[PATCH 0/10] Integrate msr.h

[PATCH 0/10] Integrate msr.h

[Glauber de Oliveira Costa]

Hello,

This series of patches integrates msr.h header.
What it really does, is a series of steps to allow us
to get rid of duplicate code between i386 and x86_64 versions

With this done, achieving paravirt for x86_64 gets really easy,
just a couple of extra code.

The first patch was already sent a while ago, but was not yet pushed
to any tree , to my knowledge. So it is sent again. Also, 
patches 9 and 10 relies on b02f15537b3bf43e347214cf14bad80aeaef1caf,
already in the x86 tree.

Other than that, it should apply fine ontop of linus today git.

Comments are welcome

[PATCH 1/10] Wipe out traditional opt from x86_64 Makefile

[Glauber de Oliveira Costa]

Among other things, using -traditional as a gcc option stops us from
using macro token pasting, which is a feature we heavily rely on.

There was still a use of -traditional in arch/x86/kernel/Makefile_64,
which this patch removes.

I don't see any problems building kernels in my x86_64 box without
-traditional.

Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Acked-by: Jeremy Fitzhardinge <jeremy@xensource.com>
---
 arch/x86/kernel/Makefile_64 |    1 -
 1 files changed, 0 insertions(+), 1 deletions(-)

diff --git a/arch/x86/kernel/Makefile_64 b/arch/x86/kernel/Makefile_64
index 5a88890..da0b1e5 100644
--- a/arch/x86/kernel/Makefile_64
+++ b/arch/x86/kernel/Makefile_64
@@ -4,7 +4,6 @@
 
 extra-y 	:= head_64.o head64.o init_task.o vmlinux.lds
 CPPFLAGS_vmlinux.lds += -Ux86_64
-EXTRA_AFLAGS	:= -traditional
 
 obj-y	:= process_64.o signal_64.o entry_64.o traps_64.o irq_64.o \
 		ptrace_64.o time_64.o ioport_64.o ldt_64.o setup_64.o i8259_64.o sys_x86_64.o \
-- 
1.4.4.2

[PATCH 1/10] Wipe out traditional opt from x86_64 Makefile

[Glauber de Oliveira Costa]

Among other things, using -traditional as a gcc option stops us from
using macro token pasting, which is a feature we heavily rely on.

There was still a use of -traditional in arch/x86/kernel/Makefile_64,
which this patch removes.

I don't see any problems building kernels in my x86_64 box without
-traditional.

Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Acked-by: Jeremy Fitzhardinge <jeremy@xensource.com>
---
 arch/x86/kernel/Makefile_64 |    1 -
 1 files changed, 0 insertions(+), 1 deletions(-)

diff --git a/arch/x86/kernel/Makefile_64 b/arch/x86/kernel/Makefile_64
index 5a88890..da0b1e5 100644
--- a/arch/x86/kernel/Makefile_64
+++ b/arch/x86/kernel/Makefile_64
@@ -4,7 +4,6 @@
 
 extra-y 	:= head_64.o head64.o init_task.o vmlinux.lds
 CPPFLAGS_vmlinux.lds += -Ux86_64
-EXTRA_AFLAGS	:= -traditional
 
 obj-y	:= process_64.o signal_64.o entry_64.o traps_64.o irq_64.o \
 		ptrace_64.o time_64.o ioport_64.o ldt_64.o setup_64.o i8259_64.o sys_x86_64.o \
-- 
1.4.4.2

[PATCH 2/10] unify msr smp funcs

[Glauber de Oliveira Costa]

The functions under #ifdef CONFIG_SMP in msr.h are the same
for both x86_64 and i386, and this patches removes one of them,
putting them in a single location

Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
---
 include/asm-x86/msr.h |   33 +++++++--------------------------
 1 files changed, 7 insertions(+), 26 deletions(-)

diff --git a/include/asm-x86/msr.h b/include/asm-x86/msr.h
index ba4b314..b3cd509 100644
--- a/include/asm-x86/msr.h
+++ b/include/asm-x86/msr.h
@@ -135,29 +135,6 @@ static inline int wrmsr_safe(u32 __msr, u32 __low, u32 __high)
 	} while(0)
 #endif	/* !CONFIG_PARAVIRT */
 
-#ifdef CONFIG_SMP
-void rdmsr_on_cpu(unsigned int cpu, u32 msr_no, u32 *l, u32 *h);
-void wrmsr_on_cpu(unsigned int cpu, u32 msr_no, u32 l, u32 h);
-int rdmsr_safe_on_cpu(unsigned int cpu, u32 msr_no, u32 *l, u32 *h);
-int wrmsr_safe_on_cpu(unsigned int cpu, u32 msr_no, u32 l, u32 h);
-#else  /*  CONFIG_SMP  */
-static inline void rdmsr_on_cpu(unsigned int cpu, u32 msr_no, u32 *l, u32 *h)
-{
-	rdmsr(msr_no, *l, *h);
-}
-static inline void wrmsr_on_cpu(unsigned int cpu, u32 msr_no, u32 l, u32 h)
-{
-	wrmsr(msr_no, l, h);
-}
-static inline int rdmsr_safe_on_cpu(unsigned int cpu, u32 msr_no, u32 *l, u32 *h)
-{
-	return rdmsr_safe(msr_no, l, h);
-}
-static inline int wrmsr_safe_on_cpu(unsigned int cpu, u32 msr_no, u32 l, u32 h)
-{
-	return wrmsr_safe(msr_no, l, h);
-}
-#endif  /*  CONFIG_SMP  */
 #endif  /* ! __ASSEMBLY__ */
 #endif  /* __KERNEL__ */
 
@@ -320,6 +297,12 @@ static inline unsigned int cpuid_edx(unsigned int op)
 	return edx;
 }
 
+#endif  /* __ASSEMBLY__ */
+
+#endif  /* !__i386__ */
+
+#ifndef __ASSEMBLY__
+
 #ifdef CONFIG_SMP
 void rdmsr_on_cpu(unsigned int cpu, u32 msr_no, u32 *l, u32 *h);
 void wrmsr_on_cpu(unsigned int cpu, u32 msr_no, u32 l, u32 h);
@@ -343,8 +326,6 @@ static inline int wrmsr_safe_on_cpu(unsigned int cpu, u32 msr_no, u32 l, u32 h)
 	return wrmsr_safe(msr_no, l, h);
 }
 #endif  /* CONFIG_SMP */
-#endif  /* __ASSEMBLY__ */
-
-#endif  /* !__i386__ */
+#endif /* __ASSEMBLY__ */
 
 #endif
-- 
1.4.4.2

[PATCH 2/10] unify msr smp funcs

[Glauber de Oliveira Costa]

The functions under #ifdef CONFIG_SMP in msr.h are the same
for both x86_64 and i386, and this patches removes one of them,
putting them in a single location

Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
---
 include/asm-x86/msr.h |   33 +++++++--------------------------
 1 files changed, 7 insertions(+), 26 deletions(-)

diff --git a/include/asm-x86/msr.h b/include/asm-x86/msr.h
index ba4b314..b3cd509 100644
--- a/include/asm-x86/msr.h
+++ b/include/asm-x86/msr.h
@@ -135,29 +135,6 @@ static inline int wrmsr_safe(u32 __msr, u32 __low, u32 __high)
 	} while(0)
 #endif	/* !CONFIG_PARAVIRT */
 
-#ifdef CONFIG_SMP
-void rdmsr_on_cpu(unsigned int cpu, u32 msr_no, u32 *l, u32 *h);
-void wrmsr_on_cpu(unsigned int cpu, u32 msr_no, u32 l, u32 h);
-int rdmsr_safe_on_cpu(unsigned int cpu, u32 msr_no, u32 *l, u32 *h);
-int wrmsr_safe_on_cpu(unsigned int cpu, u32 msr_no, u32 l, u32 h);
-#else  /*  CONFIG_SMP  */
-static inline void rdmsr_on_cpu(unsigned int cpu, u32 msr_no, u32 *l, u32 *h)
-{
-	rdmsr(msr_no, *l, *h);
-}
-static inline void wrmsr_on_cpu(unsigned int cpu, u32 msr_no, u32 l, u32 h)
-{
-	wrmsr(msr_no, l, h);
-}
-static inline int rdmsr_safe_on_cpu(unsigned int cpu, u32 msr_no, u32 *l, u32 *h)
-{
-	return rdmsr_safe(msr_no, l, h);
-}
-static inline int wrmsr_safe_on_cpu(unsigned int cpu, u32 msr_no, u32 l, u32 h)
-{
-	return wrmsr_safe(msr_no, l, h);
-}
-#endif  /*  CONFIG_SMP  */
 #endif  /* ! __ASSEMBLY__ */
 #endif  /* __KERNEL__ */
 
@@ -320,6 +297,12 @@ static inline unsigned int cpuid_edx(unsigned int op)
 	return edx;
 }
 
+#endif  /* __ASSEMBLY__ */
+
+#endif  /* !__i386__ */
+
+#ifndef __ASSEMBLY__
+
 #ifdef CONFIG_SMP
 void rdmsr_on_cpu(unsigned int cpu, u32 msr_no, u32 *l, u32 *h);
 void wrmsr_on_cpu(unsigned int cpu, u32 msr_no, u32 l, u32 h);
@@ -343,8 +326,6 @@ static inline int wrmsr_safe_on_cpu(unsigned int cpu, u32 msr_no, u32 l, u32 h)
 	return wrmsr_safe(msr_no, l, h);
 }
 #endif  /* CONFIG_SMP */
-#endif  /* __ASSEMBLY__ */
-
-#endif  /* !__i386__ */
+#endif /* __ASSEMBLY__ */
 
 #endif
-- 
1.4.4.2

[PATCH 3/10] allow sched clock to be overridden by paravirt

[Glauber de Oliveira Costa]

This patch turns the sched_clock into native_sched_clock.
sched clock becomes a weak symbol, which can then give its
place to a paravirt definition.

Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
---
 arch/x86/kernel/tsc_64.c |   15 ++++++++++++++-
 1 files changed, 14 insertions(+), 1 deletions(-)

diff --git a/arch/x86/kernel/tsc_64.c b/arch/x86/kernel/tsc_64.c
index 9c70af4..5158476 100644
--- a/arch/x86/kernel/tsc_64.c
+++ b/arch/x86/kernel/tsc_64.c
@@ -30,7 +30,7 @@ static unsigned long long cycles_2_ns(unsigned long long cyc)
 	return (cyc * cyc2ns_scale) >> NS_SCALE;
 }
 
-unsigned long long sched_clock(void)
+unsigned long long native_sched_clock(void)
 {
 	unsigned long a = 0;
 
@@ -44,6 +44,19 @@ unsigned long long sched_clock(void)
 	return cycles_2_ns(a);
 }
 
+/* We need to define a real function for sched_clock, to override the
+   weak default version */
+#ifdef CONFIG_PARAVIRT
+unsigned long long sched_clock(void)
+{
+	return paravirt_sched_clock();
+}
+#else
+unsigned long long
+sched_clock(void) __attribute__((alias("native_sched_clock")));
+#endif
+
+
 static int tsc_unstable;
 
 inline int check_tsc_unstable(void)
-- 
1.4.4.2

[PATCH 3/10] allow sched clock to be overridden by paravirt

[Glauber de Oliveira Costa]

This patch turns the sched_clock into native_sched_clock.
sched clock becomes a weak symbol, which can then give its
place to a paravirt definition.

Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
---
 arch/x86/kernel/tsc_64.c |   15 ++++++++++++++-
 1 files changed, 14 insertions(+), 1 deletions(-)

diff --git a/arch/x86/kernel/tsc_64.c b/arch/x86/kernel/tsc_64.c
index 9c70af4..5158476 100644
--- a/arch/x86/kernel/tsc_64.c
+++ b/arch/x86/kernel/tsc_64.c
@@ -30,7 +30,7 @@ static unsigned long long cycles_2_ns(unsigned long long cyc)
 	return (cyc * cyc2ns_scale) >> NS_SCALE;
 }
 
-unsigned long long sched_clock(void)
+unsigned long long native_sched_clock(void)
 {
 	unsigned long a = 0;
 
@@ -44,6 +44,19 @@ unsigned long long sched_clock(void)
 	return cycles_2_ns(a);
 }
 
+/* We need to define a real function for sched_clock, to override the
+   weak default version */
+#ifdef CONFIG_PARAVIRT
+unsigned long long sched_clock(void)
+{
+	return paravirt_sched_clock();
+}
+#else
+unsigned long long
+sched_clock(void) __attribute__((alias("native_sched_clock")));
+#endif
+
+
 static int tsc_unstable;
 
 inline int check_tsc_unstable(void)
-- 
1.4.4.2

[PATCH 4/10] split get_cycles_sync

[Glauber de Oliveira Costa]

This patch splits get_cycles_sync() into  __get_cycles_sync(),
and the rdtscll part. Paravirt guests cannot issue rdtscl directly,
as it involves a function call in vdso area.

So, using the __get_cycles_sync() base, we introduce vget_cycles_sync,
which then calls the native version of rdtscll. Ideally, however, a guest
should define its own clocksource, together with a vread function

Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
---
 arch/x86/kernel/tsc_64.c |    2 +-
 include/asm-x86/tsc.h    |   37 +++++++++++++++++++++++++++++++++----
 2 files changed, 34 insertions(+), 5 deletions(-)

diff --git a/arch/x86/kernel/tsc_64.c b/arch/x86/kernel/tsc_64.c
index 5158476..4502539 100644
--- a/arch/x86/kernel/tsc_64.c
+++ b/arch/x86/kernel/tsc_64.c
@@ -269,7 +269,7 @@ static cycle_t read_tsc(void)
 
 static cycle_t __vsyscall_fn vread_tsc(void)
 {
-	cycle_t ret = (cycle_t)get_cycles_sync();
+	cycle_t ret = (cycle_t)vget_cycles_sync();
 	return ret;
 }
 
diff --git a/include/asm-x86/tsc.h b/include/asm-x86/tsc.h
index 6baab30..d6cb056 100644
--- a/include/asm-x86/tsc.h
+++ b/include/asm-x86/tsc.h
@@ -33,14 +33,14 @@ static inline cycles_t get_cycles(void)
 }
 
 /* Like get_cycles, but make sure the CPU is synchronized. */
-static __always_inline cycles_t get_cycles_sync(void)
+static __always_inline cycles_t __get_cycles_sync(void)
 {
 	unsigned long long ret;
 	unsigned eax, edx;
 
 	/*
-  	 * Use RDTSCP if possible; it is guaranteed to be synchronous
- 	 * and doesn't cause a VMEXIT on Hypervisors
+	 * Use RDTSCP if possible; it is guaranteed to be synchronous
+	 * and doesn't cause a VMEXIT on Hypervisors
 	 */
 	alternative_io(ASM_NOP3, ".byte 0x0f,0x01,0xf9", X86_FEATURE_RDTSCP,
 		       ASM_OUTPUT2("=a" (eax), "=d" (edx)),
@@ -55,11 +55,40 @@ static __always_inline cycles_t get_cycles_sync(void)
 	 */
 	alternative_io("cpuid", ASM_NOP2, X86_FEATURE_SYNC_RDTSC,
 			  "=a" (eax), "0" (1) : "ebx","ecx","edx","memory");
-	rdtscll(ret);
 
+	return 0;
+}
+
+static __always_inline cycles_t get_cycles_sync(void)
+{
+	unsigned long long ret;
+	ret = __get_cycles_sync();
+	if (!ret)
+		rdtscll(ret);
 	return ret;
 }
 
+#ifdef CONFIG_PARAVIRT
+/*
+ * For paravirt guests, some functionalities are executed through function
+ * pointers in the various pvops structures.
+ * These function pointers exist inside the kernel and can not
+ * be accessed by user space. To avoid this, we make a copy of the
+ * get_cycles_sync (called in kernel) but force the use of native_read_tsc.
+ * Ideally, the guest should set up it's own clock and vread
+ */
+static __always_inline long long vget_cycles_sync(void)
+{
+	unsigned long long ret;
+	ret = __get_cycles_sync();
+	if (!ret)
+		ret = native_read_tsc();
+	return ret;
+}
+#else
+# define vget_cycles_sync() get_cycles_sync()
+#endif
+
 extern void tsc_init(void);
 extern void mark_tsc_unstable(char *reason);
 extern int unsynchronized_tsc(void);
-- 
1.4.4.2

[PATCH 4/10] split get_cycles_sync

[Glauber de Oliveira Costa]

This patch splits get_cycles_sync() into  __get_cycles_sync(),
and the rdtscll part. Paravirt guests cannot issue rdtscl directly,
as it involves a function call in vdso area.

So, using the __get_cycles_sync() base, we introduce vget_cycles_sync,
which then calls the native version of rdtscll. Ideally, however, a guest
should define its own clocksource, together with a vread function

Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
---
 arch/x86/kernel/tsc_64.c |    2 +-
 include/asm-x86/tsc.h    |   37 +++++++++++++++++++++++++++++++++----
 2 files changed, 34 insertions(+), 5 deletions(-)

diff --git a/arch/x86/kernel/tsc_64.c b/arch/x86/kernel/tsc_64.c
index 5158476..4502539 100644
--- a/arch/x86/kernel/tsc_64.c
+++ b/arch/x86/kernel/tsc_64.c
@@ -269,7 +269,7 @@ static cycle_t read_tsc(void)
 
 static cycle_t __vsyscall_fn vread_tsc(void)
 {
-	cycle_t ret = (cycle_t)get_cycles_sync();
+	cycle_t ret = (cycle_t)vget_cycles_sync();
 	return ret;
 }
 
diff --git a/include/asm-x86/tsc.h b/include/asm-x86/tsc.h
index 6baab30..d6cb056 100644
--- a/include/asm-x86/tsc.h
+++ b/include/asm-x86/tsc.h
@@ -33,14 +33,14 @@ static inline cycles_t get_cycles(void)
 }
 
 /* Like get_cycles, but make sure the CPU is synchronized. */
-static __always_inline cycles_t get_cycles_sync(void)
+static __always_inline cycles_t __get_cycles_sync(void)
 {
 	unsigned long long ret;
 	unsigned eax, edx;
 
 	/*
-  	 * Use RDTSCP if possible; it is guaranteed to be synchronous
- 	 * and doesn't cause a VMEXIT on Hypervisors
+	 * Use RDTSCP if possible; it is guaranteed to be synchronous
+	 * and doesn't cause a VMEXIT on Hypervisors
 	 */
 	alternative_io(ASM_NOP3, ".byte 0x0f,0x01,0xf9", X86_FEATURE_RDTSCP,
 		       ASM_OUTPUT2("=a" (eax), "=d" (edx)),
@@ -55,11 +55,40 @@ static __always_inline cycles_t get_cycles_sync(void)
 	 */
 	alternative_io("cpuid", ASM_NOP2, X86_FEATURE_SYNC_RDTSC,
 			  "=a" (eax), "0" (1) : "ebx","ecx","edx","memory");
-	rdtscll(ret);
 
+	return 0;
+}
+
+static __always_inline cycles_t get_cycles_sync(void)
+{
+	unsigned long long ret;
+	ret = __get_cycles_sync();
+	if (!ret)
+		rdtscll(ret);
 	return ret;
 }
 
+#ifdef CONFIG_PARAVIRT
+/*
+ * For paravirt guests, some functionalities are executed through function
+ * pointers in the various pvops structures.
+ * These function pointers exist inside the kernel and can not
+ * be accessed by user space. To avoid this, we make a copy of the
+ * get_cycles_sync (called in kernel) but force the use of native_read_tsc.
+ * Ideally, the guest should set up it's own clock and vread
+ */
+static __always_inline long long vget_cycles_sync(void)
+{
+	unsigned long long ret;
+	ret = __get_cycles_sync();
+	if (!ret)
+		ret = native_read_tsc();
+	return ret;
+}
+#else
+# define vget_cycles_sync() get_cycles_sync()
+#endif
+
 extern void tsc_init(void);
 extern void mark_tsc_unstable(char *reason);
 extern int unsynchronized_tsc(void);
-- 
1.4.4.2

[PATCH 5/10] unify cpuid functions

[Glauber de Oliveira Costa]

cpuid is not very different between i386 and x86_64.
We move away the x86_64 version from msr.h, and
unify them at processor.h, where they belong.

cpuid() paravirt then comes for free.

Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
---
 include/asm-x86/msr.h          |   67 ----------------------------------
 include/asm-x86/processor.h    |   78 ++++++++++++++++++++++++++++++++++++++++
 include/asm-x86/processor_32.h |   69 -----------------------------------
 3 files changed, 78 insertions(+), 136 deletions(-)

diff --git a/include/asm-x86/msr.h b/include/asm-x86/msr.h
index b3cd509..5fa30bd 100644
--- a/include/asm-x86/msr.h
+++ b/include/asm-x86/msr.h
@@ -230,73 +230,6 @@ static inline int wrmsr_safe(u32 __msr, u32 __low, u32 __high)
 			  : "=a" (low), "=d" (high) \
 			  : "c" (counter))
 
-static inline void cpuid(int op, unsigned int *eax, unsigned int *ebx,
-			 unsigned int *ecx, unsigned int *edx)
-{
-	__asm__("cpuid"
-		: "=a" (*eax),
-		  "=b" (*ebx),
-		  "=c" (*ecx),
-		  "=d" (*edx)
-		: "0" (op));
-}
-
-/* Some CPUID calls want 'count' to be placed in ecx */
-static inline void cpuid_count(int op, int count, int *eax, int *ebx, int *ecx,
-			       int *edx)
-{
-	__asm__("cpuid"
-		: "=a" (*eax),
-		  "=b" (*ebx),
-		  "=c" (*ecx),
-		  "=d" (*edx)
-		: "0" (op), "c" (count));
-}
-
-/*
- * CPUID functions returning a single datum
- */
-static inline unsigned int cpuid_eax(unsigned int op)
-{
-	unsigned int eax;
-
-	__asm__("cpuid"
-		: "=a" (eax)
-		: "0" (op)
-		: "bx", "cx", "dx");
-	return eax;
-}
-static inline unsigned int cpuid_ebx(unsigned int op)
-{
-	unsigned int eax, ebx;
-
-	__asm__("cpuid"
-		: "=a" (eax), "=b" (ebx)
-		: "0" (op)
-		: "cx", "dx" );
-	return ebx;
-}
-static inline unsigned int cpuid_ecx(unsigned int op)
-{
-	unsigned int eax, ecx;
-
-	__asm__("cpuid"
-		: "=a" (eax), "=c" (ecx)
-		: "0" (op)
-		: "bx", "dx" );
-	return ecx;
-}
-static inline unsigned int cpuid_edx(unsigned int op)
-{
-	unsigned int eax, edx;
-
-	__asm__("cpuid"
-		: "=a" (eax), "=d" (edx)
-		: "0" (op)
-		: "bx", "cx");
-	return edx;
-}
-
 #endif  /* __ASSEMBLY__ */
 
