From: Stefan Hajnoczi <stefanha@gmail.com>
To: Alex Bligh <alex@alex.org.uk>
Cc: Kevin Wolf <kwolf@redhat.com>,
Anthony Liguori <aliguori@us.ibm.com>,
qemu-devel@nongnu.org, liu ping fan <qemulist@gmail.com>,
Stefan Hajnoczi <stefanha@redhat.com>,
Paolo Bonzini <pbonzini@redhat.com>,
MORITA Kazutaka <morita.kazutaka@lab.ntt.co.jp>,
rth@twiddle.net
Subject: Re: [Qemu-devel] [RFC] [PATCHv8 11/30] aio / timers: Add a notify callback to QEMUTimerList
Date: Fri, 9 Aug 2013 11:02:40 +0200 [thread overview]
Message-ID: <20130809090240.GD22210@stefanha-thinkpad.redhat.com> (raw)
In-Reply-To: <1375998147-24292-12-git-send-email-alex@alex.org.uk>
On Thu, Aug 08, 2013 at 10:42:08PM +0100, Alex Bligh wrote:
> @@ -213,13 +214,41 @@ QEMUClock *timerlist_get_clock(QEMUTimerList *timer_list);
> bool timerlist_run_timers(QEMUTimerList *timer_list);
>
> /**
> + * timerlist_set_notify_cb:
> + * @timer_list: the timer list to use
> + * @cb: the callback to call on notification
> + * @opaque: the opaque pointer to pass to the callback
> + *
> + * Set the notify callback for a timer list. The notifier
> + * callback is called when the clock is reenabled or a timer
> + * on the list is modified.
> + */
> +void timerlist_set_notify_cb(QEMUTimerList *timer_list,
> + QEMUTimerListNotifyCB *cb, void *opaque);
When looking at thread-safety I had to think about set_notify_cb() for a
while. The issue is that we add the timerlist to the clock source's
->timerlists *before* notify_cb has been assigned.
This could be a problem is another thread re-enables the clock source
while we are still in timerlist_new().
In practice it is not an issue when AioContexts are created under the
global mutex (that's the case today).
Still, it would be a little safer to drop set_notify_cb() and instead
pass in cb/opaque in timerlist_new().
Here is a patch that does this (against the previous revision of this
patch):
>From 75096b8fcafbac598ec0a5eab7a10cfb2e571fdf Mon Sep 17 00:00:00 2001
From: Stefan Hajnoczi <stefanha@redhat.com>
Date: Wed, 7 Aug 2013 15:44:02 +0200
Subject: [PATCH] qemu-timer: set notify_cb in timerlist_new()
Eliminate the race condition between creating a QEMUTimerList and
setting its notify_cb field. This is important for multi-threading
scenarios where a timerlist can be notified before timerlist_new() has
returned.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
---
include/qemu/timer.h | 22 ++++++++--------------
qemu-timer.c | 25 ++++++++++++-------------
2 files changed, 20 insertions(+), 27 deletions(-)
diff --git a/include/qemu/timer.h b/include/qemu/timer.h
index 9989d0e..935c259 100644
--- a/include/qemu/timer.h
+++ b/include/qemu/timer.h
@@ -207,13 +207,20 @@ void qemu_clock_notify(QEMUClock *clock);
/**
* timerlist_new:
* @type: the clock type to associate with the timerlist
+ * @cb: the callback to call on notification
+ * @opaque: the opaque pointer to pass to the callback
*
* Create a new timerlist associated with the clock of
* type @type.
*
+ * The notifier callback is called when the clock is reenabled or a timer on
+ * the list is modified.
+ *
* Returns: a pointer to the QEMUTimerList created
*/
-QEMUTimerList *timerlist_new(QEMUClockType type);
+QEMUTimerList *timerlist_new(QEMUClockType type,
+ QEMUTimerListNotifyCB *cb,
+ void *opaque);
/**
* timerlist_free:
@@ -282,19 +289,6 @@ QEMUClock *timerlist_get_clock(QEMUTimerList *timer_list);
bool timerlist_run_timers(QEMUTimerList *timer_list);
/**
- * timerlist_set_notify_cb:
- * @timer_list: the timer list to use
- * @cb: the callback to call on notification
- * @opaque: the opaque pointer to pass to the callback
- *
- * Set the notify callback for a timer list. The notifier
- * callback is called when the clock is reenabled or a timer
- * on the list is modified.
