From: Chris Mason <clm@fb.com>
To: Benjamin LaHaise <bcrl@kvack.org>
Cc: Kent Overstreet <kmo@daterainc.com>,
<linux-fsdevel@vger.kernel.org>, <linux-aio@kvack.org>,
Peter Zijlstra <peterz@infradead.org>
Subject: Re: [PATCH RFC] fs/aio: fix sleeping while TASK_INTERRUPTIBLE
Date: Sun, 28 Dec 2014 20:24:26 -0500 [thread overview]
Message-ID: <1419816266.13012.15@mail.thefacebook.com> (raw)
In-Reply-To: <20141225031134.GI17185@kvack.org>
On Wed, Dec 24, 2014 at 10:11 PM, Benjamin LaHaise <bcrl@kvack.org>
wrote:
> On Wed, Dec 24, 2014 at 06:59:58PM -0800, Kent Overstreet wrote:
>> On Tue, Dec 23, 2014 at 04:58:47PM -0500, Benjamin LaHaise wrote:
>> > Hi Chris,
>> >
>> > On Tue, Dec 23, 2014 at 01:55:26PM -0500, Chris Mason wrote:
>> > > Works for me, the patch is mostly a (somewhat commented) list
>> of all
>> > > the places we're currently doing it wrong.
>> >
>> > I think the following change should suffice to fix this issue,
>> and it's a
>> > lot easier to review, too. I've given this a quick test, and it
>> works for
>> > me. I do have one concern: is it safe to call mutex_lock() when
>> the current
>> > task is already on other wait queues? If the answer is no, then
>> it may be
>> > necessary to convert ->ring_lock back into spinlock as it was
>> prior to 3.10
>> > to avoid using mutex_lock(). The same question applies to kmap()
>> and
>> > copy_to_user(), and those concerns might have implications across
>> the rest
>> > of the kernel. Thoughts?
>> >
>> > + __set_current_state(state);
>>
>> I don't think this is safe - if we race, and another thread wakes
>> us up, we're
>> setting our state back to TASK_INTERRUPTIBLE _without_ us being on
>> the waitlist.
>
> It should be -- the conditions for going to sleep are checked after
> current's
> state is set here.
We're task interruptible and may be run under a timer, so we'll catch
wakeups for signals and timers in addition to available aio events.
-chris
--
To unsubscribe, send a message with 'unsubscribe linux-aio' in
the body to majordomo@kvack.org. For more info on Linux AIO,
see: http://www.kvack.org/aio/
Don't email: <a href=mailto:"aart@kvack.org">aart@kvack.org</a>
next prev parent reply other threads:[~2014-12-29 1:24 UTC|newest]
Thread overview: 21+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-12-23 0:16 [PATCH RFC] fs/aio: fix sleeping while TASK_INTERRUPTIBLE Chris Mason
2014-12-23 18:43 ` Benjamin LaHaise
2014-12-23 18:55 ` Chris Mason
2014-12-23 21:58 ` Benjamin LaHaise
2014-12-25 2:59 ` Kent Overstreet
2014-12-25 3:11 ` Benjamin LaHaise
2014-12-25 3:29 ` Kent Overstreet
2014-12-29 1:24 ` Chris Mason [this message]
2014-12-25 2:56 ` Kent Overstreet
2014-12-25 14:27 ` Sedat Dilek
2015-01-04 10:16 ` Sedat Dilek
2014-12-29 15:08 ` Chris Mason
2014-12-29 22:08 ` Kent Overstreet
2015-01-13 16:06 ` Benjamin LaHaise
2015-01-13 16:20 ` Chris Mason
2015-01-21 10:13 ` Dave Chinner
2015-01-21 21:42 ` Chris Mason
2015-02-03 9:14 ` Sedat Dilek
2015-02-03 9:54 ` Sedat Dilek
2015-02-09 3:08 ` Sedat Dilek
2015-02-09 4:21 ` Sedat Dilek
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=1419816266.13012.15@mail.thefacebook.com \
--to=clm@fb.com \
--cc=bcrl@kvack.org \
--cc=kmo@daterainc.com \
--cc=linux-aio@kvack.org \
--cc=linux-fsdevel@vger.kernel.org \
--cc=peterz@infradead.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).