/mandos/trunk

To get this branch, use:
bzr branch http://bzr.recompile.se/loggerhead/mandos/trunk

« back to all changes in this revision

Viewing changes to mandos-monitor

  • Committer: Teddy Hogeborn
  • Date: 2015-05-23 20:18:34 UTC
  • mto: This revision was merged to the branch mainline in revision 756.
  • Revision ID: teddy@recompile.se-20150523201834-e89ex4ito93yni8x
mandos: Use multiprocessing module to run checkers.

For a long time, the Mandos server has occasionally logged the message
"ERROR: Child process vanished".  This was never a fatal error, but it
has been annoying and slightly worrying, since a definite cause was
not found.  One potential cause could be the "multiprocessing" and
"subprocess" modules conflicting w.r.t. SIGCHLD.  To avoid this,
change the running of checkers from using subprocess.Popen
asynchronously to instead first create a multiprocessing.Process()
(which is asynchronous) calling a function, and have that function
then call subprocess.call() (which is synchronous).  In this way, the
only thing using any asynchronous subprocesses is the multiprocessing
module.

This makes it necessary to change one small thing in the D-Bus API,
since the subprocesses.call() function does not expose the raw wait(2)
status value.

DBUS-API (CheckerCompleted): Change the second value provided by this
                             D-Bus signal from the raw wait(2) status
                             to the actual terminating signal number.
mandos (subprocess_call_pipe): New function to be called by
                               multiprocessing.Process (starting a
                               separate process).
(Client.last_checker signal): New attribute for signal which
                              terminated last checker.  Like
                              last_checker_status, only not accessible
                              via D-Bus.
(Client.checker_callback): Take new "connection" argument and use it
                           to get returncode; set last_checker_signal.
                           Return False so gobject does not call this
                           callback again.
(Client.start_checker): Start checker using a multiprocessing.Process
                        instead of a subprocess.Popen.
(ClientDBus.checker_callback): Take new "connection" argument.        Call
                               Client.checker_callback early to have
                               it set last_checker_status and
                               last_checker_signal; use those.  Change
                               second value provided to D-Bus signal
                               CheckerCompleted to use
                               last_checker_signal if checker was
                               terminated by signal.
mandos-monitor: Update to reflect DBus API change.
(MandosClientWidget.checker_completed): Take "signal" instead of
                                        "condition" argument.  Use it
                                        accordingly.  Remove dead code
                                        (os.WCOREDUMP case).

Show diffs side-by-side

added added

removed removed

Lines of Context:
3
3
4
4
# Mandos Monitor - Control and monitor the Mandos server
5
5
6
 
# Copyright © 2009-2014 Teddy Hogeborn
7
 
# Copyright © 2009-2014 Björn Påhlsson
 
6
# Copyright © 2009-2015 Teddy Hogeborn
 
7
# Copyright © 2009-2015 Björn Påhlsson
8
8
9
9
# This program is free software: you can redistribute it and/or modify
10
10
# it under the terms of the GNU General Public License as published by
173
173
            gobject.source_remove(self._update_timer_callback_tag)
174
174
            self._update_timer_callback_tag = None
175
175
    
176
 
    def checker_completed(self, exitstatus, condition, command):
 
176
    def checker_completed(self, exitstatus, signal, command):
177
177
        if exitstatus == 0:
178
178
            self.logger('Checker for client {} (command "{}")'
179
179
                        ' succeeded'.format(self.properties["Name"],
181
181
            self.update()
182
182
            return
183
183
        # Checker failed
184
 
        if os.WIFEXITED(condition):
 
184
        if exitstatus >= 0:
185
185
            self.logger('Checker for client {} (command "{}") failed'
186
186
                        ' with exit code {}'
187
187
                        .format(self.properties["Name"], command,
188
 
                                os.WEXITSTATUS(condition)))
189
 
        elif os.WIFSIGNALED(condition):
 
188
                                exitstatus))
 
189
        elif signal != 0:
190
190
            self.logger('Checker for client {} (command "{}") was'
191
191
                        ' killed by signal {}'
192
192
                        .format(self.properties["Name"], command,
193
 
                                os.WTERMSIG(condition)))
194
 
        elif os.WCOREDUMP(condition):
195
 
            self.logger('Checker for client {} (command "{}") dumped'
196
 
                        ' core'.format(self.properties["Name"],
197
 
                                       command))
 
193
                                signal))
198
194
        else:
199
195
            self.logger('Checker for client {} completed'
200
196
                        ' mysteriously'