed_ = GetParam();
+    sec_cache_is_compressed_ = std::get<0>(GetParam());
   }
+  const std::string& Type() override { return std::get<1>(GetParam()); }
   bool sec_cache_is_compressed_;
 };
 
-#ifndef ROCKSDB_LITE
-
 TEST_P(CompressedSecondaryCacheTestWithCompressionParam, BasicTestFromString) {
   std::shared_ptr<SecondaryCache> sec_cache{nullptr};
   std::string sec_cache_uri;
@@ -934,7 +835,6 @@ TEST_P(CompressedSecondaryCacheTestWithCompressionParam,
   BasicTestHelper(sec_cache, sec_cache_is_compressed_);
 }
 
-#endif  // ROCKSDB_LITE
 
 TEST_P(CompressedSecondaryCacheTestWithCompressionParam, FailsTest) {
   FailsTest(sec_cache_is_compressed_);
@@ -960,18 +860,92 @@ TEST_P(CompressedSecondaryCacheTestWithCompressionParam,
   IntegrationFullCapacityTest(sec_cache_is_compressed_);
 }
 
+TEST_P(CompressedSecondaryCacheTestWithCompressionParam, EntryRoles) {
+  CompressedSecondaryCacheOptions opts;
+  opts.capacity = 2048;
+  opts.num_shard_bits = 0;
+
+  if (sec_cache_is_compressed_) {
+    if (!LZ4_Supported()) {
+      ROCKSDB_GTEST_SKIP("This test requires LZ4 support.");
+      return;
+    }
+  } else {
+    opts.compression_type = CompressionType::kNoCompression;
+  }
+
+  // Select a random subset to include, for fast test
+  Random& r = *Random::GetTLSInstance();
+  CacheEntryRoleSet do_not_compress;
+  for (uint32_t i = 0; i < kNumCacheEntryRoles; ++i) {
+    // A few included on average, but decent chance of zero
+    if (r.OneIn(5)) {
+      do_not_compress.Add(static_cast<CacheEntryRole>(i));
+    }
+  }
+  opts.do_not_compress_roles = do_not_compress;
+
+  std::shared_ptr<SecondaryCache> sec_cache = NewCompressedSecondaryCache(opts);
+
+  // Fixed seed to ensure consistent compressibility (doesn't compress)
+  std::string junk(Random(301).RandomString(1000));
+
+  for (uint32_t i = 0; i < kNumCacheEntryRoles; ++i) {
+    CacheEntryRole role = static_cast<CacheEntryRole>(i);
+
+    // Uniquify `junk`
+    junk[0] = static_cast<char>(i);
+    TestItem item{junk.data(), junk.length()};
+    Slice ith_key = Slice(junk.data(), 16);
+
+    get_perf_context()->Reset();
+    ASSERT_OK(sec_cache->Insert(ith_key, &item, GetHelper(role)));
+    ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_dummy_count, 1U);
+
+    ASSERT_OK(sec_cache->Insert(ith_key, &item, GetHelper(role)));
+    ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_real_count, 1U);
+
+    bool kept_in_sec_cache{true};
+    std::unique_ptr<SecondaryCacheResultHandle> handle =
+        sec_cache->Lookup(ith_key, GetHelper(role), this, true,
+                          /*advise_erase=*/true, kept_in_sec_cache);
+    ASSERT_NE(handle, nullptr);
+
+    // Lookup returns the right data
+    std::unique_ptr<TestItem> val =
+        std::unique_ptr<TestItem>(static_cast<TestItem*>(handle->Value()));
+    ASSERT_NE(val, nullptr);
+    ASSERT_EQ(memcmp(val->Buf(), item.Buf(), item.Size()), 0);
+
+    bool compressed =
+        sec_cache_is_compressed_ && !do_not_compress.Contains(role);
+    if (compressed) {
+      ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes,
+                1000);
+      ASSERT_EQ(get_perf_context()->compressed_sec_cache_compressed_bytes,
+                1007);
+    } else {
+      ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes, 0);
+      ASSERT_EQ(get_perf_context()->compressed_sec_cache_compressed_bytes, 0);
+    }
+  }
+}
+
 INSTANTIATE_TEST_CASE_P(CompressedSecCacheTests,
                         CompressedSecondaryCacheTestWithCompressionParam,
-                        testing::Bool());
+                        testing::Combine(testing::Bool(),
+                                         GetTestingCacheTypes()));
 
 class CompressedSecCacheTestWithCompressAndSplitParam
-    : public CompressedSecondaryCacheTest,
-      public ::testing::WithParamInterface<std::tuple<bool, bool>> {
+    : public CompressedSecondaryCacheTestBase,
+      public ::testing::WithParamInterface<
+          std::tuple<bool, bool, std::string>> {
  public:
   CompressedSecCacheTestWithCompressAndSplitParam() {
     sec_cache_is_compressed_ = std::get<0>(GetParam());
     enable_custom_split_merge_ = std::get<1>(GetParam());
   }
+  const std::string& Type() override { return std::get<2>(GetParam()); }
   bool sec_cache_is_compressed_;
   bool enable_custom_split_merge_;
 };
@@ -982,17 +956,18 @@ TEST_P(CompressedSecCacheTestWithCompressAndSplitParam, BasicIntegrationTest) {
 
 INSTANTIATE_TEST_CASE_P(CompressedSecCacheTests,
                         CompressedSecCacheTestWithCompressAndSplitParam,
-                        ::testing::Combine(testing::Bool(), testing::Bool()));
+                        ::testing::Combine(testing::Bool(), testing::Bool(),
+                                           GetTestingCacheTypes()));
 
-TEST_F(CompressedSecondaryCacheTest, SplitValueIntoChunksTest) {
+TEST_P(CompressedSecondaryCacheTest, SplitValueIntoChunksTest) {
   SplitValueIntoChunksTest();
 }
 
-TEST_F(CompressedSecondaryCacheTest, MergeChunksIntoValueTest) {
+TEST_P(CompressedSecondaryCacheTest, MergeChunksIntoValueTest) {
   MergeChunksIntoValueTest();
 }
 
-TEST_F(CompressedSecondaryCacheTest, SplictValueAndMergeChunksTest) {
+TEST_P(CompressedSecondaryCacheTest, SplictValueAndMergeChunksTest) {
   SplictValueAndMergeChunksTest();
 }
 

cache/fast_lru_cache.cc

@@ -1,580 +0,0 @@
-//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
-//  This source code is licensed under both the GPLv2 (found in the
-//  COPYING file in the root directory) and Apache 2.0 License
-//  (found in the LICENSE.Apache file in the root directory).
-//
-// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file. See the AUTHORS file for names of contributors.
-
-#include "cache/fast_lru_cache.h"
-
-#include <cassert>
-#include <cstdint>
-#include <cstdio>
-#include <functional>
-
-#include "monitoring/perf_context_imp.h"
-#include "monitoring/statistics.h"
-#include "port/lang.h"
-#include "util/distributed_mutex.h"
-#include "util/hash.h"
-#include "util/math.h"
-#include "util/random.h"
-
-namespace ROCKSDB_NAMESPACE {
-
-namespace fast_lru_cache {
-
-LRUHandleTable::LRUHandleTable(int hash_bits)
-    : length_bits_(hash_bits),
-      length_bits_mask_((uint32_t{1} << length_bits_) - 1),
-      occupancy_(0),
-      occupancy_limit_(static_cast<uint32_t>((uint32_t{1} << length_bits_) *
-                                             kStrictLoadFactor)),
-      array_(new LRUHandle[size_t{1} << length_bits_]) {
-  assert(hash_bits <= 32);
-}
-
-LRUHandleTable::~LRUHandleTable() {
-  ApplyToEntriesRange([](LRUHandle* h) { h->FreeData(); }, 0, GetTableSize());
-}
-
-LRUHandle* LRUHandleTable::Lookup(const Slice& key, uint32_t hash) {
-  int probe = 0;
-  int slot = FindVisibleElement(key, hash, probe, 0);
-  return (slot == -1) ? nullptr : &array_[slot];
-}
-
-LRUHandle* LRUHandleTable::Insert(LRUHandle* h, LRUHandle** old) {
-  int probe = 0;
-  int slot = FindVisibleElementOrAvailableSlot(h->key(), h->hash, probe,
-                                               1 /*displacement*/);
-  *old = nullptr;
-  if (slot == -1) {
-    // TODO(Guido) Don't we need to roll back displacements here?
-    return nullptr;
-  }
-
-  if (array_[slot].IsEmpty() || array_[slot].IsTombstone()) {
-    bool empty = array_[slot].IsEmpty();
-    Assign(slot, h);
-    LRUHandle* new_entry = &array_[slot];
-    if (empty) {
-      // This used to be an empty slot.
-      return new_entry;
-    }
-    // It used to be a tombstone, so there may already be a copy of the
-    // key in the table.
-    slot = FindVisibleElement(h->key(), h->hash, probe, 0 /*displacement*/);
-    if (slot == -1) {
-      // No existing copy of the key.
-      return new_entry;
-    }
-    *old = &array_[slot];
-    return new_entry;
-  } else {
-    // There is an existing copy of the key.
-    *old = &array_[slot];
-    // Find an available slot for the new element.
-    array_[slot].displacements++;
-    slot = FindAvailableSlot(h->key(), probe, 1 /*displacement*/);
-    if (slot == -1) {
-      // No available slots. Roll back displacements.
-      probe = 0;
-      slot = FindVisibleElement(h->key(), h->hash, probe, -1);
-      array_[slot].displacements--;
-      FindAvailableSlot(h->key(), probe, -1);
-      return nullptr;
-    }
-    Assign(slot, h);
-    return &array_[slot];
-  }
-}
-
-void LRUHandleTable::Remove(LRUHandle* h) {
-  assert(h->next == nullptr &&
-         h->prev == nullptr);  // Already off the LRU list.
-  int probe = 0;
-  FindSlot(
-      h->key(), [&h](LRUHandle* e) { return e == h; }, probe,
-      -1 /*displacement*/);
-  h->SetIsVisible(false);
-  h->SetIsElement(false);
-  occupancy_--;
-}
-
-void LRUHandleTable::Assign(int slot, LRUHandle* h) {
-  LRUHandle* dst = &array_[slot];
-  uint32_t disp = dst->displacements;
-  *dst = *h;
-  dst->displacements = disp;
-  dst->SetIsVisible(true);
-  dst->SetIsElement(true);
-  occupancy_++;
-}
-
-void LRUHandleTable::Exclude(LRUHandle* h) { h->SetIsVisible(false); }
-
-int LRUHandleTable::FindVisibleElement(const Slice& key, uint32_t hash,
-                                       int& probe, int displacement) {
-  return FindSlot(
-      key,
-      [&](LRUHandle* h) { return h->Matches(key, hash) && h->IsVisible(); },
-      probe, displacement);
-}
-
-int LRUHandleTable::FindAvailableSlot(const Slice& key, int& probe,
-                                      int displacement) {
-  return FindSlot(
-      key, [](LRUHandle* h) { return h->IsEmpty() || h->IsTombstone(); }, probe,
-      displacement);
-}
-
-int LRUHandleTable::FindVisibleElementOrAvailableSlot(const Slice& key,
-                                                      uint32_t hash, int& probe,
-                                                      int displacement) {
-  return FindSlot(
-      key,
-      [&](LRUHandle* h) {
-        return h->IsEmpty() || h->IsTombstone() ||
-               (h->Matches(key, hash) && h->IsVisible());
-      },
-      probe, displacement);
-}
-
-inline int LRUHandleTable::FindSlot(const Slice& key,
-                                    std::function<bool(LRUHandle*)> cond,
-                                    int& probe, int displacement) {
-  uint32_t base = ModTableSize(Hash(key.data(), key.size(), kProbingSeed1));
-  uint32_t increment =
-      ModTableSize((Hash(key.data(), key.size(), kProbingSeed2) << 1) | 1);
-  uint32_t current = ModTableSize(base + probe * increment);
-  while (true) {
-    LRUHandle* h = &array_[current];
-    probe++;
-    if (current == base && probe > 1) {
-      // We looped back.
-      return -1;
-    }
-    if (cond(h)) {
-      return current;
-    }
-    if (h->IsEmpty()) {
-      // We check emptyness after the condition, because
-      // the condition may be emptyness.
-      return -1;
-    }
-    h->displacements += displacement;
-    current = ModTableSize(current + increment);
-  }
-}
-
-LRUCacheShard::LRUCacheShard(size_t capacity, size_t estimated_value_size,
-                             bool strict_capacity_limit,
-                             CacheMetadataChargePolicy metadata_charge_policy)
-    : CacheShardBase(metadata_charge_policy),
-      capacity_(capacity),
-      strict_capacity_limit_(strict_capacity_limit),
-      table_(
-          CalcHashBits(capacity, estimated_value_size, metadata_charge_policy)),
-      usage_(0),
-      lru_usage_(0) {
-  // Make empty circular linked list.
-  lru_.next = &lru_;
-  lru_.prev = &lru_;
-  lru_low_pri_ = &lru_;
-}
-
-void LRUCacheShard::EraseUnRefEntries() {
-  autovector<LRUHandle> last_reference_list;
-  {
-    DMutexLock l(mutex_);
-    while (lru_.next != &lru_) {
-      LRUHandle* old = lru_.next;
-      // LRU list contains only elements which can be evicted.
-      assert(old->IsVisible() && !old->HasRefs());
-      LRU_Remove(old);
-      table_.Remove(old);
-      assert(usage_ >= old->total_charge);
-      usage_ -= old->total_charge;
-      last_reference_list.push_back(*old);
-    }
-  }
-
-  // Free the entries here outside of mutex for performance reasons.
-  for (auto& h : last_reference_list) {
-    h.FreeData();
-  }
-}
-
-void LRUCacheShard::ApplyToSomeEntries(
-    const std::function<void(const Slice& key, void* value, size_t charge,
-                             DeleterFn deleter)>& callback,
-    size_t average_entries_per_lock, size_t* state) {
-  // The state is essentially going to be the starting hash, which works
-  // nicely even if we resize between calls because we use upper-most
-  // hash bits for table indexes.
-  DMutexLock l(mutex_);
-  size_t length_bits = table_.GetLengthBits();
-  size_t length = table_.GetTableSize();
-
-  assert(average_entries_per_lock > 0);
-  // Assuming we are called with same average_entries_per_lock repeatedly,
-  // this simplifies some logic (index_end will not overflow).
-  assert(average_entries_per_lock < length || *state == 0);
-
-  size_t index_begin = *state >> (sizeof(size_t) * 8u - length_bits);
-  size_t index_end = index_begin + average_entries_per_lock;
-  if (index_end >= length) {
-    // Going to end
-    index_end = length;
-    *state = SIZE_MAX;
-  } else {
-    *state = index_end << (sizeof(size_t) * 8u - length_bits);
-  }
-
-  table_.ApplyToEntriesRange(
-      [callback,
-       metadata_charge_policy = metadata_charge_policy_](LRUHandle* h) {
-        callback(h->key(), h->value, h->GetCharge(metadata_charge_policy),
-                 h->deleter);
-      },
-      index_begin, index_end);
-}
-
-void LRUCacheShard::LRU_Remove(LRUHandle* h) {
-  assert(h->next != nullptr);
-  assert(h->prev != nullptr);
-  h->next->prev = h->prev;
-  h->prev->next = h->next;
-  h->prev = h->next = nullptr;
-  assert(lru_usage_ >= h->total_charge);
-  lru_usage_ -= h->total_charge;
-}
-
-void LRUCacheShard::LRU_Insert(LRUHandle* h) {
-  assert(h->next == nullptr);
-  assert(h->prev == nullptr);
-  // Insert h to head of LRU list.
-  h->next = &lru_;
-  h->prev = lru_.prev;
-  h->prev->next = h;
-  h->next->prev = h;
-  lru_usage_ += h->total_charge;
-}
-
-void LRUCacheShard::EvictFromLRU(size_t charge,
-                                 autovector<LRUHandle>* deleted) {
-  while ((usage_ + charge) > capacity_ && lru_.next != &lru_) {
-    LRUHandle* old = lru_.next;
-    // LRU list contains only elements which can be evicted.
-    assert(old->IsVisible() && !old->HasRefs());
-    LRU_Remove(old);
-    table_.Remove(old);
-    assert(usage_ >= old->total_charge);
-    usage_ -= old->total_charge;
-    deleted->push_back(*old);
-  }
-}
-
-size_t LRUCacheShard::CalcEstimatedHandleCharge(
-    size_t estimated_value_size,
-    CacheMetadataChargePolicy metadata_charge_policy) {
-  LRUHandle h;
-  h.CalcTotalCharge(estimated_value_size, metadata_charge_policy);
-  return h.total_charge;
-}
-
-int LRUCacheShard::CalcHashBits(
-    size_t capacity, size_t estimated_value_size,
-    CacheMetadataChargePolicy metadata_charge_policy) {
-  size_t handle_charge =
-      CalcEstimatedHandleCharge(estimated_value_size, metadata_charge_policy);
-  assert(handle_charge > 0);
-  uint32_t num_entries =
-      static_cast<uint32_t>(capacity / (kLoadFactor * handle_charge)) + 1;
-  assert(num_entries <= uint32_t{1} << 31);
-  return FloorLog2((num_entries << 1) - 1);
-}
-
-void LRUCacheShard::SetCapacity(size_t capacity) {
-  autovector<LRUHandle> last_reference_list;
-  {
-    DMutexLock l(mutex_);
-    if (capacity > capacity_) {
-      assert(false);  // Not supported.
-    }
-    capacity_ = capacity;
-    EvictFromLRU(0, &last_reference_list);
-  }
-
-  // Free the entries here outside of mutex for performance reasons.
-  for (auto& h : last_reference_list) {
-    h.FreeData();
-  }
-}
-
-void LRUCacheShard::SetStrictCapacityLimit(bool strict_capacity_limit) {
-  DMutexLock l(mutex_);
-  strict_capacity_limit_ = strict_capacity_limit;
-}
-
-Status LRUCacheShard::Insert(const Slice& key, uint32_t hash, void* value,
-                             size_t charge, Cache::DeleterFn deleter,
-                             LRUHandle** handle, Cache::Priority /*priority*/) {
-  if (key.size() != kCacheKeySize) {
-    return Status::NotSupported("FastLRUCache only supports key size " +
-                                std::to_string(kCacheKeySize) + "B");
-  }
-
-  LRUHandle tmp;
-  tmp.value = value;
-  tmp.deleter = deleter;
-  tmp.hash = hash;
-  tmp.CalcTotalCharge(charge, metadata_charge_policy_);
-  for (int i = 0; i < kCacheKeySize; i++) {
-    tmp.key_data[i] = key.data()[i];
-  }
-
-  Status s = Status::OK();
-  autovector<LRUHandle> last_reference_list;
-  {
-    DMutexLock l(mutex_);
-    assert(table_.GetOccupancy() <= table_.GetOccupancyLimit());
-
-    // Free the space following strict LRU policy until enough space
-    // is freed or the lru list is empty.
-    EvictFromLRU(tmp.total_charge, &last_reference_list);
-    if ((usage_ + tmp.total_charge > capacity_ &&
-         (strict_capacity_limit_ || handle == nullptr)) ||
-        table_.GetOccupancy() == table_.GetOccupancyLimit()) {
-      // There are two measures of capacity:
-      // - Space (or charge) capacity: The maximum possible sum of the charges
-      //    of the elements.
-      // - Table capacity: The number of slots in the hash table.
-      // These are incomparable, in the sense that one doesn't imply the other.
-      // Typically we will reach space capacity before table capacity---
-      // if the user always inserts values with size equal to
-      // estimated_value_size, then at most a kLoadFactor fraction of slots
-      // will ever be occupied. But in some cases we may reach table capacity
-      // before space capacity---if the user initially claims a very large
-      // estimated_value_size but then inserts tiny values, more elements than
-      // initially estimated will be inserted.
-
-      // TODO(Guido) Some tests (at least two from cache_test, as well as the
-      // stress tests) currently assume the table capacity is unbounded.
-      if (handle == nullptr) {
-        // Don't insert the entry but still return ok, as if the entry inserted
-        // into cache and get evicted immediately.
-        last_reference_list.push_back(tmp);
-      } else {
-        if (table_.GetOccupancy() == table_.GetOccupancyLimit()) {
-          // TODO: Consider using a distinct status for this case, but usually
-          // it will be handled the same way as reaching charge capacity limit
-          s = Status::MemoryLimit(
-              "Insert failed because all slots in the hash table are full.");
-        } else {
-          s = Status::MemoryLimit(
-              "Insert failed because the total charge has exceeded the "
-              "capacity.");
-        }
-      }
-    } else {
-      // Insert into the cache. Note that the cache might get larger than its
-      // capacity if not enough space was freed up.
-      LRUHandle* old;
-      LRUHandle* h = table_.Insert(&tmp, &old);
-      assert(h != nullptr);  // We're below occupancy, so this insertion should
-                             // never fail.
-      usage_ += h->total_charge;
-      if (old != nullptr) {
-        s = Status::OkOverwritten();
-        assert(old->IsVisible());
-        table_.Exclude(old);
-        if (!old->HasRefs()) {
-          // old is on LRU because it's in cache and its reference count is 0.
-          LRU_Remove(old);
-          table_.Remove(old);
-          assert(usage_ >= old->total_charge);
-          usage_ -= old->total_charge;
-          last_reference_list.push_back(*old);
-        }
-      }
-      if (handle == nullptr) {
-        LRU_Insert(h);
-      } else {
-        // If caller already holds a ref, no need to take one here.
-        if (!h->HasRefs()) {
-          h->Ref();
-        }
-        *handle = h;
-      }
-    }
-  }
-
-  // Free the entries here outside of mutex for performance reasons.
-  for (auto& h : last_reference_list) {
-    h.FreeData();
-  }
-
-  return s;
-}
-
-LRUHandle* LRUCacheShard::Lookup(const Slice& key, uint32_t hash) {
-  LRUHandle* h = nullptr;
-  {
-    DMutexLock l(mutex_);
-    h = table_.Lookup(key, hash);
-    if (h != nullptr) {
-      assert(h->IsVisible());
-      if (!h->HasRefs()) {
-        // The entry is in LRU since it's in hash and has no external
-        // references.
-        LRU_Remove(h);
-      }
-      h->Ref();
-    }
-  }
-  return h;
-}
-
-bool LRUCacheShard::Ref(LRUHandle* h) {
-  DMutexLock l(mutex_);
-  // To create another reference - entry must be already externally referenced.
-  assert(h->HasRefs());
-  h->Ref();
-  return true;
-}
-
-bool LRUCacheShard::Release(LRUHandle* h, bool erase_if_last_ref) {
-  if (h == nullptr) {
-    return false;
-  }
-  LRUHandle copy;
-  bool last_reference = false;
-  {
-    DMutexLock l(mutex_);
-    last_reference = h->Unref();
-    if (last_reference && h->IsVisible()) {
-      // The item is still in cache, and nobody else holds a reference to it.
-      if (usage_ > capacity_ || erase_if_last_ref) {
-        // The LRU list must be empty since the cache is full.
-        assert(lru_.next == &lru_ || erase_if_last_ref);
-        // Take this opportunity and remove the item.
-        table_.Remove(h);
-      } else {
-        // Put the item back on the LRU list, and don't free it.
-        LRU_Insert(h);
-        last_reference = false;
-      }
-    }
-    // If it was the last reference, then decrement the cache usage.
-    if (last_reference) {
-      assert(usage_ >= h->total_charge);
-      usage_ -= h->total_charge;
-      copy = *h;
-    }
-  }
-
-  // Free the entry here outside of mutex for performance reasons.
-  if (last_reference) {
-    copy.FreeData();
-  }
-  return last_reference;
-}
-
-void LRUCacheShard::Erase(const Slice& key, uint32_t hash) {
-  LRUHandle copy;
-  bool last_reference = false;
-  {
-    DMutexLock l(mutex_);
-    LRUHandle* h = table_.Lookup(key, hash);
-    if (h != nullptr) {
-      table_.Exclude(h);
-      if (!h->HasRefs()) {
-        // The entry is in LRU since it's in cache and has no external
-        // references.
-        LRU_Remove(h);
-        table_.Remove(h);
-        assert(usage_ >= h->total_charge);
-        usage_ -= h->total_charge;
-        last_reference = true;
-        copy = *h;
-      }
-    }
-  }
-  // Free the entry here outside of mutex for performance reasons.
-  // last_reference will only be true if e != nullptr.
-  if (last_reference) {
-    copy.FreeData();
-  }
-}
-
-size_t LRUCacheShard::GetUsage() const {
-  DMutexLock l(mutex_);
-  return usage_;
-}
-
-size_t LRUCacheShard::GetPinnedUsage() const {
-  DMutexLock l(mutex_);
-  assert(usage_ >= lru_usage_);
-  return usage_ - lru_usage_;
-}
-
-size_t LRUCacheShard::GetOccupancyCount() const {
-  DMutexLock l(mutex_);
-  return table_.GetOccupancy();
-}
-
-size_t LRUCacheShard::GetTableAddressCount() const {
-  DMutexLock l(mutex_);
-  return table_.GetTableSize();
-}
-
-LRUCache::LRUCache(size_t capacity, size_t estimated_value_size,
-                   int num_shard_bits, bool strict_capacity_limit,
-                   CacheMetadataChargePolicy metadata_charge_policy)
-    : ShardedCache(capacity, num_shard_bits, strict_capacity_limit,
-                   nullptr /*allocator*/) {
-  assert(estimated_value_size > 0 ||
-         metadata_charge_policy != kDontChargeCacheMetadata);
-  size_t per_shard = GetPerShardCapacity();
-  InitShards([=](LRUCacheShard* cs) {
-    new (cs) LRUCacheShard(per_shard, estimated_value_size,
-                           strict_capacity_limit, metadata_charge_policy);
-  });
-}
-
-void* LRUCache::Value(Handle* handle) {
-  return reinterpret_cast<const LRUHandle*>(handle)->value;
-}
-
-size_t LRUCache::GetCharge(Handle* handle) const {
-  return reinterpret_cast<const LRUHandle*>(handle)->GetCharge(
-      GetShard(0).metadata_charge_policy_);
-}
-
-Cache::DeleterFn LRUCache::GetDeleter(Handle* handle) const {
-  auto h = reinterpret_cast<const LRUHandle*>(handle);
-  return h->deleter;
-}
-
-}  // namespace fast_lru_cache
-
-std::shared_ptr<Cache> NewFastLRUCache(
-    size_t capacity, size_t estimated_value_size, int num_shard_bits,
-    bool strict_capacity_limit,
-    CacheMetadataChargePolicy metadata_charge_policy) {
-  if (num_shard_bits >= 20) {
-    return nullptr;  // The cache cannot be sharded into too many fine pieces.
-  }
-  if (num_shard_bits < 0) {
-    num_shard_bits = GetDefaultCacheShardBits(capacity);
-  }
-  return std::make_shared<fast_lru_cache::LRUCache>(
-      capacity, estimated_value_size, num_shard_bits, strict_capacity_limit,
-      metadata_charge_policy);
-}
-
-}  // namespace ROCKSDB_NAMESPACE

cache/fast_lru_cache.h

@@ -1,477 +0,0 @@
-//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved
-//  This source code is licensed under both the GPLv2 (found in the
-//  COPYING file in the root directory) and Apache 2.0 License
-//  (found in the LICENSE.Apache file in the root directory).
-//
-// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file. See the AUTHORS file for names of contributors.
-#pragma once
-
-#include <array>
-#include <memory>
-#include <string>
-
-#include "cache/cache_key.h"
-#include "cache/sharded_cache.h"
-#include "port/lang.h"
-#include "port/malloc.h"
-#include "port/port.h"
-#include "rocksdb/secondary_cache.h"
-#include "util/autovector.h"
-#include "util/distributed_mutex.h"
-
-namespace ROCKSDB_NAMESPACE {
-
-namespace fast_lru_cache {
-
-// Forward declaration of friend class.
-class FastLRUCacheTest;
-
-// LRU cache implementation using an open-address hash table.
-//
-// Every slot in the hash table is an LRUHandle. Because handles can be
-// referenced externally, we can't discard them immediately once they are
-// deleted (via a delete or an LRU eviction) or replaced by a new version
-// (via an insert of the same key). The state of an element is defined by
-// the following two properties:
-// (R) Referenced: An element can be referenced externally (refs > 0), or not.
-//    Importantly, an element can be evicted if and only if it's not
-//    referenced. In particular, when an element becomes referenced, it's
-//    temporarily taken out of the LRU list until all references to it
-//    are dropped.
-// (V) Visible: An element can visible for lookups (IS_VISIBLE set), or not.
-//    Initially, every element is visible. An element that is not visible is
-//    called a ghost.
-// These properties induce 4 different states, with transitions defined as
-// follows:
-// - V --> not V: When a visible element is deleted or replaced by a new
-//    version.
-// - Not V --> V: This cannot happen. A ghost remains in that state until it's
-//    not referenced any more, at which point it's ready to be removed from the
-//    hash table. (A ghost simply waits to transition to the afterlife---it will
-//    never be visible again.)
-// - R --> not R: When all references to an element are dropped.
-// - Not R --> R: When an unreferenced element becomes referenced. This can only
-//    happen if the element is V, since references to an element can only be
-//    created when it's visible.
-//
-// Internally, the cache uses an open-addressed hash table to index the handles.
-// We use tombstone counters to keep track of displacements.
-// Because of the tombstones and the two possible visibility states of an
-// element, the table slots can be in 4 different states:
-// 1. Visible element (IS_ELEMENT set and IS_VISIBLE set): The slot contains a
-//    key-value element.
-// 2. Ghost element (IS_ELEMENT set and IS_VISIBLE unset): The slot contains an
-//    element that has been removed, but it's still referenced. It's invisible
-//    to lookups.
-// 3. Tombstone (IS_ELEMENT unset and displacements > 0): The slot contains a
-//    tombstone.
-// 4. Empty (IS_ELEMENT unset and displacements == 0): The slot is unused.
-//    A slot that is an element can further have IS_VISIBLE set or not.
-// When a ghost is removed from the table, it can either transition to being a
-// tombstone or an empty slot, depending on the number of displacements of the
-// slot. In any case, the slot becomes available. When a handle is inserted
-// into that slot, it becomes a visible element again.
-
-// The load factor p is a real number in (0, 1) such that at all
-// times at most a fraction p of all slots, without counting tombstones,
-// are occupied by elements. This means that the probability that a
-// random probe hits an empty slot is at most p, and thus at most 1/p probes
-// are required on average. For example, p = 70% implies that between 1 and 2
-// probes are needed on average (bear in mind that this reasoning doesn't
-// consider the effects of clustering over time).
-// Because the size of the hash table is always rounded up to the next
-// power of 2, p is really an upper bound on the actual load factor---the
-// actual load factor is anywhere between p/2 and p. This is a bit wasteful,
-// but bear in mind that slots only hold metadata, not actual values.
-// Since space cost is dominated by the values (the LSM blocks),
-// overprovisioning the table with metadata only increases the total cache space
-// usage by a tiny fraction.
-constexpr double kLoadFactor = 0.35;
-
-// The user can exceed kLoadFactor if the sizes of the inserted values don't
-// match estimated_value_size, or if strict_capacity_limit == false. To
-// avoid performance to plunge, we set a strict upper bound on the load factor.
-constexpr double kStrictLoadFactor = 0.7;
-
-// Arbitrary seeds.
-constexpr uint32_t kProbingSeed1 = 0xbc9f1d34;
-constexpr uint32_t kProbingSeed2 = 0x7a2bb9d5;
-
-// An experimental (under development!) alternative to LRUCache
-
-struct LRUHandle {
-  void* value;
-  Cache::DeleterFn deleter;
-  LRUHandle* next;
-  LRUHandle* prev;
-  size_t total_charge;  // TODO(opt): Only allow uint32_t?
-  // The hash of key(). Used for fast sharding and comparisons.
-  uint32_t hash;
-  // The number of external refs to this entry.
-  uint32_t refs;
-
-  enum Flags : uint8_t {
-    // Whether the handle is visible to Lookups.
-    IS_VISIBLE = (1 << 0),
-    // Whether the slot is in use by an element.
-    IS_ELEMENT = (1 << 1),
-  };
-  uint8_t flags;
-
-  // The number of elements that hash to this slot or a lower one,
-  // but wind up in a higher slot.
-  uint32_t displacements;
-
-  std::array<char, kCacheKeySize> key_data;
-
-  LRUHandle() {
-    value = nullptr;
-    deleter = nullptr;
-    next = nullptr;
-    prev = nullptr;
-    total_charge = 0;
-    hash = 0;
-    refs = 0;
-    flags = 0;
-    displacements = 0;
-    key_data.fill(0);
-  }
-
-  Slice key() const { return Slice(key_data.data(), kCacheKeySize); }
-
-  // For HandleImpl concept
-  uint32_t GetHash() const { return hash; }
-
-  // Increase the reference count by 1.
-  void Ref() { refs++; }
-
-  // Just reduce the reference count by 1. Return true if it was last reference.
-  bool Unref() {
-    assert(refs > 0);
-    refs--;
-    return refs == 0;
-  }
-
-  // Return true if there are external refs, false otherwise.
-  bool HasRefs() const { return refs > 0; }
-
-  bool IsVisible() const { return flags & IS_VISIBLE; }
-
-  void SetIsVisible(bool is_visible) {
-    if (is_visible) {
-      flags |= IS_VISIBLE;
-    } else {
-      flags &= ~IS_VISIBLE;
-    }
-  }
-
-  bool IsElement() const { return flags & IS_ELEMENT; }
-
-  void SetIsElement(bool is_element) {
-    if (is_element) {
-      flags |= IS_ELEMENT;
-    } else {
-      flags &= ~IS_ELEMENT;
-    }
-  }
-
-  void FreeData() {
-    assert(refs == 0);
-    if (deleter) {
-      (*deleter)(key(), value);
-    }
-  }
-
-  // Calculate the memory usage by metadata.
-  inline size_t CalcMetaCharge(
-      CacheMetadataChargePolicy metadata_charge_policy) const {
-    if (metadata_charge_policy != kFullChargeCacheMetadata) {
-      return 0;
-    } else {
-      // #ifdef ROCKSDB_MALLOC_USABLE_SIZE
-      //       return malloc_usable_size(
-      //           const_cast<void*>(static_cast<const void*>(this)));
-      // #else
-      // TODO(Guido) malloc_usable_size only works when we call it on
-      // a pointer allocated with malloc. Because our handles are all
-      // allocated in a single shot as an array, the user can't call
-      // CalcMetaCharge (or CalcTotalCharge or GetCharge) on a handle
-      // pointer returned by the cache. Moreover, malloc_usable_size
-      // expects a heap-allocated handle, but sometimes in our code we
-      // wish to pass a stack-allocated handle (this is only a performance
-      // concern).
-      // What is the right way to compute metadata charges with pre-allocated
-      // handles?
-      return sizeof(LRUHandle);
-      // #endif
-    }
-  }
-
-  inline void CalcTotalCharge(
-      size_t charge, CacheMetadataChargePolicy metadata_charge_policy) {
-    total_charge = charge + CalcMetaCharge(metadata_charge_policy);
-  }
-
-  inline size_t GetCharge(
-      CacheMetadataChargePolicy metadata_charge_policy) const {
-    size_t meta_charge = CalcMetaCharge(metadata_charge_policy);
-    assert(total_charge >= meta_charge);
-    return total_charge - meta_charge;
-  }
-
-  inline bool IsEmpty() {
-    return !this->IsElement() && this->displacements == 0;
-  }
-
-  inline bool IsTombstone() {
-    return !this->IsElement() && this->displacements > 0;
-  }
-
-  inline bool Matches(const Slice& some_key, uint32_t some_hash) {
-    return this->IsElement() && this->hash == some_hash &&
-           this->key() == some_key;
-  }
-};
-
-
-class LRUHandleTable {
- public:
-  explicit LRUHandleTable(int hash_bits);
-  ~LRUHandleTable();
-
-  // Returns a pointer to a visible element matching the key/hash, or
-  // nullptr if not present.
-  LRUHandle* Lookup(const Slice& key, uint32_t hash);
-
-  // Inserts a copy of h into the hash table.
-  // Returns a pointer to the inserted handle, or nullptr if no slot
-  // available was found. If an existing visible element matching the
-  // key/hash is already present in the hash table, the argument old
-  // is set to pointe to it; otherwise, it's set to nullptr.
-  LRUHandle* Insert(LRUHandle* h, LRUHandle** old);
-
-  // Removes h from the hash table. The handle must already be off
-  // the LRU list.
-  void Remove(LRUHandle* h);
-
-  // Turns a visible element h into a ghost (i.e., not visible).
-  void Exclude(LRUHandle* h);
-
-  // Assigns a copy of h to the given slot.
-  void Assign(int slot, LRUHandle* h);
-
-  template <typename T>
-  void ApplyToEntriesRange(T func, size_t index_begin, size_t index_end) {
-    for (size_t i = index_begin; i < index_end; i++) {
-      LRUHandle* h = &array_[i];
-      if (h->IsVisible()) {
-        func(h);
-      }
-    }
-  }
-
-  uint32_t GetTableSize() const { return uint32_t{1} << length_bits_; }
-
-  int GetLengthBits() const { return length_bits_; }
-
-  uint32_t GetOccupancyLimit() const { return occupancy_limit_; }
-
-  uint32_t GetOccupancy() const { return occupancy_; }
-
-  // Returns x mod 2^{length_bits_}.
-  uint32_t ModTableSize(uint32_t x) { return x & length_bits_mask_; }
-
- private:
-  int FindVisibleElement(const Slice& key, uint32_t hash, int& probe,
-                         int displacement);
-
-  int FindAvailableSlot(const Slice& key, int& probe, int displacement);
-
-  int FindVisibleElementOrAvailableSlot(const Slice& key, uint32_t hash,
-                                        int& probe, int displacement);
-
-  // Returns the index of the first slot probed (hashing with
-  // the given key) with a handle e such that cond(e) is true.
-  // Otherwise, if no match is found, returns -1.
-  // For every handle e probed except the final slot, updates
-  // e->displacements += displacement.
-  // The argument probe is modified such that consecutive calls
-  // to FindSlot continue probing right after where the previous
-  // call left.
-  int FindSlot(const Slice& key, std::function<bool(LRUHandle*)> cond,
-               int& probe, int displacement);
-
-  // Number of hash bits used for table index.
-  // The size of the table is 1 << length_bits_.
-  int length_bits_;
-
-  const uint32_t length_bits_mask_;
-
-  // Number of elements in the table.
-  uint32_t occupancy_;
-
-  // Maximum number of elements the user can store in the table.
-  uint32_t occupancy_limit_;
-
-  std::unique_ptr<LRUHandle[]> array_;
-};
-
-// A single shard of sharded cache.
-class ALIGN_AS(CACHE_LINE_SIZE) LRUCacheShard final : public CacheShardBase {
- public:
-  LRUCacheShard(size_t capacity, size_t estimated_value_size,
-                bool strict_capacity_limit,
-                CacheMetadataChargePolicy metadata_charge_policy);
-
-  // For CacheShard concept
-  using HandleImpl = LRUHandle;
-
-  // Keep 32-bit hashing for now (FIXME: upgrade to 64-bit)
-  using HashVal = uint32_t;
-  using HashCref = uint32_t;
-  static inline HashVal ComputeHash(const Slice& key) {
-    return Lower32of64(GetSliceNPHash64(key));
-  }
-  static inline uint32_t HashPieceForSharding(HashCref hash) { return hash; }
-
-  // Separate from constructor so caller can easily make an array of LRUCache
-  // if current usage is more than new capacity, the function will attempt to
-  // free the needed space.
-  void SetCapacity(size_t capacity);
-
-  // Set the flag to reject insertion if cache if full.
-  void SetStrictCapacityLimit(bool strict_capacity_limit);
-
-  // Like Cache methods, but with an extra "hash" parameter.
-  // Insert an item into the hash table and, if handle is null, insert into
-  // the LRU list. Older items are evicted as necessary. If the cache is full
-  // and free_handle_on_fail is true, the item is deleted and handle is set to
-  // nullptr.
-  Status Insert(const Slice& key, uint32_t hash, void* value, size_t charge,
-                Cache::DeleterFn deleter, LRUHandle** handle,
-                Cache::Priority priority);
-
-  Status Insert(const Slice& key, uint32_t hash, void* value,
-                const Cache::CacheItemHelper* helper, size_t charge,
-                LRUHandle** handle, Cache::Priority priority) {
-    return Insert(key, hash, value, charge, helper->del_cb, handle, priority);
-  }
-
-  LRUHandle* Lookup(const Slice& key, uint32_t hash,
-                    const Cache::CacheItemHelper* /*helper*/,
-                    const Cache::CreateCallback& /*create_cb*/,
-                    Cache::Priority /*priority*/, bool /*wait*/,
-                    Statistics* /*stats*/) {
-    return Lookup(key, hash);
-  }
-  LRUHandle* Lookup(const Slice& key, uint32_t hash);
-
-  bool Release(LRUHandle* handle, bool /*useful*/, bool erase_if_last_ref) {
-    return Release(handle, erase_if_last_ref);
-  }
-  bool IsReady(LRUHandle* /*handle*/) { return true; }
-  void Wait(LRUHandle* /*handle*/) {}
-
-  bool Ref(LRUHandle* handle);
-  bool Release(LRUHandle* handle, bool erase_if_last_ref = false);
-  void Erase(const Slice& key, uint32_t hash);
-
-  size_t GetUsage() const;
-  size_t GetPinnedUsage() const;
-  size_t GetOccupancyCount() const;
-  size_t GetTableAddressCount() const;
-
-  void ApplyToSomeEntries(
-      const std::function<void(const Slice& key, void* value, size_t charge,
-                               DeleterFn deleter)>& callback,
-      size_t average_entries_per_lock, size_t* state);
-
-  void EraseUnRefEntries();
-
- private:
-  friend class LRUCache;
-  friend class FastLRUCacheTest;
-
-  void LRU_Remove(LRUHandle* e);
-  void LRU_Insert(LRUHandle* e);
-
-  // Free some space following strict LRU policy until enough space
-  // to hold (usage_ + charge) is freed or the LRU list is empty
-  // This function is not thread safe - it needs to be executed while
-  // holding the mutex_.
-  void EvictFromLRU(size_t charge, autovector<LRUHandle>* deleted);
-
-  // Returns the charge of a single handle.
-  static size_t CalcEstimatedHandleCharge(
-      size_t estimated_value_size,
-      CacheMetadataChargePolicy metadata_charge_policy);
-
-  // Returns the number of bits used to hash an element in the hash
-  // table.
-  static int CalcHashBits(size_t capacity, size_t estimated_value_size,
-                          CacheMetadataChargePolicy metadata_charge_policy);
-
-  // Initialized before use.
-  size_t capacity_;
-
-  // Whether to reject insertion if cache reaches its full capacity.
-  bool strict_capacity_limit_;
-
-  // Dummy head of LRU list.
-  // lru.prev is newest entry, lru.next is oldest entry.
-  // LRU contains items which can be evicted, ie reference only by cache
-  LRUHandle lru_;
-
-  // Pointer to head of low-pri pool in LRU list.
-  LRUHandle* lru_low_pri_;
-
-  // ------------^^^^^^^^^^^^^-----------
-  // Not frequently modified data members
-  // ------------------------------------
-  //
-  // We separate data members that are updated frequently from the ones that
-  // are not frequently updated so that they don't share the same cache line
-  // which will lead into false cache sharing
-  //
-  // ------------------------------------
-  // Frequently modified data members
-  // ------------vvvvvvvvvvvvv-----------
-  LRUHandleTable table_;
-
-  // Memory size for entries residing in the cache.
-  size_t usage_;
-
-  // Memory size for entries residing only in the LRU list.
-  size_t lru_usage_;
-
-  // mutex_ protects the following state.
-  // We don't count mutex_ as the cache's internal state so semantically we
-  // don't mind mutex_ invoking the non-const actions.
-  mutable DMutex mutex_;
-};
-
-class LRUCache
-#ifdef NDEBUG
-    final
-#endif
-    : public ShardedCache<LRUCacheShard> {
- public:
-  LRUCache(size_t capacity, size_t estimated_value_size, int num_shard_bits,
-           bool strict_capacity_limit,
-           CacheMetadataChargePolicy metadata_charge_policy =
-               kDontChargeCacheMetadata);
-  const char* Name() const override { return "LRUCache"; }
-  void* Value(Handle* handle) override;
-  size_t GetCharge(Handle* handle) const override;
-  DeleterFn GetDeleter(Handle* handle) const override;
-};
-}  // namespace fast_lru_cache
-
-std::shared_ptr<Cache> NewFastLRUCache(
-    size_t capacity, size_t estimated_value_size, int num_shard_bits,
-    bool strict_capacity_limit,
-    CacheMetadataChargePolicy metadata_charge_policy);
-
-}  // namespace ROCKSDB_NAMESPACE

cache/lru_cache.cc

@@ -14,6 +14,7 @@
 #include <cstdio>
 #include <cstdlib>
 
+#include "cache/secondary_cache_adapter.h"
 #include "monitoring/perf_context_imp.h"
 #include "monitoring/statistics.h"
 #include "port/lang.h"
@@ -22,20 +23,20 @@
 namespace ROCKSDB_NAMESPACE {
 namespace lru_cache {
 
-// A distinct pointer value for marking "dummy" cache entries
-void* const kDummyValueMarker = const_cast<char*>("kDummyValueMarker");
-
-LRUHandleTable::LRUHandleTable(int max_upper_hash_bits)
+LRUHandleTable::LRUHandleTable(int max_upper_hash_bits,
+                               MemoryAllocator* allocator)
     : length_bits_(/* historical starting size*/ 4),
       list_(new LRUHandle* [size_t{1} << length_bits_] {}),
       elems_(0),
-      max_length_bits_(max_upper_hash_bits) {}
+      max_length_bits_(max_upper_hash_bits),
+      allocator_(allocator) {}
 
 LRUHandleTable::~LRUHandleTable() {
+  auto alloc = allocator_;
   ApplyToEntriesRange(
-      [](LRUHandle* h) {
+      [alloc](LRUHandle* h) {
         if (!h->HasRefs()) {
-          h->Free();
+          h->Free(alloc);
         }
       },
       0, size_t{1} << length_bits_);
@@ -95,7 +96,7 @@ void LRUHandleTable::Resize() {
   std::unique_ptr<LRUHandle* []> new_list {
     new LRUHandle* [size_t{1} << new_length_bits] {}
   };
-  uint32_t count = 0;
+  [[maybe_unused]] uint32_t count = 0;
   for (uint32_t i = 0; i < old_length; i++) {
     LRUHandle* h = list_[i];
     while (h != nullptr) {
@@ -118,7 +119,8 @@ LRUCacheShard::LRUCacheShard(size_t capacity, bool strict_capacity_limit,
                              double low_pri_pool_ratio, bool use_adaptive_mutex,
                              CacheMetadataChargePolicy metadata_charge_policy,
                              int max_upper_hash_bits,
-                             SecondaryCache* secondary_cache)
+                             MemoryAllocator* allocator,
+                             const Cache::EvictionCallback* eviction_callback)
     : CacheShardBase(metadata_charge_policy),
       capacity_(0),
       high_pri_pool_usage_(0),
@@ -128,11 +130,11 @@ LRUCacheShard::LRUCacheShard(size_t capacity, bool strict_capacity_limit,
       high_pri_pool_capacity_(0),
       low_pri_pool_ratio_(low_pri_pool_ratio),
       low_pri_pool_capacity_(0),
-      table_(max_upper_hash_bits),
+      table_(max_upper_hash_bits, allocator),
       usage_(0),
       lru_usage_(0),
       mutex_(use_adaptive_mutex),
-      secondary_cache_(secondary_cache) {
+      eviction_callback_(*eviction_callback) {
   // Make empty circular linked list.
   lru_.next = &lru_;
   lru_.prev = &lru_;
@@ -159,13 +161,14 @@ void LRUCacheShard::EraseUnRefEntries() {
   }
 
   for (auto entry : last_reference_list) {
-    entry->Free();
+    entry->Free(table_.GetAllocator());
   }
 }
 
 void LRUCacheShard::ApplyToSomeEntries(
-    const std::function<void(const Slice& key, void* value, size_t charge,
-                             DeleterFn deleter)>& callback,
+    const std::function<void(const Slice& key, Cache::ObjectPtr value,
+                             size_t charge,
+                             const Cache::CacheItemHelper* helper)>& callback,
     size_t average_entries_per_lock, size_t* state) {
   // The state is essentially going to be the starting hash, which works
   // nicely even if we resize between calls because we use upper-most
@@ -192,11 +195,8 @@ void LRUCacheShard::ApplyToSomeEntries(
   table_.ApplyToEntriesRange(
       [callback,
        metadata_charge_policy = metadata_charge_policy_](LRUHandle* h) {
-        DeleterFn deleter = h->IsSecondaryCacheCompatible()
-                                ? h->info_.helper->del_cb
-                                : h->info_.deleter;
         callback(h->key(), h->value, h->GetCharge(metadata_charge_policy),
-                 deleter);
+                 h->helper);
       },
       index_begin, index_end);
 }
@@ -334,16 +334,19 @@ void LRUCacheShard::EvictFromLRU(size_t charge,
   }
 }
 
-void LRUCacheShard::TryInsertIntoSecondaryCache(
-    autovector<LRUHandle*> evicted_handles) {
-  for (auto entry : evicted_handles) {
-    if (secondary_cache_ && entry->IsSecondaryCacheCompatible() &&
-        !entry->IsInSecondaryCache()) {
-      secondary_cache_->Insert(entry->key(), entry->value, entry->info_.helper)
-          .PermitUncheckedError();
+void LRUCacheShard::NotifyEvicted(
+    const autovector<LRUHandle*>& evicted_handles) {
+  MemoryAllocator* alloc = table_.GetAllocator();
+  for (LRUHandle* entry : evicted_handles) {
+    if (eviction_callback_ &&
+        eviction_callback_(entry->key(),
+                           reinterpret_cast<Cache::Handle*>(entry))) {
+      // Callback took ownership of obj; just free handle
+      free(entry);
+    } else {
+      // Free the entries here outside of mutex for performance reasons.
+      entry->Free(alloc);
     }
-    // Free the entries here outside of mutex for performance reasons.
-    entry->Free();
   }
 }
 
@@ -357,7 +360,7 @@ void LRUCacheShard::SetCapacity(size_t capacity) {
     EvictFromLRU(0, &last_reference_list);
   }
 
-  TryInsertIntoSecondaryCache(last_reference_list);
+  NotifyEvicted(last_reference_list);
 }
 
 void LRUCacheShard::SetStrictCapacityLimit(bool strict_capacity_limit) {
@@ -365,8 +368,7 @@ void LRUCacheShard::SetStrictCapacityLimit(bool strict_capacity_limit) {
   strict_capacity_limit_ = strict_capacity_limit;
 }
 
-Status LRUCacheShard::InsertItem(LRUHandle* e, LRUHandle** handle,
-                                 bool free_handle_on_fail) {
+Status LRUCacheShard::InsertItem(LRUHandle* e, LRUHandle** handle) {
   Status s = Status::OK();
   autovector<LRUHandle*> last_reference_list;
 
@@ -385,10 +387,9 @@ Status LRUCacheShard::InsertItem(LRUHandle* e, LRUHandle** handle,
         // into cache and get evicted immediately.
         last_reference_list.push_back(e);
       } else {
-        if (free_handle_on_fail) {
-          free(e);
-          *handle = nullptr;
-        }
+        free(e);
+        e = nullptr;
+        *handle = nullptr;
         s = Status::MemoryLimit("Insert failed due to LRU cache being full.");
       }
     } else {
@@ -420,192 +421,27 @@ Status LRUCacheShard::InsertItem(LRUHandle* e, LRUHandle** handle,
     }
   }
 
-  TryInsertIntoSecondaryCache(last_reference_list);
+  NotifyEvicted(last_reference_list);
 
   return s;
 }
 
-void LRUCacheShard::Promote(LRUHandle* e) {
-  SecondaryCacheResultHandle* secondary_handle = e->sec_handle;
-
-  assert(secondary_handle->IsReady());
-  // e is not thread-shared here; OK to modify "immutable" fields as well as
-  // "mutable" (normally requiring mutex)
-  e->SetIsPending(false);
-  e->value = secondary_handle->Value();
-  assert(e->total_charge == 0);
-  size_t value_size = secondary_handle->Size();
-  delete secondary_handle;
-
-  if (e->value) {
-    e->CalcTotalCharge(value_size, metadata_charge_policy_);
-    Status s;
-    if (e->IsStandalone()) {
-      assert(secondary_cache_ && secondary_cache_->SupportForceErase());
-
-      // Insert a dummy handle and return a standalone handle to caller.
-      // Charge the standalone handle.
-      autovector<LRUHandle*> last_reference_list;
-      bool free_standalone_handle{false};
-      {
-        DMutexLock l(mutex_);
-
-        // Free the space following strict LRU policy until enough space
-        // is freed or the lru list is empty.
-        EvictFromLRU(e->total_charge, &last_reference_list);
-
-        if ((usage_ + e->total_charge) > capacity_ && strict_capacity_limit_) {
-          free_standalone_handle = true;
-        } else {
-          usage_ += e->total_charge;
-        }
-      }
-
-      TryInsertIntoSecondaryCache(last_reference_list);
-      if (free_standalone_handle) {
-        e->Unref();
-        e->Free();
-        e = nullptr;
-      } else {
-        PERF_COUNTER_ADD(block_cache_standalone_handle_count, 1);
-      }
-
-      // Insert a dummy handle into the primary cache. This dummy handle is
-      // not IsSecondaryCacheCompatible().
-      // FIXME? This should not overwrite an existing non-dummy entry in the
-      // rare case that one exists
-      Cache::Priority priority =
-          e->IsHighPri() ? Cache::Priority::HIGH : Cache::Priority::LOW;
-      s = Insert(e->key(), e->hash, kDummyValueMarker, /*charge=*/0,
-                 /*deleter=*/nullptr, /*helper=*/nullptr, /*handle=*/nullptr,
-                 priority);
-    } else {
-      e->SetInCache(true);
-      LRUHandle* handle = e;
-      // This InsertItem() could fail if the cache is over capacity and
-      // strict_capacity_limit_ is true. In such a case, we don't want
-      // InsertItem() to free the handle, since the item is already in memory
-      // and the caller will most likely just read it from disk if we erase it
-      // here.
-      s = InsertItem(e, &handle, /*free_handle_on_fail=*/false);
-      if (s.ok()) {
-        PERF_COUNTER_ADD(block_cache_real_handle_count, 1);
-      }
-    }
-
-    if (!s.ok()) {
-      // Item is in memory, but not accounted against the cache capacity.
-      // When the handle is released, the item should get deleted.
-      assert(!e->InCache());
-    }
-  } else {
-    // Secondary cache lookup failed. The caller will take care of detecting
-    // this and eventually releasing e.
-    assert(!e->value);
-    assert(!e->InCache());
-  }
-}
-
 LRUHandle* LRUCacheShard::Lookup(const Slice& key, uint32_t hash,
-                                 const Cache::CacheItemHelper* helper,
-                                 const Cache::CreateCallback& create_cb,
-                                 Cache::Priority priority, bool wait,
-                                 Statistics* stats) {
-  LRUHandle* e = nullptr;
-  bool found_dummy_entry{false};
-  {
-    DMutexLock l(mutex_);
-    e = table_.Lookup(key, hash);
-    if (e != nullptr) {
-      assert(e->InCache());
-      if (e->value == kDummyValueMarker) {
-        // For a dummy handle, if it was retrieved from secondary cache,
-        // it may still exist in secondary cache.
-        // If the handle exists in secondary cache, the value should be
-        // erased from sec cache and be inserted into primary cache.
-        found_dummy_entry = true;
-        // Let the dummy entry be overwritten
-        e = nullptr;
-      } else {
-        if (!e->HasRefs()) {
-          // The entry is in LRU since it's in hash and has no external
-          // references.
-          LRU_Remove(e);
-        }
-        e->Ref();
-        e->SetHit();
-      }
-    }
-  }
-
-  // If handle table lookup failed or the handle is a dummy one, allocate
-  // a handle outside the mutex if we re going to lookup in the secondary cache.
-  //
-  // When a block is firstly Lookup from CompressedSecondaryCache, we just
-  // insert a dummy block into the primary cache (charging the actual size of
-  // the block) and don't erase the block from CompressedSecondaryCache. A
-  // standalone handle is returned to the caller. Only if the block is hit
-  // again, we erase it from CompressedSecondaryCache and add it into the
-  // primary cache.
-  if (!e && secondary_cache_ && helper && helper->saveto_cb) {
-    // For objects from the secondary cache, we expect the caller to provide
-    // a way to create/delete the primary cache object. The only case where
-    // a deleter would not be required is for dummy entries inserted for
-    // accounting purposes, which we won't demote to the secondary cache
-    // anyway.
-    assert(create_cb && helper->del_cb);
-    bool is_in_sec_cache{false};
-    std::unique_ptr<SecondaryCacheResultHandle> secondary_handle =
-        secondary_cache_->Lookup(key, create_cb, wait, found_dummy_entry,
-                                 is_in_sec_cache);
-    if (secondary_handle != nullptr) {
-      e = static_cast<LRUHandle*>(malloc(sizeof(LRUHandle) - 1 + key.size()));
-
-      e->m_flags = 0;
-      e->im_flags = 0;
-      e->SetSecondaryCacheCompatible(true);
-      e->info_.helper = helper;
-      e->key_length = key.size();
-      e->hash = hash;
-      e->refs = 0;
-      e->next = e->prev = nullptr;
-      e->SetPriority(priority);
-      memcpy(e->key_data, key.data(), key.size());
-      e->value = nullptr;
-      e->sec_handle = secondary_handle.release();
-      e->total_charge = 0;
-      e->Ref();
-      e->SetIsInSecondaryCache(is_in_sec_cache);
-      e->SetIsStandalone(secondary_cache_->SupportForceErase() &&
-                         !found_dummy_entry);
-
-      if (wait) {
-        Promote(e);
-        if (e) {
-          if (!e->value) {
-            // The secondary cache returned a handle, but the lookup failed.
-            e->Unref();
-            e->Free();
-            e = nullptr;
-          } else {
-            PERF_COUNTER_ADD(secondary_cache_hit_count, 1);
-            RecordTick(stats, SECONDARY_CACHE_HITS);
-          }
-        }
-      } else {
-        // If wait is false, we always return a handle and let the caller
-        // release the handle after checking for success or failure.
-        e->SetIsPending(true);
-        // This may be slightly inaccurate, if the lookup eventually fails.
-        // But the probability is very low.
-        PERF_COUNTER_ADD(secondary_cache_hit_count, 1);
-        RecordTick(stats, SECONDARY_CACHE_HITS);
-      }
-    } else {
-      // Caller will most likely overwrite the dummy entry with an Insert
-      // after this Lookup fails
-      assert(e == nullptr);
+                                 const Cache::CacheItemHelper* /*helper*/,
+                                 Cache::CreateContext* /*create_context*/,
+                                 Cache::Priority /*priority*/,
+                                 Statistics* /*stats*/) {
+  DMutexLock l(mutex_);
+  LRUHandle* e = table_.Lookup(key, hash);
+  if (e != nullptr) {
+    assert(e->InCache());
+    if (!e->HasRefs()) {
+      // The entry is in LRU since it's in hash and has no external
+      // references.
+      LRU_Remove(e);
     }
+    e->Ref();
+    e->SetHit();
   }
   return e;
 }
@@ -614,8 +450,6 @@ bool LRUCacheShard::Ref(LRUHandle* e) {
   DMutexLock l(mutex_);
   // To create another reference - entry must be already externally referenced.
   assert(e->HasRefs());
-  // Pending handles are not for sharing
-  assert(!e->IsPending());
   e->Ref();
   return true;
 }
@@ -639,14 +473,13 @@ bool LRUCacheShard::Release(LRUHandle* e, bool /*useful*/,
   if (e == nullptr) {
     return false;
   }
-  bool last_reference = false;
-  // Must Wait or WaitAll first on pending handles. Otherwise, would leak
-  // a secondary cache handle.
-  assert(!e->IsPending());
+  bool must_free;
+  bool was_in_cache;
   {
     DMutexLock l(mutex_);
-    last_reference = e->Unref();
-    if (last_reference && e->InCache()) {
+    must_free = e->Unref();
+    was_in_cache = e->InCache();
+    if (must_free && was_in_cache) {
       // The item is still in cache, and nobody else holds a reference to it.
       if (usage_ > capacity_ || erase_if_last_ref) {
         // The LRU list must be empty since the cache is full.
@@ -657,28 +490,39 @@ bool LRUCacheShard::Release(LRUHandle* e, bool /*useful*/,
       } else {
         // Put the item back on the LRU list, and don't free it.
         LRU_Insert(e);
-        last_reference = false;
+        must_free = false;
       }
     }
-    // If it was the last reference, then decrement the cache usage.
-    if (last_reference) {
+    // If about to be freed, then decrement the cache usage.
+    if (must_free) {
       assert(usage_ >= e->total_charge);
       usage_ -= e->total_charge;
     }
   }
 
