Received-SPF: pass (google.com: domain of linux-nfs-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Fri, 22 Jan 2021 08:20:59 +0000
From:   Christoph Hellwig <hch@infradead.org>
To:     "J. Bruce Fields" <bfields@redhat.com>
Cc:     Christoph Hellwig <hch@infradead.org>, linux-nfs@vger.kernel.org,
        linux-fsdevel@vger.kernel.org,
        Trond Myklebust <trondmy@hammerspace.com>,
        Anna Schumaker <schumakeranna@gmail.com>,
        Chuck Lever <chuck.lever@oracle.com>
Subject: Re: [PATCH 2/3] nfsd: move change attribute generation to filesystem
Message-ID: <20210122082059.GA119852@infradead.org>
References: <1611084297-27352-1-git-send-email-bfields@redhat.com>
 <1611084297-27352-3-git-send-email-bfields@redhat.com>
 <20210120084638.GA3678536@infradead.org>
 <20210121202756.GA13298@pick.fieldses.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20210121202756.GA13298@pick.fieldses.org>
Precedence: bulk

On Thu, Jan 21, 2021 at 03:27:56PM -0500, J. Bruce Fields wrote:
> > Another indirect call just to fetch the change attribute (which happens
> > a lot IIRC) does not seem very optimal to me.
> 
> In the next patch we're removing an fh_getattr (vfs_getattr) in the case
> we call the new op, so that's typically a net decrease in indirect
> calls.
> 
> Though maybe we could use a flag here and do without either.
> 
> > And the fact that we need three duplicate implementations also is not
> > very nice.
> 
> Ext4 and xfs are identical, btrfs is a little different since it doesn't
> consult I_VERSION.  (And then there's nfs, which uses the origin
> server's i_version if it can.)

I'd much prefer to just keep consulting the I_VERSION flag and only
add the new op as an override for the NFS export.

> 
> I also have a vague idea that some filesystem-specific improvements
> might be possible.  (E.g., if a filesystem had some kind of boot count
> in the superblock, maybe that would be a better way to prevent the
> change attribute from going backwards on reboot than the thing
> generic_fetch_iversion is currently doing with ctime.  But I have no
> concrete plan there, maybe I'm dreaming.)

Even without the ctime i_version never goes backward, what is the
problem here?