Bug 457067 - iwl3945 on Dell Latitude D620: keyboard does not respond after resume
Summary: iwl3945 on Dell Latitude D620: keyboard does not respond after resume
Keywords:
Status: CLOSED CURRENTRELEASE
Alias: None
Product: Fedora
Classification: Fedora
Component: kernel
Version: 9
Hardware: i686
OS: Linux
low
medium
Target Milestone: ---
Assignee: Kernel Maintainer List
QA Contact: Fedora Extras Quality Assurance
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2008-07-29 14:47 UTC by Matt McCutchen
Modified: 2008-10-15 00:23 UTC (History)
1 user (show)

Fixed In Version: kernel-2.6.26.2-14.fc9.i686
Clone Of:
Environment:
Last Closed: 2008-10-15 00:23:32 UTC
Type: ---
Embargoed:


Attachments (Terms of Use)
/var/log/messages output (422.70 KB, application/octet-stream)
2008-07-29 14:51 UTC, Matt McCutchen
no flags Details

Description Matt McCutchen 2008-07-29 14:47:29 UTC
Description of problem:
I am using iwl3945 on a Dell Latitude D620.  When NetworkManager tries to
connect to a wireless network after the computer resumes from suspend-to-RAM,
the keyboard stops responding (though the mouse still works) and the "events/1"
process uses 100% CPU.  If I turn off the wireless kill switch on the side of
the computer, the keyboard and CPU usage eventually return to normal, but when I
turn the switch back on, the bad behavior resumes.  Removing and reinserting the
iwl3945 module usually fixes the problem until the next suspend-to-RAM.

Version-Release number of selected component (if applicable):
kernel-2.6.25.11-97.fc9.i686
NetworkManager-0.7.0-0.11.svn3846.fc9.i386

How reproducible:
With perhaps 50% probability on each resume.

Steps to Reproduce:
1. Configure NetworkManager to automatically connect to a wireless network.
2. Run "pm-suspend" in a gnome-terminal.  (I'm not using the GNOME Suspend
button because of a separate issue.)
3. Make sure the kill switch is on.  Press the power button to resume the
computer and enter any necessary BIOS and gnome-screensaver passwords.
  
Actual results:
NetworkManager is stuck forever in the one-green-dot stage, occasionally
prompting for wireless secrets.  Pressing keys on the keyboard has no effect,
though the mouse works.  "top" (painstakingly started by copying the letters
from gucharmap :) indicates that "events/1" is using 100% CPU.

Expected results:
The keyboard works, CPU usage is reasonable, and NetworkManager connects to the
network after a few tries/prompts.

Additional info:
- I got some stack traces in /var/log/messages, which I will attach.
- The problem does not seem to occur with kernel-2.6.25.11-92.fc9.i686.

Comment 1 Matt McCutchen 2008-07-29 14:51:16 UTC
Created attachment 312879 [details]
/var/log/messages output

Note the three "INFO: task XXX:### blocked for more than 120 seconds" stack
traces near the bottom.

Comment 2 David 2008-07-29 22:15:20 UTC
Kyle has got two more kernels built over on koji, why not try these?

http://koji.fedoraproject.org/koji/buildinfo?buildID=57826   (2.6.25.13-103.fc9)

http://koji.fedoraproject.org/koji/buildinfo?buildID=57445   (2.6.25.12-100.fc9)

The .13 kernel appeared only the other day.



Comment 3 Matt McCutchen 2008-07-30 18:38:27 UTC
(In reply to comment #0)
> How reproducible:
> With perhaps 50% probability on each resume.

More specifically, suspending while NetworkManager is trying to connect to a
network (in the one-green-dot stage) appears to trigger the problem.

Comment 4 Matt McCutchen 2008-07-30 18:59:13 UTC
I can reproduce the problem with both kernel-2.6.25.12-100.fc9.i686 and
kernel-2.6.25.13-103.fc9.i686, but it didn't occur immediately after resume; I
had to disconnect and reconnect NetworkManager to trigger it.

What would help you fix this?  I'm no good at kernel hacking, but I can test
things / bisect.

Comment 5 Matt McCutchen 2008-07-31 14:19:31 UTC
I can reproduce the problem with kernel-2.6.25.13-104.fc9.i686.  I suspended
while NetworkManager was connecting; I resumed; NetworkManager tried and failed
to automatically connect; and I canceled the wireless secrets dialog and
selected the network again on the nm-applet menu.  Only then the keyboard
stopped responding and the "events" process started using 100% CPU.

Comment 6 Matt McCutchen 2008-08-03 18:24:34 UTC
I can still reproduce the problem with kernel-2.6.25.14-106.fc9.i686 .

I also tried bisecting with the vanilla kernel.  Kernel 94ad374a0751f40d25e22e036c37f7263569d24c has the problem and bce7f793daec3e65ec5c5705d2457b81fe7b5725 doesn't, but I ran into a separate problem with suspend-to-RAM that prevented testing of other versions.  I will "bisect" between kernel-2.6.25.11-92.fc9.i686 and kernel-2.6.25.11-97.fc9.i686 instead.

Comment 7 Dan Scholnik 2008-08-21 19:52:53 UTC
I (on a Dell Latitude D630) and a colleague (on a Thinkpad T61) are having a variation on this problem.  We can get the wireless to work after a reboot, but at some point (often after suspend to RAM, but doesn't seem entirely deterministic) a new connection attempt will fail.  (At least sometimes with the "events" process using 100% cpu.)  Removing and then reinserting the iwl3945 module at this point results in total keyboard lockup requiring a hard reboot.  I'm not actually sure what kernel version this started at, but kernel-2.6.25.11-92.fc9.x86_64 seems about the right time frame.  I wasn't using wireless much so I didn't see the problem for a while.

BTW - we mostly do network connections "by hand", not with NetworkManager.

Comment 8 Matt McCutchen 2008-08-21 19:59:18 UTC
The problem appears to be gone with kernel-2.6.26.2-14.fc9.i686 .

Comment 9 Matt McCutchen 2008-10-15 00:23:32 UTC
This problem has been fixed for a while.


Note You need to log in before you can comment on or make changes to this bug.