Fix GitHub issue #3716 : gcc-8 warnings

Summary: Fix the following gcc-8 warnings: - conflicting C language linkage declaration [-Werror] - writing to an object with no trivial copy-assignment [-Werror=class-memaccess] - array subscript -1 is below array bounds [-Werror=array-bounds] Solves https://github.com/facebook/rocksdb/issues/3716 Closes https://github.com/facebook/rocksdb/pull/3736 Differential Revision: D7684161 Pulled By: yiwu-arbug fbshipit-source-id: 47c0423d26b74add251f1d3595211eee1e41e54a
Making platform 007 (gcc 7) default in build_detect_platform.sh (#5947 )
2026-11-07 11:32:34 -05:00 · 2019-10-30 13:15:18 -07:00 · 2019-10-30 13:14:07 -07:00 · 2019-10-30 13:13:58 -07:00 · 2018-06-18 09:55:01 -07:00 · 2018-06-18 09:50:10 -07:00
337 changed files with 19704 additions and 2686 deletions
@@ -34,12 +34,15 @@

 cmake_minimum_required(VERSION 2.8.12)
 project(rocksdb)
+enable_language(CXX)
+enable_language(C)
+enable_language(ASM)

 if(POLICY CMP0042)
  cmake_policy(SET CMP0042 NEW)
 endif()

-list(APPEND CMAKE_MODULE_PATH "${CMAKE_SOURCE_DIR}/cmake/modules/")
+list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_LIST_DIR}/cmake/modules/")

 option(WITH_JEMALLOC "build with JeMalloc" OFF)
 option(WITH_SNAPPY "build with SNAPPY" OFF)
@@ -129,20 +132,19 @@ endif()

 string(REGEX REPLACE "[^0-9a-f]+" "" GIT_SHA "${GIT_SHA}")

-if(NOT WIN32)
-  execute_process(COMMAND
-      "./build_tools/version.sh" "full"
-      WORKING_DIRECTORY ${PROJECT_SOURCE_DIR}
-      OUTPUT_VARIABLE ROCKSDB_VERSION
-  )
-  string(STRIP "${ROCKSDB_VERSION}" ROCKSDB_VERSION)
-  execute_process(COMMAND
-      "./build_tools/version.sh" "major"
-      WORKING_DIRECTORY ${PROJECT_SOURCE_DIR}
-      OUTPUT_VARIABLE ROCKSDB_VERSION_MAJOR
-  )
-  string(STRIP "${ROCKSDB_VERSION_MAJOR}" ROCKSDB_VERSION_MAJOR)
-endif()
+set(SH_CMD "sh")
+execute_process(COMMAND
+  ${SH_CMD} -c "build_tools/version.sh full"
+  WORKING_DIRECTORY ${PROJECT_SOURCE_DIR}
+  OUTPUT_VARIABLE ROCKSDB_VERSION
+)
+string(STRIP "${ROCKSDB_VERSION}" ROCKSDB_VERSION)
+execute_process(COMMAND
+  ${SH_CMD} -c "build_tools/version.sh major"
+  WORKING_DIRECTORY ${PROJECT_SOURCE_DIR}
+  OUTPUT_VARIABLE ROCKSDB_VERSION_MAJOR
+)
+string(STRIP "${ROCKSDB_VERSION_MAJOR}" ROCKSDB_VERSION_MAJOR)

 option(WITH_MD_LIBRARY "build with MD" ON)
 if(WIN32 AND MSVC)
@@ -178,6 +180,18 @@ else()
  endif()
 endif()

+include(CheckCCompilerFlag)
+if(CMAKE_SYSTEM_PROCESSOR MATCHES "ppc64le")
+  CHECK_C_COMPILER_FLAG("-maltivec" HAS_ALTIVEC)
+  if(HAS_ALTIVEC)
+    message(STATUS " HAS_ALTIVEC yes")
+    set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -maltivec")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -maltivec")
+    set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mcpu=power8")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -mcpu=power8")
+  endif(HAS_ALTIVEC)
+endif(CMAKE_SYSTEM_PROCESSOR MATCHES "ppc64le")
+
 option(PORTABLE "build a portable binary" OFF)
 option(FORCE_SSE42 "force building with SSE4.2, even when PORTABLE=ON" OFF)
 if(PORTABLE)
@@ -190,7 +204,9 @@ else()
  if(MSVC)
    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /arch:AVX2")
  else()
-    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -march=native")
+    if(NOT HAVE_POWER8)
+      set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -march=native")
+    endif()
  endif()
 endif()

@@ -586,6 +602,7 @@ set(SOURCES
        utilities/leveldb_options/leveldb_options.cc
        utilities/lua/rocks_lua_compaction_filter.cc
        utilities/memory/memory_util.cc
+        utilities/merge_operators/bytesxor.cc
        utilities/merge_operators/max.cc
        utilities/merge_operators/put.cc
        utilities/merge_operators/string_append/stringappend.cc
@@ -626,6 +643,12 @@ set_source_files_properties(
  endif()
 endif()

+if(HAVE_POWER8)
+  list(APPEND SOURCES
+    util/crc32c_ppc.c
+    util/crc32c_ppc_asm.S)
+endif(HAVE_POWER8)
+
 if(WIN32)
  list(APPEND SOURCES
    port/win/io_win.cc
@@ -720,7 +743,7 @@ if(NOT WIN32 OR ROCKSDB_INSTALL_ON_WINDOWS)
  set(package_config_destination ${CMAKE_INSTALL_LIBDIR}/cmake/rocksdb)

  configure_package_config_file(
-    ${CMAKE_SOURCE_DIR}/cmake/RocksDBConfig.cmake.in RocksDBConfig.cmake
+    ${CMAKE_CURRENT_LIST_DIR}/cmake/RocksDBConfig.cmake.in RocksDBConfig.cmake
    INSTALL_DESTINATION ${package_config_destination}
  )

@@ -950,17 +973,18 @@ if(WITH_TESTS)

  foreach(sourcefile ${TEST_EXES})
      get_filename_component(exename ${sourcefile} NAME_WE)
-      add_executable(${exename}${ARTIFACT_SUFFIX} ${sourcefile}
+      add_executable(${CMAKE_PROJECT_NAME}_${exename}${ARTIFACT_SUFFIX} ${sourcefile}
        $<TARGET_OBJECTS:testharness>)
-      set_target_properties(${exename}${ARTIFACT_SUFFIX}
+      set_target_properties(${CMAKE_PROJECT_NAME}_${exename}${ARTIFACT_SUFFIX}
        PROPERTIES EXCLUDE_FROM_DEFAULT_BUILD_RELEASE 1
        EXCLUDE_FROM_DEFAULT_BUILD_MINRELEASE 1
        EXCLUDE_FROM_DEFAULT_BUILD_RELWITHDEBINFO 1
+        OUTPUT_NAME ${exename}${ARTIFACT_SUFFIX}
        )
-      target_link_libraries(${exename}${ARTIFACT_SUFFIX} testutillib${ARTIFACT_SUFFIX} gtest ${LIBS})
+      target_link_libraries(${CMAKE_PROJECT_NAME}_${exename}${ARTIFACT_SUFFIX} testutillib${ARTIFACT_SUFFIX} gtest ${LIBS})
      if(NOT "${exename}" MATCHES "db_sanity_test")
        add_test(NAME ${exename} COMMAND ${exename}${ARTIFACT_SUFFIX})
-        add_dependencies(check ${exename}${ARTIFACT_SUFFIX})
+        add_dependencies(check ${CMAKE_PROJECT_NAME}_${exename}${ARTIFACT_SUFFIX})
      endif()
  endforeach(sourcefile ${TEST_EXES})

@@ -1,5 +1,68 @@
 # Rocksdb Change Log
-## Unreleased
+## 5.12.5 (6/18/2018)
+### Bug Fixes
+* Fix regression bug of Prev() with ReadOptions.iterate_upper_bound.
+
+## 5.12.3 (04/09/2018)
+### Bug Fixes
+* Fix memory leak when pin_l0_filter_and_index_blocks_in_cache is used with partitioned filters
+* Fix memory leak in two_level_iterator
+
+## 5.12.2 (3/23/2018)
+### Bug Fixes
+* Fsync after writing global seq number to the ingestion file in ExternalSstFileIngestionJob.
+
+### Java API Changes
+* Add `BlockBasedTableConfig.setBlockCache` to allow sharing a block cache across DB instances.
+
+## 5.12.1 (3/16/2018)
+### Public API Change
+* RocksDBOptionsParser::Parse()'s `ignore_unknown_options` argument will only be effective if the option file shows it is generated using a higher version of RocksDB than the current version.
+
+### New Features
+* Avoid unnecessarily flushing in `CompactRange()` when the range specified by the user does not overlap unflushed memtables.
+
+### Bug Fixes
+* Fix WAL corruption caused by race condition between user write thread and backup/checkpoint thread.
+
+## 5.12.0 (2/14/2018)
+### Public API Change
+* Iterator::SeekForPrev is now a pure virtual method. This is to prevent user who implement the Iterator interface fail to implement SeekForPrev by mistake.
+* Add `include_end` option to make the range end exclusive when `include_end == false` in `DeleteFilesInRange()`.
+* Add `CompactRangeOptions::allow_write_stall`, which makes `CompactRange` start working immediately, even if it causes user writes to stall. The default value is false, meaning we add delay to `CompactRange` calls until stalling can be avoided when possible. Note this delay is not present in previous RocksDB versions.
+* Creating checkpoint with empty directory now returns `Status::InvalidArgument`; previously, it returned `Status::IOError`.
+* Adds a BlockBasedTableOption to turn off index block compression.
+* Close() method now returns a status when closing a db.
+
+### New Features
+* Improve the performance of iterators doing long range scans by using readahead.
+* Add new function `DeleteFilesInRanges()` to delete files in multiple ranges at once for better performance.
+* FreeBSD build support for RocksDB and RocksJava.
+* Improved performance of long range scans with readahead.
+* Updated to and now continuously tested in Visual Studio 2017.
+
+### Bug Fixes
+* Fix `DisableFileDeletions()` followed by `GetSortedWalFiles()` to not return obsolete WAL files that `PurgeObsoleteFiles()` is going to delete.
+* Fix Handle error return from WriteBuffer() during WAL file close and DB close.
+* Fix advance reservation of arena block addresses.
+* Fix handling of empty string as checkpoint directory.
+
+## 5.11.0 (01/08/2018)
+### Public API Change
+* Add `autoTune` and `getBytesPerSecond()` to RocksJava RateLimiter
+
+### New Features
+* Add a new histogram stat called rocksdb.db.flush.micros for memtable flush.
+* Add "--use_txn" option to use transactional API in db_stress.
+* Disable onboard cache for compaction output in Windows platform.
+* Improve the performance of iterators doing long range scans by using readahead.
+
+### Bug Fixes
+* Fix a stack-use-after-scope bug in ForwardIterator.
+* Fix builds on platforms including Linux, Windows, and PowerPC.
+* Fix buffer overrun in backup engine for DBs with huge number of files.
+* Fix a mislabel bug for bottom-pri compaction threads.
+* Fix DB::Flush() keep waiting after flush finish under certain condition.

 ## 5.10.0 (12/11/2017)
 ### Public API Change
@@ -107,6 +107,41 @@ to build a portable binary, add `PORTABLE=1` before your make commands, like thi
            * run `brew tap homebrew/versions; brew install gcc48 --use-llvm` to install gcc 4.8 (or higher).
    * run `brew install rocksdb`

+* **FreeBSD** (11.01):
+
+    * You can either install RocksDB from the Ports system using `cd /usr/ports/databases/rocksdb && make install`, or you can follow the details below to install dependencies and compile from source code:
+
+    * Install the dependencies for RocksDB:
+
+        export BATCH=YES
+        cd /usr/ports/devel/gmake && make install
+        cd /usr/ports/devel/gflags && make install
+
+        cd /usr/ports/archivers/snappy && make install
+        cd /usr/ports/archivers/bzip2 && make install
+        cd /usr/ports/archivers/liblz4 && make install
+        cd /usr/ports/archivesrs/zstd && make install
+
+        cd /usr/ports/devel/git && make install
+
+
+    * Install the dependencies for RocksJava (optional):
+
+        export BATCH=yes
+        cd /usr/ports/java/openjdk7 && make install
+
+    * Build RocksDB from source:
+        cd ~
+        git clone https://github.com/facebook/rocksdb.git
+        cd rocksdb
+        gmake static_lib
+
+    * Build RocksJava from source (optional):
+        cd rocksdb
+        export JAVA_HOME=/usr/local/openjdk7
+        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`.

@@ -15,3 +15,4 @@ This is the list of all known third-party language bindings for RocksDB. If some
 * D programming language - https://github.com/b1naryth1ef/rocksdb
 * Erlang - https://gitlab.com/barrel-db/erlang-rocksdb
 * Elixir - https://github.com/urbint/rox
+* Nim - https://github.com/status-im/nim-rocksdb
@@ -232,9 +232,13 @@ endif
 # USAN doesn't work well with jemalloc. If we're compiling with USAN, we should use regular malloc.
 ifdef COMPILE_WITH_UBSAN
 	DISABLE_JEMALLOC=1
-	EXEC_LDFLAGS += -fsanitize=undefined
-	PLATFORM_CCFLAGS += -fsanitize=undefined -DROCKSDB_UBSAN_RUN
-	PLATFORM_CXXFLAGS += -fsanitize=undefined -DROCKSDB_UBSAN_RUN
+	# Suppress alignment warning because murmurhash relies on casting unaligned
+	# memory to integer. Fixing it may cause performance regression. 3-way crc32
+	# relies on it too, although it can be rewritten to eliminate with minimal
+	# performance regression.
+	EXEC_LDFLAGS += -fsanitize=undefined -fno-sanitize-recover=all
+	PLATFORM_CCFLAGS += -fsanitize=undefined -fno-sanitize-recover=all -DROCKSDB_UBSAN_RUN
+	PLATFORM_CXXFLAGS += -fsanitize=undefined -fno-sanitize-recover=all -DROCKSDB_UBSAN_RUN
 endif

 ifdef ROCKSDB_VALGRIND_RUN
@@ -523,6 +527,10 @@ PARALLEL_TEST = \
 	transaction_test \
 	write_prepared_transaction_test \

+# options_settable_test doesn't pass with UBSAN as we use hack in the test
+ifdef COMPILE_WITH_UBSAN
+        TESTS := $(shell echo $(TESTS) | sed 's/\boptions_settable_test\b//g')
+endif
 SUBSET := $(TESTS)
 ifdef ROCKSDBTESTS_START
        SUBSET := $(shell echo $(SUBSET) | sed 's/^.*$(ROCKSDBTESTS_START)/$(ROCKSDBTESTS_START)/')
@@ -1553,8 +1561,8 @@ ZLIB_DOWNLOAD_BASE ?= http://zlib.net
 BZIP2_VER ?= 1.0.6
 BZIP2_SHA256 ?= a2848f34fcd5d6cf47def00461fcb528a0484d8edef8208d6d2e2909dc61d9cd
 BZIP2_DOWNLOAD_BASE ?= http://www.bzip.org
-SNAPPY_VER ?= 1.1.7
-SNAPPY_SHA256 ?= 3dfa02e873ff51a11ee02b9ca391807f0c8ea0529a4924afa645fbf97163f9d4
+SNAPPY_VER ?= 1.1.4
+SNAPPY_SHA256 ?= 134bfe122fd25599bb807bb8130e7ba6d9bdb851e0b16efcb83ac4f5d0b70057
 SNAPPY_DOWNLOAD_BASE ?= https://github.com/google/snappy/releases/download
 LZ4_VER ?= 1.8.0
 LZ4_SHA256 ?= 2ca482ea7a9bb103603108b5a7510b7592b90158c151ff50a28f1ca8389fccf6
@@ -1574,7 +1582,7 @@ else
 endif
 endif
 ifeq ($(PLATFORM), OS_FREEBSD)
-	JAVA_INCLUDE += -I$(JAVA_HOME)/include/freebsd
+	JAVA_INCLUDE = -I$(JAVA_HOME)/include -I$(JAVA_HOME)/include/freebsd
 	ROCKSDBJNILIB = librocksdbjni-freebsd$(ARCH).so
 	ROCKSDB_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-freebsd$(ARCH).jar
 endif
@@ -1600,7 +1608,7 @@ libz.a:
 		exit 1; \
 	fi
 	tar xvzf zlib-$(ZLIB_VER).tar.gz
-	cd zlib-$(ZLIB_VER) && CFLAGS='-fPIC ${EXTRA_CFLAGS}' LDFLAGS='${EXTRA_LDFLAGS}' ./configure --static && make
+	cd zlib-$(ZLIB_VER) && CFLAGS='-fPIC ${EXTRA_CFLAGS}' LDFLAGS='${EXTRA_LDFLAGS}' ./configure --static && $(MAKE)
 	cp zlib-$(ZLIB_VER)/libz.a .

 libbz2.a:
@@ -1612,7 +1620,7 @@ libbz2.a:
 		exit 1; \
 	fi
 	tar xvzf bzip2-$(BZIP2_VER).tar.gz
-	cd bzip2-$(BZIP2_VER) && make CFLAGS='-fPIC -O2 -g -D_FILE_OFFSET_BITS=64 ${EXTRA_CFLAGS}' AR='ar ${EXTRA_ARFLAGS}'
+	cd bzip2-$(BZIP2_VER) && $(MAKE) CFLAGS='-fPIC -O2 -g -D_FILE_OFFSET_BITS=64 ${EXTRA_CFLAGS}' AR='ar ${EXTRA_ARFLAGS}'
 	cp bzip2-$(BZIP2_VER)/libbz2.a .

 libsnappy.a:
@@ -1625,7 +1633,7 @@ libsnappy.a:
 	fi
 	tar xvzf snappy-$(SNAPPY_VER).tar.gz
 	cd snappy-$(SNAPPY_VER) && CFLAGS='${EXTRA_CFLAGS}' CXXFLAGS='${EXTRA_CXXFLAGS}' LDFLAGS='${EXTRA_LDFLAGS}' ./configure --with-pic --enable-static --disable-shared
-	cd snappy-$(SNAPPY_VER) && make ${SNAPPY_MAKE_TARGET}
+	cd snappy-$(SNAPPY_VER) && $(MAKE) ${SNAPPY_MAKE_TARGET}
 	cp snappy-$(SNAPPY_VER)/.libs/libsnappy.a .

 liblz4.a:
@@ -1638,7 +1646,7 @@ liblz4.a:
 		exit 1; \
 	fi
 	tar xvzf lz4-$(LZ4_VER).tar.gz
-	cd lz4-$(LZ4_VER)/lib && make CFLAGS='-fPIC -O2 ${EXTRA_CFLAGS}' all
+	cd lz4-$(LZ4_VER)/lib && $(MAKE) CFLAGS='-fPIC -O2 ${EXTRA_CFLAGS}' all
 	cp lz4-$(LZ4_VER)/lib/liblz4.a .

 libzstd.a:
@@ -1651,7 +1659,7 @@ libzstd.a:
 		exit 1; \
 	fi
 	tar xvzf zstd-$(ZSTD_VER).tar.gz
-	cd zstd-$(ZSTD_VER)/lib && DESTDIR=. PREFIX= make CFLAGS='-fPIC -O2 ${EXTRA_CFLAGS}' install
+	cd zstd-$(ZSTD_VER)/lib && DESTDIR=. PREFIX= $(MAKE) CFLAGS='-fPIC -O2 ${EXTRA_CFLAGS}' install
 	cp zstd-$(ZSTD_VER)/lib/libzstd.a .

 # A version of each $(LIBOBJECTS) compiled with -fPIC and a fixed set of static compression libraries
@@ -1750,7 +1758,7 @@ JAVA_C_LIBOBJECTS = $(patsubst %.c.o,jl/%.c.o,$(JAVA_C_OBJECTS))
 JAVA_ASM_LIBOBJECTS = $(patsubst %.S.o,jl/%.S.o,$(JAVA_ASM_OBJECTS))
 endif

-java_libobjects = $(patsubst %,jl/%,$(LIBOBJECTS))
+java_libobjects = $(patsubst %,jl/%,$(LIB_CC_OBJECTS))
 CLEAN_FILES += jl
 java_all_libobjects = $(java_libobjects)

@@ -1,7 +1,8 @@
 ## RocksDB: A Persistent Key-Value Store for Flash and RAM Storage

-[![Build Status](https://travis-ci.org/facebook/rocksdb.svg?branch=master)](https://travis-ci.org/facebook/rocksdb)
-[![Build status](https://ci.appveyor.com/api/projects/status/fbgfu0so3afcno78/branch/master?svg=true)](https://ci.appveyor.com/project/Facebook/rocksdb/branch/master)
+[![Linux/Mac Build Status](https://travis-ci.org/facebook/rocksdb.svg?branch=master)](https://travis-ci.org/facebook/rocksdb)
+[![Windows Build status](https://ci.appveyor.com/api/projects/status/fbgfu0so3afcno78/branch/master?svg=true)](https://ci.appveyor.com/project/Facebook/rocksdb/branch/master)
+[![PPC64le Build Status](http://140.211.168.68:8080/buildStatus/icon?job=Rocksdb)](http://140.211.168.68:8080/job/Rocksdb)


 RocksDB is developed and maintained by Facebook Database Engineering Team.
@@ -242,6 +242,7 @@ cpp_library(
        "utilities/leveldb_options/leveldb_options.cc",
        "utilities/lua/rocks_lua_compaction_filter.cc",
        "utilities/memory/memory_util.cc",
+        "utilities/merge_operators/bytesxor.cc",
        "utilities/merge_operators/max.cc",
        "utilities/merge_operators/put.cc",
        "utilities/merge_operators/string_append/stringappend.cc",
@@ -14,6 +14,11 @@ Vagrant.configure("2") do |config|
    box.vm.box = "chef/centos-6.5"
  end

+  config.vm.define "centos7" do |box|
+    box.vm.box = "centos/7"
+    box.vm.provision "shell", path: "build_tools/setup_centos7.sh"
+  end
+
  config.vm.define "FreeBSD10" do |box|
    box.vm.guest = :freebsd
    box.vm.box = "robin/freebsd-10"
@@ -1,9 +1,9 @@
 version: 1.0.{build}
-image: Visual Studio 2015
+image: Visual Studio 2017
 before_build:
 - md %APPVEYOR_BUILD_FOLDER%\build
 - cd %APPVEYOR_BUILD_FOLDER%\build
- cmake -G "Visual Studio 14 Win64" -DOPTDBG=1 -DWITH_XPRESS=1 -DPORTABLE=1 ..
+- cmake -G "Visual Studio 15 Win64" -DOPTDBG=1 -DWITH_XPRESS=1 -DPORTABLE=1 ..
 - cd ..
 build:
  project: build\rocksdb.sln
@@ -36,7 +36,7 @@ def parse_src_mk(repo_path):
 # get all .cc / .c files
 def get_cc_files(repo_path):
    cc_files = []
-    for root, dirnames, filenames in os.walk(repo_path):
+    for root, dirnames, filenames in os.walk(repo_path):  # noqa: B007 T25377293 Grandfathered in
        root = root[(len(repo_path) + 1):]
        if "java" in root:
            # Skip java
@@ -1,4 +1,5 @@
 #!/usr/bin/env bash
 # Create a tmp directory for the test to use
 TEST_DIR=$(mktemp -d /dev/shm/fbcode_rocksdb_XXXXXXX)
+# shellcheck disable=SC2068
 TEST_TMPDIR="$TEST_DIR" $@ && rm -rf "$TEST_DIR"
@@ -51,11 +51,13 @@ if [ -z "$ROCKSDB_NO_FBCODE" -a -d /mnt/gvfs/third-party ]; then
    FBCODE_BUILD="true"
    # If we're compiling with TSAN we need pic build
    PIC_BUILD=$COMPILE_WITH_TSAN
-    if [ -z "$ROCKSDB_FBCODE_BUILD_WITH_481" ]; then
-      source "$PWD/build_tools/fbcode_config.sh"
-    else
+    if [ -n "$ROCKSDB_FBCODE_BUILD_WITH_481" ]; then
      # we need this to build with MySQL. Don't use for other purposes.
      source "$PWD/build_tools/fbcode_config4.8.1.sh"
+    elif [ -n "$ROCKSDB_FBCODE_BUILD_WITH_5xx" ]; then
+      source "$PWD/build_tools/fbcode_config.sh"
+    else
+      source "$PWD/build_tools/fbcode_config_platform007.sh"
    fi
 fi

@@ -141,6 +143,7 @@ case "$TARGET_OS" in
        ;;
    FreeBSD)
        PLATFORM=OS_FREEBSD
+        CXX=clang++
        COMMON_FLAGS="$COMMON_FLAGS -fno-builtin-memcmp -D_REENTRANT -DOS_FREEBSD"
        PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -lpthread"
        # PORT_FILES=port/freebsd/freebsd_specific.cc
@@ -481,6 +484,9 @@ if test -z "$PORTABLE"; then
    COMMON_FLAGS="$COMMON_FLAGS -mcpu=$POWER -mtune=$POWER "
  elif test -n "`echo $TARGET_ARCHITECTURE | grep ^s390x`"; then
    COMMON_FLAGS="$COMMON_FLAGS -march=z10 "
+  elif test -n "`echo $TARGET_ARCHITECTURE | grep ^arm`"; then
+    # TODO: Handle this with approprite options.
+    COMMON_FLAGS="$COMMON_FLAGS"
  elif [ "$TARGET_OS" != AIX ] && [ "$TARGET_OS" != SunOS ]; then
    COMMON_FLAGS="$COMMON_FLAGS -march=native "
  elif test "$USE_SSE"; then
@@ -13,10 +13,12 @@ error=0

 function log {
  DATE=`date +%Y-%m-%d:%H:%M:%S`
+  # shellcheck disable=SC2068
  echo $DATE $@
 }

 function log_err {
+  # shellcheck disable=SC2145
  log "ERROR: $@ Error code: $error."
 }

@@ -1,3 +1,4 @@
+# shellcheck disable=SC2148
 GCC_BASE=/mnt/gvfs/third-party2/gcc/8219ec1bcedf8ad9da05e121e193364de2cc4f61/5.x/centos6-native/c447969
 CLANG_BASE=/mnt/gvfs/third-party2/llvm-fb/64d8d58e3d84f8bde7a029763d4f5baf39d0d5b9/stable/centos6-native/6aaf4de
 LIBGCC_BASE=/mnt/gvfs/third-party2/libgcc/ba9be983c81de7299b59fe71950c664a84dcb5f8/5.x/gcc-5-glibc-2.23/339d858
@@ -1,3 +1,4 @@
+# shellcheck disable=SC2148
 GCC_BASE=/mnt/gvfs/third-party2/gcc/cf7d14c625ce30bae1a4661c2319c5a283e4dd22/4.8.1/centos6-native/cc6c9dc
 CLANG_BASE=/mnt/gvfs/third-party2/llvm-fb/8598c375b0e94e1448182eb3df034704144a838d/stable/centos6-native/3f16ddd
 LIBGCC_BASE=/mnt/gvfs/third-party2/libgcc/d6e0a7da6faba45f5e5b1638f9edd7afc2f34e7d/4.8.1/gcc-4.8.1-glibc-2.17/8aac7fc
@@ -0,0 +1,18 @@
+GCC_BASE=/mnt/gvfs/third-party2/gcc/6e8e715624fd15256a7970073387793dfcf79b46/7.x/centos7-native/b2ef2b6
+CLANG_BASE=/mnt/gvfs/third-party2/llvm-fb/ef37e1faa1c29782abfac1ae65a291b9b7966f6d/stable/centos7-native/c9f9104
+LIBGCC_BASE=/mnt/gvfs/third-party2/libgcc/c67031f0f739ac61575a061518d6ef5038f99f90/7.x/platform007/5620abc
+GLIBC_BASE=/mnt/gvfs/third-party2/glibc/60d6f124a78798b73944f5ba87c2306ae3460153/2.26/platform007/f259413
+SNAPPY_BASE=/mnt/gvfs/third-party2/snappy/7f9bdaada18f59bc27ec2b0871eb8a6144343aef/1.1.3/platform007/ca4da3d
+ZLIB_BASE=/mnt/gvfs/third-party2/zlib/22c2d65676fb7c23cfa797c4f6937f38b026f3cf/1.2.8/platform007/ca4da3d
+BZIP2_BASE=/mnt/gvfs/third-party2/bzip2/dc49a21c5fceec6456a7a28a94dcd16690af1337/1.0.6/platform007/ca4da3d
+LZ4_BASE=/mnt/gvfs/third-party2/lz4/907b498203d297947f3bb70b9466f47e100f1873/r131/platform007/ca4da3d
+ZSTD_BASE=/mnt/gvfs/third-party2/zstd/3ee276cbacfad3074e3f07bf826ac47f06970f4e/1.3.5/platform007/15a3614
+GFLAGS_BASE=/mnt/gvfs/third-party2/gflags/0b9929d2588991c65a57168bf88aff2db87c5d48/2.2.0/platform007/ca4da3d
+JEMALLOC_BASE=/mnt/gvfs/third-party2/jemalloc/9c910d36d6235cc40e8ff559358f1833452300ca/master/platform007/5b0f53e
+NUMA_BASE=/mnt/gvfs/third-party2/numa/9cbf2460284c669ed19c3ccb200a71f7dd7e53c7/2.0.11/platform007/ca4da3d
+LIBUNWIND_BASE=/mnt/gvfs/third-party2/libunwind/bf3d7497fe4e6d007354f0adffa16ce3003f8338/1.3/platform007/6f3e0a9
+TBB_BASE=/mnt/gvfs/third-party2/tbb/ff4e0b093534704d8abab678a4fd7f5ea7b094c7/2018_U5/platform007/ca4da3d
+KERNEL_HEADERS_BASE=/mnt/gvfs/third-party2/kernel-headers/b5c4a61a5c483ba24722005ae07895971a2ac707/fb/platform007/da39a3e
+BINUTILS_BASE=/mnt/gvfs/third-party2/binutils/92ff90349e2f43ea0a8246d8b1cf17b6869013e3/2.29.1/centos7-native/da39a3e
+VALGRIND_BASE=/mnt/gvfs/third-party2/valgrind/f3f697a28122e6bcd513273dd9c1ff23852fc59f/3.13.0/platform007/ca4da3d
+LUA_BASE=/mnt/gvfs/third-party2/lua/f0cd714433206d5139df61659eb7b28b1dea6683/5.3.4/platform007/5007832
@@ -0,0 +1,157 @@
+#!/bin/sh
+#
+# 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_platform007.sh"
+
+CFLAGS=""
+
+# libgcc
+LIBGCC_INCLUDE="$LIBGCC_BASE/include/c++/7.3.0"
+LIBGCC_LIBS=" -L $LIBGCC_BASE/lib"
+
+# glibc
+GLIBC_INCLUDE="$GLIBC_BASE/include"
+GLIBC_LIBS=" -L $GLIBC_BASE/lib"
+
+# snappy
+SNAPPY_INCLUDE=" -I $SNAPPY_BASE/include/"
+if test -z $PIC_BUILD; then
+  SNAPPY_LIBS=" $SNAPPY_BASE/lib/libsnappy.a"
+else
+  SNAPPY_LIBS=" $SNAPPY_BASE/lib/libsnappy_pic.a"
+fi
+CFLAGS+=" -DSNAPPY"
+
+if test -z $PIC_BUILD; then
+  # location of zlib headers and libraries
+  ZLIB_INCLUDE=" -I $ZLIB_BASE/include/"
+  ZLIB_LIBS=" $ZLIB_BASE/lib/libz.a"
+  CFLAGS+=" -DZLIB"
+
+  # location of bzip headers and libraries
+  BZIP_INCLUDE=" -I $BZIP2_BASE/include/"
+  BZIP_LIBS=" $BZIP2_BASE/lib/libbz2.a"
+  CFLAGS+=" -DBZIP2"
+
+  LZ4_INCLUDE=" -I $LZ4_BASE/include/"
+  LZ4_LIBS=" $LZ4_BASE/lib/liblz4.a"
+  CFLAGS+=" -DLZ4"
+fi
+
+ZSTD_INCLUDE=" -I $ZSTD_BASE/include/"
+if test -z $PIC_BUILD; then
+  ZSTD_LIBS=" $ZSTD_BASE/lib/libzstd.a"
+else
+  ZSTD_LIBS=" $ZSTD_BASE/lib/libzstd_pic.a"
+fi
+CFLAGS+=" -DZSTD"
+
+# location of gflags headers and libraries
+GFLAGS_INCLUDE=" -I $GFLAGS_BASE/include/"
+if test -z $PIC_BUILD; then
+  GFLAGS_LIBS=" $GFLAGS_BASE/lib/libgflags.a"
+else
+  GFLAGS_LIBS=" $GFLAGS_BASE/lib/libgflags_pic.a"
+fi
+CFLAGS+=" -DGFLAGS=gflags"
+
+# location of jemalloc
+JEMALLOC_INCLUDE=" -I $JEMALLOC_BASE/include/"
+JEMALLOC_LIB=" $JEMALLOC_BASE/lib/libjemalloc.a"
+
+if test -z $PIC_BUILD; then
+  # location of numa
+  NUMA_INCLUDE=" -I $NUMA_BASE/include/"
+  NUMA_LIB=" $NUMA_BASE/lib/libnuma.a"
+  CFLAGS+=" -DNUMA"
+
+  # location of libunwind
+  LIBUNWIND="$LIBUNWIND_BASE/lib/libunwind.a"
+fi
+
+# location of TBB
+TBB_INCLUDE=" -isystem $TBB_BASE/include/"
+if test -z $PIC_BUILD; then
+  TBB_LIBS="$TBB_BASE/lib/libtbb.a"
+else
+  TBB_LIBS="$TBB_BASE/lib/libtbb_pic.a"
+fi
+CFLAGS+=" -DTBB"
+
+# use Intel SSE support for checksum calculations
+export USE_SSE=1
+export PORTABLE=1
+
+BINUTILS="$BINUTILS_BASE/bin"
+AR="$BINUTILS/ar"
+
+DEPS_INCLUDE="$SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP_INCLUDE $LZ4_INCLUDE $ZSTD_INCLUDE $GFLAGS_INCLUDE $NUMA_INCLUDE $TBB_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/tools/clang/tools/scan-build/bin/scan-build"
+
+if [ -z "$USE_CLANG" ]; then
+  # gcc
+  CC="$GCC_BASE/bin/gcc"
+  CXX="$GCC_BASE/bin/g++"
+
+  CFLAGS+=" -B$BINUTILS/gold"
+  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++"
+
+  KERNEL_HEADERS_INCLUDE="$KERNEL_HEADERS_BASE/include"
+
+  CFLAGS+=" -B$BINUTILS/gold -nostdinc -nostdlib"
+  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/7.x "
+  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/7.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 -DROCKSDB_SUPPORT_THREAD_LOCAL -DHAVE_SSE42"
+CXXFLAGS+=" $CFLAGS"
+
+EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB $TBB_LIBS"
+EXEC_LDFLAGS+=" -B$BINUTILS/gold"
+EXEC_LDFLAGS+=" -Wl,--dynamic-linker,/usr/local/fbcode/platform007/lib/ld.so"
+EXEC_LDFLAGS+=" $LIBUNWIND"
+EXEC_LDFLAGS+=" -Wl,-rpath=/usr/local/fbcode/platform007/lib"
+# required by libtbb
+EXEC_LDFLAGS+=" -ldl"
+
+PLATFORM_LDFLAGS="$LIBGCC_LIBS $GLIBC_LIBS $STDLIBS -lgcc -lstdc++"
+
+EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $TBB_LIBS"
+
+VALGRIND_VER="$VALGRIND_BASE/bin/"
+
+# lua not supported because it's on track for deprecation, I think
+LUA_PATH=
+LUA_LIB=
+
+export CC CXX AR CFLAGS CXXFLAGS EXEC_LDFLAGS EXEC_LDFLAGS_SHARED VALGRIND_VER JEMALLOC_LIB JEMALLOC_INCLUDE CLANG_ANALYZER CLANG_SCAN_BUILD LUA_PATH LUA_LIB
@@ -1,3 +1,4 @@
+# shellcheck disable=SC1113
 #/usr/bin/env bash

 set -e
@@ -28,12 +29,14 @@ function package() {
    if dpkg --get-selections | grep --quiet $1; then
      log "$1 is already installed. skipping."
    else
+      # shellcheck disable=SC2068
      apt-get install $@ -y
    fi
  elif [[ $OS = "centos" ]]; then
    if rpm -qa | grep --quiet $1; then
      log "$1 is already installed. skipping."
    else
+      # shellcheck disable=SC2068
      yum install $@ -y
    fi
  fi
@@ -52,6 +55,7 @@ function gem_install() {
  if gem list | grep --quiet $1; then
    log "$1 is already installed. skipping."
  else
+    # shellcheck disable=SC2068
    gem install $@
  fi
 }
@@ -125,4 +129,5 @@ function main() {
    include $LIB_DIR
 }

+# shellcheck disable=SC2068
 main $@
@@ -85,8 +85,9 @@ NON_SHM="TMPD=/tmp/rocksdb_test_tmp"
 GCC_481="ROCKSDB_FBCODE_BUILD_WITH_481=1"
 ASAN="COMPILE_WITH_ASAN=1"
 CLANG="USE_CLANG=1"
-LITE="OPT=\"-DROCKSDB_LITE -g\""
-TSAN="COMPILE_WITH_TSAN=1"
+# in gcc-5 there are known problems with TSAN like https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71090.
+# using platform007 gives us gcc-8 or higher which has that bug fixed.
+TSAN="ROCKSDB_FBCODE_BUILD_WITH_PLATFORM007=1 COMPILE_WITH_TSAN=1"
 UBSAN="COMPILE_WITH_UBSAN=1"
 DISABLE_JEMALLOC="DISABLE_JEMALLOC=1"
 HTTP_PROXY="https_proxy=http://fwdproxy.29.prn1:8080 http_proxy=http://fwdproxy.29.prn1:8080 ftp_proxy=http://fwdproxy.29.prn1:8080"
@@ -352,6 +353,22 @@ LITE_BUILD_COMMANDS="[
    }
 ]"

