[2/2] ath11k: use cumulative survey statistics
diff mbox series

Message ID 20200504154122.91862-2-markus.theil@tu-ilmenau.de
State New
Headers show
Series
  • [1/2] ath10k: use cumulative survey statistics
Related show

Commit Message

Markus Theil May 4, 2020, 3:41 p.m. UTC
ath11k currently reports survey results for the last interval between each
invocation of NL80211_CMD_GET_SURVEY. For concurrent invocations, this
can lead to unexpectedly small results, e.g. when hostapd uses survey
data and iw survey dump is invoked in parallel. Fix this by returning
cumulative results, that don't depend on the last invocation. Other
drivers, e.g. ath9k or mt76 also use this behavior.

Signed-off-by: Markus Theil <markus.theil@tu-ilmenau.de>
---
 drivers/net/wireless/ath/ath11k/wmi.c | 20 ++++++++++----------
 1 file changed, 10 insertions(+), 10 deletions(-)

Comments

Sven Eckelmann May 4, 2020, 4:36 p.m. UTC | #1
On Monday, 4 May 2020 17:41:22 CEST Markus Theil wrote:
[...]
> diff --git a/drivers/net/wireless/ath/ath11k/wmi.c b/drivers/net/wireless/ath/ath11k/wmi.c
> index c2a972377687..322ddfda5bfd 100644
> --- a/drivers/net/wireless/ath/ath11k/wmi.c
> +++ b/drivers/net/wireless/ath/ath11k/wmi.c
> @@ -5610,16 +5610,16 @@ ath11k_pdev_bss_chan_info_event(struct ath11k_base *ab, struct sk_buff *skb)
[...]

What about the survey modifying code in ath11k_chan_info_event?

Kind regards,
	Sven
Rajkumar Manoharan May 4, 2020, 11:37 p.m. UTC | #2
On 2020-05-04 08:41, Markus Theil wrote:
> ath11k currently reports survey results for the last interval between 
> each
> invocation of NL80211_CMD_GET_SURVEY. For concurrent invocations, this
> can lead to unexpectedly small results, e.g. when hostapd uses survey
> data and iw survey dump is invoked in parallel. Fix this by returning
> cumulative results, that don't depend on the last invocation. Other
> drivers, e.g. ath9k or mt76 also use this behavior.
> 
> Signed-off-by: Markus Theil <markus.theil@tu-ilmenau.de>
> ---
>  drivers/net/wireless/ath/ath11k/wmi.c | 20 ++++++++++----------
>  1 file changed, 10 insertions(+), 10 deletions(-)
> 
> diff --git a/drivers/net/wireless/ath/ath11k/wmi.c
> b/drivers/net/wireless/ath/ath11k/wmi.c
> index c2a972377687..322ddfda5bfd 100644
> --- a/drivers/net/wireless/ath/ath11k/wmi.c
> +++ b/drivers/net/wireless/ath/ath11k/wmi.c
> @@ -5610,16 +5610,16 @@ ath11k_pdev_bss_chan_info_event(struct
> ath11k_base *ab, struct sk_buff *skb)
> 
>  	survey = &ar->survey[idx];
> 
> -	survey->noise     = bss_ch_info_ev.noise_floor;
> -	survey->time      = div_u64(total, cc_freq_hz);
> -	survey->time_busy = div_u64(busy, cc_freq_hz);
> -	survey->time_rx   = div_u64(rx_bss, cc_freq_hz);
> -	survey->time_tx   = div_u64(tx, cc_freq_hz);
> -	survey->filled   |= (SURVEY_INFO_NOISE_DBM |
> -			     SURVEY_INFO_TIME |
> -			     SURVEY_INFO_TIME_BUSY |
> -			     SURVEY_INFO_TIME_RX |
> -			     SURVEY_INFO_TIME_TX);
> +	survey->noise      = bss_ch_info_ev.noise_floor;
> +	survey->time      += div_u64(total, cc_freq_hz);
> +	survey->time_busy += div_u64(busy, cc_freq_hz);
> +	survey->time_rx   += div_u64(rx_bss, cc_freq_hz);
> +	survey->time_tx   += div_u64(tx, cc_freq_hz);
> +	survey->filled    |= (SURVEY_INFO_NOISE_DBM |
> +			      SURVEY_INFO_TIME |
> +			      SURVEY_INFO_TIME_BUSY |
> +			      SURVEY_INFO_TIME_RX |
> +			      SURVEY_INFO_TIME_TX);

Markus,

It depends on type of survey request is given to firmware. In ath11k, 
firmware reports
accumulated values. So the above addition is wrong and report double 
value. Have you
tested this change?

