Merge branch 'topic/snd-device' into for-next

34 files changed, 364 insertions, 214 deletions

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
index 633532a2e7ec..25bc47f3c1cf 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
@@ -26,6 +26,7 @@
 #include <linux/slab.h>
 #include "kfd_priv.h"
 #include "kfd_device_queue_manager.h"
+#include "kfd_pm4_headers.h"
 
 #define MQD_SIZE_ALIGNED 768
 
@@ -169,9 +170,8 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
 	kfd->shared_resources = *gpu_resources;
 
 	/* calculate max size of mqds needed for queues */
-	size = max_num_of_processes *
-		max_num_of_queues_per_process *
-		kfd->device_info->mqd_size_aligned;
+	size = max_num_of_queues_per_device *
+			kfd->device_info->mqd_size_aligned;
 
 	/* add another 512KB for all other allocations on gart */
 	size += 512 * 1024;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
index 30c8fda9622e..0d8694f015c1 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
@@ -183,6 +183,13 @@ static int create_queue_nocpsch(struct device_queue_manager *dqm,
 
 	mutex_lock(&dqm->lock);
 
+	if (dqm->total_queue_count >= max_num_of_queues_per_device) {
+		pr_warn("amdkfd: Can't create new usermode queue because %d queues were already created\n",
+				dqm->total_queue_count);
+		mutex_unlock(&dqm->lock);
+		return -EPERM;
+	}
+
 	if (list_empty(&qpd->queues_list)) {
 		retval = allocate_vmid(dqm, qpd, q);
 		if (retval != 0) {
@@ -207,6 +214,14 @@ static int create_queue_nocpsch(struct device_queue_manager *dqm,
 	list_add(&q->list, &qpd->queues_list);
 	dqm->queue_count++;
 
+	/*
+	 * Unconditionally increment this counter, regardless of the queue's
+	 * type or whether the queue is active.
+	 */
+	dqm->total_queue_count++;
+	pr_debug("Total of %d queues are accountable so far\n",
+			dqm->total_queue_count);
+
 	mutex_unlock(&dqm->lock);
 	return 0;
 }
@@ -326,6 +341,15 @@ static int destroy_queue_nocpsch(struct device_queue_manager *dqm,
 	if (list_empty(&qpd->queues_list))
 		deallocate_vmid(dqm, qpd, q);
 	dqm->queue_count--;
+
+	/*
+	 * Unconditionally decrement this counter, regardless of the queue's
+	 * type
+	 */
+	dqm->total_queue_count--;
+	pr_debug("Total of %d queues are accountable so far\n",
+			dqm->total_queue_count);
+
 out:
 	mutex_unlock(&dqm->lock);
 	return retval;
@@ -541,10 +565,14 @@ static int init_pipelines(struct device_queue_manager *dqm,
 
 	for (i = 0; i < pipes_num; i++) {
 		inx = i + first_pipe;
+		/*
+		 * HPD buffer on GTT is allocated by amdkfd, no need to waste
+		 * space in GTT for pipelines we don't initialize
+		 */
 		pipe_hpd_addr = dqm->pipelines_addr + i * CIK_HPD_EOP_BYTES;
 		pr_debug("kfd: pipeline address %llX\n", pipe_hpd_addr);
 		/* = log2(bytes/4)-1 */
-		kfd2kgd->init_pipeline(dqm->dev->kgd, i,
+		kfd2kgd->init_pipeline(dqm->dev->kgd, inx,
 				CIK_HPD_EOP_BYTES_LOG2 - 3, pipe_hpd_addr);
 	}
 
@@ -560,7 +588,7 @@ static int init_scheduler(struct device_queue_manager *dqm)
 
 	pr_debug("kfd: In %s\n", __func__);
 
-	retval = init_pipelines(dqm, get_pipes_num(dqm), KFD_DQM_FIRST_PIPE);
+	retval = init_pipelines(dqm, get_pipes_num(dqm), get_first_pipe(dqm));
 	if (retval != 0)
 		return retval;
 
@@ -752,6 +780,21 @@ static int create_kernel_queue_cpsch(struct device_queue_manager *dqm,
 	pr_debug("kfd: In func %s\n", __func__);
 
 	mutex_lock(&dqm->lock);
+	if (dqm->total_queue_count >= max_num_of_queues_per_device) {
+		pr_warn("amdkfd: Can't create new kernel queue because %d queues were already created\n",
+				dqm->total_queue_count);
+		mutex_unlock(&dqm->lock);
+		return -EPERM;
+	}
+
+	/*
+	 * Unconditionally increment this counter, regardless of the queue's
+	 * type or whether the queue is active.
+	 */
+	dqm->total_queue_count++;
+	pr_debug("Total of %d queues are accountable so far\n",
+			dqm->total_queue_count);
+
 	list_add(&kq->list, &qpd->priv_queue_list);
 	dqm->queue_count++;
 	qpd->is_debug = true;
@@ -775,6 +818,13 @@ static void destroy_kernel_queue_cpsch(struct device_queue_manager *dqm,
 	dqm->queue_count--;
 	qpd->is_debug = false;
 	execute_queues_cpsch(dqm, false);
+	/*
+	 * Unconditionally decrement this counter, regardless of the queue's
+	 * type.
+	 */
+	dqm->total_queue_count++;
+	pr_debug("Total of %d queues are accountable so far\n",
+			dqm->total_queue_count);
 	mutex_unlock(&dqm->lock);
 }
 
@@ -793,6 +843,13 @@ static int create_queue_cpsch(struct device_queue_manager *dqm, struct queue *q,
 
 	mutex_lock(&dqm->lock);
 
+	if (dqm->total_queue_count >= max_num_of_queues_per_device) {
+		pr_warn("amdkfd: Can't create new usermode queue because %d queues were already created\n",
+				dqm->total_queue_count);
+		retval = -EPERM;
+		goto out;
+	}
+
 	mqd = dqm->get_mqd_manager(dqm, KFD_MQD_TYPE_CIK_CP);
 	if (mqd == NULL) {
 		mutex_unlock(&dqm->lock);
@@ -810,6 +867,15 @@ static int create_queue_cpsch(struct device_queue_manager *dqm, struct queue *q,
 		retval = execute_queues_cpsch(dqm, false);
 	}
 
+	/*
+	 * Unconditionally increment this counter, regardless of the queue's
+	 * type or whether the queue is active.
+	 */
+	dqm->total_queue_count++;
+
+	pr_debug("Total of %d queues are accountable so far\n",
+			dqm->total_queue_count);
+
 out:
 	mutex_unlock(&dqm->lock);
 	return retval;
@@ -930,6 +996,14 @@ static int destroy_queue_cpsch(struct device_queue_manager *dqm,
 
 	mqd->uninit_mqd(mqd, q->mqd, q->mqd_mem_obj);
 
+	/*
+	 * Unconditionally decrement this counter, regardless of the queue's
+	 * type
+	 */
+	dqm->total_queue_count--;
+	pr_debug("Total of %d queues are accountable so far\n",
+			dqm->total_queue_count);
+
 	mutex_unlock(&dqm->lock);
 
 	return 0;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h
index c3f189e8ae35..52035bf0c1cb 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h
@@ -130,6 +130,7 @@ struct device_queue_manager {
 	struct list_head	queues;
 	unsigned int		processes_count;
 	unsigned int		queue_count;
+	unsigned int		total_queue_count;
 	unsigned int		next_pipe_to_allocate;
 	unsigned int		*allocated_queues;
 	unsigned int		vmid_bitmap;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_module.c b/drivers/gpu/drm/amd/amdkfd/kfd_module.c
index 95d5af138e6e..a8be6df85347 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_module.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_module.c
@@ -50,15 +50,10 @@ module_param(sched_policy, int, 0444);
 MODULE_PARM_DESC(sched_policy,
 	"Kernel cmdline parameter that defines the amdkfd scheduling policy");
 
-int max_num_of_processes = KFD_MAX_NUM_OF_PROCESSES_DEFAULT;
-module_param(max_num_of_processes, int, 0444);
-MODULE_PARM_DESC(max_num_of_processes,
-	"Kernel cmdline parameter that defines the amdkfd maximum number of supported processes");
-
-int max_num_of_queues_per_process = KFD_MAX_NUM_OF_QUEUES_PER_PROCESS_DEFAULT;
-module_param(max_num_of_queues_per_process, int, 0444);
-MODULE_PARM_DESC(max_num_of_queues_per_process,
-	"Kernel cmdline parameter that defines the amdkfd maximum number of supported queues per process");
+int max_num_of_queues_per_device = KFD_MAX_NUM_OF_QUEUES_PER_DEVICE_DEFAULT;
+module_param(max_num_of_queues_per_device, int, 0444);
+MODULE_PARM_DESC(max_num_of_queues_per_device,
+	"Maximum number of supported queues per device (1 = Minimum, 4096 = default)");
 
 bool kgd2kfd_init(unsigned interface_version,
 		  const struct kfd2kgd_calls *f2g,
@@ -100,16 +95,10 @@ static int __init kfd_module_init(void)
 	}
 
 	/* Verify module parameters */
-	if ((max_num_of_processes < 0) ||
-		(max_num_of_processes > KFD_MAX_NUM_OF_PROCESSES)) {
-		pr_err("kfd: max_num_of_processes must be between 0 to KFD_MAX_NUM_OF_PROCESSES\n");
-		return -1;
-	}
-
-	if ((max_num_of_queues_per_process < 0) ||
-		(max_num_of_queues_per_process >
-			KFD_MAX_NUM_OF_QUEUES_PER_PROCESS)) {
-		pr_err("kfd: max_num_of_queues_per_process must be between 0 to KFD_MAX_NUM_OF_QUEUES_PER_PROCESS\n");
+	if ((max_num_of_queues_per_device < 0) ||
+		(max_num_of_queues_per_device >
+			KFD_MAX_NUM_OF_QUEUES_PER_DEVICE)) {
+		pr_err("kfd: max_num_of_queues_per_device must be between 0 to KFD_MAX_NUM_OF_QUEUES_PER_DEVICE\n");
 		return -1;
 	}
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_pasid.c b/drivers/gpu/drm/amd/amdkfd/kfd_pasid.c
index 4c25ef504f79..6cfe7f1f18cf 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_pasid.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_pasid.c
@@ -30,7 +30,7 @@ static DEFINE_MUTEX(pasid_mutex);
 
 int kfd_pasid_init(void)
 {
-	pasid_limit = max_num_of_processes;
+	pasid_limit = KFD_MAX_NUM_OF_PROCESSES;
 
 	pasid_bitmap = kcalloc(BITS_TO_LONGS(pasid_limit), sizeof(long), GFP_KERNEL);
 	if (!pasid_bitmap)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index b3dc13c83169..96dc10e8904a 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
@@ -52,20 +52,19 @@
 #define kfd_alloc_struct(ptr_to_struct)	\
 	((typeof(ptr_to_struct)) kzalloc(sizeof(*ptr_to_struct), GFP_KERNEL))
 
