* [Qemu-devel] [PATCH v2 0/3] aio: use g_slice_alloc() for AIOCB pooling
@ 2012-10-31 15:34 Stefan Hajnoczi
2012-10-31 15:34 ` [Qemu-devel] [PATCH v2 1/3] aio: switch aiocb_size type int -> size_t Stefan Hajnoczi
` (3 more replies)
0 siblings, 4 replies; 11+ messages in thread
From: Stefan Hajnoczi @ 2012-10-31 15:34 UTC (permalink / raw)
To: qemu-devel; +Cc: Kevin Wolf, Paolo Bonzini, Anthony Liguori, Stefan Hajnoczi
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
^ permalink raw reply [flat|nested] 11+ messages in thread* [Qemu-devel] [PATCH v2 1/3] aio: switch aiocb_size type int -> size_t 2012-10-31 15:34 [Qemu-devel] [PATCH v2 0/3] aio: use g_slice_alloc() for AIOCB pooling Stefan Hajnoczi @ 2012-10-31 15:34 ` Stefan Hajnoczi 2012-11-02 14:49 ` Paolo Bonzini 2012-10-31 15:34 ` [Qemu-devel] [PATCH v2 2/3] aio: use g_slice_alloc() for AIOCB pooling Stefan Hajnoczi ` (2 subsequent siblings) 3 siblings, 1 reply; 11+ messages in thread From: Stefan Hajnoczi @ 2012-10-31 15:34 UTC (permalink / raw) To: qemu-devel; +Cc: Kevin Wolf, Paolo Bonzini, Anthony Liguori, Stefan Hajnoczi Using appropriate types for variables is a good thing :). All users simply do sizeof(MyType) and the value is passed to a memory allocator, it should be size_t. Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> --- qemu-aio.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/qemu-aio.h b/qemu-aio.h index 1b7eb6e..111b0b3 100644 --- a/qemu-aio.h +++ b/qemu-aio.h @@ -23,7 +23,7 @@ typedef void BlockDriverCompletionFunc(void *opaque, int ret); typedef struct AIOPool { void (*cancel)(BlockDriverAIOCB *acb); - int aiocb_size; + size_t aiocb_size; BlockDriverAIOCB *free_aiocb; } AIOPool; -- 1.7.11.7 ^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [Qemu-devel] [PATCH v2 1/3] aio: switch aiocb_size type int -> size_t 2012-10-31 15:34 ` [Qemu-devel] [PATCH v2 1/3] aio: switch aiocb_size type int -> size_t Stefan Hajnoczi @ 2012-11-02 14:49 ` Paolo Bonzini 0 siblings, 0 replies; 11+ messages in thread From: Paolo Bonzini @ 2012-11-02 14:49 UTC (permalink / raw) To: Stefan Hajnoczi; +Cc: Kevin Wolf, Anthony Liguori, qemu-devel Il 31/10/2012 16:34, Stefan Hajnoczi ha scritto: > Using appropriate types for variables is a good thing :). All users > simply do sizeof(MyType) and the value is passed to a memory allocator, > it should be size_t. > > Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> > --- > qemu-aio.h | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/qemu-aio.h b/qemu-aio.h > index 1b7eb6e..111b0b3 100644 > --- a/qemu-aio.h > +++ b/qemu-aio.h > @@ -23,7 +23,7 @@ typedef void BlockDriverCompletionFunc(void *opaque, int ret); > > typedef struct AIOPool { > void (*cancel)(BlockDriverAIOCB *acb); > - int aiocb_size; > + size_t aiocb_size; > BlockDriverAIOCB *free_aiocb; > } AIOPool; > > Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> ^ permalink raw reply [flat|nested] 11+ messages in thread
* [Qemu-devel] [PATCH v2 2/3] aio: use g_slice_alloc() for AIOCB pooling 2012-10-31 15:34 [Qemu-devel] [PATCH v2 0/3] aio: use g_slice_alloc() for AIOCB pooling Stefan Hajnoczi 2012-10-31 15:34 ` [Qemu-devel] [PATCH v2 1/3] aio: switch aiocb_size type int -> size_t Stefan Hajnoczi @ 2012-10-31 15:34 ` Stefan Hajnoczi 2012-11-02 14:49 ` Paolo Bonzini 2012-11-12 10:34 ` Kevin Wolf 2012-10-31 15:34 ` [Qemu-devel] [PATCH v2 3/3] aio: rename AIOPool to AIOCBInfo Stefan Hajnoczi 2012-11-12 10:34 ` [Qemu-devel] [PATCH v2 0/3] aio: use g_slice_alloc() for AIOCB pooling Kevin Wolf 3 siblings, 2 replies; 11+ messages in thread From: Stefan Hajnoczi @ 2012-10-31 15:34 UTC (permalink / raw) To: qemu-devel; +Cc: Kevin Wolf, Paolo Bonzini, Anthony Liguori, Stefan Hajnoczi 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. Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> --- block.c | 15 ++++----------- qemu-aio.h | 2 -- 2 files changed, 4 insertions(+), 13 deletions(-) diff --git a/block.c b/block.c index da1fdca..ea0f7d8 100644 --- a/block.c +++ b/block.c @@ -3909,13 +3909,8 @@ void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs, { BlockDriverAIOCB *acb; - if (pool->free_aiocb) { - acb = pool->free_aiocb; - pool->free_aiocb = acb->next; - } else { - acb = g_malloc0(pool->aiocb_size); - acb->pool = pool; - } + acb = g_slice_alloc(pool->aiocb_size); + acb->pool = pool; acb->bs = bs; acb->cb = cb; acb->opaque = opaque; @@ -3924,10 +3919,8 @@ void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs, void qemu_aio_release(void *p) { - BlockDriverAIOCB *acb = (BlockDriverAIOCB *)p; - AIOPool *pool = acb->pool; - acb->next = pool->free_aiocb; - pool->free_aiocb = acb; + BlockDriverAIOCB *acb = p; + g_slice_free1(acb->pool->aiocb_size, acb); } /**************************************************************/ diff --git a/qemu-aio.h b/qemu-aio.h index 111b0b3..b29c509 100644 --- a/qemu-aio.h +++ b/qemu-aio.h @@ -24,7 +24,6 @@ typedef void BlockDriverCompletionFunc(void *opaque, int ret); typedef struct AIOPool { void (*cancel)(BlockDriverAIOCB *acb); size_t aiocb_size; - BlockDriverAIOCB *free_aiocb; } AIOPool; struct BlockDriverAIOCB { @@ -32,7 +31,6 @@ struct BlockDriverAIOCB { BlockDriverState *bs; BlockDriverCompletionFunc *cb; void *opaque; - BlockDriverAIOCB *next; }; void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs, -- 1.7.11.7 ^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [Qemu-devel] [PATCH v2 2/3] aio: use g_slice_alloc() for AIOCB pooling 2012-10-31 15:34 ` [Qemu-devel] [PATCH v2 2/3] aio: use g_slice_alloc() for AIOCB pooling Stefan Hajnoczi @ 2012-11-02 14:49 ` Paolo Bonzini 2012-11-12 10:34 ` Kevin Wolf 1 sibling, 0 replies; 11+ messages in thread From: Paolo Bonzini @ 2012-11-02 14:49 UTC (permalink / raw) To: Stefan Hajnoczi; +Cc: Kevin Wolf, Anthony Liguori, qemu-devel Il 31/10/2012 16:34, Stefan Hajnoczi ha scritto: > 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. > > Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> > --- > block.c | 15 ++++----------- > qemu-aio.h | 2 -- > 2 files changed, 4 insertions(+), 13 deletions(-) > > diff --git a/block.c b/block.c > index da1fdca..ea0f7d8 100644 > --- a/block.c > +++ b/block.c > @@ -3909,13 +3909,8 @@ void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs, > { > BlockDriverAIOCB *acb; > > - if (pool->free_aiocb) { > - acb = pool->free_aiocb; > - pool->free_aiocb = acb->next; > - } else { > - acb = g_malloc0(pool->aiocb_size); > - acb->pool = pool; > - } > + acb = g_slice_alloc(pool->aiocb_size); > + acb->pool = pool; > acb->bs = bs; > acb->cb = cb; > acb->opaque = opaque; > @@ -3924,10 +3919,8 @@ void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs, > > void qemu_aio_release(void *p) > { > - BlockDriverAIOCB *acb = (BlockDriverAIOCB *)p; > - AIOPool *pool = acb->pool; > - acb->next = pool->free_aiocb; > - pool->free_aiocb = acb; > + BlockDriverAIOCB *acb = p; > + g_slice_free1(acb->pool->aiocb_size, acb); > } > > /**************************************************************/ > diff --git a/qemu-aio.h b/qemu-aio.h > index 111b0b3..b29c509 100644 > --- a/qemu-aio.h > +++ b/qemu-aio.h > @@ -24,7 +24,6 @@ typedef void BlockDriverCompletionFunc(void *opaque, int ret); > typedef struct AIOPool { > void (*cancel)(BlockDriverAIOCB *acb); > size_t aiocb_size; > - BlockDriverAIOCB *free_aiocb; > } AIOPool; > > struct BlockDriverAIOCB { > @@ -32,7 +31,6 @@ struct BlockDriverAIOCB { > BlockDriverState *bs; > BlockDriverCompletionFunc *cb; > void *opaque; > - BlockDriverAIOCB *next; > }; > > void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs, > Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> ^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [Qemu-devel] [PATCH v2 2/3] aio: use g_slice_alloc() for AIOCB pooling 2012-10-31 15:34 ` [Qemu-devel] [PATCH v2 2/3] aio: use g_slice_alloc() for AIOCB pooling Stefan Hajnoczi 2012-11-02 14:49 ` Paolo Bonzini @ 2012-11-12 10:34 ` Kevin Wolf 2012-11-12 10:37 ` Paolo Bonzini 1 sibling, 1 reply; 11+ messages in thread From: Kevin Wolf @ 2012-11-12 10:34 UTC (permalink / raw) To: Stefan Hajnoczi; +Cc: Paolo Bonzini, qemu-devel Am 31.10.2012 16:34, schrieb Stefan Hajnoczi: > 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. Of course the real reason is that all fields are set anyway. We could even express this fact more clearly in the code: diff --git a/block.c b/block.c index 854ebd6..6f5a0e7 100644 --- a/block.c +++ b/block.c @@ -3910,10 +3910,13 @@ void *qemu_aio_get(const AIOCBInfo *aiocb_info, BlockDriverState *bs, BlockDriverAIOCB *acb; acb = g_slice_alloc(aiocb_info->aiocb_size); - acb->aiocb_info = aiocb_info; - acb->bs = bs; - acb->cb = cb; - acb->opaque = opaque; + *acb = (BlockDriverAIOCB) { + .aiocb_info = aiocb_info, + .bs = bs, + .cb = cb, + .opaque = opaque, + }; + return acb; } If you agree, I'll commit this on top of your series. Kevin ^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [Qemu-devel] [PATCH v2 2/3] aio: use g_slice_alloc() for AIOCB pooling 2012-11-12 10:34 ` Kevin Wolf @ 2012-11-12 10:37 ` Paolo Bonzini 2012-11-12 10:42 ` Kevin Wolf 0 siblings, 1 reply; 11+ messages in thread From: Paolo Bonzini @ 2012-11-12 10:37 UTC (permalink / raw) To: Kevin Wolf; +Cc: qemu-devel, Stefan Hajnoczi Il 12/11/2012 11:34, Kevin Wolf ha scritto: >> 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. > > Of course the real reason is that all fields are set anyway. This doesn't necessarily apply for "subclasses", though. (It does for the reason Stefan mentioned, but it is not possible to make it apparent in the code). Paolo > diff --git a/block.c b/block.c > index 854ebd6..6f5a0e7 100644 > --- a/block.c > +++ b/block.c > @@ -3910,10 +3910,13 @@ void *qemu_aio_get(const AIOCBInfo *aiocb_info, > BlockDriverState *bs, > BlockDriverAIOCB *acb; > > acb = g_slice_alloc(aiocb_info->aiocb_size); > - acb->aiocb_info = aiocb_info; > - acb->bs = bs; > - acb->cb = cb; > - acb->opaque = opaque; > + *acb = (BlockDriverAIOCB) { > + .aiocb_info = aiocb_info, > + .bs = bs, > + .cb = cb, > + .opaque = opaque, > + }; > + > return acb; > } > > If you agree, I'll commit this on top of your series. > > Kevin > ^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [Qemu-devel] [PATCH v2 2/3] aio: use g_slice_alloc() for AIOCB pooling 2012-11-12 10:37 ` Paolo Bonzini @ 2012-11-12 10:42 ` Kevin Wolf 0 siblings, 0 replies; 11+ messages in thread From: Kevin Wolf @ 2012-11-12 10:42 UTC (permalink / raw) To: Paolo Bonzini; +Cc: qemu-devel, Stefan Hajnoczi Am 12.11.2012 11:37, schrieb Paolo Bonzini: > Il 12/11/2012 11:34, Kevin Wolf ha scritto: >>> 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. >> >> Of course the real reason is that all fields are set anyway. > > This doesn't necessarily apply for "subclasses", though. (It does for > the reason Stefan mentioned, but it is not possible to make it apparent > in the code). See? I really shouldn't review code any more. :-) Kevin ^ permalink raw reply [flat|nested] 11+ messages in thread
