Breezy install fails at partitoning on Dell Inspiron 4150

[Christopher Bayliss]

When installing breezy the installer gets as far as the partioning stage where
it fails. Instead of the normal partition mangement window appearing the
following small window appears.

     [!!] Partition Disks

          ??? ???

     <Go Back> <Continue>

Using either go back or continue results in "E: Unimplemented function"
appearing before the same window reappears. The only way out of this I have
found is a hard reset. The installer appears to be functioning properly until
this happens.

Notes:

This problem is in both colony CD 2 and the daily install CD ( 6th August )

This laptop had previously been running Hoary and had a broken Breezy ( Hoary
upgrade ) install on it at the time of the install attempts.

This problem has appeared on the development mailing list with no responses
http://www.ubuntuforums.org/showthread.php?t=47190&highlight=breezy+installer

I have not tried bypassing this stage from an earlier point in the install.

[D. Heck (dheck)]

I can confirm this bug. The same happens on my Samsung P35 notebook with
the Breezy Colony 3 installer.

[Jay Camp (jayc)]

This is occuring on my Dell Inspiron 8200 as well on Colony 3 (along with
netboot from current). Attaching syslog and partman logs.

[Jay Camp (jayc)]

Created an attachment (id=3323)
partman log

[Jay Camp (jayc)]

Created an attachment (id=3324)
syslog

[grendelkhan (scottricketts)]

I can also confirm this bug - I get this every time I try to do a Breezy install
(Asus K8V based system - no SATA) - I did not have this problem with Hoary.

My partition types are all XFS with the exception of /boot

[Jay Camp (jayc)]

Using the netboot image from 9/5/2005, the partitioning was successful on an
Inspiron 8200. Anybody confirm with Colony 4?

[Colin Watson (cjwatson)]

*** Bug 20200 has been marked as a duplicate of this bug. ***

[Colin Watson (cjwatson)]

This change may help:

partman-auto (41ubuntu3) breezy; urgency=low

  * Don't attempt to automatically resize fat16 partitions, since they
    appear to cause errors when finding the resize range and probably aren't
    big enough to worry about anyway; should work around Ubuntu #13250.

 -- Colin Watson <email address hidden> Wed, 7 Sep 2005 15:00:22 +0100

I'm not convinced this addresses the real problem (something seems to be wrong
with exception handling in partman), but it should at least be a workaround.
When the Breezy preview is released, please test with that to see if the
partitioner now works.

[Colin Watson (cjwatson)]

Make that partman-auto 41ubuntu4; I collided with an earlier upload I'd
forgotten about.

[grendelkhan (scottricketts)]

*** Bug 20443 has been marked as a duplicate of this bug. ***

[Ben Andersen (ben72)]

I can also confirm this on a standard Soltek P3 motherboard using a 20Gb Maxtor
disk with unknown content. It was probably FAT32 partitions on it before.

To resolve it I booted with the Hoary install CD and created a partition as
suggested and rebooted. Then I could get into the partitioner in Breezy and the
installation went fine!

[Colin Watson (cjwatson)]

*** Bug 19418 has been marked as a duplicate of this bug. ***

[Colin Watson (cjwatson)]

Bug #22378 indicates that this is still happening, with fat32 partitions this
time (to which I can't apply the fat16 workaround, because I want to be able to
auto-resize fat32 partitions). Sigh.

[Colin Watson (cjwatson)]

*** Bug 22378 has been marked as a duplicate of this bug. ***

[Guillaume Desmottes (cassidy)]

Bug seems fixed for me (bug 20200) with Colony CD 5.

[Quim Gil (qgil-interactors)]

Created an attachment (id=4174)
partman log installing Breezy 27/sep/05

I still get this bug with the iso downloaded this morning.

/var/log/partman attached

[Colin Watson (cjwatson)]

(In reply to comment #16)
> I still get this bug with the iso downloaded this morning.
>
> /var/log/partman attached

Your log is very interesting; unlike the other reports of this bug, it's with
NTFS. The code path for NTFS is a bit different, and doesn't attempt to open the
partition at all. That suggests that, contrary to my earlier guess, we may be
dealing with a geometry problem, because there's really not much else in the
segment of code where parted_server seems to have crashed in your case other
than geometry calculations.

Would you be willing to test a custom debug build, if I put one together for you
tomorrow morning?

[Colin Watson (cjwatson)]

Quim, could you please have a look at my comment #17 on this bug and let me know
whether you'll be able to test a debug build? I hadn't realised you weren't CCed.

