* Timo Sirainen, 2010-01-02 17:48 > On 2.1.2010, at 3.45, Thomas Zajic wrote: > >> [<c1095367>] ? inotify_free_group_priv+0x12/0x21 >> [<c1093c81>] ? fsnotify_final_destroy_group+0x16/0x1e > .. >> There are no processes other than dovecot's "imap" that are ever causing >> these warnings. Does this indicate a problem in the kernel or rather in >> dovecot? > > Yes.
Hehe ... okay, so which one is it? :-) I don't even know what exactly to make of this warning - obviously, there may be a potential memory leak related to inotify and dovecot's imap process, but is it 1. a bug in the kernel that is triggered by something that dovecot's inotify code does, or 2. a bug in dovecot's imap that the kernel notices and warns about as of 2.6.31 (but didn't notice and/or warn about in 2.6.30 and earlier)? > The reason why only imap process creates these is because it's probably > the only process using inotify. You could of course compile Dovecot > without inotify support, but that's just a workaround.. Hmmm ... I don't mind the warning as such, as it doesn't have any actual bad side effects as far as I can tell, I'm just curious as to what causes it. Is there anything related I should/could tune in /proc/sys/fs/inotify? [r...@airframe]:~# grep -r . /proc/sys/fs/inotify /proc/sys/fs/inotify/max_user_instances:128 /proc/sys/fs/inotify/max_user_watches:65536 /proc/sys/fs/inotify/max_queued_events:16384 [r...@airframe]:~# Bye, Thomas