iwlagn: crash when unloading the module

iwlagn: crash when unloading the module

[Kalle Valo]

Hi,

I noticed today a crash when removing iwlagn module. I can see the
crash everytime on my lenovo x201 (x86 64bit, i5) with this wireless
chip:

[   19.484551] iwlagn 0000:02:00.0: Detected Intel(R) Centrino(R)
Ultimate-N 6300 AGN, REV=0x74

As I'm travelling I only have two bad quality screenshots and 40 MB
video of the crash :) From the stacktrace it looks that this is iommu
related:

http://www.adurom.com/~kvalo/tmp/iwlagn/rmmod-iommu-crash/1.png
http://www.adurom.com/~kvalo/tmp/iwlagn/rmmod-iommu-crash/2.png

This is a new laptop and most likely I haven't removed iwlagn module
before. So this shouldn't be a regression.

-- 
Kalle Valo

Re: iwlagn: crash when unloading the module

[wwguy]

On Tue, 2011-05-03 at 03:20 -0700, Kalle Valo wrote:
> Hi,
> 
> I noticed today a crash when removing iwlagn module. I can see the
> crash everytime on my lenovo x201 (x86 64bit, i5) with this wireless
> chip:
> 
> [   19.484551] iwlagn 0000:02:00.0: Detected Intel(R) Centrino(R)
> Ultimate-N 6300 AGN, REV=0x74
> 
> As I'm travelling I only have two bad quality screenshots and 40 MB
> video of the crash :) From the stacktrace it looks that this is iommu
> related:
> 
> http://www.adurom.com/~kvalo/tmp/iwlagn/rmmod-iommu-crash/1.png
> http://www.adurom.com/~kvalo/tmp/iwlagn/rmmod-iommu-crash/2.png
> 
> This is a new laptop and most likely I haven't removed iwlagn module
> before. So this shouldn't be a regression.
> 

it is new to me as well, I never see that happen on my system. Could you
share some more information with me? such as laptop model, kernel
version, 32 or 64 bits.

Also is there any activity while you unload the iwlagn, is it in connect
state or not? 

Thanks
Wey

Re: iwlagn: crash when unloading the module

[Johannes Berg]

On Tue, 2011-05-03 at 07:42 -0700, wwguy wrote:

> > http://www.adurom.com/~kvalo/tmp/iwlagn/rmmod-iommu-crash/1.png
> > http://www.adurom.com/~kvalo/tmp/iwlagn/rmmod-iommu-crash/2.png
> > 
> > This is a new laptop and most likely I haven't removed iwlagn module
> > before. So this shouldn't be a regression.
> > 
> 
> it is new to me as well, I never see that happen on my system. Could you
> share some more information with me? such as laptop model, kernel
> version, 32 or 64 bits.
> 
> Also is there any activity while you unload the iwlagn, is it in connect
> state or not? 

Dave looked at this and it turned out to most likely be a bug in the
intel-iommu code. A tentative patch to fix it from him is here:
http://david.woodhou.se/double-unmap-check.patch

johannes

Re: iwlagn: crash when unloading the module

[Johannes Berg]

On Tue, 2011-05-03 at 17:07 +0200, Johannes Berg wrote:
> On Tue, 2011-05-03 at 07:42 -0700, wwguy wrote:
> 
> > > http://www.adurom.com/~kvalo/tmp/iwlagn/rmmod-iommu-crash/1.png
> > > http://www.adurom.com/~kvalo/tmp/iwlagn/rmmod-iommu-crash/2.png
> > > 
> > > This is a new laptop and most likely I haven't removed iwlagn module
> > > before. So this shouldn't be a regression.
> > > 
> > 
> > it is new to me as well, I never see that happen on my system. Could you
> > share some more information with me? such as laptop model, kernel
> > version, 32 or 64 bits.
> > 
> > Also is there any activity while you unload the iwlagn, is it in connect
> > state or not? 
> 
> Dave looked at this and it turned out to most likely be a bug in the
> intel-iommu code. A tentative patch to fix it from him is here:
> http://david.woodhou.se/double-unmap-check.patch

Err, no, http://david.woodhou.se/flush-unmaps-on-unbind.patch

johannes

Re: iwlagn: crash when unloading the module

[Kalle Valo]

Johannes Berg <johannes@sipsolutions.net> writes:

>> Dave looked at this and it turned out to most likely be a bug in the
>> intel-iommu code. A tentative patch to fix it from him is here:

[...]

