Sonnet 7200 PCI programming

[BULI]

I send to you my light- partition, configured for your hardware, but I afraid that problem could be with your cpu card (I don't have it)
Try and good luck.

[gsoravil]

Originally Posted by BULI

I send to you my light- partition, configured for your hardware, but I afraid that problem could be with your cpu card (I don't have it)
Try and good luck.

Thanks Buli. I just responded to the PM.

[gsoravil]

Wrangler & Buli,

I just noticed something. When trying to start things manually, if I go to early startup menu and "Disable CPU Caches" then initPPC no longer locks up but instead returns with [CLI 2] to the prompt. This also seems to be consistent. I think this is a clue.

I saw on https://github.com/Sakura-IT/SonnetAmiga/releases that "The library supports options through ENVARC:sonnet". I'm not sure I understand what that means. If I try to call sonnet driver like this: sonnet Debug=1 ...then I get a parameter error. So I'm ignorant here.

Wrangler - to update you - I tried a clean setup (thanks to Buli) and have still been getting the exact same symptom: All PCI card seen, machine boots, hangs at initPPC. (Exception is the above "disable caches" I decided to try assuming there's a bus conflict.) Note that I did try cards in different order but that made no difference. And I did only put the two cards (PPC, VooDoo).

Originally Posted by Wrangler

gsoravil: to begin with, remove all the PCI cards except the Voodoo 3 and the PRPMC cards. These two are the most stable pair. If/when you get these working, you can add the other ones in.

Some users have reported crash/lockup on running InitPPC based on the order of the PCI cards in the slots. This is still not totally understood AFAIK but you need to keep trying those two cards in different combinations of slots (move one card at a time) until you find a combination that works. There is also some suggestion that you might need a PCI card in the final slot from the Mediator logic board to properly terminate the PCI bus.

Try this out using the no startup-sequence test described on the wiki page.

I'd also suggest re-copying the powerpc.library to your set up in case it was corrupted when you installed it. This bit me with Warp3D a while back, even though the file size was correct.

Also: keep the case lid off while doing your testing. Heat - especially from the Voodoo card - will cause spurious crashes. My cards are bad for this especially on a hot day, and it's summer here...

[jbenam]

@gsoravil

I'm pretty much having the same symptoms your Amiga shows with my Mediator v1 + Killer K1. I'm starting to think it might be an issue with v1 Mediators.

I actually haven't used my A4000 in months (I think I last tried getting this thing to work during Xmas) since it's still entirely tore apart on my working desk in the hope that I'll someday get the PPC to work.

I have pretty much tried everything - clean install with 3.1, 3.9 and 3.1.4, always the same issue. Freezing on InitPPC. I tried any and every combination I could. I also tried with a Radeon 9200 in place of my Voodoo 3. It's either a timeout or a freeze.

Sometimes I do get the RAM added to AmigaOS, only for everything to freeze as soon I open something (either a folder or an executable). PSU has been recapped by me and voltages are perfect (they actually were even before recapping, but I recapped in hope of getting the M1 to work...).

I also tried injecting 3.3V straight to the Mediator (thought I had a faulty power regulator) and with an ATX power supply by building an adaptor, nothing helped.

Not sure where to go from here, except to get another Mediator.

[gsoravil]

Yeah in my previous post I was overzealous because I inadvertently typed "run initppc" instead of "initppc" so it was kicking off the initppc in the background, but still hanging (even with cache disabled via early startup).

I'm glad I'm not alone. It must be a v1 issue. I did the same thing - modded to add 3.3v to a slot just so I could try a Killer NIC. The NIC powers up and there's these lovely twinkling red LEDs on it after it starts itself (all pre-Sonnet/pre-boot). My machine works perfectly fine outside of this but there must be something different about the v1. I wonder if we could just get a new daughter-card rather than a whole new....wait a minute. I wonder if there is any other lines - DMA maybe? on the PCI bus that are also missing on the V1.

Buli/Wrangler/Anyone - Can I get a picture of the back of your Mediator?

...I wonder if there are other disconnected lines on the V1 beyond just the 3.3v line.

Hmmm.

Also - if anyone has had success using a Mediator V1 with or without 3.3v mod, please let us know!

