[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#776994: Random DNS lookup failure when mixing IPV6 and IPV4 resolver and using rotate option on /etc/resolv.conf



Package: eglibc
Version: 2.13-38+deb7u7

When setting a Postfix server for outgoing mail using a very simple
configuration, I had issues resolving MX entries on the postfix logs like :

Feb  3 20:11:06 londo postfix/smtp[4288]: 3CB504BE8: to=<foo@bar.com>,
relay=none, delay=0.06, delays=0.04/0.01/0/0, dsn=4.4.3, status=deferred
(Host or domain name not found. Name service error for name=bar.com
type=MX: Host not found, try again).

This resolving issue was random (About one of the two tests)

The configuration in /etc/resolv.conf was the following :

nameserver <IPV6 resolver>
nameserver <IPV4 resolver>
nameserver <IPV4 resolver>
options timeout:1 attempts:3 rotate

The three resolvers are OK : I queried them independently and they
answer perfectly so it's not a resolver issue.

After lots of investigations, I isolated the following behaviour : There
is no resolving issue when removing the rotate option OR when putting
the IPV6 resolver after the IPV4 ones.

Browsing on the Web, a chat partner (Thanks Stephane) found this bug on
Redhat/Fedora distributions :
https://bugzilla.redhat.com/show_bug.cgi?id=841787 which seems to be
similar to my bug.

A patch has been applied to the Fedora / RH distributions. This patch is
described here:
http://pkgs.fedoraproject.org/cgit/glibc.git/tree/glibc-rh841787.patch?h=f17.

I updated the patch for the eglibc since a part of the patch was already
into the debian patches series (any/submitted-resolv-init.diff) and
recompiled a package.

After installation, it seems that the issue was solved : The DNS
resolution was 100 % OK.

I send the patch as attachement. (I'm not a low level C developper so i
hope this patch is not too ugly)

Thanks !

Regards,
Olivier Bonhomme






diff -rup a/resolv/res_init.c b/resolv/res_init.c
--- a/resolv/res_init.c 2012-07-26 15:10:45.655638776 -0600
+++ b/resolv/res_init.c 2012-07-26 15:11:27.731423002 -0600
@@ -314,9 +314,9 @@ __res_vinit(res_state statp, int preinit
                        cp++;
                    if ((*cp != '\0') && (*cp != '\n')
                        && __inet_aton(cp, &a)) {
-                       statp->nsaddr_list[nservall].sin_addr = a;
-                       statp->nsaddr_list[nservall].sin_family = AF_INET;
-                       statp->nsaddr_list[nservall].sin_port =
+                       statp->nsaddr_list[nserv].sin_addr = a;
+                       statp->nsaddr_list[nserv].sin_family = AF_INET;
+                       statp->nsaddr_list[nserv].sin_port =
                                htons(NAMESERVER_PORT);
                        nserv++;
 #ifdef _LIBC
diff -rup a/resolv/res_send.c b/resolv/res_send.c
--- a/resolv/res_send.c 2010-05-04 05:27:23.000000000 -0600
+++ b/resolv/res_send.c 2012-07-26 15:34:58.398261659 -0600
@@ -421,10 +421,10 @@ __libc_res_nsend(res_state statp, const
                                EXT(statp).nsmap[n] = MAXNS;
                        }
                }
-               n = statp->nscount;
-               if (statp->nscount > EXT(statp).nscount)
+               n = statp->nscount - EXT(statp).nscount6;
+               if (n > EXT(statp).nscount)
                        for (n = EXT(statp).nscount, ns = 0;
-                            n < statp->nscount; n++) {
+                            n < statp->nscount - EXT(statp).nscount6; n++) {
                                while (ns < MAXNS
                                       && EXT(statp).nsmap[ns] != MAXNS)
                                        ns++;

Reply to: