lilo fails on root raid 1 on 2.6.4

lilo fails on root raid 1 on 2.6.4

[Lewis Shobbrook]

Hi All,

I've struck an issue attempting to run lilo when booting to a raid 1
system using a stock Debain 2.6.4-1-686 unstable release kernel image.

After sucecssfully booting into the raid filesystem, attempts to run
lilo give me the error...

Fatal: Trying to map files from unnamed device 0x0000 (NFS/RAID mirror
down ?)


This error occurrs regardless of how I specify the root device (eg
root=dev/hda2 or root=/dev/md0).
The mirror is in a failed disk state, with /dev/hda2 holding on to old
file system and manually marked as failed.
(This process has never been an issue in the past) Conf files listed
below...

Has anyone come across this before?

Any suggestions much appreciated,

Cheers,

Lewis


Raidtab

raiddev /dev/md0
        raid-level              1
        nr-raid-disks           2
        nr-spare-disks          0
        persistent-superblock   1
        device                  /dev/hda2
        failed-disk             0
        device                  /dev/hdc2
        raid-disk               1

fstab

/dev/md0        /               ext3    defaults,errors=remount-ro
0       1
/dev/hda1       none            swap    sw
0       0
proc            /proc           proc    defaults
0       0
/dev/fd0        /floppy         auto    defaults,user,noauto
0       0
/dev/cdrom      /cdrom          iso9660 defaults,ro,user,noauto
0       0


lilo.conf
boot=/dev/hda
root=/dev/md0
install=/boot/boot-menu.b
map=/boot/map
prompt
timeout=100


default=Linux

image=/vmlinuz
        label=Linux
        read-only
        initrd=/initrd.img
#        append="root=/dev/md0"

image=/vmlinuz.old
        label=LinuxOLD
        read-only
        initrd=/initrd.img.old
        append="root=/dev/hda2"

Re: lilo fails on root raid 1 on 2.6.4

[Neil Brown]

On Thursday April 8, lshobbrook@fasttrack.net.au wrote:
> 
> Hi All,
> 
> I've struck an issue attempting to run lilo when booting to a raid 1
> system using a stock Debain 2.6.4-1-686 unstable release kernel image.
> 
> After sucecssfully booting into the raid filesystem, attempts to run
> lilo give me the error...
> 
> Fatal: Trying to map files from unnamed device 0x0000 (NFS/RAID mirror
> down ?)
> 

You probably need a "raid-extra-boot=" option in lilo.conf
Just list the partition(s) that are currently active and it might
work.

Check "man lilo.conf" for more details

NeilBrown

> 
> This error occurrs regardless of how I specify the root device (eg
> root=dev/hda2 or root=/dev/md0).
> The mirror is in a failed disk state, with /dev/hda2 holding on to old
> file system and manually marked as failed.
> (This process has never been an issue in the past) Conf files listed
> below...
> 
> Has anyone come across this before?
> 
> Any suggestions much appreciated,
> 
> Cheers,
> 
> Lewis
> 
> 
> Raidtab
> 
> raiddev /dev/md0
>         raid-level              1
>         nr-raid-disks           2
>         nr-spare-disks          0
>         persistent-superblock   1
>         device                  /dev/hda2
>         failed-disk             0
>         device                  /dev/hdc2
>         raid-disk               1
> 
> fstab
> 
> /dev/md0        /               ext3    defaults,errors=remount-ro
> 0       1
> /dev/hda1       none            swap    sw
> 0       0
> proc            /proc           proc    defaults
> 0       0
> /dev/fd0        /floppy         auto    defaults,user,noauto
> 0       0
> /dev/cdrom      /cdrom          iso9660 defaults,ro,user,noauto
> 0       0
> 
> 
> lilo.conf
> boot=/dev/hda
> root=/dev/md0
> install=/boot/boot-menu.b
> map=/boot/map
> prompt
> timeout=100
> 
> 
> default=Linux
> 
> image=/vmlinuz
>         label=Linux
>         read-only
>         initrd=/initrd.img
> #        append="root=/dev/md0"
> 
> image=/vmlinuz.old
>         label=LinuxOLD
>         read-only
>         initrd=/initrd.img.old
>         append="root=/dev/hda2"
> 
> 
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

RE: lilo fails on root raid 1 on 2.6.4

[Lewis Shobbrook]

Thanks for the reply Neil,
 
> > Hi All,
> > 
> > I've struck an issue attempting to run lilo when booting to 
> a raid 1 
> > system using a stock Debain 2.6.4-1-686 unstable release 
> kernel image.
> > 
> > After sucecssfully booting into the raid filesystem, 
> attempts to run 
> > lilo give me the error...
> > 
> > Fatal: Trying to map files from unnamed device 0x0000 
> (NFS/RAID mirror 
> > down ?)
> > 
> 
> You probably need a "raid-extra-boot=" option in lilo.conf
> Just list the partition(s) that are currently active and it 
> might work.


I've been able to apply the "raid-extra-boot=" option on an alternate
system and did a bit of toying around with the syntax to see what was
required and possible. All was well on the test machine, running the
same 2.6.4-1 kernel, however the problems persist.... 

With raid-extra-boot=/dev/hdc,hda  (hda being marked as failed)...

Fatal: Unable to open dev/hda

With raid-extra-boot=/dev/hdc ...
Fatal: Trying to map files from unnamed device 0x0000 (NFS/RAID mirror
down ?)

And with raid-extra-boot=auto... 

Warning: Boot sector on  /dev/md0  will depend upon the BIOS device code
  passed in the DL register being accurate.  Install Master Boot Records
  with the 'lilo -M' command, and activate the RAID1 partitions with the
  'lilo -A' command.
Fatal: Trying to map files from unnamed device 0x0000 (NFS/RAID mirror
down ?)

...==>...

# lilo -M /dev/hda
/boot/boot.0300 exists - no /dev/hda backup copy made.

 *** oops, well at least I have a boot disk....

The Master Boot Record of  /dev/hda  has been updated.
# lilo -M /dev/hdc
Backup copy of /dev/hdc in /boot/boot.1600
The Master Boot Record of  /dev/hdc  has been updated.
# lilo -A /dev/hdc
/dev/hdc2
# lilo -A /dev/hda
/dev/hda2
# lilo
Warning: Boot sector on  /dev/md0  will depend upon the BIOS device code
  passed in the DL register being accurate.  Install Master Boot Records
  with the 'lilo -M' command, and activate the RAID1 partitions with the
  'lilo -A' command.
Fatal: Trying to map files from unnamed device 0x0000 (NFS/RAID mirror
down ?)

Perhaps it may be due stale bios HDD detection as...
lilo -T geom

    bios=0x00, cylinders=80, heads=2, sectors=18
        (  1.44Mb         2,880 sectors)        C:H:S supported (IBM PC
bios)

BIOS reports 2 hard drives
    bios=0x80, cylinders=1022, heads=255, sectors=63    vol-ID: 95759575
        (  8.40Gb    16,418,430 sectors)        LBA32 supported (EDD
bios)
    bios=0x81, cylinders=1022, heads=255, sectors=63    vol-ID: 3A19FFA5
        (  8.40Gb    16,418,430 sectors)        LBA32 supported (EDD
bios)

And I know the Hard drives are not the same size and neither are 8.40 Gb

Cheers,

Lewis

Re: lilo fails on root raid 1 on 2.6.4

[Michael Tokarev]

Lewis Shobbrook wrote:
> Thanks for the reply Neil,
>  
> 
>>>Hi All,
>>>
>>>I've struck an issue attempting to run lilo when booting to a raid 1 
>>>system using a stock Debain 2.6.4-1-686 unstable release kernel image.
>>>After sucecssfully booting into the raid filesystem, attempts to run 
>>>lilo give me the error...
>>>
>>>Fatal: Trying to map files from unnamed device 0x0000 
>>(NFS/RAID mirror down ?)
>>
>>You probably need a "raid-extra-boot=" option in lilo.conf
>>Just list the partition(s) that are currently active and it 
>>might work.
> 
> I've been able to apply the "raid-extra-boot=" option on an alternate
> system and did a bit of toying around with the syntax to see what was
> required and possible. All was well on the test machine, running the
> same 2.6.4-1 kernel, however the problems persist.... 
[snip]

Quite some time ago I discovered the "ultimate solution" for this
lilo/raid stuff (even back when LILO wasn't aware of raid), and I
mentioned it on this (linux-raid) list several times.

Instead of using LILO to install to raid devices, use standard MBR
(comes in `mbr' debian package, install-mbr /dev/sda etc, and LILO
itself may ), mark your
boot raid partitions active and use "boot=/dev/md.." in lilo.conf.
For this to work, you should have your boot partitions at exactly
the same place on all disks, but this isn't a problem in most cases
(after all, boot partition shouldn't be large and it's easy to place
it somewhere at the beginning of the disk, on the same partition).

> With raid-extra-boot=/dev/hdc,hda  (hda being marked as failed)...
> 
> Fatal: Unable to open dev/hda
> 
> With raid-extra-boot=/dev/hdc ...
> Fatal: Trying to map files from unnamed device 0x0000 (NFS/RAID mirror
> down ?)

Hmm... this is a bug somewhere in lilo it seems...

/mjt

RE: lilo fails on root raid 1 on 2.6.4

[Lewis Shobbrook]

Hi Michael,
Thanks for the sugestions.  
> >>>
> >>>I've struck an issue attempting to run lilo when booting 
> to a raid 1
> >>>system using a stock Debain 2.6.4-1-686 unstable release 
> kernel image.
> >>>After sucecssfully booting into the raid filesystem, 
> attempts to run 
> >>>lilo give me the error...
> >>>
> >>>Fatal: Trying to map files from unnamed device 0x0000
> >>(NFS/RAID mirror down ?)
> >>
> >>You probably need a "raid-extra-boot=" option in lilo.conf 
> Just list 
> >>the partition(s) that are currently active and it might work.
> > 
> > I've been able to apply the "raid-extra-boot=" option on an 
> alternate 
> > system and did a bit of toying around with the syntax to 
> see what was 
> > required and possible. All was well on the test machine, 
> running the 
> > same 2.6.4-1 kernel, however the problems persist....
> [snip]
> 
> Quite some time ago I discovered the "ultimate solution" for 
> this lilo/raid stuff (even back when LILO wasn't aware of 
> raid), and I mentioned it on this (linux-raid) list several times.
> 
> Instead of using LILO to install to raid devices, use 
> standard MBR (comes in `mbr' debian package, install-mbr 
> /dev/sda etc, and LILO itself may ), mark your boot raid 
> partitions active and use "boot=/dev/md.." in lilo.conf. For 
> this to work, you should have your boot partitions at exactly 
> the same place on all disks, but this isn't a problem in most 
> cases (after all, boot partition shouldn't be large and it's 
> easy to place it somewhere at the beginning of the disk, on 
> the same partition).

I installed mbr using install-mbr to both devices, still comes back to
the issue of...

Fatal: Trying to map files from unnamed device 0x0000 (NFS/RAID mirror
down ?)

The active partitions are both the second partition, however their
location on the disks do vary. The first partition on the first drive is
2G swap, while the first partition of the second drive is 750Mb.  Once
I'd installed lilo I intended on replicating the layout of the second
drive to the first.  I'm reluctant to do this until I'm sure I don't
need to go back to the original filesystem. 

> > With raid-extra-boot=/dev/hdc,hda  (hda being marked as failed)...
> > 
> > Fatal: Unable to open dev/hda
> > 
> > With raid-extra-boot=/dev/hdc ...
> > Fatal: Trying to map files from unnamed device 0x0000 
> (NFS/RAID mirror 
> > down ?)
> 
> Hmm... this is a bug somewhere in lilo it seems...
> 

The crazy thing is that even after altering lilo.conf with
boot=/dev/hda & root=/dev/hda and removing any reference to /dev/md0...
The result is the same.... Fatal: ....
When I am in the original filesystem I can install lilo to the mbr on
both drives.  It's only once in the raid filesystem (copied from the
original) that we get the error.

I created the raid device and did some testing in another machine prior
to moving the device into place, could this be the source of the issue?



Cheers,

Lewis