It appears to be a bug with the nftables userspace utility nft. I
wrote a program to decode nftables rules and found the nft utility is
trying to fill the BPF data register with the priority value in
big-endian format. If I hack it to use little-endian format for the
meta value, the skb->priority is set appropriately.
Brad
Brad Pousland
Principle Software Engineer
WigWag, Inc.
c - 512-963-2446
www.wigwag.com
On Tue, Apr 12, 2016 at 12:13 PM, Brad Pousland <bpousland@xxxxxxxxx> wrote:
> I am trying to control the vlan qos priority of a link by using an
> egress qos map on the vlan link and setting the packet priority using
> an nftables rule.
>
> I have a vlan link defined with an egress map from 3 packet priorities
> to 3 vlan qos priorities as follows:
>
> 30: eth0.2@enx00e07cc8db5b: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500
> qdisc noqueue state UP mode DEFAULT group default
> link/ether 00:e0:7c:c8:db:5b brd ff:ff:ff:ff:ff:ff promiscuity 0
> vlan protocol 802.1Q id 2 <REORDER_HDR>
> egress-qos-map { 0:3 1:2 2:6 } addrgenmode eui64
>
> As a test I am using a simple rule to set the packet priority of any
> packet from the vlan address:
>
> table ip mangle {
> chain postrouting {
> type filter hook postrouting priority 0; policy accept;
> ip saddr 192.168.1.62 meta priority set 0200: counter
> packets 0 bytes 0
> }
> }
>
> I then send packets from eth0.2 and inspect the packets from the
> physical parent device. The QOS field is always set to 3 as if the
> priority was never modified on the outgoing packet by the rule, even
> though the rule counter shows the rule was invoked.
>
> Does anyone know if the "meta priority set" can be used for this
> purpose? I could give up and use tc for vlan priority control but
> that seems overkill since I don't have complex traffic shaping
> requirements.
>
> Thanks,
> Brad
--
To unsubscribe from this list: send the line "unsubscribe netfilter" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
