Bug 248024

+++ This bug was initially created as a clone of Bug #247265 +++

Description of problem:
When the xen block frontend driver is built as a module the module load is only
synchronous up to the point where the frontend and the backend become connected
rather than when the disk is added.

This means that there can be a race on boot between loading the module and
loading the dm-* modules and doing the scan for LVM physical volumes (all in the
initrd). In the failure case the disk is not present until after the scan for
physical volumes is complete.

Version-Release number of selected component (if applicable): 2.6.18-8.1.6.EL
Also applies to RHEL4u5 2.6.9-55.0.2.EL

Upstream fix is:
http://xenbits.xensource.com/linux-2.6.18-xen.hg?rev/11483a00c017
http://xenbits.xensource.com/kernels/rhel4x.hg?rev/156e3eaca552

How reproducible:
Not precisely determined. Seems to be related to high load in domain 0 or in
multiple other domains implying delays in scheduling the backend driver. 

Actual results:
  Loading xenblk.ko module
  Registering block device major 202
  xvda:Loading dm-mod.ko module
  <6>device-mapper: ioctl: 4.11.0-ioctl (2006-09-14) initialised:
dm-devel
  Loading dm-mirror.ko module
  Loading dm-zero.ko module
  Loading dm-snapshot.ko module
  Making device-mapper control node
  Scanning logical volumes
  Reading all physical volumes. This may take a while...
  xvda1 xvda2
  No volume groups found
  Activating logical volumes
  Volume group "VolGroup00" not found
  Creating root device.
  Mounting root filesystem.
  mount: could not find filesystem '/dev/root'
  Setting up other filesystems.
  Setting up new root fs
  setuproot: moving /dev failed: No such file or directory
  no fstab.sys, mounting internal defaults
  setuproot: error mounting /proc: No such file or directory
  setuproot: error mounting /sys: No such file or directory
  Switching to new root and running init.
  unmounting old /dev
  unmounting old /proc
  unmounting old /sys
  switchroot: mount failed: No such file or directory
  Kernel panic - not syncing: Attempted to kill init!

Expected results:
  Registering block device major 202
  xvda: xvda1 xvda2
  Loading dm-mod.ko module
  device-mapper: ioctl: 4.11.0-ioctl (2006-09-14) initialised: dm-devel
  Loading dm-mirror.ko module
  Loading dm-zero.ko module
  Loading dm-snapshot.ko module
  Making device-mapper control node
  Scanning logical volumes
  Reading all physical volumes. This may take a while...
  Found volume group "VolGroup00" using metadata type lvm2
  Activating logical volumes
  2 logical volume(s) in volume group "VolGroup00" now active

Additional info:

-- Additional comment from pm-rhel on 2007-07-06 12:23 EST --
This request was evaluated by Red Hat Product Management for inclusion in a Red
Hat Enterprise Linux maintenance release.  Product Management has requested
further review of this request by Red Hat Engineering, for potential
inclusion in a Red Hat Enterprise Linux Update release for currently deployed
products.  This request is not yet committed for inclusion in an Update
release.

-- Additional comment from rjones on 2007-07-07 03:49 EST --
AFAICS, this only solves half the problem.  Waiting for the disk to be
added gets you so far.  You also need to wait for the partitions on
the disk to be scanned.

See 241793 comment 15 (and onwards).

-- Additional comment from rjones on 2007-07-07 03:50 EST --
Bug 241793 comment 15
(Bugzilla _really_ needs a preview feature).

-- Additional comment from ijc.uk on 2007-07-09 04:39 EST --
The proposed patch causes the module load to wait until add_disk() has returned.
In 2.6.18 at least this calls down to rescan_partitions in a synchronous manner.

(add_disk->register_disk->blkdev_get->do_open->rescan_partitions).

-- Additional comment from sct on 2007-07-11 12:03 EST --
May also fix bug 241793.

Richard, does this look reasonable?  And given that it's a fairly simple,
xen-only, already-upstream fix, do we want to consider an exception for it for 5.1?


-- Additional comment from kraxel on 2007-07-12 09:46 EST --
trapped into this issue too while trying to make rhel5 boot with pv-on-hvm
drivers.  Fixed it this way:

--- /sbin/mkinitrd.kraxel       2007-07-11 13:25:06.000000000 +0200
+++ /sbin/mkinitrd      2007-07-12 12:58:16.000000000 +0200
@@ -1239,9 +1239,11 @@
 unset usb_mounted
 
 if [ -n "$scsi" ]; then
-    emit "echo Waiting for driver initialization."
+    emit "echo Waiting for driver initialization (scsi)."
     emit "stabilized --hash --interval 250 /proc/scsi/scsi"
 fi
+emit "echo Waiting for driver initialization (partitions)."
+emit "stabilized --hash --interval 250 /proc/partitions"
 
 
 if [ -n "$vg_list" ]; then


-- Additional comment from rjones on 2007-07-12 10:30 EST --
I have now tested the Xen upstream fix and it works.

-- Additional comment from sct on 2007-07-12 12:15 EST --
OK, dev-ack+ then.  The new fix doesn't need userland support, is xen-only, and
is xen-upstream, which I think makes it easier to justify squeezing into 5.1.

Requesting blocker-beta as this is on a partner hot-list to fix (XenSource).