Re: MIPS Assembly.

[Chris Jones <cjones at sutus dot com>]

Paul;

Patching uaccess.h  to wrap j 2b as follows:

.set noreorder
.set nomacro
j 2b
.set macro
.set reorder

still causes problems. 

If the code is "correct" and it is the link that is the problem, then I
would like to know so I can address the symptoms in the best fashion.

Here's a test module that you can compile on MIPS that should generate
the errors I'm talking about.

gcc -I/location_of_my_kernel_includes -mips32 -DMODULE -D__KERNEL__ -c
moduletest.c

Cheers,

   -=chris


On Mon, Jun 21, 2004 at 04:06:14PM -0400, Paul Koning wrote:
> >>>>> "Thiemo" == Thiemo Seufer <ica2_ts@csv.ica.uni-stuttgart.de> writes:
> 
>  Thiemo> ...A possible alternative might be to really use jumps in that
>  Thiemo> case, but I don't see how this can be done sanely without a
>  Thiemo> symbol as jump target.
> 
> How about using ".set nomacro"?  It already means "don't change my
> code, do what I told you".  If it doesn't work that way in ALL cases,
> it should.  This is just another such case.
> 
> Alternatively, fix the linker so it complains if the branch is out of
> range so the assembler doesn't have to reject this code?
> 
> 	paul
> 

-- 
                  Chris Jones | Software Developer | Sutus, Inc.
                 t: +1.604.987.8866 x 2204 | e: cjones@sutus.com


#include <linux/kernel.h>
#include <linux/module.h>


#include <linux/ppp_channel.h>


int init_module() {
	int j=1;
	unsigned long data;
   int foo = 1024;
   data = &foo;
   //OFFENDING k_call -> get_user() : See <asm/uaccess.h>
   // Generates "j 2b" assembly
   get_user(j,(int *)data);
	return 0;
}

void cleanup_module() {
}