Merge tag 'asoc-v3.14-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/sound into for-linus

ASoC: Fixes for v3.14

A few fixes, all driver speccific ones.  The DaVinci ones aren't as
clear as they should be from the subject lines on the commits but they
fix issues which will prevent correct operation in some use cases and
only affect that particular driver so are reasonably safe.

50 files changed, 1282 insertions, 299 deletions

diff --git a/drivers/gpu/drm/ast/ast_fb.c b/drivers/gpu/drm/ast/ast_fb.c
index 3f65dd6..a28640f 100644
--- a/drivers/gpu/drm/ast/ast_fb.c
+++ b/drivers/gpu/drm/ast/ast_fb.c
@@ -65,7 +65,7 @@ static void ast_dirty_update(struct ast_fbdev *afbdev,
 	 * then the BO is being moved and we should
 	 * store up the damage until later.
 	 */
-	if (!drm_can_sleep())
+	if (drm_can_sleep())
 		ret = ast_bo_reserve(bo, true);
 	if (ret) {
 		if (ret != -EBUSY)
diff --git a/drivers/gpu/drm/cirrus/cirrus_fbdev.c b/drivers/gpu/drm/cirrus/cirrus_fbdev.c
index 2fd4a92..32bbba0 100644
--- a/drivers/gpu/drm/cirrus/cirrus_fbdev.c
+++ b/drivers/gpu/drm/cirrus/cirrus_fbdev.c
@@ -39,7 +39,7 @@ static void cirrus_dirty_update(struct cirrus_fbdev *afbdev,
 	 * then the BO is being moved and we should
 	 * store up the damage until later.
 	 */
-	if (!drm_can_sleep())
+	if (drm_can_sleep())
 		ret = cirrus_bo_reserve(bo, true);
 	if (ret) {
 		if (ret != -EBUSY)
diff --git a/drivers/gpu/drm/exynos/Kconfig b/drivers/gpu/drm/exynos/Kconfig
index f227f54..6e1a1a2 100644
--- a/drivers/gpu/drm/exynos/Kconfig
+++ b/drivers/gpu/drm/exynos/Kconfig
@@ -51,7 +51,7 @@ config DRM_EXYNOS_G2D
 
 config DRM_EXYNOS_IPP
 	bool "Exynos DRM IPP"
-	depends on DRM_EXYNOS && !ARCH_MULTIPLATFORM
+	depends on DRM_EXYNOS
 	help
 	  Choose this option if you want to use IPP feature for DRM.
 
@@ -69,6 +69,6 @@ config DRM_EXYNOS_ROTATOR
 
 config DRM_EXYNOS_GSC
 	bool "Exynos DRM GSC"
-	depends on DRM_EXYNOS_IPP && ARCH_EXYNOS5
+	depends on DRM_EXYNOS_IPP && ARCH_EXYNOS5 && !ARCH_MULTIPLATFORM
 	help
 	  Choose this option if you want to use Exynos GSC for DRM.
diff --git a/drivers/gpu/drm/exynos/exynos_drm_drv.c b/drivers/gpu/drm/exynos/exynos_drm_drv.c
index 9d096a0..215131a 100644
--- a/drivers/gpu/drm/exynos/exynos_drm_drv.c
+++ b/drivers/gpu/drm/exynos/exynos_drm_drv.c
@@ -171,22 +171,24 @@ static int exynos_drm_open(struct drm_device *dev, struct drm_file *file)
 	file->driver_priv = file_priv;
 
 	ret = exynos_drm_subdrv_open(dev, file);
-	if (ret) {
-		kfree(file_priv);
-		file->driver_priv = NULL;
-	}
+	if (ret)
+		goto out;
 
 	anon_filp = anon_inode_getfile("exynos_gem", &exynos_drm_gem_fops,
 					NULL, 0);
 	if (IS_ERR(anon_filp)) {
-		kfree(file_priv);
-		return PTR_ERR(anon_filp);
+		ret = PTR_ERR(anon_filp);
+		goto out;
 	}
 
 	anon_filp->f_mode = FMODE_READ | FMODE_WRITE;
 	file_priv->anon_filp = anon_filp;
 
 	return ret;
+out:
+	kfree(file_priv);
+	file->driver_priv = NULL;
+	return ret;
 }
 
 static void exynos_drm_preclose(struct drm_device *dev,
diff --git a/drivers/gpu/drm/exynos/exynos_drm_g2d.c b/drivers/gpu/drm/exynos/exynos_drm_g2d.c
index 380aec2..6c1885e 100644
--- a/drivers/gpu/drm/exynos/exynos_drm_g2d.c
+++ b/drivers/gpu/drm/exynos/exynos_drm_g2d.c
@@ -607,7 +607,7 @@ static enum g2d_reg_type g2d_get_reg_type(int reg_offset)
 		reg_type = REG_TYPE_NONE;
 		DRM_ERROR("Unknown register offset![%d]\n", reg_offset);
 		break;
-	};
+	}
 
 	return reg_type;
 }
diff --git a/drivers/gpu/drm/exynos/exynos_drm_ipp.c b/drivers/gpu/drm/exynos/exynos_drm_ipp.c
index d519a4e..09312b8 100644
--- a/drivers/gpu/drm/exynos/exynos_drm_ipp.c
+++ b/drivers/gpu/drm/exynos/exynos_drm_ipp.c
@@ -16,7 +16,6 @@
 #include <linux/types.h>
 #include <linux/clk.h>
 #include <linux/pm_runtime.h>
-#include <plat/map-base.h>
 
 #include <drm/drmP.h>
 #include <drm/exynos_drm.h>
@@ -826,7 +825,7 @@ static void ipp_put_event(struct drm_exynos_ipp_cmd_node *c_node,
 		DRM_DEBUG_KMS("count[%d]e[0x%x]\n", count++, (int)e);
 
 		/*
-		 * quf == NULL condition means all event deletion.
+		 * qbuf == NULL condition means all event deletion.
 		 * stop operations want to delete all event list.
 		 * another case delete only same buf id.
 		 */
diff --git a/drivers/gpu/drm/exynos/exynos_hdmi.c b/drivers/gpu/drm/exynos/exynos_hdmi.c
index a0e10ae..c021ddc 100644
--- a/drivers/gpu/drm/exynos/exynos_hdmi.c
+++ b/drivers/gpu/drm/exynos/exynos_hdmi.c
@@ -34,6 +34,7 @@
 #include <linux/io.h>
 #include <linux/of.h>
 #include <linux/of_gpio.h>
+#include <linux/hdmi.h>
 
 #include <drm/exynos_drm.h>
 
@@ -59,19 +60,6 @@
 #define HDMI_AUI_VERSION	0x01
 #define HDMI_AUI_LENGTH	0x0A
 
-/* HDMI infoframe to configure HDMI out packet header, AUI and AVI */
-enum HDMI_PACKET_TYPE {
-	/* refer to Table 5-8 Packet Type in HDMI specification v1.4a */
-	/* InfoFrame packet type */
-	HDMI_PACKET_TYPE_INFOFRAME = 0x80,
-	/* Vendor-Specific InfoFrame */
-	HDMI_PACKET_TYPE_VSI = HDMI_PACKET_TYPE_INFOFRAME + 1,
-	/* Auxiliary Video information InfoFrame */
-	HDMI_PACKET_TYPE_AVI = HDMI_PACKET_TYPE_INFOFRAME + 2,
-	/* Audio information InfoFrame */
-	HDMI_PACKET_TYPE_AUI = HDMI_PACKET_TYPE_INFOFRAME + 4
-};
-
 enum hdmi_type {
 	HDMI_TYPE13,
 	HDMI_TYPE14,
@@ -379,12 +367,6 @@ static const struct hdmiphy_config hdmiphy_v14_configs[] = {
 	},
 };
 
-struct hdmi_infoframe {
-	enum HDMI_PACKET_TYPE type;
-	u8 ver;
-	u8 len;
-};
-
 static inline u32 hdmi_reg_read(struct hdmi_context *hdata, u32 reg_id)
 {
 	return readl(hdata->regs + reg_id);
@@ -682,7 +664,7 @@ static u8 hdmi_chksum(struct hdmi_context *hdata,
 }
 
 static void hdmi_reg_infoframe(struct hdmi_context *hdata,
-			struct hdmi_infoframe *infoframe)
+			union hdmi_infoframe *infoframe)
 {
 	u32 hdr_sum;
 	u8 chksum;
@@ -700,13 +682,15 @@ static void hdmi_reg_infoframe(struct hdmi_context *hdata,
 		return;
 	}
 
-	switch (infoframe->type) {
-	case HDMI_PACKET_TYPE_AVI:
+	switch (infoframe->any.type) {
+	case HDMI_INFOFRAME_TYPE_AVI:
 		hdmi_reg_writeb(hdata, HDMI_AVI_CON, HDMI_AVI_CON_EVERY_VSYNC);
-		hdmi_reg_writeb(hdata, HDMI_AVI_HEADER0, infoframe->type);
-		hdmi_reg_writeb(hdata, HDMI_AVI_HEADER1, infoframe->ver);
-		hdmi_reg_writeb(hdata, HDMI_AVI_HEADER2, infoframe->len);
-		hdr_sum = infoframe->type + infoframe->ver + infoframe->len;
+		hdmi_reg_writeb(hdata, HDMI_AVI_HEADER0, infoframe->any.type);
+		hdmi_reg_writeb(hdata, HDMI_AVI_HEADER1,
+				infoframe->any.version);
+		hdmi_reg_writeb(hdata, HDMI_AVI_HEADER2, infoframe->any.length);
+		hdr_sum = infoframe->any.type + infoframe->any.version +
+			  infoframe->any.length;
 
 		/* Output format zero hardcoded ,RGB YBCR selection */
 		hdmi_reg_writeb(hdata, HDMI_AVI_BYTE(1), 0 << 5 |
@@ -722,18 +706,20 @@ static void hdmi_reg_infoframe(struct hdmi_context *hdata,
 		hdmi_reg_writeb(hdata, HDMI_AVI_BYTE(4), vic);
 
 		chksum = hdmi_chksum(hdata, HDMI_AVI_BYTE(1),
-					infoframe->len, hdr_sum);
+					infoframe->any.length, hdr_sum);
 		DRM_DEBUG_KMS("AVI checksum = 0x%x\n", chksum);
 		hdmi_reg_writeb(hdata, HDMI_AVI_CHECK_SUM, chksum);
 		break;
-	case HDMI_PACKET_TYPE_AUI:
+	case HDMI_INFOFRAME_TYPE_AUDIO:
 		hdmi_reg_writeb(hdata, HDMI_AUI_CON, 0x02);
-		hdmi_reg_writeb(hdata, HDMI_AUI_HEADER0, infoframe->type);
-		hdmi_reg_writeb(hdata, HDMI_AUI_HEADER1, infoframe->ver);
-		hdmi_reg_writeb(hdata, HDMI_AUI_HEADER2, infoframe->len);
-		hdr_sum = infoframe->type + infoframe->ver + infoframe->len;
+		hdmi_reg_writeb(hdata, HDMI_AUI_HEADER0, infoframe->any.type);
+		hdmi_reg_writeb(hdata, HDMI_AUI_HEADER1,
+				infoframe->any.version);
+		hdmi_reg_writeb(hdata, HDMI_AUI_HEADER2, infoframe->any.length);
+		hdr_sum = infoframe->any.type + infoframe->any.version +
+			  infoframe->any.length;
 		chksum = hdmi_chksum(hdata, HDMI_AUI_BYTE(1),
-					infoframe->len, hdr_sum);
+					infoframe->any.length, hdr_sum);
 		DRM_DEBUG_KMS("AUI checksum = 0x%x\n", chksum);
 		hdmi_reg_writeb(hdata, HDMI_AUI_CHECK_SUM, chksum);
 		break;
@@ -985,7 +971,7 @@ static void hdmi_conf_reset(struct hdmi_context *hdata)
 
 static void hdmi_conf_init(struct hdmi_context *hdata)
 {
-	struct hdmi_infoframe infoframe;
+	union hdmi_infoframe infoframe;
 
 	/* disable HPD interrupts from HDMI IP block, use GPIO instead */
 	hdmi_reg_writemask(hdata, HDMI_INTC_CON, 0, HDMI_INTC_EN_GLOBAL |
@@ -1021,14 +1007,14 @@ static void hdmi_conf_init(struct hdmi_context *hdata)
 		hdmi_reg_writeb(hdata, HDMI_V13_AUI_CON, 0x02);
 		hdmi_reg_writeb(hdata, HDMI_V13_ACR_CON, 0x04);
 	} else {
-		infoframe.type = HDMI_PACKET_TYPE_AVI;
-		infoframe.ver = HDMI_AVI_VERSION;
-		infoframe.len = HDMI_AVI_LENGTH;
+		infoframe.any.type = HDMI_INFOFRAME_TYPE_AVI;
+		infoframe.any.version = HDMI_AVI_VERSION;
+		infoframe.any.length = HDMI_AVI_LENGTH;
 		hdmi_reg_infoframe(hdata, &infoframe);
 
-		infoframe.type = HDMI_PACKET_TYPE_AUI;
-		infoframe.ver = HDMI_AUI_VERSION;
-		infoframe.len = HDMI_AUI_LENGTH;
+		infoframe.any.type = HDMI_INFOFRAME_TYPE_AUDIO;
+		infoframe.any.version = HDMI_AUI_VERSION;
+		infoframe.any.length = HDMI_AUI_LENGTH;
 		hdmi_reg_infoframe(hdata, &infoframe);
 
 		/* enable AVI packet every vsync, fixes purple line problem */
diff --git a/drivers/gpu/drm/i2c/tda998x_drv.c b/drivers/gpu/drm/i2c/tda998x_drv.c
index 400b0c4..fa18cf3 100644
--- a/drivers/gpu/drm/i2c/tda998x_drv.c
+++ b/drivers/gpu/drm/i2c/tda998x_drv.c
@@ -208,7 +208,7 @@ struct tda998x_priv {
 # define PLL_SERIAL_1_SRL_IZ(x)   (((x) & 3) << 1)
 # define PLL_SERIAL_1_SRL_MAN_IZ  (1 << 6)
 #define REG_PLL_SERIAL_2          REG(0x02, 0x01)     /* read/write */
-# define PLL_SERIAL_2_SRL_NOSC(x) (((x) & 3) << 0)
+# define PLL_SERIAL_2_SRL_NOSC(x) ((x) << 0)
 # define PLL_SERIAL_2_SRL_PR(x)   (((x) & 0xf) << 4)
 #define REG_PLL_SERIAL_3          REG(0x02, 0x02)     /* read/write */
 # define PLL_SERIAL_3_SRL_CCIR    (1 << 0)
@@ -528,10 +528,10 @@ tda998x_write_aif(struct drm_encoder *encoder, struct tda998x_encoder_params *p)
 {
 	uint8_t buf[PB(5) + 1];
 
+	memset(buf, 0, sizeof(buf));
 	buf[HB(0)] = 0x84;
 	buf[HB(1)] = 0x01;
 	buf[HB(2)] = 10;
-	buf[PB(0)] = 0;
 	buf[PB(1)] = p->audio_frame[1] & 0x07; /* CC */
 	buf[PB(2)] = p->audio_frame[2] & 0x1c; /* SF */
 	buf[PB(4)] = p->audio_frame[4];
@@ -824,6 +824,11 @@ tda998x_encoder_mode_set(struct drm_encoder *encoder,
 	}
 
 	div = 148500 / mode->clock;
+	if (div != 0) {
+		div--;
+		if (div > 3)
+			div = 3;
+	}
 
 	/* mute the audio FIFO: */
 	reg_set(encoder, REG_AIP_CNTRL_0, AIP_CNTRL_0_RST_FIFO);
@@ -913,7 +918,7 @@ tda998x_encoder_mode_set(struct drm_encoder *encoder,
 
 	if (priv->rev == TDA19988) {
 		/* let incoming pixels fill the active space (if any) */
-		reg_write(encoder, REG_ENABLE_SPACE, 0x01);
+		reg_write(encoder, REG_ENABLE_SPACE, 0x00);
 	}
 
 	/* must be last register set: */
@@ -1094,6 +1099,8 @@ tda998x_encoder_destroy(struct drm_encoder *encoder)
 {
 	struct tda998x_priv *priv = to_tda998x_priv(encoder);
 	drm_i2c_encoder_destroy(encoder);
+	if (priv->cec)
+		i2c_unregister_device(priv->cec);
 	kfree(priv);
 }
 
