Re: [BUG] CPU hotplug, freezer: Freezing of tasks failed after 20.00 seconds

["Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com>]

On 09/06/2011 11:31 AM, Rafael J. Wysocki wrote:
> On Tuesday, September 06, 2011, Tejun Heo wrote:
>> Hello, again.
>>
>> On Mon, Sep 05, 2011 at 11:15:12PM +0900, Tejun Heo wrote:
>>>>  Freezing of tasks failed after 20.01 seconds (2 tasks refusing to freeze, wq_busy=0):
>>>>  invert_cpu_stat D 0000000000000000  5304 20435  17329 0x00000084
>>>>   ffff8801f367bab8 0000000000000046 ffff8801f367bfd8 00000000001d3a00
>>>>   ffff8801f367a010 00000000001d3a00 00000000001d3a00 00000000001d3a00
>>>>   ffff8801f367bfd8 00000000001d3a00 ffff880414cc6840 ffff8801f36783c0
>>>>  Call Trace:
>>>>   [<ffffffff81532de5>] schedule_timeout+0x235/0x320
>>>>   [<ffffffff81532a0b>] wait_for_common+0x11b/0x170
>>>>   [<ffffffff81532b3d>] wait_for_completion+0x1d/0x20
>>>>   [<ffffffff81364486>] _request_firmware+0x156/0x2c0
>>>>   [<ffffffff81364686>] request_firmware+0x16/0x20
>>>>   [<ffffffffa01f0da0>] request_microcode_fw+0x70/0xf0 [microcode]
>>>>   [<ffffffffa01f0390>] microcode_init_cpu+0xc0/0x100 [microcode]
>>>>   [<ffffffffa01f14b4>] mc_cpu_callback+0x7c/0x11f [microcode]
>>>>   [<ffffffff815393a4>] notifier_call_chain+0x94/0xd0
>>>>   [<ffffffff8109770e>] __raw_notifier_call_chain+0xe/0x10
>>>>   [<ffffffff8106d000>] __cpu_notify+0x20/0x40
>>>>   [<ffffffff8152cf5b>] _cpu_up+0xc7/0x10e
>>>>   [<ffffffff8152d07b>] cpu_up+0xd9/0xec
>>>>   [<ffffffff8151e599>] store_online+0x99/0xd0
>>>>   [<ffffffff81355eb0>] sysdev_store+0x20/0x30
>>>>   [<ffffffff811f3096>] sysfs_write_file+0xe6/0x170
>>>>   [<ffffffff8117ee50>] vfs_write+0xd0/0x1a0
>>>>   [<ffffffff8117f024>] sys_write+0x54/0xa0
>>>>   [<ffffffff8153df02>] system_call_fastpath+0x16/0x1b
>>>
>>> So, this task is trying to bring a CPU up, which triggers firmware
>>> helper to load microcode.  Firmware class currently sleeps
>>> non-interruptibly to wait for firmware load to complete, which is
>>> performed by another userland task.  Now, the PM freezer doesn't
>>> assume that there will be non-freezable wait dependencies among
>>> userland tasks.  It only knows two levels - userland and kernel tasks
>>> - and assumes that the former group may have non-freezable wait
>>> dependency on the latter but there's no such dependency among each
>>> group itself.  If there's such dependency, PM freezer may fail, which
>>> is what happened here.
>>>
>>> ie. the firmware loader userland process got frozen first.
>>> invert_cpu_stat trying to bring up CPU was waiting for the firmware
>>> loader to finish in non-interruptible sleep, so the freezer couldn't
>>> proceed.
>>
>> Hmmm... I went through the code again and usermodehelper_disable()
>> seems to be there to prevent deadlocks like this.  usermode helpers
>> are drained & plugged before freezing is tried.  Rafael, the above
>> shouldn't be happening, right?
> 
> No, it shouldn't in theory, but I'm not sure any more after the recent
> modifications of firmware loading related to the initialization.  I'll have
> a closer look tomorrow.
> 

Hi,
I have posted a fix for this bug at https://lkml.org/lkml/2011/10/2/142
With my fix, the numerous "WARNING"s at drivers/base/firmware_class.c
disappear and the task freezing failures are fixed too.
I have tested this for about 10-12 hours (much more time than what was
necessary to reproduce the bug earlier).

-- 
Regards,
Srivatsa S. Bhat  <srivatsa.bhat@linux.vnet.ibm.com>
Linux Technology Center,
IBM India Systems and Technology Lab