Linux Ethernet Bridging - Feb 2023 - [Bridge] [PATCH net-next 5/5] net: dsa: mv88e6xxx: implementation of dynamic ATU entries

On 2023-02-04 09:12, Simon Horman wrote:
> On Fri, Feb 03, 2023 at 10:44:22PM +0200, Vladimir Oltean wrote:
>> On Fri, Feb 03, 2023 at 09:20:22AM +0100, Simon Horman wrote:
>> > > else if (someflag)
>> > >         dosomething();
>> > >
>> > > For now only one flag will actually be set and they are
mutually exclusive,
>> > > as they will not make sense together with the potential flags
I know, but
>> > > that can change at some time of course.
>> >
>> > Yes, I see that is workable. I do feel that checking for other
flags would
>> > be a bit more robust. But as you say, there are none. So whichever
>> > approach you prefer is fine by me.
>> 
>> The model we have for unsupported bits in the 
>> SWITCHDEV_ATTR_ID_PORT_PRE_BRIDGE_FLAGS
>> and SWITCHDEV_ATTR_ID_PORT_BRIDGE_FLAGS handlers is essentially this:
>> 
>> 	if (flags & ~(supported_flag_mask))
>> 		return -EOPNOTSUPP;
>> 
>> 	if (flags & supported_flag_1)
>> 		...
>> 
>> 	if (flags & supported_flag_2)
>> 		...
>> 
>> I suppose applying this model here would address Simon's
extensibility
>> concern.
> 
> Yes, that is the model I had in mind.
The only thing is that we actually need to return both 0 and -EOPNOTSUPP 
for unsupported flags. The dynamic flag requires 0 when not supported 
(and supported) AFAICS.
Setting a mask as 'supported' for a feature that is not really supported
defeats the notion of 'supported' IMHO.

On Sat, Feb 04, 2023 at 09:48:24AM +0100, netdev at kapio-technology.com
wrote:
> On 2023-02-04 09:12, Simon Horman wrote:
> > On Fri, Feb 03, 2023 at 10:44:22PM +0200, Vladimir Oltean wrote:
> > > On Fri, Feb 03, 2023 at 09:20:22AM +0100, Simon Horman wrote:
> > > > > else if (someflag)
> > > > >         dosomething();
> > > > >
> > > > > For now only one flag will actually be set and they are
mutually exclusive,
> > > > > as they will not make sense together with the potential
flags I know, but
> > > > > that can change at some time of course.
> > > >
> > > > Yes, I see that is workable. I do feel that checking for
other flags would
> > > > be a bit more robust. But as you say, there are none. So
whichever
> > > > approach you prefer is fine by me.
> > > 
> > > The model we have for unsupported bits in the
> > > SWITCHDEV_ATTR_ID_PORT_PRE_BRIDGE_FLAGS
> > > and SWITCHDEV_ATTR_ID_PORT_BRIDGE_FLAGS handlers is essentially
this:
> > > 
> > > 	if (flags & ~(supported_flag_mask))
> > > 		return -EOPNOTSUPP;
> > > 
> > > 	if (flags & supported_flag_1)
> > > 		...
> > > 
> > > 	if (flags & supported_flag_2)
> > > 		...
> > > 
> > > I suppose applying this model here would address Simon's
> > > extensibility concern.
> > 
> > Yes, that is the model I had in mind.
> 
> The only thing is that we actually need to return both 0 and -EOPNOTSUPP
for
> unsupported flags. The dynamic flag requires 0 when not supported (and
> supported) AFAICS.
> Setting a mask as 'supported' for a feature that is not really
supported
> defeats the notion of 'supported' IMHO.
Just to clarify my suggestion one last time, it would be along the lines
of the following (completely untested!). I feel that it robustly covers
all cases for fdb_flags. And as a bonus doesn't need to be modified
if other (unsupported) flags are added in future.

	if (fdb_flags & ~(DSA_FDB_FLAG_DYNAMIC))
		return -EOPNOTSUPP;

	is_dynamic = !!(fdb_flags & DSA_FDB_FLAG_DYNAMIC)
	if (is_dynamic)
		state = MV88E6XXX_G1_ATU_DATA_STATE_UC_AGE_7_NEWEST;


And perhaps for other drivers:

	if (fdb_flags & ~(DSA_FDB_FLAG_DYNAMIC))
		return -EOPNOTSUPP;
	if (fdb_flags)
		return 0;

Perhaps a helper would be warranted for the above.

But in writing this I think that, perhaps drivers could return -EOPNOTSUPP
for the DSA_FDB_FLAG_DYNAMIC case and the caller can handle, rather tha
propagate, -EOPNOTSUPP.

Returning -EOPNOTSUPP is the normal way to drivers to respond to requests
for unsupported hardware offloads. Sticking to that may be clearner
in the long run. That said, I do agree your current patch is correct
given the flag that is defined (by your patchset).