This is required to handle sta.wme correctly
after update rates (eg. when using HT/VHT) when
driver using IEEE80211_HW_SUPPORT_FAST_XMIT.
In other case fast tx path will not add QOS data and
in case of ath10k driver (rate control in the HW) only
legacy rates will be used.
Signed-off-by: Janusz Dziedzic <[email protected]>
---
net/mac80211/ibss.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/net/mac80211/ibss.c b/net/mac80211/ibss.c
index bfef1b2..f76414e 100644
--- a/net/mac80211/ibss.c
+++ b/net/mac80211/ibss.c
@@ -1087,6 +1087,7 @@ static void ieee80211_update_sta_info(struct ieee80211_sub_if_data *sdata,
if (sta->sta.rx_nss != rx_nss)
changed |= IEEE80211_RC_NSS_CHANGED;
+ ieee80211_clear_fast_xmit(sta);
drv_sta_rc_update(local, sdata, &sta->sta, changed);
}
--
1.9.1
On Mon, 2015-05-11 at 11:52 +0200, Johannes Berg wrote:
> On Mon, 2015-05-11 at 11:39 +0200, Janusz Dziedzic wrote:
> > This is required to handle sta.wme correctly
> > after update rates (eg. when using HT/VHT) when
> > driver using IEEE80211_HW_SUPPORT_FAST_XMIT.
>
> Wouldn't you want to check, so it can be re-assigned?
i.e. something like
diff --git a/net/mac80211/ibss.c b/net/mac80211/ibss.c
index bfef1b215050..7e62183525aa 100644
--- a/net/mac80211/ibss.c
+++ b/net/mac80211/ibss.c
@@ -1031,8 +1031,11 @@ static void ieee80211_update_sta_info(struct ieee80211_sub_if_data *sdata,
}
}
- if (sta && elems->wmm_info && local->hw.queues >= IEEE80211_NUM_ACS)
+ if (sta && !sta->sta.wme && elems->wmm_info &&
+ local->hw.queues >= IEEE80211_NUM_ACS) {
sta->sta.wme = true;
+ ieee80211_check_fast_xmit(sta);
+ }
if (sta && elems->ht_operation && elems->ht_cap_elem &&
sdata->u.ibss.chandef.width != NL80211_CHAN_WIDTH_20_NOHT &&
johannes
On Mon, 2015-05-11 at 11:39 +0200, Janusz Dziedzic wrote:
> This is required to handle sta.wme correctly
> after update rates (eg. when using HT/VHT) when
> driver using IEEE80211_HW_SUPPORT_FAST_XMIT.
Wouldn't you want to check, so it can be re-assigned?
johannes
On 11 May 2015 at 14:01, Johannes Berg <[email protected]> wrote:
> On Mon, 2015-05-11 at 13:52 +0200, Janusz Dziedzic wrote:
>> On 11 May 2015 at 11:53, Johannes Berg <[email protected]> wrote:
>> > On Mon, 2015-05-11 at 11:52 +0200, Johannes Berg wrote:
>> >> On Mon, 2015-05-11 at 11:39 +0200, Janusz Dziedzic wrote:
>> >> > This is required to handle sta.wme correctly
>> >> > after update rates (eg. when using HT/VHT) when
>> >> > driver using IEEE80211_HW_SUPPORT_FAST_XMIT.
>> >>
>> >> Wouldn't you want to check, so it can be re-assigned?
>> >
>> > i.e. something like
>> >
>> > diff --git a/net/mac80211/ibss.c b/net/mac80211/ibss.c
>> > index bfef1b215050..7e62183525aa 100644
>> > --- a/net/mac80211/ibss.c
>> > +++ b/net/mac80211/ibss.c
>> > @@ -1031,8 +1031,11 @@ static void ieee80211_update_sta_info(struct ieee80211_sub_if_data *sdata,
>> > }
>> > }
>> >
>> > - if (sta && elems->wmm_info && local->hw.queues >= IEEE80211_NUM_ACS)
>> > + if (sta && !sta->sta.wme && elems->wmm_info &&
>> > + local->hw.queues >= IEEE80211_NUM_ACS) {
>> > sta->sta.wme = true;
>> > + ieee80211_check_fast_xmit(sta);
>> > + }
>> >
>> > if (sta && elems->ht_operation && elems->ht_cap_elem &&
>> > sdata->u.ibss.chandef.width != NL80211_CHAN_WIDTH_20_NOHT &&
>> >
>> Your version also works fine.
>> Should I send new patch or you already have this in your tree?
>
> I didn't put it in, but can do that.
>
Yes, please do that.
BR
Janusz
On Mon, 2015-05-11 at 13:52 +0200, Janusz Dziedzic wrote:
> On 11 May 2015 at 11:53, Johannes Berg <[email protected]> wrote:
> > On Mon, 2015-05-11 at 11:52 +0200, Johannes Berg wrote:
> >> On Mon, 2015-05-11 at 11:39 +0200, Janusz Dziedzic wrote:
> >> > This is required to handle sta.wme correctly
> >> > after update rates (eg. when using HT/VHT) when
> >> > driver using IEEE80211_HW_SUPPORT_FAST_XMIT.
> >>
> >> Wouldn't you want to check, so it can be re-assigned?
> >
> > i.e. something like
> >
> > diff --git a/net/mac80211/ibss.c b/net/mac80211/ibss.c
> > index bfef1b215050..7e62183525aa 100644
> > --- a/net/mac80211/ibss.c
> > +++ b/net/mac80211/ibss.c
> > @@ -1031,8 +1031,11 @@ static void ieee80211_update_sta_info(struct ieee80211_sub_if_data *sdata,
> > }
> > }
> >
> > - if (sta && elems->wmm_info && local->hw.queues >= IEEE80211_NUM_ACS)
> > + if (sta && !sta->sta.wme && elems->wmm_info &&
> > + local->hw.queues >= IEEE80211_NUM_ACS) {
> > sta->sta.wme = true;
> > + ieee80211_check_fast_xmit(sta);
> > + }
> >
> > if (sta && elems->ht_operation && elems->ht_cap_elem &&
> > sdata->u.ibss.chandef.width != NL80211_CHAN_WIDTH_20_NOHT &&
> >
> Your version also works fine.
> Should I send new patch or you already have this in your tree?
I didn't put it in, but can do that.
johannes
On 11 May 2015 at 11:53, Johannes Berg <[email protected]> wrote:
> On Mon, 2015-05-11 at 11:52 +0200, Johannes Berg wrote:
>> On Mon, 2015-05-11 at 11:39 +0200, Janusz Dziedzic wrote:
>> > This is required to handle sta.wme correctly
>> > after update rates (eg. when using HT/VHT) when
>> > driver using IEEE80211_HW_SUPPORT_FAST_XMIT.
>>
>> Wouldn't you want to check, so it can be re-assigned?
>
> i.e. something like
>
> diff --git a/net/mac80211/ibss.c b/net/mac80211/ibss.c
> index bfef1b215050..7e62183525aa 100644
> --- a/net/mac80211/ibss.c
> +++ b/net/mac80211/ibss.c
> @@ -1031,8 +1031,11 @@ static void ieee80211_update_sta_info(struct ieee80211_sub_if_data *sdata,
> }
> }
>
> - if (sta && elems->wmm_info && local->hw.queues >= IEEE80211_NUM_ACS)
> + if (sta && !sta->sta.wme && elems->wmm_info &&
> + local->hw.queues >= IEEE80211_NUM_ACS) {
> sta->sta.wme = true;
> + ieee80211_check_fast_xmit(sta);
> + }
>
> if (sta && elems->ht_operation && elems->ht_cap_elem &&
> sdata->u.ibss.chandef.width != NL80211_CHAN_WIDTH_20_NOHT &&
>
Your version also works fine.
Should I send new patch or you already have this in your tree?
BR
Janusz