A4000T Freezing Scsi/dma related?

[cehofer]

I have 2-a4000T/060s  OS3.14, equipped the same with PicassoIV and Deneb usb running in PIO.  The main difference between them is the CPU card.  One has the QPak 060 v1 the big card and the other has the Qpak XP.   The one running the Qpak v1 is rock solid but the one running the Qpak XP. 060 is not.  The Qpak XP will run fine for a long time and for whatever reason will crash hard.  On a reboot, it will be scanning the HDs and freeze with the HD light fully on.  Reboot again, it will have to scan the HDs and most of the time boot successfully.   On both systems, I am using the onboard SCSI controller.

I think this is DMA related since everything I read about the onboard SCSI controller is very dma sensitive .   I found recently that the Picasso IV had DMA related issues and had to be at the latests firmware.   Both are running 060v6.  They both had 060v1 and acted the same.   Both a4000s have been recapped.   The SCSI cables and terminations are routed and connected the same.   The scsi switch options are the same.   

Is somebody an Enforcer or other diagnostic software expert to tell me how to capture the event that causes this?   The software error is usually 80000004 which is cpu related.   

It acted the same with OS3.9 bb4.  I was advised to upgrade to 3.14 because in setpatch, it sets the right options for the 060 based on version.  I recently set a jumper only on the XP to backoff dma for interrupts but that did not help.  The quikpak V1 does not have that setting.   I believe there is some collision happening between disk I/O and another process.

[Castellen]

Have you tried a complete memory test?  Boot Amiga Test Kit from a floppy disk and run 'test all memory' for a few hours.
https://github.com/keirf/Amiga-Stuff/releases


It would be worth exchanging the two QuikPak CPU boards between the two A4000T, which would narrow down if the problem follows the CPU board or not.  That would tell you where to look next.


Not directly relevant, though known compatibility issues with the QuikPak boards:
https://bboah.com/index.php?action=artikel&cat=45&id=2093&artlang=en


Lastly, check the two AT power connectors at the main board.  I've worked on several A4000T which had random instability/freezing that was caused by power connector problems.  If there's poor contact in the connector, this can cause excessive heat dissipation, discolouring the white plastic of the connector, and results in an unreliable +5V supply at the main board, which is critical.

[cehofer]

I have tested the memory extensively.   This A4000T I purchased in 2009 new from Software hut.  That's what this is disappointing about this.  I inspected the board a couple years ago when I recapped it.  I dont remember seeing anything weird with the PS but I can look again.   I was going to swap Picassos since i recently heard about dma issues. 

The Quikpak you have linked is the cpu that works without issue.   I hold my breath every time I have to remove the CPU. 

I would like to trap or record the event that is causing this.  Is it software or hardware. 

I was working with Michael McMaster who designed the SCSI2SD.  He had me enable the debugging on the card and this is what he saw.

"Please use scsi2sd-util to set the "speed" limit option to Async 5MB/s. I can see that the logs all show synchronous data transfers, and the hang occurred during a sync transfer. It would be good to know if the problems go away in async mode."

It doesn't seem to matter whether in Async or Sync mode. It still locks up.  Also, both A4000Ts have the SCSI2SD card in it.  I have also removed the card and it still locks up as described.

[alexh]

Being an A4000Ts I imagine both have -11 busters?

[cehofer]

Yes, both have Buster 11.  From what I read that this is a problem in itself because Buster 11 can only have 1 busmaster.

[cehofer]

UPDATE: 

I'm not getting too euphoric yet but it looks like I fixed it.   Yesterday, I swapped Picasso IV cards between A4000Ts but that made no difference.  I then started looking at versions of libraries.  I got advise a couple threads and years back before OS3.14 when I was using OS3.9BB4 to install THORs MMU.  This went very badly and my tower would not boot.  I had to swap around libraries but I ended up using 68060.library  Version 46.4 05/30/18.  I have two versions of this library one named 68060-mmu.library and 68060.library that show the same version and size of 46.4.

The stable A4000T, I never installed THORs MMU.  It was using 68060 Version 2.2 Jeff Boyer 07/26/96.  I checked both A4000T install disks and this is the original 68060.library that came with both machines.  I installed this library back onto the unstable tower and so far so good.  It has been stable for several hours now.  I have been trying everything to crash it.  I have also put both A4000Ts SCSI back to "ASYNC" mode so far without issue as well.

So I had a couple separate issues that was crashing both towers.  First, I use Storm C4 and I use Magic Menu.  I found on one of the support websites that StormC4 and Magic Menu do not play well together at all.  Magic Menu Version 2.30 and 2.35 will cause a collision and lock up the Amiga.  However, H&P and Aminet both have the fix for this.  It is called "MagicMenuPatch".   Once I installed that patch on both, there was no more crashing due to Magic Menu.  If you use Magic Menu, I would advise to install the patch anyway.  It is just an icon added to WBStartup folder. 

http://aminet.net/package/dev/c/StormC_Tools

Second, the unstable A4000T had the wrong 68060.library and I replaced it with the original Jeff Boyer 2.2 version.  For some reason even though THORs MMU was suppose to be intelligent and work for ALL 060s it caused mine to crash randomly.   If Thomas is around and would like to explain or have me test for him, I would be glad to.

Another added note is that both original 060s were version 1.  I replaced them with version 6.  I don't know that really did anything but now I am suppose to have the most stable 060 processor with version 6.

[BozzerBigD]

Glad you're making progress! Just to clarify is Asynchronous faster? I really need to check my SCSI2SD version 6 board is configured correctly to get the best speeds.

[BozzerBigD]

P.S. I doubt a Rev6 060 would make any difference to stability at stock speeds but it does make a big difference with overclocking potential and stability at higher clock frequencies!

[cehofer]

Async basically doubles the speed.  I am getting Sysinfo speeds of 8MB on HD and 10MB on the SCSI2SD drives.   I will have to look to see if I am still limiting the speed on the SCSI2SD card.

One thing I found on my Picasso IV that it was in Zorro II mode instead of ZorroIII.  That doubles the Picassos RAM from 2MB to 4MB. 

[cehofer]

I did a post a couple years back on how to set up the SCSI2SD v6.  It might help you out. 

[BozzerBigD]

Cool! Do you have the link?

[cehofer]

https://forum.amiga.org/index.php?topic=73855.msg841963#msg841963