NetworkManager uses 100% CPU with bcm43xx

[David Erosa]

Binary package hint: network-manager

Just installed Feisty on my iBook G4 and upgraded to Gutsy (using update-manager).

Using Feisty, network-manager worked fine with both my wired and wireless (bcm43xx) connection. After upgrading to Gutsy (network-manager 0.6.5-0ubuntu16), NetworkManager process takes 100% CPU and nm-applet shows no device and disconnected status. Using "htop", I straced NetworkManager and found this:

...
sigreturn() = ? (mask now [ILL TRAP ABRT BUS FPE USR1 USR2 PIPE ALRM TERM STKFLT CONT STOP IO])
--- SIGSEGV (Segmentation fault) @ 0 (0) ---
sigreturn() = ? (mask now [ILL TRAP ABRT BUS FPE USR1 USR2 PIPE ALRM TERM STKFLT CONT STOP IO])
--- SIGSEGV (Segmentation fault) @ 0 (0) ---
sigreturn() = ? (mask now [ILL TRAP ABRT BUS FPE USR1 USR2 PIPE ALRM TERM STKFLT CONT STOP IO])
--- SIGSEGV (Segmentation fault) @ 0 (0) ---
sigreturn() = ? (mask now [ILL TRAP ABRT BUS FPE USR1 USR2 PIPE ALRM TERM STKFLT CONT STOP IO])
--- SIGSEGV (Segmentation fault) @ 0 (0) ---
...

So, I killed it (SIGKILL) and straced it from the beginning, I'll attach the log.

If I rmmod bcm43xx and restart network manager with /etc/dbus-1/event.d/25NetworkManager restart, everything seems to work fine, but if I modprobe bcm43xx again (without restarting nm), the CPU goes 100% again and this is shown on syslog:

Oct 18 00:51:27 peque kernel: [ 2619.734134] bcm43xx driver
Oct 18 00:51:27 peque NetworkManager: <debug> [1192661487.280961] nm_hal_device_added(): New device added (hal udi is '/org/freedesktop/Hal/devices/net_00_11_24_90_13_b2').
Oct 18 00:51:27 peque NetworkManager: <WARN> new_nl_handle(): new_nl_handle: couldn't connecto to netlink: bind() failed (errno = Operation not permitted)
Oct 18 00:51:30 peque NetworkManager: <info> eth1: Device is fully-supported using driver 'bcm43xx'.
Oct 18 00:51:30 peque NetworkManager: <info> nm_device_init(): waiting for device's worker thread to start
Oct 18 00:51:30 peque NetworkManager: <info> nm_device_init(): device's worker thread started, continuing.
Oct 18 00:51:30 peque NetworkManager: <info> Now managing wireless (802.11) device 'eth1'.
Oct 18 00:51:30 peque NetworkManager: <info> Deactivating device eth1.
Oct 18 00:51:30 peque NetworkManager: <WARN> new_nl_handle(): new_nl_handle: couldn't connecto to netlink: bind() failed (errno = Operation not permitted)
Oct 18 00:51:30 peque NetworkManager: <WARN> nm_signal_handler(): Caught signal 11. Generating backtrace...
Oct 18 00:51:30 peque NetworkManager: ******************* START **********************************
Oct 18 00:51:30 peque kernel: [ 2623.391981] ADDRCONF(NETDEV_UP): eth1: link is not ready
Oct 18 00:51:31 peque NetworkManager: Using host libthread_db library "/lib/libthread_db.so.1".
Oct 18 00:51:31 peque kernel: [ 2623.742305] SoftMAC: Open Authentication completed with 00:11:f5:5d:5c:97
Oct 18 00:51:31 peque kernel: [ 2623.754752] ADDRCONF(NETDEV_CHANGE): eth1: link becomes ready
Oct 18 00:51:31 peque NetworkManager: [Thread debugging using libthread_db enabled]
Oct 18 00:51:31 peque NetworkManager: [New Thread 805502976 (LWP 6375)]
Oct 18 00:51:31 peque NetworkManager: [New Thread 822281424 (LWP 6485)]
Oct 18 00:51:31 peque NetworkManager: [New Thread 813892816 (LWP 6381)]
Oct 18 00:51:31 peque NetworkManager: 0x0fbff994 in ?? () from /lib/libpthread.so.0
Oct 18 00:51:31 peque NetworkManager: ******************* END **********************************
Oct 18 00:51:32 peque avahi-daemon[5301]: Registering new address record for fe80::211:24ff:fe90:13b2 on eth1.*.
Oct 18 00:51:41 peque kernel: [ 2633.832799] eth1: no IPv6 routers present

And NetworkManager goes to the "eternal-SIGSEGV-status" again.

[David Erosa (erosa)]

Here's the strace. It was originally sized a few hundred MB, so I chopped it down to a reasonable size (the rest of the file was a repetition of the SIGSEGV).

[Ben Sandee (tbsandee)]

Is this a duplicate of bug 83623?

[Ben Sandee (tbsandee)]

Hmm, no it looks like this problem is not specific to coming out of suspend or hibernate.

[David Erosa (erosa)]

I dont think this is a duplicated bug. As you said, it's not happening adter suspending or hibernating.
I'm doing a fresh gutsy install (gutsy-powerpc, 16102007) to see if the feisty upgrade was the responsable of this behavior.

[David Erosa (erosa)]

Just finished the gutsy installation and the results are:

- still 100% CPU for NetworkManager
- Altought the nic is plugged, NM doesn't wake it up (no IP)
- nm-applet informs that there are no devices.

Any hint I could try?

[Alexander Sack (asac)]

this looks like a duplicate of bug 145683. please test the package named in bug summary.

Thanks,
 - Alexander

[David Erosa (erosa)]

The test package of bug 145683 solved the problem.