Re: [PATCH 07/17] writeback: Factor writeback_get_batch() out of write_cache_pages()

From: Christoph Hellwig <hch@lst.de>
To: Jan Kara <jack@suse.cz>
Cc: Christoph Hellwig <hch@lst.de>,
	linux-mm@kvack.org,
	"Matthew Wilcox (Oracle)" <willy@infradead.org>,
	Jan Kara <jack@suse.com>, David Howells <dhowells@redhat.com>,
	Brian Foster <bfoster@redhat.com>,
	linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH 07/17] writeback: Factor writeback_get_batch() out of write_cache_pages()
Date: Thu, 21 Dec 2023 13:22:33 +0100	[thread overview]
Message-ID: <20231221122233.GC17956@lst.de> (raw)
In-Reply-To: <20231221111743.sppmjkyah3u4ao6g@quack3>

On Thu, Dec 21, 2023 at 12:17:43PM +0100, Jan Kara wrote:
> > +static void writeback_get_batch(struct address_space *mapping,
> > +		struct writeback_control *wbc)
> > +{
> > +	folio_batch_release(&wbc->fbatch);
> > +	cond_resched();
> 
> I'd prefer to have cond_resched() explicitely in the writeback loop instead
> of hidden here in writeback_get_batch() where it logically does not make
> too much sense to me...

Based on the final state after this series, where would you place it?

(That beeing said there is a discussion underway on lkml to maybe
 kill cond_resched entirely as part of sorting out the preemption
 model mess, at that point this would become a moot point anyway)

> >  	} else {
> > -		index = wbc->range_start >> PAGE_SHIFT;
> > +		wbc->index = wbc->range_start >> PAGE_SHIFT;
> >  		end = wbc->range_end >> PAGE_SHIFT;
> >  	}
> 
> Maybe we should have:
> 	end = wbc_end(wbc);
> 
> when we have the helper? But I guess this gets cleaned up in later patches
> anyway so whatever.

Yeah, this end just goes away.  I can convert it here, but that feels
like pointless churn to me.