68030, 68040 and 68060 MMU support (really!)

[gilgamesh]

Awesome, simply awesome!

[andreas]

EAB. Where Miracles Happen Every Day.™

(RCK, go make this a brand, already! (before someone beats you to it))

To a both more serious and technical note:

hda: task_no_data_intr: status=0x51 { DriveReady SeekComplete Error }
hda: task_no_data_intr: error=0x04 { DriveStatusError }

Now that the main uglinesses are finally ironed out, there is time to tackle the above, obviously rather minor issues.

Andreas, it may not be a problem at all.
http://www.captain.at/howto-linux-dr...tatuserror.php

The gist of those messages is that the drive may have received a command it did not understand. Maybe the linux driver is sending something odd, or maybe the emulated drive is missing a command. Or a timing issue possibly. Basically a command sent was not understood for some reason.

The fellow from the article does go on to say that you sometimes see those messages pop up in the kernel log on occasion with a healthy system.

In any event I'm off to go try it myself. Linux on Winuae - exciting!

[andreas]

Nonetheless, these messages do look ugly, whatsoever. (Always.)

[mark_k]

Hi,

In the MAME 0.141u3 changelog there is this:

M680x0 updates [Hans Ostermeyer]
* Instruction cache emulated on '020 and later
* Fixed interaction between MMU and bfextu/bfexts/bfins
* Added fsgldiv, fsglmul, and fscc FPU instructions
* Fixed fault address in stack frame
* Fixed supervisor violation bit in MMU status register
* Add fmovem modes 1 and 3
* Various other MMU fixes

I don't know if the MMU emulation was originally developed by the same author, but might WinUAE benefit from those fixes?

[Toni Wilen]

Probably something totally different and license incompatible anyway.

[Toni Wilen]

Next emulator "Previous" (http://previous.alternative-system.com/) has 68030 MMU under development (based on Aranym 68040 MMU/UAE CPU emulator) and is GPL compatible.

UAE 68030 MMU emulation probably will be done someday..

[Toni Wilen]

Previous 68030 MMU emulation is now working, UAE merge will happen..

EDIT:

Merge done. Mostly missing bus error exception handler implemented. Still some work do to.

A3000 boot ROM and Enforcer confirmed working!

[Toni Wilen]

Thread title changed. It is time to start testing 68030 MMU.

http://www.winuae.net/files/b/winuae.zip

Tested and seems to work:

A3000 1.3 SuperKickstart
Enforcer
WHDLoad (at least it does not crash)

Debian Linux finds root drive and then it seems to get in infinite bus error loop.
Amix should now technically work, not yet tested..
NetBSD not yet tested.

[Toni Wilen]

68060 MMU implemented, at least Enforcer works.

[mark_k]

Nice work! I'll try testing that with SetCPU FASTROM/KICKROM/CARDROM options and the MMULib tools MuFastROM, MuFastZero etc. Could possibly also be tested (by someone else ) with Emplant and/or ShapeShifter MMU-using video drivers and virtual memory software.

Now you just need to add SCSI tape drive emulation and I can try my copy of AMIX. [That's not entirely a joke. Adding minimal read-only support to at least allow the AMIX installer to read the tape might not actually be hugely difficult.]

[mark_k]

I tried out mmu.library and some programs which come with it: http://aminet.net/package/util/libs/MMULib

It seems that running anything which opens mmu.library (MuScan, MuFastROM, MuForce etc.) results in the machine resetting. WinUAE log output when that happens looks something like this:

68030 MMU enabled. Page size = 32768
Function code lookup enabled, FC = 6
Table FCL at 00202240: index = 6, Function code lookup enabled, FC = 6
Table FCL at 00202240: index = 6, Function code lookup enabled, FC = 1
Table FCL at 00202240: index = 1, Function code lookup enabled, FC = 6
Table FCL at 00202240: index = 6, Function code lookup enabled, FC = 5
Table FCL at 00202240: index = 5, MMU disabled
68030 MMU enabled. Page size = 1024
MMU: page fault (logical addr=00202286 SSW=0325 read=0 size=2 fc=5 pc=00f80be0)
MMU: page fault (logical addr=00202282 SSW=0305 read=0 size=4 fc=5 pc=00f80be0)
MMU: double bus error, rebooting..

Or with 68040+MMU CPU

Illegal instruction: 4e7a at 00363A2C -> 00363A3C
68040 MMU: enabled=0 page8k=0
68040 MMU: enabled=0 page8k=0
68040 MMU: enabled=1 page8k=0
MMU: double bus error, rebooting..
68040 MMU: enabled=0 page8k=0

[Toni Wilen]

Originally Posted by mark_k

It seems that running anything which opens mmu.library (MuScan, MuFastROM, MuForce etc.) results in the machine resetting.

Works fine here (both 68030 and 68040)

MMU: page fault (logical addr=00202286 SSW=0325 read=0 size=2 fc=5 pc=00f80be0)
MMU: page fault (logical addr=00202282 SSW=0305 read=0 size=4 fc=5 pc=00f80be0)

This makes no sense if you have RAM at 00200000. These messages means 00202282 is write protected or mapped as invalid..

[mark_k]

Originally Posted by Toni Wilen

EDIT: Actually it does not get that far, it tries to find SCSI HD but it accesses DMAC/SCSI chip quite differently compared to A590/A2091/A3000 Amiga boot roms and it hangs.. Need to implement missing WD SCSI chip functionality first.

Could the files /usr/sys/amiga/alien/a2091.c and /usr/sys/amiga/alien/a3091.c in the AMIX kernel configuration package (conf.cpio) be of any help there?

I'll try testing the MMULib tools when booting from a clean floppy disk and report back.

[Toni Wilen]

Amix and wd33c93 stuff moved to separate threads.

[mark_k]

Would you be able to build a winuae.exe which pauses emulation or halts on a double bus error?

Hopefully I can use something like SnoopDOS to figure out what system functions are being called before I get the crash/reboot, and see the details on screen before the crash. As far as I remember the Mu... tools did work from this hard disk on my real 68030 A2000. (I tried booting/running them from a clean floppy disk and they do work from there in WinUAE.)

[Toni Wilen]

Originally Posted by mark_k

Would you be able to build a winuae.exe which pauses emulation or halts on a double bus error?

Done. Now CPU halts (just like real CPU would do)

[mark_k]

Thanks. That allowed me to use SnoopDOS and see ENV:MMU-Configuration was being read shortly before the hang. It turns out I had an entry to write-protect $200000-$27FFFF. The Kickstart image was loaded there by ReKick on my A2000, but I recently changed my emulated config to use 3.1 ROM not ReKick, so there was actually RAM there.

In the case when the CPU is halted, is it worth telling the user that somehow? Also I noticed the CPU indicator still reads 100% when the CPU is halted.

[Toni Wilen]

Originally Posted by mark_k

Thanks. That allowed me to use SnoopDOS and see ENV:MMU-Configuration was being read shortly before the hang. It turns out I had an entry to write-protect $200000-$27FFFF. The Kickstart image was loaded there by ReKick on my A2000, but I recently changed my emulated config to use 3.1 ROM not ReKick, so there was actually RAM there.

Bah, not an emulation bug

In the case when the CPU is halted, is it worth telling the user that somehow? Also I noticed the CPU indicator still reads 100% when the CPU is halted.

I changed it to show "HALTx" where x = halt reason code (I also changed all non-MMU related insta-resets to halts), OSD CPU% becomes yellow and shows Hx.

[Wepl]

I started to check the 68030 with WHDLoad and my qa.slave.
There is a strange fault with many (but not all) checks while whdload is leaving and saving the contents of the custom registers.

winuaelog.txt

51-703 [1498 107-010]: MMU: page fault (logical addr=00BFDE00 SSW=0315 read=0 size=1 fc=5 pc=009d5ec4)
51-703 [1498 132-011]: MMU: page fault (logical addr=00BFDF00 SSW=0315 read=0 size=1 fc=5 pc=009d5ee4)
51-703 [1498 220-013]: MMU: page fault (logical addr=00DFF002 SSW=0365 read=1 size=2 fc=5 pc=009d5f78)
51-703 [1498 108-014]: MMU: page fault (logical addr=009F7474 SSW=0325 read=0 size=2 fc=5 pc=009d5f78)
51-703 [1498 121-015]: MMU disabled

.whdl_register

************************* 04-Jan-13 13:00:43 ************************17.2.5225***
Slave='QA.Slave' (14320 bytes)
ShadowMem   52D940 -   539940 (  49152) AbsolutMem     C000 -    40000 ( 212992)
Resload     9D5000 -   9DC2D8 (  29400) at 9D5000  GL=$9EF000
Slave       9F8000 -   9FB7CC (  14284) at 9F8000  BaseMemSize=$40000
ExpMem      9DE000 -   9EE000 (  65536) at 9DE000
attn=207(UAE,30) fc=4 kn=1000 cs=FD26 rw=0 zpt=-1 ep=0 ei=0
setcpu=33D(DC,IC,SCB,ECB,BNC)

Exception "Access Fault" ($B008) PC = $9D5F78 (Resload $F78) Word Write to
$9F7474