-/* Kernel module parameter to specify maximum number of supported processes */
-extern int max_num_of_processes;
-
-#define KFD_MAX_NUM_OF_PROCESSES_DEFAULT 32
 #define KFD_MAX_NUM_OF_PROCESSES 512
+#define KFD_MAX_NUM_OF_QUEUES_PER_PROCESS 1024
 
 /*
- * Kernel module parameter to specify maximum number of supported queues
- * per process
+ * Kernel module parameter to specify maximum number of supported queues per
+ * device
  */
-extern int max_num_of_queues_per_process;
+extern int max_num_of_queues_per_device;
 
-#define KFD_MAX_NUM_OF_QUEUES_PER_PROCESS_DEFAULT 128
-#define KFD_MAX_NUM_OF_QUEUES_PER_PROCESS 1024
+#define KFD_MAX_NUM_OF_QUEUES_PER_DEVICE_DEFAULT 4096
+#define KFD_MAX_NUM_OF_QUEUES_PER_DEVICE		\
+	(KFD_MAX_NUM_OF_PROCESSES *			\
+			KFD_MAX_NUM_OF_QUEUES_PER_PROCESS)
 
 #define KFD_KERNEL_QUEUE_SIZE 2048
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
index 47526780d736..f37cf5efe642 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c
@@ -54,11 +54,11 @@ static int find_available_queue_slot(struct process_queue_manager *pqm,
 	pr_debug("kfd: in %s\n", __func__);
 
 	found = find_first_zero_bit(pqm->queue_slot_bitmap,
-			max_num_of_queues_per_process);
+			KFD_MAX_NUM_OF_QUEUES_PER_PROCESS);
 
 	pr_debug("kfd: the new slot id %lu\n", found);
 
-	if (found >= max_num_of_queues_per_process) {
+	if (found >= KFD_MAX_NUM_OF_QUEUES_PER_PROCESS) {
 		pr_info("amdkfd: Can not open more queues for process with pasid %d\n",
 				pqm->process->pasid);
 		return -ENOMEM;
@@ -76,7 +76,7 @@ int pqm_init(struct process_queue_manager *pqm, struct kfd_process *p)
 
 	INIT_LIST_HEAD(&pqm->queues);
 	pqm->queue_slot_bitmap =
-			kzalloc(DIV_ROUND_UP(max_num_of_queues_per_process,
+			kzalloc(DIV_ROUND_UP(KFD_MAX_NUM_OF_QUEUES_PER_PROCESS,
 					BITS_PER_BYTE), GFP_KERNEL);
 	if (pqm->queue_slot_bitmap == NULL)
 		return -ENOMEM;
@@ -203,6 +203,7 @@ int pqm_create_queue(struct process_queue_manager *pqm,
 		pqn->kq = NULL;
 		retval = dev->dqm->create_queue(dev->dqm, q, &pdd->qpd,
 						&q->properties.vmid);
+		pr_debug("DQM returned %d for create_queue\n", retval);
 		print_queue(q);
 		break;
 	case KFD_QUEUE_TYPE_DIQ:
@@ -222,7 +223,7 @@ int pqm_create_queue(struct process_queue_manager *pqm,
 	}
 
 	if (retval != 0) {
-		pr_err("kfd: error dqm create queue\n");
+		pr_debug("Error dqm create queue\n");
 		goto err_create_queue;
 	}
 
@@ -241,7 +242,10 @@ int pqm_create_queue(struct process_queue_manager *pqm,
 err_create_queue:
 	kfree(pqn);
 err_allocate_pqn:
+	/* check if queues list is empty unregister process from device */
 	clear_bit(*qid, pqm->queue_slot_bitmap);
+	if (list_empty(&pqm->queues))
+		dev->dqm->unregister_process(dev->dqm, &pdd->qpd);
 	return retval;
 }
 
diff --git a/drivers/gpu/drm/drm_fb_helper.c b/drivers/gpu/drm/drm_fb_helper.c
index cf775a4449c1..dc386ebe5193 100644
--- a/drivers/gpu/drm/drm_fb_helper.c
+++ b/drivers/gpu/drm/drm_fb_helper.c
@@ -145,6 +145,31 @@ int drm_fb_helper_add_one_connector(struct drm_fb_helper *fb_helper, struct drm_
 }
 EXPORT_SYMBOL(drm_fb_helper_add_one_connector);
 
+static void remove_from_modeset(struct drm_mode_set *set,
+		struct drm_connector *connector)
+{
+	int i, j;
+
+	for (i = 0; i < set->num_connectors; i++) {
+		if (set->connectors[i] == connector)
+			break;
+	}
+
+	if (i == set->num_connectors)
+		return;
+
+	for (j = i + 1; j < set->num_connectors; j++) {
+		set->connectors[j - 1] = set->connectors[j];
+	}
+	set->num_connectors--;
+
+	/* because i915 is pissy about this..
+	 * TODO maybe need to makes sure we set it back to !=NULL somewhere?
+	 */
+	if (set->num_connectors == 0)
+		set->fb = NULL;
+}
+
 int drm_fb_helper_remove_one_connector(struct drm_fb_helper *fb_helper,
 				       struct drm_connector *connector)
 {
@@ -167,6 +192,11 @@ int drm_fb_helper_remove_one_connector(struct drm_fb_helper *fb_helper,
 	}
 	fb_helper->connector_count--;
 	kfree(fb_helper_connector);
+
+	/* also cleanup dangling references to the connector: */
+	for (i = 0; i < fb_helper->crtc_count; i++)
+		remove_from_modeset(&fb_helper->crtc_info[i].mode_set, connector);
+
 	return 0;
 }
 EXPORT_SYMBOL(drm_fb_helper_remove_one_connector);
diff --git a/drivers/gpu/drm/i2c/tda998x_drv.c b/drivers/gpu/drm/i2c/tda998x_drv.c
index d4762799351d..a9041d1a8ff0 100644
--- a/drivers/gpu/drm/i2c/tda998x_drv.c
+++ b/drivers/gpu/drm/i2c/tda998x_drv.c
@@ -32,6 +32,8 @@
 struct tda998x_priv {
 	struct i2c_client *cec;
 	struct i2c_client *hdmi;
+	struct mutex mutex;
+	struct delayed_work dwork;
 	uint16_t rev;
 	uint8_t current_page;
 	int dpms;
@@ -402,9 +404,10 @@ reg_read_range(struct tda998x_priv *priv, uint16_t reg, char *buf, int cnt)
 	uint8_t addr = REG2ADDR(reg);
 	int ret;
 
+	mutex_lock(&priv->mutex);
 	ret = set_page(priv, reg);
 	if (ret < 0)
-		return ret;
+		goto out;
 
 	ret = i2c_master_send(client, &addr, sizeof(addr));
 	if (ret < 0)
@@ -414,10 +417,12 @@ reg_read_range(struct tda998x_priv *priv, uint16_t reg, char *buf, int cnt)
 	if (ret < 0)
 		goto fail;
 
-	return ret;
+	goto out;
 
 fail:
 	dev_err(&client->dev, "Error %d reading from 0x%x\n", ret, reg);
+out:
+	mutex_unlock(&priv->mutex);
 	return ret;
 }
 
@@ -431,13 +436,16 @@ reg_write_range(struct tda998x_priv *priv, uint16_t reg, uint8_t *p, int cnt)
 	buf[0] = REG2ADDR(reg);
 	memcpy(&buf[1], p, cnt);
 
+	mutex_lock(&priv->mutex);
 	ret = set_page(priv, reg);
 	if (ret < 0)
-		return;
+		goto out;
 
 	ret = i2c_master_send(client, buf, cnt + 1);
 	if (ret < 0)
 		dev_err(&client->dev, "Error %d writing to 0x%x\n", ret, reg);
+out:
+	mutex_unlock(&priv->mutex);
 }
 
 static int
@@ -459,13 +467,16 @@ reg_write(struct tda998x_priv *priv, uint16_t reg, uint8_t val)
 	uint8_t buf[] = {REG2ADDR(reg), val};
 	int ret;
 
+	mutex_lock(&priv->mutex);
 	ret = set_page(priv, reg);
 	if (ret < 0)
-		return;
+		goto out;
 
 	ret = i2c_master_send(client, buf, sizeof(buf));
 	if (ret < 0)
 		dev_err(&client->dev, "Error %d writing to 0x%x\n", ret, reg);
+out:
+	mutex_unlock(&priv->mutex);
 }
 
 static void
@@ -475,13 +486,16 @@ reg_write16(struct tda998x_priv *priv, uint16_t reg, uint16_t val)
 	uint8_t buf[] = {REG2ADDR(reg), val >> 8, val};
 	int ret;
 
+	mutex_lock(&priv->mutex);
 	ret = set_page(priv, reg);
 	if (ret < 0)
-		return;
+		goto out;
 
 	ret = i2c_master_send(client, buf, sizeof(buf));
 	if (ret < 0)
 		dev_err(&client->dev, "Error %d writing to 0x%x\n", ret, reg);
+out:
+	mutex_unlock(&priv->mutex);
 }
 
 static void
@@ -536,6 +550,17 @@ tda998x_reset(struct tda998x_priv *priv)
 	reg_write(priv, REG_MUX_VP_VIP_OUT, 0x24);
 }
 
+/* handle HDMI connect/disconnect */
+static void tda998x_hpd(struct work_struct *work)
+{
+	struct delayed_work *dwork = to_delayed_work(work);
+	struct tda998x_priv *priv =
+			container_of(dwork, struct tda998x_priv, dwork);
+
+	if (priv->encoder && priv->encoder->dev)
+		drm_kms_helper_hotplug_event(priv->encoder->dev);
+}
+
 /*
  * only 2 interrupts may occur: screen plug/unplug and EDID read
  */
@@ -559,8 +584,7 @@ static irqreturn_t tda998x_irq_thread(int irq, void *data)
 		priv->wq_edid_wait = 0;
 		wake_up(&priv->wq_edid);
 	} else if (cec != 0) {			/* HPD change */
-		if (priv->encoder && priv->encoder->dev)
-			drm_helper_hpd_irq_event(priv->encoder->dev);
+		schedule_delayed_work(&priv->dwork, HZ/10);
 	}
 	return IRQ_HANDLED;
 }
@@ -1170,8 +1194,10 @@ static void tda998x_destroy(struct tda998x_priv *priv)
 	/* disable all IRQs and free the IRQ handler */
 	cec_write(priv, REG_CEC_RXSHPDINTENA, 0);
 	reg_clear(priv, REG_INT_FLAGS_2, INT_FLAGS_2_EDID_BLK_RD);
-	if (priv->hdmi->irq)
+	if (priv->hdmi->irq) {
 		free_irq(priv->hdmi->irq, priv);
+		cancel_delayed_work_sync(&priv->dwork);
+	}
 
 	i2c_unregister_device(priv->cec);
 }
@@ -1255,6 +1281,7 @@ static int tda998x_create(struct i2c_client *client, struct tda998x_priv *priv)
 	struct device_node *np = client->dev.of_node;
 	u32 video;
 	int rev_lo, rev_hi, ret;
+	unsigned short cec_addr;
 
 	priv->vip_cntrl_0 = VIP_CNTRL_0_SWAP_A(2) | VIP_CNTRL_0_SWAP_B(3);
 	priv->vip_cntrl_1 = VIP_CNTRL_1_SWAP_C(0) | VIP_CNTRL_1_SWAP_D(1);
@@ -1262,12 +1289,16 @@ static int tda998x_create(struct i2c_client *client, struct tda998x_priv *priv)
 
 	priv->current_page = 0xff;
 	priv->hdmi = client;
-	priv->cec = i2c_new_dummy(client->adapter, 0x34);
+	/* CEC I2C address bound to TDA998x I2C addr by configuration pins */
+	cec_addr = 0x34 + (client->addr & 0x03);
+	priv->cec = i2c_new_dummy(client->adapter, cec_addr);
 	if (!priv->cec)
 		return -ENODEV;
 
 	priv->dpms = DRM_MODE_DPMS_OFF;
 
