Re: arm-softfloat gcc3.3.3 internal_relocation error

[Allen Curtis <acurtis at onz dot com>]

On Jun 9, 2004, at 11:01 PM, Dan Kegel wrote:

> Allen Curtis wrote:
> > I used the demo-arm-softfloat.sh script to build a GCC 3.3.3 cross 
> > compiler. I thought that everything was working fine until I ran into 
> > this error:
> > arm-softfloat-linux-gnu-gcc -D__ASSEMBLY__ -D__KERNEL__ 
> > -I/home/acurtis/localdev/thales/ipcu/linux-2.4.21/include -mapcs-32 
> > -D__LINUX_ARM_ARCH__=4 -march=armv4 -msoft-float   -c -o getuser.o 
> > getuser.S
> > getuser.S: Assembler messages:
> > getuser.S:47: Error: internal_relocation (type: OFFSET_IMM) not fixed 
> > up
> > make[2]: *** [getuser.o] Error 1
> > make[2]: Leaving directory 
> > `/home/acurtis/localdev/thales/ipcu/linux-2.4.21/arch/arm/lib'
> > Google searches indicate that this is a compiler problem. I am 
> > building a 2.4.21 kernel with the following patches:patch-2.4.21-rmk1
> > patch-2.4.21-rmk1-omap
> > patch-1540/1 (from arm.linux.org for float compile issues)
> > Linux 2.6 for OMAP and u-boot compiled successfully. I thought I was 
> > home free.
> > Anyone know of a patch?
>
> Hadn't heard of any.  Can you post the first 50 lines of getuser.S?
/*
 *  linux/arch/arm/lib/getuser.S
 *
 *  Copyright (C) 2001 Russell King
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 as
 * published by the Free Software Foundation.
 *
 *  Idea from x86 version, (C) Copyright 1998 Linus Torvalds
 *
 * These functions have a non-standard call interface to make them more
 * efficient, especially as they return an error value in addition to
 * the "real" return value.
 *
 * __get_user_X
 *
 * Inputs:	r0 contains the address
 * Outputs:	r0 is the error code
 *		r1, r2 contains the zero-extended value
 *		lr corrupted
 *
 * No other registers must be altered.  (see include/asm-arm/uaccess.h
 * for specific ASM register usage).
 *
 * Note that ADDR_LIMIT is either 0 or 0xc0000000.
 * Note also that it is intended that __get_user_bad is not global.
 */
#include <asm/constants.h>

	.global	__get_user_1
__get_user_1:
	bic	r1, sp, #0x1f00
	bic	r1, r1, #0x00ff
	ldr	r1, [r1, #TSK_ADDR_LIMIT]
	sub	r1, r1, #1
	cmp	r0, r1
1:	ldrlsbt	r1, [r0]
	movls	r0, #0
	movls	pc, lr
	b	__get_user_bad

	.global	__get_user_2
__get_user_2:			/* THE PROBLEM IS HERE. 2.6 USES DIFFERENT REGS, WILL 
TRY THAT */
 	bic	ip, sp, #0x1f00
 	bic	ip, ip, #0x00ff
 	ldr	ip, [ip, #TI_ADDR_LIMIT]
 	sub	ip, ip, #2
 	cmp	r0, ip
2:	ldrlsbt	r1, [r0], #1
3:	ldrlsbt	ip, [r0]
#ifndef __ARMEB__
 	orrls	r1, r1, ip, lsl #8
#else
	orrls	r1, ip, r1, lsl #8
#endif
	movls	r0, #0
	movls	pc, lr
	b	__get_user_bad


> Also, does gcc-3.4.0 fail in the same way?

GCC 3.4 failed right out the starting gate. I didn't spend much time 
debugging based on the comments in the demo-arm-softfloat.sh file. 
Changed the script to build the above mentioned configuration and 
plowed forward.


------
Want more information?  See the CrossGCC FAQ, http://www.objsw.com/CrossGCC/
Want to unsubscribe? Send a note to crossgcc-unsubscribe@sources.redhat.com