[ovs-discuss] userspace datapath (for MPLS and VLAN tagging) seems to be broken in 2.3.1

[Ben Pfaff]

OK, since it works when vswitchd is freshly started, that means it's a
bug in switching a bridge at runtime between the kernel and userspace
datapaths.  That doesn't get much testing so problems do come up from
time to time.

You restart vswitchd with its init script, which is probably
/etc/init.d/openvswitch or /etc/init.d/openvswitch-switch.

You can confirm that it's a netdev device by running "ovs-dpctl show";
if your devices don't show up there then it's not kernel-based.

On Sun, Feb 22, 2015 at 10:01:55PM +0100, Andrei Manescu wrote:
> Didn’t find how to gracefully restart vswitchd (no init script). After rebooting the hypervisor, it’s gone. I tried adding a new switch and set the dp type at creation and it still works.
> root at pve:~# ovs-vsctl add-br vmbr2  -- set bridge vmbr2 datapath_type=netdev
> root at pve:~# ovs-ofctl show vmbr2
> OFPT_FEATURES_REPLY (xid=0x2): dpid:00006a2ebc0d894a
> n_tables:254, n_buffers:256
> capabilities: FLOW_STATS TABLE_STATS PORT_STATS QUEUE_STATS ARP_MATCH_IP
> actions: OUTPUT SET_VLAN_VID SET_VLAN_PCP STRIP_VLAN SET_DL_SRC SET_DL_DST SET_NW_SRC SET_NW_DST SET_NW_TOS SET_TP_SRC SET_TP_DST ENQUEUE
>  LOCAL(vmbr2): addr:6a:2e:bc:0d:89:4a
>      config:     PORT_DOWN
>      state:      LINK_DOWN
>      current:    10MB-FD COPPER
>      speed: 10 Mbps now, 0 Mbps max
> OFPT_GET_CONFIG_REPLY (xid=0x4): frags=normal miss_send_len=0
> 
> 
> P.S.: Anyway I can restart vswitchd ? From ovs-vsctl maybe ? How do I confirm it’s still a “netdev” device ? I see MPLS label is being pushed so I don’t doubt that. Just learning bit by bit.
> 
> 
> Would you
> Thanks.
> 
> 
> > On 22 Feb 2015, at 21:42, Ben Pfaff <blp at nicira.com> wrote:
> > 
> > Does it make any difference if you set the datapath type as part of
> > creating the bridge instead of afterward?  Or if you restart vswitchd
> > after setting the datapath type?
> > 
> > On Sun, Feb 22, 2015 at 09:40:22PM +0100, Andrei Manescu wrote:
> >> Nope. That’s the whole thing AFTER I set the ‘netdev’ switch:
> >> 
> >> 
> >> 
> >> 
> >> 
> >> 
> >> 
> >> 
> >> 
> >> 
> >> 
> >> 
> >> 2015-02-22T20:39:05.099Z|00018|ofproto_dpif|INFO|netdev at ovs-netdev: Datapath supports recirculation
> >> 2015-02-22T20:39:05.099Z|00019|ofproto_dpif|INFO|netdev at ovs-netdev: MPLS label stack length probed as 3
> >> 2015-02-22T20:39:05.099Z|00020|bridge|INFO|bridge vmbr1: added interface vmbr1 on port 65534
> >> 2015-02-22T20:39:05.099Z|00021|netdev_linux|WARN|vmbr1: obtaining netdev stats via vport failed (No such device)
> >> 2015-02-22T20:39:05.102Z|00022|bridge|INFO|bridge vmbr1: added interface eth1 on port 1
> >> 2015-02-22T20:39:05.102Z|00023|netdev_linux|WARN|eth1: obtaining netdev stats via vport failed (No such device)
> >> 2015-02-22T20:39:05.102Z|00024|bridge|INFO|bridge vmbr1: using datapath ID 0000000c29e93542
> >> 2015-02-22T20:39:05.102Z|00025|connmgr|INFO|vmbr1: added service controller "punix:/var/run/openvswitch/vmbr1.mgmt"
> >> 2015-02-22T20:39:08.079Z|00026|netdev_linux|WARN|vmbr1: obtaining netdev stats via vport failed (No such device)
> >> 2015-02-22T20:39:08.079Z|00027|netdev_linux|WARN|eth1: obtaining netdev stats via vport failed (No such device)
> >> 
> >> Anyway specific debugs I can enable ?
> >> 
> >> 
> >> 
> >> 
> >> 
> >>> On 22 Feb 2015, at 21:35, Ben Pfaff <blp at nicira.com> wrote:
> >>> 
> >>> The log is probably longer than that, can you provide the whole thing?
> >> 
> > 
> > 
>