2.6.26-rc1 on x86: ld: warning: dot moved backwards before `.text'

2.6.26-rc1 on x86: ld: warning: dot moved backwards before `.text'

[Mikael Pettersson]

Building kernel 2.6.26-rc1 on FC6/i386 causes:

  ld -m elf_i386   -T arch/x86/kernel/acpi/realmode/wakeup.lds arch/x86/kernel/acpi/realmode/wakeup.o arch/x86/kernel/acpi/realmode/wakemain.o arch/x86/kernel/acpi/realmode/video-mode.o arch/x86/kernel/acpi/realmode/copy.o arch/x86/kernel/acpi/realmode/video-vga.o arch/x86/kernel/acpi/realmode/video-vesa.o arch/x86/kernel/acpi/realmode/video-bios.o -o arch/x86/kernel/acpi/realmode/wakeup.elf 
ld: warning: dot moved backwards before `.text'
ld: warning: dot moved backwards before `.text'
ld: warning: dot moved backwards before `.text'

This is with gcc-4.2.3 and binutils-2.17.50.0.6-5.fc6 20061020.

The resulting kernel hasn't broken on me yet, however.

A search through the LKML archives showed that this was reported
for 2.6.25-rc3-mm1, but I couldn't find any discussion about it
after that:

<http://marc.info/?l=linux-kernel&m=120470303728010&w=2>

The .config causing this warning is available in
<http://user.it.uu.se/~mikpe/linux/ale/config-2.6.26-rc1>

/Mikael

Re: 2.6.26-rc1 on x86: ld: warning: dot moved backwards before `.text'

[Rafael J. Wysocki]

On Saturday, 10 of May 2008, Mikael Pettersson wrote:
> Building kernel 2.6.26-rc1 on FC6/i386 causes:
> 
>   ld -m elf_i386   -T arch/x86/kernel/acpi/realmode/wakeup.lds arch/x86/kernel/acpi/realmode/wakeup.o arch/x86/kernel/acpi/realmode/wakemain.o arch/x86/kernel/acpi/realmode/video-mode.o arch/x86/kernel/acpi/realmode/copy.o arch/x86/kernel/acpi/realmode/video-vga.o arch/x86/kernel/acpi/realmode/video-vesa.o arch/x86/kernel/acpi/realmode/video-bios.o -o arch/x86/kernel/acpi/realmode/wakeup.elf 
> ld: warning: dot moved backwards before `.text'
> ld: warning: dot moved backwards before `.text'
> ld: warning: dot moved backwards before `.text'
> 
> This is with gcc-4.2.3 and binutils-2.17.50.0.6-5.fc6 20061020.
> 
> The resulting kernel hasn't broken on me yet, however.

It shouldn't.  Have you tried to suspend to RAM?

> A search through the LKML archives showed that this was reported
> for 2.6.25-rc3-mm1, but I couldn't find any discussion about it
> after that:
> 
> <http://marc.info/?l=linux-kernel&m=120470303728010&w=2>
> 
> The .config causing this warning is available in
> <http://user.it.uu.se/~mikpe/linux/ale/config-2.6.26-rc1>

Well, we don't really know why this happens.  It surely doesn't happen with
newer toolchains (as of OpenSUSE 10.3 for example).

Thanks,
Rafael

Re: 2.6.26-rc1 on x86: ld: warning: dot moved backwards before `.text'

[H. Peter Anvin]

Mikael Pettersson wrote:
> 
> The resulting kernel hasn't broken on me yet, however.
> 
> A search through the LKML archives showed that this was reported
> for 2.6.25-rc3-mm1, but I couldn't find any discussion about it
> after that:
> 
> <http://marc.info/?l=linux-kernel&m=120470303728010&w=2>
> 
> The .config causing this warning is available in
> <http://user.it.uu.se/~mikpe/linux/ale/config-2.6.26-rc1>
> 

This is because the organization of the file was changed, and the linker 
script wasn't changed to match:

SECTIONS
{
         . = HEADER_OFFSET;
         .header : {
                  *(.header)
         }

         . = 0;
         .text : {
                  *(.text*)
         }

This is crap; the sections should be listed *in order* so the linker can 
warn properly when something bad happens.

