update HISTORY.md for 8.0.0

Summary:
I couldn't figure out why this causes failures in our 8.0 release to fbcode while this issue appears to not be new in 8.0. Anyways, we can add the missing `override` keywords to these functions as the compiler insists.

Pull Request resolved: https://github.com/facebook/rocksdb/pull/11232

Reviewed By: pdillinger

Differential Revision: D43420656

Pulled By: ajkr

fbshipit-source-id: da748eeef6ba38dd113dbe4b5143d7558daf38dd

.circleci/config.yml

@@ -152,15 +152,15 @@ commands:
     steps:
       - run:
           name: "Test low-variance benchmarks"
-          command: ./tools/benchmark_ci.py --db_dir /tmp/rocksdb-benchmark-datadir --output_dir /tmp/benchmark-results --num_keys 10000000
+          command: ./tools/benchmark_ci.py --db_dir /tmp/rocksdb-benchmark-datadir --output_dir /tmp/benchmark-results --num_keys 20000000
           environment:
             LD_LIBRARY_PATH: /usr/local/lib
             # How long to run parts of the test(s)
-            DURATION_RO: 400
-            DURATION_RW: 700
+            DURATION_RO: 300
+            DURATION_RW: 500
             # Keep threads within physical capacity of server (much lower than default)
             NUM_THREADS: 1
-            MAX_BACKGROUND_JOBS: 3
+            MAX_BACKGROUND_JOBS: 4
             # Don't run a couple of "optional" initial tests
             CI_TESTS_ONLY: "true"
             # Reduce configured size of levels to ensure more levels in the leveled compaction LSM tree
@@ -170,7 +170,11 @@ commands:
             # The benchmark host has 32GB memory
             # The following values are tailored to work with that
             # Note, tests may not exercise the targeted issues if the memory is increased on new test hosts.
-
+            COMPRESSION_TYPE: "none"
+            CACHE_INDEX_AND_FILTER_BLOCKS: 1
+            MIN_LEVEL_TO_COMPRESS: 3
+            CACHE_SIZE_MB: 10240
+            MB_WRITE_PER_SEC: 2
 
   post-benchmarks:
     steps:
@@ -269,12 +273,12 @@ jobs:
           ./sst_dump --help | grep -E -q 'Supported compression types: kNoCompression$' # Verify no compiled in compression
       - post-steps
 
-  build-linux-shared_lib-alt_namespace-status_checked:
+  build-linux-static_lib-alt_namespace-status_checked:
     executor: linux-docker
     resource_class: 2xlarge
     steps:
       - pre-steps
-      - run: ASSERT_STATUS_CHECKED=1 TEST_UINT128_COMPAT=1 ROCKSDB_MODIFY_NPHASH=1 LIB_MODE=shared OPT="-DROCKSDB_NAMESPACE=alternative_rocksdb_ns" make V=1 -j32 check
+      - run: ASSERT_STATUS_CHECKED=1 TEST_UINT128_COMPAT=1 ROCKSDB_MODIFY_NPHASH=1 LIB_MODE=static OPT="-DROCKSDB_NAMESPACE=alternative_rocksdb_ns" make V=1 -j24 check
       - post-steps
 
   build-linux-release:
@@ -282,11 +286,21 @@ jobs:
     resource_class: 2xlarge
     steps:
       - checkout # check out the code in the project directory
+      - run: make V=1 -j32 LIB_MODE=shared release
+      - run: ls librocksdb.so # ensure shared lib built
+      - run: ./db_stress --version # ensure with gflags
+      - run: make clean
       - run: make V=1 -j32 release
+      - run: ls librocksdb.a # ensure static lib built
       - run: ./db_stress --version # ensure with gflags
       - run: make clean
       - run: apt-get remove -y libgflags-dev
+      - run: make V=1 -j32 LIB_MODE=shared release
+      - run: ls librocksdb.so # ensure shared lib built
+      - run: if ./db_stress --version; then false; else true; fi # ensure without gflags
+      - run: make clean
       - run: make V=1 -j32 release
+      - run: ls librocksdb.a # ensure static lib built
       - run: if ./db_stress --version; then false; else true; fi # ensure without gflags
       - post-steps
 
@@ -406,7 +420,10 @@ jobs:
     steps:
       - checkout # check out the code in the project directory
       - run: apt-get update -y && apt-get install -y libgflags-dev
-      - run: make V=1 -j8 unity_test
+      - run:
+          name: "Unity build"
+          command: make V=1 -j8 unity_test
+          no_output_timeout: 20m
       - run: make V=1 -j8 -k check-headers # could be moved to a different build
       - post-steps
 
@@ -417,7 +434,7 @@ jobs:
       - pre-steps
       - setup-folly
       - build-folly
-      - run: USE_FOLLY=1 CC=gcc-7 CXX=g++-7 V=1 make -j32 check
+      - run: USE_FOLLY=1 LIB_MODE=static CC=gcc-7 CXX=g++-7 V=1 make -j32 check # TODO: LIB_MODE only to work around unresolved linker failures
       - post-steps
 
   build-linux-gcc-7-with-folly-lite-no-test:
@@ -463,7 +480,7 @@ jobs:
     resource_class: 2xlarge
     steps:
       - pre-steps
-      - run: CC=gcc-11 CXX=g++-11 V=1 make -j32 all microbench
+      - run: LIB_MODE=static CC=gcc-11 CXX=g++-11 V=1 make -j32 all microbench # TODO: LIB_MODE only to work around unresolved linker failures
       - post-steps
 
   build-linux-clang-13-no_test_run:
@@ -482,7 +499,7 @@ jobs:
       - pre-steps
       - setup-folly
       - build-folly
-      - run: CC=clang-13 CXX=clang++-13 USE_CLANG=1 USE_FOLLY=1 COMPILE_WITH_UBSAN=1 COMPILE_WITH_ASAN=1 make -j32 check
+      - run: CC=clang-13 CXX=clang++-13 LIB_MODE=static USE_CLANG=1 USE_FOLLY=1 COMPILE_WITH_UBSAN=1 COMPILE_WITH_ASAN=1 make -j32 check # TODO: LIB_MODE only to work around unresolved linker failures
       - post-steps
 
   # This job is only to make sure the microbench tests are able to run, the benchmark result is not meaningful as the CI host is changing.
@@ -499,7 +516,7 @@ jobs:
     resource_class: large
     steps:
       - pre-steps
-      - run: ulimit -S -n `ulimit -H -n` && make V=1 -j8 CRASH_TEST_EXT_ARGS='--duration=960 --max_key=2500000' blackbox_crash_test_with_atomic_flush
+      - run: ulimit -S -n `ulimit -H -n` && make V=1 -j8 CRASH_TEST_EXT_ARGS='--duration=960 --max_key=2500000 --async_io=0' blackbox_crash_test_with_atomic_flush
       - post-steps
 
   build-linux-crashtest-tiered-storage-bb:
@@ -509,7 +526,7 @@ jobs:
       - pre-steps
       - run:
           name: "run crashtest"
-          command: ulimit -S -n `ulimit -H -n` && make V=1 -j32 CRASH_TEST_EXT_ARGS=--duration=10800 blackbox_crash_test_with_tiered_storage
+          command: ulimit -S -n `ulimit -H -n` && make V=1 -j32 CRASH_TEST_EXT_ARGS='--duration=10800 --async_io=0' blackbox_crash_test_with_tiered_storage
           no_output_timeout: 100m
       - post-steps
 
@@ -520,7 +537,7 @@ jobs:
       - pre-steps
       - run:
           name: "run crashtest"
-          command: ulimit -S -n `ulimit -H -n` && make V=1 -j32 CRASH_TEST_EXT_ARGS=--duration=10800 whitebox_crash_test_with_tiered_storage
+          command: ulimit -S -n `ulimit -H -n` && make V=1 -j32 CRASH_TEST_EXT_ARGS='--duration=10800 --async_io=0' whitebox_crash_test_with_tiered_storage
           no_output_timeout: 100m
       - post-steps
 
@@ -805,7 +822,7 @@ workflows:
       - build-linux-clang10-asan
       - build-linux-clang10-ubsan
       - build-linux-clang10-mini-tsan
-      - build-linux-shared_lib-alt_namespace-status_checked
+      - build-linux-static_lib-alt_namespace-status_checked
   jobs-linux-no-test-run:
     jobs:
       - build-linux-release

CMakeLists.txt

@@ -879,6 +879,7 @@ set(SOURCES
         util/compression_context_cache.cc
         util/concurrent_task_limiter_impl.cc
         util/crc32c.cc
+        util/data_structure.cc
         util/dynamic_bloom.cc
         util/hash.cc
         util/murmurhash.cc

HISTORY.md

@@ -1,8 +1,8 @@
 # Rocksdb Change Log
-## Unreleased
-
+## 8.0.0 (02/19/2023)
 ### Behavior changes
 * `ReadOptions::verify_checksums=false` disables checksum verification for more reads of non-`CacheEntryRole::kDataBlock` blocks.
+* In case of scan with async_io enabled, if posix doesn't support IOUring, Status::NotSupported error will be returned to the users. Initially that error was swallowed and reads were switched to synchronous reads.
 
 ### Bug Fixes
 * Fixed a data race on `ColumnFamilyData::flush_reason` caused by concurrent flushes.
@@ -10,6 +10,9 @@
 * Fixed some atypical behaviors for `LockWAL()` such as allowing concurrent/recursive use and not expecting `UnlockWAL()` after non-OK result. See API comments.
 * Fixed a feature interaction bug where for blobs `GetEntity` would expose the blob reference instead of the blob value.
 * Fixed `DisableManualCompaction()` and `CompactRangeOptions::canceled` to cancel compactions even when they are waiting on conflicting compactions to finish
+* Fixed a bug in which a successful `GetMergeOperands()` could transiently return `Status::MergeInProgress()`
+* Return the correct error (Status::NotSupported()) to MultiGet caller when ReadOptions::async_io flag is true and IO uring is not enabled. Previously, Status::Corruption() was being returned when the actual failure was lack of async IO support.
+* Fixed a bug in DB open/recovery from a compressed WAL that was caused due to incorrect handling of certain record fragments with the same offset within a WAL block.
 
 ### Feature Removal
 * Remove RocksDB Lite.
@@ -17,9 +20,22 @@
 * Remove deprecated Env::LoadEnv(). Use Env::CreateFromString() instead.
 * Remove deprecated FileSystem::Load(). Use FileSystem::CreateFromString() instead.
 * Removed the deprecated version of these utility functions and the corresponding Java bindings: `LoadOptionsFromFile`, `LoadLatestOptions`, `CheckOptionsCompatibility`.
+* Remove the FactoryFunc from the LoadObject method from the Customizable helper methods.
 
 ### Public API Changes
+* Moved rarely-needed Cache class definition to new advanced_cache.h, and added a CacheWrapper class to advanced_cache.h. Minor changes to SimCache API definitions.
 * Completely removed the following deprecated/obsolete statistics: the tickers `BLOCK_CACHE_INDEX_BYTES_EVICT`, `BLOCK_CACHE_FILTER_BYTES_EVICT`, `BLOOM_FILTER_MICROS`, `NO_FILE_CLOSES`, `STALL_L0_SLOWDOWN_MICROS`, `STALL_MEMTABLE_COMPACTION_MICROS`, `STALL_L0_NUM_FILES_MICROS`, `RATE_LIMIT_DELAY_MILLIS`, `NO_ITERATORS`, `NUMBER_FILTERED_DELETES`, `WRITE_TIMEDOUT`, `BLOB_DB_GC_NUM_KEYS_OVERWRITTEN`, `BLOB_DB_GC_NUM_KEYS_EXPIRED`, `BLOB_DB_GC_BYTES_OVERWRITTEN`, `BLOB_DB_GC_BYTES_EXPIRED`, `BLOCK_CACHE_COMPRESSION_DICT_BYTES_EVICT` as well as the histograms `STALL_L0_SLOWDOWN_COUNT`, `STALL_MEMTABLE_COMPACTION_COUNT`, `STALL_L0_NUM_FILES_COUNT`, `HARD_RATE_LIMIT_DELAY_COUNT`, `SOFT_RATE_LIMIT_DELAY_COUNT`, `BLOB_DB_GC_MICROS`, and `NUM_DATA_BLOCKS_READ_PER_LEVEL`. Note that as a result, the C++ enum values of the still supported statistics have changed. Developers are advised to not rely on the actual numeric values.
+* Deprecated IngestExternalFileOptions::write_global_seqno and change default to false. This option only needs to be set to true to generate a DB compatible with RocksDB versions before 5.16.0.
+* Remove deprecated APIs `GetColumnFamilyOptionsFrom{Map|String}(const ColumnFamilyOptions&, ..)`, `GetDBOptionsFrom{Map|String}(const DBOptions&, ..)`, `GetBlockBasedTableOptionsFrom{Map|String}(const BlockBasedTableOptions& table_options, ..)` and ` GetPlainTableOptionsFrom{Map|String}(const PlainTableOptions& table_options,..)`.
+* Added a subcode of `Status::Corruption`, `Status::SubCode::kMergeOperatorFailed`, for users to identify corruption failures originating in the merge operator, as opposed to RocksDB's internally identified data corruptions
+
+### Build Changes
+* The `make` build now builds a shared library by default instead of a static library. Use `LIB_MODE=static` to override.
+
+### New Features
+* Compaction filters are now supported for wide-column entities by means of the `FilterV3` API. See the comment of the API for more details.
+* Added `do_not_compress_roles` to `CompressedSecondaryCacheOptions` to disable compression on certain kinds of block. Filter blocks are now not compressed by CompressedSecondaryCache by default.
+* Added a new `MultiGetEntity` API that enables batched wide-column point lookups. See the API comments for more details.
 
 ## 7.10.0 (01/23/2023)
 ### Behavior changes

Makefile

@@ -44,13 +44,6 @@ quoted_perl_command = $(subst ','\'',$(perl_command))
 # Set the default DEBUG_LEVEL to 1
 DEBUG_LEVEL?=1
 
-# LIB_MODE says whether or not to use/build "shared" or "static" libraries.
-# Mode "static" means to link against static libraries (.a)
-# Mode "shared" means to link against shared libraries (.so, .sl, .dylib, etc)
-#
-# Set the default LIB_MODE to static
-LIB_MODE?=static
-
 # OBJ_DIR is where the object files reside.  Default to the current directory
 OBJ_DIR?=.
 
@@ -81,7 +74,19 @@ else ifneq ($(filter jtest rocksdbjava%, $(MAKECMDGOALS)),)
 	endif
 endif
 
-$(info $$DEBUG_LEVEL is ${DEBUG_LEVEL})
+# LIB_MODE says whether or not to use/build "shared" or "static" libraries.
+# Mode "static" means to link against static libraries (.a)
+# Mode "shared" means to link against shared libraries (.so, .sl, .dylib, etc)
+#
+ifeq ($(DEBUG_LEVEL), 0)
+# For optimized, set the default LIB_MODE to static for code size/efficiency
+	LIB_MODE?=static
+else
+# For debug, set the default LIB_MODE to shared for efficient `make check` etc.
+	LIB_MODE?=shared
+endif
+
+$(info $$DEBUG_LEVEL is $(DEBUG_LEVEL), $$LIB_MODE is $(LIB_MODE))
 
 # Figure out optimize level.
 ifneq ($(DEBUG_LEVEL), 2)
@@ -1217,9 +1222,9 @@ clean: clean-ext-libraries-all clean-rocks clean-rocksjava
 clean-not-downloaded: clean-ext-libraries-bin clean-rocks clean-not-downloaded-rocksjava
 
 clean-rocks:
-	echo shared=$(ALL_SHARED_LIBS)
-	echo static=$(ALL_STATIC_LIBS)
-	rm -f $(BENCHMARKS) $(TOOLS) $(TESTS) $(PARALLEL_TEST) $(ALL_STATIC_LIBS) $(ALL_SHARED_LIBS) $(MICROBENCHS)
+# Not practical to exactly match all versions/variants in naming (e.g. debug or not)
+	rm -f ${LIBNAME}*.so* ${LIBNAME}*.a
+	rm -f $(BENCHMARKS) $(TOOLS) $(TESTS) $(PARALLEL_TEST) $(MICROBENCHS)
 	rm -rf $(CLEAN_FILES) ios-x86 ios-arm scan_build_report
 	$(FIND) . -name "*.[oda]" -exec rm -f {} \;
 	$(FIND) . -type f \( -name "*.gcda" -o -name "*.gcno" \) -exec rm -f {} \;

TARGETS

@@ -247,6 +247,7 @@ cpp_library_wrapper(name="rocksdb_lib", srcs=[
         "util/concurrent_task_limiter_impl.cc",
         "util/crc32c.cc",
         "util/crc32c_arm64.cc",
+        "util/data_structure.cc",
         "util/dynamic_bloom.cc",
         "util/file_checksum_helper.cc",
         "util/hash.cc",
@@ -589,6 +590,7 @@ cpp_library_wrapper(name="rocksdb_whole_archive_lib", srcs=[
         "util/concurrent_task_limiter_impl.cc",
         "util/crc32c.cc",
         "util/crc32c_arm64.cc",
+        "util/data_structure.cc",
         "util/dynamic_bloom.cc",
         "util/file_checksum_helper.cc",
         "util/hash.cc",

cache/cache.cc

@@ -87,8 +87,7 @@ Status SecondaryCache::CreateFromString(
     }
     return status;
   } else {
-    return LoadSharedObject<SecondaryCache>(config_options, value, nullptr,
-                                            result);
+    return LoadSharedObject<SecondaryCache>(config_options, value, result);
   }
 }
 

cache/cache_bench_tool.cc

@@ -16,7 +16,7 @@
 #include "db/db_impl/db_impl.h"
 #include "monitoring/histogram.h"
 #include "port/port.h"
-#include "rocksdb/cache.h"
+#include "rocksdb/advanced_cache.h"
 #include "rocksdb/convenience.h"
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"

cache/cache_helpers.h

@@ -7,7 +7,7 @@
 
 #include <cassert>
 
-#include "rocksdb/cache.h"
+#include "rocksdb/advanced_cache.h"
 #include "rocksdb/rocksdb_namespace.h"
 
 namespace ROCKSDB_NAMESPACE {

cache/cache_key.cc

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

cache/charged_cache.h

@@ -8,7 +8,7 @@
 #include <string>
 
 #include "port/port.h"
-#include "rocksdb/cache.h"
+#include "rocksdb/advanced_cache.h"
 
 namespace ROCKSDB_NAMESPACE {
 

cache/clock_cache.cc

@@ -9,7 +9,6 @@
 
 #include "cache/clock_cache.h"
 
-#include <cassert>
 #include <functional>
 #include <numeric>
 
@@ -18,6 +17,7 @@
 #include "monitoring/perf_context_imp.h"
 #include "monitoring/statistics.h"
 #include "port/lang.h"
+#include "rocksdb/env.h"
 #include "util/hash.h"
 #include "util/math.h"
 #include "util/random.h"

cache/compressed_secondary_cache.cc

@@ -22,12 +22,13 @@ CompressedSecondaryCache::CompressedSecondaryCache(
     std::shared_ptr<MemoryAllocator> memory_allocator, bool use_adaptive_mutex,
     CacheMetadataChargePolicy metadata_charge_policy,
     CompressionType compression_type, uint32_t compress_format_version,
-    bool enable_custom_split_merge)
+    bool enable_custom_split_merge,
+    const CacheEntryRoleSet& do_not_compress_roles)
     : cache_options_(capacity, num_shard_bits, strict_capacity_limit,
                      high_pri_pool_ratio, low_pri_pool_ratio, memory_allocator,
                      use_adaptive_mutex, metadata_charge_policy,
                      compression_type, compress_format_version,
-                     enable_custom_split_merge) {
+                     enable_custom_split_merge, do_not_compress_roles) {
   cache_ =
       NewLRUCache(capacity, num_shard_bits, strict_capacity_limit,
                   high_pri_pool_ratio, memory_allocator, use_adaptive_mutex,
@@ -71,7 +72,8 @@ std::unique_ptr<SecondaryCacheResultHandle> CompressedSecondaryCache::Lookup(
   Status s;
   Cache::ObjectPtr value{nullptr};
   size_t charge{0};
-  if (cache_options_.compression_type == kNoCompression) {
+  if (cache_options_.compression_type == kNoCompression ||
+      cache_options_.do_not_compress_roles.Contains(helper->role)) {
     s = helper->create_cb(Slice(ptr->get(), handle_value_charge),
                           create_context, allocator, &value, &charge);
   } else {
@@ -143,7 +145,8 @@ Status CompressedSecondaryCache::Insert(const Slice& key,
   Slice val(ptr.get(), size);
 
   std::string compressed_val;
-  if (cache_options_.compression_type != kNoCompression) {
+  if (cache_options_.compression_type != kNoCompression &&
+      !cache_options_.do_not_compress_roles.Contains(helper->role)) {
     PERF_COUNTER_ADD(compressed_sec_cache_uncompressed_bytes, size);
     CompressionOptions compression_opts;
     CompressionContext compression_context(cache_options_.compression_type);
@@ -314,12 +317,13 @@ std::shared_ptr<SecondaryCache> NewCompressedSecondaryCache(
     std::shared_ptr<MemoryAllocator> memory_allocator, bool use_adaptive_mutex,
     CacheMetadataChargePolicy metadata_charge_policy,
     CompressionType compression_type, uint32_t compress_format_version,
-    bool enable_custom_split_merge) {
+    bool enable_custom_split_merge,
+    const CacheEntryRoleSet& do_not_compress_roles) {
   return std::make_shared<CompressedSecondaryCache>(
       capacity, num_shard_bits, strict_capacity_limit, high_pri_pool_ratio,
       low_pri_pool_ratio, memory_allocator, use_adaptive_mutex,
       metadata_charge_policy, compression_type, compress_format_version,
-      enable_custom_split_merge);
+      enable_custom_split_merge, do_not_compress_roles);
 }
 
 std::shared_ptr<SecondaryCache> NewCompressedSecondaryCache(
@@ -331,7 +335,7 @@ std::shared_ptr<SecondaryCache> NewCompressedSecondaryCache(
       opts.high_pri_pool_ratio, opts.low_pri_pool_ratio, opts.memory_allocator,
       opts.use_adaptive_mutex, opts.metadata_charge_policy,
       opts.compression_type, opts.compress_format_version,
-      opts.enable_custom_split_merge);
+      opts.enable_custom_split_merge, opts.do_not_compress_roles);
 }
 
 }  // namespace ROCKSDB_NAMESPACE

cache/compressed_secondary_cache.h

@@ -78,7 +78,9 @@ class CompressedSecondaryCache : public SecondaryCache {
           kDefaultCacheMetadataChargePolicy,
       CompressionType compression_type = CompressionType::kLZ4Compression,
       uint32_t compress_format_version = 2,
-      bool enable_custom_split_merge = false);
+      bool enable_custom_split_merge = false,
+      const CacheEntryRoleSet& do_not_compress_roles = {
+          CacheEntryRole::kFilterBlock});
   ~CompressedSecondaryCache() override;
 
   const char* Name() const override { return "CompressedSecondaryCache"; }

cache/compressed_secondary_cache_test.cc

@@ -5,6 +5,7 @@
 
 #include "cache/compressed_secondary_cache.h"
 
+#include <array>
 #include <iterator>
 #include <memory>
 #include <tuple>
@@ -75,12 +76,23 @@ class CompressedSecondaryCacheTest : public testing::Test,
     return Status::OK();
   }
 
-  static constexpr Cache::CacheItemHelper kHelper{
-      CacheEntryRole::kMisc, &DeletionCallback, &SizeCallback, &SaveToCallback,
-      &CreateCallback};
+  static constexpr auto GenerateHelpersByRole() {
+    std::array<Cache::CacheItemHelper, kNumCacheEntryRoles> a;
+    for (uint32_t i = 0; i < kNumCacheEntryRoles; ++i) {
+      a[i] = Cache::CacheItemHelper{static_cast<CacheEntryRole>(i),
+                                    &DeletionCallback, &SizeCallback,
+                                    &SaveToCallback, &CreateCallback};
+    }
+    return a;
+  }
+
+  static const std::array<Cache::CacheItemHelper, kNumCacheEntryRoles>
+      kHelperByRole;
+
+  static const Cache::CacheItemHelper& kHelper;
 
   static constexpr Cache::CacheItemHelper kHelperFail{
-      CacheEntryRole::kMisc, &DeletionCallback, &SizeCallback,
+      CacheEntryRole::kDataBlock, &DeletionCallback, &SizeCallback,
       &SaveToCallbackFail, &CreateCallback};
 
   void SetFailCreate(bool fail) { fail_create_ = fail; }
@@ -787,6 +799,13 @@ class CompressedSecondaryCacheTest : public testing::Test,
   bool fail_create_;
 };
 
+const std::array<Cache::CacheItemHelper, kNumCacheEntryRoles>
+    CompressedSecondaryCacheTest::kHelperByRole = GenerateHelpersByRole();
+
+const Cache::CacheItemHelper& CompressedSecondaryCacheTest::kHelper =
+    CompressedSecondaryCacheTest::kHelperByRole[static_cast<int>(
+        CacheEntryRole::kDataBlock)];
+
 class CompressedSecCacheTestWithCompressAndAllocatorParam
     : public CompressedSecondaryCacheTest,
       public ::testing::WithParamInterface<std::tuple<bool, bool>> {
@@ -906,6 +925,77 @@ TEST_P(CompressedSecondaryCacheTestWithCompressionParam,
   IntegrationFullCapacityTest(sec_cache_is_compressed_);
 }
 
+TEST_P(CompressedSecondaryCacheTestWithCompressionParam, EntryRoles) {
+  CompressedSecondaryCacheOptions opts;
+  opts.capacity = 2048;
+  opts.num_shard_bits = 0;
+
+  if (sec_cache_is_compressed_) {
+    if (!LZ4_Supported()) {
+      ROCKSDB_GTEST_SKIP("This test requires LZ4 support.");
+      return;
+    }
+  } else {
+    opts.compression_type = CompressionType::kNoCompression;
+  }
+
+  // Select a random subset to include, for fast test
+  Random& r = *Random::GetTLSInstance();
+  CacheEntryRoleSet do_not_compress;
+  for (uint32_t i = 0; i < kNumCacheEntryRoles; ++i) {
+    // A few included on average, but decent chance of zero
+    if (r.OneIn(5)) {
+      do_not_compress.Add(static_cast<CacheEntryRole>(i));
+    }
+  }
+  opts.do_not_compress_roles = do_not_compress;
+
+  std::shared_ptr<SecondaryCache> sec_cache = NewCompressedSecondaryCache(opts);
+
+  // Fixed seed to ensure consistent compressibility (doesn't compress)
+  std::string junk(Random(301).RandomString(1000));
+
+  for (uint32_t i = 0; i < kNumCacheEntryRoles; ++i) {
+    CacheEntryRole role = static_cast<CacheEntryRole>(i);
+
+    // Uniquify `junk`
+    junk[0] = static_cast<char>(i);
+    TestItem item{junk.data(), junk.length()};
+    Slice ith_key = Slice(junk.data(), 5);
+
+    get_perf_context()->Reset();
+    ASSERT_OK(sec_cache->Insert(ith_key, &item, &kHelperByRole[i]));
+    ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_dummy_count, 1U);
+
+    ASSERT_OK(sec_cache->Insert(ith_key, &item, &kHelperByRole[i]));
+    ASSERT_EQ(get_perf_context()->compressed_sec_cache_insert_real_count, 1U);
+
+    bool is_in_sec_cache{true};
+    std::unique_ptr<SecondaryCacheResultHandle> handle =
+        sec_cache->Lookup(ith_key, &kHelperByRole[i], this, true,
+                          /*advise_erase=*/true, is_in_sec_cache);
+    ASSERT_NE(handle, nullptr);
+
+    // Lookup returns the right data
+    std::unique_ptr<TestItem> val =
+        std::unique_ptr<TestItem>(static_cast<TestItem*>(handle->Value()));
+    ASSERT_NE(val, nullptr);
+    ASSERT_EQ(memcmp(val->Buf(), item.Buf(), item.Size()), 0);
+
+    bool compressed =
+        sec_cache_is_compressed_ && !do_not_compress.Contains(role);
+    if (compressed) {
+      ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes,
+                1000);
+      ASSERT_EQ(get_perf_context()->compressed_sec_cache_compressed_bytes,
+                1007);
+    } else {
+      ASSERT_EQ(get_perf_context()->compressed_sec_cache_uncompressed_bytes, 0);
+      ASSERT_EQ(get_perf_context()->compressed_sec_cache_compressed_bytes, 0);
+    }
+  }
+}
+
 INSTANTIATE_TEST_CASE_P(CompressedSecCacheTests,
                         CompressedSecondaryCacheTestWithCompressionParam,
                         testing::Bool());

cache/sharded_cache.h

@@ -15,7 +15,7 @@
 
 #include "port/lang.h"
 #include "port/port.h"
-#include "rocksdb/cache.h"
+#include "rocksdb/advanced_cache.h"
 #include "util/hash.h"
 #include "util/mutexlock.h"
 

cache/typed_cache.h

@@ -29,8 +29,8 @@
 #include <type_traits>
 
 #include "cache/cache_helpers.h"
+#include "rocksdb/advanced_cache.h"
 #include "rocksdb/advanced_options.h"
-#include "rocksdb/cache.h"
 
 namespace ROCKSDB_NAMESPACE {
 

db/blob/blob_contents.h

@@ -8,7 +8,7 @@
 #include <memory>
 
 #include "memory/memory_allocator.h"
-#include "rocksdb/cache.h"
+#include "rocksdb/advanced_cache.h"
 #include "rocksdb/rocksdb_namespace.h"
 #include "rocksdb/slice.h"
 #include "rocksdb/status.h"

db/blob/db_blob_basic_test.cc

@@ -1782,16 +1782,49 @@ TEST_F(DBBlobBasicTest, GetEntityBlob) {
   constexpr char key[] = "key";
   constexpr char blob_value[] = "blob_value";
 
+  constexpr char other_key[] = "other_key";
+  constexpr char other_blob_value[] = "other_blob_value";
+
   ASSERT_OK(Put(key, blob_value));
+  ASSERT_OK(Put(other_key, other_blob_value));
 
   ASSERT_OK(Flush());
 
-  PinnableWideColumns result;
-  ASSERT_OK(
-      db_->GetEntity(ReadOptions(), db_->DefaultColumnFamily(), key, &result));
-
   WideColumns expected_columns{{kDefaultWideColumnName, blob_value}};
-  ASSERT_EQ(result.columns(), expected_columns);
+  WideColumns other_expected_columns{
+      {kDefaultWideColumnName, other_blob_value}};
+
+  {
+    PinnableWideColumns result;
+    ASSERT_OK(db_->GetEntity(ReadOptions(), db_->DefaultColumnFamily(), key,
+                             &result));
+    ASSERT_EQ(result.columns(), expected_columns);
+  }
+
+  {
+    PinnableWideColumns result;
+    ASSERT_OK(db_->GetEntity(ReadOptions(), db_->DefaultColumnFamily(),
+                             other_key, &result));
+
+    ASSERT_EQ(result.columns(), other_expected_columns);
+  }
+
+  {
+    constexpr size_t num_keys = 2;
+
+    std::array<Slice, num_keys> keys{{key, other_key}};
+    std::array<PinnableWideColumns, num_keys> results;
+    std::array<Status, num_keys> statuses;
+
+    db_->MultiGetEntity(ReadOptions(), db_->DefaultColumnFamily(), num_keys,
+                        &keys[0], &results[0], &statuses[0]);
+
+    ASSERT_OK(statuses[0]);
+    ASSERT_EQ(results[0].columns(), expected_columns);
+
+    ASSERT_OK(statuses[1]);
+    ASSERT_EQ(results[1].columns(), other_expected_columns);
+  }
 }
 
 class DBBlobWithTimestampTest : public DBBasicTestWithTimestampBase {

db/c.cc

@@ -16,7 +16,7 @@
 #include <vector>
 
 #include "port/port.h"
-#include "rocksdb/cache.h"
+#include "rocksdb/advanced_cache.h"
 #include "rocksdb/compaction_filter.h"
 #include "rocksdb/comparator.h"
 #include "rocksdb/convenience.h"
@@ -6397,4 +6397,3 @@ void rocksdb_enable_manual_compaction(rocksdb_t* db) {
 }
 
 }  // end extern "C"
-

db/compaction/compaction_iterator.cc

@@ -13,6 +13,7 @@
 #include "db/blob/blob_index.h"
 #include "db/blob/prefetch_buffer_collection.h"
 #include "db/snapshot_checker.h"
+#include "db/wide/wide_column_serialization.h"
 #include "logging/logging.h"
 #include "port/likely.h"
 #include "rocksdb/listener.h"
@@ -225,8 +226,8 @@ bool CompactionIterator::InvokeFilterIfNeeded(bool* need_skip,
     return true;
   }
 
-  // TODO: support compaction filter for wide-column entities
-  if (ikey_.type != kTypeValue && ikey_.type != kTypeBlobIndex) {
+  if (ikey_.type != kTypeValue && ikey_.type != kTypeBlobIndex &&
+      ikey_.type != kTypeWideColumnEntity) {
     return true;
   }
 
@@ -234,7 +235,9 @@ bool CompactionIterator::InvokeFilterIfNeeded(bool* need_skip,
       CompactionFilter::Decision::kUndetermined;
   CompactionFilter::ValueType value_type =
       ikey_.type == kTypeValue ? CompactionFilter::ValueType::kValue
-                               : CompactionFilter::ValueType::kBlobIndex;
+      : ikey_.type == kTypeBlobIndex
+          ? CompactionFilter::ValueType::kBlobIndex
+          : CompactionFilter::ValueType::kWideColumnEntity;
 
   // Hack: pass internal key to BlobIndexCompactionFilter since it needs
   // to get sequence number.
@@ -248,6 +251,8 @@ bool CompactionIterator::InvokeFilterIfNeeded(bool* need_skip,
   compaction_filter_value_.clear();
   compaction_filter_skip_until_.Clear();
 
+  std::vector<std::pair<std::string, std::string>> new_columns;
+
   {
     StopWatchNano timer(clock_, report_detailed_time_);
 
@@ -303,10 +308,36 @@ bool CompactionIterator::InvokeFilterIfNeeded(bool* need_skip,
         value_type = CompactionFilter::ValueType::kValue;
       }
     }
+
     if (decision == CompactionFilter::Decision::kUndetermined) {
-      decision = compaction_filter_->FilterV2(
-          level_, filter_key, value_type,
-          blob_value_.empty() ? value_ : blob_value_, &compaction_filter_value_,
+      const Slice* existing_val = nullptr;
+      const WideColumns* existing_col = nullptr;
+
+      WideColumns existing_columns;
+
+      if (ikey_.type != kTypeWideColumnEntity) {
+        if (!blob_value_.empty()) {
+          existing_val = &blob_value_;
+        } else {
+          existing_val = &value_;
+        }
+      } else {
+        Slice value_copy = value_;
+        const Status s =
+            WideColumnSerialization::Deserialize(value_copy, existing_columns);
+
+        if (!s.ok()) {
+          status_ = s;
+          validity_info_.Invalidate();
+          return false;
+        }
+
+        existing_col = &existing_columns;
+      }
+
+      decision = compaction_filter_->FilterV3(
+          level_, filter_key, value_type, existing_val, existing_col,
+          &compaction_filter_value_, &new_columns,
           compaction_filter_skip_until_.rep());
     }
 
@@ -315,9 +346,10 @@ bool CompactionIterator::InvokeFilterIfNeeded(bool* need_skip,
   }
 
   if (decision == CompactionFilter::Decision::kUndetermined) {
-    // Should not reach here, since FilterV2 should never return kUndetermined.
-    status_ =
-        Status::NotSupported("FilterV2() should never return kUndetermined");
+    // Should not reach here, since FilterV2/FilterV3 should never return
+    // kUndetermined.
+    status_ = Status::NotSupported(
+        "FilterV2/FilterV3 should never return kUndetermined");
     validity_info_.Invalidate();
     return false;
   }
@@ -326,7 +358,7 @@ bool CompactionIterator::InvokeFilterIfNeeded(bool* need_skip,
       cmp_->Compare(*compaction_filter_skip_until_.rep(), ikey_.user_key) <=
           0) {
     // Can't skip to a key smaller than the current one.
-    // Keep the key as per FilterV2 documentation.
+    // Keep the key as per FilterV2/FilterV3 documentation.
     decision = CompactionFilter::Decision::kKeep;
   }
 
@@ -388,6 +420,35 @@ bool CompactionIterator::InvokeFilterIfNeeded(bool* need_skip,
     status_ = Status::IOError("Failed to access blob during compaction filter");
     validity_info_.Invalidate();
     return false;
+  } else if (decision == CompactionFilter::Decision::kChangeWideColumnEntity) {
+    WideColumns sorted_columns;
+
+    sorted_columns.reserve(new_columns.size());
+    for (const auto& column : new_columns) {
+      sorted_columns.emplace_back(column.first, column.second);
+    }
+
+    std::sort(sorted_columns.begin(), sorted_columns.end(),
+              [](const WideColumn& lhs, const WideColumn& rhs) {
+                return lhs.name().compare(rhs.name()) < 0;
+              });
+
+    {
+      const Status s = WideColumnSerialization::Serialize(
+          sorted_columns, compaction_filter_value_);
+      if (!s.ok()) {
+        status_ = s;
+        validity_info_.Invalidate();
+        return false;
+      }
+    }
+
+    if (ikey_.type != kTypeWideColumnEntity) {
+      ikey_.type = kTypeWideColumnEntity;
+      current_key_.UpdateInternalKey(ikey_.sequence, kTypeWideColumnEntity);
+    }
+
+    value_ = compaction_filter_value_;
   }
 
   return true;

db/compaction/sst_partitioner.cc

@@ -78,7 +78,6 @@ Status SstPartitionerFactory::CreateFromString(
   std::call_once(once, [&]() {
     RegisterSstPartitionerFactories(*(ObjectLibrary::Default().get()), "");
   });
-  return LoadSharedObject<SstPartitionerFactory>(options, value, nullptr,
-                                                 result);
+  return LoadSharedObject<SstPartitionerFactory>(options, value, result);
 }
 }  // namespace ROCKSDB_NAMESPACE

db/db_basic_test.cc

@@ -30,6 +30,9 @@
 
 namespace ROCKSDB_NAMESPACE {
 
+static bool enable_io_uring = true;
+extern "C" bool RocksDbIOUringEnable() { return enable_io_uring; }
+
 class DBBasicTest : public DBTestBase {
  public:
   DBBasicTest() : DBTestBase("db_basic_test", /*env_do_fsync=*/false) {}
@@ -2173,6 +2176,7 @@ class DBMultiGetAsyncIOTest : public DBBasicTest,
     options_.disable_auto_compactions = true;
     options_.statistics = statistics_;
     options_.table_factory.reset(NewBlockBasedTableFactory(bbto));
+    options_.env = Env::Default();
     Reopen(options_);
     int num_keys = 0;
 
@@ -2239,6 +2243,20 @@ class DBMultiGetAsyncIOTest : public DBBasicTest,
   const std::shared_ptr<Statistics>& statistics() { return statistics_; }
 
  protected:
+  void PrepareDBForTest() {
+#ifdef ROCKSDB_IOURING_PRESENT
+    Reopen(options_);
+#else   // ROCKSDB_IOURING_PRESENT
+    // Warm up the block cache so we don't need to use the IO uring
+    Iterator* iter = dbfull()->NewIterator(ReadOptions());
+    for (iter->SeekToFirst(); iter->Valid() && iter->status().ok();
+         iter->Next())
+      ;
+    EXPECT_OK(iter->status());
+    delete iter;
+#endif  // ROCKSDB_IOURING_PRESENT
+  }
+
   void ReopenDB() { Reopen(options_); }
 
  private:
@@ -2253,6 +2271,8 @@ TEST_P(DBMultiGetAsyncIOTest, GetFromL0) {
   std::vector<PinnableSlice> values(key_strs.size());
   std::vector<Status> statuses(key_strs.size());
 
+  PrepareDBForTest();
+
   ReadOptions ro;
   ro.async_io = true;
   ro.optimize_multiget_for_io = GetParam();
@@ -2271,6 +2291,7 @@ TEST_P(DBMultiGetAsyncIOTest, GetFromL0) {
   statistics()->histogramData(MULTIGET_IO_BATCH_SIZE, &multiget_io_batch_size);
 