> Err, no, http://david.woodhou.se/flush-unmaps-on-unbind.patch

Thanks. This patch fixes the issue. I was able to load and remove
iwlagn five times without a crash.

Tested-by: Kalle Valo <kvalo@adurom.com>

-- 
Kalle Valo

Re: iwlagn: crash when unloading the module

[Kalle Valo]

wwguy <wey-yi.w.guy@intel.com> writes:

> On Tue, 2011-05-03 at 03:20 -0700, Kalle Valo wrote:
>
>> I noticed today a crash when removing iwlagn module. I can see the
>> crash everytime on my lenovo x201 (x86 64bit, i5) with this wireless
>> chip:

[...]

> it is new to me as well, I never see that happen on my system. Could you
> share some more information with me? such as laptop model, kernel
> version, 32 or 64 bits.

This was in my original mail, see above :)

> Also is there any activity while you unload the iwlagn, is it in connect
> state or not? 

wlan0 interface was down.

-- 
Kalle Valo

Re: iwlagn: crash when unloading the module

[Stanislaw Gruszka]

On Tue, May 03, 2011 at 01:20:25PM +0300, Kalle Valo wrote:
> I noticed today a crash when removing iwlagn module. I can see the
> crash everytime on my lenovo x201 (x86 64bit, i5) with this wireless
> chip:
> 
> [   19.484551] iwlagn 0000:02:00.0: Detected Intel(R) Centrino(R)
> Ultimate-N 6300 AGN, REV=0x74
> 
> As I'm travelling I only have two bad quality screenshots and 40 MB
> video of the crash :) From the stacktrace it looks that this is iommu
> related:
> 
> http://www.adurom.com/~kvalo/tmp/iwlagn/rmmod-iommu-crash/1.png
> http://www.adurom.com/~kvalo/tmp/iwlagn/rmmod-iommu-crash/2.png
> 
> This is a new laptop and most likely I haven't removed iwlagn module
> before. So this shouldn't be a regression.

We can not exclude iwlagn problem only triggered with intel iommu,
because problem does not happen with other modules, or does it?
Except test patch pointed by Johannes, I suggest turning on
CONFIG_DMA_API_DEBUG and test with intel_iommu=on and intel_iommu=off.

Stanislaw

Re: iwlagn: crash when unloading the module

[Kalle Valo]

Stanislaw Gruszka <sgruszka@redhat.com> writes:

> On Tue, May 03, 2011 at 01:20:25PM +0300, Kalle Valo wrote:
>> I noticed today a crash when removing iwlagn module. I can see the
>> crash everytime on my lenovo x201 (x86 64bit, i5) with this wireless
>> chip:
>> 
>> [   19.484551] iwlagn 0000:02:00.0: Detected Intel(R) Centrino(R)
>> Ultimate-N 6300 AGN, REV=0x74

[...]

> We can not exclude iwlagn problem only triggered with intel iommu,
> because problem does not happen with other modules, or does it?

I have only seen the crash with iwlagn. But I have to admit that I
haven't tried to unload modules that much.

> Except test patch pointed by Johannes, I suggest turning on
> CONFIG_DMA_API_DEBUG

I already had this on and I haven't noticed anything wrong.

> and test with intel_iommu=on and intel_iommu=off.

I'll skip these as David's patch seem to already help.

-- 
Kalle Valo