diff --git a/drivers/gpu/drm/radeon/evergreen_cs.c b/drivers/gpu/drm/radeon/evergreen_cs.c
index cd4590a..f7442e6 100644
--- a/drivers/gpu/drm/radeon/evergreen_cs.c
+++ b/drivers/gpu/drm/radeon/evergreen_cs.c
@@ -520,7 +520,7 @@
 		break;
 	case DB_Z_INFO:
 		track->db_z_info = radeon_get_ib_value(p, idx);
-		if (!p->keep_tiling_flags) {
+		if (!(p->cs_flags & RADEON_CS_KEEP_TILING_FLAGS)) {
 			r = evergreen_cs_packet_next_reloc(p, &reloc);
 			if (r) {
 				dev_warn(p->dev, "bad SET_CONTEXT_REG "
@@ -649,7 +649,7 @@
 	case CB_COLOR7_INFO:
 		tmp = (reg - CB_COLOR0_INFO) / 0x3c;
 		track->cb_color_info[tmp] = radeon_get_ib_value(p, idx);
-		if (!p->keep_tiling_flags) {
+		if (!(p->cs_flags & RADEON_CS_KEEP_TILING_FLAGS)) {
 			r = evergreen_cs_packet_next_reloc(p, &reloc);
 			if (r) {
 				dev_warn(p->dev, "bad SET_CONTEXT_REG "
@@ -666,7 +666,7 @@
 	case CB_COLOR11_INFO:
 		tmp = ((reg - CB_COLOR8_INFO) / 0x1c) + 8;
 		track->cb_color_info[tmp] = radeon_get_ib_value(p, idx);
-		if (!p->keep_tiling_flags) {
+		if (!(p->cs_flags & RADEON_CS_KEEP_TILING_FLAGS)) {
 			r = evergreen_cs_packet_next_reloc(p, &reloc);
 			if (r) {
 				dev_warn(p->dev, "bad SET_CONTEXT_REG "
@@ -1355,7 +1355,7 @@
 					return -EINVAL;
 				}
 				ib[idx+1+(i*8)+2] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
-				if (!p->keep_tiling_flags) {
+				if (!(p->cs_flags & RADEON_CS_KEEP_TILING_FLAGS)) {
 					ib[idx+1+(i*8)+1] |=
 						TEX_ARRAY_MODE(evergreen_cs_get_aray_mode(reloc->lobj.tiling_flags));
 					if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) {
@@ -1572,3 +1572,241 @@
 	return 0;
 }
 
