From: "Michael L. Semon" <mlsemon35@gmail.com>
To: Jason Low <jason.low2@hp.com>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>,
"Michael L. Semon" <mlsemon35@gmail.com>,
Ingo Molnar <mingo@kernel.org>,
Peter Zijlstra <peterz@infradead.org>,
linux-kernel@vger.kernel.org
Subject: Re: 3.14.0+/x86: lockdep and mutexes not getting along
Date: Fri, 11 Apr 2014 09:41:19 -0400 (EDT) [thread overview]
Message-ID: <alpine.LNX.2.11.1404110923200.3964@bpserver.ds> (raw)
In-Reply-To: <1397108579.2586.15.camel@j-VirtualBox>
On Wed, 9 Apr 2014, Jason Low wrote:
> On Wed, 2014-04-09 at 15:19 +0300, Kirill A. Shutemov wrote:
> > On Sun, Apr 06, 2014 at 01:12:14AM -0400, Michael L. Semon wrote:
> > > Hi! Starting early in this merge window for 3.15, lockdep has been
> > > giving me trouble. Normally, a splat will happen, lockdep will shut
> > > itself off, and my i686 Pentium 4 PC will continue. Now, after the
> > > splat, it will allow one key of input at either a VGA console or over
> > > serial. After that, only the magic SysRq keys and KDB still work.
> > > File activity stops, and many processes are stuck in the D state.
> > >
> > > Bisect brought me here:
> > >
> > > root@plbearer:/usr/src/kernel-git/linux# git bisect good
> > > 6f008e72cd111a119b5d8de8c5438d892aae99eb is the first bad commit
> > > commit 6f008e72cd111a119b5d8de8c5438d892aae99eb
> > > Author: Peter Zijlstra <peterz@infradead.org>
> > > Date: Wed Mar 12 13:24:42 2014 +0100
> > >
> > > locking/mutex: Fix debug checks
> > >
> > > OK, so commit:
> > >
> > > 1d8fe7dc8078 ("locking/mutexes: Unlock the mutex without the wait_lock")
> > >
> > > generates this boot warning when CONFIG_DEBUG_MUTEXES=y:
> > >
> > > WARNING: CPU: 0 PID: 139 at /usr/src/linux-2.6/kernel/locking/mutex-debug.c:82 debug_mutex_unlock+0x155/0x180() DEBUG_LOCKS_WARN_ON(lock->owner != current)
> > >
> > > And that makes sense, because as soon as we release the lock a
> > > new owner can come in...
> > >
> > > One would think that !__mutex_slowpath_needs_to_unlock()
> > > implementations suffer the same, but for DEBUG we fall back to
> > > mutex-null.h which has an unconditional 1 for that.
> > >
> > > The mutex debug code requires the mutex to be unlocked after
> > > doing the debug checks, otherwise it can find inconsistent
> > > state.
> > >
> > > Reported-by: Ingo Molnar <mingo@kernel.org>
> > > Signed-off-by: Peter Zijlstra <peterz@infradead.org>
> > > Cc: jason.low2@hp.com
>
> Hello,
>
> As a starting point, would either of you like to test the following
> patch to see if it fixes the issue? This patch essentially generates the
> same code as in older kernels in the debug case. This applies on top of
> kernels with both commits 6f008e72cd11 and 1d8fe7dc8078.
>
> Thanks.
>
> -----
> diff --git a/kernel/locking/mutex-debug.c b/kernel/locking/mutex-debug.c
> index e1191c9..faf6f5b 100644
> --- a/kernel/locking/mutex-debug.c
> +++ b/kernel/locking/mutex-debug.c
> @@ -83,12 +83,6 @@ void debug_mutex_unlock(struct mutex *lock)
>
> DEBUG_LOCKS_WARN_ON(!lock->wait_list.prev && !lock->wait_list.next);
> mutex_clear_owner(lock);
> -
> - /*
> - * __mutex_slowpath_needs_to_unlock() is explicitly 0 for debug
> - * mutexes so that we can do it here after we've verified state.
