From: Linus Torvalds <torvalds@linux-foundation.org>
To: Rafael David Tinoco <inaddy@ubuntu.com>
Cc: LKML <linux-kernel@vger.kernel.org>,
Thomas Gleixner <tglx@linutronix.de>,
Jens Axboe <axboe@kernel.dk>
Subject: Re: smp_call_function_single lockups
Date: Wed, 11 Feb 2015 11:59:06 -0800 [thread overview]
Message-ID: <CA+55aFyGkOCVGD3Ds7Wt1z9Dw7cmk_yXw7YwruACQh5QAXOQvQ@mail.gmail.com> (raw)
In-Reply-To: <CA+55aFxbtyciJwhgaLMK7XH8MyQ9Nm4=5Ke-QCo3WOaacegLXw@mail.gmail.com>
On Wed, Feb 11, 2015 at 10:18 AM, Linus Torvalds
<torvalds@linux-foundation.org> wrote:
>
> I'll think about this all, but we couldn't figure anything out last
> time we looked at it, so without more clues, don't hold your breath.
So having looked at it once more, one thing struck me:
Look at smp_call_function_single_async(). The comment says
* Like smp_call_function_single(), but the call is asynchonous and
* can thus be done from contexts with disabled interrupts.
but that is *only* true if we don't have to wait for the csd lock. The
comments even clarify that:
* The caller passes his own pre-allocated data structure
* (ie: embedded in an object) and is responsible for synchronizing it
* such that the IPIs performed on the @csd are strictly serialized.
but it's not at all clear that the caller *can* do that. Since the
"csd_unlock()" is done *after* the call to the callback function, any
serialization done by the caller is fundamentally not trustworthy,
since it cannot serialize with the csd lock - if it releases things in
the callback, the csd lock will still be set after releasing things.
So the caller has a really hard time guaranteeing that CSD_LOCK isn't
set. And if the call is done in interrupt context, for all we know it
is interrupting the code that is going to clear CSD_LOCK, so CSD_LOCK
will never be cleared at all, and csd_lock() will wait forever.
So I actually think that for the async case, we really *should* unlock
before doing the callback (which is what Thomas' old patch did).
And we migth well be better off doing something like
WARN_ON_ONCE(csd->flags & CSD_LOCK);
in smp_call_function_single_async(), because that really is a hard requirement.
And it strikes me that hrtick_csd is one of these cases that do this
with interrupts disabled, and use the callback for serialization. So I
really wonder if this is part of the problem..
Thomas? Am I missing something?
Linus
next prev parent reply other threads:[~2015-02-11 19:59 UTC|newest]
Thread overview: 77+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-02-11 13:19 smp_call_function_single lockups Rafael David Tinoco
2015-02-11 18:18 ` Linus Torvalds
2015-02-11 19:59 ` Linus Torvalds [this message]
2015-02-11 20:42 ` Linus Torvalds
2015-02-12 16:38 ` Rafael David Tinoco
2015-02-18 22:25 ` Peter Zijlstra
2015-02-19 15:42 ` Rafael David Tinoco
2015-02-19 16:14 ` Linus Torvalds
2015-02-23 14:01 ` Rafael David Tinoco
2015-02-23 19:32 ` Linus Torvalds
2015-02-23 20:50 ` Peter Zijlstra
2015-02-23 21:02 ` Rafael David Tinoco
2015-02-19 16:16 ` Peter Zijlstra
2015-02-19 16:26 ` Linus Torvalds
2015-02-19 16:32 ` Rafael David Tinoco
2015-02-19 16:59 ` Linus Torvalds
2015-02-19 17:30 ` Rafael David Tinoco
2015-02-19 17:39 ` Linus Torvalds
2015-02-19 20:29 ` Linus Torvalds
2015-02-19 21:59 ` Linus Torvalds
2015-02-19 22:45 ` Linus Torvalds
2015-03-31 3:15 ` Chris J Arges
2015-03-31 4:28 ` Linus Torvalds
2015-03-31 10:56 ` [debug PATCHes] " Ingo Molnar
2015-03-31 22:38 ` Chris J Arges
2015-04-01 12:39 ` Ingo Molnar
2015-04-01 14:10 ` Chris J Arges
2015-04-01 14:55 ` Ingo Molnar
2015-03-31 4:46 ` Linus Torvalds
2015-03-31 15:08 ` Linus Torvalds
2015-03-31 22:23 ` Chris J Arges
2015-03-31 23:07 ` Linus Torvalds
2015-04-01 14:32 ` Chris J Arges
2015-04-01 15:36 ` Linus Torvalds
2015-04-02 9:55 ` Ingo Molnar
2015-04-02 17:35 ` Linus Torvalds
2015-04-01 12:43 ` Ingo Molnar
2015-04-01 16:10 ` Chris J Arges
2015-04-01 16:14 ` Linus Torvalds
2015-04-01 21:59 ` Chris J Arges
2015-04-02 17:31 ` Linus Torvalds
2015-04-02 18:26 ` Ingo Molnar
2015-04-02 18:51 ` Chris J Arges
2015-04-02 19:07 ` Ingo Molnar
2015-04-02 20:57 ` Linus Torvalds
2015-04-02 21:13 ` Chris J Arges
2015-04-03 5:43 ` [PATCH] smp/call: Detect stuck CSD locks Ingo Molnar
2015-04-03 5:47 ` Ingo Molnar
2015-04-06 16:58 ` Chris J Arges
2015-04-06 17:32 ` Linus Torvalds
2015-04-07 9:21 ` Ingo Molnar
2015-04-07 20:59 ` Chris J Arges
2015-04-07 21:15 ` Linus Torvalds
2015-04-08 6:47 ` Ingo Molnar
2015-04-13 3:56 ` Chris J Arges
2015-04-13 6:14 ` Ingo Molnar
2015-04-15 19:54 ` Chris J Arges
2015-04-16 11:04 ` Ingo Molnar
2015-04-16 15:58 ` Chris J Arges
2015-04-16 16:31 ` Ingo Molnar
2015-04-29 21:08 ` Chris J Arges
2015-05-11 14:00 ` Ingo Molnar
2015-05-20 18:19 ` Chris J Arges
2015-04-03 5:45 ` smp_call_function_single lockups Ingo Molnar
2015-04-06 17:23 ` Chris J Arges
2015-02-20 9:30 ` Ingo Molnar
2015-02-20 16:49 ` Linus Torvalds
2015-02-20 19:41 ` Ingo Molnar
2015-02-20 20:03 ` Linus Torvalds
2015-02-20 20:11 ` Ingo Molnar
2015-03-20 10:15 ` Peter Zijlstra
2015-03-20 16:26 ` Linus Torvalds
2015-03-20 17:14 ` Mike Galbraith
2015-04-01 14:22 ` Frederic Weisbecker
2015-04-18 10:13 ` [tip:locking/urgent] smp: Fix smp_call_function_single_async() locking tip-bot for Linus Torvalds
2015-02-22 8:59 smp_call_function_single lockups Daniel J Blueman
2015-02-22 10:37 ` Ingo Molnar
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=CA+55aFyGkOCVGD3Ds7Wt1z9Dw7cmk_yXw7YwruACQh5QAXOQvQ@mail.gmail.com \
--to=torvalds@linux-foundation.org \
--cc=axboe@kernel.dk \
--cc=inaddy@ubuntu.com \
--cc=linux-kernel@vger.kernel.org \
--cc=tglx@linutronix.de \
/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).