Re: [ath9k-devel] ath9k module-Unable to set channel-since ath comes

Re: [ath9k-devel] ath9k module-Unable to set channel-since ath comes

[Luis R. Rodriguez]

On Fri, May 8, 2009 at 2:53 PM, Luis R. Rodriguez <mcgrof@gmail.com> wr=
ote:
> On Fri, May 8, 2009 at 2:40 PM, Pavel Roskin <proski@gnu.org> wrote:
>> On Fri, 2009-05-08 at 23:12 +0200, Alexander Bartha wrote:
>>> Hello to everybody at linuxwireless
>>>
>>> Since the "Submodule" ath comes aht9k says in messages:
>>>
>>> May =C2=A07 22:11:34 ws1 [ 4442.304763] ath9k: Unable to set channe=
l
>>> May =C2=A07 22:11:34 ws1 [ 4442.425550] ath9k: timeout (100000 us) =
on reg 0x9860: 0x0
>>> 0049d19 & 0x00000001 !=3D 0x00000000
>>> May =C2=A07 22:11:34 ws1 [ 4442.425555] ath9k: Unable to reset chan=
nel (2422 Mhz) res
>>> et status 4294967291
>>> May =C2=A07 22:11:34 ws1 [ 4442.425560] ath9k: Unable to set channe=
l
>>> May =C2=A07 22:11:34 ws1 [ 4442.546365] ath9k: timeout (100000 us) =
on reg 0x9860: 0x0
>>> 0049d19 & 0x00000001 !=3D 0x00000000

I can reproduce, and I tried without the idle patches but I still see
the issue. I see this issue only after suspend. Did you suspend and
get this as well? I see this using AR5416. If you are seeing this on
suspend can you try with the patch I posted to linux-wireless to see
if it helps a little?

  Luis

>>> i have tried different versions of the modules but since compat-wir=
eless-2009-04-16 i get this errors.
>>>
>>> till that all works fine sometime the driver have little problems a=
fter not using the network. thanks for the drivers
>>
>> I track wireless-testing.git and I saw it in the last weeks. =C2=A0I=
 think it
>> happens after about 260 pings at the normal speed (i.e. not with flo=
od
>> ping). =C2=A0I was thinking about bisecting the problem, but I don't=
 have
>> much time right now.
>>
>> Last time I saw it was on Monday. =C2=A0There were some changes to a=
th9k
>> since then, so maybe it's fixed.
>
> This may have been caused by the idle patches for which Johannes has
> posted a fix. This should get merged soon so give it a few more days
> and try again.
>
> If you want something a little more stable you can go with the stable
> compat-wireless releases:
>
> http://wireless.kernel.org/en/users/Download/stable
>
> =C2=A0Luis
>
--
To unsubscribe from this list: send the line "unsubscribe linux-wireles=
s" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [ath9k-devel] ath9k module-Unable to set channel-since ath comes

[Pavel Roskin]

Quoting "Luis R. Rodriguez" <mcgrof@gmail.com>:

> I can reproduce, and I tried without the idle patches but I still see
> the issue. I see this issue only after suspend. Did you suspend and
> get this as well? I see this using AR5416. If you are seeing this on
> suspend can you try with the patch I posted to linux-wireless to see
> if it helps a little?

I saw this on AR9280 without any suspends.  The card doesn't support  
the 802.11a range, so it's not a direct equivalent of the problem I  
bisected with ath5k (invalid rates when scanning and transmitting).

-- 
Regards,
Pavel Roskin

Re: [ath9k-devel] ath9k module-Unable to set channel-since ath comes

[Luis R. Rodriguez]

On Sun, May 10, 2009 at 9:17 PM, Pavel Roskin <proski@gnu.org> wrote:
> Quoting "Luis R. Rodriguez" <mcgrof@gmail.com>:
>
>> I can reproduce, and I tried without the idle patches but I still se=
e
>> the issue. I see this issue only after suspend. Did you suspend and
>> get this as well? I see this using AR5416. If you are seeing this on
>> suspend can you try with the patch I posted to linux-wireless to see
>> if it helps a little?
>
> I saw this on AR9280 without any suspends. =C2=A0The card doesn't sup=
port the
> 802.11a range, so it's not a direct equivalent of the problem I bisec=
ted
> with ath5k (invalid rates when scanning and transmitting).

