Discussion:
'ath0: device timeout' findings
(too old to reply)
David Young
2007-11-29 20:49:44 UTC
Permalink
I have experimented a bit with an ***@cardbus and ***@pci to see if I
can stop them from printing 'ath0: device timeout'.

Turning off Tx interrupt mitigation with 'sysctl -w
hw.ath0.txintrperiod=1' helps widen the gap between timeouts from seconds
to minutes, but it does not stop them.

It looks like all interrupts (Rx & Tx) have ceased when a device timeout
occurs, and actual radio transmissions seem to stop, too. However,
it looks like the DMA engine continues to mark descriptors on the Tx
ring as completed.

I have experimented with both the PCI exception reporting and PCI
performance parameters on my Soekris net45xx boards, and I see my
***@pci on a net4526 report many "delayed transaction timeouts," which
indicates that one of my PCI devices does not play nice and retry PCI
transactions. I am not sure, yet, if the Atheros adapter is guilty.
I am going to see if I can enable the same exception reporting on my
net4521's Cardbus bridge.

000:00:0: Advanced Micro Devices Elan SC520 System Controller (host bridge)
000:16:0: Atheros Communications AR5212 Wireless LAN (ethernet network, revision 0x01)
000:18:0: National Semiconductor DP83815 10/100 Ethernet (ethernet network)

Dave
--
David Young OJC Technologies
***@ojctech.com Urbana, IL * (217) 278-3933 ext 24

--
Posted automagically by a mail2news gateway at muc.de e.V.
Please direct questions, flames, donations, etc. to news-***@muc.de
j***@kollasch.net
2007-11-29 21:57:32 UTC
Permalink
Post by David Young
can stop them from printing 'ath0: device timeout'.
Do we know what caused this to start? This was not
an issue when I was running 4.99.22. On 4.99.34 it is.

Jonathan Kollasch
David Young
2007-11-29 22:34:46 UTC
Permalink
Post by j***@kollasch.net
Post by David Young
can stop them from printing 'ath0: device timeout'.
Do we know what caused this to start? This was not
an issue when I was running 4.99.22. On 4.99.34 it is.
Seems to me there have always been timeouts, but some people see none,
and others see many. I just looked through most of my commits between
4.99.22 and 4.99.34. What if you back out these changes I made and test?

cvs rdiff -r1.83 -r1.84 src/sys/dev/ic/ath.c
cvs rdiff -r1.20 -r1.21 src/sys/dev/ic/athvar.h

Dave
--
David Young OJC Technologies
***@ojctech.com Urbana, IL * (217) 278-3933 ext 24

--
Posted automagically by a mail2news gateway at muc.de e.V.
Please direct questions, flames, donations, etc. to news-***@muc.de
Sean Boudreau
2007-11-30 00:43:54 UTC
Permalink
Post by David Young
I
Post by j***@kollasch.net
Post by David Young
can stop them from printing 'ath0: device timeout'.
Do we know what caused this to start? This was not
an issue when I was running 4.99.22. On 4.99.34 it is.
Seems to me there have always been timeouts, but some people see none,
and others see many. I just looked through most of my commits between
4.99.22 and 4.99.34. What if you back out these changes I made and test?
cvs rdiff -r1.83 -r1.84 src/sys/dev/ic/ath.c
cvs rdiff -r1.20 -r1.21 src/sys/dev/ic/athvar.h
I see these infrequently on 3.99.19.

-seanb

--
Posted automagically by a mail2news gateway at muc.de e.V.
Please direct questions, flames, donations, etc. to news-***@muc.de
Joerg Sonnenberger
2007-11-30 10:44:14 UTC
Permalink
Post by David Young
I have experimented with both the PCI exception reporting and PCI
performance parameters on my Soekris net45xx boards, and I see my
indicates that one of my PCI devices does not play nice and retry PCI
transactions.
Do any devices on the bus have non-zero register 0x41?

Joerg

--
Posted automagically by a mail2news gateway at muc.de e.V.
Please direct questions, flames, donations, etc. to news-***@muc.de
Loading...