Software failures 68040 CPU on Amiga 4000 EC030

[astremler]

Originally Posted by AmigaHope

What happens if you boot with caches disabled in early startup?

I have just tried with caches disabled and now I am able to boot into the Workbench 3.1.4 ! Interesting.. But only with cache disabled I can boot, without it I will get the errors I described earlier.

At least I get 1 step further now..

[AmigaHope]

Originally Posted by astremler

I have just tried with caches disabled and now I am able to boot into the Workbench 3.1.4 ! Interesting.. But only with cache disabled I can boot, without it I will get the errors I described earlier.

At least I get 1 step further now..

OK glad I could help!

It sounds like one of your drivers doesn't play nicely with caches, but the 030's caches are too small to cause it a problem. It could be that once Setpatch runs it will correct this.

Once your system is booted (i.e. once setpatch has run), can you run the command "cpu cache" in the CLI to enable all caches, does everything work fine, or does it crash again?

If everything works fine after setpatch, it means you probably need a new ROM that is prepatched.

What sort of memory is in your system? Is it just motherboard memory or do you have memory on a Zorro 3 card?

[astremler]

Originally Posted by AmigaHope

OK glad I could help!

It sounds like one of your drivers doesn't play nicely with caches, but the 030's caches are too small to cause it a problem. It could be that once Setpatch runs it will correct this.

Once your system is booted (i.e. once setpatch has run), can you run the command "cpu cache" in the CLI to enable all caches, does everything work fine, or does it crash again?

If everything works fine after setpatch, it means you probably need a new ROM that is prepatched.

What sort of memory is in your system? Is it just motherboard memory or do you have memory on a Zorro 3 card?

I will try about the cpu cache and let you know. I have 16 MB memory on the motherboard. No memory on Zorro 3 card.

I have found a link of a guy with a similar issue and he seemed to have solved it:

http://forum.hyperion-entertainment....hp?f=15&t=4119

I have followed exactly what this guy wrote from scratch, but the problem got even worse! It doesn't even recognize any 68040.library anymore.

I am going to put the computer back in the state I had this morning so that I could boot into the Workbench with the caches disabled again. Continue investigating. Really curious where this issue is coming from...

[AmigaHope]

Originally Posted by astremler

I will try about the cpu cache and let you know. I have 16 MB memory on the motherboard. No memory on Zorro 3 card.

I have found a link of a guy with a similar issue and he seemed to have solved it:

http://forum.hyperion-entertainment....hp?f=15&t=4119

I have followed exactly what this guy wrote from scratch, but the problem got even worse! It doesn't even recognize any 68040.library anymore.

I am going to put the computer back in the state I had this morning so that I could boot into the Workbench with the caches disabled again. Continue investigating. Really curious where this issue is coming from...

My guess is that some memory range isn't being mapped as noncacheable or non writeback cache, or a burst mode issue. The A3640 doesn't bursts at all onto the motherboard bus and ordinarily doesn't even try (68040 burst mode is not compatible with 68030 bus burst mode, and motherboard is 68030 bus), but you can try adjusting.

If your system is still unstable after you try "cpu cache" to turn on caches, other commands you can try "cpu cache noburst nocopyback" "cpu cache noburst copyback"

MMULib lets you look at how the MMU maps cache control to various pages. If the problem is indeed that map you can manually fix that map as a last resort, but it would help first to figure out where the problem is.

[astremler]

Originally Posted by AmigaHope

OK glad I could help!

It sounds like one of your drivers doesn't play nicely with caches, but the 030's caches are too small to cause it a problem. It could be that once Setpatch runs it will correct this.

Once your system is booted (i.e. once setpatch has run), can you run the command "cpu cache" in the CLI to enable all caches, does everything work fine, or does it crash again?

If everything works fine after setpatch, it means you probably need a new ROM that is prepatched.

What sort of memory is in your system? Is it just motherboard memory or do you have memory on a Zorro 3 card?

I have changed the boot partition to use "long file names" in stead of using directory cache, formatted it and re-installed the workbench 3.1.4.

After install I have copied from mmulib disk the 68040. library,680x0.library, mmu.library, mmu-configuration and copied loadModule from the Module disk to C:. (same what the guy did in the previous link)

Without CPU cache disabled the Workbench failed to load with error "C:Version program failed (#80000004) .

