On Tue, Sep 21, 2021 at 09:34:55AM +0100, Christoph Hellwig wrote:
> On Fri, Sep 17, 2021 at 06:30:50PM -0700, Darrick J. Wong wrote:
> > + case IOMAP_MAPPED:
> > + while (nr_pages > 0) {
> > + /* XXX function only supports one page at a time?! */
> > + ret = dax_zero_page_range(iomap->dax_dev, start_page,
> > + 1);
>
> Yes. Given that it will have to kmap every page that kinda makes sense.
> Unlike a nr_pages argument which needs to be 1, which is just silly.
> That being said it would make sense to move the trivial loop over the
> pages into the methods to reduce the indirect function call overhead
Done. AFAICT all the implementations *except* nvdimm can handle more
than one page.
--D
