diff --git a/drivers/gpu/drm/nouveau/nouveau_backlight.c b/drivers/gpu/drm/nouveau/nouveau_backlight.c
index 406228f4a2a0..b14c81110575 100644
--- a/drivers/gpu/drm/nouveau/nouveau_backlight.c
+++ b/drivers/gpu/drm/nouveau/nouveau_backlight.c
@@ -31,6 +31,7 @@
  */
 
 #include <linux/backlight.h>
+#include <linux/acpi.h>
 
 #include "drmP.h"
 #include "nouveau_drv.h"
@@ -136,6 +137,14 @@ int nouveau_backlight_init(struct drm_device *dev)
 {
 	struct drm_nouveau_private *dev_priv = dev->dev_private;
 
+#ifdef CONFIG_ACPI
+	if (acpi_video_backlight_support()) {
+		NV_INFO(dev, "ACPI backlight interface available, "
+			     "not registering our own\n");
+		return 0;
+	}
+#endif
+
 	switch (dev_priv->card_type) {
 	case NV_40:
 		return nouveau_nv40_backlight_init(dev);
diff --git a/drivers/gpu/drm/nouveau/nouveau_bios.c b/drivers/gpu/drm/nouveau/nouveau_bios.c
index 5f21030a293b..b2293576f278 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bios.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bios.c
@@ -6829,7 +6829,7 @@ nouveau_bios_posted(struct drm_device *dev)
 	struct drm_nouveau_private *dev_priv = dev->dev_private;
 	unsigned htotal;
 
-	if (dev_priv->chipset >= NV_50) {
+	if (dev_priv->card_type >= NV_50) {
 		if (NVReadVgaCrtc(dev, 0, 0x00) == 0 &&
 		    NVReadVgaCrtc(dev, 0, 0x1a) == 0)
 			return false;
diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
index 80353e2b8409..c41e1c200ef5 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bo.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -143,8 +143,10 @@ nouveau_bo_new(struct drm_device *dev, struct nouveau_channel *chan,
 	nvbo->no_vm = no_vm;
 	nvbo->tile_mode = tile_mode;
 	nvbo->tile_flags = tile_flags;
+	nvbo->bo.bdev = &dev_priv->ttm.bdev;
 
-	nouveau_bo_fixup_align(dev, tile_mode, tile_flags, &align, &size);
+	nouveau_bo_fixup_align(dev, tile_mode, nouveau_bo_tile_layout(nvbo),
+			       &align, &size);
 	align >>= PAGE_SHIFT;
 
 	nouveau_bo_placement_set(nvbo, flags, 0);
@@ -176,6 +178,31 @@ set_placement_list(uint32_t *pl, unsigned *n, uint32_t type, uint32_t flags)
 		pl[(*n)++] = TTM_PL_FLAG_SYSTEM | flags;
 }
 
+static void
+set_placement_range(struct nouveau_bo *nvbo, uint32_t type)
+{
+	struct drm_nouveau_private *dev_priv = nouveau_bdev(nvbo->bo.bdev);
+
+	if (dev_priv->card_type == NV_10 &&
+	    nvbo->tile_mode && (type & TTM_PL_FLAG_VRAM)) {
+		/*
+		 * Make sure that the color and depth buffers are handled
+		 * by independent memory controller units. Up to a 9x
+		 * speed up when alpha-blending and depth-test are enabled
+		 * at the same time.
+		 */
+		int vram_pages = dev_priv->vram_size >> PAGE_SHIFT;
+
+		if (nvbo->tile_flags & NOUVEAU_GEM_TILE_ZETA) {
+			nvbo->placement.fpfn = vram_pages / 2;
+			nvbo->placement.lpfn = ~0;
+		} else {
+			nvbo->placement.fpfn = 0;
+			nvbo->placement.lpfn = vram_pages / 2;
+		}
+	}
+}
+
 void
 nouveau_bo_placement_set(struct nouveau_bo *nvbo, uint32_t type, uint32_t busy)
 {
@@ -190,6 +217,8 @@ nouveau_bo_placement_set(struct nouveau_bo *nvbo, uint32_t type, uint32_t busy)
 	pl->busy_placement = nvbo->busy_placements;
 	set_placement_list(nvbo->busy_placements, &pl->num_busy_placement,
 			   type | busy, flags);
+
+	set_placement_range(nvbo, type);
 }
 
 int
@@ -525,7 +554,8 @@ nv50_bo_move_m2mf(struct nouveau_channel *chan, struct ttm_buffer_object *bo,
 		stride  = 16 * 4;
 		height  = amount / stride;
 
-		if (new_mem->mem_type == TTM_PL_VRAM && nvbo->tile_flags) {
+		if (new_mem->mem_type == TTM_PL_VRAM &&
+		    nouveau_bo_tile_layout(nvbo)) {
 			ret = RING_SPACE(chan, 8);
 			if (ret)
 				return ret;
@@ -546,7 +576,8 @@ nv50_bo_move_m2mf(struct nouveau_channel *chan, struct ttm_buffer_object *bo,
 			BEGIN_RING(chan, NvSubM2MF, 0x0200, 1);
 			OUT_RING  (chan, 1);
 		}
-		if (old_mem->mem_type == TTM_PL_VRAM && nvbo->tile_flags) {
+		if (old_mem->mem_type == TTM_PL_VRAM &&
+		    nouveau_bo_tile_layout(nvbo)) {
 			ret = RING_SPACE(chan, 8);
 			if (ret)
 				return ret;
@@ -753,7 +784,8 @@ nouveau_bo_vm_bind(struct ttm_buffer_object *bo, struct ttm_mem_reg *new_mem,
 	if (dev_priv->card_type == NV_50) {
 		ret = nv50_mem_vm_bind_linear(dev,
 					      offset + dev_priv->vm_vram_base,
-					      new_mem->size, nvbo->tile_flags,
+					      new_mem->size,
+					      nouveau_bo_tile_layout(nvbo),
 					      offset);
 		if (ret)
 			return ret;
@@ -894,7 +926,8 @@ nouveau_ttm_fault_reserve_notify(struct ttm_buffer_object *bo)
 	 * nothing to do here.
 	 */
 	if (bo->mem.mem_type != TTM_PL_VRAM) {
-		if (dev_priv->card_type < NV_50 || !nvbo->tile_flags)
+		if (dev_priv->card_type < NV_50 ||
+		    !nouveau_bo_tile_layout(nvbo))
 			return 0;
 	}
 
diff --git a/drivers/gpu/drm/nouveau/nouveau_connector.c b/drivers/gpu/drm/nouveau/nouveau_connector.c
index 0871495096fa..52c356e9a3d1 100644
--- a/drivers/gpu/drm/nouveau/nouveau_connector.c
+++ b/drivers/gpu/drm/nouveau/nouveau_connector.c
@@ -281,7 +281,7 @@ detect_analog:
 	nv_encoder = find_encoder_by_type(connector, OUTPUT_ANALOG);
 	if (!nv_encoder && !nouveau_tv_disable)
 		nv_encoder = find_encoder_by_type(connector, OUTPUT_TV);
-	if (nv_encoder) {
+	if (nv_encoder && force) {
 		struct drm_encoder *encoder = to_drm_encoder(nv_encoder);
 		struct drm_encoder_helper_funcs *helper =
 						encoder->helper_private;
@@ -641,11 +641,28 @@ nouveau_connector_get_modes(struct drm_connector *connector)
 	return ret;
 }
 
