Re: [PATCH 3/4] usb: typec: add TUSB320xA driver

[Heikki Krogerus <heikki.krogerus@linux.intel.com>]

On Mon, Mar 07, 2022 at 10:17:04PM +0000, Alvin Šipraga wrote:
> Hi Heikki,
> 
> Heikki Krogerus <heikki.krogerus@linux.intel.com> writes:
> 
> > Hi,
> >
> > On Tue, Mar 01, 2022 at 02:20:07PM +0100, Alvin Šipraga wrote:
> >> From: Alvin Šipraga <alsi@bang-olufsen.dk>
> >> 
> >> The TUSB320LA and TUSB320HA (or LAI, HAI) chips are I2C controlled
> >> non-PD Type-C port controllers. They support detection of cable
> >> orientation, port attachment state, and role, including Audio Accessory
> >> and Debug Accessory modes. Add a typec class driver for this family.
> >> 
> >> Note that there already exists an extcon driver for the TUSB320 (a
> >> slightly older revision that does not support setting role preference or
> >> disabling the CC state machine). This driver is loosely based on that
> >> one.
> >
> > This looked mostly OK to me. There is one question below.
> >
> > <snip>
> >
> >> +static int tusb320xa_check_signature(struct tusb320xa *tusb)
> >> +{
> >> +	static const char sig[] = { '\0', 'T', 'U', 'S', 'B', '3', '2', '0' };
> >> +	unsigned int val;
> >> +	int i, ret;
> >> +
> >> +	mutex_lock(&tusb->lock);
> >> +
> >> +	for (i = 0; i < sizeof(sig); i++) {
> >> +		ret = regmap_read(tusb->regmap, sizeof(sig) - 1 - i, &val);
> >> +		if (ret)
> >> +			goto done;
> >> +
> >> +		if (val != sig[i]) {
> >> +			dev_err(tusb->dev, "signature mismatch!\n");
> >> +			ret = -ENODEV;
> >> +			goto done;
> >> +		}
> >> +	}
> >> +
> >> +done:
> >> +	mutex_unlock(&tusb->lock);
> >> +
> >> +	return ret;
> >> +}
> >
> > Couldn't that be done with a single read?
> >
> >         char sig[8];
> >         u64 val;
> >
> >         strcpy(sig, "TUSB320")
> >
> >         mutex_lock(&tusb->lock);
> >
> >         ret = regmap_raw_read(tusb->regmap, 0, &val, sizeof(val));
> >         ...
> >         if (val != cpu_to_le64(*(u64 *)sig)) {
> >         ...
> >
> > Something like that?
> 
> I think it's a bit cryptic - are you sure it's worth it just to save 8
> one-off regmap_read()s? I could also just remove this check... I see it
> mostly as a courtesy to the user in case the I2C address in his device
> tree mistakenly points to some other unsuspecting chip.
> 
> BTW, do you have any feedback on the device tree bindings of this
> series? Rob had some questions and I am not sure that my proposed
> bindings are fully aligned with the typec subsystem expectations. Any
> feedback would be welcome.

I don't think I understand DT well enough to comment. I'm not
completely sure what he's asking..

> I will wait for more comments and send a v2 in ~a week.

thanks,

-- 
heikki