   // Free the entry here outside of mutex for performance reasons.
-  if (last_reference) {
-    e->Free();
+  if (must_free) {
+    // Only call eviction callback if we're sure no one requested erasure
+    // FIXME: disabled because of test churn
+    if (false && was_in_cache && !erase_if_last_ref && eviction_callback_ &&
+        eviction_callback_(e->key(), reinterpret_cast<Cache::Handle*>(e))) {
+      // Callback took ownership of obj; just free handle
+      free(e);
+    } else {
+      e->Free(table_.GetAllocator());
+    }
   }
-  return last_reference;
+  return must_free;
 }
 
-Status LRUCacheShard::Insert(const Slice& key, uint32_t hash, void* value,
-                             size_t charge,
-                             void (*deleter)(const Slice& key, void* value),
-                             const Cache::CacheItemHelper* helper,
-                             LRUHandle** handle, Cache::Priority priority) {
+LRUHandle* LRUCacheShard::CreateHandle(const Slice& key, uint32_t hash,
+                                       Cache::ObjectPtr value,
+                                       const Cache::CacheItemHelper* helper,
+                                       size_t charge) {
+  assert(helper);
+  // value == nullptr is reserved for indicating failure in SecondaryCache
+  assert(!(helper->IsSecondaryCacheCompatible() && value == nullptr));
+
   // Allocate the memory here outside of the mutex.
   // If the cache is full, we'll have to release it.
   // It shouldn't happen very often though.
@@ -688,27 +532,58 @@ Status LRUCacheShard::Insert(const Slice& key, uint32_t hash, void* value,
   e->value = value;
   e->m_flags = 0;
   e->im_flags = 0;
-  if (helper) {
-    // Use only one of the two parameters
-    assert(deleter == nullptr);
-    // value == nullptr is reserved for indicating failure for when secondary
-    // cache compatible
-    assert(value != nullptr);
-    e->SetSecondaryCacheCompatible(true);
-    e->info_.helper = helper;
-  } else {
-    e->info_.deleter = deleter;
-  }
+  e->helper = helper;
   e->key_length = key.size();
   e->hash = hash;
   e->refs = 0;
   e->next = e->prev = nullptr;
-  e->SetInCache(true);
-  e->SetPriority(priority);
   memcpy(e->key_data, key.data(), key.size());
   e->CalcTotalCharge(charge, metadata_charge_policy_);
 
-  return InsertItem(e, handle, /* free_handle_on_fail */ true);
+  return e;
+}
+
+Status LRUCacheShard::Insert(const Slice& key, uint32_t hash,
+                             Cache::ObjectPtr value,
+                             const Cache::CacheItemHelper* helper,
+                             size_t charge, LRUHandle** handle,
+                             Cache::Priority priority) {
+  LRUHandle* e = CreateHandle(key, hash, value, helper, charge);
+  e->SetPriority(priority);
+  e->SetInCache(true);
+  return InsertItem(e, handle);
+}
+
+LRUHandle* LRUCacheShard::CreateStandalone(const Slice& key, uint32_t hash,
+                                           Cache::ObjectPtr value,
+                                           const Cache::CacheItemHelper* helper,
+                                           size_t charge,
+                                           bool allow_uncharged) {
+  LRUHandle* e = CreateHandle(key, hash, value, helper, charge);
+  e->SetIsStandalone(true);
+  e->Ref();
+
+  autovector<LRUHandle*> last_reference_list;
+
+  {
+    DMutexLock l(mutex_);
+
+    EvictFromLRU(e->total_charge, &last_reference_list);
+
+    if (strict_capacity_limit_ && (usage_ + e->total_charge) > capacity_) {
+      if (allow_uncharged) {
+        e->total_charge = 0;
+      } else {
+        free(e);
+        e = nullptr;
+      }
+    } else {
+      usage_ += e->total_charge;
+    }
+  }
+
+  NotifyEvicted(last_reference_list);
+  return e;
 }
 
 void LRUCacheShard::Erase(const Slice& key, uint32_t hash) {
@@ -733,20 +608,10 @@ void LRUCacheShard::Erase(const Slice& key, uint32_t hash) {
   // Free the entry here outside of mutex for performance reasons.
   // last_reference will only be true if e != nullptr.
   if (last_reference) {
-    e->Free();
+    e->Free(table_.GetAllocator());
   }
 }
 
-bool LRUCacheShard::IsReady(LRUHandle* e) {
-  bool ready = true;
-  if (e->IsPending()) {
-    assert(secondary_cache_);
-    assert(e->sec_handle);
-    ready = e->sec_handle->IsReady();
-  }
-  return ready;
-}
-
 size_t LRUCacheShard::GetUsage() const {
   DMutexLock l(mutex_);
   return usage_;
@@ -786,25 +651,23 @@ LRUCache::LRUCache(size_t capacity, int num_shard_bits,
                    double low_pri_pool_ratio,
                    std::shared_ptr<MemoryAllocator> allocator,
                    bool use_adaptive_mutex,
-                   CacheMetadataChargePolicy metadata_charge_policy,
-                   std::shared_ptr<SecondaryCache> _secondary_cache)
+                   CacheMetadataChargePolicy metadata_charge_policy)
     : ShardedCache(capacity, num_shard_bits, strict_capacity_limit,
-                   std::move(allocator)),
-      secondary_cache_(std::move(_secondary_cache)) {
+                   std::move(allocator)) {
   size_t per_shard = GetPerShardCapacity();
-  SecondaryCache* secondary_cache = secondary_cache_.get();
+  MemoryAllocator* alloc = memory_allocator();
+  const EvictionCallback* eviction_callback = &eviction_callback_;
   InitShards([=](LRUCacheShard* cs) {
-    new (cs) LRUCacheShard(
-        per_shard, strict_capacity_limit, high_pri_pool_ratio,
-        low_pri_pool_ratio, use_adaptive_mutex, metadata_charge_policy,
-        /* max_upper_hash_bits */ 32 - num_shard_bits, secondary_cache);
+    new (cs) LRUCacheShard(per_shard, strict_capacity_limit,
+                           high_pri_pool_ratio, low_pri_pool_ratio,
+                           use_adaptive_mutex, metadata_charge_policy,
+                           /* max_upper_hash_bits */ 32 - num_shard_bits, alloc,
+                           eviction_callback);
   });
 }
 
-void* LRUCache::Value(Handle* handle) {
+Cache::ObjectPtr LRUCache::Value(Handle* handle) {
   auto h = reinterpret_cast<const LRUHandle*>(handle);
-  assert(!h->IsPending() || h->value == nullptr);
-  assert(h->value != kDummyValueMarker);
   return h->value;
 }
 
@@ -813,13 +676,10 @@ size_t LRUCache::GetCharge(Handle* handle) const {
       GetShard(0).metadata_charge_policy_);
 }
 
-Cache::DeleterFn LRUCache::GetDeleter(Handle* handle) const {
+const Cache::CacheItemHelper* LRUCache::GetCacheItemHelper(
+    Handle* handle) const {
   auto h = reinterpret_cast<const LRUHandle*>(handle);
-  if (h->IsSecondaryCacheCompatible()) {
-    return h->info_.helper->del_cb;
-  } else {
-    return h->info_.deleter;
-  }
+  return h->helper;
 }
 
 size_t LRUCache::TEST_GetLRUSize() {
@@ -830,42 +690,6 @@ double LRUCache::GetHighPriPoolRatio() {
   return GetShard(0).GetHighPriPoolRatio();
 }
 
-void LRUCache::WaitAll(std::vector<Handle*>& handles) {
-  if (secondary_cache_) {
-    std::vector<SecondaryCacheResultHandle*> sec_handles;
-    sec_handles.reserve(handles.size());
-    for (Handle* handle : handles) {
-      if (!handle) {
-        continue;
-      }
-      LRUHandle* lru_handle = reinterpret_cast<LRUHandle*>(handle);
-      if (!lru_handle->IsPending()) {
-        continue;
-      }
-      sec_handles.emplace_back(lru_handle->sec_handle);
-    }
-    secondary_cache_->WaitAll(sec_handles);
-    for (Handle* handle : handles) {
-      if (!handle) {
-        continue;
-      }
-      LRUHandle* lru_handle = reinterpret_cast<LRUHandle*>(handle);
-      if (!lru_handle->IsPending()) {
-        continue;
-      }
-      GetShard(lru_handle->hash).Promote(lru_handle);
-    }
-  }
-}
-
-void LRUCache::AppendPrintableOptions(std::string& str) const {
-  ShardedCache::AppendPrintableOptions(str);  // options from shard
-  if (secondary_cache_) {
-    str.append("  secondary_cache:\n");
-    str.append(secondary_cache_->GetPrintableOptions());
-  }
-}
-
 }  // namespace lru_cache
 
 std::shared_ptr<Cache> NewLRUCache(
@@ -893,10 +717,14 @@ std::shared_ptr<Cache> NewLRUCache(
   if (num_shard_bits < 0) {
     num_shard_bits = GetDefaultCacheShardBits(capacity);
   }
-  return std::make_shared<LRUCache>(
+  std::shared_ptr<Cache> cache = std::make_shared<LRUCache>(
       capacity, num_shard_bits, strict_capacity_limit, high_pri_pool_ratio,
       low_pri_pool_ratio, std::move(memory_allocator), use_adaptive_mutex,
-      metadata_charge_policy, secondary_cache);
+      metadata_charge_policy);
+  if (secondary_cache) {
+    cache = std::make_shared<CacheWithSecondaryAdapter>(cache, secondary_cache);
+  }
+  return cache;
 }
 
 std::shared_ptr<Cache> NewLRUCache(const LRUCacheOptions& cache_opts) {

cache/lru_cache.h

@@ -13,9 +13,9 @@
 
 #include "cache/sharded_cache.h"
 #include "port/lang.h"
+#include "port/likely.h"
 #include "port/malloc.h"
 #include "port/port.h"
-#include "rocksdb/secondary_cache.h"
 #include "util/autovector.h"
 #include "util/distributed_mutex.h"
 
@@ -48,19 +48,9 @@ namespace lru_cache {
 // While refs > 0, public properties like value and deleter must not change.
 
 struct LRUHandle {
-  void* value;
-  union Info {
-    Info() {}
-    ~Info() {}
-    Cache::DeleterFn deleter;
-    const Cache::CacheItemHelper* helper;
-  } info_;
-  // An entry is not added to the LRUHandleTable until the secondary cache
-  // lookup is complete, so its safe to have this union.
-  union {
-    LRUHandle* next_hash;
-    SecondaryCacheResultHandle* sec_handle;
-  };
+  Cache::ObjectPtr value;
+  const Cache::CacheItemHelper* helper;
+  LRUHandle* next_hash;
   LRUHandle* next;
   LRUHandle* prev;
   size_t total_charge;  // TODO(opt): Only allow uint32_t?
@@ -93,14 +83,8 @@ struct LRUHandle {
     IM_IS_HIGH_PRI = (1 << 0),
     // Whether this entry is low priority entry.
     IM_IS_LOW_PRI = (1 << 1),
-    // Can this be inserted into the secondary cache.
-    IM_IS_SECONDARY_CACHE_COMPATIBLE = (1 << 2),
-    // Is the handle still being read from a lower tier.
-    IM_IS_PENDING = (1 << 3),
-    // Whether this handle is still in a lower tier
-    IM_IS_IN_SECONDARY_CACHE = (1 << 4),
     // Marks result handles that should not be inserted into cache
-    IM_IS_STANDALONE = (1 << 5),
+    IM_IS_STANDALONE = (1 << 2),
   };
 
   // Beginning of the key (MUST BE THE LAST FIELD IN THIS STRUCT!)
@@ -130,13 +114,6 @@ struct LRUHandle {
   bool IsLowPri() const { return im_flags & IM_IS_LOW_PRI; }
   bool InLowPriPool() const { return m_flags & M_IN_LOW_PRI_POOL; }
   bool HasHit() const { return m_flags & M_HAS_HIT; }
-  bool IsSecondaryCacheCompatible() const {
-    return im_flags & IM_IS_SECONDARY_CACHE_COMPATIBLE;
-  }
-  bool IsPending() const { return im_flags & IM_IS_PENDING; }
-  bool IsInSecondaryCache() const {
-    return im_flags & IM_IS_IN_SECONDARY_CACHE;
-  }
   bool IsStandalone() const { return im_flags & IM_IS_STANDALONE; }
 
   void SetInCache(bool in_cache) {
@@ -178,30 +155,6 @@ struct LRUHandle {
 
   void SetHit() { m_flags |= M_HAS_HIT; }
 
-  void SetSecondaryCacheCompatible(bool compat) {
-    if (compat) {
-      im_flags |= IM_IS_SECONDARY_CACHE_COMPATIBLE;
-    } else {
-      im_flags &= ~IM_IS_SECONDARY_CACHE_COMPATIBLE;
-    }
-  }
-
-  void SetIsPending(bool pending) {
-    if (pending) {
-      im_flags |= IM_IS_PENDING;
-    } else {
-      im_flags &= ~IM_IS_PENDING;
-    }
-  }
-
-  void SetIsInSecondaryCache(bool is_in_secondary_cache) {
-    if (is_in_secondary_cache) {
-      im_flags |= IM_IS_IN_SECONDARY_CACHE;
-    } else {
-      im_flags &= ~IM_IS_IN_SECONDARY_CACHE;
-    }
-  }
-
   void SetIsStandalone(bool is_standalone) {
     if (is_standalone) {
       im_flags |= IM_IS_STANDALONE;
@@ -210,22 +163,11 @@ struct LRUHandle {
     }
   }
 
-  void Free() {
+  void Free(MemoryAllocator* allocator) {
     assert(refs == 0);
-
-    if (!IsSecondaryCacheCompatible() && info_.deleter) {
-      (*info_.deleter)(key(), value);
-    } else if (IsSecondaryCacheCompatible()) {
-      if (IsPending()) {
-        assert(sec_handle != nullptr);
-        SecondaryCacheResultHandle* tmp_sec_handle = sec_handle;
-        tmp_sec_handle->Wait();
-        value = tmp_sec_handle->Value();
-        delete tmp_sec_handle;
-      }
-      if (value) {
-        (*info_.helper->del_cb)(key(), value);
-      }
+    assert(helper);
+    if (helper->del_cb) {
+      helper->del_cb(value, allocator);
     }
 
     free(this);
@@ -267,7 +209,7 @@ struct LRUHandle {
 // 4.4.3's builtin hashtable.
 class LRUHandleTable {
  public:
-  explicit LRUHandleTable(int max_upper_hash_bits);
+  explicit LRUHandleTable(int max_upper_hash_bits, MemoryAllocator* allocator);
   ~LRUHandleTable();
 
   LRUHandle* Lookup(const Slice& key, uint32_t hash);
@@ -291,6 +233,8 @@ class LRUHandleTable {
 
   size_t GetOccupancyCount() const { return elems_; }
 
+  MemoryAllocator* GetAllocator() const { return allocator_; }
+
  private:
   // Return a pointer to slot that points to a cache entry that
   // matches key/hash.  If there is no such cache entry, return a
@@ -312,16 +256,22 @@ class LRUHandleTable {
 
   // Set from max_upper_hash_bits (see constructor).
   const int max_length_bits_;
+
+  // From Cache, needed for delete
+  MemoryAllocator* const allocator_;
 };
 
 // A single shard of sharded cache.
 class ALIGN_AS(CACHE_LINE_SIZE) LRUCacheShard final : public CacheShardBase {
  public:
+  // NOTE: the eviction_callback ptr is saved, as is it assumed to be kept
+  // alive in Cache.
   LRUCacheShard(size_t capacity, bool strict_capacity_limit,
                 double high_pri_pool_ratio, double low_pri_pool_ratio,
                 bool use_adaptive_mutex,
                 CacheMetadataChargePolicy metadata_charge_policy,
-                int max_upper_hash_bits, SecondaryCache* secondary_cache);
+                int max_upper_hash_bits, MemoryAllocator* allocator,
+                const Cache::EvictionCallback* eviction_callback);
 
  public:  // Type definitions expected as parameter to ShardedCache
   using HandleImpl = LRUHandle;
@@ -348,29 +298,21 @@ class ALIGN_AS(CACHE_LINE_SIZE) LRUCacheShard final : public CacheShardBase {
   void SetLowPriorityPoolRatio(double low_pri_pool_ratio);
 
   // Like Cache methods, but with an extra "hash" parameter.
-  inline Status Insert(const Slice& key, uint32_t hash, void* value,
-                       size_t charge, Cache::DeleterFn deleter,
-                       LRUHandle** handle, Cache::Priority priority) {
-    return Insert(key, hash, value, charge, deleter, nullptr, handle, priority);
-  }
-  inline Status Insert(const Slice& key, uint32_t hash, void* value,
-                       const Cache::CacheItemHelper* helper, size_t charge,
-                       LRUHandle** handle, Cache::Priority priority) {
-    assert(helper);
-    return Insert(key, hash, value, charge, nullptr, helper, handle, priority);
-  }
-  // If helper_cb is null, the values of the following arguments don't matter.
+  Status Insert(const Slice& key, uint32_t hash, Cache::ObjectPtr value,
+                const Cache::CacheItemHelper* helper, size_t charge,
+                LRUHandle** handle, Cache::Priority priority);
+
+  LRUHandle* CreateStandalone(const Slice& key, uint32_t hash,
+                              Cache::ObjectPtr obj,
+                              const Cache::CacheItemHelper* helper,
+                              size_t charge, bool allow_uncharged);
+
   LRUHandle* Lookup(const Slice& key, uint32_t hash,
                     const Cache::CacheItemHelper* helper,
-                    const Cache::CreateCallback& create_cb,
-                    Cache::Priority priority, bool wait, Statistics* stats);
-  inline LRUHandle* Lookup(const Slice& key, uint32_t hash) {
-    return Lookup(key, hash, nullptr, nullptr, Cache::Priority::LOW, true,
-                  nullptr);
-  }
+                    Cache::CreateContext* create_context,
+                    Cache::Priority priority, Statistics* stats);
+
   bool Release(LRUHandle* handle, bool useful, bool erase_if_last_ref);
-  bool IsReady(LRUHandle* /*handle*/);
-  void Wait(LRUHandle* /*handle*/) {}
   bool Ref(LRUHandle* handle);
   void Erase(const Slice& key, uint32_t hash);
 
@@ -384,8 +326,9 @@ class ALIGN_AS(CACHE_LINE_SIZE) LRUCacheShard final : public CacheShardBase {
   size_t GetTableAddressCount() const;
 
   void ApplyToSomeEntries(
-      const std::function<void(const Slice& key, void* value, size_t charge,
-                               DeleterFn deleter)>& callback,
+      const std::function<void(const Slice& key, Cache::ObjectPtr value,
+                               size_t charge,
+                               const Cache::CacheItemHelper* helper)>& callback,
       size_t average_entries_per_lock, size_t* state);
 
   void EraseUnRefEntries();
@@ -409,23 +352,10 @@ class ALIGN_AS(CACHE_LINE_SIZE) LRUCacheShard final : public CacheShardBase {
  private:
   friend class LRUCache;
   // Insert an item into the hash table and, if handle is null, insert into
-  // the LRU list. Older items are evicted as necessary. If the cache is full
-  // and free_handle_on_fail is true, the item is deleted and handle is set to
-  // nullptr.
-  Status InsertItem(LRUHandle* item, LRUHandle** handle,
-                    bool free_handle_on_fail);
-  Status Insert(const Slice& key, uint32_t hash, void* value, size_t charge,
-                DeleterFn deleter, const Cache::CacheItemHelper* helper,
-                LRUHandle** handle, Cache::Priority priority);
-  // Promote an item looked up from the secondary cache to the LRU cache.
-  // The item may be still in the secondary cache.
-  // It is only inserted into the hash table and not the LRU list, and only
-  // if the cache is not at full capacity, as is the case during Insert.  The
-  // caller should hold a reference on the LRUHandle. When the caller releases
-  // the last reference, the item is added to the LRU list.
-  // The item is promoted to the high pri or low pri pool as specified by the
-  // caller in Lookup.
-  void Promote(LRUHandle* e);
+  // the LRU list. Older items are evicted as necessary. Frees `item` on
+  // non-OK status.
+  Status InsertItem(LRUHandle* item, LRUHandle** handle);
+
   void LRU_Remove(LRUHandle* e);
   void LRU_Insert(LRUHandle* e);
 
@@ -439,8 +369,11 @@ class ALIGN_AS(CACHE_LINE_SIZE) LRUCacheShard final : public CacheShardBase {
   // holding the mutex_.
   void EvictFromLRU(size_t charge, autovector<LRUHandle*>* deleted);
 
-  // Try to insert the evicted handles into the secondary cache.
-  void TryInsertIntoSecondaryCache(autovector<LRUHandle*> evicted_handles);
+  void NotifyEvicted(const autovector<LRUHandle*>& evicted_handles);
+
+  LRUHandle* CreateHandle(const Slice& key, uint32_t hash,
+                          Cache::ObjectPtr value,
+                          const Cache::CacheItemHelper* helper, size_t charge);
 
   // Initialized before use.
   size_t capacity_;
@@ -503,8 +436,8 @@ class ALIGN_AS(CACHE_LINE_SIZE) LRUCacheShard final : public CacheShardBase {
   // don't mind mutex_ invoking the non-const actions.
   mutable DMutex mutex_;
 
-  // Owned by LRUCache
-  SecondaryCache* secondary_cache_;
+  // A reference to Cache::eviction_callback_
+  const Cache::EvictionCallback& eviction_callback_;
 };
 
 class LRUCache
@@ -518,23 +451,16 @@ class LRUCache
            std::shared_ptr<MemoryAllocator> memory_allocator = nullptr,
            bool use_adaptive_mutex = kDefaultToAdaptiveMutex,
            CacheMetadataChargePolicy metadata_charge_policy =
-               kDontChargeCacheMetadata,
-           std::shared_ptr<SecondaryCache> secondary_cache = nullptr);
+               kDontChargeCacheMetadata);
   const char* Name() const override { return "LRUCache"; }
-  void* Value(Handle* handle) override;
+  ObjectPtr Value(Handle* handle) override;
   size_t GetCharge(Handle* handle) const override;
-  DeleterFn GetDeleter(Handle* handle) const override;
-  void WaitAll(std::vector<Handle*>& handles) override;
+  const CacheItemHelper* GetCacheItemHelper(Handle* handle) const override;
 
   // Retrieves number of elements in LRU, for unit test purpose only.
   size_t TEST_GetLRUSize();
   // Retrieves high pri pool ratio.
   double GetHighPriPoolRatio();
-
-  void AppendPrintableOptions(std::string& str) const override;
-
- private:
-  std::shared_ptr<SecondaryCache> secondary_cache_;
 };
 
 }  // namespace lru_cache

cache/secondary_cache.cc

@@ -0,0 +1,44 @@
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under both the GPLv2 (found in the
+//  COPYING file in the root directory) and Apache 2.0 License
+//  (found in the LICENSE.Apache file in the root directory).
+
+#include "rocksdb/secondary_cache.h"
+
+#include "cache/cache_entry_roles.h"
+
+namespace ROCKSDB_NAMESPACE {
+
+namespace {
+
+void NoopDelete(Cache::ObjectPtr, MemoryAllocator*) {}
+
+size_t SliceSize(Cache::ObjectPtr obj) {
+  return static_cast<Slice*>(obj)->size();
+}
+
+Status SliceSaveTo(Cache::ObjectPtr from_obj, size_t from_offset, size_t length,
+                   char* out) {
+  const Slice& slice = *static_cast<Slice*>(from_obj);
+  std::memcpy(out, slice.data() + from_offset, length);
+  return Status::OK();
+}
+
+Status FailCreate(const Slice&, Cache::CreateContext*, MemoryAllocator*,
+                  Cache::ObjectPtr*, size_t*) {
+  return Status::NotSupported("Only for dumping data into SecondaryCache");
+}
+
+}  // namespace
+
+Status SecondaryCache::InsertSaved(const Slice& key, const Slice& saved) {
+  static Cache::CacheItemHelper helper_no_secondary{CacheEntryRole::kMisc,
+                                                    &NoopDelete};
+  static Cache::CacheItemHelper helper{
+      CacheEntryRole::kMisc, &NoopDelete, &SliceSize,
+      &SliceSaveTo,          &FailCreate, &helper_no_secondary};
+  // NOTE: depends on Insert() being synchronous, not keeping pointer `&saved`
+  return Insert(key, const_cast<Slice*>(&saved), &helper);
+}
+
+}  // namespace ROCKSDB_NAMESPACE

cache/secondary_cache_adapter.cc

@@ -0,0 +1,295 @@
+//  Copyright (c) Meta Platforms, Inc. and affiliates.
+//  This source code is licensed under both the GPLv2 (found in the
+//  COPYING file in the root directory) and Apache 2.0 License
+//  (found in the LICENSE.Apache file in the root directory).
+
+#include "cache/secondary_cache_adapter.h"
+
+#include "monitoring/perf_context_imp.h"
+
+namespace ROCKSDB_NAMESPACE {
+
+namespace {
+// A distinct pointer value for marking "dummy" cache entries
+struct Dummy {
+  char val[7] = "kDummy";
+};
+const Dummy kDummy{};
+Cache::ObjectPtr const kDummyObj = const_cast<Dummy*>(&kDummy);
+}  // namespace
+
+CacheWithSecondaryAdapter::CacheWithSecondaryAdapter(
+    std::shared_ptr<Cache> target,
+    std::shared_ptr<SecondaryCache> secondary_cache)
+    : CacheWrapper(std::move(target)),
+      secondary_cache_(std::move(secondary_cache)) {
+  target_->SetEvictionCallback([this](const Slice& key, Handle* handle) {
+    return EvictionHandler(key, handle);
+  });
+}
+
+CacheWithSecondaryAdapter::~CacheWithSecondaryAdapter() {
+  // `*this` will be destroyed before `*target_`, so we have to prevent
+  // use after free
+  target_->SetEvictionCallback({});
+}
+
+bool CacheWithSecondaryAdapter::EvictionHandler(const Slice& key,
+                                                Handle* handle) {
+  auto helper = GetCacheItemHelper(handle);
+  if (helper->IsSecondaryCacheCompatible()) {
+    auto obj = target_->Value(handle);
+    // Ignore dummy entry
+    if (obj != kDummyObj) {
+      // Spill into secondary cache.
+      secondary_cache_->Insert(key, obj, helper).PermitUncheckedError();
+    }
+  }
+  // Never takes ownership of obj
+  return false;
+}
+
+bool CacheWithSecondaryAdapter::ProcessDummyResult(Cache::Handle** handle,
+                                                   bool erase) {
+  if (*handle && target_->Value(*handle) == kDummyObj) {
+    target_->Release(*handle, erase);
+    *handle = nullptr;
+    return true;
+  } else {
+    return false;
+  }
+}
+
+void CacheWithSecondaryAdapter::CleanupCacheObject(
+    ObjectPtr obj, const CacheItemHelper* helper) {
+  if (helper->del_cb) {
+    helper->del_cb(obj, memory_allocator());
+  }
+}
+
+Cache::Handle* CacheWithSecondaryAdapter::Promote(
+    std::unique_ptr<SecondaryCacheResultHandle>&& secondary_handle,
+    const Slice& key, const CacheItemHelper* helper, Priority priority,
+    Statistics* stats, bool found_dummy_entry, bool kept_in_sec_cache) {
+  assert(secondary_handle->IsReady());
+
+  ObjectPtr obj = secondary_handle->Value();
+  if (!obj) {
+    // Nothing found.
+    return nullptr;
+  }
+  // Found something.
+  switch (helper->role) {
+    case CacheEntryRole::kFilterBlock:
+      RecordTick(stats, SECONDARY_CACHE_FILTER_HITS);
+      break;
+    case CacheEntryRole::kIndexBlock:
+      RecordTick(stats, SECONDARY_CACHE_INDEX_HITS);
+      break;
+    case CacheEntryRole::kDataBlock:
+      RecordTick(stats, SECONDARY_CACHE_DATA_HITS);
+      break;
+    default:
+      break;
+  }
+  PERF_COUNTER_ADD(secondary_cache_hit_count, 1);
+  RecordTick(stats, SECONDARY_CACHE_HITS);
+
+  // Note: SecondaryCache::Size() is really charge (from the CreateCallback)
+  size_t charge = secondary_handle->Size();
+  Handle* result = nullptr;
+  // Insert into primary cache, possibly as a standalone+dummy entries.
+  if (secondary_cache_->SupportForceErase() && !found_dummy_entry) {
+    // Create standalone and insert dummy
+    // Allow standalone to be created even if cache is full, to avoid
+    // reading the entry from storage.
+    result =
+        CreateStandalone(key, obj, helper, charge, /*allow_uncharged*/ true);
+    assert(result);
+    PERF_COUNTER_ADD(block_cache_standalone_handle_count, 1);
+
+    // Insert dummy to record recent use
+    // TODO: try to avoid case where inserting this dummy could overwrite a
+    // regular entry
+    Status s = Insert(key, kDummyObj, &kNoopCacheItemHelper, /*charge=*/0,
+                      /*handle=*/nullptr, priority);
+    s.PermitUncheckedError();
+    // Nothing to do or clean up on dummy insertion failure
+  } else {
+    // Insert regular entry into primary cache.
+    // Don't allow it to spill into secondary cache again if it was kept there.
+    Status s = Insert(
+        key, obj, kept_in_sec_cache ? helper->without_secondary_compat : helper,
+        charge, &result, priority);
+    if (s.ok()) {
+      assert(result);
+      PERF_COUNTER_ADD(block_cache_real_handle_count, 1);
+    } else {
+      // Create standalone result instead, even if cache is full, to avoid
+      // reading the entry from storage.
+      result =
+          CreateStandalone(key, obj, helper, charge, /*allow_uncharged*/ true);
+      assert(result);
+      PERF_COUNTER_ADD(block_cache_standalone_handle_count, 1);
+    }
+  }
+  return result;
+}
+
+Cache::Handle* CacheWithSecondaryAdapter::Lookup(const Slice& key,
+                                                 const CacheItemHelper* helper,
+                                                 CreateContext* create_context,
+                                                 Priority priority,
+                                                 Statistics* stats) {
+  // NOTE: we could just StartAsyncLookup() and Wait(), but this should be a bit
+  // more efficient
+  Handle* result =
+      target_->Lookup(key, helper, create_context, priority, stats);
+  bool secondary_compatible = helper && helper->IsSecondaryCacheCompatible();
+  bool found_dummy_entry =
+      ProcessDummyResult(&result, /*erase=*/secondary_compatible);
+  if (!result && secondary_compatible) {
+    // Try our secondary cache
+    bool kept_in_sec_cache = false;
+    std::unique_ptr<SecondaryCacheResultHandle> secondary_handle =
+        secondary_cache_->Lookup(key, helper, create_context, /*wait*/ true,
+                                 found_dummy_entry, /*out*/ kept_in_sec_cache);
+    if (secondary_handle) {
+      result = Promote(std::move(secondary_handle), key, helper, priority,
+                       stats, found_dummy_entry, kept_in_sec_cache);
+    }
+  }
+  return result;
+}
+
+Cache::ObjectPtr CacheWithSecondaryAdapter::Value(Handle* handle) {
+  ObjectPtr v = target_->Value(handle);
+  // TODO with stacked secondaries: might fail in EvictionHandler
+  assert(v != kDummyObj);
+  return v;
+}
+
+void CacheWithSecondaryAdapter::StartAsyncLookupOnMySecondary(
+    AsyncLookupHandle& async_handle) {
+  assert(!async_handle.IsPending());
+  assert(async_handle.result_handle == nullptr);
+
+  std::unique_ptr<SecondaryCacheResultHandle> secondary_handle =
+      secondary_cache_->Lookup(async_handle.key, async_handle.helper,
+                               async_handle.create_context, /*wait*/ false,
+                               async_handle.found_dummy_entry,
+                               /*out*/ async_handle.kept_in_sec_cache);
+  if (secondary_handle) {
+    // TODO with stacked secondaries: Check & process if already ready?
+    async_handle.pending_handle = secondary_handle.release();
+    async_handle.pending_cache = secondary_cache_.get();
+  }
+}
+
+void CacheWithSecondaryAdapter::StartAsyncLookup(
+    AsyncLookupHandle& async_handle) {
+  target_->StartAsyncLookup(async_handle);
+  if (!async_handle.IsPending()) {
+    bool secondary_compatible =
+        async_handle.helper &&
+        async_handle.helper->IsSecondaryCacheCompatible();
+    async_handle.found_dummy_entry |= ProcessDummyResult(
+        &async_handle.result_handle, /*erase=*/secondary_compatible);
+
+    if (async_handle.Result() == nullptr && secondary_compatible) {
+      // Not found and not pending on another secondary cache
+      StartAsyncLookupOnMySecondary(async_handle);
+    }
+  }
+}
+
+void CacheWithSecondaryAdapter::WaitAll(AsyncLookupHandle* async_handles,
+                                        size_t count) {
+  if (count == 0) {
+    // Nothing to do
+    return;
+  }
+  // Requests that are pending on *my* secondary cache, at the start of this
+  // function
+  std::vector<AsyncLookupHandle*> my_pending;
+  // Requests that are pending on an "inner" secondary cache (managed somewhere
+  // under target_), as of the start of this function
+  std::vector<AsyncLookupHandle*> inner_pending;
+
+  // Initial accounting of pending handles, excluding those already handled
+  // by "outer" secondary caches. (See cur->pending_cache = nullptr.)
+  for (size_t i = 0; i < count; ++i) {
+    AsyncLookupHandle* cur = async_handles + i;
+    if (cur->pending_cache) {
+      assert(cur->IsPending());
+      assert(cur->helper);
+      assert(cur->helper->IsSecondaryCacheCompatible());
+      if (cur->pending_cache == secondary_cache_.get()) {
+        my_pending.push_back(cur);
+        // Mark as "to be handled by this caller"
+        cur->pending_cache = nullptr;
+      } else {
+        // Remember as potentially needing a lookup in my secondary
+        inner_pending.push_back(cur);
+      }
+    }
+  }
+
+  // Wait on inner-most cache lookups first
+  // TODO with stacked secondaries: because we are not using proper
+  // async/await constructs here yet, there is a false synchronization point
+  // here where all the results at one level are needed before initiating
+  // any lookups at the next level. Probably not a big deal, but worth noting.
+  if (!inner_pending.empty()) {
+    target_->WaitAll(async_handles, count);
+  }
+
+  // For those that failed to find something, convert to lookup in my
+  // secondary cache.
+  for (AsyncLookupHandle* cur : inner_pending) {
+    if (cur->Result() == nullptr) {
+      // Not found, try my secondary
+      StartAsyncLookupOnMySecondary(*cur);
+      if (cur->IsPending()) {
+        assert(cur->pending_cache == secondary_cache_.get());
+        my_pending.push_back(cur);
+        // Mark as "to be handled by this caller"
+        cur->pending_cache = nullptr;
+      }
+    }
+  }
+
+  // Wait on all lookups on my secondary cache
+  {
+    std::vector<SecondaryCacheResultHandle*> my_secondary_handles;
+    for (AsyncLookupHandle* cur : my_pending) {
+      my_secondary_handles.push_back(cur->pending_handle);
+    }
+    secondary_cache_->WaitAll(my_secondary_handles);
+  }
+
+  // Process results
+  for (AsyncLookupHandle* cur : my_pending) {
+    std::unique_ptr<SecondaryCacheResultHandle> secondary_handle(
+        cur->pending_handle);
+    cur->pending_handle = nullptr;
+    cur->result_handle = Promote(
+        std::move(secondary_handle), cur->key, cur->helper, cur->priority,
+        cur->stats, cur->found_dummy_entry, cur->kept_in_sec_cache);
+    assert(cur->pending_cache == nullptr);
+  }
+}
+
+std::string CacheWithSecondaryAdapter::GetPrintableOptions() const {
+  std::string str = target_->GetPrintableOptions();
+  str.append("  secondary_cache:\n");
+  str.append(secondary_cache_->GetPrintableOptions());
+  return str;
+}
+
+const char* CacheWithSecondaryAdapter::Name() const {
+  // To the user, at least for now, configure the underlying cache with
+  // a secondary cache. So we pretend to be that cache
+  return target_->Name();
+}
+}  // namespace ROCKSDB_NAMESPACE

cache/secondary_cache_adapter.h

@@ -0,0 +1,52 @@
+//  Copyright (c) Meta Platforms, Inc. and affiliates.
+//  This source code is licensed under both the GPLv2 (found in the
+//  COPYING file in the root directory) and Apache 2.0 License
+//  (found in the LICENSE.Apache file in the root directory).
+
+#pragma once
+
+#include "rocksdb/secondary_cache.h"
+
+namespace ROCKSDB_NAMESPACE {
+
+class CacheWithSecondaryAdapter : public CacheWrapper {
+ public:
+  explicit CacheWithSecondaryAdapter(
+      std::shared_ptr<Cache> target,
+      std::shared_ptr<SecondaryCache> secondary_cache);
+
+  ~CacheWithSecondaryAdapter() override;
+
+  Handle* Lookup(const Slice& key, const CacheItemHelper* helper,
+                 CreateContext* create_context,
+                 Priority priority = Priority::LOW,
+                 Statistics* stats = nullptr) override;
+
+  ObjectPtr Value(Handle* handle) override;
+
+  void StartAsyncLookup(AsyncLookupHandle& async_handle) override;
+
+  void WaitAll(AsyncLookupHandle* async_handles, size_t count) override;
+
+  std::string GetPrintableOptions() const override;
+
+  const char* Name() const override;
+
+ private:
+  bool EvictionHandler(const Slice& key, Handle* handle);
+
+  void StartAsyncLookupOnMySecondary(AsyncLookupHandle& async_handle);
+
+  Handle* Promote(
+      std::unique_ptr<SecondaryCacheResultHandle>&& secondary_handle,
+      const Slice& key, const CacheItemHelper* helper, Priority priority,
+      Statistics* stats, bool found_dummy_entry, bool kept_in_sec_cache);
+
+  bool ProcessDummyResult(Cache::Handle** handle, bool erase);
+
+  void CleanupCacheObject(ObjectPtr obj, const CacheItemHelper* helper);
+
+  std::shared_ptr<SecondaryCache> secondary_cache_;
+};
+
+}  // namespace ROCKSDB_NAMESPACE

cache/sharded_cache.h

@@ -15,7 +15,7 @@
 
 #include "port/lang.h"
 #include "port/port.h"
-#include "rocksdb/cache.h"
+#include "rocksdb/advanced_cache.h"
 #include "util/hash.h"
 #include "util/mutexlock.h"
 
@@ -49,21 +49,19 @@ class CacheShardBase {
     HashCref GetHash() const;
     ...
   };
-  Status Insert(const Slice& key, HashCref hash, void* value, size_t charge,
-                DeleterFn deleter, HandleImpl** handle,
-                Cache::Priority priority) = 0;
-  Status Insert(const Slice& key, HashCref hash, void* value,
+  Status Insert(const Slice& key, HashCref hash, Cache::ObjectPtr value,
                 const Cache::CacheItemHelper* helper, size_t charge,
-                HandleImpl** handle, Cache::Priority priority) = 0;
-  HandleImpl* Lookup(const Slice& key, HashCref hash) = 0;
+                HandleImpl** handle, Cache::Priority priority,
+                bool standalone) = 0;
+  Handle* CreateStandalone(const Slice& key, HashCref hash, ObjectPtr obj,
+                           const CacheItemHelper* helper,
+                           size_t charge, bool allow_uncharged) = 0;
   HandleImpl* Lookup(const Slice& key, HashCref hash,
                         const Cache::CacheItemHelper* helper,
-                        const Cache::CreateCallback& create_cb,
-                        Cache::Priority priority, bool wait,
+                        Cache::CreateContext* create_context,
+                        Cache::Priority priority,
                         Statistics* stats) = 0;
   bool Release(HandleImpl* handle, bool useful, bool erase_if_last_ref) = 0;
-  bool IsReady(HandleImpl* handle) = 0;
-  void Wait(HandleImpl* handle) = 0;
   bool Ref(HandleImpl* handle) = 0;
   void Erase(const Slice& key, HashCref hash) = 0;
   void SetCapacity(size_t capacity) = 0;
@@ -77,8 +75,9 @@ class CacheShardBase {
   // *state == 0 and implementation sets *state = SIZE_MAX to indicate
   // completion.
   void ApplyToSomeEntries(
-      const std::function<void(const Slice& key, void* value, size_t charge,
-                               DeleterFn deleter)>& callback,
+      const std::function<void(const Slice& key, ObjectPtr value,
+                               size_t charge,
+                               const Cache::CacheItemHelper* helper)>& callback,
       size_t average_entries_per_lock, size_t* state) = 0;
   void EraseUnRefEntries() = 0;
   */
@@ -172,36 +171,33 @@ class ShardedCache : public ShardedCacheBase {
         [s_c_l](CacheShard* cs) { cs->SetStrictCapacityLimit(s_c_l); });
   }
 
-  Status Insert(const Slice& key, void* value, size_t charge, DeleterFn deleter,
-                Handle** handle, Priority priority) override {
-    HashVal hash = CacheShard::ComputeHash(key);
-    auto h_out = reinterpret_cast<HandleImpl**>(handle);
-    return GetShard(hash).Insert(key, hash, value, charge, deleter, h_out,
-                                 priority);
-  }
-  Status Insert(const Slice& key, void* value, const CacheItemHelper* helper,
+  Status Insert(const Slice& key, ObjectPtr obj, const CacheItemHelper* helper,
                 size_t charge, Handle** handle = nullptr,
                 Priority priority = Priority::LOW) override {
-    if (!helper) {
-      return Status::InvalidArgument();
-    }
+    assert(helper);
     HashVal hash = CacheShard::ComputeHash(key);
     auto h_out = reinterpret_cast<HandleImpl**>(handle);
-    return GetShard(hash).Insert(key, hash, value, helper, charge, h_out,
+    return GetShard(hash).Insert(key, hash, obj, helper, charge, h_out,
                                  priority);
   }
 
-  Handle* Lookup(const Slice& key, Statistics* /*stats*/) override {
+  Handle* CreateStandalone(const Slice& key, ObjectPtr obj,
+                           const CacheItemHelper* helper, size_t charge,
+                           bool allow_uncharged) override {
+    assert(helper);
     HashVal hash = CacheShard::ComputeHash(key);
-    HandleImpl* result = GetShard(hash).Lookup(key, hash);
+    HandleImpl* result = GetShard(hash).CreateStandalone(
+        key, hash, obj, helper, charge, allow_uncharged);
     return reinterpret_cast<Handle*>(result);
   }
-  Handle* Lookup(const Slice& key, const CacheItemHelper* helper,
-                 const CreateCallback& create_cb, Priority priority, bool wait,
+
+  Handle* Lookup(const Slice& key, const CacheItemHelper* helper = nullptr,
+                 CreateContext* create_context = nullptr,
+                 Priority priority = Priority::LOW,
                  Statistics* stats = nullptr) override {
     HashVal hash = CacheShard::ComputeHash(key);
-    HandleImpl* result = GetShard(hash).Lookup(key, hash, helper, create_cb,
-                                               priority, wait, stats);
+    HandleImpl* result = GetShard(hash).Lookup(key, hash, helper,
+                                               create_context, priority, stats);
     return reinterpret_cast<Handle*>(result);
   }
 
@@ -215,14 +211,6 @@ class ShardedCache : public ShardedCacheBase {
     auto h = reinterpret_cast<HandleImpl*>(handle);
     return GetShard(h->GetHash()).Release(h, useful, erase_if_last_ref);
   }
-  bool IsReady(Handle* handle) override {
-    auto h = reinterpret_cast<HandleImpl*>(handle);
-    return GetShard(h->GetHash()).IsReady(h);
-  }
-  void Wait(Handle* handle) override {
-    auto h = reinterpret_cast<HandleImpl*>(handle);
-    GetShard(h->GetHash()).Wait(h);
-  }
   bool Ref(Handle* handle) override {
     auto h = reinterpret_cast<HandleImpl*>(handle);
     return GetShard(h->GetHash()).Ref(h);
@@ -238,14 +226,14 @@ class ShardedCache : public ShardedCacheBase {
     return SumOverShards2(&CacheShard::GetPinnedUsage);
   }
   size_t GetOccupancyCount() const override {
-    return SumOverShards2(&CacheShard::GetPinnedUsage);
+    return SumOverShards2(&CacheShard::GetOccupancyCount);
   }
   size_t GetTableAddressCount() const override {
     return SumOverShards2(&CacheShard::GetTableAddressCount);
   }
   void ApplyToAllEntries(
-      const std::function<void(const Slice& key, void* value, size_t charge,
-                               DeleterFn deleter)>& callback,
+      const std::function<void(const Slice& key, ObjectPtr value, size_t charge,
+                               const CacheItemHelper* helper)>& callback,
       const ApplyToAllEntriesOptions& opts) override {
     uint32_t num_shards = GetNumShards();
     // Iterate over part of each shard, rotating between shards, to

cache/typed_cache.h

@@ -0,0 +1,375 @@
+//  Copyright (c) Meta Platforms, Inc. and affiliates.
+//  This source code is licensed under both the GPLv2 (found in the
+//  COPYING file in the root directory) and Apache 2.0 License
+//  (found in the LICENSE.Apache file in the root directory).
+
+// APIs for accessing Cache in a type-safe and convenient way. Cache is kept
+// at a low, thin level of abstraction so that different implementations can
+// be plugged in, but these wrappers provide clean, convenient access to the
+// most common operations.
+//
+// A number of template classes are needed for sharing common structure. The
+// key classes are these:
+//
+// * PlaceholderCacheInterface - Used for making cache reservations, with
+// entries that have a charge but no value.
+// * BasicTypedCacheInterface<TValue> - Used for primary cache storage of
+// objects of type TValue.
+// * FullTypedCacheHelper<TValue, TCreateContext> - Used for secondary cache
+// compatible storage of objects of type TValue.
+// * For each of these, there's a "Shared" version
+// (e.g. FullTypedSharedCacheInterface) that holds a shared_ptr to the Cache,
+// rather than assuming external ownership by holding only a raw `Cache*`.
+
+#pragma once
+
+#include <algorithm>
+#include <cstdint>
+#include <memory>
+#include <type_traits>
+
+#include "cache/cache_helpers.h"
+#include "rocksdb/advanced_cache.h"
+#include "rocksdb/advanced_options.h"
+
+namespace ROCKSDB_NAMESPACE {
+
+// For future consideration:
+// * Pass in value to Insert with std::unique_ptr& to simplify ownership
+//   transfer logic in callers
+// * Make key type a template parameter (e.g. useful for table cache)
+// * Closer integration with CacheHandleGuard (opt-in, so not always
+//   paying the extra overhead)
+
+#define CACHE_TYPE_DEFS()                     \
+  using Priority = Cache::Priority;           \
+  using Handle = Cache::Handle;               \
+  using ObjectPtr = Cache::ObjectPtr;         \
+  using CreateContext = Cache::CreateContext; \
+  using CacheItemHelper = Cache::CacheItemHelper /* caller ; */
+
+template <typename CachePtr>
+class BaseCacheInterface {
+ public:
+  CACHE_TYPE_DEFS();
+
+  /*implicit*/ BaseCacheInterface(CachePtr cache) : cache_(std::move(cache)) {}
+
+  inline void Release(Handle* handle) { cache_->Release(handle); }
+
+  inline void ReleaseAndEraseIfLastRef(Handle* handle) {
+    cache_->Release(handle, /*erase_if_last_ref*/ true);
+  }
+
+  inline void RegisterReleaseAsCleanup(Handle* handle, Cleanable& cleanable) {
+    cleanable.RegisterCleanup(&ReleaseCacheHandleCleanup, get(), handle);
+  }
+
+  inline Cache* get() const { return &*cache_; }
+
+  explicit inline operator bool() const noexcept { return cache_ != nullptr; }
+
+ protected:
+  CachePtr cache_;
+};
+
+// PlaceholderCacheInterface - Used for making cache reservations, with
+// entries that have a charge but no value. CacheEntryRole is required as
+// a template parameter.
+template <CacheEntryRole kRole, typename CachePtr = Cache*>
+class PlaceholderCacheInterface : public BaseCacheInterface<CachePtr> {
+ public:
+  CACHE_TYPE_DEFS();
+  using BaseCacheInterface<CachePtr>::BaseCacheInterface;
+
+  inline Status Insert(const Slice& key, size_t charge, Handle** handle) {
+    return this->cache_->Insert(key, /*value=*/nullptr, GetHelper(), charge,
+                                handle);
+  }
+
+  static const Cache::CacheItemHelper* GetHelper() {
+    static const Cache::CacheItemHelper kHelper{kRole};
+    return &kHelper;
+  }
+};
+
+template <CacheEntryRole kRole>
+using PlaceholderSharedCacheInterface =
+    PlaceholderCacheInterface<kRole, std::shared_ptr<Cache>>;
+
+template <class TValue>
+class BasicTypedCacheHelperFns {
+ public:
+  CACHE_TYPE_DEFS();
+  // E.g. char* for char[]
+  using TValuePtr = std::remove_extent_t<TValue>*;
+
+ protected:
+  inline static ObjectPtr UpCastValue(TValuePtr value) { return value; }
+  inline static TValuePtr DownCastValue(ObjectPtr value) {
+    return static_cast<TValuePtr>(value);
+  }
+
+  static void Delete(ObjectPtr value, MemoryAllocator* allocator) {
+    // FIXME: Currently, no callers actually allocate the ObjectPtr objects
+    // using the custom allocator, just subobjects that keep a reference to
+    // the allocator themselves (with CacheAllocationPtr).
+    if (/*DISABLED*/ false && allocator) {
+      if constexpr (std::is_destructible_v<TValue>) {
+        DownCastValue(value)->~TValue();
+      }
+      allocator->Deallocate(value);
+    } else {
+      // Like delete but properly handles TValue=char[] etc.
+      std::default_delete<TValue>{}(DownCastValue(value));
+    }
+  }
+};
+
+// In its own class to try to minimize the number of distinct CacheItemHelper
+// instances (e.g. don't vary by CachePtr)
+template <class TValue, CacheEntryRole kRole>
+class BasicTypedCacheHelper : public BasicTypedCacheHelperFns<TValue> {
+ public:
+  static const Cache::CacheItemHelper* GetBasicHelper() {
+    static const Cache::CacheItemHelper kHelper{kRole,
+                                                &BasicTypedCacheHelper::Delete};
+    return &kHelper;
+  }
+};
+
+// BasicTypedCacheInterface - Used for primary cache storage of objects of
+// type TValue, which can be cleaned up with std::default_delete<TValue>. The
+// role is provided by TValue::kCacheEntryRole or given in an optional
+// template parameter.
+template <class TValue, CacheEntryRole kRole = TValue::kCacheEntryRole,
+          typename CachePtr = Cache*>
+class BasicTypedCacheInterface : public BaseCacheInterface<CachePtr>,
+                                 public BasicTypedCacheHelper<TValue, kRole> {
+ public:
+  CACHE_TYPE_DEFS();
+  using typename BasicTypedCacheHelperFns<TValue>::TValuePtr;
+  struct TypedHandle : public Handle {};
+  using BasicTypedCacheHelper<TValue, kRole>::GetBasicHelper;
+  // ctor
+  using BaseCacheInterface<CachePtr>::BaseCacheInterface;
+  struct TypedAsyncLookupHandle : public Cache::AsyncLookupHandle {
+    TypedHandle* Result() {
+      return reinterpret_cast<TypedHandle*>(Cache::AsyncLookupHandle::Result());
+    }
+  };
+
+  inline Status Insert(const Slice& key, TValuePtr value, size_t charge,
+                       TypedHandle** handle = nullptr,
+                       Priority priority = Priority::LOW) {
+    auto untyped_handle = reinterpret_cast<Handle**>(handle);
+    return this->cache_->Insert(
+        key, BasicTypedCacheHelperFns<TValue>::UpCastValue(value),
+        GetBasicHelper(), charge, untyped_handle, priority);
+  }
+
+  inline TypedHandle* Lookup(const Slice& key, Statistics* stats = nullptr) {
+    return reinterpret_cast<TypedHandle*>(
+        this->cache_->BasicLookup(key, stats));
+  }
+
+  inline void StartAsyncLookup(TypedAsyncLookupHandle& async_handle) {
+    assert(async_handle.helper == nullptr);
+    this->cache_->StartAsyncLookup(async_handle);
+  }
+
+  inline CacheHandleGuard<TValue> Guard(TypedHandle* handle) {
+    if (handle) {
+      return CacheHandleGuard<TValue>(&*this->cache_, handle);
+    } else {
+      return {};
+    }
+  }
+
+  inline std::shared_ptr<TValue> SharedGuard(TypedHandle* handle) {
+    if (handle) {
+      return MakeSharedCacheHandleGuard<TValue>(&*this->cache_, handle);
+    } else {
+      return {};
+    }
+  }
+
+  inline TValuePtr Value(TypedHandle* handle) {
+    return BasicTypedCacheHelperFns<TValue>::DownCastValue(
+        this->cache_->Value(handle));
+  }
+};
+
+// BasicTypedSharedCacheInterface - Like BasicTypedCacheInterface but with a
+// shared_ptr<Cache> for keeping Cache alive.
+template <class TValue, CacheEntryRole kRole = TValue::kCacheEntryRole>
+using BasicTypedSharedCacheInterface =
+    BasicTypedCacheInterface<TValue, kRole, std::shared_ptr<Cache>>;
+
+// TValue must implement ContentSlice() and ~TValue
+// TCreateContext must implement Create(std::unique_ptr<TValue>*, ...)
+template <class TValue, class TCreateContext>
+class FullTypedCacheHelperFns : public BasicTypedCacheHelperFns<TValue> {
+ public:
+  CACHE_TYPE_DEFS();
+
+ protected:
+  using typename BasicTypedCacheHelperFns<TValue>::TValuePtr;
+  using BasicTypedCacheHelperFns<TValue>::DownCastValue;
+  using BasicTypedCacheHelperFns<TValue>::UpCastValue;
+
+  static size_t Size(ObjectPtr v) {
+    TValuePtr value = DownCastValue(v);
+    auto slice = value->ContentSlice();
+    return slice.size();
+  }
+
+  static Status SaveTo(ObjectPtr v, size_t from_offset, size_t length,
+                       char* out) {
+    TValuePtr value = DownCastValue(v);
+    auto slice = value->ContentSlice();
+    assert(from_offset < slice.size());
+    assert(from_offset + length <= slice.size());
+    std::copy_n(slice.data() + from_offset, length, out);
+    return Status::OK();
+  }
+
+  static Status Create(const Slice& data, CreateContext* context,
+                       MemoryAllocator* allocator, ObjectPtr* out_obj,
+                       size_t* out_charge) {
+    std::unique_ptr<TValue> value = nullptr;
+    if constexpr (sizeof(TCreateContext) > 0) {
+      TCreateContext* tcontext = static_cast<TCreateContext*>(context);
+      tcontext->Create(&value, out_charge, data, allocator);
+    } else {
+      TCreateContext::Create(&value, out_charge, data, allocator);
+    }
+    *out_obj = UpCastValue(value.release());
+    return Status::OK();
+  }
+};
+
+// In its own class to try to minimize the number of distinct CacheItemHelper
+// instances (e.g. don't vary by CachePtr)
+template <class TValue, class TCreateContext, CacheEntryRole kRole>
+class FullTypedCacheHelper
+    : public FullTypedCacheHelperFns<TValue, TCreateContext> {
+ public:
+  static const Cache::CacheItemHelper* GetFullHelper() {
+    static const Cache::CacheItemHelper kHelper{
+        kRole,
+        &FullTypedCacheHelper::Delete,
+        &FullTypedCacheHelper::Size,
+        &FullTypedCacheHelper::SaveTo,
+        &FullTypedCacheHelper::Create,
+        BasicTypedCacheHelper<TValue, kRole>::GetBasicHelper()};
+    return &kHelper;
+  }
+};
+
+// FullTypedCacheHelper - Used for secondary cache compatible storage of
+// objects of type TValue. In addition to BasicTypedCacheInterface constraints,
+// we require TValue::ContentSlice() to return persistable data. This
+// simplifies usage for the normal case of simple secondary cache compatibility
+// (can give you a Slice to the data already in memory). In addition to
+// TCreateContext performing the role of Cache::CreateContext, it is also
+// expected to provide a function Create(std::unique_ptr<TValue>* value,
+// size_t* out_charge, const Slice& data, MemoryAllocator* allocator) for
+// creating new TValue.
+template <class TValue, class TCreateContext,
+          CacheEntryRole kRole = TValue::kCacheEntryRole,
+          typename CachePtr = Cache*>
+class FullTypedCacheInterface
+    : public BasicTypedCacheInterface<TValue, kRole, CachePtr>,
+      public FullTypedCacheHelper<TValue, TCreateContext, kRole> {
+ public:
+  CACHE_TYPE_DEFS();
+  using typename BasicTypedCacheInterface<TValue, kRole, CachePtr>::TypedHandle;
+  using typename BasicTypedCacheInterface<TValue, kRole,
+                                          CachePtr>::TypedAsyncLookupHandle;
+  using typename BasicTypedCacheHelperFns<TValue>::TValuePtr;
+  using BasicTypedCacheHelper<TValue, kRole>::GetBasicHelper;
+  using FullTypedCacheHelper<TValue, TCreateContext, kRole>::GetFullHelper;
+  using BasicTypedCacheHelperFns<TValue>::UpCastValue;
+  using BasicTypedCacheHelperFns<TValue>::DownCastValue;
+  // ctor
+  using BasicTypedCacheInterface<TValue, kRole,
+                                 CachePtr>::BasicTypedCacheInterface;
+
+  // Insert with SecondaryCache compatibility (subject to CacheTier).
+  // (Basic Insert() also inherited.)
+  inline Status InsertFull(
+      const Slice& key, TValuePtr value, size_t charge,
+      TypedHandle** handle = nullptr, Priority priority = Priority::LOW,
+      CacheTier lowest_used_cache_tier = CacheTier::kNonVolatileBlockTier) {
+    auto untyped_handle = reinterpret_cast<Handle**>(handle);
+    auto helper = lowest_used_cache_tier == CacheTier::kNonVolatileBlockTier
+                      ? GetFullHelper()
+                      : GetBasicHelper();
+    return this->cache_->Insert(key, UpCastValue(value), helper, charge,
+                                untyped_handle, priority);
+  }
+
+  // Like SecondaryCache::InsertSaved, with SecondaryCache compatibility
+  // (subject to CacheTier).
+  inline Status InsertSaved(
+      const Slice& key, const Slice& data, TCreateContext* create_context,
+      Priority priority = Priority::LOW,
+      CacheTier lowest_used_cache_tier = CacheTier::kNonVolatileBlockTier,
+      size_t* out_charge = nullptr) {
+    ObjectPtr value;
+    size_t charge;
+    Status st = GetFullHelper()->create_cb(data, create_context,
+                                           this->cache_->memory_allocator(),
+                                           &value, &charge);
+    if (out_charge) {
+      *out_charge = charge;
+    }
+    if (st.ok()) {
+      st = InsertFull(key, DownCastValue(value), charge, nullptr /*handle*/,
+                      priority, lowest_used_cache_tier);
+    } else {
+      GetFullHelper()->del_cb(value, this->cache_->memory_allocator());
+    }
+    return st;
+  }
+
+  // Lookup with SecondaryCache support (subject to CacheTier).
+  // (Basic Lookup() also inherited.)
+  inline TypedHandle* LookupFull(
+      const Slice& key, TCreateContext* create_context = nullptr,
+      Priority priority = Priority::LOW, Statistics* stats = nullptr,
+      CacheTier lowest_used_cache_tier = CacheTier::kNonVolatileBlockTier) {
+    if (lowest_used_cache_tier == CacheTier::kNonVolatileBlockTier) {
+      return reinterpret_cast<TypedHandle*>(this->cache_->Lookup(
+          key, GetFullHelper(), create_context, priority, stats));
+    } else {
+      return BasicTypedCacheInterface<TValue, kRole, CachePtr>::Lookup(key,
+                                                                       stats);
+    }
+  }
+
+  inline void StartAsyncLookupFull(
+      TypedAsyncLookupHandle& async_handle,
+      CacheTier lowest_used_cache_tier = CacheTier::kNonVolatileBlockTier) {
+    if (lowest_used_cache_tier == CacheTier::kNonVolatileBlockTier) {
+      async_handle.helper = GetFullHelper();
+      this->cache_->StartAsyncLookup(async_handle);
+    } else {
+      BasicTypedCacheInterface<TValue, kRole, CachePtr>::StartAsyncLookup(
+          async_handle);
+    }
+  }
+};
+
+// FullTypedSharedCacheInterface - Like FullTypedCacheInterface but with a
+// shared_ptr<Cache> for keeping Cache alive.
+template <class TValue, class TCreateContext,
+          CacheEntryRole kRole = TValue::kCacheEntryRole>
+using FullTypedSharedCacheInterface =
+    FullTypedCacheInterface<TValue, TCreateContext, kRole,
+                            std::shared_ptr<Cache>>;
+
+#undef CACHE_TYPE_DEFS
+
+}  // namespace ROCKSDB_NAMESPACE

coverage/coverage_test.sh

@@ -12,7 +12,7 @@ fi
 ROOT=".."
 # Fetch right version of gcov
 if [ -d /mnt/gvfs/third-party -a -z "$CXX" ]; then
-  source $ROOT/build_tools/fbcode_config_platform009.sh
+  source $ROOT/build_tools/fbcode_config_platform010.sh
   GCOV=$GCC_BASE/bin/gcov
 else
   GCOV=$(which gcov)

crash_test.mk

@@ -8,7 +8,7 @@ DB_STRESS_CMD?=./db_stress
 include common.mk
 
 CRASHTEST_MAKE=$(MAKE) -f crash_test.mk
-CRASHTEST_PY=$(PYTHON) -u tools/db_crashtest.py --stress_cmd=$(DB_STRESS_CMD)
+CRASHTEST_PY=$(PYTHON) -u tools/db_crashtest.py --stress_cmd=$(DB_STRESS_CMD) --cleanup_cmd='$(DB_CLEANUP_CMD)'
 
