Skip to contentSkip to conversation

drm/atomic: Add drm_crtc_state->active

PATCH In review Daniel Vetter 4 participants 12 messages Jan 22, 2015 – Jan 23, 2015

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/5 drm/atomic: Add drm_crtc_state->active

This is the infrastructure for DPMS ported to the atomic world. Fundamental changes compare to legacy DPMS are:

- No more per-connector dpms state, instead there's just one per each display pipeline. So if you clone either you have to unclone first if you only want to switch off one screen, or you just switch of everything (like all desktops do). This massively reduces complexity for cloning since now there's no more half-enabled cloned configs to consider.

- Only on/off, dpms standby/suspend are as dead as real CRTs. Again reduces complexity a lot.

Now especially for backwards compat the really important part for dpms support is that dpms on always succeeds (except for hw death and unplugged cables ofc). Which means everything that could fail (like configuration checking, resources assignments and buffer management) must be done irrespective from ->active. ->active is really only a toggle to change the hardware state. More precisely:

- Drivers MUST NOT look at ->active in their ->atomic_check callbacks. Changes to ->active MUST always suceed if nothing else changes.

- Drivers using the atomic helpers MUST NOT look at ->active anywhere, period. The helpers will take care of calling the respective enable/modeset/disable hooks as necessary. As before the helpers will carefully keep track of the state and not call any hooks unecessarily, so still no double-disables or enables like with crtc helpers.

- ->mode_set hooks are only called when the mode or output configuration changes, not for changes in ->active state.

- Drivers which reconstruct the state objects in their ->reset hooks or through some other hw state readout infrastructure must ensure that ->active reflects actual hw state.

This just implements the core bits and helper logic, a subsequent patch will implement the helper code to implement legacy dpms with this.

v2: Rebase on top of the drm ioctl work:
- Move crtc checks to the core check function.
- Also check for ->active_changed when deciding whether a modeset might happen (for the ALLOW_MODESET mode).
- Expose the ->active state with an atomic prop.

v3: Review from Rob
- Spelling fix in comment.
- Extract needs_modeset helper to consolidate the ->mode_changed || ->active_changed checks.

v4: Fixup fumble between crtc->state and crtc_state.

---
 drivers/gpu/drm/drm_atomic.c        | 18 +++++++++++--
 drivers/gpu/drm/drm_atomic_helper.c | 54 +++++++++++++++++++++++++++++++++----
 drivers/gpu/drm/drm_crtc.c          | 10 +++++++
 include/drm/drm_crtc.h              |  3 +++
 4 files changed, 78 insertions(+), 7 deletions(-)
drivers/gpu/drm/drm_atomic.c +16 -2
index 1e38dfc8e462..ee68267bb326 100644
@@ -241,7 +241,13 @@ int drm_atomic_crtc_set_property(struct drm_crtc *crtc,
 		struct drm_crtc_state *state, struct drm_property *property,
 		uint64_t val)
 {
-	if (crtc->funcs->atomic_set_property)
+	struct drm_device *dev = crtc->dev;
+	struct drm_mode_config *config = &dev->mode_config;
+
+	/* FIXME: Mode prop is missing, which also controls ->enable. */
+	if (property == config->prop_active) {
+		state->active = val;
+	} else if (crtc->funcs->atomic_set_property)
 		return crtc->funcs->atomic_set_property(crtc, state, property, val);
 	return -EINVAL;
 }
@@ -282,6 +288,13 @@ static int drm_atomic_crtc_check(struct drm_crtc *crtc,
 	 *
 	 * TODO: Add generic modeset state checks once we support those.
 	 */
+
+	if (state->active && !state->enable) {
+		DRM_DEBUG_KMS("[CRTC:%d] active without enabled\n",
+			      crtc->base.id);
+		return -EINVAL;
+	}
+
 	return 0;
 }
 
@@ -976,7 +989,8 @@ int drm_atomic_check_only(struct drm_atomic_state *state)
 			if (!crtc)
 				continue;
 
-			if (crtc_state->mode_changed) {
+			if (crtc_state->mode_changed ||
+			    crtc_state->active_changed) {
 				DRM_DEBUG_KMS("[CRTC:%d] requires full modeset\n",
 					      crtc->base.id);
 				return -EINVAL;
drivers/gpu/drm/drm_atomic_helper.c +49 -5

127 lines not on this page · raw message

drivers/gpu/drm/drm_crtc.c +10 -0
index 56e3256d5249..30a136b8a4fc 100644
@@ -691,6 +691,10 @@ int drm_crtc_init_with_planes(struct drm_device *dev, struct drm_crtc *crtc,
 	if (cursor)
 		cursor->possible_crtcs = 1 << drm_crtc_index(crtc);
 
+	if (drm_core_check_feature(dev, DRIVER_ATOMIC)) {
+		drm_object_attach_property(&crtc->base, config->prop_active, 0);
+	}
+
 	return 0;
 }
 EXPORT_SYMBOL(drm_crtc_init_with_planes);
@@ -1391,6 +1395,12 @@ static int drm_mode_create_standard_properties(struct drm_device *dev)
 		return -ENOMEM;
 	dev->mode_config.prop_crtc_id = prop;
 
+	prop = drm_property_create_bool(dev, DRM_MODE_PROP_ATOMIC,
+			"ACTIVE");
+	if (!prop)
+		return -ENOMEM;
+	dev->mode_config.prop_active = prop;
+
 	return 0;
 }
 
include/drm/drm_crtc.h +3 -0
index 8022ab11958a..4d3f3b874dd6 100644
@@ -249,6 +249,7 @@ struct drm_atomic_state;
  * @enable: whether the CRTC should be enabled, gates all other state
  * @active: whether the CRTC is actively displaying (used for DPMS)
  * @mode_changed: for use by helpers and drivers when computing state updates
+ * @active_changed: for use by helpers and drivers when computing state updates
  * @plane_mask: bitmask of (1 << drm_plane_index(plane)) of attached planes
  * @last_vblank_count: for helpers and drivers to capture the vblank of the
  * 	update to ensure framebuffer cleanup isn't done too early
@@ -274,6 +275,7 @@ struct drm_crtc_state {
 	/* computed state bits used by helpers and drivers */
 	bool planes_changed : 1;
 	bool mode_changed : 1;
+	bool active_changed : 1;
 
