Fix win build

Summary: Fixing error with win build where we compare int64_t with size_t. Test Plan: make check Reviewers: andrewkr Reviewed By: andrewkr Subscribers: andrewkr, dhruba, leveldb Differential Revision: https://reviews.facebook.net/D57885
Fix valgrind (DBIteratorTest.ReadAhead)
2026-11-06 11:02:36 -05:00 · 2016-05-09 13:43:05 -07:00 · 2016-05-09 13:42:56 -07:00 · 2016-05-06 11:39:26 -07:00 · 2016-05-06 10:18:34 -07:00 · 2016-05-06 10:18:03 -07:00
725 changed files with 28735 additions and 18917 deletions
@@ -42,6 +42,7 @@ unity.a
 tags
 rocksdb_dump
 rocksdb_undump
+db_test2

 java/out
 java/target
@@ -1,24 +1,29 @@
 sudo: false
 language: cpp
+os:
+  - linux
+  - osx
+compiler:
+  - clang

-matrix:
-  include:
-    - os: linux
-      compiler: clang
-      env: COMPILER=clang++-3.6
-      addons:
-         apt:
-            sources: ['ubuntu-toolchain-r-test', 'llvm-toolchain-precise-3.6']
-            packages: ['clang-3.6', 'clang-format-3.6', 'zlib1g-dev', 'libbz2-dev', 'libsnappy-dev', 'curl']
-    - os: osx
-      compiler: clang
+addons:
+   apt:
+      sources: ['ubuntu-toolchain-r-test', 'llvm-toolchain-precise-3.6']
+      packages: ['clang-3.6' , 'zlib1g-dev', 'libbz2-dev', 'libsnappy-dev', 'curl']
+env:
+  # Run all tests before db_block_cache_test (db_test, db_test2)
+  - JOB_NAME=unittests ROCKSDBTESTS_END=db_block_cache_test
+  # Run all tests starting from db_block_cache_test (db_block_cache_test, db_iter_test, ...)
+  - JOB_NAME=unittests ROCKSDBTESTS_START=db_block_cache_test
+  # Run java tests
+  - JOB_NAME=java_test
+  # Build ROCKSDB_LITE
+  - JOB_NAME=lite_build

 install:
  # Build gflags
  # TODO(noetzli): Remove when gflags available through Travis
  - pushd /tmp/ && curl -L https://github.com/gflags/gflags/archive/v2.1.2.tar.gz -o gflags.tar.gz && tar xfz gflags.tar.gz && cd gflags-2.1.2 && cmake . && make && popd
-  # Download clang-format-diff.py to check source code formatting
-  - pushd /tmp/ && curl -L http://llvm.org/svn/llvm-project/cfe/trunk/tools/clang-format/clang-format-diff.py -o clang-format-diff.py && chmod +x clang-format-diff.py && popd

 before_script:
  # Add gflags to include/library paths
@@ -26,16 +31,15 @@ before_script:
  - export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/tmp/gflags-2.1.2/lib"
  - export LIBRARY_PATH="$LIBRARY_PATH:/tmp/gflags-2.1.2/lib"
  - export CPLUS_INCLUDE_PATH="$CPLUS_INCLUDE_PATH:/tmp/gflags-2.1.2/include"
-  - if [ -n "${COMPILER}" ]; then CXX=${COMPILER}; fi
+  - if [[ "${TRAVIS_OS_NAME}" == 'linux' ]]; then CXX=clang++-3.6; fi
  - if [[ "${TRAVIS_OS_NAME}" == 'osx' ]]; then brew install gflags snappy; fi
+  # Limit the maximum number of open file descriptors to 2000
  - ulimit -n 2000 || true

-# Lousy hack to disable use and testing of fallocate, which doesn't behave quite
-# as EnvPosixTest::AllocateTest expects within the Travis OpenVZ environment.
 script:
-  - if [[ "${TRAVIS_OS_NAME}" == 'linux' ]]; then OPT=-DTRAVIS CLANG_FORMAT_DIFF=/tmp/clang-format-diff.py make format || true; fi
-  - OPT=-DTRAVIS V=1 make -j4 check && OPT=-DTRAVIS V=1 make clean jclean rocksdbjava jtest && make clean && OPT="-DTRAVIS -DROCKSDB_LITE" V=1 make -j4 check
-
+  - if [[ "${JOB_NAME}" == 'unittests' ]]; then OPT=-DTRAVIS V=1 make -j4 check_some; fi
+  - if [[ "${JOB_NAME}" == 'java_test' ]]; then OPT=-DTRAVIS V=1 make clean jclean rocksdbjava jtest; fi
+  - if [[ "${JOB_NAME}" == 'lite_build' ]]; then OPT="-DTRAVIS -DROCKSDB_LITE" V=1 make -j4 static_lib; fi
 notifications:
    email:
      - leveldb@fb.com
@@ -13,8 +13,8 @@
 #        cd build
 # 3. Run cmake to generate project files for Windows, add more options to enable required third-party libraries.
 #    See thirdparty.inc for more information.
-#        sample command: cmake -G "Visual Studio 12 Win64" -DGFLAGS=1 -DSNAPPY=1 -DJEMALLOC=1 ..
-#        OR for VS Studio 15 cmake -G "Visual Studio 14 Win64" -DGFLAGS=1 -DSNAPPY=1 -DJEMALLOC=1 ..
+#        sample command: cmake -G "Visual Studio 12 Win64" -DGFLAGS=1 -DSNAPPY=1 -DJEMALLOC=1 -DJNI=1 ..
+#        OR for VS Studio 15 cmake -G "Visual Studio 14 Win64" -DGFLAGS=1 -DSNAPPY=1 -DJEMALLOC=1 -DJNI=1 ..
 # 4. Then build the project in debug mode (you may want to add /m[:<N>] flag to run msbuild in <N> parallel threads
 #                                          or simply /m ot use all avail cores)
 #        msbuild rocksdb.sln
@@ -30,10 +30,9 @@ cmake_minimum_required(VERSION 2.6)
 project(rocksdb)

 include(${CMAKE_CURRENT_SOURCE_DIR}/thirdparty.inc)
-
-execute_process(COMMAND $ENV{COMSPEC} " /C date /T" OUTPUT_VARIABLE DATE)
-execute_process(COMMAND $ENV{COMSPEC} " /C time /T" OUTPUT_VARIABLE TIME)
-string(REGEX REPLACE "(..)/(..)/..(..).*" "\\1/\\2/\\3" DATE ${DATE})
+execute_process(COMMAND powershell -Command "Get-Date -format MM_dd_yyyy" OUTPUT_VARIABLE DATE)
+execute_process(COMMAND powershell -Command "Get-Date -format HH:mm:ss" OUTPUT_VARIABLE TIME)
+string(REGEX REPLACE "(..)_(..)_..(..).*" "\\1/\\2/\\3" DATE ${DATE})
 string(REGEX REPLACE "(..):(.....).*" " \\1:\\2" TIME ${TIME})
 string(CONCAT GIT_DATE_TIME ${DATE} ${TIME})

@@ -83,7 +82,7 @@ set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} /Oxt /Zp8 /Gm- /Gy /MD")
 set(CMAKE_SHARED_LINKER_FLAGS "${CMAKE_SHARED_LINKER_FLAGS} /DEBUG")
 set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} /DEBUG")

-add_definitions(-DWIN32 -DOS_WIN -D_MBCS -DWIN64)
+add_definitions(-DWIN32 -DOS_WIN -D_MBCS -DWIN64 -DNOMINMAX)

 include_directories(${PROJECT_SOURCE_DIR})
 include_directories(${PROJECT_SOURCE_DIR}/include)
@@ -91,7 +90,7 @@ include_directories(${PROJECT_SOURCE_DIR}/third-party/gtest-1.7.0/fused-src)

 set(ROCKSDB_LIBS rocksdblib${ARTIFACT_SUFFIX})
 set(THIRDPARTY_LIBS ${THIRDPARTY_LIBS} gtest)
-set(SYSTEM_LIBS Shlwapi.lib Rpcrt4.lib)
+set(SYSTEM_LIBS ${SYSTEM_LIBS} Shlwapi.lib Rpcrt4.lib)

 set(LIBS ${ROCKSDB_LIBS} ${THIRDPARTY_LIBS} ${SYSTEM_LIBS})

@@ -99,6 +98,7 @@ add_subdirectory(third-party/gtest-1.7.0/fused-src/gtest)

 # Main library source code
 set(SOURCES
+        db/auto_roll_logger.cc
        db/builder.cc
        db/c.cc
        db/column_family.cc
@@ -114,6 +114,7 @@ set(SOURCES
        db/db_impl_debug.cc
        db/db_impl_experimental.cc
        db/db_impl_readonly.cc
+        db/db_info_dumper.cc
        db/db_iter.cc
        db/event_helpers.cc
        db/experimental.cc
@@ -145,13 +146,17 @@ set(SOURCES
        db/write_batch_base.cc
        db/write_controller.cc
        db/write_thread.cc
+        db/xfunc_test_points.cc
        memtable/hash_cuckoo_rep.cc
        memtable/hash_linklist_rep.cc
        memtable/hash_skiplist_rep.cc
+        memtable/skiplistrep.cc
+        memtable/vectorrep.cc
        port/stack_trace.cc
        port/win/env_win.cc
        port/win/port_win.cc
        port/win/win_logger.cc
+        port/win/xpress_win.cc
        table/adaptive_table_factory.cc
        table/block.cc
        table/block_based_filter_block.cc
@@ -173,7 +178,6 @@ set(SOURCES
        table/merger.cc
        table/sst_file_writer.cc
        table/meta_blocks.cc
-        table/mock_table.cc
        table/plain_table_builder.cc
        table/plain_table_factory.cc
        table/plain_table_index.cc
@@ -182,9 +186,9 @@ set(SOURCES
        table/table_properties.cc
        table/two_level_iterator.cc
        tools/sst_dump_tool.cc
+        tools/db_bench_tool.cc
        tools/dump/db_dump_tool.cc
        util/arena.cc
-        util/auto_roll_logger.cc
        util/bloom.cc
        util/build_version.cc
        util/cache.cc
@@ -193,17 +197,18 @@ set(SOURCES
        util/comparator.cc
        util/concurrent_arena.cc
        util/crc32c.cc
-        util/db_info_dumper.cc
-        util/delete_scheduler_impl.cc
+        util/delete_scheduler.cc
        util/dynamic_bloom.cc
        util/env.cc
        util/env_hdfs.cc
        util/event_logger.cc
        util/file_util.cc
        util/file_reader_writer.cc
+        util/sst_file_manager_impl.cc
        util/filter_policy.cc
        util/hash.cc
        util/histogram.cc
+        util/histogram_windowing.cc
        util/instrumented_mutex.cc
        util/iostats_context.cc
        tools/ldb_cmd.cc
@@ -211,7 +216,6 @@ set(SOURCES
        util/logging.cc
        util/log_buffer.cc
        util/memenv.cc
-        util/mock_env.cc
        util/murmurhash.cc
        util/mutable_cf_options.cc
        util/options.cc
@@ -223,7 +227,6 @@ set(SOURCES
        util/perf_level.cc
        util/random.cc
        util/rate_limiter.cc
-        util/skiplistrep.cc
        util/slice.cc
        util/statistics.cc
        util/status.cc
@@ -237,11 +240,12 @@ set(SOURCES
        util/thread_status_updater.cc
        util/thread_status_util.cc
        util/thread_status_util_debug.cc
-        util/vectorrep.cc
+        util/transaction_test_util.cc
        util/xfunc.cc
        util/xxhash.cc
        utilities/backupable/backupable_db.cc
        utilities/checkpoint/checkpoint.cc
+        utilities/compaction_filters/remove_emptyvalue_compactionfilter.cc
        utilities/document/document_db.cc
        utilities/document/json_document.cc
        utilities/document/json_document_builder.cc
@@ -275,6 +279,8 @@ set(SOURCES
 # and linked to tests. Add test only code that is not #ifdefed for Release here.
 set(TESTUTIL_SOURCE
    db/db_test_util.cc
+    table/mock_table.cc
+    util/mock_env.cc
    util/thread_status_updater_debug.cc
 )

@@ -287,13 +293,25 @@ set_target_properties(rocksdb${ARTIFACT_SUFFIX} PROPERTIES COMPILE_FLAGS "-DROCK
 add_dependencies(rocksdb${ARTIFACT_SUFFIX} GenerateBuildVersion)
 target_link_libraries(rocksdb${ARTIFACT_SUFFIX} ${LIBS})

+if (DEFINED JNI)
+  if (${JNI} EQUAL 1)
+    message(STATUS "JNI library is enabled")
+    add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/java)
+  else()
+    message(STATUS "JNI library is disabled")
+  endif()
+else()
+  message(STATUS "JNI library is disabled")
+endif()
+
 set(APPS
-        db/db_bench.cc
+        tools/db_bench.cc
        db/memtablerep_bench.cc
        table/table_reader_bench.cc
        tools/db_stress.cc
        tools/write_stress.cc
        tools/db_repl_stress.cc
+        tools/ldb.cc
        tools/sst_dump.cc
        tools/dump/rocksdb_dump.cc
        tools/dump/rocksdb_undump.cc
@@ -303,6 +321,7 @@ set(APPS
 set(C_TESTS db/c_test.c)

 set(TESTS
+        db/auto_roll_logger_test.cc
        db/column_family_test.cc
        db/compact_files_test.cc
        db/compaction_iterator_test.cc
@@ -312,17 +331,23 @@ set(TESTS
        db/comparator_db_test.cc
        db/corruption_test.cc
        db/cuckoo_table_db_test.cc
-        db/db_iter_test.cc
-        db/db_test.cc
        db/db_compaction_filter_test.cc
        db/db_compaction_test.cc
        db/db_dynamic_level_test.cc
        db/db_inplace_update_test.cc
+        db/db_iter_test.cc
        db/db_log_iter_test.cc
+        db/db_properties_test.cc
+        db/db_table_properties_test.cc
+        db/db_tailing_iter_test.cc
+        db/db_test.cc
+        db/db_test2.cc
+        db/db_block_cache_test.cc
+        db/db_bloom_filter_test.cc
+        db/db_iterator_test.cc
+        db/db_sst_test.cc
        db/db_universal_compaction_test.cc
        db/db_wal_test.cc
-        db/db_tailing_iter_test.cc
-        db/db_table_properties_test.cc
        db/dbformat_test.cc
        db/deletefile_test.cc
        db/fault_injection_test.cc
@@ -340,6 +365,7 @@ set(TESTS
        db/perf_context_test.cc
        db/plain_table_db_test.cc
        db/prefix_test.cc
+        db/repair_test.cc
        db/skiplist_test.cc
        db/table_properties_collector_test.cc
        db/version_builder_test.cc
@@ -363,7 +389,6 @@ set(TESTS
        tools/sst_dump_test.cc
        util/arena_test.cc
        util/autovector_test.cc
-        util/auto_roll_logger_test.cc
        util/bloom_test.cc
        util/cache_test.cc
        util/coding_test.cc
@@ -376,8 +401,10 @@ set(TESTS
        util/file_reader_writer_test.cc
        util/heap_test.cc
        util/histogram_test.cc
+        util/iostats_context_test.cc
        util/memenv_test.cc
        util/mock_env_test.cc
+        util/options_settable_test.cc
        util/options_test.cc
        util/rate_limiter_test.cc
        util/slice_transform_test.cc
@@ -0,0 +1,14 @@
+# RocksDB default options change log
+## 4.8.0 (5/2/2016)
+* options.max_open_files changes from 5000 to -1. It improves performance, but users need to set file descriptor limit to be large enough and watch memory usage for index and bloom filters.
+* options.base_background_compactions changes from max_background_compactions to 1. When users set higher max_background_compactions but the write throughput is not high, the writes are less spiky to disks.
+* options.wal_recovery_mode changes from kTolerateCorruptedTailRecords to kPointInTimeRecovery. Avoid some false positive when file system or hardware reorder the writes for file data and metadata.
+
+## 4.7.0 (4/8/2016)
+* options.write_buffer_size changes from 4MB to 64MB.
+* options.target_file_size_base changes from 2MB to 64MB.
+* options.max_bytes_for_level_base changes from 10MB to 256MB.
+* options.soft_pending_compaction_bytes_limit changes from 0 (disabled) to 64GB.
+* options.hard_pending_compaction_bytes_limit changes from 0 (disabled) to 256GB.
+* table_cache_numshardbits changes from 4 to 6.
+* max_file_opening_threads changes from 1 to 16.
@@ -1,4 +1,38 @@
 # Rocksdb Change Log
+## 4.8.0 (5/2/2016)
+### Public API Change
+* Allow preset compression dictionary for improved compression of block-based tables. This is supported for zlib, zstd, and lz4. The compression dictionary's size is configurable via CompressionOptions::max_dict_bytes.
+* Delete deprecated classes for creating backups (BackupableDB) and restoring from backups (RestoreBackupableDB). Now, BackupEngine should be used for creating backups, and BackupEngineReadOnly should be used for restorations. For more details, see https://github.com/facebook/rocksdb/wiki/How-to-backup-RocksDB%3F
+* Expose estimate of per-level compression ratio via DB property: "rocksdb.compression-ratio-at-levelN".
+* Added EventListener::OnTableFileCreationStarted. EventListener::OnTableFileCreated will be called on failure case. User can check creation status via TableFileCreationInfo::status.
+### New Features
+* Add ReadOptions::readahead_size. If non-zero, NewIterator will create a new table reader which performs reads of the given size.
+
+## 4.7.0 (4/8/2016)
+### Public API Change
+* rename options compaction_measure_io_stats to report_bg_io_stats and include flush too.
+* Change some default options. Now default options will optimize for server-workloads. Also enable slowdown and full stop triggers for pending compaction bytes. These changes may cause sub-optimal performance or significant increase of resource usage. To avoid these risks, users can open existing RocksDB with options extracted from RocksDB option files. See https://github.com/facebook/rocksdb/wiki/RocksDB-Options-File for how to use RocksDB option files. Or you can call Options.OldDefaults() to recover old defaults. DEFAULT_OPTIONS_HISTORY.md will track change history of default options.
+
+## 4.6.0 (3/10/2016)
+### Public API Changes
+* Change default of BlockBasedTableOptions.format_version to 2. It means default DB created by 4.6 or up cannot be opened by RocksDB version 3.9 or earlier.
+* Added strict_capacity_limit option to NewLRUCache. If the flag is set to true, insert to cache will fail if no enough capacity can be free. Signiture of Cache::Insert() is updated accordingly.
+* Tickers [NUMBER_DB_NEXT, NUMBER_DB_PREV, NUMBER_DB_NEXT_FOUND, NUMBER_DB_PREV_FOUND, ITER_BYTES_READ] are not updated immediately. The are updated when the Iterator is deleted.  
+* Add monotonically increasing counter (DB property "rocksdb.current-super-version-number") that increments upon any change to the LSM tree.
+### New Features
+* Add CompactionPri::kMinOverlappingRatio, a compaction picking mode friendly to write amplification.
+* Deprecate Iterator::IsKeyPinned() and replace it with Iterator::GetProperty() with prop_name="rocksdb.iterator.is.key.pinned"
+
+## 4.5.0 (2/5/2016)
+### Public API Changes
+* Add a new perf context level between kEnableCount and kEnableTime. Level 2 now does not include timers for mutexes.
+* Statistics of mutex operation durations will not be measured by default. If you want to have them enabled, you need to set Statistics::stats_level_ to kAll.
+* DBOptions::delete_scheduler and NewDeleteScheduler() are removed, please use DBOptions::sst_file_manager and NewSstFileManager() instead
+
+### New Features
+* ldb tool now supports operations to non-default column families.
+* Add kPersistedTier to ReadTier.  This option allows Get and MultiGet to read only the persited data and skip mem-tables if writes were done with disableWAL = true.
+* Add DBOptions::sst_file_manager. Use NewSstFileManager() in include/rocksdb/sst_file_manager.h to create a SstFileManager that can be used to track the total size of SST files and control the SST files deletion rate.  

 ## 4.4.0 (1/14/2016)
 ### Public API Changes
@@ -21,7 +21,7 @@ depend on gflags. You will need to have gflags installed to run `make all`. This
 use binaries compiled by `make all` in production.

 * By default the binary we produce is optimized for the platform you're compiling on
-(-march=native). If you want to build a portable binary, add 'PORTABLE=1' before
+(-march=native or the equivalent). If you want to build a portable binary, add 'PORTABLE=1' before
 your make commands, like this: `PORTABLE=1 make static_lib`

 ## Dependencies
@@ -0,0 +1,11 @@
+This is the list of all known third-party language bindings for RocksDB. If something is missing, please open a pull request to add it.
+
+* Java - https://github.com/facebook/rocksdb/tree/master/java
+* Python - http://pyrocksdb.readthedocs.org/en/latest/
+* Perl - https://metacpan.org/pod/RocksDB
+* Node.js - https://npmjs.org/package/rocksdb
+* Go - https://github.com/tecbot/gorocksdb
+* Ruby - http://rubygems.org/gems/rocksdb-ruby
+* Haskell - https://hackage.haskell.org/package/rocksdb-haskell
+* PHP - https://github.com/Photonios/rocksdb-php
+* C# - https://github.com/warrenfalk/rocksdb-sharp
@@ -2,7 +2,7 @@ BSD License

 For rocksdb software

-Copyright (c) 2014, Facebook, Inc.
+Copyright (c) 2011-present, Facebook, Inc.
 All rights reserved.
 ---------------------------------------------------------------------

@@ -15,7 +15,9 @@ ARFLAGS = rs

 # Transform parallel LOG output into something more readable.
 perl_command = perl -n \
-  -e '@a=split("\t",$$_,-1); $$t=$$a[8]; $$t =~ s,^\./,,;'		\
+  -e '@a=split("\t",$$_,-1); $$t=$$a[8];'				\
+  -e '$$t =~ /.*if\s\[\[\s"(.*?\.[\w\/]+)/ and $$t=$$1;'		\
+  -e '$$t =~ s,^\./,,;'							\
  -e '$$t =~ s, >.*,,; chomp $$t;'					\
  -e '$$t =~ /.*--gtest_filter=(.*?\.[\w\/]+)/ and $$t=$$1;'		\
  -e 'printf "%7.3f %s %s\n", $$a[3], $$a[6] == 0 ? "PASS" : "FAIL", $$t'
@@ -84,7 +86,8 @@ endif
 # compile with -O2 if debug level is not 2
 ifneq ($(DEBUG_LEVEL), 2)
 OPT += -O2 -fno-omit-frame-pointer
-ifneq ($(MACHINE),ppc64) # ppc64 doesn't support -momit-leaf-frame-pointer
+# Skip for archs that don't support -momit-leaf-frame-pointer
+ifeq (,$(shell $(CXX) -fsyntax-only -momit-leaf-frame-pointer -xc /dev/null 2>&1))
 OPT += -momit-leaf-frame-pointer
 endif
 endif
@@ -143,6 +146,9 @@ else
 OPT += -DNDEBUG
 endif

+ifeq ($(PLATFORM), OS_SOLARIS)
+	PLATFORM_CXXFLAGS += -D _GLIBCXX_USE_C99
+endif
 ifneq ($(filter -DROCKSDB_LITE,$(OPT)),)
 	# found
 	CFLAGS += -fno-exceptions
@@ -165,12 +171,22 @@ ifdef COMPILE_WITH_TSAN
 	PLATFORM_CXXFLAGS += -fsanitize=thread -fPIC -DROCKSDB_TSAN_RUN
        # Turn off -pg when enabling TSAN testing, because that induces
        # a link failure.  TODO: find the root cause
-	pg =
-else
-	pg = -pg
+	PROFILING_FLAGS =
+endif
+
+# USAN doesn't work well with jemalloc. If we're compiling with USAN, we should use regular malloc.
+ifdef COMPILE_WITH_UBSAN
+	DISABLE_JEMALLOC=1
+	EXEC_LDFLAGS += -fsanitize=undefined
+	PLATFORM_CCFLAGS += -fsanitize=undefined
+	PLATFORM_CXXFLAGS += -fsanitize=undefined
 endif

 ifndef DISABLE_JEMALLOC
+	ifdef JEMALLOC
+		PLATFORM_CXXFLAGS += "-DROCKSDB_JEMALLOC"
+		PLATFORM_CCFLAGS +=  "-DROCKSDB_JEMALLOC"
+	endif
 	EXEC_LDFLAGS := $(JEMALLOC_LIB) $(EXEC_LDFLAGS)
 	PLATFORM_CXXFLAGS += $(JEMALLOC_INCLUDE)
 	PLATFORM_CCFLAGS += $(JEMALLOC_INCLUDE)
@@ -237,17 +253,25 @@ VALGRIND_VER := $(join $(VALGRIND_VER),valgrind)

 VALGRIND_OPTS = --error-exitcode=$(VALGRIND_ERROR) --leak-check=full

+BENCHTOOLOBJECTS = $(BENCH_SOURCES:.cc=.o) $(LIBOBJECTS) $(TESTUTIL)
+
 TESTS = \
 	db_test \
+	db_test2 \
+	db_block_cache_test \
+	db_bloom_filter_test \
 	db_iter_test \
 	db_log_iter_test \
 	db_compaction_filter_test \
 	db_compaction_test \
 	db_dynamic_level_test \
 	db_inplace_update_test \
+	db_iterator_test \
+	db_sst_test \
 	db_tailing_iter_test \
 	db_universal_compaction_test \
 	db_wal_test \
+	db_properties_test \
 	db_table_properties_test \
 	block_hash_index_test \
 	autovector_test \
@@ -312,6 +336,7 @@ TESTS = \
 	rate_limiter_test \
 	delete_scheduler_test \
 	options_test \
+	options_settable_test \
 	options_util_test \
 	event_logger_test \
 	cuckoo_table_builder_test \
@@ -332,9 +357,29 @@ TESTS = \
 	compact_on_deletion_collector_test \
 	compaction_job_stats_test \
 	transaction_test \
-	ldb_cmd_test
+	ldb_cmd_test \
+	iostats_context_test

-SUBSET :=  $(shell echo $(TESTS) |sed s/^.*$(ROCKSDBTESTS_START)/$(ROCKSDBTESTS_START)/)
+PARALLEL_TEST = \
+	backupable_db_test \
+	compact_on_deletion_collector_test \
+	db_compaction_filter_test \
+	db_compaction_test \
+	db_test \
+	db_universal_compaction_test \
+	fault_injection_test \
+	inlineskiplist_test \
+	manual_compaction_test \
+	table_test
+
+SUBSET := $(TESTS)
+ifdef ROCKSDBTESTS_START
+        SUBSET := $(shell echo $(SUBSET) | sed 's/^.*$(ROCKSDBTESTS_START)/$(ROCKSDBTESTS_START)/')
+endif
+
+ifdef ROCKSDBTESTS_END
+        SUBSET := $(shell echo $(SUBSET) | sed 's/$(ROCKSDBTESTS_END).*//')
+endif

 TOOLS = \
 	sst_dump \
@@ -435,7 +480,26 @@ coverage:
        # Delete intermediate files
 	find . -type f -regex ".*\.\(\(gcda\)\|\(gcno\)\)" -exec rm {} \;

-# Extract the names of its tests by running db_test with --gtest_list_tests.
+ifneq (,$(filter check parallel_check,$(MAKECMDGOALS)),)
+# Use /dev/shm if it has the sticky bit set (otherwise, /tmp),
+# and create a randomly-named rocksdb.XXXX directory therein.
+# We'll use that directory in the "make check" rules.
+ifeq ($(TMPD),)
+TMPD := $(shell f=/dev/shm; test -k $$f || f=/tmp;			\
+  perl -le 'use File::Temp "tempdir";'					\
+    -e 'print tempdir("'$$f'/rocksdb.XXXX", CLEANUP => 0)')
+endif
+endif
+
+# Run all tests in parallel, accumulating per-test logs in t/log-*.
+#
+# Each t/run-* file is a tiny generated bourne shell script that invokes one of
+# sub-tests. Why use a file for this?  Because that makes the invocation of
+# parallel below simpler, which in turn makes the parsing of parallel's
+# LOG simpler (the latter is for live monitoring as parallel
+# tests run).
+#
+# Test names are extracted by running tests with --gtest_list_tests.
 # This filter removes the "#"-introduced comments, and expands to
 # fully-qualified names by changing input like this:
 #
@@ -453,52 +517,33 @@ coverage:
 #   MultiThreaded/MultiThreadedDBTest.MultiThreaded/0
 #   MultiThreaded/MultiThreadedDBTest.MultiThreaded/1
 #
-test_names = \
-  ./db_test --gtest_list_tests						\
-    | perl -n								\
-      -e 's/ *\#.*//;'							\
-      -e '/^(\s*)(\S+)/; !$$1 and do {$$p=$$2; break};'			\
-      -e 'print qq! $$p$$2!'

-ifeq ($(MAKECMDGOALS),check)
-# Use /dev/shm if it has the sticky bit set (otherwise, /tmp),
-# and create a randomly-named rocksdb.XXXX directory therein.
-# We'll use that directory in the "make check" rules.
-ifeq ($(TMPD),)
-TMPD := $(shell f=/dev/shm; test -k $$f || f=/tmp;			\
-  perl -le 'use File::Temp "tempdir";'					\
-    -e 'print tempdir("'$$f'/rocksdb.XXXX", CLEANUP => 0)')
-endif
-endif
+parallel_tests = $(patsubst %,parallel_%,$(PARALLEL_TEST))
+.PHONY: gen_parallel_tests $(parallel_tests)
+$(parallel_tests): $(PARALLEL_TEST)
+	$(AM_V_at)TEST_BINARY=$(patsubst parallel_%,%,$@); \
+  TEST_NAMES=` \
+    ./$$TEST_BINARY --gtest_list_tests \
+    | perl -n \
+      -e 's/ *\#.*//;' \
+      -e '/^(\s*)(\S+)/; !$$1 and do {$$p=$$2; break};'	\
+      -e 'print qq! $$p$$2!'`; \
+	for TEST_NAME in $$TEST_NAMES; do \
+		TEST_SCRIPT=t/run-$$TEST_BINARY-$${TEST_NAME//\//-}; \
+		echo "  GEN     " $$TEST_SCRIPT; \
+    printf '%s\n' \
+      '#!/bin/sh' \
+      "d=\$(TMPD)$$TEST_SCRIPT" \
+      'mkdir -p $$d' \
+			"TEST_TMPDIR=\$$d ./$$TEST_BINARY --gtest_filter=$$TEST_NAME" \
+		> $$TEST_SCRIPT; \
+		chmod a=rx $$TEST_SCRIPT; \
+	done

-ifneq ($(T),)
-
-# Run all tests in parallel, accumulating per-test logs in t/log-*.
-
-# t_sanitized is each $(T) with "-" in place of each "/".
-t_sanitized = $(subst /,-,$(T))
-
-# t_run is each sanitized name with a leading "t/".
-t_run = $(patsubst %,t/%,$(t_sanitized))
-
-# Each t_run file is a tiny generated bourne shell script
-# that invokes one of db_tests's sub-tests. Why use a file
-# for this?  Because that makes the invocation of parallel
-# below simpler, which in turn makes the parsing of parallel's
-# LOG simpler (the latter is for live monitoring as parallel
-# tests run).
-filter = --gtest_filter=$(subst -,/,$(@F))
-$(t_run): Makefile db_test
-	$(AM_V_GEN)mkdir -p t
-	$(AM_V_at)rm -f $@ $@-t
-	$(AM_V_at)printf '%s\n'						\
-	    '#!/bin/sh'							\
-	    'd=$(TMPD)/$(@F)'						\
-	    'mkdir -p $$d'						\
-	    'TEST_TMPDIR=$$d ./db_test $(filter)'			\
-	  > $@-t
-	$(AM_V_at)chmod a=rx $@-t
-	$(AM_V_at)mv $@-t $@
+gen_parallel_tests:
+	$(AM_V_at)mkdir -p t
+	$(AM_V_at)rm -f t/run-*
+	$(MAKE) $(parallel_tests)

 # Reorder input lines (which are one per test) so that the
 # longest-running tests appear first in the output.
@@ -517,7 +562,7 @@ $(t_run): Makefile db_test
 # 107.816 PASS t/DBTest.EncodeDecompressedBlockSizeTest
 #
 slow_test_regexp = \
-  ^t/DBTest\.(?:FileCreationRandomFailure|EncodeDecompressedBlockSizeTest)$$
+	^t/run-table_test-HarnessTest.Randomized$$|^t/run-db_test-.*(?:FileCreationRandomFailure|EncodeDecompressedBlockSizeTest)$$
 prioritize_long_running_tests =						\
  perl -pe 's,($(slow_test_regexp)),100 $$1,'				\
    | sort -k1,1gr							\
@@ -528,26 +573,43 @@ prioritize_long_running_tests =						\
 # Run with "make J=200% check" to run two parallel jobs per core.
 # The default is to run one job per core (J=100%).
 # See "man parallel" for its "-j ..." option.
-J = 100%
+J ?= 100%

 # Use this regexp to select the subset of tests whose names match.
 tests-regexp = .

+t_run = $(wildcard t/run-*)
 .PHONY: check_0
-check_0: $(t_run)
+check_0:
+	$(AM_V_GEN)export TEST_TMPDIR=$(TMPD); \
+	printf '%s\n' ''						\
+	  'To monitor subtest <duration,pass/fail,name>,'		\
+	  '  run "make watch-log" in a separate window' '';		\
+	test -t 1 && eta=--eta || eta=; \
+	{ \
+		printf './%s\n' $(filter-out $(PARALLEL_TEST),$(TESTS)); \
+		printf '%s\n' $(t_run); \
+	} \
+	  | $(prioritize_long_running_tests)				\
+	  | grep -E '$(tests-regexp)'					\
+	  | parallel -j$(J) --joblog=LOG $$eta --gnu '{} >& t/log-{/}'
+
+.PHONY: valgrind_check_0
+valgrind_check_0:
 	$(AM_V_GEN)export TEST_TMPDIR=$(TMPD);				\
 	printf '%s\n' ''						\
 	  'To monitor subtest <duration,pass/fail,name>,'		\
 	  '  run "make watch-log" in a separate window' '';		\
 	test -t 1 && eta=--eta || eta=;					\
 	{								\
-	  printf './%s\n' $(filter-out db_test, $(TESTS));		\
+	  printf './%s\n' $(filter-out $(PARALLEL_TEST) %skiplist_test options_settable_test, $(TESTS));		\
 	  printf '%s\n' $(t_run);					\
 	}								\
 	  | $(prioritize_long_running_tests)				\
 	  | grep -E '$(tests-regexp)'					\
-	  | parallel -j$(J) --joblog=LOG $$eta --gnu '{} >& t/log-{/}'
-endif
+	  | parallel -j$(J) --joblog=LOG $$eta --gnu \
+      'if [[ "{}" == "./"* ]] ; then $(DRIVER) {} >& t/valgrind_log-{/}; ' \
+      'else {} >& t/valgrind_log-{/}; fi'

 CLEAN_FILES += t LOG $(TMPD)

@@ -564,11 +626,11 @@ watch-log:
 # If J != 1 and GNU parallel is installed, run the tests in parallel,
 # via the check_0 rule above.  Otherwise, run them sequentially.
 check: all
+	$(MAKE) gen_parallel_tests
 	$(AM_V_GEN)if test "$(J)" != 1                                  \
 	    && (parallel --gnu --help 2>/dev/null) |                    \
 	        grep -q 'GNU Parallel';                                 \
 	then                                                            \
-	    t=$$($(test_names));                                        \
 	    $(MAKE) T="$$t" TMPD=$(TMPD) check_0;                       \
 	else                                                            \
 	    for t in $(TESTS); do                                       \
@@ -590,7 +652,7 @@ ldb_tests: ldb
 crash_test: whitebox_crash_test blackbox_crash_test

 blackbox_crash_test: db_stress
-	python -u tools/db_crashtest.py --simple blackbox 
+	python -u tools/db_crashtest.py --simple blackbox
 	python -u tools/db_crashtest.py blackbox

 whitebox_crash_test: db_stress
@@ -607,14 +669,80 @@ asan_crash_test:
 	COMPILE_WITH_ASAN=1 $(MAKE) crash_test
 	$(MAKE) clean

+ubsan_check:
+	$(MAKE) clean
+	COMPILE_WITH_UBSAN=1 $(MAKE) check -j32
+	$(MAKE) clean
+
+ubsan_crash_test:
+	$(MAKE) clean
+	COMPILE_WITH_UBSAN=1 $(MAKE) crash_test
+	$(MAKE) clean
+
 valgrind_check: $(TESTS)
-	for t in $(filter-out %skiplist_test,$(TESTS)); do \
-		$(VALGRIND_VER) $(VALGRIND_OPTS) ./$$t; \
+	$(MAKE) gen_parallel_tests
+	$(AM_V_GEN)if test "$(J)" != 1                                  \
+	    && (parallel --gnu --help 2>/dev/null) |                    \
+	        grep -q 'GNU Parallel';                                 \
+	then                                                            \
+      $(MAKE) TMPD=$(TMPD)                                        \
+      DRIVER="$(VALGRIND_VER) $(VALGRIND_OPTS)" valgrind_check_0; \
+	else                                                            \
+		for t in $(filter-out %skiplist_test options_settable_test,$(TESTS)); do \
+			$(VALGRIND_VER) $(VALGRIND_OPTS) ./$$t; \
+			ret_code=$$?; \
+			if [ $$ret_code -ne 0 ]; then \
+				exit $$ret_code; \
+			fi; \
+		done; \
+	fi
+
+
+ifneq ($(PAR_TEST),)
+parloop:
+	ret_bad=0;							\
+	for t in $(PAR_TEST); do		\
+		echo "===== Running $$t in parallel $(NUM_PAR)";\
+		if [ $(db_test) -eq 1 ]; then \
+			seq $(J) | v="$$t" parallel --gnu 's=$(TMPD)/rdb-{};  export TEST_TMPDIR=$$s;' \
+				'timeout 2m ./db_test --gtest_filter=$$v >> $$s/log-{} 2>1'; \
+		else\
+			seq $(J) | v="./$$t" parallel --gnu 's=$(TMPD)/rdb-{};' \
+			     'export TEST_TMPDIR=$$s; timeout 10m $$v >> $$s/log-{} 2>1'; \
+		fi; \
 		ret_code=$$?; \
 		if [ $$ret_code -ne 0 ]; then \
-			exit $$ret_code; \
+			ret_bad=$$ret_code; \
+			echo $$t exited with $$ret_code; \
 		fi; \
-	done
+	done; \
+	exit $$ret_bad;
+endif
+
+test_names = \
+  ./db_test --gtest_list_tests						\
+    | perl -n								\
+      -e 's/ *\#.*//;'							\
+      -e '/^(\s*)(\S+)/; !$$1 and do {$$p=$$2; break};'			\
+      -e 'print qq! $$p$$2!'
+
+parallel_check: $(TESTS)
+	$(AM_V_GEN)if test "$(J)" > 1                                  \
+	    && (parallel --gnu --help 2>/dev/null) |                    \
+	        grep -q 'GNU Parallel';                                 \
+	then                                                            \
+	    echo Running in parallel $(J);			\
+	else                                                            \
+	    echo "Need to have GNU Parallel and J > 1"; exit 1;		\
+	fi;								\
+	ret_bad=0;							\
+	echo $(J);\
+	echo Test Dir: $(TMPD); \
+        seq $(J) | parallel --gnu 's=$(TMPD)/rdb-{}; rm -rf $$s; mkdir $$s'; \
+	$(MAKE)  PAR_TEST="$(shell $(test_names))" TMPD=$(TMPD) \
+		J=$(J) db_test=1 parloop; \
+	$(MAKE) PAR_TEST="$(filter-out db_test, $(TESTS))" \
+		TMPD=$(TMPD) J=$(J) db_test=0 parloop;

 analyze: clean
 	$(CLANG_SCAN_BUILD) --use-analyzer=$(CLANG_ANALYZER) \
@@ -652,7 +780,7 @@ clean:

 tags:
 	ctags * -R
-	cscope -b `find . -name '*.cc'` `find . -name '*.h'`
+	cscope -b `find . -name '*.cc'` `find . -name '*.h'` `find . -name '*.c'`

 format:
 	build_tools/format-diff.sh
@@ -667,7 +795,7 @@ $(LIBRARY): $(LIBOBJECTS)
 	$(AM_V_AR)rm -f $@
 	$(AM_V_at)$(AR) $(ARFLAGS) $@ $(LIBOBJECTS)

-db_bench: db/db_bench.o $(LIBOBJECTS) $(TESTUTIL)
+db_bench: tools/db_bench.o $(BENCHTOOLOBJECTS)
 	$(AM_LINK)

 cache_bench: util/cache_bench.o $(LIBOBJECTS) $(TESTUTIL)
@@ -742,6 +870,15 @@ slice_transform_test: util/slice_transform_test.o $(LIBOBJECTS) $(TESTHARNESS)
 db_test: db/db_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

+db_test2: db/db_test2.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
+db_block_cache_test: db/db_block_cache_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
+db_bloom_filter_test: db/db_bloom_filter_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 db_log_iter_test: db/db_log_iter_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

@@ -757,6 +894,12 @@ db_dynamic_level_test: db/db_dynamic_level_test.o db/db_test_util.o $(LIBOBJECTS
 db_inplace_update_test: db/db_inplace_update_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

+db_iterator_test: db/db_iterator_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
+db_sst_test: db/db_sst_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 db_tailing_iter_test: db/db_tailing_iter_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

@@ -769,11 +912,14 @@ db_universal_compaction_test: db/db_universal_compaction_test.o db/db_test_util.
 db_wal_test: db/db_wal_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

+db_properties_test: db/db_properties_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 db_table_properties_test: db/db_table_properties_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

 log_write_bench: util/log_write_bench.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(AM_LINK) $(pg)
+	$(AM_LINK) $(PROFILING_FLAGS)

 plain_table_db_test: db/plain_table_db_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
@@ -782,7 +928,7 @@ comparator_db_test: db/comparator_db_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

 table_reader_bench: table/table_reader_bench.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(AM_LINK) $(pg)
+	$(AM_LINK) $(PROFILING_FLAGS)

 perf_context_test: db/perf_context_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_V_CCLD)$(CXX) $^ $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
@@ -946,6 +1092,9 @@ compact_files_test: db/compact_files_test.o $(LIBOBJECTS) $(TESTHARNESS)
 options_test: util/options_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

+options_settable_test: util/options_settable_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 options_util_test: utilities/options/options_util_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

@@ -970,7 +1119,7 @@ manual_compaction_test: db/manual_compaction_test.o $(LIBOBJECTS) $(TESTHARNESS)
 filelock_test: util/filelock_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

-auto_roll_logger_test: util/auto_roll_logger_test.o $(LIBOBJECTS) $(TESTHARNESS)
+auto_roll_logger_test: db/auto_roll_logger_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

 memtable_list_test: db/memtable_list_test.o $(LIBOBJECTS) $(TESTHARNESS)
@@ -988,12 +1137,18 @@ transaction_test: utilities/transactions/transaction_test.o $(LIBOBJECTS) $(TEST
 sst_dump: tools/sst_dump.o $(LIBOBJECTS)
 	$(AM_LINK)

+repair_test: db/repair_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 ldb_cmd_test: tools/ldb_cmd_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)

 ldb: tools/ldb.o $(LIBOBJECTS)
 	$(AM_LINK)

+iostats_context_test: util/iostats_context_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_V_CCLD)$(CXX) $^ $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+
 #-------------------------------------------------
 # make install related stuff
 INSTALL_PATH ?= /usr/local
@@ -1036,7 +1191,11 @@ install: install-static
 # ---------------------------------------------------------------------------

 JAVA_INCLUDE = -I$(JAVA_HOME)/include/ -I$(JAVA_HOME)/include/linux
-ARCH := $(shell getconf LONG_BIT)
+ifeq ($(PLATFORM), OS_SOLARIS)
+	ARCH := $(shell isainfo -b)
+else
+	ARCH := $(shell getconf LONG_BIT)
+endif
 ROCKSDBJNILIB = librocksdbjni-linux$(ARCH).so
 ROCKSDB_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-linux$(ARCH).jar
 ROCKSDB_JAR_ALL = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH).jar
@@ -1044,14 +1203,19 @@ ROCKSDB_JAVADOCS_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PA
 ROCKSDB_SOURCES_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-sources.jar

 ifeq ($(PLATFORM), OS_MACOSX)
-ROCKSDBJNILIB = librocksdbjni-osx.jnilib
-ROCKSDB_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-osx.jar
+	ROCKSDBJNILIB = librocksdbjni-osx.jnilib
+	ROCKSDB_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-osx.jar
 ifneq ("$(wildcard $(JAVA_HOME)/include/darwin)","")
 	JAVA_INCLUDE = -I$(JAVA_HOME)/include -I $(JAVA_HOME)/include/darwin
 else
 	JAVA_INCLUDE = -I/System/Library/Frameworks/JavaVM.framework/Headers/
 endif
 endif
+ifeq ($(PLATFORM), OS_SOLARIS)
+	ROCKSDBJNILIB = librocksdbjni-solaris$(ARCH).so
+	ROCKSDB_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-solaris$(ARCH).jar
+	JAVA_INCLUDE = -I$(JAVA_HOME)/include/ -I$(JAVA_HOME)/include/solaris
+endif

 libz.a:
 	-rm -rf zlib-1.2.8
@@ -1145,11 +1309,10 @@ jtest: rocksdbjava
 jdb_bench:
 	cd java;$(MAKE) db_bench;

-commit-prereq:
-	$(MAKE) clean && $(MAKE) all check;
+commit_prereq: build_tools/rocksdb-lego-determinator \
+               build_tools/precommit_checker.py
+	J=$(J) build_tools/precommit_checker.py unit unit_481 clang_unit tsan asan ubsan lite
 	$(MAKE) clean && $(MAKE) jclean && $(MAKE) rocksdbjava;
-	$(MAKE) clean && USE_CLANG=1 $(MAKE) all;
-	$(MAKE) clean && OPT=-DROCKSDB_LITE $(MAKE) static_lib;

 xfunc:
 	for xftest in $(XFUNC_TESTS); do \
@@ -9,7 +9,7 @@ At Facebook, we use RocksDB as a backend for many different stateful services. W
 ## LinkedIn
 Two different use cases at Linkedin are using RocksDB as a storage engine:

-1. LinkedIn's follow feed for storing user's activities
+1. LinkedIn's follow feed for storing user's activities. Check out the blog post: https://engineering.linkedin.com/blog/2016/03/followfeed--linkedin-s-feed-made-faster-and-smarter
 2. Apache Samza, open source framework for stream processing

 Learn more about those use cases in a Tech Talk by Ankit Gupta and Naveen Somasundaram: http://www.youtube.com/watch?v=plqVp_OnSzg
@@ -45,4 +45,14 @@ Airbnb is using RocksDB as a storage engine for their personalized search servic
 Pinterest's Object Retrieval System uses RocksDB for storage: https://www.youtube.com/watch?v=MtFEVEs_2Vo

 ## Smyte
-[Smyte](https://www.smyte.com/) uses RocksDB as the storage layer for their core key-value storage, high-performance counters and time-windowed HyperLogLog services.
+[Smyte](https://www.smyte.com/) uses RocksDB as the storage layer for their core key-value storage, high-performance counters and time-windowed HyperLogLog services.
+
+## Rakuten Marketing
+[Rakuten Marketing](https://marketing.rakuten.com/) uses RocksDB as the disk cache layer for the real-time bidding service in their Performance DSP.
+
+## VWO, Wingify
+[VWO's](https://vwo.com/) Smart Code checker and URL helper uses RocksDB to store all the URLs where VWO's Smart Code is installed.
+
+## quasardb
+[quasardb](https://www.quasardb.net) is a high-performance, distributed, transactional key-value database that integrates well with in-memory analytics engines such as Apache Spark. 
+quasardb uses a heavily tuned RocksDB as its persistence layer.
@@ -10,13 +10,185 @@ class FacebookArcanistConfiguration extends ArcanistConfiguration {
                                 ArcanistBaseWorkflow $workflow,
                                 $error_code) {
    if ($command == 'diff' && !$workflow->isRawDiffSource()) {
-      $this->maybePushToJenkins($workflow);
+      $this->startTestsInJenkins($workflow);
+      $this->startTestsInSandcastle($workflow);
    }
  }

+  //////////////////////////////////////////////////////////////////////
+  /*  Run tests in sandcastle */
+  function postURL($diffID, $url) {
+    $cmd = 'echo \'{"diff_id": "' . $diffID . '", '
+           . '"name":"click here for sandcastle tests for D' . $diffID . '", '
+           . '"link":"' . $url . '"}\' | '
+           . 'http_proxy=fwdproxy.any.facebook.com:8080 '
+           . 'https_proxy=fwdproxy.any.facebook.com:8080 arc call-conduit '
+           . 'differential.updateunitresults';
+    shell_exec($cmd);
+  }
+
+  function updateTestCommand($diffID, $test, $status) {
+    $cmd = 'echo \'{"diff_id": "' . $diffID . '", '
+           . '"name":"' . $test . '", '
+           . '"result":"' . $status . '"}\' | '
+           . 'http_proxy=fwdproxy.any.facebook.com:8080 '
+           . 'https_proxy=fwdproxy.any.facebook.com:8080 arc call-conduit '
+           . 'differential.updateunitresults';
+    return $cmd;
+  }
+
+  function updateTest($diffID, $test) {
+    shell_exec($this->updateTestCommand($diffID, $test, "waiting"));
+  }
+
+  function getSteps($diffID, $username, $test) {
+    $arcrc_content = exec("cat ~/.arcrc | gzip -f | base64 -w0");
+
+    // Sandcastle machines don't have arc setup. We copy the user certificate
+    // and authenticate using that in sandcastle
+    $setup = array(
+      "name" => "Setup arcrc",
+      "shell" => "echo " . $arcrc_content . " | base64 --decode"
+                 . " | gzip -d > ~/.arcrc",
+      "user" => "root"
+    );
+
+    // arc demands certain permission on its config
+    $fix_permission = array(
+      "name" => "Fix environment",
+      "shell" => "chmod 600 ~/.arcrc",
+      "user" => "root"
+    );
+
+    // fbcode is a sub-repo. We cannot patch until we add it to ignore otherwise
+    // git thinks it is uncommited change
+    $fix_git_ignore = array(
+      "name" => "Fix git ignore",
+      "shell" => "echo fbcode >> .git/info/exclude",
+      "user" => "root"
+    );
+
+    // Patch the code (keep your fingures crossed)
+    $patch = array(
+      "name" => "Patch " . $diffID,
+      "shell" => "HTTPS_PROXY=fwdproxy:8080 arc --arcrc-file ~/.arcrc "
+                  . "patch --diff " . $diffID,
+      "user" => "root"
+    );
+
+    // Clean up the user arc config we are using
+    $cleanup = array(
+      "name" => "Arc cleanup",
+      "shell" => "rm -f ~/.arcrc",
+      "user" => "root"
+    );
+
+    // Construct the steps in the order of execution
+    $steps[] = $setup;
+    $steps[] = $fix_permission;
+    $steps[] = $fix_git_ignore;
+    $steps[] = $patch;
+
+    // Run the actual command
+    $this->updateTest($diffID, $test);
+    $cmd = $this->updateTestCommand($diffID, $test, "running") . ";"
+           . "./build_tools/precommit_checker.py " . $test
+           . "; exit_code=$?; ([[ \$exit_code -eq 0 ]] &&"
+           . $this->updateTestCommand($diffID, $test, "pass") . ")"
+           . "||" . $this->updateTestCommand($diffID, $test, "fail")
+           . "; cat /tmp/precommit-check.log"
+           . "; for f in `ls t/log-*`; do echo \$f; cat \$f; done;"
+           . "[[ \$exit_code -eq 0 ]]";
+
+    $run_test = array(
+      "name" => "Run " . $test,
+      "shell" => $cmd,
+      "user" => "root",
+    );
+
+    $steps[] = $run_test;
+    $steps[] = $cleanup;
+
+    return $steps;
+  }
+
+  function startTestsInSandcastle($workflow) {
+    // extract information we need from workflow or CLI
+    $diffID = $workflow->getDiffId();
+    $username = exec("whoami");
+
+    if ($diffID == null || $username == null) {
+      // there is no diff and we can't extract username
+      // we cannot schedule sandcasstle job
+      return;
+    }
+
+    if (strcmp(getenv("ROCKSDB_CHECK_ALL"), 1) == 0) {
+      // extract all tests from the CI definition
+      $output = file_get_contents("build_tools/rocksdb-lego-determinator");
+      preg_match_all('/[ ]{2}([a-zA-Z0-9_]+)[\)]{1}/', $output, $matches);
+      $tests = $matches[1];
+    } else {
+      // manually list of tests we want to run in sandcastle
+      $tests = array(
+        "unit", "unit_481", "clang_unit", "tsan", "asan", "lite", "valgrind"
+      );
+    }
+
+    // construct a job definition for each test and add it to the master plan
+    foreach ($tests as $test) {
+      $arg[] = array(
+        "name" => "RocksDB diff " . $diffID . " test " . $test,
+        "steps" => $this->getSteps($diffID, $username, $test)
+      );
+    }
+
+    // we cannot submit the parallel execution master plan to sandcastle
+    // we need supply the job plan as a determinator
+    // so we construct a small job that will spit out the master job plan
+    // which sandcastle will parse and execute
+    // Why compress ? Otherwise we run over the max string size.
+    $cmd = "echo " . base64_encode(json_encode($arg))
+           . " | gzip -f | base64 -w0";
+    $arg_encoded = shell_exec($cmd);
+
+    $command = array(
+      "name" => "Run diff " . $diffID . "for user " . $username,
+      "steps" => array()
+    );
+
+    $command["steps"][] = array(
+      "name" => "Generate determinator",
+      "shell" => "echo " . $arg_encoded . " | base64 --decode | gzip -d"
+                 . " | base64 --decode",
+      "determinator" => true,
+      "user" => "root"
+    );
+
+    // submit to sandcastle
+    $url = 'https://interngraph.intern.facebook.com/sandcastle/generate?'
+            .'command=SandcastleUniversalCommand'
+            .'&vcs=rocksdb-git&revision=origin%2Fmaster&type=lego'
+            .'&user=krad&alias=rocksdb-precommit'
+            .'&command-args=' . urlencode(json_encode($command));
+
+    $cmd = 'https_proxy= HTTPS_PROXY= curl -s -k -F app=659387027470559 '
+            . '-F token=AeO_3f2Ya3TujjnxGD4 "' . $url . '"';
+
+    $output = shell_exec($cmd);
+
+    // extract sandcastle URL from the response
+    preg_match('/url": "(.+)"/', $output, $sandcastle_url);
+
+    echo "\nSandcastle URL: " . $sandcastle_url[1] . "\n";
+
+    // Ask phabricator to display it on the diff UI
+    $this->postURL($diffID, $sandcastle_url[1]);
+  }
+
  //////////////////////////////////////////////////////////////////////
  /* Send off builds to jenkins */
-  function maybePushToJenkins($workflow) {
+  function startTestsInJenkins($workflow) {
    $diffID = $workflow->getDiffID();
    if ($diffID === null) {
      return;
@@ -31,5 +203,4 @@ class FacebookArcanistConfiguration extends ArcanistConfiguration {
               ."buildWithParameters?token=AUTH&DIFF_ID=$diffID";
    system("curl --noproxy '*' \"$url\" > /dev/null 2>&1");
  }
-
 }
@@ -1,5 +1,5 @@
 #!/usr/bin/python
-# Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+# Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 # This source code is licensed under the BSD-style license found in the
 # LICENSE file in the root directory of this source tree. An additional grant
 # of patent rights can be found in the PATENTS file in the same directory.
@@ -2714,7 +2714,7 @@ def CheckSpacing(filename, clean_lines, linenum, nesting_state, error):

