Merge branch 'drm-fixes-5.1' of git://people.freedesktop.org/~agd5f/linux into drm-fixes

- GPUVM fixes for vega/RV and shadow buffers
- TTM fixes for hugepages
- TTM fix for refcount imbalance in error path
- DC AUX fix for some active DP-DVI dongles
- DC fix for multihead VT switch regression

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexdeucher@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190415051703.3377-1-alexander.deucher@amd.com

8 files changed, 43 insertions, 13 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index 5d8b30fd4534..79fb302fb954 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -3165,6 +3165,7 @@ static int amdgpu_device_recover_vram(struct amdgpu_device *adev)
 
 		/* No need to recover an evicted BO */
 		if (shadow->tbo.mem.mem_type != TTM_PL_TT ||
+		    shadow->tbo.mem.start == AMDGPU_BO_INVALID_OFFSET ||
 		    shadow->parent->tbo.mem.mem_type != TTM_PL_VRAM)
 			continue;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c
index d0d966d6080a..1696644ec022 100644
--- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c
@@ -182,6 +182,7 @@ static void mmhub_v1_0_init_cache_regs(struct amdgpu_device *adev)
 		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3,
 				    L2_CACHE_BIGK_FRAGMENT_SIZE, 6);
 	}
+	WREG32_SOC15(MMHUB, 0, mmVM_L2_CNTL3, tmp);
 
 	tmp = mmVM_L2_CNTL4_DEFAULT;
 	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 0);
diff --git a/drivers/gpu/drm/amd/display/dc/core/dc.c b/drivers/gpu/drm/amd/display/dc/core/dc.c
index c68fbd55db3c..a6cda201c964 100644
--- a/drivers/gpu/drm/amd/display/dc/core/dc.c
+++ b/drivers/gpu/drm/amd/display/dc/core/dc.c
@@ -1377,6 +1377,11 @@ static enum surface_update_type det_surface_update(const struct dc *dc,
 		return UPDATE_TYPE_FULL;
 	}
 
+	if (u->surface->force_full_update) {
+		update_flags->bits.full_update = 1;
+		return UPDATE_TYPE_FULL;
+	}
+
 	type = get_plane_info_update_type(u);
 	elevate_update_type(&overall_type, type);
 
@@ -1802,6 +1807,14 @@ void dc_commit_updates_for_stream(struct dc *dc,
 		}
 
 		dc_resource_state_copy_construct(state, context);
+
+		for (i = 0; i < dc->res_pool->pipe_count; i++) {
+			struct pipe_ctx *new_pipe = &context->res_ctx.pipe_ctx[i];
+			struct pipe_ctx *old_pipe = &dc->current_state->res_ctx.pipe_ctx[i];
+
+			if (new_pipe->plane_state && new_pipe->plane_state != old_pipe->plane_state)
+				new_pipe->plane_state->force_full_update = true;
+		}
 	}
 
 
@@ -1838,6 +1851,12 @@ void dc_commit_updates_for_stream(struct dc *dc,
 		dc->current_state = context;
 		dc_release_state(old);
 
+		for (i = 0; i < dc->res_pool->pipe_count; i++) {
+			struct pipe_ctx *pipe_ctx = &context->res_ctx.pipe_ctx[i];
+
+			if (pipe_ctx->plane_state && pipe_ctx->stream == stream)
+				pipe_ctx->plane_state->force_full_update = false;
+		}
 	}
 	/*let's use current_state to update watermark etc*/
 	if (update_type >= UPDATE_TYPE_FULL)
diff --git a/drivers/gpu/drm/amd/display/dc/dc.h b/drivers/gpu/drm/amd/display/dc/dc.h
index 1a7fd6aa77eb..0515095574e7 100644
--- a/drivers/gpu/drm/amd/display/dc/dc.h
+++ b/drivers/gpu/drm/amd/display/dc/dc.h
@@ -503,6 +503,9 @@ struct dc_plane_state {
 	struct dc_plane_status status;
 	struct dc_context *ctx;
 
+	/* HACK: Workaround for forcing full reprogramming under some conditions */
+	bool force_full_update;
+
 	/* private to dc_surface.c */
 	enum dc_irq_source irq_source;
 	struct kref refcount;
diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_aux.c b/drivers/gpu/drm/amd/display/dc/dce/dce_aux.c
index 4febf4ef7240..4fe3664fb495 100644
--- a/drivers/gpu/drm/amd/display/dc/dce/dce_aux.c
+++ b/drivers/gpu/drm/amd/display/dc/dce/dce_aux.c
@@ -190,6 +190,12 @@ static void submit_channel_request(
 				1,
 				0);
 	}
+
+	REG_UPDATE(AUX_INTERRUPT_CONTROL, AUX_SW_DONE_ACK, 1);
+
+	REG_WAIT(AUX_SW_STATUS, AUX_SW_DONE, 0,
+				10, aux110->timeout_period/10);
+
 	/* set the delay and the number of bytes to write */
 
