Re: [PATCH] hugetlbfs: zero partial pages during fallocate hole punch

[Mike kravetz <mike.kravetz@oracle.com>]

On Mon, Jun 13, 2022 at 07:30:22PM +0100”, Matthew Wilcox wrote:
> On Mon, Jun 13, 2022 at 11:08:58AM -0700, Mike Kravetz wrote:
> > diff --git a/fs/hugetlbfs/inode.c b/fs/hugetlbfs/inode.c
> > index eca1d0fabd7e..b0d0947a7e9b 100644
> > --- a/fs/hugetlbfs/inode.c
> > +++ b/fs/hugetlbfs/inode.c
> > @@ -584,41 +584,79 @@ static void hugetlb_vmtruncate(struct inode *inode, loff_t offset)
> >  	remove_inode_hugepages(inode, offset, LLONG_MAX);
> >  }
> >  
> > +static void hugetlbfs_zero_partial_page(struct hstate *h,
> > +					struct address_space *mapping,
> > +					unsigned long start,
> 
> This should be loff_t.  Otherwise we can truncate on 32-bit machines.
> 

Thanks!  I missed that.

> > +					unsigned long end)
> > +{
> > +	pgoff_t idx = start >> huge_page_shift(h);
> > +	struct page *page;
> > +
> > +	page = find_lock_page(mapping, idx);
> > +	if (!page)
> > +		return;
> > +
> > +	start = start & ~huge_page_mask(h);
> > +	end = end & ~huge_page_mask(h);
> > +	if (!end)
> > +		end = huge_page_size(h);
> > +
> > +	zero_user_segment(page, (unsigned int)start, (unsigned int)end);
> > +
> > +	unlock_page(page);
> > +	put_page(page);
> 
> We haven't started converting hugetlbfs to folios yet, but here's how
> that would look (which will save us converting it later):
> 
> 	folio = filemap_lock_folio(mapping, idx);
> 	if (!folio)
> 		return;
> ...
> 	folio_zero_segment(folio, start, end);
> 	folio_unlock(folio);
> 	folio_put(folio);
> 
> Pretty much a 1-for-1 replacement.
> 

Great.  The folio conversion is on my todo list, but I would not be
offended if someone beat me to it.
-- 
Mike Kravetz