On Thu, Jan 14, 2021 at 10:52:51PM +0100, Ángel wrote:
> On 2021-01-12 at 18:02 -0500, Sam Varshavchik wrote:
> > It's theoretically possible to run "ps" in a subprocess and kill
> > the gam_server process. That will result in a failed socket
> > connection. But that's not going to be reliable, there will likely be
> > multiple gam_server processes and there is no convenient way to find
> > the right one.
>
> *If* Alexey is running each user under a different uid, it would be
> simple to find the right gam_server which is not working.
Yes, in my case I made this simple patch that solved my problem. Of
course, this is not entirely true as I wanted,
but it's better to get only one error during authorization
and the next time the user tries to log in
we get a work mailbox instead of hangs unclear for how long.
maildirwatch.c
@@ -29,6 +29,14 @@
"Timeout initializing the FAM library. Your FAM library
is broken.\n";
write(2, msg, sizeof(msg)-1);
+
+ uid_t uid;
+ char cmd[255];
+
+ uid = getuid();
+ sprintf(cmd, "/usr/bin/pkill -u %d -x gam_server >/dev/null 2>&1",
uid);
+ system(cmd);
+
kill(getpid(), SIGKILL);
}
#endif
> > stracing it should be useful to forward to gam.
>
> Indeed. strace may be able to show where/why it is (apparently) stuck
> in a loop. Are the maliboxes in some kind of exotic filesystem?
mdraid (0 and 10) on local SSD with XFS
_______________________________________________
Courier-imap mailing list
Courier-imap@lists.sourceforge.net
Unsubscribe: https://lists.sourceforge.net/lists/listinfo/courier-imap
