Re: [PATCH v5 1/6] userfaultfd: convert mfill_atomic_pte_copy() to use a folio

[zhangpeng (AS)]

On 2023/4/7 5:31, Mike Kravetz wrote:

> On 03/31/23 17:39, Peng Zhang wrote:
> > From: ZhangPeng <zhangpeng362@xxxxxxxxxx>
> >
> > Call vma_alloc_folio() directly instead of alloc_page_vma() and convert
> > page_kaddr to kaddr in mfill_atomic_pte_copy(). Removes several calls to
> > compound_head().
> >
> > Signed-off-by: ZhangPeng <zhangpeng362@xxxxxxxxxx>
> > Reviewed-by: Sidhartha Kumar <sidhartha.kumar@xxxxxxxxxx>
> > ---
> >   mm/userfaultfd.c | 33 +++++++++++++++++----------------
> >   1 file changed, 17 insertions(+), 16 deletions(-)
> Looks good,
>
> Reviewed-by: Mike Kravetz <mike.kravetz@xxxxxxxxxx>
>
> >   	} else {
> > -		page = *pagep;
> > +		folio = page_folio(*pagep);
> > +		VM_BUG_ON_FOLIO(folio_test_large(folio), folio);
> >   		*pagep = NULL;
> >   	}
> However, I am still unsure about the reason for adding the VM_BUG_ON_FOLIO
> here.

VM_BUG_ON_FOLIO was added to ensure that folio is a single-page folio.
However, the folio corresponding to the foliop is always a single-page
folio. We just don't need this check. I'll drop VM_BUG_ON_FOLIO.

Thanks for your review.

Best Regards,
Peng