Hi Ben,
On Tue, Jul 24, 2012 at 4:01 AM, Ben Greear <gree...@candelatech.com> wrote:
> On 07/23/2012 03:08 PM, Ben Greear wrote:
>> On 07/23/2012 02:10 PM, Ben Greear wrote:
>>> Came back after a 1 week vacation and found the 3.3.8+ kernel spitting
>>> timeout errors, and network devices will not 'ifconfig foo up'.
>>>
>>> I rebooted into 3.5.0+, and see the same (or at least similar errors):
>>
>> Well, I replaced the NIC and the problem remains.
>>
>> Guess it's time to poke a bit deeper.
>
> Ahh..so here's what happened. I added code to set the rx-chainmask
> and tx-chainmask from user-space app (via writing to appropriate debugfs
> files).
>
> Code assumed 0x7 by default, but this particular NIC is only 2x2.
>
> When the chainmask is set wrong, the NIC gets into the broken state.
great!, this could be one root cause for chip reset failures!
>
> Changing it back to 0x3 fixes the problem.
>
> Is that worth trying to fix in the driver, or should I just
> fix it in user-space so that it never sets more than what
> the eeprom reports as supported?
Felix made a fix for broken EEPROM chainmasks.
commit 6054069a03f77ffa686e2dfd5f07cff8ee40b72d
Author: Felix Fietkau <n...@openwrt.org>
Date: Tue Jul 19 08:46:44 2011 +0200
ath9k_hw: validate and fix broken eeprom chainmask settings
Some devices (e.g. Ubiquiti AirRouter) ship with broken EEPROM chainmask
data, which breaks the initial calibration after a hardware reset.
To fix this, mask the eeprom chainmask with the chainmask of the chip,
and use the chip chainmask if the result is zero.
Signed-off-by: Felix Fietkau <n...@openwrt.org>
Signed-off-by: John W. Linville <linvi...@tuxdriver.com>
the hard coded chain mask comes into picture only when the EEPROM
chainmask settings are zero.
Incase we are validating the chainmask in the driver we got to be sure
of validating for all chipsets.
Also we need to figure it out how to differentiate AR9382 (2x2) and
AR9380 (3x3).
>
> --
> Ben Greear <gree...@candelatech.com>
> Candela Technologies Inc http://www.candelatech.com
>
>
>
> _______________________________________________
> ath9k-devel mailing list
> ath9k-devel@lists.ath9k.org
> https://lists.ath9k.org/mailman/listinfo/ath9k-devel
--
thanks,
shafi
_______________________________________________
ath9k-devel mailing list
ath9k-devel@lists.ath9k.org
https://lists.ath9k.org/mailman/listinfo/ath9k-devel
