Thank you very much for your help. Since then, there have been a
couple of developments.
I have used diskedit.exe (part of Norton Utilities) to create an
extended partition entry covering all the remaining space (up to the
1024 cylynder limitation). My Boot Magic can now see my linux
partition and I can boot from it. The problem is that Partition Magic
now complains. The reason why it complains is that it says that the
partitions within the extended partition exceed the upper limit of the
extended partition itself (again, this is because of the 1024 cylinder
limit). I feel like something was done to get around this when the
extended partition was created the first time, but since my repair,
that has kind of gone out of synch.
-> I don't know enough about extended partitions, but can anyone help
me with this and tell me what I need to fix so that it doesn't look
like the partitions within the extended partition are going beyond the
upper limit of the extended partition itself?
Also, since recreating the extended partition, I was able to run
partinfo (part of the Partition Magic suite) and obtain the exact
location of the various sectors. Here is the data:
============================================================================================
Partition Information for Disk 1: 13481.0 Megabytes
Volume PartitionType Status Size MB PartSect #
StartSect TotalSects UsedSects FreeSects
============================================================================================
*:EMPTY Hidden FAT16B Pri 2037.6 0 0 63
4,173,057 4,173,057 0
C:WIN98_HDA1 FAT32 Pri,Boot 3802.1 0 1
4,173,183 7,786,737 3,985,158 3,801,579
*:TMP98_HDA1 Hidden FAT32 Pri 1388 0 2
11,959,920 2,842,560 1,827,972 1,014,588
Extended Pri 332.2 0 3 14,802,480
680,400 680,400 0
EPBR Log 4732.4 None -- 14,802,480
9,691,920 9,691,920 0
Warning #113: EPBR partition starting at 14802480 overlaps extended
partition.
Linux Ext2 Log 4732.4 14802480 0
14,802,543 9,691,857 3,135,047 6,556,810
Error #113: Logical starting at 14802543 overlaps extended partition.
Free Space Pri 5921 None -- 15,482,880
12,126,240 0 12,126,240
EPBR Log 132.9 14802480 1 24,494,400
272,160 272,160 0
Warning #113: EPBR partition starting at 24494400 overlaps extended
partition.
Linux Swap Log 132.9 24494400 0
24,494,463 272,097 272,097 0
Error #113: Logical starting at 24494463 overlaps extended partition.
EPBR Log 1388 24494400 1 24,766,560
2,842,560 2,842,560 0
Warning #113: EPBR partition starting at 24766560 overlaps extended
partition.
D:WIN98_HDA2 FAT16B Log 1387.9 24766560 0
24,766,623 2,842,497 351,735 2,490,762
Error #113: Logical starting at 24766623 overlaps extended partition.
QEUSTION -> I guess I have the sector numbers to rebuild the table
with linux fdisk now. Are we absolutely sure that linux fdisk is NOT
like dos fdisk and it only touches the partition tables, leaving the
boot sectors alone?
PS -> I cannot dd to another hard disk. I forgot to mention that this
is a laptop (yes, with a 14Gb hard disk).
>p_tux_news (p_tux_news) schrieb in Nachricht
>>My disk structure was as follows (approximate sizes):
>> 14Gb hard disk:
>> 2.0Gb FAT16 partition (NT4 Server)
>> 3.8Gb FAT32 partition (Windows 98 SE)
>> 1.4Gb FAT32 partition (Windows 98 SE)
>> 5.7Gb externded partition:
>> 3.8Gb linux ext2 partition
>> 128Mb linux swap partition
>> 1.4Gb logical d: FAT16 partition
>Linux has a *powerful* fdisk command where you can do nearly everything.
>Another benefit of Linux's "fdisk" is the fact that it modifies only the
>partition table sector, *not* the content itself. Microsft's FDISK.EXE for
>example always overwrites the first sector so remaining data are lost. In
>Linux for example, I also often have "hiddened"/"disabled" partition (for
>example installing a Windows 98 which should not fiddle on an existing
>MS-DOS 6.22 environment or installing IBM OS/2 on a removable medium to
>influence to drive letter assignment during installation). It's also
>possible to delete any (primary only!) entries and recreating them later. If
>you use exactly the same number of cylinders and starting cylinder itself
>with the same fstype itself (for example 0x6 for DOS FAT 16bit), you will
>find your filesystem's content without damages. :-)
>Prerequisites for a successful recovery are to know the *exact* layout: What
>was primary partition? What was on a logical drive? How many cylinders
>(exact numbers!) had every partition. A good tip for preventation: Make a
>"p" (Print partition table) inside "fdisk", you will get a valueful
>documentation for printing out and storing on a safe place.
>Because you only known the approximate sizes: You only can try working by
>"trial and error": fsck will savely report "bad superblock" when you choosed
>the wrong start cylinder. Note that you may run fsck in "read only" mode!
>When you found the superblock, then you will get an appropriate message that
>the file system size differs so try different sizes until fsck does not
>longer shows any messages.
>I know that in such a process, it's easy to even destroy more than recover.
>So if you have a secound harddisk drive where there's at least so much free
>space as the size of your destroyed disk drive, then you can make a backup
>with
>dd if=/dev/hda of=/mnt/fullsave/fullsave_disk_c.bin
>Do this *before* every kind of recovery attempt! This saves *every* sector,
>so you can restore to the state like at the beginning to get a further
>chance for restoring when something went wrong by using "dd
>if=/mnt/fullsave/fullsave_disk_c.dat of=/dev/hda".
> Andreas
linux: the revenge of the nerds!!
by