Re: ath5k past 2.6.30 breaks monitor mode (and thus the aircrack suite)

["Gábor Stefanik" <netrolller.3d@gmail.com>]

2010/5/23 Gábor Stefanik <netrolller.3d@gmail.com>:
> 2010/5/23 Weedy <weedy2887@gmail.com>:
>> 2010/5/17 Gábor Stefanik <netrolller.3d@gmail.com>:
>>> What about cutting down the kernel to the absolute minimum required by
>>> your machine?
>>>
>>> Also, Kbuild supports incremental builds as long as you do not "make
>>> clean" before builds.
>>>
>> I just finished the cycle (told you it would take forever on this
>> crapbook) and this  is what I got. To be sure I have done multiple
>> boots with vmlinuz-2.6.31-rc5fffffffff-01634-geadac6b and
>> vmlinuz-2.6.31-rc5fffffffff-01635-g56d1de0 and confirmed it broken.
>> Patching todays compat-wireless tar with the output of "git diff
>> 56d1de0a21db28e41741cfa0a66e18bc8d920554
>> 56d1de0a21db28e41741cfa0a66e18bc8d920554^ >fffffffff.diff " restores
>> expected functionality (Does this mean I get my name in the kernel
>> log? :D ).
>>
>> $ git bisect bad
>> 56d1de0a21db28e41741cfa0a66e18bc8d920554 is the first bad commit
>> commit 56d1de0a21db28e41741cfa0a66e18bc8d920554
>> Author: Bob Copeland <me@bobcopeland.com>
>> Date:   Mon Aug 24 23:00:30 2009 -0400
>>
>>    ath5k: clean up filter flags setting
>>
>>    The maze of if() statements in configure_filter is confusing.
>>    Reorganizing it as a switch statement makes it more apparent what
>>    is going on and reveals several suspicious settings.  This has no
>>    functional changes, though it does remove some redundant flags
>>    that are set earlier.
>>
>>    Also now that we can sleep, protect sc->filter_flags with the
>>    sc lock.
>>
>>    Signed-off-by: Bob Copeland <me@bobcopeland.com>
>>    Signed-off-by: John W. Linville <linville@tuxdriver.com>
>>
>> :040000 040000 1ca00241b99b379f192c1ade63c082955c7dda69
>> 1a19187e11210e3c66c224dba62a1896c46d0114 M      drivers
>>
>
> Certainly a possible cause. Chances are some of the "redundant" flags
> weren't so redundant at all...
>
> However, the original code was wrong too - I highly doubt we can rely
> on sc->opmode to set monitor mode filter flags. I will look into this
> when I get to my Linux box again.
>
> --
> Vista: [V]iruses, [I]ntruders, [S]pyware, [T]rojans and [A]dware. :-)
>

In the meantime, one thing to test: Add a printk of sc->opmode. If I
am right, this will never be set to NL80211_IFTYPE_MONITOR... That
would mean the promisc flag would never be set. Prior to the patch,
this was taken care by the following code:
       if (sc->opmode != NL80211_IFTYPE_AP &&
               sc->opmode != NL80211_IFTYPE_MESH_POINT &&
               test_bit(ATH_STAT_PROMISC, sc->status))
               rfilt |= AR5K_RX_FILTER_PROM;


-- 
Vista: [V]iruses, [I]ntruders, [S]pyware, [T]rojans and [A]dware. :-)