From: Markus Armbruster <armbru@redhat.com>
To: Wolfgang Bumiller <w.bumiller@proxmox.com>
Cc: "Marc-André Lureau" <marcandre.lureau@redhat.com>,
qemu-stable@nongnu.org, qemu-devel@nongnu.org,
"Michael Roth" <mdroth@linux.vnet.ibm.com>
Subject: Re: [PATCH] monitor/qmp: resume monitor when clearing its queue
Date: Wed, 09 Oct 2019 10:39:44 +0200 [thread overview]
Message-ID: <87ftk28g8f.fsf@dusky.pond.sub.org> (raw)
In-Reply-To: 20191002083003.21556-1-w.bumiller@proxmox.com
Cc: Marc-André for additional monitor and chardev expertise.
Wolfgang Bumiller <w.bumiller@proxmox.com> writes:
> When a monitor's queue is filled up in handle_qmp_command()
> it gets suspended. It's the dispatcher bh's job currently to
> resume the monitor, which it does after processing an event
> from the queue. However, it is possible for a
> CHR_EVENT_CLOSED event to be processed before before the bh
> is scheduled, which will clear the queue without resuming
> the monitor, thereby preventing the dispatcher from reaching
> the resume() call.
Because with the request queue cleared, there's nothing for
monitor_qmp_requests_pop_any_with_lock() to pop, so
monitor_qmp_bh_dispatcher() won't look at this monitor. It stays
suspended forever. Correct?
Observable effect for the monitor's user?
> Fix this by resuming the monitor when clearing a queue which
> was filled up.
>
> Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
> ---
> @Michael, we ran into this with qemu 4.0, so if the logic in this patch
> is correct it may make sense to include it in the 4.0.1 roundup.
> A backport is at [1] as 4.0 was before the monitor/ dir split.
>
> [1] https://gitlab.com/wbumiller/qemu/commit/9d8bbb5294ed084f282174b0c91e1a614e0a0714
>
> monitor/qmp.c | 10 ++++++++++
> 1 file changed, 10 insertions(+)
>
> diff --git a/monitor/qmp.c b/monitor/qmp.c
> index 9d9e5d8b27..c1db5bf940 100644
> --- a/monitor/qmp.c
> +++ b/monitor/qmp.c
> @@ -70,9 +70,19 @@ static void qmp_request_free(QMPRequest *req)
> /* Caller must hold mon->qmp.qmp_queue_lock */
> static void monitor_qmp_cleanup_req_queue_locked(MonitorQMP *mon)
> {
> + bool need_resume = (!qmp_oob_enabled(mon) && mon->qmp_requests->length > 0)
> + || mon->qmp_requests->length == QMP_REQ_QUEUE_LEN_MAX;
Can you explain why this condition is correct?
> while (!g_queue_is_empty(mon->qmp_requests)) {
> qmp_request_free(g_queue_pop_head(mon->qmp_requests));
> }
> + if (need_resume) {
> + /*
> + * Pairs with the monitor_suspend() in handle_qmp_command() in case the
> + * queue gets cleared from a CH_EVENT_CLOSED event before the dispatch
> + * bh got scheduled.
> + */
> + monitor_resume(&mon->common);
> + }
> }
>
> static void monitor_qmp_cleanup_queues(MonitorQMP *mon)
Is monitor_qmp_cleanup_req_queue_locked() the correct place?
It's called from
* monitor_qmp_event() case CHR_EVENT_CLOSED via
monitor_qmp_cleanup_queues(), as part of destroying the monitor's
session state.
This is the case you're trying to fix. Correct?
I figure monitor_resume() is safe because we haven't really destroyed
anything, yet, we merely flushed the request queue. Correct?
* monitor_data_destroy() via monitor_data_destroy_qmp() when destroying
the monitor.
Can need_resume be true in this case? If yes, is monitor_resume()
still safe? We're in the middle of destroying the monitor...
next prev parent reply other threads:[~2019-10-09 16:55 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-10-02 8:30 [PATCH] monitor/qmp: resume monitor when clearing its queue Wolfgang Bumiller
2019-10-09 8:39 ` Markus Armbruster [this message]
2019-10-09 10:10 ` Wolfgang Bumiller
2019-10-09 19:18 ` Markus Armbruster
2019-10-10 8:12 ` Wolfgang Bumiller
2019-10-10 9:03 ` Markus Armbruster
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=87ftk28g8f.fsf@dusky.pond.sub.org \
--to=armbru@redhat.com \
--cc=marcandre.lureau@redhat.com \
--cc=mdroth@linux.vnet.ibm.com \
--cc=qemu-devel@nongnu.org \
--cc=qemu-stable@nongnu.org \
--cc=w.bumiller@proxmox.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).