Re: XAmiga Current Crash (blizzard 1230mkIV, 64MB): Info from core

[Roc Vallès <vallesroc%gmail.com@localhost>]

On Thu, Dec 5, 2013 at 12:53 PM, Radoslaw Kujawa
<radoslaw.kujawa%c0ff33.net@localhost> wrote:
>
> On 5 Dec 2013, at 12:58, Roc Vallès <vallesroc%gmail.com@localhost> wrote:
>
>> Still a load of nonsense. But this time I enabled ddb on panic, so I
>> did a few things there and took "screenshots". They contain a
>> sane-looking backtrace, so it might help.
>>
>> http://b.rvalles.net/unsorted/netbsd/Xamiga_panic1.jpg
>> http://b.rvalles.net/unsorted/netbsd/Xamiga_panic2.jpg
>> http://b.rvalles.net/unsorted/netbsd/Xamiga_panic3.jpg
>
> Hmm, it looks like either something calls VIOCREMOVE or VIOCSSIZE ioctl and 
> this causes view to be removed. However, the function removing the view fails 
> to deallocate the chip memory in free_chipmem(), possibly because of access 
> to address for which physical to virtual mapping was never added...
>
> Did you perhaps try to start X with “startx” instead of calling Xamiga 
> directly? I wonder if it would change anything.
>
Test was done with Xamiga, but I know both crash.

> Also, please try rolling back:
> src/sys/arch/amiga/amiga/cc.h to version 1.16
> src/sys/arch/amiga/amiga/cc.c to version 1.22
> and recompile the kernel.
>
> I’m not sure if it will compile, but is worth trying. If it won’t change 
> anything update them to newest version again.
>
Will look into it.

>> Upon "continue", it generated the following crash dump, which is
>> similar to the prior nonsense ones. Something must wrong with making
>> dumps, savecore or perhaps netbsd/m68k current's gdb.
>
> I have to admit I didn’t use gdb to analyse the core on m68k for years. Maybe 
> other hackers would be able to comment?
>
Perhaps I'll try using crossdev to get a working m68k toolchain on my
workstation's Gentoo Linux. Then try the cross gdb. Or perhaps I
should get somebody to try IDA for me.