Inconsistent partition table

w1k0 · 01-02-2009, 12:28 PM

January 8, 2:53 PM: Short summary of remaining problem and comparison of tar and lilo commands from Slackware CD-ROM, installed Slackware system, and Slax (see: post #13).

January 7, 7:16 PM: It seems I got rid of my problems with the first machine though there are still problems with the second machine (see: post #11).

January 6, 9:45 AM: The problem with inconsistent partition table returned (see: post #5). I don't understand it.

January 5, 11:53 PM: I solved the problems with invalid entries in partitions and Windows restoration from backup (see: post #4). The question what caused these errors in partition tables on my three machines remains unanswered.

***

I saw yesterday the message similar to that one for four times -- two times on one machine and two times on the other:

Code:

Warning: Partition 2 on /dev/sda is not marked Active. 
Warning: Device 0x0800: Inconsistent partition table, 2nd entry 
 CHS address in PT: 1000:0:1 --> LBA (16065000) 
 LBA address in PT: 15120000 --> CHS (941:45:1) 
Fatal: Either FIX-TABLE or IGNORE-TABLE must be specified 
If not sure, first try IGNORE-TABLE (-P ignore)

First case

It was when I prepared old machine for work with Windows. At the beginning I removed all partitions and created new ones: two for Windows, one for Linux, and one for swap.

Then I installed Slackware on the Linux partition. During LILO set up I added by mistake to the configuration empty first partition. When I chose LILO installation setup displayed message similar to the above. So I came back, repeated LILO set up without including first partition, installed LILO, and finalized Slackware configuration.

After reboot I ran new Slackware and unpacked backup of Windows system on the first partition. When I tried to reload LILO it displayed the same message as before. I tried lilo -P ignore and lilo -P fix commands but after reboot Windows didn't run.

So I decided to start from the beginning. I ran Slackware installation for the second time, removed and restored all partitions, etc. Amazing surprise. When I prepared partitions for the first time hard drive had 34976 cylinders. When I did it for the second time it had 70860 cylinders.

After second disk partitioning, Slackware installation, and Windows restoration I was able to reload LILO without errors.

Second case

It was when I prepared new machine for work with Slackware. On that machine on the first and the second partitions was installed Slackware: old on the second and new on the first. I decided to install new Slackware on the second partition, that is I decided to replace Slackware 12.1 by Slackware 12.2. Everything went well up to LILO installation when I saw the message exactly the same as that quoted at the beginning of that post.

I finalized Slackware configuration without installing LILO. Then I ran the machine using Slax to chroot to newly installed system. When I tried to reload LILO it displayed the same message so I ran lilo -P ignore command. At the end I started my new system and I reloaded LILO once again -- without errors.

Questions

I wonder what was wrong with these two machines and two hard disks.

Partitions on the first machine I prepared yesterday using Slackware 12.2. I did it twice. At the first attempt hard disk had 34976 cylinders -- at the second attempt it had 70860 cylinders. I don't understand it.

Partitions on the second machine I prepared a few moths ago using Slackware 12.1. Both original systems -- Slackware 12.2 on the first partition and Slackware 12.1 on the second -- worked well before I started to install Slackware 12.2 on the second partition.

During two installations of Slackware 12.2 on these two machines LILO displayed warnings similar to quoted at the beginning of that post. I can't explain it. In the first case I removed and restored all partitions. In the second case I simply told LILO to ignore that problem. In result I can now reload LILO on both these machines without errors.

I have no idea what was the cause of these problems, why first hard drive changed number of cylinders, and how to avoid such situations in the future. I don't know too if these warnings report some serious problems or only states some cosmetic irregularities.

Any assistance will be appreciated.

CJS · 01-02-2009, 08:49 PM

How about posting the output of the following commands for each of the two computers:

Code:

fdisk -l
sfdisk -luS -V
parted /dev/sda print

And replace sda above if the drive name is different. That should give a little more info about your HDD partitioning.

w1k0 · 01-04-2009, 04:31 PM

I use three machines:

1. ThinkPad T41 with 40 GB IDE HDD,

2. ThinkPad T60 with 160 GB SCSI HDD,

3. ThinkPad T40 with 80 GB IDE HDD.

In the above post I mentioned the first and the second machines (T41 and T60). Both disks installed in these machines came with these computers. On both were so called ``hidden partitions'' including installation versions of Windows XP. I kept that partition on the first machine but removed it completely from the second one. In the case of the first machine fdisk doesn't show ``hidden partition'' at all -- instead it shows that that disk has 36.5 GB. In the case of the second machine fdisk showed that partition as ``Compaq diagnostics'' (type 12) before I removed it.

First machine (T41)

I rearranged partition table on the first disk with fdisk once again. Afterwards Windows installation program changed partition table on that disk in it's own way. So the current partitioning on that disk isn't the same as it was when I encountered ``Inconsistent partition table'' error.

Now fdisk complains that ``Partition 1 does not end on cylinder boundary'' and sfdisk reports two further errors: ``end: (c,h,s) expected (1023,15,63) found (929,239,63)'' and ``start: (c,h,s) expected (1023,15,63) found (1023,0,1)''. These errors are apparently caused by Windows installation program. Moreover sfdisk displays message: ``end of partition 1 has impossible value for head: 239''. According to fdisk and sfdisk that hard drive has 16 heads. I suspect that I caused that error restoring Windows from backup and writing the image of the smaller partition on the bigger one. The output of parted seems to be correct though it reports that partition table is of type msdos.

Windows installation (T41)

Three years ago I did backup of installed and configured Windows. It was original IBM version of Windows XP installed from hidden partition on that disk. I prepared backup with the command cat /dev/hda1 | gzip > windows-backup.gz. I'm not sure that I used exactly that command. I could also use the command dd if=/dev/hda1 | gzip > windows-backup.gz.

