this commit:

==
1331e7a1bbe1f11b19c4327ba0853bee2a606543 is the first bad commit
commit 1331e7a1bbe1f11b19c4327ba0853bee2a606543
Author: Paul E. McKenney <paul.mcken...@linaro.org>
Date:   Thu Aug 2 17:43:50 2012 -0700

    rcu: Remove _rcu_barrier() dependency on __stop_machine()

    Currently, _rcu_barrier() relies on preempt_disable() to prevent
    any CPU from going offline, which in turn depends on CPU hotplug's
    use of __stop_machine().

    This patch therefore makes _rcu_barrier() use get_online_cpus() to
    block CPU-hotplug operations.  This has the added benefit of removing
    the need for _rcu_barrier() to adopt callbacks:  Because CPU-hotplug
    operations are excluded, there can be no callbacks to adopt.  This
    commit simplifies the code accordingly.

    Signed-off-by: Paul E. McKenney <paul.mcken...@linaro.org>
    Signed-off-by: Paul E. McKenney <paul...@linux.vnet.ibm.com>
    Reviewed-by: Josh Triplett <j...@joshtriplett.org>
==

is causing lockdep to complain (see the full trace below). I haven't yet
had time to analyze what exactly is happening, and probably will not have
time to do so until tomorrow, so just sending this as a heads-up in case
anyone sees the culprit immediately.

 ======================================================
 [ INFO: possible circular locking dependency detected ]
 3.6.0-rc5-00004-g0d8ee37 #143 Not tainted
 -------------------------------------------------------
 kworker/u:2/40 is trying to acquire lock:
  (rcu_sched_state.barrier_mutex){+.+...}, at: [<ffffffff810f2126>] _rcu_barrier+0x26/0x1e0

 but task is already holding lock:
  (slab_mutex){+.+.+.}, at: [<ffffffff81176e15>] kmem_cache_destroy+0x45/0xe0

 which lock already depends on the new lock.

 the existing dependency chain (in reverse order) is:

 -> #2 (slab_mutex){+.+.+.}:
        [<ffffffff810ae1e2>] validate_chain+0x632/0x720
        [<ffffffff810ae5d9>] __lock_acquire+0x309/0x530
        [<ffffffff810ae921>] lock_acquire+0x121/0x190
        [<ffffffff8155d4cc>] __mutex_lock_common+0x5c/0x450
        [<ffffffff8155d9ee>] mutex_lock_nested+0x3e/0x50
        [<ffffffff81558cb5>] cpuup_callback+0x2f/0xbe
        [<ffffffff81564b83>] notifier_call_chain+0x93/0x140
        [<ffffffff81076f89>] __raw_notifier_call_chain+0x9/0x10
        [<ffffffff8155719d>] _cpu_up+0xba/0x14e
        [<ffffffff815572ed>] cpu_up+0xbc/0x117
        [<ffffffff81ae05e3>] smp_init+0x6b/0x9f
        [<ffffffff81ac47d6>] kernel_init+0x147/0x1dc
        [<ffffffff8156ab44>] kernel_thread_helper+0x4/0x10

 -> #1 (cpu_hotplug.lock){+.+.+.}:
        [<ffffffff810ae1e2>] validate_chain+0x632/0x720
        [<ffffffff810ae5d9>] __lock_acquire+0x309/0x530
        [<ffffffff810ae921>] lock_acquire+0x121/0x190
        [<ffffffff8155d4cc>] __mutex_lock_common+0x5c/0x450
        [<ffffffff8155d9ee>] mutex_lock_nested+0x3e/0x50
        [<ffffffff81049197>] get_online_cpus+0x37/0x50
        [<ffffffff810f21bb>] _rcu_barrier+0xbb/0x1e0
        [<ffffffff810f22f0>] rcu_barrier_sched+0x10/0x20
        [<ffffffff810f2309>] rcu_barrier+0x9/0x10
        [<ffffffff8118c129>] deactivate_locked_super+0x49/0x90
        [<ffffffff8118cc01>] deactivate_super+0x61/0x70
        [<ffffffff811aaaa7>] mntput_no_expire+0x127/0x180
        [<ffffffff811ab49e>] sys_umount+0x6e/0xd0
        [<ffffffff81569979>] system_call_fastpath+0x16/0x1b