Originally Posted by jbenam

@gsoravil

I'm pretty much having the same symptoms your Amiga shows with my Mediator v1 + Killer K1. I'm starting to think it might be an issue with v1 Mediators.

I actually haven't used my A4000 in months (I think I last tried getting this thing to work during Xmas) since it's still entirely tore apart on my working desk in the hope that I'll someday get the PPC to work.

I have pretty much tried everything - clean install with 3.1, 3.9 and 3.1.4, always the same issue. Freezing on InitPPC. I tried any and every combination I could. I also tried with a Radeon 9200 in place of my Voodoo 3. It's either a timeout or a freeze.

Sometimes I do get the RAM added to AmigaOS, only for everything to freeze as soon I open something (either a folder or an executable). PSU has been recapped by me and voltages are perfect (they actually were even before recapping, but I recapped in hope of getting the M1 to work...).

I also tried injecting 3.3V straight to the Mediator (thought I had a faulty power regulator) and with an ATX power supply by building an adaptor, nothing helped.

Not sure where to go from here, except to get another Mediator.

[BULI]

Originally Posted by gsoravil

Yeah in my previous post I was overzealous because I inadvertently typed "run initppc" instead of "initppc" so it was kicking off the initppc in the background, but still hanging (even with cache disabled via early startup).

For clarity.
This System from me, booting correct on Voodoo (without init PPC)?
Do you have on start, picture of Workbench screen on GFX Voodoo3 (no PAL)?

[trixster]

I would guess that if BULI’s known-good working setup has not fixed your issues then it’s almost certainly a hardware problem and not a software/OS/patch conflict.

[jbenam]

One thing I still need to test is to build that 68060 adaptor I bought months ago. Maybe it will work with a 060, instead of just freezing?

@BULI
Do you mind sending me your known-good config? Worth a try, I suppose. Thanks!

[gsoravil]

Originally Posted by BULI

For clarity.
This System from me, booting correct on Voodoo (without init PPC)?
Do you have on start, picture of Workbench screen on GFX Voodoo3 (no PAL)?

It boots without the PPC card in with VooDoo active. I simply get a message from initPPC that there's no PPC card.

I could modify the startup-sequence to remove the initPPC to avoid lockup at that step when the PPC card is installed, just to prove it boots otherwise, but I know that's the problem. I had booted with no startup-sequence and manually ran each step until it locked up (initPPC step). This is always where I locked up when I was trying with my original setup.

jbenam saw our conversation and posted that he's also having the same problem and like me he had a Mediator V1. But somehow both him and I must have independently believed that the 3.3v was the only hurdle because we both separately added 3.3v to one slot to try the Killer NIC. The thing is, even my PMC card, which supplies its own 3.3v and powers up in any of my slots still exhibits this same behavior.

I can mod the startup-sequence and remove the initPPC if you think there's value.

Greg

[gsoravil]

Originally Posted by trixster

I would guess that if BULI’s known-good working setup has not fixed your issues then it’s almost certainly a hardware problem and not a software/OS/patch conflict.

At this point, I'm in agreement with you Trixter. I think me and jbenam are sharing an issue where maybe these don't work in a Mediator V1 even with a 3.3v mod (only needed for Killer NIC - my PMC supplies its own 3.3v).

I have reached out to Elbox to see if it's possible to upgrade just the bridge card and keep the backplane. However, if I can get a picture of the back of anyone's V2 or V3 backplane, I can see if maybe it's just a case of some other missing lines, such as address lines between cards.

[gsoravil]

Originally Posted by jbenam

One thing I still need to test is to build that 68060 adaptor I bought months ago. Maybe it will work with a 060, instead of just freezing?

@BULI
Do you mind sending me your known-good config? Worth a try, I suppose. Thanks!

I have already done this. I have custom ROMs and an 060 adapter which I had installed in my A3640. I had the exact same issue. I put back the 040 and original ROMs only to troubleshoot.

There's an outside possibility that the way I built the CF from Buli's files is "dirty" with additional files present (I installed a base WB 3.1 and unpacked his config over it). I should redo it as a last-ditch effort.

To be honest, I am pretty well convinced it's a Mediator V1 problem based on you sharing that specific bit of hardware, especially because we're using different PPC cards.

