https://bugzilla.mindrot.org/show_bug.cgi?id=2641
--- Comment #8 from Jakub Jelen <jje...@redhat.com> --- (In reply to Darren Tucker from comment #7) > (In reply to Jakub Jelen from comment #6) > > Created attachment 2950 [details] > > fixed patch > > > > Never mind. Nothing from above resolves the race condition between > > systemd reading PID file and sshd after reexec writing it, except > > adding SD_NOTIFY code so I gave up. > > What about reading the pidfile first to see if it has the correct > PID before rewriting it? I did something like that to work around a > problem in pam_loginuid (LinuxPAM ticket #23, I'd link to it but > fedorahosted.org seems to have imploded). I don't think that would help the initial race condition, when systemd tries to read the PID file before it is written after the daemon(). > Does systemd even need a pidfile? >From man systemd.service: > If this setting [Type=forking] is used, it is recommended to also use the > PIDFile= option, so that systemd can identify the main process of the daemon. > systemd will proceed with starting follow-up units as soon as the parent > process exits. There is also GuessMainPID= option, but from the documentation in the same manual page I am not convinced that it is something that we would like to use in case we can specify the PID file reliably. -- You are receiving this mail because: You are watching the assignee of the bug. You are watching someone on the CC list of the bug. _______________________________________________ openssh-bugs mailing list openssh-bugs@mindrot.org https://lists.mindrot.org/mailman/listinfo/openssh-bugs