On Dec 29, 2011, at 7:07 PM, Tobias Brunner wrote: > Hi Jan Willem, > > Thanks a lot for the config and test script. > > The lock up seen here is not actually a classic deadlock. It is caused > by too many concurrent calls of "ipsec up". Such calls are synchronous > and thus make the stroke plugin create jobs which each block a thread > from the thread pool waiting for the initiated SA to be established. If > too many of these jobs are created they clog the thread pool and no > other jobs can be handled (e.g. for incoming packets). > > I pushed two commits (see [1] and [2]) which fix this problem by > limiting the number of concurrently handled stroke commands. The second > commit adds an option to strongswan.conf that allows you to set the > number of concurrently handled commands. Depending on the number of > threads in the thread pool you might want to increase the current > default of 4. > > Regards, > Tobias > > [1] http://git.strongswan.org/?p=strongswan.git;a=commitdiff;h=8ff513a8 > [2] http://git.strongswan.org/?p=strongswan.git;a=commitdiff;h=7c0c2349
Hi Tobias, thank you for supplying the patches. This solved a blocking issue for my Masters thesis experiments. So very, very much appreciated! After tweaking the right parameters a bit I've managed to verify that the script I sent you now works without deadlock/pool exhaustion. For those googlers finding this thread useful: Bringing the connections down right/soon after initiation also mitigates this 'bug' it seems. Kind regards, Jan Willem Beusink _______________________________________________ Dev mailing list [email protected] https://lists.strongswan.org/mailman/listinfo/dev