+static unsigned
+get_tmds_link_bandwidth(struct drm_connector *connector)
+{
+	struct nouveau_connector *nv_connector = nouveau_connector(connector);
+	struct drm_nouveau_private *dev_priv = connector->dev->dev_private;
+	struct dcb_entry *dcb = nv_connector->detected_encoder->dcb;
+
+	if (dcb->location != DCB_LOC_ON_CHIP ||
+	    dev_priv->chipset >= 0x46)
+		return 165000;
+	else if (dev_priv->chipset >= 0x40)
+		return 155000;
+	else if (dev_priv->chipset >= 0x18)
+		return 135000;
+	else
+		return 112000;
+}
+
 static int
 nouveau_connector_mode_valid(struct drm_connector *connector,
 			     struct drm_display_mode *mode)
 {
-	struct drm_nouveau_private *dev_priv = connector->dev->dev_private;
 	struct nouveau_connector *nv_connector = nouveau_connector(connector);
 	struct nouveau_encoder *nv_encoder = nv_connector->detected_encoder;
 	struct drm_encoder *encoder = to_drm_encoder(nv_encoder);
@@ -663,11 +680,9 @@ nouveau_connector_mode_valid(struct drm_connector *connector,
 		max_clock = 400000;
 		break;
 	case OUTPUT_TMDS:
-		if ((dev_priv->card_type >= NV_50 && !nouveau_duallink) ||
-		    !nv_encoder->dcb->duallink_possible)
-			max_clock = 165000;
-		else
-			max_clock = 330000;
+		max_clock = get_tmds_link_bandwidth(connector);
+		if (nouveau_duallink && nv_encoder->dcb->duallink_possible)
+			max_clock *= 2;
 		break;
 	case OUTPUT_ANALOG:
 		max_clock = nv_encoder->dcb->crtconf.maxfreq;
@@ -709,44 +724,6 @@ nouveau_connector_best_encoder(struct drm_connector *connector)
 	return NULL;
 }
 
-void
-nouveau_connector_set_polling(struct drm_connector *connector)
-{
-	struct drm_device *dev = connector->dev;
-	struct drm_nouveau_private *dev_priv = dev->dev_private;
-	struct drm_crtc *crtc;
-	bool spare_crtc = false;
-
-	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head)
-		spare_crtc |= !crtc->enabled;
-
-	connector->polled = 0;
-
-	switch (connector->connector_type) {
-	case DRM_MODE_CONNECTOR_VGA:
-	case DRM_MODE_CONNECTOR_TV:
-		if (dev_priv->card_type >= NV_50 ||
-		    (nv_gf4_disp_arch(dev) && spare_crtc))
-			connector->polled = DRM_CONNECTOR_POLL_CONNECT;
-		break;
-
-	case DRM_MODE_CONNECTOR_DVII:
-	case DRM_MODE_CONNECTOR_DVID:
-	case DRM_MODE_CONNECTOR_HDMIA:
-	case DRM_MODE_CONNECTOR_DisplayPort:
-	case DRM_MODE_CONNECTOR_eDP:
-		if (dev_priv->card_type >= NV_50)
-			connector->polled = DRM_CONNECTOR_POLL_HPD;
-		else if (connector->connector_type == DRM_MODE_CONNECTOR_DVID ||
-			 spare_crtc)
-			connector->polled = DRM_CONNECTOR_POLL_CONNECT;
-		break;
-
-	default:
-		break;
-	}
-}
-
 static const struct drm_connector_helper_funcs
 nouveau_connector_helper_funcs = {
 	.get_modes = nouveau_connector_get_modes,
@@ -872,6 +849,7 @@ nouveau_connector_create(struct drm_device *dev, int index)
 					dev->mode_config.scaling_mode_property,
 					nv_connector->scaling_mode);
 		}
+		connector->polled = DRM_CONNECTOR_POLL_CONNECT;
 		/* fall-through */
 	case DCB_CONNECTOR_TV_0:
 	case DCB_CONNECTOR_TV_1:
@@ -888,11 +866,16 @@ nouveau_connector_create(struct drm_device *dev, int index)
 				dev->mode_config.dithering_mode_property,
 				nv_connector->use_dithering ?
 				DRM_MODE_DITHERING_ON : DRM_MODE_DITHERING_OFF);
+
+		if (dcb->type != DCB_CONNECTOR_LVDS) {
+			if (dev_priv->card_type >= NV_50)
+				connector->polled = DRM_CONNECTOR_POLL_HPD;
+			else
+				connector->polled = DRM_CONNECTOR_POLL_CONNECT;
+		}
 		break;
 	}
 
-	nouveau_connector_set_polling(connector);
-
 	drm_sysfs_connector_add(connector);
 	dcb->drm = connector;
 	return dcb->drm;
diff --git a/drivers/gpu/drm/nouveau/nouveau_connector.h b/drivers/gpu/drm/nouveau/nouveau_connector.h
index c21ed6b16f88..711b1e9203af 100644
--- a/drivers/gpu/drm/nouveau/nouveau_connector.h
+++ b/drivers/gpu/drm/nouveau/nouveau_connector.h
@@ -52,9 +52,6 @@ static inline struct nouveau_connector *nouveau_connector(
 struct drm_connector *
 nouveau_connector_create(struct drm_device *, int index);
 
-void
-nouveau_connector_set_polling(struct drm_connector *);
-
 int
 nouveau_connector_bpp(struct drm_connector *);
 
diff --git a/drivers/gpu/drm/nouveau/nouveau_drv.h b/drivers/gpu/drm/nouveau/nouveau_drv.h
index 3a07e580d27a..1c7db64c03bf 100644
--- a/drivers/gpu/drm/nouveau/nouveau_drv.h
+++ b/drivers/gpu/drm/nouveau/nouveau_drv.h
@@ -100,6 +100,9 @@ struct nouveau_bo {
 	int pin_refcnt;
 };
 
+#define nouveau_bo_tile_layout(nvbo)				\
+	((nvbo)->tile_flags & NOUVEAU_GEM_TILE_LAYOUT_MASK)
+
 static inline struct nouveau_bo *
 nouveau_bo(struct ttm_buffer_object *bo)
 {
@@ -304,6 +307,7 @@ struct nouveau_fifo_engine {
 	void (*destroy_context)(struct nouveau_channel *);
 	int  (*load_context)(struct nouveau_channel *);
 	int  (*unload_context)(struct drm_device *);
+	void (*tlb_flush)(struct drm_device *dev);
 };
 
 struct nouveau_pgraph_object_method {
@@ -336,6 +340,7 @@ struct nouveau_pgraph_engine {
 	void (*destroy_context)(struct nouveau_channel *);
 	int  (*load_context)(struct nouveau_channel *);
 	int  (*unload_context)(struct drm_device *);
+	void (*tlb_flush)(struct drm_device *dev);
 
 	void (*set_region_tiling)(struct drm_device *dev, int i, uint32_t addr,
 				  uint32_t size, uint32_t pitch);
@@ -485,13 +490,13 @@ enum nv04_fp_display_regs {
 };
 
 struct nv04_crtc_reg {
-	unsigned char MiscOutReg;     /* */
+	unsigned char MiscOutReg;
 	uint8_t CRTC[0xa0];
 	uint8_t CR58[0x10];
 	uint8_t Sequencer[5];
 	uint8_t Graphics[9];
 	uint8_t Attribute[21];
-	unsigned char DAC[768];       /* Internal Colorlookuptable */
+	unsigned char DAC[768];
 
 	/* PCRTC regs */
 	uint32_t fb_start;
@@ -539,43 +544,9 @@ struct nv04_output_reg {
 };
 
 struct nv04_mode_state {
-	uint32_t bpp;
-	uint32_t width;
-	uint32_t height;
-	uint32_t interlace;
-	uint32_t repaint0;
-	uint32_t repaint1;
-	uint32_t screen;
-	uint32_t scale;
-	uint32_t dither;
-	uint32_t extra;
-	uint32_t fifo;
-	uint32_t pixel;
-	uint32_t horiz;
-	int arbitration0;
-	int arbitration1;
-	uint32_t pll;
-	uint32_t pllB;
-	uint32_t vpll;
-	uint32_t vpll2;
-	uint32_t vpllB;
-	uint32_t vpll2B;
+	struct nv04_crtc_reg crtc_reg[2];
 	uint32_t pllsel;
 	uint32_t sel_clk;
-	uint32_t general;
-	uint32_t crtcOwner;
-	uint32_t head;
-	uint32_t head2;
-	uint32_t cursorConfig;
-	uint32_t cursor0;
-	uint32_t cursor1;
-	uint32_t cursor2;
-	uint32_t timingH;
-	uint32_t timingV;
-	uint32_t displayV;
-	uint32_t crtcSync;
-
-	struct nv04_crtc_reg crtc_reg[2];
 };
 
 enum nouveau_card_type {
@@ -613,6 +584,12 @@ struct drm_nouveau_private {
 	struct work_struct irq_work;
 	struct work_struct hpd_work;
 
