[PATCH] drm/i915/display: Introduce Display Metrics info

Kumar, Naveen1 naveen1.kumar at intel.com
Tue May 7 03:19:56 UTC 2024



>-----Original Message-----
>From: Vivi, Rodrigo <rodrigo.vivi at intel.com>
>Sent: Monday, May 6, 2024 10:52 PM
>To: Kumar, Naveen1 <naveen1.kumar at intel.com>
>Cc: intel-gfx at lists.freedesktop.org; Shankar, Uma <uma.shankar at intel.com>;
>Kulkarni, Vandita <vandita.kulkarni at intel.com>; Nikula, Jani
><jani.nikula at intel.com>; Belgaumkar, Vinay <vinay.belgaumkar at intel.com>;
>Borah, Chaitanya Kumar <chaitanya.kumar.borah at intel.com>
>Subject: Re: [PATCH] drm/i915/display: Introduce Display Metrics info
>
>On Mon, May 06, 2024 at 06:03:17AM -0400, Kumar, Naveen1 wrote:
>>
>>
>> >-----Original Message-----
>> >From: Vivi, Rodrigo <rodrigo.vivi at intel.com>
>> >Sent: Saturday, April 6, 2024 3:39 AM
>> >To: intel-gfx at lists.freedesktop.org
>> >Cc: Vivi, Rodrigo <rodrigo.vivi at intel.com>; Shankar, Uma
>> ><uma.shankar at intel.com>; Kulkarni, Vandita
>> ><vandita.kulkarni at intel.com>; Kumar, Naveen1
>> ><naveen1.kumar at intel.com>; Nikula, Jani <jani.nikula at intel.com>;
>> >Belgaumkar, Vinay <vinay.belgaumkar at intel.com>
>> >Subject: [PATCH] drm/i915/display: Introduce Display Metrics info
>> >
>> >Introduce a display metrics information through debugfs for a better
>> >view of the vblank and page flips, in special Async flips behavior.
>> >
>> >There is currently an overall expectation that whenever
>> >vblank_mode=0 is used with an graphics application, that
>> >automatically async_flips are happening. However, while implementing
>> >the Display Metrics for GuC SLPC, it was observed that it is not
>> >necessarily true for many of the expected cases.
>> >
>> >So, while the GuC SLPC side of the metrics doesn't get ready, let's
>> >at least bring the debugfs view of it so we can work to understand
>> >and fix any potential issue around our async vblanks.
>> >
>> >Please notice that the every struct here follows exactly the GuC
>> >shared data buffer, so the next step of the integration would be
>> >smooth and almost transparent to this intel_metrics on the display side.
>> >
>> >Cc: Uma Shankar <uma.shankar at intel.com>
>> >Cc: Vandita Kulkarni <vandita.kulkarni at intel.com>
>> >Cc: Naveen Kumar <naveen1.kumar at intel.com>
>> >Cc: Jani Nikula <jani.nikula at intel.com>
>> >Cc: Vinay Belgaumkar <vinay.belgaumkar at intel.com>
>> >Signed-off-by: Rodrigo Vivi <rodrigo.vivi at intel.com>
>> >---
>> > drivers/gpu/drm/i915/Makefile                 |   1 +
>> > drivers/gpu/drm/i915/display/intel_display.c  |  15 +-
>> > .../gpu/drm/i915/display/intel_display_core.h |   2 +
>> > .../drm/i915/display/intel_display_debugfs.c  |  12 +
>> > .../drm/i915/display/intel_display_driver.c   |   5 +
>> > .../gpu/drm/i915/display/intel_display_irq.c  |   3 +
>> > drivers/gpu/drm/i915/display/intel_metrics.c  | 356
>> >++++++++++++++++++ drivers/gpu/drm/i915/display/intel_metrics.h  |  27
>++
>> > .../drm/i915/display/skl_universal_plane.c    |   3 +
>> > drivers/gpu/drm/xe/Makefile                   |   1 +
>> > 10 files changed, 424 insertions(+), 1 deletion(-)  create mode
>> >100644 drivers/gpu/drm/i915/display/intel_metrics.c
>> > create mode 100644 drivers/gpu/drm/i915/display/intel_metrics.h
>> >
>> >diff --git a/drivers/gpu/drm/i915/Makefile
>> >b/drivers/gpu/drm/i915/Makefile index af9e871daf1d..a3c8d9f5614c
>> >100644
>> >--- a/drivers/gpu/drm/i915/Makefile
>> >+++ b/drivers/gpu/drm/i915/Makefile
