mm: Place preemption point in do_mlockall() loop
authorPaul E. McKenney <paulmck@linux.vnet.ibm.com>
Fri, 19 Jul 2013 22:09:25 +0000 (15:09 -0700)
committerPaul E. McKenney <paulmck@linux.vnet.ibm.com>
Fri, 20 Sep 2013 16:43:05 +0000 (09:43 -0700)
There is a loop in do_mlockall() that lacks a preemption point, which
means that the following can happen on non-preemptible builds of the
kernel:

> My fuzz tester keeps hitting this. Every instance shows the non-irq stack
> came in from mlockall.  I'm only seeing this on one box, but that has more
> ram (8gb) than my other machines, which might explain it.
>
>  Dave
>
> INFO: rcu_preempt self-detected stall on CPU { 3}  (t=6500 jiffies g=470344 c=470343 q=0)
> sending NMI to all CPUs:
> NMI backtrace for cpu 3
> CPU: 3 PID: 29664 Comm: trinity-child2 Not tainted 3.11.0-rc1+ #32
> task: ffff88023e743fc0 ti: ffff88022f6f2000 task.ti: ffff88022f6f2000
> RIP: 0010:[<ffffffff810bf7d1>]  [<ffffffff810bf7d1>] trace_hardirqs_off_caller+0x21/0xb0
> RSP: 0018:ffff880244e03c30  EFLAGS: 00000046
> RAX: ffff88023e743fc0 RBX: 0000000000000001 RCX: 000000000000003c
> RDX: 000000000000000f RSI: 0000000000000004 RDI: ffffffff81033cab
> RBP: ffff880244e03c38 R08: ffff880243288a80 R09: 0000000000000001
> R10: 0000000000000000 R11: 0000000000000001 R12: ffff880243288a80
> R13: ffff8802437eda40 R14: 0000000000080000 R15: 000000000000d010
> FS:  00007f50ae33b740(0000) GS:ffff880244e00000(0000) knlGS:0000000000000000
> CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> CR2: 000000000097f000 CR3: 0000000240fa0000 CR4: 00000000001407e0
> DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000600
> Stack:
>  ffffffff810bf86d ffff880244e03c98 ffffffff81033cab 0000000000000096
>  000000000000d008 0000000300000002 0000000000000004 0000000000000003
>  0000000000002710 ffffffff81c50d00 ffffffff81c50d00 ffff880244fcde00
> Call Trace:
>  <IRQ>
>  [<ffffffff810bf86d>] ? trace_hardirqs_off+0xd/0x10
>  [<ffffffff81033cab>] __x2apic_send_IPI_mask+0x1ab/0x1c0
>  [<ffffffff81033cdc>] x2apic_send_IPI_all+0x1c/0x20
>  [<ffffffff81030115>] arch_trigger_all_cpu_backtrace+0x65/0xa0
>  [<ffffffff811144b1>] rcu_check_callbacks+0x331/0x8e0
>  [<ffffffff8108bfa0>] ? hrtimer_run_queues+0x20/0x180
>  [<ffffffff8109e905>] ? sched_clock_cpu+0xb5/0x100
>  [<ffffffff81069557>] update_process_times+0x47/0x80
>  [<ffffffff810bd115>] tick_sched_handle.isra.16+0x25/0x60
>  [<ffffffff810bd231>] tick_sched_timer+0x41/0x60
>  [<ffffffff8108ace1>] __run_hrtimer+0x81/0x4e0
>  [<ffffffff810bd1f0>] ? tick_sched_do_timer+0x60/0x60
>  [<ffffffff8108b93f>] hrtimer_interrupt+0xff/0x240
>  [<ffffffff8102de84>] local_apic_timer_interrupt+0x34/0x60
>  [<ffffffff81718c5f>] smp_apic_timer_interrupt+0x3f/0x60
>  [<ffffffff817178ef>] apic_timer_interrupt+0x6f/0x80
>  [<ffffffff8170e8e0>] ? retint_restore_args+0xe/0xe
>  [<ffffffff8105f101>] ? __do_softirq+0xb1/0x440
>  [<ffffffff8105f64d>] irq_exit+0xcd/0xe0
>  [<ffffffff81718c65>] smp_apic_timer_interrupt+0x45/0x60
>  [<ffffffff817178ef>] apic_timer_interrupt+0x6f/0x80
>  <EOI>
>  [<ffffffff8170e8e0>] ? retint_restore_args+0xe/0xe
>  [<ffffffff8170b830>] ? wait_for_completion_killable+0x170/0x170
>  [<ffffffff8170c853>] ? preempt_schedule_irq+0x53/0x90
>  [<ffffffff8170e9f6>] retint_kernel+0x26/0x30
>  [<ffffffff8107a523>] ? queue_work_on+0x43/0x90
>  [<ffffffff8107c369>] schedule_on_each_cpu+0xc9/0x1a0
>  [<ffffffff81167770>] ? lru_add_drain+0x50/0x50
>  [<ffffffff811677c5>] lru_add_drain_all+0x15/0x20
>  [<ffffffff81186965>] SyS_mlockall+0xa5/0x1a0
>  [<ffffffff81716e94>] tracesys+0xdd/0xe2

This commit addresses this problem by inserting the required preemption
point.

Reported-by: Dave Jones <davej@redhat.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@gmail.com>
Cc: Michel Lespinasse <walken@google.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
mm/mlock.c

index d63802663242eb6ad13ca2ed065434f24fd7e5d1..67ba6da7d0e311b1b45007bbd0e37bd78db0f3a8 100644 (file)
@@ -736,6 +736,7 @@ static int do_mlockall(int flags)
 
                /* Ignore errors */
                mlock_fixup(vma, &prev, vma->vm_start, vma->vm_end, newflags);
+               cond_resched();
        }
 out:
        return 0;