-Rajkumar
Markus Theil May 5, 2020, 6:53 a.m. UTC | #3
Am 05.05.2020 um 01:37 schrieb Rajkumar Manoharan:
> On 2020-05-04 08:41, Markus Theil wrote:
>> ath11k currently reports survey results for the last interval between 
>> each
>> invocation of NL80211_CMD_GET_SURVEY. For concurrent invocations, this
>> can lead to unexpectedly small results, e.g. when hostapd uses survey
>> data and iw survey dump is invoked in parallel. Fix this by returning
>> cumulative results, that don't depend on the last invocation. Other
>> drivers, e.g. ath9k or mt76 also use this behavior.
>>
>> Signed-off-by: Markus Theil <markus.theil@tu-ilmenau.de>
>> ---
>>  drivers/net/wireless/ath/ath11k/wmi.c | 20 ++++++++++----------
>>  1 file changed, 10 insertions(+), 10 deletions(-)
>>
>> diff --git a/drivers/net/wireless/ath/ath11k/wmi.c
>> b/drivers/net/wireless/ath/ath11k/wmi.c
>> index c2a972377687..322ddfda5bfd 100644
>> --- a/drivers/net/wireless/ath/ath11k/wmi.c
>> +++ b/drivers/net/wireless/ath/ath11k/wmi.c
>> @@ -5610,16 +5610,16 @@ ath11k_pdev_bss_chan_info_event(struct
>> ath11k_base *ab, struct sk_buff *skb)
>>
>>      survey = &ar->survey[idx];
>>
>> -    survey->noise     = bss_ch_info_ev.noise_floor;
>> -    survey->time      = div_u64(total, cc_freq_hz);
>> -    survey->time_busy = div_u64(busy, cc_freq_hz);
>> -    survey->time_rx   = div_u64(rx_bss, cc_freq_hz);
>> -    survey->time_tx   = div_u64(tx, cc_freq_hz);
>> -    survey->filled   |= (SURVEY_INFO_NOISE_DBM |
>> -                 SURVEY_INFO_TIME |
>> -                 SURVEY_INFO_TIME_BUSY |
>> -                 SURVEY_INFO_TIME_RX |
>> -                 SURVEY_INFO_TIME_TX);
>> +    survey->noise      = bss_ch_info_ev.noise_floor;
>> +    survey->time      += div_u64(total, cc_freq_hz);
>> +    survey->time_busy += div_u64(busy, cc_freq_hz);
>> +    survey->time_rx   += div_u64(rx_bss, cc_freq_hz);
>> +    survey->time_tx   += div_u64(tx, cc_freq_hz);
>> +    survey->filled    |= (SURVEY_INFO_NOISE_DBM |
>> +                  SURVEY_INFO_TIME |
>> +                  SURVEY_INFO_TIME_BUSY |
>> +                  SURVEY_INFO_TIME_RX |
>> +                  SURVEY_INFO_TIME_TX);
>
> Markus,
>
> It depends on type of survey request is given to firmware. In ath11k, 
> firmware reports
> accumulated values. So the above addition is wrong and report double 
> value. Have you
> tested this change?
>
> -Rajkumar
Ok, so please drop both of my patches. My assumptions were incomplete 
and too humble. I just assumed, ath10k and ath11k both use incremental 
surveys.
Rajkumar Manoharan May 5, 2020, 5:31 p.m. UTC | #4
On 2020-05-04 23:53, Markus Theil wrote:
> Am 05.05.2020 um 01:37 schrieb Rajkumar Manoharan:
>> On 2020-05-04 08:41, Markus Theil wrote:
>>> ath11k currently reports survey results for the last interval between 
>>> each
>>> invocation of NL80211_CMD_GET_SURVEY. For concurrent invocations, 
>>> this
>>> can lead to unexpectedly small results, e.g. when hostapd uses survey
>>> data and iw survey dump is invoked in parallel. Fix this by returning
>>> cumulative results, that don't depend on the last invocation. Other
>>> drivers, e.g. ath9k or mt76 also use this behavior.
>>> 
[...]
>> Markus,
>> 
>> It depends on type of survey request is given to firmware. In ath11k, 
>> firmware reports
>> accumulated values. So the above addition is wrong and report double 
>> value. Have you
>> tested this change?
>> 
>> -Rajkumar
> Ok, so please drop both of my patches. My assumptions were incomplete
> and too humble. I just assumed, ath10k and ath11k both use incremental
> surveys.
> 
Markus,

The ath10k driver still counts survey stats incrementally. But it should 
be handled by sending
appropriate survey request_type to firmware.

-Rajkumar

Patch
diff mbox series

diff --git a/drivers/net/wireless/ath/ath11k/wmi.c b/drivers/net/wireless/ath/ath11k/wmi.c
index c2a972377687..322ddfda5bfd 100644
--- a/drivers/net/wireless/ath/ath11k/wmi.c
+++ b/drivers/net/wireless/ath/ath11k/wmi.c
@@ -5610,16 +5610,16 @@  ath11k_pdev_bss_chan_info_event(struct ath11k_base *ab, struct sk_buff *skb)
 
 	survey = &ar->survey[idx];
 
-	survey->noise     = bss_ch_info_ev.noise_floor;
-	survey->time      = div_u64(total, cc_freq_hz);
-	survey->time_busy = div_u64(busy, cc_freq_hz);
-	survey->time_rx   = div_u64(rx_bss, cc_freq_hz);
-	survey->time_tx   = div_u64(tx, cc_freq_hz);
-	survey->filled   |= (SURVEY_INFO_NOISE_DBM |
-			     SURVEY_INFO_TIME |
-			     SURVEY_INFO_TIME_BUSY |
-			     SURVEY_INFO_TIME_RX |
-			     SURVEY_INFO_TIME_TX);
+	survey->noise      = bss_ch_info_ev.noise_floor;
+	survey->time      += div_u64(total, cc_freq_hz);
+	survey->time_busy += div_u64(busy, cc_freq_hz);
+	survey->time_rx   += div_u64(rx_bss, cc_freq_hz);
+	survey->time_tx   += div_u64(tx, cc_freq_hz);
+	survey->filled    |= (SURVEY_INFO_NOISE_DBM |
+			      SURVEY_INFO_TIME |
+			      SURVEY_INFO_TIME_BUSY |
+			      SURVEY_INFO_TIME_RX |
+			      SURVEY_INFO_TIME_TX);
 exit:
 	spin_unlock_bh(&ar->data_lock);
 	complete(&ar->bss_survey_done);