    # Look for < that is not surrounded by spaces.  This is only
    # triggered if both sides are missing spaces, even though
-    # technically should should flag if at least one side is missing a
+    # technically should flag if at least one side is missing a
    # space.  This is done to avoid some false positives with shifts.
    match = Search(r'[^\s<]<([^\s=<].*)', reduced_line)
    if (match and
@@ -189,17 +189,19 @@ if [ "$CROSS_COMPILE" = "true" -o "$FBCODE_BUILD" = "true" ]; then
    # Also don't need any compilation tests if compiling on fbcode
    true
 else
-    # Test whether fallocate is available
-    $CXX $CFLAGS -x c++ - -o /dev/null 2>/dev/null  <<EOF
-      #include <fcntl.h>
-      #include <linux/falloc.h>
-      int main() {
-	int fd = open("/dev/null", 0);
-  fallocate(fd, FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE, 0, 1024);
-      }
+    if ! test $ROCKSDB_DISABLE_FALLOCATE; then
+        # Test whether fallocate is available
+        $CXX $CFLAGS -x c++ - -o /dev/null 2>/dev/null  <<EOF
+          #include <fcntl.h>
+          #include <linux/falloc.h>
+          int main() {
+      int fd = open("/dev/null", 0);
+      fallocate(fd, FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE, 0, 1024);
+          }
 EOF
-    if [ "$?" = 0 ]; then
-        COMMON_FLAGS="$COMMON_FLAGS -DROCKSDB_FALLOCATE_PRESENT"
+        if [ "$?" = 0 ]; then
+            COMMON_FLAGS="$COMMON_FLAGS -DROCKSDB_FALLOCATE_PRESENT"
+        fi
    fi

    # Test whether Snappy library is installed
@@ -297,13 +299,14 @@ EOF

    # Test whether jemalloc is available
    if echo 'int main() {}' | $CXX $CFLAGS -x c++ - -o /dev/null -ljemalloc \
-      2>/dev/null; then 
+      2>/dev/null; then
        PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -ljemalloc"
        JAVA_LDFLAGS="$JAVA_LDFLAGS -ljemalloc"
+        JEMALLOC=1
    else
        # jemalloc is not available. Let's try tcmalloc
        if echo 'int main() {}' | $CXX $CFLAGS -x c++ - -o /dev/null \
-          -ltcmalloc 2>/dev/null; then 
+          -ltcmalloc 2>/dev/null; then
            PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -ltcmalloc"
            JAVA_LDFLAGS="$JAVA_LDFLAGS -ltcmalloc"
        fi
@@ -320,6 +323,55 @@ EOF
    if [ "$?" = 0 ]; then
        COMMON_FLAGS="$COMMON_FLAGS -DROCKSDB_MALLOC_USABLE_SIZE"
    fi
+
+    # Test whether PTHREAD_MUTEX_ADAPTIVE_NP mutex type is available
+    $CXX $CFLAGS -x c++ - -o /dev/null 2>/dev/null  <<EOF
+      #include <pthread.h>
+      int main() {
+        int x = PTHREAD_MUTEX_ADAPTIVE_NP;
+        return 0;
+      }
+EOF
+    if [ "$?" = 0 ]; then
+        COMMON_FLAGS="$COMMON_FLAGS -DROCKSDB_PTHREAD_ADAPTIVE_MUTEX"
+    fi
+
+    # Test whether backtrace is available
+    $CXX $CFLAGS -x c++ - -o /dev/null 2>/dev/null  <<EOF
+      #include <execinfo.h>>
+      int main() {
+        void* frames[1];
+        backtrace_symbols(frames, backtrace(frames, 1));
+        return 0;
+      }
+EOF
+    if [ "$?" = 0 ]; then
+        COMMON_FLAGS="$COMMON_FLAGS -DROCKSDB_BACKTRACE"
+    else
+        # Test whether execinfo library is installed
+        $CXX $CFLAGS -lexecinfo -x c++ - -o /dev/null 2>/dev/null  <<EOF
+          #include <execinfo.h>
+          int main() {
+            void* frames[1];
+            backtrace_symbols(frames, backtrace(frames, 1));
+          }
+EOF
+        if [ "$?" = 0 ]; then
+            COMMON_FLAGS="$COMMON_FLAGS -DROCKSDB_BACKTRACE"
+            PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -lexecinfo"
+            JAVA_LDFLAGS="$JAVA_LDFLAGS -lexecinfo"
+        fi
+    fi
+
+    # Test if -pg is supported
+    $CXX $CFLAGS -pg -x c++ - -o /dev/null 2>/dev/null  <<EOF
+      int main() {
+        return 0;
+      }
+EOF
+    if [ "$?" = 0 ]; then
+        PROFILING_FLAGS=-pg
+    fi
 fi

 # TODO(tec): Fix -Wshorten-64-to-32 errors on FreeBSD and enable the warning.
@@ -359,7 +411,15 @@ if test "$USE_SSE"; then
  # if Intel SSE instruction set is supported, set USE_SSE=1
  COMMON_FLAGS="$COMMON_FLAGS -msse -msse4.2 "
 elif test -z "$PORTABLE"; then
-  COMMON_FLAGS="$COMMON_FLAGS -march=native "
+  if test -n "`echo $TARGET_ARCHITECTURE | grep ^ppc64`"; then
+    # Tune for this POWER processor, treating '+' models as base models
+    POWER=`LD_SHOW_AUXV=1 /bin/true | grep AT_PLATFORM | grep -E -o power[0-9]+`
+    COMMON_FLAGS="$COMMON_FLAGS -mcpu=$POWER -mtune=$POWER "
+  elif test -n "`echo $TARGET_ARCHITECTURE | grep ^s390x`"; then
+    COMMON_FLAGS="$COMMON_FLAGS -march=z10 "
+  else
+    COMMON_FLAGS="$COMMON_FLAGS -march=native "
+  fi
 fi

 PLATFORM_CCFLAGS="$PLATFORM_CCFLAGS $COMMON_FLAGS"
@@ -392,3 +452,7 @@ echo "ROCKSDB_MINOR=$ROCKSDB_MINOR" >> "$OUTPUT"
 echo "ROCKSDB_PATCH=$ROCKSDB_PATCH" >> "$OUTPUT"
 echo "CLANG_SCAN_BUILD=$CLANG_SCAN_BUILD" >> "$OUTPUT"
 echo "CLANG_ANALYZER=$CLANG_ANALYZER" >> "$OUTPUT"
+echo "PROFILING_FLAGS=$PROFILING_FLAGS" >> "$OUTPUT"
+if test -n "$JEMALLOC"; then
+  echo "JEMALLOC=1" >> "$OUTPUT"
+fi
@@ -0,0 +1,16 @@
+GCC_BASE=/mnt/vol/engshare/fbcode/third-party2/gcc/4.9.x/centos6-native/1317bc4/
+CLANG_BASE=/mnt/gvfs/third-party2/clang/fc904e50a9266b9d7b98cae1993afa0c5aae1440/3.7.1/centos6-native/9d9ecb9/
+LIBGCC_BASE=/mnt/gvfs/third-party2/libgcc/f97108c12512b3b0789ac4515d836bdb1eae1142/4.9.x/gcc-4.9-glibc-2.20/024dbc3
+GLIBC_BASE=/mnt/gvfs/third-party2/glibc/f5484f168c0e4d19823d41df052c5870c6e575a4/2.20/gcc-4.9-glibc-2.20/500e281
+SNAPPY_BASE=/mnt/gvfs/third-party2/snappy/cbf6f1f209e5bd160bdc5d971744e039f36b1566/1.1.3/gcc-4.9-glibc-2.20/e9936bf
+ZLIB_BASE=/mnt/gvfs/third-party2/zlib/6d39cb54708049f527e713ad19f2aadb9d3667e8/1.2.8/gcc-4.9-glibc-2.20/e9936bf
+BZIP2_BASE=/mnt/gvfs/third-party2/bzip2/2ddd45f0853bfc8bb1c27f0f447236a1a26c338a/1.0.6/gcc-4.9-glibc-2.20/e9936bf
+LZ4_BASE=/mnt/gvfs/third-party2/lz4/6858fac689e0f92e584224d91bdb0e39f6c8320d/r131/gcc-4.9-glibc-2.20/e9936bf
+ZSTD_BASE=/mnt/gvfs/third-party2/zstd/cb6c4880fcb4fee471574ba6af63a3882155a16a/0.5.1/gcc-4.9-glibc-2.20/e9936bf
+GFLAGS_BASE=/mnt/gvfs/third-party2/gflags/c7275a4ceae0aca0929e56964a31dafc53c1ee96/2.1.1/gcc-4.8.1-glibc-2.17/c3f970a
+JEMALLOC_BASE=/mnt/gvfs/third-party2/jemalloc/d9acac8a9a2c9378eb696e22ffa8bd0e97d9206b/master/gcc-4.9-glibc-2.20/a6c5e1e
+NUMA_BASE=/mnt/gvfs/third-party2/numa/ae54a5ed22cdabb1c6446dce4e8ffae5b4446d73/2.0.8/gcc-4.9-glibc-2.20/e9936bf
+LIBUNWIND_BASE=/mnt/gvfs/third-party2/libunwind/41bfa3759ce52c071f5fd547ec9ecd2522929f0a/trunk/gcc-4.9-glibc-2.20/12266b1
+KERNEL_HEADERS_BASE=/mnt/gvfs/third-party2/kernel-headers/1a48835975c66d30e47770ec419758ed3b9ba010/3.10.62-62_fbk17_03959_ge29cc63/gcc-4.9-glibc-2.20/da39a3e
+BINUTILS_BASE=/mnt/gvfs/third-party2/binutils/71454c53fffcb716a0beb9a90047aff7fb5c984a/2.26/centos6-native/da39a3e
+VALGRIND_BASE=/mnt/gvfs/third-party2/valgrind/af85c56f424cd5edfc2c97588299b44ecdec96bb/3.10.0/gcc-4.9-glibc-2.20/e9936bf
@@ -0,0 +1,16 @@
+GCC_BASE=/mnt/vol/engshare/fbcode/third-party2/gcc/4.8.1/centos6-native/cc6c9dc/
+CLANG_BASE=/mnt/gvfs/third-party2/clang/fc904e50a9266b9d7b98cae1993afa0c5aae1440/3.7.1/centos6-native/9d9ecb9/
+LIBGCC_BASE=/mnt/gvfs/third-party2/libgcc/f97108c12512b3b0789ac4515d836bdb1eae1142/4.8.1/gcc-4.8.1-glibc-2.17/8aac7fc
+GLIBC_BASE=/mnt/gvfs/third-party2/glibc/f5484f168c0e4d19823d41df052c5870c6e575a4/2.17/gcc-4.8.1-glibc-2.17/99df8fc
+SNAPPY_BASE=/mnt/gvfs/third-party2/snappy/cbf6f1f209e5bd160bdc5d971744e039f36b1566/1.1.3/gcc-4.8.1-glibc-2.17/c3f970a
+ZLIB_BASE=/mnt/gvfs/third-party2/zlib/6d39cb54708049f527e713ad19f2aadb9d3667e8/1.2.8/gcc-4.8.1-glibc-2.17/c3f970a
+BZIP2_BASE=/mnt/gvfs/third-party2/bzip2/2ddd45f0853bfc8bb1c27f0f447236a1a26c338a/1.0.6/gcc-4.8.1-glibc-2.17/c3f970a
+LZ4_BASE=/mnt/gvfs/third-party2/lz4/6858fac689e0f92e584224d91bdb0e39f6c8320d/r131/gcc-4.8.1-glibc-2.17/c3f970a
+ZSTD_BASE=/mnt/gvfs/third-party2/zstd/cb6c4880fcb4fee471574ba6af63a3882155a16a/0.5.1/gcc-4.8.1-glibc-2.17/c3f970a
+GFLAGS_BASE=/mnt/gvfs/third-party2/gflags/c7275a4ceae0aca0929e56964a31dafc53c1ee96/2.1.1/gcc-4.8.1-glibc-2.17/c3f970a
+JEMALLOC_BASE=/mnt/gvfs/third-party2/jemalloc/d9acac8a9a2c9378eb696e22ffa8bd0e97d9206b/master/gcc-4.8.1-glibc-2.17/8d31e51
+NUMA_BASE=/mnt/gvfs/third-party2/numa/ae54a5ed22cdabb1c6446dce4e8ffae5b4446d73/2.0.8/gcc-4.8.1-glibc-2.17/c3f970a
+LIBUNWIND_BASE=/mnt/gvfs/third-party2/libunwind/41bfa3759ce52c071f5fd547ec9ecd2522929f0a/trunk/gcc-4.8.1-glibc-2.17/675d945
+KERNEL_HEADERS_BASE=/mnt/gvfs/third-party2/kernel-headers/1a48835975c66d30e47770ec419758ed3b9ba010/3.10.62-62_fbk17_03959_ge29cc63/gcc-4.8.1-glibc-2.17/da39a3e
+BINUTILS_BASE=/mnt/gvfs/third-party2/binutils/71454c53fffcb716a0beb9a90047aff7fb5c984a/2.26/centos6-native/da39a3e
+VALGRIND_BASE=/mnt/gvfs/third-party2/valgrind/af85c56f424cd5edfc2c97588299b44ecdec96bb/3.8.1/gcc-4.8.1-glibc-2.17/c3f970a
@@ -0,0 +1,164 @@
+# Copyright (c) 2016-present, Facebook, Inc.  All rights reserved.
+# This source code is licensed under the BSD-style license found in the
+# LICENSE file in the root directory of this source tree. An additional grant
+# of patent rights can be found in the PATENTS file in the same directory.
+
+'''Filter for error messages in test output:
+    - Receives merged stdout/stderr from test on stdin
+    - Finds patterns of known error messages for test name (first argument)
+    - Prints those error messages to stdout
+'''
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+from __future__ import unicode_literals
+
+import re
+import sys
+
+
+class ErrorParserBase(object):
+    def parse_error(self, line):
+        '''Parses a line of test output. If it contains an error, returns a
+        formatted message describing the error; otherwise, returns None.
+        Subclasses must override this method.
+        '''
+        raise NotImplementedError
+
+
+class GTestErrorParser(ErrorParserBase):
+    '''A parser that remembers the last test that began running so it can print
+    that test's name upon detecting failure.
+    '''
+    _GTEST_NAME_PATTERN = re.compile(r'\[ RUN      \] (\S+)$')
+    # format: '<filename or "unknown file">:<line #>: Failure'
+    _GTEST_FAIL_PATTERN = re.compile(r'(unknown file|\S+:\d+): Failure$')
+
+    def __init__(self):
+        self._last_gtest_name = 'Unknown test'
+
+    def parse_error(self, line):
+        gtest_name_match = self._GTEST_NAME_PATTERN.match(line)
+        if gtest_name_match:
+            self._last_gtest_name = gtest_name_match.group(1)
+            return None
+        gtest_fail_match = self._GTEST_FAIL_PATTERN.match(line)
+        if gtest_fail_match:
+            return '%s failed: %s' % (
+                    self._last_gtest_name, gtest_fail_match.group(1))
+        return None
+
+
+class MatchErrorParser(ErrorParserBase):
+    '''A simple parser that returns the whole line if it matches the pattern.
+    '''
+    def __init__(self, pattern):
+        self._pattern = re.compile(pattern)
+
+    def parse_error(self, line):
+        if self._pattern.match(line):
+            return line
+        return None
+
+
+class CompilerErrorParser(MatchErrorParser):
+    def __init__(self):
+        # format: '<filename>:<line #>:<column #>: error: <error msg>'
+        super(CompilerErrorParser, self).__init__(r'\S+:\d+:\d+: error:')
+
+
+class ScanBuildErrorParser(MatchErrorParser):
+    def __init__(self):
+        super(ScanBuildErrorParser, self).__init__(
+                r'scan-build: \d+ bugs found.$')
+
+
+class DbCrashErrorParser(MatchErrorParser):
+    def __init__(self):
+        super(DbCrashErrorParser, self).__init__(r'\*\*\*.*\^$|TEST FAILED.')
+
+
+class WriteStressErrorParser(MatchErrorParser):
+    def __init__(self):
+        super(WriteStressErrorParser, self).__init__(
+                r'ERROR: write_stress died with exitcode=\d+')
+
+
+class AsanErrorParser(MatchErrorParser):
+    def __init__(self):
+        super(AsanErrorParser, self).__init__(
+                r'==\d+==ERROR: AddressSanitizer:')
+
+
+class UbsanErrorParser(MatchErrorParser):
+    def __init__(self):
+        # format: '<filename>:<line #>:<column #>: runtime error: <error msg>'
+        super(UbsanErrorParser, self).__init__(r'\S+:\d+:\d+: runtime error:')
+
+
+class ValgrindErrorParser(MatchErrorParser):
+    def __init__(self):
+        # just grab the summary, valgrind doesn't clearly distinguish errors
+        # from other log messages.
+        super(ValgrindErrorParser, self).__init__(r'==\d+== ERROR SUMMARY:')
+
+
+class CompatErrorParser(MatchErrorParser):
+    def __init__(self):
+        super(CompatErrorParser, self).__init__(r'==== .*[Ee]rror.* ====$')
+
+
+class TsanErrorParser(MatchErrorParser):
+    def __init__(self):
+        super(TsanErrorParser, self).__init__(r'WARNING: ThreadSanitizer:')
+
+
+_TEST_NAME_TO_PARSERS = {
+    'punit': [CompilerErrorParser, GTestErrorParser],
+    'unit': [CompilerErrorParser, GTestErrorParser],
+    'unit_481': [CompilerErrorParser, GTestErrorParser],
+    'clang_unit': [CompilerErrorParser, GTestErrorParser],
+    'clang_analyze': [CompilerErrorParser, ScanBuildErrorParser],
+    'code_cov': [CompilerErrorParser, GTestErrorParser],
+    'unity': [CompilerErrorParser, GTestErrorParser],
+    'lite': [CompilerErrorParser],
+    'lite_test': [CompilerErrorParser, GTestErrorParser],
+    'stress_crash': [CompilerErrorParser, DbCrashErrorParser],
+    'write_stress': [CompilerErrorParser, WriteStressErrorParser],
+    'asan': [CompilerErrorParser, GTestErrorParser, AsanErrorParser],
+    'asan_crash': [CompilerErrorParser, AsanErrorParser, DbCrashErrorParser],
+    'ubsan': [CompilerErrorParser, GTestErrorParser, UbsanErrorParser],
+    'ubsan_crash': [CompilerErrorParser, UbsanErrorParser, DbCrashErrorParser],
+    'valgrind': [CompilerErrorParser, GTestErrorParser, ValgrindErrorParser],
+    'tsan': [CompilerErrorParser, GTestErrorParser, TsanErrorParser],
+    'format_compatible': [CompilerErrorParser, CompatErrorParser],
+    'run_format_compatible': [CompilerErrorParser, CompatErrorParser],
+    'no_compression': [CompilerErrorParser, GTestErrorParser],
+    'run_no_compression': [CompilerErrorParser, GTestErrorParser],
+    'regression': [CompilerErrorParser],
+    'run_regression': [CompilerErrorParser],
+}
+
+
+def main():
+    if len(sys.argv) != 2:
+        return 'Usage: %s <test name>' % sys.argv[0]
+    test_name = sys.argv[1]
+    if test_name not in _TEST_NAME_TO_PARSERS:
+        return 'Unknown test name: %s' % test_name
+
+    error_parsers = []
+    for parser_cls in _TEST_NAME_TO_PARSERS[test_name]:
+        error_parsers.append(parser_cls())
+
+    for line in sys.stdin:
+        line = line.strip()
+        for error_parser in error_parsers:
+            error_msg = error_parser.parse_error(line)
+            if error_msg is not None:
+                print(error_msg)
+
+
+if __name__ == '__main__':
+    sys.exit(main())
@@ -6,103 +6,105 @@
 # Environment variables that change the behavior of this script:
 # PIC_BUILD -- if true, it will only take pic versions of libraries from fbcode. libraries that don't have pic variant will not be included

+
+BASEDIR=`dirname $BASH_SOURCE`
+source "$BASEDIR/dependencies.sh"
+
 CFLAGS=""

-# location of libgcc
-LIBGCC_BASE="/mnt/gvfs/third-party2/libgcc/0473c80518a10d6efcbe24c5eeca3fb4ec9b519c/4.9.x/gcc-4.9-glibc-2.20/e1a7e4e"
+# libgcc
 LIBGCC_INCLUDE="$LIBGCC_BASE/include"
 LIBGCC_LIBS=" -L $LIBGCC_BASE/libs"

-# location of glibc
-GLIBC_REV=7397bed99280af5d9543439cdb7d018af7542720
-GLIBC_INCLUDE="/mnt/gvfs/third-party2/glibc/$GLIBC_REV/2.20/gcc-4.9-glibc-2.20/99df8fc/include"
-GLIBC_LIBS=" -L /mnt/gvfs/third-party2/glibc/$GLIBC_REV/2.20/gcc-4.9-glibc-2.20/99df8fc/lib"
-
-SNAPPY_INCLUDE=" -I /mnt/gvfs/third-party2/snappy/b0f269b3ca47770121aa159b99e1d8d2ab260e1f/1.0.3/gcc-4.9-glibc-2.20/c32916f/include/"
+# glibc
+GLIBC_INCLUDE="$GLIBC_BASE/include"
+GLIBC_LIBS=" -L $GLIBC_BASE/lib"

+# snappy
+SNAPPY_INCLUDE=" -I $SNAPPY_BASE/include/"
 if test -z $PIC_BUILD; then
-  SNAPPY_LIBS=" /mnt/gvfs/third-party2/snappy/b0f269b3ca47770121aa159b99e1d8d2ab260e1f/1.0.3/gcc-4.9-glibc-2.20/c32916f/lib/libsnappy.a"
+  SNAPPY_LIBS=" $SNAPPY_BASE/lib/libsnappy.a"
 else
-  SNAPPY_LIBS=" /mnt/gvfs/third-party2/snappy/b0f269b3ca47770121aa159b99e1d8d2ab260e1f/1.0.3/gcc-4.9-glibc-2.20/c32916f/lib/libsnappy_pic.a"
+  SNAPPY_LIBS=" $SNAPPY_BASE/lib/libsnappy_pic.a"
 fi
-
 CFLAGS+=" -DSNAPPY"

 if test -z $PIC_BUILD; then
  # location of zlib headers and libraries
-  ZLIB_INCLUDE=" -I /mnt/gvfs/third-party2/zlib/feb983d9667f4cf5e9da07ce75abc824764b67a1/1.2.8/gcc-4.9-glibc-2.20/4230243/include/"
-  ZLIB_LIBS=" /mnt/gvfs/third-party2/zlib/feb983d9667f4cf5e9da07ce75abc824764b67a1/1.2.8/gcc-4.9-glibc-2.20/4230243/lib/libz.a"
+  ZLIB_INCLUDE=" -I $ZLIB_BASE/include/"
+  ZLIB_LIBS=" $ZLIB_BASE/lib/libz.a"
  CFLAGS+=" -DZLIB"

  # location of bzip headers and libraries
-  BZIP_INCLUDE=" -I /mnt/gvfs/third-party2/bzip2/af004cceebb2dfd173ca29933ea5915e727aad2f/1.0.6/gcc-4.9-glibc-2.20/4230243/include/"
-  BZIP_LIBS=" /mnt/gvfs/third-party2/bzip2/af004cceebb2dfd173ca29933ea5915e727aad2f/1.0.6/gcc-4.9-glibc-2.20/4230243/lib/libbz2.a"
+  BZIP_INCLUDE=" -I $BZIP2_BASE/include/"
+  BZIP_LIBS=" $BZIP2_BASE/lib/libbz2.a"
  CFLAGS+=" -DBZIP2"

-  LZ4_INCLUDE=" -I /mnt/gvfs/third-party2/lz4/6858fac689e0f92e584224d91bdb0e39f6c8320d/r131/gcc-4.9-glibc-2.20/e9936bf/include/"
-  LZ4_LIBS=" /mnt/gvfs/third-party2/lz4/6858fac689e0f92e584224d91bdb0e39f6c8320d/r131/gcc-4.9-glibc-2.20/e9936bf/lib/liblz4.a"
+  LZ4_INCLUDE=" -I $LZ4_BASE/include/"
+  LZ4_LIBS=" $LZ4_BASE/lib/liblz4.a"
  CFLAGS+=" -DLZ4"

-  ZSTD_REV=810b81b4705def5243e998b54701f3c504e4009e
-  ZSTD_INCLUDE=" -I /mnt/gvfs/third-party2/zstd/$ZSTD_REV/0.4.2/gcc-4.8.1-glibc-2.17/c3f970a/include"
-  ZSTD_LIBS=" /mnt/gvfs/third-party2/zstd/$ZSTD_REV/0.4.2/gcc-4.8.1-glibc-2.17/c3f970a/lib/libzstd.a"
+  ZSTD_INCLUDE=" -I $ZSTD_BASE/include/"
+  ZSTD_LIBS=" $ZSTD_BASE/lib/libzstd.a"
  CFLAGS+=" -DZSTD"
 fi

 # location of gflags headers and libraries
-GFLAGS_INCLUDE=" -I /mnt/gvfs/third-party2/gflags/0fa60e2b88de3e469db6c482d6e6dac72f5d65f9/1.6/gcc-4.9-glibc-2.20/4230243/include/"
+GFLAGS_INCLUDE=" -I $GFLAGS_BASE/include/"
 if test -z $PIC_BUILD; then
-  GFLAGS_LIBS=" /mnt/gvfs/third-party2/gflags/0fa60e2b88de3e469db6c482d6e6dac72f5d65f9/1.6/gcc-4.9-glibc-2.20/4230243/lib/libgflags.a"
+  GFLAGS_LIBS=" $GFLAGS_BASE/lib/libgflags.a"
 else
-  GFLAGS_LIBS=" /mnt/gvfs/third-party2/gflags/0fa60e2b88de3e469db6c482d6e6dac72f5d65f9/1.6/gcc-4.9-glibc-2.20/4230243/lib/libgflags_pic.a"
+  GFLAGS_LIBS=" $GFLAGS_BASE/lib/libgflags_pic.a"
 fi
 CFLAGS+=" -DGFLAGS=google"

 # location of jemalloc
-JEMALLOC_INCLUDE=" -I /mnt/gvfs/third-party2/jemalloc/bcd68e5e419efa4e61b9486d6854564d6d75a0b5/3.6.0/gcc-4.9-glibc-2.20/2aafc78/include/"
-JEMALLOC_LIB=" /mnt/gvfs/third-party2/jemalloc/bcd68e5e419efa4e61b9486d6854564d6d75a0b5/3.6.0/gcc-4.9-glibc-2.20/2aafc78/lib/libjemalloc.a"
+JEMALLOC_INCLUDE=" -I $JEMALLOC_BASE/include/"
+JEMALLOC_LIB=" $JEMALLOC_BASE/lib/libjemalloc.a"

 if test -z $PIC_BUILD; then
  # location of numa
-  NUMA_INCLUDE=" -I /mnt/gvfs/third-party2/numa/bbefc39ecbf31d0ca184168eb613ef8d397790ee/2.0.8/gcc-4.9-glibc-2.20/4230243/include/"
-  NUMA_LIB=" /mnt/gvfs/third-party2/numa/bbefc39ecbf31d0ca184168eb613ef8d397790ee/2.0.8/gcc-4.9-glibc-2.20/4230243/lib/libnuma.a"
+  NUMA_INCLUDE=" -I $NUMA_BASE/include/"
+  NUMA_LIB=" $NUMA_BASE/lib/libnuma.a"
  CFLAGS+=" -DNUMA"

  # location of libunwind
-  LIBUNWIND="/mnt/gvfs/third-party2/libunwind/1de3b75e0afedfe5585b231bbb340ec7a1542335/1.1/gcc-4.9-glibc-2.20/34235e8/lib/libunwind.a"
+  LIBUNWIND="$LIBUNWIND_BASE/lib/libunwind.a"
 fi

 # use Intel SSE support for checksum calculations
 export USE_SSE=1

-BINUTILS="/mnt/gvfs/third-party2/binutils/0b6ad0c88ddd903333a48ae8bff134efac468e4a/2.25/centos6-native/da39a3e/bin"
+BINUTILS="$BINUTILS_BASE/bin"
 AR="$BINUTILS/ar"

 DEPS_INCLUDE="$SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP_INCLUDE $LZ4_INCLUDE $ZSTD_INCLUDE $GFLAGS_INCLUDE $NUMA_INCLUDE"

-GCC_BASE="/mnt/gvfs/third-party2/gcc/1c67a0b88f64d4d9ced0382d141c76aaa7d62fba/4.9.x/centos6-native/1317bc4"
 STDLIBS="-L $GCC_BASE/lib64"

-CLANG_BASE="/mnt/gvfs/third-party2/clang/d81444dd214df3d2466734de45bb264a0486acc3/dev"
-CLANG_BIN="$CLANG_BASE/centos6-native/af4b1a0/bin"
+CLANG_BIN="$CLANG_BASE/bin"
+CLANG_LIB="$CLANG_BASE/lib"
+CLANG_SRC="$CLANG_BASE/../../src"
+
 CLANG_ANALYZER="$CLANG_BIN/clang++"
-CLANG_SCAN_BUILD="$CLANG_BASE/src/clang/tools/scan-build/scan-build"
+CLANG_SCAN_BUILD="$CLANG_SRC/clang/tools/scan-build/scan-build"

 if [ -z "$USE_CLANG" ]; then
  # gcc
  CC="$GCC_BASE/bin/gcc"
  CXX="$GCC_BASE/bin/g++"
-  
+
  CFLAGS+=" -B$BINUTILS/gold"
  CFLAGS+=" -isystem $GLIBC_INCLUDE"
  CFLAGS+=" -isystem $LIBGCC_INCLUDE"
+  JEMALLOC=1
 else
-  # clang 
-  CLANG_INCLUDE="$CLANG_BASE/gcc-4.9-glibc-2.20/74c386f/lib/clang/dev/include/"
+  # clang
+  CLANG_INCLUDE="$CLANG_LIB/clang/*/include"
  CC="$CLANG_BIN/clang"
  CXX="$CLANG_BIN/clang++"

-  KERNEL_HEADERS_INCLUDE="/mnt/gvfs/third-party2/kernel-headers/ffd14f660a43c4b92717986b1bba66722ef089d0/3.2.18_70_fbk11_00129_gc8882d0/gcc-4.9-glibc-2.20/da39a3e/include"
+  KERNEL_HEADERS_INCLUDE="$KERNEL_HEADERS_BASE/include"

  CFLAGS+=" -B$BINUTILS/gold -nostdinc -nostdlib"
  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/4.9.x "
@@ -128,6 +130,6 @@ PLATFORM_LDFLAGS="$LIBGCC_LIBS $GLIBC_LIBS $STDLIBS -lgcc -lstdc++"

 EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS"

-VALGRIND_VER="/mnt/gvfs/third-party2/valgrind/6c45ef049cbf11c2df593addb712cd891049e737/3.10.0/gcc-4.9-glibc-2.20/4230243/bin/"
+VALGRIND_VER="$VALGRIND_BASE/bin/"

 export CC CXX AR CFLAGS CXXFLAGS EXEC_LDFLAGS EXEC_LDFLAGS_SHARED VALGRIND_VER JEMALLOC_LIB JEMALLOC_INCLUDE CLANG_ANALYZER CLANG_SCAN_BUILD
@@ -4,85 +4,78 @@
 # fbcode settings.  It uses the latest g++ compiler and also
 # uses jemalloc

+BASEDIR=`dirname $BASH_SOURCE`
+source "$BASEDIR/dependencies_4.8.1.sh"
+
 # location of libgcc
-LIBGCC_BASE="/mnt/gvfs/third-party2/libgcc/d00277f4559e261ed0a81f30f23c0ce5564e359e/4.8.1/gcc-4.8.1-glibc-2.17/8aac7fc"
 LIBGCC_INCLUDE="$LIBGCC_BASE/include"
 LIBGCC_LIBS=" -L $LIBGCC_BASE/libs"

 # location of glibc
-GLIBC_REV=0600c95b31226b5e535614c590677d87c62d8016
-GLIBC_INCLUDE="/mnt/gvfs/third-party2/glibc/$GLIBC_REV/2.17/gcc-4.8.1-glibc-2.17/99df8fc/include"
-GLIBC_LIBS=" -L /mnt/gvfs/third-party2/glibc/$GLIBC_REV/2.17/gcc-4.8.1-glibc-2.17/99df8fc/lib"
+GLIBC_INCLUDE="$GLIBC_BASE/include"
+GLIBC_LIBS=" -L $GLIBC_BASE/lib"

 # location of snappy headers and libraries
-SNAPPY_REV=cbf6f1f209e5bd160bdc5d971744e039f36b1566
-SNAPPY_INCLUDE=" -I /mnt/gvfs/third-party2/snappy/$SNAPPY_REV/1.1.3/gcc-4.8.1-glibc-2.17/c3f970a/include"
-SNAPPY_LIBS=" /mnt/gvfs/third-party2/snappy/$SNAPPY_REV/1.1.3/gcc-4.8.1-glibc-2.17/c3f970a/lib/libsnappy.a"
+SNAPPY_INCLUDE=" -I $SNAPPY_BASE/include"
+SNAPPY_LIBS=" $SNAPPY_BASE/lib/libsnappy.a"

 # location of zlib headers and libraries
-ZLIB_REV=6d39cb54708049f527e713ad19f2aadb9d3667e8
-ZLIB_INCLUDE=" -I /mnt/gvfs/third-party2/zlib/$ZLIB_REV/1.2.8/gcc-4.8.1-glibc-2.17/c3f970a/include"
-ZLIB_LIBS=" /mnt/gvfs/third-party2/zlib/$ZLIB_REV/1.2.8/gcc-4.8.1-glibc-2.17/c3f970a/lib/libz.a"
+ZLIB_INCLUDE=" -I $ZLIB_BASE/include"
+ZLIB_LIBS=" $ZLIB_BASE/lib/libz.a"

 # location of bzip headers and libraries
-BZIP_REV=d6c789bfc2ec4c51a63d66df2878926b8158cde8
-BZIP_INCLUDE=" -I /mnt/gvfs/third-party2/bzip2/$BZIP_REV/1.0.6/gcc-4.8.1-glibc-2.17/c3f970a/include/"
-BZIP_LIBS=" /mnt/gvfs/third-party2/bzip2/$BZIP_REV/1.0.6/gcc-4.8.1-glibc-2.17/c3f970a/lib/libbz2.a"
+BZIP2_INCLUDE=" -I $BZIP2_BASE/include/"
+BZIP2_LIBS=" $BZIP2_BASE/lib/libbz2.a"

-LZ4_REV=6858fac689e0f92e584224d91bdb0e39f6c8320d
-LZ4_INCLUDE=" -I /mnt/gvfs/third-party2/lz4/$LZ4_REV/r131/gcc-4.8.1-glibc-2.17/c3f970a/include"
-LZ4_LIBS=" /mnt/gvfs/third-party2/lz4/$LZ4_REV/r131/gcc-4.8.1-glibc-2.17/c3f970a/lib/liblz4.a"
+LZ4_INCLUDE=" -I $LZ4_BASE/include"
+LZ4_LIBS=" $LZ4_BASE/lib/liblz4.a"

-ZSTD_REV=810b81b4705def5243e998b54701f3c504e4009e
-ZSTD_INCLUDE=" -I /mnt/gvfs/third-party2/zstd/$ZSTD_REV/0.4.2/gcc-4.8.1-glibc-2.17/c3f970a/include"
-ZSTD_LIBS=" /mnt/gvfs/third-party2/zstd/$ZSTD_REV/0.4.2/gcc-4.8.1-glibc-2.17/c3f970a/lib/libzstd.a"
+ZSTD_INCLUDE=" -I $ZSTD_BASE/include"
+ZSTD_LIBS=" $ZSTD_BASE/lib/libzstd.a"

 # location of gflags headers and libraries
-GFLAGS_REV=c7275a4ceae0aca0929e56964a31dafc53c1ee96
-GFLAGS_INCLUDE=" -I /mnt/gvfs/third-party2/gflags/$GFLAGS_REV/2.1.1/gcc-4.8.1-glibc-2.17/c3f970a/include/"
-GFLAGS_LIBS=" /mnt/gvfs/third-party2/gflags/$GFLAGS_REV/2.1.1/gcc-4.8.1-glibc-2.17/c3f970a/lib/libgflags.a"
+GFLAGS_INCLUDE=" -I $GFLAGS_BASE/include/"
+GFLAGS_LIBS=" $GFLAGS_BASE/lib/libgflags.a"

 # location of jemalloc
-JEMALLOC_REV=c370265e58c4b6602e798df23335a1e9913dae52
-JEMALLOC_INCLUDE=" -I /mnt/gvfs/third-party2/jemalloc/$JEMALLOC_REV/4.0.3/gcc-4.8.1-glibc-2.17/8d31e51/include"
-JEMALLOC_LIB="/mnt/gvfs/third-party2/jemalloc/$JEMALLOC_REV/4.0.3/gcc-4.8.1-glibc-2.17/8d31e51/lib/libjemalloc.a"
+JEMALLOC_INCLUDE=" -I $JEMALLOC_BASE/include"
+JEMALLOC_LIB="$JEMALLOC_BASE/lib/libjemalloc.a"

 # location of numa
-NUMA_REV=ae54a5ed22cdabb1c6446dce4e8ffae5b4446d73
-NUMA_INCLUDE=" -I /mnt/gvfs/third-party2/numa/$NUMA_REV/2.0.8/gcc-4.8.1-glibc-2.17/c3f970a/include/"
-NUMA_LIB=" /mnt/gvfs/third-party2/numa/$NUMA_REV/2.0.8/gcc-4.8.1-glibc-2.17/c3f970a/lib/libnuma.a"
+NUMA_INCLUDE=" -I $NUMA_BASE/include/"
+NUMA_LIB=" $NUMA_BASE/lib/libnuma.a"

 # location of libunwind
-LIBUNWIND_REV=121f1a75c4414683aea8c70b761bfaf187f7c1a3
-LIBUNWIND="/mnt/gvfs/third-party2/libunwind/$LIBUNWIND_REV/trunk/gcc-4.8.1-glibc-2.17/675d945/lib/libunwind.a"
+LIBUNWIND="$LIBUNWIND_BASE/lib/libunwind.a"

 # use Intel SSE support for checksum calculations
 export USE_SSE=1

-BINUTILS="/mnt/gvfs/third-party2/binutils/75670d0d8ef4891fd1ec2a7513ef01cd002c823b/2.25/centos6-native/da39a3e/bin"
+BINUTILS="$BINUTILS_BASE/bin"
 AR="$BINUTILS/ar"

-DEPS_INCLUDE="$SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP_INCLUDE $LZ4_INCLUDE $ZSTD_INCLUDE $GFLAGS_INCLUDE $NUMA_INCLUDE"
+DEPS_INCLUDE="$SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP2_INCLUDE $LZ4_INCLUDE $ZSTD_INCLUDE $GFLAGS_INCLUDE $NUMA_INCLUDE"

-GCC_BASE="/mnt/gvfs/third-party2/gcc/c0064002d2609ab649603f769f0bd110bbe48029/4.8.1/centos6-native/cc6c9dc"
 STDLIBS="-L $GCC_BASE/lib64"

 if [ -z "$USE_CLANG" ]; then
  # gcc
  CC="$GCC_BASE/bin/gcc"
  CXX="$GCC_BASE/bin/g++"
-  
+
  CFLAGS="-B$BINUTILS/gold -m64 -mtune=generic"
  CFLAGS+=" -isystem $GLIBC_INCLUDE"
  CFLAGS+=" -isystem $LIBGCC_INCLUDE"
+  JEMALLOC=1
 else
-  # clang 
-  CLANG_BASE="/mnt/gvfs/third-party2/clang/ab054e9a490a8fd4537c0b6ec56e5c91c0f81c91/3.7"
-  CLANG_INCLUDE="$CLANG_BASE/gcc-4.8.1-glibc-2.17/ee9b060/lib/clang/3.7/include"
-  CC="$CLANG_BASE/centos6-native/b2feaee/bin/clang"
-  CXX="$CLANG_BASE/centos6-native/b2feaee/bin/clang++"
+  # clang
+  CLANG_BIN="$CLANG_BASE/bin"
+  CLANG_LIB="$CLANG_BASE/lib"
+  CLANG_INCLUDE="$CLANG_LIB/clang/*/include"
+  CC="$CLANG_BIN/clang"
+  CXX="$CLANG_BIN/clang++"

-  KERNEL_HEADERS_INCLUDE="/mnt/gvfs/third-party2/kernel-headers/1a48835975c66d30e47770ec419758ed3b9ba010/3.10.62-62_fbk17_03959_ge29cc63/gcc-4.8.1-glibc-2.17/da39a3e/include/"
+  KERNEL_HEADERS_INCLUDE="$KERNEL_HEADERS_BASE/include/"

  CFLAGS="-B$BINUTILS/gold -nostdinc -nostdlib"
  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/4.8.1 "
@@ -100,16 +93,15 @@ CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PR
 CFLAGS+=" -DSNAPPY -DGFLAGS=google -DZLIB -DBZIP2 -DLZ4 -DZSTD -DNUMA"
 CXXFLAGS+=" $CFLAGS"

-EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB"
+EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $BZIP2_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB"
 EXEC_LDFLAGS+=" -Wl,--dynamic-linker,/usr/local/fbcode/gcc-4.8.1-glibc-2.17/lib/ld.so"
 EXEC_LDFLAGS+=" $LIBUNWIND"
 EXEC_LDFLAGS+=" -Wl,-rpath=/usr/local/fbcode/gcc-4.8.1-glibc-2.17/lib"

 PLATFORM_LDFLAGS="$LIBGCC_LIBS $GLIBC_LIBS $STDLIBS -lgcc -lstdc++"

-EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS"
+EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP2_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS"

-VALGRIND_REV=af85c56f424cd5edfc2c97588299b44ecdec96bb
-VALGRIND_VER="/mnt/gvfs/third-party2/valgrind/$VALGRIND_REV/3.8.1/gcc-4.8.1-glibc-2.17/c3f970a/bin/"
+VALGRIND_VER="$VALGRIND_BASE/bin/"

 export CC CXX AR CFLAGS CXXFLAGS EXEC_LDFLAGS EXEC_LDFLAGS_SHARED VALGRIND_VER JEMALLOC_LIB JEMALLOC_INCLUDE
@@ -1,5 +1,5 @@
 #!/bin/bash
-#  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+#  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 #  This source code is licensed under the BSD-style license found in the
 #  LICENSE file in the root directory of this source tree. An additional grant
 #  of patent rights can be found in the PATENTS file in the same directory.
@@ -10,7 +10,7 @@ then
  GIT="git"
 fi

-# Print out the colored progress info so that it can be brainlessly 
+# Print out the colored progress info so that it can be brainlessly
 # distinguished by users.
 function title() {
  echo -e "\033[1;32m$*\033[0m"
@@ -0,0 +1,198 @@
+#!/usr/local/fbcode/gcc-4.8.1-glibc-2.17-fb/bin/python2.7
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+from __future__ import unicode_literals
+import argparse
+import commands
+import subprocess
+import sys
+import re
+import os
+import time
+
+#
+# Simple logger
+#
+
+
+class Log:
+
+    LOG_FILE = "/tmp/precommit-check.log"
+
+    def __init__(self):
+        self.filename = Log.LOG_FILE
+        self.f = open(self.filename, 'w+', 0)
+
+    def caption(self, str):
+        line = "\n##### %s #####\n" % str
+        if self.f:
+            self.f.write("%s \n" % line)
+        else:
+            print(line)
+
+    def error(self, str):
+        data = "\n\n##### ERROR ##### %s" % str
+        if self.f:
+            self.f.write("%s \n" % data)
+        else:
+            print(data)
+
+    def log(self, str):
+        if self.f:
+            self.f.write("%s \n" % str)
+        else:
+            print(str)
+
+#
+# Shell Environment
+#
+
+
+class Env(object):
+
+    def __init__(self, tests):
+        self.tests = tests
+        self.log = Log()
+
+    def shell(self, cmd, path=os.getcwd()):
+        if path:
+            os.chdir(path)
+
+        self.log.log("==== shell session ===========================")
+        self.log.log("%s> %s" % (path, cmd))
+        status = subprocess.call("cd %s; %s" % (path, cmd), shell=True,
+                                 stdout=self.log.f, stderr=self.log.f)
+        self.log.log("status = %s" % status)
+        self.log.log("============================================== \n\n")
+        return status
+
+    def GetOutput(self, cmd, path=os.getcwd()):
+        if path:
+            os.chdir(path)
+
+        self.log.log("==== shell session ===========================")
+        self.log.log("%s> %s" % (path, cmd))
+        status, out = commands.getstatusoutput(cmd)
+        self.log.log("status = %s" % status)
+        self.log.log("out = %s" % out)
+        self.log.log("============================================== \n\n")
+        return status, out
+
+#
+# Pre-commit checker
+#
+
+
+class PreCommitChecker(Env):
+
+    def __init__(self, tests):
+        Env.__init__(self, tests)
+
+    #
+    #   Get commands for a given job from the determinator file
+    #
+    def get_commands(self, test):
+        status, out = self.GetOutput(
+            "build_tools/rocksdb-lego-determinator %s" % test, ".")
+        return status, out
+
+    #
+    # Run a specific CI job
+    #
+    def run_test(self, test):
+        self.log.caption("Running test %s locally" % test)
+
+        # get commands for the CI job determinator
+        status, cmds = self.get_commands(test)
+        if status != 0:
+            self.log.error("Error getting commands for test %s" % test)
+            return False
+
+        # Parse the JSON to extract the commands to run
+        cmds = re.findall("'shell':'([^\']*)'", cmds)
+
+        if len(cmds) == 0:
+            self.log.log("No commands found")
+            return False
+
+        # Run commands
+        for cmd in cmds:
+            # Replace J=<..> with the local environment variable
+            if "J" in os.environ:
+                cmd = cmd.replace("J=1", "J=%s" % os.environ["J"])
+                cmd = cmd.replace("make ", "make -j%s " % os.environ["J"])
+            # Run the command
+            status = self.shell(cmd, ".")
+            if status != 0:
+                self.log.error("Error running command %s for test %s"
+                               % (cmd, test))
+                return False
+
+        return True
+
+    #
+    # Run specified CI jobs
+    #
+    def run_tests(self):
+        if not self.tests:
+            self.log.error("Invalid args. Please provide tests")
+            return False
+
+        self.print_separator()
+        self.print_row("TEST", "RESULT")
+        self.print_separator()
+
+        for test in self.tests:
+            start_time = time.time()
+            self.print_test(test)
+            result = self.run_test(test)
+            elapsed_min = (time.time() - start_time) / 60
+            if not result:
+                self.log.error("Error running test %s" % test)
+                self.print_result("FAIL (%dm)" % elapsed_min)
+                return False
+            self.print_result("PASS (%dm)" % elapsed_min)
+
+        self.print_separator()
+        return True
+
+    #
+    # Print a line
+    #
+    def print_separator(self):
+        print("".ljust(60, "-"))
+
+    #
+    # Print two colums
+    #
+    def print_row(self, c0, c1):
+        print("%s%s" % (c0.ljust(40), c1.ljust(20)))
+
+    def print_test(self, test):
+        print(test.ljust(40), end="")
+        sys.stdout.flush()
+
+    def print_result(self, result):
+        print(result.ljust(20))
+
+#
+# Main
+#
+parser = argparse.ArgumentParser(description='RocksDB pre-commit checker.')
+
+# <test ....>
+parser.add_argument('test', nargs='+',
+                    help='CI test(s) to run. e.g: unit punit asan tsan ubsan')
+
+print("Please follow log %s" % Log.LOG_FILE)
+
+args = parser.parse_args()
+checker = PreCommitChecker(args.test)
+
+if not checker.run_tests():
+    print("Error running tests. Please check log file %s" % Log.LOG_FILE)
+    sys.exit(1)
+
+sys.exit(0)
@@ -67,8 +67,11 @@ ASAN="COMPILE_WITH_ASAN=1"
 CLANG="USE_CLANG=1"
 LITE="OPT=-DROCKSDB_LITE"
 TSAN="COMPILE_WITH_TSAN=1"
+UBSAN="COMPILE_WITH_UBSAN=1"
 DISABLE_JEMALLOC="DISABLE_JEMALLOC=1"
-PARSER="'parser':'egrep \'Failure|^#|Abort|Expected|Actual|GoogleTestFailure|^==\''"
+HTTP_PROXY="https_proxy=http://fwdproxy.29.prn1:8080 http_proxy=http://fwdproxy.29.prn1:8080 ftp_proxy=http://fwdproxy.29.prn1:8080"
+SETUP_JAVA_ENV="export $HTTP_PROXY; export JAVA_HOME=/usr/local/jdk-7u10-64/; export PATH=\$PATH:\$JAVA_HOME/bin"
+PARSER="'parser':'python build_tools/error_filter.py $1'"

 ARTIFACTS=" 'artifacts': [
    {
@@ -376,6 +379,48 @@ ASAN_CRASH_TEST_COMMANDS="[
    }
 ]"

