Add a sched_clock paravirt_op

The tsc-based get_scheduled_cycles interface is not a good match for
Xen's runstate accounting, which reports everything in nanoseconds.

This patch replaces this interface with a sched_clock interface, which
matches both Xen and VMI's requirements.

In order to do this, we:
   1. replace get_scheduled_cycles with sched_clock
   2. hoist cycles_2_ns into a common header
   3. update vmi accordingly

One thing to note: because sched_clock is implemented as a weak
function in kernel/sched.c, we must define a real function in order to
override this weak binding.  This means the usual paravirt_ops
technique of using an inline function won't work in this case.

Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Cc: Zachary Amsden <zach@vmware.com>
Cc: Dan Hecht <dhecht@vmware.com>
Cc: john stultz <johnstul@us.ibm.com>

7 files changed, 57 insertions, 17 deletions

diff --git a/arch/i386/kernel/paravirt.c b/arch/i386/kernel/paravirt.c
index 60e08b9b50a4..53f07a8275e3 100644
--- a/arch/i386/kernel/paravirt.c
+++ b/arch/i386/kernel/paravirt.c
@@ -302,7 +302,7 @@ struct paravirt_ops paravirt_ops = {
         .write_msr = native_write_msr_safe,
         .read_tsc = native_read_tsc,
         .read_pmc = native_read_pmc,
-        .get_scheduled_cycles = native_read_tsc,
+        .sched_clock = native_sched_clock,
         .get_cpu_khz = native_calculate_cpu_khz,
         .load_tr_desc = native_load_tr_desc,
         .set_ldt = native_set_ldt,
diff --git a/arch/i386/kernel/tsc.c b/arch/i386/kernel/tsc.c
index ea63a30ca3e8..252f9010f283 100644
--- a/arch/i386/kernel/tsc.c
+++ b/arch/i386/kernel/tsc.c
@@ -84,7 +84,7 @@ static inline int check_tsc_unstable(void)
  *
  *                      -johnstul@us.ibm.com "math is hard, lets go shopping!"
  */
-static unsigned long cyc2ns_scale __read_mostly;
+unsigned long cyc2ns_scale __read_mostly;
 
 #define CYC2NS_SCALE_FACTOR 10 /* 2^10, carefully chosen */
 
@@ -93,15 +93,10 @@ static inline void set_cyc2ns_scale(unsigned long cpu_khz)
         cyc2ns_scale = (1000000 << CYC2NS_SCALE_FACTOR)/cpu_khz;
 }
 
-static inline unsigned long long cycles_2_ns(unsigned long long cyc)
-{
-        return (cyc * cyc2ns_scale) >> CYC2NS_SCALE_FACTOR;
-}
-
 /*
  * Scheduler clock - returns current time in nanosec units.
  */
-unsigned long long sched_clock(void)
+unsigned long long native_sched_clock(void)
 {
         unsigned long long this_offset;
 
@@ -118,12 +113,24 @@ unsigned long long sched_clock(void)
                 return (jiffies_64 - INITIAL_JIFFIES) * (1000000000 / HZ);
 
         /* read the Time Stamp Counter: */
-        get_scheduled_cycles(this_offset);
+        rdtscll(this_offset);
 
         /* return the value in ns */
         return cycles_2_ns(this_offset);
 }
 
+/* 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
+
 unsigned long native_calculate_cpu_khz(void)
 {
         unsigned long long start, end;
diff --git a/arch/i386/kernel/vmi.c b/arch/i386/kernel/vmi.c
index 234bd6ff518d..72042bb7ec94 100644
--- a/arch/i386/kernel/vmi.c
+++ b/arch/i386/kernel/vmi.c
@@ -891,7 +891,7 @@ static inline int __init activate_vmi(void)
                 paravirt_ops.setup_boot_clock = vmi_time_bsp_init;
                 paravirt_ops.setup_secondary_clock = vmi_time_ap_init;
 #endif
-                paravirt_ops.get_scheduled_cycles = vmi_get_sched_cycles;
+                paravirt_ops.sched_clock = vmi_sched_clock;
                 paravirt_ops.get_cpu_khz = vmi_cpu_khz;
 
                 /* We have true wallclock functions; disable CMOS clock sync */
diff --git a/arch/i386/kernel/vmiclock.c b/arch/i386/kernel/vmiclock.c
index 26a37f8a8762..f9b845f4e692 100644
--- a/arch/i386/kernel/vmiclock.c
+++ b/arch/i386/kernel/vmiclock.c
@@ -64,10 +64,10 @@ int vmi_set_wallclock(unsigned long now)
         return 0;
 }
 
-/* paravirt_ops.get_scheduled_cycles = vmi_get_sched_cycles */
-unsigned long long vmi_get_sched_cycles(void)
+/* paravirt_ops.sched_clock = vmi_sched_clock */
+unsigned long long vmi_sched_clock(void)
 {
-        return vmi_timer_ops.get_cycle_counter(VMI_CYCLES_AVAILABLE);
+        return cycles_2_ns(vmi_timer_ops.get_cycle_counter(VMI_CYCLES_AVAILABLE));
 }
 
 /* paravirt_ops.get_cpu_khz = vmi_cpu_khz */
diff --git a/include/asm-i386/paravirt.h b/include/asm-i386/paravirt.h
index 690ada22437a..7df88be2dd9e 100644
--- a/include/asm-i386/paravirt.h
+++ b/include/asm-i386/paravirt.h
@@ -118,7 +118,7 @@ struct paravirt_ops
 
         u64 (*read_tsc)(void);
         u64 (*read_pmc)(void);
-        u64 (*get_scheduled_cycles)(void);
+        unsigned long long (*sched_clock)(void);
         unsigned long (*get_cpu_khz)(void);
 
         /* Segment descriptor handling */
@@ -566,7 +566,10 @@ static inline u64 paravirt_read_tsc(void)
 
 #define rdtscll(val) (val = paravirt_read_tsc())
 
-#define get_scheduled_cycles(val) (val = paravirt_ops.get_scheduled_cycles())
+static inline unsigned long long paravirt_sched_clock(void)
+{
+        return PVOP_CALL0(unsigned long long, sched_clock);
+}
 #define calculate_cpu_khz() (paravirt_ops.get_cpu_khz())
 
 #define write_tsc(val1,val2) wrmsr(0x10, val1, val2)
diff --git a/include/asm-i386/timer.h b/include/asm-i386/timer.h
index 153770e25faa..51a713e33a9e 100644
--- a/include/asm-i386/timer.h
+++ b/include/asm-i386/timer.h
@@ -15,8 +15,38 @@ extern int no_sync_cmos_clock;
 extern int recalibrate_cpu_khz(void);
 
 #ifndef CONFIG_PARAVIRT
-#define get_scheduled_cycles(val) rdtscll(val)
 #define calculate_cpu_khz() native_calculate_cpu_khz()
 #endif
 
+/* Accellerators for sched_clock()
+ * convert from cycles(64bits) => nanoseconds (64bits)
+ *  basic equation:
+ *              ns = cycles / (freq / ns_per_sec)
+ *              ns = cycles * (ns_per_sec / freq)
+ *              ns = cycles * (10^9 / (cpu_khz * 10^3))
+ *              ns = cycles * (10^6 / cpu_khz)
+ *
+ *      Then we use scaling math (suggested by george@mvista.com) to get:
+ *              ns = cycles * (10^6 * SC / cpu_khz) / SC
+ *              ns = cycles * cyc2ns_scale / SC
+ *
+ *      And since SC is a constant power of two, we can convert the div
+ *  into a shift.
+ *
+ *  We can use khz divisor instead of mhz to keep a better percision, since
+ *  cyc2ns_scale is limited to 10^6 * 2^10, which fits in 32 bits.
+ *  (mathieu.desnoyers@polymtl.ca)
+ *
+ *                      -johnstul@us.ibm.com "math is hard, lets go shopping!"
+ */
+extern unsigned long cyc2ns_scale __read_mostly;
+
+#define CYC2NS_SCALE_FACTOR 10 /* 2^10, carefully chosen */
+
+static inline unsigned long long cycles_2_ns(unsigned long long cyc)
+{
+        return (cyc * cyc2ns_scale) >> CYC2NS_SCALE_FACTOR;
+}
+
+
 #endif
diff --git a/include/asm-i386/vmi_time.h b/include/asm-i386/vmi_time.h
index 213930b995cb..478188130328 100644
--- a/include/asm-i386/vmi_time.h
+++ b/include/asm-i386/vmi_time.h
@@ -49,7 +49,7 @@ extern struct vmi_timer_ops {
 extern void __init vmi_time_init(void);
 extern unsigned long vmi_get_wallclock(void);
 extern int vmi_set_wallclock(unsigned long now);
-extern unsigned long long vmi_get_sched_cycles(void);
+extern unsigned long long vmi_sched_clock(void);
 extern unsigned long vmi_cpu_khz(void);
 
 #ifdef CONFIG_X86_LOCAL_APIC