On 6/30/26 4:00 AM, Toke Høiland-Jørgensen wrote:

quoted

It does not make sense to require a flag to get lookup output. vlan
proto of 0 is not valid, so it is a clear indication that the vlan
output parameters were not set during the lookup.

Okay, so we could just unconditionally set the VLAN fields, but if we
start rewriting the ifindex that would be a change of the existing
behaviour that could break existing applications, no?

Consistently dealing with upper devices is one of the reasons I never
sent patches for vlan support.

xdp support is at the driver layer for real (physical) devices. The fib
lookup is going to return the vlan device index - a virtual device.
Support for xdp should not be propagated to virtual devices; it goes
against the intent of xdp. Any trip down this path will have to decide
how to handle vlan-in-vlan use cases. Where is the line drawn for fast
networking?

Specifically, if an XDP application has a table of the interfaces it
forwards between, today they'd get a VLAN interface ifindex, which would
not be in that table, and the application would return XDP_PASS. Whereas
if we change the ifindex and populate the VLAN tag, suddenly the
interface would be in the table, but because the application doesn't
read the returned VLAN tag, it will end up sending packets out without
tagging them, leading to broken forwarding.

I have not followed developments over the past few years. Does XDP have
support for vlan acceleration in the Tx path now? You really want that
to deal with vlans and not replicating s/w processing in ebpf.

So if we don't want the flag, we'd need some other mechanism to resolve
the parent ifindex, AFAICT? Maybe a xdp_get_parent_ifindex() kfunc, say?
That could also be made generic for other stacked interface types, I
suppose.

WDYT?

dealing with stacked devices is hard :-)

What is the return is a bond device or a vlan on a bond device?