Context switch throws out

[leonp at plris.com (Leon Pollak)]

OK, I found the answer by myself :-))

The interrupt service code of mbx8xx BSP was written in assumption that memory 
translation and caching are ON. I decided to turn them OFF for the starting 
phase, but there is some code, which turns them on unconditionally.

Thanks to those who read this. :-))

On Sunday 07 September 2003 15:06, Leon Pollak wrote:
> Hello,
> 	Slowly, slowly, I came to the phase of the first context switch on my
> mbx860 derived board. It the uses new_exception_processing, as I was able
> to determine.
> 	With the bdm interfaced debugger I can follow the execution up to the end
> of _CPU_Context_switch function, where everything seems to be OK. Before
> the last instruction execution (blr) the link register contains the address
> of the _Thread_Handler function. I suppose that this is OK and make one
> stepi command.
> 	The result is very strange: the PC jumps somewhere inside
> push_normalized_frame routine (?), while the assembler code at the address
> of _Thread_Handler looks corrupted.
> 	All my attempts to determine when the _Thread_Handler becomes corrupted
> shows that this happens exactly after the blr command execution, which
> seems to be noncense...
>
> Any hint about where to look which be VERY highly appreciated...:-))

-- 
leonp at plris dot com