GNU bug report logs - #31111
Partition table wrongly identified as "atari"

Previous Next

To add a comment to this bug, you must first unarchive it, by sending
a message to control AT debbugs.gnu.org, with unarchive 31111 in the body.
You can then email your comments to 31111 AT debbugs.gnu.org in the normal way.

Toggle the display of automated, internal messages from the tracker.

Message #5 received at submit <at> debbugs.gnu.org (full text, mbox):

From: bouke_1 <bouke_1 <at> zonnet.nl>
To: bug-parted <at> gnu.org
Subject: Partition table wrongly identified as "atari"
Date: Mon, 09 Apr 2018 15:39:52 +0200

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

After restoring an image using Clonezilla (which in turn uses sfdisk to 
configure the partitions), parted identifies the partition table of my 
harddisk as "Atari" and sees only one partition, while in fact it is a 
valid msdos partition table with three NTFS partitions (boot, main and 
restore). Other tools (e.g., fdisk) correctly identify an "msdos" 
partition list. The disk contains a Windows 10 environment and that 
boots and works correctly. So I have reason to believe that parted is 
wrong (or the rest of the world is crazy, which is not impossible).

I ran parted from the Linux command line from a Clonezilla Live disk (on 
a USB drive) which is based on debian zesty. The version of parted used 
there is (of course) the latest stable, 3.2. I also tried gparted (which 
uses libparted) in its latest stable live distribution, which gives the 
same results.
Also, I restored the same Clonezilla image to another PC, again, with 
the same results. These PC's are pretty standard, neither brand new nor 
very old desktop boxes with a single 500GB disk.

The attachment shows a screenshot of the parted results together with 
those of sfdisk (which is proof but does not provide too much 
information). I also attached the extended parted output since it is 
mandatory but I don't think it provides much.

Clonezilla uses parted to check the partition layout. When I restore the 
image to a PC, I can use the PC normally but Clonezilla will not allow 
me to write a new image from it. I think I will be able to work around 
this, but it if this is a bug or quirk in parted it may merit some 
attention.

I don't know how parted determines the partition table type, but there 
must be something on the disk that makes it think it is Atari. If I can 
help by providing e.g., a dump of the boot sector of the disk, or any 
other data please let me know.

Bouke

[parted+fdisk_scrndmp.JPG (image/jpeg, attachment)]

[partedprintprintprint.JPG (image/jpeg, attachment)]

Message #8 received at submit <at> debbugs.gnu.org (full text, mbox):

From: bouke_1 <bouke_1 <at> zonnet.nl>
To: bug-parted <at> gnu.org
Subject: bug#31111: Partition table wrongly identified as "atari"
Date: Mon, 09 Apr 2018 17:10:22 +0200

Some additional observations:

Browsing the bug list archive (which I probably should have done before 
posting) I noticed that the support for Atari disks is implemented as a 
patch on top of version 3.2 
(http://lists.gnu.org/archive/html/bug-parted/2016-12/msg00004.html). No 
wonder I could not find the term 'atari' in the parted 3.2 source :) 
Looks like the function atari_probe may be fuddled by some coincidence?

<quote>
+/*
+ * The Atari disk label doesn't have any magic id
+ * so we must completely parse the layout to be sure
+ * we are really dealing with it.
+ */
</unquote>

Apparently, the version of parted in the Clonezilla distribution (and of 
partedlib used in gparted) is patched (parted --version reports 3.2).
I don't have a full Linux system at hand so I can't compile from source 
& depend on binary distributions.

Bouke

Message #11 received at submit <at> debbugs.gnu.org (full text, mbox):

From: bouke_1 <bouke_1 <at> zonnet.nl>
To: bug-parted <at> gnu.org
Subject: Re: bug#31111: Partition table wrongly identified as "atari"
Date: Tue, 10 Apr 2018 13:59:05 +0200

I quickly looked through the atari_probe routine in the Atari patch 
code, and I may have found the logic error. At the bottom:

     } /* no XGM so try ICD */
+         else if (atr_part_valid (&table.icd_part[0], rs_hd_size)
+                  && atr_part_known (&table.icd_part[0], 
atr_known_icd_pid)) {
+               for (i = 1; i < N_ICD; i++) {
+                       if (atr_part_trash (&table.icd_part[i], 
rs_hd_size))
+                               return 0;
+               }
+       }
+
+       return 1;
+ }

Correct me if I'm wrong (I know little about partitioning and less about 
atari disks), but it looks if the partition (or whatever) does not check 
out as "XGM" and either atr_part_valid  or atr_part_known  returns 
false, the routine will return 1 (meaning: this checks out as an atari 
disk).

In order to reach this part, only a few things need to check out (in the 
top part of the routine): three id values need to be alphanumeric and a 
couple of sizes need to be realistic (i.e., in my case, between 0 and 
500G). That does not seem like something that can never happen by 
coincidence for a non-atari disk.

I would guess it may be a good idea to add another "else return 0;" line 
so that if the disk is neither "XGM" nor "ICD" (whatever that may mean), 
it is not detected as atari disk.

I have the neither the stomach nor the means or the time to look into 
this but if I can assist in any way please let me know.

Bouke

Message #14 received at submit <at> debbugs.gnu.org (full text, mbox):

From: bouke_1 <bouke_1 <at> zonnet.nl>
To: bug-parted <at> gnu.org
Subject: Re: bug#31111: Partition table wrongly identified as "atari"
Date: Wed, 18 Apr 2018 17:07:20 +0200

For the record, let me add to this lonely conversation.

I tried to reproduce the problem in a simple way, by separating the 
sfdisk command from the Clonezilla operation where it first appeared. 
Clonezilla uses parted for information and sfdisk to configure the 
partitions.

Starting with a functional Windows 10 system with three partitions I ran 
some tests from the Clonezilla live distribution (parted says it is 
version 3.2).

1.parted says I have an atari disk (even though it is a functional 
Windows 10 system)
2.Instruct parted to create a new dos partition table
3.parted says I have an msdos disk with no partitions (as expected)
4.Run sfdisk with sda-pt.sf from the Clonezilla image as input
5.sfdisk creates three NTFS partitions on the disk (as expected)
6.parted says I have an atari disk with one partition

See the full log below. This reproduces on a bunch of PC's here (albeit, 
identical ones), so my guess it that it should reproduce elsewere?


Bouke


// log

sudo parted /dev/sda print

Model: ATA WDC WD5000AAKX-6 (scsi)
Disk /dev/sda: 500GB
Sector size (logical/physical): 512B/512B
Partition Table: atari
Disk Flags:

Number  Start   End     Size    Type     File system  Flags
 1      8323MB  14.8GB  6446MB  primary               boot

sudo parted -s /dev/sda mklabel msdos
sudo parted /dev/sda print

Model: ATA WDC WD5000AAKX-6 (scsi)
Disk /dev/sda: 500GB
Sector size (logical/physical): 512B/512B
Partition Table: msdos
Disk Flags:

Number  Start  End  Size  Type  File system  Flags

sudo LC_ALL=C sfdisk --force /dev/sda

label: dos
label-id: 0x22a777ed
device: /dev/sda
unit: sectors

/dev/sda1 : start=        2048, size=      716800, type=17, bootable
/dev/sda2 : start=      718848, size=   975028224, type=7
/dev/sda3 : start=   975747072, size=     1024000, type=27
Checking that no-one is using this disk right now ... OK

Disk /dev/sda: 465.8 GiB, 500107862016 bytes, 976773168 sectors
Units: sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disklabel type: dos
Disk identifier: 0xcb89b023

Old situation:

>>> Script header accepted.
>>> Script header accepted.
>>> Script header accepted.
>>> Script header accepted.
>>> Created a new DOS disklabel with disk identifier 0x22a777ed.
/dev/sda1: Created a new partition 1 of type 'Hidden HPFS/NTFS' and of 
size 350 MiB.
/dev/sda2: Created a new partition 2 of type 'HPFS/NTFS/exFAT' and of 
size 465 GiB.
/dev/sda3: Created a new partition 3 of type 'Hidden NTFS WinRE' and of 
size 500 MiB.
/dev/sda4: Done.

New situation:

Device     Boot     Start       End   Sectors  Size Id Type
/dev/sda1  *         2048    718847    716800  350M 17 Hidden HPFS/NTFS
/dev/sda2          718848 975747071 975028224  465G  7 HPFS/NTFS/exFAT
/dev/sda3       975747072 976771071   1024000  500M 27 Hidden NTFS WinRE

The partition table has been altered.
Calling ioctl() to re-read partition table.
Syncing disks.

sudo parted /dev/sda print

Model: ATA WDC WD5000AAKX-6 (scsi)
Disk /dev/sda: 500GB
Sector size (logical/physical): 512B/512B
Partition Table: atari
Disk Flags:

Number  Start   End     Size    Type     File system  Flags
 1      8323MB  14.8GB  6446MB  primary               boot

// end

Message #17 received at submit <at> debbugs.gnu.org (full text, mbox):

From: bouke_1 <bouke_1 <at> zonnet.nl>
To: bug-parted <at> gnu.org
Subject: Re: bug#31111: Partition table wrongly identified as "atari"
Date: Thu, 19 Apr 2018 17:17:40 +0200

One final observation: the size of sda2 seems to be the one that 
triggers the bug. If I change it from 9750281224 to 9750281222, the 
problem goes away (i.e., parted no longer identifies my Partition Table 
as Atari).

I rest my case.

Bouke

Message #20 received at 31111 <at> debbugs.gnu.org (full text, mbox):

From: Phil Susi <psusi <at> ubuntu.com>
To: bouke_1 <bouke_1 <at> zonnet.nl>, 31111 <at> debbugs.gnu.org
Subject: Re: bug#31111: Partition table wrongly identified as "atari"
Date: Thu, 10 May 2018 13:53:21 -0400

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

On 4/10/2018 7:59 AM, bouke_1 wrote:
> I quickly looked through the atari_probe routine in the Atari patch
> code, and I may have found the logic error. At the bottom:
> 
>      } /* no XGM so try ICD */
> +         else if (atr_part_valid (&table.icd_part[0], rs_hd_size)
> +                  && atr_part_known (&table.icd_part[0],
> atr_known_icd_pid)) {
> +               for (i = 1; i < N_ICD; i++) {
> +                       if (atr_part_trash (&table.icd_part[i],
> rs_hd_size))
> +                               return 0;
> +               }
> +       }
> +
> +       return 1;
> + }
> 
> Correct me if I'm wrong (I know little about partitioning and less about
> atari disks), but it looks if the partition (or whatever) does not check
> out as "XGM" and either atr_part_valid  or atr_part_known  returns
> false, the routine will return 1 (meaning: this checks out as an atari
> disk).

I'd say so.  I can't reproduce the problem though by simply creating an
image file and making the partitions as you describe.  Can you send your
whole MBR so I can try to reproduce it and see if fixing this logic
error fixes it?

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

Message #25 received at 31111-done <at> debbugs.gnu.org (full text, mbox):

From: Phil Susi <psusi <at> ubuntu.com>
To: bouke_1 <bouke_1 <at> zonnet.nl>, 31111-done <at> debbugs.gnu.org
Subject: Re: bug#31111: Partition table wrongly identified as "atari"
Date: Thu, 10 May 2018 14:52:52 -0400

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

On 5/10/2018 1:53 PM, Phil Susi wrote:
> I'd say so.  I can't reproduce the problem though by simply creating an
> image file and making the partitions as you describe.  Can you send your
> whole MBR so I can try to reproduce it and see if fixing this logic
> error fixes it?

Ok, after looking at the code it wasn't very hard to slap a few of the
right bytes into an empty file and trick atari_probe.  I've fixed it
now.  Will post the patch tomorrow.  Thanks.

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

Message #28 received at 31111 <at> debbugs.gnu.org (full text, mbox):

From: Phillip Susi <psusi <at> ubuntu.com>
To: bouke_1 <bouke_1 <at> zonnet.nl>, 31111 <at> debbugs.gnu.org
Subject: Re: bug#31111: Partition table wrongly identified as "atari"
Date: Fri, 11 May 2018 10:55:26 -0400

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

On 5/10/2018 2:52 PM, Phil Susi wrote:
> Ok, after looking at the code it wasn't very hard to slap a few of the
> right bytes into an empty file and trick atari_probe.  I've fixed it
> now.  Will post the patch tomorrow.  Thanks.

Nevermind; it wasn't a logical error.  As long as at least one partition
has a plausible start and end value and 3 ASCII characters in the name
field, it is a valid atari table.  I think I'll just move the atari
table to be probed last, after every other table has rejected it.

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

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.