kernel oops results in apport-gtk launching for every .crash file

[Brian Murray]

I've been testing multiple different kernels for bug 1218004, as a part of this process my system frequently experiences a kernel oops. Only occasionally, is a kernel oops crash file created for these crashes. This particular time the following happened:

Sep 24 09:47:42 localhost kernel: [ 330.255893] ubuntuone-syncd[2596]: segfault at 20 ip 00007f1e907d3300 sp 00007fff91549050 error 6 in libc-2.17.so[7f1e90752000+1bd000]
Sep 24 09:47:54 localhost kernel: [ 341.887110] BUG: unable to handle kernel NULL pointer dereference at 0000000000000070
Sep 24 09:47:54 localhost kernel: [ 341.887197] IP: [<ffffffff811fbfd7>] show_vfsmnt+0x37/0x150
Sep 24 09:47:54 localhost kernel: [ 341.887242] PGD 72b78067 PUD 6d306067 PMD 0
Sep 24 09:47:54 localhost kernel: [ 341.887276] Oops: 0000 [#1] SMP

It looks like ubuntuone-syncd crashed and shortly after that I experienced a kernel oops. Looking in /var/crash we can see the following:

 $ ls -lctr /var/crash/
total 39532
-rw-r----- 1 root whoopsie 7141 Sep 18 10:10 linux-image-3.9.0-030900-generic.229283.crash
-rw-r--r-- 1 bdmurray whoopsie 0 Sep 18 12:00 _usr_bin_xeyes.1000.upload
-rw-r--r-- 1 bdmurray whoopsie 0 Sep 18 12:02 _usr_bin_d-feet.1000.upload
-rw------- 1 whoopsie whoopsie 0 Sep 18 13:47 _usr_bin_xeyes.1000.uploaded
-rw------- 1 whoopsie whoopsie 0 Sep 18 13:47 _usr_bin_d-feet.1000.uploaded
-rw-r----- 1 bdmurray whoopsie 33668113 Sep 24 09:53 _usr_lib_ubuntuone-client_ubuntuone-syncdaemon.1000.crash
-rw-r----- 1 bdmurray whoopsie 105043 Sep 24 09:53 _usr_lib_firefox_firefox.1000.crash
-rw-r----- 1 bdmurray whoopsie 1770585 Sep 24 09:53 _usr_bin_xeyes.1000.crash
---------- 1 bdmurray whoopsie 50209 Sep 24 09:53 _usr_bin_gnome-session.1000.crash
---------- 1 bdmurray whoopsie 952256 Sep 24 09:53 _usr_bin_easystroke.1000.crash
-rw-r----- 1 bdmurray whoopsie 3851246 Sep 24 09:53 _usr_bin_d-feet.1000.crash

Something caused every crash file's modification time to change. The end result of this being that on the next reboot I received an apport-gtk notification for every crash, even ones that I experienced some time ago and that have already been uploaded. While this is an unlikely, it is rather annoying.

[Brian Murray (brian-murray)]

Looking at the upstart user job log files we see:

 $ cat ~/.cache/upstart/update-notifier-crash-_var_crash__usr_lib_firefox_firefox.1000.crash.log
Tue Sep 24 09:52:46 PDT 2013 crash report /var/crash/_usr_lib_firefox_firefox.1000.crash detected
ubuntuone-syncdaemon

 $ cat update-notifier-crash-_var_crash__usr_bin_xeyes.1000.crash.log
Tue Sep 24 09:52:46 PDT 2013 crash report /var/crash/_usr_bin_xeyes.1000.crash detected
ubuntuone-syncdaemon

Most of the other crash.log files look the same except for ubuntuone-syncdaemon.

 $ zcat update-notifier-crash-_var_crash__usr_lib_ubuntuone-client_ubuntuone-syncdaemon.1000.crash.log.1.gz
Tue Sep 24 09:48:04 PDT 2013 crash report /var/crash/_usr_lib_ubuntuone-client_ubuntuone-syncdaemon.1000.crash detected
ubuntuone-syncdaemon

(apport-gtk:2718): Gtk-WARNING **: Theme parsing error: gtk-widgets.css:1605:38: '' is not a valid color name
Tue Sep 24 09:52:46 PDT 2013 crash report /var/crash/_usr_lib_ubuntuone-client_ubuntuone-syncdaemon.1000.crash detected
ubuntuone-syncdaemon

One is from before the reboot after the kernel oops and the other is after it.

The contents of the linux crash log file are also interesting:

 $ cat update-notifier-crash-_var_crash_linux-image-3.9.0-030900-generic.229283.crash.log
Tue Sep 24 09:52:46 PDT 2013 crash report /var/crash/linux-image-3.9.0-030900-generic.229283.crash detected
ubuntuone-syncdaemon

so the linux*.crash file was detected but it remains unmodified for some reason.

[Brian Murray (brian-murray)]

The update-notifier crash notification job calls apport-checkreports before launching apport-gtk, so I'd think that if apport-checkreports were the problem just running it would cause the .crash files to be modified. It does not so I think this is actually unrelated to update-notifier and may be an upstart issue.