From: "Benoît Canet" <benoit.canet@irqsave.net>
To: Kevin Wolf <kwolf@redhat.com>
Cc: pl@kamp.de, qemu-devel@nongnu.org, mreitz@redhat.com,
stefanha@redhat.com, pbonzini@redhat.com,
xiawenc@linux.vnet.ibm.com
Subject: Re: [Qemu-devel] [PATCH v3 20/29] block: Align requests in bdrv_co_do_pwritev()
Date: Wed, 22 Jan 2014 21:29:39 +0100 [thread overview]
Message-ID: <20140122202939.GF3053@irqsave.net> (raw)
In-Reply-To: <1389968119-24771-21-git-send-email-kwolf@redhat.com>
Le Friday 17 Jan 2014 à 15:15:10 (+0100), Kevin Wolf a écrit :
> This patch changes bdrv_co_do_pwritev() to actually be what its name
> promises. If requests aren't properly aligned, it performs a RMW.
>
> Requests touching the same block are serialised against the RMW request.
> Further optimisation of this is possible by differentiating types of
> requests (concurrent reads should actually be okay here).
>
> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
> Reviewed-by: Max Reitz <mreitz@redhat.com>
> ---
> block.c | 86 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
> 1 file changed, 85 insertions(+), 1 deletion(-)
>
> diff --git a/block.c b/block.c
> index 55e8c69..859e1aa 100644
> --- a/block.c
> +++ b/block.c
> @@ -3055,6 +3055,12 @@ static int coroutine_fn bdrv_co_do_pwritev(BlockDriverState *bs,
> BdrvRequestFlags flags)
> {
> BdrvTrackedRequest req;
> + /* TODO Lift BDRV_SECTOR_SIZE restriction in BlockDriver interface */
> + uint64_t align = MAX(BDRV_SECTOR_SIZE, bs->request_alignment);
> + uint8_t *head_buf = NULL;
> + uint8_t *tail_buf = NULL;
> + QEMUIOVector local_qiov;
> + bool use_local_qiov = false;
> int ret;
>
> if (!bs->drv) {
> @@ -3073,10 +3079,88 @@ static int coroutine_fn bdrv_co_do_pwritev(BlockDriverState *bs,
> bdrv_io_limits_intercept(bs, bytes >> BDRV_SECTOR_BITS, true);
> }
>
> + /*
> + * Align write if necessary by performing a read-modify-write cycle.
> + * Pad qiov with the read parts and be sure to have a tracked request not
> + * only for bdrv_aligned_pwritev, but also for the reads of the RMW cycle.
> + */
> tracked_request_begin(&req, bs, offset, bytes, true);
> - ret = bdrv_aligned_pwritev(bs, &req, offset, bytes, qiov, flags);
> +
> + if (offset & (align - 1)) {
> + QEMUIOVector head_qiov;
> + struct iovec head_iov;
> +
> + mark_request_serialising(&req, align);
> + wait_serialising_requests(&req);
> +
> + head_buf = qemu_blockalign(bs, align);
> + head_iov = (struct iovec) {
> + .iov_base = head_buf,
> + .iov_len = align,
> + };
> + qemu_iovec_init_external(&head_qiov, &head_iov, 1);
> +
> + ret = bdrv_aligned_preadv(bs, &req, offset & ~(align - 1), align,
> + align, &head_qiov, 0);
> + if (ret < 0) {
> + goto fail;
> + }
> +
> + qemu_iovec_init(&local_qiov, qiov->niov + 2);
> + qemu_iovec_add(&local_qiov, head_buf, offset & (align - 1));
> + qemu_iovec_concat(&local_qiov, qiov, 0, qiov->size);
> + use_local_qiov = true;
> +
> + bytes += offset & (align - 1);
> + offset = offset & ~(align - 1);
> + }
> +
> + if ((offset + bytes) & (align - 1)) {
> + QEMUIOVector tail_qiov;
> + struct iovec tail_iov;
> + size_t tail_bytes;
> +
> + mark_request_serialising(&req, align);
> + wait_serialising_requests(&req);
> +
> + tail_buf = qemu_blockalign(bs, align);
> + tail_iov = (struct iovec) {
> + .iov_base = tail_buf,
> + .iov_len = align,
> + };
> + qemu_iovec_init_external(&tail_qiov, &tail_iov, 1);
> +
> + ret = bdrv_aligned_preadv(bs, &req, (offset + bytes) & ~(align - 1), align,
> + align, &tail_qiov, 0);
> + if (ret < 0) {
> + goto fail;
> + }
> +
> + if (!use_local_qiov) {
> + qemu_iovec_init(&local_qiov, qiov->niov + 1);
> + qemu_iovec_concat(&local_qiov, qiov, 0, qiov->size);
> + use_local_qiov = true;
> + }
> +
> + tail_bytes = (offset + bytes) & (align - 1);
> + qemu_iovec_add(&local_qiov, tail_buf + tail_bytes, align - tail_bytes);
> +
> + bytes = ROUND_UP(bytes, align);
> + }
> +
> + ret = bdrv_aligned_pwritev(bs, &req, offset, bytes,
> + use_local_qiov ? &local_qiov : qiov,
> + flags);
> +
> +fail:
> tracked_request_end(&req);
>
> + if (use_local_qiov) {
> + qemu_iovec_destroy(&local_qiov);
> + qemu_vfree(head_buf);
> + qemu_vfree(tail_buf);
> + }
> +
> return ret;
> }
>
> --
> 1.8.1.4
>
>
Reviewed-by: Benoit Canet <benoit@irqsave.net>
next prev parent reply other threads:[~2014-01-22 20:29 UTC|newest]
Thread overview: 70+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-01-17 14:14 [Qemu-devel] [PATCH v3 00/29] block: Support for 512b-on-4k emulation Kevin Wolf
2014-01-17 14:14 ` [Qemu-devel] [PATCH v3 01/29] block: Move initialisation of BlockLimits to bdrv_refresh_limits() Kevin Wolf
2014-01-17 22:39 ` Benoît Canet
2014-01-20 9:31 ` Kevin Wolf
2014-01-20 9:49 ` Peter Lieven
2014-01-21 12:49 ` Benoît Canet
2014-01-17 14:14 ` [Qemu-devel] [PATCH v3 02/29] block: Inherit opt_transfer_length Kevin Wolf
2014-01-17 22:42 ` Benoît Canet
2014-01-17 14:14 ` [Qemu-devel] [PATCH v3 03/29] block: Update BlockLimits when they might have changed Kevin Wolf
2014-01-17 22:47 ` Benoît Canet
2014-01-17 14:14 ` [Qemu-devel] [PATCH v3 04/29] qemu_memalign: Allow small alignments Kevin Wolf
2014-01-17 22:49 ` Benoît Canet
2014-01-17 14:14 ` [Qemu-devel] [PATCH v3 05/29] block: Detect unaligned length in bdrv_qiov_is_aligned() Kevin Wolf
2014-01-17 14:14 ` [Qemu-devel] [PATCH v3 06/29] block: Don't use guest sector size for qemu_blockalign() Kevin Wolf
2014-01-17 14:14 ` [Qemu-devel] [PATCH v3 07/29] block: rename buffer_alignment to guest_block_size Kevin Wolf
2014-01-21 12:54 ` Benoît Canet
2014-01-17 14:14 ` [Qemu-devel] [PATCH v3 08/29] raw: Probe required direct I/O alignment Kevin Wolf
2014-01-21 13:03 ` Benoît Canet
2014-01-21 13:29 ` Kevin Wolf
2014-01-17 14:14 ` [Qemu-devel] [PATCH v3 09/29] block: Introduce bdrv_aligned_preadv() Kevin Wolf
2014-01-21 13:13 ` Benoît Canet
2014-01-17 14:15 ` [Qemu-devel] [PATCH v3 10/29] block: Introduce bdrv_co_do_preadv() Kevin Wolf
2014-01-17 23:59 ` Max Reitz
2014-01-21 13:29 ` Benoît Canet
2014-01-17 14:15 ` [Qemu-devel] [PATCH v3 11/29] block: Introduce bdrv_aligned_pwritev() Kevin Wolf
2014-01-21 13:31 ` Benoît Canet
2014-01-17 14:15 ` [Qemu-devel] [PATCH v3 12/29] block: write: Handle COR dependency after I/O throttling Kevin Wolf
2014-01-21 13:33 ` Benoît Canet
2014-01-17 14:15 ` [Qemu-devel] [PATCH v3 13/29] block: Introduce bdrv_co_do_pwritev() Kevin Wolf
2014-01-18 0:00 ` Max Reitz
2014-01-21 13:36 ` Benoît Canet
2014-01-17 14:15 ` [Qemu-devel] [PATCH v3 14/29] block: Switch BdrvTrackedRequest to byte granularity Kevin Wolf
2014-01-17 23:19 ` Max Reitz
2014-01-21 13:49 ` Benoît Canet
2014-01-17 14:15 ` [Qemu-devel] [PATCH v3 15/29] block: Allow waiting for overlapping requests between begin/end Kevin Wolf
2014-01-22 19:46 ` Benoît Canet
2014-01-17 14:15 ` [Qemu-devel] [PATCH v3 16/29] block: Make zero-after-EOF work with larger alignment Kevin Wolf
2014-01-17 23:21 ` Max Reitz
2014-01-22 19:50 ` Benoît Canet
2014-01-17 14:15 ` [Qemu-devel] [PATCH v3 17/29] block: Generalise and optimise COR serialisation Kevin Wolf
2014-01-22 20:00 ` Benoît Canet
2014-01-17 14:15 ` [Qemu-devel] [PATCH v3 18/29] block: Make overlap range for serialisation dynamic Kevin Wolf
2014-01-22 20:15 ` Benoît Canet
2014-01-17 14:15 ` [Qemu-devel] [PATCH v3 19/29] block: Allow wait_serialising_requests() at any point Kevin Wolf
2014-01-22 20:21 ` Benoît Canet
2014-01-17 14:15 ` [Qemu-devel] [PATCH v3 20/29] block: Align requests in bdrv_co_do_pwritev() Kevin Wolf
2014-01-22 20:29 ` Benoît Canet [this message]
2014-01-17 14:15 ` [Qemu-devel] [PATCH v3 21/29] block: Assert serialisation assumptions in pwritev Kevin Wolf
2014-01-17 23:42 ` Max Reitz
2014-01-24 16:09 ` Benoît Canet
2014-01-24 16:18 ` Kevin Wolf
2014-01-17 14:15 ` [Qemu-devel] [PATCH v3 22/29] block: Change coroutine wrapper to byte granularity Kevin Wolf
2014-01-17 14:15 ` [Qemu-devel] [PATCH v3 23/29] block: Make bdrv_pread() a bdrv_prwv_co() wrapper Kevin Wolf
2014-01-17 14:15 ` [Qemu-devel] [PATCH v3 24/29] block: Make bdrv_pwrite() " Kevin Wolf
2014-01-17 23:43 ` Max Reitz
2014-01-17 14:15 ` [Qemu-devel] [PATCH v3 25/29] iscsi: Set bs->request_alignment Kevin Wolf
2014-01-24 16:29 ` Benoît Canet
2014-01-17 14:15 ` [Qemu-devel] [PATCH v3 26/29] blkdebug: Make required alignment configurable Kevin Wolf
2014-01-17 23:50 ` Max Reitz
2014-01-17 14:15 ` [Qemu-devel] [PATCH v3 27/29] qemu-io: New command 'sleep' Kevin Wolf
2014-01-17 23:55 ` Max Reitz
2014-01-20 9:58 ` Kevin Wolf
2014-01-17 14:15 ` [Qemu-devel] [PATCH v3 28/29] qemu-iotests: Test pwritev RMW logic Kevin Wolf
2014-01-18 16:01 ` Max Reitz
2014-01-20 9:44 ` Kevin Wolf
2014-01-17 14:15 ` [Qemu-devel] [PATCH v3 29/29] block: Switch bdrv_io_limits_intercept() to byte granularity Kevin Wolf
2014-01-17 23:59 ` Max Reitz
2014-01-22 20:30 ` [Qemu-devel] [PATCH v3 00/29] block: Support for 512b-on-4k emulation Christian Borntraeger
2014-01-23 10:29 ` Kevin Wolf
2014-01-23 11:12 ` Christian Borntraeger
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20140122202939.GF3053@irqsave.net \
--to=benoit.canet@irqsave.net \
--cc=kwolf@redhat.com \
--cc=mreitz@redhat.com \
--cc=pbonzini@redhat.com \
--cc=pl@kamp.de \
--cc=qemu-devel@nongnu.org \
--cc=stefanha@redhat.com \
--cc=xiawenc@linux.vnet.ibm.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).