[gentoo-user] initramfs, network diskless boot, init process, problems with switchroot (pivot_root)

[gentoo-user] initramfs, network diskless boot, init process, problems with switchroot (pivot_root)

[Claudinei Matos]

[-- Attachment #1: Type: text/plain, Size: 1345 bytes --]

Hi guys,

I'm not sure if it's the best place to do this question but I hope somebody
could help me...
well, I'm trying to make such a network diskless boot script for a amd64
machine, the target of this project is to boot an amd64 machine via network
and mount nfs share as the root filesystem.
I'd created an initramfs wich mount the nfs share and do the pivot_root
(actually switch_root from busybox) but the problem is exactly at this
moment, 'cause when I try to do the switch_root and start the real init from
the nfs share, the system appears to freeze but after some seconds it do
print a message "Rebooting System" and just reboot the machine.
I'd tried to reinstall sysvinit (even a static version) but the problem
persists, I als tried to change the sysinit script on inittab but again the
same problem and it feels like the script is not executed, so for a last try
I did changed the init binary with a script that just call /bin/bash and for
my surprise it did work, but none system process run course.

What I want to know is if is there a way to change this "/sbin/rc sysinit"
from inittab to a custom script wich will do some configurations like create
some directories, mount some systems and later start the normal process from
"default" rc?

tks in any advance,


-- 
Claudinei Matos
<claudineimatos@gmail.com>
55-21-81980605

[-- Attachment #2: Type: text/html, Size: 1488 bytes --]

Re: [gentoo-user] initramfs, network diskless boot, init process, problems with switchroot (pivot_root)

[Richard Fish]

On 6/16/06, Claudinei Matos <claudineimatos@gmail.com> wrote:
> I'd created an initramfs wich mount the nfs share and do the pivot_root
> (actually switch_root from busybox) but the problem is exactly at this
> moment, 'cause when I try to do the switch_root and start the real init from
> the nfs share, the system appears to freeze but after some seconds it do
> print a message "Rebooting System" and just reboot the machine.

Sorry for the late reply...I've been on offline (vacation) for several days.

pivot_root is specifically *not* allowed from an initramfs
environment.  What you want to do is simply mount the new root
filesystem, chroot into it, and execute init.  Something like:

cd /new_root ; exec ./bin/chroot . ./sbin/init "$@" >dev/console
<dev/console 2>&1

If you are *extremely* tricky, and use a symlinked /lib directory, you
can actually delete everything from the initramfs before doing the
chroot/init calls.  Let me know if you need some more details on this.

HTH,
-Richard
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] initramfs, network diskless boot, init process, problems with switchroot (pivot_root)

[Hans-Werner Hilse]

Hi,

On Wed, 21 Jun 2006 13:50:12 -0700
"Richard Fish" <bigfish@asmallpond.org> wrote:

> pivot_root is specifically *not* allowed from an initramfs
> environment.  What you want to do is simply mount the new root
> filesystem, chroot into it, and execute init.  Something like:
> 
> cd /new_root ; exec ./bin/chroot . ./sbin/init "$@" >dev/console
> <dev/console 2>&1
> 
> If you are *extremely* tricky, and use a symlinked /lib directory, you
> can actually delete everything from the initramfs before doing the
> chroot/init calls.  Let me know if you need some more details on this.

Hm, I'm pretty sure that it is well possible to pivot_root from an
initramfs. Isn't that the whole point of pivot_root? But you may be
right that it is not possible for NFS mounts, I never tried that before.

In fact, the approach you took is weak. /sbin/init wouldn't run as PID
1 in this case which is bad for the situation that the calling script
(which _has_ PID 1) dies. The kernel would recognize this and reboot
(and/or panic, not sure). To avoid this, one has to exec the init from
the script.

So basically it boils down to this /init in the initramfs:

#!/bin/sh
PATH=/bin:/sbin
modprobe supermightyrootfsprovidingmodule && \
  mount -t blahfs /dev/whereitis /mnt/stagetwo && \
  cd /mnt/stagetwo && \
  pivot_root . /mnt/initramfs || reboot -f
exec /sbin/init

Note that I assume /mnt/stagetwo exists in initramfs (as well as
modprobe, mount, pivot_root and reboot) and /mnt/initramfs exists in
the to-be-mounted fs. Note that the last point may prevent
pivot_root'ing in a scenario where an NFS root fs is desired (because
I'm not sure if it can have mounts in it, but that would be needed for
proc, sys and dev, too, so it _may_ work here, too). The call
to /sbin/init happens in the new fs because pivot_root manipulates the
namespace of the calling process. Exec'ing is important so
that /sbin/init gets PID 1 and we don't have a process which depends on
the initramfs anymore, so we can unmount it at a later point.