   // With async IO, lookups will happen in parallel for each key
+#ifdef ROCKSDB_IOURING_PRESENT
   if (GetParam()) {
     ASSERT_EQ(multiget_io_batch_size.count, 1);
     ASSERT_EQ(multiget_io_batch_size.max, 3);
@@ -2280,6 +2301,11 @@ TEST_P(DBMultiGetAsyncIOTest, GetFromL0) {
     // L0 file
     ASSERT_EQ(multiget_io_batch_size.count, 3);
   }
+#else   // ROCKSDB_IOURING_PRESENT
+  if (GetParam()) {
+    ASSERT_EQ(statistics()->getTickerCount(MULTIGET_COROUTINE_COUNT), 3);
+  }
+#endif  // ROCKSDB_IOURING_PRESENT
 }
 
 TEST_P(DBMultiGetAsyncIOTest, GetFromL1) {
@@ -2297,6 +2323,8 @@ TEST_P(DBMultiGetAsyncIOTest, GetFromL1) {
   values.resize(keys.size());
   statuses.resize(keys.size());
 
+  PrepareDBForTest();
+
   ReadOptions ro;
   ro.async_io = true;
   ro.optimize_multiget_for_io = GetParam();
@@ -2310,6 +2338,7 @@ TEST_P(DBMultiGetAsyncIOTest, GetFromL1) {
   ASSERT_EQ(values[1], "val_l1_" + std::to_string(54));
   ASSERT_EQ(values[2], "val_l1_" + std::to_string(102));
 
+#ifdef ROCKSDB_IOURING_PRESENT
   HistogramData multiget_io_batch_size;
 
   statistics()->histogramData(MULTIGET_IO_BATCH_SIZE, &multiget_io_batch_size);
@@ -2317,9 +2346,11 @@ TEST_P(DBMultiGetAsyncIOTest, GetFromL1) {
   // A batch of 3 async IOs is expected, one for each overlapping file in L1
   ASSERT_EQ(multiget_io_batch_size.count, 1);
   ASSERT_EQ(multiget_io_batch_size.max, 3);
+#endif  // ROCKSDB_IOURING_PRESENT
   ASSERT_EQ(statistics()->getTickerCount(MULTIGET_COROUTINE_COUNT), 3);
 }
 
+#ifdef ROCKSDB_IOURING_PRESENT
 TEST_P(DBMultiGetAsyncIOTest, GetFromL1Error) {
   std::vector<std::string> key_strs;
   std::vector<Slice> keys;
@@ -2335,9 +2366,9 @@ TEST_P(DBMultiGetAsyncIOTest, GetFromL1Error) {
   values.resize(keys.size());
   statuses.resize(keys.size());
 
+  int count = 0;
   SyncPoint::GetInstance()->SetCallBack(
       "TableCache::GetTableReader:BeforeOpenFile", [&](void* status) {
-        static int count = 0;
         count++;
         // Fail the last table reader open, which is the 6th SST file
         // since 3 overlapping L0 files + 3 L1 files containing the keys
@@ -2360,7 +2391,7 @@ TEST_P(DBMultiGetAsyncIOTest, GetFromL1Error) {
       });
   SyncPoint::GetInstance()->EnableProcessing();
 
-  ReopenDB();
+  PrepareDBForTest();
 
   ReadOptions ro;
   ro.async_io = true;
@@ -2382,6 +2413,7 @@ TEST_P(DBMultiGetAsyncIOTest, GetFromL1Error) {
   ASSERT_EQ(multiget_io_batch_size.max, 2);
   ASSERT_EQ(statistics()->getTickerCount(MULTIGET_COROUTINE_COUNT), 2);
 }
+#endif  // ROCKSDB_IOURING_PRESENT
 
 TEST_P(DBMultiGetAsyncIOTest, LastKeyInFile) {
   std::vector<std::string> key_strs;
@@ -2399,6 +2431,8 @@ TEST_P(DBMultiGetAsyncIOTest, LastKeyInFile) {
   values.resize(keys.size());
   statuses.resize(keys.size());
 
+  PrepareDBForTest();
+
   ReadOptions ro;
   ro.async_io = true;
   ro.optimize_multiget_for_io = GetParam();
@@ -2412,6 +2446,7 @@ TEST_P(DBMultiGetAsyncIOTest, LastKeyInFile) {
   ASSERT_EQ(values[1], "val_l1_" + std::to_string(54));
   ASSERT_EQ(values[2], "val_l1_" + std::to_string(102));
 
+#ifdef ROCKSDB_IOURING_PRESENT
   HistogramData multiget_io_batch_size;
 
   statistics()->histogramData(MULTIGET_IO_BATCH_SIZE, &multiget_io_batch_size);
@@ -2422,6 +2457,7 @@ TEST_P(DBMultiGetAsyncIOTest, LastKeyInFile) {
   // will lookup 2 files in parallel and issue 2 async reads
   ASSERT_EQ(multiget_io_batch_size.count, 2);
   ASSERT_EQ(multiget_io_batch_size.max, 2);
+#endif  // ROCKSDB_IOURING_PRESENT
 }
 
 TEST_P(DBMultiGetAsyncIOTest, GetFromL1AndL2) {
@@ -2440,6 +2476,8 @@ TEST_P(DBMultiGetAsyncIOTest, GetFromL1AndL2) {
   values.resize(keys.size());
   statuses.resize(keys.size());
 
+  PrepareDBForTest();
+
   ReadOptions ro;
   ro.async_io = true;
   ro.optimize_multiget_for_io = GetParam();
@@ -2453,6 +2491,7 @@ TEST_P(DBMultiGetAsyncIOTest, GetFromL1AndL2) {
   ASSERT_EQ(values[1], "val_l2_" + std::to_string(56));
   ASSERT_EQ(values[2], "val_l1_" + std::to_string(102));
 
+#ifdef ROCKSDB_IOURING_PRESENT
   HistogramData multiget_io_batch_size;
 
   statistics()->histogramData(MULTIGET_IO_BATCH_SIZE, &multiget_io_batch_size);
@@ -2462,6 +2501,7 @@ TEST_P(DBMultiGetAsyncIOTest, GetFromL1AndL2) {
   // Otherwise, the L2 lookup will happen after L1.
   ASSERT_EQ(multiget_io_batch_size.count, GetParam() ? 1 : 2);
   ASSERT_EQ(multiget_io_batch_size.max, GetParam() ? 3 : 2);
+#endif  // ROCKSDB_IOURING_PRESENT
 }
 
 TEST_P(DBMultiGetAsyncIOTest, GetFromL2WithRangeOverlapL0L1) {
@@ -2478,6 +2518,8 @@ TEST_P(DBMultiGetAsyncIOTest, GetFromL2WithRangeOverlapL0L1) {
   values.resize(keys.size());
   statuses.resize(keys.size());
 
+  PrepareDBForTest();
+
   ReadOptions ro;
   ro.async_io = true;
   ro.optimize_multiget_for_io = GetParam();
@@ -2493,6 +2535,7 @@ TEST_P(DBMultiGetAsyncIOTest, GetFromL2WithRangeOverlapL0L1) {
   ASSERT_EQ(statistics()->getTickerCount(MULTIGET_COROUTINE_COUNT), 2);
 }
 
+#ifdef ROCKSDB_IOURING_PRESENT
 TEST_P(DBMultiGetAsyncIOTest, GetFromL2WithRangeDelInL1) {
   std::vector<std::string> key_strs;
   std::vector<Slice> keys;
@@ -2507,6 +2550,8 @@ TEST_P(DBMultiGetAsyncIOTest, GetFromL2WithRangeDelInL1) {
   values.resize(keys.size());
   statuses.resize(keys.size());
 
+  PrepareDBForTest();
+
   ReadOptions ro;
   ro.async_io = true;
   ro.optimize_multiget_for_io = GetParam();
@@ -2536,6 +2581,8 @@ TEST_P(DBMultiGetAsyncIOTest, GetFromL1AndL2WithRangeDelInL1) {
   values.resize(keys.size());
   statuses.resize(keys.size());
 
+  PrepareDBForTest();
+
   ReadOptions ro;
   ro.async_io = true;
   ro.optimize_multiget_for_io = GetParam();
@@ -2550,6 +2597,45 @@ TEST_P(DBMultiGetAsyncIOTest, GetFromL1AndL2WithRangeDelInL1) {
   // Bloom filters in L0/L1 will avoid the coroutine calls in those levels
   ASSERT_EQ(statistics()->getTickerCount(MULTIGET_COROUTINE_COUNT), 3);
 }
+#endif  // ROCKSDB_IOURING_PRESENT
+
+TEST_P(DBMultiGetAsyncIOTest, GetNoIOUring) {
+  std::vector<std::string> key_strs;
+  std::vector<Slice> keys;
+  std::vector<PinnableSlice> values;
+  std::vector<Status> statuses;
+
+  key_strs.push_back(Key(33));
+  key_strs.push_back(Key(54));
+  key_strs.push_back(Key(102));
+  keys.push_back(key_strs[0]);
+  keys.push_back(key_strs[1]);
+  keys.push_back(key_strs[2]);
+  values.resize(keys.size());
+  statuses.resize(keys.size());
+
+  enable_io_uring = false;
+  ReopenDB();
+
+  ReadOptions ro;
+  ro.async_io = true;
+  ro.optimize_multiget_for_io = GetParam();
+  dbfull()->MultiGet(ro, dbfull()->DefaultColumnFamily(), keys.size(),
+                     keys.data(), values.data(), statuses.data());
+  ASSERT_EQ(values.size(), 3);
+  ASSERT_EQ(statuses[0], Status::NotSupported());
+  ASSERT_EQ(statuses[1], Status::NotSupported());
+  ASSERT_EQ(statuses[2], Status::NotSupported());
+
+  HistogramData multiget_io_batch_size;
+
+  statistics()->histogramData(MULTIGET_IO_BATCH_SIZE, &multiget_io_batch_size);
+
+  // A batch of 3 async IOs is expected, one for each overlapping file in L1
+  ASSERT_EQ(multiget_io_batch_size.count, 1);
+  ASSERT_EQ(multiget_io_batch_size.max, 3);
+  ASSERT_EQ(statistics()->getTickerCount(MULTIGET_COROUTINE_COUNT), 3);
+}
 
 INSTANTIATE_TEST_CASE_P(DBMultiGetAsyncIOTest, DBMultiGetAsyncIOTest,
                         testing::Bool());

db/db_block_cache_test.cc

@@ -287,9 +287,11 @@ class PersistentCacheFromCache : public PersistentCache {
 };
 
 class ReadOnlyCacheWrapper : public CacheWrapper {
+ public:
   using CacheWrapper::CacheWrapper;
 
-  using Cache::Insert;
+  const char* Name() const override { return "ReadOnlyCacheWrapper"; }
+
   Status Insert(const Slice& /*key*/, Cache::ObjectPtr /*value*/,
                 const CacheItemHelper* /*helper*/, size_t /*charge*/,
                 Handle** /*handle*/, Priority /*priority*/) override {
@@ -711,7 +713,8 @@ class LookupLiarCache : public CacheWrapper {
   explicit LookupLiarCache(std::shared_ptr<Cache> target)
       : CacheWrapper(std::move(target)) {}
 
-  using Cache::Lookup;
+  const char* Name() const override { return "LookupLiarCache"; }
+
   Handle* Lookup(const Slice& key, const CacheItemHelper* helper = nullptr,
                  CreateContext* create_context = nullptr,
                  Priority priority = Priority::LOW, bool wait = true,

db/db_impl/db_impl.cc

@@ -1539,6 +1539,14 @@ Status DBImpl::LockWAL() {
       assert(lock_wal_write_token_);
       ++lock_wal_count_;
     } else {
+      // NOTE: this will "unnecessarily" wait for other non-LockWAL() write
+      // stalls to clear before LockWAL returns, however fixing that would
+      // not be simple because if we notice the primary queue is already
+      // stalled, that stall might clear while we release DB mutex in
+      // EnterUnbatched() for the nonmem queue. And if we work around that in
+      // the naive way, we could deadlock by locking the two queues in different
+      // orders.
+
       WriteThread::Writer w;
       write_thread_.EnterUnbatched(&w, &mutex_);
       WriteThread::Writer nonmem_w;
@@ -1571,6 +1579,8 @@ Status DBImpl::LockWAL() {
 
 Status DBImpl::UnlockWAL() {
   bool signal = false;
+  uint64_t maybe_stall_begun_count = 0;
+  uint64_t nonmem_maybe_stall_begun_count = 0;
   {
     InstrumentedMutexLock lock(&mutex_);
     if (lock_wal_count_ == 0) {
@@ -1580,12 +1590,28 @@ Status DBImpl::UnlockWAL() {
     if (lock_wal_count_ == 0) {
       lock_wal_write_token_.reset();
       signal = true;
+      // For the last UnlockWAL, we don't want to return from UnlockWAL()
+      // until the thread(s) that called BeginWriteStall() have had a chance to
+      // call EndWriteStall(), so that no_slowdown writes after UnlockWAL() are
+      // guaranteed to succeed if there's no other source of stall.
+      maybe_stall_begun_count = write_thread_.GetBegunCountOfOutstandingStall();
+      if (two_write_queues_) {
+        nonmem_maybe_stall_begun_count =
+            nonmem_write_thread_.GetBegunCountOfOutstandingStall();
+      }
     }
   }
   if (signal) {
     // SignalAll outside of mutex for efficiency
     bg_cv_.SignalAll();
   }
+  // Ensure stalls have cleared
+  if (maybe_stall_begun_count) {
+    write_thread_.WaitForStallEndedCount(maybe_stall_begun_count);
+  }
+  if (nonmem_maybe_stall_begun_count) {
+    nonmem_write_thread_.WaitForStallEndedCount(nonmem_maybe_stall_begun_count);
+  }
   return Status::OK();
 }
 
@@ -2579,14 +2605,25 @@ void DBImpl::MultiGet(const ReadOptions& read_options, const size_t num_keys,
                       ColumnFamilyHandle** column_families, const Slice* keys,
                       PinnableSlice* values, Status* statuses,
                       const bool sorted_input) {
-  return MultiGet(read_options, num_keys, column_families, keys, values,
-                  /*timestamps=*/nullptr, statuses, sorted_input);
+  MultiGet(read_options, num_keys, column_families, keys, values,
+           /* timestamps */ nullptr, statuses, sorted_input);
 }
 
 void DBImpl::MultiGet(const ReadOptions& read_options, const size_t num_keys,
                       ColumnFamilyHandle** column_families, const Slice* keys,
                       PinnableSlice* values, std::string* timestamps,
                       Status* statuses, const bool sorted_input) {
+  MultiGetCommon(read_options, num_keys, column_families, keys, values,
+                 /* columns */ nullptr, timestamps, statuses, sorted_input);
+}
+
+void DBImpl::MultiGetCommon(const ReadOptions& read_options,
+                            const size_t num_keys,
+                            ColumnFamilyHandle** column_families,
+                            const Slice* keys, PinnableSlice* values,
+                            PinnableWideColumns* columns,
+                            std::string* timestamps, Status* statuses,
+                            const bool sorted_input) {
   if (num_keys == 0) {
     return;
   }
@@ -2632,8 +2669,20 @@ void DBImpl::MultiGet(const ReadOptions& read_options, const size_t num_keys,
   autovector<KeyContext*, MultiGetContext::MAX_BATCH_SIZE> sorted_keys;
   sorted_keys.resize(num_keys);
   for (size_t i = 0; i < num_keys; ++i) {
-    values[i].Reset();
-    key_context.emplace_back(column_families[i], keys[i], &values[i],
+    PinnableSlice* val = nullptr;
+    PinnableWideColumns* col = nullptr;
+
+    if (values) {
+      val = &values[i];
+      val->Reset();
+    } else {
+      assert(columns);
+
+      col = &columns[i];
+      col->Reset();
+    }
+
+    key_context.emplace_back(column_families[i], keys[i], val, col,
                              timestamps ? &timestamps[i] : nullptr,
                              &statuses[i]);
   }
@@ -2757,8 +2806,8 @@ void DBImpl::MultiGet(const ReadOptions& read_options,
                       ColumnFamilyHandle* column_family, const size_t num_keys,
                       const Slice* keys, PinnableSlice* values,
                       Status* statuses, const bool sorted_input) {
-  return MultiGet(read_options, column_family, num_keys, keys, values,
-                  /*timestamp=*/nullptr, statuses, sorted_input);
+  MultiGet(read_options, column_family, num_keys, keys, values,
+           /* timestamps */ nullptr, statuses, sorted_input);
 }
 
 void DBImpl::MultiGet(const ReadOptions& read_options,
@@ -2766,6 +2815,16 @@ void DBImpl::MultiGet(const ReadOptions& read_options,
                       const Slice* keys, PinnableSlice* values,
                       std::string* timestamps, Status* statuses,
                       const bool sorted_input) {
+  MultiGetCommon(read_options, column_family, num_keys, keys, values,
+                 /* columns */ nullptr, timestamps, statuses, sorted_input);
+}
+
+void DBImpl::MultiGetCommon(const ReadOptions& read_options,
+                            ColumnFamilyHandle* column_family,
+                            const size_t num_keys, const Slice* keys,
+                            PinnableSlice* values, PinnableWideColumns* columns,
+                            std::string* timestamps, Status* statuses,
+                            bool sorted_input) {
   if (tracer_) {
     // TODO: This mutex should be removed later, to improve performance when
     // tracing is enabled.
@@ -2779,8 +2838,20 @@ void DBImpl::MultiGet(const ReadOptions& read_options,
   autovector<KeyContext*, MultiGetContext::MAX_BATCH_SIZE> sorted_keys;
   sorted_keys.resize(num_keys);
   for (size_t i = 0; i < num_keys; ++i) {
-    values[i].Reset();
-    key_context.emplace_back(column_family, keys[i], &values[i],
+    PinnableSlice* val = nullptr;
+    PinnableWideColumns* col = nullptr;
+
+    if (values) {
+      val = &values[i];
+      val->Reset();
+    } else {
+      assert(columns);
+
+      col = &columns[i];
+      col->Reset();
+    }
+
+    key_context.emplace_back(column_family, keys[i], val, col,
                              timestamps ? &timestamps[i] : nullptr,
                              &statuses[i]);
   }
@@ -2942,8 +3013,17 @@ Status DBImpl::MultiGetImpl(
   uint64_t bytes_read = 0;
   for (size_t i = start_key; i < start_key + num_keys - keys_left; ++i) {
     KeyContext* key = (*sorted_keys)[i];
+    assert(key);
+    assert(key->s);
+
     if (key->s->ok()) {
-      bytes_read += key->value->size();
+      if (key->value) {
+        bytes_read += key->value->size();
+      } else {
+        assert(key->columns);
+        bytes_read += key->columns->serialized_size();
+      }
+
       num_found++;
     }
   }
@@ -2967,6 +3047,22 @@ Status DBImpl::MultiGetImpl(
   return s;
 }
 
+void DBImpl::MultiGetEntity(const ReadOptions& options, size_t num_keys,
+                            ColumnFamilyHandle** column_families,
+                            const Slice* keys, PinnableWideColumns* results,
+                            Status* statuses, bool sorted_input) {
+  MultiGetCommon(options, num_keys, column_families, keys, /* values */ nullptr,
+                 results, /* timestamps */ nullptr, statuses, sorted_input);
+}
+
+void DBImpl::MultiGetEntity(const ReadOptions& options,
+                            ColumnFamilyHandle* column_family, size_t num_keys,
+                            const Slice* keys, PinnableWideColumns* results,
+                            Status* statuses, bool sorted_input) {
+  MultiGetCommon(options, column_family, num_keys, keys, /* values */ nullptr,
+                 results, /* timestamps */ nullptr, statuses, sorted_input);
+}
+
 Status DBImpl::CreateColumnFamily(const ColumnFamilyOptions& cf_options,
                                   const std::string& column_family,
                                   ColumnFamilyHandle** handle) {

db/db_impl/db_impl.h

@@ -277,33 +277,40 @@ class DBImpl : public DB {
   // The values and statuses parameters are arrays with number of elements
   // equal to keys.size(). This allows the storage for those to be alloacted
   // by the caller on the stack for small batches
-  virtual void MultiGet(const ReadOptions& options,
-                        ColumnFamilyHandle* column_family,
-                        const size_t num_keys, const Slice* keys,
-                        PinnableSlice* values, Status* statuses,
-                        const bool sorted_input = false) override;
-  virtual void MultiGet(const ReadOptions& options,
-                        ColumnFamilyHandle* column_family,
-                        const size_t num_keys, const Slice* keys,
-                        PinnableSlice* values, std::string* timestamps,
-                        Status* statuses,
-                        const bool sorted_input = false) override;
+  void MultiGet(const ReadOptions& options, ColumnFamilyHandle* column_family,
+                const size_t num_keys, const Slice* keys, PinnableSlice* values,
+                Status* statuses, const bool sorted_input = false) override;
+  void MultiGet(const ReadOptions& options, ColumnFamilyHandle* column_family,
+                const size_t num_keys, const Slice* keys, PinnableSlice* values,
+                std::string* timestamps, Status* statuses,
+                const bool sorted_input = false) override;
 
-  virtual void MultiGet(const ReadOptions& options, const size_t num_keys,
-                        ColumnFamilyHandle** column_families, const Slice* keys,
-                        PinnableSlice* values, Status* statuses,
-                        const bool sorted_input = false) override;
-  virtual void MultiGet(const ReadOptions& options, const size_t num_keys,
-                        ColumnFamilyHandle** column_families, const Slice* keys,
-                        PinnableSlice* values, std::string* timestamps,
-                        Status* statuses,
-                        const bool sorted_input = false) override;
+  void MultiGet(const ReadOptions& options, const size_t num_keys,
+                ColumnFamilyHandle** column_families, const Slice* keys,
+                PinnableSlice* values, Status* statuses,
+                const bool sorted_input = false) override;
+  void MultiGet(const ReadOptions& options, const size_t num_keys,
+                ColumnFamilyHandle** column_families, const Slice* keys,
+                PinnableSlice* values, std::string* timestamps,
+                Status* statuses, const bool sorted_input = false) override;
 
-  virtual void MultiGetWithCallback(
+  void MultiGetWithCallback(
       const ReadOptions& options, ColumnFamilyHandle* column_family,
       ReadCallback* callback,
       autovector<KeyContext*, MultiGetContext::MAX_BATCH_SIZE>* sorted_keys);
 
+  using DB::MultiGetEntity;
+
+  void MultiGetEntity(const ReadOptions& options,
+                      ColumnFamilyHandle* column_family, size_t num_keys,
+                      const Slice* keys, PinnableWideColumns* results,
+                      Status* statuses, bool sorted_input) override;
+
+  void MultiGetEntity(const ReadOptions& options, size_t num_keys,
+                      ColumnFamilyHandle** column_families, const Slice* keys,
+                      PinnableWideColumns* results, Status* statuses,
+                      bool sorted_input) override;
+
   virtual Status CreateColumnFamily(const ColumnFamilyOptions& cf_options,
                                     const std::string& column_family,
                                     ColumnFamilyHandle** handle) override;
@@ -1117,6 +1124,8 @@ class DBImpl : public DB {
 
   void TEST_UnlockMutex();
 
+  void TEST_SignalAllBgCv();
+
   // REQUIRES: mutex locked
   void* TEST_BeginWrite();
 
@@ -2189,6 +2198,18 @@ class DBImpl : public DB {
       const size_t num_keys, bool sorted,
       autovector<KeyContext*, MultiGetContext::MAX_BATCH_SIZE>* key_ptrs);
 
+  void MultiGetCommon(const ReadOptions& options,
+                      ColumnFamilyHandle* column_family, const size_t num_keys,
+                      const Slice* keys, PinnableSlice* values,
+                      PinnableWideColumns* columns, std::string* timestamps,
+                      Status* statuses, bool sorted_input);
+
+  void MultiGetCommon(const ReadOptions& options, const size_t num_keys,
+                      ColumnFamilyHandle** column_families, const Slice* keys,
+                      PinnableSlice* values, PinnableWideColumns* columns,
+                      std::string* timestamps, Status* statuses,
+                      bool sorted_input);
+
   // A structure to hold the information required to process MultiGet of keys
   // belonging to one column family. For a multi column family MultiGet, there
   // will be a container of these objects.

db/db_impl/db_impl_debug.cc

@@ -198,6 +198,8 @@ void DBImpl::TEST_LockMutex() { mutex_.Lock(); }
 
 void DBImpl::TEST_UnlockMutex() { mutex_.Unlock(); }
 
+void DBImpl::TEST_SignalAllBgCv() { bg_cv_.SignalAll(); }
+
 void* DBImpl::TEST_BeginWrite() {
   auto w = new WriteThread::Writer();
   write_thread_.EnterUnbatched(w, &mutex_);

db/db_impl/db_impl_files.cc

@@ -361,8 +361,6 @@ void DBImpl::DeleteObsoleteFileImpl(int job_id, const std::string& fname,
   }
   TEST_SYNC_POINT_CALLBACK("DBImpl::DeleteObsoleteFileImpl:AfterDeletion",
                            &file_deletion_status);
-  TEST_SYNC_POINT_CALLBACK("DBImpl::DeleteObsoleteFileImpl:AfterDeletion2",
-                           const_cast<std::string*>(&fname));
   if (file_deletion_status.ok()) {
     ROCKS_LOG_DEBUG(immutable_db_options_.info_log,
                     "[JOB %d] Delete %s type=%d #%" PRIu64 " -- %s\n", job_id,

db/db_impl/db_impl_write.cc

@@ -1833,8 +1833,13 @@ Status DBImpl::DelayWrite(uint64_t num_bytes, WriteThread& write_thread,
       // Notify write_thread about the stall so it can setup a barrier and
       // fail any pending writers with no_slowdown
       write_thread.BeginWriteStall();
-      TEST_SYNC_POINT("DBImpl::DelayWrite:Wait");
+      if (&write_thread == &write_thread_) {
+        TEST_SYNC_POINT("DBImpl::DelayWrite:Wait");
+      } else {
+        TEST_SYNC_POINT("DBImpl::DelayWrite:NonmemWait");
+      }
       bg_cv_.Wait();
+      TEST_SYNC_POINT_CALLBACK("DBImpl::DelayWrite:AfterWait", &mutex_);
       write_thread.EndWriteStall();
     }
   }

db/db_merge_operator_test.cc

@@ -231,7 +231,9 @@ TEST_F(DBMergeOperatorTest, MergeOperatorFailsWithMustMerge) {
       {
         std::string value;
         ASSERT_OK(db_->Get(ReadOptions(), "k0", &value));
-        ASSERT_TRUE(db_->Get(ReadOptions(), "k1", &value).IsCorruption());
+        Status s = db_->Get(ReadOptions(), "k1", &value);
+        ASSERT_TRUE(s.IsCorruption());
+        ASSERT_EQ(Status::SubCode::kMergeOperatorFailed, s.subcode());
         ASSERT_OK(db_->Get(ReadOptions(), "k2", &value));
       }
 
@@ -243,6 +245,8 @@ TEST_F(DBMergeOperatorTest, MergeOperatorFailsWithMustMerge) {
         ASSERT_EQ("k0", iter->key());
         iter->Next();
         ASSERT_TRUE(iter->status().IsCorruption());
+        ASSERT_EQ(Status::SubCode::kMergeOperatorFailed,
+                  iter->status().subcode());
 
         iter->SeekToLast();
         ASSERT_TRUE(iter->Valid());

db/db_test_util.h

@@ -909,82 +909,6 @@ class TestPutOperator : public MergeOperator {
   virtual const char* Name() const override { return "TestPutOperator"; }
 };
 
-// A wrapper around Cache that can easily be extended with instrumentation,
-// etc.
-class CacheWrapper : public Cache {
- public:
-  explicit CacheWrapper(std::shared_ptr<Cache> target)
-      : target_(std::move(target)) {}
-
-  const char* Name() const override { return target_->Name(); }
-
-  Status Insert(const Slice& key, ObjectPtr value,
-                const CacheItemHelper* helper, size_t charge,
-                Handle** handle = nullptr,
-                Priority priority = Priority::LOW) override {
-    return target_->Insert(key, value, helper, charge, handle, priority);
-  }
-
-  Handle* Lookup(const Slice& key, const CacheItemHelper* helper,
-                 CreateContext* create_context,
-                 Priority priority = Priority::LOW, bool wait = true,
-                 Statistics* stats = nullptr) override {
-    return target_->Lookup(key, helper, create_context, priority, wait, stats);
-  }
-
-  bool Ref(Handle* handle) override { return target_->Ref(handle); }
-
-  using Cache::Release;
-  bool Release(Handle* handle, bool erase_if_last_ref = false) override {
-    return target_->Release(handle, erase_if_last_ref);
-  }
-
-  ObjectPtr Value(Handle* handle) override { return target_->Value(handle); }
-
-  void Erase(const Slice& key) override { target_->Erase(key); }
-  uint64_t NewId() override { return target_->NewId(); }
-
-  void SetCapacity(size_t capacity) override { target_->SetCapacity(capacity); }
-
-  void SetStrictCapacityLimit(bool strict_capacity_limit) override {
-    target_->SetStrictCapacityLimit(strict_capacity_limit);
-  }
-
-  bool HasStrictCapacityLimit() const override {
-    return target_->HasStrictCapacityLimit();
-  }
-
-  size_t GetCapacity() const override { return target_->GetCapacity(); }
-
-  size_t GetUsage() const override { return target_->GetUsage(); }
-
-  size_t GetUsage(Handle* handle) const override {
-    return target_->GetUsage(handle);
-  }
-
-  size_t GetPinnedUsage() const override { return target_->GetPinnedUsage(); }
-
-  size_t GetCharge(Handle* handle) const override {
-    return target_->GetCharge(handle);
-  }
-
-  const CacheItemHelper* GetCacheItemHelper(Handle* handle) const override {
-    return target_->GetCacheItemHelper(handle);
-  }
-
-  void ApplyToAllEntries(
-      const std::function<void(const Slice& key, ObjectPtr value, size_t charge,
-                               const CacheItemHelper* helper)>& callback,
-      const ApplyToAllEntriesOptions& opts) override {
-    target_->ApplyToAllEntries(callback, opts);
-  }
-
-  void EraseUnRefEntries() override { target_->EraseUnRefEntries(); }
-
- protected:
-  std::shared_ptr<Cache> target_;
-};
-
 /*
  * A cache wrapper that tracks certain CacheEntryRole's cache charge, its
  * peaks and increments
@@ -1002,6 +926,8 @@ class TargetCacheChargeTrackingCache : public CacheWrapper {
  public:
   explicit TargetCacheChargeTrackingCache(std::shared_ptr<Cache> target);
 
+  const char* Name() const override { return "TargetCacheChargeTrackingCache"; }
+
   Status Insert(const Slice& key, ObjectPtr value,
                 const CacheItemHelper* helper, size_t charge,
                 Handle** handle = nullptr,

db/db_wal_test.cc

@@ -1633,6 +1633,8 @@ TEST_F(DBWALTest, RaceInstallFlushResultsWithWalObsoletion) {
 
 TEST_F(DBWALTest, FixSyncWalOnObseletedWalWithNewManifestCausingMissingWAL) {
   Options options = CurrentOptions();
+  // Small size to force manifest creation
+  options.max_manifest_file_size = 1;
   options.track_and_verify_wals_in_manifest = true;
   DestroyAndReopen(options);
 
@@ -1649,53 +1651,33 @@ TEST_F(DBWALTest, FixSyncWalOnObseletedWalWithNewManifestCausingMissingWAL) {
   // (2) SyncWAL() proceeds with the lock. It
   // creates a new manifest and syncs all the inactive wals before the latest
   // (i.e, active log), which is 4.log. Note that SyncWAL() is not aware of the
-  // fact that 4.log has marked as to be obseleted. Prior to the fix, such wal
+  // fact that 4.log has marked as to be obseleted. Such wal
   // sync will then add a WAL addition record of 4.log to the new manifest
-  // without any special treatment.
-  // (3) BackgroundFlush() will eventually purge 4.log.
+  // without any special treatment. Prior to the fix, there is no WAL deletion
+  // record to offset it. (3) BackgroundFlush() will eventually purge 4.log.
+
   bool wal_synced = false;
   SyncPoint::GetInstance()->SetCallBack(
       "FindObsoleteFiles::PostMutexUnlock", [&](void*) {
         ASSERT_OK(env_->FileExists(wal_file_path));
-
-        SyncPoint::GetInstance()->SetCallBack(
-            "VersionSet::ProcessManifestWrites:"
-            "PostDecidingCreateNewManifestOrNot",
-            [&](void* arg) {
-              bool* new_descriptor_log = (bool*)arg;
-              *new_descriptor_log = true;
-            });
-
+        uint64_t pre_sync_wal_manifest_no =
+            dbfull()->TEST_Current_Manifest_FileNo();
         ASSERT_OK(db_->SyncWAL());
+        uint64_t post_sync_wal_manifest_no =
+            dbfull()->TEST_Current_Manifest_FileNo();
+        bool new_manifest_created =
+            post_sync_wal_manifest_no == pre_sync_wal_manifest_no + 1;
+        ASSERT_TRUE(new_manifest_created);
         wal_synced = true;
       });
 
-  SyncPoint::GetInstance()->SetCallBack(
-      "DBImpl::DeleteObsoleteFileImpl:AfterDeletion2", [&](void* arg) {
-        std::string* file_name = (std::string*)arg;
-        if (*file_name == wal_file_path) {
-          TEST_SYNC_POINT(
-              "DBWALTest::"
-              "FixSyncWalOnObseletedWalWithNewManifestCausingMissingWAL::"
-              "PostDeleteWAL");
-        }
-      });
-
-  SyncPoint::GetInstance()->LoadDependency(
-      {{"DBImpl::BackgroundCallFlush:FilesFound",
-        "PreConfrimObsoletedWALSynced"},
-       {"DBWALTest::FixSyncWalOnObseletedWalWithNewManifestCausingMissingWAL::"
-        "PostDeleteWAL",
-        "PreConfrimWALDeleted"}});
 
   SyncPoint::GetInstance()->EnableProcessing();
 
   ASSERT_OK(Flush());
+  ASSERT_OK(dbfull()->TEST_WaitForBackgroundWork());
 
-  TEST_SYNC_POINT("PreConfrimObsoletedWALSynced");
   ASSERT_TRUE(wal_synced);
-
-  TEST_SYNC_POINT("PreConfrimWALDeleted");
   // BackgroundFlush() purged 4.log
   // because the memtable associated with the WAL was flushed and new WAL was
   // created (i.e, 8.log)

db/db_write_test.cc

@@ -612,6 +612,7 @@ TEST_P(DBWriteTest, LockWALInEffect) {
   std::unique_ptr<FaultInjectionTestEnv> mock_env(
       new FaultInjectionTestEnv(env_));
   options.env = mock_env.get();
+  options.disable_auto_compactions = true;
   options.paranoid_checks = false;
   Reopen(options);
   // try the 1st WAL created during open

db/event_helpers.cc

@@ -13,7 +13,7 @@ namespace ROCKSDB_NAMESPACE {
 Status EventListener::CreateFromString(const ConfigOptions& config_options,
                                        const std::string& id,
                                        std::shared_ptr<EventListener>* result) {
-  return LoadSharedObject<EventListener>(config_options, id, nullptr, result);
+  return LoadSharedObject<EventListener>(config_options, id, result);
 }
 
 namespace {

db/external_sst_file_basic_test.cc

@@ -1182,6 +1182,7 @@ TEST_F(ExternalSSTFileBasicTest, SyncFailure) {
     ASSERT_OK(sst_file_writer->Finish());
 
     IngestExternalFileOptions ingest_opt;
+    ASSERT_FALSE(ingest_opt.write_global_seqno);  // new default
     if (i == 0) {
       ingest_opt.move_files = true;
     }

db/log_reader.cc

@@ -515,10 +515,11 @@ unsigned int Reader::ReadPhysicalRecord(Slice* result, size_t* drop_size,
 
       size_t uncompressed_size = 0;
       int remaining = 0;
+      const char* input = header + header_size;
       do {
-        remaining = uncompress_->Uncompress(header + header_size, length,
-                                            uncompressed_buffer_.get(),
-                                            &uncompressed_size);
+        remaining = uncompress_->Uncompress(
+            input, length, uncompressed_buffer_.get(), &uncompressed_size);
+        input = nullptr;
         if (remaining < 0) {
           buffer_.clear();
           return kBadRecord;
@@ -830,10 +831,11 @@ bool FragmentBufferedReader::TryReadFragment(
     uncompressed_record_.clear();
     size_t uncompressed_size = 0;
     int remaining = 0;
+    const char* input = header + header_size;
     do {
-      remaining = uncompress_->Uncompress(header + header_size, length,
-                                          uncompressed_buffer_.get(),
-                                          &uncompressed_size);
+      remaining = uncompress_->Uncompress(
+          input, length, uncompressed_buffer_.get(), &uncompressed_size);
+      input = nullptr;
       if (remaining < 0) {
         buffer_.clear();
         *fragment_type_or_err = kBadRecord;

db/log_test.cc

@@ -979,6 +979,38 @@ TEST_P(CompressionLogTest, Fragmentation) {
   ASSERT_EQ("EOF", Read());
 }
 
