A4000T SCSI, 53C710 vs 53C810

mark_k · 01 December 2013, 13:55

I noticed that the Qemu 1.7 changelog mentions that it now supports emulation of the LSI 53C810 SCSI controller. Could that be relevant/useful for possible future A4000T SCSI emulation? (A4000T SCSI uses the 53C710 chip.)

Toni Wilen · 16 December 2013, 12:31

"Support" seems to only mean 53C895A is faked as 53C810 (PCI IDs changed).

Comment says 53C895A is 53C810 + new features, so perhaps C810 is C710 + new features, without dropping any obsoleted features. Hopefully.. At least NCR SCSI SCRIPTS is same. (Programming language that chip executes, drivers can use it to program the chip for automatic operation = really complex chip compared to WD SCSI chips for example. This is the most important part.)

joeksterz · 17 December 2013, 17:26

I am NOT an expert at this, but my $0.02...
The c810 is their first PCI controller, so it makes sense that later ones keep compatibility.
But, c710 is NOT PCI, so I really doubt that even if the core is similar - the chip interface and behavior would be significantly different...

Toni Wilen · 28 January 2014, 11:06

It seems 53C720 and 53C810 are mostly identical.

Lots of changes between 53C710 -> 53C720, new registers (number of registers has doubled), some old registeres obsoleted and some register bits moved to other registers. SCRIPTS engine seems to be backwards compatible.

53C710 emulation should still be possible, I think it is time to start hacking QEMU code again

Toni Wilen · 31 January 2014, 09:15

Some success. Managed to get A4091 ROM to initialize NCR chip and it also started executing SCRIPTS code correctly (at least it looks correct, debug logging looks ok, correct looking SCRIPTS opcodes are fetched from memory etc..) and it attempted to select first SCSI device.

Next step: add missing SCSI glue code and connect NCR emulation to UAE SCSI devices and check if selection actually works..

(Yes, not very useful feature but unpatched A4000T KS ROM will hang without SCSI hardware)

EDIT: Works if no SCSI devices connected, select timouts work. Which is better than nothing, it allows original unmodified A4000T ROM to boot properly for the first time.

Toni Wilen · 08 February 2014, 09:49

It works! A4000T and A4091 NCR drivers now boot with HD mounted.

(Most likely non-AOS systems, like Linux, won't yet work, later..)

mark_k · 08 February 2014, 14:18

I'll try testing the A4000T SCSI emulation with Debian and NetBSD.

By the way, the ROM scanner dialog box which shows the found ROMs doesn't have an entry for A4000T yet.

mark_k · 08 February 2014, 14:39

I tried booting my Debian 3.1 HDF with A4000T SCSI 030 and 040 configs. I get an early crash. On 68040 there is much more diagnostic text than with 030.

Booting NetBSD 5.2, it says
scsibus0: waiting 2 seconds for devices to settle...
Then after a delay much longer than that it prints
probe(aftsc0:0:0:0): timed out
aftsc0: abort reset: dstat 80, sstat0 00 sbcl 00
aftsc0: siop id 7 reset V2

Also, even when just booting to the AmigaDOS shell, if I reset with Ctrl-A-A the SCSI functionality seems broken after that. The system sits at the purple insert disk screen. I need to quit and re-run WinUAE to be able to boot from the HDF again.

Toni Wilen · 08 February 2014, 18:55

- NetBSD should work now.
- Reset fixed.
- CD and Tape devices supported.

Debian still does not work, I don't know whats wrong. SCRIPTS code gets bogus data pointer and length after first device is selected.

EDIT: Internet seems to say Linux A4000T/A4091 support seems to be broken in most kernel versions.

EDIT2: It seems SCRIPTS relocation is broken, locations that should contain absolute addresses still contain small relative value (created by SCRIPTS compiler).

mark_k · 09 February 2014, 14:40

I tested the Debian HDF with 2.7.1b7, A4000T/040 config. The error messages are different to the previous winuae.exe, see pics. (Just for interest/information, if Debian is broken on an actual A4000T.)

Toni Wilen · 09 February 2014, 15:11

0000002f probably is another case of unrelocated (base address not added) 53c710 chip address.

Errors probably are different on real hardware because 53c710 emulation does not catch invalid SCRIPTS (if reserved bits are set and so on) that real 53c710 would detect and abort with error.

Toni Wilen · 10 February 2014, 09:41

If there is Debian RDB HDF that works standalone, I can test NCR SCSI with real A4000 + A4091 combination.

mark_k · 10 February 2014, 13:37

You can try the HDF I mentioned in the 2.7.0 beta thread here. That's a ~138MB archive.

