Wireless on Atheros 5213A drops often, is slow, and ath5k_pci consumes large amounts of CPU

Bug #267761 reported by Michael Rooney
38
This bug affects 3 people
Affects Status Importance Assigned to Milestone
linux (Ubuntu)
Fix Released
Medium
Unassigned

Bug Description

Using Intrepid Alpha 5 with an Atheros 5212/5213 (as identified by Hardware Test / lspci) and connecting to a WPA 2 Personal wireless network, the connection drops every 5-10 minutes, and when it is connected, the connection is very slow (bitwise, typically 10-100x slower than with Hardy). Often when reconnecting after dropping it will ask for the WPA key again, which is already saved, and will work after one or two times of just pressing Okay. The process ath5k_pci also consumes around 40-85% CPU during the connection process and also intermittently while the connection exists, making the computer unusable during this time (mouse is not very responsive).

This card worked fine on Gutsy and Hardy and did not experience these problems. This problem is easy to duplicate; I can boot into Hardy and have fast and stable wireless, and then boot again into Intrepid and have the experience described above on the same network. Note that I don't know that this issue is restricted to WPA 2 Personal networks; that is just my home network that I have to test.

Strangely enough this bug existed in Alpha 4, and I say strangely because that was Linux 2.6.26 which should have been using a different driver. Originally after doing a fresh install of Alpha 5, I didn't notice this bug, but now after a restart it seems to be back. I wonder if it is possible that it is using the wrong (old, restricted) driver, somehow? I have attached a screenshot of the Hardware Drivers application, which seems to display conflicting messages in the title and body.

Towards the bottom of dmesg (attached) there is all sorts of repetitive wireless output such as:

[ 601.255022] ath5k phy0: noise floor calibration timeout (2412MHz)
[ 601.255039] ath5k phy0: ath5k_chan_set: unable to reset channel (2412 Mhz)
[ 601.255048] wlan0: failed to set freq to 2412 MHz for scan

and then plenty of drops like:

[ 541.022613] wlan0: deauthenticated
[ 542.020141] wlan0: authenticate with AP 00:07:85:b3:cb:ee
[ 542.030863] wlan0: authenticated
[ 542.030881] wlan0: associate with AP 00:07:85:b3:cb:ee

Intrepid Alpha 5
Linux: 2.6.27.2.2
lspci line: 02:02.0 Ethernet controller: Atheros Communications Inc. AR5212/AR5213 Multiprotocol MAC/baseband processor (rev 01)

Tags: linux-2.6.27
Revision history for this message
Michael Rooney (mrooney) wrote :
Revision history for this message
Leann Ogasawara (leannogasawara) wrote :

Hi Mike,

Can you attach you dmesg output as well. You may also want to test installing the madwifi package that is available in one of our developers PPA's - https://launchpad.net/~timg-tpi/+archive . Thanks.

Changed in linux:
status: New → Incomplete
Revision history for this message
Michael Rooney (mrooney) wrote :

Thanks for helping me debug this Leann. Attached is the dmesg output. I'll see if I can test the madwifi PPA but I'll have to find a wired network to get it in the first place :)

Changed in linux:
status: Incomplete → New
Revision history for this message
Michael Rooney (mrooney) wrote :

I am still experiencing this after upgrading to the Linux 27-4 packages; is the PPA still useful for me to test?

My wireless seems to lag around 10-45kB whereas in Hardy on the same network speeds are closer to 500kB, and I am still noticing periods of ath5k_pci usage around 85% that make the computer unusable for a short period of time.

Revision history for this message
Leann Ogasawara (leannogasawara) wrote :

Hi Mike,

I think it would actually be good if you would still be willing to test the madwifi package in Tim's PPA. I realize that ath5k is intended to replace madwifi going forward, but it would be interesting to know if it provides a stable solution for you in the mean time.

Michael Rooney (mrooney)
description: updated
Revision history for this message
Michael Rooney (mrooney) wrote :

Thanks Leann. I don't see an Intrepid option for that PPA, can I just change Hardy to Intrepid in the sources entry and have it work? I would expect to see Intrepid listed in the drop-down and in the series below, if it was built for Intrepid. Let me know and I can test the madwifi PPA.

