can't handle ext2 resize_inode feature

According to http://www.gnu.org/software/parted/faq.html this might be fixed in version 1.7 of parted which is close to release it seems.

Looks like the FAQ is a bit dated. Version 1.7 is of course already released and in edgy: http://packages.ubuntu.com/edgy/admin/parted I will try to update the version on Knoppix and update this bug report on how things went.

> I will try to update the version on Knoppix and update this bug report on how
> things went.

Unfortunately, that did not go so well. Despite the claim on the home page of parted, I still get that error for even a simple layout as mine. I used the latest version in Debian testing (unstable has the same version) to update the latest Knoppix Live CD. The update succeeded but the failure remains. The error message changed a little.

  root@2[tmp]# No Implementation: This ext2 file system has a rather strange layout! Parted can't resize this (yet).
  No Implementation: This ext2 file system has a rather strange layout! Parted can't resize this (yet).
  No Implementation: This ext2 file system has a rather strange layout! Parted can't resize this (yet).
  No Implementation: This ext2 file system has a rather strange layout! Parted can't resize this (yet).
  No Implementation: This ext2 file system has a rather strange layout! Parted can't resize this (yet).
  Error: File system has an incompatible feature enabled.

I am not sure what kind of incompatible feature that could be.

The guys from parted ask to include the following information in a bug report to them. So I guess it makes sense to post the info here as well.

root@2[tmp]# parted /dev/hda print unit s print unit chs print

Disk /dev/hda: 40,0GB
Sector size (logical/physical): 512B/512B
Partition Table: msdos

Number Start End Size Type File system Flags
 1 32,3kB 516MB 516MB primary ext2
 2 4000MB 4251MB 250MB primary ext2
 4 4501MB 39,3GB 34,8GB extended
 5 4501MB 6501MB 2000MB logical ext2
 6 6501MB 7002MB 501MB logical ext2
 7 7002MB 37,0GB 30,0GB logical ext2
 8 37,0GB 37,3GB 256MB logical linux-swap
 9 37,3GB 39,3GB 2000MB logical ext2 boot

Disk /dev/hda: 78140159s
Sector size (logical/physical): 512B/512B
Partition Table: msdos

Number Start End Size Type File system Flags
 1 63s 1007999s 1007937s primary ext2
 2 7813008s 8301887s 488880s primary ext2
 4 8790768s 76678559s 67887792s extended
 5 8790831s 12697775s 3906945s logical ext2
 6 12697839s 13675535s 977697s logical ext2
 7 13675599s 72270575s 58594977s logical ext2
 8 72270639s 72771551s 500913s logical linux-swap
 9 72771615s 76678559s 3906945s logical ext2 boot

Disk /dev/hda: 77519,15,62
Sector size (logical/physical): 512B/512B
BIOS cylinder,head,sector geometry: 77520,16,63. Each cylinder is 516kB.
Partition Table: msdos

Number Start End Type File system Flags
 1 0,1,0 999,15,62 primary ext2
 2 7751,0,0 8235,15,62 primary ext2
 4 8721,0,0 76069,15,62 extended
 5 8721,1,0 12596,15,62 logical ext2
 6 12597,1,0 13566,15,62 logical ext2
 7 13567,1,0 71696,15,62 logical ext2
 8 71697,1,0 72193,15,62 logical linux-swap
 9 72194,1,0 76069,15,62 logical ext2 boot

Information: Don't forget to update /etc/fstab, if necessary.

I marked bug 14330 as a duplicate as, after searching on:
http://www-307.ibm.com/pc/support/site.wss/product.do?subcategoryind=0&familyind=99275&brandind=10&doccategoryind=0&modelind=0&doctypeind=8&validate=true&partnumberind=0&sitestyle=lenovo&template=%2Fproductpage%2Flandingpages%2FproductPageLandingPage.vm&operatingsystemind=49979&machineind=0
I find that the X24 is probably made around 2001, which would make it in the range of HPA or even pre-HPA with hidden by BIOS primary partition.

As a note, you may wish to look that you have latest BIOS at:
http://www-307.ibm.com/pc/support/site.wss/document.do?lndocid=MIGR-39665

I realize my previous link may be hard to begin finding about what hardware X24 is. So I add these page to the specifications:
http://www-307.ibm.com/pc/support/site.wss/document.do?lndocid=MIGR-58294

http://www.thinkwiki.org/wiki/Hidden_Protected_Area:
The HPA was introduced with the R/T/X 40 series of ThinkPads.

The X24 thus would not have HPA (but above page says on the bottom that the X31 also sports this technology). The R40 in the other bug report does.