@@ -1142,8 +1149,10 @@ tda998x_encoder_init(struct i2c_client *client,
 	priv->vip_cntrl_1 = VIP_CNTRL_1_SWAP_C(0) | VIP_CNTRL_1_SWAP_D(1);
 	priv->vip_cntrl_2 = VIP_CNTRL_2_SWAP_E(4) | VIP_CNTRL_2_SWAP_F(5);
 
-	priv->current_page = 0;
+	priv->current_page = 0xff;
 	priv->cec = i2c_new_dummy(client->adapter, 0x34);
+	if (!priv->cec)
+		return -ENODEV;
 	priv->dpms = DRM_MODE_DPMS_OFF;
 
 	encoder_slave->slave_priv = priv;
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 4a2bf8e..df77e20 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -1831,6 +1831,14 @@ struct drm_i915_file_private {
 
 /* Early gen2 have a totally busted CS tlb and require pinned batches. */
 #define HAS_BROKEN_CS_TLB(dev)		(IS_I830(dev) || IS_845G(dev))
+/*
+ * dp aux and gmbus irq on gen4 seems to be able to generate legacy interrupts
+ * even when in MSI mode. This results in spurious interrupt warnings if the
+ * legacy irq no. is shared with another device. The kernel then disables that
+ * interrupt source and so prevents the other device from working properly.
+ */
+#define HAS_AUX_IRQ(dev) (INTEL_INFO(dev)->gen >= 5)
+#define HAS_GMBUS_IRQ(dev) (INTEL_INFO(dev)->gen >= 5)
 
 /* With the 945 and later, Y tiling got adjusted so that it was 32 128-byte
  * rows, which changed the alignment requirements and fence programming.
diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c
index d7fd2fd..990cf8f 100644
--- a/drivers/gpu/drm/i915/i915_gpu_error.c
+++ b/drivers/gpu/drm/i915/i915_gpu_error.c
@@ -146,7 +146,10 @@ static void i915_error_vprintf(struct drm_i915_error_state_buf *e,
 		va_list tmp;
 
 		va_copy(tmp, args);
-		if (!__i915_error_seek(e, vsnprintf(NULL, 0, f, tmp)))
+		len = vsnprintf(NULL, 0, f, tmp);
+		va_end(tmp);
+
+		if (!__i915_error_seek(e, len))
 			return;
 	}
 
diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c
index 17d8fcb..9fec711 100644
--- a/drivers/gpu/drm/i915/i915_irq.c
+++ b/drivers/gpu/drm/i915/i915_irq.c
@@ -567,8 +567,7 @@ static u32 i915_get_vblank_counter(struct drm_device *dev, int pipe)
 
 		vbl_start = mode->crtc_vblank_start * mode->crtc_htotal;
 	} else {
-		enum transcoder cpu_transcoder =
-			intel_pipe_to_cpu_transcoder(dev_priv, pipe);
+		enum transcoder cpu_transcoder = (enum transcoder) pipe;
 		u32 htotal;
 
 		htotal = ((I915_READ(HTOTAL(cpu_transcoder)) >> 16) & 0x1fff) + 1;
diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c
index 5ede4e8..2f517b8 100644
--- a/drivers/gpu/drm/i915/intel_dp.c
+++ b/drivers/gpu/drm/i915/intel_dp.c
@@ -404,7 +404,7 @@ intel_dp_aux_ch(struct intel_dp *intel_dp,
 	int i, ret, recv_bytes;
 	uint32_t status;
 	int try, precharge, clock = 0;
-	bool has_aux_irq = true;
+	bool has_aux_irq = HAS_AUX_IRQ(dev);
 	uint32_t timeout;
 
 	/* dp aux is extremely sensitive to irq latency, hence request the
@@ -1869,10 +1869,12 @@ static void vlv_pre_enable_dp(struct intel_encoder *encoder)
 
 	mutex_unlock(&dev_priv->dpio_lock);
 
-	/* init power sequencer on this pipe and port */
-	intel_dp_init_panel_power_sequencer(dev, intel_dp, &power_seq);
-	intel_dp_init_panel_power_sequencer_registers(dev, intel_dp,
-						      &power_seq);
+	if (is_edp(intel_dp)) {
+		/* init power sequencer on this pipe and port */
+		intel_dp_init_panel_power_sequencer(dev, intel_dp, &power_seq);
+		intel_dp_init_panel_power_sequencer_registers(dev, intel_dp,
+							      &power_seq);
+	}
 
 	intel_enable_dp(encoder);
 
diff --git a/drivers/gpu/drm/i915/intel_i2c.c b/drivers/gpu/drm/i915/intel_i2c.c
index b1dc33f..d33b61d 100644
--- a/drivers/gpu/drm/i915/intel_i2c.c
+++ b/drivers/gpu/drm/i915/intel_i2c.c
@@ -258,13 +258,6 @@ intel_gpio_setup(struct intel_gmbus *bus, u32 pin)
 	algo->data = bus;
 }
 
-/*
- * gmbus on gen4 seems to be able to generate legacy interrupts even when in MSI
- * mode. This results in spurious interrupt warnings if the legacy irq no. is
- * shared with another device. The kernel then disables that interrupt source
- * and so prevents the other device from working properly.
- */
-#define HAS_GMBUS_IRQ(dev) (INTEL_INFO(dev)->gen >= 5)
 static int
 gmbus_wait_hw_status(struct drm_i915_private *dev_priv,
 		     u32 gmbus2_status,
diff --git a/drivers/gpu/drm/i915/intel_opregion.c b/drivers/gpu/drm/i915/intel_opregion.c
index 4e960ec..acde294 100644
--- a/drivers/gpu/drm/i915/intel_opregion.c
+++ b/drivers/gpu/drm/i915/intel_opregion.c
@@ -226,6 +226,8 @@ struct opregion_asle {
 #define ACPI_DIGITAL_OUTPUT (3<<8)
 #define ACPI_LVDS_OUTPUT (4<<8)
 
+#define MAX_DSLP	1500
+
 #ifdef CONFIG_ACPI
 static int swsci(struct drm_device *dev, u32 function, u32 parm, u32 *parm_out)
 {
@@ -260,10 +262,11 @@ static int swsci(struct drm_device *dev, u32 function, u32 parm, u32 *parm_out)
 		/* The spec says 2ms should be the default, but it's too small
 		 * for some machines. */
 		dslp = 50;
-	} else if (dslp > 500) {
+	} else if (dslp > MAX_DSLP) {
 		/* Hey bios, trust must be earned. */
-		WARN_ONCE(1, "excessive driver sleep timeout (DSPL) %u\n", dslp);
-		dslp = 500;
+		DRM_INFO_ONCE("ACPI BIOS requests an excessive sleep of %u ms, "
+			      "using %u ms instead\n", dslp, MAX_DSLP);
+		dslp = MAX_DSLP;
 	}
 
 	/* The spec tells us to do this, but we are the only user... */
diff --git a/drivers/gpu/drm/mgag200/mgag200_fb.c b/drivers/gpu/drm/mgag200/mgag200_fb.c
index f9adc27..13b7dd8 100644
--- a/drivers/gpu/drm/mgag200/mgag200_fb.c
+++ b/drivers/gpu/drm/mgag200/mgag200_fb.c
@@ -41,7 +41,7 @@ static void mga_dirty_update(struct mga_fbdev *mfbdev,
 	 * then the BO is being moved and we should
 	 * store up the damage until later.
 	 */
-	if (!drm_can_sleep())
+	if (drm_can_sleep())
 		ret = mgag200_bo_reserve(bo, true);
 	if (ret) {
 		if (ret != -EBUSY)
diff --git a/drivers/gpu/drm/mgag200/mgag200_mode.c b/drivers/gpu/drm/mgag200/mgag200_mode.c
index b8583f2..9683747 100644
--- a/drivers/gpu/drm/mgag200/mgag200_mode.c
+++ b/drivers/gpu/drm/mgag200/mgag200_mode.c
@@ -1519,11 +1519,11 @@ static int mga_vga_mode_valid(struct drm_connector *connector,
 		(mga_vga_calculate_mode_bandwidth(mode, bpp)
 			> (32700 * 1024))) {
 		return MODE_BANDWIDTH;
-	} else if (mode->type == G200_EH &&
+	} else if (mdev->type == G200_EH &&
 		(mga_vga_calculate_mode_bandwidth(mode, bpp)
 			> (37500 * 1024))) {
 		return MODE_BANDWIDTH;
-	} else if (mode->type == G200_ER &&
+	} else if (mdev->type == G200_ER &&
 		(mga_vga_calculate_mode_bandwidth(mode,
 			bpp) > (55000 * 1024))) {
 		return MODE_BANDWIDTH;
diff --git a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c
index 1964f4f..84c5b13 100644
--- a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c
+++ b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_crtc.c
@@ -39,6 +39,7 @@ struct mdp4_crtc {
 		spinlock_t lock;
 		bool stale;
 		uint32_t width, height;
+		uint32_t x, y;
 
 		/* next cursor to scan-out: */
 		uint32_t next_iova;
@@ -57,9 +58,16 @@ struct mdp4_crtc {
 #define PENDING_FLIP   0x2
 	atomic_t pending;
 
-	/* the fb that we currently hold a scanout ref to: */
+	/* the fb that we logically (from PoV of KMS API) hold a ref
+	 * to.  Which we may not yet be scanning out (we may still
+	 * be scanning out previous in case of page_flip while waiting
+	 * for gpu rendering to complete:
+	 */
 	struct drm_framebuffer *fb;
 
+	/* the fb that we currently hold a scanout ref to: */
+	struct drm_framebuffer *scanout_fb;
+
 	/* for unref'ing framebuffers after scanout completes: */
 	struct drm_flip_work unref_fb_work;
 
@@ -77,24 +85,73 @@ static struct mdp4_kms *get_kms(struct drm_crtc *crtc)
 	return to_mdp4_kms(to_mdp_kms(priv->kms));
 }
 
-static void update_fb(struct drm_crtc *crtc, bool async,
-		struct drm_framebuffer *new_fb)
+static void request_pending(struct drm_crtc *crtc, uint32_t pending)
 {
 	struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc);
-	struct drm_framebuffer *old_fb = mdp4_crtc->fb;
 
-	if (old_fb)
-		drm_flip_work_queue(&mdp4_crtc->unref_fb_work, old_fb);
+	atomic_or(pending, &mdp4_crtc->pending);
+	mdp_irq_register(&get_kms(crtc)->base, &mdp4_crtc->vblank);
+}
+
+static void crtc_flush(struct drm_crtc *crtc)
+{
+	struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc);
+	struct mdp4_kms *mdp4_kms = get_kms(crtc);
+	uint32_t i, flush = 0;
+
+	for (i = 0; i < ARRAY_SIZE(mdp4_crtc->planes); i++) {
+		struct drm_plane *plane = mdp4_crtc->planes[i];
+		if (plane) {
+			enum mdp4_pipe pipe_id = mdp4_plane_pipe(plane);
+			flush |= pipe2flush(pipe_id);
+		}
+	}
+	flush |= ovlp2flush(mdp4_crtc->ovlp);
+
+	DBG("%s: flush=%08x", mdp4_crtc->name, flush);
+
+	mdp4_write(mdp4_kms, REG_MDP4_OVERLAY_FLUSH, flush);
+}
+
+static void update_fb(struct drm_crtc *crtc, struct drm_framebuffer *new_fb)
+{
+	struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc);
+	struct drm_framebuffer *old_fb = mdp4_crtc->fb;
 
 	/* grab reference to incoming scanout fb: */
 	drm_framebuffer_reference(new_fb);
 	mdp4_crtc->base.fb = new_fb;
 	mdp4_crtc->fb = new_fb;
 
-	if (!async) {
-		/* enable vblank to pick up the old_fb */
-		mdp_irq_register(&get_kms(crtc)->base, &mdp4_crtc->vblank);
-	}
+	if (old_fb)
+		drm_flip_work_queue(&mdp4_crtc->unref_fb_work, old_fb);
+}
+
+/* unlike update_fb(), take a ref to the new scanout fb *before* updating
+ * plane, then call this.  Needed to ensure we don't unref the buffer that
+ * is actually still being scanned out.
+ *
+ * Note that this whole thing goes away with atomic.. since we can defer
+ * calling into driver until rendering is done.
+ */
+static void update_scanout(struct drm_crtc *crtc, struct drm_framebuffer *fb)
+{
+	struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc);
+
+	/* flush updates, to make sure hw is updated to new scanout fb,
+	 * so that we can safely queue unref to current fb (ie. next
+	 * vblank we know hw is done w/ previous scanout_fb).
+	 */
+	crtc_flush(crtc);
+
+	if (mdp4_crtc->scanout_fb)
+		drm_flip_work_queue(&mdp4_crtc->unref_fb_work,
+				mdp4_crtc->scanout_fb);
+
+	mdp4_crtc->scanout_fb = fb;
+
+	/* enable vblank to complete flip: */
+	request_pending(crtc, PENDING_FLIP);
 }
 
 /* if file!=NULL, this is preclose potential cancel-flip path */
@@ -120,34 +177,6 @@ static void complete_flip(struct drm_crtc *crtc, struct drm_file *file)
 	spin_unlock_irqrestore(&dev->event_lock, flags);
 }
 
-static void crtc_flush(struct drm_crtc *crtc)
-{
-	struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc);
-	struct mdp4_kms *mdp4_kms = get_kms(crtc);
-	uint32_t i, flush = 0;
-
-	for (i = 0; i < ARRAY_SIZE(mdp4_crtc->planes); i++) {
-		struct drm_plane *plane = mdp4_crtc->planes[i];
-		if (plane) {
-			enum mdp4_pipe pipe_id = mdp4_plane_pipe(plane);
-			flush |= pipe2flush(pipe_id);
-		}
-	}
-	flush |= ovlp2flush(mdp4_crtc->ovlp);
-
-	DBG("%s: flush=%08x", mdp4_crtc->name, flush);
-
-	mdp4_write(mdp4_kms, REG_MDP4_OVERLAY_FLUSH, flush);
-}
-
-static void request_pending(struct drm_crtc *crtc, uint32_t pending)
-{
-	struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc);
-
-	atomic_or(pending, &mdp4_crtc->pending);
-	mdp_irq_register(&get_kms(crtc)->base, &mdp4_crtc->vblank);
-}
-
 static void pageflip_cb(struct msm_fence_cb *cb)
 {
 	struct mdp4_crtc *mdp4_crtc =
@@ -158,11 +187,9 @@ static void pageflip_cb(struct msm_fence_cb *cb)
 	if (!fb)
 		return;
 
+	drm_framebuffer_reference(fb);
 	mdp4_plane_set_scanout(mdp4_crtc->plane, fb);
-	crtc_flush(crtc);
-
-	/* enable vblank to complete flip: */
-	request_pending(crtc, PENDING_FLIP);
+	update_scanout(crtc, fb);
 }
 
 static void unref_fb_worker(struct drm_flip_work *work, void *val)
@@ -320,6 +347,20 @@ static int mdp4_crtc_mode_set(struct drm_crtc *crtc,
 			mode->vsync_end, mode->vtotal,
 			mode->type, mode->flags);
 
+	/* grab extra ref for update_scanout() */
+	drm_framebuffer_reference(crtc->fb);
+
+	ret = mdp4_plane_mode_set(mdp4_crtc->plane, crtc, crtc->fb,
+			0, 0, mode->hdisplay, mode->vdisplay,
+			x << 16, y << 16,
+			mode->hdisplay << 16, mode->vdisplay << 16);
+	if (ret) {
+		drm_framebuffer_unreference(crtc->fb);
+		dev_err(crtc->dev->dev, "%s: failed to set mode on plane: %d\n",
+				mdp4_crtc->name, ret);
+		return ret;
+	}
+
 	mdp4_write(mdp4_kms, REG_MDP4_DMA_SRC_SIZE(dma),
 			MDP4_DMA_SRC_SIZE_WIDTH(mode->hdisplay) |
 			MDP4_DMA_SRC_SIZE_HEIGHT(mode->vdisplay));
@@ -341,24 +382,15 @@ static int mdp4_crtc_mode_set(struct drm_crtc *crtc,
 
 	mdp4_write(mdp4_kms, REG_MDP4_OVLP_CFG(ovlp), 1);
 
