[ath9k-devel] ath9k flaky in 3.6(.x)

[ath9k-devel] ath9k flaky in 3.6(.x)

[Sune Mølgaard]

Hiya,

During the 3.6 rcs, I have experienced progressively worse problems with 
connection drops from an AP that seems to be serving my GF's Mac just 
fine (even during these problems).

3.6.0 seems to be mostly fine for a number of hours (as did the rcs), 
but then it begins. 3.6.1 and 3.6.2 are basically unusable.

I'm hoping for someone to say "this is a known problem - we are working 
on it", but if not, I shall be happy to git bisect, though the process 
might be somewhat slow since the problem only seems to manifest itself 
after some hours in some of the earlier cases.

As we speak, I'm on a 3.5-something kernel that seems to work flawlessly.

Relevant output from lspci is:

08:01.0 Network controller: Atheros Communications Inc. AR922X Wireless 
Network Adapter (rev 01)

Best regards,

Sune M?lgaard

-- 
In those days spirits were brave, the stakes were high, men were real
men, women were real women and small furry creatures from Alpha
Centauri were real small furry creatures from Alpha Centauri.
- Douglas Adams

[ath9k-devel] ath9k flaky in 3.6(.x)

[Adrian Chadd]

Please do go through and bisect things.

The active ath9k developers will really, really appreciate it if you:

a) file a bugzilla.kernel.org bug; then
b) bisect things until you find culprits and report them



adrian


On 13 October 2012 10:35, Sune M?lgaard <sune@molgaard.org> wrote:
> Hiya,
>
> During the 3.6 rcs, I have experienced progressively worse problems with
> connection drops from an AP that seems to be serving my GF's Mac just
> fine (even during these problems).
>
> 3.6.0 seems to be mostly fine for a number of hours (as did the rcs),
> but then it begins. 3.6.1 and 3.6.2 are basically unusable.
>
> I'm hoping for someone to say "this is a known problem - we are working
> on it", but if not, I shall be happy to git bisect, though the process
> might be somewhat slow since the problem only seems to manifest itself
> after some hours in some of the earlier cases.
>
> As we speak, I'm on a 3.5-something kernel that seems to work flawlessly.
>
> Relevant output from lspci is:
>
> 08:01.0 Network controller: Atheros Communications Inc. AR922X Wireless
> Network Adapter (rev 01)
>
> Best regards,
>
> Sune M?lgaard
>
> --
> In those days spirits were brave, the stakes were high, men were real
> men, women were real women and small furry creatures from Alpha
> Centauri were real small furry creatures from Alpha Centauri.
> - Douglas Adams
> _______________________________________________
> ath9k-devel mailing list
> ath9k-devel at lists.ath9k.org
> https://lists.ath9k.org/mailman/listinfo/ath9k-devel

[ath9k-devel] ath9k flaky in 3.6(.x)

[Peter Stuge]

Sune M?lgaard wrote:
> I'm hoping for someone to say "this is a known problem - we are
> working on it"

Forget it. You will also not really get any help with debugging. If
you can find, and ideally prove, a regression then developers will
likely look into the problem.

If you perceive and can demonstrate a problem, and complain loudly,
then developers will attempt to reproduce the problem in a lab, but
that is of course absolutely useless.


> 08:01.0 Network controller: Atheros Communications Inc.
>   AR922X Wireless Network Adapter (rev 01)

This is quite old hardware by now. The driver is supposed to work
well also with this hardware, but in practise it has never for me.

I have a feeling that the AR922x PCI silicon is not so great, and
that it causes some of the problems. Perhaps the driver could work
around them, but I'm not sure if it does.


When dealing with this kernel driver your best chances at any kind of
success is to use the hardware that developers currently work on.
Those code paths are actively tested, improved, and debugged. Of
course there is a small complication here; since the developers work
on the driver before the hardware is made available on the market.

At the moment I would not use anything older than AR9380 with the
ath9k driver.


I would suggest that you give FreeBSD a go on your laptop. Adrian is
the author of that driver, it is not a copypaste of the ath9k, so it
may work better or simply differently. I believe it supports your
card. You obviously have to be open to trying out FreeBSD for that.


//Peter

[ath9k-devel] ath9k flaky in 3.6(.x)

[Sune Mølgaard]

Peter Stuge wrote:
> Sune M?lgaard wrote:
>> I'm hoping for someone to say "this is a known problem - we are
>> working on it"
>
> Forget it. You will also not really get any help with debugging. If
> you can find, and ideally prove, a regression then developers will
> likely look into the problem.
>
> If you perceive and can demonstrate a problem, and complain loudly,
> then developers will attempt to reproduce the problem in a lab, but
> that is of course absolutely useless.

Behaviour seems erratic, but without any clear cause - have temporarily 
switched to USB dongle that I remembered I had lying around...

>> 08:01.0 Network controller: Atheros Communications Inc.
>>    AR922X Wireless Network Adapter (rev 01)
>
> This is quite old hardware by now. The driver is supposed to work
> well also with this hardware, but in practise it has never for me.
>
> I have a feeling that the AR922x PCI silicon is not so great, and
> that it causes some of the problems. Perhaps the driver could work
> around them, but I'm not sure if it does.

That could very well be true. Would you happen to know a resource for 
finding consumer-level hardware with specific (new) chips?

>
> When dealing with this kernel driver your best chances at any kind of
> success is to use the hardware that developers currently work on.
> Those code paths are actively tested, improved, and debugged. Of
> course there is a small complication here; since the developers work
> on the driver before the hardware is made available on the market.
>
> At the moment I would not use anything older than AR9380 with the
> ath9k driver.
>
>
> I would suggest that you give FreeBSD a go on your laptop. Adrian is
> the author of that driver, it is not a copypaste of the ath9k, so it
> may work better or simply differently. I believe it supports your
> card. You obviously have to be open to trying out FreeBSD for that.

I'm not opposed to FreeBSD per se, but I do like me my Linux, so if you, 
or anyone else can answer my questions above, I'll buy a newer card.

>
>
> //Peter
> _______________________________________________
> ath9k-devel mailing list
> ath9k-devel at lists.ath9k.org
> https://lists.ath9k.org/mailman/listinfo/ath9k-devel
>

Thank you for answering.

Best regards,

Sune M?lgaard

-- 
No trees were harmed in the transmission of this email but trillions of 
electrons were excited to participate.

[ath9k-devel] ath9k flaky in 3.6(.x)

[Sujith Manoharan]

Sune M?lgaard wrote:
> During the 3.6 rcs, I have experienced progressively worse problems with 
> connection drops from an AP that seems to be serving my GF's Mac just 
> fine (even during these problems).
> 
> 3.6.0 seems to be mostly fine for a number of hours (as did the rcs), 
> but then it begins. 3.6.1 and 3.6.2 are basically unusable.
> 
> I'm hoping for someone to say "this is a known problem - we are working 
> on it", but if not, I shall be happy to git bisect, though the process 
> might be somewhat slow since the problem only seems to manifest itself 
> after some hours in some of the earlier cases.
> 
> As we speak, I'm on a 3.5-something kernel that seems to work flawlessly.
> 
> Relevant output from lspci is:
> 
> 08:01.0 Network controller: Atheros Communications Inc. AR922X Wireless 
> Network Adapter (rev 01)

Can you try the latest compat-wireless package ? Please enable debugging
and post the kernel log when the issue arises. Loading the driver with
debug=0x8f49 would give a fair amount of debug messages.

Also, information about the AP, connection type etc. would be useful.

https://backports.wiki.kernel.org/index.php/Releases
http://linuxwireless.org/en/users/Drivers/ath9k/debug

Sujith