ANAIIS 1.05 (28 may 2012)

[Gilloo]

If you got a basic Amiga and a Subway (or a Highway), you can now boot like A1000 did it in the past. With a boot disk!

http://aminet.net/package/driver/other/anaiis_boot

A usb->sata interface + disk is needed. (thumbs are not reliable)
The system must be prepared on a PFS3 partitioned disk.
WinUAE can do this job, if your system disks are out.
Not really speedy, but it can help.

[Gilloo]

Some weeks ago I have posted something concerning WinUAE and dlls.
I have understood how to interface a dll with WinUAE, and now I can manage the libusb.dll functions and a lot of windows stuff (like read/write .ini files, open/close/read/write windows files from inside the emulator, in 68000 mode).

It is strange (or marvelous ) to see ANAIIS working also under PC and WinUAE.
Joysticks (with rumble!), keyboards, mouse are ok!

Usb keys and consorts don't work yet, usb_bulk_read, returns -116 (timeout) but usb_bulk_write seems to write something...
So if someone is interressed by my work, it will decide me to put it on aminet or keep it jealously in my dark harddisk

[Toni Wilen]

Just a small warning: in my opinion that dll interface is badly designed horrible security hole. Interface used = unsupported state.

[Gilloo]

Originally Posted by Toni Wilen

Just a small warning: in my opinion that dll interface is badly designed horrible security hole. Interface used = unsupported state.

Yes this interface is very tricky, but very useful for me (please Toni don't remove it in WinUAE )
You can use memory or ressource outside the emulator, but at this time, I haven't seen any blue screen.

- Only D1,D2,D3,D4,D5,D6,D7,A1,A2 registers (9 pointers or unsigned long values) can be used (perhaps it's a SASC limitation... )
- Only chip memory (memory starting at $0) or Zorro3 memory can be used (like a DMA memory in fact )
- SASC has a little bug with D1 and D6 registers: they can't be pointers (the workaround is to cast all in ULONG )

[Toni Wilen]

Security problem is that DLL runs inside winuae process space = it has same privileges as winuae which is bad if winuae was run as admin. DLL can do anything it wants in this situation.

Most users assume emulators won't allow full access to host. This makes it dangerous.

Also DLL can always poke and peek any address in winuae's address space, it can do all kinds of things behind emulator's back. This makes it unsupported.

I am not going to remove it, at least unless someone abuses it too much but it is possible I'll add some confirmation dialogs before any previously unused dll gets loaded for the first time.

Only chip memory (memory starting at $0) or Zorro3 memory can be used

Why do you think so? All memory is linear mapped from base address in 68020+ modes. Last few betas have linear memory, even if 68000.

(Don't care about SAS-C)

[Gilloo]

Originally Posted by Toni Wilen

Why do you think so? All memory is linear mapped from base address in 68020+ modes. Last few betas have linear memory, even if 68000.

betas have linear memory, even if 68000.
Not a thought, to pass a pointer to a dll, you should translate the 'amiga' address to 'host' address. The function 105, passing through $00f0ffc0 is the offset between address 0 (the beginning of the chip mem) and the true block ptr in the 'host' mapping.

; input void
; output D0 return memory offset
DLLGetMemoryOffset
moveq #105,D0
jsr $00f0ffc0
rts

This routine computes the offset between address 0 and the real memory offset in Windows world.
All buffers must be allocated in CHIP, for example toto = AllocMem(512, MEMF_CHIP), then DLLGetMemoryOffset()+toto matches exactly the buffer toto in the dll.


Perhaps you want I keep all this stuff secret to not use WinUAE for hacking

[Toni Wilen]

Originally Posted by Gilloo

This routine computes the offset between address 0 and the real memory offset in Windows world.

Yes and you can add it to any Amiga-side RAM address, not just chip. (as long as it is real RAM, poking IO or similar won't work, of course)

[Gilloo]

Originally Posted by Toni Wilen

Yes and you can add it to any Amiga-side RAM address, not just chip. (as long as it is real RAM, poking IO or similar won't work, of course)

RAM, poking IO or similar won't work, of course)
You mean that DLLGetMemoryOffset()+toto, where toto is in FAST works too ?
Strange, for me it crashes the emulator...

[Toni Wilen]

Originally Posted by Gilloo

RAM, poking IO or similar won't work, of course)

You mean that DLLGetMemoryOffset()+toto, where toto is in FAST works too ?

Strange, for me it crashes the emulator...

What kind of config? Did you use latest beta?

