Skip to contentSkip to conversation

drm/i915: check the power wells on assert_{cursor, plane}

PATCH In review Paulo Zanoni 2 participants 10 messages Jul 16, 2014 – Jul 17, 2014

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

PATCH 1/1 drm/i915: check the power wells on assert_{cursor, plane}

From: Paulo Zanoni <[email protected]>

Since we merged runtime PM support for DPMS, it is possible that these functions will be called when the power wells are disabled but a mode is "set", resulting in "failed assertion" and "device suspended while reading register" WARNs.

To reproduce the bug: disable all screens using mode unset, do a modeset on one screen, disable it using DPMS, then try to do a mode unset on it again to see the WARNs.

Testcase: igt/rpm_rpm/dpms-mode-unset-lpsp
Testcase: igt/rpm_rpm/dpms-mode-unset-non-lpsp

---
 drivers/gpu/drm/i915/intel_display.c | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)
drivers/gpu/drm/i915/intel_display.c +10 -4
index 54e3af9..7ad46e2 100644
@@ -1216,7 +1216,9 @@ static void assert_cursor(struct drm_i915_private *dev_priv,
 	struct drm_device *dev = dev_priv->dev;
 	bool cur_state;
 
-	if (IS_845G(dev) || IS_I865G(dev))
+	if (!intel_display_power_enabled(dev_priv, POWER_DOMAIN_PIPE(pipe)))
+		cur_state = false;
+	else if (IS_845G(dev) || IS_I865G(dev))
 		cur_state = I915_READ(_CURACNTR) & CURSOR_ENABLE;
 	else
 		cur_state = I915_READ(CURCNTR(pipe)) & CURSOR_MODE;
@@ -1262,9 +1264,13 @@ static void assert_plane(struct drm_i915_private *dev_priv,
 	u32 val;
 	bool cur_state;
 
-	reg = DSPCNTR(plane);
-	val = I915_READ(reg);
-	cur_state = !!(val & DISPLAY_PLANE_ENABLE);
+	if (!intel_display_power_enabled(dev_priv, POWER_DOMAIN_PIPE(plane))) {
+		cur_state = false;
+	} else {
+		reg = DSPCNTR(plane);
+		val = I915_READ(reg);
+		cur_state = !!(val & DISPLAY_PLANE_ENABLE);
+	}
 	WARN(cur_state != state,
 	     "plane %c assertion failure (expected %s, current %s)\n",
 	     plane_name(plane), state_string(state), state_string(cur_state));
Signature · 1 line
2.0.0
#1 lore raw

PATCH 1/1 tests/pm_rpm: add dpms-mode-unset{, -non}-lpsp subtests

From: Paulo Zanoni <[email protected]>

These tests currently trigger WARNs on our Kernel. Let's make sure we fix the bug and it never comes back.

---
 tests/pm_rpm.c | 21 +++++++++++++++++++++
 1 file changed, 21 insertions(+)
tests/pm_rpm.c +21 -0
index 323e072..1e63bc8 100644
@@ -1390,6 +1390,23 @@ static void system_suspend_subtest(void)
 	igt_assert(wait_for_suspended());
 }
 
+/* Enable a screen, activate DPMS, then do a modeset. At some point our driver
+ * produced WARNs on this case. */
+static void dpms_mode_unset_subtest(enum screen_type type)
+{
+	disable_all_screens(&ms_data);
+	igt_assert(wait_for_suspended());
+
+	igt_require(enable_one_screen_with_type(&ms_data, type));
+	igt_assert(wait_for_active());
+
+	disable_all_screens_dpms(&ms_data);
+	igt_assert(wait_for_suspended());
+
+	disable_all_screens(&ms_data);
+	igt_assert(wait_for_suspended());
+}
+
 int main(int argc, char *argv[])
 {
 	int rounds = 50;
@@ -1462,6 +1479,10 @@ int main(int argc, char *argv[])
 		debugfs_forcewake_user_subtest();
 	igt_subtest("sysfs-read")
 		sysfs_read_subtest();
+	igt_subtest("dpms-mode-unset-lpsp")
+		dpms_mode_unset_subtest(SCREEN_TYPE_LPSP);
+	igt_subtest("dpms-mode-unset-non-lpsp")
+		dpms_mode_unset_subtest(SCREEN_TYPE_NON_LPSP);
 
 	/* Modeset stress */
 	igt_subtest("modeset-lpsp-stress")
Signature · 1 line
2.0.0
#2 lore raw
In reply to Paulo Zanoni
Quoted 7 lines · Paulo Zanoni

From: Paulo Zanoni <[email protected]>

Since we merged runtime PM support for DPMS, it is possible that these functions will be called when the power wells are disabled but a mode is "set", resulting in "failed assertion" and "device suspended while reading register" WARNs.

