x86: Move tsc_calibration to x86_init_ops

TSC calibration is modified by the vmware hypervisor and paravirt by
separate means. Moorestown wants to add its own calibration routine as
well. So make calibrate_tsc a proper x86_init_ops function and
override it by paravirt or by the early setup of the vmware
hypervisor.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

arch/x86/include/asm/hypervisor.h

@@ -20,7 +20,7 @@
 #ifndef ASM_X86__HYPERVISOR_H
 #define ASM_X86__HYPERVISOR_H
 
-extern unsigned long get_hypervisor_tsc_freq(void);
 extern void init_hypervisor(struct cpuinfo_x86 *c);
+extern void init_hypervisor_platform(void);
 
 #endif

arch/x86/include/asm/paravirt.h

@@ -210,7 +210,6 @@ static inline unsigned long long paravirt_sched_clock(void)
 {
 	return PVOP_CALL0(unsigned long long, pv_time_ops.sched_clock);
 }
-#define calibrate_tsc() (pv_time_ops.get_tsc_khz())
 
 static inline unsigned long long paravirt_read_pmc(int counter)
 {

arch/x86/include/asm/timer.h

@@ -8,7 +8,6 @@
 #define TICK_SIZE (tick_nsec / 1000)
 
 unsigned long long native_sched_clock(void);
-unsigned long native_calibrate_tsc(void);
 extern int recalibrate_cpu_khz(void);
 
 #if defined(CONFIG_X86_32) && defined(CONFIG_X86_IO_APIC)
@@ -19,10 +18,6 @@ extern int timer_ack;
 
 extern int no_timer_check;
 
-#ifndef CONFIG_PARAVIRT
-#define calibrate_tsc() native_calibrate_tsc()
-#endif
-
 /* Accelerators for sched_clock()
  * convert from cycles(64bits) => nanoseconds (64bits)
  *  basic equation:

arch/x86/include/asm/tsc.h

@@ -48,7 +48,8 @@ static __always_inline cycles_t vget_cycles(void)
 extern void tsc_init(void);
 extern void mark_tsc_unstable(char *reason);
 extern int unsynchronized_tsc(void);
-int check_tsc_unstable(void);
+extern int check_tsc_unstable(void);
+extern unsigned long native_calibrate_tsc(void);
 
 /*
  * Boot-time check whether the TSCs are synchronized across

arch/x86/include/asm/vmware.h

@@ -20,7 +20,7 @@
 #ifndef ASM_X86__VMWARE_H
 #define ASM_X86__VMWARE_H
 
-extern unsigned long vmware_get_tsc_khz(void);
+extern void vmware_platform_setup(void);
 extern int vmware_platform(void);
 extern void vmware_set_feature_bits(struct cpuinfo_x86 *c);
 

arch/x86/include/asm/x86_init.h

@@ -112,8 +112,17 @@ struct x86_cpuinit_ops {
 	void (*setup_percpu_clockev)(void);
 };
 
+/**
+ * struct x86_platform_ops - platform specific runtime functions
+ * @calibrate_tsc:		calibrate TSC
+ */
+struct x86_platform_ops {
+	unsigned long (*calibrate_tsc)(void);
+};
+
 extern struct x86_init_ops x86_init;
 extern struct x86_cpuinit_ops x86_cpuinit;
+extern struct x86_platform_ops x86_platform;
 
 extern void x86_init_noop(void);
 extern void x86_init_uint_noop(unsigned int unused);

arch/x86/kernel/cpu/hypervisor.c

@@ -34,13 +34,6 @@ detect_hypervisor_vendor(struct cpuinfo_x86 *c)
 		c->x86_hyper_vendor = X86_HYPER_VENDOR_NONE;
 }
 
-unsigned long get_hypervisor_tsc_freq(void)
-{
-	if (boot_cpu_data.x86_hyper_vendor == X86_HYPER_VENDOR_VMWARE)
-		return vmware_get_tsc_khz();
-	return 0;
-}
-
 static inline void __cpuinit
 hypervisor_set_feature_bits(struct cpuinfo_x86 *c)
 {
@@ -55,3 +48,10 @@ void __cpuinit init_hypervisor(struct cpuinfo_x86 *c)
 	detect_hypervisor_vendor(c);
 	hypervisor_set_feature_bits(c);
 }
+
+void __init init_hypervisor_platform(void)
+{
+	init_hypervisor(&boot_cpu_data);
+	if (boot_cpu_data.x86_hyper_vendor == X86_HYPER_VENDOR_VMWARE)
+		vmware_platform_setup();
+}

arch/x86/kernel/cpu/vmware.c

@@ -24,6 +24,7 @@
 #include <linux/dmi.h>
 #include <asm/div64.h>
 #include <asm/vmware.h>
+#include <asm/x86_init.h>
 