+	struct {
+		spinlock_t lock;
+		uint32_t hpd0_bits;
+		uint32_t hpd1_bits;
+	} hpd_state;
+
 	struct list_head vbl_waiting;
 
 	struct {
@@ -1045,6 +1022,7 @@ extern int  nv50_fifo_create_context(struct nouveau_channel *);
 extern void nv50_fifo_destroy_context(struct nouveau_channel *);
 extern int  nv50_fifo_load_context(struct nouveau_channel *);
 extern int  nv50_fifo_unload_context(struct drm_device *);
+extern void nv50_fifo_tlb_flush(struct drm_device *dev);
 
 /* nvc0_fifo.c */
 extern int  nvc0_fifo_init(struct drm_device *);
@@ -1122,6 +1100,8 @@ extern int  nv50_graph_load_context(struct nouveau_channel *);
 extern int  nv50_graph_unload_context(struct drm_device *);
 extern void nv50_graph_context_switch(struct drm_device *);
 extern int  nv50_grctx_init(struct nouveau_grctx *);
+extern void nv50_graph_tlb_flush(struct drm_device *dev);
+extern void nv86_graph_tlb_flush(struct drm_device *dev);
 
 /* nvc0_graph.c */
 extern int  nvc0_graph_init(struct drm_device *);
@@ -1239,7 +1219,6 @@ extern u16 nouveau_bo_rd16(struct nouveau_bo *nvbo, unsigned index);
 extern void nouveau_bo_wr16(struct nouveau_bo *nvbo, unsigned index, u16 val);
 extern u32 nouveau_bo_rd32(struct nouveau_bo *nvbo, unsigned index);
 extern void nouveau_bo_wr32(struct nouveau_bo *nvbo, unsigned index, u32 val);
-extern int nouveau_bo_sync_gpu(struct nouveau_bo *, struct nouveau_channel *);
 
 /* nouveau_fence.c */
 struct nouveau_fence;
diff --git a/drivers/gpu/drm/nouveau/nouveau_fence.c b/drivers/gpu/drm/nouveau/nouveau_fence.c
index 441b12420bb1..ab1bbfbf266e 100644
--- a/drivers/gpu/drm/nouveau/nouveau_fence.c
+++ b/drivers/gpu/drm/nouveau/nouveau_fence.c
@@ -249,6 +249,7 @@ alloc_semaphore(struct drm_device *dev)
 {
 	struct drm_nouveau_private *dev_priv = dev->dev_private;
 	struct nouveau_semaphore *sema;
+	int ret;
 
 	if (!USE_SEMA(dev))
 		return NULL;
@@ -257,10 +258,14 @@ alloc_semaphore(struct drm_device *dev)
 	if (!sema)
 		goto fail;
 
+	ret = drm_mm_pre_get(&dev_priv->fence.heap);
+	if (ret)
+		goto fail;
+
 	spin_lock(&dev_priv->fence.lock);
 	sema->mem = drm_mm_search_free(&dev_priv->fence.heap, 4, 0, 0);
 	if (sema->mem)
-		sema->mem = drm_mm_get_block(sema->mem, 4, 0);
+		sema->mem = drm_mm_get_block_atomic(sema->mem, 4, 0);
 	spin_unlock(&dev_priv->fence.lock);
 
 	if (!sema->mem)
diff --git a/drivers/gpu/drm/nouveau/nouveau_gem.c b/drivers/gpu/drm/nouveau/nouveau_gem.c
index 5c4c929d7f74..9a1fdcf400c2 100644
--- a/drivers/gpu/drm/nouveau/nouveau_gem.c
+++ b/drivers/gpu/drm/nouveau/nouveau_gem.c
@@ -107,23 +107,29 @@ nouveau_gem_info(struct drm_gem_object *gem, struct drm_nouveau_gem_info *rep)
 }
 
 static bool
-nouveau_gem_tile_flags_valid(struct drm_device *dev, uint32_t tile_flags) {
-	switch (tile_flags) {
-	case 0x0000:
-	case 0x1800:
-	case 0x2800:
-	case 0x4800:
-	case 0x7000:
-	case 0x7400:
-	case 0x7a00:
-	case 0xe000:
-		break;
-	default:
-		NV_ERROR(dev, "bad page flags: 0x%08x\n", tile_flags);
-		return false;
+nouveau_gem_tile_flags_valid(struct drm_device *dev, uint32_t tile_flags)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	if (dev_priv->card_type >= NV_50) {
+		switch (tile_flags & NOUVEAU_GEM_TILE_LAYOUT_MASK) {
+		case 0x0000:
+		case 0x1800:
+		case 0x2800:
+		case 0x4800:
+		case 0x7000:
+		case 0x7400:
+		case 0x7a00:
+		case 0xe000:
+			return true;
+		}
+	} else {
+		if (!(tile_flags & NOUVEAU_GEM_TILE_LAYOUT_MASK))
+			return true;
 	}
 
-	return true;
+	NV_ERROR(dev, "bad page flags: 0x%08x\n", tile_flags);
+	return false;
 }
 
 int
diff --git a/drivers/gpu/drm/nouveau/nouveau_hw.c b/drivers/gpu/drm/nouveau/nouveau_hw.c
index bed669a54a2d..b9672a05c411 100644
--- a/drivers/gpu/drm/nouveau/nouveau_hw.c
+++ b/drivers/gpu/drm/nouveau/nouveau_hw.c
@@ -519,11 +519,11 @@ nouveau_hw_fix_bad_vpll(struct drm_device *dev, int head)
 
 	struct pll_lims pll_lim;
 	struct nouveau_pll_vals pv;
-	uint32_t pllreg = head ? NV_RAMDAC_VPLL2 : NV_PRAMDAC_VPLL_COEFF;
+	enum pll_types pll = head ? PLL_VPLL1 : PLL_VPLL0;
 
-	if (get_pll_limits(dev, pllreg, &pll_lim))
+	if (get_pll_limits(dev, pll, &pll_lim))
 		return;
-	nouveau_hw_get_pllvals(dev, pllreg, &pv);
+	nouveau_hw_get_pllvals(dev, pll, &pv);
 
 	if (pv.M1 >= pll_lim.vco1.min_m && pv.M1 <= pll_lim.vco1.max_m &&
 	    pv.N1 >= pll_lim.vco1.min_n && pv.N1 <= pll_lim.vco1.max_n &&
@@ -536,7 +536,7 @@ nouveau_hw_fix_bad_vpll(struct drm_device *dev, int head)
 	pv.M1 = pll_lim.vco1.max_m;
 	pv.N1 = pll_lim.vco1.min_n;
 	pv.log2P = pll_lim.max_usable_log2p;
-	nouveau_hw_setpll(dev, pllreg, &pv);
+	nouveau_hw_setpll(dev, pll_lim.reg, &pv);
 }
 
 /*
diff --git a/drivers/gpu/drm/nouveau/nouveau_hw.h b/drivers/gpu/drm/nouveau/nouveau_hw.h
index 869130f83602..2989090b9434 100644
--- a/drivers/gpu/drm/nouveau/nouveau_hw.h
+++ b/drivers/gpu/drm/nouveau/nouveau_hw.h
@@ -415,6 +415,25 @@ nv_fix_nv40_hw_cursor(struct drm_device *dev, int head)
 	NVWriteRAMDAC(dev, head, NV_PRAMDAC_CU_START_POS, curpos);
 }
 
+static inline void
+nv_set_crtc_base(struct drm_device *dev, int head, uint32_t offset)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+
+	NVWriteCRTC(dev, head, NV_PCRTC_START, offset);
+
+	if (dev_priv->card_type == NV_04) {
+		/*
+		 * Hilarious, the 24th bit doesn't want to stick to
+		 * PCRTC_START...
+		 */
+		int cre_heb = NVReadVgaCrtc(dev, head, NV_CIO_CRE_HEB__INDEX);
+
+		NVWriteVgaCrtc(dev, head, NV_CIO_CRE_HEB__INDEX,
+			       (cre_heb & ~0x40) | ((offset >> 18) & 0x40));
+	}
+}
+
 static inline void
 nv_show_cursor(struct drm_device *dev, int head, bool show)
 {
diff --git a/drivers/gpu/drm/nouveau/nouveau_i2c.c b/drivers/gpu/drm/nouveau/nouveau_i2c.c
index fdd7e3de79c8..cb389d014326 100644
--- a/drivers/gpu/drm/nouveau/nouveau_i2c.c
+++ b/drivers/gpu/drm/nouveau/nouveau_i2c.c
@@ -256,7 +256,7 @@ nouveau_i2c_find(struct drm_device *dev, int index)
 	if (index >= DCB_MAX_NUM_I2C_ENTRIES)
 		return NULL;
 
