[v2,2/2] ath10k: correct legacy rate in tx stats
diff mbox series

Message ID 0101016e82883ded-63f88383-cd90-4cb0-b9bb-3dd6a1e9f4e9-000000@us-west-2.amazonses.com
State Changes Requested
Delegated to: Kalle Valo
Headers show
Series
  • ath10k: correct tx bitrate for QCA6174/QCA9377/QCA9888
Related show

Commit Message

Yu Wang Nov. 19, 2019, 7:19 a.m. UTC
When working in station mode, after connected to a legacy
AP, 11g only, for example, the tx bitrate is incorrect in
output of command 'iw wlan0 link'.

That's because the legacy tx bitrate value reported by
firmware is not well handled:
For QCA6174, the value represents rate index, but treated
as a real rate;
For QCA9888, the value is real rate, with unit 'Mbps', but
treated as '100kbps'.

To fix this issue:
1. Translate the rate index to real rate for QCA6174;
2. Translate the rate from 'Mbps' to 'kbps' for QCA9888.

Tested with:
QCA6174 PCIe with firmware WLAN.RM.4.4.1.c3-00031.
QCA6174 SDIO with firmware WLAN.RMH.4.4.1-00029.
QCA9888 PCIe with firmware 10.4-3.9.0.2-00040.

Signed-off-by: Yu Wang <yyuwang@codeaurora.org>
---
 drivers/net/wireless/ath/ath10k/htt_rx.c | 83 +++++++++++++++++++++++++++++++-
 drivers/net/wireless/ath/ath10k/mac.c    |  2 +-
 drivers/net/wireless/ath/ath10k/mac.h    |  1 +
 3 files changed, 83 insertions(+), 3 deletions(-)

Comments

Kalle Valo April 9, 2020, 2:21 p.m. UTC | #1
Yu Wang <yyuwang@codeaurora.org> wrote:

> When working in station mode, after connected to a legacy
> AP, 11g only, for example, the tx bitrate is incorrect in
> output of command 'iw wlan0 link'.
> 
> That's because the legacy tx bitrate value reported by
> firmware is not well handled:
> For QCA6174, the value represents rate index, but treated
> as a real rate;
> For QCA9888, the value is real rate, with unit 'Mbps', but
> treated as '100kbps'.
> 
> To fix this issue:
> 1. Translate the rate index to real rate for QCA6174;
> 2. Translate the rate from 'Mbps' to 'kbps' for QCA9888.
> 
> Tested with:
> QCA6174 PCIe with firmware WLAN.RM.4.4.1.c3-00031.
> QCA6174 SDIO with firmware WLAN.RMH.4.4.1-00029.
> QCA9888 PCIe with firmware 10.4-3.9.0.2-00040.
> 
> Signed-off-by: Yu Wang <yyuwang@codeaurora.org>
> Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

My comments don't seem to go to patchwork, so trying again:

What about QCA988X and WCN3990, how do they behave? Does this patch
break those?

> +	cck = (preamble == WMI_RATE_PREAMBLE_CCK);
> +	hw_rate = ATH10K_HW_LEGACY_RATE(*ratecode);
> +	for (i = 0; i < sband->n_bitrates; i++) {
> +		bitrates = &sband->bitrates[i];
> +		if (ath10k_mac_bitrate_is_cck(bitrates->bitrate) != cck)
> +			continue;
> +
> +		if (bitrates->hw_value == hw_rate ||
> +		    (bitrates->flags & IEEE80211_RATE_SHORT_PREAMBLE &&
> +		     bitrates->hw_value_short == hw_rate)) {
> +			bitrate = bitrates->bitrate;
> +
> +			/* The bitrate will be recovered in
> +			 * ath10k_update_per_peer_tx_stats().
> +			 */
> +			if (bitrate == 55)
> +				bitrate = 60;
> +
> +			bitrate = bitrate / 10;

Here you use magic value 60 but in ath10k_update_per_peer_tx_stats() you
use magic value 50:

> +		/* from 1Mbps to 100Kbps */
> +		rate = rate * 10;
> +		if (rate == 50)
> +			rate = 55;