 	/* attached planes bitmask:
 	 * WARNING: transitional helpers do not maintain plane_mask so
@@ -1128,6 +1130,7 @@ struct drm_mode_config {
 	struct drm_property *prop_crtc_h;
 	struct drm_property *prop_fb_id;
 	struct drm_property *prop_crtc_id;
+	struct drm_property *prop_active;
 
 	/* DVI-I properties */
 	struct drm_property *dvi_i_subconnector_property;
Signature · 6 lines
2.1.4

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#1 lore raw

PATCH 2/5 drm/atomic-helper: add connector->dpms() implementation

This builds on top of the crtc->active infrastructure to implement legacy DPMS. My choice of semantics is somewhat arbitrary, but the entire pipe is enabled as along as one output is still enabled.

Of course it also clamps everything that's not ON to OFF.

v2: Fix spelling in one comment.

---
 drivers/gpu/drm/drm_atomic_helper.c | 75 +++++++++++++++++++++++++++++++++++++
 include/drm/drm_atomic_helper.h     |  2 +
 2 files changed, 77 insertions(+)
drivers/gpu/drm/drm_atomic_helper.c +75 -0
index 3f17933b1d2c..736ab924ddc9 100644
@@ -1887,6 +1887,81 @@ backoff:
 EXPORT_SYMBOL(drm_atomic_helper_page_flip);
 
 /**
+ * drm_atomic_helper_connector_dpms() - connector dpms helper implementation
+ * @connector: affected connector
+ * @mode: DPMS mode
+ *
+ * This is the main helper function provided by the atomic helper framework for
+ * implementing the legacy DPMS connector interface. It computes the new desired
+ * ->active state for the corresponding CRTC (if the connector is enabled) and
+ *  updates it.
+ */
+void drm_atomic_helper_connector_dpms(struct drm_connector *connector,
+				      int mode)
+{
+	struct drm_mode_config *config = &connector->dev->mode_config;
+	struct drm_atomic_state *state;
+	struct drm_crtc_state *crtc_state;
+	struct drm_crtc *crtc;
+	struct drm_connector *tmp_connector;
+	int ret;
+	bool active = false;
+
+	if (mode != DRM_MODE_DPMS_ON)
+		mode = DRM_MODE_DPMS_OFF;
+
+	connector->dpms = mode;
+	crtc = connector->state->crtc;
+
+	if (!crtc)
+		return;
+
+	/* FIXME: ->dpms has no return value so can't forward the -ENOMEM. */
+	state = drm_atomic_state_alloc(connector->dev);
+	if (!state)
+		return;
+
+	state->acquire_ctx = drm_modeset_legacy_acquire_ctx(crtc);
+retry:
+	crtc_state = drm_atomic_get_crtc_state(state, crtc);
+
+	WARN_ON(!drm_modeset_is_locked(&config->connection_mutex));
+
+	list_for_each_entry(tmp_connector, &config->connector_list, head) {
+		if (connector->state->crtc != crtc)
+			continue;
+
+		if (connector->dpms == DRM_MODE_DPMS_ON) {
+			active = true;
+			break;
+		}
+	}
+	crtc_state->active = active;
+
+	ret = drm_atomic_async_commit(state);
+	if (ret != 0)
+		goto fail;
+
+	/* Driver takes ownership of state on successful async commit. */
+	return;
+fail:
+	if (ret == -EDEADLK)
+		goto backoff;
+
+	drm_atomic_state_free(state);
+
+	WARN(1, "Driver bug: Changing ->active failed with ret=%i\n", ret);
+
+	return;
+backoff:
+	drm_atomic_state_clear(state);
+	drm_atomic_legacy_backoff(state);
+
+	goto retry;
+}
+EXPORT_SYMBOL(drm_atomic_helper_connector_dpms);
+
+/**
  * DOC: atomic state reset and initialization
  *
  * Both the drm core and the atomic helpers assume that there is always the full
include/drm/drm_atomic_helper.h +2 -0
index 2095917ff8c7..cf501df9e513 100644
@@ -82,6 +82,8 @@ int drm_atomic_helper_page_flip(struct drm_crtc *crtc,
 				struct drm_framebuffer *fb,
 				struct drm_pending_vblank_event *event,
 				uint32_t flags);
+void drm_atomic_helper_connector_dpms(struct drm_connector *connector,
+				      int mode);
 
 /* default implementations for state handling */
 void drm_atomic_helper_crtc_reset(struct drm_crtc *crtc);
Signature · 6 lines
2.1.4

_______________________________________________
dri-devel mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/dri-devel
#2 lore raw

PATCH 1/1 drm/atomic-helper: add connector->dpms() implementation

This builds on top of the crtc->active infrastructure to implement legacy DPMS. My choice of semantics is somewhat arbitrary, but the entire pipe is enabled as along as one output is still enabled.

Of course it also clamps everything that's not ON to OFF.

v2: Fix spelling in one comment.

v3: Don't do an async commit (Thierry)

---
 drivers/gpu/drm/drm_atomic_helper.c | 75 +++++++++++++++++++++++++++++++++++++
 include/drm/drm_atomic_helper.h     |  2 +
 2 files changed, 77 insertions(+)
drivers/gpu/drm/drm_atomic_helper.c +75 -0
index 3f17933b1d2c..f693344d9573 100644
@@ -1887,6 +1887,81 @@ backoff:
 EXPORT_SYMBOL(drm_atomic_helper_page_flip);
 
