Wojciech Dubowik wrote:
> I have seen the wrong rssi in scan results but I just don't have yet
> have a proof on clean
> wireless testing. It seems that when I put a lot of printk I don't see
> the problem because
> maybe scanning/calibration/etc. is shifted more away. Anyway it's just a
> guess so I will dig
> further.
I think the initial skew in RSSI right after association happens because
NF calibration hasn't completed yet. After doing the first scan, we reset
to the home/operating channel before initiating the association procedure.
When we do this, the history is reset for the operating channel using
ath9k_init_nfcal_hist_buffer() and then after the chip/phy is reset, NF cal
is started - at the end of ath9k_hw_reset() using ath9k_hw_start_nfcal().
But, we don't wait for the NF-cal to complete and neither do we poll for
the calibration status. In this small window, any beacons that are received show
a skewed RSSI. We receive about two beacons in this period, after
which the RSSI becomes normal.
If we wait explicitly for NF-cal to complete, the issue is fixed and the
reported RSSI is normal. (See the attached hacky patch).
For example, the current behavior is:
[ 9071.755238] wlan0: Inserted STA 00:03:7f:42:08:56
[ 9071.755242] wlan0: send auth to 00:03:7f:42:08:56 (try 1/3)
[ 9071.756735] NF cal not completed
[ 9071.756743] NOISE/RSSI/SIGNAL for 00:03:7f:42:08:56: -92/68/-24
[ 9071.756829] wlan0: authenticated
[ 9071.756836] wlan0: moving STA 00:03:7f:42:08:56 to state 2
[ 9071.766123] wlan0: associate with 00:03:7f:42:08:56 (try 1/3)
[ 9071.767920] NF cal not completed
[ 9071.767925] NOISE/RSSI/SIGNAL for 00:03:7f:42:08:56: -92/68/-24
[ 9071.767960] wlan0: RX AssocResp from 00:03:7f:42:08:56 (capab=0x401 status=0 aid=1)
[ 9071.768399] wlan0: associated
...
[ 9071.777816] cfg80211: Regulatory domain changed to country: US
[ 9071.777819] cfg80211: DFS Master region: unset
[ 9071.777821] cfg80211: (start_freq - end_freq @ bandwidth), (max_antenna_gain, max_eirp)
[ 9071.777825] cfg80211: (2402000 KHz - 2472000 KHz @ 40000 KHz), (N/A, 3000 mBm)
[ 9071.777828] cfg80211: (5170000 KHz - 5250000 KHz @ 80000 KHz), (N/A, 1700 mBm)
[ 9071.777830] cfg80211: (5250000 KHz - 5330000 KHz @ 80000 KHz), (N/A, 2300 mBm)
[ 9071.777833] cfg80211: (5735000 KHz - 5835000 KHz @ 80000 KHz), (N/A, 3000 mBm)
[ 9071.777836] cfg80211: (57240000 KHz - 63720000 KHz @ 2160000 KHz), (N/A, 4000 mBm)
[ 9071.787601] NF cal completed
[ 9071.787608] NOISE/RSSI/SIGNAL for 00:03:7f:46:06:49: -92/60/-32
[ 9071.823292] NF cal completed
[ 9071.823330] NOISE/RSSI/SIGNAL for 00:03:7f:42:08:56: -92/59/-33
But, when we wait for NF-cal to complete after initiating it:
[ 9140.932811] wlan0: Inserted STA 00:03:7f:42:08:56
[ 9140.932814] wlan0: send auth to 00:03:7f:42:08:56 (try 1/3)
[ 9140.934368] NF cal completed
[ 9140.934403] NOISE/RSSI/SIGNAL for 00:03:7f:42:08:56: -94/60/-34
[ 9140.934453] wlan0: authenticated
[ 9140.934457] wlan0: moving STA 00:03:7f:42:08:56 to state 2
[ 9140.940706] wlan0: associate with 00:03:7f:42:08:56 (try 1/3)
[ 9140.942575] NF cal completed
[ 9140.942588] NOISE/RSSI/SIGNAL for 00:03:7f:42:08:56: -94/61/-33
[ 9140.942643] wlan0: RX AssocResp from 00:03:7f:42:08:56 (capab=0x401 status=0 aid=1)
[ 9140.942728] wlan0: associated
...