Am I missing something or how is this supposed to work?
Yu Wang April 14, 2020, 5:32 a.m. UTC | #2
> -----Original Message-----
> From: ath10k <ath10k-bounces@lists.infradead.org> On Behalf Of Kalle Valo
> Sent: Thursday, April 9, 2020 10:22 PM
> To: Yu Wang <yyuwang@codeaurora.org>
> Cc: linux-wireless@vger.kernel.org; ath10k@lists.infradead.org
> Subject: [EXT] Re: [PATCH v2 2/2] ath10k: correct legacy rate in tx stats
>
> Yu Wang <yyuwang@codeaurora.org> wrote:
>
> > When working in station mode, after connected to a legacy AP, 11g
> > only, for example, the tx bitrate is incorrect in output of command
> > 'iw wlan0 link'.
> >
> > That's because the legacy tx bitrate value reported by firmware is not
> > well handled:
> > For QCA6174, the value represents rate index, but treated as a real
> > rate; For QCA9888, the value is real rate, with unit 'Mbps', but
> > treated as '100kbps'.
> >
> > To fix this issue:
> > 1. Translate the rate index to real rate for QCA6174; 2. Translate the
> > rate from 'Mbps' to 'kbps' for QCA9888.
> >
> > Tested with:
> > QCA6174 PCIe with firmware WLAN.RM.4.4.1.c3-00031.
> > QCA6174 SDIO with firmware WLAN.RMH.4.4.1-00029.
> > QCA9888 PCIe with firmware 10.4-3.9.0.2-00040.
> >
> > Signed-off-by: Yu Wang <yyuwang@codeaurora.org>
> > Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
>
> My comments don't seem to go to patchwork, so trying again:
>
> What about QCA988X and WCN3990, how do they behave? Does this patch
> break those?
Since HTT_T2H_MSG_TYPE_PPDU_STATS_IND is a newly added message, suppose it's won't break the other functions.
I don’t have environment to verify the patch with QCA988X and WCN3990.
Can anyone help on this?

>
> > +cck = (preamble == WMI_RATE_PREAMBLE_CCK);
> > +hw_rate = ATH10K_HW_LEGACY_RATE(*ratecode);
> > +for (i = 0; i < sband->n_bitrates; i++) {
> > +bitrates = &sband->bitrates[i];
> > +if (ath10k_mac_bitrate_is_cck(bitrates->bitrate) != cck)
> > +continue;
> > +
> > +if (bitrates->hw_value == hw_rate ||
> > +    (bitrates->flags & IEEE80211_RATE_SHORT_PREAMBLE &&
> > +     bitrates->hw_value_short == hw_rate)) {
> > +bitrate = bitrates->bitrate;
> > +
> > +/* The bitrate will be recovered in
> > + * ath10k_update_per_peer_tx_stats().
> > + */
> > +if (bitrate == 55)
> > +bitrate = 60;
> > +
> > +bitrate = bitrate / 10;
>
> Here you use magic value 60 but in ath10k_update_per_peer_tx_stats() you use
> magic value 50:
>
> > +/* from 1Mbps to 100Kbps */
> > +rate = rate * 10;
> > +if (rate == 50)
> > +rate = 55;
>
> Am I missing something or how is this supposed to work?
In existing code, ath10k_update_per_peer_tx_stats() will check the bitrate and convert 6_CCK to 5(in the comment: FW sends CCK rate 5.5Mbps as 6), and then 5 will be recovered to 55.
That's why we need to convert bitrate 55 to 6 when processing PPDU_STATS.

if (txrate.flags == WMI_RATE_PREAMBLE_CCK ||
txrate.flags == WMI_RATE_PREAMBLE_OFDM) {
rate = ATH10K_HW_LEGACY_RATE(peer_stats->ratecode);
/* This is hacky, FW sends CCK rate 5.5Mbps as 6 */
if (rate == 6 && txrate.flags == WMI_RATE_PREAMBLE_CCK)
rate = 5;
rate_idx = ath10k_get_legacy_rate_idx(ar, rate);
if (rate_idx < 0)
return;

/* from 1Mbps to 100Kbps */
rate = rate * 10;
if (rate == 50)
rate = 55;
>
> --
> https://patchwork.kernel.org/patch/11251001/
>
> https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatch
> es
>
> _______________________________________________
> ath10k mailing list
> ath10k@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/ath10k
Tom Psyborg April 14, 2020, 8:53 a.m. UTC | #3
On 14/04/2020, Yu Wang <yyuwang@qti.qualcomm.com> wrote:
>
>
>> -----Original Message-----
>> From: ath10k <ath10k-bounces@lists.infradead.org> On Behalf Of Kalle Valo
>> Sent: Thursday, April 9, 2020 10:22 PM
>> To: Yu Wang <yyuwang@codeaurora.org>
>> Cc: linux-wireless@vger.kernel.org; ath10k@lists.infradead.org
>> Subject: [EXT] Re: [PATCH v2 2/2] ath10k: correct legacy rate in tx stats
>>
>> Yu Wang <yyuwang@codeaurora.org> wrote:
>>
>> > When working in station mode, after connected to a legacy AP, 11g
>> > only, for example, the tx bitrate is incorrect in output of command
>> > 'iw wlan0 link'.
>> >
>> > That's because the legacy tx bitrate value reported by firmware is not
>> > well handled:
>> > For QCA6174, the value represents rate index, but treated as a real
>> > rate; For QCA9888, the value is real rate, with unit 'Mbps', but
>> > treated as '100kbps'.
>> >
>> > To fix this issue:
>> > 1. Translate the rate index to real rate for QCA6174; 2. Translate the
>> > rate from 'Mbps' to 'kbps' for QCA9888.
>> >
>> > Tested with:
>> > QCA6174 PCIe with firmware WLAN.RM.4.4.1.c3-00031.
>> > QCA6174 SDIO with firmware WLAN.RMH.4.4.1-00029.
>> > QCA9888 PCIe with firmware 10.4-3.9.0.2-00040.
>> >
>> > Signed-off-by: Yu Wang <yyuwang@codeaurora.org>
>> > Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
>>
>> My comments don't seem to go to patchwork, so trying again:
>>
>> What about QCA988X and WCN3990, how do they behave? Does this patch
>> break those?
> Since HTT_T2H_MSG_TYPE_PPDU_STATS_IND is a newly added message, suppose it's
> won't break the other functions.
> I don’t have environment to verify the patch with QCA988X and WCN3990.
> Can anyone help on this?
>
>>
>> > +cck = (preamble == WMI_RATE_PREAMBLE_CCK);
>> > +hw_rate = ATH10K_HW_LEGACY_RATE(*ratecode);
>> > +for (i = 0; i < sband->n_bitrates; i++) {
>> > +bitrates = &sband->bitrates[i];
>> > +if (ath10k_mac_bitrate_is_cck(bitrates->bitrate) != cck)
>> > +continue;
>> > +
>> > +if (bitrates->hw_value == hw_rate ||
>> > +    (bitrates->flags & IEEE80211_RATE_SHORT_PREAMBLE &&
>> > +     bitrates->hw_value_short == hw_rate)) {
>> > +bitrate = bitrates->bitrate;
>> > +
>> > +/* The bitrate will be recovered in
>> > + * ath10k_update_per_peer_tx_stats().
>> > + */
>> > +if (bitrate == 55)
>> > +bitrate = 60;
>> > +
>> > +bitrate = bitrate / 10;
>>
>> Here you use magic value 60 but in ath10k_update_per_peer_tx_stats() you
>> use
>> magic value 50:
>>
>> > +/* from 1Mbps to 100Kbps */
>> > +rate = rate * 10;
>> > +if (rate == 50)
>> > +rate = 55;
>>
>> Am I missing something or how is this supposed to work?
> In existing code, ath10k_update_per_peer_tx_stats() will check the bitrate
> and convert 6_CCK to 5(in the comment: FW sends CCK rate 5.5Mbps as 6), and
> then 5 will be recovered to 55.
> That's why we need to convert bitrate 55 to 6 when processing PPDU_STATS.
>
> if (txrate.flags == WMI_RATE_PREAMBLE_CCK ||
> txrate.flags == WMI_RATE_PREAMBLE_OFDM) {
> rate = ATH10K_HW_LEGACY_RATE(peer_stats->ratecode);
> /* This is hacky, FW sends CCK rate 5.5Mbps as 6 */
> if (rate == 6 && txrate.flags == WMI_RATE_PREAMBLE_CCK)
> rate = 5;
> rate_idx = ath10k_get_legacy_rate_idx(ar, rate);
> if (rate_idx < 0)
> return;
>
> /* from 1Mbps to 100Kbps */
> rate = rate * 10;
> if (rate == 50)
> rate = 55;
>>
>> --
>> https://patchwork.kernel.org/patch/11251001/
>>
>> https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatch
>> es
>>
>> _______________________________________________
>> ath10k mailing list
>> ath10k@lists.infradead.org
>> http://lists.infradead.org/mailman/listinfo/ath10k
> _______________________________________________
> ath10k mailing list
> ath10k@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/ath10k
>

