Page fault : CR3 not page-aligned

[nimbus]

When trying to run the flat assembler, DOS version (FASM.EXE) it crashes immediately, yields the above message in a windows error box,
and it aborts vDOS (64 bit) instantly. This is with - or without - first running the DPMI server, CWSDPMI, that FASM relies upon.

Any hint of why this is happening and how to fix the issue if possible ?

I have not investigated in depth, but by the by, I'm finding the vDOS by itself is providing EMS, apparently EMS 3.2 *only*.
That might be incompatible with the CWSDPMI :=(

Can the built in EMS 3.2 "board" be removed from vDOS ?

[Jos]

The address/value of the CR3 register must be 4K page aligned, so vDos exits with this error if a program tries to write a non-aligned value to it. EMS memory is always provided, only in previous versions you could turn this off.

FASM is also available as a genuine Windows program, why then run it on an emulated PC/DOS Eventually you could try DOSBox.

Jos

[nimbus]

Hi Jos !
> The address/value of the CR3 register must be 4K page aligned, so vDos exits with this error if a program tries to write a non-aligned value to it.
Sure, but why does this condition happen in vDOS ? It never happens in DOS running on real "metal".

>FASM is also available as a genuine Windows program, why then run it on an emulated PC/DOS?
- However isn't vDOS supposed to run serious DOS programs, such as FASM ?

> Eventually you could try DOSBox.
- Isn't DOSBox, precisely, supposed to be for running GAMES (as opposed to "serious" DOS programs. Cf. previous point ;=)

Anyway I'm just reporting what I found doing casual tests. Never mind if you're not interested in digging deep inside
the particular problem re FASM. Still it raises the question , does CWSDPMI-extended programs (other than FASM then) usually run well or at all in vDOS ? What about other "extenders" ? HXDOS ?



(Edited, to add) : re. DOSBOX, precisely, the combo : CWSDPMI + FASM was reported to work on it, by a person named Rugxulo writing to some DOS-oriented news group a couple days ago. I don't have DOSbox myself, but generally I would trust Mr Rugxulo who is an active tester of FreeDOS and all DOS things in general. [later edited:] I've just confirmed CWSDPMI (and FASM) run fine in a "vanilla" DOSBOX, this seems to point to some specific bug or regression in the current vDOS ...


Nim'

[Jos]

Perhaps real CPU’s ignore the offset if found in the address.
vDos is intended to run serious programs, though end-user programs that survived and still being used. It doesn’t emulate the PC hardware to a level like DOSBox does.

Don’t know about CWSDPMI or HXDOS extended programs still in use. Only Phar Lap (FoxProX) and DOS/16M (several) that are supported.

Jos

[nimbus]

> Perhaps real CPU’s ignore the offset if found in the address.
- Indeed ! Not "perhaps" ° ! If vDOS doesn't, it is clearly a "feature" (== bug :=) in vDOS CPU emulation :=(


° From Intel's official manual : CR3 — Contains the physical address of the base of the paging-structure hierarchy and two flags (PCD and
PWT). Only the most-significant bits (less the lower 12 bits) of the base address are specified... the lower 12 bits
of the address are assumed to be 0.
°° In addition, note that since the 80486 (!) some of those 12 lowest CR3 bits ARE in use by the processor (PCD, PWT, for page-level cache control... maybe more on newer X86-64 architecture). For vDOS emulation purposes you MUST ignore those bits. Absolutely.

[Jos]

I’ll remove the test in the next version.

Jos

[nimbus]

Jul 23, 2018 14:25:50 GMT 1 Jos said:

I’ll remove the test in the next version.

Thx ! Is there already an estimated date of arrival for "next" ?

[Jos]

Would be first half next year.

Jos

[nimbus]

Jul 23, 2018 17:19:27 GMT 1 Jos said:

Would be first half next year.

Jos

Oh ! Wait till the next year to have an inappropriate check removed ? You must be kidding :=)

Well, uh, meanwhile, I'm going to cling with DOSBOX - that does the job.

Good luck, Jos !

--
Nim'