Problem with MMU again

Problem with MMU again

[Mikko Alutoin]

Hi folks!

I am stuck and need your help.
We have our own board with MPC8260 on it running Linux 2.4.18. I have been building new images for the system routinely for quite a some time, but now I am faced with a strange problem. I applied one of the USAGI projects kernel patches (in order to have IPv4/IPv6 dual-stack) and now something goes very wrong when jumping to start_kernel. I used Abatron BDI 2000 to trace over the rfi instruction at the end of start_here, after which I found myself in the middle of InstructionAddressInvalid in the exception vector - not in the virtual address of start_kernel where I wanted to be, buuaaah! (Simultaneously Abatron complained that 'step timeout detected' or something like that.)
I am confused. How come the boot code now fails to get the MMU working, as the USAGI patch should not have touched any of the kernel initialization code (to my knowledge at least). Could it be because start_kernel is now at a higher adress than it used to? I know that I am not probably giving much to go on here, but I would be glad to even get some hints where to look for more information.

Cheers,
Mikko

..............................................................
 MTV3 Laajakaista - Hauskemman elämän puolesta.
 http://www.mtv3.fi/liittyma/hankinta/laajakaista/


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Problem with MMU again

[Dan Malek]

Mikko Alutoin wrote:

> I am confused. How come the boot code now fails to get the MMU working,

You are chasing ghosts.  There is nothing wrong with the MMU initialization.
There are just times during the initialization that the system and the BDI2000
are not exactly in sync.  This is one of them.

Let the code run, do some post-mortem debugging (like dumping out the
console log buffer for starters).  Work backward from the failure instead
of trying to step into it (pun intended) :-)


	-- Dan


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Problem with MMU again

[Mikko Alutoin]

Hi,

I traced over the rfi-instruction for a reason: I was not able to break at start_kernel, but setting a breakpoint at start_here worked ok. Also, tracing over the rfi with a working kernel took me directly to start_kernel. So I thought MMU was to blame, but little did I know...

Thanks for the reply, I'll try your advice when I get back to work on Monday.

Cheers, Mikko

> You are chasing ghosts.  There is nothing wrong with the MMU
> initialization.
> There are just times during the initialization that the system and the
> BDI2000
> are not exactly in sync.  This is one of them.
>
> Let the code run, do some post-mortem debugging (like dumping out the
> console log buffer for starters).  Work backward from the failure
>  instead
> of trying to step into it (pun intended) :-)
>
>
> 	-- Dan
>
>


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/