+/* vm parser */
+static bool evergreen_vm_reg_valid(u32 reg)
+{
+	/* context regs are fine */
+	if (reg >= 0x28000)
+		return true;
+
+	/* check config regs */
+	switch (reg) {
+	case GRBM_GFX_INDEX:
+	case VGT_VTX_VECT_EJECT_REG:
+	case VGT_CACHE_INVALIDATION:
+	case VGT_GS_VERTEX_REUSE:
+	case VGT_PRIMITIVE_TYPE:
+	case VGT_INDEX_TYPE:
+	case VGT_NUM_INDICES:
+	case VGT_NUM_INSTANCES:
+	case VGT_COMPUTE_DIM_X:
+	case VGT_COMPUTE_DIM_Y:
+	case VGT_COMPUTE_DIM_Z:
+	case VGT_COMPUTE_START_X:
+	case VGT_COMPUTE_START_Y:
+	case VGT_COMPUTE_START_Z:
+	case VGT_COMPUTE_INDEX:
+	case VGT_COMPUTE_THREAD_GROUP_SIZE:
+	case VGT_HS_OFFCHIP_PARAM:
+	case PA_CL_ENHANCE:
+	case PA_SU_LINE_STIPPLE_VALUE:
+	case PA_SC_LINE_STIPPLE_STATE:
+	case PA_SC_ENHANCE:
+	case SQ_DYN_GPR_CNTL_PS_FLUSH_REQ:
+	case SQ_DYN_GPR_SIMD_LOCK_EN:
+	case SQ_CONFIG:
+	case SQ_GPR_RESOURCE_MGMT_1:
+	case SQ_GLOBAL_GPR_RESOURCE_MGMT_1:
+	case SQ_GLOBAL_GPR_RESOURCE_MGMT_2:
+	case SQ_CONST_MEM_BASE:
+	case SQ_STATIC_THREAD_MGMT_1:
+	case SQ_STATIC_THREAD_MGMT_2:
+	case SQ_STATIC_THREAD_MGMT_3:
+	case SPI_CONFIG_CNTL:
+	case SPI_CONFIG_CNTL_1:
+	case TA_CNTL_AUX:
+	case DB_DEBUG:
+	case DB_DEBUG2:
+	case DB_DEBUG3:
+	case DB_DEBUG4:
+	case DB_WATERMARKS:
+	case TD_PS_BORDER_COLOR_INDEX:
+	case TD_PS_BORDER_COLOR_RED:
+	case TD_PS_BORDER_COLOR_GREEN:
+	case TD_PS_BORDER_COLOR_BLUE:
+	case TD_PS_BORDER_COLOR_ALPHA:
+	case TD_VS_BORDER_COLOR_INDEX:
+	case TD_VS_BORDER_COLOR_RED:
+	case TD_VS_BORDER_COLOR_GREEN:
+	case TD_VS_BORDER_COLOR_BLUE:
+	case TD_VS_BORDER_COLOR_ALPHA:
+	case TD_GS_BORDER_COLOR_INDEX:
+	case TD_GS_BORDER_COLOR_RED:
+	case TD_GS_BORDER_COLOR_GREEN:
+	case TD_GS_BORDER_COLOR_BLUE:
+	case TD_GS_BORDER_COLOR_ALPHA:
+	case TD_HS_BORDER_COLOR_INDEX:
+	case TD_HS_BORDER_COLOR_RED:
+	case TD_HS_BORDER_COLOR_GREEN:
+	case TD_HS_BORDER_COLOR_BLUE:
+	case TD_HS_BORDER_COLOR_ALPHA:
+	case TD_LS_BORDER_COLOR_INDEX:
+	case TD_LS_BORDER_COLOR_RED:
+	case TD_LS_BORDER_COLOR_GREEN:
+	case TD_LS_BORDER_COLOR_BLUE:
+	case TD_LS_BORDER_COLOR_ALPHA:
+	case TD_CS_BORDER_COLOR_INDEX:
+	case TD_CS_BORDER_COLOR_RED:
+	case TD_CS_BORDER_COLOR_GREEN:
+	case TD_CS_BORDER_COLOR_BLUE:
+	case TD_CS_BORDER_COLOR_ALPHA:
+	case SQ_ESGS_RING_SIZE:
+	case SQ_GSVS_RING_SIZE:
+	case SQ_ESTMP_RING_SIZE:
+	case SQ_GSTMP_RING_SIZE:
+	case SQ_HSTMP_RING_SIZE:
+	case SQ_LSTMP_RING_SIZE:
+	case SQ_PSTMP_RING_SIZE:
+	case SQ_VSTMP_RING_SIZE:
+	case SQ_ESGS_RING_ITEMSIZE:
+	case SQ_ESTMP_RING_ITEMSIZE:
+	case SQ_GSTMP_RING_ITEMSIZE:
+	case SQ_GSVS_RING_ITEMSIZE:
+	case SQ_GS_VERT_ITEMSIZE:
+	case SQ_GS_VERT_ITEMSIZE_1:
+	case SQ_GS_VERT_ITEMSIZE_2:
+	case SQ_GS_VERT_ITEMSIZE_3:
+	case SQ_GSVS_RING_OFFSET_1:
+	case SQ_GSVS_RING_OFFSET_2:
+	case SQ_GSVS_RING_OFFSET_3:
+	case SQ_HSTMP_RING_ITEMSIZE:
+	case SQ_LSTMP_RING_ITEMSIZE:
+	case SQ_PSTMP_RING_ITEMSIZE:
+	case SQ_VSTMP_RING_ITEMSIZE:
+	case VGT_TF_RING_SIZE:
+	case SQ_ESGS_RING_BASE:
+	case SQ_GSVS_RING_BASE:
+	case SQ_ESTMP_RING_BASE:
+	case SQ_GSTMP_RING_BASE:
+	case SQ_HSTMP_RING_BASE:
+	case SQ_LSTMP_RING_BASE:
+	case SQ_PSTMP_RING_BASE:
+	case SQ_VSTMP_RING_BASE:
+	case CAYMAN_VGT_OFFCHIP_LDS_BASE:
+	case CAYMAN_SQ_EX_ALLOC_TABLE_SLOTS:
+		return true;
+	default:
+		return false;
+	}
+}
+
+static int evergreen_vm_packet3_check(struct radeon_device *rdev,
+				      u32 *ib, struct radeon_cs_packet *pkt)
+{
+	u32 idx = pkt->idx + 1;
+	u32 idx_value = ib[idx];
+	u32 start_reg, end_reg, reg, i;
+
+	switch (pkt->opcode) {
+	case PACKET3_NOP:
+	case PACKET3_SET_BASE:
+	case PACKET3_CLEAR_STATE:
+	case PACKET3_INDEX_BUFFER_SIZE:
+	case PACKET3_DISPATCH_DIRECT:
+	case PACKET3_DISPATCH_INDIRECT:
+	case PACKET3_MODE_CONTROL:
+	case PACKET3_SET_PREDICATION:
+	case PACKET3_COND_EXEC:
+	case PACKET3_PRED_EXEC:
+	case PACKET3_DRAW_INDIRECT:
+	case PACKET3_DRAW_INDEX_INDIRECT:
+	case PACKET3_INDEX_BASE:
+	case PACKET3_DRAW_INDEX_2:
+	case PACKET3_CONTEXT_CONTROL:
+	case PACKET3_DRAW_INDEX_OFFSET:
+	case PACKET3_INDEX_TYPE:
+	case PACKET3_DRAW_INDEX:
+	case PACKET3_DRAW_INDEX_AUTO:
+	case PACKET3_DRAW_INDEX_IMMD:
+	case PACKET3_NUM_INSTANCES:
+	case PACKET3_DRAW_INDEX_MULTI_AUTO:
+	case PACKET3_STRMOUT_BUFFER_UPDATE:
+	case PACKET3_DRAW_INDEX_OFFSET_2:
+	case PACKET3_DRAW_INDEX_MULTI_ELEMENT:
+	case PACKET3_MPEG_INDEX:
+	case PACKET3_WAIT_REG_MEM:
+	case PACKET3_MEM_WRITE:
+	case PACKET3_SURFACE_SYNC:
+	case PACKET3_EVENT_WRITE:
+	case PACKET3_EVENT_WRITE_EOP:
+	case PACKET3_EVENT_WRITE_EOS:
+	case PACKET3_SET_CONTEXT_REG:
+	case PACKET3_SET_BOOL_CONST:
+	case PACKET3_SET_LOOP_CONST:
+	case PACKET3_SET_RESOURCE:
+	case PACKET3_SET_SAMPLER:
+	case PACKET3_SET_CTL_CONST:
+	case PACKET3_SET_RESOURCE_OFFSET:
+	case PACKET3_SET_CONTEXT_REG_INDIRECT:
+	case PACKET3_SET_RESOURCE_INDIRECT:
+	case CAYMAN_PACKET3_DEALLOC_STATE:
+		break;
+	case PACKET3_COND_WRITE:
+		if (idx_value & 0x100) {
+			reg = ib[idx + 5] * 4;
+			if (!evergreen_vm_reg_valid(reg))
+				return -EINVAL;
+		}
+		break;
+	case PACKET3_COPY_DW:
+		if (idx_value & 0x2) {
+			reg = ib[idx + 3] * 4;
+			if (!evergreen_vm_reg_valid(reg))
+				return -EINVAL;
+		}
+		break;
+	case PACKET3_SET_CONFIG_REG:
+		start_reg = (idx_value << 2) + PACKET3_SET_CONFIG_REG_START;
+		end_reg = 4 * pkt->count + start_reg - 4;
+		if ((start_reg < PACKET3_SET_CONFIG_REG_START) ||
+		    (start_reg >= PACKET3_SET_CONFIG_REG_END) ||
+		    (end_reg >= PACKET3_SET_CONFIG_REG_END)) {
+			DRM_ERROR("bad PACKET3_SET_CONFIG_REG\n");
+			return -EINVAL;
+		}
+		for (i = 0; i < pkt->count; i++) {
+			reg = start_reg + (4 * i);
+			if (!evergreen_vm_reg_valid(reg))
+				return -EINVAL;
+		}
+		break;
+	default:
+		return -EINVAL;
+	}
+	return 0;
+}
+
+int evergreen_ib_parse(struct radeon_device *rdev, struct radeon_ib *ib)
+{
+	int ret = 0;
+	u32 idx = 0;
+	struct radeon_cs_packet pkt;
+
+	do {
+		pkt.idx = idx;
+		pkt.type = CP_PACKET_GET_TYPE(ib->ptr[idx]);
+		pkt.count = CP_PACKET_GET_COUNT(ib->ptr[idx]);
+		pkt.one_reg_wr = 0;
+		switch (pkt.type) {
+		case PACKET_TYPE0:
+			dev_err(rdev->dev, "Packet0 not allowed!\n");
+			ret = -EINVAL;
+			break;
+		case PACKET_TYPE2:
+			break;
+		case PACKET_TYPE3:
+			pkt.opcode = CP_PACKET3_GET_OPCODE(ib->ptr[idx]);
+			ret = evergreen_vm_packet3_check(rdev, ib->ptr, &pkt);
+			break;
+		default:
+			dev_err(rdev->dev, "Unknown packet type %d !\n", pkt.type);
+			ret = -EINVAL;
+			break;
+		}
+		if (ret)
+			break;
+		idx += pkt.count + 2;
+	} while (idx < ib->length_dw);
+
+	return ret;
+}
diff --git a/drivers/gpu/drm/radeon/evergreend.h b/drivers/gpu/drm/radeon/evergreend.h
index e00039e..b502216 100644
--- a/drivers/gpu/drm/radeon/evergreend.h
+++ b/drivers/gpu/drm/radeon/evergreend.h
@@ -242,6 +242,7 @@
 #define	PA_CL_ENHANCE					0x8A14
 #define		CLIP_VTX_REORDER_ENA				(1 << 0)
 #define		NUM_CLIP_SEQ(x)					((x) << 1)
+#define	PA_SC_ENHANCE					0x8BF0
 #define PA_SC_AA_CONFIG					0x28C04
 #define         MSAA_NUM_SAMPLES_SHIFT                  0
 #define         MSAA_NUM_SAMPLES_MASK                   0x3
@@ -319,6 +320,8 @@
 #define	SQ_GPR_RESOURCE_MGMT_3				0x8C0C
 #define		NUM_HS_GPRS(x)					((x) << 0)
 #define		NUM_LS_GPRS(x)					((x) << 16)
+#define	SQ_GLOBAL_GPR_RESOURCE_MGMT_1			0x8C10
+#define	SQ_GLOBAL_GPR_RESOURCE_MGMT_2			0x8C14
 #define	SQ_THREAD_RESOURCE_MGMT				0x8C18
 #define		NUM_PS_THREADS(x)				((x) << 0)
 #define		NUM_VS_THREADS(x)				((x) << 8)
@@ -337,6 +340,10 @@
 #define		NUM_HS_STACK_ENTRIES(x)				((x) << 0)
 #define		NUM_LS_STACK_ENTRIES(x)				((x) << 16)
 #define	SQ_DYN_GPR_CNTL_PS_FLUSH_REQ    		0x8D8C
+#define	SQ_DYN_GPR_SIMD_LOCK_EN    			0x8D94
+#define	SQ_STATIC_THREAD_MGMT_1    			0x8E20
+#define	SQ_STATIC_THREAD_MGMT_2    			0x8E24
+#define	SQ_STATIC_THREAD_MGMT_3    			0x8E28
 #define	SQ_LDS_RESOURCE_MGMT    			0x8E2C
 
