att6300p.cpp: Initial driver for AT&T 6300 Plus #14149
Conversation
…ork. The 6300 Plus has external logic to reset the 80286 via the reset pin, so that it can return to Real Mode from protected mode. Upon return to Real Mode, the BIOS code is clearly depending on the fact that SP register is preserved upon reset (the BIOS does however explicitly load the SS register after reset, before it is used). The 8086 and 80286 documentation call out which registers are to be initialized upon reset, and SS is one of them. For AX, BX, CX, DX, BP, SP, SI and DI, the behaviour is not specified. However, based on what the 6300 Plus BIOS is doing, it appears the actual hardware behaviour is to leave them untouched. Although we can only be fully confident that the real 286 behaves this way with respect to the SP register, it's likely that the other registers which do not have explicitly defined initialization behavior are also not touched during reset. Note that this observed behaviour only pertains to reset via the reset pin, and not to power-on initialization.
| virtual void machine_reset() override ATTR_COLD; | ||
|
|
||
| private: | ||
| /* |
There was a problem hiding this comment.
This is really quite verbose. Maybe it can moved down to where it's used and be a bit more condensed (assuming this isn't reverse engineered but also mentioned in the manual).
There was a problem hiding this comment.
Thanks for the code review!
These registers are not documented anywhere that I could find, and I wrote the descriptions based on my understanding from the schematics. I thought it best to put it near the register bit field enums, but I could move it down to where the register reads/writes are implemented.
|
|
||
|
|
||
| // YEAR NAME PARENT COMPAT MACHINE INPUT CLASS INIT COMPANY FULLNAME FLAGS | ||
| COMP( 1985, att6300p, ibm5150, 0, att6300p, att6300p, att6300p_state, empty_init, "AT&T", "6300 Plus", MACHINE_NOT_WORKING ) |
There was a problem hiding this comment.
Very likely the ibm5150 shouldn't be the parent, this seems to be a fully original machine.
Also add MACHINE_SUPPORTS_SAVE if the relevant state is saved.
There was a problem hiding this comment.
This was inherited from m24.cpp. Like the M24 and the original 6300, this machine was at least intended to be PC compatible :-)
The lack of MACHINE_SUPPORTS_SAVE also came from m24.cpp. I'll do some testing and see if saving actually works.
There was a problem hiding this comment.
Seems to work fine. I can push the change when the other issues are resolved.
| (m_nmi_enable && BIT(m_ctrlport_b, 6) && !BIT(m_ctrlport_a, 4))); | ||
|
|
||
| if (m_nmi_active) | ||
| m_maincpu->set_input_line(INPUT_LINE_NMI, ASSERT_LINE); |
There was a problem hiding this comment.
Are you sure this isn't pulsed? I. e. something like m_maincpu->pulse_input_line(INPUT_LINE_NMI, attotime::zero);.
There was a problem hiding this comment.
My understanding is that NMI is to remain asserted, and gets cleared when TRAPCE register is accessed.
Initial driver for the AT&T 6300 plus. It seems to run DOS just fine.
The 6300 Plus contains unique hardware to support a virtual 8086 environment allowing DOS to run concurrently with, and under the control of, a version of AT&T SVR 2.0 Unix that was specifically provided for this machine. While I've attempted to implement the virtualization hardware, it's largely untested and would definitely need more work. Unless someone unearths a copy of the right version of Unix (and the accompanying OS Merge software), the virtualization implementation cannot be completed and fully verified. There are more details in the code comments about the exact version of Unix that's needed.
This is not your run-of-the-mill PC clone, and I'm expecting to get a lot of feedback, especially about how I hooked into accesses from the CPU to implement the virtualization.
Note that this PR includes a proposed change to how the i86/i286 registers are initialized upon reset. This is obviously a risky change, but the BIOS fails when performing the memory check without it (since it fails to return from protected mode to real mode when testing memory above 640K).