Uncompress Ok, but cannot run linux kernel...

Hello:

    I have a 750-based custom board. I want to port linux-2.4.2 on it. The
bootrom code which initialized the hardware is working OK. I want to
download the linux kernel via TFTP(and it is OK now) and run it. But I met
some urgent problems ...
    I modified the bootloader codes according to arch/ppc/mbxboot/head.S and
other "printf" functions to suit my cases. And it runs like the following:

################################
Loading... 611056
Starting at 0x400000...

loaded at:     00400000 004952F0
zimage at:     004062C8 0048F994
avail ram:     004952F0 10000000

Linux/PPC load:
Uncompressing Linux...done.
Now booting the kernel
###################################

Then nothing happens(I have added a serial output code at the very begining
of arch/ppc/kernel/head.S and  have nothing output.). By printing the code
from 0x0 to 0x400000 I find the uncompressed code is right. After
uncompressing it jumps to 0x0(arch/ppc/mbxboot/head.S), but it cannot run .
Why?
Then I write a simple HelloWorld program  and compile it with -Ttext=0x0
compiling options, then use it to replace /usr/src/linux/vmlinux and
recompile the kernel.This time it works OK like the following:
################################
Loading... 4326
Starting at 0x400000...

loaded at:     ..........
zimage at:     ..........
avail ram:     ..........

Linux/PPC load:
Uncompressing Linux...done.
Now booting the kernel

Hello world
###################################

Here I attach my .config and arch/ppc/mbxboot/head.S. Hope you can help me
if you are free.

Thank you very much.

machael thailer

machael thailer wrote:

..... But I met
some urgent problems ...
    I modified the bootloader codes according to arch/ppc/mbxboot/head.S

For starters, you shouldn't be using this "boot loader" for a 750.....
How did you configure a kernel to get there in the first place?
A 750 should be a CHRP (or PReP, if you have to.....), which will
direct you to a working "chrpboot" boot loader.


	-- Dan

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

At 09:01 AM 5/6/2001 +0800, machael thailer wrote:

    I have a 750-based custom board. I want to port linux-2.4.2 on it. The
bootrom code which initialized the hardware is working OK. I want to
download the linux kernel via TFTP(and it is OK now) and run it. But I met
some urgent problems ...

I have the same problem on an FADS board. Tracing head.S showed me that the
code returns with
a return from interrupt instead of a branch. So it maybe works if the image
is used without bootloader or by changing head.S.

I'll test it today, please let me know if you've found the problem.


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

Michael Habermann wrote:

I have the same problem on an FADS board.

What "same" problem?  How are you loading and tracing the kernel?

......... So it maybe works if the image
is used without bootloader or by changing head.S.

The image certainly won't work without any bootloader, whether it
is PPCBoot or the code found in arch/ppc/mbx (or arch/ppc/boot/mbx
depending on your source base).  The kernel requires a bunch of
registers and information set up before it is called, and one of
those two methods must be used to ensure this happens.

I believe there is a working FADS port in the FSM Labs BitKeeper
sources.  If not, you can download the older 2.4.0-test2 LSP from
MontaVista and use it as a guide for updating a later kernel.  The
changes are very minor.


	-- Dan

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

machael thailer wrote:

quoted

..... But I met
some urgent problems ...
    I modified the bootloader codes according to arch/ppc/mbxboot/head.S

For starters, you shouldn't be using this "boot loader" for a 750.....
How did you configure a kernel to get there in the first place?
A 750 should be a CHRP (or PReP, if you have to.....), which will
direct you to a working "chrpboot" boot loader.

Yes, I choose "6xx/7xx/74xx/8260" and "/PowerMac/PRep/MTX/CHRP" when
configuring the kernel.
Now I find that something is wrong with my serial output codes. When I fix
it , I can trace into arch/ppc/kernel/head.S now.
But it hangs after running "turn_on_mmu". That is to say, it should be able
to run "start_here" (but it cannot .)
By the way , I find that "initial_bats" do 2 different branches according to
whether PVR is 601 or 604. I don't know which branch will 750 do?

Thank you very much.

machael


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

At 09:16 PM 5/7/2001 -0400, Dan Malek wrote:

quoted

I have the same problem on an FADS board.

What "same" problem?  How are you loading and tracing the kernel?

The same problem is that the kernel doesn't survive until the welcome
messge is printed.
The kernel is loaded with ppcboot and the tftpboot commando. To trace, I
used mpc8bug.

I use the Montavista linux-2.4.0-test2-1.2.0-1 kernel.
I set a breakpoint at the entry of the __start routine. After passing
the turn_on_mmu label, it returns with an rfi and hangs.
It is maybe ment to be started from a power_on interrupt, because in
this context, I don't see the sense for an rfi. But I don't know much details.


I believe there is a working FADS port in the FSM Labs BitKeeper

sources.  If not, you can download the older 2.4.0-test2 LSP from
MontaVista and use it as a guide for updating a later kernel.  The
changes are very minor.

The Montavista is the one I used. I also tried one from Bitkeeper, but I
didn't found support
for the FADS board. Even when I compiled for MBX, I got no image in
arch/ppc/coffboot.
Which is the machine type I should configure for a MPC860FADS?

It would be nice if you could help me. So far I couldn't start any kernel.


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

At 09:16 PM 5/7/2001 -0400, Dan Malek wrote:

quoted

quoted

I have the same problem on an FADS board.

What "same" problem?  How are you loading and tracing the kernel?

The same problem is that the kernel doesn't survive until the welcome
messge is printed.
The kernel is loaded with ppcboot and the tftpboot commando. To trace, I
used mpc8bug.

I use the Montavista linux-2.4.0-test2-1.2.0-1 kernel.
I set a breakpoint at the entry of the __start routine. After passing
the turn_on_mmu label, it returns with an rfi and hangs.

This is exactly what I meet now.
If you have good solutions, please let me know.

Thank you very much.

machael


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

machael thailer wrote:

quoted

..... But I met
some urgent problems ...
    I modified the bootloader codes according to arch/ppc/mbxboot/head.S

For starters, you shouldn't be using this "boot loader" for a 750.....
How did you configure a kernel to get there in the first place?
A 750 should be a CHRP (or PReP, if you have to.....), which will
direct you to a working "chrpboot" boot loader.

If I use  "chrpboot " bootloader, I have to provide many functions in my
boot codes according to CHRP specifications:

1  "prom"(promptr) functions which can provide with "/chosen" ,"claim" and
"getprop" services etc in arch/ppc/chrpboot/start.c and in
arch/ppc/kernel/prom.c .
2  device-trees for my custom board.

Are these functions all necessary for booting the linux kernel?


And do I have to  modify arch/ppc/mm/init.c=>MMU_init() and
arch/ppc/kernel/setup.c=>identify_machines() to suit my cases?
I notice that in identify_machines(), there are many types initializations
functions such as pmac_init(),prep_init(),chrp_init(),apus_init(),
oak_init(),m8xx_init(),m8260_init(). Do I have to provide my own ***_init()
functions?


Thank you  very much.

machael


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