[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20140707134521.GQ6758@twins.programming.kicks-ass.net>
Date: Mon, 7 Jul 2014 15:45:21 +0200
From: Peter Zijlstra <peterz@...radead.org>
To: Sasha Levin <sasha.levin@...cle.com>
Cc: Ingo Molnar <mingo@...nel.org>, acme@...stprotocols.net,
paulus@...ba.org, Tejun Heo <tj@...nel.org>,
LKML <linux-kernel@...r.kernel.org>,
Dave Jones <davej@...hat.com>
Subject: Re: perf/workqueue: lockdep warning on process exit
On Wed, Jun 25, 2014 at 05:17:48PM -0400, Sasha Levin wrote:
> [ 1853.817250] 2 locks held by trinity-c3/29876:
> [ 1853.817250] #0: cpu_hotplug.lock){++++++}, at: get_online_cpus (kernel/cpu.c:90)
> [ 1853.924037] #1: (&ctx->lock){-.-...}, at: perf_lock_task_context (kernel/events/core.c:983)
> [ 1853.924037]
> [ 1853.924037] Call Trace:
> [ 1853.924037] lock_acquire (./arch/x86/include/asm/current.h:14 kernel/locking/lockdep.c:3602)
> [ 1853.924037] _raw_spin_lock (include/linux/spinlock_api_smp.h:143 kernel/locking/spinlock.c:151)
> [ 1853.924037] __queue_work (kernel/workqueue.c:1339)
> [ 1853.924037] queue_work_on (kernel/workqueue.c:1417)
> [ 1853.924037] free_object (lib/debugobjects.c:209)
> [ 1853.924037] __debug_check_no_obj_freed (lib/debugobjects.c:715)
> [ 1853.924037] debug_check_no_obj_freed (lib/debugobjects.c:727)
> [ 1853.924037] kmem_cache_free (mm/slub.c:2657 mm/slub.c:2686)
> [ 1853.924037] free_task (kernel/fork.c:221)
> [ 1853.924037] __put_task_struct (kernel/fork.c:250)
> [ 1853.924037] put_ctx (include/linux/sched.h:1857 kernel/events/core.c:899)
> [ 1853.924037] find_get_context (kernel/events/core.c:908 kernel/events/core.c:3161)
> [ 1853.924037] SYSC_perf_event_open (kernel/events/core.c:7200)
> [ 1853.924037] ? preempt_count_sub (kernel/sched/core.c:2606)
> [ 1853.924037] ? context_tracking_user_exit (./arch/x86/include/asm/paravirt.h:809 (discriminator 2) kernel/context_tracking.c:184 (discriminator 2))
> [ 1853.924037] SyS_perf_event_open (kernel/events/core.c:7064)
Egads.. that shouldn't be possible.
The only put_ctx() from find_get_context() is in the !ctx branch, and if
perf_lock_task_context() returns NULL it shouldn't be holding ctx->lock.
Does this reproduce?
Content of type "application/pgp-signature" skipped
Powered by blists - more mailing lists