Re: [PATCH v2] mac80211: VHT peer STA caps

[Mahesh Palivela <maheshp@posedge.com>]

On 09/25/2012 12:55 PM, Johannes Berg wrote:
>>
>>           vht_cap->cap = le16_to_cpu(vht_cap_ie->vht_cap_info) &
>>                   (sband->ht_cap.cap |
>
> err, vht_cap

Sorry, I meant vht_cap only. Typo..

>
>>                    ~(IEEE80211_VHT_CAP_RXLDPC |
>>                      IEEE80211_VHT_CAP_SU_BEAMFORMER_CAPABLE |
>>                      IEEE80211_VHT_CAP_MU_BEAMFORMER_CAPABLE |
>>                      IEEE80211_VHT_CAP_HTC_VHT));
>
> I don't think that can be right. It seems to me it should be something
> like
>
> cap = cap_info & sband->vht_cap.cap &
> 	(MAX_MPDU_LENGTH_... |
> 	 SHORT_GI_... | HTC_VHT | ... ? )
>
> if (cap_info & SU_BEAMFORMEE_CAPABLE &&
>      sband->vht_cap.cap & SU_BEAMFORMER_CAPABLE)
> 	cap |= SU_BEAMFORMEE_CAPABLE;
>
> etc.?
>
> OTOH, this would only matter to drivers that actually support all these
> features, and those drivers could take care of not enabling the features
> if their hardware doesn't support them. However, think of a driver that
> is for different hardware, some that supports beamforming and some that
> doesn't. Then if we mask out the beamforming capability of a station if
> it's not locally supported, that driver could be simpler? Then again, if
> we don't the driver just has to check if the hardware supports it, which
> seems reasonable as well.
>
> Do you understand the issue?
>

why should we mask out remote STA caps bits. Our local hw caps are in 
sband->sta.vht_caps. remote caps sta_info->sta.vht_caps. In what way 
drivers are connected with remote STA vht caps bits.

> So maybe if you document it right, your original v2 patch is sufficient.
> Convince me :)
>
> johannes
>