Re: [PATCH] ceph: Replace DEFINE_SIMPLE_ATTRIBUTE with DEFINE_DEBUGFS_ATTRIBUTE

From: Jeff Layton <jlayton@kernel.org>
To: Ilya Dryomov <idryomov@gmail.com>,
	Jiapeng Chong <jiapeng.chong@linux.alibaba.com>
Cc: Ceph Development <ceph-devel@vger.kernel.org>,
	LKML <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH] ceph: Replace DEFINE_SIMPLE_ATTRIBUTE with DEFINE_DEBUGFS_ATTRIBUTE
Date: Tue, 02 Feb 2021 07:45:03 -0500	[thread overview]
Message-ID: <af100a9c49cca8c9da2b9656c017c38dca2a02f9.camel@kernel.org> (raw)
In-Reply-To: <CAOi1vP_ifWsyOjXo2NaDQnAT9Gn22442KuUXp1LjDCWfpH5yuQ@mail.gmail.com>

On Tue, 2021-02-02 at 13:07 +0100, Ilya Dryomov wrote:
> On Mon, Feb 1, 2021 at 8:52 AM Jiapeng Chong
> <jiapeng.chong@linux.alibaba.com> wrote:
> > 
> > Fix the following coccicheck warning:
> > 
> > ./fs/ceph/debugfs.c:347:0-23: WARNING: congestion_kb_fops should be
> > defined with DEFINE_DEBUGFS_ATTRIBUTE.
> > 
> > Reported-by: Abaci Robot <abaci@linux.alibaba.com>
> > Signed-off-by: Jiapeng Chong <jiapeng.chong@linux.alibaba.com>
> > ---
> >  fs/ceph/debugfs.c | 4 ++--
> >  1 file changed, 2 insertions(+), 2 deletions(-)
> > 
> > diff --git a/fs/ceph/debugfs.c b/fs/ceph/debugfs.c
> > index 66989c8..617327e 100644
> > --- a/fs/ceph/debugfs.c
> > +++ b/fs/ceph/debugfs.c
> > @@ -344,8 +344,8 @@ static int congestion_kb_get(void *data, u64 *val)
> >         return 0;
> >  }
> > 
> > -DEFINE_SIMPLE_ATTRIBUTE(congestion_kb_fops, congestion_kb_get,
> > -                       congestion_kb_set, "%llu\n");
> > +DEFINE_DEBUGFS_ATTRIBUTE(congestion_kb_fops, congestion_kb_get,
> > +                         congestion_kb_set, "%llu\n");
> > 
> > 
> >  void ceph_fs_debugfs_cleanup(struct ceph_fs_client *fsc)
> 
> Hi Jiapeng,
> 
> What is the benefit of this conversion?
> 
> From a quick look, with DEFINE_DEBUGFS_ATTRIBUTE writeback_congestion_kb
> file would no longer be seekable.  It may not matter much, but something
> that should have been mentioned.
> 
> Futher, debugfs_create_file() creates a full proxy for fops, protecting
> against removal races.  DEFINE_DEBUGFS_ATTRIBUTE adds its own protection
> but just for ->read() and ->write().  I don't think we need both.
> 

The coccinelle script clarifies some of this. See the commit log here:

    https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=5103068eaca290f890a30aae70085fac44cecaf6

That said, it also mentions that the file should be converted to use
debugfs_create_file_unsafe now as well, and that wasn't done in this
patch. Jiapeng, was that intentional? If so, why?

-- 
Jeff Layton <jlayton@kernel.org>