HELP : 2 MB Kickstart refuse to start...

[Cosmos Amiga]

I added a second socket on my CD32 with few components to make 2 eeproms available
==> http://leblogdecosmos.blogspot.fr/2016/02/cd32-kick-39-i.html

1 MB Kickstart 3.1 on now two eeproms works fine !

But, 2 MB failed... Zero boot, nothing...
==> http://leblogdecosmos.blogspot.fr/2016/02/cd32-kick-39-ii.html


I built a new 2 MB Kickstart like this :

- $00A80000 (512 KB) = icon.library_46.4.418
- $00B00000 (512 KB) = workbench.library_45.132
- $00E00000 (512 KB) = RomStartCD32 / checkdisk_37.21 / cdstrap_40.13 / cdui.library_40.20...
- $00F80000 (512 KB) = exec_40.9(CD32) / audio.device_37.1 / battclock.resource_39.3 / battmem.resource_39.2...


Where is my mistake ?

Any ideas ?


Thanks,

[Cosmos Amiga]

Did you verify that all the components which you put into ROM are actually designed to run out of read-only memory? I'm not entirely sure that workbench.library V45 will run properly in this context. The same goes for icon.library.

Also worth checking is the order in which the components are initialized by exec (in "exec/resident.i", the "RT_PRI" field), as well as if the flags in the Resident structure are set up properly for the task (the "RT_FLAGS" field).

If you can build a 2 MByte ROM image in memory, perform the same kind of ROM tag scan which the exec cold start code does, can you verify that all components are found, have the proper priorities and the appropriate flags set?

As for the memory ranges you are using I believe that these will work correctly with the CD32 exec.

All libraries are romable, and used since a very long time into $E0 and/or $F8 : the issue is elsewhere...

Maybe the four 512 KB blocs are in a incorrect order ?

I repeat : when I build the 1 MB CD32 Kickstart 3.1 on TWO eeproms ($E0 & $F8), all is fine, the console start and run nicely...

For the 2 MB Kickstart, I begin with $A0, $B8, $E0 and then $F8. I guess the issue is here, because I use 4 blocs now and not 2 anymore...

[Cosmos Amiga]

CD32 has a single 16 bit rom, but the schematic shows you how to hook up a second rom so it works in 32bit mode. Apparently in 16bit mode it's not laid out the same as any other Amiga 16bit rom because high and low words are stored separately.

http://eab.abime.net/showthread.php?t=71509

The thread appears to show that my idea wouldn't work as it would see the same modules in both.

Probably best way would be to hack together a small boot rom that purely sets up a screen and dumps out hex from various addresses and see if you can get that to run.

I use $B0 who is for the FMV module : when I plug it, the FMV's eprom is read...

Maybe the problem is $A8 (extended Rom CD1200)...

[Cosmos Amiga]

The order should not matter. If I remember correctly, the exec coldstart code scans a list of memory spaces for resident tags, i.e. what's defined in "exec/resident.i" and each of these goes into a list, sorted by priority. This means that no matter in which memory space a resident tag is found, the priority defines when it is initialized.

Once the resident tag list is assembled, first the components which precede expansion.library are initialized (that's usually just exec.library, if I remember correctly), then the rest are initialized one after another until eventually something tickles dos.library, kicking off the user-visible part of the system boot process.

The CD32 ROM checks the following areas for resident tags, in this specific order:

0x00F80000 - 0x00FFFFFE (512K Kickstart ROM)
0x00E00000 - 0x00E7FFFE (CD32 ROM)
0x00A80000 - 0x00B7FFFE (CD32 FMV ROM)
0x00F00000 - 0x00F7FFFE (diagnostic cartridge)

Note that only the CD32-specific Kickstart ROM looks into the CD32 ROM and FMV ROM address spaces. The desktop Kickstart ROM only checks the Kickstart ROM space and the diagnostic cartridge areas.

I understand what you write !

My main problem is : the exec don't start with my rom build !

Zero, nada, quedalle, nein, no, nothing at power on...

[Cosmos Amiga]

I do some testing on my 1200 this time, and if my understanding is good I guess the 2 eproms U6A/U6B are limited by hardware to 1 MB maximum...

$A8 seems mirrored to $E0...