6 files changed, 50 insertions, 17 deletions

diff --git a/drivers/gpu/drm/radeon/cik.c b/drivers/gpu/drm/radeon/cik.c
index e22be8458d92..bbb17841a9e5 100644
--- a/drivers/gpu/drm/radeon/cik.c
+++ b/drivers/gpu/drm/radeon/cik.c
@@ -4134,8 +4134,11 @@ static void cik_cp_compute_enable(struct radeon_device *rdev, bool enable)
 {
 	if (enable)
 		WREG32(CP_MEC_CNTL, 0);
-	else
+	else {
 		WREG32(CP_MEC_CNTL, (MEC_ME1_HALT | MEC_ME2_HALT));
+		rdev->ring[CAYMAN_RING_TYPE_CP1_INDEX].ready = false;
+		rdev->ring[CAYMAN_RING_TYPE_CP2_INDEX].ready = false;
+	}
 	udelay(50);
 }
 
diff --git a/drivers/gpu/drm/radeon/cik_sdma.c b/drivers/gpu/drm/radeon/cik_sdma.c
index 1ecb3f1070e3..94626ea90fa5 100644
--- a/drivers/gpu/drm/radeon/cik_sdma.c
+++ b/drivers/gpu/drm/radeon/cik_sdma.c
@@ -264,6 +264,8 @@ static void cik_sdma_gfx_stop(struct radeon_device *rdev)
 		WREG32(SDMA0_GFX_RB_CNTL + reg_offset, rb_cntl);
 		WREG32(SDMA0_GFX_IB_CNTL + reg_offset, 0);
 	}
+	rdev->ring[R600_RING_TYPE_DMA_INDEX].ready = false;
+	rdev->ring[CAYMAN_RING_TYPE_DMA1_INDEX].ready = false;
 }
 
 /**
@@ -291,6 +293,11 @@ void cik_sdma_enable(struct radeon_device *rdev, bool enable)
 	u32 me_cntl, reg_offset;
 	int i;
 
+	if (enable == false) {
+		cik_sdma_gfx_stop(rdev);
+		cik_sdma_rlc_stop(rdev);
+	}
+
 	for (i = 0; i < 2; i++) {
 		if (i == 0)
 			reg_offset = SDMA0_REGISTER_OFFSET;
@@ -420,10 +427,6 @@ static int cik_sdma_load_microcode(struct radeon_device *rdev)
 	if (!rdev->sdma_fw)
 		return -EINVAL;
 
-	/* stop the gfx rings and rlc compute queues */
-	cik_sdma_gfx_stop(rdev);
-	cik_sdma_rlc_stop(rdev);
-
 	/* halt the MEs */
 	cik_sdma_enable(rdev, false);
 
@@ -492,9 +495,6 @@ int cik_sdma_resume(struct radeon_device *rdev)
  */
 void cik_sdma_fini(struct radeon_device *rdev)
 {
-	/* stop the gfx rings and rlc compute queues */
-	cik_sdma_gfx_stop(rdev);
-	cik_sdma_rlc_stop(rdev);
 	/* halt the MEs */
 	cik_sdma_enable(rdev, false);
 	radeon_ring_fini(rdev, &rdev->ring[R600_RING_TYPE_DMA_INDEX]);
diff --git a/drivers/gpu/drm/radeon/radeon_kms.c b/drivers/gpu/drm/radeon/radeon_kms.c
index 2aecd6dc2610..66ed3ea71440 100644
--- a/drivers/gpu/drm/radeon/radeon_kms.c
+++ b/drivers/gpu/drm/radeon/radeon_kms.c
@@ -33,6 +33,13 @@
 #include <linux/vga_switcheroo.h>
 #include <linux/slab.h>
 #include <linux/pm_runtime.h>
+
+#if defined(CONFIG_VGA_SWITCHEROO)
+bool radeon_is_px(void);
+#else
+static inline bool radeon_is_px(void) { return false; }
+#endif
+
 /**
  * radeon_driver_unload_kms - Main unload function for KMS.
  *
@@ -130,7 +137,8 @@ int radeon_driver_load_kms(struct drm_device *dev, unsigned long flags)
 				"Error during ACPI methods call\n");
 	}
 