-	if (dev_priv->chipset >= NV_50 && (i2c->entry & 0x00000100)) {
+	if (dev_priv->card_type >= NV_50 && (i2c->entry & 0x00000100)) {
 		uint32_t reg = 0xe500, val;
 
 		if (i2c->port_type == 6) {
diff --git a/drivers/gpu/drm/nouveau/nouveau_irq.c b/drivers/gpu/drm/nouveau/nouveau_irq.c
index 6fd51a51c608..7bfd9e6c9d67 100644
--- a/drivers/gpu/drm/nouveau/nouveau_irq.c
+++ b/drivers/gpu/drm/nouveau/nouveau_irq.c
@@ -42,6 +42,13 @@
 #include "nouveau_connector.h"
 #include "nv50_display.h"
 
+static DEFINE_RATELIMIT_STATE(nouveau_ratelimit_state, 3 * HZ, 20);
+
+static int nouveau_ratelimit(void)
+{
+	return __ratelimit(&nouveau_ratelimit_state);
+}
+
 void
 nouveau_irq_preinstall(struct drm_device *dev)
 {
@@ -53,6 +60,7 @@ nouveau_irq_preinstall(struct drm_device *dev)
 	if (dev_priv->card_type >= NV_50) {
 		INIT_WORK(&dev_priv->irq_work, nv50_display_irq_handler_bh);
 		INIT_WORK(&dev_priv->hpd_work, nv50_display_irq_hotplug_bh);
+		spin_lock_init(&dev_priv->hpd_state.lock);
 		INIT_LIST_HEAD(&dev_priv->vbl_waiting);
 	}
 }
@@ -202,8 +210,8 @@ nouveau_fifo_irq_handler(struct drm_device *dev)
 		}
 
 		if (status & NV_PFIFO_INTR_DMA_PUSHER) {
-			u32 get = nv_rd32(dev, 0x003244);
-			u32 put = nv_rd32(dev, 0x003240);
+			u32 dma_get = nv_rd32(dev, 0x003244);
+			u32 dma_put = nv_rd32(dev, 0x003240);
 			u32 push = nv_rd32(dev, 0x003220);
 			u32 state = nv_rd32(dev, 0x003228);
 
@@ -213,16 +221,18 @@ nouveau_fifo_irq_handler(struct drm_device *dev)
 				u32 ib_get = nv_rd32(dev, 0x003334);
 				u32 ib_put = nv_rd32(dev, 0x003330);
 
-				NV_INFO(dev, "PFIFO_DMA_PUSHER - Ch %d Get 0x%02x%08x "
+				if (nouveau_ratelimit())
+					NV_INFO(dev, "PFIFO_DMA_PUSHER - Ch %d Get 0x%02x%08x "
 					     "Put 0x%02x%08x IbGet 0x%08x IbPut 0x%08x "
 					     "State 0x%08x Push 0x%08x\n",
-					chid, ho_get, get, ho_put, put, ib_get, ib_put,
-					state, push);
+						chid, ho_get, dma_get, ho_put,
+						dma_put, ib_get, ib_put, state,
+						push);
 
 				/* METHOD_COUNT, in DMA_STATE on earlier chipsets */
 				nv_wr32(dev, 0x003364, 0x00000000);
-				if (get != put || ho_get != ho_put) {
-					nv_wr32(dev, 0x003244, put);
+				if (dma_get != dma_put || ho_get != ho_put) {
+					nv_wr32(dev, 0x003244, dma_put);
 					nv_wr32(dev, 0x003328, ho_put);
 				} else
 				if (ib_get != ib_put) {
@@ -231,10 +241,10 @@ nouveau_fifo_irq_handler(struct drm_device *dev)
 			} else {
 				NV_INFO(dev, "PFIFO_DMA_PUSHER - Ch %d Get 0x%08x "
 					     "Put 0x%08x State 0x%08x Push 0x%08x\n",
-					chid, get, put, state, push);
+					chid, dma_get, dma_put, state, push);
 
-				if (get != put)
-					nv_wr32(dev, 0x003244, put);
+				if (dma_get != dma_put)
+					nv_wr32(dev, 0x003244, dma_put);
 			}
 
 			nv_wr32(dev, 0x003228, 0x00000000);
@@ -266,8 +276,9 @@ nouveau_fifo_irq_handler(struct drm_device *dev)
 		}
 
 		if (status) {
-			NV_INFO(dev, "PFIFO_INTR 0x%08x - Ch %d\n",
-				status, chid);
+			if (nouveau_ratelimit())
+				NV_INFO(dev, "PFIFO_INTR 0x%08x - Ch %d\n",
+					status, chid);
 			nv_wr32(dev, NV03_PFIFO_INTR_0, status);
 			status = 0;
 		}
@@ -544,13 +555,6 @@ nouveau_pgraph_intr_notify(struct drm_device *dev, uint32_t nsource)
 		nouveau_graph_dump_trap_info(dev, "PGRAPH_NOTIFY", &trap);
 }
 
-static DEFINE_RATELIMIT_STATE(nouveau_ratelimit_state, 3 * HZ, 20);
-
-static int nouveau_ratelimit(void)
-{
-	return __ratelimit(&nouveau_ratelimit_state);
-}
-
 
 static inline void
 nouveau_pgraph_intr_error(struct drm_device *dev, uint32_t nsource)
diff --git a/drivers/gpu/drm/nouveau/nouveau_mem.c b/drivers/gpu/drm/nouveau/nouveau_mem.c
index a163c7c612e7..fe4a30dc4b42 100644
--- a/drivers/gpu/drm/nouveau/nouveau_mem.c
+++ b/drivers/gpu/drm/nouveau/nouveau_mem.c
@@ -33,9 +33,9 @@
 #include "drmP.h"
 #include "drm.h"
 #include "drm_sarea.h"
-#include "nouveau_drv.h"
 
-#define MIN(a,b) a < b ? a : b
+#include "nouveau_drv.h"
+#include "nouveau_pm.h"
 
 /*
  * NV10-NV40 tiling helpers
@@ -175,11 +175,10 @@ nv50_mem_vm_bind_linear(struct drm_device *dev, uint64_t virt, uint32_t size,
 			}
 		}
 	}
-	dev_priv->engine.instmem.flush(dev);
 
-	nv50_vm_flush(dev, 5);
-	nv50_vm_flush(dev, 0);
-	nv50_vm_flush(dev, 4);
+	dev_priv->engine.instmem.flush(dev);
+	dev_priv->engine.fifo.tlb_flush(dev);
+	dev_priv->engine.graph.tlb_flush(dev);
 	nv50_vm_flush(dev, 6);
 	return 0;
 }
@@ -209,11 +208,10 @@ nv50_mem_vm_unbind(struct drm_device *dev, uint64_t virt, uint32_t size)
 			pte++;
 		}
 	}
-	dev_priv->engine.instmem.flush(dev);
 
-	nv50_vm_flush(dev, 5);
-	nv50_vm_flush(dev, 0);
-	nv50_vm_flush(dev, 4);
+	dev_priv->engine.instmem.flush(dev);
+	dev_priv->engine.fifo.tlb_flush(dev);
+	dev_priv->engine.graph.tlb_flush(dev);
 	nv50_vm_flush(dev, 6);
 }
 