-	update_fb(crtc, false, crtc->fb);
-
-	ret = mdp4_plane_mode_set(mdp4_crtc->plane, crtc, crtc->fb,
-			0, 0, mode->hdisplay, mode->vdisplay,
-			x << 16, y << 16,
-			mode->hdisplay << 16, mode->vdisplay << 16);
-	if (ret) {
-		dev_err(crtc->dev->dev, "%s: failed to set mode on plane: %d\n",
-				mdp4_crtc->name, ret);
-		return ret;
-	}
-
 	if (dma == DMA_E) {
 		mdp4_write(mdp4_kms, REG_MDP4_DMA_E_QUANT(0), 0x00ff0000);
 		mdp4_write(mdp4_kms, REG_MDP4_DMA_E_QUANT(1), 0x00ff0000);
 		mdp4_write(mdp4_kms, REG_MDP4_DMA_E_QUANT(2), 0x00ff0000);
 	}
 
+	update_fb(crtc, crtc->fb);
+	update_scanout(crtc, crtc->fb);
+
 	return 0;
 }
 
@@ -385,13 +417,24 @@ static int mdp4_crtc_mode_set_base(struct drm_crtc *crtc, int x, int y,
 	struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc);
 	struct drm_plane *plane = mdp4_crtc->plane;
 	struct drm_display_mode *mode = &crtc->mode;
+	int ret;
 
-	update_fb(crtc, false, crtc->fb);
+	/* grab extra ref for update_scanout() */
+	drm_framebuffer_reference(crtc->fb);
 
-	return mdp4_plane_mode_set(plane, crtc, crtc->fb,
+	ret = mdp4_plane_mode_set(plane, crtc, crtc->fb,
 			0, 0, mode->hdisplay, mode->vdisplay,
 			x << 16, y << 16,
 			mode->hdisplay << 16, mode->vdisplay << 16);
+	if (ret) {
+		drm_framebuffer_unreference(crtc->fb);
+		return ret;
+	}
+
+	update_fb(crtc, crtc->fb);
+	update_scanout(crtc, crtc->fb);
+
+	return 0;
 }
 
 static void mdp4_crtc_load_lut(struct drm_crtc *crtc)
@@ -419,7 +462,7 @@ static int mdp4_crtc_page_flip(struct drm_crtc *crtc,
 	mdp4_crtc->event = event;
 	spin_unlock_irqrestore(&dev->event_lock, flags);
 
-	update_fb(crtc, true, new_fb);
+	update_fb(crtc, new_fb);
 
 	return msm_gem_queue_inactive_cb(obj, &mdp4_crtc->pageflip_cb);
 }
@@ -442,12 +485,12 @@ static int mdp4_crtc_set_property(struct drm_crtc *crtc,
 static void update_cursor(struct drm_crtc *crtc)
 {
 	struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc);
+	struct mdp4_kms *mdp4_kms = get_kms(crtc);
 	enum mdp4_dma dma = mdp4_crtc->dma;
 	unsigned long flags;
 
 	spin_lock_irqsave(&mdp4_crtc->cursor.lock, flags);
 	if (mdp4_crtc->cursor.stale) {
-		struct mdp4_kms *mdp4_kms = get_kms(crtc);
 		struct drm_gem_object *next_bo = mdp4_crtc->cursor.next_bo;
 		struct drm_gem_object *prev_bo = mdp4_crtc->cursor.scanout_bo;
 		uint32_t iova = mdp4_crtc->cursor.next_iova;
@@ -479,6 +522,11 @@ static void update_cursor(struct drm_crtc *crtc)
 		mdp4_crtc->cursor.scanout_bo = next_bo;
 		mdp4_crtc->cursor.stale = false;
 	}
+
+	mdp4_write(mdp4_kms, REG_MDP4_DMA_CURSOR_POS(dma),
+			MDP4_DMA_CURSOR_POS_X(mdp4_crtc->cursor.x) |
+			MDP4_DMA_CURSOR_POS_Y(mdp4_crtc->cursor.y));
+
 	spin_unlock_irqrestore(&mdp4_crtc->cursor.lock, flags);
 }
 
@@ -530,6 +578,7 @@ static int mdp4_crtc_cursor_set(struct drm_crtc *crtc,
 		drm_gem_object_unreference_unlocked(old_bo);
 	}
 
+	crtc_flush(crtc);
 	request_pending(crtc, PENDING_CURSOR);
 
 	return 0;
@@ -542,12 +591,15 @@ fail:
 static int mdp4_crtc_cursor_move(struct drm_crtc *crtc, int x, int y)
 {
 	struct mdp4_crtc *mdp4_crtc = to_mdp4_crtc(crtc);
-	struct mdp4_kms *mdp4_kms = get_kms(crtc);
-	enum mdp4_dma dma = mdp4_crtc->dma;
+	unsigned long flags;
 
-	mdp4_write(mdp4_kms, REG_MDP4_DMA_CURSOR_POS(dma),
-			MDP4_DMA_CURSOR_POS_X(x) |
-			MDP4_DMA_CURSOR_POS_Y(y));
+	spin_lock_irqsave(&mdp4_crtc->cursor.lock, flags);
+	mdp4_crtc->cursor.x = x;
+	mdp4_crtc->cursor.y = y;
+	spin_unlock_irqrestore(&mdp4_crtc->cursor.lock, flags);
+
+	crtc_flush(crtc);
+	request_pending(crtc, PENDING_CURSOR);
 
 	return 0;
 }
@@ -713,6 +765,7 @@ struct drm_crtc *mdp4_crtc_init(struct drm_device *dev,
 	crtc = &mdp4_crtc->base;
 
 	mdp4_crtc->plane = plane;
+	mdp4_crtc->id = id;
 
 	mdp4_crtc->ovlp = ovlp_id;
 	mdp4_crtc->dma = dma_id;
diff --git a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_plane.c b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_plane.c
index 2406027..1e893dd 100644
--- a/drivers/gpu/drm/msm/mdp/mdp4/mdp4_plane.c
+++ b/drivers/gpu/drm/msm/mdp/mdp4/mdp4_plane.c
@@ -170,8 +170,8 @@ int mdp4_plane_mode_set(struct drm_plane *plane,
 			MDP4_PIPE_DST_SIZE_HEIGHT(crtc_h));
 
 	mdp4_write(mdp4_kms, REG_MDP4_PIPE_DST_XY(pipe),
-			MDP4_PIPE_SRC_XY_X(crtc_x) |
-			MDP4_PIPE_SRC_XY_Y(crtc_y));
+			MDP4_PIPE_DST_XY_X(crtc_x) |
+			MDP4_PIPE_DST_XY_Y(crtc_y));
 
 	mdp4_plane_set_scanout(plane, fb);
 
diff --git a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c
index 71a3b23..f279402 100644
--- a/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c
+++ b/drivers/gpu/drm/msm/mdp/mdp5/mdp5_crtc.c
@@ -296,6 +296,7 @@ static int mdp5_crtc_mode_set(struct drm_crtc *crtc,
 			x << 16, y << 16,
 			mode->hdisplay << 16, mode->vdisplay << 16);
 	if (ret) {
+		drm_framebuffer_unreference(crtc->fb);
 		dev_err(crtc->dev->dev, "%s: failed to set mode on plane: %d\n",
 				mdp5_crtc->name, ret);
 		return ret;
@@ -343,11 +344,15 @@ static int mdp5_crtc_mode_set_base(struct drm_crtc *crtc, int x, int y,
 			0, 0, mode->hdisplay, mode->vdisplay,
 			x << 16, y << 16,
 			mode->hdisplay << 16, mode->vdisplay << 16);
+	if (ret) {
+		drm_framebuffer_unreference(crtc->fb);
+		return ret;
+	}
 
 	update_fb(crtc, crtc->fb);
 	update_scanout(crtc, crtc->fb);
 
-	return ret;
+	return 0;
 }
 
 static void mdp5_crtc_load_lut(struct drm_crtc *crtc)
diff --git a/drivers/gpu/drm/msm/msm_gem.c b/drivers/gpu/drm/msm/msm_gem.c
index d8d60c9..3da8264 100644
--- a/drivers/gpu/drm/msm/msm_gem.c
+++ b/drivers/gpu/drm/msm/msm_gem.c
@@ -644,7 +644,7 @@ struct drm_gem_object *msm_gem_new(struct drm_device *dev,
 
 fail:
 	if (obj)
-		drm_gem_object_unreference_unlocked(obj);
+		drm_gem_object_unreference(obj);
 
 	return ERR_PTR(ret);
 }
diff --git a/drivers/gpu/drm/msm/msm_gem_submit.c b/drivers/gpu/drm/msm/msm_gem_submit.c
index 5281d4b..5423e91 100644
--- a/drivers/gpu/drm/msm/msm_gem_submit.c
+++ b/drivers/gpu/drm/msm/msm_gem_submit.c
@@ -163,7 +163,7 @@ retry:
 
 
 		/* if locking succeeded, pin bo: */
-		ret = msm_gem_get_iova(&msm_obj->base,
+		ret = msm_gem_get_iova_locked(&msm_obj->base,
 				submit->gpu->id, &iova);
 
 		/* this would break the logic in the fail path.. there is no
@@ -247,7 +247,7 @@ static int submit_reloc(struct msm_gem_submit *submit, struct msm_gem_object *ob
 	/* For now, just map the entire thing.  Eventually we probably
 	 * to do it page-by-page, w/ kmap() if not vmap()d..
 	 */
-	ptr = msm_gem_vaddr(&obj->base);
+	ptr = msm_gem_vaddr_locked(&obj->base);
 
 	if (IS_ERR(ptr)) {
 		ret = PTR_ERR(ptr);
@@ -307,14 +307,12 @@ static void submit_cleanup(struct msm_gem_submit *submit, bool fail)
 {
 	unsigned i;
 
-	mutex_lock(&submit->dev->struct_mutex);
 	for (i = 0; i < submit->nr_bos; i++) {
 		struct msm_gem_object *msm_obj = submit->bos[i].obj;
 		submit_unlock_unpin_bo(submit, i);
 		list_del_init(&msm_obj->submit_entry);
 		drm_gem_object_unreference(&msm_obj->base);
 	}
-	mutex_unlock(&submit->dev->struct_mutex);
 
 	ww_acquire_fini(&submit->ticket);
 	kfree(submit);
@@ -342,6 +340,8 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data,
 	if (args->nr_cmds > MAX_CMDS)
 		return -EINVAL;
 
+	mutex_lock(&dev->struct_mutex);
+
 	submit = submit_create(dev, gpu, args->nr_bos);
 	if (!submit) {
 		ret = -ENOMEM;
@@ -410,5 +410,6 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data,
 out:
 	if (submit)
 		submit_cleanup(submit, !!ret);
+	mutex_unlock(&dev->struct_mutex);
 	return ret;
 }
diff --git a/drivers/gpu/drm/msm/msm_gpu.c b/drivers/gpu/drm/msm/msm_gpu.c
index 4ebce8b..0cfe3f4 100644
--- a/drivers/gpu/drm/msm/msm_gpu.c
+++ b/drivers/gpu/drm/msm/msm_gpu.c
@@ -298,8 +298,6 @@ int msm_gpu_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit,
 	struct msm_drm_private *priv = dev->dev_private;
 	int i, ret;
 
-	mutex_lock(&dev->struct_mutex);
-
 	submit->fence = ++priv->next_fence;
 
 	gpu->submitted_fence = submit->fence;
@@ -331,7 +329,6 @@ int msm_gpu_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit,
 			msm_gem_move_to_active(&msm_obj->base, gpu, true, submit->fence);
 	}
 	hangcheck_timer_reset(gpu);
-	mutex_unlock(&dev->struct_mutex);
 
 	return ret;
 }
diff --git a/drivers/gpu/drm/radeon/btc_dpm.c b/drivers/gpu/drm/radeon/btc_dpm.c
index 0fbd36f..ea103cc 100644
--- a/drivers/gpu/drm/radeon/btc_dpm.c
+++ b/drivers/gpu/drm/radeon/btc_dpm.c
@@ -29,6 +29,7 @@
 #include "cypress_dpm.h"
 #include "btc_dpm.h"
 #include "atom.h"
+#include <linux/seq_file.h>
 
 #define MC_CG_ARB_FREQ_F0           0x0a
 #define MC_CG_ARB_FREQ_F1           0x0b
@@ -2756,6 +2757,37 @@ void btc_dpm_fini(struct radeon_device *rdev)
 	r600_free_extended_power_table(rdev);
 }
 
+void btc_dpm_debugfs_print_current_performance_level(struct radeon_device *rdev,
+						     struct seq_file *m)
+{
+	struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev);
+	struct radeon_ps *rps = &eg_pi->current_rps;
+	struct rv7xx_ps *ps = rv770_get_ps(rps);
+	struct rv7xx_pl *pl;
+	u32 current_index =
+		(RREG32(TARGET_AND_CURRENT_PROFILE_INDEX) & CURRENT_PROFILE_INDEX_MASK) >>
+		CURRENT_PROFILE_INDEX_SHIFT;
+
+	if (current_index > 2) {
+		seq_printf(m, "invalid dpm profile %d\n", current_index);
+	} else {
+		if (current_index == 0)
+			pl = &ps->low;
+		else if (current_index == 1)
+			pl = &ps->medium;
+		else /* current_index == 2 */
+			pl = &ps->high;
+		seq_printf(m, "uvd    vclk: %d dclk: %d\n", rps->vclk, rps->dclk);
+		if (rdev->family >= CHIP_CEDAR) {
+			seq_printf(m, "power level %d    sclk: %u mclk: %u vddc: %u vddci: %u\n",
+				   current_index, pl->sclk, pl->mclk, pl->vddc, pl->vddci);
+		} else {
+			seq_printf(m, "power level %d    sclk: %u mclk: %u vddc: %u\n",
+				   current_index, pl->sclk, pl->mclk, pl->vddc);
+		}
+	}
+}
+
 u32 btc_dpm_get_sclk(struct radeon_device *rdev, bool low)
 {
 	struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev);
diff --git a/drivers/gpu/drm/radeon/btcd.h b/drivers/gpu/drm/radeon/btcd.h
index 29e32de..9c65be2 100644
--- a/drivers/gpu/drm/radeon/btcd.h
+++ b/drivers/gpu/drm/radeon/btcd.h
@@ -44,6 +44,10 @@
 #       define DYN_SPREAD_SPECTRUM_EN                   (1 << 23)
 #       define AC_DC_SW                                 (1 << 24)
 
+#define TARGET_AND_CURRENT_PROFILE_INDEX                  0x66c
+#       define CURRENT_PROFILE_INDEX_MASK                 (0xf << 4)
+#       define CURRENT_PROFILE_INDEX_SHIFT                4
+
 #define	CG_BIF_REQ_AND_RSP				0x7f4
 #define		CG_CLIENT_REQ(x)			((x) << 0)
 #define		CG_CLIENT_REQ_MASK			(0xff << 0)
diff --git a/drivers/gpu/drm/radeon/kv_dpm.c b/drivers/gpu/drm/radeon/kv_dpm.c
index b6e01d5..351db36 100644
--- a/drivers/gpu/drm/radeon/kv_dpm.c
+++ b/drivers/gpu/drm/radeon/kv_dpm.c
@@ -1223,7 +1223,7 @@ int kv_dpm_enable(struct radeon_device *rdev)
 
 int kv_dpm_late_enable(struct radeon_device *rdev)
 {
-	int ret;
+	int ret = 0;
 
 	if (rdev->irq.installed &&
 	    r600_is_internal_thermal_sensor(rdev->pm.int_thermal_type)) {
diff --git a/drivers/gpu/drm/radeon/ni_dpm.c b/drivers/gpu/drm/radeon/ni_dpm.c
index c351226..1217fbc 100644
--- a/drivers/gpu/drm/radeon/ni_dpm.c
+++ b/drivers/gpu/drm/radeon/ni_dpm.c
@@ -3945,7 +3945,6 @@ static void ni_parse_pplib_clock_info(struct radeon_device *rdev,
 	struct rv7xx_power_info *pi = rv770_get_pi(rdev);
 	struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev);
 	struct ni_ps *ps = ni_get_ps(rps);
-	u16 vddc;
 	struct rv7xx_pl *pl = &ps->performance_levels[index];
 
 	ps->performance_level_count = index + 1;
@@ -3961,8 +3960,8 @@ static void ni_parse_pplib_clock_info(struct radeon_device *rdev,
 