- */
-void timerlist_set_notify_cb(QEMUTimerList *timer_list,
- QEMUTimerListNotifyCB *cb, void *opaque);
-
-/**
* timerlist_notify:
* @timer_list: the timer list to use
*
diff --git a/qemu-timer.c b/qemu-timer.c
index a39c4d6..8cb4fe7 100644
--- a/qemu-timer.c
+++ b/qemu-timer.c
@@ -87,7 +87,9 @@ static bool qemu_timer_expired_ns(QEMUTimer *timer_head, int64_t current_time)
return timer_head && (timer_head->expire_time <= current_time);
}
-static QEMUTimerList *timerlist_new_from_clock(QEMUClock *clock)
+static QEMUTimerList *timerlist_new_from_clock(QEMUClock *clock,
+ QEMUTimerListNotifyCB *cb,
+ void *opaque)
{
QEMUTimerList *timer_list;
@@ -101,13 +103,17 @@ static QEMUTimerList *timerlist_new_from_clock(QEMUClock *clock)
timer_list = g_malloc0(sizeof(QEMUTimerList));
timer_list->clock = clock;
+ timer_list->notify_cb = cb;
+ timer_list->notify_opaque = opaque;
QLIST_INSERT_HEAD(&clock->timerlists, timer_list, list);
return timer_list;
}
-QEMUTimerList *timerlist_new(QEMUClockType type)
+QEMUTimerList *timerlist_new(QEMUClockType type,
+ QEMUTimerListNotifyCB *cb,
+ void *opaque)
{
- return timerlist_new_from_clock(qemu_get_clock(type));
+ return timerlist_new_from_clock(qemu_get_clock(type), cb, opaque);
}
void timerlist_free(QEMUTimerList *timer_list)
@@ -131,7 +137,8 @@ QEMUClock *qemu_clock_new(QEMUClockType type)
clock->enabled = true;
clock->last = INT64_MIN;
notifier_list_init(&clock->reset_notifiers);
- clock->default_timerlist = timerlist_new_from_clock(clock);
+ QLIST_INIT(&clock->timerlists);
+ clock->default_timerlist = timerlist_new_from_clock(clock, NULL, NULL);
return clock;
}
@@ -239,13 +246,6 @@ QEMUTimerList *qemu_clock_get_default_timerlist(QEMUClock *clock)
return clock->default_timerlist;
}
-void timerlist_set_notify_cb(QEMUTimerList *timer_list,
- QEMUTimerListNotifyCB *cb, void *opaque)
-{
- timer_list->notify_cb = cb;
- timer_list->notify_opaque = opaque;
-}
-
void timerlist_notify(QEMUTimerList *timer_list)
{
if (timer_list->notify_cb) {
@@ -433,8 +433,7 @@ void timerlistgroup_init(QEMUTimerListGroup tlg,
{
QEMUClockType type;
for (type = 0; type < QEMU_CLOCK_MAX; type++) {
- tlg[type] = timerlist_new(type);
- timerlist_set_notify_cb(tlg[type], cb, opaque);
+ tlg[type] = timerlist_new(type, cb, opaque);
}
}
--
1.8.1.4
next prev parent reply other threads:[~2013-08-09 9:03 UTC|newest]
Thread overview: 91+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-08-08 21:41 [Qemu-devel] [RFC] [PATCHv8 00/30] aio / timers: Add AioContext timers and use ppoll Alex Bligh
2013-08-08 21:41 ` [Qemu-devel] [RFC] [PATCHv8 01/30] aio / timers: Rename qemu_new_clock and expose clock types Alex Bligh
2013-08-08 21:41 ` [Qemu-devel] [RFC] [PATCHv8 02/30] aio / timers: add qemu-timer.c utility functions Alex Bligh
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 03/30] aio / timers: Consistent treatment of disabled clocks for deadlines Alex Bligh
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 04/30] aio / timers: add ppoll support with qemu_poll_ns Alex Bligh
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 05/30] aio / timers: Add prctl(PR_SET_TIMERSLACK, 1, ...) to reduce timer slack Alex Bligh
2013-08-09 8:53 ` Stefan Hajnoczi
2013-08-09 14:10 ` Alex Bligh
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 06/30] aio / timers: Make qemu_run_timers and qemu_run_all_timers return progress Alex Bligh
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 07/30] aio / timers: Split QEMUClock into QEMUClock and QEMUTimerList Alex Bligh
2013-08-09 10:03 ` Paolo Bonzini
2013-08-09 14:23 ` Alex Bligh
2013-08-09 14:28 ` Paolo Bonzini
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 08/30] aio / timers: Untangle include files Alex Bligh
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 09/30] aio / timers: Add QEMUTimerListGroup and helper functions Alex Bligh
2013-08-09 10:02 ` Paolo Bonzini
2013-08-09 14:27 ` Alex Bligh
2013-08-09 14:35 ` Paolo Bonzini
2013-08-09 14:47 ` Alex Bligh
2013-08-09 14:59 ` Paolo Bonzini
2013-08-10 11:05 ` Alex Bligh
2013-08-11 7:53 ` Paolo Bonzini
2013-08-11 8:29 ` Alex Bligh
2013-08-12 17:27 ` Paolo Bonzini
2013-08-12 17:32 ` Alex Bligh
2013-08-12 6:53 ` Wenchao Xia
2013-08-12 7:08 ` Alex Bligh
2013-08-09 15:09 ` Stefan Hajnoczi
2013-08-09 15:15 ` Paolo Bonzini
2013-08-10 3:27 ` liu ping fan
2013-08-10 8:38 ` Paolo Bonzini