 -> #0 (rcu_sched_state.barrier_mutex){+.+...}:
        [<ffffffff810adb4e>] check_prev_add+0x3de/0x440
        [<ffffffff810ae1e2>] validate_chain+0x632/0x720
        [<ffffffff810ae5d9>] __lock_acquire+0x309/0x530
        [<ffffffff810ae921>] lock_acquire+0x121/0x190
        [<ffffffff8155d4cc>] __mutex_lock_common+0x5c/0x450
        [<ffffffff8155d9ee>] mutex_lock_nested+0x3e/0x50
        [<ffffffff810f2126>] _rcu_barrier+0x26/0x1e0
        [<ffffffff810f22f0>] rcu_barrier_sched+0x10/0x20
        [<ffffffff810f2309>] rcu_barrier+0x9/0x10
        [<ffffffff81176ea1>] kmem_cache_destroy+0xd1/0xe0
        [<ffffffffa04c3154>] nf_conntrack_cleanup_net+0xe4/0x110 [nf_conntrack]
        [<ffffffffa04c31aa>] nf_conntrack_cleanup+0x2a/0x70 [nf_conntrack]
        [<ffffffffa04c42ce>] nf_conntrack_net_exit+0x5e/0x80 [nf_conntrack]
        [<ffffffff81454b79>] ops_exit_list+0x39/0x60
        [<ffffffff814551ab>] cleanup_net+0xfb/0x1b0
        [<ffffffff8106917b>] process_one_work+0x26b/0x4c0
        [<ffffffff81069f3e>] worker_thread+0x12e/0x320
        [<ffffffff8106f73e>] kthread+0x9e/0xb0
        [<ffffffff8156ab44>] kernel_thread_helper+0x4/0x10

 other info that might help us debug this:

 Chain exists of:
   rcu_sched_state.barrier_mutex --> cpu_hotplug.lock --> slab_mutex

  Possible unsafe locking scenario:

        CPU0                    CPU1
        ----                    ----
   lock(slab_mutex);
                                lock(cpu_hotplug.lock);
                                lock(slab_mutex);
   lock(rcu_sched_state.barrier_mutex);

  *** DEADLOCK ***

 4 locks held by kworker/u:2/40:
  #0:  (netns){.+.+.+}, at: [<ffffffff810690b2>] process_one_work+0x1a2/0x4c0
  #1:  (net_cleanup_work){+.+.+.}, at: [<ffffffff810690b2>] process_one_work+0x1a2/0x4c0
  #2:  (net_mutex){+.+.+.}, at: [<ffffffff81455130>] cleanup_net+0x80/0x1b0
  #3:  (slab_mutex){+.+.+.}, at: [<ffffffff81176e15>] kmem_cache_destroy+0x45/0xe0

 stack backtrace:
 Pid: 40, comm: kworker/u:2 Not tainted 3.6.0-rc5-00004-g0d8ee37 #143
 Call Trace:
  [<ffffffff810ac85f>] print_circular_bug+0x10f/0x120
  [<ffffffff810adb4e>] check_prev_add+0x3de/0x440
  [<ffffffff810ad85a>] ? check_prev_add+0xea/0x440
  [<ffffffff8102c72f>] ? flat_send_IPI_mask+0x7f/0xc0
  [<ffffffff810ae1e2>] validate_chain+0x632/0x720
  [<ffffffff810ae5d9>] __lock_acquire+0x309/0x530
  [<ffffffff810ae921>] lock_acquire+0x121/0x190
  [<ffffffff810f2126>] ? _rcu_barrier+0x26/0x1e0
  [<ffffffff8155d4cc>] __mutex_lock_common+0x5c/0x450
  [<ffffffff810f2126>] ? _rcu_barrier+0x26/0x1e0
  [<ffffffff810b5e45>] ? on_each_cpu+0x65/0xc0
  [<ffffffff810f2126>] ? _rcu_barrier+0x26/0x1e0
  [<ffffffff8155d9ee>] mutex_lock_nested+0x3e/0x50
  [<ffffffff810f2126>] _rcu_barrier+0x26/0x1e0
  [<ffffffff810f22f0>] rcu_barrier_sched+0x10/0x20
  [<ffffffff810f2309>] rcu_barrier+0x9/0x10
  [<ffffffff81176ea1>] kmem_cache_destroy+0xd1/0xe0
  [<ffffffffa04c3154>] nf_conntrack_cleanup_net+0xe4/0x110 [nf_conntrack]
  [<ffffffffa04c31aa>] nf_conntrack_cleanup+0x2a/0x70 [nf_conntrack]
  [<ffffffffa04c42ce>] nf_conntrack_net_exit+0x5e/0x80 [nf_conntrack]
  [<ffffffff81454b79>] ops_exit_list+0x39/0x60
  [<ffffffff814551ab>] cleanup_net+0xfb/0x1b0
  [<ffffffff8106917b>] process_one_work+0x26b/0x4c0
  [<ffffffff810690b2>] ? process_one_work+0x1a2/0x4c0
  [<ffffffff81069e69>] ? worker_thread+0x59/0x320
  [<ffffffff814550b0>] ? net_drop_ns+0x40/0x40
  [<ffffffff81069f3e>] worker_thread+0x12e/0x320
  [<ffffffff81069e10>] ? manage_workers+0x110/0x110
  [<ffffffff8106f73e>] kthread+0x9e/0xb0
  [<ffffffff8156ab44>] kernel_thread_helper+0x4/0x10
  [<ffffffff81560b70>] ? retint_restore_args+0x13/0x13
  [<ffffffff8106f6a0>] ? __init_kthread_worker+0x70/0x70
  [<ffffffff8156ab40>] ? gs_change+0x13/0x13

--
Jiri Kosina
SUSE Labs
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/