Linux Ethernet Bridging - Feb 2012 - [Bridge] STP message age broken?

There are problems with STP and its message age calculation even after commit:
 bridge: send proper message_age in config BPDU

Using HZ=1000 one can see that received msg and transmitted msg age are 0
going back to HZ=100 it gets better, but still looks strange.

Using HW switches(RSTP) and linux bridge in series:
  HW Switch(A, master) - HW Switch(B) - Linux BR(C) - Linux BR(D) - Linux BR(E)
In C we can with tcpdump see that incoming msg age is 1, but
in D we see incoming msg age 0, should be > 1 (or so I think)
Then in E we see incoming msg age 2.

Should not msg age increase with at least 1 for every hop?

 Jocke

Joakim Tjernlund/Transmode wrote on 2012/02/27 14:57:43:
>
> There are problems with STP and its message age calculation even after
commit:
>  bridge: send proper message_age in config BPDU
>
> Using HZ=1000 one can see that received msg and transmitted msg age are 0
> going back to HZ=100 it gets better, but still looks strange.
>
> Using HW switches(RSTP) and linux bridge in series:
>   HW Switch(A, master) - HW Switch(B) - Linux BR(C) - Linux BR(D) - Linux
BR(E)
> In C we can with tcpdump see that incoming msg age is 1, but
> in D we see incoming msg age 0, should be > 1 (or so I think)
> Then in E we see incoming msg age 2.
>
> Should not msg age increase with at least 1 for every hop?
hmm, Linux bridge seem to add a RELATIVE msg age instead of adding to the
received
msg age, that is, if rx msg age is 10 and processing takes 1 then the bridge
code sends 1 instead of
11. This seems broken to me.

 Jocke