Revision history for this message
Michael Rooney (mrooney) wrote :

After updating today (the first time I updated in about a week) and restarting, network-manager doesn't show any wireless networks at all (and there are usually about 4-6 here). How interesting. I will try the madwifi PPA later tonight.

Revision history for this message
Michael Rooney (mrooney) wrote :

Okay, after a second restart ath5k was functional again (though with the same problems as above). I then installed madwifi from svn snapshot and blacklisted ath5k, and my wireless experience is much better! Much faster speeds, no CPU spikes, et cetera. However this bug is probably still good to solve since ath5k is the future anyway, as I understand it.

Revision history for this message
coubi64 (frederic-valentin) wrote :

Same problem for me, logs in a post here: http://forum.ubuntu-fr.org/viewtopic.php?id=256441

I'm surprised by a thing too: lspci give "Atheros Communications Inc. AR5212/AR5213" but dmesg gives "ath5k phy0: Atheros AR2413 chip". Is it normal?

Revision history for this message
Michael Rooney (mrooney) wrote :

Thanks coubi64, I will Confirm thanks to your input.

As an update for Leann, after more use with the madwifi it is an improvement but still a definite regression from Hardy. It is slower and doesn't seem to get as strong of a connection; I have had to reboot into my Hardy install to connect to lower signal networks or get faster speeds.

Changed in linux:
importance: Undecided → Medium
status: New → Confirmed
Revision history for this message
coubi64 (frederic-valentin) wrote :

Found a related bug report here: https://bugs.launchpad.net/ubuntu/+source/linux-restricted-modules/+bug/275423

Still have the problem with the 2.6.27-6 kernel

Revision history for this message
Neil J. Patel (njpatel) wrote :

I'm getting the same issue on a fresh install of Intrepid Beta, fully updated.

Connection keeps dropping, passwords are asked each time and performance of the network is quite bad. lspci shows:

njpatel@njp-laptop:~$ lspci | grep Atheros
0b:00.0 Network controller: Atheros Communications Inc. AR5418 802.11abgn Wireless PCI Express Adapter (rev 01)

I'll try the madwifi drivers and update.

Revision history for this message
dvhart (darren-dvhart) wrote :

Experiencing similar poor networking performance and periodic short-duration freezes with an AR5212 on Intrepid. Mike mentioned building madwifi from svn and blacklisting ath5k. As I understand it, madwifi should still be installed with Intrepid. How do I tell Ubuntu to use the madwifi driver INSTEAD of the ath5k driver? Even if I have to build madwifi, how do I blacklist ath5k ?

Revision history for this message
Russell Phillips (ignissport) wrote :

To blacklist ath5k, at this line to /etc/modprobe.d/blacklist:

blacklist ath5k

Revision history for this message
Leann Ogasawara (leannogasawara) wrote :

Hi Guys,

Care to try installing the linux-backports-modules-intrepid package as it recently pulled in an updated version of the compat-wireless stack and ath5k driver.

Revision history for this message
nclm (nclm) wrote :

The ath5k in backports-modules worked much better than the one we have in the current ubuntu kernel but I think the new compat-wireless stack is making trouble. I had the same backtraces in my dmesg as they were mentioned in https://bugs.launchpad.net/ubuntu/+source/linux-restricted-modules/+bug/275423/comments/10 so I'm back at ath_pci.

Revision history for this message
glepore70 (greg-rhobard) wrote :

For me the new ath5k works much better, however every couple of hours of use I get kicked off the network and have to rmmod ath5k, modprobe ath5k, and restart networking to get it back. I don't know where this might log a bug for tracking. The WiFi Signal Strength Meter Plasmoid is consistently reporting ca. 90% strength, where previously it was around 30%, so that's better.

Revision history for this message
Michael Rooney (mrooney) wrote :

I did a fresh install of Intrepid today and my wireless seems to work very well out of the box. Since I originally reported this I'll mark it as fixed; if it isn't for others it might be a different issue.

Changed in linux:
status: Confirmed → Fix Released
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.