lkml.org 
[lkml]   [2020]   [Mar]   [19]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
Date
SubjectRe: [PATCH net] ipv6: don't auto-add link-local address to lag ports
On Wed, Mar 18, 2020 at 4:42 PM Jay Vosburgh <jay.vosburgh@canonical.com> wrote:
>
> Jarod Wilson <jarod@redhat.com> wrote:
>
> >On Wed, Mar 18, 2020 at 2:02 PM Eric Dumazet <eric.dumazet@gmail.com> wrote:
> >>
> >> On 3/18/20 7:06 AM, Jarod Wilson wrote:
> >> > Bonding slave and team port devices should not have link-local addresses
> >> > automatically added to them, as it can interfere with openvswitch being
> >> > able to properly add tc ingress.
> >> >
> >> > Reported-by: Moshe Levi <moshele@mellanox.com>
> >> > CC: Marcelo Ricardo Leitner <mleitner@redhat.com>
> >> > CC: netdev@vger.kernel.org
> >> > Signed-off-by: Jarod Wilson <jarod@redhat.com>
> >>
> >>
> >> This does not look a net candidate to me, unless the bug has been added recently ?
> >>
> >> The absence of Fixes: tag is a red flag for a net submission.
> >>
> >> By adding a Fixes: tag, you are doing us a favor, please.
> >
> >Yeah, wasn't entirely sure on this one. It fixes a problem, but it's
> >not exactly a new one. A quick look at git history suggests this might
> >actually be something that technically pre-dates the move to git in
> >2005, but only really became a problem with some additional far more
> >recent infrastructure (tc and friends). I can resubmit it as net-next
> >if that's preferred.
>
> Commit
>
> c2edacf80e15 bonding / ipv6: no addrconf for slaves separately from master
>
> should (in theory) already prevent ipv6 link-local addrconf, at
> least for bonding slaves, and dates from 2007. If something has changed
> to break the logic in this commit, then (a) you might need to do some
> research to find a candidate for your Fixes tag, and (b) I'd suggest
> also investigating whether or not the change added by c2edacf80e15 to
> addrconf_notify() no longer serves any purpose, and should be removed if
> that is the case.
>
> Note also that the hyperv netvsc driver, in netvsc_vf_join(),
> sets IFF_SLAVE in order to trigger the addrconf prevention logic from
> c2edacf80e15; I'm not sure if your patch would affect its expectations
> (if c2edacf80e15 were removed).

Interesting. We'll keep digging over here, but that's definitely not
working for this particular use case with OVS for whatever reason.

--
Jarod Wilson
jarod@redhat.com

\
 
 \ /
  Last update: 2020-03-19 17:43    [W:0.104 / U:0.888 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site