Re: [PATCH v13 01/12] splice: Fix O_DIRECT file read splice to avoid reversion of ITER_PIPE

[David Howells]

Matthew Wilcox <willy@xxxxxxxxxxxxx> wrote:

> On Thu, Feb 09, 2023 at 10:29:43AM +0000, David Howells wrote:
> > +	npages = alloc_pages_bulk_list(GFP_USER, npages, &pages);
> 
> Please don't use alloc_pages_bulk_list().  If nobody uses it, it can go
> away again soon.  Does alloc_pages_bulk_array() work for you?  It's
> faster.

Sure.

> > +	/* Free any pages that didn't get touched at all. */
> > +	for (; reclaim >= PAGE_SIZE; reclaim -= PAGE_SIZE)
> > +		__free_page(bv[--npages].bv_page);
> 
> If you have that array, you can then use release_pages() to free
> them, which will be faster.

Um.  I would normally overlay the array on end of the bvec[] so that I could
save on an allocation (I have to fill in the bvec[] anyway) - which means I
wouldn't still have the array at release time.  But in this case I can make an
exception, though I would've thought that the expectation would be that all
the requested data would be fetched.

David