Merge tag 'drm-next-2022-01-07' of git://anongit.freedesktop.org/drm/drm

Pull drm updates from Dave Airlie:
 "Highlights are support for privacy screens found in new laptops, a
  bunch of nomodeset refactoring, and i915 enables ADL-P systems by
  default, while starting to add RPL-S support.

  vmwgfx adds GEM and support for OpenGL 4.3 features in userspace.

  Lots of internal refactorings around dma reservations, and lots of
  driver refactoring as well.

  Summary:

  core:
   - add privacy screen support
   - move nomodeset option into drm subsystem
   - clean up nomodeset handling in drivers
   - make drm_irq.c legacy
   - fix stack_depot name conflicts
   - remove DMA_BUF_SET_NAME ioctl restrictions
   - sysfs: send hotplug event
   - replace several DRM_* logging macros with drm_*
   - move hashtable to legacy code
   - add error return from gem_create_object
   - cma-helper: improve interfaces, drop CONFIG_DRM_KMS_CMA_HELPER
   - kernel.h related include cleanups
   - support XRGB2101010 source buffers

  ttm:
   - don't include drm hashtable
   - stop pruning fences after wait
   - documentation updates

  dma-buf:
   - add dma_resv selftest
   - add debugfs helpers
   - remove dma_resv_get_excl_unlocked
   - documentation
   - make fences mandatory in dma_resv_add_excl_fence

  dp:
   - add link training delay helpers

  gem:
   - link shmem/cma helpers into separate modules
   - use dma_resv iteratior
   - import dma-buf namespace into gem helper modules

  scheduler:
   - fence grab fix
   - lockdep fixes

  bridge:
   - switch to managed MIPI DSI helpers
   - register and attach during probe fixes
   - convert to YAML in several places.

  panel:
   - add bunch of new panesl

  simpledrm:
   - support FB_DAMAGE_CLIPS
   - support virtual screen sizes
   - add Apple M1 support

  amdgpu:
   - enable seamless boot for DCN 3.01
   - runtime PM fixes
   - use drm_kms_helper_connector_hotplug_event
   - get all fences at once
   - use generic drm fb helpers
   - PSR/DPCD/LTTPR/DSC/PM/RAS/OLED/SRIOV fixes
   - add smart trace buffer (STB) for supported GPUs
   - display debugfs entries
   - new SMU debug option
   - Documentation update

  amdkfd:
   - IP discovery enumeration refactor
   - interface between driver fixes
   - SVM fixes
   - kfd uapi header to define some sysfs bitfields.

  i915:
   - support VESA panel backlights
   - enable ADL-P by default
   - add eDP privacy screen support
   - add Raptor Lake S (RPL-S) support
   - DG2 page table support
   - lots of GuC/HuC fw refactoring
   - refactored i915->gt interfaces
   - CD clock squashing support
   - enable 10-bit gamma support
   - update ADL-P DMC fw to v2.14
   - enable runtime PM autosuspend by default
   - ADL-P DSI support
   - per-lane DP drive settings for ICL+
   - add support for pipe C/D DMC firmware
   - Atomic gamma LUT updates
   - remove CCS FB stride restrictions on ADL-P
   - VRR platform support for display 11
   - add support for display audio codec keepalive
   - lots of display refactoring
   - fix runtime PM handling during PXP suspend
   - improved eviction performance with async TTM moves
   - async VMA unbinding improvements
   - VMA locking refactoring
   - improved error capture robustness
   - use per device iommu checks
   - drop bits stealing from i915_sw_fence function ptr
   - remove dma_resv_prune
   - add IC cache invalidation on DG2

  nouveau:
   - crc fixes
   - validate LUTs in atomic check
   - set HDMI AVI RGB quant to full

  tegra:
   - buffer objects reworks for dma-buf compat
   - NVDEC driver uAPI support
   - power management improvements

  etnaviv:
   - IOMMU enabled system support
   - fix > 4GB command buffer mapping
   - close a DoS vector
   - fix spurious GPU resets

  ast:
   - fix i2c initialization

  rcar-du:
   - DSI output support

  exynos:
   - replace legacy gpio interface
   - implement generic GEM object mmap

  msm:
   - dpu plane state cleanup in prep for multirect
   - dpu debugfs cleanups
   - dp support for sc7280
   - a506 support
   - removal of struct_mutex
   - remove old eDP sub-driver

  anx7625:
   - support MIPI DSI input
   - support HDMI audio
   - fix reading EDID

  lvds:
   - fix bridge DT bindings

  megachips:
   - probe both bridges before registering

  dw-hdmi:
   - allow interlace on bridge

  ps8640:
   - enable runtime PM
   - support aux-bus

  tx358768:
   - enable reference clock
   - add pulse mode support

  ti-sn65dsi86:
   - use regmap bulk write
   - add PWM support

  etnaviv:
   - get all fences at once

  gma500:
   - gem object cleanups

  kmb:
   - enable fb console

  radeon:
   - use dma_resv_wait_timeout

  rockchip:
   - add DSP hold timeout
   - suspend/resume fixes
   - PLL clock fixes
   - implement mmap in GEM object functions
   - use generic fbdev emulation