 /**
+ * drm_atomic_helper_connector_dpms() - connector dpms helper implementation
+ * @connector: affected connector
+ * @mode: DPMS mode
+ *
+ * This is the main helper function provided by the atomic helper framework for
+ * implementing the legacy DPMS connector interface. It computes the new desired
+ * ->active state for the corresponding CRTC (if the connector is enabled) and
+ *  updates it.
+ */
+void drm_atomic_helper_connector_dpms(struct drm_connector *connector,
+				      int mode)
+{
+	struct drm_mode_config *config = &connector->dev->mode_config;
+	struct drm_atomic_state *state;
+	struct drm_crtc_state *crtc_state;
+	struct drm_crtc *crtc;
+	struct drm_connector *tmp_connector;
+	int ret;
+	bool active = false;
+
+	if (mode != DRM_MODE_DPMS_ON)
+		mode = DRM_MODE_DPMS_OFF;
+
+	connector->dpms = mode;
+	crtc = connector->state->crtc;
+
+	if (!crtc)
+		return;
+
+	/* FIXME: ->dpms has no return value so can't forward the -ENOMEM. */
+	state = drm_atomic_state_alloc(connector->dev);
+	if (!state)
+		return;
+
+	state->acquire_ctx = drm_modeset_legacy_acquire_ctx(crtc);
+retry:
+	crtc_state = drm_atomic_get_crtc_state(state, crtc);
+
+	WARN_ON(!drm_modeset_is_locked(&config->connection_mutex));
+
+	list_for_each_entry(tmp_connector, &config->connector_list, head) {
+		if (connector->state->crtc != crtc)
+			continue;
+
+		if (connector->dpms == DRM_MODE_DPMS_ON) {
+			active = true;
+			break;
+		}
+	}
+	crtc_state->active = active;
+
+	ret = drm_atomic_commit(state);
+	if (ret != 0)
+		goto fail;
+
+	/* Driver takes ownership of state on successful async commit. */
+	return;
+fail:
+	if (ret == -EDEADLK)
+		goto backoff;
+
+	drm_atomic_state_free(state);
+
+	WARN(1, "Driver bug: Changing ->active failed with ret=%i\n", ret);
+
+	return;
+backoff:
+	drm_atomic_state_clear(state);
+	drm_atomic_legacy_backoff(state);
+
+	goto retry;
+}
+EXPORT_SYMBOL(drm_atomic_helper_connector_dpms);
+
+/**
  * DOC: atomic state reset and initialization
  *
  * Both the drm core and the atomic helpers assume that there is always the full
include/drm/drm_atomic_helper.h +2 -0
index 2095917ff8c7..cf501df9e513 100644
@@ -82,6 +82,8 @@ int drm_atomic_helper_page_flip(struct drm_crtc *crtc,
 				struct drm_framebuffer *fb,
 				struct drm_pending_vblank_event *event,
 				uint32_t flags);
+void drm_atomic_helper_connector_dpms(struct drm_connector *connector,
+				      int mode);
 
 /* default implementations for state handling */
 void drm_atomic_helper_crtc_reset(struct drm_crtc *crtc);
Signature · 6 lines
2.1.4

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Tested-by on v1 1/1
#3 lore raw

Re: [PATCH] drm/atomic-helper: add connector->dpms() implementation

Task id: 5629 -------------------------------------Summary-------------------------------------
Platform Delta drm-intel-nightly Series Applied

PNV                                  353/353              353/353
ILK                                  353/353              353/353
SNB              +1                 399/422              400/422
IVB                                  486/487              486/487
BYT                                  296/296              296/296

HSW +1-3 507/508 505/508
BDW +1 399/402 400/402
-------------------------------------Detailed-------------------------------------
Platform Test drm-intel-nightly Series Applied

 SNB  igt_kms_flip_nonexisting-fb      NSPT(1, M35)PASS(2, M35)      PASS(1, M35)
*HSW  igt_gem_pwrite_pread_display-copy-performance      PASS(2, M19M20)      DMESG_WARN(1, M20)
*HSW  igt_gem_pwrite_pread_snooped-copy-performance      PASS(2, M19M20)      DMESG_WARN(1, M20)
*HSW  igt_gem_pwrite_pread_snooped-pwrite-blt-cpu_mmap-performance      PASS(2, M19M20)      DMESG_WARN(1, M20)
 HSW  igt_gem_storedw_loop_blt      DMESG_WARN(2, M19M20)PASS(1, M20)      PASS(1, M20)
 BDW  igt_gem_pwrite_pread_display-pwrite-blt-gtt_mmap-performance      DMESG_WARN(2, M28)PASS(1, M30)      PASS(1, M30)

Note: You need to pay more attention to line start with '*'
_______________________________________________
Intel-gfx mailing list [email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

#4 lore raw

PATCH 3/5 drm/atomic-helpers: Recover full cursor plane behaviour

In reply to Daniel Vetter

Cursor plane updates have historically been fully async and mutliple updates batched together for the next vsync. And userspace relies upon that. Since implementing a full queue of async atomic updates is a bit of work lets just recover the cursor specific behaviour with a hint flag and some hacks to drop the vblank wait.

---
 drivers/gpu/drm/drm_atomic_helper.c | 11 +++++++++++
 include/drm/drm_crtc.h              |  2 ++
 2 files changed, 13 insertions(+)
drivers/gpu/drm/drm_atomic_helper.c +11 -0
index 736ab924ddc9..a74a22db4dea 100644
@@ -909,6 +909,11 @@ drm_atomic_helper_wait_for_vblanks(struct drm_device *dev,
 		if (!crtc->state->enable)
 			continue;
 
+		/* Legacy cursor ioctls are completely unsynced, and userspace
+		 * relies on that (by doing tons of cursor updates). */
+		if (old_state->legacy_cursor_update)
+			continue;
+
 		if (!framebuffer_changed(dev, old_state, crtc))
 			continue;
 
@@ -1335,6 +1340,9 @@ retry:
 	if (ret != 0)
 		goto fail;
 
+	if (plane == crtc->cursor)
+		state->legacy_cursor_update = true;
+
 	/* Driver takes ownership of state on successful commit. */
 	return 0;
 fail:
@@ -1410,6 +1418,9 @@ retry:
 	plane_state->src_h = 0;
 	plane_state->src_w = 0;
 
+	if (plane == plane->crtc->cursor)
+		state->legacy_cursor_update = true;
+
 	ret = drm_atomic_commit(state);
 	if (ret != 0)
 		goto fail;
include/drm/drm_crtc.h +2 -0
index 4d3f3b874dd6..8b626ff3d3bd 100644
@@ -929,6 +929,7 @@ struct drm_bridge {
  * struct struct drm_atomic_state - the global state object for atomic updates
  * @dev: parent DRM device
  * @allow_modeset: allow full modeset
+ * @allow_modeset: hint to enforce legacy cursor ioctl semantics
  * @planes: pointer to array of plane pointers
  * @plane_states: pointer to array of plane states pointers
  * @crtcs: pointer to array of CRTC pointers
@@ -941,6 +942,7 @@ struct drm_bridge {
 struct drm_atomic_state {
 	struct drm_device *dev;
 	bool allow_modeset : 1;
+	bool legacy_cursor_update : 1;
 	struct drm_plane **planes;
 	struct drm_plane_state **plane_states;
 	struct drm_crtc **crtcs;
Signature · 6 lines
2.1.4

_______________________________________________
dri-devel mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/dri-devel
#5 lore raw

PATCH 4/5 drm/atomic-helpers: Saner encoder/crtc callbacks

In reply to Daniel Vetter

For historical reasons going all the way back to how the Xrandr code was implemented the semantics of the callbacks used to enable/disable crtcs and encoders are ... interesting.

But with atomic helpers all that complexity has been binned, with only a well-defined on/off action left. Unfortunately the names stuck.

Let's fix that by adding enable/disable hooks every, make them the preferred variant for atomic and update documentations.

Later on we add debug warnings when drivers have deprecated hooks. But while everything is in-flight with lots of drivers converting to atomic that's a bit too much - better wait for things to settle a bit first.

---
 drivers/gpu/drm/drm_atomic_helper.c | 17 ++++++++++++-----
 include/drm/drm_crtc_helper.h       | 20 ++++++++++++++++++--
 2 files changed, 30 insertions(+), 7 deletions(-)
drivers/gpu/drm/drm_atomic_helper.c +12 -5
index a74a22db4dea..b67754035fec 100644
@@ -599,7 +599,7 @@ disable_outputs(struct drm_device *dev, struct drm_atomic_state *old_state)
 			encoder->bridge->funcs->disable(encoder->bridge);
 
 		/* Right function depends upon target state. */
-		if (connector->state->crtc)
+		if (connector->state->crtc && funcs->prepare)
 			funcs->prepare(encoder);
 		else if (funcs->disable)
 			funcs->disable(encoder);
@@ -628,7 +628,7 @@ disable_outputs(struct drm_device *dev, struct drm_atomic_state *old_state)
 		funcs = crtc->helper_private;
 
 		/* Right function depends upon target state. */
-		if (crtc->state->enable)
+		if (crtc->state->enable && funcs->prepare)
 			funcs->prepare(crtc);
 		else if (funcs->disable)
 			funcs->disable(crtc);
@@ -792,8 +792,12 @@ void drm_atomic_helper_commit_post_planes(struct drm_device *dev,
 
 		funcs = crtc->helper_private;
 
-		if (crtc->state->enable)
-			funcs->commit(crtc);
+		if (crtc->state->enable) {
+			if (funcs->enable)
+				funcs->enable(crtc);
+			else
+				funcs->commit(crtc);
+		}
 	}
 
 	for (i = 0; i < old_state->num_connector; i++) {
@@ -819,7 +823,10 @@ void drm_atomic_helper_commit_post_planes(struct drm_device *dev,
 		if (encoder->bridge)
 			encoder->bridge->funcs->pre_enable(encoder->bridge);
 
-		funcs->commit(encoder);
+		if (funcs->enable)
+			funcs->enable(encoder);
+		else
+			funcs->commit(encoder);
 
 		if (encoder->bridge)
 			encoder->bridge->funcs->enable(encoder->bridge);
include/drm/drm_crtc_helper.h +18 -2
index e76828d81a8b..c36f1bc26837 100644
@@ -58,11 +58,19 @@ enum mode_set_atomic {
  * @mode_set_base_atomic: non-blocking mode set (used for kgdb support)
  * @load_lut: load color palette
  * @disable: disable CRTC when no longer in use
+ * @disable: enable CRTC
  * @atomic_check: check for validity of an atomic state
  * @atomic_begin: begin atomic update
  * @atomic_flush: flush atomic update
  *
  * The helper operations are called by the mid-layer CRTC helper.
+ *
+ * Note that with atomic helpers @dpms, @prepare and @commit hooks are
+ * deprecated. Used @enable and @disable instead exclusively.
+ *
+ * With legacy crtc helpers there's a big semantic difference between @disable
+ * and the other hooks: @disable also needs to release any resources acquired in
+ * @mode_set (like shared PLLs).
  */
 struct drm_crtc_helper_funcs {
 	/*
@@ -93,8 +101,8 @@ struct drm_crtc_helper_funcs {
 	/* reload the current crtc LUT */
 	void (*load_lut)(struct drm_crtc *crtc);
 
-	/* disable crtc when not in use - more explicit than dpms off */
 	void (*disable)(struct drm_crtc *crtc);
+	void (*enable)(struct drm_crtc *crtc);
 
 	/* atomic helpers */
 	int (*atomic_check)(struct drm_crtc *crtc,
@@ -115,8 +123,16 @@ struct drm_crtc_helper_funcs {
  * @get_crtc: return CRTC that the encoder is currently attached to
  * @detect: connection status detection
  * @disable: disable encoder when not in use (overrides DPMS off)
+ * @disable: enable encoder
  *
  * The helper operations are called by the mid-layer CRTC helper.
+ *
+ * Note that with atomic helpers @dpms, @prepare and @commit hooks are
+ * deprecated. Used @enable and @disable instead exclusively.
+ *
+ * With legacy crtc helpers there's a big semantic difference between @disable
+ * and the other hooks: @disable also needs to release any resources acquired in
+ * @mode_set (like shared PLLs).
  */
 struct drm_encoder_helper_funcs {
 	void (*dpms)(struct drm_encoder *encoder, int mode);
@@ -135,8 +151,8 @@ struct drm_encoder_helper_funcs {
 	/* detect for DAC style encoders */
 	enum drm_connector_status (*detect)(struct drm_encoder *encoder,
 					    struct drm_connector *connector);
-	/* disable encoder when not in use - more explicit than dpms off */
 	void (*disable)(struct drm_encoder *encoder);
+	void (*enable)(struct drm_encoder *encoder);
 };
 
 /**
Signature · 6 lines
2.1.4

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#6 lore raw

PATCH 5/5 drm/atomic-helper: debug output for modesets

In reply to Daniel Vetter

With the combination of ->enable and ->active it's a bit complicated to follow what exactly is going on sometimes within a full modeset. Add debug output to make this all traceable.

