From: Pascal Hambourg on
itba a �crit :
> I like your point of view, while thinking about you said, I made some dump of
> traffic from eth0 (which contains eth0.100,eth0.200 and eth0.300), where all
> things are messing up. Here is a log from tcpdump:
> $> tcpdump -e -i eth0 > /tmp/dump-eth0.txt
> $> grep DHCP /tmp/dump-eth0.txt
>
> 11:14:54.247429 00:0b:5d:b8:09:b0 (oui Unknown) > Broadcast, ethertype
> 802.1Q
> (0x8100), length 594: vlan 300, p 0, ethertype IPv4, 0.0.0.0.bootpc >
> 255.255.255.255.bootps: BOOTP/DHCP, Request from 00:0b:5d:b8:09:b0 (oui
> Unknown), length 548
> 11:14:54.247450 00:0b:5d:b8:09:b0 (oui Unknown) > Broadcast, ethertype
> 802.1Q
> (0x8100), length 594: vlan 100, p 0, ethertype IPv4, 0.0.0.0.bootpc >
> 255.255.255.255.bootps: BOOTP/DHCP, Request from 00:0b:5d:b8:09:b0 (oui
> Unknown), length 548
> 11:14:54.247827 00:0b:5d:b8:09:b0 (oui Unknown) > Broadcast, ethertype
> 802.1Q
> (0x8100), length 594: vlan 200, p 0, ethertype IPv4, 0.0.0.0.bootpc >
> 255.255.255.255.bootps: BOOTP/DHCP, Request from 00:0b:5d:b8:09:b0 (oui
> Unknown), length 548
[...]
> Here can we see again how this packets are received by all subinterfaces.
> What you think about this log ?

Apparently the same DHCP request is received with the 3 VLAN IDs at
about the same time on the raw interface, as I suspected in my previous
message. So my opinion is that the problem does not lie in dnsmasq nor
the Linux box but in the ethernet network outside, maybe the switch that
handles the VLANs. Unless the 3 VLAN interfaces are bridged together in
the Linux box ; you didn't do that, did you ?
Does this happen with any type of broadcast (e.g. broadcast ping or ARP
request) or only DHCP packets ?