[Gilloo]

Originally Posted by Toni Wilen

What kind of config? Did you use latest beta?

hu, not the last beta WinUAE 1.4.4.0 (2007) and also 2.4.0.0 (2012)
A1000 68000 + kick 1.1 + 512kB chip
A2000 68030 + 68882 + kick 3.1 + 2MB chip + 32MB Z3 fast + 32MB video(Picasso96)

Please continue to support windows dll, cause there is a lot of good things to do with like wazp3d (to make opengl things...) thanks

[Toni Wilen]

Originally Posted by Gilloo

hu, not the last beta WinUAE 1.4.4.0 (2007) and also 2.4.0.0 (2012)
A1000 68000 + kick 1.1 + 512kB chip
A2000 68030 + 68882 + kick 3.1 + 2MB chip + 32MB Z3 fast + 32MB video(Picasso96)

Please continue to support windows dll, cause there is a lot of good things to do with like wazp3d (to make opengl things...) thanks

Never ever ask or complain about anything without first testing latest version (and also beta if asking on eab!)..

[desiv]

To be fair:

Originally Posted by Gilloo

Strange, for me it crashes the emulator...

That's a statement, not a complaint or a question...

:-)

desiv

[FastRobPlus]

Originally Posted by Gilloo

If you got a basic Amiga and a Subway (or a Highway), you can now boot like A1000 did it in the past. With a boot disk!

http://aminet.net/package/driver/other/anaiis_boot

A usb->sata interface + disk is needed. (thumbs are not reliable)

Not really speedy, but it can help.

What's the risk of using a thumbdrive? Are they all unreliabale? Or just some?

[Gilloo]

Originally Posted by Toni Wilen

Never ever ask or complain about anything without first testing latest version (and also beta if asking on eab!)..

b...bu...but this is my thread !!
Same crash with WinUAE 2.3.1.b (not the last but not so old too)
In fact the 105 function gives the start address of the block allocated by WinUAE for the chipmem. Other memory blocks are allocated elsewhere and not follow this block. Other advanced developpers use the structure pointed by the A6 register to get the Z3 offset. So there is always a solution.

[Toni Wilen]

Originally Posted by Gilloo

b...bu...but this is my thread !!

You started to argue with me about WinUAE

Same crash with WinUAE 2.3.1.b (not the last but not so old too)

I said in non-JIT mode it is only supported in latest betas..

What Amiga address? What did 105 return?

In fact the 105 function gives the start address of the block allocated by WinUAE for the chipmem. Other memory blocks are allocated elsewhere and not follow this block. Other advanced developpers use the structure pointed by the A6 register to get the Z3 offset. So there is always a solution.

Why do you need "Z3 offset"? Any Amiga address (chip or fast or whatever RAM) + 105 function return value = correct address. Everything is linear mapped (except IO or some other non-RAM special regions).

btw, make sure you use unsigned address variables because returned address can be very high in 32-bit address space if using 64-bit Windows. (like 0x7fxxxxxx or larger) to prevent negative results and crashes..

Explain exactly what you do if you still have problems. It can't be that difficult

[Gilloo]

Originally Posted by FastRobPlus

What's the risk of using a thumbdrive?

The main risk is to lose your data, and give me a trial. But I'm not of the kind to fatten the judges and lawyers. So you are warned

Originally Posted by FastRobPlus

Are they all unreliabale? Or just some?

All thumbs use flash ram technology, which are not a model of reliability.
It is supposed to replace floppies, but sometimes floppies or disks are more robust.
So I recommend to use a sata/usb converter (20€ = 20$ according Apple) and a 2.5' sata disk (from a dammaged PC portable for example), PFS3 formatted (partition size up to 80GB) This works with Workbench 1.3 and higher.

[Gilloo]

Originally Posted by Toni Wilen

I said in non-JIT mode it is only supported in latest betas..

where is the latest beta? please, I will do the test..

Originally Posted by Toni Wilen

What Amiga address?

anything in fastmem, or local variable on stack (amiga emulated I mean)

Originally Posted by Toni Wilen

What did 105 return?

Not got the time to see if 105 returns something, the emulator was crashing... (Windows is collecting data... ... ...)

Originally Posted by Toni Wilen

Why do you need "Z3 offset"?

Not my own use, but Wazp3D use it!

Originally Posted by Toni Wilen

Any Amiga address (chip or fast or whatever RAM) + 105 function return value = correct address. Everything is linear mapped (except IO or some other non-RAM special regions).

If you say that amiga memory is linear mapped in the emulator, I thrust you. But inside the source code I got (1.4.4.0) it's not the case.

Originally Posted by Toni Wilen

btw, make sure you use unsigned address variables because returned address can be very high in 32-bit address space if using 64-bit Windows. (like 0x7fxxxxxx or larger) to prevent negative results and crashes..

ok, all in ULONG 32bits

something outside WinUAE, but related with ANAIIS.
Toni, have you fix something for PFS3 WB1.3 ? It guruhize when unmounting...

[Toni Wilen]

Originally Posted by Gilloo

where is the latest beta? please, I will do the test..

Check winuae forum, can't miss it..

anything in fastmem, or local variable on stack (amiga emulated I mean)

Anything that is RAM. (chip, slow, fast, whatever)

Not got the time to see if 105 returns something, the emulator was crashing... (Windows is collecting data... ... ...)

Then your problem is totally different.

If you say that amiga memory is linear mapped in the emulator, I thrust you. But inside the source code I got (1.4.4.0) it's not the case.

It is because JIT direct mode requires it. (JIT takes the Amiga address and adds static offset to make it PC address)

something outside WinUAE, but related with ANAIIS.
Toni, have you fix something for PFS3 WB1.3 ? It guruhize when unmounting...

I think I already said: show me a test case. Sending ACTION_DIE is enough? Does official PFS work?

[Gilloo]

ok ! got WinUAE 242beta

68030 + JIT
Must activate option Host / Miscellaneous / "allow native code"

If I put my hostusb.dll into winuae_dll it logs that:
42-154 [1444 212-262]: native open: 'D:\vieux\Gilles\amiga\WinUAE_242b\plugins\hostusb.dll' = 126
42-154 [1444 212-262]: fallback native open: 'D:\vieux\Gilles\amiga\WinUAE_242b\winuae_dll\hostusb.dll' = 126

If put hostusb.dll into Plugins
And take my program, replace MEMF_CHIP by MEMF_ANY => ok

BUT!
On 68000 512KB chip + 4MB fast
Must activate option Host / Miscellaneous / "allow native code"
It crashes the emulator...

So if I want to test A1000 stuff or non JIT configs, I must insist on MEMF_CHIP for struct allocations needed to pass data to the dll.


// for PFS3, I use this
http://aminet.net/package/disk/misc/pfs3ks13

It is quite easy, mount a removable disk with PFS3 on WB1.3 (on a true amiga 68000), then remove it. It makes a guru 3.
The official PFS3 doesn't work at all on WB1.3, because it use some 2.04 and higher exec routines.

[Toni Wilen]

Originally Posted by Gilloo

It crashes the emulator...

So if I want to test A1000 stuff or non JIT configs, I must insist on MEMF_CHIP for struct allocations needed to pass data to the dll.

You still haven't replied what are you exactly doing, what causes the crash? What address? Does the address match "special area" mentioned in winuaebootlog.txt? "Not working" or "It crashes" won't help anyone.

It is quite easy, mount a removable disk with PFS3 on WB1.3 (on a true amiga 68000), then remove it. It makes a guru 3.
The official PFS3 doesn't work at all on WB1.3, because it use some 2.04 and higher exec routines.

You never said it only crashes when running under 1.3, only that pfs3 for ks1.3 crashes in some not-that-specific media change situations.

Does diskchange <devicename>: also cause crash? Trackdisk.device only?

It could be buggy TD_ADDCHANGEINT in trackdisk.device but there are no workarounds, TD_REMOVE can't be used because it removes old change interrupt = DF0: disk changes would stop working.

[Gilloo]

You never said it only crashes when running under 1.3, only that pfs3 for ks1.3 crashes in some not-that-specific media change situations.

I never test the 1.3 version on 3.1... will do it!
The official PFS3 unmount correctly on 3.1.

[FastRobPlus]

Originally Posted by Gilloo

The main risk is to lose your data, and give me a trial. But I'm not of the kind to fatten the judges and lawyers. So you are warned

Thanks - this is good info. I may try a thumbdrive just to see how it works.

[Gilloo]

Originally Posted by Gilloo

I never test the 1.3 version on 3.1... will do it!
The official PFS3 unmount correctly on 3.1.

Verdict:
PFS13 on 3.1 does the same thing as 1.3 when unmount, but guru #8 (guru #3 on 1.3)

[Toni Wilen]

Does attached version work better?

I noticed that removable support wasn't fully enabled (TRACKDISK define) which can explain disk change problems.

WARNING: This is test version only. Do not upload anywhere else!

[Gilloo]

Originally Posted by Toni Wilen

Does attached version work better?

I noticed that removable support wasn't fully enabled (TRACKDISK define) which can explain disk change problems.

WARNING: This is test version only. Do not upload anywhere else!

Toni, I've test this version 03-Sep-2012, and it makes guru #8 when unmounting in both cases... (wb3.1 and wb1.3) (Sorry to announce bad news)

[Toni Wilen]

Originally Posted by Gilloo

Toni, I've test this version 03-Sep-2012, and it makes guru #8 when unmounting in both cases... (wb3.1 and wb1.3) (Sorry to announce bad news)

Works for me. Explain your test case _exactly_. Not just "I remove disk and it crashes".

EDIT: My test case:

Mount PFS3 floppy device (using similar mountlist used for RAD, except device and handler changed)
Format it.
Eject and insert and check that everything works. Including disk getting reported as removed, not non-dos after removal.

Previously media changes didn't work at all and it also randomly crashed after eject.

[Gilloo]

Originally Posted by Toni Wilen

Works for me. Explain your test case _exactly_. Not just "I remove disk and it crashes".

EDIT: My test case:

Mount PFS3 floppy device (using similar mountlist used for RAD, except device and handler changed)
Format it.
Eject and insert and check that everything works. Including disk getting reported as removed, not non-dos after removal.

Previously media changes didn't work at all and it also randomly crashed after eject.

aah, kiitos
mount uh2:
open the uh2: icon... display image or open a text file, close the application.
unmount uh2: then guru #8

http://aminet.net/package/disk/misc/unmount-0.1
Given there is the source code, I've done the equivalent unmount for 1.3 (which works with fat95, when fat95 is decided to work )

[Toni Wilen]

So it is ACTION_DIE problem after all. Nothing to do with removing removable media..

[Gilloo]

Originally Posted by Toni Wilen

So it is ACTION_DIE problem after all. Nothing to do with removing removable media..

Is it possible to got ACTION_DIE packet with pfs3_ks13 ? kiitos

[Toni Wilen]

Aminet unmount_m68k is broken, first instruction in executable is illegal instruction.

[Gilloo]

Originally Posted by Toni Wilen

Aminet unmount_m68k is broken, first instruction in executable is illegal instruction.

It seems to be compiled for other thing than pure 68000...
Toni, you can use the ANAIIS unmount (seek it in bin of the following package)
http://aminet.net/package/driver/other/anaiis
Good luck

[Toni Wilen]

pfs3 unmount is now working but there is still one problem remaining: under 1.3, mounting again after unmount crashes with memory list corruption guru.

More debugging to do..

Attach source of your version of unmount, thanks.

[Gilloo]

Hi Toni,
here's my unmount code with the makefile... for SASC 6.0 and up
Perhaps it's buggy. Please tell me if you find monsters

unmountc.c is the shell command source code,
unmount.c is the code to unmount function for 1.3 and for 2.04 and up.

[Toni Wilen]

I think calling c:assign is too ugly solution..

assign dismount: internally calls RemDosEntry() which is as simple as:

    dl = DeviceNode head
    while(dl->dol_Next)
    {
        struct DosList *dl2 = BADDR(dl->dol_Next);
        if(dl2 == dlist)
	{
	    dl->dol_Next = dlist->dol_Next;
	    break;
	}
	dl = dl2;
    }

(Forbid state required of course)

You should first find the correct device node and then exit the loop because doslist may have changed and "DeviceNode = (struct DeviceNode *)BADDR(DeviceNode->dn_Next) ;" may access random memory.

It is not safe to scan doslist without forbid state. (and forbid in this code would be broken anyway during packet calls)

But I don't think this is the cause for crash.

[Gilloo]

So I can remove the Execute call to "assign dismount" (I don't think it is working with 1.3) and "assign remove" then.

[Toni Wilen]

Originally Posted by Gilloo

So I can remove the Execute call to "assign dismount" (I don't think it is working with 1.3) and "assign remove" then.

Yeah, just manually unlink the doslist entry and your done.

[Toni Wilen]

Ok, this one should really work.

EDIT: also includes all SVN updates since previous pfs3ks13 Aminet release.

[Gilloo]

Kiitos Toni, I will try this weekend

I don't know why I wrote a Delay(10) before RemDosEntry... it seems to be superfluous...

[Gilloo]

hum... tested the new pfs_ks13

wb1.3: it freezes after unmount... (but no guru :-) )
wb3.1: guru #8

wb3.1, original PFS3 from the PFS3, guru #8 after unmount

Perhaps its my unmount in cause... so I broke the version
if (1 /*DOSBase->lib_Version < 37*/)
to activate always the 1.3 compatible code, but it is the same with pfs_ks13 and wb3.1/original PFS3 unmount correctly

Toni, do you also got the freeze in 1.3?

[Toni Wilen]

Works perfectly here, including 1.3, using your unmount. You sure you use correct pfs3 version?

EDIT: Run mungwall, must be wrong version if you get huge amount of memory allocation errors when unmounting. (You can use -log -serlog winuae command line parameters to see serial debugging without extra utilities)