Hi Uwe,
On Tue, Jan 18, 2022 at 10:09 AM Uwe Kleine-König
<u.kleine-koenig@xxxxxxxxxxxxxx> wrote:
> On Tue, Jan 18, 2022 at 09:25:01AM +0100, Geert Uytterhoeven wrote:
> > On Mon, Jan 17, 2022 at 6:06 PM Uwe Kleine-König
> > <u.kleine-koenig@xxxxxxxxxxxxxx> wrote:
> > > On Mon, Jan 17, 2022 at 02:08:19PM +0100, Geert Uytterhoeven wrote:
> > > > On Mon, Jan 17, 2022 at 12:49 PM Uwe Kleine-König
> > > > <u.kleine-koenig@xxxxxxxxxxxxxx> wrote:
> > > > > > So there are three reasons: because the absence of an optional IRQ
> > > > > > is not an error, and thus that should not cause (a) an error code
> > > > > > to be returned, and (b) an error message to be printed, and (c)
> > > > > > because it can simplify the logic in device drivers.
> > > > >
> > > > > I don't agree to (a). If the value signaling not-found is -ENXIO or 0
> > > > > (or -ENODEV) doesn't matter much. I wouldn't deviate from the return
> > > > > code semantics of platform_get_irq() just for having to check against 0
> > > > > instead of -ENXIO. Zero is then just another magic value.
> > > >
> > > > Zero is a natural magic value (also for pointers).
> > > > Errors are always negative.
> > > > Positive values are cookies (or pointers) associated with success.
> > >
> > > Yeah, the issue where we don't agree is if "not-found" is special enough
> > > to deserve the natural magic value. For me -ENXIO is magic enough to
> > > handle the absence of an irq line. I consider it even the better magic
> > > value.
> >
> > It differs from other subsystems (clk, gpio, reset), which do return
> > zero on not found.
>
> IMHO it doesn't matter at all that the return value is zero, relevant is
> the semantic of the returned value. For clk, gpio, reset and regulator
> NULL is a usable dummy, for irqs it's not. So what you do with the value
> returned by platform_get_irq_whatever() is: you compare it with the
> (magic?) not-found value, and if it matches, you enter a suitable
> if-block.
>
> For the (clk|gpiod|regulator)_get_optional() you don't have to check
> against the magic not-found value (so no implementation detail magic
> leaks into the caller code) and just pass it to the next API function.
> (And my expectation would be that if you chose to represent not-found by
> (void *)66 instead of NULL, you won't have to adapt any user, just the
> framework internal checks. This is a good thing!)
Ah, there is the wrong assumption: drivers sometimes do need to know
if the resource was found, and thus do need to know about (void *)66,
-ENODEV, or -ENXIO. I already gave examples for IRQ and clk before.
I can imagine these exist for gpiod and regulator, too, as soon as
you go beyond the trivial "enable" and "disable" use-cases.
And 0/NULL vs. > 0 is the natural check here: missing, but not
an error. Even for IRQ this was envisioned before, when it was
decided that vIRQ zero does not exist.
(Inconsistent) Error codes are not, as missing optional resources
are not error conditions.
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
