[pvrusb2] Still getting kernel oops on pvrusb2 driver removal

[JE Geiger]

I will recompile when the final 2.6.34 is released and rerun with more
kernel debug enabled.

It does spew quite a few kernel object debug messages on load.

Also, once I get the final, I will try to duplicate and capture the
lost oops regarding 6b poison and task structure cleanup.

Maybe clues, I don't know.

Many of the comments I read regarding similar oops from other
subsystems commented about calling the string length function with a
garbage pointer.  Typical of some dereference of some pointer in some
structure after the structure had been freed.

Good news is that the pvrusb2 works fine in 2.6.34-rc5 even if it
craters on rmmod.  Just use it and don't unload it.

On Thu, Apr 22, 2010 at 11:13 AM, Mike Isely <isely at isely.net> wrote:
>
> Thanks for the info.  I've been saving all of these messages.  This
> Fri/Sat I should be able to make some headway on the problem.  I will
> also likely be active in IRC over that time if someone wants to chat
> about it.
>
> BTW, the other oops about stack-allocated memory being used for DMA is
> something I do understand.  The driver generates USB transfers through
> several pathways.  It's entirely possible that one or more of these
> paths is using small stack allocated transfer buffers.  This is logic
> that has been this way since at least 2005 - first time I've heard that
> this could be a problem.  Likely it's some kind of new constraint with
> the later kernels.  But if this is the case it's easy to fix so I'm not
> worried about it.
>
>  -Mike
>
>
> On Thu, 22 Apr 2010, JE Geiger wrote:
>
>> Compiled 2.6.34-rc5 stock kernel.  Non-tainted with nvidia (had not
>> built the nvidia driver module at time of testing).
>>
>> Did modprobe pvrusb2 and got normal stuff:
>>
>> http://www.kd4ab.org/modin1
>>
>>
>> Did a rmmod pvrusb2 and got the same oops:
>>
>> http://www.kd4ab.org/rmout1
>>
>>
>>
>> I saw an oops that I was unable to repeat that talked about Poison
>> Overwritten.  It appears that 6b6b6b6b..... is a marker that detects a
>> random write to a defined structure and/or memory page.  There were
>> also comments about Task Structure being fixed.  These are kernel
>> issues.
>>
>>
>>
>> The more I read, the more I believe that this oops is due to some
>> underlying kernel issue and not the pvrusb2 driver.
>>
>> The pvrusb2 driver works fine when loaded.  I just can't unload it
>> without generating an oops.
>> _______________________________________________
>> pvrusb2 mailing list
>> pvrusb2 at isely.net
>> http://www.isely.net/cgi-bin/mailman/listinfo/pvrusb2
>>
>
> --
>
> Mike Isely
> isely @ isely (dot) net
> PGP: 03 54 43 4D 75 E5 CC 92 71 16 01 E2 B5 F5 C1 E8
> _______________________________________________
> pvrusb2 mailing list
> pvrusb2 at isely.net
> http://www.isely.net/cgi-bin/mailman/listinfo/pvrusb2
>