machine/8042kbdc.cpp: make aux port to be alive with PS/2 interface

[angelosa]

• Fix shutms11 recognizing mouse properly instead of indefinitely hang.

[angelosa]

Opening as a PR because I'm not sure if this is the right fix. It looks like that 8042kbdc is really untested for anything PS/2 based.

• pcipc just uses serial mouse
• nforcepc doesn't set double as interrupt type (will fail later I presume)
• pciagp still fails for another bug when it inits keyboard ...

[cuavas]

Eugh, this thing still has the PS/2 mouse simulation in the controller, duplicating the separate PS/2 mouse simulation.

First thing, I would have thought it should be impossible to get to that point if m_mouse.on is false, since the mouse shouldn’t be processing commands at all if the auxiliary interface is disabled (KBC commands 0xa7/0xa8 control this).

Secondly, the mouse reset command (0xff) should set the following state:

• 100 samples/second rate
• 4 counts/millimetre resolution
• 1:1 scaling
• Reporting disabled

(For example see here: https://eaw.app/Downloads/PS2_Mouse.pdf – search for Reset Mode to find the reset state.)

After completing reset, it should do a self-test and report 0xaa (good) or 0xfc (fail), followed by its device ID 0x00.

So on the face of it, this change is wrong on both counts. If it improves behaviour, it’s working around bigger issues elsewhere. Given the history of this device, that shouldn’t be a surprise.

[angelosa]

Thanks for the heads up, I'll investigate thru this branch (easy to revert)

My test rig is a FreeDOS 1.3 with CuteMouse driver, latter works up to pcipc but indefinitely hang with shutms11, and I presume the aux enable/disable state isn't respected at all here (reading something wrong?) Other OSes (MSDOS, Windows 98 etc.) behaves in similar fashion (just pukes at current emulation)

[angelosa]

Has no sense leaving this open after https://github.com/mamedev/mame/commit/b77af1eb3859dabb2d620e9a79afef7a4985b01c