Merge branch 'vmwgfx-next' of git://people.freedesktop.org/~thomash/l…

…inux into drm-next

Changes for vmwgfx for 4.4. If there is time, I'll follow up with a series
to move to threaded irqs.

* 'vmwgfx-next' of git://people.freedesktop.org/~thomash/linux:
  drm/vmwgfx: Replace iowrite/ioread with volatile memory accesses
  drm/vmwgfx: Turn off support for multisample count != 0 v2
  drm/vmwgfx: switch from ioremap_cache to memremap

drivers/gpu/drm/vmwgfx/vmwgfx_drv.c

@@ -752,8 +752,8 @@ static int vmw_driver_load(struct drm_device *dev, unsigned long chipset)
 	ttm_lock_set_kill(&dev_priv->fbdev_master.lock, false, SIGTERM);
 	dev_priv->active_master = &dev_priv->fbdev_master;
 
-	dev_priv->mmio_virt = ioremap_cache(dev_priv->mmio_start,
-					    dev_priv->mmio_size);
+	dev_priv->mmio_virt = memremap(dev_priv->mmio_start,
+				       dev_priv->mmio_size, MEMREMAP_WB);
 
 	if (unlikely(dev_priv->mmio_virt == NULL)) {
 		ret = -ENOMEM;
@@ -907,7 +907,7 @@ static int vmw_driver_load(struct drm_device *dev, unsigned long chipset)
 out_no_device:
 	ttm_object_device_release(&dev_priv->tdev);
 out_err4:
-	iounmap(dev_priv->mmio_virt);
+	memunmap(dev_priv->mmio_virt);
 out_err3:
 	vmw_ttm_global_release(dev_priv);
 out_err0:
@@ -958,7 +958,7 @@ static int vmw_driver_unload(struct drm_device *dev)
 		pci_release_regions(dev->pdev);
 
 	ttm_object_device_release(&dev_priv->tdev);
-	iounmap(dev_priv->mmio_virt);
+	memunmap(dev_priv->mmio_virt);
 	if (dev_priv->ctx.staged_bindings)
 		vmw_binding_state_free(dev_priv->ctx.staged_bindings);
 	vmw_ttm_global_release(dev_priv);

drivers/gpu/drm/vmwgfx/vmwgfx_drv.h

@@ -375,7 +375,7 @@ struct vmw_private {
 	uint32_t stdu_max_height;
 	uint32_t initial_width;
 	uint32_t initial_height;
-	u32 __iomem *mmio_virt;
+	u32 *mmio_virt;
 	uint32_t capabilities;
 	uint32_t max_gmr_ids;
 	uint32_t max_gmr_pages;
@@ -1206,4 +1206,30 @@ static inline void vmw_fifo_resource_dec(struct vmw_private *dev_priv)
 {
 	atomic_dec(&dev_priv->num_fifo_resources);
 }
+
+/**
+ * vmw_mmio_read - Perform a MMIO read from volatile memory
+ *
+ * @addr: The address to read from
+ *
+ * This function is intended to be equivalent to ioread32() on
+ * memremap'd memory, but without byteswapping.
+ */
+static inline u32 vmw_mmio_read(u32 *addr)
+{
+	return READ_ONCE(*addr);
+}
+
+/**
+ * vmw_mmio_write - Perform a MMIO write to volatile memory
+ *
+ * @addr: The address to write to
+ *
+ * This function is intended to be equivalent to iowrite32 on
+ * memremap'd memory, but without byteswapping.
+ */
+static inline void vmw_mmio_write(u32 value, u32 *addr)
+{
+	WRITE_ONCE(*addr, value);
+}
 #endif

drivers/gpu/drm/vmwgfx/vmwgfx_fence.c

@@ -142,8 +142,8 @@ static bool vmw_fence_enable_signaling(struct fence *f)
 	struct vmw_fence_manager *fman = fman_from_fence(fence);
 	struct vmw_private *dev_priv = fman->dev_priv;
 
