Re: [GIT PATCHES FOR 2.6.37] davinci & videobuf fixes

[Mauro Carvalho Chehab <mchehab@redhat.com>]

Em 22-09-2010 15:49, Hans Verkuil escreveu:
> On Wednesday, September 22, 2010 20:39:03 Mauro Carvalho Chehab wrote:
>> Em 07-09-2010 06:23, Hans Verkuil escreveu:
>>> Hi Mauro,
>>>
>>> The following changes since commit 50b9d21ae2ac1b85be46f1ee5aa1b5e588622361:
>>>   Jarod Wilson (1):
>>>         V4L/DVB: mceusb: add two new ASUS device IDs
>>>
>>> are available in the git repository at:
>>>
>>>   ssh://linuxtv.org/git/hverkuil/v4l-dvb.git for-2.6.37
>>>
>>> Hans Verkuil (1):
>>>       videobuf-dma-sg: set correct size in last sg element
>>
>> Ok.
>>>
>>> Mats Randgaard (5):
>>>       videobuf-core.c: Replaced BUG_ON with WARN_ON
>>
>> Why? Please provide a description to allow us to understand why this change makes sense.
>> Is there any condition where this would be acceptable?
> 
> Does it indicate a driver bug? Yes. But it's not a harmful driver bug, i.e. it
> will not cause a crash later. So BUG_ON is way overkill. WARN_ON is sufficient.
> 
> A bug in the davinci driver actually triggered this BUG_ON for us. Since it's
> a BUG_ON you are forced to reboot for no good reason. Note that we are working
> on fixing the davinci bug as well.
> 
> AFAIK BUG_ON is meant for fatal errors, i.e. continuing will cause crashes later.
> That's not at all the case here.

Ok, please include this description on the patch. I'm fine with it. I was afraid that
this could be the default behavior for the davinci driver. The better is to put
this patch in the same series where you'll be also correcting the davinci bug.

> 
> Regards,
> 
> 	Hans
> 
>>>       vpif_cap/disp: Removed section mismatch warning
>>>       vpif_cap/disp: Replaced kmalloc with kzalloc
>>> 	vpif_cap: don't ignore return code of videobuf_poll_stream()
>>
>> The better would be to provide some description for all patches, but, in this
>> specific case, they're trivial.
>>
>> Applied, thanks.
>>
>>>       vpif_cap/disp: Fixed strlcpy NULL pointer bug
>>
>> Hmm... this one doesn't make sense to me. Instead, you should be sure that
>> config->card_name (or cap->card) is always filled.
>>
>> Ok, I've applied 4 of the 6 patches (I've included the patch of the last email
>> on the above).
>>
>> Please, provide a better solution for the strlcpy NULL pointer bug, properly
>> filling the config struct in all cases.
>>
>> Cheers,
>> Mauro
>>
> 

Cheers,
Mauro