Re: [Openocd-development] How to debug a program not run at it's runtime address

[Duane E. <op...@du...>]

dswei wrote:
> Hello, all.
>   A program, before it can run at it's runtime address, it has to configure the memory, copy itself to the memory,
> or enable the MMU; after these, it can run at it's runtime address.
>    My question is, how can I debug it with OpenOCD and gdb, when it does not run at it's runtime address?  
>    This is a example, when I debug the linux kernel, the gdb command 'break' can't work before the MMU
> is enabled ,  the address of the break point does not exist yet. My method is: telnet the OpenOCD, 'step' to 
> the instuction 'turn_on_mmu. Just after the MMU is enable, I can set the break point use GDB. 
>    But this method is difficult for newbie.
>    Do somebody have a easy way?    
>   
Yes, it is quite easy. Set a HARDWARE instruction break point at that 
address, and RUN.

Note the difference - I said an instruction hardware break point - not 
soft break point, and not read or write break point.

Depending on your OS - it may need to *LOAD* the application into 
memory. Often a "write breakpoint" would happen application is loaded 
from FLASH, or from TFTP etc.

Also note: Depending on your OS - if it supports "demand page virtual 
memory" (ones I have been involved with did) you often require an OS 
hack that forces the application to be resident in memory, otherwise GDB 
tries to set a break point on an non-resident memory location and 
faults. Yuck.

BTW - this also lets you run & debug other operating systems like the 
linux kernel - that you TFTP boot because it takes too damn long to JTAG 
a 1MEG operating system image into memory.

-Duane.