+	mutex_init(&priv->mutex);	/* protect the page access */
+
 	/* wake up the device: */
 	cec_write(priv, REG_CEC_ENAMODS,
 			CEC_ENAMODS_EN_RXSENS | CEC_ENAMODS_EN_HDMI);
@@ -1323,8 +1354,9 @@ static int tda998x_create(struct i2c_client *client, struct tda998x_priv *priv)
 	if (client->irq) {
 		int irqf_trigger;
 
-		/* init read EDID waitqueue */
+		/* init read EDID waitqueue and HDP work */
 		init_waitqueue_head(&priv->wq_edid);
+		INIT_DELAYED_WORK(&priv->dwork, tda998x_hpd);
 
 		/* clear pending interrupts */
 		reg_read(priv, REG_INT_FLAGS_0);
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index cbbd23f78518..489b220af02b 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -462,19 +462,13 @@ void intel_detect_pch(struct drm_device *dev)
 			} else if (id == INTEL_PCH_LPT_DEVICE_ID_TYPE) {
 				dev_priv->pch_type = PCH_LPT;
 				DRM_DEBUG_KMS("Found LynxPoint PCH\n");
-				WARN_ON(!IS_HASWELL(dev));
-				WARN_ON(IS_HSW_ULT(dev));
-			} else if (IS_BROADWELL(dev)) {
-				dev_priv->pch_type = PCH_LPT;
-				dev_priv->pch_id =
-					INTEL_PCH_LPT_LP_DEVICE_ID_TYPE;
-				DRM_DEBUG_KMS("This is Broadwell, assuming "
-					      "LynxPoint LP PCH\n");
+				WARN_ON(!IS_HASWELL(dev) && !IS_BROADWELL(dev));
+				WARN_ON(IS_HSW_ULT(dev) || IS_BDW_ULT(dev));
 			} else if (id == INTEL_PCH_LPT_LP_DEVICE_ID_TYPE) {
 				dev_priv->pch_type = PCH_LPT;
 				DRM_DEBUG_KMS("Found LynxPoint LP PCH\n");
-				WARN_ON(!IS_HASWELL(dev));
-				WARN_ON(!IS_HSW_ULT(dev));
+				WARN_ON(!IS_HASWELL(dev) && !IS_BROADWELL(dev));
+				WARN_ON(!IS_HSW_ULT(dev) && !IS_BDW_ULT(dev));
 			} else if (id == INTEL_PCH_SPT_DEVICE_ID_TYPE) {
 				dev_priv->pch_type = PCH_SPT;
 				DRM_DEBUG_KMS("Found SunrisePoint PCH\n");
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 176afc507f7c..24cc36a9f3ff 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -2167,8 +2167,7 @@ struct drm_i915_cmd_table {
 #define IS_HSW_EARLY_SDV(dev)	(IS_HASWELL(dev) && \
 				 (INTEL_DEVID(dev) & 0xFF00) == 0x0C00)
 #define IS_BDW_ULT(dev)		(IS_BROADWELL(dev) && \
-				 ((INTEL_DEVID(dev) & 0xf) == 0x2  || \
-				 (INTEL_DEVID(dev) & 0xf) == 0x6 || \
+				 ((INTEL_DEVID(dev) & 0xf) == 0x6 ||	\
 				 (INTEL_DEVID(dev) & 0xf) == 0xe))
 #define IS_BDW_GT3(dev)		(IS_BROADWELL(dev) && \
 				 (INTEL_DEVID(dev) & 0x00F0) == 0x0020)
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 76354d3ba925..5f614828d365 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -3148,6 +3148,13 @@ static void i965_write_fence_reg(struct drm_device *dev, int reg,
 		u32 size = i915_gem_obj_ggtt_size(obj);
 		uint64_t val;
 
+		/* Adjust fence size to match tiled area */
+		if (obj->tiling_mode != I915_TILING_NONE) {
+			uint32_t row_size = obj->stride *
+				(obj->tiling_mode == I915_TILING_Y ? 32 : 8);
+			size = (size / row_size) * row_size;
+		}
+
 		val = (uint64_t)((i915_gem_obj_ggtt_offset(obj) + size - 4096) &
 				 0xfffff000) << 32;
 		val |= i915_gem_obj_ggtt_offset(obj) & 0xfffff000;
@@ -4884,25 +4891,18 @@ i915_gem_init_hw(struct drm_device *dev)
 	for (i = 0; i < NUM_L3_SLICES(dev); i++)
 		i915_gem_l3_remap(&dev_priv->ring[RCS], i);
 
-	/*
-	 * XXX: Contexts should only be initialized once. Doing a switch to the
-	 * default context switch however is something we'd like to do after
-	 * reset or thaw (the latter may not actually be necessary for HW, but
-	 * goes with our code better). Context switching requires rings (for
-	 * the do_switch), but before enabling PPGTT. So don't move this.
-	 */
-	ret = i915_gem_context_enable(dev_priv);
+	ret = i915_ppgtt_init_hw(dev);
 	if (ret && ret != -EIO) {
-		DRM_ERROR("Context enable failed %d\n", ret);
+		DRM_ERROR("PPGTT enable failed %d\n", ret);
 		i915_gem_cleanup_ringbuffer(dev);
-
-		return ret;
 	}
 
-	ret = i915_ppgtt_init_hw(dev);
+	ret = i915_gem_context_enable(dev_priv);
 	if (ret && ret != -EIO) {
-		DRM_ERROR("PPGTT enable failed %d\n", ret);
+		DRM_ERROR("Context enable failed %d\n", ret);
 		i915_gem_cleanup_ringbuffer(dev);
+
+		return ret;
 	}
 
 	return ret;
diff --git a/drivers/gpu/drm/i915/intel_panel.c b/drivers/gpu/drm/i915/intel_panel.c
index 4d63839bd9b4..dfb783a8f2c3 100644
--- a/drivers/gpu/drm/i915/intel_panel.c
+++ b/drivers/gpu/drm/i915/intel_panel.c
@@ -962,7 +962,7 @@ void intel_panel_enable_backlight(struct intel_connector *connector)
 
 	WARN_ON(panel->backlight.max == 0);
 
-	if (panel->backlight.level == 0) {
+	if (panel->backlight.level <= panel->backlight.min) {
 		panel->backlight.level = panel->backlight.max;
 		if (panel->backlight.device)
 			panel->backlight.device->props.brightness =
diff --git a/drivers/gpu/drm/radeon/cik_sdma.c b/drivers/gpu/drm/radeon/cik_sdma.c
index a0133c74f4cf..42cd0cffe210 100644
--- a/drivers/gpu/drm/radeon/cik_sdma.c
+++ b/drivers/gpu/drm/radeon/cik_sdma.c
@@ -816,7 +816,6 @@ void cik_sdma_vm_write_pages(struct radeon_device *rdev,
 		for (; ndw > 0; ndw -= 2, --count, pe += 8) {
 			if (flags & R600_PTE_SYSTEM) {
 				value = radeon_vm_map_gart(rdev, addr);
-				value &= 0xFFFFFFFFFFFFF000ULL;
 			} else if (flags & R600_PTE_VALID) {
 				value = addr;
 			} else {
diff --git a/drivers/gpu/drm/radeon/ni_dma.c b/drivers/gpu/drm/radeon/ni_dma.c
index 4be2bb7cbef3..ce787a9f12c0 100644
--- a/drivers/gpu/drm/radeon/ni_dma.c
+++ b/drivers/gpu/drm/radeon/ni_dma.c
@@ -372,7 +372,6 @@ void cayman_dma_vm_write_pages(struct radeon_device *rdev,
 		for (; ndw > 0; ndw -= 2, --count, pe += 8) {
 			if (flags & R600_PTE_SYSTEM) {
 				value = radeon_vm_map_gart(rdev, addr);
-				value &= 0xFFFFFFFFFFFFF000ULL;
 			} else if (flags & R600_PTE_VALID) {
 				value = addr;
 			} else {
diff --git a/drivers/gpu/drm/radeon/r100.c b/drivers/gpu/drm/radeon/r100.c
index 74f06d540591..279801ca5110 100644
--- a/drivers/gpu/drm/radeon/r100.c
+++ b/drivers/gpu/drm/radeon/r100.c
@@ -644,6 +644,7 @@ int r100_pci_gart_init(struct radeon_device *rdev)
 		return r;
 	rdev->gart.table_size = rdev->gart.num_gpu_pages * 4;
 	rdev->asic->gart.tlb_flush = &r100_pci_gart_tlb_flush;
+	rdev->asic->gart.get_page_entry = &r100_pci_gart_get_page_entry;
 	rdev->asic->gart.set_page = &r100_pci_gart_set_page;
 	return radeon_gart_table_ram_alloc(rdev);
 }
@@ -681,11 +682,16 @@ void r100_pci_gart_disable(struct radeon_device *rdev)
 	WREG32(RADEON_AIC_HI_ADDR, 0);
 }
 
+uint64_t r100_pci_gart_get_page_entry(uint64_t addr, uint32_t flags)
+{
+	return addr;
+}
+
 void r100_pci_gart_set_page(struct radeon_device *rdev, unsigned i,
-			    uint64_t addr, uint32_t flags)
+			    uint64_t entry)
 {
 	u32 *gtt = rdev->gart.ptr;
-	gtt[i] = cpu_to_le32(lower_32_bits(addr));
+	gtt[i] = cpu_to_le32(lower_32_bits(entry));
 }
 
 void r100_pci_gart_fini(struct radeon_device *rdev)
diff --git a/drivers/gpu/drm/radeon/r300.c b/drivers/gpu/drm/radeon/r300.c
index 064ad5569cca..08d68f3e13e9 100644
--- a/drivers/gpu/drm/radeon/r300.c
+++ b/drivers/gpu/drm/radeon/r300.c
@@ -73,11 +73,8 @@ void rv370_pcie_gart_tlb_flush(struct radeon_device *rdev)
 #define R300_PTE_WRITEABLE (1 << 2)
 #define R300_PTE_READABLE  (1 << 3)
 
-void rv370_pcie_gart_set_page(struct radeon_device *rdev, unsigned i,
-			      uint64_t addr, uint32_t flags)
+uint64_t rv370_pcie_gart_get_page_entry(uint64_t addr, uint32_t flags)
 {
-	void __iomem *ptr = rdev->gart.ptr;
-
 	addr = (lower_32_bits(addr) >> 8) |
 		((upper_32_bits(addr) & 0xff) << 24);
 	if (flags & RADEON_GART_PAGE_READ)
@@ -86,10 +83,18 @@ void rv370_pcie_gart_set_page(struct radeon_device *rdev, unsigned i,
 		addr |= R300_PTE_WRITEABLE;
 	if (!(flags & RADEON_GART_PAGE_SNOOP))
 		addr |= R300_PTE_UNSNOOPED;
+	return addr;
+}
+
+void rv370_pcie_gart_set_page(struct radeon_device *rdev, unsigned i,
+			      uint64_t entry)
+{
+	void __iomem *ptr = rdev->gart.ptr;
+
 	/* on x86 we want this to be CPU endian, on powerpc
 	 * on powerpc without HW swappers, it'll get swapped on way
 	 * into VRAM - so no need for cpu_to_le32 on VRAM tables */
-	writel(addr, ((void __iomem *)ptr) + (i * 4));
+	writel(entry, ((void __iomem *)ptr) + (i * 4));
 }
 
 int rv370_pcie_gart_init(struct radeon_device *rdev)
@@ -109,6 +114,7 @@ int rv370_pcie_gart_init(struct radeon_device *rdev)
 		DRM_ERROR("Failed to register debugfs file for PCIE gart !\n");
 	rdev->gart.table_size = rdev->gart.num_gpu_pages * 4;
 	rdev->asic->gart.tlb_flush = &rv370_pcie_gart_tlb_flush;
+	rdev->asic->gart.get_page_entry = &rv370_pcie_gart_get_page_entry;
 	rdev->asic->gart.set_page = &rv370_pcie_gart_set_page;
 	return radeon_gart_table_vram_alloc(rdev);
 }
diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h
index 54529b837afa..3f2a8d3febca 100644
--- a/drivers/gpu/drm/radeon/radeon.h
+++ b/drivers/gpu/drm/radeon/radeon.h
@@ -242,6 +242,7 @@ bool radeon_get_bios(struct radeon_device *rdev);
  * Dummy page
  */
 struct radeon_dummy_page {
+	uint64_t	entry;
 	struct page	*page;
 	dma_addr_t	addr;
 };
@@ -645,7 +646,7 @@ struct radeon_gart {
 	unsigned			num_cpu_pages;
 	unsigned			table_size;
 	struct page			**pages;
-	dma_addr_t			*pages_addr;
+	uint64_t			*pages_entry;
 	bool				ready;
 };
 
