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

On 2023-01-31 19:56, Simon Horman wrote:
>> --- a/drivers/net/dsa/mv88e6xxx/chip.c
>> +++ b/drivers/net/dsa/mv88e6xxx/chip.c
>> @@ -42,6 +42,7 @@
>>  #include "ptp.h"
>>  #include "serdes.h"
>>  #include "smi.h"
>> +#include "switchdev.h"
>> 
>>  static void assert_reg_lock(struct mv88e6xxx_chip *chip)
>>  {
>> @@ -2726,18 +2727,25 @@ static int mv88e6xxx_port_fdb_add(struct 
>> dsa_switch *ds, int port,
>>  				  const unsigned char *addr, u16 vid,
>>  				  u16 fdb_flags, struct dsa_db db)
>>  {
>> +	bool is_dynamic = !!(fdb_flags & DSA_FDB_FLAG_DYNAMIC);
>>  	struct mv88e6xxx_chip *chip = ds->priv;
>> +	u8 state;
>>  	int err;
>> 
>> -	/* Ignore entries with flags set */
>> -	if (fdb_flags)
>> -		return 0;
>> +	state = MV88E6XXX_G1_ATU_DATA_STATE_UC_STATIC;
>> +	if (is_dynamic)
>> +		state = MV88E6XXX_G1_ATU_DATA_STATE_UC_AGE_7_NEWEST;
> 
> What if flags other than DSA_FDB_FLAG_DYNAMIC are set (in future)?
They will have to be caught and handled here if there is support for it, 
e.g. something like...

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.
> 
>> +	else
>> +		if (fdb_flags)
> 
> nit: else if (fdb_flags)
> 
>> +			return 0;
>> 
> 
> ...

On Thu, Feb 02, 2023 at 06:00:00PM +0100, netdev at kapio-technology.com
wrote:
> On 2023-01-31 19:56, Simon Horman wrote:
> > > --- a/drivers/net/dsa/mv88e6xxx/chip.c
> > > +++ b/drivers/net/dsa/mv88e6xxx/chip.c
> > > @@ -42,6 +42,7 @@
> > >  #include "ptp.h"
> > >  #include "serdes.h"
> > >  #include "smi.h"
> > > +#include "switchdev.h"
> > > 
> > >  static void assert_reg_lock(struct mv88e6xxx_chip *chip)
> > >  {
> > > @@ -2726,18 +2727,25 @@ static int mv88e6xxx_port_fdb_add(struct
> > > dsa_switch *ds, int port,
> > >  				  const unsigned char *addr, u16 vid,
> > >  				  u16 fdb_flags, struct dsa_db db)
> > >  {
> > > +	bool is_dynamic = !!(fdb_flags & DSA_FDB_FLAG_DYNAMIC);
> > >  	struct mv88e6xxx_chip *chip = ds->priv;
> > > +	u8 state;
> > >  	int err;
> > > 
> > > -	/* Ignore entries with flags set */
> > > -	if (fdb_flags)
> > > -		return 0;
> > > +	state = MV88E6XXX_G1_ATU_DATA_STATE_UC_STATIC;
> > > +	if (is_dynamic)
> > > +		state = MV88E6XXX_G1_ATU_DATA_STATE_UC_AGE_7_NEWEST;
> > 
> > What if flags other than DSA_FDB_FLAG_DYNAMIC are set (in future)?
> 
> They will have to be caught and handled here if there is support for it,
> e.g. something like...
> 
> 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.
> > 
> > > +	else
> > > +		if (fdb_flags)
> > 
> > nit: else if (fdb_flags)
> > 
> > > +			return 0;
> > > 
> > 
> > ...

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.