The code should be correct; reorganizing the linker script correctly 
should fix the problem.

	-hpa

Re: 2.6.26-rc1 on x86: ld: warning: dot moved backwards before `.text'

[Rafael J. Wysocki]

On Saturday, 10 of May 2008, H. Peter Anvin wrote:
> Mikael Pettersson wrote:
> > 
> > The resulting kernel hasn't broken on me yet, however.
> > 
> > A search through the LKML archives showed that this was reported
> > for 2.6.25-rc3-mm1, but I couldn't find any discussion about it
> > after that:
> > 
> > <http://marc.info/?l=linux-kernel&m=120470303728010&w=2>
> > 
> > The .config causing this warning is available in
> > <http://user.it.uu.se/~mikpe/linux/ale/config-2.6.26-rc1>
> > 
> 
> This is because the organization of the file was changed, and the linker 
> script wasn't changed to match:
> 
> SECTIONS
> {
>          . = HEADER_OFFSET;
>          .header : {
>                   *(.header)
>          }
> 
>          . = 0;
>          .text : {
>                   *(.text*)
>          }
> 
> This is crap; the sections should be listed *in order* so the linker can 
> warn properly when something bad happens.
> 
> The code should be correct; reorganizing the linker script correctly 
> should fix the problem.

Hm.  The above is from Pavel and Sam.

Thanks,
Rafael

Re: 2.6.26-rc1 on x86: ld: warning: dot moved backwards before `.text'

[Pavel Machek]

Hi!

> > > The resulting kernel hasn't broken on me yet, however.
> > > 
> > > A search through the LKML archives showed that this was reported
> > > for 2.6.25-rc3-mm1, but I couldn't find any discussion about it
> > > after that:
> > > 
> > > <http://marc.info/?l=linux-kernel&m=120470303728010&w=2>
> > > 
> > > The .config causing this warning is available in
> > > <http://user.it.uu.se/~mikpe/linux/ale/config-2.6.26-rc1>
> > > 
> > 
> > This is because the organization of the file was changed, and the linker 
> > script wasn't changed to match:
> > 
> > SECTIONS
> > {
> >          . = HEADER_OFFSET;
> >          .header : {
> >                   *(.header)
> >          }
> > 
> >          . = 0;
> >          .text : {
> >                   *(.text*)
> >          }
> > 
> > This is crap; the sections should be listed *in order* so the linker can 
> > warn properly when something bad happens.
> > 
> > The code should be correct; reorganizing the linker script correctly 
> > should fix the problem.

I'm not good at linker magic. I needed to put code from offset zero,
while still having header at known address...

> Hm.  The above is from Pavel and Sam.

Sorry about that. Plus, I'm on holidays just now...

									Pavel

-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: 2.6.26-rc1 on x86: ld: warning: dot moved backwards before `.text'

[Sam Ravnborg]

On Sat, May 10, 2008 at 01:25:53PM -0700, H. Peter Anvin wrote:
> Mikael Pettersson wrote:
> >
> >The resulting kernel hasn't broken on me yet, however.
> >
> >A search through the LKML archives showed that this was reported
> >for 2.6.25-rc3-mm1, but I couldn't find any discussion about it
> >after that:
> >
> ><http://marc.info/?l=linux-kernel&m=120470303728010&w=2>
> >
> >The .config causing this warning is available in
> ><http://user.it.uu.se/~mikpe/linux/ale/config-2.6.26-rc1>
> >
> 
> This is because the organization of the file was changed, and the linker 
> script wasn't changed to match:
> 
> SECTIONS
> {
>         . = HEADER_OFFSET;
>         .header : {
>                  *(.header)
>         }
> 
>         . = 0;
>         .text : {
>                  *(.text*)
>         }
> 
> This is crap; the sections should be listed *in order* so the linker can 
> warn properly when something bad happens.
> 
> The code should be correct; reorganizing the linker script correctly 
> should fix the problem.

I need Pavel to comment on this.
Why we start with . equal 0x3f00 and then later as . equals 0x0000
I dunno.
I did not look into this part of the code back then - but I should anyway
have seen this flaw in the linker script :-(

	Sam

Re: 2.6.26-rc1 on x86: ld: warning: dot moved backwards before `.text'

[Cyrill Gorcunov]

[Sam Ravnborg - Sat, May 10, 2008 at 11:16:02PM +0200]
| On Sat, May 10, 2008 at 01:25:53PM -0700, H. Peter Anvin wrote:
| > Mikael Pettersson wrote:
| > >
| > >The resulting kernel hasn't broken on me yet, however.
| > >
| > >A search through the LKML archives showed that this was reported
| > >for 2.6.25-rc3-mm1, but I couldn't find any discussion about it
| > >after that:
| > >
| > ><http://marc.info/?l=linux-kernel&m=120470303728010&w=2>
| > >
| > >The .config causing this warning is available in
| > ><http://user.it.uu.se/~mikpe/linux/ale/config-2.6.26-rc1>
| > >
| > 
| > This is because the organization of the file was changed, and the linker 
| > script wasn't changed to match:
| > 
| > SECTIONS
| > {
| >         . = HEADER_OFFSET;
| >         .header : {
| >                  *(.header)
| >         }
| > 
| >         . = 0;
| >         .text : {
| >                  *(.text*)
| >         }
| > 
| > This is crap; the sections should be listed *in order* so the linker can 
| > warn properly when something bad happens.
| > 
| > The code should be correct; reorganizing the linker script correctly 
| > should fix the problem.
| 
| I need Pavel to comment on this.
| Why we start with . equal 0x3f00 and then later as . equals 0x0000
| I dunno.
| I did not look into this part of the code back then - but I should anyway
| have seen this flaw in the linker script :-(
| 
| 	Sam

Should not it be something like
---

diff --git a/arch/x86/kernel/acpi/realmode/wakeup.lds.S b/arch/x86/kernel/acpi/realmode/wakeup.lds.S
index 22fab6c..cccf62d 100644
--- a/arch/x86/kernel/acpi/realmode/wakeup.lds.S
+++ b/arch/x86/kernel/acpi/realmode/wakeup.lds.S
@@ -12,11 +12,6 @@ ENTRY(_start)
 
 SECTIONS
 {
-	. = HEADER_OFFSET;
-	.header : {
-		 *(.header)
-	}
-
 	. = 0;
 	.text : {
 		 *(.text*)
@@ -53,6 +48,11 @@ SECTIONS
 	. = ALIGN(16);
 	_end = .;
 
+	. = HEADER_OFFSET;
+	.header : {
+		 *(.header)
+	}
+
 	/DISCARD/ : {
 		*(.note*)
 	}
---
		- Cyrill -

Re: 2.6.26-rc1 on x86: ld: warning: dot moved backwards before `.text'

[H. Peter Anvin]

Cyrill Gorcunov wrote:
> 
> Should not it be something like
> ---
> 
> diff --git a/arch/x86/kernel/acpi/realmode/wakeup.lds.S b/arch/x86/kernel/acpi/realmode/wakeup.lds.S
> index 22fab6c..cccf62d 100644
> --- a/arch/x86/kernel/acpi/realmode/wakeup.lds.S
> +++ b/arch/x86/kernel/acpi/realmode/wakeup.lds.S
> @@ -12,11 +12,6 @@ ENTRY(_start)
>  
>  SECTIONS
>  {
> -	. = HEADER_OFFSET;
> -	.header : {
> -		 *(.header)
> -	}
> -
>  	. = 0;
>  	.text : {
>  		 *(.text*)
> @@ -53,6 +48,11 @@ SECTIONS
>  	. = ALIGN(16);
>  	_end = .;
>  
> +	. = HEADER_OFFSET;
> +	.header : {
> +		 *(.header)
> +	}
> +
>  	/DISCARD/ : {
>  		*(.note*)
>  	}
> ---
> 		- Cyrill -

Yes.  This was the way I originally had it, then Pavel decided to move 
the header to the end.

	-hpa

Re: 2.6.26-rc1 on x86: ld: warning: dot moved backwards before `.text'

[Cyrill Gorcunov]

[H. Peter Anvin - Sun, May 11, 2008 at 11:15:04AM -0700]
> Cyrill Gorcunov wrote:
>> Should not it be something like
>> ---
>> diff --git a/arch/x86/kernel/acpi/realmode/wakeup.lds.S 
>> b/arch/x86/kernel/acpi/realmode/wakeup.lds.S
>> index 22fab6c..cccf62d 100644
>> --- a/arch/x86/kernel/acpi/realmode/wakeup.lds.S
>> +++ b/arch/x86/kernel/acpi/realmode/wakeup.lds.S
>> @@ -12,11 +12,6 @@ ENTRY(_start)
>>   SECTIONS
>>  {
>> -	. = HEADER_OFFSET;
>> -	.header : {
>> -		 *(.header)
>> -	}
>> -
>>  	. = 0;
>>  	.text : {
>>  		 *(.text*)
>> @@ -53,6 +48,11 @@ SECTIONS
>>  	. = ALIGN(16);
>>  	_end = .;
>>  +	. = HEADER_OFFSET;
>> +	.header : {
>> +		 *(.header)
>> +	}
>> +
>>  	/DISCARD/ : {
>>  		*(.note*)
>>  	}
>> ---
>> 		- Cyrill -
>
> Yes.  This was the way I originally had it, then Pavel decided to move the 
> header to the end.
>
> 	-hpa
>

Hi Peter,

actually I only doubt about exactly position of this headers - I mean
meguess it better should be _before_ "_end" not after. Since the others
script uses (acpi/wakeup_rm.S):

	wakeup_code_start:
	.incbin	"arch/x86/kernel/acpi/realmode/wakeup.bin"
	wakeup_code_end:

and kernel uses these wakeup_code_* marks to save/restore memory I think
it should be like

	. = HEADER_OFFSET;
	.header : {
		...
	}
	. = ALIGN(16);
	_end = .;
	/DISCARD/ ...

		- Cyrill -

Re: 2.6.26-rc1 on x86: ld: warning: dot moved backwards before `.text'

[H. Peter Anvin]

Cyrill Gorcunov wrote:
> 
> Hi Peter,
> 
> actually I only doubt about exactly position of this headers - I mean
> meguess it better should be _before_ "_end" not after. Since the others
> script uses (acpi/wakeup_rm.S):
> 
> 	wakeup_code_start:
> 	.incbin	"arch/x86/kernel/acpi/realmode/wakeup.bin"
> 	wakeup_code_end:
> 
> and kernel uses these wakeup_code_* marks to save/restore memory I think
> it should be like
> 
> 	. = HEADER_OFFSET;
> 	.header : {
> 		...
> 	}
> 	. = ALIGN(16);
> 	_end = .;
> 	/DISCARD/ ...
> 

Yes, I believe that's correct.

	-hpa

Re: 2.6.26-rc1 on x86: ld: warning: dot moved backwards before `.text'

[Cyrill Gorcunov]

[H. Peter Anvin - Sun, May 11, 2008 at 11:25:54AM -0700]
> Cyrill Gorcunov wrote:
>> Hi Peter,
>> actually I only doubt about exactly position of this headers - I mean
>> meguess it better should be _before_ "_end" not after. Since the others
>> script uses (acpi/wakeup_rm.S):
>> 	wakeup_code_start:
>> 	.incbin	"arch/x86/kernel/acpi/realmode/wakeup.bin"
>> 	wakeup_code_end:
>> and kernel uses these wakeup_code_* marks to save/restore memory I think
>> it should be like
>> 	. = HEADER_OFFSET;
>> 	.header : {
>> 		...
>> 	}
>> 	. = ALIGN(16);
>> 	_end = .;
>> 	/DISCARD/ ...
>
> Yes, I believe that's correct.
>
> 	-hpa
>

Thanks for review, Peter, I'll cook a patch and wait for
Pavel's opinion ;-)

		- Cyrill -

Re: 2.6.26-rc1 on x86: ld: warning: dot moved backwards before `.text'

[Rafael J. Wysocki]

On Sunday, 11 of May 2008, Cyrill Gorcunov wrote:
> [H. Peter Anvin - Sun, May 11, 2008 at 11:25:54AM -0700]
> > Cyrill Gorcunov wrote:
> >> Hi Peter,
> >> actually I only doubt about exactly position of this headers - I mean
> >> meguess it better should be _before_ "_end" not after. Since the others
> >> script uses (acpi/wakeup_rm.S):
> >> 	wakeup_code_start:
> >> 	.incbin	"arch/x86/kernel/acpi/realmode/wakeup.bin"
> >> 	wakeup_code_end:
> >> and kernel uses these wakeup_code_* marks to save/restore memory I think
> >> it should be like
> >> 	. = HEADER_OFFSET;
> >> 	.header : {
> >> 		...
> >> 	}
> >> 	. = ALIGN(16);
> >> 	_end = .;
> >> 	/DISCARD/ ...
> >
> > Yes, I believe that's correct.
> >
> > 	-hpa
> >
> 
> Thanks for review, Peter, I'll cook a patch and wait for
> Pavel's opinion ;-)

I've tested the appended one.  If that's what you mean, we'll ask Ingo to take
it.

Thanks,
Rafael


---
 arch/x86/kernel/acpi/realmode/wakeup.lds.S |   10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

Index: linux-2.6/arch/x86/kernel/acpi/realmode/wakeup.lds.S
===================================================================
--- linux-2.6.orig/arch/x86/kernel/acpi/realmode/wakeup.lds.S
+++ linux-2.6/arch/x86/kernel/acpi/realmode/wakeup.lds.S
@@ -12,11 +12,6 @@ ENTRY(_start)
 
 SECTIONS
 {
-	. = HEADER_OFFSET;
-	.header : {
-		 *(.header)
-	}
-
 	. = 0;
 	.text : {
 		 *(.text*)
@@ -50,6 +45,11 @@ SECTIONS
 		__bss_end = .;
 	}
 
+	. = HEADER_OFFSET;
+	.header : {
+		 *(.header)
+	}
+
 	. = ALIGN(16);
 	_end = .;
 

Re: 2.6.26-rc1 on x86: ld: warning: dot moved backwards before `.text'

[Cyrill Gorcunov]

On Mon, May 12, 2008 at 1:22 AM, Rafael J. Wysocki <rjw@sisk.pl> wrote:
> On Sunday, 11 of May 2008, Cyrill Gorcunov wrote:
>> [H. Peter Anvin - Sun, May 11, 2008 at 11:25:54AM -0700]
>> > Cyrill Gorcunov wrote:
>> >> Hi Peter,
>> >> actually I only doubt about exactly position of this headers - I mean
>> >> meguess it better should be _before_ "_end" not after. Since the others
>> >> script uses (acpi/wakeup_rm.S):
>> >>    wakeup_code_start:
>> >>    .incbin "arch/x86/kernel/acpi/realmode/wakeup.bin"
>> >>    wakeup_code_end:
>> >> and kernel uses these wakeup_code_* marks to save/restore memory I think
>> >> it should be like
>> >>    . = HEADER_OFFSET;
>> >>    .header : {
>> >>            ...
>> >>    }
>> >>    . = ALIGN(16);
>> >>    _end = .;
>> >>    /DISCARD/ ...
>> >
>> > Yes, I believe that's correct.
>> >
>> >     -hpa
>> >
>>
>> Thanks for review, Peter, I'll cook a patch and wait for
>> Pavel's opinion ;-)
>
> I've tested the appended one.  If that's what you mean, we'll ask Ingo to take
> it.
>
> Thanks,
> Rafael
>
>
> ---
>  arch/x86/kernel/acpi/realmode/wakeup.lds.S |   10 +++++-----
>  1 file changed, 5 insertions(+), 5 deletions(-)
>
> Index: linux-2.6/arch/x86/kernel/acpi/realmode/wakeup.lds.S
> ===================================================================
> --- linux-2.6.orig/arch/x86/kernel/acpi/realmode/wakeup.lds.S
> +++ linux-2.6/arch/x86/kernel/acpi/realmode/wakeup.lds.S
> @@ -12,11 +12,6 @@ ENTRY(_start)
>
>  SECTIONS
>  {
> -       . = HEADER_OFFSET;
> -       .header : {
> -                *(.header)
> -       }
> -
>        . = 0;
>        .text : {
>                 *(.text*)
> @@ -50,6 +45,11 @@ SECTIONS
>                __bss_end = .;
>        }
>
> +       . = HEADER_OFFSET;
> +       .header : {
> +                *(.header)
> +       }
> +
>        . = ALIGN(16);
>        _end = .;
>
>

