[ath9k-devel] [PATCH] ath9k: Check for NULL sta in ath_tx_start

[Ben Greear <greearb@candelatech.com>]

On 12/07/2010 09:17 PM, Vasanthakumar Thiagarajan wrote:
> On Tue, Dec 07, 2010 at 10:43:38PM +0530, Ben Greear wrote:
>> On 12/07/2010 12:31 AM, Vasanthakumar Thiagarajan wrote:
>>> On Tue, Dec 07, 2010 at 01:19:22PM +0530, Ben Greear wrote:
>>>> On 12/06/2010 11:42 PM, Vasanthakumar Thiagarajan wrote:
>>>>> On Tue, Dec 07, 2010 at 11:06:24AM +0530, Ben Greear wrote:
>>>>>> On 12/06/2010 09:21 PM, Luis R. Rodriguez wrote:
>>>>>>> On Mon, Dec 6, 2010 at 9:13 PM,<greearb@gmail.com>     wrote:
>>>>>>>> From: Ben Greear<greearb@candelatech.com>
>>>>>>>>
>>>>>>>> It can be NULL according to docs, and logging showed it
>>>>>>>> to be NULL in practice.
>>>>>>>>
>>>>>>>> Signed-off-by: Ben Greear<greearb@candelatech.com>
>>>>>>>
>>>>>>> Does this fix an oops? If so can you explain and provide the trace and
>>>>>>> resubmit and cc stable in the commit log?
>>>>>>
>>>>>> I think it fixes the TID corruption I posted about earlier.  It seems
>>>>>> so obvious though, that I'm curious why no one else sees problems,
>>>>>> and why I didn't see more crashes in my setup.
>>>>>>
>>>>>> (The paprd code appears to send with null STA, for instance, and my
>>>>>> printks showed lots of NULL stas in my 16-sta test setup, though I
>>>>>> don't think I'm using the paprd code path.)
>>>>>
>>>>> paprd is used only with>= AR9003.
>>>>
>>>> Whoever coded it up hopefully had that hardware...so why didn't
>>>> they see lots of crashes?
>>>
>>> I myself tested it lot of times, but did not see any crash, weird.
>>
>> Looks like the offending change when in recently (11/4/10, one of Felix's
>> patches).
>>
>> That is probably why no one else is hitting this yet, and it
>> isn't needed for stable I'm guessing....
>
> Nice, care to send a patch?. Yeah, it is not needed for stable.

The same patch I originally posted to to start this thread
should do just fine.

If someone really wants an oops report, I can add an artificial BUG_ON
to catch this, but for whatever reason, it normally just crashes things
weirdly and perhaps sometimes screwed up the file-system, so I'd rather
just have the patch go in as it is..I think it is obviously more correct
than the existing code.

Thanks,
Ben

>
> Vasanth


-- 
Ben Greear <greearb@candelatech.com>
Candela Technologies Inc  http://www.candelatech.com