 #define	SQ_MS_FIFO_SIZES				0x8CF0
@@ -691,6 +698,7 @@
 #define	PACKET3_DRAW_INDEX_MULTI_ELEMENT		0x36
 #define	PACKET3_MEM_SEMAPHORE				0x39
 #define	PACKET3_MPEG_INDEX				0x3A
+#define	PACKET3_COPY_DW					0x3B
 #define	PACKET3_WAIT_REG_MEM				0x3C
 #define	PACKET3_MEM_WRITE				0x3D
 #define	PACKET3_INDIRECT_BUFFER				0x32
@@ -768,6 +776,8 @@
 #define			SQ_TEX_VTX_VALID_TEXTURE			0x2
 #define			SQ_TEX_VTX_VALID_BUFFER				0x3
 
+#define VGT_VTX_VECT_EJECT_REG				0x88b0
+
 #define SQ_CONST_MEM_BASE				0x8df8
 
 #define SQ_ESGS_RING_BASE				0x8c40
@@ -892,8 +902,27 @@
 #define PA_SC_SCREEN_SCISSOR_TL                         0x28030
 #define PA_SC_GENERIC_SCISSOR_TL                        0x28240
 #define PA_SC_WINDOW_SCISSOR_TL                         0x28204
-#define VGT_PRIMITIVE_TYPE                              0x8958
 
+#define VGT_PRIMITIVE_TYPE                              0x8958
+#define VGT_INDEX_TYPE                                  0x895C
+
+#define VGT_NUM_INDICES                                 0x8970
+
+#define VGT_COMPUTE_DIM_X                               0x8990
+#define VGT_COMPUTE_DIM_Y                               0x8994
+#define VGT_COMPUTE_DIM_Z                               0x8998
+#define VGT_COMPUTE_START_X                             0x899C
+#define VGT_COMPUTE_START_Y                             0x89A0
+#define VGT_COMPUTE_START_Z                             0x89A4
+#define VGT_COMPUTE_INDEX                               0x89A8
+#define VGT_COMPUTE_THREAD_GROUP_SIZE                   0x89AC
+#define VGT_HS_OFFCHIP_PARAM                            0x89B0
+
+#define DB_DEBUG					0x9830
+#define DB_DEBUG2					0x9834
+#define DB_DEBUG3					0x9838
+#define DB_DEBUG4					0x983C
+#define DB_WATERMARKS					0x9854
 #define DB_DEPTH_CONTROL				0x28800
 #define DB_DEPTH_VIEW					0x28008
 #define DB_HTILE_DATA_BASE				0x28014
@@ -1189,8 +1218,40 @@
 #define SQ_VTX_CONSTANT_WORD6_0                         0x30018
 #define SQ_VTX_CONSTANT_WORD7_0                         0x3001c
 
+#define TD_PS_BORDER_COLOR_INDEX                        0xA400
+#define TD_PS_BORDER_COLOR_RED                          0xA404
+#define TD_PS_BORDER_COLOR_GREEN                        0xA408
+#define TD_PS_BORDER_COLOR_BLUE                         0xA40C
+#define TD_PS_BORDER_COLOR_ALPHA                        0xA410
+#define TD_VS_BORDER_COLOR_INDEX                        0xA414
+#define TD_VS_BORDER_COLOR_RED                          0xA418
+#define TD_VS_BORDER_COLOR_GREEN                        0xA41C
+#define TD_VS_BORDER_COLOR_BLUE                         0xA420
+#define TD_VS_BORDER_COLOR_ALPHA                        0xA424
+#define TD_GS_BORDER_COLOR_INDEX                        0xA428
+#define TD_GS_BORDER_COLOR_RED                          0xA42C
+#define TD_GS_BORDER_COLOR_GREEN                        0xA430
+#define TD_GS_BORDER_COLOR_BLUE                         0xA434
+#define TD_GS_BORDER_COLOR_ALPHA                        0xA438
+#define TD_HS_BORDER_COLOR_INDEX                        0xA43C
+#define TD_HS_BORDER_COLOR_RED                          0xA440
+#define TD_HS_BORDER_COLOR_GREEN                        0xA444
+#define TD_HS_BORDER_COLOR_BLUE                         0xA448
+#define TD_HS_BORDER_COLOR_ALPHA                        0xA44C
+#define TD_LS_BORDER_COLOR_INDEX                        0xA450
+#define TD_LS_BORDER_COLOR_RED                          0xA454
+#define TD_LS_BORDER_COLOR_GREEN                        0xA458
+#define TD_LS_BORDER_COLOR_BLUE                         0xA45C
+#define TD_LS_BORDER_COLOR_ALPHA                        0xA460
+#define TD_CS_BORDER_COLOR_INDEX                        0xA464
+#define TD_CS_BORDER_COLOR_RED                          0xA468
+#define TD_CS_BORDER_COLOR_GREEN                        0xA46C
+#define TD_CS_BORDER_COLOR_BLUE                         0xA470
+#define TD_CS_BORDER_COLOR_ALPHA                        0xA474
+
 /* cayman 3D regs */
-#define CAYMAN_VGT_OFFCHIP_LDS_BASE			0x89B0
+#define CAYMAN_VGT_OFFCHIP_LDS_BASE			0x89B4
+#define CAYMAN_SQ_EX_ALLOC_TABLE_SLOTS			0x8E48
 #define CAYMAN_DB_EQAA					0x28804
 #define CAYMAN_DB_DEPTH_INFO				0x2803C
 #define CAYMAN_PA_SC_AA_CONFIG				0x28BE0
diff --git a/drivers/gpu/drm/radeon/ni.c b/drivers/gpu/drm/radeon/ni.c
index d89b2eb..3211372 100644
--- a/drivers/gpu/drm/radeon/ni.c
+++ b/drivers/gpu/drm/radeon/ni.c
@@ -934,7 +934,7 @@
 
 int cayman_pcie_gart_enable(struct radeon_device *rdev)
 {
-	int r;
+	int i, r;
 
 	if (rdev->gart.robj == NULL) {
 		dev_err(rdev->dev, "No VRAM object for PCIE GART.\n");
@@ -945,9 +945,12 @@
 		return r;
 	radeon_gart_restore(rdev);
 	/* Setup TLB control */
-	WREG32(MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB |
+	WREG32(MC_VM_MX_L1_TLB_CNTL,
+	       (0xA << 7) |
+	       ENABLE_L1_TLB |
 	       ENABLE_L1_FRAGMENT_PROCESSING |
 	       SYSTEM_ACCESS_MODE_NOT_IN_SYS |
+	       ENABLE_ADVANCED_DRIVER_MODEL |
 	       SYSTEM_APERTURE_UNMAPPED_ACCESS_PASS_THRU);
 	/* Setup L2 cache */
 	WREG32(VM_L2_CNTL, ENABLE_L2_CACHE |
@@ -967,9 +970,26 @@
 	WREG32(VM_CONTEXT0_CNTL2, 0);
 	WREG32(VM_CONTEXT0_CNTL, ENABLE_CONTEXT | PAGE_TABLE_DEPTH(0) |
 				RANGE_PROTECTION_FAULT_ENABLE_DEFAULT);
-	/* disable context1-7 */
+
+	WREG32(0x15D4, 0);
+	WREG32(0x15D8, 0);
+	WREG32(0x15DC, 0);
+
+	/* empty context1-7 */
+	for (i = 1; i < 8; i++) {
+		WREG32(VM_CONTEXT0_PAGE_TABLE_START_ADDR + (i << 2), 0);
+		WREG32(VM_CONTEXT0_PAGE_TABLE_END_ADDR + (i << 2), 0);
+		WREG32(VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (i << 2),
+			rdev->gart.table_addr >> 12);
+	}
+
+	/* enable context1-7 */
+	WREG32(VM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR,
+	       (u32)(rdev->dummy_page.addr >> 12));
 	WREG32(VM_CONTEXT1_CNTL2, 0);
 	WREG32(VM_CONTEXT1_CNTL, 0);
+	WREG32(VM_CONTEXT1_CNTL, ENABLE_CONTEXT | PAGE_TABLE_DEPTH(0) |
+				RANGE_PROTECTION_FAULT_ENABLE_DEFAULT);
 
 	cayman_pcie_gart_tlb_flush(rdev);
 	DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",
@@ -1024,7 +1044,10 @@
 	struct radeon_ring *ring = &rdev->ring[fence->ring];
 	u64 addr = rdev->fence_drv[fence->ring].gpu_addr;
 
-	/* flush read cache over gart */
+	/* flush read cache over gart for this vmid */
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1));
+	radeon_ring_write(ring, (CP_COHER_CNTL2 - PACKET3_SET_CONFIG_REG_START) >> 2);
+	radeon_ring_write(ring, 0);
 	radeon_ring_write(ring, PACKET3(PACKET3_SURFACE_SYNC, 3));
 	radeon_ring_write(ring, PACKET3_TC_ACTION_ENA | PACKET3_SH_ACTION_ENA);
 	radeon_ring_write(ring, 0xFFFFFFFF);
@@ -1039,6 +1062,33 @@
 	radeon_ring_write(ring, 0);
 }
 
+void cayman_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib)
+{
+	struct radeon_ring *ring = &rdev->ring[ib->fence->ring];
+
+	/* set to DX10/11 mode */
+	radeon_ring_write(ring, PACKET3(PACKET3_MODE_CONTROL, 0));
+	radeon_ring_write(ring, 1);
+	radeon_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
+	radeon_ring_write(ring,
+#ifdef __BIG_ENDIAN
+			  (2 << 0) |
+#endif
+			  (ib->gpu_addr & 0xFFFFFFFC));
+	radeon_ring_write(ring, upper_32_bits(ib->gpu_addr) & 0xFF);
+	radeon_ring_write(ring, ib->length_dw | (ib->vm_id << 24));
+
+	/* flush read cache over gart for this vmid */
+	radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1));
+	radeon_ring_write(ring, (CP_COHER_CNTL2 - PACKET3_SET_CONFIG_REG_START) >> 2);
+	radeon_ring_write(ring, ib->vm_id);
+	radeon_ring_write(ring, PACKET3(PACKET3_SURFACE_SYNC, 3));
+	radeon_ring_write(ring, PACKET3_TC_ACTION_ENA | PACKET3_SH_ACTION_ENA);
+	radeon_ring_write(ring, 0xFFFFFFFF);
+	radeon_ring_write(ring, 0);
+	radeon_ring_write(ring, 10); /* poll interval */
+}
+
 static void cayman_cp_enable(struct radeon_device *rdev, bool enable)
 {
 	if (enable)
@@ -1324,6 +1374,15 @@
 		RREG32(GRBM_STATUS_SE1));
 	dev_info(rdev->dev, "  SRBM_STATUS=0x%08X\n",
 		RREG32(SRBM_STATUS));
+	dev_info(rdev->dev, "  VM_CONTEXT0_PROTECTION_FAULT_ADDR   0x%08X\n",
+		 RREG32(0x14F8));
+	dev_info(rdev->dev, "  VM_CONTEXT0_PROTECTION_FAULT_STATUS 0x%08X\n",
+		 RREG32(0x14D8));
+	dev_info(rdev->dev, "  VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x%08X\n",
+		 RREG32(0x14FC));
+	dev_info(rdev->dev, "  VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x%08X\n",
+		 RREG32(0x14DC));
+
 	evergreen_mc_stop(rdev, &save);
 	if (evergreen_mc_wait_for_idle(rdev)) {
 		dev_warn(rdev->dev, "Wait for MC idle timedout !\n");
@@ -1354,6 +1413,7 @@
 	(void)RREG32(GRBM_SOFT_RESET);
 	/* Wait a little for things to settle down */
 	udelay(50);
+
 	dev_info(rdev->dev, "  GRBM_STATUS=0x%08X\n",
 		RREG32(GRBM_STATUS));
 	dev_info(rdev->dev, "  GRBM_STATUS_SE0=0x%08X\n",
@@ -1464,6 +1524,10 @@
 		return r;
 	}
 
+	r = radeon_vm_manager_start(rdev);
+	if (r)
+		return r;
+
 	return 0;
 }
 
@@ -1491,6 +1555,7 @@
 {
 	/* FIXME: we should wait for ring to be empty */
 	radeon_ib_pool_suspend(rdev);
+	radeon_vm_manager_suspend(rdev);
 	r600_blit_suspend(rdev);
 	cayman_cp_enable(rdev, false);
 	rdev->ring[RADEON_RING_TYPE_GFX_INDEX].ready = false;
@@ -1577,6 +1642,10 @@
 		dev_err(rdev->dev, "IB initialization failed (%d).\n", r);
 		rdev->accel_working = false;
 	}
+	r = radeon_vm_manager_init(rdev);
+	if (r) {
+		dev_err(rdev->dev, "vm manager initialization failed (%d).\n", r);
+	}
 
 	r = cayman_startup(rdev);
 	if (r) {
@@ -1585,6 +1654,7 @@
 		r600_irq_fini(rdev);
 		radeon_wb_fini(rdev);
 		r100_ib_fini(rdev);
+		radeon_vm_manager_fini(rdev);
 		radeon_irq_kms_fini(rdev);
 		cayman_pcie_gart_fini(rdev);
 		rdev->accel_working = false;
@@ -1608,6 +1678,7 @@
 	cayman_cp_fini(rdev);
 	r600_irq_fini(rdev);
 	radeon_wb_fini(rdev);
+	radeon_vm_manager_fini(rdev);
 	r100_ib_fini(rdev);
 	radeon_irq_kms_fini(rdev);
 	cayman_pcie_gart_fini(rdev);
@@ -1621,3 +1692,84 @@
 	rdev->bios = NULL;
 }
 
