Platform: x86
Compiler: gcc-3.4.6
Binutils: 2.17.50.0.8
Kernel under build: 2.6.25 (from kernel.org)
Problem: Build fails due to incompatibility of asm-offsets.h with
assembler. See below:
gcc -Wp,-MD,arch/x86/kernel/.entry_32.o.d -nostdinc -isystem
/apps/gcc/3.4.6/lib/gcc/i686-pc-linux-gnu/3.4.6/include -D__KERNEL__
-Iinclude -include include/linux/autoconf.h -D__ASSEMBLY__ -m32
-DCONFIG_AS_CFI=1 -DCONFIG_AS_CFI_SIGNAL_FRAME=1
-Iinclude/asm-x86/mach-default -c -o arch/x86/kernel/entry_32.o
arch/x86/kernel/entry_32.S
include/asm/asm-offsets.h: Assembler messages:
include/asm/asm-offsets.h:10: Error: invalid character '_' in mnemonic
include/asm/asm-offsets.h:11: Error: invalid character '_' in mnemonic
include/asm/asm-offsets.h:12: Error: invalid character '_' in mnemonic
include/asm/asm-offsets.h:13: Error: invalid character '_' in mnemonic
include/asm/asm-offsets.h:14: Error: invalid character '_' in mnemonic
include/asm/asm-offsets.h:15: Error: invalid character '_' in mnemonic
(snip - much more of the same)
Analysis:
1. All errors are related to use of offsetof() macro. The asm-offsets.h
contains lines like
IA32_SIGCONTEXT_ax $44 offsetof(struct sigcontext, ax) #
2. offsetof() macro is defined in include/linux/stddef.h. It appears
to rely on __compiler_offsetof() which maps to __builtin_offsetof()
on gcc-4.x, but on gcc-3.4.6 the definition is as follows
#define offsetof(TYPE, MEMBER) ((size_t) &((TYPE *)0)->MEMBER)
I believe this C syntax is what is causing indigestion to gas.
3. The compilation can be forced to proceed bit further by
adding to beginning of .S files #define __ASM_OFFSETS_H__
which prevents the problematic include from being included.
However, some .S files actually need the definitions.
4. In Documentation/Changes file the compiler requirement is
documented as gcc-3.2. Therefore it seems reasonable to expect
gcc-3.4.6 to work.
5. Last time I compiled kernel was 2.6.23.9. Back then this
problem did not exist. The toolchain is the same as used back then.
6. googling did not return any enlightening entries, thus I presume
this problem has not been reported yet.
Cheers,
--Sampo
--
To unsubscribe from this list: send the line "unsubscribe linux-kbuild" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