With cache disabled I am able to boot normally. After boot I did this: 'cpu cache' and it shows this:

68040 68882 68040-MMU no FastROM (INST: Cache Burst) (DATA: CopyBack)

It works fine too. In 'Sysinfo' it shows the full speed and caches enabled. So the problem is indeed something about the ROM which probably needs to be pre-patched, but I have no idea how to do.

Is there something more I need to copy from the Modules disk?

[astremler]

Originally Posted by astremler

With cache disabled I am able to boot normally. After boot I did this: 'cpu cache' and it shows this:

68040 68882 68040-MMU no FastROM (INST: Cache Burst) (DATA: CopyBack)

Hmm.. I noticed that after cache enabled I am getting problems again when starting programs : 'Program failed'. For example I did: ed s:Startup-Sequence before 'cpu cache' it works fine. After 'cpu cache' I did the same and getting the errors again..

In the Startup-Sequence there is a line of code about 'Rom update' which is there by default after installation.

[astremler]

I did another test: I booted with cache disabled into the Workbench. I copied C:Ed to DF0:

I opened the shell, open df0: and did 'CPU cache'. When 'Ed <somefile>' it still works fine. Even 'Sysinfo' which was loaded from Disk loaded fine.

But when I go to DH0: and open any program (like Ed) then I am getting 'Program Failed'. So it is definitely a compatibility problem with the IDE bus on the mainboard, because once starting to use the hard disk after cache enabled the problems will start...

[Don_Adan]

Originally Posted by astremler

I did another test: I booted with cache disabled into the Workbench. I copied C:Ed to DF0:

I opened the shell, open df0: and did 'CPU cache'. When 'Ed <somefile>' it still works fine. Even 'Sysinfo' which was loaded from Disk loaded fine.

But when I go to DH0: and open any program (like Ed) then I am getting 'Program Failed'. So it is definitely a compatibility problem with the IDE bus on the mainboard, because once starting to use the hard disk after cache enabled the problems will start...

I have similar problem, i think this is hardware problem, something is too hot if 68040 cpu caches are enabled, because this crash occured randomly after about 10-15 minutes from start cold Amiga 4000T.

[Thomas Richter]

Originally Posted by astremler

I did another test: I booted with cache disabled into the Workbench. I copied C:Ed to DF0:

I opened the shell, open df0: and did 'CPU cache'. When 'Ed <somefile>' it still works fine. Even 'Sysinfo' which was loaded from Disk loaded fine.

But when I go to DH0: and open any program (like Ed) then I am getting 'Program Failed'. So it is definitely a compatibility problem with the IDE bus on the mainboard, because once starting to use the hard disk after cache enabled the problems will start...

The A4000 IDE is a rather harmless device, it doesn't do any DMA, just PIO, and there is nothing on the bus that would easily explain such an issue. For the 68040, the IDE registers look like any other chip registers, and the 68040.library should keep care about it not being cachable. However, is this a physical 3.1.4 ROM or the software/LoadModule solution? In the latter case, please ensure that the boot partition (the one the modules are on and LoadModule is on) must be below the 4GB barrier, or all hell may break loose. Also, if you run the MuTools based solution, an output of "MuScan" *after* SetPatch might be worth to look at. Probably post it here.

[astremler]

Originally Posted by Don_Adan

I have similar problem, i think this is hardware problem, something is too hot if 68040 cpu caches are enabled, because this crash occured randomly after about 10-15 minutes from start cold Amiga 4000T.

This is not the case with my Amiga. I booted with caches disabled. I inserted a floppy disk with Ed, Version, SysInfo and DosControl and went into the shell with DF0:

I enabled the caches by 'cpu caches'. On DF0: I can still open all of these programs on the disk. I waited for more than 1 hour and tried to open these programs again (on Disk). They all work fine. I did a speed test with SysInfo again. It shows the full speed. So it's definitely nothing about overheating.

When I open the same programs on hard disk then the 'Software failures' are coming back no matter which program I open. So once I access the hard disk after enabling caches these problems are showing up...

