Skip to contentSkip to conversation

export native versions of machine_ops functions

PATCH In review Glauber Costa 4 participants 6 messages Mar 5, 2008 – Mar 6, 2008 x86

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 export native versions of machine_ops functions

People overriding machine_ops provided functions may want to call the native version after its pre-processing. It already happens for the smp_ops functions, so I don't see a reason for avoiding it here.

---
 arch/x86/kernel/reboot.c |   10 +++++-----
 include/asm-x86/reboot.h |    5 +++++
 2 files changed, 10 insertions(+), 5 deletions(-)
arch/x86/kernel/reboot.c +5 -5
index 7fd6ac4..8b577f1 100644
@@ -326,7 +326,7 @@ static inline void kb_wait(void)
 	}
 }
 
-static void native_machine_emergency_restart(void)
+void native_machine_emergency_restart(void)
 {
 	int i;
 
@@ -376,7 +376,7 @@ #endif
 	}
 }
 
-static void native_machine_shutdown(void)
+void native_machine_shutdown(void)
 {
 	/* Stop the cpus and apics */
 #ifdef CONFIG_SMP
@@ -420,7 +420,7 @@ #ifdef CONFIG_X86_64
 #endif
 }
 
-static void native_machine_restart(char *__unused)
+void native_machine_restart(char *__unused)
 {
 	printk("machine restart\n");
 
@@ -429,11 +429,11 @@ static void native_machine_restart(char 
 	machine_emergency_restart();
 }
 
-static void native_machine_halt(void)
+void native_machine_halt(void)
 {
 }
 
-static void native_machine_power_off(void)
+void native_machine_power_off(void)
 {
 	if (pm_power_off) {
 		if (!reboot_force)
include/asm-x86/reboot.h +5 -0
index e9e3ffc..fb1bdf1 100644
@@ -16,5 +16,10 @@ struct machine_ops
 extern struct machine_ops machine_ops;
 
 void machine_real_restart(unsigned char *code, int length);
+void native_machine_emergency_restart(void);
+void native_machine_shutdown(void);
+void native_machine_restart(char *__unused);
+void native_machine_halt(void);
+void native_machine_power_off(void);
 
 #endif	/* _ASM_REBOOT_H */
Signature · 1 line
1.4.2
#1 lore raw
Quoted 3 lines · Glauber Costa

People overriding machine_ops provided functions may want to call the native version after its pre-processing. It already happens for the smp_ops functions, so I don't see a reason for avoiding it here.

i'd like to do this only if there are actual users - otherwise we'll get a trickle of "make needlessly global functions static" patches a few weeks later ;-)

Ingo

#2 lore raw
Quoted 7 lines · Ingo Molnar
Quoted 3 lines · Glauber Costa

People overriding machine_ops provided functions may want to call the native version after its pre-processing. It already happens for the smp_ops functions, so I don't see a reason for avoiding it here.

Quoted 6 lines

i'd like to do this only if there are actual users - otherwise we'll get a trickle of "make needlessly global functions static" patches a few weeks later ;-)

Ingo

KVM is a user for that. Because it registers a memory area in which the host will be continually writting to (the clock), when it shutdown, the host
have no way to know it has to stop. If you're really shutting down, there's no problem. But if you're kexecing, for example, the host will keep writting to a random
memory area after the new kernel installs.

Signature · 5 lines
Glauber  Costa.
"Free as in Freedom"
http://glommer.net

"The less confident you are, the more serious you have to act."
#3 lore raw
Quoted 10 lines · Glauber Costa
Quoted 3 lines

i'd like to do this only if there are actual users - otherwise we'll get a trickle of "make needlessly global functions static" patches a few weeks later ;-)

KVM is a user for that. Because it registers a memory area in which the host will be continually writting to (the clock), when it shutdown, the host have no way to know it has to stop. If you're really shutting down, there's no problem. But if you're kexecing, for example, the host will keep writting to a random memory area after the new kernel installs.

i mean, i agree in general - but KVM is not a user of any of those currently static functions. So i've applied your patch but lets make sure those functions really get used :)

Ingo

#4 lore raw
Quoted 12 lines · Ingo Molnar
Quoted 6 lines · Glauber Costa
Quoted 4 lines

i'd like to do this only if there are actual users - otherwise we'll get a trickle of "make needlessly global functions static" patches a few weeks later ;-)

KVM is a user for that. Because it registers a memory area in which the host will be continually writting to (the clock), when it shutdown, the host have no way to know it has to stop. If you're really shutting down, there's no problem. But if you're kexecing, for example, the host will keep writting to a random memory area after the new kernel installs.

i mean, i agree in general - but KVM is not a user of any of those currently static functions. So i've applied your patch but lets make sure those functions really get used :)

The user will be in my 2.6.26 queue. Unfortunately it means a dependency between kvm.git and x86.git, again.

Signature · 1 line
Do not meddle in the internals of kernels, for they are subtle and quick to panic.
#5 lore raw
Quoted 6 lines · Avi Kivity
Quoted 3 lines

i mean, i agree in general - but KVM is not a user of any of those currently static functions. So i've applied your patch but lets make sure those functions really get used :)

The user will be in my 2.6.26 queue. Unfortunately it means a dependency between kvm.git and x86.git, again.

a couple of solutions:

- add the same patch to kvm.git (find it below), integrators ought to be able to detect and exclude it

- i could start an x86.git#core tree that would include the for-linus bits and any agreed-upon infrastructure changes [like this one]. (but it would not include any other more experimental x86.git bits) You could base kvm.git on that and track it. This makes integration even easier because the commit IDs would be the same and git would be able to resolve it. [but note, x86.git#core would not be an append-only tree, it's a daily-rebased tree]

Ingo

------------------>
Subject: x86: export native versions of machine_ops functions
From: Glauber Costa <[email protected]>
Date: Wed, 5 Mar 2008 14:44:00 -0300

People overriding machine_ops provided functions may want to call the native version after its pre-processing. It already happens for the smp_ops functions, so I don't see a reason for avoiding it here.

---
 arch/x86/kernel/reboot.c |   10 +++++-----
 include/asm-x86/reboot.h |    5 +++++
 2 files changed, 10 insertions(+), 5 deletions(-)
arch/x86/kernel/reboot.c +5 -5
Index: linux-x86.q/arch/x86/kernel/reboot.c
===================================================================
@@ -326,7 +326,7 @@ static inline void kb_wait(void)
 	}
 }
 
-static void native_machine_emergency_restart(void)
+void native_machine_emergency_restart(void)
 {
 	int i;
 
@@ -376,7 +376,7 @@ static void native_machine_emergency_res
 	}
 }
 
-static void native_machine_shutdown(void)
+void native_machine_shutdown(void)
 {
 	/* Stop the cpus and apics */
 #ifdef CONFIG_SMP
@@ -432,7 +432,7 @@ static void native_machine_shutdown(void
 #endif
 }
 
-static void native_machine_restart(char *__unused)
+void native_machine_restart(char *__unused)
 {
 	printk("machine restart\n");
 
@@ -441,11 +441,11 @@ static void native_machine_restart(char 
 	machine_emergency_restart();
 }
 
-static void native_machine_halt(void)
+void native_machine_halt(void)
 {
 }
 
-static void native_machine_power_off(void)
+void native_machine_power_off(void)
 {
 	if (pm_power_off) {
 		if (!reboot_force)
include/asm-x86/reboot.h +5 -0
Index: linux-x86.q/include/asm-x86/reboot.h
===================================================================
@@ -16,5 +16,10 @@ struct machine_ops
 extern struct machine_ops machine_ops;
 
 void machine_real_restart(unsigned char *code, int length);
+void native_machine_emergency_restart(void);
+void native_machine_shutdown(void);
+void native_machine_restart(char *__unused);
+void native_machine_halt(void);
+void native_machine_power_off(void);
 
 #endif	/* _ASM_REBOOT_H */