lkml.org 
[lkml]   [2017]   [Aug]   [23]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
SubjectRe: possible circular locking dependency detected [was: linux-next: Tree for Aug 22]
On Wed, Aug 23, 2017 at 11:49:51AM +0800, Boqun Feng wrote:
> Hi Byungchul,
>
> On Wed, Aug 23, 2017 at 09:03:04AM +0900, Byungchul Park wrote:
> > On Tue, Aug 22, 2017 at 09:43:56PM +0000, Bart Van Assche wrote:
> > > On Tue, 2017-08-22 at 19:47 +0900, Sergey Senozhatsky wrote:
> > > > ======================================================
> > > > WARNING: possible circular locking dependency detected
> > > > 4.13.0-rc6-next-20170822-dbg-00020-g39758ed8aae0-dirty #1746 Not tainted
> > > > ------------------------------------------------------
> > > > fsck.ext4/148 is trying to acquire lock:
> > > > (&bdev->bd_mutex){+.+.}, at: [<ffffffff8116e73e>] __blkdev_put+0x33/0x190
> > > >
> > > > but now in release context of a crosslock acquired at the following:
> > > > ((complete)&wait#2){+.+.}, at: [<ffffffff812159e0>] blk_execute_rq+0xbb/0xda
> > > >
> > > > which lock already depends on the new lock.
> > > >
>
> I felt this message really misleading, because the deadlock is detected
> at the commit time of "((complete)&wait#2)" rather than the acquisition
> time of "(&bdev->bd_mutex)", so I made the following improvement.
>
> Thoughts?
>
> Regards,
> Boqun
>

While I'm on this one, I think we should also add a case in @check_src
is a cross lock, i.e. we detect cross deadlock at the acquisition time
of the cross lock. How about the following?

Regards,
Boqun

--------------------------------------->8
From: Boqun Feng <boqun.feng@gmail.com>
Date: Wed, 23 Aug 2017 12:12:16 +0800
Subject: [PATCH] lockdep: Print proper scenario if cross deadlock detected at
acquisition time

For a potential deadlock about CROSSRELEASE as follow:

P1 P2
=========== =============
lock(A)
lock(X)
lock(A)
commit(X)

A: normal lock, X: cross lock

, we could detect it at two places:

1. commit time:

We have run P1 first, and have dependency A --> X in graph, and
then we run P2, and find the deadlock.

2. acquisition time:

We have run P2 first, and have dependency A --> X, in
graph(because another P3 may run previously and is acquiring for
lock X), and then we run P1 and find the deadlock.

In current print_circular_lock_scenario(), for 1) we could print the
right scenario and note that's a deadlock related to CROSSRELEASE,
however for 2) we print the scenario as a normal lockdep deadlock.

It's better to print a proper scenario related to CROSSRELEASE to help
users find their bugs more easily, so improve this.

Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
---
kernel/locking/lockdep.c | 17 +++++++++++++++++
1 file changed, 17 insertions(+)

diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
index 642fb5362507..a3709e15f609 100644
--- a/kernel/locking/lockdep.c
+++ b/kernel/locking/lockdep.c
@@ -1156,6 +1156,23 @@ print_circular_lock_scenario(struct held_lock *src,
__print_lock_name(target);
printk(KERN_CONT ");\n");
printk("\n *** DEADLOCK ***\n\n");
+ } else if (cross_lock(src->instance)) {
+ printk(" Possible unsafe locking scenario by crosslock:\n\n");
+ printk(" CPU0 CPU1\n");
+ printk(" ---- ----\n");
+ printk(" lock(");
+ __print_lock_name(target);
+ printk(KERN_CONT ");\n");
+ printk(" lock(");
+ __print_lock_name(source);
+ printk(KERN_CONT ");\n");
+ printk(" lock(");
+ __print_lock_name(parent == source ? target : parent);
+ printk(KERN_CONT ");\n");
+ printk(" unlock(");
+ __print_lock_name(source);
+ printk(KERN_CONT ");\n");
+ printk("\n *** DEADLOCK ***\n\n");
} else {
printk(" Possible unsafe locking scenario:\n\n");
printk(" CPU0 CPU1\n");
--
2.14.1
\
 
 \ /
  Last update: 2017-08-23 06:38    [W:0.255 / U:0.184 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site