From mboxrd@z Thu Jan  1 00:00:00 1970
Received: from eggs.gnu.org ([140.186.70.92]:47637)
	by lists.gnu.org with esmtp (Exim 4.71)
	(envelope-from <mtosatti@redhat.com>) id 1Qm6MT-000459-6D
	for qemu-devel@nongnu.org; Wed, 27 Jul 2011 11:49:31 -0400
Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71)
	(envelope-from <mtosatti@redhat.com>) id 1Qm6MQ-0000vL-DB
	for qemu-devel@nongnu.org; Wed, 27 Jul 2011 11:49:29 -0400
Received: from mx1.redhat.com ([209.132.183.28]:63046)
	by eggs.gnu.org with esmtp (Exim 4.71)
	(envelope-from <mtosatti@redhat.com>) id 1Qm6MP-0000vA-Ry
	for qemu-devel@nongnu.org; Wed, 27 Jul 2011 11:49:26 -0400
Date: Wed, 27 Jul 2011 12:49:13 -0300
From: Marcelo Tosatti <mtosatti@redhat.com>
Message-ID: <20110727154913.GA6334@amt.cnet>
References: <1311670746-20498-1-git-send-email-wuzhy@linux.vnet.ibm.com>
	<1311670746-20498-2-git-send-email-wuzhy@linux.vnet.ibm.com>
	<20110726192618.GA8126@amt.cnet>
	<CAEH94Liq-0jNH4mC5nCib+SBkYKoCO1vwy4LemiNQzWVSwYAsg@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
In-Reply-To: <CAEH94Liq-0jNH4mC5nCib+SBkYKoCO1vwy4LemiNQzWVSwYAsg@mail.gmail.com>
Content-Transfer-Encoding: quoted-printable
Subject: Re: [Qemu-devel] [PATCH v2 1/1] The codes V2 for QEMU disk I/O
	limits.
List-Id: <qemu-devel.nongnu.org>
List-Unsubscribe: <https://lists.nongnu.org/mailman/options/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=unsubscribe>
List-Archive: <http://lists.nongnu.org/archive/html/qemu-devel>
List-Post: <mailto:qemu-devel@nongnu.org>
List-Help: <mailto:qemu-devel-request@nongnu.org?subject=help>
List-Subscribe: <https://lists.nongnu.org/mailman/listinfo/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=subscribe>
To: Zhi Yong Wu <zwu.kernel@gmail.com>
Cc: kwolf@redhat.com, aliguori@us.ibm.com, stefanha@linux.vnet.ibm.com, kvm@vger.kernel.org, Zhi Yong Wu <wuzhy@linux.vnet.ibm.com>, qemu-devel@nongnu.org, ryanh@us.ibm.com, vgoyal@redhat.com

On Wed, Jul 27, 2011 at 06:17:15PM +0800, Zhi Yong Wu wrote:
> >> + =A0 =A0 =A0 =A0wait_time =3D 1;
> >> + =A0 =A0}
> >> +
> >> + =A0 =A0wait_time =3D wait_time + (slice_time - elapsed_time);
> >> + =A0 =A0if (wait) {
> >> + =A0 =A0 =A0 =A0*wait =3D wait_time * BLOCK_IO_SLICE_TIME * 10 + 1;
> >> + =A0 =A0}
> >
> > The guest can keep submitting requests where "wait_time =3D 1" above,
> > and the timer will be rearmed continuously in the future.

This is wrong.

>  Can't you
> > simply arm the timer to the next slice start? _Some_ data must be
> > transfered by then, anyway (and nothing can be transfered earlier tha=
n
> > that).

This is valid.

> Sorry, i have got what you mean. Can you elaborate in more detail?

Sorry, the bug i mentioned about timer being rearmed does not exist.

But arming the timer for the last request as its done is confusing/unnece=
ssary.

The timer processes queued requests, but the timer is armed accordingly
to the last queued request in the slice. For example, if a request is
submitted 1ms before the slice ends, the timer is armed 100ms +
(slice_time - elapsed_time) in the future.

> > Same for iops calculation below.
