[j-nsp] ACX5048 - 802.1q CFI DEI bit

Aaron Gould aaron1 at gvtc.com
Tue Nov 20 16:32:13 EST 2018


Had something interesting occur in my company's lab recently.  We had an
issue where we couldn't get connectivity through third party device
connected to a ACX5048 and we found the following while sniffing that
connection.

 

It seems on certain vlans, the CFI/DEI bit is set on traffic as it exits the
ACX5048.  I need to confirm that this is normal for the ACX5048 and how to
configure the CFI/DEI bit to *always* be 0

 

The ACX5048 systematically, every group of 16 vlans, cycle the CFI/DEI bit.
Also the ACX5048 will increment the 802.1p priority bits by 1 every 16th
group of vlans, except the first 32 stay 0, but after that it seems to
increment by 1 every group of 16.

 

I tested and sniffed the following vlans.  First column is vlan id.  The
reason I say "good" or "bad" is because, when the ACX5048 seemed to set the
CFI/DEI bit to 1, I called it "bad" because I couldn't ping the host behind
the third party device.  I'm thinking the 3rd party device was dropping it
as discard eligible or some legacy reason related to cfi rules of Ethernet
to token ring bridging or something like that.  I will say that I replaced
the 3rd party product with a newer one from that same vendor, and then all
was well.  .meaning, even while the ACX5048 persisted to mark cfi/dei as 1,
the newer 3rd party device would allow it to pass and I could then connected
to that test host.

 

02 - cfi 0 - good - pri 000

...untested but assuming good

05 - cfi 0 - good - pri 000

...untested but assuming good

13 - cfi 0 - good - pri 000 (0)

14 - cfi 0 - good - pri 000 (0)

15 - cfi 0 - good - pri 000 (0)

16 - cfi 1 - bad  - pri 000 (0)

17 - cfi 1 - bad  - pri 000 (0)

18 - cfi 1 - bad  - pri 000 (0)

19 - cfi 1 - bad  - pri 000 (0)

20 - cfi 1 - bad  - pri 000 (0)

21 - cfi 1 - bad  - pri 000 (0)

22 - cfi 1 - bad  - pri 000 (0)

23 - cfi 1 - bad  - pri 000 (0)

24 - cfi 1 - bad  - pri 000 (0)

25 - cfi 1 - bad  - pri 000 (0)

26 - cfi 1 - bad  - pri 000 (0)

27 - cfi 1 - bad  - pri 000 (0)

28 - cfi 1 - bad  - pri 000 (0)

29 - cfi 1 - bad  - pri 000 (0)

30 - cfi 1 - bad  - pri 000 (0)

31 - cfi 1 - bad  - pri 000 (0)

32 - cfi 0 - good - pri 001 (1)

33 - cfi 0 - good - pri 001 (1)

34 - cfi 0 - good - pri 001 (1)

...untested but assuming good

47 - cfi 0 - good - pri 001 (1)

48 - cfi 1 - bad  - pri 001 (1)

...untested but assuming bad

63 - cfi 1 - bad  - pri 001 (1)

64 - cfi 0 - good - pri 010 (2)

...untested but assuming good

79 - cfi 0 - good - pri 010 (2)

80 - cfi 1 - bad  - pri 010 (2)

...untested but assuming bad

95 - cfi 1 - bad  - pri 010 (2)

96 - cfi 0 - good - pri 011 (3)

 

seems that arp from ACX5048 doesn't set the CFI/DEI to 1, but pings do get
set to 1.

 

 

- Aaron

 

 



More information about the juniper-nsp mailing list