+TEST_P(CompressionLogTest, AlignedFragmentation) {
+  CompressionType compression_type = std::get<2>(GetParam());
+  if (!StreamingCompressionTypeSupported(compression_type)) {
+    ROCKSDB_GTEST_SKIP("Test requires support for compression type");
+    return;
+  }
+  ASSERT_OK(SetupTestEnv());
+  Random rnd(301);
+  int num_filler_records = 0;
+  // Keep writing small records until the next record will be aligned at the
+  // beginning of the block.
+  while ((WrittenBytes() & (kBlockSize - 1)) >= kHeaderSize) {
+    char entry = 'a';
+    ASSERT_OK(writer_->AddRecord(Slice(&entry, 1)));
+    num_filler_records++;
+  }
+  const std::vector<std::string> wal_entries = {
+      rnd.RandomBinaryString(3 * kBlockSize),
+  };
+  for (const std::string& wal_entry : wal_entries) {
+    Write(wal_entry);
+  }
+
+  for (int i = 0; i < num_filler_records; ++i) {
+    ASSERT_EQ("a", Read());
+  }
+  for (const std::string& wal_entry : wal_entries) {
+    ASSERT_EQ(wal_entry, Read());
+  }
+  ASSERT_EQ("EOF", Read());
+}
+
 INSTANTIATE_TEST_CASE_P(
     Compression, CompressionLogTest,
     ::testing::Combine(::testing::Values(0, 1), ::testing::Bool(),
@@ -1026,10 +1058,11 @@ TEST_P(StreamingCompressionTest, Basic) {
   for (int i = 0; i < (int)compressed_buffers.size(); i++) {
     // Call uncompress till either the entire input is consumed or the output
     // buffer size is equal to the allocated output buffer size.
+    const char* input = compressed_buffers[i].c_str();
     do {
-      ret_val = uncompress->Uncompress(compressed_buffers[i].c_str(),
-                                       compressed_buffers[i].size(),
+      ret_val = uncompress->Uncompress(input, compressed_buffers[i].size(),
                                        uncompressed_output_buffer, &output_pos);
+      input = nullptr;
       if (output_pos > 0) {
         std::string uncompressed_fragment;
         uncompressed_fragment.assign(uncompressed_output_buffer, output_pos);

db/memtable.cc

@@ -1452,18 +1452,24 @@ void MemTable::MultiGet(const ReadOptions& read_options, MultiGetRange* range,
     }
     SequenceNumber dummy_seq;
     GetFromTable(*(iter->lkey), iter->max_covering_tombstone_seq, true,
-                 callback, &iter->is_blob_index, iter->value->GetSelf(),
-                 /*columns=*/nullptr, iter->timestamp, iter->s,
-                 &(iter->merge_context), &dummy_seq, &found_final_value,
-                 &merge_in_progress);
+                 callback, &iter->is_blob_index,
+                 iter->value ? iter->value->GetSelf() : nullptr, iter->columns,
+                 iter->timestamp, iter->s, &(iter->merge_context), &dummy_seq,
+                 &found_final_value, &merge_in_progress);
 
     if (!found_final_value && merge_in_progress) {
       *(iter->s) = Status::MergeInProgress();
     }
 
     if (found_final_value) {
-      iter->value->PinSelf();
-      range->AddValueSize(iter->value->size());
+      if (iter->value) {
+        iter->value->PinSelf();
+        range->AddValueSize(iter->value->size());
+      } else {
+        assert(iter->columns);
+        range->AddValueSize(iter->columns->serialized_size());
+      }
+
       range->MarkKeyDone(iter);
       RecordTick(moptions_.statistics, MEMTABLE_HIT);
       if (range->GetValueSize() > read_options.value_size_soft_limit) {

db/merge_helper.cc

@@ -113,7 +113,7 @@ Status MergeHelper::TimedFullMerge(
 
   if (!success) {
     RecordTick(statistics, NUMBER_MERGE_FAILURES);
-    return Status::Corruption("Error: Could not perform merge.");
+    return Status::Corruption(Status::SubCode::kMergeOperatorFailed);
   }
 
   return Status::OK();
@@ -578,14 +578,15 @@ CompactionFilter::Decision MergeHelper::FilterMerge(const Slice& user_key,
   }
   compaction_filter_value_.clear();
   compaction_filter_skip_until_.Clear();
-  auto ret = compaction_filter_->FilterV2(
-      level_, user_key, CompactionFilter::ValueType::kMergeOperand, value_slice,
-      &compaction_filter_value_, compaction_filter_skip_until_.rep());
+  auto ret = compaction_filter_->FilterV3(
+      level_, user_key, CompactionFilter::ValueType::kMergeOperand,
+      &value_slice, /* existing_columns */ nullptr, &compaction_filter_value_,
+      /* new_columns */ nullptr, compaction_filter_skip_until_.rep());
   if (ret == CompactionFilter::Decision::kRemoveAndSkipUntil) {
     if (user_comparator_->Compare(*compaction_filter_skip_until_.rep(),
                                   user_key) <= 0) {
       // Invalid skip_until returned from compaction filter.
-      // Keep the key as per FilterV2 documentation.
+      // Keep the key as per FilterV2/FilterV3 documentation.
       ret = CompactionFilter::Decision::kKeep;
     } else {
       compaction_filter_skip_until_.ConvertFromUserKey(kMaxSequenceNumber,

db/version_edit.h

@@ -20,8 +20,8 @@
 #include "db/wal_edit.h"
 #include "memory/arena.h"
 #include "port/malloc.h"
+#include "rocksdb/advanced_cache.h"
 #include "rocksdb/advanced_options.h"
-#include "rocksdb/cache.h"
 #include "table/table_reader.h"
 #include "table/unique_id_impl.h"
 #include "util/autovector.h"

db/version_set.cc

@@ -23,7 +23,6 @@
 #include "db/blob/blob_fetcher.h"
 #include "db/blob/blob_file_cache.h"
 #include "db/blob/blob_file_reader.h"
-#include "db/blob/blob_index.h"
 #include "db/blob/blob_log_format.h"
 #include "db/blob/blob_source.h"
 #include "db/compaction/compaction.h"
@@ -2177,26 +2176,35 @@ void Version::MultiGetBlob(
 
     autovector<BlobReadRequest> blob_reqs_in_file;
     BlobReadContexts& blobs_in_file = ctx.second;
-    for (const auto& blob : blobs_in_file) {
-      const BlobIndex& blob_index = blob.first;
-      const KeyContext& key_context = blob.second;
+    for (auto& blob : blobs_in_file) {
+      const BlobIndex& blob_index = blob.blob_index;
+      const KeyContext* const key_context = blob.key_context;
+      assert(key_context);
+      assert(key_context->get_context);
+      assert(key_context->s);
+
+      if (key_context->value) {
+        key_context->value->Reset();
+      } else {
+        assert(key_context->columns);
+        key_context->columns->Reset();
+      }
 
       if (!blob_file_meta) {
-        *key_context.s = Status::Corruption("Invalid blob file number");
+        *key_context->s = Status::Corruption("Invalid blob file number");
         continue;
       }
 
       if (blob_index.HasTTL() || blob_index.IsInlined()) {
-        *key_context.s =
+        *key_context->s =
             Status::Corruption("Unexpected TTL/inlined blob index");
         continue;
       }
 
-      key_context.value->Reset();
       blob_reqs_in_file.emplace_back(
-          key_context.get_context->ukey_to_get_blob_value(),
+          key_context->get_context->ukey_to_get_blob_value(),
           blob_index.offset(), blob_index.size(), blob_index.compression(),
-          key_context.value, key_context.s);
+          &blob.result, key_context->s);
     }
     if (blob_reqs_in_file.size() > 0) {
       const auto file_size = blob_file_meta->GetBlobFileSize();
@@ -2211,18 +2219,29 @@ void Version::MultiGetBlob(
 
   for (auto& ctx : blob_ctxs) {
     BlobReadContexts& blobs_in_file = ctx.second;
-    for (const auto& blob : blobs_in_file) {
-      const KeyContext& key_context = blob.second;
-      if (key_context.s->ok()) {
-        range.AddValueSize(key_context.value->size());
-        if (range.GetValueSize() > read_options.value_size_soft_limit) {
-          *key_context.s = Status::Aborted();
+    for (auto& blob : blobs_in_file) {
+      const KeyContext* const key_context = blob.key_context;
+      assert(key_context);
+      assert(key_context->get_context);
+      assert(key_context->s);
+
+      if (key_context->s->ok()) {
+        if (key_context->value) {
+          *key_context->value = std::move(blob.result);
+          range.AddValueSize(key_context->value->size());
+        } else {
+          assert(key_context->columns);
+          key_context->columns->SetPlainValue(blob.result);
+          range.AddValueSize(key_context->columns->serialized_size());
         }
-      } else if (key_context.s->IsIncomplete()) {
+
+        if (range.GetValueSize() > read_options.value_size_soft_limit) {
+          *key_context->s = Status::Aborted();
+        }
+      } else if (key_context->s->IsIncomplete()) {
         // read_options.read_tier == kBlockCacheTier
         // Cannot read blob(s): no disk I/O allowed
-        assert(key_context.get_context);
-        auto& get_context = *(key_context.get_context);
+        auto& get_context = *(key_context->get_context);
         get_context.MarkKeyMayExist();
       }
     }
@@ -2388,6 +2407,9 @@ void Version::Get(const ReadOptions& read_options, const LookupKey& k,
             "Encounter unexpected blob index. Please open DB with "
             "ROCKSDB_NAMESPACE::blob_db::BlobDB instead.");
         return;
+      case GetContext::kMergeOperatorFailed:
+        *status = Status::Corruption(Status::SubCode::kMergeOperatorFailed);
+        return;
     }
     f = fp.GetNextFile();
   }
@@ -2457,7 +2479,7 @@ void Version::MultiGet(const ReadOptions& read_options, MultiGetRange* range,
     get_ctx.emplace_back(
         user_comparator(), merge_operator_, info_log_, db_statistics_,
         iter->s->ok() ? GetContext::kNotFound : GetContext::kMerge,
-        iter->ukey_with_ts, iter->value, /*columns=*/nullptr, iter->timestamp,
+        iter->ukey_with_ts, iter->value, iter->columns, iter->timestamp,
         nullptr, &(iter->merge_context), true,
         &iter->max_covering_tombstone_seq, clock_, nullptr,
         merge_operator_ ? &pinned_iters_mgr : nullptr, callback,
@@ -2657,23 +2679,29 @@ void Version::MultiGet(const ReadOptions& read_options, MultiGetRange* range,
       }
       // merge_operands are in saver and we hit the beginning of the key history
       // do a final merge of nullptr and operands;
-      std::string* str_value =
-          iter->value != nullptr ? iter->value->GetSelf() : nullptr;
+      std::string result;
+
       // `op_failure_scope` (an output parameter) is not provided (set to
       // nullptr) since a failure must be propagated regardless of its value.
       *status = MergeHelper::TimedFullMerge(
           merge_operator_, user_key, nullptr, iter->merge_context.GetOperands(),
-          str_value, info_log_, db_statistics_, clock_,
+          &result, info_log_, db_statistics_, clock_,
           /* result_operand */ nullptr, /* update_num_ops_stats */ true,
           /* op_failure_scope */ nullptr);
       if (LIKELY(iter->value != nullptr)) {
+        *iter->value->GetSelf() = std::move(result);
         iter->value->PinSelf();
         range->AddValueSize(iter->value->size());
-        range->MarkKeyDone(iter);
-        if (range->GetValueSize() > read_options.value_size_soft_limit) {
-          s = Status::Aborted();
-          break;
-        }
+      } else {
+        assert(iter->columns);
+        iter->columns->SetPlainValue(result);
+        range->AddValueSize(iter->columns->serialized_size());
+      }
+
+      range->MarkKeyDone(iter);
+      if (range->GetValueSize() > read_options.value_size_soft_limit) {
+        s = Status::Aborted();
+        break;
       }
     } else {
       range->MarkKeyDone(iter);
@@ -5077,15 +5105,10 @@ Status VersionSet::ProcessManifestWrites(
   if (!descriptor_log_ ||
       manifest_file_size_ > db_options_->max_manifest_file_size) {
     TEST_SYNC_POINT("VersionSet::ProcessManifestWrites:BeforeNewManifest");
-    TEST_SYNC_POINT_CALLBACK(
-        "VersionSet::ProcessManifestWrites:BeforeNewManifest", nullptr);
     new_descriptor_log = true;
   } else {
     pending_manifest_file_number_ = manifest_file_number_;
   }
-  TEST_SYNC_POINT_CALLBACK(
-      "VersionSet::ProcessManifestWrites:PostDecidingCreateNewManifestOrNot",
-      &new_descriptor_log);
 
   // Local cached copy of state variable(s). WriteCurrentStateToManifest()
   // reads its content after releasing db mutex to avoid race with

db/version_set.h

@@ -33,6 +33,7 @@
 
 #include "cache/cache_helpers.h"
 #include "db/blob/blob_file_meta.h"
+#include "db/blob/blob_index.h"
 #include "db/column_family.h"
 #include "db/compaction/compaction.h"
 #include "db/compaction/compaction_picker.h"
@@ -889,8 +890,15 @@ class Version {
                  FilePrefetchBuffer* prefetch_buffer, PinnableSlice* value,
                  uint64_t* bytes_read) const;
 
-  using BlobReadContext =
-      std::pair<BlobIndex, std::reference_wrapper<const KeyContext>>;
+  struct BlobReadContext {
+    BlobReadContext(const BlobIndex& blob_idx, const KeyContext* key_ctx)
+        : blob_index(blob_idx), key_context(key_ctx) {}
+
+    BlobIndex blob_index;
+    const KeyContext* key_context;
+    PinnableSlice result;
+  };
+
   using BlobReadContexts = std::vector<BlobReadContext>;
   void MultiGetBlob(const ReadOptions& read_options, MultiGetRange& range,
                     std::unordered_map<uint64_t, BlobReadContexts>& blob_ctxs);

db/version_set_sync_and_async.h

@@ -101,23 +101,39 @@ DEFINE_SYNC_AND_ASYNC(Status, Version::MultiGetFromSST)
         file_range.MarkKeyDone(iter);
 
         if (iter->is_blob_index) {
+          BlobIndex blob_index;
+          Status tmp_s;
+
           if (iter->value) {
             TEST_SYNC_POINT_CALLBACK("Version::MultiGet::TamperWithBlobIndex",
                                      &(*iter));
 
-            const Slice& blob_index_slice = *(iter->value);
-            BlobIndex blob_index;
-            Status tmp_s = blob_index.DecodeFrom(blob_index_slice);
-            if (tmp_s.ok()) {
-              const uint64_t blob_file_num = blob_index.file_number();
-              blob_ctxs[blob_file_num].emplace_back(
-                  std::make_pair(blob_index, std::cref(*iter)));
-            } else {
-              *(iter->s) = tmp_s;
-            }
+            tmp_s = blob_index.DecodeFrom(*(iter->value));
+
+          } else {
+            assert(iter->columns);
+            assert(!iter->columns->columns().empty());
+            assert(iter->columns->columns().front().name() ==
+                   kDefaultWideColumnName);
+
+            tmp_s =
+                blob_index.DecodeFrom(iter->columns->columns().front().value());
+          }
+
+          if (tmp_s.ok()) {
+            const uint64_t blob_file_num = blob_index.file_number();
+            blob_ctxs[blob_file_num].emplace_back(blob_index, &*iter);
+          } else {
+            *(iter->s) = tmp_s;
           }
         } else {
-          file_range.AddValueSize(iter->value->size());
+          if (iter->value) {
+            file_range.AddValueSize(iter->value->size());
+          } else {
+            assert(iter->columns);
+            file_range.AddValueSize(iter->columns->serialized_size());
+          }
+
           if (file_range.GetValueSize() > read_options.value_size_soft_limit) {
             s = Status::Aborted();
             break;
@@ -141,6 +157,10 @@ DEFINE_SYNC_AND_ASYNC(Status, Version::MultiGetFromSST)
             "ROCKSDB_NAMESPACE::blob_db::BlobDB instead.");
         file_range.MarkKeyDone(iter);
         continue;
+      case GetContext::kMergeOperatorFailed:
+        *status = Status::Corruption(Status::SubCode::kMergeOperatorFailed);
+        file_range.MarkKeyDone(iter);
+        continue;
     }
   }
 

db/wide/db_wide_basic_test.cc

@@ -4,6 +4,7 @@
 //  (found in the LICENSE.Apache file in the root directory).
 
 #include <array>
+#include <cctype>
 #include <memory>
 
 #include "db/db_test_util.h"
@@ -104,6 +105,26 @@ TEST_F(DBWideBasicTest, PutEntity) {
       ASSERT_EQ(values[2], third_value);
     }
 
+    {
+      constexpr size_t num_keys = 3;
+
+      std::array<Slice, num_keys> keys{{first_key, second_key, third_key}};
+      std::array<PinnableWideColumns, num_keys> results;
+      std::array<Status, num_keys> statuses;
+
+      db_->MultiGetEntity(ReadOptions(), db_->DefaultColumnFamily(), num_keys,
+                          &keys[0], &results[0], &statuses[0]);
+
+      ASSERT_OK(statuses[0]);
+      ASSERT_EQ(results[0].columns(), first_columns);
+
+      ASSERT_OK(statuses[1]);
+      ASSERT_EQ(results[1].columns(), second_columns);
+
+      ASSERT_OK(statuses[2]);
+      ASSERT_EQ(results[2].columns(), expected_third_columns);
+    }
+
     {
       std::unique_ptr<Iterator> iter(db_->NewIterator(ReadOptions()));
 
@@ -209,6 +230,40 @@ TEST_F(DBWideBasicTest, PutEntityColumnFamily) {
   ASSERT_OK(db_->Write(WriteOptions(), &batch));
 }
 
+TEST_F(DBWideBasicTest, MultiCFMultiGetEntity) {
+  Options options = GetDefaultOptions();
+  CreateAndReopenWithCF({"corinthian"}, options);
+
+  constexpr char first_key[] = "first";
+  WideColumns first_columns{{"attr_name1", "foo"}, {"attr_name2", "bar"}};
+
+  ASSERT_OK(db_->PutEntity(WriteOptions(), db_->DefaultColumnFamily(),
+                           first_key, first_columns));
+
+  constexpr char second_key[] = "second";
+  WideColumns second_columns{{"attr_one", "two"}, {"attr_three", "four"}};
+
+  ASSERT_OK(
+      db_->PutEntity(WriteOptions(), handles_[1], second_key, second_columns));
+
+  constexpr size_t num_keys = 2;
+
+  std::array<ColumnFamilyHandle*, num_keys> column_families{
+      {db_->DefaultColumnFamily(), handles_[1]}};
+  std::array<Slice, num_keys> keys{{first_key, second_key}};
+  std::array<PinnableWideColumns, num_keys> results;
+  std::array<Status, num_keys> statuses;
+
+  db_->MultiGetEntity(ReadOptions(), num_keys, &column_families[0], &keys[0],
+                      &results[0], &statuses[0]);
+
+  ASSERT_OK(statuses[0]);
+  ASSERT_EQ(results[0].columns(), first_columns);
+
+  ASSERT_OK(statuses[1]);
+  ASSERT_EQ(results[1].columns(), second_columns);
+}
+
 TEST_F(DBWideBasicTest, MergePlainKeyValue) {
   Options options = GetDefaultOptions();
   options.create_if_missing = true;
@@ -279,6 +334,26 @@ TEST_F(DBWideBasicTest, MergePlainKeyValue) {
       ASSERT_EQ(result.columns(), expected_third_columns);
     }
 
+    {
+      constexpr size_t num_keys = 3;
+
+      std::array<Slice, num_keys> keys{{first_key, second_key, third_key}};
+      std::array<PinnableWideColumns, num_keys> results;
+      std::array<Status, num_keys> statuses;
+
+      db_->MultiGetEntity(ReadOptions(), db_->DefaultColumnFamily(), num_keys,
+                          &keys[0], &results[0], &statuses[0]);
+
+      ASSERT_OK(statuses[0]);
+      ASSERT_EQ(results[0].columns(), expected_first_columns);
+
+      ASSERT_OK(statuses[1]);
+      ASSERT_EQ(results[1].columns(), expected_second_columns);
+
+      ASSERT_OK(statuses[2]);
+      ASSERT_EQ(results[2].columns(), expected_third_columns);
+    }
+
     {
       std::unique_ptr<Iterator> iter(db_->NewIterator(ReadOptions()));
 
@@ -456,6 +531,23 @@ TEST_F(DBWideBasicTest, MergeEntity) {
       ASSERT_OK(statuses[1]);
     }
 
+    {
+      constexpr size_t num_keys = 2;
+
+      std::array<Slice, num_keys> keys{{first_key, second_key}};
+      std::array<PinnableWideColumns, num_keys> results;
+      std::array<Status, num_keys> statuses;
+
+      db_->MultiGetEntity(ReadOptions(), db_->DefaultColumnFamily(), num_keys,
+                          &keys[0], &results[0], &statuses[0]);
+
+      ASSERT_OK(statuses[0]);
+      ASSERT_EQ(results[0].columns(), first_expected_columns);
+
+      ASSERT_OK(statuses[1]);
+      ASSERT_EQ(results[1].columns(), second_expected_columns);
+    }
+
     {
       std::unique_ptr<Iterator> iter(db_->NewIterator(ReadOptions()));
 
@@ -593,6 +685,302 @@ TEST_F(DBWideBasicTest, MergeEntity) {
   verify_merge_ops_post_compaction();
 }
 
+TEST_F(DBWideBasicTest, CompactionFilter) {
+  Options options = GetDefaultOptions();
+  options.create_if_missing = true;
+
+  // Wide-column entity with default column
+  constexpr char first_key[] = "first";
+  WideColumns first_columns{{kDefaultWideColumnName, "a"},
+                            {"attr_name1", "foo"},
+                            {"attr_name2", "bar"}};
+  WideColumns first_columns_uppercase{{kDefaultWideColumnName, "A"},
+                                      {"attr_name1", "FOO"},
+                                      {"attr_name2", "BAR"}};
+
+  // Wide-column entity without default column
+  constexpr char second_key[] = "second";
+  WideColumns second_columns{{"attr_one", "two"}, {"attr_three", "four"}};
+  WideColumns second_columns_uppercase{{"attr_one", "TWO"},
+                                       {"attr_three", "FOUR"}};
+
+  // Plain old key-value
+  constexpr char last_key[] = "last";
+  constexpr char last_value[] = "baz";
+  constexpr char last_value_uppercase[] = "BAZ";
+
+  auto write = [&] {
+    ASSERT_OK(db_->PutEntity(WriteOptions(), db_->DefaultColumnFamily(),
+                             first_key, first_columns));
+    ASSERT_OK(db_->PutEntity(WriteOptions(), db_->DefaultColumnFamily(),
+                             second_key, second_columns));
+
+    ASSERT_OK(Flush());
+
+    ASSERT_OK(db_->Put(WriteOptions(), db_->DefaultColumnFamily(), last_key,
+                       last_value));
+
+    ASSERT_OK(Flush());
+    ASSERT_OK(db_->CompactRange(CompactRangeOptions(), /* begin */ nullptr,
+                                /* end */ nullptr));
+  };
+
+  // Test a compaction filter that keeps all entries
+  {
+    class KeepFilter : public CompactionFilter {
+     public:
+      Decision FilterV3(
+          int /* level */, const Slice& /* key */, ValueType /* value_type */,
+          const Slice* /* existing_value */,
+          const WideColumns* /* existing_columns */,
+          std::string* /* new_value */,
+          std::vector<std::pair<std::string, std::string>>* /* new_columns */,
+          std::string* /* skip_until */) const override {
+        return Decision::kKeep;
+      }
+
+      const char* Name() const override { return "KeepFilter"; }
+    };
+
+    KeepFilter filter;
+    options.compaction_filter = &filter;
+
+    DestroyAndReopen(options);
+
+    write();
+
+    {
+      PinnableWideColumns result;
+      ASSERT_OK(db_->GetEntity(ReadOptions(), db_->DefaultColumnFamily(),
+                               first_key, &result));
+      ASSERT_EQ(result.columns(), first_columns);
+    }
+
+    {
+      PinnableWideColumns result;
+      ASSERT_OK(db_->GetEntity(ReadOptions(), db_->DefaultColumnFamily(),
+                               second_key, &result));
+      ASSERT_EQ(result.columns(), second_columns);
+    }
+
+    // Note: GetEntity should return an entity with a single default column,
+    // since last_key is a plain key-value
+    {
+      PinnableWideColumns result;
+      ASSERT_OK(db_->GetEntity(ReadOptions(), db_->DefaultColumnFamily(),
+                               last_key, &result));
+
+      WideColumns expected_columns{{kDefaultWideColumnName, last_value}};
+      ASSERT_EQ(result.columns(), expected_columns);
+    }
+  }
+
+  // Test a compaction filter that removes all entries
+  {
+    class RemoveFilter : public CompactionFilter {
+     public:
+      Decision FilterV3(
+          int /* level */, const Slice& /* key */, ValueType /* value_type */,
+          const Slice* /* existing_value */,
+          const WideColumns* /* existing_columns */,
+          std::string* /* new_value */,
+          std::vector<std::pair<std::string, std::string>>* /* new_columns */,
+          std::string* /* skip_until */) const override {
+        return Decision::kRemove;
+      }
+
+      const char* Name() const override { return "RemoveFilter"; }
+    };
+
+    RemoveFilter filter;
+    options.compaction_filter = &filter;
+
+    DestroyAndReopen(options);
+
+    write();
+
+    {
+      PinnableWideColumns result;
+      ASSERT_TRUE(db_->GetEntity(ReadOptions(), db_->DefaultColumnFamily(),
+                                 first_key, &result)
+                      .IsNotFound());
+    }
+
+    {
+      PinnableWideColumns result;
+      ASSERT_TRUE(db_->GetEntity(ReadOptions(), db_->DefaultColumnFamily(),
+                                 second_key, &result)
+                      .IsNotFound());
+    }
+
+    {
+      PinnableWideColumns result;
+      ASSERT_TRUE(db_->GetEntity(ReadOptions(), db_->DefaultColumnFamily(),
+                                 last_key, &result)
+                      .IsNotFound());
+    }
+  }
+
+  // Test a compaction filter that changes the values of entries to uppercase.
+  // The new entry is always a plain key-value; if the existing entry is a
+  // wide-column entity, only the value of its first column is kept.
+  {
+    class ChangeValueFilter : public CompactionFilter {
+     public:
+      Decision FilterV3(
+          int /* level */, const Slice& /* key */, ValueType value_type,
+          const Slice* existing_value, const WideColumns* existing_columns,
+          std::string* new_value,
+          std::vector<std::pair<std::string, std::string>>* /* new_columns */,
+          std::string* /* skip_until */) const override {
+        assert(new_value);
+
+        auto upper = [](const std::string& str) {
+          std::string result(str);
+
+          for (char& c : result) {
+            c = static_cast<char>(std::toupper(static_cast<unsigned char>(c)));
+          }
+
+          return result;
+        };
+
+        if (value_type == ValueType::kWideColumnEntity) {
+          assert(existing_columns);
+
+          if (!existing_columns->empty()) {
+            *new_value = upper(existing_columns->front().value().ToString());
+          }
+        } else {
+          assert(existing_value);
+
+          *new_value = upper(existing_value->ToString());
+        }
+
+        return Decision::kChangeValue;
+      }
+
+      const char* Name() const override { return "ChangeValueFilter"; }
+    };
+
+    ChangeValueFilter filter;
+    options.compaction_filter = &filter;
+
+    DestroyAndReopen(options);
+
+    write();
+
+    // Note: GetEntity should return entities with a single default column,
+    // since all entries are now plain key-values
+    {
+      PinnableWideColumns result;
+      ASSERT_OK(db_->GetEntity(ReadOptions(), db_->DefaultColumnFamily(),
+                               first_key, &result));
+
+      WideColumns expected_columns{
+          {kDefaultWideColumnName, first_columns_uppercase[0].value()}};
+      ASSERT_EQ(result.columns(), expected_columns);
+    }
+
+    {
+      PinnableWideColumns result;
+      ASSERT_OK(db_->GetEntity(ReadOptions(), db_->DefaultColumnFamily(),
+                               second_key, &result));
+
+      WideColumns expected_columns{
+          {kDefaultWideColumnName, second_columns_uppercase[0].value()}};
+      ASSERT_EQ(result.columns(), expected_columns);
+    }
+
+    {
+      PinnableWideColumns result;
+      ASSERT_OK(db_->GetEntity(ReadOptions(), db_->DefaultColumnFamily(),
+                               last_key, &result));
+
+      WideColumns expected_columns{
+          {kDefaultWideColumnName, last_value_uppercase}};
+      ASSERT_EQ(result.columns(), expected_columns);
+    }
+  }
+
+  // Test a compaction filter that changes the column values of entries to
+  // uppercase. The new entry is always a wide-column entity; if the existing
+  // entry is a plain key-value, it is converted to a wide-column entity with a
+  // single default column.
+  {
+    class ChangeEntityFilter : public CompactionFilter {
+     public:
+      Decision FilterV3(
+          int /* level */, const Slice& /* key */, ValueType value_type,
+          const Slice* existing_value, const WideColumns* existing_columns,
+          std::string* /* new_value */,
+          std::vector<std::pair<std::string, std::string>>* new_columns,
+          std::string* /* skip_until */) const override {
+        assert(new_columns);
+
+        auto upper = [](const std::string& str) {
+          std::string result(str);
+
+          for (char& c : result) {
+            c = static_cast<char>(std::toupper(static_cast<unsigned char>(c)));
+          }
+
+          return result;
+        };
+
+        if (value_type == ValueType::kWideColumnEntity) {
+          assert(existing_columns);
+
+          for (const auto& column : *existing_columns) {
+            new_columns->emplace_back(column.name().ToString(),
+                                      upper(column.value().ToString()));
+          }
+        } else {
+          assert(existing_value);
+
+          new_columns->emplace_back(kDefaultWideColumnName.ToString(),
+                                    upper(existing_value->ToString()));
+        }
+
+        return Decision::kChangeWideColumnEntity;
+      }
+
+      const char* Name() const override { return "ChangeEntityFilter"; }
+    };
+
+    ChangeEntityFilter filter;
+    options.compaction_filter = &filter;
+
+    DestroyAndReopen(options);
+
+    write();
+
+    {
+      PinnableWideColumns result;
+      ASSERT_OK(db_->GetEntity(ReadOptions(), db_->DefaultColumnFamily(),
+                               first_key, &result));
+      ASSERT_EQ(result.columns(), first_columns_uppercase);
+    }
+
+    {
+      PinnableWideColumns result;
+      ASSERT_OK(db_->GetEntity(ReadOptions(), db_->DefaultColumnFamily(),
+                               second_key, &result));
+      ASSERT_EQ(result.columns(), second_columns_uppercase);
+    }
+
+    {
+      PinnableWideColumns result;
+      ASSERT_OK(db_->GetEntity(ReadOptions(), db_->DefaultColumnFamily(),
+                               last_key, &result));
+
+      WideColumns expected_columns{
+          {kDefaultWideColumnName, last_value_uppercase}};
+      ASSERT_EQ(result.columns(), expected_columns);
+    }
+  }
+}
+
 TEST_F(DBWideBasicTest, PutEntityTimestampError) {
   // Note: timestamps are currently not supported
 

db/write_thread.cc

@@ -228,6 +228,7 @@ bool WriteThread::LinkOne(Writer* w, std::atomic<Writer*>* newest_writer) {
   assert(w->state == STATE_INIT);
   Writer* writers = newest_writer->load(std::memory_order_relaxed);
   while (true) {
+    assert(writers != w);
     // If write stall in effect, and w->no_slowdown is not true,
     // block here until stall is cleared. If its true, then return
     // immediately
@@ -325,6 +326,7 @@ void WriteThread::CompleteFollower(Writer* w, WriteGroup& write_group) {
 }
 
 void WriteThread::BeginWriteStall() {
+  ++stall_begun_count_;
   LinkOne(&write_stall_dummy_, &newest_writer_);
 
   // Walk writer list until w->write_group != nullptr. The current write group
@@ -367,10 +369,34 @@ void WriteThread::EndWriteStall() {
   }
   newest_writer_.exchange(write_stall_dummy_.link_older);
 
+  ++stall_ended_count_;
+
   // Wake up writers
   stall_cv_.SignalAll();
 }
 
+uint64_t WriteThread::GetBegunCountOfOutstandingStall() {
+  if (stall_begun_count_ > stall_ended_count_) {
+    // Oustanding stall in queue
+    assert(newest_writer_.load(std::memory_order_relaxed) ==
+           &write_stall_dummy_);
+    return stall_begun_count_;
+  } else {
+    // No stall in queue
+    assert(newest_writer_.load(std::memory_order_relaxed) !=
+           &write_stall_dummy_);
+    return 0;
+  }
+}
+
+void WriteThread::WaitForStallEndedCount(uint64_t stall_count) {
+  MutexLock lock(&stall_mu_);
+
+  while (stall_ended_count_ < stall_count) {
+    stall_cv_.Wait();
+  }
+}
+
 static WriteThread::AdaptationContext jbg_ctx("JoinBatchGroup");
 void WriteThread::JoinBatchGroup(Writer* w) {
   TEST_SYNC_POINT_CALLBACK("WriteThread::JoinBatchGroup:Start", w);

db/write_thread.h

@@ -357,11 +357,23 @@ class WriteThread {
 
   // Insert a dummy writer at the tail of the write queue to indicate a write
   // stall, and fail any writers in the queue with no_slowdown set to true
+  // REQUIRES: db mutex held, no other stall on this queue outstanding
   void BeginWriteStall();
 
   // Remove the dummy writer and wake up waiting writers
+  // REQUIRES: db mutex held
   void EndWriteStall();
 
+  // Number of BeginWriteStall(), or 0 if there is no active stall in the
+  // write queue.
+  // REQUIRES: db mutex held
+  uint64_t GetBegunCountOfOutstandingStall();
+
+  // Wait for number of completed EndWriteStall() to reach >= `stall_count`,
+  // which will generally have come from GetBegunCountOfOutstandingStall().
+  // (Does not require db mutex held)
+  void WaitForStallEndedCount(uint64_t stall_count);
+
  private:
   // See AwaitState.
   const uint64_t max_yield_usec_;
@@ -401,6 +413,18 @@ class WriteThread {
   port::Mutex stall_mu_;
   port::CondVar stall_cv_;
 
+  // Count the number of stalls begun, so that we can check whether
+  // a particular stall has cleared (even if caught in another stall).
+  // Controlled by DB mutex.
+  // Because of the contract on BeginWriteStall() / EndWriteStall(),
+  // stall_ended_count_ <= stall_begun_count_ <= stall_ended_count_ + 1.
+  uint64_t stall_begun_count_ = 0;
+  // Count the number of stalls ended, so that we can check whether
+  // a particular stall has cleared (even if caught in another stall).
+  // Writes controlled by DB mutex + stall_mu_, signalled by stall_cv_.
+  // Read with stall_mu or DB mutex.
+  uint64_t stall_ended_count_ = 0;
+
   // Waits for w->state & goal_mask using w->StateMutex().  Returns
   // the state that satisfies goal_mask.
   uint8_t BlockingAwaitState(Writer* w, uint8_t goal_mask);

db_stress_tool/db_stress_gflags.cc

@@ -1073,4 +1073,6 @@ DEFINE_uint64(stats_dump_period_sec,
               ROCKSDB_NAMESPACE::Options().stats_dump_period_sec,
               "Gap between printing stats to log in seconds");
 
+extern "C" bool RocksDbIOUringEnable() { return true; }
+
 #endif  // GFLAGS

db_stress_tool/db_stress_test_base.cc

@@ -2901,7 +2901,9 @@ void StressTest::MaybeUseOlderTimestampForRangeScan(ThreadState* thread,
   read_opts.timestamp = &ts_slice;
 
   // TODO (yanqin): support Merge with iter_start_ts
-  if (!thread->rand.OneInOpt(3) || FLAGS_use_merge || FLAGS_use_full_merge_v1) {
+  // TODO (yuzhangyu): support BlobDB with iter_start_ts
+  if (!thread->rand.OneInOpt(3) || FLAGS_use_merge || FLAGS_use_full_merge_v1 ||
+      FLAGS_enable_blob_files) {
     return;
   }
 
@@ -2926,10 +2928,6 @@ void CheckAndSetOptionsForUserTimestamp(Options& options) {
     fprintf(stderr, "TransactionDB does not support timestamp yet.\n");
     exit(1);
   }
-  if (FLAGS_enable_blob_files || FLAGS_use_blob_db) {
-    fprintf(stderr, "BlobDB not supported with timestamp.\n");
-    exit(1);
-  }
   if (FLAGS_test_cf_consistency || FLAGS_test_batches_snapshots) {
     fprintf(stderr,
             "Due to per-key ts-seq ordering constraint, only the (default) "

db_stress_tool/db_stress_tool.cc

@@ -99,13 +99,6 @@ int db_stress_tool(int argc, char** argv) {
 
   env_wrapper_guard = std::make_shared<CompositeEnvWrapper>(
       raw_env, std::make_shared<DbStressFSWrapper>(raw_env->GetFileSystem()));
-  if (!env_opts) {
-    // If using the default Env (Posix), wrap DbStressEnvWrapper with the
-    // legacy EnvWrapper. This is a temporary fix for the ReadAsync interface
-    // not being properly supported with Posix and db_stress. The EnvWrapper
-    // has a default implementation of ReadAsync that redirects to Read.
-    env_wrapper_guard = std::make_shared<EnvWrapper>(env_wrapper_guard);
-  }
   db_stress_env = env_wrapper_guard.get();
 
   FLAGS_rep_factory = StringToRepFactory(FLAGS_memtablerep.c_str());

db_stress_tool/no_batched_ops_stress.cc

@@ -53,6 +53,7 @@ class NonBatchedOpsStressTest : public StressTest {
         kGet,
         kGetEntity,
         kMultiGet,
+        kMultiGetEntity,
         kGetMergeOperands,
         // Add any new items above kNumberOfMethods
         kNumberOfMethods
@@ -163,11 +164,11 @@ class NonBatchedOpsStressTest : public StressTest {
           Status s =
               db_->GetEntity(options, column_families_[cf], key, &columns);
 
-          const WideColumns& columns_from_db = columns.columns();
-
           std::string from_db;
 
           if (s.ok()) {
+            const WideColumns& columns_from_db = columns.columns();
+
             if (!columns_from_db.empty() &&
                 columns_from_db[0].name() == kDefaultWideColumnName) {
               from_db = columns_from_db[0].value().ToString();
@@ -200,14 +201,14 @@ class NonBatchedOpsStressTest : public StressTest {
           size_t batch_size = thread->rand.Uniform(128) + 1;
           batch_size = std::min<size_t>(batch_size, end - i);
 
-          std::vector<std::string> keystrs(batch_size);
+          std::vector<std::string> key_strs(batch_size);
           std::vector<Slice> keys(batch_size);
           std::vector<PinnableSlice> values(batch_size);
           std::vector<Status> statuses(batch_size);
 
           for (size_t j = 0; j < batch_size; ++j) {
-            keystrs[j] = Key(i + j);
-            keys[j] = Slice(keystrs[j].data(), keystrs[j].size());
+            key_strs[j] = Key(i + j);
+            keys[j] = Slice(key_strs[j]);
           }
 
           db_->MultiGet(options, column_families_[cf], batch_size, keys.data(),
@@ -226,6 +227,61 @@ class NonBatchedOpsStressTest : public StressTest {
             }
           }
 