 	/* patch up vddc if necessary */
 	if (pl->vddc == 0xff01) {
-		if (radeon_atom_get_max_vddc(rdev, 0, 0, &vddc) == 0)
-			pl->vddc = vddc;
+		if (pi->max_vddc)
+			pl->vddc = pi->max_vddc;
 	}
 
 	if (rps->class & ATOM_PPLIB_CLASSIFICATION_ACPI) {
@@ -4322,7 +4321,8 @@ void ni_dpm_print_power_state(struct radeon_device *rdev,
 void ni_dpm_debugfs_print_current_performance_level(struct radeon_device *rdev,
 						    struct seq_file *m)
 {
-	struct radeon_ps *rps = rdev->pm.dpm.current_ps;
+	struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev);
+	struct radeon_ps *rps = &eg_pi->current_rps;
 	struct ni_ps *ps = ni_get_ps(rps);
 	struct rv7xx_pl *pl;
 	u32 current_index =
diff --git a/drivers/gpu/drm/radeon/r600.c b/drivers/gpu/drm/radeon/r600.c
index 56140b4..cdbc417 100644
--- a/drivers/gpu/drm/radeon/r600.c
+++ b/drivers/gpu/drm/radeon/r600.c
@@ -3991,6 +3991,10 @@ restart_ih:
 				break;
 			}
 			break;
+		case 124: /* UVD */
+			DRM_DEBUG("IH: UVD int: 0x%08x\n", src_data);
+			radeon_fence_process(rdev, R600_RING_TYPE_UVD_INDEX);
+			break;
 		case 176: /* CP_INT in ring buffer */
 		case 177: /* CP_INT in IB1 */
 		case 178: /* CP_INT in IB2 */
diff --git a/drivers/gpu/drm/radeon/r600_cs.c b/drivers/gpu/drm/radeon/r600_cs.c
index 7b399dc..2812c7d1a 100644
--- a/drivers/gpu/drm/radeon/r600_cs.c
+++ b/drivers/gpu/drm/radeon/r600_cs.c
@@ -1007,8 +1007,22 @@ static int r600_cs_check_reg(struct radeon_cs_parser *p, u32 reg, u32 idx)
 	case R_008C64_SQ_VSTMP_RING_SIZE:
 	case R_0288C8_SQ_GS_VERT_ITEMSIZE:
 		/* get value to populate the IB don't remove */
-		tmp =radeon_get_ib_value(p, idx);
-		ib[idx] = 0;
+		/*tmp =radeon_get_ib_value(p, idx);
+		  ib[idx] = 0;*/
+		break;
+	case SQ_ESGS_RING_BASE:
+	case SQ_GSVS_RING_BASE:
+	case SQ_ESTMP_RING_BASE:
+	case SQ_GSTMP_RING_BASE:
+	case SQ_PSTMP_RING_BASE:
+	case SQ_VSTMP_RING_BASE:
+		r = radeon_cs_packet_next_reloc(p, &reloc, 0);
+		if (r) {
+			dev_warn(p->dev, "bad SET_CONTEXT_REG "
+					"0x%04X\n", reg);
+			return -EINVAL;
+		}
+		ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
 		break;
 	case SQ_CONFIG:
 		track->sq_config = radeon_get_ib_value(p, idx);
diff --git a/drivers/gpu/drm/radeon/radeon_asic.c b/drivers/gpu/drm/radeon/radeon_asic.c
index f74db43..dda02bf 100644
--- a/drivers/gpu/drm/radeon/radeon_asic.c
+++ b/drivers/gpu/drm/radeon/radeon_asic.c
@@ -1555,7 +1555,7 @@ static struct radeon_asic btc_asic = {
 		.get_sclk = &btc_dpm_get_sclk,
 		.get_mclk = &btc_dpm_get_mclk,
 		.print_power_state = &rv770_dpm_print_power_state,
-		.debugfs_print_current_performance_level = &rv770_dpm_debugfs_print_current_performance_level,
+		.debugfs_print_current_performance_level = &btc_dpm_debugfs_print_current_performance_level,
 		.force_performance_level = &rv770_dpm_force_performance_level,
 		.vblank_too_short = &btc_dpm_vblank_too_short,
 	},
diff --git a/drivers/gpu/drm/radeon/radeon_asic.h b/drivers/gpu/drm/radeon/radeon_asic.h
index b3bc433..ae637cf 100644
--- a/drivers/gpu/drm/radeon/radeon_asic.h
+++ b/drivers/gpu/drm/radeon/radeon_asic.h
@@ -551,6 +551,8 @@ void btc_dpm_fini(struct radeon_device *rdev);
 u32 btc_dpm_get_sclk(struct radeon_device *rdev, bool low);
 u32 btc_dpm_get_mclk(struct radeon_device *rdev, bool low);
 bool btc_dpm_vblank_too_short(struct radeon_device *rdev);
+void btc_dpm_debugfs_print_current_performance_level(struct radeon_device *rdev,
+						     struct seq_file *m);
 int sumo_dpm_init(struct radeon_device *rdev);
 int sumo_dpm_enable(struct radeon_device *rdev);
 int sumo_dpm_late_enable(struct radeon_device *rdev);
diff --git a/drivers/gpu/drm/radeon/radeon_drv.c b/drivers/gpu/drm/radeon/radeon_drv.c
index ec8c388..84a1bbb7 100644
--- a/drivers/gpu/drm/radeon/radeon_drv.c
+++ b/drivers/gpu/drm/radeon/radeon_drv.c
@@ -78,9 +78,10 @@
  *   2.34.0 - Add CIK tiling mode array query
  *   2.35.0 - Add CIK macrotile mode array query
  *   2.36.0 - Fix CIK DCE tiling setup
+ *   2.37.0 - allow GS ring setup on r6xx/r7xx
  */
 #define KMS_DRIVER_MAJOR	2
-#define KMS_DRIVER_MINOR	36
+#define KMS_DRIVER_MINOR	37
 #define KMS_DRIVER_PATCHLEVEL	0
 int radeon_driver_load_kms(struct drm_device *dev, unsigned long flags);
 int radeon_driver_unload_kms(struct drm_device *dev);
diff --git a/drivers/gpu/drm/radeon/reg_srcs/r600 b/drivers/gpu/drm/radeon/reg_srcs/r600
index 20bfbda..ec0c682 100644
--- a/drivers/gpu/drm/radeon/reg_srcs/r600
+++ b/drivers/gpu/drm/radeon/reg_srcs/r600
@@ -18,6 +18,7 @@ r600 0x9400
 0x00028A3C VGT_GROUP_VECT_1_FMT_CNTL
 0x00028A40 VGT_GS_MODE
 0x00028A6C VGT_GS_OUT_PRIM_TYPE
+0x00028B38 VGT_GS_MAX_VERT_OUT
 0x000088C8 VGT_GS_PER_ES
 0x000088E8 VGT_GS_PER_VS
 0x000088D4 VGT_GS_VERTEX_REUSE
diff --git a/drivers/gpu/drm/radeon/rv770_dpm.c b/drivers/gpu/drm/radeon/rv770_dpm.c
index 80c595a..5b2ea8a 100644
--- a/drivers/gpu/drm/radeon/rv770_dpm.c
+++ b/drivers/gpu/drm/radeon/rv770_dpm.c
@@ -2174,7 +2174,6 @@ static void rv7xx_parse_pplib_clock_info(struct radeon_device *rdev,
 	struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev);
 	struct rv7xx_ps *ps = rv770_get_ps(rps);
 	u32 sclk, mclk;
-	u16 vddc;
 	struct rv7xx_pl *pl;
 
 	switch (index) {
@@ -2214,8 +2213,8 @@ static void rv7xx_parse_pplib_clock_info(struct radeon_device *rdev,
 
 	/* patch up vddc if necessary */
 	if (pl->vddc == 0xff01) {
-		if (radeon_atom_get_max_vddc(rdev, 0, 0, &vddc) == 0)
-			pl->vddc = vddc;
+		if (pi->max_vddc)
+			pl->vddc = pi->max_vddc;
 	}
 
 	if (rps->class & ATOM_PPLIB_CLASSIFICATION_ACPI) {
diff --git a/drivers/gpu/drm/radeon/si.c b/drivers/gpu/drm/radeon/si.c
index 09ec4f6..8357832 100644
--- a/drivers/gpu/drm/radeon/si.c
+++ b/drivers/gpu/drm/radeon/si.c
@@ -6338,6 +6338,10 @@ restart_ih:
 				break;
 			}
 			break;
+		case 124: /* UVD */
+			DRM_DEBUG("IH: UVD int: 0x%08x\n", src_data);
+			radeon_fence_process(rdev, R600_RING_TYPE_UVD_INDEX);
+			break;
 		case 146:
 		case 147:
 			addr = RREG32(VM_CONTEXT1_PROTECTION_FAULT_ADDR);
diff --git a/drivers/gpu/drm/radeon/si_dpm.c b/drivers/gpu/drm/radeon/si_dpm.c
index 0471501..eafb0e6 100644
--- a/drivers/gpu/drm/radeon/si_dpm.c
+++ b/drivers/gpu/drm/radeon/si_dpm.c
@@ -6472,7 +6472,8 @@ void si_dpm_fini(struct radeon_device *rdev)
 void si_dpm_debugfs_print_current_performance_level(struct radeon_device *rdev,
 						    struct seq_file *m)
 {
-	struct radeon_ps *rps = rdev->pm.dpm.current_ps;
+	struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev);
+	struct radeon_ps *rps = &eg_pi->current_rps;
 	struct ni_ps *ps = ni_get_ps(rps);
 	struct rv7xx_pl *pl;
 	u32 current_index =
diff --git a/drivers/gpu/drm/radeon/sumo_dpm.c b/drivers/gpu/drm/radeon/sumo_dpm.c
index f121efe..8b47b3cd 100644
--- a/drivers/gpu/drm/radeon/sumo_dpm.c
+++ b/drivers/gpu/drm/radeon/sumo_dpm.c
@@ -1807,7 +1807,7 @@ void sumo_dpm_debugfs_print_current_performance_level(struct radeon_device *rdev
 						      struct seq_file *m)
 {
 	struct sumo_power_info *pi = sumo_get_pi(rdev);
-	struct radeon_ps *rps = rdev->pm.dpm.current_ps;
+	struct radeon_ps *rps = &pi->current_rps;
 	struct sumo_ps *ps = sumo_get_ps(rps);
 	struct sumo_pl *pl;
 	u32 current_index =
diff --git a/drivers/gpu/drm/radeon/trinity_dpm.c b/drivers/gpu/drm/radeon/trinity_dpm.c
index 2d44719..2da0e17 100644
--- a/drivers/gpu/drm/radeon/trinity_dpm.c
+++ b/drivers/gpu/drm/radeon/trinity_dpm.c
@@ -1926,7 +1926,8 @@ void trinity_dpm_print_power_state(struct radeon_device *rdev,
 void trinity_dpm_debugfs_print_current_performance_level(struct radeon_device *rdev,
 							 struct seq_file *m)
 {
-	struct radeon_ps *rps = rdev->pm.dpm.current_ps;
+	struct trinity_power_info *pi = trinity_get_pi(rdev);
+	struct radeon_ps *rps = &pi->current_rps;
 	struct trinity_ps *ps = trinity_get_ps(rps);
 	struct trinity_pl *pl;
 	u32 current_index =
diff --git a/drivers/gpu/drm/radeon/uvd_v2_2.c b/drivers/gpu/drm/radeon/uvd_v2_2.c
index 824550d..d177100 100644
--- a/drivers/gpu/drm/radeon/uvd_v2_2.c
+++ b/drivers/gpu/drm/radeon/uvd_v2_2.c
@@ -57,7 +57,6 @@ void uvd_v2_2_fence_emit(struct radeon_device *rdev,
 	radeon_ring_write(ring, 0);
 	radeon_ring_write(ring, PACKET0(UVD_GPCOM_VCPU_CMD, 0));
 	radeon_ring_write(ring, 2);
-	return;
 }
 
 /**
diff --git a/drivers/gpu/drm/ttm/ttm_object.c b/drivers/gpu/drm/ttm/ttm_object.c
index 3707985..53b51c4 100644
--- a/drivers/gpu/drm/ttm/ttm_object.c
+++ b/drivers/gpu/drm/ttm/ttm_object.c
@@ -292,7 +292,7 @@ int ttm_ref_object_add(struct ttm_object_file *tfile,
 
 		if (ret == 0) {
 			ref = drm_hash_entry(hash, struct ttm_ref_object, hash);
-			if (!kref_get_unless_zero(&ref->kref)) {
+			if (kref_get_unless_zero(&ref->kref)) {
 				rcu_read_unlock();
 				break;
 			}
diff --git a/drivers/gpu/drm/ttm/ttm_tt.c b/drivers/gpu/drm/ttm/ttm_tt.c
index 9af9908..75f3190 100644
--- a/drivers/gpu/drm/ttm/ttm_tt.c
+++ b/drivers/gpu/drm/ttm/ttm_tt.c
@@ -380,6 +380,9 @@ static void ttm_tt_clear_mapping(struct ttm_tt *ttm)
 	pgoff_t i;
 	struct page **page = ttm->pages;
 
+	if (ttm->page_flags & TTM_PAGE_FLAG_SG)
+		return;
+
 	for (i = 0; i < ttm->num_pages; ++i) {
 		(*page)->mapping = NULL;
 		(*page++)->index = 0;
diff --git a/drivers/gpu/drm/vmwgfx/svga3d_reg.h b/drivers/gpu/drm/vmwgfx/svga3d_reg.h
index d95335c..b645647 100644
--- a/drivers/gpu/drm/vmwgfx/svga3d_reg.h
+++ b/drivers/gpu/drm/vmwgfx/svga3d_reg.h
@@ -2583,4 +2583,28 @@ typedef union {
    float  f;
 } SVGA3dDevCapResult;
 
+typedef enum {
+   SVGA3DCAPS_RECORD_UNKNOWN        = 0,
+   SVGA3DCAPS_RECORD_DEVCAPS_MIN    = 0x100,
+   SVGA3DCAPS_RECORD_DEVCAPS        = 0x100,
+   SVGA3DCAPS_RECORD_DEVCAPS_MAX    = 0x1ff,
+} SVGA3dCapsRecordType;
+
+typedef
+struct SVGA3dCapsRecordHeader {
+   uint32 length;
+   SVGA3dCapsRecordType type;
+}
+SVGA3dCapsRecordHeader;
+
+typedef
+struct SVGA3dCapsRecord {
+   SVGA3dCapsRecordHeader header;
+   uint32 data[1];
+}
+SVGA3dCapsRecord;
+
+
+typedef uint32 SVGA3dCapPair[2];
+
 #endif /* _SVGA3D_REG_H_ */
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_context.c b/drivers/gpu/drm/vmwgfx/vmwgfx_context.c
index 82c41da..9426c53 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_context.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_context.c
@@ -37,7 +37,7 @@ struct vmw_user_context {
 
 
 
-typedef int (*vmw_scrub_func)(struct vmw_ctx_bindinfo *);
+typedef int (*vmw_scrub_func)(struct vmw_ctx_bindinfo *, bool);
 
 static void vmw_user_context_free(struct vmw_resource *res);
 static struct vmw_resource *
@@ -50,9 +50,11 @@ static int vmw_gb_context_unbind(struct vmw_resource *res,
 				 bool readback,
 				 struct ttm_validate_buffer *val_buf);
 static int vmw_gb_context_destroy(struct vmw_resource *res);
-static int vmw_context_scrub_shader(struct vmw_ctx_bindinfo *bi);
-static int vmw_context_scrub_render_target(struct vmw_ctx_bindinfo *bi);
-static int vmw_context_scrub_texture(struct vmw_ctx_bindinfo *bi);
+static int vmw_context_scrub_shader(struct vmw_ctx_bindinfo *bi, bool rebind);
+static int vmw_context_scrub_render_target(struct vmw_ctx_bindinfo *bi,
+					   bool rebind);
+static int vmw_context_scrub_texture(struct vmw_ctx_bindinfo *bi, bool rebind);
+static void vmw_context_binding_state_scrub(struct vmw_ctx_binding_state *cbs);
 static void vmw_context_binding_state_kill(struct vmw_ctx_binding_state *cbs);
 static uint64_t vmw_user_context_size;
 
