https://bugzilla.mindrot.org/show_bug.cgi?id=2641
Luca Boccassi <[email protected]> changed: What |Removed |Added ---------------------------------------------------------------------------- Attachment #3802|0 |1 is obsolete| | --- Comment #27 from Luca Boccassi <[email protected]> --- Created attachment 3804 --> https://bugzilla.mindrot.org/attachment.cgi?id=3804&action=edit standalone notify and timestamp patch > That's more tricky as the reload is called from signal handler context and we > can't use snprint() there to format the usec part of the message. We'd have > to refactor how sshd manages SIGHUP restarts. > > That would make some other things easier, but it's still a bigger change. I went back and had a look at this, and unless I am missing something the reloading message is not being sent from the signal handler? The handler is sighup_handler which just sets a boolean and returns, following the usual pattern: https://anongit.mindrot.org/openssh.git/tree/sshd.c#n298 but the notification message is sent from the platform_pre_restart() hook, which is called from the main context from the main loop via sighup_restart(): https://anongit.mindrot.org/openssh.git/tree/sshd.c#n304 This already does some logging, which uses format strings. Also platform_pre_restart() already calls oom_adjust_restore() which also uses format strings. So I went ahead and did the necessary modifications in the latest version, which also simplified the message handling as it can log unconditionally now, and added the timestamp too. I've tested this and seems to work just fine on Debian testing, I can change ssh.service to Type=notify-reload and reloading works just fine, including the state transitions. -- You are receiving this mail because: You are watching someone on the CC list of the bug. You are watching the assignee of the bug. _______________________________________________ openssh-bugs mailing list [email protected] https://lists.mindrot.org/mailman/listinfo/openssh-bugs
