MTU, fragmentation and Jumbo Frames question
h.schmalzbauer at omnilan.de
Tue Feb 9 13:14:04 UTC 2010
Ilya Zhuravlev schrieb am 09.02.2010 13:22 (localtime):
> On 09.02.2010 20:12, Harald Schmalzbauer wrote:
>> Hello all,
>> I have 3 machines connected to an JumboFrame enabled switch.
>> One FreeBSD 8-stable and two windows machines.
>> I can send echo requests with payload of 8972 between the windows
>> machines, but I don't get an answer from the FreeBSD machine. At the
>> edge of 8130 bytes, the FreeBSD machine eats the packets without any
>> error notification. A payload <= 8130 bytes works!
>> All MTUs are set to 9000, interfaces are all intel em.
>> Like mentioned, the ping between the two windows machines work like
>> expected, beyond 8972 bytes payload the OS is fragmenting (resp. tells
>> me that DF bit was set but fragmentation was needed)
>> FreeBSD seems to never fragment packets, since I don't get an answer if
>> I define payload greater than MTU. But this should work, shouldn't it
>> (`ping -s 10000 host`)
>> Does anybody have any explanation why pings are working up to 8130 bytes
>> payload and are silently droped beyond that even my MTU ist set to 9000
>> (and route get confirms the MTU 9000)?
> but "ifconfig em0 -rxcsum" is enough for me.
Thanks, but disabling any sum offloading doesn't change anything.
Still, I wonder why FreeBSD doesn't fragment packets if they're too large.
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 196 bytes
Desc: OpenPGP digital signature
Url : http://lists.freebsd.org/pipermail/freebsd-questions/attachments/20100209/d332ccb5/signature.pgp
More information about the freebsd-questions