+          i += batch_size;
+        }
+      } else if (method == VerificationMethod::kMultiGetEntity) {
+        for (int64_t i = start; i < end;) {
+          if (thread->shared->HasVerificationFailedYet()) {
+            break;
+          }
+
+          // Keep the batch size to some reasonable value
+          size_t batch_size = thread->rand.Uniform(128) + 1;
+          batch_size = std::min<size_t>(batch_size, end - i);
+
+          std::vector<std::string> key_strs(batch_size);
+          std::vector<Slice> keys(batch_size);
+          std::vector<PinnableWideColumns> results(batch_size);
+          std::vector<Status> statuses(batch_size);
+
+          for (size_t j = 0; j < batch_size; ++j) {
+            key_strs[j] = Key(i + j);
+            keys[j] = Slice(key_strs[j]);
+          }
+
+          db_->MultiGetEntity(options, column_families_[cf], batch_size,
+                              keys.data(), results.data(), statuses.data());
+
+          for (size_t j = 0; j < batch_size; ++j) {
+            std::string from_db;
+
+            if (statuses[j].ok()) {
+              const WideColumns& columns_from_db = results[j].columns();
+
+              if (!columns_from_db.empty() &&
+                  columns_from_db[0].name() == kDefaultWideColumnName) {
+                from_db = columns_from_db[0].value().ToString();
+              }
+
+              const WideColumns expected_columns =
+                  GenerateExpectedWideColumns(GetValueBase(from_db), from_db);
+              if (columns_from_db != expected_columns) {
+                VerificationAbort(shared, static_cast<int>(cf), i, from_db,
+                                  columns_from_db, expected_columns);
+              }
+            }
+
+            VerifyOrSyncValue(static_cast<int>(cf), i + j, options, shared,
+                              from_db,
+                              /* msg_prefix */ "MultiGetEntity verification",
+                              statuses[j], /* strict */ true);
+
+            if (!from_db.empty()) {
+              PrintKeyValue(static_cast<int>(cf), static_cast<uint32_t>(i + j),
+                            from_db.data(), from_db.size());
+            }
+          }
+
           i += batch_size;
         }
       } else {

docs/Gemfile

@@ -1,4 +1,4 @@
 source 'https://rubygems.org'
-gem 'github-pages', '~> 225'
+gem 'github-pages', '~> 227'
 
 gem "webrick", "~> 1.7"

docs/Gemfile.lock

@@ -1,285 +0,0 @@
-GEM
-  remote: https://rubygems.org/
-  specs:
-    activesupport (6.0.4.6)
-      concurrent-ruby (~> 1.0, >= 1.0.2)
-      i18n (>= 0.7, < 2)
-      minitest (~> 5.1)
-      tzinfo (~> 1.1)
-      zeitwerk (~> 2.2, >= 2.2.2)
-    addressable (2.8.0)
-      public_suffix (>= 2.0.2, < 5.0)
-    coffee-script (2.4.1)
-      coffee-script-source
-      execjs
-    coffee-script-source (1.11.1)
-    colorator (1.1.0)
-    commonmarker (0.23.7)
-    concurrent-ruby (1.1.9)
-    dnsruby (1.61.9)
-      simpleidn (~> 0.1)
-    em-websocket (0.5.3)
-      eventmachine (>= 0.12.9)
-      http_parser.rb (~> 0)
-    ethon (0.15.0)
-      ffi (>= 1.15.0)
-    eventmachine (1.2.7)
-    execjs (2.8.1)
-    faraday (1.10.0)
-      faraday-em_http (~> 1.0)
-      faraday-em_synchrony (~> 1.0)
-      faraday-excon (~> 1.1)
-      faraday-httpclient (~> 1.0)
-      faraday-multipart (~> 1.0)
-      faraday-net_http (~> 1.0)
-      faraday-net_http_persistent (~> 1.0)
-      faraday-patron (~> 1.0)
-      faraday-rack (~> 1.0)
-      faraday-retry (~> 1.0)
-      ruby2_keywords (>= 0.0.4)
-    faraday-em_http (1.0.0)
-    faraday-em_synchrony (1.0.0)
-    faraday-excon (1.1.0)
-    faraday-httpclient (1.0.1)
-    faraday-multipart (1.0.3)
-      multipart-post (>= 1.2, < 3)
-    faraday-net_http (1.0.1)
-    faraday-net_http_persistent (1.2.0)
-    faraday-patron (1.0.0)
-    faraday-rack (1.0.0)
-    faraday-retry (1.0.3)
-    ffi (1.15.5)
-    forwardable-extended (2.6.0)
-    gemoji (3.0.1)
-    github-pages (225)
-      github-pages-health-check (= 1.17.9)
-      jekyll (= 3.9.0)
-      jekyll-avatar (= 0.7.0)
-      jekyll-coffeescript (= 1.1.1)
-      jekyll-commonmark-ghpages (= 0.2.0)
-      jekyll-default-layout (= 0.1.4)
-      jekyll-feed (= 0.15.1)
-      jekyll-gist (= 1.5.0)
-      jekyll-github-metadata (= 2.13.0)
-      jekyll-include-cache (= 0.2.1)
-      jekyll-mentions (= 1.6.0)
-      jekyll-optional-front-matter (= 0.3.2)
-      jekyll-paginate (= 1.1.0)
-      jekyll-readme-index (= 0.3.0)
-      jekyll-redirect-from (= 0.16.0)
-      jekyll-relative-links (= 0.6.1)
-      jekyll-remote-theme (= 0.4.3)
-      jekyll-sass-converter (= 1.5.2)
-      jekyll-seo-tag (= 2.8.0)
-      jekyll-sitemap (= 1.4.0)
-      jekyll-swiss (= 1.0.0)
-      jekyll-theme-architect (= 0.2.0)
-      jekyll-theme-cayman (= 0.2.0)
-      jekyll-theme-dinky (= 0.2.0)
-      jekyll-theme-hacker (= 0.2.0)
-      jekyll-theme-leap-day (= 0.2.0)
-      jekyll-theme-merlot (= 0.2.0)
-      jekyll-theme-midnight (= 0.2.0)
-      jekyll-theme-minimal (= 0.2.0)
-      jekyll-theme-modernist (= 0.2.0)
-      jekyll-theme-primer (= 0.6.0)
-      jekyll-theme-slate (= 0.2.0)
-      jekyll-theme-tactile (= 0.2.0)
-      jekyll-theme-time-machine (= 0.2.0)
-      jekyll-titles-from-headings (= 0.5.3)
-      jemoji (= 0.12.0)
-      kramdown (= 2.3.1)
-      kramdown-parser-gfm (= 1.1.0)
-      liquid (= 4.0.3)
-      mercenary (~> 0.3)
-      minima (= 2.5.1)
-      nokogiri (>= 1.12.5, < 2.0)
-      rouge (= 3.26.0)
-      terminal-table (~> 1.4)
-    github-pages-health-check (1.17.9)
-      addressable (~> 2.3)
-      dnsruby (~> 1.60)
-      octokit (~> 4.0)
-      public_suffix (>= 3.0, < 5.0)
-      typhoeus (~> 1.3)
-    html-pipeline (2.14.0)
-      activesupport (>= 2)
-      nokogiri (>= 1.4)
-    http_parser.rb (0.8.0)
-    i18n (0.9.5)
-      concurrent-ruby (~> 1.0)
-    jekyll (3.9.0)
-      addressable (~> 2.4)
-      colorator (~> 1.0)
-      em-websocket (~> 0.5)
-      i18n (~> 0.7)
-      jekyll-sass-converter (~> 1.0)
-      jekyll-watch (~> 2.0)
-      kramdown (>= 1.17, < 3)
-      liquid (~> 4.0)
-      mercenary (~> 0.3.3)
-      pathutil (~> 0.9)
-      rouge (>= 1.7, < 4)
-      safe_yaml (~> 1.0)
-    jekyll-avatar (0.7.0)
-      jekyll (>= 3.0, < 5.0)
-    jekyll-coffeescript (1.1.1)
-      coffee-script (~> 2.2)
-      coffee-script-source (~> 1.11.1)
-    jekyll-commonmark (1.4.0)
-      commonmarker (~> 0.22)
-    jekyll-commonmark-ghpages (0.2.0)
-      commonmarker (~> 0.23.4)
-      jekyll (~> 3.9.0)
-      jekyll-commonmark (~> 1.4.0)
-      rouge (>= 2.0, < 4.0)
-    jekyll-default-layout (0.1.4)
-      jekyll (~> 3.0)
-    jekyll-feed (0.15.1)
-      jekyll (>= 3.7, < 5.0)
-    jekyll-gist (1.5.0)
-      octokit (~> 4.2)
-    jekyll-github-metadata (2.13.0)
-      jekyll (>= 3.4, < 5.0)
-      octokit (~> 4.0, != 4.4.0)
-    jekyll-include-cache (0.2.1)
-      jekyll (>= 3.7, < 5.0)
-    jekyll-mentions (1.6.0)
-      html-pipeline (~> 2.3)
-      jekyll (>= 3.7, < 5.0)
-    jekyll-optional-front-matter (0.3.2)
-      jekyll (>= 3.0, < 5.0)
-    jekyll-paginate (1.1.0)
-    jekyll-readme-index (0.3.0)
-      jekyll (>= 3.0, < 5.0)
-    jekyll-redirect-from (0.16.0)
-      jekyll (>= 3.3, < 5.0)
-    jekyll-relative-links (0.6.1)
-      jekyll (>= 3.3, < 5.0)
-    jekyll-remote-theme (0.4.3)
-      addressable (~> 2.0)
-      jekyll (>= 3.5, < 5.0)
-      jekyll-sass-converter (>= 1.0, <= 3.0.0, != 2.0.0)
-      rubyzip (>= 1.3.0, < 3.0)
-    jekyll-sass-converter (1.5.2)
-      sass (~> 3.4)
-    jekyll-seo-tag (2.8.0)
-      jekyll (>= 3.8, < 5.0)
-    jekyll-sitemap (1.4.0)
-      jekyll (>= 3.7, < 5.0)
-    jekyll-swiss (1.0.0)
-    jekyll-theme-architect (0.2.0)
-      jekyll (> 3.5, < 5.0)
-      jekyll-seo-tag (~> 2.0)
-    jekyll-theme-cayman (0.2.0)
-      jekyll (> 3.5, < 5.0)
-      jekyll-seo-tag (~> 2.0)
-    jekyll-theme-dinky (0.2.0)
-      jekyll (> 3.5, < 5.0)
-      jekyll-seo-tag (~> 2.0)
-    jekyll-theme-hacker (0.2.0)
-      jekyll (> 3.5, < 5.0)
-      jekyll-seo-tag (~> 2.0)
-    jekyll-theme-leap-day (0.2.0)
-      jekyll (> 3.5, < 5.0)
-      jekyll-seo-tag (~> 2.0)
-    jekyll-theme-merlot (0.2.0)
-      jekyll (> 3.5, < 5.0)
-      jekyll-seo-tag (~> 2.0)
-    jekyll-theme-midnight (0.2.0)
-      jekyll (> 3.5, < 5.0)
-      jekyll-seo-tag (~> 2.0)
-    jekyll-theme-minimal (0.2.0)
-      jekyll (> 3.5, < 5.0)
-      jekyll-seo-tag (~> 2.0)
-    jekyll-theme-modernist (0.2.0)
-      jekyll (> 3.5, < 5.0)
-      jekyll-seo-tag (~> 2.0)
-    jekyll-theme-primer (0.6.0)
-      jekyll (> 3.5, < 5.0)
-      jekyll-github-metadata (~> 2.9)
-      jekyll-seo-tag (~> 2.0)
-    jekyll-theme-slate (0.2.0)
-      jekyll (> 3.5, < 5.0)
-      jekyll-seo-tag (~> 2.0)
-    jekyll-theme-tactile (0.2.0)
-      jekyll (> 3.5, < 5.0)
-      jekyll-seo-tag (~> 2.0)
-    jekyll-theme-time-machine (0.2.0)
-      jekyll (> 3.5, < 5.0)
-      jekyll-seo-tag (~> 2.0)
-    jekyll-titles-from-headings (0.5.3)
-      jekyll (>= 3.3, < 5.0)
-    jekyll-watch (2.2.1)
-      listen (~> 3.0)
-    jemoji (0.12.0)
-      gemoji (~> 3.0)
-      html-pipeline (~> 2.2)
-      jekyll (>= 3.0, < 5.0)
-    kramdown (2.3.1)
-      rexml
-    kramdown-parser-gfm (1.1.0)
-      kramdown (~> 2.0)
-    liquid (4.0.3)
-    listen (3.7.1)
-      rb-fsevent (~> 0.10, >= 0.10.3)
-      rb-inotify (~> 0.9, >= 0.9.10)
-    mercenary (0.3.6)
-    mini_portile2 (2.8.0)
-    minima (2.5.1)
-      jekyll (>= 3.5, < 5.0)
-      jekyll-feed (~> 0.9)
-      jekyll-seo-tag (~> 2.1)
-    minitest (5.15.0)
-    multipart-post (2.1.1)
-    nokogiri (1.13.10)
-      mini_portile2 (~> 2.8.0)
-      racc (~> 1.4)
-    octokit (4.22.0)
-      faraday (>= 0.9)
-      sawyer (~> 0.8.0, >= 0.5.3)
-    pathutil (0.16.2)
-      forwardable-extended (~> 2.6)
-    public_suffix (4.0.6)
-    racc (1.6.1)
-    rb-fsevent (0.11.1)
-    rb-inotify (0.10.1)
-      ffi (~> 1.0)
-    rexml (3.2.5)
-    rouge (3.26.0)
-    ruby2_keywords (0.0.5)
-    rubyzip (2.3.2)
-    safe_yaml (1.0.5)
-    sass (3.7.4)
-      sass-listen (~> 4.0.0)
-    sass-listen (4.0.0)
-      rb-fsevent (~> 0.9, >= 0.9.4)
-      rb-inotify (~> 0.9, >= 0.9.7)
-    sawyer (0.8.2)
-      addressable (>= 2.3.5)
-      faraday (> 0.8, < 2.0)
-    simpleidn (0.2.1)
-      unf (~> 0.1.4)
-    terminal-table (1.8.0)
-      unicode-display_width (~> 1.1, >= 1.1.1)
-    thread_safe (0.3.6)
-    typhoeus (1.4.0)
-      ethon (>= 0.9.0)
-    tzinfo (1.2.10)
-      thread_safe (~> 0.1)
-    unf (0.1.4)
-      unf_ext
-    unf_ext (0.0.8)
-    unicode-display_width (1.8.0)
-    webrick (1.7.0)
-    zeitwerk (2.5.4)
-
-PLATFORMS
-  ruby
-
-DEPENDENCIES
-  github-pages (~> 225)
-  webrick (~> 1.7)
-
-BUNDLED WITH
-   2.2.3

env/env.cc

@@ -641,7 +641,7 @@ Status Env::CreateFromString(const ConfigOptions& config_options,
   } else {
     RegisterSystemEnvs();
     Env* env = *result;
-    Status s = LoadStaticObject<Env>(config_options, value, nullptr, &env);
+    Status s = LoadStaticObject<Env>(config_options, value, &env);
     if (s.ok()) {
       *result = env;
     }
@@ -1227,8 +1227,7 @@ Status SystemClock::CreateFromString(const ConfigOptions& config_options,
     std::call_once(once, [&]() {
       RegisterBuiltinSystemClocks(*(ObjectLibrary::Default().get()), "");
     });
-    return LoadSharedObject<SystemClock>(config_options, value, nullptr,
-                                         result);
+    return LoadSharedObject<SystemClock>(config_options, value, result);
   }
 }
 }  // namespace ROCKSDB_NAMESPACE

env/env_encryption.cc

@@ -1332,15 +1332,14 @@ Status BlockCipher::CreateFromString(const ConfigOptions& config_options,
                                      const std::string& value,
                                      std::shared_ptr<BlockCipher>* result) {
   RegisterEncryptionBuiltins();
-  return LoadSharedObject<BlockCipher>(config_options, value, nullptr, result);
+  return LoadSharedObject<BlockCipher>(config_options, value, result);
 }
 
 Status EncryptionProvider::CreateFromString(
     const ConfigOptions& config_options, const std::string& value,
     std::shared_ptr<EncryptionProvider>* result) {
   RegisterEncryptionBuiltins();
-  return LoadSharedObject<EncryptionProvider>(config_options, value, nullptr,
-                                              result);
+  return LoadSharedObject<EncryptionProvider>(config_options, value, result);
 }
 
 

env/file_system.cc

@@ -91,7 +91,7 @@ Status FileSystem::CreateFromString(const ConfigOptions& config_options,
     std::call_once(once, [&]() {
       RegisterBuiltinFileSystems(*(ObjectLibrary::Default().get()), "");
     });
-    return LoadSharedObject<FileSystem>(config_options, value, nullptr, result);
+    return LoadSharedObject<FileSystem>(config_options, value, result);
   }
 }
 

file/file_prefetch_buffer.cc

@@ -569,10 +569,11 @@ Status FilePrefetchBuffer::PrefetchAsyncInternal(
     bufs_[second].offset_ = rounddown_start2;
     assert(roundup_len2 >= chunk_len2);
     uint64_t read_len2 = static_cast<size_t>(roundup_len2 - chunk_len2);
-    Status tmp_s = ReadAsync(opts, reader, read_len2, rounddown_start2, second);
-    if (!tmp_s.ok()) {
+    s = ReadAsync(opts, reader, read_len2, rounddown_start2, second);
+    if (!s.ok()) {
       DestroyAndClearIOHandle(second);
       bufs_[second].buffer_.Clear();
+      return s;
     }
   }
 

file/prefetch_test.cc

@@ -1609,27 +1609,23 @@ TEST_P(PrefetchTest, ReadAsyncWithPosixFS) {
       num_keys++;
     }
 
-    ASSERT_EQ(num_keys, total_keys);
-    ASSERT_GT(buff_prefetch_count, 0);
-
-    // Check stats to make sure async prefetch is done.
-    {
+    if (read_async_called) {
+      ASSERT_EQ(num_keys, total_keys);
+      ASSERT_GT(buff_prefetch_count, 0);
+      // Check stats to make sure async prefetch is done.
       HistogramData async_read_bytes;
       options.statistics->histogramData(ASYNC_READ_BYTES, &async_read_bytes);
       HistogramData prefetched_bytes_discarded;
       options.statistics->histogramData(PREFETCHED_BYTES_DISCARDED,
                                         &prefetched_bytes_discarded);
-
+      ASSERT_GT(async_read_bytes.count, 0);
+      ASSERT_GT(prefetched_bytes_discarded.count, 0);
+      ASSERT_EQ(get_perf_context()->number_async_seek, 0);
+    } else {
       // Not all platforms support iouring. In that case, ReadAsync in posix
       // won't submit async requests.
-      if (read_async_called) {
-        ASSERT_GT(async_read_bytes.count, 0);
-      } else {
-        ASSERT_EQ(async_read_bytes.count, 0);
-      }
-      ASSERT_GT(prefetched_bytes_discarded.count, 0);
+      ASSERT_EQ(iter->status(), Status::NotSupported());
     }
-    ASSERT_EQ(get_perf_context()->number_async_seek, 0);
   }
 
   SyncPoint::GetInstance()->DisableProcessing();
@@ -1740,22 +1736,19 @@ TEST_P(PrefetchTest, MultipleSeekWithPosixFS) {
         num_keys++;
         iter->Next();
       }
-      ASSERT_OK(iter->status());
-      ASSERT_EQ(num_keys, num_keys_first_batch);
-      // Check stats to make sure async prefetch is done.
-      {
+
+      if (read_async_called) {
+        ASSERT_OK(iter->status());
+        ASSERT_EQ(num_keys, num_keys_first_batch);
+        // Check stats to make sure async prefetch is done.
         HistogramData async_read_bytes;
         options.statistics->histogramData(ASYNC_READ_BYTES, &async_read_bytes);
-
+        ASSERT_GT(async_read_bytes.count, 0);
+        ASSERT_GT(get_perf_context()->number_async_seek, 0);
+      } else {
         // Not all platforms support iouring. In that case, ReadAsync in posix
         // won't submit async requests.
-        if (read_async_called) {
-          ASSERT_GT(async_read_bytes.count, 0);
-          ASSERT_GT(get_perf_context()->number_async_seek, 0);
-        } else {
-          ASSERT_EQ(async_read_bytes.count, 0);
-          ASSERT_EQ(get_perf_context()->number_async_seek, 0);
-        }
+        ASSERT_EQ(iter->status(), Status::NotSupported());
       }
     }
 
@@ -1771,29 +1764,26 @@ TEST_P(PrefetchTest, MultipleSeekWithPosixFS) {
         num_keys++;
         iter->Next();
       }
-      ASSERT_OK(iter->status());
-      ASSERT_EQ(num_keys, num_keys_second_batch);
 
-      ASSERT_GT(buff_prefetch_count, 0);
+      if (read_async_called) {
+        ASSERT_OK(iter->status());
+        ASSERT_EQ(num_keys, num_keys_second_batch);
 
-      // Check stats to make sure async prefetch is done.
-      {
+        ASSERT_GT(buff_prefetch_count, 0);
+
+        // Check stats to make sure async prefetch is done.
         HistogramData async_read_bytes;
         options.statistics->histogramData(ASYNC_READ_BYTES, &async_read_bytes);
         HistogramData prefetched_bytes_discarded;
         options.statistics->histogramData(PREFETCHED_BYTES_DISCARDED,
                                           &prefetched_bytes_discarded);
-
+        ASSERT_GT(async_read_bytes.count, 0);
+        ASSERT_GT(get_perf_context()->number_async_seek, 0);
+        ASSERT_GT(prefetched_bytes_discarded.count, 0);
+      } else {
         // Not all platforms support iouring. In that case, ReadAsync in posix
         // won't submit async requests.
-        if (read_async_called) {
-          ASSERT_GT(async_read_bytes.count, 0);
-          ASSERT_GT(get_perf_context()->number_async_seek, 0);
-        } else {
-          ASSERT_EQ(async_read_bytes.count, 0);
-          ASSERT_EQ(get_perf_context()->number_async_seek, 0);
-        }
-        ASSERT_GT(prefetched_bytes_discarded.count, 0);
+        ASSERT_EQ(iter->status(), Status::NotSupported());
       }
     }
   }
@@ -1872,51 +1862,52 @@ TEST_P(PrefetchTest, SeekParallelizationTestWithPosix) {
     // Each block contains around 4 keys.
     auto iter = std::unique_ptr<Iterator>(db_->NewIterator(ro));
     iter->Seek(BuildKey(0));  // Prefetch data because of seek parallelization.
-    ASSERT_TRUE(iter->Valid());
-    iter->Next();
-    ASSERT_TRUE(iter->Valid());
-    iter->Next();
-    ASSERT_TRUE(iter->Valid());
-    iter->Next();
-    ASSERT_TRUE(iter->Valid());
+    if (std::get<1>(GetParam()) && !read_async_called) {
+      ASSERT_EQ(iter->status(), Status::NotSupported());
+    } else {
+      ASSERT_TRUE(iter->Valid());
+      iter->Next();
+      ASSERT_TRUE(iter->Valid());
+      iter->Next();
+      ASSERT_TRUE(iter->Valid());
+      iter->Next();
+      ASSERT_TRUE(iter->Valid());
 
-    // New data block. Since num_file_reads in FilePrefetch after this read is
-    // 2, it won't go for prefetching.
-    iter->Next();
-    ASSERT_TRUE(iter->Valid());
-    iter->Next();
-    ASSERT_TRUE(iter->Valid());
-    iter->Next();
-    ASSERT_TRUE(iter->Valid());
-    iter->Next();
-    ASSERT_TRUE(iter->Valid());
+      // New data block. Since num_file_reads in FilePrefetch after this read is
+      // 2, it won't go for prefetching.
+      iter->Next();
+      ASSERT_TRUE(iter->Valid());
+      iter->Next();
+      ASSERT_TRUE(iter->Valid());
+      iter->Next();
+      ASSERT_TRUE(iter->Valid());
+      iter->Next();
+      ASSERT_TRUE(iter->Valid());
 
-    // Prefetch data.
-    iter->Next();
-    ASSERT_TRUE(iter->Valid());
+      // Prefetch data.
+      iter->Next();
 
-    // Check stats to make sure async prefetch is done.
-    {
-      HistogramData async_read_bytes;
-      options.statistics->histogramData(ASYNC_READ_BYTES, &async_read_bytes);
-      // Not all platforms support iouring. In that case, ReadAsync in posix
-      // won't submit async requests.
       if (read_async_called) {
-        ASSERT_GT(async_read_bytes.count, 0);
-        ASSERT_GT(get_perf_context()->number_async_seek, 0);
-        if (std::get<1>(GetParam())) {
-          ASSERT_EQ(buff_prefetch_count, 1);
-        } else {
-          ASSERT_EQ(buff_prefetch_count, 2);
+        ASSERT_TRUE(iter->Valid());
+        // Check stats to make sure async prefetch is done.
+        {
+          HistogramData async_read_bytes;
+          options.statistics->histogramData(ASYNC_READ_BYTES,
+                                            &async_read_bytes);
+          ASSERT_GT(async_read_bytes.count, 0);
+          ASSERT_GT(get_perf_context()->number_async_seek, 0);
+          if (std::get<1>(GetParam())) {
+            ASSERT_EQ(buff_prefetch_count, 1);
+          } else {
+            ASSERT_EQ(buff_prefetch_count, 2);
+          }
         }
       } else {
-        ASSERT_EQ(async_read_bytes.count, 0);
-        ASSERT_EQ(get_perf_context()->number_async_seek, 0);
-        ASSERT_EQ(buff_prefetch_count, 1);
+        // Not all platforms support iouring. In that case, ReadAsync in posix
+        // won't submit async requests.
+        ASSERT_EQ(iter->status(), Status::NotSupported());
       }
     }
-
-    buff_prefetch_count = 0;
   }
   Close();
 }
@@ -2009,20 +2000,17 @@ TEST_P(PrefetchTest, TraceReadAsyncWithCallbackWrapper) {
     ASSERT_OK(db_->EndIOTrace());
     ASSERT_OK(env_->FileExists(trace_file_path));
 
-    ASSERT_EQ(num_keys, total_keys);
-    ASSERT_GT(buff_prefetch_count, 0);
-
-    // Check stats to make sure async prefetch is done.
-    {
+    if (read_async_called) {
+      ASSERT_EQ(num_keys, total_keys);
+      ASSERT_GT(buff_prefetch_count, 0);
+      // Check stats to make sure async prefetch is done.
       HistogramData async_read_bytes;
       options.statistics->histogramData(ASYNC_READ_BYTES, &async_read_bytes);
+      ASSERT_GT(async_read_bytes.count, 0);
+    } else {
       // Not all platforms support iouring. In that case, ReadAsync in posix
       // won't submit async requests.
-      if (read_async_called) {
-        ASSERT_GT(async_read_bytes.count, 0);
-      } else {
-        ASSERT_EQ(async_read_bytes.count, 0);
-      }
+      ASSERT_EQ(iter->status(), Status::NotSupported());
     }
 
     // Check the file to see if ReadAsync is logged.
@@ -2102,8 +2090,13 @@ TEST_F(FilePrefetchBufferTest, SeekWithBlockCacheHit) {
   // Simulate a seek of 4096 bytes at offset 0. Due to the readahead settings,
   // it will do two reads of 4096+8192 and 8192
   Status s = fpb.PrefetchAsync(IOOptions(), r.get(), 0, 4096, &result);
-  // Platforms that don't have IO uring may not support async IO
-  ASSERT_TRUE(s.IsTryAgain() || s.IsNotSupported());
+
+  // Platforms that don't have IO uring may not support async IO.
+  if (s.IsNotSupported()) {
+    return;
+  }
+
+  ASSERT_TRUE(s.IsTryAgain());
   // Simulate a block cache hit
   fpb.UpdateReadPattern(0, 4096, false);
   // Now read some data that straddles the two prefetch buffers - offset 8192 to
@@ -2137,9 +2130,13 @@ TEST_F(FilePrefetchBufferTest, NoSyncWithAsyncIO) {
   // Simulate a seek of 4000 bytes at offset 3000. Due to the readahead
   // settings, it will do two reads of 4000+4096 and 4096
   Status s = fpb.PrefetchAsync(IOOptions(), r.get(), 3000, 4000, &async_result);
-  // Platforms that don't have IO uring may not support async IO
-  ASSERT_TRUE(s.IsTryAgain() || s.IsNotSupported());
 
+  // Platforms that don't have IO uring may not support async IO
+  if (s.IsNotSupported()) {
+    return;
+  }
+
+  ASSERT_TRUE(s.IsTryAgain());
   ASSERT_TRUE(fpb.TryReadFromCacheAsync(IOOptions(), r.get(), /*offset=*/3000,
                                         /*length=*/4000, &async_result, &s,
                                         Env::IOPriority::IO_LOW));

include/rocksdb/advanced_cache.h

@@ -0,0 +1,527 @@
+//  Copyright (c) Meta Platforms, Inc. and affiliates.
+//  This source code is licensed under both the GPLv2 (found in the
+//  COPYING file in the root directory) and Apache 2.0 License
+//  (found in the LICENSE.Apache file in the root directory).
+//
+// APIs for customizing read caches in RocksDB.
+
+#pragma once
+
+#include <cstdint>
+#include <functional>
+#include <memory>
+#include <string>
+
+#include "rocksdb/cache.h"
+#include "rocksdb/memory_allocator.h"
+#include "rocksdb/slice.h"
+#include "rocksdb/status.h"
+
+namespace ROCKSDB_NAMESPACE {
+
+class Logger;
+class Statistics;
+
+// A Cache maps keys to objects resident in memory, tracks reference counts
+// on those key-object entries, and is able to remove unreferenced entries
+// whenever it wants. All operations are fully thread safe except as noted.
+// Inserted entries have a specified "charge" which is some quantity in
+// unspecified units, typically bytes of memory used. A Cache will typically
+// have a finite capacity in units of charge, and evict entries as needed
+// to stay at or below that capacity.
+//
+// NOTE: This API is for expert use only and is intended more for customizing
+// cache behavior than for calling into outside of RocksDB. It is subject to
+// change as RocksDB evolves, especially the RocksDB block cache. Overriding
+// CacheWrapper is the preferred way of customizing some operations on an
+// existing implementation.
+//
+// INTERNAL: See typed_cache.h for convenient wrappers on top of this API.
+// New virtual functions must also be added to CacheWrapper below.
+class Cache {
+ public:  // types hidden from API client
+  // Opaque handle to an entry stored in the cache.
+  struct Handle {};
+
+ public:  // types hidden from Cache implementation
+  // Pointer to cached object of unspecified type. (This type alias is
+  // provided for clarity, not really for type checking.)
+  using ObjectPtr = void*;
+
+  // Opaque object providing context (settings, etc.) to create objects
+  // for primary cache from saved (serialized) secondary cache entries.
+  struct CreateContext {};
+
+ public:  // type defs
+  // Depending on implementation, cache entries with higher priority levels
+  // could be less likely to get evicted than entries with lower priority
+  // levels. The "high" priority level applies to certain SST metablocks (e.g.
+  // index and filter blocks) if the option
+  // cache_index_and_filter_blocks_with_high_priority is set. The "low" priority
+  // level is used for other kinds of SST blocks (most importantly, data
+  // blocks), as well as the above metablocks in case
+  // cache_index_and_filter_blocks_with_high_priority is
+  // not set. The "bottom" priority level is for BlobDB's blob values.
+  enum class Priority { HIGH, LOW, BOTTOM };
+
+  // A set of callbacks to allow objects in the primary block cache to be
+  // be persisted in a secondary cache. The purpose of the secondary cache
+  // is to support other ways of caching the object, such as persistent or
+  // compressed data, that may require the object to be parsed and transformed
+  // in some way. Since the primary cache holds C++ objects and the secondary
+  // cache may only hold flat data that doesn't need relocation, these
+  // callbacks need to be provided by the user of the block
+  // cache to do the conversion.
+  // The CacheItemHelper is passed to Insert() and Lookup(). It has pointers
+  // to callback functions for size, saving and deletion of the
+  // object. The callbacks are defined in C-style in order to make them
+  // stateless and not add to the cache metadata size.
+  // Saving multiple std::function objects will take up 32 bytes per
+  // function, even if its not bound to an object and does no capture.
+  //
+  // All the callbacks are C-style function pointers in order to simplify
+  // lifecycle management. Objects in the cache can outlive the parent DB,
+  // so anything required for these operations should be contained in the
+  // object itself.
+  //
+  // The SizeCallback takes a pointer to the object and returns the size
+  // of the persistable data. It can be used by the secondary cache to allocate
+  // memory if needed.
+  //
+  // RocksDB callbacks are NOT exception-safe. A callback completing with an
+  // exception can lead to undefined behavior in RocksDB, including data loss,
+  // unreported corruption, deadlocks, and more.
+  using SizeCallback = size_t (*)(ObjectPtr obj);
+
+  // The SaveToCallback takes an object pointer and saves the persistable
+  // data into a buffer. The secondary cache may decide to not store it in a
+  // contiguous buffer, in which case this callback will be called multiple
+  // times with increasing offset
+  using SaveToCallback = Status (*)(ObjectPtr from_obj, size_t from_offset,
+                                    size_t length, char* out_buf);
+
+  // A function pointer type for destruction of a cache object. This will
+  // typically call the destructor for the appropriate type of the object.
+  // The Cache is responsible for copying and reclaiming space for the key,
+  // but objects are managed in part using this callback. Generally a DeleterFn
+  // can be nullptr if the ObjectPtr does not need destruction (e.g. nullptr or
+  // pointer into static data).
+  using DeleterFn = void (*)(ObjectPtr obj, MemoryAllocator* allocator);
+
+  // The CreateCallback is takes in a buffer from the NVM cache and constructs
+  // an object using it. The callback doesn't have ownership of the buffer and
+  // should copy the contents into its own buffer. The CreateContext* is
+  // provided by Lookup and may be used to follow DB- or CF-specific settings.
+  // In case of some error, non-OK is returned and the caller should ignore
+  // any result in out_obj. (The implementation must clean up after itself.)
+  using CreateCallback = Status (*)(const Slice& data, CreateContext* context,
+                                    MemoryAllocator* allocator,
+                                    ObjectPtr* out_obj, size_t* out_charge);
+
+  // A struct with pointers to helper functions for spilling items from the
+  // cache into the secondary cache. May be extended in the future. An
+  // instance of this struct is expected to outlive the cache.
+  struct CacheItemHelper {
+    // Function for deleting an object on its removal from the Cache.
+    // nullptr is only for entries that require no destruction, such as
+    // "placeholder" cache entries with nullptr object.
+    DeleterFn del_cb;  // (<- Most performance critical)
+    // Next three are used for persisting values as described above.
+    // If any is nullptr, then all three should be nullptr and persisting the
+    // entry to/from secondary cache is not supported.
+    SizeCallback size_cb;
+    SaveToCallback saveto_cb;
+    CreateCallback create_cb;
+    // Classification of the entry for monitoring purposes in block cache.
+    CacheEntryRole role;
+
+    constexpr CacheItemHelper()
+        : del_cb(nullptr),
+          size_cb(nullptr),
+          saveto_cb(nullptr),
+          create_cb(nullptr),
+          role(CacheEntryRole::kMisc) {}
+
+    explicit constexpr CacheItemHelper(CacheEntryRole _role,
+                                       DeleterFn _del_cb = nullptr,
+                                       SizeCallback _size_cb = nullptr,
+                                       SaveToCallback _saveto_cb = nullptr,
+                                       CreateCallback _create_cb = nullptr)
+        : del_cb(_del_cb),
+          size_cb(_size_cb),
+          saveto_cb(_saveto_cb),
+          create_cb(_create_cb),
+          role(_role) {
+      // Either all three secondary cache callbacks are non-nullptr or
+      // all three are nullptr
+      assert((size_cb != nullptr) == (saveto_cb != nullptr));
+      assert((size_cb != nullptr) == (create_cb != nullptr));
+    }
+    inline bool IsSecondaryCacheCompatible() const {
+      return size_cb != nullptr;
+    }
+  };
+
+ public:  // ctor/dtor/create
+  Cache(std::shared_ptr<MemoryAllocator> allocator = nullptr)
+      : memory_allocator_(std::move(allocator)) {}
+  // No copying allowed
+  Cache(const Cache&) = delete;
+  Cache& operator=(const Cache&) = delete;
+
+  // Destroys all remaining entries by calling the associated "deleter"
+  virtual ~Cache() {}
+
+  // Creates a new Cache based on the input value string and returns the result.
+  // Currently, this method can be used to create LRUCaches only
+  // @param config_options
+  // @param value  The value might be:
+  //   - an old-style cache ("1M") -- equivalent to NewLRUCache(1024*102(
+  //   - Name-value option pairs -- "capacity=1M; num_shard_bits=4;
+  //     For the LRUCache, the values are defined in LRUCacheOptions.
+  // @param result The new Cache object
+  // @return OK if the cache was successfully created
+  // @return NotFound if an invalid name was specified in the value
+  // @return InvalidArgument if either the options were not valid
+  static Status CreateFromString(const ConfigOptions& config_options,
+                                 const std::string& value,
+                                 std::shared_ptr<Cache>* result);
+
+ public:  // functions
+  // The type of the Cache
+  virtual const char* Name() const = 0;
+
+  // The Insert and Lookup APIs below are intended to allow cached objects
+  // to be demoted/promoted between the primary block cache and a secondary
+  // cache. The secondary cache could be a non-volatile cache, and will
+  // likely store the object in a different representation. They rely on a
+  // per object CacheItemHelper to do the conversions.
+  // The secondary cache may persist across process and system restarts,
+  // and may even be moved between hosts. Therefore, the cache key must
+  // be repeatable across restarts/reboots, and globally unique if
+  // multiple DBs share the same cache and the set of DBs can change
+  // over time.
+
+  // Insert a mapping from key->object into the cache and assign it
+  // the specified charge against the total cache capacity. If
+  // strict_capacity_limit is true and cache reaches its full capacity,
+  // return Status::MemoryLimit. `obj` must be non-nullptr if compatible
+  // with secondary cache (helper->size_cb != nullptr), because Value() ==
+  // nullptr is reserved for indicating some secondary cache failure cases.
+  // On success, returns OK and takes ownership of `obj`, eventually deleting
+  // it with helper->del_cb. On non-OK return, the caller maintains ownership
+  // of `obj` so will often need to delete it in such cases.
+  //
+  // The helper argument is saved by the cache and will be used when the
+  // inserted object is evicted or considered for promotion to the secondary
+  // cache. Promotion to secondary cache is only enabled if helper->size_cb
+  // != nullptr. The helper must outlive the cache. Callers may use
+  // &kNoopCacheItemHelper as a trivial helper (no deleter for the object,
+  // no secondary cache). `helper` must not be nullptr (efficiency).
+  //
+  // If `handle` is not nullptr and return status is OK, `handle` is set
+  // to a Handle* for the entry. The caller must call this->Release(handle)
+  // when the returned entry is no longer needed. If `handle` is nullptr, it is
+  // as if Release is called immediately after Insert.
+  //
+  // Regardless of whether the item was inserted into the cache,
+  // it will attempt to insert it into the secondary cache if one is
+  // configured, and the helper supports it.
+  // The cache implementation must support a secondary cache, otherwise
+  // the item is only inserted into the primary cache. It may
+  // defer the insertion to the secondary cache as it sees fit.
+  //
+  // When the inserted entry is no longer needed, it will be destroyed using
+  // helper->del_cb (if non-nullptr).
+  virtual Status Insert(const Slice& key, ObjectPtr obj,
+                        const CacheItemHelper* helper, size_t charge,
+                        Handle** handle = nullptr,
+                        Priority priority = Priority::LOW) = 0;
+
+  // Lookup the key, returning nullptr if not found. If found, returns
+  // a handle to the mapping that must eventually be passed to Release().
+  //
+  // If a non-nullptr helper argument is provided with a non-nullptr
+  // create_cb, and a secondary cache is configured, then the secondary
+  // cache is also queried if lookup in the primary cache fails. If found
+  // in secondary cache, the provided create_db and create_context are
+  // used to promote the entry to an object in the primary cache.
+  // In that case, the helper may be saved and used later when the object
+  // is evicted, so as usual, the pointed-to helper must outlive the cache.
+  //
+  // ======================== Async Lookup (wait=false) ======================
+  // When wait=false, the handle returned might be in any of three states:
+  // * Present - If Value() != nullptr, then the result is present and
+  // the handle can be used just as if wait=true.
+  // * Pending, not ready (IsReady() == false) - secondary cache is still
+  // working to retrieve the value. Might become ready any time.
+  // * Pending, ready (IsReady() == true) - secondary cache has the value
+  // but it has not been loaded as an object into primary cache. Call to
+  // Wait()/WaitAll() will not block.
+  //
+  // IMPORTANT: Pending handles are not thread-safe, and only these functions
+  // are allowed on them: Value(), IsReady(), Wait(), WaitAll(). Even Release()
+  // can only come after Wait() or WaitAll() even though a reference is held.
+  //
+  // Only Wait()/WaitAll() gets a Handle out of a Pending state. (Waiting is
+  // safe and has no effect on other handle states.) After waiting on a Handle,
+  // it is in one of two states:
+  // * Present - if Value() != nullptr
+  // * Failed - if Value() == nullptr, such as if the secondary cache
+  // initially thought it had the value but actually did not.
+  //
+  // Note that given an arbitrary Handle, the only way to distinguish the
+  // Pending+ready state from the Failed state is to Wait() on it. A cache
+  // entry not compatible with secondary cache can also have Value()==nullptr
+  // like the Failed state, but this is not generally a concern.
+  virtual Handle* Lookup(const Slice& key,
+                         const CacheItemHelper* helper = nullptr,
+                         CreateContext* create_context = nullptr,
+                         Priority priority = Priority::LOW, bool wait = true,
+                         Statistics* stats = nullptr) = 0;
+
+  // Convenience wrapper when secondary cache not supported
+  inline Handle* BasicLookup(const Slice& key, Statistics* stats) {
+    return Lookup(key, nullptr, nullptr, Priority::LOW, true, stats);
+  }
+
+  // Increments the reference count for the handle if it refers to an entry in
+  // the cache. Returns true if refcount was incremented; otherwise, returns
+  // false.
+  // REQUIRES: handle must have been returned by a method on *this.
+  virtual bool Ref(Handle* handle) = 0;
+
+  /**
+   * Release a mapping returned by a previous Lookup(). A released entry might
+   * still remain in cache in case it is later looked up by others. If
+   * erase_if_last_ref is set then it also erases it from the cache if there is
+   * no other reference to  it. Erasing it should call the deleter function that
+   * was provided when the entry was inserted.
+   *
+   * Returns true if the entry was also erased.
+   */
+  // REQUIRES: handle must not have been released yet.
+  // REQUIRES: handle must have been returned by a method on *this.
+  virtual bool Release(Handle* handle, bool erase_if_last_ref = false) = 0;
+
+  // Return the object assiciated with a handle returned by a successful
+  // Lookup(). For historical reasons, this is also known at the "value"
+  // associated with the key.
+  // REQUIRES: handle must not have been released yet.
+  // REQUIRES: handle must have been returned by a method on *this.
+  virtual ObjectPtr Value(Handle* handle) = 0;
+
+  // If the cache contains the entry for the key, erase it.  Note that the
+  // underlying entry will be kept around until all existing handles
+  // to it have been released.
+  virtual void Erase(const Slice& key) = 0;
+  // Return a new numeric id.  May be used by multiple clients who are
+  // sharding the same cache to partition the key space.  Typically the
+  // client will allocate a new id at startup and prepend the id to
+  // its cache keys.
+  virtual uint64_t NewId() = 0;
+
+  // sets the maximum configured capacity of the cache. When the new
+  // capacity is less than the old capacity and the existing usage is
+  // greater than new capacity, the implementation will do its best job to
+  // purge the released entries from the cache in order to lower the usage
+  virtual void SetCapacity(size_t capacity) = 0;
+
+  // Set whether to return error on insertion when cache reaches its full
+  // capacity.
+  virtual void SetStrictCapacityLimit(bool strict_capacity_limit) = 0;
+
+  // Get the flag whether to return error on insertion when cache reaches its
+  // full capacity.
+  virtual bool HasStrictCapacityLimit() const = 0;
+
+  // Returns the maximum configured capacity of the cache
+  virtual size_t GetCapacity() const = 0;
+
+  // Returns the memory size for the entries residing in the cache.
+  virtual size_t GetUsage() const = 0;
+
+  // Returns the number of entries currently tracked in the table. SIZE_MAX
+  // means "not supported." This is used for inspecting the load factor, along
+  // with GetTableAddressCount().
+  virtual size_t GetOccupancyCount() const { return SIZE_MAX; }
+
+  // Returns the number of ways the hash function is divided for addressing
+  // entries. Zero means "not supported." This is used for inspecting the load
+  // factor, along with GetOccupancyCount().
+  virtual size_t GetTableAddressCount() const { return 0; }
+
+  // Returns the memory size for a specific entry in the cache.
+  virtual size_t GetUsage(Handle* handle) const = 0;
+
+  // Returns the memory size for the entries in use by the system
+  virtual size_t GetPinnedUsage() const = 0;
+
+  // Returns the charge for the specific entry in the cache.
+  virtual size_t GetCharge(Handle* handle) const = 0;
+
+  // Returns the helper for the specified entry.
+  virtual const CacheItemHelper* GetCacheItemHelper(Handle* handle) const = 0;
+
+  // Call this on shutdown if you want to speed it up. Cache will disown
+  // any underlying data and will not free it on delete. This call will leak
+  // memory - call this only if you're shutting down the process.
+  // Any attempts of using cache after this call will fail terribly.
+  // Always delete the DB object before calling this method!
+  virtual void DisownData() {
+    // default implementation is noop
+  }
+
+  struct ApplyToAllEntriesOptions {
+    // If the Cache uses locks, setting `average_entries_per_lock` to
+    // a higher value suggests iterating over more entries each time a lock
+    // is acquired, likely reducing the time for ApplyToAllEntries but
+    // increasing latency for concurrent users of the Cache. Setting
+    // `average_entries_per_lock` to a smaller value could be helpful if
+    // callback is relatively expensive, such as using large data structures.
+    size_t average_entries_per_lock = 256;
+  };
+
+  // Apply a callback to all entries in the cache. The Cache must ensure
+  // thread safety but does not guarantee that a consistent snapshot of all
+  // entries is iterated over if other threads are operating on the Cache
+  // also.
+  virtual void ApplyToAllEntries(
+      const std::function<void(const Slice& key, ObjectPtr obj, size_t charge,
+                               const CacheItemHelper* helper)>& callback,
+      const ApplyToAllEntriesOptions& opts) = 0;
+
+  // Remove all entries.
+  // Prerequisite: no entry is referenced.
+  virtual void EraseUnRefEntries() = 0;
+
+  virtual std::string GetPrintableOptions() const { return ""; }
+
+  // Check for any warnings or errors in the operation of the cache and
+  // report them to the logger. This is intended only to be called
+  // periodically so does not need to be very efficient. (Obscure calling
+  // conventions for Logger inherited from env.h)
+  virtual void ReportProblems(
+      const std::shared_ptr<Logger>& /*info_log*/) const {}
+
+  MemoryAllocator* memory_allocator() const { return memory_allocator_.get(); }
+
+  // EXPERIMENTAL
+  // The following APIs are experimental and might change in the future.
+
+  // Release a mapping returned by a previous Lookup(). The "useful"
+  // parameter specifies whether the data was actually used or not,
+  // which may be used by the cache implementation to decide whether
+  // to consider it as a hit for retention purposes. As noted elsewhere,
+  // "pending" handles require Wait()/WaitAll() before Release().
+  virtual bool Release(Handle* handle, bool /*useful*/,
+                       bool erase_if_last_ref) {
+    return Release(handle, erase_if_last_ref);
+  }
+
+  // Determines if the handle returned by Lookup() can give a value without
+  // blocking, though Wait()/WaitAll() might be required to publish it to
+  // Value(). See secondary cache compatible Lookup() above for details.
+  // This call is not thread safe on "pending" handles.
+  virtual bool IsReady(Handle* /*handle*/) { return true; }
+
+  // Convert a "pending" handle into a full thread-shareable handle by
+  // * If necessary, wait until secondary cache finishes loading the value.
+  // * Construct the object for primary cache and set it in the handle.
+  // Even after Wait() on a pending handle, the caller must check for
+  // Value() == nullptr in case of failure. This call is not thread-safe
+  // on pending handles. This call has no effect on non-pending handles.
+  // See secondary cache compatible Lookup() above for details.
+  virtual void Wait(Handle* /*handle*/) {}
+
+  // Wait for a vector of handles to become ready. As with Wait(), the user
+  // should check the Value() of each handle for nullptr. This call is not
+  // thread-safe on pending handles.
+  virtual void WaitAll(std::vector<Handle*>& /*handles*/) {}
+
+ private:
+  std::shared_ptr<MemoryAllocator> memory_allocator_;
+};
+
+// A wrapper around Cache that can easily be extended with instrumentation,
+// etc.
+class CacheWrapper : public Cache {
+ public:
+  explicit CacheWrapper(std::shared_ptr<Cache> target)
+      : target_(std::move(target)) {}
+
+  // Only function that derived class must provide
+  // const char* Name() const override { ... }
+
+  Status Insert(const Slice& key, ObjectPtr value,
+                const CacheItemHelper* helper, size_t charge,
+                Handle** handle = nullptr,
+                Priority priority = Priority::LOW) override {
+    return target_->Insert(key, value, helper, charge, handle, priority);
+  }
+
+  Handle* Lookup(const Slice& key, const CacheItemHelper* helper,
+                 CreateContext* create_context,
+                 Priority priority = Priority::LOW, bool wait = true,
+                 Statistics* stats = nullptr) override {
+    return target_->Lookup(key, helper, create_context, priority, wait, stats);
+  }
+
+  bool Ref(Handle* handle) override { return target_->Ref(handle); }
+
+  using Cache::Release;
+  bool Release(Handle* handle, bool erase_if_last_ref = false) override {
+    return target_->Release(handle, erase_if_last_ref);
+  }
+
+  ObjectPtr Value(Handle* handle) override { return target_->Value(handle); }
+
+  void Erase(const Slice& key) override { target_->Erase(key); }
+  uint64_t NewId() override { return target_->NewId(); }
+
+  void SetCapacity(size_t capacity) override { target_->SetCapacity(capacity); }
+
+  void SetStrictCapacityLimit(bool strict_capacity_limit) override {
+    target_->SetStrictCapacityLimit(strict_capacity_limit);
+  }
+
+  bool HasStrictCapacityLimit() const override {
+    return target_->HasStrictCapacityLimit();
+  }
+
+  size_t GetCapacity() const override { return target_->GetCapacity(); }
+
+  size_t GetUsage() const override { return target_->GetUsage(); }
+
+  size_t GetUsage(Handle* handle) const override {
+    return target_->GetUsage(handle);
+  }
+
+  size_t GetPinnedUsage() const override { return target_->GetPinnedUsage(); }
+
+  size_t GetCharge(Handle* handle) const override {
+    return target_->GetCharge(handle);
+  }
+
+  const CacheItemHelper* GetCacheItemHelper(Handle* handle) const override {
+    return target_->GetCacheItemHelper(handle);
+  }
+
+  void ApplyToAllEntries(
+      const std::function<void(const Slice& key, ObjectPtr value, size_t charge,
+                               const CacheItemHelper* helper)>& callback,
+      const ApplyToAllEntriesOptions& opts) override {
+    target_->ApplyToAllEntries(callback, opts);
+  }
+
+  void EraseUnRefEntries() override { target_->EraseUnRefEntries(); }
+
+ protected:
+  std::shared_ptr<Cache> target_;
+};
+
+// Useful for cache entries requiring no clean-up, such as for cache
+// reservations
+inline constexpr Cache::CacheItemHelper kNoopCacheItemHelper{};
+
+}  // namespace ROCKSDB_NAMESPACE

include/rocksdb/cache.h

@@ -7,26 +7,22 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 //
-// Various APIs for creating and customizing read caches in RocksDB.
+// Various APIs for configuring, creating, and monitoring read caches.
 
 #pragma once
 
 #include <cstdint>
-#include <functional>
 #include <memory>
 #include <string>
 
 #include "rocksdb/compression_type.h"
+#include "rocksdb/data_structure.h"
 #include "rocksdb/memory_allocator.h"
-#include "rocksdb/slice.h"
-#include "rocksdb/statistics.h"
-#include "rocksdb/status.h"
 
 namespace ROCKSDB_NAMESPACE {
 
-class Cache;
+class Cache;  // defined in advanced_cache.h
 struct ConfigOptions;
-class Logger;
 class SecondaryCache;
 
 // Classifications of block cache entries.
@@ -75,6 +71,9 @@ constexpr uint32_t kNumCacheEntryRoles =
 // Obtain a hyphen-separated, lowercase name of a `CacheEntryRole`.
 const std::string& GetCacheEntryRoleName(CacheEntryRole);
 
+// A fast bit set for CacheEntryRoles
+using CacheEntryRoleSet = SmallEnumSet<CacheEntryRole, CacheEntryRole::kMisc>;
+
 // For use with `GetMapProperty()` for property
 // `DB::Properties::kBlockCacheEntryStats`. On success, the map will
 // be populated with all keys that can be obtained from these functions.
@@ -240,6 +239,10 @@ struct CompressedSecondaryCacheOptions : LRUCacheOptions {
   // into chunks so that they may better fit jemalloc bins.
   bool enable_custom_split_merge = false;
 
+  // Kinds of entries that should not be compressed, but can be stored.
+  // (Filter blocks are essentially non-compressible but others usually are.)
+  CacheEntryRoleSet do_not_compress_roles = {CacheEntryRole::kFilterBlock};
+
   CompressedSecondaryCacheOptions() {}
   CompressedSecondaryCacheOptions(
       size_t _capacity, int _num_shard_bits, bool _strict_capacity_limit,
@@ -250,14 +253,17 @@ struct CompressedSecondaryCacheOptions : LRUCacheOptions {
           kDefaultCacheMetadataChargePolicy,
       CompressionType _compression_type = CompressionType::kLZ4Compression,
       uint32_t _compress_format_version = 2,
-      bool _enable_custom_split_merge = false)
+      bool _enable_custom_split_merge = false,
+      const CacheEntryRoleSet& _do_not_compress_roles =
+          {CacheEntryRole::kFilterBlock})
       : LRUCacheOptions(_capacity, _num_shard_bits, _strict_capacity_limit,
                         _high_pri_pool_ratio, std::move(_memory_allocator),
                         _use_adaptive_mutex, _metadata_charge_policy,
                         _low_pri_pool_ratio),
         compression_type(_compression_type),
         compress_format_version(_compress_format_version),
-        enable_custom_split_merge(_enable_custom_split_merge) {}
+        enable_custom_split_merge(_enable_custom_split_merge),
+        do_not_compress_roles(_do_not_compress_roles) {}
 };
 
 // EXPERIMENTAL
@@ -272,7 +278,9 @@ extern std::shared_ptr<SecondaryCache> NewCompressedSecondaryCache(
         kDefaultCacheMetadataChargePolicy,
     CompressionType compression_type = CompressionType::kLZ4Compression,
     uint32_t compress_format_version = 2,
-    bool enable_custom_split_merge = false);
+    bool enable_custom_split_merge = false,
+    const CacheEntryRoleSet& _do_not_compress_roles = {
+        CacheEntryRole::kFilterBlock});
 
 extern std::shared_ptr<SecondaryCache> NewCompressedSecondaryCache(
     const CompressedSecondaryCacheOptions& opts);
@@ -352,426 +360,4 @@ extern std::shared_ptr<Cache> NewClockCache(
     CacheMetadataChargePolicy metadata_charge_policy =
         kDefaultCacheMetadataChargePolicy);
 
-// A Cache maps keys to objects resident in memory, tracks reference counts
-// on those key-object entries, and is able to remove unreferenced entries
-// whenever it wants. All operations are fully thread safe except as noted.
-// Inserted entries have a specified "charge" which is some quantity in
-// unspecified units, typically bytes of memory used. A Cache will typically
-// have a finite capacity in units of charge, and evict entries as needed
-// to stay at or below that capacity.
-//
-// NOTE: This API is for expert use only and is more intended for providing
-// custom implementations than for calling into. It is subject to change
-// as RocksDB evolves, especially the RocksDB block cache.
-//
-// INTERNAL: See typed_cache.h for convenient wrappers on top of this API.
-class Cache {
- public:  // types hidden from API client
-  // Opaque handle to an entry stored in the cache.
-  struct Handle {};
-
- public:  // types hidden from Cache implementation
-  // Pointer to cached object of unspecified type. (This type alias is
-  // provided for clarity, not really for type checking.)
-  using ObjectPtr = void*;
-
-  // Opaque object providing context (settings, etc.) to create objects
-  // for primary cache from saved (serialized) secondary cache entries.
-  struct CreateContext {};
-
- public:  // type defs
-  // Depending on implementation, cache entries with higher priority levels
-  // could be less likely to get evicted than entries with lower priority
-  // levels. The "high" priority level applies to certain SST metablocks (e.g.
-  // index and filter blocks) if the option
-  // cache_index_and_filter_blocks_with_high_priority is set. The "low" priority
-  // level is used for other kinds of SST blocks (most importantly, data
-  // blocks), as well as the above metablocks in case
-  // cache_index_and_filter_blocks_with_high_priority is
-  // not set. The "bottom" priority level is for BlobDB's blob values.
-  enum class Priority { HIGH, LOW, BOTTOM };
-
-  // A set of callbacks to allow objects in the primary block cache to be
-  // be persisted in a secondary cache. The purpose of the secondary cache
-  // is to support other ways of caching the object, such as persistent or
-  // compressed data, that may require the object to be parsed and transformed
-  // in some way. Since the primary cache holds C++ objects and the secondary
-  // cache may only hold flat data that doesn't need relocation, these
-  // callbacks need to be provided by the user of the block
-  // cache to do the conversion.
-  // The CacheItemHelper is passed to Insert() and Lookup(). It has pointers
-  // to callback functions for size, saving and deletion of the
-  // object. The callbacks are defined in C-style in order to make them
-  // stateless and not add to the cache metadata size.
-  // Saving multiple std::function objects will take up 32 bytes per
-  // function, even if its not bound to an object and does no capture.
-  //
-  // All the callbacks are C-style function pointers in order to simplify
-  // lifecycle management. Objects in the cache can outlive the parent DB,
-  // so anything required for these operations should be contained in the
-  // object itself.
-  //
-  // The SizeCallback takes a pointer to the object and returns the size
-  // of the persistable data. It can be used by the secondary cache to allocate
-  // memory if needed.
-  //
-  // RocksDB callbacks are NOT exception-safe. A callback completing with an
-  // exception can lead to undefined behavior in RocksDB, including data loss,
-  // unreported corruption, deadlocks, and more.
-  using SizeCallback = size_t (*)(ObjectPtr obj);
-
-  // The SaveToCallback takes an object pointer and saves the persistable
-  // data into a buffer. The secondary cache may decide to not store it in a
-  // contiguous buffer, in which case this callback will be called multiple
-  // times with increasing offset
-  using SaveToCallback = Status (*)(ObjectPtr from_obj, size_t from_offset,
-                                    size_t length, char* out_buf);
-
-  // A function pointer type for destruction of a cache object. This will
-  // typically call the destructor for the appropriate type of the object.
-  // The Cache is responsible for copying and reclaiming space for the key,
-  // but objects are managed in part using this callback. Generally a DeleterFn
-  // can be nullptr if the ObjectPtr does not need destruction (e.g. nullptr or
-  // pointer into static data).
-  using DeleterFn = void (*)(ObjectPtr obj, MemoryAllocator* allocator);
-
-  // The CreateCallback is takes in a buffer from the NVM cache and constructs
-  // an object using it. The callback doesn't have ownership of the buffer and
-  // should copy the contents into its own buffer. The CreateContext* is
-  // provided by Lookup and may be used to follow DB- or CF-specific settings.
-  // In case of some error, non-OK is returned and the caller should ignore
-  // any result in out_obj. (The implementation must clean up after itself.)
-  using CreateCallback = Status (*)(const Slice& data, CreateContext* context,
-                                    MemoryAllocator* allocator,
-                                    ObjectPtr* out_obj, size_t* out_charge);
-
-  // A struct with pointers to helper functions for spilling items from the
-  // cache into the secondary cache. May be extended in the future. An
-  // instance of this struct is expected to outlive the cache.
-  struct CacheItemHelper {
-    // Function for deleting an object on its removal from the Cache.
-    // nullptr is only for entries that require no destruction, such as
-    // "placeholder" cache entries with nullptr object.
-    DeleterFn del_cb;  // (<- Most performance critical)
-    // Next three are used for persisting values as described above.
-    // If any is nullptr, then all three should be nullptr and persisting the
-    // entry to/from secondary cache is not supported.
-    SizeCallback size_cb;
-    SaveToCallback saveto_cb;
-    CreateCallback create_cb;
-    // Classification of the entry for monitoring purposes in block cache.
-    CacheEntryRole role;
-
-    constexpr CacheItemHelper()
-        : del_cb(nullptr),
-          size_cb(nullptr),
-          saveto_cb(nullptr),
-          create_cb(nullptr),
-          role(CacheEntryRole::kMisc) {}
-
-    explicit constexpr CacheItemHelper(CacheEntryRole _role,
-                                       DeleterFn _del_cb = nullptr,
-                                       SizeCallback _size_cb = nullptr,
-                                       SaveToCallback _saveto_cb = nullptr,
-                                       CreateCallback _create_cb = nullptr)
-        : del_cb(_del_cb),
-          size_cb(_size_cb),
-          saveto_cb(_saveto_cb),
-          create_cb(_create_cb),
-          role(_role) {
-      // Either all three secondary cache callbacks are non-nullptr or
-      // all three are nullptr
-      assert((size_cb != nullptr) == (saveto_cb != nullptr));
-      assert((size_cb != nullptr) == (create_cb != nullptr));
-    }
-    inline bool IsSecondaryCacheCompatible() const {
-      return size_cb != nullptr;
-    }
-  };
-
- public:  // ctor/dtor/create
-  Cache(std::shared_ptr<MemoryAllocator> allocator = nullptr)
-      : memory_allocator_(std::move(allocator)) {}
-  // No copying allowed
-  Cache(const Cache&) = delete;
-  Cache& operator=(const Cache&) = delete;
-
-  // Destroys all remaining entries by calling the associated "deleter"
-  virtual ~Cache() {}
-
-  // Creates a new Cache based on the input value string and returns the result.
-  // Currently, this method can be used to create LRUCaches only
-  // @param config_options
-  // @param value  The value might be:
-  //   - an old-style cache ("1M") -- equivalent to NewLRUCache(1024*102(
-  //   - Name-value option pairs -- "capacity=1M; num_shard_bits=4;
-  //     For the LRUCache, the values are defined in LRUCacheOptions.
-  // @param result The new Cache object
-  // @return OK if the cache was successfully created
-  // @return NotFound if an invalid name was specified in the value
-  // @return InvalidArgument if either the options were not valid
-  static Status CreateFromString(const ConfigOptions& config_options,
-                                 const std::string& value,
-                                 std::shared_ptr<Cache>* result);
-
- public:  // functions
-  // The type of the Cache
-  virtual const char* Name() const = 0;
-
-  // The Insert and Lookup APIs below are intended to allow cached objects
-  // to be demoted/promoted between the primary block cache and a secondary
-  // cache. The secondary cache could be a non-volatile cache, and will
-  // likely store the object in a different representation. They rely on a
-  // per object CacheItemHelper to do the conversions.
-  // The secondary cache may persist across process and system restarts,
-  // and may even be moved between hosts. Therefore, the cache key must
-  // be repeatable across restarts/reboots, and globally unique if
-  // multiple DBs share the same cache and the set of DBs can change
-  // over time.
-
-  // Insert a mapping from key->object into the cache and assign it
-  // the specified charge against the total cache capacity. If
-  // strict_capacity_limit is true and cache reaches its full capacity,
-  // return Status::MemoryLimit. `obj` must be non-nullptr if compatible
-  // with secondary cache (helper->size_cb != nullptr), because Value() ==
-  // nullptr is reserved for indicating some secondary cache failure cases.
-  // On success, returns OK and takes ownership of `obj`, eventually deleting
-  // it with helper->del_cb. On non-OK return, the caller maintains ownership
-  // of `obj` so will often need to delete it in such cases.
-  //
-  // The helper argument is saved by the cache and will be used when the
-  // inserted object is evicted or considered for promotion to the secondary
-  // cache. Promotion to secondary cache is only enabled if helper->size_cb
-  // != nullptr. The helper must outlive the cache. Callers may use
-  // &kNoopCacheItemHelper as a trivial helper (no deleter for the object,
-  // no secondary cache). `helper` must not be nullptr (efficiency).
-  //
-  // If `handle` is not nullptr and return status is OK, `handle` is set
-  // to a Handle* for the entry. The caller must call this->Release(handle)
-  // when the returned entry is no longer needed. If `handle` is nullptr, it is
-  // as if Release is called immediately after Insert.
-  //
-  // Regardless of whether the item was inserted into the cache,
-  // it will attempt to insert it into the secondary cache if one is
-  // configured, and the helper supports it.
-  // The cache implementation must support a secondary cache, otherwise
-  // the item is only inserted into the primary cache. It may
-  // defer the insertion to the secondary cache as it sees fit.
-  //
-  // When the inserted entry is no longer needed, it will be destroyed using
-  // helper->del_cb (if non-nullptr).
-  virtual Status Insert(const Slice& key, ObjectPtr obj,
-                        const CacheItemHelper* helper, size_t charge,
-                        Handle** handle = nullptr,
-                        Priority priority = Priority::LOW) = 0;
-
-  // Lookup the key, returning nullptr if not found. If found, returns
-  // a handle to the mapping that must eventually be passed to Release().
-  //
-  // If a non-nullptr helper argument is provided with a non-nullptr
-  // create_cb, and a secondary cache is configured, then the secondary
-  // cache is also queried if lookup in the primary cache fails. If found
-  // in secondary cache, the provided create_db and create_context are
-  // used to promote the entry to an object in the primary cache.
-  // In that case, the helper may be saved and used later when the object
-  // is evicted, so as usual, the pointed-to helper must outlive the cache.
-  //
-  // ======================== Async Lookup (wait=false) ======================
-  // When wait=false, the handle returned might be in any of three states:
-  // * Present - If Value() != nullptr, then the result is present and
-  // the handle can be used just as if wait=true.
-  // * Pending, not ready (IsReady() == false) - secondary cache is still
-  // working to retrieve the value. Might become ready any time.
-  // * Pending, ready (IsReady() == true) - secondary cache has the value
-  // but it has not been loaded as an object into primary cache. Call to
-  // Wait()/WaitAll() will not block.
-  //
-  // IMPORTANT: Pending handles are not thread-safe, and only these functions
-  // are allowed on them: Value(), IsReady(), Wait(), WaitAll(). Even Release()
-  // can only come after Wait() or WaitAll() even though a reference is held.
-  //
-  // Only Wait()/WaitAll() gets a Handle out of a Pending state. (Waiting is
-  // safe and has no effect on other handle states.) After waiting on a Handle,
-  // it is in one of two states:
-  // * Present - if Value() != nullptr
-  // * Failed - if Value() == nullptr, such as if the secondary cache
-  // initially thought it had the value but actually did not.
-  //
-  // Note that given an arbitrary Handle, the only way to distinguish the
-  // Pending+ready state from the Failed state is to Wait() on it. A cache
-  // entry not compatible with secondary cache can also have Value()==nullptr
-  // like the Failed state, but this is not generally a concern.
-  virtual Handle* Lookup(const Slice& key,
-                         const CacheItemHelper* helper = nullptr,
-                         CreateContext* create_context = nullptr,
-                         Priority priority = Priority::LOW, bool wait = true,
-                         Statistics* stats = nullptr) = 0;
-
-  // Convenience wrapper when secondary cache not supported
-  inline Handle* BasicLookup(const Slice& key, Statistics* stats) {
-    return Lookup(key, nullptr, nullptr, Priority::LOW, true, stats);
-  }
-
-  // Increments the reference count for the handle if it refers to an entry in
-  // the cache. Returns true if refcount was incremented; otherwise, returns
-  // false.
-  // REQUIRES: handle must have been returned by a method on *this.
-  virtual bool Ref(Handle* handle) = 0;
-
-  /**
-   * Release a mapping returned by a previous Lookup(). A released entry might
-   * still remain in cache in case it is later looked up by others. If
-   * erase_if_last_ref is set then it also erases it from the cache if there is
-   * no other reference to  it. Erasing it should call the deleter function that
-   * was provided when the entry was inserted.
-   *
-   * Returns true if the entry was also erased.
-   */
-  // REQUIRES: handle must not have been released yet.
-  // REQUIRES: handle must have been returned by a method on *this.
-  virtual bool Release(Handle* handle, bool erase_if_last_ref = false) = 0;
-
-  // Return the object assiciated with a handle returned by a successful
-  // Lookup(). For historical reasons, this is also known at the "value"
-  // associated with the key.
-  // REQUIRES: handle must not have been released yet.
-  // REQUIRES: handle must have been returned by a method on *this.
-  virtual ObjectPtr Value(Handle* handle) = 0;
-
-  // If the cache contains the entry for the key, erase it.  Note that the
-  // underlying entry will be kept around until all existing handles
-  // to it have been released.
-  virtual void Erase(const Slice& key) = 0;
-  // Return a new numeric id.  May be used by multiple clients who are
-  // sharding the same cache to partition the key space.  Typically the
-  // client will allocate a new id at startup and prepend the id to
-  // its cache keys.
-  virtual uint64_t NewId() = 0;
-
-  // sets the maximum configured capacity of the cache. When the new
-  // capacity is less than the old capacity and the existing usage is
-  // greater than new capacity, the implementation will do its best job to
-  // purge the released entries from the cache in order to lower the usage
-  virtual void SetCapacity(size_t capacity) = 0;
-
-  // Set whether to return error on insertion when cache reaches its full
-  // capacity.
-  virtual void SetStrictCapacityLimit(bool strict_capacity_limit) = 0;
-
-  // Get the flag whether to return error on insertion when cache reaches its
-  // full capacity.
-  virtual bool HasStrictCapacityLimit() const = 0;
-
-  // Returns the maximum configured capacity of the cache
-  virtual size_t GetCapacity() const = 0;
-
-  // Returns the memory size for the entries residing in the cache.
-  virtual size_t GetUsage() const = 0;
-
-  // Returns the number of entries currently tracked in the table. SIZE_MAX
-  // means "not supported." This is used for inspecting the load factor, along
-  // with GetTableAddressCount().
-  virtual size_t GetOccupancyCount() const { return SIZE_MAX; }
-
-  // Returns the number of ways the hash function is divided for addressing
-  // entries. Zero means "not supported." This is used for inspecting the load
-  // factor, along with GetOccupancyCount().
-  virtual size_t GetTableAddressCount() const { return 0; }
-
-  // Returns the memory size for a specific entry in the cache.
-  virtual size_t GetUsage(Handle* handle) const = 0;
-
-  // Returns the memory size for the entries in use by the system
-  virtual size_t GetPinnedUsage() const = 0;
-
-  // Returns the charge for the specific entry in the cache.
-  virtual size_t GetCharge(Handle* handle) const = 0;
-
-  // Returns the helper for the specified entry.
-  virtual const CacheItemHelper* GetCacheItemHelper(Handle* handle) const = 0;
-
-  // Call this on shutdown if you want to speed it up. Cache will disown
-  // any underlying data and will not free it on delete. This call will leak
-  // memory - call this only if you're shutting down the process.
-  // Any attempts of using cache after this call will fail terribly.
-  // Always delete the DB object before calling this method!
-  virtual void DisownData() {
-    // default implementation is noop
-  }
-
-  struct ApplyToAllEntriesOptions {
-    // If the Cache uses locks, setting `average_entries_per_lock` to
-    // a higher value suggests iterating over more entries each time a lock
-    // is acquired, likely reducing the time for ApplyToAllEntries but
-    // increasing latency for concurrent users of the Cache. Setting
-    // `average_entries_per_lock` to a smaller value could be helpful if
-    // callback is relatively expensive, such as using large data structures.
-    size_t average_entries_per_lock = 256;
-  };
-
-  // Apply a callback to all entries in the cache. The Cache must ensure
-  // thread safety but does not guarantee that a consistent snapshot of all
-  // entries is iterated over if other threads are operating on the Cache
-  // also.
-  virtual void ApplyToAllEntries(
-      const std::function<void(const Slice& key, ObjectPtr obj, size_t charge,
-                               const CacheItemHelper* helper)>& callback,
-      const ApplyToAllEntriesOptions& opts) = 0;
-
-  // Remove all entries.
-  // Prerequisite: no entry is referenced.
-  virtual void EraseUnRefEntries() = 0;
-
-  virtual std::string GetPrintableOptions() const { return ""; }
-
-  // Check for any warnings or errors in the operation of the cache and
-  // report them to the logger. This is intended only to be called
-  // periodically so does not need to be very efficient. (Obscure calling
-  // conventions for Logger inherited from env.h)
-  virtual void ReportProblems(
-      const std::shared_ptr<Logger>& /*info_log*/) const {}
-
-  MemoryAllocator* memory_allocator() const { return memory_allocator_.get(); }
-
-  // EXPERIMENTAL
-  // The following APIs are experimental and might change in the future.
-
-  // Release a mapping returned by a previous Lookup(). The "useful"
-  // parameter specifies whether the data was actually used or not,
-  // which may be used by the cache implementation to decide whether
-  // to consider it as a hit for retention purposes. As noted elsewhere,
-  // "pending" handles require Wait()/WaitAll() before Release().
-  virtual bool Release(Handle* handle, bool /*useful*/,
-                       bool erase_if_last_ref) {
-    return Release(handle, erase_if_last_ref);
-  }
-
-  // Determines if the handle returned by Lookup() can give a value without
-  // blocking, though Wait()/WaitAll() might be required to publish it to
-  // Value(). See secondary cache compatible Lookup() above for details.
-  // This call is not thread safe on "pending" handles.
-  virtual bool IsReady(Handle* /*handle*/) { return true; }
-
-  // Convert a "pending" handle into a full thread-shareable handle by
-  // * If necessary, wait until secondary cache finishes loading the value.
-  // * Construct the object for primary cache and set it in the handle.
-  // Even after Wait() on a pending handle, the caller must check for
-  // Value() == nullptr in case of failure. This call is not thread-safe
-  // on pending handles. This call has no effect on non-pending handles.
-  // See secondary cache compatible Lookup() above for details.
-  virtual void Wait(Handle* /*handle*/) {}
-
-  // Wait for a vector of handles to become ready. As with Wait(), the user
-  // should check the Value() of each handle for nullptr. This call is not
-  // thread-safe on pending handles.
-  virtual void WaitAll(std::vector<Handle*>& /*handles*/) {}
-
- private:
-  std::shared_ptr<MemoryAllocator> memory_allocator_;
-};
-
-// Useful for cache entries requiring no clean-up, such as for cache
-// reservations
-inline constexpr Cache::CacheItemHelper kNoopCacheItemHelper{};
-
 }  // namespace ROCKSDB_NAMESPACE

include/rocksdb/compaction_filter.h

@@ -11,11 +11,13 @@
 #include <cassert>
 #include <memory>
 #include <string>
+#include <utility>
 #include <vector>
 
 #include "rocksdb/customizable.h"
 #include "rocksdb/rocksdb_namespace.h"
 #include "rocksdb/types.h"
+#include "rocksdb/wide_columns.h"
 
 namespace ROCKSDB_NAMESPACE {
 
@@ -34,6 +36,7 @@ class CompactionFilter : public Customizable {
     kValue,
     kMergeOperand,
     kBlobIndex,  // used internally by BlobDB.
+    kWideColumnEntity,
   };
 
   enum class Decision {
@@ -44,6 +47,7 @@ class CompactionFilter : public Customizable {
     kChangeBlobIndex,  // used internally by BlobDB.
     kIOError,          // used internally by BlobDB.
     kPurge,            // used for keys that can only be SingleDelete'ed
+    kChangeWideColumnEntity,
     kUndetermined,
   };
 
@@ -176,15 +180,57 @@ class CompactionFilter : public Customizable {
         }
         return value_changed ? Decision::kChangeValue : Decision::kKeep;
       }
+
       case ValueType::kMergeOperand: {
         bool rv = FilterMergeOperand(level, key, existing_value);
         return rv ? Decision::kRemove : Decision::kKeep;
       }
+
       case ValueType::kBlobIndex:
         return Decision::kKeep;
+
+      default:
+        assert(false);
+        return Decision::kKeep;
     }
-    assert(false);
-    return Decision::kKeep;
+  }
+
+  // Wide column aware API. Called for plain values, merge operands, and
+  // wide-column entities; the `value_type` parameter indicates the type of the
+  // key-value. When the key-value is a plain value or a merge operand, the
+  // `existing_value` parameter contains the existing value and the
+  // `existing_columns` parameter is invalid (nullptr). When the key-value is a
+  // wide-column entity, the `existing_columns` parameter contains the wide
+  // columns of the existing entity and the `existing_value` parameter is
+  // invalid (nullptr). The output parameters `new_value` and `new_columns` can
+  // be used to change the value or wide columns of the key-value when
+  // `kChangeValue` or `kChangeWideColumnEntity` is returned. See above for more
+  // information on the semantics of the potential return values.
+  //
+  // For compatibility, the default implementation keeps all wide-column
+  // entities, and falls back to FilterV2 for plain values and merge operands.
+  // If you override this method, there is no need to override FilterV2 (or
+  // Filter/FilterMergeOperand).
+  virtual Decision FilterV3(
+      int level, const Slice& key, ValueType value_type,
+      const Slice* existing_value, const WideColumns* existing_columns,
+      std::string* new_value,
+      std::vector<std::pair<std::string, std::string>>* /* new_columns */,
+      std::string* skip_until) const {
+#ifdef NDEBUG
+    (void)existing_columns;
+#endif
+
+    assert(!existing_value || !existing_columns);
+    assert(value_type == ValueType::kWideColumnEntity || existing_value);
+    assert(value_type != ValueType::kWideColumnEntity || existing_columns);
+
+    if (value_type == ValueType::kWideColumnEntity) {
+      return Decision::kKeep;
+    }
+
+    return FilterV2(level, key, value_type, *existing_value, new_value,
+                    skip_until);
   }
 
   // Internal (BlobDB) use only. Do not override in application code.
@@ -211,7 +257,7 @@ class CompactionFilter : public Customizable {
   // In the case of BlobDB, it may be possible to reach a decision with only
   // the key without reading the actual value. Keys whose value_type is
   // kBlobIndex will be checked by this method.
-  // Returning kUndetermined will cause FilterV2() to be called to make a
+  // Returning kUndetermined will cause FilterV3() to be called to make a
   // decision as usual.
   virtual Decision FilterBlobByKey(int /*level*/, const Slice& /*key*/,
                                    std::string* /*new_value*/,

include/rocksdb/convenience.h

@@ -160,9 +160,10 @@ struct ConfigOptions {
 //     "kCompactionStyleNone".
 //
 
-// Take a default ColumnFamilyOptions "base_options" in addition to a
-// map "opts_map" of option name to option value to construct the new
-// ColumnFamilyOptions "new_options".
+// Take a ConfigOptions `config_options` and a ColumnFamilyOptions
+// "base_options" as the default option in addition to a map "opts_map" of
+// option name to option value to construct the new ColumnFamilyOptions
+// "new_options".
 //
 // Below are the instructions of how to config some non-primitive-typed
 // options in ColumnFamilyOptions:
@@ -238,11 +239,6 @@ struct ConfigOptions {
 //        cf_opt.compression_opts.strategy = 6;
 //        cf_opt.compression_opts.max_dict_bytes = 7;
 //
-// The GetColumnFamilyOptionsFromMap(ConfigOptions, ...) should be used; the
-// alternative signature may be deprecated in a future release.  The equivalent
-// functionality can be achieved by setting the corresponding options in
-// the ConfigOptions parameter.
-//
 // @param config_options controls how the map is processed.
 // @param base_options the default options of the output "new_options".
 // @param opts_map an option name to value map for specifying how "new_options"
@@ -267,15 +263,10 @@ Status GetColumnFamilyOptionsFromMap(
     const ColumnFamilyOptions& base_options,
     const std::unordered_map<std::string, std::string>& opts_map,
     ColumnFamilyOptions* new_options);
-Status GetColumnFamilyOptionsFromMap(
-    const ColumnFamilyOptions& base_options,
-    const std::unordered_map<std::string, std::string>& opts_map,
-    ColumnFamilyOptions* new_options, bool input_strings_escaped = false,
-    bool ignore_unknown_options = false);
 
-// Take a default DBOptions "base_options" in addition to a
-// map "opts_map" of option name to option value to construct the new
-// DBOptions "new_options".
+// Take a ConfigOptions `config_options` and a DBOptions "base_options" as the
+// default option in addition to a map "opts_map" of option name to option value
+// to construct the new DBOptions "new_options".
 //
 // Below are the instructions of how to config some non-primitive-typed
 // options in DBOptions:
@@ -286,11 +277,6 @@ Status GetColumnFamilyOptionsFromMap(
 //   - Passing {"rate_limiter_bytes_per_sec", "1024"} is equivalent to
 //     passing NewGenericRateLimiter(1024) to rate_limiter_bytes_per_sec.
 //
-// The GetDBOptionsFromMap(ConfigOptions, ...) should be used; the
-// alternative signature may be deprecated in a future release. The equivalent
-// functionality can be achieved by setting the corresponding options in
-// the ConfigOptions parameter.
-//
 // @param config_options controls how the map is processed.
 // @param base_options the default options of the output "new_options".
 // @param opts_map an option name to value map for specifying how "new_options"
@@ -314,15 +300,11 @@ Status GetDBOptionsFromMap(
     const ConfigOptions& cfg_options, const DBOptions& base_options,
     const std::unordered_map<std::string, std::string>& opts_map,
     DBOptions* new_options);
-Status GetDBOptionsFromMap(
-    const DBOptions& base_options,
-    const std::unordered_map<std::string, std::string>& opts_map,
-    DBOptions* new_options, bool input_strings_escaped = false,
-    bool ignore_unknown_options = false);
 
-// Take a default BlockBasedTableOptions "table_options" in addition to a
-// map "opts_map" of option name to option value to construct the new
-// BlockBasedTableOptions "new_table_options".
+// Take a ConfigOptions `config_options` and a BlockBasedTableOptions
+// "table_options" as the default option in addition to a map "opts_map" of
+// option name to option value to construct the new BlockBasedTableOptions
+// "new_table_options".
 //
 // Below are the instructions of how to config some non-primitive-typed
 // options in BlockBasedTableOptions:
@@ -345,11 +327,6 @@ Status GetDBOptionsFromMap(
 //   - Passing {"block_cache", "1M"} in GetBlockBasedTableOptionsFromMap is
 //     equivalent to setting block_cache using NewLRUCache(1024 * 1024).
 //
-// The GetBlockBasedTableOptionsFromMap(ConfigOptions, ...) should be used;
-// the alternative signature may be deprecated in a future release. The
-// equivalent functionality can be achieved by setting the corresponding
-// options in the ConfigOptions parameter.
-//
 // @param config_options controls how the map is processed.
 // @param table_options the default options of the output "new_table_options".
 // @param opts_map an option name to value map for specifying how
@@ -369,20 +346,11 @@ Status GetBlockBasedTableOptionsFromMap(
     const BlockBasedTableOptions& table_options,
     const std::unordered_map<std::string, std::string>& opts_map,
     BlockBasedTableOptions* new_table_options);
-Status GetBlockBasedTableOptionsFromMap(
-    const BlockBasedTableOptions& table_options,
-    const std::unordered_map<std::string, std::string>& opts_map,
-    BlockBasedTableOptions* new_table_options,
-    bool input_strings_escaped = false, bool ignore_unknown_options = false);
 
