[LARTC] Bridge + TC

[Jon Anderson]

Jeroen Vriesman wrote:

>So I would suggest testing:
>
>1) no filter rule for 1:10 which is default
>  
>
This shouldn't affect things in the end though, correct? (I.e. it's 
overkill, but it won't hurt anything, right?)

(I've also had it pass by default through the 1:1, in which case nothing 
passed through 1:10, nor 1:20 - everything went through 1:1.)

>2) no filters with handle 1, (I start at 101 for the filters)
>  
>
>> tc filter add dev eth0 parent 1:0 protocol ip prio 1 handle 1 fw classid 1:10

"handle 1" - is that not how it picks up on nfmark? If nfmark is set to 
1, is this not the part that picks up that nfmark?

>3) marking with iptables in mangle PREROUTING
>  
>

I have also tried that. I'm using 2.6.3-mm3 -> packets don't seem to 
pass through iptables anymore unless they're specifically routed rather 
than bridged (can anyone confirm this?). I have another (2-if, no QoS) 
bridge running 2.4, and iptables commands filter fine. With this new 
bridge running 2.6, dropping everything with iptables doesn't work 
'iptables -A FORWARD -j DROP' doesn't affect the bridge in the least - 
hosts continue to talk through the bridge. (Same in ebtables works as 
expected though.)

>should work, it's working fine here on 2.4.24+ebtables 
>
>  
>
Perhaps the key here is 2.4. I might have to revert...

Thanks for the input. Now I have something else to try!

Cheers,

jon