Hi

I don't understand why are you fixing this in driver when the comment
clearly says "FW sends CCK rate 5.5Mbps as 6"
Shouldn't a proper way be to fix this in firmware?

Patch
diff mbox series

diff --git a/drivers/net/wireless/ath/ath10k/htt_rx.c b/drivers/net/wireless/ath/ath10k/htt_rx.c
index cde2d9b..796c8c0 100644
--- a/drivers/net/wireless/ath/ath10k/htt_rx.c
+++ b/drivers/net/wireless/ath/ath10k/htt_rx.c
@@ -3455,7 +3455,7 @@  ath10k_update_per_peer_tx_stats(struct ath10k *ar,
 	struct ath10k_sta *arsta = (struct ath10k_sta *)sta->drv_priv;
 	struct ieee80211_chanctx_conf *conf = NULL;
 	u8 rate = 0, sgi;
-	s8 rate_idx = 0;
+	s8 rate_idx = -1;
 	bool skip_auto_rate;
 	struct rate_info txrate;
 
@@ -3497,6 +3497,12 @@  ath10k_update_per_peer_tx_stats(struct ath10k *ar,
 		rate_idx = ath10k_get_legacy_rate_idx(ar, rate);
 		if (rate_idx < 0)
 			return;
+
+		/* from 1Mbps to 100Kbps */
+		rate = rate * 10;
+		if (rate == 50)
+			rate = 55;
+
 		arsta->txrate.legacy = rate;
 	} else if (txrate.flags == WMI_RATE_PREAMBLE_HT) {
 		arsta->txrate.flags = RATE_INFO_FLAGS_MCS;
@@ -3723,6 +3729,74 @@  static void ath10k_htt_rx_sec_ind_handler(struct ath10k *ar,
 	spin_unlock_bh(&ar->data_lock);
 }
 
+static int
+ath10k_htt_update_ratecode(struct ath10k *ar, struct ath10k_sta *arsta,
+			   u8 *ratecode)
+{
+	u8 hw_rate, preamble;
+	u16 bitrate;
+	int i;
+	const struct ieee80211_rate *bitrates;
+	bool cck;
+	struct ieee80211_chanctx_conf *conf = NULL;
+	enum nl80211_band band;
+	struct ieee80211_supported_band *sband;
+
+	if (!ratecode)
+		return -EINVAL;
+
+	/* only for legacy ratecode */
+	preamble = ATH10K_HW_PREAMBLE(*ratecode);
+	if (preamble != WMI_RATE_PREAMBLE_CCK &&
+	    preamble != WMI_RATE_PREAMBLE_OFDM)
+		return 0;
+
+	if (!arsta->arvif || !arsta->arvif->vif)
+		return -EINVAL;
+
+	WARN_ON(!rcu_read_lock_held());
+	conf = rcu_dereference(arsta->arvif->vif->chanctx_conf);
+	if (!conf)
+		return -EINVAL;
+
+	band = conf->def.chan->band;
+	sband = &ar->mac.sbands[band];
+	if (!sband->bitrates)
+		return -EINVAL;
+
+	if (WARN_ON_ONCE(sband->n_bitrates > S8_MAX))
+		return -EINVAL;
+
+	cck = (preamble == WMI_RATE_PREAMBLE_CCK);
+	hw_rate = ATH10K_HW_LEGACY_RATE(*ratecode);
+	for (i = 0; i < sband->n_bitrates; i++) {
+		bitrates = &sband->bitrates[i];
+		if (ath10k_mac_bitrate_is_cck(bitrates->bitrate) != cck)
+			continue;
+
+		if (bitrates->hw_value == hw_rate ||
+		    (bitrates->flags & IEEE80211_RATE_SHORT_PREAMBLE &&
+		     bitrates->hw_value_short == hw_rate)) {
+			bitrate = bitrates->bitrate;
+
+			/* The bitrate will be recovered in
+			 * ath10k_update_per_peer_tx_stats().
+			 */
+			if (bitrate == 55)
+				bitrate = 60;
+
+			bitrate = bitrate / 10;
+
+			/* replace hw_value with bitrate in ratecode */
+			*ratecode = ATH10K_HW_RATECODE(bitrate, 0, preamble);
+			return 0;
+		}
+	}
+
+	ath10k_warn(ar, "Invalid legacy ratecode %hhd ppdu stats", *ratecode);
+	return -EINVAL;
+}
+
 static void
 ath10k_htt_process_ppdu_stats(struct ath10k *ar, struct sk_buff *skb)
 {
@@ -3734,6 +3808,7 @@  ath10k_htt_process_ppdu_stats(struct ath10k *ar, struct sk_buff *skb)
 	struct ath10k_peer *peer;
 	u32 peer_id, i;
 	u8 num_ppdu;
+	u8 ratecode;
 