Or an HDF without X (smaller ~119MB archive) here. I can't edit that posting to add a multiupload link (because the thread is closed?), so: multiupload link

Both HDFs unpack to 4,000,000,000 bytes. If you don't have a hard disk that large to connect to your A4000 I can remake a smaller installed Debian HDF.

By the way: is the A4000T SCSI non-AutoConfig (at fixed address), whereas the A4091 is AutoConfig?

Toni Wilen · 10 February 2014, 16:12

Copying 4G to old SCSI disk would take forever.. 500M or so would be usable. Largest SCSI-2 drive I have is 2G. I also have ACard SCSI to IDE adapters but I'd prefer not to have unknown variables in device chain.

A4000T is at fixed address (0xdd0040), A4091 is Z3 autoconfig card.

mark_k · 10 February 2014, 18:54

I've created a 16MB hardfile which shows the problem.
, Multiupload

By the way, A4000T SCSI emulation doesn't seem to boot from a bootable FFS CD image. Someone reported to me that a real A4000T can boot from that.

Toni Wilen · 10 February 2014, 19:25

Quote:

Originally Posted by mark_k

I've created a 16MB hardfile which shows the problem.

Thanks.

Quote:

By the way, A4000T SCSI emulation doesn't seem to boot from a bootable FFS CD image. Someone reported to me that a real A4000T can boot from that.

Fixed. Commands that fail (MODE SENSE with mode parameter that only HDs support) returned incorrect status back to NCR emulation.

mark_k · 11 February 2014, 21:06

I tested an A4000/040 + A4091 config (as opposed to A4000T), and the emulated machine repeatedly resets. Also happens with no SCSI HDF. Part of the log output:

Code:

Reset at 00F800D0. Chipset mask = 00000007
00000000    2048K/1 =    2048K Chip memory
00200000   10176K/0 =   10176K <none>
00BF0000      64K/0 =      64K CIA
00C00000    1024K/0 =    1024K Custom chipset
00D00000     768K/0 =     768K <none>
00DC0000      64K/0 =      64K Battery backed up clock (RF5C01A)
00DD0000      64K/0 =      64K A4000 IDE
00DE0000      64K/0 =      64K Motherboard Resources
00DF0000      64K/0 =      64K Custom chipset
00E00000     512K/0 =     512K <none>
00E80000      64K/0 =      64K Autoconfig [reset]
00E90000     960K/0 =     960K <none>
00F80000     512K/1 =     512K Kickstart ROM (D6BAE334)
=KS ROM v3.1 (A4000) rev 40.68 (512k)
01000000    4080M/0 =    4080M <none>
PAL mode V=50.0804Hz H=15625.0881Hz (227x312+0) IDX=10 (PAL) D=1 RTG=0/0
RTGFREQ: 312*50.0804 = 15625.0879 / 50.1 = 312
Z3MEM (32bit): mapped @$10000000: 128 MB Zorro III fast memory 
   Card 1 remapped 40000000 -> 10000000
   Card 1 (ZorroIII) done.
A4091 BOOT ROM 'A4091 ROM 40.13'
ncr_wput 0044=4800 PC=00FCA5F0
A4091 Z3 autoconfigured at 10000000
Memory map after autoconfig:
00000000    2048K/1 =    2048K Chip memory
00200000   10176K/0 =   10176K <none>
00BF0000      64K/0 =      64K CIA
00C00000    1024K/0 =    1024K Custom chipset
00D00000     768K/0 =     768K <none>
00DC0000      64K/0 =      64K Battery backed up clock (RF5C01A)
00DD0000      64K/0 =      64K A4000 IDE
00DE0000      64K/0 =      64K Motherboard Resources
00DF0000      64K/0 =      64K Custom chipset
00E00000    1536K/0 =    1536K <none>
00F80000     512K/1 =     512K Kickstart ROM (D6BAE334)
=KS ROM v3.1 (A4000) rev 40.68 (512k)
01000000     240M/0 =     240M <none>
10000000      16M/0 =      16M A4091
11000000     112M/1 =     112M Zorro III Fast RAM
18000000    3712M/0 =    3712M <none>
ncr_lput 00000000=00000000 PC=00F82132
ncr_lput 00000004=00000000 PC=00F82134
ncr_lput 0000000A=00FCA38E PC=00F82142
ncr_wput 000E=0005 PC=00F82146
ncr_lput 00000010=10000020 PC=00F82166
ncr_lput 00000014=10000020 PC=00F8216A
ncr_lput 00000018=18000000 PC=00F82172
ncr_lput 0000001C=07FFFFE0 PC=00F82176
ncr_lput 00000020=00000000 PC=00F8217A
ncr_lput 00000024=07FFFFE0 PC=00F8217C
ncr_lput 00000000=00000400 PC=00F819E8
ncr_lput 00000004=0000089A PC=00F819EA
ncr_lget 00000000=00000000 PC=00F81EA8
.. that line repeats 42 times ...
ncr_lget 00000000=00000000 PC=00F81EA8
ncr_lget 00000000=00000000 PC=00F806E0
ncr_lget 00000000=00000000 PC=00F81EA8
ncr_lget 00000000=00000000 PC=00F81EA8
GUIBase 803x631 (6x14)
Requested GUI size = 841x631 (768x576)
Got GUI size = 840x631
*: C:\users\Public\Documents\Amiga Files\WinUAE\kick40068_a4000.rom:524288 = KS ROM v3.1 (A4000) rev 40.68 (512k)
CRC32=D6BAE334 SHA1=5FE04842D04A489720F0F4BB0E46948199406F49
clipboard: reset (00000000)
memory init end
Reset at 00F83642. Chipset mask = 00000007

