Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Feedback-ID: i31e841b0:Fastmail
Message-ID: <3364aae7-9247-21aa-9ea4-36348462df4c@themaw.net>
Date:   Wed, 10 Aug 2022 21:01:26 +0800
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101
 Thunderbird/91.11.0
Subject: Re: [RFC][PATCH] uapi: Remove the inclusion of linux/mount.h from
 uapi/linux/fs.h
Content-Language: en-US
To:     Florian Weimer <fweimer@redhat.com>,
        David Howells <dhowells@redhat.com>
Cc:     linux-fsdevel@vger.kernel.org,
        Alexander Viro <viro@zeniv.linux.org.uk>,
        Christian Brauner <christian@brauner.io>,
        linux-api@vger.kernel.org, linux-kernel@vger.kernel.org
References: <163410.1659964655@warthog.procyon.org.uk>
 <87zggce9fd.fsf@oldenburg.str.redhat.com>
From:   Ian Kent <raven@themaw.net>
In-Reply-To: <87zggce9fd.fsf@oldenburg.str.redhat.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 10/8/22 17:26, Florian Weimer wrote:
> * David Howells:
>
>> We're seeing issues in autofs and xfstests whereby linux/mount.h (the UAPI
>> version) as included indirectly by linux/fs.h is conflicting with
>> sys/mount.h (there's a struct and an enum).
>>
>> Would it be possible to just remove the #include from linux/fs.h (as patch
>> below) and rely on those hopefully few things that need mount flags that don't
>> use the glibc header for them working around it by configuration?
> Wasn't <linux/mount.h> split from <linux/fs.h> relatively recently, and
> userspace is probably using <linux/fs.h> to get the mount flag
> definitions?

Not sure myself but this is in the user space kernel includes

and sys/mount.h has pretty much what linux/mount.h has plus a

few function declarations. It's almost a complete duplication.


The reality is that the enum declaration could be changed to

#defines (not a preferred approach) which leaves only the

struct mount_attr which is the difficult one to resolve.


>
> In retrospect, it would have been better to add the new fsmount stuff to
> a separate header file, so that we could include that easily from
> <sys/mount.h> on the glibc side.  Adhemerval posted a glibc patch to
> fake that (for recent compilers):
>
>    [PATCH] linux: Fix sys/mount.h usage with kernel headers
>    <https://sourceware.org/pipermail/libc-alpha/2022-August/141316.html>
>
> I think it should work reliably, so that's probably the direction we are
> going to move in.

Looked a lot more complicated than I thought it could be, enough

that I can't say if it will work so I'll take your word for it.


Ian