| 1 |
Hi, |
| 2 |
|
| 3 |
On Wed, 21 Jun 2006 13:50:12 -0700 |
| 4 |
"Richard Fish" <bigfish@××××××××××.org> wrote: |
| 5 |
|
| 6 |
> pivot_root is specifically *not* allowed from an initramfs |
| 7 |
> environment. What you want to do is simply mount the new root |
| 8 |
> filesystem, chroot into it, and execute init. Something like: |
| 9 |
> |
| 10 |
> cd /new_root ; exec ./bin/chroot . ./sbin/init "$@" >dev/console |
| 11 |
> <dev/console 2>&1 |
| 12 |
> |
| 13 |
> If you are *extremely* tricky, and use a symlinked /lib directory, you |
| 14 |
> can actually delete everything from the initramfs before doing the |
| 15 |
> chroot/init calls. Let me know if you need some more details on this. |
| 16 |
|
| 17 |
Hm, I'm pretty sure that it is well possible to pivot_root from an |
| 18 |
initramfs. Isn't that the whole point of pivot_root? But you may be |
| 19 |
right that it is not possible for NFS mounts, I never tried that before. |
| 20 |
|
| 21 |
In fact, the approach you took is weak. /sbin/init wouldn't run as PID |
| 22 |
1 in this case which is bad for the situation that the calling script |
| 23 |
(which _has_ PID 1) dies. The kernel would recognize this and reboot |
| 24 |
(and/or panic, not sure). To avoid this, one has to exec the init from |
| 25 |
the script. |
| 26 |
|
| 27 |
So basically it boils down to this /init in the initramfs: |
| 28 |
|
| 29 |
#!/bin/sh |
| 30 |
PATH=/bin:/sbin |
| 31 |
modprobe supermightyrootfsprovidingmodule && \ |
| 32 |
mount -t blahfs /dev/whereitis /mnt/stagetwo && \ |
| 33 |
cd /mnt/stagetwo && \ |
| 34 |
pivot_root . /mnt/initramfs || reboot -f |
| 35 |
exec /sbin/init |
| 36 |
|
| 37 |
Note that I assume /mnt/stagetwo exists in initramfs (as well as |
| 38 |
modprobe, mount, pivot_root and reboot) and /mnt/initramfs exists in |
| 39 |
the to-be-mounted fs. Note that the last point may prevent |
| 40 |
pivot_root'ing in a scenario where an NFS root fs is desired (because |
| 41 |
I'm not sure if it can have mounts in it, but that would be needed for |
| 42 |
proc, sys and dev, too, so it _may_ work here, too). The call |
| 43 |
to /sbin/init happens in the new fs because pivot_root manipulates the |
| 44 |
namespace of the calling process. Exec'ing is important so |
| 45 |
that /sbin/init gets PID 1 and we don't have a process which depends on |
| 46 |
the initramfs anymore, so we can unmount it at a later point. |
| 47 |
|
| 48 |
-hwh |
| 49 |
-- |
| 50 |
gentoo-user@g.o mailing list |
Archives