| Message ID | 1540991486-1948-1-git-send-email-sgruszka@redhat.com (mailing list archive) |
|---|---|
| State | Changes Requested |
| Delegated to: | Kalle Valo |
| Headers | show |
| Series | [v2] mt76x0: run calibration after scanning | expand |
On Wed, Oct 31, 2018 at 02:11:26PM +0100, Stanislaw Gruszka wrote: > If we are associated and scanning is performed, sw_scan_complete callback > is done after we get back to operating channel, so we do not perform > queue cal work. Fix this queue cal work from sw_scan_complete(). > > We have to restore gain in MT_BBP(AGC, 8) register after scanning, as > it was multiple times modified by channel switch code. So queue cal work > without any delay and reset dev->cal.low_gain to assure calibration > code will AGC gain value. > > Note patch was not tested on mt76x2, but should work and be needed > fix as well. > > Fixes: bbd10586f0df ("mt76x0: phy: do not run calibration during channel switch") > Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com> > --- > v1 -> v2: > only queue cal work with 0 delay an reset dev->cal.low_gain > > drivers/net/wireless/mediatek/mt76/mt76x02_util.c | 6 ++++++ > 1 file changed, 6 insertions(+) > > diff --git a/drivers/net/wireless/mediatek/mt76/mt76x02_util.c b/drivers/net/wireless/mediatek/mt76/mt76x02_util.c > index 9f083008dbd4..747fd2c9ec45 100644 > --- a/drivers/net/wireless/mediatek/mt76/mt76x02_util.c > +++ b/drivers/net/wireless/mediatek/mt76/mt76x02_util.c > @@ -553,6 +553,12 @@ void mt76x02_sw_scan_complete(struct ieee80211_hw *hw, > clear_bit(MT76_SCANNING, &dev->mt76.state); > if (mt76_is_mmio(dev)) > tasklet_enable(&dev->pre_tbtt_tasklet); > + > + if (vif->bss_conf.assoc) { > + /* Restore AGC gain and resume calibration after scanning. */ > + dev->cal.low_gain = -1; > + ieee80211_queue_delayed_work(hw, &dev->cal_work, 0); > + } > } > EXPORT_SYMBOL_GPL(mt76x02_sw_scan_complete); I think this can not be sufficient fix. Something like seems to be needed as wall to force AGC gain write to register: diff --git a/drivers/net/wireless/mediatek/mt76/mt76x0/phy.c b/drivers/net/wireless/mediatek/mt76/mt76x0/phy.c index 1af2a1227924..fc4085ef3e16 100644 --- a/drivers/net/wireless/mediatek/mt76/mt76x0/phy.c +++ b/drivers/net/wireless/mediatek/mt76/mt76x0/phy.c @@ -833,7 +833,7 @@ mt76x0_phy_update_channel_gain(struct mt76x02_dev *dev) low_gain = (dev->cal.avg_rssi_all > mt76x02_get_rssi_gain_thresh(dev)) + (dev->cal.avg_rssi_all > mt76x02_get_low_rssi_gain_thresh(dev)); - gain_change = (dev->cal.low_gain & 2) ^ (low_gain & 2); + gain_change = log_gain != dev->cal.low_gain; dev->cal.low_gain = low_gain; but I do not fully understand this xor logic, so I'm not sure. Thanks Stanislaw
On 2018-10-31 14:11, Stanislaw Gruszka wrote: > If we are associated and scanning is performed, sw_scan_complete callback > is done after we get back to operating channel, so we do not perform > queue cal work. Fix this queue cal work from sw_scan_complete(). > > We have to restore gain in MT_BBP(AGC, 8) register after scanning, as > it was multiple times modified by channel switch code. So queue cal work > without any delay and reset dev->cal.low_gain to assure calibration > code will AGC gain value. > > Note patch was not tested on mt76x2, but should work and be needed > fix as well. > > Fixes: bbd10586f0df ("mt76x0: phy: do not run calibration during channel switch") > Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com> > --- > v1 -> v2: > only queue cal work with 0 delay an reset dev->cal.low_gain > > drivers/net/wireless/mediatek/mt76/mt76x02_util.c | 6 ++++++ > 1 file changed, 6 insertions(+) > > diff --git a/drivers/net/wireless/mediatek/mt76/mt76x02_util.c b/drivers/net/wireless/mediatek/mt76/mt76x02_util.c > index 9f083008dbd4..747fd2c9ec45 100644 > --- a/drivers/net/wireless/mediatek/mt76/mt76x02_util.c > +++ b/drivers/net/wireless/mediatek/mt76/mt76x02_util.c > @@ -553,6 +553,12 @@ void mt76x02_sw_scan_complete(struct ieee80211_hw *hw, > clear_bit(MT76_SCANNING, &dev->mt76.state); > if (mt76_is_mmio(dev)) > tasklet_enable(&dev->pre_tbtt_tasklet); > + > + if (vif->bss_conf.assoc) { > + /* Restore AGC gain and resume calibration after scanning. */ > + dev->cal.low_gain = -1; > + ieee80211_queue_delayed_work(hw, &dev->cal_work, 0); > + } I think checking vif->bss_conf.assoc here is not enough, since we might be using multiple interfaces (e.g. AP+STA) - Felix
diff --git a/drivers/net/wireless/mediatek/mt76/mt76x02_util.c b/drivers/net/wireless/mediatek/mt76/mt76x02_util.c index 9f083008dbd4..747fd2c9ec45 100644 --- a/drivers/net/wireless/mediatek/mt76/mt76x02_util.c +++ b/drivers/net/wireless/mediatek/mt76/mt76x02_util.c @@ -553,6 +553,12 @@ void mt76x02_sw_scan_complete(struct ieee80211_hw *hw, clear_bit(MT76_SCANNING, &dev->mt76.state); if (mt76_is_mmio(dev)) tasklet_enable(&dev->pre_tbtt_tasklet); + + if (vif->bss_conf.assoc) { + /* Restore AGC gain and resume calibration after scanning. */ + dev->cal.low_gain = -1; + ieee80211_queue_delayed_work(hw, &dev->cal_work, 0); + } } EXPORT_SYMBOL_GPL(mt76x02_sw_scan_complete);
If we are associated and scanning is performed, sw_scan_complete callback is done after we get back to operating channel, so we do not perform queue cal work. Fix this queue cal work from sw_scan_complete(). We have to restore gain in MT_BBP(AGC, 8) register after scanning, as it was multiple times modified by channel switch code. So queue cal work without any delay and reset dev->cal.low_gain to assure calibration code will AGC gain value. Note patch was not tested on mt76x2, but should work and be needed fix as well. Fixes: bbd10586f0df ("mt76x0: phy: do not run calibration during channel switch") Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com> --- v1 -> v2: only queue cal work with 0 delay an reset dev->cal.low_gain drivers/net/wireless/mediatek/mt76/mt76x02_util.c | 6 ++++++ 1 file changed, 6 insertions(+)