-// Take a default PlainTableOptions "table_options" in addition to a
-// map "opts_map" of option name to option value to construct the new
-// PlainTableOptions "new_table_options".
-//
-// The GetPlainTableOptionsFromMap(ConfigOptions, ...) should be used; the
-// alternative signature may be deprecated in a future release. The equivalent
-// functionality can be achieved by setting the corresponding options in
-// the ConfigOptions parameter.
+// Take a ConfigOptions `config_options` and a default PlainTableOptions
+// "table_options" as the default option in addition to a map "opts_map" of
+// option name to option value to construct the new PlainTableOptions
+// "new_table_options".
 //
 // @param config_options controls how the map is processed.
 // @param table_options the default options of the output "new_table_options".
@@ -402,43 +370,26 @@ Status GetPlainTableOptionsFromMap(
     const ConfigOptions& config_options, const PlainTableOptions& table_options,
     const std::unordered_map<std::string, std::string>& opts_map,
     PlainTableOptions* new_table_options);
-Status GetPlainTableOptionsFromMap(
-    const PlainTableOptions& table_options,
-    const std::unordered_map<std::string, std::string>& opts_map,
-    PlainTableOptions* new_table_options, bool input_strings_escaped = false,
-    bool ignore_unknown_options = false);
 
-// Take a string representation of option names and values, apply them into the
-// base_options, and return the new options as a result. The string has the
-// following format:
+// Take a ConfigOptions `config_options`, a string representation of option
+// names and values, apply them into the base_options, and return the new
+// options as a result. The string has the following format:
 //   "write_buffer_size=1024;max_write_buffer_number=2"
 // Nested options config is also possible. For example, you can define
 // BlockBasedTableOptions as part of the string for block-based table factory:
 //   "write_buffer_size=1024;block_based_table_factory={block_size=4k};"
 //   "max_write_buffer_num=2"
 //
-//
-// The GetColumnFamilyOptionsFromString(ConfigOptions, ...) should be used; the
-// alternative signature may be deprecated in a future release. The equivalent
-// functionality can be achieved by setting the corresponding options in
-// the ConfigOptions parameter.
 Status GetColumnFamilyOptionsFromString(const ConfigOptions& config_options,
                                         const ColumnFamilyOptions& base_options,
                                         const std::string& opts_str,
                                         ColumnFamilyOptions* new_options);
-Status GetColumnFamilyOptionsFromString(const ColumnFamilyOptions& base_options,
-                                        const std::string& opts_str,
-                                        ColumnFamilyOptions* new_options);
 
 Status GetDBOptionsFromString(const ConfigOptions& config_options,
                               const DBOptions& base_options,
                               const std::string& opts_str,
                               DBOptions* new_options);
 
-Status GetDBOptionsFromString(const DBOptions& base_options,
-                              const std::string& opts_str,
-                              DBOptions* new_options);
-
 Status GetStringFromDBOptions(const ConfigOptions& config_options,
                               const DBOptions& db_options,
                               std::string* opts_str);
@@ -458,17 +409,11 @@ Status GetStringFromCompressionType(std::string* compression_str,
 
 std::vector<CompressionType> GetSupportedCompressions();
 
-Status GetBlockBasedTableOptionsFromString(
-    const BlockBasedTableOptions& table_options, const std::string& opts_str,
-    BlockBasedTableOptions* new_table_options);
 Status GetBlockBasedTableOptionsFromString(
     const ConfigOptions& config_options,
     const BlockBasedTableOptions& table_options, const std::string& opts_str,
     BlockBasedTableOptions* new_table_options);
 
-Status GetPlainTableOptionsFromString(const PlainTableOptions& table_options,
-                                      const std::string& opts_str,
-                                      PlainTableOptions* new_table_options);
 Status GetPlainTableOptionsFromString(const ConfigOptions& config_options,
                                       const PlainTableOptions& table_options,
                                       const std::string& opts_str,

include/rocksdb/data_structure.h

@@ -15,37 +15,172 @@
 
 namespace ROCKSDB_NAMESPACE {
 
-// This is a data structure specifically designed as a "Set" for a
-// pretty small scale of Enum structure. For now, it can support up
-// to 64 element, and it is expandable in the future.
-template <typename ENUM_TYPE, ENUM_TYPE MAX_VALUE>
+namespace detail {
+int CountTrailingZeroBitsForSmallEnumSet(uint64_t);
+}  // namespace detail
+
+// Represents a set of values of some enum type with a small number of
+// possible enumerators. For now, it supports enums where no enumerator
+// exceeds 63 when converted to int.
+template <typename ENUM_TYPE, ENUM_TYPE MAX_ENUMERATOR>
 class SmallEnumSet {
+ private:
+  using StateT = uint64_t;
+  static constexpr int kStateBits = sizeof(StateT) * 8;
+  static constexpr int kMaxMax = kStateBits - 1;
+  static constexpr int kMaxValue = static_cast<int>(MAX_ENUMERATOR);
+  static_assert(kMaxValue >= 0);
+  static_assert(kMaxValue <= kMaxMax);
+
  public:
+  // construct / create
   SmallEnumSet() : state_(0) {}
 
-  ~SmallEnumSet() {}
-
-  // Return true if the input enum is included in the "Set" (i.e., changes the
-  // internal scalar state successfully), otherwise, it will return false.
-  bool Add(const ENUM_TYPE value) {
-    static_assert(MAX_VALUE <= 63, "Size currently limited to 64");
-    assert(value >= 0 && value <= MAX_VALUE);
-    uint64_t old_state = state_;
-    uint64_t tmp = 1;
-    state_ |= (tmp << value);
-    return old_state != state_;
+  template <class... TRest>
+  /*implicit*/ constexpr SmallEnumSet(const ENUM_TYPE e, TRest... rest) {
+    *this = SmallEnumSet(rest...).With(e);
   }
 
+  // Return the set that includes all valid values, assuming the enum
+  // is "dense" (includes all values converting to 0 through kMaxValue)
+  static constexpr SmallEnumSet All() {
+    StateT tmp = StateT{1} << kMaxValue;
+    return SmallEnumSet(RawStateMarker(), tmp | (tmp - 1));
+  }
+
+  // equality
+  bool operator==(const SmallEnumSet& that) const {
+    return this->state_ == that.state_;
+  }
+  bool operator!=(const SmallEnumSet& that) const { return !(*this == that); }
+
+  // query
+
   // Return true if the input enum is contained in the "Set".
-  bool Contains(const ENUM_TYPE value) {
-    static_assert(MAX_VALUE <= 63, "Size currently limited to 64");
-    assert(value >= 0 && value <= MAX_VALUE);
-    uint64_t tmp = 1;
+  bool Contains(const ENUM_TYPE e) const {
+    int value = static_cast<int>(e);
+    assert(value >= 0 && value <= kMaxValue);
+    StateT tmp = 1;
     return state_ & (tmp << value);
   }
 
+  bool empty() const { return state_ == 0; }
+
+  // iterator
+  class const_iterator {
+   public:
+    // copy
+    const_iterator(const const_iterator& that) = default;
+    const_iterator& operator=(const const_iterator& that) = default;
+
+    // move
+    const_iterator(const_iterator&& that) noexcept = default;
+    const_iterator& operator=(const_iterator&& that) noexcept = default;
+
+    // equality
+    bool operator==(const const_iterator& that) const {
+      assert(set_ == that.set_);
+      return this->pos_ == that.pos_;
+    }
+
+    bool operator!=(const const_iterator& that) const {
+      return !(*this == that);
+    }
+
+    // ++iterator
+    const_iterator& operator++() {
+      if (pos_ < kMaxValue) {
+        pos_ = set_->SkipUnset(pos_ + 1);
+      } else {
+        pos_ = kStateBits;
+      }
+      return *this;
+    }
+
+    // iterator++
+    const_iterator operator++(int) {
+      auto old = *this;
+      ++*this;
+      return old;
+    }
+
+    ENUM_TYPE operator*() const {
+      assert(pos_ <= kMaxValue);
+      return static_cast<ENUM_TYPE>(pos_);
+    }
+
+   private:
+    friend class SmallEnumSet;
+    const_iterator(const SmallEnumSet* set, int pos) : set_(set), pos_(pos) {}
+    const SmallEnumSet* set_;
+    int pos_;
+  };
+
+  const_iterator begin() const { return const_iterator(this, SkipUnset(0)); }
+
+  const_iterator end() const { return const_iterator(this, kStateBits); }
+
+  // mutable ops
+
+  // Modifies the set (if needed) to include the given value. Returns true
+  // iff the set was modified.
+  bool Add(const ENUM_TYPE e) {
+    int value = static_cast<int>(e);
+    assert(value >= 0 && value <= kMaxValue);
+    StateT old_state = state_;
+    state_ |= (StateT{1} << value);
+    return old_state != state_;
+  }
+
+  // Modifies the set (if needed) not to include the given value. Returns true
+  // iff the set was modified.
+  bool Remove(const ENUM_TYPE e) {
+    int value = static_cast<int>(e);
+    assert(value >= 0 && value <= kMaxValue);
+    StateT old_state = state_;
+    state_ &= ~(StateT{1} << value);
+    return old_state != state_;
+  }
+
+  // applicative ops
+
+  // Return a new set based on this one with the additional value(s) inserted
+  constexpr SmallEnumSet With(const ENUM_TYPE e) const {
+    int value = static_cast<int>(e);
+    assert(value >= 0 && value <= kMaxValue);
+    return SmallEnumSet(RawStateMarker(), state_ | (StateT{1} << value));
+  }
+  template <class... TRest>
+  constexpr SmallEnumSet With(const ENUM_TYPE e1, const ENUM_TYPE e2,
+                              TRest... rest) const {
+    return With(e1).With(e2, rest...);
+  }
+
+  // Return a new set based on this one excluding the given value(s)
+  constexpr SmallEnumSet Without(const ENUM_TYPE e) const {
+    int value = static_cast<int>(e);
+    assert(value >= 0 && value <= kMaxValue);
+    return SmallEnumSet(RawStateMarker(), state_ & ~(StateT{1} << value));
+  }
+  template <class... TRest>
+  constexpr SmallEnumSet Without(const ENUM_TYPE e1, const ENUM_TYPE e2,
+                                 TRest... rest) const {
+    return Without(e1).Without(e2, rest...);
+  }
+
  private:
-  uint64_t state_;
+  int SkipUnset(int pos) const {
+    StateT tmp = state_ >> pos;
+    if (tmp == 0) {
+      return kStateBits;
+    } else {
+      return pos + detail::CountTrailingZeroBitsForSmallEnumSet(tmp);
+    }
+  }
+  struct RawStateMarker {};
+  explicit SmallEnumSet(RawStateMarker, StateT state) : state_(state) {}
+
+  StateT state_;
 };
 
 }  // namespace ROCKSDB_NAMESPACE

include/rocksdb/db.h

@@ -769,6 +769,68 @@ class DB {
     }
   }
 
+  // Batched MultiGet-like API that returns wide-column entities from a single
+  // column family. For any given "key[i]" in "keys" (where 0 <= "i" <
+  // "num_keys"), if the column family specified by "column_family" contains an
+  // entry, it is returned it as a wide-column entity in "results[i]". If the
+  // entry is a wide-column entity, it is returned as-is; if it is a plain
+  // key-value, it is returned as an entity with a single anonymous column (see
+  // kDefaultWideColumnName) which contains the value.
+  //
+  // "statuses[i]" is set to OK if "keys[i]" is successfully retrieved. It is
+  // set to NotFound and an empty wide-column entity is returned in "results[i]"
+  // if there is no entry for "keys[i]". Finally, "statuses[i]" is set to some
+  // other non-OK status on error.
+  //
+  // If "keys" are sorted according to the column family's comparator, the
+  // "sorted_input" flag can be set for a small performance improvement.
+  //
+  // Note that it is the caller's responsibility to ensure that "keys",
+  // "results", and "statuses" point to "num_keys" number of contiguous objects
+  // (Slices, PinnableWideColumns, and Statuses respectively).
+  virtual void MultiGetEntity(const ReadOptions& /* options */,
+                              ColumnFamilyHandle* /* column_family */,
+                              size_t num_keys, const Slice* /* keys */,
+                              PinnableWideColumns* /* results */,
+                              Status* statuses,
+                              bool /* sorted_input */ = false) {
+    for (size_t i = 0; i < num_keys; ++i) {
+      statuses[i] = Status::NotSupported("MultiGetEntity not supported");
+    }
+  }
+
+  // Batched MultiGet-like API that returns wide-column entities potentially
+  // from multiple column families. For any given "key[i]" in "keys" (where 0 <=
+  // "i" < "num_keys"), if the column family specified by "column_families[i]"
+  // contains an entry, it is returned it as a wide-column entity in
+  // "results[i]". If the entry is a wide-column entity, it is returned as-is;
+  // if it is a plain key-value, it is returned as an entity with a single
+  // anonymous column (see kDefaultWideColumnName) which contains the value.
+  //
+  // "statuses[i]" is set to OK if "keys[i]" is successfully retrieved. It is
+  // set to NotFound and an empty wide-column entity is returned in "results[i]"
+  // if there is no entry for "keys[i]". Finally, "statuses[i]" is set to some
+  // other non-OK status on error.
+  //
+  // If "keys" are sorted by column family id and within each column family,
+  // according to the column family's comparator, the "sorted_input" flag can be
+  // set for a small performance improvement.
+  //
+  // Note that it is the caller's responsibility to ensure that
+  // "column_families", "keys", "results", and "statuses" point to "num_keys"
+  // number of contiguous objects (ColumnFamilyHandle pointers, Slices,
+  // PinnableWideColumns, and Statuses respectively).
+  virtual void MultiGetEntity(const ReadOptions& /* options */, size_t num_keys,
+                              ColumnFamilyHandle** /* column_families */,
+                              const Slice* /* keys */,
+                              PinnableWideColumns* /* results */,
+                              Status* statuses,
+                              bool /* sorted_input */ = false) {
+    for (size_t i = 0; i < num_keys; ++i) {
+      statuses[i] = Status::NotSupported("MultiGetEntity not supported");
+    }
+  }
+
   // If the key definitely does not exist in the database, then this method
   // returns false, else true. If the caller wants to obtain value when the key
   // is found in memory, a bool for 'value_found' must be passed. 'value_found'

include/rocksdb/env.h

@@ -27,6 +27,7 @@
 
 #include "rocksdb/customizable.h"
 #include "rocksdb/functor_wrapper.h"
+#include "rocksdb/port_defs.h"
 #include "rocksdb/status.h"
 #include "rocksdb/thread_status.h"
 
@@ -69,13 +70,6 @@ struct ConfigOptions;
 
 const size_t kDefaultPageSize = 4 * 1024;
 
-enum class CpuPriority {
-  kIdle = 0,
-  kLow = 1,
-  kNormal = 2,
-  kHigh = 3,
-};
-
 // Options while opening a file to read/write
 struct EnvOptions {
   // Construct with default Options

include/rocksdb/options.h

@@ -1923,7 +1923,8 @@ struct IngestExternalFileOptions {
   bool snapshot_consistency = true;
   // If set to false, IngestExternalFile() will fail if the file key range
   // overlaps with existing keys or tombstones or output of ongoing compaction
-  // during file ingestion in the DB.
+  // during file ingestion in the DB (the conditions under which a global_seqno
+  // must be assigned to the ingested file).
   bool allow_global_seqno = true;
   // If set to false and the file key range overlaps with the memtable key range
   // (memtable flush required), IngestExternalFile will fail.
@@ -1936,18 +1937,14 @@ struct IngestExternalFileOptions {
   // with allow_ingest_behind=true since the dawn of time.
   // All files will be ingested at the bottommost level with seqno=0.
   bool ingest_behind = false;
-  // Set to true if you would like to write global_seqno to a given offset in
-  // the external SST file for backward compatibility. Older versions of
-  // RocksDB writes a global_seqno to a given offset within ingested SST files,
-  // and new versions of RocksDB do not. If you ingest an external SST using
-  // new version of RocksDB and would like to be able to downgrade to an
-  // older version of RocksDB, you should set 'write_global_seqno' to true. If
-  // your service is just starting to use the new RocksDB, we recommend that
-  // you set this option to false, which brings two benefits:
-  // 1. No extra random write for global_seqno during ingestion.
-  // 2. Without writing external SST file, it's possible to do checksum.
-  // We have a plan to set this option to false by default in the future.
-  bool write_global_seqno = true;
+  // DEPRECATED - Set to true if you would like to write global_seqno to
+  // the external SST file on ingestion for backward compatibility before
+  // RocksDB 5.16.0. Such old versions of RocksDB expect any global_seqno to
+  // be written to the SST file rather than recorded in the DB manifest.
+  // This functionality was deprecated because (a) random writes might be
+  // costly or unsupported on some FileSystems, and (b) the file checksum
+  // changes with such a write.
+  bool write_global_seqno = false;
   // Set to true if you would like to verify the checksums of each block of the
   // external SST file before ingestion.
   // Warning: setting this to true causes slowdown in file ingestion because

include/rocksdb/port_defs.h

@@ -0,0 +1,22 @@
+//  Copyright (c) Meta Platforms, Inc. and affiliates.
+//  This source code is licensed under both the GPLv2 (found in the
+//  COPYING file in the root directory) and Apache 2.0 License
+//  (found in the LICENSE.Apache file in the root directory).
+//
+// This file includes the common definitions used in the port/,
+// the public API (this directory), and other directories
+
+#pragma once
+
+#include "rocksdb/rocksdb_namespace.h"
+
+namespace ROCKSDB_NAMESPACE {
+
+enum class CpuPriority {
+  kIdle = 0,
+  kLow = 1,
+  kNormal = 2,
+  kHigh = 3,
+};
+
+}  // namespace ROCKSDB_NAMESPACE

include/rocksdb/secondary_cache.h

@@ -9,7 +9,7 @@
 #include <memory>
 #include <string>
 
-#include "rocksdb/cache.h"
+#include "rocksdb/advanced_cache.h"
 #include "rocksdb/customizable.h"
 #include "rocksdb/slice.h"
 #include "rocksdb/statistics.h"

include/rocksdb/status.h

@@ -113,6 +113,7 @@ class Status {
     kOverwritten = 12,
     kTxnNotPrepared = 13,
     kIOFenced = 14,
+    kMergeOperatorFailed = 15,
     kMaxSubCode
   };
 

include/rocksdb/utilities/customizable_util.h

@@ -13,7 +13,6 @@
 // for more information on how to develop and use customizable objects
 
 #pragma once
-#include <functional>
 #include <memory>
 #include <unordered_map>
 
@@ -24,24 +23,6 @@
 #include "rocksdb/utilities/object_registry.h"
 
 namespace ROCKSDB_NAMESPACE {
-// The FactoryFunc functions are used to create a new customizable object
-// without going through the ObjectRegistry.  This methodology is especially
-// useful in LITE mode, where there is no ObjectRegistry.  The methods take
-// in an ID of the object to create and a pointer to store the created object.
-// If the factory successfully recognized the input ID, the method should return
-// success; otherwise false should be returned.  On success, the object
-// parameter contains the new object.
-template <typename T>
-using SharedFactoryFunc =
-    std::function<bool(const std::string&, std::shared_ptr<T>*)>;
-
-template <typename T>
-using UniqueFactoryFunc =
-    std::function<bool(const std::string&, std::unique_ptr<T>*)>;
-
-template <typename T>
-using StaticFactoryFunc = std::function<bool(const std::string&, T**)>;
-
 // Creates a new shared customizable instance object based on the
 // input parameters using the object registry.
 //
@@ -156,12 +137,10 @@ static Status NewManagedObject(
 // handled
 // @param value Either the simple name of the instance to create, or a set of
 // name-value pairs to create and initailize the object
-// @param func  Optional function to call to attempt to create an instance
 // @param result The newly created instance.
 template <typename T>
 static Status LoadSharedObject(const ConfigOptions& config_options,
                                const std::string& value,
-                               const SharedFactoryFunc<T>& func,
                                std::shared_ptr<T>* result) {
   std::string id;
   std::unordered_map<std::string, std::string> opt_map;
@@ -170,12 +149,8 @@ static Status LoadSharedObject(const ConfigOptions& config_options,
                                               value, &id, &opt_map);
   if (!status.ok()) {  // GetOptionsMap failed
     return status;
-  } else if (func == nullptr ||
-             !func(id, result)) {  // No factory, or it failed
-    return NewSharedObject(config_options, id, opt_map, result);
   } else {
-    return Customizable::ConfigureNewObject(config_options, result->get(),
-                                            opt_map);
+    return NewSharedObject(config_options, id, opt_map, result);
   }
 }
 
@@ -204,7 +179,6 @@ static Status LoadSharedObject(const ConfigOptions& config_options,
 // handled
 // @param value Either the simple name of the instance to create, or a set of
 // name-value pairs to create and initailize the object
-// @param func  Optional function to call to attempt to create an instance
 // @param result The newly created instance.
 template <typename T>
 static Status LoadManagedObject(const ConfigOptions& config_options,
@@ -270,12 +244,10 @@ static Status NewUniqueObject(
 // handled
 // @param value Either the simple name of the instance to create, or a set of
 // name-value pairs to create and initailize the object
-// @param func  Optional function to call to attempt to create an instance
 // @param result The newly created instance.
 template <typename T>
 static Status LoadUniqueObject(const ConfigOptions& config_options,
                                const std::string& value,
-                               const UniqueFactoryFunc<T>& func,
                                std::unique_ptr<T>* result) {
   std::string id;
   std::unordered_map<std::string, std::string> opt_map;
@@ -283,12 +255,8 @@ static Status LoadUniqueObject(const ConfigOptions& config_options,
                                               value, &id, &opt_map);
   if (!status.ok()) {  // GetOptionsMap failed
     return status;
-  } else if (func == nullptr ||
-             !func(id, result)) {  // No factory, or it failed
-    return NewUniqueObject(config_options, id, opt_map, result);
   } else {
-    return Customizable::ConfigureNewObject(config_options, result->get(),
-                                            opt_map);
+    return NewUniqueObject(config_options, id, opt_map, result);
   }
 }
 
@@ -337,23 +305,18 @@ static Status NewStaticObject(
 // handled
 // @param value Either the simple name of the instance to create, or a set of
 // name-value pairs to create and initailize the object
-// @param func  Optional function to call to attempt to create an instance
 // @param result The newly created instance.
 template <typename T>
 static Status LoadStaticObject(const ConfigOptions& config_options,
-                               const std::string& value,
-                               const StaticFactoryFunc<T>& func, T** result) {
+                               const std::string& value, T** result) {
   std::string id;
   std::unordered_map<std::string, std::string> opt_map;
   Status status = Customizable::GetOptionsMap(config_options, *result, value,
                                               &id, &opt_map);
   if (!status.ok()) {  // GetOptionsMap failed
     return status;
-  } else if (func == nullptr ||
-             !func(id, result)) {  // No factory, or it failed
-    return NewStaticObject(config_options, id, opt_map, result);
   } else {
-    return Customizable::ConfigureNewObject(config_options, *result, opt_map);
+    return NewStaticObject(config_options, id, opt_map, result);
   }
 }
 }  // namespace ROCKSDB_NAMESPACE

include/rocksdb/utilities/sim_cache.h

@@ -10,7 +10,7 @@
 #include <memory>
 #include <string>
 
-#include "rocksdb/cache.h"
+#include "rocksdb/advanced_cache.h"
 #include "rocksdb/env.h"
 #include "rocksdb/slice.h"
 #include "rocksdb/statistics.h"
@@ -42,13 +42,10 @@ extern std::shared_ptr<SimCache> NewSimCache(std::shared_ptr<Cache> sim_cache,
                                              std::shared_ptr<Cache> cache,
                                              int num_shard_bits);
 
-class SimCache : public Cache {
+// An abstract base class (public interface) to the SimCache implementation
+class SimCache : public CacheWrapper {
  public:
-  SimCache() {}
-
-  ~SimCache() override {}
-
-  const char* Name() const override { return "SimCache"; }
+  using CacheWrapper::CacheWrapper;
 
   // returns the maximum configured capacity of the simcache for simulation
   virtual size_t GetSimCapacity() const = 0;

include/rocksdb/utilities/stackable_db.h

@@ -136,6 +136,24 @@ class StackableDB : public DB {
                          statuses, sorted_input);
   }
 
+  using DB::MultiGetEntity;
+
+  void MultiGetEntity(const ReadOptions& options,
+                      ColumnFamilyHandle* column_family, size_t num_keys,
+                      const Slice* keys, PinnableWideColumns* results,
+                      Status* statuses, bool sorted_input) override {
+    db_->MultiGetEntity(options, column_family, num_keys, keys, results,
+                        statuses, sorted_input);
+  }
+
+  void MultiGetEntity(const ReadOptions& options, size_t num_keys,
+                      ColumnFamilyHandle** column_families, const Slice* keys,
+                      PinnableWideColumns* results, Status* statuses,
+                      bool sorted_input) override {
+    db_->MultiGetEntity(options, num_keys, column_families, keys, results,
+                        statuses, sorted_input);
+  }
+
   using DB::IngestExternalFile;
   virtual Status IngestExternalFile(
       ColumnFamilyHandle* column_family,

java/Makefile

@@ -110,6 +110,7 @@ JAVA_TESTS = \
 	org.rocksdb.BlobOptionsTest\
 	org.rocksdb.BlockBasedTableConfigTest\
 	org.rocksdb.BuiltinComparatorTest\
+	org.rocksdb.ByteBufferUnsupportedOperationTest\
 	org.rocksdb.BytewiseComparatorRegressionTest\
 	org.rocksdb.util.BytewiseComparatorTest\
 	org.rocksdb.util.BytewiseComparatorIntTest\

java/rocksjni/cache.cc

@@ -6,11 +6,10 @@
 // This file implements the "bridge" between Java and C++ for
 // ROCKSDB_NAMESPACE::Cache.
 
-#include "rocksdb/cache.h"
-
 #include <jni.h>
 
 #include "include/org_rocksdb_Cache.h"
+#include "rocksdb/advanced_cache.h"
 
 /*
  * Class:     org_rocksdb_Cache

java/rocksjni/comparatorjnicallback.cc

@@ -14,7 +14,8 @@ namespace ROCKSDB_NAMESPACE {
 ComparatorJniCallback::ComparatorJniCallback(
     JNIEnv* env, jobject jcomparator,
     const ComparatorJniCallbackOptions* options)
-    : JniCallback(env, jcomparator), m_options(options) {
+    : JniCallback(env, jcomparator),
+      m_options(std::make_unique<ComparatorJniCallbackOptions>(*options)) {
   // cache the AbstractComparatorJniBridge class as we will reuse it many times
   // for each callback
   m_abstract_comparator_jni_bridge_clazz = static_cast<jclass>(

java/rocksjni/comparatorjnicallback.h

@@ -45,15 +45,12 @@ enum ReusedSynchronisationType {
 struct ComparatorJniCallbackOptions {
   // Set the synchronisation type used to guard the reused buffers.
   // Only used if max_reused_buffer_size > 0.
-  // Default: ADAPTIVE_MUTEX
-  ReusedSynchronisationType reused_synchronisation_type =
-      ReusedSynchronisationType::ADAPTIVE_MUTEX;
+  ReusedSynchronisationType reused_synchronisation_type = ADAPTIVE_MUTEX;
 
   // Indicates if a direct byte buffer (i.e. outside of the normal
   // garbage-collected heap) is used for the callbacks to Java,
   // as opposed to a non-direct byte buffer which is a wrapper around
   // an on-heap byte[].
-  // Default: true
   bool direct_buffer = true;
 
   // Maximum size of a buffer (in bytes) that will be reused.
@@ -63,7 +60,6 @@ struct ComparatorJniCallbackOptions {
   // if it requires less than max_reused_buffer_size, then an
   // existing buffer will be reused, else a new buffer will be
   // allocated just for that callback. -1 to disable.
-  // Default: 64 bytes
   int32_t max_reused_buffer_size = 64;
 };
 
@@ -92,7 +88,7 @@ class ComparatorJniCallback : public JniCallback, public Comparator {
   virtual void FindShortestSeparator(std::string* start,
                                      const Slice& limit) const;
   virtual void FindShortSuccessor(std::string* key) const;
-  const ComparatorJniCallbackOptions* m_options;
+  const std::unique_ptr<ComparatorJniCallbackOptions> m_options;
 
  private:
   struct ThreadLocalBuf {

java/rocksjni/options.cc

@@ -3990,9 +3990,13 @@ jlong Java_org_rocksdb_ColumnFamilyOptions_getColumnFamilyOptionsFromProps__Ljav
   }
 
   auto* cf_options = new ROCKSDB_NAMESPACE::ColumnFamilyOptions();
+  ROCKSDB_NAMESPACE::ConfigOptions config_options;
+  config_options.input_strings_escaped = false;
+  config_options.ignore_unknown_options = false;
   ROCKSDB_NAMESPACE::Status status =
       ROCKSDB_NAMESPACE::GetColumnFamilyOptionsFromString(
-          ROCKSDB_NAMESPACE::ColumnFamilyOptions(), opt_string, cf_options);
+          config_options, ROCKSDB_NAMESPACE::ColumnFamilyOptions(), opt_string,
+          cf_options);
 
   env->ReleaseStringUTFChars(jopt_string, opt_string);
 
@@ -5848,9 +5852,13 @@ jlong Java_org_rocksdb_DBOptions_getDBOptionsFromProps__Ljava_lang_String_2(
     return 0;
   }
 
+  const ROCKSDB_NAMESPACE::DBOptions base_options;
   auto* db_options = new ROCKSDB_NAMESPACE::DBOptions();
+  ROCKSDB_NAMESPACE::ConfigOptions config_options(base_options);
+  config_options.input_strings_escaped = false;
+  config_options.ignore_unknown_options = false;
   ROCKSDB_NAMESPACE::Status status = ROCKSDB_NAMESPACE::GetDBOptionsFromString(
-      ROCKSDB_NAMESPACE::DBOptions(), opt_string, db_options);
+      config_options, base_options, opt_string, db_options);
 
   env->ReleaseStringUTFChars(jopt_string, opt_string);
 

java/rocksjni/write_buffer_manager.cc

@@ -7,6 +7,8 @@
 
 #include <jni.h>
 
+#include <cassert>
+
 #include "include/org_rocksdb_WriteBufferManager.h"
 #include "rocksdb/cache.h"
 #include "rocksjni/cplusplus_to_java_convert.h"

java/src/main/java/org/rocksdb/AbstractComparator.java

@@ -20,8 +20,8 @@ public abstract class AbstractComparator
     super();
   }
 
-  protected AbstractComparator(final ComparatorOptions copt) {
-    super(copt.nativeHandle_);
+  protected AbstractComparator(final ComparatorOptions comparatorOptions) {
+    super(comparatorOptions.nativeHandle_);
   }
 
   @Override

java/src/main/java/org/rocksdb/ColumnFamilyOptions.java

@@ -11,8 +11,8 @@ import java.util.*;
 /**
  * ColumnFamilyOptions to control the behavior of a database.  It will be used
  * during the creation of a {@link org.rocksdb.RocksDB} (i.e., RocksDB.open()).
- *
- * As a descendent of {@link AbstractNativeReference}, this class is {@link AutoCloseable}
+ * <p>
+ * As a descendant of {@link AbstractNativeReference}, this class is {@link AutoCloseable}
  * and will be automatically released if opened in the preamble of a try with resources block.
  */
 public class ColumnFamilyOptions extends RocksObject
@@ -24,7 +24,7 @@ public class ColumnFamilyOptions extends RocksObject
 
   /**
    * Construct ColumnFamilyOptions.
-   *
+   * <p>
    * This constructor will create (by allocating a block of memory)
    * an {@code rocksdb::ColumnFamilyOptions} in the c++ side.
    */
@@ -34,13 +34,13 @@ public class ColumnFamilyOptions extends RocksObject
 
   /**
    * Copy constructor for ColumnFamilyOptions.
-   *
+   * <p>
    * NOTE: This does a shallow copy, which means comparator, merge_operator, compaction_filter,
    * compaction_filter_factory and other pointers will be cloned!
    *
    * @param other The ColumnFamilyOptions to copy.
    */
-  public ColumnFamilyOptions(ColumnFamilyOptions other) {
+  public ColumnFamilyOptions(final ColumnFamilyOptions other) {
     super(copyColumnFamilyOptions(other.nativeHandle_));
     this.memTableConfig_ = other.memTableConfig_;
     this.tableFormatConfig_ = other.tableFormatConfig_;
@@ -707,7 +707,7 @@ public class ColumnFamilyOptions extends RocksObject
   }
 
   @Override
-  public ColumnFamilyOptions setBloomLocality(int bloomLocality) {
+  public ColumnFamilyOptions setBloomLocality(final int bloomLocality) {
     setBloomLocality(nativeHandle_, bloomLocality);
     return this;
   }
@@ -742,9 +742,7 @@ public class ColumnFamilyOptions extends RocksObject
   }
 
   @Override
-  public ColumnFamilyOptions
-  setMemtableHugePageSize(
-      long memtableHugePageSize) {
+  public ColumnFamilyOptions setMemtableHugePageSize(final long memtableHugePageSize) {
     setMemtableHugePageSize(nativeHandle_,
         memtableHugePageSize);
     return this;
@@ -756,7 +754,8 @@ public class ColumnFamilyOptions extends RocksObject
   }
 
   @Override
-  public ColumnFamilyOptions setSoftPendingCompactionBytesLimit(long softPendingCompactionBytesLimit) {
+  public ColumnFamilyOptions setSoftPendingCompactionBytesLimit(
+      final long softPendingCompactionBytesLimit) {
     setSoftPendingCompactionBytesLimit(nativeHandle_,
         softPendingCompactionBytesLimit);
     return this;
@@ -768,7 +767,8 @@ public class ColumnFamilyOptions extends RocksObject
   }
 
   @Override
-  public ColumnFamilyOptions setHardPendingCompactionBytesLimit(long hardPendingCompactionBytesLimit) {
+  public ColumnFamilyOptions setHardPendingCompactionBytesLimit(
+      final long hardPendingCompactionBytesLimit) {
     setHardPendingCompactionBytesLimit(nativeHandle_, hardPendingCompactionBytesLimit);
     return this;
   }
@@ -779,7 +779,8 @@ public class ColumnFamilyOptions extends RocksObject
   }
 
   @Override
-  public ColumnFamilyOptions setLevel0FileNumCompactionTrigger(int level0FileNumCompactionTrigger) {
+  public ColumnFamilyOptions setLevel0FileNumCompactionTrigger(
+      final int level0FileNumCompactionTrigger) {
     setLevel0FileNumCompactionTrigger(nativeHandle_, level0FileNumCompactionTrigger);
     return this;
   }
@@ -790,7 +791,7 @@ public class ColumnFamilyOptions extends RocksObject
   }
 
   @Override
-  public ColumnFamilyOptions setLevel0SlowdownWritesTrigger(int level0SlowdownWritesTrigger) {
+  public ColumnFamilyOptions setLevel0SlowdownWritesTrigger(final int level0SlowdownWritesTrigger) {
     setLevel0SlowdownWritesTrigger(nativeHandle_, level0SlowdownWritesTrigger);
     return this;
   }
@@ -801,7 +802,7 @@ public class ColumnFamilyOptions extends RocksObject
   }
 
   @Override
-  public ColumnFamilyOptions setLevel0StopWritesTrigger(int level0StopWritesTrigger) {
+  public ColumnFamilyOptions setLevel0StopWritesTrigger(final int level0StopWritesTrigger) {
     setLevel0StopWritesTrigger(nativeHandle_, level0StopWritesTrigger);
     return this;
   }
@@ -812,7 +813,8 @@ public class ColumnFamilyOptions extends RocksObject
   }
 
   @Override
-  public ColumnFamilyOptions setMaxBytesForLevelMultiplierAdditional(int[] maxBytesForLevelMultiplierAdditional) {
+  public ColumnFamilyOptions setMaxBytesForLevelMultiplierAdditional(
+      final int[] maxBytesForLevelMultiplierAdditional) {
     setMaxBytesForLevelMultiplierAdditional(nativeHandle_, maxBytesForLevelMultiplierAdditional);
     return this;
   }
@@ -823,7 +825,7 @@ public class ColumnFamilyOptions extends RocksObject
   }
 
   @Override
-  public ColumnFamilyOptions setParanoidFileChecks(boolean paranoidFileChecks) {
+  public ColumnFamilyOptions setParanoidFileChecks(final boolean paranoidFileChecks) {
     setParanoidFileChecks(nativeHandle_, paranoidFileChecks);
     return this;
   }
@@ -931,7 +933,8 @@ public class ColumnFamilyOptions extends RocksObject
   }
 
   @Override
-  public ColumnFamilyOptions setSstPartitionerFactory(SstPartitionerFactory sstPartitionerFactory) {
+  public ColumnFamilyOptions setSstPartitionerFactory(
+      final SstPartitionerFactory sstPartitionerFactory) {
     setSstPartitionerFactory(nativeHandle_, sstPartitionerFactory.nativeHandle_);
     this.sstPartitionerFactory_ = sstPartitionerFactory;
     return this;
@@ -967,9 +970,9 @@ public class ColumnFamilyOptions extends RocksObject
    * for reads. See also the options min_blob_size, blob_file_size,
    * blob_compression_type, enable_blob_garbage_collection, and
    * blob_garbage_collection_age_cutoff below.
-   *
+   * <p>
    * Default: false
-   *
+   * <p>
    * Dynamically changeable through
    * {@link RocksDB#setOptions(ColumnFamilyHandle, MutableColumnFamilyOptions)}.
    *
@@ -990,9 +993,9 @@ public class ColumnFamilyOptions extends RocksObject
    * for reads. See also the options min_blob_size, blob_file_size,
    * blob_compression_type, enable_blob_garbage_collection, and
    * blob_garbage_collection_age_cutoff below.
-   *
+   * <p>
    * Default: false
-   *
+   * <p>
    * Dynamically changeable through
    * {@link RocksDB#setOptions(ColumnFamilyHandle, MutableColumnFamilyOptions)}.
    *
@@ -1008,9 +1011,9 @@ public class ColumnFamilyOptions extends RocksObject
    * alongside the keys in SST files in the usual fashion. A value of zero for
    * this option means that all values are stored in blob files. Note that
    * enable_blob_files has to be set in order for this option to have any effect.
-   *
+   * <p>
    * Default: 0
-   *
+   * <p>
    * Dynamically changeable through
    * {@link RocksDB#setOptions(ColumnFamilyHandle, MutableColumnFamilyOptions)}.
    *
@@ -1029,9 +1032,9 @@ public class ColumnFamilyOptions extends RocksObject
    * alongside the keys in SST files in the usual fashion. A value of zero for
    * this option means that all values are stored in blob files. Note that
    * enable_blob_files has to be set in order for this option to have any effect.
-   *
+   * <p>
    * Default: 0
-   *
+   * <p>
    * Dynamically changeable through
    * {@link RocksDB#setOptions(ColumnFamilyHandle, MutableColumnFamilyOptions)}.
    *
@@ -1046,9 +1049,9 @@ public class ColumnFamilyOptions extends RocksObject
    * Set the size limit for blob files. When writing blob files, a new file is opened
    * once this limit is reached. Note that enable_blob_files has to be set in
    * order for this option to have any effect.
-   *
+   * <p>
    * Default: 256 MB
-   *
+   * <p>
    * Dynamically changeable through
    * {@link RocksDB#setOptions(ColumnFamilyHandle, MutableColumnFamilyOptions)}.
    *
@@ -1066,9 +1069,9 @@ public class ColumnFamilyOptions extends RocksObject
    * Get the size limit for blob files. When writing blob files, a new file is opened
    * once this limit is reached. Note that enable_blob_files has to be set in
    * order for this option to have any effect.
-   *
+   * <p>
    * Default: 256 MB
-   *
+   * <p>
    * Dynamically changeable through
    * {@link RocksDB#setOptions(ColumnFamilyHandle, MutableColumnFamilyOptions)}.
    *
@@ -1083,9 +1086,9 @@ public class ColumnFamilyOptions extends RocksObject
    * Set the compression algorithm to use for large values stored in blob files. Note
    * that enable_blob_files has to be set in order for this option to have any
    * effect.
-   *
+   * <p>
    * Default: no compression
-   *
+   * <p>
    * Dynamically changeable through
    * {@link RocksDB#setOptions(ColumnFamilyHandle, MutableColumnFamilyOptions)}.
    *
@@ -1103,9 +1106,9 @@ public class ColumnFamilyOptions extends RocksObject
    * Get the compression algorithm to use for large values stored in blob files. Note
    * that enable_blob_files has to be set in order for this option to have any
    * effect.
-   *
+   * <p>
    * Default: no compression
-   *
+   * <p>
    * Dynamically changeable through
    * {@link RocksDB#setOptions(ColumnFamilyHandle, MutableColumnFamilyOptions)}.
    *
@@ -1122,7 +1125,7 @@ public class ColumnFamilyOptions extends RocksObject
    * relocated to new files as they are encountered during compaction, which makes
    * it possible to clean up blob files once they contain nothing but
    * obsolete/garbage blobs. See also blob_garbage_collection_age_cutoff below.
-   *
+   * <p>
    * Default: false
    *
    * @param enableBlobGarbageCollection true iff blob garbage collection is to be enabled
@@ -1142,7 +1145,7 @@ public class ColumnFamilyOptions extends RocksObject
    * relocated to new files as they are encountered during compaction, which makes
    * it possible to clean up blob files once they contain nothing but
    * obsolete/garbage blobs. See also blob_garbage_collection_age_cutoff below.
-   *
+   * <p>
    * Default: false
    *
    * @return true iff blob garbage collection is currently enabled
@@ -1158,7 +1161,7 @@ public class ColumnFamilyOptions extends RocksObject
    * where N = garbage_collection_cutoff * number_of_blob_files. Note that
    * enable_blob_garbage_collection has to be set in order for this option to have
    * any effect.
-   *
+   * <p>
    * Default: 0.25
    *
    * @param blobGarbageCollectionAgeCutoff the new blob garbage collection age cutoff
@@ -1178,7 +1181,7 @@ public class ColumnFamilyOptions extends RocksObject
    * where N = garbage_collection_cutoff * number_of_blob_files. Note that
    * enable_blob_garbage_collection has to be set in order for this option to have
    * any effect.
-   *
+   * <p>
    * Default: 0.25
    *
    * @return the current blob garbage collection age cutoff
@@ -1194,12 +1197,12 @@ public class ColumnFamilyOptions extends RocksObject
    *  the blob files in question, assuming they are all eligible based on the
    *  value of {@link #blobGarbageCollectionAgeCutoff} above. This option is
    *  currently only supported with leveled compactions.
-   *
+   * <p>
    *  Note that {@link #enableBlobGarbageCollection} has to be set in order for this
    *  option to have any effect.
-   *
+   * <p>
    *  Default: 1.0
-   *
+   * <p>
    * Dynamically changeable through the SetOptions() API
    *
    * @param blobGarbageCollectionForceThreshold new value for the threshold
@@ -1223,9 +1226,9 @@ public class ColumnFamilyOptions extends RocksObject
 