 .PHONY: crash_test crash_test_with_atomic_flush crash_test_with_txn \
 	crash_test_with_best_efforts_recovery crash_test_with_ts \

db/arena_wrapped_db_iter.cc

@@ -8,6 +8,7 @@
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 
 #include "db/arena_wrapped_db_iter.h"
+
 #include "memory/arena.h"
 #include "rocksdb/env.h"
 #include "rocksdb/iterator.h"
@@ -46,6 +47,9 @@ void ArenaWrappedDBIter::Init(
   read_options_ = read_options;
   allow_refresh_ = allow_refresh;
   memtable_range_tombstone_iter_ = nullptr;
+  if (!env->GetFileSystem()->use_async_io()) {
+    read_options_.async_io = false;
+  }
 }
 
 Status ArenaWrappedDBIter::Refresh() {

db/arena_wrapped_db_iter.h

@@ -9,7 +9,9 @@
 
 #pragma once
 #include <stdint.h>
+
 #include <string>
+
 #include "db/db_impl/db_impl.h"
 #include "db/db_iter.h"
 #include "db/range_del_aggregator.h"

db/blob/blob_contents.cc

@@ -13,12 +13,6 @@
 
 namespace ROCKSDB_NAMESPACE {
 
-std::unique_ptr<BlobContents> BlobContents::Create(
-    CacheAllocationPtr&& allocation, size_t size) {
-  return std::unique_ptr<BlobContents>(
-      new BlobContents(std::move(allocation), size));
-}
-
 size_t BlobContents::ApproximateMemoryUsage() const {
   size_t usage = 0;
 
@@ -45,46 +39,4 @@ size_t BlobContents::ApproximateMemoryUsage() const {
   return usage;
 }
 
-size_t BlobContents::SizeCallback(void* obj) {
-  assert(obj);
-
-  return static_cast<const BlobContents*>(obj)->size();
-}
-
-Status BlobContents::SaveToCallback(void* from_obj, size_t from_offset,
-                                    size_t length, void* out) {
-  assert(from_obj);
-
-  const BlobContents* buf = static_cast<const BlobContents*>(from_obj);
-  assert(buf->size() >= from_offset + length);
-
-  memcpy(out, buf->data().data() + from_offset, length);
-
-  return Status::OK();
-}
-
-Cache::CacheItemHelper* BlobContents::GetCacheItemHelper() {
-  static Cache::CacheItemHelper cache_helper(
-      &SizeCallback, &SaveToCallback,
-      GetCacheEntryDeleterForRole<BlobContents, CacheEntryRole::kBlobValue>());
-
-  return &cache_helper;
-}
-
-Status BlobContents::CreateCallback(CacheAllocationPtr&& allocation,
-                                    const void* buf, size_t size,
-                                    void** out_obj, size_t* charge) {
-  assert(allocation);
-
-  memcpy(allocation.get(), buf, size);
-
-  std::unique_ptr<BlobContents> obj = Create(std::move(allocation), size);
-  BlobContents* const contents = obj.release();
-
-  *out_obj = contents;
-  *charge = contents->ApproximateMemoryUsage();
-
-  return Status::OK();
-}
-
 }  // namespace ROCKSDB_NAMESPACE

db/blob/blob_contents.h

@@ -8,7 +8,7 @@
 #include <memory>
 
 #include "memory/memory_allocator.h"
-#include "rocksdb/cache.h"
+#include "rocksdb/advanced_cache.h"
 #include "rocksdb/rocksdb_namespace.h"
 #include "rocksdb/slice.h"
 #include "rocksdb/status.h"
@@ -18,8 +18,8 @@ namespace ROCKSDB_NAMESPACE {
 // A class representing a single uncompressed value read from a blob file.
 class BlobContents {
  public:
-  static std::unique_ptr<BlobContents> Create(CacheAllocationPtr&& allocation,
-                                              size_t size);
+  BlobContents(CacheAllocationPtr&& allocation, size_t size)
+      : allocation_(std::move(allocation)), data_(allocation_.get(), size) {}
 
   BlobContents(const BlobContents&) = delete;
   BlobContents& operator=(const BlobContents&) = delete;
@@ -34,23 +34,26 @@ class BlobContents {
 
   size_t ApproximateMemoryUsage() const;
 
-  // Callbacks for secondary cache
-  static size_t SizeCallback(void* obj);
-
-  static Status SaveToCallback(void* from_obj, size_t from_offset,
-                               size_t length, void* out);
-
-  static Cache::CacheItemHelper* GetCacheItemHelper();
-
-  static Status CreateCallback(CacheAllocationPtr&& allocation, const void* buf,
-                               size_t size, void** out_obj, size_t* charge);
+  // For TypedCacheInterface
+  const Slice& ContentSlice() const { return data_; }
+  static constexpr CacheEntryRole kCacheEntryRole = CacheEntryRole::kBlobValue;
 
  private:
-  BlobContents(CacheAllocationPtr&& allocation, size_t size)
-      : allocation_(std::move(allocation)), data_(allocation_.get(), size) {}
-
   CacheAllocationPtr allocation_;
   Slice data_;
 };
 
+class BlobContentsCreator : public Cache::CreateContext {
+ public:
+  static void Create(std::unique_ptr<BlobContents>* out, size_t* out_charge,
+                     const Slice& contents, MemoryAllocator* alloc) {
+    auto raw = new BlobContents(AllocateAndCopyBlock(contents, alloc),
+                                contents.size());
+    out->reset(raw);
+    if (out_charge) {
+      *out_charge = raw->ApproximateMemoryUsage();
+    }
+  }
+};
+
 }  // namespace ROCKSDB_NAMESPACE

db/blob/blob_counting_iterator.h

@@ -123,6 +123,10 @@ class BlobCountingIterator : public InternalIterator {
     return iter_->GetProperty(prop_name, prop);
   }
 
+  bool IsDeleteRangeSentinelKey() const override {
+    return iter_->IsDeleteRangeSentinelKey();
+  }
+
  private:
   void UpdateAndCountBlobIfNeeded() {
     assert(!iter_->Valid() || iter_->status().ok());

db/blob/blob_file_builder.cc

@@ -13,6 +13,7 @@
 #include "db/blob/blob_index.h"
 #include "db/blob/blob_log_format.h"
 #include "db/blob/blob_log_writer.h"
+#include "db/blob/blob_source.h"
 #include "db/event_helpers.h"
 #include "db/version_set.h"
 #include "file/filename.h"
@@ -258,6 +259,7 @@ Status BlobFileBuilder::CompressBlobIfNeeded(
     return Status::OK();
   }
 
+  // TODO: allow user CompressionOptions, including max_compressed_bytes_per_kb
   CompressionOptions opts;
   CompressionContext context(blob_compression_type_);
   constexpr uint64_t sample_for_compression = 0;
@@ -393,7 +395,7 @@ Status BlobFileBuilder::PutBlobIntoCacheIfNeeded(const Slice& blob,
                                                  uint64_t blob_offset) const {
   Status s = Status::OK();
 
-  auto blob_cache = immutable_options_->blob_cache;
+  BlobSource::SharedCacheInterface blob_cache{immutable_options_->blob_cache};
   auto statistics = immutable_options_->statistics.get();
   bool warm_cache =
       prepopulate_blob_cache_ == PrepopulateBlobCache::kFlushOnly &&
@@ -407,34 +409,12 @@ Status BlobFileBuilder::PutBlobIntoCacheIfNeeded(const Slice& blob,
 
     const Cache::Priority priority = Cache::Priority::BOTTOM;
 
-    // Objects to be put into the cache have to be heap-allocated and
-    // self-contained, i.e. own their contents. The Cache has to be able to
-    // take unique ownership of them.
-    CacheAllocationPtr allocation =
-        AllocateBlock(blob.size(), blob_cache->memory_allocator());
-    memcpy(allocation.get(), blob.data(), blob.size());
-    std::unique_ptr<BlobContents> buf =
-        BlobContents::Create(std::move(allocation), blob.size());
-
-    Cache::CacheItemHelper* const cache_item_helper =
-        BlobContents::GetCacheItemHelper();
-    assert(cache_item_helper);
-
-    if (immutable_options_->lowest_used_cache_tier ==
-        CacheTier::kNonVolatileBlockTier) {
-      s = blob_cache->Insert(key, buf.get(), cache_item_helper,
-                             buf->ApproximateMemoryUsage(),
-                             nullptr /* cache_handle */, priority);
-    } else {
-      s = blob_cache->Insert(key, buf.get(), buf->ApproximateMemoryUsage(),
-                             cache_item_helper->del_cb,
-                             nullptr /* cache_handle */, priority);
-    }
+    s = blob_cache.InsertSaved(key, blob, nullptr /*context*/, priority,
+                               immutable_options_->lowest_used_cache_tier);
 
     if (s.ok()) {
       RecordTick(statistics, BLOB_DB_CACHE_ADD);
-      RecordTick(statistics, BLOB_DB_CACHE_BYTES_WRITE, buf->size());
-      buf.release();
+      RecordTick(statistics, BLOB_DB_CACHE_BYTES_WRITE, blob.size());
     } else {
       RecordTick(statistics, BLOB_DB_CACHE_ADD_FAILURES);
     }

db/blob/blob_file_cache.cc

@@ -37,18 +37,18 @@ BlobFileCache::BlobFileCache(Cache* cache,
 }
 
 Status BlobFileCache::GetBlobFileReader(
-    uint64_t blob_file_number,
+    const ReadOptions& read_options, uint64_t blob_file_number,
     CacheHandleGuard<BlobFileReader>* blob_file_reader) {
   assert(blob_file_reader);
   assert(blob_file_reader->IsEmpty());
 
-  const Slice key = GetSlice(&blob_file_number);
+  const Slice key = GetSliceForKey(&blob_file_number);
 
   assert(cache_);
 
-  Cache::Handle* handle = cache_->Lookup(key);
+  TypedHandle* handle = cache_.Lookup(key);
   if (handle) {
-    *blob_file_reader = CacheHandleGuard<BlobFileReader>(cache_, handle);
+    *blob_file_reader = cache_.Guard(handle);
     return Status::OK();
   }
 
@@ -57,9 +57,9 @@ Status BlobFileCache::GetBlobFileReader(
   // Check again while holding mutex
   MutexLock lock(mutex_.get(key));
 
-  handle = cache_->Lookup(key);
+  handle = cache_.Lookup(key);
   if (handle) {
-    *blob_file_reader = CacheHandleGuard<BlobFileReader>(cache_, handle);
+    *blob_file_reader = cache_.Guard(handle);
     return Status::OK();
   }
 
@@ -73,7 +73,7 @@ Status BlobFileCache::GetBlobFileReader(
   {
     assert(file_options_);
     const Status s = BlobFileReader::Create(
-        *immutable_options_, *file_options_, column_family_id_,
+        *immutable_options_, read_options, *file_options_, column_family_id_,
         blob_file_read_hist_, blob_file_number, io_tracer_, &reader);
     if (!s.ok()) {
       RecordTick(statistics, NO_FILE_ERRORS);
@@ -84,8 +84,7 @@ Status BlobFileCache::GetBlobFileReader(
   {
     constexpr size_t charge = 1;
 
-    const Status s = cache_->Insert(key, reader.get(), charge,
-                                    &DeleteCacheEntry<BlobFileReader>, &handle);
+    const Status s = cache_.Insert(key, reader.get(), charge, &handle);
     if (!s.ok()) {
       RecordTick(statistics, NO_FILE_ERRORS);
       return s;
@@ -94,7 +93,7 @@ Status BlobFileCache::GetBlobFileReader(
 
   reader.release();
 
-  *blob_file_reader = CacheHandleGuard<BlobFileReader>(cache_, handle);
+  *blob_file_reader = cache_.Guard(handle);
 
   return Status::OK();
 }

db/blob/blob_file_cache.h

@@ -7,7 +7,8 @@
 
 #include <cinttypes>
 
-#include "cache/cache_helpers.h"
+#include "cache/typed_cache.h"
+#include "db/blob/blob_file_reader.h"
 #include "rocksdb/rocksdb_namespace.h"
 #include "util/mutexlock.h"
 
@@ -18,7 +19,6 @@ struct ImmutableOptions;
 struct FileOptions;
 class HistogramImpl;
 class Status;
-class BlobFileReader;
 class Slice;
 class IOTracer;
 
@@ -32,11 +32,15 @@ class BlobFileCache {
   BlobFileCache(const BlobFileCache&) = delete;
   BlobFileCache& operator=(const BlobFileCache&) = delete;
 
-  Status GetBlobFileReader(uint64_t blob_file_number,
+  Status GetBlobFileReader(const ReadOptions& read_options,
+                           uint64_t blob_file_number,
                            CacheHandleGuard<BlobFileReader>* blob_file_reader);
 
  private:
-  Cache* cache_;
+  using CacheInterface =
+      BasicTypedCacheInterface<BlobFileReader, CacheEntryRole::kMisc>;
+  using TypedHandle = CacheInterface::TypedHandle;
+  CacheInterface cache_;
   // Note: mutex_ below is used to guard against multiple threads racing to open
   // the same file.
   Striped<port::Mutex, Slice> mutex_;

db/blob/blob_file_cache_test.cc

@@ -118,7 +118,9 @@ TEST_F(BlobFileCacheTest, GetBlobFileReader) {
   // First try: reader should be opened and put in cache
   CacheHandleGuard<BlobFileReader> first;
 
-  ASSERT_OK(blob_file_cache.GetBlobFileReader(blob_file_number, &first));
+  const ReadOptions read_options;
+  ASSERT_OK(blob_file_cache.GetBlobFileReader(read_options, blob_file_number,
+                                              &first));
   ASSERT_NE(first.GetValue(), nullptr);
   ASSERT_EQ(options.statistics->getTickerCount(NO_FILE_OPENS), 1);
   ASSERT_EQ(options.statistics->getTickerCount(NO_FILE_ERRORS), 0);
@@ -126,7 +128,8 @@ TEST_F(BlobFileCacheTest, GetBlobFileReader) {
   // Second try: reader should be served from cache
   CacheHandleGuard<BlobFileReader> second;
 
-  ASSERT_OK(blob_file_cache.GetBlobFileReader(blob_file_number, &second));
+  ASSERT_OK(blob_file_cache.GetBlobFileReader(read_options, blob_file_number,
+                                              &second));
   ASSERT_NE(second.GetValue(), nullptr);
   ASSERT_EQ(options.statistics->getTickerCount(NO_FILE_OPENS), 1);
   ASSERT_EQ(options.statistics->getTickerCount(NO_FILE_ERRORS), 0);
@@ -163,19 +166,21 @@ TEST_F(BlobFileCacheTest, GetBlobFileReader_Race) {
   CacheHandleGuard<BlobFileReader> first;
   CacheHandleGuard<BlobFileReader> second;
 
+  const ReadOptions read_options;
   SyncPoint::GetInstance()->SetCallBack(
       "BlobFileCache::GetBlobFileReader:DoubleCheck", [&](void* /* arg */) {
         // Disabling sync points to prevent infinite recursion
         SyncPoint::GetInstance()->DisableProcessing();
-
-        ASSERT_OK(blob_file_cache.GetBlobFileReader(blob_file_number, &second));
+        ASSERT_OK(blob_file_cache.GetBlobFileReader(read_options,
+                                                    blob_file_number, &second));
         ASSERT_NE(second.GetValue(), nullptr);
         ASSERT_EQ(options.statistics->getTickerCount(NO_FILE_OPENS), 1);
         ASSERT_EQ(options.statistics->getTickerCount(NO_FILE_ERRORS), 0);
       });
   SyncPoint::GetInstance()->EnableProcessing();
 
-  ASSERT_OK(blob_file_cache.GetBlobFileReader(blob_file_number, &first));
+  ASSERT_OK(blob_file_cache.GetBlobFileReader(read_options, blob_file_number,
+                                              &first));
   ASSERT_NE(first.GetValue(), nullptr);
   ASSERT_EQ(options.statistics->getTickerCount(NO_FILE_OPENS), 1);
   ASSERT_EQ(options.statistics->getTickerCount(NO_FILE_ERRORS), 0);
@@ -213,8 +218,10 @@ TEST_F(BlobFileCacheTest, GetBlobFileReader_IOError) {
 
   CacheHandleGuard<BlobFileReader> reader;
 
+  const ReadOptions read_options;
   ASSERT_TRUE(
-      blob_file_cache.GetBlobFileReader(blob_file_number, &reader).IsIOError());
+      blob_file_cache.GetBlobFileReader(read_options, blob_file_number, &reader)
+          .IsIOError());
   ASSERT_EQ(reader.GetValue(), nullptr);
   ASSERT_EQ(options.statistics->getTickerCount(NO_FILE_OPENS), 1);
   ASSERT_EQ(options.statistics->getTickerCount(NO_FILE_ERRORS), 1);
@@ -253,8 +260,10 @@ TEST_F(BlobFileCacheTest, GetBlobFileReader_CacheFull) {
   // strict_capacity_limit is set
   CacheHandleGuard<BlobFileReader> reader;
 
-  ASSERT_TRUE(blob_file_cache.GetBlobFileReader(blob_file_number, &reader)
-                  .IsMemoryLimit());
+  const ReadOptions read_options;
+  ASSERT_TRUE(
+      blob_file_cache.GetBlobFileReader(read_options, blob_file_number, &reader)
+          .IsMemoryLimit());
   ASSERT_EQ(reader.GetValue(), nullptr);
   ASSERT_EQ(options.statistics->getTickerCount(NO_FILE_OPENS), 1);
   ASSERT_EQ(options.statistics->getTickerCount(NO_FILE_ERRORS), 1);

db/blob/blob_file_completion_callback.h

@@ -23,32 +23,19 @@ class BlobFileCompletionCallback {
       const std::vector<std::shared_ptr<EventListener>>& listeners,
       const std::string& dbname)
       : event_logger_(event_logger), listeners_(listeners), dbname_(dbname) {
-#ifndef ROCKSDB_LITE
     sst_file_manager_ = sst_file_manager;
     mutex_ = mutex;
     error_handler_ = error_handler;
-#else
-    (void)sst_file_manager;
-    (void)mutex;
-    (void)error_handler;
-#endif  // ROCKSDB_LITE
   }
 
   void OnBlobFileCreationStarted(const std::string& file_name,
                                  const std::string& column_family_name,
                                  int job_id,
                                  BlobFileCreationReason creation_reason) {
-#ifndef ROCKSDB_LITE
     // Notify the listeners.
     EventHelpers::NotifyBlobFileCreationStarted(listeners_, dbname_,
                                                 column_family_name, file_name,
                                                 job_id, creation_reason);
-#else
-    (void)file_name;
-    (void)column_family_name;
-    (void)job_id;
-    (void)creation_reason;
-#endif
   }
 
   Status OnBlobFileCompleted(const std::string& file_name,
@@ -61,7 +48,6 @@ class BlobFileCompletionCallback {
                              uint64_t blob_count, uint64_t blob_bytes) {
     Status s;
 
-#ifndef ROCKSDB_LITE
     auto sfm = static_cast<SstFileManagerImpl*>(sst_file_manager_);
     if (sfm) {
       // Report new blob files to SstFileManagerImpl
@@ -74,7 +60,6 @@ class BlobFileCompletionCallback {
         error_handler_->SetBGError(s, BackgroundErrorReason::kFlush);
       }
     }
-#endif  // !ROCKSDB_LITE
 
     // Notify the listeners.
     EventHelpers::LogAndNotifyBlobFileCreationFinished(
@@ -89,11 +74,9 @@ class BlobFileCompletionCallback {
   }
 
  private:
-#ifndef ROCKSDB_LITE
   SstFileManager* sst_file_manager_;
   InstrumentedMutex* mutex_;
   ErrorHandler* error_handler_;
-#endif  // ROCKSDB_LITE
   EventLogger* event_logger_;
   std::vector<std::shared_ptr<EventListener>> listeners_;
   std::string dbname_;

db/blob/blob_file_reader.cc

@@ -26,9 +26,10 @@
 namespace ROCKSDB_NAMESPACE {
 
 Status BlobFileReader::Create(
-    const ImmutableOptions& immutable_options, const FileOptions& file_options,
-    uint32_t column_family_id, HistogramImpl* blob_file_read_hist,
-    uint64_t blob_file_number, const std::shared_ptr<IOTracer>& io_tracer,
+    const ImmutableOptions& immutable_options, const ReadOptions& read_options,
+    const FileOptions& file_options, uint32_t column_family_id,
+    HistogramImpl* blob_file_read_hist, uint64_t blob_file_number,
+    const std::shared_ptr<IOTracer>& io_tracer,
     std::unique_ptr<BlobFileReader>* blob_file_reader) {
   assert(blob_file_reader);
   assert(!*blob_file_reader);
@@ -52,15 +53,17 @@ Status BlobFileReader::Create(
   CompressionType compression_type = kNoCompression;
 
   {
-    const Status s = ReadHeader(file_reader.get(), column_family_id, statistics,
-                                &compression_type);
+    const Status s =
+        ReadHeader(file_reader.get(), read_options, column_family_id,
+                   statistics, &compression_type);
     if (!s.ok()) {
       return s;
     }
   }
 
   {
-    const Status s = ReadFooter(file_reader.get(), file_size, statistics);
+    const Status s =
+        ReadFooter(file_reader.get(), read_options, file_size, statistics);
     if (!s.ok()) {
       return s;
     }
@@ -134,6 +137,7 @@ Status BlobFileReader::OpenFile(
 }
 
 Status BlobFileReader::ReadHeader(const RandomAccessFileReader* file_reader,
+                                  const ReadOptions& read_options,
                                   uint32_t column_family_id,
                                   Statistics* statistics,
                                   CompressionType* compression_type) {
@@ -151,9 +155,10 @@ Status BlobFileReader::ReadHeader(const RandomAccessFileReader* file_reader,
     constexpr size_t read_size = BlobLogHeader::kSize;
 
     // TODO: rate limit reading headers from blob files.
-    const Status s = ReadFromFile(file_reader, read_offset, read_size,
-                                  statistics, &header_slice, &buf, &aligned_buf,
-                                  Env::IO_TOTAL /* rate_limiter_priority */);
+    const Status s =
+        ReadFromFile(file_reader, read_options, read_offset, read_size,
+                     statistics, &header_slice, &buf, &aligned_buf,
+                     Env::IO_TOTAL /* rate_limiter_priority */);
     if (!s.ok()) {
       return s;
     }
@@ -187,6 +192,7 @@ Status BlobFileReader::ReadHeader(const RandomAccessFileReader* file_reader,
 }
 
 Status BlobFileReader::ReadFooter(const RandomAccessFileReader* file_reader,
+                                  const ReadOptions& read_options,
                                   uint64_t file_size, Statistics* statistics) {
   assert(file_size >= BlobLogHeader::kSize + BlobLogFooter::kSize);
   assert(file_reader);
@@ -202,9 +208,10 @@ Status BlobFileReader::ReadFooter(const RandomAccessFileReader* file_reader,
     constexpr size_t read_size = BlobLogFooter::kSize;
 
     // TODO: rate limit reading footers from blob files.
-    const Status s = ReadFromFile(file_reader, read_offset, read_size,
-                                  statistics, &footer_slice, &buf, &aligned_buf,
-                                  Env::IO_TOTAL /* rate_limiter_priority */);
+    const Status s =
+        ReadFromFile(file_reader, read_options, read_offset, read_size,
+                     statistics, &footer_slice, &buf, &aligned_buf,
+                     Env::IO_TOTAL /* rate_limiter_priority */);
     if (!s.ok()) {
       return s;
     }
@@ -232,6 +239,7 @@ Status BlobFileReader::ReadFooter(const RandomAccessFileReader* file_reader,
 }
 
 Status BlobFileReader::ReadFromFile(const RandomAccessFileReader* file_reader,
+                                    const ReadOptions& read_options,
                                     uint64_t read_offset, size_t read_size,
                                     Statistics* statistics, Slice* slice,
                                     Buffer* buf, AlignedBuf* aligned_buf,
@@ -246,17 +254,23 @@ Status BlobFileReader::ReadFromFile(const RandomAccessFileReader* file_reader,
 
   Status s;
 
+  IOOptions io_options;
+  s = file_reader->PrepareIOOptions(read_options, io_options);
+  if (!s.ok()) {
+    return s;
+  }
+
   if (file_reader->use_direct_io()) {
     constexpr char* scratch = nullptr;
 
-    s = file_reader->Read(IOOptions(), read_offset, read_size, slice, scratch,
+    s = file_reader->Read(io_options, read_offset, read_size, slice, scratch,
                           aligned_buf, rate_limiter_priority);
   } else {
     buf->reset(new char[read_size]);
     constexpr AlignedBuf* aligned_scratch = nullptr;
 
-    s = file_reader->Read(IOOptions(), read_offset, read_size, slice,
-                          buf->get(), aligned_scratch, rate_limiter_priority);
+    s = file_reader->Read(io_options, read_offset, read_size, slice, buf->get(),
+                          aligned_scratch, rate_limiter_priority);
   }
 
   if (!s.ok()) {
@@ -324,8 +338,13 @@ Status BlobFileReader::GetBlob(
     Status s;
     constexpr bool for_compaction = true;
 
+    IOOptions io_options;
+    s = file_reader_->PrepareIOOptions(read_options, io_options);
+    if (!s.ok()) {
+      return s;
+    }
     prefetched = prefetch_buffer->TryReadFromCache(
-        IOOptions(), file_reader_.get(), record_offset,
+        io_options, file_reader_.get(), record_offset,
         static_cast<size_t>(record_size), &record_slice, &s,
         read_options.rate_limiter_priority, for_compaction);
     if (!s.ok()) {
@@ -338,10 +357,10 @@ Status BlobFileReader::GetBlob(
     PERF_COUNTER_ADD(blob_read_count, 1);
     PERF_COUNTER_ADD(blob_read_byte, record_size);
     PERF_TIMER_GUARD(blob_read_time);
-    const Status s = ReadFromFile(file_reader_.get(), record_offset,
-                                  static_cast<size_t>(record_size), statistics_,
-                                  &record_slice, &buf, &aligned_buf,
-                                  read_options.rate_limiter_priority);
+    const Status s = ReadFromFile(
+        file_reader_.get(), read_options, record_offset,
+        static_cast<size_t>(record_size), statistics_, &record_slice, &buf,
+        &aligned_buf, read_options.rate_limiter_priority);
     if (!s.ok()) {
       return s;
     }
@@ -569,12 +588,7 @@ Status BlobFileReader::UncompressBlobIfNeeded(
   assert(result);
 
   if (compression_type == kNoCompression) {
-    CacheAllocationPtr allocation =
-        AllocateBlock(value_slice.size(), allocator);
-    memcpy(allocation.get(), value_slice.data(), value_slice.size());
-
-    *result = BlobContents::Create(std::move(allocation), value_slice.size());
-
+    BlobContentsCreator::Create(result, nullptr, value_slice, allocator);
     return Status::OK();
   }
 
@@ -602,7 +616,7 @@ Status BlobFileReader::UncompressBlobIfNeeded(
     return Status::Corruption("Unable to uncompress blob");
   }
 
-  *result = BlobContents::Create(std::move(output), uncompressed_size);
+  result->reset(new BlobContents(std::move(output), uncompressed_size));
 
   return Status::OK();
 }

db/blob/blob_file_reader.h

@@ -29,6 +29,7 @@ class Statistics;
 class BlobFileReader {
  public:
   static Status Create(const ImmutableOptions& immutable_options,
+                       const ReadOptions& read_options,
                        const FileOptions& file_options,
                        uint32_t column_family_id,
                        HistogramImpl* blob_file_read_hist,
@@ -74,15 +75,18 @@ class BlobFileReader {
                          std::unique_ptr<RandomAccessFileReader>* file_reader);
 
   static Status ReadHeader(const RandomAccessFileReader* file_reader,
+                           const ReadOptions& read_options,
                            uint32_t column_family_id, Statistics* statistics,
                            CompressionType* compression_type);
 
   static Status ReadFooter(const RandomAccessFileReader* file_reader,
-                           uint64_t file_size, Statistics* statistics);
+                           const ReadOptions& read_options, uint64_t file_size,
+                           Statistics* statistics);
 
   using Buffer = std::unique_ptr<char[]>;
 
   static Status ReadFromFile(const RandomAccessFileReader* file_reader,
+                             const ReadOptions& read_options,
                              uint64_t read_offset, size_t read_size,
                              Statistics* statistics, Slice* slice, Buffer* buf,
                              AlignedBuf* aligned_buf,

db/blob/blob_file_reader_test.cc

@@ -172,12 +172,12 @@ TEST_F(BlobFileReaderTest, CreateReaderAndGetBlob) {
 
   std::unique_ptr<BlobFileReader> reader;
 
+  ReadOptions read_options;
   ASSERT_OK(BlobFileReader::Create(
-      immutable_options, FileOptions(), column_family_id, blob_file_read_hist,
-      blob_file_number, nullptr /*IOTracer*/, &reader));
+      immutable_options, read_options, FileOptions(), column_family_id,
+      blob_file_read_hist, blob_file_number, nullptr /*IOTracer*/, &reader));
 
   // Make sure the blob can be retrieved with and without checksum verification
-  ReadOptions read_options;
   read_options.verify_checksums = false;
 
   constexpr FilePrefetchBuffer* prefetch_buffer = nullptr;
@@ -479,11 +479,11 @@ TEST_F(BlobFileReaderTest, Malformed) {
   constexpr HistogramImpl* blob_file_read_hist = nullptr;
 
   std::unique_ptr<BlobFileReader> reader;
-
-  ASSERT_TRUE(BlobFileReader::Create(immutable_options, FileOptions(),
-                                     column_family_id, blob_file_read_hist,
-                                     blob_file_number, nullptr /*IOTracer*/,
-                                     &reader)
+  const ReadOptions read_options;
+  ASSERT_TRUE(BlobFileReader::Create(immutable_options, read_options,
+                                     FileOptions(), column_family_id,
+                                     blob_file_read_hist, blob_file_number,
+                                     nullptr /*IOTracer*/, &reader)
                   .IsCorruption());
 }
 
@@ -513,11 +513,11 @@ TEST_F(BlobFileReaderTest, TTL) {
   constexpr HistogramImpl* blob_file_read_hist = nullptr;
 
   std::unique_ptr<BlobFileReader> reader;
-
-  ASSERT_TRUE(BlobFileReader::Create(immutable_options, FileOptions(),
-                                     column_family_id, blob_file_read_hist,
-                                     blob_file_number, nullptr /*IOTracer*/,
-                                     &reader)
+  const ReadOptions read_options;
+  ASSERT_TRUE(BlobFileReader::Create(immutable_options, read_options,
+                                     FileOptions(), column_family_id,
+                                     blob_file_read_hist, blob_file_number,
+                                     nullptr /*IOTracer*/, &reader)
                   .IsCorruption());
 }
 
@@ -552,11 +552,11 @@ TEST_F(BlobFileReaderTest, ExpirationRangeInHeader) {
   constexpr HistogramImpl* blob_file_read_hist = nullptr;
 
   std::unique_ptr<BlobFileReader> reader;
-
-  ASSERT_TRUE(BlobFileReader::Create(immutable_options, FileOptions(),
-                                     column_family_id, blob_file_read_hist,
-                                     blob_file_number, nullptr /*IOTracer*/,
-                                     &reader)
+  const ReadOptions read_options;
+  ASSERT_TRUE(BlobFileReader::Create(immutable_options, read_options,
+                                     FileOptions(), column_family_id,
+                                     blob_file_read_hist, blob_file_number,
+                                     nullptr /*IOTracer*/, &reader)
                   .IsCorruption());
 }
 
@@ -591,11 +591,11 @@ TEST_F(BlobFileReaderTest, ExpirationRangeInFooter) {
   constexpr HistogramImpl* blob_file_read_hist = nullptr;
 
   std::unique_ptr<BlobFileReader> reader;
-
-  ASSERT_TRUE(BlobFileReader::Create(immutable_options, FileOptions(),
-                                     column_family_id, blob_file_read_hist,
-                                     blob_file_number, nullptr /*IOTracer*/,
-                                     &reader)
+  const ReadOptions read_options;
+  ASSERT_TRUE(BlobFileReader::Create(immutable_options, read_options,
+                                     FileOptions(), column_family_id,
+                                     blob_file_read_hist, blob_file_number,
+                                     nullptr /*IOTracer*/, &reader)
                   .IsCorruption());
 }
 
@@ -629,9 +629,9 @@ TEST_F(BlobFileReaderTest, IncorrectColumnFamily) {
   std::unique_ptr<BlobFileReader> reader;
 
   constexpr uint32_t incorrect_column_family_id = 2;
-
-  ASSERT_TRUE(BlobFileReader::Create(immutable_options, FileOptions(),
-                                     incorrect_column_family_id,
+  const ReadOptions read_options;
+  ASSERT_TRUE(BlobFileReader::Create(immutable_options, read_options,
+                                     FileOptions(), incorrect_column_family_id,
                                      blob_file_read_hist, blob_file_number,
                                      nullptr /*IOTracer*/, &reader)
                   .IsCorruption());
@@ -664,10 +664,10 @@ TEST_F(BlobFileReaderTest, BlobCRCError) {
   constexpr HistogramImpl* blob_file_read_hist = nullptr;
 
   std::unique_ptr<BlobFileReader> reader;
-
+  const ReadOptions read_options;
   ASSERT_OK(BlobFileReader::Create(
-      immutable_options, FileOptions(), column_family_id, blob_file_read_hist,
-      blob_file_number, nullptr /*IOTracer*/, &reader));
+      immutable_options, read_options, FileOptions(), column_family_id,
+      blob_file_read_hist, blob_file_number, nullptr /*IOTracer*/, &reader));
 
   SyncPoint::GetInstance()->SetCallBack(
       "BlobFileReader::VerifyBlob:CheckBlobCRC", [](void* arg) {
@@ -728,13 +728,12 @@ TEST_F(BlobFileReaderTest, Compression) {
   constexpr HistogramImpl* blob_file_read_hist = nullptr;
 
   std::unique_ptr<BlobFileReader> reader;
-
+  ReadOptions read_options;
   ASSERT_OK(BlobFileReader::Create(
-      immutable_options, FileOptions(), column_family_id, blob_file_read_hist,
-      blob_file_number, nullptr /*IOTracer*/, &reader));
+      immutable_options, read_options, FileOptions(), column_family_id,
+      blob_file_read_hist, blob_file_number, nullptr /*IOTracer*/, &reader));
 
   // Make sure the blob can be retrieved with and without checksum verification
-  ReadOptions read_options;
   read_options.verify_checksums = false;
 
   constexpr FilePrefetchBuffer* prefetch_buffer = nullptr;
@@ -803,10 +802,10 @@ TEST_F(BlobFileReaderTest, UncompressionError) {
   constexpr HistogramImpl* blob_file_read_hist = nullptr;
 
   std::unique_ptr<BlobFileReader> reader;
-
+  const ReadOptions read_options;
   ASSERT_OK(BlobFileReader::Create(
-      immutable_options, FileOptions(), column_family_id, blob_file_read_hist,
-      blob_file_number, nullptr /*IOTracer*/, &reader));
+      immutable_options, read_options, FileOptions(), column_family_id,
+      blob_file_read_hist, blob_file_number, nullptr /*IOTracer*/, &reader));
 
   SyncPoint::GetInstance()->SetCallBack(
       "BlobFileReader::UncompressBlobIfNeeded:TamperWithResult", [](void* arg) {
@@ -895,10 +894,10 @@ TEST_P(BlobFileReaderIOErrorTest, IOError) {
   constexpr HistogramImpl* blob_file_read_hist = nullptr;
 
   std::unique_ptr<BlobFileReader> reader;
-
+  const ReadOptions read_options;
   const Status s = BlobFileReader::Create(
-      immutable_options, FileOptions(), column_family_id, blob_file_read_hist,
-      blob_file_number, nullptr /*IOTracer*/, &reader);
+      immutable_options, read_options, FileOptions(), column_family_id,
+      blob_file_read_hist, blob_file_number, nullptr /*IOTracer*/, &reader);
 
   const bool fail_during_create =
       (sync_point_ != "BlobFileReader::GetBlob:ReadFromFile");
@@ -983,10 +982,10 @@ TEST_P(BlobFileReaderDecodingErrorTest, DecodingError) {
   constexpr HistogramImpl* blob_file_read_hist = nullptr;
 
   std::unique_ptr<BlobFileReader> reader;
-
+  const ReadOptions read_options;
   const Status s = BlobFileReader::Create(
-      immutable_options, FileOptions(), column_family_id, blob_file_read_hist,
-      blob_file_number, nullptr /*IOTracer*/, &reader);
+      immutable_options, read_options, FileOptions(), column_family_id,
+      blob_file_read_hist, blob_file_number, nullptr /*IOTracer*/, &reader);
 
   const bool fail_during_create =
       sync_point_ != "BlobFileReader::GetBlob:TamperWithResult";

db/blob/blob_log_format.h

@@ -22,6 +22,8 @@ constexpr uint32_t kVersion1 = 1;
 
 using ExpirationRange = std::pair<uint64_t, uint64_t>;
 
+// clang-format off
+
 // Format of blob log file header (30 bytes):
 //
 //    +--------------+---------+---------+-------+-------------+-------------------+
@@ -35,6 +37,9 @@ using ExpirationRange = std::pair<uint64_t, uint64_t>;
 //
 // Expiration range in the header is a rough range based on
 // blob_db_options.ttl_range_secs.
+
+// clang-format on
+
 struct BlobLogHeader {
   static constexpr size_t kSize = 30;
 
@@ -57,6 +62,8 @@ struct BlobLogHeader {
   Status DecodeFrom(Slice slice);
 };
 
+// clang-format off
+
 // Format of blob log file footer (32 bytes):
 //
 //    +--------------+------------+-------------------+------------+
@@ -69,6 +76,9 @@ struct BlobLogHeader {
 //
 // Unlike the same field in file header, expiration range in the footer is the
 // range of smallest and largest expiration of the data in this file.
+
+// clang-format on
+
 struct BlobLogFooter {
   static constexpr size_t kSize = 32;
 
@@ -81,6 +91,8 @@ struct BlobLogFooter {
   Status DecodeFrom(Slice slice);
 };
 
+// clang-format off
+
 // Blob record format (32 bytes header + key + value):
 //
 //    +------------+--------------+------------+------------+----------+---------+-----------+
@@ -100,6 +112,9 @@ struct BlobLogFooter {
 //
 // We could use variable length encoding (Varint64) to save more space, but it
 // make reader more complicated.
+
+// clang-format on
+
 struct BlobLogRecord {
   // header include fields up to blob CRC
   static constexpr size_t kHeaderSize = 32;

db/blob/blob_source.cc

@@ -30,16 +30,14 @@ BlobSource::BlobSource(const ImmutableOptions* immutable_options,
       blob_file_cache_(blob_file_cache),
       blob_cache_(immutable_options->blob_cache),
       lowest_used_cache_tier_(immutable_options->lowest_used_cache_tier) {
-#ifndef ROCKSDB_LITE
   auto bbto =
       immutable_options->table_factory->GetOptions<BlockBasedTableOptions>();
   if (bbto &&
       bbto->cache_usage_options.options_overrides.at(CacheEntryRole::kBlobCache)
               .charged == CacheEntryRoleOptions::Decision::kEnabled) {
-    blob_cache_ = std::make_shared<ChargedCache>(immutable_options->blob_cache,
-                                                 bbto->block_cache);
+    blob_cache_ = SharedCacheInterface{std::make_shared<ChargedCache>(
+        immutable_options->blob_cache, bbto->block_cache)};
   }
-#endif  // ROCKSDB_LITE
 }
 
 BlobSource::~BlobSource() = default;
@@ -82,9 +80,8 @@ Status BlobSource::PutBlobIntoCache(
   assert(cached_blob);
   assert(cached_blob->IsEmpty());
 
-  Cache::Handle* cache_handle = nullptr;
+  TypedHandle* cache_handle = nullptr;
   const Status s = InsertEntryIntoCache(cache_key, blob->get(),
-                                        (*blob)->ApproximateMemoryUsage(),
                                         &cache_handle, Cache::Priority::BOTTOM);
   if (s.ok()) {
     blob->release();
@@ -106,26 +103,10 @@ Status BlobSource::PutBlobIntoCache(
   return s;
 }
 
-Cache::Handle* BlobSource::GetEntryFromCache(const Slice& key) const {
-  Cache::Handle* cache_handle = nullptr;
-
-  if (lowest_used_cache_tier_ == CacheTier::kNonVolatileBlockTier) {
-    Cache::CreateCallback create_cb =
-        [allocator = blob_cache_->memory_allocator()](
-            const void* buf, size_t size, void** out_obj,
-            size_t* charge) -> Status {
-      return BlobContents::CreateCallback(AllocateBlock(size, allocator), buf,
-                                          size, out_obj, charge);
-    };
-
-    cache_handle = blob_cache_->Lookup(key, BlobContents::GetCacheItemHelper(),
-                                       create_cb, Cache::Priority::BOTTOM,
-                                       true /* wait_for_cache */, statistics_);
-  } else {
-    cache_handle = blob_cache_->Lookup(key, statistics_);
-  }
-
-  return cache_handle;
+BlobSource::TypedHandle* BlobSource::GetEntryFromCache(const Slice& key) const {
+  return blob_cache_.LookupFull(key, nullptr /* context */,
+                                Cache::Priority::BOTTOM, statistics_,
+                                lowest_used_cache_tier_);
 }
 
 void BlobSource::PinCachedBlob(CacheHandleGuard<BlobContents>* cached_blob,
@@ -166,24 +147,11 @@ void BlobSource::PinOwnedBlob(std::unique_ptr<BlobContents>* owned_blob,
 }
 
 Status BlobSource::InsertEntryIntoCache(const Slice& key, BlobContents* value,
-                                        size_t charge,
-                                        Cache::Handle** cache_handle,
+                                        TypedHandle** cache_handle,
                                         Cache::Priority priority) const {
-  Status s;
-
-  Cache::CacheItemHelper* const cache_item_helper =
-      BlobContents::GetCacheItemHelper();
-  assert(cache_item_helper);
-
-  if (lowest_used_cache_tier_ == CacheTier::kNonVolatileBlockTier) {
-    s = blob_cache_->Insert(key, value, cache_item_helper, charge, cache_handle,
-                            priority);
-  } else {
-    s = blob_cache_->Insert(key, value, charge, cache_item_helper->del_cb,
-                            cache_handle, priority);
-  }
-
-  return s;
+  return blob_cache_.InsertFull(key, value, value->ApproximateMemoryUsage(),
+                                cache_handle, priority,
+                                lowest_used_cache_tier_);
 }
 
 Status BlobSource::GetBlob(const ReadOptions& read_options,
@@ -241,7 +209,8 @@ Status BlobSource::GetBlob(const ReadOptions& read_options,
 
   {
     CacheHandleGuard<BlobFileReader> blob_file_reader;
-    s = blob_file_cache_->GetBlobFileReader(file_number, &blob_file_reader);
+    s = blob_file_cache_->GetBlobFileReader(read_options, file_number,
+                                            &blob_file_reader);
     if (!s.ok()) {
       return s;
     }
@@ -252,9 +221,10 @@ Status BlobSource::GetBlob(const ReadOptions& read_options,
       return Status::Corruption("Compression type mismatch when reading blob");
     }
 
-    MemoryAllocator* const allocator = (blob_cache_ && read_options.fill_cache)
-                                           ? blob_cache_->memory_allocator()
-                                           : nullptr;
+    MemoryAllocator* const allocator =
+        (blob_cache_ && read_options.fill_cache)
+            ? blob_cache_.get()->memory_allocator()
+            : nullptr;
 
     uint64_t read_size = 0;
     s = blob_file_reader.GetValue()->GetBlob(
@@ -403,8 +373,8 @@ void BlobSource::MultiGetBlobFromOneFile(const ReadOptions& read_options,
     }
 
     CacheHandleGuard<BlobFileReader> blob_file_reader;
-    Status s =
-        blob_file_cache_->GetBlobFileReader(file_number, &blob_file_reader);
+    Status s = blob_file_cache_->GetBlobFileReader(read_options, file_number,
+                                                   &blob_file_reader);
     if (!s.ok()) {
       for (size_t i = 0; i < _blob_reqs.size(); ++i) {
         BlobReadRequest* const req = _blob_reqs[i].first;
@@ -418,9 +388,10 @@ void BlobSource::MultiGetBlobFromOneFile(const ReadOptions& read_options,
 
     assert(blob_file_reader.GetValue());
 
-    MemoryAllocator* const allocator = (blob_cache_ && read_options.fill_cache)
-                                           ? blob_cache_->memory_allocator()
-                                           : nullptr;
+    MemoryAllocator* const allocator =
+        (blob_cache_ && read_options.fill_cache)
+            ? blob_cache_.get()->memory_allocator()
+            : nullptr;
 
     blob_file_reader.GetValue()->MultiGetBlob(read_options, allocator,
                                               _blob_reqs, &_bytes_read);

db/blob/blob_source.h

@@ -8,8 +8,9 @@
 #include <cinttypes>
 #include <memory>
 
-#include "cache/cache_helpers.h"
 #include "cache/cache_key.h"
+#include "cache/typed_cache.h"
+#include "db/blob/blob_contents.h"
 #include "db/blob/blob_file_cache.h"
 #include "db/blob/blob_read_request.h"
 #include "rocksdb/cache.h"
@@ -23,7 +24,6 @@ struct ImmutableOptions;
 class Status;
 class FilePrefetchBuffer;
 class Slice;
-class BlobContents;
 
 // BlobSource is a class that provides universal access to blobs, regardless of
 // whether they are in the blob cache, secondary cache, or (remote) storage.
@@ -95,9 +95,9 @@ class BlobSource {
                                uint64_t* bytes_read);
 
   inline Status GetBlobFileReader(
-      uint64_t blob_file_number,
+      const ReadOptions& read_options, uint64_t blob_file_number,
       CacheHandleGuard<BlobFileReader>* blob_file_reader) {
-    return blob_file_cache_->GetBlobFileReader(blob_file_number,
+    return blob_file_cache_->GetBlobFileReader(read_options, blob_file_number,
                                                blob_file_reader);
   }
 
@@ -106,6 +106,14 @@ class BlobSource {
   bool TEST_BlobInCache(uint64_t file_number, uint64_t file_size,
                         uint64_t offset, size_t* charge = nullptr) const;
 
+  // For TypedSharedCacheInterface
+  void Create(BlobContents** out, const char* buf, size_t size,
+              MemoryAllocator* alloc);
+
+  using SharedCacheInterface =
+      FullTypedSharedCacheInterface<BlobContents, BlobContentsCreator>;
+  using TypedHandle = SharedCacheInterface::TypedHandle;
+
  private:
   Status GetBlobFromCache(const Slice& cache_key,
                           CacheHandleGuard<BlobContents>* cached_blob) const;
@@ -120,10 +128,10 @@ class BlobSource {
   static void PinOwnedBlob(std::unique_ptr<BlobContents>* owned_blob,
                            PinnableSlice* value);
 
-  Cache::Handle* GetEntryFromCache(const Slice& key) const;
+  TypedHandle* GetEntryFromCache(const Slice& key) const;
 
   Status InsertEntryIntoCache(const Slice& key, BlobContents* value,
-                              size_t charge, Cache::Handle** cache_handle,
+                              TypedHandle** cache_handle,
                               Cache::Priority priority) const;
 
   inline CacheKey GetCacheKey(uint64_t file_number, uint64_t /*file_size*/,
@@ -141,7 +149,7 @@ class BlobSource {
   BlobFileCache* blob_file_cache_;
 
   // A cache to store uncompressed blobs.
-  std::shared_ptr<Cache> blob_cache_;
+  mutable SharedCacheInterface blob_cache_;
 
   // The control option of how the cache tiers will be used. Currently rocksdb
   // support block/blob cache (volatile tier) and secondary cache (this tier

db/blob/blob_source_test.cc

@@ -517,7 +517,8 @@ TEST_F(BlobSourceTest, GetCompressedBlobs) {
                   compression, blob_offsets, blob_sizes);
 
     CacheHandleGuard<BlobFileReader> blob_file_reader;
-    ASSERT_OK(blob_source.GetBlobFileReader(file_number, &blob_file_reader));
+    ASSERT_OK(blob_source.GetBlobFileReader(read_options, file_number,
+                                            &blob_file_reader));
     ASSERT_NE(blob_file_reader.GetValue(), nullptr);
 
     const uint64_t file_size = blob_file_reader.GetValue()->GetFileSize();
@@ -1139,26 +1140,18 @@ TEST_F(BlobSecondaryCacheTest, GetBlobsFromSecondaryCache) {
                          blob_file_cache.get());
 
   CacheHandleGuard<BlobFileReader> file_reader;
-  ASSERT_OK(blob_source.GetBlobFileReader(file_number, &file_reader));
+  ReadOptions read_options;
+  ASSERT_OK(
+      blob_source.GetBlobFileReader(read_options, file_number, &file_reader));
   ASSERT_NE(file_reader.GetValue(), nullptr);
   const uint64_t file_size = file_reader.GetValue()->GetFileSize();
   ASSERT_EQ(file_reader.GetValue()->GetCompressionType(), kNoCompression);
 
-  ReadOptions read_options;
   read_options.verify_checksums = true;
 
   auto blob_cache = options_.blob_cache;
   auto secondary_cache = lru_cache_opts_.secondary_cache;
 
-  Cache::CreateCallback create_cb = [](const void* buf, size_t size,
-                                       void** out_obj,
-                                       size_t* charge) -> Status {
-    CacheAllocationPtr allocation(new char[size]);
-
-    return BlobContents::CreateCallback(std::move(allocation), buf, size,
-                                        out_obj, charge);
-  };
-
   {
     // GetBlob
     std::vector<PinnableSlice> values(keys.size());
@@ -1219,15 +1212,16 @@ TEST_F(BlobSecondaryCacheTest, GetBlobsFromSecondaryCache) {
     {
       CacheKey cache_key = base_cache_key.WithOffset(blob_offsets[0]);
       const Slice key0 = cache_key.AsSlice();
-      auto handle0 = blob_cache->Lookup(key0, statistics);
+      auto handle0 = blob_cache->BasicLookup(key0, statistics);
       ASSERT_EQ(handle0, nullptr);
 
       // key0's item should be in the secondary cache.
-      bool is_in_sec_cache = false;
-      auto sec_handle0 =
-          secondary_cache->Lookup(key0, create_cb, true,
-                                  /*advise_erase=*/true, is_in_sec_cache);
-      ASSERT_FALSE(is_in_sec_cache);
+      bool kept_in_sec_cache = false;
+      auto sec_handle0 = secondary_cache->Lookup(
+          key0, BlobSource::SharedCacheInterface::GetFullHelper(),
+          /*context*/ nullptr, true,
+          /*advise_erase=*/true, kept_in_sec_cache);
+      ASSERT_FALSE(kept_in_sec_cache);
       ASSERT_NE(sec_handle0, nullptr);
       ASSERT_TRUE(sec_handle0->IsReady());
       auto value = static_cast<BlobContents*>(sec_handle0->Value());
@@ -1246,15 +1240,16 @@ TEST_F(BlobSecondaryCacheTest, GetBlobsFromSecondaryCache) {
     {
       CacheKey cache_key = base_cache_key.WithOffset(blob_offsets[1]);
       const Slice key1 = cache_key.AsSlice();
-      auto handle1 = blob_cache->Lookup(key1, statistics);
+      auto handle1 = blob_cache->BasicLookup(key1, statistics);
       ASSERT_NE(handle1, nullptr);
       blob_cache->Release(handle1);
 
