Re: [PATCH net-next v3 3/3] net: dsa: rzn1-a5psw: add vlan support

["Clément Léger" <clement.leger@bootlin.com>]

Le Thu, 9 Feb 2023 00:02:19 +0200,
Vladimir Oltean <olteanv@gmail.com> a écrit :

> On Wed, Feb 08, 2023 at 09:38:04AM -0800, Florian Fainelli wrote:
> > > +	/* Enable TAG always mode for the port, this is actually controlled
> > > +	 * by VLAN_IN_MODE_ENA field which will be used for PVID insertion
> > > +	 */
> > > +	reg = A5PSW_VLAN_IN_MODE_TAG_ALWAYS;
> > > +	reg <<= A5PSW_VLAN_IN_MODE_PORT_SHIFT(port);
> > > +	a5psw_reg_rmw(a5psw, A5PSW_VLAN_IN_MODE, A5PSW_VLAN_IN_MODE_PORT(port),
> > > +		      reg);  
> > 
> > If we always enable VLAN mode, which VLAN ID do switch ports not part of a
> > VLAN aware bridge get classified into?  
> 
> Good question. I'd guess 0, since otherwise, the VLAN-unaware FDB
> entries added with a5psw_port_fdb_add() wouldn't work.

The name of the mode is probably missleading. When setting VLAN_IN_MODE
with A5PSW_VLAN_IN_MODE_TAG_ALWAYS, the input packet will be tagged
_only_ if VLAN_IN_MODE_ENA port bit is set. If this bit is not set,
packet will passthrough transparently. This bit is actually enabled in
a5psw_port_vlan_add() when a PVID is set and unset when the PVID is
removed. Maybe the comment above these lines was not clear enough.

> 
> But the driver has to survive the following chain of commands, which, by
> looking at the current code structure, it doesn't:
> 
> ip link add br0 type bridge vlan_filtering 0
> ip link set swp0 master br0 # PVID should remain at a value chosen privately by the driver
> bridge vlan add dev swp0 vid 100 pvid untagged # PVID should not change in hardware yet
> ip link set br0 type bridge vlan_filtering 1 # PVID should change to 100 now
> ip link set br0 type bridge vlan_filtering 0 # PVID should change to the value chosen by the driver
> 
> Essentially, what I'm saying is that VLANs added with "bridge vlan add"
> should only be active while vlan_filtering=1.
> 
> If you search for "commit_pvid" in drivers/net/dsa, you'll find a number
> of drivers which have a more elaborate code structure which allows the
> commands above to work properly.



-- 
Clément Léger,
Embedded Linux and Kernel engineer at Bootlin
https://bootlin.com