Received-SPF: pass (google.com: domain of linux-kernel+bounces-9144-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45e3:2400::1 as permitted sender) client-ip=2604:1380:45e3:2400::1;
Precedence: bulk
MIME-Version: 1.0
References: <CAPnZJGDcNwPLbzC99qNQ+bRMwxPU-Z0xe=TD6DWQU=0MNyeftA@mail.gmail.com>
 <d4b227de-d609-aef2-888b-203dbcf06707@landley.net>
In-Reply-To: <d4b227de-d609-aef2-888b-203dbcf06707@landley.net>
From: Askar Safin <safinaskar@gmail.com>
Date: Fri, 22 Dec 2023 01:58:19 +0300
Message-ID: <CAPnZJGBeV-E_AN8GnTfkaJvRtBmCeMYYCt+O0XMsc3kDULRuKg@mail.gmail.com>
Subject: Re: [PATCH v3] rootfs: Fix support for rootfstype= when root= is given
To: Rob Landley <rob@landley.net>, stefanb@linux.ibm.com
Cc: gregkh@linuxfoundation.org, initramfs@vger.kernel.org, 
	linux-kernel@vger.kernel.org, stable@vger.kernel.org, zohar@linux.ibm.com
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

Hi, Rob. And Stefan.

First of all, this patch got to linux-next (
https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/log/?qt=
=3Dauthor&q=3DStefan+Berger
), so it seems it soon will be in mainline.

On Thu, Dec 21, 2023 at 12:24=E2=80=AFPM Rob Landley <rob@landley.net> wrot=
e:
> Can you build tmpfs on a nommu system? Last I checked the plumbing expect=
s swap,
> but it's been a while...
Okay, I agree, let's not remove ramfs.

Still, I don't like this (already applied) patch. init=3D and rdinit=3D
are two different options, and this is good. So, I think we should
have two different options. Analogously they should be rootfstype=3D and
rdrootfstype=3D. rootfstype=3D should be read by kernel when deciding how
to mount real root (i. e. not initramfs or initrd) only and
rdrootfstype=3D when deciding how to mount initramfs only. This will
make everything cleaner. Also note that userspace tools read
rootfstype=3D and assume that it always applies to real root. For
example, this is Debian's rdinit:

https://salsa.debian.org/kernel-team/initramfs-tools/-/blob/cf964bfb4362019=
fd7fba1e839e403ff950dca8e/init#L103

As you can see, this shell script parses /proc/cmdline and assumes
that rootfstype=3D always applies to real root. So, if someone sets
rootfstype=3D to tmpfs or ramfs, this will likely break this script.

So, I think the code should look so:

+if (IS_ENABLED(CONFIG_TMPFS)) {
+        if (!rd_root_fs_names) // We assume rd_root_fs_names is set
by rdrootfstype=3D
+                is_tmpfs =3D true; // Use tmpfs if rdrootfstype=3D is not
set. To get all tmpfs benefits
+        else if (rd_root_fs_names && !!strstr(rd_root_fs_names, "tmpfs"))
+                is_tmpfs =3D true;
+}

Yes, this will slightly break compatibility. I. e. this will make
Linux always choose tmpfs if rdrootfstype=3D is not present. But I think
there is nothing wrong with it. If a user cares, he will set
rdrootfstype=3D . And early boot code will become a lot more clean and
logical.

Rob, do you agree? Stefan, do you agree? Then I will write a patch,
with doc changes (currently I use gmail web interface, of course I
will use git send-email when I sent actual patch)

--=20
Askar Safin