Skip to contentSkip to conversation

drm/i915/psr: Lockless version of psr_wait_for_idle

PATCH In review Tarun Vyas 3 participants 6 messages Jun 27, 2018 – Jul 2, 2018

Subsystems (2)
lore
Share & tools
lore

Tarun Vyas

author
#0 lore raw

PATCH v8 1/2 drm/i915/psr: Lockless version of psr_wait_for_idle

This is a lockless version of the exisiting psr_wait_for_idle(). We want to wait for PSR to idle out inside intel_pipe_update_start. At the time of a pipe update, we should never race with any psr enable or disable code, which is a part of crtc enable/disable. The follow up patch will use this lockless wait inside pipe_update_ start to wait for PSR to idle out before checking for vblank evasion. We need to keep the wait in pipe_update_start to as less as it can be. So,we can live and flourish w/o taking any psr locks at all.

Even if psr is never enabled, psr2_enabled will be false and this function will wait for PSR1 to idle out, which should just return immediately, so a very short (~1-2 usec) wait for cases where PSR is disabled.

v2: Add comment to explain the 25msec timeout (DK)

v3: Rename psr_wait_for_idle to __psr_wait_for_idle_locked to avoid
naming conflicts and propagate err (if any) to the caller (Chris)

v5: Form a series with the next patch

v7: Better explain the need for lockless wait and increase the max
timeout to handle refresh rates < 60 Hz (Daniel Vetter)

v8: Rebase

---
 drivers/gpu/drm/i915/intel_drv.h |  1 +
 drivers/gpu/drm/i915/intel_psr.c | 36 ++++++++++++++++++++++++++++++++++--
 2 files changed, 35 insertions(+), 2 deletions(-)
drivers/gpu/drm/i915/intel_drv.h +1 -0
index a6ff2600a3a0..b9b70321c054 100644
@@ -1922,6 +1922,7 @@ void intel_psr_compute_config(struct intel_dp *intel_dp,
 void intel_psr_irq_control(struct drm_i915_private *dev_priv, bool debug);
 void intel_psr_irq_handler(struct drm_i915_private *dev_priv, u32 psr_iir);
 void intel_psr_short_pulse(struct intel_dp *intel_dp);
+int intel_psr_wait_for_idle(struct drm_i915_private *dev_priv);
 
 /* intel_runtime_pm.c */
 int intel_power_domains_init(struct drm_i915_private *);
drivers/gpu/drm/i915/intel_psr.c +34 -2
index 45f1cb7d6c04..23acc9ac8d4d 100644
@@ -717,7 +717,39 @@ void intel_psr_disable(struct intel_dp *intel_dp,
 	cancel_work_sync(&dev_priv->psr.work);
 }
 
-static bool psr_wait_for_idle(struct drm_i915_private *dev_priv)
+int intel_psr_wait_for_idle(struct drm_i915_private *dev_priv)
+{
+	i915_reg_t reg;
+	u32 mask;
+
+	/*
+	 * The sole user right now is intel_pipe_update_start(),
+	 * which won't race with psr_enable/disable, which is
+	 * where psr2_enabled is written to. So, we don't need
+	 * to acquire the psr.lock. More importantly, we want the
+	 * latency inside intel_pipe_update_start() to be as low
+	 * as possible, so no need to acquire psr.lock when it is
+	 * not needed and will induce latencies in the atomic
+	 * update path.
+	 */
+	if (dev_priv->psr.psr2_enabled) {
+		reg = EDP_PSR2_STATUS;
+		mask = EDP_PSR2_STATUS_STATE_MASK;
+	} else {
+		reg = EDP_PSR_STATUS;
+		mask = EDP_PSR_STATUS_STATE_MASK;
+	}
+
+	/*
+	 * Max time for PSR to idle = Inverse of the refresh rate +
+	 * 6 ms of exit training time + 1.5 ms of aux channel
+	 * handshake. 50 msec is defesive enough to cover everything.
+	 */
+	return intel_wait_for_register(dev_priv, reg, mask,
+				       EDP_PSR_STATUS_STATE_IDLE, 50);
+}
+
+static bool __psr_wait_for_idle_locked(struct drm_i915_private *dev_priv)
 {
 	struct intel_dp *intel_dp;
 	i915_reg_t reg;
@@ -763,7 +795,7 @@ static void intel_psr_work(struct work_struct *work)
 	 * PSR might take some time to get fully disabled
 	 * and be ready for re-enable.
 	 */
-	if (!psr_wait_for_idle(dev_priv))
+	if (!__psr_wait_for_idle_locked(dev_priv))
 		goto unlock;
 
 	/*
Signature · 6 lines
2.13.5

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

PATCH v8 2/2 drm/i915: Wait for PSR exit before checking for vblank evasion

The PIPEDSL freezes on PSR entry and if PSR hasn't fully exited, then the pipe_update_start call schedules itself out to check back later.

On ChromeOS-4.4 kernel, which is fairly up-to-date w.r.t drm/i915 but lags w.r.t core kernel code, hot plugging an external display triggers tons of "potential atomic update errors" in the dmesg, on *pipe A*. A closer analysis reveals that we try to read the scanline 3 times and eventually timeout, b/c PSR hasn't exited fully leading to a PIPEDSL stuck @ 1599. This issue is not seen on upstream kernels, b/c for *some* reason we loop inside intel_pipe_update start for ~2+ msec which in this case is more than enough to exit PSR fully, hence an *unstuck* PIPEDSL counter, hence no error. On the other hand, the ChromeOS kernel spends ~1.1 msec looping inside intel_pipe_update_start and hence errors out b/c the source is still in PSR.

Regardless, we should wait for PSR exit (if PSR is disabled, we incur a ~1-2 usec penalty) before reading the PIPEDSL, b/c if we haven't fully exited PSR, then checking for vblank evasion isn't actually applicable.

v4: Comment explaining psr_wait after enabling VBL interrupts (DK)

v5: CAN_PSR() to handle platforms that don't support PSR.

v6: Handle local_irq_disable on early return (Chris)

---
 drivers/gpu/drm/i915/intel_sprite.c | 20 ++++++++++++++++----
 1 file changed, 16 insertions(+), 4 deletions(-)
drivers/gpu/drm/i915/intel_sprite.c +16 -4
index 344c0e709b19..4990d6e84ddf 100644
@@ -107,13 +107,21 @@ void intel_pipe_update_start(const struct intel_crtc_state *new_crtc_state)
 						      VBLANK_EVASION_TIME_US);
 	max = vblank_start - 1;
 
-	local_irq_disable();
-
 	if (min <= 0 || max <= 0)
-		return;
+		goto irq_disable;
 
 	if (WARN_ON(drm_crtc_vblank_get(&crtc->base)))
-		return;
+		goto irq_disable;
+
+	/*
+	 * Wait for psr to idle out after enabling the VBL interrupts
+	 * VBL interrupts will start the PSR exit and prevent a PSR
+	 * re-entry as well.
+	 */
+	if (CAN_PSR(dev_priv) && intel_psr_wait_for_idle(dev_priv))
+		DRM_ERROR("PSR idle timed out, atomic update may fail\n");
+
+	local_irq_disable();
 
 	crtc->debug.min_vbl = min;
 	crtc->debug.max_vbl = max;
@@ -171,6 +179,10 @@ void intel_pipe_update_start(const struct intel_crtc_state *new_crtc_state)
 	crtc->debug.start_vbl_count = intel_crtc_get_vblank_counter(crtc);
 
 	trace_i915_pipe_update_vblank_evaded(crtc);
+	return;
+
+irq_disable:
+	local_irq_disable();
 }
 
 /**
Signature · 6 lines
2.13.5

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

Re: [PATCH v8 2/2] drm/i915: Wait for PSR exit before checking for vblank evasion

Quoted 25 lines · Tarun Vyas

The PIPEDSL freezes on PSR entry and if PSR hasn't fully exited, then the pipe_update_start call schedules itself out to check back later.