-      bool is_in_sec_cache = false;
-      auto sec_handle1 =
-          secondary_cache->Lookup(key1, create_cb, true,
-                                  /*advise_erase=*/true, is_in_sec_cache);
-      ASSERT_FALSE(is_in_sec_cache);
+      bool kept_in_sec_cache = false;
+      auto sec_handle1 = secondary_cache->Lookup(
+          key1, BlobSource::SharedCacheInterface::GetFullHelper(),
+          /*context*/ nullptr, true,
+          /*advise_erase=*/true, kept_in_sec_cache);
+      ASSERT_FALSE(kept_in_sec_cache);
       ASSERT_EQ(sec_handle1, nullptr);
 
       ASSERT_TRUE(blob_source.TEST_BlobInCache(file_number, file_size,
@@ -1276,7 +1271,7 @@ TEST_F(BlobSecondaryCacheTest, GetBlobsFromSecondaryCache) {
       // key0 should be in the primary cache.
       CacheKey cache_key0 = base_cache_key.WithOffset(blob_offsets[0]);
       const Slice key0 = cache_key0.AsSlice();
-      auto handle0 = blob_cache->Lookup(key0, statistics);
+      auto handle0 = blob_cache->BasicLookup(key0, statistics);
       ASSERT_NE(handle0, nullptr);
       auto value = static_cast<BlobContents*>(blob_cache->Value(handle0));
       ASSERT_NE(value, nullptr);
@@ -1286,12 +1281,12 @@ TEST_F(BlobSecondaryCacheTest, GetBlobsFromSecondaryCache) {
       // key1 is not in the primary cache and is in the secondary cache.
       CacheKey cache_key1 = base_cache_key.WithOffset(blob_offsets[1]);
       const Slice key1 = cache_key1.AsSlice();
-      auto handle1 = blob_cache->Lookup(key1, statistics);
+      auto handle1 = blob_cache->BasicLookup(key1, statistics);
       ASSERT_EQ(handle1, nullptr);
 
       // erase key0 from the primary cache.
       blob_cache->Erase(key0);
-      handle0 = blob_cache->Lookup(key0, statistics);
+      handle0 = blob_cache->BasicLookup(key0, statistics);
       ASSERT_EQ(handle0, nullptr);
 
       // key1 promotion should succeed due to the primary cache being empty. we
@@ -1307,7 +1302,7 @@ TEST_F(BlobSecondaryCacheTest, GetBlobsFromSecondaryCache) {
       // in the secondary cache. So, the primary cache's Lookup() without
       // secondary cache support cannot see it. (NOTE: The dummy handle used
       // to be a leaky abstraction but not anymore.)
-      handle1 = blob_cache->Lookup(key1, statistics);
+      handle1 = blob_cache->BasicLookup(key1, statistics);
       ASSERT_EQ(handle1, nullptr);
 
       // But after another access, it is promoted to primary cache
@@ -1315,7 +1310,7 @@ TEST_F(BlobSecondaryCacheTest, GetBlobsFromSecondaryCache) {
                                                blob_offsets[1]));
 
       // And Lookup() can find it (without secondary cache support)
-      handle1 = blob_cache->Lookup(key1, statistics);
+      handle1 = blob_cache->BasicLookup(key1, statistics);
       ASSERT_NE(handle1, nullptr);
       ASSERT_NE(blob_cache->Value(handle1), nullptr);
       blob_cache->Release(handle1);
@@ -1379,7 +1374,7 @@ class BlobSourceCacheReservationTest : public DBTestBase {
 
   static constexpr std::size_t kSizeDummyEntry = CacheReservationManagerImpl<
       CacheEntryRole::kBlobCache>::GetDummyEntrySize();
-  static constexpr std::size_t kCacheCapacity = 1 * kSizeDummyEntry;
+  static constexpr std::size_t kCacheCapacity = 2 * kSizeDummyEntry;
   static constexpr int kNumShardBits = 0;  // 2^0 shard
 
   static constexpr uint32_t kColumnFamilyId = 1;
@@ -1398,7 +1393,6 @@ class BlobSourceCacheReservationTest : public DBTestBase {
   std::string db_session_id_;
 };
 
-#ifndef ROCKSDB_LITE
 TEST_F(BlobSourceCacheReservationTest, SimpleCacheReservation) {
   options_.cf_paths.emplace_back(
       test::PerThreadDBPath(
@@ -1513,11 +1507,10 @@ TEST_F(BlobSourceCacheReservationTest, SimpleCacheReservation) {
   }
 }
 
-TEST_F(BlobSourceCacheReservationTest, IncreaseCacheReservationOnFullCache) {
+TEST_F(BlobSourceCacheReservationTest, IncreaseCacheReservation) {
   options_.cf_paths.emplace_back(
       test::PerThreadDBPath(
-          env_,
-          "BlobSourceCacheReservationTest_IncreaseCacheReservationOnFullCache"),
+          env_, "BlobSourceCacheReservationTest_IncreaseCacheReservation"),
       0);
 
   GenerateKeysAndBlobs();
@@ -1525,7 +1518,7 @@ TEST_F(BlobSourceCacheReservationTest, IncreaseCacheReservationOnFullCache) {
   DestroyAndReopen(options_);
 
   ImmutableOptions immutable_options(options_);
-  constexpr size_t blob_size = kSizeDummyEntry / (kNumBlobs / 2);
+  constexpr size_t blob_size = 24 << 10;  // 24KB
   for (size_t i = 0; i < kNumBlobs; ++i) {
     blob_file_size_ -= blobs_[i].size();  // old blob size
     blob_strs_[i].resize(blob_size, '@');
@@ -1583,11 +1576,6 @@ TEST_F(BlobSourceCacheReservationTest, IncreaseCacheReservationOnFullCache) {
 
     std::vector<PinnableSlice> values(keys_.size());
 
-    // Since we resized each blob to be kSizeDummyEntry / (num_blobs / 2), we
-    // can't fit all the blobs in the cache at the same time, which means we
-    // should observe cache evictions once we reach the cache's capacity.
-    // Due to the overhead of the cache and the BlobContents objects, as well as
-    // jemalloc bin sizes, this happens after inserting seven blobs.
     uint64_t blob_bytes = 0;
     for (size_t i = 0; i < kNumBlobs; ++i) {
       ASSERT_OK(blob_source.GetBlob(
@@ -1598,22 +1586,21 @@ TEST_F(BlobSourceCacheReservationTest, IncreaseCacheReservationOnFullCache) {
       // Release cache handle
       values[i].Reset();
 
-      if (i < kNumBlobs / 2 - 1) {
-        size_t charge = 0;
-        ASSERT_TRUE(blob_source.TEST_BlobInCache(
-            kBlobFileNumber, blob_file_size_, blob_offsets[i], &charge));
+      size_t charge = 0;
+      ASSERT_TRUE(blob_source.TEST_BlobInCache(kBlobFileNumber, blob_file_size_,
+                                               blob_offsets[i], &charge));
 
-        blob_bytes += charge;
-      }
+      blob_bytes += charge;
 
-      ASSERT_EQ(cache_res_mgr->GetTotalReservedCacheSize(), kSizeDummyEntry);
+      ASSERT_EQ(cache_res_mgr->GetTotalReservedCacheSize(),
+                (blob_bytes <= kSizeDummyEntry) ? kSizeDummyEntry
+                                                : (2 * kSizeDummyEntry));
       ASSERT_EQ(cache_res_mgr->GetTotalMemoryUsed(), blob_bytes);
       ASSERT_EQ(cache_res_mgr->GetTotalMemoryUsed(),
                 options_.blob_cache->GetUsage());
     }
   }
 }
-#endif  // ROCKSDB_LITE
 
 }  // namespace ROCKSDB_NAMESPACE
 

db/blob/db_blob_basic_test.cc

@@ -11,6 +11,7 @@
 #include "db/blob/blob_index.h"
 #include "db/blob/blob_log_format.h"
 #include "db/db_test_util.h"
+#include "db/db_with_timestamp_test_util.h"
 #include "port/stack_trace.h"
 #include "test_util/sync_point.h"
 #include "utilities/fault_injection_env.h"
@@ -584,7 +585,6 @@ TEST_F(DBBlobBasicTest, MultiGetBlobsFromCache) {
   }
 }
 
-#ifndef ROCKSDB_LITE
 TEST_F(DBBlobBasicTest, MultiGetWithDirectIO) {
   Options options = GetDefaultOptions();
 
@@ -773,13 +773,12 @@ TEST_F(DBBlobBasicTest, MultiGetWithDirectIO) {
     ASSERT_EQ(values[2], second_blob);
   }
 }
-#endif  // !ROCKSDB_LITE
 
 TEST_F(DBBlobBasicTest, MultiGetBlobsFromMultipleFiles) {
   Options options = GetDefaultOptions();
 
   LRUCacheOptions co;
-  co.capacity = 2 << 20; // 2MB
+  co.capacity = 2 << 20;  // 2MB
   co.num_shard_bits = 2;
   co.metadata_charge_policy = kDontChargeCacheMetadata;
   auto backing_cache = NewLRUCache(co);
@@ -1062,7 +1061,6 @@ TEST_F(DBBlobBasicTest, GetBlob_IndexWithInvalidFileNumber) {
                   .IsCorruption());
 }
 
-#ifndef ROCKSDB_LITE
 TEST_F(DBBlobBasicTest, GenerateIOTracing) {
   Options options = GetDefaultOptions();
   options.enable_blob_files = true;
@@ -1117,7 +1115,6 @@ TEST_F(DBBlobBasicTest, GenerateIOTracing) {
     ASSERT_GT(blob_files_op_count, 2);
   }
 }
-#endif  // !ROCKSDB_LITE
 
 TEST_F(DBBlobBasicTest, BestEffortsRecovery_MissingNewestBlobFile) {
   Options options = GetDefaultOptions();
@@ -1219,7 +1216,6 @@ TEST_F(DBBlobBasicTest, MultiGetMergeBlobWithPut) {
   ASSERT_EQ(values[2], "v2_0");
 }
 
-#ifndef ROCKSDB_LITE
 TEST_F(DBBlobBasicTest, Properties) {
   Options options = GetDefaultOptions();
   options.enable_blob_files = true;
@@ -1382,7 +1378,6 @@ TEST_F(DBBlobBasicTest, PropertiesMultiVersion) {
                  BlobLogRecord::CalculateAdjustmentForRecordHeader(key_size) +
                  blob_size + BlobLogFooter::kSize));
 }
-#endif  // !ROCKSDB_LITE
 
 class DBBlobBasicIOErrorTest : public DBBlobBasicTest,
                                public testing::WithParamInterface<std::string> {
@@ -1632,7 +1627,6 @@ TEST_F(DBBlobBasicTest, WarmCacheWithBlobsDuringFlush) {
             options.statistics->getTickerCount(BLOB_DB_CACHE_ADD));
 }
 
-#ifndef ROCKSDB_LITE
 TEST_F(DBBlobBasicTest, DynamicallyWarmCacheDuringFlush) {
   Options options = GetDefaultOptions();
 
@@ -1700,7 +1694,6 @@ TEST_F(DBBlobBasicTest, DynamicallyWarmCacheDuringFlush) {
                               /*end=*/nullptr));
   EXPECT_EQ(0, options.statistics->getTickerCount(BLOB_DB_CACHE_ADD));
 }
-#endif  // !ROCKSDB_LITE
 
 TEST_F(DBBlobBasicTest, WarmCacheWithBlobsSecondary) {
   CompressedSecondaryCacheOptions secondary_cache_opts;
@@ -1779,6 +1772,461 @@ TEST_F(DBBlobBasicTest, WarmCacheWithBlobsSecondary) {
             1);
 }
 
+TEST_F(DBBlobBasicTest, GetEntityBlob) {
+  Options options = GetDefaultOptions();
+  options.enable_blob_files = true;
+  options.min_blob_size = 0;
+
+  Reopen(options);
+
+  constexpr char key[] = "key";
+  constexpr char blob_value[] = "blob_value";
+
+  constexpr char other_key[] = "other_key";
+  constexpr char other_blob_value[] = "other_blob_value";
+
+  ASSERT_OK(Put(key, blob_value));
+  ASSERT_OK(Put(other_key, other_blob_value));
+
+  ASSERT_OK(Flush());
+
+  WideColumns expected_columns{{kDefaultWideColumnName, blob_value}};
+  WideColumns other_expected_columns{
+      {kDefaultWideColumnName, other_blob_value}};
+
+  {
+    PinnableWideColumns result;
+    ASSERT_OK(db_->GetEntity(ReadOptions(), db_->DefaultColumnFamily(), key,
+                             &result));
+    ASSERT_EQ(result.columns(), expected_columns);
+  }
+
+  {
+    PinnableWideColumns result;
+    ASSERT_OK(db_->GetEntity(ReadOptions(), db_->DefaultColumnFamily(),
+                             other_key, &result));
+
+    ASSERT_EQ(result.columns(), other_expected_columns);
+  }
+
+  {
+    constexpr size_t num_keys = 2;
+
+    std::array<Slice, num_keys> keys{{key, other_key}};
+    std::array<PinnableWideColumns, num_keys> results;
+    std::array<Status, num_keys> statuses;
+
+    db_->MultiGetEntity(ReadOptions(), db_->DefaultColumnFamily(), num_keys,
+                        &keys[0], &results[0], &statuses[0]);
+
+    ASSERT_OK(statuses[0]);
+    ASSERT_EQ(results[0].columns(), expected_columns);
+
+    ASSERT_OK(statuses[1]);
+    ASSERT_EQ(results[1].columns(), other_expected_columns);
+  }
+}
+
+class DBBlobWithTimestampTest : public DBBasicTestWithTimestampBase {
+ protected:
+  DBBlobWithTimestampTest()
+      : DBBasicTestWithTimestampBase("db_blob_with_timestamp_test") {}
+};
+
+TEST_F(DBBlobWithTimestampTest, GetBlob) {
+  Options options = GetDefaultOptions();
+  options.create_if_missing = true;
+  options.enable_blob_files = true;
+  options.min_blob_size = 0;
+  const size_t kTimestampSize = Timestamp(0, 0).size();
+  TestComparator test_cmp(kTimestampSize);
+  options.comparator = &test_cmp;
+
+  DestroyAndReopen(options);
+  WriteOptions write_opts;
+  const std::string ts = Timestamp(1, 0);
+  constexpr char key[] = "key";
+  constexpr char blob_value[] = "blob_value";
+
+  ASSERT_OK(db_->Put(write_opts, key, ts, blob_value));
+
+  ASSERT_OK(Flush());
+
+  const std::string read_ts = Timestamp(2, 0);
+  Slice read_ts_slice(read_ts);
+  ReadOptions read_opts;
+  read_opts.timestamp = &read_ts_slice;
+  std::string value;
+  ASSERT_OK(db_->Get(read_opts, key, &value));
+  ASSERT_EQ(value, blob_value);
+}
+
+TEST_F(DBBlobWithTimestampTest, MultiGetBlobs) {
+  constexpr size_t min_blob_size = 6;
+
+  Options options = GetDefaultOptions();
+  options.enable_blob_files = true;
+  options.min_blob_size = min_blob_size;
+  options.create_if_missing = true;
+  const size_t kTimestampSize = Timestamp(0, 0).size();
+  TestComparator test_cmp(kTimestampSize);
+  options.comparator = &test_cmp;
+
+  DestroyAndReopen(options);
+
+  // Put then retrieve three key-values. The first value is below the size limit
+  // and is thus stored inline; the other two are stored separately as blobs.
+  constexpr size_t num_keys = 3;
+
+  constexpr char first_key[] = "first_key";
+  constexpr char first_value[] = "short";
+  static_assert(sizeof(first_value) - 1 < min_blob_size,
+                "first_value too long to be inlined");
+
+  DestroyAndReopen(options);
+  WriteOptions write_opts;
+  const std::string ts = Timestamp(1, 0);
+  ASSERT_OK(db_->Put(write_opts, first_key, ts, first_value));
+
+  constexpr char second_key[] = "second_key";
+  constexpr char second_value[] = "long_value";
+  static_assert(sizeof(second_value) - 1 >= min_blob_size,
+                "second_value too short to be stored as blob");
+
+  ASSERT_OK(db_->Put(write_opts, second_key, ts, second_value));
+
+  constexpr char third_key[] = "third_key";
+  constexpr char third_value[] = "other_long_value";
+  static_assert(sizeof(third_value) - 1 >= min_blob_size,
+                "third_value too short to be stored as blob");
+
+  ASSERT_OK(db_->Put(write_opts, third_key, ts, third_value));
+
+  ASSERT_OK(Flush());
+
+  ReadOptions read_options;
+  const std::string read_ts = Timestamp(2, 0);
+  Slice read_ts_slice(read_ts);
+  read_options.timestamp = &read_ts_slice;
+  std::array<Slice, num_keys> keys{{first_key, second_key, third_key}};
+
+  {
+    std::array<PinnableSlice, num_keys> values;
+    std::array<Status, num_keys> statuses;
+
+    db_->MultiGet(read_options, db_->DefaultColumnFamily(), num_keys, &keys[0],
+                  &values[0], &statuses[0]);
+
+    ASSERT_OK(statuses[0]);
+    ASSERT_EQ(values[0], first_value);
+
+    ASSERT_OK(statuses[1]);
+    ASSERT_EQ(values[1], second_value);
+
+    ASSERT_OK(statuses[2]);
+    ASSERT_EQ(values[2], third_value);
+  }
+}
+
+TEST_F(DBBlobWithTimestampTest, GetMergeBlobWithPut) {
+  Options options = GetDefaultOptions();
+  options.merge_operator = MergeOperators::CreateStringAppendOperator();
+  options.enable_blob_files = true;
+  options.min_blob_size = 0;
+  options.create_if_missing = true;
+  const size_t kTimestampSize = Timestamp(0, 0).size();
+  TestComparator test_cmp(kTimestampSize);
+  options.comparator = &test_cmp;
+
+  DestroyAndReopen(options);
+
+  WriteOptions write_opts;
+  const std::string ts = Timestamp(1, 0);
+  ASSERT_OK(db_->Put(write_opts, "Key1", ts, "v1"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(
+      db_->Merge(write_opts, db_->DefaultColumnFamily(), "Key1", ts, "v2"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(
+      db_->Merge(write_opts, db_->DefaultColumnFamily(), "Key1", ts, "v3"));
+  ASSERT_OK(Flush());
+
+  std::string value;
+  const std::string read_ts = Timestamp(2, 0);
+  Slice read_ts_slice(read_ts);
+  ReadOptions read_opts;
+  read_opts.timestamp = &read_ts_slice;
+  ASSERT_OK(db_->Get(read_opts, "Key1", &value));
+  ASSERT_EQ(value, "v1,v2,v3");
+}
+
+TEST_F(DBBlobWithTimestampTest, MultiGetMergeBlobWithPut) {
+  constexpr size_t num_keys = 3;
+
+  Options options = GetDefaultOptions();
+  options.merge_operator = MergeOperators::CreateStringAppendOperator();
+  options.enable_blob_files = true;
+  options.min_blob_size = 0;
+  options.create_if_missing = true;
+  const size_t kTimestampSize = Timestamp(0, 0).size();
+  TestComparator test_cmp(kTimestampSize);
+  options.comparator = &test_cmp;
+
+  DestroyAndReopen(options);
+
+  WriteOptions write_opts;
+  const std::string ts = Timestamp(1, 0);
+
+  ASSERT_OK(db_->Put(write_opts, "Key0", ts, "v0_0"));
+  ASSERT_OK(db_->Put(write_opts, "Key1", ts, "v1_0"));
+  ASSERT_OK(db_->Put(write_opts, "Key2", ts, "v2_0"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(
+      db_->Merge(write_opts, db_->DefaultColumnFamily(), "Key0", ts, "v0_1"));
+  ASSERT_OK(
+      db_->Merge(write_opts, db_->DefaultColumnFamily(), "Key1", ts, "v1_1"));
+  ASSERT_OK(Flush());
+  ASSERT_OK(
+      db_->Merge(write_opts, db_->DefaultColumnFamily(), "Key0", ts, "v0_2"));
+  ASSERT_OK(Flush());
+
+  const std::string read_ts = Timestamp(2, 0);
+  Slice read_ts_slice(read_ts);
+  ReadOptions read_opts;
+  read_opts.timestamp = &read_ts_slice;
+  std::array<Slice, num_keys> keys{{"Key0", "Key1", "Key2"}};
+  std::array<PinnableSlice, num_keys> values;
+  std::array<Status, num_keys> statuses;
+
+  db_->MultiGet(read_opts, db_->DefaultColumnFamily(), num_keys, &keys[0],
+                &values[0], &statuses[0]);
+
+  ASSERT_OK(statuses[0]);
+  ASSERT_EQ(values[0], "v0_0,v0_1,v0_2");
+
+  ASSERT_OK(statuses[1]);
+  ASSERT_EQ(values[1], "v1_0,v1_1");
+
+  ASSERT_OK(statuses[2]);
+  ASSERT_EQ(values[2], "v2_0");
+}
+
+TEST_F(DBBlobWithTimestampTest, IterateBlobs) {
+  Options options = GetDefaultOptions();
+  options.enable_blob_files = true;
+  options.create_if_missing = true;
+  const size_t kTimestampSize = Timestamp(0, 0).size();
+  TestComparator test_cmp(kTimestampSize);
+  options.comparator = &test_cmp;
+
+  DestroyAndReopen(options);
+
+  int num_blobs = 5;
+  std::vector<std::string> keys;
+  std::vector<std::string> blobs;
+
+  WriteOptions write_opts;
+  std::vector<std::string> write_timestamps = {Timestamp(1, 0),
+                                               Timestamp(2, 0)};
+
+  // For each key in ["key0", ... "keyi", ...], write two versions:
+  // Timestamp(1, 0), "blobi0"
+  // Timestamp(2, 0), "blobi1"
+  for (int i = 0; i < num_blobs; i++) {
+    keys.push_back("key" + std::to_string(i));
+    blobs.push_back("blob" + std::to_string(i));
+    for (size_t j = 0; j < write_timestamps.size(); j++) {
+      ASSERT_OK(db_->Put(write_opts, keys[i], write_timestamps[j],
+                         blobs[i] + std::to_string(j)));
+    }
+  }
+  ASSERT_OK(Flush());
+
+  ReadOptions read_options;
+  std::vector<std::string> read_timestamps = {Timestamp(0, 0), Timestamp(3, 0)};
+  Slice ts_upper_bound(read_timestamps[1]);
+  read_options.timestamp = &ts_upper_bound;
+
+  auto check_iter_entry =
+      [](const Iterator* iter, const std::string& expected_key,
+         const std::string& expected_ts, const std::string& expected_value,
+         bool key_is_internal = true) {
+        ASSERT_OK(iter->status());
+        if (key_is_internal) {
+          std::string expected_ukey_and_ts;
+          expected_ukey_and_ts.assign(expected_key.data(), expected_key.size());
+          expected_ukey_and_ts.append(expected_ts.data(), expected_ts.size());
+
+          ParsedInternalKey parsed_ikey;
+          ASSERT_OK(ParseInternalKey(iter->key(), &parsed_ikey,
+                                     true /* log_err_key */));
+          ASSERT_EQ(parsed_ikey.user_key, expected_ukey_and_ts);
+        } else {
+          ASSERT_EQ(iter->key(), expected_key);
+        }
+        ASSERT_EQ(iter->timestamp(), expected_ts);
+        ASSERT_EQ(iter->value(), expected_value);
+      };
+
+  // Forward iterating one version of each key, get in this order:
+  // [("key0", Timestamp(2, 0), "blob01"),
+  //  ("key1", Timestamp(2, 0), "blob11")...]
+  {
+    std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
+    ASSERT_OK(iter->status());
+
+    iter->SeekToFirst();
+    for (int i = 0; i < num_blobs; i++) {
+      check_iter_entry(iter.get(), keys[i], write_timestamps[1],
+                       blobs[i] + std::to_string(1), /*key_is_internal*/ false);
+      iter->Next();
+    }
+  }
+
+  // Forward iteration, then reverse to backward.
+  {
+    std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
+    ASSERT_OK(iter->status());
+
+    iter->SeekToFirst();
+    for (int i = 0; i < num_blobs * 2 - 1; i++) {
+      if (i < num_blobs) {
+        check_iter_entry(iter.get(), keys[i], write_timestamps[1],
+                         blobs[i] + std::to_string(1),
+                         /*key_is_internal*/ false);
+        if (i != num_blobs - 1) {
+          iter->Next();
+        }
+      } else {
+        if (i != num_blobs) {
+          check_iter_entry(iter.get(), keys[num_blobs * 2 - 1 - i],
+                           write_timestamps[1],
+                           blobs[num_blobs * 2 - 1 - i] + std::to_string(1),
+                           /*key_is_internal*/ false);
+        }
+        iter->Prev();
+      }
+    }
+  }
+
+  // Backward iterating one versions of each key, get in this order:
+  // [("key4", Timestamp(2, 0), "blob41"),
+  //  ("key3", Timestamp(2, 0), "blob31")...]
+  {
+    std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
+    ASSERT_OK(iter->status());
+
+    iter->SeekToLast();
+    for (int i = 0; i < num_blobs; i++) {
+      check_iter_entry(iter.get(), keys[num_blobs - 1 - i], write_timestamps[1],
+                       blobs[num_blobs - 1 - i] + std::to_string(1),
+                       /*key_is_internal*/ false);
+      iter->Prev();
+    }
+  }
+
+  // Backward iteration, then reverse to forward.
+  {
+    std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
+    ASSERT_OK(iter->status());
+
+    iter->SeekToLast();
+    for (int i = 0; i < num_blobs * 2 - 1; i++) {
+      if (i < num_blobs) {
+        check_iter_entry(iter.get(), keys[num_blobs - 1 - i],
+                         write_timestamps[1],
+                         blobs[num_blobs - 1 - i] + std::to_string(1),
+                         /*key_is_internal*/ false);
+        if (i != num_blobs - 1) {
+          iter->Prev();
+        }
+      } else {
+        if (i != num_blobs) {
+          check_iter_entry(iter.get(), keys[i - num_blobs], write_timestamps[1],
+                           blobs[i - num_blobs] + std::to_string(1),
+                           /*key_is_internal*/ false);
+        }
+        iter->Next();
+      }
+    }
+  }
+
+  Slice ts_lower_bound(read_timestamps[0]);
+  read_options.iter_start_ts = &ts_lower_bound;
+  // Forward iterating multiple versions of the same key, get in this order:
+  // [("key0", Timestamp(2, 0), "blob01"),
+  //  ("key0", Timestamp(1, 0), "blob00"),
+  //  ("key1", Timestamp(2, 0), "blob11")...]
+  {
+    std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
+    ASSERT_OK(iter->status());
+
+    iter->SeekToFirst();
+    for (int i = 0; i < num_blobs; i++) {
+      for (size_t j = write_timestamps.size(); j > 0; --j) {
+        check_iter_entry(iter.get(), keys[i], write_timestamps[j - 1],
+                         blobs[i] + std::to_string(j - 1));
+        iter->Next();
+      }
+    }
+  }
+
+  // Backward iterating multiple versions of the same key, get in this order:
+  // [("key4", Timestamp(1, 0), "blob00"),
+  //  ("key4", Timestamp(2, 0), "blob01"),
+  //  ("key3", Timestamp(1, 0), "blob10")...]
+  {
+    std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
+    ASSERT_OK(iter->status());
+
+    iter->SeekToLast();
+    for (int i = num_blobs; i > 0; i--) {
+      for (size_t j = 0; j < write_timestamps.size(); j++) {
+        check_iter_entry(iter.get(), keys[i - 1], write_timestamps[j],
+                         blobs[i - 1] + std::to_string(j));
+        iter->Prev();
+      }
+    }
+  }
+
+  int upper_bound_idx = num_blobs - 2;
+  int lower_bound_idx = 1;
+  Slice upper_bound_slice(keys[upper_bound_idx]);
+  Slice lower_bound_slice(keys[lower_bound_idx]);
+  read_options.iterate_upper_bound = &upper_bound_slice;
+  read_options.iterate_lower_bound = &lower_bound_slice;
+
+  // Forward iteration with upper and lower bound.
+  {
+    std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
+    ASSERT_OK(iter->status());
+
+    iter->SeekToFirst();
+    for (int i = lower_bound_idx; i < upper_bound_idx; i++) {
+      for (size_t j = write_timestamps.size(); j > 0; --j) {
+        check_iter_entry(iter.get(), keys[i], write_timestamps[j - 1],
+                         blobs[i] + std::to_string(j - 1));
+        iter->Next();
+      }
+    }
+  }
+
+  // Backward iteration with upper and lower bound.
+  {
+    std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
+    ASSERT_OK(iter->status());
+
+    iter->SeekToLast();
+    for (int i = upper_bound_idx; i > lower_bound_idx; i--) {
+      for (size_t j = 0; j < write_timestamps.size(); j++) {
+        check_iter_entry(iter.get(), keys[i - 1], write_timestamps[j],
+                         blobs[i - 1] + std::to_string(j));
+        iter->Prev();
+      }
+    }
+  }
+}
+
 }  // namespace ROCKSDB_NAMESPACE
 
 int main(int argc, char** argv) {

db/blob/db_blob_compaction_test.cc

@@ -16,7 +16,6 @@ class DBBlobCompactionTest : public DBTestBase {
   explicit DBBlobCompactionTest()
       : DBTestBase("db_blob_compaction_test", /*env_do_fsync=*/false) {}
 
-#ifndef ROCKSDB_LITE
   const std::vector<InternalStats::CompactionStats>& GetCompactionStats() {
     VersionSet* const versions = dbfull()->GetVersionSet();
     assert(versions);
@@ -30,7 +29,6 @@ class DBBlobCompactionTest : public DBTestBase {
 
     return internal_stats->TEST_GetCompactionStats();
   }
-#endif  // ROCKSDB_LITE
 };
 
 namespace {
@@ -250,7 +248,6 @@ TEST_F(DBBlobCompactionTest, FilterByKeyLength) {
   ASSERT_OK(db_->Get(ReadOptions(), long_key, &value));
   ASSERT_EQ("value", value);
 
-#ifndef ROCKSDB_LITE
   const auto& compaction_stats = GetCompactionStats();
   ASSERT_GE(compaction_stats.size(), 2);
 
@@ -258,7 +255,6 @@ TEST_F(DBBlobCompactionTest, FilterByKeyLength) {
   // this involves neither reading nor writing blobs
   ASSERT_EQ(compaction_stats[1].bytes_read_blob, 0);
   ASSERT_EQ(compaction_stats[1].bytes_written_blob, 0);
-#endif  // ROCKSDB_LITE
 
   Close();
 }
@@ -299,7 +295,6 @@ TEST_F(DBBlobCompactionTest, FilterByValueLength) {
     ASSERT_EQ(long_value, value);
   }
 
-#ifndef ROCKSDB_LITE
   const auto& compaction_stats = GetCompactionStats();
   ASSERT_GE(compaction_stats.size(), 2);
 
@@ -307,12 +302,10 @@ TEST_F(DBBlobCompactionTest, FilterByValueLength) {
   // this involves reading but not writing blobs
   ASSERT_GT(compaction_stats[1].bytes_read_blob, 0);
   ASSERT_EQ(compaction_stats[1].bytes_written_blob, 0);
-#endif  // ROCKSDB_LITE
 
   Close();
 }
 
-#ifndef ROCKSDB_LITE
 TEST_F(DBBlobCompactionTest, BlobCompactWithStartingLevel) {
   Options options = GetDefaultOptions();
 
@@ -388,7 +381,6 @@ TEST_F(DBBlobCompactionTest, BlobCompactWithStartingLevel) {
 
   Close();
 }
-#endif
 
 TEST_F(DBBlobCompactionTest, BlindWriteFilter) {
   Options options = GetDefaultOptions();
@@ -413,7 +405,6 @@ TEST_F(DBBlobCompactionTest, BlindWriteFilter) {
     ASSERT_EQ(new_blob_value, Get(key));
   }
 
-#ifndef ROCKSDB_LITE
   const auto& compaction_stats = GetCompactionStats();
   ASSERT_GE(compaction_stats.size(), 2);
 
@@ -421,7 +412,6 @@ TEST_F(DBBlobCompactionTest, BlindWriteFilter) {
   // this involves writing but not reading blobs
   ASSERT_EQ(compaction_stats[1].bytes_read_blob, 0);
   ASSERT_GT(compaction_stats[1].bytes_written_blob, 0);
-#endif  // ROCKSDB_LITE
 
   Close();
 }
@@ -540,7 +530,6 @@ TEST_F(DBBlobCompactionTest, CompactionFilter) {
     ASSERT_EQ(kv.second + std::string(padding), Get(kv.first));
   }
 
-#ifndef ROCKSDB_LITE
   const auto& compaction_stats = GetCompactionStats();
   ASSERT_GE(compaction_stats.size(), 2);
 
@@ -548,7 +537,6 @@ TEST_F(DBBlobCompactionTest, CompactionFilter) {
   // this involves reading and writing blobs
   ASSERT_GT(compaction_stats[1].bytes_read_blob, 0);
   ASSERT_GT(compaction_stats[1].bytes_written_blob, 0);
-#endif  // ROCKSDB_LITE
 
   Close();
 }
@@ -606,7 +594,6 @@ TEST_F(DBBlobCompactionTest, CompactionFilterReadBlobAndKeep) {
                               /*end=*/nullptr));
   ASSERT_EQ(blob_files, GetBlobFileNumbers());
 
-#ifndef ROCKSDB_LITE
   const auto& compaction_stats = GetCompactionStats();
   ASSERT_GE(compaction_stats.size(), 2);
 
@@ -614,7 +601,6 @@ TEST_F(DBBlobCompactionTest, CompactionFilterReadBlobAndKeep) {
   // this involves reading but not writing blobs
   ASSERT_GT(compaction_stats[1].bytes_read_blob, 0);
   ASSERT_EQ(compaction_stats[1].bytes_written_blob, 0);
-#endif  // ROCKSDB_LITE
 
   Close();
 }

db/blob/db_blob_corruption_test.cc

@@ -34,7 +34,6 @@ class DBBlobCorruptionTest : public DBTestBase {
   }
 };
 
-#ifndef ROCKSDB_LITE
 TEST_F(DBBlobCorruptionTest, VerifyWholeBlobFileChecksum) {
   Options options = GetDefaultOptions();
   options.enable_blob_files = true;
@@ -71,7 +70,6 @@ TEST_F(DBBlobCorruptionTest, VerifyWholeBlobFileChecksum) {
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
 }
-#endif  // !ROCKSDB_LITE
 }  // namespace ROCKSDB_NAMESPACE
 
 int main(int argc, char** argv) {

db/blob/db_blob_index_test.cc

@@ -131,9 +131,7 @@ class DBBlobIndexTest : public DBTestBase {
         ASSERT_OK(Flush());
         ASSERT_OK(
             dbfull()->CompactRange(CompactRangeOptions(), nullptr, nullptr));
-#ifndef ROCKSDB_LITE
         ASSERT_EQ("0,1", FilesPerLevel());
-#endif  // !ROCKSDB_LITE
         break;
     }
   }
@@ -459,7 +457,6 @@ TEST_F(DBBlobIndexTest, Iterate) {
     verify(15, Status::kOk, get_value(16, 0), get_value(14, 0),
            create_blob_iterator, check_is_blob(false));
 
-#ifndef ROCKSDB_LITE
     // Iterator with blob support and using seek.
     ASSERT_OK(dbfull()->SetOptions(
         cfh(), {{"max_sequential_skip_in_iterations", "0"}}));
@@ -484,7 +481,6 @@ TEST_F(DBBlobIndexTest, Iterate) {
            create_blob_iterator, check_is_blob(false));
     verify(15, Status::kOk, get_value(16, 0), get_value(14, 0),
            create_blob_iterator, check_is_blob(false));
-#endif  // !ROCKSDB_LITE
 
     for (auto* snapshot : snapshots) {
       dbfull()->ReleaseSnapshot(snapshot);
@@ -584,12 +580,10 @@ TEST_F(DBBlobIndexTest, IntegratedBlobIterate) {
   Status expected_status;
   verify(1, expected_status, expected_value);
 
-#ifndef ROCKSDB_LITE
   // Test DBIter::FindValueForCurrentKeyUsingSeek flow.
   ASSERT_OK(dbfull()->SetOptions(cfh(),
                                  {{"max_sequential_skip_in_iterations", "0"}}));
   verify(1, expected_status, expected_value);
-#endif  // !ROCKSDB_LITE
 }
 
 }  // namespace ROCKSDB_NAMESPACE

db/builder.cc

@@ -56,8 +56,8 @@ TableBuilder* NewTableBuilder(const TableBuilderOptions& tboptions,
 Status BuildTable(
     const std::string& dbname, VersionSet* versions,
     const ImmutableDBOptions& db_options, const TableBuilderOptions& tboptions,
-    const FileOptions& file_options, TableCache* table_cache,
-    InternalIterator* iter,
+    const FileOptions& file_options, const ReadOptions& read_options,
+    TableCache* table_cache, InternalIterator* iter,
     std::vector<std::unique_ptr<FragmentedRangeTombstoneIterator>>
         range_del_iters,
     FileMetaData* meta, std::vector<BlobFileAddition>* blob_file_additions,
@@ -71,8 +71,9 @@ Status BuildTable(
     int job_id, const Env::IOPriority io_priority,
     TableProperties* table_properties, Env::WriteLifeTimeHint write_hint,
     const std::string* full_history_ts_low,
-    BlobFileCompletionCallback* blob_callback, uint64_t* num_input_entries,
-    uint64_t* memtable_payload_bytes, uint64_t* memtable_garbage_bytes) {
+    BlobFileCompletionCallback* blob_callback, Version* version,
+    uint64_t* num_input_entries, uint64_t* memtable_payload_bytes,
+    uint64_t* memtable_garbage_bytes) {
   assert((tboptions.column_family_id ==
           TablePropertiesCollectorFactory::Context::kUnknownColumnFamily) ==
          tboptions.column_family_name.empty());
@@ -106,11 +107,9 @@ Status BuildTable(
   std::vector<std::string> blob_file_paths;
   std::string file_checksum = kUnknownFileChecksum;
   std::string file_checksum_func_name = kUnknownFileChecksumFuncName;
-#ifndef ROCKSDB_LITE
   EventHelpers::NotifyTableFileCreationStarted(ioptions.listeners, dbname,
                                                tboptions.column_family_name,
                                                fname, job_id, tboptions.reason);
-#endif  // !ROCKSDB_LITE
   Env* env = db_options.env;
   assert(env);
   FileSystem* fs = db_options.fs.get();
@@ -175,10 +174,10 @@ Status BuildTable(
       builder = NewTableBuilder(tboptions, file_writer.get());
     }
 
+    auto ucmp = tboptions.internal_comparator.user_comparator();
     MergeHelper merge(
-        env, tboptions.internal_comparator.user_comparator(),
-        ioptions.merge_operator.get(), compaction_filter.get(), ioptions.logger,
-        true /* internal key corruption is not ok */,
+        env, ucmp, ioptions.merge_operator.get(), compaction_filter.get(),
+        ioptions.logger, true /* internal key corruption is not ok */,
         snapshots.empty() ? 0 : snapshots.back(), snapshot_checker);
 
     std::unique_ptr<BlobFileBuilder> blob_file_builder(
@@ -196,9 +195,8 @@ Status BuildTable(
 
     const std::atomic<bool> kManualCompactionCanceledFalse{false};
     CompactionIterator c_iter(
-        iter, tboptions.internal_comparator.user_comparator(), &merge,
-        kMaxSequenceNumber, &snapshots, earliest_write_conflict_snapshot,
-        job_snapshot, snapshot_checker, env,
+        iter, ucmp, &merge, kMaxSequenceNumber, &snapshots,
+        earliest_write_conflict_snapshot, job_snapshot, snapshot_checker, env,
         ShouldReportDetailedTime(env, ioptions.stats),
         true /* internal key corruption is not ok */, range_del_agg.get(),
         blob_file_builder.get(), ioptions.allow_data_in_errors,
@@ -241,14 +239,28 @@ Status BuildTable(
 
     if (s.ok()) {
       auto range_del_it = range_del_agg->NewIterator();
+      Slice last_tombstone_start_user_key{};
       for (range_del_it->SeekToFirst(); range_del_it->Valid();
            range_del_it->Next()) {
         auto tombstone = range_del_it->Tombstone();
         auto kv = tombstone.Serialize();
         builder->Add(kv.first.Encode(), kv.second);
-        meta->UpdateBoundariesForRange(kv.first, tombstone.SerializeEndKey(),
-                                       tombstone.seq_,
+        InternalKey tombstone_end = tombstone.SerializeEndKey();
+        meta->UpdateBoundariesForRange(kv.first, tombstone_end, tombstone.seq_,
                                        tboptions.internal_comparator);
+        if (version) {
+          if (last_tombstone_start_user_key.empty() ||
+              ucmp->CompareWithoutTimestamp(last_tombstone_start_user_key,
+                                            range_del_it->start_key()) < 0) {
+            SizeApproximationOptions approx_opts;
+            approx_opts.files_size_error_margin = 0.1;
+            meta->compensated_range_deletion_size += versions->ApproximateSize(
+                approx_opts, read_options, version, kv.first.Encode(),
+                tombstone_end.Encode(), 0 /* start_level */, -1 /* end_level */,
+                TableReaderCaller::kFlush);
+          }
+          last_tombstone_start_user_key = range_del_it->start_key();
+        }
       }
     }
 
@@ -281,7 +293,8 @@ Status BuildTable(
       meta->fd.file_size = file_size;
       meta->marked_for_compaction = builder->NeedCompact();
       assert(meta->fd.GetFileSize() > 0);
-      tp = builder->GetTableProperties(); // refresh now that builder is finished
+      tp = builder
+               ->GetTableProperties();  // refresh now that builder is finished
       if (memtable_payload_bytes != nullptr &&
           memtable_garbage_bytes != nullptr) {
         const CompactionIterationStats& ci_stats = c_iter.iter_stats();
@@ -356,7 +369,6 @@ Status BuildTable(
       // here because this is a special case after we finish the table building.
       // No matter whether use_direct_io_for_flush_and_compaction is true,
       // the goal is to cache it here for further user reads.
-      ReadOptions read_options;
       std::unique_ptr<InternalIterator> it(table_cache->NewIterator(
           read_options, file_options, tboptions.internal_comparator, *meta,
           nullptr /* range_del_agg */, mutable_cf_options.prefix_extractor,

db/builder.h

@@ -13,6 +13,7 @@
 #include "db/range_tombstone_fragmenter.h"
 #include "db/seqno_to_time_mapping.h"
 #include "db/table_properties_collector.h"
+#include "db/version_set.h"
 #include "logging/event_logger.h"
 #include "options/cf_options.h"
 #include "rocksdb/comparator.h"
@@ -52,8 +53,8 @@ TableBuilder* NewTableBuilder(const TableBuilderOptions& tboptions,
 extern Status BuildTable(
     const std::string& dbname, VersionSet* versions,
     const ImmutableDBOptions& db_options, const TableBuilderOptions& tboptions,
-    const FileOptions& file_options, TableCache* table_cache,
-    InternalIterator* iter,
+    const FileOptions& file_options, const ReadOptions& read_options,
+    TableCache* table_cache, InternalIterator* iter,
     std::vector<std::unique_ptr<FragmentedRangeTombstoneIterator>>
         range_del_iters,
     FileMetaData* meta, std::vector<BlobFileAddition>* blob_file_additions,
@@ -70,7 +71,7 @@ extern Status BuildTable(
     Env::WriteLifeTimeHint write_hint = Env::WLTH_NOT_SET,
     const std::string* full_history_ts_low = nullptr,
     BlobFileCompletionCallback* blob_callback = nullptr,
-    uint64_t* num_input_entries = nullptr,
+    Version* version = nullptr, uint64_t* num_input_entries = nullptr,
     uint64_t* memtable_payload_bytes = nullptr,
     uint64_t* memtable_garbage_bytes = nullptr);
 

db/c_test.c

@@ -5,8 +5,6 @@
 
 #include <stdio.h>
 
-#ifndef ROCKSDB_LITE  // Lite does not support C API
-
 #include <assert.h>
 #include <stddef.h>
 #include <stdlib.h>
@@ -49,32 +47,32 @@ static void StartPhase(const char* name) {
 }
 #ifdef _MSC_VER
 #pragma warning(push)
-#pragma warning (disable: 4996) // getenv security warning
+#pragma warning(disable : 4996)  // getenv security warning
 #endif
 static const char* GetTempDir(void) {
-    const char* ret = getenv("TEST_TMPDIR");
-    if (ret == NULL || ret[0] == '\0')
+  const char* ret = getenv("TEST_TMPDIR");
+  if (ret == NULL || ret[0] == '\0')
 #ifdef OS_WIN
-      ret = getenv("TEMP");
+    ret = getenv("TEMP");
 #else
-      ret = "/tmp";
+    ret = "/tmp";
 #endif
-    return ret;
+  return ret;
 }
 #ifdef _MSC_VER
 #pragma warning(pop)
 #endif
 
-#define CheckNoError(err)                                               \
-  if ((err) != NULL) {                                                  \
+#define CheckNoError(err)                                                 \
+  if ((err) != NULL) {                                                    \
     fprintf(stderr, "%s:%d: %s: %s\n", __FILE__, __LINE__, phase, (err)); \
-    abort();                                                            \
+    abort();                                                              \
   }
 
-#define CheckCondition(cond)                                            \
-  if (!(cond)) {                                                        \
+#define CheckCondition(cond)                                              \
+  if (!(cond)) {                                                          \
     fprintf(stderr, "%s:%d: %s: %s\n", __FILE__, __LINE__, phase, #cond); \
-    abort();                                                            \
+    abort();                                                              \
   }
 
 static void CheckEqual(const char* expected, const char* v, size_t n) {
@@ -98,21 +96,15 @@ static void Free(char** ptr) {
   }
 }
 
-static void CheckValue(
-    char* err,
-    const char* expected,
-    char** actual,
-    size_t actual_length) {
+static void CheckValue(char* err, const char* expected, char** actual,
+                       size_t actual_length) {
   CheckNoError(err);
   CheckEqual(expected, *actual, actual_length);
   Free(actual);
 }
 
-static void CheckGet(
-    rocksdb_t* db,
-    const rocksdb_readoptions_t* options,
-    const char* key,
-    const char* expected) {
+static void CheckGet(rocksdb_t* db, const rocksdb_readoptions_t* options,
+                     const char* key, const char* expected) {
   char* err = NULL;
   size_t val_len;
   char* val;
@@ -122,12 +114,9 @@ static void CheckGet(
   Free(&val);
 }
 
-static void CheckGetCF(
-    rocksdb_t* db,
-    const rocksdb_readoptions_t* options,
-    rocksdb_column_family_handle_t* handle,
-    const char* key,
-    const char* expected) {
+static void CheckGetCF(rocksdb_t* db, const rocksdb_readoptions_t* options,
+                       rocksdb_column_family_handle_t* handle, const char* key,
+                       const char* expected) {
   char* err = NULL;
   size_t val_len;
   char* val;
@@ -174,8 +163,8 @@ static void CheckMultiGetValues(size_t num_keys, char** values,
   }
 }
 
-static void CheckIter(rocksdb_iterator_t* iter,
-                      const char* key, const char* val) {
+static void CheckIter(rocksdb_iterator_t* iter, const char* key,
+                      const char* val) {
   size_t len;
   const char* str;
   str = rocksdb_iter_key(iter, &len);
@@ -185,10 +174,9 @@ static void CheckIter(rocksdb_iterator_t* iter,
 }
 
 // Callback from rocksdb_writebatch_iterate()
-static void CheckPut(void* ptr,
-                     const char* k, size_t klen,
-                     const char* v, size_t vlen) {
-  int* state = (int*) ptr;
+static void CheckPut(void* ptr, const char* k, size_t klen, const char* v,
+                     size_t vlen) {
+  int* state = (int*)ptr;
   CheckCondition(*state < 2);
   switch (*state) {
     case 0:
@@ -205,7 +193,7 @@ static void CheckPut(void* ptr,
 
 // Callback from rocksdb_writebatch_iterate()
 static void CheckDel(void* ptr, const char* k, size_t klen) {
-  int* state = (int*) ptr;
+  int* state = (int*)ptr;
   CheckCondition(*state == 2);
   CheckEqual("bar", k, klen);
   (*state)++;
@@ -213,14 +201,16 @@ static void CheckDel(void* ptr, const char* k, size_t klen) {
 
 static void CmpDestroy(void* arg) { (void)arg; }
 
-static int CmpCompare(void* arg, const char* a, size_t alen,
-                      const char* b, size_t blen) {
+static int CmpCompare(void* arg, const char* a, size_t alen, const char* b,
+                      size_t blen) {
   (void)arg;
   size_t n = (alen < blen) ? alen : blen;
   int r = memcmp(a, b, n);
   if (r == 0) {
-    if (alen < blen) r = -1;
-    else if (alen > blen) r = +1;
+    if (alen < blen)
+      r = -1;
+    else if (alen > blen)
+      r = +1;
   }
   return r;
 }
@@ -405,11 +395,9 @@ static const char* MergeOperatorName(void* arg) {
   return "TestMergeOperator";
 }
 static char* MergeOperatorFullMerge(
-    void* arg,
-    const char* key, size_t key_length,
-    const char* existing_value, size_t existing_value_length,
-    const char* const* operands_list, const size_t* operands_list_length,
-    int num_operands,
+    void* arg, const char* key, size_t key_length, const char* existing_value,
+    size_t existing_value_length, const char* const* operands_list,
+    const size_t* operands_list_length, int num_operands,
     unsigned char* success, size_t* new_value_length) {
   (void)arg;
   (void)key;
@@ -425,12 +413,12 @@ static char* MergeOperatorFullMerge(
   memcpy(result, "fake", 4);
   return result;
 }
-static char* MergeOperatorPartialMerge(
-    void* arg,
-    const char* key, size_t key_length,
-    const char* const* operands_list, const size_t* operands_list_length,
-    int num_operands,
-    unsigned char* success, size_t* new_value_length) {
+static char* MergeOperatorPartialMerge(void* arg, const char* key,
+                                       size_t key_length,
+                                       const char* const* operands_list,
+                                       const size_t* operands_list_length,
+                                       int num_operands, unsigned char* success,
+                                       size_t* new_value_length) {
   (void)arg;
   (void)key;
   (void)key_length;
@@ -444,18 +432,16 @@ static char* MergeOperatorPartialMerge(
   return result;
 }
 
-static void CheckTxnGet(
-        rocksdb_transaction_t* txn,
-        const rocksdb_readoptions_t* options,
-        const char* key,
-        const char* expected) {
-        char* err = NULL;
-        size_t val_len;
-        char* val;
-        val = rocksdb_transaction_get(txn, options, key, strlen(key), &val_len, &err);
-        CheckNoError(err);
-        CheckEqual(expected, val, val_len);
-        Free(&val);
+static void CheckTxnGet(rocksdb_transaction_t* txn,
+                        const rocksdb_readoptions_t* options, const char* key,
+                        const char* expected) {
+  char* err = NULL;
+  size_t val_len;
+  char* val;
+  val = rocksdb_transaction_get(txn, options, key, strlen(key), &val_len, &err);
+  CheckNoError(err);
+  CheckEqual(expected, val, val_len);
+  Free(&val);
 }
 
 static void CheckTxnGetCF(rocksdb_transaction_t* txn,
@@ -502,11 +488,22 @@ static void CheckTxnPinGetCF(rocksdb_transaction_t* txn,
   rocksdb_pinnableslice_destroy(p);
 }
 
-static void CheckTxnDBGet(
-        rocksdb_transactiondb_t* txn_db,
-        const rocksdb_readoptions_t* options,
-        const char* key,
-        const char* expected) {
+static void CheckTxnGetForUpdate(rocksdb_transaction_t* txn,
+                                 const rocksdb_readoptions_t* options,
+                                 const char* key, const char* expected) {
+  char* err = NULL;
+  size_t val_len;
+  char* val;
+  val = rocksdb_transaction_get_for_update(txn, options, key, strlen(key),
+                                           &val_len, true, &err);
+  CheckNoError(err);
+  CheckEqual(expected, val, val_len);
+  Free(&val);
+}
+
+static void CheckTxnDBGet(rocksdb_transactiondb_t* txn_db,
+                          const rocksdb_readoptions_t* options, const char* key,
+                          const char* expected) {
   char* err = NULL;
   size_t val_len;
   char* val;
@@ -531,6 +528,20 @@ static void CheckTxnDBGetCF(rocksdb_transactiondb_t* txn_db,
   Free(&val);
 }
 
+static void CheckTxnGetForUpdateCF(
+    rocksdb_transaction_t* txn, const rocksdb_readoptions_t* options,
+    rocksdb_column_family_handle_t* column_family, const char* key,
+    const char* expected) {
+  char* err = NULL;
+  size_t val_len;
+  char* val;
+  val = rocksdb_transaction_get_for_update_cf(
+      txn, options, column_family, key, strlen(key), &val_len, true, &err);
+  CheckNoError(err);
+  CheckEqual(expected, val, val_len);
+  Free(&val);
+}
+
 static void CheckTxnDBPinGet(rocksdb_transactiondb_t* txn_db,
                              const rocksdb_readoptions_t* options,
                              const char* key, const char* expected) {
@@ -632,7 +643,7 @@ int main(int argc, char** argv) {
   rocksdb_t* db;
   rocksdb_comparator_t* cmp;
   rocksdb_cache_t* cache;
-  rocksdb_dbpath_t *dbpath;
+  rocksdb_dbpath_t* dbpath;
   rocksdb_env_t* env;
   rocksdb_options_t* options;
   rocksdb_compactoptions_t* coptions;
@@ -649,30 +660,20 @@ int main(int argc, char** argv) {
   char* err = NULL;
   int run = -1;
 
-  snprintf(dbname, sizeof(dbname),
-           "%s/rocksdb_c_test-%d",
-           GetTempDir(),
-           ((int) geteuid()));
-
-  snprintf(dbbackupname, sizeof(dbbackupname),
-           "%s/rocksdb_c_test-%d-backup",
-           GetTempDir(),
-           ((int) geteuid()));
-
-  snprintf(dbcheckpointname, sizeof(dbcheckpointname),
-           "%s/rocksdb_c_test-%d-checkpoint",
-           GetTempDir(),
-           ((int) geteuid()));
-
-  snprintf(sstfilename, sizeof(sstfilename),
-           "%s/rocksdb_c_test-%d-sst",
-           GetTempDir(),
+  snprintf(dbname, sizeof(dbname), "%s/rocksdb_c_test-%d", GetTempDir(),
            ((int)geteuid()));
 
-  snprintf(dbpathname, sizeof(dbpathname),
-           "%s/rocksdb_c_test-%d-dbpath",
-           GetTempDir(),
-           ((int) geteuid()));
+  snprintf(dbbackupname, sizeof(dbbackupname), "%s/rocksdb_c_test-%d-backup",
+           GetTempDir(), ((int)geteuid()));
+
+  snprintf(dbcheckpointname, sizeof(dbcheckpointname),
+           "%s/rocksdb_c_test-%d-checkpoint", GetTempDir(), ((int)geteuid()));
+
+  snprintf(sstfilename, sizeof(sstfilename), "%s/rocksdb_c_test-%d-sst",
+           GetTempDir(), ((int)geteuid()));
+
+  snprintf(dbpathname, sizeof(dbpathname), "%s/rocksdb_c_test-%d-dbpath",
+           GetTempDir(), ((int)geteuid()));
 
   StartPhase("create_objects");
   cmp = rocksdb_comparator_create(NULL, CmpDestroy, CmpCompare, CmpName);
@@ -746,7 +747,8 @@ int main(int argc, char** argv) {
     rocksdb_destroy_db(options, dbbackupname, &err);
     CheckNoError(err);
 
-    rocksdb_backup_engine_t *be = rocksdb_backup_engine_open(options, dbbackupname, &err);
+    rocksdb_backup_engine_t* be =
+        rocksdb_backup_engine_open(options, dbbackupname, &err);
     CheckNoError(err);
 
     rocksdb_backup_engine_create_new_backup(be, db, &err);
@@ -759,7 +761,8 @@ int main(int argc, char** argv) {
     rocksdb_backup_engine_create_new_backup(be, db, &err);
     CheckNoError(err);
 
-    const rocksdb_backup_engine_info_t* bei = rocksdb_backup_engine_get_backup_info(be);
+    const rocksdb_backup_engine_info_t* bei =
+        rocksdb_backup_engine_get_backup_info(be);
     CheckCondition(rocksdb_backup_engine_info_count(bei) > 1);
     rocksdb_backup_engine_info_destroy(bei);
 
@@ -778,9 +781,11 @@ int main(int argc, char** argv) {
     rocksdb_destroy_db(options, dbname, &err);
     CheckNoError(err);
 
-    rocksdb_restore_options_t *restore_options = rocksdb_restore_options_create();
+    rocksdb_restore_options_t* restore_options =
+        rocksdb_restore_options_create();
     rocksdb_restore_options_set_keep_log_files(restore_options, 0);
-    rocksdb_backup_engine_restore_db_from_latest_backup(be, dbname, dbname, restore_options, &err);
+    rocksdb_backup_engine_restore_db_from_latest_backup(be, dbname, dbname,
+                                                        restore_options, &err);
     CheckNoError(err);
     rocksdb_restore_options_destroy(restore_options);
 
@@ -799,7 +804,8 @@ int main(int argc, char** argv) {
     rocksdb_destroy_db(options, dbcheckpointname, &err);
     CheckNoError(err);
 
