Compiling assembler file with ppc_405-gcc

[Garcia Jérémie]

Hi everybody,
although I'm a newbie in Linux developement and even more in LinuxPPC development, I have to work on a Linux PPC assembler based application.
In order to achieve this, I need to compile an assembler file (ppcsiafn.s) which now gives nightmares ...
Below this line you can dicover this file:

>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
/*DESCRIPTION
This module contains chip-dependent routines written in assembly language.
*/

#include <linux/config.h>
#include <asm/processor.h>
#include <asm/cache.h>
#include <asm-ppc/ppc_asm.h>

#define _ASMLANGUAGE

.globl sidma_AsDmaRegSet_F
.globl sidma_AsDmaRegGet_F
.globl AsfnDmapolGet
[etc...  ]

.text

/*******************************************************************************
* sidmai_AsDmaRegSet_F(reg, val) --> Set specified "reg" with value "val"
*
* Return : N/A
*
* void sidma_AsDmaRegSet_F(uLong reg, uLong val)
*/
sidma_AsDmaRegSet_F:
       mflr r5
       bl next_inst
       next_inst:
       mflr r6
      add r6,r6,r3
      addi r6,r6,20

[etc...  ]

<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<

To compile this, I use the following lines in a makefile:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
AS = ppc_405-gcc
CFLAGS_AS = -gdwarf -mcpu=405 -mstrict-align -nostdinc -fvolatile -fno-builtin -fno-for-scope -P -x assembler - D__KERNEL__ -DCPU=PPC405 -DPRJ_BUILD $(ALLINCLUDE) -I$(SRC_PATH)/basic/services/include -I$(SRC_PATH)/basic/ascii/include -I$(SRC_PATH)/basic/hdwctl/include -I$(SRC_PATH)/basic/core/include -I$(SRC_PATH)/basic/hscx/include -I$(SRC_PATH)/basic/enet/include -I$(SRC_PATH)/eq/include -I$(SRC_PATH)/pzm/include -I$(SRC_PATH)/startup/init -I$(SRC_PATH)/startup/maint -O0 -DEQ_CODE
ppcsiafn.o:
      $(AS) $(CFLAGS_AS) -c $(SRC_PATH)/startup/init/ppcsiafn.s -o ppcsiafn.o
 <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<

nb: don't worry the CFLAGS_AS is on a single line in my makefile

All that things to get the following compilation result:
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
$> make all
/home/jgarcia/Desktop/kernel/ssc/startup/init/ppcsiafn.s: Assembler messages:
/home/jgarcia/Desktop/kernel/ssc/startup/init/ppcsiafn.s:83: Error: unsupported relocation against r5
/home/jgarcia/Desktop/kernel/ssc/startup/init/ppcsiafn.s:86: Error: unsupported relocation against r6
/home/jgarcia/Desktop/kernel/ssc/startup/init/ppcsiafn.s:87: Error: unsupported relocation against r6
/home/jgarcia/Desktop/kernel/ssc/startup/init/ppcsiafn.s:87: Error: unsupported relocation against r6
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<

After a lot of tests, here are my conclusions: the different "#include" are not considered by the compiler. Indeed, I tried to introduce compilation errors in the different ".h" files --> no difference in the result
If I suppress the # in the #include --> the compiler tells me that is not a valid op code
So I guess that but I don't know why : the character "#" is understood as a comment (equivalent to /* ...*/).

To conclude : is there a specific compil option to give to gcc for it to understand the #include command?
I checked the different linux assembler file and the #include is used. So why am I not able to use it?

Thanks for your help !

                                                                Jérémie


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