---
 drivers/gpu/drm/drm_atomic_helper.c | 22 +++++++++++++++++++++-
 1 file changed, 21 insertions(+), 1 deletion(-)
drivers/gpu/drm/drm_atomic_helper.c +21 -1
index b67754035fec..b7ac83555b11 100644
@@ -591,6 +591,9 @@ disable_outputs(struct drm_device *dev, struct drm_atomic_state *old_state)
 
 		funcs = encoder->helper_private;
 
+		DRM_DEBUG_KMS("disabling [ENCODER:%d:%s]\n",
+			      encoder->base.id, encoder->name);
+
 		/*
 		 * Each encoder has at most one connector (since we always steal
 		 * it away), so we won't call call disable hooks twice.
@@ -627,6 +630,10 @@ disable_outputs(struct drm_device *dev, struct drm_atomic_state *old_state)
 
 		funcs = crtc->helper_private;
 
+		DRM_DEBUG_KMS("disabling [CRTC:%d]\n",
+			      crtc->base.id);
+
+
 		/* Right function depends upon target state. */
 		if (crtc->state->enable && funcs->prepare)
 			funcs->prepare(crtc);
@@ -707,8 +714,12 @@ crtc_set_mode(struct drm_device *dev, struct drm_atomic_state *old_state)
 
 		funcs = crtc->helper_private;
 
-		if (crtc->state->enable)
+		if (crtc->state->enable) {
+			DRM_DEBUG_KMS("modeset on [CRTC:%d]\n",
+				      crtc->base.id);
+
 			funcs->mode_set_nofb(crtc);
+		}
 	}
 
 	for (i = 0; i < old_state->num_connector; i++) {
@@ -732,6 +743,9 @@ crtc_set_mode(struct drm_device *dev, struct drm_atomic_state *old_state)
 		if (!new_crtc_state->mode_changed)
 			continue;
 
+		DRM_DEBUG_KMS("modeset on [ENCODER:%d:%s]\n",
+			      encoder->base.id, encoder->name);
+
 		/*
 		 * Each encoder has at most one connector (since we always steal
 		 * it away), so we won't call call mode_set hooks twice.
@@ -793,6 +807,9 @@ void drm_atomic_helper_commit_post_planes(struct drm_device *dev,
 		funcs = crtc->helper_private;
 
 		if (crtc->state->enable) {
+			DRM_DEBUG_KMS("enabling [CRTC:%d]\n",
+				      crtc->base.id);
+
 			if (funcs->enable)
 				funcs->enable(crtc);
 			else
@@ -816,6 +833,9 @@ void drm_atomic_helper_commit_post_planes(struct drm_device *dev,
 		encoder = connector->state->best_encoder;
 		funcs = encoder->helper_private;
 
+		DRM_DEBUG_KMS("enabling [ENCODER:%d:%s]\n",
+			      encoder->base.id, encoder->name);
+
 		/*
 		 * Each encoder has at most one connector (since we always steal
 		 * it away), so we won't call call enable hooks twice.
Signature · 6 lines
2.1.4

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#7 lore raw
In reply to Daniel Vetter
Quoted 33 lines · Daniel Vetter

This is the infrastructure for DPMS ported to the atomic world. Fundamental changes compare to legacy DPMS are:

- No more per-connector dpms state, instead there's just one per each display pipeline. So if you clone either you have to unclone first if you only want to switch off one screen, or you just switch of everything (like all desktops do). This massively reduces complexity for cloning since now there's no more half-enabled cloned configs to consider.

- Only on/off, dpms standby/suspend are as dead as real CRTs. Again reduces complexity a lot.

Now especially for backwards compat the really important part for dpms support is that dpms on always succeeds (except for hw death and unplugged cables ofc). Which means everything that could fail (like configuration checking, resources assignments and buffer management) must be done irrespective from ->active. ->active is really only a toggle to change the hardware state. More precisely:

- Drivers MUST NOT look at ->active in their ->atomic_check callbacks. Changes to ->active MUST always suceed if nothing else changes.

- Drivers using the atomic helpers MUST NOT look at ->active anywhere, period. The helpers will take care of calling the respective enable/modeset/disable hooks as necessary. As before the helpers will carefully keep track of the state and not call any hooks unecessarily, so still no double-disables or enables like with crtc helpers.

- ->mode_set hooks are only called when the mode or output configuration changes, not for changes in ->active state.

Quoted 7 lines

- Drivers which reconstruct the state objects in their ->reset hooks or through some other hw state readout infrastructure must ensure that ->active reflects actual hw state.

This just implements the core bits and helper logic, a subsequent patch will implement the helper code to implement legacy dpms with this.

So we now have multiple conflicting properties for the same thing? I don't particularly relish that idea.

I suppose a rather easy way to way to deal with that would be to hide the dpms properties from non-atomic clients, and conversly hide the active property from legacy clients.

Quoted 271 lines

v2: Rebase on top of the drm ioctl work:
- Move crtc checks to the core check function.
- Also check for ->active_changed when deciding whether a modeset might happen (for the ALLOW_MODESET mode).
- Expose the ->active state with an atomic prop.

v3: Review from Rob
- Spelling fix in comment.
- Extract needs_modeset helper to consolidate the ->mode_changed || ->active_changed checks.

v4: Fixup fumble between crtc->state and crtc_state.

