frnkblk at iname
May 18, 2012, 12:08 PM
Post #3 of 3
Re: Preserving double-tagged traffic on ICX6610
[In reply to]
I'm not in a position to test this as you've asked, but I can confirm from
what the Brocade support engineer told me is that the issue was with how the
CPU handled double-tagged broadcast traffic (ARP). The problem has been
replicated on 1G ports, unstacked, and non-LAG.
From: igor [at] ergens [mailto:igor [at] ergens] On Behalf Of Igor Ybema
Sent: Friday, May 18, 2012 1:23 PM
To: frnkblk [at] iname
Cc: foundry-nsp [at] puck
Subject: Re: [f-nsp] Preserving double-tagged traffic on ICX6610
Hi Frank and all,
Today I had called our Brocade SE. For a case he was testing this bug also.
He could reproduce it on ICX 6450 and ICX 6610. However I still can
not reproduce it on the 6430.
We used the same configuration. The only differences are that:
- I am using 6430 instead of 6450 or 6610 (so no 10G ports)
- I am using a 24 ports model (so only one packer processor)
- I used the same 'port region' (because there is only one) for all
ports (uplink and downlinks) in the test
We concluded that the bug somehow is related to the packet processors,
how the 10G is connected internally or CPU core related. The Brocade
SE did not had time to test if using the same port region also fixed
the issue on the 6450 and the 6610. Can someone test this? If you use
port 1/1/1 to 1/1/4 for the test you should be in the same port region
for all types.
foundry-nsp mailing list
foundry-nsp [at] puck