According to http://www.thinkwiki.org/wiki/Predesktop_Area, IBM used the so called Predesktop Area for reinstallation purposes before HPA. http://www.thinkwiki.org/wiki/Rescue_and_Recovery: "As opposed to a Hidden Protected Area Recovery partitions are ordinary partitions, accessible through the partition table. As they are ordinary partitions they are accessible by ordinary partitioning tools."

Such a partition is not on my computer, if it were I would have wiped it. Actually, I swapped the HD when I got the X24 so there really should not be any kind of recovery partitions on this thing.

I just rebooted and checked that I have 1.32, the newest version of the BIOS.

Ok, I think I have found:
"> No Implementation: This ext2 filesystem has a rather strange
>layout! Parted can't resize this (yet).
Parted cannot handle newer ext2/3 filesystems, and this results in the message you see. upstream knows about this, but i didn't see any clear intentions to solve this, so i don't think there will be a quick fix for this.
Friendly,
Sven Luther" from http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=291234
Also as a workaround: "For ext2/3, resize2fs works fine."

This would make this bug a duplicate of bug #7993.

I which I was able to remove bug #14330 as a duplicate, before marking this one as a duplicate of bug #7983. I'll probably have to ask someone in QA Team for that.

Setting as confirmed, because it seems to be known by Debian and by upstream.

Could be a dupe of 7973 (how did I miss that one?).

But the debian stuff is more or less irrelevant because a) that bug is about resizing which I never did and never intended plus b) it is old and talks about parted pre-1.7, it is marked as fixed in 1.7 as suggested in the parted FAQ. I did use a post-1.7 parted but still got this error despite the claims in the FAQ that it should be gone.

In that sense I also feel that the issues in 7973 have been dealt with but this is something else that does look similar.

Correctings typos, bug #7993, not 7973 or 7983.

>But the debian stuff is more or less irrelevant because a) that bug is >about resizing which I never did and never intended plus
But your bug description was citing:
"This ext2 filesystem has a rather strange layout! Parted can't resize this (yet)."
I insist about can't resize this (yet) part.
Note that you thought it was your partition layout, but it is about the layout of the filesystem itself.

>b) it is old and talks about parted pre-1.7, it is marked as fixed in 1.7
Which version dpkg -l parted gives you.

According to http://packages.ubuntu.com/dapper/gnome/gparted,
Dapper version is gparted (0.1-0ubuntu9)
which seems to indicate that it would be based on parted 1.0 !?

Are you able to tell if you are using ext2 or ext3 filesystems?
I guess cat /etc/fstab is what I would want to see.

Oh, yeah, bad package.
Dapper version of parted was parted (1.6.25.1-1ubuntu2).

And you did report quite extensively your partition. But I had the feeling parted could show ext3 as ext2, but it does indeed show ext3 for me.

Still, if you could really make sure what parted --version gives.
Just to make sure.

Paul, sorry to say, but you are doing an abyssmal job here.

You mark as dupe for the wrong bug second time around. You fail to read the stuff that I already posted. All the information you needed was AFAIK in the original report.

Facts you you choose to continue to ignore

1) I never asked for resizing.
2) In fact I never asked for parted to come into the picture when installing. I already had all partitions set up. parted bringing the ubuntu install to a halt is IMNSHO another, serious flaw in the ubuntu installation process. Unfortunately, I do not know what package that should be reported against. There was no job for parted to be done, really.
3) I tried version 1.7 of parted via an updated Knoppix Live CD. The other reports were 1.6 and upstream claims all these have been dealt with in 1.7. I went to the trouble of confirming this is not true only to have you mark my report as dupe falsely twice.
4) You keep asking already answered and irrelevant question (like "give me the output of dpkg -l". Like I said, I stumbled across this when installing ubuntu, I confirmed this with Knoppix. What on earth does my current ubuntu installation have to do with this???)

Do as you please. I'm outta here.

Paul, please don't mark bugs as duplicates that are not entirely understood. I've read this bug over twice and I still don't entirely understand what's going on. If you have a better understanding than me, by all means share it, but until then remember that this is not a competition to mark the most bugs as duplicates - the objective is to make sure that as many bugs as possible get fixed, and incorrectly marking bugs as duplicates is a problem for meeting that objective. I would greatly appreciate it if you would confine your efforts to suggesting possible duplicate bugs rather than actually marking them as such. Now you appear to have alienated a bug reporter who might have been able to give more useful information once we figured out what information to ask for!

The current version of parted in Edgy has the "rather strange layout" message commented out, so it's not possible to get this error with the version of parted in Edgy. However, the same problem will cause a different error, so I'd still like to investigate.