> >
> >> +
> >> + =A0 =A0return true;
> >> +}
> >> +
> >> +static bool bdrv_exceed_iops_limits(BlockDriverState *bs, bool is_w=
rite,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 double ela=
psed_time, uint64_t *wait) {
> >> + =A0 =A0uint64_t iops_limit =3D 0;
> >> + =A0 =A0double =A0 ios_limit, ios_disp;
> >> + =A0 =A0double =A0 slice_time =3D 0.1, wait_time;
> >> +
> >> + =A0 =A0if (bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]) {
> >> + =A0 =A0 =A0 =A0iops_limit =3D bs->io_limits.iops[BLOCK_IO_LIMIT_TO=
TAL];
> >> + =A0 =A0} else if (bs->io_limits.iops[is_write]) {
> >> + =A0 =A0 =A0 =A0iops_limit =3D bs->io_limits.iops[is_write];
> >> + =A0 =A0} else {
> >> + =A0 =A0 =A0 =A0if (wait) {
> >> + =A0 =A0 =A0 =A0 =A0 =A0*wait =3D 0;
> >> + =A0 =A0 =A0 =A0}
> >> +
> >> + =A0 =A0 =A0 =A0return false;
> >> + =A0 =A0}
> >> +
> >> + =A0 =A0ios_limit =3D iops_limit * slice_time;
> >> + =A0 =A0ios_disp =A0=3D bs->io_disps.ios[is_write];
> >> + =A0 =A0if (bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]) {
> >> + =A0 =A0 =A0 =A0ios_disp +=3D bs->io_disps.ios[!is_write];
> >> + =A0 =A0}
> >> +
> >> + =A0 =A0if (ios_disp + 1 <=3D ios_limit) {
> >> + =A0 =A0 if (wait) {
> >> + =A0 =A0 =A0 =A0 *wait =3D 0;
> >> + =A0 =A0 }
> >> +
> >> + =A0 =A0 =A0 =A0return false;
> >> + =A0 =A0}
> >> +
> >> + =A0 =A0/* Calc approx time to dispatch */
> >> + =A0 =A0wait_time =3D (ios_disp + 1) / iops_limit;
> >> + =A0 =A0if (wait_time > elapsed_time) {
> >> + =A0 =A0 wait_time =3D wait_time - elapsed_time;
> >> + =A0 =A0} else {
> >> + =A0 =A0 wait_time =3D 0;
> >> + =A0 =A0}
> >> +
> >> + =A0 =A0if (wait) {
> >> + =A0 =A0 =A0 =A0*wait =3D wait_time * BLOCK_IO_SLICE_TIME * 10 + 1;
> >> + =A0 =A0}
> >> +
> >> + =A0 =A0return true;
> >> +}
> >> +
> >> +static bool bdrv_exceed_io_limits(BlockDriverState *bs, int nb_sect=
ors,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 bool is_write,=
 uint64_t *wait) {
> >> + =A0 =A0int64_t =A0real_time;
> >> + =A0 =A0uint64_t bps_wait =3D 0, iops_wait =3D 0, max_wait;
> >> + =A0 =A0double =A0 elapsed_time;
> >> + =A0 =A0int =A0 =A0 =A0bps_ret, iops_ret;
> >> +
> >> + =A0 =A0real_time =3D qemu_get_clock_ns(vm_clock);
> >> + =A0 =A0if (bs->slice_start[is_write] + BLOCK_IO_SLICE_TIME <=3D re=
al_time) {
> >> + =A0 =A0 =A0 =A0bs->slice_start[is_write] =3D real_time;
> >> +
> >> + =A0 =A0 =A0 =A0bs->io_disps.bytes[is_write] =A0 =3D 0;
> >> + =A0 =A0 =A0 =A0bs->io_disps.bytes[!is_write] =A0=3D 0;
> >> +
> >> + =A0 =A0 =A0 =A0bs->io_disps.ios[is_write] =A0 =A0 =3D 0;
> >> + =A0 =A0 =A0 =A0bs->io_disps.ios[!is_write] =A0 =A0=3D 0;
> >> + =A0 =A0}
> >> +
> >> + =A0 =A0/* If a limit was exceeded, immediately queue this request =
*/
> >> + =A0 =A0if ((bs->req_from_queue =3D=3D false)
> >> + =A0 =A0 =A0 =A0&& !QTAILQ_EMPTY(&bs->block_queue->requests)) {
> >> + =A0 =A0 =A0 =A0if (bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]
> >> + =A0 =A0 =A0 =A0 =A0 =A0|| bs->io_limits.bps[is_write] || bs->io_li=
mits.iops[is_write]
> >> + =A0 =A0 =A0 =A0 =A0 =A0|| bs->io_limits.iops[BLOCK_IO_LIMIT_TOTAL]=
) {
> >> + =A0 =A0 =A0 =A0 =A0 =A0if (wait) {
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0*wait =3D 0;
> >> + =A0 =A0 =A0 =A0 =A0 =A0}
> >> +
> >> + =A0 =A0 =A0 =A0 =A0 =A0return true;
> >> + =A0 =A0 =A0 =A0}
> >> + =A0 =A0}
> >> +
> >> + =A0 =A0elapsed_time =A0=3D real_time - bs->slice_start[is_write];
> >> + =A0 =A0elapsed_time =A0/=3D (BLOCK_IO_SLICE_TIME * 10.0);
> >> +
> >> + =A0 =A0bps_ret =A0=3D bdrv_exceed_bps_limits(bs, nb_sectors,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0=
 =A0 =A0is_write, elapsed_time, &bps_wait);
> >> + =A0 =A0iops_ret =3D bdrv_exceed_iops_limits(bs, is_write,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0=
 =A0 =A0elapsed_time, &iops_wait);
> >> + =A0 =A0if (bps_ret || iops_ret) {
> >> + =A0 =A0 =A0 =A0max_wait =3D bps_wait > iops_wait ? bps_wait : iops=