@@ -653,6 +651,7 @@ nouveau_mem_gart_init(struct drm_device *dev)
 void
 nouveau_mem_timing_init(struct drm_device *dev)
 {
+	/* cards < NVC0 only */
 	struct drm_nouveau_private *dev_priv = dev->dev_private;
 	struct nouveau_pm_engine *pm = &dev_priv->engine.pm;
 	struct nouveau_pm_memtimings *memtimings = &pm->memtimings;
@@ -719,14 +718,14 @@ nouveau_mem_timing_init(struct drm_device *dev)
 		tUNK_19 = 1;
 		tUNK_20 = 0;
 		tUNK_21 = 0;
-		switch (MIN(recordlen,21)) {
-		case 21:
+		switch (min(recordlen, 22)) {
+		case 22:
 			tUNK_21 = entry[21];
-		case 20:
+		case 21:
 			tUNK_20 = entry[20];
-		case 19:
+		case 20:
 			tUNK_19 = entry[19];
-		case 18:
+		case 19:
 			tUNK_18 = entry[18];
 		default:
 			tUNK_0  = entry[0];
@@ -756,24 +755,30 @@ nouveau_mem_timing_init(struct drm_device *dev)
 		timing->reg_100228 = (tUNK_12 << 16 | tUNK_11 << 8 | tUNK_10);
 		if(recordlen > 19) {
 			timing->reg_100228 += (tUNK_19 - 1) << 24;
-		} else {
+		}/* I cannot back-up this else-statement right now
+			 else {
 			timing->reg_100228 += tUNK_12 << 24;
-		}
+		}*/
 
 		/* XXX: reg_10022c */
+		timing->reg_10022c = tUNK_2 - 1;
 
 		timing->reg_100230 = (tUNK_20 << 24 | tUNK_21 << 16 |
 				      tUNK_13 << 8  | tUNK_13);
 
 		/* XXX: +6? */
 		timing->reg_100234 = (tRAS << 24 | (tUNK_19 + 6) << 8 | tRC);
-		if(tUNK_10 > tUNK_11) {
-			timing->reg_100234 += tUNK_10 << 16;
-		} else {
-			timing->reg_100234 += tUNK_11 << 16;
+		timing->reg_100234 += max(tUNK_10,tUNK_11) << 16;
+
+		/* XXX; reg_100238, reg_10023c
+		 * reg: 0x00??????
+		 * reg_10023c:
+		 *      0 for pre-NV50 cards
+		 *      0x????0202 for NV50+ cards (empirical evidence) */
+		if(dev_priv->card_type >= NV_50) {
+			timing->reg_10023c = 0x202;
 		}
 
-		/* XXX; reg_100238, reg_10023c */
 		NV_DEBUG(dev, "Entry %d: 220: %08x %08x %08x %08x\n", i,
 			 timing->reg_100220, timing->reg_100224,
 			 timing->reg_100228, timing->reg_10022c);
diff --git a/drivers/gpu/drm/nouveau/nouveau_object.c b/drivers/gpu/drm/nouveau/nouveau_object.c
index 896cf8634144..dd572adca02a 100644
--- a/drivers/gpu/drm/nouveau/nouveau_object.c
+++ b/drivers/gpu/drm/nouveau/nouveau_object.c
@@ -129,7 +129,7 @@ nouveau_gpuobj_new(struct drm_device *dev, struct nouveau_channel *chan,
 			if (ramin == NULL) {
 				spin_unlock(&dev_priv->ramin_lock);
 				nouveau_gpuobj_ref(NULL, &gpuobj);
-				return ret;
+				return -ENOMEM;
 			}
 
 			ramin = drm_mm_get_block_atomic(ramin, size, align);
diff --git a/drivers/gpu/drm/nouveau/nouveau_pm.c b/drivers/gpu/drm/nouveau/nouveau_pm.c
index 1c99c55d6d46..9f7b158f5825 100644
--- a/drivers/gpu/drm/nouveau/nouveau_pm.c
+++ b/drivers/gpu/drm/nouveau/nouveau_pm.c
@@ -284,6 +284,7 @@ nouveau_sysfs_fini(struct drm_device *dev)
 	}
 }
 
+#ifdef CONFIG_HWMON
 static ssize_t
 nouveau_hwmon_show_temp(struct device *d, struct device_attribute *a, char *buf)
 {
@@ -395,10 +396,12 @@ static struct attribute *hwmon_attributes[] = {
 static const struct attribute_group hwmon_attrgroup = {
 	.attrs = hwmon_attributes,
 };
+#endif
 
 static int
 nouveau_hwmon_init(struct drm_device *dev)
 {
+#ifdef CONFIG_HWMON
 	struct drm_nouveau_private *dev_priv = dev->dev_private;
 	struct nouveau_pm_engine *pm = &dev_priv->engine.pm;
 	struct device *hwmon_dev;
@@ -425,13 +428,14 @@ nouveau_hwmon_init(struct drm_device *dev)
 	}
 
 	pm->hwmon = hwmon_dev;
-
+#endif
 	return 0;
 }
 
 static void
 nouveau_hwmon_fini(struct drm_device *dev)
 {
+#ifdef CONFIG_HWMON
 	struct drm_nouveau_private *dev_priv = dev->dev_private;
 	struct nouveau_pm_engine *pm = &dev_priv->engine.pm;
 
@@ -439,6 +443,7 @@ nouveau_hwmon_fini(struct drm_device *dev)
 		sysfs_remove_group(&pm->hwmon->kobj, &hwmon_attrgroup);
 		hwmon_device_unregister(pm->hwmon);
 	}
+#endif
 }
 
 int
diff --git a/drivers/gpu/drm/nouveau/nouveau_ramht.c b/drivers/gpu/drm/nouveau/nouveau_ramht.c
index 7f16697cc96c..2d8580927ca4 100644
--- a/drivers/gpu/drm/nouveau/nouveau_ramht.c
+++ b/drivers/gpu/drm/nouveau/nouveau_ramht.c
@@ -153,26 +153,42 @@ nouveau_ramht_insert(struct nouveau_channel *chan, u32 handle,
 	return -ENOMEM;
 }
 
+static struct nouveau_ramht_entry *
+nouveau_ramht_remove_entry(struct nouveau_channel *chan, u32 handle)
+{
+	struct nouveau_ramht *ramht = chan ? chan->ramht : NULL;
+	struct nouveau_ramht_entry *entry;
+	unsigned long flags;
+
+	if (!ramht)
+		return NULL;
+
+	spin_lock_irqsave(&ramht->lock, flags);
+	list_for_each_entry(entry, &ramht->entries, head) {
+		if (entry->channel == chan &&
+		    (!handle || entry->handle == handle)) {
+			list_del(&entry->head);
+			spin_unlock_irqrestore(&ramht->lock, flags);
+
+			return entry;
+		}
+	}
+	spin_unlock_irqrestore(&ramht->lock, flags);
+
+	return NULL;
+}
+
 static void
-nouveau_ramht_remove_locked(struct nouveau_channel *chan, u32 handle)
+nouveau_ramht_remove_hash(struct nouveau_channel *chan, u32 handle)
 {
 	struct drm_device *dev = chan->dev;
 	struct drm_nouveau_private *dev_priv = dev->dev_private;
 	struct nouveau_instmem_engine *instmem = &dev_priv->engine.instmem;
 	struct nouveau_gpuobj *ramht = chan->ramht->gpuobj;
-	struct nouveau_ramht_entry *entry, *tmp;
+	unsigned long flags;
 	u32 co, ho;
 
-	list_for_each_entry_safe(entry, tmp, &chan->ramht->entries, head) {
-		if (entry->channel != chan || entry->handle != handle)
-			continue;
-
-		nouveau_gpuobj_ref(NULL, &entry->gpuobj);
-		list_del(&entry->head);
-		kfree(entry);
-		break;
-	}
-
+	spin_lock_irqsave(&chan->ramht->lock, flags);
 	co = ho = nouveau_ramht_hash_handle(chan, handle);
 	do {
 		if (nouveau_ramht_entry_valid(dev, ramht, co) &&
@@ -184,7 +200,7 @@ nouveau_ramht_remove_locked(struct nouveau_channel *chan, u32 handle)
 			nv_wo32(ramht, co + 0, 0x00000000);
 			nv_wo32(ramht, co + 4, 0x00000000);
 			instmem->flush(dev);
-			return;
+			goto out;
 		}
 
 		co += 8;
@@ -194,17 +210,22 @@ nouveau_ramht_remove_locked(struct nouveau_channel *chan, u32 handle)
 