[astremler]

Originally Posted by Thomas Richter

The A4000 IDE is a rather harmless device, it doesn't do any DMA, just PIO, and there is nothing on the bus that would easily explain such an issue. For the 68040, the IDE registers look like any other chip registers, and the 68040.library should keep care about it not being cachable. However, is this a physical 3.1.4 ROM or the software/LoadModule solution? In the latter case, please ensure that the boot partition (the one the modules are on and LoadModule is on) must be below the 4GB barrier, or all hell may break loose. Also, if you run the MuTools based solution, an output of "MuScan" *after* SetPatch might be worth to look at. Probably post it here.

Yes, I am using physical 3.1.4 roms and the boot partition is less than 4 GB. MuTools I will check later.

[amigasith]

Originally Posted by astremler

I think more it's a hardware/compatibility issues with the IDE bus since there were some customization done by the previous owner. It has changeable IC's which I was told it is not standard. I am not familiar with this.

Perhaps this is the "PIO 2 Mod" for the A4K? Here you can find some info: http://eab.abime.net/showpost.php?p=603519&postcount=40 However, I am not sure if this mod is known to cause issues with 68040's... Perhaps somebody else knows more about the combination of this particular mod and 040's?

[meynaf]

If you boot with caches disabled, and they type :
cpu datacache databurst copyback
how does it behave ?

[astremler]

Originally Posted by meynaf

If you boot with caches disabled, and they type :
cpu datacache databurst copyback
how does it behave ?

It seems to work! At least I am not getting any software failures anymore when opening programs on hard disk. I have opened 'SysInfo' and see attachment for the result.

Since iCache is disabled the speed is still slow, but at least we know this iCache is the problem.

[meynaf]

So it appears it's probably not some memory failure, bad i/o, or software problem.

I am not 68040 specialist but it looks like if the 040 is damaged in some way so its instruction cache doesn't work properly. At least, there appears to be some hardware problem on the board.

[astremler]

Originally Posted by meynaf

So it appears it's probably not some memory failure, bad i/o, or software problem.

I am not 68040 specialist but it looks like if the 040 is damaged in some way so its instruction cache doesn't work properly. At least, there appears to be some hardware problem on the board.

No, this is not the case. See my previous post earlier. With 'cache enabled' it works fine on Floppy. I can start every program including SysInfo (see attachment). Only when it needs to read a program from hard disk then it fails. If there would be a damage on the 68040 then it would not work at all, regardless on floppy, hard disk etc..

[meynaf]

And what if you attempt to load only data from hard disk ?
I mean, do everything from floppy, but have a program load some data from hard disk ?
F.e. you might try to make some archive (on the 030) with a program such as lha (or just download one), and then check it with the 040. Does it report corrupt data ?
Or view some pics, play some mods, with program from floppy and data from hd ?

[chb]

Just a wild guess, but maybe it's a power issue? The 040 draws more power than the 68030, so when doing a HDD access it may overload the (slightly faulty?) power supply and result in unstable voltages and CPU/HDD errors. With icache off probably the power consumption is lower, because the CPU has to wait for the instructions from RAM.

Not very likely probably, but you never know...

EDIT: Do you have the possibility to hook your HDD to an external power supply?

[AmigaHope]

Definitely try the external power supply as mentioned. It could just be that your old PSU is ass and needs recapping or replacement.

Also have you tried enabling all caches but disabling burst and copyback? "cpu cache noburst nocopyback"

[astremler]

Originally Posted by AmigaHope

Definitely try the external power supply as mentioned. It could just be that your old PSU is ass and needs recapping or replacement.

Also have you tried enabling all caches but disabling burst and copyback? "cpu cache noburst nocopyback"

Yes, the suggestion about the PSU is a good one and I will try later.. I ever bought a PSU converter cable from AmigaKIT and I have an ATX power supply laying around somewhere. Currently the original Amiga PSU is inside.

I have tried "cpu cache noburst nocopyback" but the result is the same : software failure. Once the "cache" is turned on then the problems will start with the hard disk.

I have also tried to install Workbench 3.1.4 and the MuTools afterwards. I was able to get it working and boot from hard disk with the cache on. It still wasn't perfect because there was a delay in the boot up of 20 seconds and 6 MB of my Fast memory was already consumed after booting in the Workbench. (only 10 MB left of the 16 MB) But anyway, it worked (for a while).. Wow... It ran very fast and when running 'SysInfo' it showed the full speed.

Cool, I thought.. I turned off the computer (didn't change anything) and turned on the computer again the next day: same problem came back ! Lol... I can only boot with cache disabled again.. Really bizarre..

I will investigate more..