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 10/30] aio / timers: Add QEMUTimerListGroup to AioContext Alex Bligh
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 11/30] aio / timers: Add a notify callback to QEMUTimerList Alex Bligh
2013-08-09 9:02 ` Stefan Hajnoczi [this message]
2013-08-09 14:16 ` Alex Bligh
2013-08-09 10:03 ` Paolo Bonzini
2013-08-09 14:28 ` Alex Bligh
2013-08-10 11:13 ` Alex Bligh
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 12/30] aio / timers: aio_ctx_prepare sets timeout from AioContext timers Alex Bligh
2013-08-09 10:04 ` Paolo Bonzini
2013-08-09 14:30 ` Alex Bligh
2013-08-09 14:41 ` Paolo Bonzini
2013-08-09 14:55 ` Alex Bligh
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 13/30] aio / timers: Add aio_timer_new wrapper Alex Bligh
2013-08-09 10:06 ` Paolo Bonzini
2013-08-09 14:32 ` Alex Bligh
2013-08-09 14:36 ` Paolo Bonzini
2013-08-09 14:51 ` Alex Bligh
2013-08-09 14:51 ` Paolo Bonzini
2013-08-09 14:57 ` Alex Bligh
2013-08-09 14:58 ` Paolo Bonzini
2013-08-09 22:57 ` Alex Bligh
2013-08-10 8:36 ` Paolo Bonzini
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 14/30] aio / timers: Convert aio_poll to use AioContext timers' deadline Alex Bligh
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 15/30] aio / timers: Convert mainloop to use timeout Alex Bligh
2013-08-09 10:07 ` Paolo Bonzini
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 16/30] aio / timers: On timer modification, qemu_notify or aio_notify Alex Bligh
2013-08-09 10:08 ` Paolo Bonzini
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 17/30] aio / timers: Introduce new API qemu_timer_new and friends Alex Bligh
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 18/30] aio / timers: Use all timerlists in icount warp calculations Alex Bligh
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 19/30] aio / timers: Add documentation and new format calls Alex Bligh
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 20/30] aio / timers: Remove alarm timers Alex Bligh
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 21/30] aio / timers: Remove legacy qemu_clock_deadline & qemu_timerlist_deadline Alex Bligh
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 22/30] aio / timers: Add qemu_clock_get_ms and qemu_clock_get_ms Alex Bligh
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 23/30] aio / timers: Rearrange timer.h & make legacy functions call non-legacy Alex Bligh
2013-08-08 22:12 ` Alex Bligh
2013-08-09 9:23 ` Stefan Hajnoczi
2013-08-09 14:18 ` Alex Bligh
2013-08-09 14:26 ` Paolo Bonzini
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 24/30] aio / timers: Remove main_loop_timerlist Alex Bligh
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 25/30] aio / timers: Convert rtc_clock to be a QEMUClockType Alex Bligh
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 26/30] aio / timers: convert block_job_sleep_ns and co_sleep_ns to new API Alex Bligh
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 27/30] aio / timers: Add test harness for AioContext timers Alex Bligh
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 28/30] aio / timers: Add scripts/switch-timer-api Alex Bligh
2013-08-09 10:23 ` Paolo Bonzini
2013-08-09 14:34 ` Alex Bligh
2013-08-09 10:26 ` Paolo Bonzini
2013-08-09 14:35 ` Alex Bligh
2013-08-09 14:37 ` Paolo Bonzini
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 29/30] aio / timers: Switch entire codebase to the new timer API Alex Bligh
2013-08-08 21:42 ` [Qemu-devel] [RFC] [PATCHv8 30/30] aio / timers: Remove legacy interface Alex Bligh
2013-08-08 21:46 ` [Qemu-devel] [RFC] [PATCHv8 00/30] aio / timers: Add AioContext timers and use ppoll Alex Bligh
2013-08-09 8:12 ` Jan Kiszka
2013-08-09 8:24 ` liu ping fan
2013-08-09 8:33 ` Jan Kiszka
2013-08-09 10:35 ` Alex Bligh
2013-08-09 9:41 ` Stefan Hajnoczi
2013-08-09 14:19 ` Alex Bligh
2013-08-09 14:27 ` Paolo Bonzini
2013-08-09 10:27 ` Paolo Bonzini
2013-08-09 11:18 ` 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=20130809090240.GD22210@stefanha-thinkpad.redhat.com \
--to=stefanha@gmail.com \
--cc=alex@alex.org.uk \
--cc=aliguori@us.ibm.com \
--cc=kwolf@redhat.com \
--cc=morita.kazutaka@lab.ntt.co.jp \
--cc=pbonzini@redhat.com \
--cc=qemu-devel@nongnu.org \
--cc=qemulist@gmail.com \
--cc=rth@twiddle.net \
--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).