+#
+# RocksDB test under undefined behavior sanitizer
+#
+UBSAN_TEST_COMMANDS="[
+    {
+        'name':'Rocksdb Unit Test under UBSAN',
+        'oncall':'$ONCALL',
+        'steps': [
+            $CLEANUP_ENV,
+            {
+                'name':'Test RocksDB debug under UBSAN',
+                'shell':'set -o pipefail && $SHM $UBSAN $DEBUG make J=1 ubsan_check',
+                'user':'root',
+                $PARSER
+            }
+        ],
+        $REPORT
+    }
+]"
+
+#
+# RocksDB crash testing under udnefined behavior sanitizer
+#
+UBSAN_CRASH_TEST_COMMANDS="[
+    {
+        'name':'Rocksdb crash test under UBSAN',
+        'oncall':'$ONCALL',
+        'timeout': 86400,
+        'steps': [
+            $CLEANUP_ENV,
+            {
+                'name':'Build and run RocksDB debug ubsan_crash_test',
+                'timeout': 86400,
+                'shell':'$SHM $DEBUG make J=1 ubsan_crash_test',
+                'user':'root',
+                $PARSER
+            },
+        ],
+        $REPORT
+    }
+]"
+
 #
 # RocksDB unit test under valgrind
 #
@@ -462,9 +507,9 @@ run_format_compatible()
  if [ -e "build_detect_platform" ]
  then
    sed "s/tcmalloc/nothingnothingnothing/g" build_detect_platform > $TEST_TMPDIR/temp_build_file
-    rm -rf build_detect_platform 
+    rm -rf build_detect_platform
    cp $TEST_TMPDIR/temp_build_file build_detect_platform
-    chmod +x build_detect_platform 
+    chmod +x build_detect_platform
  fi

  make ldb -j32
@@ -592,6 +637,27 @@ REGRESSION_COMMANDS="[
    }
 ]"

+#
+# RocksDB Java build
+#
+JAVA_BUILD_TEST_COMMANDS="[
+    {
+        'name':'Rocksdb Java Build',
+        'oncall':'$ONCALL',
+        'steps': [
+            $CLEANUP_ENV,
+            {
+                'name':'Build RocksDB for Java',
+                'shell':'$SETUP_JAVA_ENV; $SHM make rocksdbjava',
+                'user':'root',
+                $PARSER
+            },
+        ],
+        $REPORT
+    }
+]"
+
+
 case $1 in
  punit)
    echo $PARALLEL_UNIT_TEST_COMMANDS
@@ -632,6 +698,12 @@ case $1 in
  asan_crash)
    echo $ASAN_CRASH_TEST_COMMANDS
    ;;
+  ubsan)
+    echo $UBSAN_TEST_COMMANDS
+    ;;
+  ubsan_crash)
+    echo $UBSAN_CRASH_TEST_COMMANDS
+    ;;
  valgrind)
    echo $VALGRIND_TEST_COMMANDS
    ;;
@@ -659,6 +731,9 @@ case $1 in
  run_regression)
    run_regression
    ;;
+  java_build)
+    echo $JAVA_BUILD_TEST_COMMANDS
+    ;;
  *)
    echo "Invalid determinator command"
    ;;
@@ -9,7 +9,7 @@ Param(
  [string]$WorkFolder = "",  # Direct tests to use that folder
  [int]$Limit = -1, # -1 means run all otherwise limit for testing purposes
  [string]$Exclude = "", # Expect a comma separated list, no spaces
-  [string]$Run = "db_test",  # Run db_test|tests|testname1,testname2...
+  [string]$Run = "db_test",  # Run db_test|db_test2|tests|testname1,testname2...
   # Number of async tasks that would run concurrently. Recommend a number below 64.
   # However, CPU utlization really depends on the storage media. Recommend ram based disk.
   # a value of 1 will run everything serially
@@ -46,15 +46,25 @@ if($WorkFolder -eq "") {
  $Env:TEST_TMPDIR = $WorkFolder
 }

+Write-Output "Root: $RootFolder, WorkFolder: $WorkFolder"
+
 # Use JEMALLOC executables
-if($EnableJE) {
-    $db_test = -Join ($BinariesFolder, "db_test_je.exe")
-} else {
-    $db_test = -Join ($BinariesFolder, "db_test.exe")
+if($Run -ceq "db_test" -or
+   $Run -ceq "db_test2" ) {
+
+   $file_name = $Run
+
+   if($EnableJE) {
+     $file_name += "_je"
+   }
+
+   $file_name += ".exe"
+
+   $db_test = -Join ($BinariesFolder, $file_name)
+
+   Write-Output "Binaries: $BinariesFolder db_test: $db_test"
 }

-Write-Output "Root: $RootFolder, WorkFolder: $WorkFolder"
-Write-Output "Binaries: $BinariesFolder db_test: $db_test"

 #Exclusions that we do not want to run
 $ExcludeTests = New-Object System.Collections.Generic.HashSet[string]
@@ -153,7 +163,8 @@ function Discover-TestBinaries([string]$Pattern, $HashTable) {

 $TestsToRun = [ordered]@{}

-if($Run -ceq "db_test") {
+if($Run -ceq "db_test" -or
+   $Run -ceq "db_test2") {
    Normalize-DbTests -HashTable $TestsToRun
 } elseif($Run -ceq "tests") {
    if($EnableJE) {
@@ -225,7 +236,8 @@ function RunJobs($TestToLog, [int]$ConcurrencyVal, [bool]$AddForRerun)
            Write-Host "Starting $k"
            $log_path = ($TestToLog.$k)

-            if($Run -ceq "db_test") {
+            if($Run -ceq "db_test" -or
+               $Run -ceq "db_test2") {
              $job = Start-Job -Name $k -ScriptBlock $InvokeTestCase -ArgumentList @($db_test,$k,$log_path)
            } else {
              [string]$Exe =  -Join ($BinariesFolder, $k)
@@ -0,0 +1,127 @@
+#!/bin/sh
+#
+# Update dependencies.sh file with the latest avaliable versions
+
+BASEDIR=$(dirname $0)
+OUTPUT=""
+
+function log_variable()
+{
+  echo "$1=${!1}" >> "$OUTPUT"
+}
+
+
+TP2_LATEST="/mnt/vol/engshare/fbcode/third-party2"
+## $1 => lib name
+## $2 => lib version (if not provided, will try to pick latest)
+## $3 => platform (if not provided, will try to pick latest gcc)
+##
+## get_lib_base will set a variable named ${LIB_NAME}_BASE to the lib location
+function get_lib_base()
+{
+  local lib_name=$1
+  local lib_version=$2
+  local lib_platform=$3
+
+  local result="$TP2_LATEST/$lib_name/"
+  
+  # Lib Version
+  if [ -z "$lib_version" ] || [ "$lib_version" = "LATEST" ]; then
+    # version is not provided, use latest
+    result=`ls -dr1v $result/*/ | head -n1`
+  else
+    result="$result/$lib_version/"
+  fi
+  
+  # Lib Platform
+  if [ -z "$lib_platform" ]; then
+    # platform is not provided, use latest gcc
+    result=`ls -dr1v $result/gcc-*[^fb]/ | head -n1`
+  else
+    result="$result/$lib_platform/"
+  fi
+  
+  result=`ls -1d $result/*/ | head -n1`
+  
+  # lib_name => LIB_NAME_BASE
+  local __res_var=${lib_name^^}"_BASE"
+  __res_var=`echo $__res_var | tr - _`
+  # LIB_NAME_BASE=$result
+  eval $__res_var=`readlink -f $result`
+  
+  log_variable $__res_var
+}
+
+###########################################################
+#                   4.9.x dependencies                    #
+###########################################################
+
+OUTPUT="$BASEDIR/dependencies.sh"
+
+rm -f "$OUTPUT"
+touch "$OUTPUT"
+
+echo "Writing dependencies to $OUTPUT"
+
+# Compilers locations
+GCC_BASE=`ls -d1 $TP2_LATEST/gcc/4.9.x/centos6-native/*/ | head -n1`
+CLANG_BASE=`ls -d1 /mnt/gvfs/third-party2/clang/fc904e50a9266b9d7b98cae1993afa0c5aae1440/3.7.1/centos6-native/*/ | head -n1`
+
+log_variable GCC_BASE
+log_variable CLANG_BASE
+
+# Libraries locations
+get_lib_base libgcc     4.9.x
+get_lib_base glibc      2.20 
+get_lib_base snappy     LATEST
+get_lib_base zlib       LATEST
+get_lib_base bzip2      LATEST
+get_lib_base lz4        LATEST
+get_lib_base zstd       LATEST
+get_lib_base gflags     LATEST
+get_lib_base jemalloc   LATEST
+get_lib_base numa       LATEST
+get_lib_base libunwind  LATEST
+
+get_lib_base kernel-headers LATEST 
+get_lib_base binutils   LATEST centos6-native 
+get_lib_base valgrind   LATEST
+
+git diff $OUTPUT
+
+###########################################################
+#                   4.8.1 dependencies                    #
+###########################################################
+
+OUTPUT="$BASEDIR/dependencies_4.8.1.sh"
+
+rm -f "$OUTPUT"
+touch "$OUTPUT"
+
+echo "Writing 4.8.1 dependencies to $OUTPUT"
+
+# Compilers locations
+GCC_BASE=`ls -d1 $TP2_LATEST/gcc/4.8.1/centos6-native/*/ | head -n1`
+CLANG_BASE=`ls -d1 /mnt/gvfs/third-party2/clang/fc904e50a9266b9d7b98cae1993afa0c5aae1440/3.7.1/centos6-native/*/ | head -n1`
+
+log_variable GCC_BASE
+log_variable CLANG_BASE
+
+# Libraries locations
+get_lib_base libgcc     4.8.1  gcc-4.8.1-glibc-2.17
+get_lib_base glibc      2.17   gcc-4.8.1-glibc-2.17  
+get_lib_base snappy     LATEST gcc-4.8.1-glibc-2.17
+get_lib_base zlib       LATEST gcc-4.8.1-glibc-2.17
+get_lib_base bzip2      LATEST gcc-4.8.1-glibc-2.17
+get_lib_base lz4        LATEST gcc-4.8.1-glibc-2.17
+get_lib_base zstd       LATEST gcc-4.8.1-glibc-2.17
+get_lib_base gflags     LATEST gcc-4.8.1-glibc-2.17
+get_lib_base jemalloc   LATEST gcc-4.8.1-glibc-2.17
+get_lib_base numa       LATEST gcc-4.8.1-glibc-2.17
+get_lib_base libunwind  LATEST gcc-4.8.1-glibc-2.17
+
+get_lib_base kernel-headers LATEST gcc-4.8.1-glibc-2.17 
+get_lib_base binutils   LATEST centos6-native 
+get_lib_base valgrind   3.8.1  gcc-4.8.1-glibc-2.17
+
+git diff $OUTPUT
@@ -1,9 +1,9 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
 //
-#include "util/auto_roll_logger.h"
+#include "db/auto_roll_logger.h"
 #include "util/mutexlock.h"

 using namespace std;
@@ -12,7 +12,9 @@ namespace rocksdb {

 // -- AutoRollLogger
 Status AutoRollLogger::ResetLogger() {
+  TEST_SYNC_POINT("AutoRollLogger::ResetLogger:BeforeNewLogger");
  status_ = env_->NewLogger(log_fname_, &logger_);
+  TEST_SYNC_POINT("AutoRollLogger::ResetLogger:AfterNewLogger");

  if (!status_.ok()) {
    return status_;
@@ -32,8 +34,16 @@ Status AutoRollLogger::ResetLogger() {
 }

 void AutoRollLogger::RollLogFile() {
-  std::string old_fname = OldInfoLogFileName(
-      dbname_, env_->NowMicros(), db_absolute_path_, db_log_dir_);
+  // This function is called when log is rotating. Two rotations
+  // can happen quickly (NowMicro returns same value). To not overwrite
+  // previous log file we increment by one micro second and try again.
+  uint64_t now = env_->NowMicros();
+  std::string old_fname;
+  do {
+    old_fname = OldInfoLogFileName(
+      dbname_, now, db_absolute_path_, db_log_dir_);
+    now++;
+  } while (env_->FileExists(old_fname).ok());
  env_->RenameFile(log_fname_, old_fname);
 }

@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -13,6 +13,8 @@
 #include "db/filename.h"
 #include "port/port.h"
 #include "port/util_logger.h"
+#include "util/sync_point.h"
+#include "util/mutexlock.h"

 namespace rocksdb {

@@ -53,11 +55,26 @@ class AutoRollLogger : public Logger {
    return status_;
  }

-  size_t GetLogFileSize() const override { return logger_->GetLogFileSize(); }
+  size_t GetLogFileSize() const override {
+    std::shared_ptr<Logger> logger;
+    {
+      MutexLock l(&mutex_);
+      // pin down the current logger_ instance before releasing the mutex.
+      logger = logger_;
+    }
+    return logger->GetLogFileSize();
+  }

  void Flush() override {
-    if (logger_) {
-      logger_->Flush();
+    std::shared_ptr<Logger> logger;
+    {
+      MutexLock l(&mutex_);
+      // pin down the current logger_ instance before releasing the mutex.
+      logger = logger_;
+    }
+    TEST_SYNC_POINT("AutoRollLogger::Flush:PinnedLogger");
+    if (logger) {
+      logger->Flush();
    }
  }

@@ -101,7 +118,7 @@ class AutoRollLogger : public Logger {
  uint64_t ctime_;
  uint64_t cached_now_access_count;
  uint64_t call_NowMicros_every_N_records_;
-  port::Mutex mutex_;
+  mutable port::Mutex mutex_;
 };

 // Facade to craete logger automatically
@@ -1,17 +1,20 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
 //
 #include <string>
+#include <thread>
 #include <vector>
 #include <cmath>
 #include <iostream>
 #include <fstream>
 #include <iterator>
 #include <algorithm>
+#include "db/auto_roll_logger.h"
+#include "port/port.h"
+#include "util/sync_point.h"
 #include "util/testharness.h"
-#include "util/auto_roll_logger.h"
 #include "rocksdb/db.h"
 #include <sys/stat.h>
 #include <errno.h>
@@ -260,7 +263,60 @@ TEST_F(AutoRollLoggerTest, CreateLoggerFromOptions) {
      auto_roll_logger, options.log_file_time_to_roll,
      kSampleMessage + ":CreateLoggerFromOptions - both");
 }
-#endif
+
+TEST_F(AutoRollLoggerTest, LogFlushWhileRolling) {
+  DBOptions options;
+  shared_ptr<Logger> logger;
+
+  InitTestDb();
+  options.max_log_file_size = 1024 * 5;
+  ASSERT_OK(CreateLoggerFromOptions(kTestDir, options, &logger));
+  AutoRollLogger* auto_roll_logger =
+      dynamic_cast<AutoRollLogger*>(logger.get());
+  ASSERT_TRUE(auto_roll_logger);
+  std::thread flush_thread;
+
+  rocksdb::SyncPoint::GetInstance()->LoadDependency({
+      // Need to pin the old logger before beginning the roll, as rolling grabs
+      // the mutex, which would prevent us from accessing the old logger.
+      {"AutoRollLogger::Flush:PinnedLogger",
+       "AutoRollLoggerTest::LogFlushWhileRolling:PreRollAndPostThreadInit"},
+      // Need to finish the flush thread init before this callback because the
+      // callback accesses flush_thread.get_id() in order to apply certain sync
+      // points only to the flush thread.
+      {"AutoRollLoggerTest::LogFlushWhileRolling:PreRollAndPostThreadInit",
+       "AutoRollLoggerTest::LogFlushWhileRolling:FlushCallbackBegin"},
+      // Need to reset logger at this point in Flush() to exercise a race
+      // condition case, which is executing the flush with the pinned (old)
+      // logger after the roll has cut over to a new logger.
+      {"AutoRollLoggerTest::LogFlushWhileRolling:FlushCallback1",
+       "AutoRollLogger::ResetLogger:BeforeNewLogger"},
+      {"AutoRollLogger::ResetLogger:AfterNewLogger",
+       "AutoRollLoggerTest::LogFlushWhileRolling:FlushCallback2"},
+  });
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "PosixLogger::Flush:BeginCallback", [&](void* arg) {
+        TEST_SYNC_POINT(
+            "AutoRollLoggerTest::LogFlushWhileRolling:FlushCallbackBegin");
+        if (std::this_thread::get_id() == flush_thread.get_id()) {
+          TEST_SYNC_POINT(
+              "AutoRollLoggerTest::LogFlushWhileRolling:FlushCallback1");
+          TEST_SYNC_POINT(
+              "AutoRollLoggerTest::LogFlushWhileRolling:FlushCallback2");
+        }
+      });
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  flush_thread = std::thread([&]() { auto_roll_logger->Flush(); });
+  TEST_SYNC_POINT(
+      "AutoRollLoggerTest::LogFlushWhileRolling:PreRollAndPostThreadInit");
+  RollLogFileBySizeTest(auto_roll_logger, options.max_log_file_size,
+                        kSampleMessage + ":LogFlushWhileRolling");
+  flush_thread.join();
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+}
+
+#endif  // OS_WIN

 TEST_F(AutoRollLoggerTest, InfoLogLevel) {
  InitTestDb();
@@ -402,7 +458,16 @@ TEST_F(AutoRollLoggerTest, LogHeaderTest) {
 TEST_F(AutoRollLoggerTest, LogFileExistence) {
  rocksdb::DB* db;
  rocksdb::Options options;
+#ifdef OS_WIN
+  // Replace all slashes in the path so windows CompSpec does not
+  // become confused
+  std::string testDir(kTestDir);
+  std::replace_if(testDir.begin(), testDir.end(),
+    [](char ch) { return ch == '/'; }, '\\');
+  std::string deleteCmd = "if exist " + testDir + " rd /s /q " + testDir;
+#else
  string deleteCmd = "rm -rf " + kTestDir;
+#endif
  ASSERT_EQ(system(deleteCmd.c_str()), 0);
  options.max_log_file_size = 100 * 1024 * 1024;
  options.create_if_missing = true;
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -15,6 +15,7 @@

 #include "db/compaction_iterator.h"
 #include "db/dbformat.h"
+#include "db/event_helpers.h"
 #include "db/filename.h"
 #include "db/internal_stats.h"
 #include "db/merge_helper.h"
@@ -41,13 +42,18 @@ TableBuilder* NewTableBuilder(
    const InternalKeyComparator& internal_comparator,
    const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
        int_tbl_prop_collector_factories,
-    uint32_t column_family_id, WritableFileWriter* file,
-    const CompressionType compression_type,
-    const CompressionOptions& compression_opts, const bool skip_filters) {
+    uint32_t column_family_id, const std::string& column_family_name,
+    WritableFileWriter* file, const CompressionType compression_type,
+    const CompressionOptions& compression_opts,
+    const std::string* compression_dict, const bool skip_filters) {
+  assert((column_family_id ==
+          TablePropertiesCollectorFactory::Context::kUnknownColumnFamily) ==
+         column_family_name.empty());
  return ioptions.table_factory->NewTableBuilder(
      TableBuilderOptions(ioptions, internal_comparator,
                          int_tbl_prop_collector_factories, compression_type,
-                          compression_opts, skip_filters),
+                          compression_opts, compression_dict, skip_filters,
+                          column_family_name),
      column_family_id, file);
 }

@@ -58,12 +64,17 @@ Status BuildTable(
    const InternalKeyComparator& internal_comparator,
    const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
        int_tbl_prop_collector_factories,
-    uint32_t column_family_id, std::vector<SequenceNumber> snapshots,
+    uint32_t column_family_id, const std::string& column_family_name,
+    std::vector<SequenceNumber> snapshots,
    SequenceNumber earliest_write_conflict_snapshot,
    const CompressionType compression,
    const CompressionOptions& compression_opts, bool paranoid_file_checks,
-    InternalStats* internal_stats, const Env::IOPriority io_priority,
-    TableProperties* table_properties) {
+    InternalStats* internal_stats, TableFileCreationReason reason,
+    EventLogger* event_logger, int job_id, const Env::IOPriority io_priority,
+    TableProperties* table_properties, int level) {
+  assert((column_family_id ==
+          TablePropertiesCollectorFactory::Context::kUnknownColumnFamily) ==
+         column_family_name.empty());
  // Reports the IOStats for flush for every following bytes.
  const size_t kReportFlushIOStatsEvery = 1048576;
  Status s;
@@ -72,6 +83,12 @@ Status BuildTable(

  std::string fname = TableFileName(ioptions.db_paths, meta->fd.GetNumber(),
                                    meta->fd.GetPathId());
+#ifndef ROCKSDB_LITE
+  EventHelpers::NotifyTableFileCreationStarted(
+      ioptions.listeners, dbname, column_family_name, fname, job_id, reason);
+#endif  // !ROCKSDB_LITE
+  TableProperties tp;
+
  if (iter->Valid()) {
    TableBuilder* builder;
    unique_ptr<WritableFileWriter> file_writer;
@@ -79,6 +96,9 @@ Status BuildTable(
      unique_ptr<WritableFile> file;
      s = NewWritableFile(env, fname, &file, env_options);
      if (!s.ok()) {
+        EventHelpers::LogAndNotifyTableFileCreationFinished(
+            event_logger, ioptions.listeners, dbname, column_family_name, fname,
+            job_id, meta->fd, tp, reason, s);
        return s;
      }
      file->SetIOPriority(io_priority);
@@ -87,7 +107,8 @@ Status BuildTable(

      builder = NewTableBuilder(
          ioptions, internal_comparator, int_tbl_prop_collector_factories,
-          column_family_id, file_writer.get(), compression, compression_opts);
+          column_family_id, column_family_name, file_writer.get(), compression,
+          compression_opts);
    }

    MergeHelper merge(env, internal_comparator.user_comparator(),
@@ -125,11 +146,13 @@ Status BuildTable(
    }

    if (s.ok() && !empty) {
-      meta->fd.file_size = builder->FileSize();
+      uint64_t file_size = builder->FileSize();
+      meta->fd.file_size = file_size;
      meta->marked_for_compaction = builder->NeedCompact();
      assert(meta->fd.GetFileSize() > 0);
+      tp = builder->GetTableProperties();
      if (table_properties) {
-        *table_properties = builder->GetTableProperties();
+        *table_properties = tp;
      }
    }
    delete builder;
@@ -149,7 +172,8 @@ Status BuildTable(
          ReadOptions(), env_options, internal_comparator, meta->fd, nullptr,
          (internal_stats == nullptr) ? nullptr
                                      : internal_stats->GetFileReadHist(0),
-          false));
+          false /* for_compaction */, nullptr /* arena */,
+          false /* skip_filter */, level));
      s = it->status();
      if (s.ok() && paranoid_file_checks) {
        for (it->SeekToFirst(); it->Valid(); it->Next()) {
@@ -167,6 +191,12 @@ Status BuildTable(
  if (!s.ok() || meta->fd.GetFileSize() == 0) {
    env->DeleteFile(fname);
  }
+
+  // Output to event logger and fire events.
+  EventHelpers::LogAndNotifyTableFileCreationFinished(
+      event_logger, ioptions.listeners, dbname, column_family_name, fname,
+      job_id, meta->fd, tp, reason, s);
+
  return s;
 }

@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -12,11 +12,13 @@
 #include "db/table_properties_collector.h"
 #include "rocksdb/comparator.h"
 #include "rocksdb/env.h"
-#include "rocksdb/status.h"
-#include "rocksdb/types.h"
-#include "rocksdb/options.h"
 #include "rocksdb/immutable_options.h"
+#include "rocksdb/listener.h"
+#include "rocksdb/options.h"
+#include "rocksdb/status.h"
 #include "rocksdb/table_properties.h"
+#include "rocksdb/types.h"
+#include "util/event_logger.h"

 namespace rocksdb {

@@ -33,14 +35,20 @@ class WritableFileWriter;
 class InternalStats;
 class InternalIterator;

+// @param column_family_name Name of the column family that is also identified
+//    by column_family_id, or empty string if unknown. It must outlive the
+//    TableBuilder returned by this function.
+// @param compression_dict Data for presetting the compression library's
+//    dictionary, or nullptr.
 TableBuilder* NewTableBuilder(
    const ImmutableCFOptions& options,
    const InternalKeyComparator& internal_comparator,
    const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
        int_tbl_prop_collector_factories,
-    uint32_t column_family_id, WritableFileWriter* file,
-    const CompressionType compression_type,
+    uint32_t column_family_id, const std::string& column_family_name,
+    WritableFileWriter* file, const CompressionType compression_type,
    const CompressionOptions& compression_opts,
+    const std::string* compression_dict = nullptr,
    const bool skip_filters = false);

 // Build a Table file from the contents of *iter.  The generated file
@@ -48,6 +56,9 @@ TableBuilder* NewTableBuilder(
 // *meta will be filled with metadata about the generated table.
 // If no data is present in *iter, meta->file_size will be set to
 // zero, and no Table file will be produced.
+//
+// @param column_family_name Name of the column family that is also identified
+//    by column_family_id, or empty string if unknown.
 extern Status BuildTable(
    const std::string& dbname, Env* env, const ImmutableCFOptions& options,
    const EnvOptions& env_options, TableCache* table_cache,
@@ -55,12 +66,14 @@ extern Status BuildTable(
    const InternalKeyComparator& internal_comparator,
    const std::vector<std::unique_ptr<IntTblPropCollectorFactory>>*
        int_tbl_prop_collector_factories,
-    uint32_t column_family_id, std::vector<SequenceNumber> snapshots,
+    uint32_t column_family_id, const std::string& column_family_name,
+    std::vector<SequenceNumber> snapshots,
    SequenceNumber earliest_write_conflict_snapshot,
    const CompressionType compression,
    const CompressionOptions& compression_opts, bool paranoid_file_checks,
-    InternalStats* internal_stats,
+    InternalStats* internal_stats, TableFileCreationReason reason,
+    EventLogger* event_logger = nullptr, int job_id = 0,
    const Env::IOPriority io_priority = Env::IO_HIGH,
-    TableProperties* table_properties = nullptr);
+    TableProperties* table_properties = nullptr, int level = -1);

 }  // namespace rocksdb
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -446,6 +446,12 @@ void rocksdb_backup_engine_create_new_backup(rocksdb_backup_engine_t* be,
  SaveError(errptr, be->rep->CreateNewBackup(db->rep));
 }

+void rocksdb_backup_engine_purge_old_backups(rocksdb_backup_engine_t* be,
+                                             uint32_t num_backups_to_keep,
+                                             char** errptr) {
+  SaveError(errptr, be->rep->PurgeOldBackups(num_backups_to_keep));
+}
+
 rocksdb_restore_options_t* rocksdb_restore_options_create() {
  return new rocksdb_restore_options_t;
 }
@@ -825,6 +831,31 @@ rocksdb_iterator_t* rocksdb_create_iterator_cf(
  return result;
 }

+void rocksdb_create_iterators(
+    rocksdb_t *db,
+    rocksdb_readoptions_t* opts,
+    rocksdb_column_family_handle_t** column_families,
+    rocksdb_iterator_t** iterators,
+    size_t size,
+    char** errptr) {
+  std::vector<ColumnFamilyHandle*> column_families_vec;
+  for (size_t i = 0; i < size; i++) {
+    column_families_vec.push_back(column_families[i]->rep);
+  }
+
+  std::vector<Iterator*> res;
+  Status status = db->rep->NewIterators(opts->rep, column_families_vec, &res);
+  assert(res.size() == size);
+  if (SaveError(errptr, status)) {
+    return;
+  }
+
+  for (size_t i = 0; i < size; i++) {
+    iterators[i] = new rocksdb_iterator_t;
+    iterators[i]->rep = res[i];
+  }
+}
+
 const rocksdb_snapshot_t* rocksdb_create_snapshot(
    rocksdb_t* db) {
  rocksdb_snapshot_t* result = new rocksdb_snapshot_t;
@@ -1288,6 +1319,11 @@ void rocksdb_block_based_options_set_cache_index_and_filter_blocks(
  options->rep.cache_index_and_filter_blocks = v;
 }

+void rocksdb_block_based_options_set_pin_l0_filter_and_index_blocks_in_cache(
+    rocksdb_block_based_table_options_t* options, unsigned char v) {
+  options->rep.pin_l0_filter_and_index_blocks_in_cache = v;
+}
+
 void rocksdb_block_based_options_set_skip_table_builder_flush(
    rocksdb_block_based_table_options_t* options, unsigned char v) {
  options->rep.skip_table_builder_flush = v;
@@ -1388,6 +1424,11 @@ void rocksdb_options_set_compaction_filter_factory(
      std::shared_ptr<CompactionFilterFactory>(factory);
 }

+void rocksdb_options_compaction_readahead_size(
+    rocksdb_options_t* opt, size_t s) {
+  opt->rep.compaction_readahead_size = s;
+}
+
 void rocksdb_options_set_comparator(
    rocksdb_options_t* opt,
    rocksdb_comparator_t* cmp) {
@@ -1531,11 +1572,13 @@ void rocksdb_options_set_compression_per_level(rocksdb_options_t* opt,
  }
 }

-void rocksdb_options_set_compression_options(
-    rocksdb_options_t* opt, int w_bits, int level, int strategy) {
+void rocksdb_options_set_compression_options(rocksdb_options_t* opt, int w_bits,
+                                             int level, int strategy,
+                                             int max_dict_bytes) {
  opt->rep.compression_opts.window_bits = w_bits;
  opt->rep.compression_opts.level = level;
  opt->rep.compression_opts.strategy = strategy;
+  opt->rep.compression_opts.max_dict_bytes = max_dict_bytes;
 }

 void rocksdb_options_set_prefix_extractor(
@@ -2127,6 +2170,13 @@ rocksdb_env_t* rocksdb_create_default_env() {
  return result;
 }

+rocksdb_env_t* rocksdb_create_mem_env() {
+  rocksdb_env_t* result = new rocksdb_env_t;
+  result->rep = rocksdb::NewMemEnv(Env::Default());
+  result->is_default = false;
+  return result;
+}
+
 void rocksdb_env_set_background_threads(rocksdb_env_t* env, int n) {
  env->rep->SetBackgroundThreads(n);
 }
@@ -322,7 +322,7 @@ int main(int argc, char** argv) {
  rocksdb_options_set_block_based_table_factory(options, table_options);

  rocksdb_options_set_compression(options, rocksdb_no_compression);
-  rocksdb_options_set_compression_options(options, -14, -1, 0);
+  rocksdb_options_set_compression_options(options, -14, -1, 0, 0);
  int compression_levels[] = {rocksdb_no_compression, rocksdb_no_compression,
                              rocksdb_no_compression, rocksdb_no_compression};
  rocksdb_options_set_compression_per_level(options, compression_levels, 4);
@@ -365,6 +365,24 @@ int main(int argc, char** argv) {
    rocksdb_backup_engine_create_new_backup(be, db, &err);
    CheckNoError(err);

+    // need a change to trigger a new backup
+    rocksdb_delete(db, woptions, "does-not-exist", 14, &err);
+    CheckNoError(err);
+
+    rocksdb_backup_engine_create_new_backup(be, db, &err);
+    CheckNoError(err);
+
+    const rocksdb_backup_engine_info_t* bei = rocksdb_backup_engine_get_backup_info(be);
+    CheckCondition(rocksdb_backup_engine_info_count(bei) > 1);
+    rocksdb_backup_engine_info_destroy(bei);
+
+    rocksdb_backup_engine_purge_old_backups(be, 1, &err);
+    CheckNoError(err);
+
+    bei = rocksdb_backup_engine_get_backup_info(be);
+    CheckCondition(rocksdb_backup_engine_info_count(bei) == 1);
+    rocksdb_backup_engine_info_destroy(bei);
+
    rocksdb_delete(db, woptions, "foo", 3, &err);
    CheckNoError(err);

@@ -768,6 +786,30 @@ int main(int argc, char** argv) {
    CheckNoError(err);
    rocksdb_iter_destroy(iter);

+    rocksdb_column_family_handle_t* iters_cf_handles[2] = { handles[0], handles[1] };
+    rocksdb_iterator_t* iters_handles[2];
+    rocksdb_create_iterators(db, roptions, iters_cf_handles, iters_handles, 2, &err);
+    CheckNoError(err);
+
+    iter = iters_handles[0];
+    CheckCondition(!rocksdb_iter_valid(iter));
+    rocksdb_iter_seek_to_first(iter);
+    CheckCondition(!rocksdb_iter_valid(iter));
+    rocksdb_iter_destroy(iter);
+
+    iter = iters_handles[1];
+    CheckCondition(!rocksdb_iter_valid(iter));
+    rocksdb_iter_seek_to_first(iter);
+    CheckCondition(rocksdb_iter_valid(iter));
+
+    for (i = 0; rocksdb_iter_valid(iter) != 0; rocksdb_iter_next(iter)) {
+      i++;
+    }
+    CheckCondition(i == 1);
+    rocksdb_iter_get_error(iter, &err);
+    CheckNoError(err);
+    rocksdb_iter_destroy(iter);
+
    rocksdb_drop_column_family(db, handles[1], &err);
    CheckNoError(err);
    for (i = 0; i < 2; i++) {
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -135,6 +135,10 @@ Status CheckConcurrentWritesSupported(const ColumnFamilyOptions& cf_options) {
        "Delete filtering (filter_deletes) is not compatible with concurrent "
        "memtable writes (allow_concurrent_memtable_writes)");
  }
+  if (!cf_options.memtable_factory->IsInsertConcurrentlySupported()) {
+    return Status::InvalidArgument(
+        "Memtable doesn't concurrent writes (allow_concurrent_memtable_write)");
+  }
  return Status::OK();
 }

@@ -143,13 +147,10 @@ ColumnFamilyOptions SanitizeOptions(const DBOptions& db_options,
                                    const ColumnFamilyOptions& src) {
  ColumnFamilyOptions result = src;
  result.comparator = icmp;
-#ifdef OS_MACOSX
-  // TODO(icanadi) make write_buffer_size uint64_t instead of size_t
-  ClipToRange(&result.write_buffer_size, ((size_t)64) << 10, ((size_t)1) << 30);
-#else
-  ClipToRange(&result.write_buffer_size,
-              ((size_t)64) << 10, ((size_t)64) << 30);
-#endif
+  size_t clamp_max = std::conditional<
+      sizeof(size_t) == 4, std::integral_constant<size_t, 0xffffffff>,
+      std::integral_constant<uint64_t, 64ull << 30>>::type::value;
+  ClipToRange(&result.write_buffer_size, ((size_t)64) << 10, clamp_max);
  // if user sets arena_block_size, we trust user to use this value. Otherwise,
  // calculate a proper value from writer_buffer_size;
  if (result.arena_block_size <= 0) {
@@ -239,6 +240,17 @@ ColumnFamilyOptions SanitizeOptions(const DBOptions& db_options,
         result.level0_slowdown_writes_trigger,
         result.level0_file_num_compaction_trigger);
  }
+
+  if (result.soft_pending_compaction_bytes_limit == 0) {
+    result.soft_pending_compaction_bytes_limit =
+        result.hard_pending_compaction_bytes_limit;
+  } else if (result.hard_pending_compaction_bytes_limit > 0 &&
+             result.soft_pending_compaction_bytes_limit >
+                 result.hard_pending_compaction_bytes_limit) {
+    result.soft_pending_compaction_bytes_limit =
+        result.hard_pending_compaction_bytes_limit;
+  }
+
  if (result.level_compaction_dynamic_level_bytes) {
    if (result.compaction_style != kCompactionStyleLevel ||
        db_options.db_paths.size() > 1U) {
@@ -513,6 +525,21 @@ std::unique_ptr<WriteControllerToken> SetupDelay(
  }
  return write_controller->GetDelayToken(write_rate);
 }
+
+int GetL0ThresholdSpeedupCompaction(int level0_file_num_compaction_trigger,
+                                    int level0_slowdown_writes_trigger) {
+  // SanitizeOptions() ensures it.
+  assert(level0_file_num_compaction_trigger <= level0_slowdown_writes_trigger);
+
+  // 1/4 of the way between L0 compaction trigger threshold and slowdown
+  // condition.
+  // Or twice as compaction trigger, if it is smaller.
+  return std::min(level0_file_num_compaction_trigger * 2,
+                  level0_file_num_compaction_trigger +
+                      (level0_slowdown_writes_trigger -
+                       level0_file_num_compaction_trigger) /
+                          4);
+}
 }  // namespace

 void ColumnFamilyData::RecalculateWriteStallConditions(
@@ -531,21 +558,6 @@ void ColumnFamilyData::RecalculateWriteStallConditions(
          "(waiting for flush), max_write_buffer_number is set to %d",
          name_.c_str(), imm()->NumNotFlushed(),
          mutable_cf_options.max_write_buffer_number);
-    } else if (mutable_cf_options.max_write_buffer_number > 3 &&
-               imm()->NumNotFlushed() >=
-                   mutable_cf_options.max_write_buffer_number - 1) {
-      write_controller_token_ =
-          SetupDelay(ioptions_.delayed_write_rate, write_controller,
-                     compaction_needed_bytes, prev_compaction_needed_bytes_,
-                     mutable_cf_options.disable_auto_compactions);
-      internal_stats_->AddCFStats(InternalStats::MEMTABLE_SLOWDOWN, 1);
-      Log(InfoLogLevel::WARN_LEVEL, ioptions_.info_log,
-          "[%s] Stalling writes because we have %d immutable memtables "
-          "(waiting for flush), max_write_buffer_number is set to %d "
-          "rate %" PRIu64,
-          name_.c_str(), imm()->NumNotFlushed(),
-          mutable_cf_options.max_write_buffer_number,
-          write_controller->delayed_write_rate());
    } else if (vstorage->l0_delay_trigger_count() >=
               mutable_cf_options.level0_stop_writes_trigger) {
      write_controller_token_ = write_controller->GetStopToken();
@@ -567,6 +579,21 @@ void ColumnFamilyData::RecalculateWriteStallConditions(
          "[%s] Stopping writes because of estimated pending compaction "
          "bytes %" PRIu64,
          name_.c_str(), compaction_needed_bytes);
+    } else if (mutable_cf_options.max_write_buffer_number > 3 &&
+               imm()->NumNotFlushed() >=
+                   mutable_cf_options.max_write_buffer_number - 1) {
+      write_controller_token_ =
+          SetupDelay(ioptions_.delayed_write_rate, write_controller,
+                     compaction_needed_bytes, prev_compaction_needed_bytes_,
+                     mutable_cf_options.disable_auto_compactions);
+      internal_stats_->AddCFStats(InternalStats::MEMTABLE_SLOWDOWN, 1);
+      Log(InfoLogLevel::WARN_LEVEL, ioptions_.info_log,
+          "[%s] Stalling writes because we have %d immutable memtables "
+          "(waiting for flush), max_write_buffer_number is set to %d "
+          "rate %" PRIu64,
+          name_.c_str(), imm()->NumNotFlushed(),
+          mutable_cf_options.max_write_buffer_number,
+          write_controller->delayed_write_rate());
    } else if (mutable_cf_options.level0_slowdown_writes_trigger >= 0 &&
               vstorage->l0_delay_trigger_count() >=
                   mutable_cf_options.level0_slowdown_writes_trigger) {
@@ -598,6 +625,29 @@ void ColumnFamilyData::RecalculateWriteStallConditions(
          "bytes %" PRIu64 " rate %" PRIu64,
          name_.c_str(), vstorage->estimated_compaction_needed_bytes(),
          write_controller->delayed_write_rate());
+    } else if (vstorage->l0_delay_trigger_count() >=
+               GetL0ThresholdSpeedupCompaction(
+                   mutable_cf_options.level0_file_num_compaction_trigger,
+                   mutable_cf_options.level0_slowdown_writes_trigger)) {
+      write_controller_token_ = write_controller->GetCompactionPressureToken();
+      Log(InfoLogLevel::WARN_LEVEL, ioptions_.info_log,
+          "[%s] Increasing compaction threads because we have %d level-0 "
+          "files ",
+          name_.c_str(), vstorage->l0_delay_trigger_count());
+    } else if (vstorage->estimated_compaction_needed_bytes() >=
+               mutable_cf_options.soft_pending_compaction_bytes_limit / 4) {
+      // Increase compaction threads if bytes needed for compaction exceeds
+      // 1/4 of threshold for slowing down.
+      // If soft pending compaction byte limit is not set, always speed up
+      // compaction.
+      write_controller_token_ = write_controller->GetCompactionPressureToken();
+      if (mutable_cf_options.soft_pending_compaction_bytes_limit > 0) {
+        Log(InfoLogLevel::WARN_LEVEL, ioptions_.info_log,
+            "[%s] Increasing compaction threads because of estimated pending "
+            "compaction "
+            "bytes %" PRIu64,
+            name_.c_str(), vstorage->estimated_compaction_needed_bytes());
+      }
    } else {
      write_controller_token_.reset();
    }
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -209,7 +209,7 @@ class ColumnFamilyData {
  const ImmutableCFOptions* ioptions() const { return &ioptions_; }
  // REQUIRES: DB mutex held
  // This returns the MutableCFOptions used by current SuperVersion
-  // You shoul use this API to reference MutableCFOptions most of the time.
+  // You should use this API to reference MutableCFOptions most of the time.
  const MutableCFOptions* GetCurrentMutableCFOptions() const {
    return &(super_version_->mutable_cf_options);
  }
@@ -465,6 +465,8 @@ class ColumnFamilySet {
  // Don't call while iterating over ColumnFamilySet
  void FreeDeadColumnFamilies();

+  Cache* get_table_cache() { return table_cache_; }
+
 private:
  friend class ColumnFamilyData;
  // helper function that gets called from cfd destructor
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -65,6 +65,7 @@ class ColumnFamilyTest : public testing::Test {

  ~ColumnFamilyTest() {
    Close();
+    rocksdb::SyncPoint::GetInstance()->DisableProcessing();
    Destroy();
    delete env_;
  }
@@ -1041,6 +1042,7 @@ TEST_F(ColumnFamilyTest, AutomaticAndManualCompactions) {
  db_options_.max_open_files = 20;  // only 10 files in file cache
  db_options_.disableDataSync = true;
  db_options_.max_background_compactions = 3;
+  db_options_.base_background_compactions = 3;

  default_cf.compaction_style = kCompactionStyleLevel;
  default_cf.num_levels = 3;
@@ -1133,6 +1135,7 @@ TEST_F(ColumnFamilyTest, ManualAndAutomaticCompactions) {
  db_options_.max_open_files = 20;  // only 10 files in file cache
  db_options_.disableDataSync = true;
  db_options_.max_background_compactions = 3;
+  db_options_.base_background_compactions = 3;

  default_cf.compaction_style = kCompactionStyleLevel;
  default_cf.num_levels = 3;
@@ -1228,6 +1231,7 @@ TEST_F(ColumnFamilyTest, SameCFManualManualCompactions) {
  db_options_.max_open_files = 20;  // only 10 files in file cache
  db_options_.disableDataSync = true;
  db_options_.max_background_compactions = 3;
+  db_options_.base_background_compactions = 3;

  default_cf.compaction_style = kCompactionStyleLevel;
  default_cf.num_levels = 3;
@@ -1326,6 +1330,7 @@ TEST_F(ColumnFamilyTest, SameCFManualAutomaticCompactions) {
  db_options_.max_open_files = 20;  // only 10 files in file cache
  db_options_.disableDataSync = true;
  db_options_.max_background_compactions = 3;
+  db_options_.base_background_compactions = 3;

  default_cf.compaction_style = kCompactionStyleLevel;
  default_cf.num_levels = 3;
@@ -1415,6 +1420,7 @@ TEST_F(ColumnFamilyTest, SameCFManualAutomaticCompactionsLevel) {
  db_options_.max_open_files = 20;  // only 10 files in file cache
  db_options_.disableDataSync = true;
  db_options_.max_background_compactions = 3;
+  db_options_.base_background_compactions = 3;

  default_cf.compaction_style = kCompactionStyleLevel;
  default_cf.num_levels = 3;
@@ -1511,6 +1517,7 @@ TEST_F(ColumnFamilyTest, SameCFManualAutomaticConflict) {
  db_options_.max_open_files = 20;  // only 10 files in file cache
  db_options_.disableDataSync = true;
  db_options_.max_background_compactions = 3;
+  db_options_.base_background_compactions = 3;

  default_cf.compaction_style = kCompactionStyleLevel;
  default_cf.num_levels = 3;
@@ -1630,6 +1637,7 @@ TEST_F(ColumnFamilyTest, SameCFAutomaticManualCompactions) {
  db_options_.max_open_files = 20;  // only 10 files in file cache
  db_options_.disableDataSync = true;
  db_options_.max_background_compactions = 3;
+  db_options_.base_background_compactions = 3;

  default_cf.compaction_style = kCompactionStyleLevel;
  default_cf.num_levels = 3;
@@ -1951,13 +1959,27 @@ TEST_F(ColumnFamilyTest, ReadDroppedColumnFamily) {
    PutRandomData(1, kKeysNum, 100);
    PutRandomData(2, kKeysNum, 100);

-    if (iter == 0) {
-      // Drop CF two
-      ASSERT_OK(db_->DropColumnFamily(handles_[2]));
-    } else {
-      // delete CF two
-      delete handles_[2];
-      handles_[2] = nullptr;
+    {
+      std::unique_ptr<Iterator> iterator(
+          db_->NewIterator(ReadOptions(), handles_[2]));
+      iterator->SeekToFirst();
+
+      if (iter == 0) {
+        // Drop CF two
+        ASSERT_OK(db_->DropColumnFamily(handles_[2]));
+      } else {
+        // delete CF two
+        delete handles_[2];
+        handles_[2] = nullptr;
+      }
+      // Make sure iterator created can still be used.
+      int count = 0;
+      for (; iterator->Valid(); iterator->Next()) {
+        ASSERT_OK(iterator->status());
+        ++count;
+      }
+      ASSERT_OK(iterator->status());
+      ASSERT_EQ(count, kKeysNum);
    }

    // Add bunch more data to other CFs
@@ -1999,7 +2021,9 @@ TEST_F(ColumnFamilyTest, FlushAndDropRaceCondition) {
  Reopen({options, options});

  rocksdb::SyncPoint::GetInstance()->LoadDependency(
-      {{"VersionSet::LogAndApply::ColumnFamilyDrop:1",
+      {{"VersionSet::LogAndApply::ColumnFamilyDrop:0",
+        "FlushJob::WriteLevel0Table"},
+       {"VersionSet::LogAndApply::ColumnFamilyDrop:1",
        "FlushJob::InstallResults"},
       {"FlushJob::InstallResults",
        "VersionSet::LogAndApply::ColumnFamilyDrop:2"}});
@@ -2045,7 +2069,6 @@ TEST_F(ColumnFamilyTest, FlushAndDropRaceCondition) {

  Close();
  Destroy();
-  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
 }

 #ifndef ROCKSDB_LITE
@@ -2123,7 +2146,6 @@ TEST_F(ColumnFamilyTest, CreateAndDropRace) {
  drop_cf_thread.join();
  Close();
  Destroy();
-  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
  for (auto* comparator : comparators) {
    if (comparator) {
      delete comparator;
@@ -2135,6 +2157,9 @@ TEST_F(ColumnFamilyTest, CreateAndDropRace) {
 TEST_F(ColumnFamilyTest, WriteStallSingleColumnFamily) {
  const uint64_t kBaseRate = 810000u;
  db_options_.delayed_write_rate = kBaseRate;
+  db_options_.base_background_compactions = 2;
+  db_options_.max_background_compactions = 6;
+
  Open({"default"});
  ColumnFamilyData* cfd =
      static_cast<ColumnFamilyHandleImpl*>(db_->DefaultColumnFamily())->cfd();
@@ -2160,6 +2185,7 @@ TEST_F(ColumnFamilyTest, WriteStallSingleColumnFamily) {
  ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
  ASSERT_EQ(kBaseRate, dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());

  vstorage->TEST_set_estimated_compaction_needed_bytes(400);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
@@ -2167,6 +2193,7 @@ TEST_F(ColumnFamilyTest, WriteStallSingleColumnFamily) {
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
  ASSERT_EQ(kBaseRate / 1.2,
            dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());

  vstorage->TEST_set_estimated_compaction_needed_bytes(500);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
@@ -2222,6 +2249,7 @@ TEST_F(ColumnFamilyTest, WriteStallSingleColumnFamily) {
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
  ASSERT_TRUE(dbfull()->TEST_write_controler().IsStopped());
  ASSERT_TRUE(!dbfull()->TEST_write_controler().NeedsDelay());
+  ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());

  vstorage->TEST_set_estimated_compaction_needed_bytes(3001);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
@@ -2246,6 +2274,7 @@ TEST_F(ColumnFamilyTest, WriteStallSingleColumnFamily) {
  ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
  ASSERT_EQ(kBaseRate / 1.2,
            dbfull()->TEST_write_controler().delayed_write_rate());
+  ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());

  vstorage->set_l0_delay_trigger_count(101);
  cfd->RecalculateWriteStallConditions(mutable_cf_options);
@@ -2318,6 +2347,73 @@ TEST_F(ColumnFamilyTest, WriteStallSingleColumnFamily) {
            dbfull()->TEST_write_controler().delayed_write_rate());
 }

+TEST_F(ColumnFamilyTest, CompactionSpeedupSingleColumnFamily) {
+  db_options_.base_background_compactions = 2;
+  db_options_.max_background_compactions = 6;
+  Open({"default"});
+  ColumnFamilyData* cfd =
+      static_cast<ColumnFamilyHandleImpl*>(db_->DefaultColumnFamily())->cfd();
+
+  VersionStorageInfo* vstorage = cfd->current()->storage_info();
+
+  MutableCFOptions mutable_cf_options(
+      Options(db_options_, column_family_options_),
+      ImmutableCFOptions(Options(db_options_, column_family_options_)));
+
+  // Speed up threshold = min(4 * 2, 4 + (36 - 4)/4) = 8
+  mutable_cf_options.level0_file_num_compaction_trigger = 4;
+  mutable_cf_options.level0_slowdown_writes_trigger = 36;
+  mutable_cf_options.level0_stop_writes_trigger = 50;
+  // Speedup threshold = 200 / 4 = 50
+  mutable_cf_options.soft_pending_compaction_bytes_limit = 200;
+  mutable_cf_options.hard_pending_compaction_bytes_limit = 2000;
+
+  vstorage->TEST_set_estimated_compaction_needed_bytes(40);
+  cfd->RecalculateWriteStallConditions(mutable_cf_options);
+  ASSERT_EQ(2, dbfull()->BGCompactionsAllowed());
+
+  vstorage->TEST_set_estimated_compaction_needed_bytes(50);
+  cfd->RecalculateWriteStallConditions(mutable_cf_options);
+  ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());
+
+  vstorage->TEST_set_estimated_compaction_needed_bytes(300);
+  cfd->RecalculateWriteStallConditions(mutable_cf_options);
+  ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());
+
+  vstorage->TEST_set_estimated_compaction_needed_bytes(45);
+  cfd->RecalculateWriteStallConditions(mutable_cf_options);
+  ASSERT_EQ(2, dbfull()->BGCompactionsAllowed());
+
+  vstorage->set_l0_delay_trigger_count(7);
+  cfd->RecalculateWriteStallConditions(mutable_cf_options);
+  ASSERT_EQ(2, dbfull()->BGCompactionsAllowed());
+
+  vstorage->set_l0_delay_trigger_count(9);
+  cfd->RecalculateWriteStallConditions(mutable_cf_options);
+  ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());
+
+  vstorage->set_l0_delay_trigger_count(6);
+  cfd->RecalculateWriteStallConditions(mutable_cf_options);
+  ASSERT_EQ(2, dbfull()->BGCompactionsAllowed());
+
+  // Speed up threshold = min(4 * 2, 4 + (12 - 4)/4) = 6
+  mutable_cf_options.level0_file_num_compaction_trigger = 4;
+  mutable_cf_options.level0_slowdown_writes_trigger = 16;
+  mutable_cf_options.level0_stop_writes_trigger = 30;
+
+  vstorage->set_l0_delay_trigger_count(5);
+  cfd->RecalculateWriteStallConditions(mutable_cf_options);
+  ASSERT_EQ(2, dbfull()->BGCompactionsAllowed());
+
+  vstorage->set_l0_delay_trigger_count(7);
+  cfd->RecalculateWriteStallConditions(mutable_cf_options);
+  ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());
+
+  vstorage->set_l0_delay_trigger_count(3);
+  cfd->RecalculateWriteStallConditions(mutable_cf_options);
+  ASSERT_EQ(2, dbfull()->BGCompactionsAllowed());
+}
+
 TEST_F(ColumnFamilyTest, WriteStallTwoColumnFamilies) {
  const uint64_t kBaseRate = 810000u;
  db_options_.delayed_write_rate = kBaseRate;
@@ -2399,6 +2495,108 @@ TEST_F(ColumnFamilyTest, WriteStallTwoColumnFamilies) {
  ASSERT_EQ(kBaseRate / 1.2,
            dbfull()->TEST_write_controler().delayed_write_rate());
 }
+
+TEST_F(ColumnFamilyTest, CompactionSpeedupTwoColumnFamilies) {
+  db_options_.base_background_compactions = 2;
+  db_options_.max_background_compactions = 6;
+  column_family_options_.soft_pending_compaction_bytes_limit = 200;
+  column_family_options_.hard_pending_compaction_bytes_limit = 2000;
+  Open();
+  CreateColumnFamilies({"one"});
+  ColumnFamilyData* cfd =
+      static_cast<ColumnFamilyHandleImpl*>(db_->DefaultColumnFamily())->cfd();
+  VersionStorageInfo* vstorage = cfd->current()->storage_info();
+
+  ColumnFamilyData* cfd1 =
+      static_cast<ColumnFamilyHandleImpl*>(handles_[1])->cfd();
+  VersionStorageInfo* vstorage1 = cfd1->current()->storage_info();
+
+  MutableCFOptions mutable_cf_options(
+      Options(db_options_, column_family_options_),
+      ImmutableCFOptions(Options(db_options_, column_family_options_)));
+  // Speed up threshold = min(4 * 2, 4 + (36 - 4)/4) = 8
+  mutable_cf_options.level0_file_num_compaction_trigger = 4;
+  mutable_cf_options.level0_slowdown_writes_trigger = 36;
+  mutable_cf_options.level0_stop_writes_trigger = 30;
+  // Speedup threshold = 200 / 4 = 50
+  mutable_cf_options.soft_pending_compaction_bytes_limit = 200;
+  mutable_cf_options.hard_pending_compaction_bytes_limit = 2000;
+
+  MutableCFOptions mutable_cf_options1 = mutable_cf_options;
+  mutable_cf_options1.level0_slowdown_writes_trigger = 16;
+
+  vstorage->TEST_set_estimated_compaction_needed_bytes(40);
+  cfd->RecalculateWriteStallConditions(mutable_cf_options);
+  ASSERT_EQ(2, dbfull()->BGCompactionsAllowed());
+
+  vstorage->TEST_set_estimated_compaction_needed_bytes(60);
+  cfd1->RecalculateWriteStallConditions(mutable_cf_options);
+  ASSERT_EQ(2, dbfull()->BGCompactionsAllowed());
+  cfd->RecalculateWriteStallConditions(mutable_cf_options);
+  ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());
+
+  vstorage1->TEST_set_estimated_compaction_needed_bytes(30);
+  cfd1->RecalculateWriteStallConditions(mutable_cf_options);
+  ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());
+
+  vstorage1->TEST_set_estimated_compaction_needed_bytes(70);
+  cfd1->RecalculateWriteStallConditions(mutable_cf_options);
+  ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());
+
+  vstorage->TEST_set_estimated_compaction_needed_bytes(20);
+  cfd->RecalculateWriteStallConditions(mutable_cf_options);
+  ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());
+
+  vstorage1->TEST_set_estimated_compaction_needed_bytes(3);
+  cfd1->RecalculateWriteStallConditions(mutable_cf_options);
+  ASSERT_EQ(2, dbfull()->BGCompactionsAllowed());
+
+  vstorage->set_l0_delay_trigger_count(9);
+  cfd->RecalculateWriteStallConditions(mutable_cf_options);
+  ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());
+
+  vstorage1->set_l0_delay_trigger_count(2);
+  cfd1->RecalculateWriteStallConditions(mutable_cf_options);
+  ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());
+
+  vstorage->set_l0_delay_trigger_count(0);
+  cfd->RecalculateWriteStallConditions(mutable_cf_options);
+  ASSERT_EQ(2, dbfull()->BGCompactionsAllowed());
+}
+
+// Disable on windows because SyncWAL requires env->IsSyncThreadSafe()
+// to return true which is not so in unbuffered mode.
+#ifndef OS_WIN
+TEST_F(ColumnFamilyTest, LogSyncConflictFlush) {
+  Open();
+  CreateColumnFamiliesAndReopen({"one", "two"});
+
+  Put(0, "", "");
+  Put(1, "foo", "bar");
+
+  rocksdb::SyncPoint::GetInstance()->LoadDependency(
+      {{"DBImpl::SyncWAL:BeforeMarkLogsSynced:1",
+        "ColumnFamilyTest::LogSyncConflictFlush:1"},
+       {"ColumnFamilyTest::LogSyncConflictFlush:2",
+        "DBImpl::SyncWAL:BeforeMarkLogsSynced:2"}});
+
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  std::thread thread([&] { db_->SyncWAL(); });
+
+  TEST_SYNC_POINT("ColumnFamilyTest::LogSyncConflictFlush:1");
+  Flush(1);
+  Put(1, "foo", "bar");
+  Flush(1);
+
+  TEST_SYNC_POINT("ColumnFamilyTest::LogSyncConflictFlush:2");
+
+  thread.join();
+
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+  Close();
+}
+#endif
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -1,4 +1,4 @@
-//  Copyright (c) 2014, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -7,6 +7,7 @@

 #include <mutex>
 #include <string>
+#include <thread>
 #include <vector>

 #include "rocksdb/db.h"
@@ -107,6 +108,7 @@ TEST_F(CompactFilesTest, L0ConflictsFiles) {
      break;
    }
  }
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
  delete db;
 }

@@ -141,9 +143,6 @@ TEST_F(CompactFilesTest, ObsoleteFiles) {
  }

  auto l0_files = collector->GetFlushedFiles();
-  CompactionOptions compact_opt;
-  compact_opt.compression = kNoCompression;
-  compact_opt.output_file_size_limit = kWriteBufferSize * 5;
  ASSERT_OK(db->CompactFiles(CompactionOptions(), l0_files, 1));

  // verify all compaction input files are deleted
@@ -153,6 +152,62 @@ TEST_F(CompactFilesTest, ObsoleteFiles) {
  delete db;
 }

+TEST_F(CompactFilesTest, CapturingPendingFiles) {
+  Options options;
+  options.create_if_missing = true;
+  // Disable RocksDB background compaction.
+  options.compaction_style = kCompactionStyleNone;
+  // Always do full scans for obsolete files (needed to reproduce the issue).
+  options.delete_obsolete_files_period_micros = 0;
+
+  // Add listener.
+  FlushedFileCollector* collector = new FlushedFileCollector();
+  options.listeners.emplace_back(collector);
+
+  DB* db = nullptr;
+  DestroyDB(db_name_, options);
+  Status s = DB::Open(options, db_name_, &db);
+  assert(s.ok());
+  assert(db);
+
+  // Create 5 files.
+  for (int i = 0; i < 5; ++i) {
+    db->Put(WriteOptions(), "key" + ToString(i), "value");
+    db->Flush(FlushOptions());
+  }
+
+  auto l0_files = collector->GetFlushedFiles();
+  EXPECT_EQ(5, l0_files.size());
+
+  rocksdb::SyncPoint::GetInstance()->LoadDependency({
+      {"CompactFilesImpl:2", "CompactFilesTest.CapturingPendingFiles:0"},
+      {"CompactFilesTest.CapturingPendingFiles:1", "CompactFilesImpl:3"},
+  });
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  // Start compacting files.
+  std::thread compaction_thread(
+      [&] { EXPECT_OK(db->CompactFiles(CompactionOptions(), l0_files, 1)); });
+
+  // In the meantime flush another file.
+  TEST_SYNC_POINT("CompactFilesTest.CapturingPendingFiles:0");
+  db->Put(WriteOptions(), "key5", "value");
+  db->Flush(FlushOptions());
+  TEST_SYNC_POINT("CompactFilesTest.CapturingPendingFiles:1");
+
+  compaction_thread.join();
+
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+
+  delete db;
+
+  // Make sure we can reopen the DB.
+  s = DB::Open(options, db_name_, &db);
+  ASSERT_TRUE(s.ok());
+  assert(db);
+  delete db;
+}
+
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -1,4 +1,4 @@
-//  Copyright (c) 2014, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -93,7 +93,7 @@ Status CompactedDBImpl::Init(const Options& options) {
  mutex_.Lock();
  ColumnFamilyDescriptor cf(kDefaultColumnFamilyName,
                            ColumnFamilyOptions(options));
-  Status s = Recover({ cf }, true /* read only */, false);
+  Status s = Recover({cf}, true /* read only */, false, true);
  if (s.ok()) {
    cfd_ = reinterpret_cast<ColumnFamilyHandleImpl*>(
              DefaultColumnFamily())->cfd();
@@ -1,4 +1,4 @@
-//  Copyright (c) 2014, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -162,9 +162,6 @@ Compaction::Compaction(VersionStorageInfo* vstorage,
      deletion_compaction_(_deletion_compaction),
      inputs_(std::move(_inputs)),
      grandparents_(std::move(_grandparents)),
-      grandparent_index_(0),
-      seen_key_(false),
-      overlapped_bytes_(0),
      score_(_score),
      bottommost_level_(IsBottommostLevel(output_level_, vstorage, inputs_)),
      is_full_compaction_(IsFullCompaction(vstorage, inputs_)),
@@ -290,32 +287,6 @@ bool Compaction::KeyNotExistsBeyondOutputLevel(
  return true;
 }

-bool Compaction::ShouldStopBefore(const Slice& internal_key) {
-  // Scan to find earliest grandparent file that contains key.
-  const InternalKeyComparator* icmp = &cfd_->internal_comparator();
-  while (grandparent_index_ < grandparents_.size() &&
-      icmp->Compare(internal_key,
-                    grandparents_[grandparent_index_]->largest.Encode()) > 0) {
-    if (seen_key_) {
-      overlapped_bytes_ += grandparents_[grandparent_index_]->fd.GetFileSize();
-    }
-    assert(grandparent_index_ + 1 >= grandparents_.size() ||
-           icmp->Compare(grandparents_[grandparent_index_]->largest.Encode(),
-                         grandparents_[grandparent_index_+1]->smallest.Encode())
-                         < 0);
-    grandparent_index_++;
-  }
-  seen_key_ = true;
-
-  if (overlapped_bytes_ > max_grandparent_overlap_bytes_) {
-    // Too much overlap for current output; start new output
-    overlapped_bytes_ = 0;
-    return true;
-  } else {
-    return false;
-  }
-}
-
 // Mark (or clear) each file that is being compacted
 void Compaction::MarkFilesBeingCompacted(bool mark_as_compacted) {
  for (size_t i = 0; i < num_input_levels(); i++) {
@@ -421,7 +392,7 @@ void Compaction::Summary(char* output, int len) {
  snprintf(output + write, len - write, "]");
 }

-uint64_t Compaction::OutputFilePreallocationSize() {
+uint64_t Compaction::OutputFilePreallocationSize() const {
  uint64_t preallocation_size = 0;

  if (cfd_->ioptions()->compaction_style == kCompactionStyleLevel ||
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -104,7 +104,7 @@ class Compaction {
  }

  // Returns the LevelFilesBrief of the specified compaction input level.
-  LevelFilesBrief* input_levels(size_t compaction_input_level) {
+  const LevelFilesBrief* input_levels(size_t compaction_input_level) const {
    return &input_levels_[compaction_input_level];
  }

@@ -132,12 +132,10 @@ class Compaction {
  bool KeyNotExistsBeyondOutputLevel(const Slice& user_key,
                                     std::vector<size_t>* level_ptrs) const;

-  // Returns true iff we should stop building the current output
-  // before processing "internal_key".
-  bool ShouldStopBefore(const Slice& internal_key);
-
  // Clear all files to indicate that they are not being compacted
  // Delete this compaction from the list of running compactions.
+  //
+  // Requirement: DB mutex held
  void ReleaseCompactionFiles(Status status);

  // Returns the summary of the compaction in "output" with maximum "len"
@@ -149,13 +147,13 @@ class Compaction {
  double score() const { return score_; }

  // Is this compaction creating a file in the bottom most level?
-  bool bottommost_level() { return bottommost_level_; }
+  bool bottommost_level() const { return bottommost_level_; }

  // Does this compaction include all sst files?
-  bool is_full_compaction() { return is_full_compaction_; }
+  bool is_full_compaction() const { return is_full_compaction_; }

  // Was this compaction triggered manually by the client?
-  bool is_manual_compaction() { return is_manual_compaction_; }
+  bool is_manual_compaction() const { return is_manual_compaction_; }

  // Used when allow_trivial_move option is set in
  // Universal compaction. If all the input files are
@@ -168,19 +166,21 @@ class Compaction {
  // Used when allow_trivial_move option is set in
  // Universal compaction. Returns true, if the input files
  // are non-overlapping and can be trivially moved.
-  bool is_trivial_move() { return is_trivial_move_; }
+  bool is_trivial_move() const { return is_trivial_move_; }

  // How many total levels are there?
  int number_levels() const { return number_levels_; }

  // Return the MutableCFOptions that should be used throughout the compaction
  // procedure
-  const MutableCFOptions* mutable_cf_options() { return &mutable_cf_options_; }
+  const MutableCFOptions* mutable_cf_options() const {
+    return &mutable_cf_options_;
+  }

  // Returns the size in bytes that the output file should be preallocated to.
  // In level compaction, that is max_file_size_. In universal compaction, that
  // is the sum of all input file sizes.
-  uint64_t OutputFilePreallocationSize();
+  uint64_t OutputFilePreallocationSize() const;

  void SetInputVersion(Version* input_version);

@@ -223,6 +223,14 @@ class Compaction {

  CompactionReason compaction_reason() { return compaction_reason_; }

+  const std::vector<FileMetaData*>& grandparents() const {
+    return grandparents_;
+  }
+
+  uint64_t max_grandparent_overlap_bytes() const {
+    return max_grandparent_overlap_bytes_;
+  }
+
 private:
  // mark (or clear) all files that are being compacted
  void MarkFilesBeingCompacted(bool mark_as_compacted);
@@ -263,13 +271,9 @@ class Compaction {
  // A copy of inputs_, organized more closely in memory
  autovector<LevelFilesBrief, 2> input_levels_;

-  // State used to check for number of of overlapping grandparent files
+  // State used to check for number of overlapping grandparent files
  // (grandparent == "output_level_ + 1")
  std::vector<FileMetaData*> grandparents_;
-  size_t grandparent_index_;   // Index in grandparent_starts_
-  bool seen_key_;              // Some output key has been seen
-  uint64_t overlapped_bytes_;  // Bytes of overlap between current output
-                               // and grandparent files
  const double score_;         // score that was used to pick this compaction.

  // Is this compaction creating a file in the bottom most level?
@@ -1,6 +1,6 @@
 // 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.
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -14,7 +14,7 @@ CompactionIterator::CompactionIterator(
    InternalIterator* input, const Comparator* cmp, MergeHelper* merge_helper,
    SequenceNumber last_sequence, std::vector<SequenceNumber>* snapshots,
    SequenceNumber earliest_write_conflict_snapshot, Env* env,
-    bool expect_valid_internal_key, Compaction* compaction,
+    bool expect_valid_internal_key, const Compaction* compaction,
    const CompactionFilter* compaction_filter, LogBuffer* log_buffer)
    : input_(input),
      cmp_(cmp),
@@ -423,15 +423,15 @@ void CompactionIterator::PrepareOutput() {
 inline SequenceNumber CompactionIterator::findEarliestVisibleSnapshot(
    SequenceNumber in, SequenceNumber* prev_snapshot) {
  assert(snapshots_->size());
-  SequenceNumber prev __attribute__((unused)) = 0;
+  SequenceNumber prev __attribute__((__unused__)) = kMaxSequenceNumber;
  for (const auto cur : *snapshots_) {
-    assert(prev <= cur);
+    assert(prev == kMaxSequenceNumber || prev <= cur);
    if (cur >= in) {
-      *prev_snapshot = prev;
+      *prev_snapshot = prev == kMaxSequenceNumber ? 0 : prev;
      return cur;
    }
    prev = cur;
-    assert(prev);
+    assert(prev < kMaxSequenceNumber);
  }
  *prev_snapshot = prev;
  return kMaxSequenceNumber;
@@ -1,6 +1,6 @@
 // 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.
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -42,7 +42,7 @@ class CompactionIterator {
                     std::vector<SequenceNumber>* snapshots,
                     SequenceNumber earliest_write_conflict_snapshot, Env* env,
                     bool expect_valid_internal_key,
-                     Compaction* compaction = nullptr,
+                     const Compaction* compaction = nullptr,
                     const CompactionFilter* compaction_filter = nullptr,
                     LogBuffer* log_buffer = nullptr);

@@ -92,7 +92,7 @@ class CompactionIterator {
  const SequenceNumber earliest_write_conflict_snapshot_;
  Env* env_;
  bool expect_valid_internal_key_;
-  Compaction* compaction_;
+  const Compaction* compaction_;
  const CompactionFilter* compaction_filter_;
  LogBuffer* log_buffer_;
  bool bottommost_level_;
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -16,12 +16,13 @@
 #include <inttypes.h>
 #include <algorithm>
 #include <functional>
-#include <vector>
-#include <memory>
 #include <list>
+#include <memory>
+#include <random>
 #include <set>
 #include <thread>
 #include <utility>
+#include <vector>

 #include "db/builder.h"
 #include "db/db_iter.h"
@@ -51,6 +52,7 @@
 #include "util/iostats_context_imp.h"
 #include "util/log_buffer.h"
 #include "util/logging.h"
+#include "util/sst_file_manager_impl.h"
 #include "util/mutexlock.h"
 #include "util/perf_context_imp.h"
 #include "util/stop_watch.h"
@@ -62,7 +64,7 @@ namespace rocksdb {

 // Maintains state for each sub-compaction
 struct CompactionJob::SubcompactionState {
-  Compaction* compaction;
+  const Compaction* compaction;
  std::unique_ptr<CompactionIterator> c_iter;

  // The boundaries of the key-range this compaction is interested in. No two
@@ -103,6 +105,14 @@ struct CompactionJob::SubcompactionState {
  uint64_t num_output_records;
  CompactionJobStats compaction_job_stats;
  uint64_t approx_size;
+  // An index that used to speed up ShouldStopBefore().
+  size_t grandparent_index = 0;
+  // The number of bytes overlapping between the current output and
+  // grandparent files used in ShouldStopBefore().
+  uint64_t overlapped_bytes = 0;
+  // A flag determine whether the key has been seen in ShouldStopBefore()
+  bool seen_key = false;
+  std::string compression_dict;

  SubcompactionState(Compaction* c, Slice* _start, Slice* _end,
                     uint64_t size = 0)
@@ -114,7 +124,11 @@ struct CompactionJob::SubcompactionState {
        total_bytes(0),
        num_input_records(0),
        num_output_records(0),
-        approx_size(size) {
+        approx_size(size),
+        grandparent_index(0),
+        overlapped_bytes(0),
+        seen_key(false),
+        compression_dict() {
    assert(compaction != nullptr);
  }

@@ -133,6 +147,10 @@ struct CompactionJob::SubcompactionState {
    num_output_records = std::move(o.num_output_records);
    compaction_job_stats = std::move(o.compaction_job_stats);
    approx_size = std::move(o.approx_size);
+    grandparent_index = std::move(o.grandparent_index);
+    overlapped_bytes = std::move(o.overlapped_bytes);
+    seen_key = std::move(o.seen_key);
+    compression_dict = std::move(o.compression_dict);
    return *this;
  }

@@ -140,6 +158,38 @@ struct CompactionJob::SubcompactionState {
  SubcompactionState(const SubcompactionState&) = delete;

  SubcompactionState& operator=(const SubcompactionState&) = delete;
+
+  // Returns true iff we should stop building the current output
+  // before processing "internal_key".
+  bool ShouldStopBefore(const Slice& internal_key) {
+    const InternalKeyComparator* icmp =
+        &compaction->column_family_data()->internal_comparator();
+    const std::vector<FileMetaData*>& grandparents = compaction->grandparents();
+
+    // Scan to find earliest grandparent file that contains key.
+    while (grandparent_index < grandparents.size() &&
+           icmp->Compare(internal_key,
+                         grandparents[grandparent_index]->largest.Encode()) >
+               0) {
+      if (seen_key) {
+        overlapped_bytes += grandparents[grandparent_index]->fd.GetFileSize();
+      }
+      assert(grandparent_index + 1 >= grandparents.size() ||
+             icmp->Compare(
+                 grandparents[grandparent_index]->largest.Encode(),
+                 grandparents[grandparent_index + 1]->smallest.Encode()) <= 0);
+      grandparent_index++;
+    }
+    seen_key = true;
+
+    if (overlapped_bytes > compaction->max_grandparent_overlap_bytes()) {
+      // Too much overlap for current output; start new output
+      overlapped_bytes = 0;
+      return true;
+    }
+
+    return false;
+  }
 };

 // Maintains state for the entire compaction
@@ -211,6 +261,7 @@ CompactionJob::CompactionJob(
    const EnvOptions& env_options, VersionSet* versions,
    std::atomic<bool>* shutting_down, LogBuffer* log_buffer,
    Directory* db_directory, Directory* output_directory, Statistics* stats,
+    InstrumentedMutex* db_mutex, Status* db_bg_error,
    std::vector<SequenceNumber> existing_snapshots,
    SequenceNumber earliest_write_conflict_snapshot,
    std::shared_ptr<Cache> table_cache, EventLogger* event_logger,
@@ -230,6 +281,8 @@ CompactionJob::CompactionJob(
      db_directory_(db_directory),
      output_directory_(output_directory),
      stats_(stats),
+      db_mutex_(db_mutex),
+      db_bg_error_(db_bg_error),
      existing_snapshots_(std::move(existing_snapshots)),
      earliest_write_conflict_snapshot_(earliest_write_conflict_snapshot),
      table_cache_(std::move(table_cache)),
@@ -237,7 +290,9 @@ CompactionJob::CompactionJob(
      paranoid_file_checks_(paranoid_file_checks),
      measure_io_stats_(measure_io_stats) {
  assert(log_buffer_ != nullptr);
-  ThreadStatusUtil::SetColumnFamily(compact_->compaction->column_family_data());
+  const auto* cfd = compact_->compaction->column_family_data();
+  ThreadStatusUtil::SetColumnFamily(cfd, cfd->ioptions()->env,
+                                    cfd->options()->enable_thread_tracking);
  ThreadStatusUtil::SetThreadOperation(ThreadStatus::OP_COMPACTION);
  ReportStartedCompaction(compaction);
 }
@@ -249,8 +304,9 @@ CompactionJob::~CompactionJob() {

 void CompactionJob::ReportStartedCompaction(
    Compaction* compaction) {
-  ThreadStatusUtil::SetColumnFamily(
-      compact_->compaction->column_family_data());
+  const auto* cfd = compact_->compaction->column_family_data();
+  ThreadStatusUtil::SetColumnFamily(cfd, cfd->ioptions()->env,
+                                    cfd->options()->enable_thread_tracking);

  ThreadStatusUtil::SetThreadOperationProperty(
      ThreadStatus::COMPACTION_JOB_ID,
@@ -356,7 +412,7 @@ void CompactionJob::GenSubcompactionBoundaries() {
      size_t num_files = flevel->num_files;

      if (num_files == 0) {
-        break;
+        continue;
      }

      if (lvl == 0) {
@@ -415,12 +471,9 @@ void CompactionJob::GenSubcompactionBoundaries() {

  // Group the ranges into subcompactions
  const double min_file_fill_percent = 4.0 / 5;
-  uint64_t max_output_files = 
-    static_cast<uint64_t>(
-        std::ceil(
-          sum / min_file_fill_percent /
-          cfd->GetCurrentMutableCFOptions()->MaxFileSizeForLevel(out_lvl))
-          );
+  uint64_t max_output_files = static_cast<uint64_t>(std::ceil(
+      sum / min_file_fill_percent /
+      cfd->GetCurrentMutableCFOptions()->MaxFileSizeForLevel(out_lvl)));
  uint64_t subcompactions =
      std::min({static_cast<uint64_t>(ranges.size()),
                static_cast<uint64_t>(db_options_.max_subcompactions),
@@ -518,18 +571,17 @@ Status CompactionJob::Run() {
  return status;
 }

-Status CompactionJob::Install(const MutableCFOptions& mutable_cf_options,
-                              InstrumentedMutex* db_mutex) {
+Status CompactionJob::Install(const MutableCFOptions& mutable_cf_options) {
  AutoThreadOperationStageUpdater stage_updater(
      ThreadStatus::STAGE_COMPACTION_INSTALL);
-  db_mutex->AssertHeld();
+  db_mutex_->AssertHeld();
  Status status = compact_->status;
  ColumnFamilyData* cfd = compact_->compaction->column_family_data();
  cfd->internal_stats()->AddCompactionStats(
      compact_->compaction->output_level(), compaction_stats_);

  if (status.ok()) {
-    status = InstallCompactionResults(mutable_cf_options, db_mutex);
+    status = InstallCompactionResults(mutable_cf_options);
  }
  VersionStorageInfo::LevelSummaryStorage tmp;
  auto vstorage = cfd->current()->storage_info();
@@ -617,6 +669,31 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
  }

  ColumnFamilyData* cfd = sub_compact->compaction->column_family_data();
+  const MutableCFOptions* mutable_cf_options =
+      sub_compact->compaction->mutable_cf_options();
+
+  // To build compression dictionary, we sample the first output file, assuming
+  // it'll reach the maximum length, and then use the dictionary for compressing
+  // subsequent output files. The dictionary may be less than max_dict_bytes if
+  // the first output file's length is less than the maximum.
+  const int kSampleLenShift = 6;  // 2^6 = 64-byte samples
+  std::set<size_t> sample_begin_offsets;
+  if (bottommost_level_ &&
+      cfd->ioptions()->compression_opts.max_dict_bytes > 0) {
+    const size_t kMaxSamples =
+        cfd->ioptions()->compression_opts.max_dict_bytes >> kSampleLenShift;
+    const size_t kOutFileLen = mutable_cf_options->MaxFileSizeForLevel(
+        compact_->compaction->output_level());
+    if (kOutFileLen != port::kMaxSizet) {
+      const size_t kOutFileNumSamples = kOutFileLen >> kSampleLenShift;
+      Random64 generator{versions_->NewFileNumber()};
+      for (size_t i = 0; i < kMaxSamples; ++i) {
+        sample_begin_offsets.insert(generator.Uniform(kOutFileNumSamples)
+                                    << kSampleLenShift);
+      }
+    }
+  }
+
  auto compaction_filter = cfd->ioptions()->compaction_filter;
  std::unique_ptr<CompactionFilter> compaction_filter_from_factory = nullptr;
  if (compaction_filter == nullptr) {
@@ -652,6 +729,13 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
  auto c_iter = sub_compact->c_iter.get();
  c_iter->SeekToFirst();
  const auto& c_iter_stats = c_iter->iter_stats();
+  auto sample_begin_offset_iter = sample_begin_offsets.cbegin();
+  // data_begin_offset and compression_dict are only valid while generating
+  // dictionary from the first output file.
+  size_t data_begin_offset = 0;
+  std::string compression_dict;
+  compression_dict.reserve(cfd->ioptions()->compression_opts.max_dict_bytes);
+
  // TODO(noetzli): check whether we could check !shutting_down_->... only
  // only occasionally (see diff D42687)
  while (status.ok() && !shutting_down_->load(std::memory_order_acquire) &&
@@ -666,7 +750,7 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
    if (end != nullptr &&
        cfd->user_comparator()->Compare(c_iter->user_key(), *end) >= 0) {
      break;
-    } else if (sub_compact->compaction->ShouldStopBefore(key) &&
+    } else if (sub_compact->ShouldStopBefore(key) &&
               sub_compact->builder != nullptr) {
      status = FinishCompactionOutputFile(input->status(), sub_compact);
      if (!status.ok()) {
@@ -695,6 +779,55 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
        key, c_iter->ikey().sequence);
    sub_compact->num_output_records++;

+    if (sub_compact->outputs.size() == 1) {  // first output file
+      // Check if this key/value overlaps any sample intervals; if so, appends
+      // overlapping portions to the dictionary.
+      for (const auto& data_elmt : {key, value}) {
+        size_t data_end_offset = data_begin_offset + data_elmt.size();
+        while (sample_begin_offset_iter != sample_begin_offsets.cend() &&
+               *sample_begin_offset_iter < data_end_offset) {
+          size_t sample_end_offset =
+              *sample_begin_offset_iter + (1 << kSampleLenShift);
+          // Invariant: Because we advance sample iterator while processing the
+          // data_elmt containing the sample's last byte, the current sample
+          // cannot end before the current data_elmt.
+          assert(data_begin_offset < sample_end_offset);
+
+          size_t data_elmt_copy_offset, data_elmt_copy_len;
+          if (*sample_begin_offset_iter <= data_begin_offset) {
+            // The sample starts before data_elmt starts, so take bytes starting
+            // at the beginning of data_elmt.
+            data_elmt_copy_offset = 0;
+          } else {
+            // data_elmt starts before the sample starts, so take bytes starting
+            // at the below offset into data_elmt.
+            data_elmt_copy_offset =
+                *sample_begin_offset_iter - data_begin_offset;
+          }
+          if (sample_end_offset <= data_end_offset) {
+            // The sample ends before data_elmt ends, so take as many bytes as
+            // needed.
+            data_elmt_copy_len =
+                sample_end_offset - (data_begin_offset + data_elmt_copy_offset);
+          } else {
+            // data_elmt ends before the sample ends, so take all remaining
+            // bytes in data_elmt.
+            data_elmt_copy_len =
+                data_end_offset - (data_begin_offset + data_elmt_copy_offset);
+          }
+          compression_dict.append(&data_elmt.data()[data_elmt_copy_offset],
+                                  data_elmt_copy_len);
+          if (sample_end_offset > data_end_offset) {
+            // Didn't finish sample. Try to finish it with the next data_elmt.
+            break;
+          }
+          // Next sample may require bytes from same data_elmt.
+          sample_begin_offset_iter++;
+        }
+        data_begin_offset = data_end_offset;
+      }
+    }
+
    // Close output file if it is big enough
    // TODO(aekmekji): determine if file should be closed earlier than this
    // during subcompactions (i.e. if output size, estimated by input size, is
@@ -703,8 +836,12 @@ void CompactionJob::ProcessKeyValueCompaction(SubcompactionState* sub_compact) {
    if (sub_compact->builder->FileSize() >=
        sub_compact->compaction->max_output_file_size()) {
      status = FinishCompactionOutputFile(input->status(), sub_compact);
+      if (sub_compact->outputs.size() == 1) {
+        // Use dictionary from first output file for compression of subsequent
+        // files.
+        sub_compact->compression_dict = std::move(compression_dict);
+      }
    }
-
    c_iter->Next();
  }

@@ -817,9 +954,10 @@ Status CompactionJob::FinishCompactionOutputFile(
  }
  sub_compact->outfile.reset();

+  ColumnFamilyData* cfd = sub_compact->compaction->column_family_data();
+  TableProperties tp;
  if (s.ok() && current_entries > 0) {
    // Verify that the table is usable
-    ColumnFamilyData* cfd = sub_compact->compaction->column_family_data();
    InternalIterator* iter = cfd->table_cache()->NewIterator(
        ReadOptions(), env_options_, cfd->internal_comparator(), meta->fd,
        nullptr, cfd->internal_stats()->GetFileReadHist(
@@ -833,35 +971,51 @@ Status CompactionJob::FinishCompactionOutputFile(
    }

    delete iter;
+
+    // Output to event logger and fire events.
    if (s.ok()) {
-      auto tp = sub_compact->builder->GetTableProperties();
+      tp = sub_compact->builder->GetTableProperties();
      sub_compact->current_output()->table_properties =
          std::make_shared<TableProperties>(tp);
-      TableFileCreationInfo info(std::move(tp));
-      info.db_name = dbname_;
-      info.cf_name = cfd->GetName();
-      info.file_path =
-          TableFileName(cfd->ioptions()->db_paths, meta->fd.GetNumber(),
-                        meta->fd.GetPathId());
-      info.file_size = meta->fd.GetFileSize();
-      info.job_id = job_id_;
      Log(InfoLogLevel::INFO_LEVEL, db_options_.info_log,
          "[%s] [JOB %d] Generated table #%" PRIu64 ": %" PRIu64
          " keys, %" PRIu64 " bytes%s",
          cfd->GetName().c_str(), job_id_, output_number, current_entries,
          current_bytes,
          meta->marked_for_compaction ? " (need compaction)" : "");
-      EventHelpers::LogAndNotifyTableFileCreation(
-          event_logger_, cfd->ioptions()->listeners, meta->fd, info);
    }
  }
+  std::string fname = TableFileName(db_options_.db_paths, meta->fd.GetNumber(),
+                                    meta->fd.GetPathId());
+  EventHelpers::LogAndNotifyTableFileCreationFinished(
+      event_logger_, cfd->ioptions()->listeners, dbname_, cfd->GetName(), fname,
+      job_id_, meta->fd, tp, TableFileCreationReason::kCompaction, s);
+
+  // Report new file to SstFileManagerImpl
+  auto sfm =
+      static_cast<SstFileManagerImpl*>(db_options_.sst_file_manager.get());
+  if (sfm && meta->fd.GetPathId() == 0) {
+    auto fn = TableFileName(cfd->ioptions()->db_paths, meta->fd.GetNumber(),
+                            meta->fd.GetPathId());
+    sfm->OnAddFile(fn);
+    if (sfm->IsMaxAllowedSpaceReached()) {
+      InstrumentedMutexLock l(db_mutex_);
+      if (db_bg_error_->ok()) {
+        s = Status::IOError("Max allowed space was reached");
+        *db_bg_error_ = s;
+        TEST_SYNC_POINT(
+            "CompactionJob::FinishCompactionOutputFile:MaxAllowedSpaceReached");
+      }
+    }
+  }
+
  sub_compact->builder.reset();
  return s;
 }

 Status CompactionJob::InstallCompactionResults(
-    const MutableCFOptions& mutable_cf_options, InstrumentedMutex* db_mutex) {
-  db_mutex->AssertHeld();
+    const MutableCFOptions& mutable_cf_options) {
+  db_mutex_->AssertHeld();

  auto* compaction = compact_->compaction;
  // paranoia: verify that the files that we started with
@@ -896,7 +1050,7 @@ Status CompactionJob::InstallCompactionResults(
  }
  return versions_->LogAndApply(compaction->column_family_data(),
                                mutable_cf_options, compaction->edit(),
-                                db_mutex, db_directory_);
+                                db_mutex_, db_directory_);
 }

 void CompactionJob::RecordCompactionIOStats() {
@@ -916,10 +1070,17 @@ Status CompactionJob::OpenCompactionOutputFile(
  assert(sub_compact->builder == nullptr);
  // no need to lock because VersionSet::next_file_number_ is atomic
  uint64_t file_number = versions_->NewFileNumber();
-  // Make the output file
-  unique_ptr<WritableFile> writable_file;
  std::string fname = TableFileName(db_options_.db_paths, file_number,
                                    sub_compact->compaction->output_path_id());
+  // Fire events.
+  ColumnFamilyData* cfd = sub_compact->compaction->column_family_data();
+#ifndef ROCKSDB_LITE
+  EventHelpers::NotifyTableFileCreationStarted(
+      cfd->ioptions()->listeners, dbname_, cfd->GetName(), fname, job_id_,
+      TableFileCreationReason::kCompaction);
+#endif  // !ROCKSDB_LITE
+  // Make the output file
+  unique_ptr<WritableFile> writable_file;
  Status s = NewWritableFile(env_, fname, &writable_file, env_options_);
  if (!s.ok()) {
    Log(InfoLogLevel::ERROR_LEVEL, db_options_.info_log,
@@ -928,8 +1089,13 @@ Status CompactionJob::OpenCompactionOutputFile(
        sub_compact->compaction->column_family_data()->GetName().c_str(),
        job_id_, file_number, s.ToString().c_str());
    LogFlush(db_options_.info_log);
+    EventHelpers::LogAndNotifyTableFileCreationFinished(
+        event_logger_, cfd->ioptions()->listeners, dbname_, cfd->GetName(),
+        fname, job_id_, FileDescriptor(), TableProperties(),
+        TableFileCreationReason::kCompaction, s);
    return s;
  }
+
  SubcompactionState::Output out;
  out.meta.fd =
      FileDescriptor(file_number, sub_compact->compaction->output_path_id(), 0);
@@ -942,7 +1108,6 @@ Status CompactionJob::OpenCompactionOutputFile(
  sub_compact->outfile.reset(
      new WritableFileWriter(std::move(writable_file), env_options_));

-  ColumnFamilyData* cfd = sub_compact->compaction->column_family_data();
  // If the Column family flag is to only optimize filters for hits,
  // we can skip creating filters if this is the bottommost_level where
  // data is going to be found
@@ -950,9 +1115,10 @@ Status CompactionJob::OpenCompactionOutputFile(
      cfd->ioptions()->optimize_filters_for_hits && bottommost_level_;
  sub_compact->builder.reset(NewTableBuilder(
      *cfd->ioptions(), cfd->internal_comparator(),
-      cfd->int_tbl_prop_collector_factories(), cfd->GetID(),
+      cfd->int_tbl_prop_collector_factories(), cfd->GetID(), cfd->GetName(),
      sub_compact->outfile.get(), sub_compact->compaction->output_compression(),
-      cfd->ioptions()->compression_opts, skip_filters));
+      cfd->ioptions()->compression_opts, &sub_compact->compression_dict,
+      skip_filters));
  LogFlush(db_options_.info_log);
  return s;
 }
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -56,7 +56,8 @@ class CompactionJob {
                const EnvOptions& env_options, VersionSet* versions,
                std::atomic<bool>* shutting_down, LogBuffer* log_buffer,
                Directory* db_directory, Directory* output_directory,
-                Statistics* stats,
+                Statistics* stats, InstrumentedMutex* db_mutex,
+                Status* db_bg_error,
                std::vector<SequenceNumber> existing_snapshots,
                SequenceNumber earliest_write_conflict_snapshot,
                std::shared_ptr<Cache> table_cache, EventLogger* event_logger,
@@ -77,8 +78,7 @@ class CompactionJob {
  Status Run();

  // REQUIRED: mutex held
-  Status Install(const MutableCFOptions& mutable_cf_options,
-                 InstrumentedMutex* db_mutex);
+  Status Install(const MutableCFOptions& mutable_cf_options);

 private:
  struct SubcompactionState;
@@ -95,8 +95,7 @@ class CompactionJob {

  Status FinishCompactionOutputFile(const Status& input_status,
                                    SubcompactionState* sub_compact);
-  Status InstallCompactionResults(const MutableCFOptions& mutable_cf_options,
-                                  InstrumentedMutex* db_mutex);
+  Status InstallCompactionResults(const MutableCFOptions& mutable_cf_options);
  void RecordCompactionIOStats();
  Status OpenCompactionOutputFile(SubcompactionState* sub_compact);
  void CleanupCompaction();
@@ -130,6 +129,8 @@ class CompactionJob {
  Directory* db_directory_;
  Directory* output_directory_;
  Statistics* stats_;
+  InstrumentedMutex* db_mutex_;
+  Status* db_bg_error_;
  // If there were two snapshots with seq numbers s1 and
  // s2 and s1 < s2, and if we find two instances of a key k1 then lies
  // entirely within s1 and s2, then the earlier version of k1 can be safely
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -624,7 +624,10 @@ CompressionType GetAnyCompression() {
    return kBZip2Compression;
  } else if (LZ4_Supported()) {
    return kLZ4Compression;
+  } else if (XPRESS_Supported()) {
+    return kXpressCompression;
  }
+
  return kNoCompression;
 }

@@ -660,7 +663,7 @@ TEST_P(CompactionJobStatsTest, CompactionJobStatsTest) {
  options.max_subcompactions = max_subcompactions_;
  options.bytes_per_sync = 512 * 1024;

-  options.compaction_measure_io_stats = true;
+  options.report_bg_io_stats = true;
  for (int test = 0; test < 2; ++test) {
    DestroyAndReopen(options);
    CreateAndReopenWithCF({"pikachu"}, options);
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -250,9 +250,9 @@ class CompactionJobTest : public testing::Test {
    EventLogger event_logger(db_options_.info_log.get());
    CompactionJob compaction_job(
        0, &compaction, db_options_, env_options_, versions_.get(),
-        &shutting_down_, &log_buffer, nullptr, nullptr, nullptr, snapshots,
-        earliest_write_conflict_snapshot, table_cache_, &event_logger, false,
-        false, dbname_, &compaction_job_stats_);
+        &shutting_down_, &log_buffer, nullptr, nullptr, nullptr, &mutex_,
+        &bg_error_, snapshots, earliest_write_conflict_snapshot, table_cache_,
+        &event_logger, false, false, dbname_, &compaction_job_stats_);

    VerifyInitializationOfCompactionJobStats(compaction_job_stats_);

@@ -262,8 +262,7 @@ class CompactionJobTest : public testing::Test {
    s = compaction_job.Run();
    ASSERT_OK(s);
    mutex_.Lock();
-    ASSERT_OK(compaction_job.Install(*cfd->GetLatestMutableCFOptions(),
-                                     &mutex_));
+    ASSERT_OK(compaction_job.Install(*cfd->GetLatestMutableCFOptions()));
    mutex_.Unlock();

    if (expected_results.size() == 0) {
@@ -295,6 +294,7 @@ class CompactionJobTest : public testing::Test {
  ColumnFamilyData* cfd_;
  std::unique_ptr<CompactionFilter> compaction_filter_;
  std::shared_ptr<MergeOperator> merge_op_;
+  Status bg_error_;
 };

 TEST_F(CompactionJobTest, Simple) {
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -141,7 +141,8 @@ CompactionPicker::~CompactionPicker() {}

 // Delete this compaction from the list of running compactions.
 void CompactionPicker::ReleaseCompactionFiles(Compaction* c, Status status) {
-  if (c->start_level() == 0) {
+  if (c->start_level() == 0 ||
+      ioptions_.compaction_style == kCompactionStyleUniversal) {
    level0_compactions_in_progress_.erase(c);
  }
  if (!status.ok()) {
@@ -612,6 +613,17 @@ Compaction* CompactionPicker::CompactRange(
  if (input_level == 0) {
    level0_compactions_in_progress_.insert(compaction);
  }
+
+  // Creating a compaction influences the compaction score because the score
+  // takes running compactions into account (by skipping files that are already
+  // being compacted). Since we just changed compaction score, we recalculate it
+  // here
+  {  // this piece of code recomputes compaction score
+    CompactionOptionsFIFO dummy_compaction_options_fifo;
+    vstorage->ComputeCompactionScore(mutable_cf_options,
+                                     dummy_compaction_options_fifo);
+  }
+
  return compaction;
 }

@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -83,6 +83,8 @@ class CompactionPicker {
 #endif  // ROCKSDB_LITE

  // Free up the files that participated in a compaction
+  //
+  // Requirement: DB mutex held
  void ReleaseCompactionFiles(Compaction* c, Status status);

  // Returns true if any one of the specified files are being compacted
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -79,7 +79,7 @@ class CompactionPickerTest : public testing::Test {
  }

  void Add(int level, uint32_t file_number, const char* smallest,
-           const char* largest, uint64_t file_size = 0, uint32_t path_id = 0,
+           const char* largest, uint64_t file_size = 1, uint32_t path_id = 0,
           SequenceNumber smallest_seq = 100,
           SequenceNumber largest_seq = 100) {
    assert(level < vstorage_->num_levels());
@@ -195,6 +195,7 @@ TEST_F(CompactionPickerTest, LevelMaxScore) {
  NewVersionStorage(6, kCompactionStyleLevel);
  mutable_cf_options_.target_file_size_base = 10000000;
  mutable_cf_options_.target_file_size_multiplier = 10;
+  mutable_cf_options_.max_bytes_for_level_base = 10 * 1024 * 1024;
  Add(0, 1U, "150", "200", 1000000000U);
  // Level 1 score 1.2
  Add(1, 66U, "150", "200", 6000000U);
@@ -319,6 +320,7 @@ TEST_F(CompactionPickerTest, Level0TriggerDynamic4) {
  mutable_cf_options_.level0_file_num_compaction_trigger = 2;
  mutable_cf_options_.max_bytes_for_level_base = 200;
  mutable_cf_options_.max_bytes_for_level_multiplier = 10;
+
  NewVersionStorage(num_levels, kCompactionStyleLevel);
  Add(0, 1U, "150", "200");
  Add(0, 2U, "200", "250");
@@ -351,6 +353,7 @@ TEST_F(CompactionPickerTest, LevelTriggerDynamic4) {
  mutable_cf_options_.level0_file_num_compaction_trigger = 2;
  mutable_cf_options_.max_bytes_for_level_base = 200;
  mutable_cf_options_.max_bytes_for_level_multiplier = 10;
+  mutable_cf_options_.compaction_pri = kMinOverlappingRatio;
  NewVersionStorage(num_levels, kCompactionStyleLevel);
  Add(0, 1U, "150", "200");
  Add(num_levels - 1, 3U, "200", "250", 300U);
@@ -366,11 +369,9 @@ TEST_F(CompactionPickerTest, LevelTriggerDynamic4) {
      cf_name_, mutable_cf_options_, vstorage_.get(), &log_buffer_));
  ASSERT_TRUE(compaction.get() != nullptr);
  ASSERT_EQ(1U, compaction->num_input_files(0));
-  ASSERT_EQ(6U, compaction->input(0, 0)->fd.GetNumber());
-  ASSERT_EQ(2U, compaction->num_input_files(1));
-  ASSERT_EQ(3U, compaction->input(1, 0)->fd.GetNumber());
-  ASSERT_EQ(4U, compaction->input(1, 1)->fd.GetNumber());
-  ASSERT_EQ(2U, compaction->num_input_levels());
+  ASSERT_EQ(5U, compaction->input(0, 0)->fd.GetNumber());
+  ASSERT_EQ(0, compaction->num_input_files(1));
+  ASSERT_EQ(1U, compaction->num_input_levels());
  ASSERT_EQ(num_levels - 1, compaction->output_level());
 }

@@ -487,6 +488,90 @@ TEST_F(CompactionPickerTest, NeedsCompactionFIFO) {
 }
 #endif  // ROCKSDB_LITE

+TEST_F(CompactionPickerTest, CompactionPriMinOverlapping1) {
+  NewVersionStorage(6, kCompactionStyleLevel);
+  mutable_cf_options_.target_file_size_base = 10000000;
+  mutable_cf_options_.target_file_size_multiplier = 10;
+  mutable_cf_options_.max_bytes_for_level_base = 10 * 1024 * 1024;
+  mutable_cf_options_.compaction_pri = kMinOverlappingRatio;
+
+  Add(2, 6U, "150", "179", 50000000U);
+  Add(2, 7U, "180", "220", 50000000U);
+  Add(2, 8U, "321", "400", 50000000U);  // File not overlapping
+  Add(2, 9U, "721", "800", 50000000U);
+
+  Add(3, 26U, "150", "170", 260000000U);
+  Add(3, 27U, "171", "179", 260000000U);
+  Add(3, 28U, "191", "220", 260000000U);
+  Add(3, 29U, "221", "300", 260000000U);
+  Add(3, 30U, "750", "900", 260000000U);
+  UpdateVersionStorageInfo();
+
+  std::unique_ptr<Compaction> compaction(level_compaction_picker.PickCompaction(
+      cf_name_, mutable_cf_options_, vstorage_.get(), &log_buffer_));
+  ASSERT_TRUE(compaction.get() != nullptr);
+  ASSERT_EQ(1U, compaction->num_input_files(0));
+  // Pick file 8 because it overlaps with 0 files on level 3.
+  ASSERT_EQ(8U, compaction->input(0, 0)->fd.GetNumber());
+}
+
+TEST_F(CompactionPickerTest, CompactionPriMinOverlapping2) {
+  NewVersionStorage(6, kCompactionStyleLevel);
+  mutable_cf_options_.target_file_size_base = 10000000;
+  mutable_cf_options_.target_file_size_multiplier = 10;
+  mutable_cf_options_.max_bytes_for_level_base = 10 * 1024 * 1024;
+  mutable_cf_options_.compaction_pri = kMinOverlappingRatio;
+
+  Add(2, 6U, "150", "175",
+      60000000U);  // Overlaps with file 26, 27, total size 521M
+  Add(2, 7U, "176", "200", 60000000U);  // Overlaps with file 27, 28, total size
+                                        // 520M, the smalelst overlapping
+  Add(2, 8U, "201", "300",
+      60000000U);  // Overlaps with file 28, 29, total size 521M
+
+  Add(3, 26U, "100", "110", 261000000U);
+  Add(3, 26U, "150", "170", 261000000U);
+  Add(3, 27U, "171", "179", 260000000U);
+  Add(3, 28U, "191", "220", 260000000U);
+  Add(3, 29U, "221", "300", 261000000U);
+  Add(3, 30U, "321", "400", 261000000U);
+  UpdateVersionStorageInfo();
+
+  std::unique_ptr<Compaction> compaction(level_compaction_picker.PickCompaction(
+      cf_name_, mutable_cf_options_, vstorage_.get(), &log_buffer_));
+  ASSERT_TRUE(compaction.get() != nullptr);
+  ASSERT_EQ(1U, compaction->num_input_files(0));
+  // Picking file 7 because overlapping ratio is the biggest.
+  ASSERT_EQ(7U, compaction->input(0, 0)->fd.GetNumber());
+}
+
+TEST_F(CompactionPickerTest, CompactionPriMinOverlapping3) {
+  NewVersionStorage(6, kCompactionStyleLevel);
+  mutable_cf_options_.max_bytes_for_level_base = 10000000;
+  mutable_cf_options_.max_bytes_for_level_multiplier = 10;
+  mutable_cf_options_.compaction_pri = kMinOverlappingRatio;
+
+  // file 7 and 8 over lap with the same file, but file 8 is smaller so
+  // it will be picked.
+  Add(2, 6U, "150", "167", 60000000U);  // Overlaps with file 26, 27
+  Add(2, 7U, "168", "169", 60000000U);  // Overlaps with file 27
+  Add(2, 8U, "201", "300", 61000000U);  // Overlaps with file 28, but the file
+                                        // itself is larger. Should be picked.
+
+  Add(3, 26U, "160", "165", 260000000U);
+  Add(3, 27U, "166", "170", 260000000U);
+  Add(3, 28U, "180", "400", 260000000U);
+  Add(3, 29U, "401", "500", 260000000U);
+  UpdateVersionStorageInfo();
+
+  std::unique_ptr<Compaction> compaction(level_compaction_picker.PickCompaction(
+      cf_name_, mutable_cf_options_, vstorage_.get(), &log_buffer_));
+  ASSERT_TRUE(compaction.get() != nullptr);
+  ASSERT_EQ(1U, compaction->num_input_files(0));
+  // Picking file 8 because overlapping ratio is the biggest.
+  ASSERT_EQ(8U, compaction->input(0, 0)->fd.GetNumber());
+}
+
 // This test exhibits the bug where we don't properly reset parent_index in
 // PickCompaction()
 TEST_F(CompactionPickerTest, ParentIndexResetBug) {
@@ -514,6 +599,8 @@ TEST_F(CompactionPickerTest, ParentIndexResetBug) {
 // ranges (with different sequence numbers) in the input files.
 TEST_F(CompactionPickerTest, OverlappingUserKeys) {
  NewVersionStorage(6, kCompactionStyleLevel);
+  mutable_cf_options_.compaction_pri = kByCompensatedSize;
+
  Add(1, 1U, "100", "150", 1U);
  // Overlapping user keys
  Add(1, 2U, "200", "400", 1U);
@@ -1,6 +1,6 @@
 // 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.
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -42,6 +42,7 @@ class CorruptionTest : public testing::Test {