@@ -111,10 +113,14 @@ static void vmw_hw_context_destroy(struct vmw_resource *res)
 
 	if (res->func->destroy == vmw_gb_context_destroy) {
 		mutex_lock(&dev_priv->cmdbuf_mutex);
+		mutex_lock(&dev_priv->binding_mutex);
+		(void) vmw_context_binding_state_kill
+			(&container_of(res, struct vmw_user_context, res)->cbs);
 		(void) vmw_gb_context_destroy(res);
 		if (dev_priv->pinned_bo != NULL &&
 		    !dev_priv->query_cid_valid)
 			__vmw_execbuf_release_pinned_bo(dev_priv, NULL);
+		mutex_unlock(&dev_priv->binding_mutex);
 		mutex_unlock(&dev_priv->cmdbuf_mutex);
 		return;
 	}
@@ -328,7 +334,7 @@ static int vmw_gb_context_unbind(struct vmw_resource *res,
 	BUG_ON(bo->mem.mem_type != VMW_PL_MOB);
 
 	mutex_lock(&dev_priv->binding_mutex);
-	vmw_context_binding_state_kill(&uctx->cbs);
+	vmw_context_binding_state_scrub(&uctx->cbs);
 
 	submit_size = sizeof(*cmd2) + (readback ? sizeof(*cmd1) : 0);
 
@@ -378,10 +384,6 @@ static int vmw_gb_context_destroy(struct vmw_resource *res)
 		SVGA3dCmdHeader header;
 		SVGA3dCmdDestroyGBContext body;
 	} *cmd;
-	struct vmw_user_context *uctx =
-		container_of(res, struct vmw_user_context, res);
-
-	BUG_ON(!list_empty(&uctx->cbs.list));
 
 	if (likely(res->id == -1))
 		return 0;
@@ -528,8 +530,9 @@ out_unlock:
  * vmw_context_scrub_shader - scrub a shader binding from a context.
  *
  * @bi: single binding information.
+ * @rebind: Whether to issue a bind instead of scrub command.
  */
-static int vmw_context_scrub_shader(struct vmw_ctx_bindinfo *bi)
+static int vmw_context_scrub_shader(struct vmw_ctx_bindinfo *bi, bool rebind)
 {
 	struct vmw_private *dev_priv = bi->ctx->dev_priv;
 	struct {
@@ -548,7 +551,8 @@ static int vmw_context_scrub_shader(struct vmw_ctx_bindinfo *bi)
 	cmd->header.size = sizeof(cmd->body);
 	cmd->body.cid = bi->ctx->id;
 	cmd->body.type = bi->i1.shader_type;
-	cmd->body.shid = SVGA3D_INVALID_ID;
+	cmd->body.shid =
+		cpu_to_le32((rebind) ? bi->res->id : SVGA3D_INVALID_ID);
 	vmw_fifo_commit(dev_priv, sizeof(*cmd));
 
 	return 0;
@@ -559,8 +563,10 @@ static int vmw_context_scrub_shader(struct vmw_ctx_bindinfo *bi)
  * from a context.
  *
  * @bi: single binding information.
+ * @rebind: Whether to issue a bind instead of scrub command.
  */
-static int vmw_context_scrub_render_target(struct vmw_ctx_bindinfo *bi)
+static int vmw_context_scrub_render_target(struct vmw_ctx_bindinfo *bi,
+					   bool rebind)
 {
 	struct vmw_private *dev_priv = bi->ctx->dev_priv;
 	struct {
@@ -579,7 +585,8 @@ static int vmw_context_scrub_render_target(struct vmw_ctx_bindinfo *bi)
 	cmd->header.size = sizeof(cmd->body);
 	cmd->body.cid = bi->ctx->id;
 	cmd->body.type = bi->i1.rt_type;
-	cmd->body.target.sid = SVGA3D_INVALID_ID;
+	cmd->body.target.sid =
+		cpu_to_le32((rebind) ? bi->res->id : SVGA3D_INVALID_ID);
 	cmd->body.target.face = 0;
 	cmd->body.target.mipmap = 0;
 	vmw_fifo_commit(dev_priv, sizeof(*cmd));
@@ -591,11 +598,13 @@ static int vmw_context_scrub_render_target(struct vmw_ctx_bindinfo *bi)
  * vmw_context_scrub_texture - scrub a texture binding from a context.
  *
  * @bi: single binding information.
+ * @rebind: Whether to issue a bind instead of scrub command.
  *
  * TODO: Possibly complement this function with a function that takes
  * a list of texture bindings and combines them to a single command.
  */
-static int vmw_context_scrub_texture(struct vmw_ctx_bindinfo *bi)
+static int vmw_context_scrub_texture(struct vmw_ctx_bindinfo *bi,
+				     bool rebind)
 {
 	struct vmw_private *dev_priv = bi->ctx->dev_priv;
 	struct {
@@ -619,7 +628,8 @@ static int vmw_context_scrub_texture(struct vmw_ctx_bindinfo *bi)
 	cmd->body.c.cid = bi->ctx->id;
 	cmd->body.s1.stage = bi->i1.texture_stage;
 	cmd->body.s1.name = SVGA3D_TS_BIND_TEXTURE;
-	cmd->body.s1.value = (uint32) SVGA3D_INVALID_ID;
+	cmd->body.s1.value =
+		cpu_to_le32((rebind) ? bi->res->id : SVGA3D_INVALID_ID);
 	vmw_fifo_commit(dev_priv, sizeof(*cmd));
 
 	return 0;
@@ -692,6 +702,7 @@ int vmw_context_binding_add(struct vmw_ctx_binding_state *cbs,
 		vmw_context_binding_drop(loc);
 
 	loc->bi = *bi;
+	loc->bi.scrubbed = false;
 	list_add_tail(&loc->ctx_list, &cbs->list);
 	INIT_LIST_HEAD(&loc->res_list);
 
@@ -727,12 +738,11 @@ static void vmw_context_binding_transfer(struct vmw_ctx_binding_state *cbs,
 	if (loc->bi.ctx != NULL)
 		vmw_context_binding_drop(loc);
 
-	loc->bi = *bi;
-	list_add_tail(&loc->ctx_list, &cbs->list);
-	if (bi->res != NULL)
+	if (bi->res != NULL) {
+		loc->bi = *bi;
+		list_add_tail(&loc->ctx_list, &cbs->list);
 		list_add_tail(&loc->res_list, &bi->res->binding_head);
-	else
-		INIT_LIST_HEAD(&loc->res_list);
+	}
 }
 
 /**
@@ -746,7 +756,10 @@ static void vmw_context_binding_transfer(struct vmw_ctx_binding_state *cbs,
  */
 static void vmw_context_binding_kill(struct vmw_ctx_binding *cb)
 {
-	(void) vmw_scrub_funcs[cb->bi.bt](&cb->bi);
+	if (!cb->bi.scrubbed) {
+		(void) vmw_scrub_funcs[cb->bi.bt](&cb->bi, false);
+		cb->bi.scrubbed = true;
+	}
 	vmw_context_binding_drop(cb);
 }
 
@@ -768,6 +781,27 @@ static void vmw_context_binding_state_kill(struct vmw_ctx_binding_state *cbs)
 }
 
 /**
+ * vmw_context_binding_state_scrub - Scrub all bindings associated with a
+ * struct vmw_ctx_binding state structure.
+ *
+ * @cbs: Pointer to the context binding state tracker.
+ *
+ * Emits commands to scrub all bindings associated with the
+ * context binding state tracker.
+ */
+static void vmw_context_binding_state_scrub(struct vmw_ctx_binding_state *cbs)
+{
+	struct vmw_ctx_binding *entry;
+
+	list_for_each_entry(entry, &cbs->list, ctx_list) {
+		if (!entry->bi.scrubbed) {
+			(void) vmw_scrub_funcs[entry->bi.bt](&entry->bi, false);
+			entry->bi.scrubbed = true;
+		}
+	}
+}
+
+/**
  * vmw_context_binding_res_list_kill - Kill all bindings on a
  * resource binding list
  *
@@ -785,6 +819,27 @@ void vmw_context_binding_res_list_kill(struct list_head *head)
 }
 
 /**
+ * vmw_context_binding_res_list_scrub - Scrub all bindings on a
+ * resource binding list
+ *
+ * @head: list head of resource binding list
+ *
+ * Scrub all bindings associated with a specific resource. Typically
+ * called before the resource is evicted.
+ */
+void vmw_context_binding_res_list_scrub(struct list_head *head)
+{
+	struct vmw_ctx_binding *entry;
+
+	list_for_each_entry(entry, head, res_list) {
+		if (!entry->bi.scrubbed) {
+			(void) vmw_scrub_funcs[entry->bi.bt](&entry->bi, false);
+			entry->bi.scrubbed = true;
+		}
+	}
+}
+
+/**
  * vmw_context_binding_state_transfer - Commit staged binding info
  *
  * @ctx: Pointer to context to commit the staged binding info to.
@@ -803,3 +858,50 @@ void vmw_context_binding_state_transfer(struct vmw_resource *ctx,
 	list_for_each_entry_safe(entry, next, &from->list, ctx_list)
 		vmw_context_binding_transfer(&uctx->cbs, &entry->bi);
 }
+
+/**
+ * vmw_context_rebind_all - Rebind all scrubbed bindings of a context
+ *
+ * @ctx: The context resource
+ *
+ * Walks through the context binding list and rebinds all scrubbed
+ * resources.
+ */
+int vmw_context_rebind_all(struct vmw_resource *ctx)
+{
+	struct vmw_ctx_binding *entry;
+	struct vmw_user_context *uctx =
+		container_of(ctx, struct vmw_user_context, res);
+	struct vmw_ctx_binding_state *cbs = &uctx->cbs;
+	int ret;
+
+	list_for_each_entry(entry, &cbs->list, ctx_list) {
+		if (likely(!entry->bi.scrubbed))
+			continue;
+
+		if (WARN_ON(entry->bi.res == NULL || entry->bi.res->id ==
+			    SVGA3D_INVALID_ID))
+			continue;
+
+		ret = vmw_scrub_funcs[entry->bi.bt](&entry->bi, true);
+		if (unlikely(ret != 0))
+			return ret;
+
+		entry->bi.scrubbed = false;
+	}
+
+	return 0;
+}
+
+/**
+ * vmw_context_binding_list - Return a list of context bindings
+ *
+ * @ctx: The context resource
+ *
+ * Returns the current list of bindings of the given context. Note that
+ * this list becomes stale as soon as the dev_priv::binding_mutex is unlocked.
+ */
+struct list_head *vmw_context_binding_list(struct vmw_resource *ctx)
+{
+	return &(container_of(ctx, struct vmw_user_context, res)->cbs.list);
+}
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
index 9893328..3bdc0ad 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
@@ -941,6 +941,7 @@ static void vmw_postclose(struct drm_device *dev,
 		drm_master_put(&vmw_fp->locked_master);
 	}
 
+	vmw_compat_shader_man_destroy(vmw_fp->shman);
 	ttm_object_file_release(&vmw_fp->tfile);
 	kfree(vmw_fp);
 }
@@ -960,11 +961,17 @@ static int vmw_driver_open(struct drm_device *dev, struct drm_file *file_priv)
 	if (unlikely(vmw_fp->tfile == NULL))
 		goto out_no_tfile;
 
+	vmw_fp->shman = vmw_compat_shader_man_create(dev_priv);
+	if (IS_ERR(vmw_fp->shman))
+		goto out_no_shman;
+
 	file_priv->driver_priv = vmw_fp;
 	dev_priv->bdev.dev_mapping = dev->dev_mapping;
 
 	return 0;
 
+out_no_shman:
+	ttm_object_file_release(&vmw_fp->tfile);
 out_no_tfile:
 	kfree(vmw_fp);
 	return ret;
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
index 554e7fa..ecaa302 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
@@ -75,10 +75,14 @@
 #define VMW_RES_FENCE ttm_driver_type3
 #define VMW_RES_SHADER ttm_driver_type4
 
+struct vmw_compat_shader_manager;
+
 struct vmw_fpriv {
 	struct drm_master *locked_master;
 	struct ttm_object_file *tfile;
 	struct list_head fence_events;
+	bool gb_aware;
+	struct vmw_compat_shader_manager *shman;
 };
 
 struct vmw_dma_buffer {
@@ -272,6 +276,7 @@ struct vmw_ctx_bindinfo {
 	struct vmw_resource *ctx;
 	struct vmw_resource *res;
 	enum vmw_ctx_binding_type bt;
+	bool scrubbed;
 	union {
 		SVGA3dShaderType shader_type;
 		SVGA3dRenderTargetType rt_type;
@@ -318,7 +323,7 @@ struct vmw_sw_context{
 	struct drm_open_hash res_ht;
 	bool res_ht_initialized;
 	bool kernel; /**< is the called made from the kernel */
-	struct ttm_object_file *tfile;
+	struct vmw_fpriv *fp;
 	struct list_head validate_nodes;
 	struct vmw_relocation relocs[VMWGFX_MAX_RELOCATIONS];
 	uint32_t cur_reloc;
@@ -336,6 +341,7 @@ struct vmw_sw_context{
 	bool needs_post_query_barrier;
 	struct vmw_resource *error_resource;
 	struct vmw_ctx_binding_state staged_bindings;
+	struct list_head staged_shaders;
 };
 
 struct vmw_legacy_display;
@@ -569,6 +575,8 @@ struct vmw_user_resource_conv;
 
 extern void vmw_resource_unreference(struct vmw_resource **p_res);
 extern struct vmw_resource *vmw_resource_reference(struct vmw_resource *res);
+extern struct vmw_resource *
+vmw_resource_reference_unless_doomed(struct vmw_resource *res);
 extern int vmw_resource_validate(struct vmw_resource *res);
 extern int vmw_resource_reserve(struct vmw_resource *res, bool no_backup);
 extern bool vmw_resource_needs_backup(const struct vmw_resource *res);
@@ -957,6 +965,9 @@ extern void
 vmw_context_binding_state_transfer(struct vmw_resource *res,
 				   struct vmw_ctx_binding_state *cbs);
 extern void vmw_context_binding_res_list_kill(struct list_head *head);
+extern void vmw_context_binding_res_list_scrub(struct list_head *head);
+extern int vmw_context_rebind_all(struct vmw_resource *ctx);
+extern struct list_head *vmw_context_binding_list(struct vmw_resource *ctx);
 
 /*
  * Surface management - vmwgfx_surface.c
@@ -991,6 +1002,28 @@ extern int vmw_shader_define_ioctl(struct drm_device *dev, void *data,
 				   struct drm_file *file_priv);
 extern int vmw_shader_destroy_ioctl(struct drm_device *dev, void *data,
 				    struct drm_file *file_priv);
+extern int vmw_compat_shader_lookup(struct vmw_compat_shader_manager *man,
+				    SVGA3dShaderType shader_type,
+				    u32 *user_key);
+extern void vmw_compat_shaders_commit(struct vmw_compat_shader_manager *man,
+				      struct list_head *list);
+extern void vmw_compat_shaders_revert(struct vmw_compat_shader_manager *man,
+				      struct list_head *list);
+extern int vmw_compat_shader_remove(struct vmw_compat_shader_manager *man,
+				    u32 user_key,
+				    SVGA3dShaderType shader_type,
+				    struct list_head *list);
+extern int vmw_compat_shader_add(struct vmw_compat_shader_manager *man,
+				 u32 user_key, const void *bytecode,
+				 SVGA3dShaderType shader_type,
+				 size_t size,
+				 struct ttm_object_file *tfile,
+				 struct list_head *list);
+extern struct vmw_compat_shader_manager *
+vmw_compat_shader_man_create(struct vmw_private *dev_priv);
+extern void
+vmw_compat_shader_man_destroy(struct vmw_compat_shader_manager *man);
+
 
 /**
  * Inline helper functions
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c b/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c
index 7a5f1eb..269b85c 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c
@@ -114,8 +114,10 @@ static void vmw_resource_list_unreserve(struct list_head *list,
 		 * persistent context binding tracker.
 		 */
 		if (unlikely(val->staged_bindings)) {
-			vmw_context_binding_state_transfer
-				(val->res, val->staged_bindings);
+			if (!backoff) {
+				vmw_context_binding_state_transfer
+					(val->res, val->staged_bindings);
+			}
 			kfree(val->staged_bindings);
 			val->staged_bindings = NULL;
 		}
@@ -178,6 +180,44 @@ static int vmw_resource_val_add(struct vmw_sw_context *sw_context,
 }
 