[ 9140.948155] cfg80211: Regulatory domain changed to country: US
[ 9140.948156] cfg80211: DFS Master region: unset
[ 9140.948158] cfg80211: (start_freq - end_freq @ bandwidth), (max_antenna_gain, max_eirp)
[ 9140.948160] cfg80211: (2402000 KHz - 2472000 KHz @ 40000 KHz), (N/A, 3000 mBm)
[ 9140.948161] cfg80211: (5170000 KHz - 5250000 KHz @ 80000 KHz), (N/A, 1700 mBm)
[ 9140.948162] cfg80211: (5250000 KHz - 5330000 KHz @ 80000 KHz), (N/A, 2300 mBm)
[ 9140.948164] cfg80211: (5735000 KHz - 5835000 KHz @ 80000 KHz), (N/A, 3000 mBm)
[ 9140.948165] cfg80211: (57240000 KHz - 63720000 KHz @ 2160000 KHz), (N/A, 4000 mBm)
[ 9140.962224] NF cal completed
[ 9140.962258] NOISE/RSSI/SIGNAL for 00:03:7f:46:06:49: -94/59/-35
[ 9141.010795] NF cal completed
[ 9141.010822] NOISE/RSSI/SIGNAL for 00:03:7f:42:08:56: -94/61/-33
We could fix this by checking if we are initializing the history for the operating
channel and waiting for NF-cal completion, but I am not sure if this needs
to be done. The window is really small and subsequent NF calibrations using
the ANI routine work properly.
The only problem that I can see if running a scan when associated distorts the
reported RSSI, but I can't see such an issue. If you can provide logs for this
case, it would be helpful.
Sujith
diff --git a/drivers/net/wireless/ath/ath9k/calib.c b/drivers/net/wireless/ath/ath9k/calib.c
index 278365b..84e93d5 100644
--- a/drivers/net/wireless/ath/ath9k/calib.c
+++ b/drivers/net/wireless/ath/ath9k/calib.c
@@ -218,6 +218,8 @@ EXPORT_SYMBOL(ath9k_hw_reset_calvalid);
void ath9k_hw_start_nfcal(struct ath_hw *ah, bool update)
{
+ int j;
+
if (ah->caldata)
set_bit(NFCAL_PENDING, &ah->caldata->cal_flags);
@@ -232,6 +234,13 @@ void ath9k_hw_start_nfcal(struct ath_hw *ah, bool update)
AR_PHY_AGC_CONTROL_NO_UPDATE_NF);
REG_SET_BIT(ah, AR_PHY_AGC_CONTROL, AR_PHY_AGC_CONTROL_NF);
+
+ for (j = 0; j < 10000; j++) {
+ if ((REG_READ(ah, AR_PHY_AGC_CONTROL) &
+ AR_PHY_AGC_CONTROL_NF) == 0)
+ break;
+ udelay(10);
+ }
}
void ath9k_hw_loadnf(struct ath_hw *ah, struct ath9k_channel *chan)
diff --git a/drivers/net/wireless/ath/ath9k/recv.c b/drivers/net/wireless/ath/ath9k/recv.c
index 5bf3243..e017b01 100644
--- a/drivers/net/wireless/ath/ath9k/recv.c
+++ b/drivers/net/wireless/ath/ath9k/recv.c
@@ -1088,6 +1088,14 @@ static int ath9k_rx_skb_preprocess(struct ath_softc *sc,
ath9k_process_rssi(common, hw, rx_stats, rx_status);
+ if ((REG_READ(ah, AR_PHY_AGC_CONTROL) & AR_PHY_AGC_CONTROL_NF) != 0)
+ printk("NF cal not completed\n");
+ else
+ printk("NF cal completed\n");
+
+ printk("NOISE/RSSI/SIGNAL for %pM: %d/%d/%d\n",
+ hdr->addr3, ah->noise, rx_stats->rs_rssi, rx_status->signal);
+
rx_status->band = ah->curchan->chan->band;
rx_status->freq = ah->curchan->chan->center_freq;
rx_status->antenna = rx_stats->rs_antenna;
