Linux Ethernet Bridging - Jan 2023 - [Bridge] [RFC PATCH net-next 2/5] net: dsa: propagate flags down towards drivers

On 2023-01-18 00:17, Vladimir Oltean wrote:
> On Tue, Jan 17, 2023 at 07:57:11PM +0100, Hans J. Schultz wrote:
>> Dynamic FDB flag needs to be propagated through the DSA layer to be
>> added to drivers.
>> Use a u16 for fdb flags for future use, so that other flags can also 
>> be
>> sent the same way without having to change function interfaces.
>> 
>> Signed-off-by: Hans J. Schultz <netdev at kapio-technology.com>
>> ---
>> @@ -3364,6 +3368,7 @@ static int dsa_slave_fdb_event(struct net_device 
>> *dev,
>>  	struct dsa_port *dp = dsa_slave_to_port(dev);
>>  	bool host_addr = fdb_info->is_local;
>>  	struct dsa_switch *ds = dp->ds;
>> +	u16 fdb_flags = 0;
>> 
>>  	if (ctx && ctx != dp)
>>  		return 0;
>> @@ -3410,6 +3415,9 @@ static int dsa_slave_fdb_event(struct net_device 
>> *dev,
>>  		   orig_dev->name, fdb_info->addr, fdb_info->vid,
>>  		   host_addr ? " as host address" : "");
>> 
>> +	if (fdb_info->is_dyn)
>> +		fdb_flags |= DSA_FDB_FLAG_DYNAMIC;
>> +
> 
> Hmm, I don't think this is going to work with the 
> assisted_learning_on_cpu_port
> feature ("if (switchdev_fdb_is_dynamically_learned(fdb_info))").
The
> reason being
> that a "dynamically learned" FDB entry (defined as this):
> 
> static inline bool
> switchdev_fdb_is_dynamically_learned(const struct
> switchdev_notifier_fdb_info *fdb_info)
> {
> 	return !fdb_info->added_by_user && !fdb_info->is_local;
> }
> 
> is also dynamic in the DSA_FDB_FLAG_DYNAMIC sense. But we install a
> static FDB entry for it on the CPU port.
> 
> And in your follow-up patch 3/5, you make all drivers except mv88e6xxx
> ignore all DSA_FDB_FLAG_DYNAMIC entries (including the ones snooped 
> from
> address learning on software interfaces). So this breaks those drivers
> which don't implement DSA_FDB_FLAG_DYNAMIC but do set
> ds->assisted_learning_on_cpu_port
> to true.
I am not sure I understand you entirely.
 From my standpoint I see it as so: that until now any fdb entry coming 
to port_fdb_add() (or port_fdb_del()) are seen as static entries. And 
this changes nothing with respect to those static entries as how drivers 
handle them.
When the new dynamic flag is true, all drivers will ignore it in patch 
#3, so basically nothing will change by that. Then in patch #5 the 
dynamic flag is handled by the mv88e6xxx driver.

I don't know the assisted_learning_on_cpu_port feature you mention, but 
there has still not been anything but static entries going towards 
port_fdb_add() yet...
> 
> I think you also want to look at the added_by_user flag to disambiguate
> between a dynamic FDB entry added from learning (which it's ok to
> offload as static, because software ageing will remove it) and one 
> added
> by the user.
> 
>>  	INIT_WORK(&switchdev_work->work,
dsa_slave_switchdev_event_work);
>>  	switchdev_work->event = event;
>>  	switchdev_work->dev = dev;
>> @@ -3418,6 +3426,7 @@ static int dsa_slave_fdb_event(struct net_device 
>> *dev,
>>  	ether_addr_copy(switchdev_work->addr, fdb_info->addr);
>>  	switchdev_work->vid = fdb_info->vid;
>>  	switchdev_work->host_addr = host_addr;
>> +	switchdev_work->fdb_flags = fdb_flags;
>> 
>>  	dsa_schedule_work(&switchdev_work->work);
>>

On Wed, Jan 18, 2023 at 11:35:08PM +0100, netdev at kapio-technology.com
wrote:
> I am not sure I understand you entirely.
> From my standpoint I see it as so: that until now any fdb entry coming to
> port_fdb_add() (or port_fdb_del()) are seen as static entries. And this
> changes nothing with respect to those static entries as how drivers handle
> them.
This is true; it is implicit that the port_fdb_add() and port_fdb_del()
DSA methods request switches to operate on static FDB entries (in hardware).
> When the new dynamic flag is true, all drivers will ignore it in patch #3,
> so basically nothing will change by that.
This is not true, because it assumes that DSA never called port_fdb_add()
up until now for bridge FDB entries with the BR_FDB_STATIC flag unset,
which is incorrect (it did).

So what will change is that drivers which used to react to those bridge
FDB entries will stop doing so.
> Then in patch #5 the dynamic flag is handled by the mv88e6xxx driver.
> 
> I don't know the assisted_learning_on_cpu_port feature you mention, but
> there has still not been anything but static entries going towards
> port_fdb_add() yet...
For starters, you can read the commit message of the patch that
introduced it, which is d5f19486cee7 ("net: dsa: listen for
SWITCHDEV_{FDB,DEL}_ADD_TO_DEVICE on foreign bridge neighbors").