+#
+# Report RocksDB lite binary size to scuba
+REPORT_LITE_BINARY_SIZE_COMMANDS="[
+    {
+        'name':'Rocksdb Lite Binary Size',
+        'oncall':'$ONCALL',
+        'steps': [
+            $CLEANUP_ENV,
+            {
+                'name':'Report RocksDB Lite binary size to scuba',
+                'shell':'tools/report_lite_binary_size.sh',
+                'user':'root',
+            },
+        ],
+]"
+
 #
 # RocksDB stress/crash test
 #
@@ -728,6 +745,9 @@ case $1 in
  lite)
    echo $LITE_BUILD_COMMANDS
    ;;
+  report_lite_binary_size)
+    echo $REPORT_LITE_BINARY_SIZE_COMMANDS
+    ;;
  stress_crash)
    echo $STRESS_CRASH_TEST_COMMANDS
    ;;
@@ -0,0 +1,43 @@
+#!/bin/bash
+set -e
+
+ROCKSDB_VERSION="5.10.3"
+ZSTD_VERSION="1.1.3"
+
+echo "This script configures CentOS with everything needed to build and run RocksDB"
+
+yum update -y && yum install epel-release -y
+
+yum install -y \
+  wget \
+  gcc-c++ \
+  snappy snappy-devel \
+  zlib zlib-devel \
+  bzip2 bzip2-devel \
+  lz4-devel \
+  libasan \
+  gflags
+
+mkdir -pv /usr/local/rocksdb-${ROCKSDB_VERSION}
+ln -sfT /usr/local/rocksdb-${ROCKSDB_VERSION} /usr/local/rocksdb
+
+wget -qO /tmp/zstd-${ZSTD_VERSION}.tar.gz https://github.com/facebook/zstd/archive/v${ZSTD_VERSION}.tar.gz
+wget -qO /tmp/rocksdb-${ROCKSDB_VERSION}.tar.gz https://github.com/facebook/rocksdb/archive/v${ROCKSDB_VERSION}.tar.gz
+
+cd /tmp
+
+tar xzvf zstd-${ZSTD_VERSION}.tar.gz
+tar xzvf rocksdb-${ROCKSDB_VERSION}.tar.gz -C /usr/local/
+
+echo "Installing ZSTD..."
+pushd zstd-${ZSTD_VERSION}
+make && make install
+popd
+
+echo "Compiling RocksDB..."
+cd /usr/local/rocksdb
+chown -R vagrant:vagrant /usr/local/rocksdb/
+sudo -u vagrant make static_lib
+cd examples/
+sudo -u vagrant make all
+sudo -u vagrant ./c_simple_example
@@ -53,6 +53,45 @@ function get_lib_base()
  log_variable $__res_var
 }

+###########################################################
+#                platform007 dependencies                 #
+###########################################################
+
+OUTPUT="$BASEDIR/dependencies_platform007.sh"
+
+rm -f "$OUTPUT"
+touch "$OUTPUT"
+
+echo "Writing dependencies to $OUTPUT"
+
+# Compilers locations
+GCC_BASE=`readlink -f $TP2_LATEST/gcc/7.x/centos7-native/*/`
+CLANG_BASE=`readlink -f $TP2_LATEST/llvm-fb/stable/centos7-native/*/`
+
+log_variable GCC_BASE
+log_variable CLANG_BASE
+
+# Libraries locations
+get_lib_base libgcc     7.x     platform007
+get_lib_base glibc      2.26    platform007
+get_lib_base snappy     LATEST  platform007
+get_lib_base zlib       LATEST  platform007
+get_lib_base bzip2      LATEST  platform007
+get_lib_base lz4        LATEST  platform007
+get_lib_base zstd       LATEST  platform007
+get_lib_base gflags     LATEST  platform007
+get_lib_base jemalloc   LATEST  platform007
+get_lib_base numa       LATEST  platform007
+get_lib_base libunwind  LATEST  platform007
+get_lib_base tbb        LATEST  platform007
+
+get_lib_base kernel-headers fb platform007
+get_lib_base binutils   LATEST centos7-native
+get_lib_base valgrind   LATEST platform007
+get_lib_base lua        5.3.4  platform007
+
+git diff $OUTPUT
+
 ###########################################################
 #                   5.x dependencies                      #
 ###########################################################
@@ -57,7 +57,7 @@ struct LRUHandle {
  // Include the following flags:
  //   in_cache:    whether this entry is referenced by the hash table.
  //   is_high_pri: whether this entry is high priority entry.
-  //   in_high_pro_pool: whether this entry is in high-pri pool.
+  //   in_high_pri_pool: whether this entry is in high-pri pool.
  char flags;

  uint32_t hash;     // Hash of key(); used for fast sharding and comparisons
@@ -72,7 +72,7 @@ def display_file_coverage(per_file_coverage, total_coverage):
    header_template = \
        "%" + str(max_file_name_length) + "s\t%s\t%s"
    separator = "-" * (max_file_name_length + 10 + 20)
-    print header_template % ("Filename", "Coverage", "Lines")
+    print header_template % ("Filename", "Coverage", "Lines")  # noqa: E999 T25377293 Grandfathered in
    print separator

    # -- Print body
@@ -131,7 +131,8 @@ Status BuildTable(
    MergeHelper merge(env, internal_comparator.user_comparator(),
                      ioptions.merge_operator, nullptr, ioptions.info_log,
                      true /* internal key corruption is not ok */,
-                      snapshots.empty() ? 0 : snapshots.back());
+                      snapshots.empty() ? 0 : snapshots.back(),
+                      snapshot_checker);

    CompactionIterator c_iter(
        iter, internal_comparator.user_comparator(), &merge, kMaxSequenceNumber,
@@ -1388,23 +1388,24 @@ void rocksdb_writebatch_put_log_data(
  b->rep.PutLogData(Slice(blob, len));
 }

+class H : public WriteBatch::Handler {
+ public:
+  void* state_;
+  void (*put_)(void*, const char* k, size_t klen, const char* v, size_t vlen);
+  void (*deleted_)(void*, const char* k, size_t klen);
+  virtual void Put(const Slice& key, const Slice& value) override {
+    (*put_)(state_, key.data(), key.size(), value.data(), value.size());
+  }
+  virtual void Delete(const Slice& key) override {
+    (*deleted_)(state_, key.data(), key.size());
+  }
+};
+
 void rocksdb_writebatch_iterate(
    rocksdb_writebatch_t* b,
    void* state,
    void (*put)(void*, const char* k, size_t klen, const char* v, size_t vlen),
    void (*deleted)(void*, const char* k, size_t klen)) {
-  class H : public WriteBatch::Handler {
-   public:
-    void* state_;
-    void (*put_)(void*, const char* k, size_t klen, const char* v, size_t vlen);
-    void (*deleted_)(void*, const char* k, size_t klen);
-    virtual void Put(const Slice& key, const Slice& value) override {
-      (*put_)(state_, key.data(), key.size(), value.data(), value.size());
-    }
-    virtual void Delete(const Slice& key) override {
-      (*deleted_)(state_, key.data(), key.size());
-    }
-  };
  H handler;
  handler.state_ = state;
  handler.put_ = put;
@@ -1649,18 +1650,6 @@ void rocksdb_writebatch_wi_iterate(
    void* state,
    void (*put)(void*, const char* k, size_t klen, const char* v, size_t vlen),
    void (*deleted)(void*, const char* k, size_t klen)) {
-  class H : public WriteBatch::Handler {
-   public:
-    void* state_;
-    void (*put_)(void*, const char* k, size_t klen, const char* v, size_t vlen);
-    void (*deleted_)(void*, const char* k, size_t klen);
-    virtual void Put(const Slice& key, const Slice& value) override {
-      (*put_)(state_, key.data(), key.size(), value.data(), value.size());
-    }
-    virtual void Delete(const Slice& key) override {
-      (*deleted_)(state_, key.data(), key.size());
-    }
-  };
  H handler;
  handler.state_ = state;
  handler.put_ = put;
@@ -3109,20 +3098,21 @@ void rocksdb_slicetransform_destroy(rocksdb_slicetransform_t* st) {
  delete st;
 }

+struct Wrapper : public rocksdb_slicetransform_t {
+  const SliceTransform* rep_;
+  ~Wrapper() { delete rep_; }
+  const char* Name() const override { return rep_->Name(); }
+  Slice Transform(const Slice& src) const override {
+    return rep_->Transform(src);
+  }
+  bool InDomain(const Slice& src) const override {
+    return rep_->InDomain(src);
+  }
+  bool InRange(const Slice& src) const override { return rep_->InRange(src); }
+  static void DoNothing(void*) { }
+};
+
 rocksdb_slicetransform_t* rocksdb_slicetransform_create_fixed_prefix(size_t prefixLen) {
-  struct Wrapper : public rocksdb_slicetransform_t {
-    const SliceTransform* rep_;
-    ~Wrapper() { delete rep_; }
-    const char* Name() const override { return rep_->Name(); }
-    Slice Transform(const Slice& src) const override {
-      return rep_->Transform(src);
-    }
-    bool InDomain(const Slice& src) const override {
-      return rep_->InDomain(src);
-    }
-    bool InRange(const Slice& src) const override { return rep_->InRange(src); }
-    static void DoNothing(void*) { }
-  };
  Wrapper* wrapper = new Wrapper;
  wrapper->rep_ = rocksdb::NewFixedPrefixTransform(prefixLen);
  wrapper->state_ = nullptr;
@@ -3131,19 +3121,6 @@ rocksdb_slicetransform_t* rocksdb_slicetransform_create_fixed_prefix(size_t pref
 }

 rocksdb_slicetransform_t* rocksdb_slicetransform_create_noop() {
-  struct Wrapper : public rocksdb_slicetransform_t {
-    const SliceTransform* rep_;
-    ~Wrapper() { delete rep_; }
-    const char* Name() const override { return rep_->Name(); }
-    Slice Transform(const Slice& src) const override {
-      return rep_->Transform(src);
-    }
-    bool InDomain(const Slice& src) const override {
-      return rep_->InDomain(src);
-    }
-    bool InRange(const Slice& src) const override { return rep_->InRange(src); }
-    static void DoNothing(void*) { }
-  };
  Wrapper* wrapper = new Wrapper;
  wrapper->rep_ = rocksdb::NewNoopTransform();
  wrapper->state_ = nullptr;
@@ -31,6 +31,7 @@
 #include "monitoring/thread_status_util.h"
 #include "options/options_helper.h"
 #include "table/block_based_table_factory.h"
+#include "table/merging_iterator.h"
 #include "util/autovector.h"
 #include "util/compression.h"

@@ -344,12 +345,13 @@ void SuperVersionUnrefHandle(void* ptr) {
  // When latter happens, we are in ~ColumnFamilyData(), no get should happen as
  // well.
  SuperVersion* sv = static_cast<SuperVersion*>(ptr);
-  if (sv->Unref()) {
-    sv->db_mutex->Lock();
-    sv->Cleanup();
-    sv->db_mutex->Unlock();
-    delete sv;
-  }
+  bool was_last_ref __attribute__((__unused__));
+  was_last_ref = sv->Unref();
+  // Thread-local SuperVersions can't outlive ColumnFamilyData::super_version_.
+  // This is important because we can't do SuperVersion cleanup here.
+  // That would require locking DB mutex, which would deadlock because
+  // SuperVersionUnrefHandle is called with locked ThreadLocalPtr mutex.
+  assert(!was_last_ref);
 }
 }  // anonymous namespace

@@ -381,11 +383,13 @@ ColumnFamilyData::ColumnFamilyData(
      next_(nullptr),
      prev_(nullptr),
      log_number_(0),
+      flush_reason_(FlushReason::kUnknown),
      column_family_set_(column_family_set),
      pending_flush_(false),
      pending_compaction_(false),
      prev_compaction_needed_bytes_(0),
-      allow_2pc_(db_options.allow_2pc) {
+      allow_2pc_(db_options.allow_2pc),
+      last_memtable_id_(0) {
  Ref();

  // Convert user defined table properties collector factories to internal ones.
@@ -468,7 +472,7 @@ ColumnFamilyData::~ColumnFamilyData() {
    local_sv_.reset();
    super_version_->db_mutex->Lock();

-    bool is_last_reference __attribute__((unused));
+    bool is_last_reference __attribute__((__unused__));
    is_last_reference = super_version_->Unref();
    assert(is_last_reference);
    super_version_->Cleanup();
@@ -479,7 +483,7 @@ ColumnFamilyData::~ColumnFamilyData() {
  if (dummy_versions_ != nullptr) {
    // List must be empty
    assert(dummy_versions_->TEST_Next() == dummy_versions_);
-    bool deleted __attribute__((unused));
+    bool deleted __attribute__((__unused__));
    deleted = dummy_versions_->Unref();
    assert(deleted);
  }
@@ -630,6 +634,41 @@ int GetL0ThresholdSpeedupCompaction(int level0_file_num_compaction_trigger,
 }
 }  // namespace

+std::pair<WriteStallCondition, ColumnFamilyData::WriteStallCause>
+ColumnFamilyData::GetWriteStallConditionAndCause(
+    int num_unflushed_memtables, int num_l0_files,
+    uint64_t num_compaction_needed_bytes,
+    const MutableCFOptions& mutable_cf_options) {
+  if (num_unflushed_memtables >= mutable_cf_options.max_write_buffer_number) {
+    return {WriteStallCondition::kStopped, WriteStallCause::kMemtableLimit};
+  } else if (!mutable_cf_options.disable_auto_compactions &&
+             num_l0_files >= mutable_cf_options.level0_stop_writes_trigger) {
+    return {WriteStallCondition::kStopped, WriteStallCause::kL0FileCountLimit};
+  } else if (!mutable_cf_options.disable_auto_compactions &&
+             mutable_cf_options.hard_pending_compaction_bytes_limit > 0 &&
+             num_compaction_needed_bytes >=
+                 mutable_cf_options.hard_pending_compaction_bytes_limit) {
+    return {WriteStallCondition::kStopped,
+            WriteStallCause::kPendingCompactionBytes};
+  } else if (mutable_cf_options.max_write_buffer_number > 3 &&
+             num_unflushed_memtables >=
+                 mutable_cf_options.max_write_buffer_number - 1) {
+    return {WriteStallCondition::kDelayed, WriteStallCause::kMemtableLimit};
+  } else if (!mutable_cf_options.disable_auto_compactions &&
+             mutable_cf_options.level0_slowdown_writes_trigger >= 0 &&
+             num_l0_files >=
+                 mutable_cf_options.level0_slowdown_writes_trigger) {
+    return {WriteStallCondition::kDelayed, WriteStallCause::kL0FileCountLimit};
+  } else if (!mutable_cf_options.disable_auto_compactions &&
+             mutable_cf_options.soft_pending_compaction_bytes_limit > 0 &&
+             num_compaction_needed_bytes >=
+                 mutable_cf_options.soft_pending_compaction_bytes_limit) {
+    return {WriteStallCondition::kDelayed,
+            WriteStallCause::kPendingCompactionBytes};
+  }
+  return {WriteStallCondition::kNormal, WriteStallCause::kNone};
+}
+
 WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
      const MutableCFOptions& mutable_cf_options) {
  auto write_stall_condition = WriteStallCondition::kNormal;
@@ -639,25 +678,29 @@ WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
    uint64_t compaction_needed_bytes =
        vstorage->estimated_compaction_needed_bytes();

+    auto write_stall_condition_and_cause = GetWriteStallConditionAndCause(
+        imm()->NumNotFlushed(), vstorage->l0_delay_trigger_count(),
+        vstorage->estimated_compaction_needed_bytes(), mutable_cf_options);
+    write_stall_condition = write_stall_condition_and_cause.first;
+    auto write_stall_cause = write_stall_condition_and_cause.second;
+
    bool was_stopped = write_controller->IsStopped();
    bool needed_delay = write_controller->NeedsDelay();

-    if (imm()->NumNotFlushed() >= mutable_cf_options.max_write_buffer_number) {
+    if (write_stall_condition == WriteStallCondition::kStopped &&
+        write_stall_cause == WriteStallCause::kMemtableLimit) {
      write_controller_token_ = write_controller->GetStopToken();
      internal_stats_->AddCFStats(InternalStats::MEMTABLE_LIMIT_STOPS, 1);
-      write_stall_condition = WriteStallCondition::kStopped;
      ROCKS_LOG_WARN(
          ioptions_.info_log,
          "[%s] Stopping writes because we have %d immutable memtables "
          "(waiting for flush), max_write_buffer_number is set to %d",
          name_.c_str(), imm()->NumNotFlushed(),
          mutable_cf_options.max_write_buffer_number);
-    } else if (!mutable_cf_options.disable_auto_compactions &&
-               vstorage->l0_delay_trigger_count() >=
-                   mutable_cf_options.level0_stop_writes_trigger) {
+    } else if (write_stall_condition == WriteStallCondition::kStopped &&
+               write_stall_cause == WriteStallCause::kL0FileCountLimit) {
      write_controller_token_ = write_controller->GetStopToken();
      internal_stats_->AddCFStats(InternalStats::L0_FILE_COUNT_LIMIT_STOPS, 1);
-      write_stall_condition = WriteStallCondition::kStopped;
      if (compaction_picker_->IsLevel0CompactionInProgress()) {
        internal_stats_->AddCFStats(
            InternalStats::LOCKED_L0_FILE_COUNT_LIMIT_STOPS, 1);
@@ -665,28 +708,23 @@ WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
      ROCKS_LOG_WARN(ioptions_.info_log,
                     "[%s] Stopping writes because we have %d level-0 files",
                     name_.c_str(), vstorage->l0_delay_trigger_count());
-    } else if (!mutable_cf_options.disable_auto_compactions &&
-               mutable_cf_options.hard_pending_compaction_bytes_limit > 0 &&
-               compaction_needed_bytes >=
-                   mutable_cf_options.hard_pending_compaction_bytes_limit) {
+    } else if (write_stall_condition == WriteStallCondition::kStopped &&
+               write_stall_cause == WriteStallCause::kPendingCompactionBytes) {
      write_controller_token_ = write_controller->GetStopToken();
      internal_stats_->AddCFStats(
          InternalStats::PENDING_COMPACTION_BYTES_LIMIT_STOPS, 1);
-      write_stall_condition = WriteStallCondition::kStopped;
      ROCKS_LOG_WARN(
          ioptions_.info_log,
          "[%s] Stopping writes because of estimated pending compaction "
          "bytes %" PRIu64,
          name_.c_str(), compaction_needed_bytes);
-    } else if (mutable_cf_options.max_write_buffer_number > 3 &&
-               imm()->NumNotFlushed() >=
-                   mutable_cf_options.max_write_buffer_number - 1) {
+    } else if (write_stall_condition == WriteStallCondition::kDelayed &&
+               write_stall_cause == WriteStallCause::kMemtableLimit) {
      write_controller_token_ =
          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);
-      write_stall_condition = WriteStallCondition::kDelayed;
      ROCKS_LOG_WARN(
          ioptions_.info_log,
          "[%s] Stalling writes because we have %d immutable memtables "
@@ -695,10 +733,8 @@ WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
          name_.c_str(), imm()->NumNotFlushed(),
          mutable_cf_options.max_write_buffer_number,
          write_controller->delayed_write_rate());
-    } else if (!mutable_cf_options.disable_auto_compactions &&
-               mutable_cf_options.level0_slowdown_writes_trigger >= 0 &&
-               vstorage->l0_delay_trigger_count() >=
-                   mutable_cf_options.level0_slowdown_writes_trigger) {
+    } else if (write_stall_condition == WriteStallCondition::kDelayed &&
+               write_stall_cause == WriteStallCause::kL0FileCountLimit) {
      // L0 is the last two files from stopping.
      bool near_stop = vstorage->l0_delay_trigger_count() >=
                       mutable_cf_options.level0_stop_writes_trigger - 2;
@@ -708,7 +744,6 @@ WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
                     mutable_cf_options.disable_auto_compactions);
      internal_stats_->AddCFStats(InternalStats::L0_FILE_COUNT_LIMIT_SLOWDOWNS,
                                  1);
-      write_stall_condition = WriteStallCondition::kDelayed;
      if (compaction_picker_->IsLevel0CompactionInProgress()) {
        internal_stats_->AddCFStats(
            InternalStats::LOCKED_L0_FILE_COUNT_LIMIT_SLOWDOWNS, 1);
@@ -718,10 +753,8 @@ WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
                     "rate %" PRIu64,
                     name_.c_str(), vstorage->l0_delay_trigger_count(),
                     write_controller->delayed_write_rate());
-    } else if (!mutable_cf_options.disable_auto_compactions &&
-               mutable_cf_options.soft_pending_compaction_bytes_limit > 0 &&
-               vstorage->estimated_compaction_needed_bytes() >=
-                   mutable_cf_options.soft_pending_compaction_bytes_limit) {
+    } else if (write_stall_condition == WriteStallCondition::kDelayed &&
+               write_stall_cause == WriteStallCause::kPendingCompactionBytes) {
      // If the distance to hard limit is less than 1/4 of the gap between soft
      // and
      // hard bytes limit, we think it is near stop and speed up the slowdown.
@@ -739,7 +772,6 @@ WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
                     mutable_cf_options.disable_auto_compactions);
      internal_stats_->AddCFStats(
          InternalStats::PENDING_COMPACTION_BYTES_LIMIT_SLOWDOWNS, 1);
-      write_stall_condition = WriteStallCondition::kDelayed;
      ROCKS_LOG_WARN(
          ioptions_.info_log,
          "[%s] Stalling writes because of estimated pending compaction "
@@ -747,6 +779,7 @@ WriteStallCondition ColumnFamilyData::RecalculateWriteStallConditions(
          name_.c_str(), vstorage->estimated_compaction_needed_bytes(),
          write_controller->delayed_write_rate());
    } else {
+      assert(write_stall_condition == WriteStallCondition::kNormal);
      if (vstorage->l0_delay_trigger_count() >=
          GetL0ThresholdSpeedupCompaction(
              mutable_cf_options.level0_file_num_compaction_trigger,
@@ -849,6 +882,68 @@ bool ColumnFamilyData::RangeOverlapWithCompaction(
      smallest_user_key, largest_user_key, level);
 }

+Status ColumnFamilyData::RangesOverlapWithMemtables(
+    const autovector<Range>& ranges, SuperVersion* super_version,
+    bool* overlap) {
+  assert(overlap != nullptr);
+  *overlap = false;
+  // Create an InternalIterator over all unflushed memtables
+  Arena arena;
+  ReadOptions read_opts;
+  read_opts.total_order_seek = true;
+  MergeIteratorBuilder merge_iter_builder(&internal_comparator_, &arena);
+  merge_iter_builder.AddIterator(
+      super_version->mem->NewIterator(read_opts, &arena));
+  super_version->imm->AddIterators(read_opts, &merge_iter_builder);
+  ScopedArenaIterator memtable_iter(merge_iter_builder.Finish());
+
+  std::vector<InternalIterator*> memtable_range_del_iters;
+  auto* active_range_del_iter =
+      super_version->mem->NewRangeTombstoneIterator(read_opts);
+  if (active_range_del_iter != nullptr) {
+    memtable_range_del_iters.push_back(active_range_del_iter);
+  }
+  super_version->imm->AddRangeTombstoneIterators(read_opts,
+                                                 &memtable_range_del_iters);
+  RangeDelAggregator range_del_agg(internal_comparator_, {} /* snapshots */,
+                                   false /* collapse_deletions */);
+  Status status;
+  {
+    std::unique_ptr<InternalIterator> memtable_range_del_iter(
+        NewMergingIterator(&internal_comparator_,
+                           memtable_range_del_iters.empty()
+                               ? nullptr
+                               : &memtable_range_del_iters[0],
+                           static_cast<int>(memtable_range_del_iters.size())));
+    status = range_del_agg.AddTombstones(std::move(memtable_range_del_iter));
+  }
+  for (size_t i = 0; i < ranges.size() && status.ok() && !*overlap; ++i) {
+    auto* vstorage = super_version->current->storage_info();
+    auto* ucmp = vstorage->InternalComparator()->user_comparator();
+    InternalKey range_start(ranges[i].start, kMaxSequenceNumber,
+                            kValueTypeForSeek);
+    memtable_iter->Seek(range_start.Encode());
+    status = memtable_iter->status();
+    ParsedInternalKey seek_result;
+    if (status.ok()) {
+      if (memtable_iter->Valid() &&
+          !ParseInternalKey(memtable_iter->key(), &seek_result)) {
+        status = Status::Corruption("DB have corrupted keys");
+      }
+    }
+    if (status.ok()) {
+      if (memtable_iter->Valid() &&
+          ucmp->Compare(seek_result.user_key, ranges[i].limit) <= 0) {
+        *overlap = true;
+      } else if (range_del_agg.IsRangeOverlapped(ranges[i].start,
+                                                 ranges[i].limit)) {
+        *overlap = true;
+      }
+    }
+  }
+  return status;
+}
+
 const int ColumnFamilyData::kCompactAllLevels = -1;
 const int ColumnFamilyData::kCompactToBaseLevel = -2;

@@ -867,8 +962,7 @@ Compaction* ColumnFamilyData::CompactRange(

 SuperVersion* ColumnFamilyData::GetReferencedSuperVersion(
    InstrumentedMutex* db_mutex) {
-  SuperVersion* sv = nullptr;
-  sv = GetThreadLocalSuperVersion(db_mutex);
+  SuperVersion* sv = GetThreadLocalSuperVersion(db_mutex);
  sv->Ref();
  if (!ReturnThreadLocalSuperVersion(sv)) {
    // This Unref() corresponds to the Ref() in GetThreadLocalSuperVersion()
@@ -882,7 +976,6 @@ SuperVersion* ColumnFamilyData::GetReferencedSuperVersion(

 SuperVersion* ColumnFamilyData::GetThreadLocalSuperVersion(
    InstrumentedMutex* db_mutex) {
-  SuperVersion* sv = nullptr;
  // The SuperVersion is cached in thread local storage to avoid acquiring
  // mutex when SuperVersion does not change since the last use. When a new
  // SuperVersion is installed, the compaction or flush thread cleans up
@@ -901,7 +994,7 @@ SuperVersion* ColumnFamilyData::GetThreadLocalSuperVersion(
  // should only keep kSVInUse before ReturnThreadLocalSuperVersion call
  // (if no Scrape happens).
  assert(ptr != SuperVersion::kSVInUse);
-  sv = static_cast<SuperVersion*>(ptr);
+  SuperVersion* sv = static_cast<SuperVersion*>(ptr);
  if (sv == SuperVersion::kSVObsolete ||
      sv->version_number != super_version_number_.load()) {
    RecordTick(ioptions_.statistics, NUMBER_SUPERVERSION_ACQUIRES);
@@ -965,6 +1058,12 @@ void ColumnFamilyData::InstallSuperVersion(
      RecalculateWriteStallConditions(mutable_cf_options);

  if (old_superversion != nullptr) {
+    // Reset SuperVersions cached in thread local storage.
+    // This should be done before old_superversion->Unref(). That's to ensure
+    // that local_sv_ never holds the last reference to SuperVersion, since
+    // it has no means to safely do SuperVersion cleanup.
+    ResetThreadLocalSuperVersions();
+
    if (old_superversion->mutable_cf_options.write_buffer_size !=
        mutable_cf_options.write_buffer_size) {
      mem_->UpdateWriteBufferSize(mutable_cf_options.write_buffer_size);
@@ -980,9 +1079,6 @@ void ColumnFamilyData::InstallSuperVersion(
      sv_context->superversions_to_free.push_back(old_superversion);
    }
  }
-
-  // Reset SuperVersions cached in thread local storage
-  ResetThreadLocalSuperVersions();
 }

 void ColumnFamilyData::ResetThreadLocalSuperVersions() {
@@ -994,10 +1090,12 @@ void ColumnFamilyData::ResetThreadLocalSuperVersions() {
      continue;
    }
    auto sv = static_cast<SuperVersion*>(ptr);
-    if (sv->Unref()) {
-      sv->Cleanup();
-      delete sv;
-    }
+    bool was_last_ref __attribute__((__unused__));
+    was_last_ref = sv->Unref();
+    // sv couldn't have been the last reference because
+    // ResetThreadLocalSuperVersions() is called before
+    // unref'ing super_version_.
+    assert(!was_last_ref);
  }
 }

@@ -1059,10 +1157,14 @@ ColumnFamilySet::~ColumnFamilySet() {
  while (column_family_data_.size() > 0) {
    // cfd destructor will delete itself from column_family_data_
    auto cfd = column_family_data_.begin()->second;
-    cfd->Unref();
+    bool last_ref __attribute__((__unused__));
+    last_ref = cfd->Unref();
+    assert(last_ref);
    delete cfd;
  }
-  dummy_cfd_->Unref();
+  bool dummy_last_ref __attribute__((__unused__));
+  dummy_last_ref = dummy_cfd_->Unref();
+  assert(dummy_last_ref);
  delete dummy_cfd_;
 }

@@ -30,6 +30,7 @@ namespace rocksdb {

 class Version;
 class VersionSet;
+class VersionStorageInfo;
 class MemTable;
 class MemTableListVersion;
 class CompactionPicker;
@@ -202,6 +203,10 @@ 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 EnvOptions* soptions() const;
  const ImmutableCFOptions* ioptions() const { return &ioptions_; }
@@ -239,7 +244,11 @@ class ColumnFamilyData {
  void SetCurrent(Version* _current);
  uint64_t GetNumLiveVersions() const;  // REQUIRE: DB mutex held
  uint64_t GetTotalSstFilesSize() const;  // REQUIRE: DB mutex held
-  void SetMemtable(MemTable* new_mem) { mem_ = new_mem; }
+  void SetMemtable(MemTable* new_mem) {
+    uint64_t memtable_id = last_memtable_id_.fetch_add(1) + 1;
+    new_mem->SetID(memtable_id);
+    mem_ = new_mem;
+  }

  // calculate the oldest log needed for the durability of this column family
  uint64_t OldestLogToKeep();
@@ -265,6 +274,16 @@ class ColumnFamilyData {
                                  const Slice& largest_user_key,
                                  int level) const;

+  // Check if the passed ranges overlap with any unflushed memtables
+  // (immutable or mutable).
+  //
+  // @param super_version A referenced SuperVersion that will be held for the
+  //    duration of this function.
+  //
+  // Thread-safe
+  Status RangesOverlapWithMemtables(const autovector<Range>& ranges,
+                                    SuperVersion* super_version, bool* overlap);
+
  // A flag to tell a manual compaction is to compact all levels together
  // instead of a specific level.
  static const int kCompactAllLevels;
@@ -327,6 +346,17 @@ class ColumnFamilyData {
  bool pending_flush() { return pending_flush_; }
  bool pending_compaction() { return pending_compaction_; }

+  enum class WriteStallCause {
+    kNone,
+    kMemtableLimit,
+    kL0FileCountLimit,
+    kPendingCompactionBytes,
+  };
+  static std::pair<WriteStallCondition, WriteStallCause>
+  GetWriteStallConditionAndCause(int num_unflushed_memtables, int num_l0_files,
+                                 uint64_t num_compaction_needed_bytes,
+                                 const MutableCFOptions& mutable_cf_options);
+
  // Recalculate some small conditions, which are changed only during
  // compaction, adding new memtable and/or
  // recalculation of compaction score. These values are used in
@@ -400,6 +430,8 @@ 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_;
@@ -419,6 +451,9 @@ class ColumnFamilyData {

  // if the database was opened with 2pc enabled
  bool allow_2pc_;
+
+  // Memtable id to track flush.
+  std::atomic<uint64_t> last_memtable_id_;
 };

 // ColumnFamilySet has interesting thread-safety requirements
@@ -383,6 +383,9 @@ class ColumnFamilyTest : public testing::Test {
  void AssertFilesPerLevel(const std::string& value, int cf) {
 #ifndef ROCKSDB_LITE
    ASSERT_EQ(value, FilesPerLevel(cf));
+#else
+    (void) value;
+    (void) cf;
 #endif
  }

@@ -397,6 +400,8 @@ class ColumnFamilyTest : public testing::Test {
  void AssertCountLiveFiles(int expected_value) {
 #ifndef ROCKSDB_LITE
    ASSERT_EQ(expected_value, CountLiveFiles());
+#else
+    (void) expected_value;
 #endif
  }

@@ -445,6 +450,8 @@ class ColumnFamilyTest : public testing::Test {
  void AssertCountLiveLogFiles(int value) {
 #ifndef ROCKSDB_LITE  // GetSortedWalFiles is not supported
    ASSERT_EQ(value, CountLiveLogFiles());
+#else
+    (void) value;
 #endif  // !ROCKSDB_LITE
  }

@@ -1868,126 +1875,6 @@ TEST_F(ColumnFamilyTest, SameCFManualAutomaticCompactionsLevel) {
  }
 }

-// This test checks for automatic getting a conflict if there is a
-// manual which has not yet been scheduled.
-// The manual compaction waits in NotScheduled
-// We generate more files and then trigger an automatic compaction
-// This will wait because there is an unscheduled manual compaction.
-// Once the conflict is hit, the manual compaction starts and ends
-// Then another automatic will start and end.
-TEST_F(ColumnFamilyTest, SameCFManualAutomaticConflict) {
-  Open();
-  CreateColumnFamilies({"one"});
-  ColumnFamilyOptions default_cf, one;
-  db_options_.max_open_files = 20;  // only 10 files in file cache
-  db_options_.max_background_compactions = 3;
-
-  default_cf.compaction_style = kCompactionStyleLevel;
-  default_cf.num_levels = 3;
-  default_cf.write_buffer_size = 64 << 10;  // 64KB
-  default_cf.target_file_size_base = 30 << 10;
-  default_cf.max_compaction_bytes = default_cf.target_file_size_base * 1100;
-  BlockBasedTableOptions table_options;
-  table_options.no_block_cache = true;
-  default_cf.table_factory.reset(NewBlockBasedTableFactory(table_options));
-
-  one.compaction_style = kCompactionStyleUniversal;
-
-  one.num_levels = 1;
-  // trigger compaction if there are >= 4 files
-  one.level0_file_num_compaction_trigger = 4;
-  one.write_buffer_size = 120000;
-
-  Reopen({default_cf, one});
-  // make sure all background compaction jobs can be scheduled
-  auto stop_token =
-      dbfull()->TEST_write_controler().GetCompactionPressureToken();
-
-  // SETUP column family "one" -- universal style
-  for (int i = 0; i < one.level0_file_num_compaction_trigger - 2; ++i) {
-    PutRandomData(1, 10, 12000, true);
-    PutRandomData(1, 1, 10, true);
-    WaitForFlush(1);
-    AssertFilesPerLevel(ToString(i + 1), 1);
-  }
-  bool cf_1_1 = true;
-  bool cf_1_2 = true;
-  rocksdb::SyncPoint::GetInstance()->LoadDependency(
-      {{"DBImpl::BackgroundCompaction()::Conflict",
-        "ColumnFamilyTest::ManualAutoCon:7"},
-       {"ColumnFamilyTest::ManualAutoCon:9",
-        "ColumnFamilyTest::ManualAutoCon:8"},
-       {"ColumnFamilyTest::ManualAutoCon:2",
-        "ColumnFamilyTest::ManualAutoCon:6"},
-       {"ColumnFamilyTest::ManualAutoCon:4",
-        "ColumnFamilyTest::ManualAutoCon:5"},
-       {"ColumnFamilyTest::ManualAutoCon:1",
-        "ColumnFamilyTest::ManualAutoCon:2"},
-       {"ColumnFamilyTest::ManualAutoCon:1",
-        "ColumnFamilyTest::ManualAutoCon:3"}});
-  rocksdb::SyncPoint::GetInstance()->SetCallBack(
-      "DBImpl::BackgroundCompaction:NonTrivial:AfterRun", [&](void* arg) {
-        if (cf_1_1) {
-          TEST_SYNC_POINT("ColumnFamilyTest::ManualAutoCon:4");
-          cf_1_1 = false;
-          TEST_SYNC_POINT("ColumnFamilyTest::ManualAutoCon:3");
-        } else if (cf_1_2) {
-          cf_1_2 = false;
-          TEST_SYNC_POINT("ColumnFamilyTest::ManualAutoCon:2");
-        }
-      });
-  rocksdb::SyncPoint::GetInstance()->SetCallBack(
-      "DBImpl::RunManualCompaction:NotScheduled", [&](void* arg) {
-        InstrumentedMutex* mutex = static_cast<InstrumentedMutex*>(arg);
-        mutex->Unlock();
-        TEST_SYNC_POINT("ColumnFamilyTest::ManualAutoCon:9");
-        TEST_SYNC_POINT("ColumnFamilyTest::ManualAutoCon:7");
-        mutex->Lock();
-      });
-
-  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
-  rocksdb::port::Thread threads([&] {
-    CompactRangeOptions compact_options;
-    compact_options.exclusive_manual_compaction = false;
-    ASSERT_OK(
-        db_->CompactRange(compact_options, handles_[1], nullptr, nullptr));
-    TEST_SYNC_POINT("ColumnFamilyTest::ManualAutoCon:6");
-  });
-
-  TEST_SYNC_POINT("ColumnFamilyTest::ManualAutoCon:8");
-  WaitForFlush(1);
-
-  // Add more L0 files and force automatic compaction
-  for (int i = 0; i < one.level0_file_num_compaction_trigger; ++i) {
-    PutRandomData(1, 10, 12000, true);
-    PutRandomData(1, 1, 10, true);
-    WaitForFlush(1);
-    AssertFilesPerLevel(ToString(one.level0_file_num_compaction_trigger + i),
-                        1);
-  }
-
-  TEST_SYNC_POINT("ColumnFamilyTest::ManualAutoCon:5");
-  // Add more L0 files and force automatic compaction
-  for (int i = 0; i < one.level0_file_num_compaction_trigger; ++i) {
-    PutRandomData(1, 10, 12000, true);
-    PutRandomData(1, 1, 10, true);
-    WaitForFlush(1);
-  }
-  TEST_SYNC_POINT("ColumnFamilyTest::ManualAutoCon:1");
-
-  threads.join();
-  WaitForCompaction();
-  // VERIFY compaction "one"
-  ASSERT_LE(NumTableFilesAtLevel(0, 1), 3);
-
-  // Compare against saved keys
-  std::set<std::string>::iterator key_iter = keys_.begin();
-  while (key_iter != keys_.end()) {
-    ASSERT_NE("NOT_FOUND", Get(1, *key_iter));
-    key_iter++;
-  }
-}
-
 // In this test, we generate enough files to trigger automatic compactions.
 // The automatic compaction waits in NonTrivial:AfterRun
 // We generate more files and then trigger an automatic compaction
@@ -319,7 +319,7 @@ int main(int argc, char** argv) {
 #else
 #include <stdio.h>

-int main(int argc, char** argv) {
+int main(int /*argc*/, char** /*argv*/) {
  fprintf(stderr,
          "SKIPPED as DBImpl::CompactFiles is not supported in ROCKSDB_LITE\n");
  return 0;
@@ -405,20 +405,23 @@ void Compaction::Summary(char* output, int len) {
 uint64_t Compaction::OutputFilePreallocationSize() const {
  uint64_t preallocation_size = 0;

