[patch V2 00/25] timekeeping/ptp: Expand snapshot functionality

[patch V2 00/25] timekeeping/ptp: Expand snapshot functionality

[Thomas Gleixner]

This is an update to V1 which can be found here:

   https://lore.kernel.org/lkml/20260526165826.392227559@kernel.org

PTP wants to grow new snapshot functionality, which provides not only the
captured CLOCK* values, but also the underlying clocksource counter value.

   https://lore.kernel.org/20260515164033.6403-1-akiyano@amazon.com

There was quite some discussion in seemingly related threads how to capture
these values and how to provide core infrastructure so that driver writers
have something to work with

   https://lore.kernel.org/20260514225842.110706-1-hramamurthy@google.com
   https://lore.kernel.org/20260520135207.37826-1-dwmw2@infradead.org

This series implements the timekeeping related mechanisms to:

     1) Capture CLOCK values along with the clocksource counter value for
     	non-hardware based sampling

     2) Expanding the hardware cross time stamp mechanism to hand back the
     	clocksource counter value, which was captured by the device, along
     	with the related CLOCK values

     3) Adding AUX clock support to the hardware cross timestamping core

     4) Add support for derived clocksources to the snapshot mechanism (New
     	in V2)

Changes vs. V1:

  - Fixed the ptp_ocp typo - 0-day, Jakub

  - Renamed the system_time_snapshot members sys and raw so systime and
    monoraw to make them less ambigous.

  - Fixed the error case return values of get_device_system_crosststamp()

  - Made ktime_snapshot_id() void as there is no point for the return
    value, which is nowhere checked and cannot be propagated.
    system_time_snapshot::valid has to be evaluated at the call sites
    anyway. - Jacob

  - Picked up the first patch from Davids follow up series, which extends
    the snapshot mechanism so that derived clocksources (like kvmclock and
    Hyper-V scaled TSC) can return the actual underlying hardware counter
    value (TSC for the two examples).

  - Collected Reviewed/Acked/Tested-by tags

Delta patch against v1 below.

The series is based on v7.1-rc2 and also available from git:

    git://git.kernel.org/pub/scm/linux/kernel/git/tglx/devel.git timekeeping-ptp-extend-v2

Thanks,

	tglx
---
diff --git a/arch/arm64/kvm/hyp_trace.c b/arch/arm64/kvm/hyp_trace.c
index 616062587510..b056c652ff01 100644
--- a/arch/arm64/kvm/hyp_trace.c
+++ b/arch/arm64/kvm/hyp_trace.c
@@ -51,8 +51,8 @@ static void __hyp_clock_work(struct work_struct *work)
 
 	hyp_clock = container_of(dwork, struct hyp_trace_clock, work);
 
-	ktime_get_snapshot_id(&snap, CLOCK_BOOTTIME);
-	boot = ktime_to_ns(snap.sys);
+	ktime_get_snapshot_id(CLOCK_BOOTTIME, &snap);
+	boot = ktime_to_ns(snap.systime);
 
 	delta_boot = boot - hyp_clock->boot;
 	delta_cycles = snap.cycles - hyp_clock->cycles;
@@ -120,7 +120,7 @@ static void hyp_trace_clock_enable(struct hyp_trace_clock *hyp_clock, bool enabl
 
 	ktime_get_snapshot_id(&snap, CLOCK_BOOTTIME);
 
-	hyp_clock->boot = ktime_to_ns(snap.sys);
+	hyp_clock->boot = ktime_to_ns(snap.systime);
 	hyp_clock->cycles = snap.cycles;
 	hyp_clock->mult = 0;
 
diff --git a/arch/arm64/kvm/hypercalls.c b/arch/arm64/kvm/hypercalls.c
index e60cc7ed3e70..b11b8821c9fb 100644
--- a/arch/arm64/kvm/hypercalls.c
+++ b/arch/arm64/kvm/hypercalls.c
@@ -28,7 +28,7 @@ static void kvm_ptp_get_time(struct kvm_vcpu *vcpu, u64 *val)
 	 * system time and counter value must captured at the same
 	 * time to keep consistency and precision.
 	 */
-	ktime_get_snapshot_id(&systime_snapshot, CLOCK_REALTIME);
+	ktime_get_snapshot_id(CLOCK_REALTIME, &systime_snapshot);
 
 	/*
 	 * This is only valid if the current clocksource is the
@@ -61,8 +61,8 @@ static void kvm_ptp_get_time(struct kvm_vcpu *vcpu, u64 *val)
 	 * in the future (about 292 years from 1970, and at that stage
 	 * nobody will give a damn about it).
 	 */
-	val[0] = upper_32_bits(systime_snapshot.sys);
-	val[1] = lower_32_bits(systime_snapshot.sys);
+	val[0] = upper_32_bits(systime_snapshot.systime);
+	val[1] = lower_32_bits(systime_snapshot.systime);
 	val[2] = upper_32_bits(cycles);
 	val[3] = lower_32_bits(cycles);
 }
diff --git a/drivers/net/dsa/sja1105/sja1105_main.c b/drivers/net/dsa/sja1105/sja1105_main.c
index 1d5fef4df560..2697073dbf90 100644
--- a/drivers/net/dsa/sja1105/sja1105_main.c
+++ b/drivers/net/dsa/sja1105/sja1105_main.c
@@ -2310,10 +2310,10 @@ int sja1105_static_config_reload(struct sja1105_private *priv,
 		goto out;
 	}
 
-	t1 = ktime_to_ns(ptp_sts_before.pre_sts.sys);
-	t2 = ktime_to_ns(ptp_sts_before.post_sts.sys);
-	t3 = ktime_to_ns(ptp_sts_after.pre_sts.sys);
-	t4 = ktime_to_ns(ptp_sts_after.post_sts.sys);
+	t1 = ktime_to_ns(ptp_sts_before.pre_sts.systime);
+	t2 = ktime_to_ns(ptp_sts_before.post_sts.systime);
+	t3 = ktime_to_ns(ptp_sts_after.pre_sts.systime);
+	t4 = ktime_to_ns(ptp_sts_after.post_sts.systime);
 	/* Mid point, corresponds to pre-reset PTPCLKVAL */
 	t12 = t1 + (t2 - t1) / 2;
 	/* Mid point, corresponds to post-reset PTPCLKVAL, aka 0 */
diff --git a/drivers/net/ethernet/intel/ice/ice_ptp.c b/drivers/net/ethernet/intel/ice/ice_ptp.c
index 5023fc1587f9..f9e4ec6f7ebb 100644
--- a/drivers/net/ethernet/intel/ice/ice_ptp.c
+++ b/drivers/net/ethernet/intel/ice/ice_ptp.c
@@ -2117,7 +2117,7 @@ static int ice_capture_crosststamp(ktime_t *device,
 	}
 
 	/* Snapshot system time for historic interpolation */
-	ktime_get_snapshot_id(&ctx->snapshot, ctx->snapshot_clock_id);
+	ktime_get_snapshot_id(ctx->snapshot_clock_id, &ctx->snapshot);
 
 	/* Program cmd to master timer */
 	ice_ptp_src_cmd(hw, ICE_PTP_READ_TIME);
diff --git a/drivers/net/ethernet/intel/igc/igc_ptp.c b/drivers/net/ethernet/intel/igc/igc_ptp.c
index 9b8b4a04e32d..b40aba9ab685 100644
--- a/drivers/net/ethernet/intel/igc/igc_ptp.c
+++ b/drivers/net/ethernet/intel/igc/igc_ptp.c
@@ -1049,7 +1049,7 @@ static int igc_phc_get_syncdevicetime(ktime_t *device,
 	 */
 	do {
 		/* Get a snapshot of system clocks to use as historic value. */
-		ktime_get_snapshot_id(&adapter->snapshot, adapter->snapshot_clock_id);
+		ktime_get_snapshot_id(adapter->snapshot_clock_id, &adapter->snapshot);
 
 		igc_ptm_trigger(hw);
 
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
index beb80912b9d5..5df786133e4b 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
@@ -340,7 +340,7 @@ static int mlx5_ptp_getcrosststamp(struct ptp_clock_info *ptp,
 		goto unlock;
 	}
 
-	ktime_get_snapshot_id(&history_begin, cts->clock_id);
+	ktime_get_snapshot_id(cts->clock_id, &history_begin);
 
 	err = get_device_system_crosststamp(mlx5_mtctr_syncdevicetime, mdev,
 					    &history_begin, cts);
@@ -366,7 +366,7 @@ static int mlx5_ptp_getcrosscycles(struct ptp_clock_info *ptp,
 		goto unlock;
 	}
 
-	ktime_get_snapshot_id(&history_begin, cts->clock_id);
+	ktime_get_snapshot_id(cts->clock_id, &history_begin);
 
 	err = get_device_system_crosststamp(mlx5_mtctr_syncdevicecyclestime,
 					    mdev, &history_begin, cts);
diff --git a/drivers/ptp/ptp_chardev.c b/drivers/ptp/ptp_chardev.c
index aed5c13cd1be..dc23cd708cfe 100644
--- a/drivers/ptp/ptp_chardev.c
+++ b/drivers/ptp/ptp_chardev.c
@@ -392,11 +392,11 @@ static long ptp_sys_offset_extended(struct ptp_clock *ptp, void __user *arg,
 		extoff->ts[i][1].sec = ts.tv_sec;
 		extoff->ts[i][1].nsec = ts.tv_nsec;
 
-		ts = ktime_to_timespec64(sts.pre_sts.sys);
+		ts = ktime_to_timespec64(sts.pre_sts.systime);
 		extoff->ts[i][0].sec = ts.tv_sec;
 		extoff->ts[i][0].nsec = ts.tv_nsec;
 
-		ts = ktime_to_timespec64(sts.post_sts.sys);
+		ts = ktime_to_timespec64(sts.post_sts.systime);
 		extoff->ts[i][2].sec = ts.tv_sec;
 		extoff->ts[i][2].nsec = ts.tv_nsec;
 	}
diff --git a/drivers/ptp/ptp_ocp.c b/drivers/ptp/ptp_ocp.c
index b7a23936a44d..28b0302c6250 100644
--- a/drivers/ptp/ptp_ocp.c
+++ b/drivers/ptp/ptp_ocp.c
@@ -1492,7 +1492,7 @@ __ptp_ocp_gettime_locked(struct ptp_ocp *bp, struct timespec64 *ts,
 	ptp_read_system_postts(sts);
 
 	if (sts && bp->ts_window_adjust)
-		sts->post_ts.sys -= bp->ts_window_adjust;
+		sts->post_sts.systime -= bp->ts_window_adjust;
 
 	time_ns = ioread32(&bp->reg->time_ns);
 	time_sec = ioread32(&bp->reg->time_sec);
@@ -4592,8 +4592,8 @@ ptp_ocp_summary_show(struct seq_file *s, void *data)
 		struct timespec64 sys_ts;
 		s64 pre_ns, post_ns, ns;
 
-		pre_ns = ktime_to_ns(sts.pre_sts.sys);
-		post_ns = ktime_to_ns(sts.post_sts.sys);
+		pre_ns = ktime_to_ns(sts.pre_sts.systime);
+		post_ns = ktime_to_ns(sts.post_sts.systime);
 		ns = (pre_ns + post_ns) / 2;
 		ns += (s64)bp->utc_tai_offset * NSEC_PER_SEC;
 		sys_ts = ns_to_timespec64(ns);
diff --git a/drivers/ptp/ptp_vmclock.c b/drivers/ptp/ptp_vmclock.c
index cb18c15a4697..d6a5a533164a 100644
--- a/drivers/ptp/ptp_vmclock.c
+++ b/drivers/ptp/ptp_vmclock.c
@@ -263,7 +263,7 @@ static int ptp_vmclock_getcrosststamp(struct ptp_clock_info *ptp,
 	if (ret == -ENODEV) {
 		struct system_time_snapshot systime_snapshot;
 
-		ktime_get_snapshot_id(&systime_snapshot, CLOCK_REALTIME);
+		ktime_get_snapshot_id(CLOCK_REALTIME, &systime_snapshot);
 
 		if (systime_snapshot.cs_id == CSID_X86_TSC ||
 		    systime_snapshot.cs_id == CSID_X86_KVM_CLK) {
diff --git a/drivers/virtio/virtio_rtc_ptp.c b/drivers/virtio/virtio_rtc_ptp.c
index e15d00aeb01d..ff8d834493dc 100644
--- a/drivers/virtio/virtio_rtc_ptp.c
+++ b/drivers/virtio/virtio_rtc_ptp.c
@@ -139,7 +139,7 @@ static int viortc_ptp_getcrosststamp(struct ptp_clock_info *ptp,
 	if (ret)
 		return ret;
 
-	ktime_get_snapshot_id(&history_begin, xtstamp->clock_id);
+	ktime_get_snapshot_id(xtstamp->clock_id, &history_begin);
 	if (history_begin.cs_id != cs_id)
 		return -EOPNOTSUPP;
 
diff --git a/include/linux/clocksource.h b/include/linux/clocksource.h
index 7c38190b10bf..6d9ddf1587a2 100644
--- a/include/linux/clocksource.h
+++ b/include/linux/clocksource.h
@@ -31,6 +31,21 @@ struct module;
 
 #include <vdso/clocksource.h>
 
+/**
+ * struct clocksource_hw_snapshot - Snapshot for the underlying hardware counter of derived
+ *				    clocksources like kvmclock or Hyper-V scaled TSC
+ * @hw_cycles:		The hardware counter value
+ * @hw_csid:		Clocksource ID of the hardware counter
+ *
+ * Such clocksources must implement the read_snapshot() callback and fill in the
+ * hardware counter value, the clocksource ID of the hardware counter and derive
+ * the actual clocksource cycles from @hw_cycles to provide an atomic snapshot
+ */
+struct clocksource_hw_snapshot {
+	u64			hw_cycles;
+	enum clocksource_ids	hw_csid;
+};
+
 /**
  * struct clocksource - hardware abstraction for a free running counter
  *	Provides mostly state-free accessors to the underlying hardware.
@@ -72,6 +87,14 @@ struct module;
  * @flags:		Flags describing special properties
  * @base:		Hardware abstraction for clock on which a clocksource
  *			is based
+ * @read_snapshot:	Extended @read() function for clocksources such as
+ *			kvmclock or the Hyper-V scaled TSC where the actual
+ *			clocksource value for timekeeping is calculated from an
+ *			underlying hardware counter. Returns the timekeeping
+ *			relevant cycle value and stores the raw value of the
+ *			underlying counter from which it was calculated
+ *			including the clocksource ID of that counter in the
+ *			clocksource hardware snapshot.
  * @enable:		Optional function to enable the clocksource
  * @disable:		Optional function to disable the clocksource
  * @suspend:		Optional suspend function for the clocksource
@@ -113,6 +136,7 @@ struct clocksource {
 	unsigned long		flags;
 	struct clocksource_base *base;
 
+	u64			(*read_snapshot)(struct clocksource *cs, struct clocksource_hw_snapshot *chs);
 	int			(*enable)(struct clocksource *cs);
 	void			(*disable)(struct clocksource *cs);
 	void			(*suspend)(struct clocksource *cs);
diff --git a/include/linux/pps_kernel.h b/include/linux/pps_kernel.h
index cd80f1cb96a9..9f088c9023b1 100644
--- a/include/linux/pps_kernel.h
+++ b/include/linux/pps_kernel.h
@@ -102,9 +102,9 @@ static inline void pps_get_ts(struct pps_event_time *ts)
 #ifdef CONFIG_NTP_PPS
 	struct system_time_snapshot snap;
 
-	ktime_get_snapshot_id(&snap, CLOCK_REALTIME);
-	ts->ts_real = ktime_to_timespec64(snap.sys);
-	ts->ts_raw = ktime_to_timespec64(snap.raw);
+	ktime_get_snapshot_id(CLOCK_REALTIME, &snap);
+	ts->ts_real = ktime_to_timespec64(snap.systime);
+	ts->ts_raw = ktime_to_timespec64(snap.monoraw);
 #else
 	ktime_get_real_ts64(&ts->ts_real);
 #endif
diff --git a/include/linux/ptp_clock_kernel.h b/include/linux/ptp_clock_kernel.h
index df6c9aac458b..36a27a910595 100644
--- a/include/linux/ptp_clock_kernel.h
+++ b/include/linux/ptp_clock_kernel.h
@@ -511,13 +511,13 @@ static inline ktime_t ptp_convert_timestamp(const ktime_t *hwtstamp,
 static inline void ptp_read_system_prets(struct ptp_system_timestamp *sts)
 {
 	if (sts)
-		ktime_get_snapshot_id(&sts->pre_sts, sts->clockid);
+		ktime_get_snapshot_id(sts->clockid, &sts->pre_sts);
 }
 
 static inline void ptp_read_system_postts(struct ptp_system_timestamp *sts)
 {
 	if (sts)
-		ktime_get_snapshot_id(&sts->post_sts, sts->clockid);
+		ktime_get_snapshot_id(sts->clockid, &sts->post_sts);
 }
 
 #endif
diff --git a/include/linux/timekeeping.h b/include/linux/timekeeping.h
index f7945f1048fc..984a866d293b 100644
--- a/include/linux/timekeeping.h
+++ b/include/linux/timekeeping.h
@@ -279,18 +279,24 @@ static inline bool ktime_get_aux_ts64(clockid_t id, struct timespec64 *kt) { ret
  * struct system_time_snapshot - Simultaneous time capture of CLOCK_MONOTONIC_RAW,
  *				 a selected CLOCK_* and the clocksource counter value
  * @cycles:		Clocksource counter value to produce the system times
- * @sys:		The system time of the selected CLOCK ID
- * @raw:		Monotonic raw system time
+ * @hw_cycles:		For derived clocksources, the hardware counter value from
+ *			which @cycles was derived
+ * @systime:		The system time of the selected CLOCK ID
+ * @monoraw:		Monotonic raw system time
  * @cs_id:		Clocksource ID
+ * @hw_csid:		Clocksource ID of the underlying hardware counter for derived
+ *			clocksources which implement the read_snapshot() callback.
  * @clock_was_set_seq:	The sequence number of clock-was-set events
  * @cs_was_changed_seq:	The sequence number of clocksource change events
  * @valid:		True if the snapshot is valid
  */
 struct system_time_snapshot {
 	u64			cycles;
-	ktime_t			sys;
-	ktime_t			raw;
+	u64			hw_cycles;
+	ktime_t			systime;
+	ktime_t			monoraw;
 	enum clocksource_ids	cs_id;
+	enum clocksource_ids	hw_csid;
 	unsigned int		clock_was_set_seq;
 	u8			cs_was_changed_seq;
 	u8			valid;
@@ -348,8 +354,7 @@ extern int get_device_system_crosststamp(
  * Simultaneously snapshot a given clock with MONOTONIC_RAW and the underlying
  * clocksource counter value.
  */
-extern bool ktime_get_snapshot_id(struct system_time_snapshot *systime_snapshot,
-				  clockid_t clock_id);
+extern void ktime_get_snapshot_id(clockid_t clock_id, struct system_time_snapshot *systime_snapshot);
 
 /*
  * Persistent clock related interfaces
diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c
index c4fd7229b7da..0d5b67f609bb 100644
--- a/kernel/time/timekeeping.c
+++ b/kernel/time/timekeeping.c
@@ -320,6 +320,7 @@ static __always_inline u64 tk_clock_read(const struct tk_read_base *tkr)
 
 	return clock->read(clock);
 }
+
 static inline void clocksource_disable_inline_read(void) { }
 static inline void clocksource_enable_inline_read(void) { }
 #endif
@@ -1187,14 +1188,26 @@ noinstr time64_t __ktime_get_real_seconds(void)
 	return tk->xtime_sec;
 }
 
+static inline u64 tk_clock_read_snapshot(const struct tk_read_base *tkr,
+					 struct clocksource_hw_snapshot *chs)
+{
+	struct clocksource *clock = READ_ONCE(tkr->clock);
+
+	if (unlikely(clock->read_snapshot))
+		return clock->read_snapshot(clock, chs);
+
+	return clock->read(clock);
+}
+
+
 /**
  * ktime_get_snapshot_id -  Simultaneously snapshot a given clock ID with
  *			    CLOCK_MONOTONIC_RAW and the underlying
  *			    clocksource counter value.
- * @systime_snapshot:	Pointer to struct receiving the system time snapshot
  * @clock_id:		The clock ID to snapshot
+ * @systime_snapshot:	Pointer to struct receiving the system time snapshot
  */
-bool ktime_get_snapshot_id(struct system_time_snapshot *systime_snapshot, clockid_t clock_id)
+void ktime_get_snapshot_id(clockid_t clock_id, struct system_time_snapshot *systime_snapshot)
 {
 	ktime_t base_raw, base_sys, offs_sys, *offs, offs_zero = 0;
 	u64 nsec_raw, nsec_sys, now;
@@ -1206,7 +1219,7 @@ bool ktime_get_snapshot_id(struct system_time_snapshot *systime_snapshot, clocki
 	systime_snapshot->valid = false;
 
 	if (WARN_ON_ONCE(timekeeping_suspended))
-		return false;
+		return;
 
 	switch (clock_id) {
 	case CLOCK_REALTIME:
@@ -1226,25 +1239,31 @@ bool ktime_get_snapshot_id(struct system_time_snapshot *systime_snapshot, clocki
 	case CLOCK_AUX ... CLOCK_AUX_LAST:
 		tkd = aux_get_tk_data(clock_id);
 		if (!tkd)
-			return false;
+			return;
 		offs = &tkd->timekeeper.offs_aux;
 		break;
 	default:
 		WARN_ON_ONCE(1);
-		return false;
+		return;
 	}
 
 	tk = &tkd->timekeeper;
 
 	do {
+		struct clocksource_hw_snapshot chs = { };
+
 		seq = read_seqcount_begin(&tkd->seq);
 
 		/* Aux clocks can be invalid */
 		if (!tk->clock_valid)
-			return false;
+			return;
 
-		now = tk_clock_read(&tk->tkr_mono);
+		now = tk_clock_read_snapshot(&tk->tkr_mono, &chs);
 		systime_snapshot->cs_id = tk->tkr_mono.clock->id;
+
+		systime_snapshot->hw_cycles = chs.hw_cycles;
+		systime_snapshot->hw_csid = chs.hw_csid;
+
 		systime_snapshot->cs_was_changed_seq = tk->cs_was_changed_seq;
 		systime_snapshot->clock_was_set_seq = tk->clock_was_set_seq;
 
@@ -1257,18 +1276,17 @@ bool ktime_get_snapshot_id(struct system_time_snapshot *systime_snapshot, clocki
 	} while (read_seqcount_retry(&tkd->seq, seq));
 
 	systime_snapshot->cycles = now;
-	systime_snapshot->sys = ktime_add_ns(base_sys, offs_sys + nsec_sys);
-	systime_snapshot->raw = ktime_add_ns(base_raw, nsec_raw);
+	systime_snapshot->systime = ktime_add_ns(base_sys, offs_sys + nsec_sys);
+	systime_snapshot->monoraw = ktime_add_ns(base_raw, nsec_raw);
 
 	/*
 	 * Special case for PTP. Just transfer the raw time into sys,
-	 * so the call sites can consistently use snap::sys.
+	 * so the call sites can consistently use snap::systime.
 	 */
 	if (clock_id == CLOCK_MONOTONIC_RAW)
-		systime_snapshot->sys = systime_snapshot->raw;
+		systime_snapshot->systime = systime_snapshot->monoraw;
 	/* Tell the consumer that this snapshot is valid */
 	systime_snapshot->valid = true;
-	return true;
 }
 EXPORT_SYMBOL_GPL(ktime_get_snapshot_id);
 
@@ -1330,7 +1348,7 @@ static int adjust_historical_crosststamp(struct system_time_snapshot *history,
 	 * Scale the monotonic raw time delta by:
 	 *	partial_history_cycles / total_history_cycles
 	 */
-	corr_raw = (u64)ktime_to_ns(ktime_sub(ts->sys_monoraw, history->raw));
+	corr_raw = (u64)ktime_to_ns(ktime_sub(ts->sys_monoraw, history->monoraw));
 	ret = scale64_check_overflow(partial_history_cycles,
 				     total_history_cycles, &corr_raw);
 	if (ret)
@@ -1347,7 +1365,7 @@ static int adjust_historical_crosststamp(struct system_time_snapshot *history,
 	if (discontinuity) {
 		corr_sys = mul_u64_u32_div(corr_raw, tk->tkr_mono.mult, tk->tkr_raw.mult);
 	} else {
-		corr_sys = (u64)ktime_to_ns(ktime_sub(ts->sys_systime, history->sys));
+		corr_sys = (u64)ktime_to_ns(ktime_sub(ts->sys_systime, history->systime));
 		ret = scale64_check_overflow(partial_history_cycles, total_history_cycles,
 					     &corr_sys);
 		if (ret)
@@ -1356,8 +1374,8 @@ static int adjust_historical_crosststamp(struct system_time_snapshot *history,
 
 	/* Fixup monotonic raw and system time time values */
 	if (interp_forward) {
-		ts->sys_monoraw = ktime_add_ns(history->raw, corr_raw);
-		ts->sys_systime = ktime_add_ns(history->sys, corr_sys);
+		ts->sys_monoraw = ktime_add_ns(history->monoraw, corr_raw);
+		ts->sys_systime = ktime_add_ns(history->systime, corr_sys);
 	} else {
 		ts->sys_monoraw = ktime_sub_ns(ts->sys_monoraw, corr_raw);
 		ts->sys_systime = ktime_sub_ns(ts->sys_systime, corr_sys);
@@ -1521,12 +1539,12 @@ int get_device_system_crosststamp(int (*get_time_fn)
 	case CLOCK_AUX ... CLOCK_AUX_LAST:
 		tkd = aux_get_tk_data(xtstamp->clock_id);
 		if (!tkd)
-			return false;
+			return -ENODEV;
 		offs = &tkd->timekeeper.offs_aux;
 		break;
 	default:
 		WARN_ON_ONCE(1);
-		return false;
+		return -ENODEV;
 	}
 
 	tk = &tkd->timekeeper;

[patch V2 01/25] timekeeping: Provide ktime_get_snapshot_id()

[Thomas Gleixner]

From: Thomas Gleixner <tglx@kernel.org>

ktime_get_snapshot() provides a snapshot of the underlying clocksource
counter value and the corresponding CLOCK_MONOTONIC_RAW, CLOCK_REALTIME and
CLOCK_BOOTTIME timestamps.