 #endif  /* !__i386__ */
diff --git a/include/asm-x86/processor.h b/include/asm-x86/processor.h
index 46e1c04..dea81b7 100644
--- a/include/asm-x86/processor.h
+++ b/include/asm-x86/processor.h
@@ -1,5 +1,83 @@
+#ifndef __ASM_X86_PROCESSOR_H
+#define __ASM_X86_PROCESSOR_H
+
+static inline void native_cpuid(unsigned int *eax, unsigned int *ebx,
+					 unsigned int *ecx, unsigned int *edx)
+{
+	/* ecx is often an input as well as an output. */
+	__asm__("cpuid"
+		: "=a" (*eax),
+		  "=b" (*ebx),
+		  "=c" (*ecx),
+		  "=d" (*edx)
+		: "0" (*eax), "2" (*ecx));
+}
+
+
 #ifdef CONFIG_X86_32
 # include "processor_32.h"
 #else
 # include "processor_64.h"
 #endif
+
+#ifndef CONFIG_PARAVIRT
+#define __cpuid native_cpuid
+#endif
+
+/*
+ * Generic CPUID function
+ * clear %ecx since some cpus (Cyrix MII) do not set or clear %ecx
+ * resulting in stale register contents being returned.
+ */
+static inline void cpuid(unsigned int op,
+			 unsigned int *eax, unsigned int *ebx,
+			 unsigned int *ecx, unsigned int *edx)
+{
+	*eax = op;
+	*ecx = 0;
+	__cpuid(eax, ebx, ecx, edx);
+}
+
+/* Some CPUID calls want 'count' to be placed in ecx */
+static inline void cpuid_count(unsigned int op, int count,
+			       unsigned int *eax, unsigned int *ebx,
+			       unsigned int *ecx, unsigned int *edx)
+{
+	*eax = op;
+	*ecx = count;
+	__cpuid(eax, ebx, ecx, edx);
+}
+
+/*
+ * CPUID functions returning a single datum
+ */
+static inline unsigned int cpuid_eax(unsigned int op)
+{
+	unsigned int eax, ebx, ecx, edx;
+
+	cpuid(op, &eax, &ebx, &ecx, &edx);
+	return eax;
+}
+static inline unsigned int cpuid_ebx(unsigned int op)
+{
+	unsigned int eax, ebx, ecx, edx;
+
+	cpuid(op, &eax, &ebx, &ecx, &edx);
+	return ebx;
+}
+static inline unsigned int cpuid_ecx(unsigned int op)
+{
+	unsigned int eax, ebx, ecx, edx;
+
+	cpuid(op, &eax, &ebx, &ecx, &edx);
+	return ecx;
+}
+static inline unsigned int cpuid_edx(unsigned int op)
+{
+	unsigned int eax, ebx, ecx, edx;
+
+	cpuid(op, &eax, &ebx, &ecx, &edx);
+	return edx;
+}
+
+#endif
diff --git a/include/asm-x86/processor_32.h b/include/asm-x86/processor_32.h
index 13976b0..12a4f83 100644
--- a/include/asm-x86/processor_32.h
+++ b/include/asm-x86/processor_32.h
@@ -134,18 +134,6 @@ extern void detect_ht(struct cpuinfo_x86 *c);
 static inline void detect_ht(struct cpuinfo_x86 *c) {}
 #endif
 
-static inline void native_cpuid(unsigned int *eax, unsigned int *ebx,
-					 unsigned int *ecx, unsigned int *edx)
-{
-	/* ecx is often an input as well as an output. */
-	__asm__("cpuid"
-		: "=a" (*eax),
-		  "=b" (*ebx),
-		  "=c" (*ecx),
-		  "=d" (*edx)
-		: "0" (*eax), "2" (*ecx));
-}
-
 #define load_cr3(pgdir) write_cr3(__pa(pgdir))
 
 /*
@@ -578,7 +566,6 @@ static inline void native_set_iopl_mask(unsigned mask)
 #include <asm/paravirt.h>
 #else
 #define paravirt_enabled() 0
-#define __cpuid native_cpuid
 
 static inline void load_esp0(struct tss_struct *tss, struct thread_struct *thread)
 {
@@ -596,62 +583,6 @@ static inline void load_esp0(struct tss_struct *tss, struct thread_struct *threa
 #define set_iopl_mask native_set_iopl_mask
 #endif /* CONFIG_PARAVIRT */
 
-/*
- * Generic CPUID function
- * clear %ecx since some cpus (Cyrix MII) do not set or clear %ecx
- * resulting in stale register contents being returned.
- */
-static inline void cpuid(unsigned int op,
-			 unsigned int *eax, unsigned int *ebx,
-			 unsigned int *ecx, unsigned int *edx)
-{
-	*eax = op;
-	*ecx = 0;
-	__cpuid(eax, ebx, ecx, edx);
-}
-
-/* Some CPUID calls want 'count' to be placed in ecx */
-static inline void cpuid_count(unsigned int op, int count,
-			       unsigned int *eax, unsigned int *ebx,
-			       unsigned int *ecx, unsigned int *edx)
-{
-	*eax = op;
-	*ecx = count;
-	__cpuid(eax, ebx, ecx, edx);
-}
-
-/*
- * CPUID functions returning a single datum
- */
-static inline unsigned int cpuid_eax(unsigned int op)
-{
-	unsigned int eax, ebx, ecx, edx;
-
-	cpuid(op, &eax, &ebx, &ecx, &edx);
-	return eax;
-}
-static inline unsigned int cpuid_ebx(unsigned int op)
-{
-	unsigned int eax, ebx, ecx, edx;
-
-	cpuid(op, &eax, &ebx, &ecx, &edx);
-	return ebx;
-}
-static inline unsigned int cpuid_ecx(unsigned int op)
-{
-	unsigned int eax, ebx, ecx, edx;
-
-	cpuid(op, &eax, &ebx, &ecx, &edx);
-	return ecx;
-}
-static inline unsigned int cpuid_edx(unsigned int op)
-{
-	unsigned int eax, ebx, ecx, edx;
-
-	cpuid(op, &eax, &ebx, &ecx, &edx);
-	return edx;
-}
-
 /* generic versions from gas */
 #define GENERIC_NOP1	".byte 0x90\n"
 #define GENERIC_NOP2    	".byte 0x89,0xf6\n"
-- 
1.4.4.2

[PATCH 5/10] unify cpuid functions

[Glauber de Oliveira Costa]

cpuid is not very different between i386 and x86_64.
We move away the x86_64 version from msr.h, and
unify them at processor.h, where they belong.

cpuid() paravirt then comes for free.

Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
---
 include/asm-x86/msr.h          |   67 ----------------------------------
 include/asm-x86/processor.h    |   78 ++++++++++++++++++++++++++++++++++++++++
 include/asm-x86/processor_32.h |   69 -----------------------------------
 3 files changed, 78 insertions(+), 136 deletions(-)

diff --git a/include/asm-x86/msr.h b/include/asm-x86/msr.h
index b3cd509..5fa30bd 100644
--- a/include/asm-x86/msr.h
+++ b/include/asm-x86/msr.h
@@ -230,73 +230,6 @@ static inline int wrmsr_safe(u32 __msr, u32 __low, u32 __high)
 			  : "=a" (low), "=d" (high) \
 			  : "c" (counter))
 
-static inline void cpuid(int op, unsigned int *eax, unsigned int *ebx,
-			 unsigned int *ecx, unsigned int *edx)
-{
-	__asm__("cpuid"
-		: "=a" (*eax),
-		  "=b" (*ebx),
-		  "=c" (*ecx),
-		  "=d" (*edx)
-		: "0" (op));
-}
-
-/* Some CPUID calls want 'count' to be placed in ecx */
-static inline void cpuid_count(int op, int count, int *eax, int *ebx, int *ecx,
-			       int *edx)
-{
-	__asm__("cpuid"
-		: "=a" (*eax),
-		  "=b" (*ebx),
-		  "=c" (*ecx),
-		  "=d" (*edx)
-		: "0" (op), "c" (count));
-}
-
-/*
- * CPUID functions returning a single datum
- */
-static inline unsigned int cpuid_eax(unsigned int op)
-{
-	unsigned int eax;
-
-	__asm__("cpuid"
-		: "=a" (eax)
-		: "0" (op)
-		: "bx", "cx", "dx");
-	return eax;
-}
-static inline unsigned int cpuid_ebx(unsigned int op)
-{
-	unsigned int eax, ebx;
-
-	__asm__("cpuid"
-		: "=a" (eax), "=b" (ebx)
-		: "0" (op)
-		: "cx", "dx" );
-	return ebx;
-}
-static inline unsigned int cpuid_ecx(unsigned int op)
-{
-	unsigned int eax, ecx;
-
-	__asm__("cpuid"
-		: "=a" (eax), "=c" (ecx)
-		: "0" (op)
-		: "bx", "dx" );
-	return ecx;
-}
-static inline unsigned int cpuid_edx(unsigned int op)
-{
-	unsigned int eax, edx;
-
-	__asm__("cpuid"
-		: "=a" (eax), "=d" (edx)
-		: "0" (op)
-		: "bx", "cx");
-	return edx;
-}
-
 #endif  /* __ASSEMBLY__ */
 
 #endif  /* !__i386__ */
diff --git a/include/asm-x86/processor.h b/include/asm-x86/processor.h
index 46e1c04..dea81b7 100644
--- a/include/asm-x86/processor.h
+++ b/include/asm-x86/processor.h
@@ -1,5 +1,83 @@
+#ifndef __ASM_X86_PROCESSOR_H
+#define __ASM_X86_PROCESSOR_H
+
+static inline void native_cpuid(unsigned int *eax, unsigned int *ebx,
+					 unsigned int *ecx, unsigned int *edx)
+{
+	/* ecx is often an input as well as an output. */
+	__asm__("cpuid"
+		: "=a" (*eax),
+		  "=b" (*ebx),
+		  "=c" (*ecx),
+		  "=d" (*edx)
+		: "0" (*eax), "2" (*ecx));
+}
+
+
 #ifdef CONFIG_X86_32
 # include "processor_32.h"
 #else
 # include "processor_64.h"
 #endif
+
+#ifndef CONFIG_PARAVIRT
+#define __cpuid native_cpuid
+#endif
+
+/*
+ * Generic CPUID function
+ * clear %ecx since some cpus (Cyrix MII) do not set or clear %ecx
+ * resulting in stale register contents being returned.
+ */
+static inline void cpuid(unsigned int op,
+			 unsigned int *eax, unsigned int *ebx,
+			 unsigned int *ecx, unsigned int *edx)
+{
+	*eax = op;
+	*ecx = 0;
+	__cpuid(eax, ebx, ecx, edx);
+}
+
+/* Some CPUID calls want 'count' to be placed in ecx */
+static inline void cpuid_count(unsigned int op, int count,
+			       unsigned int *eax, unsigned int *ebx,
+			       unsigned int *ecx, unsigned int *edx)
+{
+	*eax = op;
+	*ecx = count;
+	__cpuid(eax, ebx, ecx, edx);
+}
+
+/*
+ * CPUID functions returning a single datum
+ */
+static inline unsigned int cpuid_eax(unsigned int op)
+{
+	unsigned int eax, ebx, ecx, edx;
+
+	cpuid(op, &eax, &ebx, &ecx, &edx);
+	return eax;
+}
+static inline unsigned int cpuid_ebx(unsigned int op)
+{
+	unsigned int eax, ebx, ecx, edx;
+
+	cpuid(op, &eax, &ebx, &ecx, &edx);
+	return ebx;
+}
+static inline unsigned int cpuid_ecx(unsigned int op)
+{
+	unsigned int eax, ebx, ecx, edx;
+
+	cpuid(op, &eax, &ebx, &ecx, &edx);
+	return ecx;
+}
+static inline unsigned int cpuid_edx(unsigned int op)
+{
+	unsigned int eax, ebx, ecx, edx;
+
+	cpuid(op, &eax, &ebx, &ecx, &edx);
+	return edx;
+}
+
+#endif
diff --git a/include/asm-x86/processor_32.h b/include/asm-x86/processor_32.h
index 13976b0..12a4f83 100644
--- a/include/asm-x86/processor_32.h
+++ b/include/asm-x86/processor_32.h
@@ -134,18 +134,6 @@ extern void detect_ht(struct cpuinfo_x86 *c);
 static inline void detect_ht(struct cpuinfo_x86 *c) {}
 #endif
 
-static inline void native_cpuid(unsigned int *eax, unsigned int *ebx,
-					 unsigned int *ecx, unsigned int *edx)
-{
-	/* ecx is often an input as well as an output. */
-	__asm__("cpuid"
-		: "=a" (*eax),
-		  "=b" (*ebx),
-		  "=c" (*ecx),
-		  "=d" (*edx)
-		: "0" (*eax), "2" (*ecx));
-}
-
 #define load_cr3(pgdir) write_cr3(__pa(pgdir))
 