The reason that resizing is showing up is probably the automatic partitioner checking to see if it's possible to do automatic resizing (not actually doing the resize). A full copy of /var/log/syslog and /var/log/partman might be useful. I'll create a partman-auto task on this bug because this shouldn't cause partman to fall over and die.

Unassigning Paul from this bug, unless he actually plans to write the code necessary to fix this.

Oh, hmm, it may be that parted throws this exception when you ask it to simply check the partition, not even resize it. That would be unfortunate.

Rolf: it is possible that this has been partially fixed in 1.7.1 (the version released with Edgy). One of the sets of changes was this:

        * fs/ext2/ext2.c (ext2_open): removed call to ext2_determine_itoffset;
        also moving it from this file to ext2_resize.c.
        * fs/ext2/ext2_resize.c (ext2_resize): added call to
        ext2_determine_itoffset and show a warning if not successful.

I think that's the "rather strange layout" warning. This change will arrange for that code only to be run when you actually try to resize a partition, not merely when libparted opens the partition, which should make us a lot less likely to run into this in the installer.

Would it be possible for me to get the output of:

  for x in 1 2 5 6 7 9; do sudo tune2fs -l /dev/hda$x; done

... run from e.g. an Edgy live CD (or Knoppix would probably do as well)? That should tell me what features are enabled on each filesystem, and thus which ones libparted is unable to handle.

First of all, let me say that I do appreciate the great effort that Paul was making. Quite obviously he was very enthusiastic about his task which is testified to by the amount of related information he dug up. I hope he continues with his passion to make ubuntu and FOSS in general a better and more bug-free place.

Back to the task and thanks for taking over, Collin. I am now booted into Knoppix. I believe it is the latest version.

knoppix@1[knoppix]$ cat /etc/knoppix-version
5.0.1 2006-06-01

Knoppix still has version 1.6.9 of parted. Thus I again updated this via "wget http://ftp.debian.org/debian/pool/main/p/parted/parted_1.7.1-2.1_i386.deb;wget http://ftp.debian.org/debian/pool/main/p/parted/libparted1.7-1_1.7.1-2.1_i386.deb;dpkg -r parted-bf;dpkg -i libparted*;dpkg -i parted_1.7.1-2.1_i386.deb" to version 1.7.1. Looking at the developper information for debian parted, I believe this is the same version I had used when writing this report.

OK, here the dissection. While I still have 1.6.9, I get

root@1[knoppix]# parted /dev/hda check 1
No Implementation: This ext2 filesystem has a rather strange layout! Parted
can't resize this (yet).
Information: Don't forget to update /etc/fstab, if necessary.

for partitions 1,2,3,5,6 and 7. Partition 9 gives

root@1[knoppix]# parted /dev/hda check 9
Error: Filesystem has incompatible feature enabled
Information: Don't forget to update /etc/fstab, if necessary.

After the upgrade to 1.7.1 the errors for partitions 1 to 8 are gone, the parted check passes. Partition 9 still has the same error. I will also attach the information from tune2fs that you requested. It seems I added hda3 partition in the meantime, so I include that one as well. hda2 now runs from cylinder 1001 to 1485 and hda3 from 1486 to 8721. Furthermore, I have cleared all bootable flags. The rest of the partition are as per the original report.

hda9 has the following features which are not present on the other partitioins: resize_inode dir_index. I admit to not knowing what they do and how they got there ;-)

>First of all, let me say that I do appreciate the great effort that Paul >was making. ... I hope he continues with his passion to make ubuntu >and FOSS in general a better and more bug-free place.
Thanks, it does makes me feels better to read this.
Don't worry, I'll continue, trying to avoid previous mistakes.
I would have wished that my mistakes would not have to be fixed by
Colin, AGAIN. I hope my future work will be good enough to compensate the overwork I have put on his shoulders by mistakes I have done lately while l am learning to become a better bug triager.

I'm going to reject the Debian task on this bug; while the bug undoubtedly exists in Debian too, it's not very useful to have this task here. If we can come up with a patch set it can more easily be forwarded to Debian in other ways.

The main "offending" feature is resize_inode. I spent a while the other day trying to beat support for this feature into libparted, but it's not easy work. I may get there, or I may not ...

One thought I had was that perhaps partman could detect the presence of unsupported features and degrade to using resize2fs instead of libparted. This doesn't offer a progress bar in a way that we can sanely handle in partman, which isn't great, but it would be better than not being able to resize at all.

Also, while dir_index doesn't cause libparted to fall over completely like resize_inode, it's not supported by the resize code. We probably need to make libparted at least tolerate resize_inode outside the resize code even if it can't deal with resizing it.