Afterwards I restored Windows from that image for a few times using the command zcat -c windows-backup.gz | dd of=/dev/hda1. In the meantime I changed small 40 GB disk to bigger 80 GB one and successfully restored Windows from the same image on the new disk for a few further times. The first partition on the new disk was bigger than the first partition on the old one. I used that partition usually for Linux installation because I need Windows very seldom. When I needed Windows I restored it in a few minutes.

Now I tried to restore Windows from that backup on the original 40 GB hard drive. I could restore system and run it for the first time. Windows discovered new devices, because that laptop has now new main board, and rebooted the machine. After restart during boot procedures it displayed blue screen with long message:

Code:

 
A problem has been detected and Windows has been shut down to prevent damage to your computer.

IRQL_NOT_LESS_OR_EQUAL

.
.
.

Technical information:

*** STOP: 0x0000000A (0xF895D1E4, 0x00000002, 0x00000001, 0x80527470)

It was impossible to run restored Windows on that machine for the second time.

So I tried to install Windows using ``hidden partition'' on that disk. Installer reported an error:

Code:

The program has detected an error 110 on the partition starting at sector 63 on disk 1.

The length of the partition in the partition table is incorrect. 
The CHS length is 15468705, the LBA length is 14011137, and the File System length is 13607937.
The program has determined that the length can be changed to the correct value of 13607937.

Would you like The program to fix this error?

I said ``Yes'' and installer replied:

Code:

The partition table error was successfully fixed!

Unfortunately after some installation and a few reboots installer displayed similar blue screen as above and ran simplified version of Windows. It was possible to use it but it was trimmed. Moreover installer changed two partition of type c (W95 FAT32 (LBA)) to type 6 (FAT16) for the first partition and type 1c (Hidden W95 FAT32) for the second one.

So I started from the beginning and after rearranging partitions I managed to restore Windows from the backup. After first reboot it repaired file system displaying a lot of messages such as:

Code:

.
.
.
\WINDOWS\inf\msnike.PNF   first allocation unit is invalid, entry truncated.
\WINDOWS\inf\msmusb.PNF   first allocation unit is invalid, entry truncated.
\WINDOWS\inf\msmscsi.PNF   first allocation unit is invalid, entry truncated.
.
.
.

Now I can run restored and repaired Windows on that machine but I suspect the system is incomplete because scandisk removed a lot of allocation units.

T41$ fdisk -l /dev/hda

Code:

Disk /dev/hda: 36.5 GB, 36570653696 bytes
16 heads, 63 sectors/track, 70860 cylinders
Units = cylinders of 1008 * 512 = 516096 bytes
Disk identifier: 0xcccdcccd

   Device Boot      Start         End      Blocks   Id  System
/dev/hda1   *           1       13950     7030768+   c  W95 FAT32 (LBA)
Partition 1 does not end on cylinder boundary.
/dev/hda2           13951       33900    10054800   83  Linux
/dev/hda3           33901       69800    18093600    c  W95 FAT32 (LBA)
/dev/hda4           69801       70860      534240   82  Linux swap

T41$ sfdisk -luS -V /dev/hda

Code:

Disk /dev/hda: 70860 cylinders, 16 heads, 63 sectors/track
Units = sectors of 512 bytes, counting from 0

   Device Boot    Start       End   #sectors  Id  System
/dev/hda1   *        63  14061599   14061537   c  W95 FAT32 (LBA)
                end: (c,h,s) expected (1023,15,63) found (929,239,63)
/dev/hda2      14061600  34171199   20109600  83  Linux
/dev/hda3      34171200  70358399   36187200   c  W95 FAT32 (LBA)
                start: (c,h,s) expected (1023,15,63) found (1023,0,1)
/dev/hda4      70358400  71426879    1068480  82  Linux swap
end of partition 1 has impossible value for head: 239 (should be in 0-15)

T41$ parted /dev/hda print

Code:

Model: FUJITSU MHT2040AH (ide)
Disk /dev/hda: 36.6GB
Sector size (logical/physical): 512B/512B
Partition Table: msdos

Number  Start   End     Size    Type     File system  Flags
 1      32.3kB  7200MB  7200MB  primary  fat32        boot, lba
 2      7200MB  17.5GB  10.3GB  primary  reiserfs
 3      17.5GB  36.0GB  18.5GB  primary  fat32        lba
 4      36.0GB  36.6GB  547MB   primary  linux-swap

Second machine (T60)

The partition scheme on the second disk is the same as it was when I encountered ``Inconsistent partition table'' error. Program sfdisk displays two warnings: ``The partition table looks like it was made for C/H/S=*/240/63 (instead of 19457/255/63)'' and ``partition 1 does not end at a cylinder boundary''. Maybe these errors are remnants of my attempts of restoring Windows from the backup on that machine. Program parted reports that partition table is of type msdos.

Windows installation (T60)

As I wrote at the beginning I removed ``hidden partition'' with Windows installation program from the second disk so I couldn't install Windows on that machine in a regular way. When I couldn't restore nor install Windows on the first machine I tried to restore it from the same backup on the second one but it displayed the message similar to the quoted above: ``A problem has been detected and Windows has been shut down to prevent damage to your computer''. In result it was impossible to run Windows on that machine.

T60$ fdisk -l /dev/sda

Code:

Disk /dev/sda: 160.0 GB, 160041885696 bytes
240 heads, 63 sectors/track, 20673 cylinders
Units = cylinders of 15120 * 512 = 7741440 bytes
Disk identifier: 0xcccdcccd

   Device Boot      Start         End      Blocks   Id  System
/dev/sda1   *           1        1000     7559968+  83  Linux
/dev/sda2            1001        2000     7560000   83  Linux
/dev/sda3            2001       20400   139104000   83  Linux
/dev/sda4           20401       20673     2063880   82  Linux swap

T60$ sfdisk -luS -V /dev/sda

Code:

Disk /dev/sda: 19457 cylinders, 255 heads, 63 sectors/track
Warning: The partition table looks like it was made
  for C/H/S=*/240/63 (instead of 19457/255/63).
For this listing I'll assume that geometry.
Units = sectors of 512 bytes, counting from 0

