Messages in this thread | | | Date | Wed, 26 Jul 2023 15:15:20 +0200 | From | Simon Horman <> | Subject | Re: [net-next PATCH 2/3] net: dsa: qca8k: make learning configurable and keep off if standalone |
| |
On Wed, Jul 26, 2023 at 03:14:35PM +0300, Vladimir Oltean wrote: > On Wed, Jul 26, 2023 at 10:19:34AM +0200, Simon Horman wrote: > > > +int qca8k_port_pre_bridge_flags(struct dsa_switch *ds, int port, > > > + struct switchdev_brport_flags flags, > > > + struct netlink_ext_ack *extack) > > > +{ > > > + if (flags.mask & ~BR_LEARNING) > > > + return -EINVAL; > > > > If I am reading things right then some implementation of this callback > > return -EINVAL when they see unexpected flags. And some seem not to > > - possibly because all flags are expected. > > > > So I'm slightly unsure if this is correct or not. > > Which ones don't? All handlers of SWITCHDEV_ATTR_ID_PORT_PRE_BRIDGE_FLAGS > should return -EINVAL for changes made to bridge port flags that aren't > supported.
Sorry, on a second look I see that my statement above is incorrect. I do wonder what it was I saw this morning. But this afternoon I see that all users check flags and return -EINVAL as appropriate.
| |