[ath9k-devel] [PATCH] ath9k: Avoid acktimeout wraps around at bootstrap

[ath9k-devel] [PATCH] ath9k: Avoid acktimeout wraps around at bootstrap

[Lorenzo Bianconi]

Hi all,

I am using ath9k/mac80211 on a PC Engines Alix with a Mikrotik R52n
card (AR9280 chipset), OpenWrt r21030 (kernel 2.6.32.10) and
compat-wireless-2010-04-21. I am injecting packets using a VAP in
monitor mode.
I have a doubt on ath9k_hw_init_defaults() in hw.c. In particular
during the bootstrap ah->slottime in ath9k_hw_init_defaults() is set
to " (u32) -1", so after a HW reset (for example as a result of a
channel change) in the function ath9k_hw_init_global_settings() the
acktimeout variable wraps around as ah->slottime is 0xFFFFFFFF and

acktimeout = sifstime + ah->slottime + 3 * ah->coverage_class

In this way we obtain that acktimout is set to 15 us (assuming
ah->coverage_class set to 0) in the 5GHZ band so less than one of its
summands (sifstime = 16).
I wrote this simple patch in order to set ah->slottime to standard
value for OFDM PHY layer during bootstrap.

Signed-off-by: Lorenzo Bianconi <lorenzo.bianconi83@gmail.com>
---
--- a/drivers/net/wireless/ath/ath9k/hw.c
+++ b/drivers/net/wireless/ath/ath9k/hw.c
@@ -425,7 +425,7 @@
 	ah->sta_id1_defaults = AR_STA_ID1_CRPT_MIC_ENABLE;
 	ah->beacon_interval = 100;
 	ah->enable_32kHz_clock = DONT_USE_32KHZ;
-	ah->slottime = (u32) -1;
+	ah->slottime = 20;
 	ah->globaltxtimeout = (u32) -1;
 	ah->power_mode = ATH9K_PM_UNDEFINED;
 }
--

Regards

Lorenzo

[ath9k-devel] [PATCH] ath9k: Avoid acktimeout wraps around at bootstrap

[Benoit PAPILLAULT]

Lorenzo Bianconi a ?crit :
> Hi all,
>
> I am using ath9k/mac80211 on a PC Engines Alix with a Mikrotik R52n
> card (AR9280 chipset), OpenWrt r21030 (kernel 2.6.32.10) and
> compat-wireless-2010-04-21. I am injecting packets using a VAP in
> monitor mode.
> I have a doubt on ath9k_hw_init_defaults() in hw.c. In particular
> during the bootstrap ah->slottime in ath9k_hw_init_defaults() is set
> to " (u32) -1", so after a HW reset (for example as a result of a
> channel change) in the function ath9k_hw_init_global_settings() the
> acktimeout variable wraps around as ah->slottime is 0xFFFFFFFF and
>
> acktimeout = sifstime + ah->slottime + 3 * ah->coverage_class
>
> In this way we obtain that acktimout is set to 15 us (assuming
> ah->coverage_class set to 0) in the 5GHZ band so less than one of its
> summands (sifstime = 16).
> I wrote this simple patch in order to set ah->slottime to standard
> value for OFDM PHY layer during bootstrap.
>   
Does this have any visible effect or is this patch for strict 
compliance? I am asking this since I see very strange stuff regarding 
ACK (delayed, missing, out of order, ...)

Regards,
Benoit

[ath9k-devel] [PATCH] ath9k: Avoid acktimeout wraps around at bootstrap

[Lorenzo Bianconi]

> Lorenzo Bianconi a ?crit :
>>
>> Hi all,
>>
>> I am using ath9k/mac80211 on a PC Engines Alix with a Mikrotik R52n
>> card (AR9280 chipset), OpenWrt r21030 (kernel 2.6.32.10) and
>> compat-wireless-2010-04-21. I am injecting packets using a VAP in
>> monitor mode.
>> I have a doubt on ath9k_hw_init_defaults() in hw.c. In particular
>> during the bootstrap ah->slottime in ath9k_hw_init_defaults() is set
>> to " (u32) -1", so after a HW reset (for example as a result of a
>> channel change) in the function ath9k_hw_init_global_settings() the
>> acktimeout variable wraps around as ah->slottime is 0xFFFFFFFF and
>>
>> acktimeout = sifstime + ah->slottime + 3 * ah->coverage_class
>>
>> In this way we obtain that acktimout is set to 15 us (assuming
>> ah->coverage_class set to 0) in the 5GHZ band so less than one of its
>> summands (sifstime = 16).
>> I wrote this simple patch in order to set ah->slottime to standard
>> value for OFDM PHY layer during bootstrap.
>>
>
> Does this have any visible effect or is this patch for strict compliance? I
> am asking this since I see very strange stuff regarding ACK (delayed,
> missing, out of order, ...)
>
> Regards,
> Benoit
>

Hi Benoit,

In the simple network configuration I am testing (just two devices
which transmit through VAPs in monitor mode on a 5GHz channel), I
found out that stations are not able to ping each other if ack timeout
is less than 20us, I have to explicitly set a greater value.

Regards.

Lorenzo