All of lore.kernel.org
 help / color / mirror / Atom feed
* RE: [Bridge] Small packets
@ 2007-10-29  5:40 Leigh Sharpe
  2007-10-29 13:06 ` [Bridge] " Benny Amorsen
  0 siblings, 1 reply; 5+ messages in thread
From: Leigh Sharpe @ 2007-10-29  5:40 UTC (permalink / raw)
  To: Stephen Hemminger; +Cc: bridge

 

-----Original Message-----
From: Stephen Hemminger [mailto:shemminger@linux-foundation.org] 
Sent: Monday, 29 October 2007 4:12 PM
To: Leigh Sharpe
Cc: bridge@lists.linux-foundation.org
Subject: Re: [Bridge] Small packets

On Mon, 29 Oct 2007 15:37:52 +1100
"Leigh Sharpe" <lsharpe@pacificwireless.com.au> wrote:

>> Which seems to bridge OK. However, I have noticed that the bridge
>> appears to be truncating some packets. In particular, I am worried
>> about packets such as IP ACK's, which are entering the bridge as
>> 64-byte packets, but coming out the other side as 60-byte packets.
>> As I recall, the smallest valid ehternet packet is 64 bytes. The
>> bridge seems to be removing any padding in the packet and sending it
>> on it's way as a packet smaller than the minimum ethernet packet
>> size. Looking at the packet distribution on other devices connected
>> to the bridge, it seems to be doing the same to lots of other
> packets, too. Anybody got any idea why the bridge might be truncating
> packets? Any idea why it's sending 60-byte packets, when the minimu
> should be 64? Or where to tell it to pad all packets out to 64 bytes?
>  
>  
> Leigh.
>  

>What hardware? It could be a driver bug.
>Also if you use VLAN's then the driver may send 60 byte packet (because
>of the 4 byte VLAN tag).

Standard e1000 hardware.
The packets being bridged contain a VLAN tag, which is included in the
60 bytes.





^ permalink raw reply	[flat|nested] 5+ messages in thread

* [Bridge] Re: Small packets
  2007-10-29  5:40 [Bridge] Small packets Leigh Sharpe
@ 2007-10-29 13:06 ` Benny Amorsen
  2007-10-29 15:06   ` Andy Gospodarek
  0 siblings, 1 reply; 5+ messages in thread
From: Benny Amorsen @ 2007-10-29 13:06 UTC (permalink / raw)
  To: bridge

>>>>> "LS" == Leigh Sharpe <lsharpe@pacificwireless.com.au> writes:

LS> Standard e1000 hardware. The packets being bridged contain a VLAN
LS> tag, which is included in the 60 bytes.

The e1000 has VLAN acceleration. The VLAN tag is sent in a separate
register. If you do packet capture on the sender, the packet will
likely look 60 bytes long, even if it is 64 bytes on the wire.

The same thing happens on receive. Packet dumping with VLAN's is a bit
of a mess in Linux. If you're lucky you can find a card without VLAN
acceleration to do the packet dump.


/Benny



^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [Bridge] Re: Small packets
  2007-10-29 13:06 ` [Bridge] " Benny Amorsen
@ 2007-10-29 15:06   ` Andy Gospodarek
  2007-10-29 15:48     ` Ben Greear
  0 siblings, 1 reply; 5+ messages in thread
From: Andy Gospodarek @ 2007-10-29 15:06 UTC (permalink / raw)
  To: Benny Amorsen; +Cc: bridge

On 10/29/07, Benny Amorsen <benny+usenet@amorsen.dk> wrote:
> >>>>> "LS" == Leigh Sharpe <lsharpe@pacificwireless.com.au> writes:
>
> LS> Standard e1000 hardware. The packets being bridged contain a VLAN
> LS> tag, which is included in the 60 bytes.
>
> The e1000 has VLAN acceleration. The VLAN tag is sent in a separate
> register. If you do packet capture on the sender, the packet will
> likely look 60 bytes long, even if it is 64 bytes on the wire.
>
> The same thing happens on receive. Packet dumping with VLAN's is a bit
> of a mess in Linux. If you're lucky you can find a card without VLAN
> acceleration to do the packet dump.
>
>
> /Benny
>

Are these the lengths on the wire or when captured on the host?   The
smallest VLAN tagged frame should be 68 bytes IIRC.  A tagged frame
that is 64 bytes seems too small.

