DHCPDISCOVER with bad udp checksum issue

dhcp-question at kraitch.net dhcp-question at kraitch.net
Wed Oct 6 19:07:04 UTC 2021


Hello.

We are wondering if a bad udp checksum on
a DHCPDISCOVER prevents the fault ISC dhcpd from
providing dhcp support?   If so what are are
the options for dealing with this issue s while
we wait for our network vendor to
reproduce and fix the bad udp checksum issue?

We posted a message on dhcp-users at lists.isc.org
in September 2021 but received no replies.

Here is the content of that was sent to dhcp-users -


Date: Thu, 09 Sep 2021 15:07:18 -0700
Subject: dhcpd not seeing DHCPDISCOVER traffic

Hello.

We have a stable working single VRF
*virtual routing and Forwarding) network technology
with routers perfuming dhcp relaying for
multiple networks to ISC  4.4.2-P1 DHCP
revers running on RHEL (Red Hat Enterprise Linux) 7 servers.

We are migrating to a multiple VRF environment.

We were surprised when we encountered an issue
with our first test network on a new VRF with
the router dhcp relaying to a dhcp server on the
original VRF did not work.   The dhcp server does
not appear to see and does not log or process the
DHCPDISCOVER traffic being relayed form the new
VRF.   When we run tcpdump on the dhcp server,
tcpdump does see the DHCPDISCOVER traffic being
relayed form the new VRF.   We also noticed that
tcpdump showed the DHCPDISCOVER traffic has a
"bad udp cksum."

Could the bad checksum be why the dhcp server is
not logging or processing the DHCPDISCOVER requests?

We tried compiling dhcpd with DEBUG_CHECKSUM_VERBOSE
defined in in includes/site.h but that did not
appears to provide any helpful insight.  BTW, we
could not compile with DEBUG_CHECKSUM defined.

"ethtool -k" on a RHEL dhcp server showed that offload of
checksums was enabled.  Disabling offloading
did not change tcpdump shewing bad udp checksums and
did not result in dhcpd seeing and processing the
DHCPDISCOVERs with bad udp checksums.

Assuming the bad udp checksums are why dhcpod is
not processing the DHCPDISCOVERs, what can be done from
the dhcpd end?   There does not be a #define to
ignore bad udp checksums.   We are not sure what
the code in common//packet.c does but it might be a
candidate for modification if ignoring the bad udp checksums
is possible,

Our main question remains are the bad udp checksums on
the DHCPDISCOVERs preventing the fault dhcpd from
providing dhcp support?

Thanks,

mark



More information about the dhcp-workers mailing list