-hwh
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] initramfs, network diskless boot, init process, problems with switchroot (pivot_root)

[Richard Fish]

On 6/21/06, Hans-Werner Hilse <hilse@web.de> wrote:
> On Wed, 21 Jun 2006 13:50:12 -0700
>
> Hm, I'm pretty sure that it is well possible to pivot_root from an
> initramfs. Isn't that the whole point of pivot_root? But you may be
> right that it is not possible for NFS mounts, I never tried that before.

http://bugzilla.kernel.org/show_bug.cgi?id=4857

My final comments on that bug are based on the attached email that I
received from Andrew Morton that made it clear that Al Viro was
opposed to pivot_root being used from an initramfs.  (BTW, viro's
comments are not very polite, but you have to expect such directness
from kernel hackers!)

Note, be sure we are talking about an initramfs here, not an initrd.
>From an initrd it is still possible and supported to use pivot_root.

> In fact, the approach you took is weak. /sbin/init wouldn't run as PID
> 1 in this case which is bad for the situation that the calling script
> (which _has_ PID 1) dies. The kernel would recognize this and reboot
> (and/or panic, not sure). To avoid this, one has to exec the init from
> the script.

No, this works perfectly.  I use it every time I boot my kernel, and
my init *is* PID 1.

The "exec ./bin/chroot" part replaces the shell executing the /init
script with the chroot command, so chroot then becomes PID 1.  Chroot
then does an exec of ./sbin/init so that init becomes PID 1.

>
> So basically it boils down to this /init in the initramfs:
>
> #!/bin/sh
> PATH=/bin:/sbin
> modprobe supermightyrootfsprovidingmodule && \
>   mount -t blahfs /dev/whereitis /mnt/stagetwo && \
>   cd /mnt/stagetwo && \
>   pivot_root . /mnt/initramfs || reboot -f
> exec /sbin/init

Be warned that if you do this, and then you try to "mount --move"
anything, your kernel will probably hang.

But again, in recent kernels, pivot_root *should* be returning -EINVAL
if you do this.  Indeed I just checked my 2.6.16 sources, and the code
that returns EINVAL in sys_pivot_root for unattached mounts is still
there:

        error = -EINVAL;
        if (user_nd.mnt->mnt_root != user_nd.dentry)
                goto out2; /* not a mountpoint */
        if (user_nd.mnt->mnt_parent == user_nd.mnt)
                goto out2; /* not attached */
        if (new_nd.mnt->mnt_root != new_nd.dentry)
                goto out2; /* not a mountpoint */
        if (new_nd.mnt->mnt_parent == new_nd.mnt)
                goto out2; /* not attached */

So unless your the pivot_root program is doing something other than
sys_pivot_root(), I don't see how this can possibly work on a recent
kernel.

-Richard
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] initramfs, network diskless boot, init process, problems with switchroot (pivot_root)

[Richard Fish]

[-- Attachment #1: Type: text/plain, Size: 233 bytes --]

On 6/21/06, Richard Fish <bigfish@asmallpond.org> wrote:
> http://bugzilla.kernel.org/show_bug.cgi?id=4857
>
> My final comments on that bug are based on the attached email

-1,000,000 points for not including the email...

-Richard

[-- Attachment #2: pivot_root_mail.txt --]
[-- Type: text/plain, Size: 2792 bytes --]

From - Thu Aug 18 13:01:31 2005
X-Account-Key: account1
X-UIDL: f9050c2168cec49d57cb44456faba2d8
X-Mozilla-Status: 0001
X-Mozilla-Status2: 00000000
X-Apparently-To: bigfish@asmallpond.org via 66.218.93.224; Thu, 18 Aug 2005 10:24:05 -0700
X-Originating-IP: [65.172.181.4]
Return-Path: <akpm@osdl.org>
Authentication-Results: mta107.pa.mail.re2.yahoo.com
  from=osdl.org; domainkeys=neutral (no sig)
Received: from 65.172.181.4  (EHLO smtp.osdl.org) (65.172.181.4)
  by mta107.pa.mail.re2.yahoo.com with SMTP; Thu, 18 Aug 2005 10:24:05 -0700
Received: from shell0.pdx.osdl.net (fw.osdl.org [65.172.181.6])
	by smtp.osdl.org (8.12.8/8.12.8) with ESMTP id j7IHO3jA023758
	(version=TLSv1/SSLv3 cipher=EDH-RSA-DES-CBC3-SHA bits=168 verify=NO);
	Thu, 18 Aug 2005 10:24:03 -0700
Received: from bix (shell0.pdx.osdl.net [10.9.0.31])
	by shell0.pdx.osdl.net (8.13.1/8.11.6) with SMTP id j7IHO2iP028972;
	Thu, 18 Aug 2005 10:24:02 -0700
Date: Thu, 18 Aug 2005 10:22:42 -0700
From: Andrew Morton <akpm@osdl.org>
To: Miklos Szeredi <miklos@szeredi.hu>, <bigfish@asmallpond.org>
Subject: pivot_root-circular-reference-fix-2.patch
Message-Id: <20050818102242.24df31e1.akpm@osdl.org>
X-Mailer: Sylpheed version 1.0.4 (GTK+ 1.2.10; i386-redhat-linux-gnu)
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
X-Spam-Status: No, hits=0 required=5 tests=
X-Spam-Checker-Version: SpamAssassin 2.63-osdl_revision__1.45__
X-MIMEDefang-Filter: osdl$Revision: 1.114 $
X-Scanned-By: MIMEDefang 2.36

