68030, 68040 and 68060 MMU support (really!)

[Toni Wilen]

winuae.zip updated yet again. Sarge's installer now runs (until it asks for non-existing CD..)

I guess Gayle IDE ATAPI CDROM emulation is needed next... (at least it is simpler than SCSI adapter emulation)

[gilgamesh]

Yeah, I've seen the installer running. I think the progress is simply incredible.

[Toni Wilen]

Originally Posted by gilgamesh

Yeah, I've seen the installer running. I think the progress is simply incredible.

Progress happened years ago in Aranym, I "only" needed to find how and what to merge without breaking everything else and without slowing down other emulation modes.

btw, you can press ALT+function key to open another console when installer is open and play with basic shell commands/look for /proc entries without need for real installation. Pointless and useless but so is running Linux on emulated Amiga

Remember that there is no:

- CDROM hardware emulation (except CD32/CDTV but I don't think Linux supports those..)
- Ethernet (uaenet is AmigaOS driver)
- Graphics card (Picasso96 is AmigaOS driver, only framebuffer is hardware emulated)

[mark_k]

Originally Posted by Toni Wilen

Bus type does not really matter much, hardware level SCSI/IDE CDROM is not worth the trouble. (Just for installing Linux = not worth the trouble. MMU at least can be used in AmigaOS side too)

[I guess this is getting off-topic for this thread, but...]

Rather than SCSI CD-ROM drive emulation, is there any possibility to "own" a real SCSI device connected to the PC's SCSI controller? In other words, all commands sent by the emulated Amiga are passed straight through to the drive and Windows wouldn't send any commands to the drive itself since WinUAE owns it.

[Toni Wilen]

Originally Posted by mark_k

[I guess this is getting off-topic for this thread, but...]

Rather than SCSI CD-ROM drive emulation, is there any possibility to "own" a real SCSI device connected to the PC's SCSI controller? In other words, all commands sent by the emulated Amiga are passed straight through to the drive and Windows wouldn't send any commands to the drive itself since WinUAE owns it.

It is possible (and off topic) but it requires real hardware access in separate thread or winuae would always pause when real CDROM is seeking or doing any other slow operation and I am not really interested just for mostly useless "real" cdrom emulation.

[Toni Wilen]

Network stuff moved, http://eab.abime.net/showthread.php?t=47388

[jotd]

Toni I just tested a little more today.

Looks like MMU is working in the WHDLoad context but with 2 bugs:

1) when there's an access fault, the PC is after the faulty instruction. On a real amiga, the PC is on the instruction (so resuming execution triggers access fault again)

2) I was able to use it only when I start with MMU on. If I enable it after having booted, it is not active. Maybe something to do with 68040.library at startup?

Anyway, this is not very convenient at the time being but looks very promising!!!

thanks

[Toni Wilen]

Originally Posted by jotd

Toni I just tested a little more today.

Looks like MMU is working in the WHDLoad context but with 2 bugs:

1) when there's an access fault, the PC is after the faulty instruction. On a real amiga, the PC is on the instruction (so resuming execution triggers access fault again)

Perhaps you tested write faults because they don't restart on 68040? (only on 68060 all faults restart. btw, 68060 MMU is nearly identical to 68040 MMU, only status registers and exception vectors are different. Perhaps 68060 MMU option is possible in future..)

EDIT: AFAIK 68040 can even delay write faults (writes are async), PC may not always point to immediately next instruction.. ("write" = instruction that does "normal" single write at the end, this does not include special cases like MOVEM to memory)

What was the exact instruction?

2) I was able to use it only when I start with MMU on. If I enable it after having booted, it is not active. Maybe something to do with 68040.library at startup?

I can guarantee that this is possible but I'll try

[Wepl]

my 68040 test machine is now configured, will next add full 68040 mmu support (Snoop/Protect) and afterwards test it with WinUAE

[jotd]

Wouldn't that be a dream if finding whdload access faults with WinUAE was possible?