* [Qemu-devel] [PATCH v2 3/3] aio: rename AIOPool to AIOCBInfo 2012-10-31 15:34 [Qemu-devel] [PATCH v2 0/3] aio: use g_slice_alloc() for AIOCB pooling Stefan Hajnoczi 2012-10-31 15:34 ` [Qemu-devel] [PATCH v2 1/3] aio: switch aiocb_size type int -> size_t Stefan Hajnoczi 2012-10-31 15:34 ` [Qemu-devel] [PATCH v2 2/3] aio: use g_slice_alloc() for AIOCB pooling Stefan Hajnoczi @ 2012-10-31 15:34 ` Stefan Hajnoczi 2012-11-02 14:50 ` Paolo Bonzini 2012-11-12 10:34 ` [Qemu-devel] [PATCH v2 0/3] aio: use g_slice_alloc() for AIOCB pooling Kevin Wolf 3 siblings, 1 reply; 11+ messages in thread From: Stefan Hajnoczi @ 2012-10-31 15:34 UTC (permalink / raw) To: qemu-devel; +Cc: Kevin Wolf, Paolo Bonzini, Anthony Liguori, Stefan Hajnoczi Now that AIOPool no longer keeps a freelist, it isn't really a "pool" anymore. Rename it to AIOCBInfo and make it const since it no longer needs to be modified. Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> --- block.c | 22 +++++++++++----------- 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 | 8 ++++---- thread-pool.c | 4 ++-- 15 files changed, 46 insertions(+), 46 deletions(-) diff --git a/block.c b/block.c index ea0f7d8..854ebd6 100644 --- a/block.c +++ b/block.c @@ -3521,7 +3521,7 @@ int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs) void bdrv_aio_cancel(BlockDriverAIOCB *acb) { - acb->pool->cancel(acb); + acb->aiocb_info->cancel(acb); } /* block I/O throttling */ @@ -3711,7 +3711,7 @@ static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb) qemu_aio_release(acb); } -static AIOPool bdrv_em_aio_pool = { +static const AIOCBInfo bdrv_em_aiocb_info = { .aiocb_size = sizeof(BlockDriverAIOCBSync), .cancel = bdrv_aio_cancel_em, }; @@ -3740,7 +3740,7 @@ static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs, { BlockDriverAIOCBSync *acb; - acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque); + acb = qemu_aio_get(&bdrv_em_aiocb_info, bs, cb, opaque); acb->is_write = is_write; acb->qiov = qiov; acb->bounce = qemu_blockalign(bs, qiov->size); @@ -3785,7 +3785,7 @@ static void bdrv_aio_co_cancel_em(BlockDriverAIOCB *blockacb) qemu_aio_flush(); } -static AIOPool bdrv_em_co_aio_pool = { +static const AIOCBInfo bdrv_em_co_aiocb_info = { .aiocb_size = sizeof(BlockDriverAIOCBCoroutine), .cancel = bdrv_aio_co_cancel_em, }; @@ -3828,7 +3828,7 @@ static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, Coroutine *co; BlockDriverAIOCBCoroutine *acb; - acb = qemu_aio_get(&bdrv_em_co_aio_pool, bs, cb, opaque); + acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); acb->req.sector = sector_num; acb->req.nb_sectors = nb_sectors; acb->req.qiov = qiov; @@ -3858,7 +3858,7 @@ BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs, Coroutine *co; BlockDriverAIOCBCoroutine *acb; - acb = qemu_aio_get(&bdrv_em_co_aio_pool, bs, cb, opaque); + acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); co = qemu_coroutine_create(bdrv_aio_flush_co_entry); qemu_coroutine_enter(co, acb); @@ -3884,7 +3884,7 @@ BlockDriverAIOCB *bdrv_aio_discard(BlockDriverState *bs, trace_bdrv_aio_discard(bs, sector_num, nb_sectors, opaque); - acb = qemu_aio_get(&bdrv_em_co_aio_pool, bs, cb, opaque); + acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); acb->req.sector = sector_num; acb->req.nb_sectors = nb_sectors; co = qemu_coroutine_create(bdrv_aio_discard_co_entry); @@ -3904,13 +3904,13 @@ void bdrv_init_with_whitelist(void) bdrv_init(); } -void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs, +void *qemu_aio_get(const AIOCBInfo *aiocb_info, BlockDriverState *bs, BlockDriverCompletionFunc *cb, void *opaque) { BlockDriverAIOCB *acb; - acb = g_slice_alloc(pool->aiocb_size); - acb->pool = pool; + acb = g_slice_alloc(aiocb_info->aiocb_size); + acb->aiocb_info = aiocb_info; acb->bs = bs; acb->cb = cb; acb->opaque = opaque; @@ -3920,7 +3920,7 @@ void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs, void qemu_aio_release(void *p) { BlockDriverAIOCB *acb = p; - g_slice_free1(acb->pool->aiocb_size, acb); + g_slice_free1(acb->aiocb_info->aiocb_size, acb); } /**************************************************************/ diff --git a/block/blkdebug.c b/block/blkdebug.c index 1206d52..d61ece8 100644 --- a/block/blkdebug.c +++ b/block/blkdebug.c @@ -41,7 +41,7 @@ typedef struct BlkdebugAIOCB { static void blkdebug_aio_cancel(BlockDriverAIOCB *blockacb); -static AIOPool blkdebug_aio_pool = { +static const AIOCBInfo blkdebug_aiocb_info = { .aiocb_size = sizeof(BlkdebugAIOCB), .cancel = blkdebug_aio_cancel, }; @@ -335,7 +335,7 @@ static BlockDriverAIOCB *inject_error(BlockDriverState *bs, return NULL; } - acb = qemu_aio_get(&blkdebug_aio_pool, bs, cb, opaque); + acb = qemu_aio_get(&blkdebug_aiocb_info, bs, cb, opaque); acb->ret = -error; bh = qemu_bh_new(error_callback_bh, acb); diff --git a/block/blkverify.c b/block/blkverify.c index 9d5f1ec..4beede7 100644 --- a/block/blkverify.c +++ b/block/blkverify.c @@ -48,7 +48,7 @@ static void blkverify_aio_cancel(BlockDriverAIOCB *blockacb) } } -static AIOPool blkverify_aio_pool = { +static const AIOCBInfo blkverify_aiocb_info = { .aiocb_size = sizeof(BlkverifyAIOCB), .cancel = blkverify_aio_cancel, }; @@ -233,7 +233,7 @@ static BlkverifyAIOCB *blkverify_aio_get(BlockDriverState *bs, bool