-    rocksdb_checkpoint_t* checkpoint = rocksdb_checkpoint_object_create(db, &err);
+    rocksdb_checkpoint_t* checkpoint =
+        rocksdb_checkpoint_object_create(db, &err);
     CheckNoError(err);
 
     rocksdb_checkpoint_create(checkpoint, dbcheckpointname, 0, &err);
@@ -976,10 +982,10 @@ int main(int argc, char** argv) {
   StartPhase("writebatch_vectors");
   {
     rocksdb_writebatch_t* wb = rocksdb_writebatch_create();
-    const char* k_list[2] = { "z", "ap" };
-    const size_t k_sizes[2] = { 1, 2 };
-    const char* v_list[3] = { "x", "y", "z" };
-    const size_t v_sizes[3] = { 1, 1, 1 };
+    const char* k_list[2] = {"z", "ap"};
+    const size_t k_sizes[2] = {1, 2};
+    const char* v_list[3] = {"x", "y", "z"};
+    const size_t v_sizes[3] = {1, 1, 1};
     rocksdb_writebatch_putv(wb, 2, k_list, k_sizes, 3, v_list, v_sizes);
     rocksdb_write(db, woptions, wb, &err);
     CheckNoError(err);
@@ -1041,13 +1047,17 @@ int main(int argc, char** argv) {
     CheckCondition(count == 3);
     size_t size;
     char* value;
-    value = rocksdb_writebatch_wi_get_from_batch(wbi, options, "box", 3, &size, &err);
+    value = rocksdb_writebatch_wi_get_from_batch(wbi, options, "box", 3, &size,
+                                                 &err);
     CheckValue(err, "c", &value, size);
-    value = rocksdb_writebatch_wi_get_from_batch(wbi, options, "bar", 3, &size, &err);
+    value = rocksdb_writebatch_wi_get_from_batch(wbi, options, "bar", 3, &size,
+                                                 &err);
     CheckValue(err, NULL, &value, size);
-    value = rocksdb_writebatch_wi_get_from_batch_and_db(wbi, db, roptions, "foo", 3, &size, &err);
+    value = rocksdb_writebatch_wi_get_from_batch_and_db(wbi, db, roptions,
+                                                        "foo", 3, &size, &err);
     CheckValue(err, "hello", &value, size);
-    value = rocksdb_writebatch_wi_get_from_batch_and_db(wbi, db, roptions, "box", 3, &size, &err);
+    value = rocksdb_writebatch_wi_get_from_batch_and_db(wbi, db, roptions,
+                                                        "box", 3, &size, &err);
     CheckValue(err, "c", &value, size);
     rocksdb_write_writebatch_wi(db, woptions, wbi, &err);
     CheckNoError(err);
@@ -1064,10 +1074,10 @@ int main(int argc, char** argv) {
   StartPhase("writebatch_wi_vectors");
   {
     rocksdb_writebatch_wi_t* wb = rocksdb_writebatch_wi_create(0, 1);
-    const char* k_list[2] = { "z", "ap" };
-    const size_t k_sizes[2] = { 1, 2 };
-    const char* v_list[3] = { "x", "y", "z" };
-    const size_t v_sizes[3] = { 1, 1, 1 };
+    const char* k_list[2] = {"z", "ap"};
+    const size_t k_sizes[2] = {1, 2};
+    const char* v_list[3] = {"x", "y", "z"};
+    const size_t v_sizes[3] = {1, 1, 1};
     rocksdb_writebatch_wi_putv(wb, 2, k_list, k_sizes, 3, v_list, v_sizes);
     rocksdb_write_writebatch_wi(db, woptions, wb, &err);
     CheckNoError(err);
@@ -1156,13 +1166,14 @@ int main(int argc, char** argv) {
 
   StartPhase("multiget");
   {
-    const char* keys[3] = { "box", "foo", "notfound" };
-    const size_t keys_sizes[3] = { 3, 3, 8 };
+    const char* keys[3] = {"box", "foo", "notfound"};
+    const size_t keys_sizes[3] = {3, 3, 8};
     char* vals[3];
     size_t vals_sizes[3];
     char* errs[3];
     const char* expected[3] = {"c", "hello", NULL};
-    rocksdb_multi_get(db, roptions, 3, keys, keys_sizes, vals, vals_sizes, errs);
+    rocksdb_multi_get(db, roptions, 3, keys, keys_sizes, vals, vals_sizes,
+                      errs);
     CheckMultiGetValues(3, vals, vals_sizes, errs, expected);
   }
 
@@ -1180,10 +1191,10 @@ int main(int argc, char** argv) {
     char keybuf[100];
     char valbuf[100];
     uint64_t sizes[2];
-    const char* start[2] = { "a", "k00000000000000010000" };
-    size_t start_len[2] = { 1, 21 };
-    const char* limit[2] = { "k00000000000000010000", "z" };
-    size_t limit_len[2] = { 21, 1 };
+    const char* start[2] = {"a", "k00000000000000010000"};
+    size_t start_len[2] = {1, 21};
+    const char* limit[2] = {"k00000000000000010000", "z"};
+    size_t limit_len[2] = {21, 1};
     rocksdb_writeoptions_set_sync(woptions, 0);
     for (i = 0; i < n; i++) {
       snprintf(keybuf, sizeof(keybuf), "k%020d", i);
@@ -1393,8 +1404,8 @@ int main(int argc, char** argv) {
                                                   factory);
     db = CheckCompaction(db, options_with_filter_factory, roptions, woptions);
 
-    rocksdb_options_set_compaction_filter_factory(
-        options_with_filter_factory, NULL);
+    rocksdb_options_set_compaction_filter_factory(options_with_filter_factory,
+                                                  NULL);
     rocksdb_options_destroy(options_with_filter_factory);
   }
 
@@ -1449,7 +1460,8 @@ int main(int argc, char** argv) {
     rocksdb_close(db);
 
     size_t cflen;
-    char** column_fams = rocksdb_list_column_families(db_options, dbname, &cflen, &err);
+    char** column_fams =
+        rocksdb_list_column_families(db_options, dbname, &cflen, &err);
     CheckNoError(err);
     CheckEqual("default", column_fams[0], 7);
     CheckEqual("cf1", column_fams[1], 3);
@@ -1465,7 +1477,8 @@ int main(int argc, char** argv) {
     LoadAndCheckLatestOptions(dbname, env, false, cache, NULL, 2, cf_names,
                               NULL);
 
-    db = rocksdb_open_column_families(db_options, dbname, 2, cf_names, cf_opts, handles, &err);
+    db = rocksdb_open_column_families(db_options, dbname, 2, cf_names, cf_opts,
+                                      handles, &err);
     CheckNoError(err);
 
     rocksdb_put_cf(db, woptions, handles[1], "foo", 3, "hello", 5, &err);
@@ -1483,11 +1496,10 @@ int main(int argc, char** argv) {
                                      &err);
     CheckNoError(err);
 
-    rocksdb_flushoptions_t *flush_options = rocksdb_flushoptions_create();
+    rocksdb_flushoptions_t* flush_options = rocksdb_flushoptions_create();
     rocksdb_flushoptions_set_wait(flush_options, 1);
     rocksdb_flush_cf(db, flush_options, handles[1], &err);
-    CheckNoError(err)
-    rocksdb_flushoptions_destroy(flush_options);
+    CheckNoError(err) rocksdb_flushoptions_destroy(flush_options);
 
     CheckGetCF(db, roptions, handles[1], "foo", "hello");
     CheckPinGetCF(db, roptions, handles[1], "foo", "hello");
@@ -1524,27 +1536,29 @@ int main(int argc, char** argv) {
     rocksdb_flush_wal(db, 1, &err);
     CheckNoError(err);
 
-    const char* keys[3] = { "box", "box", "barfooxx" };
-    const rocksdb_column_family_handle_t* get_handles[3] = { handles[0], handles[1], handles[1] };
-    const size_t keys_sizes[3] = { 3, 3, 8 };
+    const char* keys[3] = {"box", "box", "barfooxx"};
+    const rocksdb_column_family_handle_t* get_handles[3] = {
+        handles[0], handles[1], handles[1]};
+    const size_t keys_sizes[3] = {3, 3, 8};
     char* vals[3];
     size_t vals_sizes[3];
     char* errs[3];
-    rocksdb_multi_get_cf(db, roptions, get_handles, 3, keys, keys_sizes, vals, vals_sizes, errs);
+    rocksdb_multi_get_cf(db, roptions, get_handles, 3, keys, keys_sizes, vals,
+                         vals_sizes, errs);
 
     int i;
     for (i = 0; i < 3; i++) {
       CheckEqual(NULL, errs[i], 0);
       switch (i) {
-      case 0:
-        CheckEqual(NULL, vals[i], vals_sizes[i]); // wrong cf
-        break;
-      case 1:
-        CheckEqual("c", vals[i], vals_sizes[i]); // bingo
-        break;
-      case 2:
-        CheckEqual(NULL, vals[i], vals_sizes[i]); // normal not found
-        break;
+        case 0:
+          CheckEqual(NULL, vals[i], vals_sizes[i]);  // wrong cf
+          break;
+        case 1:
+          CheckEqual("c", vals[i], vals_sizes[i]);  // bingo
+          break;
+        case 2:
+          CheckEqual(NULL, vals[i], vals_sizes[i]);  // normal not found
+          break;
       }
       Free(&vals[i]);
     }
@@ -1592,7 +1606,8 @@ int main(int argc, char** argv) {
       }
     }
 
-    rocksdb_iterator_t* iter = rocksdb_create_iterator_cf(db, roptions, handles[1]);
+    rocksdb_iterator_t* iter =
+        rocksdb_create_iterator_cf(db, roptions, handles[1]);
     CheckCondition(!rocksdb_iter_valid(iter));
     rocksdb_iter_seek_to_first(iter);
     CheckCondition(rocksdb_iter_valid(iter));
@@ -1605,9 +1620,11 @@ int main(int argc, char** argv) {
     CheckNoError(err);
     rocksdb_iter_destroy(iter);
 
-    rocksdb_column_family_handle_t* iters_cf_handles[2] = { handles[0], handles[1] };
+    rocksdb_column_family_handle_t* iters_cf_handles[2] = {handles[0],
+                                                           handles[1]};
     rocksdb_iterator_t* iters_handles[2];
-    rocksdb_create_iterators(db, roptions, iters_cf_handles, iters_handles, 2, &err);
+    rocksdb_create_iterators(db, roptions, iters_cf_handles, iters_handles, 2,
+                             &err);
     CheckNoError(err);
 
     iter = iters_handles[0];
@@ -1652,7 +1669,8 @@ int main(int argc, char** argv) {
   {
     // Create new database
     rocksdb_options_set_allow_mmap_reads(options, 1);
-    rocksdb_options_set_prefix_extractor(options, rocksdb_slicetransform_create_fixed_prefix(3));
+    rocksdb_options_set_prefix_extractor(
+        options, rocksdb_slicetransform_create_fixed_prefix(3));
     rocksdb_options_set_hash_skip_list_rep(options, 5000, 4, 4);
     rocksdb_options_set_plain_table_factory(options, 4, 10, 0.75, 16);
     rocksdb_options_set_allow_concurrent_memtable_write(options, 0);
@@ -1747,8 +1765,9 @@ int main(int argc, char** argv) {
     // amount of memory used within memtables should grow
     CheckCondition(rocksdb_approximate_memory_usage_get_mem_table_total(mu2) >=
                    rocksdb_approximate_memory_usage_get_mem_table_total(mu1));
-    CheckCondition(rocksdb_approximate_memory_usage_get_mem_table_unflushed(mu2) >=
-                   rocksdb_approximate_memory_usage_get_mem_table_unflushed(mu1));
+    CheckCondition(
+        rocksdb_approximate_memory_usage_get_mem_table_unflushed(mu2) >=
+        rocksdb_approximate_memory_usage_get_mem_table_unflushed(mu1));
 
     rocksdb_memory_consumers_destroy(consumers);
     rocksdb_approximate_memory_usage_destroy(mu1);
@@ -2578,6 +2597,9 @@ int main(int argc, char** argv) {
     rocksdb_readoptions_set_io_timeout(ro, 400);
     CheckCondition(400 == rocksdb_readoptions_get_io_timeout(ro));
 
+    rocksdb_readoptions_set_async_io(ro, 1);
+    CheckCondition(1 == rocksdb_readoptions_get_async_io(ro));
+
     rocksdb_readoptions_destroy(ro);
   }
 
@@ -2839,53 +2861,57 @@ int main(int argc, char** argv) {
     db = rocksdb_open(options, dbname, &err);
     CheckNoError(err);
 
-    rocksdb_put(db, woptions, "a",    1, "0",    1, &err); CheckNoError(err);
-    rocksdb_put(db, woptions, "foo",  3, "bar",  3, &err); CheckNoError(err);
-    rocksdb_put(db, woptions, "foo1", 4, "bar1", 4, &err); CheckNoError(err);
-    rocksdb_put(db, woptions, "g1",   2, "0",    1, &err); CheckNoError(err);
+    rocksdb_put(db, woptions, "a", 1, "0", 1, &err);
+    CheckNoError(err);
+    rocksdb_put(db, woptions, "foo", 3, "bar", 3, &err);
+    CheckNoError(err);
+    rocksdb_put(db, woptions, "foo1", 4, "bar1", 4, &err);
+    CheckNoError(err);
+    rocksdb_put(db, woptions, "g1", 2, "0", 1, &err);
+    CheckNoError(err);
 
     // testing basic case with no iterate_upper_bound and no prefix_extractor
     {
-       rocksdb_readoptions_set_iterate_upper_bound(roptions, NULL, 0);
-       rocksdb_iterator_t* iter = rocksdb_create_iterator(db, roptions);
+      rocksdb_readoptions_set_iterate_upper_bound(roptions, NULL, 0);
+      rocksdb_iterator_t* iter = rocksdb_create_iterator(db, roptions);
 
-       rocksdb_iter_seek(iter, "foo", 3);
-       CheckCondition(rocksdb_iter_valid(iter));
-       CheckIter(iter, "foo", "bar");
+      rocksdb_iter_seek(iter, "foo", 3);
+      CheckCondition(rocksdb_iter_valid(iter));
+      CheckIter(iter, "foo", "bar");
 
-       rocksdb_iter_next(iter);
-       CheckCondition(rocksdb_iter_valid(iter));
-       CheckIter(iter, "foo1", "bar1");
+      rocksdb_iter_next(iter);
+      CheckCondition(rocksdb_iter_valid(iter));
+      CheckIter(iter, "foo1", "bar1");
 
-       rocksdb_iter_next(iter);
-       CheckCondition(rocksdb_iter_valid(iter));
-       CheckIter(iter, "g1", "0");
+      rocksdb_iter_next(iter);
+      CheckCondition(rocksdb_iter_valid(iter));
+      CheckIter(iter, "g1", "0");
 
-       rocksdb_iter_destroy(iter);
+      rocksdb_iter_destroy(iter);
     }
 
     // testing iterate_upper_bound and forward iterator
     // to make sure it stops at bound
     {
-       // iterate_upper_bound points beyond the last expected entry
-       rocksdb_readoptions_set_iterate_upper_bound(roptions, "foo2", 4);
+      // iterate_upper_bound points beyond the last expected entry
+      rocksdb_readoptions_set_iterate_upper_bound(roptions, "foo2", 4);
 
-       rocksdb_iterator_t* iter = rocksdb_create_iterator(db, roptions);
+      rocksdb_iterator_t* iter = rocksdb_create_iterator(db, roptions);
 
-       rocksdb_iter_seek(iter, "foo", 3);
-       CheckCondition(rocksdb_iter_valid(iter));
-       CheckIter(iter, "foo", "bar");
+      rocksdb_iter_seek(iter, "foo", 3);
+      CheckCondition(rocksdb_iter_valid(iter));
+      CheckIter(iter, "foo", "bar");
 
-       rocksdb_iter_next(iter);
-       CheckCondition(rocksdb_iter_valid(iter));
-       CheckIter(iter, "foo1", "bar1");
+      rocksdb_iter_next(iter);
+      CheckCondition(rocksdb_iter_valid(iter));
+      CheckIter(iter, "foo1", "bar1");
 
-       rocksdb_iter_next(iter);
-       // should stop here...
-       CheckCondition(!rocksdb_iter_valid(iter));
+      rocksdb_iter_next(iter);
+      // should stop here...
+      CheckCondition(!rocksdb_iter_valid(iter));
 
-       rocksdb_iter_destroy(iter);
-       rocksdb_readoptions_set_iterate_upper_bound(roptions, NULL, 0);
+      rocksdb_iter_destroy(iter);
+      rocksdb_readoptions_set_iterate_upper_bound(roptions, NULL, 0);
     }
   }
 
@@ -3009,7 +3035,7 @@ int main(int argc, char** argv) {
     snapshot = rocksdb_transactiondb_create_snapshot(txn_db);
     rocksdb_readoptions_set_snapshot(roptions, snapshot);
 
-    rocksdb_transactiondb_put(txn_db, woptions, "foo", 3, "hey", 3,  &err);
+    rocksdb_transactiondb_put(txn_db, woptions, "foo", 3, "hey", 3, &err);
     CheckNoError(err);
 
     CheckTxnDBGet(txn_db, roptions, "foo", "hello");
@@ -3021,7 +3047,8 @@ int main(int argc, char** argv) {
 
     // iterate
     rocksdb_transaction_put(txn, "bar", 3, "hi", 2, &err);
-    rocksdb_iterator_t* iter = rocksdb_transaction_create_iterator(txn, roptions);
+    rocksdb_iterator_t* iter =
+        rocksdb_transaction_create_iterator(txn, roptions);
     CheckCondition(!rocksdb_iter_valid(iter));
     rocksdb_iter_seek_to_first(iter);
     CheckCondition(rocksdb_iter_valid(iter));
@@ -3204,6 +3231,120 @@ int main(int argc, char** argv) {
     rocksdb_transactiondb_options_destroy(txn_db_options);
   }
 
+  StartPhase("transactions_multi_get_for_update");
+  {
+    // open a TransactionDB
+    txn_db_options = rocksdb_transactiondb_options_create();
+    rocksdb_transactiondb_options_set_transaction_lock_timeout(txn_db_options,
+                                                               0);
+    txn_options = rocksdb_transaction_options_create();
+    rocksdb_options_set_create_if_missing(options, 1);
+    txn_db = rocksdb_transactiondb_open(options, txn_db_options, dbname, &err);
+    CheckNoError(err);
+
+    rocksdb_transactiondb_put(txn_db, woptions, "foo", 3, "hey", 3, &err);
+    CheckNoError(err);
+    rocksdb_transactiondb_put(txn_db, woptions, "bar", 3, "hello", 5, &err);
+    CheckNoError(err);
+
+    // begin transactions
+    txn = rocksdb_transaction_begin(txn_db, woptions, txn_options, NULL);
+    rocksdb_transaction_t* txn2 =
+        rocksdb_transaction_begin(txn_db, woptions, txn_options, NULL);
+
+    // multi get
+    {
+      const char* keys[2] = {"foo", "bar"};
+      const size_t keys_sizes[2] = {3, 3};
+      char* vals[2];
+      size_t vals_sizes[2];
+      char* errs[2];
+      const char* expected[2] = {"hey", "hello"};
+      rocksdb_transaction_multi_get_for_update(
+          txn, roptions, 2, keys, keys_sizes, vals, vals_sizes, errs);
+      CheckMultiGetValues(2, vals, vals_sizes, errs, expected);
+    }
+
+    char* conflict_err = NULL;
+    size_t val_len;
+    rocksdb_transaction_get_for_update(txn2, roptions, "foo", 3, &val_len, true,
+                                       &conflict_err);
+    // get-for-update conflict
+    CheckCondition(conflict_err != NULL);
+    Free(&conflict_err);
+
+    // commit
+    rocksdb_transaction_commit(txn, &err);
+    CheckNoError(err);
+
+    // should work after first tx is commited
+    CheckTxnGetForUpdate(txn2, roptions, "foo", "hey");
+
+    // commit the second one
+    rocksdb_transaction_commit(txn2, &err);
+    CheckNoError(err);
+
+    // destroy txns
+    rocksdb_transaction_destroy(txn);
+    rocksdb_transaction_destroy(txn2);
+
+    // same for column families
+
+    rocksdb_column_family_handle_t* cfh;
+    cfh = rocksdb_transactiondb_create_column_family(txn_db, options,
+                                                     "txn_db_cf", &err);
+    CheckNoError(err);
+
+    rocksdb_transactiondb_put_cf(txn_db, woptions, cfh, "cf_foo", 6, "cf_hello",
+                                 8, &err);
+    CheckNoError(err);
+    rocksdb_transactiondb_put_cf(txn_db, woptions, cfh, "cf_bar", 6, "cf_hey",
+                                 6, &err);
+    CheckNoError(err);
+
+    txn = rocksdb_transaction_begin(txn_db, woptions, txn_options, NULL);
+    txn2 = rocksdb_transaction_begin(txn_db, woptions, txn_options, NULL);
+
+    {
+      const rocksdb_column_family_handle_t* get_handles[2] = {cfh, cfh};
+      const char* keys[2] = {"cf_foo", "cf_bar"};
+      const size_t keys_sizes[2] = {6, 6};
+      char* vals[2];
+      size_t vals_sizes[2];
+      char* errs[2];
+      const char* expected[2] = {"cf_hello", "cf_hey"};
+      rocksdb_transaction_multi_get_for_update_cf(txn, roptions, get_handles, 2,
+                                                  keys, keys_sizes, vals,
+                                                  vals_sizes, errs);
+      CheckMultiGetValues(2, vals, vals_sizes, errs, expected);
+    }
+
+    char* conflict_err_cf = NULL;
+    size_t val_len_cf;
+    rocksdb_transaction_get_for_update_cf(txn2, roptions, cfh, "cf_foo", 6,
+                                          &val_len_cf, true, &conflict_err_cf);
+    CheckCondition(conflict_err_cf != NULL);
+    Free(&conflict_err_cf);
+
+    rocksdb_transaction_commit(txn, &err);
+    CheckNoError(err);
+
+    CheckTxnGetForUpdateCF(txn2, roptions, cfh, "cf_foo", "cf_hello");
+
+    rocksdb_transaction_commit(txn2, &err);
+    CheckNoError(err);
+
+    // close and destroy
+    rocksdb_column_family_handle_destroy(cfh);
+    rocksdb_transaction_destroy(txn);
+    rocksdb_transaction_destroy(txn2);
+    rocksdb_transactiondb_close(txn_db);
+    rocksdb_destroy_db(options, dbname, &err);
+    CheckNoError(err);
+    rocksdb_transaction_options_destroy(txn_options);
+    rocksdb_transactiondb_options_destroy(txn_db_options);
+  }
+
   StartPhase("optimistic_transactions");
   {
     rocksdb_options_t* db_options = rocksdb_options_create();
@@ -3466,12 +3607,3 @@ int main(int argc, char** argv) {
   fprintf(stderr, "PASS\n");
   return 0;
 }
-
-#else
-
-int main(void) {
-  fprintf(stderr, "SKIPPED\n");
-  return 0;
-}
-
-#endif  // !ROCKSDB_LITE

db/column_family.cc

@@ -53,11 +53,9 @@ ColumnFamilyHandleImpl::ColumnFamilyHandleImpl(
 
 ColumnFamilyHandleImpl::~ColumnFamilyHandleImpl() {
   if (cfd_ != nullptr) {
-#ifndef ROCKSDB_LITE
     for (auto& listener : cfd_->ioptions()->listeners) {
       listener->OnColumnFamilyHandleDeletionStarted(this);
     }
-#endif  // ROCKSDB_LITE
     // Job id == 0 means that this is not our background process, but rather
     // user thread
     // Need to hold some shared pointers owned by the initial_cf_options
@@ -88,15 +86,10 @@ const std::string& ColumnFamilyHandleImpl::GetName() const {
 }
 
 Status ColumnFamilyHandleImpl::GetDescriptor(ColumnFamilyDescriptor* desc) {
-#ifndef ROCKSDB_LITE
   // accessing mutable cf-options requires db mutex.
   InstrumentedMutexLock l(mutex_);
   *desc = ColumnFamilyDescriptor(cfd()->GetName(), cfd()->GetLatestCFOptions());
   return Status::OK();
-#else
-  (void)desc;
-  return Status::NotSupported();
-#endif  // !ROCKSDB_LITE
 }
 
 const Comparator* ColumnFamilyHandleImpl::GetComparator() const {
@@ -192,8 +185,7 @@ Status CheckCFPathsSupported(const DBOptions& db_options,
       return Status::NotSupported(
           "More than one CF paths are only supported in "
           "universal and level compaction styles. ");
-    } else if (cf_options.cf_paths.empty() &&
-               db_options.db_paths.size() > 1) {
+    } else if (cf_options.cf_paths.empty() && db_options.db_paths.size() > 1) {
       return Status::NotSupported(
           "More than one DB paths are only supported in "
           "universal and level compaction styles. ");
@@ -205,7 +197,7 @@ Status CheckCFPathsSupported(const DBOptions& db_options,
 namespace {
 const uint64_t kDefaultTtl = 0xfffffffffffffffe;
 const uint64_t kDefaultPeriodicCompSecs = 0xfffffffffffffffe;
-}  // namespace
+}  // anonymous namespace
 
 ColumnFamilyOptions SanitizeOptions(const ImmutableDBOptions& db_options,
                                     const ColumnFamilyOptions& src) {
@@ -348,18 +340,17 @@ ColumnFamilyOptions SanitizeOptions(const ImmutableDBOptions& db_options,
         result.hard_pending_compaction_bytes_limit;
   }
 
-#ifndef ROCKSDB_LITE
   // When the DB is stopped, it's possible that there are some .trash files that
   // were not deleted yet, when we open the DB we will find these .trash files
   // and schedule them to be deleted (or delete immediately if SstFileManager
   // was not used)
-  auto sfm = static_cast<SstFileManagerImpl*>(db_options.sst_file_manager.get());
+  auto sfm =
+      static_cast<SstFileManagerImpl*>(db_options.sst_file_manager.get());
   for (size_t i = 0; i < result.cf_paths.size(); i++) {
     DeleteScheduler::CleanupDirectory(db_options.env, sfm,
                                       result.cf_paths[i].path)
         .PermitUncheckedError();
   }
-#endif
 
   if (result.cf_paths.empty()) {
     result.cf_paths = db_options.db_paths;
@@ -557,7 +548,6 @@ ColumnFamilyData::ColumnFamilyData(
       next_(nullptr),
       prev_(nullptr),
       log_number_(0),
-      flush_reason_(FlushReason::kOthers),
       column_family_set_(column_family_set),
       queued_for_flush_(false),
       queued_for_compaction_(false),
@@ -565,7 +555,8 @@ ColumnFamilyData::ColumnFamilyData(
       allow_2pc_(db_options.allow_2pc),
       last_memtable_id_(0),
       db_paths_registered_(false),
-      mempurge_used_(false) {
+      mempurge_used_(false),
+      next_epoch_number_(1) {
   if (id_ != kDummyColumnFamilyDataId) {
     // TODO(cc): RegisterDbPaths can be expensive, considering moving it
     // outside of this constructor which might be called with db mutex held.
@@ -602,7 +593,6 @@ ColumnFamilyData::ColumnFamilyData(
     if (ioptions_.compaction_style == kCompactionStyleLevel) {
       compaction_picker_.reset(
           new LevelCompactionPicker(ioptions_, &internal_comparator_));
-#ifndef ROCKSDB_LITE
     } else if (ioptions_.compaction_style == kCompactionStyleUniversal) {
       compaction_picker_.reset(
           new UniversalCompactionPicker(ioptions_, &internal_comparator_));
@@ -610,13 +600,12 @@ ColumnFamilyData::ColumnFamilyData(
       compaction_picker_.reset(
           new FIFOCompactionPicker(ioptions_, &internal_comparator_));
     } else if (ioptions_.compaction_style == kCompactionStyleNone) {
-      compaction_picker_.reset(new NullCompactionPicker(
-          ioptions_, &internal_comparator_));
+      compaction_picker_.reset(
+          new NullCompactionPicker(ioptions_, &internal_comparator_));
       ROCKS_LOG_WARN(ioptions_.logger,
                      "Column family %s does not use any background compaction. "
                      "Compactions can only be done via CompactFiles\n",
                      GetName().c_str());
-#endif  // !ROCKSDB_LITE
     } else {
       ROCKS_LOG_ERROR(ioptions_.logger,
                       "Unable to recognize the specified compaction style %d. "
@@ -878,9 +867,9 @@ int GetL0ThresholdSpeedupCompaction(int level0_file_num_compaction_trigger,
     return static_cast<int>(res);
   }
 }
-}  // namespace
+}  // anonymous namespace
 
-std::pair<WriteStallCondition, ColumnFamilyData::WriteStallCause>
+std::pair<WriteStallCondition, WriteStallCause>
 ColumnFamilyData::GetWriteStallConditionAndCause(
     int num_unflushed_memtables, int num_l0_files,
     uint64_t num_compaction_needed_bytes,
@@ -919,7 +908,7 @@ ColumnFamilyData::GetWriteStallConditionAndCause(
 }
 
 WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
-      const MutableCFOptions& mutable_cf_options) {
+    const MutableCFOptions& mutable_cf_options) {
   auto write_stall_condition = WriteStallCondition::kNormal;
   if (current_ != nullptr) {
     auto* vstorage = current_->storage_info();
@@ -953,7 +942,8 @@ WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
       internal_stats_->AddCFStats(InternalStats::L0_FILE_COUNT_LIMIT_STOPS, 1);
       if (compaction_picker_->IsLevel0CompactionInProgress()) {
         internal_stats_->AddCFStats(
-            InternalStats::LOCKED_L0_FILE_COUNT_LIMIT_STOPS, 1);
+            InternalStats::L0_FILE_COUNT_LIMIT_STOPS_WITH_ONGOING_COMPACTION,
+            1);
       }
       ROCKS_LOG_WARN(ioptions_.logger,
                      "[%s] Stopping writes because we have %d level-0 files",
@@ -974,7 +964,7 @@ WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
           SetupDelay(write_controller, compaction_needed_bytes,
                      prev_compaction_needed_bytes_, was_stopped,
                      mutable_cf_options.disable_auto_compactions);
-      internal_stats_->AddCFStats(InternalStats::MEMTABLE_LIMIT_SLOWDOWNS, 1);
+      internal_stats_->AddCFStats(InternalStats::MEMTABLE_LIMIT_DELAYS, 1);
       ROCKS_LOG_WARN(
           ioptions_.logger,
           "[%s] Stalling writes because we have %d immutable memtables "
@@ -992,11 +982,11 @@ WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
           SetupDelay(write_controller, compaction_needed_bytes,
                      prev_compaction_needed_bytes_, was_stopped || near_stop,
                      mutable_cf_options.disable_auto_compactions);
-      internal_stats_->AddCFStats(InternalStats::L0_FILE_COUNT_LIMIT_SLOWDOWNS,
-                                  1);
+      internal_stats_->AddCFStats(InternalStats::L0_FILE_COUNT_LIMIT_DELAYS, 1);
       if (compaction_picker_->IsLevel0CompactionInProgress()) {
         internal_stats_->AddCFStats(
-            InternalStats::LOCKED_L0_FILE_COUNT_LIMIT_SLOWDOWNS, 1);
+            InternalStats::L0_FILE_COUNT_LIMIT_DELAYS_WITH_ONGOING_COMPACTION,
+            1);
       }
       ROCKS_LOG_WARN(ioptions_.logger,
                      "[%s] Stalling writes because we have %d level-0 files "
@@ -1012,7 +1002,8 @@ WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
           mutable_cf_options.hard_pending_compaction_bytes_limit > 0 &&
           (compaction_needed_bytes -
            mutable_cf_options.soft_pending_compaction_bytes_limit) >
-              3 * (mutable_cf_options.hard_pending_compaction_bytes_limit -
+              3 *
+                  (mutable_cf_options.hard_pending_compaction_bytes_limit -
                    mutable_cf_options.soft_pending_compaction_bytes_limit) /
                   4;
 
@@ -1021,7 +1012,7 @@ WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
                      prev_compaction_needed_bytes_, was_stopped || near_stop,
                      mutable_cf_options.disable_auto_compactions);
       internal_stats_->AddCFStats(
-          InternalStats::PENDING_COMPACTION_BYTES_LIMIT_SLOWDOWNS, 1);
+          InternalStats::PENDING_COMPACTION_BYTES_LIMIT_DELAYS, 1);
       ROCKS_LOG_WARN(
           ioptions_.logger,
           "[%s] Stalling writes because of estimated pending compaction "
@@ -1127,12 +1118,9 @@ bool ColumnFamilyData::NeedsCompaction() const {
 Compaction* ColumnFamilyData::PickCompaction(
     const MutableCFOptions& mutable_options,
     const MutableDBOptions& mutable_db_options, LogBuffer* log_buffer) {
-  SequenceNumber earliest_mem_seqno =
-      std::min(mem_->GetEarliestSequenceNumber(),
-               imm_.current()->GetEarliestSequenceNumber(false));
   auto* result = compaction_picker_->PickCompaction(
       GetName(), mutable_options, mutable_db_options, current_->storage_info(),
-      log_buffer, earliest_mem_seqno);
+      log_buffer);
   if (result != nullptr) {
     result->SetInputVersion(current_);
   }
@@ -1153,6 +1141,7 @@ Status ColumnFamilyData::RangesOverlapWithMemtables(
   *overlap = false;
   // Create an InternalIterator over all unflushed memtables
   Arena arena;
+  // TODO: plumb Env::IOActivity
   ReadOptions read_opts;
   read_opts.total_order_seek = true;
   MergeIteratorBuilder merge_iter_builder(&internal_comparator_, &arena);
@@ -1219,6 +1208,7 @@ Compaction* ColumnFamilyData::CompactRange(
   if (result != nullptr) {
     result->SetInputVersion(current_);
   }
+  TEST_SYNC_POINT("ColumnFamilyData::CompactRange:Return");
   return result;
 }
 
@@ -1302,8 +1292,8 @@ bool ColumnFamilyData::ReturnThreadLocalSuperVersion(SuperVersion* sv) {
   return false;
 }
 
-void ColumnFamilyData::InstallSuperVersion(
-    SuperVersionContext* sv_context, InstrumentedMutex* db_mutex) {
+void ColumnFamilyData::InstallSuperVersion(SuperVersionContext* sv_context,
+                                           InstrumentedMutex* db_mutex) {
   db_mutex->AssertHeld();
   return InstallSuperVersion(sv_context, mutable_cf_options_);
 }
@@ -1441,7 +1431,6 @@ Status ColumnFamilyData::ValidateOptions(
   return s;
 }
 
-#ifndef ROCKSDB_LITE
 Status ColumnFamilyData::SetOptions(
     const DBOptions& db_opts,
     const std::unordered_map<std::string, std::string>& options_map) {
@@ -1460,7 +1449,6 @@ Status ColumnFamilyData::SetOptions(
   }
   return s;
 }
-#endif  // ROCKSDB_LITE
 
 // REQUIRES: DB mutex held
 Env::WriteLifeTimeHint ColumnFamilyData::CalculateSSTWriteHint(int level) {
@@ -1480,8 +1468,8 @@ Env::WriteLifeTimeHint ColumnFamilyData::CalculateSSTWriteHint(int level) {
     // than base_level.
     return Env::WLTH_MEDIUM;
   }
-  return static_cast<Env::WriteLifeTimeHint>(level - base_level +
-                            static_cast<int>(Env::WLTH_MEDIUM));
+  return static_cast<Env::WriteLifeTimeHint>(
+      level - base_level + static_cast<int>(Env::WLTH_MEDIUM));
 }
 
 Status ColumnFamilyData::AddDirectories(
@@ -1519,6 +1507,13 @@ FSDirectory* ColumnFamilyData::GetDataDir(size_t path_id) const {
   return data_dirs_[path_id].get();
 }
 
+void ColumnFamilyData::RecoverEpochNumbers() {
+  assert(current_);
+  auto* vstorage = current_->storage_info();
+  assert(vstorage);
+  vstorage->RecoverEpochNumbers(this);
+}
+
 ColumnFamilySet::ColumnFamilySet(const std::string& dbname,
                                  const ImmutableDBOptions* db_options,
                                  const FileOptions& file_options,
@@ -1577,8 +1572,8 @@ ColumnFamilyData* ColumnFamilySet::GetColumnFamily(uint32_t id) const {
   }
 }
 
-ColumnFamilyData* ColumnFamilySet::GetColumnFamily(const std::string& name)
-    const {
+ColumnFamilyData* ColumnFamilySet::GetColumnFamily(
+    const std::string& name) const {
   auto cfd_iter = column_families_.find(name);
   if (cfd_iter != column_families_.end()) {
     auto cfd = GetColumnFamily(cfd_iter->second);

db/column_family.h

@@ -163,8 +163,8 @@ extern const double kIncSlowdownRatio;
 class ColumnFamilyHandleImpl : public ColumnFamilyHandle {
  public:
   // create while holding the mutex
-  ColumnFamilyHandleImpl(
-      ColumnFamilyData* cfd, DBImpl* db, InstrumentedMutex* mutex);
+  ColumnFamilyHandleImpl(ColumnFamilyData* cfd, DBImpl* db,
+                         InstrumentedMutex* mutex);
   // destroy without mutex
   virtual ~ColumnFamilyHandleImpl();
   virtual ColumnFamilyData* cfd() const { return cfd_; }
@@ -189,7 +189,8 @@ class ColumnFamilyHandleImpl : public ColumnFamilyHandle {
 class ColumnFamilyHandleInternal : public ColumnFamilyHandleImpl {
  public:
   ColumnFamilyHandleInternal()
-      : ColumnFamilyHandleImpl(nullptr, nullptr, nullptr), internal_cfd_(nullptr) {}
+      : ColumnFamilyHandleImpl(nullptr, nullptr, nullptr),
+        internal_cfd_(nullptr) {}
 
   void SetCFD(ColumnFamilyData* _cfd) { internal_cfd_ = _cfd; }
   virtual ColumnFamilyData* cfd() const override { return internal_cfd_; }
@@ -309,10 +310,6 @@ class ColumnFamilyData {
   void SetLogNumber(uint64_t log_number) { log_number_ = log_number; }
   uint64_t GetLogNumber() const { return log_number_; }
 
-  void SetFlushReason(FlushReason flush_reason) {
-    flush_reason_ = flush_reason;
-  }
-  FlushReason GetFlushReason() const { return flush_reason_; }
   // thread-safe
   const FileOptions* soptions() const;
   const ImmutableOptions* ioptions() const { return &ioptions_; }
@@ -338,12 +335,10 @@ class ColumnFamilyData {
   // Validate CF options against DB options
   static Status ValidateOptions(const DBOptions& db_options,
                                 const ColumnFamilyOptions& cf_options);
-#ifndef ROCKSDB_LITE
   // REQUIRES: DB mutex held
   Status SetOptions(
       const DBOptions& db_options,
       const std::unordered_map<std::string, std::string>& options_map);
-#endif  // ROCKSDB_LITE
 
   InternalStats* internal_stats() { return internal_stats_.get(); }
 
@@ -357,7 +352,7 @@ class ColumnFamilyData {
   Version* current() { return current_; }
   Version* dummy_versions() { return dummy_versions_; }
   void SetCurrent(Version* _current);
-  uint64_t GetNumLiveVersions() const;  // REQUIRE: DB mutex held
+  uint64_t GetNumLiveVersions() const;    // REQUIRE: DB mutex held
   uint64_t GetTotalSstFilesSize() const;  // REQUIRE: DB mutex held
   uint64_t GetLiveSstFilesSize() const;   // REQUIRE: DB mutex held
   uint64_t GetTotalBlobFileSize() const;  // REQUIRE: DB mutex held
@@ -467,12 +462,6 @@ class ColumnFamilyData {
   bool queued_for_flush() { return queued_for_flush_; }
   bool queued_for_compaction() { return queued_for_compaction_; }
 
-  enum class WriteStallCause {
-    kNone,
-    kMemtableLimit,
-    kL0FileCountLimit,
-    kPendingCompactionBytes,
-  };
   static std::pair<WriteStallCondition, WriteStallCause>
   GetWriteStallConditionAndCause(
       int num_unflushed_memtables, int num_l0_files,
@@ -532,6 +521,24 @@ class ColumnFamilyData {
   void SetMempurgeUsed() { mempurge_used_ = true; }
   bool GetMempurgeUsed() { return mempurge_used_; }
 
+  // Allocate and return a new epoch number
+  uint64_t NewEpochNumber() { return next_epoch_number_.fetch_add(1); }
+
+  // Get the next epoch number to be assigned
+  uint64_t GetNextEpochNumber() const { return next_epoch_number_.load(); }
+
+  // Set the next epoch number to be assigned
+  void SetNextEpochNumber(uint64_t next_epoch_number) {
+    next_epoch_number_.store(next_epoch_number);
+  }
+
+  // Reset the next epoch number to be assigned
+  void ResetNextEpochNumber() { next_epoch_number_.store(1); }
+
+  // Recover the next epoch number of this CF and epoch number
+  // of its files (if missing)
+  void RecoverEpochNumbers();
+
  private:
   friend class ColumnFamilySet;
   ColumnFamilyData(uint32_t id, const std::string& name,
@@ -552,7 +559,7 @@ class ColumnFamilyData {
   Version* dummy_versions_;  // Head of circular doubly-linked list of versions.
   Version* current_;         // == dummy_versions->prev_
 
-  std::atomic<int> refs_;      // outstanding references to ColumnFamilyData
+  std::atomic<int> refs_;  // outstanding references to ColumnFamilyData
   std::atomic<bool> initialized_;
   std::atomic<bool> dropped_;  // true if client dropped it
 
@@ -597,8 +604,6 @@ class ColumnFamilyData {
   // recovered from
   uint64_t log_number_;
 
-  std::atomic<FlushReason> flush_reason_;
-
   // An object that keeps all the compaction stats
   // and picks the next compaction
   std::unique_ptr<CompactionPicker> compaction_picker_;
@@ -633,6 +638,8 @@ class ColumnFamilyData {
   // a Version associated with this CFD
   std::shared_ptr<CacheReservationManager> file_metadata_cache_res_mgr_;
   bool mempurge_used_;
+
+  std::atomic<uint64_t> next_epoch_number_;
 };
 
 // ColumnFamilySet has interesting thread-safety requirements
@@ -656,8 +663,7 @@ class ColumnFamilySet {
   // ColumnFamilySet supports iteration
   class iterator {
    public:
-    explicit iterator(ColumnFamilyData* cfd)
-        : current_(cfd) {}
+    explicit iterator(ColumnFamilyData* cfd) : current_(cfd) {}
     // NOTE: minimum operators for for-loop iteration
     iterator& operator++() {
       current_ = current_->next_;

db/column_family_test.cc

@@ -39,9 +39,7 @@ class EnvCounter : public SpecialEnv {
  public:
   explicit EnvCounter(Env* base)
       : SpecialEnv(base), num_new_writable_file_(0) {}
-  int GetNumberOfNewWritableFileCalls() {
-    return num_new_writable_file_;
-  }
+  int GetNumberOfNewWritableFileCalls() { return num_new_writable_file_; }
   Status NewWritableFile(const std::string& f, std::unique_ptr<WritableFile>* r,
                          const EnvOptions& soptions) override {
     ++num_new_writable_file_;
@@ -73,11 +71,7 @@ class ColumnFamilyTestBase : public testing::Test {
     for (auto h : handles_) {
       ColumnFamilyDescriptor cfdescriptor;
       Status s = h->GetDescriptor(&cfdescriptor);
-#ifdef ROCKSDB_LITE
-      EXPECT_TRUE(s.IsNotSupported());
-#else
       EXPECT_OK(s);
-#endif  // ROCKSDB_LITE
       column_families.push_back(cfdescriptor);
     }
     ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
@@ -187,7 +181,7 @@ class ColumnFamilyTestBase : public testing::Test {
   }
 
   Status OpenReadOnly(std::vector<std::string> cf,
-                         std::vector<ColumnFamilyOptions> options = {}) {
+                      std::vector<ColumnFamilyOptions> options = {}) {
     std::vector<ColumnFamilyDescriptor> column_families;
     names_.clear();
     for (size_t i = 0; i < cf.size(); ++i) {
@@ -199,22 +193,17 @@ class ColumnFamilyTestBase : public testing::Test {
                                &db_);
   }
 
-#ifndef ROCKSDB_LITE  // ReadOnlyDB is not supported
   void AssertOpenReadOnly(std::vector<std::string> cf,
-                    std::vector<ColumnFamilyOptions> options = {}) {
+                          std::vector<ColumnFamilyOptions> options = {}) {
     ASSERT_OK(OpenReadOnly(cf, options));
   }
-#endif  // !ROCKSDB_LITE
-
 
   void Open(std::vector<std::string> cf,
             std::vector<ColumnFamilyOptions> options = {}) {
     ASSERT_OK(TryOpen(cf, options));
   }
 
-  void Open() {
-    Open({"default"});
-  }
+  void Open() { Open({"default"}); }
 
   DBImpl* dbfull() { return static_cast_with_check<DBImpl>(db_); }
 
@@ -229,31 +218,20 @@ class ColumnFamilyTestBase : public testing::Test {
   }
 
   bool IsDbWriteStopped() {
-#ifndef ROCKSDB_LITE
     uint64_t v;
     EXPECT_TRUE(dbfull()->GetIntProperty("rocksdb.is-write-stopped", &v));
     return (v == 1);
-#else
-    return dbfull()->TEST_write_controler().IsStopped();
-#endif  // !ROCKSDB_LITE
   }
 
   uint64_t GetDbDelayedWriteRate() {
-#ifndef ROCKSDB_LITE
     uint64_t v;
     EXPECT_TRUE(
         dbfull()->GetIntProperty("rocksdb.actual-delayed-write-rate", &v));
     return v;
-#else
-    if (!dbfull()->TEST_write_controler().NeedsDelay()) {
-      return 0;
-    }
-    return dbfull()->TEST_write_controler().delayed_write_rate();
-#endif  // !ROCKSDB_LITE
   }
 
   void Destroy(const std::vector<ColumnFamilyDescriptor>& column_families =
-                  std::vector<ColumnFamilyDescriptor>()) {
+                   std::vector<ColumnFamilyDescriptor>()) {
     Close();
     ASSERT_OK(DestroyDB(dbname_, Options(db_options_, column_family_options_),
                         column_families));
@@ -272,7 +250,6 @@ class ColumnFamilyTestBase : public testing::Test {
           db_->CreateColumnFamily(current_cf_opt, cfs[i], &handles_[cfi]));
       names_[cfi] = cfs[i];
 
-#ifndef ROCKSDB_LITE  // RocksDBLite does not support GetDescriptor
       // Verify the CF options of the returned CF handle.
       ColumnFamilyDescriptor desc;
       ASSERT_OK(handles_[cfi]->GetDescriptor(&desc));
@@ -281,7 +258,6 @@ class ColumnFamilyTestBase : public testing::Test {
       ASSERT_OK(RocksDBOptionsParser::VerifyCFOptions(
           ConfigOptions(), desc.options,
           SanitizeOptions(dbfull()->immutable_db_options(), current_cf_opt)));
-#endif  // !ROCKSDB_LITE
       cfi++;
     }
   }
@@ -330,14 +306,11 @@ class ColumnFamilyTestBase : public testing::Test {
     ASSERT_OK(db_->FlushWAL(/*sync=*/false));
   }
 
-#ifndef ROCKSDB_LITE  // TEST functions in DB are not supported in lite
   void WaitForFlush(int cf) {
     ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable(handles_[cf]));
   }
 
-  void WaitForCompaction() {
-    ASSERT_OK(dbfull()->TEST_WaitForCompact());
-  }
+  void WaitForCompaction() { ASSERT_OK(dbfull()->TEST_WaitForCompact()); }
 
   uint64_t MaxTotalInMemoryState() {
     return dbfull()->TEST_MaxTotalInMemoryState();
@@ -346,7 +319,6 @@ class ColumnFamilyTestBase : public testing::Test {
   void AssertMaxTotalInMemoryState(uint64_t value) {
     ASSERT_EQ(value, MaxTotalInMemoryState());
   }
-#endif  // !ROCKSDB_LITE
 
   Status Put(int cf, const std::string& key, const std::string& value) {
     return db_->Put(WriteOptions(), handles_[cf], Slice(key), Slice(value));
@@ -354,9 +326,7 @@ class ColumnFamilyTestBase : public testing::Test {
   Status Merge(int cf, const std::string& key, const std::string& value) {
     return db_->Merge(WriteOptions(), handles_[cf], Slice(key), Slice(value));
   }
-  Status Flush(int cf) {
-    return db_->Flush(FlushOptions(), handles_[cf]);
-  }
+  Status Flush(int cf) { return db_->Flush(FlushOptions(), handles_[cf]); }
 
   std::string Get(int cf, const std::string& key) {
     ReadOptions options;
@@ -386,7 +356,6 @@ class ColumnFamilyTestBase : public testing::Test {
                        "rocksdb.num-files-at-level" + std::to_string(level));
   }
 
-#ifndef ROCKSDB_LITE
   // Return spread of files per level
   std::string FilesPerLevel(int cf) {
     std::string result;
@@ -403,31 +372,19 @@ class ColumnFamilyTestBase : public testing::Test {
     result.resize(last_non_zero_offset);
     return result;
   }
-#endif
 
   void AssertFilesPerLevel(const std::string& value, int cf) {
-#ifndef ROCKSDB_LITE
     ASSERT_EQ(value, FilesPerLevel(cf));
-#else
-    (void) value;
-    (void) cf;
-#endif
   }
 
-#ifndef ROCKSDB_LITE  // GetLiveFilesMetaData is not supported
   int CountLiveFiles() {
     std::vector<LiveFileMetaData> metadata;
     db_->GetLiveFilesMetaData(&metadata);
     return static_cast<int>(metadata.size());
   }
-#endif  // !ROCKSDB_LITE
 
   void AssertCountLiveFiles(int expected_value) {
-#ifndef ROCKSDB_LITE
     ASSERT_EQ(expected_value, CountLiveFiles());
-#else
-    (void) expected_value;
-#endif
   }
 
   // Do n memtable flushes, each of which produces an sstable
@@ -441,7 +398,6 @@ class ColumnFamilyTestBase : public testing::Test {
     }
   }
 
-#ifndef ROCKSDB_LITE  // GetSortedWalFiles is not supported
   int CountLiveLogFiles() {
     int micros_wait_for_log_deletion = 20000;
     env_->SleepForMicroseconds(micros_wait_for_log_deletion);
@@ -470,25 +426,18 @@ class ColumnFamilyTestBase : public testing::Test {
     return ret;
     return 0;
   }
-#endif  // !ROCKSDB_LITE
 
   void AssertCountLiveLogFiles(int value) {
-#ifndef ROCKSDB_LITE  // GetSortedWalFiles is not supported
     ASSERT_EQ(value, CountLiveLogFiles());
-#else
-    (void) value;
-#endif  // !ROCKSDB_LITE
   }
 
   void AssertNumberOfImmutableMemtables(std::vector<int> num_per_cf) {
     assert(num_per_cf.size() == handles_.size());
 
-#ifndef ROCKSDB_LITE  // GetProperty is not supported in lite
     for (size_t i = 0; i < num_per_cf.size(); ++i) {
       ASSERT_EQ(num_per_cf[i], GetProperty(static_cast<int>(i),
                                            "rocksdb.num-immutable-mem-table"));
     }
-#endif  // !ROCKSDB_LITE
   }
 
   void CopyFile(const std::string& source, const std::string& destination,
@@ -521,14 +470,14 @@ class ColumnFamilyTestBase : public testing::Test {
     return static_cast<int>(files.size());
   }
 
-  void RecalculateWriteStallConditions(ColumnFamilyData* cfd,
-      const MutableCFOptions& mutable_cf_options)  {
+  void RecalculateWriteStallConditions(
+      ColumnFamilyData* cfd, const MutableCFOptions& mutable_cf_options) {
     // add lock to avoid race condition between
     // `RecalculateWriteStallConditions` which writes to CFStats and
     // background `DBImpl::DumpStats()` threads which read CFStats
     dbfull()->TEST_LockMutex();
     cfd->RecalculateWriteStallConditions(mutable_cf_options);
-    dbfull()-> TEST_UnlockMutex();
+    dbfull()->TEST_UnlockMutex();
   }
 
   std::vector<ColumnFamilyHandle*> handles_;
@@ -584,7 +533,6 @@ TEST_P(ColumnFamilyTest, DontReuseColumnFamilyID) {
   }
 }
 
-#ifndef ROCKSDB_LITE
 TEST_P(ColumnFamilyTest, CreateCFRaceWithGetAggProperty) {
   Open();
 
@@ -607,7 +555,6 @@ TEST_P(ColumnFamilyTest, CreateCFRaceWithGetAggProperty) {
 
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
 }
-#endif  // !ROCKSDB_LITE
 
 class FlushEmptyCFTestWithParam
     : public ColumnFamilyTestBase,
@@ -951,7 +898,6 @@ TEST_P(ColumnFamilyTest, IgnoreRecoveredLog) {
   }
 }
 
-#ifndef ROCKSDB_LITE  // TEST functions used are not supported
 TEST_P(ColumnFamilyTest, FlushTest) {
   Open();
   CreateColumnFamiliesAndReopen({"one", "two"});
@@ -970,8 +916,7 @@ TEST_P(ColumnFamilyTest, FlushTest) {
     }
 
     for (int i = 0; i < 3; ++i) {
-      uint64_t max_total_in_memory_state =
-          MaxTotalInMemoryState();
+      uint64_t max_total_in_memory_state = MaxTotalInMemoryState();
       ASSERT_OK(Flush(i));
       AssertMaxTotalInMemoryState(max_total_in_memory_state);
     }
@@ -1067,7 +1012,6 @@ TEST_P(ColumnFamilyTest, LogDeletionTest) {
   AssertCountLiveLogFiles(4);
   Close();
 }
-#endif  // !ROCKSDB_LITE
 
 TEST_P(ColumnFamilyTest, CrashAfterFlush) {
   std::unique_ptr<FaultInjectionTestEnv> fault_env(
@@ -1107,7 +1051,6 @@ TEST_P(ColumnFamilyTest, OpenNonexistentColumnFamily) {
   ASSERT_TRUE(TryOpen({"default", "dne"}).IsInvalidArgument());
 }
 
-#ifndef ROCKSDB_LITE  // WaitForFlush() is not supported
 // Makes sure that obsolete log files get deleted
 TEST_P(ColumnFamilyTest, DifferentWriteBufferSizes) {
   // disable flushing stale column families
@@ -1209,20 +1152,18 @@ TEST_P(ColumnFamilyTest, DifferentWriteBufferSizes) {
   WaitForFlush(2);
   AssertNumberOfImmutableMemtables({0, 0, 0, 0});
   AssertCountLiveLogFiles(12);
-  PutRandomData(1, 2*200, 1000);
+  PutRandomData(1, 2 * 200, 1000);
   WaitForFlush(1);
   AssertNumberOfImmutableMemtables({0, 0, 0, 0});
   AssertCountLiveLogFiles(7);
   Close();
 }
-#endif  // !ROCKSDB_LITE
 
 // The test is commented out because we want to test that snapshot is
 // not created for memtables not supported it, but There isn't a memtable
 // that doesn't support snapshot right now. If we have one later, we can
 // re-enable the test.
 //
-// #ifndef ROCKSDB_LITE  // Cuckoo is not supported in lite
 //   TEST_P(ColumnFamilyTest, MemtableNotSupportSnapshot) {
 //   db_options_.allow_concurrent_memtable_write = false;
 //   Open();
@@ -1242,7 +1183,6 @@ TEST_P(ColumnFamilyTest, DifferentWriteBufferSizes) {
 //   {second}); auto* s3 = dbfull()->GetSnapshot(); ASSERT_TRUE(s3 == nullptr);
 //   Close();
 // }
-// #endif  // !ROCKSDB_LITE
 
 class TestComparator : public Comparator {
   int Compare(const ROCKSDB_NAMESPACE::Slice& /*a*/,
@@ -1309,7 +1249,6 @@ TEST_P(ColumnFamilyTest, DifferentMergeOperators) {
   Close();
 }
 
-#ifndef ROCKSDB_LITE  // WaitForFlush() is not supported
 TEST_P(ColumnFamilyTest, DifferentCompactionStyles) {
   Open();
   CreateColumnFamilies({"one", "two"});
@@ -1377,9 +1316,7 @@ TEST_P(ColumnFamilyTest, DifferentCompactionStyles) {
 
   Close();
 }
-#endif  // !ROCKSDB_LITE
 
-#ifndef ROCKSDB_LITE
 // Sync points not supported in RocksDB Lite
 
 TEST_P(ColumnFamilyTest, MultipleManualCompactions) {
@@ -2043,9 +1980,7 @@ TEST_P(ColumnFamilyTest, SameCFAutomaticManualCompactions) {
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
 }
-#endif  // !ROCKSDB_LITE
 
-#ifndef ROCKSDB_LITE  // Tailing iterator not supported
 namespace {
 std::string IterStatus(Iterator* iter) {
   std::string result;
@@ -2103,9 +2038,7 @@ TEST_P(ColumnFamilyTest, NewIteratorsTest) {
     Destroy();
   }
 }
-#endif  // !ROCKSDB_LITE
 
-#ifndef ROCKSDB_LITE  // ReadOnlyDB is not supported
 TEST_P(ColumnFamilyTest, ReadOnlyDBTest) {
   Open();
   CreateColumnFamiliesAndReopen({"one", "two", "three", "four"});
@@ -2123,7 +2056,6 @@ TEST_P(ColumnFamilyTest, ReadOnlyDBTest) {
   ASSERT_EQ("bla", Get(1, "foo"));
   ASSERT_EQ("blablablabla", Get(2, "foo"));
 
-
   // test newiterators
   {
     std::vector<Iterator*> iterators;
@@ -2155,9 +2087,7 @@ TEST_P(ColumnFamilyTest, ReadOnlyDBTest) {
   s = OpenReadOnly({"one", "four"});
   ASSERT_TRUE(!s.ok());
 }
-#endif  // !ROCKSDB_LITE
 
-#ifndef ROCKSDB_LITE  //  WaitForFlush() is not supported in lite
 TEST_P(ColumnFamilyTest, DontRollEmptyLogs) {
   Open();
   CreateColumnFamiliesAndReopen({"one", "two", "three", "four"});
@@ -2179,9 +2109,7 @@ TEST_P(ColumnFamilyTest, DontRollEmptyLogs) {
   ASSERT_EQ(static_cast<size_t>(total_new_writable_files), handles_.size() + 1);
   Close();
 }
-#endif  // !ROCKSDB_LITE
 
-#ifndef ROCKSDB_LITE  //  WaitForCompaction() is not supported in lite
 TEST_P(ColumnFamilyTest, FlushStaleColumnFamilies) {
   Open();
   CreateColumnFamilies({"one", "two"});
@@ -2228,7 +2156,6 @@ TEST_P(ColumnFamilyTest, FlushStaleColumnFamilies) {
   ASSERT_EQ(0, dbfull()->TEST_total_log_size());
   Close();
 }
-#endif  // !ROCKSDB_LITE
 
 TEST_P(ColumnFamilyTest, CreateMissingColumnFamilies) {
   Status s = TryOpen({"one", "two"});
@@ -2468,8 +2395,6 @@ TEST_P(ColumnFamilyTest, FlushAndDropRaceCondition) {
   Destroy();
 }
 
-#ifndef ROCKSDB_LITE
-// skipped as persisting options is not supported in ROCKSDB_LITE
 namespace {
 std::atomic<int> test_stage(0);
 std::atomic<bool> ordered_by_writethread(false);
@@ -2488,7 +2413,7 @@ void DropSingleColumnFamily(ColumnFamilyTest* cf_test, int cf_id,
   }
   test_stage = kChildThreadFinishDroppingColumnFamily;
 }
-}  // namespace
+}  // anonymous namespace
 
 TEST_P(ColumnFamilyTest, CreateAndDropRace) {
   const int kCfCount = 5;
@@ -2551,7 +2476,6 @@ TEST_P(ColumnFamilyTest, CreateAndDropRace) {
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
   ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
 }
-#endif  // !ROCKSDB_LITE
 
 TEST_P(ColumnFamilyTest, WriteStallSingleColumnFamily) {
   const uint64_t kBaseRate = 800000u;
@@ -2961,7 +2885,6 @@ TEST_P(ColumnFamilyTest, CreateDropAndDestroy) {
   ASSERT_OK(db_->DestroyColumnFamilyHandle(cfh));
 }
 
-#ifndef ROCKSDB_LITE
 TEST_P(ColumnFamilyTest, CreateDropAndDestroyWithoutFileDeletion) {
   ColumnFamilyHandle* cfh;
   Open();
@@ -3016,9 +2939,7 @@ TEST_P(ColumnFamilyTest, FlushCloseWALFiles) {
   db_options_.env = env_;
   Close();
 }
-#endif  // !ROCKSDB_LITE
 
-#ifndef ROCKSDB_LITE  // WaitForFlush() is not supported
 TEST_P(ColumnFamilyTest, IteratorCloseWALFile1) {
   SpecialEnv env(Env::Default());
   db_options_.env = &env;
@@ -3125,9 +3046,7 @@ TEST_P(ColumnFamilyTest, IteratorCloseWALFile2) {
   db_options_.env = env_;
   Close();
 }
-#endif  // !ROCKSDB_LITE
 
-#ifndef ROCKSDB_LITE  // TEST functions are not supported in lite
 TEST_P(ColumnFamilyTest, ForwardIteratorCloseWALFile) {
   SpecialEnv env(Env::Default());
   // Allow both of flush and purge job to schedule.
@@ -3203,7 +3122,6 @@ TEST_P(ColumnFamilyTest, ForwardIteratorCloseWALFile) {
   db_options_.env = env_;
   Close();
 }
-#endif  // !ROCKSDB_LITE
 
 // Disable on windows because SyncWAL requires env->IsSyncThreadSafe()
 // to return true which is not so in unbuffered mode.

db/compact_files_test.cc

@@ -3,7 +3,6 @@
 //  COPYING file in the root directory) and Apache 2.0 License
 //  (found in the LICENSE.Apache file in the root directory).
 
