Skip to contentSkip to conversation

build fix for smp

PATCH In review Glauber de Oliveira Costa 3 participants 6 messages Mar 19, 2008 – Apr 28, 2008 x86

lore
Share & tools
lore

Glauber de Oliveira Costa

author
#0 lore raw

PATCH 1/1 build fix for smp

From: Glauber Costa <[email protected]>

smp.c won't build when reboot.c is not compiled in, due to the reboot_force dependency. It is always okay for x86_64, but should be conditional on x86_32

Signed-off-by Glauber Costa <[email protected]>
---
 arch/x86/kernel/smp.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)
arch/x86/kernel/smp.c +2 -0
index 16c52aa..0ee8ad5 100644
@@ -274,8 +274,10 @@ static void native_smp_send_stop(void)
 	int nolock;
 	unsigned long flags;
 
+#if defined(CONFIG_X86_64) || defined(CONFIG_X86_BIOS_REBOOT)
 	if (reboot_force)
 		return;
+#endif
 
 	/* Don't deadlock on the call lock in panic */
 	nolock = !spin_trylock(&call_lock);
Signature · 1 line
1.5.0.6
#1 lore raw

arch/x86/kernel/smp.c static void native_smp_send_stop(void) · line 280

Quoted 4 lines · Glauber de Oliveira Costa
arch/x86/kernel/smp.c +2 -0
+#if defined(CONFIG_X86_64) || defined(CONFIG_X86_BIOS_REBOOT)
 	if (reboot_force)
 		return;
+#endif

please solve it slightly differently: introduce a uniformly usable reboot_force flag, that will always default to the constant of 1 if !X86_64 && !X86_BIOS_REBOOT. That makes the code cleaner - and not the least i can also merge the fix earlier in the series, without having to redo the big smp.c movement patch :-)

Ingo

#2 lore raw
Quoted 14 lines · Ingo Molnar

arch/x86/kernel/smp.c static void native_smp_send_stop(void) · line 280

Quoted 4 lines · Glauber de Oliveira Costa
arch/x86/kernel/smp.c +2 -0
+#if defined(CONFIG_X86_64) || defined(CONFIG_X86_BIOS_REBOOT)
      if (reboot_force)
              return;
+#endif

please solve it slightly differently: introduce a uniformly usable reboot_force flag, that will always default to the constant of 1 if !X86_64 && !X86_BIOS_REBOOT. That makes the code cleaner - and not the least i can also merge the fix earlier in the series, without having to redo the big smp.c movement patch :-)

Quoted 2 lines

Ingo

Hello,
what's the final fix for this issue? 2.6.25-git11 is still broken, just gave me

arch/x86/kernel/built-in.o: In function `native_smp_send_stop':
smp.c:(.text+0xc751): undefined reference to `reboot_force'
make: *** [.tmp_vmlinux1] Error 1

for a randconfig. Same reported at http://lkml.org/lkml/2008/4/27/39

thanks,
-sergio

#3 lore raw
Quoted 8 lines · Sergio Luis

Hello, what's the final fix for this issue? 2.6.25-git11 is still broken, just gave me

arch/x86/kernel/built-in.o: In function `native_smp_send_stop':
smp.c:(.text+0xc751): undefined reference to `reboot_force'
make: *** [.tmp_vmlinux1] Error 1

for a randconfig. Same reported at http://lkml.org/lkml/2008/4/27/39

btw., that's a VISWS config - that wont boot on any PC. The patch below should fix this.

Ingo

--------------------------->
Subject: x86 VISWS: build fix
From: Ingo Molnar <[email protected]>
Date: Mon Apr 28 10:46:58 CEST 2008

the 'reboot_force' flag is a notion that non-PC subarchitectures do not have.

Signed-off-by Ingo Molnar <[email protected]>
---
 include/asm-x86/proto.h |    4 ++++
 1 file changed, 4 insertions(+)
include/asm-x86/proto.h +4 -0
Index: linux/include/asm-x86/proto.h
===================================================================
@@ -20,7 +20,11 @@ extern void syscall32_cpu_init(void);
 
 extern void check_efer(void);
 
+#ifdef X86_BIOS_REBOOT
 extern int reboot_force;
+#else
+static const int reboot_force = 0;
+#endif
 
 long do_arch_prctl(struct task_struct *task, int code, unsigned long addr);
#4 lore raw
Quoted 6 lines · Ingo Molnar
Quoted 2 lines · Sergio Luis

Hello, what's the final fix for this issue? 2.6.25-git11 is still broken, just gave me

Quoted 8 lines
Quoted 5 lines

arch/x86/kernel/built-in.o: In function `native_smp_send_stop':
smp.c:(.text+0xc751): undefined reference to `reboot_force'
make: *** [.tmp_vmlinux1] Error 1

for a randconfig. Same reported at http://lkml.org/lkml/2008/4/27/39

btw., that's a VISWS config - that wont boot on any PC. The patch below should fix this.

Yep, it does fix it thanks.
-sergio

Quoted 33 lines

Ingo

--------------------------->
Subject: x86 VISWS: build fix
From: Ingo Molnar <[email protected]>
Date: Mon Apr 28 10:46:58 CEST 2008

the 'reboot_force' flag is a notion that non-PC subarchitectures do not have.

Signed-off-by Ingo Molnar <[email protected]>

---

  include/asm-x86/proto.h |    4 ++++
  1 file changed, 4 insertions(+)

Index: linux/include/asm-x86/proto.h
===================================================================
--- linux.orig/include/asm-x86/proto.h
+++ linux/include/asm-x86/proto.h
@@ -20,7 +20,11 @@ extern void syscall32_cpu_init(void);

  extern void check_efer(void);

+#ifdef X86_BIOS_REBOOT
extern int reboot_force;
+#else
+static const int reboot_force = 0;
+#endif

long do_arch_prctl(struct task_struct *task, int code, unsigned long addr);

author
#5 lore raw
In reply to Sergio Luis
Quoted 14 lines · Sergio Luis
Quoted 13 lines · Ingo Molnar

arch/x86/kernel/smp.c static void native_smp_send_stop(void) · line 280

Quoted 4 lines · Glauber de Oliveira Costa
arch/x86/kernel/smp.c +2 -0
+#if defined(CONFIG_X86_64) || defined(CONFIG_X86_BIOS_REBOOT)
      if (reboot_force)
              return;
+#endif

please solve it slightly differently: introduce a uniformly usable reboot_force flag, that will always default to the constant of 1 if !X86_64 && !X86_BIOS_REBOOT. That makes the code cleaner - and not the least i can also merge the fix earlier in the series, without having to redo the big smp.c movement patch :-)

Quoted 5 lines
Quoted 2 lines

Ingo

Hello,
what's the final fix for this issue? 2.6.25-git11 is still broken, just gave me

Doh. I had the patch, but completely forgot to send it to ingo. Thanks for raising this issue again. (But I saw ingo already put a fix for this in)

Quoted 8 lines

arch/x86/kernel/built-in.o: In function `native_smp_send_stop':
smp.c:(.text+0xc751): undefined reference to `reboot_force'
make: *** [.tmp_vmlinux1] Error 1

for a randconfig. Same reported at http://lkml.org/lkml/2008/4/27/39

thanks,
-sergio