Compatibility problems with existing SSD using ide.device

[TorstenLang68]

Hi,

another project implementing an AIDE compatible IDE interface (PR77/A500_ACCEL_RAM_IDE) points to MHeinrichs' project for the ide.device driver. I did build a custom ROM with his driver and could partition the SSD (2x <2GB partitions and some larger partitions) with HDToolBox and install the system on the first two partitions but when booting only these first two partitions show up. I've tried an updated FastFilesystem and PFS on the other partitions with no success. At least PFS to my understanding should support large disks and large partitions.

To my understanding his driver supports LBA48 which is only required for large disk support (>128GiB), so I assumed that it should support large disks >4GB.

As I couldn't find a solution I continued searching and found that you also support the AIDE controller with your project and clearly state that you support large disks. I did build a custom ROM and tried your lide.device, but it doesn't recognize the existing data at all. Nevertheless it seems to detect the SSD just fine. I will do further tests with a fresh SSD to avoid losing my work but I have no idea what I did wrong here. I couldn't find any hint that your or MHeinrichs' driver do something special to the data like byte swapping so I assumed that your driver should be capable to access the pre-existing data.

I could use any help I can get...

Regards,

Torsten

[LIV2]

I made a mistake with some of the changes in 40.11 that broke compatibility, please try the attached version

I don't have the actual hardware to test though, I recommend making a backup of the drive first if possible
aide-lide-40.11-fix.zip

[TorstenLang68]

Thanks a lot. I hope I can test it soon. Currently nothing works as during the repeated swapping of components something broke and the new machine is currently dead. I think I will tear it down to the basic configuration and do extensive testing with the DiagROM and add the extensions board by board again.

Can you please tell in a few words what was fixes and what symtoms that fixed? When the machine was still alive I had the problem that with your driver writing wasn't possible. Depending on the HDToolBox I used I got some error -6 or 44 (I think) around about 10s after an attempt to save the disk and partition data.

[LIV2]

On the AIDE interface the registers are spaced 4 bytes apart, on RIPPLE, AT-Bus etc they're 512 bytes apart
The driver was selecting the wrong transfer routine when running on AIDE devices (it selected the one for at-bus etc) which caused it to read/write the first data word but then would read/write to other drive registers instead of the data register

The exception to this is that the IDENTIFY DRIVE function did not have this issue, so you would be able to detect the drive in HDToolbox for example but as soon as you try to write it would fail

The fix was to lock out this transfer mode completely if the hardware cannot support it as you can see here: 62015cd

[TorstenLang68]

OK, that explains the errors I had of not detecting the existing disk and when accessing the new disk for writing the partition data. As the transfer routine is disabled at compile time that shall work fine with the A500_ACCEL_RAM_IDE project from Paul.

[TorstenLang68]

Just a short feedback - I built a custom ROM with your new driver and it now works (with the existing data on the SSD). Furthermore, I see the full disk and could install a large PFS3 partition beyond the 4GB limit. Copying the existing AmigaOS installation over and running the PFS doctor afterwards did not show any problems, so all in all it seems to work fine.

[LIV2]

Good to hear :)

Out of interest I took a look at the ide.device code to see why you might be having issues and noticed a couple of things

• It's support of TD64 seems correct, the driver itself should be able to address up to 2TB however,
• If I understand correctly the mounter does not load Filesystems from the RDB - meaning you're stuck with the version of FFS in your kickstart (if 3.1, it is limited to 4GB as it only supports 32-bit trackdisk commands)
• The Direct SCSI emulation only supports SCSI READ (6) - the 6 byte commands can only address up to 4GB and it doesn't emulate the scsi write command at all

It's possible that with Kickstart 3.2 you could use large drives (with FFS) as it may support TD64 but I'm not certain

[TorstenLang68]

Thanks for your analysis. I have licenses for OS 3.5, 3.9, but my 3.2 disk did not arrive before I installed the system.

The restriction of the ide.device boot process regarding the RDB could explain a lot of my trouble. I have licensed physical Cloanto 3.1 ROMs which I used this as a basis for my custom ROM. When partitioning the SSD the RDB received an updated FFS (I would have to look up the exact version).

As the official FFS to my knowledge has a 4GB limitation per partition even in the modern 3.x OS versions I went for PFS3 for the large data partition. From what I've read PFS seems to use direct SCSI.

OS 3.5 and 3.9 require at least a 68020 although some parts are still plain 68000, so I didn't try to install one of these. I don't know how memory hungry 3.2 is, so I didn't try it out yet on this machine after I finally had a working installation.

Conclusion: When installing OS 3.1 I was stuck with the outdated FFS from the harddisk installation (which I didn't update), the updated version in the RDB was ignored. PFS3 uses direct SCSI which only is supported with the short commands and thus limited to 4GB harddisk size by ide.device.

My decision to try out A500_ACCEL_RAM_IDE was based on the fact that it is relatively cheap to build and using an 68000 should be as compatible as possible when it comes to older games and demos what this machine is intended for.

Maybe I will swap components in future, especially as the accelerator board and a direct RGBtoHDMI do not fit very well together in the Denise board.

[CCKMik]

I confirm,
version 40.9 works fine,
version 40.11 does not work,
version 40.11 fix (May 21, 2026) works fine

[LIV2]

Thanks for confirming, I will release 40.12 to include the fix after doing more compatibility testing