_wait;
> >> + =A0 =A0 =A0 =A0if (wait) {
> >> + =A0 =A0 =A0 =A0 =A0 =A0*wait =3D max_wait;
> >> + =A0 =A0 =A0 =A0}
> >> +
> >> + =A0 =A0 =A0 =A0return true;
> >> + =A0 =A0}
> >> +
> >> + =A0 =A0if (wait) {
> >> + =A0 =A0 =A0 =A0*wait =3D 0;
> >> + =A0 =A0}
> >> +
> >> + =A0 =A0return false;
> >> +}
> >>
> >> =A0/**************************************************************/
> >> =A0/* async I/Os */
> >> @@ -2121,13 +2343,28 @@ BlockDriverAIOCB *bdrv_aio_readv(BlockDriver=
State *bs, int64_t sector_num,
> >> =A0{
> >> =A0 =A0 =A0BlockDriver *drv =3D bs->drv;
> >> =A0 =A0 =A0BlockDriverAIOCB *ret;
> >> + =A0 =A0uint64_t wait_time =3D 0;
> >>
> >> =A0 =A0 =A0trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque);
> >>
> >> - =A0 =A0if (!drv)
> >> - =A0 =A0 =A0 =A0return NULL;
> >> - =A0 =A0if (bdrv_check_request(bs, sector_num, nb_sectors))
> >> + =A0 =A0if (!drv || bdrv_check_request(bs, sector_num, nb_sectors))=
 {
> >> + =A0 =A0 =A0 =A0if (bdrv_io_limits_enable(&bs->io_limits)) {
> >> + =A0 =A0 =A0 =A0 =A0 =A0bs->req_from_queue =3D false;
> >> + =A0 =A0 =A0 =A0}
> >> =A0 =A0 =A0 =A0 =A0return NULL;
> >> + =A0 =A0}
> >> +
> >> + =A0 =A0/* throttling disk read I/O */
> >> + =A0 =A0if (bdrv_io_limits_enable(&bs->io_limits)) {
> >> + =A0 =A0 =A0 =A0if (bdrv_exceed_io_limits(bs, nb_sectors, false, &w=
ait_time)) {
> >> + =A0 =A0 =A0 =A0 =A0 =A0ret =3D qemu_block_queue_enqueue(bs->block_=
queue, bs, bdrv_aio_readv,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0sector_=
num, qiov, nb_sectors, cb, opaque);
> >> + =A0 =A0 =A0 =A0 =A0 =A0qemu_mod_timer(bs->block_timer,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 wait_time + qemu_get_c=
lock_ns(vm_clock));
> >> + =A0 =A0 =A0 =A0 =A0 =A0bs->req_from_queue =3D false;
> >> + =A0 =A0 =A0 =A0 =A0 =A0return ret;
> >> + =A0 =A0 =A0 =A0}
> >> + =A0 =A0}
> >>
> >> =A0 =A0 =A0ret =3D drv->bdrv_aio_readv(bs, sector_num, qiov, nb_sect=
ors,
> >> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0cb, o=
paque);
> >> @@ -2136,6 +2373,16 @@ BlockDriverAIOCB *bdrv_aio_readv(BlockDriverS=
tate *bs, int64_t sector_num,
> >> =A0 =A0 =A0 /* Update stats even though technically transfer has not=
 happened. */
> >> =A0 =A0 =A0 bs->rd_bytes +=3D (unsigned) nb_sectors * BDRV_SECTOR_SI=
ZE;
> >> =A0 =A0 =A0 bs->rd_ops ++;
> >> +
> >> + =A0 =A0 =A0 =A0if (bdrv_io_limits_enable(&bs->io_limits)) {
> >> + =A0 =A0 =A0 =A0 =A0 =A0bs->io_disps.bytes[BLOCK_IO_LIMIT_READ] +=3D
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0(unsign=
ed) nb_sectors * BDRV_SECTOR_SIZE;
> >> + =A0 =A0 =A0 =A0 =A0 =A0bs->io_disps.ios[BLOCK_IO_LIMIT_READ]++;
> >> + =A0 =A0 =A0 =A0}
> >> + =A0 =A0}
> >> +
> >> + =A0 =A0if (bdrv_io_limits_enable(&bs->io_limits)) {
> >> + =A0 =A0 =A0 =A0bs->req_from_queue =3D false;
> >> =A0 =A0 =A0}
> >>
> >> =A0 =A0 =A0return ret;
> >> @@ -2184,15 +2431,18 @@ BlockDriverAIOCB *bdrv_aio_writev(BlockDrive=
rState *bs, int64_t sector_num,
> >> =A0 =A0 =A0BlockDriver *drv =3D bs->drv;
> >> =A0 =A0 =A0BlockDriverAIOCB *ret;
> >> =A0 =A0 =A0BlockCompleteData *blk_cb_data;
> >> + =A0 =A0uint64_t wait_time =3D 0;
> >>
> >> =A0 =A0 =A0trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque)=
;
> >>
> >> - =A0 =A0if (!drv)
> >> - =A0 =A0 =A0 =A0return NULL;
> >> - =A0 =A0if (bs->read_only)
> >> - =A0 =A0 =A0 =A0return NULL;
> >> - =A0 =A0if (bdrv_check_request(bs, sector_num, nb_sectors))
> >> + =A0 =A0if (!drv || bs->read_only
> >> + =A0 =A0 =A0 =A0|| bdrv_check_request(bs, sector_num, nb_sectors)) =
{
> >> + =A0 =A0 =A0 =A0if (bdrv_io_limits_enable(&bs->io_limits)) {
> >> + =A0 =A0 =A0 =A0 =A0 =A0bs->req_from_queue =3D false;
> >> + =A0 =A0 =A0 =A0}
> >> +
> >> =A0 =A0 =A0 =A0 =A0return NULL;
> >> + =A0 =A0}
> >>
> >> =A0 =A0 =A0if (bs->dirty_bitmap) {
> >> =A0 =A0 =A0 =A0 =A0blk_cb_data =3D blk_dirty_cb_alloc(bs, sector_num=
, nb_sectors, cb,
> >> @@ -2201,6 +2451,18 @@ BlockDriverAIOCB *bdrv_aio_writev(BlockDriver=
State *bs, int64_t sector_num,
> >> =A0 =A0 =A0 =A0 =A0opaque =3D blk_cb_data;
> >> =A0 =A0 =A0}
> >>
> >> + =A0 =A0/* throttling disk write I/O */
> >> + =A0 =A0if (bdrv_io_limits_enable(&bs->io_limits)) {
> >> + =A0 =A0 =A0 =A0if (bdrv_exceed_io_limits(bs, nb_sectors, true, &wa=
it_time)) {
> >> + =A0 =A0 =A0 =A0 =A0 =A0ret =3D qemu_block_queue_enqueue(bs->block_=
queue, bs, bdrv_aio_writev,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0=
sector_num, qiov, nb_sectors, cb, opaque);
> >> + =A0 =A0 =A0 =A0 =A0 =A0qemu_mod_timer(bs->block_timer,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0=
wait_time + qemu_get_clock_ns(vm_clock));
> >> + =A0 =A0 =A0 =A0 =A0 =A0bs->req_from_queue =3D false;
> >> + =A0 =A0 =A0 =A0 =A0 =A0return ret;
> >> + =A0 =A0 =A0 =A0}
> >> + =A0 =A0}
> >> +
> >> =A0 =A0 =A0ret =3D drv->bdrv_aio_writev(bs, sector_num, qiov, nb_sec=
tors,
> >> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 cb, =
opaque);
> >>
> >> @@ -2211,6 +2473,16 @@ BlockDriverAIOCB *bdrv_aio_writev(BlockDriver=
State *bs, int64_t sector_num,
> >> =A0 =A0 =A0 =A0 =A0if (bs->wr_highest_sector < sector_num + nb_secto=
rs - 1) {
> >> =A0 =A0 =A0 =A0 =A0 =A0 =A0bs->wr_highest_sector =3D sector_num + nb=
_sectors - 1;
> >> =A0 =A0 =A0 =A0 =A0}
> >> +
> >> + =A0 =A0 =A0 =A0if (bdrv_io_limits_enable(&bs->io_limits)) {
> >> + =A0 =A0 =A0 =A0 =A0 =A0bs->io_disps.bytes[BLOCK_IO_LIMIT_WRITE] +=3D
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 (unsig=
ned) nb_sectors * BDRV_SECTOR_SIZE;
> >> + =A0 =A0 =A0 =A0 =A0 =A0bs->io_disps.ios[BLOCK_IO_LIMIT_WRITE]++;
> >> + =A0 =A0 =A0 =A0}
> >> + =A0 =A0}
> >> +
> >> + =A0 =A0if (bdrv_io_limits_enable(&bs->io_limits)) {
> >> + =A0 =A0 =A0 =A0bs->req_from_queue =3D false;
> >> =A0 =A0 =A0}
> >>
> >> =A0 =A0 =A0return ret;
> >> diff --git a/block.h b/block.h
> >> index 859d1d9..f0dac62 100644
> >> --- a/block.h
> >> +++ b/block.h
> >> @@ -97,7 +97,6 @@ int bdrv_change_backing_file(BlockDriverState *bs,
> >> =A0 =A0 =A0const char *backing_file, const char *backing_fmt);
> >> =A0void bdrv_register(BlockDriver *bdrv);
> >>
> >> -
> >> =A0typedef struct BdrvCheckResult {
> >> =A0 =A0 =A0int corruptions;
> >> =A0 =A0 =A0int leaks;
> >> diff --git a/block/blk-queue.c b/block/blk-queue.c
> >> new file mode 100644
> >> index 0000000..09fcfe9
> >> --- /dev/null
> >> +++ b/block/blk-queue.c
> >> @@ -0,0 +1,116 @@
> >> +/*
> >> + * QEMU System Emulator queue definition for block layer
> >> + *
> >> + * Copyright (c) 2011 Zhi Yong Wu =A0<zwu.kernel@gmail.com>
> >> + *
> >> + * Permission is hereby granted, free of charge, to any person obta=
ining a copy
> >> + * of this software and associated documentation files (the "Softwa=
re"), to deal
> >> + * in the Software without restriction, including without limitatio=
n the rights
> >> + * to use, copy, modify, merge, publish, distribute, sublicense, an=
d/or sell
> >> + * copies of the Software, and to permit persons to whom the Softwa=
re is
> >> + * furnished to do so, subject to the following conditions:
> >> + *
> >> + * The above copyright notice and this permission notice shall be i=
ncluded in
> >> + * all copies or substantial portions of the Software.
> >> + *
> >> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, =
EXPRESS OR
> >> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANT=
ABILITY,
> >> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVEN=
T SHALL
> >> + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGE=
S OR OTHER
> >> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, =
ARISING FROM,
> >> + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DE=
ALINGS IN
> >> + * THE SOFTWARE.
> >> + */
> >> +
> >> +#include "block_int.h"
> >> +#include "qemu-queue.h"
> >> +#include "block/blk-queue.h"
> >> +
> >> +/* The APIs for block request queue on qemu block layer.
> >> + */
> >> +
> >> +static void qemu_block_queue_cancel(BlockDriverAIOCB *acb)
> >> +{
> >> + =A0 =A0qemu_aio_release(acb);
> >> +}
> >> +
> >> +static AIOPool block_queue_pool =3D {
> >> + =A0 =A0.aiocb_size =A0 =A0 =A0 =A0 =3D sizeof(struct BlockDriverAI=
OCB),
> >> + =A0 =A0.cancel =A0 =A0 =A0 =A0 =A0 =A0 =3D qemu_block_queue_cancel=
,
> >> +};
> >> +
> >> +static void qemu_block_queue_callback(void *opaque, int ret)
> >> +{
> >> + =A0 =A0BlockDriverAIOCB *acb =3D opaque;
> >> +
> >> + =A0 =A0qemu_aio_release(acb);
> >> +}
> >> +
> >> +BlockQueue *qemu_new_block_queue(void)
> >> +{
> >> + =A0 =A0BlockQueue *queue;
> >> +
> >> + =A0 =A0queue =3D qemu_mallocz(sizeof(BlockQueue));
> >> +
> >> + =A0 =A0QTAILQ_INIT(&queue->requests);
> >> +
> >> + =A0 =A0return queue;
> >> +}
> >> +
> >> +void qemu_del_block_queue(BlockQueue *queue)
> >> +{
> >> + =A0 =A0BlockIORequest *request, *next;
> >> +
> >> + =A0 =A0QTAILQ_FOREACH_SAFE(request, &queue->requests, entry, next)=
 {
> >> + =A0 =A0 =A0 =A0QTAILQ_REMOVE(&queue->requests, request, entry);
> >> + =A0 =A0 =A0 =A0qemu_free(request);
> >> + =A0 =A0}
> >> +
> >> + =A0 =A0qemu_free(queue);
> >> +}
> >> +
> >> +BlockDriverAIOCB *qemu_block_queue_enqueue(BlockQueue *queue,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 BlockDriverState *bs,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 BlockRequestHandler *handl=
er,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 int64_t sector_num,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 QEMUIOVector *qiov,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 int nb_sectors,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 BlockDriverCompletionFunc =
*cb,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 void *opaque)
> >> +{
> >> + =A0 =A0BlockIORequest *request;
> >> + =A0 =A0BlockDriverAIOCB *acb;
> >> +
> >> + =A0 =A0request =3D qemu_malloc(sizeof(BlockIORequest));
> >> + =A0 =A0request->bs =3D bs;
> >> + =A0 =A0request->handler =3D handler;
> >> + =A0 =A0request->sector_num =3D sector_num;
> >> + =A0 =A0request->qiov =3D qiov;
> >> + =A0 =A0request->nb_sectors =3D nb_sectors;
> >> + =A0 =A0request->cb =3D cb;
> >> + =A0 =A0request->opaque =3D opaque;
> >> +
> >> + =A0 =A0QTAILQ_INSERT_TAIL(&queue->requests, request, entry);
> >> +
> >> + =A0 =A0acb =3D qemu_aio_get(&block_queue_pool, bs,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 qemu_block_queue_callb=
ack, opaque);
> >> +
> >> + =A0 =A0return acb;
> >> +}
> >> +
> >> +int qemu_block_queue_handler(BlockIORequest *request)
> >> +{
> >> + =A0 =A0int ret;
> >> + =A0 =A0BlockDriverAIOCB *res;
> >> +
> >> + =A0 =A0/* indicate this req is from block queue */
> >> + =A0 =A0request->bs->req_from_queue =3D true;
> >> +
> >> + =A0 =A0res =3D request->handler(request->bs, request->sector_num,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 request->q=
iov, request->nb_sectors,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 request->c=
b, request->opaque);
> >> +
> >> + =A0 =A0ret =3D (res =3D=3D NULL) ? 0 : 1;
> >> +
> >> + =A0 =A0return ret;
> >> +}
> >> diff --git a/block/blk-queue.h b/block/blk-queue.h
> >> new file mode 100644
> >> index 0000000..47f8a36
> >> --- /dev/null
> >> +++ b/block/blk-queue.h
> >> @@ -0,0 +1,70 @@
> >> +/*
> >> + * QEMU System Emulator queue declaration for block layer
> >> + *
> >> + * Copyright (c) 2011 Zhi Yong Wu =A0<zwu.kernel@gmail.com>
> >> + *
> >> + * Permission is hereby granted, free of charge, to any person obta=
ining a copy
> >> + * of this software and associated documentation files (the "Softwa=
re"), to deal
> >> + * in the Software without restriction, including without limitatio=
n the rights
> >> + * to use, copy, modify, merge, publish, distribute, sublicense, an=
d/or sell
> >> + * copies of the Software, and to permit persons to whom the Softwa=
re is
> >> + * furnished to do so, subject to the following conditions:
> >> + *
> >> + * The above copyright notice and this permission notice shall be i=
ncluded in
> >> + * all copies or substantial portions of the Software.
> >> + *
> >> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, =
EXPRESS OR
> >> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANT=
ABILITY,
> >> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVEN=
T SHALL
> >> + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGE=
S OR OTHER
> >> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, =
ARISING FROM,
> >> + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DE=
ALINGS IN
> >> + * THE SOFTWARE.