-  if (max_output_file_size_ != port::kMaxUint64 &&
-      (cfd_->ioptions()->compaction_style == kCompactionStyleLevel ||
-       output_level() > 0)) {
-    preallocation_size = max_output_file_size_;
-  } else {
-    for (const auto& level_files : inputs_) {
-      for (const auto& file : level_files.files) {
-        preallocation_size += file->fd.GetFileSize();
-      }
+  for (const auto& level_files : inputs_) {
+    for (const auto& file : level_files.files) {
+      preallocation_size += file->fd.GetFileSize();
    }
  }
+
+  if (max_output_file_size_ != port::kMaxUint64 &&
+      (immutable_cf_options_.compaction_style == kCompactionStyleLevel ||
+       output_level() > 0)) {
+    preallocation_size = std::min(max_output_file_size_, preallocation_size);
+  }
+
  // Over-estimate slightly so we don't end up just barely crossing
  // the threshold
-  return preallocation_size + (preallocation_size / 10);
+  // No point to prellocate more than 1GB.
+  return std::min(uint64_t{1073741824},
+                  preallocation_size + (preallocation_size / 10));
 }

 std::unique_ptr<CompactionFilter> Compaction::CreateCompactionFilter() const {
@@ -537,6 +537,11 @@ void CompactionIterator::NextFromInput() {
      //
      // Note:  Dropping this Delete will not affect TransactionDB
      // write-conflict checking since it is earlier than any snapshot.
+      //
+      // It seems that we can also drop deletion later than earliest snapshot
+      // given that:
+      // (1) The deletion is earlier than earliest_write_conflict_snapshot, and
+      // (2) No value exist earlier than the deletion.
      ++iter_stats_.num_record_drop_obsolete;
      if (!bottommost_level_) {
        ++iter_stats_.num_optimized_del_drop_obsolete;
@@ -554,10 +559,6 @@ void CompactionIterator::NextFromInput() {
      // have hit (A)
      // We encapsulate the merge related state machine in a different
      // object to minimize change to the existing flow.
-      // In case snapshot_checker is present, we can probably merge further
-      // beyond prev_snapshot, since there could be more keys with sequence
-      // smaller than prev_snapshot, but reported by snapshot_checker as not
-      // visible by prev_snapshot. But it will make the logic more complicated.
      Status s = merge_helper_->MergeUntil(input_, range_del_agg_,
                                           prev_snapshot, bottommost_level_);
      merge_out_iter_.SeekToFirst();
@@ -621,9 +622,12 @@ void CompactionIterator::PrepareOutput() {
  // and the earliest snapshot is larger than this seqno
  // and the userkey differs from the last userkey in compaction
  // then we can squash the seqno to zero.
-
+  //
  // This is safe for TransactionDB write-conflict checking since transactions
  // only care about sequence number larger than any active snapshots.
+  //
+  // Can we do the same for levels above bottom level as long as
+  // KeyNotExistsBeyondOutputLevel() return true?
  if ((compaction_ != nullptr &&
      !compaction_->allow_ingest_behind()) &&
      ikeyNotNeededForIncrementalSnapshot() &&
@@ -9,8 +9,10 @@
 #include <vector>

 #include "port/port.h"
+#include "util/string_util.h"
 #include "util/testharness.h"
 #include "util/testutil.h"
+#include "utilities/merge_operators.h"

 namespace rocksdb {

@@ -39,9 +41,9 @@ class NoMergingMergeOp : public MergeOperator {
 // Always returns Decition::kRemove.
 class StallingFilter : public CompactionFilter {
 public:
-  virtual Decision FilterV2(int level, const Slice& key, ValueType t,
-                            const Slice& existing_value, std::string* new_value,
-                            std::string* skip_until) const override {
+  Decision FilterV2(int /*level*/, const Slice& key, ValueType /*type*/,
+                    const Slice& /*existing_value*/, std::string* /*new_value*/,
+                    std::string* /*skip_until*/) const override {
    int k = std::atoi(key.ToString().c_str());
    last_seen.store(k);
    while (k >= stall_at.load()) {
@@ -72,6 +74,18 @@ class StallingFilter : public CompactionFilter {
  mutable std::atomic<int> last_seen{0};
 };

+// Compaction filter that filter out all keys.
+class FilterAllKeysCompactionFilter : public CompactionFilter {
+ public:
+  Decision FilterV2(int /*level*/, const Slice& /*key*/, ValueType /*type*/,
+                    const Slice& /*existing_value*/, std::string* /*new_value*/,
+                    std::string* /*skip_until*/) const override {
+    return Decision::kRemove;
+  }
+
+  const char* Name() const override { return "AllKeysCompactionFilter"; }
+};
+
 class LoggingForwardVectorIterator : public InternalIterator {
 public:
  struct Action {
@@ -144,76 +158,146 @@ class FakeCompaction : public CompactionIterator::CompactionProxy {
 public:
  FakeCompaction() = default;

-  virtual int level(size_t compaction_input_level) const { return 0; }
+  virtual int level(size_t compaction_input_level) const override { return 0; }
  virtual bool KeyNotExistsBeyondOutputLevel(
-      const Slice& user_key, std::vector<size_t>* level_ptrs) const {
-    return key_not_exists_beyond_output_level;
+      const Slice& user_key, std::vector<size_t>* level_ptrs) const override {
+    return is_bottommost_level || key_not_exists_beyond_output_level;
  }
-  virtual bool bottommost_level() const { return false; }
-  virtual int number_levels() const { return 1; }
-  virtual Slice GetLargestUserKey() const {
+  virtual bool bottommost_level() const override { return is_bottommost_level; }
+  virtual int number_levels() const override { return 1; }
+  virtual Slice GetLargestUserKey() const override {
    return "\xff\xff\xff\xff\xff\xff\xff\xff\xff";
  }
-  virtual bool allow_ingest_behind() const { return false; }
+  virtual bool allow_ingest_behind() const override { return false; }

-  virtual bool preserve_deletes() const {return false; }
+  virtual bool preserve_deletes() const override { return false; }

  bool key_not_exists_beyond_output_level = false;
+
+  bool is_bottommost_level = false;
 };

-class CompactionIteratorTest : public testing::Test {
+// A simplifed snapshot checker which assumes each snapshot has a global
+// last visible sequence.
+class TestSnapshotChecker : public SnapshotChecker {
+ public:
+  explicit TestSnapshotChecker(
+      SequenceNumber last_committed_sequence,
+      const std::unordered_map<SequenceNumber, SequenceNumber>& snapshots = {})
+      : last_committed_sequence_(last_committed_sequence),
+        snapshots_(snapshots) {}
+
+  bool IsInSnapshot(SequenceNumber seq,
+                    SequenceNumber snapshot_seq) const override {
+    if (snapshot_seq == kMaxSequenceNumber) {
+      return seq <= last_committed_sequence_;
+    }
+    assert(snapshots_.count(snapshot_seq) > 0);
+    return seq <= snapshots_.at(snapshot_seq);
+  }
+
+ private:
+  SequenceNumber last_committed_sequence_;
+  // A map of valid snapshot to last visible sequence to the snapshot.
+  std::unordered_map<SequenceNumber, SequenceNumber> snapshots_;
+};
+
+// Test param:
+//   bool: whether to pass snapshot_checker to compaction iterator.
+class CompactionIteratorTest : public testing::TestWithParam<bool> {
 public:
  CompactionIteratorTest()
      : cmp_(BytewiseComparator()), icmp_(cmp_), snapshots_({}) {}

-  void InitIterators(const std::vector<std::string>& ks,
-                     const std::vector<std::string>& vs,
-                     const std::vector<std::string>& range_del_ks,
-                     const std::vector<std::string>& range_del_vs,
-                     SequenceNumber last_sequence,
-                     MergeOperator* merge_op = nullptr,
-                     CompactionFilter* filter = nullptr) {
+  void InitIterators(
+      const std::vector<std::string>& ks, const std::vector<std::string>& vs,
+      const std::vector<std::string>& range_del_ks,
+      const std::vector<std::string>& range_del_vs,
+      SequenceNumber last_sequence,
+      SequenceNumber last_committed_sequence = kMaxSequenceNumber,
+      MergeOperator* merge_op = nullptr, CompactionFilter* filter = nullptr,
+      bool bottommost_level = false,
+      SequenceNumber earliest_write_conflict_snapshot = kMaxSequenceNumber) {
    std::unique_ptr<InternalIterator> range_del_iter(
        new test::VectorIterator(range_del_ks, range_del_vs));
    range_del_agg_.reset(new RangeDelAggregator(icmp_, snapshots_));
    ASSERT_OK(range_del_agg_->AddTombstones(std::move(range_del_iter)));

    std::unique_ptr<CompactionIterator::CompactionProxy> compaction;
-    if (filter) {
+    if (filter || bottommost_level) {
      compaction_proxy_ = new FakeCompaction();
+      compaction_proxy_->is_bottommost_level = bottommost_level;
      compaction.reset(compaction_proxy_);
    }
-    // TODO(yiwu) add a mock snapshot checker and add test for it.
-    SnapshotChecker* snapshot_checker = nullptr;
+    bool use_snapshot_checker = UseSnapshotChecker() || GetParam();
+    if (use_snapshot_checker || last_committed_sequence < kMaxSequenceNumber) {
+      snapshot_checker_.reset(
+          new TestSnapshotChecker(last_committed_sequence, snapshot_map_));
+    }
+    merge_helper_.reset(
+        new MergeHelper(Env::Default(), cmp_, merge_op, filter, nullptr, false,
+                        0 /*latest_snapshot*/, snapshot_checker_.get(),
+                        0 /*level*/, nullptr /*statistics*/, &shutting_down_));

-    merge_helper_.reset(new MergeHelper(Env::Default(), cmp_, merge_op, filter,
-                                        nullptr, false, 0, 0, nullptr,
-                                        &shutting_down_));
    iter_.reset(new LoggingForwardVectorIterator(ks, vs));
    iter_->SeekToFirst();
    c_iter_.reset(new CompactionIterator(
        iter_.get(), cmp_, merge_helper_.get(), last_sequence, &snapshots_,
-        kMaxSequenceNumber, snapshot_checker, Env::Default(), false,
-        range_del_agg_.get(), std::move(compaction), filter, nullptr,
-        &shutting_down_));
+        earliest_write_conflict_snapshot, snapshot_checker_.get(),
+        Env::Default(), false, range_del_agg_.get(), std::move(compaction),
+        filter, nullptr, &shutting_down_));
  }

-  void AddSnapshot(SequenceNumber snapshot) { snapshots_.push_back(snapshot); }
+  void AddSnapshot(SequenceNumber snapshot,
+                   SequenceNumber last_visible_seq = kMaxSequenceNumber) {
+    snapshots_.push_back(snapshot);
+    snapshot_map_[snapshot] = last_visible_seq;
+  }
+
+  virtual bool UseSnapshotChecker() const { return false; }
+
+  void RunTest(
+      const std::vector<std::string>& input_keys,
+      const std::vector<std::string>& input_values,
+      const std::vector<std::string>& expected_keys,
+      const std::vector<std::string>& expected_values,
+      SequenceNumber last_committed_seq = kMaxSequenceNumber,
+      MergeOperator* merge_operator = nullptr,
+      CompactionFilter* compaction_filter = nullptr,
+      bool bottommost_level = false,
+      SequenceNumber earliest_write_conflict_snapshot = kMaxSequenceNumber) {
+    InitIterators(input_keys, input_values, {}, {}, kMaxSequenceNumber,
+                  last_committed_seq, merge_operator, compaction_filter,
+                  bottommost_level, earliest_write_conflict_snapshot);
+    c_iter_->SeekToFirst();
+    for (size_t i = 0; i < expected_keys.size(); i++) {
+      std::string info = "i = " + ToString(i);
+      ASSERT_TRUE(c_iter_->Valid()) << info;
+      ASSERT_OK(c_iter_->status()) << info;
+      ASSERT_EQ(expected_keys[i], c_iter_->key().ToString()) << info;
+      ASSERT_EQ(expected_values[i], c_iter_->value().ToString()) << info;
+      c_iter_->Next();
+    }
+    ASSERT_FALSE(c_iter_->Valid());
+  }

  const Comparator* cmp_;
  const InternalKeyComparator icmp_;
  std::vector<SequenceNumber> snapshots_;
+  // A map of valid snapshot to last visible sequence to the snapshot.
+  std::unordered_map<SequenceNumber, SequenceNumber> snapshot_map_;
  std::unique_ptr<MergeHelper> merge_helper_;
  std::unique_ptr<LoggingForwardVectorIterator> iter_;
  std::unique_ptr<CompactionIterator> c_iter_;
  std::unique_ptr<RangeDelAggregator> range_del_agg_;
+  std::unique_ptr<SnapshotChecker> snapshot_checker_;
  std::atomic<bool> shutting_down_{false};
  FakeCompaction* compaction_proxy_;
 };

 // It is possible that the output of the compaction iterator is empty even if
 // the input is not.
-TEST_F(CompactionIteratorTest, EmptyResult) {
+TEST_P(CompactionIteratorTest, EmptyResult) {
  InitIterators({test::KeyStr("a", 5, kTypeSingleDeletion),
                 test::KeyStr("a", 3, kTypeValue)},
                {"", "val"}, {}, {}, 5);
@@ -223,7 +307,7 @@ TEST_F(CompactionIteratorTest, EmptyResult) {

 // If there is a corruption after a single deletion, the corrupted key should
 // be preserved.
-TEST_F(CompactionIteratorTest, CorruptionAfterSingleDeletion) {
+TEST_P(CompactionIteratorTest, CorruptionAfterSingleDeletion) {
  InitIterators({test::KeyStr("a", 5, kTypeSingleDeletion),
                 test::KeyStr("a", 3, kTypeValue, true),
                 test::KeyStr("b", 10, kTypeValue)},
@@ -242,7 +326,7 @@ TEST_F(CompactionIteratorTest, CorruptionAfterSingleDeletion) {
  ASSERT_FALSE(c_iter_->Valid());
 }

-TEST_F(CompactionIteratorTest, SimpleRangeDeletion) {
+TEST_P(CompactionIteratorTest, SimpleRangeDeletion) {
  InitIterators({test::KeyStr("morning", 5, kTypeValue),
                 test::KeyStr("morning", 2, kTypeValue),
                 test::KeyStr("night", 3, kTypeValue)},
@@ -258,7 +342,7 @@ TEST_F(CompactionIteratorTest, SimpleRangeDeletion) {
  ASSERT_FALSE(c_iter_->Valid());
 }

-TEST_F(CompactionIteratorTest, RangeDeletionWithSnapshots) {
+TEST_P(CompactionIteratorTest, RangeDeletionWithSnapshots) {
  AddSnapshot(10);
  std::vector<std::string> ks1;
  ks1.push_back(test::KeyStr("ma", 28, kTypeRangeDeletion));
@@ -279,7 +363,7 @@ TEST_F(CompactionIteratorTest, RangeDeletionWithSnapshots) {
  ASSERT_FALSE(c_iter_->Valid());
 }

-TEST_F(CompactionIteratorTest, CompactionFilterSkipUntil) {
+TEST_P(CompactionIteratorTest, CompactionFilterSkipUntil) {
  class Filter : public CompactionFilter {
    virtual Decision FilterV2(int level, const Slice& key, ValueType t,
                              const Slice& existing_value,
@@ -354,7 +438,7 @@ TEST_F(CompactionIteratorTest, CompactionFilterSkipUntil) {
       test::KeyStr("j", 99, kTypeValue)},
      {"av50", "am45", "bv60", "bv40", "cv35", "dm70", "em71", "fm65", "fm30",
       "fv25", "gv90", "hv91", "im95", "jv99"},
-      {}, {}, kMaxSequenceNumber, &merge_op, &filter);
+      {}, {}, kMaxSequenceNumber, kMaxSequenceNumber, &merge_op, &filter);

  // Compaction should output just "a", "e" and "h" keys.
  c_iter_->SeekToFirst();
@@ -389,13 +473,14 @@ TEST_F(CompactionIteratorTest, CompactionFilterSkipUntil) {
  ASSERT_EQ(expected_actions, iter_->log);
 }

-TEST_F(CompactionIteratorTest, ShuttingDownInFilter) {
+TEST_P(CompactionIteratorTest, ShuttingDownInFilter) {
  NoMergingMergeOp merge_op;
  StallingFilter filter;
  InitIterators(
      {test::KeyStr("1", 1, kTypeValue), test::KeyStr("2", 2, kTypeValue),
       test::KeyStr("3", 3, kTypeValue), test::KeyStr("4", 4, kTypeValue)},
-      {"v1", "v2", "v3", "v4"}, {}, {}, kMaxSequenceNumber, &merge_op, &filter);
+      {"v1", "v2", "v3", "v4"}, {}, {}, kMaxSequenceNumber, kMaxSequenceNumber,
+      &merge_op, &filter);
  // Don't leave tombstones (kTypeDeletion) for filtered keys.
  compaction_proxy_->key_not_exists_beyond_output_level = true;

@@ -426,13 +511,14 @@ TEST_F(CompactionIteratorTest, ShuttingDownInFilter) {

 // Same as ShuttingDownInFilter, but shutdown happens during filter call for
 // a merge operand, not for a value.
-TEST_F(CompactionIteratorTest, ShuttingDownInMerge) {
+TEST_P(CompactionIteratorTest, ShuttingDownInMerge) {
  NoMergingMergeOp merge_op;
  StallingFilter filter;
  InitIterators(
      {test::KeyStr("1", 1, kTypeValue), test::KeyStr("2", 2, kTypeMerge),
       test::KeyStr("3", 3, kTypeMerge), test::KeyStr("4", 4, kTypeValue)},
-      {"v1", "v2", "v3", "v4"}, {}, {}, kMaxSequenceNumber, &merge_op, &filter);
+      {"v1", "v2", "v3", "v4"}, {}, {}, kMaxSequenceNumber, kMaxSequenceNumber,
+      &merge_op, &filter);
  compaction_proxy_->key_not_exists_beyond_output_level = true;

  std::atomic<bool> seek_done{false};
@@ -460,7 +546,7 @@ TEST_F(CompactionIteratorTest, ShuttingDownInMerge) {
  EXPECT_EQ(2, filter.last_seen.load());
 }

-TEST_F(CompactionIteratorTest, SingleMergeOperand) {
+TEST_P(CompactionIteratorTest, SingleMergeOperand) {
  class Filter : public CompactionFilter {
    virtual Decision FilterV2(int level, const Slice& key, ValueType t,
                              const Slice& existing_value,
@@ -552,7 +638,7 @@ TEST_F(CompactionIteratorTest, SingleMergeOperand) {
       // c should invoke FullMerge due to kTypeValue at the beginning.
       test::KeyStr("c", 90, kTypeMerge), test::KeyStr("c", 80, kTypeValue)},
      {"av1", "bv2", "bv1", "cv2", "cv1"}, {}, {}, kMaxSequenceNumber,
-      &merge_op, &filter);
+      kMaxSequenceNumber, &merge_op, &filter);

  c_iter_->SeekToFirst();
  ASSERT_TRUE(c_iter_->Valid());
@@ -565,6 +651,298 @@ TEST_F(CompactionIteratorTest, SingleMergeOperand) {
  ASSERT_EQ("cv1cv2", c_iter_->value().ToString());
 }

+// In bottommost level, values earlier than earliest snapshot can be output
+// with sequence = 0.
+TEST_P(CompactionIteratorTest, ZeroOutSequenceAtBottomLevel) {
+  AddSnapshot(1);
+  RunTest({test::KeyStr("a", 1, kTypeValue), test::KeyStr("b", 2, kTypeValue)},
+          {"v1", "v2"},
+          {test::KeyStr("a", 0, kTypeValue), test::KeyStr("b", 2, kTypeValue)},
+          {"v1", "v2"}, kMaxSequenceNumber /*last_commited_seq*/,
+          nullptr /*merge_operator*/, nullptr /*compaction_filter*/,
+          true /*bottommost_level*/);
+}
+
+// In bottommost level, deletions earlier than earliest snapshot can be removed
+// permanently.
+TEST_P(CompactionIteratorTest, RemoveDeletionAtBottomLevel) {
+  AddSnapshot(1);
+  RunTest({test::KeyStr("a", 1, kTypeDeletion),
+           test::KeyStr("b", 2, kTypeDeletion)},
+          {"", ""}, {test::KeyStr("b", 2, kTypeDeletion)}, {""},
+          kMaxSequenceNumber /*last_commited_seq*/, nullptr /*merge_operator*/,
+          nullptr /*compaction_filter*/, true /*bottommost_level*/);
+}
+
+// In bottommost level, single deletions earlier than earliest snapshot can be
+// removed permanently.
+TEST_P(CompactionIteratorTest, RemoveSingleDeletionAtBottomLevel) {
+  AddSnapshot(1);
+  RunTest({test::KeyStr("a", 1, kTypeSingleDeletion),
+           test::KeyStr("b", 2, kTypeSingleDeletion)},
+          {"", ""}, {test::KeyStr("b", 2, kTypeSingleDeletion)}, {""},
+          kMaxSequenceNumber /*last_commited_seq*/, nullptr /*merge_operator*/,
+          nullptr /*compaction_filter*/, true /*bottommost_level*/);
+}
+
+INSTANTIATE_TEST_CASE_P(CompactionIteratorTestInstance, CompactionIteratorTest,
+                        testing::Values(true, false));
+
+// Tests how CompactionIterator work together with SnapshotChecker.
+class CompactionIteratorWithSnapshotCheckerTest
+    : public CompactionIteratorTest {
+ public:
+  bool UseSnapshotChecker() const override { return true; }
+};
+
+// Uncommitted keys (keys with seq > last_committed_seq) should be output as-is
+// while committed version of these keys should get compacted as usual.
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest,
+       PreserveUncommittedKeys_Value) {
+  RunTest(
+      {test::KeyStr("foo", 3, kTypeValue), test::KeyStr("foo", 2, kTypeValue),
+       test::KeyStr("foo", 1, kTypeValue)},
+      {"v3", "v2", "v1"},
+      {test::KeyStr("foo", 3, kTypeValue), test::KeyStr("foo", 2, kTypeValue)},
+      {"v3", "v2"}, 2 /*last_committed_seq*/);
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest,
+       PreserveUncommittedKeys_Deletion) {
+  RunTest({test::KeyStr("foo", 2, kTypeDeletion),
+           test::KeyStr("foo", 1, kTypeValue)},
+          {"", "v1"},
+          {test::KeyStr("foo", 2, kTypeDeletion),
+           test::KeyStr("foo", 1, kTypeValue)},
+          {"", "v1"}, 1 /*last_committed_seq*/);
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest,
+       PreserveUncommittedKeys_Merge) {
+  auto merge_op = MergeOperators::CreateStringAppendOperator();
+  RunTest(
+      {test::KeyStr("foo", 3, kTypeMerge), test::KeyStr("foo", 2, kTypeMerge),
+       test::KeyStr("foo", 1, kTypeValue)},
+      {"v3", "v2", "v1"},
+      {test::KeyStr("foo", 3, kTypeMerge), test::KeyStr("foo", 2, kTypeValue)},
+      {"v3", "v1,v2"}, 2 /*last_committed_seq*/, merge_op.get());
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest,
+       PreserveUncommittedKeys_SingleDelete) {
+  RunTest({test::KeyStr("foo", 2, kTypeSingleDeletion),
+           test::KeyStr("foo", 1, kTypeValue)},
+          {"", "v1"},
+          {test::KeyStr("foo", 2, kTypeSingleDeletion),
+           test::KeyStr("foo", 1, kTypeValue)},
+          {"", "v1"}, 1 /*last_committed_seq*/);
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest,
+       PreserveUncommittedKeys_BlobIndex) {
+  RunTest({test::KeyStr("foo", 3, kTypeBlobIndex),
+           test::KeyStr("foo", 2, kTypeBlobIndex),
+           test::KeyStr("foo", 1, kTypeBlobIndex)},
+          {"v3", "v2", "v1"},
+          {test::KeyStr("foo", 3, kTypeBlobIndex),
+           test::KeyStr("foo", 2, kTypeBlobIndex)},
+          {"v3", "v2"}, 2 /*last_committed_seq*/);
+}
+
+// Test compaction iterator dedup keys visible to the same snapshot.
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest, DedupSameSnapshot_Value) {
+  AddSnapshot(2, 1);
+  RunTest(
+      {test::KeyStr("foo", 4, kTypeValue), test::KeyStr("foo", 3, kTypeValue),
+       test::KeyStr("foo", 2, kTypeValue), test::KeyStr("foo", 1, kTypeValue)},
+      {"v4", "v3", "v2", "v1"},
+      {test::KeyStr("foo", 4, kTypeValue), test::KeyStr("foo", 3, kTypeValue),
+       test::KeyStr("foo", 1, kTypeValue)},
+      {"v4", "v3", "v1"}, 3 /*last_committed_seq*/);
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest, DedupSameSnapshot_Deletion) {
+  AddSnapshot(2, 1);
+  RunTest(
+      {test::KeyStr("foo", 4, kTypeValue),
+       test::KeyStr("foo", 3, kTypeDeletion),
+       test::KeyStr("foo", 2, kTypeValue), test::KeyStr("foo", 1, kTypeValue)},
+      {"v4", "", "v2", "v1"},
+      {test::KeyStr("foo", 4, kTypeValue),
+       test::KeyStr("foo", 3, kTypeDeletion),
+       test::KeyStr("foo", 1, kTypeValue)},
+      {"v4", "", "v1"}, 3 /*last_committed_seq*/);
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest, DedupSameSnapshot_Merge) {
+  AddSnapshot(2, 1);
+  AddSnapshot(4, 3);
+  auto merge_op = MergeOperators::CreateStringAppendOperator();
+  RunTest(
+      {test::KeyStr("foo", 5, kTypeMerge), test::KeyStr("foo", 4, kTypeMerge),
+       test::KeyStr("foo", 3, kTypeMerge), test::KeyStr("foo", 2, kTypeMerge),
+       test::KeyStr("foo", 1, kTypeValue)},
+      {"v5", "v4", "v3", "v2", "v1"},
+      {test::KeyStr("foo", 5, kTypeMerge), test::KeyStr("foo", 4, kTypeMerge),
+       test::KeyStr("foo", 3, kTypeMerge), test::KeyStr("foo", 1, kTypeValue)},
+      {"v5", "v4", "v2,v3", "v1"}, 4 /*last_committed_seq*/, merge_op.get());
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest,
+       DedupSameSnapshot_SingleDeletion) {
+  AddSnapshot(2, 1);
+  RunTest(
+      {test::KeyStr("foo", 4, kTypeValue),
+       test::KeyStr("foo", 3, kTypeSingleDeletion),
+       test::KeyStr("foo", 2, kTypeValue), test::KeyStr("foo", 1, kTypeValue)},
+      {"v4", "", "v2", "v1"},
+      {test::KeyStr("foo", 4, kTypeValue), test::KeyStr("foo", 1, kTypeValue)},
+      {"v4", "v1"}, 3 /*last_committed_seq*/);
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest, DedupSameSnapshot_BlobIndex) {
+  AddSnapshot(2, 1);
+  RunTest({test::KeyStr("foo", 4, kTypeBlobIndex),
+           test::KeyStr("foo", 3, kTypeBlobIndex),
+           test::KeyStr("foo", 2, kTypeBlobIndex),
+           test::KeyStr("foo", 1, kTypeBlobIndex)},
+          {"v4", "v3", "v2", "v1"},
+          {test::KeyStr("foo", 4, kTypeBlobIndex),
+           test::KeyStr("foo", 3, kTypeBlobIndex),
+           test::KeyStr("foo", 1, kTypeBlobIndex)},
+          {"v4", "v3", "v1"}, 3 /*last_committed_seq*/);
+}
+
+// At bottom level, sequence numbers can be zero out, and deletions can be
+// removed, but only when they are visible to earliest snapshot.
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest,
+       NotZeroOutSequenceIfNotVisibleToEarliestSnapshot) {
+  AddSnapshot(2, 1);
+  RunTest({test::KeyStr("a", 1, kTypeValue), test::KeyStr("b", 2, kTypeValue),
+           test::KeyStr("c", 3, kTypeValue)},
+          {"v1", "v2", "v3"},
+          {test::KeyStr("a", 0, kTypeValue), test::KeyStr("b", 2, kTypeValue),
+           test::KeyStr("c", 3, kTypeValue)},
+          {"v1", "v2", "v3"}, kMaxSequenceNumber /*last_commited_seq*/,
+          nullptr /*merge_operator*/, nullptr /*compaction_filter*/,
+          true /*bottommost_level*/);
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest,
+       NotRemoveDeletionIfNotVisibleToEarliestSnapshot) {
+  AddSnapshot(2, 1);
+  RunTest(
+      {test::KeyStr("a", 1, kTypeDeletion), test::KeyStr("b", 2, kTypeDeletion),
+       test::KeyStr("c", 3, kTypeDeletion)},
+      {"", "", ""},
+      {test::KeyStr("b", 2, kTypeDeletion),
+       test::KeyStr("c", 3, kTypeDeletion)},
+      {"", ""}, kMaxSequenceNumber /*last_commited_seq*/,
+      nullptr /*merge_operator*/, nullptr /*compaction_filter*/,
+      true /*bottommost_level*/);
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest,
+       NotRemoveSingleDeletionIfNotVisibleToEarliestSnapshot) {
+  AddSnapshot(2, 1);
+  RunTest({test::KeyStr("a", 1, kTypeSingleDeletion),
+           test::KeyStr("b", 2, kTypeSingleDeletion),
+           test::KeyStr("c", 3, kTypeSingleDeletion)},
+          {"", "", ""},
+          {test::KeyStr("b", 2, kTypeSingleDeletion),
+           test::KeyStr("c", 3, kTypeSingleDeletion)},
+          {"", ""}, kMaxSequenceNumber /*last_commited_seq*/,
+          nullptr /*merge_operator*/, nullptr /*compaction_filter*/,
+          true /*bottommost_level*/);
+}
+
+// Single delete should not cancel out values that not visible to the
+// same set of snapshots
+TEST_F(CompactionIteratorWithSnapshotCheckerTest,
+       SingleDeleteAcrossSnapshotBoundary) {
+  AddSnapshot(2, 1);
+  RunTest({test::KeyStr("a", 2, kTypeSingleDeletion),
+           test::KeyStr("a", 1, kTypeValue)},
+          {"", "v1"},
+          {test::KeyStr("a", 2, kTypeSingleDeletion),
+           test::KeyStr("a", 1, kTypeValue)},
+          {"", "v1"}, 2 /*last_committed_seq*/);
+}
+
+// Single delete should be kept in case it is not visible to the
+// earliest write conflict snapshot. If a single delete is kept for this reason,
+// corresponding value can be trimmed to save space.
+TEST_F(CompactionIteratorWithSnapshotCheckerTest,
+       KeepSingleDeletionForWriteConflictChecking) {
+  AddSnapshot(2, 0);
+  RunTest({test::KeyStr("a", 2, kTypeSingleDeletion),
+           test::KeyStr("a", 1, kTypeValue)},
+          {"", "v1"},
+          {test::KeyStr("a", 2, kTypeSingleDeletion),
+           test::KeyStr("a", 1, kTypeValue)},
+          {"", ""}, 2 /*last_committed_seq*/, nullptr /*merge_operator*/,
+          nullptr /*compaction_filter*/, false /*bottommost_level*/,
+          2 /*earliest_write_conflict_snapshot*/);
+}
+
+// Compaction filter should keep uncommitted key as-is, and
+//   * Convert the latest velue to deletion, and/or
+//   * if latest value is a merge, apply filter to all suequent merges.
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest, CompactionFilter_Value) {
+  std::unique_ptr<CompactionFilter> compaction_filter(
+      new FilterAllKeysCompactionFilter());
+  RunTest(
+      {test::KeyStr("a", 2, kTypeValue), test::KeyStr("a", 1, kTypeValue),
+       test::KeyStr("b", 3, kTypeValue), test::KeyStr("c", 1, kTypeValue)},
+      {"v2", "v1", "v3", "v4"},
+      {test::KeyStr("a", 2, kTypeValue), test::KeyStr("a", 1, kTypeDeletion),
+       test::KeyStr("b", 3, kTypeValue), test::KeyStr("c", 1, kTypeDeletion)},
+      {"v2", "", "v3", ""}, 1 /*last_committed_seq*/,
+      nullptr /*merge_operator*/, compaction_filter.get());
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest, CompactionFilter_Deletion) {
+  std::unique_ptr<CompactionFilter> compaction_filter(
+      new FilterAllKeysCompactionFilter());
+  RunTest(
+      {test::KeyStr("a", 2, kTypeDeletion), test::KeyStr("a", 1, kTypeValue)},
+      {"", "v1"},
+      {test::KeyStr("a", 2, kTypeDeletion),
+       test::KeyStr("a", 1, kTypeDeletion)},
+      {"", ""}, 1 /*last_committed_seq*/, nullptr /*merge_operator*/,
+      compaction_filter.get());
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest,
+       CompactionFilter_PartialMerge) {
+  std::shared_ptr<MergeOperator> merge_op =
+      MergeOperators::CreateStringAppendOperator();
+  std::unique_ptr<CompactionFilter> compaction_filter(
+      new FilterAllKeysCompactionFilter());
+  RunTest({test::KeyStr("a", 3, kTypeMerge), test::KeyStr("a", 2, kTypeMerge),
+           test::KeyStr("a", 1, kTypeMerge)},
+          {"v3", "v2", "v1"}, {test::KeyStr("a", 3, kTypeMerge)}, {"v3"},
+          2 /*last_committed_seq*/, merge_op.get(), compaction_filter.get());
+}
+
+TEST_F(CompactionIteratorWithSnapshotCheckerTest, CompactionFilter_FullMerge) {
+  std::shared_ptr<MergeOperator> merge_op =
+      MergeOperators::CreateStringAppendOperator();
+  std::unique_ptr<CompactionFilter> compaction_filter(
+      new FilterAllKeysCompactionFilter());
+  RunTest(
+      {test::KeyStr("a", 3, kTypeMerge), test::KeyStr("a", 2, kTypeMerge),
+       test::KeyStr("a", 1, kTypeValue)},
+      {"v3", "v2", "v1"},
+      {test::KeyStr("a", 3, kTypeMerge), test::KeyStr("a", 1, kTypeDeletion)},
+      {"v3", ""}, 2 /*last_committed_seq*/, merge_op.get(),
+      compaction_filter.get());
+}
+
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -747,8 +747,8 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
      compaction_filter, db_options_.info_log.get(),
      false /* internal key corruption is expected */,
      existing_snapshots_.empty() ? 0 : existing_snapshots_.back(),
-      compact_->compaction->level(), db_options_.statistics.get(),
-      shutting_down_);
+      snapshot_checker_, compact_->compaction->level(),
+      db_options_.statistics.get(), shutting_down_);

  TEST_SYNC_POINT("CompactionJob::Run():Inprogress");

@@ -1034,7 +1034,7 @@ int main(int argc, char** argv) {
 #else
 #include <stdio.h>

-int main(int argc, char** argv) {
+int main(int /*argc*/, char** /*argv*/) {
  fprintf(stderr, "SKIPPED, not supported in ROCKSDB_LITE\n");
  return 0;
 }
@@ -1043,5 +1043,5 @@ int main(int argc, char** argv) {

 #else

-int main(int argc, char** argv) { return 0; }
+int main(int /*argc*/, char** /*argv*/) { return 0; }
 #endif  // !defined(IOS_CROSS_COMPILE)
@@ -946,7 +946,7 @@ int main(int argc, char** argv) {
 #else
 #include <stdio.h>

-int main(int argc, char** argv) {
+int main(int /*argc*/, char** /*argv*/) {
  fprintf(stderr,
          "SKIPPED as CompactionJobStats is not supported in ROCKSDB_LITE\n");
  return 0;
@@ -175,6 +175,8 @@ TEST_F(CompactionPickerTest, Level1Trigger) {
 }

 TEST_F(CompactionPickerTest, Level1Trigger2) {
+  mutable_cf_options_.target_file_size_base = 10000000000;
+  mutable_cf_options_.RefreshDerivedOptions(ioptions_);
  NewVersionStorage(6, kCompactionStyleLevel);
  Add(1, 66U, "150", "200", 1000000001U);
  Add(1, 88U, "201", "300", 1000000000U);
@@ -191,13 +193,14 @@ TEST_F(CompactionPickerTest, Level1Trigger2) {
  ASSERT_EQ(66U, compaction->input(0, 0)->fd.GetNumber());
  ASSERT_EQ(6U, compaction->input(1, 0)->fd.GetNumber());
  ASSERT_EQ(7U, compaction->input(1, 1)->fd.GetNumber());
+  ASSERT_EQ(uint64_t{1073741824}, compaction->OutputFilePreallocationSize());
 }

 TEST_F(CompactionPickerTest, LevelMaxScore) {
  NewVersionStorage(6, kCompactionStyleLevel);
  mutable_cf_options_.target_file_size_base = 10000000;
-  mutable_cf_options_.target_file_size_multiplier = 10;
  mutable_cf_options_.max_bytes_for_level_base = 10 * 1024 * 1024;
+  mutable_cf_options_.RefreshDerivedOptions(ioptions_);
  Add(0, 1U, "150", "200", 1000000U);
  // Level 1 score 1.2
  Add(1, 66U, "150", "200", 6000000U);
@@ -218,6 +221,9 @@ TEST_F(CompactionPickerTest, LevelMaxScore) {
  ASSERT_TRUE(compaction.get() != nullptr);
  ASSERT_EQ(1U, compaction->num_input_files(0));
  ASSERT_EQ(7U, compaction->input(0, 0)->fd.GetNumber());
+  ASSERT_EQ(mutable_cf_options_.target_file_size_base +
+                mutable_cf_options_.target_file_size_base / 10,
+            compaction->OutputFilePreallocationSize());
 }

 TEST_F(CompactionPickerTest, NeedsCompactionLevel) {
@@ -521,9 +527,10 @@ TEST_F(CompactionPickerTest, NeedsCompactionFIFO) {
 TEST_F(CompactionPickerTest, CompactionPriMinOverlapping1) {
  NewVersionStorage(6, kCompactionStyleLevel);
  ioptions_.compaction_pri = kMinOverlappingRatio;
-  mutable_cf_options_.target_file_size_base = 10000000;
+  mutable_cf_options_.target_file_size_base = 100000000000;
  mutable_cf_options_.target_file_size_multiplier = 10;
  mutable_cf_options_.max_bytes_for_level_base = 10 * 1024 * 1024;
+  mutable_cf_options_.RefreshDerivedOptions(ioptions_);

  Add(2, 6U, "150", "179", 50000000U);
  Add(2, 7U, "180", "220", 50000000U);
@@ -543,6 +550,8 @@ TEST_F(CompactionPickerTest, CompactionPriMinOverlapping1) {
  ASSERT_EQ(1U, compaction->num_input_files(0));
  // Pick file 8 because it overlaps with 0 files on level 3.
  ASSERT_EQ(8U, compaction->input(0, 0)->fd.GetNumber());
+  // Compaction input size * 1.1
+  ASSERT_GE(uint64_t{55000000}, compaction->OutputFilePreallocationSize());
 }

 TEST_F(CompactionPickerTest, CompactionPriMinOverlapping2) {
@@ -603,9 +603,9 @@ Compaction* UniversalCompactionPicker::PickCompactionToReduceSortedRuns(