On ChromeOS-4.4 kernel, which is fairly up-to-date w.r.t drm/i915 but lags w.r.t core kernel code, hot plugging an external display triggers
tons of "potential atomic update errors" in the dmesg, on *pipe A*. A closer analysis reveals that we try to read the scanline 3 times and eventually timeout, b/c PSR hasn't exited fully leading to a PIPEDSL stuck @ 1599. This issue is not seen on upstream kernels, b/c for *some*
reason we loop inside intel_pipe_update start for ~2+ msec which in this
case is more than enough to exit PSR fully, hence an *unstuck* PIPEDSL
counter, hence no error. On the other hand, the ChromeOS kernel spends
~1.1 msec looping inside intel_pipe_update_start and hence errors out b/c the source is still in PSR.

Regardless, we should wait for PSR exit (if PSR is disabled, we incur a ~1-2 usec penalty) before reading the PIPEDSL, b/c if we haven't fully exited PSR, then checking for vblank evasion isn't actually applicable.

Quoted 5 lines

v4: Comment explaining psr_wait after enabling VBL interrupts (DK)

v5: CAN_PSR() to handle platforms that don't support PSR.

v6: Handle local_irq_disable on early return (Chris)


Series Reviewed-by: Dhinakaran Pandiyan
<[email protected]>

Daniel's questions were addressed over IRC, I'll push this version if they aren't any other concerns.

-DK

Quoted 51 lines

---
 drivers/gpu/drm/i915/intel_sprite.c | 20 ++++++++++++++++----

 1 file changed, 16 insertions(+), 4 deletions(-)
drivers/gpu/drm/i915/intel_sprite.c +0 -0
index 344c0e709b19..4990d6e84ddf 100644
@@ -107,13 +107,21 @@ void intel_pipe_update_start(const struct
 intel_crtc_state *new_crtc_state)

        VBLANK_EVASION _TIME_US);
  max = vblank_start - 1;

drivers/gpu/drm/i915/intel_sprite.c +0 -2
-	local_irq_disable();
-

  if (min <= 0 || max <= 0)

drivers/gpu/drm/i915/intel_sprite.c +1 -1
-		return;
+		goto irq_disable;


  if (WARN_ON(drm_crtc_vblank_get(&crtc->base)))

drivers/gpu/drm/i915/intel_sprite.c +4 -1
-		return;
+		goto irq_disable;
+
+	/*
+	 * Wait for psr to idle out after enabling the VBL

interrupts

drivers/gpu/drm/i915/intel_sprite.c +5 -0
+	 * VBL interrupts will start the PSR exit and prevent a PSR
+	 * re-entry as well.
+	 */
+	if (CAN_PSR(dev_priv) && intel_psr_wait_for_idle(dev_priv))
+		DRM_ERROR("PSR idle timed out, atomic update may

fail\n");

drivers/gpu/drm/i915/intel_sprite.c +2 -0
+
+	local_irq_disable();


  crtc->debug.min_vbl = min;
  crtc->debug.max_vbl = max;

drivers/gpu/drm/i915/intel_sprite.c +0 -0
@@ -171,6 +179,10 @@ void intel_pipe_update_start(const struct
 intel_crtc_state *new_crtc_state)

  crtc->debug.start_vbl_count = intel_crtc_get_vblank_counter(crtc);
trace_i915_pipe_update_vblank_evaded(crtc);

drivers/gpu/drm/i915/intel_sprite.c +4 -0
+	return;
+
+irq_disable:
+	local_irq_disable();
Quoted 3 lines

 }

 /**

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

#3 lore raw

Re: [PATCH v8 2/2] drm/i915: Wait for PSR exit before checking for vblank evasion

Quoted 39 lines · Dhinakaran Pandiyan
Quoted 38 lines · Tarun Vyas

The PIPEDSL freezes on PSR entry and if PSR hasn't fully exited, then
the pipe_update_start call schedules itself out to check back later.

On ChromeOS-4.4 kernel, which is fairly up-to-date w.r.t drm/i915 but
lags w.r.t core kernel code, hot plugging an external display triggers
tons of "potential atomic update errors" in the dmesg, on *pipe A*. A
closer analysis reveals that we try to read the scanline 3 times and
eventually timeout, b/c PSR hasn't exited fully leading to a PIPEDSL
stuck @ 1599. This issue is not seen on upstream kernels, b/c for *some*
reason we loop inside intel_pipe_update start for ~2+ msec which in this
case is more than enough to exit PSR fully, hence an *unstuck* PIPEDSL
counter, hence no error. On the other hand, the ChromeOS kernel spends
~1.1 msec looping inside intel_pipe_update_start and hence errors out
b/c the source is still in PSR.

