From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([140.186.70.92]:44299) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1RdOcQ-0007le-3g for qemu-devel@nongnu.org; Wed, 21 Dec 2011 11:02:20 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1RdOcG-0001sZ-Av for qemu-devel@nongnu.org; Wed, 21 Dec 2011 11:02:14 -0500 Received: from e06smtp13.uk.ibm.com ([195.75.94.109]:42465) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1RdOcF-0001sJ-Tl for qemu-devel@nongnu.org; Wed, 21 Dec 2011 11:02:04 -0500 Received: from /spool/local by e06smtp13.uk.ibm.com with IBM ESMTP SMTP Gateway: Authorized Use Only! Violators will be prosecuted for from ; Wed, 21 Dec 2011 16:02:02 -0000 Received: from d06av12.portsmouth.uk.ibm.com (d06av12.portsmouth.uk.ibm.com [9.149.37.247]) by d06nrmr1707.portsmouth.uk.ibm.com (8.13.8/8.13.8/NCO v10.0) with ESMTP id pBLG21xl2711608 for ; Wed, 21 Dec 2011 16:02:01 GMT Received: from d06av12.portsmouth.uk.ibm.com (loopback [127.0.0.1]) by d06av12.portsmouth.uk.ibm.com (8.14.4/8.13.1/NCO v10.0 AVout) with ESMTP id pBLG21pt022284 for ; Wed, 21 Dec 2011 09:02:01 -0700 From: Stefan Hajnoczi Date: Wed, 21 Dec 2011 16:00:36 +0000 Message-Id: <1324483240-31726-3-git-send-email-stefanha@linux.vnet.ibm.com> In-Reply-To: <1324483240-31726-1-git-send-email-stefanha@linux.vnet.ibm.com> References: <1324483240-31726-1-git-send-email-stefanha@linux.vnet.ibm.com> Subject: [Qemu-devel] [PATCH v3 2/6] block: add .bdrv_co_write_zeroes() interface List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: qemu-devel@nongnu.org Cc: Kevin Wolf , Marcelo Tosatti , Stefan Hajnoczi The ability to zero regions of an image file is a useful primitive for higher-level features such as image streaming or zero write detection. Image formats may support an optimized metadata representation instead of writing zeroes into the image file. This allows zero writes to be potentially faster than regular write operations and also preserve sparseness of the image file. The .bdrv_co_write_zeroes() interface should be implemented by block drivers that wish to provide efficient zeroing. Note that this operation is different from the discard operation, which may leave the contents of the region indeterminate. That means discarded blocks are not guaranteed to contain zeroes and may contain junk data instead. Signed-off-by: Stefan Hajnoczi --- block.c | 50 ++++++++++++++++++++++++++++++++++++++++++++------ block.h | 7 +++++++ block_int.h | 8 ++++++++ trace-events | 1 + 4 files changed, 60 insertions(+), 6 deletions(-) diff --git a/block.c b/block.c index 3f072f6..5ebbd4d 100644 --- a/block.c +++ b/block.c @@ -64,7 +64,7 @@ static int coroutine_fn bdrv_co_writev_em(BlockDriverState *bs, static int coroutine_fn bdrv_co_do_readv(BlockDriverState *bs, int64_t sector_num, int nb_sectors, QEMUIOVector *qiov); static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs, - int64_t sector_num, int nb_sectors, QEMUIOVector *qiov); + int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, bool write_zeroes); static BlockDriverAIOCB *bdrv_co_aio_rw_vector(BlockDriverState *bs, int64_t sector_num, QEMUIOVector *qiov, @@ -1291,7 +1291,7 @@ static void coroutine_fn bdrv_rw_co_entry(void *opaque) rwco->nb_sectors, rwco->qiov); } else { rwco->ret = bdrv_co_do_writev(rwco->bs, rwco->sector_num, - rwco->nb_sectors, rwco->qiov); + rwco->nb_sectors, rwco->qiov, false); } } @@ -1608,11 +1608,37 @@ int coroutine_fn bdrv_co_readv(BlockDriverState *bs, int64_t sector_num, return bdrv_co_do_readv(bs, sector_num, nb_sectors, qiov); } +static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, + int64_t sector_num, int nb_sectors) +{ + BlockDriver *drv = bs->drv; + QEMUIOVector qiov; + struct iovec iov; + int ret; + + /* First try the efficient write zeroes operation */ + if (drv->bdrv_co_write_zeroes) { + return drv->bdrv_co_write_zeroes(bs, sector_num, nb_sectors); + } + + /* Fall back to bounce buffer if write zeroes is unsupported */ + iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE; + iov.iov_base = qemu_blockalign(bs, iov.iov_len); + memset(iov.iov_base, 0, iov.iov_len); + qemu_iovec_init_external(&qiov, &iov, 1); + + ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, &qiov); + + qemu_vfree(iov.iov_base); + return