[Bug 206721] FreeBSDs DHCP client(dhclient) does not support the interface-mtu option(option 26).

bugzilla-noreply at freebsd.org bugzilla-noreply at freebsd.org
Fri Jan 29 03:35:08 UTC 2016


https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=206721

            Bug ID: 206721
           Summary: FreeBSDs DHCP client(dhclient) does not support the
                    interface-mtu option(option 26).
           Product: Base System
           Version: 11.0-CURRENT
          Hardware: Any
                OS: Any
            Status: New
          Severity: Affects Some People
          Priority: ---
         Component: bin
          Assignee: freebsd-bugs at FreeBSD.org
          Reporter: jlpetz at gmail.com

After getting the Intel ixv driver working for AWS instances. I noticed there
was some severe packet loss in iPerf tests. Looking at ifconfig I suspected
there was an MTU blackhole between the virtual ethernet interface(ixv) with MTU
of 1500 and the physical card MTU 9001. Confirmed by manually changing it and a
few ping tests with 'Don't fragment' flags set.

Changing the MTU manually corrected the packet loss. But then I was wondering
why this wasn't automatically done like it is for other operating systems which
I have run on AWS EC2.

Turn out it looks like EC2 sends the MTU which should be configured for the
interface as part of the DHCP response and FreeBSDs default client.
1. Doesn't request the interface-mtu option(option 26). Needed for it to be in
the response
2. Even if it did request it this option and got it in the response. I don't
think it would process it and configure the MTU on the NIC. Based on the source
code I looked at.

I was able to work around this by using the isc-dhcp43-client-4.3.3
package/port. But am logging this as a bug/feature-request as I feel this
should be part of the base system. MTU blackholes are quite common when using
jumbo frame(~9000 MTU) systems and this seems like a good way of avoiding them
which a lot of other operating systems are doing by default.

More details about this are in the phabricator review which has my testing
details(Link below). Make sure you click on the "Show Older Changes" to see all
the text on the page below.
https://reviews.freebsd.org/D4788

-- 
You are receiving this mail because:
You are the assignee for the bug.


More information about the freebsd-bugs mailing list