---
 drivers/gpu/drm/drm_atomic.c        | 18 +++++++++++--
 drivers/gpu/drm/drm_atomic_helper.c | 54 +++++++++++++++++++++++++++++++++----
 drivers/gpu/drm/drm_crtc.c          | 10 +++++++
 include/drm/drm_crtc.h              |  3 +++
 4 files changed, 78 insertions(+), 7 deletions(-)
drivers/gpu/drm/drm_atomic.c +16 -2
index 1e38dfc8e462..ee68267bb326 100644
@@ -241,7 +241,13 @@ int drm_atomic_crtc_set_property(struct drm_crtc *crtc,
 		struct drm_crtc_state *state, struct drm_property *property,
 		uint64_t val)
 {
-	if (crtc->funcs->atomic_set_property)
+	struct drm_device *dev = crtc->dev;
+	struct drm_mode_config *config = &dev->mode_config;
+
+	/* FIXME: Mode prop is missing, which also controls ->enable. */
+	if (property == config->prop_active) {
+		state->active = val;
+	} else if (crtc->funcs->atomic_set_property)
 		return crtc->funcs->atomic_set_property(crtc, state, property, val);
 	return -EINVAL;
 }
@@ -282,6 +288,13 @@ static int drm_atomic_crtc_check(struct drm_crtc *crtc,
 	 *
 	 * TODO: Add generic modeset state checks once we support those.
 	 */
+
+	if (state->active && !state->enable) {
+		DRM_DEBUG_KMS("[CRTC:%d] active without enabled\n",
+			      crtc->base.id);
+		return -EINVAL;
+	}
+
 	return 0;
 }
 
@@ -976,7 +989,8 @@ int drm_atomic_check_only(struct drm_atomic_state *state)
 			if (!crtc)
 				continue;
 
-			if (crtc_state->mode_changed) {
+			if (crtc_state->mode_changed ||
+			    crtc_state->active_changed) {
 				DRM_DEBUG_KMS("[CRTC:%d] requires full modeset\n",
 					      crtc->base.id);
 				return -EINVAL;
drivers/gpu/drm/drm_atomic_helper.c +49 -5
index 541ba833ed36..3f17933b1d2c 100644
@@ -330,6 +330,12 @@ mode_fixup(struct drm_atomic_state *state)
 	return 0;
 }
 
+static bool
+needs_modeset(struct drm_crtc_state *state)
+{
+	return state->mode_changed || state->active_changed;
+}
+
 /**
  * drm_atomic_helper_check - validate state object for modeset changes
  * @dev: DRM device
@@ -404,12 +410,27 @@ drm_atomic_helper_check_modeset(struct drm_device *dev,
 		crtc = state->crtcs[i];
 		crtc_state = state->crtc_states[i];
 
-		if (!crtc || !crtc_state->mode_changed)
+		if (!crtc)
 			continue;
 
-		DRM_DEBUG_KMS("[CRTC:%d] needs full modeset, enable: %c\n",
+		/*
+		 * We must set ->active_changed after walking connectors for
+		 * otherwise an update that only changes active would result in
+		 * a full modeset because update_connector_routing force that.
+		 */
+		if (crtc->state->active != crtc_state->active) {
+			DRM_DEBUG_KMS("[CRTC:%d] active changed\n",
+				      crtc->base.id);
+			crtc_state->active_changed = true;
+		}
+
+		if (!needs_modeset(crtc_state))
+			continue;
+
+		DRM_DEBUG_KMS("[CRTC:%d] needs all connectors, enable: %c, active: %c\n",
 			      crtc->base.id,
-			      crtc_state->enable ? 'y' : 'n');
+			      crtc_state->enable ? 'y' : 'n',
+			      crtc_state->active ? 'y' : 'n');
 
 		ret = drm_atomic_add_affected_connectors(state, crtc);
 		if (ret != 0)
@@ -545,6 +566,7 @@ disable_outputs(struct drm_device *dev, struct drm_atomic_state *old_state)
 		struct drm_connector *connector;
 		struct drm_encoder_helper_funcs *funcs;
 		struct drm_encoder *encoder;
+		struct drm_crtc_state *old_crtc_state;
 
 		old_conn_state = old_state->connector_states[i];
 		connector = old_state->connectors[i];
@@ -554,6 +576,11 @@ disable_outputs(struct drm_device *dev, struct drm_atomic_state *old_state)
 		if (!old_conn_state || !old_conn_state->crtc)
 			continue;
 
+		old_crtc_state = old_state->crtc_states[drm_crtc_index(old_conn_state->crtc)];
+
+		if (!old_crtc_state->active)
+			continue;
+
 		encoder = old_conn_state->best_encoder;
 
 		/* We shouldn't get this far if we didn't previously have
@@ -586,11 +613,16 @@ disable_outputs(struct drm_device *dev, struct drm_atomic_state *old_state)
 	for (i = 0; i < ncrtcs; i++) {
 		struct drm_crtc_helper_funcs *funcs;
 		struct drm_crtc *crtc;
+		struct drm_crtc_state *old_crtc_state;
 
 		crtc = old_state->crtcs[i];
+		old_crtc_state = old_state->crtc_states[i];
 
 		/* Shut down everything that needs a full modeset. */
-		if (!crtc || !crtc->state->mode_changed)
+		if (!crtc || !needs_modeset(crtc->state))
+			continue;
+
+		if (!old_crtc_state->active)
 			continue;
 
 		funcs = crtc->helper_private;
@@ -697,6 +729,9 @@ crtc_set_mode(struct drm_device *dev, struct drm_atomic_state *old_state)
 		mode = &new_crtc_state->mode;
 		adjusted_mode = &new_crtc_state->adjusted_mode;
 
+		if (!new_crtc_state->mode_changed)
+			continue;
+
 		/*
 		 * Each encoder has at most one connector (since we always steal
 		 * it away), so we won't call call mode_set hooks twice.
@@ -749,7 +784,10 @@ void drm_atomic_helper_commit_post_planes(struct drm_device *dev,
 		crtc = old_state->crtcs[i];
 
 		/* Need to filter out CRTCs where only planes change. */
-		if (!crtc || !crtc->state->mode_changed)
+		if (!crtc || !needs_modeset(crtc->state))
+			continue;
+
+		if (!crtc->state->active)
 			continue;
 
 		funcs = crtc->helper_private;
@@ -768,6 +806,9 @@ void drm_atomic_helper_commit_post_planes(struct drm_device *dev,
 		if (!connector || !connector->state->best_encoder)
 			continue;
 
