arch/cris/arch-v10/vmlinux.lds.S fix boot problem

PATCH In review Yuusei KUWANA 3 participants 6 messages Dec 15, 2007 – Jan 3, 2008

Share & tools

Yuusei KUWANA

author

2007-12-15 06:15 UTC #0 lore raw

PATCH 1/1 arch/cris/arch-v10/vmlinux.lds.S fix boot problem

arch/cris/arch-v10/vmlinux.lds.S
fix boot problem
* too old initcall style. replace INITCALLS macro
* __init_begin, __init_end move for free_initmem()

Note: with this patch kernel boot and mount root, but after init done, kernel panic at do_signal() ...

ryu

Signed-off-by Yuusei KUWANA <[email protected]>

---
 arch/cris/arch-v10/vmlinux.lds.S |   13 ++++---------
 1 files changed, 4 insertions(+), 9 deletions(-)

arch/cris/arch-v10/vmlinux.lds.S +4 -9

index 9859d49..c5bab0e 100644
@@ -53,26 +53,22 @@ SECTIONS
   	.data.init_task : { *(.data.init_task) }
 
   	. = ALIGN(8192);              /* Init code and data */
-  	__init_begin = .;
 	.init.text : { 
 		   _sinittext = .;
 		   *(.init.text)
 		   _einittext = .;
 	}
+  	__init_begin = .;
   	.init.data : { *(.init.data) }
+  	__init_end = .;
+
   	. = ALIGN(16);
   	__setup_start = .;
   	.init.setup : { *(.init.setup) }
   	__setup_end = .;
   	.initcall.init : {
 		__initcall_start = .;
-		*(.initcall1.init);
-		*(.initcall2.init);
-		*(.initcall3.init);
-		*(.initcall4.init);
-		*(.initcall5.init);
-		*(.initcall6.init);
-		*(.initcall7.init);
+			INITCALLS
 		__initcall_end = .;	
 	}

@@ -97,7 +93,6 @@ SECTIONS
 #endif
 	
 	__vmlinux_end = .;            /* last address of the physical file */
-  	__init_end = .;
 
 	__data_end = . ;              /* Move to _edata ? */
 	__bss_start = .;              /* BSS */

Signature · 1 line

1.5.3

Jesper Nilsson

2007-12-21 15:14 UTC #1 lore raw

Quoted 4 lines · Yuusei KUWANA

arch/cris/arch-v10/vmlinux.lds.S
fix boot problem
* too old initcall style. replace INITCALLS macro
* __init_begin, __init_end move for free_initmem()

Hi,

The conversion to INITCALLS is ok (I have the same change here), but I'm curious what problem you are solving with moving __init_begin and __init_end to only include the .init.data.

If there is use of initcalls/data outside init then that is a bug and should be fixed, not worked around.

There might however be another problem that Mikael Starvik told me about, the padding between initdata and the next section had gone AWOL, which made the free_initmem() destroy useful data.

With the attached patch (and a lot of other patches) I am successful in booting an Axis Devboard82+.
Could you please try the included patch and see if that solves your problem?

Quoted 2 lines

Note: with this patch kernel boot and mount root, but after init done, kernel panic at do_signal() ...

I have a patch in the pipe that should fix this, will send it in just a small while.

Quoted 1 line

ryu

Thank you for your help!

Best regards,

/Jesper

Quoted 50 lines

Signed-off-by Yuusei KUWANA <[email protected]>

---
 arch/cris/arch-v10/vmlinux.lds.S |   13 ++++---------
 1 files changed, 4 insertions(+), 9 deletions(-)

arch/cris/arch-v10/vmlinux.lds.S +4 -9

index 9859d49..c5bab0e 100644
@@ -53,26 +53,22 @@ SECTIONS
   	.data.init_task : { *(.data.init_task) }
 
   	. = ALIGN(8192);              /* Init code and data */
-  	__init_begin = .;
 	.init.text : { 
 		   _sinittext = .;
 		   *(.init.text)
 		   _einittext = .;
 	}
+  	__init_begin = .;
   	.init.data : { *(.init.data) }
+  	__init_end = .;
+
   	. = ALIGN(16);
   	__setup_start = .;
   	.init.setup : { *(.init.setup) }
   	__setup_end = .;
   	.initcall.init : {
 		__initcall_start = .;
-		*(.initcall1.init);
-		*(.initcall2.init);
-		*(.initcall3.init);
-		*(.initcall4.init);
-		*(.initcall5.init);
-		*(.initcall6.init);
-		*(.initcall7.init);
+			INITCALLS
 		__initcall_end = .;	
 	}

@@ -97,7 +93,6 @@ SECTIONS
 #endif
 
 	__vmlinux_end = .;            /* last address of the physical file */
-  	__init_end = .;
 
 	__data_end = . ;              /* Move to _edata ? */
 	__bss_start = .;              /* BSS */

Signature · 1 line

1.5.3

arch/cris/arch-v10/vmlinux.lds.S +11 ~~-13~~

index 9859d49..30fa872 100644
@@ -23,7 +23,7 @@ SECTIONS
 	_stext = .;
 	__stext = .;
 	.text : {
-		*(.text)
+		TEXT_TEXT
 		SCHED_TEXT
 		LOCK_TEXT
 		*(.fixup)

@@ -66,13 +66,7 @@ SECTIONS
   	__setup_end = .;
   	.initcall.init : {
 		__initcall_start = .;
-		*(.initcall1.init);
-		*(.initcall2.init);
-		*(.initcall3.init);
-		*(.initcall4.init);
-		*(.initcall5.init);
-		*(.initcall6.init);
-		*(.initcall7.init);
+		INIT_CALLS
 		__initcall_end = .;	
 	}

@@ -88,11 +82,15 @@ SECTIONS
 		__initramfs_start = .;
 		*(.init.ramfs)
 		__initramfs_end = .;
-		/* We fill to the next page, so we can discard all init
-		   pages without needing to consider what payload might be
-		   appended to the kernel image.  */
-		FILL (0); 
-		. = ALIGN (8192);
+		/*
+		 * We fill to the next page, so we can discard all init
+		 * pages without needing to consider what payload might be
+		 * appended to the kernel image.
+		 */
+		.init.fill : {
+			FILL (0);
+			. = ALIGN (8192);
+		}
 	}
 #endif

/^JN - Jesper Nilsson
--
Jesper Nilsson -- [email protected]

Jesper Nilsson

2007-12-21 15:55 UTC #2 lore raw

...and of course I got a spurious "_" pasted into INITCALLS. Argh. Corrected patch below.

arch/cris/arch-v10/vmlinux.lds.S +12 ~~-13~~

index 9859d49..ac98b2a 100644
@@ -23,7 +23,7 @@ SECTIONS
 	_stext = .;
 	__stext = .;
 	.text : {
-		*(.text)
+		TEXT_TEXT
 		SCHED_TEXT
 		LOCK_TEXT
 		*(.fixup)

@@ -66,13 +66,7 @@ SECTIONS
   	__setup_end = .;
   	.initcall.init : {
 		__initcall_start = .;
-		*(.initcall1.init);
-		*(.initcall2.init);
-		*(.initcall3.init);
-		*(.initcall4.init);
-		*(.initcall5.init);
-		*(.initcall6.init);
-		*(.initcall7.init);
+		INITCALLS
 		__initcall_end = .;	
 	}

@@ -88,11 +82,16 @@ SECTIONS
 		__initramfs_start = .;
 		*(.init.ramfs)
 		__initramfs_end = .;
-		/* We fill to the next page, so we can discard all init
-		   pages without needing to consider what payload might be
-		   appended to the kernel image.  */
-		FILL (0); 
-		. = ALIGN (8192);
+
+		/*
+		 * We fill to the next page, so we can discard all init
+		 * pages without needing to consider what payload might be
+		 * appended to the kernel image.
+		 */
+		.init.fill : {
+			FILL (0);
+			. = ALIGN (8192);
+		}
 	}
 #endif

Quoted 30 lines · Jesper Nilsson

Quoted 4 lines · Yuusei KUWANA
arch/cris/arch-v10/vmlinux.lds.S
fix boot problem
* too old initcall style. replace INITCALLS macro
* __init_begin, __init_end move for free_initmem()
Hi,
The conversion to INITCALLS is ok (I have the same change here), but I'm curious what problem you are solving with moving __init_begin and __init_end to only include the .init.data.
If there is use of initcalls/data outside init then that is a bug and should be fixed, not worked around.
There might however be another problem that Mikael Starvik told me about, the padding between initdata and the next section had gone AWOL, which made the free_initmem() destroy useful data.
With the attached patch (and a lot of other patches) I am successful in booting an Axis Devboard82+.
Could you please try the included patch and see if that solves your problem?
Quoted 2 lines
Note: with this patch kernel boot and mount root, but after init done, kernel panic at do_signal() ...
I have a patch in the pipe that should fix this, will send it in just a small while.

Quoted 7 lines

Quoted 1 line
ryu
Thank you for your help!
Best regards,
/Jesper

/^JN - Jesper Nilsson
--
Jesper Nilsson -- [email protected]

Jesper Nilsson

2007-12-21 16:07 UTC #3 lore raw

Argh. Not my day, hopefully this is the last revision of the patch.

Best regards,

/Jesper

arch/cris/arch-v10/vmlinux.lds.S +14 ~~-15~~

index 9859d49..2afe1bd 100644
@@ -23,7 +23,7 @@ SECTIONS
 	_stext = .;
 	__stext = .;
 	.text : {
-		*(.text)
+		TEXT_TEXT
 		SCHED_TEXT
 		LOCK_TEXT
 		*(.fixup)

@@ -66,13 +66,7 @@ SECTIONS
   	__setup_end = .;
   	.initcall.init : {
 		__initcall_start = .;
-		*(.initcall1.init);
-		*(.initcall2.init);
-		*(.initcall3.init);
-		*(.initcall4.init);
-		*(.initcall5.init);
-		*(.initcall6.init);
-		*(.initcall7.init);
+		INITCALLS
 		__initcall_end = .;	
 	}

@@ -88,16 +82,21 @@ SECTIONS
 		__initramfs_start = .;
 		*(.init.ramfs)
 		__initramfs_end = .;
-		/* We fill to the next page, so we can discard all init
-		   pages without needing to consider what payload might be
-		   appended to the kernel image.  */
-		FILL (0); 
-		. = ALIGN (8192);
 	}
 #endif
-	
 	__vmlinux_end = .;            /* last address of the physical file */
-  	__init_end = .;
+
+	/*
+	 * We fill to the next page, so we can discard all init
+	 * pages without needing to consider what payload might be
+	 * appended to the kernel image.
+	 */
+	.init.fill : {
+		FILL (0);
+		. = ALIGN (8192);
+	}
+
+	__init_end = .;
 
 	__data_end = . ;              /* Move to _edata ? */
 	__bss_start = .;              /* BSS */

Quoted 39 lines · Jesper Nilsson

Quoted 30 lines · Yuusei KUWANA
Quoted 4 lines
arch/cris/arch-v10/vmlinux.lds.S
fix boot problem
* too old initcall style. replace INITCALLS macro
* __init_begin, __init_end move for free_initmem()
Hi,
The conversion to INITCALLS is ok (I have the same change here), but I'm curious what problem you are solving with moving __init_begin and __init_end to only include the .init.data.
If there is use of initcalls/data outside init then that is a bug and should be fixed, not worked around.
There might however be another problem that Mikael Starvik told me about, the padding between initdata and the next section had gone AWOL, which made the free_initmem() destroy useful data.
With the attached patch (and a lot of other patches) I am successful in booting an Axis Devboard82+.
Could you please try the included patch and see if that solves your problem?
Quoted 2 lines
Note: with this patch kernel boot and mount root, but after init done, kernel panic at do_signal() ...
I have a patch in the pipe that should fix this, will send it in just a small while.
Quoted 7 lines
Quoted 1 line
ryu
Thank you for your help!
Best regards,
/Jesper

Quoted 8 lines

/^JN - Jesper Nilsson
--
Jesper Nilsson -- [email protected]
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

/^JN - Jesper Nilsson
--
Jesper Nilsson -- [email protected]

Sam Ravnborg

2007-12-22 08:35 UTC #4 lore raw

Hi Jesper.

Quoted 13 lines

arch/cris/arch-v10/vmlinux.lds.S +2 -7

 		__initramfs_end = .;
-		/* We fill to the next page, so we can discard all init
-		   pages without needing to consider what payload might be
-		   appended to the kernel image.  */
-		FILL (0); 
-		. = ALIGN (8192);
 	}
 #endif
-	
 	__vmlinux_end = .;            /* last address of the physical file */
-  	__init_end = .;
+
+	/*

arch/cris/arch-v10/vmlinux.lds.S SECTIONS · line 99

Quoted 10 lines

arch/cris/arch-v10/vmlinux.lds.S +10 -0

+	 * We fill to the next page, so we can discard all init
+	 * pages without needing to consider what payload might be
+	 * appended to the kernel image.
+	 */
+	.init.fill : {
+		FILL (0);
+		. = ALIGN (8192);
+	}
+
+	__init_end = .;

Can we please have the hardcoded 8192 replaced with a proper constant. I assume this is PAGESIZE?
See hwat other archs does to make PAGESIZE available for assembly files and ldscripts. The macro AC is the key point here.

And then please explain why it make a diffrence to put the alignment in a section like you do.
I do not see it used in the other archs.

Sam

Jesper Nilsson

2008-01-03 17:02 UTC #5 lore raw

Quoted 20 lines · Sam Ravnborg

Hi Jesper.

Quoted 18 lines

arch/cris/arch-v10/vmlinux.lds.S +7 -7

 		__initramfs_end = .;
-		/* We fill to the next page, so we can discard all init
-		   pages without needing to consider what payload might be
-		   appended to the kernel image.  */
-		FILL (0); 
-		. = ALIGN (8192);
 	}
 #endif
-	
 	__vmlinux_end = .;            /* last address of the physical file */
-  	__init_end = .;
+
+	/*
+	 * We fill to the next page, so we can discard all init
+	 * pages without needing to consider what payload might be
+	 * appended to the kernel image.
+	 */
+	.init.fill : {

Quoted 8 lines

arch/cris/arch-v10/vmlinux.lds.S SECTIONS · line 99
Quoted 5 lines
arch/cris/arch-v10/vmlinux.lds.S +5 -0
+		FILL (0);
+		. = ALIGN (8192);
+	}
+
+	__init_end = .;
Can we please have the hardcoded 8192 replaced with a proper constant. I assume this is PAGESIZE?

Yes, quite so.

Quoted 2 lines

See hwat other archs does to make PAGESIZE available for assembly files and ldscripts. The macro AC is the key point here.

Ok, will do.

Quoted 3 lines

And then please explain why it make a diffrence to put the alignment in a section like you do.
I do not see it used in the other archs.

I'm not an expert on ld-scripts, so I only know that the linker didn't respect the other way of doing the alignment, but did when added in a section.

Perhaps Mikael has a better explanation when he's back from vacation.

Quoted 1 line

Sam

Best regards,

/^JN - Jesper Nilsson
--
Jesper Nilsson -- [email protected]