  CompactionReason compaction_reason;
  if (max_number_of_files_to_compact == UINT_MAX) {
-    compaction_reason = CompactionReason::kUniversalSortedRunNum;
-  } else {
    compaction_reason = CompactionReason::kUniversalSizeRatio;
+  } else {
+    compaction_reason = CompactionReason::kUniversalSortedRunNum;
  }
  return new Compaction(
      vstorage, ioptions_, mutable_cf_options, std::move(inputs), output_level,
@@ -19,9 +19,17 @@ void CancelAllBackgroundWork(DB* db, bool wait) {
 }

 Status DeleteFilesInRange(DB* db, ColumnFamilyHandle* column_family,
-                          const Slice* begin, const Slice* end) {
+                          const Slice* begin, const Slice* end,
+                          bool include_end) {
+  RangePtr range(begin, end);
+  return DeleteFilesInRanges(db, column_family, &range, 1, include_end);
+}
+
+Status DeleteFilesInRanges(DB* db, ColumnFamilyHandle* column_family,
+                           const RangePtr* ranges, size_t n,
+                           bool include_end) {
  return (static_cast_with_check<DBImpl, DB>(db->GetRootDB()))
-      ->DeleteFilesInRange(column_family, begin, end);
+      ->DeleteFilesInRanges(column_family, ranges, n, include_end);
 }

 Status VerifySstFileChecksum(const Options& options,
@@ -510,7 +510,7 @@ int main(int argc, char** argv) {
 #else
 #include <stdio.h>

-int main(int argc, char** argv) {
+int main(int /*argc*/, char** /*argv*/) {
  fprintf(stderr, "SKIPPED as RepairDB() is not supported in ROCKSDB_LITE\n");
  return 0;
 }
@@ -333,7 +333,7 @@ int main(int argc, char** argv) {
 #else
 #include <stdio.h>

-int main(int argc, char** argv) {
+int main(int /*argc*/, char** /*argv*/) {
  fprintf(stderr, "SKIPPED as Cuckoo table is not supported in ROCKSDB_LITE\n");
  return 0;
 }
@@ -9,6 +9,7 @@
 #include "db/db_test_util.h"
 #include "port/stack_trace.h"
 #include "rocksdb/perf_context.h"
+#include "util/fault_injection_test_env.h"
 #if !defined(ROCKSDB_LITE)
 #include "util/sync_point.h"
 #endif
@@ -570,19 +571,19 @@ TEST_F(DBBasicTest, CompactBetweenSnapshots) {

 TEST_F(DBBasicTest, DBOpen_Options) {
  Options options = CurrentOptions();
-  std::string dbname = test::TmpDir(env_) + "/db_options_test";
-  ASSERT_OK(DestroyDB(dbname, options));
+  Close();
+  Destroy(options);

  // Does not exist, and create_if_missing == false: error
  DB* db = nullptr;
  options.create_if_missing = false;
-  Status s = DB::Open(options, dbname, &db);
+  Status s = DB::Open(options, dbname_, &db);
  ASSERT_TRUE(strstr(s.ToString().c_str(), "does not exist") != nullptr);
  ASSERT_TRUE(db == nullptr);

  // Does not exist, and create_if_missing == true: OK
  options.create_if_missing = true;
-  s = DB::Open(options, dbname, &db);
+  s = DB::Open(options, dbname_, &db);
  ASSERT_OK(s);
  ASSERT_TRUE(db != nullptr);

@@ -592,14 +593,14 @@ TEST_F(DBBasicTest, DBOpen_Options) {
  // Does exist, and error_if_exists == true: error
  options.create_if_missing = false;
  options.error_if_exists = true;
-  s = DB::Open(options, dbname, &db);
+  s = DB::Open(options, dbname_, &db);
  ASSERT_TRUE(strstr(s.ToString().c_str(), "exists") != nullptr);
  ASSERT_TRUE(db == nullptr);

  // Does exist, and error_if_exists == false: OK
  options.create_if_missing = true;
  options.error_if_exists = false;
-  s = DB::Open(options, dbname, &db);
+  s = DB::Open(options, dbname_, &db);
  ASSERT_OK(s);
  ASSERT_TRUE(db != nullptr);

@@ -847,6 +848,113 @@ TEST_F(DBBasicTest, MmapAndBufferOptions) {
 }
 #endif

+class TestEnv : public EnvWrapper {
+  public:
+    explicit TestEnv() : EnvWrapper(Env::Default()),
+                close_count(0) { }
+
+    class TestLogger : public Logger {
+      public:
+        using Logger::Logv;
+        TestLogger(TestEnv *env_ptr) : Logger() { env = env_ptr; }
+        ~TestLogger() {
+          if (!closed_) {
+            CloseHelper();
+          }
+        }
+        virtual void Logv(const char *format, va_list ap) override { };
+      protected:
+        virtual Status CloseImpl() override {
+          return CloseHelper();
+        }
+      private:
+        Status CloseHelper() {
+          env->CloseCountInc();;
+          return Status::IOError();
+        }
+        TestEnv *env;
+    };
+
+    void CloseCountInc() { close_count++; }
+
+    int GetCloseCount() { return close_count; }
+
+    virtual Status NewLogger(const std::string& fname,
+                             shared_ptr<Logger>* result) {
+      result->reset(new TestLogger(this));
+      return Status::OK();
+    }
+
+  private:
+    int close_count;
+};
+
+TEST_F(DBBasicTest, DBClose) {
+  Options options = GetDefaultOptions();
+  std::string dbname = test::TmpDir(env_) + "/db_close_test";
+  ASSERT_OK(DestroyDB(dbname, options));
+
+  DB* db = nullptr;
+  TestEnv *env = new TestEnv();
+  options.create_if_missing = true;
+  options.env = env;
+  Status s = DB::Open(options, dbname, &db);
+  ASSERT_OK(s);
+  ASSERT_TRUE(db != nullptr);
+
+  s = db->Close();
+  ASSERT_EQ(env->GetCloseCount(), 1);
+  ASSERT_EQ(s, Status::IOError());
+
+  delete db;
+  ASSERT_EQ(env->GetCloseCount(), 1);
+
+  // Do not call DB::Close() and ensure our logger Close() still gets called
+  s = DB::Open(options, dbname, &db);
+  ASSERT_OK(s);
+  ASSERT_TRUE(db != nullptr);
+  delete db;
+  ASSERT_EQ(env->GetCloseCount(), 2);
+
+  // Provide our own logger and ensure DB::Close() does not close it
+  options.info_log.reset(new TestEnv::TestLogger(env));
+  options.create_if_missing = false;
+  s = DB::Open(options, dbname, &db);
+  ASSERT_OK(s);
+  ASSERT_TRUE(db != nullptr);
+
+  s = db->Close();
+  ASSERT_EQ(s, Status::OK());
+  delete db;
+  ASSERT_EQ(env->GetCloseCount(), 2);
+  options.info_log.reset();
+  ASSERT_EQ(env->GetCloseCount(), 3);
+
+  delete options.env;
+}
+
+TEST_F(DBBasicTest, DBCloseFlushError) {
+  std::unique_ptr<FaultInjectionTestEnv> fault_injection_env(
+      new FaultInjectionTestEnv(Env::Default()));
+  Options options = GetDefaultOptions();
+  options.create_if_missing = true;
+  options.manual_wal_flush = true;
+  options.write_buffer_size=100;
+  options.env = fault_injection_env.get();
+
+  Reopen(options);
+  ASSERT_OK(Put("key1", "value1"));
+  ASSERT_OK(Put("key2", "value2"));
+  ASSERT_OK(dbfull()->TEST_SwitchMemtable());
+  ASSERT_OK(Put("key3", "value3"));
+  fault_injection_env->SetFilesystemActive(false);
+  Status s = dbfull()->Close();
+  fault_injection_env->SetFilesystemActive(true);
+  ASSERT_NE(s, Status::OK());
+
+  Destroy(options);
+}
+
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -111,6 +111,31 @@ class DBBlockCacheTest : public DBTestBase {
  }
 };

+TEST_F(DBBlockCacheTest, IteratorBlockCacheUsage) {
+  ReadOptions read_options;
+  read_options.fill_cache = false;
+  auto table_options = GetTableOptions();
+  auto options = GetOptions(table_options);
+  InitTable(options);
+
+  std::shared_ptr<Cache> cache = NewLRUCache(0, 0, false);
+  table_options.block_cache = cache;
+  options.table_factory.reset(new BlockBasedTableFactory(table_options));
+  Reopen(options);
+  RecordCacheCounters(options);
+
+  std::vector<std::unique_ptr<Iterator>> iterators(kNumBlocks - 1);
+  Iterator* iter = nullptr;
+
+  ASSERT_EQ(0, cache->GetUsage());
+  iter = db_->NewIterator(read_options);
+  iter->Seek(ToString(0));
+  ASSERT_LT(0, cache->GetUsage());
+  delete iter;
+  iter = nullptr;
+  ASSERT_EQ(0, cache->GetUsage());
+}
+
 TEST_F(DBBlockCacheTest, TestWithoutCompressedBlockCache) {
  ReadOptions read_options;
  auto table_options = GetTableOptions();
@@ -1517,6 +1517,122 @@ TEST_F(DBCompactionTest, DeleteFileRange) {
  ASSERT_GT(old_num_files, new_num_files);
 }

+TEST_F(DBCompactionTest, DeleteFilesInRanges) {
+  Options options = CurrentOptions();
+  options.write_buffer_size = 10 * 1024 * 1024;
+  options.max_bytes_for_level_multiplier = 2;
+  options.num_levels = 4;
+  options.max_background_compactions = 3;
+  options.disable_auto_compactions = true;
+
+  DestroyAndReopen(options);
+  int32_t value_size = 10 * 1024;  // 10 KB
+
+  Random rnd(301);
+  std::map<int32_t, std::string> values;
+
+  // file [0 => 100), [100 => 200), ... [900, 1000)
+  for (auto i = 0; i < 10; i++) {
+    for (auto j = 0; j < 100; j++) {
+      auto k = i * 100 + j;
+      values[k] = RandomString(&rnd, value_size);
+      ASSERT_OK(Put(Key(k), values[k]));
+    }
+    ASSERT_OK(Flush());
+  }
+  ASSERT_EQ("10", FilesPerLevel(0));
+  CompactRangeOptions compact_options;
+  compact_options.change_level = true;
+  compact_options.target_level = 2;
+  ASSERT_OK(db_->CompactRange(compact_options, nullptr, nullptr));
+  ASSERT_EQ("0,0,10", FilesPerLevel(0));
+
+  // file [0 => 100), [200 => 300), ... [800, 900)
+  for (auto i = 0; i < 10; i+=2) {
+    for (auto j = 0; j < 100; j++) {
+      auto k = i * 100 + j;
+      ASSERT_OK(Put(Key(k), values[k]));
+    }
+    ASSERT_OK(Flush());
+  }
+  ASSERT_EQ("5,0,10", FilesPerLevel(0));
+  ASSERT_OK(dbfull()->TEST_CompactRange(0, nullptr, nullptr));
+  ASSERT_EQ("0,5,10", FilesPerLevel(0));
+
+  // Delete files in range [0, 299] (inclusive)
+  {
+    auto begin_str1 = Key(0), end_str1 = Key(100);
+    auto begin_str2 = Key(100), end_str2 = Key(200);
+    auto begin_str3 = Key(200), end_str3 = Key(299);
+    Slice begin1(begin_str1), end1(end_str1);
+    Slice begin2(begin_str2), end2(end_str2);
+    Slice begin3(begin_str3), end3(end_str3);
+    std::vector<RangePtr> ranges;
+    ranges.push_back(RangePtr(&begin1, &end1));
+    ranges.push_back(RangePtr(&begin2, &end2));
+    ranges.push_back(RangePtr(&begin3, &end3));
+    ASSERT_OK(DeleteFilesInRanges(db_, db_->DefaultColumnFamily(),
+                                  ranges.data(), ranges.size()));
+    ASSERT_EQ("0,3,7", FilesPerLevel(0));
+
+    // Keys [0, 300) should not exist.
+    for (auto i = 0; i < 300; i++) {
+      ReadOptions ropts;
+      std::string result;
+      auto s = db_->Get(ropts, Key(i), &result);
+      ASSERT_TRUE(s.IsNotFound());
+    }
+    for (auto i = 300; i < 1000; i++) {
+      ASSERT_EQ(Get(Key(i)), values[i]);
+    }
+  }
+
+  // Delete files in range [600, 999) (exclusive)
+  {
+    auto begin_str1 = Key(600), end_str1 = Key(800);
+    auto begin_str2 = Key(700), end_str2 = Key(900);
+    auto begin_str3 = Key(800), end_str3 = Key(999);
+    Slice begin1(begin_str1), end1(end_str1);
+    Slice begin2(begin_str2), end2(end_str2);
+    Slice begin3(begin_str3), end3(end_str3);
+    std::vector<RangePtr> ranges;
+    ranges.push_back(RangePtr(&begin1, &end1));
+    ranges.push_back(RangePtr(&begin2, &end2));
+    ranges.push_back(RangePtr(&begin3, &end3));
+    ASSERT_OK(DeleteFilesInRanges(db_, db_->DefaultColumnFamily(),
+                                  ranges.data(), ranges.size(), false));
+    ASSERT_EQ("0,1,4", FilesPerLevel(0));
+
+    // Keys [600, 900) should not exist.
+    for (auto i = 600; i < 900; i++) {
+      ReadOptions ropts;
+      std::string result;
+      auto s = db_->Get(ropts, Key(i), &result);
+      ASSERT_TRUE(s.IsNotFound());
+    }
+    for (auto i = 300; i < 600; i++) {
+      ASSERT_EQ(Get(Key(i)), values[i]);
+    }
+    for (auto i = 900; i < 1000; i++) {
+      ASSERT_EQ(Get(Key(i)), values[i]);
+    }
+  }
+
+  // Delete all files.
+  {
+    RangePtr range;
+    ASSERT_OK(DeleteFilesInRanges(db_, db_->DefaultColumnFamily(), &range, 1));
+    ASSERT_EQ("", FilesPerLevel(0));
+
+    for (auto i = 0; i < 1000; i++) {
+      ReadOptions ropts;
+      std::string result;
+      auto s = db_->Get(ropts, Key(i), &result);
+      ASSERT_TRUE(s.IsNotFound());
+    }
+  }
+}
+
 TEST_F(DBCompactionTest, DeleteFileRangeFileEndpointsOverlapBug) {
  // regression test for #2833: groups of files whose user-keys overlap at the
  // endpoints could be split by `DeleteFilesInRange`. This caused old data to
@@ -2989,6 +3105,276 @@ TEST_F(DBCompactionTest, CompactBottomLevelFilesWithDeletions) {
  }
 }

+TEST_F(DBCompactionTest, CompactRangeDelayedByL0FileCount) {
+  // Verify that, when `CompactRangeOptions::allow_write_stall == false`, manual
+  // compaction only triggers flush after it's sure stall won't be triggered for
+  // L0 file count going too high.
+  const int kNumL0FilesTrigger = 4;
+  const int kNumL0FilesLimit = 8;
+  // i == 0: verifies normal case where stall is avoided by delay
+  // i == 1: verifies no delay in edge case where stall trigger is same as
+  //         compaction trigger, so stall can't be avoided
+  for (int i = 0; i < 2; ++i) {
+    Options options = CurrentOptions();
+    options.level0_slowdown_writes_trigger = kNumL0FilesLimit;
+    if (i == 0) {
+      options.level0_file_num_compaction_trigger = kNumL0FilesTrigger;
+    } else {
+      options.level0_file_num_compaction_trigger = kNumL0FilesLimit;
+    }
+    Reopen(options);
+
+    if (i == 0) {
+      // ensure the auto compaction doesn't finish until manual compaction has
+      // had a chance to be delayed.
+      rocksdb::SyncPoint::GetInstance()->LoadDependency(
+          {{"DBImpl::CompactRange:StallWait", "CompactionJob::Run():End"}});
+    } else {
+      // ensure the auto-compaction doesn't finish until manual compaction has
+      // continued without delay.
+      rocksdb::SyncPoint::GetInstance()->LoadDependency(
+          {{"DBImpl::CompactRange:StallWaitDone", "CompactionJob::Run():End"}});
+    }
+    rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+    Random rnd(301);
+    for (int j = 0; j < kNumL0FilesLimit - 1; ++j) {
+      for (int k = 0; k < 2; ++k) {
+        ASSERT_OK(Put(Key(k), RandomString(&rnd, 1024)));
+      }
+      Flush();
+    }
+    auto manual_compaction_thread = port::Thread([this]() {
+      CompactRangeOptions cro;
+      cro.allow_write_stall = false;
+      db_->CompactRange(cro, nullptr, nullptr);
+    });
+
+    manual_compaction_thread.join();
+    dbfull()->TEST_WaitForCompact();
+    ASSERT_EQ(0, NumTableFilesAtLevel(0));
+    ASSERT_GT(NumTableFilesAtLevel(1), 0);
+    rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+  }
+}
+
+TEST_F(DBCompactionTest, CompactRangeDelayedByImmMemTableCount) {
+  // Verify that, when `CompactRangeOptions::allow_write_stall == false`, manual
+  // compaction only triggers flush after it's sure stall won't be triggered for
+  // immutable memtable count going too high.
+  const int kNumImmMemTableLimit = 8;
+  // i == 0: verifies normal case where stall is avoided by delay
+  // i == 1: verifies no delay in edge case where stall trigger is same as flush
+  //         trigger, so stall can't be avoided
+  for (int i = 0; i < 2; ++i) {
+    Options options = CurrentOptions();
+    options.disable_auto_compactions = true;
+    // the delay limit is one less than the stop limit. This test focuses on
+    // avoiding delay limit, but this option sets stop limit, so add one.
+    options.max_write_buffer_number = kNumImmMemTableLimit + 1;
+    if (i == 1) {
+      options.min_write_buffer_number_to_merge = kNumImmMemTableLimit;
+    }
+    Reopen(options);
+
+    if (i == 0) {
+      // ensure the flush doesn't finish until manual compaction has had a
+      // chance to be delayed.
+      rocksdb::SyncPoint::GetInstance()->LoadDependency(
+          {{"DBImpl::CompactRange:StallWait", "FlushJob::WriteLevel0Table"}});
+    } else {
+      // ensure the flush doesn't finish until manual compaction has continued
+      // without delay.
+      rocksdb::SyncPoint::GetInstance()->LoadDependency(
+          {{"DBImpl::CompactRange:StallWaitDone",
+            "FlushJob::WriteLevel0Table"}});
+    }
+    rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+    Random rnd(301);
+    for (int j = 0; j < kNumImmMemTableLimit - 1; ++j) {
+      ASSERT_OK(Put(Key(0), RandomString(&rnd, 1024)));
+      FlushOptions flush_opts;
+      flush_opts.wait = false;
+      dbfull()->Flush(flush_opts);
+    }
+
+    auto manual_compaction_thread = port::Thread([this]() {
+      CompactRangeOptions cro;
+      cro.allow_write_stall = false;
+      db_->CompactRange(cro, nullptr, nullptr);
+    });
+
+    manual_compaction_thread.join();
+    dbfull()->TEST_WaitForFlushMemTable();
+    ASSERT_EQ(0, NumTableFilesAtLevel(0));
+    ASSERT_GT(NumTableFilesAtLevel(1), 0);
+    rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+  }
+}
+
+TEST_F(DBCompactionTest, CompactRangeShutdownWhileDelayed) {
+  // Verify that, when `CompactRangeOptions::allow_write_stall == false`, delay
+  // does not hang if CF is dropped or DB is closed
+  const int kNumL0FilesTrigger = 4;
+  const int kNumL0FilesLimit = 8;
+  Options options = CurrentOptions();
+  options.level0_file_num_compaction_trigger = kNumL0FilesTrigger;
+  options.level0_slowdown_writes_trigger = kNumL0FilesLimit;
+  // i == 0: DB::DropColumnFamily() on CompactRange's target CF unblocks it
+  // i == 1: DB::CancelAllBackgroundWork() unblocks CompactRange. This is to
+  //         simulate what happens during Close as we can't call Close (it
+  //         blocks on the auto-compaction, making a cycle).
+  for (int i = 0; i < 2; ++i) {
+    CreateAndReopenWithCF({"one"}, options);
+    // The calls to close CF/DB wait until the manual compaction stalls.
+    // The auto-compaction waits until the manual compaction finishes to ensure
+    // the signal comes from closing CF/DB, not from compaction making progress.
+    rocksdb::SyncPoint::GetInstance()->LoadDependency(
+        {{"DBImpl::CompactRange:StallWait",
+          "DBCompactionTest::CompactRangeShutdownWhileDelayed:PreShutdown"},
+         {"DBCompactionTest::CompactRangeShutdownWhileDelayed:PostManual",
+          "CompactionJob::Run():End"}});
+    rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+    Random rnd(301);
+    for (int j = 0; j < kNumL0FilesLimit - 1; ++j) {
+      for (int k = 0; k < 2; ++k) {
+        ASSERT_OK(Put(1, Key(k), RandomString(&rnd, 1024)));
+      }
+      Flush(1);
+    }
+    auto manual_compaction_thread = port::Thread([this]() {
+      CompactRangeOptions cro;
+      cro.allow_write_stall = false;
+      ASSERT_TRUE(db_->CompactRange(cro, handles_[1], nullptr, nullptr)
+                      .IsShutdownInProgress());
+    });
+
+    TEST_SYNC_POINT(
+        "DBCompactionTest::CompactRangeShutdownWhileDelayed:PreShutdown");
+    if (i == 0) {
+      ASSERT_OK(db_->DropColumnFamily(handles_[1]));
+    } else {
+      dbfull()->CancelAllBackgroundWork(false /* wait */);
+    }
+    manual_compaction_thread.join();
+    TEST_SYNC_POINT(
+        "DBCompactionTest::CompactRangeShutdownWhileDelayed:PostManual");
+    dbfull()->TEST_WaitForCompact();
+    rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+  }
+}
+
+TEST_F(DBCompactionTest, CompactRangeSkipFlushAfterDelay) {
+  // Verify that, when `CompactRangeOptions::allow_write_stall == false`,
+  // CompactRange skips its flush if the delay is long enough that the memtables
+  // existing at the beginning of the call have already been flushed.
+  const int kNumL0FilesTrigger = 4;
+  const int kNumL0FilesLimit = 8;
+  Options options = CurrentOptions();
+  options.level0_slowdown_writes_trigger = kNumL0FilesLimit;
+  options.level0_file_num_compaction_trigger = kNumL0FilesTrigger;
+  Reopen(options);
+
+  Random rnd(301);
+  // The manual flush includes the memtable that was active when CompactRange
+  // began. So it unblocks CompactRange and precludes its flush. Throughout the
+  // test, stall conditions are upheld via high L0 file count.
+  rocksdb::SyncPoint::GetInstance()->LoadDependency(
+      {{"DBImpl::CompactRange:StallWait",
+        "DBCompactionTest::CompactRangeSkipFlushAfterDelay:PreFlush"},
+       {"DBCompactionTest::CompactRangeSkipFlushAfterDelay:PostFlush",
+        "DBImpl::CompactRange:StallWaitDone"},
+       {"DBImpl::CompactRange:StallWaitDone", "CompactionJob::Run():End"}});
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  for (int i = 0; i < kNumL0FilesLimit - 1; ++i) {
+    for (int j = 0; j < 2; ++j) {
+      ASSERT_OK(Put(Key(j), RandomString(&rnd, 1024)));
+    }
+    Flush();
+  }
+  auto manual_compaction_thread = port::Thread([this]() {
+    CompactRangeOptions cro;
+    cro.allow_write_stall = false;
+    db_->CompactRange(cro, nullptr, nullptr);
+  });
+
+  TEST_SYNC_POINT("DBCompactionTest::CompactRangeSkipFlushAfterDelay:PreFlush");
+  Put(ToString(0), RandomString(&rnd, 1024));
+  Flush();
+  Put(ToString(0), RandomString(&rnd, 1024));
+  TEST_SYNC_POINT("DBCompactionTest::CompactRangeSkipFlushAfterDelay:PostFlush");
+  manual_compaction_thread.join();
+
+  // If CompactRange's flush was skipped, the final Put above will still be
+  // in the active memtable.
+  std::string num_keys_in_memtable;
+  db_->GetProperty(DB::Properties::kNumEntriesActiveMemTable, &num_keys_in_memtable);
+  ASSERT_EQ(ToString(1), num_keys_in_memtable);
+
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+}
+
+TEST_F(DBCompactionTest, CompactRangeFlushOverlappingMemtable) {
+  // Verify memtable only gets flushed if it contains data overlapping the range
+  // provided to `CompactRange`. Tests all kinds of overlap/non-overlap.
+  const int kNumEndpointKeys = 5;
+  std::string keys[kNumEndpointKeys] = {"a", "b", "c", "d", "e"};
+  Options options = CurrentOptions();
+  options.disable_auto_compactions = true;
+  Reopen(options);
+
+  // One extra iteration for nullptr, which means left side of interval is
+  // unbounded.
+  for (int i = 0; i <= kNumEndpointKeys; ++i) {
+    Slice begin;
+    Slice* begin_ptr;
+    if (i == 0) {
+      begin_ptr = nullptr;
+    } else {
+      begin = keys[i - 1];
+      begin_ptr = &begin;
+    }
+    // Start at `i` so right endpoint comes after left endpoint. One extra
+    // iteration for nullptr, which means right side of interval is unbounded.
+    for (int j = std::max(0, i - 1); j <= kNumEndpointKeys; ++j) {
+      Slice end;
+      Slice* end_ptr;
+      if (j == kNumEndpointKeys) {
+        end_ptr = nullptr;
+      } else {
+        end = keys[j];
+        end_ptr = &end;
+      }
+      ASSERT_OK(Put("b", "val"));
+      ASSERT_OK(Put("d", "val"));
+      CompactRangeOptions compact_range_opts;
+      ASSERT_OK(db_->CompactRange(compact_range_opts, begin_ptr, end_ptr));
+
+      uint64_t get_prop_tmp, num_memtable_entries = 0;
+      ASSERT_TRUE(db_->GetIntProperty(DB::Properties::kNumEntriesImmMemTables,
+                                      &get_prop_tmp));
+      num_memtable_entries += get_prop_tmp;
+      ASSERT_TRUE(db_->GetIntProperty(DB::Properties::kNumEntriesActiveMemTable,
+                                      &get_prop_tmp));
+      num_memtable_entries += get_prop_tmp;
+      if (begin_ptr == nullptr || end_ptr == nullptr ||
+          (i <= 4 && j >= 1 && (begin != "c" || end != "c"))) {
+        // In this case `CompactRange`'s range overlapped in some way with the
+        // memtable's range, so flush should've happened. Then "b" and "d" won't
+        // be in the memtable.
+        ASSERT_EQ(0, num_memtable_entries);
+      } else {
+        ASSERT_EQ(2, num_memtable_entries);
+        // flush anyways to prepare for next iteration
+        db_->Flush(FlushOptions());
+      }
+    }
+  }
+}
+
 INSTANTIATE_TEST_CASE_P(DBCompactionTestWithParam, DBCompactionTestWithParam,
                        ::testing::Values(std::make_tuple(1, true),
                                          std::make_tuple(1, false),
@@ -3095,6 +3481,8 @@ int main(int argc, char** argv) {
  ::testing::InitGoogleTest(&argc, argv);
  return RUN_ALL_TESTS();
 #else
+  (void) argc;
+  (void) argv;
  return 0;
 #endif
 }
@@ -501,6 +501,8 @@ int main(int argc, char** argv) {
  ::testing::InitGoogleTest(&argc, argv);
  return RUN_ALL_TESTS();
 #else
+  (void) argc;
+  (void) argv;
  return 0;
 #endif
 }
@@ -57,6 +57,7 @@ Status DBImpl::EnableFileDeletions(bool force) {
      ROCKS_LOG_INFO(immutable_db_options_.info_log, "File Deletions Enabled");
      should_purge_files = true;
      FindObsoleteFiles(&job_context, true);
+      bg_cv_.SignalAll();
    } else {
      ROCKS_LOG_WARN(
          immutable_db_options_.info_log,
@@ -92,7 +93,7 @@ Status DBImpl::GetLiveFiles(std::vector<std::string>& ret,
      }
      cfd->Ref();
      mutex_.Unlock();
-      status = FlushMemTable(cfd, FlushOptions());
+      status = FlushMemTable(cfd, FlushOptions(), FlushReason::kGetLiveFiles);
      TEST_SYNC_POINT("DBImpl::GetLiveFiles:1");
      TEST_SYNC_POINT("DBImpl::GetLiveFiles:2");
      mutex_.Lock();
@@ -141,6 +142,18 @@ Status DBImpl::GetLiveFiles(std::vector<std::string>& ret,
 }

 Status DBImpl::GetSortedWalFiles(VectorLogPtr& files) {
+  {
+    // If caller disabled deletions, this function should return files that are
+    // guaranteed not to be deleted until deletions are re-enabled. We need to
+    // wait for pending purges to finish since WalManager doesn't know which
+    // files are going to be purged. Additional purges won't be scheduled as
+    // long as deletions are disabled (so the below loop must terminate).
+    InstrumentedMutexLock l(&mutex_);
+    while (disable_delete_obsolete_files_ > 0 &&
+           pending_purge_obsolete_files_ > 0) {
+      bg_cv_.Wait();
+    }
+  }
  return wal_manager_.GetSortedWalFiles(files);
 }

@@ -126,6 +126,41 @@ TEST_F(DBFlushTest, FlushInLowPriThreadPool) {
  ASSERT_EQ(1, num_compactions);
 }

+TEST_F(DBFlushTest, ManualFlushWithMinWriteBufferNumberToMerge) {
+  Options options = CurrentOptions();
+  options.write_buffer_size = 100;
+  options.max_write_buffer_number = 4;
+  options.min_write_buffer_number_to_merge = 3;
+  Reopen(options);
+
+  SyncPoint::GetInstance()->LoadDependency(
+      {{"DBImpl::BGWorkFlush",
+        "DBFlushTest::ManualFlushWithMinWriteBufferNumberToMerge:1"},
+       {"DBFlushTest::ManualFlushWithMinWriteBufferNumberToMerge:2",
+        "FlushJob::WriteLevel0Table"}});
+  SyncPoint::GetInstance()->EnableProcessing();
+
+  ASSERT_OK(Put("key1", "value1"));
+
+  port::Thread t([&]() {
+    // The call wait for flush to finish, i.e. with flush_options.wait = true.
+    ASSERT_OK(Flush());
+  });
+
+  // Wait for flush start.
+  TEST_SYNC_POINT("DBFlushTest::ManualFlushWithMinWriteBufferNumberToMerge:1");
+  // Insert a second memtable before the manual flush finish.
+  // At the end of the manual flush job, it will check if further flush
+  // is needed, but it will not trigger flush of the second memtable because
+  // min_write_buffer_number_to_merge is not reached.
+  ASSERT_OK(Put("key2", "value2"));
+  ASSERT_OK(dbfull()->TEST_SwitchMemtable());
+  TEST_SYNC_POINT("DBFlushTest::ManualFlushWithMinWriteBufferNumberToMerge:2");
+
+  // Manual flush should return, without waiting for flush indefinitely.
+  t.join();
+}
+
 TEST_P(DBFlushDirectIOTest, DirectIO) {
  Options options;
  options.create_if_missing = true;
@@ -150,6 +185,26 @@ TEST_P(DBFlushDirectIOTest, DirectIO) {
  delete options.env;
 }

+TEST_F(DBFlushTest, FlushError) {
+  Options options;
+  std::unique_ptr<FaultInjectionTestEnv> fault_injection_env(
+      new FaultInjectionTestEnv(env_));
+  options.write_buffer_size = 100;
+  options.max_write_buffer_number = 4;
+  options.min_write_buffer_number_to_merge = 3;
+  options.disable_auto_compactions = true;
+  options.env = fault_injection_env.get();
+  Reopen(options);
+
+  ASSERT_OK(Put("key1", "value1"));
+  ASSERT_OK(Put("key2", "value2"));
+  fault_injection_env->SetFilesystemActive(false);
+  Status s = dbfull()->TEST_SwitchMemtable();
+  fault_injection_env->SetFilesystemActive(true);
+  Destroy(options);
+  ASSERT_NE(s, Status::OK());
+}
+
 INSTANTIATE_TEST_CASE_P(DBFlushDirectIOTest, DBFlushDirectIOTest,
                        testing::Bool());

@@ -123,13 +123,13 @@ CompressionType GetCompressionFlush(
 namespace {
 void DumpSupportInfo(Logger* logger) {
  ROCKS_LOG_HEADER(logger, "Compression algorithms supported:");
-  ROCKS_LOG_HEADER(logger, "\tSnappy supported: %d", Snappy_Supported());
-  ROCKS_LOG_HEADER(logger, "\tZlib supported: %d", Zlib_Supported());
-  ROCKS_LOG_HEADER(logger, "\tBzip supported: %d", BZip2_Supported());
-  ROCKS_LOG_HEADER(logger, "\tLZ4 supported: %d", LZ4_Supported());
-  ROCKS_LOG_HEADER(logger, "\tZSTDNotFinal supported: %d",
-                   ZSTDNotFinal_Supported());
-  ROCKS_LOG_HEADER(logger, "\tZSTD supported: %d", ZSTD_Supported());
+  for (auto& compression : OptionsHelper::compression_type_string_map) {
+    if (compression.second != kNoCompression &&
+        compression.second != kDisableCompressionOption) {
+      ROCKS_LOG_HEADER(logger, "\t%s supported: %d", compression.first.c_str(),
+                       CompressionTypeSupported(compression.second));
+    }
+  }
  ROCKS_LOG_HEADER(logger, "Fast CRC32 supported: %s",
                   crc32c::IsFastCrc32Supported().c_str());
 }
@@ -141,6 +141,7 @@ DBImpl::DBImpl(const DBOptions& options, const std::string& dbname,
               const bool seq_per_batch)
    : env_(options.env),
      dbname_(dbname),
+      own_info_log_(options.info_log == nullptr),
      initial_db_options_(SanitizeOptions(dbname, options)),
      immutable_db_options_(initial_db_options_),
      mutable_db_options_(initial_db_options_),
@@ -177,6 +178,7 @@ DBImpl::DBImpl(const DBOptions& options, const std::string& dbname,
      num_running_flushes_(0),
      bg_purge_scheduled_(0),
      disable_delete_obsolete_files_(0),
+      pending_purge_obsolete_files_(0),
      delete_obsolete_files_last_run_(env_->NowMicros()),
      last_stats_dump_time_microsec_(0),
      next_job_id_(1),
@@ -212,7 +214,8 @@ DBImpl::DBImpl(const DBOptions& options, const std::string& dbname,
      // requires a custom gc for compaction, we use that to set use_custom_gc_
      // as well.
      use_custom_gc_(seq_per_batch),
-      preserve_deletes_(options.preserve_deletes) {
+      preserve_deletes_(options.preserve_deletes),
+      closed_(false) {
  env_->GetAbsolutePath(dbname, &db_absolute_path_);

  // Reserve ten files or so for other uses and give the rest to TableCache.
@@ -255,7 +258,7 @@ void DBImpl::CancelAllBackgroundWork(bool wait) {
      if (!cfd->IsDropped() && cfd->initialized() && !cfd->mem()->IsEmpty()) {
        cfd->Ref();
        mutex_.Unlock();
-        FlushMemTable(cfd, FlushOptions());
+        FlushMemTable(cfd, FlushOptions(), FlushReason::kShutDown);
        mutex_.Lock();
        cfd->Unref();
      }
@@ -275,7 +278,7 @@ void DBImpl::CancelAllBackgroundWork(bool wait) {
  }
 }

-DBImpl::~DBImpl() {
+Status DBImpl::CloseHelper() {
  // CancelAllBackgroundWork called with false means we just set the shutdown
  // marker. After this we do a variant of the waiting and unschedule work
  // (to consider: moving all the waiting into CancelAllBackgroundWork(true))
@@ -284,6 +287,7 @@ DBImpl::~DBImpl() {
      env_->UnSchedule(this, Env::Priority::BOTTOM);
  int compactions_unscheduled = env_->UnSchedule(this, Env::Priority::LOW);
  int flushes_unscheduled = env_->UnSchedule(this, Env::Priority::HIGH);
+  Status ret;
  mutex_.Lock();
  bg_bottom_compaction_scheduled_ -= bottom_compactions_unscheduled;
  bg_compaction_scheduled_ -= compactions_unscheduled;
@@ -291,7 +295,8 @@ DBImpl::~DBImpl() {

  // Wait for background work to finish
  while (bg_bottom_compaction_scheduled_ || bg_compaction_scheduled_ ||
-         bg_flush_scheduled_ || bg_purge_scheduled_) {
+         bg_flush_scheduled_ || bg_purge_scheduled_ ||
+         pending_purge_obsolete_files_) {
    TEST_SYNC_POINT("DBImpl::~DBImpl:WaitJob");
    bg_cv_.Wait();
  }
@@ -345,7 +350,18 @@ DBImpl::~DBImpl() {
    delete l;
  }
  for (auto& log : logs_) {
-    log.ClearWriter();
+    uint64_t log_number = log.writer->get_log_number();
+    Status s = log.ClearWriter();
+    if (!s.ok()) {
+      ROCKS_LOG_WARN(immutable_db_options_.info_log,
+                     "Unable to Sync WAL file %s with error -- %s",
+                     LogFileName(immutable_db_options_.wal_dir, log_number).c_str(),
+                     s.ToString().c_str());
+      // Retain the first error
+      if (ret.ok()) {
+        ret = s;
+      }
+    }
  }
  logs_.clear();

@@ -378,6 +394,25 @@ DBImpl::~DBImpl() {

  ROCKS_LOG_INFO(immutable_db_options_.info_log, "Shutdown complete");
  LogFlush(immutable_db_options_.info_log);
+
+  if (immutable_db_options_.info_log && own_info_log_) {
+    Status s = immutable_db_options_.info_log->Close();
+    if (ret.ok()) {
+      ret = s;
+    }
+  }
+  return ret;
+}
+
+Status DBImpl::CloseImpl() {
+  return CloseHelper();
+}
+
+DBImpl::~DBImpl() {
+  if (!closed_) {
+    closed_ = true;
+    CloseHelper();
+  }
 }

 void DBImpl::MaybeIgnoreError(Status* s) const {
@@ -523,6 +558,7 @@ Status DBImpl::SetOptions(ColumnFamilyHandle* column_family,

      persist_options_status = WriteOptionsFile(
          false /*need_mutex_lock*/, true /*need_enter_write_thread*/);
+      bg_cv_.SignalAll();
    }
  }
  sv_context.Clean();
@@ -1407,6 +1443,7 @@ Status DBImpl::DropColumnFamilyImpl(ColumnFamilyHandle* column_family) {
      }
      is_snapshot_supported_ = new_is_snapshot_supported;
    }
+    bg_cv_.SignalAll();
  }

  if (s.ok()) {
@@ -2047,8 +2084,7 @@ Status DBImpl::DeleteFile(std::string name) {
                      name.c_str());
      return Status::NotSupported("Delete only supported for archived logs");
    }
-    status =
-        env_->DeleteFile(immutable_db_options_.wal_dir + "/" + name.c_str());
+    status = wal_manager_.DeleteFile(name, number);
    if (!status.ok()) {
      ROCKS_LOG_ERROR(immutable_db_options_.info_log,
                      "DeleteFile %s failed -- %s.\n", name.c_str(),
@@ -2127,52 +2163,63 @@ Status DBImpl::DeleteFile(std::string name) {
  return status;
 }