Quoted 7 lines

To reproduce the bug: disable all screens using mode unset, do a modeset on one screen, disable it using DPMS, then try to do a mode unset on it again to see the WARNs.

Testcase: igt/rpm_rpm/dpms-mode-unset-lpsp
Testcase: igt/rpm_rpm/dpms-mode-unset-non-lpsp

Hm, where do we call these asserts while the pipe is off? Do you have some example backtraces at hand?
-Daniel

Quoted 43 lines
---
 drivers/gpu/drm/i915/intel_display.c | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)
drivers/gpu/drm/i915/intel_display.c +10 -4
index 54e3af9..7ad46e2 100644
@@ -1216,7 +1216,9 @@ static void assert_cursor(struct drm_i915_private *dev_priv,
 	struct drm_device *dev = dev_priv->dev;
 	bool cur_state;
 
-	if (IS_845G(dev) || IS_I865G(dev))
+	if (!intel_display_power_enabled(dev_priv, POWER_DOMAIN_PIPE(pipe)))
+		cur_state = false;
+	else if (IS_845G(dev) || IS_I865G(dev))
 		cur_state = I915_READ(_CURACNTR) & CURSOR_ENABLE;
 	else
 		cur_state = I915_READ(CURCNTR(pipe)) & CURSOR_MODE;
@@ -1262,9 +1264,13 @@ static void assert_plane(struct drm_i915_private *dev_priv,
 	u32 val;
 	bool cur_state;
 
-	reg = DSPCNTR(plane);
-	val = I915_READ(reg);
-	cur_state = !!(val & DISPLAY_PLANE_ENABLE);
+	if (!intel_display_power_enabled(dev_priv, POWER_DOMAIN_PIPE(plane))) {
+		cur_state = false;
+	} else {
+		reg = DSPCNTR(plane);
+		val = I915_READ(reg);
+		cur_state = !!(val & DISPLAY_PLANE_ENABLE);
+	}
 	WARN(cur_state != state,
 	     "plane %c assertion failure (expected %s, current %s)\n",
 	     plane_name(plane), state_string(state), state_string(cur_state));
Signature · 6 lines
2.0.0

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
#3 lore raw

2014-07-17 5:38 GMT-03:00 Daniel Vetter <[email protected]>:

Quoted 12 lines · Paulo Zanoni
Quoted 11 lines

From: Paulo Zanoni <[email protected]>

Since we merged runtime PM support for DPMS, it is possible that these functions will be called when the power wells are disabled but a mode is "set", resulting in "failed assertion" and "device suspended while reading register" WARNs.

To reproduce the bug: disable all screens using mode unset, do a modeset on one screen, disable it using DPMS, then try to do a mode unset on it again to see the WARNs.

Quoted 6 lines
Quoted 3 lines

Testcase: igt/rpm_rpm/dpms-mode-unset-lpsp
Testcase: igt/rpm_rpm/dpms-mode-unset-non-lpsp

Hm, where do we call these asserts while the pipe is off? Do you have some example backtraces at hand?

Function __intel_set_mode() directly calls intel_crtc_disable(), which calls both assert_plane_disabled() and assert_cursor_disabled().

Quoted 50 lines

-Daniel

Quoted 39 lines
---
 drivers/gpu/drm/i915/intel_display.c | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)
drivers/gpu/drm/i915/intel_display.c +10 -5
index 54e3af9..7ad46e2 100644
@@ -1216,7 +1216,9 @@ static void assert_cursor(struct drm_i915_private *dev_priv,
      struct drm_device *dev = dev_priv->dev;
      bool cur_state;
 
