Synchronize rdtsc instructions

The CPU can arbitrarily reorder calls to rdtsc, significantly
reducing the precision of timing using the CPUs time stamp counter.
Unfortunately the method of synchronizing rdtsc is different
on AMD and Intel CPUs. There is a generic method, using the cpuid
instruction, but that uses up a lot of registers, and is very slow.
Hence, use the correct lfence/mfence instructions (for CPUs that
we know support it)

Change-Id: I17ecb48d283f38f23148c13159aceda704c64ea5
Signed-off-by: Stefan Reinauer <reinauer@google.com>
Reviewed-on: http://review.coreboot.org/1422
Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
Tested-by: build bot (Jenkins)

src/cpu/amd/agesa/Kconfig

@@ -25,6 +25,7 @@ config CPU_AMD_AGESA
         default y if CPU_AMD_AGESA_FAMILY15
         default y if CPU_AMD_AGESA_FAMILY15_TN
         default n
+	select TSC_SYNC_LFENCE
 
 if CPU_AMD_AGESA
 

src/cpu/amd/model_10xxx/Kconfig

@@ -3,6 +3,7 @@ config CPU_AMD_MODEL_10XXX
 	select SSE
 	select SSE2
 	select MMCONF_SUPPORT_DEFAULT
+	select TSC_SYNC_LFENCE
 
 if CPU_AMD_MODEL_10XXX
 config CPU_ADDR_BITS

src/cpu/amd/model_fxx/Kconfig

@@ -3,6 +3,7 @@ config CPU_AMD_MODEL_FXX
 	select MMX
 	select SSE
 	select SSE2
+	select TSC_SYNC_LFENCE
 
 if CPU_AMD_MODEL_FXX
 config UDELAY_IO

src/cpu/intel/model_1067x/Kconfig

@@ -2,3 +2,4 @@ config CPU_INTEL_MODEL_1067X
 	bool
 	select SMP
 	select SSE2
+	select TSC_SYNC_MFENCE

src/cpu/intel/model_106cx/Kconfig

@@ -4,9 +4,9 @@ config CPU_INTEL_MODEL_106CX
 	select SSE2
 	select UDELAY_LAPIC
 	select AP_IN_SIPI_WAIT
+	select TSC_SYNC_MFENCE
 
 config CPU_ADDR_BITS
 	int
 	default 32
 
-

src/cpu/intel/model_206ax/Kconfig

@@ -14,6 +14,7 @@ config CPU_SPECIFIC_OPTIONS
 	select SMM_TSEG
 	select MICROCODE_IN_CBFS
 	#select AP_IN_SIPI_WAIT
+	select TSC_SYNC_MFENCE
 
 config BOOTBLOCK_CPU_INIT
 	string

src/cpu/intel/model_6ex/Kconfig

@@ -4,3 +4,4 @@ config CPU_INTEL_MODEL_6EX
 	select SSE2
 	select UDELAY_LAPIC
 	select AP_IN_SIPI_WAIT
+	select TSC_SYNC_MFENCE

src/cpu/intel/model_6fx/Kconfig

@@ -4,3 +4,4 @@ config CPU_INTEL_MODEL_6FX
 	select SSE2
 	select UDELAY_LAPIC
 	select AP_IN_SIPI_WAIT
+	select TSC_SYNC_MFENCE

src/cpu/x86/Kconfig

@@ -23,6 +23,22 @@ config TSC_CALIBRATE_WITH_IO
 	bool
 	default n
 
+config TSC_SYNC_LFENCE
+	bool
+	default n
+	help
+	  The CPU driver should select this if the CPU needs
+	  to execute an lfence instruction in order to synchronize
+	  rdtsc. This is true for all modern AMD CPUs.
+
+config TSC_SYNC_MFENCE
+	bool
+	default n
+	help
+	  The CPU driver should select this if the CPU needs
+	  to execute an mfence instruction in order to synchronize
+	  rdtsc. This is true for all modern Intel CPUs.
+
 config XIP_ROM_SIZE
 	hex
 	default ROM_SIZE if ROMCC

src/include/cpu/x86/tsc.h

@@ -1,6 +1,14 @@
 #ifndef CPU_X86_TSC_H
 #define CPU_X86_TSC_H
 
+#if CONFIG_TSC_SYNC_MFENCE
+#define TSC_SYNC "mfence\n"
+#elif CONFIG_TSC_SYNC_LFENCE
+#define TSC_SYNC "lfence\n"
+#else
+#define TSC_SYNC
+#endif
+
 struct tsc_struct {
 	unsigned lo;
 	unsigned hi;
@@ -10,10 +18,11 @@ typedef struct tsc_struct tsc_t;
 static inline tsc_t rdtsc(void)
 {
 	tsc_t res;
-	__asm__ __volatile__ (
+	asm volatile (
+		TSC_SYNC
 		"rdtsc"
 		: "=a" (res.lo), "=d"(res.hi) /* outputs */
-		);
+	);
 	return res;
 }
 
@@ -22,7 +31,11 @@ static inline tsc_t rdtsc(void)
 static inline unsigned long long rdtscll(void)
 {
 	unsigned long long val;
-	asm volatile ("rdtsc" : "=A" (val));
+	asm volatile (
+		TSC_SYNC
+		"rdtsc"
+		: "=A" (val)
+	);
 	return val;
 }
 #endif