  CorruptionTest() {
    tiny_cache_ = NewLRUCache(100);
+    options_.wal_recovery_mode = WALRecoveryMode::kTolerateCorruptedTailRecords;
    options_.env = &env_;
    dbname_ = test::TmpDir() + "/corruption_test";
    DestroyDB(dbname_, options_);
@@ -1,4 +1,4 @@
-//  Copyright (c) 2014, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -0,0 +1,454 @@
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+//
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+#include <cstdlib>
+#include "db/db_test_util.h"
+#include "port/stack_trace.h"
+
+namespace rocksdb {
+
+class DBBlockCacheTest : public DBTestBase {
+ private:
+  size_t miss_count_ = 0;
+  size_t hit_count_ = 0;
+  size_t insert_count_ = 0;
+  size_t failure_count_ = 0;
+  size_t compressed_miss_count_ = 0;
+  size_t compressed_hit_count_ = 0;
+  size_t compressed_insert_count_ = 0;
+  size_t compressed_failure_count_ = 0;
+
+ public:
+  const size_t kNumBlocks = 10;
+  const size_t kValueSize = 100;
+
+  DBBlockCacheTest() : DBTestBase("/db_block_cache_test") {}
+
+  BlockBasedTableOptions GetTableOptions() {
+    BlockBasedTableOptions table_options;
+    // Set a small enough block size so that each key-value get its own block.
+    table_options.block_size = 1;
+    return table_options;
+  }
+
+  Options GetOptions(const BlockBasedTableOptions& table_options) {
+    Options options = CurrentOptions();
+    options.create_if_missing = true;
+    // options.compression = kNoCompression;
+    options.statistics = rocksdb::CreateDBStatistics();
+    options.table_factory.reset(new BlockBasedTableFactory(table_options));
+    return options;
+  }
+
+  void InitTable(const Options& options) {
+    std::string value(kValueSize, 'a');
+    for (size_t i = 0; i < kNumBlocks; i++) {
+      ASSERT_OK(Put(ToString(i), value.c_str()));
+    }
+  }
+
+  void RecordCacheCounters(const Options& options) {
+    miss_count_ = TestGetTickerCount(options, BLOCK_CACHE_MISS);
+    hit_count_ = TestGetTickerCount(options, BLOCK_CACHE_HIT);
+    insert_count_ = TestGetTickerCount(options, BLOCK_CACHE_ADD);
+    failure_count_ = TestGetTickerCount(options, BLOCK_CACHE_ADD_FAILURES);
+    compressed_miss_count_ =
+        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_MISS);
+    compressed_hit_count_ =
+        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_HIT);
+    compressed_insert_count_ =
+        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_ADD);
+    compressed_failure_count_ =
+        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_ADD_FAILURES);
+  }
+
+  void CheckCacheCounters(const Options& options, size_t expected_misses,
+                          size_t expected_hits, size_t expected_inserts,
+                          size_t expected_failures) {
+    size_t new_miss_count = TestGetTickerCount(options, BLOCK_CACHE_MISS);
+    size_t new_hit_count = TestGetTickerCount(options, BLOCK_CACHE_HIT);
+    size_t new_insert_count = TestGetTickerCount(options, BLOCK_CACHE_ADD);
+    size_t new_failure_count =
+        TestGetTickerCount(options, BLOCK_CACHE_ADD_FAILURES);
+    ASSERT_EQ(miss_count_ + expected_misses, new_miss_count);
+    ASSERT_EQ(hit_count_ + expected_hits, new_hit_count);
+    ASSERT_EQ(insert_count_ + expected_inserts, new_insert_count);
+    ASSERT_EQ(failure_count_ + expected_failures, new_failure_count);
+    miss_count_ = new_miss_count;
+    hit_count_ = new_hit_count;
+    insert_count_ = new_insert_count;
+    failure_count_ = new_failure_count;
+  }
+
+  void CheckCompressedCacheCounters(const Options& options,
+                                    size_t expected_misses,
+                                    size_t expected_hits,
+                                    size_t expected_inserts,
+                                    size_t expected_failures) {
+    size_t new_miss_count =
+        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_MISS);
+    size_t new_hit_count =
+        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_HIT);
+    size_t new_insert_count =
+        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_ADD);
+    size_t new_failure_count =
+        TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_ADD_FAILURES);
+    ASSERT_EQ(compressed_miss_count_ + expected_misses, new_miss_count);
+    ASSERT_EQ(compressed_hit_count_ + expected_hits, new_hit_count);
+    ASSERT_EQ(compressed_insert_count_ + expected_inserts, new_insert_count);
+    ASSERT_EQ(compressed_failure_count_ + expected_failures, new_failure_count);
+    compressed_miss_count_ = new_miss_count;
+    compressed_hit_count_ = new_hit_count;
+    compressed_insert_count_ = new_insert_count;
+    compressed_failure_count_ = new_failure_count;
+  }
+};
+
+TEST_F(DBBlockCacheTest, TestWithoutCompressedBlockCache) {
+  ReadOptions read_options;
+  auto table_options = GetTableOptions();
+  auto options = GetOptions(table_options);
+  InitTable(options);
+
+  std::shared_ptr<Cache> cache = NewLRUCache(0, 0, false);
+  table_options.block_cache = cache;
+  options.table_factory.reset(new BlockBasedTableFactory(table_options));
+  Reopen(options);
+  RecordCacheCounters(options);
+
+  std::vector<std::unique_ptr<Iterator>> iterators(kNumBlocks - 1);
+  Iterator* iter = nullptr;
+
+  // Load blocks into cache.
+  for (size_t i = 0; i < kNumBlocks - 1; i++) {
+    iter = db_->NewIterator(read_options);
+    iter->Seek(ToString(i));
+    ASSERT_OK(iter->status());
+    CheckCacheCounters(options, 1, 0, 1, 0);
+    iterators[i].reset(iter);
+  }
+  size_t usage = cache->GetUsage();
+  ASSERT_LT(0, usage);
+  cache->SetCapacity(usage);
+  ASSERT_EQ(usage, cache->GetPinnedUsage());
+
+  // Test with strict capacity limit.
+  cache->SetStrictCapacityLimit(true);
+  iter = db_->NewIterator(read_options);
+  iter->Seek(ToString(kNumBlocks - 1));
+  ASSERT_TRUE(iter->status().IsIncomplete());
+  CheckCacheCounters(options, 1, 0, 0, 1);
+  delete iter;
+  iter = nullptr;
+
+  // Release interators and access cache again.
+  for (size_t i = 0; i < kNumBlocks - 1; i++) {
+    iterators[i].reset();
+    CheckCacheCounters(options, 0, 0, 0, 0);
+  }
+  ASSERT_EQ(0, cache->GetPinnedUsage());
+  for (size_t i = 0; i < kNumBlocks - 1; i++) {
+    iter = db_->NewIterator(read_options);
+    iter->Seek(ToString(i));
+    ASSERT_OK(iter->status());
+    CheckCacheCounters(options, 0, 1, 0, 0);
+    iterators[i].reset(iter);
+  }
+}
+
+#ifdef SNAPPY
+TEST_F(DBBlockCacheTest, TestWithCompressedBlockCache) {
+  ReadOptions read_options;
+  auto table_options = GetTableOptions();
+  auto options = GetOptions(table_options);
+  options.compression = CompressionType::kSnappyCompression;
+  InitTable(options);
+
+  std::shared_ptr<Cache> cache = NewLRUCache(0, 0, false);
+  std::shared_ptr<Cache> compressed_cache = NewLRUCache(0, 0, false);
+  table_options.block_cache = cache;
+  table_options.block_cache_compressed = compressed_cache;
+  options.table_factory.reset(new BlockBasedTableFactory(table_options));
+  Reopen(options);
+  RecordCacheCounters(options);
+
+  std::vector<std::unique_ptr<Iterator>> iterators(kNumBlocks - 1);
+  Iterator* iter = nullptr;
+
+  // Load blocks into cache.
+  for (size_t i = 0; i < kNumBlocks - 1; i++) {
+    iter = db_->NewIterator(read_options);
+    iter->Seek(ToString(i));
+    ASSERT_OK(iter->status());
+    CheckCacheCounters(options, 1, 0, 1, 0);
+    CheckCompressedCacheCounters(options, 1, 0, 1, 0);
+    iterators[i].reset(iter);
+  }
+  size_t usage = cache->GetUsage();
+  ASSERT_LT(0, usage);
+  ASSERT_EQ(usage, cache->GetPinnedUsage());
+  size_t compressed_usage = compressed_cache->GetUsage();
+  ASSERT_LT(0, compressed_usage);
+  // Compressed block cache cannot be pinned.
+  ASSERT_EQ(0, compressed_cache->GetPinnedUsage());
+
+  // Set strict capacity limit flag. Now block will only load into compressed
+  // block cache.
+  cache->SetCapacity(usage);
+  cache->SetStrictCapacityLimit(true);
+  ASSERT_EQ(usage, cache->GetPinnedUsage());
+  // compressed_cache->SetCapacity(compressed_usage);
+  compressed_cache->SetCapacity(0);
+  // compressed_cache->SetStrictCapacityLimit(true);
+  iter = db_->NewIterator(read_options);
+  iter->Seek(ToString(kNumBlocks - 1));
+  ASSERT_TRUE(iter->status().IsIncomplete());
+  CheckCacheCounters(options, 1, 0, 0, 1);
+  CheckCompressedCacheCounters(options, 1, 0, 1, 0);
+  delete iter;
+  iter = nullptr;
+
+  // Clear strict capacity limit flag. This time we shall hit compressed block
+  // cache.
+  cache->SetStrictCapacityLimit(false);
+  iter = db_->NewIterator(read_options);
+  iter->Seek(ToString(kNumBlocks - 1));
+  ASSERT_OK(iter->status());
+  CheckCacheCounters(options, 1, 0, 1, 0);
+  CheckCompressedCacheCounters(options, 0, 1, 0, 0);
+  delete iter;
+  iter = nullptr;
+}
+#endif  // SNAPPY
+
+#ifndef ROCKSDB_LITE
+
+// Make sure that when options.block_cache is set, after a new table is
+// created its index/filter blocks are added to block cache.
+TEST_F(DBBlockCacheTest, IndexAndFilterBlocksOfNewTableAddedToCache) {
+  Options options = CurrentOptions();
+  options.create_if_missing = true;
+  options.statistics = rocksdb::CreateDBStatistics();
+  BlockBasedTableOptions table_options;
+  table_options.cache_index_and_filter_blocks = true;
+  table_options.filter_policy.reset(NewBloomFilterPolicy(20));
+  options.table_factory.reset(new BlockBasedTableFactory(table_options));
+  CreateAndReopenWithCF({"pikachu"}, options);
+
+  ASSERT_OK(Put(1, "key", "val"));
+  // Create a new table.
+  ASSERT_OK(Flush(1));
+
+  // index/filter blocks added to block cache right after table creation.
+  ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_INDEX_MISS));
+  ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
+  ASSERT_EQ(2, /* only index/filter were added */
+            TestGetTickerCount(options, BLOCK_CACHE_ADD));
+  ASSERT_EQ(0, TestGetTickerCount(options, BLOCK_CACHE_DATA_MISS));
+  uint64_t int_num;
+  ASSERT_TRUE(
+      dbfull()->GetIntProperty("rocksdb.estimate-table-readers-mem", &int_num));
+  ASSERT_EQ(int_num, 0U);
+
+  // Make sure filter block is in cache.
+  std::string value;
+  ReadOptions ropt;
+  db_->KeyMayExist(ReadOptions(), handles_[1], "key", &value);
+
+  // Miss count should remain the same.
+  ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
+  ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT));
+
+  db_->KeyMayExist(ReadOptions(), handles_[1], "key", &value);
+  ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
+  ASSERT_EQ(2, TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT));
+
+  // Make sure index block is in cache.
+  auto index_block_hit = TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT);
+  value = Get(1, "key");
+  ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
+  ASSERT_EQ(index_block_hit + 1,
+            TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT));
+
+  value = Get(1, "key");
+  ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
+  ASSERT_EQ(index_block_hit + 2,
+            TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT));
+}
+
+TEST_F(DBBlockCacheTest, ParanoidFileChecks) {
+  Options options = CurrentOptions();
+  options.create_if_missing = true;
+  options.statistics = rocksdb::CreateDBStatistics();
+  options.level0_file_num_compaction_trigger = 2;
+  options.paranoid_file_checks = true;
+  BlockBasedTableOptions table_options;
+  table_options.cache_index_and_filter_blocks = false;
+  table_options.filter_policy.reset(NewBloomFilterPolicy(20));
+  options.table_factory.reset(new BlockBasedTableFactory(table_options));
+  CreateAndReopenWithCF({"pikachu"}, options);
+
+  ASSERT_OK(Put(1, "1_key", "val"));
+  ASSERT_OK(Put(1, "9_key", "val"));
+  // Create a new table.
+  ASSERT_OK(Flush(1));
+  ASSERT_EQ(1, /* read and cache data block */
+            TestGetTickerCount(options, BLOCK_CACHE_ADD));
+
+  ASSERT_OK(Put(1, "1_key2", "val2"));
+  ASSERT_OK(Put(1, "9_key2", "val2"));
+  // Create a new SST file. This will further trigger a compaction
+  // and generate another file.
+  ASSERT_OK(Flush(1));
+  dbfull()->TEST_WaitForCompact();
+  ASSERT_EQ(3, /* Totally 3 files created up to now */
+            TestGetTickerCount(options, BLOCK_CACHE_ADD));
+
+  // After disabling options.paranoid_file_checks. NO further block
+  // is added after generating a new file.
+  ASSERT_OK(
+      dbfull()->SetOptions(handles_[1], {{"paranoid_file_checks", "false"}}));
+
+  ASSERT_OK(Put(1, "1_key3", "val3"));
+  ASSERT_OK(Put(1, "9_key3", "val3"));
+  ASSERT_OK(Flush(1));
+  ASSERT_OK(Put(1, "1_key4", "val4"));
+  ASSERT_OK(Put(1, "9_key4", "val4"));
+  ASSERT_OK(Flush(1));
+  dbfull()->TEST_WaitForCompact();
+  ASSERT_EQ(3, /* Totally 3 files created up to now */
+            TestGetTickerCount(options, BLOCK_CACHE_ADD));
+}
+
+TEST_F(DBBlockCacheTest, CompressedCache) {
+  if (!Snappy_Supported()) {
+    return;
+  }
+  int num_iter = 80;
+
+  // Run this test three iterations.
+  // Iteration 1: only a uncompressed block cache
+  // Iteration 2: only a compressed block cache
+  // Iteration 3: both block cache and compressed cache
+  // Iteration 4: both block cache and compressed cache, but DB is not
+  // compressed
+  for (int iter = 0; iter < 4; iter++) {
+    Options options = CurrentOptions();
+    options.write_buffer_size = 64 * 1024;  // small write buffer
+    options.statistics = rocksdb::CreateDBStatistics();
+
+    BlockBasedTableOptions table_options;
+    switch (iter) {
+      case 0:
+        // only uncompressed block cache
+        table_options.block_cache = NewLRUCache(8 * 1024);
+        table_options.block_cache_compressed = nullptr;
+        options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+        break;
+      case 1:
+        // no block cache, only compressed cache
+        table_options.no_block_cache = true;
+        table_options.block_cache = nullptr;
+        table_options.block_cache_compressed = NewLRUCache(8 * 1024);
+        options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+        break;
+      case 2:
+        // both compressed and uncompressed block cache
+        table_options.block_cache = NewLRUCache(1024);
+        table_options.block_cache_compressed = NewLRUCache(8 * 1024);
+        options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+        break;
+      case 3:
+        // both block cache and compressed cache, but DB is not compressed
+        // also, make block cache sizes bigger, to trigger block cache hits
+        table_options.block_cache = NewLRUCache(1024 * 1024);
+        table_options.block_cache_compressed = NewLRUCache(8 * 1024 * 1024);
+        options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+        options.compression = kNoCompression;
+        break;
+      default:
+        ASSERT_TRUE(false);
+    }
+    CreateAndReopenWithCF({"pikachu"}, options);
+    // default column family doesn't have block cache
+    Options no_block_cache_opts;
+    no_block_cache_opts.statistics = options.statistics;
+    no_block_cache_opts = CurrentOptions(no_block_cache_opts);
+    BlockBasedTableOptions table_options_no_bc;
+    table_options_no_bc.no_block_cache = true;
+    no_block_cache_opts.table_factory.reset(
+        NewBlockBasedTableFactory(table_options_no_bc));
+    ReopenWithColumnFamilies(
+        {"default", "pikachu"},
+        std::vector<Options>({no_block_cache_opts, options}));
+
+    Random rnd(301);
+
+    // Write 8MB (80 values, each 100K)
+    ASSERT_EQ(NumTableFilesAtLevel(0, 1), 0);
+    std::vector<std::string> values;
+    std::string str;
+    for (int i = 0; i < num_iter; i++) {
+      if (i % 4 == 0) {  // high compression ratio
+        str = RandomString(&rnd, 1000);
+      }
+      values.push_back(str);
+      ASSERT_OK(Put(1, Key(i), values[i]));
+    }
+
+    // flush all data from memtable so that reads are from block cache
+    ASSERT_OK(Flush(1));
+
+    for (int i = 0; i < num_iter; i++) {
+      ASSERT_EQ(Get(1, Key(i)), values[i]);
+    }
+
+    // check that we triggered the appropriate code paths in the cache
+    switch (iter) {
+      case 0:
+        // only uncompressed block cache
+        ASSERT_GT(TestGetTickerCount(options, BLOCK_CACHE_MISS), 0);
+        ASSERT_EQ(TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_MISS), 0);
+        break;
+      case 1:
+        // no block cache, only compressed cache
+        ASSERT_EQ(TestGetTickerCount(options, BLOCK_CACHE_MISS), 0);
+        ASSERT_GT(TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_MISS), 0);
+        break;
+      case 2:
+        // both compressed and uncompressed block cache
+        ASSERT_GT(TestGetTickerCount(options, BLOCK_CACHE_MISS), 0);
+        ASSERT_GT(TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_MISS), 0);
+        break;
+      case 3:
+        // both compressed and uncompressed block cache
+        ASSERT_GT(TestGetTickerCount(options, BLOCK_CACHE_MISS), 0);
+        ASSERT_GT(TestGetTickerCount(options, BLOCK_CACHE_HIT), 0);
+        ASSERT_GT(TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_MISS), 0);
+        // compressed doesn't have any hits since blocks are not compressed on
+        // storage
+        ASSERT_EQ(TestGetTickerCount(options, BLOCK_CACHE_COMPRESSED_HIT), 0);
+        break;
+      default:
+        ASSERT_TRUE(false);
+    }
+
+    options.create_if_missing = true;
+    DestroyAndReopen(options);
+  }
+}
+
+#endif  // ROCKSDB_LITE
+
+}  // namespace rocksdb
+
+int main(int argc, char** argv) {
+  rocksdb::port::InstallStackTraceHandler();
+  ::testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -368,11 +368,10 @@ TEST_F(DBTestCompactionFilter, CompactionFilter) {
 // entries are deleted. The compaction should create bunch of 'DeleteFile'
 // entries in VersionEdit, but none of the 'AddFile's.
 TEST_F(DBTestCompactionFilter, CompactionFilterDeletesAll) {
-  Options options;
+  Options options = CurrentOptions();
  options.compaction_filter_factory = std::make_shared<DeleteFilterFactory>();
  options.disable_auto_compactions = true;
  options.create_if_missing = true;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);

  // put some data
@@ -400,11 +399,10 @@ TEST_F(DBTestCompactionFilter, CompactionFilterDeletesAll) {

 TEST_F(DBTestCompactionFilter, CompactionFilterWithValueChange) {
  do {
-    Options options;
+    Options options = CurrentOptions();
    options.num_levels = 3;
    options.compaction_filter_factory =
      std::make_shared<ChangeFilterFactory>();
-    options = CurrentOptions(options);
    CreateAndReopenWithCF({"pikachu"}, options);

    // Write 100K+1 keys, these are written to a few files
@@ -467,8 +465,7 @@ TEST_F(DBTestCompactionFilter, CompactionFilterWithMergeOperator) {
  PutFixed64(&three, 3);
  PutFixed64(&four, 4);

-  Options options;
-  options = CurrentOptions(options);
+  Options options = CurrentOptions();
  options.create_if_missing = true;
  options.merge_operator = MergeOperators::CreateUInt64AddOperator();
  options.num_levels = 3;
@@ -621,11 +618,10 @@ TEST_F(DBTestCompactionFilter, CompactionFilterContextCfId) {
 // Compaction filters should only be applied to records that are newer than the
 // latest snapshot. This test inserts records and applies a delete filter.
 TEST_F(DBTestCompactionFilter, CompactionFilterSnapshot) {
-  Options options;
+  Options options = CurrentOptions();
  options.compaction_filter_factory = std::make_shared<DeleteFilterFactory>();
  options.disable_auto_compactions = true;
  options.create_if_missing = true;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);

  // Put some data.
@@ -659,11 +655,10 @@ TEST_F(DBTestCompactionFilter, CompactionFilterSnapshot) {
 // records newer than the snapshot will also be processed
 TEST_F(DBTestCompactionFilter, CompactionFilterIgnoreSnapshot) {
  std::string five = ToString(5);
-  Options options;
+  Options options = CurrentOptions();
  options.compaction_filter_factory = std::make_shared<DeleteISFilterFactory>();
  options.disable_auto_compactions = true;
  options.create_if_missing = true;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);

  // Put some data.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -40,34 +40,6 @@ class DBCompactionTestWithParam
 };

 namespace {
-class OnFileDeletionListener : public EventListener {
- public:
-  OnFileDeletionListener() :
-      matched_count_(0),
-      expected_file_name_("") {}
-
-  void SetExpectedFileName(
-      const std::string file_name) {
-    expected_file_name_ = file_name;
-  }
-
-  void VerifyMatchedCount(size_t expected_value) {
-    ASSERT_EQ(matched_count_, expected_value);
-  }
-
-  void OnTableFileDeleted(
-      const TableFileDeletionInfo& info) override {
-    if (expected_file_name_ != "") {
-      ASSERT_EQ(expected_file_name_, info.file_path);
-      expected_file_name_ = "";
-      matched_count_++;
-    }
-  }
-
- private:
-  size_t matched_count_;
-  std::string expected_file_name_;
-};

 class FlushedFileCollector : public EventListener {
 public:
@@ -98,8 +70,7 @@ class FlushedFileCollector : public EventListener {
 static const int kCDTValueSize = 1000;
 static const int kCDTKeysPerBuffer = 4;
 static const int kCDTNumLevels = 8;
-Options DeletionTriggerOptions() {
-  Options options;
+Options DeletionTriggerOptions(Options options) {
  options.compression = kNoCompression;
  options.write_buffer_size = kCDTKeysPerBuffer * (kCDTValueSize + 24);
  options.min_write_buffer_number_to_merge = 1;
@@ -202,7 +173,7 @@ const SstFileMetaData* PickFileRandomly(
 TEST_P(DBCompactionTestWithParam, CompactionDeletionTrigger) {
  for (int tid = 0; tid < 3; ++tid) {
    uint64_t db_size[2];
-    Options options = CurrentOptions(DeletionTriggerOptions());
+    Options options = DeletionTriggerOptions(CurrentOptions());
    options.max_subcompactions = max_subcompactions_;

    if (tid == 1) {
@@ -245,8 +216,7 @@ TEST_F(DBCompactionTest, SkipStatsUpdateTest) {
  // the compaction behavior when there are many of deletion entries.
  // The test will need to be updated if the internal behavior changes.

-  Options options = DeletionTriggerOptions();
-  options = CurrentOptions(options);
+  Options options = DeletionTriggerOptions(CurrentOptions());
  options.env = env_;
  DestroyAndReopen(options);
  Random rnd(301);
@@ -288,8 +258,7 @@ TEST_F(DBCompactionTest, SkipStatsUpdateTest) {
 }

 TEST_F(DBCompactionTest, TestTableReaderForCompaction) {
-  Options options;
-  options = CurrentOptions(options);
+  Options options = CurrentOptions();
  options.env = env_;
  options.new_table_reader_for_compaction_inputs = true;
  options.max_open_files = 100;
@@ -378,7 +347,7 @@ TEST_F(DBCompactionTest, TestTableReaderForCompaction) {
 TEST_P(DBCompactionTestWithParam, CompactionDeletionTriggerReopen) {
  for (int tid = 0; tid < 2; ++tid) {
    uint64_t db_size[3];
-    Options options = CurrentOptions(DeletionTriggerOptions());
+    Options options = DeletionTriggerOptions(CurrentOptions());
    options.max_subcompactions = max_subcompactions_;

    if (tid == 1) {
@@ -435,7 +404,7 @@ TEST_P(DBCompactionTestWithParam, CompactionDeletionTriggerReopen) {
 TEST_F(DBCompactionTest, DisableStatsUpdateReopen) {
  uint64_t db_size[3];
  for (int test = 0; test < 2; ++test) {
-    Options options = CurrentOptions(DeletionTriggerOptions());
+    Options options = DeletionTriggerOptions(CurrentOptions());
    options.skip_stats_update_on_db_open = (test == 0);

    env_->random_read_counter_.Reset();
@@ -493,14 +462,13 @@ TEST_F(DBCompactionTest, DisableStatsUpdateReopen) {
 TEST_P(DBCompactionTestWithParam, CompactionTrigger) {
  const int kNumKeysPerFile = 100;

-  Options options;
+  Options options = CurrentOptions();
  options.write_buffer_size = 110 << 10;  // 110KB
  options.arena_block_size = 4 << 10;
  options.num_levels = 3;
  options.level0_file_num_compaction_trigger = 3;
  options.max_subcompactions = max_subcompactions_;
  options.memtable_factory.reset(new SpecialSkipListFactory(kNumKeysPerFile));
-  options = CurrentOptions(options);
  CreateAndReopenWithCF({"pikachu"}, options);

  Random rnd(301);
@@ -533,11 +501,107 @@ TEST_P(DBCompactionTestWithParam, CompactionTrigger) {
  ASSERT_EQ(NumTableFilesAtLevel(1, 1), 1);
 }

+TEST_F(DBCompactionTest, BGCompactionsAllowed) {
+  // Create several column families. Make compaction triggers in all of them
+  // and see number of compactions scheduled to be less than allowed.
+  const int kNumKeysPerFile = 100;
+
+  Options options = CurrentOptions();
+  options.write_buffer_size = 110 << 10;  // 110KB
+  options.arena_block_size = 4 << 10;
+  options.num_levels = 3;
+  // Should speed up compaction when there are 4 files.
+  options.level0_file_num_compaction_trigger = 2;
+  options.level0_slowdown_writes_trigger = 20;
+  options.soft_pending_compaction_bytes_limit = 1 << 30;  // Infinitely large
+  options.base_background_compactions = 1;
+  options.max_background_compactions = 3;
+  options.memtable_factory.reset(new SpecialSkipListFactory(kNumKeysPerFile));
+
+  // Block all threads in thread pool.
+  const size_t kTotalTasks = 4;
+  env_->SetBackgroundThreads(4, Env::LOW);
+  test::SleepingBackgroundTask sleeping_tasks[kTotalTasks];
+  for (size_t i = 0; i < kTotalTasks; i++) {
+    env_->Schedule(&test::SleepingBackgroundTask::DoSleepTask,
+                   &sleeping_tasks[i], Env::Priority::LOW);
+    sleeping_tasks[i].WaitUntilSleeping();
+  }
+
+  CreateAndReopenWithCF({"one", "two", "three"}, options);
+
+  Random rnd(301);
+  for (int cf = 0; cf < 4; cf++) {
+    for (int num = 0; num < options.level0_file_num_compaction_trigger; num++) {
+      for (int i = 0; i < kNumKeysPerFile; i++) {
+        ASSERT_OK(Put(cf, Key(i), ""));
+      }
+      // put extra key to trigger flush
+      ASSERT_OK(Put(cf, "", ""));
+      dbfull()->TEST_WaitForFlushMemTable(handles_[cf]);
+      ASSERT_EQ(NumTableFilesAtLevel(0, cf), num + 1);
+    }
+  }
+
+  // Now all column families qualify compaction but only one should be
+  // scheduled, because no column family hits speed up condition.
+  ASSERT_EQ(1, env_->GetThreadPoolQueueLen(Env::Priority::LOW));
+
+  // Create two more files for one column family, which triggers speed up
+  // condition, three compactions will be scheduled.
+  for (int num = 0; num < options.level0_file_num_compaction_trigger; num++) {
+    for (int i = 0; i < kNumKeysPerFile; i++) {
+      ASSERT_OK(Put(2, Key(i), ""));
+    }
+    // put extra key to trigger flush
+    ASSERT_OK(Put(2, "", ""));
+    dbfull()->TEST_WaitForFlushMemTable(handles_[2]);
+    ASSERT_EQ(options.level0_file_num_compaction_trigger + num + 1,
+              NumTableFilesAtLevel(0, 2));
+  }
+  ASSERT_EQ(3, env_->GetThreadPoolQueueLen(Env::Priority::LOW));
+
+  // Unblock all threads to unblock all compactions.
+  for (size_t i = 0; i < kTotalTasks; i++) {
+    sleeping_tasks[i].WakeUp();
+    sleeping_tasks[i].WaitUntilDone();
+  }
+  dbfull()->TEST_WaitForCompact();
+
+  // Verify number of compactions allowed will come back to 1.
+
+  for (size_t i = 0; i < kTotalTasks; i++) {
+    sleeping_tasks[i].Reset();
+    env_->Schedule(&test::SleepingBackgroundTask::DoSleepTask,
+                   &sleeping_tasks[i], Env::Priority::LOW);
+    sleeping_tasks[i].WaitUntilSleeping();
+  }
+  for (int cf = 0; cf < 4; cf++) {
+    for (int num = 0; num < options.level0_file_num_compaction_trigger; num++) {
+      for (int i = 0; i < kNumKeysPerFile; i++) {
+        ASSERT_OK(Put(cf, Key(i), ""));
+      }
+      // put extra key to trigger flush
+      ASSERT_OK(Put(cf, "", ""));
+      dbfull()->TEST_WaitForFlushMemTable(handles_[cf]);
+      ASSERT_EQ(NumTableFilesAtLevel(0, cf), num + 1);
+    }
+  }
+
+  // Now all column families qualify compaction but only one should be
+  // scheduled, because no column family hits speed up condition.
+  ASSERT_EQ(1, env_->GetThreadPoolQueueLen(Env::Priority::LOW));
+
+  for (size_t i = 0; i < kTotalTasks; i++) {
+    sleeping_tasks[i].WakeUp();
+    sleeping_tasks[i].WaitUntilDone();
+  }
+}
+
 TEST_P(DBCompactionTestWithParam, CompactionsGenerateMultipleFiles) {
-  Options options;
+  Options options = CurrentOptions();
  options.write_buffer_size = 100000000;        // Large write buffer
  options.max_subcompactions = max_subcompactions_;
-  options = CurrentOptions(options);
  CreateAndReopenWithCF({"pikachu"}, options);

  Random rnd(301);
@@ -564,9 +628,8 @@ TEST_P(DBCompactionTestWithParam, CompactionsGenerateMultipleFiles) {

 TEST_F(DBCompactionTest, MinorCompactionsHappen) {
  do {
-    Options options;
+    Options options = CurrentOptions();
    options.write_buffer_size = 10000;
-    options = CurrentOptions(options);
    CreateAndReopenWithCF({"pikachu"}, options);

    const int N = 500;
@@ -591,7 +654,7 @@ TEST_F(DBCompactionTest, MinorCompactionsHappen) {
 }

 TEST_F(DBCompactionTest, ZeroSeqIdCompaction) {
-  Options options;
+  Options options = CurrentOptions();
  options.compaction_style = kCompactionStyleLevel;
  options.level0_file_num_compaction_trigger = 3;

@@ -605,7 +668,6 @@ TEST_F(DBCompactionTest, ZeroSeqIdCompaction) {
  const size_t key_len =
    static_cast<size_t>(compact_opt.output_file_size_limit) / 5;

-  options = CurrentOptions(options);
  DestroyAndReopen(options);

  std::vector<const Snapshot*> snaps;
@@ -649,10 +711,8 @@ TEST_F(DBCompactionTest, ZeroSeqIdCompaction) {
 // if the database is shutdown during the memtable compaction.
 TEST_F(DBCompactionTest, RecoverDuringMemtableCompaction) {
  do {
-    Options options;
+    Options options = CurrentOptions();
    options.env = env_;
-    options.write_buffer_size = 1000000;
-    options = CurrentOptions(options);
    CreateAndReopenWithCF({"pikachu"}, options);

    // Trigger a long memtable compaction and reopen the database during it
@@ -676,10 +736,9 @@ TEST_P(DBCompactionTestWithParam, TrivialMoveOneFile) {
      [&](void* arg) { trivial_move++; });
  rocksdb::SyncPoint::GetInstance()->EnableProcessing();

-  Options options;
+  Options options = CurrentOptions();
  options.write_buffer_size = 100000000;
  options.max_subcompactions = max_subcompactions_;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);

  int32_t num_keys = 80;
@@ -897,7 +956,9 @@ TEST_P(DBCompactionTestWithParam, ManualCompactionPartial) {
  bool second = true;
  rocksdb::SyncPoint::GetInstance()->LoadDependency(
      {{"DBCompaction::ManualPartial:4", "DBCompaction::ManualPartial:1"},
-       {"DBCompaction::ManualPartial:2", "DBCompaction::ManualPartial:3"}});
+       {"DBCompaction::ManualPartial:2", "DBCompaction::ManualPartial:3"},
+       {"DBCompaction::ManualPartial:5",
+        "DBImpl::BackgroundCompaction:NonTrivial:AfterRun"}});
  rocksdb::SyncPoint::GetInstance()->SetCallBack(
      "DBImpl::BackgroundCompaction:NonTrivial:AfterRun", [&](void* arg) {
        if (first) {
@@ -1008,6 +1069,8 @@ TEST_P(DBCompactionTestWithParam, ManualCompactionPartial) {

  // 3 files in L0
  ASSERT_EQ("3,0,0,0,0,1,2", FilesPerLevel(0));
+  TEST_SYNC_POINT("DBCompaction::ManualPartial:5");
+
  // 1 file in L6, 1 file in L1
  dbfull()->TEST_WaitForFlushMemTable();
  dbfull()->TEST_WaitForCompact();
@@ -1271,10 +1334,9 @@ TEST_P(DBCompactionTestWithParam, TrivialMoveToLastLevelWithFiles) {
      [&](void* arg) { non_trivial_move++; });
  rocksdb::SyncPoint::GetInstance()->EnableProcessing();

-  Options options;
+  Options options = CurrentOptions();
  options.write_buffer_size = 100000000;
  options.max_subcompactions = max_subcompactions_;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);

  int32_t value_size = 10 * 1024;  // 10 KB
@@ -1563,7 +1625,7 @@ TEST_P(DBCompactionTestWithParam, ConvertCompactionStyle) {
  int max_key_universal_insert = 600;

  // Stage 1: generate a db with level compaction
-  Options options;
+  Options options = CurrentOptions();
  options.write_buffer_size = 110 << 10;  // 110KB
  options.arena_block_size = 4 << 10;
  options.num_levels = 4;
@@ -1573,7 +1635,6 @@ TEST_P(DBCompactionTestWithParam, ConvertCompactionStyle) {
  options.target_file_size_base = 200 << 10;  // 200KB
  options.target_file_size_multiplier = 1;
  options.max_subcompactions = max_subcompactions_;
-  options = CurrentOptions(options);
  CreateAndReopenWithCF({"pikachu"}, options);

  for (int i = 0; i <= max_key_level_insert; i++) {
@@ -1898,7 +1959,7 @@ TEST_P(DBCompactionTestWithParam, DISABLED_CompactFilesOnLevelCompaction) {
    std::set<std::string> overlapping_file_names;
    std::vector<std::string> compaction_input_file_names;
    for (int f = 0; f < file_picked; ++f) {
-      int level;
+      int level = 0;
      auto file_meta = PickFileRandomly(cf_meta, &rnd, &level);
      compaction_input_file_names.push_back(file_meta->name);
      GetOverlappingFileNumbersForLevelCompaction(
@@ -2198,6 +2259,25 @@ TEST_P(DBCompactionTestWithParam, CompressLevelCompaction) {
  Destroy(options);
 }

+TEST_F(DBCompactionTest, SanitizeCompactionOptionsTest) {
+  Options options = CurrentOptions();
+  options.max_background_compactions = 5;
+  options.soft_pending_compaction_bytes_limit = 0;
+  options.hard_pending_compaction_bytes_limit = 100;
+  options.create_if_missing = true;
+  DestroyAndReopen(options);
+  ASSERT_EQ(5, db_->GetOptions().base_background_compactions);
+  ASSERT_EQ(100, db_->GetOptions().soft_pending_compaction_bytes_limit);
+
+  options.base_background_compactions = 4;
+  options.max_background_compactions = 3;
+  options.soft_pending_compaction_bytes_limit = 200;
+  options.hard_pending_compaction_bytes_limit = 150;
+  DestroyAndReopen(options);
+  ASSERT_EQ(3, db_->GetOptions().base_background_compactions);
+  ASSERT_EQ(150, db_->GetOptions().soft_pending_compaction_bytes_limit);
+}
+
 // This tests for a bug that could cause two level0 compactions running
 // concurrently
 // TODO(aekmekji): Make sure that the reason this fails when run with
@@ -2269,10 +2349,9 @@ TEST_P(DBCompactionTestWithParam, ForceBottommostLevelCompaction) {
      [&](void* arg) { non_trivial_move++; });
  rocksdb::SyncPoint::GetInstance()->EnableProcessing();

-  Options options;
+  Options options = CurrentOptions();
  options.write_buffer_size = 100000000;
  options.max_subcompactions = max_subcompactions_;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);

  int32_t value_size = 10 * 1024;  // 10 KB
@@ -2362,14 +2441,14 @@ class CompactionPriTest : public DBTestBase,
 };

 TEST_P(CompactionPriTest, Test) {
-  Options options;
+  Options options = CurrentOptions();
  options.write_buffer_size = 16 * 1024;
  options.compaction_pri = static_cast<CompactionPri>(compaction_pri_);
  options.hard_pending_compaction_bytes_limit = 256 * 1024;
  options.max_bytes_for_level_base = 64 * 1024;
  options.max_bytes_for_level_multiplier = 4;
  options.compression = kNoCompression;
-  options = CurrentOptions(options);
+
  DestroyAndReopen(options);

  Random rnd(301);
@@ -2390,8 +2469,12 @@ TEST_P(CompactionPriTest, Test) {
  }
 }

-INSTANTIATE_TEST_CASE_P(CompactionPriTest, CompactionPriTest,
-                        ::testing::Values(0, 1, 2));
+INSTANTIATE_TEST_CASE_P(
+    CompactionPriTest, CompactionPriTest,
+    ::testing::Values(CompactionPri::kByCompensatedSize,
+                      CompactionPri::kOldestLargestSeqFirst,
+                      CompactionPri::kOldestSmallestSeqFirst,
+                      CompactionPri::kMinOverlappingRatio));

 #endif // !defined(ROCKSDB_LITE)
 }  // namespace rocksdb
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -232,7 +232,7 @@ class DBImpl : public DB {
  // SST files will also be checked.
  //
  // If a key is found, *found_record_for_key will be set to true and
-  // *seq will will be set to the stored sequence number for the latest
+  // *seq will be set to the stored sequence number for the latest
  // operation on this key or kMaxSequenceNumber if unknown.
  // If no key is found, *found_record_for_key will be set to false.
  //
@@ -347,6 +347,10 @@ class DBImpl : public DB {

 #endif  // NDEBUG

+  // Return maximum background compaction alowed to be scheduled based on
+  // compaction status.
+  int BGCompactionsAllowed() const;
+
  // Returns the list of live files in 'live' and the list
  // of all files in the filesystem in 'candidate_files'.
  // If force == false and the last call was less than
@@ -479,7 +483,8 @@ class DBImpl : public DB {
  // amount of work to recover recently logged updates.  Any changes to
  // be made to the descriptor are added to *edit.
  Status Recover(const std::vector<ColumnFamilyDescriptor>& column_families,
-                 bool read_only = false, bool error_if_log_file_exist = false);
+                 bool read_only = false, bool error_if_log_file_exist = false,
+                 bool error_if_data_exists_in_logs = false);

  void MaybeIgnoreError(Status* s) const;

@@ -490,10 +495,11 @@ class DBImpl : public DB {

  // Background process needs to call
  //     auto x = CaptureCurrentFileNumberInPendingOutputs()
+  //     auto file_num = versions_->NewFileNumber();
  //     <do something>
  //     ReleaseFileNumberFromPendingOutputs(x)
-  // This will protect any temporary files created while <do something> is
-  // executing from being deleted.
+  // This will protect any file with number `file_num` or greater from being
+  // deleted while <do something> is running.
  // -----------
  // This function will capture current file number and append it to
  // pending_outputs_. This will prevent any background process to delete any
@@ -518,7 +524,7 @@ class DBImpl : public DB {
                         SequenceNumber* max_sequence, bool read_only);

  // The following two methods are used to flush a memtable to
-  // storage. The first one is used atdatabase RecoveryTime (when the
+  // storage. The first one is used at database RecoveryTime (when the
  // database is opened) and is heavyweight because it holds the mutex
  // for the entire period. The second method WriteLevel0Table supports
  // concurrent flush memtables to storage.
@@ -539,9 +545,6 @@ class DBImpl : public DB {
  // Wait for memtable flushed
  Status WaitForFlushMemTable(ColumnFamilyData* cfd);

-  void RecordFlushIOStats();
-  void RecordCompactionIOStats();
-
 #ifndef ROCKSDB_LITE
  Status CompactFilesImpl(
      const CompactionOptions& compact_options, ColumnFamilyData* cfd,
@@ -818,7 +821,10 @@ class DBImpl : public DB {
  // they're unique
  std::atomic<int> next_job_id_;

-  bool flush_on_destroy_; // Used when disableWAL is true.
+  // A flag indicating whether the current rocksdb database has any
+  // data that is not yet persisted into either WAL or SST file.
+  // Used when disableWAL is true.
+  bool has_unpersisted_data_;

  static const int KEEP_LOG_FILE_NUM = 1000;
  // MSVC version 1800 still does not have constexpr for ::max()
@@ -836,9 +842,12 @@ class DBImpl : public DB {
  // Unified interface for logging events
  EventLogger event_logger_;

-  // A value of >0 temporarily disables scheduling of background work
+  // A value of > 0 temporarily disables scheduling of background work
  int bg_work_paused_;

+  // A value of > 0 temporarily disables scheduling of background compaction
+  int bg_compaction_paused_;
+
  // Guard against multiple concurrent refitting
  bool refitting_level_;

@@ -889,9 +898,8 @@ class DBImpl : public DB {
                 bool* value_found = nullptr);

  bool GetIntPropertyInternal(ColumnFamilyData* cfd,
-                              DBPropertyType property_type,
-                              bool need_out_of_mutex, bool is_locked,
-                              uint64_t* value);
+                              const DBPropertyInfo& property_info,
+                              bool is_locked, uint64_t* value);

  bool HasPendingManualCompaction();
  bool HasExclusiveManualCompaction();
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -54,10 +54,11 @@ Iterator* DBImplReadOnly::NewIterator(const ReadOptions& read_options,
  auto db_iter = NewArenaWrappedDbIterator(
      env_, *cfd->ioptions(), cfd->user_comparator(),
      (read_options.snapshot != nullptr
-           ? reinterpret_cast<const SnapshotImpl*>(
-                read_options.snapshot)->number_
+           ? reinterpret_cast<const SnapshotImpl*>(read_options.snapshot)
+                 ->number_
           : latest_snapshot),
-      super_version->mutable_cf_options.max_sequential_skip_in_iterations);
+      super_version->mutable_cf_options.max_sequential_skip_in_iterations,
+      super_version->version_number);
  auto internal_iter = NewInternalIterator(
      read_options, cfd, super_version, db_iter->GetArena());
  db_iter->SetIterUnderDBIter(internal_iter);
@@ -81,10 +82,11 @@ Status DBImplReadOnly::NewIterators(
    auto* db_iter = NewArenaWrappedDbIterator(
        env_, *cfd->ioptions(), cfd->user_comparator(),
        (read_options.snapshot != nullptr
-            ? reinterpret_cast<const SnapshotImpl*>(
-                  read_options.snapshot)->number_
-            : latest_snapshot),
-        sv->mutable_cf_options.max_sequential_skip_in_iterations);
+             ? reinterpret_cast<const SnapshotImpl*>(read_options.snapshot)
+                   ->number_
+             : latest_snapshot),
+        sv->mutable_cf_options.max_sequential_skip_in_iterations,
+        sv->version_number);
    auto* internal_iter = NewInternalIterator(
        read_options, cfd, sv, db_iter->GetArena());
    db_iter->SetIterUnderDBIter(internal_iter);
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -7,6 +7,8 @@
 #define __STDC_FORMAT_MACROS
 #endif

+#include "db/db_info_dumper.h"
+
 #include <inttypes.h>
 #include <stdio.h>
 #include <string>
@@ -16,7 +18,6 @@
 #include "db/filename.h"
 #include "rocksdb/options.h"
 #include "rocksdb/env.h"
-#include "util/db_info_dumper.h"

 namespace rocksdb {

@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -18,12 +18,11 @@ class DBTestInPlaceUpdate : public DBTestBase {

 TEST_F(DBTestInPlaceUpdate, InPlaceUpdate) {
  do {
-    Options options;
+    Options options = CurrentOptions();
    options.create_if_missing = true;
    options.inplace_update_support = true;
    options.env = env_;
    options.write_buffer_size = 100000;
-    options = CurrentOptions(options);
    CreateAndReopenWithCF({"pikachu"}, options);

    // Update key with values of smaller size
@@ -41,12 +40,11 @@ TEST_F(DBTestInPlaceUpdate, InPlaceUpdate) {

 TEST_F(DBTestInPlaceUpdate, InPlaceUpdateLargeNewValue) {
  do {
-    Options options;
+    Options options = CurrentOptions();
    options.create_if_missing = true;
    options.inplace_update_support = true;
    options.env = env_;
    options.write_buffer_size = 100000;
-    options = CurrentOptions(options);
    CreateAndReopenWithCF({"pikachu"}, options);

    // Update key with values of larger size
@@ -64,7 +62,7 @@ TEST_F(DBTestInPlaceUpdate, InPlaceUpdateLargeNewValue) {

 TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackSmallerSize) {
  do {
-    Options options;
+    Options options = CurrentOptions();
    options.create_if_missing = true;
    options.inplace_update_support = true;

@@ -72,7 +70,6 @@ TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackSmallerSize) {
    options.write_buffer_size = 100000;
    options.inplace_callback =
      rocksdb::DBTestInPlaceUpdate::updateInPlaceSmallerSize;
-    options = CurrentOptions(options);
    CreateAndReopenWithCF({"pikachu"}, options);

    // Update key with values of smaller size
@@ -92,7 +89,7 @@ TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackSmallerSize) {

 TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackSmallerVarintSize) {
  do {
-    Options options;
+    Options options = CurrentOptions();
    options.create_if_missing = true;
    options.inplace_update_support = true;

@@ -100,7 +97,6 @@ TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackSmallerVarintSize) {
    options.write_buffer_size = 100000;
    options.inplace_callback =
      rocksdb::DBTestInPlaceUpdate::updateInPlaceSmallerVarintSize;
-    options = CurrentOptions(options);
    CreateAndReopenWithCF({"pikachu"}, options);

    // Update key with values of smaller varint size
@@ -120,7 +116,7 @@ TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackSmallerVarintSize) {

 TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackLargeNewValue) {
  do {
-    Options options;
+    Options options = CurrentOptions();
    options.create_if_missing = true;
    options.inplace_update_support = true;

@@ -128,7 +124,6 @@ TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackLargeNewValue) {
    options.write_buffer_size = 100000;
    options.inplace_callback =
      rocksdb::DBTestInPlaceUpdate::updateInPlaceLargerSize;
-    options = CurrentOptions(options);
    CreateAndReopenWithCF({"pikachu"}, options);

    // Update key with values of larger size
@@ -146,7 +141,7 @@ TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackLargeNewValue) {

 TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackNoAction) {
  do {
-    Options options;
+    Options options = CurrentOptions();
    options.create_if_missing = true;
    options.inplace_update_support = true;

@@ -154,7 +149,6 @@ TEST_F(DBTestInPlaceUpdate, InPlaceUpdateCallbackNoAction) {
    options.write_buffer_size = 100000;
    options.inplace_callback =
      rocksdb::DBTestInPlaceUpdate::updateInPlaceNoAction;
-    options = CurrentOptions(options);
    CreateAndReopenWithCF({"pikachu"}, options);

    // Callback function requests no actions from db
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -13,18 +13,21 @@
 #include <string>
 #include <limits>

-#include "db/filename.h"
 #include "db/dbformat.h"
+#include "db/filename.h"
+#include "db/merge_context.h"
+#include "db/pinned_iterators_manager.h"
 #include "port/port.h"
 #include "rocksdb/env.h"
-#include "rocksdb/options.h"
 #include "rocksdb/iterator.h"
 #include "rocksdb/merge_operator.h"
+#include "rocksdb/options.h"
 #include "table/internal_iterator.h"
 #include "util/arena.h"
 #include "util/logging.h"
 #include "util/mutexlock.h"
 #include "util/perf_context_imp.h"
+#include "util/string_util.h"

 namespace rocksdb {

@@ -59,11 +62,49 @@ class DBIter: public Iterator {
    kReverse
  };

+  // LocalStatistics contain Statistics counters that will be aggregated per
+  // each iterator instance and then will be sent to the global statistics when
+  // the iterator is destroyed.
+  //
+  // The purpose of this approach is to avoid perf regression happening
+  // when multiple threads bump the atomic counters from a DBIter::Next().
+  struct LocalStatistics {
+    explicit LocalStatistics() { ResetCounters(); }
+
+    void ResetCounters() {
+      next_count_ = 0;
+      next_found_count_ = 0;
+      prev_count_ = 0;
+      prev_found_count_ = 0;
+      bytes_read_ = 0;
+    }
+
+    void BumpGlobalStatistics(Statistics* global_statistics) {
+      RecordTick(global_statistics, NUMBER_DB_NEXT, next_count_);
+      RecordTick(global_statistics, NUMBER_DB_NEXT_FOUND, next_found_count_);
+      RecordTick(global_statistics, NUMBER_DB_PREV, prev_count_);
+      RecordTick(global_statistics, NUMBER_DB_PREV_FOUND, prev_found_count_);
+      RecordTick(global_statistics, ITER_BYTES_READ, bytes_read_);
+      ResetCounters();
+    }
+
+    // Map to Tickers::NUMBER_DB_NEXT
+    uint64_t next_count_;
+    // Map to Tickers::NUMBER_DB_NEXT_FOUND
+    uint64_t next_found_count_;
+    // Map to Tickers::NUMBER_DB_PREV
+    uint64_t prev_count_;
+    // Map to Tickers::NUMBER_DB_PREV_FOUND
+    uint64_t prev_found_count_;
+    // Map to Tickers::ITER_BYTES_READ
+    uint64_t bytes_read_;
+  };
+
  DBIter(Env* env, const ImmutableCFOptions& ioptions, const Comparator* cmp,
         InternalIterator* iter, SequenceNumber s, bool arena_mode,
-         uint64_t max_sequential_skip_in_iterations,
+         uint64_t max_sequential_skip_in_iterations, uint64_t version_number,
         const Slice* iterate_upper_bound = nullptr,
-         bool prefix_same_as_start = false)
+         bool prefix_same_as_start = false, bool pin_data = false)
      : arena_mode_(arena_mode),
        env_(env),
        logger_(ioptions.info_log),
@@ -75,15 +116,25 @@ class DBIter: public Iterator {
        valid_(false),
        current_entry_is_merged_(false),
        statistics_(ioptions.statistics),
+        version_number_(version_number),
        iterate_upper_bound_(iterate_upper_bound),
        prefix_same_as_start_(prefix_same_as_start),
-        iter_pinned_(false) {
+        pin_thru_lifetime_(pin_data) {
    RecordTick(statistics_, NO_ITERATORS);
    prefix_extractor_ = ioptions.prefix_extractor;
    max_skip_ = max_sequential_skip_in_iterations;
+    if (pin_thru_lifetime_) {
+      pinned_iters_mgr_.StartPinning();
+    }
+    if (iter_) {
+      iter_->SetPinnedItersMgr(&pinned_iters_mgr_);
+    }
  }
  virtual ~DBIter() {
+    // Release pinned data if any
+    pinned_iters_mgr_.ReleasePinnedIterators();
    RecordTick(statistics_, NO_ITERATORS, -1);
+    local_stats_.BumpGlobalStatistics(statistics_);
    if (!arena_mode_) {
      delete iter_;
    } else {
@@ -93,9 +144,7 @@ class DBIter: public Iterator {
  virtual void SetIter(InternalIterator* iter) {
    assert(iter_ == nullptr);
    iter_ = iter;
-    if (iter_ && iter_pinned_) {
-      iter_->PinData();
-    }
+    iter_->SetPinnedItersMgr(&pinned_iters_mgr_);
  }
  virtual bool Valid() const override { return valid_; }
  virtual Slice key() const override {
@@ -104,8 +153,13 @@ class DBIter: public Iterator {
  }
  virtual Slice value() const override {
    assert(valid_);
-    return (direction_ == kForward && !current_entry_is_merged_) ?
-      iter_->value() : saved_value_;
+    if (current_entry_is_merged_) {
+      return saved_value_;
+    } else if (direction_ == kReverse) {
+      return pinned_value_;
+    } else {
+      return iter_->value();
+    }
  }
  virtual Status status() const override {
    if (status_.ok()) {
@@ -114,31 +168,27 @@ class DBIter: public Iterator {
      return status_;
    }
  }
-  virtual Status PinData() {
-    Status s;
-    if (iter_) {
-      s = iter_->PinData();
+
+  virtual Status GetProperty(std::string prop_name,
+                             std::string* prop) override {
+    if (prop == nullptr) {
+      return Status::InvalidArgument("prop is nullptr");
    }
-    if (s.ok()) {
-      // Even if iter_ is nullptr, we set iter_pinned_ to true so that when
-      // iter_ is updated using SetIter, we Pin it.
-      iter_pinned_ = true;
+    if (prop_name == "rocksdb.iterator.super-version-number") {
+      // First try to pass the value returned from inner iterator.
+      if (!iter_->GetProperty(prop_name, prop).ok()) {
+        *prop = ToString(version_number_);
+      }
+      return Status::OK();
+    } else if (prop_name == "rocksdb.iterator.is-key-pinned") {
+      if (valid_) {
+        *prop = (pin_thru_lifetime_ && saved_key_.IsKeyPinned()) ? "1" : "0";
+      } else {
+        *prop = "Iterator is not valid.";
+      }
+      return Status::OK();
    }
-    return s;
-  }
-  virtual Status ReleasePinnedData() {
-    Status s;
-    if (iter_) {
-      s = iter_->ReleasePinnedData();
-    }
-    if (s.ok()) {
-      iter_pinned_ = false;
-    }
-    return s;
-  }
-  virtual bool IsKeyPinned() const override {
-    assert(valid_);
-    return iter_pinned_ && saved_key_.IsKeyPinned();
+    return Status::InvalidArgument("Undentified property.");
  }

  virtual void Next() override;
@@ -155,11 +205,26 @@ class DBIter: public Iterator {
  bool FindValueForCurrentKeyUsingSeek();
  void FindPrevUserKey();
  void FindNextUserKey();
-  inline void FindNextUserEntry(bool skipping);
-  void FindNextUserEntryInternal(bool skipping);
+  inline void FindNextUserEntry(bool skipping, bool prefix_check);
+  void FindNextUserEntryInternal(bool skipping, bool prefix_check);
  bool ParseKey(ParsedInternalKey* key);
  void MergeValuesNewToOld();

