Merge branch 'macvlan-broadcast-queue-bypass'
authorDavid S. Miller <davem@davemloft.net>
Wed, 29 Mar 2023 08:03:33 +0000 (09:03 +0100)
committerDavid S. Miller <davem@davemloft.net>
Wed, 29 Mar 2023 08:03:33 +0000 (09:03 +0100)
commit37018b5a291032b8c8c9f5e1b6292fff75cd9329
treebea8507cb6952108f9fba4decbc3ecd19cdcf6eb
parent6fc5f5bcc0c34138d955a51642f1c4fbb353f0af
parent954d1fa1ac93aa8a66f7d9a9ba545cf7f020d348
Merge branch 'macvlan-broadcast-queue-bypass'

Herbert Xu says:

====================
macvlan: Allow some packets to bypass broadcast queue

This patch series allows some packets to bypass the broadcast
queue on receive.  Currently all multicast packets are queued
on receive and then processed in a work queue.  This is to avoid
an unbounded amount of work occurring in the receive path, as
one broadcast packet could easily translate into 4,000 packets.

However, for multicast packets with just one receiver (possible
for IPv6 ND), this introduces unnecessary latency as the packet
will go to exactly one device.

This series allows such multicast packets to be processed inline.
It also adds a toggle which lets the admin control what threshold
to set between queueing and not queueing.  A follow-up patch for
iproute will be posted.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>