The check that there's enough space to store the DHCP agent-id
at the end of the packet could succeed when it should fail
if the END option is in either of the oprion-overload areas.
That could overwrite legit options in the request and cause
bad behaviour. It's highly unlikely that any sane DHCP client
would trigger this bug, and it's never been seen, but this
fixes the problem.
Also fix off-by-one in bounds checking of option processing.
Worst case scenario on that is a read one byte beyond the
end off a buffer with a crafted packet, and maybe therefore
a SIGV crash if the memory after the buffer is not mapped.
Thanks to Timothy Becker for spotting these.
From f3d832b41f44c856003517c583fbd7af4dca722c Mon Sep 17 00:00:00 2001
From: Neil Jerram <Neil.Jerram@metaswitch.com>
Date: Fri, 8 Apr 2016 19:23:47 +0100
Subject: [PATCH] Fix DHCPv4 reply via --bridge-interface alias interface
Sending a DHCPv4 reply through a --bridge-interface alias interface
was inadvertently broken by
commit 65c7212000
Author: Lung-Pin Chang <changlp@cs.nctu.edu.tw>
Date: Thu Mar 19 23:22:21 2015 +0000
dhcp: set outbound interface via cmsg in unicast reply
If multiple routes to the same network exist, Linux blindly picks
the first interface (route) based on destination address, which might not be
the one we're actually offering leases. Rather than relying on this,
always set the interface for outgoing unicast DHCP packets.
because in the aliasing case, iface_index is changed from the index of
the interface on which the packet was received, to be the interface
index of the 'bridge' interface (where the DHCP context is expected to
be defined, and so needs to be looked up).
For the cmsg code that the cited commit added, we need the original
iface_index; so this commit saves that off before the aliasing code
can change it, as rcvd_iface_index, and then uses rcvd_iface_index
instead of iface_index for the cmsg code.
Commit 51967f9807 began treating SERVFAIL
as a successful response from an upstream server (thus ignoring future
responses to the query from other upstream servers), but a typo in that
commit means that REFUSED responses are accidentally being treated as
successful instead of SERVFAIL responses.
This commit corrects this typo and provides the behaviour intended by
commit 51967f9: SERVFAIL responses are considered successful (and will
be sent back to the requester), while REFUSED responses are considered
unsuccessful (and dnsmasq will wait for responses from other upstream
servers that haven't responded yet).