+		if (!connector->state->crtc->state->active)
+			continue;
+
 		encoder = connector->state->best_encoder;
 		funcs = encoder->helper_private;
 
@@ -1518,6 +1559,7 @@ retry:
 		WARN_ON(set->num_connectors);
 
 		crtc_state->enable = false;
+		crtc_state->active = false;
 
 		ret = drm_atomic_set_crtc_for_plane(primary_state, NULL);
 		if (ret != 0)
@@ -1532,6 +1574,7 @@ retry:
 	WARN_ON(!set->num_connectors);
 
 	crtc_state->enable = true;
+	crtc_state->active = true;
 	drm_mode_copy(&crtc_state->mode, set->mode);
 
 	ret = drm_atomic_set_crtc_for_plane(primary_state, crtc);
@@ -1894,6 +1937,7 @@ drm_atomic_helper_crtc_duplicate_state(struct drm_crtc *crtc)
 
 	if (state) {
 		state->mode_changed = false;
+		state->active_changed = false;
 		state->planes_changed = false;
 		state->event = NULL;
 	}
drivers/gpu/drm/drm_crtc.c +10 -0
index 56e3256d5249..30a136b8a4fc 100644
@@ -691,6 +691,10 @@ int drm_crtc_init_with_planes(struct drm_device *dev, struct drm_crtc *crtc,
 	if (cursor)
 		cursor->possible_crtcs = 1 << drm_crtc_index(crtc);
 
+	if (drm_core_check_feature(dev, DRIVER_ATOMIC)) {
+		drm_object_attach_property(&crtc->base, config->prop_active, 0);
+	}
+
 	return 0;
 }
 EXPORT_SYMBOL(drm_crtc_init_with_planes);
@@ -1391,6 +1395,12 @@ static int drm_mode_create_standard_properties(struct drm_device *dev)
 		return -ENOMEM;
 	dev->mode_config.prop_crtc_id = prop;
 
+	prop = drm_property_create_bool(dev, DRM_MODE_PROP_ATOMIC,
+			"ACTIVE");
+	if (!prop)
+		return -ENOMEM;
+	dev->mode_config.prop_active = prop;
+
 	return 0;
 }
 
include/drm/drm_crtc.h +3 -0
index 8022ab11958a..4d3f3b874dd6 100644
@@ -249,6 +249,7 @@ struct drm_atomic_state;
  * @enable: whether the CRTC should be enabled, gates all other state
  * @active: whether the CRTC is actively displaying (used for DPMS)
  * @mode_changed: for use by helpers and drivers when computing state updates
+ * @active_changed: for use by helpers and drivers when computing state updates
  * @plane_mask: bitmask of (1 << drm_plane_index(plane)) of attached planes
  * @last_vblank_count: for helpers and drivers to capture the vblank of the
  * 	update to ensure framebuffer cleanup isn't done too early
@@ -274,6 +275,7 @@ struct drm_crtc_state {
 	/* computed state bits used by helpers and drivers */
 	bool planes_changed : 1;
 	bool mode_changed : 1;
+	bool active_changed : 1;
 
 	/* attached planes bitmask:
 	 * WARNING: transitional helpers do not maintain plane_mask so
@@ -1128,6 +1130,7 @@ struct drm_mode_config {
 	struct drm_property *prop_crtc_h;
 	struct drm_property *prop_fb_id;
 	struct drm_property *prop_crtc_id;
+	struct drm_property *prop_active;
 
 	/* DVI-I properties */
 	struct drm_property *dvi_i_subconnector_property;
Signature · 6 lines
2.1.4

_______________________________________________
dri-devel mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/dri-devel
Signature · 6 lines
Ville Syrjälä
Intel OTC
_______________________________________________
dri-devel mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/dri-devel
#8 lore raw
Quoted 42 lines · Ville Syrjälä
Quoted 33 lines · Daniel Vetter

This is the infrastructure for DPMS ported to the atomic world. Fundamental changes compare to legacy DPMS are:

- No more per-connector dpms state, instead there's just one per each display pipeline. So if you clone either you have to unclone first if you only want to switch off one screen, or you just switch of everything (like all desktops do). This massively reduces complexity for cloning since now there's no more half-enabled cloned configs to consider.

- Only on/off, dpms standby/suspend are as dead as real CRTs. Again reduces complexity a lot.

Now especially for backwards compat the really important part for dpms support is that dpms on always succeeds (except for hw death and unplugged cables ofc). Which means everything that could fail (like configuration checking, resources assignments and buffer management) must be done irrespective from ->active. ->active is really only a toggle to change the hardware state. More precisely:

- Drivers MUST NOT look at ->active in their ->atomic_check callbacks. Changes to ->active MUST always suceed if nothing else changes.

- Drivers using the atomic helpers MUST NOT look at ->active anywhere, period. The helpers will take care of calling the respective enable/modeset/disable hooks as necessary. As before the helpers will carefully keep track of the state and not call any hooks unecessarily, so still no double-disables or enables like with crtc helpers.

- ->mode_set hooks are only called when the mode or output configuration changes, not for changes in ->active state.

Quoted 7 lines

- Drivers which reconstruct the state objects in their ->reset hooks or through some other hw state readout infrastructure must ensure that ->active reflects actual hw state.

This just implements the core bits and helper logic, a subsequent patch will implement the helper code to implement legacy dpms with this.

Quoted 6 lines

So we now have multiple conflicting properties for the same thing? I don't particularly relish that idea.

I suppose a rather easy way to way to deal with that would be to hide the dpms properties from non-atomic clients, and conversly hide the active property from legacy clients.

Which is pretty much what I do - you can only access the per-crtc ACTIVE property from the atomic ioctl, the per-connector dpms property is _not_ exposed through atomic. Vice-versa legacy clients wont see atomic properties (and hence this new one) either.

Is that good enough?
-Daniel

Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#9 lore raw
Quoted 46 lines · Daniel Vetter
Quoted 45 lines · Ville Syrjälä
Quoted 33 lines · Daniel Vetter

This is the infrastructure for DPMS ported to the atomic world. Fundamental changes compare to legacy DPMS are:

- No more per-connector dpms state, instead there's just one per each display pipeline. So if you clone either you have to unclone first if you only want to switch off one screen, or you just switch of everything (like all desktops do). This massively reduces complexity for cloning since now there's no more half-enabled cloned configs to consider.