I attempted to do a separate resize2fs/e2fsck, but parted refuses to move or resize a partition that has this flag in the fs. So falling back to resize2fs probably isn't likely to help.

I ran into the same problem when trying to install Edgy together on my Feisty installation. Resizing the ext3 partition (created by Feisty) during Edgy install failed with some non-verbose error like "unable to resize partition".
I've opened another terminal (Alt+F2) and tried parted. It gave me the following error message: "Error: File system has an incompatible feature enabled".

According to "parted --version", Edgy's using v1.7.1.

I've worked around this for Feisty as follows:

partman-partitioning (47ubuntu2) feisty; urgency=low

  * Use tune2fs/resize2fs to handle ext2/ext3 resizing, since libparted
    can't deal with the resize_inode feature (LP: #59620).

 -- Colin Watson <email address hidden> Mon, 19 Mar 2007 11:41:11 +0000

I understand that eventually the amount of filesystem code duplication in libparted will be decreased, making it easier to fix this bug properly there.

partman-auto (62ubuntu7) feisty; urgency=low

  * Make use of changes in partman-partitioning 47ubuntu2 to use
    tune2fs/resize2fs to handle ext2/ext3 resizing (LP: #59620).

 -- Colin Watson <email address hidden> Mon, 19 Mar 2007 12:04:46 +0000

I am happy to say that this report I made just saved my a**. On the weekend I somehow managed to completely kill my partition table (still unsure as to what did it). Thanks to the information I was able to restore the drive quickly.

Thanks, Launchpad!

"Thanks to the information about my HD layout that I recorded in this report I was able to restore it quickly" should be more understandable.

The thing is I believe resize_inode is out of the way as I already "disabled" it with tune2fs.

The only 2 attributes left are ext_attr and sparse_super, both of them I cannot remove with tune2fs.
dumpe2fs /dev/sda4 | grep feature
dumpe2fs 1.40.8 (13-Mar-2008)
Filesystem features: ext_attr sparse_super

The problem is the same for the resize options of parted but this I simply overtake the problem by resizing the FS and then deleting/recreating the partition.

There is also the problem of not being able to set the resize_inode feature again without formatting the partition (or at least I haven't found how)

(using hardy btw)

Looks like partman-basicfilesystems needs to work around this too (for ext2 with unusual features); Carl Karsten is going to attach some logs. (This is quite rare FWIW.)

/tmp/misc # tune2fs -l /dev/sdc1
tune2fs 1.41.0 (10-Jul-2008)
Filesystem volume name: <none>
Last mounted on: <not available>
Filesystem UUID: c8d7d2a8-ce89-4fdb-b000-225bb3baeaa2
Filesystem magic number: 0xEF53
Filesystem revision #: 1 (dynamic)
Filesystem features: ext_attr resize_inode dir_index filetype sparse_super large_file
Filesystem flags: signed_directory_hash
Default mount options: (none)
Filesystem state: clean
Errors behavior: Continue
Filesystem OS type: Linux
Inode count: 655776
Block count: 2622603
Reserved block count: 131130
Free blocks: 2141345
Free inodes: 655756
First block: 0
Block size: 4096
Fragment size: 4096
Reserved GDT blocks: 640
Blocks per group: 32768
Fragments per group: 32768
Inodes per group: 8096
Inode blocks per group: 506
Filesystem created: Thu Oct 2 14:54:21 2008
Last mount time: Thu Oct 2 14:56:12 2008
Last write time: Thu Oct 2 18:24:49 2008
Mount count: 1
Maximum mount count: 36
Last checked: Thu Oct 2 14:54:21 2008
Check interval: 15552000 (6 months)
Next check after: Tue Mar 31 14:54:21 2009
Reserved blocks uid: 0 (user root)
Reserved blocks gid: 0 (group root)
First inode: 11
Inode size: 256
Required extra isize: 28
Desired extra isize: 28
Default directory hash: tea
Directory Hash Seed: 51325a80-f2a4-4b86-b12d-e7895bad659c

I have exactly the same problem as in bug #37137 which is reported as a duplicate of this bug, with the same messages "file system has an incompatible feature enabled" and then "Ext2 partition has unrecoverable system error" using the installer of both Ubuntu and Kubuntu 9.04 JAUNTY.

The ext2 partition is my boot partition. If I ignore the messages and go on for the installation, it seems that everything works well. But for a newcomer to linux this may be dissuasive and crippling before he could have a chance to see anything of Linux and Ubuntu!

Why are some of the tasks for this fixed, and some are still open? It seems that this issue was fixed in Ubuntu, wasn't it? The ability for parted to resize is depreciated and going to be removed so the installer should not be using it.