I'm now almost certain that I'm seeing corruption when trying to access data beyond approximately 8GB.

My last test was to partition a 16GB card into:

512MB OS
7.3GB-ish Data 1
7.3GB-ish Data 2

The OS partition is fine, as is everything on the 1st data partition, but there's almost no directories on the 2nd data partition that I can access: they either come back empty, with gibberish filenames or lock up the computer when accessing them.

I also tried a 2.5" IDE drive of ~60GB capacity and had similar issues to my 16GB and 64GB CF cards - hit and miss after more than 8GB or so.

Of course, if I pull the card out and put it in my card reader all the content is absolutely fine.

It just so happens that I'm tesing an ISA 3Com card that I intend to also use as XTIDE carrier and I refreshed my memory by re-reading some of the docs. The XTIDE pages states that "MS-DOS 7.x (Windows 9x) or FreeDOS is required to access more than 8.4 GB" - so your corruption probably happens when you try to access the higher sector addresses via older 16-bit DOS.

Do you create the >8.4GB partitions on the retro PC or on your modern PC ?

The address translation algorithm used by your card reader may not match up to the one used by the XTIDE .

See this excerpt from https://code.google.com/archive/p/xtideuniver … kis/Manual.wiki :

1Drive configuration information is displayed for XTIDE Universal BIOS controlled drives. Information includes:
2Addressing (Addr.)
3This can be L-CHS, P-CHS, LBA28 or LBA48. CHS addressing is the old type of addressing method where cylinder, head and sector numbers will be handled separately. Original PC BIOS functions are designed for CHS addressing with maximum hard disk size being 7.8 GiB (8.4 GB). LBA addressing is modern addressing method where every sector has its own address. There are no cylinders or heads anymore. Enhanced BIOS functions were introduced for LBA drives but they are not supported before Windows 95 (DOS 7). These EBIOS function are not yet supported by XTIDE Universal BIOS. CHS address must be translated to LBA address when using old CHS BIOS functions with LBA addressing.
4
5L-CHS (known as NORMAL on many old BIOSes) is used for drives <= 504 MiB that can accept the CHS parameters without translation. That makes L-CHS the fastest addressing method.
6P-CHS (known as LARGE on many old BIOSes) is used for drives from 504 MiB up to 7.8 GiB. This is a bit slower than L-CHS since simple translation is required to make BIOS L-CHS parameters compatible with IDE P-CHS parameters.
7LBA28 (28-bit address) allows drive sizes up to 128 GiB (137 GB) but maximum accessible size is 7.8 GiB when old BIOS functions are used. L-CHS to LBA translation is more complex and slower than L-CHS to P-CHS conversion.
8LBA48 (48-bit address) work just like LBA28 but with 20 more address bits. This makes possible to use drives with over 128 GiB capacity.

Also worth reading if you are using an old version of XTIDE (excerpt from http://www.xtideuniversalbios.org/ ) :

1Important if you are upgrading from any previous XTIDE Universal BIOS version ¶
2The v2.0.0 beta 2 and later versions, like most other BIOSes, adheres to the Phoenix Enhanced Disk Drive Specification. The older v1.x.x versions and v2.0.0 beta 1 do NOT - they may generate different L-CHS parameters for many drives. If you move a drive handled by a v1.x.x or v2.0.0 beta 1 BIOS to another system or upgrade to v2.x.x you risk data corruption if different L-CHS parameters are used.
3
4IMPORTANT! This means that, after upgrading to XTIDE Universal BIOS v2.0.0 beta 2 or later, you need to re-create and format any partitions on drives handled by this BIOS.

I'm using r605 (and only have with this system).

Partitions were created and formatted on the XTIDE system itself, though I also tried the opposite; creating them on a modern Linux system first. Same behaviour both ways.

Multiple partitions or, with PC-DOS 7.1, one big partition of up to 60gb, same behaviour observed; in that beyond a certain point of data being copied to th drive, subsequent data becomes inaccessible.

I did read that there was a bug in the CHS to LBA translation found in XTide recently (and fixed in r606), so will give that a try, but I'm not sure as described whether that's the specific case I'm hitting.

Also, that "information shown on boot menu" description doesn't match what I see - all of these drives just show "LBA" addressing, not LBA28 or LBA48.

Have flashed the latest XTIDE (r606), and recreated the partitions on the card:

All partitions check out normally, as they did previously; correct size/free-space etc:

The XTIDE biosdrvs.com tool reports sensible information for the card:

I'm now on copying a couple of GB's of data to the last partition, which start just before the 8GB boundary - so if it's going to have the same behaviour as before, this should trigger it.

I am always using the old PowerQuest/Norton Partition Magic for DOS to create partitions on the target system itself - never failed me in all these years.
Could be an option to try - it needs a 386+ to work.

It's still corrupting/not accessing content beyond 8GB.

Viewing the root directory of the third partition, which occupies from ~7.8-16GB looks fine (there should be one folder there, and there is):

But trying to access or list content from within that folder, and boom:

Inaccessible and corrupt directory entries, identical behaviour to before. The card and content of that partition still reads absolutely perfect if I take it back out and put it in the reader again.

keropi wrote on 2021-05-20, 08:59:

I am always using the old PowerQuest/Norton Partition Magic for DOS to create partitions on the target system itself - never failed me in all these years.
Could be an option to try - it needs a 386+ to work.

It's got to be worth a try. I'm running out of ideas.

Since you are running out of ideas you can also try to partition, format and boot the win9x ms-dos instead of pc-dos.

konc wrote on 2021-05-20, 10:46:

Since you are running out of ideas you can also try to partition, format and boot the win9x ms-dos instead of pc-dos.

I have - the same behaviour is observed with both the Win98SE / Dos 7.1 as well as with PC-DOS 7.1 🙁

megatron-uk wrote on 2021-05-20, 10:49:

konc wrote on 2021-05-20, 10:46:

Since you are running out of ideas you can also try to partition, format and boot the win9x ms-dos instead of pc-dos.

I have - the same behaviour is observed with both the Win98SE / Dos 7.1 as well as with PC-DOS 7.1 🙁

Yes sorry I just noticed it, the pc-dos was for trying one big partition

Tried re-partitioning with Partition Magic 4 and I get the same result with the partition which extends beyond 8GB:

damn.... I have no other ideas...

I suppose I'm going to have to give a DDO a try and see how messed up the geometry is when I try to read it from Linux. If I can still copy stuff to/from it, then I'll live with the overlay software, but it's not ideal.

megatron-uk wrote on 2021-05-20, 09:21:

Inaccessible and corrupt directory entries, identical behaviour to before. The card and content of that partition still reads absolutely perfect if I take it back out and put it in the reader again.

Am I interpreting this correctly?
The data you copied to the card that isn't accesible on the old pc, is readable on your card reader ?

Yes, that is correct.

After having partitioned, formatted and sys-ed the card in the 386 with XTIDE I took it back out.

With the card in my Linux PC, I copy ~3GB of data to the E: partition which runs from ~7.8 - 16GB. This is in addition to content I copy to C: and D: in the same manner.

With the card then back in the 386 with XTIDE, all content on C: and D: is fine. The majority of the data/directories on E: is inaccessible or corrupt.

Put card back in Linux PC, everything on that third partition is absolutely perfect.

Are you using AT or 386 build? If you use 386 build, try AT build instead.

I've tried the following:

r605 ide_atl.bin (as borrowed from my 286 with a 4GB CF card)
r605 ide_386l.bin
t606 ide_386l.bin

All have the same symptoms - content is inaccessible/appears corrupt beyond 8GB, regardless of the tool used to partition or format the drives.