Merge tag 'drm-next-2020-04-01' of git://anongit.freedesktop.org/drm/drm

Pull drm updates from Dave Airlie:
 "This is the main drm pull request for 5.7-rc1.

  Highlights:

   - i915 enables Tigerlake by default

   - i915 and amdgpu have initial OLED backlight support

     [ Jani Nikula pipes up and points out that we've had a bunch of
       "initial support" code for a long time already, but only now
       Lyude made it actually work on real world machines ]

   - vmwgfx add support to enable OpenGL 4 userspace

   - zero length arrays are mostly removed.

  Detailed summary:

  new driver:
   - tidss: TI Keystone platform display subsystem

  core:
   - new drm device warn macros
   - mode config valid for memory constrained devices
   - bridge bus format negotation
   - consolidated fake vblank event handling
   - dma_alloc related cleanups
   - drop get_crtc callback
   - dp: DP1.4 EDID corruption test
   - EDID CEA detailed timings improvements
   - relicense some code to dual GPL2/MIT
   - convert core vblank support to per-crtc support
   - rework drm_global_mutex
   - bridge rework to allow omap_dss custom driver removeal
   - remove drm_fb_helper connector interrfaces
   - zero-length array removal

  scheduler:
   - support for modifying the sched list
   - revert job distribution optimization
   - helper to pick least loaded scheduler
   - race condition fix

  mst:
   - various fixes
   - remove register_connector callback

  i915:
   - uapi to allows userspace specific CS ring buffer sizes
   - Tigerlake enablement patches + Tigerlake enabled by default
   - new sysfs entries for engine properties
   - display/logging refactors
   - eDP/DP fixes for DPCD
   - Gen7 back to aliasing-ppgtt
   - Gen8+ irq refactor
   - Avoid globals
   - GEM locking fixes and simplifications
   - Ice Lake and Elkhart Lake fixes and workarounds
   - Baytrail/Haswell instability fix
   - GVT - VFIO edid better support

  amdgpu:
   - Rework VM update handling in preparation for HMM support
   - drm load/unload removal fixups
   - USB-C PD firmware updates
   - HDCP srm support
   - Navi/renoir PM watermark fixes
   - OLED panel support
   - Optimize debugging vram access
   - Use BACO for runtime pm
   - DC clock programming optimizations and fixes
   - PSP fw loading sequence updates
   - Drop DRIVER_USE_AGP
   - Remove legacy drm load and unload callbacks
   - ACP Kconfig fix
   - Lots of fixes across the driver

  amdkfd:
   - runtime pm support
   - more gfx config details in amdgpu

  radeon:
   - drop DRIVER_USE_AGP

  vmwgfx:
   - Disable DMA when SEV encryption in use
   - Shader Model 5 support - needed for GL4 support

  msm:
   - DPU resource manager refactor
   - dpu using atomic global state

  mediatek:
   - MT8183 DPI support

  etnaviv:
   - out-of-bounds read fix
   - expose feature flags for GC400 STM32MP1 SoC
   - runtime suspend entry fix
   - dma32 zone fix

  hisilicon:
   - mode selection fixes

  meson:
   - YUV420 support

  lima:
   - add support for heap buffers

  tinydrm:
   - removal of owner field
   - explicit DT dependency removal
   - YAML schema conversion

  tegra:
   - misc cleanups

  tidss:
   - new driver

  virtio:
   - better batching of notifications to host
   - memory handling reworked
   - shmem + gpu context fixes

  hibmc:
   - add gamma_set support
   - improve DPMS support

  pl111:
   - Integrator IM-PD1 support

  sun4i:
   - LVDS support for A20 + A33
   - DSI panel handling improvements"

* tag 'drm-next-2020-04-01' of git://anongit.freedesktop.org/drm/drm: (1537 commits)
  drm/i915/display: Fix mode private_flags comparison at atomic_check
  drm/i915/gt: Stage the transfer of the virtual breadcrumb
  drm/i915/gt: Select the deepest available parking mode for rc6
  drm/i915: Avoid live-lock with i915_vma_parked()
  drm/i915/gt: Treat idling as a RPS downclock event
  drm/i915/gt: Cancel a hung context if already closed
  drm/i915: Use explicit flag to mark unreachable intel_context
  drm/amdgpu: don't try to reserve training bo for sriov (v2)
  drm/amdgpu/smu11: add support for SMU AC/DC interrupts
  drm/amdgpu/swSMU: handle manual AC/DC notifications
  drm/amdgpu/swSMU: handle DC controlled by GPIO for navi1x
  drm/amdgpu/swSMU: set AC/DC mode based on the current system state (v2)
  drm/amdgpu/swSMU: correct the bootup power source for Navi1X (v2)
  drm/amdgpu/swSMU: use the smu11 power source helper for navi1x
  drm/amdgpu/smu11: add a helper to set the power source
  drm/amd/swSMU: add callback to set AC/DC power source (v2)
  drm/scheduler: fix rare NULL ptr race
  drm/amdgpu: fix the coverage issue to clear ArcVPGRs
  drm/amd/display: Fix pageflip event race condition for DCN.
  drm/[radeon|amdgpu]: Remove HAINAN board from max_sclk override check
  ...

