[PATCH] ARM: Stricten arguments to virt_to_pfn()

[Geert Uytterhoeven]

Hi Linus,

Thanks for your patch!

On Mon, May 16, 2022 at 2:18 PM Linus Walleij <linus.walleij at linaro.org> wrote:
> The kernel convention is that the argument passed to virt_to_pfn()
> (virtual address to page frame number) needs to be an unsigned long,
> but ARM has avoided this by including a cast to (unsigned long)
> inside the definition of virt_to_pfn() itself. This creates problems
> if we want to make virt_to_pfn() an actual function with a typed argument,
> so fix up the offending call sites.

IMHO this is a step in the wrong direction: a virtual address should
use "void *".

See also e.g. virt_to_phys() on both arm32 and arm64:

    static inline phys_addr_t virt_to_phys(const volatile void *x)
    {
            return __virt_to_phys((unsigned long)(x));
    }

BTW, virt_to_pfn() differs, though.  Arm32 has:

    #define virt_to_pfn(kaddr) \
            ((((unsigned long)(kaddr) - PAGE_OFFSET) >> PAGE_SHIFT) + \
             PHYS_PFN_OFFSET)

while arm64 has:

    #define virt_to_pfn(x)
__phys_to_pfn(__virt_to_phys((unsigned long)(x)))

In the latter, it is more ibvious the cast to "unsigned long" is there
because __virt_to_phys() takes "unsigned long".  Perhaps this should
be made clearer by calling virt_to_phys() instead?.

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert at linux-m68k.org

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