On Tuesday 26 of May 2009, Rafał Kupka wrote:
> Something very strange happens to signals in your system. Some time
> ago there was similar problem with lost SIGALRM:
> http://www.mail-archive.com/exim-users@exim.org/msg23913.html
> but on your system exim process signals masks looks correct.
I'm currently running exim with this patch and so far it "fired" once per 2 days
"2009-06-03 00:15:52 [19289] daemon: queue-runner not run for 360s. Forcing. Is delivery of SIGALRM broken on this system ?"
Is such patch sane enough for upstream inclusion? Does anyone see bugs
in this approach for different configurations?
--- exim-4.69.org/src/daemon.c 2009-06-01 23:02:02.505119117 +0200
+++ exim-4.69/src/daemon.c 2009-06-01 23:09:58.088404461 +0200
@@ -25,7 +25,7 @@
static smtp_slot empty_smtp_slot = { 0, NULL };
-
+static time_t sigalrm_seen_last;
/*************************************************
* Local static variables *
@@ -1603,6 +1603,8 @@
smtp_input = TRUE;
+time(&sigalrm_seen_last);
+
/* Enter the never-ending loop... */
for (;;)
@@ -1624,6 +1626,8 @@
{
DEBUG(D_any) debug_printf("SIGALRM received\n");
+ time(&sigalrm_seen_last);
+
/* Do a full queue run in a child process, if required, unless we already
have enough queue runners on the go. If we are not running as root, a
re-exec is required. */
@@ -1885,11 +1889,19 @@
else
{
+ int time_diff;
struct timeval tv;
tv.tv_sec = queue_interval;
tv.tv_usec = 0;
select(0, NULL, NULL, NULL, &tv);
handle_ending_processes();
+
+ time_diff = (int)difftime(time(NULL), sigalrm_seen_last);
+ if ((queue_interval > 0) && (time_diff > (2*queue_interval)))
+ {
+ sigalrm_seen = TRUE;
+ log_write(0, LOG_MAIN|LOG_PANIC, "daemon: queue-runner not run for %lds. Forcing. Is SIGALRM delivery broken on this system ?", time_diff);
+ }
}
/* Re-enable the SIGCHLD handler if it has been run. It can't do it
--
Arkadiusz Miśkiewicz PLD/Linux Team
arekm / maven.pl http://ftp.pld-linux.org/
