Hi Alex,

Hi Andrea,

in s3c44b0x's app03-test.pdf, the stacks in boot code are initialized like this:

;****************************************************
;* Initialize stacks *
;****************************************************
ldr sp, =SVCStack ;Why?
bl InitStacks
......
;****************************************************
;* Initialize stacks    *
;****************************************************

InitStacks ;Don't use DRAM,such as stmfd,ldmfd...... ;SVCstack is initialized before ;Under toolkit ver 2.50, 'msr cpsr,r1' can be used instead of 'msr cpsr_cxsf,r1' mrs r0,cpsr bic r0,r0,#MODEMASK orr r1,r0,#UNDEFMODE|NOINT msr cpsr_cxsf,r1 ;UndefMode ldr sp,=UndefStack

orr     r1,r0,#ABORTMODE|NOINT
msr     cpsr_cxsf,r1       ;AbortMode
ldr     sp,=AbortStack

orr     r1,r0,#IRQMODE|NOINT
msr     cpsr_cxsf,r1       ;IRQMode
ldr     sp,=IRQStack

orr     r1,r0,#FIQMODE|NOINT
msr     cpsr_cxsf,r1       ;FIQMode
ldr     sp,=FIQStack

orr     r1,r0,#SVCMODE|NOINT
msr     cpsr_cxsf,r1       ;SVCMode
ldr     sp,=SVCStack

......

Thanks you very much for the piece of code, I have find the original Samsung document and now I can understand what you are talking about! (well.. it has been a long time since I have read that doc! :-)

But in U-boot there isn't any code like this.

In u-boot there is no need of that code because the only thing u-boot does when there is an exception is to reset the board (ARM architecture does not need stack because it uses banked register on exception, so only user code needs it? am I wrong?). BTW I think the reset code is not right because it just jump to the reset vector, which may not be what you want to reset the board (I think the better way is to call cpu_reset() or something like that and use the watchdog, don't you think so?)

BTW, we have run a uclinux in our 44b0x system, but there isn't any code for initializing stacks in uclinux. What should the cpu do if any of the exceptions happen?

So I think the code like this are not very reliable. Is my comprehension right? Or if we should initialize these stacks before call start_kernel() in uclinux?

Don't really know how the ARM Linux kernel works regarding interrupt and exception stack... are you really sure that it does not initialize stack? Now I don't allocate stack for interrupt but interrupts works pretty well :-) Anyway I'll have a close look at the source code as soon as I can...

BTW did you found some ARMLinux docs where it says that ARMLinux bootloader have to set up stack before calling the kernel? (I'll have another look at them too as soon as I can...)

Best Regards, Andrea

Best Regards,

Alex Zu ----- Original Message ----- From: "Andrea Scian" andrea.scian@dave-tech.it To: "Alex Zu" Seu_huizu@tom.com Cc: u-boot-users@lists.sourceforge.net Sent: Friday, January 14, 2005 7:13 PM Subject: Re: Questions about startup code for S3C44B0 CPU-core

...

Hi Alex,

...

Hi Andrea, thank you for your explanation but the stacks of EXCEPTION MODEs which we discussed about is still not very clear. U-boot here does not use interrupts,so there is no need for such initialization for interrupts,but we can"t insure that there is no core exception.In case of these exception,the system may go wrong. So I think that we should initialize the stacks for those exceptions.

BTW, even if not used into the bootloader, interrupt vector are

initialized

...

by u-boot (take a look at cpu/s3c44b0/start.S)

...

And there is no initialization in linux, either.What should linux do in case of these exceptions. So there is the initialization in other common start codes of s3c44b0x.

Something is still not clear to me... could you please tell me where I can find some example of such code (or where this initialization is explained into the ARM or Samsung manual)? Thanks.

Best Regards,

Andrea Scian