 	NV_ERROR(dev, "RAMHT entry not found. ch=%d, handle=0x%08x\n",
 		 chan->id, handle);
+out:
+	spin_unlock_irqrestore(&chan->ramht->lock, flags);
 }
 
 void
 nouveau_ramht_remove(struct nouveau_channel *chan, u32 handle)
 {
-	struct nouveau_ramht *ramht = chan->ramht;
-	unsigned long flags;
+	struct nouveau_ramht_entry *entry;
 
-	spin_lock_irqsave(&ramht->lock, flags);
-	nouveau_ramht_remove_locked(chan, handle);
-	spin_unlock_irqrestore(&ramht->lock, flags);
+	entry = nouveau_ramht_remove_entry(chan, handle);
+	if (!entry)
+		return;
+
+	nouveau_ramht_remove_hash(chan, entry->handle);
+	nouveau_gpuobj_ref(NULL, &entry->gpuobj);
+	kfree(entry);
 }
 
 struct nouveau_gpuobj *
@@ -265,23 +286,19 @@ void
 nouveau_ramht_ref(struct nouveau_ramht *ref, struct nouveau_ramht **ptr,
 		  struct nouveau_channel *chan)
 {
-	struct nouveau_ramht_entry *entry, *tmp;
+	struct nouveau_ramht_entry *entry;
 	struct nouveau_ramht *ramht;
-	unsigned long flags;
 
 	if (ref)
 		kref_get(&ref->refcount);
 
 	ramht = *ptr;
 	if (ramht) {
-		spin_lock_irqsave(&ramht->lock, flags);
-		list_for_each_entry_safe(entry, tmp, &ramht->entries, head) {
-			if (entry->channel != chan)
-				continue;
-
-			nouveau_ramht_remove_locked(chan, entry->handle);
+		while ((entry = nouveau_ramht_remove_entry(chan, 0))) {
+			nouveau_ramht_remove_hash(chan, entry->handle);
+			nouveau_gpuobj_ref(NULL, &entry->gpuobj);
+			kfree(entry);
 		}
-		spin_unlock_irqrestore(&ramht->lock, flags);
 
 		kref_put(&ramht->refcount, nouveau_ramht_del);
 	}
diff --git a/drivers/gpu/drm/nouveau/nouveau_sgdma.c b/drivers/gpu/drm/nouveau/nouveau_sgdma.c
index 288bacac7e5a..d4ac97007038 100644
--- a/drivers/gpu/drm/nouveau/nouveau_sgdma.c
+++ b/drivers/gpu/drm/nouveau/nouveau_sgdma.c
@@ -120,8 +120,8 @@ nouveau_sgdma_bind(struct ttm_backend *be, struct ttm_mem_reg *mem)
 	dev_priv->engine.instmem.flush(nvbe->dev);
 
 	if (dev_priv->card_type == NV_50) {
-		nv50_vm_flush(dev, 5); /* PGRAPH */
-		nv50_vm_flush(dev, 0); /* PFIFO */
+		dev_priv->engine.fifo.tlb_flush(dev);
+		dev_priv->engine.graph.tlb_flush(dev);
 	}
 
 	nvbe->bound = true;
@@ -162,8 +162,8 @@ nouveau_sgdma_unbind(struct ttm_backend *be)
 	dev_priv->engine.instmem.flush(nvbe->dev);
 
 	if (dev_priv->card_type == NV_50) {
-		nv50_vm_flush(dev, 5);
-		nv50_vm_flush(dev, 0);
+		dev_priv->engine.fifo.tlb_flush(dev);
+		dev_priv->engine.graph.tlb_flush(dev);
 	}
 
 	nvbe->bound = false;
@@ -224,7 +224,11 @@ nouveau_sgdma_init(struct drm_device *dev)
 	int i, ret;
 
 	if (dev_priv->card_type < NV_50) {
-		aper_size = (64 * 1024 * 1024);
+		if(dev_priv->ramin_rsvd_vram < 2 * 1024 * 1024)
+			aper_size = 64 * 1024 * 1024;
+		else
+			aper_size = 512 * 1024 * 1024;
+
 		obj_size  = (aper_size >> NV_CTXDMA_PAGE_SHIFT) * 4;
 		obj_size += 8; /* ctxdma header */
 	} else {
diff --git a/drivers/gpu/drm/nouveau/nouveau_state.c b/drivers/gpu/drm/nouveau/nouveau_state.c
index ed7757f14083..049f755567e5 100644
--- a/drivers/gpu/drm/nouveau/nouveau_state.c
+++ b/drivers/gpu/drm/nouveau/nouveau_state.c
@@ -354,6 +354,15 @@ static int nouveau_init_engine_ptrs(struct drm_device *dev)
 		engine->graph.destroy_context	= nv50_graph_destroy_context;
 		engine->graph.load_context	= nv50_graph_load_context;
 		engine->graph.unload_context	= nv50_graph_unload_context;
+		if (dev_priv->chipset != 0x86)
+			engine->graph.tlb_flush	= nv50_graph_tlb_flush;
+		else {
+			/* from what i can see nvidia do this on every
+			 * pre-NVA3 board except NVAC, but, we've only
+			 * ever seen problems on NV86
+			 */
+			engine->graph.tlb_flush	= nv86_graph_tlb_flush;
+		}
 		engine->fifo.channels		= 128;
 		engine->fifo.init		= nv50_fifo_init;
 		engine->fifo.takedown		= nv50_fifo_takedown;
@@ -365,6 +374,7 @@ static int nouveau_init_engine_ptrs(struct drm_device *dev)
 		engine->fifo.destroy_context	= nv50_fifo_destroy_context;
 		engine->fifo.load_context	= nv50_fifo_load_context;
 		engine->fifo.unload_context	= nv50_fifo_unload_context;
+		engine->fifo.tlb_flush		= nv50_fifo_tlb_flush;
 		engine->display.early_init	= nv50_display_early_init;
 		engine->display.late_takedown	= nv50_display_late_takedown;
 		engine->display.create		= nv50_display_create;
@@ -1041,6 +1051,9 @@ int nouveau_ioctl_getparam(struct drm_device *dev, void *data,
 	case NOUVEAU_GETPARAM_PTIMER_TIME:
 		getparam->value = dev_priv->engine.timer.read(dev);
 		break;
+	case NOUVEAU_GETPARAM_HAS_BO_USAGE:
+		getparam->value = 1;
+		break;
 	case NOUVEAU_GETPARAM_GRAPH_UNITS:
 		/* NV40 and NV50 versions are quite different, but register
 		 * address is the same. User is supposed to know the card
@@ -1051,7 +1064,7 @@ int nouveau_ioctl_getparam(struct drm_device *dev, void *data,
 		}
 		/* FALLTHRU */
 	default:
-		NV_ERROR(dev, "unknown parameter %lld\n", getparam->param);
+		NV_DEBUG(dev, "unknown parameter %lld\n", getparam->param);
 		return -EINVAL;
 	}
 
@@ -1066,7 +1079,7 @@ nouveau_ioctl_setparam(struct drm_device *dev, void *data,
 
 	switch (setparam->param) {
 	default:
-		NV_ERROR(dev, "unknown parameter %lld\n", setparam->param);
+		NV_DEBUG(dev, "unknown parameter %lld\n", setparam->param);
 		return -EINVAL;
 	}
 
diff --git a/drivers/gpu/drm/nouveau/nouveau_temp.c b/drivers/gpu/drm/nouveau/nouveau_temp.c
index 16bbbf1eff63..7ecc4adc1e45 100644
--- a/drivers/gpu/drm/nouveau/nouveau_temp.c
+++ b/drivers/gpu/drm/nouveau/nouveau_temp.c
@@ -191,7 +191,7 @@ nv40_temp_get(struct drm_device *dev)
 	int offset = sensor->offset_mult / sensor->offset_div;
 	int core_temp;
 
-	if (dev_priv->chipset >= 0x50) {
+	if (dev_priv->card_type >= NV_50) {
 		core_temp = nv_rd32(dev, 0x20008);
 	} else {
 		core_temp = nv_rd32(dev, 0x0015b4) & 0x1fff;
diff --git a/drivers/gpu/drm/nouveau/nv04_crtc.c b/drivers/gpu/drm/nouveau/nv04_crtc.c
index c71abc2a34d5..40e180741629 100644
--- a/drivers/gpu/drm/nouveau/nv04_crtc.c
+++ b/drivers/gpu/drm/nouveau/nv04_crtc.c
@@ -158,7 +158,6 @@ nv_crtc_dpms(struct drm_crtc *crtc, int mode)
 {
 	struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
 	struct drm_device *dev = crtc->dev;
-	struct drm_connector *connector;
 	unsigned char seq1 = 0, crtc17 = 0;
 	unsigned char crtc1A;
 