 /**
+ * vmw_resource_context_res_add - Put resources previously bound to a context on
+ * the validation list
+ *
+ * @dev_priv: Pointer to a device private structure
+ * @sw_context: Pointer to a software context used for this command submission
+ * @ctx: Pointer to the context resource
+ *
+ * This function puts all resources that were previously bound to @ctx on
+ * the resource validation list. This is part of the context state reemission
+ */
+static int vmw_resource_context_res_add(struct vmw_private *dev_priv,
+					struct vmw_sw_context *sw_context,
+					struct vmw_resource *ctx)
+{
+	struct list_head *binding_list;
+	struct vmw_ctx_binding *entry;
+	int ret = 0;
+	struct vmw_resource *res;
+
+	mutex_lock(&dev_priv->binding_mutex);
+	binding_list = vmw_context_binding_list(ctx);
+
+	list_for_each_entry(entry, binding_list, ctx_list) {
+		res = vmw_resource_reference_unless_doomed(entry->bi.res);
+		if (unlikely(res == NULL))
+			continue;
+
+		ret = vmw_resource_val_add(sw_context, entry->bi.res, NULL);
+		vmw_resource_unreference(&res);
+		if (unlikely(ret != 0))
+			break;
+	}
+
+	mutex_unlock(&dev_priv->binding_mutex);
+	return ret;
+}
+
+/**
  * vmw_resource_relocation_add - Add a relocation to the relocation list
  *
  * @list: Pointer to head of relocation list.
@@ -233,8 +273,12 @@ static void vmw_resource_relocations_apply(uint32_t *cb,
 {
 	struct vmw_resource_relocation *rel;
 
-	list_for_each_entry(rel, list, head)
-		cb[rel->offset] = rel->res->id;
+	list_for_each_entry(rel, list, head) {
+		if (likely(rel->res != NULL))
+			cb[rel->offset] = rel->res->id;
+		else
+			cb[rel->offset] = SVGA_3D_CMD_NOP;
+	}
 }
 
 static int vmw_cmd_invalid(struct vmw_private *dev_priv,
@@ -379,22 +423,27 @@ static int vmw_resources_validate(struct vmw_sw_context *sw_context)
 }
 
 /**
- * vmw_cmd_res_check - Check that a resource is present and if so, put it
+ * vmw_cmd_compat_res_check - Check that a resource is present and if so, put it
  * on the resource validate list unless it's already there.
  *
  * @dev_priv: Pointer to a device private structure.
  * @sw_context: Pointer to the software context.
  * @res_type: Resource type.
  * @converter: User-space visisble type specific information.
- * @id: Pointer to the location in the command buffer currently being
+ * @id: user-space resource id handle.
+ * @id_loc: Pointer to the location in the command buffer currently being
  * parsed from where the user-space resource id handle is located.
+ * @p_val: Pointer to pointer to resource validalidation node. Populated
+ * on exit.
  */
-static int vmw_cmd_res_check(struct vmw_private *dev_priv,
-			     struct vmw_sw_context *sw_context,
-			     enum vmw_res_type res_type,
-			     const struct vmw_user_resource_conv *converter,
-			     uint32_t *id,
-			     struct vmw_resource_val_node **p_val)
+static int
+vmw_cmd_compat_res_check(struct vmw_private *dev_priv,
+			 struct vmw_sw_context *sw_context,
+			 enum vmw_res_type res_type,
+			 const struct vmw_user_resource_conv *converter,
+			 uint32_t id,
+			 uint32_t *id_loc,
+			 struct vmw_resource_val_node **p_val)
 {
 	struct vmw_res_cache_entry *rcache =
 		&sw_context->res_cache[res_type];
@@ -402,7 +451,7 @@ static int vmw_cmd_res_check(struct vmw_private *dev_priv,
 	struct vmw_resource_val_node *node;
 	int ret;
 
-	if (*id == SVGA3D_INVALID_ID) {
+	if (id == SVGA3D_INVALID_ID) {
 		if (p_val)
 			*p_val = NULL;
 		if (res_type == vmw_res_context) {
@@ -417,7 +466,7 @@ static int vmw_cmd_res_check(struct vmw_private *dev_priv,
 	 * resource
 	 */
 
-	if (likely(rcache->valid && *id == rcache->handle)) {
+	if (likely(rcache->valid && id == rcache->handle)) {
 		const struct vmw_resource *res = rcache->res;
 
 		rcache->node->first_usage = false;
@@ -426,28 +475,28 @@ static int vmw_cmd_res_check(struct vmw_private *dev_priv,
 
 		return vmw_resource_relocation_add
 			(&sw_context->res_relocations, res,
-			 id - sw_context->buf_start);
+			 id_loc - sw_context->buf_start);
 	}
 
 	ret = vmw_user_resource_lookup_handle(dev_priv,
-					      sw_context->tfile,
-					      *id,
+					      sw_context->fp->tfile,
+					      id,
 					      converter,
 					      &res);
 	if (unlikely(ret != 0)) {
 		DRM_ERROR("Could not find or use resource 0x%08x.\n",
-			  (unsigned) *id);
+			  (unsigned) id);
 		dump_stack();
 		return ret;
 	}
 
 	rcache->valid = true;
 	rcache->res = res;
-	rcache->handle = *id;
+	rcache->handle = id;
 
 	ret = vmw_resource_relocation_add(&sw_context->res_relocations,
 					  res,
-					  id - sw_context->buf_start);
+					  id_loc - sw_context->buf_start);
 	if (unlikely(ret != 0))
 		goto out_no_reloc;
 
@@ -459,7 +508,11 @@ static int vmw_cmd_res_check(struct vmw_private *dev_priv,
 	if (p_val)
 		*p_val = node;
 
-	if (node->first_usage && res_type == vmw_res_context) {
+	if (dev_priv->has_mob && node->first_usage &&
+	    res_type == vmw_res_context) {
+		ret = vmw_resource_context_res_add(dev_priv, sw_context, res);
+		if (unlikely(ret != 0))
+			goto out_no_reloc;
 		node->staged_bindings =
 			kzalloc(sizeof(*node->staged_bindings), GFP_KERNEL);
 		if (node->staged_bindings == NULL) {
@@ -481,6 +534,59 @@ out_no_reloc:
 }
 
 /**
+ * vmw_cmd_res_check - Check that a resource is present and if so, put it
+ * on the resource validate list unless it's already there.
+ *
+ * @dev_priv: Pointer to a device private structure.
+ * @sw_context: Pointer to the software context.
+ * @res_type: Resource type.
+ * @converter: User-space visisble type specific information.
+ * @id_loc: Pointer to the location in the command buffer currently being
+ * parsed from where the user-space resource id handle is located.
+ * @p_val: Pointer to pointer to resource validalidation node. Populated
+ * on exit.
+ */
+static int
+vmw_cmd_res_check(struct vmw_private *dev_priv,
+		  struct vmw_sw_context *sw_context,
+		  enum vmw_res_type res_type,
+		  const struct vmw_user_resource_conv *converter,
+		  uint32_t *id_loc,
+		  struct vmw_resource_val_node **p_val)
+{
+	return vmw_cmd_compat_res_check(dev_priv, sw_context, res_type,
+					converter, *id_loc, id_loc, p_val);
+}
+
+/**
+ * vmw_rebind_contexts - Rebind all resources previously bound to
+ * referenced contexts.
+ *
+ * @sw_context: Pointer to the software context.
+ *
+ * Rebind context binding points that have been scrubbed because of eviction.
+ */
+static int vmw_rebind_contexts(struct vmw_sw_context *sw_context)
+{
+	struct vmw_resource_val_node *val;
+	int ret;
+
+	list_for_each_entry(val, &sw_context->resource_list, head) {
+		if (likely(!val->staged_bindings))
+			continue;
+
+		ret = vmw_context_rebind_all(val->res);
+		if (unlikely(ret != 0)) {
+			if (ret != -ERESTARTSYS)
+				DRM_ERROR("Failed to rebind context.\n");
+			return ret;
+		}
+	}
+
+	return 0;
+}
+
+/**
  * vmw_cmd_cid_check - Check a command header for valid context information.
  *
  * @dev_priv: Pointer to a device private structure.
@@ -767,7 +873,7 @@ static int vmw_translate_mob_ptr(struct vmw_private *dev_priv,
 	struct vmw_relocation *reloc;
 	int ret;
 
-	ret = vmw_user_dmabuf_lookup(sw_context->tfile, handle, &vmw_bo);
+	ret = vmw_user_dmabuf_lookup(sw_context->fp->tfile, handle, &vmw_bo);
 	if (unlikely(ret != 0)) {
 		DRM_ERROR("Could not find or use MOB buffer.\n");
 		return -EINVAL;
@@ -828,7 +934,7 @@ static int vmw_translate_guest_ptr(struct vmw_private *dev_priv,
 	struct vmw_relocation *reloc;
 	int ret;
 
-	ret = vmw_user_dmabuf_lookup(sw_context->tfile, handle, &vmw_bo);
+	ret = vmw_user_dmabuf_lookup(sw_context->fp->tfile, handle, &vmw_bo);
 	if (unlikely(ret != 0)) {
 		DRM_ERROR("Could not find or use GMR region.\n");
 		return -EINVAL;
@@ -1127,7 +1233,8 @@ static int vmw_cmd_dma(struct vmw_private *dev_priv,
 
 	srf = vmw_res_to_srf(sw_context->res_cache[vmw_res_surface].res);
 
-	vmw_kms_cursor_snoop(srf, sw_context->tfile, &vmw_bo->base, header);
+	vmw_kms_cursor_snoop(srf, sw_context->fp->tfile, &vmw_bo->base,
+			     header);
 
 out_no_surface:
 	vmw_dmabuf_unreference(&vmw_bo);
@@ -1478,6 +1585,98 @@ static int vmw_cmd_invalidate_gb_surface(struct vmw_private *dev_priv,
 				 &cmd->body.sid, NULL);
 }
 
+
+/**
+ * vmw_cmd_shader_define - Validate an SVGA_3D_CMD_SHADER_DEFINE
+ * command
+ *
+ * @dev_priv: Pointer to a device private struct.
+ * @sw_context: The software context being used for this batch.
+ * @header: Pointer to the command header in the command stream.
+ */
+static int vmw_cmd_shader_define(struct vmw_private *dev_priv,
+				 struct vmw_sw_context *sw_context,
+				 SVGA3dCmdHeader *header)
+{
+	struct vmw_shader_define_cmd {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdDefineShader body;
+	} *cmd;
+	int ret;
+	size_t size;
+
+	cmd = container_of(header, struct vmw_shader_define_cmd,
+			   header);
+
+	ret = vmw_cmd_res_check(dev_priv, sw_context, vmw_res_context,
+				user_context_converter, &cmd->body.cid,
+				NULL);
+	if (unlikely(ret != 0))
+		return ret;
+
+	if (unlikely(!dev_priv->has_mob))
+		return 0;
+
+	size = cmd->header.size - sizeof(cmd->body);
+	ret = vmw_compat_shader_add(sw_context->fp->shman,
+				    cmd->body.shid, cmd + 1,
+				    cmd->body.type, size,
+				    sw_context->fp->tfile,
+				    &sw_context->staged_shaders);
+	if (unlikely(ret != 0))
+		return ret;
+
+	return vmw_resource_relocation_add(&sw_context->res_relocations,
+					   NULL, &cmd->header.id -
+					   sw_context->buf_start);
+
+	return 0;
+}
+
+/**
+ * vmw_cmd_shader_destroy - Validate an SVGA_3D_CMD_SHADER_DESTROY
+ * command
+ *
+ * @dev_priv: Pointer to a device private struct.
+ * @sw_context: The software context being used for this batch.
+ * @header: Pointer to the command header in the command stream.
+ */
+static int vmw_cmd_shader_destroy(struct vmw_private *dev_priv,
+				  struct vmw_sw_context *sw_context,
+				  SVGA3dCmdHeader *header)
+{
+	struct vmw_shader_destroy_cmd {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdDestroyShader body;
+	} *cmd;
+	int ret;
+
+	cmd = container_of(header, struct vmw_shader_destroy_cmd,
+			   header);
+
+	ret = vmw_cmd_res_check(dev_priv, sw_context, vmw_res_context,
+				user_context_converter, &cmd->body.cid,
+				NULL);
+	if (unlikely(ret != 0))
+		return ret;
+
+	if (unlikely(!dev_priv->has_mob))
+		return 0;
+
+	ret = vmw_compat_shader_remove(sw_context->fp->shman,
+				       cmd->body.shid,
+				       cmd->body.type,
+				       &sw_context->staged_shaders);
+	if (unlikely(ret != 0))
+		return ret;
+
+	return vmw_resource_relocation_add(&sw_context->res_relocations,
+					   NULL, &cmd->header.id -
+					   sw_context->buf_start);
+
+	return 0;
+}
+
 /**
  * vmw_cmd_set_shader - Validate an SVGA_3D_CMD_SET_SHADER
  * command
@@ -1509,10 +1708,18 @@ static int vmw_cmd_set_shader(struct vmw_private *dev_priv,
 	if (dev_priv->has_mob) {
 		struct vmw_ctx_bindinfo bi;
 		struct vmw_resource_val_node *res_node;
-
-		ret = vmw_cmd_res_check(dev_priv, sw_context, vmw_res_shader,
-					user_shader_converter,
-					&cmd->body.shid, &res_node);
+		u32 shid = cmd->body.shid;
+
+		if (shid != SVGA3D_INVALID_ID)
+			(void) vmw_compat_shader_lookup(sw_context->fp->shman,
+							cmd->body.type,
+							&shid);
+
+		ret = vmw_cmd_compat_res_check(dev_priv, sw_context,
+					       vmw_res_shader,
+					       user_shader_converter,
+					       shid,
+					       &cmd->body.shid, &res_node);
 		if (unlikely(ret != 0))
 			return ret;
 
@@ -1527,6 +1734,39 @@ static int vmw_cmd_set_shader(struct vmw_private *dev_priv,
 }
 
 /**
+ * vmw_cmd_set_shader_const - Validate an SVGA_3D_CMD_SET_SHADER_CONST
+ * command
+ *
+ * @dev_priv: Pointer to a device private struct.
+ * @sw_context: The software context being used for this batch.
+ * @header: Pointer to the command header in the command stream.
+ */
+static int vmw_cmd_set_shader_const(struct vmw_private *dev_priv,
+				    struct vmw_sw_context *sw_context,
+				    SVGA3dCmdHeader *header)
+{
+	struct vmw_set_shader_const_cmd {
+		SVGA3dCmdHeader header;
+		SVGA3dCmdSetShaderConst body;
+	} *cmd;
+	int ret;
+
+	cmd = container_of(header, struct vmw_set_shader_const_cmd,
+			   header);
+
+	ret = vmw_cmd_res_check(dev_priv, sw_context, vmw_res_context,
+				user_context_converter, &cmd->body.cid,
+				NULL);
+	if (unlikely(ret != 0))
+		return ret;
+
+	if (dev_priv->has_mob)
+		header->id = SVGA_3D_CMD_SET_GB_SHADERCONSTS_INLINE;
+
+	return 0;
+}
+
+/**
  * vmw_cmd_bind_gb_shader - Validate an SVGA_3D_CMD_BIND_GB_SHADER
  * command
  *
@@ -1634,14 +1874,14 @@ static const struct vmw_cmd_entry const vmw_cmd_entries[SVGA_3D_CMD_MAX] = {
 		    true, false, false),
 	VMW_CMD_DEF(SVGA_3D_CMD_PRESENT, &vmw_cmd_present_check,
 		    false, false, false),
-	VMW_CMD_DEF(SVGA_3D_CMD_SHADER_DEFINE, &vmw_cmd_cid_check,
-		    true, true, false),
-	VMW_CMD_DEF(SVGA_3D_CMD_SHADER_DESTROY, &vmw_cmd_cid_check,
-		    true, true, false),
+	VMW_CMD_DEF(SVGA_3D_CMD_SHADER_DEFINE, &vmw_cmd_shader_define,
+		    true, false, false),
+	VMW_CMD_DEF(SVGA_3D_CMD_SHADER_DESTROY, &vmw_cmd_shader_destroy,
+		    true, false, false),
 	VMW_CMD_DEF(SVGA_3D_CMD_SET_SHADER, &vmw_cmd_set_shader,
 		    true, false, false),
-	VMW_CMD_DEF(SVGA_3D_CMD_SET_SHADER_CONST, &vmw_cmd_cid_check,
-		    true, true, false),
+	VMW_CMD_DEF(SVGA_3D_CMD_SET_SHADER_CONST, &vmw_cmd_set_shader_const,
+		    true, false, false),
 	VMW_CMD_DEF(SVGA_3D_CMD_DRAW_PRIMITIVES, &vmw_cmd_draw,
 		    true, false, false),
 	VMW_CMD_DEF(SVGA_3D_CMD_SETSCISSORRECT, &vmw_cmd_cid_check,
@@ -2171,7 +2411,7 @@ int vmw_execbuf_process(struct drm_file *file_priv,
 	} else
 		sw_context->kernel = true;
 