1 files changed, 120 insertions, 49 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
index 94a6c42f29ea..6ed36a2c5f73 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
@@ -61,12 +61,24 @@ static int amdgpu_ctx_priority_permit(struct drm_file *filp,
 	return -EACCES;
 }
 
+static enum gfx_pipe_priority amdgpu_ctx_sched_prio_to_compute_prio(enum drm_sched_priority prio)
+{
+	switch (prio) {
+	case DRM_SCHED_PRIORITY_HIGH_HW:
+	case DRM_SCHED_PRIORITY_KERNEL:
+		return AMDGPU_GFX_PIPE_PRIO_HIGH;
+	default:
+		return AMDGPU_GFX_PIPE_PRIO_NORMAL;
+	}
+}
+
 static int amdgpu_ctx_init_entity(struct amdgpu_ctx *ctx, const u32 hw_ip, const u32 ring)
 {
 	struct amdgpu_device *adev = ctx->adev;
 	struct amdgpu_ctx_entity *entity;
 	struct drm_gpu_scheduler **scheds = NULL, *sched = NULL;
 	unsigned num_scheds = 0;
+	enum gfx_pipe_priority hw_prio;
 	enum drm_sched_priority priority;
 	int r;
 
@@ -79,46 +91,51 @@ static int amdgpu_ctx_init_entity(struct amdgpu_ctx *ctx, const u32 hw_ip, const
 	priority = (ctx->override_priority == DRM_SCHED_PRIORITY_UNSET) ?
 				ctx->init_priority : ctx->override_priority;
 	switch (hw_ip) {
-		case AMDGPU_HW_IP_GFX:
-			sched = &adev->gfx.gfx_ring[0].sched;
-			scheds = &sched;
-			num_scheds = 1;
-			break;
-		case AMDGPU_HW_IP_COMPUTE:
-			scheds = adev->gfx.compute_sched;
-			num_scheds = adev->gfx.num_compute_sched;
-			break;
-		case AMDGPU_HW_IP_DMA:
-			scheds = adev->sdma.sdma_sched;
-			num_scheds = adev->sdma.num_sdma_sched;
-			break;
-		case AMDGPU_HW_IP_UVD:
-			sched = &adev->uvd.inst[0].ring.sched;
-			scheds = &sched;
-			num_scheds = 1;
-			break;
-		case AMDGPU_HW_IP_VCE:
-			sched = &adev->vce.ring[0].sched;
-			scheds = &sched;
-			num_scheds = 1;
-			break;
-		case AMDGPU_HW_IP_UVD_ENC:
-			sched = &adev->uvd.inst[0].ring_enc[0].sched;
-			scheds = &sched;
-			num_scheds = 1;
-			break;
-		case AMDGPU_HW_IP_VCN_DEC:
-			scheds = adev->vcn.vcn_dec_sched;
-			num_scheds =  adev->vcn.num_vcn_dec_sched;
-			break;
-		case AMDGPU_HW_IP_VCN_ENC:
-			scheds = adev->vcn.vcn_enc_sched;
-			num_scheds =  adev->vcn.num_vcn_enc_sched;
-			break;
-		case AMDGPU_HW_IP_VCN_JPEG:
-			scheds = adev->jpeg.jpeg_sched;
-			num_scheds =  adev->jpeg.num_jpeg_sched;
-			break;
+	case AMDGPU_HW_IP_GFX:
+		sched = &adev->gfx.gfx_ring[0].sched;
+		scheds = &sched;
+		num_scheds = 1;
+		break;
+	case AMDGPU_HW_IP_COMPUTE:
+		hw_prio = amdgpu_ctx_sched_prio_to_compute_prio(priority);
+		scheds = adev->gfx.compute_prio_sched[hw_prio];
+		num_scheds = adev->gfx.num_compute_sched[hw_prio];
+		break;
+	case AMDGPU_HW_IP_DMA:
+		scheds = adev->sdma.sdma_sched;
+		num_scheds = adev->sdma.num_sdma_sched;
+		break;
+	case AMDGPU_HW_IP_UVD:
+		sched = &adev->uvd.inst[0].ring.sched;
+		scheds = &sched;
+		num_scheds = 1;
+		break;
+	case AMDGPU_HW_IP_VCE:
+		sched = &adev->vce.ring[0].sched;
+		scheds = &sched;
+		num_scheds = 1;
+		break;
+	case AMDGPU_HW_IP_UVD_ENC:
+		sched = &adev->uvd.inst[0].ring_enc[0].sched;
+		scheds = &sched;
+		num_scheds = 1;
+		break;
+	case AMDGPU_HW_IP_VCN_DEC:
+		sched = drm_sched_pick_best(adev->vcn.vcn_dec_sched,
+					    adev->vcn.num_vcn_dec_sched);
+		scheds = &sched;
+		num_scheds = 1;
+		break;
+	case AMDGPU_HW_IP_VCN_ENC:
+		sched = drm_sched_pick_best(adev->vcn.vcn_enc_sched,
+					    adev->vcn.num_vcn_enc_sched);
+		scheds = &sched;
+		num_scheds = 1;
+		break;
+	case AMDGPU_HW_IP_VCN_JPEG:
+		scheds = adev->jpeg.jpeg_sched;
+		num_scheds =  adev->jpeg.num_jpeg_sched;
+		break;
 	}
 
 	r = drm_sched_entity_init(&entity->entity, priority, scheds, num_scheds,
@@ -502,6 +519,29 @@ struct dma_fence *amdgpu_ctx_get_fence(struct amdgpu_ctx *ctx,
 	return fence;
 }
 