-     if (IS_845G(dev) || IS_I865G(dev))
+     if (!intel_display_power_enabled(dev_priv, POWER_DOMAIN_PIPE(pipe)))
+             cur_state = false;
+     else if (IS_845G(dev) || IS_I865G(dev))
              cur_state = I915_READ(_CURACNTR) & CURSOR_ENABLE;
      else
              cur_state = I915_READ(CURCNTR(pipe)) & CURSOR_MODE;
@@ -1262,9 +1264,13 @@ static void assert_plane(struct drm_i915_private *dev_priv,
      u32 val;
      bool cur_state;
 
-     reg = DSPCNTR(plane);
-     val = I915_READ(reg);
-     cur_state = !!(val & DISPLAY_PLANE_ENABLE);
+     if (!intel_display_power_enabled(dev_priv, POWER_DOMAIN_PIPE(plane))) {
+             cur_state = false;
+     } else {
+             reg = DSPCNTR(plane);
+             val = I915_READ(reg);
+             cur_state = !!(val & DISPLAY_PLANE_ENABLE);
+     }
      WARN(cur_state != state,
           "plane %c assertion failure (expected %s, current %s)\n",
           plane_name(plane), state_string(state), state_string(cur_state));
--

2.0.0

Quoted 4 lines

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

--
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch

Signature · 1 line
Paulo Zanoni
#4 lore raw
Quoted 17 lines · Paulo Zanoni

2014-07-17 5:38 GMT-03:00 Daniel Vetter <[email protected]>:

Quoted 16 lines · Paulo Zanoni
Quoted 14 lines

From: Paulo Zanoni <[email protected]>

Since we merged runtime PM support for DPMS, it is possible that these functions will be called when the power wells are disabled but a mode is "set", resulting in "failed assertion" and "device suspended while reading register" WARNs.

To reproduce the bug: disable all screens using mode unset, do a modeset on one screen, disable it using DPMS, then try to do a mode unset on it again to see the WARNs.

Testcase: igt/rpm_rpm/dpms-mode-unset-lpsp
Testcase: igt/rpm_rpm/dpms-mode-unset-non-lpsp

Quoted 5 lines
Quoted 2 lines

Hm, where do we call these asserts while the pipe is off? Do you have some example backtraces at hand?

Function __intel_set_mode() directly calls intel_crtc_disable(), which calls both assert_plane_disabled() and assert_cursor_disabled().

Hm, I think it makes more sense to drop the three asserts in there. The modeset state checker will already notice when we've failed to turn off the pipe. And we check cursors and plane state in the enable sequence, too.

Since we use these asserts a lot to lock down the precise modeset sequence I actually prefer if they're a bit dumb and don't check the power wells.
-Daniel

Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
#5 lore raw

2014-07-17 10:23 GMT-03:00 Daniel Vetter <[email protected]>:

Quoted 24 lines · Paulo Zanoni
Quoted 17 lines

2014-07-17 5:38 GMT-03:00 Daniel Vetter <[email protected]>:

Quoted 16 lines · Paulo Zanoni
Quoted 14 lines

From: Paulo Zanoni <[email protected]>

Since we merged runtime PM support for DPMS, it is possible that these functions will be called when the power wells are disabled but a mode is "set", resulting in "failed assertion" and "device suspended while reading register" WARNs.

To reproduce the bug: disable all screens using mode unset, do a modeset on one screen, disable it using DPMS, then try to do a mode unset on it again to see the WARNs.

Testcase: igt/rpm_rpm/dpms-mode-unset-lpsp
Testcase: igt/rpm_rpm/dpms-mode-unset-non-lpsp

Quoted 5 lines
Quoted 2 lines

Hm, where do we call these asserts while the pipe is off? Do you have some example backtraces at hand?

Function __intel_set_mode() directly calls intel_crtc_disable(), which calls both assert_plane_disabled() and assert_cursor_disabled().

Quoted 8 lines

Hm, I think it makes more sense to drop the three asserts in there. The modeset state checker will already notice when we've failed to turn off the pipe. And we check cursors and plane state in the enable sequence, too.

Since we use these asserts a lot to lock down the precise modeset sequence I actually prefer if they're a bit dumb and don't check the power wells.

I can do this, but please notice that we already have power-well-checking code in many of the other assertions on our driver... And it will probably be just a matter of time since someone starts using the assertions again on a place where the power well can be disabled :)

Quoted 5 lines

-Daniel
--
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch

Signature · 1 line
Paulo Zanoni
#8 lore raw

PATCH 1/1 drm/i915: remove plane/cursor/pipe assertions from intel_crtc_disable

In reply to Daniel Vetter

From: Paulo Zanoni <[email protected]>

Since we merged runtime PM support for DPMS, it is possible that these assertions will be called when the power wells are disabled but a mode is "set", resulting in "failed assertion" and "device suspended while reading register" WARNs.

To reproduce the bug: disable all screens using mode unset, do a modeset on one screen, disable it using DPMS, then try to do a mode unset on it again to see the WARNs.

v2: The first version of this patch changed the assertions to also check the power domains. Daniel suggested that it would be better to just remove the assertions: "The modeset state checker will already notice when we've failed to turn off the pipe. And we check cursors and plane state in the enable sequence, too. Since we use these asserts a lot to lock down the precise modeset sequence I actually prefer if they're a bit dumb and don't check the power wells."

Testcase: igt/rpm_rpm/dpms-mode-unset-lpsp
Testcase: igt/rpm_rpm/dpms-mode-unset-non-lpsp

---
 drivers/gpu/drm/i915/intel_display.c | 4 ----
 1 file changed, 4 deletions(-)
drivers/gpu/drm/i915/intel_display.c +0 -4
index cb40a94..f4ef59e 100644
@@ -4930,10 +4930,6 @@ static void intel_crtc_disable(struct drm_crtc *crtc)
 	intel_crtc_update_sarea(crtc, false);
 	dev_priv->display.off(crtc);
 
-	assert_plane_disabled(dev->dev_private, to_intel_crtc(crtc)->plane);
-	assert_cursor_disabled(dev_priv, pipe);
-	assert_pipe_disabled(dev->dev_private, pipe);
-
 	if (crtc->primary->fb) {
 		mutex_lock(&dev->struct_mutex);
 		intel_unpin_fb_obj(old_obj);
Signature · 1 line
2.0.0
#6 lore raw
In reply to Paulo Zanoni
Quoted 5 lines · Paulo Zanoni

I can do this, but please notice that we already have power-well-checking code in many of the other assertions on our driver... And it will probably be just a matter of time since someone starts using the assertions again on a place where the power well can be disabled :)

The only one I've found outside of the hw state readout code and error capture code (and there we obviously need them) is in assert_pipe. This was added in 693101618a4be. Tbh I wonder whether we could revert that with the patch to ditch the assert from intel_crtc_disable? -Daniel

Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
#9 lore raw

Re: [PATCH] drm/i915: remove plane/cursor/pipe assertions from intel_crtc_disable

In reply to Paulo Zanoni
Quoted 20 lines · Paulo Zanoni

From: Paulo Zanoni <[email protected]>

Since we merged runtime PM support for DPMS, it is possible that these assertions will be called when the power wells are disabled but a mode is "set", resulting in "failed assertion" and "device suspended while reading register" WARNs.

To reproduce the bug: disable all screens using mode unset, do a modeset on one screen, disable it using DPMS, then try to do a mode unset on it again to see the WARNs.

v2: The first version of this patch changed the assertions to also check the power domains. Daniel suggested that it would be better to just remove the assertions: "The modeset state checker will already notice when we've failed to turn off the pipe. And we check cursors and plane state in the enable sequence, too. Since we use these asserts a lot to lock down the precise modeset sequence I actually prefer if they're a bit dumb and don't check the power wells."

Quoted 3 lines

Testcase: igt/rpm_rpm/dpms-mode-unset-lpsp
Testcase: igt/rpm_rpm/dpms-mode-unset-non-lpsp

Queued for -next, thanks for the patch.
-Daniel

Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
#7 lore raw
In reply to Daniel Vetter

2014-07-17 13:58 GMT-03:00 Daniel Vetter <[email protected]>:

Quoted 11 lines · Paulo Zanoni
Quoted 5 lines

I can do this, but please notice that we already have power-well-checking code in many of the other assertions on our driver... And it will probably be just a matter of time since someone starts using the assertions again on a place where the power well can be disabled :)

The only one I've found outside of the hw state readout code and error capture code (and there we obviously need them) is in assert_pipe. This was added in 693101618a4be. Tbh I wonder whether we could revert that with the patch to ditch the assert from intel_crtc_disable?

I was talking about all that hw state readout code (which are also used as assertions) and basically every single caller of intel_display_power_enabled().

Quoted 5 lines

-Daniel
--
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch

Signature · 1 line
Paulo Zanoni