Re: [BUGFIX 1/2] PCI/PM: Fix deadlock when unbind device if its parent in D3cold

[Bjorn Helgaas <bhelgaas@google.com>]

On Sat, Nov 3, 2012 at 9:38 PM, Huang Ying <ying.huang@intel.com> wrote:
> On Sat, 2012-11-03 at 11:22 -0600, Bjorn Helgaas wrote:
>> On Fri, Nov 2, 2012 at 11:06 PM, Huang Ying <ying.huang@intel.com> wrote:
>> > On Fri, 2012-11-02 at 10:50 -0600, Bjorn Helgaas wrote:
>> >> On Wed, Oct 24, 2012 at 12:54 AM, Huang Ying <ying.huang@intel.com> wrote:
>> >> > If a PCI device and its parents are put into D3cold, unbinding the
>> >> > device will trigger deadlock as follow:
>> >> >
>> >> > - driver_unbind
>> >> >   - device_release_driver
>> >> >     - device_lock(dev)                          <--- previous lock here
>> >> >     - __device_release_driver
>> >> >       - pm_runtime_get_sync
>> >> >         ...
>> >> >           - rpm_resume(dev)
>> >> >             - rpm_resume(dev->parent)
>> >> >               ...
>> >> >                 - pci_pm_runtime_resume
>> >> >                   ...
>> >> >                   - pci_set_power_state
>> >> >                     - __pci_start_power_transition
>> >> >                       - pci_wakeup_bus(dev->parent->subordinate)
>> >> >                         - pci_walk_bus
>> >> >                           - device_lock(dev)    <--- dead lock here
>> >> >
>> >> >
>> >> > If we do not do device_lock in pci_walk_bus, we can avoid dead lock.
>> >> > Device_lock in pci_walk_bus is introduced in commit:
>> >> > d71374dafbba7ec3f67371d3b7e9f6310a588808, corresponding email thread
>> >> > is: https://lkml.org/lkml/2006/5/26/38.  The patch author Zhang Yanmin
>> >> > said device_lock is added to pci_walk_bus because:
>> >> >
>> >> >   Some error handling functions call pci_walk_bus. For example, PCIe
>> >> >   aer. Here we lock the device, so the driver wouldn't detach from the
>> >> >   device, as the cb might call driver's callback function.
>> >> >
>> >> > So I fixed the dead lock as follow:
>> >> >
>> >> > - remove device_lock from pci_walk_bus
>> >> > - add device_lock into callback if callback will call driver's callback
>> >> >
>> >> > I checked pci_walk_bus users one by one, and found only PCIe aer needs
>> >> > device lock.
>> >>
>> >> Is there a problem report or bugzilla for this issue?
>> >
>> > No.  I found this issue when I developed kexec fixing.
>> >
>> > Best Regards,
>> > Huang Ying
>> >
>> >> > Signed-off-by: Huang Ying <ying.huang@intel.com>
>> >> > Cc: Zhang Yanmin <yanmin.zhang@intel.com>
>> >>
>> >> Should this go to stable as well?  The D3cold support appeared in
>> >> v3.6, so my guess is that this fix could go to v3.6.x.
>>
>> What about the stable tree?
>
> You are right.  This fix should go to v3.6.x stable tree.

I applied these two patches to my for-linus branch as v3.7 material.  Thanks!