- Only on/off, dpms standby/suspend are as dead as real CRTs. Again reduces complexity a lot.

Now especially for backwards compat the really important part for dpms support is that dpms on always succeeds (except for hw death and unplugged cables ofc). Which means everything that could fail (like configuration checking, resources assignments and buffer management) must be done irrespective from ->active. ->active is really only a toggle to change the hardware state. More precisely:

- Drivers MUST NOT look at ->active in their ->atomic_check callbacks. Changes to ->active MUST always suceed if nothing else changes.

- Drivers using the atomic helpers MUST NOT look at ->active anywhere, period. The helpers will take care of calling the respective enable/modeset/disable hooks as necessary. As before the helpers will carefully keep track of the state and not call any hooks unecessarily, so still no double-disables or enables like with crtc helpers.

- ->mode_set hooks are only called when the mode or output configuration changes, not for changes in ->active state.

Quoted 7 lines

- Drivers which reconstruct the state objects in their ->reset hooks or through some other hw state readout infrastructure must ensure that ->active reflects actual hw state.

This just implements the core bits and helper logic, a subsequent patch will implement the helper code to implement legacy dpms with this.

So we now have multiple conflicting properties for the same thing? I don't particularly relish that idea.

Quoted 8 lines
Quoted 3 lines

I suppose a rather easy way to way to deal with that would be to hide the dpms properties from non-atomic clients, and conversly hide the active property from legacy clients.

Which is pretty much what I do - you can only access the per-crtc ACTIVE property from the atomic ioctl, the per-connector dpms property is _not_ exposed through atomic. Vice-versa legacy clients wont see atomic properties (and hence this new one) either.

Oh, OK. I didn't see anything obvious that would filter out the dpms prop for non-atomic, nor do I see code to reject stuff in setprop/getprop ioctls etc. But I suppose such stuff may be in flight and I've just not paid attention.

Quoted 1 line

Is that good enough?

I suppose.

Another thing that came to mind wrt. the 'this can't fail rule' was fb pinning. So is the rule now going to be that we need to pin even when ACTIVE==false, or otherwise make sure all the FBs can be pinned simultaneosly? If we don't want to everything pinned all the time when ACTIVE==false, then we would need to prevent pinning of anything else in the meantime to make sure we don't run out of address space.

Signature · 6 lines
Ville Syrjälä
Intel OTC
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#10 lore raw
Quoted 9 lines · Ville Syrjälä
Quoted 4 lines

Which is pretty much what I do - you can only access the per-crtc ACTIVE property from the atomic ioctl, the per-connector dpms property is _not_ exposed through atomic. Vice-versa legacy clients wont see atomic properties (and hence this new one) either.

Oh, OK. I didn't see anything obvious that would filter out the dpms prop for non-atomic, nor do I see code to reject stuff in setprop/getprop ioctls etc. But I suppose such stuff may be in flight and I've just not paid attention.

On the atomic side the dpms prop is simple not decoded. The driver /could/ do that itself, but that would be really pointless. In getprops atomic ioctls are filtered out for non-atomic clients. Which means that an atomic client could do a dpms on the connector through the legacy setprop ioctl, but that would be rather stupid.

Quoted 10 lines
Quoted 1 line

Is that good enough?

I suppose.

Another thing that came to mind wrt. the 'this can't fail rule' was fb pinning. So is the rule now going to be that we need to pin even when ACTIVE==false, or otherwise make sure all the FBs can be pinned simultaneosly? If we don't want to everything pinned all the time when ACTIVE==false, then we would need to prevent pinning of anything else in the meantime to make sure we don't run out of address space.

fb pinning is done irrespective of the state of active. So if you update the fb while the display pipe is off the helpers will upin/pin correctly. Imo that's totally fine, since failing to pin when setting the display back to active really isn't a great thing. And we need to be able to tell userspace when something has gone wrong with the pinning (e.g. due to giant framebuffer or something). -Daniel

Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
dri-devel mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/dri-devel
#11 lore raw
Quoted 12 lines · Daniel Vetter
Quoted 9 lines · Ville Syrjälä
Quoted 4 lines

Which is pretty much what I do - you can only access the per-crtc ACTIVE property from the atomic ioctl, the per-connector dpms property is _not_ exposed through atomic. Vice-versa legacy clients wont see atomic properties (and hence this new one) either.

Oh, OK. I didn't see anything obvious that would filter out the dpms prop for non-atomic, nor do I see code to reject stuff in setprop/getprop ioctls etc. But I suppose such stuff may be in flight and I've just not paid attention.

Quoted 5 lines

On the atomic side the dpms prop is simple not decoded. The driver /could/ do that itself, but that would be really pointless. In getprops atomic ioctls are filtered out for non-atomic clients. Which means that an atomic client could do a dpms on the connector through the legacy setprop ioctl, but that would be rather stupid.

Well I'd rather it refused the entire thing. Less weird interactions to worry about if we're strict and block all the silly stuff at the top.

Quoted 12 lines
Quoted 11 lines
Quoted 2 lines

Is that good enough?

I suppose.

Another thing that came to mind wrt. the 'this can't fail rule' was fb pinning. So is the rule now going to be that we need to pin even when ACTIVE==false, or otherwise make sure all the FBs can be pinned simultaneosly? If we don't want to everything pinned all the time when ACTIVE==false, then we would need to prevent pinning of anything else in the meantime to make sure we don't run out of address space.

Quoted 6 lines

fb pinning is done irrespective of the state of active. So if you update the fb while the display pipe is off the helpers will upin/pin correctly. Imo that's totally fine, since failing to pin when setting the display back to active really isn't a great thing. And we need to be able to tell userspace when something has gone wrong with the pinning (e.g. due to giant framebuffer or something).

Yeah that was pretty much my original idea too. I suppose now that's the call comes from the core it's a bit less likely to be messed up again.

Signature · 6 lines
Ville Syrjälä
Intel OTC
_______________________________________________
dri-devel mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/dri-devel