linux-kernel - Re: [PATCH] ovl: Allow mount options to be parsed on remount

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <c555a382-fd74-4d9b-ab3e-995049d2947f@igalia.com>
Date: Thu, 22 May 2025 12:22:44 -0300
From: André Almeida <andrealmeid@...lia.com>
To: Amir Goldstein <amir73il@...il.com>, Karel Zak <kzak@...hat.com>
Cc: Christian Brauner <brauner@...nel.org>, Miklos Szeredi
 <miklos@...redi.hu>, linux-unionfs@...r.kernel.org,
 linux-kernel@...r.kernel.org, kernel-dev@...lia.com,
 linux-fsdevel@...r.kernel.org
Subject: Re: [PATCH] ovl: Allow mount options to be parsed on remount

Em 22/05/2025 12:13, Amir Goldstein escreveu:
> cc libfuse maintainer
> 
> On Thu, May 22, 2025 at 4:30 PM André Almeida <andrealmeid@...lia.com> wrote:
>>
>> Em 22/05/2025 06:52, Amir Goldstein escreveu:
>>> On Thu, May 22, 2025 at 8:20 AM André Almeida <andrealmeid@...lia.com> wrote:
>>>>
>>>> Hi Christian, Amir,
>>>>
>>>> Thanks for the feedback :)
>>>>
>>>> Em 21/05/2025 08:20, Christian Brauner escreveu:
>>>>> On Wed, May 21, 2025 at 12:35:57PM +0200, Amir Goldstein wrote:
>>>>>> On Wed, May 21, 2025 at 8:45 AM André Almeida <andrealmeid@...lia.com> wrote:
>>>>>>>
>>>>
>>>> [...]
>>>>
>>>>>>
>>>>>> I see the test generic/623 failure - this test needs to be fixed for overlay
>>>>>> or not run on overlayfs.
>>>>>>
>>>>>> I do not see those other 5 failures although before running the test I did:
>>>>>> export LIBMOUNT_FORCE_MOUNT2=always
>>>>>>
>>>>>> Not sure what I am doing differently.
>>>>>>
>>>>
>>>> I have created a smaller reproducer for this, have a look:
>>>>
>>>>     mkdir -p ovl/lower ovl/upper ovl/merge ovl/work ovl/mnt
>>>>     sudo mount -t overlay overlay -o lowerdir=ovl/lower,upperdir=ovl/
>>>> upper,workdir=ovl/work ovl/mnt
>>>>     sudo mount ovl/mnt -o remount,ro
>>>>
>>>
>>> Why would you use this command?
>>> Why would you want to re-specify the lower/upperdir when remounting ro?
>>> And more specifically, fstests does not use this command in the tests
>>> that you mention that they fail, so what am I missing?
>>>
>>
>> I've added "set -x" to tests/generic/294 to see exactly which mount
>> parameters were being used and I got this from the output:
>>
>> + _try_scratch_mount -o remount,ro
>> + local mount_ret
>> + '[' overlay == overlay ']'
>> + _overlay_scratch_mount -o remount,ro
>> + echo '-o remount,ro'
>> + grep -q remount
>> + /usr/bin/mount /tmp/dir2/ovl-mnt -o remount,ro
>> mount: /tmp/dir2/ovl-mnt: fsconfig() failed: ...
>>
>> So, from what I can see, fstests is using this command. Not sure if I
>> did something wrong when setting up fstests.
>>
> 
> No you are right, I misread your reproducer.
> The problem is that my test machine has older libmount 2.38.1
> without the new mount API.
> 
> 
>>>> And this returns:
>>>>
>>>>     mount: /tmp/ovl/mnt: fsconfig() failed: overlay: No changes allowed in
>>>>     reconfigure.
>>>>           dmesg(1) may have more information after failed mount system call.
>>>>
>>>> However, when I use mount like this:
>>>>
>>>>     sudo mount -t overlay overlay -o remount,ro ovl/mnt
>>>>
>>>> mount succeeds. Having a look at strace, I found out that the first
>>>> mount command tries to set lowerdir to "ovl/lower" again, which will to
>>>> return -EINVAL from ovl_parse_param():
>>>>
>>>>       fspick(3, "", FSPICK_NO_AUTOMOUNT|FSPICK_EMPTY_PATH) = 4
>>>>       fsconfig(4, FSCONFIG_SET_STRING, "lowerdir", "/tmp/ovl/lower", 0) =
>>>> -1 EINVAL (Invalid argument)
>>>>
>>>> Now, the second mount command sets just the "ro" flag, which will return
>>>> after vfs_parse_sb_flag(), before getting to ovl_parse_param():
>>>>
>>>>       fspick(3, "", FSPICK_NO_AUTOMOUNT|FSPICK_EMPTY_PATH) = 4
>>>>       fsconfig(4, FSCONFIG_SET_FLAG, "ro", NULL, 0) = 0
>>>>
>>>> After applying my patch and running the first mount command again, we
>>>> can set that this flag is set only after setting all the strings:
>>>>
>>>>       fsconfig(4, FSCONFIG_SET_STRING, "lowerdir", "/tmp/ovl/lower", 0) = 0
>>>>       fsconfig(4, FSCONFIG_SET_STRING, "upperdir", "/tmp/ovl/upper", 0) = 0
>>>>       fsconfig(4, FSCONFIG_SET_STRING, "workdir", "/tmp/ovl/work", 0) = 0
>>>>       fsconfig(4, FSCONFIG_SET_STRING, "uuid", "on", 0) = 0
>>>>       fsconfig(4, FSCONFIG_SET_FLAG, "ro", NULL, 0) = 0
>>>>
>>>> I understood that the patch that I proposed is wrong, and now I wonder
>>>> if the kernel needs to be fixed at all, or if the bug is how mount is
>>>> using fsconfig() in the first mount command?
>>>
> 
> If you ask me, when a user does:
> /usr/bin/mount /tmp/dir2/ovl-mnt -o remount,ro
> 
> The library only needs to do the FSCONFIG_SET_FLAG command and has no
> business re-sending the other config commands, but that's just me.
> 

Yes, this makes sense to me as well.

> BTW, which version of libmount (mount --version) are you using?
> I think there were a few problematic versions when the new mount api
> was first introduced.
> 

mount from util-linux 2.41 (libmount 2.41.0: btrfs, verity, namespaces, 
idmapping, fd-based-mount, statmount, assert, debug)

> Thanks,
> Amir.