[Pkg-sysvinit-devel] Bug#623174: Bug#621803: Add support for /run directory

[rleigh]

On Tue, Apr 26, 2011 at 12:06:14PM +0100, Roger Leigh wrote:
> block 623174 by 621803
> thanks
> 
> On Fri, Apr 22, 2011 at 08:14:47PM +0100, Roger Leigh wrote:
> > On Mon, Apr 18, 2011 at 07:38:41PM +0100, Roger Leigh wrote:
> > > By the way, could you consider adding this patch to your branch:
> > 
> > diff --git a/init b/init
> > index 38c8a5d..cdbfe50 100755
> > --- a/init
> > +++ b/init
> > @@ -25,7 +25,7 @@ if ! mount -t devtmpfs -o mode=0755 none /dev; then
> >  fi
> >  mkdir /dev/pts
> >  mount -t devpts -o noexec,nosuid,gid=5,mode=0620 none /dev/pts || true
> > -mount -t tmpfs -o nodev,noexec,nosuid,mode=0755 none /run
> > +mount -t tmpfs -o "nosuid,size=20%,mode=0755" tmpfs /run
> >  mkdir /run/initramfs
> >  
> >  # Export the dpkg architecture
> > 
> > Slightly simpler than the previous patch, but does the same job.
> > Please note that I've now patched initscripts
> > (http://paste.debian.net/114826/ and
> > http://lists.alioth.debian.org/pipermail/pkg-sysvinit-devel/2011-April/004990.html) so that the filesystems mounted in the initramfs and later moved
> > over to the root filesystem are remounted with the mount options
> > specified in /etc/fstab (or /etc/default/tmpfs if no fstab entry
> > exists).  In consequence, it's no longer critical that the mount
> > option defaults in the initramfs match the "real" root filesystem
> > defaults, because they will always be applied by mountkernfs/
> > mountdevsubfs.  So if you choose not to apply this, that's fine--it's
> > just a "nice to have" addition; the patch as it stands without the
> > above change is equally acceptable.
> > 
> > Thanks for doing this.  Do we have any ETA on an upload?
> 
> The updated initscripts should be uploaded to experimental later today.
> This includes all the changes listed above.
> 
> I've added the block above because we can't make an upload of sysvinit/
> initscripts to unstable until we have an updated initramfs-tools in
> unstable, or else udev will break.  Therefore in order to have a smooth
> transition we need initramfs-tools uploading first, followed by
> initscripts.
> 
> Marco, have you tested this upgrade path?  That is /run in the
> initramfs and no /run on the rootfs?  Is udev checking for that and
> not using /run in this scenario (or moving the files from /run to
> /dev/.udev)?  We will need to cope with this for a brief transitional
> period when initramfs-tools is updated and initscripts is not.

Testing with initramfs-tools (maks/run) with current unstable shows
udev appearing to work correctly with it using /dev/.udev when /run
is not present on the host system.  And also with /run present on
the host system in addition.  In the former case, it uses /run in
the initramfs and moved it to /dev/.udev on the host.  In the latter
it uses /run in the initramfs and on the host as well.  So from the
udev side, it does appear to be working correctly in all cases.  So
from this POV, the maks/run branch is working just fine.


In a kvm VM, I do see some errors from udev at startup:

--------------------------
Loading Linux 2.6.38-2-amd64 ...
Loading initial ramdisk ...
Loading, please wait...
udevd[125]: exec of program 'net.agent' failed

udevd[126]: exec of program 'net.agent' failed

udevd[137]: exec of program 'cdrom_id --export /dev/sr0' failed

udevd[142]: exec of program 'cdrom_id --export /dev/sr0' failed

INIT: version 2.88 booting
Using makefile-style concurrent boot in runlevel S.
Starting the hotplug events dispatcher: udevd
--------------------------

This is with udev 168-1.  However, I see the above without any changes
to initramfs-tools or initscripts, and this doesn't changes with /run
present in the initramfs or both the initramfs and rootfs, so it's a
separate issue, and so far specific to the VM image.


Regards,
Roger

-- 
  .''`.  Roger Leigh
 : :' :  Debian GNU/Linux             http://people.debian.org/~rleigh/
 `. `'   Printing on GNU/Linux?       http://gutenprint.sourceforge.net/
   `-    GPG Public Key: 0x25BFB848   Please GPG sign your mail.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://lists.alioth.debian.org/pipermail/pkg-sysvinit-devel/attachments/20110426/c844ab51/attachment.pgp>