-	sw_context->tfile = vmw_fpriv(file_priv)->tfile;
+	sw_context->fp = vmw_fpriv(file_priv);
 	sw_context->cur_reloc = 0;
 	sw_context->cur_val_buf = 0;
 	sw_context->fence_flags = 0;
@@ -2188,16 +2428,17 @@ int vmw_execbuf_process(struct drm_file *file_priv,
 			goto out_unlock;
 		sw_context->res_ht_initialized = true;
 	}
+	INIT_LIST_HEAD(&sw_context->staged_shaders);
 
 	INIT_LIST_HEAD(&resource_list);
 	ret = vmw_cmd_check_all(dev_priv, sw_context, kernel_commands,
 				command_size);
 	if (unlikely(ret != 0))
-		goto out_err;
+		goto out_err_nores;
 
 	ret = vmw_resources_reserve(sw_context);
 	if (unlikely(ret != 0))
-		goto out_err;
+		goto out_err_nores;
 
 	ret = ttm_eu_reserve_buffers(&ticket, &sw_context->validate_nodes);
 	if (unlikely(ret != 0))
@@ -2225,6 +2466,12 @@ int vmw_execbuf_process(struct drm_file *file_priv,
 		goto out_err;
 	}
 
+	if (dev_priv->has_mob) {
+		ret = vmw_rebind_contexts(sw_context);
+		if (unlikely(ret != 0))
+			goto out_err;
+	}
+
 	cmd = vmw_fifo_reserve(dev_priv, command_size);
 	if (unlikely(cmd == NULL)) {
 		DRM_ERROR("Failed reserving fifo space for commands.\n");
@@ -2276,6 +2523,8 @@ int vmw_execbuf_process(struct drm_file *file_priv,
 	}
 
 	list_splice_init(&sw_context->resource_list, &resource_list);
+	vmw_compat_shaders_commit(sw_context->fp->shman,
+				  &sw_context->staged_shaders);
 	mutex_unlock(&dev_priv->cmdbuf_mutex);
 
 	/*
@@ -2289,10 +2538,11 @@ int vmw_execbuf_process(struct drm_file *file_priv,
 out_unlock_binding:
 	mutex_unlock(&dev_priv->binding_mutex);
 out_err:
-	vmw_resource_relocations_free(&sw_context->res_relocations);
-	vmw_free_relocations(sw_context);
 	ttm_eu_backoff_reservation(&ticket, &sw_context->validate_nodes);
+out_err_nores:
 	vmw_resource_list_unreserve(&sw_context->resource_list, true);
+	vmw_resource_relocations_free(&sw_context->res_relocations);
+	vmw_free_relocations(sw_context);
 	vmw_clear_validations(sw_context);
 	if (unlikely(dev_priv->pinned_bo != NULL &&
 		     !dev_priv->query_cid_valid))
@@ -2301,6 +2551,8 @@ out_unlock:
 	list_splice_init(&sw_context->resource_list, &resource_list);
 	error_resource = sw_context->error_resource;
 	sw_context->error_resource = NULL;
+	vmw_compat_shaders_revert(sw_context->fp->shman,
+				  &sw_context->staged_shaders);
 	mutex_unlock(&dev_priv->cmdbuf_mutex);
 
 	/*
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ioctl.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ioctl.c
index 116c497..f9881f9 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_ioctl.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ioctl.c
@@ -29,12 +29,18 @@
 #include <drm/vmwgfx_drm.h>
 #include "vmwgfx_kms.h"
 
+struct svga_3d_compat_cap {
+	SVGA3dCapsRecordHeader header;
+	SVGA3dCapPair pairs[SVGA3D_DEVCAP_MAX];
+};
+
 int vmw_getparam_ioctl(struct drm_device *dev, void *data,
 		       struct drm_file *file_priv)
 {
 	struct vmw_private *dev_priv = vmw_priv(dev);
 	struct drm_vmw_getparam_arg *param =
 	    (struct drm_vmw_getparam_arg *)data;
+	struct vmw_fpriv *vmw_fp = vmw_fpriv(file_priv);
 
 	switch (param->param) {
 	case DRM_VMW_PARAM_NUM_STREAMS:
@@ -60,6 +66,11 @@ int vmw_getparam_ioctl(struct drm_device *dev, void *data,
 		__le32 __iomem *fifo_mem = dev_priv->mmio_virt;
 		const struct vmw_fifo_state *fifo = &dev_priv->fifo;
 
+		if ((dev_priv->capabilities & SVGA_CAP_GBOBJECTS)) {
+			param->value = SVGA3D_HWVERSION_WS8_B1;
+			break;
+		}
+
 		param->value =
 			ioread32(fifo_mem +
 				 ((fifo->capabilities &
@@ -69,17 +80,26 @@ int vmw_getparam_ioctl(struct drm_device *dev, void *data,
 		break;
 	}
 	case DRM_VMW_PARAM_MAX_SURF_MEMORY:
-		param->value = dev_priv->memory_size;
+		if ((dev_priv->capabilities & SVGA_CAP_GBOBJECTS) &&
+		    !vmw_fp->gb_aware)
+			param->value = dev_priv->max_mob_pages * PAGE_SIZE / 2;
+		else
+			param->value = dev_priv->memory_size;
 		break;
 	case DRM_VMW_PARAM_3D_CAPS_SIZE:
-		if (dev_priv->capabilities & SVGA_CAP_GBOBJECTS)
-			param->value = SVGA3D_DEVCAP_MAX;
+		if ((dev_priv->capabilities & SVGA_CAP_GBOBJECTS) &&
+		    vmw_fp->gb_aware)
+			param->value = SVGA3D_DEVCAP_MAX * sizeof(uint32_t);
+		else if (dev_priv->capabilities & SVGA_CAP_GBOBJECTS)
+			param->value = sizeof(struct svga_3d_compat_cap) +
+				sizeof(uint32_t);
 		else
 			param->value = (SVGA_FIFO_3D_CAPS_LAST -
-					SVGA_FIFO_3D_CAPS + 1);
-		param->value *= sizeof(uint32_t);
+					SVGA_FIFO_3D_CAPS + 1) *
+				sizeof(uint32_t);
 		break;
 	case DRM_VMW_PARAM_MAX_MOB_MEMORY:
+		vmw_fp->gb_aware = true;
 		param->value = dev_priv->max_mob_pages * PAGE_SIZE;
 		break;
 	default:
@@ -91,6 +111,38 @@ int vmw_getparam_ioctl(struct drm_device *dev, void *data,
 	return 0;
 }
 
+static int vmw_fill_compat_cap(struct vmw_private *dev_priv, void *bounce,
+			       size_t size)
+{
+	struct svga_3d_compat_cap *compat_cap =
+		(struct svga_3d_compat_cap *) bounce;
+	unsigned int i;
+	size_t pair_offset = offsetof(struct svga_3d_compat_cap, pairs);
+	unsigned int max_size;
+
+	if (size < pair_offset)
+		return -EINVAL;
+
+	max_size = (size - pair_offset) / sizeof(SVGA3dCapPair);
+
+	if (max_size > SVGA3D_DEVCAP_MAX)
+		max_size = SVGA3D_DEVCAP_MAX;
+
+	compat_cap->header.length =
+		(pair_offset + max_size * sizeof(SVGA3dCapPair)) / sizeof(u32);
+	compat_cap->header.type = SVGA3DCAPS_RECORD_DEVCAPS;
+
+	mutex_lock(&dev_priv->hw_mutex);
+	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);
+
+	return 0;
+}
+
 
 int vmw_get_cap_3d_ioctl(struct drm_device *dev, void *data,
 			 struct drm_file *file_priv)
@@ -104,41 +156,49 @@ int vmw_get_cap_3d_ioctl(struct drm_device *dev, void *data,
 	void *bounce;
 	int ret;
 	bool gb_objects = !!(dev_priv->capabilities & SVGA_CAP_GBOBJECTS);
+	struct vmw_fpriv *vmw_fp = vmw_fpriv(file_priv);
 
 	if (unlikely(arg->pad64 != 0)) {
 		DRM_ERROR("Illegal GET_3D_CAP argument.\n");
 		return -EINVAL;
 	}
 
-	if (gb_objects)
-		size = SVGA3D_DEVCAP_MAX;
+	if (gb_objects && vmw_fp->gb_aware)
+		size = SVGA3D_DEVCAP_MAX * sizeof(uint32_t);
+	else if (gb_objects)
+		size = sizeof(struct svga_3d_compat_cap) + sizeof(uint32_t);
 	else
-		size = (SVGA_FIFO_3D_CAPS_LAST - SVGA_FIFO_3D_CAPS + 1);
-
-	size *= sizeof(uint32_t);
+		size = (SVGA_FIFO_3D_CAPS_LAST - SVGA_FIFO_3D_CAPS + 1) *
+			sizeof(uint32_t);
 
 	if (arg->max_size < size)
 		size = arg->max_size;
 
-	bounce = vmalloc(size);
+	bounce = vzalloc(size);
 	if (unlikely(bounce == NULL)) {
 		DRM_ERROR("Failed to allocate bounce buffer for 3D caps.\n");
 		return -ENOMEM;
 	}
 
-	if (gb_objects) {
-		int i;
+	if (gb_objects && vmw_fp->gb_aware) {
+		int i, num;
 		uint32_t *bounce32 = (uint32_t *) bounce;
 
+		num = size / sizeof(uint32_t);
+		if (num > SVGA3D_DEVCAP_MAX)
+			num = SVGA3D_DEVCAP_MAX;
+
 		mutex_lock(&dev_priv->hw_mutex);
-		for (i = 0; i < SVGA3D_DEVCAP_MAX; ++i) {
+		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);
-
+	} else if (gb_objects) {
+		ret = vmw_fill_compat_cap(dev_priv, bounce, size);
+		if (unlikely(ret != 0))
+			goto out_err;
 	} else {
-
 		fifo_mem = dev_priv->mmio_virt;
 		memcpy_fromio(bounce, &fifo_mem[SVGA_FIFO_3D_CAPS], size);
 	}
@@ -146,6 +206,7 @@ int vmw_get_cap_3d_ioctl(struct drm_device *dev, void *data,
 	ret = copy_to_user(buffer, bounce, size);
 	if (ret)
 		ret = -EFAULT;
+out_err:
 	vfree(bounce);
 
 	if (unlikely(ret != 0))
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_mob.c b/drivers/gpu/drm/vmwgfx/vmwgfx_mob.c
index 4910e7b..d4a5a19 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_mob.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_mob.c
@@ -134,6 +134,7 @@ static int vmw_setup_otable_base(struct vmw_private *dev_priv,
 	cmd = vmw_fifo_reserve(dev_priv, sizeof(*cmd));
 	if (unlikely(cmd == NULL)) {
 		DRM_ERROR("Failed reserving FIFO space for OTable setup.\n");
+		ret = -ENOMEM;
 		goto out_no_fifo;
 	}
 
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c
index 6fdd82d..2aa4bc6 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c
@@ -88,6 +88,11 @@ struct vmw_resource *vmw_resource_reference(struct vmw_resource *res)
 	return res;
 }
 
+struct vmw_resource *
+vmw_resource_reference_unless_doomed(struct vmw_resource *res)
+{
+	return kref_get_unless_zero(&res->kref) ? res : NULL;
+}
 
 /**
  * vmw_resource_release_id - release a resource id to the id manager.
@@ -136,8 +141,12 @@ static void vmw_resource_release(struct kref *kref)
 		vmw_dmabuf_unreference(&res->backup);
 	}
 
-	if (likely(res->hw_destroy != NULL))
+	if (likely(res->hw_destroy != NULL)) {
 		res->hw_destroy(res);
+		mutex_lock(&dev_priv->binding_mutex);
+		vmw_context_binding_res_list_kill(&res->binding_head);
+		mutex_unlock(&dev_priv->binding_mutex);
+	}
 
 	id = res->id;
 	if (res->res_free != NULL)
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_shader.c b/drivers/gpu/drm/vmwgfx/vmwgfx_shader.c
index 1457ec4b..217d941 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_shader.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_shader.c
@@ -29,6 +29,8 @@
 #include "vmwgfx_resource_priv.h"
 #include "ttm/ttm_placement.h"
 
+#define VMW_COMPAT_SHADER_HT_ORDER 12
+
 struct vmw_shader {
 	struct vmw_resource res;
 	SVGA3dShaderType type;
@@ -40,6 +42,50 @@ struct vmw_user_shader {
 	struct vmw_shader shader;
 };
 
+/**
+ * enum vmw_compat_shader_state - Staging state for compat shaders
+ */
+enum vmw_compat_shader_state {
+	VMW_COMPAT_COMMITED,
+	VMW_COMPAT_ADD,
+	VMW_COMPAT_DEL
+};
+
+/**
+ * struct vmw_compat_shader - Metadata for compat shaders.
+ *
+ * @handle: The TTM handle of the guest backed shader.
+ * @tfile: The struct ttm_object_file the guest backed shader is registered
+ * with.
+ * @hash: Hash item for lookup.
+ * @head: List head for staging lists or the compat shader manager list.
+ * @state: Staging state.
+ *
+ * The structure is protected by the cmdbuf lock.
+ */
+struct vmw_compat_shader {
+	u32 handle;
+	struct ttm_object_file *tfile;
+	struct drm_hash_item hash;
+	struct list_head head;
+	enum vmw_compat_shader_state state;
+};
+
+/**
+ * struct vmw_compat_shader_manager - Compat shader manager.
+ *
+ * @shaders: Hash table containing staged and commited compat shaders
+ * @list: List of commited shaders.
+ * @dev_priv: Pointer to a device private structure.
+ *
+ * @shaders and @list are protected by the cmdbuf mutex for now.
+ */
+struct vmw_compat_shader_manager {
+	struct drm_open_hash shaders;
+	struct list_head list;
+	struct vmw_private *dev_priv;
+};
+
 static void vmw_user_shader_free(struct vmw_resource *res);
 static struct vmw_resource *
 vmw_user_shader_base_to_res(struct ttm_base_object *base);
@@ -258,7 +304,7 @@ static int vmw_gb_shader_destroy(struct vmw_resource *res)
 		return 0;
 
 	mutex_lock(&dev_priv->binding_mutex);
-	vmw_context_binding_res_list_kill(&res->binding_head);
+	vmw_context_binding_res_list_scrub(&res->binding_head);
 
 	cmd = vmw_fifo_reserve(dev_priv, sizeof(*cmd));
 	if (unlikely(cmd == NULL)) {
@@ -325,13 +371,81 @@ int vmw_shader_destroy_ioctl(struct drm_device *dev, void *data,
 					 TTM_REF_USAGE);
 }
 
