Re: [hatari-devel] Possible bug in 060 exception stack frames

[Toni Wilen <twilen@xxxxxxxxxx>]

>> 08123BE0 4280 CLR.L D0
>
>> 08123BE2 4e75 RTS
>
>
> Ah, damn. Must have accidently checked that in. No wonder that the
> tests work now :/ With those tests active again, i get the crashes
> also, with or with out softfloat. I currently cannot update the repo
> with a fix, because i have other changes pending, but changed *.sa
> are again attached below.
Thanks. Now it crashes :)
Same happens with real 68060. Address error crash in same test set.

This is what happens:

fmovem_out_ctrl_done:
	move.l		%a1,L_SCR1(%a6)

	lea		(%sp),%a0		/*  pass: supervisor src */
	move.l		%d0,-(%sp)		/*  save size */
	bsr.l		_dmem_write		/*  copy data to user mem */

->

	.globl		_dmem_write
_dmem_write:
	move.l		%d0,-(%sp)
	move.l		(_060FPSP_TABLE-0x80+_off_dmw,%pc),%d0
	pea.l		((_060FPSP_TABLE-0x80).l,%pc,%d0.l)
	move.l		0x4(%sp),%d0
	rtd		&0x4

->

x060_dmem_read:
        dc.w      0x4efb,0x0522,0x6,0         /* jmp
([mov_tab,pc,d0.w*4],0) */
mov_tab:
        dc.l      mov0,mov1,mov2,mov3,mov4,mov5
        dc.l      mov6,mov7,mov8,mov9,mov10,mov11,mov12

-> Out of bounds access. D0 = dynamic FMOVEM data transfer size = 4 * 12
= $30.

I guess you didn't expect that big data transfer :)