 	num_ppdu = resp->ppdu_stats.num_ppdu;
 	tx_stats = &resp->ppdu_stats.tx_ppdu_stats[0];
@@ -3753,6 +3828,10 @@  ath10k_htt_process_ppdu_stats(struct ath10k *ar, struct sk_buff *skb)
 	for (i = 0; i < num_ppdu; i++) {
 		tx_stats = &resp->ppdu_stats.tx_ppdu_stats[i];
 		arsta = (struct ath10k_sta *)sta->drv_priv;
+		ratecode = tx_stats->tx_ratecode;
+
+		if (ath10k_htt_update_ratecode(ar, arsta, &ratecode))
+			goto err;
 
 		p_tx_stats->succ_bytes =
 			__le32_to_cpu(tx_stats->tx_success_bytes);
@@ -3760,7 +3839,7 @@  ath10k_htt_process_ppdu_stats(struct ath10k *ar, struct sk_buff *skb)
 			__le32_to_cpu(tx_stats->tx_retry_bytes);
 		p_tx_stats->failed_bytes =
 			__le32_to_cpu(tx_stats->tx_failed_bytes);
-		p_tx_stats->ratecode = tx_stats->tx_ratecode;
+		p_tx_stats->ratecode = ratecode;
 		p_tx_stats->flags = tx_stats->flags;
 		p_tx_stats->succ_pkts =
 			__le16_to_cpu(tx_stats->tx_success_msdus);
diff --git a/drivers/net/wireless/ath/ath10k/mac.c b/drivers/net/wireless/ath/ath10k/mac.c
index 83cc877..95a5069 100644
--- a/drivers/net/wireless/ath/ath10k/mac.c
+++ b/drivers/net/wireless/ath/ath10k/mac.c
@@ -94,7 +94,7 @@  static struct ieee80211_rate ath10k_rates_rev2[] = {
 
 #define ath10k_wmi_legacy_rates ath10k_rates
 
-static bool ath10k_mac_bitrate_is_cck(int bitrate)
+bool ath10k_mac_bitrate_is_cck(int bitrate)
 {
 	switch (bitrate) {
 	case 10:
diff --git a/drivers/net/wireless/ath/ath10k/mac.h b/drivers/net/wireless/ath/ath10k/mac.h
index 98d83a2..cc245b1 100644
--- a/drivers/net/wireless/ath/ath10k/mac.h
+++ b/drivers/net/wireless/ath/ath10k/mac.h
@@ -28,6 +28,7 @@  struct rfc1042_hdr {
 	__be16 snap_type;
 } __packed;
 
+bool ath10k_mac_bitrate_is_cck(int bitrate);
 struct ath10k *ath10k_mac_create(size_t priv_size);
 void ath10k_mac_destroy(struct ath10k *ar);
 int ath10k_mac_register(struct ath10k *ar);