is_write, BlockDriverCompletionFunc *cb, void *opaque) { - BlkverifyAIOCB *acb = qemu_aio_get(&blkverify_aio_pool, bs, cb, opaque); + BlkverifyAIOCB *acb = qemu_aio_get(&blkverify_aiocb_info, bs, cb, opaque); acb->bh = NULL; acb->is_write = is_write; diff --git a/block/curl.c b/block/curl.c index c1074cd..1179484 100644 --- a/block/curl.c +++ b/block/curl.c @@ -438,7 +438,7 @@ static void curl_aio_cancel(BlockDriverAIOCB *blockacb) // Do we have to implement canceling? Seems to work without... } -static AIOPool curl_aio_pool = { +static const AIOCBInfo curl_aiocb_info = { .aiocb_size = sizeof(CURLAIOCB), .cancel = curl_aio_cancel, }; @@ -505,7 +505,7 @@ static BlockDriverAIOCB *curl_aio_readv(BlockDriverState *bs, { CURLAIOCB *acb; - acb = qemu_aio_get(&curl_aio_pool, bs, cb, opaque); + acb = qemu_aio_get(&curl_aiocb_info, bs, cb, opaque); acb->qiov = qiov; acb->sector_num = sector_num; diff --git a/block/gluster.c b/block/gluster.c index 3588d73..1c90174 100644 --- a/block/gluster.c +++ b/block/gluster.c @@ -388,7 +388,7 @@ static void qemu_gluster_aio_cancel(BlockDriverAIOCB *blockacb) } } -static AIOPool gluster_aio_pool = { +static const AIOCBInfo gluster_aiocb_info = { .aiocb_size = sizeof(GlusterAIOCB), .cancel = qemu_gluster_aio_cancel, }; @@ -439,7 +439,7 @@ static BlockDriverAIOCB *qemu_gluster_aio_rw(BlockDriverState *bs, size = nb_sectors * BDRV_SECTOR_SIZE; s->qemu_aio_count++; - acb = qemu_aio_get(&gluster_aio_pool, bs, cb, opaque); + acb = qemu_aio_get(&gluster_aiocb_info, bs, cb, opaque); acb->size = size; acb->ret = 0; acb->finished = NULL; @@ -484,7 +484,7 @@ static BlockDriverAIOCB *qemu_gluster_aio_flush(BlockDriverState *bs, GlusterAIOCB *acb; BDRVGlusterState *s = bs->opaque; - acb = qemu_aio_get(&gluster_aio_pool, bs, cb, opaque); + acb = qemu_aio_get(&gluster_aiocb_info, bs, cb, opaque); acb->size = 0; acb->ret = 0; acb->finished = NULL; diff --git a/block/iscsi.c b/block/iscsi.c index d0b1a10..a6a819d 100644 --- a/block/iscsi.c +++ b/block/iscsi.c @@ -133,7 +133,7 @@ iscsi_aio_cancel(BlockDriverAIOCB *blockacb) } } -static AIOPool iscsi_aio_pool = { +static const AIOCBInfo iscsi_aiocb_info = { .aiocb_size = sizeof(IscsiAIOCB), .cancel = iscsi_aio_cancel, }; @@ -234,7 +234,7 @@ iscsi_aio_writev(BlockDriverState *bs, int64_t sector_num, uint64_t lba; struct iscsi_data data; - acb = qemu_aio_get(&iscsi_aio_pool, bs, cb, opaque); + acb = qemu_aio_get(&iscsi_aiocb_info, bs, cb, opaque); trace_iscsi_aio_writev(iscsi, sector_num, nb_sectors, opaque, acb); acb->iscsilun = iscsilun; @@ -325,7 +325,7 @@ iscsi_aio_readv(BlockDriverState *bs, int64_t sector_num, qemu_read_size = BDRV_SECTOR_SIZE * (size_t)nb_sectors; - acb = qemu_aio_get(&iscsi_aio_pool, bs, cb, opaque); + acb = qemu_aio_get(&iscsi_aiocb_info, bs, cb, opaque); trace_iscsi_aio_readv(iscsi, sector_num, nb_sectors, opaque, acb); acb->iscsilun = iscsilun; @@ -430,7 +430,7 @@ iscsi_aio_flush(BlockDriverState *bs, struct iscsi_context *iscsi = iscsilun->iscsi; IscsiAIOCB *acb; - acb = qemu_aio_get(&iscsi_aio_pool, bs, cb, opaque); + acb = qemu_aio_get(&iscsi_aiocb_info, bs, cb, opaque); acb->iscsilun = iscsilun; acb->canceled = 0; @@ -483,7 +483,7 @@ iscsi_aio_discard(BlockDriverState *bs, IscsiAIOCB *acb; struct unmap_list list[1]; - acb = qemu_aio_get(&iscsi_aio_pool, bs, cb, opaque); + acb = qemu_aio_get(&iscsi_aiocb_info, bs, cb, opaque); acb->iscsilun = iscsilun; acb->canceled = 0; @@ -558,7 +558,7 @@ static BlockDriverAIOCB *iscsi_aio_ioctl(BlockDriverState *bs, assert(req == SG_IO); - acb = qemu_aio_get(&iscsi_aio_pool, bs, cb, opaque); + acb = qemu_aio_get(&iscsi_aiocb_info, bs, cb, opaque); acb->iscsilun = iscsilun; acb->canceled = 0; diff --git a/block/linux-aio.c b/block/linux-aio.c index 6ca984d..91ef863 100644 --- a/block/linux-aio.c +++ b/block/linux-aio.c @@ -140,7 +140,7 @@ static void laio_cancel(BlockDriverAIOCB *blockacb) } } -static AIOPool laio_pool = { +static const AIOCBInfo laio_aiocb_info = { .aiocb_size = sizeof(struct qemu_laiocb), .cancel = laio_cancel, }; @@ -154,7 +154,7 @@ BlockDriverAIOCB *laio_submit(BlockDriverState *bs, void *aio_ctx, int fd, struct iocb *iocbs; off_t offset = sector_num * 512; - laiocb = qemu_aio_get(&laio_pool, bs, cb, opaque); + laiocb = qemu_aio_get(&laio_aiocb_info, bs, cb, opaque); laiocb->nbytes = nb_sectors * 512; laiocb->ctx = s; laiocb->ret = -EINPROGRESS; diff --git a/block/qed.c b/block/qed.c index 6c182ca..0b5374a 100644 --- a/block/qed.c +++ b/block/qed.c @@ -30,7 +30,7 @@ static void qed_aio_cancel(BlockDriverAIOCB *blockacb) } } -static AIOPool qed_aio_pool = { +static const AIOCBInfo qed_aiocb_info = { .aiocb_size = sizeof(QEDAIOCB), .cancel = qed_aio_cancel, }; @@ -1311,7 +1311,7 @@ static BlockDriverAIOCB *qed_aio_setup(BlockDriverState *bs, BlockDriverCompletionFunc *cb, void *opaque, int flags) { - QEDAIOCB *acb = qemu_aio_get(&qed_aio_pool, bs, cb, opaque); + QEDAIOCB *acb = qemu_aio_get(&qed_aiocb_info, bs, cb, opaque); trace_qed_aio_setup(bs->opaque, acb, sector_num, nb_sectors, opaque, flags); diff --git a/block/rbd.c b/block/rbd.c index 015a9db..0aaacaf 100644 --- a/block/rbd.c +++ b/block/rbd.c @@ -570,7 +570,7 @@ static void qemu_rbd_aio_cancel(BlockDriverAIOCB *blockacb) acb->cancelled = 1; } -static AIOPool rbd_aio_pool = { +static const AIOCBInfo rbd_aiocb_info = { .aiocb_size = sizeof(RBDAIOCB), .cancel = qemu_rbd_aio_cancel, }; @@ -672,7 +672,7 @@ static