From: Stefan Hajnoczi <stefanha@redhat.com>
To: qemu-devel@nongnu.org, Peter Maydell <peter.maydell@linaro.org>
Cc: Fam Zheng <fam@euphon.net>, Paolo Bonzini <pbonzini@redhat.com>,
Jason Wang <jasowang@redhat.com>,
qemu-block@nongnu.org, Stefan Hajnoczi <stefanha@redhat.com>
Subject: [PULL for-7.0 2/2] aio-posix: fix spurious ->poll_ready() callbacks in main loop
Date: Thu, 17 Mar 2022 16:57:43 +0000 [thread overview]
Message-ID: <20220317165743.238662-3-stefanha@redhat.com> (raw)
In-Reply-To: <20220317165743.238662-1-stefanha@redhat.com>
When ->poll() succeeds the AioHandler is placed on the ready list with
revents set to the magic value 0. This magic value causes
aio_dispatch_handler() to invoke ->poll_ready() instead of ->io_read()
for G_IO_IN or ->io_write() for G_IO_OUT.
This magic value 0 hack works for the IOThread where AioHandlers are
placed on ->ready_list and processed by aio_dispatch_ready_handlers().
It does not work for the main loop where all AioHandlers are processed
by aio_dispatch_handlers(), even those that are not ready and have a
revents value of 0.
As a result the main loop invokes ->poll_ready() on AioHandlers that are
not ready. These spurious ->poll_ready() calls waste CPU cycles and
could lead to crashes if the code assumes ->poll() must have succeeded
before ->poll_ready() is called (a reasonable asumption but I haven't
seen it in practice).
Stop using revents to track whether ->poll_ready() will be called on an
AioHandler. Introduce a separate AioHandler->poll_ready field instead.
This eliminates spurious ->poll_ready() calls in the main loop.
Fixes: 826cc32423db2a99d184dbf4f507c737d7e7a4ae ("aio-posix: split poll check from ready handler")
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reported-by: Jason Wang <jasowang@redhat.com>
Tested-by: Jason Wang <jasowang@redhat.com>
Message-id: 20220223155703.136833-1-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
---
util/aio-posix.h | 1 +
util/aio-posix.c | 32 ++++++++++++++++++--------------
2 files changed, 19 insertions(+), 14 deletions(-)
diff --git a/util/aio-posix.h b/util/aio-posix.h
index 7f2c37a684..80b927c7f4 100644
--- a/util/aio-posix.h
+++ b/util/aio-posix.h
@@ -37,6 +37,7 @@ struct AioHandler {
unsigned flags; /* see fdmon-io_uring.c */
#endif
int64_t poll_idle_timeout; /* when to stop userspace polling */
+ bool poll_ready; /* has polling detected an event? */
bool is_external;
};
diff --git a/util/aio-posix.c b/util/aio-posix.c
index 7b9f629218..be0182a3c6 100644
--- a/util/aio-posix.c
+++ b/util/aio-posix.c
@@ -23,15 +23,6 @@
#include "trace.h"
#include "aio-posix.h"
-/*
- * G_IO_IN and G_IO_OUT are not appropriate revents values for polling, since
- * the handler may not need to access the file descriptor. For example, the
- * handler doesn't need to read from an EventNotifier if it polled a memory
- * location and a read syscall would be slow. Define our own unique revents
- * value to indicate that polling determined this AioHandler is ready.