+/*
+ * vm
+ */
+int cayman_vm_init(struct radeon_device *rdev)
+{
+	/* number of VMs */
+	rdev->vm_manager.nvm = 8;
+	/* base offset of vram pages */
+	rdev->vm_manager.vram_base_offset = 0;
+	return 0;
+}
+
+void cayman_vm_fini(struct radeon_device *rdev)
+{
+}
+
+int cayman_vm_bind(struct radeon_device *rdev, struct radeon_vm *vm, int id)
+{
+	WREG32(VM_CONTEXT0_PAGE_TABLE_START_ADDR + (id << 2), 0);
+	WREG32(VM_CONTEXT0_PAGE_TABLE_END_ADDR + (id << 2), vm->last_pfn);
+	WREG32(VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (id << 2), vm->pt_gpu_addr >> 12);
+	/* flush hdp cache */
+	WREG32(HDP_MEM_COHERENCY_FLUSH_CNTL, 0x1);
+	/* bits 0-7 are the VM contexts0-7 */
+	WREG32(VM_INVALIDATE_REQUEST, 1 << id);
+	return 0;
+}
+
+void cayman_vm_unbind(struct radeon_device *rdev, struct radeon_vm *vm)
+{
+	WREG32(VM_CONTEXT0_PAGE_TABLE_START_ADDR + (vm->id << 2), 0);
+	WREG32(VM_CONTEXT0_PAGE_TABLE_END_ADDR + (vm->id << 2), 0);
+	WREG32(VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (vm->id << 2), 0);
+	/* flush hdp cache */
+	WREG32(HDP_MEM_COHERENCY_FLUSH_CNTL, 0x1);
+	/* bits 0-7 are the VM contexts0-7 */
+	WREG32(VM_INVALIDATE_REQUEST, 1 << vm->id);
+}
+
+void cayman_vm_tlb_flush(struct radeon_device *rdev, struct radeon_vm *vm)
+{
+	if (vm->id == -1)
+		return;
+
+	/* flush hdp cache */
+	WREG32(HDP_MEM_COHERENCY_FLUSH_CNTL, 0x1);
+	/* bits 0-7 are the VM contexts0-7 */
+	WREG32(VM_INVALIDATE_REQUEST, 1 << vm->id);
+}
+
+#define R600_PTE_VALID     (1 << 0)
+#define R600_PTE_SYSTEM    (1 << 1)
+#define R600_PTE_SNOOPED   (1 << 2)
+#define R600_PTE_READABLE  (1 << 5)
+#define R600_PTE_WRITEABLE (1 << 6)
+
+uint32_t cayman_vm_page_flags(struct radeon_device *rdev,
+			      struct radeon_vm *vm,
+			      uint32_t flags)
+{
+	uint32_t r600_flags = 0;
+
+	r600_flags |= (flags & RADEON_VM_PAGE_VALID) ? R600_PTE_VALID : 0;
+	r600_flags |= (flags & RADEON_VM_PAGE_READABLE) ? R600_PTE_READABLE : 0;
+	r600_flags |= (flags & RADEON_VM_PAGE_WRITEABLE) ? R600_PTE_WRITEABLE : 0;
+	if (flags & RADEON_VM_PAGE_SYSTEM) {
+		r600_flags |= R600_PTE_SYSTEM;
+		r600_flags |= (flags & RADEON_VM_PAGE_SNOOPED) ? R600_PTE_SNOOPED : 0;
+	}
+	return r600_flags;
+}
+
+void cayman_vm_set_page(struct radeon_device *rdev, struct radeon_vm *vm,
+			unsigned pfn, uint64_t addr, uint32_t flags)
+{
+	void __iomem *ptr = (void *)vm->pt;
+
+	addr = addr & 0xFFFFFFFFFFFFF000ULL;
+	addr |= flags;
+	writeq(addr, ptr + (pfn * 8));
+}
diff --git a/drivers/gpu/drm/radeon/nid.h b/drivers/gpu/drm/radeon/nid.h
index 0d3f52c..f9df2a6 100644
--- a/drivers/gpu/drm/radeon/nid.h
+++ b/drivers/gpu/drm/radeon/nid.h
@@ -222,6 +222,7 @@
 #define	SCRATCH_UMSK					0x8540
 #define	SCRATCH_ADDR					0x8544
 #define	CP_SEM_WAIT_TIMER				0x85BC
+#define	CP_COHER_CNTL2					0x85E8
 #define CP_ME_CNTL					0x86D8
 #define		CP_ME_HALT					(1 << 28)
 #define		CP_PFP_HALT					(1 << 26)
@@ -458,6 +459,7 @@
 #define	PACKET3_DISPATCH_DIRECT				0x15
 #define	PACKET3_DISPATCH_INDIRECT			0x16
 #define	PACKET3_INDIRECT_BUFFER_END			0x17
+#define	PACKET3_MODE_CONTROL				0x18
 #define	PACKET3_SET_PREDICATION				0x20
 #define	PACKET3_REG_RMW					0x21
 #define	PACKET3_COND_EXEC				0x22
diff --git a/drivers/gpu/drm/radeon/r300.c b/drivers/gpu/drm/radeon/r300.c
index 8ad5c64..3fc0d29 100644
--- a/drivers/gpu/drm/radeon/r300.c
+++ b/drivers/gpu/drm/radeon/r300.c
@@ -704,7 +704,7 @@
 			return r;
 		}
 
-		if (p->keep_tiling_flags) {
+		if (p->cs_flags & RADEON_CS_KEEP_TILING_FLAGS) {
 			ib[idx] = (idx_value & 31) | /* keep the 1st 5 bits */
 				  ((idx_value & ~31) + (u32)reloc->lobj.gpu_offset);
 		} else {
@@ -768,7 +768,7 @@
 		/* RB3D_COLORPITCH1 */
 		/* RB3D_COLORPITCH2 */
 		/* RB3D_COLORPITCH3 */
-		if (!p->keep_tiling_flags) {
+		if (!(p->cs_flags & RADEON_CS_KEEP_TILING_FLAGS)) {
 			r = r100_cs_packet_next_reloc(p, &reloc);
 			if (r) {
 				DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
@@ -853,7 +853,7 @@
 		break;
 	case 0x4F24:
 		/* ZB_DEPTHPITCH */
-		if (!p->keep_tiling_flags) {
+		if (!(p->cs_flags & RADEON_CS_KEEP_TILING_FLAGS)) {
 			r = r100_cs_packet_next_reloc(p, &reloc);
 			if (r) {
 				DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
diff --git a/drivers/gpu/drm/radeon/r600_cs.c b/drivers/gpu/drm/radeon/r600_cs.c
index cb1acff..38ce5d04 100644
--- a/drivers/gpu/drm/radeon/r600_cs.c
+++ b/drivers/gpu/drm/radeon/r600_cs.c
@@ -941,7 +941,7 @@
 		track->db_depth_control = radeon_get_ib_value(p, idx);
 		break;
 	case R_028010_DB_DEPTH_INFO:
-		if (!p->keep_tiling_flags &&
+		if (!(p->cs_flags & RADEON_CS_KEEP_TILING_FLAGS) &&
 		    r600_cs_packet_next_is_pkt3_nop(p)) {
 			r = r600_cs_packet_next_reloc(p, &reloc);
 			if (r) {
@@ -993,7 +993,7 @@
 	case R_0280B4_CB_COLOR5_INFO:
 	case R_0280B8_CB_COLOR6_INFO:
 	case R_0280BC_CB_COLOR7_INFO:
-		if (!p->keep_tiling_flags &&
+		if (!(p->cs_flags & RADEON_CS_KEEP_TILING_FLAGS) &&
 		     r600_cs_packet_next_is_pkt3_nop(p)) {
 			r = r600_cs_packet_next_reloc(p, &reloc);
 			if (r) {
@@ -1293,7 +1293,7 @@
 	mip_offset <<= 8;
 
 	word0 = radeon_get_ib_value(p, idx + 0);
-	if (!p->keep_tiling_flags) {
+	if (!(p->cs_flags & RADEON_CS_KEEP_TILING_FLAGS)) {
 		if (tiling_flags & RADEON_TILING_MACRO)
 			word0 |= S_038000_TILE_MODE(V_038000_ARRAY_2D_TILED_THIN1);
 		else if (tiling_flags & RADEON_TILING_MICRO)
@@ -1625,7 +1625,7 @@
 					return -EINVAL;
 				}
 				base_offset = (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
-				if (!p->keep_tiling_flags) {
+				if (!(p->cs_flags & RADEON_CS_KEEP_TILING_FLAGS)) {
 					if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO)
 						ib[idx+1+(i*7)+0] |= S_038000_TILE_MODE(V_038000_ARRAY_2D_TILED_THIN1);
 					else if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO)
diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h
index 374f9a4..5e35423 100644
--- a/drivers/gpu/drm/radeon/radeon.h
+++ b/drivers/gpu/drm/radeon/radeon.h
@@ -118,6 +118,10 @@
 #define CAYMAN_RING_TYPE_CP1_INDEX 1
 #define CAYMAN_RING_TYPE_CP2_INDEX 2
 
+/* hardcode those limit for now */
+#define RADEON_VA_RESERVED_SIZE		(8 << 20)
+#define RADEON_IB_VM_MAX_SIZE		(64 << 10)
+
 /*
  * Errata workarounds.
  */
@@ -262,6 +266,21 @@
 	bool				initialized;
 };
 
