Bug#497110: boot loader installation failed when dmraid=true
On Friday 05 September 2008, Giuseppe Iuculano wrote:
> This is the ls -la after the first "Write changes in SATA RAID
> partitioning to disk" # ls -la /dev/mapper/
> drwxr-xr-x 2 root root 160 Sep 5 08:33 .
> drwxr-xr-x 10 root root 2840 Sep 5 08:33 ..
> crw-rw---- 1 root root 10, 60 Sep 5 08:28 control
> brw------- 1 root root 254, 0 Sep 5 08:32 sil_aiahbgbgaaaj
> brw------- 1 root root 254, 3 Sep 5 08:33 sil_aiahbgbgaaaj1
> brw------- 1 root root 254, 4 Sep 5 08:33 sil_aiahbgbgaaaj2
> brw------- 1 root root 254, 1 Sep 5 08:32 sil_aiahbgbgaaajp1
> brw------- 1 root root 254, 2 Sep 5 08:32 sil_aiahbgbgaaajp2
OK, this is definitely not what used to happen with earlier versions of
parted, so I think Jérémy was correct and this is a change in libparted.
The parted log shows for example:
parted_server: Read command: NEW_PARTITION
parted_server: command_new_partition()
parted_server: Note =dev=mapper=sil_aiahbgbgaaaj as changed
parted_server: Opening outfifo
parted_server: requested partition with type Primary
parted_server: requested partition with file system ext2
parted_server: add_primary_partition(disk(156299440),0-154296875)
parted_server: OUT: OK
parted_server: OUT: 1 32256-79003814399 79003782144 primary
ext2 /dev/mapper/sil_aiahbgbgaaajp1
What seems to happen now is that libparted creates the pX devices and that
dmraid creates the additional devices without "p". At least this duplication
is obviously wrong.
Because the whole current dmraid support is a bit of a hack, it may be
that the new libparted actually provides better dmraid support, but it seems
there are at least inconsistencies between what libparted does and what
dmraid expects.
I think I have done what I can here to identify the issues. From here on
it is up to Jérémy and Otavio to actually fix the bugs that have been
identified.
I really do not know at this point whether that means reverting some change
in libparted, or fixing up things in partman-dmraid.
What could still help is maybe carefully comparing *exactly* what happens
with the Lenny Beta2 version of the installer (which had the old libparted)
and the current version. You'd need to compare what's done both by
libparted and when the "Write SATA RAID changes" option is selected.
Main problem here is that so few people have the hardware for this (my
desktop does support it, but I'd need at least two extra harddisks to be
able to test dmraid installs with it).
I have not tried to use those "empty images" you pointed at. I doubt that
usging those with qemu really emulates a real SATA RAID system as I'd
expect the dmraid utility to interface with the BIOS RAID support.
After all, something needs to keep the disks in sync...
But if you can convince me otherwise and provide detailed instructions
I may take another look.
Cheers,
FJP
Reply to: