[v2,14/15] drm/i915: Mechanism to forward clock monotonic raw time in perf samples

Commit Message

[e-mail] Nov. 7, 2016, 10:32 a.m. UTC

From: Sourab Gupta <[e-mail]>

Currently, we have the ability to only forward the GPU timestamps in the
samples (which are generated via OA reports or PIPE_CONTROL commands
inserted in the ring). This limits the ability to correlate these samples
with the system events. If we scale the GPU timestamps according the
timestamp base/frequency info present in bspec, it is observed that the
timestamps drift really quickly from the system time.

An ability is therefore needed to report timestamps in different clock
domains, such as CLOCK_MONOTONIC (or _MONO_RAW), in the perf samples to
be of more practical use to the userspace. This ability becomes important
when we want to correlate/plot GPU events/samples with other system events
on the same timeline (e.g. vblank events, or timestamps when work was
submitted to kernel, etc.)

The patch here proposes a mechanism to achieve this. The correlation between
gpu time and system time is established using the cross timestamp framework.
For this purpose, the timestamp clock associated with the command stream, is
abstracted as timecounter/cyclecounter, before utilizing cross timestamp
framework to retrieve gpu/system time correlated values.
Different such gpu/system time values are then used to detect and correct
the error in published gpu timestamp clock frequency. The userspace can
request CLOCK_MONOTONIC_RAW timestamps in samples by requesting the
corresponding property while opening the stream.

v2: Added i915_driver_init_late() function to capture the new late init phase
for perf (Chris)

Signed-off-by: Sourab Gupta <[e-mail]>
---
 drivers/gpu/drm/i915/i915_drv.c  |  14 ++
 drivers/gpu/drm/i915/i915_drv.h  |  24 +++-
 drivers/gpu/drm/i915/i915_perf.c | 273 +++++++++++++++++++++++++++++++++++----
 include/uapi/drm/i915_drm.h      |   9 +-
 4 files changed, 296 insertions(+), 24 deletions(-)

Patch

diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index 06c7b55..2140fa7 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -876,6 +876,18 @@  static void i915_driver_cleanup_early(struct drm_i915_private *dev_priv)
 	i915_workqueues_cleanup(dev_priv);
 }
 
+/**
+ * i915_driver_init_late - initialize late stage driver components
+ * @dev_priv: device private
+ *
+ * Setup the driver components, which need to be inited after driver state has
+ * been registered and device enabled.
+ */
+static int i915_driver_init_late(struct drm_i915_private *dev_priv)
+{
+	i915_perf_init_late(dev_priv);
+}
+
 static int i915_mmio_setup(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = to_i915(dev);
@@ -1254,6 +1266,8 @@  int i915_driver_load(struct pci_dev *pdev, const struct pci_device_id *ent)
 
 	i915_driver_register(dev_priv);
 
+	i915_driver_init_late(dev_priv);
+
 	intel_runtime_pm_enable(dev_priv);
 
 	/* Everything is in place, we can now relax! */
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index e912679..557a124 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -42,6 +42,9 @@ 
 #include <linux/kref.h>
 #include <linux/pm_qos.h>
 #include <linux/shmem_fs.h>
+#include <linux/timecounter.h>
+#include <linux/clocksource.h>
+#include <linux/timekeeping.h>
 
 #include <drm/drmP.h>
 #include <drm/intel-gtt.h>
@@ -1843,6 +1846,9 @@  struct i915_perf_stream {
 	/* Whether the OA unit is in use */
 	bool using_oa;
 
+	/* monotonic_raw clk timestamp (in ns) for last sample */
+	u64 last_sample_ts;
+
 	const struct i915_perf_stream_ops *ops;
 };
 
@@ -1889,6 +1895,20 @@  struct i915_perf_cs_data_node {
 	u32 tag;
 };
 
+/**
+ * struct i915_clock_info - decribes i915 timestamp clock
+ *
+ */
+struct i915_clock_info {
+	struct cyclecounter cc;
+	struct timecounter tc;
+	struct system_device_crosststamp xtstamp;
+	ktime_t clk_offset; /* Offset (in ns) between monoraw clk and gpu time */
+	u32 timestamp_frequency;
+	u32 resync_period; /* in msecs */
+	struct delayed_work clk_sync_work;
+};
+
 struct drm_i915_private {
 	struct drm_device drm;
 
@@ -2189,6 +2209,8 @@  struct drm_i915_private {
 
 	struct i915_runtime_pm pm;
 
+	struct i915_clock_info ts_clk_info;
+
 	struct {
 		bool initialized;
 
@@ -2213,7 +2235,6 @@  struct drm_i915_private {
 
 			bool periodic;
 			int period_exponent;
-			int timestamp_frequency;
 
 			int tail_margin;
 
@@ -3796,6 +3817,7 @@  int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 
 /* i915_perf.c */
 extern void i915_perf_init(struct drm_i915_private *dev_priv);
+extern void i915_perf_init_late(struct drm_i915_private *dev_priv);
 extern void i915_perf_fini(struct drm_i915_private *dev_priv);
 extern void i915_perf_register(struct drm_i915_private *dev_priv);
 extern void i915_perf_unregister(struct drm_i915_private *dev_priv);
diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
index 3e39d69..9bf0f90 100644
--- a/drivers/gpu/drm/i915/i915_perf.c
+++ b/drivers/gpu/drm/i915/i915_perf.c
@@ -189,6 +189,7 @@ 
 
 #include <linux/anon_inodes.h>
 #include <linux/sizes.h>
+#include <linux/ktime.h>
 
 #include "i915_drv.h"
 #include "intel_ringbuffer.h"
@@ -228,6 +229,9 @@ 
 #define POLL_FREQUENCY 200
 #define POLL_PERIOD (NSEC_PER_SEC / POLL_FREQUENCY)
 
+#define MAX_CLK_SYNC_PERIOD (60*MSEC_PER_SEC)
+#define INIT_CLK_SYNC_PERIOD (20) /* in msecs */
+
 static u32 i915_perf_stream_paranoid = true;
 
 /* The maximum exponent the hardware accepts is 63 (essentially it selects one
@@ -254,13 +258,24 @@  static u32 i915_perf_stream_paranoid = true;
 #define TS_ADDR_ALIGN 8
 #define I915_PERF_TS_SAMPLE_SIZE 8
 
+/* Published frequency of GT command stream timestamp clock */
+#define FREQUENCY_12_5_MHZ	(12500000)
+#define FREQUENCY_12_0_MHZ	(12000000)
+#define FREQUENCY_19_2_MHZ	(19200000)
+#define GT_CS_TIMESTAMP_FREQUENCY(dev_priv) (IS_GEN9(dev_priv) ? \
+				(IS_BROXTON(dev_priv) ? \
+				FREQUENCY_19_2_MHZ : \
+				FREQUENCY_12_0_MHZ) : \
+				FREQUENCY_12_5_MHZ)
+
 /* Data common to all samples (periodic OA / CS based OA / Timestamps) */
 struct sample_data {
 	u32 source;
 	u32 ctx_id;
 	u32 pid;
 	u32 tag;
-	u64 ts;
+	u64 gpu_ts;
+	u64 clk_monoraw;
 	const u8 *report;
 };
 
@@ -319,6 +334,7 @@  static const enum intel_engine_id user_ring_map[I915_USER_RINGS + 1] = {
 #define SAMPLE_PID		(1<<3)
 #define SAMPLE_TAG		(1<<4)
 #define SAMPLE_TS		(1<<5)
+#define SAMPLE_CLK_MONO_RAW	(1<<6)
 
 struct perf_open_properties {
 	u32 sample_flags;
@@ -337,6 +353,136 @@  struct perf_open_properties {
 	enum intel_engine_id engine;
 };
 
+/**
+ * i915_tstamp_clk_cyclecounter_read - read raw cycle counter
+ * @cc: cyclecounter structure
+ **/
+static cycle_t i915_tstamp_clk_cyclecounter_read(
+				const struct cyclecounter *cc)
+{
+	struct drm_i915_private *dev_priv =
+		container_of(cc, typeof(*dev_priv),
+			     ts_clk_info.cc);
+
+	return I915_READ64_2x32(GT_TIMESTAMP_COUNT,
+					GT_TIMESTAMP_COUNT_UDW);
+}
+
+static void i915_tstamp_clk_calc_update_mult_shift(struct cyclecounter *cc,
+					u32 frequency)
+{
+	clocks_calc_mult_shift(&cc->mult, &cc->shift,
+			frequency, NSEC_PER_SEC, 3600);
+}
+
+static void i915_tstamp_clk_init_base_freq(struct drm_i915_private *dev_priv,
+				struct cyclecounter *cc)
+{
+	cc->read = i915_tstamp_clk_cyclecounter_read;
+	cc->mask = CYCLECOUNTER_MASK(64);
+	i915_tstamp_clk_calc_update_mult_shift(cc,
+			dev_priv->ts_clk_info.timestamp_frequency);
+}
+
+/**
+ * i915_get_syncdevicetime - Callback given to timekeeping code to read
+	device time and sytem clk counter value
+ * @device_time: current device time
+ * @system: system counter value read synchronously with device time
+ * @ctx: context provided by timekeeping code
+ *
+ **/
+static int i915_get_syncdevicetime(ktime_t *device_time,
+					 struct system_counterval_t *system,
+					 void *ctx)
+{
+	struct drm_i915_private *dev_priv = (struct drm_i915_private *)ctx;
+	struct timecounter *tc = &dev_priv->ts_clk_info.tc;
+	struct clocksource *curr_clksource;
+
+	*device_time = ns_to_ktime(timecounter_read(tc));
+
+	curr_clksource = get_current_clocksource();
+	system->cycles = curr_clksource->read(curr_clksource);
+	system->cs = curr_clksource;
+
+	return 0;
+}
+
+static void i915_perf_clock_sync_work(struct work_struct *work)
+{
+	struct drm_i915_private *dev_priv =
+		container_of(work, typeof(*dev_priv),
+		ts_clk_info.clk_sync_work.work);
+	struct system_device_crosststamp *xtstamp =
+			&dev_priv->ts_clk_info.xtstamp;
+	ktime_t last_sys_time = xtstamp->sys_monoraw;
+	ktime_t last_gpu_time = xtstamp->device;
+	ktime_t clk_mono_offset, gpu_time_offset;
+	s64 delta;
+	u32 gpu_freq = dev_priv->ts_clk_info.timestamp_frequency;
+	u32 freq_delta = 0;
+
+	get_device_system_crosststamp(i915_get_syncdevicetime, dev_priv,
+					NULL, xtstamp);
+
+	clk_mono_offset = ktime_sub(xtstamp->sys_monoraw, last_sys_time);
+	gpu_time_offset = ktime_sub(xtstamp->device, last_gpu_time);
+
+	/* delta time in ns */
+	delta = ktime_to_ns(ktime_sub(gpu_time_offset, clk_mono_offset));
+
+	/* If time delta < 1 us, we can assume gpu frequency is correct */
+	if (abs(delta) < NSEC_PER_USEC)
+		goto out;
+
+	/* The two clocks shouldn't deviate more than 1 second during the
+	 * resync period. If this is the case (which may happen due to
+	 * suspend/resume), then don't apply frequency correction, and
+	 * fast forward/rewind the clocks to resync immediately
+	 */
+	if (abs(delta) > NSEC_PER_SEC)
+		goto out;
+
+	/* Calculate frequency delta */
+	freq_delta = abs(delta)*gpu_freq;
+	do_div(freq_delta, ktime_to_ns(clk_mono_offset));
+
+	if (freq_delta == 0)
+		goto out;
+
+	if (delta < 0)
+		freq_delta = -freq_delta;
+
+	dev_priv->ts_clk_info.timestamp_frequency += freq_delta;
+	i915_tstamp_clk_calc_update_mult_shift(&dev_priv->ts_clk_info.cc,
+			dev_priv->ts_clk_info.timestamp_frequency);
+
+	/*
+	 * Get updated device/system times based on corrected frequency.
+	 * Note that this may cause jumps in device time depending on whether
+	 * frequency delta is positive or negative.
+	 * NB: Take care that monotonicity of sample timestamps is maintained
+	 * even with these jumps.
+	 */
+	get_device_system_crosststamp(i915_get_syncdevicetime, dev_priv,
+					NULL, xtstamp);
+
+out:
+	dev_priv->ts_clk_info.clk_offset = ktime_sub(xtstamp->sys_monoraw,
+						xtstamp->device);
+
+	/* We can schedule next synchronization at incrementally higher
+	 * durations, so that the accuracy of our calculated frequency
+	 * can improve over time. The max resync period is arbitrarily
+	 * set as one hour.
+	 */
+	dev_priv->ts_clk_info.resync_period *= 2;
+	if (dev_priv->ts_clk_info.resync_period < MAX_CLK_SYNC_PERIOD)
+		schedule_delayed_work(&dev_priv->ts_clk_info.clk_sync_work,
+			msecs_to_jiffies(dev_priv->ts_clk_info.resync_period));
+}
+
 /*
  * Emit the commands to capture metrics, into the command stream. This function
  * can be called concurrently with the stream operations and doesn't require
@@ -411,7 +557,7 @@  static int insert_perf_entry(struct drm_i915_private *dev_priv,
 
 	if (stream->sample_flags & SAMPLE_OA_REPORT)
 		entry_size += dev_priv->perf.oa.oa_buffer.format_size;
-	else if (sample_flags & SAMPLE_TS) {
+	else if (sample_flags & (SAMPLE_TS|SAMPLE_CLK_MONO_RAW)) {
 		/*
 		 * XXX: Since TS data can anyways be derived from OA report, so
 		 * no need to capture it for RCS engine, if capture oa data is
@@ -627,7 +773,7 @@  static void i915_ring_stream_cs_hook(struct i915_perf_stream *stream,
 		ret = i915_ring_stream_capture_oa(req, entry->oa_offset);
 		if (ret)
 			goto err_unref;
-	} else if (sample_flags & SAMPLE_TS) {
+	} else if (sample_flags & (SAMPLE_TS|SAMPLE_CLK_MONO_RAW)) {
 		/*
 		 * XXX: Since TS data can anyways be derived from OA report, so
 		 * no need to capture it for RCS engine, if capture oa data is
@@ -887,7 +1033,14 @@  static int append_sample(struct i915_perf_stream *stream,
 	}
 
 	if (sample_flags & SAMPLE_TS) {
-		if (copy_to_user(buf, &data->ts, I915_PERF_TS_SAMPLE_SIZE))
+		if (copy_to_user(buf, &data->gpu_ts, I915_PERF_TS_SAMPLE_SIZE))
+			return -EFAULT;
+		buf += I915_PERF_TS_SAMPLE_SIZE;
+	}
+
+	if (sample_flags & SAMPLE_CLK_MONO_RAW) {
+		if (copy_to_user(buf, &data->clk_monoraw,
+					I915_PERF_TS_SAMPLE_SIZE))
 			return -EFAULT;
 		buf += I915_PERF_TS_SAMPLE_SIZE;
 	}
@@ -903,6 +1056,27 @@  static int append_sample(struct i915_perf_stream *stream,
 	return 0;
 }
 
+static u64 get_clk_monoraw_from_gpu_ts(struct i915_perf_stream *stream,
+					u64 gpu_ts)
+{
+	struct drm_i915_private *dev_priv = stream->dev_priv;
+	struct timecounter *tc = &dev_priv->ts_clk_info.tc;
+	u64 gpu_time, clk_monoraw;
+
+	gpu_time = cyclecounter_cyc2ns(tc->cc, gpu_ts, tc->mask, &tc->frac);
+
+	clk_monoraw = gpu_time + ktime_to_ns(dev_priv->ts_clk_info.clk_offset);
+
+	/* Ensure monotonicity by clamping the system time in case it goes
+	 * backwards.
+	 */
+	if (clk_monoraw < stream->last_sample_ts)
+		clk_monoraw = stream->last_sample_ts;
+
+	stream->last_sample_ts = clk_monoraw;
+	return clk_monoraw;
+}
+
 static u64 get_gpu_ts_from_oa_report(struct drm_i915_private *dev_priv,
 					const u8 *report)
 {
@@ -960,7 +1134,13 @@  static int append_oa_buffer_sample(struct i915_perf_stream *stream,
 
 	/* Derive timestamp from OA report */
 	if (sample_flags & SAMPLE_TS)
-		data.ts = get_gpu_ts_from_oa_report(dev_priv, report);
+		data.gpu_ts = get_gpu_ts_from_oa_report(dev_priv, report);
+
+	if (sample_flags & SAMPLE_CLK_MONO_RAW) {
+		u64 gpu_ts = get_gpu_ts_from_oa_report(dev_priv, report);
+
+		data.clk_monoraw = get_clk_monoraw_from_gpu_ts(stream, gpu_ts);
+	}
 
 	if (sample_flags & SAMPLE_OA_REPORT)
 		data.report = report;
@@ -1478,7 +1658,7 @@  static int append_one_cs_sample(struct i915_perf_stream *stream,
 		if (ret)
 			return ret;
 
-		if (sample_flags & SAMPLE_TS)
+		if (sample_flags & (SAMPLE_TS|SAMPLE_CLK_MONO_RAW))
 			gpu_ts = get_gpu_ts_from_oa_report(dev_priv, report);
 	}
 
@@ -1500,7 +1680,7 @@  static int append_one_cs_sample(struct i915_perf_stream *stream,
 		dev_priv->perf.last_tag = node->tag;
 	}
 
-	if (sample_flags & SAMPLE_TS) {
+	if (sample_flags & (SAMPLE_TS|SAMPLE_CLK_MONO_RAW)) {
 		/* If OA sampling is enabled, derive the ts from OA report.
 		 * Else, forward the timestamp collected via command stream.
 		 */
@@ -1508,7 +1688,12 @@  static int append_one_cs_sample(struct i915_perf_stream *stream,
 			gpu_ts = *(u64 *)
 				(dev_priv->perf.command_stream_buf[id].addr +
 					node->ts_offset);
-		data.ts = gpu_ts;
+
+		if (sample_flags & SAMPLE_TS)
+			data.gpu_ts = gpu_ts;
+		if (sample_flags & SAMPLE_CLK_MONO_RAW)
+			data.clk_monoraw = get_clk_monoraw_from_gpu_ts(
+							stream,	gpu_ts);
 	}
 
 	return append_sample(stream, buf, count, offset, &data);
@@ -2299,12 +2484,28 @@  static void i915_ring_stream_enable(struct i915_perf_stream *stream)
 	if (stream->sample_flags & SAMPLE_OA_REPORT) {
 		dev_priv->perf.oa.ops.oa_enable(dev_priv);
 
-		if (stream->sample_flags & SAMPLE_TS)
+		if (stream->sample_flags & (SAMPLE_TS|SAMPLE_CLK_MONO_RAW))
 			dev_priv->perf.oa.last_gpu_ts =
 				I915_READ64_2x32(GT_TIMESTAMP_COUNT,
 					GT_TIMESTAMP_COUNT_UDW);
 	}
 
+	if (stream->sample_flags & SAMPLE_CLK_MONO_RAW) {
+		struct system_device_crosststamp *xtstamp =
+			&dev_priv->ts_clk_info.xtstamp;
+
+		get_device_system_crosststamp(i915_get_syncdevicetime, dev_priv,
+					NULL, xtstamp);
+		dev_priv->ts_clk_info.clk_offset = ktime_sub(xtstamp->sys_monoraw,
+						xtstamp->device);
+
+		if (dev_priv->ts_clk_info.resync_period < MAX_CLK_SYNC_PERIOD)
+			schedule_delayed_work(
+				&dev_priv->ts_clk_info.clk_sync_work,
+				msecs_to_jiffies(
+					dev_priv->ts_clk_info.resync_period));
+	}
+
 	if (stream->cs_mode || dev_priv->perf.oa.periodic)
 		hrtimer_start(&dev_priv->perf.poll_check_timer,
 			      ns_to_ktime(POLL_PERIOD),
@@ -2325,6 +2526,8 @@  static void i915_ring_stream_disable(struct i915_perf_stream *stream)
 {
 	struct drm_i915_private *dev_priv = stream->dev_priv;
 
+	cancel_delayed_work_sync(&dev_priv->ts_clk_info.clk_sync_work);
+
 	if (stream->cs_mode || dev_priv->perf.oa.periodic)
 		hrtimer_cancel(&dev_priv->perf.poll_check_timer);
 
@@ -2340,7 +2543,7 @@  static void i915_ring_stream_disable(struct i915_perf_stream *stream)
 static u64 oa_exponent_to_ns(struct drm_i915_private *dev_priv, int exponent)
 {
 	return 1000000000ULL * (2ULL << exponent) /
-		dev_priv->perf.oa.timestamp_frequency;
+		dev_priv->ts_clk_info.timestamp_frequency;
 }
 
 static const struct i915_perf_stream_ops i915_oa_stream_ops = {
@@ -2363,7 +2566,8 @@  static int i915_ring_stream_init(struct i915_perf_stream *stream,
 	bool require_cs_mode = props->sample_flags & (SAMPLE_PID |
 						      SAMPLE_TAG);
 	bool cs_sample_data = props->sample_flags & (SAMPLE_OA_REPORT |
-							SAMPLE_TS);
+							SAMPLE_TS |
+							SAMPLE_CLK_MONO_RAW);
 	int ret;
 
 	if ((props->sample_flags & SAMPLE_CTX_ID) && !props->cs_mode) {
@@ -2525,6 +2729,19 @@  static int i915_ring_stream_init(struct i915_perf_stream *stream,
 			require_cs_mode = true;
 	}
 
+	if (props->sample_flags & SAMPLE_CLK_MONO_RAW) {
+		stream->sample_flags |= SAMPLE_CLK_MONO_RAW;
+		stream->sample_size += I915_PERF_TS_SAMPLE_SIZE;
+
+		/*
+		 * NB: it's meaningful to request SAMPLE_CLK_MONO with just CS
+		 * mode or periodic OA mode sampling but we don't allow
+		 * SAMPLE_CLK_MONO without either mode
+		 */
+		if (!require_oa_unit)
+			require_cs_mode = true;
+	}
+
 	if (require_cs_mode && !props->cs_mode) {
 		DRM_ERROR(
 			"PID, TAG or TS sampling require a ring to be specified");
@@ -2549,11 +2766,13 @@  static int i915_ring_stream_init(struct i915_perf_stream *stream,
 
 		/*
 		 * The only time we should allow enabling CS mode if it's not
-		 * strictly required, is if SAMPLE_CTX_ID  or SAMPLE_TS has been
-		 * requested, as they're usable with periodic OA or CS sampling.
+		 * strictly required, is if SAMPLE_CTX_ID, SAMPLE_TS or
+		 * SAMPLE_CLK_MONO_RAW has been requested, as they're usable
+		 * with periodic OA or CS sampling.
 		 */
 		if (!require_cs_mode &&
-		    !(props->sample_flags & (SAMPLE_CTX_ID|SAMPLE_TS))) {
+		    !(props->sample_flags &
+				(SAMPLE_CTX_ID|SAMPLE_TS|SAMPLE_CLK_MONO_RAW))) {
 			DRM_ERROR(
 				"Ring given without requesting any CS specific property");
 			ret = -EINVAL;
@@ -3247,6 +3466,9 @@  static int read_properties_unlocked(struct drm_i915_private *dev_priv,
 		case DRM_I915_PERF_PROP_SAMPLE_TS:
 			props->sample_flags |= SAMPLE_TS;
 			break;
+		case DRM_I915_PERF_PROP_SAMPLE_CLOCK_MONOTONIC_RAW:
+			props->sample_flags |= SAMPLE_CLK_MONO_RAW;
+			break;
 		case DRM_I915_PERF_PROP_MAX:
 			BUG();
 		}
@@ -3394,6 +3616,7 @@  static struct ctl_table dev_root[] = {
 
 void i915_perf_init(struct drm_i915_private *dev_priv)
 {
+	struct i915_clock_info *clk_info = &dev_priv->ts_clk_info;
 	int i;
 
 	if (!(IS_HASWELL(dev_priv) ||
@@ -3401,6 +3624,12 @@  void i915_perf_init(struct drm_i915_private *dev_priv)
 	      IS_SKYLAKE(dev_priv)))
 		return;
 
+	clk_info->timestamp_frequency = GT_CS_TIMESTAMP_FREQUENCY(dev_priv);
+	clk_info->resync_period = INIT_CLK_SYNC_PERIOD;
+	INIT_DELAYED_WORK(&clk_info->clk_sync_work, i915_perf_clock_sync_work);
+
+	i915_tstamp_clk_init_base_freq(dev_priv, &clk_info->cc);
+
 	hrtimer_init(&dev_priv->perf.poll_check_timer,
 		     CLOCK_MONOTONIC, HRTIMER_MODE_REL);
 	dev_priv->perf.poll_check_timer.function = poll_check_timer_cb;
@@ -3430,8 +3659,6 @@  void i915_perf_init(struct drm_i915_private *dev_priv)
 		dev_priv->perf.oa.ops.oa_buffer_get_ctx_id =
 					gen7_oa_buffer_get_ctx_id;
 