Thanks Rafael,

it's exactly what I meant - http://lkml.org/lkml/2008/5/11/120 ;)

Re: 2.6.26-rc1 on x86: ld: warning: dot moved backwards before `.text'

[Cyrill Gorcunov]

On Mon, May 12, 2008 at 8:15 AM, Cyrill Gorcunov <gorcunov@gmail.com> wrote:
> On Mon, May 12, 2008 at 1:22 AM, Rafael J. Wysocki <rjw@sisk.pl> wrote:
>> On Sunday, 11 of May 2008, Cyrill Gorcunov wrote:
>>> [H. Peter Anvin - Sun, May 11, 2008 at 11:25:54AM -0700]
>>> > Cyrill Gorcunov wrote:
>>> >> Hi Peter,
>>> >> actually I only doubt about exactly position of this headers - I mean
>>> >> meguess it better should be _before_ "_end" not after. Since the others
>>> >> script uses (acpi/wakeup_rm.S):
>>> >>    wakeup_code_start:
>>> >>    .incbin "arch/x86/kernel/acpi/realmode/wakeup.bin"
>>> >>    wakeup_code_end:
>>> >> and kernel uses these wakeup_code_* marks to save/restore memory I think
>>> >> it should be like
>>> >>    . = HEADER_OFFSET;
>>> >>    .header : {
>>> >>            ...
>>> >>    }
>>> >>    . = ALIGN(16);
>>> >>    _end = .;
>>> >>    /DISCARD/ ...
>>> >
>>> > Yes, I believe that's correct.
>>> >
>>> >     -hpa
>>> >
>>>
>>> Thanks for review, Peter, I'll cook a patch and wait for
>>> Pavel's opinion ;-)
>>
>> I've tested the appended one.  If that's what you mean, we'll ask Ingo to take
>> it.
>>
>> Thanks,
>> Rafael
>>
>>
>> ---
>>  arch/x86/kernel/acpi/realmode/wakeup.lds.S |   10 +++++-----
>>  1 file changed, 5 insertions(+), 5 deletions(-)
>>
>> Index: linux-2.6/arch/x86/kernel/acpi/realmode/wakeup.lds.S
>> ===================================================================
>> --- linux-2.6.orig/arch/x86/kernel/acpi/realmode/wakeup.lds.S
>> +++ linux-2.6/arch/x86/kernel/acpi/realmode/wakeup.lds.S
>> @@ -12,11 +12,6 @@ ENTRY(_start)
>>
>>  SECTIONS
>>  {
>> -       . = HEADER_OFFSET;
>> -       .header : {
>> -                *(.header)
>> -       }
>> -
>>        . = 0;
>>        .text : {
>>                 *(.text*)
>> @@ -50,6 +45,11 @@ SECTIONS
>>                __bss_end = .;
>>        }
>>
>> +       . = HEADER_OFFSET;
>> +       .header : {
>> +                *(.header)
>> +       }
>> +
>>        . = ALIGN(16);
>>        _end = .;
>>
>>
>
> Thanks Rafael,
>
> it's exactly what I meant - http://lkml.org/lkml/2008/5/11/120 ;)
>

Heh, now we have even two patches (exactly the same in contents). So, Ingo,
could you pick up any patch, please? ;-) (i'm in office now without access to
my linux machine,  so I can't send the patch with my and Rafael's sign-offs)

Re: 2.6.26-rc1 on x86: ld: warning: dot moved backwards before `.text'

[Pavel Machek]

On Sun 2008-05-11 23:22:52, Rafael J. Wysocki wrote:
> On Sunday, 11 of May 2008, Cyrill Gorcunov wrote:
> > [H. Peter Anvin - Sun, May 11, 2008 at 11:25:54AM -0700]
> > > Cyrill Gorcunov wrote:
> > >> Hi Peter,
> > >> actually I only doubt about exactly position of this headers - I mean
> > >> meguess it better should be _before_ "_end" not after. Since the others
> > >> script uses (acpi/wakeup_rm.S):
> > >> 	wakeup_code_start:
> > >> 	.incbin	"arch/x86/kernel/acpi/realmode/wakeup.bin"
> > >> 	wakeup_code_end:
> > >> and kernel uses these wakeup_code_* marks to save/restore memory I think
> > >> it should be like
> > >> 	. = HEADER_OFFSET;
> > >> 	.header : {
> > >> 		...
> > >> 	}
> > >> 	. = ALIGN(16);
> > >> 	_end = .;
> > >> 	/DISCARD/ ...
> > >
> > > Yes, I believe that's correct.
> > >
> > > 	-hpa
> > >
> > 
> > Thanks for review, Peter, I'll cook a patch and wait for
> > Pavel's opinion ;-)
> 
> I've tested the appended one.  If that's what you mean, we'll ask Ingo to take
> it.

ACK... and thanks!
								Pavel

-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: 2.6.26-rc1 on x86: ld: warning: dot moved backwards before `.text'

[Pavel Machek]

On Sun 2008-05-11 19:44:11, Cyrill Gorcunov wrote:
> [Sam Ravnborg - Sat, May 10, 2008 at 11:16:02PM +0200]
> | On Sat, May 10, 2008 at 01:25:53PM -0700, H. Peter Anvin wrote:
> | > Mikael Pettersson wrote:
> | > >
> | > >The resulting kernel hasn't broken on me yet, however.
> | > >
> | > >A search through the LKML archives showed that this was reported
> | > >for 2.6.25-rc3-mm1, but I couldn't find any discussion about it
> | > >after that:
> | > >
> | > ><http://marc.info/?l=linux-kernel&m=120470303728010&w=2>
> | > >
> | > >The .config causing this warning is available in
> | > ><http://user.it.uu.se/~mikpe/linux/ale/config-2.6.26-rc1>
> | > >
> | > 
> | > This is because the organization of the file was changed, and the linker 
> | > script wasn't changed to match:
> | > 
> | > SECTIONS
> | > {
> | >         . = HEADER_OFFSET;
> | >         .header : {
> | >                  *(.header)
> | >         }
> | > 
> | >         . = 0;
> | >         .text : {
> | >                  *(.text*)
> | >         }
> | > 
> | > This is crap; the sections should be listed *in order* so the linker can 
> | > warn properly when something bad happens.
> | > 
> | > The code should be correct; reorganizing the linker script correctly 
> | > should fix the problem.
> | 
> | I need Pavel to comment on this.
> | Why we start with . equal 0x3f00 and then later as . equals 0x0000
> | I dunno.
> | I did not look into this part of the code back then - but I should anyway
> | have seen this flaw in the linker script :-(
> | 
> | 	Sam
> 
> Should not it be something like
> ---
> 
> diff --git a/arch/x86/kernel/acpi/realmode/wakeup.lds.S b/arch/x86/kernel/acpi/realmode/wakeup.lds.S
> index 22fab6c..cccf62d 100644
> --- a/arch/x86/kernel/acpi/realmode/wakeup.lds.S
> +++ b/arch/x86/kernel/acpi/realmode/wakeup.lds.S
> @@ -12,11 +12,6 @@ ENTRY(_start)
>  
>  SECTIONS
>  {
> -	. = HEADER_OFFSET;
> -	.header : {
> -		 *(.header)
> -	}
> -
>  	. = 0;
>  	.text : {
>  		 *(.text*)
> @@ -53,6 +48,11 @@ SECTIONS
>  	. = ALIGN(16);
>  	_end = .;
>  
> +	. = HEADER_OFFSET;
> +	.header : {
> +		 *(.header)
> +	}
> +
>  	/DISCARD/ : {
>  		*(.note*)
>  	}

If it passes s2ram/resume test, it is probably correct ;-).
									Pavel

-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html