From: Kevin Wolf <kwolf@redhat.com>
To: qemu-block@nongnu.org
Cc: kwolf@redhat.com, mreitz@redhat.com, pbonzini@redhat.com,
famz@redhat.com, stefanha@redhat.com, qemu-devel@nongnu.org
Subject: [Qemu-devel] [PATCH 06/19] block: Avoid unnecessary aio_poll() in AIO_WAIT_WHILE()
Date: Wed, 11 Apr 2018 18:39:27 +0200 [thread overview]
Message-ID: <20180411163940.2523-7-kwolf@redhat.com> (raw)
In-Reply-To: <20180411163940.2523-1-kwolf@redhat.com>
Commit 91af091f923 added an additional aio_poll() to BDRV_POLL_WHILE()
in order to make sure that all pending BHs are executed on drain. This
was the wrong place to make the fix, as it is useless overhead for all
other users of the macro and unnecessarily complicates the mechanism.
This patch effectively reverts said commit (the context has changed a
bit and the code has moved to AIO_WAIT_WHILE()) and instead polls in the
loop condition for drain.
The effect is probably hard to measure in any real-world use case
because actual I/O will dominate, but if I run only the initialisation
part of 'qemu-img convert' where it calls bdrv_block_status() for the
whole image to find out how much data there is copy, this phase actually
needs only roughly half the time after this patch.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
---
include/block/aio-wait.h | 22 ++++++++--------------
block/io.c | 11 ++++++++++-
2 files changed, 18 insertions(+), 15 deletions(-)
diff --git a/include/block/aio-wait.h b/include/block/aio-wait.h
index 8c90a2e66e..783d3678dd 100644
--- a/include/block/aio-wait.h
+++ b/include/block/aio-wait.h
@@ -73,29 +73,23 @@ typedef struct {
*/
#define AIO_WAIT_WHILE(wait, ctx, cond) ({ \
bool waited_ = false; \
- bool busy_ = true; \
AioWait *wait_ = (wait); \
AioContext *ctx_ = (ctx); \
if (in_aio_context_home_thread(ctx_)) { \
- while ((cond) || busy_) { \
- busy_ = aio_poll(ctx_, (cond)); \
- waited_ |= !!(cond) | busy_; \
+ while ((cond)) { \
+ aio_poll(ctx_, true); \
+ waited_ = true; \
} \
} else { \
assert(qemu_get_current_aio_context() == \
qemu_get_aio_context()); \
/* Increment wait_->num_waiters before evaluating cond. */ \
atomic_inc(&wait_->num_waiters); \
- while (busy_) { \
- if ((cond)) { \
- waited_ = busy_ = true; \
- aio_context_release(ctx_); \
- aio_poll(qemu_get_aio_context(), true); \
- aio_context_acquire(ctx_); \
- } else { \
- busy_ = aio_poll(ctx_, false); \
- waited_ |= busy_; \
- } \
+ while ((cond)) { \
+ aio_context_release(ctx_); \
+ aio_poll(qemu_get_aio_context(), true); \
+ aio_context_acquire(ctx_); \
+ waited_ = true; \
} \
atomic_dec(&wait_->num_waiters); \
} \
diff --git a/block/io.c b/block/io.c
index ea6f9f023a..6f580f49ff 100644
--- a/block/io.c
+++ b/block/io.c
@@ -181,13 +181,22 @@ static void bdrv_drain_invoke(BlockDriverState *bs, bool begin)
BDRV_POLL_WHILE(bs, !data.done);
}
+/* Returns true if BDRV_POLL_WHILE() should go into a blocking aio_poll() */
+static bool bdrv_drain_poll(BlockDriverState *bs)
+{
+ /* Execute pending BHs first and check everything else only after the BHs
+ * have executed. */
+ while (aio_poll(bs->aio_context, false));
+ return atomic_read(&bs->in_flight);
+}
+
static bool bdrv_drain_recurse(BlockDriverState *bs)
{
BdrvChild *child, *tmp;
bool waited;
/* Wait for drained requests to finish */
- waited = BDRV_POLL_WHILE(bs, atomic_read(&bs->in_flight) > 0);
+ waited = BDRV_POLL_WHILE(bs, bdrv_drain_poll(bs));
QLIST_FOREACH_SAFE(child, &bs->children, next, tmp) {
BlockDriverState *bs = child->bs;
--
2.13.6
next prev parent reply other threads:[~2018-04-11 16:40 UTC|newest]
Thread overview: 52+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-04-11 16:39 [Qemu-devel] [PATCH 00/19] Drain fixes and cleanups, part 3 Kevin Wolf