@@ -213,10 +212,6 @@ nv_crtc_dpms(struct drm_crtc *crtc, int mode)
 	NVVgaSeqReset(dev, nv_crtc->index, false);
 
 	NVWriteVgaCrtc(dev, nv_crtc->index, NV_CIO_CRE_RPC1_INDEX, crtc1A);
-
-	/* Update connector polling modes */
-	list_for_each_entry(connector, &dev->mode_config.connector_list, head)
-		nouveau_connector_set_polling(connector);
 }
 
 static bool
@@ -831,7 +826,7 @@ nv04_crtc_do_mode_set_base(struct drm_crtc *crtc,
 	/* Update the framebuffer location. */
 	regp->fb_start = nv_crtc->fb.offset & ~3;
 	regp->fb_start += (y * drm_fb->pitch) + (x * drm_fb->bits_per_pixel / 8);
-	NVWriteCRTC(dev, nv_crtc->index, NV_PCRTC_START, regp->fb_start);
+	nv_set_crtc_base(dev, nv_crtc->index, regp->fb_start);
 
 	/* Update the arbitration parameters. */
 	nouveau_calc_arb(dev, crtc->mode.clock, drm_fb->bits_per_pixel,
diff --git a/drivers/gpu/drm/nouveau/nv04_dfp.c b/drivers/gpu/drm/nouveau/nv04_dfp.c
index c936403b26e2..ef23550407b5 100644
--- a/drivers/gpu/drm/nouveau/nv04_dfp.c
+++ b/drivers/gpu/drm/nouveau/nv04_dfp.c
@@ -185,14 +185,15 @@ static bool nv04_dfp_mode_fixup(struct drm_encoder *encoder,
 	struct nouveau_encoder *nv_encoder = nouveau_encoder(encoder);
 	struct nouveau_connector *nv_connector = nouveau_encoder_connector_get(nv_encoder);
 
-	/* For internal panels and gpu scaling on DVI we need the native mode */
-	if (nv_connector->scaling_mode != DRM_MODE_SCALE_NONE) {
-		if (!nv_connector->native_mode)
-			return false;
+	if (!nv_connector->native_mode ||
+	    nv_connector->scaling_mode == DRM_MODE_SCALE_NONE ||
+	    mode->hdisplay > nv_connector->native_mode->hdisplay ||
+	    mode->vdisplay > nv_connector->native_mode->vdisplay) {
+		nv_encoder->mode = *adjusted_mode;
+
+	} else {
 		nv_encoder->mode = *nv_connector->native_mode;
 		adjusted_mode->clock = nv_connector->native_mode->clock;
-	} else {
-		nv_encoder->mode = *adjusted_mode;
 	}
 
 	return true;
diff --git a/drivers/gpu/drm/nouveau/nv04_pm.c b/drivers/gpu/drm/nouveau/nv04_pm.c
index 6a6eb697d38e..eb1c70dd82ed 100644
--- a/drivers/gpu/drm/nouveau/nv04_pm.c
+++ b/drivers/gpu/drm/nouveau/nv04_pm.c
@@ -76,6 +76,15 @@ nv04_pm_clock_set(struct drm_device *dev, void *pre_state)
 		reg += 4;
 
 	nouveau_hw_setpll(dev, reg, &state->calc);
+
+	if (dev_priv->card_type < NV_30 && reg == NV_PRAMDAC_MPLL_COEFF) {
+		if (dev_priv->card_type == NV_20)
+			nv_mask(dev, 0x1002c4, 0, 1 << 20);
+
+		/* Reset the DLLs */
+		nv_mask(dev, 0x1002c0, 0, 1 << 8);
+	}
+
 	kfree(state);
 }
 
diff --git a/drivers/gpu/drm/nouveau/nv50_calc.c b/drivers/gpu/drm/nouveau/nv50_calc.c
index 2cdc2bfe7179..de81151648f8 100644
--- a/drivers/gpu/drm/nouveau/nv50_calc.c
+++ b/drivers/gpu/drm/nouveau/nv50_calc.c
@@ -51,24 +51,28 @@ nv50_calc_pll2(struct drm_device *dev, struct pll_lims *pll, int clk,
 	       int *N, int *fN, int *M, int *P)
 {
 	fixed20_12 fb_div, a, b;
+	u32 refclk = pll->refclk / 10;
+	u32 max_vco_freq = pll->vco1.maxfreq / 10;
+	u32 max_vco_inputfreq = pll->vco1.max_inputfreq / 10;
+	clk /= 10;
 
-	*P = pll->vco1.maxfreq / clk;
+	*P = max_vco_freq / clk;
 	if (*P > pll->max_p)
 		*P = pll->max_p;
 	if (*P < pll->min_p)
 		*P = pll->min_p;
 
-	/* *M = ceil(refclk / pll->vco.max_inputfreq); */
-	a.full = dfixed_const(pll->refclk);
-	b.full = dfixed_const(pll->vco1.max_inputfreq);
+	/* *M = floor((refclk + max_vco_inputfreq) / max_vco_inputfreq); */
+	a.full = dfixed_const(refclk + max_vco_inputfreq);
+	b.full = dfixed_const(max_vco_inputfreq);
 	a.full = dfixed_div(a, b);
-	a.full = dfixed_ceil(a);
+	a.full = dfixed_floor(a);
 	*M = dfixed_trunc(a);
 
 	/* fb_div = (vco * *M) / refclk; */
 	fb_div.full = dfixed_const(clk * *P);
 	fb_div.full = dfixed_mul(fb_div, a);
-	a.full = dfixed_const(pll->refclk);
+	a.full = dfixed_const(refclk);
 	fb_div.full = dfixed_div(fb_div, a);
 
 	/* *N = floor(fb_div); */
diff --git a/drivers/gpu/drm/nouveau/nv50_crtc.c b/drivers/gpu/drm/nouveau/nv50_crtc.c
index 16380d52cd88..56476d0c6de8 100644
--- a/drivers/gpu/drm/nouveau/nv50_crtc.c
+++ b/drivers/gpu/drm/nouveau/nv50_crtc.c
@@ -546,7 +546,7 @@ nv50_crtc_do_mode_set_base(struct drm_crtc *crtc,
 	}
 
 	nv_crtc->fb.offset = fb->nvbo->bo.offset - dev_priv->vm_vram_base;
-	nv_crtc->fb.tile_flags = fb->nvbo->tile_flags;
+	nv_crtc->fb.tile_flags = nouveau_bo_tile_layout(fb->nvbo);
 	nv_crtc->fb.cpp = drm_fb->bits_per_pixel / 8;
 	if (!nv_crtc->fb.blanked && dev_priv->chipset != 0x50) {
 		ret = RING_SPACE(evo, 2);
@@ -578,7 +578,7 @@ nv50_crtc_do_mode_set_base(struct drm_crtc *crtc,
 				  fb->nvbo->tile_mode);
 	}
 	if (dev_priv->chipset == 0x50)
-		OUT_RING(evo, (fb->nvbo->tile_flags << 8) | format);
+		OUT_RING(evo, (nv_crtc->fb.tile_flags << 8) | format);
 	else
 		OUT_RING(evo, format);
 
diff --git a/drivers/gpu/drm/nouveau/nv50_display.c b/drivers/gpu/drm/nouveau/nv50_display.c
index 55c9663ef2bf..f624c611ddea 100644
--- a/drivers/gpu/drm/nouveau/nv50_display.c
+++ b/drivers/gpu/drm/nouveau/nv50_display.c
@@ -1032,11 +1032,18 @@ nv50_display_irq_hotplug_bh(struct work_struct *work)
 	struct drm_connector *connector;
 	const uint32_t gpio_reg[4] = { 0xe104, 0xe108, 0xe280, 0xe284 };
 	uint32_t unplug_mask, plug_mask, change_mask;
-	uint32_t hpd0, hpd1 = 0;
+	uint32_t hpd0, hpd1;
 
