Skip to contentSkip to conversation

Fix hard_smp_processor_id compile error

ARM Simon Horman 4 participants 9 messages May 15, 2007

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

"Remove hardcoding of hard_smp_processor_id on UP systems", 2f4dfe206a2fc07099dfad77a8ea2f4b4ae2140f in Linus' tree, moved the definition of hard_smp_processor_id linux/smp.h to asm/smp.h for UP systems. This causes a regression on ARM as the definition was not added to asm-arm/smp.h.

Cc Fernando Luis Vazquez Cao <[email protected]>

---

 arch/arm/mach-integrator/core.c |    1 +
 include/asm-arm/smp.h           |   10 ++++++----
 2 files changed, 7 insertions(+), 4 deletions(-)

arm-unknown-linux-gnu-gcc (GCC) 4.1.1
Copyright (C) 2006 Free Software Foundation, Inc.
This is free software; see the source for copying conditions. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

CC arch/arm/mach-integrator/core.o
arch/arm/mach-integrator/core.c: In function 'integrator_timer_interrupt':
arch/arm/mach-integrator/core.c:264: warning: implicit declaration of function 'hard_smp_processor_id' [snip]
LD .tmp_vmlinux1
arch/arm/mach-integrator/built-in.o: In function `integrator_timer_interrupt':
cpu.c:(.text+0x398): undefined reference to `hard_smp_processor_id'
make: *** [.tmp_vmlinux1] error 1

This was produced by running the following in a crosstool cross-compiling environment:
cp arch/arm/configs/integrator_defconfig .config
yes "" | make modconfig
make

arch/arm/mach-integrator/core.c +1 -0
Index: linux-2.6/arch/arm/mach-integrator/core.c
===================================================================
@@ -28,6 +28,7 @@
 #include <asm/system.h>
 #include <asm/leds.h>
 #include <asm/mach/time.h>
+#include <asm/smp.h>
 
 #include "common.h"
 
include/asm-arm/smp.h +6 -4
Index: linux-2.6/include/asm-arm/smp.h
===================================================================
@@ -10,16 +10,16 @@
 #ifndef __ASM_ARM_SMP_H
 #define __ASM_ARM_SMP_H
 
+#ifndef CONFIG_SMP
+#define hard_smp_processor_id()		0
+#else
+
 #include <linux/threads.h>
 #include <linux/cpumask.h>
 #include <linux/thread_info.h>
 
 #include <asm/arch/smp.h>
 
-#ifndef CONFIG_SMP
-# error "<asm-arm/smp.h> included in non-SMP build"
-#endif
-
 #define raw_smp_processor_id() (current_thread_info()->cpu)
 
 /*
@@ -134,4 +134,6 @@ extern void show_local_irqs(struct seq_f
  */
 asmlinkage void do_local_timer(struct pt_regs *);
 
+#endif /* CONFIG_SMP */
+
 #endif /* ifndef __ASM_ARM_SMP_H */
#1 lore raw
Quoted 5 lines · Simon Horman

"Remove hardcoding of hard_smp_processor_id on UP systems", 2f4dfe206a2fc07099dfad77a8ea2f4b4ae2140f in Linus' tree, moved the definition of hard_smp_processor_id linux/smp.h to asm/smp.h for UP systems. This causes a regression on ARM as the definition was not added to asm-arm/smp.h.

Hi Simon!

Thank you for catching and fixing these compile errors. I should install a cross-compiler ASAP.

- Fernando

Quoted 72 lines
Cc Fernando Luis Vazquez Cao <[email protected]>

---

 arch/arm/mach-integrator/core.c |    1 +
 include/asm-arm/smp.h           |   10 ++++++----
 2 files changed, 7 insertions(+), 4 deletions(-)

arm-unknown-linux-gnu-gcc (GCC) 4.1.1
Copyright (C) 2006 Free Software Foundation, Inc.
This is free software; see the source for copying conditions. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

CC arch/arm/mach-integrator/core.o
arch/arm/mach-integrator/core.c: In function 'integrator_timer_interrupt':
arch/arm/mach-integrator/core.c:264: warning: implicit declaration of function 'hard_smp_processor_id' [snip]
LD .tmp_vmlinux1
arch/arm/mach-integrator/built-in.o: In function `integrator_timer_interrupt':
cpu.c:(.text+0x398): undefined reference to `hard_smp_processor_id'
make: *** [.tmp_vmlinux1] error 1

This was produced by running the following in a crosstool cross-compiling environment:
cp arch/arm/configs/integrator_defconfig .config
yes "" | make modconfig
make

Index: linux-2.6/arch/arm/mach-integrator/core.c
===================================================================

arch/arm/mach-integrator/core.c +1 -0
@@ -28,6 +28,7 @@
 #include <asm/system.h>
 #include <asm/leds.h>
 #include <asm/mach/time.h>
+#include <asm/smp.h>
 
 #include "common.h"
 

Index: linux-2.6/include/asm-arm/smp.h
===================================================================

include/asm-arm/smp.h +6 -4
@@ -10,16 +10,16 @@
 #ifndef __ASM_ARM_SMP_H
 #define __ASM_ARM_SMP_H
 
+#ifndef CONFIG_SMP
+#define hard_smp_processor_id()		0
+#else
+
 #include <linux/threads.h>
 #include <linux/cpumask.h>
 #include <linux/thread_info.h>
 
 #include <asm/arch/smp.h>
 
-#ifndef CONFIG_SMP
-# error "<asm-arm/smp.h> included in non-SMP build"
-#endif
-
 #define raw_smp_processor_id() (current_thread_info()->cpu)
 
 /*
@@ -134,4 +134,6 @@ extern void show_local_irqs(struct seq_f
  */
 asmlinkage void do_local_timer(struct pt_regs *);
 
+#endif /* CONFIG_SMP */
+
 #endif /* ifndef __ASM_ARM_SMP_H */
#2 lore raw
Quoted 10 lines · On Tue, 15 May 2007 11:59:41 +0900 Fernando Luis Vázquez Cao…
Quoted 5 lines · Simon Horman

"Remove hardcoding of hard_smp_processor_id on UP systems", 2f4dfe206a2fc07099dfad77a8ea2f4b4ae2140f in Linus' tree, moved the definition of hard_smp_processor_id linux/smp.h to asm/smp.h for UP systems. This causes a regression on ARM as the definition was not added to asm-arm/smp.h.

Hi Simon!

Thank you for catching and fixing these compile errors. I should install a cross-compiler ASAP.

Mine are at http://userweb.kernel.org/~akpm/cross-compilers/

#3 lore raw
Quoted 8 lines · Andrew Morton
Quoted 6 lines · On Tue, 15 May 2007 11:59:41 +0900 Fernando Luis Vázquez Cao…
Quoted 5 lines · Simon Horman

"Remove hardcoding of hard_smp_processor_id on UP systems", 2f4dfe206a2fc07099dfad77a8ea2f4b4ae2140f in Linus' tree, moved the definition of hard_smp_processor_id linux/smp.h to asm/smp.h for UP systems. This causes a regression on ARM as the definition was not added to asm-arm/smp.h.

Quoted 6 lines
Quoted 4 lines

Hi Simon!

Thank you for catching and fixing these compile errors. I should install a cross-compiler ASAP.

Mine are at http://userweb.kernel.org/~akpm/cross-compilers/

Do you have / know of tools to create a set of configs for testing? What I am thinking of is something that takes the default configs, then toggles SMP, PREEMT and stuff like that, testing all the permutations.

Signature · 3 lines
Horms
  H: http://www.vergenet.net/~horms/
  W: http://www.valinux.co.jp/en/
#4 lore raw
Quoted 14 lines · On Tue, 15 May 2007 13:54:23 +0900 Simon Horman <[email protected]> wrote:
Quoted 13 lines · Andrew Morton
Quoted 10 lines · On Tue, 15 May 2007 11:59:41 +0900 Fernando Luis Vázquez Cao…
Quoted 5 lines · Simon Horman

"Remove hardcoding of hard_smp_processor_id on UP systems", 2f4dfe206a2fc07099dfad77a8ea2f4b4ae2140f in Linus' tree, moved the definition of hard_smp_processor_id linux/smp.h to asm/smp.h for UP systems. This causes a regression on ARM as the definition was not added to asm-arm/smp.h.

Hi Simon!

Thank you for catching and fixing these compile errors. I should install a cross-compiler ASAP.

Quoted 6 lines
Quoted 1 line

Mine are at http://userweb.kernel.org/~akpm/cross-compilers/

Do you have / know of tools to create a set of configs for testing? What I am thinking of is something that takes the default configs, then toggles SMP, PREEMT and stuff like that, testing all the permutations.

`make randconfig'?

#5 lore raw
Quoted 16 lines · Andrew Morton
Quoted 14 lines · On Tue, 15 May 2007 13:54:23 +0900 Simon Horman <[email protected]> wrote:
Quoted 13 lines · Andrew Morton
Quoted 10 lines · On Tue, 15 May 2007 11:59:41 +0900 Fernando Luis Vázquez Cao…
Quoted 5 lines · Simon Horman

"Remove hardcoding of hard_smp_processor_id on UP systems", 2f4dfe206a2fc07099dfad77a8ea2f4b4ae2140f in Linus' tree, moved the definition of hard_smp_processor_id linux/smp.h to asm/smp.h for UP systems. This causes a regression on ARM as the definition was not added to asm-arm/smp.h.

Hi Simon!

Thank you for catching and fixing these compile errors. I should install a cross-compiler ASAP.

Quoted 8 lines
Quoted 6 lines
Quoted 1 line

Mine are at http://userweb.kernel.org/~akpm/cross-compilers/

Do you have / know of tools to create a set of configs for testing? What I am thinking of is something that takes the default configs, then toggles SMP, PREEMT and stuff like that, testing all the permutations.

`make randconfig'?

I was thinking of something a little more predictable. But that end up could working just as well.

Signature · 3 lines
Horms
  H: http://www.vergenet.net/~horms/
  W: http://www.valinux.co.jp/en/
#6 lore raw
In reply to Simon Horman
Quoted 8 lines · Simon Horman

"Remove hardcoding of hard_smp_processor_id on UP systems", 2f4dfe206a2fc07099dfad77a8ea2f4b4ae2140f in Linus' tree, moved the definition of hard_smp_processor_id linux/smp.h to asm/smp.h for UP systems. This causes a regression on ARM as the definition was not added to asm-arm/smp.h.

Cc Fernando Luis Vazquez Cao <[email protected]>

I took an alternative approach - since Integrator never had the complete SMP support merged, I ripped it out of that platform. Whether realview suffers from the same thing or not I'm not sure - uniprocessor realview is not something covered by the ARM default configuration files, and therefore kautobuild won't check that configuration.

However, the things I find annoying about this is:

1. asm/smp.h was _never_ included in ARM UP builds prior to this change.
2. we have linux/smp.h and the general rule is that if asm/foo.h and linux/foo.h are present, linux/foo.h is included in preference to asm/foo.h

Given the amount of janitors we now have looking at the kernel code,
(2) is a big concern - it could mean a constant stream of patches trying to "fix" files including asm/smp.h.

I would suggest that if there is a desire to include asm/smp.h on UP builds, the inclusion of it by linux/smp.h is made unconditional.

Signature · 3 lines
Russell King
 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/
 maintainer of:
#7 lore raw
Quoted 7 lines · Russell King
Quoted 6 lines · Simon Horman

"Remove hardcoding of hard_smp_processor_id on UP systems", 2f4dfe206a2fc07099dfad77a8ea2f4b4ae2140f in Linus' tree, moved the definition of hard_smp_processor_id linux/smp.h to asm/smp.h for UP systems. This causes a regression on ARM as the definition was not added to asm-arm/smp.h.

Quoted 8 lines
Quoted 2 lines
Cc Fernando Luis Vazquez Cao <[email protected]>

I took an alternative approach - since Integrator never had the complete SMP support merged, I ripped it out of that platform. Whether realview suffers from the same thing or not I'm not sure - uniprocessor realview is not something covered by the ARM default configuration files, and therefore kautobuild won't check that configuration.

I can take a look into this. Is there a tree other than Linus' I should be working with?

Quoted 7 lines

However, the things I find annoying about this is:

1. asm/smp.h was _never_ included in ARM UP builds prior to this change.
2. we have linux/smp.h and the general rule is that if asm/foo.h and linux/foo.h are present, linux/foo.h is included in preference to asm/foo.h

Quoted 6 lines

Given the amount of janitors we now have looking at the kernel code,
(2) is a big concern - it could mean a constant stream of patches trying to "fix" files including asm/smp.h.

I would suggest that if there is a desire to include asm/smp.h on UP builds, the inclusion of it by linux/smp.h is made unconditional.

I was thinking that myself, though I'm kind of fearful of what kind of (additional) breakage it might cause.

#8 lore raw
Quoted 11 lines · Simon Horman
Quoted 10 lines · Russell King
Quoted 8 lines · Simon Horman

"Remove hardcoding of hard_smp_processor_id on UP systems", 2f4dfe206a2fc07099dfad77a8ea2f4b4ae2140f in Linus' tree, moved the definition of hard_smp_processor_id linux/smp.h to asm/smp.h for UP systems. This causes a regression on ARM as the definition was not added to asm-arm/smp.h.

Cc Fernando Luis Vazquez Cao <[email protected]>
Quoted 8 lines
Quoted 5 lines

I took an alternative approach - since Integrator never had the complete SMP support merged, I ripped it out of that platform. Whether realview suffers from the same thing or not I'm not sure - uniprocessor realview is not something covered by the ARM default configuration files, and therefore kautobuild won't check that configuration.

I can take a look into this. Is there a tree other than Linus' I should be working with?

Looking at Linus' it seems that for realview UP builds based on realview_defconfig don't suffer hard_smp_processor_id breakage.