Message-ID: <20170410085501.GC26743@calimero.vinschen.de>
Date:   Mon, 10 Apr 2017 10:55:01 +0200
From:   Corinna Vinschen <vinschen@...hat.com>
To:     Jeff Kirsher <jeffrey.t.kirsher@...el.com>
Cc:     intel-wired-lan@...ts.osuosl.org, netdev@...r.kernel.org,
        Alexander Duyck <alexander.duyck@...il.com>,
        "Duyck, Alexander H" <alexander.h.duyck@...el.com>,
        Corinna Vinschen <vinschen@...hat.com>
Subject: Re: [Intel-wired-lan] [PATCH v2] igb: Allow to remove
 administratively set MAC on VFs

On Apr  7 12:06, Jeff Kirsher wrote:
> On Wed, 2017-04-05 at 15:46 +0200, Corinna Vinschen wrote:
> >   Before libvirt modifies the MAC address and vlan tag for an SRIOV
> > VF
> >   for use by a virtual machine (either using vfio device assignment
> > or
> >   macvtap passthru mode), it saves the current MAC address and vlan
> > tag
> >   so that it can reset them to their original value when the guest is
> >   done.  Libvirt can't leave the VF MAC set to the value used by the
> >   now-defunct guest since it may be started again later using a
> >   different VF, but it certainly shouldn't just pick any random
> > value,
> >   either. So it saves the state of everything prior to using the VF,
> > and
> >   resets it to that.
> > 
> >   The igb driver initializes the MAC addresses of all VFs to
> >   00:00:00:00:00:00, and reports that when asked (via an RTM_GETLINK
> >   netlink message, also visible in the list of VFs in the output of
> > "ip
> >   link show"). But when libvirt attempts to restore the MAC address
> > back
> >   to 00:00:00:00:00:00 (using an RTM_SETLINK netlink message) the
> > kernel
> >   responds with "Invalid argument".
> > 
> >   Forbidding a reset back to the original value leaves the VF MAC at
> > the
> >   value set for the now-defunct virtual machine. Especially on a
> > system
> >   with NetworkManager enabled, this has very bad consequences, since
> >   NetworkManager forces all interfacess to be IFF_UP all the time -
> > if
> >   the same virtual machine is restarted using a different VF (or even
> > on
> >   a different host), there will be multiple interfaces watching for
> >   traffic with the same MAC address.
> > 
> >   To allow libvirt to revert to the original state, we need a way to
> >   remove the administrative set MAC on a VF, to allow normal host
> >   operation again, and to reset/overwrite the VF MAC via VF netdev.
> > 
> >   This patch implements the outlined scenario by allowing to set the
> >   VF MAC to 00:00:00:00:00:00 via RTM_SETLINK on the PF.
> >   igb_ndo_set_vf_mac resets the IGB_VF_FLAG_PF_SET_MAC flag to 0,
> >   so it's possible to reset the VF MAC back to the original value via
> >   the VF netdev.
> > 
> >   Note: Recent patches to libvirt allow for a workaround if the NIC
> >   isn't capable of resetting the administrative MAC back to all 0,
> > but
> >   in theory the NIC should allow resetting the MAC in the first
> > place.
> > 
> > Signed-off-by: Corinna Vinschen <vinschen@...hat.com>
> > ---
> >  drivers/net/ethernet/intel/igb/igb_main.c | 42
> > +++++++++++++++++++++++--------
> >  1 file changed, 31 insertions(+), 11 deletions(-)
> 
> This patch does not apply (not even close).  Please make sure to base
> you patch off my dev-queue branch of my next-queue tree on kernel.org.

Right, I applied the patch against net-next.  Actually, only the
first hunk didn't apply to dev-queue because of the change from
igb_rar_set_qsel to igb_rar_set_index.

I'll send a v3 agains dev-queue in a bit.


Corinna

Download attachment "signature.asc" of type "application/pgp-signature" (820 bytes)

Powered by blists - more mailing lists