>> >@@ -291,6 +291,7 @@ i915-y += \
>> > 	display/intel_link_bw.o \
>> > 	display/intel_load_detect.o \
>> > 	display/intel_lpe_audio.o \
>> >+	display/intel_metrics.o \
>> > 	display/intel_modeset_lock.o \
>> > 	display/intel_modeset_setup.o \
>> > 	display/intel_modeset_verify.o \
>> >diff --git a/drivers/gpu/drm/i915/display/intel_display.c
>> >b/drivers/gpu/drm/i915/display/intel_display.c
>> >index a481c9218138..ca30b8d48e1f 100644
>> >--- a/drivers/gpu/drm/i915/display/intel_display.c
>> >+++ b/drivers/gpu/drm/i915/display/intel_display.c
>> >@@ -94,6 +94,7 @@
>> > #include "intel_link_bw.h"
>> > #include "intel_lvds.h"
>> > #include "intel_lvds_regs.h"
>> >+#include "intel_metrics.h"
>> > #include "intel_modeset_setup.h"
>> > #include "intel_modeset_verify.h"
>> > #include "intel_overlay.h"
>> >@@ -1021,11 +1022,15 @@ static void intel_post_plane_update(struct
>> >intel_atomic_state *state,
>> > 				    struct intel_crtc *crtc)
>> > {
>> > 	struct drm_i915_private *dev_priv = to_i915(state->base.dev);
>> >+	struct intel_display *display = &dev_priv->display;
>> > 	const struct intel_crtc_state *old_crtc_state =
>> > 		intel_atomic_get_old_crtc_state(state, crtc);
>> > 	const struct intel_crtc_state *new_crtc_state =
>> > 		intel_atomic_get_new_crtc_state(state, crtc);
>> > 	enum pipe pipe = crtc->pipe;
>> >+	const struct intel_plane_state __maybe_unused *plane_state;
>> >+	struct intel_plane *plane;
>> >+	int i;
>> >
>> > 	intel_psr_post_plane_update(state, crtc);
>> >
>> >@@ -1057,6 +1062,12 @@ static void intel_post_plane_update(struct
>> >intel_atomic_state *state,
>> >
>> > 	if (audio_enabling(old_crtc_state, new_crtc_state))
>> > 		intel_encoders_audio_enable(state, crtc);
>> >+
>> >+	if (!new_crtc_state->do_async_flip) {
>> >+		for_each_new_intel_plane_in_state(state, plane, plane_state, i)
>> >+			intel_metrics_flip(display, new_crtc_state, plane->id,
>> >+					   false);
>> >+	}
>> > }
>> >
>> > static void intel_crtc_enable_flip_done(struct intel_atomic_state
>> >*state, @@ -
>> >7139,6 +7150,7 @@ static void intel_atomic_commit_tail(struct
>> >intel_atomic_state *state)  {
>> > 	struct drm_device *dev = state->base.dev;
>> > 	struct drm_i915_private *dev_priv = to_i915(dev);
>> >+	struct intel_display *display = &dev_priv->display;
>> > 	struct intel_crtc_state *new_crtc_state, *old_crtc_state;
>> > 	struct intel_crtc *crtc;
>> > 	struct intel_power_domain_mask put_domains[I915_MAX_PIPES] = {};
>@@
>> >-7261,7 +7273,6 @@ static void intel_atomic_commit_tail(struct
>> >intel_atomic_state *state)
>> > 	for_each_new_intel_crtc_in_state(state, crtc, new_crtc_state, i) {
>> > 		if (new_crtc_state->do_async_flip)
>> > 			intel_crtc_disable_flip_done(state, crtc);
>> >-
>> > 		intel_color_wait_commit(new_crtc_state);
>> > 	}
>> >
>> >@@ -7314,6 +7325,8 @@ static void intel_atomic_commit_tail(struct
>> >intel_atomic_state *state)
>> > 		 * FIXME get rid of this funny new->old swapping
>> > 		 */
>> > 		old_crtc_state->dsb = fetch_and_zero(&new_crtc_state->dsb);
>> >+
>> >+		intel_metrics_refresh_info(display, new_crtc_state);
>> > 	}
>> >
>> > 	/* Underruns don't always raise interrupts, so check manually */
>> >diff -- git a/drivers/gpu/drm/i915/display/intel_display_core.h
>> >b/drivers/gpu/drm/i915/display/intel_display_core.h
>> >index 2167dbee5eea..992db9098566 100644
>> >--- a/drivers/gpu/drm/i915/display/intel_display_core.h
>> >+++ b/drivers/gpu/drm/i915/display/intel_display_core.h
>> >@@ -42,6 +42,7 @@ struct intel_cdclk_vals;  struct intel_color_funcs;
>> >struct intel_crtc;  struct intel_crtc_state;
>> >+struct intel_display_metrics;
>> > struct intel_dmc;
>> > struct intel_dpll_funcs;
>> > struct intel_dpll_mgr;
>> >@@ -530,6 +531,7 @@ struct intel_display {
>> > 	struct intel_fbc *fbc[I915_MAX_FBCS];
>> > 	struct intel_frontbuffer_tracking fb_tracking;
>> > 	struct intel_hotplug hotplug;
>> >+	struct intel_display_metrics *metrics;
>> > 	struct intel_opregion *opregion;
>> > 	struct intel_overlay *overlay;
>> > 	struct intel_display_params params; diff --git
>> >a/drivers/gpu/drm/i915/display/intel_display_debugfs.c
>> >b/drivers/gpu/drm/i915/display/intel_display_debugfs.c
>> >index 3e364891dcd0..f05b9a9ddee0 100644
>> >--- a/drivers/gpu/drm/i915/display/intel_display_debugfs.c
>> >+++ b/drivers/gpu/drm/i915/display/intel_display_debugfs.c
>> >@@ -30,6 +30,7 @@
>> > #include "intel_hdcp.h"
>> > #include "intel_hdmi.h"
>> > #include "intel_hotplug.h"
>> >+#include "intel_metrics.h"
>> > #include "intel_panel.h"
>> > #include "intel_psr.h"
>> > #include "intel_psr_regs.h"
>> >@@ -642,6 +643,16 @@ static int i915_display_capabilities(struct
>> >seq_file *m, void *unused)
>> > 	return 0;
>> > }
>> >
>> >+static int i915_display_metrics(struct seq_file *m, void *unused) {
>> >+	struct drm_i915_private *i915 = node_to_i915(m->private);
>> >+	struct drm_printer p = drm_seq_file_printer(m);
>> >+
>> >+	intel_metrics_show(&i915->display, &p);
>> >+
>> >+	return 0;
>> >+}
>> >+
>> > static int i915_shared_dplls_info(struct seq_file *m, void *unused)  {
>> > 	struct drm_i915_private *dev_priv = node_to_i915(m->private); @@ -
>> >1059,6 +1070,7 @@ static const struct drm_info_list
>> >intel_display_debugfs_list[] = {
>> > 	{"i915_power_domain_info", i915_power_domain_info, 0},
>> > 	{"i915_display_info", i915_display_info, 0},
>> > 	{"i915_display_capabilities", i915_display_capabilities, 0},
>> >+	{"i915_display_metrics", i915_display_metrics, 0},
>> > 	{"i915_shared_dplls_info", i915_shared_dplls_info, 0},
>> > 	{"i915_dp_mst_info", i915_dp_mst_info, 0},
>> > 	{"i915_ddb_info", i915_ddb_info, 0}, diff --git
>> >a/drivers/gpu/drm/i915/display/intel_display_driver.c
>> >b/drivers/gpu/drm/i915/display/intel_display_driver.c
>> >index 87dd07e0d138..767b2d5b3826 100644
>> >--- a/drivers/gpu/drm/i915/display/intel_display_driver.c
>> >+++ b/drivers/gpu/drm/i915/display/intel_display_driver.c
>> >@@ -46,6 +46,7 @@
>> > #include "intel_hdcp.h"
>> > #include "intel_hotplug.h"
>> > #include "intel_hti.h"
>> >+#include "intel_metrics.h"
>> > #include "intel_modeset_lock.h"
>> > #include "intel_modeset_setup.h"
>> > #include "intel_opregion.h"
>> >@@ -514,6 +515,8 @@ int intel_display_driver_probe(struct
>> >drm_i915_private *i915)
>> >
>> > 	intel_overlay_setup(i915);
>> >
>> >+	intel_metrics_init(&i915->display);
>> >+
>> > 	ret = intel_fbdev_init(&i915->drm);
>> > 	if (ret)
>> > 		return ret;
>> >@@ -611,6 +614,8 @@ void intel_display_driver_remove_noirq(struct
>> >drm_i915_private *i915)
>> >
>> > 	intel_dp_tunnel_mgr_cleanup(i915);
>> >
>> >+	intel_metrics_fini(&i915->display);
>> >+
>> > 	intel_overlay_cleanup(i915);
>> >
>> > 	intel_gmbus_teardown(i915);
>> >diff --git a/drivers/gpu/drm/i915/display/intel_display_irq.c
>> >b/drivers/gpu/drm/i915/display/intel_display_irq.c
>> >index f846c5b108b5..202400a771b2 100644
>> >--- a/drivers/gpu/drm/i915/display/intel_display_irq.c
>> >+++ b/drivers/gpu/drm/i915/display/intel_display_irq.c
>> >@@ -18,6 +18,7 @@
>> > #include "intel_fifo_underrun.h"
>> > #include "intel_gmbus.h"
>> > #include "intel_hotplug_irq.h"
>> >+#include "intel_metrics.h"
>> > #include "intel_pmdemand.h"
>> > #include "intel_psr.h"
>> > #include "intel_psr_regs.h"
>> >@@ -25,8 +26,10 @@
>> > static void
>> > intel_handle_vblank(struct drm_i915_private *dev_priv, enum pipe
>> >pipe)  {
>> >+	struct intel_display *display = &dev_priv->display;
>> > 	struct intel_crtc *crtc = intel_crtc_for_pipe(dev_priv, pipe);
>> >
>> >+	intel_metrics_vblank(display, crtc);
>> > 	drm_crtc_handle_vblank(&crtc->base);
>> > }
>> >
>> >diff --git a/drivers/gpu/drm/i915/display/intel_metrics.c
>> >b/drivers/gpu/drm/i915/display/intel_metrics.c
>> >new file mode 100644
>> >index 000000000000..2cb2b8189b25
>> >--- /dev/null
>> >+++ b/drivers/gpu/drm/i915/display/intel_metrics.c
>> >@@ -0,0 +1,356 @@
>> >+// SPDX-License-Identifier: MIT
>> >+/*
>> >+ * Copyright (c) 2024 Intel Corporation  */
>> >+
>> >+#include "intel_metrics.h"
>> >+
>> >+#include <drm/drm_modes.h>
>> >+#include <drm/drm_print.h>
>> >+
>> >+#include "i915_drv.h"
>> >+#include "intel_de.h"
>> >+#include "intel_display_types.h"
>> >+
>> >+/**
>> >+ * Display Metrics
>> >+ *
>> >+ * Provide some display activity overview such as active refresh
>> >+rates,
>> >+ * vblank activity and page flip activities.
>> >+ * For now it is informative debug only, but later it will be
>> >+expanded
>> >+ * to be used for GT frequency selection by GuC SLPC.
>> >+ */
>> >+
>> >+/*
>> >+ * An event using an work queue is used to avoid any disturbance in
>> >+the
>> >+ * critical path that could cause performance impacts.
>> >+ */
>> >+struct display_event {
>> >+	struct work_struct work;
>> >+	struct drm_i915_private *i915;
>> >+	struct intel_display *display;
>> >+	bool is_vblank;
>> >+	int pipe;
>> >+	int plane;
>> >+	bool async_flip;
>> >+};
>> >+
>> >+/*
>> >+ * Although we could simply save this inside our crtc structs, we
>> >+are
>> >+ * already mimicking the GuC SLPC defition of the display data, for
>> >+future
>> >+ * usage.
>> >+ */
>> >+#define MAX_PIPES		8
>> >+#define MAX_PLANES_PER_PIPE	8
>> >+
>> >+struct display_global_info {
>> >+	u32 version:8;
>> >+	u32 num_pipes:4;
>> >+	u32 num_planes_per_pipe:4;
>> >+	u32 reserved_1:16;
>> >+	u32 refresh_count:16;
>> >+	u32 vblank_count:16;
>> >+	u32 flip_count:16;
>> >+	u32 reserved_2:16;
>> >+	u32 reserved_3[13];
>> >+} __packed;
>> >+
>> >+struct display_refresh_info {
>> >+	u32 refresh_interval:16;
>> >+	u32 is_variable:1;
>> >+	u32 reserved:15;
>> >+} __packed;
>> >+
>> >+/*
>> >+ * When used with GuC SLPC, the host must update each 32-bit part
>> >+with a single
>> >+ * atomic write so that SLPC will read the contained bit fields together.
>> >+ * The host must update the two parts in order - total flip count
>> >+and timestamp
>> >+ * first, vsync and async flip counts second.
>> >+ * Hence, these items are not defined with individual bitfields.
>> >+ */
>> >+#define FLIP_P1_LAST		REG_GENMASK(31, 7)
>> >+#define FLIP_P1_TOTAL_COUNT	REG_GENMASK(6, 0)
>> >+#define FLIP_P2_ASYNC_COUNT	REG_GENMASK(31, 16)
>> >+#define FLIP_P2_VSYNC_COUNT	REG_GENMASK(15, 0)
>> >+
>> >+struct display_flip_metrics {
>> >+	u32 part1;
>> >+	u32 part2;
>> >+} __packed;
>> >+
>> >+/*
>> >+ * When used with GuC SLPC, the host must update each 32-bit part
>> >+with a single
>> >+ * atomic write, so that SLPC will read the count and timestamp together.
>> >+ * Hence, this item is not defined with individual bitfields.
>> >+ */
>> >+#define VBLANK_LAST	REG_GENMASK(31, 7)
>> >+#define VBLANK_COUNT	REG_GENMASK(6, 0)
>> >+
>> >+struct intel_display_metrics {
>> >+	struct display_global_info global_info;
>> >+	struct display_refresh_info refresh_info[MAX_PIPES];
>> >+	u32 vblank_metrics[MAX_PIPES];
>> >+	struct display_flip_metrics
>> >+	flip_metrics[MAX_PIPES][MAX_PLANES_PER_PIPE];
>> >+} __packed;
>> >+
>> >+/**
>> >+ * intel_metrics_refresh_info - Refresh rate information
>> >+ * @display: Pointer to the intel_display struct that is in use by the gfx
>device.
>> >+ * @crtc_state: New CRTC state upon a modeset.
>> >+ *
>> >+ * To be called on a modeset. It then saves the current refresh
>> >+interval in
>> >+ * micro seconds.
>> >+ */
>> >+void intel_metrics_refresh_info(struct intel_display *display,
>> >+				struct intel_crtc_state *crtc_state) {
>> >+	struct intel_display_metrics *metrics = display->metrics;
>> >+	struct intel_crtc *crtc = to_intel_crtc(crtc_state->uapi.crtc);
>> >+	struct drm_display_mode *mode = &crtc_state->hw.adjusted_mode;
>> >+	u32 interval_us;
>> >+
>> >+	if (!display->metrics)
>> >+		return;
>> >+
>> >+	interval_us = crtc_state->hw.active ? DIV_ROUND_UP(1000000,
>> >+						drm_mode_vrefresh(mode)) :
>> >0;
>> >+
>> >+	metrics->refresh_info[crtc->pipe].refresh_interval = interval_us;
>> >+	metrics->refresh_info[crtc->pipe].is_variable = crtc_state-
>> >>uapi.vrr_enabled;
>> >+	metrics->global_info.refresh_count += 1; }
>> >+
>> >+static void metrics_update_vblank(struct intel_display_metrics
>> >+*metrics, int
>> >pipe,
>> >+				  u32 timestamp)
>> >+{
>> >+	u32 vblank;
>> >+
>> >+	vblank = metrics->vblank_metrics[pipe];
>> >+
>> >+	vblank = REG_FIELD_GET(VBLANK_COUNT, vblank);
>> >+	vblank = REG_FIELD_PREP(VBLANK_COUNT, vblank + 1);
>> >+	vblank |= REG_FIELD_PREP(VBLANK_LAST, timestamp);
>> >+
>> >+	/* Write everything at once in preparation for the GuC SLPC
>> >requirement */
>> >+	metrics->vblank_metrics[pipe] = vblank;
>> >+	metrics->global_info.vblank_count += 1; }
>> >+
>> >+static void metrics_update_flip(struct intel_display_metrics *metrics, int
>pipe,
>> >+				int plane, bool async_flip, u32 timestamp) {
>> >+	u32 part1, part2, count;
>> >+
>> >+	part1 = metrics->flip_metrics[pipe][plane].part1;
>> >+	part2 = metrics->flip_metrics[pipe][plane].part2;
>> >+
>> >+	part1 = REG_FIELD_GET(FLIP_P1_TOTAL_COUNT, part1);
>> >+	part1 = REG_FIELD_PREP(FLIP_P1_TOTAL_COUNT, part1 + 1);
>> >+	part1 |= REG_FIELD_PREP(FLIP_P1_LAST, timestamp);
>> >+
>> >+	if (async_flip) {
>> >+		count = REG_FIELD_GET(FLIP_P2_ASYNC_COUNT, part2);
>> >+		part2 &= ~FLIP_P2_ASYNC_COUNT;
>> >+		part2 |= REG_FIELD_PREP(FLIP_P2_ASYNC_COUNT, count + 1);
>> >+	} else {
>> >+		count = REG_FIELD_GET(FLIP_P2_VSYNC_COUNT, part2);
>> >+		part2 &= ~FLIP_P2_VSYNC_COUNT;
>> >+		part2 |= REG_FIELD_PREP(FLIP_P2_VSYNC_COUNT, count + 1);
>> >+	}
>> >+
>> >+	/*
>> >+	 * Write everything in this way and this order in preparation for the
>> >+	 * GuC SLPC requirement
>> >+	 */
>> >+	metrics->flip_metrics[pipe][plane].part1 = part1;
>> >+	metrics->flip_metrics[pipe][plane].part2 = part2;
>> >+
>> >+	metrics->global_info.flip_count += 1; }
>> >+
>> >+/*
>> >+ * Let's use the same register GuC SLPC uses for timestamp.
>> >+ * It uses a register that is outside GT domain so GuC doesn't need
>> >+ * to wake the GT for reading during SLPC loop.
>> >+ * This is a single register regarding the GT, so we can read
>> >+directly
>> >+ * from here, regarding the GT GuC is in.
>> >+ */
>> >+#define MCHBAR_MIRROR_BASE_SNB	0x140000
>> >+#define MCHBAR_BCLK_COUNT	_MMIO(MCHBAR_MIRROR_BASE_SNB
>> >+ 0x5984)
>> >+#define MTL_BCLK_COUNT		_MMIO(0xc28)
>> >+#define   TIMESTAMP_MASK	REG_GENMASK(30, 6)
>> >+
>> >+static u32 bclk_read_timestamp(struct drm_i915_private *i915) {
>> >+	u32 timestamp;
>> >+
>> >+	if (DISPLAY_VER(i915) >= 14)
>> >+		timestamp = intel_de_read(i915, MTL_BCLK_COUNT);
>> >+	else
>> >+		timestamp = intel_de_read(i915, MCHBAR_BCLK_COUNT);
>> >+
>> >+	return REG_FIELD_GET(TIMESTAMP_MASK, timestamp); }
>> >+
>> >+static void display_event_work(struct work_struct *work) {
>> >+	struct display_event *event = container_of(work, struct
>> >+display_event,
>> >work);
>> >+	struct intel_display *display = event->display;
>> >+	u32 timestamp;
>> >+
>> >+	timestamp = bclk_read_timestamp(event->i915);
>> >+
>> >+	if (event->is_vblank) {
>> >+		metrics_update_vblank(display->metrics, event->pipe,
>> >timestamp);
>> >+	} else {
>> >+		metrics_update_flip(display->metrics, event->pipe, event-
>> >>plane,
>> >+				    event->async_flip, timestamp);
>> >+	}
>> >+
>> >+	kfree(event);
>> >+}
>> >+
>> >+void intel_metrics_init(struct intel_display *display) {
>> >+	struct intel_display_metrics *metrics;
>> >+
>> >+	metrics = kzalloc(sizeof(*metrics), GFP_KERNEL);
>> >+	if (!metrics)
>> >+		return;
>> >+
>> >+	metrics->global_info.version = 1;
>> >+	metrics->global_info.num_pipes = MAX_PIPES;
>> >+	metrics->global_info.num_planes_per_pipe = MAX_PLANES_PER_PIPE;
>> >+
>> >+	display->metrics = metrics;
>> >+}
>> >+
>> >+void intel_metrics_fini(struct intel_display *display) {
>> >+	if (!display->metrics)
>> >+		return;
>> >+
>> >+	kfree(display->metrics);
>> >+}
>> >+
>> >+/**
>> >+ * intel_metrics_vblank - Vblank information
>> >+ * @display: Pointer to the intel_display struct that is in use by the gfx
>device.
>> >+ * @crtc: The Intel CRTC that received the vblank interrupt.
>> >+ *
>> >+ * To be called when a vblank is passed.
>> >+ */
>> >+void intel_metrics_vblank(struct intel_display *display,
>> >+			  struct intel_crtc *crtc)
>> >+{
>> >+	struct display_event *event;
>> >+
>> >+	if (!display->metrics)
>> >+		return;
>> >+
>> >+	event = kmalloc(sizeof(*event), GFP_ATOMIC);
>>
>> HI Rodrigo, after changing kmalloc -> kzalloc, it works fine for us. It does
>correctly prints Global Flip count and Async Flip count.
>> Prior to this change, event->is_vblank in function display_event_work is
>always true and hence it never increases flip count.
>
>thanks for pointing that out. kzalloc is indeed better.
>There's also another kmalloc down that needs to be changed to kzalloc.
>
>Anyway, when you mentioned that you saw the async flip count increasing,
>you are telling about the igt test right?
>or do you have any special compositor change required?

Hi Rodrigo, async flip count is increasing using both IGT and Weston/wayland client app (https://gitlab.freedesktop.org/wayland/weston/-/blob/main/clients/simple-egl.c#L1294)
Additionally, we had to hack mesa code to use async flip supported modifiers. Mesa needs to handle async flip scenarios and select supported modifiers.

>In my environment here I still only see the async flip increasing with IGT or with
>a very limited cases...
>
>>
>> Thanks Chaitanya for pointing this out.
>> Regards,
>> Naveen Kumar
>>
>> >+	if (!event)
>> >+		return;
>> >+
>> >+	INIT_WORK(&event->work, display_event_work);
>> >+	event->i915 = to_i915(crtc->base.dev);
>> >+	event->display = display;
>> >+	event->is_vblank = true;
>> >+	event->pipe = crtc->pipe;
>> >+	queue_work(system_highpri_wq, &event->work); }
>> >+
>> >+/**
>> >+ * intel_metrics_flip - Flip information
>> >+ * @display: Pointer to the intel_display struct that is in use by the gfx
>device.
>> >+ * @crtc_state: New CRTC state upon a page flip.
>> >+ * @plane: Which plane ID got the page flip.
>> >+ * @async_flip: A boolean to indicate if the page flip was async.
>> >+ *
>> >+ * To be called on a page flip.
>> >+ */
>> >+void intel_metrics_flip(struct intel_display *display,
>> >+			const struct intel_crtc_state *crtc_state,
>> >+			int plane, bool async_flip)
>> >+{
>> >+	struct intel_crtc *crtc = to_intel_crtc(crtc_state->uapi.crtc);
>> >+	struct display_event *event;
>> >+
>> >+	if (!display->metrics)
>> >+		return;
>> >+
>> >+	event = kmalloc(sizeof(*event), GFP_ATOMIC);
>> Same here, had to change kmalloc -> kzalloc
>>
>> >+	if (!event)
>> >+		return;
>> >+
>> >+	INIT_WORK(&event->work, display_event_work);
>> >+	event->i915 = to_i915(crtc->base.dev);
>> >+	event->display = display;
>> >+	event->pipe = crtc->pipe;
>> >+	event->plane = plane;
>> >+	event->async_flip = async_flip;
>> >+	queue_work(system_highpri_wq, &event->work); }
>> >+
>> >+void intel_metrics_show(struct intel_display *display, struct
>> >+drm_printer *p) {
>> >+	struct intel_display_metrics *metrics = display->metrics;
>> >+	int pipe, plane;
>> >+	u32 val;
>> >+
>> >+	if (!metrics)
>> >+		return;
>> >+
>> >+	drm_printf(p, "\nDisplay Metrics - Globals:\n");
>> >+	drm_printf(p, "\tVersion: %d\n", metrics->global_info.version);
>> >+	drm_printf(p, "\tNum Pipes: %d\n", metrics->global_info.num_pipes);
>> >+	drm_printf(p, "\tNum Planes per Pipe: %d\n",
>> >+		   metrics->global_info.num_planes_per_pipe);
>> >+	drm_printf(p, "\tGlobal Refresh Info Count: %d\n",
>> >+		   metrics->global_info.refresh_count);
>> >+	drm_printf(p, "\tGlobal Vblank Count: %d\n",
>> >+		   metrics->global_info.vblank_count);
>> >+	drm_printf(p, "\tGlobal Flip Count: %d\n",
>> >+		   metrics->global_info.flip_count);
>> >+
>> >+	for (pipe = 0; pipe < MAX_PIPES; pipe++) {
>> >+		if (!metrics->refresh_info[pipe].refresh_interval)
>> >+			continue;
>> >+
>> >+		drm_printf(p, "\nDisplay Metrics - Refresh Info - Pipe[%d]:\n",
>> >+			   pipe);
>> >+		drm_printf(p, "\tRefresh Interval: %d\n",
>> >+			   metrics->refresh_info[pipe].refresh_interval);
>> >+		drm_printf(p, "\tIS VRR: %d\n",
>> >+			   metrics->refresh_info[pipe].is_variable);
>> >+
>> >+		drm_printf(p, "Display Metrics - Vblank Info - Pipe[%d]:\n",
>> >+			   pipe);
>> >+		val = metrics->vblank_metrics[pipe];
>> >+		drm_printf(p, "\tVBlank Last Timestamp: %x\n",
>> >+			   REG_FIELD_GET(VBLANK_LAST, val));
>> >+		drm_printf(p, "\tVBlank Count: %d\n",
>> >+			   REG_FIELD_GET(VBLANK_COUNT, val));
>> >+
>> >+		drm_printf(p, "Display Metrics - Flip Info - Pipe[%d]:\n", pipe);
>> >+		for (plane = 0; plane < MAX_PLANES_PER_PIPE; plane++) {
>> >+			val = metrics->flip_metrics[pipe][plane].part1;
>> >+			if (!val)
>> >+				continue;
>> >+
>> >+			drm_printf(p, "\tFlip Info - Plane[%d]:\n", plane);
>> >+			drm_printf(p, "\t\tFlip Last Timestamp: %x\n",
>> >+				   REG_FIELD_GET(FLIP_P1_LAST, val));
>> >+			drm_printf(p, "\t\tFlip Total Count: %d\n",
>> >+				   REG_FIELD_GET(FLIP_P1_TOTAL_COUNT,
>> >val));
>> >+
>> >+			val = metrics->flip_metrics[pipe][plane].part2;
>> >+
>> >+			drm_printf(p, "\t\tFlip Async Count: %d\n",
>> >+				   REG_FIELD_GET(FLIP_P2_ASYNC_COUNT,
>> >val));
>> >+			drm_printf(p, "\t\tFlip Vsync Count: %d\n",
>> >+				   REG_FIELD_GET(FLIP_P2_VSYNC_COUNT,
>> >val));
>> >+		}
>> >+	}
>> >+}
>> >diff --git a/drivers/gpu/drm/i915/display/intel_metrics.h
>> >b/drivers/gpu/drm/i915/display/intel_metrics.h
>> >new file mode 100644
>> >index 000000000000..9e41dc9522f3
>> >--- /dev/null
>> >+++ b/drivers/gpu/drm/i915/display/intel_metrics.h
>> >@@ -0,0 +1,27 @@
>> >+/* SPDX-License-Identifier: MIT */
>> >+/*
>> >+ * Copyright (c) 2024 Intel Corporation  */
>> >+
>> >+#ifndef __INTEL_METRICS_H__
>> >+#define __INTEL_METRICS_H__
>> >+
>> >+#include <linux/types.h>
>> >+
>> >+struct drm_printer;
>> >+struct intel_crtc;
>> >+struct intel_crtc_state;
>> >+struct intel_display;
>> >+
>> >+void intel_metrics_refresh_info(struct intel_display *display,
>> >+				struct intel_crtc_state *crtc_state); void
>> >+intel_metrics_vblank(struct intel_display *display,
>> >+			  struct intel_crtc *intel_crtc); void
>intel_metrics_flip(struct
>> >+intel_display *display,
>> >+			const struct intel_crtc_state *crtc_state,
>> >+			int plane, bool async_flip);
>> >+void intel_metrics_init(struct intel_display *display); void
>> >+intel_metrics_fini(struct intel_display *display); void
>> >+intel_metrics_show(struct intel_display *display, struct drm_printer
>> >+*p);
>> >+
>> >+#endif
>> >diff --git a/drivers/gpu/drm/i915/display/skl_universal_plane.c
>> >b/drivers/gpu/drm/i915/display/skl_universal_plane.c
>> >index 860574d04f88..fdd21a41d79d 100644
>> >--- a/drivers/gpu/drm/i915/display/skl_universal_plane.c
>> >+++ b/drivers/gpu/drm/i915/display/skl_universal_plane.c
>> >@@ -17,6 +17,7 @@
>> > #include "intel_fb.h"
>> > #include "intel_fbc.h"
>> > #include "intel_frontbuffer.h"
>> >+#include "intel_metrics.h"
>> > #include "intel_psr.h"
>> > #include "intel_psr_regs.h"
>> > #include "skl_scaler.h"
>> >@@ -1392,6 +1393,7 @@ skl_plane_async_flip(struct intel_plane *plane,
>> > 		     bool async_flip)
>> > {
>> > 	struct drm_i915_private *dev_priv = to_i915(plane->base.dev);
>> >+	struct intel_display *display = &dev_priv->display;
>> > 	enum plane_id plane_id = plane->id;
>> > 	enum pipe pipe = plane->pipe;
>> > 	u32 plane_ctl = plane_state->ctl;
>> >@@ -1404,6 +1406,7 @@ skl_plane_async_flip(struct intel_plane *plane,
>> > 	intel_de_write_fw(dev_priv, PLANE_CTL(pipe, plane_id), plane_ctl);
>> > 	intel_de_write_fw(dev_priv, PLANE_SURF(pipe, plane_id),
>> > 			  skl_plane_surf(plane_state, 0));
>> >+	intel_metrics_flip(display, crtc_state, plane_id, async_flip);
>> > }
>> >
>> > static bool intel_format_is_p01x(u32 format) diff --git
>> >a/drivers/gpu/drm/xe/Makefile b/drivers/gpu/drm/xe/Makefile index
>> >e5b1715f721e..5133dba2c7de 100644
>> >--- a/drivers/gpu/drm/xe/Makefile
>> >+++ b/drivers/gpu/drm/xe/Makefile
>> >@@ -263,6 +263,7 @@ xe-$(CONFIG_DRM_XE_DISPLAY) += \
>> > 	i915-display/intel_hti.o \
>> > 	i915-display/intel_link_bw.o \
>> > 	i915-display/intel_lspcon.o \
>> >+	i915-display/intel_metrics.o \
>> > 	i915-display/intel_modeset_lock.o \
>> > 	i915-display/intel_modeset_setup.o \
>> > 	i915-display/intel_modeset_verify.o \
>> >--
>> >2.44.0
>>


More information about the Intel-gfx mailing list