6.2 mtu now limits size of incomming packet

[Tom Judge]

Stephen Clark wrote:
> Mike Karels wrote:
> 
>>> A related change that should probably be discussed if we want to 
>>> think more about asymmetry in maximum transmission unit is this one:
>>>   
>>
>>  
>>
>>>   ----------------------------
>>>   revision 1.98
>>>   date: 2006/06/26 17:54:53;  author: andre;  state: Exp;  lines: +2 -0
>>>   In syncache_respond() do not reply with a MSS that is larger than what
>>>   the peer announced to us but make it at least tcp_minmss in size.
>>>   
>>
>>  
>>
>>>   Sponsored by:   TCP/IP Optimization Fundraise 2005
>>>   ----------------------------
>>>   
>>
>>  
>>
>>> In this change, we cap the advertised MSS in SYN/ACK to the received 
>>> advertised MSS, which presumably avoids an extra PMTU round trip if 
>>> jumbograms are enabled on the receiving endpoint.  However, it also 
>>> prevents use of larger packet sizes if asymmetric MTU is supported.  
>>> I think I suggested after this was committed that we at least add an 
>>> administrative twiddle to enable/disable this mode of operation, but 
>>> don't see one in there currently. Does the Secure Computing scenario 
>>> use TCP in this way, and is the potential win in avoiding a PMTU 
>>> round-trip worth disallowing asymmetric MSS at the TCP layer?
>>>   
>>
>> In our case, TCP isn't aware of the MRU, and bases its MSS on the MTU 
>> values.
>> However, I don't see any reason for TCP to cap the MSS at the received 
>> MSS.
>> If the other end doesn't want to receive more than 1024 bytes, that's no
>> reason to refuse to accept more.
>>
>>         Mike
>>
>>  
>>
> So was any decision reached on this issue - will FreeBSD changed to 
> accept a packet on an
> interface that is larger than the mtu on that interface?
> 
> Steve
> 
There are also things to consider such as if a GigE card is connected to 
a GigE device (switch/card etc) and the card supports jumbo frames 
should the MRU be set to the max jumbo receive size for the card? This 
could cause confusion when people plug jumbo capable devices in with 
hardware limitations making the MRU lower than other devices on the network.

Just some thoughts.

Tom