@@ -1847,8 +1848,9 @@ struct radeon_asic {
 	/* gart */
 	struct {
 		void (*tlb_flush)(struct radeon_device *rdev);
+		uint64_t (*get_page_entry)(uint64_t addr, uint32_t flags);
 		void (*set_page)(struct radeon_device *rdev, unsigned i,
-				 uint64_t addr, uint32_t flags);
+				 uint64_t entry);
 	} gart;
 	struct {
 		int (*init)(struct radeon_device *rdev);
@@ -2852,7 +2854,8 @@ static inline void radeon_ring_write(struct radeon_ring *ring, uint32_t v)
 #define radeon_vga_set_state(rdev, state) (rdev)->asic->vga_set_state((rdev), (state))
 #define radeon_asic_reset(rdev) (rdev)->asic->asic_reset((rdev))
 #define radeon_gart_tlb_flush(rdev) (rdev)->asic->gart.tlb_flush((rdev))
-#define radeon_gart_set_page(rdev, i, p, f) (rdev)->asic->gart.set_page((rdev), (i), (p), (f))
+#define radeon_gart_get_page_entry(a, f) (rdev)->asic->gart.get_page_entry((a), (f))
+#define radeon_gart_set_page(rdev, i, e) (rdev)->asic->gart.set_page((rdev), (i), (e))
 #define radeon_asic_vm_init(rdev) (rdev)->asic->vm.init((rdev))
 #define radeon_asic_vm_fini(rdev) (rdev)->asic->vm.fini((rdev))
 #define radeon_asic_vm_copy_pages(rdev, ib, pe, src, count) ((rdev)->asic->vm.copy_pages((rdev), (ib), (pe), (src), (count)))
diff --git a/drivers/gpu/drm/radeon/radeon_asic.c b/drivers/gpu/drm/radeon/radeon_asic.c
index 121aff6a3b41..ed0e10eee2dc 100644
--- a/drivers/gpu/drm/radeon/radeon_asic.c
+++ b/drivers/gpu/drm/radeon/radeon_asic.c
@@ -159,11 +159,13 @@ void radeon_agp_disable(struct radeon_device *rdev)
 		DRM_INFO("Forcing AGP to PCIE mode\n");
 		rdev->flags |= RADEON_IS_PCIE;
 		rdev->asic->gart.tlb_flush = &rv370_pcie_gart_tlb_flush;
+		rdev->asic->gart.get_page_entry = &rv370_pcie_gart_get_page_entry;
 		rdev->asic->gart.set_page = &rv370_pcie_gart_set_page;
 	} else {
 		DRM_INFO("Forcing AGP to PCI mode\n");
 		rdev->flags |= RADEON_IS_PCI;
 		rdev->asic->gart.tlb_flush = &r100_pci_gart_tlb_flush;
+		rdev->asic->gart.get_page_entry = &r100_pci_gart_get_page_entry;
 		rdev->asic->gart.set_page = &r100_pci_gart_set_page;
 	}
 	rdev->mc.gtt_size = radeon_gart_size * 1024 * 1024;
@@ -199,6 +201,7 @@ static struct radeon_asic r100_asic = {
 	.mc_wait_for_idle = &r100_mc_wait_for_idle,
 	.gart = {
 		.tlb_flush = &r100_pci_gart_tlb_flush,
+		.get_page_entry = &r100_pci_gart_get_page_entry,
 		.set_page = &r100_pci_gart_set_page,
 	},
 	.ring = {
@@ -265,6 +268,7 @@ static struct radeon_asic r200_asic = {
 	.mc_wait_for_idle = &r100_mc_wait_for_idle,
 	.gart = {
 		.tlb_flush = &r100_pci_gart_tlb_flush,
+		.get_page_entry = &r100_pci_gart_get_page_entry,
 		.set_page = &r100_pci_gart_set_page,
 	},
 	.ring = {
@@ -359,6 +363,7 @@ static struct radeon_asic r300_asic = {
 	.mc_wait_for_idle = &r300_mc_wait_for_idle,
 	.gart = {
 		.tlb_flush = &r100_pci_gart_tlb_flush,
+		.get_page_entry = &r100_pci_gart_get_page_entry,
 		.set_page = &r100_pci_gart_set_page,
 	},
 	.ring = {
@@ -425,6 +430,7 @@ static struct radeon_asic r300_asic_pcie = {
 	.mc_wait_for_idle = &r300_mc_wait_for_idle,
 	.gart = {
 		.tlb_flush = &rv370_pcie_gart_tlb_flush,
+		.get_page_entry = &rv370_pcie_gart_get_page_entry,
 		.set_page = &rv370_pcie_gart_set_page,
 	},
 	.ring = {
@@ -491,6 +497,7 @@ static struct radeon_asic r420_asic = {
 	.mc_wait_for_idle = &r300_mc_wait_for_idle,
 	.gart = {
 		.tlb_flush = &rv370_pcie_gart_tlb_flush,
+		.get_page_entry = &rv370_pcie_gart_get_page_entry,
 		.set_page = &rv370_pcie_gart_set_page,
 	},
 	.ring = {
@@ -557,6 +564,7 @@ static struct radeon_asic rs400_asic = {
 	.mc_wait_for_idle = &rs400_mc_wait_for_idle,
 	.gart = {
 		.tlb_flush = &rs400_gart_tlb_flush,
+		.get_page_entry = &rs400_gart_get_page_entry,
 		.set_page = &rs400_gart_set_page,
 	},
 	.ring = {
@@ -623,6 +631,7 @@ static struct radeon_asic rs600_asic = {
 	.mc_wait_for_idle = &rs600_mc_wait_for_idle,
 	.gart = {
 		.tlb_flush = &rs600_gart_tlb_flush,
+		.get_page_entry = &rs600_gart_get_page_entry,
 		.set_page = &rs600_gart_set_page,
 	},
 	.ring = {
@@ -691,6 +700,7 @@ static struct radeon_asic rs690_asic = {
 	.mc_wait_for_idle = &rs690_mc_wait_for_idle,
 	.gart = {
 		.tlb_flush = &rs400_gart_tlb_flush,
+		.get_page_entry = &rs400_gart_get_page_entry,
 		.set_page = &rs400_gart_set_page,
 	},
 	.ring = {
@@ -759,6 +769,7 @@ static struct radeon_asic rv515_asic = {
 	.mc_wait_for_idle = &rv515_mc_wait_for_idle,
 	.gart = {
 		.tlb_flush = &rv370_pcie_gart_tlb_flush,
+		.get_page_entry = &rv370_pcie_gart_get_page_entry,
 		.set_page = &rv370_pcie_gart_set_page,
 	},
 	.ring = {
@@ -825,6 +836,7 @@ static struct radeon_asic r520_asic = {
 	.mc_wait_for_idle = &r520_mc_wait_for_idle,
 	.gart = {
 		.tlb_flush = &rv370_pcie_gart_tlb_flush,
+		.get_page_entry = &rv370_pcie_gart_get_page_entry,
 		.set_page = &rv370_pcie_gart_set_page,
 	},
 	.ring = {
@@ -919,6 +931,7 @@ static struct radeon_asic r600_asic = {
 	.get_gpu_clock_counter = &r600_get_gpu_clock_counter,
 	.gart = {
 		.tlb_flush = &r600_pcie_gart_tlb_flush,
+		.get_page_entry = &rs600_gart_get_page_entry,
 		.set_page = &rs600_gart_set_page,
 	},
 	.ring = {
@@ -1004,6 +1017,7 @@ static struct radeon_asic rv6xx_asic = {
 	.get_gpu_clock_counter = &r600_get_gpu_clock_counter,
 	.gart = {
 		.tlb_flush = &r600_pcie_gart_tlb_flush,
+		.get_page_entry = &rs600_gart_get_page_entry,
 		.set_page = &rs600_gart_set_page,
 	},
 	.ring = {
@@ -1095,6 +1109,7 @@ static struct radeon_asic rs780_asic = {
 	.get_gpu_clock_counter = &r600_get_gpu_clock_counter,
 	.gart = {
 		.tlb_flush = &r600_pcie_gart_tlb_flush,
+		.get_page_entry = &rs600_gart_get_page_entry,
 		.set_page = &rs600_gart_set_page,
 	},
 	.ring = {
@@ -1199,6 +1214,7 @@ static struct radeon_asic rv770_asic = {
 	.get_gpu_clock_counter = &r600_get_gpu_clock_counter,
 	.gart = {
 		.tlb_flush = &r600_pcie_gart_tlb_flush,
+		.get_page_entry = &rs600_gart_get_page_entry,
 		.set_page = &rs600_gart_set_page,
 	},
 	.ring = {
@@ -1317,6 +1333,7 @@ static struct radeon_asic evergreen_asic = {
 	.get_gpu_clock_counter = &r600_get_gpu_clock_counter,
 	.gart = {
 		.tlb_flush = &evergreen_pcie_gart_tlb_flush,
+		.get_page_entry = &rs600_gart_get_page_entry,
 		.set_page = &rs600_gart_set_page,
 	},
 	.ring = {
@@ -1409,6 +1426,7 @@ static struct radeon_asic sumo_asic = {
 	.get_gpu_clock_counter = &r600_get_gpu_clock_counter,
 	.gart = {
 		.tlb_flush = &evergreen_pcie_gart_tlb_flush,
+		.get_page_entry = &rs600_gart_get_page_entry,
 		.set_page = &rs600_gart_set_page,
 	},
 	.ring = {
@@ -1500,6 +1518,7 @@ static struct radeon_asic btc_asic = {
 	.get_gpu_clock_counter = &r600_get_gpu_clock_counter,
 	.gart = {
 		.tlb_flush = &evergreen_pcie_gart_tlb_flush,
+		.get_page_entry = &rs600_gart_get_page_entry,
 		.set_page = &rs600_gart_set_page,
 	},
 	.ring = {
@@ -1635,6 +1654,7 @@ static struct radeon_asic cayman_asic = {
 	.get_gpu_clock_counter = &r600_get_gpu_clock_counter,
 	.gart = {
 		.tlb_flush = &cayman_pcie_gart_tlb_flush,
+		.get_page_entry = &rs600_gart_get_page_entry,
 		.set_page = &rs600_gart_set_page,
 	},
 	.vm = {
@@ -1738,6 +1758,7 @@ static struct radeon_asic trinity_asic = {
 	.get_gpu_clock_counter = &r600_get_gpu_clock_counter,
 	.gart = {
 		.tlb_flush = &cayman_pcie_gart_tlb_flush,
+		.get_page_entry = &rs600_gart_get_page_entry,
 		.set_page = &rs600_gart_set_page,
 	},
 	.vm = {
@@ -1871,6 +1892,7 @@ static struct radeon_asic si_asic = {
 	.get_gpu_clock_counter = &si_get_gpu_clock_counter,
 	.gart = {
 		.tlb_flush = &si_pcie_gart_tlb_flush,
+		.get_page_entry = &rs600_gart_get_page_entry,
 		.set_page = &rs600_gart_set_page,
 	},
 	.vm = {
@@ -2032,6 +2054,7 @@ static struct radeon_asic ci_asic = {
 	.get_gpu_clock_counter = &cik_get_gpu_clock_counter,
 	.gart = {
 		.tlb_flush = &cik_pcie_gart_tlb_flush,
+		.get_page_entry = &rs600_gart_get_page_entry,
 		.set_page = &rs600_gart_set_page,
 	},
 	.vm = {
@@ -2139,6 +2162,7 @@ static struct radeon_asic kv_asic = {
 	.get_gpu_clock_counter = &cik_get_gpu_clock_counter,
 	.gart = {
 		.tlb_flush = &cik_pcie_gart_tlb_flush,
+		.get_page_entry = &rs600_gart_get_page_entry,
 		.set_page = &rs600_gart_set_page,
 	},
 	.vm = {
diff --git a/drivers/gpu/drm/radeon/radeon_asic.h b/drivers/gpu/drm/radeon/radeon_asic.h
index 2a45d548d5ec..8d787d115653 100644
--- a/drivers/gpu/drm/radeon/radeon_asic.h
+++ b/drivers/gpu/drm/radeon/radeon_asic.h
@@ -67,8 +67,9 @@ bool r100_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *cp);
 int r100_asic_reset(struct radeon_device *rdev);
 u32 r100_get_vblank_counter(struct radeon_device *rdev, int crtc);
 void r100_pci_gart_tlb_flush(struct radeon_device *rdev);
+uint64_t r100_pci_gart_get_page_entry(uint64_t addr, uint32_t flags);
 void r100_pci_gart_set_page(struct radeon_device *rdev, unsigned i,
-			    uint64_t addr, uint32_t flags);
+			    uint64_t entry);
 void r100_ring_start(struct radeon_device *rdev, struct radeon_ring *ring);
 int r100_irq_set(struct radeon_device *rdev);
 int r100_irq_process(struct radeon_device *rdev);
@@ -172,8 +173,9 @@ extern void r300_fence_ring_emit(struct radeon_device *rdev,
 				struct radeon_fence *fence);
 extern int r300_cs_parse(struct radeon_cs_parser *p);
 extern void rv370_pcie_gart_tlb_flush(struct radeon_device *rdev);
+extern uint64_t rv370_pcie_gart_get_page_entry(uint64_t addr, uint32_t flags);
 extern void rv370_pcie_gart_set_page(struct radeon_device *rdev, unsigned i,
-				     uint64_t addr, uint32_t flags);
+				     uint64_t entry);
 extern void rv370_set_pcie_lanes(struct radeon_device *rdev, int lanes);
 extern int rv370_get_pcie_lanes(struct radeon_device *rdev);
 extern void r300_set_reg_safe(struct radeon_device *rdev);
@@ -208,8 +210,9 @@ extern void rs400_fini(struct radeon_device *rdev);
 extern int rs400_suspend(struct radeon_device *rdev);
 extern int rs400_resume(struct radeon_device *rdev);
 void rs400_gart_tlb_flush(struct radeon_device *rdev);
+uint64_t rs400_gart_get_page_entry(uint64_t addr, uint32_t flags);
 void rs400_gart_set_page(struct radeon_device *rdev, unsigned i,
-			 uint64_t addr, uint32_t flags);
+			 uint64_t entry);
 uint32_t rs400_mc_rreg(struct radeon_device *rdev, uint32_t reg);
 void rs400_mc_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v);
 int rs400_gart_init(struct radeon_device *rdev);
@@ -232,8 +235,9 @@ int rs600_irq_process(struct radeon_device *rdev);
 void rs600_irq_disable(struct radeon_device *rdev);
 u32 rs600_get_vblank_counter(struct radeon_device *rdev, int crtc);
 void rs600_gart_tlb_flush(struct radeon_device *rdev);
+uint64_t rs600_gart_get_page_entry(uint64_t addr, uint32_t flags);
 void rs600_gart_set_page(struct radeon_device *rdev, unsigned i,
-			 uint64_t addr, uint32_t flags);
+			 uint64_t entry);
 uint32_t rs600_mc_rreg(struct radeon_device *rdev, uint32_t reg);
 void rs600_mc_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v);
 void rs600_bandwidth_update(struct radeon_device *rdev);
diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c
index 0ec65168f331..bd7519fdd3f4 100644
--- a/drivers/gpu/drm/radeon/radeon_device.c
+++ b/drivers/gpu/drm/radeon/radeon_device.c
@@ -774,6 +774,8 @@ int radeon_dummy_page_init(struct radeon_device *rdev)
 		rdev->dummy_page.page = NULL;
 		return -ENOMEM;
 	}
+	rdev->dummy_page.entry = radeon_gart_get_page_entry(rdev->dummy_page.addr,
+							    RADEON_GART_PAGE_DUMMY);
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/radeon/radeon_gart.c b/drivers/gpu/drm/radeon/radeon_gart.c
index 84146d5901aa..5450fa95a47e 100644
--- a/drivers/gpu/drm/radeon/radeon_gart.c
+++ b/drivers/gpu/drm/radeon/radeon_gart.c
@@ -165,6 +165,19 @@ int radeon_gart_table_vram_pin(struct radeon_device *rdev)
 		radeon_bo_unpin(rdev->gart.robj);
 	radeon_bo_unreserve(rdev->gart.robj);
 	rdev->gart.table_addr = gpu_addr;
+
+	if (!r) {
+		int i;
+
+		/* We might have dropped some GART table updates while it wasn't
+		 * mapped, restore all entries
+		 */
+		for (i = 0; i < rdev->gart.num_gpu_pages; i++)
+			radeon_gart_set_page(rdev, i, rdev->gart.pages_entry[i]);
+		mb();
+		radeon_gart_tlb_flush(rdev);
+	}
+
 	return r;
 }
 
@@ -228,7 +241,6 @@ void radeon_gart_unbind(struct radeon_device *rdev, unsigned offset,
 	unsigned t;
 	unsigned p;
 	int i, j;
-	u64 page_base;
 
 	if (!rdev->gart.ready) {
 		WARN(1, "trying to unbind memory from uninitialized GART !\n");
@@ -239,14 +251,12 @@ void radeon_gart_unbind(struct radeon_device *rdev, unsigned offset,
 	for (i = 0; i < pages; i++, p++) {
 		if (rdev->gart.pages[p]) {
 			rdev->gart.pages[p] = NULL;
-			rdev->gart.pages_addr[p] = rdev->dummy_page.addr;
-			page_base = rdev->gart.pages_addr[p];
 			for (j = 0; j < (PAGE_SIZE / RADEON_GPU_PAGE_SIZE); j++, t++) {
+				rdev->gart.pages_entry[t] = rdev->dummy_page.entry;
 				if (rdev->gart.ptr) {
-					radeon_gart_set_page(rdev, t, page_base,
-							     RADEON_GART_PAGE_DUMMY);
+					radeon_gart_set_page(rdev, t,
+							     rdev->dummy_page.entry);
 				}
-				page_base += RADEON_GPU_PAGE_SIZE;
 			}
 		}
 	}
@@ -274,7 +284,7 @@ int radeon_gart_bind(struct radeon_device *rdev, unsigned offset,
 {
 	unsigned t;
 	unsigned p;
-	uint64_t page_base;
+	uint64_t page_base, page_entry;
 	int i, j;
 
 	if (!rdev->gart.ready) {
@@ -285,14 +295,15 @@ int radeon_gart_bind(struct radeon_device *rdev, unsigned offset,
 	p = t / (PAGE_SIZE / RADEON_GPU_PAGE_SIZE);
 
 	for (i = 0; i < pages; i++, p++) {
-		rdev->gart.pages_addr[p] = dma_addr[i];
 		rdev->gart.pages[p] = pagelist[i];
-		if (rdev->gart.ptr) {
-			page_base = rdev->gart.pages_addr[p];
-			for (j = 0; j < (PAGE_SIZE / RADEON_GPU_PAGE_SIZE); j++, t++) {
-				radeon_gart_set_page(rdev, t, page_base, flags);
-				page_base += RADEON_GPU_PAGE_SIZE;
+		page_base = dma_addr[i];
+		for (j = 0; j < (PAGE_SIZE / RADEON_GPU_PAGE_SIZE); j++, t++) {
+			page_entry = radeon_gart_get_page_entry(page_base, flags);
+			rdev->gart.pages_entry[t] = page_entry;
+			if (rdev->gart.ptr) {
+				radeon_gart_set_page(rdev, t, page_entry);
 			}
+			page_base += RADEON_GPU_PAGE_SIZE;
 		}
 	}
 	mb();
@@ -334,16 +345,15 @@ int radeon_gart_init(struct radeon_device *rdev)
 		radeon_gart_fini(rdev);
 		return -ENOMEM;
 	}
-	rdev->gart.pages_addr = vzalloc(sizeof(dma_addr_t) *
-					rdev->gart.num_cpu_pages);
-	if (rdev->gart.pages_addr == NULL) {
+	rdev->gart.pages_entry = vmalloc(sizeof(uint64_t) *
+					 rdev->gart.num_gpu_pages);
+	if (rdev->gart.pages_entry == NULL) {
 		radeon_gart_fini(rdev);
 		return -ENOMEM;
 	}
 	/* set GART entry to point to the dummy page by default */
-	for (i = 0; i < rdev->gart.num_cpu_pages; i++) {
-		rdev->gart.pages_addr[i] = rdev->dummy_page.addr;
-	}
+	for (i = 0; i < rdev->gart.num_gpu_pages; i++)
+		rdev->gart.pages_entry[i] = rdev->dummy_page.entry;
 	return 0;
 }
 