-#ifndef ROCKSDB_LITE
 
 #include <mutex>
 #include <string>
@@ -121,7 +120,9 @@ TEST_F(CompactFilesTest, L0ConflictsFiles) {
 TEST_F(CompactFilesTest, MultipleLevel) {
   Options options;
   options.create_if_missing = true;
-  options.level_compaction_dynamic_level_bytes = true;
+  // Otherwise background compaction can happen to
+  // drain unnecessary level
+  options.level_compaction_dynamic_level_bytes = false;
   options.num_levels = 6;
   // Add listener
   FlushedFileCollector* collector = new FlushedFileCollector();
@@ -182,7 +183,6 @@ TEST_F(CompactFilesTest, MultipleLevel) {
   for (int invalid_output_level = 0; invalid_output_level < 5;
        invalid_output_level++) {
     s = db->CompactFiles(CompactionOptions(), files, invalid_output_level);
-    std::cout << s.ToString() << std::endl;
     ASSERT_TRUE(s.IsInvalidArgument());
   }
 
@@ -348,9 +348,7 @@ TEST_F(CompactFilesTest, CompactionFilterWithGetSv) {
       return true;
     }
 
-    void SetDB(DB* db) {
-      db_ = db;
-    }
+    void SetDB(DB* db) { db_ = db; }
 
     const char* Name() const override { return "FilterWithGet"; }
 
@@ -358,7 +356,6 @@ TEST_F(CompactFilesTest, CompactionFilterWithGetSv) {
     DB* db_;
   };
 
-
   std::shared_ptr<FilterWithGet> cf(new FilterWithGet());
 
   Options options;
@@ -385,7 +382,6 @@ TEST_F(CompactFilesTest, CompactionFilterWithGetSv) {
         db->CompactFiles(ROCKSDB_NAMESPACE::CompactionOptions(), {fname}, 0));
   }
 
-
   delete db;
 }
 
@@ -400,10 +396,9 @@ TEST_F(CompactFilesTest, SentinelCompressionType) {
   }
   // Check that passing `CompressionType::kDisableCompressionOption` to
   // `CompactFiles` causes it to use the column family compression options.
-  for (auto compaction_style :
-       {CompactionStyle::kCompactionStyleLevel,
-        CompactionStyle::kCompactionStyleUniversal,
-        CompactionStyle::kCompactionStyleNone}) {
+  for (auto compaction_style : {CompactionStyle::kCompactionStyleLevel,
+                                CompactionStyle::kCompactionStyleUniversal,
+                                CompactionStyle::kCompactionStyleNone}) {
     ASSERT_OK(DestroyDB(db_name_, Options()));
     Options options;
     options.compaction_style = compaction_style;
@@ -495,13 +490,3 @@ int main(int argc, char** argv) {
   return RUN_ALL_TESTS();
 }
 
-#else
-#include <stdio.h>
-
-int main(int /*argc*/, char** /*argv*/) {
-  fprintf(stderr,
-          "SKIPPED as DBImpl::CompactFiles is not supported in ROCKSDB_LITE\n");
-  return 0;
-}
-
-#endif  // !ROCKSDB_LITE

db/compaction/clipping_iterator.h

@@ -188,6 +188,11 @@ class ClippingIterator : public InternalIterator {
     return iter_->GetProperty(prop_name, prop);
   }
 
+  bool IsDeleteRangeSentinelKey() const override {
+    assert(valid_);
+    return iter_->IsDeleteRangeSentinelKey();
+  }
+
  private:
   void UpdateValid() {
     assert(!iter_->Valid() || iter_->status().ok());

db/compaction/compaction.cc

@@ -238,12 +238,19 @@ Compaction::Compaction(
       inputs_(PopulateWithAtomicBoundaries(vstorage, std::move(_inputs))),
       grandparents_(std::move(_grandparents)),
       score_(_score),
-      bottommost_level_(IsBottommostLevel(output_level_, vstorage, inputs_)),
+      bottommost_level_(
+          // For simplicity, we don't support the concept of "bottommost level"
+          // with
+          // `CompactionReason::kExternalSstIngestion` and
+          // `CompactionReason::kRefitLevel`
+          (_compaction_reason == CompactionReason::kExternalSstIngestion ||
+           _compaction_reason == CompactionReason::kRefitLevel)
+              ? false
+              : IsBottommostLevel(output_level_, vstorage, inputs_)),
       is_full_compaction_(IsFullCompaction(vstorage, inputs_)),
       is_manual_compaction_(_manual_compaction),
       trim_ts_(_trim_ts),
       is_trivial_move_(false),
-
       compaction_reason_(_compaction_reason),
       notify_on_compaction_completion_(false),
       enable_blob_garbage_collection_(
@@ -258,8 +265,15 @@ Compaction::Compaction(
                   _blob_garbage_collection_age_cutoff > 1
               ? mutable_cf_options()->blob_garbage_collection_age_cutoff
               : _blob_garbage_collection_age_cutoff),
-      penultimate_level_(EvaluatePenultimateLevel(
-          vstorage, immutable_options_, start_level_, output_level_)) {
+      penultimate_level_(
+          // For simplicity, we don't support the concept of "penultimate level"
+          // with `CompactionReason::kExternalSstIngestion` and
+          // `CompactionReason::kRefitLevel`
+          _compaction_reason == CompactionReason::kExternalSstIngestion ||
+                  _compaction_reason == CompactionReason::kRefitLevel
+              ? Compaction::kInvalidLevel
+              : EvaluatePenultimateLevel(vstorage, immutable_options_,
+                                         start_level_, output_level_)) {
   MarkFilesBeingCompacted(true);
   if (is_manual_compaction_) {
     compaction_reason_ = CompactionReason::kManualCompaction;
@@ -332,6 +346,7 @@ void Compaction::PopulatePenultimateLevelOutputRange() {
   // the case that the penultimate level is empty).
   if (immutable_options_.compaction_style == kCompactionStyleUniversal) {
     exclude_level = kInvalidLevel;
+    penultimate_output_range_type_ = PenultimateOutputRangeType::kFullRange;
     std::set<uint64_t> penultimate_inputs;
     for (const auto& input_lvl : inputs_) {
       if (input_lvl.level == penultimate_level_) {
@@ -345,7 +360,8 @@ void Compaction::PopulatePenultimateLevelOutputRange() {
       if (penultimate_inputs.find(file->fd.GetNumber()) ==
           penultimate_inputs.end()) {
         exclude_level = number_levels_ - 1;
-        penultimate_output_range_type_ = PenultimateOutputRangeType::kFullRange;
+        penultimate_output_range_type_ =
+            PenultimateOutputRangeType::kNonLastRange;
         break;
       }
     }
@@ -354,35 +370,6 @@ void Compaction::PopulatePenultimateLevelOutputRange() {
   GetBoundaryKeys(input_vstorage_, inputs_,
                   &penultimate_level_smallest_user_key_,
                   &penultimate_level_largest_user_key_, exclude_level);
-
-  // If there's a case that the penultimate level output range is overlapping
-  // with the existing files, disable the penultimate level output by setting
-  // the range to empty. One example is the range delete could have overlap
-  // boundary with the next file. (which is actually a false overlap)
-  // TODO: Exclude such false overlap, so it won't disable the penultimate
-  //  output.
-  std::set<uint64_t> penultimate_inputs;
-  for (const auto& input_lvl : inputs_) {
-    if (input_lvl.level == penultimate_level_) {
-      for (const auto& file : input_lvl.files) {
-        penultimate_inputs.emplace(file->fd.GetNumber());
-      }
-    }
-  }
-
-  auto penultimate_files = input_vstorage_->LevelFiles(penultimate_level_);
-  for (const auto& file : penultimate_files) {
-    if (penultimate_inputs.find(file->fd.GetNumber()) ==
-            penultimate_inputs.end() &&
-        OverlapPenultimateLevelOutputRange(file->smallest.user_key(),
-                                           file->largest.user_key())) {
-      // basically disable the penultimate range output. which should be rare
-      // or a false overlap caused by range del
-      penultimate_level_smallest_user_key_ = "";
-      penultimate_level_largest_user_key_ = "";
-      penultimate_output_range_type_ = PenultimateOutputRangeType::kDisabled;
-    }
-  }
 }
 
 Compaction::~Compaction() {
@@ -807,6 +794,16 @@ uint64_t Compaction::MinInputFileOldestAncesterTime(
   return min_oldest_ancester_time;
 }
 
+uint64_t Compaction::MinInputFileEpochNumber() const {
+  uint64_t min_epoch_number = std::numeric_limits<uint64_t>::max();
+  for (const auto& inputs_per_level : inputs_) {
+    for (const auto& file : inputs_per_level.files) {
+      min_epoch_number = std::min(min_epoch_number, file->epoch_number);
+    }
+  }
+  return min_epoch_number;
+}
+
 int Compaction::EvaluatePenultimateLevel(
     const VersionStorageInfo* vstorage,
     const ImmutableOptions& immutable_options, const int start_level,

db/compaction/compaction.h

@@ -378,6 +378,9 @@ class Compaction {
   // This is used to filter out some input files' ancester's time range.
   uint64_t MinInputFileOldestAncesterTime(const InternalKey* start,
                                           const InternalKey* end) const;
+  // Return the minimum epoch number among
+  // input files' associated with this compaction
+  uint64_t MinInputFileEpochNumber() const;
 
   // Called by DBImpl::NotifyOnCompactionCompleted to make sure number of
   // compaction begin and compaction completion callbacks match.
@@ -543,13 +546,16 @@ struct PerKeyPlacementContext {
   const Slice value;
   const SequenceNumber seq_num;
 
-  bool output_to_penultimate_level;
+  bool& output_to_penultimate_level;
 
   PerKeyPlacementContext(int _level, Slice _key, Slice _value,
-                         SequenceNumber _seq_num)
-      : level(_level), key(_key), value(_value), seq_num(_seq_num) {
-    output_to_penultimate_level = false;
-  }
+                         SequenceNumber _seq_num,
+                         bool& _output_to_penultimate_level)
+      : level(_level),
+        key(_key),
+        value(_value),
+        seq_num(_seq_num),
+        output_to_penultimate_level(_output_to_penultimate_level) {}
 };
 #endif /* !NDEBUG */
 

db/compaction/compaction_iterator.cc

@@ -13,6 +13,7 @@
 #include "db/blob/blob_index.h"
 #include "db/blob/prefetch_buffer_collection.h"
 #include "db/snapshot_checker.h"
+#include "db/wide/wide_column_serialization.h"
 #include "logging/logging.h"
 #include "port/likely.h"
 #include "rocksdb/listener.h"
@@ -124,6 +125,9 @@ CompactionIterator::CompactionIterator(
          timestamp_size_ == full_history_ts_low_->size());
 #endif
   input_.SetPinnedItersMgr(&pinned_iters_mgr_);
+  // The default `merge_until_status_` does not need to be checked since it is
+  // overwritten as soon as `MergeUntil()` is called
+  merge_until_status_.PermitUncheckedError();
   TEST_SYNC_POINT_CALLBACK("CompactionIterator:AfterInit", compaction_.get());
 }
 
@@ -168,11 +172,30 @@ void CompactionIterator::Next() {
       }
 
       // Keep current_key_ in sync.
-      current_key_.UpdateInternalKey(ikey_.sequence, ikey_.type);
+      if (0 == timestamp_size_) {
+        current_key_.UpdateInternalKey(ikey_.sequence, ikey_.type);
+      } else {
+        Slice ts = ikey_.GetTimestamp(timestamp_size_);
+        current_key_.UpdateInternalKey(ikey_.sequence, ikey_.type, &ts);
+      }
       key_ = current_key_.GetInternalKey();
       ikey_.user_key = current_key_.GetUserKey();
       validity_info_.SetValid(ValidContext::kMerge1);
     } else {
+      if (merge_until_status_.IsMergeInProgress()) {
+        // `Status::MergeInProgress()` tells us that the previous `MergeUntil()`
+        // produced only merge operands. Those merge operands were accessed and
+        // written out using `merge_out_iter_`. Since `merge_out_iter_` is
+        // exhausted at this point, all merge operands have been written out.
+        //
+        // Still, there may be a base value (PUT, DELETE, SINGLEDEL, etc.) that
+        // needs to be written out. Normally, `CompactionIterator` would skip it
+        // on the basis that it has already output something in the same
+        // snapshot stripe. To prevent this, we reset `has_current_user_key_` to
+        // trick the future iteration from finding out the snapshot stripe is
+        // unchanged.
+        has_current_user_key_ = false;
+      }
       // We consumed all pinned merge operands, release pinned iterators
       pinned_iters_mgr_.ReleasePinnedData();
       // MergeHelper moves the iterator to the first record after the merged
@@ -199,39 +222,47 @@ void CompactionIterator::Next() {
 
 bool CompactionIterator::InvokeFilterIfNeeded(bool* need_skip,
                                               Slice* skip_until) {
-  // TODO: support compaction filter for wide-column entities
-  if (!compaction_filter_ ||
-      (ikey_.type != kTypeValue && ikey_.type != kTypeBlobIndex)) {
+  if (!compaction_filter_) {
     return true;
   }
-  bool error = false;
-  // If the user has specified a compaction filter and the sequence
-  // number is greater than any external snapshot, then invoke the
-  // filter. If the return value of the compaction filter is true,
-  // replace the entry with a deletion marker.
-  CompactionFilter::Decision filter = CompactionFilter::Decision::kUndetermined;
-  compaction_filter_value_.clear();
-  compaction_filter_skip_until_.Clear();
+
+  if (ikey_.type != kTypeValue && ikey_.type != kTypeBlobIndex &&
+      ikey_.type != kTypeWideColumnEntity) {
+    return true;
+  }
+
+  CompactionFilter::Decision decision =
+      CompactionFilter::Decision::kUndetermined;
   CompactionFilter::ValueType value_type =
       ikey_.type == kTypeValue ? CompactionFilter::ValueType::kValue
-                               : CompactionFilter::ValueType::kBlobIndex;
+      : ikey_.type == kTypeBlobIndex
+          ? CompactionFilter::ValueType::kBlobIndex
+          : CompactionFilter::ValueType::kWideColumnEntity;
+
   // Hack: pass internal key to BlobIndexCompactionFilter since it needs
   // to get sequence number.
   assert(compaction_filter_);
-  Slice& filter_key =
-      (ikey_.type == kTypeValue ||
+  const Slice& filter_key =
+      (ikey_.type != kTypeBlobIndex ||
        !compaction_filter_->IsStackedBlobDbInternalCompactionFilter())
           ? ikey_.user_key
           : key_;
+
+  compaction_filter_value_.clear();
+  compaction_filter_skip_until_.Clear();
+
+  std::vector<std::pair<std::string, std::string>> new_columns;
+
   {
     StopWatchNano timer(clock_, report_detailed_time_);
-    if (kTypeBlobIndex == ikey_.type) {
-      filter = compaction_filter_->FilterBlobByKey(
+
+    if (ikey_.type == kTypeBlobIndex) {
+      decision = compaction_filter_->FilterBlobByKey(
           level_, filter_key, &compaction_filter_value_,
           compaction_filter_skip_until_.rep());
-      if (CompactionFilter::Decision::kUndetermined == filter &&
+      if (decision == CompactionFilter::Decision::kUndetermined &&
           !compaction_filter_->IsStackedBlobDbInternalCompactionFilter()) {
-        if (compaction_ == nullptr) {
+        if (!compaction_) {
           status_ =
               Status::Corruption("Unexpected blob index outside of compaction");
           validity_info_.Invalidate();
@@ -277,33 +308,61 @@ bool CompactionIterator::InvokeFilterIfNeeded(bool* need_skip,
         value_type = CompactionFilter::ValueType::kValue;
       }
     }
-    if (CompactionFilter::Decision::kUndetermined == filter) {
-      filter = compaction_filter_->FilterV2(
-          level_, filter_key, value_type,
-          blob_value_.empty() ? value_ : blob_value_, &compaction_filter_value_,
+
+    if (decision == CompactionFilter::Decision::kUndetermined) {
+      const Slice* existing_val = nullptr;
+      const WideColumns* existing_col = nullptr;
+
+      WideColumns existing_columns;
+
+      if (ikey_.type != kTypeWideColumnEntity) {
+        if (!blob_value_.empty()) {
+          existing_val = &blob_value_;
+        } else {
+          existing_val = &value_;
+        }
+      } else {
+        Slice value_copy = value_;
+        const Status s =
+            WideColumnSerialization::Deserialize(value_copy, existing_columns);
+
+        if (!s.ok()) {
+          status_ = s;
+          validity_info_.Invalidate();
+          return false;
+        }
+
+        existing_col = &existing_columns;
+      }
+
+      decision = compaction_filter_->FilterV3(
+          level_, filter_key, value_type, existing_val, existing_col,
+          &compaction_filter_value_, &new_columns,
           compaction_filter_skip_until_.rep());
     }
+
     iter_stats_.total_filter_time +=
         env_ != nullptr && report_detailed_time_ ? timer.ElapsedNanos() : 0;
   }
 
-  if (CompactionFilter::Decision::kUndetermined == filter) {
-    // Should not reach here, since FilterV2 should never return kUndetermined.
-    status_ =
-        Status::NotSupported("FilterV2() should never return kUndetermined");
+  if (decision == CompactionFilter::Decision::kUndetermined) {
+    // Should not reach here, since FilterV2/FilterV3 should never return
+    // kUndetermined.
+    status_ = Status::NotSupported(
+        "FilterV2/FilterV3 should never return kUndetermined");
     validity_info_.Invalidate();
     return false;
   }
 
-  if (filter == CompactionFilter::Decision::kRemoveAndSkipUntil &&
+  if (decision == CompactionFilter::Decision::kRemoveAndSkipUntil &&
       cmp_->Compare(*compaction_filter_skip_until_.rep(), ikey_.user_key) <=
           0) {
     // Can't skip to a key smaller than the current one.
-    // Keep the key as per FilterV2 documentation.
-    filter = CompactionFilter::Decision::kKeep;
+    // Keep the key as per FilterV2/FilterV3 documentation.
+    decision = CompactionFilter::Decision::kKeep;
   }
 
-  if (filter == CompactionFilter::Decision::kRemove) {
+  if (decision == CompactionFilter::Decision::kRemove) {
     // convert the current key to a delete; key_ is pointing into
     // current_key_ at this point, so updating current_key_ updates key()
     ikey_.type = kTypeDeletion;
@@ -311,7 +370,7 @@ bool CompactionIterator::InvokeFilterIfNeeded(bool* need_skip,
     // no value associated with delete
     value_.clear();
     iter_stats_.num_record_drop_user++;
-  } else if (filter == CompactionFilter::Decision::kPurge) {
+  } else if (decision == CompactionFilter::Decision::kPurge) {
     // convert the current key to a single delete; key_ is pointing into
     // current_key_ at this point, so updating current_key_ updates key()
     ikey_.type = kTypeSingleDeletion;
@@ -319,19 +378,19 @@ bool CompactionIterator::InvokeFilterIfNeeded(bool* need_skip,
     // no value associated with single delete
     value_.clear();
     iter_stats_.num_record_drop_user++;
-  } else if (filter == CompactionFilter::Decision::kChangeValue) {
-    if (ikey_.type == kTypeBlobIndex) {
-      // value transfer from blob file to inlined data
+  } else if (decision == CompactionFilter::Decision::kChangeValue) {
+    if (ikey_.type != kTypeValue) {
       ikey_.type = kTypeValue;
-      current_key_.UpdateInternalKey(ikey_.sequence, ikey_.type);
+      current_key_.UpdateInternalKey(ikey_.sequence, kTypeValue);
     }
+
     value_ = compaction_filter_value_;
-  } else if (filter == CompactionFilter::Decision::kRemoveAndSkipUntil) {
+  } else if (decision == CompactionFilter::Decision::kRemoveAndSkipUntil) {
     *need_skip = true;
     compaction_filter_skip_until_.ConvertFromUserKey(kMaxSequenceNumber,
                                                      kValueTypeForSeek);
     *skip_until = compaction_filter_skip_until_.Encode();
-  } else if (filter == CompactionFilter::Decision::kChangeBlobIndex) {
+  } else if (decision == CompactionFilter::Decision::kChangeBlobIndex) {
     // Only the StackableDB-based BlobDB impl's compaction filter should return
     // kChangeBlobIndex. Decision about rewriting blob and changing blob index
     // in the integrated BlobDB impl is made in subsequent call to
@@ -343,23 +402,56 @@ bool CompactionIterator::InvokeFilterIfNeeded(bool* need_skip,
       validity_info_.Invalidate();
       return false;
     }
-    if (ikey_.type == kTypeValue) {
-      // value transfer from inlined data to blob file
+
+    if (ikey_.type != kTypeBlobIndex) {
       ikey_.type = kTypeBlobIndex;
-      current_key_.UpdateInternalKey(ikey_.sequence, ikey_.type);
+      current_key_.UpdateInternalKey(ikey_.sequence, kTypeBlobIndex);
     }
+
     value_ = compaction_filter_value_;
-  } else if (filter == CompactionFilter::Decision::kIOError) {
+  } else if (decision == CompactionFilter::Decision::kIOError) {
     if (!compaction_filter_->IsStackedBlobDbInternalCompactionFilter()) {
       status_ = Status::NotSupported(
           "CompactionFilter for integrated BlobDB should not return kIOError");
       validity_info_.Invalidate();
       return false;
     }
+
     status_ = Status::IOError("Failed to access blob during compaction filter");
-    error = true;
+    validity_info_.Invalidate();
+    return false;
+  } else if (decision == CompactionFilter::Decision::kChangeWideColumnEntity) {
+    WideColumns sorted_columns;
+
+    sorted_columns.reserve(new_columns.size());
+    for (const auto& column : new_columns) {
+      sorted_columns.emplace_back(column.first, column.second);
+    }
+
+    std::sort(sorted_columns.begin(), sorted_columns.end(),
+              [](const WideColumn& lhs, const WideColumn& rhs) {
+                return lhs.name().compare(rhs.name()) < 0;
+              });
+
+    {
+      const Status s = WideColumnSerialization::Serialize(
+          sorted_columns, compaction_filter_value_);
+      if (!s.ok()) {
+        status_ = s;
+        validity_info_.Invalidate();
+        return false;
+      }
+    }
+
+    if (ikey_.type != kTypeWideColumnEntity) {
+      ikey_.type = kTypeWideColumnEntity;
+      current_key_.UpdateInternalKey(ikey_.sequence, kTypeWideColumnEntity);
+    }
+
+    value_ = compaction_filter_value_;
   }
-  return !error;
+
+  return true;
 }
 
 void CompactionIterator::NextFromInput() {
@@ -372,6 +464,7 @@ void CompactionIterator::NextFromInput() {
     value_ = input_.value();
     blob_value_.Reset();
     iter_stats_.num_input_records++;
+    is_range_del_ = input_.IsDeleteRangeSentinelKey();
 
     Status pik_status = ParseInternalKey(key_, &ikey_, allow_data_in_errors_);
     if (!pik_status.ok()) {
@@ -391,7 +484,10 @@ void CompactionIterator::NextFromInput() {
       break;
     }
     TEST_SYNC_POINT_CALLBACK("CompactionIterator:ProcessKV", &ikey_);
-
+    if (is_range_del_) {
+      validity_info_.SetValid(kRangeDeletion);
+      break;
+    }
     // Update input statistics
     if (ikey_.type == kTypeDeletion || ikey_.type == kTypeSingleDeletion ||
         ikey_.type == kTypeDeletionWithTimestamp) {
@@ -613,6 +709,14 @@ void CompactionIterator::NextFromInput() {
 
       ParsedInternalKey next_ikey;
       AdvanceInputIter();
+      while (input_.Valid() && input_.IsDeleteRangeSentinelKey() &&
+             ParseInternalKey(input_.key(), &next_ikey, allow_data_in_errors_)
+                 .ok() &&
+             cmp_->EqualWithoutTimestamp(ikey_.user_key, next_ikey.user_key)) {
+        // skip range tombstone start keys with the same user key
+        // since they are not "real" point keys.
+        AdvanceInputIter();
+      }
 
       // Check whether the next key exists, is not corrupt, and is the same key
       // as the single delete.
@@ -620,6 +724,7 @@ void CompactionIterator::NextFromInput() {
           ParseInternalKey(input_.key(), &next_ikey, allow_data_in_errors_)
               .ok() &&
           cmp_->EqualWithoutTimestamp(ikey_.user_key, next_ikey.user_key)) {
+        assert(!input_.IsDeleteRangeSentinelKey());
 #ifndef NDEBUG
         const Compaction* c =
             compaction_ ? compaction_->real_compaction() : nullptr;
@@ -825,8 +930,8 @@ void CompactionIterator::NextFromInput() {
                  cmp_with_history_ts_low_ < 0)) &&
                bottommost_level_) {
       // Handle the case where we have a delete key at the bottom most level
-      // We can skip outputting the key iff there are no subsequent puts for this
-      // key
+      // We can skip outputting the key iff there are no subsequent puts for
+      // this key
       assert(!compaction_ || compaction_->KeyNotExistsBeyondOutputLevel(
                                  ikey_.user_key, &level_ptrs_));
       ParsedInternalKey next_ikey;
@@ -844,17 +949,19 @@ void CompactionIterator::NextFromInput() {
       // Note that a deletion marker of type kTypeDeletionWithTimestamp will be
       // considered to have a different user key unless the timestamp is older
       // than *full_history_ts_low_.
+      //
+      // Range tombstone start keys are skipped as they are not "real" keys.
       while (!IsPausingManualCompaction() && !IsShuttingDown() &&
              input_.Valid() &&
              (ParseInternalKey(input_.key(), &next_ikey, allow_data_in_errors_)
                   .ok()) &&
              cmp_->EqualWithoutTimestamp(ikey_.user_key, next_ikey.user_key) &&
-             (prev_snapshot == 0 ||
+             (prev_snapshot == 0 || input_.IsDeleteRangeSentinelKey() ||
               DefinitelyNotInSnapshot(next_ikey.sequence, prev_snapshot))) {
         AdvanceInputIter();
       }
-      // If you find you still need to output a row with this key, we need to output the
-      // delete too
+      // If you find you still need to output a row with this key, we need to
+      // output the delete too
       if (input_.Valid() &&
           (ParseInternalKey(input_.key(), &next_ikey, allow_data_in_errors_)
                .ok()) &&
@@ -875,14 +982,15 @@ void CompactionIterator::NextFromInput() {
       // have hit (A)
       // We encapsulate the merge related state machine in a different
       // object to minimize change to the existing flow.
-      Status s = merge_helper_->MergeUntil(
+      merge_until_status_ = merge_helper_->MergeUntil(
           &input_, range_del_agg_, prev_snapshot, bottommost_level_,
-          allow_data_in_errors_, blob_fetcher_.get(), prefetch_buffers_.get(),
-          &iter_stats_);
+          allow_data_in_errors_, blob_fetcher_.get(), full_history_ts_low_,
+          prefetch_buffers_.get(), &iter_stats_);
       merge_out_iter_.SeekToFirst();
 
-      if (!s.ok() && !s.IsMergeInProgress()) {
-        status_ = s;
+      if (!merge_until_status_.ok() &&
+          !merge_until_status_.IsMergeInProgress()) {
+        status_ = merge_until_status_;
         return;
       } else if (merge_out_iter_.Valid()) {
         // NOTE: key, value, and ikey_ refer to old entries.
@@ -1093,17 +1201,7 @@ void CompactionIterator::GarbageCollectBlobIfNeeded() {
 
 void CompactionIterator::DecideOutputLevel() {
   assert(compaction_->SupportsPerKeyPlacement());
-#ifndef NDEBUG
-  // Could be overridden by unittest
-  PerKeyPlacementContext context(level_, ikey_.user_key, value_,
-                                 ikey_.sequence);
-  TEST_SYNC_POINT_CALLBACK("CompactionIterator::PrepareOutput.context",
-                           &context);
-  output_to_penultimate_level_ = context.output_to_penultimate_level;
-#else
   output_to_penultimate_level_ = false;
-#endif  // NDEBUG
-
   // if the key is newer than the cutoff sequence or within the earliest
   // snapshot, it should output to the penultimate level.
   if (ikey_.sequence > preclude_last_level_min_seqno_ ||
@@ -1111,6 +1209,17 @@ void CompactionIterator::DecideOutputLevel() {
     output_to_penultimate_level_ = true;
   }
 
+#ifndef NDEBUG
+  // Could be overridden by unittest
+  PerKeyPlacementContext context(level_, ikey_.user_key, value_, ikey_.sequence,
+                                 output_to_penultimate_level_);
+  TEST_SYNC_POINT_CALLBACK("CompactionIterator::PrepareOutput.context",
+                           &context);
+  if (ikey_.sequence > earliest_snapshot_) {
+    output_to_penultimate_level_ = true;
+  }
+#endif  // NDEBUG
+
   if (output_to_penultimate_level_) {
     // If it's decided to output to the penultimate level, but unsafe to do so,
     // still output to the last level. For example, moving the data from a lower
@@ -1142,10 +1251,12 @@ void CompactionIterator::DecideOutputLevel() {
 
 void CompactionIterator::PrepareOutput() {
   if (Valid()) {
-    if (ikey_.type == kTypeValue) {
-      ExtractLargeValueIfNeeded();
-    } else if (ikey_.type == kTypeBlobIndex) {
-      GarbageCollectBlobIfNeeded();
+    if (LIKELY(!is_range_del_)) {
+      if (ikey_.type == kTypeValue) {
+        ExtractLargeValueIfNeeded();
+      } else if (ikey_.type == kTypeBlobIndex) {
+        GarbageCollectBlobIfNeeded();
+      }
     }
 
     if (compaction_ != nullptr && compaction_->SupportsPerKeyPlacement()) {
@@ -1168,7 +1279,7 @@ void CompactionIterator::PrepareOutput() {
         DefinitelyInSnapshot(ikey_.sequence, earliest_snapshot_) &&
         ikey_.type != kTypeMerge && current_key_committed_ &&
         !output_to_penultimate_level_ &&
-        ikey_.sequence < preserve_time_min_seqno_) {
+        ikey_.sequence < preserve_time_min_seqno_ && !is_range_del_) {
       if (ikey_.type == kTypeDeletion ||
           (ikey_.type == kTypeSingleDeletion && timestamp_size_ == 0)) {
         ROCKS_LOG_FATAL(
@@ -1214,8 +1325,8 @@ inline SequenceNumber CompactionIterator::findEarliestVisibleSnapshot(
     ROCKS_LOG_FATAL(info_log_,
                     "No snapshot left in findEarliestVisibleSnapshot");
   }
-  auto snapshots_iter = std::lower_bound(
-      snapshots_->begin(), snapshots_->end(), in);
+  auto snapshots_iter =
+      std::lower_bound(snapshots_->begin(), snapshots_->end(), in);
   assert(prev_snapshot != nullptr);
   if (snapshots_iter == snapshots_->begin()) {
     *prev_snapshot = 0;
@@ -1230,8 +1341,8 @@ inline SequenceNumber CompactionIterator::findEarliestVisibleSnapshot(
     }
   }
   if (snapshot_checker_ == nullptr) {
-    return snapshots_iter != snapshots_->end()
-      ? *snapshots_iter : kMaxSequenceNumber;
+    return snapshots_iter != snapshots_->end() ? *snapshots_iter
+                                               : kMaxSequenceNumber;
   }
   bool has_released_snapshot = !released_snapshots_.empty();
   for (; snapshots_iter != snapshots_->end(); ++snapshots_iter) {
@@ -1301,6 +1412,7 @@ std::unique_ptr<BlobFetcher> CompactionIterator::CreateBlobFetcherIfNeeded(
   }
 
   ReadOptions read_options;
+  read_options.io_activity = Env::IOActivity::kCompaction;
   read_options.fill_cache = false;
 
   return std::unique_ptr<BlobFetcher>(new BlobFetcher(version, read_options));

db/compaction/compaction_iterator.h

@@ -63,6 +63,10 @@ class SequenceIterWrapper : public InternalIterator {
   void SeekToLast() override { assert(false); }
 
   uint64_t num_itered() const { return num_itered_; }
+  bool IsDeleteRangeSentinelKey() const override {
+    assert(Valid());
+    return inner_iter_->IsDeleteRangeSentinelKey();
+  }
 
  private:
   InternalKeyComparator icmp_;
@@ -242,7 +246,12 @@ class CompactionIterator {
   const Status& status() const { return status_; }
   const ParsedInternalKey& ikey() const { return ikey_; }
   inline bool Valid() const { return validity_info_.IsValid(); }
-  const Slice& user_key() const { return current_user_key_; }
+  const Slice& user_key() const {
+    if (UNLIKELY(is_range_del_)) {
+      return ikey_.user_key;
+    }
+    return current_user_key_;
+  }
   const CompactionIterationStats& iter_stats() const { return iter_stats_; }
   uint64_t num_input_entry_scanned() const { return input_.num_itered(); }
   // If the current key should be placed on penultimate level, only valid if
@@ -252,6 +261,8 @@ class CompactionIterator {
   }
   Status InputStatus() const { return input_.status(); }
 
+  bool IsDeleteRangeSentinelKey() const { return is_range_del_; }
+
  private:
   // Processes the input stream to find the next output
   void NextFromInput();
@@ -385,6 +396,7 @@ class CompactionIterator {
     kKeepSD = 8,
     kKeepDel = 9,
     kNewUserKey = 10,
+    kRangeDeletion = 11,
   };
 
   struct ValidityInfo {
@@ -432,6 +444,7 @@ class CompactionIterator {
   bool clear_and_output_next_key_ = false;
 
   MergeOutputIterator merge_out_iter_;
+  Status merge_until_status_;
   // PinnedIteratorsManager used to pin input_ Iterator blocks while reading
   // merge operands and then releasing them after consuming them.
   PinnedIteratorsManager pinned_iters_mgr_;
@@ -492,6 +505,10 @@ class CompactionIterator {
     // This is a best-effort facility, so memory_order_relaxed is sufficient.
     return manual_compaction_canceled_.load(std::memory_order_relaxed);
   }
+
+  // Stores whether the current compaction iterator output
+  // is a range tombstone start key.
+  bool is_range_del_{false};
 };
 
 inline bool CompactionIterator::DefinitelyInSnapshot(SequenceNumber seq,

db/compaction/compaction_iterator_test.cc

@@ -203,7 +203,8 @@ class TestSnapshotChecker : public SnapshotChecker {
  public:
   explicit TestSnapshotChecker(
       SequenceNumber last_committed_sequence,
-      const std::unordered_map<SequenceNumber, SequenceNumber>& snapshots = {{}})
+      const std::unordered_map<SequenceNumber, SequenceNumber>& snapshots =
+          {{}})
       : last_committed_sequence_(last_committed_sequence),
         snapshots_(snapshots) {}
 
@@ -811,6 +812,8 @@ TEST_P(PerKeyPlacementCompIteratorTest, SplitLastLevelData) {
   c_iter_->Next();
   ASSERT_OK(c_iter_->status());
   ASSERT_FALSE(c_iter_->Valid());
+  SyncPoint::GetInstance()->DisableProcessing();
+  SyncPoint::GetInstance()->ClearAllCallBacks();
 }
 
 TEST_P(PerKeyPlacementCompIteratorTest, SnapshotData) {
@@ -876,6 +879,8 @@ TEST_P(PerKeyPlacementCompIteratorTest, ConflictWithSnapshot) {
   // output_to_penultimate_level.
   c_iter_->Next();
   ASSERT_TRUE(c_iter_->status().IsCorruption());
+  SyncPoint::GetInstance()->DisableProcessing();
+  SyncPoint::GetInstance()->ClearAllCallBacks();
 }
 
 INSTANTIATE_TEST_CASE_P(PerKeyPlacementCompIteratorTest,
@@ -1038,7 +1043,7 @@ TEST_F(CompactionIteratorWithSnapshotCheckerTest,
 
 TEST_F(CompactionIteratorWithSnapshotCheckerTest,
        NotRemoveDeletionIfValuePresentToEarlierSnapshot) {
-  AddSnapshot(2,1);
+  AddSnapshot(2, 1);
   RunTest({test::KeyStr("a", 4, kTypeDeletion),
            test::KeyStr("a", 1, kTypeValue), test::KeyStr("b", 3, kTypeValue)},
           {"", "", ""},
@@ -1250,6 +1255,31 @@ TEST_P(CompactionIteratorTsGcTest, NoKeyEligibleForGC) {
   }
 }
 
+TEST_P(CompactionIteratorTsGcTest, NoMergeEligibleForGc) {
+  constexpr char user_key[] = "a";
+  const std::vector<std::string> input_keys = {
+      test::KeyStr(10002, user_key, 102, kTypeMerge),
+      test::KeyStr(10001, user_key, 101, kTypeMerge),
+      test::KeyStr(10000, user_key, 100, kTypeValue)};
+  const std::vector<std::string> input_values = {"2", "1", "a0"};
+  std::shared_ptr<MergeOperator> merge_op =
+      MergeOperators::CreateStringAppendTESTOperator();
+  const std::vector<std::string>& expected_keys = input_keys;
+  const std::vector<std::string>& expected_values = input_values;
+  const std::vector<std::pair<bool, bool>> params = {
+      {false, false}, {false, true}, {true, true}};
+  for (const auto& param : params) {
+    const bool bottommost_level = param.first;
+    const bool key_not_exists_beyond_output_level = param.second;
+    RunTest(input_keys, input_values, expected_keys, expected_values,
+            /*last_committed_seq=*/kMaxSequenceNumber, merge_op.get(),
+            /*compaction_filter=*/nullptr, bottommost_level,
+            /*earliest_write_conflict_snapshot=*/kMaxSequenceNumber,
+            key_not_exists_beyond_output_level,
+            /*full_history_ts_low=*/nullptr);
+  }
+}
+
 TEST_P(CompactionIteratorTsGcTest, AllKeysOlderThanThreshold) {
   constexpr char user_key[][2] = {{'a', '\0'}, {'b', '\0'}};
   const std::vector<std::string> input_keys = {
@@ -1303,6 +1333,91 @@ TEST_P(CompactionIteratorTsGcTest, AllKeysOlderThanThreshold) {
   }
 }
 
+TEST_P(CompactionIteratorTsGcTest, SomeMergesOlderThanThreshold) {
+  constexpr char user_key[][2] = {"a", "f"};
+  const std::vector<std::string> input_keys = {
+      test::KeyStr(/*ts=*/25000, user_key[0], /*seq=*/2500, kTypeMerge),
+      test::KeyStr(/*ts=*/19000, user_key[0], /*seq=*/2300, kTypeMerge),
+      test::KeyStr(/*ts=*/18000, user_key[0], /*seq=*/1800, kTypeMerge),
+      test::KeyStr(/*ts=*/16000, user_key[0], /*seq=*/1600, kTypeValue),
+      test::KeyStr(/*ts=*/19000, user_key[1], /*seq=*/2000, kTypeMerge),
+      test::KeyStr(/*ts=*/17000, user_key[1], /*seq=*/1700, kTypeMerge),
+      test::KeyStr(/*ts=*/15000, user_key[1], /*seq=*/1600,
+                   kTypeDeletionWithTimestamp)};
+  const std::vector<std::string> input_values = {"25", "19", "18", "16",
+                                                 "19", "17", ""};
+  std::shared_ptr<MergeOperator> merge_op =
+      MergeOperators::CreateStringAppendTESTOperator();
+  std::string full_history_ts_low;
+  PutFixed64(&full_history_ts_low, 20000);
+
+  const std::vector<std::pair<bool, bool>> params = {
+      {false, false}, {false, true}, {true, true}};
+
+  {
+    AddSnapshot(1600);
+    AddSnapshot(1900);
+    const std::vector<std::string> expected_keys = {
+        test::KeyStr(/*ts=*/25000, user_key[0], /*seq=*/2500, kTypeMerge),
+        test::KeyStr(/*ts=*/19000, user_key[0], /*seq=*/2300, kTypeMerge),
+        test::KeyStr(/*ts=*/18000, user_key[0], /*seq=*/1800, kTypeMerge),
+        test::KeyStr(/*ts=*/16000, user_key[0], /*seq=*/1600, kTypeValue),
+        test::KeyStr(/*ts=*/19000, user_key[1], /*seq=*/2000, kTypeMerge),
+        test::KeyStr(/*ts=*/17000, user_key[1], /*seq=*/1700, kTypeMerge),
+        test::KeyStr(/*ts=*/15000, user_key[1], /*seq=*/1600,
+                     kTypeDeletionWithTimestamp)};
+    const std::vector<std::string> expected_values = {"25", "19", "18", "16",
+                                                      "19", "17", ""};
+    for (const auto& param : params) {
+      const bool bottommost_level = param.first;
+      const bool key_not_exists_beyond_output_level = param.second;
+      auto expected_keys_copy = expected_keys;
+      auto expected_values_copy = expected_values;
+      if (bottommost_level || key_not_exists_beyond_output_level) {
+        // the kTypeDeletionWithTimestamp will be dropped
+        expected_keys_copy.pop_back();
+        expected_values_copy.pop_back();
+        if (bottommost_level) {
+          // seq zero
+          expected_keys_copy[3] =
+              test::KeyStr(/*ts=*/0, user_key[0], /*seq=*/0, kTypeValue);
+        }
+      }
+      RunTest(input_keys, input_values, expected_keys_copy,
+              expected_values_copy,
+              /*last_committed_seq=*/kMaxSequenceNumber, merge_op.get(),
+              /*compaction_filter=*/nullptr, bottommost_level,
+              /*earliest_write_conflict_snapshot=*/kMaxSequenceNumber,
+              key_not_exists_beyond_output_level, &full_history_ts_low);
+    }
+    ClearSnapshots();
+  }
+
+  // No snapshots
+  {
+    const std::vector<std::string> expected_keys = {
+        test::KeyStr(/*ts=*/25000, user_key[0], /*seq=*/2500, kTypeValue),
+        test::KeyStr(/*ts=*/19000, user_key[1], /*seq=*/2000, kTypeValue)};
+    const std::vector<std::string> expected_values = {"16,18,19,25", "17,19"};
+    for (const auto& param : params) {
+      const bool bottommost_level = param.first;
+      const bool key_not_exists_beyond_output_level = param.second;
+      auto expected_keys_copy = expected_keys;
+      auto expected_values_copy = expected_values;
+      if (bottommost_level) {
+        expected_keys_copy[1] =
+            test::KeyStr(/*ts=*/0, user_key[1], /*seq=*/0, kTypeValue);
+      }
+      RunTest(input_keys, input_values, expected_keys_copy,
+              expected_values_copy,
+              /*last_committed_seq=*/kMaxSequenceNumber, merge_op.get(),
+              /*compaction_filter=*/nullptr, bottommost_level,
+              /*earliest_write_conflict_snapshot=*/kMaxSequenceNumber,
+              key_not_exists_beyond_output_level, &full_history_ts_low);
+    }
+  }
+}
+
 TEST_P(CompactionIteratorTsGcTest, NewHidesOldSameSnapshot) {
   constexpr char user_key[] = "a";
   const std::vector<std::string> input_keys = {

db/compaction/compaction_job.cc

@@ -99,6 +99,8 @@ const char* GetCompactionReasonString(CompactionReason compaction_reason) {
       return "ForcedBlobGC";
     case CompactionReason::kRoundRobinTtl:
       return "RoundRobinTtl";
+    case CompactionReason::kRefitLevel:
+      return "RefitLevel";
     case CompactionReason::kNumOfReasons:
       // fall through
     default:
@@ -190,8 +192,8 @@ CompactionJob::CompactionJob(
   assert(log_buffer_ != nullptr);
 
   const auto* cfd = compact_->compaction->column_family_data();
-  ThreadStatusUtil::SetColumnFamily(cfd, cfd->ioptions()->env,
-                                    db_options_.enable_thread_tracking);
+  ThreadStatusUtil::SetEnableTracking(db_options_.enable_thread_tracking);
+  ThreadStatusUtil::SetColumnFamily(cfd);
   ThreadStatusUtil::SetThreadOperation(ThreadStatus::OP_COMPACTION);
   ReportStartedCompaction(compaction);
 }
@@ -202,10 +204,6 @@ CompactionJob::~CompactionJob() {
 }
 
 void CompactionJob::ReportStartedCompaction(Compaction* compaction) {
-  const auto* cfd = compact_->compaction->column_family_data();
-  ThreadStatusUtil::SetColumnFamily(cfd, cfd->ioptions()->env,
-                                    db_options_.enable_thread_tracking);
-
   ThreadStatusUtil::SetThreadOperationProperty(ThreadStatus::COMPACTION_JOB_ID,
                                                job_id_);
 
@@ -289,12 +287,14 @@ void CompactionJob::Prepare() {
                c->immutable_options()->preclude_last_level_data_seconds);
 
   if (preserve_time_duration > 0) {
+    const ReadOptions read_options(Env::IOActivity::kCompaction);
     // setup seqno_time_mapping_
     seqno_time_mapping_.SetMaxTimeDuration(preserve_time_duration);
     for (const auto& each_level : *c->inputs()) {
       for (const auto& fmd : each_level.files) {
         std::shared_ptr<const TableProperties> tp;
-        Status s = cfd->current()->GetTableProperties(&tp, fmd, nullptr);
+        Status s =
+            cfd->current()->GetTableProperties(read_options, &tp, fmd, nullptr);
         if (s.ok()) {
           seqno_time_mapping_.Add(tp->seqno_to_time_mapping)
               .PermitUncheckedError();
@@ -470,7 +470,7 @@ void CompactionJob::GenSubcompactionBoundaries() {
   // overlap with N-1 other ranges. Since we requested a relatively large number
   // (128) of ranges from each input files, even N range overlapping would
   // cause relatively small inaccuracy.
-
+  const ReadOptions read_options(Env::IOActivity::kCompaction);
   auto* c = compact_->compaction;
   if (c->max_subcompactions() <= 1 &&
       !(c->immutable_options()->compaction_pri == kRoundRobin &&
@@ -504,7 +504,7 @@ void CompactionJob::GenSubcompactionBoundaries() {
         FileMetaData* f = flevel->files[i].file_metadata;
         std::vector<TableReader::Anchor> my_anchors;
         Status s = cfd->table_cache()->ApproximateKeyAnchors(
-            ReadOptions(), icomp, *f, my_anchors);
+            read_options, icomp, *f, my_anchors);
         if (!s.ok() || my_anchors.empty()) {
           my_anchors.emplace_back(f->largest.user_key(), f->fd.GetFileSize());
         }
@@ -714,16 +714,18 @@ Status CompactionJob::Run() {
           break;
         }
         // Verify that the table is usable
-        // We set for_compaction to false and don't OptimizeForCompactionTableRead
-        // here because this is a special case after we finish the table building
-        // No matter whether use_direct_io_for_flush_and_compaction is true,
-        // we will regard this verification as user reads since the goal is
-        // to cache it here for further user reads
-        ReadOptions read_options;
+        // We set for_compaction to false and don't
+        // OptimizeForCompactionTableRead here because this is a special case
+        // after we finish the table building No matter whether
+        // use_direct_io_for_flush_and_compaction is true, we will regard this
+        // verification as user reads since the goal is to cache it here for
+        // further user reads
+        const ReadOptions verify_table_read_options(
+            Env::IOActivity::kCompaction);
         InternalIterator* iter = cfd->table_cache()->NewIterator(
-            read_options, file_options_, cfd->internal_comparator(),
-            files_output[file_idx]->meta, /*range_del_agg=*/nullptr,
-            prefix_extractor,
+            verify_table_read_options, file_options_,
+            cfd->internal_comparator(), files_output[file_idx]->meta,
+            /*range_del_agg=*/nullptr, prefix_extractor,
             /*table_reader_ptr=*/nullptr,
             cfd->internal_stats()->GetFileReadHist(
                 compact_->compaction->output_level()),
@@ -764,8 +766,8 @@ Status CompactionJob::Run() {
       }
     };
     for (size_t i = 1; i < compact_->sub_compact_states.size(); i++) {
-      thread_pool.emplace_back(verify_table,
-                               std::ref(compact_->sub_compact_states[i].status));
+      thread_pool.emplace_back(
+          verify_table, std::ref(compact_->sub_compact_states[i].status));
     }
     verify_table(compact_->sub_compact_states[0].status);
     for (auto& thread : thread_pool) {
@@ -975,7 +977,6 @@ Status CompactionJob::Install(const MutableCFOptions& mutable_cf_options) {
 
 void CompactionJob::NotifyOnSubcompactionBegin(
     SubcompactionState* sub_compact) {
-#ifndef ROCKSDB_LITE
   Compaction* c = compact_->compaction;
 
   if (db_options_.listeners.empty()) {
@@ -1001,14 +1002,10 @@ void CompactionJob::NotifyOnSubcompactionBegin(
   }
   info.status.PermitUncheckedError();
 
-#else
-  (void)sub_compact;
-#endif  // ROCKSDB_LITE
 }
 
 void CompactionJob::NotifyOnSubcompactionCompleted(
     SubcompactionState* sub_compact) {
-#ifndef ROCKSDB_LITE
 
   if (db_options_.listeners.empty()) {
     return;
@@ -1029,16 +1026,11 @@ void CompactionJob::NotifyOnSubcompactionCompleted(
   for (const auto& listener : db_options_.listeners) {
     listener->OnSubcompactionCompleted(info);
   }
-#else
-  (void)sub_compact;
-#endif  // ROCKSDB_LITE
 }
 
 void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
   assert(sub_compact);
   assert(sub_compact->compaction);
-
-#ifndef ROCKSDB_LITE
   if (db_options_.compaction_service) {
     CompactionServiceJobStatus comp_status =
         ProcessKeyValueCompactionWithCompactionService(sub_compact);
@@ -1049,7 +1041,6 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
     // fallback to local compaction
     assert(comp_status == CompactionServiceJobStatus::kUseLocal);
   }
-#endif  // !ROCKSDB_LITE
 
   uint64_t prev_cpu_micros = db_options_.clock->CPUMicros();
 
@@ -1090,6 +1081,7 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
   read_options.verify_checksums = true;
   read_options.fill_cache = false;
   read_options.rate_limiter_priority = GetRateLimiterPriority();
+  read_options.io_activity = Env::IOActivity::kCompaction;
   // Compaction iterators shouldn't be confined to a single prefix.
   // Compactions use Seek() for
   // (a) concurrent compactions,
@@ -1125,6 +1117,8 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
   IterKey end_ikey;
   Slice start_slice;
   Slice end_slice;
+  Slice start_user_key{};
+  Slice end_user_key{};
 
   static constexpr char kMaxTs[] =
       "\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff\xff";
@@ -1147,6 +1141,7 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
                                    &ts_slice);
     }
     start_slice = start_ikey.GetInternalKey();
+    start_user_key = start_ikey.GetUserKey();
   }
   if (end.has_value()) {
     end_ikey.SetInternalKey(end.value(), kMaxSequenceNumber, kValueTypeForSeek);
@@ -1155,6 +1150,7 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
                                  &ts_slice);
     }
     end_slice = end_ikey.GetInternalKey();
+    end_user_key = end_ikey.GetUserKey();
   }
 
   std::unique_ptr<InternalIterator> clip;
@@ -1270,11 +1266,15 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
       [this, sub_compact](CompactionOutputs& outputs) {
         return this->OpenCompactionOutputFile(sub_compact, outputs);
       };
+
   const CompactionFileCloseFunc close_file_func =
-      [this, sub_compact](CompactionOutputs& outputs, const Status& status,
-                          const Slice& next_table_min_key) {
-        return this->FinishCompactionOutputFile(status, sub_compact, outputs,
-                                                next_table_min_key);
+      [this, sub_compact, start_user_key, end_user_key](
+          CompactionOutputs& outputs, const Status& status,
+          const Slice& next_table_min_key) {
+        return this->FinishCompactionOutputFile(
+            status, sub_compact, outputs, next_table_min_key,
+            sub_compact->start.has_value() ? &start_user_key : nullptr,
+            sub_compact->end.has_value() ? &end_user_key : nullptr);
       };
 
   Status status;
@@ -1285,7 +1285,6 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
   while (status.ok() && !cfd->IsDropped() && c_iter->Valid()) {
     // Invariant: c_iter.status() is guaranteed to be OK if c_iter->Valid()
     // returns true.
-
     assert(!end.has_value() || cfd->user_comparator()->Compare(
                                    c_iter->user_key(), end.value()) < 0);
 
@@ -1465,7 +1464,8 @@ void CompactionJob::RecordDroppedKeys(
 
 Status CompactionJob::FinishCompactionOutputFile(
     const Status& input_status, SubcompactionState* sub_compact,
-    CompactionOutputs& outputs, const Slice& next_table_min_key) {
+    CompactionOutputs& outputs, const Slice& next_table_min_key,
+    const Slice* comp_start_user_key, const Slice* comp_end_user_key) {
   AutoThreadOperationStageUpdater stage_updater(
       ThreadStatus::STAGE_COMPACTION_SYNC_FILE);
   assert(sub_compact != nullptr);
@@ -1495,12 +1495,10 @@ Status CompactionJob::FinishCompactionOutputFile(
     // output_to_penultimate_level compaction here, as it's only used to decide
     // if range dels could be dropped.
     if (outputs.HasRangeDel()) {
-      s = outputs.AddRangeDels(
-          sub_compact->start.has_value() ? &(sub_compact->start.value())
-                                         : nullptr,
-          sub_compact->end.has_value() ? &(sub_compact->end.value()) : nullptr,
-          range_del_out_stats, bottommost_level_, cfd->internal_comparator(),
-          earliest_snapshot, next_table_min_key, full_history_ts_low_);
+      s = outputs.AddRangeDels(comp_start_user_key, comp_end_user_key,
+                               range_del_out_stats, bottommost_level_,
+                               cfd->internal_comparator(), earliest_snapshot,
+                               next_table_min_key, full_history_ts_low_);
     }
     RecordDroppedKeys(range_del_out_stats, &sub_compact->compaction_job_stats);
     TEST_SYNC_POINT("CompactionJob::FinishCompactionOutputFile1");
@@ -1612,7 +1610,6 @@ Status CompactionJob::FinishCompactionOutputFile(
       TableFileCreationReason::kCompaction, status_for_listener, file_checksum,
       file_checksum_func_name);
 
-#ifndef ROCKSDB_LITE
   // Report new file to SstFileManagerImpl
   auto sfm =
       static_cast<SstFileManagerImpl*>(db_options_.sst_file_manager.get());
@@ -1631,7 +1628,6 @@ Status CompactionJob::FinishCompactionOutputFile(
       db_error_handler_->SetBGError(s, BackgroundErrorReason::kCompaction);
     }
   }
-#endif
 
   outputs.ResetBuilder();
   return s;
@@ -1643,6 +1639,7 @@ Status CompactionJob::InstallCompactionResults(
 
   db_mutex_->AssertHeld();
 
+  const ReadOptions read_options(Env::IOActivity::kCompaction);
   auto* compaction = compact_->compaction;
   assert(compaction);
 
@@ -1720,8 +1717,8 @@ Status CompactionJob::InstallCompactionResults(
   }
 
   return versions_->LogAndApply(compaction->column_family_data(),
-                                mutable_cf_options, edit, db_mutex_,
-                                db_directory_);
+                                mutable_cf_options, read_options, edit,
+                                db_mutex_, db_directory_);
 }
 
 void CompactionJob::RecordCompactionIOStats() {
@@ -1756,11 +1753,9 @@ Status CompactionJob::OpenCompactionOutputFile(SubcompactionState* sub_compact,
   std::string fname = GetTableFileName(file_number);
   // Fire events.
   ColumnFamilyData* cfd = sub_compact->compaction->column_family_data();
-#ifndef ROCKSDB_LITE
   EventHelpers::NotifyTableFileCreationStarted(
       cfd->ioptions()->listeners, dbname_, cfd->GetName(), fname, job_id_,
       TableFileCreationReason::kCompaction);
-#endif  // !ROCKSDB_LITE
   // Make the output file
   std::unique_ptr<FSWritableFile> writable_file;
 #ifndef NDEBUG
@@ -1833,12 +1828,14 @@ Status CompactionJob::OpenCompactionOutputFile(SubcompactionState* sub_compact,
   }
 
