Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
From:   ebiederm@xmission.com (Eric W. Biederman)
To:     David Howells <dhowells@redhat.com>
Cc:     Karel Zak <kzak@redhat.com>, linux-api@vger.kernel.org,
        linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org,
        Al Viro <viro@ZenIV.linux.org.uk>,
        Miklos Szeredi <miklos@szeredi.hu>,
        Linus Torvalds <torvalds@linux-foundation.org>,
        util-linux@vger.kernel.org, Andy Lutomirski <luto@amacapital.net>
References: <20190130120654.q5zqcexquca7u337@ws.net.home>
        <87va2716mh.fsf@xmission.com> <9871.1548853314@warthog.procyon.org.uk>
Date:   Wed, 30 Jan 2019 07:35:39 -0600
In-Reply-To: <9871.1548853314@warthog.procyon.org.uk> (David Howells's message
        of "Wed, 30 Jan 2019 13:01:54 +0000")
Message-ID: <87h8dqs1xw.fsf@xmission.com>
User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/25.1 (gnu/linux)
MIME-Version: 1.0
Content-Type: text/plain
Subject: Re: [RFD] A mount api that notices previous mounts
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

David Howells <dhowells@redhat.com> writes:

> Karel Zak <kzak@redhat.com> wrote:
>
>> It seems more elegant is to ask for Nth option as expected by fsinfo().
>
> More elegant yes, but there's an issue with atomiticity[*].  I'm in the
> process of switching to something that returns you a single buffer with all
> the options in, but each key and each value is preceded by a length count.
>
> The reasons for not using separator characters are:
>
>  (1) There's no separator char that cannot validly occur within an option[**].

*Blink*  I had missed the cifs issue.  So yes we certainly need
a better way to encode things in the buffer.  I just used a single
string as an easy way to place everything in a buffer.

>  (2) Makes it possible to return binary values if we need to.

I don't totally disagree with this.  But I will point out that
except for coda passing a file descriptor there are no filesystems
that currently take or need binary options.

I suspect that as long as userspace supports /etc/fstab and we in turn
support /proc/mounts there is going to be a lot of pressure to keep
the majority of options so they  can be encoded in a string separated by
commas.

> David
>
> [*] Atomic with respect to remount calls, that is.

There are also mount options that depend on each other and whose order
matters with respect to other mount options extN's ("sb=<NNNN>") for
example.

> [**] Oh, and look at cifs where you can *change* the separator char during
>      option parsing ("sep=<char>").