@@ -356,15 +366,15 @@ int radeon_gart_init(struct radeon_device *rdev)
  */
 void radeon_gart_fini(struct radeon_device *rdev)
 {
-	if (rdev->gart.pages && rdev->gart.pages_addr && rdev->gart.ready) {
+	if (rdev->gart.ready) {
 		/* unbind pages */
 		radeon_gart_unbind(rdev, 0, rdev->gart.num_cpu_pages);
 	}
 	rdev->gart.ready = false;
 	vfree(rdev->gart.pages);
-	vfree(rdev->gart.pages_addr);
+	vfree(rdev->gart.pages_entry);
 	rdev->gart.pages = NULL;
-	rdev->gart.pages_addr = NULL;
+	rdev->gart.pages_entry = NULL;
 
 	radeon_dummy_page_fini(rdev);
 }
diff --git a/drivers/gpu/drm/radeon/radeon_kfd.c b/drivers/gpu/drm/radeon/radeon_kfd.c
index 8bf87f1203cc..bef9a0953284 100644
--- a/drivers/gpu/drm/radeon/radeon_kfd.c
+++ b/drivers/gpu/drm/radeon/radeon_kfd.c
@@ -436,7 +436,7 @@ static int kgd_init_memory(struct kgd_dev *kgd)
 static int kgd_init_pipeline(struct kgd_dev *kgd, uint32_t pipe_id,
 				uint32_t hpd_size, uint64_t hpd_gpu_addr)
 {
-	uint32_t mec = (++pipe_id / CIK_PIPE_PER_MEC) + 1;
+	uint32_t mec = (pipe_id / CIK_PIPE_PER_MEC) + 1;
 	uint32_t pipe = (pipe_id % CIK_PIPE_PER_MEC);
 
 	lock_srbm(kgd, mec, pipe, 0, 0);
diff --git a/drivers/gpu/drm/radeon/radeon_vm.c b/drivers/gpu/drm/radeon/radeon_vm.c
index cde48c42b30a..06d2246d07f1 100644
--- a/drivers/gpu/drm/radeon/radeon_vm.c
+++ b/drivers/gpu/drm/radeon/radeon_vm.c
@@ -587,10 +587,8 @@ uint64_t radeon_vm_map_gart(struct radeon_device *rdev, uint64_t addr)
 	uint64_t result;
 
 	/* page table offset */
-	result = rdev->gart.pages_addr[addr >> PAGE_SHIFT];
-
-	/* in case cpu page size != gpu page size*/
-	result |= addr & (~PAGE_MASK);
+	result = rdev->gart.pages_entry[addr >> RADEON_GPU_PAGE_SHIFT];
+	result &= ~RADEON_GPU_PAGE_MASK;
 
 	return result;
 }
diff --git a/drivers/gpu/drm/radeon/rs400.c b/drivers/gpu/drm/radeon/rs400.c
index c5799f16aa4b..34e3235f41d2 100644
--- a/drivers/gpu/drm/radeon/rs400.c
+++ b/drivers/gpu/drm/radeon/rs400.c
@@ -212,11 +212,9 @@ void rs400_gart_fini(struct radeon_device *rdev)
 #define RS400_PTE_WRITEABLE (1 << 2)
 #define RS400_PTE_READABLE  (1 << 3)
 
-void rs400_gart_set_page(struct radeon_device *rdev, unsigned i,
-			 uint64_t addr, uint32_t flags)
+uint64_t rs400_gart_get_page_entry(uint64_t addr, uint32_t flags)
 {
 	uint32_t entry;
-	u32 *gtt = rdev->gart.ptr;
 
 	entry = (lower_32_bits(addr) & PAGE_MASK) |
 		((upper_32_bits(addr) & 0xff) << 4);
@@ -226,8 +224,14 @@ void rs400_gart_set_page(struct radeon_device *rdev, unsigned i,
 		entry |= RS400_PTE_WRITEABLE;
 	if (!(flags & RADEON_GART_PAGE_SNOOP))
 		entry |= RS400_PTE_UNSNOOPED;
-	entry = cpu_to_le32(entry);
-	gtt[i] = entry;
+	return entry;
+}
+
+void rs400_gart_set_page(struct radeon_device *rdev, unsigned i,
+			 uint64_t entry)
+{
+	u32 *gtt = rdev->gart.ptr;
+	gtt[i] = cpu_to_le32(lower_32_bits(entry));
 }
 
 int rs400_mc_wait_for_idle(struct radeon_device *rdev)
diff --git a/drivers/gpu/drm/radeon/rs600.c b/drivers/gpu/drm/radeon/rs600.c
index 9acb1c3c005b..74bce91aecc1 100644
--- a/drivers/gpu/drm/radeon/rs600.c
+++ b/drivers/gpu/drm/radeon/rs600.c
@@ -625,11 +625,8 @@ static void rs600_gart_fini(struct radeon_device *rdev)
 	radeon_gart_table_vram_free(rdev);
 }
 
-void rs600_gart_set_page(struct radeon_device *rdev, unsigned i,
-			 uint64_t addr, uint32_t flags)
+uint64_t rs600_gart_get_page_entry(uint64_t addr, uint32_t flags)
 {
-	void __iomem *ptr = (void *)rdev->gart.ptr;
-
 	addr = addr & 0xFFFFFFFFFFFFF000ULL;
 	addr |= R600_PTE_SYSTEM;
 	if (flags & RADEON_GART_PAGE_VALID)
@@ -640,7 +637,14 @@ void rs600_gart_set_page(struct radeon_device *rdev, unsigned i,
 		addr |= R600_PTE_WRITEABLE;
 	if (flags & RADEON_GART_PAGE_SNOOP)
 		addr |= R600_PTE_SNOOPED;
-	writeq(addr, ptr + (i * 8));
+	return addr;
+}
+
+void rs600_gart_set_page(struct radeon_device *rdev, unsigned i,
+			 uint64_t entry)
+{
+	void __iomem *ptr = (void *)rdev->gart.ptr;
+	writeq(entry, ptr + (i * 8));
 }
 
 int rs600_irq_set(struct radeon_device *rdev)
diff --git a/drivers/gpu/drm/radeon/si_dma.c b/drivers/gpu/drm/radeon/si_dma.c
index aa7b872b2c43..83207929fc62 100644
--- a/drivers/gpu/drm/radeon/si_dma.c
+++ b/drivers/gpu/drm/radeon/si_dma.c
@@ -123,7 +123,6 @@ void si_dma_vm_write_pages(struct radeon_device *rdev,
 		for (; ndw > 0; ndw -= 2, --count, pe += 8) {
 			if (flags & R600_PTE_SYSTEM) {
 				value = radeon_vm_map_gart(rdev, addr);
-				value &= 0xFFFFFFFFFFFFF000ULL;
 			} else if (flags & R600_PTE_VALID) {
 				value = addr;
 			} else {
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
index 7b5d22110f25..6c6b655defcf 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
@@ -406,11 +406,9 @@ int vmw_3d_resource_inc(struct vmw_private *dev_priv,
 		if (unlikely(ret != 0))
 			--dev_priv->num_3d_resources;
 	} else if (unhide_svga) {
-		mutex_lock(&dev_priv->hw_mutex);
 		vmw_write(dev_priv, SVGA_REG_ENABLE,
 			  vmw_read(dev_priv, SVGA_REG_ENABLE) &
 			  ~SVGA_REG_ENABLE_HIDE);
-		mutex_unlock(&dev_priv->hw_mutex);
 	}
 
 	mutex_unlock(&dev_priv->release_mutex);
@@ -433,13 +431,10 @@ void vmw_3d_resource_dec(struct vmw_private *dev_priv,
 	mutex_lock(&dev_priv->release_mutex);
 	if (unlikely(--dev_priv->num_3d_resources == 0))
 		vmw_release_device(dev_priv);
-	else if (hide_svga) {
-		mutex_lock(&dev_priv->hw_mutex);
+	else if (hide_svga)
 		vmw_write(dev_priv, SVGA_REG_ENABLE,
 			  vmw_read(dev_priv, SVGA_REG_ENABLE) |
 			  SVGA_REG_ENABLE_HIDE);
-		mutex_unlock(&dev_priv->hw_mutex);
-	}
 
 	n3d = (int32_t) dev_priv->num_3d_resources;
 	mutex_unlock(&dev_priv->release_mutex);
@@ -600,12 +595,14 @@ static int vmw_driver_load(struct drm_device *dev, unsigned long chipset)
 	dev_priv->dev = dev;
 	dev_priv->vmw_chipset = chipset;
 	dev_priv->last_read_seqno = (uint32_t) -100;
-	mutex_init(&dev_priv->hw_mutex);
 	mutex_init(&dev_priv->cmdbuf_mutex);
 	mutex_init(&dev_priv->release_mutex);
 	mutex_init(&dev_priv->binding_mutex);
 	rwlock_init(&dev_priv->resource_lock);
 	ttm_lock_init(&dev_priv->reservation_sem);
+	spin_lock_init(&dev_priv->hw_lock);
+	spin_lock_init(&dev_priv->waiter_lock);
+	spin_lock_init(&dev_priv->cap_lock);
 
 	for (i = vmw_res_context; i < vmw_res_max; ++i) {
 		idr_init(&dev_priv->res_idr[i]);
@@ -626,14 +623,11 @@ static int vmw_driver_load(struct drm_device *dev, unsigned long chipset)
 
 	dev_priv->enable_fb = enable_fbdev;
 
