[RFC PATCH 2/2] ath10k: add testmode

Michal Kazior michal.kazior at tieto.com
Fri May 30 03:40:49 PDT 2014


On 30 May 2014 12:35, Kalle Valo <kvalo at qca.qualcomm.com> wrote:
> Michal Kazior <michal.kazior at tieto.com> writes:
>
>> On 30 May 2014 12:06, Kalle Valo <kvalo at qca.qualcomm.com> wrote:
>>> Michal Kazior <michal.kazior at tieto.com> writes:
>>>
>>>> On 29 May 2014 14:51, Kalle Valo <kvalo at qca.qualcomm.com> wrote:
>>>>
>>>>> BTW, please edit your quotes. It's pain to find your comments from a 700
>>>>> line monster :)
>>>
>>> This request still stands, even though 170 lines is much better than 700
>>> lines ;)
>>
>> I intended to trim it this time but I've hit 'send' before actually
>> doing that.. sorry!
>
> No problem :)
>
>>>> We could also depend on cmd_id to accept/reject skbuffs in
>>>> ath10k_tm_event_wmi() since, at least for now, there aren't any
>>>> special cases to handle. But then.. will we really ever need to have a
>>>> passthrough (i.e. pass buffers to userspace but still process them in
>>>> the driver afterwards)? Does that even make sense?
>>>
>>> I think that's getting a bit too complicated. Basically we just need
>>> ath10k act as a pipe for WMI packets.
>>
>> Hmm..
>>
>> With v1 code if you assume UTF firmware sends an SWBA event then
>> ath10k would send the testmode event AND try to actually handle it as
>> if it was regular operation.. but how is that supposed to work if you
>> have no mac80211 interface structures around at all? My point is
>> ath10k_tm_event_wmi() should _consume_ buffers and the caller should
>> not processes consumed buffers (assuming you want to keep
>> ath10k_tm_event_wmi call where it is now, i.e. before switch-case).
>
> I see your point now, thanks for banging it to my thick head with a
> sledge hammer :)
>
> I'll rethink this part and try to solve it for v2. The reason I would
> like to avoid your consuming idea is that at some point we might want to
> enable this WMI pipe also in "normal mode".

I'm not really sure what you mean by that. Can you give an example
where this pass-through behaviour would be desired, please?


Michał



More information about the ath10k mailing list