Re: [PATCH 2/3] afs: Fix afs_write_end() to handle short writes

From: Matthew Wilcox <willy@infradead.org>
To: David Howells <dhowells@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>,
	jlayton@kernel.org, linux-afs@lists.infradead.org,
	ceph-devel@vger.kernel.org, linux-fsdevel@vger.kernel.org,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH 2/3] afs: Fix afs_write_end() to handle short writes
Date: Mon, 14 Jun 2021 14:46:03 +0100	[thread overview]
Message-ID: <YMddm2P0vD+4edBu@casper.infradead.org> (raw)
In-Reply-To: <466590.1623677832@warthog.procyon.org.uk>

On Mon, Jun 14, 2021 at 02:37:12PM +0100, David Howells wrote:
> Matthew Wilcox <willy@infradead.org> wrote:
> 
> > >  (1) If the page is not up to date, then we should just return 0
> > >      (ie. indicating a zero-length copy).  The loop in
> > >      generic_perform_write() will go around again, possibly breaking up the
> > >      iterator into discrete chunks.
> > 
> > Does this actually work?  What about the situation where you're reading
> > the last page of a file and thus (almost) always reading fewer bytes
> > than a PAGE_SIZE?
> 
> Al Viro made such a change for Ceph - and we're writing, not reading.

I'd feel better if you said "xfstests doesn't show any new problems"
than arguing to authority.

I know the operation which triggers this path is a call to write(),
but if, say, the file is 32 bytes long, not in cache, and you write
bytes 32-63, the client must READ bytes 0-31 from the server, which
is less than a full page.