Bug 151947

Summary: Installation: The partitioner is broken in SL 10.1
Product: [openSUSE] SUSE Linux 10.1 Reporter: Michael Gross <mgross>
Component: InstallationAssignee: Thomas Fehr <fehr>
Status: RESOLVED FIXED QA Contact: Klaus Kämpf <kkaempf>
Severity: Blocker    
Priority: P5 - None CC: aj, brian.b, Daniel.Ottey, fmfischer, jdouglas, kernel01, kukuk, marcel, michael.raskey, suse-beta, tiwai, trenn, ug, varkoly
Version: Beta 5   
Target Milestone: ---   
Hardware: i586   
OS: Other   
Whiteboard:
Found By: Other Services Priority:
Business Priority: Blocker: ---
Marketing QA Status: --- IT Deployment: ---
Attachments: yast logs for the second installation
Very huge as I turned on debugging info and the installation was a bit progressed...

Description Michael Gross 2006-02-17 20:45:49 UTC
This report is a summary of the following reports:
bug #181783, bug #151808, bug #151783, bug #151884 (and their duplicates)
which seem to be different results from the same problem with the partitioner in 10.1 Beta4.

----------------------------------------------
So far, the following problems were reported: 
The partitioner...

(1) mixes up filesystem types - for example one chooses ext2 and
    the partition is getting formatted with reiserfs (bug #151884) or it even
    tries to format an NTFS fs (bug #151808) - which would fail naturally.

These are the results for this issue so far: This whole behaviour does not occurr if the disk is unpartitioned. I also figured that the wrongly chosen fs does not depend on your selection within the packager, it seems to depend on the already existing partitioning. I for example also chose ext2 for /boot (like Thomas in bug #151808) but it was then formatted as reiserfs.

(2) creates double or totally obscure entries within the fstab of the system
    (like reported in bug #151884).
    This is tremendously bad as the initrd for the later system might lack
    the correct fs-module for the rootfs which results in a kernel panic
    within the second stage of the installation (because the rootfs cannot
    be mounted due to lacking filesystem support). Unversed users are probably
    totally lost here.

(3) makes inproper proposals for a standard partitioning. This is the least
    problem of all but compromises the usability of the system. As a result,
    no valid default boot-setup can be proposed (no /boot partition is created
    by default - only swap is getting adapted). Other reports do relate to this
    problem as SCSI disks take precedence over IDE-disks even if they are
    attached via USB or firewire. Moreover, on systems with multiple disks only
    the last disk seems to be honored at all.
----------------------------------------------

That much for the summary. Right now I am installing a system that had hdd without a partition table and I nilled out the first 10 Megs of the system. The result so far was a perfectly smooth installation, the fstab did neither contain double nor wrong entries and filesystems were not formatted with the wrong format.

In bug#151808 comment #6 Thomas reports that creating the partition table manually with fdisk and let YaST do the rest, works. So far I could not assert this and it would somehow narrow down the problem to an existing fstab, as I suspsected earlier. Hence the problem is most likely located within the part the partitioner is parsing the existing fstab.

In addition to the referenced bugs I will add the following information from my approach:

+ Logfiles and resulting files of the first (at the moment appearing flawless installation)

as well as logfiles and results from an installation performed after that changing some of the partitioning that was created before.

For a better handling, I will mark the mentioned bugs as duplicates of this one.
Comment 1 Michael Gross 2006-02-17 20:51:41 UTC
Revisions for the first comment (sorry but there is no preview available ):

> This report is a summary of the following reports:
  bug #181783 ...

This means of course bug #151873

Comment 2 Michael Gross 2006-02-17 21:01:26 UTC
*** Bug 150744 has been marked as a duplicate of this bug. ***
Comment 3 Michael Gross 2006-02-17 21:11:05 UTC
*** Bug 151808 has been marked as a duplicate of this bug. ***
Comment 4 Michael Gross 2006-02-17 21:12:38 UTC
*** Bug 151783 has been marked as a duplicate of this bug. ***
Comment 5 Michael Gross 2006-02-17 21:13:18 UTC
*** Bug 151884 has been marked as a duplicate of this bug. ***
Comment 6 Michael Gross 2006-02-17 21:57:20 UTC
We decided that 150744 should go independently.
Comment 7 Michael Gross 2006-02-17 22:05:43 UTC
This is the result of the first (successful) installation. I did not attach the logs because this time it worked:
Comment 8 Michael Gross 2006-02-17 22:07:00 UTC
fstab:

/dev/hda3            /                    ext3       acl,user_xattr        1 1
/dev/hda2            /boot                ext2       acl,user_xattr        1 2
/dev/hda1            swap                 swap       defaults              0 0
proc                 /proc                proc       defaults              0 0
sysfs                /sys                 sysfs      noauto                0 0
usbfs                /proc/bus/usb        usbfs      noauto                0 0
devpts               /dev/pts             devpts     mode=0620,gid=5       0 0

fdisk -l:

Disk /dev/hda: 20.0 GB, 20020396032 bytes
255 heads, 63 sectors/track, 2434 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes

   Device Boot      Start         End      Blocks   Id  System
/dev/hda1               1          66      530113+  82  Linux swap / Solaris
/dev/hda2   *          67          75       72292+  83  Linux
/dev/hda3              76        2434    18948667+  83  Linux

everything looks fine.
Comment 9 Michael Gross 2006-02-17 23:10:42 UTC
Interesting...

This relates to issue (3):

after I started a new installation based on the partitioning in comment #8, YaST made another proposal for the partitioning:

Basically deleting hda3 (which is was primary partition) and use the rest of the space for its default partition scheme.

Conclusion: This problem (3) obviously does only occurr with an existing extended partition (which will usually be the case).

I accepted the proposal without a change and after that which resulted in a working system.

For this installation `fdisk -l', `fstab' and  YaST-logs follow.
I'm no longer sure if YaST really proposed to create no extra partiotion for boot - but this can be extracted from the logfiles, they are just a bit to cryptic for me right now.
Comment 10 Michael Gross 2006-02-17 23:19:51 UTC
Created attachment 69152 [details]
yast logs for the second installation
Comment 11 Michael Gross 2006-02-17 23:21:10 UTC
The partitioning and fstab after the second installation:

Disk /dev/hda: 20.0 GB, 20020396032 bytes
255 heads, 63 sectors/track, 2434 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes

   Device Boot      Start         End      Blocks   Id  System
/dev/hda1               1          66      530113+  82  Linux swap / Solaris
/dev/hda2   *          67          75       72292+  83  Linux
/dev/hda3              76        2434    18948667+   f  W95 Ext'd (LBA)
/dev/hda5              76        1490    11365956   83  Linux
/dev/hda6            1491        2434     7582648+  83  Linux

---

/dev/hda6            /                    reiserfs   acl,user_xattr        1 1
/dev/hda5            /home                reiserfs   acl,user_xattr        1 2
/dev/hda1            swap                 swap       defaults              0 0
proc                 /proc                proc       defaults              0 0
sysfs                /sys                 sysfs      noauto                0 0
usbfs                /proc/bus/usb        usbfs      noauto                0 0
devpts               /dev/pts             devpts     mode=0620,gid=5       0 0
/dev/hda2            /data1               auto       noauto,user           0 0


Still - looks good.
It has to be checked if the proposal for /boot is correct, though.
Comment 12 Michael Gross 2006-02-17 23:50:29 UTC
OK, now... making an installation based on this partitioning, YaST fails again to make a good proposal exactly as described in (3). Note: hda2 was /boot in the installation before and stayed as primary partition.

So this is definetly about the existance of an partitioning scheme with an extended partition.