 /*
@@ -578,7 +566,6 @@ static inline void native_set_iopl_mask(unsigned mask)
 #include <asm/paravirt.h>
 #else
 #define paravirt_enabled() 0
-#define __cpuid native_cpuid
 
 static inline void load_esp0(struct tss_struct *tss, struct thread_struct *thread)
 {
@@ -596,62 +583,6 @@ static inline void load_esp0(struct tss_struct *tss, struct thread_struct *threa
 #define set_iopl_mask native_set_iopl_mask
 #endif /* CONFIG_PARAVIRT */
 
-/*
- * Generic CPUID function
- * clear %ecx since some cpus (Cyrix MII) do not set or clear %ecx
- * resulting in stale register contents being returned.
- */
-static inline void cpuid(unsigned int op,
-			 unsigned int *eax, unsigned int *ebx,
-			 unsigned int *ecx, unsigned int *edx)
-{
-	*eax = op;
-	*ecx = 0;
-	__cpuid(eax, ebx, ecx, edx);
-}
-
-/* Some CPUID calls want 'count' to be placed in ecx */
-static inline void cpuid_count(unsigned int op, int count,
-			       unsigned int *eax, unsigned int *ebx,
-			       unsigned int *ecx, unsigned int *edx)
-{
-	*eax = op;
-	*ecx = count;
-	__cpuid(eax, ebx, ecx, edx);
-}
-
-/*
- * CPUID functions returning a single datum
- */
-static inline unsigned int cpuid_eax(unsigned int op)
-{
-	unsigned int eax, ebx, ecx, edx;
-
-	cpuid(op, &eax, &ebx, &ecx, &edx);
-	return eax;
-}
-static inline unsigned int cpuid_ebx(unsigned int op)
-{
-	unsigned int eax, ebx, ecx, edx;
-
-	cpuid(op, &eax, &ebx, &ecx, &edx);
-	return ebx;
-}
-static inline unsigned int cpuid_ecx(unsigned int op)
-{
-	unsigned int eax, ebx, ecx, edx;
-
-	cpuid(op, &eax, &ebx, &ecx, &edx);
-	return ecx;
-}
-static inline unsigned int cpuid_edx(unsigned int op)
-{
-	unsigned int eax, ebx, ecx, edx;
-
-	cpuid(op, &eax, &ebx, &ecx, &edx);
-	return edx;
-}
-
 /* generic versions from gas */
 #define GENERIC_NOP1	".byte 0x90\n"
 #define GENERIC_NOP2    	".byte 0x89,0xf6\n"
-- 
1.4.4.2

[PATCH 6/10] introduce native_read_tscp

[Glauber de Oliveira Costa]

Targetting paravirt, this patch introduces native_read_tscp, in
place of rdtscp() macro. When in a paravirt guest, this will
involve a function call, and thus, cannot be done in the vdso area.
These users then have to call the native version directly

Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
---
 arch/x86/kernel/vsyscall_64.c |    4 ++--
 arch/x86/vdso/vgetcpu.c       |    4 ++--
 include/asm-x86/msr.h         |   29 +++++++++++++++++++++--------
 3 files changed, 25 insertions(+), 12 deletions(-)

diff --git a/arch/x86/kernel/vsyscall_64.c b/arch/x86/kernel/vsyscall_64.c
index ad4005c..1425d02 100644
--- a/arch/x86/kernel/vsyscall_64.c
+++ b/arch/x86/kernel/vsyscall_64.c
@@ -190,7 +190,7 @@ time_t __vsyscall(1) vtime(time_t *t)
 long __vsyscall(2)
 vgetcpu(unsigned *cpu, unsigned *node, struct getcpu_cache *tcache)
 {
-	unsigned int dummy, p;
+	unsigned int p;
 	unsigned long j = 0;
 
 	/* Fast cache - only recompute value once per jiffies and avoid
@@ -205,7 +205,7 @@ vgetcpu(unsigned *cpu, unsigned *node, struct getcpu_cache *tcache)
 		p = tcache->blob[1];
 	} else if (__vgetcpu_mode == VGETCPU_RDTSCP) {
 		/* Load per CPU data from RDTSCP */
-		rdtscp(dummy, dummy, p);
+		native_read_tscp(&p);
 	} else {
 		/* Load per CPU data from GDT */
 		asm("lsl %1,%0" : "=r" (p) : "r" (__PER_CPU_SEG));
diff --git a/arch/x86/vdso/vgetcpu.c b/arch/x86/vdso/vgetcpu.c
index 3b1ae1a..c8097f1 100644
--- a/arch/x86/vdso/vgetcpu.c
+++ b/arch/x86/vdso/vgetcpu.c
@@ -15,11 +15,11 @@
 
 long __vdso_getcpu(unsigned *cpu, unsigned *node, struct getcpu_cache *unused)
 {
-	unsigned int dummy, p;
+	unsigned int p;
 
 	if (*vdso_vgetcpu_mode == VGETCPU_RDTSCP) {
 		/* Load per CPU data from RDTSCP */
-		rdtscp(dummy, dummy, p);
+		native_read_tscp(&p);
 	} else {
 		/* Load per CPU data from GDT */
 		asm("lsl %1,%0" : "=r" (p) : "r" (__PER_CPU_SEG));
diff --git a/include/asm-x86/msr.h b/include/asm-x86/msr.h
index 5fa30bd..0296358 100644
--- a/include/asm-x86/msr.h
+++ b/include/asm-x86/msr.h
@@ -3,6 +3,27 @@
 
 #include <asm/msr-index.h>
 
+#ifdef __KERNEL__
+#ifndef __ASSEMBLY__
+static inline unsigned long long native_read_tscp(int *aux)
+{
+	unsigned long low, high;
+	asm volatile (".byte 0x0f,0x01,0xf9"
+		      : "=a" (low), "=d" (high), "=c" (*aux));
+	return low | ((u64)high >> 32);
+}
+
+#define rdtscp(low, high, aux)						\
+       do {                                                            \
+               unsigned long long _val = native_read_tscp(&(aux));     \
+               (low) = (u32)_val;                                      \
+               (high) = (u32)(_val >> 32);                             \
+       } while (0)
+
+#define rdtscpll(val, aux) (val) = native_read_tscp(&(aux))
+#endif
+#endif
+
 #ifdef __i386__
 
 #ifdef __KERNEL__
@@ -206,8 +227,6 @@ static inline int wrmsr_safe(u32 __msr, u32 __low, u32 __high)
 #define rdtscl(low) \
      __asm__ __volatile__ ("rdtsc" : "=a" (low) : : "edx")
 
-#define rdtscp(low,high,aux) \
-     asm volatile (".byte 0x0f,0x01,0xf9" : "=a" (low), "=d" (high), "=c" (aux))
 
 #define rdtscll(val) do { \
      unsigned int __a,__d; \
@@ -215,12 +234,6 @@ static inline int wrmsr_safe(u32 __msr, u32 __low, u32 __high)
      (val) = ((unsigned long)__a) | (((unsigned long)__d)<<32); \
 } while(0)
 
-#define rdtscpll(val, aux) do { \
-     unsigned long __a, __d; \
-     asm volatile (".byte 0x0f,0x01,0xf9" : "=a" (__a), "=d" (__d), "=c" (aux)); \
-     (val) = (__d << 32) | __a; \
-} while (0)
-
 #define write_tsc(val1,val2) wrmsr(0x10, val1, val2)
 
 #define write_rdtscp_aux(val) wrmsr(0xc0000103, val, 0)
-- 
1.4.4.2

[PATCH 6/10] introduce native_read_tscp

[Glauber de Oliveira Costa]

Targetting paravirt, this patch introduces native_read_tscp, in
place of rdtscp() macro. When in a paravirt guest, this will
involve a function call, and thus, cannot be done in the vdso area.
These users then have to call the native version directly

Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
---
 arch/x86/kernel/vsyscall_64.c |    4 ++--
 arch/x86/vdso/vgetcpu.c       |    4 ++--
 include/asm-x86/msr.h         |   29 +++++++++++++++++++++--------
 3 files changed, 25 insertions(+), 12 deletions(-)

diff --git a/arch/x86/kernel/vsyscall_64.c b/arch/x86/kernel/vsyscall_64.c
index ad4005c..1425d02 100644
--- a/arch/x86/kernel/vsyscall_64.c
+++ b/arch/x86/kernel/vsyscall_64.c
@@ -190,7 +190,7 @@ time_t __vsyscall(1) vtime(time_t *t)
 long __vsyscall(2)
 vgetcpu(unsigned *cpu, unsigned *node, struct getcpu_cache *tcache)
 {
-	unsigned int dummy, p;
+	unsigned int p;
 	unsigned long j = 0;
 
 	/* Fast cache - only recompute value once per jiffies and avoid
@@ -205,7 +205,7 @@ vgetcpu(unsigned *cpu, unsigned *node, struct getcpu_cache *tcache)
 		p = tcache->blob[1];
 	} else if (__vgetcpu_mode == VGETCPU_RDTSCP) {
 		/* Load per CPU data from RDTSCP */
-		rdtscp(dummy, dummy, p);
+		native_read_tscp(&p);
 	} else {
 		/* Load per CPU data from GDT */
 		asm("lsl %1,%0" : "=r" (p) : "r" (__PER_CPU_SEG));
diff --git a/arch/x86/vdso/vgetcpu.c b/arch/x86/vdso/vgetcpu.c
index 3b1ae1a..c8097f1 100644
--- a/arch/x86/vdso/vgetcpu.c
+++ b/arch/x86/vdso/vgetcpu.c
@@ -15,11 +15,11 @@
 
 long __vdso_getcpu(unsigned *cpu, unsigned *node, struct getcpu_cache *unused)
 {
-	unsigned int dummy, p;
+	unsigned int p;
 
 	if (*vdso_vgetcpu_mode == VGETCPU_RDTSCP) {
 		/* Load per CPU data from RDTSCP */
-		rdtscp(dummy, dummy, p);
+		native_read_tscp(&p);
 	} else {
 		/* Load per CPU data from GDT */
 		asm("lsl %1,%0" : "=r" (p) : "r" (__PER_CPU_SEG));
diff --git a/include/asm-x86/msr.h b/include/asm-x86/msr.h
index 5fa30bd..0296358 100644
--- a/include/asm-x86/msr.h
+++ b/include/asm-x86/msr.h
@@ -3,6 +3,27 @@
 
 #include <asm/msr-index.h>
 
+#ifdef __KERNEL__
+#ifndef __ASSEMBLY__
+static inline unsigned long long native_read_tscp(int *aux)
+{
+	unsigned long low, high;
+	asm volatile (".byte 0x0f,0x01,0xf9"
+		      : "=a" (low), "=d" (high), "=c" (*aux));
+	return low | ((u64)high >> 32);
+}
+
+#define rdtscp(low, high, aux)						\
+       do {                                                            \
+               unsigned long long _val = native_read_tscp(&(aux));     \
+               (low) = (u32)_val;                                      \
+               (high) = (u32)(_val >> 32);                             \
+       } while (0)
+
+#define rdtscpll(val, aux) (val) = native_read_tscp(&(aux))
+#endif
+#endif
+
 #ifdef __i386__
 
 #ifdef __KERNEL__
@@ -206,8 +227,6 @@ static inline int wrmsr_safe(u32 __msr, u32 __low, u32 __high)
 #define rdtscl(low) \
      __asm__ __volatile__ ("rdtsc" : "=a" (low) : : "edx")
 
-#define rdtscp(low,high,aux) \
-     asm volatile (".byte 0x0f,0x01,0xf9" : "=a" (low), "=d" (high), "=c" (aux))
 
 #define rdtscll(val) do { \
      unsigned int __a,__d; \
@@ -215,12 +234,6 @@ static inline int wrmsr_safe(u32 __msr, u32 __low, u32 __high)
      (val) = ((unsigned long)__a) | (((unsigned long)__d)<<32); \
 } while(0)
 
-#define rdtscpll(val, aux) do { \
-     unsigned long __a, __d; \
-     asm volatile (".byte 0x0f,0x01,0xf9" : "=a" (__a), "=d" (__d), "=c" (aux)); \
-     (val) = (__d << 32) | __a; \
-} while (0)
-
 #define write_tsc(val1,val2) wrmsr(0x10, val1, val2)
 
 #define write_rdtscp_aux(val) wrmsr(0xc0000103, val, 0)
-- 
1.4.4.2

[PATCH 7/10] change rdpmc interface

[Glauber de Oliveira Costa]

the rdpmc instruction gets a counter argument in rcx. However,
the i386 version was ignoring it. To make both x86_64 and i386 versions
the same, as well as to comply with the instruction semantics, this
parameter is added in the i386 version

Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
---
 include/asm-x86/msr.h      |    6 +++---
 include/asm-x86/paravirt.h |    2 +-
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/include/asm-x86/msr.h b/include/asm-x86/msr.h
index 0296358..ce2a257 100644
--- a/include/asm-x86/msr.h
+++ b/include/asm-x86/msr.h
@@ -90,10 +90,10 @@ static inline unsigned long long native_read_tsc(void)
 	return val;
 }
 
-static inline unsigned long long native_read_pmc(void)
+static inline unsigned long long native_read_pmc(int counter)
 {
 	unsigned long long val;
-	asm volatile("rdpmc" : "=A" (val));
+	asm volatile("rdpmc" : "=A" (val) : "c" (counter));
 	return val;
 }
 
@@ -150,7 +150,7 @@ static inline int wrmsr_safe(u32 __msr, u32 __low, u32 __high)
 
 #define rdpmc(counter,low,high)					\
 	do {							\
-		u64 _l = native_read_pmc();			\
+		u64 _l = native_read_pmc(counter);		\
 		(low)  = (u32)_l;				\
 		(high) = (u32)(_l >> 32);			\
 	} while(0)
diff --git a/include/asm-x86/paravirt.h b/include/asm-x86/paravirt.h
index f59d370..4a7be97 100644
--- a/include/asm-x86/paravirt.h
+++ b/include/asm-x86/paravirt.h
@@ -118,7 +118,7 @@ struct pv_cpu_ops {
 	int (*write_msr)(unsigned int msr, u64 val);
 
 	u64 (*read_tsc)(void);
-	u64 (*read_pmc)(void);
+	u64 (*read_pmc)(int counter);
 
 	/* These two are jmp to, not actually called. */
 	void (*irq_enable_sysexit)(void);
-- 
1.4.4.2

[PATCH 7/10] change rdpmc interface

[Glauber de Oliveira Costa]

the rdpmc instruction gets a counter argument in rcx. However,
the i386 version was ignoring it. To make both x86_64 and i386 versions
the same, as well as to comply with the instruction semantics, this
parameter is added in the i386 version

Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
---
 include/asm-x86/msr.h      |    6 +++---
 include/asm-x86/paravirt.h |    2 +-
 2 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/include/asm-x86/msr.h b/include/asm-x86/msr.h
index 0296358..ce2a257 100644
--- a/include/asm-x86/msr.h
+++ b/include/asm-x86/msr.h
@@ -90,10 +90,10 @@ static inline unsigned long long native_read_tsc(void)
 	return val;
 }
 
-static inline unsigned long long native_read_pmc(void)
+static inline unsigned long long native_read_pmc(int counter)
 {
 	unsigned long long val;
-	asm volatile("rdpmc" : "=A" (val));
+	asm volatile("rdpmc" : "=A" (val) : "c" (counter));
 	return val;
 }
 
@@ -150,7 +150,7 @@ static inline int wrmsr_safe(u32 __msr, u32 __low, u32 __high)
 
 #define rdpmc(counter,low,high)					\
 	do {							\
-		u64 _l = native_read_pmc();			\
+		u64 _l = native_read_pmc(counter);		\
 		(low)  = (u32)_l;				\
 		(high) = (u32)(_l >> 32);			\
 	} while(0)
diff --git a/include/asm-x86/paravirt.h b/include/asm-x86/paravirt.h
index f59d370..4a7be97 100644
--- a/include/asm-x86/paravirt.h
+++ b/include/asm-x86/paravirt.h
@@ -118,7 +118,7 @@ struct pv_cpu_ops {
 	int (*write_msr)(unsigned int msr, u64 val);
 
 	u64 (*read_tsc)(void);
-	u64 (*read_pmc)(void);
+	u64 (*read_pmc)(int counter);
 
 	/* These two are jmp to, not actually called. */
 	void (*irq_enable_sysexit)(void);
-- 
1.4.4.2

[PATCH 8/10] change write msr functions interface

[Glauber de Oliveira Costa]

This patche changes the native_write_msr() and friends interface
to explicitly take 2 32-bit registers instead of a 64-bit value.
The change will ease the merge with 64-bit code. As the 64-bit
value will be passed as two registers anyway in i386,
the PVOP_CALL interface has to account for that and use low/high parameters
It would force the x86_64 version to be different.

The change does not make i386 generated code less efficient. As said above,
it would get the values from two registers anyway.

Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
---
 include/asm-x86/msr.h      |   19 ++++++++++---------
 include/asm-x86/paravirt.h |    2 +-
 2 files changed, 11 insertions(+), 10 deletions(-)

diff --git a/include/asm-x86/msr.h b/include/asm-x86/msr.h
index ce2a257..9c695c7 100644
--- a/include/asm-x86/msr.h
+++ b/include/asm-x86/msr.h
@@ -59,13 +59,14 @@ static inline unsigned long long native_read_msr_safe(unsigned int msr,
 	return val;
 }
 
-static inline void native_write_msr(unsigned int msr, unsigned long long val)
+static inline void native_write_msr(unsigned int msr,
+				    unsigned low, unsigned high)
 {
-	asm volatile("wrmsr" : : "c" (msr), "A"(val));
+	asm volatile("wrmsr" : : "c" (msr), "a"(low), "d" (high));
 }
 
 static inline int native_write_msr_safe(unsigned int msr,
-					unsigned long long val)
+					unsigned low, unsigned high)
 {
 	int err;
 	asm volatile("2: wrmsr ; xorl %0,%0\n"
@@ -78,7 +79,7 @@ static inline int native_write_msr_safe(unsigned int msr,
 		     "   .long	2b,3b\n\t"
 		     ".previous"
 		     : "=a" (err)
-		     : "c" (msr), "0" ((u32)val), "d" ((u32)(val>>32)),
+		     : "c" (msr), "0" (low), "d" (high),
 		       "i" (-EFAULT));
 	return err;
 }
@@ -114,20 +115,20 @@ static inline unsigned long long native_read_pmc(int counter)
 		(val2) = (u32)(__val >> 32);				\
 	} while(0)
 
-static inline void wrmsr(u32 __msr, u32 __low, u32 __high)
+static inline void wrmsr(unsigned msr, unsigned low, unsigned high)
 {
-	native_write_msr(__msr, ((u64)__high << 32) | __low);
+	native_write_msr(msr, low, high);
 }
 
 #define rdmsrl(msr,val)							\
 	((val) = native_read_msr(msr))
 
-#define wrmsrl(msr,val)	native_write_msr(msr, val)
+#define wrmsrl(msr,val)	native_write_msr(msr, (u32)val, (u32)(val >> 32))
 
 /* wrmsr with exception handling */
-static inline int wrmsr_safe(u32 __msr, u32 __low, u32 __high)
+static inline int wrmsr_safe(unsigned msr, unsigned low, unsigned high)
 {
-	return native_write_msr_safe(__msr, ((u64)__high << 32) | __low);
+	return native_write_msr_safe(msr, low, high);
 }
 
 /* rdmsr with exception handling */
diff --git a/include/asm-x86/paravirt.h b/include/asm-x86/paravirt.h
index 4a7be97..6f837bb 100644
--- a/include/asm-x86/paravirt.h
+++ b/include/asm-x86/paravirt.h
@@ -115,7 +115,7 @@ struct pv_cpu_ops {
 	/* MSR, PMC and TSR operations.
 	   err = 0/-EFAULT.  wrmsr returns 0/-EFAULT. */
 	u64 (*read_msr)(unsigned int msr, int *err);
-	int (*write_msr)(unsigned int msr, u64 val);
+	int (*write_msr)(unsigned int msr, unsigned low, unsigned high);
 
 	u64 (*read_tsc)(void);
 	u64 (*read_pmc)(int counter);
-- 
1.4.4.2

[PATCH 8/10] change write msr functions interface

[Glauber de Oliveira Costa]

This patche changes the native_write_msr() and friends interface
to explicitly take 2 32-bit registers instead of a 64-bit value.
The change will ease the merge with 64-bit code. As the 64-bit
value will be passed as two registers anyway in i386,
the PVOP_CALL interface has to account for that and use low/high parameters
It would force the x86_64 version to be different.

The change does not make i386 generated code less efficient. As said above,
it would get the values from two registers anyway.

Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
---
 include/asm-x86/msr.h      |   19 ++++++++++---------
 include/asm-x86/paravirt.h |    2 +-
 2 files changed, 11 insertions(+), 10 deletions(-)

diff --git a/include/asm-x86/msr.h b/include/asm-x86/msr.h
index ce2a257..9c695c7 100644
--- a/include/asm-x86/msr.h
+++ b/include/asm-x86/msr.h
@@ -59,13 +59,14 @@ static inline unsigned long long native_read_msr_safe(unsigned int msr,
 	return val;
 }
 
-static inline void native_write_msr(unsigned int msr, unsigned long long val)
+static inline void native_write_msr(unsigned int msr,
+				    unsigned low, unsigned high)
 {
-	asm volatile("wrmsr" : : "c" (msr), "A"(val));
+	asm volatile("wrmsr" : : "c" (msr), "a"(low), "d" (high));
 }
 
 static inline int native_write_msr_safe(unsigned int msr,
-					unsigned long long val)
+					unsigned low, unsigned high)
 {
 	int err;
 	asm volatile("2: wrmsr ; xorl %0,%0\n"
@@ -78,7 +79,7 @@ static inline int native_write_msr_safe(unsigned int msr,
 		     "   .long	2b,3b\n\t"
 		     ".previous"
 		     : "=a" (err)
-		     : "c" (msr), "0" ((u32)val), "d" ((u32)(val>>32)),
+		     : "c" (msr), "0" (low), "d" (high),
 		       "i" (-EFAULT));
 	return err;
 }
@@ -114,20 +115,20 @@ static inline unsigned long long native_read_pmc(int counter)
 		(val2) = (u32)(__val >> 32);				\
 	} while(0)
 
-static inline void wrmsr(u32 __msr, u32 __low, u32 __high)
+static inline void wrmsr(unsigned msr, unsigned low, unsigned high)
 {
-	native_write_msr(__msr, ((u64)__high << 32) | __low);
+	native_write_msr(msr, low, high);
 }
 
 #define rdmsrl(msr,val)							\
 	((val) = native_read_msr(msr))
 
-#define wrmsrl(msr,val)	native_write_msr(msr, val)
+#define wrmsrl(msr,val)	native_write_msr(msr, (u32)val, (u32)(val >> 32))
 
 /* wrmsr with exception handling */
-static inline int wrmsr_safe(u32 __msr, u32 __low, u32 __high)
+static inline int wrmsr_safe(unsigned msr, unsigned low, unsigned high)
 {
-	return native_write_msr_safe(__msr, ((u64)__high << 32) | __low);
+	return native_write_msr_safe(msr, low, high);
 }
 
 /* rdmsr with exception handling */
diff --git a/include/asm-x86/paravirt.h b/include/asm-x86/paravirt.h
index 4a7be97..6f837bb 100644
--- a/include/asm-x86/paravirt.h
+++ b/include/asm-x86/paravirt.h
@@ -115,7 +115,7 @@ struct pv_cpu_ops {
 	/* MSR, PMC and TSR operations.
 	   err = 0/-EFAULT.  wrmsr returns 0/-EFAULT. */
 	u64 (*read_msr)(unsigned int msr, int *err);
-	int (*write_msr)(unsigned int msr, u64 val);
+	int (*write_msr)(unsigned int msr, unsigned low, unsigned high);
 
 	u64 (*read_tsc)(void);
 	u64 (*read_pmc)(int counter);
-- 
1.4.4.2

[PATCH 9/10] make fixups wordsize agnostic

[Glauber de Oliveira Costa]

This patch uses the _ASM_ALIGN and _ASM_PTR macros
to make the fixups in native_read/write_msr_safe look the same
for x86_64 and i386. Besides using this macros, we also have to
take the explicit instruction suffixes out. It's okay
because all this instructions uses registers, and can be sized by
them.

Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
---
 include/asm-x86/msr.h |   17 +++++++++--------
 1 files changed, 9 insertions(+), 8 deletions(-)

diff --git a/include/asm-x86/msr.h b/include/asm-x86/msr.h
index 9c695c7..f3a01ca 100644
--- a/include/asm-x86/msr.h
+++ b/include/asm-x86/msr.h
@@ -29,6 +29,7 @@ static inline unsigned long long native_read_tscp(int *aux)
 #ifdef __KERNEL__
 #ifndef __ASSEMBLY__
 
+#include <asm/asm.h>
 #include <asm/errno.h>
 
 static inline unsigned long long native_read_msr(unsigned int msr)
@@ -44,14 +45,14 @@ static inline unsigned long long native_read_msr_safe(unsigned int msr,
 {
 	unsigned long long val;
 
-	asm volatile("2: rdmsr ; xorl %0,%0\n"
+	asm volatile("2: rdmsr ; xor %0,%0\n"
 		     "1:\n\t"
 		     ".section .fixup,\"ax\"\n\t"
-		     "3:  movl %3,%0 ; jmp 1b\n\t"
+		     "3:  mov %3,%0 ; jmp 1b\n\t"
 		     ".previous\n\t"
 		     ".section __ex_table,\"a\"\n"
-		     "   .align 4\n\t"
-		     "   .long	2b,3b\n\t"
+		     _ASM_ALIGN "\n\t"
+		     _ASM_PTR " 2b,3b\n\t"
 		     ".previous"
 		     : "=r" (*err), "=A" (val)
 		     : "c" (msr), "i" (-EFAULT));
@@ -69,14 +70,14 @@ static inline int native_write_msr_safe(unsigned int msr,
 					unsigned low, unsigned high)
 {
 	int err;
-	asm volatile("2: wrmsr ; xorl %0,%0\n"
+	asm volatile("2: wrmsr ; xor %0,%0\n"
 		     "1:\n\t"
 		     ".section .fixup,\"ax\"\n\t"
-		     "3:  movl %4,%0 ; jmp 1b\n\t"
+		     "3:  mov %4,%0 ; jmp 1b\n\t"
 		     ".previous\n\t"
 		     ".section __ex_table,\"a\"\n"
-		     "   .align 4\n\t"
-		     "   .long	2b,3b\n\t"
+		     _ASM_ALIGN "\n\t"
+		     _ASM_PTR " 2b,3b\n\t"
 		     ".previous"
 		     : "=a" (err)
 		     : "c" (msr), "0" (low), "d" (high),
-- 
1.4.4.2

[PATCH 9/10] make fixups wordsize agnostic

[Glauber de Oliveira Costa]

This patch uses the _ASM_ALIGN and _ASM_PTR macros
to make the fixups in native_read/write_msr_safe look the same
for x86_64 and i386. Besides using this macros, we also have to
take the explicit instruction suffixes out. It's okay
because all this instructions uses registers, and can be sized by
them.

Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
---
 include/asm-x86/msr.h |   17 +++++++++--------
 1 files changed, 9 insertions(+), 8 deletions(-)

diff --git a/include/asm-x86/msr.h b/include/asm-x86/msr.h
index 9c695c7..f3a01ca 100644
--- a/include/asm-x86/msr.h
+++ b/include/asm-x86/msr.h
@@ -29,6 +29,7 @@ static inline unsigned long long native_read_tscp(int *aux)
 #ifdef __KERNEL__
 #ifndef __ASSEMBLY__
 
+#include <asm/asm.h>
 #include <asm/errno.h>
 
 static inline unsigned long long native_read_msr(unsigned int msr)
@@ -44,14 +45,14 @@ static inline unsigned long long native_read_msr_safe(unsigned int msr,
 {
 	unsigned long long val;
 
-	asm volatile("2: rdmsr ; xorl %0,%0\n"
+	asm volatile("2: rdmsr ; xor %0,%0\n"
 		     "1:\n\t"
 		     ".section .fixup,\"ax\"\n\t"
-		     "3:  movl %3,%0 ; jmp 1b\n\t"
+		     "3:  mov %3,%0 ; jmp 1b\n\t"
 		     ".previous\n\t"
 		     ".section __ex_table,\"a\"\n"
-		     "   .align 4\n\t"
-		     "   .long	2b,3b\n\t"
+		     _ASM_ALIGN "\n\t"
+		     _ASM_PTR " 2b,3b\n\t"
 		     ".previous"
 		     : "=r" (*err), "=A" (val)
 		     : "c" (msr), "i" (-EFAULT));
@@ -69,14 +70,14 @@ static inline int native_write_msr_safe(unsigned int msr,
 					unsigned low, unsigned high)
 {
 	int err;
-	asm volatile("2: wrmsr ; xorl %0,%0\n"
+	asm volatile("2: wrmsr ; xor %0,%0\n"
 		     "1:\n\t"
 		     ".section .fixup,\"ax\"\n\t"
-		     "3:  movl %4,%0 ; jmp 1b\n\t"
+		     "3:  mov %4,%0 ; jmp 1b\n\t"
 		     ".previous\n\t"
 		     ".section __ex_table,\"a\"\n"
-		     "   .align 4\n\t"
-		     "   .long	2b,3b\n\t"
+		     _ASM_ALIGN "\n\t"
+		     _ASM_PTR " 2b,3b\n\t"
 		     ".previous"
 		     : "=a" (err)
 		     : "c" (msr), "0" (low), "d" (high),
-- 
1.4.4.2

[PATCH 10/10] integrate i386 and x86_64 code in msr.h

[Glauber de Oliveira Costa]

This patches proceeds with the integration of msr.h, making
the code unified, instead of having a version for each architecture.
We stick with the native_* functions, and then paravirt comes for free.

Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
---
 include/asm-x86/msr.h |  161 ++++++++++++++-----------------------------------
 1 files changed, 46 insertions(+), 115 deletions(-)

diff --git a/include/asm-x86/msr.h b/include/asm-x86/msr.h
index f3a01ca..84116fa 100644
--- a/include/asm-x86/msr.h
+++ b/include/asm-x86/msr.h
@@ -5,6 +5,10 @@
 
 #ifdef __KERNEL__
 #ifndef __ASSEMBLY__
+
+#include <asm/asm.h>
+#include <asm/errno.h>
+
 static inline unsigned long long native_read_tscp(int *aux)
 {
 	unsigned long low, high;
@@ -13,37 +17,36 @@ static inline unsigned long long native_read_tscp(int *aux)
 	return low | ((u64)high >> 32);
 }
 
-#define rdtscp(low, high, aux)						\
-       do {                                                            \
-               unsigned long long _val = native_read_tscp(&(aux));     \
-               (low) = (u32)_val;                                      \
-               (high) = (u32)(_val >> 32);                             \
-       } while (0)
-
-#define rdtscpll(val, aux) (val) = native_read_tscp(&(aux))
-#endif
+/*
+ * i386 calling convention returns 64-bit value in edx:eax, while
+ * x86_64 returns at rax. Also, the "A" constraint does not really
+ * mean rdx:rax in x86_64, so we need specialized behaviour for each
+ * architecture
+ */
+#ifdef CONFIG_X86_64
+#define DECLARE_ARGS(val, low, high)	unsigned low, high
+#define EAX_EDX_VAL(val, low, high)	(low | ((u64)(high) << 32))
+#define EAX_EDX_ARGS(val, low, high)	"a" (low), "d" (high)
+#define EAX_EDX_RET(val, low, high)	"=a" (low), "=d" (high)
+#else
+#define DECLARE_ARGS(val, low, high)	unsigned long long val
+#define EAX_EDX_VAL(val, low, high)	(val)
+#define EAX_EDX_ARGS(val, low, high)	"A" (val)
+#define EAX_EDX_RET(val, low, high)	"=A" (val)
 #endif
 
-#ifdef __i386__
-
-#ifdef __KERNEL__
-#ifndef __ASSEMBLY__
-
-#include <asm/asm.h>
-#include <asm/errno.h>
-
 static inline unsigned long long native_read_msr(unsigned int msr)
 {
-	unsigned long long val;
+	DECLARE_ARGS(val, low, high);
 
-	asm volatile("rdmsr" : "=A" (val) : "c" (msr));
-	return val;
+	asm volatile("rdmsr" : EAX_EDX_RET(val, low, high) : "c" (msr));
+	return EAX_EDX_VAL(val, low, high);
 }
 
 static inline unsigned long long native_read_msr_safe(unsigned int msr,
 						      int *err)
 {
-	unsigned long long val;
+	DECLARE_ARGS(val, low, high);
 
 	asm volatile("2: rdmsr ; xor %0,%0\n"
 		     "1:\n\t"
@@ -54,10 +57,9 @@ static inline unsigned long long native_read_msr_safe(unsigned int msr,
 		     _ASM_ALIGN "\n\t"
 		     _ASM_PTR " 2b,3b\n\t"
 		     ".previous"
-		     : "=r" (*err), "=A" (val)
+		     : "=r" (*err), EAX_EDX_RET(val, low, high)
 		     : "c" (msr), "i" (-EFAULT));
-
-	return val;
+	return EAX_EDX_VAL(val, low, high);
 }
 
 static inline void native_write_msr(unsigned int msr,
@@ -87,16 +89,18 @@ static inline int native_write_msr_safe(unsigned int msr,
 
 static inline unsigned long long native_read_tsc(void)
 {
-	unsigned long long val;
-	asm volatile("rdtsc" : "=A" (val));
-	return val;
+	DECLARE_ARGS(val, low, high);
+
+	asm volatile("rdtsc" : EAX_EDX_RET(val, low, high));
+	return EAX_EDX_VAL(val, low, high);
 }
 
 static inline unsigned long long native_read_pmc(int counter)
 {
-	unsigned long long val;
-	asm volatile("rdpmc" : "=A" (val) : "c" (counter));
-	return val;
+	DECLARE_ARGS(val, low, high);
+
+	asm volatile("rdpmc" : EAX_EDX_RET(val, low, high) : "c" (counter));
+	return EAX_EDX_VAL(val, low, high);
 }
 
 #ifdef CONFIG_PARAVIRT
@@ -124,7 +128,8 @@ static inline void wrmsr(unsigned msr, unsigned low, unsigned high)
 #define rdmsrl(msr,val)							\
 	((val) = native_read_msr(msr))
 
-#define wrmsrl(msr,val)	native_write_msr(msr, (u32)val, (u32)(val >> 32))
+#define wrmsrl(msr,val)							\
+	native_write_msr(msr, (u32)((u64)(val)), (u32)((u64)(val) >> 32))
 
 /* wrmsr with exception handling */
 static inline int wrmsr_safe(unsigned msr, unsigned low, unsigned high)
@@ -156,101 +161,25 @@ static inline int wrmsr_safe(unsigned msr, unsigned low, unsigned high)
 		(low)  = (u32)_l;				\
 		(high) = (u32)(_l >> 32);			\
 	} while(0)
-#endif	/* !CONFIG_PARAVIRT */
-
-#endif  /* ! __ASSEMBLY__ */
-#endif  /* __KERNEL__ */
 
-#else   /* __i386__ */
-
-#ifndef __ASSEMBLY__
-#include <linux/errno.h>
-/*
- * Access to machine-specific registers (available on 586 and better only)
- * Note: the rd* operations modify the parameters directly (without using
- * pointer indirection), this allows gcc to optimize better
- */
-
-#define rdmsr(msr,val1,val2) \
-       __asm__ __volatile__("rdmsr" \
-			    : "=a" (val1), "=d" (val2) \
-			    : "c" (msr))
-
-
-#define rdmsrl(msr,val) do { unsigned long a__,b__; \
-       __asm__ __volatile__("rdmsr" \
-			    : "=a" (a__), "=d" (b__) \
-			    : "c" (msr)); \
-       val = a__ | (b__<<32); \
-} while(0)
+#define rdtscp(low, high, aux)						\
+       do {                                                            \
+               unsigned long long _val = native_read_tscp(&(aux));     \
+               (low) = (u32)_val;                                      \
+               (high) = (u32)(_val >> 32);                             \
+       } while (0)
 
-#define wrmsr(msr,val1,val2) \
-     __asm__ __volatile__("wrmsr" \
-			  : /* no outputs */ \
-			  : "c" (msr), "a" (val1), "d" (val2))
+#define rdtscpll(val, aux) (val) = native_read_tscp(&(aux))
 
-#define wrmsrl(msr,val) wrmsr(msr,(__u32)((__u64)(val)),((__u64)(val))>>32)
+#endif	/* !CONFIG_PARAVIRT */
 
-/* wrmsr with exception handling */
-#define wrmsr_safe(msr,a,b) ({ int ret__;			\
-	asm volatile("2: wrmsr ; xorl %0,%0\n"			\
-		     "1:\n\t"					\
-		     ".section .fixup,\"ax\"\n\t"		\
-		     "3:  movl %4,%0 ; jmp 1b\n\t"		\
-		     ".previous\n\t"				\
-		     ".section __ex_table,\"a\"\n"		\
-		     "   .align 8\n\t"				\
-		     "   .quad	2b,3b\n\t"			\
-		     ".previous"				\
-		     : "=a" (ret__)				\
-		     : "c" (msr), "0" (a), "d" (b), "i" (-EFAULT)); \
-	ret__; })
 
 #define checking_wrmsrl(msr,val) wrmsr_safe(msr,(u32)(val),(u32)((val)>>32))
 
-#define rdmsr_safe(msr,a,b) \
-	({ int ret__;						\
-	  asm volatile ("1:       rdmsr\n"			\
-			"2:\n"					\
-			".section .fixup,\"ax\"\n"		\
-			"3:       movl %4,%0\n"			\
-			" jmp 2b\n"				\
-			".previous\n"				\
-			".section __ex_table,\"a\"\n"		\
-			" .align 8\n"				\
-			" .quad 1b,3b\n"				\
-			".previous":"=&bDS" (ret__), "=a"(*(a)), "=d"(*(b)) \
-			:"c"(msr), "i"(-EIO), "0"(0));			\
-	  ret__; })
-
-#define rdtsc(low,high) \
-     __asm__ __volatile__("rdtsc" : "=a" (low), "=d" (high))
-
-#define rdtscl(low) \
-     __asm__ __volatile__ ("rdtsc" : "=a" (low) : : "edx")
-
-
-#define rdtscll(val) do { \
-     unsigned int __a,__d; \
-     asm volatile("rdtsc" : "=a" (__a), "=d" (__d)); \
-     (val) = ((unsigned long)__a) | (((unsigned long)__d)<<32); \
-} while(0)
-
 #define write_tsc(val1,val2) wrmsr(0x10, val1, val2)
 
 #define write_rdtscp_aux(val) wrmsr(0xc0000103, val, 0)
 
-#define rdpmc(counter,low,high) \
-     __asm__ __volatile__("rdpmc" \
-			  : "=a" (low), "=d" (high) \
-			  : "c" (counter))
-
-#endif  /* __ASSEMBLY__ */
-
-#endif  /* !__i386__ */
-
-#ifndef __ASSEMBLY__
-
 #ifdef CONFIG_SMP
 void rdmsr_on_cpu(unsigned int cpu, u32 msr_no, u32 *l, u32 *h);
 void wrmsr_on_cpu(unsigned int cpu, u32 msr_no, u32 l, u32 h);
@@ -275,5 +204,7 @@ static inline int wrmsr_safe_on_cpu(unsigned int cpu, u32 msr_no, u32 l, u32 h)
 }
 #endif  /* CONFIG_SMP */
 #endif /* __ASSEMBLY__ */
+#endif /* __KERNEL__ */
+
 
 #endif
-- 
1.4.4.2

[PATCH 10/10] integrate i386 and x86_64 code in msr.h

[Glauber de Oliveira Costa]

This patches proceeds with the integration of msr.h, making
the code unified, instead of having a version for each architecture.
We stick with the native_* functions, and then paravirt comes for free.

Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
---
 include/asm-x86/msr.h |  161 ++++++++++++++-----------------------------------
 1 files changed, 46 insertions(+), 115 deletions(-)

diff --git a/include/asm-x86/msr.h b/include/asm-x86/msr.h
index f3a01ca..84116fa 100644
--- a/include/asm-x86/msr.h
+++ b/include/asm-x86/msr.h
@@ -5,6 +5,10 @@
 
 #ifdef __KERNEL__
 #ifndef __ASSEMBLY__
+
+#include <asm/asm.h>
+#include <asm/errno.h>
+
 static inline unsigned long long native_read_tscp(int *aux)
 {
 	unsigned long low, high;
@@ -13,37 +17,36 @@ static inline unsigned long long native_read_tscp(int *aux)
 	return low | ((u64)high >> 32);
 }
 
-#define rdtscp(low, high, aux)						\
-       do {                                                            \
-               unsigned long long _val = native_read_tscp(&(aux));     \
-               (low) = (u32)_val;                                      \
-               (high) = (u32)(_val >> 32);                             \
-       } while (0)
-
-#define rdtscpll(val, aux) (val) = native_read_tscp(&(aux))
-#endif
+/*
+ * i386 calling convention returns 64-bit value in edx:eax, while
+ * x86_64 returns at rax. Also, the "A" constraint does not really
+ * mean rdx:rax in x86_64, so we need specialized behaviour for each
+ * architecture
+ */
+#ifdef CONFIG_X86_64
+#define DECLARE_ARGS(val, low, high)	unsigned low, high
+#define EAX_EDX_VAL(val, low, high)	(low | ((u64)(high) << 32))
+#define EAX_EDX_ARGS(val, low, high)	"a" (low), "d" (high)
+#define EAX_EDX_RET(val, low, high)	"=a" (low), "=d" (high)
+#else
+#define DECLARE_ARGS(val, low, high)	unsigned long long val
+#define EAX_EDX_VAL(val, low, high)	(val)
+#define EAX_EDX_ARGS(val, low, high)	"A" (val)
+#define EAX_EDX_RET(val, low, high)	"=A" (val)
 #endif
 
-#ifdef __i386__
-
-#ifdef __KERNEL__
-#ifndef __ASSEMBLY__
-
-#include <asm/asm.h>
-#include <asm/errno.h>
-
 static inline unsigned long long native_read_msr(unsigned int msr)
 {
-	unsigned long long val;
+	DECLARE_ARGS(val, low, high);
 
-	asm volatile("rdmsr" : "=A" (val) : "c" (msr));
-	return val;
+	asm volatile("rdmsr" : EAX_EDX_RET(val, low, high) : "c" (msr));
+	return EAX_EDX_VAL(val, low, high);
 }
 
 static inline unsigned long long native_read_msr_safe(unsigned int msr,
 						      int *err)
 {
-	unsigned long long val;
+	DECLARE_ARGS(val, low, high);
 
 	asm volatile("2: rdmsr ; xor %0,%0\n"
 		     "1:\n\t"
@@ -54,10 +57,9 @@ static inline unsigned long long native_read_msr_safe(unsigned int msr,
 		     _ASM_ALIGN "\n\t"
 		     _ASM_PTR " 2b,3b\n\t"
 		     ".previous"
-		     : "=r" (*err), "=A" (val)
+		     : "=r" (*err), EAX_EDX_RET(val, low, high)
 		     : "c" (msr), "i" (-EFAULT));
-
-	return val;
+	return EAX_EDX_VAL(val, low, high);
 }
 
 static inline void native_write_msr(unsigned int msr,
@@ -87,16 +89,18 @@ static inline int native_write_msr_safe(unsigned int msr,
 
 static inline unsigned long long native_read_tsc(void)
 {
-	unsigned long long val;
-	asm volatile("rdtsc" : "=A" (val));
-	return val;
+	DECLARE_ARGS(val, low, high);
+
+	asm volatile("rdtsc" : EAX_EDX_RET(val, low, high));
+	return EAX_EDX_VAL(val, low, high);
 }
 
 static inline unsigned long long native_read_pmc(int counter)
 {
-	unsigned long long val;
-	asm volatile("rdpmc" : "=A" (val) : "c" (counter));
-	return val;
+	DECLARE_ARGS(val, low, high);
+
+	asm volatile("rdpmc" : EAX_EDX_RET(val, low, high) : "c" (counter));
+	return EAX_EDX_VAL(val, low, high);
 }
 
 #ifdef CONFIG_PARAVIRT
@@ -124,7 +128,8 @@ static inline void wrmsr(unsigned msr, unsigned low, unsigned high)
 #define rdmsrl(msr,val)							\
 	((val) = native_read_msr(msr))
 
-#define wrmsrl(msr,val)	native_write_msr(msr, (u32)val, (u32)(val >> 32))
+#define wrmsrl(msr,val)							\
+	native_write_msr(msr, (u32)((u64)(val)), (u32)((u64)(val) >> 32))
 
 /* wrmsr with exception handling */
 static inline int wrmsr_safe(unsigned msr, unsigned low, unsigned high)
@@ -156,101 +161,25 @@ static inline int wrmsr_safe(unsigned msr, unsigned low, unsigned high)
 		(low)  = (u32)_l;				\
 		(high) = (u32)(_l >> 32);			\
 	} while(0)
-#endif	/* !CONFIG_PARAVIRT */
-
-#endif  /* ! __ASSEMBLY__ */
-#endif  /* __KERNEL__ */
 
-#else   /* __i386__ */
-
-#ifndef __ASSEMBLY__
-#include <linux/errno.h>
-/*
- * Access to machine-specific registers (available on 586 and better only)
- * Note: the rd* operations modify the parameters directly (without using
- * pointer indirection), this allows gcc to optimize better
- */
-
-#define rdmsr(msr,val1,val2) \
-       __asm__ __volatile__("rdmsr" \
-			    : "=a" (val1), "=d" (val2) \
-			    : "c" (msr))
-
-
-#define rdmsrl(msr,val) do { unsigned long a__,b__; \
-       __asm__ __volatile__("rdmsr" \
-			    : "=a" (a__), "=d" (b__) \
-			    : "c" (msr)); \
-       val = a__ | (b__<<32); \
-} while(0)
+#define rdtscp(low, high, aux)						\
+       do {                                                            \
+               unsigned long long _val = native_read_tscp(&(aux));     \
+               (low) = (u32)_val;                                      \
+               (high) = (u32)(_val >> 32);                             \
+       } while (0)
 
-#define wrmsr(msr,val1,val2) \
-     __asm__ __volatile__("wrmsr" \
-			  : /* no outputs */ \
-			  : "c" (msr), "a" (val1), "d" (val2))
+#define rdtscpll(val, aux) (val) = native_read_tscp(&(aux))
 
-#define wrmsrl(msr,val) wrmsr(msr,(__u32)((__u64)(val)),((__u64)(val))>>32)
+#endif	/* !CONFIG_PARAVIRT */
 
-/* wrmsr with exception handling */
-#define wrmsr_safe(msr,a,b) ({ int ret__;			\
-	asm volatile("2: wrmsr ; xorl %0,%0\n"			\
-		     "1:\n\t"					\
-		     ".section .fixup,\"ax\"\n\t"		\
-		     "3:  movl %4,%0 ; jmp 1b\n\t"		\
-		     ".previous\n\t"				\
-		     ".section __ex_table,\"a\"\n"		\
-		     "   .align 8\n\t"				\
-		     "   .quad	2b,3b\n\t"			\
-		     ".previous"				\
-		     : "=a" (ret__)				\
-		     : "c" (msr), "0" (a), "d" (b), "i" (-EFAULT)); \
-	ret__; })
 
 #define checking_wrmsrl(msr,val) wrmsr_safe(msr,(u32)(val),(u32)((val)>>32))
 
-#define rdmsr_safe(msr,a,b) \
-	({ int ret__;						\
-	  asm volatile ("1:       rdmsr\n"			\
-			"2:\n"					\
-			".section .fixup,\"ax\"\n"		\
-			"3:       movl %4,%0\n"			\
-			" jmp 2b\n"				\
-			".previous\n"				\
-			".section __ex_table,\"a\"\n"		\
-			" .align 8\n"				\
-			" .quad 1b,3b\n"				\
-			".previous":"=&bDS" (ret__), "=a"(*(a)), "=d"(*(b)) \
-			:"c"(msr), "i"(-EIO), "0"(0));			\
-	  ret__; })
-
-#define rdtsc(low,high) \
-     __asm__ __volatile__("rdtsc" : "=a" (low), "=d" (high))
-
-#define rdtscl(low) \
-     __asm__ __volatile__ ("rdtsc" : "=a" (low) : : "edx")
-
-
-#define rdtscll(val) do { \
-     unsigned int __a,__d; \
-     asm volatile("rdtsc" : "=a" (__a), "=d" (__d)); \
-     (val) = ((unsigned long)__a) | (((unsigned long)__d)<<32); \
-} while(0)
-
 #define write_tsc(val1,val2) wrmsr(0x10, val1, val2)
 
 #define write_rdtscp_aux(val) wrmsr(0xc0000103, val, 0)
 
-#define rdpmc(counter,low,high) \
-     __asm__ __volatile__("rdpmc" \
-			  : "=a" (low), "=d" (high) \
-			  : "c" (counter))
-
-#endif  /* __ASSEMBLY__ */
-
-#endif  /* !__i386__ */
-
-#ifndef __ASSEMBLY__
-
 #ifdef CONFIG_SMP
 void rdmsr_on_cpu(unsigned int cpu, u32 msr_no, u32 *l, u32 *h);
 void wrmsr_on_cpu(unsigned int cpu, u32 msr_no, u32 l, u32 h);
@@ -275,5 +204,7 @@ static inline int wrmsr_safe_on_cpu(unsigned int cpu, u32 msr_no, u32 l, u32 h)
 }
 #endif  /* CONFIG_SMP */
 #endif /* __ASSEMBLY__ */
+#endif /* __KERNEL__ */
+
 
 #endif
-- 
1.4.4.2

Re: [PATCH 3/10] allow sched clock to be overridden by paravirt

[Andi Kleen]

On Tue, Dec 04, 2007 at 09:09:57AM -0200, Glauber de Oliveira Costa wrote:
> This patch turns the sched_clock into native_sched_clock.
> sched clock becomes a weak symbol, which can then give its
> place to a paravirt definition.

I ended up regretting that solution on i386, it would be better
to solve that more clearly with ifdefs.

-Andi

Re: [PATCH 3/10] allow sched clock to be overridden by paravirt

[Andi Kleen]

On Tue, Dec 04, 2007 at 09:09:57AM -0200, Glauber de Oliveira Costa wrote:
> This patch turns the sched_clock into native_sched_clock.
> sched clock becomes a weak symbol, which can then give its
> place to a paravirt definition.

I ended up regretting that solution on i386, it would be better
to solve that more clearly with ifdefs.

-Andi

Re: [PATCH 0/10] Integrate msr.h

[Ingo Molnar]

* Glauber de Oliveira Costa <gcosta@redhat.com> wrote:

> Hello,
> 
> This series of patches integrates msr.h header. What it really does, 
> is a series of steps to allow us to get rid of duplicate code between 
> i386 and x86_64 versions
> 
> With this done, achieving paravirt for x86_64 gets really easy, just a 
> couple of extra code.

thanks, the patches look certainly sane and i've picked them up into 
x86.git. A quick build & boot test was successful on both 64-bit and 
32-bit, with paravirt on/off. I've pushed out the latest tree into the 
'mm' branch of:

   git://git.kernel.org/pub/scm/linux/kernel/git/mingo/linux-2.6-x86.git

minor nit: some of your patches had scripts/checkpatch.pl failures, 
could you try to fix them up? (If possible then as a follow-up second 
patch series.) Thanks!

[ btw., how far away are you from having Kconfig selectable 64-bit
  paravirt support? Just curious :-) ]

	Ingo

Re: [PATCH 0/10] Integrate msr.h

[Ingo Molnar]

* Glauber de Oliveira Costa <gcosta@redhat.com> wrote:

> Hello,
> 
> This series of patches integrates msr.h header. What it really does, 
> is a series of steps to allow us to get rid of duplicate code between 
> i386 and x86_64 versions
> 
> With this done, achieving paravirt for x86_64 gets really easy, just a 
> couple of extra code.

thanks, the patches look certainly sane and i've picked them up into 
x86.git. A quick build & boot test was successful on both 64-bit and 
32-bit, with paravirt on/off. I've pushed out the latest tree into the 
'mm' branch of:

   git://git.kernel.org/pub/scm/linux/kernel/git/mingo/linux-2.6-x86.git

minor nit: some of your patches had scripts/checkpatch.pl failures, 
could you try to fix them up? (If possible then as a follow-up second 
patch series.) Thanks!

[ btw., how far away are you from having Kconfig selectable 64-bit
  paravirt support? Just curious :-) ]

	Ingo

Re: [PATCH 0/10] Integrate msr.h

[Glauber de Oliveira Costa]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Ingo Molnar escreveu:
> * Glauber de Oliveira Costa <gcosta@redhat.com> wrote:
> 
>> Hello,
>>
>> This series of patches integrates msr.h header. What it really does, 
>> is a series of steps to allow us to get rid of duplicate code between 
>> i386 and x86_64 versions
>>
>> With this done, achieving paravirt for x86_64 gets really easy, just a 
>> couple of extra code.
> 
> thanks, the patches look certainly sane and i've picked them up into 
> x86.git. A quick build & boot test was successful on both 64-bit and 
> 32-bit, with paravirt on/off. I've pushed out the latest tree into the 
> 'mm' branch of:
> 
>    git://git.kernel.org/pub/scm/linux/kernel/git/mingo/linux-2.6-x86.git
> 
> minor nit: some of your patches had scripts/checkpatch.pl failures, 
> could you try to fix them up? (If possible then as a follow-up second 
> patch series.) Thanks!
Sure. I'll resend them.

> [ btw., how far away are you from having Kconfig selectable 64-bit
>   paravirt support? Just curious :-) ]
> 
Not too much. Jeremy is handling the page table code, and I'm basically
taking care of the rest. I don't go to far in integration in other
areas, so the patches are the same that were already previously sent
some days ago. I'll just rebase and send again.

The only rebase I'm expecting to be troublesome is the paravirt.c /
paravirt.h files. So I'd say we're close.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (GNU/Linux)
Comment: Using GnuPG with Remi - http://enigmail.mozdev.org

iD8DBQFHVXagjYI8LaFUWXMRAhHjAKDVDesU/iRyDXDgQB6Pxx23PTizyQCgzzcU
rZsECAdjqaGeZqAxWY2lhro=
=qrHF
-----END PGP SIGNATURE-----

Re: [PATCH 0/10] Integrate msr.h

[Glauber de Oliveira Costa]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Ingo Molnar escreveu:
> * Glauber de Oliveira Costa <gcosta@redhat.com> wrote:
> 
>> Hello,
>>
>> This series of patches integrates msr.h header. What it really does, 
>> is a series of steps to allow us to get rid of duplicate code between 
>> i386 and x86_64 versions
>>
>> With this done, achieving paravirt for x86_64 gets really easy, just a 
>> couple of extra code.
> 
> thanks, the patches look certainly sane and i've picked them up into 
> x86.git. A quick build & boot test was successful on both 64-bit and 
> 32-bit, with paravirt on/off. I've pushed out the latest tree into the 
> 'mm' branch of:
> 
>    git://git.kernel.org/pub/scm/linux/kernel/git/mingo/linux-2.6-x86.git
> 
> minor nit: some of your patches had scripts/checkpatch.pl failures, 
> could you try to fix them up? (If possible then as a follow-up second 
> patch series.) Thanks!
Sure. I'll resend them.

> [ btw., how far away are you from having Kconfig selectable 64-bit
>   paravirt support? Just curious :-) ]
> 
Not too much. Jeremy is handling the page table code, and I'm basically
taking care of the rest. I don't go to far in integration in other
areas, so the patches are the same that were already previously sent
some days ago. I'll just rebase and send again.

The only rebase I'm expecting to be troublesome is the paravirt.c /
paravirt.h files. So I'd say we're close.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (GNU/Linux)
Comment: Using GnuPG with Remi - http://enigmail.mozdev.org

iD8DBQFHVXagjYI8LaFUWXMRAhHjAKDVDesU/iRyDXDgQB6Pxx23PTizyQCgzzcU
rZsECAdjqaGeZqAxWY2lhro=
=qrHF
-----END PGP SIGNATURE-----

[PATCH 3/10] allow sched clock to be overridden by paravirt

[Glauber de Oliveira Costa]

This patch turns the sched_clock into native_sched_clock.
sched clock becomes a weak symbol, which can then give its
place to a paravirt definition.

Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
---
 arch/x86/kernel/tsc_64.c |   15 ++++++++++++++-
 1 files changed, 14 insertions(+), 1 deletions(-)

diff --git a/arch/x86/kernel/tsc_64.c b/arch/x86/kernel/tsc_64.c
index 9c70af4..5158476 100644
--- a/arch/x86/kernel/tsc_64.c
+++ b/arch/x86/kernel/tsc_64.c
@@ -30,7 +30,7 @@ static unsigned long long cycles_2_ns(unsigned long long cyc)
 	return (cyc * cyc2ns_scale) >> NS_SCALE;
 }
 
-unsigned long long sched_clock(void)
+unsigned long long native_sched_clock(void)
 {
 	unsigned long a = 0;
 
@@ -44,6 +44,19 @@ unsigned long long sched_clock(void)
 	return cycles_2_ns(a);
 }
 
+/* We need to define a real function for sched_clock, to override the
+   weak default version */
+#ifdef CONFIG_PARAVIRT
+unsigned long long sched_clock(void)
+{
+	return paravirt_sched_clock();
+}
+#else
+unsigned long long
+sched_clock(void) __attribute__((alias("native_sched_clock")));
+#endif
+
+
 static int tsc_unstable;
 
 inline int check_tsc_unstable(void)
-- 
1.4.4.2

[PATCH 3/10] allow sched clock to be overridden by paravirt

[Glauber de Oliveira Costa]

This patch turns the sched_clock into native_sched_clock.
sched clock becomes a weak symbol, which can then give its
place to a paravirt definition.

Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
---
 arch/x86/kernel/tsc_64.c |   15 ++++++++++++++-
 1 files changed, 14 insertions(+), 1 deletions(-)

diff --git a/arch/x86/kernel/tsc_64.c b/arch/x86/kernel/tsc_64.c
index 9c70af4..5158476 100644
--- a/arch/x86/kernel/tsc_64.c
+++ b/arch/x86/kernel/tsc_64.c
@@ -30,7 +30,7 @@ static unsigned long long cycles_2_ns(unsigned long long cyc)
 	return (cyc * cyc2ns_scale) >> NS_SCALE;
 }
 
-unsigned long long sched_clock(void)
+unsigned long long native_sched_clock(void)
 {
 	unsigned long a = 0;
 
@@ -44,6 +44,19 @@ unsigned long long sched_clock(void)
 	return cycles_2_ns(a);
 }
 
+/* We need to define a real function for sched_clock, to override the
+   weak default version */
+#ifdef CONFIG_PARAVIRT
+unsigned long long sched_clock(void)
+{
+	return paravirt_sched_clock();
+}
+#else
+unsigned long long
+sched_clock(void) __attribute__((alias("native_sched_clock")));
+#endif
+
+
 static int tsc_unstable;
 
 inline int check_tsc_unstable(void)
-- 
1.4.4.2