Thanks for the heads up Pavel. So can you guys test with this option
set as follows:

CONFIG_MAC80211_DEFAULT_PS_VALUE=3D0

If using compat-wireless just stuff it into config.mk.

This fixes my issues and points towards PS code. This could be how we
handle PS in ath9k (more likely) or mac80211. For now please disable
this if using ath9k until we figure out what causes this.

  Luis
--
To unsubscribe from this list: send the line "unsubscribe linux-wireles=
s" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [ath9k-devel] ath9k module-Unable to set channel-since ath comes

[Pavel Roskin]

On Mon, 2009-05-11 at 00:40 -0700, Luis R. Rodriguez wrote:
> On Sun, May 10, 2009 at 9:17 PM, Pavel Roskin <proski@gnu.org> wrote:
> > Quoting "Luis R. Rodriguez" <mcgrof@gmail.com>:
> >
> >> I can reproduce, and I tried without the idle patches but I still see
> >> the issue. I see this issue only after suspend. Did you suspend and
> >> get this as well? I see this using AR5416. If you are seeing this on
> >> suspend can you try with the patch I posted to linux-wireless to see
> >> if it helps a little?
> >
> > I saw this on AR9280 without any suspends.  The card doesn't support the
> > 802.11a range, so it's not a direct equivalent of the problem I bisected
> > with ath5k (invalid rates when scanning and transmitting).
> 
> Thanks for the heads up Pavel. So can you guys test with this option
> set as follows:
> 
> CONFIG_MAC80211_DEFAULT_PS_VALUE=0
> 
> If using compat-wireless just stuff it into config.mk.
> 
> This fixes my issues and points towards PS code. This could be how we
> handle PS in ath9k (more likely) or mac80211. For now please disable
> this if using ath9k until we figure out what causes this.

It looks like there are two issues, and I confused them.  One is
characterized by the kernel messages about the channel, the one
originally reported in this thread.  I have seen it, but I don't
remember the exact circumstances.

Another issue is connection loss to a WPA enabled AP after several
minutes.  That's what I have seen with AR9280.  iwconfig still shows an
association, but no traffic can flow.

I just want to tell that it's two different issues.  The later is much
more critical for me, but it would take about 15 minutes to verify every
step if bisecting.  Unfortunately, I don't have time to bisect it now,
but maybe I'll do it later during this week.

-- 
Regards,
Pavel Roskin

Re: [ath9k-devel] ath9k module-Unable to set channel-since ath comes

[Luis R. Rodriguez]

On Mon, May 11, 2009 at 3:06 PM, Pavel Roskin <proski@gnu.org> wrote:
> On Mon, 2009-05-11 at 00:40 -0700, Luis R. Rodriguez wrote:
>> On Sun, May 10, 2009 at 9:17 PM, Pavel Roskin <proski@gnu.org> wrote=
:
>> > Quoting "Luis R. Rodriguez" <mcgrof@gmail.com>:
>> >
>> >> I can reproduce, and I tried without the idle patches but I still=
 see
>> >> the issue. I see this issue only after suspend. Did you suspend a=
nd
>> >> get this as well? I see this using AR5416. If you are seeing this=
 on
>> >> suspend can you try with the patch I posted to linux-wireless to =
see
>> >> if it helps a little?
>> >
>> > I saw this on AR9280 without any suspends. =C2=A0The card doesn't =
support the
>> > 802.11a range, so it's not a direct equivalent of the problem I bi=
sected
>> > with ath5k (invalid rates when scanning and transmitting).
>>
>> Thanks for the heads up Pavel. So can you guys test with this option
>> set as follows:
>>
>> CONFIG_MAC80211_DEFAULT_PS_VALUE=3D0
>>
>> If using compat-wireless just stuff it into config.mk.
>>
>> This fixes my issues and points towards PS code. This could be how w=
e
>> handle PS in ath9k (more likely) or mac80211. For now please disable
>> this if using ath9k until we figure out what causes this.
>
> It looks like there are two issues, and I confused them. =C2=A0One is
> characterized by the kernel messages about the channel, the one
> originally reported in this thread. =C2=A0I have seen it, but I don't
> remember the exact circumstances.
>
> Another issue is connection loss to a WPA enabled AP after several
> minutes. =C2=A0That's what I have seen with AR9280. =C2=A0iwconfig st=
ill shows an
> association, but no traffic can flow.
>
> I just want to tell that it's two different issues. =C2=A0The later i=
s much
> more critical for me, but it would take about 15 minutes to verify ev=
ery
> step if bisecting. =C2=A0Unfortunately, I don't have time to bisect i=
t now,
> but maybe I'll do it later during this week.

Thanks for the heads up, lets address this in another thread then.

  Luis
--
To unsubscribe from this list: send the line "unsubscribe linux-wireles=
s" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Re: [ath9k-devel] ath9k module-Unable to set channel-since ath comes

[Luis R. Rodriguez]

On Mon, May 11, 2009 at 11:38 PM,  <alexander.bartha@drz-wien.at> wrote:
>
> Hello to linuxwireless
>
> thanks for response. The stabel compat-wireless-2.6.30-rc4.tar.bz2 release
> works well. The bleeding edge compat-wireless works to but only till the new
> submodule ath comes in live (compat-wireless-2009-04-12.tar.bz2 is the last
> version which i have tasted and which works well).

I've tested this and its not the "ath" module which introduces an
issue its a patch which enables PS by default. You can disable that by
recompiling with:

CONFIG_MAC80211_DEFAULT_PS_VALUE=0

If using compat-wireless just stuff it into config.mk.

  Luis

Re: Re: [ath9k-devel] ath9k module-Unable to set channel-since ath comes

[Johannes Berg]

[-- Attachment #1: Type: text/plain, Size: 740 bytes --]

On Mon, 2009-05-11 at 23:55 -0700, Luis R. Rodriguez wrote:

> I've tested this and its not the "ath" module which introduces an
> issue its a patch which enables PS by default. You can disable that by
> recompiling with:
> 
> CONFIG_MAC80211_DEFAULT_PS_VALUE=0
> 
> If using compat-wireless just stuff it into config.mk.

I object to that assertion, or at least this way of putting it.

If powersave by default causes issues then it's not my patch that's
causing the problem, it's the fact that ath9k can't properly deal with
powersave in this case. Had the user enabled powersave manually they
would have run into the same issue. As such, I think ath9k should
probably simply stop claiming it can support PS.

johannes

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 801 bytes --]

Re: Re: [ath9k-devel] ath9k module-Unable to set channel-since ath comes

[Luis R. Rodriguez]

On Tue, May 12, 2009 at 12:11 AM, Johannes Berg
<johannes@sipsolutions.net> wrote:
> On Mon, 2009-05-11 at 23:55 -0700, Luis R. Rodriguez wrote:
>
>> I've tested this and its not the "ath" module which introduces an
>> issue its a patch which enables PS by default. You can disable that by
>> recompiling with:
>>
>> CONFIG_MAC80211_DEFAULT_PS_VALUE=0
>>
>> If using compat-wireless just stuff it into config.mk.
>
> I object to that assertion, or at least this way of putting it.
>
> If powersave by default causes issues then it's not my patch that's
> causing the problem, it's the fact that ath9k can't properly deal with
> powersave in this case. Had the user enabled powersave manually they
> would have run into the same issue. As such, I think ath9k should
> probably simply stop claiming it can support PS.

Heh yes, what I meant was ath9k's PS support has some issues which
still need to be resolved (at least as I've tested it).

  Luis

Re: Re: [ath9k-devel] ath9k module-Unable to set channel-since ath comes

[Davide Pesavento]

