Re: [PATCH] i2c: mpc: Use of_property_read_reg() to parse "reg"

[Rob Herring]

On Mon, Jun 12, 2023 at 3:08 PM Andi Shyti <andi.shyti@xxxxxxxxxx> wrote:
>
> Hi Rob,
>
> On Mon, Jun 12, 2023 at 01:27:03PM -0600, Rob Herring wrote:
> > On Sat, Jun 10, 2023 at 3:36 AM Andi Shyti <andi.shyti@xxxxxxxxxx> wrote:
> > >
> > > Hi Rob,
> > >
> > > On Fri, Jun 09, 2023 at 12:30:44PM -0600, Rob Herring wrote:
> > > > Use the recently added of_property_read_reg() helper to get the
> > > > untranslated "reg" address value.
> > > >
> > > > Signed-off-by: Rob Herring <robh@xxxxxxxxxx>
> > > > ---
> > > >  drivers/i2c/busses/i2c-mpc.c | 5 +++--
> > > >  1 file changed, 3 insertions(+), 2 deletions(-)
> > > >
> > > > diff --git a/drivers/i2c/busses/i2c-mpc.c b/drivers/i2c/busses/i2c-mpc.c
> > > > index cfd074ee6d54..595dce9218ad 100644
> > > > --- a/drivers/i2c/busses/i2c-mpc.c
> > > > +++ b/drivers/i2c/busses/i2c-mpc.c
> > > > @@ -316,9 +316,10 @@ static void mpc_i2c_setup_512x(struct device_node *node,
> > > >       if (node_ctrl) {
> > > >               ctrl = of_iomap(node_ctrl, 0);
> > > >               if (ctrl) {
> > > > +                     u64 addr;
> > > >                       /* Interrupt enable bits for i2c-0/1/2: bit 24/26/28 */
> > > > -                     pval = of_get_property(node, "reg", NULL);
> > > > -                     idx = (*pval & 0xff) / 0x20;
> > > > +                     of_property_read_reg(node, 0, &addr, NULL);
> > >
> > > because of_property_read_reg() can return error, can we check
> > > also the error value here?
> >
> > Why?
>
> Because if a function can return an error, the error must be
> checked. Even if the property is "reg" and the binding says that
> it's required. Otherwise let's make those functions void.

Then every function should have a must_check annotation, but they
don't as the function is designed to work with optional properties
where we want to ignore errors.

> > The old code wasn't worried about of_get_property() returning
> > NULL on the same possible errors.
>
> Sure! Checking the error comes for free. The patch is fine as it
> is, mine was a little improvement I asked for. I can still ack
> it and add the error handling later myself :)
>
> > If anyone is still actually using
> > mpc512x, I don't think their DTB will have an error at this point.
> > IOW, is improving the error handling on this really worth it?
>
> In my view, every error needs to be checked as every error is
> unlikely to happen: it makes the code future proof and makes sure
> other components failure don't impact the normal functioning of
> this driver.

An error in this case is a bad DT. It's not the kernel's job to ensure
DT is correct. If it is, then it is doing a terrible job. The reason
we have dtschema is to ensure correctness.

Rob