BlockDriverAIOCB *rbd_start_aio(BlockDriverState *bs, BDRVRBDState *s = bs->opaque; - acb = qemu_aio_get(&rbd_aio_pool, bs, cb, opaque); + acb = qemu_aio_get(&rbd_aiocb_info, bs, cb, opaque); acb->cmd = cmd; acb->qiov = qiov; if (cmd == RBD_AIO_DISCARD) { diff --git a/block/sheepdog.c b/block/sheepdog.c index 9306174..a48f58c 100644 --- a/block/sheepdog.c +++ b/block/sheepdog.c @@ -420,7 +420,7 @@ static void sd_aio_cancel(BlockDriverAIOCB *blockacb) acb->canceled = true; } -static AIOPool sd_aio_pool = { +static const AIOCBInfo sd_aiocb_info = { .aiocb_size = sizeof(SheepdogAIOCB), .cancel = sd_aio_cancel, }; @@ -431,7 +431,7 @@ static SheepdogAIOCB *sd_aio_setup(BlockDriverState *bs, QEMUIOVector *qiov, { SheepdogAIOCB *acb; - acb = qemu_aio_get(&sd_aio_pool, bs, cb, opaque); + acb = qemu_aio_get(&sd_aiocb_info, bs, cb, opaque); acb->qiov = qiov; diff --git a/block/win32-aio.c b/block/win32-aio.c index c34dc73..444c228 100644 --- a/block/win32-aio.c +++ b/block/win32-aio.c @@ -131,7 +131,7 @@ static void win32_aio_cancel(BlockDriverAIOCB *blockacb) } } -static AIOPool win32_aio_pool = { +static const AIOCBInfo win32_aiocb_info = { .aiocb_size = sizeof(QEMUWin32AIOCB), .cancel = win32_aio_cancel, }; @@ -145,7 +145,7 @@ BlockDriverAIOCB *win32_aio_submit(BlockDriverState *bs, uint64_t offset = sector_num * 512; DWORD rc; - waiocb = qemu_aio_get(&win32_aio_pool, bs, cb, opaque); + waiocb = qemu_aio_get(&win32_aiocb_info, bs, cb, opaque); waiocb->nbytes = nb_sectors * 512; waiocb->qiov = qiov; waiocb->is_read = (type == QEMU_AIO_READ); diff --git a/dma-helpers.c b/dma-helpers.c index 0c18e9e..4f5fb64 100644 --- a/dma-helpers.c +++ b/dma-helpers.c @@ -195,7 +195,7 @@ static void dma_aio_cancel(BlockDriverAIOCB *acb) dma_complete(dbs, 0); } -static AIOPool dma_aio_pool = { +static const AIOCBInfo dma_aiocb_info = { .aiocb_size = sizeof(DMAAIOCB), .cancel = dma_aio_cancel, }; @@ -205,7 +205,7 @@ BlockDriverAIOCB *dma_bdrv_io( DMAIOFunc *io_func, BlockDriverCompletionFunc *cb, void *opaque, DMADirection dir) { - DMAAIOCB *dbs = qemu_aio_get(&dma_aio_pool, bs, cb, opaque); + DMAAIOCB *dbs = qemu_aio_get(&dma_aiocb_info, bs, cb, opaque); trace_dma_bdrv_io(dbs, bs, sector_num, (dir == DMA_DIRECTION_TO_DEVICE)); diff --git a/hw/ide/core.c b/hw/ide/core.c index d683a8c..7d6b0fa 100644 --- a/hw/ide/core.c +++ b/hw/ide/core.c @@ -336,7 +336,7 @@ static void trim_aio_cancel(BlockDriverAIOCB *acb) qemu_aio_release(iocb); } -static AIOPool trim_aio_pool = { +static const AIOCBInfo trim_aiocb_info = { .aiocb_size = sizeof(TrimAIOCB), .cancel = trim_aio_cancel, }; @@ -360,7 +360,7 @@ BlockDriverAIOCB *ide_issue_trim(BlockDriverState *bs, TrimAIOCB *iocb; int i, j, ret; - iocb = qemu_aio_get(&trim_aio_pool, bs, cb, opaque); + iocb = qemu_aio_get(&trim_aiocb_info, bs, cb, opaque); iocb->bh = qemu_bh_new(ide_trim_bh_cb, iocb); iocb->ret = 0; diff --git a/qemu-aio.h b/qemu-aio.h index b29c509..3889fe9 100644 --- a/qemu-aio.h +++ b/qemu-aio.h @@ -21,19 +21,19 @@ typedef struct BlockDriverAIOCB BlockDriverAIOCB; typedef void BlockDriverCompletionFunc(void *opaque, int ret); -typedef struct AIOPool { +typedef struct AIOCBInfo { void (*cancel)(BlockDriverAIOCB *acb); size_t aiocb_size; -} AIOPool; +} AIOCBInfo; struct BlockDriverAIOCB { - AIOPool *pool; + const AIOCBInfo *aiocb_info; BlockDriverState *bs; BlockDriverCompletionFunc *cb; void *opaque; }; -void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs, +void *qemu_aio_get(const AIOCBInfo *aiocb_info, BlockDriverState *bs, BlockDriverCompletionFunc *cb, void *opaque); void qemu_aio_release(void *p); diff --git a/thread-pool.c b/thread-pool.c index 266f12f..894cde6 100644 --- a/thread-pool.c +++ b/thread-pool.c @@ -206,7 +206,7 @@ static void thread_pool_cancel(BlockDriverAIOCB *acb) qemu_mutex_unlock(&lock); } -static AIOPool thread_pool_cb_pool = { +static const AIOCBInfo thread_pool_aiocb_info = { .aiocb_size = sizeof(ThreadPoolElement), .cancel = thread_pool_cancel, }; @@ -216,7 +216,7 @@ BlockDriverAIOCB *thread_pool_submit_aio(ThreadPoolFunc *func, void *arg, { ThreadPoolElement *req; - req = qemu_aio_get(&thread_pool_cb_pool, NULL, cb, opaque); + req = qemu_aio_get(&thread_pool_aiocb_info, NULL, cb, opaque); req->func = func; req->arg = arg; req->state = THREAD_QUEUED; -- 1.7.11.7 ^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [Qemu-devel] [PATCH v2 3/3] aio: rename AIOPool to AIOCBInfo 2012-10-31 15:34 ` [Qemu-devel] [PATCH v2 3/3] aio: rename AIOPool to AIOCBInfo Stefan Hajnoczi @ 2012-11-02 14:50 ` Paolo Bonzini 0 siblings, 0 replies; 11+ messages in thread From: Paolo Bonzini @ 2012-11-02 14:50 UTC (permalink / raw) To: Stefan Hajnoczi; +Cc: Kevin Wolf, Anthony Liguori, qemu-devel Il 31/10/2012 16:34, Stefan Hajnoczi ha scritto: > Now that AIOPool no longer keeps a freelist, it isn't really a "pool" > anymore. Rename it to AIOCBInfo and make it const since it no longer > needs to be modified. > > Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> > --- > block.c | 22 +++++++++++----------- > 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 | 8 ++++---- > thread-pool.c | 4 ++-- > 15 files changed, 46 insertions(+), 46 deletions(-) > > diff --git a/block.c b/block.c > index ea0f7d8..854ebd6 100644 > --- a/block.c > +++ b/block.c > @@ -3521,7 +3521,7 @@ int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs) > > void bdrv_aio_cancel(BlockDriverAIOCB *acb) > { > - acb->pool->cancel(acb); > + acb->aiocb_info->cancel(acb); > } > > /* block I/O throttling */ > @@ -3711,7 +3711,7 @@ static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb) > qemu_aio_release(acb); > } > > -static AIOPool bdrv_em_aio_pool = { > +static const AIOCBInfo bdrv_em_aiocb_info = { > .aiocb_size = sizeof(BlockDriverAIOCBSync), > .cancel = bdrv_aio_cancel_em, > }; > @@ -3740,7 +3740,7 @@ static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs, > { > BlockDriverAIOCBSync *acb; > > - acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque); > + acb = qemu_aio_get(&bdrv_em_aiocb_info, bs, cb, opaque); > acb->is_write = is_write; > acb->qiov = qiov; > acb->bounce = qemu_blockalign(bs, qiov->size); > @@ -3785,7 +3785,7 @@ static void bdrv_aio_co_cancel_em(BlockDriverAIOCB *blockacb) > qemu_aio_flush(); > } > > -static AIOPool bdrv_em_co_aio_pool = { > +static const AIOCBInfo bdrv_em_co_aiocb_info = { > .aiocb_size = sizeof(BlockDriverAIOCBCoroutine), > .cancel = bdrv_aio_co_cancel_em, > }; > @@ -3828,7 +3828,7 @@ static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, > Coroutine *co; > BlockDriverAIOCBCoroutine *acb; > > - acb = qemu_aio_get(&bdrv_em_co_aio_pool, bs, cb, opaque); > + acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); > acb->req.sector = sector_num; > acb->req.nb_sectors = nb_sectors; > acb->req.qiov = qiov; > @@ -3858,7 +3858,7 @@ BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs, > Coroutine *co; > BlockDriverAIOCBCoroutine *acb; > > - acb = qemu_aio_get(&bdrv_em_co_aio_pool, bs, cb, opaque); > + acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); > co = qemu_coroutine_create(bdrv_aio_flush_co_entry); > qemu_coroutine_enter(co, acb); > > @@ -3884,7 +3884,7 @@ BlockDriverAIOCB *bdrv_aio_discard(BlockDriverState *bs, > > trace_bdrv_aio_discard(bs, sector_num, nb_sectors, opaque); > > - acb = qemu_aio_get(&bdrv_em_co_aio_pool, bs, cb, opaque); > + acb = qemu_aio_get(&bdrv_em_co_aiocb_info, bs, cb, opaque); > acb->req.sector = sector_num; > acb->req.nb_sectors = nb_sectors; > co = qemu_coroutine_create(bdrv_aio_discard_co_entry); > @@ -3904,13 +3904,13 @@ void bdrv_init_with_whitelist(void) > bdrv_init(); > } > > -void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs, > +void *qemu_aio_get(const AIOCBInfo *aiocb_info, BlockDriverState *bs, > BlockDriverCompletionFunc *cb, void *opaque) > { > BlockDriverAIOCB *acb; > > - acb = g_slice_alloc(pool->aiocb_size); > - acb->pool = pool; > + acb = g_slice_alloc(aiocb_info->aiocb_size); > + acb->aiocb_info = aiocb_info; > acb->bs = bs; > acb->cb = cb; > acb->opaque = opaque; > @@ -3920,7 +3920,7 @@ void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs, > void qemu_aio_release(void *p) > { > BlockDriverAIOCB *acb = p; > - g_slice_free1(acb->pool->aiocb_size, acb); > + g_slice_free1(acb->aiocb_info->aiocb_size, acb); > } > > /**************************************************************/ > diff --git a/block/blkdebug.c b/block/blkdebug.c > index 1206d52..d61ece8 100644 > --- a/block/blkdebug.c > +++ b/block/blkdebug.c > @@ -41,7 +41,7 @@ typedef struct BlkdebugAIOCB { > > static void blkdebug_aio_cancel(BlockDriverAIOCB *blockacb); > > -static AIOPool blkdebug_aio_pool = { > +static const AIOCBInfo blkdebug_aiocb_info = { > .aiocb_size = sizeof(BlkdebugAIOCB), > .cancel = blkdebug_aio_cancel, > }; > @@ -335,7 +335,7 @@ static BlockDriverAIOCB *inject_error(BlockDriverState *bs, > return NULL; > } > > - acb = qemu_aio_get(&blkdebug_aio_pool, bs, cb, opaque); > + acb = qemu_aio_get(&blkdebug_aiocb_info, bs, cb, opaque); > acb->ret = -error; > > bh = qemu_bh_new(error_callback_bh, acb); > diff --git a/block/blkverify.c b/block/blkverify.c > index 9d5f1ec..4beede7 100644 > --- a/block/blkverify.c > +++ b/block/blkverify.c > @@ -48,7 +48,7 @@ static void blkverify_aio_cancel(BlockDriverAIOCB *blockacb) > } > } > > -static AIOPool blkverify_aio_pool = { > +static const AIOCBInfo blkverify_aiocb_info = { > .aiocb_size = sizeof(BlkverifyAIOCB), > .cancel = blkverify_aio_cancel, > }; > @@ -233,7 +233,7 @@ static BlkverifyAIOCB *blkverify_aio_get(BlockDriverState *bs, bool is_write, > BlockDriverCompletionFunc *cb, > void *opaque) > { > - BlkverifyAIOCB *acb = qemu_aio_get(&blkverify_aio_pool, bs, cb, opaque); > + BlkverifyAIOCB *acb = qemu_aio_get(&blkverify_aiocb_info, bs, cb, opaque); > > acb->bh = NULL; > acb->is_write = is_write; > diff --git a/block/curl.c b/block/curl.c > index c1074cd..1179484 100644 > --- a/block/curl.c > +++ b/block/curl.c > @@ -438,7 +438,7 @@ static void curl_aio_cancel(BlockDriverAIOCB *blockacb) > // Do we have to implement canceling? Seems to work without... > } > > -static AIOPool curl_aio_pool = { > +static const AIOCBInfo curl_aiocb_info = { > .aiocb_size = sizeof(CURLAIOCB), > .cancel = curl_aio_cancel, > }; > @@ -505,7 +505,7 @@ static BlockDriverAIOCB *curl_aio_readv(BlockDriverState *bs, > { > CURLAIOCB *acb; > > - acb = qemu_aio_get(&curl_aio_pool, bs, cb, opaque); > + acb = qemu_aio_get(&curl_aiocb_info, bs, cb, opaque); > > acb->qiov = qiov; > acb->sector_num = sector_num; > diff --git a/block/gluster.c b/block/gluster.c > index 3588d73..1c90174 100644 > --- a/block/gluster.c > +++ b/block/gluster.c > @@ -388,7 +388,7 @@ static void qemu_gluster_aio_cancel(BlockDriverAIOCB *blockacb) > } > } > > -static AIOPool gluster_aio_pool = { > +static const AIOCBInfo