Ubuntu
apport package

can't boot after synaptic segmentation fault (pb with apport ?)

Bug #58873 reported by samotham
2
Affects Status Importance Assigned to Milestone
apport (Ubuntu)
Fix Released
High
Martin Pitt

Bug Description

Binary package hint: apport

following a synaptic segmentation fault, my ubuntu/edgy was going insane (mouse was kinda frozen, nothing was responding...). I managed to switch to some console. 'top' command indicated that there was loads of 'apport' running, eating all the ressources. (console mode was very very slow). Then from now, i can't do anything, even if i reboot cause it gets kinda frozen very fast.
I checked some /var/log files, i found this one, which can help, i hope.

[code]
apport (pid 18586) Sat Sep 2 16:40:00 2006: called with: ['/usr/share/apport/apport', '18514', '11', 'core'], using core file /home/thomas/core
apport (pid 18586) Sat Sep 2 16:40:21 2006: wrote report /var/crash/_usr_sbin_synaptic.0.crash for package synaptic 0.57.11ubuntu2
apport (pid 31215) Sat Sep 2 20:23:11 2006: called with: ['/usr/share/apport/apport', '22533', '11', 'core'], using core file /home/thomas/core
apport (pid 31215) Sat Sep 2 20:24:43 2006: wrote report /var/crash/_usr_lib_firefox_firefox-bin.1000.crash for package firefox 1.99+2.0b1+dfsg-1ubuntu3
apport (pid 6104) Sun Sep 3 20:02:33 2006: called with: ['/usr/share/apport/apport', '5897', '11', 'core'], using core file /home/thomas/core
apport (pid 6104) Sun Sep 3 20:03:54 2006: wrote report /var/crash/_usr_lib_firefox_firefox-bin.1000.crash for package firefox 1.99+2.0b1+dfsg-1ubuntu3
apport (pid 5506) Mon Sep 4 10:35:26 2006: called with: ['/usr/share/apport/apport', '5504', '11', 'core'], using core file /home/thomas/core
apport (pid 6302) Mon Sep 4 15:45:07 2006: called with: ['/usr/share/apport/apport', '6300', '11', 'core'], using core file /home/thomas/core
apport (pid 6311) Mon Sep 4 15:45:10 2006: called with: ['/usr/share/apport/apport', '6309', '11', 'core'], using core file /core
apport (pid 6302) Mon Sep 4 15:45:27 2006: apport: report /var/crash/_usr_sbin_synaptic.0.crash already exists and unseen, doing nothing to avoid disk usage DoS
apport (pid 6326) Mon Sep 4 15:45:27 2006: called with: ['/usr/share/apport/apport', '6311', '11', 'core'], using core file /core
apport (pid 6329) Mon Sep 4 15:45:27 2006: called with: ['/usr/share/apport/apport', '6326', '11', 'core'], using core file /core
apport (pid 6332) Mon Sep 4 15:45:27 2006: called with: ['/usr/share/apport/apport', '6329', '11', 'core'], using core file /core
apport (pid 6335) Mon Sep 4 15:45:28 2006: called with: ['/usr/share/apport/apport', '6332', '11', 'core'], using core file /core
apport (pid 6338) Mon Sep 4 15:45:28 2006: called with: ['/usr/share/apport/apport', '6335', '11', 'core'], using core file /core
...
[/code]

following those, i have more than **5000** of lines similar to this one.

Each time i want to boot, this file is getting filled very fast, while the system is very slow (i can't use x and console mode is very very slow...)

Revision history for this message
Martin Pitt (pitti) wrote :

So apport somehow crashes consecutively and is called on itself. I'll try to create a reproducible case for this and will add some protections against this case.

Changed in apport:
assignee: nobody → pitti
importance: Untriaged → High
Martin Pitt (pitti)
Changed in apport:
status: Unconfirmed → Fix Committed
Revision history for this message
Martin Pitt (pitti) wrote :

I believe this version should prevent recursive apport calls. Unfortunately I was unable to reproduce this situation, so can you please report back if it still screws up? Thank you!

 apport (0.20) edgy; urgency=low
 .
   * apport: Renice ourself to priority 5 to not slow down the user's processes
     so heavily.
   * Add manpages for apport-retrace(1) and apport-unpack(1) and install them
     into apport. Closes: LP#58463
   * problem_report.py: Test attaching two files instead of one in the
     test_write_file() regression check to assert correct key sorting.
   * problem_report.py: Alter write() method to sort binary data to the end of
     the report. This makes reports easier to read, and also shows relevant
     information more quickly when progressively loading them in a web browser.
     Adapt regression tests accordingly.
   * Move setting of ExecutablePath from apport's InformationCollector ctor to
     apport_utils' report_add_proc_info(), where it belongs to. Check
     ExecutablePath in apport_utils' regression tests.
   * apport-unpack: Support '-' as report argument to read from stdin.
   * apport_utils.py, report_add_proc_info():
     - Apply some heuristics to determine whether the crashed process is an
       interpreted script (check if the Name in /proc/pid/status matches
       the second /proc/pid/cmdline part, and if that command line argument is
       an existing executable file). In the case of an interpreted script, set
       ExecutablePath to the script and InterpreterPath to the actually crashed
       ELF binary.
     - Test this with a shell (/bin/zgrep) and a Python (./apport-unpack)
       script in the test suite.
     - Closes: LP#58859
   * Add debian/apport.logrotate to add a daily 7-step /var/log/apport
     log rotation.
   * test-apport: Fix WCOREDUMP() and pidof checks in check_crash().
   * apport: Install a signal handler for all 'crashy' signals, which just logs
     the signal and stack info and exits. This should avoid a crashing apport
     examining itself, possibly in an endless loop. Closes: LP#58873

Changed in apport:
status: Fix Committed → Fix Released
Revision history for this message
samotham (samotham) wrote :

unfortunately, i won't be able to help anymore,
cause of this error, my computer was completely stuck, keyboard nearly not responding, cpu 100% busy...
I couldn't do anything else than reinstalling ubuntu.
That's what i did...

I had checked and pasted the var/log outputs with a live cd.

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.
Subscribing...

Other bug subscribers

Remote bug watches

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