Re: [PATCH v9 1/7] treewide: provide a generic clear_user_page() variant

[Ankur Arora <ankur.a.arora@oracle.com>]

David Hildenbrand (Red Hat) <david@kernel.org> writes:

> On 11/23/25 12:53, Christophe Leroy (CS GROUP) wrote:
>>
>> Le 21/11/2025 à 21:23, Ankur Arora a écrit :
>>> From: David Hildenbrand <david@redhat.com>
>>>
>>> Let's drop all variants that effectively map to clear_page() and
>>> provide it in a generic variant instead.
>>>
>>> We'll use the macro clear_user_page to indicate whether an architecture
>>> provides it's own variant.
>>>
>>> We have to be a bit careful if an architecture provides a custom
>>> clear_user_highpage(), because then it's very likely that some special
>>> flushing magic is happening behind the scenes.
>>>
>>> Maybe at some point these should be CONFIG_ options.
>>>
>>> Note that for parisc, clear_page() and clear_user_page() map to
>>> clear_page_asm(), so we can just get rid of the custom clear_user_page()
>>> implementation. There is a clear_user_page_asm() function on parisc,
>>> that seems to be unused. Not sure what's up with that.
>>>
>>> Signed-off-by: David Hildenbrand <david@redhat.com>
>>> Signed-off-by: Ankur Arora <ankur.a.arora@oracle.com>
>> ...
>>
>>> diff --git a/include/linux/mm.h b/include/linux/mm.h
>>> index 7c79b3369b82..6fa6c188f99a 100644
>>> --- a/include/linux/mm.h
>>> +++ b/include/linux/mm.h
>>> @@ -3879,6 +3879,28 @@ static inline void clear_page_guard(struct zone *zone, struct page *page,
>>>    				unsigned int order) {}
>>>    #endif	/* CONFIG_DEBUG_PAGEALLOC */
>>>    +#ifndef clear_user_page
>>> +/**
>>> + * clear_user_page() - clear a page to be mapped to user space
>>> + * @addr: the address of the page
>>> + * @vaddr: the address of the user mapping
>>> + * @page: the page
>>> + */
>>> +static inline void clear_user_page(void *addr, unsigned long vaddr, struct page *page)
>>> +{
>>> +#ifdef clear_user_highpage
>>> +	/*
>>> +	 * If an architecture defines its own clear_user_highpage() variant,
>>> +	 * then we have to be a bit more careful here and cannot simply
>>> +	 * rely on clear_page().
>>> +	 */
>>> +	clear_user_highpage(page, vaddr);
>>> +#else
>>> +	clear_page(addr);
>>> +#endif
>>> +}
>>> +#endif
>>> +
>>>    #ifdef __HAVE_ARCH_GATE_AREA
>>>    extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
>>>    extern int in_gate_area_no_mm(unsigned long addr);
>>
>> Isn't it chicken and egg with clear_user_highpage() in linux/highmem.h ? :
>
> No really, because we make use of clear_user_highpage() only when the arch
> defines it, so the highmem.h variant is ignored?
>
> Not that I particularly enjoy this way of handling it, so something cleaner
> would be nice :)
>
> (in particular, relying on highmem.h defines in mm.h is a bit suboptimal)
>
>> #ifndef clear_user_highpage
>> static inline void clear_user_highpage(struct page *page, unsigned long
>> vaddr)
>> {
>> 	void *addr = kmap_local_page(page);
>> 	clear_user_page(addr, vaddr, page);
>> 	kunmap_local(addr);
>> }
>> #endif
>> And at the end this function is the only caller of clear_user_page() so
>> there is apparently no need for a generic clear_user_page(), at least
>> not when clear_user_highpage() is defined.
>> I think is would be simpler and cleaner to instead add the following in
>> linux/highmem.c:
>
> I assume you mean highmem.h
>
> It's not really highmem.h material, but if it makes things cleaner, sure.
>
> Might be that the compiler will not be happy about that.
>
> @Ankur can you play with that and see if we can make compilers happy one way or
> the other?

This looks like a good change (and from my tests on a couple of
configs seems to build fine.)

Though clear_user_pages() is also only called from clear_user_highpages().
Would be good to treat that similarly.

(I don't think it can be done because how the arch code treats
both quite differently.)

Anyway I'll play with that a bit more.

--
ankur