ret; +} + /* * Handle a write request in coroutine context */ static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs, - int64_t sector_num, int nb_sectors, QEMUIOVector *qiov) + int64_t sector_num, int nb_sectors, QEMUIOVector *qiov, + bool write_zeroes) { BlockDriver *drv = bs->drv; BdrvTrackedRequest req; @@ -1639,7 +1665,11 @@ static int coroutine_fn bdrv_co_do_writev(BlockDriverState *bs, tracked_request_begin(&req, bs, sector_num, nb_sectors, true); - ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, qiov); + if (write_zeroes) { + ret = bdrv_co_do_write_zeroes(bs, sector_num, nb_sectors); + } else { + ret = drv->bdrv_co_writev(bs, sector_num, nb_sectors, qiov); + } if (bs->dirty_bitmap) { set_dirty_bitmap(bs, sector_num, nb_sectors, 1); @@ -1659,7 +1689,15 @@ int coroutine_fn bdrv_co_writev(BlockDriverState *bs, int64_t sector_num, { trace_bdrv_co_writev(bs, sector_num, nb_sectors); - return bdrv_co_do_writev(bs, sector_num, nb_sectors, qiov); + return bdrv_co_do_writev(bs, sector_num, nb_sectors, qiov, false); +} + +int coroutine_fn bdrv_co_write_zeroes(BlockDriverState *bs, + int64_t sector_num, int nb_sectors) +{ + trace_bdrv_co_write_zeroes(bs, sector_num, nb_sectors); + + return bdrv_co_do_writev(bs, sector_num, nb_sectors, NULL, true); } /** @@ -3143,7 +3181,7 @@ static void coroutine_fn bdrv_co_do_rw(void *opaque) acb->req.nb_sectors, acb->req.qiov); } else { acb->req.error = bdrv_co_do_writev(bs, acb->req.sector, - acb->req.nb_sectors, acb->req.qiov); + acb->req.nb_sectors, acb->req.qiov, false); } acb->bh = qemu_bh_new(bdrv_co_em_bh, acb); diff --git a/block.h b/block.h index 3bd4398..51b90c7 100644 --- a/block.h +++ b/block.h @@ -144,6 +144,13 @@ int coroutine_fn bdrv_co_readv(BlockDriverState *bs, int64_t sector_num, int nb_sectors, QEMUIOVector *qiov); int coroutine_fn bdrv_co_writev(BlockDriverState *bs, int64_t sector_num, int nb_sectors, QEMUIOVector *qiov); +/* + * Efficiently zero a region of the disk image. Note that this is a regular + * I/O request like read or write and should have a reasonable size. This + * function is not suitable for zeroing the entire image in a single request. + */ +int coroutine_fn bdrv_co_write_zeroes(BlockDriverState *bs, int64_t sector_num, + int nb_sectors); int coroutine_fn bdrv_co_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors, int *pnum); int bdrv_truncate(BlockDriverState *bs, int64_t offset); diff --git a/block_int.h b/block_int.h index 311bd2a..5362180 100644 --- a/block_int.h +++ b/block_int.h @@ -101,6 +101,14 @@ struct BlockDriver { int64_t sector_num, int nb_sectors, QEMUIOVector *qiov); int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs, int64_t sector_num, int nb_sectors, QEMUIOVector *qiov); + /* + * Efficiently zero a region of the disk image. Typically an image format + * would use a compact metadata representation to implement this. This + * function pointer may be NULL and .bdrv_co_writev() will be called + * instead. + */ + int coroutine_fn (*bdrv_co_write_zeroes)(BlockDriverState *bs, + int64_t sector_num, int nb_sectors); int coroutine_fn (*bdrv_co_discard)(BlockDriverState *bs, int64_t sector_num, int nb_sectors); int coroutine_fn (*bdrv_co_is_allocated)(BlockDriverState *bs, diff --git a/trace-events b/trace-events index 514849a..fd2d7d9 100644 --- a/trace-events +++ b/trace-events @@ -66,6 +66,7 @@ bdrv_aio_writev(void *bs, int64_t sector_num, int nb_sectors, void *opaque) "bs bdrv_lock_medium(void *bs, bool locked) "bs %p locked %d" bdrv_co_readv(void *bs, int64_t sector_num, int nb_sector) "bs %p sector_num %"PRId64" nb_sectors %d" bdrv_co_writev(void *bs, int64_t sector_num, int nb_sector) "bs %p sector_num %"PRId64" nb_sectors %d" +bdrv_co_write_zeroes(void *bs, int64_t sector_num, int nb_sector) "bs %p sector_num %"PRId64" nb_sectors %d" bdrv_co_io_em(void *bs, int64_t sector_num, int nb_sectors, int is_write, void *acb) "bs %p sector_num %"PRId64" nb_sectors %d is_write %d acb %p" bdrv_co_copy_on_readv(void *bs, int64_t sector_num, int nb_sectors, int64_t cluster_sector_num, int cluster_nb_sectors) "bs %p sector_num %"PRId64" nb_sectors %d cluster_sector_num %"PRId64" cluster_nb_sectors %d" -- 1.7.7.3