There is no usage of CLOCK_REALTIME and CLOCK_BOOTTIME at the same time and
CLOCK_BOOTTIME support was just added for the ARM64 KVM tracing mechanism,
which needs CLOCK_BOOTTIME and the underlying clocksource counter value.

ktime_get_snapshot() is also not suitable for usage with CLOCK_AUX, but
that's a prerequisite to support PTP hardware timestamping for CLOCK_AUX
steering.

As a first step, rename ktime_get_snapshot() to ktime_get_snapshot_id(),
which now takes a clockid argument to select the clock which needs to be
captured. The result is stored in system_time_snapshot::systime, which will
replace the system_time_snapshot::real/boot members once all usage sites
have been converted.

ktime_get_snapshot() is a simple wrapper which hands in CLOCK_REALTIME as
clockid argument for the conversion period. That means CLOCK_REALTIME is
now captured twice, but that redunancy is only temporary.

As all usage sites of struct system_time_snapshot has to be updated anyway,
rename the 'raw' member to 'monoraw' for clarity.

No functional change vs. current users of ktime_get_snapshot()

Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Tested-by: David Woodhouse <dwmw@amazon.co.uk>
Tested-by: Arthur Kiyanovski <akiyano@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Thomas Weißschuh <thomas.weissschuh@linutronix.de>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
---
 include/linux/timekeeping.h |   34 ++++++++++++-----
 kernel/time/timekeeping.c   |   85 +++++++++++++++++++++++++++++++++-----------
 2 files changed, 88 insertions(+), 31 deletions(-)
--- a/include/linux/timekeeping.h
+++ b/include/linux/timekeeping.h
@@ -276,24 +276,32 @@ static inline bool ktime_get_aux_ts64(cl
 #endif
 
 /**
- * struct system_time_snapshot - simultaneous raw/real time capture with
- *				 counter value
- * @cycles:	Clocksource counter value to produce the system times
- * @real:	Realtime system time
- * @boot:	Boot time
- * @raw:	Monotonic raw system time
- * @cs_id:	Clocksource ID
+ * struct system_time_snapshot - Simultaneous time capture of CLOCK_MONOTONIC_RAW,
+ *				 a selected CLOCK_* and the clocksource counter value
+ * @cycles:		Clocksource counter value to produce the system times
+ * @systime:		The system time of the selected CLOCK ID
+ * @real:		Realtime system time
+ * @boot:		Boot time
+ * @raw:		Monotonic raw system time
+ * @monoraw:		Monotonic raw system time
+ * @cs_id:		Clocksource ID
  * @clock_was_set_seq:	The sequence number of clock-was-set events
  * @cs_was_changed_seq:	The sequence number of clocksource change events
+ * @valid:		True if the snapshot is valid
  */
 struct system_time_snapshot {
 	u64			cycles;
+	ktime_t			systime;
 	ktime_t			real;
 	ktime_t			boot;
-	ktime_t			raw;
+	union {
+		ktime_t		raw;
+		ktime_t		monoraw;
+	};
 	enum clocksource_ids	cs_id;
 	unsigned int		clock_was_set_seq;
 	u8			cs_was_changed_seq;
+	u8			valid;
 };
 
 /**
@@ -341,9 +349,15 @@ extern int get_device_system_crosststamp
 			struct system_device_crosststamp *xtstamp);
 
 /*
- * Simultaneously snapshot realtime and monotonic raw clocks
+ * Simultaneously snapshot a given clock with MONOTONIC_RAW and the underlying
+ * clocksource counter value.
  */
-extern void ktime_get_snapshot(struct system_time_snapshot *systime_snapshot);
+extern void ktime_get_snapshot_id(clockid_t clock_id, struct system_time_snapshot *systime_snapshot);
+
+static inline void ktime_get_snapshot(struct system_time_snapshot *systime_snapshot)
+{
+	ktime_get_snapshot_id(CLOCK_REALTIME, systime_snapshot);
+}
 
 /*
  * Persistent clock related interfaces
--- a/kernel/time/timekeeping.c
+++ b/kernel/time/timekeeping.c
@@ -1183,43 +1183,86 @@ noinstr time64_t __ktime_get_real_second
 }
 
 /**
- * ktime_get_snapshot - snapshots the realtime/monotonic raw clocks with counter
- * @systime_snapshot:	pointer to struct receiving the system time snapshot
+ * ktime_get_snapshot_id -  Simultaneously snapshot a given clock ID with
+ *			    CLOCK_MONOTONIC_RAW and the underlying
+ *			    clocksource counter value.
+ * @clock_id:		The clock ID to snapshot
+ * @systime_snapshot:	Pointer to struct receiving the system time snapshot
  */
-void ktime_get_snapshot(struct system_time_snapshot *systime_snapshot)
+void ktime_get_snapshot_id(clockid_t clock_id, struct system_time_snapshot *systime_snapshot)
 {
-	struct timekeeper *tk = &tk_core.timekeeper;
+	ktime_t base_raw, base_sys, offs_sys, *offs, offs_zero = 0;
+	u64 nsec_raw, nsec_sys, now;
+	struct timekeeper *tk;
+	struct tk_data *tkd;
 	unsigned int seq;
-	ktime_t base_raw;
 	ktime_t base_real;
 	ktime_t base_boot;
-	u64 nsec_raw;
-	u64 nsec_real;
-	u64 now;
 
-	WARN_ON_ONCE(timekeeping_suspended);
+	/* Invalidate the snapshot for all failure cases */
+	systime_snapshot->valid = false;
+
+	if (WARN_ON_ONCE(timekeeping_suspended))
+		return;
+
+	switch (clock_id) {
+	case CLOCK_REALTIME:
+		tkd = &tk_core;
+		offs = &tk_core.timekeeper.offs_real;
+		break;
+	/* Map RAW to MONOTONIC so the loop below is trivial */
+	case CLOCK_MONOTONIC_RAW:
+	case CLOCK_MONOTONIC:
+		tkd = &tk_core;
+		offs = &offs_zero;
+		break;
+	case CLOCK_BOOTTIME:
+		tkd = &tk_core;
+		offs = &tk_core.timekeeper.offs_boot;
+		break;
+	default:
+		WARN_ON_ONCE(1);
+		return;
+	}
+
+	tk = &tkd->timekeeper;
 
 	do {
-		seq = read_seqcount_begin(&tk_core.seq);
+		seq = read_seqcount_begin(&tkd->seq);
+
 		now = tk_clock_read(&tk->tkr_mono);
 		systime_snapshot->cs_id = tk->tkr_mono.clock->id;
 		systime_snapshot->cs_was_changed_seq = tk->cs_was_changed_seq;
 		systime_snapshot->clock_was_set_seq = tk->clock_was_set_seq;
-		base_real = ktime_add(tk->tkr_mono.base,
-				      tk_core.timekeeper.offs_real);
-		base_boot = ktime_add(tk->tkr_mono.base,
-				      tk_core.timekeeper.offs_boot);
+
+		base_sys = tk->tkr_mono.base;
+		offs_sys = *offs;
 		base_raw = tk->tkr_raw.base;
-		nsec_real = timekeeping_cycles_to_ns(&tk->tkr_mono, now);
-		nsec_raw  = timekeeping_cycles_to_ns(&tk->tkr_raw, now);
-	} while (read_seqcount_retry(&tk_core.seq, seq));
+
+		/* Kept around until the callers are fixed up */
+		base_real = ktime_add(base_sys, tk_core.timekeeper.offs_real);
+		base_boot = ktime_add(base_sys, tk_core.timekeeper.offs_boot);
+
+		nsec_sys = timekeeping_cycles_to_ns(&tk->tkr_mono, now);
+		nsec_raw = timekeeping_cycles_to_ns(&tk->tkr_raw, now);
+	} while (read_seqcount_retry(&tkd->seq, seq));
 
 	systime_snapshot->cycles = now;
-	systime_snapshot->real = ktime_add_ns(base_real, nsec_real);
-	systime_snapshot->boot = ktime_add_ns(base_boot, nsec_real);
-	systime_snapshot->raw = ktime_add_ns(base_raw, nsec_raw);
+	systime_snapshot->systime = ktime_add_ns(base_sys, offs_sys + nsec_sys);
+	systime_snapshot->real = ktime_add_ns(base_real, nsec_sys);
+	systime_snapshot->boot = ktime_add_ns(base_boot, nsec_sys);
+	systime_snapshot->monoraw = ktime_add_ns(base_raw, nsec_raw);
+
+	/*
+	 * Special case for PTP. Just transfer the raw time into sys,
+	 * so the call sites can consistently use snap::systime.
+	 */
+	if (clock_id == CLOCK_MONOTONIC_RAW)
+		systime_snapshot->systime = systime_snapshot->monoraw;
+	/* Tell the consumer that this snapshot is valid */
+	systime_snapshot->valid = true;
 }
-EXPORT_SYMBOL_GPL(ktime_get_snapshot);
+EXPORT_SYMBOL_GPL(ktime_get_snapshot_id);
 
 /* Scale base by mult/div checking for overflow */
 static int scale64_check_overflow(u64 mult, u64 div, u64 *base)

[patch V2 02/25] timekeeping: Use system_time_snapshot::systime/monoraw instead of ::real/raw

[Thomas Gleixner]

From: Thomas Gleixner <tglx@kernel.org>

system_time_snapshot::systime provides the same information as
system_time_snapshot::real when the snapshot was taken with
ktime_get_snapshot_id(CLOCK_REALTIME).

Convert the history interpolation over to use 'systime' and 'monoraw' as
'real/raw' are going away once all users are converted.

As a side effect this is the first step to support CLOCK_AUX with
get_device_crosstime_stamp() and the history interpolation.

Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Tested-by: David Woodhouse <dwmw@amazon.co.uk>
Tested-by: Arthur Kiyanovski <akiyano@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Thomas Weißschuh <thomas.weissschuh@linutronix.de>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
---
 kernel/time/timekeeping.c |    8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
--- a/kernel/time/timekeeping.c
+++ b/kernel/time/timekeeping.c
@@ -1323,7 +1323,7 @@ static int adjust_historical_crosststamp
 	 *	partial_history_cycles / total_history_cycles
 	 */
 	corr_raw = (u64)ktime_to_ns(
-		ktime_sub(ts->sys_monoraw, history->raw));
+		ktime_sub(ts->sys_monoraw, history->monoraw));
 	ret = scale64_check_overflow(partial_history_cycles,
 				     total_history_cycles, &corr_raw);
 	if (ret)
@@ -1341,7 +1341,7 @@ static int adjust_historical_crosststamp
 			(corr_raw, tk->tkr_mono.mult, tk->tkr_raw.mult);
 	} else {
 		corr_real = (u64)ktime_to_ns(
-			ktime_sub(ts->sys_realtime, history->real));
+			ktime_sub(ts->sys_realtime, history->systime));
 		ret = scale64_check_overflow(partial_history_cycles,
 					     total_history_cycles, &corr_real);
 		if (ret)
@@ -1350,8 +1350,8 @@ static int adjust_historical_crosststamp
 
 	/* Fixup monotonic raw and real time time values */
 	if (interp_forward) {
-		ts->sys_monoraw = ktime_add_ns(history->raw, corr_raw);
-		ts->sys_realtime = ktime_add_ns(history->real, corr_real);
+		ts->sys_monoraw = ktime_add_ns(history->monoraw, corr_raw);
+		ts->sys_realtime = ktime_add_ns(history->systime, corr_real);
 	} else {
 		ts->sys_monoraw = ktime_sub_ns(ts->sys_monoraw, corr_raw);
 		ts->sys_realtime = ktime_sub_ns(ts->sys_realtime, corr_real);

[patch V2 03/25] pps: generators: Use ktime_get_real_ts64() instead of ktime_get_snapshot()

[Thomas Gleixner]

From: Thomas Gleixner <tglx@kernel.org>

There is no reason to use the more complex ktime_get_snapshot() for
retrieving CLOCK_REALTIME.

Just use ktime_get_real_ts64(), which avoids the extra timespec64
conversion as a bonus.

No functional change intended.

Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Tested-by: Arthur Kiyanovski <akiyano@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Thomas Weißschuh <thomas.weissschuh@linutronix.de>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
---
 drivers/pps/generators/pps_gen-dummy.c |    6 +-----
 drivers/pps/generators/pps_gen_tio.c   |    6 +-----
 2 files changed, 2 insertions(+), 10 deletions(-)
--- a/drivers/pps/generators/pps_gen-dummy.c
+++ b/drivers/pps/generators/pps_gen-dummy.c
@@ -39,11 +39,7 @@ static void pps_gen_ktimer_event(struct
 static int pps_gen_dummy_get_time(struct pps_gen_device *pps_gen,
 					struct timespec64 *time)
 {
-	struct system_time_snapshot snap;
-
-	ktime_get_snapshot(&snap);
-	*time = ktime_to_timespec64(snap.real);
-
+	ktime_get_real_ts64(time);
 	return 0;
 }
 
--- a/drivers/pps/generators/pps_gen_tio.c
+++ b/drivers/pps/generators/pps_gen_tio.c
@@ -189,11 +189,7 @@ static int pps_tio_gen_enable(struct pps
 static int pps_tio_get_time(struct pps_gen_device *pps_gen,
 			    struct timespec64 *time)
 {
-	struct system_time_snapshot snap;
-
-	ktime_get_snapshot(&snap);
-	*time = ktime_to_timespec64(snap.real);
-
+	ktime_get_real_ts64(time);
 	return 0;
 }
 

[patch V2 04/25] pps: Convert to ktime_get_snapshot_id()

[Thomas Gleixner]

From: Thomas Gleixner <tglx@kernel.org>

ktime_get_snapshot() resolves to ktime_get_snapshot_id(CLOCK_REALTIME).

Make it obvious in the code and convert the readout to use the
snapshot::systime and monoraw fields instead of snapshot::real and raw,
which aregoing away.

Similar to the PPS generators, avoid the more expensive snapshot when
CONFIG_NTP_PPS is disabled.

No functional change intended.

Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Tested-by: Arthur Kiyanovski <akiyano@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Thomas Weißschuh <thomas.weissschuh@linutronix.de>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
---
 include/linux/pps_kernel.h |   10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)
--- a/include/linux/pps_kernel.h
+++ b/include/linux/pps_kernel.h
@@ -99,12 +99,14 @@ static inline void timespec_to_pps_ktime
 
 static inline void pps_get_ts(struct pps_event_time *ts)
 {
+#ifdef CONFIG_NTP_PPS
 	struct system_time_snapshot snap;
 
-	ktime_get_snapshot(&snap);
-	ts->ts_real = ktime_to_timespec64(snap.real);
-#ifdef CONFIG_NTP_PPS
-	ts->ts_raw = ktime_to_timespec64(snap.raw);
+	ktime_get_snapshot_id(CLOCK_REALTIME, &snap);
+	ts->ts_real = ktime_to_timespec64(snap.systime);
+	ts->ts_raw = ktime_to_timespec64(snap.monoraw);
+#else
+	ktime_get_real_ts64(&ts->ts_real);
 #endif
 }
 

[patch V2 05/25] KVM: arm64: Use ktime_get_snapshot_id() to retrieve CLOCK_BOOTTIME

[Thomas Gleixner]

From: Thomas Gleixner <tglx@kernel.org>

ktime_get_snapshot() is replaced by ktime_get_snapshot_id() which allows to
request a particular CLOCK ID to be captured along with the clocksource
counter.

Convert the tracing mechanism over and use the new
system_time_snapshot::systime field, which holds the system timestamp
selected by the CLOCK ID argument.

No functional change intended.

Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Tested-by: Arthur Kiyanovski <akiyano@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Thomas Weißschuh <thomas.weissschuh@linutronix.de>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Acked-by: Marc Zyngier <maz@kernel.org>
Acked-by: Vincent Donnefort <vdonnefort@google.com>
---
 arch/arm64/kvm/hyp_trace.c |    8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
--- a/arch/arm64/kvm/hyp_trace.c
+++ b/arch/arm64/kvm/hyp_trace.c
@@ -51,8 +51,8 @@ static void __hyp_clock_work(struct work
 
 	hyp_clock = container_of(dwork, struct hyp_trace_clock, work);
 
-	ktime_get_snapshot(&snap);
-	boot = ktime_to_ns(snap.boot);
+	ktime_get_snapshot_id(CLOCK_BOOTTIME, &snap);
+	boot = ktime_to_ns(snap.systime);
 
 	delta_boot = boot - hyp_clock->boot;
 	delta_cycles = snap.cycles - hyp_clock->cycles;
@@ -118,9 +118,9 @@ static void hyp_trace_clock_enable(struc
 		hyp_clock->running = false;
 	}
 
-	ktime_get_snapshot(&snap);
+	ktime_get_snapshot_id(&snap, CLOCK_BOOTTIME);
 
-	hyp_clock->boot = ktime_to_ns(snap.boot);
+	hyp_clock->boot = ktime_to_ns(snap.systime);
 	hyp_clock->cycles = snap.cycles;
 	hyp_clock->mult = 0;
 

[patch V2 06/25] KVM: arm64: Use ktime_get_snapshot_id() to snapshot CLOCK_REALTIME

[Thomas Gleixner]

From: Thomas Gleixner <tglx@kernel.org>

ktime_get_snapshot() is replaced by ktime_get_snapshot_id() which allows to
request a particular CLOCK ID to be captured along with the clocksource
counter.

Convert the usage in kvm_get_ptp_time() over and use the new
system_time_snapshot::systime field, which holds the system timestamp
selected by the CLOCK ID argument.

No functional change intended.

Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Tested-by: Arthur Kiyanovski <akiyano@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Thomas Weißschuh <thomas.weissschuh@linutronix.de>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Acked-by: Marc Zyngier <maz@kernel.org>
---
 arch/arm64/kvm/hypercalls.c |    6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
--- a/arch/arm64/kvm/hypercalls.c
+++ b/arch/arm64/kvm/hypercalls.c
@@ -28,7 +28,7 @@ static void kvm_ptp_get_time(struct kvm_
 	 * system time and counter value must captured at the same
 	 * time to keep consistency and precision.
 	 */
-	ktime_get_snapshot(&systime_snapshot);
+	ktime_get_snapshot_id(CLOCK_REALTIME, &systime_snapshot);
 
 	/*
 	 * This is only valid if the current clocksource is the
@@ -61,8 +61,8 @@ static void kvm_ptp_get_time(struct kvm_
 	 * in the future (about 292 years from 1970, and at that stage
 	 * nobody will give a damn about it).
 	 */
-	val[0] = upper_32_bits(systime_snapshot.real);
-	val[1] = lower_32_bits(systime_snapshot.real);
+	val[0] = upper_32_bits(systime_snapshot.systime);
+	val[1] = lower_32_bits(systime_snapshot.systime);
 	val[2] = upper_32_bits(cycles);
 	val[3] = lower_32_bits(cycles);
 }

[patch V2 07/25] ptp: ptp_vmclock: Convert to ktime_get_snapshot_id()

[Thomas Gleixner]

From: Thomas Gleixner <tglx@kernel.org>

ktime_get_snapshot() is replaced by ktime_get_snapshot_id() which allows to
request a particular CLOCK ID to be captured along with the clocksource
counter.

Convert vmclock over and use the new system_time_snapshot::systime field,
which holds the system timestamp selected by the CLOCK ID argument.

No functional change intended.

Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Tested-by: David Woodhouse <dwmw@amazon.co.uk>
Tested-by: Arthur Kiyanovski <akiyano@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Thomas Weißschuh <thomas.weissschuh@linutronix.de>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
---
 drivers/ptp/ptp_vmclock.c |    8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
--- a/drivers/ptp/ptp_vmclock.c
+++ b/drivers/ptp/ptp_vmclock.c
@@ -132,7 +132,7 @@ static int vmclock_get_crosststamp(struc
 		 * will be derived from the *same* counter value.
 		 *
 		 * If the system isn't using the same counter, then the value
-		 * from ktime_get_snapshot() will still be used as pre_ts, and
+		 * from ktime_get_snapshot_id() will still be used as pre_ts, and
 		 * ptp_read_system_postts() is called to populate postts after
 		 * calling get_cycles().
 		 *
@@ -140,7 +140,7 @@ static int vmclock_get_crosststamp(struc
 		 * the seq_count loop.
 		 */
 		if (sts) {
-			ktime_get_snapshot(&systime_snapshot);
+			ktime_get_snapshot_id(CLOCK_REALTIME, &systime_snapshot);
 			if (systime_snapshot.cs_id == st->cs_id) {
 				cycle = systime_snapshot.cycles;
 			} else {
@@ -181,7 +181,7 @@ static int vmclock_get_crosststamp(struc
 	}
 
 	if (sts) {
-		sts->pre_ts = ktime_to_timespec64(systime_snapshot.real);
+		sts->pre_ts = ktime_to_timespec64(systime_snapshot.systime);
 		if (systime_snapshot.cs_id == st->cs_id)
 			sts->post_ts = sts->pre_ts;
 	}
@@ -272,7 +272,7 @@ static int ptp_vmclock_getcrosststamp(st
 	if (ret == -ENODEV) {
 		struct system_time_snapshot systime_snapshot;
 
-		ktime_get_snapshot(&systime_snapshot);
+		ktime_get_snapshot_id(CLOCK_REALTIME, &systime_snapshot);
 
 		if (systime_snapshot.cs_id == CSID_X86_TSC ||
 		    systime_snapshot.cs_id == CSID_X86_KVM_CLK) {

[patch V2 08/25] timekeeping: Remove system_time_snapshot::real/boot/raw

[Thomas Gleixner]

From: Thomas Gleixner <tglx@kernel.org>

All users are converted over to ktime_get_snapshot_id() and
system_time_snapshot::systime and ::monoraw.

Remove the leftovers.

Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Tested-by: David Woodhouse <dwmw@amazon.co.uk>
Tested-by: Arthur Kiyanovski <akiyano@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Thomas Weißschuh <thomas.weissschuh@linutronix.de>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
---
 include/linux/timekeeping.h |   10 +---------
 kernel/time/timekeeping.c   |    8 --------
 2 files changed, 1 insertion(+), 17 deletions(-)
--- a/include/linux/timekeeping.h
+++ b/include/linux/timekeeping.h
@@ -280,9 +280,6 @@ static inline bool ktime_get_aux_ts64(cl
  *				 a selected CLOCK_* and the clocksource counter value
  * @cycles:		Clocksource counter value to produce the system times
  * @systime:		The system time of the selected CLOCK ID
- * @real:		Realtime system time
- * @boot:		Boot time
- * @raw:		Monotonic raw system time
  * @monoraw:		Monotonic raw system time
  * @cs_id:		Clocksource ID
  * @clock_was_set_seq:	The sequence number of clock-was-set events
@@ -292,12 +289,7 @@ static inline bool ktime_get_aux_ts64(cl
 struct system_time_snapshot {
 	u64			cycles;
 	ktime_t			systime;
-	ktime_t			real;
-	ktime_t			boot;
-	union {
-		ktime_t		raw;
-		ktime_t		monoraw;
-	};
+	ktime_t			monoraw;
 	enum clocksource_ids	cs_id;
 	unsigned int		clock_was_set_seq;
 	u8			cs_was_changed_seq;
--- a/kernel/time/timekeeping.c
+++ b/kernel/time/timekeeping.c
@@ -1196,8 +1196,6 @@ void ktime_get_snapshot_id(clockid_t clo
 	struct timekeeper *tk;
 	struct tk_data *tkd;
 	unsigned int seq;
-	ktime_t base_real;
-	ktime_t base_boot;
 
 	/* Invalidate the snapshot for all failure cases */
 	systime_snapshot->valid = false;
@@ -1239,18 +1237,12 @@ void ktime_get_snapshot_id(clockid_t clo
 		offs_sys = *offs;
 		base_raw = tk->tkr_raw.base;
 
-		/* Kept around until the callers are fixed up */
-		base_real = ktime_add(base_sys, tk_core.timekeeper.offs_real);
-		base_boot = ktime_add(base_sys, tk_core.timekeeper.offs_boot);
-
 		nsec_sys = timekeeping_cycles_to_ns(&tk->tkr_mono, now);
 		nsec_raw = timekeeping_cycles_to_ns(&tk->tkr_raw, now);
 	} while (read_seqcount_retry(&tkd->seq, seq));
 
 	systime_snapshot->cycles = now;
 	systime_snapshot->systime = ktime_add_ns(base_sys, offs_sys + nsec_sys);
-	systime_snapshot->real = ktime_add_ns(base_real, nsec_sys);
-	systime_snapshot->boot = ktime_add_ns(base_boot, nsec_sys);
 	systime_snapshot->monoraw = ktime_add_ns(base_raw, nsec_raw);
 
 	/*

[patch V2 09/25] timekeeping: Add CLOCK_AUX support for ktime_get_snapshot_id()

[Thomas Gleixner]

From: Thomas Gleixner <tglx@kernel.org>

Now that all users are converted it's possible to enable snapshotting of
CLOCK_AUX time. The underlying clocksource is the same as for all other
CLOCK variants.

Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Tested-by: Arthur Kiyanovski <akiyano@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Thomas Weißschuh <thomas.weissschuh@linutronix.de>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
---
 kernel/time/timekeeping.c |   15 +++++++++++++++
 1 file changed, 15 insertions(+)
--- a/kernel/time/timekeeping.c
+++ b/kernel/time/timekeeping.c
@@ -67,6 +67,7 @@ static inline bool tk_is_aux(const struc
 {
 	return tk->id >= TIMEKEEPER_AUX_FIRST && tk->id <= TIMEKEEPER_AUX_LAST;
 }
+static inline struct tk_data *aux_get_tk_data(clockid_t id);
 #else
 static inline bool tk_get_aux_ts64(unsigned int tkid, struct timespec64 *ts)
 {
@@ -77,6 +78,10 @@ static inline bool tk_is_aux(const struc
 {
 	return false;
 }
+static inline struct tk_data *aux_get_tk_data(clockid_t id)
+{
+	return NULL;
+}
 #endif
 
 static inline void tk_update_aux_offs(struct timekeeper *tk, ktime_t offs)
@@ -1218,6 +1223,12 @@ void ktime_get_snapshot_id(clockid_t clo
 		tkd = &tk_core;
 		offs = &tk_core.timekeeper.offs_boot;
 		break;
+	case CLOCK_AUX ... CLOCK_AUX_LAST:
+		tkd = aux_get_tk_data(clock_id);
+		if (!tkd)
+			return;
+		offs = &tkd->timekeeper.offs_aux;
+		break;
 	default:
 		WARN_ON_ONCE(1);
 		return;
@@ -1228,6 +1239,10 @@ void ktime_get_snapshot_id(clockid_t clo
 	do {
 		seq = read_seqcount_begin(&tkd->seq);
 
+		/* Aux clocks can be invalid */
+		if (!tk->clock_valid)
+			return;
+
 		now = tk_clock_read(&tk->tkr_mono);
 		systime_snapshot->cs_id = tk->tkr_mono.clock->id;
 		systime_snapshot->cs_was_changed_seq = tk->cs_was_changed_seq;

[patch V2 10/25] timekeeping: Add system_counterval_t to struct system_device_crosststamp

[Thomas Gleixner]

From: Thomas Gleixner <tglx@kernel.org>

An upcoming extension to the PTP IOCTL requires to return the system counter
value and the clocksource ID to user space. get_device_system_crosststamp() has
this information already.

Extend struct system_device_crosststamp with a system_counterval_t member
and fill in the data.

Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Tested-by: David Woodhouse <dwmw@amazon.co.uk>
Tested-by: Arthur Kiyanovski <akiyano@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
---
 include/linux/timekeeping.h |   28 ++++++++++++++------------
 kernel/time/timekeeping.c   |   46 ++++++++++++++++++++------------------------
 2 files changed, 36 insertions(+), 38 deletions(-)
--- a/include/linux/timekeeping.h
+++ b/include/linux/timekeeping.h
@@ -297,19 +297,6 @@ struct system_time_snapshot {
 };
 
 /**
- * struct system_device_crosststamp - system/device cross-timestamp
- *				      (synchronized capture)
- * @device:		Device time
- * @sys_realtime:	Realtime simultaneous with device time
- * @sys_monoraw:	Monotonic raw simultaneous with device time
- */
-struct system_device_crosststamp {
-	ktime_t device;
-	ktime_t sys_realtime;
-	ktime_t sys_monoraw;
-};
-
-/**
  * struct system_counterval_t - system counter value with the ID of the
  *				corresponding clocksource
  * @cycles:	System counter value
@@ -325,6 +312,21 @@ struct system_counterval_t {
 	bool			use_nsecs;
 };
 
+/**
+ * struct system_device_crosststamp - system/device cross-timestamp
+ *				      (synchronized capture)
+ * @device:		Device time
+ * @sys_counter:	Clocksource counter value simultaneous with device time
+ * @sys_realtime:	Realtime simultaneous with device time
+ * @sys_monoraw:	Monotonic raw simultaneous with device time
+ */
+struct system_device_crosststamp {
+	ktime_t				device;
+	struct system_counterval_t	sys_counter;
+	ktime_t				sys_realtime;
+	ktime_t				sys_monoraw;
+};
+
 extern bool ktime_real_to_base_clock(ktime_t treal,
 				     enum clocksource_ids base_id, u64 *cycles);
 extern bool timekeeping_clocksource_has_base(enum clocksource_ids id);
--- a/kernel/time/timekeeping.c
+++ b/kernel/time/timekeeping.c
@@ -1418,6 +1418,8 @@ static bool convert_base_to_cs(struct sy
 		return false;
 
 	scv->cycles += base->offset;
+	/* Set the clocksource ID as scv::cycles is now clocksource based */
+	scv->cs_id = cs->id;
 	return true;
 }
 
@@ -1485,11 +1487,11 @@ EXPORT_SYMBOL_GPL(ktime_real_to_base_clo
 
 /**
  * get_device_system_crosststamp - Synchronously capture system/device timestamp
- * @get_time_fn:	Callback to get simultaneous device time and
- *	system counter from the device driver
+ * @get_time_fn:	Callback to get simultaneous device time and system counter
+ *			from the device driver
  * @ctx:		Context passed to get_time_fn()
- * @history_begin:	Historical reference point used to interpolate system
- *	time when counter provided by the driver is before the current interval
+ * @history_begin:	Historical reference point used to interpolate system time when
+ *			the counter value provided by the driver is before the current interval
  * @xtstamp:		Receives simultaneously captured system and device time
  *
  * Reads a timestamp from a device and correlates it to system time
@@ -1502,14 +1504,12 @@ int get_device_system_crosststamp(int (*
 				  struct system_time_snapshot *history_begin,
 				  struct system_device_crosststamp *xtstamp)
 {
-	struct system_counterval_t system_counterval = {};
+	u64 syscnt_cycles, cycles, now, interval_start;
 	struct timekeeper *tk = &tk_core.timekeeper;
-	u64 cycles, now, interval_start;
-	unsigned int clock_was_set_seq = 0;
+	unsigned int seq, clock_was_set_seq = 0;
 	ktime_t base_real, base_raw;
 	u64 nsec_real, nsec_raw;
 	u8 cs_was_changed_seq;
-	unsigned int seq;
 	bool do_interp;
 	int ret;
 
@@ -1519,19 +1519,20 @@ int get_device_system_crosststamp(int (*
 		 * Try to synchronously capture device time and a system
 		 * counter value calling back into the device driver
 		 */
-		ret = get_time_fn(&xtstamp->device, &system_counterval, ctx);
+		ret = get_time_fn(&xtstamp->device, &xtstamp->sys_counter, ctx);
 		if (ret)
 			return ret;
 
 		/*
 		 * Verify that the clocksource ID associated with the captured
 		 * system counter value is the same as for the currently
-		 * installed timekeeper clocksource
+		 * installed timekeeper clocksource and convert to it.
 		 */
-		if (system_counterval.cs_id == CSID_GENERIC ||
-		    !convert_base_to_cs(&system_counterval))
+		if (xtstamp->sys_counter.cs_id == CSID_GENERIC ||
+		    !convert_base_to_cs(&xtstamp->sys_counter))
 			return -ENODEV;
-		cycles = system_counterval.cycles;
+
+		cycles = syscnt_cycles = xtstamp->sys_counter.cycles;
 
 		/*
 		 * Check whether the system counter value provided by the
@@ -1573,24 +1574,19 @@ int get_device_system_crosststamp(int (*
 		 * clocksource change
 		 */
 		if (!history_begin ||
-		    !timestamp_in_interval(history_begin->cycles,
-					   cycles, system_counterval.cycles) ||
+		    !timestamp_in_interval(history_begin->cycles, cycles, syscnt_cycles) ||
 		    history_begin->cs_was_changed_seq != cs_was_changed_seq)
 			return -EINVAL;
-		partial_history_cycles = cycles - system_counterval.cycles;
+
+		partial_history_cycles = cycles - syscnt_cycles;
 		total_history_cycles = cycles - history_begin->cycles;
-		discontinuity =
-			history_begin->clock_was_set_seq != clock_was_set_seq;
+		discontinuity = history_begin->clock_was_set_seq != clock_was_set_seq;
 
-		ret = adjust_historical_crosststamp(history_begin,
-						    partial_history_cycles,
-						    total_history_cycles,
-						    discontinuity, xtstamp);
-		if (ret)
-			return ret;
+		ret = adjust_historical_crosststamp(history_begin, partial_history_cycles,
+						    total_history_cycles, discontinuity, xtstamp);
 	}
 
-	return 0;
+	return ret;
 }
 EXPORT_SYMBOL_GPL(get_device_system_crosststamp);
 

[patch V2 11/25] timekeeping: Add CLOCK ID to system_device_crosststamp

[Thomas Gleixner]

From: Thomas Gleixner <tglx@kernel.org>

The normal capture for system/device cross timestamps is CLOCK_REALTIME,
but that's meaningless for AUX clocks.

Add a clock_id field to struct system_device_crosststamp and initialize it
with CLOCK_REALTIME at the two places which prepare for cross
timestamps.

After the related code has been cleaned up, the core code will honor the
clock_id field when calculating the system time from the system counter
snapshot.

No functional change.

Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Tested-by: David Woodhouse <dwmw@amazon.co.uk>
Tested-by: Arthur Kiyanovski <akiyano@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
---
 drivers/ptp/ptp_chardev.c     |    2 +-
 include/linux/timekeeping.h   |    2 ++
 sound/hda/common/controller.c |    2 +-
 3 files changed, 4 insertions(+), 2 deletions(-)
--- a/drivers/ptp/ptp_chardev.c
+++ b/drivers/ptp/ptp_chardev.c
@@ -317,8 +317,8 @@ typedef int (*ptp_crosststamp_fn)(struct
 static long ptp_sys_offset_precise(struct ptp_clock *ptp, void __user *arg,
 				   ptp_crosststamp_fn crosststamp_fn)
 {
+	struct system_device_crosststamp xtstamp = { .clock_id = CLOCK_REALTIME };
 	struct ptp_sys_offset_precise precise_offset;
-	struct system_device_crosststamp xtstamp;
 	struct timespec64 ts;
 	int err;
 
--- a/include/linux/timekeeping.h
+++ b/include/linux/timekeeping.h
@@ -315,12 +315,14 @@ struct system_counterval_t {
 /**
  * struct system_device_crosststamp - system/device cross-timestamp
  *				      (synchronized capture)
+ * @clock_id:		System time Clock ID to capture
  * @device:		Device time
  * @sys_counter:	Clocksource counter value simultaneous with device time
  * @sys_realtime:	Realtime simultaneous with device time
  * @sys_monoraw:	Monotonic raw simultaneous with device time
  */
 struct system_device_crosststamp {
+	clockid_t			clock_id;
 	ktime_t				device;
 	struct system_counterval_t	sys_counter;
 	ktime_t				sys_realtime;
--- a/sound/hda/common/controller.c
+++ b/sound/hda/common/controller.c
@@ -489,9 +489,9 @@ static int azx_get_time_info(struct snd_
 			struct snd_pcm_audio_tstamp_config *audio_tstamp_config,
 			struct snd_pcm_audio_tstamp_report *audio_tstamp_report)
 {
+	struct system_device_crosststamp xtstamp = { .clock_id = CLOCK_REALTIME };
 	struct azx_dev *azx_dev = get_azx_dev(substream);
 	struct snd_pcm_runtime *runtime = substream->runtime;
-	struct system_device_crosststamp xtstamp;
 	int ret;
 	u64 nsec;
 

[patch V2 12/25] wifi: iwlwifi: Adopt PTP cross timestamps to core changes

[Thomas Gleixner]

From: Thomas Gleixner <tglx@kernel.org>

iwlwifi only supports CLOCK_REALTIME timestamps and provides an incomplete
result without system counter values etc.

It also zeros struct system_device_crosststamp, which is already zeroed in
the core and initialized with the clock ID.

Remove the zeroing and reject any request for a clock ID other than REALTIME.

Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Tested-by: Arthur Kiyanovski <akiyano@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
---
 drivers/net/wireless/intel/iwlwifi/mld/ptp.c |    3 ++-
 drivers/net/wireless/intel/iwlwifi/mvm/ptp.c |    5 +++--
 2 files changed, 5 insertions(+), 3 deletions(-)
--- a/drivers/net/wireless/intel/iwlwifi/mld/ptp.c
+++ b/drivers/net/wireless/intel/iwlwifi/mld/ptp.c
@@ -250,7 +250,8 @@ iwl_mld_phc_get_crosstimestamp(struct pt
 	/* System (wall) time */
 	ktime_t sys_time;
 
-	memset(xtstamp, 0, sizeof(struct system_device_crosststamp));
+	if (xtstamp->clock_id != CLOCK_REALTIME)
+		return -ENOTSUPP;
 
 	ret = iwl_mld_get_crosstimestamp_fw(mld, &gp2, &sys_time);
 	if (ret) {
--- a/drivers/net/wireless/intel/iwlwifi/mvm/ptp.c
+++ b/drivers/net/wireless/intel/iwlwifi/mvm/ptp.c
@@ -160,13 +160,14 @@ iwl_mvm_phc_get_crosstimestamp(struct pt
 	/* System (wall) time */
 	ktime_t sys_time;
 
-	memset(xtstamp, 0, sizeof(struct system_device_crosststamp));
-
 	if (!mvm->ptp_data.ptp_clock) {
 		IWL_ERR(mvm, "No PHC clock registered\n");
 		return -ENODEV;
 	}
 
+	if (xtstamp->clock_id != CLOCK_REALTIME)
+		return -ENOTSUPP;
+
 	mutex_lock(&mvm->mutex);
 	if (fw_has_capa(&mvm->fw->ucode_capa, IWL_UCODE_TLV_CAPA_SYNCED_TIME)) {
 		ret = iwl_mvm_get_crosstimestamp_fw(mvm, &gp2, &sys_time);

[patch V2 13/25] ice/ptp: Use provided clock ID for history snapshot

[Thomas Gleixner]

From: Thomas Gleixner <tglx@kernel.org>

The PTP core indicates in system_device_crosststamp::clock_id the clock ID
for which then system time stamp should be taken. That allows to utilize
hardware timestamps with e.g. AUX clocks.

