Rare project developers

[Alexander Gall <>]

On Tue, 14 Jun 2022 09:46:26 +0200, mc36 <> said:

> hi,
> excellent questions... let me start from the easiest to the hardest:
> -freerouter starts an sdn interface in up when you issue "int sdn1"
> -for this experiment, "int sdn1" / "log-link" should be configured btw... :)
> -upon exporting under "server p4lang", it get's it's underlay so it goes 
> down

That doesn't happen. When the port is configured, the oper state is
never reported back. All I see is (apart from some other "rx" messages
unrelated to the port state)

rx: ['ports_add', '0', '10', '2', '1', '0', '\n']
WARNING:bf_forwarder.py:rx: ['ports_add', '0', '10', '2', '1', '0', '\n']
Port[0] added
WARNING:bf_forwarder.py:Port[0] added
rx: ['state', '0', '1', '10', '2', '1', '0', '\n']
WARNING:bf_forwarder.py:rx: ['state', '0', '1', '10', '2', '1', '0', '\n']
Port[0] administrative status set to up
WARNING:bf_forwarder.py:Port[0] administrative status set to up

"sh int" shows the port as "up" indefinitely. The forwarder never
sends a status. My guess is that this is because there are no state
*transistions* in the data-plane. The oper status is always down.

Probably a corner-case we haven't covered? I.e. maybe the forwarder
should report the initial oper state?

> -as long as the forwarder reports a state from the dataplane...
> -to see the above, debug serv p4 rx/tx should be enabled too...
> now the part i clearly not understand is
> " After a warm reload, it still briefly works, but then sdn1 changes state 
> to down and never comes back up."
> because freerouter reports N if an interface is down and no other route 
> exists...

Let's assume this problem is fixed. What actually worries me more is
the fact that the model appears to never mark the port as UP. That
would mean that freerouter can't work with the model.

However, since you do use freerouter on the model and you didn't run
into this problem makes me wonder if there is anything different on my
system :/

-- 
Alex

> thanks,
> cs



> On 6/14/22 08:24, Alexander Gall wrote:
>> Hi
>> 
>> I'm reproducing the demo I've made running the RARE release on the
>> Tofino model and some questions popped up.
>> 
>> For now, I'm still using the built-in JSON ports file with the veths,
>> i.e. veth0 is connected to Tofino port #0 etc.
>> 
>> Baically, what I'm doing is add an interface sdn1 on frontpanel-1/0,
>> configure an address on it and on veth1 and do a ping. It works, but
>> I'm actually confused why it works :/
>> 
>> The thing is that I've realized for the first time that the OPR
>> status of the Tofino ports in the model is always DWN, e.g.
>> 
bf-sde.pm> show -p 1/0
>> -----+----+---+----+-------+----+--+--+---+---+---+--------+----------------+----------------+-
>> PORT |MAC |D_P|P/PT|SPEED  |FEC |AN|KR|RDY|ADM|OPR|LPBK    |FRAMES RX 
>> |FRAMES TX       |E
>> -----+----+---+----+-------+----+--+--+---+---+---+--------+----------------+----------------+-
>> 1/0  |15/0|  0|0/ 0|10G    |NONE|Ds|Au|YES|ENB|DWN|  NONE  | 8|            
>>   10|*
>> 
>> This is despite the model logging
>> 
>> :06-14 08:07:26.582773:    PORT UP asic 0 port 0
>> 
>> AFAICT, this is happening at least for SDE 9.6.x through 9.7.x. For SDE
>> 9.8.0 and 9.9.0, the OPR status is reported as UP when a port is
>> connected to it via the JSON port mapping. Can you confirm this?
>> 
>> The model seems to ignore the OPR status and forwards packets in both
>> direction. What confuses me with freerouter is that the port is
>> reported to be up anyway right after adding the config for sdn1
>> 
>> rare(cfg)#vrf definition foo
>> rare(cfg-vrf)#exit
>> rare(cfg)#interface sdn1
>> rare(cfg-if)#vrf forwarding foo
>> rare(cfg-if)#ipv4 address 192.168.1.1 /24
>> rare(cfg-if)#exit
>> rare(cfg)#server p4lang p4
>> rare(cfg-server)#export-port sdn1 frontpanel-1/0 10 2 1 0
>> rare(cfg-server)#export-vrf foo 2
>> rare(cfg-server)#sh int sdn1
>> sdn1 is up
>> description:
>> state changed 3 times, last at 2022-06-14 08:14:27, 00:00:23 ago
>> last packet input never ago, output never ago, drop never ago
>> type is sdn, hwaddr=004b.4853.5536, mtu=1500, bw=8000kbps, vrf=foo
>> ipv4 address=192.168.1.1/24, mask=255.255.255.0, ifcid=545466702
>> received 0 packets (0 bytes) dropped 0 packets (0 bytes)
>> transmitted 0 packets (0 bytes) macsec=false sgt=false
>> 
>> And ping works
>> 
>> rare(cfg)#exit
>> rare#ping 192.168.1.2 vrf foo
>> pinging 192.168.1.2, src=null, vrf=foo, cnt=5, len=64, tim=1000,
>> gap=0, ttl=255, tos=0, sgt=0, flow=0, fill=0, sweep=false, multi=false
>> .!!!!
>> result=80%, recv/sent/lost/err=4/5/1/0, rtt
>> min/avg/max/sum=16/252/956/2037, ttl min/avg/max=64/64/64, tos
>> min/avg/max=0/0/0
>> 
>> After a warm reload, it still briefly works, but then sdn1 changes
>> state to down and never comes back up.
>> 
>> So my questions are
>> 
>> * Is this a bug in the Tofino model up to 9.7?
>> * Why does freerouter (sort of) still work (i.e. ignoring the
>> actual oper status)?
>>