+  // Temporarily pin the blocks that we encounter until ReleaseTempPinnedData()
+  // is called
+  void TempPinData() {
+    if (!pin_thru_lifetime_) {
+      pinned_iters_mgr_.StartPinning();
+    }
+  }
+
+  // Release blocks pinned by TempPinData()
+  void ReleaseTempPinnedData() {
+    if (!pin_thru_lifetime_) {
+      pinned_iters_mgr_.ReleasePinnedIterators();
+    }
+  }
+
  inline void ClearSavedValue() {
    if (saved_value_.capacity() > 1048576) {
      std::string empty;
@@ -181,17 +246,24 @@ class DBIter: public Iterator {
  Status status_;
  IterKey saved_key_;
  std::string saved_value_;
+  Slice pinned_value_;
  Direction direction_;
  bool valid_;
  bool current_entry_is_merged_;
  Statistics* statistics_;
  uint64_t max_skip_;
+  uint64_t version_number_;
  const Slice* iterate_upper_bound_;
-  IterKey prefix_start_;
-  bool prefix_same_as_start_;
-  bool iter_pinned_;
+  IterKey prefix_start_buf_;
+  Slice prefix_start_key_;
+  const bool prefix_same_as_start_;
+  // Means that we will pin all data blocks we read as long the Iterator
+  // is not deleted, will be true if ReadOptions::pin_data is true
+  const bool pin_thru_lifetime_;
  // List of operands for merge operator.
-  std::deque<std::string> merge_operands_;
+  MergeContext merge_context_;
+  LocalStatistics local_stats_;
+  PinnedIteratorsManager pinned_iters_mgr_;

  // No copying allowed
  DBIter(const DBIter&);
@@ -214,6 +286,8 @@ void DBIter::Next() {
  assert(valid_);

  if (direction_ == kReverse) {
+    // We only pin blocks when doing kReverse
+    ReleaseTempPinnedData();
    FindNextUserKey();
    direction_ = kForward;
    if (!iter_->Valid()) {
@@ -229,24 +303,19 @@ void DBIter::Next() {
    PERF_COUNTER_ADD(internal_key_skipped_count, 1);
  }

+  if (statistics_ != nullptr) {
+    local_stats_.next_count_++;
+  }
  // Now we point to the next internal position, for both of merge and
  // not merge cases.
  if (!iter_->Valid()) {
    valid_ = false;
    return;
  }
-  FindNextUserEntry(true /* skipping the current user key */);
-  if (statistics_ != nullptr) {
-    RecordTick(statistics_, NUMBER_DB_NEXT);
-    if (valid_) {
-      RecordTick(statistics_, NUMBER_DB_NEXT_FOUND);
-      RecordTick(statistics_, ITER_BYTES_READ, key().size() + value().size());
-    }
-  }
-  if (valid_ && prefix_extractor_ && prefix_same_as_start_ &&
-      prefix_extractor_->Transform(saved_key_.GetKey())
-              .compare(prefix_start_.GetKey()) != 0) {
-    valid_ = false;
+  FindNextUserEntry(true /* skipping the current user key */, prefix_same_as_start_);
+  if (statistics_ != nullptr && valid_) {
+    local_stats_.next_found_count_++;
+    local_stats_.bytes_read_ += (key().size() + value().size());
  }
 }

@@ -258,13 +327,18 @@ void DBIter::Next() {
 //
 // NOTE: In between, saved_key_ can point to a user key that has
 //       a delete marker
-inline void DBIter::FindNextUserEntry(bool skipping) {
+//
+// The prefix_check parameter controls whether we check the iterated
+// keys against the prefix of the seeked key. Set to false when
+// performing a seek without a key (e.g. SeekToFirst). Set to
+// prefix_same_as_start_ for other iterations.
+inline void DBIter::FindNextUserEntry(bool skipping, bool prefix_check) {
  PERF_TIMER_GUARD(find_next_user_entry_time);
-  FindNextUserEntryInternal(skipping);
+  FindNextUserEntryInternal(skipping, prefix_check);
 }

 // Actual implementation of DBIter::FindNextUserEntry()
-void DBIter::FindNextUserEntryInternal(bool skipping) {
+void DBIter::FindNextUserEntryInternal(bool skipping, bool prefix_check) {
  // Loop until we hit an acceptable entry to yield
  assert(iter_->Valid());
  assert(direction_ == kForward);
@@ -275,7 +349,12 @@ void DBIter::FindNextUserEntryInternal(bool skipping) {

    if (ParseKey(&ikey)) {
      if (iterate_upper_bound_ != nullptr &&
-          ikey.user_key.compare(*iterate_upper_bound_) >= 0) {
+          user_comparator_->Compare(ikey.user_key, *iterate_upper_bound_) >= 0) {
+        break;
+      }
+
+      if (prefix_extractor_ && prefix_check &&
+          prefix_extractor_->Transform(ikey.user_key).compare(prefix_start_key_) != 0) {
        break;
      }

@@ -290,21 +369,24 @@ void DBIter::FindNextUserEntryInternal(bool skipping) {
            case kTypeSingleDeletion:
              // Arrange to skip all upcoming entries for this key since
              // they are hidden by this deletion.
-              saved_key_.SetKey(ikey.user_key,
-                                !iter_->IsKeyPinned() /* copy */);
+              saved_key_.SetKey(
+                  ikey.user_key,
+                  !iter_->IsKeyPinned() || !pin_thru_lifetime_ /* copy */);
              skipping = true;
              num_skipped = 0;
              PERF_COUNTER_ADD(internal_delete_skipped_count, 1);
              break;
            case kTypeValue:
              valid_ = true;
-              saved_key_.SetKey(ikey.user_key,
-                                !iter_->IsKeyPinned() /* copy */);
+              saved_key_.SetKey(
+                  ikey.user_key,
+                  !iter_->IsKeyPinned() || !pin_thru_lifetime_ /* copy */);
              return;
            case kTypeMerge:
              // By now, we are sure the current ikey is going to yield a value
-              saved_key_.SetKey(ikey.user_key,
-                                !iter_->IsKeyPinned() /* copy */);
+              saved_key_.SetKey(
+                  ikey.user_key,
+                  !iter_->IsKeyPinned() || !pin_thru_lifetime_ /* copy */);
              current_entry_is_merged_ = true;
              valid_ = true;
              MergeValuesNewToOld();  // Go to a different state machine
@@ -350,9 +432,9 @@ void DBIter::MergeValuesNewToOld() {
    return;
  }

+  merge_context_.Clear();
  // Start the merge process by pushing the first operand
-  std::deque<std::string> operands;
-  operands.push_front(iter_->value().ToString());
+  merge_context_.PushOperand(iter_->value());

  ParsedInternalKey ikey;
  for (iter_->Next(); iter_->Valid(); iter_->Next()) {
@@ -377,7 +459,8 @@ void DBIter::MergeValuesNewToOld() {
      {
        StopWatchNano timer(env_, statistics_ != nullptr);
        PERF_TIMER_GUARD(merge_operator_time_nanos);
-        user_merge_operator_->FullMerge(ikey.user_key, &val, operands,
+        user_merge_operator_->FullMerge(ikey.user_key, &val,
+                                        merge_context_.GetOperands(),
                                        &saved_value_, logger_);
        RecordTick(statistics_, MERGE_OPERATION_TOTAL_TIME,
                   timer.ElapsedNanos());
@@ -389,7 +472,7 @@ void DBIter::MergeValuesNewToOld() {
      // hit a merge, add the value as an operand and run associative merge.
      // when complete, add result to operands and continue.
      const Slice& val = iter_->value();
-      operands.push_front(val.ToString());
+      merge_context_.PushOperand(val);
    } else {
      assert(false);
    }
@@ -402,8 +485,9 @@ void DBIter::MergeValuesNewToOld() {
    // a deletion marker.
    // feed null as the existing value to the merge operator, such that
    // client can differentiate this scenario and do things accordingly.
-    user_merge_operator_->FullMerge(saved_key_.GetKey(), nullptr, operands,
-                                    &saved_value_, logger_);
+    user_merge_operator_->FullMerge(saved_key_.GetKey(), nullptr,
+                                    merge_context_.GetOperands(), &saved_value_,
+                                    logger_);
    RecordTick(statistics_, MERGE_OPERATION_TOTAL_TIME, timer.ElapsedNanos());
  }
 }
@@ -413,17 +497,18 @@ void DBIter::Prev() {
  if (direction_ == kForward) {
    ReverseToBackward();
  }
+  ReleaseTempPinnedData();
  PrevInternal();
  if (statistics_ != nullptr) {
-    RecordTick(statistics_, NUMBER_DB_PREV);
+    local_stats_.prev_count_++;
    if (valid_) {
-      RecordTick(statistics_, NUMBER_DB_PREV_FOUND);
-      RecordTick(statistics_, ITER_BYTES_READ, key().size() + value().size());
+      local_stats_.prev_found_count_++;
+      local_stats_.bytes_read_ += (key().size() + value().size());
    }
  }
  if (valid_ && prefix_extractor_ && prefix_same_as_start_ &&
      prefix_extractor_->Transform(saved_key_.GetKey())
-              .compare(prefix_start_.GetKey()) != 0) {
+              .compare(prefix_start_key_) != 0) {
    valid_ = false;
  }
 }
@@ -465,7 +550,7 @@ void DBIter::PrevInternal() {

  while (iter_->Valid()) {
    saved_key_.SetKey(ExtractUserKey(iter_->key()),
-                      !iter_->IsKeyPinned() /* copy */);
+                      !iter_->IsKeyPinned() || !pin_thru_lifetime_ /* copy */);
    if (FindValueForCurrentKey()) {
      valid_ = true;
      if (!iter_->Valid()) {
@@ -495,7 +580,8 @@ void DBIter::PrevInternal() {
 // saved_value_
 bool DBIter::FindValueForCurrentKey() {
  assert(iter_->Valid());
-  merge_operands_.clear();
+  merge_context_.Clear();
+  current_entry_is_merged_ = false;
  // last entry before merge (could be kTypeDeletion, kTypeSingleDeletion or
  // kTypeValue)
  ValueType last_not_merge_type = kTypeDeletion;
@@ -515,19 +601,21 @@ bool DBIter::FindValueForCurrentKey() {
    last_key_entry_type = ikey.type;
    switch (last_key_entry_type) {
      case kTypeValue:
-        merge_operands_.clear();
-        saved_value_ = iter_->value().ToString();
+        merge_context_.Clear();
+        ReleaseTempPinnedData();
+        TempPinData();
+        pinned_value_ = iter_->value();
        last_not_merge_type = kTypeValue;
        break;
      case kTypeDeletion:
      case kTypeSingleDeletion:
-        merge_operands_.clear();
+        merge_context_.Clear();
        last_not_merge_type = last_key_entry_type;
        PERF_COUNTER_ADD(internal_delete_skipped_count, 1);
        break;
      case kTypeMerge:
        assert(user_merge_operator_ != nullptr);
-        merge_operands_.push_back(iter_->value().ToString());
+        merge_context_.PushOperandBack(iter_->value());
        break;
      default:
        assert(false);
@@ -546,24 +634,23 @@ bool DBIter::FindValueForCurrentKey() {
      valid_ = false;
      return false;
    case kTypeMerge:
+      current_entry_is_merged_ = true;
      if (last_not_merge_type == kTypeDeletion) {
        StopWatchNano timer(env_, statistics_ != nullptr);
        PERF_TIMER_GUARD(merge_operator_time_nanos);
        user_merge_operator_->FullMerge(saved_key_.GetKey(), nullptr,
-                                        merge_operands_, &saved_value_,
-                                        logger_);
+                                        merge_context_.GetOperands(),
+                                        &saved_value_, logger_);
        RecordTick(statistics_, MERGE_OPERATION_TOTAL_TIME,
                   timer.ElapsedNanos());
      } else {
        assert(last_not_merge_type == kTypeValue);
-        std::string last_put_value = saved_value_;
-        Slice temp_slice(last_put_value);
        {
          StopWatchNano timer(env_, statistics_ != nullptr);
          PERF_TIMER_GUARD(merge_operator_time_nanos);
-          user_merge_operator_->FullMerge(saved_key_.GetKey(), &temp_slice,
-                                          merge_operands_, &saved_value_,
-                                          logger_);
+          user_merge_operator_->FullMerge(saved_key_.GetKey(), &pinned_value_,
+                                          merge_context_.GetOperands(),
+                                          &saved_value_, logger_);
          RecordTick(statistics_, MERGE_OPERATION_TOTAL_TIME,
                     timer.ElapsedNanos());
        }
@@ -596,7 +683,9 @@ bool DBIter::FindValueForCurrentKeyUsingSeek() {
  if (ikey.type == kTypeValue || ikey.type == kTypeDeletion ||
      ikey.type == kTypeSingleDeletion) {
    if (ikey.type == kTypeValue) {
-      saved_value_ = iter_->value().ToString();
+      ReleaseTempPinnedData();
+      TempPinData();
+      pinned_value_ = iter_->value();
      valid_ = true;
      return true;
    }
@@ -606,11 +695,12 @@ bool DBIter::FindValueForCurrentKeyUsingSeek() {

  // kTypeMerge. We need to collect all kTypeMerge values and save them
  // in operands
-  std::deque<std::string> operands;
+  current_entry_is_merged_ = true;
+  merge_context_.Clear();
  while (iter_->Valid() &&
         user_comparator_->Equal(ikey.user_key, saved_key_.GetKey()) &&
         ikey.type == kTypeMerge) {
-    operands.push_front(iter_->value().ToString());
+    merge_context_.PushOperand(iter_->value());
    iter_->Next();
    FindParseableKey(&ikey, kForward);
  }
@@ -621,7 +711,8 @@ bool DBIter::FindValueForCurrentKeyUsingSeek() {
    {
      StopWatchNano timer(env_, statistics_ != nullptr);
      PERF_TIMER_GUARD(merge_operator_time_nanos);
-      user_merge_operator_->FullMerge(saved_key_.GetKey(), nullptr, operands,
+      user_merge_operator_->FullMerge(saved_key_.GetKey(), nullptr,
+                                      merge_context_.GetOperands(),
                                      &saved_value_, logger_);
      RecordTick(statistics_, MERGE_OPERATION_TOTAL_TIME, timer.ElapsedNanos());
    }
@@ -639,8 +730,9 @@ bool DBIter::FindValueForCurrentKeyUsingSeek() {
  {
    StopWatchNano timer(env_, statistics_ != nullptr);
    PERF_TIMER_GUARD(merge_operator_time_nanos);
-    user_merge_operator_->FullMerge(saved_key_.GetKey(), &val, operands,
-                                    &saved_value_, logger_);
+    user_merge_operator_->FullMerge(saved_key_.GetKey(), &val,
+                                    merge_context_.GetOperands(), &saved_value_,
+                                    logger_);
    RecordTick(statistics_, MERGE_OPERATION_TOTAL_TIME, timer.ElapsedNanos());
  }
  valid_ = true;
@@ -706,6 +798,7 @@ void DBIter::FindParseableKey(ParsedInternalKey* ikey, Direction direction) {

 void DBIter::Seek(const Slice& target) {
  StopWatch sw(env_, statistics_, DB_SEEK);
+  ReleaseTempPinnedData();
  saved_key_.Clear();
  // now savved_key is used to store internal key.
  saved_key_.SetInternalKey(target, sequence_);
@@ -717,9 +810,15 @@ void DBIter::Seek(const Slice& target) {

  RecordTick(statistics_, NUMBER_DB_SEEK);
  if (iter_->Valid()) {
+    if (prefix_extractor_ && prefix_same_as_start_) {
+      prefix_start_key_ = prefix_extractor_->Transform(target);
+    }
    direction_ = kForward;
    ClearSavedValue();
-    FindNextUserEntry(false /* not skipping */);
+    FindNextUserEntry(false /* not skipping */, prefix_same_as_start_);
+    if (!valid_) {
+      prefix_start_key_.clear();
+    }
    if (statistics_ != nullptr) {
      if (valid_) {
        RecordTick(statistics_, NUMBER_DB_SEEK_FOUND);
@@ -730,7 +829,8 @@ void DBIter::Seek(const Slice& target) {
    valid_ = false;
  }
  if (valid_ && prefix_extractor_ && prefix_same_as_start_) {
-    prefix_start_.SetKey(prefix_extractor_->Transform(target));
+    prefix_start_buf_.SetKey(prefix_start_key_);
+    prefix_start_key_ = prefix_start_buf_.GetKey();
  }
 }

@@ -741,6 +841,7 @@ void DBIter::SeekToFirst() {
    max_skip_ = std::numeric_limits<uint64_t>::max();
  }
  direction_ = kForward;
+  ReleaseTempPinnedData();
  ClearSavedValue();

  {
@@ -750,7 +851,7 @@ void DBIter::SeekToFirst() {

  RecordTick(statistics_, NUMBER_DB_SEEK);
  if (iter_->Valid()) {
-    FindNextUserEntry(false /* not skipping */);
+    FindNextUserEntry(false /* not skipping */, false /* no prefix check */);
    if (statistics_ != nullptr) {
      if (valid_) {
        RecordTick(statistics_, NUMBER_DB_SEEK_FOUND);
@@ -761,7 +862,8 @@ void DBIter::SeekToFirst() {
    valid_ = false;
  }
  if (valid_ && prefix_extractor_ && prefix_same_as_start_) {
-    prefix_start_.SetKey(prefix_extractor_->Transform(saved_key_.GetKey()));
+    prefix_start_buf_.SetKey(prefix_extractor_->Transform(saved_key_.GetKey()));
+    prefix_start_key_ = prefix_start_buf_.GetKey();
  }
 }

@@ -772,6 +874,7 @@ void DBIter::SeekToLast() {
    max_skip_ = std::numeric_limits<uint64_t>::max();
  }
  direction_ = kReverse;
+  ReleaseTempPinnedData();
  ClearSavedValue();

  {
@@ -809,7 +912,8 @@ void DBIter::SeekToLast() {
    }
  }
  if (valid_ && prefix_extractor_ && prefix_same_as_start_) {
-    prefix_start_.SetKey(prefix_extractor_->Transform(saved_key_.GetKey()));
+    prefix_start_buf_.SetKey(prefix_extractor_->Transform(saved_key_.GetKey()));
+    prefix_start_key_ = prefix_start_buf_.GetKey();
  }
 }

@@ -818,15 +922,13 @@ Iterator* NewDBIterator(Env* env, const ImmutableCFOptions& ioptions,
                        InternalIterator* internal_iter,
                        const SequenceNumber& sequence,
                        uint64_t max_sequential_skip_in_iterations,
+                        uint64_t version_number,
                        const Slice* iterate_upper_bound,
                        bool prefix_same_as_start, bool pin_data) {
  DBIter* db_iter =
      new DBIter(env, ioptions, user_key_comparator, internal_iter, sequence,
-                 false, max_sequential_skip_in_iterations, iterate_upper_bound,
-                 prefix_same_as_start);
-  if (pin_data) {
-    db_iter->PinData();
-  }
+                 false, max_sequential_skip_in_iterations, version_number,
+                 iterate_upper_bound, prefix_same_as_start, pin_data);
  return db_iter;
 }

@@ -849,12 +951,9 @@ inline void ArenaWrappedDBIter::Prev() { db_iter_->Prev(); }
 inline Slice ArenaWrappedDBIter::key() const { return db_iter_->key(); }
 inline Slice ArenaWrappedDBIter::value() const { return db_iter_->value(); }
 inline Status ArenaWrappedDBIter::status() const { return db_iter_->status(); }
-inline Status ArenaWrappedDBIter::PinData() { return db_iter_->PinData(); }
-inline Status ArenaWrappedDBIter::ReleasePinnedData() {
-  return db_iter_->ReleasePinnedData();
-}
-inline bool ArenaWrappedDBIter::IsKeyPinned() const {
-  return db_iter_->IsKeyPinned();
+inline Status ArenaWrappedDBIter::GetProperty(std::string prop_name,
+                                              std::string* prop) {
+  return db_iter_->GetProperty(prop_name, prop);
 }
 void ArenaWrappedDBIter::RegisterCleanup(CleanupFunction function, void* arg1,
                                         void* arg2) {
@@ -864,7 +963,7 @@ void ArenaWrappedDBIter::RegisterCleanup(CleanupFunction function, void* arg1,
 ArenaWrappedDBIter* NewArenaWrappedDbIterator(
    Env* env, const ImmutableCFOptions& ioptions,
    const Comparator* user_key_comparator, const SequenceNumber& sequence,
-    uint64_t max_sequential_skip_in_iterations,
+    uint64_t max_sequential_skip_in_iterations, uint64_t version_number,
    const Slice* iterate_upper_bound, bool prefix_same_as_start,
    bool pin_data) {
  ArenaWrappedDBIter* iter = new ArenaWrappedDBIter();
@@ -872,13 +971,10 @@ ArenaWrappedDBIter* NewArenaWrappedDbIterator(
  auto mem = arena->AllocateAligned(sizeof(DBIter));
  DBIter* db_iter =
      new (mem) DBIter(env, ioptions, user_key_comparator, nullptr, sequence,
-                       true, max_sequential_skip_in_iterations,
-                       iterate_upper_bound, prefix_same_as_start);
+                       true, max_sequential_skip_in_iterations, version_number,
+                       iterate_upper_bound, prefix_same_as_start, pin_data);

  iter->SetDBIter(db_iter);
-  if (pin_data) {
-    iter->PinData();
-  }

  return iter;
 }
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -9,6 +9,7 @@

 #pragma once
 #include <stdint.h>
+#include <string>
 #include "rocksdb/db.h"
 #include "rocksdb/iterator.h"
 #include "db/dbformat.h"
@@ -24,14 +25,12 @@ class InternalIterator;
 // Return a new iterator that converts internal keys (yielded by
 // "*internal_iter") that were live at the specified "sequence" number
 // into appropriate user keys.
-extern Iterator* NewDBIterator(Env* env, const ImmutableCFOptions& options,
-                               const Comparator* user_key_comparator,
-                               InternalIterator* internal_iter,
-                               const SequenceNumber& sequence,
-                               uint64_t max_sequential_skip_in_iterations,
-                               const Slice* iterate_upper_bound = nullptr,
-                               bool prefix_same_as_start = false,
-                               bool pin_data = false);
+extern Iterator* NewDBIterator(
+    Env* env, const ImmutableCFOptions& options,
+    const Comparator* user_key_comparator, InternalIterator* internal_iter,
+    const SequenceNumber& sequence, uint64_t max_sequential_skip_in_iterations,
+    uint64_t version_number, const Slice* iterate_upper_bound = nullptr,
+    bool prefix_same_as_start = false, bool pin_data = false);

 // A wrapper iterator which wraps DB Iterator and the arena, with which the DB
 // iterator is supposed be allocated. This class is used as an entry point of
@@ -64,9 +63,7 @@ class ArenaWrappedDBIter : public Iterator {
  virtual Status status() const override;

  void RegisterCleanup(CleanupFunction function, void* arg1, void* arg2);
-  virtual Status PinData();
-  virtual Status ReleasePinnedData();
-  virtual bool IsKeyPinned() const override;
+  virtual Status GetProperty(std::string prop_name, std::string* prop) override;

 private:
  DBIter* db_iter_;
@@ -77,7 +74,7 @@ class ArenaWrappedDBIter : public Iterator {
 extern ArenaWrappedDBIter* NewArenaWrappedDbIterator(
    Env* env, const ImmutableCFOptions& options,
    const Comparator* user_key_comparator, const SequenceNumber& sequence,
-    uint64_t max_sequential_skip_in_iterations,
+    uint64_t max_sequential_skip_in_iterations, uint64_t version_number,
    const Slice* iterate_upper_bound = nullptr,
    bool prefix_same_as_start = false, bool pin_data = false);

@@ -1,4 +1,4 @@
-//  Copyright (c) 2014, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -181,10 +181,9 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
    internal_iter->AddPut("b", "val_b");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(
-        NewDBIterator(env_, ImmutableCFOptions(options),
-                      BytewiseComparator(), internal_iter, 10,
-                      options.max_sequential_skip_in_iterations));
+    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
+        10, options.max_sequential_skip_in_iterations, 0));

    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
@@ -215,7 +214,7 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        10, options.max_sequential_skip_in_iterations));
+        10, options.max_sequential_skip_in_iterations, 0));

    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
@@ -241,7 +240,8 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        10, options.max_sequential_skip_in_iterations, ro.iterate_upper_bound));
+        10, options.max_sequential_skip_in_iterations, 0,
+        ro.iterate_upper_bound));

    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
@@ -273,7 +273,8 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        10, options.max_sequential_skip_in_iterations, ro.iterate_upper_bound));
+        10, options.max_sequential_skip_in_iterations, 0,
+        ro.iterate_upper_bound));

    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
@@ -308,7 +309,8 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        10, options.max_sequential_skip_in_iterations, ro.iterate_upper_bound));
+        10, options.max_sequential_skip_in_iterations, 0,
+        ro.iterate_upper_bound));

    db_iter->SeekToLast();
    ASSERT_TRUE(!db_iter->Valid());
@@ -337,7 +339,8 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        7, options.max_sequential_skip_in_iterations, ro.iterate_upper_bound));
+        7, options.max_sequential_skip_in_iterations, 0,
+        ro.iterate_upper_bound));

    SetPerfLevel(kEnableCount);
    ASSERT_TRUE(GetPerfLevel() == kEnableCount);
@@ -374,7 +377,8 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        4, options.max_sequential_skip_in_iterations, ro.iterate_upper_bound));
+        4, options.max_sequential_skip_in_iterations, 0,
+        ro.iterate_upper_bound));

    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
@@ -399,7 +403,8 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        10, options.max_sequential_skip_in_iterations, ro.iterate_upper_bound));
+        10, options.max_sequential_skip_in_iterations, 0,
+        ro.iterate_upper_bound));

    db_iter->SeekToLast();
    ASSERT_TRUE(!db_iter->Valid());
@@ -421,7 +426,8 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        10, options.max_sequential_skip_in_iterations, ro.iterate_upper_bound));
+        10, options.max_sequential_skip_in_iterations, 0,
+        ro.iterate_upper_bound));

    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
@@ -456,7 +462,8 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        7, options.max_sequential_skip_in_iterations, ro.iterate_upper_bound));
+        7, options.max_sequential_skip_in_iterations, 0,
+        ro.iterate_upper_bound));

    SetPerfLevel(kEnableCount);
    ASSERT_TRUE(GetPerfLevel() == kEnableCount);
@@ -482,10 +489,9 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
    internal_iter->AddPut("b", "val_b");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(
-        NewDBIterator(env_, ImmutableCFOptions(options),
-                      BytewiseComparator(), internal_iter, 10,
-                      options.max_sequential_skip_in_iterations));
+    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
+        10, options.max_sequential_skip_in_iterations, 0));

    db_iter->SeekToFirst();
    ASSERT_TRUE(db_iter->Valid());
@@ -524,10 +530,9 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
    internal_iter->AddPut("b", "val_b");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(
-        NewDBIterator(env_, ImmutableCFOptions(options),
-                      BytewiseComparator(), internal_iter, 2,
-                      options.max_sequential_skip_in_iterations));
+    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
+        2, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "b");
@@ -555,10 +560,9 @@ TEST_F(DBIteratorTest, DBIteratorPrevNext) {
    internal_iter->AddPut("c", "val_c");
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(
-        NewDBIterator(env_, ImmutableCFOptions(options),
-                      BytewiseComparator(), internal_iter, 10,
-                      options.max_sequential_skip_in_iterations));
+    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
+        10, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "c");
@@ -583,10 +587,9 @@ TEST_F(DBIteratorTest, DBIteratorEmpty) {
    TestIterator* internal_iter = new TestIterator(BytewiseComparator());
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(
-        NewDBIterator(env_, ImmutableCFOptions(options),
-                      BytewiseComparator(), internal_iter, 0,
-                      options.max_sequential_skip_in_iterations));
+    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
+        0, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(!db_iter->Valid());
  }
@@ -595,10 +598,9 @@ TEST_F(DBIteratorTest, DBIteratorEmpty) {
    TestIterator* internal_iter = new TestIterator(BytewiseComparator());
    internal_iter->Finish();

-    std::unique_ptr<Iterator> db_iter(
-        NewDBIterator(env_, ImmutableCFOptions(options),
-                      BytewiseComparator(), internal_iter, 0,
-                      options.max_sequential_skip_in_iterations));
+    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
+        0, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToFirst();
    ASSERT_TRUE(!db_iter->Valid());
  }
@@ -617,10 +619,9 @@ TEST_F(DBIteratorTest, DBIteratorUseSkipCountSkips) {
  }
  internal_iter->Finish();

-  std::unique_ptr<Iterator> db_iter(
-      NewDBIterator(env_, ImmutableCFOptions(options),
-                    BytewiseComparator(), internal_iter, 2,
-                    options.max_sequential_skip_in_iterations));
+  std::unique_ptr<Iterator> db_iter(NewDBIterator(
+      env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter, 2,
+      options.max_sequential_skip_in_iterations, 0));
  db_iter->SeekToLast();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "c");
@@ -659,9 +660,8 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {

      options.statistics = rocksdb::CreateDBStatistics();
      std::unique_ptr<Iterator> db_iter(NewDBIterator(
-          env_, ImmutableCFOptions(options),
-          BytewiseComparator(), internal_iter, i + 2,
-          options.max_sequential_skip_in_iterations));
+          env_, ImmutableCFOptions(options), BytewiseComparator(),
+          internal_iter, i + 2, options.max_sequential_skip_in_iterations, 0));
      db_iter->SeekToLast();
      ASSERT_TRUE(db_iter->Valid());

@@ -695,9 +695,8 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
      internal_iter->Finish();

      std::unique_ptr<Iterator> db_iter(NewDBIterator(
-          env_, ImmutableCFOptions(options),
-          BytewiseComparator(), internal_iter, i + 2,
-          options.max_sequential_skip_in_iterations));
+          env_, ImmutableCFOptions(options), BytewiseComparator(),
+          internal_iter, i + 2, options.max_sequential_skip_in_iterations, 0));
      db_iter->SeekToLast();
      ASSERT_TRUE(db_iter->Valid());

@@ -724,9 +723,8 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
      internal_iter->Finish();

      std::unique_ptr<Iterator> db_iter(NewDBIterator(
-          env_, ImmutableCFOptions(options),
-          BytewiseComparator(), internal_iter, 202,
-          options.max_sequential_skip_in_iterations));
+          env_, ImmutableCFOptions(options), BytewiseComparator(),
+          internal_iter, 202, options.max_sequential_skip_in_iterations, 0));
      db_iter->SeekToLast();
      ASSERT_TRUE(db_iter->Valid());

@@ -756,10 +754,9 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
      }
      internal_iter->AddPut("c", "200");
      internal_iter->Finish();
-      std::unique_ptr<Iterator> db_iter(
-          NewDBIterator(env_, ImmutableCFOptions(options),
-                        BytewiseComparator(), internal_iter, i,
-                        options.max_sequential_skip_in_iterations));
+      std::unique_ptr<Iterator> db_iter(NewDBIterator(
+          env_, ImmutableCFOptions(options), BytewiseComparator(),
+          internal_iter, i, options.max_sequential_skip_in_iterations, 0));
      db_iter->SeekToLast();
      ASSERT_TRUE(!db_iter->Valid());

@@ -773,10 +770,9 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
    }
    internal_iter->AddPut("c", "200");
    internal_iter->Finish();
-    std::unique_ptr<Iterator> db_iter(
-        NewDBIterator(env_, ImmutableCFOptions(options),
-                      BytewiseComparator(), internal_iter, 200,
-                      options.max_sequential_skip_in_iterations));
+    std::unique_ptr<Iterator> db_iter(NewDBIterator(
+        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
+        200, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "c");
@@ -809,9 +805,8 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
      internal_iter->Finish();

      std::unique_ptr<Iterator> db_iter(NewDBIterator(
-          env_, ImmutableCFOptions(options),
-          BytewiseComparator(), internal_iter, i + 2,
-          options.max_sequential_skip_in_iterations));
+          env_, ImmutableCFOptions(options), BytewiseComparator(),
+          internal_iter, i + 2, options.max_sequential_skip_in_iterations, 0));
      db_iter->SeekToLast();
      ASSERT_TRUE(db_iter->Valid());

@@ -844,9 +839,8 @@ TEST_F(DBIteratorTest, DBIteratorUseSkip) {
      internal_iter->Finish();

      std::unique_ptr<Iterator> db_iter(NewDBIterator(
-          env_, ImmutableCFOptions(options),
-          BytewiseComparator(), internal_iter, i + 2,
-          options.max_sequential_skip_in_iterations));
+          env_, ImmutableCFOptions(options), BytewiseComparator(),
+          internal_iter, i + 2, options.max_sequential_skip_in_iterations, 0));
      db_iter->SeekToLast();
      ASSERT_TRUE(db_iter->Valid());

@@ -887,7 +881,7 @@ TEST_F(DBIteratorTest, DBIterator1) {

  std::unique_ptr<Iterator> db_iter(NewDBIterator(
      env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter, 1,
-      options.max_sequential_skip_in_iterations));
+      options.max_sequential_skip_in_iterations, 0));
  db_iter->SeekToFirst();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -913,7 +907,7 @@ TEST_F(DBIteratorTest, DBIterator2) {

  std::unique_ptr<Iterator> db_iter(NewDBIterator(
      env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter, 0,
-      options.max_sequential_skip_in_iterations));
+      options.max_sequential_skip_in_iterations, 0));
  db_iter->SeekToFirst();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -936,7 +930,7 @@ TEST_F(DBIteratorTest, DBIterator3) {

  std::unique_ptr<Iterator> db_iter(NewDBIterator(
      env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter, 2,
-      options.max_sequential_skip_in_iterations));
+      options.max_sequential_skip_in_iterations, 0));
  db_iter->SeekToFirst();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -958,7 +952,7 @@ TEST_F(DBIteratorTest, DBIterator4) {

  std::unique_ptr<Iterator> db_iter(NewDBIterator(
      env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter, 4,
-      options.max_sequential_skip_in_iterations));
+      options.max_sequential_skip_in_iterations, 0));
  db_iter->SeekToFirst();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -987,7 +981,7 @@ TEST_F(DBIteratorTest, DBIterator5) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        0, options.max_sequential_skip_in_iterations));
+        0, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1009,7 +1003,7 @@ TEST_F(DBIteratorTest, DBIterator5) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        1, options.max_sequential_skip_in_iterations));
+        1, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1031,7 +1025,7 @@ TEST_F(DBIteratorTest, DBIterator5) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        2, options.max_sequential_skip_in_iterations));
+        2, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1053,7 +1047,7 @@ TEST_F(DBIteratorTest, DBIterator5) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        3, options.max_sequential_skip_in_iterations));
+        3, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1075,7 +1069,7 @@ TEST_F(DBIteratorTest, DBIterator5) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        4, options.max_sequential_skip_in_iterations));
+        4, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1097,7 +1091,7 @@ TEST_F(DBIteratorTest, DBIterator5) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        5, options.max_sequential_skip_in_iterations));
+        5, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1119,7 +1113,7 @@ TEST_F(DBIteratorTest, DBIterator5) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        6, options.max_sequential_skip_in_iterations));
+        6, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1145,7 +1139,7 @@ TEST_F(DBIteratorTest, DBIterator6) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        0, options.max_sequential_skip_in_iterations));
+        0, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1167,7 +1161,7 @@ TEST_F(DBIteratorTest, DBIterator6) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        1, options.max_sequential_skip_in_iterations));
+        1, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1189,7 +1183,7 @@ TEST_F(DBIteratorTest, DBIterator6) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        2, options.max_sequential_skip_in_iterations));
+        2, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1211,7 +1205,7 @@ TEST_F(DBIteratorTest, DBIterator6) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        3, options.max_sequential_skip_in_iterations));
+        3, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(!db_iter->Valid());
  }
@@ -1229,7 +1223,7 @@ TEST_F(DBIteratorTest, DBIterator6) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        4, options.max_sequential_skip_in_iterations));
+        4, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1251,7 +1245,7 @@ TEST_F(DBIteratorTest, DBIterator6) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        5, options.max_sequential_skip_in_iterations));
+        5, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1273,7 +1267,7 @@ TEST_F(DBIteratorTest, DBIterator6) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        6, options.max_sequential_skip_in_iterations));
+        6, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1311,7 +1305,7 @@ TEST_F(DBIteratorTest, DBIterator7) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        0, options.max_sequential_skip_in_iterations));
+        0, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
    ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1345,7 +1339,7 @@ TEST_F(DBIteratorTest, DBIterator7) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        2, options.max_sequential_skip_in_iterations));
+        2, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());

@@ -1385,7 +1379,7 @@ TEST_F(DBIteratorTest, DBIterator7) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        4, options.max_sequential_skip_in_iterations));
+        4, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());

@@ -1425,7 +1419,7 @@ TEST_F(DBIteratorTest, DBIterator7) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        5, options.max_sequential_skip_in_iterations));
+        5, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());

@@ -1470,7 +1464,7 @@ TEST_F(DBIteratorTest, DBIterator7) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        6, options.max_sequential_skip_in_iterations));
+        6, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());

@@ -1516,7 +1510,7 @@ TEST_F(DBIteratorTest, DBIterator7) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        7, options.max_sequential_skip_in_iterations));
+        7, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());

@@ -1556,7 +1550,7 @@ TEST_F(DBIteratorTest, DBIterator7) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        9, options.max_sequential_skip_in_iterations));
+        9, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());

@@ -1602,7 +1596,7 @@ TEST_F(DBIteratorTest, DBIterator7) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        13, options.max_sequential_skip_in_iterations));
+        13, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());

@@ -1649,7 +1643,7 @@ TEST_F(DBIteratorTest, DBIterator7) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        14, options.max_sequential_skip_in_iterations));
+        14, options.max_sequential_skip_in_iterations, 0));
    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());

@@ -1678,7 +1672,7 @@ TEST_F(DBIteratorTest, DBIterator8) {

  std::unique_ptr<Iterator> db_iter(NewDBIterator(
      env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-      10, options.max_sequential_skip_in_iterations));
+      10, options.max_sequential_skip_in_iterations, 0));
  db_iter->SeekToLast();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "b");
@@ -1707,7 +1701,7 @@ TEST_F(DBIteratorTest, DBIterator9) {

    std::unique_ptr<Iterator> db_iter(NewDBIterator(
        env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-        10, options.max_sequential_skip_in_iterations));
+        10, options.max_sequential_skip_in_iterations, 0));

    db_iter->SeekToLast();
    ASSERT_TRUE(db_iter->Valid());
@@ -1754,7 +1748,7 @@ TEST_F(DBIteratorTest, DBIterator10) {

  std::unique_ptr<Iterator> db_iter(NewDBIterator(
      env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-      10, options.max_sequential_skip_in_iterations));
+      10, options.max_sequential_skip_in_iterations, 0));

  db_iter->Seek("c");
  ASSERT_TRUE(db_iter->Valid());
@@ -1778,9 +1772,9 @@ TEST_F(DBIteratorTest, SeekToLastOccurrenceSeq0) {
  internal_iter->AddPut("b", "2");
  internal_iter->Finish();

-  std::unique_ptr<Iterator> db_iter(NewDBIterator(
-      env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter,
-      10, 0 /* force seek */));
+  std::unique_ptr<Iterator> db_iter(
+      NewDBIterator(env_, ImmutableCFOptions(options), BytewiseComparator(),
+                    internal_iter, 10, 0 /* force seek */, 0));
  db_iter->SeekToFirst();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1807,7 +1801,7 @@ TEST_F(DBIteratorTest, DBIterator11) {

  std::unique_ptr<Iterator> db_iter(NewDBIterator(
      env_, ImmutableCFOptions(options), BytewiseComparator(), internal_iter, 1,
-      options.max_sequential_skip_in_iterations));
+      options.max_sequential_skip_in_iterations, 0));
  db_iter->SeekToFirst();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "a");
@@ -1832,7 +1826,7 @@ TEST_F(DBIteratorTest, DBIterator12) {

  std::unique_ptr<Iterator> db_iter(
      NewDBIterator(env_, ImmutableCFOptions(options), BytewiseComparator(),
-                    internal_iter, 10, 0));
+                    internal_iter, 10, 0, 0));
  db_iter->SeekToLast();
  ASSERT_TRUE(db_iter->Valid());
  ASSERT_EQ(db_iter->key().ToString(), "c");
@@ -1874,7 +1868,7 @@ class DBIterWithMergeIterTest : public testing::Test {
    db_iter_.reset(NewDBIterator(env_, ImmutableCFOptions(options_),
                                 BytewiseComparator(), merge_iter,
                                 8 /* read data earlier than seqId 8 */,
-                                 3 /* max iterators before reseek */));
+                                 3 /* max iterators before reseek */, 0));
  }

  Env* env_;
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -215,6 +215,34 @@ TEST_F(DBTablePropertiesTest, GetPropertiesOfTablesInRange) {
    TestGetPropertiesOfTablesInRange(std::move(ranges));
  }
 }
+
+TEST_F(DBTablePropertiesTest, GetColumnFamilyNameProperty) {
+  std::string kExtraCfName = "pikachu";
+  CreateAndReopenWithCF({kExtraCfName}, CurrentOptions());
+
+  // Create one table per CF, then verify it was created with the column family
+  // name property.
+  for (int cf = 0; cf < 2; ++cf) {
+    Put(cf, "key", "val");
+    Flush(cf);
+
+    TablePropertiesCollection fname_to_props;
+    ASSERT_OK(db_->GetPropertiesOfAllTables(handles_[cf], &fname_to_props));
+    ASSERT_EQ(1U, fname_to_props.size());
+
+    std::string expected_cf_name;
+    if (cf > 0) {
+      expected_cf_name = kExtraCfName;
+    } else {
+      expected_cf_name = kDefaultColumnFamilyName;
+    }
+    ASSERT_EQ(expected_cf_name,
+              fname_to_props.begin()->second->column_family_name);
+    ASSERT_EQ(cf, static_cast<uint32_t>(
+                      fname_to_props.begin()->second->column_family_id));
+  }
+}
+
 }  // namespace rocksdb

 #endif  // ROCKSDB_LITE
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -654,6 +654,51 @@ TEST_F(DBTestTailingIterator, ManagedTailingIteratorSeekToSame) {
  ASSERT_EQ(found, iter->key().ToString());
 }

+TEST_F(DBTestTailingIterator, ForwardIteratorVersionProperty) {
+  Options options = CurrentOptions();
+  options.write_buffer_size = 1000;
+
+  ReadOptions read_options;
+  read_options.tailing = true;
+
+  Put("foo", "bar");
+
+  uint64_t v1, v2, v3, v4;
+  {
+    std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
+    iter->Seek("foo");
+    std::string prop_value;
+    ASSERT_OK(iter->GetProperty("rocksdb.iterator.super-version-number",
+                                &prop_value));
+    v1 = static_cast<uint64_t>(std::atoi(prop_value.c_str()));
+
+    Put("foo1", "bar1");
+    Flush();
+
+    ASSERT_OK(iter->GetProperty("rocksdb.iterator.super-version-number",
+                                &prop_value));
+    v2 = static_cast<uint64_t>(std::atoi(prop_value.c_str()));
+
+    iter->Seek("f");
+
+    ASSERT_OK(iter->GetProperty("rocksdb.iterator.super-version-number",
+                                &prop_value));
+    v3 = static_cast<uint64_t>(std::atoi(prop_value.c_str()));
+
+    ASSERT_EQ(v1, v2);
+    ASSERT_GT(v3, v2);
+  }
+
+  {
+    std::unique_ptr<Iterator> iter(db_->NewIterator(read_options));
+    iter->Seek("foo");
+    std::string prop_value;
+    ASSERT_OK(iter->GetProperty("rocksdb.iterator.super-version-number",
+                                &prop_value));
+    v4 = static_cast<uint64_t>(std::atoi(prop_value.c_str()));
+  }
+  ASSERT_EQ(v3, v4);
+}
 }  // namespace rocksdb