> >> + */
> >> +
> >> +#ifndef QEMU_BLOCK_QUEUE_H
> >> +#define QEMU_BLOCK_QUEUE_H
> >> +
> >> +#include "block.h"
> >> +#include "qemu-queue.h"
> >> +#include "qemu-common.h"
> >> +
> >> +typedef BlockDriverAIOCB* (BlockRequestHandler) (BlockDriverState *=
bs,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0int=
64_t sector_num, QEMUIOVector *qiov,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0int=
 nb_sectors, BlockDriverCompletionFunc *cb,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0voi=
d *opaque);
> >> +
> >> +struct BlockIORequest {
> >> + =A0 =A0QTAILQ_ENTRY(BlockIORequest) entry;
> >> + =A0 =A0BlockDriverState *bs;
> >> + =A0 =A0BlockRequestHandler *handler;
> >> + =A0 =A0int64_t sector_num;
> >> + =A0 =A0QEMUIOVector *qiov;
> >> + =A0 =A0int nb_sectors;
> >> + =A0 =A0BlockDriverCompletionFunc *cb;
> >> + =A0 =A0void *opaque;
> >> +};
> >> +
> >> +typedef struct BlockIORequest BlockIORequest;
> >> +
> >> +struct BlockQueue {
> >> + =A0 =A0QTAILQ_HEAD(requests, BlockIORequest) requests;
> >> +};
> >> +
> >> +typedef struct BlockQueue BlockQueue;
> >> +
> >> +BlockQueue *qemu_new_block_queue(void);
> >> +
> >> +void qemu_del_block_queue(BlockQueue *queue);
> >> +
> >> +BlockDriverAIOCB *qemu_block_queue_enqueue(BlockQueue *queue,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0BlockDriverState *b=
s,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0BlockRequestHandler=
 *handler,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0int64_t sector_num,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0QEMUIOVector *qiov,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0int nb_sectors,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0BlockDriverCompleti=
onFunc *cb,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0void *opaque);
> >> +
> >> +int qemu_block_queue_handler(BlockIORequest *request);
> >> +#endif /* QEMU_BLOCK_QUEUE_H */
> >> diff --git a/block_int.h b/block_int.h
> >> index 1e265d2..1587171 100644
> >> --- a/block_int.h
> >> +++ b/block_int.h
> >> @@ -27,10 +27,17 @@
> >> =A0#include "block.h"
> >> =A0#include "qemu-option.h"
> >> =A0#include "qemu-queue.h"
> >> +#include "block/blk-queue.h"
> >>
> >> =A0#define BLOCK_FLAG_ENCRYPT =A0 1
> >> =A0#define BLOCK_FLAG_COMPAT6 =A0 4
> >>
> >> +#define BLOCK_IO_LIMIT_READ =A0 =A0 0
> >> +#define BLOCK_IO_LIMIT_WRITE =A0 =A01
> >> +#define BLOCK_IO_LIMIT_TOTAL =A0 =A02
> >> +
> >> +#define BLOCK_IO_SLICE_TIME =A0 =A0 100000000
> >> +
> >> =A0#define BLOCK_OPT_SIZE =A0 =A0 =A0 =A0 =A0"size"
> >> =A0#define BLOCK_OPT_ENCRYPT =A0 =A0 =A0 "encryption"
> >> =A0#define BLOCK_OPT_COMPAT6 =A0 =A0 =A0 "compat6"
> >> @@ -46,6 +53,16 @@ typedef struct AIOPool {
> >> =A0 =A0 =A0BlockDriverAIOCB *free_aiocb;
> >> =A0} AIOPool;
> >>
> >> +typedef struct BlockIOLimit {
> >> + =A0 =A0uint64_t bps[3];
> >> + =A0 =A0uint64_t iops[3];
> >> +} BlockIOLimit;
> >> +
> >> +typedef struct BlockIODisp {
> >> + =A0 =A0uint64_t bytes[2];
> >> + =A0 =A0uint64_t ios[2];
> >> +} BlockIODisp;
> >> +
> >> =A0struct BlockDriver {
> >> =A0 =A0 =A0const char *format_name;
> >> =A0 =A0 =A0int instance_size;
> >> @@ -175,6 +192,14 @@ struct BlockDriverState {
> >>
> >> =A0 =A0 =A0void *sync_aiocb;
> >>
> >> + =A0 =A0/* the time for latest disk I/O */
> >> + =A0 =A0int64_t slice_start[2];
> >> + =A0 =A0BlockIOLimit io_limits;
> >> + =A0 =A0BlockIODisp =A0io_disps;
> >> + =A0 =A0BlockQueue =A0 *block_queue;
> >> + =A0 =A0QEMUTimer =A0 =A0*block_timer;
> >> + =A0 =A0bool =A0 =A0 =A0 =A0 req_from_queue;
> >> +
> >> =A0 =A0 =A0/* I/O stats (display with "info blockstats"). */
> >> =A0 =A0 =A0uint64_t rd_bytes;
> >> =A0 =A0 =A0uint64_t wr_bytes;
> >> @@ -222,6 +247,9 @@ void qemu_aio_release(void *p);
> >>
> >> =A0void *qemu_blockalign(BlockDriverState *bs, size_t size);
> >>
> >> +void bdrv_set_io_limits(BlockDriverState *bs,
> >> + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0BlockIOLimi=
t *io_limits);
> >> +
> >> =A0#ifdef _WIN32
> >> =A0int is_windows_drive(const char *filename);
> >> =A0#endif
> >> diff --git a/blockdev.c b/blockdev.c
> >> index c263663..45602f4 100644
> >> --- a/blockdev.c
> >> +++ b/blockdev.c
> >> @@ -238,6 +238,9 @@ DriveInfo *drive_init(QemuOpts *opts, int defaul=
t_to_scsi)
> >> =A0 =A0 =A0int on_read_error, on_write_error;
> >> =A0 =A0 =A0const char *devaddr;
> >> =A0 =A0 =A0DriveInfo *dinfo;
> >> + =A0 =A0BlockIOLimit io_limits;
> >> + =A0 =A0bool iol_flag =3D false;
> >> + =A0 =A0const char *iol_opts[7] =3D {"bps", "bps_rd", "bps_wr", "io=
ps", "iops_rd", "iops_wr"};
> >> =A0 =A0 =A0int is_extboot =3D 0;
> >> =A0 =A0 =A0int snapshot =3D 0;
> >> =A0 =A0 =A0int ret;
> >> @@ -372,6 +375,19 @@ DriveInfo *drive_init(QemuOpts *opts, int defau=
lt_to_scsi)
> >> =A0 =A0 =A0 =A0 =A0return NULL;
> >> =A0 =A0 =A0}
> >>
> >> + =A0 =A0/* disk io limits */
> >> + =A0 =A0iol_flag =3D qemu_opt_io_limits_enable_flag(opts, iol_opts)=
;
> >> + =A0 =A0if (iol_flag) {
> >> + =A0 =A0 =A0 =A0memset(&io_limits, 0, sizeof(BlockIOLimit));
> >> +
> >> + =A0 =A0 =A0 =A0io_limits.bps[2] =A0=3D qemu_opt_get_number(opts, "=
bps", 0);
> >> + =A0 =A0 =A0 =A0io_limits.bps[0] =A0=3D qemu_opt_get_number(opts, "=
bps_rd", 0);
> >> + =A0 =A0 =A0 =A0io_limits.bps[1] =A0=3D qemu_opt_get_number(opts, "=
bps_wr", 0);
> >> + =A0 =A0 =A0 =A0io_limits.iops[2] =3D qemu_opt_get_number(opts, "io=
ps", 0);
> >> + =A0 =A0 =A0 =A0io_limits.iops[0] =3D qemu_opt_get_number(opts, "io=
ps_rd", 0);
> >> + =A0 =A0 =A0 =A0io_limits.iops[1] =3D qemu_opt_get_number(opts, "io=
ps_wr", 0);
> >> + =A0 =A0}
> >> +
> >> =A0 =A0 =A0on_write_error =3D BLOCK_ERR_STOP_ENOSPC;
> >> =A0 =A0 =A0if ((buf =3D qemu_opt_get(opts, "werror")) !=3D NULL) {
> >> =A0 =A0 =A0 =A0 =A0if (type !=3D IF_IDE && type !=3D IF_SCSI && type=
 !=3D IF_VIRTIO && type !=3D IF_NONE) {
> >> @@ -483,6 +499,11 @@ DriveInfo *drive_init(QemuOpts *opts, int defau=
lt_to_scsi)
> >>
> >> =A0 =A0 =A0bdrv_set_on_error(dinfo->bdrv, on_read_error, on_write_er=
ror);
> >>
> >> + =A0 =A0/* throttling disk io limits */
> >> + =A0 =A0if (iol_flag) {
> >> + =A0 =A0 =A0 =A0bdrv_set_io_limits(dinfo->bdrv, &io_limits);
> >> + =A0 =A0}
> >> +
> >> =A0 =A0 =A0switch(type) {
> >> =A0 =A0 =A0case IF_IDE:
> >> =A0 =A0 =A0case IF_SCSI:
> >> diff --git a/qemu-config.c b/qemu-config.c
> >> index efa892c..9232bbb 100644
> >> --- a/qemu-config.c
> >> +++ b/qemu-config.c
> >> @@ -82,6 +82,30 @@ static QemuOptsList qemu_drive_opts =3D {
> >> =A0 =A0 =A0 =A0 =A0 =A0 =A0.name =3D "boot",
> >> =A0 =A0 =A0 =A0 =A0 =A0 =A0.type =3D QEMU_OPT_BOOL,
> >> =A0 =A0 =A0 =A0 =A0 =A0 =A0.help =3D "make this a boot drive",
> >> + =A0 =A0 =A0 =A0},{
> >> + =A0 =A0 =A0 =A0 =A0 =A0.name =3D "iops",
> >> + =A0 =A0 =A0 =A0 =A0 =A0.type =3D QEMU_OPT_NUMBER,
> >> + =A0 =A0 =A0 =A0 =A0 =A0.help =3D "limit total I/O operations per s=
econd",
> >> + =A0 =A0 =A0 =A0},{
> >> + =A0 =A0 =A0 =A0 =A0 =A0.name =3D "iops_rd",
> >> + =A0 =A0 =A0 =A0 =A0 =A0.type =3D QEMU_OPT_NUMBER,
> >> + =A0 =A0 =A0 =A0 =A0 =A0.help =3D "limit read operations per second=
",
> >> + =A0 =A0 =A0 =A0},{
> >> + =A0 =A0 =A0 =A0 =A0 =A0.name =3D "iops_wr",
> >> + =A0 =A0 =A0 =A0 =A0 =A0.type =3D QEMU_OPT_NUMBER,