2018-04-11 16:39 ` [Qemu-devel] [PATCH 01/19] test-bdrv-drain: bdrv_drain() works with cross-AioContext events Kevin Wolf
2018-04-11 16:39 ` [Qemu-devel] [PATCH 02/19] block: Use bdrv_do_drain_begin/end in bdrv_drain_all() Kevin Wolf
2018-04-20 7:07 ` [Qemu-devel] [Qemu-block] " Stefan Hajnoczi
2018-04-11 16:39 ` [Qemu-devel] [PATCH 03/19] block: Remove 'recursive' parameter from bdrv_drain_invoke() Kevin Wolf
2018-04-20 7:09 ` [Qemu-devel] [Qemu-block] " Stefan Hajnoczi
2018-04-11 16:39 ` [Qemu-devel] [PATCH 04/19] block: Don't manually poll in bdrv_drain_all() Kevin Wolf
2018-04-11 18:32 ` Eric Blake
2018-04-20 7:11 ` [Qemu-devel] [Qemu-block] " Stefan Hajnoczi
2018-04-11 16:39 ` [Qemu-devel] [PATCH 05/19] tests/test-bdrv-drain: bdrv_drain_all() works in coroutines now Kevin Wolf
2018-04-11 18:33 ` Eric Blake
2018-04-20 7:12 ` [Qemu-devel] [Qemu-block] " Stefan Hajnoczi
2018-04-11 16:39 ` Kevin Wolf [this message]
2018-04-11 17:33 ` [Qemu-devel] [PATCH 06/19] block: Avoid unnecessary aio_poll() in AIO_WAIT_WHILE() Su Hang
2018-04-20 7:17 ` [Qemu-devel] [Qemu-block] " Stefan Hajnoczi
2018-04-11 16:39 ` [Qemu-devel] [PATCH 07/19] block: Really pause block jobs on drain Kevin Wolf
2018-04-12 8:37 ` Paolo Bonzini
2018-04-12 9:51 ` Kevin Wolf
2018-04-12 10:12 ` Paolo Bonzini
2018-04-12 11:11 ` Kevin Wolf
2018-04-12 11:30 ` Paolo Bonzini
2018-04-12 11:53 ` Kevin Wolf
2018-04-12 12:02 ` Paolo Bonzini
2018-04-12 13:27 ` Kevin Wolf
2018-04-12 13:42 ` Paolo Bonzini
2018-04-12 14:25 ` Kevin Wolf
2018-04-12 20:44 ` Paolo Bonzini
2018-04-13 8:01 ` Kevin Wolf
2018-04-13 11:05 ` [Qemu-devel] [Qemu-block] " Paolo Bonzini
2018-04-13 12:40 ` Kevin Wolf
2018-04-11 16:39 ` [Qemu-devel] [PATCH 08/19] block: Remove bdrv_drain_recurse() Kevin Wolf
2018-04-12 8:39 ` Paolo Bonzini
2018-04-20 7:20 ` [Qemu-devel] [Qemu-block] " Stefan Hajnoczi
2018-04-11 16:39 ` [Qemu-devel] [PATCH 09/19] test-bdrv-drain: Add test for node deletion Kevin Wolf
2018-04-20 7:32 ` [Qemu-devel] [Qemu-block] " Stefan Hajnoczi
2018-04-11 16:39 ` [Qemu-devel] [PATCH 10/19] block: Drain recursively with a single BDRV_POLL_WHILE() Kevin Wolf
2018-04-12 8:41 ` Paolo Bonzini
2018-04-11 16:39 ` [Qemu-devel] [PATCH 11/19] test-bdrv-drain: Test node deletion in subtree recursion Kevin Wolf
2018-04-11 16:39 ` [Qemu-devel] [PATCH 12/19] block: Don't poll in parent drain callbacks Kevin Wolf
2018-04-11 16:39 ` [Qemu-devel] [PATCH 13/19] test-bdrv-drain: Graph change through parent callback Kevin Wolf
2018-04-11 16:39 ` [Qemu-devel] [PATCH 14/19] block: Defer .bdrv_drain_begin callback to polling phase Kevin Wolf
2018-06-27 14:30 ` Max Reitz
2018-06-29 15:14 ` Kevin Wolf
2018-04-11 16:39 ` [Qemu-devel] [PATCH 15/19] test-bdrv-drain: Test that bdrv_drain_invoke() doesn't poll Kevin Wolf
2018-04-11 16:39 ` [Qemu-devel] [PATCH 16/19] block: Allow AIO_WAIT_WHILE with NULL ctx Kevin Wolf
2018-04-12 8:43 ` Paolo Bonzini
2018-04-11 16:39 ` [Qemu-devel] [PATCH 17/19] block: Move bdrv_drain_all_begin() out of coroutine context Kevin Wolf
2018-04-11 16:39 ` [Qemu-devel] [PATCH 18/19] block: Allow graph changes in bdrv_drain_all_begin/end sections Kevin Wolf
2018-04-12 8:47 ` Paolo Bonzini
2018-04-11 16:39 ` [Qemu-devel] [PATCH 19/19] test-bdrv-drain: Test graph changes in drain_all section Kevin Wolf
2018-04-11 17:05 ` [Qemu-devel] [PATCH 00/19] Drain fixes and cleanups, part 3 no-reply
2018-04-20 7:35 ` Stefan Hajnoczi
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=20180411163940.2523-7-kwolf@redhat.com \
--to=kwolf@redhat.com \
--cc=famz@redhat.com \
--cc=mreitz@redhat.com \
--cc=pbonzini@redhat.com \
--cc=qemu-block@nongnu.org \
--cc=qemu-devel@nongnu.org \
--cc=stefanha@redhat.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).