Regardless, we should wait for PSR exit (if PSR is disabled, we incur
a ~1-2 usec penalty) before reading the PIPEDSL, b/c if we haven't fully exited PSR, then checking for vblank evasion isn't actually applicable.

v4: Comment explaining psr_wait after enabling VBL interrupts (DK)

v5: CAN_PSR() to handle platforms that don't support PSR.

Quoted 7 lines
Quoted 1 line

v6: Handle local_irq_disable on early return (Chris)


Series Reviewed-by: Dhinakaran Pandiyan
<[email protected]>

Daniel's questions were addressed over IRC, I'll push this version if they aren't any other concerns.

and pushed to -dinq with Daniel's irc ack.

"<dhnkrn> ickle: danvet ack on the PSR wait_for_idle series? <danvet> sure"

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

#4 lore raw

✓ Fi.CI.BAT: success for series starting with [v8,1/2] drm/i915/psr: Lockless version of psr_wait_for_idle

In reply to Tarun Vyas
== Series Details ==

== Series Details ==

Series: series starting with [v8,1/2] drm/i915/psr: Lockless version of psr_wait_for_idle
URL : https://patchwork.freedesktop.org/series/45524/
State : success

== Summary ==

= CI Bug Log - changes from CI_DRM_4391 -> Patchwork_9453 =

== Summary - SUCCESS ==

No regressions found.

External URL: https://patchwork.freedesktop.org/api/1.0/series/45524/revisions/1/mbox/

== Known issues ==

Here are the changes found in Patchwork_9453 that come from known issues:

=== IGT changes ===

    ==== Possible fixes ====

    igt@gem_ctx_create@basic-files:
      fi-glk-dsi:         DMESG-WARN (fdo#106954) -> PASS

fdo#106954 https://bugs.freedesktop.org/show_bug.cgi?id=106954

== Participating hosts (44 -> 40) ==

Additional (1): fi-byt-j1900
Missing (5): fi-ctg-p8600 fi-ilk-m540 fi-byt-squawks fi-bsw-cyan fi-hsw-4200u

== Build changes ==

* Linux: CI_DRM_4391 -> Patchwork_9453

CI_DRM_4391: 9134bde2e681ec232f4d8bec56aa3700177630fe @ git://anongit.freedesktop.org/gfx-ci/linux
IGT_4530: 0e98bf69f146eb72fe3a7c3b19a049b5786f0ca3 @ git://anongit.freedesktop.org/xorg/app/intel-gpu-tools
Patchwork_9453: 7a0d28ca8e878141795a1f3044fd373ad5593ac9 @ git://anongit.freedesktop.org/gfx-ci/linux

== Linux commits ==

7a0d28ca8e87 drm/i915: Wait for PSR exit before checking for vblank evasion f6ed4df15922 drm/i915/psr: Lockless version of psr_wait_for_idle

== Logs ==

For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_9453/issues.html
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#5 lore raw

✓ Fi.CI.IGT: success for series starting with [v8,1/2] drm/i915/psr: Lockless version of psr_wait_for_idle

In reply to Tarun Vyas
== Series Details ==

== Series Details ==

Series: series starting with [v8,1/2] drm/i915/psr: Lockless version of psr_wait_for_idle
URL : https://patchwork.freedesktop.org/series/45524/
State : success

== Summary ==

= CI Bug Log - changes from CI_DRM_4391_full -> Patchwork_9453_full =

== Summary - WARNING ==

Minor unknown changes coming with Patchwork_9453_full need to be verified manually.

If you think the reported changes have nothing to do with the changes introduced in Patchwork_9453_full, please notify your bug team to allow them to document this new failure mode, which will reduce false positives in CI.

