weird 8.2.1 crash
Paul A Vixie
vixie at mibh.net
Mon Aug 9 20:06:57 UTC 1999
here's the diff if you want to just fix this without upgrading to 8.2.2:
Index: src/include/isc/list.h
===================================================================
RCS file: /proj/cvs/isc/bind/src/include/isc/list.h,v
retrieving revision 8.2
retrieving revision 8.3
diff -u -r8.2 -r8.3
--- src/include/isc/list.h 1999/01/08 19:23:05 8.2
+++ src/include/isc/list.h 1999/08/08 02:36:47 8.3
@@ -66,6 +66,7 @@
(elt)->link.prev->link.next = (elt)->link.next; \
else \
(list).head = (elt)->link.next; \
+ INIT_LINK(elt, link); \
} while (0)
#define PREV(elt, link) ((elt)->link.prev)
(i'm sending this to bind-workers rather than just jim in case others care.)
> To: Paul A Vixie <vixie at mibh.net>
> Subject: Re: weird 8.2.1 crash
> In-reply-to: Your message of "Mon, 09 Aug 1999 10:49:20 PDT."
> <199908091749.KAA04829 at bb.rc.vix.com>
> Date: Mon, 09 Aug 1999 20:27:50 +0200
> Message-ID: <29883.934223270 at kludge.mpn.cp.philips.com>
> From: Jim Reid <jim at mpn.cp.philips.com>
>
> >>>>> "Paul" == Paul A Vixie <vixie at mibh.net> writes:
>
> >> Has anyone got an idea how nzones got to >> 135 million? There are
> >> only 784 zones in its named.conf.
>
> Paul> this sounds exactly like a bug in 8.2.1 that has to do with
> Paul> zone removal and which is supposed to be fixed in 8.2.2.
>
> Thanks. We did *add* a new zone earlier in the day before the server
> crashed. I thought that might have been significant - "ndc reload
> new-zone" - but when the data structs were still messed up after
> restarting the name server, I had my doubts.... FWIW I couldn't
> reproduce the crash after an "ndc reload new-zone" on a test
> system. It's core image looked OK. Maybe it depends on the number of
> zones in named.conf? The test system I tried only had 10 zones on it,
> the production one had 700 or so. I'll try adding a new zone to a test
> system with that number of zones and see what happens. And compare the
> result with 8.2.2T1A while I'm at it.
>
> 8.2.2's just a wee bit too raw for going on production servers here,
> but I'll keep this in mind. Thanks again.
>
More information about the bind-workers
mailing list