Public SourceMage Discussion List

[Thomas Orgis <thomas-forum AT orgis.org>]

Am Wed, 15 Nov 2006 18:59:18 -0800
schrieb Eric Sandall <eric AT sandall.us>: 

> OR, a simpler approach. ;)
> 
> Sync up with test, and revert these:
> commit adbab1098ff1a2dd9f1eff2882cb6b47c1066ca7
> Author: Thomas Orgis <sobukus AT sourcemage.org>
> Date:   Thu Nov 2 20:12:10 2006 +0100
> 
>     dmsetup in init.d/mountall.sh failing is not fatal (simply no dm in
> kernel),
> 
> commit 772f276d53fecceaf68441af953afc041887da62
> Author: Thomas Orgis <sobukus AT sourcemage.org>
> Date:   Sun Oct 29 14:39:03 2006 +0100
> 
>     enable device mapper nodes discovery in init.d/mountall.sh
> 
> Assuming those are the two changes that break LVM.

Can someone please explain me what I did in October / November to cause
that LVM breaking?
I am aware of http://bugs.sourcemage.org/show_bug.cgi?id=9177 and that has 
its last comment from July!

Also please someone point out the point where LVM is cracked in these diffs::

--- a/smgl/init.d/init.d/mountall.sh
+++ b/smgl/init.d/init.d/mountall.sh
@@ -64,6 +64,14 @@ # bug 7311
     builtin echo 'none /proc proc rw 0 0' >> /etc/mtab
   fi
 
+  # find devices that were set up in initramfs/rd
+  if optional_executable /sbin/dmsetup
+  then 
+    echo "(re)creating device mapper nodes"
+    /sbin/dmsetup mknodes
+    evaluate_retval
+  fi
+
   if optional_executable /sbin/vgscan && optional_executable /sbin/vgchange 
; then
     echo -n "Scanning for and initializing all available LVM volume 
groups..."
     /sbin/vgscan       --ignorelockingfailure  --mknodes  &&


--- a/smgl/init.d/init.d/mountall.sh
+++ b/smgl/init.d/init.d/mountall.sh
@@ -68,8 +68,8 @@ # bug 7311
   if optional_executable /sbin/dmsetup
   then 
     echo "(re)creating device mapper nodes"
+    # won't check retval because kernel just may lack device mapper
     /sbin/dmsetup mknodes
-    evaluate_retval
   fi
 
   if optional_executable /sbin/vgscan && optional_executable /sbin/vgchange 
; then


I must assume that you picked the wrong commits by accident... or is there 
indeed a dmsetup mknodes preventing vgscan to succeed?


Thomas.