Subject: Re: [PATCH 1/3] autofs - fix AT_NO_AUTOMOUNT not being honored
From: Ian Kent <raven@themaw.net>
To: David Howells <dhowells@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>,
        autofs mailing list <autofs@vger.kernel.org>,
        Ondrej Holy <oholy@redhat.com>, Colin Walters <walters@redhat.com>,
        Kernel Mailing List <linux-kernel@vger.kernel.org>,
        linux-fsdevel <linux-fsdevel@vger.kernel.org>
References: <150216641255.11652.4204561328197919771.stgit@pluto.themaw.net>
 <4689.1502267949@warthog.procyon.org.uk>
 <f7ea302d-2e17-b8bf-242e-fc06c4a0adcb@themaw.net>
Message-ID: <253371b7-ce7d-1b97-bbb4-cf2263d6dd28@themaw.net>
Date: Thu, 10 Aug 2017 10:16:02 +0800
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101
 Thunderbird/52.2.1
MIME-Version: 1.0
In-Reply-To: <f7ea302d-2e17-b8bf-242e-fc06c4a0adcb@themaw.net>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 866
Lines: 24

On 09/08/17 17:51, Ian Kent wrote:
> On 09/08/17 16:39, David Howells wrote:
>> Ian Kent <raven@themaw.net> wrote:
>>
>>> In order to handle the AT_NO_AUTOMOUNT for both system calls the
>>> negative dentry case in follow_automount() needs to be changed to
>>> return ENOENT when the LOOKUP_AUTOMOUNT flag is clear (and the other
>>> required flags are clear).
>>
>> Should the be EREMOTE instead of ENOENT?
> 
> I thought about that and ended up thinking ENOENT was more sensible
> but I'll look at it again.

I think EREMOTE and ENOENT both are inaccurate.

There's no way to know if the negative dentry corresponds to a valid map
key, and we've seen increasing lookups from userspace applications for
invalid directories, so I'm not sure.

I went with ENOENT but I guess we could use EREMOTE, what's your thinking
on why EREMOTE might be better than ENOENT?

Ian