A600 won't stop using CF card as system memory

[jimbob]

I agree it shouldn't make a difference but something is up and this will tell if loading ROM from file is the problem.

I don't know if or how skick verifies the ROM files.

One thing I'm not clear on, does the amiga boot ok on 3.1 when no PCMCIA card is inserted? And then cards work normally?

[demolition]

Originally Posted by jimbob

I agree it shouldn't make a difference but something is up and this will tell if loading ROM from file is the problem.

I don't know if or how skick verifies the ROM files.

One thing I'm not clear on, does the amiga boot ok on 3.1 when no PCMCIA card is inserted? And then cards work normally?

If I take a wrong kick rom file and rename it, Skick will not accept it, so I think it is checking the checksum. If it weren't 100% correct, the patch that it applies would not be right.

And yes, if I boot it with kick3.1 without anything in the PCMCIA, I can use the PCMCIA slot normally afterwards to mount my FAT formatted CF card.

[jimbob]

Hi, I just tested this with an ACA630. Seems that something funny does happen when a 3.1 kickstart is mapROMed from file with a card in pcmcia.

with no maprom, cold boot is always normal

with -maprom * cold boot is always normal

with -maprom <3.1file> cold boot is normal with card slot empty but fails with something in PCMCIA, (tried both a memory stick adaptor and a wireless network card.)

I can see the disk activity as it loads the ROM file, then an automatic reset and it stays at a black screen until I do an additional soft reset then everything is normal. I couldn't get to an early startup that had CC0 missing so I didn't see what you described as wrongly allocated PCMCIA fast RAM. (maybe we see slightly different effects here because of different ACA cards or because I am going from kick 3.1->kick 3.1 whereas you are going from kick2.05->kick3.1)

With -maprom <2.05file> cold boot works normally, (I think, it doesn't actually load workbench because it is a 3.1 classicWB but a screen comes up with an error and a shell prompt.) You should check loading a 2.05 file see if the problem goes away. Then we will know it is some weird interaction between PCMCIA/KS3.1/maprom from file.

For me one soft reset fixes the problem without removing the card, did you say you had to remove the card to get it to boot 3.1 at all?

[demolition]

Originally Posted by jimbob

Hi, I just tested this with an ACA630. Seems that something funny does happen when a 3.1 kickstart is mapROMed from file with a card in pcmcia.

I can see the disk activity as it loads the ROM file, then an automatic reset and it stays at a black screen until I do an additional soft reset then everything is normal.

With -maprom <2.05file> cold boot works normally, (I think, it doesn't actually load workbench because it is a 3.1 classicWB but a screen comes up with an error and a shell prompt.) You should check loading a 2.05 file see if the problem goes away. Then we will know it is some weird interaction between PCMCIA/KS3.1/maprom from file.

For me one soft reset fixes the problem without removing the card, did you say you had to remove the card to get it to boot 3.1 at all?

When I had the 4GB card in PCMCIA it would not boot at all with 3.1. The screen just stayed black after it loaded the kick31 file and did a reset. If I then do a soft reboot, it loses the maprom and reboots in 2.05. You should try and see if it actually reboots with maprom enabled or if you're simply seeing your hard-ROM.

I will try loading a 2.05 file later to see if that will reproduce the error as currently it's running some stability tests.

[jimbob]

Originally Posted by demolition

When I had the 4GB card in PCMCIA it would not boot at all with 3.1. The screen just stayed black after it loaded the kick31 file and did a reset. If I then do a soft reboot, it loses the maprom and reboots in 2.05. You should try and see if it actually reboots with maprom enabled or if you're simply seeing your hard-ROM.

I will try loading a 2.05 file later to see if that will reproduce the error as currently it's running some stability tests.

According to acatune -status, mapROM is activated for me after that soft reset.

[demolition]

So I just did a transrom of my 2.05 rom into the HDD. Using maprom (both * and from file) results in mapping the PCMCIA CF card as system memory. Doesn't happen if maprom is turned off, so there is definately something up with that maprom function.

[jimbob]

Oh wait, didn't you say maprom * worked ok before?

Also I don't think you mentioned which version of ACAtune you are using. I just realised I was using version 1.0. V1.7 seems to work perfectly for me.
Sorry, I'm not sure how helpful I'm being, we're not really comparing apples with apples anyway since we have difference ACA cards.

[demolition]

Originally Posted by jimbob

Oh wait, didn't you say maprom * worked ok before?

Also I don't think you mentioned which version of ACAtune you are using. I just realised I was using version 1.0. V1.7 seems to work perfectly for me.
Sorry, I'm not sure how helpful I'm being, we're not really comparing apples with apples anyway since we have difference ACA cards.

Yeah, I did write that maprom * worked before, but I think that must have been with the other CF card in the PCMCIA slot.

I'm using v 1.7 and yes, it's not easy to compare since our cards are different as well as the pcmcia peripherals we're testing with. I don't have anything else but CF cards to test with in the PCMCIA port.

The only thing we can conclude is that maprom seems to mess with the pcmcia port (or maybe the fact that maprom makes the KS boot a lot faster?).

[jimbob]

Originally Posted by demolition

(or maybe the fact that maprom makes the KS boot a lot faster?).

Have you tried with the maprom <p> option?

[Amiga1992]

I think you should write to Jens Schoenefeld or better, lure him into this thread by PM, to solve this strange issue. I am interested myself. I want an ACA620 but if it doens't maprom I am not sure.

[demolition]

Originally Posted by jimbob

Have you tried with the maprom <p> option?

The <p> option doesn't change the issue unfortunately. I can see the delay inserted in the KS as it shows a different color on screen.

Now that I also have an A1200 with ACA1232 I tried whether the problem was the same with maprom, but there is no problem at all with the PCMCIA port and my CF card here.

[Schoenfeld]

Funny - Demolition is already in contact with me, but only pointed me to this thread through the news thread.

Obviously, Kickstart is making a mistake here: As soon as "something PCMCIA" is found, it's checking 0x60.0000 for memory. With the ACA620, it will actually find memory there, because the accelerator will not switch it off without the maxmem option. It's just not added to the Freemem list. This was the only way to get all the options into the logic chip - with a full switch-off, it doesn't fit.

Demolition: I have just sent a message to my girlfriend who was supposed to take your test-board of the ACA620 to the post office today. I'll try to cram the "switching off" of the accelerator memory into the logic - it still won't switch the memory off, but we may be able to work around this Kickstart bug by write-protecting the upper 4M memory block.

Jens

[demolition]

Originally Posted by Schoenfeld

Funny - Demolition is already in contact with me, but only pointed me to this thread through the news thread.
Obviously, Kickstart is making a mistake here: As soon as "something PCMCIA" is found, it's checking 0x60.0000 for memory. With the ACA620, it will actually find memory there, because the accelerator will not switch it off without the maxmem option. It's just not added to the Freemem list. This was the only way to get all the options into the logic chip - with a full switch-off, it doesn't fit.

How come it is not a problem then when booting from the hardware ROM with the ACA620 installed? It only happens when KS is Maprom'ed.

[Schoenfeld]

I haven't read through the whole thread, but from what I see, you're using kickstart files. Does the behaviour change if you're using * (the ROM contents)?

In any case, I'm currently on that logic part that protects the upper 4M of Zorro space against writing, so you can further investigate. I wouldn't be surprised if some open bus is recognized as memory, because a fast access is reading an open bus, which is mistaken for RAM due to the high speed it's being accessed: If kickstart is in Fastram, there is no other access between the w/r pairs to an open bus, so it will read back it's own data word that was written just before. The difference is that with maprom, there's no action on the bus, and therefore the old data stays there.

Kickstart's memory recognition routines are buggy, that's a fact. ACA620 already has some workarounds in place, so the computer will stop looking for memory at $d0.0000. The "maxmem" option can be considered another workaround, and while I've been looking for empty memory spaces, I found that the "reserved" area for a DMA controller should not contain memory any time, because the computer would crash in that case. Isn't that funny? Kickstart is crashing because a certain address space contains memory.

More stories about Kickstart's memory recognition: Back when we developed ZorRAM, it was a 128M card with a 256M option. Worked fine with the 030 processor, but not with 040 processor: The 128M card would be mistaken for a 256M model, because the buggy routines of Kickstart would recognize memory if an open bus was found. We had to implement a discharge-cycle for the local bus on the card in order to work around this Kickstart bug.

I'd expect something similar here, and I'm pretty sure that it'll be fixed with the write-protect idea.

Jens

[demolition]

Originally Posted by Schoenfeld

I haven't read through the whole thread, but from what I see, you're using kickstart files. Does the behaviour change if you're using * (the ROM contents)?

In any case, I'm currently on that logic part that protects the upper 4M of Zorro space against writing, so you can further investigate. I wouldn't be surprised if some open bus is recognized as memory, because a fast access is reading an open bus, which is mistaken for RAM due to the high speed it's being accessed: If kickstart is in Fastram, there is no other access between the w/r pairs to an open bus, so it will read back it's own data word that was written just before. The difference is that with maprom, there's no action on the bus, and therefore the old data stays there.

Yes, the problem is the same if I use -maprom *, but your open bus explanation seems plausible as that would explain the difference between running from a real ROM and maprom.

[Schoenfeld]

The modded ACA620 has reached Demolition's place and I just got word that this problem is fixed now.

Jens