- */
-#define REVENTS_POLL_READY 0
-
/* Stop userspace polling on a handler if it isn't active for some time */
#define POLL_IDLE_INTERVAL_NS (7 * NANOSECONDS_PER_SECOND)
@@ -49,6 +40,14 @@ void aio_add_ready_handler(AioHandlerList *ready_list,
QLIST_INSERT_HEAD(ready_list, node, node_ready);
}
+static void aio_add_poll_ready_handler(AioHandlerList *ready_list,
+ AioHandler *node)
+{
+ QLIST_SAFE_REMOVE(node, node_ready); /* remove from nested parent's list */
+ node->poll_ready = true;
+ QLIST_INSERT_HEAD(ready_list, node, node_ready);
+}
+
static AioHandler *find_aio_handler(AioContext *ctx, int fd)
{
AioHandler *node;
@@ -76,6 +75,7 @@ static bool aio_remove_fd_handler(AioContext *ctx, AioHandler *node)
}
node->pfd.revents = 0;
+ node->poll_ready = false;
/* If the fd monitor has already marked it deleted, leave it alone */
if (QLIST_IS_INSERTED(node, node_deleted)) {
@@ -247,7 +247,7 @@ static bool poll_set_started(AioContext *ctx, AioHandlerList *ready_list,
/* Poll one last time in case ->io_poll_end() raced with the event */
if (!started && node->io_poll(node->opaque)) {
- aio_add_ready_handler(ready_list, node, REVENTS_POLL_READY);
+ aio_add_poll_ready_handler(ready_list, node);
progress = true;
}
}
@@ -282,6 +282,7 @@ bool aio_pending(AioContext *ctx)
QLIST_FOREACH_RCU(node, &ctx->aio_handlers, node) {
int revents;
+ /* TODO should this check poll ready? */
revents = node->pfd.revents & node->pfd.events;
if (revents & (G_IO_IN | G_IO_HUP | G_IO_ERR) && node->io_read &&
aio_node_check(ctx, node->is_external)) {
@@ -323,11 +324,15 @@ static void aio_free_deleted_handlers(AioContext *ctx)
static bool aio_dispatch_handler(AioContext *ctx, AioHandler *node)
{
bool progress = false;
+ bool poll_ready;
int revents;
revents = node->pfd.revents & node->pfd.events;
node->pfd.revents = 0;
+ poll_ready = node->poll_ready;
+ node->poll_ready = false;
+
/*
* Start polling AioHandlers when they become ready because activity is
* likely to continue. Note that starvation is theoretically possible when
@@ -344,7 +349,7 @@ static bool aio_dispatch_handler(AioContext *ctx, AioHandler *node)
QLIST_INSERT_HEAD(&ctx->poll_aio_handlers, node, node_poll);
}
if (!QLIST_IS_INSERTED(node, node_deleted) &&
- revents == 0 &&
+ poll_ready && revents == 0 &&
aio_node_check(ctx, node->is_external) &&
node->io_poll_ready) {
node->io_poll_ready(node->opaque);
@@ -432,7 +437,7 @@ static bool run_poll_handlers_once(AioContext *ctx,
QLIST_FOREACH_SAFE(node, &ctx->poll_aio_handlers, node_poll, tmp) {
if (aio_node_check(ctx, node->is_external) &&
node->io_poll(node->opaque)) {
- aio_add_ready_handler(ready_list, node, REVENTS_POLL_READY);
+ aio_add_poll_ready_handler(ready_list, node);
node->poll_idle_timeout = now + POLL_IDLE_INTERVAL_NS;
@@ -491,8 +496,7 @@ static bool remove_idle_poll_handlers(AioContext *ctx,
* this causes progress.
*/
if (node->io_poll(node->opaque)) {
- aio_add_ready_handler(ready_list, node,
- REVENTS_POLL_READY);
+ aio_add_poll_ready_handler(ready_list, node);
progress = true;
}
}
--
2.35.1
next prev parent reply other threads:[~2022-03-17 17:03 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-03-17 16:57 [PULL for-7.0 0/2] Block patches Stefan Hajnoczi
2022-03-17 16:57 ` [PULL for-7.0 1/2] aio-posix: fix build failure io_uring 2.2 Stefan Hajnoczi
2022-03-17 17:14 ` Daniel P. Berrangé
2022-03-18 2:57 ` Wang, Haiyue
2022-03-21 13:41 ` Stefan Hajnoczi
2022-03-17 16:57 ` Stefan Hajnoczi [this message]
2022-03-17 18:36 ` [PULL for-7.0 0/2] Block patches Peter Maydell
2022-03-21 14:44 ` Stefan Hajnoczi
2022-03-21 17:46 ` Peter Maydell
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=20220317165743.238662-3-stefanha@redhat.com \
--to=stefanha@redhat.com \
--cc=fam@euphon.net \
--cc=jasowang@redhat.com \
--cc=pbonzini@redhat.com \
--cc=peter.maydell@linaro.org \
--cc=qemu-block@nongnu.org \
--cc=qemu-devel@nongnu.org \
/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).