== Possible new issues ==

Here are the unknown changes that may have been introduced in Patchwork_9453_full:

=== IGT changes ===

    ==== Warnings ====

    igt@gem_exec_schedule@deep-bsd2:
      shard-kbl:          SKIP -> PASS

    igt@gem_exec_schedule@deep-vebox:
      shard-kbl:          PASS -> SKIP

== Known issues ==

Here are the changes found in Patchwork_9453_full that come from known issues:

=== IGT changes ===

    ==== Issues hit ====

    igt@drv_selftest@live_gtt:
      shard-glk:          PASS -> FAIL (fdo#105347)

    igt@drv_suspend@shrink:
      shard-apl:          PASS -> FAIL (fdo#106886, fdo#107054)

    igt@gem_exec_big:
      shard-hsw:          PASS -> INCOMPLETE (fdo#103540)

    igt@gem_exec_schedule@pi-ringfull-bsd1:
      shard-kbl:          NOTRUN -> FAIL (fdo#103158)

    igt@gem_exec_schedule@pi-ringfull-render:
      {shard-glk9}:       NOTRUN -> FAIL (fdo#103158)

    igt@gem_exec_schedule@preemptive-hang-render:
      shard-snb:          SKIP -> INCOMPLETE (fdo#105411)

    igt@kms_flip@plain-flip-ts-check-interruptible:
      shard-glk:          NOTRUN -> FAIL (fdo#100368)

    igt@kms_flip_tiling@flip-to-x-tiled:
      {shard-glk9}:       NOTRUN -> FAIL (fdo#104724)

    igt@kms_flip_tiling@flip-to-y-tiled:
      shard-glk:          PASS -> FAIL (fdo#104724, fdo#103822)

    igt@kms_frontbuffer_tracking@fbc-2p-primscrn-pri-shrfb-draw-mmap-wc:
      {shard-glk9}:       NOTRUN -> FAIL (fdo#103167, fdo#104724)

    igt@kms_setmode@basic:
      {shard-glk9}:       NOTRUN -> FAIL (fdo#99912)

    
    ==== Possible fixes ====

    igt@gem_ctx_isolation@rcs0-s3:
      shard-kbl:          INCOMPLETE (fdo#103665) -> PASS

    igt@kms_flip_tiling@flip-y-tiled:
      shard-glk:          FAIL (fdo#104724, fdo#103822) -> PASS

{name}: This element is suppressed. This means it is ignored when computing
the status of the difference (SUCCESS, WARNING, or FAILURE).

fdo#100368 https://bugs.freedesktop.org/show_bug.cgi?id=100368
fdo#103158 https://bugs.freedesktop.org/show_bug.cgi?id=103158
fdo#103167 https://bugs.freedesktop.org/show_bug.cgi?id=103167
fdo#103540 https://bugs.freedesktop.org/show_bug.cgi?id=103540
fdo#103665 https://bugs.freedesktop.org/show_bug.cgi?id=103665
fdo#103822 https://bugs.freedesktop.org/show_bug.cgi?id=103822
fdo#104724 https://bugs.freedesktop.org/show_bug.cgi?id=104724
fdo#105347 https://bugs.freedesktop.org/show_bug.cgi?id=105347
fdo#105411 https://bugs.freedesktop.org/show_bug.cgi?id=105411
fdo#106886 https://bugs.freedesktop.org/show_bug.cgi?id=106886
fdo#107054 https://bugs.freedesktop.org/show_bug.cgi?id=107054
fdo#99912 https://bugs.freedesktop.org/show_bug.cgi?id=99912

== Participating hosts (6 -> 6) ==

No changes in participating hosts

== Build changes ==

* Linux: CI_DRM_4391 -> Patchwork_9453

CI_DRM_4391: 9134bde2e681ec232f4d8bec56aa3700177630fe @ git://anongit.freedesktop.org/gfx-ci/linux
IGT_4530: 0e98bf69f146eb72fe3a7c3b19a049b5786f0ca3 @ git://anongit.freedesktop.org/xorg/app/intel-gpu-tools
Patchwork_9453: 7a0d28ca8e878141795a1f3044fd373ad5593ac9 @ git://anongit.freedesktop.org/gfx-ci/linux
piglit_4509: fdc5a4ca11124ab8413c7988896eec4c97336694 @ git://anongit.freedesktop.org/piglit

== Logs ==

For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_9453/shards.html
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx