Add a gflag for IO uring enable/disable (#8931)

Summary:
In case of IO uring bugs, we need to provide a way for users to turn it off.

Pull Request resolved: https://github.com/facebook/rocksdb/pull/8931

Test Plan: Manually run db_bench with/without the option and verify the behavior

Reviewed By: pdillinger

Differential Revision: D31040252

Pulled By: anand1976

fbshipit-source-id: 56f2537d6ac8488c9e126296d8190ad9e0158f70

diff --git a/HISTORY.md b/HISTORY.md
index 0cbf74ab4c48bc083fc4ddc87e3cf2a06eef9444..97596eb15476141861eba8d6babee80e75a928e2 100644 (file)
--- a/HISTORY.md
+++ b/HISTORY.md
@@ -12,6 +12,7 @@
 * Fix the implementation of `prepopulate_block_cache = kFlushOnly` to only apply to flushes rather than to all generated files.
 * Fix WAL log data corruption when using DBOptions.manual_wal_flush(true) and WriteOptions.sync(true) together. The sync WAL should work with locked log_write_mutex_.
 * Add checks for validity of the IO uring completion queue entries, and fail the BlockBasedTableReader MultiGet sub-batch if there's an invalid completion
+* Add an interface RocksDbIOUringEnable() that, if defined by the user, will allow them to enable/disable the use of IO uring by RocksDB
 
 ### New Features
 * RemoteCompaction's interface now includes `db_name`, `db_id`, `session_id`, which could help the user uniquely identify compaction job between db instances and sessions.

diff --git a/env/fs_posix.cc b/env/fs_posix.cc
index 6e967b2badc0a416c7419ea1a8b8b7b9b2ed7e3a..d321a746fe62bdf1ad21324b8e5053d7cdc2cbbb 100644 (file)
--- a/env/fs_posix.cc
+++ b/env/fs_posix.cc
@@ -73,6 +73,8 @@
 #define EXT4_SUPER_MAGIC 0xEF53
 #endif
 
+extern "C" bool RocksDbIOUringEnable() __attribute__((__weak__));
+
 namespace ROCKSDB_NAMESPACE {
 
 namespace {
@@ -267,7 +269,7 @@ class PosixFileSystem : public FileSystem {
           options
 #if defined(ROCKSDB_IOURING_PRESENT)
           ,
-          thread_local_io_urings_.get()
+          !IsIOUringEnabled() ? nullptr : thread_local_io_urings_.get()
 #endif
               ));
     }
@@ -1025,6 +1027,16 @@ class PosixFileSystem : public FileSystem {
 #endif
   }
 
+#ifdef ROCKSDB_IOURING_PRESENT
+  bool IsIOUringEnabled() {
+    if (RocksDbIOUringEnable && RocksDbIOUringEnable()) {
+      return true;
+    } else {
+      return false;
+    }
+  }
+#endif  // ROCKSDB_IOURING_PRESENT
+
 #if defined(ROCKSDB_IOURING_PRESENT)
   // io_uring instance
   std::unique_ptr<ThreadLocalPtr> thread_local_io_urings_;

diff --git a/tools/db_bench_tool.cc b/tools/db_bench_tool.cc
index 09848baf27ecd28a85cd479b8cf14a6b0bfd6272..0fe1c1e938561b833872631df098d46d507d51fd 100644 (file)
--- a/tools/db_bench_tool.cc
+++ b/tools/db_bench_tool.cc
@@ -1021,6 +1021,9 @@ DEFINE_string(block_cache_trace_file, "", "Block cache trace file path.");
 DEFINE_int32(trace_replay_threads, 1,
              "The number of threads to replay, must >=1.");
 
+DEFINE_bool(io_uring_enabled, true,
+            "If true, enable the use of IO uring if the platform supports it");
+extern "C" bool RocksDbIOUringEnable() { return FLAGS_io_uring_enabled; }
 #endif  // ROCKSDB_LITE
 
 static enum ROCKSDB_NAMESPACE::CompressionType StringToCompressionType(