> >> + =A0 =A0 =A0 =A0 =A0 =A0.help =3D "limit write operations per secon=
d",
> >> + =A0 =A0 =A0 =A0},{
> >> + =A0 =A0 =A0 =A0 =A0 =A0.name =3D "bps",
> >> + =A0 =A0 =A0 =A0 =A0 =A0.type =3D QEMU_OPT_NUMBER,
> >> + =A0 =A0 =A0 =A0 =A0 =A0.help =3D "limit total bytes per second",
> >> + =A0 =A0 =A0 =A0},{
> >> + =A0 =A0 =A0 =A0 =A0 =A0.name =3D "bps_rd",
> >> + =A0 =A0 =A0 =A0 =A0 =A0.type =3D QEMU_OPT_NUMBER,
> >> + =A0 =A0 =A0 =A0 =A0 =A0.help =3D "limit read bytes per second",
> >> + =A0 =A0 =A0 =A0},{
> >> + =A0 =A0 =A0 =A0 =A0 =A0.name =3D "bps_wr",
> >> + =A0 =A0 =A0 =A0 =A0 =A0.type =3D QEMU_OPT_NUMBER,
> >> + =A0 =A0 =A0 =A0 =A0 =A0.help =3D "limit write bytes per second",
> >> =A0 =A0 =A0 =A0 =A0},
> >> =A0 =A0 =A0 =A0 =A0{ /* end of list */ }
> >> =A0 =A0 =A0},
> >> diff --git a/qemu-option.c b/qemu-option.c
> >> index 65db542..9fe234d 100644
> >> --- a/qemu-option.c
> >> +++ b/qemu-option.c
> >> @@ -562,6 +562,23 @@ uint64_t qemu_opt_get_number(QemuOpts *opts, co=
nst char *name, uint64_t defval)
> >> =A0 =A0 =A0return opt->value.uint;
> >> =A0}
> >>
> >> +bool qemu_opt_io_limits_enable_flag(QemuOpts *opts, const char **io=
l_opts)
> >> +{
> >> + =A0 =A0 int i;
> >> + =A0 =A0 uint64_t opt_val =A0 =3D 0;
> >> + =A0 =A0 bool iol_flag =3D false;
> >> +
> >> + =A0 =A0 for (i =3D 0; iol_opts[i]; i++) {
> >> + =A0 =A0 =A0opt_val =3D qemu_opt_get_number(opts, iol_opts[i], 0);
> >> + =A0 =A0 =A0if (opt_val !=3D 0) {
> >> + =A0 =A0 =A0 =A0 =A0iol_flag =3D true;
> >> + =A0 =A0 =A0 =A0 =A0break;
> >> + =A0 =A0 =A0}
> >> + =A0 =A0 }
> >> +
> >> + =A0 =A0 return iol_flag;
> >> +}
> >> +
> >> =A0uint64_t qemu_opt_get_size(QemuOpts *opts, const char *name, uint=
64_t defval)
> >> =A0{
> >> =A0 =A0 =A0QemuOpt *opt =3D qemu_opt_find(opts, name);
> >> diff --git a/qemu-option.h b/qemu-option.h
> >> index b515813..fc909f9 100644
> >> --- a/qemu-option.h
> >> +++ b/qemu-option.h
> >> @@ -107,6 +107,7 @@ struct QemuOptsList {
> >> =A0const char *qemu_opt_get(QemuOpts *opts, const char *name);
> >> =A0int qemu_opt_get_bool(QemuOpts *opts, const char *name, int defva=
l);
> >> =A0uint64_t qemu_opt_get_number(QemuOpts *opts, const char *name, ui=
nt64_t defval);
> >> +bool qemu_opt_io_limits_enable_flag(QemuOpts *opts, const char **io=
l_opts);
> >> =A0uint64_t qemu_opt_get_size(QemuOpts *opts, const char *name, uint=
64_t defval);
> >> =A0int qemu_opt_set(QemuOpts *opts, const char *name, const char *va=
lue);
> >> =A0typedef int (*qemu_opt_loopfunc)(const char *name, const char *va=
lue, void *opaque);
> >> diff --git a/qemu-options.hx b/qemu-options.hx
> >> index cb3347e..ae219f5 100644
> >> --- a/qemu-options.hx
> >> +++ b/qemu-options.hx
> >> @@ -121,6 +121,7 @@ DEF("drive", HAS_ARG, QEMU_OPTION_drive,
> >> =A0 =A0 =A0" =A0 =A0 =A0 [,cache=3Dwritethrough|writeback|none|unsaf=
e][,format=3Df]\n"
> >> =A0 =A0 =A0" =A0 =A0 =A0 [,serial=3Ds][,addr=3DA][,id=3Dname][,aio=3D=
threads|native]\n"
> >> =A0 =A0 =A0" =A0 =A0 =A0 [,readonly=3Don|off][,boot=3Don|off]\n"
> >> + =A0 =A0" =A0 =A0 =A0 [[,bps=3Db]|[[,bps_rd=3Dr][,bps_wr=3Dw]]][[,i=
ops=3Di]|[[,iops_rd=3Dr][,iops_wr=3Dw]]\n"
> >> =A0 =A0 =A0" =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0use 'file' as a drive im=
age\n", QEMU_ARCH_ALL)
> >> =A0STEXI
> >> =A0@item -drive @var{option}[,@var{option}[,@var{option}[,...]]]
> >> --
> >> 1.7.2.3
> >>
> >> --
> >> To unsubscribe from this list: send the line "unsubscribe kvm" in
> >> the body of a message to majordomo@vger.kernel.org
> >> More majordomo info at =A0http://vger.kernel.org/majordomo-info.html
> >
>=20
>=20
>=20
> --=20
> Regards,
>=20
> Zhi Yong Wu