-	if (radeon_runtime_pm != 0) {
+	if ((radeon_runtime_pm == 1) ||
+	    ((radeon_runtime_pm == -1) && radeon_is_px())) {
 		pm_runtime_use_autosuspend(dev->dev);
 		pm_runtime_set_autosuspend_delay(dev->dev, 5000);
 		pm_runtime_set_active(dev->dev);
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
index a06651309388..214b7992a3aa 100644
--- a/drivers/gpu/drm/ttm/ttm_bo.c
+++ b/drivers/gpu/drm/ttm/ttm_bo.c
@@ -351,9 +351,11 @@ static int ttm_bo_handle_move_mem(struct ttm_buffer_object *bo,
 
 moved:
 	if (bo->evicted) {
-		ret = bdev->driver->invalidate_caches(bdev, bo->mem.placement);
-		if (ret)
-			pr_err("Can not flush read caches\n");
+		if (bdev->driver->invalidate_caches) {
+			ret = bdev->driver->invalidate_caches(bdev, bo->mem.placement);
+			if (ret)
+				pr_err("Can not flush read caches\n");
+		}
 		bo->evicted = false;
 	}
 
diff --git a/drivers/gpu/drm/ttm/ttm_bo_vm.c b/drivers/gpu/drm/ttm/ttm_bo_vm.c
index 801231c9ae48..0ce48e5a9cb4 100644
--- a/drivers/gpu/drm/ttm/ttm_bo_vm.c
+++ b/drivers/gpu/drm/ttm/ttm_bo_vm.c
@@ -339,11 +339,13 @@ int ttm_bo_mmap(struct file *filp, struct vm_area_struct *vma,
 	vma->vm_private_data = bo;
 
 	/*
-	 * PFNMAP is faster than MIXEDMAP due to reduced page
-	 * administration. So use MIXEDMAP only if private VMA, where
-	 * we need to support COW.
+	 * We'd like to use VM_PFNMAP on shared mappings, where
+	 * (vma->vm_flags & VM_SHARED) != 0, for performance reasons,
+	 * but for some reason VM_PFNMAP + x86 PAT + write-combine is very
+	 * bad for performance. Until that has been sorted out, use
+	 * VM_MIXEDMAP on all mappings. See freedesktop.org bug #75719
 	 */
-	vma->vm_flags |= (vma->vm_flags & VM_SHARED) ? VM_PFNMAP : VM_MIXEDMAP;
+	vma->vm_flags |= VM_MIXEDMAP;
 	vma->vm_flags |= VM_IO | VM_DONTEXPAND | VM_DONTDUMP;
 	return 0;
 out_unref:
@@ -359,7 +361,7 @@ int ttm_fbdev_mmap(struct vm_area_struct *vma, struct ttm_buffer_object *bo)
 
 	vma->vm_ops = &ttm_bo_vm_ops;
 	vma->vm_private_data = ttm_bo_reference(bo);
-	vma->vm_flags |= (vma->vm_flags & VM_SHARED) ? VM_PFNMAP : VM_MIXEDMAP;
+	vma->vm_flags |= VM_MIXEDMAP;
 	vma->vm_flags |= VM_IO | VM_DONTEXPAND;
 	return 0;
 }
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_surface.c b/drivers/gpu/drm/vmwgfx/vmwgfx_surface.c
index 82468d902915..e7af580ab977 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_surface.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_surface.c
@@ -830,6 +830,24 @@ int vmw_surface_define_ioctl(struct drm_device *dev, void *data,
 	if (unlikely(ret != 0))
 		goto out_unlock;
 
+	/*
+	 * A gb-aware client referencing a shared surface will
+	 * expect a backup buffer to be present.
+	 */
+	if (dev_priv->has_mob && req->shareable) {
+		uint32_t backup_handle;
+
+		ret = vmw_user_dmabuf_alloc(dev_priv, tfile,
+					    res->backup_size,
+					    true,
+					    &backup_handle,
+					    &res->backup);
+		if (unlikely(ret != 0)) {
+			vmw_resource_unreference(&res);
+			goto out_unlock;
+		}
+	}
+
 	tmp = vmw_resource_reference(&srf->res);
 	ret = ttm_prime_object_init(tfile, res->backup_size, &user_srf->prime,
 				    req->shareable, VMW_RES_SURFACE,