Re: [PATCH v2 2/7] i3c: Add core I3C infrastructure

From: Boris Brezillon
Date: Tue Feb 27 2018 - 16:14:29 EST


On Tue, 27 Feb 2018 20:24:43 +0000
Przemyslaw Sroka <psroka@xxxxxxxxxxx> wrote:


> > > SETDASA is simply faster than ENTDAA, but only if there is no
> > > need to collect BCR/DCR/PID of such devices. I think most
> > > applications would like to have them as an status information so
> > > after all ENTDAA can be regarded as an generic approach (unless
> > > I'm mistaken).
> >
> > Actually, we could retrieve BCR/DCR/PID (and all other relevant
> > information, like MAXDS) even with the SETDASA approach. We just
> > need to send the according CCC commands after SETDASA.
> >
> I agree, what I meant by "SETDASA is simply faster than ENTDAA, but
> only if there is no need to collect BCR/DCR/PID of such devices." Is
> that it is faster than DAA but only if not followed by GET CCC
> commands to gather BCR/DCR/PID. I think we are on the same page here.

Yes, but right now it's not the case, see my other reply ;-).

>
> > But that's also my understanding that ENTDAA should always work, and
> > SETDASA usage is only needed if you want to reserve a dynamic
> > address and assign it to a device before DAA takes place. This way
> > you can enforce the device priority (WRT IBIs). But honestly,
> > that's the only use case I can think of, and to me, it sounds like
> > an advanced feature we may want to support at some point, but don't
> > need in the initial implementation.
> Still ENTDAA seems to be sufficient for IBI prioritization but I can
> imagine some use cases where people would like to use it for such
> purposes. Note that SETDASA is applicable only for devices with SA so
> it is self-explanatory that it cannot be considered as utility to
> define priorities for all devices before ENTDAA.

We have SETNEWDA for other use cases: say you want one of your device to
have an higher priority, you can just manually set a new dynamic
address that is lower than any other devices on the bus (I plan to
expose that through sysfs, by making the address file writable).

--
Boris Brezillon, Bootlin (formerly Free Electrons)
Embedded Linux and Kernel engineering
https://bootlin.com