2012-12-01 20:20:22

by Bernhard Urban

Subject: [PATCH] ath9k_htc: update RSSI values only when the device is associated

add an if-guard, otherwise iw(8) reports weird signal strengths.

The behaviour was fine before this commit:
7c277349ecbd66e19fad3d949fa6ef6c131a3b62
This patch is therefore a partially revert of it.

Tested with "TP-Link TL-WN722N"

Thanks to indoo.rs http://indoo.rs/ for sponsoring

Reported-by: Markus Krainz <[email protected]>
Tested-by: Markus Krainz <[email protected]>
Signed-off-by: Bernhard Urban <[email protected]>
---
drivers/net/wireless/ath/ath9k/htc.h | 1 +
drivers/net/wireless/ath/ath9k/htc_drv_txrx.c | 29 +++++++++++++++----------
2 files changed, 19 insertions(+), 11 deletions(-)

diff --git a/drivers/net/wireless/ath/ath9k/htc.h b/drivers/net/wireless/ath/ath9k/htc.h
index 936e920..00ebf1c 100644
--- a/drivers/net/wireless/ath/ath9k/htc.h
+++ b/drivers/net/wireless/ath/ath9k/htc.h
@@ -22,6 +22,7 @@
#include <linux/firmware.h>
#include <linux/skbuff.h>
#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
#include <linux/leds.h>
#include <linux/slab.h>
#include <net/mac80211.h>
diff --git a/drivers/net/wireless/ath/ath9k/htc_drv_txrx.c b/drivers/net/wireless/ath/ath9k/htc_drv_txrx.c
index 47e61d0..8b4da3d 100644
--- a/drivers/net/wireless/ath/ath9k/htc_drv_txrx.c
+++ b/drivers/net/wireless/ath/ath9k/htc_drv_txrx.c
@@ -969,6 +969,7 @@ static bool ath9k_rx_prepare(struct ath9k_htc_priv *priv,
int hdrlen, padpos, padsize;
int last_rssi = ATH_RSSI_DUMMY_MARKER;
__le16 fc;
+ u8 is_mybeacon;

if (skb->len < HTC_RX_FRAME_HEADER_SIZE) {
ath_err(common, "Corrupted RX frame, dropping (len: %d)\n",
@@ -1060,22 +1061,28 @@ static bool ath9k_rx_prepare(struct ath9k_htc_priv *priv,
ath9k_process_rate(hw, rx_status, rxbuf->rxstatus.rs_rate,
rxbuf->rxstatus.rs_flags);

- if (rxbuf->rxstatus.rs_rssi != ATH9K_RSSI_BAD &&
- !rxbuf->rxstatus.rs_moreaggr)
- ATH_RSSI_LPF(priv->rx.last_rssi,
- rxbuf->rxstatus.rs_rssi);
+ is_mybeacon = ieee80211_is_beacon(fc) &&
+ !is_zero_ether_addr(common->curbssid) &&
+ ether_addr_equal(hdr->addr3, common->curbssid);

- last_rssi = priv->rx.last_rssi;
+ if (is_mybeacon && priv->ah->opmode == NL80211_IFTYPE_STATION) {
+ if (rxbuf->rxstatus.rs_rssi != ATH9K_RSSI_BAD &&
+ !rxbuf->rxstatus.rs_moreaggr)
+ ATH_RSSI_LPF(priv->rx.last_rssi,
+ rxbuf->rxstatus.rs_rssi);

- if (likely(last_rssi != ATH_RSSI_DUMMY_MARKER))
- rxbuf->rxstatus.rs_rssi = ATH_EP_RND(last_rssi,
- ATH_RSSI_EP_MULTIPLIER);
+ last_rssi = priv->rx.last_rssi;

- if (rxbuf->rxstatus.rs_rssi < 0)
- rxbuf->rxstatus.rs_rssi = 0;
+ if (likely(last_rssi != ATH_RSSI_DUMMY_MARKER)) {
+ s8 rssi = ATH_EP_RND(last_rssi, ATH_RSSI_EP_MULTIPLIER);
+ rxbuf->rxstatus.rs_rssi = rssi;
+ }
+
+ if (rxbuf->rxstatus.rs_rssi < 0)
+ rxbuf->rxstatus.rs_rssi = 0;

- if (ieee80211_is_beacon(fc))
priv->ah->stats.avgbrssi = rxbuf->rxstatus.rs_rssi;
+ }

rx_status->mactime = be64_to_cpu(rxbuf->rxstatus.rs_tstamp);
rx_status->band = hw->conf.channel->band;
--
1.7.9.5

2013-02-05 22:05:39

by Felix Fietkau

Subject: Re: [PATCH] ath9k_htc: update RSSI values only when the device is associated

On 2013-02-01 1:50 PM, Bernhard Urban wrote:
> add an if-guard, otherwise iw(8) reports weird signal strengths.
>
> The behaviour was fine before this commit:
> 7c277349ecbd66e19fad3d949fa6ef6c131a3b62
> Therefore, this patch is a partially revert of it.
I think your commit message is a bit misleading. The main problem with
signal strength reporting is not that it's missing some checks for if
the device is associated, or if the received frame is a local beacon.

The main problem is in these lines below:
> + if (likely(last_rssi != ATH_RSSI_DUMMY_MARKER)) {
> + s8 rssi = ATH_EP_RND(last_rssi, ATH_RSSI_EP_MULTIPLIER);
> + rxbuf->rxstatus.rs_rssi = rssi;
> + }
It does not make any sense to update the per-frame RSSI value with any
sort of average. ath9k filters out invalid RSSI values by ignoring rssi
for ANI when rx_stats->rs_moreaggr is set. In that case it also sets the
RX_FLAG_NO_SIGNAL_VAL flag to tell mac80211 to ignore the value for its
own signal strength averages.

- Felix