Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Tue, 27 Jun 2023 16:31:17 +0100
From:   Matthew Wilcox <willy@infradead.org>
To:     Christoph Hellwig <hch@infradead.org>
Cc:     linux-mm@kvack.org, linux-fsdevel@vger.kernel.org,
        linux-kernel@vger.kernel.org, Jan Kara <jack@suse.com>,
        David Howells <dhowells@redhat.com>
Subject: Re: [PATCH 09/12] writeback: Factor writeback_iter_next() out of
 write_cache_pages()
Message-ID: <ZJsAxVRZSEvdmlfB@casper.infradead.org>
References: <20230626173521.459345-1-willy@infradead.org>
 <20230626173521.459345-10-willy@infradead.org>
 <ZJpoCy7oWtqy2FoW@infradead.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <ZJpoCy7oWtqy2FoW@infradead.org>
Precedence: bulk

On Mon, Jun 26, 2023 at 09:39:39PM -0700, Christoph Hellwig wrote:
> On Mon, Jun 26, 2023 at 06:35:18PM +0100, Matthew Wilcox (Oracle) wrote:
> > Pull the post-processing of the writepage_t callback into a
> > separate function.  That means changing writeback_finish() to
> > return NULL, and writeback_get_next() to call writeback_finish()
> > when we naturally run out of folios.
> > 
> > Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
> > ---
> >  mm/page-writeback.c | 84 ++++++++++++++++++++++++---------------------
> >  1 file changed, 44 insertions(+), 40 deletions(-)
> > 
> > diff --git a/mm/page-writeback.c b/mm/page-writeback.c
> > index 659df2b5c7c0..ef61d7006c5e 100644
> > --- a/mm/page-writeback.c
> > +++ b/mm/page-writeback.c
> > @@ -2360,7 +2360,7 @@ void tag_pages_for_writeback(struct address_space *mapping,
> >  }
> >  EXPORT_SYMBOL(tag_pages_for_writeback);
> >  
> > -static int writeback_finish(struct address_space *mapping,
> > +static struct folio *writeback_finish(struct address_space *mapping,
> >  		struct writeback_control *wbc, bool done)
> >  {
> >  	folio_batch_release(&wbc->fbatch);
> > @@ -2375,7 +2375,7 @@ static int writeback_finish(struct address_space *mapping,
> >  	if (wbc->range_cyclic || (wbc->range_whole && wbc->nr_to_write > 0))
> >  		mapping->writeback_index = wbc->done_index;
> >  
> > -	return wbc->err;
> > +	return NULL;
> 
> Having a return value that is always NULL feels a bit weird vs just
> doing that return in the caller.

It makes the callers neater.  Compare:

               if (!folio)
                        return writeback_finish(mapping, wbc, false);
vs
		if (!folio) {
			writeback_finish(mapping, wbc, false);
			return NULL;
		}

Similarly for the other two callers.

> > +		if (error == AOP_WRITEPAGE_ACTIVATE) {
> > +			folio_unlock(folio);
> > +			error = 0;
> 
> Note there really shouldn't be any need for this in outside of the
> legacy >writepage case.  But it might make sense to delay the removal
> until after ->writepage is gone to avoid bugs in conversions.

ext4_journalled_writepage_callback() still returns
AOP_WRITEPAGE_ACTIVATE, and that's used by a direct call to
write_cache_pages().

> > +		} else if (wbc->sync_mode != WB_SYNC_ALL) {
> > +			wbc->err = error;
> > +			wbc->done_index = folio->index +
> > +					folio_nr_pages(folio);
> 
> Btw, I wonder if a folio_next_index helper for this might be useful
> as it's a pattern we have in a few places.

I think that's a reasonable addition to the API.