Re: [PATCH 1/4] mm/mprotect: encourage inlining with __always_inline

[Lorenzo Stoakes (Oracle)]

On Thu, Mar 19, 2026 at 06:59:37PM +0000, Lorenzo Stoakes (Oracle) wrote:
> On Thu, Mar 19, 2026 at 06:31:05PM +0000, Pedro Falcato wrote:
> > Encourage the compiler to inline batch PTE logic and resolve constant
> > branches by adding __always_inline strategically.
>
> >
> > Signed-off-by: Pedro Falcato <pfalcato@xxxxxxx>
>
> Does this vary by compiler/arch that much?
>
> I wonder about how much ends up on the stack here too given the HUGE number of
> arguments passed around but I guess you'd be pushing and popping some even if
> these weren't inlined.
>
> I wonder if we wouldn't want to carefully check different arches for this
> though!

BTW have previously seen how compilers can be suuuuper picky as to what inlines
or not based on hueristics so can see why doing this would move the needle if we
were sure it'd universally help/at least not cause issues.

Cheers, Lorenzo