[systemd-devel] [PATCH] fstab-generator: handle mount units with "x-rootfs.mount"

[Tom Gundersen]

On Sat, Mar 23, 2013 at 2:35 PM, Harald Hoyer <harald.hoyer at gmail.com> wrote:
> Am 23.03.13 12:59, schrieb Kay Sievers:
>> On Sat, Mar 23, 2013 at 5:49 AM, Harald Hoyer <harald at redhat.com> wrote:
>>> Am 23.03.2013 03:05, schrieb Lennart Poettering:
>>>> On Thu, 14.03.13 13:15, harald at redhat.com (harald at redhat.com) wrote:
>>>>
>>>>> From: Harald Hoyer <harald at redhat.com>
>>>>>
>>>>> Mount units with "x-rootfs.mount" are now ordered before root-fs.target.
>>>>> As we sometimes construct /sysroot mounts in /etc/fstab in the initrd,
>>>>> we want these to be mounted before the root-fs.target is active.
>>>>
>>>> Hmm, I don't get this, what is this for? Where's the destinction to
>>>> "x-initrd.mount"? Why do we need both? Examples?
>>>>
>>>> Apparently a patch like this got merged, but it uses
>>>> "x-initrd-rootfs.mount" as identifier, is that the same thing? If so,
>>>> the thing really should be called "x-initrd.foobar", i.e. "x-initrd" is
>>>> supposed to be the 'namespace' the setting is in.
>>>>
>>>> Anyway, totally not grokking this, please enlighten me.
>>>>
>>>
>>> Sometimes we need to create an /etc/fstab entry for /sysroot in the initramfs
>>> manually. So, we could either hardcode "/sysroot" to result in dependencies to
>>> SPECIAL_ROOT_FS_TARGET, or mark it with "x-initrd-rootfs.mount".
>>
>> What Lennart meant was: all the initrd things should stick to one and
>> the same prefix x-initrd.* and not invent new top-level prefixes with
>> x-initrd-*.
>
> well, so, x-initrd.rootfs-mount ??
>
>>
>> This can only be one entry ever, and it will always be /sysroot,
>> right? This flexibility is probably not needed then and we can
>> hardcode /sysroot?
>
> I just wanted to be flexible, because initrd-root-fs.target and
> initrd-fs.target are different targets.
>
> initrd-root-fs.target is defined by fstab in the initrd.
> initrd-fs.target is defined by fstab in the real root.
>
> One use case could be, that you want to mount /sysroot/etc before you
> parse /sysroot/etc/fstab for xinitrd.mount.

This flexibility sort of makes sense to me. Especially if/when people
start having / on tmpfs (and therefore needs to mount /sysroot/etc as
Harald says above).

-t