   // Initialize a SubcompactionState::Output and add it to sub_compact->outputs
+  uint64_t epoch_number = sub_compact->compaction->MinInputFileEpochNumber();
   {
     FileMetaData meta;
     meta.fd = FileDescriptor(file_number,
                              sub_compact->compaction->output_path_id(), 0);
     meta.oldest_ancester_time = oldest_ancester_time;
     meta.file_creation_time = current_time;
+    meta.epoch_number = epoch_number;
     meta.temperature = temperature;
     assert(!db_id_.empty());
     assert(!db_session_id_.empty());
@@ -1895,7 +1892,6 @@ void CompactionJob::CleanupCompaction() {
   compact_ = nullptr;
 }
 
-#ifndef ROCKSDB_LITE
 namespace {
 void CopyPrefix(const Slice& src, size_t prefix_length, std::string* dst) {
   assert(prefix_length > 0);
@@ -1904,7 +1900,6 @@ void CopyPrefix(const Slice& src, size_t prefix_length, std::string* dst) {
 }
 }  // namespace
 
-#endif  // !ROCKSDB_LITE
 
 void CompactionJob::UpdateCompactionStats() {
   assert(compact_);
@@ -1951,7 +1946,6 @@ void CompactionJob::UpdateCompactionInputStatsHelper(int* num_files,
 
 void CompactionJob::UpdateCompactionJobStats(
     const InternalStats::CompactionStats& stats) const {
-#ifndef ROCKSDB_LITE
   compaction_job_stats_->elapsed_micros = stats.micros;
 
   // input information
@@ -1978,9 +1972,6 @@ void CompactionJob::UpdateCompactionJobStats(
     CopyPrefix(compact_->LargestUserKey(), CompactionJobStats::kMaxPrefixLength,
                &compaction_job_stats_->largest_output_key_prefix);
   }
-#else
-  (void)stats;
-#endif  // !ROCKSDB_LITE
 }
 
 void CompactionJob::LogCompaction() {

db/compaction/compaction_job.h

@@ -256,17 +256,19 @@ class CompactionJob {
   Status FinishCompactionOutputFile(const Status& input_status,
                                     SubcompactionState* sub_compact,
                                     CompactionOutputs& outputs,
-                                    const Slice& next_table_min_key);
+                                    const Slice& next_table_min_key,
+                                    const Slice* comp_start_user_key,
+                                    const Slice* comp_end_user_key);
   Status InstallCompactionResults(const MutableCFOptions& mutable_cf_options);
   Status OpenCompactionOutputFile(SubcompactionState* sub_compact,
                                   CompactionOutputs& outputs);
   void UpdateCompactionJobStats(
-    const InternalStats::CompactionStats& stats) const;
+      const InternalStats::CompactionStats& stats) const;
   void RecordDroppedKeys(const CompactionIterationStats& c_iter_stats,
                          CompactionJobStats* compaction_job_stats = nullptr);
 
-  void UpdateCompactionInputStatsHelper(
-      int* num_files, uint64_t* bytes_read, int input_level);
+  void UpdateCompactionInputStatsHelper(int* num_files, uint64_t* bytes_read,
+                                        int input_level);
 
   void NotifyOnSubcompactionBegin(SubcompactionState* sub_compact);
 
@@ -402,6 +404,7 @@ struct CompactionServiceOutputFile {
   std::string largest_internal_key;
   uint64_t oldest_ancester_time;
   uint64_t file_creation_time;
+  uint64_t epoch_number;
   uint64_t paranoid_hash;
   bool marked_for_compaction;
   UniqueId64x2 unique_id;
@@ -411,8 +414,8 @@ struct CompactionServiceOutputFile {
       const std::string& name, SequenceNumber smallest, SequenceNumber largest,
       std::string _smallest_internal_key, std::string _largest_internal_key,
       uint64_t _oldest_ancester_time, uint64_t _file_creation_time,
-      uint64_t _paranoid_hash, bool _marked_for_compaction,
-      UniqueId64x2 _unique_id)
+      uint64_t _epoch_number, uint64_t _paranoid_hash,
+      bool _marked_for_compaction, UniqueId64x2 _unique_id)
       : file_name(name),
         smallest_seqno(smallest),
         largest_seqno(largest),
@@ -420,6 +423,7 @@ struct CompactionServiceOutputFile {
         largest_internal_key(std::move(_largest_internal_key)),
         oldest_ancester_time(_oldest_ancester_time),
         file_creation_time(_file_creation_time),
+        epoch_number(_epoch_number),
         paranoid_hash(_paranoid_hash),
         marked_for_compaction(_marked_for_compaction),
         unique_id(std::move(_unique_id)) {}

db/compaction/compaction_job_stats_test.cc

@@ -59,7 +59,6 @@
 #include "utilities/merge_operators.h"
 
 #if !defined(IOS_CROSS_COMPILE)
-#ifndef ROCKSDB_LITE
 namespace ROCKSDB_NAMESPACE {
 
 static std::string RandomString(Random* rnd, int len, double ratio) {
@@ -155,9 +154,8 @@ class CompactionJobStatsTest : public testing::Test,
     ASSERT_OK(TryReopenWithColumnFamilies(cfs, options));
   }
 
-  Status TryReopenWithColumnFamilies(
-      const std::vector<std::string>& cfs,
-      const std::vector<Options>& options) {
+  Status TryReopenWithColumnFamilies(const std::vector<std::string>& cfs,
+                                     const std::vector<Options>& options) {
     Close();
     EXPECT_EQ(cfs.size(), options.size());
     std::vector<ColumnFamilyDescriptor> column_families;
@@ -175,9 +173,7 @@ class CompactionJobStatsTest : public testing::Test,
     return TryReopenWithColumnFamilies(cfs, v_opts);
   }
 
-  void Reopen(const Options& options) {
-    ASSERT_OK(TryReopen(options));
-  }
+  void Reopen(const Options& options) { ASSERT_OK(TryReopen(options)); }
 
   void Close() {
     for (auto h : handles_) {
@@ -226,9 +222,7 @@ class CompactionJobStatsTest : public testing::Test,
     return db_->Put(wo, handles_[cf], k, v);
   }
 
-  Status Delete(const std::string& k) {
-    return db_->Delete(WriteOptions(), k);
-  }
+  Status Delete(const std::string& k) { return db_->Delete(WriteOptions(), k); }
 
   Status Delete(int cf, const std::string& k) {
     return db_->Delete(WriteOptions(), handles_[cf], k);
@@ -338,21 +332,21 @@ class CompactionJobStatsTest : public testing::Test,
     }
   }
 
-  static void SetDeletionCompactionStats(
-      CompactionJobStats *stats, uint64_t input_deletions,
-      uint64_t expired_deletions, uint64_t records_replaced) {
+  static void SetDeletionCompactionStats(CompactionJobStats* stats,
+                                         uint64_t input_deletions,
+                                         uint64_t expired_deletions,
+                                         uint64_t records_replaced) {
     stats->num_input_deletion_records = input_deletions;
     stats->num_expired_deletion_records = expired_deletions;
     stats->num_records_replaced = records_replaced;
   }
 
-  void MakeTableWithKeyValues(
-    Random* rnd, uint64_t smallest, uint64_t largest,
-    int key_size, int value_size, uint64_t interval,
-    double ratio, int cf = 0) {
+  void MakeTableWithKeyValues(Random* rnd, uint64_t smallest, uint64_t largest,
+                              int key_size, int value_size, uint64_t interval,
+                              double ratio, int cf = 0) {
     for (auto key = smallest; key < largest; key += interval) {
       ASSERT_OK(Put(cf, Slice(Key(key, key_size)),
-                        Slice(RandomString(rnd, value_size, ratio))));
+                    Slice(RandomString(rnd, value_size, ratio))));
     }
     ASSERT_OK(Flush(cf));
   }
@@ -361,9 +355,9 @@ class CompactionJobStatsTest : public testing::Test,
   // rounds of keys are inserted into the database, as per the behavior
   // of the DeletionStatsTest.
   void SelectivelyDeleteKeys(uint64_t smallest, uint64_t largest,
-    uint64_t interval, int deletion_interval, int key_size,
-    uint64_t cutoff_key_num, CompactionJobStats* stats, int cf = 0) {
-
+                             uint64_t interval, int deletion_interval,
+                             int key_size, uint64_t cutoff_key_num,
+                             CompactionJobStats* stats, int cf = 0) {
     // interval needs to be >= 2 so that deletion entries can be inserted
     // that are intended to not result in an actual key deletion by using
     // an offset of 1 from another existing key
@@ -387,20 +381,19 @@ class CompactionJobStatsTest : public testing::Test,
 
     // Insert some deletions for keys that don't exist that
     // are both in and out of the key range
-    ASSERT_OK(Delete(cf, Key(smallest+1, key_size)));
+    ASSERT_OK(Delete(cf, Key(smallest + 1, key_size)));
     deletions_made++;
 
-    ASSERT_OK(Delete(cf, Key(smallest-1, key_size)));
+    ASSERT_OK(Delete(cf, Key(smallest - 1, key_size)));
     deletions_made++;
     num_expired++;
 
-    ASSERT_OK(Delete(cf, Key(smallest-9, key_size)));
+    ASSERT_OK(Delete(cf, Key(smallest - 9, key_size)));
     deletions_made++;
     num_expired++;
 
     ASSERT_OK(Flush(cf));
-    SetDeletionCompactionStats(stats, deletions_made, num_expired,
-      num_deleted);
+    SetDeletionCompactionStats(stats, deletions_made, num_expired, num_deleted);
   }
 };
 
@@ -440,25 +433,20 @@ class CompactionJobStatsChecker : public EventListener {
   // use ASSERT_GE and ASSERT_LE with a reasonable bias ---
   // 10% in uncompressed case and 20% when compression is used.
   virtual void Verify(const CompactionJobStats& current_stats,
-              const CompactionJobStats& stats) {
+                      const CompactionJobStats& stats) {
     // time
     ASSERT_GT(current_stats.elapsed_micros, 0U);
 
-    ASSERT_EQ(current_stats.num_input_records,
-        stats.num_input_records);
-    ASSERT_EQ(current_stats.num_input_files,
-        stats.num_input_files);
+    ASSERT_EQ(current_stats.num_input_records, stats.num_input_records);
+    ASSERT_EQ(current_stats.num_input_files, stats.num_input_files);
     ASSERT_EQ(current_stats.num_input_files_at_output_level,
-        stats.num_input_files_at_output_level);
+              stats.num_input_files_at_output_level);
 
-    ASSERT_EQ(current_stats.num_output_records,
-        stats.num_output_records);
-    ASSERT_EQ(current_stats.num_output_files,
-        stats.num_output_files);
+    ASSERT_EQ(current_stats.num_output_records, stats.num_output_records);
+    ASSERT_EQ(current_stats.num_output_files, stats.num_output_files);
 
     ASSERT_EQ(current_stats.is_full_compaction, stats.is_full_compaction);
-    ASSERT_EQ(current_stats.is_manual_compaction,
-        stats.is_manual_compaction);
+    ASSERT_EQ(current_stats.is_manual_compaction, stats.is_manual_compaction);
 
     // file size
     double kFileSizeBias = compression_enabled_ ? 0.20 : 0.10;
@@ -475,18 +463,14 @@ class CompactionJobStatsChecker : public EventListener {
     ASSERT_EQ(current_stats.total_input_raw_value_bytes,
               stats.total_input_raw_value_bytes);
 
-    ASSERT_EQ(current_stats.num_records_replaced,
-        stats.num_records_replaced);
+    ASSERT_EQ(current_stats.num_records_replaced, stats.num_records_replaced);
 
-    ASSERT_EQ(current_stats.num_corrupt_keys,
-        stats.num_corrupt_keys);
+    ASSERT_EQ(current_stats.num_corrupt_keys, stats.num_corrupt_keys);
 
-    ASSERT_EQ(
-        std::string(current_stats.smallest_output_key_prefix),
-        std::string(stats.smallest_output_key_prefix));
-    ASSERT_EQ(
-        std::string(current_stats.largest_output_key_prefix),
-        std::string(stats.largest_output_key_prefix));
+    ASSERT_EQ(std::string(current_stats.smallest_output_key_prefix),
+              std::string(stats.smallest_output_key_prefix));
+    ASSERT_EQ(std::string(current_stats.largest_output_key_prefix),
+              std::string(stats.largest_output_key_prefix));
   }
 
   // Add an expected compaction stats, which will be used to
@@ -497,9 +481,7 @@ class CompactionJobStatsChecker : public EventListener {
     expected_stats_.push(stats);
   }
 
-  void EnableCompression(bool flag) {
-    compression_enabled_ = flag;
-  }
+  void EnableCompression(bool flag) { compression_enabled_ = flag; }
 
   bool verify_next_comp_io_stats() const { return verify_next_comp_io_stats_; }
 
@@ -517,45 +499,37 @@ class CompactionJobDeletionStatsChecker : public CompactionJobStatsChecker {
   // Verifies whether two CompactionJobStats match.
   void Verify(const CompactionJobStats& current_stats,
               const CompactionJobStats& stats) override {
-    ASSERT_EQ(
-      current_stats.num_input_deletion_records,
-      stats.num_input_deletion_records);
-    ASSERT_EQ(
-        current_stats.num_expired_deletion_records,
-        stats.num_expired_deletion_records);
-    ASSERT_EQ(
-        current_stats.num_records_replaced,
-        stats.num_records_replaced);
+    ASSERT_EQ(current_stats.num_input_deletion_records,
+              stats.num_input_deletion_records);
+    ASSERT_EQ(current_stats.num_expired_deletion_records,
+              stats.num_expired_deletion_records);
+    ASSERT_EQ(current_stats.num_records_replaced, stats.num_records_replaced);
 
-    ASSERT_EQ(current_stats.num_corrupt_keys,
-        stats.num_corrupt_keys);
+    ASSERT_EQ(current_stats.num_corrupt_keys, stats.num_corrupt_keys);
   }
 };
 
 namespace {
 
-uint64_t EstimatedFileSize(
-    uint64_t num_records, size_t key_size, size_t value_size,
-    double compression_ratio = 1.0,
-    size_t block_size = 4096,
-    int bloom_bits_per_key = 10) {
+uint64_t EstimatedFileSize(uint64_t num_records, size_t key_size,
+                           size_t value_size, double compression_ratio = 1.0,
+                           size_t block_size = 4096,
+                           int bloom_bits_per_key = 10) {
   const size_t kPerKeyOverhead = 8;
   const size_t kFooterSize = 512;
 
-  uint64_t data_size =
-    static_cast<uint64_t>(
-      num_records * (key_size + value_size * compression_ratio +
-                     kPerKeyOverhead));
+  uint64_t data_size = static_cast<uint64_t>(
+      num_records *
+      (key_size + value_size * compression_ratio + kPerKeyOverhead));
 
-  return data_size + kFooterSize
-         + num_records * bloom_bits_per_key / 8      // filter block
+  return data_size + kFooterSize +
+         num_records * bloom_bits_per_key / 8        // filter block
          + data_size * (key_size + 8) / block_size;  // index block
 }
 
 namespace {
 
-void CopyPrefix(
-    const Slice& src, size_t prefix_length, std::string* dst) {
+void CopyPrefix(const Slice& src, size_t prefix_length, std::string* dst) {
   assert(prefix_length > 0);
   size_t length = src.size() > prefix_length ? prefix_length : src.size();
   dst->assign(src.data(), length);
@@ -581,28 +555,24 @@ CompactionJobStats NewManualCompactionJobStats(
   stats.num_output_files = num_output_files;
 
   stats.total_input_bytes =
-      EstimatedFileSize(
-          num_input_records / num_input_files,
-          key_size, value_size, compression_ratio) * num_input_files;
+      EstimatedFileSize(num_input_records / num_input_files, key_size,
+                        value_size, compression_ratio) *
+      num_input_files;
   stats.total_output_bytes =
-      EstimatedFileSize(
-          num_output_records / num_output_files,
-          key_size, value_size, compression_ratio) * num_output_files;
-  stats.total_input_raw_key_bytes =
-      num_input_records * (key_size + 8);
-  stats.total_input_raw_value_bytes =
-      num_input_records * value_size;
+      EstimatedFileSize(num_output_records / num_output_files, key_size,
+                        value_size, compression_ratio) *
+      num_output_files;
+  stats.total_input_raw_key_bytes = num_input_records * (key_size + 8);
+  stats.total_input_raw_value_bytes = num_input_records * value_size;
 
   stats.is_full_compaction = is_full;
   stats.is_manual_compaction = is_manual;
 
   stats.num_records_replaced = num_records_replaced;
 
-  CopyPrefix(smallest_key,
-             CompactionJobStats::kMaxPrefixLength,
+  CopyPrefix(smallest_key, CompactionJobStats::kMaxPrefixLength,
              &stats.smallest_output_key_prefix);
-  CopyPrefix(largest_key,
-             CompactionJobStats::kMaxPrefixLength,
+  CopyPrefix(largest_key, CompactionJobStats::kMaxPrefixLength,
              &stats.largest_output_key_prefix);
 
   return stats;
@@ -662,13 +632,11 @@ TEST_P(CompactionJobStatsTest, CompactionJobStatsTest) {
 
     // 1st Phase: generate "num_L0_files" L0 files.
     int num_L0_files = 0;
-    for (uint64_t start_key = key_base;
-                  start_key <= key_base * kTestScale;
-                  start_key += key_base) {
-      MakeTableWithKeyValues(
-          &rnd, start_key, start_key + key_base - 1,
-          kKeySize, kValueSize, key_interval,
-          compression_ratio, 1);
+    for (uint64_t start_key = key_base; start_key <= key_base * kTestScale;
+         start_key += key_base) {
+      MakeTableWithKeyValues(&rnd, start_key, start_key + key_base - 1,
+                             kKeySize, kValueSize, key_interval,
+                             compression_ratio, 1);
       snprintf(buf, kBufSize, "%d", ++num_L0_files);
       ASSERT_EQ(std::string(buf), FilesPerLevel(1));
     }
@@ -684,13 +652,9 @@ TEST_P(CompactionJobStatsTest, CompactionJobStatsTest) {
          start_key += key_base, count++) {
       smallest_key = Key(start_key, 10);
       largest_key = Key(start_key + key_base - key_interval, 10);
-      stats_checker->AddExpectedStats(
-          NewManualCompactionJobStats(
-              smallest_key, largest_key,
-              1, 0, num_keys_per_L0_file,
-              kKeySize, kValueSize,
-              1, num_keys_per_L0_file,
-              compression_ratio, 0));
+      stats_checker->AddExpectedStats(NewManualCompactionJobStats(
+          smallest_key, largest_key, 1, 0, num_keys_per_L0_file, kKeySize,
+          kValueSize, 1, num_keys_per_L0_file, compression_ratio, 0));
       ASSERT_EQ(stats_checker->NumberOfUnverifiedStats(), 1U);
       TEST_Compact(0, 1, smallest_key, largest_key);
       snprintf(buf, kBufSize, "%d,%d", num_L0_files - count, count);
@@ -701,14 +665,10 @@ TEST_P(CompactionJobStatsTest, CompactionJobStatsTest) {
     int num_remaining_L0 = num_L0_files - L0_compaction_count;
     smallest_key = Key(key_base * (L0_compaction_count + 1), 10);
     largest_key = Key(key_base * (kTestScale + 1) - key_interval, 10);
-    stats_checker->AddExpectedStats(
-        NewManualCompactionJobStats(
-            smallest_key, largest_key,
-            num_remaining_L0,
-            0, num_keys_per_L0_file * num_remaining_L0,
-            kKeySize, kValueSize,
-            1, num_keys_per_L0_file * num_remaining_L0,
-            compression_ratio, 0));
+    stats_checker->AddExpectedStats(NewManualCompactionJobStats(
+        smallest_key, largest_key, num_remaining_L0, 0,
+        num_keys_per_L0_file * num_remaining_L0, kKeySize, kValueSize, 1,
+        num_keys_per_L0_file * num_remaining_L0, compression_ratio, 0));
     ASSERT_EQ(stats_checker->NumberOfUnverifiedStats(), 1U);
     TEST_Compact(0, 1, smallest_key, largest_key);
 
@@ -719,13 +679,11 @@ TEST_P(CompactionJobStatsTest, CompactionJobStatsTest) {
 
     // 3rd Phase: generate sparse L0 files (wider key-range, same num of keys)
     int sparseness = 2;
-    for (uint64_t start_key = key_base;
-                  start_key <= key_base * kTestScale;
-                  start_key += key_base * sparseness) {
+    for (uint64_t start_key = key_base; start_key <= key_base * kTestScale;
+         start_key += key_base * sparseness) {
       MakeTableWithKeyValues(
-          &rnd, start_key, start_key + key_base * sparseness - 1,
-          kKeySize, kValueSize,
-          key_base * sparseness / num_keys_per_L0_file,
+          &rnd, start_key, start_key + key_base * sparseness - 1, kKeySize,
+          kValueSize, key_base * sparseness / num_keys_per_L0_file,
           compression_ratio, 1);
       snprintf(buf, kBufSize, "%d,%d", ++num_L0_files, num_L1_files);
       ASSERT_EQ(std::string(buf), FilesPerLevel(1));
@@ -737,21 +695,15 @@ TEST_P(CompactionJobStatsTest, CompactionJobStatsTest) {
     // output files without coordinating to see if the output could fit into
     // a smaller number of files like it does when it runs sequentially
     int num_output_files = options.max_subcompactions > 1 ? 2 : 1;
-    for (uint64_t start_key = key_base;
-         num_L0_files > 1;
+    for (uint64_t start_key = key_base; num_L0_files > 1;
          start_key += key_base * sparseness) {
       smallest_key = Key(start_key, 10);
-      largest_key =
-          Key(start_key + key_base * sparseness - key_interval, 10);
-      stats_checker->AddExpectedStats(
-          NewManualCompactionJobStats(
-              smallest_key, largest_key,
-              3, 2, num_keys_per_L0_file * 3,
-              kKeySize, kValueSize,
-              num_output_files,
-              num_keys_per_L0_file * 2,  // 1/3 of the data will be updated.
-              compression_ratio,
-              num_keys_per_L0_file));
+      largest_key = Key(start_key + key_base * sparseness - key_interval, 10);
+      stats_checker->AddExpectedStats(NewManualCompactionJobStats(
+          smallest_key, largest_key, 3, 2, num_keys_per_L0_file * 3, kKeySize,
+          kValueSize, num_output_files,
+          num_keys_per_L0_file * 2,  // 1/3 of the data will be updated.
+          compression_ratio, num_keys_per_L0_file));
       ASSERT_EQ(stats_checker->NumberOfUnverifiedStats(), 1U);
       Compact(1, smallest_key, largest_key);
       if (options.max_subcompactions == 1) {
@@ -766,14 +718,10 @@ TEST_P(CompactionJobStatsTest, CompactionJobStatsTest) {
     // In the first sub-compaction, we expect L0 compaction.
     smallest_key = Key(key_base, 10);
     largest_key = Key(key_base * (kTestScale + 1) - key_interval, 10);
-    stats_checker->AddExpectedStats(
-        NewManualCompactionJobStats(
-            Key(key_base * (kTestScale + 1 - sparseness), 10), largest_key,
-            2, 1, num_keys_per_L0_file * 3,
-            kKeySize, kValueSize,
-            1, num_keys_per_L0_file * 2,
-            compression_ratio,
-            num_keys_per_L0_file));
+    stats_checker->AddExpectedStats(NewManualCompactionJobStats(
+        Key(key_base * (kTestScale + 1 - sparseness), 10), largest_key, 2, 1,
+        num_keys_per_L0_file * 3, kKeySize, kValueSize, 1,
+        num_keys_per_L0_file * 2, compression_ratio, num_keys_per_L0_file));
     ASSERT_EQ(stats_checker->NumberOfUnverifiedStats(), 1U);
     Compact(1, smallest_key, largest_key);
 
@@ -869,7 +817,7 @@ TEST_P(CompactionJobStatsTest, DeletionStatsTest) {
   Options options;
   options.listeners.emplace_back(stats_checker);
   options.create_if_missing = true;
-  options.level0_file_num_compaction_trigger = kTestScale+1;
+  options.level0_file_num_compaction_trigger = kTestScale + 1;
   options.num_levels = 3;
   options.compression = kNoCompression;
   options.max_bytes_for_level_multiplier = 2;
@@ -881,13 +829,10 @@ TEST_P(CompactionJobStatsTest, DeletionStatsTest) {
   // Stage 1: Generate several L0 files and then send them to L2 by
   // using CompactRangeOptions and CompactRange(). These files will
   // have a strict subset of the keys from the full key-range
-  for (uint64_t start_key = key_base;
-                start_key <= key_base * kTestScale / 2;
-                start_key += key_base) {
-    MakeTableWithKeyValues(
-        &rnd, start_key, start_key + key_base - 1,
-        kKeySize, kValueSize, key_interval,
-        compression_ratio, 1);
+  for (uint64_t start_key = key_base; start_key <= key_base * kTestScale / 2;
+       start_key += key_base) {
+    MakeTableWithKeyValues(&rnd, start_key, start_key + key_base - 1, kKeySize,
+                           kValueSize, key_interval, compression_ratio, 1);
   }
 
   CompactRangeOptions cr_options;
@@ -897,13 +842,10 @@ TEST_P(CompactionJobStatsTest, DeletionStatsTest) {
   ASSERT_GT(NumTableFilesAtLevel(2, 1), 0);
 
   // Stage 2: Generate files including keys from the entire key range
-  for (uint64_t start_key = key_base;
-                start_key <= key_base * kTestScale;
-                start_key += key_base) {
-    MakeTableWithKeyValues(
-        &rnd, start_key, start_key + key_base - 1,
-        kKeySize, kValueSize, key_interval,
-        compression_ratio, 1);
+  for (uint64_t start_key = key_base; start_key <= key_base * kTestScale;
+       start_key += key_base) {
+    MakeTableWithKeyValues(&rnd, start_key, start_key + key_base - 1, kKeySize,
+                           kValueSize, key_interval, compression_ratio, 1);
   }
 
   // Send these L0 files to L1
@@ -912,16 +854,16 @@ TEST_P(CompactionJobStatsTest, DeletionStatsTest) {
 
   // Add a new record and flush so now there is a L0 file
   // with a value too (not just deletions from the next step)
-  ASSERT_OK(Put(1, Key(key_base-6, kKeySize), "test"));
+  ASSERT_OK(Put(1, Key(key_base - 6, kKeySize), "test"));
   ASSERT_OK(Flush(1));
 
   // Stage 3: Generate L0 files with some deletions so now
   // there are files with the same key range in L0, L1, and L2
   int deletion_interval = 3;
   CompactionJobStats first_compaction_stats;
-  SelectivelyDeleteKeys(key_base, largest_key_num,
-      key_interval, deletion_interval, kKeySize, cutoff_key_num,
-      &first_compaction_stats, 1);
+  SelectivelyDeleteKeys(key_base, largest_key_num, key_interval,
+                        deletion_interval, kKeySize, cutoff_key_num,
+                        &first_compaction_stats, 1);
 
   stats_checker->AddExpectedStats(first_compaction_stats);
 
@@ -932,8 +874,7 @@ TEST_P(CompactionJobStatsTest, DeletionStatsTest) {
 namespace {
 int GetUniversalCompactionInputUnits(uint32_t num_flushes) {
   uint32_t compaction_input_units;
-  for (compaction_input_units = 1;
-       num_flushes >= compaction_input_units;
+  for (compaction_input_units = 1; num_flushes >= compaction_input_units;
        compaction_input_units *= 2) {
     if ((num_flushes & compaction_input_units) != 0) {
       return compaction_input_units > 1 ? compaction_input_units : 0;
@@ -998,13 +939,10 @@ TEST_P(CompactionJobStatsTest, UniversalCompactionTest) {
   }
   ASSERT_EQ(stats_checker->NumberOfUnverifiedStats(), 3U);
 
-  for (uint64_t start_key = key_base;
-                start_key <= key_base * kTestScale;
-                start_key += key_base) {
-    MakeTableWithKeyValues(
-        &rnd, start_key, start_key + key_base - 1,
-        kKeySize, kValueSize, key_interval,
-        compression_ratio, 1);
+  for (uint64_t start_key = key_base; start_key <= key_base * kTestScale;
+       start_key += key_base) {
+    MakeTableWithKeyValues(&rnd, start_key, start_key + key_base - 1, kKeySize,
+                           kValueSize, key_interval, compression_ratio, 1);
     ASSERT_OK(static_cast_with_check<DBImpl>(db_)->TEST_WaitForCompact());
   }
   ASSERT_EQ(stats_checker->NumberOfUnverifiedStats(), 0U);
@@ -1020,15 +958,6 @@ int main(int argc, char** argv) {
   return RUN_ALL_TESTS();
 }
 
-#else
-#include <stdio.h>
-
-int main(int /*argc*/, char** /*argv*/) {
-  fprintf(stderr, "SKIPPED, not supported in ROCKSDB_LITE\n");
-  return 0;
-}
-
-#endif  // !ROCKSDB_LITE
 
 #else
 

db/compaction/compaction_job_test.cc

@@ -3,7 +3,6 @@
 //  COPYING file in the root directory) and Apache 2.0 License
 //  (found in the LICENSE.Apache file in the root directory).
 
-#ifndef ROCKSDB_LITE
 
 #include "db/compaction/compaction_job.h"
 
@@ -40,7 +39,7 @@ namespace ROCKSDB_NAMESPACE {
 namespace {
 
 void VerifyInitializationOfCompactionJobStats(
-      const CompactionJobStats& compaction_job_stats) {
+    const CompactionJobStats& compaction_job_stats) {
 #if !defined(IOS_CROSS_COMPILE)
   ASSERT_EQ(compaction_job_stats.elapsed_micros, 0U);
 
@@ -380,16 +379,19 @@ class CompactionJobTestBase : public testing::Test {
     }
 
     VersionEdit edit;
-    edit.AddFile(level, file_number, 0, file_size, smallest_key, largest_key,
-                 smallest_seqno, largest_seqno, false, Temperature::kUnknown,
-                 oldest_blob_file_number, kUnknownOldestAncesterTime,
-                 kUnknownFileCreationTime, kUnknownFileChecksum,
-                 kUnknownFileChecksumFuncName, kNullUniqueId64x2);
+    edit.AddFile(
+        level, file_number, 0, file_size, smallest_key, largest_key,
+        smallest_seqno, largest_seqno, false, Temperature::kUnknown,
+        oldest_blob_file_number, kUnknownOldestAncesterTime,
+        kUnknownFileCreationTime,
+        versions_->GetColumnFamilySet()->GetDefault()->NewEpochNumber(),
+        kUnknownFileChecksum, kUnknownFileChecksumFuncName, kNullUniqueId64x2,
+        0);
 
     mutex_.Lock();
-    EXPECT_OK(
-        versions_->LogAndApply(versions_->GetColumnFamilySet()->GetDefault(),
-                               mutable_cf_options_, &edit, &mutex_, nullptr));
+    EXPECT_OK(versions_->LogAndApply(
+        versions_->GetColumnFamilySet()->GetDefault(), mutable_cf_options_,
+        read_options_, &edit, &mutex_, nullptr));
     mutex_.Unlock();
   }
 
@@ -499,8 +501,7 @@ class CompactionJobTestBase : public testing::Test {
 
         // This is how the key will look like once it's written in bottommost
         // file
-        InternalKey bottommost_internal_key(
-            key, 0, kTypeValue);
+        InternalKey bottommost_internal_key(key, 0, kTypeValue);
 
         if (corrupt_id(k)) {
           test::CorruptKeyType(&internal_key);
@@ -620,7 +621,7 @@ class CompactionJobTestBase : public testing::Test {
       CompactionInputFiles compaction_level;
       compaction_level.level = input_levels[i];
       compaction_level.files.insert(compaction_level.files.end(),
-          level_files.begin(), level_files.end());
+                                    level_files.begin(), level_files.end());
       compaction_input_files.push_back(compaction_level);
       num_input_files += level_files.size();
     }
@@ -726,6 +727,7 @@ class CompactionJobTestBase : public testing::Test {
   ColumnFamilyOptions cf_options_;
   MutableCFOptions mutable_cf_options_;
   MutableDBOptions mutable_db_options_;
+  const ReadOptions read_options_;
   std::shared_ptr<Cache> table_cache_;
   WriteController write_controller_;
   WriteBufferManager write_buffer_manager_;
@@ -1656,7 +1658,7 @@ TEST_F(CompactionJobTest, ResultSerialization) {
         rnd.RandomBinaryString(rnd.Uniform(kStrMaxLen)),
         rnd.RandomBinaryString(rnd.Uniform(kStrMaxLen)),
         rnd64.Uniform(UINT64_MAX), rnd64.Uniform(UINT64_MAX),
-        rnd64.Uniform(UINT64_MAX), rnd.OneIn(2), id);
+        rnd64.Uniform(UINT64_MAX), rnd64.Uniform(UINT64_MAX), rnd.OneIn(2), id);
   }
   result.output_level = rnd.Uniform(10);
   result.output_path = rnd.RandomString(rnd.Uniform(kStrMaxLen));
@@ -2439,14 +2441,3 @@ int main(int argc, char** argv) {
   RegisterCustomObjects(argc, argv);
   return RUN_ALL_TESTS();
 }
-
-#else
-#include <stdio.h>
-
-int main(int /*argc*/, char** /*argv*/) {
-  fprintf(stderr,
-          "SKIPPED as CompactionJobStats is not supported in ROCKSDB_LITE\n");
-  return 0;
-}
-
-#endif  // ROCKSDB_LITE

db/compaction/compaction_outputs.cc

@@ -76,6 +76,46 @@ IOStatus CompactionOutputs::WriterSyncClose(const Status& input_status,
   return io_s;
 }
 
+bool CompactionOutputs::UpdateFilesToCutForTTLStates(
+    const Slice& internal_key) {
+  if (!files_to_cut_for_ttl_.empty()) {
+    const InternalKeyComparator* icmp =
+        &compaction_->column_family_data()->internal_comparator();
+    if (cur_files_to_cut_for_ttl_ != -1) {
+      // Previous key is inside the range of a file
+      if (icmp->Compare(internal_key,
+                        files_to_cut_for_ttl_[cur_files_to_cut_for_ttl_]
+                            ->largest.Encode()) > 0) {
+        next_files_to_cut_for_ttl_ = cur_files_to_cut_for_ttl_ + 1;
+        cur_files_to_cut_for_ttl_ = -1;
+        return true;
+      }
+    } else {
+      // Look for the key position
+      while (next_files_to_cut_for_ttl_ <
+             static_cast<int>(files_to_cut_for_ttl_.size())) {
+        if (icmp->Compare(internal_key,
+                          files_to_cut_for_ttl_[next_files_to_cut_for_ttl_]
+                              ->smallest.Encode()) >= 0) {
+          if (icmp->Compare(internal_key,
+                            files_to_cut_for_ttl_[next_files_to_cut_for_ttl_]
+                                ->largest.Encode()) <= 0) {
+            // With in the current file
+            cur_files_to_cut_for_ttl_ = next_files_to_cut_for_ttl_;
+            return true;
+          }
+          // Beyond the current file
+          next_files_to_cut_for_ttl_++;
+        } else {
+          // Still fall into the gap
+          break;
+        }
+      }
+    }
+  }
+  return false;
+}
+
 size_t CompactionOutputs::UpdateGrandparentBoundaryInfo(
     const Slice& internal_key) {
   size_t curr_key_boundary_switched_num = 0;
@@ -185,18 +225,39 @@ uint64_t CompactionOutputs::GetCurrentKeyGrandparentOverlappedBytes(
 
 bool CompactionOutputs::ShouldStopBefore(const CompactionIterator& c_iter) {
   assert(c_iter.Valid());
-
-  // always update grandparent information like overlapped file number, size
-  // etc.
   const Slice& internal_key = c_iter.key();
+#ifndef NDEBUG
+  bool should_stop = false;
+  std::pair<bool*, const Slice> p{&should_stop, internal_key};
+  TEST_SYNC_POINT_CALLBACK(
+      "CompactionOutputs::ShouldStopBefore::manual_decision", (void*)&p);
+  if (should_stop) {
+    return true;
+  }
+#endif  // NDEBUG
   const uint64_t previous_overlapped_bytes = grandparent_overlapped_bytes_;
-  size_t num_grandparent_boundaries_crossed =
-      UpdateGrandparentBoundaryInfo(internal_key);
+  const InternalKeyComparator* icmp =
+      &compaction_->column_family_data()->internal_comparator();
+  size_t num_grandparent_boundaries_crossed = 0;
+  bool should_stop_for_ttl = false;
+  // Always update grandparent information like overlapped file number, size
+  // etc., and TTL states.
+  // If compaction_->output_level() == 0, there is no need to update grandparent
+  // info, and that `grandparent` should be empty.
+  if (compaction_->output_level() > 0) {
+    num_grandparent_boundaries_crossed =
+        UpdateGrandparentBoundaryInfo(internal_key);
+    should_stop_for_ttl = UpdateFilesToCutForTTLStates(internal_key);
+  }
 
   if (!HasBuilder()) {
     return false;
   }
 
+  if (should_stop_for_ttl) {
+    return true;
+  }
+
   // If there's user defined partitioner, check that first
   if (partitioner_ && partitioner_->ShouldPartition(PartitionerRequest(
                           last_key_for_partitioner_, c_iter.user_key(),
@@ -214,9 +275,6 @@ bool CompactionOutputs::ShouldStopBefore(const CompactionIterator& c_iter) {
     return true;
   }
 
-  const InternalKeyComparator* icmp =
-      &compaction_->column_family_data()->internal_comparator();
-
   // Check if it needs to split for RoundRobin
   // Invalid local_output_split_key indicates that we do not need to split
   if (local_output_split_key_ != nullptr && !is_split_) {
@@ -290,41 +348,6 @@ bool CompactionOutputs::ShouldStopBefore(const CompactionIterator& c_iter) {
     }
   }
 
-  // check ttl file boundaries if there's any
-  if (!files_to_cut_for_ttl_.empty()) {
-    if (cur_files_to_cut_for_ttl_ != -1) {
-      // Previous key is inside the range of a file
-      if (icmp->Compare(internal_key,
-                        files_to_cut_for_ttl_[cur_files_to_cut_for_ttl_]
-                            ->largest.Encode()) > 0) {
-        next_files_to_cut_for_ttl_ = cur_files_to_cut_for_ttl_ + 1;
-        cur_files_to_cut_for_ttl_ = -1;
-        return true;
-      }
-    } else {
-      // Look for the key position
-      while (next_files_to_cut_for_ttl_ <
-             static_cast<int>(files_to_cut_for_ttl_.size())) {
-        if (icmp->Compare(internal_key,
-                          files_to_cut_for_ttl_[next_files_to_cut_for_ttl_]
-                              ->smallest.Encode()) >= 0) {
-          if (icmp->Compare(internal_key,
-                            files_to_cut_for_ttl_[next_files_to_cut_for_ttl_]
-                                ->largest.Encode()) <= 0) {
-            // With in the current file
-            cur_files_to_cut_for_ttl_ = next_files_to_cut_for_ttl_;
-            return true;
-          }
-          // Beyond the current file
-          next_files_to_cut_for_ttl_++;
-        } else {
-          // Still fall into the gap
-          break;
-        }
-      }
-    }
-  }
-
   return false;
 }
 
@@ -333,8 +356,14 @@ Status CompactionOutputs::AddToOutput(
     const CompactionFileOpenFunc& open_file_func,
     const CompactionFileCloseFunc& close_file_func) {
   Status s;
+  bool is_range_del = c_iter.IsDeleteRangeSentinelKey();
+  if (is_range_del && compaction_->bottommost_level()) {
+    // We don't consider range tombstone for bottommost level since:
+    // 1. there is no grandparent and hence no overlap to consider
+    // 2. range tombstone may be dropped at bottommost level.
+    return s;
+  }
   const Slice& key = c_iter.key();
-
   if (ShouldStopBefore(c_iter) && HasBuilder()) {
     s = close_file_func(*this, c_iter.InputStatus(), key);
     if (!s.ok()) {
@@ -344,6 +373,13 @@ Status CompactionOutputs::AddToOutput(
     grandparent_boundary_switched_num_ = 0;
     grandparent_overlapped_bytes_ =
         GetCurrentKeyGrandparentOverlappedBytes(key);
+    if (UNLIKELY(is_range_del)) {
+      // lower bound for this new output file, this is needed as the lower bound
+      // does not come from the smallest point key in this case.
+      range_tombstone_lower_bound_.DecodeFrom(key);
+    } else {
+      range_tombstone_lower_bound_.Clear();
+    }
   }
 
   // Open output file if necessary
@@ -354,6 +390,17 @@ Status CompactionOutputs::AddToOutput(
     }
   }
 
+  // c_iter may emit range deletion keys, so update `last_key_for_partitioner_`
+  // here before returning below when `is_range_del` is true
+  if (partitioner_) {
+    last_key_for_partitioner_.assign(c_iter.user_key().data_,
+                                     c_iter.user_key().size_);
+  }
+
+  if (UNLIKELY(is_range_del)) {
+    return s;
+  }
+
   assert(builder_ != nullptr);
   const Slice& value = c_iter.value();
   s = current_output().validator.Add(key, value);
@@ -377,106 +424,173 @@ Status CompactionOutputs::AddToOutput(
   s = current_output().meta.UpdateBoundaries(key, value, ikey.sequence,
                                              ikey.type);
 
-  if (partitioner_) {
-    last_key_for_partitioner_.assign(c_iter.user_key().data_,
-                                     c_iter.user_key().size_);
-  }
-
   return s;
 }
 
+namespace {
+void SetMaxSeqAndTs(InternalKey& internal_key, const Slice& user_key,
+                    const size_t ts_sz) {
+  if (ts_sz) {
+    static constexpr char kTsMax[] = "\xff\xff\xff\xff\xff\xff\xff\xff\xff";
+    if (ts_sz <= strlen(kTsMax)) {
+      internal_key = InternalKey(user_key, kMaxSequenceNumber,
+                                 kTypeRangeDeletion, Slice(kTsMax, ts_sz));
+    } else {
+      internal_key =
+          InternalKey(user_key, kMaxSequenceNumber, kTypeRangeDeletion,
+                      std::string(ts_sz, '\xff'));
+    }
+  } else {
+    internal_key.Set(user_key, kMaxSequenceNumber, kTypeRangeDeletion);
+  }
+}
+}  // namespace
+
 Status CompactionOutputs::AddRangeDels(
     const Slice* comp_start_user_key, const Slice* comp_end_user_key,
     CompactionIterationStats& range_del_out_stats, bool bottommost_level,
     const InternalKeyComparator& icmp, SequenceNumber earliest_snapshot,
     const Slice& next_table_min_key, const std::string& full_history_ts_low) {
+  // The following example does not happen since
+  // CompactionOutput::ShouldStopBefore() always return false for the first
+  // point key. But we should consider removing this dependency. Suppose for the
+  // first compaction output file,
+  //  - next_table_min_key.user_key == comp_start_user_key
+  //  - no point key is in the output file
+  //  - there is a range tombstone @seqno to be added that covers
+  //  comp_start_user_key
+  // Then meta.smallest will be set to comp_start_user_key@seqno
+  // and meta.largest will be set to comp_start_user_key@kMaxSequenceNumber
+  // which violates the assumption that meta.smallest should be <= meta.largest.
   assert(HasRangeDel());
   FileMetaData& meta = current_output().meta;
   const Comparator* ucmp = icmp.user_comparator();
-
+  InternalKey lower_bound_buf, upper_bound_buf;
   Slice lower_bound_guard, upper_bound_guard;
   std::string smallest_user_key;
   const Slice *lower_bound, *upper_bound;
-  bool lower_bound_from_sub_compact = false;
 
+  // We first determine the internal key lower_bound and upper_bound for
+  // this output file. All and only range tombstones that overlap with
+  // [lower_bound, upper_bound] should be added to this file. File
+  // boundaries (meta.smallest/largest) should be updated accordingly when
+  // extended by range tombstones.
   size_t output_size = outputs_.size();
   if (output_size == 1) {
-    // For the first output table, include range tombstones before the min
-    // key but after the subcompaction boundary.
-    lower_bound = comp_start_user_key;
-    lower_bound_from_sub_compact = true;
-  } else if (meta.smallest.size() > 0) {
+    // This is the first file in the subcompaction.
+    //
+    // When outputting a range tombstone that spans a subcompaction boundary,
+    // the files on either side of that boundary need to include that
+    // boundary's user key. Otherwise, the spanning range tombstone would lose
+    // coverage.
+    //
+    // To achieve this while preventing files from overlapping in internal key
+    // (an LSM invariant violation), we allow the earlier file to include the
+    // boundary user key up to `kMaxSequenceNumber,kTypeRangeDeletion`. The
+    // later file can begin at the boundary user key at the newest key version
+    // it contains. At this point that version number is unknown since we have
+    // not processed the range tombstones yet, so permit any version. Same story
+    // applies to timestamp, and a non-nullptr `comp_start_user_key` should have
+    // `kMaxTs` here, which similarly permits any timestamp.
+    if (comp_start_user_key) {
+      lower_bound_buf.Set(*comp_start_user_key, kMaxSequenceNumber,
+                          kTypeRangeDeletion);
+      lower_bound_guard = lower_bound_buf.Encode();
+      lower_bound = &lower_bound_guard;
+    } else {
+      lower_bound = nullptr;
+    }
+  } else {
     // For subsequent output tables, only include range tombstones from min
     // key onwards since the previous file was extended to contain range
     // tombstones falling before min key.
-    smallest_user_key = meta.smallest.user_key().ToString(false /*hex*/);
-    lower_bound_guard = Slice(smallest_user_key);
-    lower_bound = &lower_bound_guard;
-  } else {
-    lower_bound = nullptr;
-  }
-  if (!next_table_min_key.empty()) {
-    // This may be the last file in the subcompaction in some cases, so we
-    // need to compare the end key of subcompaction with the next file start
-    // key. When the end key is chosen by the subcompaction, we know that
-    // it must be the biggest key in output file. Therefore, it is safe to
-    // use the smaller key as the upper bound of the output file, to ensure
-    // that there is no overlapping between different output files.
-    upper_bound_guard = ExtractUserKey(next_table_min_key);
-    if (comp_end_user_key != nullptr &&
-        ucmp->CompareWithoutTimestamp(upper_bound_guard, *comp_end_user_key) >=
-            0) {
-      upper_bound = comp_end_user_key;
+    if (range_tombstone_lower_bound_.size() > 0) {
+      assert(meta.smallest.size() == 0 ||
+             icmp.Compare(range_tombstone_lower_bound_, meta.smallest) < 0);
+      lower_bound_guard = range_tombstone_lower_bound_.Encode();
     } else {
-      upper_bound = &upper_bound_guard;
+      assert(meta.smallest.size() > 0);
+      lower_bound_guard = meta.smallest.Encode();
     }
-  } else {
-    // This is the last file in the subcompaction, so extend until the
-    // subcompaction ends.
-    upper_bound = comp_end_user_key;
-  }
-  bool has_overlapping_endpoints;
-  if (upper_bound != nullptr && meta.largest.size() > 0) {
-    has_overlapping_endpoints = ucmp->CompareWithoutTimestamp(
-                                    meta.largest.user_key(), *upper_bound) == 0;
-  } else {
-    has_overlapping_endpoints = false;
+    lower_bound = &lower_bound_guard;
   }
 
+  const size_t ts_sz = ucmp->timestamp_size();
+  if (next_table_min_key.empty()) {
+    // Last file of the subcompaction.
+    if (comp_end_user_key) {
+      upper_bound_buf.Set(*comp_end_user_key, kMaxSequenceNumber,
+                          kTypeRangeDeletion);
+      upper_bound_guard = upper_bound_buf.Encode();
+      upper_bound = &upper_bound_guard;
+    } else {
+      upper_bound = nullptr;
+    }
+  } else {
+    // There is another file coming whose coverage will begin at
+    // `next_table_min_key`. The current file needs to extend range tombstone
+    // coverage through its own keys (through `meta.largest`) and through user
+    // keys preceding `next_table_min_key`'s user key.
+    ParsedInternalKey next_table_min_key_parsed;
+    ParseInternalKey(next_table_min_key, &next_table_min_key_parsed,
+                     false /* log_err_key */)
+        .PermitUncheckedError();
+    assert(next_table_min_key_parsed.sequence < kMaxSequenceNumber);
+    assert(meta.largest.size() == 0 ||
+           icmp.Compare(meta.largest.Encode(), next_table_min_key) < 0);
+    assert(!lower_bound || icmp.Compare(*lower_bound, next_table_min_key) <= 0);
+    if (meta.largest.size() > 0 &&
+        ucmp->EqualWithoutTimestamp(meta.largest.user_key(),
+                                    next_table_min_key_parsed.user_key)) {
+      // Caution: this assumes meta.largest.Encode() lives longer than
+      // upper_bound, which is only true if meta.largest is never updated.
+      // This just happens to be the case here since meta.largest serves
+      // as the upper_bound.
+      upper_bound_guard = meta.largest.Encode();
+    } else {
+      SetMaxSeqAndTs(upper_bound_buf, next_table_min_key_parsed.user_key,
+                     ts_sz);
+      upper_bound_guard = upper_bound_buf.Encode();
+    }
+    upper_bound = &upper_bound_guard;
+  }
+  if (lower_bound && upper_bound &&
+      icmp.Compare(*lower_bound, *upper_bound) > 0) {
+    assert(meta.smallest.size() == 0 &&
+           ucmp->EqualWithoutTimestamp(ExtractUserKey(*lower_bound),
+                                       ExtractUserKey(*upper_bound)));
+    // This can only happen when lower_bound have the same user key as
+    // next_table_min_key and that there is no point key in the current
+    // compaction output file.
+    return Status::OK();
+  }
   // The end key of the subcompaction must be bigger or equal to the upper
   // bound. If the end of subcompaction is null or the upper bound is null,
   // it means that this file is the last file in the compaction. So there
   // will be no overlapping between this file and others.
   assert(comp_end_user_key == nullptr || upper_bound == nullptr ||
-         ucmp->CompareWithoutTimestamp(*upper_bound, *comp_end_user_key) <= 0);
-  auto it = range_del_agg_->NewIterator(lower_bound, upper_bound,
-                                        has_overlapping_endpoints);
-  // Position the range tombstone output iterator. There may be tombstone
-  // fragments that are entirely out of range, so make sure that we do not
-  // include those.
-  if (lower_bound != nullptr) {
-    it->Seek(*lower_bound);
-  } else {
-    it->SeekToFirst();
-  }
-  for (; it->Valid(); it->Next()) {
+         ucmp->CompareWithoutTimestamp(ExtractUserKey(*upper_bound),
+                                       *comp_end_user_key) <= 0);
+  auto it = range_del_agg_->NewIterator(lower_bound, upper_bound);
+  Slice last_tombstone_start_user_key{};
+  bool reached_lower_bound = false;
+  const ReadOptions read_options(Env::IOActivity::kCompaction);
+  for (it->SeekToFirst(); it->Valid(); it->Next()) {
     auto tombstone = it->Tombstone();
-    if (upper_bound != nullptr) {
-      int cmp =
-          ucmp->CompareWithoutTimestamp(*upper_bound, tombstone.start_key_);
-      if ((has_overlapping_endpoints && cmp < 0) ||
-          (!has_overlapping_endpoints && cmp <= 0)) {
-        // Tombstones starting after upper_bound only need to be included in
-        // the next table. If the current SST ends before upper_bound, i.e.,
-        // `has_overlapping_endpoints == false`, we can also skip over range
-        // tombstones that start exactly at upper_bound. Such range
-        // tombstones will be included in the next file and are not relevant
-        // to the point keys or endpoints of the current file.
-        break;
-      }
+    auto kv = tombstone.Serialize();
+    InternalKey tombstone_end = tombstone.SerializeEndKey();
+    // TODO: the underlying iterator should support clamping the bounds.
+    // tombstone_end.Encode is of form user_key@kMaxSeqno
+    // if it is equal to lower_bound, there is no need to include
+    // such range tombstone.
+    if (!reached_lower_bound && lower_bound &&
+        icmp.Compare(tombstone_end.Encode(), *lower_bound) <= 0) {
+      continue;
     }
+    assert(!lower_bound ||
+           icmp.Compare(*lower_bound, tombstone_end.Encode()) <= 0);
+    reached_lower_bound = true;
 
-    const size_t ts_sz = ucmp->timestamp_size();
     // Garbage collection for range tombstones.
     // If user-defined timestamp is enabled, range tombstones are dropped if
     // they are at bottommost_level, below full_history_ts_low and not visible
@@ -495,96 +609,119 @@ Status CompactionOutputs::AddRangeDels(
       continue;
     }
 
-    auto kv = tombstone.Serialize();
     assert(lower_bound == nullptr ||
-           ucmp->CompareWithoutTimestamp(*lower_bound, kv.second) < 0);
+           ucmp->CompareWithoutTimestamp(ExtractUserKey(*lower_bound),
+                                         kv.second) < 0);
+    InternalKey tombstone_start = kv.first;
+    if (lower_bound &&
+        ucmp->CompareWithoutTimestamp(tombstone_start.user_key(),
+                                      ExtractUserKey(*lower_bound)) < 0) {
+      // This just updates the non-timestamp portion of `tombstone_start`'s user
+      // key. Ideally there would be a simpler API usage
+      ParsedInternalKey tombstone_start_parsed;
+      ParseInternalKey(tombstone_start.Encode(), &tombstone_start_parsed,
+                       false /* log_err_key */)
+          .PermitUncheckedError();
+      // timestamp should be from where sequence number is from, which is from
+      // tombstone in this case
+      std::string ts =
+          tombstone_start_parsed.GetTimestamp(ucmp->timestamp_size())
+              .ToString();
+      tombstone_start_parsed.user_key = ExtractUserKey(*lower_bound);
+      tombstone_start.SetFrom(tombstone_start_parsed, ts);
+    }
+    if (upper_bound != nullptr &&
+        icmp.Compare(*upper_bound, tombstone_start.Encode()) < 0) {
+      break;
+    }
+    // Here we show that *only* range tombstones that overlap with
+    // [lower_bound, upper_bound] are added to the current file, and
+    // sanity checking invariants that should hold:
+    // - [tombstone_start, tombstone_end] overlaps with [lower_bound,
+    // upper_bound]
+    // - meta.smallest <= meta.largest
+    // Corresponding assertions are made, the proof is broken is any of them
+    // fails.
+    // TODO: show that *all* range tombstones that overlap with
+    //  [lower_bound, upper_bound] are added.
+    // TODO: some invariant about boundaries are correctly updated.
+    //
+    // Note that `tombstone_start` is updated in the if condition above, we use
+    // tombstone_start to refer to its initial value, i.e.,
+    // it->Tombstone().first, and use tombstone_start* to refer to its value
+    // after the update.
+    //
+    // To show [lower_bound, upper_bound] overlaps with [tombstone_start,
+    // tombstone_end]:
+    // lower_bound <= upper_bound from the if condition right after all
+    // bounds are initialized. We assume each tombstone fragment has
+    // start_key.user_key < end_key.user_key, so
+    // tombstone_start < tombstone_end by
+    // FragmentedTombstoneIterator::Tombstone(). So these two ranges are both
+    // non-emtpy. The flag `reached_lower_bound` and the if logic before it
+    // ensures lower_bound <= tombstone_end. tombstone_start is only updated
+    // if it has a smaller user_key than lower_bound user_key, so
+    // tombstone_start <= tombstone_start*. The above if condition implies
+    // tombstone_start* <= upper_bound. So we have
+    // tombstone_start <= upper_bound and lower_bound <= tombstone_end
+    // and the two ranges overlap.
+    //
+    // To show meta.smallest <= meta.largest:
+    // From the implementation of UpdateBoundariesForRange(), it suffices to
+    // prove that when it is first called in this function, its parameters
+    // satisfy `start <= end`, where start = max(tombstone_start*, lower_bound)
+    // and end = min(tombstone_end, upper_bound). From the above proof we have
+    // lower_bound <= tombstone_end and lower_bound <= upper_bound. We only need
+    // to show that tombstone_start* <= min(tombstone_end, upper_bound).
+    // Note that tombstone_start*.user_key = max(tombstone_start.user_key,
+    // lower_bound.user_key). Assuming tombstone_end always has
+    // kMaxSequenceNumber and lower_bound.seqno < kMaxSequenceNumber.
+    // Since lower_bound <= tombstone_end and lower_bound.seqno <
+    // tombstone_end.seqno (in absolute number order, not internal key order),
+    // lower_bound.user_key < tombstone_end.user_key.
+    // Since lower_bound.user_key < tombstone_end.user_key and
+    // tombstone_start.user_key < tombstone_end.user_key, tombstone_start* <
+    // tombstone_end. Since tombstone_start* <= upper_bound from the above proof
+    // and tombstone_start* < tombstone_end, tombstone_start* <=
+    // min(tombstone_end, upper_bound), so the two ranges overlap.
+
     // Range tombstone is not supported by output validator yet.
     builder_->Add(kv.first.Encode(), kv.second);
-    InternalKey smallest_candidate = std::move(kv.first);
-    if (lower_bound != nullptr &&
-        ucmp->CompareWithoutTimestamp(smallest_candidate.user_key(),
-                                      *lower_bound) <= 0) {
-      // Pretend the smallest key has the same user key as lower_bound
-      // (the max key in the previous table or subcompaction) in order for
-      // files to appear key-space partitioned.
-      //
-      // When lower_bound is chosen by a subcompaction, we know that
-      // subcompactions over smaller keys cannot contain any keys at
-      // lower_bound. We also know that smaller subcompactions exist,
-      // because otherwise the subcompaction woud be unbounded on the left.
-      // As a result, we know that no other files on the output level will
-      // contain actual keys at lower_bound (an output file may have a
-      // largest key of lower_bound@kMaxSequenceNumber, but this only
-      // indicates a large range tombstone was truncated). Therefore, it is
-      // safe to use the tombstone's sequence number, to ensure that keys at
-      // lower_bound at lower levels are covered by truncated tombstones.
-      //
-      // If lower_bound was chosen by the smallest data key in the file,
-      // choose lowest seqnum so this file's smallest internal key comes
-      // after the previous file's largest. The fake seqnum is OK because
-      // the read path's file-picking code only considers user key.
-      if (lower_bound_from_sub_compact) {
-        if (ts_sz) {
-          assert(tombstone.ts_.size() == ts_sz);
-          smallest_candidate = InternalKey(*lower_bound, tombstone.seq_,
-                                           kTypeRangeDeletion, tombstone.ts_);
-        } else {
-          smallest_candidate =
-              InternalKey(*lower_bound, tombstone.seq_, kTypeRangeDeletion);
-        }
-      } else {
-        smallest_candidate = InternalKey(*lower_bound, 0, kTypeRangeDeletion);
-      }
+    if (lower_bound &&
+        icmp.Compare(tombstone_start.Encode(), *lower_bound) < 0) {
+      tombstone_start.DecodeFrom(*lower_bound);
     }
-    InternalKey largest_candidate = tombstone.SerializeEndKey();
-    if (upper_bound != nullptr &&
-        ucmp->CompareWithoutTimestamp(*upper_bound,
-                                      largest_candidate.user_key()) <= 0) {
-      // Pretend the largest key has the same user key as upper_bound (the
-      // min key in the following table or subcompaction) in order for files
-      // to appear key-space partitioned.
-      //
-      // Choose highest seqnum so this file's largest internal key comes
-      // before the next file's/subcompaction's smallest. The fake seqnum is
-      // OK because the read path's file-picking code only considers the
-      // user key portion.
-      //
-      // Note Seek() also creates InternalKey with (user_key,
-      // kMaxSequenceNumber), but with kTypeDeletion (0x7) instead of
-      // kTypeRangeDeletion (0xF), so the range tombstone comes before the
-      // Seek() key in InternalKey's ordering. So Seek() will look in the
-      // next file for the user key
-      if (ts_sz) {
-        static constexpr char kTsMax[] = "\xff\xff\xff\xff\xff\xff\xff\xff\xff";
-        if (ts_sz <= strlen(kTsMax)) {
-          largest_candidate =
-              InternalKey(*upper_bound, kMaxSequenceNumber, kTypeRangeDeletion,
-                          Slice(kTsMax, ts_sz));
-        } else {
-          largest_candidate =
-              InternalKey(*upper_bound, kMaxSequenceNumber, kTypeRangeDeletion,
-                          std::string(ts_sz, '\xff'));
-        }
-      } else {
-        largest_candidate =
-            InternalKey(*upper_bound, kMaxSequenceNumber, kTypeRangeDeletion);
-      }
+    if (upper_bound && icmp.Compare(*upper_bound, tombstone_end.Encode()) < 0) {
+      tombstone_end.DecodeFrom(*upper_bound);
     }
-#ifndef NDEBUG
-    SequenceNumber smallest_ikey_seqnum = kMaxSequenceNumber;
-    if (meta.smallest.size() > 0) {
-      smallest_ikey_seqnum = GetInternalKeySeqno(meta.smallest.Encode());
-    }
-#endif
-    meta.UpdateBoundariesForRange(smallest_candidate, largest_candidate,
+    assert(icmp.Compare(tombstone_start, tombstone_end) <= 0);
+    meta.UpdateBoundariesForRange(tombstone_start, tombstone_end,
                                   tombstone.seq_, icmp);
-    // The smallest key in a file is used for range tombstone truncation, so
-    // it cannot have a seqnum of 0 (unless the smallest data key in a file
-    // has a seqnum of 0). Otherwise, the truncated tombstone may expose
-    // deleted keys at lower levels.
-    assert(smallest_ikey_seqnum == 0 ||
-           ExtractInternalKeyFooter(meta.smallest.Encode()) !=
-               PackSequenceAndType(0, kTypeRangeDeletion));
+    if (!bottommost_level) {
+      bool start_user_key_changed =
+          last_tombstone_start_user_key.empty() ||
+          ucmp->CompareWithoutTimestamp(last_tombstone_start_user_key,
+                                        it->start_key()) < 0;
+      last_tombstone_start_user_key = it->start_key();
+      if (start_user_key_changed) {
+        // If tombstone_start >= tombstone_end, then either no key range is
+        // covered, or that they have the same user key. If they have the same
+        // user key, then the internal key range should only be within this
+        // level, and no keys from older levels is covered.
+        if (ucmp->CompareWithoutTimestamp(tombstone_start.user_key(),
+                                          tombstone_end.user_key()) < 0) {
+          SizeApproximationOptions approx_opts;
+          approx_opts.files_size_error_margin = 0.1;
+          auto approximate_covered_size =
+              compaction_->input_version()->version_set()->ApproximateSize(
+                  approx_opts, read_options, compaction_->input_version(),
+                  tombstone_start.Encode(), tombstone_end.Encode(),
+                  compaction_->output_level() + 1 /* start_level */,
+                  -1 /* end_level */, kCompaction);
+          meta.compensated_range_deletion_size += approximate_covered_size;
+        }
+      }
+    }
   }
   return Status::OK();
 }

db/compaction/compaction_outputs.h

@@ -167,9 +167,15 @@ class CompactionOutputs {
     current_output_file_size_ = 0;
   }
 
-  // Add range-dels from the aggregator to the current output file
+  // Add range deletions from the range_del_agg_ to the current output file.
+  // Input parameters, `range_tombstone_lower_bound_` and current output's
+  // metadata determine the bounds on range deletions to add. Updates output
+  // file metadata boundary if extended by range tombstones.
+  //
   // @param comp_start_user_key and comp_end_user_key include timestamp if
-  // user-defined timestamp is enabled.
+  // user-defined timestamp is enabled. Their timestamp should be max timestamp.
+  // @param next_table_min_key internal key lower bound for the next compaction
+  // output.
   // @param full_history_ts_low used for range tombstone garbage collection.
   Status AddRangeDels(const Slice* comp_start_user_key,
                       const Slice* comp_end_user_key,
@@ -221,6 +227,13 @@ class CompactionOutputs {
     }
   }
 
+  // Updates states related to file cutting for TTL.
+  // Returns a boolean value indicating whether the current
+  // compaction output file should be cut before `internal_key`.
+  //
+  // @param internal_key the current key to be added to output.
+  bool UpdateFilesToCutForTTLStates(const Slice& internal_key);
+
   // update tracked grandparents information like grandparent index, if it's
   // in the gap between 2 grandparent files, accumulated grandparent files size
   // etc.
@@ -307,6 +320,7 @@ class CompactionOutputs {
   std::unique_ptr<SstPartitioner> partitioner_;
 
   // A flag determines if this subcompaction has been split by the cursor
+  // for RoundRobin compaction
   bool is_split_ = false;
 
   // We also maintain the output split key for each subcompaction to avoid
@@ -338,6 +352,10 @@ class CompactionOutputs {
   // for the current output file, how many file boundaries has it crossed,
   // basically number of files overlapped * 2
   size_t grandparent_boundary_switched_num_ = 0;
+
+  // The smallest key of the current output file, this is set when current
+  // output file's smallest key is a range tombstone start key.
+  InternalKey range_tombstone_lower_bound_;
 };
 
 // helper struct to concatenate the last level and penultimate level outputs

db/compaction/compaction_picker.cc

@@ -31,27 +31,15 @@ bool FindIntraL0Compaction(const std::vector<FileMetaData*>& level_files,
                            size_t min_files_to_compact,
                            uint64_t max_compact_bytes_per_del_file,
                            uint64_t max_compaction_bytes,
-                           CompactionInputFiles* comp_inputs,
-                           SequenceNumber earliest_mem_seqno) {
-  // Do not pick ingested file when there is at least one memtable not flushed
-  // which of seqno is overlap with the sst.
+                           CompactionInputFiles* comp_inputs) {
   TEST_SYNC_POINT("FindIntraL0Compaction");
+
   size_t start = 0;
-  for (; start < level_files.size(); start++) {
-    if (level_files[start]->being_compacted) {
-      return false;
-    }
-    // If there is no data in memtable, the earliest sequence number would the
-    // largest sequence number in last memtable.
-    // Because all files are sorted in descending order by largest_seqno, so we
-    // only need to check the first one.
-    if (level_files[start]->fd.largest_seqno <= earliest_mem_seqno) {
-      break;
-    }
-  }
-  if (start >= level_files.size()) {
+
+  if (level_files.size() == 0 || level_files[start]->being_compacted) {
     return false;
   }
+
   size_t compact_bytes = static_cast<size_t>(level_files[start]->fd.file_size);
   size_t compact_bytes_per_del_file = std::numeric_limits<size_t>::max();
   // Compaction range will be [start, limit).
@@ -889,7 +877,6 @@ Compaction* CompactionPicker::CompactRange(
   return compaction;
 }
 
-#ifndef ROCKSDB_LITE
 namespace {
 // Test whether two files have overlapping key-ranges.
 bool HaveOverlappingKeyRanges(const Comparator* c, const SstFileMetaData& a,
@@ -995,6 +982,7 @@ Status CompactionPicker::SanitizeCompactionInputFilesForAllLevels(
                                current_files[f].name +
                                " is currently being compacted.");
       }
+
       input_files->insert(TableFileNameToNumber(current_files[f].name));
     }
 
@@ -1127,7 +1115,6 @@ Status CompactionPicker::SanitizeCompactionInputFiles(
 
   return Status::OK();
 }
-#endif  // !ROCKSDB_LITE
 
 void CompactionPicker::RegisterCompaction(Compaction* c) {
   if (c == nullptr) {
@@ -1137,7 +1124,11 @@ void CompactionPicker::RegisterCompaction(Compaction* c) {
          c->output_level() == 0 ||
          !FilesRangeOverlapWithCompaction(*c->inputs(), c->output_level(),
                                           c->GetPenultimateLevel()));
-  if (c->start_level() == 0 ||
+  // CompactionReason::kExternalSstIngestion's start level is just a placeholder
+  // number without actual meaning as file ingestion technically does not have
+  // an input level like other compactions
+  if ((c->start_level() == 0 &&
+       c->compaction_reason() != CompactionReason::kExternalSstIngestion) ||
       ioptions_.compaction_style == kCompactionStyleUniversal) {
     level0_compactions_in_progress_.insert(c);
   }

db/compaction/compaction_picker.h

@@ -51,14 +51,15 @@ class CompactionPicker {
   virtual ~CompactionPicker();
 
   // Pick level and inputs for a new compaction.
+  //
   // Returns nullptr if there is no compaction to be done.
   // Otherwise returns a pointer to a heap-allocated object that
   // describes the compaction.  Caller should delete the result.
-  virtual Compaction* PickCompaction(
-      const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
-      const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,
-      LogBuffer* log_buffer,
-      SequenceNumber earliest_memtable_seqno = kMaxSequenceNumber) = 0;
+  virtual Compaction* PickCompaction(const std::string& cf_name,
+                                     const MutableCFOptions& mutable_cf_options,
+                                     const MutableDBOptions& mutable_db_options,
+                                     VersionStorageInfo* vstorage,
+                                     LogBuffer* log_buffer) = 0;
 
   // Return a compaction object for compacting the range [begin,end] in
   // the specified level.  Returns nullptr if there is nothing in that
@@ -91,11 +92,10 @@ class CompactionPicker {
 // files.  If it's not possible to conver an invalid input_files
 // into a valid one by adding more files, the function will return a
 // non-ok status with specific reason.
-#ifndef ROCKSDB_LITE
+//
   Status SanitizeCompactionInputFiles(std::unordered_set<uint64_t>* input_files,
                                       const ColumnFamilyMetaData& cf_meta,
                                       const int output_level) const;
-#endif  // ROCKSDB_LITE
 
   // Free up the files that participated in a compaction
   //
@@ -227,11 +227,9 @@ class CompactionPicker {
 
 // A helper function to SanitizeCompactionInputFiles() that
 // sanitizes "input_files" by adding necessary files.
-#ifndef ROCKSDB_LITE
   virtual Status SanitizeCompactionInputFilesForAllLevels(
       std::unordered_set<uint64_t>* input_files,
       const ColumnFamilyMetaData& cf_meta, const int output_level) const;
-#endif  // ROCKSDB_LITE
 
   // Keeps track of all compactions that are running on Level0.
   // Protected by DB mutex
@@ -244,7 +242,6 @@ class CompactionPicker {
   const InternalKeyComparator* const icmp_;
 };
 
-#ifndef ROCKSDB_LITE
 // A dummy compaction that never triggers any automatic
 // compaction.
 class NullCompactionPicker : public CompactionPicker {
@@ -255,12 +252,11 @@ class NullCompactionPicker : public CompactionPicker {
   virtual ~NullCompactionPicker() {}
 
   // Always return "nullptr"
-  Compaction* PickCompaction(
-      const std::string& /*cf_name*/,
-      const MutableCFOptions& /*mutable_cf_options*/,
-      const MutableDBOptions& /*mutable_db_options*/,
-      VersionStorageInfo* /*vstorage*/, LogBuffer* /* log_buffer */,
-      SequenceNumber /* earliest_memtable_seqno */) override {
+  Compaction* PickCompaction(const std::string& /*cf_name*/,
+                             const MutableCFOptions& /*mutable_cf_options*/,
+                             const MutableDBOptions& /*mutable_db_options*/,
+                             VersionStorageInfo* /*vstorage*/,
+                             LogBuffer* /* log_buffer */) override {
     return nullptr;
   }
 
@@ -286,7 +282,6 @@ class NullCompactionPicker : public CompactionPicker {
     return false;
   }
 };
-#endif  // !ROCKSDB_LITE
 
 // Attempts to find an intra L0 compaction conforming to the given parameters.
 //
@@ -304,11 +299,11 @@ class NullCompactionPicker : public CompactionPicker {
 //                                        files. Cannot be nullptr.
 //
 // @return                                true iff compaction was found.
-bool FindIntraL0Compaction(
-    const std::vector<FileMetaData*>& level_files, size_t min_files_to_compact,
-    uint64_t max_compact_bytes_per_del_file, uint64_t max_compaction_bytes,
-    CompactionInputFiles* comp_inputs,
-    SequenceNumber earliest_mem_seqno = kMaxSequenceNumber);
+bool FindIntraL0Compaction(const std::vector<FileMetaData*>& level_files,
+                           size_t min_files_to_compact,
+                           uint64_t max_compact_bytes_per_del_file,
+                           uint64_t max_compaction_bytes,
+                           CompactionInputFiles* comp_inputs);
 
 CompressionType GetCompressionType(const VersionStorageInfo* vstorage,
                                    const MutableCFOptions& mutable_cf_options,

db/compaction/compaction_picker_fifo.cc

@@ -8,7 +8,6 @@
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 
 #include "db/compaction/compaction_picker_fifo.h"
-#ifndef ROCKSDB_LITE
 
 #include <cinttypes>
 #include <string>
@@ -232,7 +231,8 @@ Compaction* FIFOCompactionPicker::PickSizeCompaction(
         break;
       }
     }
-  } else {
+  } else if (total_size >
+             mutable_cf_options.compaction_options_fifo.max_table_files_size) {
     // If the last level is non-L0, we actually don't know which file is
     // logically the oldest since the file creation time only represents
     // when this file was compacted to this level, which is independent
@@ -247,15 +247,27 @@ Compaction* FIFOCompactionPicker::PickSizeCompaction(
       inputs[0].files.push_back(f);
       char tmp_fsize[16];
       AppendHumanBytes(f->fd.GetFileSize(), tmp_fsize, sizeof(tmp_fsize));
-      ROCKS_LOG_BUFFER(log_buffer,
-                       "[%s] FIFO compaction: picking file %" PRIu64
-                       " with size %s for deletion",
-                       cf_name.c_str(), f->fd.GetNumber(), tmp_fsize);
+      ROCKS_LOG_BUFFER(
+          log_buffer,
+          "[%s] FIFO compaction: picking file %" PRIu64
+          " with size %s for deletion under total size %" PRIu64
+          " vs max table files size %" PRIu64,
+          cf_name.c_str(), f->fd.GetNumber(), tmp_fsize, total_size,
+          mutable_cf_options.compaction_options_fifo.max_table_files_size);
+
       if (total_size <=
           mutable_cf_options.compaction_options_fifo.max_table_files_size) {
         break;
       }
     }
+  } else {
+    ROCKS_LOG_BUFFER(
+        log_buffer,
+        "[%s] FIFO compaction: nothing to do. Total size %" PRIu64
+        ", max size %" PRIu64 "\n",
+        cf_name.c_str(), total_size,
+        mutable_cf_options.compaction_options_fifo.max_table_files_size);
+    return nullptr;
   }
 
   Compaction* c = new Compaction(
@@ -389,7 +401,7 @@ Compaction* FIFOCompactionPicker::PickCompactionToWarm(
 Compaction* FIFOCompactionPicker::PickCompaction(
     const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
     const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,
-    LogBuffer* log_buffer, SequenceNumber /*earliest_memtable_seqno*/) {
+    LogBuffer* log_buffer) {
   Compaction* c = nullptr;
   if (mutable_cf_options.ttl > 0) {
     c = PickTTLCompaction(cf_name, mutable_cf_options, mutable_db_options,
@@ -430,4 +442,3 @@ Compaction* FIFOCompactionPicker::CompactRange(
 }
 
 }  // namespace ROCKSDB_NAMESPACE
-#endif  // !ROCKSDB_LITE

db/compaction/compaction_picker_fifo.h

@@ -8,7 +8,6 @@
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 
 #pragma once
-#ifndef ROCKSDB_LITE
 
 #include "db/compaction/compaction_picker.h"
 
@@ -19,11 +18,11 @@ class FIFOCompactionPicker : public CompactionPicker {
                        const InternalKeyComparator* icmp)
       : CompactionPicker(ioptions, icmp) {}
 
-  virtual Compaction* PickCompaction(
-      const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
-      const MutableDBOptions& mutable_db_options, VersionStorageInfo* version,
-      LogBuffer* log_buffer,
-      SequenceNumber earliest_memtable_seqno = kMaxSequenceNumber) override;
+  virtual Compaction* PickCompaction(const std::string& cf_name,
+                                     const MutableCFOptions& mutable_cf_options,
+                                     const MutableDBOptions& mutable_db_options,
+                                     VersionStorageInfo* version,
+                                     LogBuffer* log_buffer) override;
 
   virtual Compaction* CompactRange(
       const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
@@ -60,4 +59,3 @@ class FIFOCompactionPicker : public CompactionPicker {
                                    LogBuffer* log_buffer);
 };
 }  // namespace ROCKSDB_NAMESPACE
-#endif  // !ROCKSDB_LITE

db/compaction/compaction_picker_level.cc

@@ -50,7 +50,6 @@ class LevelCompactionBuilder {
  public:
   LevelCompactionBuilder(const std::string& cf_name,
                          VersionStorageInfo* vstorage,
-                         SequenceNumber earliest_mem_seqno,
                          CompactionPicker* compaction_picker,
                          LogBuffer* log_buffer,
                          const MutableCFOptions& mutable_cf_options,
@@ -58,7 +57,6 @@ class LevelCompactionBuilder {
                          const MutableDBOptions& mutable_db_options)
       : cf_name_(cf_name),
         vstorage_(vstorage),
-        earliest_mem_seqno_(earliest_mem_seqno),
         compaction_picker_(compaction_picker),
         log_buffer_(log_buffer),
         mutable_cf_options_(mutable_cf_options),
@@ -85,7 +83,7 @@ class LevelCompactionBuilder {
 
   Compaction* GetCompaction();
 
-  // For the specfied level, pick a file that we want to compact.
+  // From `start_level_`, pick files to compact to `output_level_`.
   // Returns false if there is no file to compact.
   // If it returns true, inputs->files.size() will be exactly one for
   // all compaction priorities except round-robin. For round-robin,
@@ -109,8 +107,9 @@ class LevelCompactionBuilder {
   bool PickIntraL0Compaction();
 
   // Return true if TrivialMove is extended. `start_index` is the index of
-  // the intiial file picked, which should already be in `start_level_inputs_`.
-  bool TryExtendNonL0TrivialMove(int start_index);
+  // the initial file picked, which should already be in `start_level_inputs_`.
+  bool TryExtendNonL0TrivialMove(int start_index,
+                                 bool only_expand_right = false);
 
   // Picks a file from level_files to compact.
   // level_files is a vector of (level, file metadata) in ascending order of
@@ -122,7 +121,6 @@ class LevelCompactionBuilder {
 
   const std::string& cf_name_;
   VersionStorageInfo* vstorage_;
-  SequenceNumber earliest_mem_seqno_;
   CompactionPicker* compaction_picker_;
   LogBuffer* log_buffer_;
   int start_level_ = -1;
@@ -196,7 +194,10 @@ void LevelCompactionBuilder::SetupInitialFiles() {
       }
       output_level_ =
           (start_level_ == 0) ? vstorage_->base_level() : start_level_ + 1;
-      if (PickFileToCompact()) {
+      bool picked_file_to_compact = PickFileToCompact();
+      TEST_SYNC_POINT_CALLBACK("PostPickFileToCompact",
+                               &picked_file_to_compact);
+      if (picked_file_to_compact) {
         // found the compaction!
         if (start_level_ == 0) {
           // L0 score = `num L0 files` / `level0_file_num_compaction_trigger`
@@ -355,7 +356,8 @@ void LevelCompactionBuilder::SetupOtherFilesWithRoundRobinExpansion() {
   vstorage_->GetOverlappingInputs(output_level_, &smallest, &largest,
                                   &output_level_inputs.files);
   if (output_level_inputs.empty()) {
-    if (TryExtendNonL0TrivialMove((int)start_index)) {
+    if (TryExtendNonL0TrivialMove((int)start_index,
+                                  true /* only_expand_right */)) {
       return;
     }
   }
@@ -447,21 +449,21 @@ bool LevelCompactionBuilder::SetupOtherInputsIfNeeded() {
       compaction_inputs_.push_back(output_level_inputs_);
     }
 
+    // In some edge cases we could pick a compaction that will be compacting
+    // a key range that overlap with another running compaction, and both
+    // of them have the same output level. This could happen if
+    // (1) we are running a non-exclusive manual compaction
+    // (2) AddFile ingest a new file into the LSM tree
+    // We need to disallow this from happening.
+    if (compaction_picker_->FilesRangeOverlapWithCompaction(
+            compaction_inputs_, output_level_,
+            Compaction::EvaluatePenultimateLevel(
+                vstorage_, ioptions_, start_level_, output_level_))) {
+      // This compaction output could potentially conflict with the output
+      // of a currently running compaction, we cannot run it.
+      return false;
+    }
     if (!is_l0_trivial_move_) {
-      // In some edge cases we could pick a compaction that will be compacting
-      // a key range that overlap with another running compaction, and both
-      // of them have the same output level. This could happen if
-      // (1) we are running a non-exclusive manual compaction
-      // (2) AddFile ingest a new file into the LSM tree
-      // We need to disallow this from happening.
-      if (compaction_picker_->FilesRangeOverlapWithCompaction(
-              compaction_inputs_, output_level_,
-              Compaction::EvaluatePenultimateLevel(
-                  vstorage_, ioptions_, start_level_, output_level_))) {
-        // This compaction output could potentially conflict with the output
-        // of a currently running compaction, we cannot run it.
-        return false;
-      }
       compaction_picker_->GetGrandparents(vstorage_, start_level_inputs_,
                                           output_level_inputs_, &grandparents_);
     }
@@ -501,6 +503,16 @@ Compaction* LevelCompactionBuilder::PickCompaction() {
 }
 
 Compaction* LevelCompactionBuilder::GetCompaction() {
+  // TryPickL0TrivialMove() does not apply to the case when compacting L0 to an
+  // empty output level. So L0 files is picked in PickFileToCompact() by
+  // compaction score. We may still be able to do trivial move when this file
+  // does not overlap with other L0s. This happens when
+  // compaction_inputs_[0].size() == 1 since SetupOtherL0FilesIfNeeded() did not
+  // pull in more L0s.
+  assert(!compaction_inputs_.empty());
+  bool l0_files_might_overlap =
+      start_level_ == 0 && !is_l0_trivial_move_ &&
+      (compaction_inputs_.size() > 1 || compaction_inputs_[0].size() > 1);
   auto c = new Compaction(
       vstorage_, ioptions_, mutable_cf_options_, mutable_db_options_,
       std::move(compaction_inputs_), output_level_,
@@ -515,8 +527,7 @@ Compaction* LevelCompactionBuilder::GetCompaction() {
       Temperature::kUnknown,
       /* max_subcompactions */ 0, std::move(grandparents_), is_manual_,
       /* trim_ts */ "", start_level_score_, false /* deletion_compaction */,
-      /* l0_files_might_overlap */ start_level_ == 0 && !is_l0_trivial_move_,
-      compaction_reason_);
+      l0_files_might_overlap, compaction_reason_);
 
   // If it's level 0 compaction, make sure we don't execute any other level 0
   // compactions in parallel
@@ -653,7 +664,8 @@ bool LevelCompactionBuilder::TryPickL0TrivialMove() {
   return false;
 }
 
-bool LevelCompactionBuilder::TryExtendNonL0TrivialMove(int start_index) {
+bool LevelCompactionBuilder::TryExtendNonL0TrivialMove(int start_index,
+                                                       bool only_expand_right) {
   if (start_level_inputs_.size() == 1 &&
       (ioptions_.db_paths.empty() || ioptions_.db_paths.size() == 1) &&
       (mutable_cf_options_.compression_per_level.empty())) {
@@ -670,6 +682,7 @@ bool LevelCompactionBuilder::TryExtendNonL0TrivialMove(int start_index) {
     size_t total_size = initial_file->fd.GetFileSize();
     CompactionInputFiles output_level_inputs;
     output_level_inputs.level = output_level_;
+    // Expand towards right
     for (int i = start_index + 1;
          i < static_cast<int>(level_files.size()) &&
          start_level_inputs_.size() < kMaxMultiTrivialMove;
@@ -702,6 +715,37 @@ bool LevelCompactionBuilder::TryExtendNonL0TrivialMove(int start_index) {
       }
       start_level_inputs_.files.push_back(next_file);
     }
+    // Expand towards left
+    if (!only_expand_right) {
+      for (int i = start_index - 1;
+           i >= 0 && start_level_inputs_.size() < kMaxMultiTrivialMove; i--) {
+        FileMetaData* next_file = level_files[i];
+        if (next_file->being_compacted) {
+          break;
+        }
+        vstorage_->GetOverlappingInputs(output_level_, &(next_file->smallest),
+                                        &(initial_file->largest),
+                                        &output_level_inputs.files);
+        if (!output_level_inputs.empty()) {
+          break;
+        }
+        if (i > 0 && compaction_picker_->icmp()
+                             ->user_comparator()
+                             ->CompareWithoutTimestamp(
+                                 next_file->smallest.user_key(),
+                                 level_files[i - 1]->largest.user_key()) == 0) {
+          // Not a clean up after adding the next file. Skip.
+          break;
+        }
+        total_size += next_file->fd.GetFileSize();
+        if (total_size > mutable_cf_options_.max_compaction_bytes) {
+          break;
+        }
+        // keep `files` sorted in increasing order by key range
+        start_level_inputs_.files.insert(start_level_inputs_.files.begin(),
+                                         next_file);
+      }
+    }
     return start_level_inputs_.size() > 1;
   }
   return false;
@@ -785,7 +829,10 @@ bool LevelCompactionBuilder::PickFileToCompact() {
     vstorage_->GetOverlappingInputs(output_level_, &smallest, &largest,
                                     &output_level_inputs.files);
     if (output_level_inputs.empty()) {
-      if (TryExtendNonL0TrivialMove(index)) {
+      if (start_level_ > 0 &&
+          TryExtendNonL0TrivialMove(index,
+                                    ioptions_.compaction_pri ==
+                                        kRoundRobin /* only_expand_right */)) {
         break;
       }
     } else {
@@ -825,16 +872,16 @@ bool LevelCompactionBuilder::PickIntraL0Compaction() {
   return FindIntraL0Compaction(level_files, kMinFilesForIntraL0Compaction,
                                std::numeric_limits<uint64_t>::max(),
                                mutable_cf_options_.max_compaction_bytes,
-                               &start_level_inputs_, earliest_mem_seqno_);
+                               &start_level_inputs_);
 }
 }  // namespace
 
 Compaction* LevelCompactionPicker::PickCompaction(
     const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
     const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,
-    LogBuffer* log_buffer, SequenceNumber earliest_mem_seqno) {
-  LevelCompactionBuilder builder(cf_name, vstorage, earliest_mem_seqno, this,
-                                 log_buffer, mutable_cf_options, ioptions_,
+    LogBuffer* log_buffer) {
+  LevelCompactionBuilder builder(cf_name, vstorage, this, log_buffer,
+                                 mutable_cf_options, ioptions_,
                                  mutable_db_options);
   return builder.PickCompaction();
 }

db/compaction/compaction_picker_level.h

@@ -20,11 +20,11 @@ class LevelCompactionPicker : public CompactionPicker {
   LevelCompactionPicker(const ImmutableOptions& ioptions,
                         const InternalKeyComparator* icmp)
       : CompactionPicker(ioptions, icmp) {}
-  virtual Compaction* PickCompaction(
-      const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
-      const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,
-      LogBuffer* log_buffer,
-      SequenceNumber earliest_memtable_seqno = kMaxSequenceNumber) override;
+  virtual Compaction* PickCompaction(const std::string& cf_name,
+                                     const MutableCFOptions& mutable_cf_options,
+                                     const MutableDBOptions& mutable_db_options,
+                                     VersionStorageInfo* vstorage,
+                                     LogBuffer* log_buffer) override;
 
   virtual bool NeedsCompaction(
       const VersionStorageInfo* vstorage) const override;

db/compaction/compaction_picker_test.cc

@@ -77,8 +77,9 @@ class CompactionPickerTestBase : public testing::Test {
   void NewVersionStorage(int num_levels, CompactionStyle style) {
     DeleteVersionStorage();
     options_.num_levels = num_levels;
-    vstorage_.reset(new VersionStorageInfo(&icmp_, ucmp_, options_.num_levels,
-                                           style, nullptr, false));
+    vstorage_.reset(new VersionStorageInfo(
+        &icmp_, ucmp_, options_.num_levels, style, nullptr, false,
+        EpochNumberRequirement::kMustPresent));
     vstorage_->PrepareForVersionAppend(ioptions_, mutable_cf_options_);
   }
 
@@ -87,7 +88,7 @@ class CompactionPickerTestBase : public testing::Test {
   void AddVersionStorage() {
     temp_vstorage_.reset(new VersionStorageInfo(
         &icmp_, ucmp_, options_.num_levels, ioptions_.compaction_style,
-        vstorage_.get(), false));
+        vstorage_.get(), false, EpochNumberRequirement::kMustPresent));
   }
 
   void DeleteVersionStorage() {
@@ -105,7 +106,8 @@ class CompactionPickerTestBase : public testing::Test {
            size_t compensated_file_size = 0, bool marked_for_compact = false,
            Temperature temperature = Temperature::kUnknown,
            uint64_t oldest_ancestor_time = kUnknownOldestAncesterTime,
-           Slice ts_of_smallest = Slice(), Slice ts_of_largest = Slice()) {
+           Slice ts_of_smallest = Slice(), Slice ts_of_largest = Slice(),
+           uint64_t epoch_number = kUnknownEpochNumber) {
     assert(ts_of_smallest.size() == ucmp_->timestamp_size());
     assert(ts_of_largest.size() == ucmp_->timestamp_size());
 
@@ -145,8 +147,8 @@ class CompactionPickerTestBase : public testing::Test {
         file_number, path_id, file_size, smallest_ikey, largest_ikey,
         smallest_seq, largest_seq, marked_for_compact, temperature,
         kInvalidBlobFileNumber, kUnknownOldestAncesterTime,
-        kUnknownFileCreationTime, kUnknownFileChecksum,
-        kUnknownFileChecksumFuncName, kNullUniqueId64x2);
+        kUnknownFileCreationTime, epoch_number, kUnknownFileChecksum,
+        kUnknownFileChecksumFuncName, kNullUniqueId64x2, 0);
     f->compensated_file_size =
         (compensated_file_size != 0) ? compensated_file_size : file_size;
     f->oldest_ancester_time = oldest_ancestor_time;
@@ -480,8 +482,6 @@ TEST_F(CompactionPickerTest, LevelTriggerDynamic4) {
   ASSERT_EQ(num_levels - 1, compaction->output_level());
 }
 
-// Universal and FIFO Compactions are not supported in ROCKSDB_LITE
-#ifndef ROCKSDB_LITE
 TEST_F(CompactionPickerTest, NeedsCompactionUniversal) {
   NewVersionStorage(1, kCompactionStyleUniversal);
   UniversalCompactionPicker universal_compaction_picker(ioptions_, &icmp_);
@@ -1243,7 +1243,6 @@ TEST_F(CompactionPickerTest, FIFOToWarmWithHotBetweenWarms) {
   ASSERT_EQ(2U, compaction->input(0, 0)->fd.GetNumber());
 }
 
-#endif  // ROCKSDB_LITE
 
 TEST_F(CompactionPickerTest, CompactionPriMinOverlapping1) {
   NewVersionStorage(6, kCompactionStyleLevel);
@@ -2521,6 +2520,61 @@ TEST_F(CompactionPickerTest, L0TrivialMoveWholeL0) {
   ASSERT_TRUE(compaction->IsTrivialMove());
 }
 
+TEST_F(CompactionPickerTest, NonL0TrivialMoveExtendBothDirection) {
+  mutable_cf_options_.max_bytes_for_level_base = 5000;
+  mutable_cf_options_.level0_file_num_compaction_trigger = 4;
+  mutable_cf_options_.max_compaction_bytes = 10000000u;
+  ioptions_.level_compaction_dynamic_level_bytes = false;
+  NewVersionStorage(6, kCompactionStyleLevel);
+
+  Add(1, 1U, "300", "350", 3000U, 0, 710, 800, 3000U);
+  Add(1, 2U, "600", "651", 3001U, 0, 610, 700, 3001U);
+  Add(1, 3U, "700", "750", 3000U, 0, 500, 550, 3000U);
+  Add(2, 4U, "800", "850", 4000U, 0, 150, 200, 4000U);
+
+  UpdateVersionStorageInfo();
+  // File #2 should be picked first, and expand both directions to include
+  // files #1 and #3.
+  std::unique_ptr<Compaction> compaction(level_compaction_picker.PickCompaction(
+      cf_name_, mutable_cf_options_, mutable_db_options_, vstorage_.get(),
+      &log_buffer_));
+  ASSERT_TRUE(compaction.get() != nullptr);
+  ASSERT_EQ(1, compaction->num_input_levels());
+  ASSERT_EQ(3, compaction->num_input_files(0));
+  ASSERT_EQ(1, compaction->input(0, 0)->fd.GetNumber());
+  ASSERT_EQ(2, compaction->input(0, 1)->fd.GetNumber());
+  ASSERT_EQ(3, compaction->input(0, 2)->fd.GetNumber());
+  ASSERT_TRUE(compaction->IsTrivialMove());
+}
+
+TEST_F(CompactionPickerTest, L0TrivialMoveToEmptyLevel) {
+  mutable_cf_options_.max_bytes_for_level_base = 5000;
+  mutable_cf_options_.level0_file_num_compaction_trigger = 4;
+  mutable_cf_options_.max_compaction_bytes = 10000000u;
+  ioptions_.level_compaction_dynamic_level_bytes = false;
+  NewVersionStorage(6, kCompactionStyleLevel);
+
+  // File 2 will be picked first, which by itself is trivial movable.
+  // There was a bug before where compaction also picks file 3 and 4,
+  // (and then file 1 since it overlaps with the key range),
+  // which makes the compaction not trivial movable.
+  Add(0, 1U, "450", "599", 3000U, 0, 710, 800, 3000U);
+  Add(0, 2U, "600", "651", 3001U, 0, 610, 700, 3001U);
+  Add(0, 3U, "300", "350", 3000U, 0, 500, 550, 3000U);
+  Add(0, 4U, "500", "550", 2999U, 0, 300, 350, 2999U);
+
+  UpdateVersionStorageInfo();
+
+  std::unique_ptr<Compaction> compaction(level_compaction_picker.PickCompaction(
+      cf_name_, mutable_cf_options_, mutable_db_options_, vstorage_.get(),
+      &log_buffer_));
+  ASSERT_TRUE(compaction.get() != nullptr);
+  ASSERT_EQ(1, compaction->num_input_levels());
+  ASSERT_EQ(1, compaction->num_input_files(0));
+  ASSERT_EQ(2, compaction->input(0, 0)->fd.GetNumber());
+  ASSERT_TRUE(compaction->IsTrivialMove());
+}
+
 TEST_F(CompactionPickerTest, IsTrivialMoveOffSstPartitioned) {
   mutable_cf_options_.max_bytes_for_level_base = 10000u;
   mutable_cf_options_.max_compaction_bytes = 10001u;
@@ -2871,40 +2925,6 @@ TEST_F(CompactionPickerTest, IntraL0MaxCompactionBytesHit) {
   ASSERT_EQ(0, compaction->output_level());
 }
 
-TEST_F(CompactionPickerTest, IntraL0ForEarliestSeqno) {
-  // Intra L0 compaction triggers only if there are at least
-  // level0_file_num_compaction_trigger + 2 L0 files.
-  mutable_cf_options_.level0_file_num_compaction_trigger = 3;
-  mutable_cf_options_.max_compaction_bytes = 999999u;
-  NewVersionStorage(6, kCompactionStyleLevel);
-
-  // 4 out of 6 L0 files will be picked for intra L0 compaction due to
-  // being_compact limit. And the latest one L0 will be skipped due to earliest
-  // seqno. The one L1 file spans entire L0 key range and is marked as being
-  // compacted to avoid L0->L1 compaction.
-  Add(1, 1U, "100", "350", 200000U, 0, 110, 111);
-  Add(0, 2U, "301", "350", 1U, 0, 108, 109);
-  Add(0, 3U, "251", "300", 1U, 0, 106, 107);
-  Add(0, 4U, "201", "250", 1U, 0, 104, 105);
-  Add(0, 5U, "151", "200", 1U, 0, 102, 103);
-  Add(0, 6U, "100", "150", 1U, 0, 100, 101);
-  Add(0, 7U, "100", "100", 1U, 0, 99, 100);
-  vstorage_->LevelFiles(0)[5]->being_compacted = true;
-  vstorage_->LevelFiles(1)[0]->being_compacted = true;
-  UpdateVersionStorageInfo();
-
-  std::unique_ptr<Compaction> compaction(level_compaction_picker.PickCompaction(
-      cf_name_, mutable_cf_options_, mutable_db_options_, vstorage_.get(),
-      &log_buffer_, 107));
-  ASSERT_TRUE(compaction.get() != nullptr);
-  ASSERT_EQ(1U, compaction->num_input_levels());
-  ASSERT_EQ(4U, compaction->num_input_files(0));
-  ASSERT_EQ(CompactionReason::kLevelL0FilesNum,
-            compaction->compaction_reason());
-  ASSERT_EQ(0, compaction->output_level());
-}
-
-#ifndef ROCKSDB_LITE
 TEST_F(CompactionPickerTest, UniversalMarkedCompactionFullOverlap) {
   const uint64_t kFileSize = 100000;
 
@@ -2916,9 +2936,23 @@ TEST_F(CompactionPickerTest, UniversalMarkedCompactionFullOverlap) {
   // should fail
   NewVersionStorage(5, kCompactionStyleUniversal);
 
-  Add(0, 1U, "150", "200", kFileSize, 0, 500, 550);
-  Add(0, 2U, "201", "250", 2 * kFileSize, 0, 401, 450);
-  Add(0, 4U, "260", "300", 4 * kFileSize, 0, 260, 300);
+  Add(0, 1U, "150", "200", kFileSize, 0, 500, 550, /*compensated_file_size*/ 0,
+      /*marked_for_compact*/ false, /* temperature*/ Temperature::kUnknown,
+      /*oldest_ancestor_time*/ kUnknownOldestAncesterTime,
+      /*ts_of_smallest*/ Slice(), /*ts_of_largest*/ Slice(),
+      /*epoch_number*/ 3);
+  Add(0, 2U, "201", "250", 2 * kFileSize, 0, 401, 450,
+      /*compensated_file_size*/ 0, /*marked_for_compact*/ false,
+      /* temperature*/ Temperature::kUnknown,
+      /*oldest_ancestor_time*/ kUnknownOldestAncesterTime,
+      /*ts_of_smallest*/ Slice(), /*ts_of_largest*/ Slice(),
+      /*epoch_number*/ 2);
+  Add(0, 4U, "260", "300", 4 * kFileSize, 0, 260, 300,
+      /*compensated_file_size*/ 0, /*marked_for_compact*/ false,
+      /* temperature*/ Temperature::kUnknown,
+      /*oldest_ancestor_time*/ kUnknownOldestAncesterTime,
+      /*ts_of_smallest*/ Slice(), /*ts_of_largest*/ Slice(),
+      /*epoch_number*/ 1);
   Add(3, 5U, "010", "080", 8 * kFileSize, 0, 200, 251);
   Add(4, 3U, "301", "350", 8 * kFileSize, 0, 101, 150);
   Add(4, 6U, "501", "750", 8 * kFileSize, 0, 101, 150);
@@ -2940,7 +2974,11 @@ TEST_F(CompactionPickerTest, UniversalMarkedCompactionFullOverlap) {
 
   AddVersionStorage();
   // Simulate a flush and mark the file for compaction
-  Add(0, 7U, "150", "200", kFileSize, 0, 551, 600, 0, true);
+  Add(0, 7U, "150", "200", kFileSize, 0, 551, 600, 0, true,
+      /* temperature*/ Temperature::kUnknown,
+      /*oldest_ancestor_time*/ kUnknownOldestAncesterTime,
+      /*ts_of_smallest*/ Slice(), /*ts_of_largest*/ Slice(),
+      /*epoch_number*/ 4);
   UpdateVersionStorageInfo();
 
   std::unique_ptr<Compaction> compaction2(
@@ -2962,7 +3000,11 @@ TEST_F(CompactionPickerTest, UniversalMarkedCompactionFullOverlap2) {
   NewVersionStorage(5, kCompactionStyleUniversal);
 
   // Mark file number 4 for compaction
-  Add(0, 4U, "260", "300", 4 * kFileSize, 0, 260, 300, 0, true);
+  Add(0, 4U, "260", "300", 4 * kFileSize, 0, 260, 300, 0, true,
+      /* temperature*/ Temperature::kUnknown,
+      /*oldest_ancestor_time*/ kUnknownOldestAncesterTime,
+      /*ts_of_smallest*/ Slice(), /*ts_of_largest*/ Slice(),
+      /*epoch_number*/ 1);
   Add(3, 5U, "240", "290", 8 * kFileSize, 0, 201, 250);
   Add(4, 3U, "301", "350", 8 * kFileSize, 0, 101, 150);
   Add(4, 6U, "501", "750", 8 * kFileSize, 0, 101, 150);
@@ -2983,8 +3025,17 @@ TEST_F(CompactionPickerTest, UniversalMarkedCompactionFullOverlap2) {
   ASSERT_EQ(1U, compaction->num_input_files(1));
 
   AddVersionStorage();
-  Add(0, 1U, "150", "200", kFileSize, 0, 500, 550);
-  Add(0, 2U, "201", "250", 2 * kFileSize, 0, 401, 450);
+  Add(0, 1U, "150", "200", kFileSize, 0, 500, 550, /*compensated_file_size*/ 0,
+      /*marked_for_compact*/ false, /* temperature*/ Temperature::kUnknown,
+      /*oldest_ancestor_time*/ kUnknownOldestAncesterTime,
+      /*ts_of_smallest*/ Slice(), /*ts_of_largest*/ Slice(),
+      /*epoch_number*/ 3);
+  Add(0, 2U, "201", "250", 2 * kFileSize, 0, 401, 450,
+      /*compensated_file_size*/ 0, /*marked_for_compact*/ false,
+      /* temperature*/ Temperature::kUnknown,
+      /*oldest_ancestor_time*/ kUnknownOldestAncesterTime,
+      /*ts_of_smallest*/ Slice(), /*ts_of_largest*/ Slice(),
+      /*epoch_number*/ 2);
   UpdateVersionStorageInfo();
 
   std::unique_ptr<Compaction> compaction2(
@@ -3150,10 +3201,29 @@ TEST_F(CompactionPickerTest, UniversalMarkedL0Overlap2) {
   NewVersionStorage(1, kCompactionStyleUniversal);
 
   // Mark file number 5 for compaction
-  Add(0, 4U, "260", "300", 1 * kFileSize, 0, 260, 300);
-  Add(0, 5U, "240", "290", 2 * kFileSize, 0, 201, 250, 0, true);
-  Add(0, 3U, "301", "350", 4 * kFileSize, 0, 101, 150);
-  Add(0, 6U, "501", "750", 8 * kFileSize, 0, 50, 100);
+  Add(0, 4U, "260", "300", 1 * kFileSize, 0, 260, 300,
+      /*compensated_file_size*/ 0, /*marked_for_compact*/ false,
+      /* temperature*/ Temperature::kUnknown,
+      /*oldest_ancestor_time*/ kUnknownOldestAncesterTime,
+      /*ts_of_smallest*/ Slice(), /*ts_of_largest*/ Slice(),
+      /*epoch_number*/ 4);
+  Add(0, 5U, "240", "290", 2 * kFileSize, 0, 201, 250, 0, true,
+      /* temperature*/ Temperature::kUnknown,
+      /*oldest_ancestor_time*/ kUnknownOldestAncesterTime,
+      /*ts_of_smallest*/ Slice(), /*ts_of_largest*/ Slice(),
+      /*epoch_number*/ 3);
+  Add(0, 3U, "301", "350", 4 * kFileSize, 0, 101, 150,
+      /*compensated_file_size*/ 0, /*marked_for_compact*/ false,
+      /* temperature*/ Temperature::kUnknown,
+      /*oldest_ancestor_time*/ kUnknownOldestAncesterTime,
+      /*ts_of_smallest*/ Slice(), /*ts_of_largest*/ Slice(),
+      /*epoch_number*/ 2);
+  Add(0, 6U, "501", "750", 8 * kFileSize, 0, 50, 100,
+      /*compensated_file_size*/ 0, /*marked_for_compact*/ false,
+      /* temperature*/ Temperature::kUnknown,
+      /*oldest_ancestor_time*/ kUnknownOldestAncesterTime,
+      /*ts_of_smallest*/ Slice(), /*ts_of_largest*/ Slice(),
+      /*epoch_number*/ 1);
   UpdateVersionStorageInfo();
 
   std::unique_ptr<Compaction> compaction(
@@ -3173,8 +3243,18 @@ TEST_F(CompactionPickerTest, UniversalMarkedL0Overlap2) {
   ASSERT_TRUE(file_map_[6].first->being_compacted);
 
   AddVersionStorage();
-  Add(0, 1U, "150", "200", kFileSize, 0, 500, 550);
-  Add(0, 2U, "201", "250", kFileSize, 0, 401, 450);
+  Add(0, 1U, "150", "200", kFileSize, 0, 500, 550, /*compensated_file_size*/ 0,
+      /*marked_for_compact*/ false,
+      /* temperature*/ Temperature::kUnknown,
+      /*oldest_ancestor_time*/ kUnknownOldestAncesterTime,
+      /*ts_of_smallest*/ Slice(), /*ts_of_largest*/ Slice(),
+      /*epoch_number*/ 6);
+  Add(0, 2U, "201", "250", kFileSize, 0, 401, 450, /*compensated_file_size*/ 0,
+      /*marked_for_compact*/ false,
+      /* temperature*/ Temperature::kUnknown,
+      /*oldest_ancestor_time*/ kUnknownOldestAncesterTime,
+      /*ts_of_smallest*/ Slice(), /*ts_of_largest*/ Slice(),
+      /*epoch_number*/ 5);
   UpdateVersionStorageInfo();
 
   std::unique_ptr<Compaction> compaction2(
@@ -3953,7 +4033,6 @@ TEST_P(PerKeyPlacementCompactionPickerTest,
 INSTANTIATE_TEST_CASE_P(PerKeyPlacementCompactionPickerTest,
                         PerKeyPlacementCompactionPickerTest, ::testing::Bool());
 
-#endif  // ROCKSDB_LITE
 
 }  // namespace ROCKSDB_NAMESPACE
 

db/compaction/compaction_picker_universal.cc

@@ -8,7 +8,6 @@
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 
 #include "db/compaction/compaction_picker_universal.h"
-#ifndef ROCKSDB_LITE
 
 #include <cinttypes>
 #include <limits>
@@ -293,7 +292,7 @@ bool UniversalCompactionPicker::NeedsCompaction(
 Compaction* UniversalCompactionPicker::PickCompaction(
     const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
     const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,
-    LogBuffer* log_buffer, SequenceNumber /* earliest_memtable_seqno */) {
+    LogBuffer* log_buffer) {
   UniversalCompactionBuilder builder(ioptions_, icmp_, cf_name,
                                      mutable_cf_options, mutable_db_options,
                                      vstorage, this, log_buffer);
@@ -308,9 +307,10 @@ void UniversalCompactionBuilder::SortedRun::Dump(char* out_buf,
     if (file->fd.GetPathId() == 0 || !print_path) {
       snprintf(out_buf, out_buf_size, "file %" PRIu64, file->fd.GetNumber());
     } else {
-      snprintf(out_buf, out_buf_size, "file %" PRIu64
-                                      "(path "
-                                      "%" PRIu32 ")",
+      snprintf(out_buf, out_buf_size,
+               "file %" PRIu64
+               "(path "
+               "%" PRIu32 ")",
                file->fd.GetNumber(), file->fd.GetPathId());
     }
   } else {
@@ -399,6 +399,7 @@ Compaction* UniversalCompactionBuilder::PickCompaction() {
   if (!vstorage_->FilesMarkedForPeriodicCompaction().empty()) {
     // Always need to do a full compaction for periodic compaction.
     c = PickPeriodicCompaction();
+    TEST_SYNC_POINT_CALLBACK("PostPickPeriodicCompaction", c);
   }
 
   // Check for size amplification.
@@ -407,6 +408,7 @@ Compaction* UniversalCompactionBuilder::PickCompaction() {
           static_cast<size_t>(
               mutable_cf_options_.level0_file_num_compaction_trigger)) {
     if ((c = PickCompactionToReduceSizeAmp()) != nullptr) {
+      TEST_SYNC_POINT("PickCompactionToReduceSizeAmpReturnNonnullptr");
       ROCKS_LOG_BUFFER(log_buffer_, "[%s] Universal: compacting for size amp\n",
                        cf_name_.c_str());
     } else {
@@ -416,6 +418,7 @@ Compaction* UniversalCompactionBuilder::PickCompaction() {
           mutable_cf_options_.compaction_options_universal.size_ratio;
 
       if ((c = PickCompactionToReduceSortedRuns(ratio, UINT_MAX)) != nullptr) {
+        TEST_SYNC_POINT("PickCompactionToReduceSortedRunsReturnNonnullptr");
         ROCKS_LOG_BUFFER(log_buffer_,
                          "[%s] Universal: compacting for size ratio\n",
                          cf_name_.c_str());
@@ -456,6 +459,7 @@ Compaction* UniversalCompactionBuilder::PickCompaction() {
 
   if (c == nullptr) {
     if ((c = PickDeleteTriggeredCompaction()) != nullptr) {
+      TEST_SYNC_POINT("PickDeleteTriggeredCompactionReturnNonnullptr");
       ROCKS_LOG_BUFFER(log_buffer_,
                        "[%s] Universal: delete triggered compaction\n",
                        cf_name_.c_str());
@@ -1446,4 +1450,3 @@ uint64_t UniversalCompactionBuilder::GetMaxOverlappingBytes() const {
 }
 }  // namespace ROCKSDB_NAMESPACE
 
-#endif  // !ROCKSDB_LITE

db/compaction/compaction_picker_universal.h

@@ -8,7 +8,6 @@
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 
 #pragma once
-#ifndef ROCKSDB_LITE
 
 #include "db/compaction/compaction_picker.h"
 
@@ -18,15 +17,14 @@ class UniversalCompactionPicker : public CompactionPicker {
   UniversalCompactionPicker(const ImmutableOptions& ioptions,
                             const InternalKeyComparator* icmp)
       : CompactionPicker(ioptions, icmp) {}
-  virtual Compaction* PickCompaction(
-      const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
-      const MutableDBOptions& mutable_db_options, VersionStorageInfo* vstorage,
-      LogBuffer* log_buffer,
-      SequenceNumber earliest_memtable_seqno = kMaxSequenceNumber) override;
+  virtual Compaction* PickCompaction(const std::string& cf_name,
+                                     const MutableCFOptions& mutable_cf_options,
+                                     const MutableDBOptions& mutable_db_options,
+                                     VersionStorageInfo* vstorage,
+                                     LogBuffer* log_buffer) override;
   virtual int MaxOutputLevel() const override { return NumberLevels() - 1; }
 
   virtual bool NeedsCompaction(
       const VersionStorageInfo* vstorage) const override;
 };
 }  // namespace ROCKSDB_NAMESPACE
-#endif  // !ROCKSDB_LITE

db/compaction/compaction_service_job.cc

@@ -16,7 +16,6 @@
 #include "options/options_helper.h"
 #include "rocksdb/utilities/options_type.h"
 
-#ifndef ROCKSDB_LITE
 namespace ROCKSDB_NAMESPACE {
 class SubcompactionState;
 
@@ -190,6 +189,7 @@ CompactionJob::ProcessKeyValueCompactionWithCompactionService(
     meta.largest.DecodeFrom(file.largest_internal_key);
     meta.oldest_ancester_time = file.oldest_ancester_time;
     meta.file_creation_time = file.file_creation_time;
+    meta.epoch_number = file.epoch_number;
     meta.marked_for_compaction = file.marked_for_compaction;
     meta.unique_id = file.unique_id;
 
@@ -333,8 +333,9 @@ Status CompactionServiceCompactionJob::Run() {
         MakeTableFileName(meta.fd.GetNumber()), meta.fd.smallest_seqno,
         meta.fd.largest_seqno, meta.smallest.Encode().ToString(),
         meta.largest.Encode().ToString(), meta.oldest_ancester_time,
-        meta.file_creation_time, output_file.validator.GetHash(),
-        meta.marked_for_compaction, meta.unique_id);
+        meta.file_creation_time, meta.epoch_number,
+        output_file.validator.GetHash(), meta.marked_for_compaction,
+        meta.unique_id);
   }
   InternalStats::CompactionStatsFull compaction_stats;
   sub_compact->AggregateCompactionStats(compaction_stats);
@@ -489,6 +490,10 @@ static std::unordered_map<std::string, OptionTypeInfo>
          {offsetof(struct CompactionServiceOutputFile, file_creation_time),
           OptionType::kUInt64T, OptionVerificationType::kNormal,
           OptionTypeFlags::kNone}},
+        {"epoch_number",
+         {offsetof(struct CompactionServiceOutputFile, epoch_number),
+          OptionType::kUInt64T, OptionVerificationType::kNormal,
+          OptionTypeFlags::kNone}},
         {"paranoid_hash",
          {offsetof(struct CompactionServiceOutputFile, paranoid_hash),
           OptionType::kUInt64T, OptionVerificationType::kNormal,
@@ -826,4 +831,3 @@ bool CompactionServiceInput::TEST_Equals(CompactionServiceInput* other,
 #endif  // NDEBUG
 }  // namespace ROCKSDB_NAMESPACE
 
-#endif  // !ROCKSDB_LITE

db/compaction/compaction_service_test.cc

@@ -3,7 +3,6 @@
 //  COPYING file in the root directory) and Apache 2.0 License
 //  (found in the LICENSE.Apache file in the root directory).
 
-#ifndef ROCKSDB_LITE
 
 #include "db/db_test_util.h"
 #include "port/stack_trace.h"
@@ -954,13 +953,3 @@ int main(int argc, char** argv) {
   return RUN_ALL_TESTS();
 }
 
-#else
-#include <stdio.h>
-
-int main(int /*argc*/, char** /*argv*/) {
-  fprintf(stderr,
-          "SKIPPED as CompactionService is not supported in ROCKSDB_LITE\n");
-  return 0;
-}
-
-#endif  // ROCKSDB_LITE

db/compaction/sst_partitioner.cc

@@ -15,11 +15,9 @@
 namespace ROCKSDB_NAMESPACE {
 static std::unordered_map<std::string, OptionTypeInfo>
     sst_fixed_prefix_type_info = {
-#ifndef ROCKSDB_LITE
         {"length",
          {0, OptionType::kSizeT, OptionVerificationType::kNormal,
           OptionTypeFlags::kNone}},
-#endif  // ROCKSDB_LITE
 };
 
 SstPartitionerFixedPrefixFactory::SstPartitionerFixedPrefixFactory(size_t len)
@@ -58,7 +56,6 @@ std::shared_ptr<SstPartitionerFactory> NewSstPartitionerFixedPrefixFactory(
   return std::make_shared<SstPartitionerFixedPrefixFactory>(prefix_len);
 }
 
-#ifndef ROCKSDB_LITE
 namespace {
 static int RegisterSstPartitionerFactories(ObjectLibrary& library,
                                            const std::string& /*arg*/) {
@@ -73,18 +70,14 @@ static int RegisterSstPartitionerFactories(ObjectLibrary& library,
   return 1;
 }
 }  // namespace
-#endif  // ROCKSDB_LITE
 
 Status SstPartitionerFactory::CreateFromString(
     const ConfigOptions& options, const std::string& value,
     std::shared_ptr<SstPartitionerFactory>* result) {
-#ifndef ROCKSDB_LITE
   static std::once_flag once;
   std::call_once(once, [&]() {
     RegisterSstPartitionerFactories(*(ObjectLibrary::Default().get()), "");
   });
-#endif  // ROCKSDB_LITE
-  return LoadSharedObject<SstPartitionerFactory>(options, value, nullptr,
-                                                 result);
+  return LoadSharedObject<SstPartitionerFactory>(options, value, result);
 }
 }  // namespace ROCKSDB_NAMESPACE