   /**
    * Set compaction readahead for blob files.
-   *
+   * <p>
    * Default: 0
-   *
+   * <p>
    * Dynamically changeable through
    * {@link RocksDB#setOptions(ColumnFamilyHandle, MutableColumnFamilyOptions)}.
    *
@@ -1252,9 +1255,9 @@ public class ColumnFamilyOptions extends RocksObject
 
   /**
    * Set a certain LSM tree level to enable blob files.
-   *
+   * <p>
    * Default: 0
-   *
+   * <p>
    * Dynamically changeable through
    * {@link RocksDB#setOptions(ColumnFamilyHandle, MutableColumnFamilyOptions)}.
    *
@@ -1270,7 +1273,7 @@ public class ColumnFamilyOptions extends RocksObject
 
   /**
    * Get the starting LSM tree level to enable blob files.
-   *
+   * <p>
    * Default: 0
    *
    * @return the current LSM tree level to enable blob files.
@@ -1282,9 +1285,9 @@ public class ColumnFamilyOptions extends RocksObject
 
   /**
    * Set a certain prepopulate blob cache option.
-   *
+   * <p>
    * Default: 0
-   *
+   * <p>
    * Dynamically changeable through
    * {@link RocksDB#setOptions(ColumnFamilyHandle, MutableColumnFamilyOptions)}.
    *
@@ -1301,7 +1304,7 @@ public class ColumnFamilyOptions extends RocksObject
 
   /**
    * Get the prepopulate blob cache option.
-   *
+   * <p>
    * Default: 0
    *
    * @return the current prepopulate blob cache option.

java/src/main/java/org/rocksdb/CompactRangeOptions.java

@@ -14,6 +14,7 @@ public class CompactRangeOptions extends RocksObject {
   private final static byte VALUE_kSkip = 0;
   private final static byte VALUE_kIfHaveCompactionFilter = 1;
   private final static byte VALUE_kForce = 2;
+  private final static byte VALUE_kForceOptimized = 3;
 
   // For level based compaction, we can configure if we want to skip/force bottommost level
   // compaction. The order of this enum MUST follow the C++ layer. See BottommostLevelCompaction in
@@ -30,7 +31,12 @@ public class CompactRangeOptions extends RocksObject {
     /**
      * Always compact bottommost level
      */
-    kForce(VALUE_kForce);
+    kForce(VALUE_kForce),
+    /**
+     * Always compact bottommost level but in bottommost level avoid
+     * double-compacting files created in the same compaction
+     */
+    kForceOptimized(VALUE_kForceOptimized);
 
     private final byte value;
 
@@ -57,6 +63,8 @@ public class CompactRangeOptions extends RocksObject {
         case VALUE_kSkip: return kSkip;
         case VALUE_kIfHaveCompactionFilter: return kIfHaveCompactionFilter;
         case VALUE_kForce: return kForce;
+        case VALUE_kForceOptimized:
+          return kForceOptimized;
         default: return null;
       }
     }

java/src/test/java/org/rocksdb/ByteBufferUnsupportedOperationTest.java

@@ -0,0 +1,132 @@
+// Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under both the GPLv2 (found in the
+//  COPYING file in the root directory) and Apache 2.0 License
+//  (found in the LICENSE.Apache file in the root directory).
+
+package org.rocksdb;
+
+import java.nio.charset.StandardCharsets;
+import java.util.*;
+import java.util.concurrent.ConcurrentHashMap;
+import org.junit.ClassRule;
+import org.junit.Rule;
+import org.junit.Test;
+import org.junit.rules.TemporaryFolder;
+import org.rocksdb.util.ReverseBytewiseComparator;
+
+public class ByteBufferUnsupportedOperationTest {
+  @ClassRule
+  public static final RocksNativeLibraryResource ROCKS_NATIVE_LIBRARY_RESOURCE =
+      new RocksNativeLibraryResource();
+
+  @Rule public TemporaryFolder dbFolder = new TemporaryFolder();
+
+  public static class Handler {
+    private final RocksDB database;
+    private final Map<UUID, ColumnFamilyHandle> columnFamilies;
+
+    public Handler(final String path, final Options options) throws RocksDBException {
+      RocksDB.destroyDB(path, options);
+      this.database = RocksDB.open(options, path);
+      this.columnFamilies = new ConcurrentHashMap<>();
+    }
+
+    public void addTable(final UUID streamID) throws RocksDBException {
+      final ColumnFamilyOptions tableOptions = new ColumnFamilyOptions();
+      tableOptions.optimizeUniversalStyleCompaction();
+      try (final ComparatorOptions comparatorOptions = new ComparatorOptions()) {
+        // comparatorOptions.setReusedSynchronisationType(ReusedSynchronisationType.ADAPTIVE_MUTEX);
+        tableOptions.setComparator(new ReverseBytewiseComparator(comparatorOptions));
+        final ColumnFamilyDescriptor tableDescriptor = new ColumnFamilyDescriptor(
+            streamID.toString().getBytes(StandardCharsets.UTF_8), tableOptions);
+        final ColumnFamilyHandle tableHandle = database.createColumnFamily(tableDescriptor);
+        columnFamilies.put(streamID, tableHandle);
+      }
+    }
+
+    public void updateAll(final List<byte[][]> keyValuePairs, final UUID streamID)
+        throws RocksDBException {
+      final ColumnFamilyHandle currTable = columnFamilies.get(streamID);
+      try (final WriteBatch batchedWrite = new WriteBatch();
+           final WriteOptions writeOptions = new WriteOptions()) {
+        for (final byte[][] pair : keyValuePairs) {
+          final byte[] keyBytes = pair[0];
+          final byte[] valueBytes = pair[1];
+          batchedWrite.put(currTable, keyBytes, valueBytes);
+        }
+        database.write(writeOptions, batchedWrite);
+      }
+    }
+    public boolean containsValue(final byte[] encodedValue, final UUID streamID) {
+      try (final RocksIterator iter = database.newIterator(columnFamilies.get(streamID))) {
+        iter.seekToFirst();
+        while (iter.isValid()) {
+          final byte[] val = iter.value();
+          if (Arrays.equals(val, encodedValue)) {
+            return true;
+          }
+          iter.next();
+        }
+      }
+      return false;
+    }
+
+    public void close() {
+      for (final ColumnFamilyHandle handle : columnFamilies.values()) {
+        handle.close();
+      }
+      database.close();
+    }
+  }
+
+  private void inner(final int numRepeats) throws RocksDBException {
+    final Options opts = new Options();
+    opts.setCreateIfMissing(true);
+    final Handler handler = new Handler("testDB", opts);
+    final UUID stream1 = UUID.randomUUID();
+
+    final List<byte[][]> entries = new ArrayList<>();
+    for (int i = 0; i < numRepeats; i++) {
+      final byte[] value = value(i);
+      final byte[] key = key(i);
+      entries.add(new byte[][] {key, value});
+    }
+    handler.addTable(stream1);
+    handler.updateAll(entries, stream1);
+
+    for (int i = 0; i < numRepeats; i++) {
+      final byte[] val = value(i);
+      final boolean hasValue = handler.containsValue(val, stream1);
+      if (!hasValue) {
+        throw new IllegalStateException("not has value " + i);
+      }
+    }
+
+    handler.close();
+  }
+
+  private static byte[] key(final int i) {
+    return ("key" + i).getBytes(StandardCharsets.UTF_8);
+  }
+
+  private static byte[] value(final int i) {
+    return ("value" + i).getBytes(StandardCharsets.UTF_8);
+  }
+
+  @Test
+  public void unsupportedOperation() throws RocksDBException {
+    final int numRepeats = 1000;
+    final int repeatTest = 10;
+
+    // the error is not always reproducible... let's try to increase the odds by repeating the main
+    // test body
+    for (int i = 0; i < repeatTest; i++) {
+      try {
+        inner(numRepeats);
+      } catch (final RuntimeException runtimeException) {
+        System.out.println("Exception on repeat " + i);
+        throw runtimeException;
+      }
+    }
+  }
+}

java/src/test/java/org/rocksdb/CompactRangeOptionsTest.java

@@ -39,6 +39,9 @@ public class CompactRangeOptionsTest {
     value = BottommostLevelCompaction.kIfHaveCompactionFilter;
     opt.setBottommostLevelCompaction(value);
     assertThat(opt.bottommostLevelCompaction()).isEqualTo(value);
+    value = BottommostLevelCompaction.kForceOptimized;
+    opt.setBottommostLevelCompaction(value);
+    assertThat(opt.bottommostLevelCompaction()).isEqualTo(value);
   }
 
   @Test

java/src/test/java/org/rocksdb/IngestExternalFileOptionsTest.java

@@ -99,9 +99,9 @@ public class IngestExternalFileOptionsTest {
   public void writeGlobalSeqno() {
     try (final IngestExternalFileOptions options =
              new IngestExternalFileOptions()) {
-      assertThat(options.writeGlobalSeqno()).isTrue();
-      options.setWriteGlobalSeqno(false);
       assertThat(options.writeGlobalSeqno()).isFalse();
+      options.setWriteGlobalSeqno(true);
+      assertThat(options.writeGlobalSeqno()).isTrue();
     }
   }
 }

memtable/write_buffer_manager_test.cc

@@ -9,6 +9,7 @@
 
 #include "rocksdb/write_buffer_manager.h"
 
+#include "rocksdb/advanced_cache.h"
 #include "test_util/testharness.h"
 
 namespace ROCKSDB_NAMESPACE {

monitoring/statistics.cc

@@ -300,7 +300,7 @@ Status Statistics::CreateFromString(const ConfigOptions& config_options,
   } else if (id == kNullptrString) {
     result->reset();
   } else {
-    s = LoadSharedObject<Statistics>(config_options, id, nullptr, result);
+    s = LoadSharedObject<Statistics>(config_options, id, result);
   }
   return s;
 }

options/cf_options.cc

@@ -16,6 +16,7 @@
 #include "options/options_helper.h"
 #include "options/options_parser.h"
 #include "port/port.h"
+#include "rocksdb/advanced_cache.h"
 #include "rocksdb/compaction_filter.h"
 #include "rocksdb/concurrent_task_limiter.h"
 #include "rocksdb/configurable.h"

options/customizable_test.cc

@@ -160,19 +160,6 @@ class BCustomizable : public TestCustomizable {
   BOptions opts_;
 };
 
-static bool LoadSharedB(const std::string& id,
-                        std::shared_ptr<TestCustomizable>* result) {
-  if (id == "B") {
-    result->reset(new BCustomizable(id));
-    return true;
-  } else if (id.empty()) {
-    result->reset();
-    return true;
-  } else {
-    return false;
-  }
-}
-
 static int A_count = 0;
 static int RegisterCustomTestObjects(ObjectLibrary& library,
                                      const std::string& /*arg*/) {
@@ -184,6 +171,12 @@ static int RegisterCustomTestObjects(ObjectLibrary& library,
         A_count++;
         return guard->get();
       });
+  library.AddFactory<TestCustomizable>(
+      "B", [](const std::string& name, std::unique_ptr<TestCustomizable>* guard,
+              std::string* /* msg */) {
+        guard->reset(new BCustomizable(name));
+        return guard->get();
+      });
 
   library.AddFactory<TestCustomizable>(
       "S", [](const std::string& name,
@@ -252,46 +245,19 @@ static void GetMapFromProperties(
 Status TestCustomizable::CreateFromString(
     const ConfigOptions& config_options, const std::string& value,
     std::shared_ptr<TestCustomizable>* result) {
-  return LoadSharedObject<TestCustomizable>(config_options, value, LoadSharedB,
-                                            result);
+  return LoadSharedObject<TestCustomizable>(config_options, value, result);
 }
 
 Status TestCustomizable::CreateFromString(
     const ConfigOptions& config_options, const std::string& value,
     std::unique_ptr<TestCustomizable>* result) {
-  return LoadUniqueObject<TestCustomizable>(
-      config_options, value,
-      [](const std::string& id, std::unique_ptr<TestCustomizable>* u) {
-        if (id == "B") {
-          u->reset(new BCustomizable(id));
-          return true;
-        } else if (id.empty()) {
-          u->reset();
-          return true;
-        } else {
-          return false;
-        }
-      },
-      result);
+  return LoadUniqueObject<TestCustomizable>(config_options, value, result);
 }
 
 Status TestCustomizable::CreateFromString(const ConfigOptions& config_options,
                                           const std::string& value,
                                           TestCustomizable** result) {
-  return LoadStaticObject<TestCustomizable>(
-      config_options, value,
-      [](const std::string& id, TestCustomizable** ptr) {
-        if (id == "B") {
-          *ptr = new BCustomizable(id);
-          return true;
-        } else if (id.empty()) {
-          *ptr = nullptr;
-          return true;
-        } else {
-          return false;
-        }
-      },
-      result);
+  return LoadStaticObject<TestCustomizable>(config_options, value, result);
 }
 
 class CustomizableTest : public testing::Test {
@@ -980,72 +946,36 @@ TEST_F(CustomizableTest, IgnoreUnknownObjects) {
   std::unique_ptr<TestCustomizable> unique;
   TestCustomizable* pointer = nullptr;
   ignore.ignore_unsupported_options = false;
-  ASSERT_NOK(
-      LoadSharedObject<TestCustomizable>(ignore, "Unknown", nullptr, &shared));
-  ASSERT_NOK(
-      LoadUniqueObject<TestCustomizable>(ignore, "Unknown", nullptr, &unique));
-  ASSERT_NOK(
-      LoadStaticObject<TestCustomizable>(ignore, "Unknown", nullptr, &pointer));
+  ASSERT_NOK(LoadSharedObject<TestCustomizable>(ignore, "Unknown", &shared));
+  ASSERT_NOK(LoadUniqueObject<TestCustomizable>(ignore, "Unknown", &unique));
+  ASSERT_NOK(LoadStaticObject<TestCustomizable>(ignore, "Unknown", &pointer));
   ASSERT_EQ(shared.get(), nullptr);
   ASSERT_EQ(unique.get(), nullptr);
   ASSERT_EQ(pointer, nullptr);
   ignore.ignore_unsupported_options = true;
-  ASSERT_OK(
-      LoadSharedObject<TestCustomizable>(ignore, "Unknown", nullptr, &shared));
-  ASSERT_OK(
-      LoadUniqueObject<TestCustomizable>(ignore, "Unknown", nullptr, &unique));
-  ASSERT_OK(
-      LoadStaticObject<TestCustomizable>(ignore, "Unknown", nullptr, &pointer));
+  ASSERT_OK(LoadSharedObject<TestCustomizable>(ignore, "Unknown", &shared));
+  ASSERT_OK(LoadUniqueObject<TestCustomizable>(ignore, "Unknown", &unique));
+  ASSERT_OK(LoadStaticObject<TestCustomizable>(ignore, "Unknown", &pointer));
   ASSERT_EQ(shared.get(), nullptr);
   ASSERT_EQ(unique.get(), nullptr);
   ASSERT_EQ(pointer, nullptr);
-  ASSERT_OK(LoadSharedObject<TestCustomizable>(ignore, "id=Unknown", nullptr,
-                                               &shared));
-  ASSERT_OK(LoadUniqueObject<TestCustomizable>(ignore, "id=Unknown", nullptr,
-                                               &unique));
-  ASSERT_OK(LoadStaticObject<TestCustomizable>(ignore, "id=Unknown", nullptr,
-                                               &pointer));
+  ASSERT_OK(LoadSharedObject<TestCustomizable>(ignore, "id=Unknown", &shared));
+  ASSERT_OK(LoadUniqueObject<TestCustomizable>(ignore, "id=Unknown", &unique));
+  ASSERT_OK(LoadStaticObject<TestCustomizable>(ignore, "id=Unknown", &pointer));
   ASSERT_EQ(shared.get(), nullptr);
   ASSERT_EQ(unique.get(), nullptr);
   ASSERT_EQ(pointer, nullptr);
   ASSERT_OK(LoadSharedObject<TestCustomizable>(ignore, "id=Unknown;option=bad",
-                                               nullptr, &shared));
+                                               &shared));
   ASSERT_OK(LoadUniqueObject<TestCustomizable>(ignore, "id=Unknown;option=bad",
-                                               nullptr, &unique));
+                                               &unique));
   ASSERT_OK(LoadStaticObject<TestCustomizable>(ignore, "id=Unknown;option=bad",
-                                               nullptr, &pointer));
+                                               &pointer));
   ASSERT_EQ(shared.get(), nullptr);
   ASSERT_EQ(unique.get(), nullptr);
   ASSERT_EQ(pointer, nullptr);
 }
 
-TEST_F(CustomizableTest, FactoryFunctionTest) {
-  std::shared_ptr<TestCustomizable> shared;
-  std::unique_ptr<TestCustomizable> unique;
-  TestCustomizable* pointer = nullptr;
-  ConfigOptions ignore = config_options_;
-  ignore.ignore_unsupported_options = false;
-  ASSERT_OK(TestCustomizable::CreateFromString(ignore, "B", &shared));
-  ASSERT_OK(TestCustomizable::CreateFromString(ignore, "B", &unique));
-  ASSERT_OK(TestCustomizable::CreateFromString(ignore, "B", &pointer));
-  ASSERT_NE(shared.get(), nullptr);
-  ASSERT_NE(unique.get(), nullptr);
-  ASSERT_NE(pointer, nullptr);
-  delete pointer;
-  pointer = nullptr;
-  ASSERT_OK(TestCustomizable::CreateFromString(ignore, "id=", &shared));
-  ASSERT_OK(TestCustomizable::CreateFromString(ignore, "id=", &unique));
-  ASSERT_OK(TestCustomizable::CreateFromString(ignore, "id=", &pointer));
-  ASSERT_EQ(shared.get(), nullptr);
-  ASSERT_EQ(unique.get(), nullptr);
-  ASSERT_EQ(pointer, nullptr);
-  ASSERT_NOK(TestCustomizable::CreateFromString(ignore, "option=bad", &shared));
-  ASSERT_NOK(TestCustomizable::CreateFromString(ignore, "option=bad", &unique));
-  ASSERT_NOK(
-      TestCustomizable::CreateFromString(ignore, "option=bad", &pointer));
-  ASSERT_EQ(pointer, nullptr);
-}
-
 TEST_F(CustomizableTest, URLFactoryTest) {
   std::unique_ptr<TestCustomizable> unique;
   config_options_.registry->AddLibrary("URL")->AddFactory<TestCustomizable>(

options/db_options.cc

@@ -12,6 +12,7 @@
 #include "options/options_helper.h"
 #include "options/options_parser.h"
 #include "port/port.h"
+#include "rocksdb/advanced_cache.h"
 #include "rocksdb/configurable.h"
 #include "rocksdb/env.h"
 #include "rocksdb/file_system.h"

options/options_helper.cc

@@ -658,18 +658,6 @@ Status GetStringFromCompressionType(std::string* compression_str,
   }
 }
 
-Status GetColumnFamilyOptionsFromMap(
-    const ColumnFamilyOptions& base_options,
-    const std::unordered_map<std::string, std::string>& opts_map,
-    ColumnFamilyOptions* new_options, bool input_strings_escaped,
-    bool ignore_unknown_options) {
-  ConfigOptions config_options;
-  config_options.ignore_unknown_options = ignore_unknown_options;
-  config_options.input_strings_escaped = input_strings_escaped;
-  return GetColumnFamilyOptionsFromMap(config_options, base_options, opts_map,
-                                       new_options);
-}
-
 Status GetColumnFamilyOptionsFromMap(
     const ConfigOptions& config_options,
     const ColumnFamilyOptions& base_options,
@@ -691,17 +679,6 @@ Status GetColumnFamilyOptionsFromMap(
   }
 }
 
-Status GetColumnFamilyOptionsFromString(
-    const ColumnFamilyOptions& base_options,
-    const std::string& opts_str,
-    ColumnFamilyOptions* new_options) {
-  ConfigOptions config_options;
-  config_options.input_strings_escaped = false;
-  config_options.ignore_unknown_options = false;
-  return GetColumnFamilyOptionsFromString(config_options, base_options,
-                                          opts_str, new_options);
-}
-
 Status GetColumnFamilyOptionsFromString(const ConfigOptions& config_options,
                                         const ColumnFamilyOptions& base_options,
                                         const std::string& opts_str,
@@ -716,18 +693,6 @@ Status GetColumnFamilyOptionsFromString(const ConfigOptions& config_options,
                                        new_options);
 }
 
-Status GetDBOptionsFromMap(
-    const DBOptions& base_options,
-    const std::unordered_map<std::string, std::string>& opts_map,
-    DBOptions* new_options, bool input_strings_escaped,
-    bool ignore_unknown_options) {
-  ConfigOptions config_options(base_options);
-  config_options.input_strings_escaped = input_strings_escaped;
-  config_options.ignore_unknown_options = ignore_unknown_options;
-  return GetDBOptionsFromMap(config_options, base_options, opts_map,
-                             new_options);
-}
-
 Status GetDBOptionsFromMap(
     const ConfigOptions& config_options, const DBOptions& base_options,
     const std::unordered_map<std::string, std::string>& opts_map,
@@ -746,17 +711,6 @@ Status GetDBOptionsFromMap(
   }
 }
 
-Status GetDBOptionsFromString(const DBOptions& base_options,
-                              const std::string& opts_str,
-                              DBOptions* new_options) {
-  ConfigOptions config_options(base_options);
-  config_options.input_strings_escaped = false;
-  config_options.ignore_unknown_options = false;
-
-  return GetDBOptionsFromString(config_options, base_options, opts_str,
-                                new_options);
-}
-
 Status GetDBOptionsFromString(const ConfigOptions& config_options,
                               const DBOptions& base_options,
                               const std::string& opts_str,

options/options_settable_test.cc

@@ -167,8 +167,13 @@ TEST_F(OptionsSettableTest, BlockBasedTableOptionsAllFieldsSettable) {
                       kBbtoExcluded);
 
   // Need to update the option string if a new option is added.
+  ConfigOptions config_options;
+  config_options.input_strings_escaped = false;
+  config_options.ignore_unknown_options = false;
+  config_options.invoke_prepare_options = false;
+  config_options.ignore_unsupported_options = false;
   ASSERT_OK(GetBlockBasedTableOptionsFromString(
-      *bbto,
+      config_options, *bbto,
       "cache_index_and_filter_blocks=1;"
       "cache_index_and_filter_blocks_with_high_priority=true;"
       "metadata_cache_options={top_level_index_pinning=kFallback;"
@@ -273,8 +278,11 @@ TEST_F(OptionsSettableTest, DBOptionsAllFieldsSettable) {
   FillWithSpecialChar(new_options_ptr, sizeof(DBOptions), kDBOptionsExcluded);
 
   // Need to update the option string if a new option is added.
+  ConfigOptions config_options(*options);
+  config_options.input_strings_escaped = false;
+  config_options.ignore_unknown_options = false;
   ASSERT_OK(
-      GetDBOptionsFromString(*options,
+      GetDBOptionsFromString(config_options, *options,
                              "wal_bytes_per_sync=4295048118;"
                              "delete_obsolete_files_period_micros=4294967758;"
                              "WAL_ttl_seconds=4295008036;"
@@ -461,8 +469,11 @@ TEST_F(OptionsSettableTest, ColumnFamilyOptionsAllFieldsSettable) {
                       kColumnFamilyOptionsExcluded);
 
   // Need to update the option string if a new option is added.
+  ConfigOptions config_options;
+  config_options.input_strings_escaped = false;
+  config_options.ignore_unknown_options = false;
   ASSERT_OK(GetColumnFamilyOptionsFromString(
-      *options,
+      config_options, *options,
       "compaction_filter_factory=mpudlojcujCompactionFilterFactory;"
       "table_factory=PlainTable;"
       "prefix_extractor=rocksdb.CappedPrefix.13;"

options/options_test.cc

@@ -811,9 +811,11 @@ TEST_F(OptionsTest, OldInterfaceTest) {
   ColumnFamilyOptions base_cf_opt;
   ColumnFamilyOptions new_cf_opt;
   ConfigOptions exact;
-
+  ConfigOptions cf_config_options;
+  cf_config_options.input_strings_escaped = false;
+  cf_config_options.ignore_unknown_options = false;
   ASSERT_OK(GetColumnFamilyOptionsFromString(
-      base_cf_opt,
+      cf_config_options, base_cf_opt,
       "write_buffer_size=18;prefix_extractor=capped:8;"
       "arena_block_size=19",
       &new_cf_opt));
@@ -824,7 +826,7 @@ TEST_F(OptionsTest, OldInterfaceTest) {
 
   // And with a bad option
   ASSERT_NOK(GetColumnFamilyOptionsFromString(
-      base_cf_opt,
+      cf_config_options, base_cf_opt,
       "write_buffer_size=10;max_write_buffer_number=16;"
       "block_based_table_factory={xx_block_size=4;}",
       &new_cf_opt));
@@ -836,15 +838,17 @@ TEST_F(OptionsTest, OldInterfaceTest) {
       {"max_write_buffer_number", "2"},
       {"min_write_buffer_number_to_merge", "3"},
   };
-  ASSERT_OK(
-      GetColumnFamilyOptionsFromMap(base_cf_opt, cf_options_map, &new_cf_opt));
+  ASSERT_OK(GetColumnFamilyOptionsFromMap(cf_config_options, base_cf_opt,
+                                          cf_options_map, &new_cf_opt));
   cf_options_map["unknown_option"] = "1";
-  ASSERT_NOK(
-      GetColumnFamilyOptionsFromMap(base_cf_opt, cf_options_map, &new_cf_opt));
+  ASSERT_NOK(GetColumnFamilyOptionsFromMap(cf_config_options, base_cf_opt,
+                                           cf_options_map, &new_cf_opt));
   ASSERT_OK(
       RocksDBOptionsParser::VerifyCFOptions(exact, base_cf_opt, new_cf_opt));
-  ASSERT_OK(GetColumnFamilyOptionsFromMap(base_cf_opt, cf_options_map,
-                                          &new_cf_opt, true, true));
+  cf_config_options.input_strings_escaped = true;
+  cf_config_options.ignore_unknown_options = true;
+  ASSERT_OK(GetColumnFamilyOptionsFromMap(cf_config_options, base_cf_opt,
+                                          cf_options_map, &new_cf_opt));
 
   DBOptions base_db_opt;
   DBOptions new_db_opt;
@@ -857,7 +861,12 @@ TEST_F(OptionsTest, OldInterfaceTest) {
       {"verify_sst_unique_id_in_manifest", "true"},
       {"max_open_files", "32"},
   };
-  ASSERT_OK(GetDBOptionsFromMap(base_db_opt, db_options_map, &new_db_opt));
+
+  ConfigOptions db_config_options(base_db_opt);
+  db_config_options.input_strings_escaped = false;
+  db_config_options.ignore_unknown_options = false;
+  ASSERT_OK(GetDBOptionsFromMap(db_config_options, base_db_opt, db_options_map,
+                                &new_db_opt));
   ASSERT_EQ(new_db_opt.create_if_missing, false);
   ASSERT_EQ(new_db_opt.create_missing_column_families, true);
   ASSERT_EQ(new_db_opt.error_if_exists, false);
@@ -866,23 +875,28 @@ TEST_F(OptionsTest, OldInterfaceTest) {
   ASSERT_EQ(new_db_opt.verify_sst_unique_id_in_manifest, true);
   ASSERT_EQ(new_db_opt.max_open_files, 32);
   db_options_map["unknown_option"] = "1";
-  Status s = GetDBOptionsFromMap(base_db_opt, db_options_map, &new_db_opt);
+  Status s = GetDBOptionsFromMap(db_config_options, base_db_opt, db_options_map,
+                                 &new_db_opt);
   ASSERT_NOK(s);
   ASSERT_TRUE(s.IsInvalidArgument());
 
   ASSERT_OK(
       RocksDBOptionsParser::VerifyDBOptions(exact, base_db_opt, new_db_opt));
-  ASSERT_OK(GetDBOptionsFromMap(base_db_opt, db_options_map, &new_db_opt, true,
-                                true));
+  db_config_options.input_strings_escaped = true;
+  db_config_options.ignore_unknown_options = true;
+  ASSERT_OK(GetDBOptionsFromMap(db_config_options, base_db_opt, db_options_map,
+                                &new_db_opt));
+  db_config_options.input_strings_escaped = false;
+  db_config_options.ignore_unknown_options = false;
   ASSERT_OK(GetDBOptionsFromString(
-      base_db_opt,
+      db_config_options, base_db_opt,
       "create_if_missing=false;error_if_exists=false;max_open_files=42;",
       &new_db_opt));
   ASSERT_EQ(new_db_opt.create_if_missing, false);
   ASSERT_EQ(new_db_opt.error_if_exists, false);
   ASSERT_EQ(new_db_opt.max_open_files, 42);
   s = GetDBOptionsFromString(
-      base_db_opt,
+      db_config_options, base_db_opt,
       "create_if_missing=false;error_if_exists=false;max_open_files=42;"
       "unknown_option=1;",
       &new_db_opt);
@@ -2355,8 +2369,11 @@ TEST_F(OptionsOldApiTest, GetOptionsFromMapTest) {
 
   ColumnFamilyOptions base_cf_opt;
   ColumnFamilyOptions new_cf_opt;
-  ASSERT_OK(GetColumnFamilyOptionsFromMap(
-            base_cf_opt, cf_options_map, &new_cf_opt));
+  ConfigOptions cf_config_options;
+  cf_config_options.ignore_unknown_options = false;
+  cf_config_options.input_strings_escaped = false;
+  ASSERT_OK(GetColumnFamilyOptionsFromMap(cf_config_options, base_cf_opt,
+                                          cf_options_map, &new_cf_opt));
   ASSERT_EQ(new_cf_opt.write_buffer_size, 1U);
   ASSERT_EQ(new_cf_opt.max_write_buffer_number, 2);
   ASSERT_EQ(new_cf_opt.min_write_buffer_number_to_merge, 3);
@@ -2444,8 +2461,8 @@ TEST_F(OptionsOldApiTest, GetOptionsFromMapTest) {
   ASSERT_EQ(new_cf_opt.bottommost_temperature, Temperature::kWarm);
 
   cf_options_map["write_buffer_size"] = "hello";
-  ASSERT_NOK(GetColumnFamilyOptionsFromMap(
-             base_cf_opt, cf_options_map, &new_cf_opt));
+  ASSERT_NOK(GetColumnFamilyOptionsFromMap(cf_config_options, base_cf_opt,
+                                           cf_options_map, &new_cf_opt));
   ConfigOptions exact, loose;
   exact.sanity_level = ConfigOptions::kSanityLevelExactMatch;
   loose.sanity_level = ConfigOptions::kSanityLevelLooselyCompatible;
@@ -2453,18 +2470,18 @@ TEST_F(OptionsOldApiTest, GetOptionsFromMapTest) {
   ASSERT_OK(RocksDBOptionsParser::VerifyCFOptions(exact, base_cf_opt, new_cf_opt));
 
   cf_options_map["write_buffer_size"] = "1";
-  ASSERT_OK(GetColumnFamilyOptionsFromMap(
-            base_cf_opt, cf_options_map, &new_cf_opt));
+  ASSERT_OK(GetColumnFamilyOptionsFromMap(cf_config_options, base_cf_opt,
+                                          cf_options_map, &new_cf_opt));
 
   cf_options_map["unknown_option"] = "1";
-  ASSERT_NOK(GetColumnFamilyOptionsFromMap(
-             base_cf_opt, cf_options_map, &new_cf_opt));
+  ASSERT_NOK(GetColumnFamilyOptionsFromMap(cf_config_options, base_cf_opt,
+                                           cf_options_map, &new_cf_opt));
   ASSERT_OK(RocksDBOptionsParser::VerifyCFOptions(exact, base_cf_opt, new_cf_opt));
 
-  ASSERT_OK(GetColumnFamilyOptionsFromMap(base_cf_opt, cf_options_map,
-                                          &new_cf_opt,
-                                          false, /* input_strings_escaped  */
-                                          true /* ignore_unknown_options */));
+  cf_config_options.input_strings_escaped = false;
+  cf_config_options.ignore_unknown_options = true;
+  ASSERT_OK(GetColumnFamilyOptionsFromMap(cf_config_options, base_cf_opt,
+                                          cf_options_map, &new_cf_opt));
   ASSERT_OK(RocksDBOptionsParser::VerifyCFOptions(
       loose, base_cf_opt, new_cf_opt, nullptr /* new_opt_map */));
   ASSERT_NOK(RocksDBOptionsParser::VerifyCFOptions(
@@ -2472,7 +2489,11 @@ TEST_F(OptionsOldApiTest, GetOptionsFromMapTest) {
 
   DBOptions base_db_opt;
   DBOptions new_db_opt;
-  ASSERT_OK(GetDBOptionsFromMap(base_db_opt, db_options_map, &new_db_opt));
+  ConfigOptions db_config_options(base_db_opt);
+  db_config_options.input_strings_escaped = false;
+  db_config_options.ignore_unknown_options = false;
+  ASSERT_OK(GetDBOptionsFromMap(db_config_options, base_db_opt, db_options_map,
+                                &new_db_opt));
   ASSERT_EQ(new_db_opt.create_if_missing, false);
   ASSERT_EQ(new_db_opt.create_missing_column_families, true);
   ASSERT_EQ(new_db_opt.error_if_exists, false);
@@ -2515,18 +2536,21 @@ TEST_F(OptionsOldApiTest, GetOptionsFromMapTest) {
   ASSERT_EQ(new_db_opt.strict_bytes_per_sync, true);
 
   db_options_map["max_open_files"] = "hello";
-  ASSERT_NOK(GetDBOptionsFromMap(base_db_opt, db_options_map, &new_db_opt));
+  ASSERT_NOK(GetDBOptionsFromMap(db_config_options, base_db_opt, db_options_map,
+                                 &new_db_opt));
   ASSERT_OK(RocksDBOptionsParser::VerifyDBOptions(exact, base_db_opt, new_db_opt));
   ASSERT_OK(RocksDBOptionsParser::VerifyDBOptions(loose, base_db_opt, new_db_opt));
 
   // unknow options should fail parsing without ignore_unknown_options = true
   db_options_map["unknown_db_option"] = "1";
-  ASSERT_NOK(GetDBOptionsFromMap(base_db_opt, db_options_map, &new_db_opt));
+  ASSERT_NOK(GetDBOptionsFromMap(db_config_options, base_db_opt, db_options_map,
+                                 &new_db_opt));
   ASSERT_OK(RocksDBOptionsParser::VerifyDBOptions(exact, base_db_opt, new_db_opt));
 
-  ASSERT_OK(GetDBOptionsFromMap(base_db_opt, db_options_map, &new_db_opt,
-                                false, /* input_strings_escaped  */
-                                true /* ignore_unknown_options */));
+  db_config_options.input_strings_escaped = false;
+  db_config_options.ignore_unknown_options = true;
+  ASSERT_OK(GetDBOptionsFromMap(db_config_options, base_db_opt, db_options_map,
+                                &new_db_opt));
   ASSERT_OK(RocksDBOptionsParser::VerifyDBOptions(loose, base_db_opt, new_db_opt));
   ASSERT_NOK(RocksDBOptionsParser::VerifyDBOptions(exact, base_db_opt, new_db_opt));
 }
@@ -2535,33 +2559,38 @@ TEST_F(OptionsOldApiTest, GetColumnFamilyOptionsFromStringTest) {
   ColumnFamilyOptions base_cf_opt;
   ColumnFamilyOptions new_cf_opt;
   base_cf_opt.table_factory.reset();
-  ASSERT_OK(GetColumnFamilyOptionsFromString(base_cf_opt, "", &new_cf_opt));
-  ASSERT_OK(GetColumnFamilyOptionsFromString(base_cf_opt,
-            "write_buffer_size=5", &new_cf_opt));
+  ConfigOptions config_options;
+  config_options.input_strings_escaped = false;
+  config_options.ignore_unknown_options = false;
+  ASSERT_OK(GetColumnFamilyOptionsFromString(config_options, base_cf_opt, "",
+                                             &new_cf_opt));
+  ASSERT_OK(GetColumnFamilyOptionsFromString(
+      config_options, base_cf_opt, "write_buffer_size=5", &new_cf_opt));
   ASSERT_EQ(new_cf_opt.write_buffer_size, 5U);
   ASSERT_TRUE(new_cf_opt.table_factory == nullptr);
-  ASSERT_OK(GetColumnFamilyOptionsFromString(base_cf_opt,
-            "write_buffer_size=6;", &new_cf_opt));
+  ASSERT_OK(GetColumnFamilyOptionsFromString(
+      config_options, base_cf_opt, "write_buffer_size=6;", &new_cf_opt));
   ASSERT_EQ(new_cf_opt.write_buffer_size, 6U);
-  ASSERT_OK(GetColumnFamilyOptionsFromString(base_cf_opt,
-            "  write_buffer_size =  7  ", &new_cf_opt));
+  ASSERT_OK(GetColumnFamilyOptionsFromString(
+      config_options, base_cf_opt, "  write_buffer_size =  7  ", &new_cf_opt));
   ASSERT_EQ(new_cf_opt.write_buffer_size, 7U);
-  ASSERT_OK(GetColumnFamilyOptionsFromString(base_cf_opt,
-            "  write_buffer_size =  8 ; ", &new_cf_opt));
+  ASSERT_OK(GetColumnFamilyOptionsFromString(
+      config_options, base_cf_opt, "  write_buffer_size =  8 ; ", &new_cf_opt));
   ASSERT_EQ(new_cf_opt.write_buffer_size, 8U);
-  ASSERT_OK(GetColumnFamilyOptionsFromString(base_cf_opt,
-            "write_buffer_size=9;max_write_buffer_number=10", &new_cf_opt));
+  ASSERT_OK(GetColumnFamilyOptionsFromString(
+      config_options, base_cf_opt,
+      "write_buffer_size=9;max_write_buffer_number=10", &new_cf_opt));
   ASSERT_EQ(new_cf_opt.write_buffer_size, 9U);
   ASSERT_EQ(new_cf_opt.max_write_buffer_number, 10);
-  ASSERT_OK(GetColumnFamilyOptionsFromString(base_cf_opt,
-            "write_buffer_size=11; max_write_buffer_number  =  12 ;",
-            &new_cf_opt));
+  ASSERT_OK(GetColumnFamilyOptionsFromString(
+      config_options, base_cf_opt,
+      "write_buffer_size=11; max_write_buffer_number  =  12 ;", &new_cf_opt));
   ASSERT_EQ(new_cf_opt.write_buffer_size, 11U);
   ASSERT_EQ(new_cf_opt.max_write_buffer_number, 12);
   // Wrong name "max_write_buffer_number_"
-  ASSERT_NOK(GetColumnFamilyOptionsFromString(base_cf_opt,
-             "write_buffer_size=13;max_write_buffer_number_=14;",
-              &new_cf_opt));
+  ASSERT_NOK(GetColumnFamilyOptionsFromString(
+      config_options, base_cf_opt,
+      "write_buffer_size=13;max_write_buffer_number_=14;", &new_cf_opt));
   ConfigOptions exact;
   exact.sanity_level = ConfigOptions::kSanityLevelExactMatch;
   ASSERT_OK(RocksDBOptionsParser::VerifyCFOptions(exact, base_cf_opt, new_cf_opt));
@@ -2574,30 +2603,34 @@ TEST_F(OptionsOldApiTest, GetColumnFamilyOptionsFromStringTest) {
          std::unique_ptr<const Comparator>* /*guard*/,
          std::string* /* errmsg */) { return ReverseBytewiseComparator(); });
 
