Re: [syzbot ci] Re: net: move netdev_compute_master_upper_features to ndo_set_features

[Sabrina Dubroca <sd@queasysnail.net>]

2026-03-12, 17:47:45 +0100, Paolo Abeni wrote:
> 
> 
> On 3/12/26 4:58 PM, Sabrina Dubroca wrote:
> > 2026-03-12, 14:34:44 +0000, Hangbin Liu wrote:
> >> On Thu, Mar 12, 2026 at 12:13:52PM +0100, Sabrina Dubroca wrote:
> >>> Proper fix (so that the notification we're sending during
> >>> upper_dev_link has full linkinfo) would be to move
> >>> netdev_upper_dev_link() to after macsec_changelink_common() and fix up
> >>> the error handling. I don't see anything in macsec_add_dev or
> >>> macsec_changelink_common that needs the device to be linked. But
> >>
> >> If we move the netdev_upper_dev_link() after macsec_changelink_common(),
> >> we will not goto nla_put_failure via default, right?
> > 
> > Yes.
> > 
> >>> anyway it doesn't make sense for macsec_fill_info to return -EMSGSIZE
> >>> on invalid data, so the "bandaid" should be included as well.
> >>>
> >>> Should this be part of this series (either just the "bandaid" or the
> >>> "proper fix"+bandaid), since we never saw a problem before?
> >>
> >> Since macsec need the "bandaid" fix either way. How about you post the
> >> "bandaid" fix to net. And I include the "proper fix" in this series for
> >> net-next?
> > 
> > But I don't think it's needed in net. Am I missing a codepath (before
> > your series) where macsec_fill_info could be called for the new device
> > before macsec_newlink returns? If not, it doesn't really qualify as a
> > fix, that's why I was asking Paolo.
> 
> FWIW, I don't see a codepath calling into rtmsg_ifinfo_build_skb()
> before device initialization, so I would be fine targeting net-next with
> the EMSGSIZE-related change.
> 
> Side note, it looks like that the WARN() in the rnnetlink code here
> helped identifying a real problem and correctly returning 0 when the
> key_len is not yet initialized will silence it forever, what about
> preserving a warning for this kind of race? something alike:

Right, I was thinking about putting a DEBUG_NET_WARN_ON_ONCE there.

-- 
Sabrina