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [Bridge] Re: Small packets
  2007-10-29 15:06   ` Andy Gospodarek
@ 2007-10-29 15:48     ` Ben Greear
  2007-10-29 16:15       ` Andy Gospodarek
  0 siblings, 1 reply; 5+ messages in thread
From: Ben Greear @ 2007-10-29 15:48 UTC (permalink / raw)
  To: Andy Gospodarek; +Cc: bridge, Benny Amorsen

Andy Gospodarek wrote:
> On 10/29/07, Benny Amorsen <benny+usenet@amorsen.dk> wrote:
>>>>>>> "LS" == Leigh Sharpe <lsharpe@pacificwireless.com.au> writes:
>> LS> Standard e1000 hardware. The packets being bridged contain a VLAN
>> LS> tag, which is included in the 60 bytes.
>>
>> The e1000 has VLAN acceleration. The VLAN tag is sent in a separate
>> register. If you do packet capture on the sender, the packet will
>> likely look 60 bytes long, even if it is 64 bytes on the wire.
>>
>> The same thing happens on receive. Packet dumping with VLAN's is a bit
>> of a mess in Linux. If you're lucky you can find a card without VLAN
>> acceleration to do the packet dump.
>>
>>
>> /Benny
>>
> 
> Are these the lengths on the wire or when captured on the host?   The
> smallest VLAN tagged frame should be 68 bytes IIRC.  A tagged frame
> that is 64 bytes seems too small.

That is not correct per the 802.1Q VLAN RFC, though
I don't have the reference handy at the moment.

64 bytes is fine, vlan tagged or otherwise.


Thanks,
Ben



> _______________________________________________
> Bridge mailing list
> Bridge@lists.linux-foundation.org
> https://lists.linux-foundation.org/mailman/listinfo/bridge


-- 
Ben Greear <greearb@candelatech.com>
Candela Technologies Inc  http://www.candelatech.com


^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [Bridge] Re: Small packets
  2007-10-29 15:48     ` Ben Greear
@ 2007-10-29 16:15       ` Andy Gospodarek
  0 siblings, 0 replies; 5+ messages in thread
From: Andy Gospodarek @ 2007-10-29 16:15 UTC (permalink / raw)
  To: Ben Greear; +Cc: bridge, Benny Amorsen

On 10/29/07, Ben Greear <greearb@candelatech.com> wrote:
> Andy Gospodarek wrote:
> > On 10/29/07, Benny Amorsen <benny+usenet@amorsen.dk> wrote:
> >>>>>>> "LS" == Leigh Sharpe <lsharpe@pacificwireless.com.au> writes:
> >> LS> Standard e1000 hardware. The packets being bridged contain a VLAN
> >> LS> tag, which is included in the 60 bytes.
> >>
> >> The e1000 has VLAN acceleration. The VLAN tag is sent in a separate
> >> register. If you do packet capture on the sender, the packet will
> >> likely look 60 bytes long, even if it is 64 bytes on the wire.
> >>
> >> The same thing happens on receive. Packet dumping with VLAN's is a bit
> >> of a mess in Linux. If you're lucky you can find a card without VLAN
> >> acceleration to do the packet dump.
> >>
> >>
> >> /Benny
> >>
> >
> > Are these the lengths on the wire or when captured on the host?   The
> > smallest VLAN tagged frame should be 68 bytes IIRC.  A tagged frame
> > that is 64 bytes seems too small.
>
> That is not correct per the 802.1Q VLAN RFC, though
> I don't have the reference handy at the moment.
>
> 64 bytes is fine, vlan tagged or otherwise.
>
>
> Thanks,
> Ben
>
>

Ah, you are correct, Ben.  Thanks for pointing that out.  Section
C.4.4.1 of the IEEE 802.1Q spec does state that 64-byte tagged frames
are OK, but that is only if you are able to drop padding from a
64-byte untagged one.  Transmitting a 60-byte frame on the wire is
incorrect.

^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2007-10-29 16:15 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2007-10-29  5:40 [Bridge] Small packets Leigh Sharpe
2007-10-29 13:06 ` [Bridge] " Benny Amorsen
2007-10-29 15:06   ` Andy Gospodarek
2007-10-29 15:48     ` Ben Greear
2007-10-29 16:15       ` Andy Gospodarek

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.