Merge: CNB96: timekeeping: add clocksource ID and infrastructure for converting to/from a base clock

MR: https://gitlab.com/redhat/centos-stream/src/kernel/centos-stream-9/-/merge_requests/5380

JIRA: https://issues.redhat.com/browse/RHEL-61639  

First, the cross-timestamp mechanism which allows to correlate hardware clocks uses clocksource pointers for describing the correlation.

That's suboptimal as drivers need to obtain the pointer, which requires needless exports and exposing internals. This can be avoided by adding clocksource IDs to all clocksources in the tree which can be exposed to this mechanism.

Second, hardware time stamps like provided by PTP clock implementations are based on a clock which feeds both the PCIe device and the system clock. For further processing the underlying hardware clock timestamp must be converted to the system clock. Right now this requires drivers to invoke an architecture specific conversion function, e.g. to convert the ART (Always Running Timer) timestamp to a TSC timestamp.

This can be avoided by moving this functionality to the core code by providing a base clock property for the system clock which contains the conversion factors and assigning a clocksource ID to the base clock.

Commits:
```
5f10376 ("add missing includes and forward declarations to networking includes under linux/")
c55cbfc ("x86/tsc: Correct kernel-doc notation")
93630d6 ("timekeeping: Add clocksource ID to struct system_counterval_t")
a2c1fe7 ("x86/tsc: Add clocksource ID, set system_counterval_t.cs_id")
576bd49 ("x86/kvm, ptp/kvm: Add clocksource ID, set system_counterval_t.cs_id")
9be3b2f ("ptp/kvm, arm_arch_timer: Set system_counterval_t.cs_id to constant")
4b7f521 ("timekeeping: Evaluate system_counterval_t.cs_id instead of .cs")
b152688 ("treewide: Remove system_counterval_t.cs, which is never read")
27f6a9c ("kvmclock: Unexport kvmclock clocksource")
6b2e299 ("timekeeping: Provide infrastructure for converting to/from a base clock")
3a52886 ("x86/tsc: Provide ART base clock information for TSC")
bd48b50 ("e1000e: Replace convert_art_to_tsc()")
fcb0591 ("igc: Remove convert_art_ns_to_tsc()")
f5e1d0d ("stmmac: intel: Remove convert_art_to_tsc()")
b3266ed ("ALSA: hda: Remove convert_art_to_tsc()")
d4bea54 ("ice/ptp: Remove convert_art_to_tsc()")
0f532a7 ("x86/tsc: Remove obsolete ART to TSC conversion functions")
02ecee0 ("timekeeping: Add function to convert realtime to base clock")
```

Signed-off-by: Ivan Vecera <ivecera@redhat.com>

Approved-by: Phil Auld <pauld@redhat.com>
Approved-by: Tony Camuso <tcamuso@redhat.com>
Approved-by: Corinna Vinschen <vinschen@redhat.com>
Approved-by: Chris von Recklinghausen <crecklin@redhat.com>
Approved-by: Mark Salter <msalter@redhat.com>
Approved-by: Davide Caratti <dcaratti@redhat.com>
Approved-by: Prarit Bhargava <prarit@redhat.com>
Approved-by: CKI KWF Bot <cki-ci-bot+kwf-gitlab-com@redhat.com>

Merged-by: Rado Vrbovsky <rvrbovsk@redhat.com>

Author: emmil

arch/x86/include/asm/kvmclock.h

@@ -4,8 +4,6 @@
 
 #include <linux/percpu.h>
 
-extern struct clocksource kvm_clock;
-
 DECLARE_PER_CPU(struct pvclock_vsyscall_time_info *, hv_clock_per_cpu);
 
 static __always_inline struct pvclock_vcpu_time_info *this_cpu_pvti(void)

arch/x86/include/asm/tsc.h

@@ -27,9 +27,6 @@ static inline cycles_t get_cycles(void)
 }
 #define get_cycles get_cycles
 
-extern struct system_counterval_t convert_art_to_tsc(u64 art);
-extern struct system_counterval_t convert_art_ns_to_tsc(u64 art_ns);
-
 extern void tsc_early_init(void);
 extern void tsc_init(void);
 extern void mark_tsc_unstable(char *reason);

arch/x86/kernel/kvmclock.c

@@ -154,15 +154,15 @@ static int kvm_cs_enable(struct clocksource *cs)
 	return 0;
 }
 
-struct clocksource kvm_clock = {
+static struct clocksource kvm_clock = {
 	.name	= "kvm-clock",
 	.read	= kvm_clock_get_cycles,
 	.rating	= 400,
 	.mask	= CLOCKSOURCE_MASK(64),
 	.flags	= CLOCK_SOURCE_IS_CONTINUOUS,
+	.id     = CSID_X86_KVM_CLK,
 	.enable	= kvm_cs_enable,
 };
-EXPORT_SYMBOL_GPL(kvm_clock);
 
 static void kvm_register_clock(char *txt)
 {

arch/x86/kernel/tsc.c

@@ -50,10 +50,10 @@ int tsc_clocksource_reliable;
 
 static int __read_mostly tsc_force_recalibrate;
 
-static u32 art_to_tsc_numerator;
-static u32 art_to_tsc_denominator;
-static u64 art_to_tsc_offset;
-static struct clocksource *art_related_clocksource;
+static struct clocksource_base art_base_clk = {
+	.id    = CSID_X86_ART,
+};
+static bool have_art;
 
 struct cyc2ns {
 	struct cyc2ns_data data[2];	/*  0 + 2*16 = 32 */
@@ -652,7 +652,7 @@ static unsigned long quick_pit_calibrate(void)
 }
 
 /**
- * native_calibrate_tsc
+ * native_calibrate_tsc - determine TSC frequency
  * Determine TSC frequency via CPUID, else return 0.
  */
 unsigned long native_calibrate_tsc(void)
@@ -1074,7 +1074,7 @@ core_initcall(cpufreq_register_tsc_scaling);
  */
 static void __init detect_art(void)
 {
-	unsigned int unused[2];
+	unsigned int unused;
 
 	if (boot_cpu_data.cpuid_level < ART_CPUID_LEAF)
 		return;
@@ -1089,13 +1089,14 @@ static void __init detect_art(void)
 	    tsc_async_resets)
 		return;
 
-	cpuid(ART_CPUID_LEAF, &art_to_tsc_denominator,
-	      &art_to_tsc_numerator, unused, unused+1);
+	cpuid(ART_CPUID_LEAF, &art_base_clk.denominator,
+	      &art_base_clk.numerator, &art_base_clk.freq_khz, &unused);
 
-	if (art_to_tsc_denominator < ART_MIN_DENOMINATOR)
+	art_base_clk.freq_khz /= KHZ;
+	if (art_base_clk.denominator < ART_MIN_DENOMINATOR)
 		return;
 
-	rdmsrl(MSR_IA32_TSC_ADJUST, art_to_tsc_offset);
+	rdmsrl(MSR_IA32_TSC_ADJUST, art_base_clk.offset);
 
 	/* Make this sticky over multiple CPU init calls */
 	setup_force_cpu_cap(X86_FEATURE_ART);
@@ -1168,6 +1169,7 @@ static struct clocksource clocksource_tsc_early = {
 	.mask			= CLOCKSOURCE_MASK(64),
 	.flags			= CLOCK_SOURCE_IS_CONTINUOUS |
 				  CLOCK_SOURCE_MUST_VERIFY,
+	.id			= CSID_X86_TSC_EARLY,
 	.vdso_clock_mode	= VDSO_CLOCKMODE_TSC,
 	.enable			= tsc_cs_enable,
 	.resume			= tsc_resume,
@@ -1190,6 +1192,7 @@ static struct clocksource clocksource_tsc = {
 				  CLOCK_SOURCE_VALID_FOR_HRES |
 				  CLOCK_SOURCE_MUST_VERIFY |
 				  CLOCK_SOURCE_VERIFY_PERCPU,
+	.id			= CSID_X86_TSC,
 	.vdso_clock_mode	= VDSO_CLOCKMODE_TSC,
 	.enable			= tsc_cs_enable,
 	.resume			= tsc_resume,
@@ -1294,70 +1297,11 @@ int unsynchronized_tsc(void)
 	return 0;
 }
 
-/*
- * Convert ART to TSC given numerator/denominator found in detect_art()
- */
-struct system_counterval_t convert_art_to_tsc(u64 art)
-{
-	u64 tmp, res, rem;
-
-	rem = do_div(art, art_to_tsc_denominator);
-
-	res = art * art_to_tsc_numerator;
-	tmp = rem * art_to_tsc_numerator;
-
-	do_div(tmp, art_to_tsc_denominator);
-	res += tmp + art_to_tsc_offset;
-
-	return (struct system_counterval_t) {.cs = art_related_clocksource,
-			.cycles = res};
-}
-EXPORT_SYMBOL(convert_art_to_tsc);
-
-/**
- * convert_art_ns_to_tsc() - Convert ART in nanoseconds to TSC.
- * @art_ns: ART (Always Running Timer) in unit of nanoseconds
- *
- * PTM requires all timestamps to be in units of nanoseconds. When user
- * software requests a cross-timestamp, this function converts system timestamp
- * to TSC.
- *
- * This is valid when CPU feature flag X86_FEATURE_TSC_KNOWN_FREQ is set
- * indicating the tsc_khz is derived from CPUID[15H]. Drivers should check
- * that this flag is set before conversion to TSC is attempted.
- *
- * Return:
- * struct system_counterval_t - system counter value with the pointer to the
- *	corresponding clocksource
- *	@cycles:	System counter value
- *	@cs:		Clocksource corresponding to system counter value. Used
- *			by timekeeping code to verify comparability of two cycle
- *			values.
- */
-
-struct system_counterval_t convert_art_ns_to_tsc(u64 art_ns)
-{
-	u64 tmp, res, rem;
-
-	rem = do_div(art_ns, USEC_PER_SEC);
-
-	res = art_ns * tsc_khz;
-	tmp = rem * tsc_khz;
-
-	do_div(tmp, USEC_PER_SEC);
-	res += tmp;
-
-	return (struct system_counterval_t) { .cs = art_related_clocksource,
-					      .cycles = res};
-}
-EXPORT_SYMBOL(convert_art_ns_to_tsc);
-
-
 static void tsc_refine_calibration_work(struct work_struct *work);
 static DECLARE_DELAYED_WORK(tsc_irqwork, tsc_refine_calibration_work);
 /**
  * tsc_refine_calibration_work - Further refine tsc freq calibration
- * @work - ignored.
+ * @work: ignored.
  *
  * This functions uses delayed work over a period of a
  * second to further refine the TSC freq value. Since this is
@@ -1454,8 +1398,10 @@ static void tsc_refine_calibration_work(struct work_struct *work)
 	if (tsc_unstable)
 		goto unreg;
 
-	if (boot_cpu_has(X86_FEATURE_ART))
-		art_related_clocksource = &clocksource_tsc;
+	if (boot_cpu_has(X86_FEATURE_ART)) {
+		have_art = true;
+		clocksource_tsc.base = &art_base_clk;
+	}
 	clocksource_register_khz(&clocksource_tsc, tsc_khz);
 unreg:
 	clocksource_unregister(&clocksource_tsc_early);
@@ -1480,8 +1426,10 @@ static int __init init_tsc_clocksource(void)
 	 * the refined calibration and directly register it as a clocksource.
 	 */
 	if (boot_cpu_has(X86_FEATURE_TSC_KNOWN_FREQ)) {
-		if (boot_cpu_has(X86_FEATURE_ART))
-			art_related_clocksource = &clocksource_tsc;
+		if (boot_cpu_has(X86_FEATURE_ART)) {
+			have_art = true;
+			clocksource_tsc.base = &art_base_clk;
+		}
 		clocksource_register_khz(&clocksource_tsc, tsc_khz);
 		clocksource_unregister(&clocksource_tsc_early);
 
@@ -1505,10 +1453,12 @@ static bool __init determine_cpu_tsc_frequencies(bool early)
 
 	if (early) {
 		cpu_khz = x86_platform.calibrate_cpu();
-		if (tsc_early_khz)
+		if (tsc_early_khz) {
 			tsc_khz = tsc_early_khz;
-		else
+		} else {
 			tsc_khz = x86_platform.calibrate_tsc();
+			clocksource_tsc.freq_khz = tsc_khz;
+		}
 	} else {
 		/* We should not be here with non-native cpu calibration */
 		WARN_ON(x86_platform.calibrate_cpu != native_calibrate_cpu);

drivers/clocksource/arm_arch_timer.c

@@ -1777,7 +1777,7 @@ TIMER_ACPI_DECLARE(arch_timer, ACPI_SIG_GTDT, arch_timer_acpi_init);
 #endif
 
