[PATCH v2 4/6] drm/pagemap: Allocate folios when possible

[Francois Dugast]

On Fri, Jul 25, 2025 at 02:43:48PM -0700, Matthew Brost wrote:
> On Fri, Jul 25, 2025 at 05:39:28PM +0200, Francois Dugast wrote:
> > If the order is greater than zero, allocate a folio when populating the
> > RAM PFNs instead of allocating individual pages one after the other. For
> > example if 2MB folios are used instead of 4KB pages, this reduces the
> > number of calls to the allocation API by 512.
> > 
> > v2:
> > - Use page order instead of extr argument (Matthew Brost)
> > - Allocate with folio_alloc() (Matthew Brost)
> > - Loop for mpages and free_pages based on order (Matthew Brost)
> > 
> > Signed-off-by: Francois Dugast <francois.dugast at intel.com>
> > Cc: Matthew Brost <matthew.brost at intel.com>
> > ---
> >  drivers/gpu/drm/drm_pagemap.c | 38 ++++++++++++++++++++++++++---------
> >  1 file changed, 28 insertions(+), 10 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/drm_pagemap.c b/drivers/gpu/drm/drm_pagemap.c
> > index b6f5d3a8e445..38753ce2bca7 100644
> > --- a/drivers/gpu/drm/drm_pagemap.c
> > +++ b/drivers/gpu/drm/drm_pagemap.c
> > @@ -458,54 +458,72 @@ static int drm_pagemap_migrate_populate_ram_pfn(struct vm_area_struct *vas,
> >  {
> >  	unsigned long i;
> >  
> > -	for (i = 0; i < npages; ++i, addr += PAGE_SIZE) {
> > +	for (i = 0; i < npages;) {
> >  		struct page *page, *src_page;
> > +		unsigned int order;
> > +		struct folio *folio;
> >  
> >  		if (!(src_mpfn[i] & MIGRATE_PFN_MIGRATE))
> > -			continue;
> > +			goto next;
> >  
> >  		src_page = migrate_pfn_to_page(src_mpfn[i]);
> >  		if (!src_page)
> > -			continue;
> > +			goto next;
> >  
> >  		if (fault_page) {
> >  			if (src_page->zone_device_data !=
> >  			    fault_page->zone_device_data)
> > -				continue;
> > +				goto next;
> >  		}
> >  
> > +		order = folio_order(page_folio(src_page));
> > +
> > +		/* TODO: Support fallback to single pages if THP allocation fails */
> >  		if (vas)
> > -			page = alloc_page_vma(GFP_HIGHUSER, vas, addr);
> > +			folio = vma_alloc_folio(GFP_HIGHUSER, order, vas, addr);
> >  		else
> > -			page = alloc_page(GFP_HIGHUSER);
> > +			folio = folio_alloc(GFP_HIGHUSER, order);
> 
> I haven't checked but I assume if order == 0, then alloc_page_vma ==
> vma_alloc_folio? So goes for alloc_page / folio_alloc. If so, then this
> looks good.

Yes according to include/linux/gfp.h:
- alloc_page_vma()->alloc_page_vma_noprof()->vma_alloc_folio_noprof(order=0)
- vma_alloc_folio()->vma_alloc_folio_noprof(order)

> 
> >  
> > -		if (!page)
> > +		if (!folio)
> >  			goto free_pages;
> >  
> > +		page = folio_page(folio, 0);
> >  		mpfn[i] = migrate_pfn(page_to_pfn(page));
> > +
> > +next:
> > +		addr += page_size(page);
> > +		i += NR_PAGES(order);
> >  	}
> >  
> > -	for (i = 0; i < npages; ++i) {
> > +	for (i = 0; i < npages;) {
> >  		struct page *page = migrate_pfn_to_page(mpfn[i]);
> > +		unsigned int order;
> 
> unsigned int order = 0;
> 
> >  
> >  		if (!page)
> >  			continue;
> 
> s/continue/goto next;
> 
> 
> >  
> >  		WARN_ON_ONCE(!trylock_page(page));
> > -		++*mpages;
> > +
> > +		order = folio_order(page_folio(page));
> 
> next:
> 
> > +		*mpages += NR_PAGES(order);
> > +		i += NR_PAGES(order);
> >  	}
> >  
> >  	return 0;
> >  
> >  free_pages:
> > -	for (i = 0; i < npages; ++i) {
> > +	for (i = 0; i < npages;) {
> 
> Same logic changes to this loop as the one above.

Will fix.

Francois

> 
> Matt
> 
> >  		struct page *page = migrate_pfn_to_page(mpfn[i]);
> > +		unsigned int order;
> >  
> >  		if (!page)
> >  			continue;
> >  
> >  		put_page(page);
> >  		mpfn[i] = 0;
> > +
> > +		order = folio_order(page_folio(page));
> > +		i += NR_PAGES(order);
> >  	}
> >  	return -ENOMEM;
> >  }
> > -- 
> > 2.43.0
> >