From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:59532) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1VoyP5-0006zL-BR for qemu-devel@nongnu.org; Fri, 06 Dec 2013 11:37:28 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1VoyP0-0003LT-EA for qemu-devel@nongnu.org; Fri, 06 Dec 2013 11:37:23 -0500 Received: from mx1.redhat.com ([209.132.183.28]:31085) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1VoyP0-0003LN-6h for qemu-devel@nongnu.org; Fri, 06 Dec 2013 11:37:18 -0500 From: Stefan Hajnoczi Date: Fri, 6 Dec 2013 17:36:08 +0100 Message-Id: <1386347807-27359-10-git-send-email-stefanha@redhat.com> In-Reply-To: <1386347807-27359-1-git-send-email-stefanha@redhat.com> References: <1386347807-27359-1-git-send-email-stefanha@redhat.com> Subject: [Qemu-devel] [PULL 09/48] block: make bdrv_co_do_write_zeroes stricter in producing aligned requests List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: qemu-devel@nongnu.org Cc: Paolo Bonzini , Stefan Hajnoczi , Anthony Liguori From: Paolo Bonzini Right now, bdrv_co_do_write_zeroes will only try to align the beginning of the request. However, it is simpler for many formats to expect the block layer to separate both the head *and* the tail. This makes sure that the format's bdrv_co_write_zeroes function will be called with aligned sector_num and nb_sectors for the bulk of the request. Signed-off-by: Paolo Bonzini Reviewed-by: Peter Lieven Signed-off-by: Stefan Hajnoczi --- block.c | 35 +++++++++++++++++++++++------------ 1 file changed, 23 insertions(+), 12 deletions(-) diff --git a/block.c b/block.c index 9818052..831c150 100644 --- a/block.c +++ b/block.c @@ -2771,14 +2771,21 @@ static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, while (nb_sectors > 0 && !ret) { int num = nb_sectors; - /* align request */ - if (bs->bl.write_zeroes_alignment && - num >= bs->bl.write_zeroes_alignment && - sector_num % bs->bl.write_zeroes_alignment) { - if (num > bs->bl.write_zeroes_alignment) { + /* Align request. Block drivers can expect the "bulk" of the request + * to be aligned. + */ + if (bs->bl.write_zeroes_alignment + && num > bs->bl.write_zeroes_alignment) { + if (sector_num % bs->bl.write_zeroes_alignment != 0) { + /* Make a small request up to the first aligned sector. */ num = bs->bl.write_zeroes_alignment; + num -= sector_num % bs->bl.write_zeroes_alignment; + } else if ((sector_num + num) % bs->bl.write_zeroes_alignment != 0) { + /* Shorten the request to the last aligned sector. num cannot + * underflow because num > bs->bl.write_zeroes_alignment. + */ + num -= (sector_num + num) % bs->bl.write_zeroes_alignment; } - num -= sector_num % bs->bl.write_zeroes_alignment; } /* limit request size */ @@ -2796,16 +2803,20 @@ static int coroutine_fn bdrv_co_do_write_zeroes(BlockDriverState *bs, /* Fall back to bounce buffer if write zeroes is unsupported */ iov.iov_len = num * BDRV_SECTOR_SIZE; if (iov.iov_base == NULL) { - /* allocate bounce buffer only once and ensure that it - * is big enough for this and all future requests. - */ - size_t bufsize = num <= nb_sectors ? num : max_write_zeroes; - iov.iov_base = qemu_blockalign(bs, bufsize * BDRV_SECTOR_SIZE); - memset(iov.iov_base, 0, bufsize * BDRV_SECTOR_SIZE); + iov.iov_base = qemu_blockalign(bs, num * BDRV_SECTOR_SIZE); + memset(iov.iov_base, 0, num * BDRV_SECTOR_SIZE); } qemu_iovec_init_external(&qiov, &iov, 1); ret = drv->bdrv_co_writev(bs, sector_num, num, &qiov); + + /* Keep bounce buffer around if it is big enough for all + * all future requests. + */ + if (num < max_write_zeroes) { + qemu_vfree(iov.iov_base); + iov.iov_base = NULL; + } } sector_num += num; -- 1.8.4.2