OSDev.org

ATA-5 LBA limits hard drive size to 28**2 * 512 bytes (28bit block size).

I found the ATA-6 specification, which extends block numbers to 48bit. Nice (apart from it being another 500-page document to add to the reference library).

But one question remains unanswered. The partition table in the MBR doesn't have more than 32bit for LBA addressing reserved...

I assume that they changed the specs for the MBR partition tables in the process of hammering out ATA-6, but couldn't find any documentation on that. Does anyone know about this?

Hi,

Solar wrote:I assume that they changed the specs for the MBR partition tables in the process of hammering out ATA-6, but couldn't find any documentation on that. Does anyone know about this?

AFAIK the partition table and MBR isn't covered by ATA standards (which only really covers how the controller and device interact). For partitions over 2 terabytes you'd need to look into EFI.

See:
http://www.intel.com/technology/efi/

And for the standard itself:
http://www.intel.com/technology/efi/agree.htm

Note: It's the second in the list "Version 1.10" not the "EFI Specification Update" at the top (which you might want to grab anyway).

Specifically, see section 11.2.2 "Partition Discovery". Please remember that I haven't had much time to look through the (6.82 MB) EFI specification, so there might be better places to look in this standard.


Cheers,

Brendan

Gods...

Could it be that they are deliberately making interfaces - and especially their descriptions - more complicated and obfuscated with every iteration, to keep upcoming OS' out?

::)

That is an excellent question. I found this on Microsoft's site: http://www.microsoft.com/resources/docu ... och32.mspx

My understanding of this is the file system itself handles anything above 32-bit LBA.

If I'm doing my math right, 2 Terabytes is max with 32-bits, and 144,115,188,075,855,872 bytes for 48-bit. The maximum drives I can find at the moment are 1.6 TB drives. I would imagine that this could be an issue within the next 18 months...

But this was a quick search...

Solar wrote: Gods...

Could it be that they are deliberately making interfaces - and especially their descriptions - more complicated and obfuscated with every iteration, to keep upcoming OS' out?

::)

Well, if you are the implementor it is possible to cicumvent competition by writing the specifications.

Hi,

Solar wrote:Could it be that they are deliberately making interfaces - and especially their descriptions - more complicated and obfuscated with every iteration, to keep upcoming OS' out?

This has been my opinion for a while now. What I find disturbing is that it's Intel this time, without too much involvement from Microsoft.

I think Intel's intention is for EFI to cover everything up to and including ACPI and make all 80x86 OSs comply with EFI, so that eventually these OSs can be ported easily to EFI on Itanium. In this way they hope to establish an "upgrade" path that makes 80x86 obsolete and leaves AMD with a much reduced market share. I'm only guessing though....

I guess the next question is: What is Microsoft's UGA (Universal Graphics Adapter) initiative, and UGA specification? Dare I hope... (on second thought I better not).

[EDIT]
I found a little UGA information to make things worse:

http://www.freshpatents.com/Universal-g ... 160191.php

I also found a good 64-bit, BIOS, EFI, UGA, etc overview that I though people might find interesting:

http://bisnowden.com/bios_and_64bit.htm
[/EDIT]


Cheers,

Brendan

Here's another link on 48-bits, enjoy!

http://www.48bitlba.com/faq.htm