[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20210120174644.083de7fa@kicinski-fedora-pc1c0hjn.dhcp.thefacebook.com>
Date: Wed, 20 Jan 2021 17:46:51 -0800
From: Jakub Kicinski <kuba@...nel.org>
To: Ido Schimmel <idosch@...sch.org>
Cc: Hangbin Liu <liuhangbin@...il.com>,
Ido Schimmel <idosch@...dia.com>, netdev@...r.kernel.org,
Jiri Pirko <jiri@...dia.com>,
Nikolay Aleksandrov <nikolay@...dia.com>,
"David S . Miller" <davem@...emloft.net>
Subject: Re: [PATCH net] selftests/net: set link down before enslave
On Wed, 20 Jan 2021 21:43:28 +0200 Ido Schimmel wrote:
> On Wed, Jan 20, 2021 at 10:38:47PM +0800, Hangbin Liu wrote:
> > Hi Ido,
> >
> > On Wed, Jan 20, 2021 at 12:42:10PM +0200, Ido Schimmel wrote:
> > > > diff --git a/tools/testing/selftests/net/rtnetlink.sh b/tools/testing/selftests/net/rtnetlink.sh
> > > > index c9ce3dfa42ee..a26fddc63992 100755
> > > > --- a/tools/testing/selftests/net/rtnetlink.sh
> > > > +++ b/tools/testing/selftests/net/rtnetlink.sh
> > > > @@ -1205,6 +1205,8 @@ kci_test_bridge_parent_id()
> > > > dev20=`ls ${sysfsnet}20/net/`
> > > >
> > > > ip link add name test-bond0 type bond mode 802.3ad
> > > > + ip link set dev $dev10 down
> > > > + ip link set dev $dev20 down
> > >
> > > But these netdevs are created with their administrative state set to
> > > 'DOWN'. Who is setting them to up?
> >
> > Would you please point me where we set the state to 'DOWN'? Cause on my
> > host it is init as UP:
> >
> > ++ ls /sys/bus/netdevsim/devices/netdevsim10/net/
> > + dev10=eth3
> > ++ ls /sys/bus/netdevsim/devices/netdevsim20/net/
> > + dev20=eth4
> > + ip link add name test-bond0 type bond mode 802.3ad
> > + ip link show eth3
> > 66: eth3: <BROADCAST,NOARP,UP,LOWER_UP> mtu 1500 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1000
> > link/ether 1e:52:27:5f:a5:3c brd ff:ff:ff:ff:ff:ff
>
> I didn't have time to look into this today, but I suspect the problem is
> either:
>
> 1. Some interface manager on your end that is setting these interfaces
> up after they are created
This must be the case, the kernel doesn't open/up devices by itself.
> 2. A bug in netdevsim that does not initialize the carrier to off.
> Maybe try with this patch (didn't test):
Yeah, but that's fine, SW devices don't have to manage carrier state.
Powered by blists - more mailing lists