 	/* The length include
@@ -242,9 +248,6 @@ static void submit_channel_request(
 		}
 	}
 
-	REG_UPDATE(AUX_INTERRUPT_CONTROL, AUX_SW_DONE_ACK, 1);
-	REG_WAIT(AUX_SW_STATUS, AUX_SW_DONE, 0,
-				10, aux110->timeout_period/10);
 	REG_UPDATE(AUX_SW_CONTROL, AUX_SW_GO, 1);
 }
 
diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_aux.h b/drivers/gpu/drm/amd/display/dc/dce/dce_aux.h
index d27f22c05e4b..e28ed6a00ff4 100644
--- a/drivers/gpu/drm/amd/display/dc/dce/dce_aux.h
+++ b/drivers/gpu/drm/amd/display/dc/dce/dce_aux.h
@@ -71,11 +71,11 @@ enum {	/* This is the timeout as defined in DP 1.2a,
 	 * at most within ~240usec. That means,
 	 * increasing this timeout will not affect normal operation,
 	 * and we'll timeout after
-	 * SW_AUX_TIMEOUT_PERIOD_MULTIPLIER * AUX_TIMEOUT_PERIOD = 1600usec.
+	 * SW_AUX_TIMEOUT_PERIOD_MULTIPLIER * AUX_TIMEOUT_PERIOD = 2400usec.
 	 * This timeout is especially important for
-	 * resume from S3 and CTS.
+	 * converters, resume from S3, and CTS.
 	 */
-	SW_AUX_TIMEOUT_PERIOD_MULTIPLIER = 4
+	SW_AUX_TIMEOUT_PERIOD_MULTIPLIER = 6
 };
 
 struct dce_aux {
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
index 3f56647cdb35..0fa5034b9f9e 100644
--- a/drivers/gpu/drm/ttm/ttm_bo.c
+++ b/drivers/gpu/drm/ttm/ttm_bo.c
@@ -876,8 +876,10 @@ static int ttm_bo_add_move_fence(struct ttm_buffer_object *bo,
 		reservation_object_add_shared_fence(bo->resv, fence);
 
 		ret = reservation_object_reserve_shared(bo->resv, 1);
-		if (unlikely(ret))
+		if (unlikely(ret)) {
+			dma_fence_put(fence);
 			return ret;
+		}
 
 		dma_fence_put(bo->moving);
 		bo->moving = fence;
diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc.c b/drivers/gpu/drm/ttm/ttm_page_alloc.c
index f841accc2c00..627f8dc91d0e 100644
--- a/drivers/gpu/drm/ttm/ttm_page_alloc.c
+++ b/drivers/gpu/drm/ttm/ttm_page_alloc.c
@@ -730,9 +730,10 @@ static void ttm_put_pages(struct page **pages, unsigned npages, int flags,
 			}
 
 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
-			if (!(flags & TTM_PAGE_FLAG_DMA32)) {
-				for (j = 0; j < HPAGE_PMD_NR; ++j)
-					if (p++ != pages[i + j])
+			if (!(flags & TTM_PAGE_FLAG_DMA32) &&
+			    (npages - i) >= HPAGE_PMD_NR) {
+				for (j = 1; j < HPAGE_PMD_NR; ++j)
+					if (++p != pages[i + j])
 					    break;
 
 				if (j == HPAGE_PMD_NR)
@@ -759,15 +760,15 @@ static void ttm_put_pages(struct page **pages, unsigned npages, int flags,
 		unsigned max_size, n2free;
 
 		spin_lock_irqsave(&huge->lock, irq_flags);
-		while (i < npages) {
+		while ((npages - i) >= HPAGE_PMD_NR) {
 			struct page *p = pages[i];
 			unsigned j;
 
 			if (!p)
 				break;
 
-			for (j = 0; j < HPAGE_PMD_NR; ++j)
-				if (p++ != pages[i + j])
+			for (j = 1; j < HPAGE_PMD_NR; ++j)
+				if (++p != pages[i + j])
 				    break;
 
 			if (j != HPAGE_PMD_NR)