Toni Wilen · 11 February 2014, 21:09

Disable Z3 RAM until it is fixed.

mark_k · 11 February 2014, 21:52

Thanks, I'm now using an A4000/040/A4091 config with 64MB motherboard RAM, no Z3 RAM.

Booting NetBSD 5.2 gives an error which doesn't happen with A4000T config.

amilo3438 · 12 February 2014, 16:18

JFI ...
just found the following page and on the bottom of the same an A4091 install disk ...
in case if someone need it -> http://amiga.resource.cx/exp/a4091

01 December 2013, 13:55	#1
mark_k Registered User Join Date: Aug 2004 Location: Posts: 3,344	A4000T SCSI, 53C710 vs 53C810 I noticed that the Qemu 1.7 changelog mentions that it now supports emulation of the LSI 53C810 SCSI controller. Could that be relevant/useful for possible future A4000T SCSI emulation? (A4000T SCSI uses the 53C710 chip.)

31 January 2014, 09:15	#5
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,534	Some success. Managed to get A4091 ROM to initialize NCR chip and it also started executing SCRIPTS code correctly (at least it looks correct, debug logging looks ok, correct looking SCRIPTS opcodes are fetched from memory etc..) and it attempted to select first SCSI device. Next step: add missing SCSI glue code and connect NCR emulation to UAE SCSI devices and check if selection actually works.. (Yes, not very useful feature but unpatched A4000T KS ROM will hang without SCSI hardware) EDIT: Works if no SCSI devices connected, select timouts work. Which is better than nothing, it allows original unmodified A4000T ROM to boot properly for the first time. Last edited by Toni Wilen; 01 February 2014 at 14:13.

08 February 2014, 14:39	#8
mark_k Registered User Join Date: Aug 2004 Location: Posts: 3,344	I tried booting my Debian 3.1 HDF with A4000T SCSI 030 and 040 configs. I get an early crash. On 68040 there is much more diagnostic text than with 030. Booting NetBSD 5.2, it says scsibus0: waiting 2 seconds for devices to settle... Then after a delay much longer than that it prints probe(aftsc0:0:0:0): timed out aftsc0: abort reset: dstat 80, sstat0 00 sbcl 00 aftsc0: siop id 7 reset V2 Also, even when just booting to the AmigaDOS shell, if I reset with Ctrl-A-A the SCSI functionality seems broken after that. The system sits at the purple insert disk screen. I need to quit and re-run WinUAE to be able to boot from the HDF again. Attached Thumbnails

08 February 2014, 18:55	#9
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,534	- NetBSD should work now. - Reset fixed. - CD and Tape devices supported. Debian still does not work, I don't know whats wrong. SCRIPTS code gets bogus data pointer and length after first device is selected. EDIT: Internet seems to say Linux A4000T/A4091 support seems to be broken in most kernel versions. EDIT2: It seems SCRIPTS relocation is broken, locations that should contain absolute addresses still contain small relative value (created by SCRIPTS compiler). Last edited by Toni Wilen; 09 February 2014 at 10:56.

09 February 2014, 14:40	#10
mark_k Registered User Join Date: Aug 2004 Location: Posts: 3,344	I tested the Debian HDF with 2.7.1b7, A4000T/040 config. The error messages are different to the previous winuae.exe, see pics. (Just for interest/information, if Debian is broken on an actual A4000T.) Attached Thumbnails

16 December 2013, 12:31	#2
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,534	"Support" seems to only mean 53C895A is faked as 53C810 (PCI IDs changed). Comment says 53C895A is 53C810 + new features, so perhaps C810 is C710 + new features, without dropping any obsoleted features. Hopefully.. At least NCR SCSI SCRIPTS is same. (Programming language that chip executes, drivers can use it to program the chip for automatic operation = really complex chip compared to WD SCSI chips for example. This is the most important part.)

17 December 2013, 17:26	#3
joeksterz Registered User Join Date: Jun 2007 Location: Florida, USA Posts: 57	I am NOT an expert at this, but my $0.02... The c810 is their first PCI controller, so it makes sense that later ones keep compatibility. But, c710 is NOT PCI, so I really doubt that even if the core is similar - the chip interface and behavior would be significantly different...

28 January 2014, 11:06	#4
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,534	It seems 53C720 and 53C810 are mostly identical. Lots of changes between 53C710 -> 53C720, new registers (number of registers has doubled), some old registeres obsoleted and some register bits moved to other registers. SCRIPTS engine seems to be backwards compatible. 53C710 emulation should still be possible, I think it is time to start hacking QEMU code again

08 February 2014, 09:49	#6
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,534	It works! A4000T and A4091 NCR drivers now boot with HD mounted. (Most likely non-AOS systems, like Linux, won't yet work, later..)

08 February 2014, 14:18	#7
mark_k Registered User Join Date: Aug 2004 Location: Posts: 3,344	I'll try testing the A4000T SCSI emulation with Debian and NetBSD. By the way, the ROM scanner dialog box which shows the found ROMs doesn't have an entry for A4000T yet.

09 February 2014, 15:11	#11
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,534	0000002f probably is another case of unrelocated (base address not added) 53c710 chip address. Errors probably are different on real hardware because 53c710 emulation does not catch invalid SCRIPTS (if reserved bits are set and so on) that real 53c710 would detect and abort with error.

10 February 2014, 09:41	#12
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,534	If there is Debian RDB HDF that works standalone, I can test NCR SCSI with real A4000 + A4091 combination.

10 February 2014, 13:37	#13
mark_k Registered User Join Date: Aug 2004 Location: Posts: 3,344	You can try the HDF I mentioned in the 2.7.0 beta thread here. That's a ~138MB archive. Or an HDF without X (smaller ~119MB archive) here. I can't edit that posting to add a multiupload link (because the thread is closed?), so: multiupload link Both HDFs unpack to 4,000,000,000 bytes. If you don't have a hard disk that large to connect to your A4000 I can remake a smaller installed Debian HDF. By the way: is the A4000T SCSI non-AutoConfig (at fixed address), whereas the A4091 is AutoConfig?

10 February 2014, 16:12	#14
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,534	Copying 4G to old SCSI disk would take forever.. 500M or so would be usable. Largest SCSI-2 drive I have is 2G. I also have ACard SCSI to IDE adapters but I'd prefer not to have unknown variables in device chain. A4000T is at fixed address (0xdd0040), A4091 is Z3 autoconfig card.

11 February 2014, 21:09	#18
Toni Wilen WinUAE developer Join Date: Aug 2001 Location: Hämeenlinna/Finland Age: 49 Posts: 26,534	Disable Z3 RAM until it is fixed.

12 February 2014, 16:18	#20
amilo3438 Amiga 500 User Join Date: Jun 2013 Location: EU Posts: 1,511	JFI ... just found the following page and on the bottom of the same an A4091 install disk ... in case if someone need it -> http://amiga.resource.cx/exp/a4091

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Pcd-50b SCSI Card Reader with A4000T LUNS Problem	crazyegg	support.Hardware	2	10 February 2013 19:01
For Sale: Cyberstorm Mk2 060 + 128MB RAM + SCSI II Module + 18GB SCSI Drive	Zetr0	MarketPlace	9	27 June 2012 22:41
The scsi interface metc vs squirrel scsi conflict	rampartsagain	support.Hardware	29	24 June 2012 17:50
WTB - SCSI Cable Blizzard MkIV SCSI - 50Pin	fitzsteve	MarketPlace	1	30 August 2010 23:10
Blizzard 1260 & SCSI Kit - 1230.scsi upgrade question	Tommo1975	support.Hardware	0	14 August 2010 19:43