Read GPT hard disk on Windows XP (solved)

[ruthan]

11 hours ago, Cixert said:

With PAE, the limit without data corruption is 2 TiB.
Without PAE, the theoretical limit is 256 TiB for SATA.

  Ok, i used it without PAE back in the day, for sure.. But there is not just 1 PAE patch, but multiple flavours of it.. some may work with PAE and GPT loader fine.
Also i wrote with Win2000, i got crash after GPT loader - not 8.x.x install, even without GPT disk connected.
  Im not aware of PAE patch for Windows 2000, it would be handy, i case of Win2000 im using only /PAE boot.ini switch for Advanced server version because DataCenter is problematic and advnaced server license is limited to 8 GB, datacenter is 32 GB.. PAE patches for WinXP+ are 64 GB or 128 GB.

 

Quote

I don't recommend following the final link you posted to the Chinese solutions from 2011. That link recommends following all the incorrect steps we've already ruled out, which limit GPT to 2 TiB. Following those steps leads to data corruption starting at 2 TiB.

  Thanks for testing, its allway good to know and test, dead ends too.

Quote

I'm pretty sure that when I did my tests, I went well over 2TB on my 3TB disk, with PAE enabled as it always has been, and there were no issues.

Ok there seems to be and the way after all.. 
There is more than 1 variable.. Used PAE patch there is multiple flavours of it. Chipset / Sata controller drivers, could matter were already know that Intel AHCI is not go.. Its make boot bluescreen on WIndows 2000 even with extended kernel and without PAE and GPT loader 10 - not 2015, even with not GPT disk connected.  There could be also disk cluster size dependency.. if there is hardcoded limti for 2TB and number of clusters, maybe using bigger cluster size could bypas it, and especially some external 2TB HDD could came with some clever preformating..

Back in the day, i was using Dybay version of PAE patch i my memory serves, or version from Rayer site..
https://web.archive.org/web/20191108052121/https://ryanvm.net/forum/viewtopic.php?f=7&t=10547

Edited yesterday at 12:03 PM by ruthan

[Cixert]

11 hours ago, Dave-H said:

I'm pretty sure that when I did my tests, I went well over 2TB on my 3TB disk, with PAE enabled as it always has been, and there were no issues.

The PAE limit was commented by @Andalu on January 18, 2025
https://msfn.org/board/topic/181911-read-gpt-hard-disk-on-windows-xp-solved/page/27/#findComment-1276911
I ran the tests with @Dibya Path PAE and confirmed the results in the following posts. If you know of a PAE patch that works with GPT, please comment.

 

[Cixert]

9 hours ago, ruthan said:

There could be also disk cluster size dependency.. if there is hardcoded limti for 2TB and number of clusters, maybe using bigger cluster size could bypas it, and especially some external 2TB HDD could came with some clever preformating..

Increasing the cluster size does not allow you to exceed the 2 TiB limit in either GPT or MBR.
In MBR, to exceed this limit up to 16 TiB, you must increase the logical sector size from 512 bytes to 4096 bytes.
This requires hardware that performs the conversion in its firmware, which is not currently possible with IDE mode, much less with the Paragon GPT Loader.

Here is a list of devices that support this with MBR.
https://msfn.org/board/topic/186645-devices-list-compatibles-mbr-hard-disk-2tb

[user57]

someone should correct me if im wrong

but PAE is something for memory managment, is actually involved ? for a disc it dont have a 32 bit offset so directly, rather it makes use of 3 register of 16 bit size

 

LBAlo - 16-bit

LBAmid - 16-bit

LBAhi - 16-bit

= LBA48

https://wiki.osdev.org/ATA_PIO_Mode#Registers

(the 8 bit are longly outdated - they are now always 16 bit) 

 

after these are set (LBAlo,LBAmid,LBAhi) it just use a I/O code - it dont use a 32 bit offset - it rather tells the disc where to write its data over these 3 filled registers

 

its a bit like with memory paging, the old norm where 4k pages what are normed with a 32 bit PTE entry (and maybe a PDE)

-> then PSE (page size extension (similiar to PAE)) apeared and made these 4 k pages to 4 MB pages 

thus for 4k you need less bits in 32 bit norm (20 bits to represent the highest 4 GB page) (4 k * 20 bit = 4 GB) 

for a 4 MB page it would be only 10 bit´s (4 mb * 1024 = 4 GB) to represent the highest 4 GB page, the rest of 4 MB pages would point to the upper pages above 4 GB 

i heared for PAE its only the half, they said its because PAE has the NX bit that means -1 bit = 2 MB pages, this is the result they often describe (also they say the 4k pages are ignored, aka PSE is always on)

https://en.wikipedia.org/wiki/Physical_Address_Extension#/media/File:X86_Paging_PAE_2M.svg

i do not know it exactly for that PAE mode, but even if it would be like that i could programm it like that

 

 

the thing with the sector-size for a HDD/SSD/drive is very similiar if you have a 4 k sector instead of a 512 sector you have 8 times as much data

 

it might not be the PAE extension itself, it might be the licence check like starter, home, pro - but that is only a guess

 

just trying to make things better, if someone have knowledge let me know i also can improve myself

[ruthan]

51 minutes ago, Cixert said:

Increasing the cluster size does not allow you to exceed the 2 TiB limit in either GPT or MBR.
In MBR, to exceed this limit up to 16 TiB, you must increase the logical sector size from 512 bytes to 4096 bytes.
This requires hardware that performs the conversion in its firmware, which is not currently possible with IDE mode, much less with the Paragon GPT Loader.

   Yeah sector is right name for it not cluster, these terms where allways a bit fuzzy for me.. i allways after some time forget where is difference.
   So this is another dead end.
   How these 2 TB+ unlockers from Gigabyte and other did the trick, its not definitelly through changing firmware and was for MBR disks.

   I have found something else if are in the mood to continue research:
https://archive.org/details/paragon-gpt-loader-10.0.16.12944-8.0.1.2-driver-update
   There are 2 interesting things,
   1) Reg file fix for xp with some magic values, without any description.
   2) GPT loader v8.x - but its described as unofficial, maybe someone somewhere did some fixing.. 
   PAE is not mentioned.

[Dave-H]

20 hours ago, Cixert said:

The PAE limit was commented by @Andalu on January 18, 2025
https://msfn.org/board/topic/181911-read-gpt-hard-disk-on-windows-xp-solved/page/27/#findComment-1276911
I ran the tests with @Dibya Path PAE and confirmed the results in the following posts. If you know of a PAE patch that works with GPT, please comment.

 

We may be talking about different things here regarding PAE.
Standard PAE is enabled on my machine, and always has been.
Until recently though, I had not experimented with 'PAE patches' to extend the amount of RAM available on my 32-bit system.
I do now have one of those available as an option on boot, and it is possible that my access to the 3TB disk will fail if I use that, if I go over 2TB on the disk.
I haven't actually tried that, but certainly with the standard PAE configuration, I went over 2TB on my tests without any apparent problems.

[user57]

i would suspect that paragon driver is a filter driver

filter drivers are something between the OS and the HARDWARE
it can perform the same actions as the OS can do with the hardware

there are those IRP requests where filter drivers often would be

if the controlment is taken from that spot the paragon driver has the control

so putting things right, the paragon driver actually could handle the harddrive instead of the OS (in this case xp)
then interact with the OS and show it as if the harddrive can control more then 2 TB of discspace

this might have a problem but, it needs that paragon driver in that case 

to figure out what the inner XP code is doing is more of work, that require a lot of debugging - i think we might have candidates to do that