Re: [PATCH] arm64: patching: avoid early page_to_phys()

[Mike Rapoport <rppt@kernel.org>]

On Tue, Dec 03, 2024 at 01:46:11AM +0100, Christoph Hellwig wrote:
> On Mon, Dec 02, 2024 at 05:03:59PM +0000, Mark Rutland wrote:
> > +	phys_addr_t phys;
> > +
> > +	if (is_image_text((unsigned long)addr)) {
> > +		phys = __pa_symbol(addr);
> > +	} else {
> > +		struct page *page = vmalloc_to_page(addr);
> > +		BUG_ON(!page);
> > +		phys = page_to_phys(page) + offset_in_page(addr);
> > +	}
> > +
> > +	return (void *)set_fixmap_offset(fixmap, phys);
> 
> This looks much better.  Btw, I recently open coded the vmalloc to phys
> logic above in an unrelated fix, and noticed that
> arch/powerpc/mm/pgtable.c has a nice vmalloc_to_phys that could
> be generalized and move to common code.  Does anyone have an opinion
> on that?

Maybe fold that check for !page into vmalloc_to_pfn() and make
vmalloc_to_phys() a static inline?
 
-- 
Sincerely yours,
Mike.