  sun4i:
   - use CMA helpers without vmap support

  vc4:
   - fix HDMI-CEC hang with display is off
   - power on HDMI controller while disabling
   - support 4K@60Hz modes
   - support 10-bit YUV 4:2:0 output

  vmwgfx:
   - fix leak on probe errors
   - fail probing on broken hosts
   - new placement for MOB page tables
   - hide internal BOs from userspace
   - implement GEM support
   - implement GL 4.3 support

  virtio:
   - overflow fixes

  xen:
   - implement mmap as GEM object function

  omapdrm:
   - fix scatterlist export
   - support virtual planes

  mediatek:
   - MT8192 support
   - CMDQ refinement"

* tag 'drm-next-2022-01-07' of git://anongit.freedesktop.org/drm/drm: (1241 commits)
  drm/amdgpu: no DC support for headless chips
  drm/amd/display: fix dereference before NULL check
  drm/amdgpu: always reset the asic in suspend (v2)
  drm/amdgpu: put SMU into proper state on runpm suspending for BOCO capable platform
  drm/amd/display: Fix the uninitialized variable in enable_stream_features()
  drm/amdgpu: fix runpm documentation
  amdgpu/pm: Make sysfs pm attributes as read-only for VFs
  drm/amdgpu: save error count in RAS poison handler
  drm/amdgpu: drop redundant semicolon
  drm/amd/display: get and restore link res map
  drm/amd/display: support dynamic HPO DP link encoder allocation
  drm/amd/display: access hpo dp link encoder only through link resource
  drm/amd/display: populate link res in both detection and validation
  drm/amd/display: define link res and make it accessible to all link interfaces
  drm/amd/display: 3.2.167
  drm/amd/display: [FW Promotion] Release 0.0.98
  drm/amd/display: Undo ODM combine
  drm/amd/display: Add reg defs for DCN303
  drm/amd/display: Changed pipe split policy to allow for multi-display pipe split
  drm/amd/display: Set optimize_pwr_state for DCN31
  ...

1 files changed, 49 insertions, 42 deletions

diff --git a/drivers/gpu/drm/msm/msm_gpu_devfreq.c b/drivers/gpu/drm/msm/msm_gpu_devfreq.c
index 384e90c4b2a7..62405e980925 100644
--- a/drivers/gpu/drm/msm/msm_gpu_devfreq.c
+++ b/drivers/gpu/drm/msm/msm_gpu_devfreq.c
@@ -9,6 +9,7 @@
 
 #include <linux/devfreq.h>
 #include <linux/devfreq_cooling.h>
+#include <linux/units.h>
 
 /*
  * Power Management:
@@ -25,17 +26,6 @@ static int msm_devfreq_target(struct device *dev, unsigned long *freq,
 	 * to something that actually is in the opp table:
 	 */
 	opp = devfreq_recommended_opp(dev, freq, flags);
-
-	/*
-	 * If the GPU is idle, devfreq is not aware, so just ignore
-	 * it's requests
-	 */
-	if (gpu->devfreq.idle_freq) {
-		gpu->devfreq.idle_freq = *freq;
-		dev_pm_opp_put(opp);
-		return 0;
-	}
-
 	if (IS_ERR(opp))
 		return PTR_ERR(opp);
 
@@ -53,9 +43,6 @@ static int msm_devfreq_target(struct device *dev, unsigned long *freq,
 
 static unsigned long get_freq(struct msm_gpu *gpu)
 {
-	if (gpu->devfreq.idle_freq)
-		return gpu->devfreq.idle_freq;
-
 	if (gpu->funcs->gpu_get_freq)
 		return gpu->funcs->gpu_get_freq(gpu);
 
@@ -93,6 +80,7 @@ static struct devfreq_dev_profile msm_devfreq_profile = {
 	.get_cur_freq = msm_devfreq_get_cur_freq,
 };
 
+static void msm_devfreq_boost_work(struct kthread_work *work);
 static void msm_devfreq_idle_work(struct kthread_work *work);
 
 void msm_devfreq_init(struct msm_gpu *gpu)
@@ -103,6 +91,12 @@ void msm_devfreq_init(struct msm_gpu *gpu)
 	if (!gpu->funcs->gpu_busy)
 		return;
 
+	dev_pm_qos_add_request(&gpu->pdev->dev, &df->idle_freq,
+			       DEV_PM_QOS_MAX_FREQUENCY,
+			       PM_QOS_MAX_FREQUENCY_DEFAULT_VALUE);
+	dev_pm_qos_add_request(&gpu->pdev->dev, &df->boost_freq,
+			       DEV_PM_QOS_MIN_FREQUENCY, 0);
+
 	msm_devfreq_profile.initial_freq = gpu->fast_rate;
 
 	/*
@@ -133,13 +127,19 @@ void msm_devfreq_init(struct msm_gpu *gpu)
 		gpu->cooling = NULL;
 	}
 
