I wrote:
The attached patch changes partman-md to keep the settings of the md
partitions when partman restarts which is enough to allow crypto-on-raid
setups to work again, thus fixing #393728, #397872 and #398464.
In order to minimize the impact of the patch this close to the Etch
release, I've implemented it so that most of it only applies when the
md-device is used for crypto.
Tested with crypto-on-md and with regular fs-on-md.
Would it be ok to commit this?
And this time...with patch included :)
--
David Härdeman
Index: debian/changelog
===
--- debian/changelog(revision 42618)
+++ debian/changelog(working copy)
@@ -1,3 +1,11 @@
+partman-md (31) UNRELEASED; urgency=low
+
+ * Make sure that settings are not reset during partman init stage if
+the device is used as a crypto volume.
+Closes: #393728, #397872, #398464.
+
+ -- David Härdeman <[EMAIL PROTECTED]> Tue, 14 Nov 2006 23:02:30 +0100
+
partman-md (30) unstable; urgency=low
[ Colin Watson ]
Index: init.d/md-devices
===
--- init.d/md-devices (revision 42618)
+++ init.d/md-devices (working copy)
@@ -27,7 +27,14 @@
sed -e 's/^\(md.*\) : active \([[:alnum:]]*\).*/\1/'); do
NUMBER=$(echo ${i}|sed -e "s/^md//")
DEVICE="/var/lib/partman/devices/=dev=md=${NUMBER}"
- mkdir -p ${DEVICE}
+ OLDDEVICE="/var/lib/partman/old_devices/=dev=md=${NUMBER}"
+
+ if [ -d "$OLDDEVICE" ]; then
+ mv "$OLDDEVICE" "$DEVICE"
+ else
+ mkdir -p ${DEVICE}
+ fi
+
cd ${DEVICE}
echo "/dev/md/${NUMBER}" > ${DEVICE}/device
@@ -54,6 +61,12 @@
fi
fi
+ if [ -e "$DEVICE/method" ] && [ "$(cat "$DEVICE/method")" = "crypto" ];
then
+ open_dialog DISK_UNCHANGED
+ close_dialog
+ continue
+ fi
+
open_dialog NEW_LABEL loop
close_dialog