mirrors/haproxy
1
0
Fork 0
mirror of https://git.haproxy.org/git/haproxy.git/ synced 2026-02-10 20:01:20 +01:00
Code Issues Projects Releases Wiki Activity

BUG/MAJOR: fix haproxy crash when using server tracking instead of checks

Browse Source 
Igor at owind reported a very recent bug (just present in latest snapshot).
Commit "4a741432 MEDIUM: Paramatise functions over the check of a server"
causes up/down to die with tracked servers due to a typo.

The following call in set_server_down causes the server to put itself
down recurseively because "check" is the current server's check, so once
fed to the function again, it will pass through the exact same path (note
we have the exact symmetry in set_server_up) :

	for (srv = s->tracknext; srv; srv = srv->tracknext)
		if (!(srv->state & SRV_MAINTAIN))
			/* Only notify tracking servers that are not already in maintenance. */
			set_server_down(check);

Instead we should stop the tracking server being visited in the loop :

	for (srv = s->tracknext; srv; srv = srv->tracknext)
		if (!(srv->state & SRV_MAINTAIN))
			/* Only notify tracking servers that are not already in maintenance. */
			set_server_down(&srv->check);

But that's not exactly enough because srv->check->server is only set when
checks are enabled, so ->server is NULL for tracking servers, still causing a
crash upon first iteration. The fix is easy and consists in always initializing
check->server when creating a new server, which is what was already done a few
patches later by 69d29f9 (MEDIUM: cfgparse: Factor out check initialisation).

With the fix above alone on top of current version or snapshot 20131122, the
problem disappears.

Thanks to Igor for testing and reporting the issue.
This commit is contained in:
Willy Tarreau 2013-11-27 16:52:23 +01:00
parent 86a446e685
commit bc16cd81c4
1 changed files with 6 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
src/checks.c
View File

@ -454,10 +454,10 @@ void set_server_down(struct check *check)
s->counters.down_trans++;
if (s->state & SRV_CHECKED)
for(srv = s->tracknext; srv; srv = srv->tracknext)
if (! (srv->state & SRV_MAINTAIN))
for (srv = s->tracknext; srv; srv = srv->tracknext)
if (!(srv->state & SRV_MAINTAIN))
/* Only notify tracking servers that are not already in maintenance. */
set_server_down(check);
set_server_down(&srv->check);
}
check->health = 0; /* failure */
@ -530,10 +530,10 @@ void set_server_up(struct check *check) {
send_log(s->proxy, LOG_NOTICE, "%s.\n", trash.str);
if (s->state & SRV_CHECKED)
for(srv = s->tracknext; srv; srv = srv->tracknext)
if (! (srv->state & SRV_MAINTAIN))
for (srv = s->tracknext; srv; srv = srv->tracknext)
if (!(srv->state & SRV_MAINTAIN))
/* Only notify tracking servers if they're not in maintenance. */
set_server_up(check);
set_server_up(&srv->check);
}
if (check->health >= check->rise)