-	mutex_lock(&dev_priv->hw_mutex);
-
 	vmw_write(dev_priv, SVGA_REG_ID, SVGA_ID_2);
 	svga_id = vmw_read(dev_priv, SVGA_REG_ID);
 	if (svga_id != SVGA_ID_2) {
 		ret = -ENOSYS;
 		DRM_ERROR("Unsupported SVGA ID 0x%x\n", svga_id);
-		mutex_unlock(&dev_priv->hw_mutex);
 		goto out_err0;
 	}
 
@@ -683,10 +677,8 @@ static int vmw_driver_load(struct drm_device *dev, unsigned long chipset)
 		dev_priv->prim_bb_mem = dev_priv->vram_size;
 
 	ret = vmw_dma_masks(dev_priv);
-	if (unlikely(ret != 0)) {
-		mutex_unlock(&dev_priv->hw_mutex);
+	if (unlikely(ret != 0))
 		goto out_err0;
-	}
 
 	/*
 	 * Limit back buffer size to VRAM size.  Remove this once
@@ -695,8 +687,6 @@ static int vmw_driver_load(struct drm_device *dev, unsigned long chipset)
 	if (dev_priv->prim_bb_mem > dev_priv->vram_size)
 		dev_priv->prim_bb_mem = dev_priv->vram_size;
 
-	mutex_unlock(&dev_priv->hw_mutex);
-
 	vmw_print_capabilities(dev_priv->capabilities);
 
 	if (dev_priv->capabilities & SVGA_CAP_GMR2) {
@@ -1160,9 +1150,7 @@ static int vmw_master_set(struct drm_device *dev,
 		if (unlikely(ret != 0))
 			return ret;
 		vmw_kms_save_vga(dev_priv);
-		mutex_lock(&dev_priv->hw_mutex);
 		vmw_write(dev_priv, SVGA_REG_TRACES, 0);
-		mutex_unlock(&dev_priv->hw_mutex);
 	}
 
 	if (active) {
@@ -1196,9 +1184,7 @@ out_no_active_lock:
 	if (!dev_priv->enable_fb) {
 		vmw_kms_restore_vga(dev_priv);
 		vmw_3d_resource_dec(dev_priv, true);
-		mutex_lock(&dev_priv->hw_mutex);
 		vmw_write(dev_priv, SVGA_REG_TRACES, 1);
-		mutex_unlock(&dev_priv->hw_mutex);
 	}
 	return ret;
 }
@@ -1233,9 +1219,7 @@ static void vmw_master_drop(struct drm_device *dev,
 			DRM_ERROR("Unable to clean VRAM on master drop.\n");
 		vmw_kms_restore_vga(dev_priv);
 		vmw_3d_resource_dec(dev_priv, true);
-		mutex_lock(&dev_priv->hw_mutex);
 		vmw_write(dev_priv, SVGA_REG_TRACES, 1);
-		mutex_unlock(&dev_priv->hw_mutex);
 	}
 
 	dev_priv->active_master = &dev_priv->fbdev_master;
@@ -1367,10 +1351,8 @@ static void vmw_pm_complete(struct device *kdev)
 	struct drm_device *dev = pci_get_drvdata(pdev);
 	struct vmw_private *dev_priv = vmw_priv(dev);
 
-	mutex_lock(&dev_priv->hw_mutex);
 	vmw_write(dev_priv, SVGA_REG_ID, SVGA_ID_2);
 	(void) vmw_read(dev_priv, SVGA_REG_ID);
-	mutex_unlock(&dev_priv->hw_mutex);
 
 	/**
 	 * Reclaim 3d reference held by fbdev and potentially
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
index 4ee799b43d5d..d26a6daa9719 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
@@ -399,7 +399,8 @@ struct vmw_private {
 	uint32_t memory_size;
 	bool has_gmr;
 	bool has_mob;
-	struct mutex hw_mutex;
+	spinlock_t hw_lock;
+	spinlock_t cap_lock;
 
 	/*
 	 * VGA registers.
@@ -449,8 +450,9 @@ struct vmw_private {
 	atomic_t marker_seq;
 	wait_queue_head_t fence_queue;
 	wait_queue_head_t fifo_queue;
-	int fence_queue_waiters; /* Protected by hw_mutex */
-	int goal_queue_waiters; /* Protected by hw_mutex */
+	spinlock_t waiter_lock;
+	int fence_queue_waiters; /* Protected by waiter_lock */
+	int goal_queue_waiters; /* Protected by waiter_lock */
 	atomic_t fifo_queue_waiters;
 	uint32_t last_read_seqno;
 	spinlock_t irq_lock;
@@ -553,20 +555,35 @@ static inline struct vmw_master *vmw_master(struct drm_master *master)
 	return (struct vmw_master *) master->driver_priv;
 }
 
+/*
+ * The locking here is fine-grained, so that it is performed once
+ * for every read- and write operation. This is of course costly, but we
+ * don't perform much register access in the timing critical paths anyway.
+ * Instead we have the extra benefit of being sure that we don't forget
+ * the hw lock around register accesses.
+ */
 static inline void vmw_write(struct vmw_private *dev_priv,
 			     unsigned int offset, uint32_t value)
 {
+	unsigned long irq_flags;
+
+	spin_lock_irqsave(&dev_priv->hw_lock, irq_flags);
 	outl(offset, dev_priv->io_start + VMWGFX_INDEX_PORT);
 	outl(value, dev_priv->io_start + VMWGFX_VALUE_PORT);
+	spin_unlock_irqrestore(&dev_priv->hw_lock, irq_flags);
 }
 
 static inline uint32_t vmw_read(struct vmw_private *dev_priv,
 				unsigned int offset)
 {
-	uint32_t val;
+	unsigned long irq_flags;
+	u32 val;
 
+	spin_lock_irqsave(&dev_priv->hw_lock, irq_flags);
 	outl(offset, dev_priv->io_start + VMWGFX_INDEX_PORT);
 	val = inl(dev_priv->io_start + VMWGFX_VALUE_PORT);
+	spin_unlock_irqrestore(&dev_priv->hw_lock, irq_flags);
+
 	return val;
 }
 
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c b/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c
index b7594cb758af..945f1e0dad92 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_fence.c
@@ -35,7 +35,7 @@ struct vmw_fence_manager {
 	struct vmw_private *dev_priv;
 	spinlock_t lock;
 	struct list_head fence_list;
-	struct work_struct work, ping_work;
+	struct work_struct work;
 	u32 user_fence_size;
 	u32 fence_size;
 	u32 event_fence_action_size;
@@ -134,14 +134,6 @@ static const char *vmw_fence_get_timeline_name(struct fence *f)
 	return "svga";
 }
 
-static void vmw_fence_ping_func(struct work_struct *work)
-{
-	struct vmw_fence_manager *fman =
-		container_of(work, struct vmw_fence_manager, ping_work);
-
-	vmw_fifo_ping_host(fman->dev_priv, SVGA_SYNC_GENERIC);
-}
-
 static bool vmw_fence_enable_signaling(struct fence *f)
 {
 	struct vmw_fence_obj *fence =
@@ -155,11 +147,7 @@ static bool vmw_fence_enable_signaling(struct fence *f)
 	if (seqno - fence->base.seqno < VMW_FENCE_WRAP)
 		return false;
 
-	if (mutex_trylock(&dev_priv->hw_mutex)) {
-		vmw_fifo_ping_host_locked(dev_priv, SVGA_SYNC_GENERIC);
-		mutex_unlock(&dev_priv->hw_mutex);
-	} else
-		schedule_work(&fman->ping_work);
+	vmw_fifo_ping_host(dev_priv, SVGA_SYNC_GENERIC);
 
 	return true;
 }
@@ -305,7 +293,6 @@ struct vmw_fence_manager *vmw_fence_manager_init(struct vmw_private *dev_priv)
 	INIT_LIST_HEAD(&fman->fence_list);
 	INIT_LIST_HEAD(&fman->cleanup_list);
 	INIT_WORK(&fman->work, &vmw_fence_work_func);
-	INIT_WORK(&fman->ping_work, &vmw_fence_ping_func);
 	fman->fifo_down = true;
 	fman->user_fence_size = ttm_round_pot(sizeof(struct vmw_user_fence));
 	fman->fence_size = ttm_round_pot(sizeof(struct vmw_fence_obj));
@@ -323,7 +310,6 @@ void vmw_fence_manager_takedown(struct vmw_fence_manager *fman)
 	bool lists_empty;
 
 	(void) cancel_work_sync(&fman->work);
-	(void) cancel_work_sync(&fman->ping_work);
 
 	spin_lock_irqsave(&fman->lock, irq_flags);
 	lists_empty = list_empty(&fman->fence_list) &&
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_fifo.c b/drivers/gpu/drm/vmwgfx/vmwgfx_fifo.c
index 09e10aefcd8e..39f2b03888e7 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_fifo.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_fifo.c
@@ -44,10 +44,10 @@ bool vmw_fifo_have_3d(struct vmw_private *dev_priv)
 		if (!dev_priv->has_mob)
 			return false;
 
-		mutex_lock(&dev_priv->hw_mutex);
+		spin_lock(&dev_priv->cap_lock);
 		vmw_write(dev_priv, SVGA_REG_DEV_CAP, SVGA3D_DEVCAP_3D);
 		result = vmw_read(dev_priv, SVGA_REG_DEV_CAP);
-		mutex_unlock(&dev_priv->hw_mutex);
+		spin_unlock(&dev_priv->cap_lock);
 
 		return (result != 0);
 	}
@@ -120,7 +120,6 @@ int vmw_fifo_init(struct vmw_private *dev_priv, struct vmw_fifo_state *fifo)
 	DRM_INFO("height %d\n", vmw_read(dev_priv, SVGA_REG_HEIGHT));
 	DRM_INFO("bpp %d\n", vmw_read(dev_priv, SVGA_REG_BITS_PER_PIXEL));
 
-	mutex_lock(&dev_priv->hw_mutex);
 	dev_priv->enable_state = vmw_read(dev_priv, SVGA_REG_ENABLE);
 	dev_priv->config_done_state = vmw_read(dev_priv, SVGA_REG_CONFIG_DONE);
 	dev_priv->traces_state = vmw_read(dev_priv, SVGA_REG_TRACES);
@@ -143,7 +142,6 @@ int vmw_fifo_init(struct vmw_private *dev_priv, struct vmw_fifo_state *fifo)
 	mb();
 
 	vmw_write(dev_priv, SVGA_REG_CONFIG_DONE, 1);
-	mutex_unlock(&dev_priv->hw_mutex);
 
 	max = ioread32(fifo_mem + SVGA_FIFO_MAX);
 	min = ioread32(fifo_mem  + SVGA_FIFO_MIN);
@@ -160,31 +158,28 @@ int vmw_fifo_init(struct vmw_private *dev_priv, struct vmw_fifo_state *fifo)
 	return vmw_fifo_send_fence(dev_priv, &dummy);
 }
 
-void vmw_fifo_ping_host_locked(struct vmw_private *dev_priv, uint32_t reason)
+void vmw_fifo_ping_host(struct vmw_private *dev_priv, uint32_t reason)
 {
 	__le32 __iomem *fifo_mem = dev_priv->mmio_virt;
+	static DEFINE_SPINLOCK(ping_lock);
+	unsigned long irq_flags;
 
