🚀 go-pugleaf

Svend Olaf Mikkelsen <svolaf@inet.uni2.dk> wrote in
message news:3979e6a0.798510@news.inet.tele.dk...
> Rod Speed <rod_speed@yahoo.com> wrote

>>> This is a really nasty bug.

>> You appear to be confusing the undoubted problem with changes
>> to the mapping between the CHS values and the LBA block numbers
>> that can certainly vary with the OS level mapping, and what the bios
>> does at boot time with an AUTO drive type, what was being discussed.

> The OS has nothing to do with this.

Wrong. Not all OSs even use CHS, let alone map them to
block numbers the same way. It cant just be the bios, most
obviously with drive overlays which get around bios limitations.

> Let me give you an example.

> Award BIOS with extended interrupt 13h (disk) support.

Thats just one way an OS can use a hard drive.

> A 9671 MB IDE harddisk. The disk is set to User/LBA,

Soorree, we were specifically discussing the
use of the AUTO drive type, not that mode.

> and the disk has a BIOS translation of
> 1232 cylinders, 255 heads, 63 sectors.

And thats not even getting the geometry details
out of the first physical sector on the hard drive
anyway, thats getting it from the drive type values.

> With this translation 9664 MB are useable.

> A 9664 MB partition is created. The partition table
> is correct, and assumes a 255 heads translation.

There aint even just one format of partition
table. That also varys with the OS.

And as Tpsa said, even if you choose to use a custom OS
with its own filesystem, the Award bios manages fine, so it
cant be getting that geometry data out of the first physical
sector on the drive when the drive type AUTO is used.

And the bios certainly cant be getting it out of the first physical
sector of the drive when that sector just contains zeros.

> The disk has 19,807,200 sectors. We look at LBA
> sector 10,000,000 (numbered from 0). This will be

> cylinder 622, head 120, sector 11

> Then we reboot, and enter BIOS setup. LBA mode is changed
> to NORMAL mode. A translation of 19650 cylinders, 16 heads,
> 63 sectors is displayed. We save the change and boot.

> Then we request the disk geometry by calling BIOS interrupt 13h.
> The answer is 1232 cylinders, 255 heads, 63 sectors.

> Then we call BIOS interrupt 13h, function 42h, and
> request LBA sector 10,000,000. A sector is returned.

> Everything looks OK, but it is definitely
> not sector 10,000,000 that is returned.

No one has ever disputed that that mapping does
vary with the  drive type for some drive type changes.

What was being discussed was whether a drive with
an AUTO drive type can suddenly start using a different
mapping, just due to a change in the contents of the first
physical sector of the drive, and that different mapping
result in the drive audibly recalibrating as if finds bad sectors
and is very slow to copy existing data for the same reason.

Yes, when the drive type is changed, the mapping between
CHS and logical block numbers can change, but that just
makes the data structure on the drive all wrong and you
can 'lose' partitions and data, but not get read errors.

> Then we use interrupt 13h, function 02h,
> to request cylinder 622, head 120, sector 11.

> Again: Everything looks OK, but it is definitely
> not sector 10,000,000 that is returned.

> Please note that this is only one of the
> Award bugs. The one that I called nasty.

Got absolutely NOTHING to do with what was being discussed,
whether the bios, with an AUTO drive type, gets the geometry
detail from the first physical sector on the drive and that data
going bad will produce bad sectors. You are just talking about the
mapping changing. No news to anyone who knows anything about it.