Bug#392764: partman: [powerpc] RAID support is broken on powermac hardware (64bit, XServe G5)
On Fri, Oct 13, 2006 at 10:56:08PM +0200, Frans Pop wrote:
> On Friday 13 October 2006 13:43, Sven Luther wrote:
> > As the title says, an install on my XServe G5, with two brand new 320GB
> > sata disks, fails. I can manually setup the RAID partitions, but when i
> > go into the raid tool, it tells me something about the RAID flag not
> > being found or something.
>
> That is not very helpful. Please provide exact error message.
>
> Please add 'set -x' in /lib/partman/definitions.sh, reproduce error and
> send full log (gzipped) to BR.
Ok, i think i found the culprit of this bug, the message is comming from
package mdcfg, file mdcfg.sh. its the mdcfg/noparts message.
# this will set PARTITIONS and NUM_PART global variables.
get_partitions() {
PARTITIONS=""
# Get a list of RAID partitions. This only works if there is no
# filesystem on the partitions, which is fine by us.
RAW_PARTITIONS=`/usr/lib/partconf/find-partitions --ignore-fstype 2>/dev/null | grep "[[:space:]]RAID[[:space:]]" |cut -f1`;
# Convert it into a proper list form for a select question
# (comma seperated)
NUM_PART=0
for i in ${RAW_PARTITIONS}; do
DEV=`echo ${i}|sed -e "s/\/dev\///"`
REALDEV=$(mapdevfs "$i")
MAPPEDDEV=$(echo "$REALDEV" | sed -e "s/\/dev\///")
if grep -q "\(${DEV}\|${MAPPEDDEV}\)" /proc/mdstat; then
continue
fi
if [ -z "${PARTITIONS}" ] ; then
PARTITIONS="$REALDEV"
else
PARTITIONS="${PARTITIONS}, $REALDEV"
fi
NUM_PART=$(($NUM_PART + 1))
done
if [ -z "${PARTITIONS}" ] ; then
db_input critical mdcfg/noparts
db_go
return 1
fi
return 0
}
As you can see, the find-partitions call returns nothing on the box, even
though the partitions are listed in partman.
Since find-partitions is using parted, this may be correlated to the bug about
parted not listing the partitions on the device :
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=392767
#392767: [mac] parted is unable to reread partition tables created by d-i/partman.
the partition tables show up after the reboot though, which is rather strange.
Maybe some problem with devfs compatibility stuff ? Or a missing kernel module ?
Friendly,
Sven Luther
Reply to: