Re: can't get Windows to boot on secondary disk

[Andrej Mikus]

I am sorry to update you again, but I believe that the bug is confirmed.
According to FAQ answer 1 I am planning to try latest version, because
currently used copy is 0.92 from RedHat distribution.

Unfortunately, I cannot browse current bugs via BTS,
http://bugcomm.enbug.org/?project=grub&mode=project gives an error
message: Connection refused - "connect(2)" (Errno::ECONNREFUSED


Full description of the problem in case it is not known yet.

System is installed with 3 hard drives as follows:
1. PM, 40GB, Linux only
2. PS, 8GB, Windows 98 only, 3 FAT partitions.
3. SS, 4GB, 2 FAT partitions.

Secondary master is an ATAPI CD Rom, not sure if it matters.
I also upgraded BIOS to the recent release from Intel, while
troubleshooting this problem.

Problem with booting Windows happens every time Grub is used,
regardless of the system booted. It appears to write something to the
secondary disk partition table that prevents Windows to access the
data in seconary partition correctly. Linux has no problems, most likely
because it does not rely on CHS data in partition table.

As mentioned already, there is no problem if disk 3 is removed from
system. I reformatted it and made it Win95 bootable and it boots when
remapped from BIOS.

Grub can now boot up to DOS prompt of Windows of both disk 2 and 3.
Problem starts if trying to access secondary partition of disk 3 from
either of them, command dir shows just garbage (seems like accessing
some subdirectory of the same) and just hangs at the end trying to
display disk free space. Perhaps one copy of FAT is broken?

If disk 2 is set as boot disk in BIOS, Windows GUI does not start,
because of a problem with secondary partition of disk 3.

If disk 3 is made primary boot source, it boots just fine, and
as a side effect it fixes the issue of this disk, because from now
on it is possible to set disk 2 or 3 as primary boot source, and no
problem occurs, until Grub is used again.

To conlude, IMO the above results confirm that Grub damages the disk
structure of disk 3 regardless of the disk used to load system. The
damage gets corrected somehow if the Win95 MBR of this disk is executed.
I understand that it is hard to believe, as I do not expect that a boot
loader would need to perform disk write access.

I apologize, but I do not have enough resources to fix the problem
myself, I did no do any coding on this low level so far. The
troubleshooting above can hopefully help some experienced maintainer
to confirm the root cause. If the problem remains in the version 0.93
I will need to look after alternatives, like loadlin. If useful, I can
contribute with further troubleshooting, but will need an advise.

Regards
Andrej



On Thu, 15.May.03 10:25:29 +0200, Andrej Mikus wrote:
> Team,
> 
> Just a small udpate, the problem is fixed and has been isolated to
> damaged partition table on the third disk. Linux reported problems
> there and apparently grub used incorrect geometry, but Windows did not
> like the correction. After the disk removed, grub works as expected.
> 
> I plan to reformat the disk before adding back to system.
> 
> Regards
> Andrej
> 
> 
> 
> On Mon, 05.May.03 08:01:09 +0900, Yoshinori K. Okuji wrote:
> > At Sun, 4 May 2003 20:27:34 +0200,
> > Andrej Mikus wrote:
> > > Perhaps a bug in the map command?
> > 
> > I think you guess is correct. I don't know what is wrong in the
> > command, but according to bug reports submitted earlier, it is obvious
> > that the command doesn't work perfectly. I appreciate if you could fix
> > the problem.
> > 
> > Thanks,
> > Okuji