 #endif  // !defined(ROCKSDB_LITE)
@@ -0,0 +1,939 @@
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+//
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+#include <cstdlib>
+#include "db/db_test_util.h"
+#include "port/stack_trace.h"
+#include "rocksdb/wal_filter.h"
+
+namespace rocksdb {
+
+class DBTest2 : public DBTestBase {
+ public:
+  DBTest2() : DBTestBase("/db_test2") {}
+};
+
+class PrefixFullBloomWithReverseComparator
+    : public DBTestBase,
+      public ::testing::WithParamInterface<bool> {
+ public:
+  PrefixFullBloomWithReverseComparator()
+      : DBTestBase("/prefix_bloom_reverse") {}
+  virtual void SetUp() override { if_cache_filter_ = GetParam(); }
+  bool if_cache_filter_;
+};
+
+TEST_P(PrefixFullBloomWithReverseComparator,
+       PrefixFullBloomWithReverseComparator) {
+  Options options = last_options_;
+  options.comparator = ReverseBytewiseComparator();
+  options.prefix_extractor.reset(NewCappedPrefixTransform(3));
+  options.statistics = rocksdb::CreateDBStatistics();
+  BlockBasedTableOptions bbto;
+  if (if_cache_filter_) {
+    bbto.no_block_cache = false;
+    bbto.cache_index_and_filter_blocks = true;
+    bbto.block_cache = NewLRUCache(1);
+  }
+  bbto.filter_policy.reset(NewBloomFilterPolicy(10, false));
+  bbto.whole_key_filtering = false;
+  options.table_factory.reset(NewBlockBasedTableFactory(bbto));
+  DestroyAndReopen(options);
+
+  ASSERT_OK(dbfull()->Put(WriteOptions(), "bar123", "foo"));
+  ASSERT_OK(dbfull()->Put(WriteOptions(), "bar234", "foo2"));
+  ASSERT_OK(dbfull()->Put(WriteOptions(), "foo123", "foo3"));
+
+  dbfull()->Flush(FlushOptions());
+
+  if (bbto.block_cache) {
+    bbto.block_cache->EraseUnRefEntries();
+  }
+
+  unique_ptr<Iterator> iter(db_->NewIterator(ReadOptions()));
+  iter->Seek("bar345");
+  ASSERT_OK(iter->status());
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_EQ("bar234", iter->key().ToString());
+  ASSERT_EQ("foo2", iter->value().ToString());
+  iter->Next();
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_EQ("bar123", iter->key().ToString());
+  ASSERT_EQ("foo", iter->value().ToString());
+
+  iter->Seek("foo234");
+  ASSERT_OK(iter->status());
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_EQ("foo123", iter->key().ToString());
+  ASSERT_EQ("foo3", iter->value().ToString());
+
+  iter->Seek("bar");
+  ASSERT_OK(iter->status());
+  ASSERT_TRUE(!iter->Valid());
+}
+
+INSTANTIATE_TEST_CASE_P(PrefixFullBloomWithReverseComparator,
+                        PrefixFullBloomWithReverseComparator, testing::Bool());
+
+TEST_F(DBTest2, IteratorPropertyVersionNumber) {
+  Put("", "");
+  Iterator* iter1 = db_->NewIterator(ReadOptions());
+  std::string prop_value;
+  ASSERT_OK(
+      iter1->GetProperty("rocksdb.iterator.super-version-number", &prop_value));
+  uint64_t version_number1 =
+      static_cast<uint64_t>(std::atoi(prop_value.c_str()));
+
+  Put("", "");
+  Flush();
+
+  Iterator* iter2 = db_->NewIterator(ReadOptions());
+  ASSERT_OK(
+      iter2->GetProperty("rocksdb.iterator.super-version-number", &prop_value));
+  uint64_t version_number2 =
+      static_cast<uint64_t>(std::atoi(prop_value.c_str()));
+
+  ASSERT_GT(version_number2, version_number1);
+
+  Put("", "");
+
+  Iterator* iter3 = db_->NewIterator(ReadOptions());
+  ASSERT_OK(
+      iter3->GetProperty("rocksdb.iterator.super-version-number", &prop_value));
+  uint64_t version_number3 =
+      static_cast<uint64_t>(std::atoi(prop_value.c_str()));
+
+  ASSERT_EQ(version_number2, version_number3);
+
+  iter1->SeekToFirst();
+  ASSERT_OK(
+      iter1->GetProperty("rocksdb.iterator.super-version-number", &prop_value));
+  uint64_t version_number1_new =
+      static_cast<uint64_t>(std::atoi(prop_value.c_str()));
+  ASSERT_EQ(version_number1, version_number1_new);
+
+  delete iter1;
+  delete iter2;
+  delete iter3;
+}
+
+TEST_F(DBTest2, CacheIndexAndFilterWithDBRestart) {
+  Options options = CurrentOptions();
+  options.create_if_missing = true;
+  options.statistics = rocksdb::CreateDBStatistics();
+  BlockBasedTableOptions table_options;
+  table_options.cache_index_and_filter_blocks = true;
+  table_options.filter_policy.reset(NewBloomFilterPolicy(20));
+  options.table_factory.reset(new BlockBasedTableFactory(table_options));
+  CreateAndReopenWithCF({"pikachu"}, options);
+
+  Put(1, "a", "begin");
+  Put(1, "z", "end");
+  ASSERT_OK(Flush(1));
+  TryReopenWithColumnFamilies({"default", "pikachu"}, options);
+
+  std::string value;
+  value = Get(1, "a");
+}
+
+#ifndef ROCKSDB_LITE
+namespace {
+  void ValidateKeyExistence(DB* db, const std::vector<Slice>& keys_must_exist,
+    const std::vector<Slice>& keys_must_not_exist) {
+    // Ensure that expected keys exist
+    std::vector<std::string> values;
+    if (keys_must_exist.size() > 0) {
+      std::vector<Status> status_list =
+        db->MultiGet(ReadOptions(), keys_must_exist, &values);
+      for (size_t i = 0; i < keys_must_exist.size(); i++) {
+        ASSERT_OK(status_list[i]);
+      }
+    }
+
+    // Ensure that given keys don't exist
+    if (keys_must_not_exist.size() > 0) {
+      std::vector<Status> status_list =
+        db->MultiGet(ReadOptions(), keys_must_not_exist, &values);
+      for (size_t i = 0; i < keys_must_not_exist.size(); i++) {
+        ASSERT_TRUE(status_list[i].IsNotFound());
+      }
+    }
+  }
+
+}  // namespace
+
+TEST_F(DBTest2, WalFilterTest) {
+  class TestWalFilter : public WalFilter {
+  private:
+    // Processing option that is requested to be applied at the given index
+    WalFilter::WalProcessingOption wal_processing_option_;
+    // Index at which to apply wal_processing_option_
+    // At other indexes default wal_processing_option::kContinueProcessing is
+    // returned.
+    size_t apply_option_at_record_index_;
+    // Current record index, incremented with each record encountered.
+    size_t current_record_index_;
+
+  public:
+    TestWalFilter(WalFilter::WalProcessingOption wal_processing_option,
+      size_t apply_option_for_record_index)
+      : wal_processing_option_(wal_processing_option),
+      apply_option_at_record_index_(apply_option_for_record_index),
+      current_record_index_(0) {}
+
+    virtual WalProcessingOption LogRecord(const WriteBatch& batch,
+      WriteBatch* new_batch,
+      bool* batch_changed) const override {
+      WalFilter::WalProcessingOption option_to_return;
+
+      if (current_record_index_ == apply_option_at_record_index_) {
+        option_to_return = wal_processing_option_;
+      }
+      else {
+        option_to_return = WalProcessingOption::kContinueProcessing;
+      }
+
+      // Filter is passed as a const object for RocksDB to not modify the
+      // object, however we modify it for our own purpose here and hence
+      // cast the constness away.
+      (const_cast<TestWalFilter*>(this)->current_record_index_)++;
+
+      return option_to_return;
+    }
+
+    virtual const char* Name() const override { return "TestWalFilter"; }
+  };
+
+  // Create 3 batches with two keys each
+  std::vector<std::vector<std::string>> batch_keys(3);
+
+  batch_keys[0].push_back("key1");
+  batch_keys[0].push_back("key2");
+  batch_keys[1].push_back("key3");
+  batch_keys[1].push_back("key4");
+  batch_keys[2].push_back("key5");
+  batch_keys[2].push_back("key6");
+
+  // Test with all WAL processing options
+  for (int option = 0;
+    option < static_cast<int>(
+    WalFilter::WalProcessingOption::kWalProcessingOptionMax);
+  option++) {
+    Options options = OptionsForLogIterTest();
+    DestroyAndReopen(options);
+    CreateAndReopenWithCF({ "pikachu" }, options);
+
+    // Write given keys in given batches
+    for (size_t i = 0; i < batch_keys.size(); i++) {
+      WriteBatch batch;
+      for (size_t j = 0; j < batch_keys[i].size(); j++) {
+        batch.Put(handles_[0], batch_keys[i][j], DummyString(1024));
+      }
+      dbfull()->Write(WriteOptions(), &batch);
+    }
+
+    WalFilter::WalProcessingOption wal_processing_option =
+      static_cast<WalFilter::WalProcessingOption>(option);
+
+    // Create a test filter that would apply wal_processing_option at the first
+    // record
+    size_t apply_option_for_record_index = 1;
+    TestWalFilter test_wal_filter(wal_processing_option,
+      apply_option_for_record_index);
+
+    // Reopen database with option to use WAL filter
+    options = OptionsForLogIterTest();
+    options.wal_filter = &test_wal_filter;
+    Status status =
+      TryReopenWithColumnFamilies({ "default", "pikachu" }, options);
+    if (wal_processing_option ==
+      WalFilter::WalProcessingOption::kCorruptedRecord) {
+      assert(!status.ok());
+      // In case of corruption we can turn off paranoid_checks to reopen
+      // databse
+      options.paranoid_checks = false;
+      ReopenWithColumnFamilies({ "default", "pikachu" }, options);
+    }
+    else {
+      assert(status.ok());
+    }
+
+    // Compute which keys we expect to be found
+    // and which we expect not to be found after recovery.
+    std::vector<Slice> keys_must_exist;
+    std::vector<Slice> keys_must_not_exist;
+    switch (wal_processing_option) {
+    case WalFilter::WalProcessingOption::kCorruptedRecord:
+    case WalFilter::WalProcessingOption::kContinueProcessing: {
+      fprintf(stderr, "Testing with complete WAL processing\n");
+      // we expect all records to be processed
+      for (size_t i = 0; i < batch_keys.size(); i++) {
+        for (size_t j = 0; j < batch_keys[i].size(); j++) {
+          keys_must_exist.push_back(Slice(batch_keys[i][j]));
+        }
+      }
+      break;
+    }
+    case WalFilter::WalProcessingOption::kIgnoreCurrentRecord: {
+      fprintf(stderr,
+        "Testing with ignoring record %" ROCKSDB_PRIszt " only\n",
+        apply_option_for_record_index);
+      // We expect the record with apply_option_for_record_index to be not
+      // found.
+      for (size_t i = 0; i < batch_keys.size(); i++) {
+        for (size_t j = 0; j < batch_keys[i].size(); j++) {
+          if (i == apply_option_for_record_index) {
+            keys_must_not_exist.push_back(Slice(batch_keys[i][j]));
+          }
+          else {
+            keys_must_exist.push_back(Slice(batch_keys[i][j]));
+          }
+        }
+      }
+      break;
+    }
+    case WalFilter::WalProcessingOption::kStopReplay: {
+      fprintf(stderr,
+        "Testing with stopping replay from record %" ROCKSDB_PRIszt
+        "\n",
+        apply_option_for_record_index);
+      // We expect records beyond apply_option_for_record_index to be not
+      // found.
+      for (size_t i = 0; i < batch_keys.size(); i++) {
+        for (size_t j = 0; j < batch_keys[i].size(); j++) {
+          if (i >= apply_option_for_record_index) {
+            keys_must_not_exist.push_back(Slice(batch_keys[i][j]));
+          }
+          else {
+            keys_must_exist.push_back(Slice(batch_keys[i][j]));
+          }
+        }
+      }
+      break;
+    }
+    default:
+      assert(false);  // unhandled case
+    }
+
+    bool checked_after_reopen = false;
+
+    while (true) {
+      // Ensure that expected keys exists
+      // and not expected keys don't exist after recovery
+      ValidateKeyExistence(db_, keys_must_exist, keys_must_not_exist);
+
+      if (checked_after_reopen) {
+        break;
+      }
+
+      // reopen database again to make sure previous log(s) are not used
+      //(even if they were skipped)
+      // reopn database with option to use WAL filter
+      options = OptionsForLogIterTest();
+      ReopenWithColumnFamilies({ "default", "pikachu" }, options);
+
+      checked_after_reopen = true;
+    }
+  }
+}
+
+TEST_F(DBTest2, WalFilterTestWithChangeBatch) {
+  class ChangeBatchHandler : public WriteBatch::Handler {
+  private:
+    // Batch to insert keys in
+    WriteBatch* new_write_batch_;
+    // Number of keys to add in the new batch
+    size_t num_keys_to_add_in_new_batch_;
+    // Number of keys added to new batch
+    size_t num_keys_added_;
+
+  public:
+    ChangeBatchHandler(WriteBatch* new_write_batch,
+      size_t num_keys_to_add_in_new_batch)
+      : new_write_batch_(new_write_batch),
+      num_keys_to_add_in_new_batch_(num_keys_to_add_in_new_batch),
+      num_keys_added_(0) {}
+    virtual void Put(const Slice& key, const Slice& value) override {
+      if (num_keys_added_ < num_keys_to_add_in_new_batch_) {
+        new_write_batch_->Put(key, value);
+        ++num_keys_added_;
+      }
+    }
+  };
+
+  class TestWalFilterWithChangeBatch : public WalFilter {
+  private:
+    // Index at which to start changing records
+    size_t change_records_from_index_;
+    // Number of keys to add in the new batch
+    size_t num_keys_to_add_in_new_batch_;
+    // Current record index, incremented with each record encountered.
+    size_t current_record_index_;
+
+  public:
+    TestWalFilterWithChangeBatch(size_t change_records_from_index,
+      size_t num_keys_to_add_in_new_batch)
+      : change_records_from_index_(change_records_from_index),
+      num_keys_to_add_in_new_batch_(num_keys_to_add_in_new_batch),
+      current_record_index_(0) {}
+
+    virtual WalProcessingOption LogRecord(const WriteBatch& batch,
+      WriteBatch* new_batch,
+      bool* batch_changed) const override {
+      if (current_record_index_ >= change_records_from_index_) {
+        ChangeBatchHandler handler(new_batch, num_keys_to_add_in_new_batch_);
+        batch.Iterate(&handler);
+        *batch_changed = true;
+      }
+
+      // Filter is passed as a const object for RocksDB to not modify the
+      // object, however we modify it for our own purpose here and hence
+      // cast the constness away.
+      (const_cast<TestWalFilterWithChangeBatch*>(this)
+        ->current_record_index_)++;
+
+      return WalProcessingOption::kContinueProcessing;
+    }
+
+    virtual const char* Name() const override {
+      return "TestWalFilterWithChangeBatch";
+    }
+  };
+
+  std::vector<std::vector<std::string>> batch_keys(3);
+
+  batch_keys[0].push_back("key1");
+  batch_keys[0].push_back("key2");
+  batch_keys[1].push_back("key3");
+  batch_keys[1].push_back("key4");
+  batch_keys[2].push_back("key5");
+  batch_keys[2].push_back("key6");
+
+  Options options = OptionsForLogIterTest();
+  DestroyAndReopen(options);
+  CreateAndReopenWithCF({ "pikachu" }, options);
+
+  // Write given keys in given batches
+  for (size_t i = 0; i < batch_keys.size(); i++) {
+    WriteBatch batch;
+    for (size_t j = 0; j < batch_keys[i].size(); j++) {
+      batch.Put(handles_[0], batch_keys[i][j], DummyString(1024));
+    }
+    dbfull()->Write(WriteOptions(), &batch);
+  }
+
+  // Create a test filter that would apply wal_processing_option at the first
+  // record
+  size_t change_records_from_index = 1;
+  size_t num_keys_to_add_in_new_batch = 1;
+  TestWalFilterWithChangeBatch test_wal_filter_with_change_batch(
+    change_records_from_index, num_keys_to_add_in_new_batch);
+
+  // Reopen database with option to use WAL filter
+  options = OptionsForLogIterTest();
+  options.wal_filter = &test_wal_filter_with_change_batch;
+  ReopenWithColumnFamilies({ "default", "pikachu" }, options);
+
+  // Ensure that all keys exist before change_records_from_index_
+  // And after that index only single key exists
+  // as our filter adds only single key for each batch
+  std::vector<Slice> keys_must_exist;
+  std::vector<Slice> keys_must_not_exist;
+
+  for (size_t i = 0; i < batch_keys.size(); i++) {
+    for (size_t j = 0; j < batch_keys[i].size(); j++) {
+      if (i >= change_records_from_index && j >= num_keys_to_add_in_new_batch) {
+        keys_must_not_exist.push_back(Slice(batch_keys[i][j]));
+      }
+      else {
+        keys_must_exist.push_back(Slice(batch_keys[i][j]));
+      }
+    }
+  }
+
+  bool checked_after_reopen = false;
+
+  while (true) {
+    // Ensure that expected keys exists
+    // and not expected keys don't exist after recovery
+    ValidateKeyExistence(db_, keys_must_exist, keys_must_not_exist);
+
+    if (checked_after_reopen) {
+      break;
+    }
+
+    // reopen database again to make sure previous log(s) are not used
+    //(even if they were skipped)
+    // reopn database with option to use WAL filter
+    options = OptionsForLogIterTest();
+    ReopenWithColumnFamilies({ "default", "pikachu" }, options);
+
+    checked_after_reopen = true;
+  }
+}
+
+TEST_F(DBTest2, WalFilterTestWithChangeBatchExtraKeys) {
+  class TestWalFilterWithChangeBatchAddExtraKeys : public WalFilter {
+  public:
+    virtual WalProcessingOption LogRecord(const WriteBatch& batch,
+      WriteBatch* new_batch,
+      bool* batch_changed) const override {
+      *new_batch = batch;
+      new_batch->Put("key_extra", "value_extra");
+      *batch_changed = true;
+      return WalProcessingOption::kContinueProcessing;
+    }
+
+    virtual const char* Name() const override {
+      return "WalFilterTestWithChangeBatchExtraKeys";
+    }
+  };
+
+  std::vector<std::vector<std::string>> batch_keys(3);
+
+  batch_keys[0].push_back("key1");
+  batch_keys[0].push_back("key2");
+  batch_keys[1].push_back("key3");
+  batch_keys[1].push_back("key4");
+  batch_keys[2].push_back("key5");
+  batch_keys[2].push_back("key6");
+
+  Options options = OptionsForLogIterTest();
+  DestroyAndReopen(options);
+  CreateAndReopenWithCF({ "pikachu" }, options);
+
+  // Write given keys in given batches
+  for (size_t i = 0; i < batch_keys.size(); i++) {
+    WriteBatch batch;
+    for (size_t j = 0; j < batch_keys[i].size(); j++) {
+      batch.Put(handles_[0], batch_keys[i][j], DummyString(1024));
+    }
+    dbfull()->Write(WriteOptions(), &batch);
+  }
+
+  // Create a test filter that would add extra keys
+  TestWalFilterWithChangeBatchAddExtraKeys test_wal_filter_extra_keys;
+
+  // Reopen database with option to use WAL filter
+  options = OptionsForLogIterTest();
+  options.wal_filter = &test_wal_filter_extra_keys;
+  Status status = TryReopenWithColumnFamilies({"default", "pikachu"}, options);
+  ASSERT_TRUE(status.IsNotSupported());
+
+  // Reopen without filter, now reopen should succeed - previous
+  // attempt to open must not have altered the db.
+  options = OptionsForLogIterTest();
+  ReopenWithColumnFamilies({ "default", "pikachu" }, options);
+
+  std::vector<Slice> keys_must_exist;
+  std::vector<Slice> keys_must_not_exist;  // empty vector
+
+  for (size_t i = 0; i < batch_keys.size(); i++) {
+    for (size_t j = 0; j < batch_keys[i].size(); j++) {
+      keys_must_exist.push_back(Slice(batch_keys[i][j]));
+    }
+  }
+
+  ValidateKeyExistence(db_, keys_must_exist, keys_must_not_exist);
+}
+
+TEST_F(DBTest2, WalFilterTestWithColumnFamilies) {
+  class TestWalFilterWithColumnFamilies : public WalFilter {
+  private:
+    // column_family_id -> log_number map (provided to WALFilter)
+    std::map<uint32_t, uint64_t> cf_log_number_map_;
+    // column_family_name -> column_family_id map (provided to WALFilter)
+    std::map<std::string, uint32_t> cf_name_id_map_;
+    // column_family_name -> keys_found_in_wal map
+    // We store keys that are applicable to the column_family
+    // during recovery (i.e. aren't already flushed to SST file(s))
+    // for verification against the keys we expect.
+    std::map<uint32_t, std::vector<std::string>> cf_wal_keys_;
+  public:
+    virtual void ColumnFamilyLogNumberMap(
+      const std::map<uint32_t, uint64_t>& cf_lognumber_map,
+      const std::map<std::string, uint32_t>& cf_name_id_map) override {
+      cf_log_number_map_ = cf_lognumber_map;
+      cf_name_id_map_ = cf_name_id_map;
+    }
+
+    virtual WalProcessingOption LogRecordFound(unsigned long long log_number,
+      const std::string& log_file_name,
+      const WriteBatch& batch,
+      WriteBatch* new_batch,
+      bool* batch_changed) override {
+      class LogRecordBatchHandler : public WriteBatch::Handler {
+      private:
+        const std::map<uint32_t, uint64_t> & cf_log_number_map_;
+        std::map<uint32_t, std::vector<std::string>> & cf_wal_keys_;
+        unsigned long long log_number_;
+      public:
+        LogRecordBatchHandler(unsigned long long current_log_number,
+          const std::map<uint32_t, uint64_t> & cf_log_number_map,
+          std::map<uint32_t, std::vector<std::string>> & cf_wal_keys) :
+          cf_log_number_map_(cf_log_number_map),
+          cf_wal_keys_(cf_wal_keys),
+          log_number_(current_log_number){}
+
+        virtual Status PutCF(uint32_t column_family_id, const Slice& key,
+          const Slice& /*value*/) override {
+          auto it = cf_log_number_map_.find(column_family_id);
+          assert(it != cf_log_number_map_.end());
+          unsigned long long log_number_for_cf = it->second;
+          // If the current record is applicable for column_family_id
+          // (i.e. isn't flushed to SST file(s) for column_family_id)
+          // add it to the cf_wal_keys_ map for verification.
+          if (log_number_ >= log_number_for_cf) {
+            cf_wal_keys_[column_family_id].push_back(std::string(key.data(),
+              key.size()));
+          }
+          return Status::OK();
+        }
+      } handler(log_number, cf_log_number_map_, cf_wal_keys_);
+
+      batch.Iterate(&handler);
+
+      return WalProcessingOption::kContinueProcessing;
+    }
+
+    virtual const char* Name() const override {
+      return "WalFilterTestWithColumnFamilies";
+    }
+
+    const std::map<uint32_t, std::vector<std::string>>& GetColumnFamilyKeys() {
+      return cf_wal_keys_;
+    }
+
+    const std::map<std::string, uint32_t> & GetColumnFamilyNameIdMap() {
+      return cf_name_id_map_;
+    }
+  };
+
+  std::vector<std::vector<std::string>> batch_keys_pre_flush(3);
+
+  batch_keys_pre_flush[0].push_back("key1");
+  batch_keys_pre_flush[0].push_back("key2");
+  batch_keys_pre_flush[1].push_back("key3");
+  batch_keys_pre_flush[1].push_back("key4");
+  batch_keys_pre_flush[2].push_back("key5");
+  batch_keys_pre_flush[2].push_back("key6");
+
+  Options options = OptionsForLogIterTest();
+  DestroyAndReopen(options);
+  CreateAndReopenWithCF({ "pikachu" }, options);
+
+  // Write given keys in given batches
+  for (size_t i = 0; i < batch_keys_pre_flush.size(); i++) {
+    WriteBatch batch;
+    for (size_t j = 0; j < batch_keys_pre_flush[i].size(); j++) {
+      batch.Put(handles_[0], batch_keys_pre_flush[i][j], DummyString(1024));
+      batch.Put(handles_[1], batch_keys_pre_flush[i][j], DummyString(1024));
+    }
+    dbfull()->Write(WriteOptions(), &batch);
+  }
+
+  //Flush default column-family
+  db_->Flush(FlushOptions(), handles_[0]);
+
+  // Do some more writes
+  std::vector<std::vector<std::string>> batch_keys_post_flush(3);
+
+  batch_keys_post_flush[0].push_back("key7");
+  batch_keys_post_flush[0].push_back("key8");
+  batch_keys_post_flush[1].push_back("key9");
+  batch_keys_post_flush[1].push_back("key10");
+  batch_keys_post_flush[2].push_back("key11");
+  batch_keys_post_flush[2].push_back("key12");
+
+  // Write given keys in given batches
+  for (size_t i = 0; i < batch_keys_post_flush.size(); i++) {
+    WriteBatch batch;
+    for (size_t j = 0; j < batch_keys_post_flush[i].size(); j++) {
+      batch.Put(handles_[0], batch_keys_post_flush[i][j], DummyString(1024));
+      batch.Put(handles_[1], batch_keys_post_flush[i][j], DummyString(1024));
+    }
+    dbfull()->Write(WriteOptions(), &batch);
+  }
+
+  // On Recovery we should only find the second batch applicable to default CF
+  // But both batches applicable to pikachu CF
+
+  // Create a test filter that would add extra keys
+  TestWalFilterWithColumnFamilies test_wal_filter_column_families;
+
+  // Reopen database with option to use WAL filter
+  options = OptionsForLogIterTest();
+  options.wal_filter = &test_wal_filter_column_families;
+  Status status =
+    TryReopenWithColumnFamilies({ "default", "pikachu" }, options);
+  ASSERT_TRUE(status.ok());
+
+  // verify that handles_[0] only has post_flush keys
+  // while handles_[1] has pre and post flush keys
+  auto cf_wal_keys = test_wal_filter_column_families.GetColumnFamilyKeys();
+  auto name_id_map = test_wal_filter_column_families.GetColumnFamilyNameIdMap();
+  size_t index = 0;
+  auto keys_cf = cf_wal_keys[name_id_map[kDefaultColumnFamilyName]];
+  //default column-family, only post_flush keys are expected
+  for (size_t i = 0; i < batch_keys_post_flush.size(); i++) {
+    for (size_t j = 0; j < batch_keys_post_flush[i].size(); j++) {
+      Slice key_from_the_log(keys_cf[index++]);
+      Slice batch_key(batch_keys_post_flush[i][j]);
+      ASSERT_TRUE(key_from_the_log.compare(batch_key) == 0);
+    }
+  }
+  ASSERT_TRUE(index == keys_cf.size());
+
+  index = 0;
+  keys_cf = cf_wal_keys[name_id_map["pikachu"]];
+  //pikachu column-family, all keys are expected
+  for (size_t i = 0; i < batch_keys_pre_flush.size(); i++) {
+    for (size_t j = 0; j < batch_keys_pre_flush[i].size(); j++) {
+      Slice key_from_the_log(keys_cf[index++]);
+      Slice batch_key(batch_keys_pre_flush[i][j]);
+      ASSERT_TRUE(key_from_the_log.compare(batch_key) == 0);
+    }
+  }
+
+  for (size_t i = 0; i < batch_keys_post_flush.size(); i++) {
+    for (size_t j = 0; j < batch_keys_post_flush[i].size(); j++) {
+      Slice key_from_the_log(keys_cf[index++]);
+      Slice batch_key(batch_keys_post_flush[i][j]);
+      ASSERT_TRUE(key_from_the_log.compare(batch_key) == 0);
+    }
+  }
+  ASSERT_TRUE(index == keys_cf.size());
+}
+
+TEST_F(DBTest2, PresetCompressionDict) {
+  const size_t kBlockSizeBytes = 4 << 10;
+  const size_t kL0FileBytes = 128 << 10;
+  const size_t kApproxPerBlockOverheadBytes = 50;
+  const int kNumL0Files = 5;
+
+  Options options;
+  options.arena_block_size = kBlockSizeBytes;
+  options.compaction_style = kCompactionStyleUniversal;
+  options.create_if_missing = true;
+  options.disable_auto_compactions = true;
+  options.level0_file_num_compaction_trigger = kNumL0Files;
+  options.memtable_factory.reset(
+      new SpecialSkipListFactory(kL0FileBytes / kBlockSizeBytes));
+  options.num_levels = 2;
+  options.target_file_size_base = kL0FileBytes;
+  options.target_file_size_multiplier = 2;
+  options.write_buffer_size = kL0FileBytes;
+  BlockBasedTableOptions table_options;
+  table_options.block_size = kBlockSizeBytes;
+  std::vector<CompressionType> compression_types;
+  if (Zlib_Supported()) {
+    compression_types.push_back(kZlibCompression);
+  }
+#if LZ4_VERSION_NUMBER >= 10400  // r124+
+  compression_types.push_back(kLZ4Compression);
+  compression_types.push_back(kLZ4HCCompression);
+#endif                          // LZ4_VERSION_NUMBER >= 10400
+#if ZSTD_VERSION_NUMBER >= 500  // v0.5.0+
+  compression_types.push_back(kZSTDNotFinalCompression);
+#endif  // ZSTD_VERSION_NUMBER >= 500
+
+  for (auto compression_type : compression_types) {
+    options.compression = compression_type;
+    size_t prev_out_bytes;
+    for (int i = 0; i < 2; ++i) {
+      // First iteration: compress without preset dictionary
+      // Second iteration: compress with preset dictionary
+      // To make sure the compression dictionary was actually used, we verify
+      // the compressed size is smaller in the second iteration. Also in the
+      // second iteration, verify the data we get out is the same data we put
+      // in.
+      if (i) {
+        options.compression_opts.max_dict_bytes = kBlockSizeBytes;
+      } else {
+        options.compression_opts.max_dict_bytes = 0;
+      }
+
+      options.statistics = rocksdb::CreateDBStatistics();
+      options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+      CreateAndReopenWithCF({"pikachu"}, options);
+      Random rnd(301);
+      std::string seq_data =
+          RandomString(&rnd, kBlockSizeBytes - kApproxPerBlockOverheadBytes);
+
+      ASSERT_EQ(0, NumTableFilesAtLevel(0, 1));
+      for (int j = 0; j < kNumL0Files; ++j) {
+        for (size_t k = 0; k < kL0FileBytes / kBlockSizeBytes + 1; ++k) {
+          ASSERT_OK(Put(1, Key(static_cast<int>(
+                               j * (kL0FileBytes / kBlockSizeBytes) + k)),
+                        seq_data));
+        }
+        dbfull()->TEST_WaitForFlushMemTable(handles_[1]);
+        ASSERT_EQ(j + 1, NumTableFilesAtLevel(0, 1));
+      }
+      db_->CompactRange(CompactRangeOptions(), handles_[1], nullptr, nullptr);
+      ASSERT_EQ(0, NumTableFilesAtLevel(0, 1));
+      ASSERT_GT(NumTableFilesAtLevel(1, 1), 0);
+
+      size_t out_bytes = 0;
+      std::vector<std::string> files;
+      GetSstFiles(dbname_, &files);
+      for (const auto& file : files) {
+        uint64_t curr_bytes;
+        env_->GetFileSize(dbname_ + "/" + file, &curr_bytes);
+        out_bytes += static_cast<size_t>(curr_bytes);
+      }
+
+      for (size_t j = 0; j < kNumL0Files * (kL0FileBytes / kBlockSizeBytes);
+           j++) {
+        ASSERT_EQ(seq_data, Get(1, Key(static_cast<int>(j))));
+      }
+      if (i) {
+        ASSERT_GT(prev_out_bytes, out_bytes);
+      }
+      prev_out_bytes = out_bytes;
+      DestroyAndReopen(options);
+    }
+  }
+}
+#endif  // ROCKSDB_LITE
+
+TEST_F(DBTest2, FirstSnapshotTest) {
+  Options options;
+  options.write_buffer_size = 100000;  // Small write buffer
+  options = CurrentOptions(options);
+  CreateAndReopenWithCF({"pikachu"}, options);
+
+  // This snapshot will have sequence number 0 what is expected behaviour.
+  const Snapshot* s1 = db_->GetSnapshot();
+
+  Put(1, "k1", std::string(100000, 'x'));  // Fill memtable
+  Put(1, "k2", std::string(100000, 'y'));  // Trigger flush
+
+  db_->ReleaseSnapshot(s1);
+}
+
+class PinL0IndexAndFilterBlocksTest : public DBTestBase,
+                                      public testing::WithParamInterface<bool> {
+ public:
+  PinL0IndexAndFilterBlocksTest() : DBTestBase("/db_pin_l0_index_bloom_test") {}
+  virtual void SetUp() override { infinite_max_files_ = GetParam(); }
+
+  bool infinite_max_files_;
+};
+
+TEST_P(PinL0IndexAndFilterBlocksTest,
+       IndexAndFilterBlocksOfNewTableAddedToCacheWithPinning) {
+  Options options = CurrentOptions();
+  if (infinite_max_files_) {
+    options.max_open_files = -1;
+  }
+  options.create_if_missing = true;
+  options.statistics = rocksdb::CreateDBStatistics();
+  BlockBasedTableOptions table_options;
+  table_options.cache_index_and_filter_blocks = true;
+  table_options.pin_l0_filter_and_index_blocks_in_cache = true;
+  table_options.filter_policy.reset(NewBloomFilterPolicy(20));
+  options.table_factory.reset(new BlockBasedTableFactory(table_options));
+  CreateAndReopenWithCF({"pikachu"}, options);
+
+  ASSERT_OK(Put(1, "key", "val"));
+  // Create a new table.
+  ASSERT_OK(Flush(1));
+
+  // index/filter blocks added to block cache right after table creation.
+  ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
+  ASSERT_EQ(0, TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT));
+  ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_INDEX_MISS));
+  ASSERT_EQ(0, TestGetTickerCount(options, BLOCK_CACHE_INDEX_HIT));
+
+  // only index/filter were added
+  ASSERT_EQ(2, TestGetTickerCount(options, BLOCK_CACHE_ADD));
+  ASSERT_EQ(0, TestGetTickerCount(options, BLOCK_CACHE_DATA_MISS));
+
+  std::string value;
+  // Miss and hit count should remain the same, they're all pinned.
+  db_->KeyMayExist(ReadOptions(), handles_[1], "key", &value);
+  ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
+  ASSERT_EQ(0, TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT));
+  ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_INDEX_MISS));
+  ASSERT_EQ(0, TestGetTickerCount(options, BLOCK_CACHE_INDEX_HIT));
+
+  // Miss and hit count should remain the same, they're all pinned.
+  value = Get(1, "key");
+  ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
+  ASSERT_EQ(0, TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT));
+  ASSERT_EQ(1, TestGetTickerCount(options, BLOCK_CACHE_INDEX_MISS));
+  ASSERT_EQ(0, TestGetTickerCount(options, BLOCK_CACHE_INDEX_HIT));
+}
+
+TEST_P(PinL0IndexAndFilterBlocksTest,
+       MultiLevelIndexAndFilterBlocksCachedWithPinning) {
+  Options options = CurrentOptions();
+  if (infinite_max_files_) {
+    options.max_open_files = -1;
+  }
+  options.create_if_missing = true;
+  options.statistics = rocksdb::CreateDBStatistics();
+  BlockBasedTableOptions table_options;
+  table_options.cache_index_and_filter_blocks = true;
+  table_options.pin_l0_filter_and_index_blocks_in_cache = true;
+  table_options.filter_policy.reset(NewBloomFilterPolicy(20));
+  options.table_factory.reset(new BlockBasedTableFactory(table_options));
+  CreateAndReopenWithCF({"pikachu"}, options);
+
+  Put(1, "a", "begin");
+  Put(1, "z", "end");
+  ASSERT_OK(Flush(1));
+  // move this table to L1
+  dbfull()->TEST_CompactRange(0, nullptr, nullptr, handles_[1]);
+
+  // reset block cache
+  table_options.block_cache = NewLRUCache(64 * 1024);
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+  TryReopenWithColumnFamilies({"default", "pikachu"}, options);
+  // create new table at L0
+  Put(1, "a2", "begin2");
+  Put(1, "z2", "end2");
+  ASSERT_OK(Flush(1));
+
+  table_options.block_cache->EraseUnRefEntries();
+
+  // get base cache values
+  uint64_t fm = TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS);
+  uint64_t fh = TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT);
+  uint64_t im = TestGetTickerCount(options, BLOCK_CACHE_INDEX_MISS);
+  uint64_t ih = TestGetTickerCount(options, BLOCK_CACHE_INDEX_HIT);
+
+  std::string value;
+  // this should be read from L0
+  // so cache values don't change
+  value = Get(1, "a2");
+  ASSERT_EQ(fm, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
+  ASSERT_EQ(fh, TestGetTickerCount(options, BLOCK_CACHE_FILTER_HIT));
+  ASSERT_EQ(im, TestGetTickerCount(options, BLOCK_CACHE_INDEX_MISS));
+  ASSERT_EQ(ih, TestGetTickerCount(options, BLOCK_CACHE_INDEX_HIT));
+
+  // this should be read from L1
+  // the file is opened, prefetching results in a cache filter miss
+  // the block is loaded and added to the cache,
+  // then the get results in a cache hit for L1
+  value = Get(1, "a");
+  ASSERT_EQ(fm + 1, TestGetTickerCount(options, BLOCK_CACHE_FILTER_MISS));
+  ASSERT_EQ(im + 1, TestGetTickerCount(options, BLOCK_CACHE_INDEX_MISS));
+}
+
+INSTANTIATE_TEST_CASE_P(PinL0IndexAndFilterBlocksTest,
+                        PinL0IndexAndFilterBlocksTest, ::testing::Bool());
+
+}  // namespace rocksdb
+
+int main(int argc, char** argv) {
+  rocksdb::port::InstallStackTraceHandler();
+  ::testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}
@@ -1,4 +1,4 @@
-// Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+// Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 // This source code is licensed under the BSD-style license found in the
 // LICENSE file in the root directory of this source tree. An additional grant
 // of patent rights can be found in the PATENTS file in the same directory.
@@ -73,63 +73,69 @@ DBTestBase::~DBTestBase() {
  delete env_;
 }

-// Switch to a fresh database with the next option configuration to
-// test.  Return false if there are no more configurations to test.
-bool DBTestBase::ChangeOptions(int skip_mask) {
-  for (option_config_++; option_config_ < kEnd; option_config_++) {
+bool DBTestBase::ShouldSkipOptions(int option_config, int skip_mask) {
 #ifdef ROCKSDB_LITE
    // These options are not supported in ROCKSDB_LITE
-    if (option_config_ == kHashSkipList ||
-        option_config_ == kPlainTableFirstBytePrefix ||
-        option_config_ == kPlainTableCappedPrefix ||
-        option_config_ == kPlainTableCappedPrefixNonMmap ||
-        option_config_ == kPlainTableAllBytesPrefix ||
-        option_config_ == kVectorRep || option_config_ == kHashLinkList ||
-        option_config_ == kHashCuckoo ||
-        option_config_ == kUniversalCompaction ||
-        option_config_ == kUniversalCompactionMultiLevel ||
-        option_config_ == kUniversalSubcompactions ||
-        option_config_ == kFIFOCompaction) {
-      continue;
+  if (option_config == kHashSkipList ||
+      option_config == kPlainTableFirstBytePrefix ||
+      option_config == kPlainTableCappedPrefix ||
+      option_config == kPlainTableCappedPrefixNonMmap ||
+      option_config == kPlainTableAllBytesPrefix ||
+      option_config == kVectorRep || option_config == kHashLinkList ||
+      option_config == kHashCuckoo || option_config == kUniversalCompaction ||
+      option_config == kUniversalCompactionMultiLevel ||
+      option_config == kUniversalSubcompactions ||
+      option_config == kFIFOCompaction ||
+      option_config == kConcurrentSkipList) {
+    return true;
    }
 #endif

    if ((skip_mask & kSkipDeletesFilterFirst) &&
-        option_config_ == kDeletesFilterFirst) {
-      continue;
+        option_config == kDeletesFilterFirst) {
+      return true;
    }
    if ((skip_mask & kSkipUniversalCompaction) &&
-        (option_config_ == kUniversalCompaction ||
-         option_config_ == kUniversalCompactionMultiLevel)) {
-      continue;
+        (option_config == kUniversalCompaction ||
+         option_config == kUniversalCompactionMultiLevel)) {
+      return true;
    }
-    if ((skip_mask & kSkipMergePut) && option_config_ == kMergePut) {
-      continue;
+    if ((skip_mask & kSkipMergePut) && option_config == kMergePut) {
+      return true;
    }
    if ((skip_mask & kSkipNoSeekToLast) &&
-        (option_config_ == kHashLinkList || option_config_ == kHashSkipList)) {
-      continue;
+        (option_config == kHashLinkList || option_config == kHashSkipList)) {
+      return true;
    }
    if ((skip_mask & kSkipPlainTable) &&
-        (option_config_ == kPlainTableAllBytesPrefix ||
-         option_config_ == kPlainTableFirstBytePrefix ||
-         option_config_ == kPlainTableCappedPrefix ||
-         option_config_ == kPlainTableCappedPrefixNonMmap)) {
-      continue;
+        (option_config == kPlainTableAllBytesPrefix ||
+         option_config == kPlainTableFirstBytePrefix ||
+         option_config == kPlainTableCappedPrefix ||
+         option_config == kPlainTableCappedPrefixNonMmap)) {
+      return true;
    }
    if ((skip_mask & kSkipHashIndex) &&
-        (option_config_ == kBlockBasedTableWithPrefixHashIndex ||
-         option_config_ == kBlockBasedTableWithWholeKeyHashIndex)) {
-      continue;
+        (option_config == kBlockBasedTableWithPrefixHashIndex ||
+         option_config == kBlockBasedTableWithWholeKeyHashIndex)) {
+      return true;
    }
-    if ((skip_mask & kSkipHashCuckoo) && (option_config_ == kHashCuckoo)) {
-      continue;
+    if ((skip_mask & kSkipHashCuckoo) && (option_config == kHashCuckoo)) {
+      return true;
    }
-    if ((skip_mask & kSkipFIFOCompaction) &&
-        option_config_ == kFIFOCompaction) {
-      continue;
+    if ((skip_mask & kSkipFIFOCompaction) && option_config == kFIFOCompaction) {
+      return true;
    }
-    if ((skip_mask & kSkipMmapReads) && option_config_ == kWalDirAndMmapReads) {
+    if ((skip_mask & kSkipMmapReads) && option_config == kWalDirAndMmapReads) {
+      return true;
+    }
+    return false;
+}
+
+// Switch to a fresh database with the next option configuration to
+// test.  Return false if there are no more configurations to test.
+bool DBTestBase::ChangeOptions(int skip_mask) {
+  for (option_config_++; option_config_ < kEnd; option_config_++) {
+    if (ShouldSkipOptions(option_config_, skip_mask)) {
      continue;
    }
    break;
@@ -204,6 +210,13 @@ Options DBTestBase::CurrentOptions(
    const anon::OptionsOverride& options_override) {
  Options options;
  options.write_buffer_size = 4090 * 4096;
+  options.target_file_size_base = 2 * 1024 * 1024;
+  options.max_bytes_for_level_base = 10 * 1024 * 1024;
+  options.max_open_files = 5000;
+  options.base_background_compactions = -1;
+  options.wal_recovery_mode = WALRecoveryMode::kTolerateCorruptedTailRecords;
+  options.compaction_pri = CompactionPri::kByCompensatedSize;
+
  return CurrentOptions(options, options_override);
 }

@@ -295,6 +308,7 @@ Options DBTestBase::CurrentOptions(
    case kPerfOptions:
      options.soft_rate_limit = 2.0;
      options.delayed_write_rate = 8 * 1024 * 1024;
+      options.report_bg_io_stats = true;
      // TODO(3.13) -- test more options
      break;
    case kDeletesFilterFirst:
@@ -333,6 +347,10 @@ Options DBTestBase::CurrentOptions(
      options.prefix_extractor.reset(NewNoopTransform());
      break;
    }
+    case kBlockBasedTableWithIndexRestartInterval: {
+      table_options.index_block_restart_interval = 8;
+      break;
+    }
    case kOptimizeFiltersForHits: {
      options.optimize_filters_for_hits = true;
      set_block_based_table_factory = true;
@@ -356,6 +374,11 @@ Options DBTestBase::CurrentOptions(
      options.max_subcompactions = 4;
      break;
    }
+    case kConcurrentSkipList: {
+      options.allow_concurrent_memtable_write = true;
+      options.enable_write_thread_adaptive_yield = true;
+      break;
+    }

    default:
      break;
@@ -696,6 +719,22 @@ int DBTestBase::NumTableFilesAtLevel(int level, int cf) {
  return atoi(property.c_str());
 }

+double DBTestBase::CompressionRatioAtLevel(int level, int cf) {
+  std::string property;
+  if (cf == 0) {
+    // default cfd
+    EXPECT_TRUE(db_->GetProperty(
+        "rocksdb.compression-ratio-at-level" + NumberToString(level),
+        &property));
+  } else {
+    EXPECT_TRUE(db_->GetProperty(
+        handles_[cf],
+        "rocksdb.compression-ratio-at-level" + NumberToString(level),
+        &property));
+  }
+  return std::stod(property);
+}
+
 int DBTestBase::TotalTableFiles(int cf, int levels) {
  if (levels == -1) {
    levels = CurrentOptions().num_levels;
@@ -1000,4 +1039,59 @@ void DBTestBase::CopyFile(const std::string& source,
  ASSERT_OK(destfile->Close());
 }

+std::unordered_map<std::string, uint64_t> DBTestBase::GetAllSSTFiles(
+    uint64_t* total_size) {
+  std::unordered_map<std::string, uint64_t> res;
+
+  if (total_size) {
+    *total_size = 0;
+  }
+  std::vector<std::string> files;
+  env_->GetChildren(dbname_, &files);
+  for (auto& file_name : files) {
+    uint64_t number;
+    FileType type;
+    std::string file_path = dbname_ + "/" + file_name;
+    if (ParseFileName(file_name, &number, &type) && type == kTableFile) {
+      uint64_t file_size = 0;
+      env_->GetFileSize(file_path, &file_size);
+      res[file_path] = file_size;
+      if (total_size) {
+        *total_size += file_size;
+      }
+    }
+  }
+  return res;
+}
+
+std::vector<std::uint64_t> DBTestBase::ListTableFiles(Env* env,
+                                                      const std::string& path) {
+  std::vector<std::string> files;
+  std::vector<uint64_t> file_numbers;
+  env->GetChildren(path, &files);
+  uint64_t number;
+  FileType type;
+  for (size_t i = 0; i < files.size(); ++i) {
+    if (ParseFileName(files[i], &number, &type)) {
+      if (type == kTableFile) {
+        file_numbers.push_back(number);
+      }
+    }
+  }
+  return file_numbers;
+}
+
+#ifndef ROCKSDB_LITE
+uint64_t DBTestBase::GetNumberOfSstFilesForColumnFamily(
+    DB* db, std::string column_family_name) {
+  std::vector<LiveFileMetaData> metadata;
+  db->GetLiveFilesMetaData(&metadata);
+  uint64_t result = 0;
+  for (auto& fileMetadata : metadata) {
+    result += (fileMetadata.column_family_name == column_family_name);
+  }
+  return result;
+}
+#endif  // ROCKSDB_LITE
+
 }  // namespace rocksdb
@@ -1,4 +1,4 @@
-// Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+// Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 // This source code is licensed under the BSD-style license found in the
 // LICENSE file in the root directory of this source tree. An additional grant
 // of patent rights can be found in the PATENTS file in the same directory.
@@ -19,6 +19,7 @@
 #endif

 #include <algorithm>
+#include <map>
 #include <set>
 #include <string>
 #include <thread>
@@ -27,7 +28,6 @@
 #include <vector>

 #include "db/db_impl.h"
-#include "db/db_test_util.h"
 #include "db/dbformat.h"
 #include "db/filename.h"
 #include "memtable/hash_linklist_rep.h"
@@ -39,6 +39,7 @@
 #include "rocksdb/filter_policy.h"
 #include "rocksdb/options.h"
 #include "rocksdb/slice.h"
+#include "rocksdb/statistics.h"
 #include "rocksdb/table.h"
 #include "rocksdb/utilities/checkpoint.h"
 #include "table/block_based_table_factory.h"
@@ -357,23 +358,30 @@ class SpecialEnv : public EnvWrapper {
    class CountingFile : public RandomAccessFile {
     public:
      CountingFile(unique_ptr<RandomAccessFile>&& target,
-                   anon::AtomicCounter* counter)
-          : target_(std::move(target)), counter_(counter) {}
+                   anon::AtomicCounter* counter,
+                   std::atomic<size_t>* bytes_read)
+          : target_(std::move(target)),
+            counter_(counter),
+            bytes_read_(bytes_read) {}
      virtual Status Read(uint64_t offset, size_t n, Slice* result,
                          char* scratch) const override {
        counter_->Increment();
-        return target_->Read(offset, n, result, scratch);
+        Status s = target_->Read(offset, n, result, scratch);
+        *bytes_read_ += result->size();
+        return s;
      }

     private:
      unique_ptr<RandomAccessFile> target_;
      anon::AtomicCounter* counter_;
+      std::atomic<size_t>* bytes_read_;
    };

    Status s = target()->NewRandomAccessFile(f, r, soptions);
    random_file_open_counter_++;
    if (s.ok() && count_random_reads_) {
-      r->reset(new CountingFile(std::move(*r), &random_read_counter_));
+      r->reset(new CountingFile(std::move(*r), &random_read_counter_,
+                                &random_read_bytes_counter_));
    }
    return s;
  }
@@ -463,6 +471,7 @@ class SpecialEnv : public EnvWrapper {

  bool count_random_reads_;
  anon::AtomicCounter random_read_counter_;
+  std::atomic<size_t> random_read_bytes_counter_;
  std::atomic<int> random_file_open_counter_;

  bool count_sequential_reads_;
@@ -491,6 +500,33 @@ class SpecialEnv : public EnvWrapper {
  std::atomic<bool> is_wal_sync_thread_safe_{true};
 };

+#ifndef ROCKSDB_LITE
+class OnFileDeletionListener : public EventListener {
+ public:
+  OnFileDeletionListener() : matched_count_(0), expected_file_name_("") {}
+
+  void SetExpectedFileName(const std::string file_name) {
+    expected_file_name_ = file_name;
+  }
+
+  void VerifyMatchedCount(size_t expected_value) {
+    ASSERT_EQ(matched_count_, expected_value);
+  }
+
+  void OnTableFileDeleted(const TableFileDeletionInfo& info) override {
+    if (expected_file_name_ != "") {
+      ASSERT_EQ(expected_file_name_, info.file_path);
+      expected_file_name_ = "";
+      matched_count_++;
+    }
+  }
+
+ private:
+  size_t matched_count_;
+  std::string expected_file_name_;
+};
+#endif
+
 class DBTestBase : public testing::Test {
 protected:
  // Sequence of option configurations to try
@@ -525,9 +561,11 @@ class DBTestBase : public testing::Test {
    kOptimizeFiltersForHits = 27,
    kRowCache = 28,
    kRecycleLogFiles = 29,
-    kLevelSubcompactions = 30,
-    kUniversalSubcompactions = 31,
-    kEnd = 30
+    kConcurrentSkipList = 30,
+    kEnd = 31,
+    kLevelSubcompactions = 31,
+    kUniversalSubcompactions = 32,
+    kBlockBasedTableWithIndexRestartInterval = 33,
  };
  int option_config_;

@@ -573,6 +611,8 @@ class DBTestBase : public testing::Test {
    return std::string(buf);
  }

+  static bool ShouldSkipOptions(int option_config, int skip_mask = kNoSkip);
+
  // Switch to a fresh database with the next option configuration to
  // test.  Return false if there are no more configurations to test.
  bool ChangeOptions(int skip_mask = kNoSkip);
@@ -668,6 +708,8 @@ class DBTestBase : public testing::Test {

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

+  double CompressionRatioAtLevel(int level, int cf = 0);
+
  int TotalTableFiles(int cf = 0, int levels = -1);

  // Return spread of files per level
@@ -749,6 +791,20 @@ class DBTestBase : public testing::Test {

  void CopyFile(const std::string& source, const std::string& destination,
                uint64_t size = 0);
+
+  std::unordered_map<std::string, uint64_t> GetAllSSTFiles(
+      uint64_t* total_size = nullptr);
+
+  std::vector<std::uint64_t> ListTableFiles(Env* env, const std::string& path);
+
+#ifndef ROCKSDB_LITE
+  uint64_t GetNumberOfSstFilesForColumnFamily(DB* db,
+                                              std::string column_family_name);
+#endif  // ROCKSDB_LITE
+
+  uint64_t TestGetTickerCount(const Options& options, Tickers ticker_type) {
+    return options.statistics->getTickerCount(ticker_type);
+  }
 };

 }  // namespace rocksdb
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -119,8 +119,7 @@ class DelayFilterFactory : public CompactionFilterFactory {
 // Make sure we don't trigger a problem if the trigger conditon is given
 // to be 0, which is invalid.
 TEST_P(DBTestUniversalCompaction, UniversalCompactionSingleSortedRun) {
-  Options options;
-  options = CurrentOptions(options);
+  Options options = CurrentOptions();

  options.compaction_style = kCompactionStyleUniversal;
  options.num_levels = num_levels_;
@@ -154,6 +153,71 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionSingleSortedRun) {
  }
 }

+TEST_P(DBTestUniversalCompaction, OptimizeFiltersForHits) {
+  Options options = CurrentOptions();
+  options.compaction_style = kCompactionStyleUniversal;
+  options.compaction_options_universal.size_ratio = 5;
+  options.num_levels = num_levels_;
+  options.write_buffer_size = 105 << 10;  // 105KB
+  options.arena_block_size = 4 << 10;
+  options.target_file_size_base = 32 << 10;  // 32KB
+  // trigger compaction if there are >= 4 files
+  options.level0_file_num_compaction_trigger = 4;
+  BlockBasedTableOptions bbto;
+  bbto.cache_index_and_filter_blocks = true;
+  bbto.filter_policy.reset(NewBloomFilterPolicy(10, false));
+  bbto.whole_key_filtering = true;
+  options.table_factory.reset(NewBlockBasedTableFactory(bbto));
+  options.optimize_filters_for_hits = true;
+  options.statistics = rocksdb::CreateDBStatistics();
+  options.memtable_factory.reset(new SpecialSkipListFactory(3));
+
+  DestroyAndReopen(options);
+
+  // block compaction from happening
+  env_->SetBackgroundThreads(1, Env::LOW);
+  test::SleepingBackgroundTask sleeping_task_low;
+  env_->Schedule(&test::SleepingBackgroundTask::DoSleepTask, &sleeping_task_low,
+                 Env::Priority::LOW);
+
+  for (int num = 0; num < options.level0_file_num_compaction_trigger; num++) {
+    Put(Key(num * 10), "val");
+    if (num) {
+      dbfull()->TEST_WaitForFlushMemTable();
+    }
+    Put(Key(30 + num * 10), "val");
+    Put(Key(60 + num * 10), "val");
+  }
+  Put("", "");
+  dbfull()->TEST_WaitForFlushMemTable();
+
+  // Query set of non existing keys
+  for (int i = 5; i < 90; i += 10) {
+    ASSERT_EQ(Get(Key(i)), "NOT_FOUND");
+  }
+
+  // Make sure bloom filter is used at least once.
+  ASSERT_GT(TestGetTickerCount(options, BLOOM_FILTER_USEFUL), 0);
+  auto prev_counter = TestGetTickerCount(options, BLOOM_FILTER_USEFUL);
+
+  // Make sure bloom filter is used for all but the last L0 file when looking
+  // up a non-existent key that's in the range of all L0 files.
+  ASSERT_EQ(Get(Key(35)), "NOT_FOUND");
+  ASSERT_EQ(prev_counter + NumTableFilesAtLevel(0) - 1,
+            TestGetTickerCount(options, BLOOM_FILTER_USEFUL));
+  prev_counter = TestGetTickerCount(options, BLOOM_FILTER_USEFUL);
+
+  // Unblock compaction and wait it for happening.
+  sleeping_task_low.WakeUp();
+  dbfull()->TEST_WaitForCompact();
+
+  // The same queries will not trigger bloom filter
+  for (int i = 5; i < 90; i += 10) {
+    ASSERT_EQ(Get(Key(i)), "NOT_FOUND");
+  }
+  ASSERT_EQ(prev_counter, TestGetTickerCount(options, BLOOM_FILTER_USEFUL));
+}
+
 // TODO(kailiu) The tests on UniversalCompaction has some issues:
 //  1. A lot of magic numbers ("11" or "12").
 //  2. Made assumption on the memtable flush conditions, which may change from
@@ -266,13 +330,12 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionTrigger) {
 }

 TEST_P(DBTestUniversalCompaction, UniversalCompactionSizeAmplification) {
-  Options options;
+  Options options = CurrentOptions();
  options.compaction_style = kCompactionStyleUniversal;
  options.num_levels = num_levels_;
  options.write_buffer_size = 100 << 10;     // 100KB
  options.target_file_size_base = 32 << 10;  // 32KB
  options.level0_file_num_compaction_trigger = 3;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);
  CreateAndReopenWithCF({"pikachu"}, options);

@@ -317,12 +380,12 @@ TEST_P(DBTestUniversalCompaction, CompactFilesOnUniversalCompaction) {
  ChangeCompactOptions();
  Options options;
  options.create_if_missing = true;
-  options.write_buffer_size = kEntrySize * kEntriesPerBuffer;
  options.compaction_style = kCompactionStyleLevel;
  options.num_levels = 1;
  options.target_file_size_base = options.write_buffer_size;
  options.compression = kNoCompression;
  options = CurrentOptions(options);
+  options.write_buffer_size = kEntrySize * kEntriesPerBuffer;
  CreateAndReopenWithCF({"pikachu"}, options);
  ASSERT_EQ(options.compaction_style, kCompactionStyleUniversal);
  Random rnd(301);
@@ -380,12 +443,11 @@ TEST_P(DBTestUniversalCompaction, CompactFilesOnUniversalCompaction) {
 }

 TEST_P(DBTestUniversalCompaction, UniversalCompactionTargetLevel) {
-  Options options;
+  Options options = CurrentOptions();
  options.compaction_style = kCompactionStyleUniversal;
  options.write_buffer_size = 100 << 10;     // 100KB
  options.num_levels = 7;
  options.disable_auto_compactions = true;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);

  // Generate 3 overlapping files
@@ -425,14 +487,13 @@ class DBTestUniversalCompactionMultiLevels
 };

 TEST_P(DBTestUniversalCompactionMultiLevels, UniversalCompactionMultiLevels) {
-  Options options;
+  Options options = CurrentOptions();
  options.compaction_style = kCompactionStyleUniversal;
  options.num_levels = num_levels_;
  options.write_buffer_size = 100 << 10;  // 100KB
  options.level0_file_num_compaction_trigger = 8;
  options.max_background_compactions = 3;
  options.target_file_size_base = 32 * 1024;
-  options = CurrentOptions(options);
  CreateAndReopenWithCF({"pikachu"}, options);

  // Trigger compaction if size amplification exceeds 110%
@@ -468,7 +529,7 @@ TEST_P(DBTestUniversalCompactionMultiLevels, UniversalCompactionTrivialMove) {
      });
  rocksdb::SyncPoint::GetInstance()->EnableProcessing();

-  Options options;
+  Options options = CurrentOptions();
  options.compaction_style = kCompactionStyleUniversal;
  options.compaction_options_universal.allow_trivial_move = true;
  options.num_levels = 3;
@@ -476,7 +537,6 @@ TEST_P(DBTestUniversalCompactionMultiLevels, UniversalCompactionTrivialMove) {
  options.level0_file_num_compaction_trigger = 3;
  options.max_background_compactions = 2;
  options.target_file_size_base = 32 * 1024;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);
  CreateAndReopenWithCF({"pikachu"}, options);

@@ -515,7 +575,7 @@ class DBTestUniversalCompactionParallel :
 };

 TEST_P(DBTestUniversalCompactionParallel, UniversalCompactionParallel) {
-  Options options;
+  Options options = CurrentOptions();
  options.compaction_style = kCompactionStyleUniversal;
  options.num_levels = num_levels_;
  options.write_buffer_size = 1 << 10;  // 1KB
@@ -524,7 +584,6 @@ TEST_P(DBTestUniversalCompactionParallel, UniversalCompactionParallel) {
  options.max_background_flushes = 3;
  options.target_file_size_base = 1 * 1024;
  options.compaction_options_universal.max_size_amplification_percent = 110;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);
  CreateAndReopenWithCF({"pikachu"}, options);

@@ -581,7 +640,7 @@ INSTANTIATE_TEST_CASE_P(DBTestUniversalCompactionParallel,
                                           ::testing::Bool()));

 TEST_P(DBTestUniversalCompaction, UniversalCompactionOptions) {
-  Options options;
+  Options options = CurrentOptions();
  options.compaction_style = kCompactionStyleUniversal;
  options.write_buffer_size = 105 << 10;    // 105KB
  options.arena_block_size = 4 << 10;       // 4KB
@@ -589,7 +648,6 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionOptions) {
  options.level0_file_num_compaction_trigger = 4;
  options.num_levels = num_levels_;
  options.compaction_options_universal.compression_size_percent = -1;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);
  CreateAndReopenWithCF({"pikachu"}, options);

@@ -702,14 +760,13 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionCompressRatio1) {
    return;
  }

-  Options options;
+  Options options = CurrentOptions();
  options.compaction_style = kCompactionStyleUniversal;
  options.write_buffer_size = 100 << 10;     // 100KB
  options.target_file_size_base = 32 << 10;  // 32KB
  options.level0_file_num_compaction_trigger = 2;
  options.num_levels = num_levels_;
  options.compaction_options_universal.compression_size_percent = 70;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);

  Random rnd(301);
@@ -770,14 +827,13 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionCompressRatio2) {
  if (!Snappy_Supported()) {
    return;
  }
-  Options options;
+  Options options = CurrentOptions();
  options.compaction_style = kCompactionStyleUniversal;
  options.write_buffer_size = 100 << 10;     // 100KB
  options.target_file_size_base = 32 << 10;  // 32KB
  options.level0_file_num_compaction_trigger = 2;
  options.num_levels = num_levels_;
  options.compaction_options_universal.compression_size_percent = 95;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);

  Random rnd(301);
@@ -813,7 +869,7 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionTrivialMoveTest1) {
      });
  rocksdb::SyncPoint::GetInstance()->EnableProcessing();

-  Options options;
+  Options options = CurrentOptions();
  options.compaction_style = kCompactionStyleUniversal;
  options.compaction_options_universal.allow_trivial_move = true;
  options.num_levels = 2;
@@ -821,7 +877,6 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionTrivialMoveTest1) {
  options.level0_file_num_compaction_trigger = 3;
  options.max_background_compactions = 1;
  options.target_file_size_base = 32 * 1024;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);
  CreateAndReopenWithCF({"pikachu"}, options);

@@ -858,7 +913,7 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionTrivialMoveTest2) {

  rocksdb::SyncPoint::GetInstance()->EnableProcessing();

-  Options options;
+  Options options = CurrentOptions();
  options.compaction_style = kCompactionStyleUniversal;
  options.compaction_options_universal.allow_trivial_move = true;
  options.num_levels = 15;
@@ -866,7 +921,6 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionTrivialMoveTest2) {
  options.level0_file_num_compaction_trigger = 8;
  options.max_background_compactions = 4;
  options.target_file_size_base = 64 * 1024;
-  options = CurrentOptions(options);
  DestroyAndReopen(options);
  CreateAndReopenWithCF({"pikachu"}, options);

@@ -892,7 +946,7 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionTrivialMoveTest2) {
 }

 TEST_P(DBTestUniversalCompaction, UniversalCompactionFourPaths) {
-  Options options;
+  Options options = CurrentOptions();
  options.db_paths.emplace_back(dbname_, 300 * 1024);
  options.db_paths.emplace_back(dbname_ + "_2", 300 * 1024);
  options.db_paths.emplace_back(dbname_ + "_3", 500 * 1024);
@@ -901,11 +955,10 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionFourPaths) {
      new SpecialSkipListFactory(KNumKeysByGenerateNewFile - 1));
  options.compaction_style = kCompactionStyleUniversal;
  options.compaction_options_universal.size_ratio = 5;
-  options.write_buffer_size = 110 << 10;  // 105KB
+  options.write_buffer_size = 111 << 10;  // 114KB
  options.arena_block_size = 4 << 10;
  options.level0_file_num_compaction_trigger = 2;
  options.num_levels = 1;
-  options = CurrentOptions(options);

  std::vector<std::string> filenames;
  env_->GetChildren(options.db_paths[1].path, &filenames);
@@ -1032,16 +1085,12 @@ TEST_P(DBTestUniversalCompaction, IncreaseUniversalCompactionNumLevels) {
  for (int i = 0; i <= max_key1; i++) {
    // each value is 10K
    ASSERT_OK(Put(1, Key(i), RandomString(&rnd, 10000)));
+    dbfull()->TEST_WaitForFlushMemTable(handles_[1]);
+    dbfull()->TEST_WaitForCompact();
  }
  ASSERT_OK(Flush(1));
  dbfull()->TEST_WaitForCompact();

-  int non_level0_num_files = 0;
-  for (int i = 1; i < options.num_levels; i++) {
-    non_level0_num_files += NumTableFilesAtLevel(i, 1);
-  }
-  ASSERT_EQ(non_level0_num_files, 0);
-
  // Stage 2: reopen with universal compaction, num_levels=4
  options.compaction_style = kCompactionStyleUniversal;
  options.num_levels = 4;
@@ -1054,6 +1103,8 @@ TEST_P(DBTestUniversalCompaction, IncreaseUniversalCompactionNumLevels) {
  for (int i = max_key1 + 1; i <= max_key2; i++) {
    // each value is 10K
    ASSERT_OK(Put(1, Key(i), RandomString(&rnd, 10000)));
+    dbfull()->TEST_WaitForFlushMemTable(handles_[1]);
+    dbfull()->TEST_WaitForCompact();
  }
  ASSERT_OK(Flush(1));
  dbfull()->TEST_WaitForCompact();
@@ -1084,6 +1135,8 @@ TEST_P(DBTestUniversalCompaction, IncreaseUniversalCompactionNumLevels) {
  for (int i = max_key2 + 1; i <= max_key3; i++) {
    // each value is 10K
    ASSERT_OK(Put(1, Key(i), RandomString(&rnd, 10000)));
+    dbfull()->TEST_WaitForFlushMemTable(handles_[1]);
+    dbfull()->TEST_WaitForCompact();
  }
  ASSERT_OK(Flush(1));
  dbfull()->TEST_WaitForCompact();
@@ -1095,19 +1148,17 @@ TEST_P(DBTestUniversalCompaction, UniversalCompactionSecondPathRatio) {
  if (!Snappy_Supported()) {
    return;
  }
-  Options options;
+  Options options = CurrentOptions();
  options.db_paths.emplace_back(dbname_, 500 * 1024);
  options.db_paths.emplace_back(dbname_ + "_2", 1024 * 1024 * 1024);
  options.compaction_style = kCompactionStyleUniversal;
  options.compaction_options_universal.size_ratio = 5;
-  options.write_buffer_size = 110 << 10;  // 105KB
-  options.arena_block_size = 4 * 1024;
+  options.write_buffer_size = 111 << 10;  // 114KB
  options.arena_block_size = 4 << 10;
  options.level0_file_num_compaction_trigger = 2;
  options.num_levels = 1;
  options.memtable_factory.reset(
      new SpecialSkipListFactory(KNumKeysByGenerateNewFile - 1));
-  options = CurrentOptions(options);

  std::vector<std::string> filenames;
  env_->GetChildren(options.db_paths[1].path, &filenames);
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -127,6 +127,637 @@ TEST_F(DBWALTest, SyncWALNotWaitWrite) {
  ASSERT_EQ(Get("foo2"), "bar2");
  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
 }
+
+TEST_F(DBWALTest, Recover) {
+  do {
+    CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
+    ASSERT_OK(Put(1, "foo", "v1"));
+    ASSERT_OK(Put(1, "baz", "v5"));
+
+    ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
+    ASSERT_EQ("v1", Get(1, "foo"));
+
+    ASSERT_EQ("v1", Get(1, "foo"));
+    ASSERT_EQ("v5", Get(1, "baz"));
+    ASSERT_OK(Put(1, "bar", "v2"));
+    ASSERT_OK(Put(1, "foo", "v3"));
+
+    ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
+    ASSERT_EQ("v3", Get(1, "foo"));
+    ASSERT_OK(Put(1, "foo", "v4"));
+    ASSERT_EQ("v4", Get(1, "foo"));
+    ASSERT_EQ("v2", Get(1, "bar"));
+    ASSERT_EQ("v5", Get(1, "baz"));
+  } while (ChangeOptions());
+}
+
+TEST_F(DBWALTest, RecoverWithTableHandle) {
+  do {
+    Options options = CurrentOptions();
+    options.create_if_missing = true;
+    options.disable_auto_compactions = true;
+    DestroyAndReopen(options);
+    CreateAndReopenWithCF({"pikachu"}, options);
+
+    ASSERT_OK(Put(1, "foo", "v1"));
+    ASSERT_OK(Put(1, "bar", "v2"));
+    ASSERT_OK(Flush(1));
+    ASSERT_OK(Put(1, "foo", "v3"));
+    ASSERT_OK(Put(1, "bar", "v4"));
+    ASSERT_OK(Flush(1));
+    ASSERT_OK(Put(1, "big", std::string(100, 'a')));
+    ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
+
+    std::vector<std::vector<FileMetaData>> files;
+    dbfull()->TEST_GetFilesMetaData(handles_[1], &files);
+    size_t total_files = 0;
+    for (const auto& level : files) {
+      total_files += level.size();
+    }
+    ASSERT_EQ(total_files, 3);
+    for (const auto& level : files) {
+      for (const auto& file : level) {
+        if (kInfiniteMaxOpenFiles == option_config_) {
+          ASSERT_TRUE(file.table_reader_handle != nullptr);
+        } else {
+          ASSERT_TRUE(file.table_reader_handle == nullptr);
+        }
+      }
+    }
+  } while (ChangeOptions());
+}
+
+TEST_F(DBWALTest, IgnoreRecoveredLog) {
+  std::string backup_logs = dbname_ + "/backup_logs";
+
+  // delete old files in backup_logs directory
+  env_->CreateDirIfMissing(backup_logs);
+  std::vector<std::string> old_files;
+  env_->GetChildren(backup_logs, &old_files);
+  for (auto& file : old_files) {
+    if (file != "." && file != "..") {
+      env_->DeleteFile(backup_logs + "/" + file);
+    }
+  }
+
+  do {
+    Options options = CurrentOptions();
+    options.create_if_missing = true;
+    options.merge_operator = MergeOperators::CreateUInt64AddOperator();
+    options.wal_dir = dbname_ + "/logs";
+    DestroyAndReopen(options);
+
+    // fill up the DB
+    std::string one, two;
+    PutFixed64(&one, 1);
+    PutFixed64(&two, 2);
+    ASSERT_OK(db_->Merge(WriteOptions(), Slice("foo"), Slice(one)));
+    ASSERT_OK(db_->Merge(WriteOptions(), Slice("foo"), Slice(one)));
+    ASSERT_OK(db_->Merge(WriteOptions(), Slice("bar"), Slice(one)));
+
+    // copy the logs to backup
+    std::vector<std::string> logs;
+    env_->GetChildren(options.wal_dir, &logs);
+    for (auto& log : logs) {
+      if (log != ".." && log != ".") {
+        CopyFile(options.wal_dir + "/" + log, backup_logs + "/" + log);
+      }
+    }
+
+    // recover the DB
+    Reopen(options);
+    ASSERT_EQ(two, Get("foo"));
+    ASSERT_EQ(one, Get("bar"));
+    Close();
+
+    // copy the logs from backup back to wal dir
+    for (auto& log : logs) {
+      if (log != ".." && log != ".") {
+        CopyFile(backup_logs + "/" + log, options.wal_dir + "/" + log);
+      }
+    }
+    // this should ignore the log files, recovery should not happen again
+    // if the recovery happens, the same merge operator would be called twice,
+    // leading to incorrect results
+    Reopen(options);
+    ASSERT_EQ(two, Get("foo"));
+    ASSERT_EQ(one, Get("bar"));
+    Close();
+    Destroy(options);
+    Reopen(options);
+    Close();
+
+    // copy the logs from backup back to wal dir
+    env_->CreateDirIfMissing(options.wal_dir);
+    for (auto& log : logs) {
+      if (log != ".." && log != ".") {
+        CopyFile(backup_logs + "/" + log, options.wal_dir + "/" + log);
+      }
+    }
+    // assert that we successfully recovered only from logs, even though we
+    // destroyed the DB
+    Reopen(options);
+    ASSERT_EQ(two, Get("foo"));
+    ASSERT_EQ(one, Get("bar"));
+
+    // Recovery will fail if DB directory doesn't exist.
+    Destroy(options);
+    // copy the logs from backup back to wal dir
+    env_->CreateDirIfMissing(options.wal_dir);
+    for (auto& log : logs) {
+      if (log != ".." && log != ".") {
+        CopyFile(backup_logs + "/" + log, options.wal_dir + "/" + log);
+        // we won't be needing this file no more
+        env_->DeleteFile(backup_logs + "/" + log);
+      }
+    }
+    Status s = TryReopen(options);
+    ASSERT_TRUE(!s.ok());
+  } while (ChangeOptions(kSkipHashCuckoo));
+}
+
+TEST_F(DBWALTest, RecoveryWithEmptyLog) {
+  do {
+    CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
+    ASSERT_OK(Put(1, "foo", "v1"));
+    ASSERT_OK(Put(1, "foo", "v2"));
+    ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
+    ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
+    ASSERT_OK(Put(1, "foo", "v3"));
+    ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
+    ASSERT_EQ("v3", Get(1, "foo"));
+  } while (ChangeOptions());
+}
+
+#ifndef ROCKSDB_LITE
+TEST_F(DBWALTest, RecoverWithLargeLog) {
+  do {
+    {
+      Options options = CurrentOptions();
+      CreateAndReopenWithCF({"pikachu"}, options);
+      ASSERT_OK(Put(1, "big1", std::string(200000, '1')));
+      ASSERT_OK(Put(1, "big2", std::string(200000, '2')));
+      ASSERT_OK(Put(1, "small3", std::string(10, '3')));
+      ASSERT_OK(Put(1, "small4", std::string(10, '4')));
+      ASSERT_EQ(NumTableFilesAtLevel(0, 1), 0);
+    }
+
+    // Make sure that if we re-open with a small write buffer size that
+    // we flush table files in the middle of a large log file.
+    Options options;
+    options.write_buffer_size = 100000;
+    options = CurrentOptions(options);
+    ReopenWithColumnFamilies({"default", "pikachu"}, options);
+    ASSERT_EQ(NumTableFilesAtLevel(0, 1), 3);
+    ASSERT_EQ(std::string(200000, '1'), Get(1, "big1"));
+    ASSERT_EQ(std::string(200000, '2'), Get(1, "big2"));
+    ASSERT_EQ(std::string(10, '3'), Get(1, "small3"));
+    ASSERT_EQ(std::string(10, '4'), Get(1, "small4"));
+    ASSERT_GT(NumTableFilesAtLevel(0, 1), 1);
+  } while (ChangeCompactOptions());
+}
+
+// In https://reviews.facebook.net/D20661 we change
+// recovery behavior: previously for each log file each column family
+// memtable was flushed, even it was empty. Now it's changed:
+// we try to create the smallest number of table files by merging
+// updates from multiple logs
+TEST_F(DBWALTest, RecoverCheckFileAmountWithSmallWriteBuffer) {
+  Options options = CurrentOptions();
+  options.write_buffer_size = 5000000;
+  CreateAndReopenWithCF({"pikachu", "dobrynia", "nikitich"}, options);
+
+  // Since we will reopen DB with smaller write_buffer_size,
+  // each key will go to new SST file
+  ASSERT_OK(Put(1, Key(10), DummyString(1000000)));
+  ASSERT_OK(Put(1, Key(10), DummyString(1000000)));
+  ASSERT_OK(Put(1, Key(10), DummyString(1000000)));
+  ASSERT_OK(Put(1, Key(10), DummyString(1000000)));
+
+  ASSERT_OK(Put(3, Key(10), DummyString(1)));
+  // Make 'dobrynia' to be flushed and new WAL file to be created
+  ASSERT_OK(Put(2, Key(10), DummyString(7500000)));
+  ASSERT_OK(Put(2, Key(1), DummyString(1)));
+  dbfull()->TEST_WaitForFlushMemTable(handles_[2]);
+  {
+    auto tables = ListTableFiles(env_, dbname_);
+    ASSERT_EQ(tables.size(), static_cast<size_t>(1));
+    // Make sure 'dobrynia' was flushed: check sst files amount
+    ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "dobrynia"),
+              static_cast<uint64_t>(1));
+  }
+  // New WAL file
+  ASSERT_OK(Put(1, Key(1), DummyString(1)));
+  ASSERT_OK(Put(1, Key(1), DummyString(1)));
+  ASSERT_OK(Put(3, Key(10), DummyString(1)));
+  ASSERT_OK(Put(3, Key(10), DummyString(1)));
+  ASSERT_OK(Put(3, Key(10), DummyString(1)));
+
+  options.write_buffer_size = 4096;
+  options.arena_block_size = 4096;
+  ReopenWithColumnFamilies({"default", "pikachu", "dobrynia", "nikitich"},
+                           options);
+  {
+    // No inserts => default is empty
+    ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "default"),
+              static_cast<uint64_t>(0));
+    // First 4 keys goes to separate SSTs + 1 more SST for 2 smaller keys
+    ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "pikachu"),
+              static_cast<uint64_t>(5));
+    // 1 SST for big key + 1 SST for small one
+    ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "dobrynia"),
+              static_cast<uint64_t>(2));
+    // 1 SST for all keys
+    ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "nikitich"),
+              static_cast<uint64_t>(1));
+  }
+}
+
+// In https://reviews.facebook.net/D20661 we change
+// recovery behavior: previously for each log file each column family
+// memtable was flushed, even it wasn't empty. Now it's changed:
+// we try to create the smallest number of table files by merging
+// updates from multiple logs
+TEST_F(DBWALTest, RecoverCheckFileAmount) {
+  Options options = CurrentOptions();
+  options.write_buffer_size = 100000;
+  options.arena_block_size = 4 * 1024;
+  CreateAndReopenWithCF({"pikachu", "dobrynia", "nikitich"}, options);
+
+  ASSERT_OK(Put(0, Key(1), DummyString(1)));
+  ASSERT_OK(Put(1, Key(1), DummyString(1)));
+  ASSERT_OK(Put(2, Key(1), DummyString(1)));
+
+  // Make 'nikitich' memtable to be flushed
+  ASSERT_OK(Put(3, Key(10), DummyString(1002400)));
+  ASSERT_OK(Put(3, Key(1), DummyString(1)));
+  dbfull()->TEST_WaitForFlushMemTable(handles_[3]);
+  // 4 memtable are not flushed, 1 sst file
+  {
+    auto tables = ListTableFiles(env_, dbname_);
+    ASSERT_EQ(tables.size(), static_cast<size_t>(1));
+    ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "nikitich"),
+              static_cast<uint64_t>(1));
+  }
+  // Memtable for 'nikitich' has flushed, new WAL file has opened
+  // 4 memtable still not flushed
+
+  // Write to new WAL file
+  ASSERT_OK(Put(0, Key(1), DummyString(1)));
+  ASSERT_OK(Put(1, Key(1), DummyString(1)));
+  ASSERT_OK(Put(2, Key(1), DummyString(1)));
+
+  // Fill up 'nikitich' one more time
+  ASSERT_OK(Put(3, Key(10), DummyString(1002400)));
+  // make it flush
+  ASSERT_OK(Put(3, Key(1), DummyString(1)));
+  dbfull()->TEST_WaitForFlushMemTable(handles_[3]);
+  // There are still 4 memtable not flushed, and 2 sst tables
+  ASSERT_OK(Put(0, Key(1), DummyString(1)));
+  ASSERT_OK(Put(1, Key(1), DummyString(1)));
+  ASSERT_OK(Put(2, Key(1), DummyString(1)));
+
+  {
+    auto tables = ListTableFiles(env_, dbname_);
+    ASSERT_EQ(tables.size(), static_cast<size_t>(2));
+    ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "nikitich"),
+              static_cast<uint64_t>(2));
+  }
+
+  ReopenWithColumnFamilies({"default", "pikachu", "dobrynia", "nikitich"},
+                           options);
+  {
+    std::vector<uint64_t> table_files = ListTableFiles(env_, dbname_);
+    // Check, that records for 'default', 'dobrynia' and 'pikachu' from
+    // first, second and third WALs  went to the same SST.
+    // So, there is 6 SSTs: three  for 'nikitich', one for 'default', one for
+    // 'dobrynia', one for 'pikachu'
+    ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "default"),
+              static_cast<uint64_t>(1));
+    ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "nikitich"),
+              static_cast<uint64_t>(3));
+    ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "dobrynia"),
+              static_cast<uint64_t>(1));
+    ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "pikachu"),
+              static_cast<uint64_t>(1));
+  }
+}
+
+TEST_F(DBWALTest, SyncMultipleLogs) {
+  const uint64_t kNumBatches = 2;
+  const int kBatchSize = 1000;
+
+  Options options = CurrentOptions();
+  options.create_if_missing = true;
+  options.write_buffer_size = 4096;
+  Reopen(options);
+
+  WriteBatch batch;
+  WriteOptions wo;
+  wo.sync = true;
+
+  for (uint64_t b = 0; b < kNumBatches; b++) {
+    batch.Clear();
+    for (int i = 0; i < kBatchSize; i++) {
+      batch.Put(Key(i), DummyString(128));
+    }
+
+    dbfull()->Write(wo, &batch);
+  }
+
+  ASSERT_OK(dbfull()->SyncWAL());
+}
+
+//
+// Test WAL recovery for the various modes available
+//
+class RecoveryTestHelper {
+ public:
+  // Number of WAL files to generate
+  static const int kWALFilesCount = 10;
+  // Starting number for the WAL file name like 00010.log
+  static const int kWALFileOffset = 10;
+  // Keys to be written per WAL file
+  static const int kKeysPerWALFile = 1024;
+  // Size of the value
+  static const int kValueSize = 10;
+
+  // Create WAL files with values filled in
+  static void FillData(DBWALTest* test, const Options& options,
+                       const size_t wal_count, size_t* count) {
+    const DBOptions& db_options = options;
+
+    *count = 0;
+
+    shared_ptr<Cache> table_cache = NewLRUCache(50000, 16);
+    EnvOptions env_options;
+    WriteBuffer write_buffer(db_options.db_write_buffer_size);
+
+    unique_ptr<VersionSet> versions;
+    unique_ptr<WalManager> wal_manager;
+    WriteController write_controller;
+
+    versions.reset(new VersionSet(test->dbname_, &db_options, env_options,
+                                  table_cache.get(), &write_buffer,
+                                  &write_controller));
+
+    wal_manager.reset(new WalManager(db_options, env_options));
+
+    std::unique_ptr<log::Writer> current_log_writer;
+
+    for (size_t j = kWALFileOffset; j < wal_count + kWALFileOffset; j++) {
+      uint64_t current_log_number = j;
+      std::string fname = LogFileName(test->dbname_, current_log_number);
+      unique_ptr<WritableFile> file;
+      ASSERT_OK(db_options.env->NewWritableFile(fname, &file, env_options));
+      unique_ptr<WritableFileWriter> file_writer(
+          new WritableFileWriter(std::move(file), env_options));
+      current_log_writer.reset(
+          new log::Writer(std::move(file_writer), current_log_number,
+                          db_options.recycle_log_file_num > 0));
+
+      for (int i = 0; i < kKeysPerWALFile; i++) {
+        std::string key = "key" + ToString((*count)++);
+        std::string value = test->DummyString(kValueSize);
+        assert(current_log_writer.get() != nullptr);
+        uint64_t seq = versions->LastSequence() + 1;
+        WriteBatch batch;
+        batch.Put(key, value);
+        WriteBatchInternal::SetSequence(&batch, seq);
+        current_log_writer->AddRecord(WriteBatchInternal::Contents(&batch));
+        versions->SetLastSequence(seq);
+      }
+    }
+  }
+
+  // Recreate and fill the store with some data
+  static size_t FillData(DBWALTest* test, Options* options) {
+    options->create_if_missing = true;
+    test->DestroyAndReopen(*options);
+    test->Close();
+
+    size_t count = 0;
+    FillData(test, *options, kWALFilesCount, &count);
+    return count;
+  }
+
+  // Read back all the keys we wrote and return the number of keys found
+  static size_t GetData(DBWALTest* test) {
+    size_t count = 0;
+    for (size_t i = 0; i < kWALFilesCount * kKeysPerWALFile; i++) {
+      if (test->Get("key" + ToString(i)) != "NOT_FOUND") {
+        ++count;
+      }
+    }
+    return count;
+  }
+
+  // Manuall corrupt the specified WAL
+  static void CorruptWAL(DBWALTest* test, const Options& options,
+                         const double off, const double len,
+                         const int wal_file_id, const bool trunc = false) {
+    Env* env = options.env;
+    std::string fname = LogFileName(test->dbname_, wal_file_id);
+    uint64_t size;
+    ASSERT_OK(env->GetFileSize(fname, &size));
+    ASSERT_GT(size, 0);
+#ifdef OS_WIN
+    // Windows disk cache behaves differently. When we truncate
+    // the original content is still in the cache due to the original
+    // handle is still open. Generally, in Windows, one prohibits
+    // shared access to files and it is not needed for WAL but we allow
+    // it to induce corruption at various tests.
+    test->Close();
+#endif
+    if (trunc) {
+      ASSERT_EQ(0, truncate(fname.c_str(), static_cast<int64_t>(size * off)));
+    } else {
+      InduceCorruption(fname, static_cast<size_t>(size * off),
+                       static_cast<size_t>(size * len));
+    }
+  }
+
+  // Overwrite data with 'a' from offset for length len
+  static void InduceCorruption(const std::string& filename, size_t offset,
+                               size_t len) {
+    ASSERT_GT(len, 0U);
+
+    int fd = open(filename.c_str(), O_RDWR);
+
+    // On windows long is 32-bit
+    ASSERT_LE(offset, std::numeric_limits<long>::max());
+
+    ASSERT_GT(fd, 0);
+    ASSERT_EQ(offset, lseek(fd, static_cast<long>(offset), SEEK_SET));
+
+    void* buf = alloca(len);
+    memset(buf, 'a', len);
+    ASSERT_EQ(len, write(fd, buf, static_cast<unsigned int>(len)));
+
+    close(fd);
+  }
+};
+
+// Test scope:
+// - We expect to open the data store when there is incomplete trailing writes
+// at the end of any of the logs
+// - We do not expect to open the data store for corruption
+TEST_F(DBWALTest, kTolerateCorruptedTailRecords) {
+  const int jstart = RecoveryTestHelper::kWALFileOffset;
+  const int jend = jstart + RecoveryTestHelper::kWALFilesCount;
+
+  for (auto trunc : {true, false}) {        /* Corruption style */
+    for (int i = 0; i < 4; i++) {           /* Corruption offset position */
+      for (int j = jstart; j < jend; j++) { /* WAL file */
+        // Fill data for testing
+        Options options = CurrentOptions();
+        const size_t row_count = RecoveryTestHelper::FillData(this, &options);
+        // test checksum failure or parsing
+        RecoveryTestHelper::CorruptWAL(this, options, /*off=*/i * .3,
+                                       /*len%=*/.1, /*wal=*/j, trunc);
+
+        if (trunc) {
+          options.wal_recovery_mode =
+              WALRecoveryMode::kTolerateCorruptedTailRecords;
+          options.create_if_missing = false;
+          ASSERT_OK(TryReopen(options));
+          const size_t recovered_row_count = RecoveryTestHelper::GetData(this);
+          ASSERT_TRUE(i == 0 || recovered_row_count > 0);
+          ASSERT_LT(recovered_row_count, row_count);
+        } else {
+          options.wal_recovery_mode =
+              WALRecoveryMode::kTolerateCorruptedTailRecords;
+          ASSERT_NOK(TryReopen(options));
+        }
+      }
+    }
+  }
+}
+
+// Test scope:
+// We don't expect the data store to be opened if there is any corruption
+// (leading, middle or trailing -- incomplete writes or corruption)
+TEST_F(DBWALTest, kAbsoluteConsistency) {
+  const int jstart = RecoveryTestHelper::kWALFileOffset;
+  const int jend = jstart + RecoveryTestHelper::kWALFilesCount;
+
+  // Verify clean slate behavior
+  Options options = CurrentOptions();
+  const size_t row_count = RecoveryTestHelper::FillData(this, &options);
+  options.wal_recovery_mode = WALRecoveryMode::kAbsoluteConsistency;
+  options.create_if_missing = false;
+  ASSERT_OK(TryReopen(options));
+  ASSERT_EQ(RecoveryTestHelper::GetData(this), row_count);
+
+  for (auto trunc : {true, false}) { /* Corruption style */
+    for (int i = 0; i < 4; i++) {    /* Corruption offset position */
+      if (trunc && i == 0) {
+        continue;
+      }
+
+      for (int j = jstart; j < jend; j++) { /* wal files */
+        // fill with new date
+        RecoveryTestHelper::FillData(this, &options);
+        // corrupt the wal
+        RecoveryTestHelper::CorruptWAL(this, options, /*off=*/i * .3,
+                                       /*len%=*/.1, j, trunc);
+        // verify
+        options.wal_recovery_mode = WALRecoveryMode::kAbsoluteConsistency;
+        options.create_if_missing = false;
+        ASSERT_NOK(TryReopen(options));
+      }
+    }
+  }
+}
+
+// Test scope:
+// - We expect to open data store under all circumstances
+// - We expect only data upto the point where the first error was encountered
+TEST_F(DBWALTest, kPointInTimeRecovery) {
+  const int jstart = RecoveryTestHelper::kWALFileOffset;
+  const int jend = jstart + RecoveryTestHelper::kWALFilesCount;
+  const int maxkeys =
+      RecoveryTestHelper::kWALFilesCount * RecoveryTestHelper::kKeysPerWALFile;
+
+  for (auto trunc : {true, false}) {        /* Corruption style */
+    for (int i = 0; i < 4; i++) {           /* Offset of corruption */
+      for (int j = jstart; j < jend; j++) { /* WAL file */
+        // Fill data for testing
+        Options options = CurrentOptions();
+        const size_t row_count = RecoveryTestHelper::FillData(this, &options);
+
+        // Corrupt the wal
+        RecoveryTestHelper::CorruptWAL(this, options, /*off=*/i * .3,
+                                       /*len%=*/.1, j, trunc);
+
+        // Verify
+        options.wal_recovery_mode = WALRecoveryMode::kPointInTimeRecovery;
+        options.create_if_missing = false;
+        ASSERT_OK(TryReopen(options));
+
+        // Probe data for invariants
+        size_t recovered_row_count = RecoveryTestHelper::GetData(this);
+        ASSERT_LT(recovered_row_count, row_count);
+
+        bool expect_data = true;
+        for (size_t k = 0; k < maxkeys; ++k) {
+          bool found = Get("key" + ToString(i)) != "NOT_FOUND";
+          if (expect_data && !found) {
+            expect_data = false;
+          }
+          ASSERT_EQ(found, expect_data);
+        }
+
+        const size_t min = RecoveryTestHelper::kKeysPerWALFile *
+                           (j - RecoveryTestHelper::kWALFileOffset);
+        ASSERT_GE(recovered_row_count, min);
+        if (!trunc && i != 0) {
+          const size_t max = RecoveryTestHelper::kKeysPerWALFile *
+                             (j - RecoveryTestHelper::kWALFileOffset + 1);
+          ASSERT_LE(recovered_row_count, max);
+        }
+      }
+    }
+  }
+}
+
+// Test scope:
+// - We expect to open the data store under all scenarios
+// - We expect to have recovered records past the corruption zone
+TEST_F(DBWALTest, kSkipAnyCorruptedRecords) {
+  const int jstart = RecoveryTestHelper::kWALFileOffset;
+  const int jend = jstart + RecoveryTestHelper::kWALFilesCount;
+
+  for (auto trunc : {true, false}) {        /* Corruption style */
+    for (int i = 0; i < 4; i++) {           /* Corruption offset */
+      for (int j = jstart; j < jend; j++) { /* wal files */
+        // Fill data for testing
+        Options options = CurrentOptions();
+        const size_t row_count = RecoveryTestHelper::FillData(this, &options);
+
+        // Corrupt the WAL
+        RecoveryTestHelper::CorruptWAL(this, options, /*off=*/i * .3,
+                                       /*len%=*/.1, j, trunc);
+
+        // Verify behavior
+        options.wal_recovery_mode = WALRecoveryMode::kSkipAnyCorruptedRecords;
+        options.create_if_missing = false;
+        ASSERT_OK(TryReopen(options));
+
+        // Probe data for invariants
+        size_t recovered_row_count = RecoveryTestHelper::GetData(this);
+        ASSERT_LT(recovered_row_count, row_count);
+
+        if (!trunc) {
+          ASSERT_TRUE(i != 0 || recovered_row_count > 0);
+        }
+      }
+    }
+  }
+}
+
+#endif  // ROCKSDB_LITE
+
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -465,6 +465,12 @@ class InternalKeySliceTransform : public SliceTransform {
  const SliceTransform* const transform_;
 };

+// Read the key of a record from a write batch.
+// if this record represent the default column family then cf_record
+// must be passed as false, otherwise it must be passed as true.
+extern bool ReadKeyFromWriteBatchEntry(Slice* input, Slice* key,
+                                       bool cf_record);
+
 // Read record from a write batch piece from input.
 // tag, column_family, key, value and blob are return values. Callers own the
 // Slice they point to.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -92,6 +92,30 @@ TEST_F(FormatTest, InternalKeyShortSeparator) {
            Shorten(IKey("foo", 100, kTypeValue),
                    IKey("hello", 200, kTypeValue)));

+  ASSERT_EQ(IKey("ABC2", kMaxSequenceNumber, kValueTypeForSeek),
+            Shorten(IKey("ABC1AAAAA", 100, kTypeValue),
+                    IKey("ABC2ABB", 200, kTypeValue)));
+
+  ASSERT_EQ(IKey("AAA2", kMaxSequenceNumber, kValueTypeForSeek),
+            Shorten(IKey("AAA1AAA", 100, kTypeValue),
+                    IKey("AAA2AA", 200, kTypeValue)));
+
+  ASSERT_EQ(
+      IKey("AAA2", kMaxSequenceNumber, kValueTypeForSeek),
+      Shorten(IKey("AAA1AAA", 100, kTypeValue), IKey("AAA4", 200, kTypeValue)));
+
+  ASSERT_EQ(
+      IKey("AAA1B", kMaxSequenceNumber, kValueTypeForSeek),
+      Shorten(IKey("AAA1AAA", 100, kTypeValue), IKey("AAA2", 200, kTypeValue)));
+
+  ASSERT_EQ(IKey("AAA2", kMaxSequenceNumber, kValueTypeForSeek),
+            Shorten(IKey("AAA1AAA", 100, kTypeValue),
+                    IKey("AAA2A", 200, kTypeValue)));
+
+  ASSERT_EQ(
+      IKey("AAA1", 100, kTypeValue),
+      Shorten(IKey("AAA1", 100, kTypeValue), IKey("AAA2", 200, kTypeValue)));
+
  // When start user key is prefix of limit user key
  ASSERT_EQ(IKey("foo", 100, kTypeValue),
            Shorten(IKey("foo", 100, kTypeValue),
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -37,6 +37,7 @@ class DeleteFileTest : public testing::Test {
  DeleteFileTest() {
    db_ = nullptr;
    env_ = Env::Default();
+    options_.delete_obsolete_files_period_micros = 0;  // always do full purge
    options_.enable_thread_tracking = true;
    options_.write_buffer_size = 1024*1024*1000;
    options_.target_file_size_base = 1024*1024*1000;
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -20,57 +20,83 @@ void EventHelpers::AppendCurrentTime(JSONWriter* jwriter) {
                  std::chrono::system_clock::now().time_since_epoch()).count();
 }

-void EventHelpers::LogAndNotifyTableFileCreation(
+#ifndef ROCKSDB_LITE
+void EventHelpers::NotifyTableFileCreationStarted(
+    const std::vector<std::shared_ptr<EventListener>>& listeners,
+    const std::string& db_name, const std::string& cf_name,
+    const std::string& file_path, int job_id, TableFileCreationReason reason) {
+  TableFileCreationBriefInfo info;
+  info.db_name = db_name;
+  info.cf_name = cf_name;
+  info.file_path = file_path;
+  info.job_id = job_id;
+  info.reason = reason;
+  for (auto& listener : listeners) {
+    listener->OnTableFileCreationStarted(info);
+  }
+}
+#endif  // !ROCKSDB_LITE
+
+void EventHelpers::LogAndNotifyTableFileCreationFinished(
    EventLogger* event_logger,
    const std::vector<std::shared_ptr<EventListener>>& listeners,
-    const FileDescriptor& fd, const TableFileCreationInfo& info) {
-  assert(event_logger);
-  JSONWriter jwriter;
-  AppendCurrentTime(&jwriter);
-  jwriter << "cf_name" << info.cf_name
-          << "job" << info.job_id
-          << "event" << "table_file_creation"
-          << "file_number" << fd.GetNumber()
-          << "file_size" << fd.GetFileSize();
+    const std::string& db_name, const std::string& cf_name,
+    const std::string& file_path, int job_id, const FileDescriptor& fd,
+    const TableProperties& table_properties, TableFileCreationReason reason,
+    const Status& s) {
+  if (s.ok() && event_logger) {
+    JSONWriter jwriter;
+    AppendCurrentTime(&jwriter);
+    jwriter << "cf_name" << cf_name << "job" << job_id << "event"
+            << "table_file_creation"
+            << "file_number" << fd.GetNumber() << "file_size"
+            << fd.GetFileSize();

-  // table_properties
-  {
-    jwriter << "table_properties";
-    jwriter.StartObject();
+    // table_properties
+    {
+      jwriter << "table_properties";
+      jwriter.StartObject();

-    // basic properties:
-    jwriter << "data_size" << info.table_properties.data_size
-            << "index_size" << info.table_properties.index_size
-            << "filter_size" << info.table_properties.filter_size
-            << "raw_key_size" << info.table_properties.raw_key_size
-            << "raw_average_key_size" << SafeDivide(
-                info.table_properties.raw_key_size,
-                info.table_properties.num_entries)
-            << "raw_value_size" << info.table_properties.raw_value_size
-            << "raw_average_value_size" << SafeDivide(
-               info.table_properties.raw_value_size,
-               info.table_properties.num_entries)
-            << "num_data_blocks" << info.table_properties.num_data_blocks
-            << "num_entries" << info.table_properties.num_entries
-            << "filter_policy_name" <<
-                info.table_properties.filter_policy_name;
+      // basic properties:
+      jwriter << "data_size" << table_properties.data_size << "index_size"
+              << table_properties.index_size << "filter_size"
+              << table_properties.filter_size << "raw_key_size"
+              << table_properties.raw_key_size << "raw_average_key_size"
+              << SafeDivide(table_properties.raw_key_size,
+                            table_properties.num_entries)
+              << "raw_value_size" << table_properties.raw_value_size
+              << "raw_average_value_size"
+              << SafeDivide(table_properties.raw_value_size,
+                            table_properties.num_entries)
+              << "num_data_blocks" << table_properties.num_data_blocks
+              << "num_entries" << table_properties.num_entries
+              << "filter_policy_name" << table_properties.filter_policy_name;

-    // user collected properties
-    for (const auto& prop : info.table_properties.readable_properties) {
-      jwriter << prop.first << prop.second;
+      // user collected properties
+      for (const auto& prop : table_properties.readable_properties) {
+        jwriter << prop.first << prop.second;
+      }
+      jwriter.EndObject();
    }
    jwriter.EndObject();
-  }
-  jwriter.EndObject();

-  event_logger->Log(jwriter);
+    event_logger->Log(jwriter);
+  }

 #ifndef ROCKSDB_LITE
  if (listeners.size() == 0) {
    return;
  }
-
-  for (auto listener : listeners) {
+  TableFileCreationInfo info;
+  info.db_name = db_name;
+  info.cf_name = cf_name;
+  info.file_path = file_path;
+  info.file_size = fd.file_size;
+  info.job_id = job_id;
+  info.table_properties = table_properties;
+  info.reason = reason;
+  info.status = s;
+  for (auto& listener : listeners) {
    listener->OnTableFileCreated(info);
  }
 #endif  // !ROCKSDB_LITE
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -19,15 +19,30 @@ namespace rocksdb {
 class EventHelpers {
 public:
  static void AppendCurrentTime(JSONWriter* json_writer);
-  static void LogAndNotifyTableFileCreation(
+#ifndef ROCKSDB_LITE
+  static void NotifyTableFileCreationStarted(
+      const std::vector<std::shared_ptr<EventListener>>& listeners,
+      const std::string& db_name, const std::string& cf_name,
+      const std::string& file_path, int job_id, TableFileCreationReason reason);
+#endif  // !ROCKSDB_LITE
+  static void LogAndNotifyTableFileCreationFinished(
      EventLogger* event_logger,
      const std::vector<std::shared_ptr<EventListener>>& listeners,
-      const FileDescriptor& fd, const TableFileCreationInfo& info);
+      const std::string& db_name, const std::string& cf_name,
+      const std::string& file_path, int job_id, const FileDescriptor& fd,
+      const TableProperties& table_properties, TableFileCreationReason reason,
+      const Status& s);
  static void LogAndNotifyTableFileDeletion(
      EventLogger* event_logger, int job_id,
      uint64_t file_number, const std::string& file_path,
      const Status& status, const std::string& db_name,
      const std::vector<std::shared_ptr<EventListener>>& listeners);
+
+ private:
+  static void LogAndNotifyTableFileCreation(
+      EventLogger* event_logger,
+      const std::vector<std::shared_ptr<EventListener>>& listeners,
+      const FileDescriptor& fd, const TableFileCreationInfo& info);
 };

 }  // namespace rocksdb
@@ -1,4 +1,4 @@
-//  Copyright (c) 2014, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2015, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -66,7 +66,7 @@ FlushJob::FlushJob(const std::string& dbname, ColumnFamilyData* cfd,
                   JobContext* job_context, LogBuffer* log_buffer,
                   Directory* db_directory, Directory* output_file_directory,
                   CompressionType output_compression, Statistics* stats,
-                   EventLogger* event_logger)
+                   EventLogger* event_logger, bool measure_io_stats)
    : dbname_(dbname),
      cfd_(cfd),
      db_options_(db_options),
@@ -83,7 +83,8 @@ FlushJob::FlushJob(const std::string& dbname, ColumnFamilyData* cfd,
      output_file_directory_(output_file_directory),
      output_compression_(output_compression),
      stats_(stats),
-      event_logger_(event_logger) {
+      event_logger_(event_logger),
+      measure_io_stats_(measure_io_stats) {
  // Update the thread status to indicate flush.
  ReportStartedFlush();
  TEST_SYNC_POINT("FlushJob::FlushJob()");
@@ -94,7 +95,8 @@ FlushJob::~FlushJob() {
 }

 void FlushJob::ReportStartedFlush() {
-  ThreadStatusUtil::SetColumnFamily(cfd_);
+  ThreadStatusUtil::SetColumnFamily(cfd_, cfd_->ioptions()->env,
+                                    cfd_->options()->enable_thread_tracking);
  ThreadStatusUtil::SetThreadOperation(ThreadStatus::OP_FLUSH);
  ThreadStatusUtil::SetThreadOperationProperty(
      ThreadStatus::COMPACTION_JOB_ID,
@@ -113,13 +115,30 @@ void FlushJob::ReportFlushInputSize(const autovector<MemTable*>& mems) {
 }

 void FlushJob::RecordFlushIOStats() {
-  ThreadStatusUtil::SetThreadOperationProperty(
+  RecordTick(stats_, FLUSH_WRITE_BYTES, IOSTATS(bytes_written));
+  ThreadStatusUtil::IncreaseThreadOperationProperty(
      ThreadStatus::FLUSH_BYTES_WRITTEN, IOSTATS(bytes_written));
+  IOSTATS_RESET(bytes_written);
 }

 Status FlushJob::Run(FileMetaData* file_meta) {
  AutoThreadOperationStageUpdater stage_run(
      ThreadStatus::STAGE_FLUSH_RUN);
+  // I/O measurement variables
+  PerfLevel prev_perf_level = PerfLevel::kEnableTime;
+  uint64_t prev_write_nanos = 0;
+  uint64_t prev_fsync_nanos = 0;
+  uint64_t prev_range_sync_nanos = 0;
+  uint64_t prev_prepare_write_nanos = 0;
+  if (measure_io_stats_) {
+    prev_perf_level = GetPerfLevel();
+    SetPerfLevel(PerfLevel::kEnableTime);
+    prev_write_nanos = IOSTATS(write_nanos);
+    prev_fsync_nanos = IOSTATS(fsync_nanos);
+    prev_range_sync_nanos = IOSTATS(range_sync_nanos);
+    prev_prepare_write_nanos = IOSTATS(prepare_write_nanos);
+  }
+
  // Save the contents of the earliest memtable as a new Table
  FileMetaData meta;
  autovector<MemTable*> mems;
@@ -179,6 +198,18 @@ Status FlushJob::Run(FileMetaData* file_meta) {
  }
  stream.EndArray();

+  if (measure_io_stats_) {
+    if (prev_perf_level != PerfLevel::kEnableTime) {
+      SetPerfLevel(prev_perf_level);
+    }
+    stream << "file_write_nanos" << (IOSTATS(write_nanos) - prev_write_nanos);
+    stream << "file_range_sync_nanos"
+           << (IOSTATS(range_sync_nanos) - prev_range_sync_nanos);
+    stream << "file_fsync_nanos" << (IOSTATS(fsync_nanos) - prev_fsync_nanos);
+    stream << "file_prepare_write_nanos"
+           << (IOSTATS(prepare_write_nanos) - prev_prepare_write_nanos);
+  }
+
  return s;
 }

@@ -222,7 +253,6 @@ Status FlushJob::WriteLevel0Table(const autovector<MemTable*>& mems,
                         << total_num_deletes << "memory_usage"
                         << total_memory_usage;

-    TableFileCreationInfo info;
    {
      ScopedArenaIterator iter(
          NewMergingIterator(&cfd_->internal_comparator(), &memtables[0],
@@ -233,15 +263,16 @@ Status FlushJob::WriteLevel0Table(const autovector<MemTable*>& mems,

      TEST_SYNC_POINT_CALLBACK("FlushJob::WriteLevel0Table:output_compression",
                               &output_compression_);
-      s = BuildTable(dbname_, db_options_.env, *cfd_->ioptions(), env_options_,
-                     cfd_->table_cache(), iter.get(), meta,
-                     cfd_->internal_comparator(),
-                     cfd_->int_tbl_prop_collector_factories(), cfd_->GetID(),
-                     existing_snapshots_, earliest_write_conflict_snapshot_,
-                     output_compression_, cfd_->ioptions()->compression_opts,
-                     mutable_cf_options_.paranoid_file_checks,
-                     cfd_->internal_stats(), Env::IO_HIGH, &table_properties_);
-      info.table_properties = table_properties_;
+      s = BuildTable(
+          dbname_, db_options_.env, *cfd_->ioptions(), env_options_,
+          cfd_->table_cache(), iter.get(), meta, cfd_->internal_comparator(),
+          cfd_->int_tbl_prop_collector_factories(), cfd_->GetID(),
+          cfd_->GetName(), existing_snapshots_,
+          earliest_write_conflict_snapshot_, output_compression_,
+          cfd_->ioptions()->compression_opts,
+          mutable_cf_options_.paranoid_file_checks, cfd_->internal_stats(),
+          TableFileCreationReason::kFlush, event_logger_, job_context_->job_id,
+          Env::IO_HIGH, &table_properties_, 0 /* level */);
      LogFlush(db_options_.info_log);
    }
    Log(InfoLogLevel::INFO_LEVEL, db_options_.info_log,
@@ -252,24 +283,10 @@ Status FlushJob::WriteLevel0Table(const autovector<MemTable*>& mems,
        meta->fd.GetFileSize(), s.ToString().c_str(),
        meta->marked_for_compaction ? " (needs compaction)" : "");

-    // output to event logger
-    if (s.ok()) {
-      info.db_name = dbname_;
-      info.cf_name = cfd_->GetName();
-      info.file_path = TableFileName(db_options_.db_paths,
-                                     meta->fd.GetNumber(),
-                                     meta->fd.GetPathId());
-      info.file_size = meta->fd.GetFileSize();
-      info.job_id = job_context_->job_id;
-      EventHelpers::LogAndNotifyTableFileCreation(
-          event_logger_, db_options_.listeners,
-          meta->fd, info);
-      TEST_SYNC_POINT("FlushJob::LogAndNotifyTableFileCreation()");
-    }
-
    if (!db_options_.disableDataSync && output_file_directory_ != nullptr) {
      output_file_directory_->Fsync();
    }
+    TEST_SYNC_POINT("FlushJob::WriteLevel0Table");
    db_mutex_->Lock();
  }
  base->Unref();
@@ -291,13 +308,14 @@ Status FlushJob::WriteLevel0Table(const autovector<MemTable*>& mems,
                  meta->marked_for_compaction);
  }

+  // Note that here we treat flush as level 0 compaction in internal stats
  InternalStats::CompactionStats stats(1);
  stats.micros = db_options_.env->NowMicros() - start_micros;
  stats.bytes_written = meta->fd.GetFileSize();
  cfd_->internal_stats()->AddCompactionStats(0 /* level */, stats);
  cfd_->internal_stats()->AddCFStats(InternalStats::BYTES_FLUSHED,
                                     meta->fd.GetFileSize());
-  RecordTick(stats_, COMPACT_WRITE_BYTES, meta->fd.GetFileSize());
+  RecordFlushIOStats();
  return s;
 }

@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -62,7 +62,7 @@ class FlushJob {
           JobContext* job_context, LogBuffer* log_buffer,
           Directory* db_directory, Directory* output_file_directory,
           CompressionType output_compression, Statistics* stats,
-           EventLogger* event_logger);
+           EventLogger* event_logger, bool measure_io_stats);

  ~FlushJob();

@@ -93,6 +93,7 @@ class FlushJob {
  Statistics* stats_;
  EventLogger* event_logger_;
  TableProperties table_properties_;
+  bool measure_io_stats_;
 };

 }  // namespace rocksdb
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
@@ -93,7 +93,7 @@ TEST_F(FlushJobTest, Empty) {
                     db_options_, *cfd->GetLatestMutableCFOptions(),
                     env_options_, versions_.get(), &mutex_, &shutting_down_,
                     {}, kMaxSequenceNumber, &job_context, nullptr, nullptr,
-                     nullptr, kNoCompression, nullptr, &event_logger);
+                     nullptr, kNoCompression, nullptr, &event_logger, false);
  ASSERT_OK(flush_job.Run());
  job_context.Clean();
 }
@@ -132,7 +132,7 @@ TEST_F(FlushJobTest, NonEmpty) {
                     db_options_, *cfd->GetLatestMutableCFOptions(),
                     env_options_, versions_.get(), &mutex_, &shutting_down_,
                     {}, kMaxSequenceNumber, &job_context, nullptr, nullptr,
-                     nullptr, kNoCompression, nullptr, &event_logger);
+                     nullptr, kNoCompression, nullptr, &event_logger, true);
  FileMetaData fd;
  mutex_.Lock();
  ASSERT_OK(flush_job.Run(&fd));
@@ -192,11 +192,11 @@ TEST_F(FlushJobTest, Snapshots) {
  }

  EventLogger event_logger(db_options_.info_log.get());
-  FlushJob flush_job(dbname_, versions_->GetColumnFamilySet()->GetDefault(),
-                     db_options_, *cfd->GetLatestMutableCFOptions(),
-                     env_options_, versions_.get(), &mutex_, &shutting_down_,
-                     snapshots, kMaxSequenceNumber, &job_context, nullptr,
-                     nullptr, nullptr, kNoCompression, nullptr, &event_logger);
+  FlushJob flush_job(
+      dbname_, versions_->GetColumnFamilySet()->GetDefault(), db_options_,
+      *cfd->GetLatestMutableCFOptions(), env_options_, versions_.get(), &mutex_,
+      &shutting_down_, snapshots, kMaxSequenceNumber, &job_context, nullptr,
+      nullptr, nullptr, kNoCompression, nullptr, &event_logger, true);
  mutex_.Lock();
  ASSERT_OK(flush_job.Run());
  mutex_.Unlock();
@@ -1,4 +1,4 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.