+static void amdgpu_ctx_set_entity_priority(struct amdgpu_ctx *ctx,
+					    struct amdgpu_ctx_entity *aentity,
+					    int hw_ip,
+					    enum drm_sched_priority priority)
+{
+	struct amdgpu_device *adev = ctx->adev;
+	enum gfx_pipe_priority hw_prio;
+	struct drm_gpu_scheduler **scheds = NULL;
+	unsigned num_scheds;
+
+	/* set sw priority */
+	drm_sched_entity_set_priority(&aentity->entity, priority);
+
+	/* set hw priority */
+	if (hw_ip == AMDGPU_HW_IP_COMPUTE) {
+		hw_prio = amdgpu_ctx_sched_prio_to_compute_prio(priority);
+		scheds = adev->gfx.compute_prio_sched[hw_prio];
+		num_scheds = adev->gfx.num_compute_sched[hw_prio];
+		drm_sched_entity_modify_sched(&aentity->entity, scheds,
+					      num_scheds);
+	}
+}
+
 void amdgpu_ctx_priority_override(struct amdgpu_ctx *ctx,
 				  enum drm_sched_priority priority)
 {
@@ -514,13 +554,11 @@ void amdgpu_ctx_priority_override(struct amdgpu_ctx *ctx,
 			ctx->init_priority : ctx->override_priority;
 	for (i = 0; i < AMDGPU_HW_IP_NUM; ++i) {
 		for (j = 0; j < amdgpu_ctx_num_entities[i]; ++j) {
-			struct drm_sched_entity *entity;
-
 			if (!ctx->entities[i][j])
 				continue;
 
-			entity = &ctx->entities[i][j]->entity;
-			drm_sched_entity_set_priority(entity, ctx_prio);
+			amdgpu_ctx_set_entity_priority(ctx, ctx->entities[i][j],
+						       i, ctx_prio);
 		}
 	}
 }
@@ -628,20 +666,53 @@ void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr)
 	mutex_destroy(&mgr->lock);
 }
 
+
+static void amdgpu_ctx_init_compute_sched(struct amdgpu_device *adev)
+{
+	int num_compute_sched_normal = 0;
+	int num_compute_sched_high = AMDGPU_MAX_COMPUTE_RINGS - 1;
+	int i;
+
+	/* use one drm sched array, gfx.compute_sched to store both high and
+	 * normal priority drm compute schedulers */
+	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+		if (!adev->gfx.compute_ring[i].has_high_prio)
+			adev->gfx.compute_sched[num_compute_sched_normal++] =
+				&adev->gfx.compute_ring[i].sched;
+		else
+			adev->gfx.compute_sched[num_compute_sched_high--] =
+				&adev->gfx.compute_ring[i].sched;
+	}
+
+	/* compute ring only has two priority for now */
+	i = AMDGPU_GFX_PIPE_PRIO_NORMAL;
+	adev->gfx.compute_prio_sched[i] = &adev->gfx.compute_sched[0];
+	adev->gfx.num_compute_sched[i] = num_compute_sched_normal;
+
+	i = AMDGPU_GFX_PIPE_PRIO_HIGH;
+	if (num_compute_sched_high == (AMDGPU_MAX_COMPUTE_RINGS - 1)) {
+		/* When compute has no high priority rings then use */
+		/* normal priority sched array */
+		adev->gfx.compute_prio_sched[i] = &adev->gfx.compute_sched[0];
+		adev->gfx.num_compute_sched[i] = num_compute_sched_normal;
+	} else {
+		adev->gfx.compute_prio_sched[i] =
+			&adev->gfx.compute_sched[num_compute_sched_high - 1];
+		adev->gfx.num_compute_sched[i] =
+			adev->gfx.num_compute_rings - num_compute_sched_normal;
+	}
+}
+
 void amdgpu_ctx_init_sched(struct amdgpu_device *adev)
 {
 	int i, j;
 
+	amdgpu_ctx_init_compute_sched(adev);
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
 		adev->gfx.gfx_sched[i] = &adev->gfx.gfx_ring[i].sched;
 		adev->gfx.num_gfx_sched++;
 	}
 
-	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
-		adev->gfx.compute_sched[i] = &adev->gfx.compute_ring[i].sched;
-		adev->gfx.num_compute_sched++;
-	}
-
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		adev->sdma.sdma_sched[i] = &adev->sdma.instance[i].ring.sched;
 		adev->sdma.num_sdma_sched++;