Re: [hatari-devel] Hatari and OUTSIDE (virtual memory manager)

[Nicolas Pomarède <npomarede@xxxxxxxxxxxx>]

Le 02/10/2018 à 17:25, Uwe Seimet a écrit :
> I'm afraid this change does not make a difference on my machine. The
> behavior is exactly as before. Even complete removing the code block and
> always returning "true" did not have any effect.

If I try with ROMSPEED, it does change something.

        lea $00e00000,a0
..nom1:  ptestr #2,([8,a0]),#7,a0        ;refer to user program address space
        move.l d0,(a0)  ;enter in descriptor table
        pflusha


Before that "ptestr #2,([8,a0]),#7,a0" gave an exception 11 when 
executing this instruction.
If I remove "eamode==6" then the instruction completes but the program 
crashes after that with a bus error on "move.l d0,(a0)" because a0=0 
(and RAM location 0.l and 4.l can't be written in Atari)


> > > > > According to TEMPLMON the instruction is F036 9E01 0161 0010
> > > > >
> > > > > PTESTR DFC,([$0010,A6]),#7
> >
> > mmu_op30_invea() incorrectly rejects effective address mode 6. I added
> > the test because previously it didn't reject all invalid addressing
> > modes like real CPU does and for some reason I also added one valid
> > addressing mode..
> >
> > Change:
> >
> > if (eamode == 0 || eamode == 1 || eamode == 3 || eamode == 4 || eamode
> > == 6 || (eamode == 7 && rreg > 1))
> >
> > To:
> >
> > if (eamode == 0 || eamode == 1 || eamode == 3 || eamode == 4 || (eamode
> > == 7 && rreg > 1))
> >
> > It should fix the unexpected f-line.