 int kvm_arch_ptp_get_crosststamp(u64 *cycle, struct timespec64 *ts,
-				 struct clocksource **cs)
+				 enum clocksource_ids *cs_id)
 {
 	struct arm_smccc_res hvc_res;
 	u32 ptp_counter;
@@ -1801,8 +1801,8 @@ int kvm_arch_ptp_get_crosststamp(u64 *cycle, struct timespec64 *ts,
 	*ts = ktime_to_timespec64(ktime);
 	if (cycle)
 		*cycle = (u64)hvc_res.a2 << 32 | hvc_res.a3;
-	if (cs)
-		*cs = &clocksource_counter;
+	if (cs_id)
+		*cs_id = CSID_ARM_ARCH_COUNTER;
 
 	return 0;
 }

drivers/net/ethernet/intel/e1000e/ptp.c

@@ -124,7 +124,8 @@ static int e1000e_phc_get_syncdevicetime(ktime_t *device,
 	sys_cycles = er32(PLTSTMPH);
 	sys_cycles <<= 32;
 	sys_cycles |= er32(PLTSTMPL);
-	*system = convert_art_to_tsc(sys_cycles);
+	system->cycles = sys_cycles;
+	system->cs_id = CSID_X86_ART;
 
 	return 0;
 }

drivers/net/ethernet/intel/ice/ice_ptp.c

@@ -2218,7 +2218,8 @@ ice_ptp_get_syncdevicetime(ktime_t *device,
 			hh_ts_lo = rd32(hw, GLHH_ART_TIME_L);
 			hh_ts_hi = rd32(hw, GLHH_ART_TIME_H);
 			hh_ts = ((u64)hh_ts_hi << 32) | hh_ts_lo;
-			*system = convert_art_ns_to_tsc(hh_ts);
+			system->cycles = hh_ts;
+			system->cs_id = CSID_X86_ART;
 			/* Read Device source clock time */
 			hh_ts_lo = rd32(hw, GLTSYN_HHTIME_L(tmr_idx));
 			hh_ts_hi = rd32(hw, GLTSYN_HHTIME_H(tmr_idx));

drivers/net/ethernet/intel/igc/igc_ptp.c

@@ -901,7 +901,11 @@ static bool igc_is_crosststamp_supported(struct igc_adapter *adapter)
 static struct system_counterval_t igc_device_tstamp_to_system(u64 tstamp)
 {
 #if IS_ENABLED(CONFIG_X86_TSC) && !defined(CONFIG_UML)
-	return convert_art_ns_to_tsc(tstamp);
+	return (struct system_counterval_t) {
+		.cs_id		= CSID_X86_ART,
+		.cycles		= tstamp,
+		.use_nsecs	= true,
+	};
 #else
 	return (struct system_counterval_t) { };
 #endif

drivers/net/ethernet/stmicro/stmmac/dwmac-intel.c

@@ -390,10 +390,11 @@ static int intel_crosststamp(ktime_t *device,
 		*device = ns_to_ktime(ptp_time);
 		read_unlock_irqrestore(&priv->ptp_lock, flags);
 		get_arttime(priv->mii, intel_priv->mdio_adhoc_addr, &art_time);
-		*system = convert_art_to_tsc(art_time);
+		system->cycles = art_time;
 	}
 
 	system->cycles *= intel_priv->crossts_adj;
+	system->cs_id = CSID_X86_ART;
 	priv->plat->flags &= ~STMMAC_FLAG_INT_SNAPSHOT_EN;
 
 	return 0;

drivers/ptp/ptp_kvm_common.c

@@ -28,15 +28,15 @@ static int ptp_kvm_get_time_fn(ktime_t *device_time,
 			       struct system_counterval_t *system_counter,
 			       void *ctx)
 {
-	long ret;
-	u64 cycle;
+	enum clocksource_ids cs_id;
 	struct timespec64 tspec;
-	struct clocksource *cs;
+	u64 cycle;
+	int ret;
 
 	spin_lock(&kvm_ptp_lock);
 
 	preempt_disable_notrace();
-	ret = kvm_arch_ptp_get_crosststamp(&cycle, &tspec, &cs);
+	ret = kvm_arch_ptp_get_crosststamp(&cycle, &tspec, &cs_id);
 	if (ret) {
 		spin_unlock(&kvm_ptp_lock);
 		preempt_enable_notrace();
@@ -46,7 +46,7 @@ static int ptp_kvm_get_time_fn(ktime_t *device_time,
 	preempt_enable_notrace();
 
 	system_counter->cycles = cycle;
-	system_counter->cs = cs;
+	system_counter->cs_id = cs_id;
 
 	*device_time = timespec64_to_ktime(tspec);