 #define CPUID_VMWARE_INFO_LEAF	0x40000000
 #define VMWARE_HYPERVISOR_MAGIC	0x564D5868
@@ -47,21 +48,29 @@ static inline int __vmware_platform(void)
 	return eax != (uint32_t)-1 && ebx == VMWARE_HYPERVISOR_MAGIC;
 }
 
-static unsigned long __vmware_get_tsc_khz(void)
+static unsigned long vmware_get_tsc_khz(void)
 {
 	uint64_t tsc_hz;
 	uint32_t eax, ebx, ecx, edx;
 
 	VMWARE_PORT(GETHZ, eax, ebx, ecx, edx);
 
-	if (ebx == UINT_MAX)
-		return 0;
 	tsc_hz = eax | (((uint64_t)ebx) << 32);
 	do_div(tsc_hz, 1000);
 	BUG_ON(tsc_hz >> 32);
 	return tsc_hz;
 }
 
+void __init vmware_platform_setup(void)
+{
+	uint32_t eax, ebx, ecx, edx;
+
+	VMWARE_PORT(GETHZ, eax, ebx, ecx, edx);
+
+	if (ebx != UINT_MAX)
+		x86_platform.calibrate_tsc = vmware_get_tsc_khz;
+}
+
 /*
  * While checking the dmi string infomation, just checking the product
  * serial key should be enough, as this will always have a VMware
@@ -87,12 +96,6 @@ int vmware_platform(void)
 	return 0;
 }
 
-unsigned long vmware_get_tsc_khz(void)
-{
-	BUG_ON(!vmware_platform());
-	return __vmware_get_tsc_khz();
-}
-
 /*
  * VMware hypervisor takes care of exporting a reliable TSC to the guest.
  * Still, due to timing difference when running on virtual cpus, the TSC can

arch/x86/kernel/kvmclock.c

@@ -187,7 +187,7 @@ void __init kvmclock_init(void)
 		pv_time_ops.get_wallclock = kvm_get_wallclock;
 		pv_time_ops.set_wallclock = kvm_set_wallclock;
 		pv_time_ops.sched_clock = kvm_clock_read;
-		pv_time_ops.get_tsc_khz = kvm_get_tsc_khz;
+		x86_platform.calibrate_tsc = kvm_get_tsc_khz;
 #ifdef CONFIG_X86_LOCAL_APIC
 		x86_cpuinit.setup_percpu_clockev =
 			kvm_setup_secondary_clock;

arch/x86/kernel/paravirt.c

@@ -309,7 +309,6 @@ struct pv_time_ops pv_time_ops = {
 	.get_wallclock = native_get_wallclock,
 	.set_wallclock = native_set_wallclock,
 	.sched_clock = native_sched_clock,
-	.get_tsc_khz = native_calibrate_tsc,
 };
 
 struct pv_irq_ops pv_irq_ops = {

arch/x86/kernel/setup.c

@@ -818,7 +818,7 @@ void __init setup_arch(char **cmdline_p)
 	 * VMware detection requires dmi to be available, so this
 	 * needs to be done after dmi_scan_machine, for the BP.
 	 */
-	init_hypervisor(&boot_cpu_data);
+	init_hypervisor_platform();
 
 	x86_init.resources.probe_roms();
 

arch/x86/kernel/tsc.c

@@ -18,6 +18,7 @@
 #include <asm/delay.h>
 #include <asm/hypervisor.h>
 #include <asm/nmi.h>
+#include <asm/x86_init.h>
 
 unsigned int __read_mostly cpu_khz;	/* TSC clocks / usec, not used here */
 EXPORT_SYMBOL(cpu_khz);
@@ -401,15 +402,9 @@ unsigned long native_calibrate_tsc(void)
 {
 	u64 tsc1, tsc2, delta, ref1, ref2;
 	unsigned long tsc_pit_min = ULONG_MAX, tsc_ref_min = ULONG_MAX;
-	unsigned long flags, latch, ms, fast_calibrate, hv_tsc_khz;
+	unsigned long flags, latch, ms, fast_calibrate;
 	int hpet = is_hpet_enabled(), i, loopmin;
 
-	hv_tsc_khz = get_hypervisor_tsc_freq();
-	if (hv_tsc_khz) {
-		printk(KERN_INFO "TSC: Frequency read from the hypervisor\n");
-		return hv_tsc_khz;
-	}
-
 	local_irq_save(flags);
 	fast_calibrate = quick_pit_calibrate();
 	local_irq_restore(flags);
@@ -567,7 +562,7 @@ int recalibrate_cpu_khz(void)
 	unsigned long cpu_khz_old = cpu_khz;
 
 	if (cpu_has_tsc) {
-		tsc_khz = calibrate_tsc();
+		tsc_khz = x86_platform.calibrate_tsc();
 		cpu_khz = tsc_khz;
 		cpu_data(0).loops_per_jiffy =
 			cpufreq_scale(cpu_data(0).loops_per_jiffy,
@@ -917,7 +912,7 @@ void __init tsc_init(void)
 	if (!cpu_has_tsc)
 		return;
 
-	tsc_khz = calibrate_tsc();
+	tsc_khz = x86_platform.calibrate_tsc();
 	cpu_khz = tsc_khz;
 
 	if (!tsc_khz) {

arch/x86/kernel/vmi_32.c

@@ -825,7 +825,7 @@ static inline int __init activate_vmi(void)
 		x86_cpuinit.setup_percpu_clockev = vmi_time_ap_init;
 #endif
 		pv_time_ops.sched_clock = vmi_sched_clock;
-		pv_time_ops.get_tsc_khz = vmi_tsc_khz;
+		x86_platform.calibrate_tsc = vmi_tsc_khz;
 
 		/* We have true wallclock functions; disable CMOS clock sync */
 		no_sync_cmos_clock = 1;

arch/x86/kernel/vmiclock_32.c

@@ -68,7 +68,7 @@ unsigned long long vmi_sched_clock(void)
 	return cycles_2_ns(vmi_timer_ops.get_cycle_counter(VMI_CYCLES_AVAILABLE));
 }
 
-/* paravirt_ops.get_tsc_khz = vmi_tsc_khz */
+/* x86_platform.calibrate_tsc = vmi_tsc_khz */
 unsigned long vmi_tsc_khz(void)
 {
 	unsigned long long khz;

arch/x86/kernel/x86_init.c

@@ -13,6 +13,7 @@
 #include <asm/e820.h>
 #include <asm/time.h>
 #include <asm/irq.h>
+#include <asm/tsc.h>
 
 void __cpuinit x86_init_noop(void) { }
 void __init x86_init_uint_noop(unsigned int unused) { }
@@ -67,3 +68,7 @@ struct __initdata x86_init_ops x86_init = {
 __cpuinitdata struct x86_cpuinit_ops x86_cpuinit = {
 	.setup_percpu_clockev		= setup_secondary_APIC_clock,
 };
+
+struct x86_platform_ops x86_platform = {
+	.calibrate_tsc			= native_calibrate_tsc,
+};

arch/x86/lguest/boot.c

@@ -1320,11 +1320,11 @@ __init void lguest_init(void)
 
 	/* Time operations */
 	pv_time_ops.get_wallclock = lguest_get_wallclock;
-	pv_time_ops.get_tsc_khz = lguest_tsc_khz;
 
 	x86_init.resources.memory_setup = lguest_memory_setup;
 	x86_init.irqs.intr_init = lguest_init_IRQ;
 	x86_init.timers.timer_init = lguest_time_init;
+	x86_platform.calibrate_tsc = lguest_tsc_khz;
 
 	/*
 	 * Now is a good time to look at the implementations of these functions

arch/x86/xen/enlighten.c

@@ -844,7 +844,6 @@ static const struct pv_init_ops xen_init_ops __initdata = {
 static const struct pv_time_ops xen_time_ops __initdata = {
 	.set_wallclock = xen_set_wallclock,
 	.get_wallclock = xen_get_wallclock,
-	.get_tsc_khz = xen_tsc_khz,
 	.sched_clock = xen_sched_clock,
 };
 
@@ -980,6 +979,8 @@ asmlinkage void __init xen_start_kernel(void)
 	x86_init.timers.setup_percpu_clockev = x86_init_noop;
 	x86_cpuinit.setup_percpu_clockev = x86_init_noop;
 
+	x86_platform.calibrate_tsc = xen_tsc_khz;
+
 #ifdef CONFIG_X86_64
 	/*
 	 * Setup percpu state.  We only need to do this for 64-bit