> - */
> - atomic_set(&lock->count, 1);
> }
>
> void debug_mutex_init(struct mutex *lock, const char *name,
> diff --git a/kernel/locking/mutex.c b/kernel/locking/mutex.c
> index bc73d33..f1f672e 100644
> --- a/kernel/locking/mutex.c
> +++ b/kernel/locking/mutex.c
> @@ -34,13 +34,6 @@
> #ifdef CONFIG_DEBUG_MUTEXES
> # include "mutex-debug.h"
> # include <asm-generic/mutex-null.h>
> -/*
> - * Must be 0 for the debug case so we do not do the unlock outside of the
> - * wait_lock region. debug_mutex_unlock() will do the actual unlock in this
> - * case.
> - */
> -# undef __mutex_slowpath_needs_to_unlock
> -# define __mutex_slowpath_needs_to_unlock() 0
> #else
> # include "mutex.h"
> # include <asm/mutex.h>
> @@ -688,6 +681,17 @@ __mutex_unlock_common_slowpath(atomic_t *lock_count, int nested)
> unsigned long flags;
>
> /*
> + * In the debug cases, obtain the wait_lock first
> + * before calling the following debugging functions.
> + */
> +#if defined(CONFIG_DEBUG_MUTEXES) || defined(CONFIG_DEBUG_LOCK_ALLOC)
> + spin_lock_mutex(&lock->wait_lock, flags);
> +#endif
> +
> + mutex_release(&lock->dep_map, nested, _RET_IP_);
> + debug_mutex_unlock(lock);
> +
> + /*
> * some architectures leave the lock unlocked in the fastpath failure
> * case, others need to leave it locked. In the later case we have to
> * unlock it here
> @@ -695,9 +699,9 @@ __mutex_unlock_common_slowpath(atomic_t *lock_count, int nested)
> if (__mutex_slowpath_needs_to_unlock())
> atomic_set(&lock->count, 1);
>
> +#if !defined(CONFIG_DEBUG_MUTEXES) && !defined(CONFIG_DEBUG_LOCK_ALLOC)
> spin_lock_mutex(&lock->wait_lock, flags);
> - mutex_release(&lock->dep_map, nested, _RET_IP_);
> - debug_mutex_unlock(lock);
> +#endif
>
> if (!list_empty(&lock->wait_list)) {
> /* get the first entry from the wait-list: */
>
>
>
This works and was given an overnight xfstests run on XFS to prove
itself. The other patch worked, too, but it was sent on a failing
mission to find an elusive JFS lockdep splat by find'ing, untarring,
and fs_mark'ing everything in sight. The other patch did fine on
the original xfstests generic/113 on devel/debug XFS.
Thanks!
Michael
next prev parent reply other threads:[~2014-04-11 13:41 UTC|newest]
Thread overview: 25+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-04-06 5:12 3.14.0+/x86: lockdep and mutexes not getting along Michael L. Semon
2014-04-09 12:19 ` Kirill A. Shutemov
2014-04-10 5:42 ` Jason Low
2014-04-10 8:14 ` Peter Zijlstra
2014-04-10 9:15 ` Kirill A. Shutemov
2014-04-10 11:42 ` Peter Zijlstra
2014-04-10 9:18 ` Peter Zijlstra
2014-04-10 14:15 ` Peter Zijlstra
2014-04-11 13:59 ` Valdis.Kletnieks
2014-04-14 7:22 ` [tip:core/urgent] locking/mutex: Fix debug_mutexes tip-bot for Peter Zijlstra
2014-04-10 17:14 ` 3.14.0+/x86: lockdep and mutexes not getting along Jason Low
2014-04-10 17:28 ` Peter Zijlstra
2014-04-10 19:04 ` Jason Low
2014-04-10 23:26 ` Dave Jones
2014-04-10 23:30 ` Dave Jones
2014-04-11 3:48 ` Paul E. McKenney
2014-04-11 13:41 ` Michael L. Semon [this message]
2014-04-10 8:12 ` Peter Zijlstra
2014-04-10 8:13 ` Peter Zijlstra
2014-04-10 14:29 ` cred_guard_mutex vs seq_file::lock [was: Re: 3.14.0+/x86: lockdep and mutexes not getting along] Peter Zijlstra
2014-04-11 14:50 ` David Howells
2014-04-11 15:07 ` Al Viro
2014-07-30 22:31 ` Kirill A. Shutemov
2014-07-30 23:03 ` Kirill A. Shutemov
2014-07-31 7:26 ` Cyrill Gorcunov
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=alpine.LNX.2.11.1404110923200.3964@bpserver.ds \
--to=mlsemon35@gmail.com \
--cc=jason.low2@hp.com \
--cc=kirill@shutemov.name \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@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).