   Device Boot    Start       End   #sectors  Id  System
/dev/sda1   *        63  15119999   15119937  83  Linux
/dev/sda2      15120000  30239999   15120000  83  Linux
/dev/sda3      30240000 308447999  278208000  83  Linux
/dev/sda4     308448000 312575759    4127760  82  Linux swap
Warning: partition 1 does not end at a cylinder boundary

T60$ parted /dev/sda print

Code:

Model: ATA WDC WD1600BEVS-0 (scsi)
Disk /dev/sda: 160GB
Sector size (logical/physical): 512B/512B
Partition Table: msdos

Number  Start   End     Size    Type     File system  Flags
 1      32.3kB  7741MB  7741MB  primary  reiserfs     boot 
 2      7741MB  15.5GB  7741MB  primary  reiserfs          
 3      15.5GB  158GB   142GB   primary  reiserfs          
 4      158GB   160GB   2113MB  primary  linux-swap

Third machine (T40)

On that hard disk sfdisk displays only one warning: ``The partition table looks like it was made for C/H/S=*/240/63 (instead of 155061/16/63)''. It's apparently caused by my attempts to restore Windows. As in two former cases parted reports that partition table is of type msdos.

Windows installation (T40)

That disk was formerly in the first machine. I tried to restore Windows from backup on it but I encountered blue screen with the mentioned above message: ``A problem has been detected and Windows has been shut down to prevent damage to your computer''. So I removed that disk from the first machine, inserted it into the third machine, took original old disk with ``hidden partition'', and inserted it into the first machine. What was further with the first machine I wrote in the first section of that post.

When I was unable to restore nor install Windows on the first machine I ran the third machine with hard disk taken from the first machine. Surprisingly Windows started without errors though it declined to start on the first machine. So I removed Windows and restored it once again from the backup. It still works. The third machine is the only one on which I can run Windows from my backup. The problem is that I don't need Windows on the third machine. I need it on the first and the second machines.

T40# fdisk -l /dev/hda

Code:

Disk /dev/hda: 80.0 GB, 80026361856 bytes
240 heads, 63 sectors/track, 10337 cylinders
Units = cylinders of 15120 * 512 = 7741440 bytes
Disk identifier: 0x3db012b3

   Device Boot      Start         End      Blocks   Id  System
/dev/hda1               1        1025     7748968+   c  W95 FAT32 (LBA)
/dev/hda2            1026        2050     7749000   83  Linux
/dev/hda3            2051       10250    61992000   83  Linux
/dev/hda4           10251       10337      657720   82  Linux swap

T40# sfdisk -luS -V /dev/hda

Code:

Disk /dev/hda: 155061 cylinders, 16 heads, 63 sectors/track
Warning: The partition table looks like it was made
  for C/H/S=*/240/63 (instead of 155061/16/63).
For this listing I'll assume that geometry.
Units = sectors of 512 bytes, counting from 0

   Device Boot    Start       End   #sectors  Id  System
/dev/hda1            63  15497999   15497937   c  W95 FAT32 (LBA)
/dev/hda2      15498000  30995999   15498000  83  Linux
/dev/hda3      30996000 154979999  123984000  83  Linux
/dev/hda4     154980000 156295439    1315440  82  Linux swap
Warning: no primary partition is marked bootable (active)
This does not matter for LILO, but the DOS MBR will not boot this disk.
end of partition 1 has impossible value for head: 239 (should be in 0-15)

T40# parted /dev/hda print

Code:

Model: TOSHIBA MK8026GAX (ide)
Disk /dev/hda: 80.0GB
Sector size (logical/physical): 512B/512B
Partition Table: msdos

Number  Start   End     Size    Type     File system  Flags
 1      32.3kB  7935MB  7935MB  primary  fat32        lba
 2      7935MB  15.9GB  7935MB  primary  reiserfs
 3      15.9GB  79.3GB  63.5GB  primary  reiserfs
 4      79.3GB  80.0GB  674MB   primary  linux-swap

A few comments and one question

Both the first and the second disks have the same identifier: 0xcccdcccd. I sought for that string in Internet and it seems it's common identifier for disks dedicated to ThinkPads.

I suspect most of these errors and warnings are effects of my attempts of restoring Windows from backup on these three machines. The system stored three years ago on the first machine works well on the third machine. It works on the first machine after truncating of a lot of allocation units. It doesn't work on the second machine -- perhaps because the second hard disk is SCSI while the first and the third disks are IDE or because the second machine has two processors while the first and the third have one processor unit.

In the past I restored Windows from that backup on the first machine for a few times and never encountered any problem. Now I can't invent the method of restoring or installing Windows on the first and the second machines. I can merely restore it on the third machine but as it turns out it's the only machine on which I don't need Windows.

Of course I can backup now all data once again and try to arrange partitions on these three machines anew but there is none warranty that it'll help. It might as well lead to the same situation: I'll try to install or restore Windows and I'll cause the same errors. It might also be worse than now: I can lose Windows from the third machine and be unable to restore it on the first and on the second ones.

What can I do in such situation?

w1k0 · 01-05-2009, 05:17 PM

CJS, thank you for your valuable hint. Thanks to you and thanks to my researches I repaired partition tables on all machines and I installed Windows on the first one.

***

First machine -- successful partitioning

I decided to repair partition tables on my machines. I started from the first machine.

1. I ran Slax.

2. I removed all partitions from the first disk with fdisk and wrote an empty partition.

3. I restored the same set of partitions with fdisk.

4. I rebooted machine and ran Slackware.

5. I tested repaired partition table with fdisk, sfdisk, and parted (all outputs are all right now):

Code:

root@home5:~# fdisk -l /dev/hda

Disk /dev/hda: 36.5 GB, 36570653696 bytes
16 heads, 63 sectors/track, 70860 cylinders
Units = cylinders of 1008 * 512 = 516096 bytes
Disk identifier: 0xcccdcccd

   Device Boot      Start         End      Blocks   Id  System
/dev/hda1   *           1       13950     7030768+   c  W95 FAT32 (LBA)
/dev/hda2           13951       33900    10054800   83  Linux
/dev/hda3           33901       69800    18093600    c  W95 FAT32 (LBA)
/dev/hda4           69801       70860      534240   82  Linux swap

root@home5:~# sfdisk -luS -V /dev/hda

Disk /dev/hda: 70860 cylinders, 16 heads, 63 sectors/track
Units = sectors of 512 bytes, counting from 0

   Device Boot    Start       End   #sectors  Id  System
/dev/hda1   *        63  14061599   14061537   c  W95 FAT32 (LBA)
/dev/hda2      14061600  34171199   20109600  83  Linux
/dev/hda3      34171200  70358399   36187200   c  W95 FAT32 (LBA)
/dev/hda4      70358400  71426879    1068480  82  Linux swap
/dev/hda: OK

root@home5:~# parted /dev/hda print

Model: FUJITSU MHT2040AH (ide)
Disk /dev/hda: 36.6GB
Sector size (logical/physical): 512B/512B
Partition Table: msdos

Number  Start   End     Size    Type     File system  Flags
 1      32.3kB  7200MB  7200MB  primary  fat32        boot, lba
 2      7200MB  17.5GB  10.3GB  primary  reiserfs
 3      17.5GB  36.0GB  18.5GB  primary  fat32        lba
 4      36.0GB  36.6GB  547MB   primary  linux-swap

Windows installation -- successful attempt

Before I started to repair partition tables I read a bit about errors reported by Windows which I described in the previous post. I established that ``IRQL_NOT_LESS_OR_EQUAL'' error means: ``This bug check is issued if paged memory (or invalid memory) is accessed when the IRQL is too high'' and that ``0x0000000A'' stop error means: ``If you receive a "Stop: 0x0000000A" error message during the installation of Windows XP, there might be a problem with the hardware components installed on your computer''.

So I restored Windows from backup on the first partition, I ran it for the first time, and I observed what will happen. During discovering devices just after Windows found``Intel(R) PRO/Wireless LAN 2100 3B Mini PCI Adapter'' I saw blue screen with the above error codes.

So I restored Windows from backup once again, disabled in BIOS option ``Internal Wireless WAN Device'', and ran Windows anew for the first time. It discovered all devices without problem. I rebooted machine and ran Windows for the next time. It started without problems. I rebooted machine once again, enabled the above option in BIOS, and ran Windows for the further time. Just after discovering Wireless LAN Adapter it displayed blue screen and rebooted machine. Then Windows started to check file system but found merely four invalid allocation units (formerly it founded hundreds of errors). After system start the only not working device was Wireless LAN Adapter.

For some reason my version of Windows XP can't cope with that device. It applies to my backup as well as to installation version on ``hidden partition''. I suppose I could install the appropriate driver from Internet.

Third machine -- successful partitioning with minor problem

Then I repaired partitions on the third machine.

As before I ran Slax, removed all partitions from the third disk with fdisk, and wrote an empty partition. At that step I stated that fdisk changed numbers of heads and cylinders to proper values.

Before repartitioning it was:

Code:

Disk /dev/hda: 80.0 GB, 80026361856 bytes
240 heads, 63 sectors/track, 10337 cylinders
Units = cylinders of 15120 * 512 = 7741440 bytes
Disk identifier: 0x3db012b3

After it was:

Code:

Disk /dev/hda: 80.0 GB, 80026361856 bytes
16 heads, 63 sectors/track, 155061 cylinders
Units = cylinders of 1008 * 512 = 516096 bytes
Disk identifier: 0x3db012b3

Fortunately 240 / 16 = 15 so I could increase the numbers of cylinders ending partitions multiplying former values by 15:

Code:

   Device Boot      Start         End      Blocks   Id  System
/dev/hda1   *           1       15375     7748968+   c  W95 FAT32 (LBA)
/dev/hda2           15376       30750     7749000   83  Linux
/dev/hda3           30751      153750    61992000   83  Linux
/dev/hda4          153751      155061      660744   82  Linux swap

Though partitions ends and starts are now different than were before first three partitions have exactly the same size as they had before.

After reboot I was able to run Linux as well as Windows. Partition table is now all right:

Code:

root@home4:~# fdisk -l /dev/hda

Disk /dev/hda: 80.0 GB, 80026361856 bytes
16 heads, 63 sectors/track, 155061 cylinders
Units = cylinders of 1008 * 512 = 516096 bytes
Disk identifier: 0x3db012b3

   Device Boot      Start         End      Blocks   Id  System
/dev/hda1   *           1       15375     7748968+   c  W95 FAT32 (LBA)
/dev/hda2           15376       30750     7749000   83  Linux
/dev/hda3           30751      153750    61992000   83  Linux
/dev/hda4          153751      155061      660744   82  Linux swap

root@home4:~# sfdisk -luS -V /dev/hda

Disk /dev/hda: 155061 cylinders, 16 heads, 63 sectors/track
Units = sectors of 512 bytes, counting from 0

   Device Boot    Start       End   #sectors  Id  System
/dev/hda1   *        63  15497999   15497937   c  W95 FAT32 (LBA)
/dev/hda2      15498000  30995999   15498000  83  Linux
/dev/hda3      30996000 154979999  123984000  83  Linux
/dev/hda4     154980000 156301487    1321488  82  Linux swap
/dev/hda: OK

root@home4:~# parted /dev/hda print

Model: TOSHIBA MK8026GAX (ide)
Disk /dev/hda: 80.0GB
Sector size (logical/physical): 512B/512B
Partition Table: msdos

Number  Start   End     Size    Type     File system  Flags
 1      32.3kB  7935MB  7935MB  primary  fat32        boot, lba
 2      7935MB  15.9GB  7935MB  primary  reiserfs
 3      15.9GB  79.3GB  63.5GB  primary  reiserfs
 4      79.3GB  80.0GB  677MB   primary  linux-swap

Second machine -- successful partitioning with major problem

That case was more complicated because program sfdisk reported:

Code:

Disk /dev/sda: 19457 cylinders, 255 heads, 63 sectors/track
Warning: The partition table looks like it was made
  for C/H/S=*/240/63 (instead of 19457/255/63).

Since 240 / 255 = 0.94 there was no possible to decrease partitions sizes by 0.94. Because I had all data stored in backups I decided to repair partition table on the second disk loosing data on the second and the third partitions.

After reboot I was able to run Slackware 12.2 from first partition. It changed size but the beginning of that partition is still on the same position as it was before:

Code:

root@home5:~# fdisk -l /dev/hda

Disk /dev/hda: 36.5 GB, 36570653696 bytes
16 heads, 63 sectors/track, 70860 cylinders
Units = cylinders of 1008 * 512 = 516096 bytes
Disk identifier: 0xcccdcccd

   Device Boot      Start         End      Blocks   Id  System
/dev/hda1   *           1       13950     7030768+   c  W95 FAT32 (LBA)
/dev/hda2           13951       33900    10054800   83  Linux
/dev/hda3           33901       69800    18093600    c  W95 FAT32 (LBA)
/dev/hda4           69801       70860      534240   82  Linux swap

root@home5:~# sfdisk -luS -V /dev/hda

Disk /dev/hda: 70860 cylinders, 16 heads, 63 sectors/track
Units = sectors of 512 bytes, counting from 0

   Device Boot    Start       End   #sectors  Id  System
/dev/hda1   *        63  14061599   14061537   c  W95 FAT32 (LBA)
/dev/hda2      14061600  34171199   20109600  83  Linux
/dev/hda3      34171200  70358399   36187200   c  W95 FAT32 (LBA)
/dev/hda4      70358400  71426879    1068480  82  Linux swap
/dev/hda: OK

root@home5:~# parted /dev/hda print

Model: FUJITSU MHT2040AH (ide)
Disk /dev/hda: 36.6GB
Sector size (logical/physical): 512B/512B
Partition Table: msdos

Number  Start   End     Size    Type     File system  Flags
 1      32.3kB  7200MB  7200MB  primary  fat32        boot, lba
 2      7200MB  17.5GB  10.3GB  primary  reiserfs
 3      17.5GB  36.0GB  18.5GB  primary  fat32        lba
 4      36.0GB  36.6GB  547MB   primary  linux-swap

Conclusions

I don't know what caused all these errors in partition tables on these three disks. I suspected my backup of Windows XP but after repairing partitions on the first and the third machines I restored Windows from backup and the partition tables on these disks are still valid. Though I resolved these problems my first question is still up-to-date: I have no idea what was the cause of these problems, why first hard drive changed number of cylinders, and how to avoid such situations in the future. I'll be grateful for any suggestions.

As I guessed in the previous post the problems with Windows backup were caused by hardware. The first machine has new main board and the second has two processors and SCSI hard disk. Switching the appropriate option in BIOS I reduced problem with Windows restoration from backup to minimum. As for the second machine I need call service guys and ask them for spare installation disks with Windows XP.

w1k0 · 01-06-2009, 02:40 AM

The problem mentioned in post #1 reappeared though partition table seems to be good now according to fdisk, sfdisk, and parted. I'm pretty sure there was no that problem yesterday when I restored fresh version of Slackware on the second partition and it returned today. At the moment it concerns the second machine (T60, home6). I don't understand it.

root@home6:/etc$ lilo

Code:

Warning: LBA32 addressing assumed
Added linux.generic *
Added linux.huge
Warning: Device 0x0800: Inconsistent partition table, 2nd entry
  CHS address in PT:  965:0:1  -->  LBA (14590800)
  LBA address in PT:  15502725  -->  CHS (1025:75:1)
Fatal: Either FIX-TABLE or IGNORE-TABLE must be specified
If not sure, first try IGNORE-TABLE (-P ignore)

root@home6:/etc$ fdisk -l /dev/sda

Code:

Disk /dev/sda: 160.0 GB, 160041885696 bytes
255 heads, 63 sectors/track, 19457 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Disk identifier: 0xcccdcccd

   Device Boot      Start         End      Blocks   Id  System
/dev/sda1   *           1         965     7751331   83  Linux
/dev/sda2             966        1930     7751362+  83  Linux
/dev/sda3            1931       19200   138721275   83  Linux
/dev/sda4           19201       19457     2064352+  82  Linux swap

root@home6:/etc$ sfdisk -luS -V /dev/sda

Code:

Disk /dev/sda: 19457 cylinders, 255 heads, 63 sectors/track
Units = sectors of 512 bytes, counting from 0

   Device Boot    Start       End   #sectors  Id  System
/dev/sda1   *        63  15502724   15502662  83  Linux
/dev/sda2      15502725  31005449   15502725  83  Linux
/dev/sda3      31005450 308447999  277442550  83  Linux
/dev/sda4     308448000 312576704    4128705  82  Linux swap
/dev/sda: OK

root@home6:/etc$ parted /dev/sda print

Code:

Model: ATA WDC WD1600BEVS-0 (scsi)
Disk /dev/sda: 160GB
Sector size (logical/physical): 512B/512B
Partition Table: msdos

Number  Start   End     Size    Type     File system  Flags
 1      32.3kB  7937MB  7937MB  primary  reiserfs     boot
 2      7937MB  15.9GB  7937MB  primary  reiserfs
 3      15.9GB  158GB   142GB   primary  reiserfs
 4      158GB   160GB   2114MB  primary  linux-swap

root@home6:/etc$ cat lilo.conf | grep -v ^#

Code:

boot = /dev/sda
compact        # faster, but won't work on all systems.
append=" vt.default_utf8=0 resume=/dev/sda4 acpi_sleep=s3_bios,s3_mode libata.atapi_enabled=1 combined_mode=libata"
prompt
timeout = 50
vga = 791
image = /boot/vmlinuz-generic-smp-2.6.27.7-smp
  initrd = /boot/initrd.gz
  root = /dev/sda1
  label = linux.generic
  read-only
image = /boot/vmlinuz-huge-smp-2.6.27.7-smp
  root = /dev/sda1
  label = linux.huge
  read-only  # Partitions should be mounted read-only for checking
other = /dev/sda2
  label = linux.new

root@home6:/etc$ lilo -P ignore

Code:

Warning: LBA32 addressing assumed
Added linux.generic *
Added linux.huge
Warning: Device 0x0800: Inconsistent partition table, 2nd entry
  CHS address in PT:  965:0:1  -->  LBA (14590800)
  LBA address in PT:  15502725  -->  CHS (1025:75:1)
Warning: The partition table is *NOT* being adjusted.
Added linux.new
3 warnings were issued.

root@home6:/etc$ lilo -P fix

Code:

Warning: LBA32 addressing assumed
Added linux.generic *
Added linux.huge
Warning: Device 0x0800: Inconsistent partition table, 2nd entry
  CHS address in PT:  965:0:1  -->  LBA (14590800)
  LBA address in PT:  15502725  -->  CHS (1025:75:1)
Writing modified partition table to device 0x0800
Added linux.new
2 warnings were issued.

w1k0 · 01-06-2009, 11:44 AM

Tomorrow morning I installed Slackware 12.2 on the second partition and changed some options in lilo.conf. Because I was unable to reload LILO using lilo command I tried both lilo -P ignore and lilo -P fix commands.

Tomorrow afternoon I modified lilo.conf for the consecutive time. I was unable to reload LILO using lilo or lilo -P ignore commands so I used lilo -P fix once again. That last command modified partition table causing errors similar to these described in post #3:

Code:

Disk /dev/sda: 19457 cylinders, 255 heads, 63 sectors/track
Units = sectors of 512 bytes, counting from 0

   Device Boot    Start       End   #sectors  Id  System
/dev/sda1   *        63  15502724   15502662  83  Linux
/dev/sda2      15502725  31005449   15502725  83  Linux
                start: (c,h,s) expected (965,0,1) found (1023,75,1)
/dev/sda3      31005450 308447999  277442550  83  Linux
/dev/sda4     308448000 312576704    4128705  82  Linux swap
partition 2: start: (c,h,s) expected (965,0,1) found (1023,75,1)
/dev/sda: OK

The main difference is that formerly I registered similar errors on the first machine (T41, home5) and now I registered these errors on the second machine (T60, home6).

There is something wrong either with my machines or with Slackware 12.2. I'm starting to suspect Slackware.

Every help will be welcomed.

bgeddy · 01-06-2009, 12:21 PM

Quote:

Tomorrow morning I installed Slackware 12.2 on the second partition and changed some options in lilo.conf.

That's a neat trick - lend me your time machine

(Just kidding - you probably meant yesterday or today).

Anyway you seem to be having a hell of a time with this. Just a few points..

Your method of backing up and restoring partitions seems faulty and may well be causing problems here. Unless all you machines have simultaneously developed hardware faults.

Windows doesn't really like clone backups unless special software is used and can throw strange fits if partition sizes change.

So - here's what will probably solve you're problems if possible and save 2009 from becoming your worst year yet !

1) Just wipe out all your partitions and recreate target partitons with parted/gparted for Windows and Linux.

2) Install fresh Windows from install media to suitable target partitons (helps if you let Windows be the first OS on the disk). Check Windows boots

3) Install fresh Slackware to reserved partitions and install lilo to MBR of your disks - let lilo handle loading Windows.

4) I noticed you have Thinkpads - make sure the disks are being correctly picked up as hda or sda depending on what controller you have.

5) Have a Happy New Year because this should all work now.. Relax...

CJS · 01-06-2009, 12:33 PM

Hi w1k0, you might want to consider using "sfdisk" to try and fix your partition table geometry errors. We used the technique here:

http://ubuntuforums.org/showthread.php?t=1013167

If you want to try that, the first thing to do is backup the current partition table by doing:

Code:

sfdisk -d /dev/sda > partition_table.txt

And put the partition_table.txt in a safe place. Then try running:

Code:

sfdisk -d /dev/sda | sfdisk --force  /dev/sda

What that does is output the original partition table, and then sfdisk rewrites the same partition table. But in the process of rewriting the table, sfdisk always uses the correct CHS values, because the original CHS values are not outputted by sfdisk; only the LBA values are outputted (start and size of the partitions). But what was a little confusing about the above thread is that it wasn't immediately clear whether that had fixed the problem, so I had the poster also reinstate the original partition table by doing:

Code:

sfdisk --force /dev/sda < partition_table.txt

With the partition_table.txt that was their original partition table. And that for sure fixed his problem. You could give it a shot on your setup and see if it works, and please let me know the results if you decide to give it a try.

w1k0 · 01-07-2009, 10:25 AM

bgeddy:

Quote:

That's a neat trick - lend me your time machine (Just kidding - you probably meant yesterday or today).

In fact it should be ``today'' not ``tomorrow'' regardless of my machine time. On the other hand I work so hard on these problems that I feel I borrow some time from tomorrow.

Quote:

Your method of backing up and restoring partitions seems faulty and may well be causing problems here.

I use two methods of backing up and restoring systems. In the case of Windows I use dd and gzip to read the entire partition and zcat and dd to write it back. In the case of Linux I use tar and gzip to store the system and tar and gzip to restore it.

The first method turns out to be useful in the case of Windows. I used it for years and I never encountered any problems. It's enough to prepare for backup restoration the partition which isn't smaller than the image of the partition stored in the backup. The problems with Windows described in the post #3 were caused by new hardware on both these machines. My backup has three years. I tried to restore partition image from it on the machine with new main board and on the machine with two processors and SATA drive. After some reading in Internet I managed to narrow down the problem in the case of the first machine to one driver for wireless device and I don't try already to restore Windows on the second machine from the backup made on the first machine because I know now that that's impossible. I wrote about it in the post #4.

Quote:

So - here's what will probably solve you're problems if possible

In fact I did it already before I read your post -- I'll write about it in my next post. As to the first step I used fdisk and prepared Linux partitions only. I change them to Windows partitions when I need restore Windows and I change them back to Linux partitions when I don't need Windows but Linux. As for the second step I didn't install Windows as you suggest because my Windows problem is already established and I know what to do to restore that system on the first machine from my old backup successfully. (Surprisingly it's impossible now to install Windows on that machine using original IBM installer stored on ``the hidden partition'' though it's possible to restore Windows from the backup I made three years ago.) In the third step I installed Slackware anew once again, made backup of that installation, and cloned successfully the system from that backup on three remaining partitions on two machines. As to fourth step all my machines recognize hard disks correctly. As for fifth step: thank you very much and Happy New Year for all Slackers all over the world!

CJS:

Quote:

you might want to consider using "sfdisk" to try and fix your partition table geometry errors

I'll consider it. I read sfdisk's manual and I noticed the warning: ``BE EXTREMELY CAREFUL - ONE TYPING MISTAKE AND ALL YOUR DATA IS LOST''. I know fdisk well and I don't know sfdisk at all. So far I flinched from using sfdisk. Your method seems to be secure. If my attempts with fdisk will fail I'll try sfdisk.

In the next post I'll describe what I did yesterday evening and today morning. I narrowed my problems down though I encountered one new problem with LILO.

Thank you for reply.

CJS · 01-07-2009, 10:45 AM

Quote:

Originally Posted by w1k0

I read sfdisk's manual and I noticed the warning: ``BE EXTREMELY CAREFUL - ONE TYPING MISTAKE AND ALL YOUR DATA IS LOST''. I know fdisk well and I don't know sfdisk at all. So far I flinched from using sfdisk. Your method seems to be secure. If my attempts with fdisk will fail I'll try sfdisk.

IMO I don't think that quote in the manual is quite accurate, because all sfdisk does is write to the partition table found in the MBR (Master Boot Record), and also to the partition table in each of the EBRs (Extended Boot Records) that are associated with each logical partition. Even if you accidentally use sfdisk to set up a completely bogus partition table, the worst that can happen is for each logical partition created, there will be one sector (usually located 63 sectors before the logical partition) that will be written to the disk to contain the EBR for that logical partition. So in the worst case scenario, if you create N bogus logical partitions, only a total of N sectors will be overwritten by sfdisk, and depending on how you set up the partition table will determine where those sectors on the HDD will be overwritten. So for the manual to say that all your data will be lost is simply not true; the data is still there, but it won't be easily accessible until you correct the partition table so you can mount the partitions. I've used "sfdisk -d" to back up and restore partition tables without any problems, and there are of course great tools like testdisk that can make recovering a deleted or corrupted partition table quite easy. Anyway, please keep me posted about what you decide to do.

w1k0 · 01-07-2009, 11:57 AM

Yesterday I started to suspect that there is some kind of virus on my pendrive with Slax. I used it once on the Windows machine of my friend and now I use it to prepare partitions and to restore Linuces from backups. So I wiped my pendrive out with dd if=/dev/zero of=/dev/sdb command, restored partition with fdisk, restored filesystem with mkdosfs, and restored Slax.

Second machine (/dev/sda)

Then I started Slackware installation on the second machine, removed and restored all partitions with fdisk, installed Slackware on the first partition, used my set of scripts to configure it, and installed a lot of additional software. Next I did backup of newly installed and configured system with tar and gzip, and afterwards I restored Slackware from that backup on the second partition of the second machine as well as on the first and the second partitions on the first machine.

There weren't problems with LILO up to that moment.

Then I tried to configure LILO on the second machine to allow it to see the other system. So I added to lilo.conf entries:

Code:

other = /dev/sda2
  label /dev/sda2

on the first partition and:

Code:

other = /dev/sda1
  label /dev/sda1

on the second one.

When I tried to reload LILO it claimed:

root@home6:~# lilo

Code:

Warning: LBA32 addressing assumed
Added linux.generic *
Added linux.huge
Warning: Device 0x0800: Inconsistent partition table, 2nd entry
  CHS address in PT:  965:0:1  -->  LBA (14590800)
  LBA address in PT:  15502725  -->  CHS (1025:75:1)
Fatal: Either FIX-TABLE or IGNORE-TABLE must be specified
If not sure, first try IGNORE-TABLE (-P ignore)

It added two partitions in the case of the first partition's installation as in the example above but refused to add any partition in the case of the second partition's installation.

So I ran Slackware CD-ROM, mounted both these partitions, chrooted to these systems, and reloaded LILO with lilo command. It reloaded it without errors. In the case of the second machine I'm unable to reload LILO from Slackware installations without errors but I'm able to do it using Slackware CD-ROM.

First machine (/dev/hda)

Then I added these entries to lilo.conf on the first machine:

Code:

other = /dev/hda2
  label /dev/hda2

and:

Code:

other = /dev/hda1
  label /dev/hda1

I was able to reload LILO from the second partition without errors. When I tried to do it from the first partition it claimed:

root@home4:/etc$ lilo

Code:

Warning: LBA32 addressing assumed
Warning: partition type 0x0C on device 0x0301 is a dangerous place for
    a boot sector.  A DOS/Windows/OS2 system may be rendered unbootable.
  The backup copy of this boot sector should be retained.

Proceed? [N/y]y
Added linux.generic *
Added linux.huge
Added /dev/hda2
2 warnings were issued.

So I encountered a new kind of problem.

Then I started Slackware installation on the first machine, mounted both partitions, chrooted to both systems, and reloaded LILO without errors.

Finally I started both systems and managed to reload LILO without problems. The output of sfdisk -luS -V command in the case of the first machine's systems is correct too.

[EDIT]
I caused the above error restoring Linux backup on Windows partition (W95 FAT32 (LBA) -- type c).
[/EDIT]

At the moment there is no problem with the first machine.

