Re: pmu_dev_alloc; warning at at kernel/locking/lockdep.c:3002 lockdep_init_map

From: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
To: Tejun Heo <tj@kernel.org>, Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org, Ingo Molnar <mingo@redhat.com>,
	Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	NeilBrown <neilb@suse.de>
Subject: Re: pmu_dev_alloc; warning at at kernel/locking/lockdep.c:3002 lockdep_init_map
Date: Thu, 07 May 2015 17:29:57 +0100	[thread overview]
Message-ID: <554B9305.301@linux.intel.com> (raw)
In-Reply-To: <20150507162523.GB12245@htj.duckdns.org>

On 05/07/2015 05:25 PM, Tejun Heo wrote:
> Hello,
> 
> On Thu, May 07, 2015 at 04:05:57PM +0200, Peter Zijlstra wrote:
>> On Thu, May 07, 2015 at 02:54:00PM +0100, Tvrtko Ursulin wrote:
>>>
>>> Hi,
>>>
>>> 4.1.0-rc2 spews the below warning and disables lockdep for me at boot.
>>>
>>> Any ideas?
>>>
>>> May  7 11:58:20 skl kernel: [    6.066696] futex hash table entries: 1024 (order: 5, 131072 bytes)
>>> May  7 11:58:20 skl kernel: [    6.075774] BUG: key ffff88014973b850 not in .data!
> ...
>>> May  7 11:58:20 skl kernel: [    6.188202]  [<ffffffff811dc06a>] __kernfs_create_file+0x7a/0xf0
>>> May  7 11:58:20 skl kernel: [    6.195069]  [<ffffffff811dcb4e>] sysfs_add_file_mode_ns+0x16e/0x1b0
>>> May  7 11:58:20 skl kernel: [    6.202346]  [<ffffffff811dd92a>] internal_create_group+0x19a/0x280
>>> May  7 11:58:20 skl kernel: [    6.209515]  [<ffffffff811dda43>] sysfs_create_group+0x13/0x20
>>> May  7 11:58:20 skl kernel: [    6.216185]  [<ffffffff811dda88>] sysfs_create_groups+0x38/0x90
>>> May  7 11:58:20 skl kernel: [    6.222958]  [<ffffffff8145c3d6>] device_add+0x276/0x5d0
>>> May  7 11:58:20 skl kernel: [    6.229040]  [<ffffffff8110a80b>] pmu_dev_alloc+0x8b/0xd0
>>> May  7 11:58:20 skl kernel: [    6.242184]  [<ffffffff81d44cd9>] perf_event_sysfs_init+0x46/0x95
>>> May  7 11:58:20 skl kernel: [    6.256118]  [<ffffffff81000352>] do_one_initcall+0x122/0x1f0
>>> May  7 11:58:20 skl kernel: [    6.268980]  [<ffffffff81d27126>] kernel_init_freeable+0x106/0x19a
>>> May  7 11:58:20 skl kernel: [    6.288120]  [<ffffffff8161a38e>] kernel_init+0xe/0xf0
>>> May  7 11:58:20 skl kernel: [    6.294004]  [<ffffffff81629e92>] ret_from_fork+0x42/0x70
>>> May  7 11:58:20 skl kernel: [    6.306367] ---[ end trace 98f2b03275198a56 ]---
>>> May  7 11:58:20 skl kernel: [    6.311703] BUG: key ffff88014973b888 not in .data!
>>> May  7 11:58:20 skl kernel: [    6.317300] BUG: key ffff88014973b8c0 not in .data!
>>> May  7 11:58:20 skl kernel: [    6.322895] BUG: key ffff88014973b8f8 not in .data!
>>> May  7 11:58:20 skl kernel: [    6.329549] HugeTLB registered 2 MB page size, pre-allocated 0 pages
>>
>> That's in sysfs magic mushroom land, lets Cc people who know about this.
> 
> sysfs by default uses lockdep key embedded in attrs.  It looks like
> somebody is making on-heap copies of attrs and using them as the group
> attrs.  What's the offending pmu type?

I am not familiar with this area, is this the answer?

May  7 11:58:20 skl kernel: [    0.849877] Performance Events: no PEBS fmt3+, generic architected perfmon, full-width counters, Intel PMU driver.
May  7 11:58:20 skl kernel: [    0.861827] ... version:                4
May  7 11:58:20 skl kernel: [    0.866406] ... bit width:              48
May  7 11:58:20 skl kernel: [    0.871083] ... generic registers:      4
May  7 11:58:20 skl kernel: [    0.875664] ... value mask:             0000ffffffffffff
May  7 11:58:20 skl kernel: [    0.881724] ... max period:             0000ffffffffffff
May  7 11:58:20 skl kernel: [    0.887778] ... fixed-purpose events:   3
May  7 11:58:20 skl kernel: [    0.892352] ... event mask:             000000070000000f
May  7 11:58:20 skl kernel: [    0.909665] NMI watchdog: enabled on all CPUs, permanently consumes one hw-PMU counter.

Regards,

Tvrtko