+	/*
+	 * The ping_lock is needed because we don't have an atomic
+	 * test-and-set of the SVGA_FIFO_BUSY register.
+	 */
+	spin_lock_irqsave(&ping_lock, irq_flags);
 	if (unlikely(ioread32(fifo_mem + SVGA_FIFO_BUSY) == 0)) {
 		iowrite32(1, fifo_mem + SVGA_FIFO_BUSY);
 		vmw_write(dev_priv, SVGA_REG_SYNC, reason);
 	}
-}
-
-void vmw_fifo_ping_host(struct vmw_private *dev_priv, uint32_t reason)
-{
-	mutex_lock(&dev_priv->hw_mutex);
-
-	vmw_fifo_ping_host_locked(dev_priv, reason);
-
-	mutex_unlock(&dev_priv->hw_mutex);
+	spin_unlock_irqrestore(&ping_lock, irq_flags);
 }
 
 void vmw_fifo_release(struct vmw_private *dev_priv, struct vmw_fifo_state *fifo)
 {
 	__le32 __iomem *fifo_mem = dev_priv->mmio_virt;
 
-	mutex_lock(&dev_priv->hw_mutex);
-
 	vmw_write(dev_priv, SVGA_REG_SYNC, SVGA_SYNC_GENERIC);
 	while (vmw_read(dev_priv, SVGA_REG_BUSY) != 0)
 		;
@@ -198,7 +193,6 @@ void vmw_fifo_release(struct vmw_private *dev_priv, struct vmw_fifo_state *fifo)
 	vmw_write(dev_priv, SVGA_REG_TRACES,
 		  dev_priv->traces_state);
 
-	mutex_unlock(&dev_priv->hw_mutex);
 	vmw_marker_queue_takedown(&fifo->marker_queue);
 
 	if (likely(fifo->static_buffer != NULL)) {
@@ -271,7 +265,7 @@ static int vmw_fifo_wait(struct vmw_private *dev_priv,
 		return vmw_fifo_wait_noirq(dev_priv, bytes,
 					   interruptible, timeout);
 
-	mutex_lock(&dev_priv->hw_mutex);
+	spin_lock(&dev_priv->waiter_lock);
 	if (atomic_add_return(1, &dev_priv->fifo_queue_waiters) > 0) {
 		spin_lock_irqsave(&dev_priv->irq_lock, irq_flags);
 		outl(SVGA_IRQFLAG_FIFO_PROGRESS,
@@ -280,7 +274,7 @@ static int vmw_fifo_wait(struct vmw_private *dev_priv,
 		vmw_write(dev_priv, SVGA_REG_IRQMASK, dev_priv->irq_mask);
 		spin_unlock_irqrestore(&dev_priv->irq_lock, irq_flags);
 	}
-	mutex_unlock(&dev_priv->hw_mutex);
+	spin_unlock(&dev_priv->waiter_lock);
 
 	if (interruptible)
 		ret = wait_event_interruptible_timeout
@@ -296,14 +290,14 @@ static int vmw_fifo_wait(struct vmw_private *dev_priv,
 	else if (likely(ret > 0))
 		ret = 0;
 
-	mutex_lock(&dev_priv->hw_mutex);
+	spin_lock(&dev_priv->waiter_lock);
 	if (atomic_dec_and_test(&dev_priv->fifo_queue_waiters)) {
 		spin_lock_irqsave(&dev_priv->irq_lock, irq_flags);
 		dev_priv->irq_mask &= ~SVGA_IRQFLAG_FIFO_PROGRESS;
 		vmw_write(dev_priv, SVGA_REG_IRQMASK, dev_priv->irq_mask);
 		spin_unlock_irqrestore(&dev_priv->irq_lock, irq_flags);
 	}
-	mutex_unlock(&dev_priv->hw_mutex);
+	spin_unlock(&dev_priv->waiter_lock);
 
 	return ret;
 }
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ioctl.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ioctl.c
index 37881ecf5d7a..69c8ce23123c 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_ioctl.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ioctl.c
@@ -135,13 +135,13 @@ static int vmw_fill_compat_cap(struct vmw_private *dev_priv, void *bounce,
 		(pair_offset + max_size * sizeof(SVGA3dCapPair)) / sizeof(u32);
 	compat_cap->header.type = SVGA3DCAPS_RECORD_DEVCAPS;
 
-	mutex_lock(&dev_priv->hw_mutex);
+	spin_lock(&dev_priv->cap_lock);
 	for (i = 0; i < max_size; ++i) {
 		vmw_write(dev_priv, SVGA_REG_DEV_CAP, i);
 		compat_cap->pairs[i][0] = i;
 		compat_cap->pairs[i][1] = vmw_read(dev_priv, SVGA_REG_DEV_CAP);
 	}
-	mutex_unlock(&dev_priv->hw_mutex);
+	spin_unlock(&dev_priv->cap_lock);
 
 	return 0;
 }
@@ -191,12 +191,12 @@ int vmw_get_cap_3d_ioctl(struct drm_device *dev, void *data,
 		if (num > SVGA3D_DEVCAP_MAX)
 			num = SVGA3D_DEVCAP_MAX;
 
-		mutex_lock(&dev_priv->hw_mutex);
+		spin_lock(&dev_priv->cap_lock);
 		for (i = 0; i < num; ++i) {
 			vmw_write(dev_priv, SVGA_REG_DEV_CAP, i);
 			*bounce32++ = vmw_read(dev_priv, SVGA_REG_DEV_CAP);
 		}
-		mutex_unlock(&dev_priv->hw_mutex);
+		spin_unlock(&dev_priv->cap_lock);
 	} else if (gb_objects) {
 		ret = vmw_fill_compat_cap(dev_priv, bounce, size);
 		if (unlikely(ret != 0))
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_irq.c b/drivers/gpu/drm/vmwgfx/vmwgfx_irq.c
index 0c423766c441..9fe9827ee499 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_irq.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_irq.c
@@ -62,13 +62,8 @@ irqreturn_t vmw_irq_handler(int irq, void *arg)
 
 static bool vmw_fifo_idle(struct vmw_private *dev_priv, uint32_t seqno)
 {
-	uint32_t busy;
 
-	mutex_lock(&dev_priv->hw_mutex);
-	busy = vmw_read(dev_priv, SVGA_REG_BUSY);
-	mutex_unlock(&dev_priv->hw_mutex);
-
-	return (busy == 0);
+	return (vmw_read(dev_priv, SVGA_REG_BUSY) == 0);
 }
 
 void vmw_update_seqno(struct vmw_private *dev_priv,
@@ -184,7 +179,7 @@ int vmw_fallback_wait(struct vmw_private *dev_priv,
 
 void vmw_seqno_waiter_add(struct vmw_private *dev_priv)
 {
-	mutex_lock(&dev_priv->hw_mutex);
+	spin_lock(&dev_priv->waiter_lock);
 	if (dev_priv->fence_queue_waiters++ == 0) {
 		unsigned long irq_flags;
 
@@ -195,12 +190,12 @@ void vmw_seqno_waiter_add(struct vmw_private *dev_priv)
 		vmw_write(dev_priv, SVGA_REG_IRQMASK, dev_priv->irq_mask);
 		spin_unlock_irqrestore(&dev_priv->irq_lock, irq_flags);
 	}
-	mutex_unlock(&dev_priv->hw_mutex);
+	spin_unlock(&dev_priv->waiter_lock);
 }
 
 void vmw_seqno_waiter_remove(struct vmw_private *dev_priv)
 {
-	mutex_lock(&dev_priv->hw_mutex);
+	spin_lock(&dev_priv->waiter_lock);
 	if (--dev_priv->fence_queue_waiters == 0) {
 		unsigned long irq_flags;
 
@@ -209,13 +204,13 @@ void vmw_seqno_waiter_remove(struct vmw_private *dev_priv)
 		vmw_write(dev_priv, SVGA_REG_IRQMASK, dev_priv->irq_mask);
 		spin_unlock_irqrestore(&dev_priv->irq_lock, irq_flags);
 	}
-	mutex_unlock(&dev_priv->hw_mutex);
+	spin_unlock(&dev_priv->waiter_lock);
 }
 
 
 void vmw_goal_waiter_add(struct vmw_private *dev_priv)
 {
-	mutex_lock(&dev_priv->hw_mutex);
+	spin_lock(&dev_priv->waiter_lock);
 	if (dev_priv->goal_queue_waiters++ == 0) {
 		unsigned long irq_flags;
 
@@ -226,12 +221,12 @@ void vmw_goal_waiter_add(struct vmw_private *dev_priv)
 		vmw_write(dev_priv, SVGA_REG_IRQMASK, dev_priv->irq_mask);
 		spin_unlock_irqrestore(&dev_priv->irq_lock, irq_flags);
 	}
-	mutex_unlock(&dev_priv->hw_mutex);
+	spin_unlock(&dev_priv->waiter_lock);
 }
 
 void vmw_goal_waiter_remove(struct vmw_private *dev_priv)
 {
-	mutex_lock(&dev_priv->hw_mutex);
+	spin_lock(&dev_priv->waiter_lock);
 	if (--dev_priv->goal_queue_waiters == 0) {
 		unsigned long irq_flags;
 
@@ -240,7 +235,7 @@ void vmw_goal_waiter_remove(struct vmw_private *dev_priv)
 		vmw_write(dev_priv, SVGA_REG_IRQMASK, dev_priv->irq_mask);
 		spin_unlock_irqrestore(&dev_priv->irq_lock, irq_flags);
 	}
-	mutex_unlock(&dev_priv->hw_mutex);
+	spin_unlock(&dev_priv->waiter_lock);
 }
 
 int vmw_wait_seqno(struct vmw_private *dev_priv,
@@ -315,9 +310,7 @@ void vmw_irq_uninstall(struct drm_device *dev)
 	if (!(dev_priv->capabilities & SVGA_CAP_IRQMASK))
 		return;
 
-	mutex_lock(&dev_priv->hw_mutex);
 	vmw_write(dev_priv, SVGA_REG_IRQMASK, 0);
-	mutex_unlock(&dev_priv->hw_mutex);
 
 	status = inl(dev_priv->io_start + VMWGFX_IRQSTATUS_PORT);
 	outl(status, dev_priv->io_start + VMWGFX_IRQSTATUS_PORT);
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c
index 3725b521d931..8725b79e7847 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c
@@ -1828,9 +1828,7 @@ vmw_du_connector_detect(struct drm_connector *connector, bool force)
 	struct vmw_private *dev_priv = vmw_priv(dev);
 	struct vmw_display_unit *du = vmw_connector_to_du(connector);
 
-	mutex_lock(&dev_priv->hw_mutex);
 	num_displays = vmw_read(dev_priv, SVGA_REG_NUM_DISPLAYS);
-	mutex_unlock(&dev_priv->hw_mutex);
 
 	return ((vmw_connector_to_du(connector)->unit < num_displays &&
 		 du->pref_active) ?