Hi,

On Tue, May 12, 2009 at 09:14, Luis R. Rodriguez <mcgrof@gmail.com> wrote:
> On Tue, May 12, 2009 at 12:11 AM, Johannes Berg
> <johannes@sipsolutions.net> wrote:
>> On Mon, 2009-05-11 at 23:55 -0700, Luis R. Rodriguez wrote:
>>
>>> I've tested this and its not the "ath" module which introduces an
>>> issue its a patch which enables PS by default. You can disable that by
>>> recompiling with:
>>>
>>> CONFIG_MAC80211_DEFAULT_PS_VALUE=0
>>>
>>> If using compat-wireless just stuff it into config.mk.
>>
>> I object to that assertion, or at least this way of putting it.
>>
>> If powersave by default causes issues then it's not my patch that's
>> causing the problem, it's the fact that ath9k can't properly deal with
>> powersave in this case. Had the user enabled powersave manually they
>> would have run into the same issue. As such, I think ath9k should
>> probably simply stop claiming it can support PS.
>
> Heh yes, what I meant was ath9k's PS support has some issues which
> still need to be resolved (at least as I've tested it).
>

I've already reported ath9k's issues with PS multiple times [1] [2]
[3], but the driver hasn't been fixed yet. You can find a verbose
debugging output at [4], I hope that helps in fixing the problem (I
can provide even more info if needed and I'm willing to test patches).

Anyway, if the issue isn't going to be resolved quickly (e.g. before
Johannes' "enable PS by default" patch lands in mainline), I think
it's better to disable PS in ath9k as previously suggested.

Regards,
Davide

[1] http://marc.info/?l=linux-wireless&m=123816406913599&w=2
[2] http://marc.info/?l=linux-wireless&m=123999864507104&w=2
[3] http://marc.info/?l=linux-wireless&m=124025926613089&w=2
[4] http://marc.info/?l=linux-wireless&m=124026772102023&w=2

Re: [ath9k-devel] ath9k module-Unable to set channel-since ath comes

[Jouni Malinen]

On Tue, 2009-05-12 at 01:35 -0700, Davide Pesavento wrote:
> I've already reported ath9k's issues with PS multiple times [1] [2]
> [3], but the driver hasn't been fixed yet. You can find a verbose
> debugging output at [4], I hope that helps in fixing the problem (I
> can provide even more info if needed and I'm willing to test patches).

I submitted a set of patches to resolve some of the PS issues in ath9k.
I would expect these to hit wireless-testing.git eventually, but if you
are willing to test the changes before that, you can find the patches
from http://w1.fi/p/ps/

One thing to note here is that there are still some issues with the
timeout=0 case (use of PS-Poll to fetch buffered frames) which is
enabled with "iwconfig wlan0 power on", so at this point, I would
recommend running the test either with the default wireless-testing.git
PS mode setting or by enabling power saving with "iwconfig wlan0 power
timeout 500ms" which should result in the same configuration.

I have not tested these on AR5418, but at least with AR9280, these
patches do improve stability and general functionality of the default PS
mode considerably. I have been able to run multi-hour tests without
losing connection.

- Jouni

Re: [ath9k-devel] ath9k module-Unable to set channel-since ath comes

[Kalle Valo]

Jouni Malinen <jouni.malinen@atheros.com> writes:

> On Tue, 2009-05-12 at 01:35 -0700, Davide Pesavento wrote:
>> I've already reported ath9k's issues with PS multiple times [1] [2]
>> [3], but the driver hasn't been fixed yet. You can find a verbose
>> debugging output at [4], I hope that helps in fixing the problem (I
>> can provide even more info if needed and I'm willing to test patches).
>
> I submitted a set of patches to resolve some of the PS issues in ath9k.
> I would expect these to hit wireless-testing.git eventually, but if you
> are willing to test the changes before that, you can find the patches
> from http://w1.fi/p/ps/
>
> One thing to note here is that there are still some issues with the
> timeout=0 case (use of PS-Poll to fetch buffered frames) which is
> enabled with "iwconfig wlan0 power on", so at this point, I would
> recommend running the test either with the default wireless-testing.git
> PS mode setting or by enabling power saving with "iwconfig wlan0 power
> timeout 500ms" which should result in the same configuration.

Most probably that's because mac80211 does not disable power save mode
when sending PS-Poll frames. So stlc45xx does not require disable of PSM
but then again ath9k requires it, we need to solve this somehow.

I have been thinking about separating power save mode and firmware
wakeup/sleep modes from each other in mac80211 API. To me it looks like
ath9k does more like firmware wakeup/sleep when power save mode is
changing. But this is just a wild idea, I have no idea if it's practical
or not.

-- 
Kalle Valo