Re: [PATCH v3 1/3] usb: misc: onboard-hub: support multiple power supplies

[Matthias Kaehlcke]

On Wed, Jun 21, 2023 at 06:22:41PM +0200, Benjamin Bara wrote:
> On Wed, 21 Jun 2023 at 18:07, Greg Kroah-Hartman
> <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
> > On Wed, Jun 21, 2023 at 05:58:30PM +0200, Benjamin Bara wrote:
> > > From: Benjamin Bara <benjamin.bara@xxxxxxxxxxx>
> > >
> > > As some of the onboard hubs require multiple power supplies, provide the
> > > environment to support them.
> > >
> > > Signed-off-by: Benjamin Bara <benjamin.bara@xxxxxxxxxxx>
> > > ---
> > > v3:
> > > - fix nits mentioned in v2
> > >
> > > v2:
> > > - replace (err != 0) with (err)
> > > ---
> > >  drivers/usb/misc/onboard_usb_hub.c | 39 ++++++++++++++++++++++++++++++--------
> > >  drivers/usb/misc/onboard_usb_hub.h |  1 +
> > >  2 files changed, 32 insertions(+), 8 deletions(-)
> > >
> > > diff --git a/drivers/usb/misc/onboard_usb_hub.c b/drivers/usb/misc/onboard_usb_hub.c
> > > index 12fc6eb67c3b..a56e712d3a45 100644
> > > --- a/drivers/usb/misc/onboard_usb_hub.c
> > > +++ b/drivers/usb/misc/onboard_usb_hub.c
> > > @@ -27,6 +27,13 @@
> > >
> > >  #include "onboard_usb_hub.h"
> > >
> > > +#define MAX_SUPPLIES 2
> >
> > Why 2?
> 
> I picked 2 because with 3/3, this is the maximum of "required" supplies. The
> currently implemented ones require only one (up to now just named "vdd"). The
> new one added in 3/3 requires 2, therefore I tried to be generic if some future
> hub might require 3 or more.
> 
> > > +
> > > +static const char * const supply_names[] = {
> > > +     "vdd",
> > > +     "vdd2",
> > > +};
> >
> > Do those names have anything to do with the number above?  If so, please
> > document it!
> 
> I picked "vdd" for the first to be compatible with the existing device-trees. As
> the actual names differ between hubs, I thought it might be generic to just use
> "vdd2" here. If I should add some comment like "if you increase MAX_SUPPLIES,
> please also add a supply_name below", I can do that. I could also implement
> "vdd${i+1}" for i>0 instead.
> 
> >
> > >  struct onboard_hub_pdata {
> > >       unsigned long reset_us;         /* reset pulse width in us */
> > > +     unsigned int num_supplies;      /* number of supplies: 0 considered as 1 */
> >
> > I can not understand that comment at all :(
> 
> This should just indicate that leaving the field empty means one supply is
> required. Maybe "defaults to 1" is better?

Whatever the comment, I suggest to put it in parentheses, rather than after a
colon. If you keep the current style maybe s/considered/is interpreted/.

Another option would be to initialize all existing hubs to num_supplies = 1,
then there is no need for a comment and the check in _probe(). That might
be the clearest thing to do, even though it adds a few extra lines.