Re: scsi and windows boot problem

[Szakacsits Szabolcs]

On Sun, 25 Jul 2004, my wrote:

> The non-coder, non-geek take is as follows, with all deference to everyone
> who has written on the topic.

You indeed don't understand the issue and unfortunately wrote a lot of
confusing things. You also wrote right things but partly they were
irrelevant. You involved several things that has nothing or not much to do
with the problem. The reality is much-much simpler than most think (John
Gilmore, Alan Cox, Andries Brouwer seem to understand the best). Thanks
for your attempt anyway. I still maintain the real problem is not
technical for a while because the software solution/fix is very easy.

Given a perfectly working, prepartitioned computer. When partitioners try
to use the buggy [lib]parted on the buggy 2.6 kernel (I tell later why
both are buggy) then before anything useful could be done, parted often
asks to choose from and only from,

  1) overwrite partition table (including Windows partition entries
     parted has _NOTHING_ to do)

  2) exit

since distros and users can't "exit" because that would mean a total
failure, moreover parted wording about the "fixing" suggests it's smart,
safe to do ("easily fixable" what's actually also not true) and doing
something good for you hence people go on and "fix", overwrite the CHS
values in the partition table to be able to read it at all. This is due to
parted and not by some technical limitation parted should enforce. So the
damage is made.

Thus even if you don't do any partitioning, just read it, you already had
corrupted the partition table. This explains why so many people have this
problem even if they didn't partition anything, just read the partition
table (e.g. upgrade, prepartitioned scenarios).

What this issue has to do with bootloaders, Windows bootloader, cylinders,
CHS, BIOS, LBA, laptops, etc? Nothing.

What parted does, is like if Linux filesystems would randomly corrupt
FAT32 or NTFS on a different partition.

Why parted is buggy? It mustn't use the kernel the way it currently use.
Andrew was told several times by kernel developers over the last 8 months,
still parted uses the function it mustn't use thus it still corrupts users
data. Unfortunately current parted CVS has even more non-sense code.

Why the kernel is buggy? Because it incorrectly broke the semantic of a
critical kernel ioctl two years ago what tools use. If it returned error
("I don't know the geometry") then that would have been pretty OK.

But here is a question for those who disagree with all of this: 

  Why parted needs to change the "number of heads" byte in the Windows
  partition entry (that's why no boot anymore) to be able to read the
  partition table?

It's totally non-sense.

        Szaka