Bug 1545574

I tried the following configuration using teamd directly without NM on
a PPC64 machine:

{
 	"device":       "team0",
	"runner":       {
                        "name": "activebackup",
			"hwaddr_policy": "by_active"
	},
        "ports":        {
		"veth0": {
                        "prio": -10,
			"sticky": true
		},
                "veth2": {
                        "prio": 100
		}
	}
}

and indeed the second port becomes the active one, while on x86_64 it
is the first port. I'm not sure this is expected, I'm reassigning the bug
to teamd for investigation. I'll attach activation logs and the teamd
state dump on PPC64.

Created attachment 1488355 [details]
teamd log

Created attachment 1488356 [details]
teamd state dump

update: After debug, it turns out that function team_get_active_port() could not get current active port and set it to 0. Then in function ab_link_watch_handler(), it clears the current active port veth0 and set to veth2.

Function get_options() also could not get correct priority. We can use examples/team_manual_control.c to reproduce this issue easily.

The odd thing is that I wrote a program with libnl and it works well. I can get correct active port and priority.

Via print kernel message, I found libteam send request correctly and kernel replies correct message. So there should be something wrong when libteam receive and deal with the data.

Hi Xin Long,

The patch[1] has been applied for upstream. Please help back port it to fix the bug.

[1] https://github.com/jpirko/libteam/commit/c35bece57a499036

Thanks
Hangbin

Hi Hangbin,

1. Does this only occur on ppc6e arch?

2. From comment #0, i see this relevant to hwaddr_policy, but from comment #3, i see this is a active port relevant issue, so how should i reprodece and test this bz?  Would you please give me a reproducer?

(In reply to LiLiang from comment #8)
> Hi Hangbin,
> 
> 1. Does this only occur on ppc6e arch?

yes
> 
> 2. From comment #0, i see this relevant to hwaddr_policy, but from comment
> #3, i see this is a active port relevant issue, so how should i reprodece
> and test this bz?  Would you please give me a reproducer?

It's a active port relevant issue. But you can just use the config file in comment 3 to reproduce.

reproduced:
[root@ibm-p720-02-lp2 ~]# teamdctl team0 state
setup:
  runner: activebackup
ports:
  veth0
    link watches:
      link summary: up
      instance[link_watch_0]:
        name: ethtool
        link: up
        down count: 0
  veth2
    link watches:
      link summary: up
      instance[link_watch_0]:
        name: ethtool
        link: up
        down count: 0
runner:
  active port: veth2
[root@ibm-p720-02-lp2 ~]# 
[root@ibm-p720-02-lp2 ~]# 
[root@ibm-p720-02-lp2 ~]# rpm -q libteam
libteam-1.27-4.el7.ppc64

verified:
[root@ibm-p720-02-lp2 ~]# teamdctl team0 state
setup:
  runner: activebackup
ports:
  veth0
    link watches:
      link summary: up
      instance[link_watch_0]:
        name: ethtool
        link: up
        down count: 0
  veth2
    link watches:
      link summary: up
      instance[link_watch_0]:
        name: ethtool
        link: up
        down count: 0
runner:
  active port: veth0
[root@ibm-p720-02-lp2 ~]# rpm -q libteam
libteam-1.27-9.el7.ppc64

Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory, and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://access.redhat.com/errata/RHBA-2019:2310