exxos's Atari, Amiga & retro forum

terriblefire wrote: ↑Thu Jan 31, 2019 7:44 am I agree. Its not the biggest issue.. The biggest issue is the disable jumpers. I'd like to figure it out though.

I agree too as you know what some Amigan`s are like Lol

x4 TF330 spoiler below

Ok.

I put one of those 8Gb cards (clean) into my TF330.. and the SDCard adapter you gave me before xmas. Both have the same issue. Soft reboot always hangs.

To be clear. I open "Shell" and type reboot. Thats how i reproduce the issue.

EDIT: I'm using a machine without a CD Drive attached.

terriblefire wrote: ↑Thu Jan 31, 2019 8:54 am Ok.

I put one of those 8Gb cards (clean) into my TF330.. and the SDCard adapter you gave me before xmas. Both have the same issue. Soft reboot always hangs.

To be clear. I open "Shell" and type reboot. Thats how i reproduce the issue.

EDIT: I'm using a machine without a CD Drive attached.

Thanks Steve

Right Did the same as you Clean cf same as you have opened shell ...then REBOOT and it rebooted , did it again and it REBOOTED, third Time it did Hang !

So can now 100 % can confirm even on a clean system the REBOOT will not work...But very infrequenty it may , so it must be close as it will REBOOT on the very odd ocassion, so you need to tweek something very slightly Steve

Ok summary of findings so far.

Basically there is an INT2 source stuck on the CD32 when you do a soft reboot.. and its not the TF330.

Digging a bit further it seems that the interrupt acknowledge doesnt work after a soft reboot... highly odd.
Really hoping this isnt a new PCB revision time

OK found the issue i think.... not good.

I need to trace where AVEC comes from on the CD32 motherboard ... it should just be a pass through to the 030 AVEC pin..
For reasons beyond me the CD32 chipset is continuing to assert IPL lines after the interrupt acknowledge cycle has completed. I've tried a few things like delays to these cycles. Nada.

I'm also very very suspicious of the CD32 AVEC line. Its pulled low and never never goes high.

terriblefire wrote: ↑Thu Jan 31, 2019 1:14 pm OK found the issue i think.... not good.

I need to trace where AVEC comes from on the CD32 motherboard ... it should just be a pass through to the 030 AVEC pin..

akiko_avec.JPG

For reasons beyond me the CD32 chipset is continuing to assert IPL lines after the interrupt acknowledge cycle has completed. I've tried a few things like delays to these cycles. Nada.

I'm also very very suspicious of the CD32 AVEC line. Its pulled low and never never goes high.

Just got my fluke multimeter for fast continuity check and the AVEC is connected to pin 16 on the 020 CPU

My next tests will need to be disabling the 030 and looking at what happens during an IACK with the 020 running on the scope.

supaduper wrote: ↑Thu Jan 31, 2019 9:33 am So can now 100 % can confirm even on a clean system the REBOOT will not work...But very infrequenty it may

For me it's 100% of the time, weird.

It does not occur on the 328, oddly enough...

It doesnt occur on the TF330 with the 030 disabled.

Interrupt acknowledge cycles look identical. I really would have expected Akiko to have disasserted the IPL lines after the IACK.
What i know is that INT2 is being asserted continuously when you soft reboot. This interrupt is NOT coming from the TF330 (i checked with the scope by echoing this to a different pin).

When the interrupt occurs the OS is calling CardRes->StatusChangeInt() which cannot service the interrupt because its some other interrupt source and then returns from interrupt.. rinse and repeat.

Why this doesnt occur on the 020.. no idea....

When the 030 is accessing the Motherboard it just looks like a 14Mhz processor on the bus. Locking the machine to 14mhz makes no difference either.

One thing I was wondering from the hanging code were the addresses DA9000 DAA000 (which seem to be for Gayle interrupt), is how your IDE works.

Assuming it's still Gayle emulation in the CPLD, how does that interact with the expectation that Akiko is probably set up to handle those addresses instead of Gayle?

With the 020 only maybe the the gayle emulation isn't fully intercepting the interrupt request, meaning Akiko gets them as well. But when the 030 is in use, maybe they are hidden from Akiko and the hand is waiting on Akiko to do the interrupt acknowledge?

Just a thought on an area to consider - I wish I actually knew what was going on... Hope you crack it!