+int vmw_shader_alloc(struct vmw_private *dev_priv,
+		     struct vmw_dma_buffer *buffer,
+		     size_t shader_size,
+		     size_t offset,
+		     SVGA3dShaderType shader_type,
+		     struct ttm_object_file *tfile,
+		     u32 *handle)
+{
+	struct vmw_user_shader *ushader;
+	struct vmw_resource *res, *tmp;
+	int ret;
+
+	/*
+	 * Approximate idr memory usage with 128 bytes. It will be limited
+	 * by maximum number_of shaders anyway.
+	 */
+	if (unlikely(vmw_user_shader_size == 0))
+		vmw_user_shader_size =
+			ttm_round_pot(sizeof(struct vmw_user_shader)) + 128;
+
+	ret = ttm_mem_global_alloc(vmw_mem_glob(dev_priv),
+				   vmw_user_shader_size,
+				   false, true);
+	if (unlikely(ret != 0)) {
+		if (ret != -ERESTARTSYS)
+			DRM_ERROR("Out of graphics memory for shader "
+				  "creation.\n");
+		goto out;
+	}
+
+	ushader = kzalloc(sizeof(*ushader), GFP_KERNEL);
+	if (unlikely(ushader == NULL)) {
+		ttm_mem_global_free(vmw_mem_glob(dev_priv),
+				    vmw_user_shader_size);
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	res = &ushader->shader.res;
+	ushader->base.shareable = false;
+	ushader->base.tfile = NULL;
+
+	/*
+	 * From here on, the destructor takes over resource freeing.
+	 */
+
+	ret = vmw_gb_shader_init(dev_priv, res, shader_size,
+				 offset, shader_type, buffer,
+				 vmw_user_shader_free);
+	if (unlikely(ret != 0))
+		goto out;
+
+	tmp = vmw_resource_reference(res);
+	ret = ttm_base_object_init(tfile, &ushader->base, false,
+				   VMW_RES_SHADER,
+				   &vmw_user_shader_base_release, NULL);
+
+	if (unlikely(ret != 0)) {
+		vmw_resource_unreference(&tmp);
+		goto out_err;
+	}
+
+	if (handle)
+		*handle = ushader->base.hash.key;
+out_err:
+	vmw_resource_unreference(&res);
+out:
+	return ret;
+}
+
+
 int vmw_shader_define_ioctl(struct drm_device *dev, void *data,
 			     struct drm_file *file_priv)
 {
 	struct vmw_private *dev_priv = vmw_priv(dev);
-	struct vmw_user_shader *ushader;
-	struct vmw_resource *res;
-	struct vmw_resource *tmp;
 	struct drm_vmw_shader_create_arg *arg =
 		(struct drm_vmw_shader_create_arg *)data;
 	struct ttm_object_file *tfile = vmw_fpriv(file_priv)->tfile;
@@ -373,69 +487,324 @@ int vmw_shader_define_ioctl(struct drm_device *dev, void *data,
 		goto out_bad_arg;
 	}
 
-	/*
-	 * Approximate idr memory usage with 128 bytes. It will be limited
-	 * by maximum number_of shaders anyway.
-	 */
+	ret = ttm_read_lock(&vmaster->lock, true);
+	if (unlikely(ret != 0))
+		goto out_bad_arg;
 
-	if (unlikely(vmw_user_shader_size == 0))
-		vmw_user_shader_size = ttm_round_pot(sizeof(*ushader))
-			+ 128;
+	ret = vmw_shader_alloc(dev_priv, buffer, arg->size, arg->offset,
+			       shader_type, tfile, &arg->shader_handle);
 
-	ret = ttm_read_lock(&vmaster->lock, true);
+	ttm_read_unlock(&vmaster->lock);
+out_bad_arg:
+	vmw_dmabuf_unreference(&buffer);
+	return ret;
+}
+
+/**
+ * vmw_compat_shader_lookup - Look up a compat shader
+ *
+ * @man: Pointer to the compat shader manager.
+ * @shader_type: The shader type, that combined with the user_key identifies
+ * the shader.
+ * @user_key: On entry, this should be a pointer to the user_key.
+ * On successful exit, it will contain the guest-backed shader's TTM handle.
+ *
+ * Returns 0 on success. Non-zero on failure, in which case the value pointed
+ * to by @user_key is unmodified.
+ */
+int vmw_compat_shader_lookup(struct vmw_compat_shader_manager *man,
+			     SVGA3dShaderType shader_type,
+			     u32 *user_key)
+{
+	struct drm_hash_item *hash;
+	int ret;
+	unsigned long key = *user_key | (shader_type << 24);
+
+	ret = drm_ht_find_item(&man->shaders, key, &hash);
 	if (unlikely(ret != 0))
 		return ret;
 
-	ret = ttm_mem_global_alloc(vmw_mem_glob(dev_priv),
-				   vmw_user_shader_size,
-				   false, true);
-	if (unlikely(ret != 0)) {
-		if (ret != -ERESTARTSYS)
-			DRM_ERROR("Out of graphics memory for shader"
-				  " creation.\n");
-		goto out_unlock;
+	*user_key = drm_hash_entry(hash, struct vmw_compat_shader,
+				   hash)->handle;
+
+	return 0;
+}
+
+/**
+ * vmw_compat_shader_free - Free a compat shader.
+ *
+ * @man: Pointer to the compat shader manager.
+ * @entry: Pointer to a struct vmw_compat_shader.
+ *
+ * Frees a struct vmw_compat_shder entry and drops its reference to the
+ * guest backed shader.
+ */
+static void vmw_compat_shader_free(struct vmw_compat_shader_manager *man,
+				   struct vmw_compat_shader *entry)
+{
+	list_del(&entry->head);
+	WARN_ON(drm_ht_remove_item(&man->shaders, &entry->hash));
+	WARN_ON(ttm_ref_object_base_unref(entry->tfile, entry->handle,
+					  TTM_REF_USAGE));
+	kfree(entry);
+}
+
+/**
+ * vmw_compat_shaders_commit - Commit a list of compat shader actions.
+ *
+ * @man: Pointer to the compat shader manager.
+ * @list: Caller's list of compat shader actions.
+ *
+ * This function commits a list of compat shader additions or removals.
+ * It is typically called when the execbuf ioctl call triggering these
+ * actions has commited the fifo contents to the device.
+ */
+void vmw_compat_shaders_commit(struct vmw_compat_shader_manager *man,
+			       struct list_head *list)
+{
+	struct vmw_compat_shader *entry, *next;
+
+	list_for_each_entry_safe(entry, next, list, head) {
+		list_del(&entry->head);
+		switch (entry->state) {
+		case VMW_COMPAT_ADD:
+			entry->state = VMW_COMPAT_COMMITED;
+			list_add_tail(&entry->head, &man->list);
+			break;
+		case VMW_COMPAT_DEL:
+			ttm_ref_object_base_unref(entry->tfile, entry->handle,
+						  TTM_REF_USAGE);
+			kfree(entry);
+			break;
+		default:
+			BUG();
+			break;
+		}
 	}
+}
 
-	ushader = kzalloc(sizeof(*ushader), GFP_KERNEL);
-	if (unlikely(ushader == NULL)) {
-		ttm_mem_global_free(vmw_mem_glob(dev_priv),
-				    vmw_user_shader_size);
-		ret = -ENOMEM;
-		goto out_unlock;
+/**
+ * vmw_compat_shaders_revert - Revert a list of compat shader actions
+ *
+ * @man: Pointer to the compat shader manager.
+ * @list: Caller's list of compat shader actions.
+ *
+ * This function reverts a list of compat shader additions or removals.
+ * It is typically called when the execbuf ioctl call triggering these
+ * actions failed for some reason, and the command stream was never
+ * submitted.
+ */
+void vmw_compat_shaders_revert(struct vmw_compat_shader_manager *man,
+			       struct list_head *list)
+{
+	struct vmw_compat_shader *entry, *next;
+	int ret;
+
+	list_for_each_entry_safe(entry, next, list, head) {
+		switch (entry->state) {
+		case VMW_COMPAT_ADD:
+			vmw_compat_shader_free(man, entry);
+			break;
+		case VMW_COMPAT_DEL:
+			ret = drm_ht_insert_item(&man->shaders, &entry->hash);
+			list_del(&entry->head);
+			list_add_tail(&entry->head, &man->list);
+			entry->state = VMW_COMPAT_COMMITED;
+			break;
+		default:
+			BUG();
+			break;
+		}
 	}
+}
 
-	res = &ushader->shader.res;
-	ushader->base.shareable = false;
-	ushader->base.tfile = NULL;
+/**
+ * vmw_compat_shader_remove - Stage a compat shader for removal.
+ *
+ * @man: Pointer to the compat shader manager
+ * @user_key: The key that is used to identify the shader. The key is
+ * unique to the shader type.
+ * @shader_type: Shader type.
+ * @list: Caller's list of staged shader actions.
+ *
+ * This function stages a compat shader for removal and removes the key from
+ * the shader manager's hash table. If the shader was previously only staged
+ * for addition it is completely removed (But the execbuf code may keep a
+ * reference if it was bound to a context between addition and removal). If
+ * it was previously commited to the manager, it is staged for removal.
+ */
+int vmw_compat_shader_remove(struct vmw_compat_shader_manager *man,
+			     u32 user_key, SVGA3dShaderType shader_type,
+			     struct list_head *list)
+{
+	struct vmw_compat_shader *entry;
+	struct drm_hash_item *hash;
+	int ret;
 
-	/*
-	 * From here on, the destructor takes over resource freeing.
-	 */
+	ret = drm_ht_find_item(&man->shaders, user_key | (shader_type << 24),
+			       &hash);
+	if (likely(ret != 0))
+		return -EINVAL;
 
-	ret = vmw_gb_shader_init(dev_priv, res, arg->size,
-				 arg->offset, shader_type, buffer,
-				 vmw_user_shader_free);
+	entry = drm_hash_entry(hash, struct vmw_compat_shader, hash);
+
+	switch (entry->state) {
+	case VMW_COMPAT_ADD:
+		vmw_compat_shader_free(man, entry);
+		break;
+	case VMW_COMPAT_COMMITED:
+		(void) drm_ht_remove_item(&man->shaders, &entry->hash);
+		list_del(&entry->head);
+		entry->state = VMW_COMPAT_DEL;
+		list_add_tail(&entry->head, list);
+		break;
+	default:
+		BUG();
+		break;
+	}
+
+	return 0;
+}
+
+/**
+ * vmw_compat_shader_add - Create a compat shader and add the
+ * key to the manager
+ *
+ * @man: Pointer to the compat shader manager
+ * @user_key: The key that is used to identify the shader. The key is
+ * unique to the shader type.
+ * @bytecode: Pointer to the bytecode of the shader.
+ * @shader_type: Shader type.
+ * @tfile: Pointer to a struct ttm_object_file that the guest-backed shader is
+ * to be created with.
+ * @list: Caller's list of staged shader actions.
+ *
+ * Note that only the key is added to the shader manager's hash table.
+ * The shader is not yet added to the shader manager's list of shaders.
+ */
+int vmw_compat_shader_add(struct vmw_compat_shader_manager *man,
+			  u32 user_key, const void *bytecode,
+			  SVGA3dShaderType shader_type,
+			  size_t size,
+			  struct ttm_object_file *tfile,
+			  struct list_head *list)
+{
+	struct vmw_dma_buffer *buf;
+	struct ttm_bo_kmap_obj map;
+	bool is_iomem;
+	struct vmw_compat_shader *compat;
+	u32 handle;
+	int ret;
+
+	if (user_key > ((1 << 24) - 1) || (unsigned) shader_type > 16)
+		return -EINVAL;
+
+	/* Allocate and pin a DMA buffer */
+	buf = kzalloc(sizeof(*buf), GFP_KERNEL);
+	if (unlikely(buf == NULL))
+		return -ENOMEM;
+
+	ret = vmw_dmabuf_init(man->dev_priv, buf, size, &vmw_sys_ne_placement,
+			      true, vmw_dmabuf_bo_free);
 	if (unlikely(ret != 0))
-		goto out_unlock;
+		goto out;
 
-	tmp = vmw_resource_reference(res);
-	ret = ttm_base_object_init(tfile, &ushader->base, false,
-				   VMW_RES_SHADER,
-				   &vmw_user_shader_base_release, NULL);
+	ret = ttm_bo_reserve(&buf->base, false, true, false, NULL);
+	if (unlikely(ret != 0))
+		goto no_reserve;
 
+	/* Map and copy shader bytecode. */
+	ret = ttm_bo_kmap(&buf->base, 0, PAGE_ALIGN(size) >> PAGE_SHIFT,
+			  &map);
 	if (unlikely(ret != 0)) {
-		vmw_resource_unreference(&tmp);
-		goto out_err;
+		ttm_bo_unreserve(&buf->base);
+		goto no_reserve;
 	}
 
-	arg->shader_handle = ushader->base.hash.key;
-out_err:
-	vmw_resource_unreference(&res);
-out_unlock:
-	ttm_read_unlock(&vmaster->lock);
-out_bad_arg:
-	vmw_dmabuf_unreference(&buffer);
+	memcpy(ttm_kmap_obj_virtual(&map, &is_iomem), bytecode, size);
+	WARN_ON(is_iomem);
+
+	ttm_bo_kunmap(&map);
+	ret = ttm_bo_validate(&buf->base, &vmw_sys_placement, false, true);
+	WARN_ON(ret != 0);
+	ttm_bo_unreserve(&buf->base);
+
+	/* Create a guest-backed shader container backed by the dma buffer */
+	ret = vmw_shader_alloc(man->dev_priv, buf, size, 0, shader_type,
+			       tfile, &handle);
+	vmw_dmabuf_unreference(&buf);
+	if (unlikely(ret != 0))
+		goto no_reserve;
+	/*
+	 * Create a compat shader structure and stage it for insertion
+	 * in the manager
+	 */
+	compat = kzalloc(sizeof(*compat), GFP_KERNEL);
+	if (compat == NULL)
+		goto no_compat;
+
+	compat->hash.key = user_key |  (shader_type << 24);
+	ret = drm_ht_insert_item(&man->shaders, &compat->hash);
+	if (unlikely(ret != 0))
+		goto out_invalid_key;
+
+	compat->state = VMW_COMPAT_ADD;
+	compat->handle = handle;
+	compat->tfile = tfile;
+	list_add_tail(&compat->head, list);
 
+	return 0;
+
+out_invalid_key:
+	kfree(compat);
+no_compat:
+	ttm_ref_object_base_unref(tfile, handle, TTM_REF_USAGE);
+no_reserve:
+out:
 	return ret;
+}
+
+/**
+ * vmw_compat_shader_man_create - Create a compat shader manager
+ *
+ * @dev_priv: Pointer to a device private structure.
+ *
+ * Typically done at file open time. If successful returns a pointer to a
+ * compat shader manager. Otherwise returns an error pointer.
+ */
+struct vmw_compat_shader_manager *
+vmw_compat_shader_man_create(struct vmw_private *dev_priv)
+{
+	struct vmw_compat_shader_manager *man;
+	int ret;
+
+	man = kzalloc(sizeof(*man), GFP_KERNEL);
+
+	man->dev_priv = dev_priv;
+	INIT_LIST_HEAD(&man->list);
+	ret = drm_ht_create(&man->shaders, VMW_COMPAT_SHADER_HT_ORDER);
+	if (ret == 0)
+		return man;
+
+	kfree(man);
+	return ERR_PTR(ret);
+}
+
+/**
+ * vmw_compat_shader_man_destroy - Destroy a compat shader manager
+ *
+ * @man: Pointer to the shader manager to destroy.
+ *
+ * Typically done at file close time.
+ */
+void vmw_compat_shader_man_destroy(struct vmw_compat_shader_manager *man)
+{
+	struct vmw_compat_shader *entry, *next;
+
+	mutex_lock(&man->dev_priv->cmdbuf_mutex);
+	list_for_each_entry_safe(entry, next, &man->list, head)
+		vmw_compat_shader_free(man, entry);
 
+	mutex_unlock(&man->dev_priv->cmdbuf_mutex);
+	kfree(man);
 }
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_surface.c b/drivers/gpu/drm/vmwgfx/vmwgfx_surface.c
index 979da1c..82468d9 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_surface.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_surface.c
@@ -908,8 +908,8 @@ int vmw_surface_reference_ioctl(struct drm_device *dev, void *data,
 	    rep->size_addr;
 
 	if (user_sizes)
-		ret = copy_to_user(user_sizes, srf->sizes,
-				   srf->num_sizes * sizeof(*srf->sizes));
+		ret = copy_to_user(user_sizes, &srf->base_size,
+				   sizeof(srf->base_size));
 	if (unlikely(ret != 0)) {
 		DRM_ERROR("copy_to_user failed %p %u\n",
 			  user_sizes, srf->num_sizes);
@@ -1111,7 +1111,7 @@ static int vmw_gb_surface_destroy(struct vmw_resource *res)
 		return 0;
 
 	mutex_lock(&dev_priv->binding_mutex);
-	vmw_context_binding_res_list_kill(&res->binding_head);
+	vmw_context_binding_res_list_scrub(&res->binding_head);
 
 	cmd = vmw_fifo_reserve(dev_priv, sizeof(*cmd));
 	if (unlikely(cmd == NULL)) {