-	u32 __iomem *fifo_mem = dev_priv->mmio_virt;
-	u32 seqno = ioread32(fifo_mem + SVGA_FIFO_FENCE);
+	u32 *fifo_mem = dev_priv->mmio_virt;
+	u32 seqno = vmw_mmio_read(fifo_mem + SVGA_FIFO_FENCE);
 	if (seqno - fence->base.seqno < VMW_FENCE_WRAP)
 		return false;
 
@@ -386,23 +386,23 @@ static bool vmw_fence_goal_new_locked(struct vmw_fence_manager *fman,
 				      u32 passed_seqno)
 {
 	u32 goal_seqno;
-	u32 __iomem *fifo_mem;
+	u32 *fifo_mem;
 	struct vmw_fence_obj *fence;
 
 	if (likely(!fman->seqno_valid))
 		return false;
 
 	fifo_mem = fman->dev_priv->mmio_virt;
-	goal_seqno = ioread32(fifo_mem + SVGA_FIFO_FENCE_GOAL);
+	goal_seqno = vmw_mmio_read(fifo_mem + SVGA_FIFO_FENCE_GOAL);
 	if (likely(passed_seqno - goal_seqno >= VMW_FENCE_WRAP))
 		return false;
 
 	fman->seqno_valid = false;
 	list_for_each_entry(fence, &fman->fence_list, head) {
 		if (!list_empty(&fence->seq_passed_actions)) {
 			fman->seqno_valid = true;
-			iowrite32(fence->base.seqno,
-				  fifo_mem + SVGA_FIFO_FENCE_GOAL);
+			vmw_mmio_write(fence->base.seqno,
+				       fifo_mem + SVGA_FIFO_FENCE_GOAL);
 			break;
 		}
 	}
@@ -430,18 +430,18 @@ static bool vmw_fence_goal_check_locked(struct vmw_fence_obj *fence)
 {
 	struct vmw_fence_manager *fman = fman_from_fence(fence);
 	u32 goal_seqno;
-	u32 __iomem *fifo_mem;
+	u32 *fifo_mem;
 
 	if (fence_is_signaled_locked(&fence->base))
 		return false;
 
 	fifo_mem = fman->dev_priv->mmio_virt;
-	goal_seqno = ioread32(fifo_mem + SVGA_FIFO_FENCE_GOAL);
+	goal_seqno = vmw_mmio_read(fifo_mem + SVGA_FIFO_FENCE_GOAL);
 	if (likely(fman->seqno_valid &&
 		   goal_seqno - fence->base.seqno < VMW_FENCE_WRAP))
 		return false;
 
-	iowrite32(fence->base.seqno, fifo_mem + SVGA_FIFO_FENCE_GOAL);
+	vmw_mmio_write(fence->base.seqno, fifo_mem + SVGA_FIFO_FENCE_GOAL);
 	fman->seqno_valid = true;
 
 	return true;
@@ -453,9 +453,9 @@ static void __vmw_fences_update(struct vmw_fence_manager *fman)
 	struct list_head action_list;
 	bool needs_rerun;
 	uint32_t seqno, new_seqno;
-	u32 __iomem *fifo_mem = fman->dev_priv->mmio_virt;
+	u32 *fifo_mem = fman->dev_priv->mmio_virt;
 
-	seqno = ioread32(fifo_mem + SVGA_FIFO_FENCE);
+	seqno = vmw_mmio_read(fifo_mem + SVGA_FIFO_FENCE);
 rerun:
 	list_for_each_entry_safe(fence, next_fence, &fman->fence_list, head) {
 		if (seqno - fence->base.seqno < VMW_FENCE_WRAP) {
@@ -477,7 +477,7 @@ static void __vmw_fences_update(struct vmw_fence_manager *fman)
 
 	needs_rerun = vmw_fence_goal_new_locked(fman, seqno);
 	if (unlikely(needs_rerun)) {
-		new_seqno = ioread32(fifo_mem + SVGA_FIFO_FENCE);
+		new_seqno = vmw_mmio_read(fifo_mem + SVGA_FIFO_FENCE);
 		if (new_seqno != seqno) {
 			seqno = new_seqno;
 			goto rerun;