Hi,
I've been seeing frequent kernel panics on wifi activity
(scp-ing a lot of files) with 4.13 on 2 different systems
which both use a brcmfmac4356-pcie wifi chip.
This is with this fix:
https://www.spinics.net/lists/linux-wireless/msg164178.html
already applied.
Here is a picture of the panic:
https://fedorapeople.org/~jwrdegoede/brcmfmac4356-pcie-4.13-panic.jpg
Reverting commit 270a6c1f65fe ("brcmfmac: rework headroom check in .start_xmit()"):
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit?id=270a6c1f65fe68a28a5d39cd405592c550b496c7
seems to fix this (I can no longer quickly reproduce the panic
by scp-ing a large amount of files).
Note I've first reverted the:
https://www.spinics.net/lists/linux-wireless/msg164178.html
fix so that I could cleanly revert commit 270a6c1f65fe.
Given that no code seems to use the statics commit 270a6c1f65fe
offers and it has been the cause of 2 different regressions now,
it might be best to just revert 270a6c1f65fe for 4.13.
Regards,
Hans
On 26-07-17 23:03, Hans de Goede wrote:
> Hi,
>
> I've been seeing frequent kernel panics on wifi activity
> (scp-ing a lot of files) with 4.13 on 2 different systems
> which both use a brcmfmac4356-pcie wifi chip.
>
> This is with this fix:
> https://www.spinics.net/lists/linux-wireless/msg164178.html
> already applied.
>
> Here is a picture of the panic:
>
> https://fedorapeople.org/~jwrdegoede/brcmfmac4356-pcie-4.13-panic.jpg
>
> Reverting commit 270a6c1f65fe ("brcmfmac: rework headroom check in
> .start_xmit()"):
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit?id=270a6c1f65fe68a28a5d39cd405592c550b496c7
>
>
> seems to fix this (I can no longer quickly reproduce the panic
> by scp-ing a large amount of files).
>
> Note I've first reverted the:
> https://www.spinics.net/lists/linux-wireless/msg164178.html
> fix so that I could cleanly revert commit 270a6c1f65fe.
>
> Given that no code seems to use the statics commit 270a6c1f65fe
> offers and it has been the cause of 2 different regressions now,
> it might be best to just revert 270a6c1f65fe for 4.13.
Hi Hans,
I suspect this patch [1] fixes that. Can you verify? And yes, I messed
up. Sorry.
Regards,
Arend
Hi,
On 26-07-17 23:08, Arend van Spriel wrote:
> + ref
>
> On 26-07-17 23:08, Arend van Spriel wrote:
>> On 26-07-17 23:03, Hans de Goede wrote:
>>> Hi,
>>>
>>> I've been seeing frequent kernel panics on wifi activity
>>> (scp-ing a lot of files) with 4.13 on 2 different systems
>>> which both use a brcmfmac4356-pcie wifi chip.
>>>
>>> This is with this fix:
>>> https://www.spinics.net/lists/linux-wireless/msg164178.html
>>> already applied.
>>>
>>> Here is a picture of the panic:
>>>
>>> https://fedorapeople.org/~jwrdegoede/brcmfmac4356-pcie-4.13-panic.jpg
>>>
>>> Reverting commit 270a6c1f65fe ("brcmfmac: rework headroom check in
>>> .start_xmit()"):
>>> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit?id=270a6c1f65fe68a28a5d39cd405592c550b496c7
>>>
>>>
>>> seems to fix this (I can no longer quickly reproduce the panic
>>> by scp-ing a large amount of files).
>>>
>>> Note I've first reverted the:
>>> https://www.spinics.net/lists/linux-wireless/msg164178.html
>>> fix so that I could cleanly revert commit 270a6c1f65fe.
>>>
>>> Given that no code seems to use the statics commit 270a6c1f65fe
>>> offers and it has been the cause of 2 different regressions now,
>>> it might be best to just revert 270a6c1f65fe for 4.13.
>>
>> Hi Hans,
>>
>> I suspect this patch [1] fixes that. Can you verify? And yes, I messed
>> up. Sorry.
>>
>> Regards,
>> Arend
>
> [1] https://patchwork.kernel.org/patch/9864575/
Ah yes I can confirm that that patch seems to fix things for me too
and its description matches with the panic.
Thank you for the quick reply.
Regards,
Hans
+ ref
On 26-07-17 23:08, Arend van Spriel wrote:
> On 26-07-17 23:03, Hans de Goede wrote:
>> Hi,
>>
>> I've been seeing frequent kernel panics on wifi activity
>> (scp-ing a lot of files) with 4.13 on 2 different systems
>> which both use a brcmfmac4356-pcie wifi chip.
>>
>> This is with this fix:
>> https://www.spinics.net/lists/linux-wireless/msg164178.html
>> already applied.
>>
>> Here is a picture of the panic:
>>
>> https://fedorapeople.org/~jwrdegoede/brcmfmac4356-pcie-4.13-panic.jpg
>>
>> Reverting commit 270a6c1f65fe ("brcmfmac: rework headroom check in
>> .start_xmit()"):
>> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit?id=270a6c1f65fe68a28a5d39cd405592c550b496c7
>>
>>
>> seems to fix this (I can no longer quickly reproduce the panic
>> by scp-ing a large amount of files).
>>
>> Note I've first reverted the:
>> https://www.spinics.net/lists/linux-wireless/msg164178.html
>> fix so that I could cleanly revert commit 270a6c1f65fe.
>>
>> Given that no code seems to use the statics commit 270a6c1f65fe
>> offers and it has been the cause of 2 different regressions now,
>> it might be best to just revert 270a6c1f65fe for 4.13.
>
> Hi Hans,
>
> I suspect this patch [1] fixes that. Can you verify? And yes, I messed
> up. Sorry.
>
> Regards,
> Arend
[1] https://patchwork.kernel.org/patch/9864575/
