[ath9k] Kernel panic in 2.6.27-11

[Aaron Kelley]

Hello, folks. I have an Atheros wireless card with the AR5008 chipset (same as used in some MacBook Pros). Performance in Ubuntu 8.10 has generally been poor. However, there are already other bug reports about that.

One of these bug reports is #297965 (https://bugs.launchpad.net/ubuntu/+source/linux/+bug/297965), where John Pugh said that he could not reproduce the performance problem in the latest kernel for intrepid. Still having the performance problem in 2.6.27-9, I decided to grab the latest kernel update and related packages from intrepid-proposed (2.6.27-11) and see if anything would improve.

Performance was better in the new version of the kernel, but after using the wireless for a while, my machine crashed with kernel panic. That is the subject of this bug report.

I suspect the ath9k module for a couple of reasons.
 - First of all, I had been using the machine on 2.6.27-11 for several days, but I usually don't use the wireless, but it crashed after maybe two hours of wireless use.
 - Also, I've tried the compat-wireless modules before (http://wireless.kernel.org/en/users/Download) with the same results: better performance, but frequent kernel panic.

I'd like to hear if anyone else has tried 2.6.27-11 who has an AR5008-based wireless card.

I'm still a relative newcomer to Linux, so if there is more information that you would like me to provide, either tell me what to do or point me towards some directions.

[Luis R. Rodriguez (mcgrof)]

We need a stack trace (oops message). Do you have that? Please try to take a picture if you cannot get it. You can also try to get the panic by not using X, simply go into a virtual terminal.

[Aaron Kelley (aaronkelley)]

I just discovered and set up linux-crashdump, so hopefully next time it panics, I'll be able to capture some useful information.

[Aaron Kelley (aaronkelley)]

Happened again today. linux-crashdump did not seem to do its job. I will try to repro from a tty so I can see if there is any useful output there.

[maxauthority (stubenschrott)]

I can confirm this bug. I have this card:

09:00.0 Network controller: Atheros Communications Inc. AR5008 Wireless Network Adapter (rev 01)
 Subsystem: Atheros Communications Inc. Device 3071
 Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B- DisINTx-
 Status: Cap+ 66MHz+ UDF- FastB2B+ ParErr- DEVSEL=medium >TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx-
 Latency: 96, Cache Line Size: 64 bytes
 Interrupt: pin A routed to IRQ 16
 Region 0: Memory at fd2e0000 (32-bit, non-prefetchable) [size=64K]
 Capabilities: [40] #80 [0000]
 Kernel driver in use: ath_pci
 Kernel modules: ath_pci, ath9k

(as you can see, using madwifi currently, not perfect but seems to work).

Nearly everytime when nm-applet tries to connect with ath9k it spins its icons in the system tray for a few seconds, and then the system freezes (no mouse, etc. - need to resart). The few times where it does connect, it usually hangs itself a few minutes afterwards. Also tried today's compat-wireless drivers, same result.

Ubuntu 8.10 with uname: Linux maxauth 2.6.27-11-generic #1 SMP Thu Jan 29 19:28:32 UTC 2009 x86_64 GNU/Linux

> dmesg | grep -i ath9k
[ 11.911350] ath9k 0000:09:00.0: PCI INT A -> GSI 16 (level, low) -> IRQ 16
[ 12.343578] phy0: Selected rate control algorithm 'ath9k_rate_control'
[ 12.420150] Registered led device: ath9k-phy0::radio
[ 12.420169] Registered led device: ath9k-phy0::assoc
[ 12.420185] Registered led device: ath9k-phy0::tx
[ 12.420202] Registered led device: ath9k-phy0::rx

I know, without a stack trace there is not much to do, I really don't know how to get that after my system just stops responding to mouse/keyboard.

[Aaron Kelley (aaronkelley)]

I still haven't been able to catch a stack trace. (I'm assuming one will pop up on the screen if I happen to be in a tty when the crash occurs, but since it is so infrequent, and I haven't been able to reproduce it "on purpose"...)

[Luis R. Rodriguez (mcgrof)]

If the crash was due to ath9k and you are on an SMP system then your issue may be fixed with the latest patches I have posted which fixes this issue in wireless-testing down to 2.6.27. These patches have been posted to be merged into wireless-testing however they have not yet been merged so if you'd like to try then you'll have to apply patches yourself. Note that these patches are not yet merged into wireless-testing but will likely be so as of this week.

The upstream kernel.org bug report is:

http://bugzilla.kernel.org/show_bug.cgi?id=12110

The patches you can use to test:

http://www.kernel.org/pub/linux/kernel/people/mcgrof/patches/ath9k/2009-03-12/serialization-v6/

As far as performance is concerned please keep in mind 802.11n aggregation support was only enabled as of 2.6.29 so you'll need at least compat-wireless for that (but the patches for serialization fix need to yet be applied on top of that).

[Luis R. Rodriguez (mcgrof)]

Can this be closed?

[Bryan Wu (cooloney)]

Is this bug fixed or not?

@Luis, Thanks for your upstream patch, do you think it is ok for us to cherry pick it to our Jaunty Stable Release Updates.

-Bryan

[Bryan Wu (cooloney)]

Assign it to me, I will keep eyes on it.

[Luis R. Rodriguez (mcgrof)]

On Thu, Jun 11, 2009 at 2:28 AM, Bryan Wu<email address hidden> wrote:
> Is this bug fixed or not?
>
> @Luis, Thanks for your upstream patch, do you think it is ok for us to
> cherry pick it to our Jaunty Stable Release Updates.

This should already be part of jaunty, I made sure to inform the
Ubuntu kernel team of the importance of the serialization patches.

On the ath9k wiki we also keep track of known issues and the kernels
when they were fixed. Let me paste the relevant section here:

http://wireless.kernel.org/en/users/Drivers/ath9k
----
Hangs with PCI devices and SMP systems

This issue was due to a hardware limitation on our PCI devices whereby
our device's FIFO queue got filled up and caused a loop. This issue is
only present on our 11n PCI devices, its not present on our
PCI-express devices nor our legacy 802.11abg devices.

The fix was to serialize read/writes – a spin_lock around each read
and write. This fix is not performed for PCI devices or non-SMP
systems. This fix is present in 2.6.29, 2.6.28.10 and 2.6.27.22.
----

In the future just please refer to the wiki page.

  Luis

[maxauthority (stubenschrott)]

With default ubuntu 9.04 I still had hangs, when upgrading to backport-modules (which basically is compat-wireless from end of March) they went away.

Sidenote: I still had random re-/disconnects though with my card, which I just could not accept anymore, as it drove me nuts ;) (no offense, I am not paying for Ubuntu/Linux, and all devs really do great things fixing the bugs!), so i just bought another draft-n usb stick (some with a RaLink chip, rt2870sta driver), that one works great, even with N speeds which somehow i could not get with my atheros PCI card.

[Luis R. Rodriguez (mcgrof)]

On Thu, Jun 11, 2009 at 12:43 PM,
maxauthority<email address hidden> wrote:
> With default ubuntu 9.04 I still had hangs,

Did you read the wiki? 9.04 will ship with a default kernel, that will
never get updated. To get the newer stable release kernels you have no
other option but to upgrade the stable kernel. Jaunty should have
these updated. Keep in mind stable kernels don't automatically trickle
down to the default repository on Ubunut, they first hit the proposed
repository and then eventually hit the main repository.

For details please read:

https://wiki.ubuntu.com/KernelTeam/LinuxWireless

> when upgrading to backport-
> modules (which basically is compat-wireless from end of March) they went
> away.

Upgrading to a stable kernel should fix it too.

  Luis

[Bryan Wu (cooloney)]

@Luis, too bad. currently 2.6.28.10 stable kernel updates are still in the jaunty kernel git tree. we've not upload it to proposed repo for testing. but it might be soon.

Currently, latest 2.6.28-13.44 in jaunty-proposed is 2.6.28.9 stable kernel.
http://kernel.ubuntu.com/~kernel-ppa/info/kernel-version-map.html

@maxauthority: could you please testing our mainline kernel building on your machine and let us know the result.
http://kernel.ubuntu.com/~kernel-ppa/mainline/v2.6.30/

Thanks a lot
-Bryan

[Bryan Wu (cooloney)]

@maxauthority: 2.6.28.10 stable kernel is in jaunty-proposed as 2.6.28-14.46. Could you please test that? Thanks

-Bryan

[Jeremy Foshee (jeremyfoshee)]

This bug report was marked as Incomplete and has not had any updated comments for quite some time. As a result this bug is being closed. Please reopen if this is still an issue in the current Ubuntu release http://www.ubuntu.com/getubuntu/download . Also, please be sure to provide any requested information that may have been missing. To reopen the bug, click on the current status under the Status column and change the status back to "New". Thanks.

[This is an automated message. Apologies if it has reached you inappropriately; please just reply to this message indicating so.]