Re: BUG triggers running lsof

[Randy Dunlap <rdunlap@infradead.org>]

On 11/20/20 12:59 PM, K.R. Foley wrote:
> 
> 
> 
> On 2020-11-20 13:51, Jeff Moyer wrote:
>> Randy Dunlap <rdunlap@infradead.org> writes:
>>
>>> On 11/20/20 11:16 AM, K.R. Foley wrote:
>>>> I have found an issue that triggers by running lsof. The problem is
>>>> reproducible, but not consistently. I have seen this issue occur on
>>>> multiple versions of the kernel (5.0.10, 5.2.8 and now 5.4.77). It
>>>> looks like it could be a race condition or the file pointer is being
>>>> corrupted. Any pointers on how to track this down? What additional
>>>> information can I provide?
>>>
>>> Hi,
>>>
>>> 2 things in general:
>>>
>>> a) Can you test with a more recent kernel?
>>>
>>> b) Can you reproduce this without loading the proprietary & out-of-tree
>>> kernel modules?  They should never have been loaded after bootup.
>>> I.e., don't just unload them -- that could leave something bad behind.
>>
>> Heh, the EIP contains part of the name of one of the modules:
>>
>>>
>>>> [ 8057.297159] BUG: unable to handle page fault for address: 31376f63
>>                                                                 ^^^^^^^^

Thanks for noticing that, Jeff.  I should have seen it.

>>>> [ 8057.297219] Modules linked in: ITXico7100Module(O)
>>                                          ^^^^
> 
> Perhaps this is a dumb question, but how could this happen?


We don't know what is in that loadable kernel module, so we can't
give a definitive answer to your question, other than it's buggy.
Or maybe it was just written for an older kernel version.
Or a kernel with different build options/settings.

Have you contacted IT support?

It would (will) be interesting to see if you can reproduce the problem
without these modules being loaded...
I kind of doubt it, but if it does still fail, it will give us something
to look at.

-- 
~Randy