AIO control blocks are frequently acquired and released because each aio request involves at least one AIOCB. Therefore, we pool them to avoid heap allocation overhead.
The problem with the freelist approach in AIOPool is thread-safety. If we want BlockDriverStates to associate with AioContexts that execute in multiple threads, then a global freelist becomes a problem. This patch drops the freelist and instead uses g_slice_alloc() which is tuned for per-thread fixed-size object pools. qemu_aio_get() and qemu_aio_release() are now thread-safe. Note that the change from g_malloc0() to g_slice_alloc() should be safe since the freelist reuse case doesn't zero the AIOCB either. v2: * Split into 3 patches (I still took the liberty of combining the AIOPool -> AIOCBInfo rename with constification because I didn't want to touch all those files twice) [Paolo] Stefan Hajnoczi (3): aio: switch aiocb_size type int -> size_t aio: use g_slice_alloc() for AIOCB pooling aio: rename AIOPool to AIOCBInfo block.c | 31 ++++++++++++------------------- block/blkdebug.c | 4 ++-- block/blkverify.c | 4 ++-- block/curl.c | 4 ++-- block/gluster.c | 6 +++--- block/iscsi.c | 12 ++++++------ block/linux-aio.c | 4 ++-- block/qed.c | 4 ++-- block/rbd.c | 4 ++-- block/sheepdog.c | 4 ++-- block/win32-aio.c | 4 ++-- dma-helpers.c | 4 ++-- hw/ide/core.c | 4 ++-- qemu-aio.h | 12 +++++------- thread-pool.c | 4 ++-- 15 files changed, 48 insertions(+), 57 deletions(-) -- 1.7.11.7
