Debian Bug report logs - #597538
grub-ieee1275: grub-install needs to pass a --prefix option to grub-mkimage

Reply or subscribe to this bug.

Toggle useless messages

Message #5 received at submit@bugs.debian.org (full text, mbox, reply):

From: Daniel Kahn Gillmor <dkg@fifthhorseman.net>

To: Debian Bug Tracking System <submit@bugs.debian.org>

Subject: grub-ieee1275: grub-install needs to pass a --prefix option to grub-mkimage

Date: Mon, 20 Sep 2010 12:33:00 -0400

[Message part 1 (text/plain, inline)]

Package: grub-ieee1275
Version: 1.98+20100804-4
Severity: important
Tags: patch

a standard powerpc mac mini is unable to boot with grub after running

  grub-install /dev/hda

(the machine's only disk uses an apple partitioning scheme, with an HFS 
volume as (hd,apple2) mounted at /boot/grub)

during boot drops into an emergency rescue, saying:

Welcome to GRUB!
error: no device is set.
Entering rescue mode...


If i say:

set prefix=(hd,apple2)/
insmod normal
normal

then everything boots as expected.

it looks like grub-install isn't passing a --prefix option to 
grub-mkimage.

Re-running /usr/sbin/grub-install /dev/hda with the attached patch 
applied makes it so that this ugly workaround is no longer necessary.

Thanks for keeping grub in debian!

	--dkg

 
-- Package-specific info:

*********************** BEGIN /proc/mounts
/dev/mapper/vg_faith0-root / ext3 rw,relatime,errors=remount-ro,commit=5,data=ordered 0 0
/dev/mapper/vg_faith0-home /home ext3 rw,relatime,errors=continue,commit=5,data=ordered 0 0
/dev/mapper/vg_faith0-tmp /tmp ext3 rw,relatime,errors=continue,commit=5,data=ordered 0 0
/dev/mapper/vg_faith0-usr /usr ext3 rw,relatime,errors=continue,commit=5,data=ordered 0 0
/dev/mapper/vg_faith0-var /var ext3 rw,relatime,errors=continue,commit=5,data=ordered 0 0
/dev/hda3 /boot ext3 rw,nosuid,nodev,noexec,relatime,errors=continue,commit=5,data=ordered 0 0
/dev/hda2 /boot/grub hfs rw,nosuid,nodev,noexec,relatime,uid=0,gid=0 0 0
*********************** END /proc/mounts

*********************** BEGIN /boot/grub/device.map
(hd)	/dev/disk/by-id/ata-ST940110A_3KW5FGZX
*********************** END /boot/grub/device.map

*********************** BEGIN /boot/grub/grub.cfg
#
# DO NOT EDIT THIS FILE
#
# It is automatically generated by grub-mkconfig using templates
# from /etc/grub.d and settings from /etc/default/grub
#

### BEGIN /etc/grub.d/00_header ###
if [ -s $prefix/grubenv ]; then
  load_env
fi
set default="0"
if [ "${prev_saved_entry}" ]; then
  set saved_entry="${prev_saved_entry}"
  save_env saved_entry
  set prev_saved_entry=
  save_env prev_saved_entry
  set boot_once=true
fi

function savedefault {
  if [ -z "${boot_once}" ]; then
    saved_entry="${chosen}"
    save_env saved_entry
  fi
}

function load_video {
  insmod ieee1275_fb
}

insmod lvm
insmod part_apple
insmod ext2
set root='(vg_faith0-usr)'
search --no-floppy --fs-uuid --set 5944e2ad-2ea9-4c59-a058-9b9fe5ff4d57
if loadfont /share/grub/unicode.pf2 ; then
  set gfxmode=640x480
  load_video
  insmod gfxterm
fi
terminal_output gfxterm
set timeout=5
### END /etc/grub.d/00_header ###

### BEGIN /etc/grub.d/05_debian_theme ###
insmod lvm
insmod part_apple
insmod ext2
set root='(vg_faith0-usr)'
search --no-floppy --fs-uuid --set 5944e2ad-2ea9-4c59-a058-9b9fe5ff4d57
insmod png
if background_image /share/images/desktop-base/moreblue-orbit-grub.png ; then
  set color_normal=black/black
  set color_highlight=magenta/black
else
  set menu_color_normal=cyan/blue
  set menu_color_highlight=white/blue
fi
### END /etc/grub.d/05_debian_theme ###