Save the provided clock ID and use it in ice_capture_crosststamp() for
taking the history snapshot.

No functional change.

Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Tested-by: Arthur Kiyanovski <akiyano@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Acked-by: Jacob Keller <jacob.e.keller@intel.com>
---
 drivers/net/ethernet/intel/ice/ice_ptp.c |    5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
--- a/drivers/net/ethernet/intel/ice/ice_ptp.c
+++ b/drivers/net/ethernet/intel/ice/ice_ptp.c
@@ -2065,11 +2065,13 @@ static const struct ice_crosststamp_cfg
 /**
  * struct ice_crosststamp_ctx - Device cross timestamp context
  * @snapshot: snapshot of system clocks for historic interpolation
+ * @snapshot_clock_id: System clock ID for @snapshot
  * @pf: pointer to the PF private structure
  * @cfg: pointer to hardware configuration for cross timestamp
  */
 struct ice_crosststamp_ctx {
 	struct system_time_snapshot snapshot;
+	clockid_t snapshot_clock_id;
 	struct ice_pf *pf;
 	const struct ice_crosststamp_cfg *cfg;
 };
@@ -2115,7 +2117,7 @@ static int ice_capture_crosststamp(ktime
 	}
 
 	/* Snapshot system time for historic interpolation */
-	ktime_get_snapshot(&ctx->snapshot);
+	ktime_get_snapshot_id(ctx->snapshot_clock_id, &ctx->snapshot);
 
 	/* Program cmd to master timer */
 	ice_ptp_src_cmd(hw, ICE_PTP_READ_TIME);
@@ -2176,6 +2178,7 @@ static int ice_ptp_getcrosststamp(struct
 {
 	struct ice_pf *pf = ptp_info_to_pf(info);
 	struct ice_crosststamp_ctx ctx = {
+		.snapshot_clock_id = cts->clock_id,
 		.pf = pf,
 	};
 

[patch V2 14/25] igc: Use provided clock ID for history snapshot

[Thomas Gleixner]

From: Thomas Gleixner <tglx@kernel.org>

The PTP core indicates in system_device_crosststamp::clock_id the clock ID
for which the system time stamp should be taken. That allows to utilize
hardware timestamps with e.g. AUX clocks.

Save the provided clock ID and use it in igc_phc_get_syncdevicetime() for
taking the history snapshot.

No functional change.

Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Tested-by: Arthur Kiyanovski <akiyano@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Acked-by: Jacob Keller <jacob.e.keller@intel.com>
---
 drivers/net/ethernet/intel/igc/igc.h     |    1 +
 drivers/net/ethernet/intel/igc/igc_ptp.c |    4 +++-
 2 files changed, 4 insertions(+), 1 deletion(-)
--- a/drivers/net/ethernet/intel/igc/igc.h
+++ b/drivers/net/ethernet/intel/igc/igc.h
@@ -326,6 +326,7 @@ struct igc_adapter {
 	struct timespec64 prev_ptp_time; /* Pre-reset PTP clock */
 	ktime_t ptp_reset_start; /* Reset time in clock mono */
 	struct system_time_snapshot snapshot;
+	clockid_t snapshot_clock_id;
 	struct mutex ptm_lock; /* Only allow one PTM transaction at a time */
 
 	char fw_version[32];
--- a/drivers/net/ethernet/intel/igc/igc_ptp.c
+++ b/drivers/net/ethernet/intel/igc/igc_ptp.c
@@ -1049,7 +1049,7 @@ static int igc_phc_get_syncdevicetime(kt
 	 */
 	do {
 		/* Get a snapshot of system clocks to use as historic value. */
-		ktime_get_snapshot(&adapter->snapshot);
+		ktime_get_snapshot_id(adapter->snapshot_clock_id, &adapter->snapshot);
 
 		igc_ptm_trigger(hw);
 
@@ -1103,6 +1103,8 @@ static int igc_ptp_getcrosststamp(struct
 	/* This blocks until any in progress PTM transactions complete */
 	mutex_lock(&adapter->ptm_lock);
 
+	adapter->snapshot_clock_id = cts->clock_id;
+
 	ret = get_device_system_crosststamp(igc_phc_get_syncdevicetime,
 					    adapter, &adapter->snapshot, cts);
 	mutex_unlock(&adapter->ptm_lock);

[patch V2 15/25] net/mlx5: Use provided clock ID for history snapshot

[Thomas Gleixner]

From: Thomas Gleixner <tglx@kernel.org>

The PTP core indicates in system_device_crosststamp::clock_id the clock ID
for which the system time stamp should be taken. That allows to utilize
hardware timestamps with e.g. AUX clocks.

Use ktime_get_snapshot_id() and hand the provided clock ID in.

No functional change.

Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Tested-by: Arthur Kiyanovski <akiyano@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
---
 drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
--- a/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/lib/clock.c
@@ -340,7 +340,7 @@ static int mlx5_ptp_getcrosststamp(struc
 		goto unlock;
 	}
 
-	ktime_get_snapshot(&history_begin);
+	ktime_get_snapshot_id(cts->clock_id, &history_begin);
 
 	err = get_device_system_crosststamp(mlx5_mtctr_syncdevicetime, mdev,
 					    &history_begin, cts);
@@ -366,7 +366,7 @@ static int mlx5_ptp_getcrosscycles(struc
 		goto unlock;
 	}
 
-	ktime_get_snapshot(&history_begin);
+	ktime_get_snapshot_id(cts->clock_id, &history_begin);
 
 	err = get_device_system_crosststamp(mlx5_mtctr_syncdevicecyclestime,
 					    mdev, &history_begin, cts);

[patch V2 16/25] virtio_rtc: Use provided clock ID for history snapshot

[Thomas Gleixner]

From: Thomas Gleixner <tglx@kernel.org>

The PTP core indicates in system_device_crosststamp::clock_id the clock ID
for which the system time stamp should be taken. That allows to utilize
hardware timestamps with e.g. AUX clocks.

Use ktime_get_snapshot_id() and hand the provided clock ID in.

No functional change.

Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Tested-by: Arthur Kiyanovski <akiyano@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
---
 drivers/virtio/virtio_rtc_ptp.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
--- a/drivers/virtio/virtio_rtc_ptp.c
+++ b/drivers/virtio/virtio_rtc_ptp.c
@@ -139,7 +139,7 @@ static int viortc_ptp_getcrosststamp(str
 	if (ret)
 		return ret;
 
-	ktime_get_snapshot(&history_begin);
+	ktime_get_snapshot_id(xtstamp->clock_id, &history_begin);
 	if (history_begin.cs_id != cs_id)
 		return -EOPNOTSUPP;
 

[patch V2 17/25] timekeeping: Remove ktime_get_snapshot()

[Thomas Gleixner]

From: Thomas Gleixner <tglx@kernel.org>

All users have been converted to ktime_get_snapshot_id().

Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Tested-by: David Woodhouse <dwmw@amazon.co.uk>
Tested-by: Arthur Kiyanovski <akiyano@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Thomas Weißschuh <thomas.weissschuh@linutronix.de>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
---
 include/linux/timekeeping.h |    5 -----
 1 file changed, 5 deletions(-)
--- a/include/linux/timekeeping.h
+++ b/include/linux/timekeeping.h
@@ -350,11 +350,6 @@ extern int get_device_system_crosststamp
  */
 extern void ktime_get_snapshot_id(clockid_t clock_id, struct system_time_snapshot *systime_snapshot);
 
-static inline void ktime_get_snapshot(struct system_time_snapshot *systime_snapshot)
-{
-	ktime_get_snapshot_id(CLOCK_REALTIME, systime_snapshot);
-}
-
 /*
  * Persistent clock related interfaces
  */

[patch V2 18/25] timekeeping: Prepare for cross timestamps on arbitrary clock IDs

[Thomas Gleixner]

From: Thomas Gleixner <tglx@kernel.org>

PTP device system crosstime stamps support only CLOCK_REALTIME, which is
meaningless for AUX clocks. The PTP core hands in the clock ID already, so
prepare the core code to honor it.

 - Add a new sys_systime field to struct system_device_crosststamp which
   aliases the sys_realtime field. Once all users are converted
   sys_realtime can be removed.

 - Prepare get_device_system_crosststamp() and the related code for it by
   switching to sys_systime and providing the initial changes to utilize
   different time keepers.

No functional change intended.

Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Tested-by: David Woodhouse <dwmw@amazon.co.uk>
Tested-by: Arthur Kiyanovski <akiyano@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Thomas Weißschuh <thomas.weissschuh@linutronix.de>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
---
 include/linux/timekeeping.h |    7 ++++-
 kernel/time/timekeeping.c   |   60 +++++++++++++++++++++++++-------------------
 2 files changed, 41 insertions(+), 26 deletions(-)
--- a/include/linux/timekeeping.h
+++ b/include/linux/timekeeping.h
@@ -319,13 +319,18 @@ struct system_counterval_t {
  * @device:		Device time
  * @sys_counter:	Clocksource counter value simultaneous with device time
  * @sys_realtime:	Realtime simultaneous with device time
+ * @sys_systime:	System time for @clock_id
  * @sys_monoraw:	Monotonic raw simultaneous with device time
  */
 struct system_device_crosststamp {
 	clockid_t			clock_id;
 	ktime_t				device;
 	struct system_counterval_t	sys_counter;
-	ktime_t				sys_realtime;
+	union {
+		/* realtime goes away once all users are converted */
+		ktime_t			sys_realtime;
+		ktime_t			sys_systime;
+	};
 	ktime_t				sys_monoraw;
 };
 
--- a/kernel/time/timekeeping.c
+++ b/kernel/time/timekeeping.c
@@ -1312,7 +1312,7 @@ static int adjust_historical_crosststamp
 					 struct system_device_crosststamp *ts)
 {
 	struct timekeeper *tk = &tk_core.timekeeper;
-	u64 corr_raw, corr_real;
+	u64 corr_raw, corr_sys;
 	bool interp_forward;
 	int ret;
 
@@ -1329,8 +1329,7 @@ static int adjust_historical_crosststamp
 	 * Scale the monotonic raw time delta by:
 	 *	partial_history_cycles / total_history_cycles
 	 */
-	corr_raw = (u64)ktime_to_ns(
-		ktime_sub(ts->sys_monoraw, history->monoraw));
+	corr_raw = (u64)ktime_to_ns(ktime_sub(ts->sys_monoraw, history->monoraw));
 	ret = scale64_check_overflow(partial_history_cycles,
 				     total_history_cycles, &corr_raw);
 	if (ret)
@@ -1338,30 +1337,29 @@ static int adjust_historical_crosststamp
 
 	/*
 	 * If there is a discontinuity in the history, scale monotonic raw
-	 *	correction by:
-	 *	mult(real)/mult(raw) yielding the realtime correction
-	 * Otherwise, calculate the realtime correction similar to monotonic
-	 *	raw calculation
+	 * correction by:
+	 *	mult(sys)/mult(raw) yielding the system time correction
+	 *
+	 * Otherwise, calculate the system time correction similar to monotonic
+	 * raw calculation
 	 */
 	if (discontinuity) {
-		corr_real = mul_u64_u32_div
-			(corr_raw, tk->tkr_mono.mult, tk->tkr_raw.mult);
+		corr_sys = mul_u64_u32_div(corr_raw, tk->tkr_mono.mult, tk->tkr_raw.mult);
 	} else {
-		corr_real = (u64)ktime_to_ns(
-			ktime_sub(ts->sys_realtime, history->systime));
-		ret = scale64_check_overflow(partial_history_cycles,
-					     total_history_cycles, &corr_real);
+		corr_sys = (u64)ktime_to_ns(ktime_sub(ts->sys_systime, history->systime));
+		ret = scale64_check_overflow(partial_history_cycles, total_history_cycles,
+					     &corr_sys);
 		if (ret)
 			return ret;
 	}
 
-	/* Fixup monotonic raw and real time time values */
+	/* Fixup monotonic raw and system time time values */
 	if (interp_forward) {
 		ts->sys_monoraw = ktime_add_ns(history->monoraw, corr_raw);
-		ts->sys_realtime = ktime_add_ns(history->systime, corr_real);
+		ts->sys_systime = ktime_add_ns(history->systime, corr_sys);
 	} else {
 		ts->sys_monoraw = ktime_sub_ns(ts->sys_monoraw, corr_raw);
-		ts->sys_realtime = ktime_sub_ns(ts->sys_realtime, corr_real);
+		ts->sys_systime = ktime_sub_ns(ts->sys_systime, corr_sys);
 	}
 
 	return 0;
@@ -1505,16 +1503,29 @@ int get_device_system_crosststamp(int (*
 				  struct system_device_crosststamp *xtstamp)
 {
 	u64 syscnt_cycles, cycles, now, interval_start;
-	struct timekeeper *tk = &tk_core.timekeeper;
 	unsigned int seq, clock_was_set_seq = 0;
-	ktime_t base_real, base_raw;
-	u64 nsec_real, nsec_raw;
+	ktime_t base_sys, base_raw, *offs;
+	u64 nsec_sys, nsec_raw;
 	u8 cs_was_changed_seq;
 	bool do_interp;
+	struct timekeeper *tk;
+	struct tk_data *tkd;
 	int ret;
 