Conclusions

It seems I got rid of my problems with the first machine though there are still problems with the second machine.

Now I'll try to repair partitions on the second machine once again in my own way. I'm determined to remove all these problems for ever.

w1k0 · 01-07-2009, 12:04 PM

Quote:

Anyway, please keep me posted about what you decide to do.

I will.

I suspect there is some bogus data on the partition table on my second machine. Earlier I wanted to wipe out the beginnings of /dev/sda, /dev/sda1, and /dev/sda2 with dd if=/dev/zero commands, then restore partitions and file systems using Slackware CD-ROM, and finally restore Slackware on both partitions from my new backup using Slax. Now I intend to use your method first. It's simpler and can be efficacious.

w1k0 · 01-07-2009, 06:44 PM

Second machine

I dumped the image of the partitions on pendrive and restored partitions using sfdisk and that image. I used Slackware CD-ROM to do it. Result was the same as before. I was able to reload LILO from within Slackware CD-ROM but when I tried to do it from within Slackware system LILO displayed the error ``Inconsistent partition table''.

***

Then I started Slackware CD-ROM, wiped out 1 MB of data from /dev/sda2, /dev/sda1, and /dev/sda with dd commands, restored all partitions with fdisk, restored file systems on two first partitions with mkreiserfs, and restored Slackware on the first and the second partitions from the backup stored on the third partition using the command tar xzf slackware-backup.tgz. That command complained:

Code:

tar: incorrect length

When I restore the system from that backup using Slax or running Slackware tar doesn't complain. It complains when I use Slackware CD-ROM.

Next I started both restored systems and I tried to reload LILO. Results were exactly the same as before. I saw in both cases the error ``Inconsistent partition table''.

LILO from Slackware from the first partition was able to add first partition but complained about the second:

root@home6:~# lilo

Code:

Warning: LBA32 addressing assumed
Added linux.generic *
Added linux.huge
Warning: Device 0x0800: Inconsistent partition table, 2nd entry
  CHS address in PT:  965:0:1  -->  LBA (14590800)
  LBA address in PT:  15502725  -->  CHS (1025:75:1)
Fatal: Either FIX-TABLE or IGNORE-TABLE must be specified
If not sure, first try IGNORE-TABLE (-P ignore)

LILO from Slackware from the second partition was unable to add any partition:

root@home6:~$ lilo

Code:

Warning: LBA32 addressing assumed
Warning: Partition 2 on /dev/sda is not marked Active.
Warning: Device 0x0800: Inconsistent partition table, 2nd entry
  CHS address in PT:  965:0:1  -->  LBA (14590800)
  LBA address in PT:  15502725  -->  CHS (1025:75:1)
Fatal: Either FIX-TABLE or IGNORE-TABLE must be specified
If not sure, first try IGNORE-TABLE (-P ignore)

When I include in lilo.conf the section ``other'' pointing Slackware on the other partition I can reload LILO without errors only from within Slackware CD-ROM or Slax.

root@home6:~$ cat /etc/lilo.conf | grep -v ^#

Code:

boot = /dev/sda
compact        # faster, but won't work on all systems.
append=" vt.default_utf8=0 resume=/dev/sda4 acpi_sleep=s3_bios,s3_mode"
prompt
timeout = 50
vga = 791
image = /boot/vmlinuz-generic-smp-2.6.27.7-smp
  initrd = /boot/initrd.gz
  root = /dev/sda1
  label = linux.generic
  read-only
image = /boot/vmlinuz
  root = /dev/sda1
  label = linux.huge
  read-only  # Partitions should be mounted read-only for checking
other = /dev/sda2
  label = /dev/sda2

***

Finally I tried to create backup of the system using Slackware CD-ROM. I did backup of the first partition and restored the system from it on the second one. The program again complained:

Code:

tar: incorrect length

When I use Slackware CD-ROM to restore backup tar complains. When I use installed Slackware or Slax it remains silent.

After reboot I stated that some restored files aren't the same as stored ones. At least one file -- /usr/bin/X11/Xorg -- was devoid of SUID bit. Program tar from Slackware CD-ROM changes properties of some files.

So I removed new backup and restored the system once again using former backup and Slax.

Summary

Programs lilo and tar from Slackware CD-ROM work different than these programs from installed Slackware or Slax:

1. I can't reload LILO with the command lilo without error ``Inconsistent partition table'' using installed Slackware but I can do it using Slackware CD-ROM or Slax

2. I can't restore system from backup with tar and gzip using Slackware CD-ROM without error ``tar: incorrect length'' but I can do it using Slax or Slackware installed on the other partition.

3. I can't prepare valid backup using tar and gzip from Slackware CD-ROM (it changes at least one file properties) but I'm able to do it with Slax or Slackware installed on the other partition.

***

After six days of hard work I have still problems with ``Inconsistent partition table'' errors when I use Slackware on my second machine. Fortunately I resolved in the meantime other problems. I'm able now to install Windows on the first machine and I repaired partition tables on my three laptops. Moreover I gained some knowledge about Windows, partitioning the disk, and invalid tar command from Slackware CD-ROM.

I don't know still which LILO is valid: from Slackware CD-ROM and Slax or from installed Slackware. Two first are silent -- the third complains about ``Inconsistent partition table''. There are two possibilities: either LILO from installed Slackware discovers some bug and is better than LILO from Slackware CD-ROM and Slax or it lies and is worse.

I have no idea how I can avoid ``Inconsistent partition table'' errors. In fact I don't understand what ``CHS address in PT: 965:0:1 --> LBA (14590800)'' and ``LBA address in PT: 15502725 --> CHS (1025:75:1)'' mean. Maybe both these addresses lay farther and I should wipe out more than 1 MB of both partitions and hard disk or maybe everything is all right and LILO from installed Slackware finds non existing bug.

***

At the moment I reconciled that I need Slackware CD-ROM or Slax to reload LILO without errors on that machine.

Thank you guys for your assistance.