Re: [PATCH 1/2] vfs: Convert debugfs to use the new mount API

From: Miklos Szeredi <miklos@szeredi.hu>
To: Christian Brauner <brauner@kernel.org>
Cc: Eric Sandeen <sandeen@sandeen.net>,
	Eric Sandeen <sandeen@redhat.com>,
	 linux-fsdevel@vger.kernel.org,
	Steven Rostedt <rostedt@goodmis.org>,
	 Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	"Rafael J. Wysocki" <rafael@kernel.org>,
	 "Bill O'Donnell" <billodo@redhat.com>,
	David Howells <dhowells@redhat.com>
Subject: Re: [PATCH 1/2] vfs: Convert debugfs to use the new mount API
Date: Fri, 8 Mar 2024 15:54:09 +0100	[thread overview]
Message-ID: <CAJfpegvW1-rn6dSoQ4j4qqQS9fSnNCRZNE6D_xw-iNokpHt+cQ@mail.gmail.com> (raw)
In-Reply-To: <20240307-winkelmesser-funkkontakt-845889326073@brauner>

On Thu, 7 Mar 2024 at 13:04, Christian Brauner <brauner@kernel.org> wrote:

> But I'm not yet clear whether FSOPEN/FSPICK_IGNORE_UNKNOWN wouldn't make
> more sense than FSOPEN/FSPICK_REJECT_UNKNOWN. IOW, invert the logic.

I think there needs to be a mode for fsopen/fspick/fconfig API that
allows implementing full backward compatibility with the old behavior
of mount(8), both in case of new mount and remount.  By old I mean
before any of the API conversions started.  If some filesystems
rejected unknown options and some ignored them, then that is what this
mode should continue to do.  This is what we currently have, so
without additional flags this is what the API should continue to
support.

And I think there needs to be a new "strict" mode for fsopen/fspick
that has clear rules for how filesystems should handle options, which
as you say most filesystem already do.   Since this is a new mode, I
think it needs a new flag, that is rejected if the API or the fs
doesn't support this mode.  Filesystems which already have sane
behavior need not care, they would work the same in both modes.
Filesystems that are currently inconsistent would have to implement
both modes.

Thanks,
Miklos