ld:arch/x86/kernel/vmlinux.lds:678: parse error

ld:arch/x86/kernel/vmlinux.lds:678: parse error

[Jean Delvare]

Hi Sam,

I am not able to build Linus' latest kernel on my oldest test system. I
get the following linking error at the end:

  LD      .tmp_vmlinux1
ld:arch/x86/kernel/vmlinux.lds:678: parse error

Which is:

ASSERT((_end - 0xC0000000 <= (512 * 1024 * 1024)),
        "kernel image bigger than KERNEL_IMAGE_SIZE")

Apparently this assertion is relatively new, as I was able to build
kernel 2.6.29 on the same system. The system has binutils 2.14.90.0.6,
which is more recent than what is listed in Documentation/Changes
(2.12).

Any idea?

Thanks,
-- 
Jean Delvare

Re: ld:arch/x86/kernel/vmlinux.lds:678: parse error

[H. Peter Anvin]

[-- Attachment #1: Type: text/plain, Size: 954 bytes --]

On 08/03/2009 05:04 AM, Jean Delvare wrote:
> Hi Sam,
> 
> I am not able to build Linus' latest kernel on my oldest test system. I
> get the following linking error at the end:
> 
>   LD      .tmp_vmlinux1
> ld:arch/x86/kernel/vmlinux.lds:678: parse error
> 
> Which is:
> 
> ASSERT((_end - 0xC0000000 <= (512 * 1024 * 1024)),
>         "kernel image bigger than KERNEL_IMAGE_SIZE")
> 
> Apparently this assertion is relatively new, as I was able to build
> kernel 2.6.29 on the same system. The system has binutils 2.14.90.0.6,
> which is more recent than what is listed in Documentation/Changes
> (2.12).
> 
> Any idea?
> 

Old binutils doesn't accept the naked ASSERT(); there has to be a sink
for the data; so one has to do crap like:

__junk_symbol = ASSERT(...);

or

. = ASSERT(...);

near the end.

Does this patch fix it for you?

	-hpa

-- 
H. Peter Anvin, Intel Open Source Technology Center
I work for Intel.  I don't speak on their behalf.


[-- Attachment #2: diff --]
[-- Type: text/plain, Size: 1089 bytes --]

diff --git a/arch/x86/kernel/vmlinux.lds.S b/arch/x86/kernel/vmlinux.lds.S
index 59f31d2..2eb6874 100644
--- a/arch/x86/kernel/vmlinux.lds.S
+++ b/arch/x86/kernel/vmlinux.lds.S
@@ -407,12 +407,12 @@ INIT_PER_CPU(irq_stack_union);
 /*
  * Build-time check on the image size:
  */
-ASSERT((_end - _text <= KERNEL_IMAGE_SIZE),
-	"kernel image bigger than KERNEL_IMAGE_SIZE")
+. = ASSERT((_end - _text <= KERNEL_IMAGE_SIZE),
+	   "kernel image bigger than KERNEL_IMAGE_SIZE");
 
 #ifdef CONFIG_SMP
-ASSERT((per_cpu__irq_stack_union == 0),
-        "irq_stack_union is not at start of per-cpu area");
+. = ASSERT((per_cpu__irq_stack_union == 0),
+           "irq_stack_union is not at start of per-cpu area");
 #endif
 
 #endif /* CONFIG_X86_32 */
@@ -420,7 +420,6 @@ ASSERT((per_cpu__irq_stack_union == 0),
 #ifdef CONFIG_KEXEC
 #include <asm/kexec.h>
 
-ASSERT(kexec_control_code_size <= KEXEC_CONTROL_CODE_MAX_SIZE,
-       "kexec control code size is too big")
+. = ASSERT(kexec_control_code_size <= KEXEC_CONTROL_CODE_MAX_SIZE,
+           "kexec control code size is too big");
 #endif
-

Re: ld:arch/x86/kernel/vmlinux.lds:678: parse error

[Jean Delvare]

Hi Peter,

Thanks for your fast answer.

On Mon, 03 Aug 2009 08:31:37 -0700, H. Peter Anvin wrote:
> On 08/03/2009 05:04 AM, Jean Delvare wrote:
> > Hi Sam,
> > 
> > I am not able to build Linus' latest kernel on my oldest test system. I
> > get the following linking error at the end:
> > 
> >   LD      .tmp_vmlinux1
> > ld:arch/x86/kernel/vmlinux.lds:678: parse error
> > 
> > Which is:
> > 
> > ASSERT((_end - 0xC0000000 <= (512 * 1024 * 1024)),
> >         "kernel image bigger than KERNEL_IMAGE_SIZE")
> > 
> > Apparently this assertion is relatively new, as I was able to build
> > kernel 2.6.29 on the same system. The system has binutils 2.14.90.0.6,
> > which is more recent than what is listed in Documentation/Changes
> > (2.12).
> > 
> > Any idea?
> > 
> 
> Old binutils doesn't accept the naked ASSERT(); there has to be a sink
> for the data; so one has to do crap like:
> 
> __junk_symbol = ASSERT(...);
> 
> or
> 
> . = ASSERT(...);
> 
> near the end.
> 
> Does this patch fix it for you?

You patch doesn't, because you forgot to fix the one ASSERT() call that
failed for me ;) but the updated version below does indeed fix the
build for me:

Index: linux-2.6.31-rc5/arch/x86/kernel/vmlinux.lds.S
===================================================================
--- linux-2.6.31-rc5.orig/arch/x86/kernel/vmlinux.lds.S	2009-08-01 22:24:12.000000000 +0200
+++ linux-2.6.31-rc5/arch/x86/kernel/vmlinux.lds.S	2009-08-03 21:22:31.000000000 +0200
@@ -393,8 +393,8 @@
 
 
 #ifdef CONFIG_X86_32
-ASSERT((_end - LOAD_OFFSET <= KERNEL_IMAGE_SIZE),
-        "kernel image bigger than KERNEL_IMAGE_SIZE")
+. = ASSERT((_end - LOAD_OFFSET <= KERNEL_IMAGE_SIZE),
+	   "kernel image bigger than KERNEL_IMAGE_SIZE");
 #else
 /*
  * Per-cpu symbols which need to be offset from __per_cpu_load
@@ -407,12 +407,12 @@
 /*
  * Build-time check on the image size:
  */
-ASSERT((_end - _text <= KERNEL_IMAGE_SIZE),
-	"kernel image bigger than KERNEL_IMAGE_SIZE")
+. = ASSERT((_end - _text <= KERNEL_IMAGE_SIZE),
+	   "kernel image bigger than KERNEL_IMAGE_SIZE");
 
 #ifdef CONFIG_SMP
-ASSERT((per_cpu__irq_stack_union == 0),
-        "irq_stack_union is not at start of per-cpu area");
+. = ASSERT((per_cpu__irq_stack_union == 0),
+           "irq_stack_union is not at start of per-cpu area");
 #endif
 
 #endif /* CONFIG_X86_32 */
@@ -420,7 +420,7 @@
 #ifdef CONFIG_KEXEC
 #include <asm/kexec.h>
 
-ASSERT(kexec_control_code_size <= KEXEC_CONTROL_CODE_MAX_SIZE,
-       "kexec control code size is too big")
+. = ASSERT(kexec_control_code_size <= KEXEC_CONTROL_CODE_MAX_SIZE,
+           "kexec control code size is too big");
 #endif
 

So I guess you can send this to Linus. Thanks!

-- 
Jean Delvare