I will let you know how the CF rebuild goes as soon as I can.

[trixster]

it's not a good idea to try and startup the ppc with initppc from the startup-sequence unless you have a fully stable system. some cards seem to require a few minutes 'warm-up' time before they behave.

Wrangler found with one of his systems is that there was a problem with the elbox logic board, so i wouldnt be surprised if your issues stem from the mediator itself

[gsoravil]

Buli/jbenam/Trixter,

To be absolutely certain, I just cleared the CF card and re-tried the config from Buli and it's exactly the same. Additionally, I commented out the initPPC and it boots, and if I then launch a shell and run initPPC from the sonnet folder (same as startup-sequence does) then InitPPC hangs.

To be clear, though, apparently the rest of the system does not hang. i.e. Mouse still moves, and I can open windows. Just InitPPC sits there indefinitely. It's as if the PPC card isn't able to communicate back on the PCI bus on its own. This would make sense to me if it's a limitation of the Mediator V1.

My understanding of all of this is that InitPPC copies PPC code to the VGA memory and tells the PPC card to run it, and that InitPPC then waits for the memory to change (the 68K CPU polls for the change). I don't know if that's right, but it makes me believe that the PPC isn't able to do its job and InitPPC is just patiently waiting for a response. Either that or the PPC is never running the code.

[gsoravil]

Originally Posted by trixster

it's not a good idea to try and startup the ppc with initppc from the startup-sequence unless you have a fully stable system. some cards seem to require a few minutes 'warm-up' time before they behave.

Wrangler found with one of his systems is that there was a problem with the elbox logic board, so i wouldnt be surprised if your issues stem from the mediator itself

I can try this modified setup and wait a little longer (for card to init). Good to know about the logic board.

Wrangler - On the system that you had a problem with, what was your symptom?

[Hedeon]

try the latest beta library of the rewrite. It has debugging output.

[jbenam]

Originally Posted by gsoravil

I have reached out to Elbox to see if it's possible to upgrade just the bridge card and keep the backplane.

I've tried contacting them as well, and after one initial reply they just disappeared.

To be fair my enquiry was about their upgrade service for the V1s. It seems that at one point they upgraded the CPLDs on the V1s for free with a newer version.

Maybe we both have the same CPLDs revisions?

@Hedeon
I've been waiting for the newer version since you told me about it months ago - hopefully it will bring us closer to the discovery of this long standing issue.

I'll see if I can squeeze a bit of time and sit at my desk tomorrow.

[gsoravil]

Originally Posted by Hedeon

try the latest beta library of the rewrite. It has debugging output.

Do you mean download the latest beta from https://jenkins.sakura-it.pl/job/sonnetamiga/ ? I just tried this now and same symptom (unless I did it wrong).

Would like jbenam to try it though, just to rule me out as a problem.

jbenam, let me know if you do ever hear of a way to get updated CPLDs for the V1. I didn't know that was once an option. Maybe if we both pester Elbox...

[BULI]

Originally Posted by gsoravil

Do you mean download the latest beta from https://jenkins.sakura-it.pl/job/sonnetamiga/ ? I just tried this now and same symptom (unless I did it wrong).

Would like jbenam to try it though, just to rule me out as a problem.

jbenam, let me know if you do ever hear of a way to get updated CPLDs for the V1. I didn't know that was once an option. Maybe if we both pester Elbox...

Try this one:
https://eab.abime.net/showpost.php?p...postcount=3986

This beta library is on my system too.
In drawer sys:expansion/sonnet/sonnet
name:
powerpc.libraryNewBeta
Just change the name powerpc.library to whatever and rename powerpc.libraryNewBeta to powerpc.library, and init it like trixster wrote, after fully start Workbench.

I don't have A3640, but I had in the past, and I know that this card may generated problem with MMU table/cache, so if you have option, try check Mediator, and different CPU card too (like WarpEngine, or CeberStorm)

@jbenam I sent link to my system via mail, if you want to try.

[jbenam]

@BULI
With your config I get
ramlib
Program failed (error #80000004)

Which AFAIR is something that happened on my config as well.

[trixster]

What processor library are you using? Have you tried mmulibs? What does BULI’s Config use?