On 2022-10-18 08:22, Ido Schimmel wrote:
On Thu, Oct 13, 2022 at 08:58:57PM +0200, netdev@xxxxxxxxxxxxxxxxxxxx
wrote:
On 2022-10-13 16:06, Ido Schimmel wrote:
> > diff --git a/net/dsa/port.c b/net/dsa/port.c
> > index e4a0513816bb..eab32b7a945a 100644
> > --- a/net/dsa/port.c
> > +++ b/net/dsa/port.c
> > @@ -304,7 +304,7 @@ static int dsa_port_inherit_brport_flags(struct
> > dsa_port *dp,
> > struct netlink_ext_ack *extack)
> > {
> > const unsigned long mask = BR_LEARNING | BR_FLOOD | BR_MCAST_FLOOD |
> > - BR_BCAST_FLOOD | BR_PORT_LOCKED;
> > + BR_BCAST_FLOOD;
>
> Not sure how this is related to the patchset.
>
In general it is needed as a fix because of the way learning with
locked
port is handled in the driver,
so as with MAB and also locked port in the future needing a non-zero
Port
Association Vector (PAV)
for refresh etc to work, inheritance of the locked port flag is a bad
idea
(say bug) and shouldn't have
been in the first place.
If it's a fix, then it needs to be submitted to 'net' tree.
It is a 'fix' for this patch set, as it changes
mv88e6xxx_port_set_lock() to need this change.
It is not strictly necessary to change it for earlier hehaviour.
