Merge lp:~serge-hallyn/ubuntu/maverick/multipath-tools/backport-fixes into lp:ubuntu/maverick/multipath-tools

[Serge Hallyn]

These fixes are needed in lucid and maverick. I've done a separate merge request of the more uptodate Debian experimental package for natty, which is why I'm not requesting this be merged for natty first.

        a. device map not ready before root mount (use bug 686832)
        b. modprobe -q not -Q (#644481)
        c. fix pathnames in uevents (#660597)
        d. scsi_wait_scan needed

I suspect I need to create a separate SRU with targeted debdiffs for each one, but am hoping there is a way to do it from the bzr tree.

[Chuck Short (zulcss)]

If this is for an SRU in maverick, then the changelog probably needs to be collapsed into one entry and the release have to be maverick-propsed and the version has to be ubuntu4.1

review: Needs Fixing

Unmerged revisions

38. By Serge Hallyn on 2011-01-04

Have the initramfs/local-top script use the scsi_wait_scan module to
make sure all LUNs are found before we proceed.

37. By Serge Hallyn on 2011-01-04

Add patch to fix the expected pathname from multipath uevents
(LP: #660597)

36. By Serge Hallyn on 2011-01-04

* Fix in debian/kpartx.udev:
    - modprobe -q instead of incorrect modprobe -Q (LP: #644481)

35. By Serge Hallyn on 2011-01-04

Make the initramfs local-top script wait until all devices
are mapped, to make sure no /dev/sdget mounted before
multipath has a chance to take it. (LP: #686832)

=== modified file 'debian/changelog'
--- debian/changelog	2010-03-08 14:50:04 +0000
+++ debian/changelog	2011-01-04 22:42:18 +0000
@@ -1,3 +1,32 @@
+multipath-tools (0.4.8-14ubuntu8) natty; urgency=low
+
+  * Have the initramfs/local-top script use the scsi_wait_scan module to
+    make sure all LUNs are found before we proceed.
+
+ -- Serge Hallyn <[e-mail]>  Tue, 04 Jan 2011 16:34:56 -0600
+
+multipath-tools (0.4.8-14ubuntu7) natty; urgency=low
+
+  * Add patch to fix the expected pathname from multipath uevents
+    (LP: #660597)
+
+ -- Serge Hallyn <[e-mail]>  Tue, 04 Jan 2011 16:06:29 -0600
+
+multipath-tools (0.4.8-14ubuntu6) maverick; urgency=low
+
+  * Fix in debian/kpartx.udev:
+      - modprobe -q instead of incorrect modprobe -Q (LP: #644481)
+
+ -- Serge Hallyn <[e-mail]>  Tue, 04 Jan 2011 15:58:13 -0600
+
+multipath-tools (0.4.8-14ubuntu5) maverick; urgency=low
+
+  * Make the initramfs local-top script wait until all devices
+    are mapped, to make sure no /dev/sd* get mounted before
+    multipath has a chance to take it. (LP: #686832)
+
+ -- Serge Hallyn <[e-mail]>  Tue, 04 Jan 2011 15:55:48 -0600
+
 multipath-tools (0.4.8-14ubuntu4) lucid; urgency=low
 
   * debian/control: Move libreadline5-dev build dependency to libreadline-dev.

=== modified file 'debian/initramfs/local-top'
--- debian/initramfs/local-top	2009-02-12 15:07:42 +0000
+++ debian/initramfs/local-top	2011-01-04 22:42:18 +0000
@@ -21,9 +21,19 @@
 modprobe -q dm-multipath
 modprobe -q dm-emc
 
+# Sync waiting for storage.
+{ rmmod scsi_wait_scan ; modprobe scsi_wait_scan ; rmmod scsi_wait_scan ; } >/dev/null 2>&1
+
 echo "Discovering multipaths..."
 
 /sbin/multipath -v $VERBOSITY
 
+if [ -x /sbin/kpartx ]; then
+	/sbin/dmsetup ls --target multipath --exec "/sbin/kpartx -a -p p" >/dev/null
+fi
+udevadm settle
+
+echo "Completed (discovering multipaths)."
+
 exit 0
 

=== modified file 'debian/kpartx.udev'
--- debian/kpartx.udev	2009-03-10 11:54:01 +0000
+++ debian/kpartx.udev	2011-01-04 22:42:18 +0000
@@ -7,7 +7,7 @@
 # Always attempt to load dm-multipath. This way we can get rid of the
 # init script.
 ACTION=="add|change", SUBSYSTEM=="block", \
-	 RUN+="/sbin/modprobe -Qba dm-multipath"
+	 RUN+="/sbin/modprobe -qba dm-multipath"
 
 KERNEL!="dm-*", GOTO="kpartx_end"
 ACTION=="remove", GOTO="kpartx_end"

=== added file 'debian/patches/1002--Fix-for-uevent-devpath-handling.patch'
--- debian/patches/1002--Fix-for-uevent-devpath-handling.patch	1970-01-01 00:00:00 +0000
+++ debian/patches/1002--Fix-for-uevent-devpath-handling.patch	2011-01-04 22:42:18 +0000
@@ -0,0 +1,47 @@
+commit 7fa7affc3d23dd9dc906804d22a61144bca9f9b9
+Author: Benjamin Marzinski <[e-mail]>
+Date:   Thu Dec 11 16:03:28 2008 -0600
+
+    Fix for uevent devpath handling
+    
+    This is necessary to make uevents work on fedora, since devpath appears as
+    something like:
+    '/devices/pci0000:00/0000:00:0a.0/0000:06:00.0/host11/rport-11:0-1/target11:0:1/11:0:1:0/block/sdi'
+    
+    It simply strips off the everything up to the /block.
+
+    [ also pull in trivial fix of potential stack overflow,
+    commit 15d4bdddcb9b71e0ec6fecc3c37a1b8cae8f51ff, by making
+    a and b length 11 not 10 ]
+    
+    Signed-off-by: Benjamin Marzinski <[e-mail]>
+
+Index: delme1/multipathd/main.c
+===================================================================
+--- delme1.orig/multipathd/main.c	2011-01-04 16:00:58.936899000 -0600
++++ delme1/multipathd/main.c	2011-01-04 16:03:04.783628002 -0600
+@@ -615,14 +615,20 @@
+ static int
+ uev_discard(char * devpath)
+ {
+-	char a[10], b[10];
++	char *tmp;
++	char a[11], b[11];
+ 
+ 	/*
+ 	 * keep only block devices, discard partitions
+ 	 */
+-	if (sscanf(devpath, "/block/%10s", a) != 1 ||
+-	    sscanf(devpath, "/block/%10[^/]/%10s", a, b) == 2) {
+-		condlog(4, "discard event on %s", devpath);
++	tmp = strstr(devpath, "/block/");
++	if (tmp == NULL){
++		condlog(0, "no /block/ in '%s'", devpath);
++		return 1;
++	}
++	if (sscanf(tmp, "/block/%10s", a) != 1 ||
++	    sscanf(tmp, "/block/%10[^/]/%10s", a, b) == 2) {
++		condlog(0, "discard event on %s", devpath);
+ 		return 1;
+ 	}
+ 	return 0;

=== modified file 'debian/patches/series'
--- debian/patches/series	2010-02-12 11:52:22 +0000
+++ debian/patches/series	2011-01-04 22:42:18 +0000
@@ -12,3 +12,4 @@
 0012--libmultipath-Update-discovery-to-work-with-new-sys.patch
 1000--set-umask-in-multipathd.patch
 1001--intel-mpath-prio-alua.patch
+1002--Fix-for-uevent-devpath-handling.patch