<cw> viro: http://www.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.13-rc5/2.6.13-rc5-mm1/broken-out/pivot_root-circular-reference-fix-2.patch
<viro> WTF?
<cw> -EWTFWTF?
<viro> akpm: patch is an utter bullshit
<viro> akpm: we are _NOT_ allowing to change namespace root
<viro> not with pivot_root(), not with anything else
<viro> moreover, I distinctly remember having such check in sys_pivot_root()
<viro> where the hell had it gone?
<cw> viro: what about depricate pivot_root since it's the source of so much angst? (i used it myself and im not sure how to avoid it here but there must be a way)
* viro goes to look through history
<viro> LARTs are needed, apparently
<olaf> or -einval for pivot_root on rootfs (if thats detectable)
<viro> of course it is
--> plai_ (~plai@gatekeeper.intransa.com) has joined #kernel
<viro> sigh...
<viro> what we should do is prohibit any mount activity on detached vfsmounts, *period*
<viro> akpm: please, consider that patch in its current form strongly vetoed
<crlf> viro: if you restructure vfsmount trees to not depend on namespaces, there's no need for that restriction
<akpm> viro: OK.  You were cc'ed on it a lot at the time..
<viro> akpm: -*-Mutt: /var/spool/mail/viro [Msgs:23748 New:21604 Del:83 Post:31 128M]---(threa

Re: [gentoo-user] initramfs, network diskless boot, init process, problems with switchroot (pivot_root)

[Hans-Werner Hilse]

Hi,

On Wed, 21 Jun 2006 15:51:51 -0700
"Richard Fish" <bigfish@asmallpond.org> wrote:

> My final comments on that bug are based on the attached email that I
> received from Andrew Morton that made it clear that Al Viro was
> opposed to pivot_root being used from an initramfs.  (BTW, viro's
> comments are not very polite, but you have to expect such directness
> from kernel hackers!)
> 
> Note, be sure we are talking about an initramfs here, not an initrd.
> From an initrd it is still possible and supported to use pivot_root.

OK, I have to apologize. I totally missed that you were exec'ing
chroot. OK, I definately stand currected. But I was pretty sure I once
did this from initramfs. But that was admittedly back in the 2.6.10
days, I think. Of course in this case /sbin/init also gets PID 1.

-hwh
(happy to not have been the one complaining to the kernel devs :-) and
also forgetting to send attachments. I need a plugin that saves me from
doing this, e.g. searching for the word "attachment" and complaining if
there isn't one...)
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] initramfs, network diskless boot, init process, problems with switchroot (pivot_root)

[Richard Fish]

On 6/21/06, Hans-Werner Hilse <hilse@web.de> wrote:
> chroot. OK, I definately stand currected. But I was pretty sure I once
> did this from initramfs. But that was admittedly back in the 2.6.10
> days, I think.

Yeah, I did it too! ;->  Worked great until I tried to use
fbsplash/bootsplash, which did a 'mount --move'  as part of it's
setup.

Cheers,
-Richard
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] initramfs, network diskless boot, init process, problems with switchroot (pivot_root)

[Claudinei Matos]

[-- Attachment #1: Type: text/plain, Size: 1078 bytes --]

Well, too many replies, thank you all :)
I could solve my problem using strace to detect where init stops (of course
I had to recompile init allowing to not be PID 1). Looking at the errors
reported on strace log I could see that the problem was with the /dev/null
and /dev/console doesn't exists so I've changed my init script to  create
they both and now everything works like a charm :)
Oh, and actually I'm using switch_root from busybox which is very similar to
pivot_root.

Again, tks for all replies,

Claudinei Matos

On 6/21/06, Richard Fish <bigfish@asmallpond.org> wrote:
>
> On 6/21/06, Hans-Werner Hilse <hilse@web.de> wrote:
> > chroot. OK, I definately stand currected. But I was pretty sure I once
> > did this from initramfs. But that was admittedly back in the 2.6.10
> > days, I think.
>
> Yeah, I did it too! ;->  Worked great until I tried to use
> fbsplash/bootsplash, which did a 'mount --move'  as part of it's
> setup.
>
> Cheers,
> -Richard
> --
> gentoo-user@gentoo.org mailing list
>
>


-- 
Claudinei Matos
<claudineimatos@gmail.com>
55-21-81980605

[-- Attachment #2: Type: text/html, Size: 1591 bytes --]