+/* bo virtual address in a specific vm */
+struct radeon_bo_va {
+	/* bo list is protected by bo being reserved */
+	struct list_head		bo_list;
+	/* vm list is protected by vm mutex */
+	struct list_head		vm_list;
+	/* constant after initialization */
+	struct radeon_vm		*vm;
+	struct radeon_bo		*bo;
+	uint64_t			soffset;
+	uint64_t			eoffset;
+	uint32_t			flags;
+	bool				valid;
+};
+
 struct radeon_bo {
 	/* Protected by gem.mutex */
 	struct list_head		list;
@@ -275,6 +294,10 @@
 	u32				tiling_flags;
 	u32				pitch;
 	int				surface_reg;
+	/* list of all virtual address to which this bo
+	 * is associated to
+	 */
+	struct list_head		va;
 	/* Constant after initialization */
 	struct radeon_device		*rdev;
 	struct drm_gem_object		gem_base;
@@ -408,6 +431,7 @@
 #define RADEON_GPU_PAGE_SIZE 4096
 #define RADEON_GPU_PAGE_MASK (RADEON_GPU_PAGE_SIZE - 1)
 #define RADEON_GPU_PAGE_SHIFT 12
+#define RADEON_GPU_PAGE_ALIGN(a) (((a) + RADEON_GPU_PAGE_MASK) & ~RADEON_GPU_PAGE_MASK)
 
 struct radeon_gart {
 	dma_addr_t			table_addr;
@@ -565,6 +589,7 @@
 	uint64_t		gpu_addr;
 	uint32_t		*ptr;
 	struct radeon_fence	*fence;
+	unsigned		vm_id;
 };
 
 /*
@@ -602,6 +627,56 @@
 };
 
 /*
+ * VM
+ */
+struct radeon_vm {
+	struct list_head		list;
+	struct list_head		va;
+	int				id;
+	unsigned			last_pfn;
+	u64				pt_gpu_addr;
+	u64				*pt;
+	struct radeon_sa_bo		sa_bo;
+	struct mutex			mutex;
+	/* last fence for cs using this vm */
+	struct radeon_fence		*fence;
+};
+
+struct radeon_vm_funcs {
+	int (*init)(struct radeon_device *rdev);
+	void (*fini)(struct radeon_device *rdev);
+	/* cs mutex must be lock for schedule_ib */
+	int (*bind)(struct radeon_device *rdev, struct radeon_vm *vm, int id);
+	void (*unbind)(struct radeon_device *rdev, struct radeon_vm *vm);
+	void (*tlb_flush)(struct radeon_device *rdev, struct radeon_vm *vm);
+	uint32_t (*page_flags)(struct radeon_device *rdev,
+			       struct radeon_vm *vm,
+			       uint32_t flags);
+	void (*set_page)(struct radeon_device *rdev, struct radeon_vm *vm,
+			unsigned pfn, uint64_t addr, uint32_t flags);
+};
+
+struct radeon_vm_manager {
+	struct list_head		lru_vm;
+	uint32_t			use_bitmap;
+	struct radeon_sa_manager	sa_manager;
+	uint32_t			max_pfn;
+	/* fields constant after init */
+	const struct radeon_vm_funcs	*funcs;
+	/* number of VMIDs */
+	unsigned			nvm;
+	/* vram base address for page table entry  */
+	u64				vram_base_offset;
+};
+
+/*
+ * file private structure
+ */
+struct radeon_fpriv {
+	struct radeon_vm		vm;
+};
+
+/*
  * R6xx+ IH ring
  */
 struct r600_ih {
@@ -691,12 +766,12 @@
 struct radeon_cs_chunk {
 	uint32_t		chunk_id;
 	uint32_t		length_dw;
-	int kpage_idx[2];
-	uint32_t                *kpage[2];
+	int			kpage_idx[2];
+	uint32_t		*kpage[2];
 	uint32_t		*kdata;
-	void __user *user_ptr;
-	int last_copied_page;
-	int last_page_index;
+	void __user		*user_ptr;
+	int			last_copied_page;
+	int			last_page_index;
 };
 
 struct radeon_cs_parser {
@@ -717,11 +792,14 @@
 	/* indices of various chunks */
 	int			chunk_ib_idx;
 	int			chunk_relocs_idx;
+	int			chunk_flags_idx;
 	struct radeon_ib	*ib;
 	void			*track;
 	unsigned		family;
 	int			parser_error;
-	bool			keep_tiling_flags;
+	u32			cs_flags;
+	u32			ring;
+	s32			priority;
 };
 
 extern int radeon_cs_update_pages(struct radeon_cs_parser *p, int pg_idx);
@@ -1018,6 +1096,7 @@
 
 	struct {
 		void (*ib_execute)(struct radeon_device *rdev, struct radeon_ib *ib);
+		int (*ib_parse)(struct radeon_device *rdev, struct radeon_ib *ib);
 		void (*emit_fence)(struct radeon_device *rdev, struct radeon_fence *fence);
 		void (*emit_semaphore)(struct radeon_device *rdev, struct radeon_ring *cp,
 				       struct radeon_semaphore *semaphore, bool emit_wait);
@@ -1255,6 +1334,8 @@
 			  struct drm_file *filp);
 int radeon_gem_wait_idle_ioctl(struct drm_device *dev, void *data,
 			      struct drm_file *filp);
+int radeon_gem_va_ioctl(struct drm_device *dev, void *data,
+			  struct drm_file *filp);
 int radeon_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp);
 int radeon_gem_set_tiling_ioctl(struct drm_device *dev, void *data,
 				struct drm_file *filp);
@@ -1404,6 +1485,8 @@
 	/* debugfs */
 	struct radeon_debugfs	debugfs[RADEON_DEBUGFS_MAX_COMPONENTS];
 	unsigned 		debugfs_count;
+	/* virtual memory */
+	struct radeon_vm_manager	vm_manager;
 };
 
 int radeon_device_init(struct radeon_device *rdev,
@@ -1568,6 +1651,7 @@
 #define radeon_ring_start(rdev) (rdev)->asic->ring_start((rdev))
 #define radeon_ring_test(rdev, cp) (rdev)->asic->ring_test((rdev), (cp))
 #define radeon_ring_ib_execute(rdev, r, ib) (rdev)->asic->ring[(r)].ib_execute((rdev), (ib))
+#define radeon_ring_ib_parse(rdev, r, ib) (rdev)->asic->ring[(r)].ib_parse((rdev), (ib))
 #define radeon_irq_set(rdev) (rdev)->asic->irq_set((rdev))
 #define radeon_irq_process(rdev) (rdev)->asic->irq_process((rdev))
 #define radeon_get_vblank_counter(rdev, crtc) (rdev)->asic->get_vblank_counter((rdev), (crtc))
@@ -1627,6 +1711,33 @@
 extern void radeon_ttm_set_active_vram_size(struct radeon_device *rdev, u64 size);
 
 /*
+ * vm
+ */
+int radeon_vm_manager_init(struct radeon_device *rdev);
+void radeon_vm_manager_fini(struct radeon_device *rdev);
+int radeon_vm_manager_start(struct radeon_device *rdev);
+int radeon_vm_manager_suspend(struct radeon_device *rdev);
+int radeon_vm_init(struct radeon_device *rdev, struct radeon_vm *vm);
+void radeon_vm_fini(struct radeon_device *rdev, struct radeon_vm *vm);
+int radeon_vm_bind(struct radeon_device *rdev, struct radeon_vm *vm);
+void radeon_vm_unbind(struct radeon_device *rdev, struct radeon_vm *vm);
+int radeon_vm_bo_update_pte(struct radeon_device *rdev,
+			    struct radeon_vm *vm,
+			    struct radeon_bo *bo,
+			    struct ttm_mem_reg *mem);
+void radeon_vm_bo_invalidate(struct radeon_device *rdev,
+			     struct radeon_bo *bo);
+int radeon_vm_bo_add(struct radeon_device *rdev,
+		     struct radeon_vm *vm,
+		     struct radeon_bo *bo,
+		     uint64_t offset,
+		     uint32_t flags);
+int radeon_vm_bo_rmv(struct radeon_device *rdev,
+		     struct radeon_vm *vm,
+		     struct radeon_bo *bo);
+
+
+/*
  * R600 vram scratch functions
  */
 int r600_vram_scratch_init(struct radeon_device *rdev);
diff --git a/drivers/gpu/drm/radeon/radeon_asic.c b/drivers/gpu/drm/radeon/radeon_asic.c
index 8493d40..123a196 100644
--- a/drivers/gpu/drm/radeon/radeon_asic.c
+++ b/drivers/gpu/drm/radeon/radeon_asic.c
@@ -952,6 +952,16 @@
 	.post_page_flip = &evergreen_post_page_flip,
 };
 
+static const struct radeon_vm_funcs cayman_vm_funcs = {
+	.init = &cayman_vm_init,
+	.fini = &cayman_vm_fini,
+	.bind = &cayman_vm_bind,
+	.unbind = &cayman_vm_unbind,
+	.tlb_flush = &cayman_vm_tlb_flush,
+	.page_flags = &cayman_vm_page_flags,
+	.set_page = &cayman_vm_set_page,
+};
+
 static struct radeon_asic cayman_asic = {
 	.init = &cayman_init,
 	.fini = &cayman_fini,
@@ -965,17 +975,20 @@
 	.ring_test = &r600_ring_test,
 	.ring = {
 		[RADEON_RING_TYPE_GFX_INDEX] = {
-			.ib_execute = &evergreen_ring_ib_execute,
+			.ib_execute = &cayman_ring_ib_execute,
+			.ib_parse = &evergreen_ib_parse,
 			.emit_fence = &cayman_fence_ring_emit,
 			.emit_semaphore = &r600_semaphore_ring_emit,
 		},
 		[CAYMAN_RING_TYPE_CP1_INDEX] = {
-			.ib_execute = &r600_ring_ib_execute,
+			.ib_execute = &cayman_ring_ib_execute,
+			.ib_parse = &evergreen_ib_parse,
 			.emit_fence = &cayman_fence_ring_emit,
 			.emit_semaphore = &r600_semaphore_ring_emit,
 		},
 		[CAYMAN_RING_TYPE_CP2_INDEX] = {
-			.ib_execute = &r600_ring_ib_execute,
+			.ib_execute = &cayman_ring_ib_execute,
+			.ib_parse = &evergreen_ib_parse,
 			.emit_fence = &cayman_fence_ring_emit,
 			.emit_semaphore = &r600_semaphore_ring_emit,
 		}
@@ -1128,6 +1141,7 @@
 		rdev->asic = &cayman_asic;
 		/* set num crtcs */
 		rdev->num_crtc = 6;
+		rdev->vm_manager.funcs = &cayman_vm_funcs;
 		break;
 	default:
 		/* FIXME: not supported yet */
diff --git a/drivers/gpu/drm/radeon/radeon_asic.h b/drivers/gpu/drm/radeon/radeon_asic.h
index c002ed1c..6304aef 100644
--- a/drivers/gpu/drm/radeon/radeon_asic.h
+++ b/drivers/gpu/drm/radeon/radeon_asic.h
@@ -438,5 +438,17 @@
 int cayman_resume(struct radeon_device *rdev);
 bool cayman_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *cp);
 int cayman_asic_reset(struct radeon_device *rdev);
+void cayman_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib);
+int cayman_vm_init(struct radeon_device *rdev);
+void cayman_vm_fini(struct radeon_device *rdev);
+int cayman_vm_bind(struct radeon_device *rdev, struct radeon_vm *vm, int id);
+void cayman_vm_unbind(struct radeon_device *rdev, struct radeon_vm *vm);
+void cayman_vm_tlb_flush(struct radeon_device *rdev, struct radeon_vm *vm);
+uint32_t cayman_vm_page_flags(struct radeon_device *rdev,
+			      struct radeon_vm *vm,
+			      uint32_t flags);
+void cayman_vm_set_page(struct radeon_device *rdev, struct radeon_vm *vm,
+			unsigned pfn, uint64_t addr, uint32_t flags);
+int evergreen_ib_parse(struct radeon_device *rdev, struct radeon_ib *ib);
 
 #endif
diff --git a/drivers/gpu/drm/radeon/radeon_cs.c b/drivers/gpu/drm/radeon/radeon_cs.c
index 6559cc4..4d59540 100644
--- a/drivers/gpu/drm/radeon/radeon_cs.c
+++ b/drivers/gpu/drm/radeon/radeon_cs.c
@@ -90,11 +90,32 @@
 	return radeon_bo_list_validate(&p->validated);
 }
 
+static int radeon_cs_get_ring(struct radeon_cs_parser *p, u32 ring, s32 priority)
+{
+	p->priority = priority;
+
+	switch (ring) {
+	default:
+		DRM_ERROR("unknown ring id: %d\n", ring);
+		return -EINVAL;
+	case RADEON_CS_RING_GFX:
+		p->ring = RADEON_RING_TYPE_GFX_INDEX;
+		break;
+	case RADEON_CS_RING_COMPUTE:
+		/* for now */
+		p->ring = RADEON_RING_TYPE_GFX_INDEX;
+		break;
+	}
+	return 0;
+}
+
 int radeon_cs_parser_init(struct radeon_cs_parser *p, void *data)
 {
 	struct drm_radeon_cs *cs = data;
 	uint64_t *chunk_array_ptr;
-	unsigned size, i, flags = 0;
+	unsigned size, i;
+	u32 ring = RADEON_CS_RING_GFX;
+	s32 priority = 0;
 
 	if (!cs->num_chunks) {
 		return 0;
@@ -104,6 +125,7 @@
 	p->idx = 0;
 	p->chunk_ib_idx = -1;
 	p->chunk_relocs_idx = -1;
+	p->chunk_flags_idx = -1;
 	p->chunks_array = kcalloc(cs->num_chunks, sizeof(uint64_t), GFP_KERNEL);
 	if (p->chunks_array == NULL) {
 		return -ENOMEM;
@@ -113,6 +135,7 @@
 			       sizeof(uint64_t)*cs->num_chunks)) {
 		return -EFAULT;
 	}
+	p->cs_flags = 0;
 	p->nchunks = cs->num_chunks;
 	p->chunks = kcalloc(p->nchunks, sizeof(struct radeon_cs_chunk), GFP_KERNEL);
 	if (p->chunks == NULL) {
@@ -141,16 +164,19 @@
 			if (p->chunks[i].length_dw == 0)
 				return -EINVAL;
 		}
-		if (p->chunks[i].chunk_id == RADEON_CHUNK_ID_FLAGS &&
-		    !p->chunks[i].length_dw) {
-			return -EINVAL;
+		if (p->chunks[i].chunk_id == RADEON_CHUNK_ID_FLAGS) {
+			p->chunk_flags_idx = i;
+			/* zero length flags aren't useful */
+			if (p->chunks[i].length_dw == 0)
+				return -EINVAL;
 		}
 
 		p->chunks[i].length_dw = user_chunk.length_dw;
 		p->chunks[i].user_ptr = (void __user *)(unsigned long)user_chunk.chunk_data;
 