[gilgamesh]

Finally found a way to install without that unsatisfied demand for a CDROM. Just use the initrd22.gz from hd-media folder (Sarge) as initial ramdisk. Now the installation hangs at the network adapter, so I will describe the problem in the appropriate thread.

[NewDeli]

I am not sure this will help much, but CheckMMU reports no cookie.

With a 040, SET040 is needed in the path to run it.
After installing
I simply typed

• SET040 fastrom
• checkMMU

and here is the result :



Oh, I am lamulating in case you hadn't notice yet .


To answer Kyon's question, the MMU is also used by WHDload.
So you might want to enable it. Check that your game icon doesn't have the nommu set, if that's the case put it into ().
In WHDload prefs, remove the ; in front of MMU and you're done.

This is especially useful with the latest WinUAE betas, in which the emulation is more accurate,
and a missing MMU is reported as an error by the slave (missing in the meaning : unticked in WinUAE).

[StingRay]

Originally Posted by Deleauvive

To answer Kyon's question, the MMU is also used by WHDload.
So you might want to enable it.

To quote Toni:

Originally Posted by Toni Wilen

It is only (possibly) useful if you are programmer or want to play with Linux (or other Unix based stuff). Just ignore it. Completely.

And this is 100% correct. You don't need any MMU support in WHDLoad if you aren't coding slaves.

[Toni Wilen]

Originally Posted by Deleauvive

I am not sure this will help much, but CheckMMU reports no cookie.

Works for me. Too old version? (latest beta is not new enough)

[Predabot]

Hey guys, did the development and testing of the MMU-support stop here? Seems like some people lost interest... But surely, MMU-support could lead to some cool stuff?

So... anything new to report?

[Toni Wilen]

MMU emulation works except there is some unknown corner case that is practically impossible to debug using current method (in Linux some user space programs crash randomly)

I don't really care about MMU in emulation, for me it was yet another interesting learning experience (how exactly 68040 MMU works etc..)

[gilgamesh]

Btw, did anybody succeed in installing NetBSD yet? You need a disk partitioning tool, which I don't have, for that. It's probably not hard with WB 2.0 or newer.

[Toni Wilen]

Possible good news: I was cleaning up MMU code (put back C++ exception handler now that WinUAE compiles as C++) and decided to disable some MMU table caching in MMU code (speed optimization)

NetBSD MMU error crash disappeared.

It seems cache gets out of sync in some situations ("slow" MMU table routines return "access ok" but cache says "nothing mapped here")

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

I'll contact Aranym developers if I can't find any obvious reasons for this error (perhaps it is some Atari-specific assumption somewhere that simply won't work with Amiga or something..)

Hopefully it is now possible to create installed netbsd or linux hardfile

[alexh]

Heh, cool. Good work. I'm sure some nutters on here will be trying it as I type.

[Toni Wilen]

It works, Debian Sarge successfully installed.

dmesg:

Linux version 2.4.27-amiga (root@dl360-g3) (gcc version 2.95.4 20010319 (prerelease)) #1 Wed May 31 16:30:52 MDT 2006
Amiga hardware found: [A4000] VIDEO BLITTER AUDIO FLOPPY A4000_IDE KEYBOARD MOUSE SERIAL PARALLEL A3000_CLK CHIP_RAM PAULA LISA ALICE_PAL ZORRO3 
On node 0 totalpages: 131072
zone(0): 131072 pages.
zone(1): 0 pages.
zone(2): 0 pages.
Kernel command line: root=/dev/hda1
Console: colour dummy device 80x25
Calibrating delay loop... 0.97 BogoMIPS
Memory: 515572k/524288k available (1544k kernel code, 7056k data, 116k init)
kmem_create: Forcing size word alignment - mm_struct
kmem_create: Forcing size word alignment - filp
kmem_create: Forcing size word alignment - dquot
Dentry cache hash table entries: 65536 (order: 7, 524288 bytes)
Inode cache hash table entries: 32768 (order: 6, 262144 bytes)
kmem_create: Forcing size word alignment - inode_cache
Mount cache hash table entries: 512 (order: 0, 4096 bytes)
kmem_create: Forcing size word alignment - bdev_cache
kmem_create: Forcing size word alignment - cdev_cache
kmem_create: Forcing size word alignment - kiobuf
Buffer cache hash table entries: 32768 (order: 5, 131072 bytes)
Page-cache hash table entries: 131072 (order: 7, 524288 bytes)
POSIX conformance testing by UNIFIX
Zorro: Probing AutoConfig expansion devices: 3 devices
Zorro: Address space collision on device Zorro device 07db0300 (Hacker Test Board) [10000000:2fffffff]
Linux NET4.0 for Linux 2.4
Based upon Swansea University Computer Society NET3.039
Initializing RT netlink socket
Starting kswapd
kmem_create: Forcing size word alignment - file_lock_cache
VFS: Disk quotas vdquot_6.5.1
Journalled Block Device driver loaded
devfs: v1.12c (20020818) Richard Gooch (rgooch@atnf.csiro.au)
devfs: boot_options: 0x0
Console: switching to colour frame buffer device 80x25
fb0: Amiga AGA frame buffer device, using 1280K of video memory
clgen: Driver for Cirrus Logic based graphic boards, v1.9.9.1
clgen: no supported board found.
pty: 256 Unix98 ptys configured
Amiga-builtin serial driver version 4.30
ttyS00 is the amiga builtin serial port
Amiga mouse installed.
kmem_create: Forcing size word alignment - blkdev_requests
FD: probing units
found fd0 
Unable to get major 2 for floppy
RAMDISK driver initialized: 16 RAM disks of 4096K size 1024 blocksize
loop: loaded (max 8 devices)
Uniform Multi-Platform E-IDE driver Revision: 7.00beta4-2.4
ide: Assuming 50MHz system bus speed for PIO modes; override with idebus=xx
ide0: Gayle IDE interface (A4000 style)
hda: UAE-IDE debian.vhd, ATA DISK drive
ide0 at 0x80dd2020 on irq 12
hda: attached ide-disk driver.
hda: 4096000 sectors (2097 MB) w/256KiB Cache, CHS=16253/4/63
hda: task_no_data_intr: status=0x51 { DriveReady SeekComplete Error }
hda: task_no_data_intr: error=0x04 { DriveStatusError }
Partition check:
 /dev/ide/host0/bus0/target0/lun0: RDSK p1 p2
SCSI subsystem driver Revision: 1.00
kmod: failed to exec /sbin/modprobe -s -k scsi_hostadapter, errno = 2
kmod: failed to exec /sbin/modprobe -s -k scsi_hostadapter, errno = 2
Initializing Cryptographic API
NET4: Linux TCP/IP 1.0 for NET4.0
IP: routing cache hash table of 4096 buckets, 32Kbytes
TCP: Hash tables configured (established 131072 bind 65536)
NET4: Unix domain sockets 1.0/SMP for Linux NET4.0.
EXT3-fs: INFO: recovery required on readonly filesystem.
EXT3-fs: write access will be enabled during recovery.
kjournald starting.  Commit interval 5 seconds
EXT3-fs: recovery complete.
EXT3-fs: mounted filesystem with ordered data mode.
VFS: Mounted root (ext3 filesystem) readonly.
Adding Swap: 122968k swap-space (priority -1)
EXT3 FS 2.4-0.9.19, 19 August 2002 on ide0(3,1), internal journal
eth0: A2065 at 0x00e90000, Ethernet Address 00:80:10:4d:5b:61
fd0: accessing DD 3.5-disk with Amiga-layout
fd0: accessing DD 3.5-disk with Amiga-layout
FAT: bogus logical sector size 17264
VFS: Can't find a valid FAT filesystem on dev 02:00.
fd0: accessing DD 3.5-disk with Amiga-layout