Why hold device_lock when calling callback in pci_walk_bus?

Why hold device_lock when calling callback in pci_walk_bus?

[Huang Ying]

Hi, All,

If my understanding were correct, device_lock is used to provide mutual
exclusion between device probe/remove/suspend/resume etc.  Why hold
device_lock when calling callback in pci_walk_bus.

This is introduced by the following commit.

commit d71374dafbba7ec3f67371d3b7e9f6310a588808
Author: Zhang Yanmin <yanmin.zhang@intel.com>
Date:   Fri Jun 2 12:35:43 2006 +0800

    [PATCH] PCI: fix race with pci_walk_bus and pci_destroy_dev
    
    pci_walk_bus has a race with pci_destroy_dev. When cb is called
    in pci_walk_bus, pci_destroy_dev might unlink the dev pointed by next.
    Later on in the next loop, pointer next becomes NULL and cause
    kernel panic.
    
    Below patch against 2.6.17-rc4 fixes it by changing pci_bus_lock (spin_lock)
    to pci_bus_sem (rw_semaphore).
    
    Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com>
    Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

Corresponding email thread is: https://lkml.org/lkml/2006/5/26/38

But from the commit and email thread, I can not find why we need to do
that.

I ask this question because I want to use pci_walk_bus in a function (in
pci runtime resume path) which may be called with device_lock held.

Can anyone help me on that?

Best Regards,
Huang Ying

RE: Why hold device_lock when calling callback in pci_walk_bus?

[Zhang, Yanmin]
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Re: Why hold device_lock when calling callback in pci_walk_bus?

[Huang Ying]

Hi, Yanmin,

Thanks for your explain.

On Fri, 2012-09-28 at 02:29 -0600, Zhang, Yanmin wrote:
> Some error handling functions call pci_walk_bus. For example, pci-e aer. Here we lock the device, so the driver wouldn't detach from the device, as the cb might call driver's callback function.

Still has two question.

1. Is it a good practice to hold device_lock when calling driver
callback to prevent driver be unbind?

2. Is it a good idea to let callback of pci_walk_bus to acquire
device_lock when necessary.  Because pci_walk_bus may be used by driver
callback too.

Best Regards,
Huang Ying

> -----Original Message-----
> From: Huang, Ying 
> Sent: Friday, September 28, 2012 4:15 PM
> To: bhelgaas@google.com
> Cc: Greg Kroah-Hartman; Zhang, Yanmin; linux-pci@vger.kernel.org; linux-kernel@vger.kernel.org; rjw@sisk.pl
> Subject: Why hold device_lock when calling callback in pci_walk_bus?
> 
> Hi, All,
> 
> If my understanding were correct, device_lock is used to provide mutual exclusion between device probe/remove/suspend/resume etc.  Why hold device_lock when calling callback in pci_walk_bus.
> 
> This is introduced by the following commit.
> 
> commit d71374dafbba7ec3f67371d3b7e9f6310a588808
> Author: Zhang Yanmin <yanmin.zhang@intel.com>
> Date:   Fri Jun 2 12:35:43 2006 +0800
> 
>     [PATCH] PCI: fix race with pci_walk_bus and pci_destroy_dev
>     
>     pci_walk_bus has a race with pci_destroy_dev. When cb is called
>     in pci_walk_bus, pci_destroy_dev might unlink the dev pointed by next.
>     Later on in the next loop, pointer next becomes NULL and cause
>     kernel panic.
>     
>     Below patch against 2.6.17-rc4 fixes it by changing pci_bus_lock (spin_lock)
>     to pci_bus_sem (rw_semaphore).
>     
>     Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com>
>     Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
> 
> Corresponding email thread is: https://lkml.org/lkml/2006/5/26/38
> 
> But from the commit and email thread, I can not find why we need to do that.
> 
> I ask this question because I want to use pci_walk_bus in a function (in pci runtime resume path) which may be called with device_lock held.
> 
> Can anyone help me on that?
> 
> Best Regards,
> Huang Ying
> 
>