Originally Posted By: pca
The IDE port has three address lines giving 8 16 bit registers directly addressable. Not all of them are decoded, there are some holes. I can't remember off the top of my head which ones, but I exploited that fact ten years ago to produce what was effectively an IDE-connected VGA card for the empeg, which could be connected to the IDE bus without interfering with the drives smile

Now, it may be that the problem is essentially that the converter is nobbling one or more address lines that don't affect the IDE port, being a non-decoded/non-used location, but the same lines mapped into the ethernet address space prevent it from working correctly. As I remember the ethernet controller uses a lot of registers, and it would only take one to screw it up completely.


Also, I very vaguely seem to recall that some of the IDE operations are 8 bit only, although I may be misremembering that. So again, if the higher data bits had an issue that might not affect the IDE to the point of killing it completely, although this seems much more unlikely to me.


The way we used IDE it used all 16 databits to get anything to/from the disk.

All 3 address bits are used for the bottom CS (sector count, LBA address, etc). The top CS only uses the top addresses for the status register, but you still need all 3 address bits intact otherwise that won't work either as it's fully decoded in the drive. Nothing from ethernet lives in these holes.

The only way I can see anything bad happening here is if the converter is pulling on address or databits when both IDE CS lines are inactive. That *would* cause problems.

...but one sanity check first. Are you sure that you loaded the mk2 disk image? Got a boot spew? That might give more clues.