Andrew Morton wrote:
> On Wed, 6 Jun 2007 10:58:34 -0700 "Miles Lane" <miles.lane at gmail.com> wrote:
>
>
>> RELOCS arch/i386/boot/compressed/vmlinux.relocs
>> WARNING: Absolute relocations present
>> Offset Info Type Sym.Value Sym.Name
>> c0102029 00535201 R_386_32 00000000 xen_irq_disable_direct_reloc
>> c0102043 00722e01 R_386_32 00000000 xen_save_fl_direct_reloc
>>
I don't know why these two are coming out absolute. I think they should
be relative. I'll see what I can do.
>> c040f81c 0064f101 R_386_32 c045a4a4 __per_cpu_end
>> c040f821 00588001 R_386_32 c0453000 __per_cpu_start
>> c040f82d 00588001 R_386_32 c0453000 __per_cpu_start
>>
These are due to the definition of PERCPU in
include/asm-generic/vmlinux.lds.h. It should read:
#define PERCPU(align) \
. = ALIGN(align); \
.data.percpu : AT(ADDR(.data.percpu) - LOAD_OFFSET) { \
__per_cpu_start = .; \
*(.data.percpu) \
*(.data.percpu.shared_aligned) \
__per_cpu_end = .; \
}
ie, __per_cpu_start/end should be inside the section, rather than
floating about between sections.
> The others _might_ have
> been triggered by use-the-new-percpu-interface-for-shared-data.patch and
> define-new-percpu-interface-for-shared-data.patch,
define-new-percpu-interface-for-shared-data.patch is the source of the
bad definition of PERCPU.
J