$009d5f3a move.b       ($200,a5),(4,a1)               ;$009f1200 $00dff004
$009d5f40 move.b       (a5),(2,a1)                    ;$009f1000 $00dff002
$009d5f44 move.b       ($300,a5),(5,a1)               ;$009f1300 $00dff005
$009d5f4a move.b       ($100,a5),(3,a1)               ;$009f1100 $00dff003
$009d5f50 move.b       ($c00,a6),($16,a0)             ;$009f2910 $009f6442
$009d5f56 move.b       ($c00,a5),($16,a1)             ;$009f1c00 $00dff016
$009d5f5c move.l       d4,a1
$009d5f5e lea          ($2d10,a4),a6                  ;$009f1d10
$009d5f62 lea          ($2000,a4),a5                  ;$009f1000
$009d5f66 moveq        #0,d0
$009d5f68 btst         #0,(0,a6,d0.w)                 ;$009f1d12
$009d5f6e beq.b        $9d5f7e
$009d5f70 btst         #1,(0,a6,d0.w)                 ;$009f1d12
$009d5f76 bne.b        $9d5f7e
$009d5f78 move.w       (0,a1,d0.w),(0,a5,d0.w)        ;$00dff002 $009f1002
$009d5f7e addq.w       #2,d0
$009d5f80 cmpi.w       #$1fe,d0
$009d5f84 bls.b        $9d5f68
$009d5f86 move.w       (2,a1),d1                      ;$00dff002
$009d5f8a andi.w       #$180,d1
$009d5f8e beq.b        $9d5fa4
$009d5f90 move.w       #$f00,d0
$009d5f94 bsr.w        $9d65b4
$009d5f98 move.w       #$7fff,($96,a1)                ;$00dff096
$009d5f9e move.w       #$7fff,($9a,a1)                ;$00dff09a
$009d5fa4 move.w       ($1c,a1),d1                    ;$00dff01c
$009d5fa8 andi.w       #$7fff,d1
$009d5fac beq.b        $9d5fc2
$009d5fae move.w       #$f,d0

exception stackframe:
$009EBF44 2700009D 5F78B008 00000325 00000000 009F7474 00003BB1 00002000 00000000
$009EBF64 00000000 009F7474 00008000 009F7474 00000000 00000003 00000000 00000000
$009EBF84 00000000 00000000 00000000 00002000 0000B008 00DF2000 00000000
regular stack:
$009EBFA0 2000009F A670B008 00000325 00000000 00DFF084 00003D41 00000001 00000000
$009EBFC0 00000000 00DFF084 00000000 00DFF084 00000000 00000001 00000000 00000000
$009EBFE0 00000000 00000000 00000000 00000000 00000000 00000001 00000084 009F82D6
$009EC000 DDDDDDDD DDDDDDDD DDDDDDDD DDDDDDDD DDDDDDDD DDDDDDDD DDDDDDDD DDDDDDDD
$009EC020 DDDDDDDD DDDDDDDD DDDDDDDD DDDDDDDD DDDDDDDD DDDDDDDD DDDDDDDD DDDDDDDD

  ----0---- ----1---- ----2---- ----3---- ----4---- ----5---- ----6---- ----7----
Dx        2         0         0         2    DFF000  D5D5D5D5    9D5C1A  D7D70207
Ax   9F642C    DFF000    9F6318    9F646C    9EF000    9F1000    9F1D10

                 TTSM III   XNZVC
›1mPC=  9D5F78›22m  SR %0010011100000000  USP=3FC00  ›1mISP=9EBF44  ›22mMSP=3F800
VBR=9F0000  SFC=4  DFC=4  CACR=2101  CAAR=0  TT0=00FF8540  TT1=00000000
TC=00000000  CRP=80000002009FE000  SRP=80000002009FE000  MMUSR=0000

         MEDRAAAABVCPSDT          YZ  PMRCBSDAAAA
intena=%›1m0000000000000000›22m dmacon=%›1m0010000000000000›22m adkcon=$1100
intreq=%›1m0000000000100000›22m
vposr....A100  vhposr...01DC  joy0dat..0200  joy1dat..0000  clxdat...8000  
pot0dat..0000  pot1dat..0000  potinp...5500  serdatr..3000  dskbytr..0000  
vposw....F004  bltafwm..FFFF  bltalwm..FFFF  cop1lc...0000  cop1lcl..5000  
cop2lc...0000  cop2lcl..5000  spr0ctl..0000  spr0data.0000  spr0datb.0000  
spr1ctl..0000  spr1data.0000  spr1datb.0000  spr2ctl..0000  spr2data.0000  
spr2datb.0000  spr3ctl..0000  spr3data.0000  spr3datb.0000  spr4ctl..0000  
spr4data.0000  spr4datb.0000  spr5ctl..0000  spr5data.0000  spr5datb.0000  
spr6ctl..0000  spr6data.0000  spr6datb.0000  spr7ctl..0000  spr7data.0000  
spr7datb.0000

but (0,a5,d0.w) should point to $009f1002 and not to $9F7474 as the access fault tells

is there maybe a bug?

to reproduce:

whdload qa.slave mmu custom1=92325 SplashDelay=0

qa.slave is in WHDLoad_dev at WHDLoad/Src/qa