gluster_aiocb_info = { > .aiocb_size = sizeof(GlusterAIOCB), > .cancel = qemu_gluster_aio_cancel, > }; > @@ -439,7 +439,7 @@ static BlockDriverAIOCB *qemu_gluster_aio_rw(BlockDriverState *bs, > size = nb_sectors * BDRV_SECTOR_SIZE; > s->qemu_aio_count++; > > - acb = qemu_aio_get(&gluster_aio_pool, bs, cb, opaque); > + acb = qemu_aio_get(&gluster_aiocb_info, bs, cb, opaque); > acb->size = size; > acb->ret = 0; > acb->finished = NULL; > @@ -484,7 +484,7 @@ static BlockDriverAIOCB *qemu_gluster_aio_flush(BlockDriverState *bs, > GlusterAIOCB *acb; > BDRVGlusterState *s = bs->opaque; > > - acb = qemu_aio_get(&gluster_aio_pool, bs, cb, opaque); > + acb = qemu_aio_get(&gluster_aiocb_info, bs, cb, opaque); > acb->size = 0; > acb->ret = 0; > acb->finished = NULL; > diff --git a/block/iscsi.c b/block/iscsi.c > index d0b1a10..a6a819d 100644 > --- a/block/iscsi.c > +++ b/block/iscsi.c > @@ -133,7 +133,7 @@ iscsi_aio_cancel(BlockDriverAIOCB *blockacb) > } > } > > -static AIOPool iscsi_aio_pool = { > +static const AIOCBInfo iscsi_aiocb_info = { > .aiocb_size = sizeof(IscsiAIOCB), > .cancel = iscsi_aio_cancel, > }; > @@ -234,7 +234,7 @@ iscsi_aio_writev(BlockDriverState *bs, int64_t sector_num, > uint64_t lba; > struct iscsi_data data; > > - acb = qemu_aio_get(&iscsi_aio_pool, bs, cb, opaque); > + acb = qemu_aio_get(&iscsi_aiocb_info, bs, cb, opaque); > trace_iscsi_aio_writev(iscsi, sector_num, nb_sectors, opaque, acb); > > acb->iscsilun = iscsilun; > @@ -325,7 +325,7 @@ iscsi_aio_readv(BlockDriverState *bs, int64_t sector_num, > > qemu_read_size = BDRV_SECTOR_SIZE * (size_t)nb_sectors; > > - acb = qemu_aio_get(&iscsi_aio_pool, bs, cb, opaque); > + acb = qemu_aio_get(&iscsi_aiocb_info, bs, cb, opaque); > trace_iscsi_aio_readv(iscsi, sector_num, nb_sectors, opaque, acb); > > acb->iscsilun = iscsilun; > @@ -430,7 +430,7 @@ iscsi_aio_flush(BlockDriverState *bs, > struct iscsi_context *iscsi = iscsilun->iscsi; > IscsiAIOCB *acb; > > - acb = qemu_aio_get(&iscsi_aio_pool, bs, cb, opaque); > + acb = qemu_aio_get(&iscsi_aiocb_info, bs, cb, opaque); > > acb->iscsilun = iscsilun; > acb->canceled = 0; > @@ -483,7 +483,7 @@ iscsi_aio_discard(BlockDriverState *bs, > IscsiAIOCB *acb; > struct unmap_list list[1]; > > - acb = qemu_aio_get(&iscsi_aio_pool, bs, cb, opaque); > + acb = qemu_aio_get(&iscsi_aiocb_info, bs, cb, opaque); > > acb->iscsilun = iscsilun; > acb->canceled = 0; > @@ -558,7 +558,7 @@ static BlockDriverAIOCB *iscsi_aio_ioctl(BlockDriverState *bs, > > assert(req == SG_IO); > > - acb = qemu_aio_get(&iscsi_aio_pool, bs, cb, opaque); > + acb = qemu_aio_get(&iscsi_aiocb_info, bs, cb, opaque); > > acb->iscsilun = iscsilun; > acb->canceled = 0; > diff --git a/block/linux-aio.c b/block/linux-aio.c > index 6ca984d..91ef863 100644 > --- a/block/linux-aio.c > +++ b/block/linux-aio.c > @@ -140,7 +140,7 @@ static void laio_cancel(BlockDriverAIOCB *blockacb) > } > } > > -static AIOPool laio_pool = { > +static const AIOCBInfo laio_aiocb_info = { > .aiocb_size = sizeof(struct qemu_laiocb), > .cancel = laio_cancel, > }; > @@ -154,7 +154,7 @@ BlockDriverAIOCB *laio_submit(BlockDriverState *bs, void *aio_ctx, int fd, > struct iocb *iocbs; > off_t offset = sector_num * 512; > > - laiocb = qemu_aio_get(&laio_pool, bs, cb, opaque); > + laiocb = qemu_aio_get(&laio_aiocb_info, bs, cb, opaque); > laiocb->nbytes = nb_sectors * 512; > laiocb->ctx = s; > laiocb->ret = -EINPROGRESS; > diff --git a/block/qed.c b/block/qed.c > index 6c182ca..0b5374a 100644 > --- a/block/qed.c > +++ b/block/qed.c > @@ -30,7 +30,7 @@ static void qed_aio_cancel(BlockDriverAIOCB *blockacb) > } > } > > -static AIOPool qed_aio_pool = { > +static const AIOCBInfo qed_aiocb_info = { > .aiocb_size = sizeof(QEDAIOCB), > .cancel = qed_aio_cancel, > }; > @@ -1311,7 +1311,7 @@ static BlockDriverAIOCB *qed_aio_setup(BlockDriverState *bs, > BlockDriverCompletionFunc *cb, > void *opaque, int flags) > { > - QEDAIOCB *acb = qemu_aio_get(&qed_aio_pool, bs, cb, opaque); > + QEDAIOCB *acb = qemu_aio_get(&qed_aiocb_info, bs, cb, opaque); > > trace_qed_aio_setup(bs->opaque, acb, sector_num, nb_sectors, > opaque, flags); > diff --git a/block/rbd.c b/block/rbd.c > index 015a9db..0aaacaf 100644 > --- a/block/rbd.c > +++ b/block/rbd.c > @@ -570,7 +570,7 @@ static void qemu_rbd_aio_cancel(BlockDriverAIOCB *blockacb) > acb->cancelled = 1; > } > > -static AIOPool rbd_aio_pool = { > +static const AIOCBInfo rbd_aiocb_info = { > .aiocb_size = sizeof(RBDAIOCB), > .cancel = qemu_rbd_aio_cancel, > }; > @@ -672,7 +672,7 @@ static BlockDriverAIOCB *rbd_start_aio(BlockDriverState *bs, > > BDRVRBDState *s = bs->opaque; > > - acb = qemu_aio_get(&rbd_aio_pool, bs, cb, opaque); > + acb = qemu_aio_get(&rbd_aiocb_info, bs, cb, opaque); > acb->cmd = cmd; > acb->qiov = qiov; > if (cmd == RBD_AIO_DISCARD) { > diff --git a/block/sheepdog.c b/block/sheepdog.c > index 9306174..a48f58c 100644 > --- a/block/sheepdog.c > +++ b/block/sheepdog.c > @@ -420,7 +420,7 @@ static void sd_aio_cancel(BlockDriverAIOCB *blockacb) > acb->canceled = true; > } > > -static AIOPool sd_aio_pool = { > +static const AIOCBInfo sd_aiocb_info = { > .aiocb_size = sizeof(SheepdogAIOCB), > .cancel = sd_aio_cancel, > }; > @@ -431,7 +431,7 @@ static SheepdogAIOCB *sd_aio_setup(BlockDriverState *bs, QEMUIOVector *qiov, > { > SheepdogAIOCB *acb; > > - acb = qemu_aio_get(&sd_aio_pool, bs, cb, opaque); > + acb = qemu_aio_get(&sd_aiocb_info, bs, cb, opaque); > > acb->qiov = qiov; > > diff --git a/block/win32-aio.c b/block/win32-aio.c > index c34dc73..444c228 100644 > --- a/block/win32-aio.c > +++ b/block/win32-aio.c > @@ -131,7 +131,7 @@ static void win32_aio_cancel(BlockDriverAIOCB *blockacb) > } > } > > -static AIOPool win32_aio_pool = { > +static const AIOCBInfo win32_aiocb_info = { > .aiocb_size = sizeof(QEMUWin32AIOCB), > .cancel = win32_aio_cancel, > }; > @@ -145,7 +145,7 @@ BlockDriverAIOCB *win32_aio_submit(BlockDriverState *bs, > uint64_t offset = sector_num * 512; > DWORD rc; > > - waiocb = qemu_aio_get(&win32_aio_pool, bs, cb, opaque); > + waiocb = qemu_aio_get(&win32_aiocb_info, bs, cb, opaque); > waiocb->nbytes = nb_sectors * 512; > waiocb->qiov = qiov; > waiocb->is_read = (type == QEMU_AIO_READ); > diff --git a/dma-helpers.c b/dma-helpers.c > index 0c18e9e..4f5fb64 100644 > --- a/dma-helpers.c > +++ b/dma-helpers.c > @@ -195,7 +195,7 @@ static void dma_aio_cancel(BlockDriverAIOCB *acb) > dma_complete(dbs, 0); > } > > -static AIOPool dma_aio_pool = { > +static const AIOCBInfo dma_aiocb_info = { > .aiocb_size = sizeof(DMAAIOCB), > .cancel = dma_aio_cancel, > }; > @@ -205,7 +205,7 @@ BlockDriverAIOCB *dma_bdrv_io( > DMAIOFunc *io_func, BlockDriverCompletionFunc *cb, > void *opaque, DMADirection dir) > { > - DMAAIOCB *dbs = qemu_aio_get(&dma_aio_pool, bs, cb, opaque); > + DMAAIOCB *dbs = qemu_aio_get(&dma_aiocb_info, bs, cb, opaque); > > trace_dma_bdrv_io(dbs, bs, sector_num, (dir == DMA_DIRECTION_TO_DEVICE)); > > diff --git a/hw/ide/core.c b/hw/ide/core.c > index d683a8c..7d6b0fa 100644 > --- a/hw/ide/core.c > +++ b/hw/ide/core.c > @@ -336,7 +336,7 @@ static void trim_aio_cancel(BlockDriverAIOCB *acb) > qemu_aio_release(iocb); > } > > -static AIOPool trim_aio_pool = { > +static const AIOCBInfo trim_aiocb_info = { > .aiocb_size = sizeof(TrimAIOCB), > .cancel = trim_aio_cancel, > }; > @@ -360,7 +360,7 @@ BlockDriverAIOCB *ide_issue_trim(BlockDriverState *bs, > TrimAIOCB *iocb; > int i, j, ret; > > - iocb = qemu_aio_get(&trim_aio_pool, bs, cb, opaque); > + iocb = qemu_aio_get(&trim_aiocb_info, bs, cb, opaque); > iocb->bh = qemu_bh_new(ide_trim_bh_cb, iocb); > iocb->ret = 0; > > diff --git a/qemu-aio.h b/qemu-aio.h > index b29c509..3889fe9 100644 > --- a/qemu-aio.h > +++ b/qemu-aio.h > @@ -21,19 +21,19 @@ > typedef struct BlockDriverAIOCB BlockDriverAIOCB; > typedef void BlockDriverCompletionFunc(void *opaque, int ret); > > -typedef struct AIOPool { > +typedef struct AIOCBInfo { > void (*cancel)(BlockDriverAIOCB *acb); > size_t aiocb_size; > -} AIOPool; > +} AIOCBInfo; > > struct BlockDriverAIOCB { > - AIOPool *pool; > + const AIOCBInfo *aiocb_info; > BlockDriverState *bs; > BlockDriverCompletionFunc *cb; > void *opaque; > }; > > -void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs, > +void *qemu_aio_get(const AIOCBInfo *aiocb_info, BlockDriverState *bs, > BlockDriverCompletionFunc *cb, void *opaque); > void qemu_aio_release(void *p); > > diff --git a/thread-pool.c b/thread-pool.c > index 266f12f..894cde6 100644 > --- a/thread-pool.c > +++ b/thread-pool.c > @@ -206,7 +206,7 @@ static void thread_pool_cancel(BlockDriverAIOCB *acb) > qemu_mutex_unlock(&lock); > } > > -static AIOPool thread_pool_cb_pool = { > +static const AIOCBInfo thread_pool_aiocb_info = { > .aiocb_size = sizeof(ThreadPoolElement), > .cancel = thread_pool_cancel, > }; > @@ -216,7 +216,7 @@ BlockDriverAIOCB *thread_pool_submit_aio(ThreadPoolFunc *func, void *arg, > { > ThreadPoolElement *req; > > - req = qemu_aio_get(&thread_pool_cb_pool, NULL, cb, opaque); > + req = qemu_aio_get(&thread_pool_aiocb_info, NULL, cb, opaque); > req->func = func; > req->arg = arg; > req->state = THREAD_QUEUED; > Didn't really review it, but if it compiles... :) Paolo ^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [Qemu-devel] [PATCH v2 0/3] aio: use g_slice_alloc() for AIOCB pooling 2012-10-31 15:34 [Qemu-devel] [PATCH v2 0/3] aio: use g_slice_alloc() for AIOCB pooling Stefan Hajnoczi ` (2 preceding siblings ...) 2012-10-31 15:34 ` [Qemu-devel] [PATCH v2 3/3] aio: rename AIOPool to AIOCBInfo Stefan Hajnoczi @ 2012-11-12 10:34 ` Kevin Wolf 3 siblings, 0 replies; 11+ messages in thread From: Kevin Wolf @ 2012-11-12 10:34 UTC (permalink / raw) To: Stefan Hajnoczi; +Cc: Paolo Bonzini, Anthony Liguori, qemu-devel Am 31.10.2012 16:34, schrieb Stefan Hajnoczi: > 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(-) Thanks, applied all to the block branch. Kevin ^ permalink raw reply [flat|nested] 11+ messages in thread
end of thread, other threads:[~2012-11-12 10:42 UTC | newest] Thread overview: 11+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2012-10-31 15:34 [Qemu-devel] [PATCH v2 0/3] aio: use g_slice_alloc() for AIOCB pooling Stefan Hajnoczi 2012-10-31 15:34 ` [Qemu-devel] [PATCH v2 1/3] aio: switch aiocb_size type int -> size_t Stefan Hajnoczi 2012-11-02 14:49 ` Paolo Bonzini 2012-10-31 15:34 ` [Qemu-devel] [PATCH v2 2/3] aio: use g_slice_alloc() for AIOCB pooling Stefan Hajnoczi 2012-11-02 14:49 ` Paolo Bonzini 2012-11-12 10:34 ` Kevin Wolf 2012-11-12 10:37 ` Paolo Bonzini 2012-11-12 10:42 ` Kevin Wolf 2012-10-31 15:34 ` [Qemu-devel] [PATCH v2 3/3] aio: rename AIOPool to AIOCBInfo Stefan Hajnoczi 2012-11-02 14:50 ` Paolo Bonzini 2012-11-12 10:34 ` [Qemu-devel] [PATCH v2 0/3] aio: use g_slice_alloc() for AIOCB pooling Kevin Wolf
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).