+	switch (xtstamp->clock_id) {
+	case CLOCK_REALTIME:
+		tkd = &tk_core;
+		offs = &tk_core.timekeeper.offs_real;
+		break;
+	default:
+		WARN_ON_ONCE(1);
+		return -ENODEV;
+	}
+
+	tk = &tkd->timekeeper;
+
 	do {
-		seq = read_seqcount_begin(&tk_core.seq);
+		seq = read_seqcount_begin(&tkd->seq);
 		/*
 		 * Try to synchronously capture device time and a system
 		 * counter value calling back into the device driver
@@ -1549,15 +1560,14 @@ int get_device_system_crosststamp(int (*
 			do_interp = false;
 		}
 
-		base_real = ktime_add(tk->tkr_mono.base,
-				      tk_core.timekeeper.offs_real);
+		base_sys = ktime_add(tk->tkr_mono.base, *offs);
 		base_raw = tk->tkr_raw.base;
 
-		nsec_real = timekeeping_cycles_to_ns(&tk->tkr_mono, cycles);
+		nsec_sys = timekeeping_cycles_to_ns(&tk->tkr_mono, cycles);
 		nsec_raw = timekeeping_cycles_to_ns(&tk->tkr_raw, cycles);
-	} while (read_seqcount_retry(&tk_core.seq, seq));
+	} while (read_seqcount_retry(&tkd->seq, seq));
 
-	xtstamp->sys_realtime = ktime_add_ns(base_real, nsec_real);
+	xtstamp->sys_systime = ktime_add_ns(base_sys, nsec_sys);
 	xtstamp->sys_monoraw = ktime_add_ns(base_raw, nsec_raw);
 
 	/*

[patch V2 19/25] ptp: Use system_device_crosststamp::sys_systime

[Thomas Gleixner]

From: Thomas Gleixner <tglx@kernel.org>

.. to prepare for cross timestamps with variable clock IDs.

No functional change.

Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Tested-by: David Woodhouse <dwmw@amazon.co.uk>
Tested-by: Arthur Kiyanovski <akiyano@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
---
 drivers/ptp/ptp_chardev.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
--- a/drivers/ptp/ptp_chardev.c
+++ b/drivers/ptp/ptp_chardev.c
@@ -333,7 +333,7 @@ static long ptp_sys_offset_precise(struc
 	ts = ktime_to_timespec64(xtstamp.device);
 	precise_offset.device.sec = ts.tv_sec;
 	precise_offset.device.nsec = ts.tv_nsec;
-	ts = ktime_to_timespec64(xtstamp.sys_realtime);
+	ts = ktime_to_timespec64(xtstamp.sys_systime);
 	precise_offset.sys_realtime.sec = ts.tv_sec;
 	precise_offset.sys_realtime.nsec = ts.tv_nsec;
 	ts = ktime_to_timespec64(xtstamp.sys_monoraw);

[patch V2 20/25] wifi: iwlwifi: Use system_device_crosststamp::sys_systime

[Thomas Gleixner]

From: Thomas Gleixner <tglx@kernel.org>

sys_systime is an alias for sys_realtime. The latter will be removed so
switch the code over to the new naming scheme.

No functional change.

Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Tested-by: Arthur Kiyanovski <akiyano@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
---
 drivers/net/wireless/intel/iwlwifi/mld/ptp.c |    2 +-
 drivers/net/wireless/intel/iwlwifi/mvm/ptp.c |    2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)
--- a/drivers/net/wireless/intel/iwlwifi/mld/ptp.c
+++ b/drivers/net/wireless/intel/iwlwifi/mld/ptp.c
@@ -271,7 +271,7 @@ iwl_mld_phc_get_crosstimestamp(struct pt
 
 	/* System monotonic raw time is not used */
 	xtstamp->device = ns_to_ktime(gp2_ns);
-	xtstamp->sys_realtime = sys_time;
+	xtstamp->sys_systime = sys_time;
 
 	return ret;
 }
--- a/drivers/net/wireless/intel/iwlwifi/mvm/ptp.c
+++ b/drivers/net/wireless/intel/iwlwifi/mvm/ptp.c
@@ -185,7 +185,7 @@ iwl_mvm_phc_get_crosstimestamp(struct pt
 
 	/* System monotonic raw time is not used */
 	xtstamp->device = (ktime_t)gp2_ns;
-	xtstamp->sys_realtime = sys_time;
+	xtstamp->sys_systime = sys_time;
 
 out:
 	mutex_unlock(&mvm->mutex);

[patch V2 21/25] ALSA: hda/common: Use system_device_crosststamp::sys_systime

[Thomas Gleixner]

From: Thomas Gleixner <tglx@kernel.org>

sys_systime is an alias for sys_realtime. The latter will be removed so
switch the code over to the new naming scheme.

No functional change.

Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Tested-by: Arthur Kiyanovski <akiyano@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
---
 sound/hda/common/controller.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
--- a/sound/hda/common/controller.c
+++ b/sound/hda/common/controller.c
@@ -525,7 +525,7 @@ static int azx_get_time_info(struct snd_
 			break;
 
 		default:
-			*system_ts = ktime_to_timespec64(xtstamp.sys_realtime);
+			*system_ts = ktime_to_timespec64(xtstamp.sys_systime);
 			break;
 
 		}

[patch V2 22/25] timekeeping: Remove system_device_crosststamp::sys_realtime

[Thomas Gleixner]

From: Thomas Gleixner <tglx@kernel.org>

All users are converted to sys_systime.

Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Tested-by: David Woodhouse <dwmw@amazon.co.uk>
Tested-by: Arthur Kiyanovski <akiyano@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Thomas Weißschuh <thomas.weissschuh@linutronix.de>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
---
 include/linux/timekeeping.h |    7 +------
 1 file changed, 1 insertion(+), 6 deletions(-)
--- a/include/linux/timekeeping.h
+++ b/include/linux/timekeeping.h
@@ -318,7 +318,6 @@ struct system_counterval_t {
  * @clock_id:		System time Clock ID to capture
  * @device:		Device time
  * @sys_counter:	Clocksource counter value simultaneous with device time
- * @sys_realtime:	Realtime simultaneous with device time
  * @sys_systime:	System time for @clock_id
  * @sys_monoraw:	Monotonic raw simultaneous with device time
  */
@@ -326,11 +325,7 @@ struct system_device_crosststamp {
 	clockid_t			clock_id;
 	ktime_t				device;
 	struct system_counterval_t	sys_counter;
-	union {
-		/* realtime goes away once all users are converted */
-		ktime_t			sys_realtime;
-		ktime_t			sys_systime;
-	};
+	ktime_t				sys_systime;
 	ktime_t				sys_monoraw;
 };
 

[patch V2 23/25] timekeeping: Add support for AUX clock cross timestamping

[Thomas Gleixner]

From: Thomas Gleixner <tglx@kernel.org>

Now that all prerequisites are in place add the final support for AUX
clocks in get_device_system_crosststamp(), which enables the PTP layer to
support hardware cross timestamps with a new IOTCL.

Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Tested-by: Arthur Kiyanovski <akiyano@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Thomas Weißschuh <thomas.weissschuh@linutronix.de>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
---
 kernel/time/timekeeping.c |    6 ++++++
 1 file changed, 6 insertions(+)
--- a/kernel/time/timekeeping.c
+++ b/kernel/time/timekeeping.c
@@ -1517,6 +1517,12 @@ int get_device_system_crosststamp(int (*
 		tkd = &tk_core;
 		offs = &tk_core.timekeeper.offs_real;
 		break;
+	case CLOCK_AUX ... CLOCK_AUX_LAST:
+		tkd = aux_get_tk_data(xtstamp->clock_id);
+		if (!tkd)
+			return -ENODEV;
+		offs = &tkd->timekeeper.offs_aux;
+		break;
 	default:
 		WARN_ON_ONCE(1);
 		return -ENODEV;

[patch V2 24/25] ptp: Switch to ktime_get_snapshot_id() for pre/post timestamps

[Thomas Gleixner]

From: Thomas Gleixner <tglx@kernel.org>

To prepare for a new PTP IOCTL, which exposes the raw counter value along
with the requested system time snapshot, switch the pre/post time stamp
sampling over to use ktime_get_snapshot_id() and fix up all usage sites.

No functional change intended.

The ptp_vmclock conversion was simplified by David Woodhouse.

Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Tested-by: David Woodhouse <dwmw@amazon.co.uk>
Tested-by: Arthur Kiyanovski <akiyano@amazon.com>
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Acked-by: Vadim Fedorenko <vadim.fedorenko@linux.dev>
---
 drivers/net/dsa/sja1105/sja1105_main.c |    8 ++++----
 drivers/ptp/ptp_chardev.c              |   14 +++++++++-----
 drivers/ptp/ptp_ocp.c                  |   11 ++++-------
 drivers/ptp/ptp_vmclock.c              |   23 +++++++----------------
 include/linux/ptp_clock_kernel.h       |   15 ++++++++-------
 5 files changed, 32 insertions(+), 39 deletions(-)
--- a/drivers/net/dsa/sja1105/sja1105_main.c
+++ b/drivers/net/dsa/sja1105/sja1105_main.c
@@ -2310,10 +2310,10 @@ int sja1105_static_config_reload(struct
 		goto out;
 	}
 
-	t1 = timespec64_to_ns(&ptp_sts_before.pre_ts);
-	t2 = timespec64_to_ns(&ptp_sts_before.post_ts);
-	t3 = timespec64_to_ns(&ptp_sts_after.pre_ts);
-	t4 = timespec64_to_ns(&ptp_sts_after.post_ts);
+	t1 = ktime_to_ns(ptp_sts_before.pre_sts.systime);
+	t2 = ktime_to_ns(ptp_sts_before.post_sts.systime);
+	t3 = ktime_to_ns(ptp_sts_after.pre_sts.systime);
+	t4 = ktime_to_ns(ptp_sts_after.post_sts.systime);
 	/* Mid point, corresponds to pre-reset PTPCLKVAL */
 	t12 = t1 + (t2 - t1) / 2;
 	/* Mid point, corresponds to post-reset PTPCLKVAL, aka 0 */
--- a/drivers/ptp/ptp_chardev.c
+++ b/drivers/ptp/ptp_chardev.c
@@ -386,15 +386,19 @@ static long ptp_sys_offset_extended(stru
 			return err;
 
 		/* Filter out disabled or unavailable clocks */
-		if (sts.pre_ts.tv_sec < 0 || sts.post_ts.tv_sec < 0)
+		if (!sts.pre_sts.valid || !sts.post_sts.valid)
 			return -EINVAL;
 
-		extoff->ts[i][0].sec = sts.pre_ts.tv_sec;
-		extoff->ts[i][0].nsec = sts.pre_ts.tv_nsec;
 		extoff->ts[i][1].sec = ts.tv_sec;
 		extoff->ts[i][1].nsec = ts.tv_nsec;
-		extoff->ts[i][2].sec = sts.post_ts.tv_sec;
-		extoff->ts[i][2].nsec = sts.post_ts.tv_nsec;
+
+		ts = ktime_to_timespec64(sts.pre_sts.systime);
+		extoff->ts[i][0].sec = ts.tv_sec;
+		extoff->ts[i][0].nsec = ts.tv_nsec;
+
+		ts = ktime_to_timespec64(sts.post_sts.systime);
+		extoff->ts[i][2].sec = ts.tv_sec;
+		extoff->ts[i][2].nsec = ts.tv_nsec;
 	}
 
 	return copy_to_user(arg, extoff, sizeof(*extoff)) ? -EFAULT : 0;
--- a/drivers/ptp/ptp_ocp.c
+++ b/drivers/ptp/ptp_ocp.c
@@ -1491,11 +1491,8 @@ static int
 	}
 	ptp_read_system_postts(sts);
 
-	if (sts && bp->ts_window_adjust) {
-		s64 ns = timespec64_to_ns(&sts->post_ts);
-
-		sts->post_ts = ns_to_timespec64(ns - bp->ts_window_adjust);
-	}
+	if (sts && bp->ts_window_adjust)
+		sts->post_sts.systime -= bp->ts_window_adjust;
 
 	time_ns = ioread32(&bp->reg->time_ns);
 	time_sec = ioread32(&bp->reg->time_sec);
@@ -4595,8 +4592,8 @@ ptp_ocp_summary_show(struct seq_file *s,
 		struct timespec64 sys_ts;
 		s64 pre_ns, post_ns, ns;
 
-		pre_ns = timespec64_to_ns(&sts.pre_ts);
-		post_ns = timespec64_to_ns(&sts.post_ts);
+		pre_ns = ktime_to_ns(sts.pre_sts.systime);
+		post_ns = ktime_to_ns(sts.post_sts.systime);
 		ns = (pre_ns + post_ns) / 2;
 		ns += (s64)bp->utc_tai_offset * NSEC_PER_SEC;
 		sys_ts = ns_to_timespec64(ns);
--- a/drivers/ptp/ptp_vmclock.c
+++ b/drivers/ptp/ptp_vmclock.c
@@ -101,7 +101,6 @@ static int vmclock_get_crosststamp(struc
 				   struct timespec64 *tspec)
 {
 	ktime_t deadline = ktime_add(ktime_get(), VMCLOCK_MAX_WAIT);
-	struct system_time_snapshot systime_snapshot;
 	uint64_t cycle, delta, seq, frac_sec;
 
 #ifdef CONFIG_X86
@@ -132,17 +131,15 @@ static int vmclock_get_crosststamp(struc
 		 * will be derived from the *same* counter value.
 		 *
 		 * If the system isn't using the same counter, then the value
-		 * from ktime_get_snapshot_id() will still be used as pre_ts, and
-		 * ptp_read_system_postts() is called to populate postts after
-		 * calling get_cycles().
-		 *
-		 * The conversion to timespec64 happens further down, outside
-		 * the seq_count loop.
+		 * from ptp_read_system_prets() will still be used as pre_ts,
+		 * and ptp_read_system_postts() is called to populate postts
+		 * after calling get_cycles().
 		 */
 		if (sts) {
-			ktime_get_snapshot_id(CLOCK_REALTIME, &systime_snapshot);
-			if (systime_snapshot.cs_id == st->cs_id) {
-				cycle = systime_snapshot.cycles;
+			ptp_read_system_prets(sts);
+			if (sts->pre_sts.cs_id == st->cs_id) {
+				cycle = sts->pre_sts.cycles;
+				sts->post_sts = sts->pre_sts;
 			} else {
 				cycle = get_cycles();
 				ptp_read_system_postts(sts);
@@ -180,12 +177,6 @@ static int vmclock_get_crosststamp(struc
 		system_counter->cs_id = st->cs_id;
 	}
 
-	if (sts) {
-		sts->pre_ts = ktime_to_timespec64(systime_snapshot.systime);
-		if (systime_snapshot.cs_id == st->cs_id)
-			sts->post_ts = sts->pre_ts;
-	}
-
 	return 0;
 }
 
--- a/include/linux/ptp_clock_kernel.h
+++ b/include/linux/ptp_clock_kernel.h
@@ -12,6 +12,7 @@
 #include <linux/pps_kernel.h>
 #include <linux/ptp_clock.h>
 #include <linux/timecounter.h>
+#include <linux/timekeeping.h>
 #include <linux/skbuff.h>
 
 #define PTP_CLOCK_NAME_LEN	32
@@ -45,13 +46,13 @@ struct system_device_crosststamp;
 
 /**
  * struct ptp_system_timestamp - system time corresponding to a PHC timestamp
- * @pre_ts: system timestamp before capturing PHC
- * @post_ts: system timestamp after capturing PHC
- * @clockid: clock-base used for capturing the system timestamps
+ * @pre_sts:	system time snapshot before capturing PHC
+ * @post_sts:	system time snapshot after capturing PHC
+ * @clockid:	clock-base used for capturing the system timestamps
  */
 struct ptp_system_timestamp {
-	struct timespec64 pre_ts;
-	struct timespec64 post_ts;
+	struct system_time_snapshot pre_sts;
+	struct system_time_snapshot post_sts;
 	clockid_t clockid;
 };
 
@@ -510,13 +511,13 @@ static inline ktime_t ptp_convert_timest
 static inline void ptp_read_system_prets(struct ptp_system_timestamp *sts)
 {
 	if (sts)
-		ktime_get_clock_ts64(sts->clockid, &sts->pre_ts);
+		ktime_get_snapshot_id(sts->clockid, &sts->pre_sts);
 }
 
 static inline void ptp_read_system_postts(struct ptp_system_timestamp *sts)
 {
 	if (sts)
-		ktime_get_clock_ts64(sts->clockid, &sts->post_ts);
+		ktime_get_snapshot_id(sts->clockid, &sts->post_sts);
 }
 
 #endif

[patch V2 25/25] timekeeping: Add clocksource read_snapshot() method and hw_cycles to snapshot

[Thomas Gleixner]

From: David Woodhouse <dwmw@amazon.co.uk>

Add a read_snapshot() callback to struct clocksource which returns the
derived clocksource value while also providing the underlying hardware
counter reading and the related clocksource ID.

This allows ktime_get_snapshot_id() to populate new hw_cycles and hw_csid
fields in struct system_time_snapshot.

For clocksources that are derived from an underlying counter (e.g., Hyper-V
TSC page scales TSC to 10MHz, kvmclock scales TSC to 1GHz), this provides
atomic access to both the derived value needed for timekeeping
calculations, and the raw hardware counter needed by consumers like KVM's
master clock and the vmclock PTP driver.

[ tglx: Reworked it slightly ]

Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Thomas Gleixner <tglx@kernel.org>
Assisted-by: Kiro:claude-opus-4.6-1m
Link: https://patch.msgid.link/20260526230635.136914-1-dwmw2@infradead.org
---
 include/linux/clocksource.h |   24 ++++++++++++++++++++++++
 include/linux/timekeeping.h |    6 ++++++
 kernel/time/timekeeping.c   |   21 ++++++++++++++++++++-
 3 files changed, 50 insertions(+), 1 deletion(-)
--- a/include/linux/clocksource.h
+++ b/include/linux/clocksource.h
@@ -32,6 +32,21 @@ struct module;
 #include <vdso/clocksource.h>
 
 /**
+ * struct clocksource_hw_snapshot - Snapshot for the underlying hardware counter of derived
+ *				    clocksources like kvmclock or Hyper-V scaled TSC
+ * @hw_cycles:		The hardware counter value
+ * @hw_csid:		Clocksource ID of the hardware counter
+ *
+ * Such clocksources must implement the read_snapshot() callback and fill in the
+ * hardware counter value, the clocksource ID of the hardware counter and derive
+ * the actual clocksource cycles from @hw_cycles to provide an atomic snapshot
+ */
+struct clocksource_hw_snapshot {
+	u64			hw_cycles;
+	enum clocksource_ids	hw_csid;
+};
+
+/**
  * struct clocksource - hardware abstraction for a free running counter
  *	Provides mostly state-free accessors to the underlying hardware.
  *	This is the structure used for system time.
@@ -72,6 +87,14 @@ struct module;
  * @flags:		Flags describing special properties
  * @base:		Hardware abstraction for clock on which a clocksource
  *			is based
+ * @read_snapshot:	Extended @read() function for clocksources such as
+ *			kvmclock or the Hyper-V scaled TSC where the actual
+ *			clocksource value for timekeeping is calculated from an
+ *			underlying hardware counter. Returns the timekeeping
+ *			relevant cycle value and stores the raw value of the
+ *			underlying counter from which it was calculated
+ *			including the clocksource ID of that counter in the
+ *			clocksource hardware snapshot.
  * @enable:		Optional function to enable the clocksource
  * @disable:		Optional function to disable the clocksource
  * @suspend:		Optional suspend function for the clocksource
@@ -113,6 +136,7 @@ struct clocksource {
 	unsigned long		flags;
 	struct clocksource_base *base;
 
+	u64			(*read_snapshot)(struct clocksource *cs, struct clocksource_hw_snapshot *chs);
 	int			(*enable)(struct clocksource *cs);
 	void			(*disable)(struct clocksource *cs);
 	void			(*suspend)(struct clocksource *cs);
--- a/include/linux/timekeeping.h
+++ b/include/linux/timekeeping.h
@@ -279,18 +279,24 @@ static inline bool ktime_get_aux_ts64(cl
  * struct system_time_snapshot - Simultaneous time capture of CLOCK_MONOTONIC_RAW,
  *				 a selected CLOCK_* and the clocksource counter value
  * @cycles:		Clocksource counter value to produce the system times
+ * @hw_cycles:		For derived clocksources, the hardware counter value from
+ *			which @cycles was derived
  * @systime:		The system time of the selected CLOCK ID
  * @monoraw:		Monotonic raw system time
  * @cs_id:		Clocksource ID
+ * @hw_csid:		Clocksource ID of the underlying hardware counter for derived
+ *			clocksources which implement the read_snapshot() callback.
  * @clock_was_set_seq:	The sequence number of clock-was-set events
  * @cs_was_changed_seq:	The sequence number of clocksource change events
  * @valid:		True if the snapshot is valid
  */
 struct system_time_snapshot {
 	u64			cycles;
+	u64			hw_cycles;
 	ktime_t			systime;
 	ktime_t			monoraw;
 	enum clocksource_ids	cs_id;
+	enum clocksource_ids	hw_csid;
 	unsigned int		clock_was_set_seq;
 	u8			cs_was_changed_seq;
 	u8			valid;
--- a/kernel/time/timekeeping.c
+++ b/kernel/time/timekeeping.c
@@ -320,6 +320,7 @@ static __always_inline u64 tk_clock_read
 
 	return clock->read(clock);
 }
+
 static inline void clocksource_disable_inline_read(void) { }
 static inline void clocksource_enable_inline_read(void) { }
 #endif
@@ -1187,6 +1188,18 @@ noinstr time64_t __ktime_get_real_second
 	return tk->xtime_sec;
 }
 
+static inline u64 tk_clock_read_snapshot(const struct tk_read_base *tkr,
+					 struct clocksource_hw_snapshot *chs)
+{
+	struct clocksource *clock = READ_ONCE(tkr->clock);
+
+	if (unlikely(clock->read_snapshot))
+		return clock->read_snapshot(clock, chs);
+
+	return clock->read(clock);
+}
+
+
 /**
  * ktime_get_snapshot_id -  Simultaneously snapshot a given clock ID with
  *			    CLOCK_MONOTONIC_RAW and the underlying
@@ -1237,14 +1250,20 @@ void ktime_get_snapshot_id(clockid_t clo
 	tk = &tkd->timekeeper;
 
 	do {
+		struct clocksource_hw_snapshot chs = { };
+
 		seq = read_seqcount_begin(&tkd->seq);
 
 		/* Aux clocks can be invalid */
 		if (!tk->clock_valid)
 			return;
 
-		now = tk_clock_read(&tk->tkr_mono);
+		now = tk_clock_read_snapshot(&tk->tkr_mono, &chs);
 		systime_snapshot->cs_id = tk->tkr_mono.clock->id;
+
+		systime_snapshot->hw_cycles = chs.hw_cycles;
+		systime_snapshot->hw_csid = chs.hw_csid;
+
 		systime_snapshot->cs_was_changed_seq = tk->cs_was_changed_seq;
 		systime_snapshot->clock_was_set_seq = tk->clock_was_set_seq;
 

Re: [patch V2 00/25] timekeeping/ptp: Expand snapshot functionality

[Jacob Keller]

On 5/29/2026 12:59 PM, Thomas Gleixner wrote:
> This is an update to V1 which can be found here:
> 
>    https://lore.kernel.org/lkml/20260526165826.392227559@kernel.org
> 
> PTP wants to grow new snapshot functionality, which provides not only the
> captured CLOCK* values, but also the underlying clocksource counter value.
> 
>    https://lore.kernel.org/20260515164033.6403-1-akiyano@amazon.com
> 
> There was quite some discussion in seemingly related threads how to capture
> these values and how to provide core infrastructure so that driver writers
> have something to work with
> 
>    https://lore.kernel.org/20260514225842.110706-1-hramamurthy@google.com
>    https://lore.kernel.org/20260520135207.37826-1-dwmw2@infradead.org
> 
> This series implements the timekeeping related mechanisms to:
> 
>      1) Capture CLOCK values along with the clocksource counter value for
>      	non-hardware based sampling
> 
>      2) Expanding the hardware cross time stamp mechanism to hand back the
>      	clocksource counter value, which was captured by the device, along
>      	with the related CLOCK values
> 
>      3) Adding AUX clock support to the hardware cross timestamping core
> 
>      4) Add support for derived clocksources to the snapshot mechanism (New
>      	in V2)
> 
> Changes vs. V1:
> 
>   - Fixed the ptp_ocp typo - 0-day, Jakub
> 
>   - Renamed the system_time_snapshot members sys and raw so systime and
>     monoraw to make them less ambigous.
> 
>   - Fixed the error case return values of get_device_system_crosststamp()
> 
>   - Made ktime_snapshot_id() void as there is no point for the return
>     value, which is nowhere checked and cannot be propagated.
>     system_time_snapshot::valid has to be evaluated at the call sites
>     anyway. - Jacob
> 
>   - Picked up the first patch from Davids follow up series, which extends
>     the snapshot mechanism so that derived clocksources (like kvmclock and
>     Hyper-V scaled TSC) can return the actual underlying hardware counter
>     value (TSC for the two examples).
> 
>   - Collected Reviewed/Acked/Tested-by tags
> 
> Delta patch against v1 below.
> 
> The series is based on v7.1-rc2 and also available from git:
> 
>     git://git.kernel.org/pub/scm/linux/kernel/git/tglx/devel.git timekeeping-ptp-extend-v2
> 
> Thanks,

The changes in v2 are great! Appreciate it.

Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>