-	hpd0 = nv_rd32(dev, 0xe054) & nv_rd32(dev, 0xe050);
+	spin_lock_irq(&dev_priv->hpd_state.lock);
+	hpd0 = dev_priv->hpd_state.hpd0_bits;
+	dev_priv->hpd_state.hpd0_bits = 0;
+	hpd1 = dev_priv->hpd_state.hpd1_bits;
+	dev_priv->hpd_state.hpd1_bits = 0;
+	spin_unlock_irq(&dev_priv->hpd_state.lock);
+
+	hpd0 &= nv_rd32(dev, 0xe050);
 	if (dev_priv->chipset >= 0x90)
-		hpd1 = nv_rd32(dev, 0xe074) & nv_rd32(dev, 0xe070);
+		hpd1 &= nv_rd32(dev, 0xe070);
 
 	plug_mask   = (hpd0 & 0x0000ffff) | (hpd1 << 16);
 	unplug_mask = (hpd0 >> 16) | (hpd1 & 0xffff0000);
@@ -1078,10 +1085,6 @@ nv50_display_irq_hotplug_bh(struct work_struct *work)
 			helper->dpms(connector->encoder, DRM_MODE_DPMS_OFF);
 	}
 
-	nv_wr32(dev, 0xe054, nv_rd32(dev, 0xe054));
-	if (dev_priv->chipset >= 0x90)
-		nv_wr32(dev, 0xe074, nv_rd32(dev, 0xe074));
-
 	drm_helper_hpd_irq_event(dev);
 }
 
@@ -1092,8 +1095,22 @@ nv50_display_irq_handler(struct drm_device *dev)
 	uint32_t delayed = 0;
 
 	if (nv_rd32(dev, NV50_PMC_INTR_0) & NV50_PMC_INTR_0_HOTPLUG) {
-		if (!work_pending(&dev_priv->hpd_work))
-			queue_work(dev_priv->wq, &dev_priv->hpd_work);
+		uint32_t hpd0_bits, hpd1_bits = 0;
+
+		hpd0_bits = nv_rd32(dev, 0xe054);
+		nv_wr32(dev, 0xe054, hpd0_bits);
+
+		if (dev_priv->chipset >= 0x90) {
+			hpd1_bits = nv_rd32(dev, 0xe074);
+			nv_wr32(dev, 0xe074, hpd1_bits);
+		}
+
+		spin_lock(&dev_priv->hpd_state.lock);
+		dev_priv->hpd_state.hpd0_bits |= hpd0_bits;
+		dev_priv->hpd_state.hpd1_bits |= hpd1_bits;
+		spin_unlock(&dev_priv->hpd_state.lock);
+
+		queue_work(dev_priv->wq, &dev_priv->hpd_work);
 	}
 
 	while (nv_rd32(dev, NV50_PMC_INTR_0) & NV50_PMC_INTR_0_DISPLAY) {
diff --git a/drivers/gpu/drm/nouveau/nv50_fifo.c b/drivers/gpu/drm/nouveau/nv50_fifo.c
index a46a961102f3..1da65bd60c10 100644
--- a/drivers/gpu/drm/nouveau/nv50_fifo.c
+++ b/drivers/gpu/drm/nouveau/nv50_fifo.c
@@ -464,3 +464,8 @@ nv50_fifo_unload_context(struct drm_device *dev)
 	return 0;
 }
 
+void
+nv50_fifo_tlb_flush(struct drm_device *dev)
+{
+	nv50_vm_flush(dev, 5);
+}
diff --git a/drivers/gpu/drm/nouveau/nv50_graph.c b/drivers/gpu/drm/nouveau/nv50_graph.c
index cbf5ae2f67d4..8b669d0af610 100644
--- a/drivers/gpu/drm/nouveau/nv50_graph.c
+++ b/drivers/gpu/drm/nouveau/nv50_graph.c
@@ -402,3 +402,55 @@ struct nouveau_pgraph_object_class nv50_graph_grclass[] = {
 	{ 0x8597, false, NULL }, /* tesla (nva3, nva5, nva8) */
 	{}
 };
+
+void
+nv50_graph_tlb_flush(struct drm_device *dev)
+{
+	nv50_vm_flush(dev, 0);
+}
+
+void
+nv86_graph_tlb_flush(struct drm_device *dev)
+{
+	struct drm_nouveau_private *dev_priv = dev->dev_private;
+	struct nouveau_timer_engine *ptimer = &dev_priv->engine.timer;
+	bool idle, timeout = false;
+	unsigned long flags;
+	u64 start;
+	u32 tmp;
+
+	spin_lock_irqsave(&dev_priv->context_switch_lock, flags);
+	nv_mask(dev, 0x400500, 0x00000001, 0x00000000);
+
+	start = ptimer->read(dev);
+	do {
+		idle = true;
+
+		for (tmp = nv_rd32(dev, 0x400380); tmp && idle; tmp >>= 3) {
+			if ((tmp & 7) == 1)
+				idle = false;
+		}
+
+		for (tmp = nv_rd32(dev, 0x400384); tmp && idle; tmp >>= 3) {
+			if ((tmp & 7) == 1)
+				idle = false;
+		}
+
+		for (tmp = nv_rd32(dev, 0x400388); tmp && idle; tmp >>= 3) {
+			if ((tmp & 7) == 1)
+				idle = false;
+		}
+	} while (!idle && !(timeout = ptimer->read(dev) - start > 2000000000));
+
+	if (timeout) {
+		NV_ERROR(dev, "PGRAPH TLB flush idle timeout fail: "
+			      "0x%08x 0x%08x 0x%08x 0x%08x\n",
+			 nv_rd32(dev, 0x400700), nv_rd32(dev, 0x400380),
+			 nv_rd32(dev, 0x400384), nv_rd32(dev, 0x400388));
+	}
+
+	nv50_vm_flush(dev, 0);
+
+	nv_mask(dev, 0x400500, 0x00000001, 0x00000001);
+	spin_unlock_irqrestore(&dev_priv->context_switch_lock, flags);
+}
diff --git a/drivers/gpu/drm/nouveau/nv50_instmem.c b/drivers/gpu/drm/nouveau/nv50_instmem.c
index a53fc974332b..b773229b7647 100644
--- a/drivers/gpu/drm/nouveau/nv50_instmem.c
+++ b/drivers/gpu/drm/nouveau/nv50_instmem.c
@@ -402,7 +402,6 @@ nv50_instmem_bind(struct drm_device *dev, struct nouveau_gpuobj *gpuobj)
 	}
 	dev_priv->engine.instmem.flush(dev);
 
-	nv50_vm_flush(dev, 4);
 	nv50_vm_flush(dev, 6);
 
 	gpuobj->im_bound = 1;
diff --git a/include/drm/nouveau_drm.h b/include/drm/nouveau_drm.h
index 01a714119506..bc5590b1a1ac 100644
--- a/include/drm/nouveau_drm.h
+++ b/include/drm/nouveau_drm.h
@@ -80,6 +80,7 @@ struct drm_nouveau_gpuobj_free {
 #define NOUVEAU_GETPARAM_VM_VRAM_BASE    12
 #define NOUVEAU_GETPARAM_GRAPH_UNITS     13
 #define NOUVEAU_GETPARAM_PTIMER_TIME     14
+#define NOUVEAU_GETPARAM_HAS_BO_USAGE    15
 struct drm_nouveau_getparam {
 	uint64_t param;
 	uint64_t value;
@@ -95,6 +96,12 @@ struct drm_nouveau_setparam {
 #define NOUVEAU_GEM_DOMAIN_GART      (1 << 2)
 #define NOUVEAU_GEM_DOMAIN_MAPPABLE  (1 << 3)
 
+#define NOUVEAU_GEM_TILE_LAYOUT_MASK 0x0000ff00
+#define NOUVEAU_GEM_TILE_16BPP       0x00000001
+#define NOUVEAU_GEM_TILE_32BPP       0x00000002
+#define NOUVEAU_GEM_TILE_ZETA        0x00000004
+#define NOUVEAU_GEM_TILE_NONCONTIG   0x00000008
+
 struct drm_nouveau_gem_info {
 	uint32_t handle;
 	uint32_t domain;