Bug 1810 – IP packets can be sent on NetDevices not respecting the minimum MTU requirements

Bug 1810 - IP packets can be sent on NetDevices not respecting the minimum MTU requirements


Summary:	IP packets can be sent on NetDevices not respecting the minimum MTU requirements

Status:	RESOLVED FIXED

Product:	ns-3
Classification:	Unclassified
Component:	internet
Version:	ns-3-dev
Hardware:	All All

Importance:	P5 normal
Assigned To:	Tommaso Pecorella

URL:
Keywords:

Depends on:
Blocks:
	Show dependency tree / graph

Reported:	2013-12-04 16:41 UTC by Tommaso Pecorella
Modified:	2013-12-12 13:37 UTC (History)
CC List:	2 users (show)

See Also:

Attachments
patch (3.76 KB, patch) 2013-12-04 16:42 UTC, Tommaso Pecorella	Details \| Diff
New patch (2.24 KB, patch) 2013-12-11 13:25 UTC, Tommaso Pecorella	Details \| Diff
Show Obsolete (1) Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.

Description Tommaso Pecorella 2013-12-04 16:41:56 UTC

Note that this bug also affects IPv6.

An IP packet might be sent over NetDevices with an MTU as small as 28 bytes, but RFC 791 clearly states:

    Every internet module must be able to forward a datagram of 68
    octets without further fragmentation.  This is because an internet
    header may be up to 60 octets, and the minimum fragment is 8 octets.

The proposed patch should fix it. A new Drop reason is introduced: DROP_MTU_TOO_LITTLE.

IPv4, IPv6 and FlowMon have been modified accordingly.

Comment 1 Tommaso Pecorella 2013-12-04 16:42:56 UTC

Created attachment 1731 [details]
patch

The proposed patch should fix it. A new Drop reason is introduced: DROP_MTU_TOO_LITTLE.

IPv4, IPv6 and FlowMon have been modified accordingly.

Comment 2 Tom Henderson 2013-12-11 11:06:30 UTC

I wonder whether this could be instead enforced when an IP interface is added (or MTU modified) rather than a per-packet check of the interface MTU.  I seem to recall the behavior in Linux for IPv6 is to disable the interface for IPv6 usage if the MTU is too small, rather than check on a per-packet basis.

Also, I might suggest "TOO_SMALL" instead of "TOO_LITTLE" (nit).

Comment 3 Tommaso Pecorella 2013-12-11 11:08:48 UTC

Very good idea. The patch was done kinda in a rush.

Moreover, we'll not need the new drop reason, as the interface will be down (at least for IP).

I'll modify it accordingly.

Comment 4 Tommaso Pecorella 2013-12-11 13:25:01 UTC

Created attachment 1732 [details]
New patch

Much cleaner approach (thanks Tom).

The Ipv4Interface (or Ipv6Interface) is simply set as "down" if the associated NetDevice has a too short MTU.

Removed the drop reason cause, as packets will be dropped with an "interface down" reason automatically.

Of course it is possible to manually force the Ipv4/6Interface to be up, but I guess nobody would do that without knowing what he/she is doing.

Comment 5 Tom Henderson 2013-12-12 09:25:30 UTC

Looks good to merge, IMO

Comment 6 Tommaso Pecorella 2013-12-12 13:37:50 UTC

Fixed in changeset: 10512:f222d52c616f