I used the expert partitioning tool and changed exactly the following:

+ deleted logial partitions hda5 and hda6
+ created a logical partition with the whole space in hda3 to be mounted as /    
  with ext3.
+ All partitions were marked for formatting

So the new setup should exactly look like this:

  /dev/hda1   swap
  /dev/hda2   /boot (ext2)
  /dev/hda5   /     (ext3)

Expectably, it formatted hda5 with the wrong fs again (reiser this time). I will attach the complete logs as well as the fstab and partition table after that.
Comment 13 Michael Gross 2006-02-17 23:57:24 UTC
This time, there was no doubled entry for / in the fstab...
I'm attaching the logfiles later. Reassigning to Thomas now.
Comment 14 Michael Gross 2006-02-20 09:57:17 UTC
*** Bug 151758 has been marked as a duplicate of this bug. ***
Comment 15 Michael Gross 2006-02-20 15:17:53 UTC
*** Bug 152257 has been marked as a duplicate of this bug. ***
Comment 16 Michael Gross 2006-02-21 18:57:06 UTC
*** Bug 152564 has been marked as a duplicate of this bug. ***
Comment 17 Michael Gross 2006-02-21 19:20:43 UTC
*** Bug 152568 has been marked as a duplicate of this bug. ***
Comment 18 Murlin Wenzel 2006-02-22 16:39:32 UTC
Michael

I've managed to get a system here to show the same behavior.

1.  Previous OS - SLES 10 p4

Partitioning

/dev/hda1 1GB as swap
/dev/hda2 39GB as / (reiser)


The first attempt at Beta4, I got the following proposal

Delete existing partitions.

create:
/dev/hda1 39GB as / (reiser)
/dev/hda2 1GB as swap

I accepted this proposal and ended up with 2 swap partitions.  The / filesystem never really mounted (/dev/hda2).  I think this is where the error about disk space being exhausted comes from.  The install completes but never really copies anything to the partition.  Grub fails install.

I deleted all partitions manually before starting the install again.  I got the same proposal as above with the same results.

2.  I used the 'custom' partitioning section of Yast to delete all partitions and manually created the following entries.

/dev/hda1 1GB as swap
/dev/hda2 39GB as / (reiser)

This install completed with no errors and the system is up and running now.

If I can get another system to show the same behavior, are there any particular logs etc... that would be most useful to you?
Comment 19 Michael Gross 2006-02-23 11:50:04 UTC
Murlin: I suppose your / was formatted with swap or some other (wrong) fs and therefore failed to mount. You are lucky because not even the manual setip (expert partitioner) does work correctly (see this comment). You might attach your logfiles for the failed installation if you still have them. If not, it's not as wild as we have plenty of material by now.

In Beta5:
This time I used the manual partitioner to delete all existings partitions and create three new ones (/boot as ext3,  512M swap and / as ext3) and the only error this time was that /boot was also created as swap space. Consistent however, as the fstab was created with 

/dev/hda1 /boot   swap (...)

I verified that the partition was actually formatted as swap space. Later at the bootloader configuration this would of course fail completely and result in a non-bootable system.

For this installation, I will attach the yast-logs.
Comment 20 Michael Gross 2006-02-23 12:15:00 UTC
Created attachment 69956 [details]
Very huge as I turned on debugging info and the installation was a bit progressed...
Comment 21 Michael Gross 2006-02-23 13:15:05 UTC
Well now this is strange. I observed the same problem pictured in comment #19 happening even if I deleted the partitions with fdisk before proceeding. The first time I assumed it always works with non-partitioned disks (non-partitioned was ment with an empty partition table here). This is not the case (anymore).
Comment 22 Uwe Gansert 2006-02-23 16:49:30 UTC
I have just fixed the problem that a partition gets formated with a wrong (old) filesystem in libstorage 2.13.30
I needed to revert another fix from Thomas for that but at least we can do partitioning now that differs from the partitiong that is already on the harddisk.

I think you had hit that bug too and it should work with the next beta.
Please test.
Comment 23 Michael Gross 2006-02-23 16:58:44 UTC
I will keep testing this. Thanks so far.
Comment 27 Michael Gross 2006-02-24 11:25:19 UTC
*** Bug 153127 has been marked as a duplicate of this bug. ***
Comment 28 Michael Gross 2006-02-24 11:27:31 UTC
153127: Wrongly crated fstab entries, this time (AFAIK yet) with wrong attributes.
Comment 31 Michael Gross 2006-02-24 14:20:12 UTC
*** Bug 153352 has been marked as a duplicate of this bug. ***
Comment 32 Frank-Michael Fischer 2006-02-25 03:47:15 UTC
Disk /dev/hda: 60.0 GB, 60011642880 bytes
here is my partition table:

255 heads, 63 sectors/track, 7296 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes

   Device Boot      Start         End      Blocks   Id  System
/dev/hda1   *           1        3648    29302528+   7  HPFS/NTFS
/dev/hda2            3649        7296    29302560    f  W95 Ext'd (LBA)
/dev/hda5            3649        5062    11357923+   7  HPFS/NTFS
/dev/hda6            5063        5193     1052226   82  Linux swap / Solaris
/dev/hda7            5194        6034     6755301   83  Linux
/dev/hda8            6035        7296    10136983+  83  Linux

Both hda7 and hda8 carry an ext3 file system. The installation routine suggests reformatting hda7 with reiserfs and mounting it at "/" (I could live with that) and suggests to reformat hda8 with reiserfs, too, and mounting it at "/home". Moreover when trying to mount hda8 by hand I get: "unknown file system type ext3". So clearly the file system type on this partition was not recognized properly.

Reformatting a "working" partition beside the root partition is useless and dangerous. 
Comment 33 Michael Gross 2006-02-27 09:21:05 UTC
*** Bug 153186 has been marked as a duplicate of this bug. ***
Comment 37 Michael Gross 2006-02-28 12:27:36 UTC
*** Bug 154006 has been marked as a duplicate of this bug. ***
Comment 38 Michael Gross 2006-02-28 12:36:11 UTC
*** Bug 153951 has been marked as a duplicate of this bug. ***
Comment 39 Michael Gross 2006-02-28 14:39:36 UTC
*** Bug 152264 has been marked as a duplicate of this bug. ***
Comment 40 Michael Gross 2006-03-01 12:11:57 UTC
*** Bug 154259 has been marked as a duplicate of this bug. ***
Comment 42 Andreas Jaeger 2006-03-02 11:13:46 UTC
MIchael, can your retest this with Beta6 and check which of these are fixed?
Comment 43 Takashi Iwai 2006-03-02 15:22:44 UTC
The bug #153352 still exists on beta6.

The partitioner doesn't use the whole disk space even if "use entire disk" is selected.  20GB remain unused.

Comment 44 Peter Varkoly 2006-03-03 10:25:42 UTC
STATE Beta6: The problem described in #153127 exist still. The /etc/fstab does not contains the options acl,user_xattr but defults for /. (Default settings, nothig changed in the partitioner)
Comment 45 Michael Gross 2006-03-07 22:22:25 UTC
Andreas: I've installed Beta6 however this time I went with the proposal, which worked fine (however I'm not sure if there was a partitioning before). I will try changing the partitioning, based on the existing one and do some more specific testing, though. I'll provide a comment as soon as it's done.
Comment 46 Uwe Gansert 2006-03-08 09:02:44 UTC
*** Bug 155904 has been marked as a duplicate of this bug. ***
Comment 47 Michael Gross 2006-03-08 14:59:59 UTC
*** Bug 155114 has been marked as a duplicate of this bug. ***
Comment 48 Takashi Iwai 2006-03-09 12:11:59 UTC
The bug #153352 (doesn't use entire disk) still remains on beta7.
Comment 49 Michael Gross 2006-03-09 15:30:14 UTC
Good news: I'm installing Beta7, deleted the old partitioning scheme with the expert partitioner, created 3 primary partitions (one for /boot [ext3], one swap and one for / [reiserfs]) and this resulted in a correct installation, correct fstab and the filesystems were formatted correctly.
Comment 50 Thomas Fehr 2006-03-15 15:40:57 UTC
Problems should be gone in beta#9.
Comment 51 Michael Gross 2006-03-15 19:25:19 UTC
Thanks Thomas.
Comment 52 Thomas Fehr 2006-03-16 10:27:15 UTC
*** Bug 153941 has been marked as a duplicate of this bug. ***
Comment 53 Peter Varkoly 2006-03-19 22:03:53 UTC
STATE Beta8: The problem described in #153127 exist still. The resulting /etc/fstab does not contains the options acl,user_xattr for /. (Default settings, nothig changed in the partitioner)
Comment 54 Thomas Fehr 2006-03-21 10:34:13 UTC
You are right the issue with wrong default fstab options for xfs should be
fixed with beta#9.