-  ASSERT_OK(GetColumnFamilyOptionsFromString(
-      base_cf_opt, "comparator=" + kCompName + ";", &new_cf_opt));
+  ASSERT_OK(GetColumnFamilyOptionsFromString(config_options, base_cf_opt,
+                                             "comparator=" + kCompName + ";",
+                                             &new_cf_opt));
   ASSERT_EQ(new_cf_opt.comparator, ReverseBytewiseComparator());
 
   // MergeOperator from object registry
   std::unique_ptr<BytesXOROperator> bxo(new BytesXOROperator());
   std::string kMoName = bxo->Name();
-  ASSERT_OK(GetColumnFamilyOptionsFromString(
-      base_cf_opt, "merge_operator=" + kMoName + ";", &new_cf_opt));
+  ASSERT_OK(GetColumnFamilyOptionsFromString(config_options, base_cf_opt,
+                                             "merge_operator=" + kMoName + ";",
+                                             &new_cf_opt));
   ASSERT_EQ(kMoName, std::string(new_cf_opt.merge_operator->Name()));
 
   // Wrong key/value pair
-  ASSERT_NOK(GetColumnFamilyOptionsFromString(base_cf_opt,
-             "write_buffer_size=13;max_write_buffer_number;", &new_cf_opt));
+  ASSERT_NOK(GetColumnFamilyOptionsFromString(
+      config_options, base_cf_opt,
+      "write_buffer_size=13;max_write_buffer_number;", &new_cf_opt));
   ASSERT_OK(RocksDBOptionsParser::VerifyCFOptions(exact, base_cf_opt, new_cf_opt));
 
   // Error Paring value
-  ASSERT_NOK(GetColumnFamilyOptionsFromString(base_cf_opt,
-             "write_buffer_size=13;max_write_buffer_number=;", &new_cf_opt));
+  ASSERT_NOK(GetColumnFamilyOptionsFromString(
+      config_options, base_cf_opt,
+      "write_buffer_size=13;max_write_buffer_number=;", &new_cf_opt));
   ASSERT_OK(RocksDBOptionsParser::VerifyCFOptions(exact, base_cf_opt, new_cf_opt));
 
   // Missing option name
-  ASSERT_NOK(GetColumnFamilyOptionsFromString(base_cf_opt,
-             "write_buffer_size=13; =100;", &new_cf_opt));
+  ASSERT_NOK(GetColumnFamilyOptionsFromString(
+      config_options, base_cf_opt, "write_buffer_size=13; =100;", &new_cf_opt));
   ASSERT_OK(RocksDBOptionsParser::VerifyCFOptions(exact, base_cf_opt, new_cf_opt));
 
   const uint64_t kilo = 1024UL;
@@ -2607,17 +2640,17 @@ TEST_F(OptionsOldApiTest, GetColumnFamilyOptionsFromStringTest) {
 
   // Units (k)
   ASSERT_OK(GetColumnFamilyOptionsFromString(
-      base_cf_opt, "max_write_buffer_number=15K", &new_cf_opt));
+      config_options, base_cf_opt, "max_write_buffer_number=15K", &new_cf_opt));
   ASSERT_EQ(new_cf_opt.max_write_buffer_number, 15 * kilo);
   // Units (m)
-  ASSERT_OK(GetColumnFamilyOptionsFromString(base_cf_opt,
-            "max_write_buffer_number=16m;inplace_update_num_locks=17M",
-            &new_cf_opt));
+  ASSERT_OK(GetColumnFamilyOptionsFromString(
+      config_options, base_cf_opt,
+      "max_write_buffer_number=16m;inplace_update_num_locks=17M", &new_cf_opt));
   ASSERT_EQ(new_cf_opt.max_write_buffer_number, 16 * mega);
   ASSERT_EQ(new_cf_opt.inplace_update_num_locks, 17u * mega);
   // Units (g)
   ASSERT_OK(GetColumnFamilyOptionsFromString(
-      base_cf_opt,
+      config_options, base_cf_opt,
       "write_buffer_size=18g;prefix_extractor=capped:8;"
       "arena_block_size=19G",
       &new_cf_opt));
@@ -2628,107 +2661,119 @@ TEST_F(OptionsOldApiTest, GetColumnFamilyOptionsFromStringTest) {
   ASSERT_EQ(new_cf_opt.prefix_extractor->AsString(), "rocksdb.CappedPrefix.8");
 
   // Units (t)
-  ASSERT_OK(GetColumnFamilyOptionsFromString(base_cf_opt,
-            "write_buffer_size=20t;arena_block_size=21T", &new_cf_opt));
+  ASSERT_OK(GetColumnFamilyOptionsFromString(
+      config_options, base_cf_opt, "write_buffer_size=20t;arena_block_size=21T",
+      &new_cf_opt));
   ASSERT_EQ(new_cf_opt.write_buffer_size, 20 * tera);
   ASSERT_EQ(new_cf_opt.arena_block_size, 21 * tera);
 
   // Nested block based table options
   // Empty
-  ASSERT_OK(GetColumnFamilyOptionsFromString(base_cf_opt,
-            "write_buffer_size=10;max_write_buffer_number=16;"
-            "block_based_table_factory={};arena_block_size=1024",
-            &new_cf_opt));
+  ASSERT_OK(GetColumnFamilyOptionsFromString(
+      config_options, base_cf_opt,
+      "write_buffer_size=10;max_write_buffer_number=16;"
+      "block_based_table_factory={};arena_block_size=1024",
+      &new_cf_opt));
   ASSERT_TRUE(new_cf_opt.table_factory != nullptr);
   // Non-empty
-  ASSERT_OK(GetColumnFamilyOptionsFromString(base_cf_opt,
-            "write_buffer_size=10;max_write_buffer_number=16;"
-            "block_based_table_factory={block_cache=1M;block_size=4;};"
-            "arena_block_size=1024",
-            &new_cf_opt));
+  ASSERT_OK(GetColumnFamilyOptionsFromString(
+      config_options, base_cf_opt,
+      "write_buffer_size=10;max_write_buffer_number=16;"
+      "block_based_table_factory={block_cache=1M;block_size=4;};"
+      "arena_block_size=1024",
+      &new_cf_opt));
   ASSERT_TRUE(new_cf_opt.table_factory != nullptr);
   // Last one
-  ASSERT_OK(GetColumnFamilyOptionsFromString(base_cf_opt,
-            "write_buffer_size=10;max_write_buffer_number=16;"
-            "block_based_table_factory={block_cache=1M;block_size=4;}",
-            &new_cf_opt));
+  ASSERT_OK(GetColumnFamilyOptionsFromString(
+      config_options, base_cf_opt,
+      "write_buffer_size=10;max_write_buffer_number=16;"
+      "block_based_table_factory={block_cache=1M;block_size=4;}",
+      &new_cf_opt));
   ASSERT_TRUE(new_cf_opt.table_factory != nullptr);
   // Mismatch curly braces
-  ASSERT_NOK(GetColumnFamilyOptionsFromString(base_cf_opt,
-             "write_buffer_size=10;max_write_buffer_number=16;"
-             "block_based_table_factory={{{block_size=4;};"
-             "arena_block_size=1024",
-             &new_cf_opt));
+  ASSERT_NOK(GetColumnFamilyOptionsFromString(
+      config_options, base_cf_opt,
+      "write_buffer_size=10;max_write_buffer_number=16;"
+      "block_based_table_factory={{{block_size=4;};"
+      "arena_block_size=1024",
+      &new_cf_opt));
   ASSERT_OK(RocksDBOptionsParser::VerifyCFOptions(exact, base_cf_opt, new_cf_opt));
 
   // Unexpected chars after closing curly brace
-  ASSERT_NOK(GetColumnFamilyOptionsFromString(base_cf_opt,
-             "write_buffer_size=10;max_write_buffer_number=16;"
-             "block_based_table_factory={block_size=4;}};"
-             "arena_block_size=1024",
-             &new_cf_opt));
+  ASSERT_NOK(GetColumnFamilyOptionsFromString(
+      config_options, base_cf_opt,
+      "write_buffer_size=10;max_write_buffer_number=16;"
+      "block_based_table_factory={block_size=4;}};"
+      "arena_block_size=1024",
+      &new_cf_opt));
   ASSERT_OK(RocksDBOptionsParser::VerifyCFOptions(exact, base_cf_opt, new_cf_opt));
 
-  ASSERT_NOK(GetColumnFamilyOptionsFromString(base_cf_opt,
-             "write_buffer_size=10;max_write_buffer_number=16;"
-             "block_based_table_factory={block_size=4;}xdfa;"
-             "arena_block_size=1024",
-             &new_cf_opt));
+  ASSERT_NOK(GetColumnFamilyOptionsFromString(
+      config_options, base_cf_opt,
+      "write_buffer_size=10;max_write_buffer_number=16;"
+      "block_based_table_factory={block_size=4;}xdfa;"
+      "arena_block_size=1024",
+      &new_cf_opt));
   ASSERT_OK(RocksDBOptionsParser::VerifyCFOptions(exact, base_cf_opt, new_cf_opt));
 
-  ASSERT_NOK(GetColumnFamilyOptionsFromString(base_cf_opt,
-             "write_buffer_size=10;max_write_buffer_number=16;"
-             "block_based_table_factory={block_size=4;}xdfa",
-             &new_cf_opt));
+  ASSERT_NOK(GetColumnFamilyOptionsFromString(
+      config_options, base_cf_opt,
+      "write_buffer_size=10;max_write_buffer_number=16;"
+      "block_based_table_factory={block_size=4;}xdfa",
+      &new_cf_opt));
   ASSERT_OK(RocksDBOptionsParser::VerifyCFOptions(exact, base_cf_opt, new_cf_opt));
 
   // Invalid block based table option
-  ASSERT_NOK(GetColumnFamilyOptionsFromString(base_cf_opt,
-             "write_buffer_size=10;max_write_buffer_number=16;"
-             "block_based_table_factory={xx_block_size=4;}",
-             &new_cf_opt));
+  ASSERT_NOK(GetColumnFamilyOptionsFromString(
+      config_options, base_cf_opt,
+      "write_buffer_size=10;max_write_buffer_number=16;"
+      "block_based_table_factory={xx_block_size=4;}",
+      &new_cf_opt));
   ASSERT_OK(RocksDBOptionsParser::VerifyCFOptions(exact, base_cf_opt, new_cf_opt));
 
-  ASSERT_OK(GetColumnFamilyOptionsFromString(base_cf_opt,
-           "optimize_filters_for_hits=true",
-           &new_cf_opt));
-  ASSERT_OK(GetColumnFamilyOptionsFromString(base_cf_opt,
-            "optimize_filters_for_hits=false",
-            &new_cf_opt));
+  ASSERT_OK(GetColumnFamilyOptionsFromString(config_options, base_cf_opt,
+                                             "optimize_filters_for_hits=true",
+                                             &new_cf_opt));
+  ASSERT_OK(GetColumnFamilyOptionsFromString(config_options, base_cf_opt,
+                                             "optimize_filters_for_hits=false",
+                                             &new_cf_opt));
 
-  ASSERT_NOK(GetColumnFamilyOptionsFromString(base_cf_opt,
-              "optimize_filters_for_hits=junk",
-              &new_cf_opt));
+  ASSERT_NOK(GetColumnFamilyOptionsFromString(config_options, base_cf_opt,
+                                              "optimize_filters_for_hits=junk",
+                                              &new_cf_opt));
   ASSERT_OK(RocksDBOptionsParser::VerifyCFOptions(exact, base_cf_opt, new_cf_opt));
 
   // Nested plain table options
   // Empty
-  ASSERT_OK(GetColumnFamilyOptionsFromString(base_cf_opt,
-            "write_buffer_size=10;max_write_buffer_number=16;"
-            "plain_table_factory={};arena_block_size=1024",
-            &new_cf_opt));
+  ASSERT_OK(GetColumnFamilyOptionsFromString(
+      config_options, base_cf_opt,
+      "write_buffer_size=10;max_write_buffer_number=16;"
+      "plain_table_factory={};arena_block_size=1024",
+      &new_cf_opt));
   ASSERT_TRUE(new_cf_opt.table_factory != nullptr);
   ASSERT_EQ(std::string(new_cf_opt.table_factory->Name()), "PlainTable");
   // Non-empty
-  ASSERT_OK(GetColumnFamilyOptionsFromString(base_cf_opt,
-            "write_buffer_size=10;max_write_buffer_number=16;"
-            "plain_table_factory={user_key_len=66;bloom_bits_per_key=20;};"
-            "arena_block_size=1024",
-            &new_cf_opt));
+  ASSERT_OK(GetColumnFamilyOptionsFromString(
+      config_options, base_cf_opt,
+      "write_buffer_size=10;max_write_buffer_number=16;"
+      "plain_table_factory={user_key_len=66;bloom_bits_per_key=20;};"
+      "arena_block_size=1024",
+      &new_cf_opt));
   ASSERT_TRUE(new_cf_opt.table_factory != nullptr);
   ASSERT_EQ(std::string(new_cf_opt.table_factory->Name()), "PlainTable");
 
   // memtable factory
-  ASSERT_OK(GetColumnFamilyOptionsFromString(base_cf_opt,
-            "write_buffer_size=10;max_write_buffer_number=16;"
-            "memtable=skip_list:10;arena_block_size=1024",
-            &new_cf_opt));
+  ASSERT_OK(GetColumnFamilyOptionsFromString(
+      config_options, base_cf_opt,
+      "write_buffer_size=10;max_write_buffer_number=16;"
+      "memtable=skip_list:10;arena_block_size=1024",
+      &new_cf_opt));
   ASSERT_TRUE(new_cf_opt.memtable_factory != nullptr);
   ASSERT_TRUE(new_cf_opt.memtable_factory->IsInstanceOf("SkipListFactory"));
 
   // blob cache
   ASSERT_OK(GetColumnFamilyOptionsFromString(
-      base_cf_opt,
+      config_options, base_cf_opt,
       "blob_cache={capacity=1M;num_shard_bits=4;"
       "strict_capacity_limit=true;high_pri_pool_ratio=0.5;};",
       &new_cf_opt));
@@ -2832,9 +2877,15 @@ TEST_F(OptionsTest, SliceTransformCreateFromString) {
 TEST_F(OptionsOldApiTest, GetBlockBasedTableOptionsFromString) {
   BlockBasedTableOptions table_opt;
   BlockBasedTableOptions new_opt;
+  ConfigOptions config_options;
+  config_options.input_strings_escaped = false;
+  config_options.ignore_unknown_options = false;
+  config_options.invoke_prepare_options = false;
+  config_options.ignore_unsupported_options = false;
+
   // make sure default values are overwritten by something else
   ASSERT_OK(GetBlockBasedTableOptionsFromString(
-      table_opt,
+      config_options, table_opt,
       "cache_index_and_filter_blocks=1;index_type=kHashSearch;"
       "checksum=kxxHash;no_block_cache=1;"
       "block_cache=1M;block_cache_compressed=1k;block_size=1024;"
@@ -2860,54 +2911,57 @@ TEST_F(OptionsOldApiTest, GetBlockBasedTableOptionsFromString) {
   EXPECT_EQ(bfp->GetWholeBitsPerKey(), 5);
 
   // unknown option
-  ASSERT_NOK(GetBlockBasedTableOptionsFromString(table_opt,
-             "cache_index_and_filter_blocks=1;index_type=kBinarySearch;"
-             "bad_option=1",
-             &new_opt));
+  ASSERT_NOK(GetBlockBasedTableOptionsFromString(
+      config_options, table_opt,
+      "cache_index_and_filter_blocks=1;index_type=kBinarySearch;"
+      "bad_option=1",
+      &new_opt));
   ASSERT_EQ(static_cast<bool>(table_opt.cache_index_and_filter_blocks),
             new_opt.cache_index_and_filter_blocks);
   ASSERT_EQ(table_opt.index_type, new_opt.index_type);
 
   // unrecognized index type
-  ASSERT_NOK(GetBlockBasedTableOptionsFromString(table_opt,
-             "cache_index_and_filter_blocks=1;index_type=kBinarySearchXX",
-             &new_opt));
+  ASSERT_NOK(GetBlockBasedTableOptionsFromString(
+      config_options, table_opt,
+      "cache_index_and_filter_blocks=1;index_type=kBinarySearchXX", &new_opt));
   ASSERT_EQ(table_opt.cache_index_and_filter_blocks,
             new_opt.cache_index_and_filter_blocks);
   ASSERT_EQ(table_opt.index_type, new_opt.index_type);
 
   // unrecognized checksum type
-  ASSERT_NOK(GetBlockBasedTableOptionsFromString(table_opt,
-             "cache_index_and_filter_blocks=1;checksum=kxxHashXX",
-             &new_opt));
+  ASSERT_NOK(GetBlockBasedTableOptionsFromString(
+      config_options, table_opt,
+      "cache_index_and_filter_blocks=1;checksum=kxxHashXX", &new_opt));
   ASSERT_EQ(table_opt.cache_index_and_filter_blocks,
             new_opt.cache_index_and_filter_blocks);
   ASSERT_EQ(table_opt.index_type, new_opt.index_type);
 
   // unrecognized filter policy name
-  ASSERT_NOK(GetBlockBasedTableOptionsFromString(table_opt,
-             "cache_index_and_filter_blocks=1;"
-             "filter_policy=bloomfilterxx:4:true",
-             &new_opt));
+  ASSERT_NOK(
+      GetBlockBasedTableOptionsFromString(config_options, table_opt,
+                                          "cache_index_and_filter_blocks=1;"
+                                          "filter_policy=bloomfilterxx:4:true",
+                                          &new_opt));
   ASSERT_EQ(table_opt.cache_index_and_filter_blocks,
             new_opt.cache_index_and_filter_blocks);
   ASSERT_EQ(table_opt.filter_policy, new_opt.filter_policy);
 
   // Used to be rejected, now accepted
   ASSERT_OK(GetBlockBasedTableOptionsFromString(
-      table_opt, "filter_policy=bloomfilter:4", &new_opt));
+      config_options, table_opt, "filter_policy=bloomfilter:4", &new_opt));
   bfp = dynamic_cast<const BloomFilterPolicy*>(new_opt.filter_policy.get());
   EXPECT_EQ(bfp->GetMillibitsPerKey(), 4000);
   EXPECT_EQ(bfp->GetWholeBitsPerKey(), 4);
 
   // Check block cache options are overwritten when specified
   // in new format as a struct.
-  ASSERT_OK(GetBlockBasedTableOptionsFromString(table_opt,
-             "block_cache={capacity=1M;num_shard_bits=4;"
-             "strict_capacity_limit=true;high_pri_pool_ratio=0.5;};"
-             "block_cache_compressed={capacity=1M;num_shard_bits=4;"
-             "strict_capacity_limit=true;high_pri_pool_ratio=0.5;}",
-             &new_opt));
+  ASSERT_OK(GetBlockBasedTableOptionsFromString(
+      config_options, table_opt,
+      "block_cache={capacity=1M;num_shard_bits=4;"
+      "strict_capacity_limit=true;high_pri_pool_ratio=0.5;};"
+      "block_cache_compressed={capacity=1M;num_shard_bits=4;"
+      "strict_capacity_limit=true;high_pri_pool_ratio=0.5;}",
+      &new_opt));
   ASSERT_TRUE(new_opt.block_cache != nullptr);
   ASSERT_EQ(new_opt.block_cache->GetCapacity(), 1024UL*1024UL);
   ASSERT_EQ(std::dynamic_pointer_cast<ShardedCacheBase>(new_opt.block_cache)
@@ -2919,10 +2973,11 @@ TEST_F(OptionsOldApiTest, GetBlockBasedTableOptionsFromString) {
 
   // Set only block cache capacity. Check other values are
   // reset to default values.
-  ASSERT_OK(GetBlockBasedTableOptionsFromString(table_opt,
-             "block_cache={capacity=2M};"
-             "block_cache_compressed={capacity=2M}",
-             &new_opt));
+  ASSERT_OK(GetBlockBasedTableOptionsFromString(
+      config_options, table_opt,
+      "block_cache={capacity=2M};"
+      "block_cache_compressed={capacity=2M}",
+      &new_opt));
   ASSERT_TRUE(new_opt.block_cache != nullptr);
   ASSERT_EQ(new_opt.block_cache->GetCapacity(), 2*1024UL*1024UL);
   // Default values
@@ -2936,7 +2991,7 @@ TEST_F(OptionsOldApiTest, GetBlockBasedTableOptionsFromString) {
 
   // Set couple of block cache options.
   ASSERT_OK(GetBlockBasedTableOptionsFromString(
-      table_opt,
+      config_options, table_opt,
       "block_cache={num_shard_bits=5;high_pri_pool_ratio=0.5;};"
       "block_cache_compressed={num_shard_bits=5;"
       "high_pri_pool_ratio=0.0;}",
@@ -2950,12 +3005,13 @@ TEST_F(OptionsOldApiTest, GetBlockBasedTableOptionsFromString) {
                 new_opt.block_cache)->GetHighPriPoolRatio(), 0.5);
 
   // Set couple of block cache options.
-  ASSERT_OK(GetBlockBasedTableOptionsFromString(table_opt,
-             "block_cache={capacity=1M;num_shard_bits=4;"
-             "strict_capacity_limit=true;};"
-             "block_cache_compressed={capacity=1M;num_shard_bits=4;"
-             "strict_capacity_limit=true;}",
-             &new_opt));
+  ASSERT_OK(GetBlockBasedTableOptionsFromString(
+      config_options, table_opt,
+      "block_cache={capacity=1M;num_shard_bits=4;"
+      "strict_capacity_limit=true;};"
+      "block_cache_compressed={capacity=1M;num_shard_bits=4;"
+      "strict_capacity_limit=true;}",
+      &new_opt));
   ASSERT_TRUE(new_opt.block_cache != nullptr);
   ASSERT_EQ(new_opt.block_cache->GetCapacity(), 1024UL*1024UL);
   ASSERT_EQ(std::dynamic_pointer_cast<ShardedCacheBase>(new_opt.block_cache)
@@ -2971,11 +3027,16 @@ TEST_F(OptionsOldApiTest, GetPlainTableOptionsFromString) {
   PlainTableOptions table_opt;
   PlainTableOptions new_opt;
   // make sure default values are overwritten by something else
-  ASSERT_OK(GetPlainTableOptionsFromString(table_opt,
-            "user_key_len=66;bloom_bits_per_key=20;hash_table_ratio=0.5;"
-            "index_sparseness=8;huge_page_tlb_size=4;encoding_type=kPrefix;"
-            "full_scan_mode=true;store_index_in_file=true",
-            &new_opt));
+  ConfigOptions config_options_from_string;
+  config_options_from_string.input_strings_escaped = false;
+  config_options_from_string.ignore_unknown_options = false;
+  config_options_from_string.invoke_prepare_options = false;
+  ASSERT_OK(GetPlainTableOptionsFromString(
+      config_options_from_string, table_opt,
+      "user_key_len=66;bloom_bits_per_key=20;hash_table_ratio=0.5;"
+      "index_sparseness=8;huge_page_tlb_size=4;encoding_type=kPrefix;"
+      "full_scan_mode=true;store_index_in_file=true",
+      &new_opt));
   ASSERT_EQ(new_opt.user_key_len, 66u);
   ASSERT_EQ(new_opt.bloom_bits_per_key, 20);
   ASSERT_EQ(new_opt.hash_table_ratio, 0.5);
@@ -2988,22 +3049,28 @@ TEST_F(OptionsOldApiTest, GetPlainTableOptionsFromString) {
   std::unordered_map<std::string, std::string> opt_map;
   ASSERT_OK(StringToMap(
       "user_key_len=55;bloom_bits_per_key=10;huge_page_tlb_size=8;", &opt_map));
-  ASSERT_OK(GetPlainTableOptionsFromMap(table_opt, opt_map, &new_opt));
+  ConfigOptions config_options_from_map;
+  config_options_from_map.input_strings_escaped = false;
+  config_options_from_map.ignore_unknown_options = false;
+  ASSERT_OK(GetPlainTableOptionsFromMap(config_options_from_map, table_opt,
+                                        opt_map, &new_opt));
   ASSERT_EQ(new_opt.user_key_len, 55u);
   ASSERT_EQ(new_opt.bloom_bits_per_key, 10);
   ASSERT_EQ(new_opt.huge_page_tlb_size, 8);
 
   // unknown option
-  ASSERT_NOK(GetPlainTableOptionsFromString(table_opt,
-             "user_key_len=66;bloom_bits_per_key=20;hash_table_ratio=0.5;"
-             "bad_option=1",
-             &new_opt));
+  ASSERT_NOK(GetPlainTableOptionsFromString(
+      config_options_from_string, table_opt,
+      "user_key_len=66;bloom_bits_per_key=20;hash_table_ratio=0.5;"
+      "bad_option=1",
+      &new_opt));
 
   // unrecognized EncodingType
-  ASSERT_NOK(GetPlainTableOptionsFromString(table_opt,
-             "user_key_len=66;bloom_bits_per_key=20;hash_table_ratio=0.5;"
-             "encoding_type=kPrefixXX",
-             &new_opt));
+  ASSERT_NOK(GetPlainTableOptionsFromString(
+      config_options_from_string, table_opt,
+      "user_key_len=66;bloom_bits_per_key=20;hash_table_ratio=0.5;"
+      "encoding_type=kPrefixXX",
+      &new_opt));
 }
 
 TEST_F(OptionsOldApiTest, GetOptionsFromStringTest) {
@@ -3085,10 +3152,15 @@ TEST_F(OptionsOldApiTest, DBOptionsSerialization) {
 
   // Phase 3: Set new_options from the derived string and expect
   //          new_options == base_options
-  ASSERT_OK(GetDBOptionsFromString(DBOptions(), base_options_file_content,
-                                   &new_options));
-  ConfigOptions config_options;
-  ASSERT_OK(RocksDBOptionsParser::VerifyDBOptions(config_options, base_options, new_options));
+  const DBOptions base_db_options;
+  ConfigOptions db_config_options(base_db_options);
+  db_config_options.input_strings_escaped = false;
+  db_config_options.ignore_unknown_options = false;
+  ASSERT_OK(GetDBOptionsFromString(db_config_options, base_db_options,
+                                   base_options_file_content, &new_options));
+  ConfigOptions verify_db_config_options;
+  ASSERT_OK(RocksDBOptionsParser::VerifyDBOptions(verify_db_config_options,
+                                                  base_options, new_options));
 }
 
 TEST_F(OptionsOldApiTest, ColumnFamilyOptionsSerialization) {
@@ -3106,10 +3178,15 @@ TEST_F(OptionsOldApiTest, ColumnFamilyOptionsSerialization) {
 
   // Phase 3: Set new_opt from the derived string and expect
   //          new_opt == base_opt
-  ASSERT_OK(GetColumnFamilyOptionsFromString(
-      ColumnFamilyOptions(), base_options_file_content, &new_opt));
-  ConfigOptions config_options;
-  ASSERT_OK(RocksDBOptionsParser::VerifyCFOptions(config_options, base_opt, new_opt));
+  ConfigOptions cf_config_options;
+  cf_config_options.input_strings_escaped = false;
+  cf_config_options.ignore_unknown_options = false;
+  ASSERT_OK(
+      GetColumnFamilyOptionsFromString(cf_config_options, ColumnFamilyOptions(),
+                                       base_options_file_content, &new_opt));
+  ConfigOptions verify_cf_config_options;
+  ASSERT_OK(RocksDBOptionsParser::VerifyCFOptions(verify_cf_config_options,
+                                                  base_opt, new_opt));
   if (base_opt.compaction_filter) {
     delete base_opt.compaction_filter;
   }

port/port_posix.h

@@ -13,7 +13,7 @@
 
 #include <thread>
 
-#include "rocksdb/options.h"
+#include "rocksdb/port_defs.h"
 #include "rocksdb/rocksdb_namespace.h"
 
 // size_t printf formatting named in the manner of C99 standard formatting

port/stack_trace.cc

@@ -31,12 +31,18 @@ void* SaveStack(int* /*num_frames*/, int /*first_frames_to_skip*/) {
 #include <string.h>
 #include <unistd.h>
 
-#if defined(OS_FREEBSD)
+#ifdef OS_FREEBSD
 #include <sys/sysctl.h>
-#endif
+#endif  // OS_FREEBSD
 #ifdef OS_LINUX
 #include <sys/prctl.h>
-#endif
+#include <sys/types.h>
+#include <sys/wait.h>
+#if __GLIBC__ < 2 || (__GLIBC__ == 2 && __GLIBC_MINOR__ < 30)
+#include <sys/syscall.h>
+#define gettid() syscall(SYS_gettid)
+#endif  // GLIBC version
+#endif  // OS_LINUX
 
 #include "port/lang.h"
 
@@ -135,6 +141,80 @@ void PrintStack(void* frames[], int num_frames) {
 }
 
 void PrintStack(int first_frames_to_skip) {
+#if defined(ROCKSDB_DLL) && defined(OS_LINUX)
+  // LIB_MODE=shared build produces mediocre information from the above
+  // backtrace+addr2line stack trace method. Try to use GDB in that case, but
+  // only on Linux where we know how to attach to a particular thread.
+  bool linux_dll = true;
+#else
+  bool linux_dll = false;
+#endif
+  // Also support invoking interactive debugger on stack trace, with this
+  // envvar set to non-empty
+  char* debug_env = getenv("ROCKSDB_DEBUG");
+  bool debug = debug_env != nullptr && strlen(debug_env) > 0;
+
+  if (linux_dll || debug) {
+    // Allow ouside debugger to attach, even with Yama security restrictions
+#ifdef PR_SET_PTRACER_ANY
+    (void)prctl(PR_SET_PTRACER, PR_SET_PTRACER_ANY, 0, 0, 0);
+#endif
+    // Try to invoke GDB, either for stack trace or debugging.
+    long long attach_id = getpid();
+
+    // `gdb -p PID` seems to always attach to main thread, but `gdb -p TID`
+    // seems to be able to attach to a particular thread in a process, which
+    // makes sense as the main thread TID == PID of the process.
+    // But I haven't found that gdb capability documented anywhere, so leave
+    // a back door to attach to main thread.
+#ifdef OS_LINUX
+    if (getenv("ROCKSDB_DEBUG_USE_PID") == nullptr) {
+      attach_id = gettid();
+    }
+#endif
+    char attach_id_str[20];
+    snprintf(attach_id_str, sizeof(attach_id_str), "%lld", attach_id);
+    pid_t child_pid = fork();
+    if (child_pid == 0) {
+      // child process
+      if (debug) {
+        fprintf(stderr, "Invoking GDB for debugging (ROCKSDB_DEBUG=%s)...\n",
+                debug_env);
+        execlp(/*cmd in PATH*/ "gdb", /*arg0*/ "gdb", "-p", attach_id_str,
+               (char*)nullptr);
+        return;
+      } else {
+        fprintf(stderr, "Invoking GDB for stack trace...\n");
+
+        // Skip top ~4 frames here in PrintStack
+        // See https://stackoverflow.com/q/40991943/454544
+        auto bt_in_gdb =
+            "frame apply level 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 "
+            "21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 "
+            "42 43 44 -q frame";
+        // Redirect child stdout to original stderr
+        dup2(2, 1);
+        // No child stdin (don't use pager)
+        close(0);
+        // -n : Loading config files can apparently cause failures with the
+        // other options here.
+        // -batch : non-interactive; suppress banners as much as possible
+        execlp(/*cmd in PATH*/ "gdb", /*arg0*/ "gdb", "-n", "-batch", "-p",
+               attach_id_str, "-ex", bt_in_gdb, (char*)nullptr);
+        return;
+      }
+    } else {
+      // parent process; wait for child
+      int wstatus;
+      waitpid(child_pid, &wstatus, 0);
+      if (WIFEXITED(wstatus) && WEXITSTATUS(wstatus) == EXIT_SUCCESS) {
+        // Good
+        return;
+      }
+    }
+    fprintf(stderr, "GDB failed; falling back on backtrace+addr2line...\n");
+  }
+
   const int kMaxFrames = 100;
   void* frames[kMaxFrames];
 
@@ -189,7 +269,9 @@ void InstallStackTraceHandler() {
   signal(SIGSEGV, StackTraceHandler);
   signal(SIGBUS, StackTraceHandler);
   signal(SIGABRT, StackTraceHandler);
-  // Allow ouside debugger to attach, even with Yama security restrictions
+  // Allow ouside debugger to attach, even with Yama security restrictions.
+  // This is needed even outside of PrintStack() so that external mechanisms
+  // can dump stacks if they suspect that a test has hung.
 #ifdef PR_SET_PTRACER_ANY
   (void)prctl(PR_SET_PTRACER, PR_SET_PTRACER_ANY, 0, 0, 0);
 #endif

port/win/env_win.h

@@ -97,7 +97,7 @@ class WinClock : public SystemClock {
 
   Status GetCurrentTime(int64_t* unix_time) override;
   // Converts seconds-since-Jan-01-1970 to a printable string
-  virtual std::string TimeToString(uint64_t time);
+  std::string TimeToString(uint64_t time) override;
 
   uint64_t GetPerfCounterFrequency() const { return perf_counter_frequency_; }
 
@@ -116,7 +116,7 @@ class WinFileSystem : public FileSystem {
   ~WinFileSystem() {}
   static const char* kClassName() { return "WinFS"; }
   const char* Name() const override { return kClassName(); }
-  const char* NickName() const { return kDefaultName(); }
+  const char* NickName() const override { return kDefaultName(); }
 
   static size_t GetSectorSize(const std::string& fname);
   size_t GetPageSize() const { return page_size_; }

port/win/port_win.h

@@ -17,21 +17,22 @@
 #endif
 
 #include <windows.h>
+//^^ <windows.h> should be included first before other system lib
+#include <intrin.h>
+#include <malloc.h>
+#include <process.h>
+#include <stdint.h>
+#include <string.h>
+
+#include <cassert>
+#include <condition_variable>
+#include <limits>
+#include <mutex>
 #include <string>
 #include <thread>
-#include <string.h>
-#include <mutex>
-#include <limits>
-#include <condition_variable>
-#include <malloc.h>
-#include <intrin.h>
-#include <process.h>
-
-#include <stdint.h>
 
 #include "port/win/win_thread.h"
-
-#include "rocksdb/options.h"
+#include "rocksdb/port_defs.h"
 
 #undef min
 #undef max

src.mk

@@ -234,6 +234,7 @@ LIB_SOURCES =                                                   \
   util/concurrent_task_limiter_impl.cc                          \
   util/crc32c.cc                                                \
   util/crc32c_arm64.cc                                          \
+  util/data_structure.cc                                        \
   util/dynamic_bloom.cc                                         \
   util/hash.cc                                                  \
   util/murmurhash.cc                                            \

table/block_based/block.h

@@ -16,6 +16,7 @@
 
 #include "db/pinned_iterators_manager.h"
 #include "port/malloc.h"
+#include "rocksdb/advanced_cache.h"
 #include "rocksdb/iterator.h"
 #include "rocksdb/options.h"
 #include "rocksdb/statistics.h"

table/block_based/block_based_table_factory.cc

@@ -909,18 +909,6 @@ Status BlockBasedTableFactory::ParseOption(const ConfigOptions& config_options,
   return status;
 }
 
-Status GetBlockBasedTableOptionsFromString(
-    const BlockBasedTableOptions& table_options, const std::string& opts_str,
-    BlockBasedTableOptions* new_table_options) {
-  ConfigOptions config_options;
-  config_options.input_strings_escaped = false;
-  config_options.ignore_unknown_options = false;
-  config_options.invoke_prepare_options = false;
-  config_options.ignore_unsupported_options = false;
-
-  return GetBlockBasedTableOptionsFromString(config_options, table_options,
-                                             opts_str, new_table_options);
-}
 Status GetBlockBasedTableOptionsFromString(
     const ConfigOptions& config_options,
     const BlockBasedTableOptions& table_options, const std::string& opts_str,
@@ -940,20 +928,6 @@ Status GetBlockBasedTableOptionsFromString(
   }
 }
 
-Status GetBlockBasedTableOptionsFromMap(
-    const BlockBasedTableOptions& table_options,
-    const std::unordered_map<std::string, std::string>& opts_map,
-    BlockBasedTableOptions* new_table_options, bool input_strings_escaped,
-    bool ignore_unknown_options) {
-  ConfigOptions config_options;
-  config_options.input_strings_escaped = input_strings_escaped;
-  config_options.ignore_unknown_options = ignore_unknown_options;
-  config_options.invoke_prepare_options = false;
-
-  return GetBlockBasedTableOptionsFromMap(config_options, table_options,
-                                          opts_map, new_table_options);
-}
-
 Status GetBlockBasedTableOptionsFromMap(
     const ConfigOptions& config_options,
     const BlockBasedTableOptions& table_options,

table/block_based/block_based_table_reader.cc

@@ -1929,6 +1929,16 @@ Status BlockBasedTable::ApproximateKeyAnchors(const ReadOptions& read_options,
   // likely not to be a problem. We are compacting the whole file, so all
   // keys will be read out anyway. An extra read to index block might be
   // a small share of the overhead. We can try to optimize if needed.
+  //
+  // `CacheDependencies()` brings all the blocks into cache using one I/O. That
+  // way the full index scan usually finds the index data it is looking for in
+  // cache rather than doing an I/O for each "dependency" (partition).
+  Status s =
+      rep_->index_reader->CacheDependencies(read_options, false /* pin */);
+  if (!s.ok()) {
+    return s;
+  }
+
   IndexBlockIter iiter_on_stack;
   auto iiter = NewIndexIterator(
       read_options, /*disable_prefix_seek=*/false, &iiter_on_stack,

table/block_based/block_based_table_reader_test.cc

@@ -253,7 +253,7 @@ TEST_P(BlockBasedTableReaderTest, MultiGet) {
                              nullptr, nullptr, nullptr, nullptr,
                              true /* do_merge */, nullptr, nullptr, nullptr,
                              nullptr, nullptr, nullptr);
-    key_context.emplace_back(nullptr, keys[i], &values[i], nullptr,
+    key_context.emplace_back(nullptr, keys[i], &values[i], nullptr, nullptr,
                              &statuses.back());
     key_context.back().get_context = &get_context.back();
   }

table/block_based/cachable_entry.h

@@ -13,7 +13,7 @@
 #include <type_traits>
 
 #include "port/likely.h"
-#include "rocksdb/cache.h"
+#include "rocksdb/advanced_cache.h"
 #include "rocksdb/cleanable.h"
 
 namespace ROCKSDB_NAMESPACE {