Read GPT hard disk on Windows XP (solved)

[Cixert]

On 8/6/2025 at 2:29 AM, ages2001 said:

It seems Vista SP1 32-bit partmgr.sys and disk.sys should be backported for Win2000.

XP32 ones has 2 TB limit and can corrupt data above 1 TB written. Buıt Vista SP1 32-bit ones don't have 2 TB limit and no corruption (not 100% sure but I didn't hear any corruption from anyone).
 

Good info, if you manage to port it and get it working, please report.

[Cixert]

On 8/5/2025 at 1:08 AM, user57 said:

so just using the right paragon driver solved it up ? thats good news that the /pae question related to that problem would be finally closed

@Andalu is referring to file corruption on his system due to the PAE driver and the GPT loader.
The correct version of the GPT loader refers to the fact that without the PAE driver, other versions of the Paragon loader also cause file corruption above 2 TiB.

At this time, there is no known PAE driver that will not cause corruption above 2 TiB with any version of Paragon GPT Loader.

Edited 5 hours ago by Cixert

[user57]

from my understanding (is not absolute - or maybe someone just correct me up)

it has 5 main things related to the harddrive
cylinder cluster track sector and head

those get translated by the firmware of the harddrive to a LBA address
the LBA registers seems 48 bit in total what makes sence (since its 3 times 16 bit = 48 = 48bit LBA) (LBAlo, LBAmid, LBAhigh)
those are a combination, the real thing not neccesary always needs 32 wires, it actually could use less wires to transfer the data either
the speed rather would be limited to what these wires can progress by frequency
a USB stick for example dont have so many wires (even tho running with a 32 bit os or maybe 64 bit os)
https://techdiagrammer.com/wp-content/images/usb_a_wire_diagram_hihdn.jpg
but we getting the point - we not neccesary need always 2, 32 or 64 bit wires

this formula often came up:
LBA = (C × HPC + H) × SPT + (S − 1)

C, H and S are the cylinder number, the head number, and the sector number
LBA is the logical block address
HPC is the maximum number of heads per cylinder
SPT is the maximum number of sectors per track 

lets say the sector is 512 bytes then you can multiply the indicator by 512 with 4096(4k) by 4096
that would also not be limited if you have a track containing lets say 1000 sectors
the next track then would point into a next 1000 sector track
if then the sector 0 would have 512 bytes 
then the indicator 0 would point into track 0 and sector 0 

the same things would be the case if its a head a cluster or a cylinder 

this cylinder sector head would actually be translated by the firmware to a pysical position - that would be like its job
it also can do this software side - it actually dont need to have these things - as long they are transtated correctly by the firmware of the harddrive

for a drive it seems that you have to set these things before you output your data stream (outsb (byte/8 bit), outsw (word/16 bit), outsd(dword/32bit))
if you see a code like "move me 4 mb of ram" - no its not doing that that way
for memory there are useally 32 bit moves (this is a small piece of data) - this then is repeated as soon 4 mb are reached
for the harddrive it seems that you actually dont have just an "offset" you have to set where first - then you write the output again either 8 bit 16 bit or 32 bit wise

the reason 8 16 32 bit maybe 64 bit or even 128 bit (sse) are not slower is because the cpu/hardware can detect this progress and translate this to a faster progress (aka faster speed)

what xp (or windows useally give us) at usermode is the overlappend structure - that one is made of 64 bits 
it probaly gets translated into the right progress that the harddrive is needing ...
here the logic would play a role - maybe they limited it to a dword (32 bit) * 512 (common sector size) = ~ 2,19 TB
that with the 4k sectors are not a bad idea because then the data just increases 8 times 2,2 * 8 = ~ 17,52 TB

it would be nice to have insight into that paragon driver, there was also a 3 tb+ unlocker from asrock, the information might be out there somewhere