-Status DBImpl::DeleteFilesInRange(ColumnFamilyHandle* column_family,
-                                  const Slice* begin, const Slice* end) {
+Status DBImpl::DeleteFilesInRanges(ColumnFamilyHandle* column_family,
+                                   const RangePtr* ranges, size_t n,
+                                   bool include_end) {
  Status status;
  auto cfh = reinterpret_cast<ColumnFamilyHandleImpl*>(column_family);
  ColumnFamilyData* cfd = cfh->cfd();
  VersionEdit edit;
-  std::vector<FileMetaData*> deleted_files;
+  std::set<FileMetaData*> deleted_files;
  JobContext job_context(next_job_id_.fetch_add(1), true);
  {
    InstrumentedMutexLock l(&mutex_);
    Version* input_version = cfd->current();

    auto* vstorage = input_version->storage_info();
-    for (int i = 1; i < cfd->NumberLevels(); i++) {
-      if (vstorage->LevelFiles(i).empty() ||
-          !vstorage->OverlapInLevel(i, begin, end)) {
-        continue;
-      }
-      std::vector<FileMetaData*> level_files;
-      InternalKey begin_storage, end_storage, *begin_key, *end_key;
-      if (begin == nullptr) {
-        begin_key = nullptr;
-      } else {
-        begin_storage.SetMinPossibleForUserKey(*begin);
-        begin_key = &begin_storage;
-      }
-      if (end == nullptr) {
-        end_key = nullptr;
-      } else {
-        end_storage.SetMaxPossibleForUserKey(*end);
-        end_key = &end_storage;
-      }
-
-      vstorage->GetCleanInputsWithinInterval(i, begin_key, end_key,
-                                             &level_files, -1 /* hint_index */,
-                                             nullptr /* file_index */);
-      FileMetaData* level_file;
-      for (uint32_t j = 0; j < level_files.size(); j++) {
-        level_file = level_files[j];
-        if (level_file->being_compacted) {
+    for (size_t r = 0; r < n; r++) {
+      auto begin = ranges[r].start, end = ranges[r].limit;
+      for (int i = 1; i < cfd->NumberLevels(); i++) {
+        if (vstorage->LevelFiles(i).empty() ||
+            !vstorage->OverlapInLevel(i, begin, end)) {
          continue;
        }
-        edit.SetColumnFamily(cfd->GetID());
-        edit.DeleteFile(i, level_file->fd.GetNumber());
-        deleted_files.push_back(level_file);
-        level_file->being_compacted = true;
+        std::vector<FileMetaData*> level_files;
+        InternalKey begin_storage, end_storage, *begin_key, *end_key;
+        if (begin == nullptr) {
+          begin_key = nullptr;
+        } else {
+          begin_storage.SetMinPossibleForUserKey(*begin);
+          begin_key = &begin_storage;
+        }
+        if (end == nullptr) {
+          end_key = nullptr;
+        } else {
+          end_storage.SetMaxPossibleForUserKey(*end);
+          end_key = &end_storage;
+        }
+
+        vstorage->GetCleanInputsWithinInterval(i, begin_key, end_key,
+                                               &level_files, -1 /* hint_index */,
+                                               nullptr /* file_index */);
+        FileMetaData* level_file;
+        for (uint32_t j = 0; j < level_files.size(); j++) {
+          level_file = level_files[j];
+          if (level_file->being_compacted) {
+            continue;
+          }
+          if (deleted_files.find(level_file) != deleted_files.end()) {
+            continue;
+          }
+          if (!include_end && end != nullptr &&
+              cfd->user_comparator()->Compare(level_file->largest.user_key(), *end) == 0) {
+            continue;
+          }
+          edit.SetColumnFamily(cfd->GetID());
+          edit.DeleteFile(i, level_file->fd.GetNumber());
+          deleted_files.insert(level_file);
+          level_file->being_compacted = true;
+        }
      }
    }
    if (edit.GetDeletedFiles().empty()) {
@@ -2321,7 +2368,15 @@ Status DB::DestroyColumnFamilyHandle(ColumnFamilyHandle* column_family) {
  return Status::OK();
 }

-DB::~DB() { }
+DB::~DB() {}
+
+Status DBImpl::Close() {
+  if (!closed_) {
+    closed_ = true;
+    return CloseImpl();
+  }
+  return Status::OK();
+}

 Status DB::ListColumnFamilies(const DBOptions& db_options,
                              const std::string& name,
@@ -2777,12 +2832,14 @@ Status DBImpl::IngestExternalFile(
    // Figure out if we need to flush the memtable first
    if (status.ok()) {
      bool need_flush = false;
-      status = ingestion_job.NeedsFlush(&need_flush);
+      status = ingestion_job.NeedsFlush(&need_flush, cfd->GetSuperVersion());
      TEST_SYNC_POINT_CALLBACK("DBImpl::IngestExternalFile:NeedFlush",
                               &need_flush);
      if (status.ok() && need_flush) {
        mutex_.Unlock();
-        status = FlushMemTable(cfd, FlushOptions(), true /* writes_stopped */);
+        status = FlushMemTable(cfd, FlushOptions(),
+                               FlushReason::kExternalFileIngestion,
+                               true /* writes_stopped */);
        mutex_.Lock();
      }
    }
@@ -247,8 +247,9 @@ class DBImpl : public DB {
      const TransactionLogIterator::ReadOptions&
          read_options = TransactionLogIterator::ReadOptions()) override;
  virtual Status DeleteFile(std::string name) override;
-  Status DeleteFilesInRange(ColumnFamilyHandle* column_family,
-                            const Slice* begin, const Slice* end);
+  Status DeleteFilesInRanges(ColumnFamilyHandle* column_family,
+                             const RangePtr* ranges, size_t n,
+                             bool include_end = true);

  virtual void GetLiveFilesMetaData(
      std::vector<LiveFileMetaData>* metadata) override;
@@ -468,6 +469,8 @@ class DBImpl : public DB {
  // belong to live files are posibly removed. Also, removes all the
  // files in sst_delete_files and log_delete_files.
  // It is not necessary to hold the mutex when invoking this method.
+  // If FindObsoleteFiles() was run, we need to also run
+  // PurgeObsoleteFiles(), even if disable_delete_obsolete_files_ is true
  void PurgeObsoleteFiles(const JobContext& background_contet,
                          bool schedule_only = false);

@@ -611,10 +614,14 @@ class DBImpl : public DB {
                     std::vector<ColumnFamilyHandle*>* handles, DB** dbptr,
                     const bool seq_per_batch);

+  virtual Status Close() override;
+
 protected:
  Env* const env_;
  const std::string dbname_;
  unique_ptr<VersionSet> versions_;
+  // Flag to check whether we allocated and own the info log file
+  bool own_info_log_;
  const DBOptions initial_db_options_;
  const ImmutableDBOptions immutable_db_options_;
  MutableDBOptions mutable_db_options_;
@@ -666,10 +673,14 @@ class DBImpl : public DB {
  // batch which will be written to memtable later during the commit, and in
  // WritePrepared it is guaranteed since it will be used only for WAL markers
  // which will never be written to memtable.
+  // batch_cnt is expected to be non-zero in seq_per_batch mode and indicates
+  // the number of sub-patches. A sub-patch is a subset of the write batch that
+  // does not have duplicate keys.
  Status WriteImpl(const WriteOptions& options, WriteBatch* updates,
                   WriteCallback* callback = nullptr,
                   uint64_t* log_used = nullptr, uint64_t log_ref = 0,
                   bool disable_memtable = false, uint64_t* seq_used = nullptr,
+                   size_t batch_cnt = 0,
                   PreReleaseCallback* pre_release_callback = nullptr);

  Status PipelinedWriteImpl(const WriteOptions& options, WriteBatch* updates,
@@ -678,10 +689,13 @@ class DBImpl : public DB {
                            bool disable_memtable = false,
                            uint64_t* seq_used = nullptr);

+  // batch_cnt is expected to be non-zero in seq_per_batch mode and indicates
+  // the number of sub-patches. A sub-patch is a subset of the write batch that
+  // does not have duplicate keys.
  Status WriteImplWALOnly(const WriteOptions& options, WriteBatch* updates,
                          WriteCallback* callback = nullptr,
                          uint64_t* log_used = nullptr, uint64_t log_ref = 0,
-                          uint64_t* seq_used = nullptr,
+                          uint64_t* seq_used = nullptr, size_t batch_cnt = 0,
                          PreReleaseCallback* pre_release_callback = nullptr);

  uint64_t FindMinLogContainingOutstandingPrep();
@@ -690,6 +704,9 @@ class DBImpl : public DB {
  // The writer must be the leader in write_thread_ and holding mutex_
  Status WriteRecoverableState();

+  // Actual implementation of Close()
+  Status CloseImpl();
+
 private:
  friend class DB;
  friend class InternalStats;
@@ -805,10 +822,14 @@ class DBImpl : public DB {

  // Force current memtable contents to be flushed.
  Status FlushMemTable(ColumnFamilyData* cfd, const FlushOptions& options,
-                       bool writes_stopped = false);
+                       FlushReason flush_reason, bool writes_stopped = false);

-  // Wait for memtable flushed
-  Status WaitForFlushMemTable(ColumnFamilyData* cfd);
+  // Wait for memtable flushed.
+  // If flush_memtable_id is non-null, wait until the memtable with the ID
+  // gets flush. Otherwise, wait until the column family don't have any
+  // memtable pending flush.
+  Status WaitForFlushMemTable(ColumnFamilyData* cfd,
+                              const uint64_t* flush_memtable_id = nullptr);

  // REQUIRES: mutex locked
  Status SwitchWAL(WriteContext* write_context);
@@ -863,7 +884,7 @@ class DBImpl : public DB {
  ColumnFamilyData* GetColumnFamilyDataByName(const std::string& cf_name);

  void MaybeScheduleFlushOrCompaction();
-  void SchedulePendingFlush(ColumnFamilyData* cfd);
+  void SchedulePendingFlush(ColumnFamilyData* cfd, FlushReason flush_reason);
  void SchedulePendingCompaction(ColumnFamilyData* cfd);
  void SchedulePendingPurge(std::string fname, FileType type, uint64_t number,
                            uint32_t path_id, int job_id);
@@ -902,7 +923,7 @@ class DBImpl : public DB {
  // helper functions for adding and removing from flush & compaction queues
  void AddToCompactionQueue(ColumnFamilyData* cfd);
  ColumnFamilyData* PopFirstFromCompactionQueue();
-  void AddToFlushQueue(ColumnFamilyData* cfd);
+  void AddToFlushQueue(ColumnFamilyData* cfd, FlushReason flush_reason);
  ColumnFamilyData* PopFirstFromFlushQueue();

  // helper function to call after some of the logs_ were synced
@@ -912,6 +933,8 @@ class DBImpl : public DB {

  uint64_t GetMaxTotalWalSize() const;

+  Status CloseHelper();
+
  // table_cache_ provides its own synchronization
  std::shared_ptr<Cache> table_cache_;

@@ -940,6 +963,10 @@ class DBImpl : public DB {
  // (i.e. whenever a flush is done, even if it didn't make any progress)
  // * whenever there is an error in background purge, flush or compaction
  // * whenever num_running_ingest_file_ goes to 0.
+  // * whenever pending_purge_obsolete_files_ goes to 0.
+  // * whenever disable_delete_obsolete_files_ goes to 0.
+  // * whenever SetOptions successfully updates options.
+  // * whenever a column family is dropped.
  InstrumentedCondVar bg_cv_;
  // Writes are protected by locking both mutex_ and log_write_mutex_, and reads
  // must be under either mutex_ or log_write_mutex_. Since after ::Open,
@@ -977,9 +1004,11 @@ class DBImpl : public DB {
      writer = nullptr;
      return w;
    }
-    void ClearWriter() {
+    Status ClearWriter() {
+      Status s = writer->WriteBuffer();
      delete writer;
      writer = nullptr;
+      return s;
    }

    uint64_t number;
@@ -1206,6 +1235,10 @@ class DBImpl : public DB {
  // without any synchronization
  int disable_delete_obsolete_files_;

+  // Number of times FindObsoleteFiles has found deletable files and the
+  // corresponding call to PurgeObsoleteFiles has not yet finished.
+  int pending_purge_obsolete_files_;
+
  // last time when DeleteObsoleteFiles with full scan was executed. Originaly
  // initialized with startup time.
  uint64_t delete_obsolete_files_last_run_;
@@ -1334,7 +1367,7 @@ class DBImpl : public DB {
    return Env::WLTH_SHORT;
  }

-  // When set, we use a seprate queue for writes that dont write to memtable. In
+  // When set, we use a separate queue for writes that dont write to memtable. In
  // 2PC these are the writes at Prepare phase.
  const bool two_write_queues_;
  const bool manual_wal_flush_;
@@ -1359,6 +1392,9 @@ class DBImpl : public DB {
  // is set to false.
  std::atomic<SequenceNumber> preserve_deletes_seqnum_;
  const bool preserve_deletes_;
+
+  // Flag to check whether Close() has been called on this DB
+  bool closed_;
 };

 extern Options SanitizeOptions(const std::string& db,
@@ -23,6 +23,7 @@
 #include "util/sync_point.h"

 namespace rocksdb {
+
 Status DBImpl::SyncClosedLogs(JobContext* job_context) {
  TEST_SYNC_POINT("DBImpl::SyncClosedLogs:Start");
  mutex_.AssertHeld();
@@ -222,6 +223,7 @@ void DBImpl::NotifyOnFlushBegin(ColumnFamilyData* cfd, FileMetaData* file_meta,
    info.smallest_seqno = file_meta->smallest_seqno;
    info.largest_seqno = file_meta->largest_seqno;
    info.table_properties = prop;
+    info.flush_reason = cfd->GetFlushReason();
    for (auto listener : immutable_db_options_.listeners) {
      listener->OnFlushBegin(this, info);
    }
@@ -266,6 +268,7 @@ void DBImpl::NotifyOnFlushCompleted(ColumnFamilyData* cfd,
    info.smallest_seqno = file_meta->smallest_seqno;
    info.largest_seqno = file_meta->largest_seqno;
    info.table_properties = prop;
+    info.flush_reason = cfd->GetFlushReason();
    for (auto listener : immutable_db_options_.listeners) {
      listener->OnFlushCompleted(this, info);
    }
@@ -287,10 +290,75 @@ Status DBImpl::CompactRange(const CompactRangeOptions& options,
  auto cfd = cfh->cfd();
  bool exclusive = options.exclusive_manual_compaction;

-  Status s = FlushMemTable(cfd, FlushOptions());
-  if (!s.ok()) {
-    LogFlush(immutable_db_options_.info_log);
-    return s;
+  bool flush_needed = true;
+  if (begin != nullptr && end != nullptr) {
+    // TODO(ajkr): We could also optimize away the flush in certain cases where
+    // one/both sides of the interval are unbounded. But it requires more
+    // changes to RangesOverlapWithMemtables.
+    Range range(*begin, *end);
+    SuperVersion* super_version = cfd->GetReferencedSuperVersion(&mutex_);
+    cfd->RangesOverlapWithMemtables({range}, super_version, &flush_needed);
+    CleanupSuperVersion(super_version);
+  }
+
+  if (!options.allow_write_stall && flush_needed) {
+    InstrumentedMutexLock l(&mutex_);
+    uint64_t orig_active_memtable_id = cfd->mem()->GetID();
+    WriteStallCondition write_stall_condition = WriteStallCondition::kNormal;
+    do {
+      if (write_stall_condition != WriteStallCondition::kNormal) {
+        TEST_SYNC_POINT("DBImpl::CompactRange:StallWait");
+        ROCKS_LOG_INFO(immutable_db_options_.info_log,
+                       "[%s] CompactRange waiting on stall conditions to clear",
+                       cfd->GetName().c_str());
+        bg_cv_.Wait();
+      }
+      if (cfd->IsDropped() || shutting_down_.load(std::memory_order_acquire)) {
+        return Status::ShutdownInProgress();
+      }
+
+      uint64_t earliest_memtable_id =
+          std::min(cfd->mem()->GetID(), cfd->imm()->GetEarliestMemTableID());
+      if (earliest_memtable_id > orig_active_memtable_id) {
+        // We waited so long that the memtable we were originally waiting on was
+        // flushed.
+        flush_needed = false;
+        break;
+      }
+
+      const auto& mutable_cf_options = *cfd->GetLatestMutableCFOptions();
+      const auto* vstorage = cfd->current()->storage_info();
+
+      // Skip stalling check if we're below auto-flush and auto-compaction
+      // triggers. If it stalled in these conditions, that'd mean the stall
+      // triggers are so low that stalling is needed for any background work. In
+      // that case we shouldn't wait since background work won't be scheduled.
+      if (cfd->imm()->NumNotFlushed() <
+              cfd->ioptions()->min_write_buffer_number_to_merge &&
+          vstorage->l0_delay_trigger_count() <
+              mutable_cf_options.level0_file_num_compaction_trigger) {
+        break;
+      }
+
+      // check whether one extra immutable memtable or an extra L0 file would
+      // cause write stalling mode to be entered. It could still enter stall
+      // mode due to pending compaction bytes, but that's less common
+      write_stall_condition =
+          ColumnFamilyData::GetWriteStallConditionAndCause(
+              cfd->imm()->NumNotFlushed() + 1,
+              vstorage->l0_delay_trigger_count() + 1,
+              vstorage->estimated_compaction_needed_bytes(), mutable_cf_options)
+              .first;
+    } while (write_stall_condition != WriteStallCondition::kNormal);
+  }
+  TEST_SYNC_POINT("DBImpl::CompactRange:StallWaitDone");
+  Status s;
+  if (flush_needed) {
+    s = FlushMemTable(cfd, FlushOptions(), FlushReason::kManualCompaction);
+    if (!s.ok()) {
+      LogFlush(immutable_db_options_.info_log);
+      return s;
+    }
  }

  int max_level_with_files = 0;
@@ -441,7 +509,8 @@ Status DBImpl::CompactFiles(
  }  // release the mutex

  // delete unnecessary files if any, this is done outside the mutex
-  if (job_context.HaveSomethingToDelete() || !log_buffer.IsEmpty()) {
+  if (job_context.HaveSomethingToClean() ||
+      job_context.HaveSomethingToDelete() || !log_buffer.IsEmpty()) {
    // Have to flush the info logs before bg_compaction_scheduled_--
    // because if bg_flush_scheduled_ becomes 0 and the lock is
    // released, the deconstructor of DB can kick in and destroy all the
@@ -809,7 +878,14 @@ int DBImpl::Level0StopWriteTrigger(ColumnFamilyHandle* column_family) {
 Status DBImpl::Flush(const FlushOptions& flush_options,
                     ColumnFamilyHandle* column_family) {
  auto cfh = reinterpret_cast<ColumnFamilyHandleImpl*>(column_family);
-  return FlushMemTable(cfh->cfd(), flush_options);
+  ROCKS_LOG_INFO(immutable_db_options_.info_log, "[%s] Manual flush start.",
+                 cfh->GetName().c_str());
+  Status s =
+      FlushMemTable(cfh->cfd(), flush_options, FlushReason::kManualCompaction);
+  ROCKS_LOG_INFO(immutable_db_options_.info_log,
+                 "[%s] Manual flush finished, status: %s\n",
+                 cfh->GetName().c_str(), s.ToString().c_str());
+  return s;
 }

 Status DBImpl::RunManualCompaction(ColumnFamilyData* cfd, int input_level,
@@ -942,8 +1018,9 @@ Status DBImpl::RunManualCompaction(ColumnFamilyData* cfd, int input_level,

 Status DBImpl::FlushMemTable(ColumnFamilyData* cfd,
                             const FlushOptions& flush_options,
-                             bool writes_stopped) {
+                             FlushReason flush_reason, bool writes_stopped) {
  Status s;
+  uint64_t flush_memtable_id = 0;
  {
    WriteContext context;
    InstrumentedMutexLock guard_lock(&mutex_);
@@ -961,6 +1038,7 @@ Status DBImpl::FlushMemTable(ColumnFamilyData* cfd,

    // SwitchMemtable() will release and reacquire mutex during execution
    s = SwitchMemtable(cfd, &context);
+    flush_memtable_id = cfd->imm()->GetLatestMemTableID();

    if (!writes_stopped) {
      write_thread_.ExitUnbatched(&w);
@@ -969,22 +1047,25 @@ Status DBImpl::FlushMemTable(ColumnFamilyData* cfd,
    cfd->imm()->FlushRequested();

    // schedule flush
-    SchedulePendingFlush(cfd);
+    SchedulePendingFlush(cfd, flush_reason);
    MaybeScheduleFlushOrCompaction();
  }

  if (s.ok() && flush_options.wait) {
    // Wait until the compaction completes
-    s = WaitForFlushMemTable(cfd);
+    s = WaitForFlushMemTable(cfd, &flush_memtable_id);
  }
  return s;
 }

-Status DBImpl::WaitForFlushMemTable(ColumnFamilyData* cfd) {
+Status DBImpl::WaitForFlushMemTable(ColumnFamilyData* cfd,
+                                    const uint64_t* flush_memtable_id) {
  Status s;
  // Wait until the compaction completes
  InstrumentedMutexLock l(&mutex_);
-  while (cfd->imm()->NumNotFlushed() > 0 && bg_error_.ok()) {
+  while (cfd->imm()->NumNotFlushed() > 0 && bg_error_.ok() &&
+         (flush_memtable_id == nullptr ||
+          cfd->imm()->GetEarliestMemTableID() <= *flush_memtable_id)) {
    if (shutting_down_.load(std::memory_order_acquire)) {
      return Status::ShutdownInProgress();
    }
@@ -1059,6 +1140,7 @@ void DBImpl::MaybeScheduleFlushOrCompaction() {
  if (HasExclusiveManualCompaction()) {
    // only manual compactions are allowed to run. don't schedule automatic
    // compactions
+    TEST_SYNC_POINT("DBImpl::MaybeScheduleFlushOrCompaction:Conflict");
    return;
  }

@@ -1121,11 +1203,12 @@ ColumnFamilyData* DBImpl::PopFirstFromCompactionQueue() {
  return cfd;
 }

-void DBImpl::AddToFlushQueue(ColumnFamilyData* cfd) {
+void DBImpl::AddToFlushQueue(ColumnFamilyData* cfd, FlushReason flush_reason) {
  assert(!cfd->pending_flush());
  cfd->Ref();
  flush_queue_.push_back(cfd);
  cfd->set_pending_flush(true);
+  cfd->SetFlushReason(flush_reason);
 }

 ColumnFamilyData* DBImpl::PopFirstFromFlushQueue() {
@@ -1134,12 +1217,14 @@ ColumnFamilyData* DBImpl::PopFirstFromFlushQueue() {
  flush_queue_.pop_front();
  assert(cfd->pending_flush());
  cfd->set_pending_flush(false);
+  // TODO: need to unset flush reason?
  return cfd;
 }

-void DBImpl::SchedulePendingFlush(ColumnFamilyData* cfd) {
+void DBImpl::SchedulePendingFlush(ColumnFamilyData* cfd,
+                                  FlushReason flush_reason) {
  if (!cfd->pending_flush() && cfd->imm()->IsFlushPending()) {
-    AddToFlushQueue(cfd);
+    AddToFlushQueue(cfd, flush_reason);
    ++unscheduled_flushes_;
  }
 }
@@ -1303,8 +1388,10 @@ void DBImpl::BackgroundCallFlush() {
    // created. Thus, we force full scan in FindObsoleteFiles()
    FindObsoleteFiles(&job_context, !s.ok() && !s.IsShutdownInProgress());
    // delete unnecessary files if any, this is done outside the mutex
-    if (job_context.HaveSomethingToDelete() || !log_buffer.IsEmpty()) {
+    if (job_context.HaveSomethingToClean() ||
+        job_context.HaveSomethingToDelete() || !log_buffer.IsEmpty()) {
      mutex_.Unlock();
+      TEST_SYNC_POINT("DBImpl::BackgroundCallFlush:FilesFound");
      // Have to flush the info logs before bg_flush_scheduled_--
      // because if bg_flush_scheduled_ becomes 0 and the lock is
      // released, the deconstructor of DB can kick in and destroy all the
@@ -1384,7 +1471,8 @@ void DBImpl::BackgroundCallCompaction(PrepickedCompaction* prepicked_compaction,
    FindObsoleteFiles(&job_context, !s.ok() && !s.IsShutdownInProgress());

    // delete unnecessary files if any, this is done outside the mutex
-    if (job_context.HaveSomethingToDelete() || !log_buffer.IsEmpty()) {
+    if (job_context.HaveSomethingToClean() ||
+        job_context.HaveSomethingToDelete() || !log_buffer.IsEmpty()) {
      mutex_.Unlock();
      // Have to flush the info logs before bg_compaction_scheduled_--
      // because if bg_flush_scheduled_ becomes 0 and the lock is
@@ -1503,7 +1591,7 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
               : m->manual_end->DebugString().c_str()));
    }
  } else if (!is_prepicked && !compaction_queue_.empty()) {
-    if (HaveManualCompaction(compaction_queue_.front())) {
+    if (HasExclusiveManualCompaction()) {
      // Can't compact right now, but try again later
      TEST_SYNC_POINT("DBImpl::BackgroundCompaction()::Conflict");

@@ -1679,7 +1767,7 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
    env_->Schedule(&DBImpl::BGWorkBottomCompaction, ca, Env::Priority::BOTTOM,
                   this, &DBImpl::UnscheduleCallback);
  } else {
-    int output_level  __attribute__((unused));
+    int output_level  __attribute__((__unused__));
    output_level = c->output_level();
    TEST_SYNC_POINT_CALLBACK("DBImpl::BackgroundCompaction:NonTrivial",
                             &output_level);
@@ -1913,7 +2001,7 @@ void DBImpl::InstallSuperVersionAndScheduleWork(

  // Whenever we install new SuperVersion, we might need to issue new flushes or
  // compactions.
-  SchedulePendingFlush(cfd);
+  SchedulePendingFlush(cfd, FlushReason::kSuperVersionChange);
  SchedulePendingCompaction(cfd);
  MaybeScheduleFlushOrCompaction();

@@ -103,7 +103,7 @@ Status DBImpl::TEST_FlushMemTable(bool wait, ColumnFamilyHandle* cfh) {
    auto cfhi = reinterpret_cast<ColumnFamilyHandleImpl*>(cfh);
    cfd = cfhi->cfd();
  }
-  return FlushMemTable(cfd, fo);
+  return FlushMemTable(cfd, fo, FlushReason::kTest);
 }

 Status DBImpl::TEST_WaitForFlushMemTable(ColumnFamilyHandle* column_family) {
@@ -285,6 +285,9 @@ void DBImpl::FindObsoleteFiles(JobContext* job_context, bool force,
  job_context->logs_to_free = logs_to_free_;
  job_context->log_recycle_files.assign(log_recycle_files_.begin(),
                                        log_recycle_files_.end());
+  if (job_context->HaveSomethingToDelete()) {
+    ++pending_purge_obsolete_files_;
+  }
  logs_to_free_.clear();
 }

@@ -342,15 +345,12 @@ void DBImpl::DeleteObsoleteFileImpl(int job_id, const std::string& fname,
 // files in sst_delete_files and log_delete_files.
 // It is not necessary to hold the mutex when invoking this method.
 void DBImpl::PurgeObsoleteFiles(const JobContext& state, bool schedule_only) {
+  TEST_SYNC_POINT("DBImpl::PurgeObsoleteFiles:Begin");
  // we'd better have sth to delete
  assert(state.HaveSomethingToDelete());

-  // this checks if FindObsoleteFiles() was run before. If not, don't do
-  // PurgeObsoleteFiles(). If FindObsoleteFiles() was run, we need to also
-  // run PurgeObsoleteFiles(), even if disable_delete_obsolete_files_ is true
-  if (state.manifest_file_number == 0) {
-    return;
-  }
+  // FindObsoleteFiles() should've populated this so nonzero
+  assert(state.manifest_file_number != 0);

  // Now, convert live list to an unordered map, WITHOUT mutex held;
  // set is slow.
@@ -533,6 +533,13 @@ void DBImpl::PurgeObsoleteFiles(const JobContext& state, bool schedule_only) {
  wal_manager_.PurgeObsoleteWALFiles();
 #endif  // ROCKSDB_LITE
  LogFlush(immutable_db_options_.info_log);
+  InstrumentedMutexLock l(&mutex_);
+  --pending_purge_obsolete_files_;
+  assert(pending_purge_obsolete_files_ >= 0);
+  if (pending_purge_obsolete_files_ == 0) {
+    bg_cv_.SignalAll();
+  }
+  TEST_SYNC_POINT("DBImpl::PurgeObsoleteFiles:End");
 }

 void DBImpl::DeleteObsoleteFiles() {
@@ -356,6 +356,29 @@ Status DBImpl::Recover(
      assert(s.IsIOError());
      return s;
    }
+    // Verify compatibility of env_options_ and filesystem
+    {
+      unique_ptr<RandomAccessFile> idfile;
+      EnvOptions customized_env(env_options_);
+      customized_env.use_direct_reads |=
+          immutable_db_options_.use_direct_io_for_flush_and_compaction;
+      s = env_->NewRandomAccessFile(IdentityFileName(dbname_), &idfile,
+                                    customized_env);
+      if (!s.ok()) {
+        const char* error_msg = s.ToString().c_str();
+        // Check if unsupported Direct I/O is the root cause
+        customized_env.use_direct_reads = false;
+        s = env_->NewRandomAccessFile(IdentityFileName(dbname_), &idfile,
+                                      customized_env);
+        if (s.ok()) {
+          return Status::InvalidArgument(
+              "Direct I/O is not supported by the specified DB.");
+        } else {
+          return Status::InvalidArgument(
+              "Found options incompatible with filesystem", error_msg);
+        }
+      }
+    }
  }

  Status s = versions_->Recover(column_families, read_only);
@@ -64,7 +64,9 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
                         WriteBatch* my_batch, WriteCallback* callback,
                         uint64_t* log_used, uint64_t log_ref,
                         bool disable_memtable, uint64_t* seq_used,
+                         size_t batch_cnt,
                         PreReleaseCallback* pre_release_callback) {
+  assert(!seq_per_batch_ || batch_cnt != 0);
  if (my_batch == nullptr) {
    return Status::Corruption("Batch is nullptr!");
  }
@@ -76,6 +78,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
        "pipelined_writes is not compatible with concurrent prepares");
  }
  if (seq_per_batch_ && immutable_db_options_.enable_pipelined_write) {
+    // TODO(yiwu): update pipeline write with seq_per_batch and batch_cnt
    return Status::NotSupported(
        "pipelined_writes is not compatible with seq_per_batch");
  }
@@ -93,7 +96,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,

  if (two_write_queues_ && disable_memtable) {
    return WriteImplWALOnly(write_options, my_batch, callback, log_used,
-                            log_ref, seq_used, pre_release_callback);
+                            log_ref, seq_used, batch_cnt, pre_release_callback);
  }

  if (immutable_db_options_.enable_pipelined_write) {
@@ -103,7 +106,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,

  PERF_TIMER_GUARD(write_pre_and_post_process_time);
  WriteThread::Writer w(write_options, my_batch, callback, log_ref,
-                        disable_memtable, pre_release_callback);
+                        disable_memtable, batch_cnt, pre_release_callback);

  if (!write_options.disableWAL) {
    RecordTick(stats_, WRITE_WITH_WAL);
@@ -122,7 +125,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
      w.status = WriteBatchInternal::InsertInto(
          &w, w.sequence, &column_family_memtables, &flush_scheduler_,
          write_options.ignore_missing_column_families, 0 /*log_number*/, this,
-          true /*concurrent_memtable_writes*/, seq_per_batch_);
+          true /*concurrent_memtable_writes*/, seq_per_batch_, w.batch_cnt);
    }

    if (write_thread_.CompleteParallelMemTableWriter(&w)) {
@@ -214,7 +217,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
    uint64_t total_byte_size = 0;
    for (auto* writer : write_group) {
      if (writer->CheckCallback(this)) {
-        valid_batches++;
+        valid_batches += writer->batch_cnt;
        if (writer->ShouldWriteToMemtable()) {
          total_count += WriteBatchInternal::Count(writer->batch);
          parallel = parallel && !writer->batch->HasMerge();
@@ -303,7 +306,8 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
          }
          writer->sequence = next_sequence;
          if (seq_per_batch_) {
-            next_sequence++;
+            assert(writer->batch_cnt);
+            next_sequence += writer->batch_cnt;
          } else if (writer->ShouldWriteToMemtable()) {
            next_sequence += WriteBatchInternal::Count(writer->batch);
          }
@@ -323,7 +327,8 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
          w.status = WriteBatchInternal::InsertInto(
              &w, w.sequence, &column_family_memtables, &flush_scheduler_,
              write_options.ignore_missing_column_families, 0 /*log_number*/,
-              this, true /*concurrent_memtable_writes*/, seq_per_batch_);
+              this, true /*concurrent_memtable_writes*/, seq_per_batch_,
+              w.batch_cnt);
        }
      }
      if (seq_used != nullptr) {
@@ -515,12 +520,13 @@ Status DBImpl::PipelinedWriteImpl(const WriteOptions& write_options,
 Status DBImpl::WriteImplWALOnly(const WriteOptions& write_options,
                                WriteBatch* my_batch, WriteCallback* callback,
                                uint64_t* log_used, uint64_t log_ref,
-                                uint64_t* seq_used,
+                                uint64_t* seq_used, size_t batch_cnt,
                                PreReleaseCallback* pre_release_callback) {
  Status status;
  PERF_TIMER_GUARD(write_pre_and_post_process_time);
  WriteThread::Writer w(write_options, my_batch, callback, log_ref,
-                        true /* disable_memtable */, pre_release_callback);
+                        true /* disable_memtable */, batch_cnt,
+                        pre_release_callback);
  RecordTick(stats_, WRITE_WITH_WAL);
  StopWatch write_sw(env_, immutable_db_options_.statistics.get(), DB_WRITE);

@@ -576,7 +582,15 @@ Status DBImpl::WriteImplWALOnly(const WriteOptions& write_options,
  PERF_TIMER_GUARD(write_wal_time);
  // LastAllocatedSequence is increased inside WriteToWAL under
  // wal_write_mutex_ to ensure ordered events in WAL
-  size_t seq_inc = seq_per_batch_ ? write_group.size : 0 /*total_count*/;
+  size_t seq_inc = 0 /* total_count */;
+  if (seq_per_batch_) {
+    size_t total_batch_cnt = 0;
+    for (auto* writer : write_group) {
+      assert(writer->batch_cnt);
+      total_batch_cnt += writer->batch_cnt;
+    }
+    seq_inc = total_batch_cnt;
+  }
  if (!write_options.disableWAL) {
    status =
        ConcurrentWriteToWAL(write_group, log_used, &last_sequence, seq_inc);
@@ -591,7 +605,8 @@ Status DBImpl::WriteImplWALOnly(const WriteOptions& write_options,
    }
    writer->sequence = curr_seq;
    if (seq_per_batch_) {
-      curr_seq++;
+      assert(writer->batch_cnt);
+      curr_seq += writer->batch_cnt;
    }
    // else seq advances only by memtable writes
  }
@@ -1004,7 +1019,7 @@ Status DBImpl::SwitchWAL(WriteContext* write_context) {
        break;
      }
      cfd->imm()->FlushRequested();
-      SchedulePendingFlush(cfd);
+      SchedulePendingFlush(cfd, FlushReason::kWriteBufferManager);
    }
  }
  MaybeScheduleFlushOrCompaction();
@@ -1050,7 +1065,7 @@ Status DBImpl::HandleWriteBufferFull(WriteContext* write_context) {
    status = SwitchMemtable(cfd_picked, write_context);
    if (status.ok()) {
      cfd_picked->imm()->FlushRequested();
-      SchedulePendingFlush(cfd_picked);
+      SchedulePendingFlush(cfd_picked, FlushReason::kWriteBufferFull);
      MaybeScheduleFlushOrCompaction();
    }
  }
@@ -1292,6 +1307,29 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context) {
                 ". Immutable memtables: %d.\n",
                 cfd->GetName().c_str(), new_log_number, num_imm_unflushed);
  mutex_.Lock();
+  if (s.ok() && creating_new_log) {
+    log_write_mutex_.Lock();
+    logfile_number_ = new_log_number;
+    assert(new_log != nullptr);
+    log_empty_ = true;
+    log_dir_synced_ = false;
+    if (!logs_.empty()) {
+      // Alway flush the buffer of the last log before switching to a new one
+      log::Writer* cur_log_writer = logs_.back().writer;
+      s = cur_log_writer->WriteBuffer();
+      if (!s.ok()) {
+        ROCKS_LOG_WARN(immutable_db_options_.info_log,
+                       "[%s] Failed to switch from #%" PRIu64 " to #%" PRIu64
+                       "  WAL file -- %s\n",
+                       cfd->GetName().c_str(), cur_log_writer->get_log_number(),
+                       new_log_number);
+      }
+    }
+    logs_.emplace_back(logfile_number_, new_log);
+    alive_log_files_.push_back(LogFileNumberSize(logfile_number_));
+    log_write_mutex_.Unlock();
+  }
+
  if (!s.ok()) {
    // how do we fail if we're not creating new log?
    assert(creating_new_log);
@@ -1302,21 +1340,7 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context) {
    }
    return s;
  }
-  if (creating_new_log) {
-    log_write_mutex_.Lock();
-    logfile_number_ = new_log_number;
-    assert(new_log != nullptr);
-    log_empty_ = true;
-    log_dir_synced_ = false;
-    if (!logs_.empty()) {
-      // Alway flush the buffer of the last log before switching to a new one
-      log::Writer* cur_log_writer = logs_.back().writer;
-      cur_log_writer->WriteBuffer();
-    }
-    logs_.emplace_back(logfile_number_, new_log);
-    alive_log_files_.push_back(LogFileNumberSize(logfile_number_));
-    log_write_mutex_.Unlock();
-  }
+
  for (auto loop_cfd : *versions_->GetColumnFamilySet()) {
    // all this is just optimization to delete logs that
    // are no longer needed -- if CF is empty, that means it
@@ -1371,7 +1395,10 @@ Status DB::Put(const WriteOptions& opt, ColumnFamilyHandle* column_family,
  // 8 bytes are taken by header, 4 bytes for count, 1 byte for type,
  // and we allocate 11 extra bytes for key length, as well as value length.
  WriteBatch batch(key.size() + value.size() + 24);
-  batch.Put(column_family, key, value);
+  Status s = batch.Put(column_family, key, value);
+  if (!s.ok()) {
+    return s;
+  }
  return Write(opt, &batch);
 }

