Re: [PATCH net-next 1/2] net: dsa: rzn1-a5psw: enable DPBU for CPU port and fix STP states

[Clément Léger]

Le Thu, 30 Mar 2023 19:51:23 +0300,
Vladimir Oltean <olteanv@xxxxxxxxx> a écrit :

> On Thu, Mar 30, 2023 at 05:44:27PM +0200, Clément Léger wrote:
> > Le Thu, 30 Mar 2023 18:16:53 +0300,
> > Vladimir Oltean <olteanv@xxxxxxxxx> a écrit :
> >   
> > > Have you considered adding some Fixes: tags and sending to the "net" tree?  
> > 
> > I wasn't sure if due to the refactoring that should go directly to the
> > net tree but I'll do that. But since they are fixes, that's the way to
> > go.  
> 
> My common sense says that code quality comes first, and so, the code
> looks however it needs to look, keeping in mind that it still needs to
> be a punctual fix for the problem. This doesn't change the fact that
> it's a fix for an an observable bug, and so, it's a candidate for 'net'.

Agreed.

> 
> That's just my opinion though, others may disagree.
> 
> > > To be absolutely clear, when talking about BPDUs, is it applicable
> > > effectively only to STP protocol frames, or to any management traffic
> > > sent by tag_rzn1_a5psw.c which has A5PSW_CTRL_DATA_FORCE_FORWARD set?  
> > 
> > The documentation uses BPDUs but this is to be understood as in a
> > broader sense for "management frames" since it matches all the MAC with
> > "01-80-c2-00-00-XX".   
> 
> And even so, is it just for frames sent to "01-80-c2-00-00-XX", or for
> all frames sent with A5PSW_CTRL_DATA_FORCE_FORWARD? Other switch
> families can inject whatever they want into ports that are in the
> BLOCKING STP state.

Forced forwarded to disabled ports will only apply to management
frames. At least this is what the documentation says for forced
forwarding (section 4.5.5.4, Table 4.234):

Normal frames will be filtered always (i.e. can never be transmitted to
disabled ports).

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