### BEGIN /etc/grub.d/10_linux ###
menuentry 'Debian GNU/Linux, with Linux 2.6.32-5-powerpc' --class debian --class gnu-linux --class gnu --class os {
	insmod part_apple
	insmod ext2
	set root='(hd,apple3)'
	search --no-floppy --fs-uuid --set 217d06ca-b92e-4b8b-9519-efed0b1368f0
	echo	'Loading Linux 2.6.32-5-powerpc ...'
	linux	/vmlinux-2.6.32-5-powerpc root=/dev/mapper/vg_faith0-root ro  quiet
	echo	'Loading initial ramdisk ...'
	initrd	/initrd.img-2.6.32-5-powerpc
}
menuentry 'Debian GNU/Linux, with Linux 2.6.32-5-powerpc (recovery mode)' --class debian --class gnu-linux --class gnu --class os {
	insmod part_apple
	insmod ext2
	set root='(hd,apple3)'
	search --no-floppy --fs-uuid --set 217d06ca-b92e-4b8b-9519-efed0b1368f0
	echo	'Loading Linux 2.6.32-5-powerpc ...'
	linux	/vmlinux-2.6.32-5-powerpc root=/dev/mapper/vg_faith0-root ro single 
	echo	'Loading initial ramdisk ...'
	initrd	/initrd.img-2.6.32-5-powerpc
}
### END /etc/grub.d/10_linux ###

### BEGIN /etc/grub.d/20_linux_xen ###
### END /etc/grub.d/20_linux_xen ###

### BEGIN /etc/grub.d/30_os-prober ###
### END /etc/grub.d/30_os-prober ###

### BEGIN /etc/grub.d/40_custom ###
# This file provides an easy way to add custom menu entries.  Simply type the
# menu entries you want to add after this comment.  Be careful not to change
# the 'exec tail' line above.
### END /etc/grub.d/40_custom ###

### BEGIN /etc/grub.d/41_custom ###
if [ -f  $prefix/custom.cfg ]; then
  source $prefix/custom.cfg;
fi
### END /etc/grub.d/41_custom ###
*********************** END /boot/grub/grub.cfg

-- System Information:
Debian Release: squeeze/sid
  APT prefers testing
  APT policy: (500, 'testing')
Architecture: powerpc (ppc)

Kernel: Linux 2.6.32-5-powerpc
Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/dash

Versions of packages grub-ieee1275 depends on:
ii  bc                       1.06.95-2       The GNU bc arbitrary precision cal
ii  debconf [debconf-2.0]    1.5.35          Debian configuration management sy
ii  grub-common              1.98+20100804-4 GRand Unified Bootloader, version 
ii  powerpc-ibm-utils        1.1.0-1         utilities for maintenance of IBM P
ii  powerpc-utils            1.1.3-24        Various utilities for Linux/PowerP
ii  ucf                      3.0025+nmu1     Update Configuration File: preserv

grub-ieee1275 recommends no packages.

Versions of packages grub-ieee1275 suggests:
ii  genisoimage                   9:1.1.10-1 Creates ISO-9660 CD-ROM filesystem
ii  os-prober                     1.39       utility to detect other OSes on a 

-- debconf information excluded

[update-grub-install.diff (text/plain, attachment)]

Message #10 received at 597538@bugs.debian.org (full text, mbox, reply):

From: Colin Watson <cjwatson@debian.org>

To: Daniel Kahn Gillmor <dkg@fifthhorseman.net>, 597538@bugs.debian.org

Subject: Re: Bug#597538: grub-ieee1275: grub-install needs to pass a --prefix option to grub-mkimage

Date: Tue, 21 Sep 2010 11:51:58 +0100

[Message part 1 (text/plain, inline)]

On Mon, Sep 20, 2010 at 12:33:00PM -0400, Daniel Kahn Gillmor wrote:
> a standard powerpc mac mini is unable to boot with grub after running
> 
>   grub-install /dev/hda
> 
> (the machine's only disk uses an apple partitioning scheme, with an HFS 
> volume as (hd,apple2) mounted at /boot/grub)
[...]
> it looks like grub-install isn't passing a --prefix option to 
> grub-mkimage.
> 
> Re-running /usr/sbin/grub-install /dev/hda with the attached patch 
> applied makes it so that this ugly workaround is no longer necessary.

Thanks for your report.

> --- /usr/sbin/grub-install	2010-08-23 08:23:50.000000000 -0400
> +++ /usr/sbin/grub-install.new	2010-09-20 11:47:30.000000000 -0400
> @@ -222,7 +222,7 @@
>  modules="$modules $fs_module $partmap_module $devabstraction_module"
>  
>  # Now perform the installation.
> -"$grub_mkimage" -O ${target_cpu}-ieee1275 --directory=${pkglibdir} --output=${grubdir}/grub $modules || exit 1
> +"$grub_mkimage" -O ${target_cpu}-ieee1275 --directory=${pkglibdir} --output=${grubdir}/grub --prefix=${prefix_drive}${relative_grubdir} $modules || exit 1
>  
>  if test $update_nvram = yes; then
>      set $ofpathname dummy

I don't see how this can be sufficient, since neither prefix_drive nor
relative_grubdir is set in the current version of the script.

Does the attached patch work?  It's a partial resync with the main
version of grub-install, anticipating the next snapshot/release where
all the different grub-installs are merged into a single version.

Thanks,

-- 
Colin Watson                                       [cjwatson@debian.org]

[ieee1275_install_prefix.patch (text/x-diff, attachment)]

Message #15 received at 597538@bugs.debian.org (full text, mbox, reply):

From: Daniel Kahn Gillmor <dkg@fifthhorseman.net>

To: Colin Watson <cjwatson@debian.org>

Cc: 597538@bugs.debian.org

Subject: Re: Bug#597538: grub-ieee1275: grub-install needs to pass a --prefix option to grub-mkimage

Date: Tue, 21 Sep 2010 08:41:17 -0400

[Message part 1 (text/plain, inline)]

On 09/21/2010 06:51 AM, Colin Watson wrote:
> I don't see how this can be sufficient, since neither prefix_drive nor
> relative_grubdir is set in the current version of the script.

hrm.  i tried this after trying --prefix='(hd,apple2)/' (which worked
fine) -- so maybe grub-mkimage ended up interpreting the empty string as
"keep what was there before", or it found the right thing in some other
clever way, triggered by the explicit setting of --prefix= ?  i confess
i don't really know.

> Does the attached patch work?

I don't have access to the machine to test it today, but i'll try to
look at it later this week.

> It's a partial resync with the main
> version of grub-install, anticipating the next snapshot/release where
> all the different grub-installs are merged into a single version.

Getting all grub-installs synchronized sounds like it would be a good thing.

thanks for the quick response,

	--dkg

[signature.asc (application/pgp-signature, attachment)]

Message #20 received at 597538@bugs.debian.org (full text, mbox, reply):

From: Daniel Kahn Gillmor <dkg@fifthhorseman.net>

To: Colin Watson <cjwatson@debian.org>

Cc: 597538@bugs.debian.org

Subject: Re: Bug#597538: grub-ieee1275: grub-install needs to pass a --prefix option to grub-mkimage

Date: Sun, 03 Oct 2010 19:42:10 -0400

[Message part 1 (text/plain, inline)]

On 09/21/2010 06:51 AM, Colin Watson wrote:

> Does the attached patch work?  It's a partial resync with the main
> version of grub-install, anticipating the next snapshot/release where
> all the different grub-installs are merged into a single version.

hrm, no, this doesn't work.  i get:

>> 0 attica:/usr/sbin# grub-install /dev/hda
>> /usr/sbin/grub-install: 229: make_system_path_relative_to_its_root: not found
>> 1 attica:/usr/sbin# 

can i help debug it further somehow?  maybe other files need syncing?

(to be clear: i applied your patch directly to /usr/sbin/grub-install
with 1.98+20100804-4 installed. i did *not* rebuild the package)

	--dkg

[signature.asc (application/pgp-signature, attachment)]

Message #25 received at 597538@bugs.debian.org (full text, mbox, reply):

From: Colin Watson <cjwatson@debian.org>

To: Daniel Kahn Gillmor <dkg@fifthhorseman.net>

Cc: 597538@bugs.debian.org

Subject: Re: Bug#597538: grub-ieee1275: grub-install needs to pass a --prefix option to grub-mkimage

Date: Mon, 4 Oct 2010 18:21:35 +0100

[Message part 1 (text/plain, inline)]

On Sun, Oct 03, 2010 at 07:42:10PM -0400, Daniel Kahn Gillmor wrote:
> On 09/21/2010 06:51 AM, Colin Watson wrote:
> > Does the attached patch work?  It's a partial resync with the main
> > version of grub-install, anticipating the next snapshot/release where
> > all the different grub-installs are merged into a single version.
> 
> hrm, no, this doesn't work.  i get:
> 
> >> 0 attica:/usr/sbin# grub-install /dev/hda
> >> /usr/sbin/grub-install: 229: make_system_path_relative_to_its_root: not found
> >> 1 attica:/usr/sbin# 

Whoops, I forgot a piece.  Try this one.

-- 
Colin Watson                                       [cjwatson@debian.org]

[ieee1275_install_prefix.patch (text/x-diff, attachment)]

Message #30 received at 597538@bugs.debian.org (full text, mbox, reply):

From: Daniel Kahn Gillmor <dkg@fifthhorseman.net>

To: Colin Watson <cjwatson@debian.org>

Cc: 597538@bugs.debian.org

Subject: Re: Bug#597538: grub-ieee1275: grub-install needs to pass a --prefix option to grub-mkimage

Date: Tue, 05 Oct 2010 13:47:00 -0400

[Message part 1 (text/plain, inline)]

Hi Colin--

On 10/04/2010 01:21 PM, Colin Watson wrote:
> On Sun, Oct 03, 2010 at 07:42:10PM -0400, Daniel Kahn Gillmor wrote:
>> On 09/21/2010 06:51 AM, Colin Watson wrote:
>>> Does the attached patch work?  It's a partial resync with the main
>>> version of grub-install, anticipating the next snapshot/release where
>>> all the different grub-installs are merged into a single version.
>>
>> hrm, no, this doesn't work.  i get:
>>
>>>> 0 attica:/usr/sbin# grub-install /dev/hda
>>>> /usr/sbin/grub-install: 229: make_system_path_relative_to_its_root: not found
>>>> 1 attica:/usr/sbin# 
> 
> Whoops, I forgot a piece.  Try this one.

hrm.  trying this one lets grub-install complete properly, but it leaves
me unable to boot smoothly (i get dropped into the grub rescue prompt,
and have to use this workaround):

  set prefix=(hd,apple3)/
  insmod normal
  normal


I it looks to me like the prefix is stored very early in the created
grub image.  So i took different variants of grub-install, generated the
grub image, and then looked at the strings in /boot/grub/grub.

the three versions are:

 grub-install
  (as shipped in grub-ieee1275 version 1.98+20100804-4)

 grub-install.patched
  (shipped version + this latest patch applied)

 grub-install.dkg
  (shipped version + --prefix='(hd,apple3)/' hard-coded on the
   "$grub_mkimage" invocation line)


of these, only grub-install.dkg works for me to get all the way through
the boot process without manual intervention (obviously, my hard-coded
change won't work for every system, though).

So here's what strings reports about the prefix string in
/boot/grub/grub after running each variant:

root@bigpuff:/usr/sbin# grub-install /dev/hda
Installation finished. No error reported.
This is the contents of the device map /boot/grub/device.map.
Check if this is correct or not. If any of the lines is incorrect,
fix it and re-run the script `grub-install'.

(hd0)	/dev/disk/by-id/ata-ST940110A_3KW5HWVC
root@bigpuff:/usr/sbin# strings /boot/grub/grub | head -2
D/boot/grub
|cH8N
root@bigpuff:/usr/sbin# grub-install.dkg /dev/hda
Installation finished. No error reported.
This is the contents of the device map /boot/grub/device.map.
Check if this is correct or not. If any of the lines is incorrect,
fix it and re-run the script `grub-install.dkg'.

(hd0)	/dev/disk/by-id/ata-ST940110A_3KW5HWVC
root@bigpuff:/usr/sbin# strings /boot/grub/grub | head -2
D(hd,apple3)/
|cH8N
root@bigpuff:/usr/sbin# grub-install.patched /dev/hda
Installation finished. No error reported.
This is the contents of the device map /boot/grub/device.map.
Check if this is correct or not. If any of the lines is incorrect,
fix it and re-run the script `grub-install.patched'.

(hd0)	/dev/disk/by-id/ata-ST940110A_3KW5HWVC
root@bigpuff:/usr/sbin# strings /boot/grub/grub | head -2
|cH8N
[9|~
root@bigpuff:/usr/sbin#


I can give you shell access to a comparable machine if you'd like to
experiment.  all these tests are done on up-to-date squeeze
installations, fwiw.  I'm dkg on irc.oftc.net, if you want to chat
realtime about it.

thanks for looking into this,

	--dkg

[signature.asc (application/pgp-signature, attachment)]

Message #35 received at 597538@bugs.debian.org (full text, mbox, reply):

From: Daniel Kahn Gillmor <dkg@fifthhorseman.net>

To: 597538@bugs.debian.org

Cc: control@bugs.debian.org

Subject: grub-install still fails to pass the right --prefix to mkimage

Date: Wed, 12 Jan 2011 13:17:18 -0500

[Message part 1 (text/plain, inline)]

found 597538 1.99~20110111-1
thanks

I built 1.99~20110111-1 from the sources in unstable for powerpc, and
tried installing it on a 1st-gen iBook.  It still fails in the same way
i was seeing with #597538 earlier.

I get dropped into rescue mode, and i have to explicitly set the prefix
to be able to insmod anything.

Any attempt to insmod gets this answer:

  error: no device is set

unless i do this first:

 set prefix=(hd,apple2)

at which point, i can insmod for the usual modules.

	--dkg

[signature.asc (application/pgp-signature, attachment)]

Message #42 received at 597538@bugs.debian.org (full text, mbox, reply):

From: Daniel Kahn Gillmor <dkg@fifthhorseman.net>

To: 597538@bugs.debian.org, control@bugs.debian.org

Subject: #597538 still found in 1.99~20110112-1

Date: Sat, 15 Jan 2011 22:43:32 -0500

[Message part 1 (text/plain, inline)]

found 597538 1.99~20110112-1
thanks

I'm still seeing this problem with the latest version of grub-ieee1275
in experimental (1.99~20110112-1).

I continue to be able to work around it with:

 set prefix=(hd,apple2)
 insmod normal
 normal

	--dkg

[signature.asc (application/pgp-signature, attachment)]

Message #49 received at 597538@bugs.debian.org (full text, mbox, reply):

From: Daniel Kahn Gillmor <dkg@fifthhorseman.net>

To: 597538@bugs.debian.org

Subject: more notes on #597538

Date: Sun, 16 Jan 2011 00:50:44 -0500

[Message part 1 (text/plain, inline)]

Re: http://bugs.debian.org/597538, using grub-common and grub-ieee1275
version 1.99~20110112-1, on this old iBook:

Attached is a transcript of running "grub-install /dev/hda" with the
shell's "set -x" added.

As you can see, it calculates that it should send --prefix=/

But in fact, for this machine, it should send --prefix=(hd,apple2)/


After generating this transcript, i tried applying the two-line patch
(also attached), and it chose --prefix=(hd0,apple2)/

This also fails to boot cleanly, in spite of the fact that grub appears
to accept (hd0,...) in later stages.  The earliest stage seems to only
want to see the device listed as (hd,...) .


It appears that grub-instal chose this value because of the device.map
that it auto-generated (i tried removing the file, and the same file was
re-generated):

(hd0)	/dev/disk/by-id/ata-IBM-DBCA-203240_HP0RPLQ9068

However, if i manually update the device.map to (hd) instead of (hd0),
keep the patch applied, and then re-run "grub-install /dev/hda", it
chooses the correct --prefix.

After that, i can successfully reboot the machine without trouble.

(if i remove the patch, and re-run "grub-install /dev/hda" with the
corrected device.map, grub fails to boot again, and needs the earlier
workaround)

hth,

	--dkg

[grub-install.out (text/plain, attachment)]

[597538.patch (text/x-diff, attachment)]

[signature.asc (application/pgp-signature, attachment)]

Message #54 received at 597538@bugs.debian.org (full text, mbox, reply):

From: Daniel Kahn Gillmor <dkg@fifthhorseman.net>

To: 597538@bugs.debian.org

Subject: Re: more notes on #597538

Date: Sun, 16 Jan 2011 11:51:22 -0500

[Message part 1 (text/plain, inline)]

On 01/16/2011 12:50 AM, Daniel Kahn Gillmor wrote:
> (if i remove the patch, and re-run "grub-install /dev/hda" with the
> corrected device.map, grub fails to boot again, and needs the earlier
> workaround)


On irc.freenode.org#grub, phcoder said:

07:42 < phcoder> dkg: I see where your trouble comes from. Try
                 launching grub-install with no argument


So i unapplied my patch, and ran "grub-install" (no additional
arguments).  grub-mkimage was invoked with --prefix=/, but it now boots
properly.  Presumably some other part of the configuration was fixed by
omitting that argument.

I then replaced device.map with the original, auto-generated device.map,
and tried "grub-install" and "update-grub" again, and it still works.  yay!

So the problem seems to be that "grub-install /dev/hda" fails on this
machine, but "grub-install" works (at least with version 1.99~20110112-1).

the man page for grub-install suggests that it's acceptable to supply a
device name as an argument:

NAME
       grub-install - install GRUB to a device

SYNOPSIS
       grub-install [OPTION] [install_device]


Though i note that install_device is *not* in brackets (i.e. is not
optional) on the man page from an i386 installation of grub-common
1.98+20100804-11.

So i'm not sure what the right way to address this is.  As powerpc users
start to migrate to grub, they'll probably carry over their habits for
invoking grub-install with an argument (as i did).  Can that case be
handled correctly?

Thanks for the advice.  I'm now happily using grub on this machine, and
plan to migrate the other powerpc systems that i maintain to grub as the
bootloader.

Regards,

	--dkg

[signature.asc (application/pgp-signature, attachment)]

Send a report that this bug log contains spam.