@@ -1400,7 +1427,10 @@ Status DB::DeleteRange(const WriteOptions& opt,
 Status DB::Merge(const WriteOptions& opt, ColumnFamilyHandle* column_family,
                 const Slice& key, const Slice& value) {
  WriteBatch batch;
-  batch.Merge(column_family, key, value);
+  Status s = batch.Merge(column_family, key, value);
+  if (!s.ok()) {
+    return s;
+  }
  return Write(opt, &batch);
 }
 }  // namespace rocksdb
@@ -217,6 +217,9 @@ class DBIter final: public Iterator {
        *prop = "Iterator is not valid.";
      }
      return Status::OK();
+    } else if (prop_name == "rocksdb.iterator.internal-key") {
+      *prop = saved_key_.GetUserKey().ToString();
+      return Status::OK();
    }
    return Status::InvalidArgument("Undentified property.");
  }
@@ -611,10 +614,12 @@ void DBIter::MergeValuesNewToOld() {
  // Start the merge process by pushing the first operand
  merge_context_.PushOperand(iter_->value(),
                             iter_->IsValuePinned() /* operand_pinned */);
+  TEST_SYNC_POINT("DBIter::MergeValuesNewToOld:PushedFirstOperand");

  ParsedInternalKey ikey;
  Status s;
  for (iter_->Next(); iter_->Valid(); iter_->Next()) {
+    TEST_SYNC_POINT("DBIter::MergeValuesNewToOld:SteppedToNextOperand");
    if (!ParseKey(&ikey)) {
      // skip corrupted key
      continue;
@@ -972,6 +977,17 @@ bool DBIter::FindValueForCurrentKeyUsingSeek() {
  // assume there is at least one parseable key for this user key
  ParsedInternalKey ikey;
  FindParseableKey(&ikey, kForward);
+  assert(iter_->Valid());
+  assert(user_comparator_->Equal(ikey.user_key, saved_key_.GetUserKey()));
+
+  // In case read_callback presents, the value we seek to may not be visible.
+  // Seek for the next value that's visible.
+  while (!IsVisible(ikey.sequence)) {
+    iter_->Next();
+    FindParseableKey(&ikey, kForward);
+    assert(iter_->Valid());
+    assert(user_comparator_->Equal(ikey.user_key, saved_key_.GetUserKey()));
+  }

  if (ikey.type == kTypeDeletion || ikey.type == kTypeSingleDeletion ||
      range_del_agg_.ShouldDelete(
@@ -36,7 +36,9 @@ class TestIterator : public InternalIterator {
        valid_(false),
        sequence_number_(0),
        iter_(0),
-        cmp(comparator) {}
+        cmp(comparator) {
+    data_.reserve(16);
+  }

  void AddPut(std::string argkey, std::string argvalue) {
    Add(argkey, kTypeValue, argvalue);
@@ -2549,7 +2551,7 @@ TEST_F(DBIterWithMergeIterTest, InnerMergeIteratorDataRace1) {
  // MergeIterator::Prev() realized the mem table iterator is at its end
  // and before an SeekToLast() is called.
  rocksdb::SyncPoint::GetInstance()->SetCallBack(
-      "MergeIterator::Prev:BeforeSeekToLast",
+      "MergeIterator::Prev:BeforePrev",
      [&](void* arg) { internal_iter2_->Add("z", kTypeValue, "7", 12u); });
  rocksdb::SyncPoint::GetInstance()->EnableProcessing();

@@ -2585,7 +2587,7 @@ TEST_F(DBIterWithMergeIterTest, InnerMergeIteratorDataRace2) {
  // mem table after MergeIterator::Prev() realized the mem tableiterator is at
  // its end and before an SeekToLast() is called.
  rocksdb::SyncPoint::GetInstance()->SetCallBack(
-      "MergeIterator::Prev:BeforeSeekToLast", [&](void* arg) {
+      "MergeIterator::Prev:BeforePrev", [&](void* arg) {
        internal_iter2_->Add("z", kTypeValue, "7", 12u);
        internal_iter2_->Add("z", kTypeValue, "7", 11u);
      });
@@ -2623,7 +2625,7 @@ TEST_F(DBIterWithMergeIterTest, InnerMergeIteratorDataRace3) {
  // mem table after MergeIterator::Prev() realized the mem table iterator is at
  // its end and before an SeekToLast() is called.
  rocksdb::SyncPoint::GetInstance()->SetCallBack(
-      "MergeIterator::Prev:BeforeSeekToLast", [&](void* arg) {
+      "MergeIterator::Prev:BeforePrev", [&](void* arg) {
        internal_iter2_->Add("z", kTypeValue, "7", 16u, true);
        internal_iter2_->Add("z", kTypeValue, "7", 15u, true);
        internal_iter2_->Add("z", kTypeValue, "7", 14u, true);
@@ -289,6 +289,8 @@ int main(int argc, char** argv) {
  ::testing::InitGoogleTest(&argc, argv);
  return RUN_ALL_TESTS();
 #else
+  (void) argc;
+  (void) argv;
  return 0;
 #endif
 }
@@ -28,13 +28,11 @@ class MockMemTableRep : public MemTableRep {
    return rep_->Allocate(len, buf);
  }

-  virtual void Insert(KeyHandle handle) override {
-    return rep_->Insert(handle);
-  }
+  virtual void Insert(KeyHandle handle) override { rep_->Insert(handle); }

  virtual void InsertWithHint(KeyHandle handle, void** hint) override {
    num_insert_with_hint_++;
-    ASSERT_NE(nullptr, hint);
+    EXPECT_NE(nullptr, hint);
    last_hint_in_ = *hint;
    rep_->InsertWithHint(handle, hint);
    last_hint_out_ = *hint;
@@ -129,6 +127,84 @@ class TestPrefixExtractor : public SliceTransform {
  }
 };

+// Test that ::Add properly returns false when inserting duplicate keys
+TEST_F(DBMemTableTest, DuplicateSeq) {
+  SequenceNumber seq = 123;
+  std::string value;
+  Status s;
+  MergeContext merge_context;
+  Options options;
+  InternalKeyComparator ikey_cmp(options.comparator);
+  RangeDelAggregator range_del_agg(ikey_cmp, {} /* snapshots */);
+
+  // Create a MemTable
+  InternalKeyComparator cmp(BytewiseComparator());
+  auto factory = std::make_shared<SkipListFactory>();
+  options.memtable_factory = factory;
+  ImmutableCFOptions ioptions(options);
+  WriteBufferManager wb(options.db_write_buffer_size);
+  MemTable* mem = new MemTable(cmp, ioptions, MutableCFOptions(options), &wb,
+                               kMaxSequenceNumber, 0 /* column_family_id */);
+
+  // Write some keys and make sure it returns false on duplicates
+  bool res;
+  res = mem->Add(seq, kTypeValue, "key", "value2");
+  ASSERT_TRUE(res);
+  res = mem->Add(seq, kTypeValue, "key", "value2");
+  ASSERT_FALSE(res);
+  // Changing the type should still cause the duplicatae key
+  res = mem->Add(seq, kTypeMerge, "key", "value2");
+  ASSERT_FALSE(res);
+  // Changing the seq number will make the key fresh
+  res = mem->Add(seq + 1, kTypeMerge, "key", "value2");
+  ASSERT_TRUE(res);
+  // Test with different types for duplicate keys
+  res = mem->Add(seq, kTypeDeletion, "key", "");
+  ASSERT_FALSE(res);
+  res = mem->Add(seq, kTypeSingleDeletion, "key", "");
+  ASSERT_FALSE(res);
+
+  // Test the duplicate keys under stress
+  for (int i = 0; i < 10000; i++) {
+    bool insert_dup = i % 10 == 1;
+    if (!insert_dup) {
+      seq++;
+    }
+    res = mem->Add(seq, kTypeValue, "foo", "value" + ToString(seq));
+    if (insert_dup) {
+      ASSERT_FALSE(res);
+    } else {
+      ASSERT_TRUE(res);
+    }
+  }
+  delete mem;
+
+  // Test with InsertWithHint
+  options.memtable_insert_with_hint_prefix_extractor.reset(
+      new TestPrefixExtractor());  // which uses _ to extract the prefix
+  ioptions = ImmutableCFOptions(options);
+  mem = new MemTable(cmp, ioptions, MutableCFOptions(options), &wb,
+                     kMaxSequenceNumber, 0 /* column_family_id */);
+  // Insert a duplicate key with _ in it
+  res = mem->Add(seq, kTypeValue, "key_1", "value");
+  ASSERT_TRUE(res);
+  res = mem->Add(seq, kTypeValue, "key_1", "value");
+  ASSERT_FALSE(res);
+  delete mem;
+
+  // Test when InsertConcurrently will be invoked
+  options.allow_concurrent_memtable_write = true;
+  ioptions = ImmutableCFOptions(options);
+  mem = new MemTable(cmp, ioptions, MutableCFOptions(options), &wb,
+                     kMaxSequenceNumber, 0 /* column_family_id */);
+  MemTablePostProcessInfo post_process_info;
+  res = mem->Add(seq, kTypeValue, "key", "value", true, &post_process_info);
+  ASSERT_TRUE(res);
+  res = mem->Add(seq, kTypeValue, "key", "value", true, &post_process_info);
+  ASSERT_FALSE(res);
+  delete mem;
+}
+
 TEST_F(DBMemTableTest, InsertWithHint) {
  Options options;
  options.allow_concurrent_memtable_write = false;
@@ -14,10 +14,43 @@

 namespace rocksdb {

+class TestReadCallback : public ReadCallback {
+ public:
+  TestReadCallback(SnapshotChecker* snapshot_checker,
+                   SequenceNumber snapshot_seq)
+      : snapshot_checker_(snapshot_checker), snapshot_seq_(snapshot_seq) {}
+
+  bool IsCommitted(SequenceNumber seq) override {
+    return snapshot_checker_->IsInSnapshot(seq, snapshot_seq_);
+  }
+
+ private:
+  SnapshotChecker* snapshot_checker_;
+  SequenceNumber snapshot_seq_;
+};
+
 // Test merge operator functionality.
 class DBMergeOperatorTest : public DBTestBase {
 public:
  DBMergeOperatorTest() : DBTestBase("/db_merge_operator_test") {}
+
+  std::string GetWithReadCallback(SnapshotChecker* snapshot_checker,
+                                  const Slice& key,
+                                  const Snapshot* snapshot = nullptr) {
+    SequenceNumber seq = snapshot == nullptr ? db_->GetLatestSequenceNumber()
+                                             : snapshot->GetSequenceNumber();
+    TestReadCallback read_callback(snapshot_checker, seq);
+    ReadOptions read_opt;
+    read_opt.snapshot = snapshot;
+    PinnableSlice value;
+    Status s =
+        dbfull()->GetImpl(read_opt, db_->DefaultColumnFamily(), key, &value,
+                          nullptr /*value_found*/, &read_callback);
+    if (!s.ok()) {
+      return s.ToString();
+    }
+    return value.ToString();
+  }
 };

 TEST_F(DBMergeOperatorTest, LimitMergeOperands) {
@@ -449,8 +482,152 @@ TEST_P(MergeOperatorPinningTest, TailingIterator) {
  writer_thread.join();
  reader_thread.join();
 }
+
+TEST_F(DBMergeOperatorTest, TailingIteratorMemtableUnrefedBySomeoneElse) {
+  Options options = CurrentOptions();
+  options.merge_operator = MergeOperators::CreateStringAppendOperator();
+  DestroyAndReopen(options);
+
+  // Overview of the test:
+  //  * There are two merge operands for the same key: one in an sst file,
+  //    another in a memtable.
+  //  * Seek a tailing iterator to this key.
+  //  * As part of the seek, the iterator will:
+  //      (a) first visit the operand in the memtable and tell ForwardIterator
+  //          to pin this operand, then
+  //      (b) move on to the operand in the sst file, then pass both operands
+  //          to merge operator.
+  //  * The memtable may get flushed and unreferenced by another thread between
+  //    (a) and (b). The test simulates it by flushing the memtable inside a
+  //    SyncPoint callback located between (a) and (b).
+  //  * In this case it's ForwardIterator's responsibility to keep the memtable
+  //    pinned until (b) is complete. There used to be a bug causing
+  //    ForwardIterator to not pin it in some circumstances. This test
+  //    reproduces it.
+
+  db_->Merge(WriteOptions(), "key", "sst");
+  db_->Flush(FlushOptions()); // Switch to SuperVersion A
+  db_->Merge(WriteOptions(), "key", "memtable");
+
+  // Pin SuperVersion A
+  std::unique_ptr<Iterator> someone_else(db_->NewIterator(ReadOptions()));
+
+  bool pushed_first_operand = false;
+  bool stepped_to_next_operand = false;
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "DBIter::MergeValuesNewToOld:PushedFirstOperand", [&](void*) {
+        EXPECT_FALSE(pushed_first_operand);
+        pushed_first_operand = true;
+        db_->Flush(FlushOptions()); // Switch to SuperVersion B
+      });
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "DBIter::MergeValuesNewToOld:SteppedToNextOperand", [&](void*) {
+        EXPECT_FALSE(stepped_to_next_operand);
+        stepped_to_next_operand = true;
+        someone_else.reset(); // Unpin SuperVersion A
+      });
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  ReadOptions ro;
+  ro.tailing = true;
+  std::unique_ptr<Iterator> iter(db_->NewIterator(ro));
+  iter->Seek("key");
+
+  ASSERT_TRUE(iter->status().ok());
+  ASSERT_TRUE(iter->Valid());
+  EXPECT_EQ(std::string("sst,memtable"), iter->value().ToString());
+  EXPECT_TRUE(pushed_first_operand);
+  EXPECT_TRUE(stepped_to_next_operand);
+}
 #endif  // ROCKSDB_LITE

+TEST_F(DBMergeOperatorTest, SnapshotCheckerAndReadCallback) {
+  Options options = CurrentOptions();
+  options.merge_operator = MergeOperators::CreateStringAppendOperator();
+  DestroyAndReopen(options);
+
+  class TestSnapshotChecker : public SnapshotChecker {
+    bool IsInSnapshot(SequenceNumber seq,
+                      SequenceNumber snapshot_seq) const override {
+      switch (snapshot_seq) {
+        case 0:
+          return seq == 0;
+        case 1:
+          return seq <= 1;
+        case 2:
+          // seq = 2 not visible to snapshot with seq = 2
+          return seq <= 1;
+        case 3:
+          return seq <= 3;
+        case 4:
+          // seq = 4 not visible to snpahost with seq = 4
+          return seq <= 3;
+        default:
+          // seq >=4 is uncommitted
+          return seq <= 4;
+      };
+    }
+  };
+  TestSnapshotChecker* snapshot_checker = new TestSnapshotChecker();
+  dbfull()->SetSnapshotChecker(snapshot_checker);
+
+  std::string value;
+  ASSERT_OK(Merge("foo", "v1"));
+  ASSERT_EQ(1, db_->GetLatestSequenceNumber());
+  ASSERT_EQ("v1", GetWithReadCallback(snapshot_checker, "foo"));
+  ASSERT_OK(Merge("foo", "v2"));
+  ASSERT_EQ(2, db_->GetLatestSequenceNumber());
+  // v2 is not visible to latest snapshot, which has seq = 2.
+  ASSERT_EQ("v1", GetWithReadCallback(snapshot_checker, "foo"));
+  // Take a snapshot with seq = 2.
+  const Snapshot* snapshot1 = db_->GetSnapshot();
+  ASSERT_EQ(2, snapshot1->GetSequenceNumber());
+  // v2 is not visible to snapshot1, which has seq = 2
+  ASSERT_EQ("v1", GetWithReadCallback(snapshot_checker, "foo", snapshot1));
+
+  // Verify flush doesn't alter the result.
+  ASSERT_OK(Flush());
+  ASSERT_EQ("v1", GetWithReadCallback(snapshot_checker, "foo", snapshot1));
+  ASSERT_EQ("v1", GetWithReadCallback(snapshot_checker, "foo"));
+
+  ASSERT_OK(Merge("foo", "v3"));
+  ASSERT_EQ(3, db_->GetLatestSequenceNumber());
+  ASSERT_EQ("v1,v2,v3", GetWithReadCallback(snapshot_checker, "foo"));
+  ASSERT_OK(Merge("foo", "v4"));
+  ASSERT_EQ(4, db_->GetLatestSequenceNumber());
+  // v4 is not visible to latest snapshot, which has seq = 4.
+  ASSERT_EQ("v1,v2,v3", GetWithReadCallback(snapshot_checker, "foo"));
+  const Snapshot* snapshot2 = db_->GetSnapshot();
+  ASSERT_EQ(4, snapshot2->GetSequenceNumber());
+  // v4 is not visible to snapshot2, which has seq = 4.
+  ASSERT_EQ("v1,v2,v3",
+            GetWithReadCallback(snapshot_checker, "foo", snapshot2));
+
+  // Verify flush doesn't alter the result.
+  ASSERT_OK(Flush());
+  ASSERT_EQ("v1", GetWithReadCallback(snapshot_checker, "foo", snapshot1));
+  ASSERT_EQ("v1,v2,v3",
+            GetWithReadCallback(snapshot_checker, "foo", snapshot2));
+  ASSERT_EQ("v1,v2,v3", GetWithReadCallback(snapshot_checker, "foo"));
+
+  ASSERT_OK(Merge("foo", "v5"));
+  ASSERT_EQ(5, db_->GetLatestSequenceNumber());
+  // v5 is uncommitted
+  ASSERT_EQ("v1,v2,v3,v4", GetWithReadCallback(snapshot_checker, "foo"));
+
+  // full manual compaction.
+  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+
+  // Verify compaction doesn't alter the result.
+  ASSERT_EQ("v1", GetWithReadCallback(snapshot_checker, "foo", snapshot1));
+  ASSERT_EQ("v1,v2,v3",
+            GetWithReadCallback(snapshot_checker, "foo", snapshot2));
+  ASSERT_EQ("v1,v2,v3,v4", GetWithReadCallback(snapshot_checker, "foo"));
+
+  db_->ReleaseSnapshot(snapshot1);
+  db_->ReleaseSnapshot(snapshot2);
+}
+
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -518,8 +518,13 @@ static void assert_candidate_files_empty(DBImpl* dbfull, const bool empty) {
  JobContext job_context(0);
  dbfull->FindObsoleteFiles(&job_context, false);
  ASSERT_EQ(empty, job_context.full_scan_candidate_files.empty());
-  job_context.Clean();
  dbfull->TEST_UnlockMutex();
+  if (job_context.HaveSomethingToDelete()) {
+    // fulfill the contract of FindObsoleteFiles by calling PurgeObsoleteFiles
+    // afterwards; otherwise the test may hang on shutdown
+    dbfull->PurgeObsoleteFiles(job_context);
+  }
+  job_context.Clean();
 }

 TEST_F(DBOptionsTest, DeleteObsoleteFilesPeriodChange) {
@@ -809,6 +809,8 @@ int main(int argc, char** argv) {
  ::testing::InitGoogleTest(&argc, argv);
  return RUN_ALL_TESTS();
 #else
+  (void) argc;
+  (void) argv;
  return 0;
 #endif
 }
@@ -480,6 +480,36 @@ TEST_F(DBTest, SingleDeletePutFlush) {
                         kSkipUniversalCompaction | kSkipMergePut));
 }

+// Disable because not all platform can run it.
+// It requires more than 9GB memory to run it, With single allocation
+// of more than 3GB.
+TEST_F(DBTest, DISABLED_SanitizeVeryVeryLargeValue) {
+  const size_t kValueSize = 4 * size_t{1024 * 1024 * 1024};  // 4GB value
+  std::string raw(kValueSize, 'v');
+  Options options = CurrentOptions();
+  options.env = env_;
+  options.merge_operator = MergeOperators::CreatePutOperator();
+  options.write_buffer_size = 100000;  // Small write buffer
+  options.paranoid_checks = true;
+  DestroyAndReopen(options);
+
+  ASSERT_OK(Put("boo", "v1"));
+  ASSERT_TRUE(Put("foo", raw).IsInvalidArgument());
+  ASSERT_TRUE(Merge("foo", raw).IsInvalidArgument());
+
+  WriteBatch wb;
+  ASSERT_TRUE(wb.Put("foo", raw).IsInvalidArgument());
+  ASSERT_TRUE(wb.Merge("foo", raw).IsInvalidArgument());
+
+  Slice value_slice = raw;
+  Slice key_slice = "foo";
+  SliceParts sp_key(&key_slice, 1);
+  SliceParts sp_value(&value_slice, 1);
+
+  ASSERT_TRUE(wb.Put(sp_key, sp_value).IsInvalidArgument());
+  ASSERT_TRUE(wb.Merge(sp_key, sp_value).IsInvalidArgument());
+}
+
 // Disable because not all platform can run it.
 // It requires more than 9GB memory to run it, With single allocation
 // of more than 3GB.
@@ -503,7 +533,9 @@ TEST_F(DBTest, DISABLED_VeryLargeValue) {
  ASSERT_OK(Put(key2, raw));
  dbfull()->TEST_WaitForFlushMemTable();

+#ifndef ROCKSDB_LITE
  ASSERT_EQ(1, NumTableFilesAtLevel(0));
+#endif  // !ROCKSDB_LITE

  std::string value;
  Status s = db_->Get(ReadOptions(), key1, &value);
@@ -2191,6 +2223,8 @@ class ModelDB : public DB {
    batch.Put(cf, k, v);
    return Write(o, &batch);
  }
+  using DB::Close;
+  virtual Status Close() override { return Status::OK(); }
  using DB::Delete;
  virtual Status Delete(const WriteOptions& o, ColumnFamilyHandle* cf,
                        const Slice& key) override {
@@ -5054,55 +5088,84 @@ TEST_F(DBTest, PromoteL0Failure) {
  status = experimental::PromoteL0(db_, db_->DefaultColumnFamily());
  ASSERT_TRUE(status.IsInvalidArgument());
 }
-#endif  // ROCKSDB_LITE

 // Github issue #596
-TEST_F(DBTest, HugeNumberOfLevels) {
+TEST_F(DBTest, CompactRangeWithEmptyBottomLevel) {
+  const int kNumLevels = 2;
+  const int kNumL0Files = 2;
  Options options = CurrentOptions();
-  options.write_buffer_size = 2 * 1024 * 1024;         // 2MB
-  options.max_bytes_for_level_base = 2 * 1024 * 1024;  // 2MB
-  options.num_levels = 12;
-  options.max_background_compactions = 10;
-  options.max_bytes_for_level_multiplier = 2;
-  options.level_compaction_dynamic_level_bytes = true;
+  options.disable_auto_compactions = true;
+  options.num_levels = kNumLevels;
  DestroyAndReopen(options);

  Random rnd(301);
-  for (int i = 0; i < 300000; ++i) {
-    ASSERT_OK(Put(Key(i), RandomString(&rnd, 1024)));
+  for (int i = 0; i < kNumL0Files; ++i) {
+    ASSERT_OK(Put(Key(0), RandomString(&rnd, 1024)));
+    Flush();
  }
+  ASSERT_EQ(NumTableFilesAtLevel(0), kNumL0Files);
+  ASSERT_EQ(NumTableFilesAtLevel(1), 0);

  ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
+  ASSERT_EQ(NumTableFilesAtLevel(0), 0);
+  ASSERT_EQ(NumTableFilesAtLevel(1), kNumL0Files);
 }
+#endif  // ROCKSDB_LITE

 TEST_F(DBTest, AutomaticConflictsWithManualCompaction) {
+  const int kNumL0Files = 50;
  Options options = CurrentOptions();
-  options.write_buffer_size = 2 * 1024 * 1024;         // 2MB
-  options.max_bytes_for_level_base = 2 * 1024 * 1024;  // 2MB
-  options.num_levels = 12;
+  options.level0_file_num_compaction_trigger = 4;
+  // never slowdown / stop
+  options.level0_slowdown_writes_trigger = 999999;
+  options.level0_stop_writes_trigger = 999999;
  options.max_background_compactions = 10;
-  options.max_bytes_for_level_multiplier = 2;
-  options.level_compaction_dynamic_level_bytes = true;
  DestroyAndReopen(options);

-  Random rnd(301);
-  for (int i = 0; i < 300000; ++i) {
-    ASSERT_OK(Put(Key(i), RandomString(&rnd, 1024)));
-  }
-
+  // schedule automatic compactions after the manual one starts, but before it
+  // finishes to ensure conflict.
+  rocksdb::SyncPoint::GetInstance()->LoadDependency(
+      {{"DBImpl::BackgroundCompaction:Start",
+        "DBTest::AutomaticConflictsWithManualCompaction:PrePuts"},
+       {"DBTest::AutomaticConflictsWithManualCompaction:PostPuts",
+        "DBImpl::BackgroundCompaction:NonTrivial:AfterRun"}});
  std::atomic<int> callback_count(0);
  rocksdb::SyncPoint::GetInstance()->SetCallBack(
-      "DBImpl::BackgroundCompaction()::Conflict",
+      "DBImpl::MaybeScheduleFlushOrCompaction:Conflict",
      [&](void* arg) { callback_count.fetch_add(1); });
  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
-  CompactRangeOptions croptions;
-  croptions.exclusive_manual_compaction = false;
-  ASSERT_OK(db_->CompactRange(croptions, nullptr, nullptr));
+
+  Random rnd(301);
+  for (int i = 0; i < 2; ++i) {
+    // put two keys to ensure no trivial move
+    for (int j = 0; j < 2; ++j) {
+      ASSERT_OK(Put(Key(j), RandomString(&rnd, 1024)));
+    }
+    ASSERT_OK(Flush());
+  }
+  std::thread manual_compaction_thread([this]() {
+    CompactRangeOptions croptions;
+    croptions.exclusive_manual_compaction = true;
+    ASSERT_OK(db_->CompactRange(croptions, nullptr, nullptr));
+  });
+
+  TEST_SYNC_POINT("DBTest::AutomaticConflictsWithManualCompaction:PrePuts");
+  for (int i = 0; i < kNumL0Files; ++i) {
+    // put two keys to ensure no trivial move
+    for (int j = 0; j < 2; ++j) {
+      ASSERT_OK(Put(Key(j), RandomString(&rnd, 1024)));
+    }
+    ASSERT_OK(Flush());
+  }
+  TEST_SYNC_POINT("DBTest::AutomaticConflictsWithManualCompaction:PostPuts");
+
  ASSERT_GE(callback_count.load(), 1);
-  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
-  for (int i = 0; i < 300000; ++i) {
+  for (int i = 0; i < 2; ++i) {
    ASSERT_NE("NOT_FOUND", Get(Key(i)));
  }
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+  manual_compaction_thread.join();
+  dbfull()->TEST_WaitForCompact();
 }

 // Github issue #595
@@ -5322,12 +5385,15 @@ class WriteStallListener : public EventListener {
 public:
  WriteStallListener() : condition_(WriteStallCondition::kNormal) {}
  void OnStallConditionsChanged(const WriteStallInfo& info) override {
+    MutexLock l(&mutex_);
    condition_ = info.condition.cur;
  }
  bool CheckCondition(WriteStallCondition expected) {
+    MutexLock l(&mutex_);
    return expected == condition_;
  }
 private:
+  port::Mutex mutex_;
  WriteStallCondition condition_;
 };

@@ -5656,6 +5722,50 @@ TEST_F(DBTest, PauseBackgroundWorkTest) {
  // now it's done
  ASSERT_TRUE(done.load());
 }
+
+// Keep spawning short-living threads that create an iterator and quit.
+// Meanwhile in another thread keep flushing memtables.
+// This used to cause a deadlock.
+TEST_F(DBTest, ThreadLocalPtrDeadlock) {
+  std::atomic<int> flushes_done{0};
+  std::atomic<int> threads_destroyed{0};
+  auto done = [&] {
+    return flushes_done.load() > 10;
+  };
+
+  std::thread flushing_thread([&] {
+    for (int i = 0; !done(); ++i) {
+      ASSERT_OK(db_->Put(WriteOptions(), Slice("hi"),
+                         Slice(std::to_string(i).c_str())));
+      ASSERT_OK(db_->Flush(FlushOptions()));
+      int cnt = ++flushes_done;
+      fprintf(stderr, "Flushed %d times\n", cnt);
+    }
+  });
+
+  std::vector<std::thread> thread_spawning_threads(10);
+  for (auto& t: thread_spawning_threads) {
+    t = std::thread([&] {
+      while (!done()) {
+        {
+          std::thread tmp_thread([&] {
+            auto it = db_->NewIterator(ReadOptions());
+            delete it;
+          });
+          tmp_thread.join();
+        }
+        ++threads_destroyed;
+      }
+    });
+  }
+
+  for (auto& t: thread_spawning_threads) {
+    t.join();
+  }
+  flushing_thread.join();
+  fprintf(stderr, "Done. Flushed %d times, destroyed %d threads\n",
+          flushes_done.load(), threads_destroyed.load());
+}
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -1820,11 +1820,17 @@ TEST_F(DBTest2, ReadAmpBitmap) {

 #ifndef OS_SOLARIS // GetUniqueIdFromFile is not implemented
 TEST_F(DBTest2, ReadAmpBitmapLiveInCacheAfterDBClose) {
-  if (dbname_.find("dev/shm") != std::string::npos) {
-    // /dev/shm dont support getting a unique file id, this mean that
-    // running this test on /dev/shm will fail because lru_cache will load
-    // the blocks again regardless of them being already in the cache
-    return;
+  {
+    const int kIdBufLen = 100;
+    char id_buf[kIdBufLen];
+    std::unique_ptr<RandomAccessFile> file;
+    env_->NewRandomAccessFile(dbname_, &file, EnvOptions());
+    if (file->GetUniqueId(id_buf, kIdBufLen) == 0) {
+      // fs holding db directory doesn't support getting a unique file id,
+      // this means that running this test will fail because lru_cache will load
+      // the blocks again regardless of them being already in the cache
+      return;
+    }
  }
  uint32_t bytes_per_bit[2] = {1, 16};
  for (size_t k = 0; k < 2; k++) {
@@ -2430,6 +2436,60 @@ TEST_F(DBTest2, ReadCallbackTest) {
  }
 }

+#ifndef ROCKSDB_LITE
+
+TEST_F(DBTest2, LiveFilesOmitObsoleteFiles) {
+  // Regression test for race condition where an obsolete file is returned to
+  // user as a "live file" but then deleted, all while file deletions are
+  // disabled.
+  //
+  // It happened like this:
+  //
+  // 1. [flush thread] Log file "x.log" found by FindObsoleteFiles
+  // 2. [user thread] DisableFileDeletions, GetSortedWalFiles are called and the
+  //    latter returned "x.log"
+  // 3. [flush thread] PurgeObsoleteFiles deleted "x.log"
+  // 4. [user thread] Reading "x.log" failed
+  //
+  // Unfortunately the only regression test I can come up with involves sleep.
+  // We cannot set SyncPoints to repro since, once the fix is applied, the
+  // SyncPoints would cause a deadlock as the repro's sequence of events is now
+  // prohibited.
+  //
+  // Instead, if we sleep for a second between Find and Purge, and ensure the
+  // read attempt happens after purge, then the sequence of events will almost
+  // certainly happen on the old code.
+  rocksdb::SyncPoint::GetInstance()->LoadDependency({
+      {"DBImpl::BackgroundCallFlush:FilesFound",
+       "DBTest2::LiveFilesOmitObsoleteFiles:FlushTriggered"},
+      {"DBImpl::PurgeObsoleteFiles:End",
+       "DBTest2::LiveFilesOmitObsoleteFiles:LiveFilesCaptured"},
+  });
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "DBImpl::PurgeObsoleteFiles:Begin",
+      [&](void* arg) { env_->SleepForMicroseconds(1000000); });
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  Put("key", "val");
+  FlushOptions flush_opts;
+  flush_opts.wait = false;
+  db_->Flush(flush_opts);
+  TEST_SYNC_POINT("DBTest2::LiveFilesOmitObsoleteFiles:FlushTriggered");
+
+  db_->DisableFileDeletions();
+  VectorLogPtr log_files;
+  db_->GetSortedWalFiles(log_files);
+  TEST_SYNC_POINT("DBTest2::LiveFilesOmitObsoleteFiles:LiveFilesCaptured");
+  for (const auto& log_file : log_files) {
+    ASSERT_OK(env_->FileExists(LogFileName(dbname_, log_file->LogNumber())));
+  }
+
+  db_->EnableFileDeletions();
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+}
+
+#endif  // ROCKSDB_LITE
+
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -867,7 +867,6 @@ size_t DBTestBase::CountLiveFiles() {
  db_->GetLiveFilesMetaData(&metadata);
  return metadata.size();
 }
-#endif  // ROCKSDB_LITE

 int DBTestBase::NumTableFilesAtLevel(int level, int cf) {
  std::string property;
@@ -928,6 +927,7 @@ std::string DBTestBase::FilesPerLevel(int cf) {
  result.resize(last_non_zero_offset);
  return result;
 }
+#endif  // !ROCKSDB_LITE

 size_t DBTestBase::CountFiles() {
  std::vector<std::string> files;
@@ -997,6 +997,7 @@ void DBTestBase::MoveFilesToLevel(int level, int cf) {
  }
 }

+#ifndef ROCKSDB_LITE
 void DBTestBase::DumpFileCounts(const char* label) {
  fprintf(stderr, "---\n%s:\n", label);
  fprintf(stderr, "maxoverlap: %" PRIu64 "\n",
@@ -1008,6 +1009,7 @@ void DBTestBase::DumpFileCounts(const char* label) {
    }
  }
 }
+#endif  // !ROCKSDB_LITE

 std::string DBTestBase::DumpSSTableList() {
  std::string property;
@@ -137,8 +137,8 @@ class SpecialMemTableRep : public MemTableRep {
  // Insert key into the list.
  // REQUIRES: nothing that compares equal to key is currently in the list.
  virtual void Insert(KeyHandle handle) override {
-    memtable_->Insert(handle);
    num_entries_++;
+    memtable_->Insert(handle);
  }

  // Returns true iff an entry that compares equal to key is in the list.
@@ -863,13 +863,13 @@ class DBTestBase : public testing::Test {
  size_t TotalLiveFiles(int cf = 0);

  size_t CountLiveFiles();
-#endif  // ROCKSDB_LITE

  int NumTableFilesAtLevel(int level, int cf = 0);

  double CompressionRatioAtLevel(int level, int cf = 0);

  int TotalTableFiles(int cf = 0, int levels = -1);
+#endif  // ROCKSDB_LITE

  // Return spread of files per level
  std::string FilesPerLevel(int cf = 0);
@@ -897,7 +897,9 @@ class DBTestBase : public testing::Test {

  void MoveFilesToLevel(int level, int cf = 0);

+#ifndef ROCKSDB_LITE
  void DumpFileCounts(const char* label);
+#endif  // ROCKSDB_LITE

  std::string DumpSSTableList();

@@ -388,10 +388,16 @@ TEST_P(DBTestUniversalCompaction, DynamicUniversalCompactionSizeAmplification) {
  DestroyAndReopen(options);

  int total_picked_compactions = 0;
+  int total_size_amp_compactions = 0;
  rocksdb::SyncPoint::GetInstance()->SetCallBack(
      "UniversalCompactionPicker::PickCompaction:Return", [&](void* arg) {
        if (arg) {
          total_picked_compactions++;
+          Compaction* c = static_cast<Compaction*>(arg);
+          if (c->compaction_reason() ==
+              CompactionReason::kUniversalSizeAmplification) {
+            total_size_amp_compactions++;
+          }
        }
      });
  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
@@ -445,6 +451,102 @@ TEST_P(DBTestUniversalCompaction, DynamicUniversalCompactionSizeAmplification) {
  // Verify that size amplification did happen
  ASSERT_EQ(NumSortedRuns(1), 1);
  ASSERT_EQ(total_picked_compactions, 1);
+  ASSERT_EQ(total_size_amp_compactions, 1);
+}
+
+TEST_P(DBTestUniversalCompaction, DynamicUniversalCompactionReadAmplification) {
+  Options options = CurrentOptions();
+  options.compaction_style = kCompactionStyleUniversal;
+  options.num_levels = 1;
+  options.write_buffer_size = 100 << 10;     // 100KB
+  options.target_file_size_base = 32 << 10;  // 32KB
+  options.level0_file_num_compaction_trigger = 3;
+  // Initial setup of compaction_options_universal will prevent universal
+  // compaction from happening
+  options.compaction_options_universal.max_size_amplification_percent = 2000;
+  options.compaction_options_universal.size_ratio = 0;
+  options.compaction_options_universal.min_merge_width = 100;
+  DestroyAndReopen(options);
+
+  int total_picked_compactions = 0;
+  int total_size_ratio_compactions = 0;
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "UniversalCompactionPicker::PickCompaction:Return", [&](void* arg) {
+        if (arg) {
+          total_picked_compactions++;
+          Compaction* c = static_cast<Compaction*>(arg);
+          if (c->compaction_reason() == CompactionReason::kUniversalSizeRatio) {
+            total_size_ratio_compactions++;
+          }
+        }
+      });
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  MutableCFOptions mutable_cf_options;
+  CreateAndReopenWithCF({"pikachu"}, options);
+
+  Random rnd(301);
+  int key_idx = 0;
+
+  // Generate three files in Level 0. All files are approx the same size.
+  for (int num = 0; num < options.level0_file_num_compaction_trigger; num++) {
+    // Write 110KB (11 values, each 10K)
+    for (int i = 0; i < 11; i++) {
+      ASSERT_OK(Put(1, Key(key_idx), RandomString(&rnd, 10000)));
+      key_idx++;
+    }
+    dbfull()->TEST_WaitForFlushMemTable(handles_[1]);
+    ASSERT_EQ(NumSortedRuns(1), num + 1);
+  }
+  ASSERT_EQ(NumSortedRuns(1), options.level0_file_num_compaction_trigger);
+
+  // Flush whatever is remaining in memtable. This is typically small, about
+  // 30KB.
+  ASSERT_OK(Flush(1));
+  dbfull()->TEST_WaitForCompact();
+  // Verify compaction did not happen
+  ASSERT_EQ(NumSortedRuns(1), options.level0_file_num_compaction_trigger + 1);
+  ASSERT_EQ(total_picked_compactions, 0);
+
+  ASSERT_OK(dbfull()->SetOptions(
+      handles_[1],
+      {{"compaction_options_universal",
+        "{min_merge_width=2;max_merge_width=2;size_ratio=100;}"}}));
+  ASSERT_EQ(dbfull()
+                ->GetOptions(handles_[1])
+                .compaction_options_universal.min_merge_width,
+            2);
+  ASSERT_EQ(dbfull()
+                ->GetOptions(handles_[1])
+                .compaction_options_universal.max_merge_width,
+            2);
+  ASSERT_EQ(
+      dbfull()->GetOptions(handles_[1]).compaction_options_universal.size_ratio,
+      100);
+
+  ASSERT_OK(dbfull()->TEST_GetLatestMutableCFOptions(handles_[1],
+                                                     &mutable_cf_options));
+  ASSERT_EQ(mutable_cf_options.compaction_options_universal.size_ratio, 100);
+  ASSERT_EQ(mutable_cf_options.compaction_options_universal.min_merge_width, 2);
+  ASSERT_EQ(mutable_cf_options.compaction_options_universal.max_merge_width, 2);
+
+  dbfull()->TEST_WaitForCompact();
+
+  // Files in L0 are approx: 0.3 (30KB), 1, 1, 1.
+  // On compaction: the files are below the size amp threshold, so we
+  // fallthrough to checking read amp conditions. The configured size ratio is
+  // not big enough to take 0.3 into consideration. So the next files 1 and 1
+  // are compacted together first as they satisfy size ratio condition and
+  // (min_merge_width, max_merge_width) condition, to give out a file size of 2.
+  // Next, the newly generated 2 and the last file 1 are compacted together. So
+  // at the end: #sortedRuns = 2, #picked_compactions = 2, and all the picked
+  // ones are size ratio based compactions.
+  ASSERT_EQ(NumSortedRuns(1), 2);
+  // If max_merge_width had not been changed dynamically above, and if it
+  // continued to be the default value of UINIT_MAX, total_picked_compactions
+  // would have been 1.
+  ASSERT_EQ(total_picked_compactions, 2);
+  ASSERT_EQ(total_size_ratio_compactions, 2);
 }

 TEST_P(DBTestUniversalCompaction, CompactFilesOnUniversalCompaction) {
@@ -1657,6 +1759,8 @@ int main(int argc, char** argv) {
  ::testing::InitGoogleTest(&argc, argv);
  return RUN_ALL_TESTS();
 #else
+  (void) argc;
+  (void) argv;
  return 0;
 #endif
 }
@@ -125,6 +125,26 @@ int InternalKeyComparator::Compare(const Slice& akey, const Slice& bkey) const {
  return r;
 }

+int InternalKeyComparator::CompareKeySeq(const Slice& akey,
+                                         const Slice& bkey) const {
+  // Order by:
+  //    increasing user key (according to user-supplied comparator)
+  //    decreasing sequence number
+  int r = user_comparator_->Compare(ExtractUserKey(akey), ExtractUserKey(bkey));
+  PERF_COUNTER_ADD(user_key_comparison_count, 1);
+  if (r == 0) {
+    // Shift the number to exclude the last byte which contains the value type
+    const uint64_t anum = DecodeFixed64(akey.data() + akey.size() - 8) >> 8;
+    const uint64_t bnum = DecodeFixed64(bkey.data() + bkey.size() - 8) >> 8;
+    if (anum > bnum) {
+      r = -1;
+    } else if (anum < bnum) {
+      r = +1;
+    }
+  }
+  return r;
+}
+
 int InternalKeyComparator::Compare(const ParsedInternalKey& a,
                                   const ParsedInternalKey& b) const {
  // Order by:
@@ -162,6 +162,8 @@ class InternalKeyComparator

  virtual const char* Name() const override;
  virtual int Compare(const Slice& a, const Slice& b) const override;
+  // Same as Compare except that it excludes the value type from comparison
+  virtual int CompareKeySeq(const Slice& a, const Slice& b) const;
  virtual void FindShortestSeparator(std::string* start,
                                     const Slice& limit) const override;
  virtual void FindShortSuccessor(std::string* key) const override;
@@ -500,7 +500,7 @@ int main(int argc, char** argv) {
 #else
 #include <stdio.h>

-int main(int argc, char** argv) {
+int main(int /*argc*/, char** /*argv*/) {
  fprintf(stderr,
          "SKIPPED as DBImpl::DeleteFile is not supported in ROCKSDB_LITE\n");
  return 0;
@@ -132,11 +132,15 @@ Status ExternalSstFileIngestionJob::Prepare(
  return status;
 }

-Status ExternalSstFileIngestionJob::NeedsFlush(bool* flush_needed) {
-  SuperVersion* super_version = cfd_->GetSuperVersion();
+Status ExternalSstFileIngestionJob::NeedsFlush(bool* flush_needed,
+                                               SuperVersion* super_version) {
+  autovector<Range> ranges;
+  for (const IngestedFileInfo& file_to_ingest : files_to_ingest_) {
+    ranges.emplace_back(file_to_ingest.smallest_user_key,
+                        file_to_ingest.largest_user_key);
+  }
  Status status =
-      IngestedFilesOverlapWithMemtables(super_version, flush_needed);
-
+      cfd_->RangesOverlapWithMemtables(ranges, super_version, flush_needed);
  if (status.ok() && *flush_needed &&
      !ingestion_options_.allow_blocking_flush) {
    status = Status::InvalidArgument("External file requires flush");
@@ -148,11 +152,12 @@ Status ExternalSstFileIngestionJob::NeedsFlush(bool* flush_needed) {
 // nonmem_write_thread_
 Status ExternalSstFileIngestionJob::Run() {
  Status status;
+  SuperVersion* super_version = cfd_->GetSuperVersion();
 #ifndef NDEBUG
  // We should never run the job with a memtable that is overlapping
  // with the files we are ingesting
  bool need_flush = false;
-  status = NeedsFlush(&need_flush);
+  status = NeedsFlush(&need_flush, super_version);
  assert(status.ok() && need_flush == false);
 #endif

@@ -167,7 +172,6 @@ Status ExternalSstFileIngestionJob::Run() {
  // It is safe to use this instead of LastAllocatedSequence since we are
  // the only active writer, and hence they are equal
  const SequenceNumber last_seqno = versions_->LastSequence();
-  SuperVersion* super_version = cfd_->GetSuperVersion();
  edit_.SetColumnFamily(cfd_->GetID());
  // The levels that the files will be ingested into

@@ -375,54 +379,6 @@ Status ExternalSstFileIngestionJob::GetIngestedFileInfo(
  return status;
 }

-Status ExternalSstFileIngestionJob::IngestedFilesOverlapWithMemtables(
-    SuperVersion* sv, bool* overlap) {
-  *overlap = false;
-  // Create an InternalIterator over all memtables
-  Arena arena;
-  ReadOptions ro;
-  ro.total_order_seek = true;
-  MergeIteratorBuilder merge_iter_builder(&cfd_->internal_comparator(), &arena);
-  merge_iter_builder.AddIterator(sv->mem->NewIterator(ro, &arena));
-  sv->imm->AddIterators(ro, &merge_iter_builder);
-  ScopedArenaIterator memtable_iter(merge_iter_builder.Finish());
-
-  std::vector<InternalIterator*> memtable_range_del_iters;
-  auto* active_range_del_iter = sv->mem->NewRangeTombstoneIterator(ro);
-  if (active_range_del_iter != nullptr) {
-    memtable_range_del_iters.push_back(active_range_del_iter);
-  }
-  sv->imm->AddRangeTombstoneIterators(ro, &memtable_range_del_iters);
-  RangeDelAggregator range_del_agg(cfd_->internal_comparator(),
-                                   {} /* snapshots */,
-                                   false /* collapse_deletions */);
-  Status status;
-  {
-    std::unique_ptr<InternalIterator> memtable_range_del_iter(
-        NewMergingIterator(&cfd_->internal_comparator(),
-                           memtable_range_del_iters.empty()
-                               ? nullptr
-                               : &memtable_range_del_iters[0],
-                           static_cast<int>(memtable_range_del_iters.size())));
-    status = range_del_agg.AddTombstones(std::move(memtable_range_del_iter));
-  }
-  if (status.ok()) {
-    for (IngestedFileInfo& f : files_to_ingest_) {
-      status = IngestedFileOverlapWithIteratorRange(&f, memtable_iter.get(),
-                                                    overlap);
-      if (!status.ok() || *overlap == true) {
-        break;
-      }
-      if (range_del_agg.IsRangeOverlapped(f.smallest_user_key,
-                                          f.largest_user_key)) {
-        *overlap = true;
-        break;
-      }
-    }
-  }
-  return status;
-}
-
 Status ExternalSstFileIngestionJob::AssignLevelAndSeqnoForIngestedFile(
    SuperVersion* sv, bool force_global_seqno, CompactionStyle compaction_style,
    IngestedFileInfo* file_to_ingest, SequenceNumber* assigned_seqno) {
@@ -473,7 +429,10 @@ Status ExternalSstFileIngestionJob::AssignLevelAndSeqnoForIngestedFile(
                            return f1->largest_seqno < f2->largest_seqno;
                          }))
                ->largest_seqno;
-        if (level_largest_seqno != 0) {
+        // should only assign seqno to current level's largest seqno when
+        // the file fits
+        if (level_largest_seqno != 0 &&
+            IngestedFileFitInLevel(file_to_ingest, lvl)) {
          *assigned_seqno = level_largest_seqno;
        } else {
          continue;
@@ -550,6 +509,9 @@ Status ExternalSstFileIngestionJob::AssignGlobalSeqnoForIngestedFile(
  std::string seqno_val;
  PutFixed64(&seqno_val, seqno);
  status = rwfile->Write(file_to_ingest->global_seqno_offset, seqno_val);
+  if (status.ok()) {
+    status = rwfile->Fsync();
+  }
  if (status.ok()) {
    file_to_ingest->assigned_seqno = seqno;
  }
@@ -84,8 +84,12 @@ class ExternalSstFileIngestionJob {
  // Check if we need to flush the memtable before running the ingestion job
  // This will be true if the files we are ingesting are overlapping with any
  // key range in the memtable.
-  // REQUIRES: Mutex held
-  Status NeedsFlush(bool* flush_needed);
+  //
+  // @param super_version A referenced SuperVersion that will be held for the
+  //    duration of this function.
+  //
+  // Thread-safe
+  Status NeedsFlush(bool* flush_needed, SuperVersion* super_version);

  // Will execute the ingestion job and prepare edit() to be applied.
  // REQUIRES: Mutex held
@@ -110,10 +114,6 @@ class ExternalSstFileIngestionJob {
  Status GetIngestedFileInfo(const std::string& external_file,
                             IngestedFileInfo* file_to_ingest);

-  // Check if the files we are ingesting overlap with any memtable.
-  // REQUIRES: Mutex held
-  Status IngestedFilesOverlapWithMemtables(SuperVersion* sv, bool* overlap);
-
  // Assign `file_to_ingest` the appropriate sequence number and  the lowest
  // possible level that it can be ingested to according to compaction_style.
  // REQUIRES: Mutex held
@@ -388,9 +388,10 @@ class SstFileWriterCollector : public TablePropertiesCollector {
  const char* Name() const override { return name_.c_str(); }

  Status Finish(UserCollectedProperties* properties) override {
+    std::string count = std::to_string(count_);
    *properties = UserCollectedProperties{
        {prefix_ + "_SstFileWriterCollector", "YES"},
-        {prefix_ + "_Count", std::to_string(count_)},
+        {prefix_ + "_Count", count},
    };
    return Status::OK();
  }
@@ -1940,6 +1941,55 @@ TEST_F(ExternalSSTFileTest, IngestBehind) {
  size_t kcnt = 0;
  VerifyDBFromMap(true_data, &kcnt, false);
 }
+
+TEST_F(ExternalSSTFileTest, SkipBloomFilter) {
+  Options options = CurrentOptions();
+
+  BlockBasedTableOptions table_options;
+  table_options.filter_policy.reset(NewBloomFilterPolicy(10));
+  table_options.cache_index_and_filter_blocks = true;
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+
+
+  // Create external SST file and include bloom filters
+  options.statistics = rocksdb::CreateDBStatistics();
+  DestroyAndReopen(options);
+  {
+    std::string file_path = sst_files_dir_ + "sst_with_bloom.sst";
+    SstFileWriter sst_file_writer(EnvOptions(), options);
+    ASSERT_OK(sst_file_writer.Open(file_path));
+    ASSERT_OK(sst_file_writer.Put("Key1", "Value1"));
+    ASSERT_OK(sst_file_writer.Finish());
+
+    ASSERT_OK(
+        db_->IngestExternalFile({file_path}, IngestExternalFileOptions()));
+
+    ASSERT_EQ(Get("Key1"), "Value1");
+    ASSERT_GE(
+        options.statistics->getTickerCount(Tickers::BLOCK_CACHE_FILTER_ADD), 1);
+  }
+
+  // Create external SST file but skip bloom filters
+  options.statistics = rocksdb::CreateDBStatistics();
+  DestroyAndReopen(options);
+  {
+    std::string file_path = sst_files_dir_ + "sst_with_no_bloom.sst";
+    SstFileWriter sst_file_writer(EnvOptions(), options, nullptr, true,
+                                  Env::IOPriority::IO_TOTAL,
+                                  true /* skip_filters */);
+    ASSERT_OK(sst_file_writer.Open(file_path));
+    ASSERT_OK(sst_file_writer.Put("Key1", "Value1"));
+    ASSERT_OK(sst_file_writer.Finish());
+
+    ASSERT_OK(
+        db_->IngestExternalFile({file_path}, IngestExternalFileOptions()));
+
+    ASSERT_EQ(Get("Key1"), "Value1");
+    ASSERT_EQ(
+        options.statistics->getTickerCount(Tickers::BLOCK_CACHE_FILTER_ADD), 0);
+  }
+}
+
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -1951,7 +2001,7 @@ int main(int argc, char** argv) {
 #else
 #include <stdio.h>

-int main(int argc, char** argv) {
+int main(int /*argc*/, char** /*argv*/) {
  fprintf(stderr,
          "SKIPPED as External SST File Writer and Ingestion are not supported "
          "in ROCKSDB_LITE\n");
@@ -228,16 +228,9 @@ class FaultInjectionTest : public testing::Test,
    return Status::OK();
  }

-#ifdef ROCKSDB_UBSAN_RUN
-#if defined(__clang__)
-__attribute__((__no_sanitize__("shift"), no_sanitize("signed-integer-overflow")))
-#elif defined(__GNUC__)
-__attribute__((__no_sanitize_undefined__))
-#endif
-#endif
  // Return the ith key
  Slice Key(int i, std::string* storage) const {
-    int num = i;
+    unsigned long long num = i;
    if (!sequential_order_) {
      // random transfer
      const int m = 0x5bd1e995;
@@ -245,7 +238,7 @@ __attribute__((__no_sanitize_undefined__))
      num ^= num << 24;
    }
    char buf[100];
-    snprintf(buf, sizeof(buf), "%016d", num);
+    snprintf(buf, sizeof(buf), "%016d", static_cast<int>(num));
    storage->assign(buf, strlen(buf));
    return Slice(*storage);
  }
@@ -54,6 +54,32 @@

 namespace rocksdb {

+const char* GetFlushReasonString (FlushReason flush_reason) {
+  switch (flush_reason) {
+    case FlushReason::kUnknown:
+      return "Unknown";
+    case FlushReason::kGetLiveFiles:
+      return "Get Live Files";
+    case FlushReason::kShutDown:
+      return "Shut down";
+    case FlushReason::kExternalFileIngestion:
+      return "External File Ingestion";
+    case FlushReason::kManualCompaction:
+      return "Manual Compaction";
+    case FlushReason::kWriteBufferManager:
+      return "Write Buffer Manager";
+    case FlushReason::kWriteBufferFull:
+      return "Write Buffer Full";
+    case FlushReason::kTest:
+      return "Test";
+    case FlushReason::kSuperVersionChange:
+      return "SuperVersion Change";
+    default:
+      return "Invalid";
+  }
+}
+
+
 FlushJob::FlushJob(const std::string& dbname, ColumnFamilyData* cfd,
                   const ImmutableDBOptions& db_options,
                   const MutableCFOptions& mutable_cf_options,
@@ -278,12 +304,13 @@ Status FlushJob::WriteLevel0Table() {
      total_memory_usage += m->ApproximateMemoryUsage();
    }

-    event_logger_->Log() << "job" << job_context_->job_id << "event"
-                         << "flush_started"
-                         << "num_memtables" << mems_.size() << "num_entries"
-                         << total_num_entries << "num_deletes"
-                         << total_num_deletes << "memory_usage"
-                         << total_memory_usage;
+    event_logger_->Log()
+        << "job" << job_context_->job_id << "event"
+        << "flush_started"
+        << "num_memtables" << mems_.size() << "num_entries" << total_num_entries
+        << "num_deletes" << total_num_deletes << "memory_usage"
+        << total_memory_usage << "flush_reason"
+        << GetFlushReasonString(cfd_->GetFlushReason());

    {
      ScopedArenaIterator iter(
@@ -25,15 +25,15 @@
 namespace rocksdb {

 // Usage:
-//     LevelIterator iter;
+//     ForwardLevelIterator iter;
 //     iter.SetFileIndex(file_index);
 //     iter.Seek(target);
 //     iter.Next()
-class LevelIterator : public InternalIterator {
+class ForwardLevelIterator : public InternalIterator {
 public:
-  LevelIterator(const ColumnFamilyData* const cfd,
-                const ReadOptions& read_options,
-                const std::vector<FileMetaData*>& files)
+  ForwardLevelIterator(const ColumnFamilyData* const cfd,
+                       const ReadOptions& read_options,
+                       const std::vector<FileMetaData*>& files)
      : cfd_(cfd),
        read_options_(read_options),
        files_(files),
@@ -42,7 +42,7 @@ class LevelIterator : public InternalIterator {
        file_iter_(nullptr),
        pinned_iters_mgr_(nullptr) {}

-  ~LevelIterator() {
+  ~ForwardLevelIterator() {
    // Reset current pointer
    if (pinned_iters_mgr_ && pinned_iters_mgr_->PinningEnabled()) {
      pinned_iters_mgr_->PinIterator(file_iter_);
@@ -84,11 +84,11 @@ class LevelIterator : public InternalIterator {
    }
  }
  void SeekToLast() override {
-    status_ = Status::NotSupported("LevelIterator::SeekToLast()");
+    status_ = Status::NotSupported("ForwardLevelIterator::SeekToLast()");
    valid_ = false;
  }
  void Prev() override {
-    status_ = Status::NotSupported("LevelIterator::Prev()");
+    status_ = Status::NotSupported("ForwardLevelIterator::Prev()");
    valid_ = false;
  }
  bool Valid() const override {
@@ -105,7 +105,7 @@ class LevelIterator : public InternalIterator {
    valid_ = file_iter_->Valid();
  }
  void SeekForPrev(const Slice& internal_key) override {
-    status_ = Status::NotSupported("LevelIterator::SeekForPrev()");
+    status_ = Status::NotSupported("ForwardLevelIterator::SeekForPrev()");
    valid_ = false;
  }
  void Next() override {
@@ -196,38 +196,60 @@ ForwardIterator::~ForwardIterator() {
  Cleanup(true);
 }

-namespace {
-// Used in PinnedIteratorsManager to release pinned SuperVersion
-static void ReleaseSuperVersionFunc(void* sv) {
-  delete reinterpret_cast<SuperVersion*>(sv);
-}
-}  // namespace
-
-void ForwardIterator::SVCleanup() {
-  if (sv_ != nullptr && sv_->Unref()) {
+void ForwardIterator::SVCleanup(DBImpl* db, SuperVersion* sv,
+                                bool background_purge_on_iterator_cleanup) {
+  if (sv->Unref()) {
    // Job id == 0 means that this is not our background process, but rather
    // user thread
    JobContext job_context(0);
-    db_->mutex_.Lock();
-    sv_->Cleanup();
-    db_->FindObsoleteFiles(&job_context, false, true);
-    if (read_options_.background_purge_on_iterator_cleanup) {
-      db_->ScheduleBgLogWriterClose(&job_context);
-    }
-    db_->mutex_.Unlock();
-    if (pinned_iters_mgr_ && pinned_iters_mgr_->PinningEnabled()) {
-      pinned_iters_mgr_->PinPtr(sv_, &ReleaseSuperVersionFunc);
-    } else {
-      delete sv_;
+    db->mutex_.Lock();
+    sv->Cleanup();
+    db->FindObsoleteFiles(&job_context, false, true);
+    if (background_purge_on_iterator_cleanup) {
+      db->ScheduleBgLogWriterClose(&job_context);
    }
+    db->mutex_.Unlock();
+    delete sv;
    if (job_context.HaveSomethingToDelete()) {
-      db_->PurgeObsoleteFiles(
-          job_context, read_options_.background_purge_on_iterator_cleanup);
+      db->PurgeObsoleteFiles(job_context, background_purge_on_iterator_cleanup);
    }
    job_context.Clean();
  }
 }

+namespace {
+struct SVCleanupParams {
+  DBImpl* db;
+  SuperVersion* sv;
+  bool background_purge_on_iterator_cleanup;
+};
+}
+
+// Used in PinnedIteratorsManager to release pinned SuperVersion
+void ForwardIterator::DeferredSVCleanup(void* arg) {
+  auto d = reinterpret_cast<SVCleanupParams*>(arg);
+  ForwardIterator::SVCleanup(
+    d->db, d->sv, d->background_purge_on_iterator_cleanup);
+  delete d;
+}
+
+void ForwardIterator::SVCleanup() {
+  if (sv_ == nullptr) {
+    return;
+  }
+  if (pinned_iters_mgr_ && pinned_iters_mgr_->PinningEnabled()) {
+    // pinned_iters_mgr_ tells us to make sure that all visited key-value slices
+    // are alive until pinned_iters_mgr_->ReleasePinnedData() is called.
+    // The slices may point into some memtables owned by sv_, so we need to keep
+    // sv_ referenced until pinned_iters_mgr_ unpins everything.
+    auto p = new SVCleanupParams{
+      db_, sv_, read_options_.background_purge_on_iterator_cleanup};
+    pinned_iters_mgr_->PinPtr(p, &ForwardIterator::DeferredSVCleanup);
+  } else {
+    SVCleanup(db_, sv_, read_options_.background_purge_on_iterator_cleanup);
+  }
+}
+
 void ForwardIterator::Cleanup(bool release_sv) {
  if (mutable_iter_ != nullptr) {
    DeleteIterator(mutable_iter_, true /* is_arena */);
@@ -361,14 +383,13 @@ void ForwardIterator::SeekInternal(const Slice& internal_key,

      if (!l0_iters_[i]->status().ok()) {
        immutable_status_ = l0_iters_[i]->status();
-      } else if (l0_iters_[i]->Valid()) {
-        if (!IsOverUpperBound(l0_iters_[i]->key())) {
-          immutable_min_heap_.push(l0_iters_[i]);
-        } else {
-          has_iter_trimmed_for_upper_bound_ = true;
-          DeleteIterator(l0_iters_[i]);
-          l0_iters_[i] = nullptr;
-        }
+      } else if (l0_iters_[i]->Valid() &&
+                 !IsOverUpperBound(l0_iters_[i]->key())) {
+        immutable_min_heap_.push(l0_iters_[i]);
+      } else {
+        has_iter_trimmed_for_upper_bound_ = true;
+        DeleteIterator(l0_iters_[i]);
+        l0_iters_[i] = nullptr;
      }
    }

@@ -395,15 +416,14 @@ void ForwardIterator::SeekInternal(const Slice& internal_key,

        if (!level_iters_[level - 1]->status().ok()) {
          immutable_status_ = level_iters_[level - 1]->status();
-        } else if (level_iters_[level - 1]->Valid()) {
-          if (!IsOverUpperBound(level_iters_[level - 1]->key())) {
-            immutable_min_heap_.push(level_iters_[level - 1]);
-          } else {
-            // Nothing in this level is interesting. Remove.
-            has_iter_trimmed_for_upper_bound_ = true;
-            DeleteIterator(level_iters_[level - 1]);
-            level_iters_[level - 1] = nullptr;
-          }
+        } else if (level_iters_[level - 1]->Valid() &&
+                   !IsOverUpperBound(level_iters_[level - 1]->key())) {
+          immutable_min_heap_.push(level_iters_[level - 1]);
+        } else {
+          // Nothing in this level is interesting. Remove.
+          has_iter_trimmed_for_upper_bound_ = true;
+          DeleteIterator(level_iters_[level - 1]);
+          level_iters_[level - 1] = nullptr;
        }
      }
    }
@@ -541,7 +561,7 @@ void ForwardIterator::UpdateChildrenPinnedItersMgr() {
  }

  // Set PinnedIteratorsManager for L1+ levels iterators.
-  for (LevelIterator* child_iter : level_iters_) {
+  for (ForwardLevelIterator* child_iter : level_iters_) {
    if (child_iter) {
      child_iter->SetPinnedItersMgr(pinned_iters_mgr_);
    }
@@ -703,7 +723,7 @@ void ForwardIterator::BuildLevelIterators(const VersionStorageInfo* vstorage) {
      }
    } else {
      level_iters_.push_back(
-          new LevelIterator(cfd_, read_options_, level_files));
+          new ForwardLevelIterator(cfd_, read_options_, level_files));
    }
  }
 }
@@ -23,7 +23,7 @@ class DBImpl;
 class Env;
 struct SuperVersion;
 class ColumnFamilyData;
-class LevelIterator;
+class ForwardLevelIterator;
 class VersionStorageInfo;
 struct FileMetaData;

@@ -85,7 +85,14 @@ class ForwardIterator : public InternalIterator {

 private:
  void Cleanup(bool release_sv);
+  // Unreference and, if needed, clean up the current SuperVersion. This is
+  // either done immediately or deferred until this iterator is unpinned by
+  // PinnedIteratorsManager.
  void SVCleanup();
+  static void SVCleanup(
+    DBImpl* db, SuperVersion* sv, bool background_purge_on_iterator_cleanup);
+  static void DeferredSVCleanup(void* arg);
+
  void RebuildIterators(bool refresh_sv);
  void RenewIterators();
  void BuildLevelIterators(const VersionStorageInfo* vstorage);
@@ -119,7 +126,7 @@ class ForwardIterator : public InternalIterator {
  InternalIterator* mutable_iter_;
  std::vector<InternalIterator*> imm_iters_;
  std::vector<InternalIterator*> l0_iters_;
-  std::vector<LevelIterator*> level_iters_;
+  std::vector<ForwardLevelIterator*> level_iters_;
  InternalIterator* current_;
  bool valid_;

@@ -81,8 +81,11 @@ struct SuperVersionContext {
 struct JobContext {
  inline bool HaveSomethingToDelete() const {
    return full_scan_candidate_files.size() || sst_delete_files.size() ||
-           log_delete_files.size() || manifest_delete_files.size() ||
-           memtables_to_free.size() > 0 || logs_to_free.size() > 0 ||
+           log_delete_files.size() || manifest_delete_files.size();
+  }
+
+  inline bool HaveSomethingToClean() const {
+    return memtables_to_free.size() > 0 || logs_to_free.size() > 0 ||
           superversion_context.HaveSomethingToDelete();
  }

@@ -528,7 +528,7 @@ TEST_F(EventListenerTest, CompactionReasonUniversal) {

  ASSERT_GT(listener->compaction_reasons_.size(), 0);
  for (auto compaction_reason : listener->compaction_reasons_) {
-    ASSERT_EQ(compaction_reason, CompactionReason::kUniversalSortedRunNum);
+    ASSERT_EQ(compaction_reason, CompactionReason::kUniversalSizeRatio);
  }

  options.level0_file_num_compaction_trigger = 8;
@@ -54,6 +54,7 @@ class Reader {
  // The Reader will start reading at the first record located at physical
  // position >= initial_offset within the file.
  Reader(std::shared_ptr<Logger> info_log,
+  // @lint-ignore TXT2 T25377293 Grandfathered in
 	 unique_ptr<SequentialFileReader>&& file,
         Reporter* reporter, bool checksum, uint64_t initial_offset,
         uint64_t log_num);
@@ -16,7 +16,12 @@
 namespace rocksdb {

 #ifdef ROCKSDB_JEMALLOC
+#ifdef __FreeBSD__
+#include <malloc_np.h>
+#define je_malloc_stats_print malloc_stats_print
+#else
 #include "jemalloc/jemalloc.h"
+#endif

 typedef struct {
  char* cur;
@@ -36,8 +41,8 @@ static void GetJemallocStatus(void* mstat_arg, const char* status) {
 }
 #endif  // ROCKSDB_JEMALLOC

-void DumpMallocStats(std::string* stats) {
 #ifdef ROCKSDB_JEMALLOC
+void DumpMallocStats(std::string* stats) {
  MallocStatus mstat;
  const unsigned int kMallocStatusLen = 1000000;
  std::unique_ptr<char[]> buf{new char[kMallocStatusLen + 1]};
@@ -45,8 +50,10 @@ void DumpMallocStats(std::string* stats) {
  mstat.end = buf.get() + kMallocStatusLen;
  je_malloc_stats_print(GetJemallocStatus, &mstat, "");
  stats->append(buf.get());
-#endif  // ROCKSDB_JEMALLOC
 }
+#else
+void DumpMallocStats(std::string*) {}
+#endif  // ROCKSDB_JEMALLOC

 }
 #endif  // !ROCKSDB_LITE
@@ -225,7 +225,7 @@ int MemTable::KeyComparator::operator()(const char* prefix_len_key1,
  // Internal keys are encoded as length-prefixed strings.
  Slice k1 = GetLengthPrefixedSlice(prefix_len_key1);
  Slice k2 = GetLengthPrefixedSlice(prefix_len_key2);
-  return comparator.Compare(k1, k2);
+  return comparator.CompareKeySeq(k1, k2);
 }

 int MemTable::KeyComparator::operator()(const char* prefix_len_key,
@@ -233,7 +233,15 @@ int MemTable::KeyComparator::operator()(const char* prefix_len_key,
    const {
  // Internal keys are encoded as length-prefixed strings.
  Slice a = GetLengthPrefixedSlice(prefix_len_key);
-  return comparator.Compare(a, key);
+  return comparator.CompareKeySeq(a, key);
+}
+
+void MemTableRep::InsertConcurrently(KeyHandle handle) {
+#ifndef ROCKSDB_LITE
+  throw std::runtime_error("concurrent insert not supported");
+#else
+  abort();
+#endif
 }

 Slice MemTableRep::UserKey(const char* key) const {
@@ -281,8 +289,7 @@ class MemTableIterator : public InternalIterator {
 #ifndef NDEBUG
    // Assert that the MemTableIterator is never deleted while
    // Pinning is Enabled.
-    assert(!pinned_iters_mgr_ ||
-           (pinned_iters_mgr_ && !pinned_iters_mgr_->PinningEnabled()));
+    assert(!pinned_iters_mgr_ || !pinned_iters_mgr_->PinningEnabled());
 #endif
    if (arena_mode_) {
      iter_->~Iterator();
@@ -436,7 +443,7 @@ MemTable::MemTableStats MemTable::ApproximateStats(const Slice& start_ikey,
  return {entry_count * (data_size / n), entry_count};
 }

-void MemTable::Add(SequenceNumber s, ValueType type,
+bool MemTable::Add(SequenceNumber s, ValueType type,
                   const Slice& key, /* user key */
                   const Slice& value, bool allow_concurrent,
                   MemTablePostProcessInfo* post_process_info) {
@@ -471,9 +478,15 @@ void MemTable::Add(SequenceNumber s, ValueType type,
    if (insert_with_hint_prefix_extractor_ != nullptr &&
        insert_with_hint_prefix_extractor_->InDomain(key_slice)) {
      Slice prefix = insert_with_hint_prefix_extractor_->Transform(key_slice);
-      table->InsertWithHint(handle, &insert_hints_[prefix]);
+      bool res = table->InsertKeyWithHint(handle, &insert_hints_[prefix]);
+      if (!res) {
+        return res;
+      }
    } else {
-      table->Insert(handle);
+      bool res = table->InsertKey(handle);
+      if (!res) {
+        return res;
+      }
    }

    // this is a bit ugly, but is the way to avoid locked instructions
@@ -506,7 +519,10 @@ void MemTable::Add(SequenceNumber s, ValueType type,
    assert(post_process_info == nullptr);
    UpdateFlushState();
  } else {
-    table->InsertConcurrently(handle);
+    bool res = table->InsertKeyConcurrently(handle);
+    if (!res) {
+      return res;
+    }

    assert(post_process_info != nullptr);
    post_process_info->num_entries++;
@@ -536,6 +552,7 @@ void MemTable::Add(SequenceNumber s, ValueType type,
    is_range_del_table_empty_ = false;
  }
  UpdateOldestKeyTime();
+  return true;
 }

 // Callback from MemTable::Get()
@@ -571,11 +588,12 @@ struct Saver {

 static bool SaveValue(void* arg, const char* entry) {
  Saver* s = reinterpret_cast<Saver*>(arg);
+  assert(s != nullptr);
  MergeContext* merge_context = s->merge_context;
  RangeDelAggregator* range_del_agg = s->range_del_agg;
  const MergeOperator* merge_operator = s->merge_operator;

-  assert(s != nullptr && merge_context != nullptr && range_del_agg != nullptr);
+  assert(merge_context != nullptr && range_del_agg != nullptr);

  // entry format is:
  //    klength  varint32
@@ -791,8 +809,9 @@ void MemTable::Update(SequenceNumber seq,
      // Correct user key
      const uint64_t tag = DecodeFixed64(key_ptr + key_length - 8);
      ValueType type;
-      SequenceNumber unused;
-      UnPackSequenceAndType(tag, &unused, &type);
+      SequenceNumber existing_seq;
+      UnPackSequenceAndType(tag, &existing_seq, &type);
+      assert(existing_seq != seq);
      if (type == kTypeValue) {
        Slice prev_value = GetLengthPrefixedSlice(key_ptr + key_length);
        uint32_t prev_size = static_cast<uint32_t>(prev_value.size());
@@ -815,7 +834,10 @@ void MemTable::Update(SequenceNumber seq,
  }

  // key doesn't exist
-  Add(seq, kTypeValue, key, value);
+  bool add_res __attribute__((__unused__));
+  add_res = Add(seq, kTypeValue, key, value);
+  // We already checked unused != seq above. In that case, Add should not fail.
+  assert(add_res);
 }

 bool MemTable::UpdateCallback(SequenceNumber seq,
@@ -167,7 +167,10 @@ class MemTable {
  //
  // REQUIRES: if allow_concurrent = false, external synchronization to prevent
  // simultaneous operations on the same MemTable.
-  void Add(SequenceNumber seq, ValueType type, const Slice& key,
+  //
+  // Returns false if MemTableRepFactory::CanHandleDuplicatedKey() is true and
+  // the <key, seq> already exists.
+  bool Add(SequenceNumber seq, ValueType type, const Slice& key,
           const Slice& value, bool allow_concurrent = false,
           MemTablePostProcessInfo* post_process_info = nullptr);

@@ -368,6 +371,11 @@ class MemTable {
    return oldest_key_time_.load(std::memory_order_relaxed);
  }

+  // REQUIRES: db_mutex held.
+  void SetID(uint64_t id) { id_ = id; }
+
+  uint64_t GetID() const { return id_; }
+
 private:
  enum FlushStateEnum { FLUSH_NOT_REQUESTED, FLUSH_REQUESTED, FLUSH_SCHEDULED };

@@ -437,6 +445,9 @@ class MemTable {
  // Timestamp of oldest key
  std::atomic<uint64_t> oldest_key_time_;

+  // Memtable id to track flush.
+  uint64_t id_ = 0;
+
  // Returns a heuristic flush decision
  bool ShouldFlushNow() const;

@@ -5,11 +5,12 @@
 //
 #pragma once

-#include <string>
-#include <list>
-#include <vector>
-#include <set>
 #include <deque>
+#include <limits>
+#include <list>
+#include <set>
+#include <string>
+#include <vector>

 #include "db/dbformat.h"
 #include "db/memtable.h"
@@ -244,6 +245,22 @@ class MemTableList {

  uint64_t GetMinLogContainingPrepSection();

+  uint64_t GetEarliestMemTableID() const {
+    auto& memlist = current_->memlist_;
+    if (memlist.empty()) {
+      return std::numeric_limits<uint64_t>::max();
+    }
+    return memlist.back()->GetID();
+  }
+
+  uint64_t GetLatestMemTableID() const {
+    auto& memlist = current_->memlist_;
+    if (memlist.empty()) {
+      return 0;
+    }
+    return memlist.front()->GetID();
+  }
+
 private:
  // DB mutex held
  void InstallNewVersion();
@@ -5,12 +5,12 @@

 #include "db/merge_helper.h"

-#include <stdio.h>
 #include <string>

 #include "db/dbformat.h"
 #include "monitoring/perf_context_imp.h"
 #include "monitoring/statistics.h"
+#include "port/likely.h"
 #include "rocksdb/comparator.h"
 #include "rocksdb/db.h"
 #include "rocksdb/merge_operator.h"
@@ -22,7 +22,8 @@ MergeHelper::MergeHelper(Env* env, const Comparator* user_comparator,
                         const MergeOperator* user_merge_operator,
                         const CompactionFilter* compaction_filter,
                         Logger* logger, bool assert_valid_internal_key,
-                         SequenceNumber latest_snapshot, int level,
+                         SequenceNumber latest_snapshot,
+                         const SnapshotChecker* snapshot_checker, int level,
                         Statistics* stats,
                         const std::atomic<bool>* shutting_down)
    : env_(env),
@@ -34,6 +35,7 @@ MergeHelper::MergeHelper(Env* env, const Comparator* user_comparator,
      assert_valid_internal_key_(assert_valid_internal_key),
      allow_single_operand_(false),
      latest_snapshot_(latest_snapshot),
+      snapshot_checker_(snapshot_checker),
      level_(level),
      keys_(),
      filter_timer_(env_),
@@ -158,7 +160,10 @@ Status MergeHelper::MergeUntil(InternalIterator* iter,
      // hit a different user key, stop right here
      hit_the_next_user_key = true;
      break;
-    } else if (stop_before && ikey.sequence <= stop_before) {
+    } else if (stop_before > 0 && ikey.sequence <= stop_before &&
+               LIKELY(snapshot_checker_ == nullptr ||
+                      snapshot_checker_->IsInSnapshot(ikey.sequence,
+                                                      stop_before))) {
      // hit an entry that's visible by the previous snapshot, can't touch that
      break;
    }
@@ -13,6 +13,7 @@
 #include "db/dbformat.h"
 #include "db/merge_context.h"
 #include "db/range_del_aggregator.h"
+#include "db/snapshot_checker.h"
 #include "rocksdb/compaction_filter.h"
 #include "rocksdb/env.h"
 #include "rocksdb/slice.h"
@@ -33,7 +34,8 @@ class MergeHelper {
              const MergeOperator* user_merge_operator,
              const CompactionFilter* compaction_filter, Logger* logger,
              bool assert_valid_internal_key, SequenceNumber latest_snapshot,
-              int level = 0, Statistics* stats = nullptr,
+              const SnapshotChecker* snapshot_checker = nullptr, int level = 0,
+              Statistics* stats = nullptr,
              const std::atomic<bool>* shutting_down = nullptr);

  // Wrapper around MergeOperator::FullMergeV2() that records perf statistics.
@@ -145,6 +147,7 @@ class MergeHelper {
  bool assert_valid_internal_key_; // enforce no internal key corruption?
  bool allow_single_operand_;
  SequenceNumber latest_snapshot_;
+  const SnapshotChecker* const snapshot_checker_;
  int level_;

  // the scratch area that holds the result of MergeUntil
@@ -112,7 +112,7 @@ int main(int argc, char** argv) {

 #include <cstdio>

-int main(int argc, char** argv) {
+int main(int /*argc*/, char** /*argv*/) {
  printf("Skipped as Options file is not supported in RocksDBLite.\n");
  return 0;
 }
@@ -1170,7 +1170,7 @@ int main(int argc, char** argv) {
 #else
 #include <stdio.h>

-int main(int argc, char** argv) {
+int main(int /*argc*/, char** /*argv*/) {
  fprintf(stderr, "SKIPPED as plain table is not supported in ROCKSDB_LITE\n");
  return 0;
 }
@@ -887,7 +887,7 @@ int main(int argc, char** argv) {
 #else
 #include <stdio.h>

-int main(int argc, char** argv) {
+int main(int /*argc*/, char** /*argv*/) {
  fprintf(stderr,
          "SKIPPED as HashSkipList and HashLinkList are not supported in "
          "ROCKSDB_LITE\n");
@@ -356,7 +356,7 @@ int main(int argc, char** argv) {
 #else
 #include <stdio.h>

-int main(int argc, char** argv) {
+int main(int /*argc*/, char** /*argv*/) {
  fprintf(stderr, "SKIPPED as RepairDB is not supported in ROCKSDB_LITE\n");
  return 0;
 }
@@ -45,6 +45,10 @@ class SnapshotList {
    list_.prev_ = &list_;
    list_.next_ = &list_;
    list_.number_ = 0xFFFFFFFFL;      // placeholder marker, for debugging
+    // Set all the variables to make UBSAN happy.
+    list_.list_ = nullptr;
+    list_.unix_time_ = 0;
+    list_.is_write_conflict_boundary_ = false;
    count_ = 0;
  }

@@ -272,9 +272,8 @@ InternalIterator* TableCache::NewRangeTombstoneIterator(
    const InternalKeyComparator& icomparator, const FileDescriptor& fd,
    HistogramImpl* file_read_hist, bool skip_filters, int level) {
  Status s;
-  TableReader* table_reader = nullptr;
  Cache::Handle* handle = nullptr;
-  table_reader = fd.table_reader;
+  TableReader* table_reader = fd.table_reader;
  if (table_reader == nullptr) {
    s = FindTable(env_options, icomparator, fd, &handle,
                  options.read_tier == kBlockCacheTier /* no_io */,
@@ -60,7 +60,7 @@ InternalKeyPropertiesCollector::GetReadableProperties() const {
 namespace {

 uint64_t GetUint64Property(const UserCollectedProperties& props,
-                           const std::string property_name,
+                           const std::string& property_name,
                           bool* property_present) {
  auto pos = props.find(property_name);
  if (pos == props.end()) {
@@ -424,129 +424,261 @@ bool SomeFileOverlapsRange(
 }

 namespace {
-
-// An internal iterator.  For a given version/level pair, yields
-// information about the files in the level.  For a given entry, key()
-// is the largest key that occurs in the file, and value() is an
-// 16-byte value containing the file number and file size, both
-// encoded using EncodeFixed64.
-class LevelFileNumIterator : public InternalIterator {
+class LevelIterator final : public InternalIterator {
 public:
-  LevelFileNumIterator(const InternalKeyComparator& icmp,
-                       const LevelFilesBrief* flevel, bool should_sample)
-      : icmp_(icmp),
-        flevel_(flevel),
-        index_(static_cast<uint32_t>(flevel->num_files)),
-        current_value_(0, 0, 0),  // Marks as invalid
-        should_sample_(should_sample) {}
-  virtual bool Valid() const override { return index_ < flevel_->num_files; }
-  virtual void Seek(const Slice& target) override {
-    index_ = FindFile(icmp_, *flevel_, target);
-  }
-  virtual void SeekForPrev(const Slice& target) override {
-    SeekForPrevImpl(target, &icmp_);
-  }
-
-  virtual void SeekToFirst() override { index_ = 0; }
-  virtual void SeekToLast() override {
-    index_ = (flevel_->num_files == 0)
-                 ? 0
-                 : static_cast<uint32_t>(flevel_->num_files) - 1;
-  }
-  virtual void Next() override {
-    assert(Valid());
-    index_++;
-  }
-  virtual void Prev() override {
-    assert(Valid());
-    if (index_ == 0) {
-      index_ = static_cast<uint32_t>(flevel_->num_files);  // Marks as invalid
-    } else {
-      index_--;
-    }
-  }
-  Slice key() const override {
-    assert(Valid());
-    return flevel_->files[index_].largest_key;
-  }
-  Slice value() const override {
-    assert(Valid());
-
-    auto file_meta = flevel_->files[index_];
-    if (should_sample_) {
-      sample_file_read_inc(file_meta.file_metadata);
-    }
-    current_value_ = file_meta.fd;
-    return Slice(reinterpret_cast<const char*>(&current_value_),
-                 sizeof(FileDescriptor));
-  }
-  virtual Status status() const override { return Status::OK(); }
-
- private:
-  const InternalKeyComparator icmp_;
-  const LevelFilesBrief* flevel_;
-  uint32_t index_;
-  mutable FileDescriptor current_value_;
-  bool should_sample_;
-};
-
-class LevelFileIteratorState : public TwoLevelIteratorState {
- public:
-  // @param skip_filters Disables loading/accessing the filter block
-  LevelFileIteratorState(TableCache* table_cache,
-                         const ReadOptions& read_options,
-                         const EnvOptions& env_options,
-                         const InternalKeyComparator& icomparator,
-                         HistogramImpl* file_read_hist, bool for_compaction,
-                         bool prefix_enabled, bool skip_filters, int level,
-                         RangeDelAggregator* range_del_agg)
-      : TwoLevelIteratorState(prefix_enabled),
-        table_cache_(table_cache),
+  LevelIterator(TableCache* table_cache, const ReadOptions& read_options,
+                const EnvOptions& env_options,
+                const InternalKeyComparator& icomparator,
+                const LevelFilesBrief* flevel, bool should_sample,
+                HistogramImpl* file_read_hist, bool for_compaction,
+                bool skip_filters, int level, RangeDelAggregator* range_del_agg)
+      : table_cache_(table_cache),
        read_options_(read_options),
        env_options_(env_options),
        icomparator_(icomparator),
+        flevel_(flevel),
        file_read_hist_(file_read_hist),
+        should_sample_(should_sample),
        for_compaction_(for_compaction),
        skip_filters_(skip_filters),
+        file_index_(flevel_->num_files),
        level_(level),
-        range_del_agg_(range_del_agg) {}
+        range_del_agg_(range_del_agg),
+        pinned_iters_mgr_(nullptr) {
+    // Empty level is not supported.
+    assert(flevel_ != nullptr && flevel_->num_files > 0);
+  }

-  InternalIterator* NewSecondaryIterator(const Slice& meta_handle) override {
-    if (meta_handle.size() != sizeof(FileDescriptor)) {
-      return NewErrorInternalIterator(
-          Status::Corruption("FileReader invoked with unexpected value"));
+  virtual ~LevelIterator() { delete file_iter_.Set(nullptr); }
+
+  virtual void Seek(const Slice& target) override;
+  virtual void SeekForPrev(const Slice& target) override;
+  virtual void SeekToFirst() override;
+  virtual void SeekToLast() override;
+  virtual void Next() override;
+  virtual void Prev() override;
+
+  virtual bool Valid() const override { return file_iter_.Valid(); }
+  virtual Slice key() const override {
+    assert(Valid());
+    return file_iter_.key();
+  }
+  virtual Slice value() const override {
+    assert(Valid());
+    return file_iter_.value();
+  }
+  virtual Status status() const override {
+    // It'd be nice if status() returned a const Status& instead of a Status
+    if (!status_.ok()) {
+      return status_;
+    } else if (file_iter_.iter() != nullptr) {
+      return file_iter_.status();
    }
-    const FileDescriptor* fd =
-        reinterpret_cast<const FileDescriptor*>(meta_handle.data());
-    return table_cache_->NewIterator(
-        read_options_, env_options_, icomparator_, *fd, range_del_agg_,
-        nullptr /* don't need reference to table */, file_read_hist_,
-        for_compaction_, nullptr /* arena */, skip_filters_, level_);
+    return Status::OK();
+  }
+  virtual void SetPinnedItersMgr(
+      PinnedIteratorsManager* pinned_iters_mgr) override {
+    pinned_iters_mgr_ = pinned_iters_mgr;
+    if (file_iter_.iter()) {
+      file_iter_.SetPinnedItersMgr(pinned_iters_mgr);
+    }
+  }
+  virtual bool IsKeyPinned() const override {
+    return pinned_iters_mgr_ && pinned_iters_mgr_->PinningEnabled() &&
+           file_iter_.iter() && file_iter_.IsKeyPinned();
+  }
+  virtual bool IsValuePinned() const override {
+    return pinned_iters_mgr_ && pinned_iters_mgr_->PinningEnabled() &&
+           file_iter_.iter() && file_iter_.IsValuePinned();
  }

-  bool PrefixMayMatch(const Slice& internal_key) override {
-    return true;
+ private:
+  void SkipEmptyFileForward();
+  void SkipEmptyFileBackward();
+  void SetFileIterator(InternalIterator* iter);
+  void InitFileIterator(size_t new_file_index);
+
+  const Slice& file_smallest_key(size_t file_index) {
+    assert(file_index < flevel_->num_files);
+    return flevel_->files[file_index].smallest_key;
  }

-  bool KeyReachedUpperBound(const Slice& internal_key) override {
+  bool KeyReachedUpperBound(const Slice& internal_key) {
    return read_options_.iterate_upper_bound != nullptr &&
           icomparator_.user_comparator()->Compare(
               ExtractUserKey(internal_key),
               *read_options_.iterate_upper_bound) >= 0;
  }

- private:
+  InternalIterator* NewFileIterator() {
+    assert(file_index_ < flevel_->num_files);
+    auto file_meta = flevel_->files[file_index_];
+    if (should_sample_) {
+      sample_file_read_inc(file_meta.file_metadata);
+    }
+
+    return table_cache_->NewIterator(
+        read_options_, env_options_, icomparator_, file_meta.fd, range_del_agg_,
+        nullptr /* don't need reference to table */, file_read_hist_,
+        for_compaction_, nullptr /* arena */, skip_filters_, level_);
+  }
+
  TableCache* table_cache_;
  const ReadOptions read_options_;
  const EnvOptions& env_options_;
  const InternalKeyComparator& icomparator_;
+  const LevelFilesBrief* flevel_;
+  mutable FileDescriptor current_value_;
+
  HistogramImpl* file_read_hist_;
+  bool should_sample_;
  bool for_compaction_;
  bool skip_filters_;
+  size_t file_index_;
  int level_;
  RangeDelAggregator* range_del_agg_;
+  IteratorWrapper file_iter_;  // May be nullptr
+  PinnedIteratorsManager* pinned_iters_mgr_;
+  Status status_;
 };

+void LevelIterator::Seek(const Slice& target) {
+  size_t new_file_index = FindFile(icomparator_, *flevel_, target);
+
+  InitFileIterator(new_file_index);
+  if (file_iter_.iter() != nullptr) {
+    file_iter_.Seek(target);
+  }
+  SkipEmptyFileForward();
+}
+
+void LevelIterator::SeekForPrev(const Slice& target) {
+  size_t new_file_index = FindFile(icomparator_, *flevel_, target);
+  if (new_file_index >= flevel_->num_files) {
+    new_file_index = flevel_->num_files - 1;
+  }
+
+  InitFileIterator(new_file_index);
+  if (file_iter_.iter() != nullptr) {
+    file_iter_.SeekForPrev(target);
+    SkipEmptyFileBackward();
+  }
+}
+
+void LevelIterator::SeekToFirst() {
+  InitFileIterator(0);
+  if (file_iter_.iter() != nullptr) {
+    file_iter_.SeekToFirst();
+  }
+  SkipEmptyFileForward();
+}
+
+void LevelIterator::SeekToLast() {
+  InitFileIterator(flevel_->num_files - 1);
+  if (file_iter_.iter() != nullptr) {
+    file_iter_.SeekToLast();
+  }
+  SkipEmptyFileBackward();
+}
+
+void LevelIterator::Next() {
+  assert(Valid());
+  file_iter_.Next();
+  SkipEmptyFileForward();
+}
+
+void LevelIterator::Prev() {
+  assert(Valid());
+  file_iter_.Prev();
+  SkipEmptyFileBackward();
+}
+
+void LevelIterator::SkipEmptyFileForward() {
+  // For an error (IO error, checksum mismatch, etc), we skip the file
+  // and move to the next one and continue reading data.
+  // TODO this behavior is from LevelDB. We should revisit it.
+  while (file_iter_.iter() == nullptr ||
+         (!file_iter_.Valid() && !file_iter_.status().IsIncomplete())) {
+    if (file_iter_.iter() != nullptr && !file_iter_.Valid() &&
+        file_iter_.iter()->IsOutOfBound()) {
+      return;
+    }
+
+    // Move to next file
+    if (file_index_ >= flevel_->num_files - 1) {
+      // Already at the last file
+      SetFileIterator(nullptr);
+      return;
+    }
+    if (KeyReachedUpperBound(file_smallest_key(file_index_ + 1))) {
+      SetFileIterator(nullptr);
+      return;
+    }
+    InitFileIterator(file_index_ + 1);
+    if (file_iter_.iter() != nullptr) {
+      file_iter_.SeekToFirst();
+    }
+  }
+}
+
+void LevelIterator::SkipEmptyFileBackward() {
+  while (file_iter_.iter() == nullptr ||
+         (!file_iter_.Valid() && !file_iter_.status().IsIncomplete())) {
+    // Move to previous file
+    if (file_index_ == 0) {
+      // Already the first file
+      SetFileIterator(nullptr);
+      return;
+    }
+    InitFileIterator(file_index_ - 1);
+    if (file_iter_.iter() != nullptr) {
+      file_iter_.SeekToLast();
+    }
+  }
+}
+
+void LevelIterator::SetFileIterator(InternalIterator* iter) {
+  if (file_iter_.iter() != nullptr && status_.ok()) {
+    // TODO right now we don't invalidate the iterator even if the status is
+    // not OK. We should consider to do that so that it is harder for users to
+    // skip errors.
+    status_ = file_iter_.status();
+  }
+
+  if (pinned_iters_mgr_ && iter) {
+    iter->SetPinnedItersMgr(pinned_iters_mgr_);
+  }
+
+  InternalIterator* old_iter = file_iter_.Set(iter);
+  if (pinned_iters_mgr_ && pinned_iters_mgr_->PinningEnabled()) {
+    pinned_iters_mgr_->PinIterator(old_iter);
+  } else {
+    delete old_iter;
+  }
+}
+
+void LevelIterator::InitFileIterator(size_t new_file_index) {
+  if (new_file_index >= flevel_->num_files) {
+    file_index_ = new_file_index;
+    SetFileIterator(nullptr);
+    return;
+  } else {
+    // If the file iterator shows incomplete, we try it again if users seek
+    // to the same file, as this time we may go to a different data block
+    // which is cached in block cache.
+    //
+    if (file_iter_.iter() != nullptr && !file_iter_.status().IsIncomplete() &&
+        new_file_index == file_index_) {
+      // file_iter_ is already constructed with this iterator, so
+      // no need to change anything
+    } else {
+      file_index_ = new_file_index;
+      InternalIterator* iter = NewFileIterator();
+      SetFileIterator(iter);
+    }
+  }
+}
+
 // A wrapper of version builder which references the current version in
 // constructor and unref it in the destructor.
 // Both of the constructor and destructor need to be called inside DB Mutex.
@@ -854,24 +986,18 @@ void Version::AddIteratorsForLevel(const ReadOptions& read_options,
        sample_file_read_inc(meta);
      }
    }
-  } else {
+  } else if (storage_info_.LevelFilesBrief(level).num_files > 0) {
    // For levels > 0, we can use a concatenating iterator that sequentially
    // walks through the non-overlapping files in the level, opening them
    // lazily.
-    auto* mem = arena->AllocateAligned(sizeof(LevelFileIteratorState));
-    auto* state = new (mem)
-        LevelFileIteratorState(cfd_->table_cache(), read_options, soptions,
-                               cfd_->internal_comparator(),
-                               cfd_->internal_stats()->GetFileReadHist(level),
-                               false /* for_compaction */,
-                               cfd_->ioptions()->prefix_extractor != nullptr,
-                               IsFilterSkipped(level), level, range_del_agg);
-    mem = arena->AllocateAligned(sizeof(LevelFileNumIterator));
-    auto* first_level_iter = new (mem) LevelFileNumIterator(
+    auto* mem = arena->AllocateAligned(sizeof(LevelIterator));
+    merge_iter_builder->AddIterator(new (mem) LevelIterator(
+        cfd_->table_cache(), read_options, soptions,
        cfd_->internal_comparator(), &storage_info_.LevelFilesBrief(level),
-        should_sample_file_read());
-    merge_iter_builder->AddIterator(
-        NewTwoLevelIterator(state, first_level_iter, arena, false));
+        should_sample_file_read(),
+        cfd_->internal_stats()->GetFileReadHist(level),
+        false /* for_compaction */, IsFilterSkipped(level), level,
+        range_del_agg));
  }
 }

@@ -1976,7 +2102,7 @@ void VersionStorageInfo::ExtendFileRangeOverlappingInterval(
 #endif
  *start_index = mid_index + 1;
  *end_index = mid_index;
-  int count __attribute__((unused));
+  int count __attribute__((__unused__));
  count = 0;

  // check backwards from 'mid' to lower indices
@@ -3732,17 +3858,13 @@ InternalIterator* VersionSet::MakeInputIterator(
        }
      } else {
        // Create concatenating iterator for the files from this level
-        list[num++] = NewTwoLevelIterator(
-            new LevelFileIteratorState(
-                cfd->table_cache(), read_options, env_options_compactions,
-                cfd->internal_comparator(),
-                nullptr /* no per level latency histogram */,
-                true /* for_compaction */, false /* prefix enabled */,
-                false /* skip_filters */, (int)which /* level */,
-                range_del_agg),
-            new LevelFileNumIterator(cfd->internal_comparator(),
-                                     c->input_levels(which),
-                                     false /* don't sample compaction */));
+        list[num++] = new LevelIterator(
+            cfd->table_cache(), read_options, env_options_compactions,
+            cfd->internal_comparator(), c->input_levels(which),
+            false /* should_sample */,
+            nullptr /* no per level latency histogram */,
+            true /* for_compaction */, false /* skip_filters */,
+            (int)which /* level */, range_del_agg);
      }
    }
  }
@@ -39,6 +39,15 @@ namespace rocksdb {

 #ifndef ROCKSDB_LITE

+Status WalManager::DeleteFile(const std::string& fname, uint64_t number) {
+  auto s = env_->DeleteFile(db_options_.wal_dir + "/" + fname);
+  if (s.ok()) {
+    MutexLock l(&read_first_record_cache_mutex_);
+    read_first_record_cache_.erase(number);
+  }
+  return s;
+}
+
 Status WalManager::GetSortedWalFiles(VectorLogPtr& files) {
  // First get sorted files in db dir, then get sorted files from archived
  // dir, to avoid a race condition where a log file is moved to archived
@@ -49,6 +49,8 @@ class WalManager {

  void ArchiveWALFile(const std::string& fname, uint64_t number);

+  Status DeleteFile(const std::string& fname, uint64_t number);
+
  Status TEST_ReadFirstRecord(const WalFileType type, const uint64_t number,
                              SequenceNumber* sequence) {
    return ReadFirstRecord(type, number, sequence);
@@ -303,7 +303,7 @@ int main(int argc, char** argv) {
 #else
 #include <stdio.h>

-int main(int argc, char** argv) {
+int main(int /*argc*/, char** /*argv*/) {
  fprintf(stderr, "SKIPPED as WalManager is not supported in ROCKSDB_LITE\n");
  return 0;
 }
@@ -15,10 +15,12 @@
 //    kTypeValue varstring varstring
 //    kTypeDeletion varstring
 //    kTypeSingleDeletion varstring
+//    kTypeRangeDeletion varstring varstring
 //    kTypeMerge varstring varstring
 //    kTypeColumnFamilyValue varint32 varstring varstring
-//    kTypeColumnFamilyDeletion varint32 varstring varstring
-//    kTypeColumnFamilySingleDeletion varint32 varstring varstring
+//    kTypeColumnFamilyDeletion varint32 varstring
+//    kTypeColumnFamilySingleDeletion varint32 varstring
+//    kTypeColumnFamilyRangeDeletion varint32 varstring varstring
 //    kTypeColumnFamilyMerge varint32 varstring varstring
 //    kTypeBeginPrepareXID varstring
 //    kTypeEndPrepareXID
@@ -144,6 +146,12 @@ WriteBatch::WriteBatch(const std::string& rep)
      max_bytes_(0),
      rep_(rep) {}

+WriteBatch::WriteBatch(std::string&& rep)
+    : save_points_(nullptr),
+      content_flags_(ContentFlags::DEFERRED),
+      max_bytes_(0),
+      rep_(std::move(rep)) {}
+
 WriteBatch::WriteBatch(const WriteBatch& src)
    : save_points_(src.save_points_),
      wal_term_point_(src.wal_term_point_),
@@ -394,14 +402,22 @@ Status WriteBatch::Iterate(Handler* handler) const {
  bool empty_batch = true;
  int found = 0;
  Status s;
-  while (s.ok() && !input.empty() && handler->Continue()) {
-    char tag = 0;
-    uint32_t column_family = 0;  // default
+  char tag = 0;
+  uint32_t column_family = 0;  // default
+  while ((s.ok() || UNLIKELY(s.IsTryAgain())) && !input.empty() &&
+         handler->Continue()) {
+    if (LIKELY(!s.IsTryAgain())) {
+      tag = 0;
+      column_family = 0;  // default

-    s = ReadRecordFromWriteBatch(&input, &tag, &column_family, &key, &value,
-                                 &blob, &xid);
-    if (!s.ok()) {
-      return s;
+      s = ReadRecordFromWriteBatch(&input, &tag, &column_family, &key, &value,
+                                   &blob, &xid);
+      if (!s.ok()) {
+        return s;
+      }
+    } else {
+      assert(s.IsTryAgain());
+      s = Status::OK();
    }

    switch (tag) {
@@ -410,47 +426,59 @@ Status WriteBatch::Iterate(Handler* handler) const {
        assert(content_flags_.load(std::memory_order_relaxed) &
               (ContentFlags::DEFERRED | ContentFlags::HAS_PUT));
        s = handler->PutCF(column_family, key, value);
-        empty_batch = false;
-        found++;
+        if (LIKELY(s.ok())) {
+          empty_batch = false;
+          found++;
+        }
        break;
      case kTypeColumnFamilyDeletion:
      case kTypeDeletion:
        assert(content_flags_.load(std::memory_order_relaxed) &
               (ContentFlags::DEFERRED | ContentFlags::HAS_DELETE));
        s = handler->DeleteCF(column_family, key);
-        empty_batch = false;
-        found++;
+        if (LIKELY(s.ok())) {
+          empty_batch = false;
+          found++;
+        }
        break;
      case kTypeColumnFamilySingleDeletion:
      case kTypeSingleDeletion:
        assert(content_flags_.load(std::memory_order_relaxed) &
               (ContentFlags::DEFERRED | ContentFlags::HAS_SINGLE_DELETE));
        s = handler->SingleDeleteCF(column_family, key);
-        empty_batch = false;
-        found++;
+        if (LIKELY(s.ok())) {
+          empty_batch = false;
+          found++;
+        }
        break;
      case kTypeColumnFamilyRangeDeletion:
      case kTypeRangeDeletion:
        assert(content_flags_.load(std::memory_order_relaxed) &
               (ContentFlags::DEFERRED | ContentFlags::HAS_DELETE_RANGE));
        s = handler->DeleteRangeCF(column_family, key, value);
-        empty_batch = false;
-        found++;
+        if (LIKELY(s.ok())) {
+          empty_batch = false;
+          found++;
+        }
        break;
      case kTypeColumnFamilyMerge:
      case kTypeMerge:
        assert(content_flags_.load(std::memory_order_relaxed) &
               (ContentFlags::DEFERRED | ContentFlags::HAS_MERGE));
        s = handler->MergeCF(column_family, key, value);
-        empty_batch = false;
-        found++;
+        if (LIKELY(s.ok())) {
+          empty_batch = false;
+          found++;
+        }
        break;
      case kTypeColumnFamilyBlobIndex:
      case kTypeBlobIndex:
        assert(content_flags_.load(std::memory_order_relaxed) &
               (ContentFlags::DEFERRED | ContentFlags::HAS_BLOB_INDEX));
        s = handler->PutBlobIndexCF(column_family, key, value);
-        found++;
+        if (LIKELY(s.ok())) {
+          found++;
+        }
        break;
      case kTypeLogData:
        handler->LogData(blob);
@@ -547,6 +575,13 @@ size_t WriteBatchInternal::GetFirstOffset(WriteBatch* b) {

 Status WriteBatchInternal::Put(WriteBatch* b, uint32_t column_family_id,
                               const Slice& key, const Slice& value) {
+  if (key.size() > size_t{port::kMaxUint32}) {
+    return Status::InvalidArgument("key is too large");
+  }
+  if (value.size() > size_t{port::kMaxUint32}) {
+    return Status::InvalidArgument("value is too large");
+  }
+
  LocalSavePoint save(b);
  WriteBatchInternal::SetCount(b, WriteBatchInternal::Count(b) + 1);
  if (column_family_id == 0) {
@@ -569,8 +604,33 @@ Status WriteBatch::Put(ColumnFamilyHandle* column_family, const Slice& key,
                                 value);
 }

+Status WriteBatchInternal::CheckSlicePartsLength(const SliceParts& key,
+                                                 const SliceParts& value) {
+  size_t total_key_bytes = 0;
+  for (int i = 0; i < key.num_parts; ++i) {
+    total_key_bytes += key.parts[i].size();
+  }
+  if (total_key_bytes >= size_t{port::kMaxUint32}) {
+    return Status::InvalidArgument("key is too large");
+  }
+
+  size_t total_value_bytes = 0;
+  for (int i = 0; i < value.num_parts; ++i) {
+    total_value_bytes += value.parts[i].size();
+  }
+  if (total_value_bytes >= size_t{port::kMaxUint32}) {
+    return Status::InvalidArgument("value is too large");
+  }
+  return Status::OK();
+}
+
 Status WriteBatchInternal::Put(WriteBatch* b, uint32_t column_family_id,
                               const SliceParts& key, const SliceParts& value) {
+  Status s = CheckSlicePartsLength(key, value);
+  if (!s.ok()) {
+    return s;
+  }
+
  LocalSavePoint save(b);
  WriteBatchInternal::SetCount(b, WriteBatchInternal::Count(b) + 1);
  if (column_family_id == 0) {
@@ -787,6 +847,13 @@ Status WriteBatch::DeleteRange(ColumnFamilyHandle* column_family,

 Status WriteBatchInternal::Merge(WriteBatch* b, uint32_t column_family_id,
                                 const Slice& key, const Slice& value) {
+  if (key.size() > size_t{port::kMaxUint32}) {
+    return Status::InvalidArgument("key is too large");
+  }
+  if (value.size() > size_t{port::kMaxUint32}) {
+    return Status::InvalidArgument("value is too large");
+  }
+
  LocalSavePoint save(b);
  WriteBatchInternal::SetCount(b, WriteBatchInternal::Count(b) + 1);
  if (column_family_id == 0) {
@@ -812,6 +879,11 @@ Status WriteBatch::Merge(ColumnFamilyHandle* column_family, const Slice& key,
 Status WriteBatchInternal::Merge(WriteBatch* b, uint32_t column_family_id,
                                 const SliceParts& key,
                                 const SliceParts& value) {
+  Status s = CheckSlicePartsLength(key, value);
+  if (!s.ok()) {
+    return s;
+  }
+
  LocalSavePoint save(b);
  WriteBatchInternal::SetCount(b, WriteBatchInternal::Count(b) + 1);
  if (column_family_id == 0) {
@@ -1076,12 +1148,23 @@ class MemTableInserter : public WriteBatch::Handler {
      MaybeAdvanceSeq();
      return seek_status;
    }
+    Status ret_status;

    MemTable* mem = cf_mems_->GetMemTable();
    auto* moptions = mem->GetImmutableMemTableOptions();
+    // inplace_update_support is inconsistent with snapshots, and therefore with
+    // any kind of transactions including the ones that use seq_per_batch
+    assert(!seq_per_batch_ || !moptions->inplace_update_support);
    if (!moptions->inplace_update_support) {
-      mem->Add(sequence_, value_type, key, value, concurrent_memtable_writes_,
-               get_post_process_info(mem));
+      bool mem_res =
+          mem->Add(sequence_, value_type, key, value,
+                   concurrent_memtable_writes_, get_post_process_info(mem));
+      if (UNLIKELY(!mem_res)) {
+        assert(seq_per_batch_);
+        ret_status = Status::TryAgain("key+seq exists");
+        const bool BATCH_BOUNDRY = true;
+        MaybeAdvanceSeq(BATCH_BOUNDRY);
+      }
    } else if (moptions->inplace_callback == nullptr) {
      assert(!concurrent_memtable_writes_);
      mem->Update(sequence_, key, value);
@@ -1117,11 +1200,17 @@ class MemTableInserter : public WriteBatch::Handler {
                                                 value, &merged_value);
        if (status == UpdateStatus::UPDATED_INPLACE) {
          // prev_value is updated in-place with final value.
-          mem->Add(sequence_, value_type, key, Slice(prev_buffer, prev_size));
+          bool mem_res __attribute__((__unused__));
+          mem_res = mem->Add(
+              sequence_, value_type, key, Slice(prev_buffer, prev_size));
+          assert(mem_res);
          RecordTick(moptions->statistics, NUMBER_KEYS_WRITTEN);
        } else if (status == UpdateStatus::UPDATED) {
          // merged_value contains the final value.
-          mem->Add(sequence_, value_type, key, Slice(merged_value));
+          bool mem_res __attribute__((__unused__));
+          mem_res =
+              mem->Add(sequence_, value_type, key, Slice(merged_value));
+          assert(mem_res);
          RecordTick(moptions->statistics, NUMBER_KEYS_WRITTEN);
        }
      }
@@ -1131,7 +1220,7 @@ class MemTableInserter : public WriteBatch::Handler {
    // in memtable add/update.
    MaybeAdvanceSeq();
    CheckMemtableFull();
-    return Status::OK();
+    return ret_status;
  }

  virtual Status PutCF(uint32_t column_family_id, const Slice& key,
@@ -1141,12 +1230,20 @@ class MemTableInserter : public WriteBatch::Handler {

  Status DeleteImpl(uint32_t column_family_id, const Slice& key,
                    const Slice& value, ValueType delete_type) {
+    Status ret_status;
    MemTable* mem = cf_mems_->GetMemTable();
-    mem->Add(sequence_, delete_type, key, value, concurrent_memtable_writes_,
-             get_post_process_info(mem));
+    bool mem_res =
+        mem->Add(sequence_, delete_type, key, value,
+                 concurrent_memtable_writes_, get_post_process_info(mem));
+    if (UNLIKELY(!mem_res)) {
+      assert(seq_per_batch_);
+      ret_status = Status::TryAgain("key+seq exists");
+      const bool BATCH_BOUNDRY = true;
+      MaybeAdvanceSeq(BATCH_BOUNDRY);
+    }
    MaybeAdvanceSeq();
    CheckMemtableFull();
-    return Status::OK();
+    return ret_status;
  }

  virtual Status DeleteCF(uint32_t column_family_id,
@@ -1238,6 +1335,7 @@ class MemTableInserter : public WriteBatch::Handler {
      return seek_status;
    }

+    Status ret_status;
    MemTable* mem = cf_mems_->GetMemTable();
    auto* moptions = mem->GetImmutableMemTableOptions();
    bool perform_merge = false;
@@ -1293,18 +1391,30 @@ class MemTableInserter : public WriteBatch::Handler {
        perform_merge = false;
      } else {
        // 3) Add value to memtable
-        mem->Add(sequence_, kTypeValue, key, new_value);
+        bool mem_res = mem->Add(sequence_, kTypeValue, key, new_value);
+        if (UNLIKELY(!mem_res)) {
+          assert(seq_per_batch_);
+          ret_status = Status::TryAgain("key+seq exists");
+          const bool BATCH_BOUNDRY = true;
+          MaybeAdvanceSeq(BATCH_BOUNDRY);
+        }
      }
    }

    if (!perform_merge) {
      // Add merge operator to memtable
-      mem->Add(sequence_, kTypeMerge, key, value);
+      bool mem_res = mem->Add(sequence_, kTypeMerge, key, value);
+      if (UNLIKELY(!mem_res)) {
+        assert(seq_per_batch_);
+        ret_status = Status::TryAgain("key+seq exists");
+        const bool BATCH_BOUNDRY = true;
+        MaybeAdvanceSeq(BATCH_BOUNDRY);
+      }
    }

    MaybeAdvanceSeq();
    CheckMemtableFull();
-    return Status::OK();
+    return ret_status;
  }

  virtual Status PutBlobIndexCF(uint32_t column_family_id, const Slice& key,
@@ -1488,6 +1598,8 @@ Status WriteBatchInternal::InsertInto(
    if (!w->status.ok()) {
      return w->status;
    }
+    assert(!seq_per_batch || w->batch_cnt != 0);
+    assert(!seq_per_batch || inserter.sequence() - w->sequence == w->batch_cnt);
  }
  return Status::OK();
 }
@@ -1496,7 +1608,7 @@ Status WriteBatchInternal::InsertInto(
    WriteThread::Writer* writer, SequenceNumber sequence,
    ColumnFamilyMemTables* memtables, FlushScheduler* flush_scheduler,
    bool ignore_missing_column_families, uint64_t log_number, DB* db,
-    bool concurrent_memtable_writes, bool seq_per_batch) {
+    bool concurrent_memtable_writes, bool seq_per_batch, size_t batch_cnt) {
  assert(writer->ShouldWriteToMemtable());
  MemTableInserter inserter(sequence, memtables, flush_scheduler,
                            ignore_missing_column_families, log_number, db,
@@ -1505,6 +1617,8 @@ Status WriteBatchInternal::InsertInto(
  SetSequence(writer->batch, sequence);
  inserter.set_log_number_ref(writer->log_ref);
  Status s = writer->batch->Iterate(&inserter);
+  assert(!seq_per_batch || batch_cnt != 0);
+  assert(!seq_per_batch || inserter.sequence() - sequence == batch_cnt);
  if (concurrent_memtable_writes) {
    inserter.PostProcess();
  }
@@ -138,6 +138,9 @@ class WriteBatchInternal {

  static Status SetContents(WriteBatch* batch, const Slice& contents);

+  static Status CheckSlicePartsLength(const SliceParts& key,
+                                      const SliceParts& value);
+
  // Inserts batches[i] into memtable, for i in 0..num_batches-1 inclusive.
  //
  // If ignore_missing_column_families == true. WriteBatch
@@ -182,7 +185,7 @@ class WriteBatchInternal {
                           bool ignore_missing_column_families = false,
                           uint64_t log_number = 0, DB* db = nullptr,
                           bool concurrent_memtable_writes = false,
-                           bool seq_per_batch = false);
+                           bool seq_per_batch = false, size_t batch_cnt = 0);

  static Status Append(WriteBatch* dst, const WriteBatch* src,
                       const bool WAL_only = false);
@@ -291,7 +291,7 @@ TEST_F(WriteCallbackTest, WriteWithCallbackTest) {
                woptions.disableWAL = !enable_WAL;
                woptions.sync = enable_WAL;
                Status s;
-                if (seq_per_batch && two_queues) {
+                if (seq_per_batch) {
                  class PublishSeqCallback : public PreReleaseCallback {
                   public:
                    PublishSeqCallback(DBImpl* db_impl_in)
@@ -302,9 +302,13 @@ TEST_F(WriteCallbackTest, WriteWithCallbackTest) {
                    }
                    DBImpl* db_impl_;
                  } publish_seq_callback(db_impl);
-                  s = db_impl->WriteImpl(woptions, &write_op.write_batch_,
-                                         &write_op.callback_, nullptr, 0, false,
-                                         nullptr, &publish_seq_callback);
+                  // seq_per_batch requires a natural batch separator or Noop
+                  WriteBatchInternal::InsertNoop(&write_op.write_batch_);
+                  const size_t ONE_BATCH = 1;
+                  s = db_impl->WriteImpl(
+                      woptions, &write_op.write_batch_, &write_op.callback_,
+                      nullptr, 0, false, nullptr, ONE_BATCH,
+                      two_queues ? &publish_seq_callback : nullptr);
                } else {
                  s = db_impl->WriteWithCallback(
                      woptions, &write_op.write_batch_, &write_op.callback_);
@@ -429,7 +433,7 @@ int main(int argc, char** argv) {
 #else
 #include <stdio.h>

-int main(int argc, char** argv) {
+int main(int /*argc*/, char** /*argv*/) {
  fprintf(stderr,
          "SKIPPED as WriteWithCallback is not supported in ROCKSDB_LITE\n");
  return 0;
@@ -118,6 +118,7 @@ class WriteThread {
    bool no_slowdown;
    bool disable_wal;
    bool disable_memtable;
+    size_t batch_cnt;  // if non-zero, number of sub-batches in the write batch
    PreReleaseCallback* pre_release_callback;
    uint64_t log_used;  // log number that this batch was inserted into
    uint64_t log_ref;   // log number that memtable insert should reference
@@ -128,6 +129,7 @@ class WriteThread {
    SequenceNumber sequence;  // the sequence number to use for the first key
    Status status;            // status of memtable inserter
    Status callback_status;   // status returned by callback->Callback()
+
    std::aligned_storage<sizeof(std::mutex)>::type state_mutex_bytes;
    std::aligned_storage<sizeof(std::condition_variable)>::type state_cv_bytes;
    Writer* link_older;  // read/write only before linking, or as leader
@@ -139,6 +141,7 @@ class WriteThread {
          no_slowdown(false),
          disable_wal(false),
          disable_memtable(false),
+          batch_cnt(0),
          pre_release_callback(nullptr),
          log_used(0),
          log_ref(0),
@@ -152,12 +155,14 @@ class WriteThread {

    Writer(const WriteOptions& write_options, WriteBatch* _batch,
           WriteCallback* _callback, uint64_t _log_ref, bool _disable_memtable,
+           size_t _batch_cnt = 0,
           PreReleaseCallback* _pre_release_callback = nullptr)
        : batch(_batch),
          sync(write_options.sync),
          no_slowdown(write_options.no_slowdown),
          disable_wal(write_options.disableWAL),
          disable_memtable(_disable_memtable),
+          batch_cnt(_batch_cnt),
          pre_release_callback(_pre_release_callback),
          log_used(0),
          log_ref(_log_ref),