+	msm_hrtimer_work_init(&df->boost_work, gpu->worker, msm_devfreq_boost_work,
+			      CLOCK_MONOTONIC, HRTIMER_MODE_REL);
 	msm_hrtimer_work_init(&df->idle_work, gpu->worker, msm_devfreq_idle_work,
 			      CLOCK_MONOTONIC, HRTIMER_MODE_REL);
 }
 
 void msm_devfreq_cleanup(struct msm_gpu *gpu)
 {
+	struct msm_gpu_devfreq *df = &gpu->devfreq;
+
 	devfreq_cooling_unregister(gpu->cooling);
+	dev_pm_qos_remove_request(&df->boost_freq);
+	dev_pm_qos_remove_request(&df->idle_freq);
 }
 
 void msm_devfreq_resume(struct msm_gpu *gpu)
@@ -155,12 +155,40 @@ void msm_devfreq_suspend(struct msm_gpu *gpu)
 	devfreq_suspend_device(gpu->devfreq.devfreq);
 }
 
+static void msm_devfreq_boost_work(struct kthread_work *work)
+{
+	struct msm_gpu_devfreq *df = container_of(work,
+			struct msm_gpu_devfreq, boost_work.work);
+
+	dev_pm_qos_update_request(&df->boost_freq, 0);
+}
+
+void msm_devfreq_boost(struct msm_gpu *gpu, unsigned factor)
+{
+	struct msm_gpu_devfreq *df = &gpu->devfreq;
+	uint64_t freq;
+
+	freq = get_freq(gpu);
+	freq *= factor;
+
+	/*
+	 * A nice little trap is that PM QoS operates in terms of KHz,
+	 * while devfreq operates in terms of Hz:
+	 */
+	do_div(freq, HZ_PER_KHZ);
+
+	dev_pm_qos_update_request(&df->boost_freq, freq);
+
+	msm_hrtimer_queue_work(&df->boost_work,
+			       ms_to_ktime(msm_devfreq_profile.polling_ms),
+			       HRTIMER_MODE_REL);
+}
+
 void msm_devfreq_active(struct msm_gpu *gpu)
 {
 	struct msm_gpu_devfreq *df = &gpu->devfreq;
 	struct devfreq_dev_status status;
 	unsigned int idle_time;
-	unsigned long target_freq = df->idle_freq;
 
 	if (!df->devfreq)
 		return;
@@ -170,12 +198,6 @@ void msm_devfreq_active(struct msm_gpu *gpu)
 	 */
 	hrtimer_cancel(&df->idle_work.timer);
 
-	/*
-	 * Hold devfreq lock to synchronize with get_dev_status()/
-	 * target() callbacks
-	 */
-	mutex_lock(&df->devfreq->lock);
-
 	idle_time = ktime_to_ms(ktime_sub(ktime_get(), df->idle_time));
 
 	/*
@@ -183,21 +205,18 @@ void msm_devfreq_active(struct msm_gpu *gpu)
 	 * interval, then we won't meet the threshold of busyness for
 	 * the governor to ramp up the freq.. so give some boost
 	 */
-	if (idle_time > msm_devfreq_profile.polling_ms/2) {
-		target_freq *= 2;
+	if (idle_time > msm_devfreq_profile.polling_ms) {
+		msm_devfreq_boost(gpu, 2);
 	}
 
-	df->idle_freq = 0;
-
-	msm_devfreq_target(&gpu->pdev->dev, &target_freq, 0);
+	dev_pm_qos_update_request(&df->idle_freq,
+				  PM_QOS_MAX_FREQUENCY_DEFAULT_VALUE);
 
 	/*
 	 * Reset the polling interval so we aren't inconsistent
 	 * about freq vs busy/total cycles
 	 */
 	msm_devfreq_get_dev_status(&gpu->pdev->dev, &status);
-
-	mutex_unlock(&df->devfreq->lock);
 }
 
 
@@ -206,23 +225,11 @@ static void msm_devfreq_idle_work(struct kthread_work *work)
 	struct msm_gpu_devfreq *df = container_of(work,
 			struct msm_gpu_devfreq, idle_work.work);
 	struct msm_gpu *gpu = container_of(df, struct msm_gpu, devfreq);
-	unsigned long idle_freq, target_freq = 0;
-
-	/*
-	 * Hold devfreq lock to synchronize with get_dev_status()/
-	 * target() callbacks
-	 */
-	mutex_lock(&df->devfreq->lock);
-
-	idle_freq = get_freq(gpu);
-
-	if (gpu->clamp_to_idle)
-		msm_devfreq_target(&gpu->pdev->dev, &target_freq, 0);
 
 	df->idle_time = ktime_get();
-	df->idle_freq = idle_freq;
 
-	mutex_unlock(&df->devfreq->lock);
+	if (gpu->clamp_to_idle)
+		dev_pm_qos_update_request(&df->idle_freq, 0);
 }
 
 void msm_devfreq_idle(struct msm_gpu *gpu)