[Dnsmasq-discuss] DNS request timeout problem on dial-in account with dynamic ip

[Jean Wolter]

Simon Kelley <simon at thekelleys.org.uk> writes:

> The netlink code we talked about went together very easily:
>
> The attached code will re-transmit the last query it sent which has not
> seen a reply, if it sees an "interface up" event.

It looks like this is the wrong event to look for in our situation. In
dial on demand the interface is already up, there is a route going
through this interface and packets going over this route trigger a
dial in.

At least this is what I see here, when a normal dial in is triggered
dnsmasq doesn't get any notifications from the kernel. If I tell pppd
to terminate and restart it, dnsmasq gets the following notifications
(generated by       
     syslog (LOG_ERR, "msg type: %x (%s)\n",
             h->nlmsg_type, 
             h->nlmsg_type == RTM_NEWLINK ? "RTM_NEWLINK" : h->"unknown");
 in netlink_multicast() ):

# kill -TERM 571
Feb 19 22:13:16 foo pppd[571]: Terminating on signal 15.
...
Feb 19 22:13:19 foo pppd[571]: Exit.
Feb 19 22:13:19 foo dnsmasq[1254]: msg type: 10 (RTM_NEWLINK) 
Feb 19 22:13:19 foo dnsmasq[1254]: msg type: 10 (RTM_NEWLINK) 
Feb 19 22:13:19 foo imond[1280]: setting default route again to
circuit 'DSL' 
Feb 19 22:13:19 foo dnsmasq[1254]: msg type: 11 (unknown) 
Feb 19 22:13:22 foo pppd[1971]: Plugin rp-pppoe.so loaded.
Feb 19 22:13:22 foo pppd[1971]: RP-PPPoE plugin version 3.3 compiled
against pppd 2.4.2
Feb 19 22:13:22 foo pppd[1971]: pppd 2.4.2 started by root, uid 0
Feb 19 22:13:22 foo dnsmasq[1254]: msg type: 10 (RTM_NEWLINK) 
Feb 19 22:13:22 foo pppd[1971]: Using interface ppp0
Feb 19 22:13:22 foo pppd[1971]: Couldn't increase MTU to 1500
Feb 19 22:13:22 foo pppd[1971]: Couldn't increase MRU to 1500
Feb 19 22:13:22 foo dnsmasq[1254]: msg type: 10 (RTM_NEWLINK) 
Feb 19 22:13:22 foo dnsmasq[1254]: msg type: 10 (RTM_NEWLINK) 

It doesn't get any notification during a normal dial in

Regards,
Jean