[Quim Gil (qgil-interactors)]

(In reply to comment #17)
> Would you be willing to test a custom debug build, if I put one together for you
> tomorrow morning?

I'm happy to collaborate and help this bug being solved. You only need to
consider that I need VERY clear and verbose instructions in order to be able to
follow the process. You don't know how much I've sweat to succeed taking out
/var/log/partman out of that PC in order to attach it here. :)

[Colin Watson (cjwatson)]

Hi all,

If you are suffering from this bug, please follow these instructions to get me
debugging information:

1) If you can spare the machine in question for a while, please try not to "fix"
the partitioning such that this bug goes away. I consider this the most serious
installer bug currently open, and I'd like to make sure that people can continue
to reproduce this bug until it's been properly fixed.

2) Download
http://people.ubuntu.com/~cjwatson/tmp/13250/breezy-install-i386-debug.iso; this
is not a real installation CD, because I've removed all the .debs from it in
order to make it reasonably quick for people to download (it's 35MB). However,
it does contain the installer, with a modified partman-base package that prints
extra debugging information to /var/log/partman during the code paths that seem
to be crashing.

3) Burn this to CD, and boot from it.

4) Run through the installer in the usual way until you reach the "Partition
disks" stage. At this point, either you will see the main menu (in which case
the bug isn't reproducible for you) or it will fail as before.

5) If it fails, please extract /var/log/partman as before and attach it to this
bug. You can do this by, for example, pressing Alt-F2, entering 'anna-install
openssh-client-udeb' to install scp in the installer's ramdisk, and then using
scp to copy that file to another machine.

Thanks in advance.

[Quim Gil (qgil-interactors)]

Created an attachment (id=4204)
partman log with debug iso v1

Done and still fails. :(

Log attached.

[Colin Watson (cjwatson)]

(In reply to comment #21)
> Created an attachment (id=4204) [edit]
> partman log with debug iso v1
>
> Done and still fails. :(
>
> Log attached.

Thanks for that; this was invaluable information! After some analysis, it
appears that parted doesn't like the geometry of one of your partitions, and
therefore fails in a place where partman is not expecting it to fail. It's
possible that a newer version of parted might fix this, but it's too late to fix
that now.

I'll have a partman-base fix for you shortly; tomorrow's daily build will
include it.

[Colin Watson (cjwatson)]

*** Bug 21456 has been marked as a duplicate of this bug. ***

[Colin Watson (cjwatson)]

*** Bug 22229 has been marked as a duplicate of this bug. ***

[Colin Watson (cjwatson)]

*** Bug 21437 has been marked as a duplicate of this bug. ***

[Colin Watson (cjwatson)]

This should fix the problem as of tomorrow's daily CD builds. If it doesn't,
please follow the directions in comment #20; if you've already done that, let me
know and I'll investigate further.

partman-base (70ubuntu2) breezy; urgency=low

  * ped_disk_get_max_partition_geometry returns NULL if the partition's
    current geometry is invalid in some way. Deal with this in
    GET_RESIZE_RANGE and GET_VIRTUAL_RESIZE_RANGE by just claiming that the
    partition cannot be enlarged (closes: Ubuntu #13250).

 -- Colin Watson <email address hidden> Fri, 30 Sep 2005 15:22:17 +0100

[Dave Rosky (d-rosky)]

This sounds very encouraging; however, where may one find
the daily CD builds? I don't see them at ftp.ubuntu.com,
nor do I see them on the download page (I only see the Sept. 8
CD's there).

Thanks in advance.
Dave

[Dave Rosky (d-rosky)]

(In reply to comment #27)
> This sounds very encouraging; however, where may one find
> the daily CD builds? I don't see them at ftp.ubuntu.com,
> nor do I see them on the download page (I only see the Sept. 8
> CD's there).
>

Nevermind, I believe I found it by googling around:

http://cdimage.ubuntulinux.org/daily/current/

Dave

[Colin Watson (cjwatson)]

(In reply to comment #28)
> Nevermind, I believe I found it by googling around:
>
> http://cdimage.ubuntulinux.org/daily/current/

That's right, although the canonical (ahem) hostname is cdimage.ubuntu.com
rather than cdimage.ubuntulinux.org.

[Dave Rosky (d-rosky)]

(In reply to comment #29)
>
> That's right, although the canonical (ahem) hostname is cdimage.ubuntu.com
> rather than cdimage.ubuntulinux.org.

Very good, thanks. I will download it tomorrow and let you know
if it works (unless you will be making another test ISO, then I will
download that instead, to save bandwidth).

BTW, if parted fails when run by partman during the install, why does
it not fail when I run it manually after booting from the live
CD? Just curious...

Dave

[Colin Watson (cjwatson)]

partman uses libparted, not the parted command-line tool; it uses libparted
calls that parted itself doesn't make, because it's presenting different
information.

[Dave Rosky (d-rosky)]

(In reply to comment #31)
> partman uses libparted, not the parted command-line tool; it uses libparted
> calls that parted itself doesn't make, because it's presenting different
> information.

I see, thanks. In my experience there have been times when the parted
command-line tool has claimed that a partition is not resizable even
though I would have expected it should be. Perhaps this was because
of a similar invalid geometry problem.

At the risk of getting off-topic for this bug, is there an easy
way of fixing such an invalid geometry so that the partation *can*
be resized by parted? (I assume it's a parameter in the partition table
that's invalid.)

Dave

[Colin Watson (cjwatson)]

(In reply to comment #32)
> At the risk of getting off-topic for this bug, is there an easy
> way of fixing such an invalid geometry so that the partation *can*
> be resized by parted? (I assume it's a parameter in the partition table
> that's invalid.)

I'm afraid I'm not sure; I haven't actually been able to set up a situation for
myself in which I could reproduce the bug, although the log files you and Quim
provided were enough to let me narrow it down to a small enough segment of code
to make me pretty certain that it was a geometry problem.

My best guess at the moment is that the problem occurs when a partition ends in
the middle of a cylinder on the disk rather than on a cylinder boundary. For
now, though, I'm content to arrange for the partitioner not to blow up and worry
about the rest after the Breezy release.

[Dave Rosky (d-rosky)]

(In reply to comment #33)
> For
> now, though, I'm content to arrange for the partitioner not to blow up and worry
> about the rest after the Breezy release.

I agree 100%. Don't hesitate to let me know at some time in the
future if I can provide any information from my system that can
help improve parted or partman; I don't plan on changing the
partitioning of this particular disk in the near future.

FWIW, these partitions were created several years ago by Partition
Magic. In the course of trying to debug this problem, I ran Partition
Magic, and it complained about the CHS values not matching the LBA values
for many of the partitions. It offered to correct this, which I
allowed it to do, but that didn't seem to help partman.

Perhaps Partition Magic allows one to create partitions that end in the
middle of a cylinder when more properly it should not.

Dave

[Quim Gil (qgil-interactors)]

Created an attachment (id=4260)
partman log current build 1/oct/05

Today at midday I have downloaded the current build and I get the same error
message as before, with no way to go back/forth as before.

[Dave Rosky (d-rosky)]

Colin,

The Oct. 1 build seems to work on my system. I am presented with
the normal partitioning menu, although I haven't yet had time
to carry the installation any farther. Because it worked,
I have not attached a log file; I can do so if it might be of any
additional help.

Since parted was detecting geometry irregularities, is it safe
to use partman to delete and create partitions, or should I only
format and assign existing partitions?

Dave

[Colin Watson (cjwatson)]

(In reply to comment #35)
> Created an attachment (id=4260) [edit]
> partman log current build 1/oct/05
>
> Today at midday I have downloaded the current build and I get the same error
> message as before, with no way to go back/forth as before.

Please check the version of the partman-base package in
/cdrom/pool/main/p/partman-base, to make sure it's really 70ubuntu2. If so, I'll
put together a new test image for you on Monday.

[Quim Gil (qgil-interactors)]

> Please check the version of the partman-base package in
> /cdrom/pool/main/p/partman-base, to make sure it's really 70ubuntu2.

Hum. 70ubuntu1 here. I downloaded it too early. :)

Downloading current build right now... Will report later.

[Colin Watson (cjwatson)]

The 20051001 and 20051002 builds hung while initialising partman when installing
from a DVD-RW (at least) for a reason unrelated to this bug. If you encounter
this, try 20051003 when it appears, and make sure that it has partman-base
70ubuntu3 in /cdrom/pool/main/p/partman-base/.

partman-base (70ubuntu3) breezy; urgency=low

  * Backport from trunk:
    - Stop USES_EXTENDED command from segfaulting if called on a device
      without a partition table (e.g. a DVD-RW).

 -- Colin Watson <email address hidden> Sun, 2 Oct 2005 20:38:56 +0100

[Quim Gil (qgil-interactors)]

> Downloading current build right now... Will report later.

This time it worked fine!

Partitions were detected and I could reformat manually one of them in order to
get space for Breezy.

Thank you.