-		dev_priv->perf.oa.timestamp_frequency = 12500000;
-
 		dev_priv->perf.oa.oa_formats = hsw_oa_formats;
 
 		dev_priv->perf.oa.n_builtin_sets =
@@ -3462,8 +3689,6 @@  void i915_perf_init(struct drm_i915_private *dev_priv)
 			dev_priv->perf.oa.ctx_flexeu0_off = 0x2ce;
 			dev_priv->perf.oa.n_builtin_sets =
 				i915_oa_n_builtin_metric_sets_bdw;
-
-			dev_priv->perf.oa.timestamp_frequency = 12500000;
 		} else if (IS_CHERRYVIEW(dev_priv)) {
 			dev_priv->perf.oa.ops.enable_metric_set =
 				chv_enable_metric_set;
@@ -3473,8 +3698,6 @@  void i915_perf_init(struct drm_i915_private *dev_priv)
 			dev_priv->perf.oa.ctx_flexeu0_off = 0x2ce;
 			dev_priv->perf.oa.n_builtin_sets =
 				i915_oa_n_builtin_metric_sets_chv;
-
-			dev_priv->perf.oa.timestamp_frequency = 12500000;
 		} else if (IS_SKYLAKE(dev_priv)) {
 			dev_priv->perf.oa.ops.enable_metric_set =
 				skl_enable_metric_set;
@@ -3484,8 +3707,6 @@  void i915_perf_init(struct drm_i915_private *dev_priv)
 			dev_priv->perf.oa.ctx_flexeu0_off = 0x3de;
 			dev_priv->perf.oa.n_builtin_sets =
 				i915_oa_n_builtin_metric_sets_skl;
-
-			dev_priv->perf.oa.timestamp_frequency = 12000000;
 		}
 	}
 
@@ -3494,6 +3715,14 @@  void i915_perf_init(struct drm_i915_private *dev_priv)
 	dev_priv->perf.initialized = true;
 }
 
+void i915_perf_init_late(struct drm_i915_private *dev_priv)
+{
+	struct i915_clock_info *clk_info = &dev_priv->ts_clk_info;
+
+	timecounter_init(&clk_info->tc,	&clk_info->cc,
+				ktime_to_ns(ktime_get_boottime()));
+}
+
 void i915_perf_fini(struct drm_i915_private *dev_priv)
 {
 	if (!dev_priv->perf.initialized)
diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
index 35c2814..096bad8 100644
--- a/include/uapi/drm/i915_drm.h
+++ b/include/uapi/drm/i915_drm.h
@@ -1330,6 +1330,12 @@  enum drm_i915_perf_property_id {
 	 */
 	DRM_I915_PERF_PROP_SAMPLE_TS,
 
+	/**
+	 * This property requests inclusion of CLOCK_MONOTONIC system time in
+	 * the perf sample data.
+	 */
+	DRM_I915_PERF_PROP_SAMPLE_CLOCK_MONOTONIC_RAW,
+
 	DRM_I915_PERF_PROP_MAX /* non-ABI */
 };
 
@@ -1398,7 +1404,8 @@  enum drm_i915_perf_record_type {
 	 *     { u32 ctx_id; } && DRM_I915_PERF_PROP_SAMPLE_CTX_ID
 	 *     { u32 pid; } && DRM_I915_PERF_PROP_SAMPLE_PID
 	 *     { u32 tag; } && DRM_I915_PERF_PROP_SAMPLE_TAG
-	 *     { u64 timestamp; } && DRM_I915_PERF_PROP_SAMPLE_TS
+	 *     { u64 gpu_ts; } && DRM_I915_PERF_PROP_SAMPLE_TS
+	 *     { u64 clk_mono; } && DRM_I915_PERF_PROP_SAMPLE_CLOCK_MONOTONIC
 	 *     { u32 oa_report[]; } && DRM_I915_PERF_PROP_SAMPLE_OA
 	 * };
 	 */