Re: [hatari-devel] New WinUAE CPU core debugging support

[Douglas Little <doug694@xxxxxxxxxxxxxx>]

Hi,

The default builds of BadMood can't properly use 32bit addressing because it sets up the PMMU tree to map 0x0e000000 as a clone of the first 16mb of STRam.

The PMMU should intercept those writes and map them back to <14MB before they reach the external bus, but if the PMMU is disabled or not correctly emulated you may get some weird problems in Hatari.

D

On 28 December 2014 at 21:16, Eero Tamminen <oak@xxxxxxxxxxxxxx> wrote:

Hi,

I'm getting huge/unreasonable amounts of "Illegal *put" messages from
new WinUAE CPU core with 32-bit addressing.  They won't invoke any
exceptions, so they cannot be trapped by debugger.

Grepping sources shows them to come from dummy_*put() functions.

I'm getting them with BadMood started in this way:
        $ hatari -s 14 --addr24 false --machine falcon ....

(I assume BM never finishing loading / outputting sound is
it just being incompatible with 32-bit addressing, not an
issue with Hatari.  It happens regardless of whether any
TT-RAM is specified.)


After adding attached patch to Hatari (and using "--debug-except bus"),
I'm seeing messagees coming from stuff like this:
-------------
Illegal lput at 0eb91df0=00000000 PC=15e12

CPU=$15e12, VBL=982, FrameCycles=148822, HBL=290, LineCycles=342, DSP=$1046
$00015e12 : 4298                               clr.l     (a0)+
> disasm $00015e00
$00015e00 : caf0 2400                          mulu.w    (a0,d2.w*4),d5
$00015e04 : 0682 0000 2000                     addi.l    #$2000,d2
$00015e0a : 508f                               addq.l    #8,sp
$00015e0c : 22c0                               move.l    d0,(a1)+
$00015e0e : 2040                               movea.l   d0,a0
$00015e10 : 4241                               clr.w     d1
$00015e12 : 4298                               clr.l     (a0)+
$00015e14 : 5241                               addq.w    #1,d1
> r
  D0 0EB91DF0   D1 00000000   D2 0EB93DF0   D3 00000010
  D4 000FC6E8   D5 8F000000   D6 00000801   D7 FFFFFFFC
  A0 0EB91DF4   A1 000FCAF4   A2 0005ED28   A3 000000CF
  A4 0000A0BC   A5 00000800   A6 0000000B   A7 00200684
USP  0023FD7E ISP  00200684 SFC  00000000 DFC  00000000
CACR 00002101 VBR  00000000 CAAR 00000000 MSP  00000000
T=00 S=1 M=0 X=0 N=0 Z=1 V=0 C=0 IMASK=3 STP=0
FP0: nan FP1: nan FP2: nan FP3: nan
FP4: nan FP5: nan FP6: nan FP7: nan
FPSR: 4000000 FPCR: 00000000 FPIAR: 00000000 N=0 Z=1 I=0 NAN=0
Prefetch 5241 (ADD) 0000 (OR) Chip latch 00000000
00015E12 4298                     CLR.L (A0)+
Next PC: 00015e14
-------------

Why addresses like that (0xeb91df0 = 235MB) go to dummy banks
instead of causing bus errors?


Btw. When looking more into exception debugging support, I noticed that
new WinUAE CPU core is also missing this stuff from old UAE CPU core:
----------------------
            /* Check for double bus errors: */
            if (regs.spcflags & SPCFLAG_BUSERROR) {
              fprintf(stderr, "Detected double bus error at address $%x,
PC=$%lx => CPU halted!\n",
                      BusErrorAddress, (long)currpc);
              unset_special(SPCFLAG_BUSERROR);
              if (ExceptionDebugMask & EXCEPT_BUS)
                DebugUI(REASON_CPU_EXCEPTION);
              else
                DlgAlert_Notice("Detected double bus error => CPU halted!
\nEmulation needs to be reset.\n");
              regs.intmask = 7;
              m68k_setstopped(true);
              return;
            }
            if ((ExceptionDebugMask & EXCEPT_BUS) &&
BusErrorAddress!=0xff8a00) {
              fprintf(stderr,"Bus Error at address $%x, PC=$%lx\n",
BusErrorAddress, (long)currpc);
              DebugUI(REASON_CPU_EXCEPTION);
            }
        }
    }

    /* Set PC and flags */
    if ((ExceptionDebugMask & EXCEPT_NOHANDLER) && (regs.vbr + 4*nr) == 0) {
        fprintf(stderr,"Uninitialized exception handler #%i!\n", nr);
        DebugUI(REASON_CPU_EXCEPTION);
    }
    newpc = get_long (regs.vbr + 4*nr);
    if ( newpc & 1)                             /* check new pc is even */
      {
        if ( nr==2 || nr==3 )                   /* address error during
bus/address error -> stop emulation */
            {
              fprintf(stderr,"Address Error during exception 2/3, aborting
new PC=$%x\n",newpc);
              if (ExceptionDebugMask & (EXCEPT_BUS|EXCEPT_ADDRESS))
                DebugUI(REASON_CPU_EXCEPTION);
              else
                DlgAlert_Notice("Address Error during exception 2/3 => CPU
halted!\nEmulation needs to be reset.\n");
            }
        else
            {
                fprintf(stderr,"Address Error during exception, new
PC=$%x\n",newpc);
                Exception ( 3 , m68k_getpc() , M68000_EXC_SRC_CPU );
            }
        return;
      }
----------------------


Old UAE core:

* prevents Hatari from taking all CPU looping unrecoverable
  exceptions and instead notifies user about it (either by dialog,
  or by invoking debugger if exception debugging is enabled).

* supports "no exception handler" debugger case


        - Eero