 		cdata = (uint32_t *)(unsigned long)user_chunk.chunk_data;
-		if (p->chunks[i].chunk_id != RADEON_CHUNK_ID_IB) {
+		if ((p->chunks[i].chunk_id == RADEON_CHUNK_ID_RELOCS) ||
+		    (p->chunks[i].chunk_id == RADEON_CHUNK_ID_FLAGS)) {
 			size = p->chunks[i].length_dw * sizeof(uint32_t);
 			p->chunks[i].kdata = kmalloc(size, GFP_KERNEL);
 			if (p->chunks[i].kdata == NULL) {
@@ -161,29 +187,58 @@
 				return -EFAULT;
 			}
 			if (p->chunks[i].chunk_id == RADEON_CHUNK_ID_FLAGS) {
-				flags = p->chunks[i].kdata[0];
+				p->cs_flags = p->chunks[i].kdata[0];
+				if (p->chunks[i].length_dw > 1)
+					ring = p->chunks[i].kdata[1];
+				if (p->chunks[i].length_dw > 2)
+					priority = (s32)p->chunks[i].kdata[2];
 			}
-		} else {
-			p->chunks[i].kpage[0] = kmalloc(PAGE_SIZE, GFP_KERNEL);
-			p->chunks[i].kpage[1] = kmalloc(PAGE_SIZE, GFP_KERNEL);
-			if (p->chunks[i].kpage[0] == NULL || p->chunks[i].kpage[1] == NULL) {
-				kfree(p->chunks[i].kpage[0]);
-				kfree(p->chunks[i].kpage[1]);
-				return -ENOMEM;
-			}
-			p->chunks[i].kpage_idx[0] = -1;
-			p->chunks[i].kpage_idx[1] = -1;
-			p->chunks[i].last_copied_page = -1;
-			p->chunks[i].last_page_index = ((p->chunks[i].length_dw * 4) - 1) / PAGE_SIZE;
 		}
 	}
-	if (p->chunks[p->chunk_ib_idx].length_dw > (16 * 1024)) {
-		DRM_ERROR("cs IB too big: %d\n",
-			  p->chunks[p->chunk_ib_idx].length_dw);
+
+	if ((p->cs_flags & RADEON_CS_USE_VM) &&
+	    (p->rdev->family < CHIP_CAYMAN)) {
+		DRM_ERROR("VM not supported on asic!\n");
+		if (p->chunk_relocs_idx != -1)
+			kfree(p->chunks[p->chunk_relocs_idx].kdata);
+		if (p->chunk_flags_idx != -1)
+			kfree(p->chunks[p->chunk_flags_idx].kdata);
 		return -EINVAL;
 	}
 
-	p->keep_tiling_flags = (flags & RADEON_CS_KEEP_TILING_FLAGS) != 0;
+	if (radeon_cs_get_ring(p, ring, priority)) {
+		if (p->chunk_relocs_idx != -1)
+			kfree(p->chunks[p->chunk_relocs_idx].kdata);
+		if (p->chunk_flags_idx != -1)
+			kfree(p->chunks[p->chunk_flags_idx].kdata);
+		return -EINVAL;
+	}
+
+
+	/* deal with non-vm */
+	if ((p->chunk_ib_idx != -1) &&
+	    ((p->cs_flags & RADEON_CS_USE_VM) == 0) &&
+	    (p->chunks[p->chunk_ib_idx].chunk_id == RADEON_CHUNK_ID_IB)) {
+		if (p->chunks[p->chunk_ib_idx].length_dw > (16 * 1024)) {
+			DRM_ERROR("cs IB too big: %d\n",
+				  p->chunks[p->chunk_ib_idx].length_dw);
+			return -EINVAL;
+		}
+		p->chunks[p->chunk_ib_idx].kpage[0] = kmalloc(PAGE_SIZE, GFP_KERNEL);
+		p->chunks[p->chunk_ib_idx].kpage[1] = kmalloc(PAGE_SIZE, GFP_KERNEL);
+		if (p->chunks[p->chunk_ib_idx].kpage[0] == NULL ||
+		    p->chunks[p->chunk_ib_idx].kpage[1] == NULL) {
+			kfree(p->chunks[p->chunk_ib_idx].kpage[0]);
+			kfree(p->chunks[p->chunk_ib_idx].kpage[1]);
+			return -ENOMEM;
+		}
+		p->chunks[p->chunk_ib_idx].kpage_idx[0] = -1;
+		p->chunks[p->chunk_ib_idx].kpage_idx[1] = -1;
+		p->chunks[p->chunk_ib_idx].last_copied_page = -1;
+		p->chunks[p->chunk_ib_idx].last_page_index =
+			((p->chunks[p->chunk_ib_idx].length_dw * 4) - 1) / PAGE_SIZE;
+	}
+
 	return 0;
 }
 
@@ -225,11 +280,131 @@
 	radeon_ib_free(parser->rdev, &parser->ib);
 }
 
+static int radeon_cs_ib_chunk(struct radeon_device *rdev,
+			      struct radeon_cs_parser *parser)
+{
+	struct radeon_cs_chunk *ib_chunk;
+	int r;
+
+	if (parser->chunk_ib_idx == -1)
+		return 0;
+
+	if (parser->cs_flags & RADEON_CS_USE_VM)
+		return 0;
+
+	ib_chunk = &parser->chunks[parser->chunk_ib_idx];
+	/* Copy the packet into the IB, the parser will read from the
+	 * input memory (cached) and write to the IB (which can be
+	 * uncached).
+	 */
+	r =  radeon_ib_get(rdev, parser->ring, &parser->ib,
+			   ib_chunk->length_dw * 4);
+	if (r) {
+		DRM_ERROR("Failed to get ib !\n");
+		return r;
+	}
+	parser->ib->length_dw = ib_chunk->length_dw;
+	r = radeon_cs_parse(parser);
+	if (r || parser->parser_error) {
+		DRM_ERROR("Invalid command stream !\n");
+		return r;
+	}
+	r = radeon_cs_finish_pages(parser);
+	if (r) {
+		DRM_ERROR("Invalid command stream !\n");
+		return r;
+	}
+	parser->ib->vm_id = 0;
+	r = radeon_ib_schedule(rdev, parser->ib);
+	if (r) {
+		DRM_ERROR("Failed to schedule IB !\n");
+	}
+	return 0;
+}
+
+static int radeon_bo_vm_update_pte(struct radeon_cs_parser *parser,
+				   struct radeon_vm *vm)
+{
+	struct radeon_bo_list *lobj;
+	struct radeon_bo *bo;
+	int r;
+
+	list_for_each_entry(lobj, &parser->validated, tv.head) {
+		bo = lobj->bo;
+		r = radeon_vm_bo_update_pte(parser->rdev, vm, bo, &bo->tbo.mem);
+		if (r) {
+			return r;
+		}
+	}
+	return 0;
+}
+
+static int radeon_cs_ib_vm_chunk(struct radeon_device *rdev,
+				 struct radeon_cs_parser *parser)
+{
+	struct radeon_cs_chunk *ib_chunk;
+	struct radeon_fpriv *fpriv = parser->filp->driver_priv;
+	struct radeon_vm *vm = &fpriv->vm;
+	int r;
+
+	if (parser->chunk_ib_idx == -1)
+		return 0;
+
+	if ((parser->cs_flags & RADEON_CS_USE_VM) == 0)
+		return 0;
+
+	ib_chunk = &parser->chunks[parser->chunk_ib_idx];
+	if (ib_chunk->length_dw > RADEON_IB_VM_MAX_SIZE) {
+		DRM_ERROR("cs IB too big: %d\n", ib_chunk->length_dw);
+		return -EINVAL;
+	}
+	r =  radeon_ib_get(rdev, parser->ring, &parser->ib,
+			   ib_chunk->length_dw * 4);
+	if (r) {
+		DRM_ERROR("Failed to get ib !\n");
+		return r;
+	}
+	parser->ib->length_dw = ib_chunk->length_dw;
+	/* Copy the packet into the IB */
+	if (DRM_COPY_FROM_USER(parser->ib->ptr, ib_chunk->user_ptr,
+			       ib_chunk->length_dw * 4)) {
+		return -EFAULT;
+	}
+	r = radeon_ring_ib_parse(rdev, parser->ring, parser->ib);
+	if (r) {
+		return r;
+	}
+
+	mutex_lock(&vm->mutex);
+	r = radeon_vm_bind(rdev, vm);
+	if (r) {
+		goto out;
+	}
+	r = radeon_bo_vm_update_pte(parser, vm);
+	if (r) {
+		goto out;
+	}
+	parser->ib->vm_id = vm->id;
+	/* ib pool is bind at 0 in virtual address space to gpu_addr is the
+	 * offset inside the pool bo
+	 */
+	parser->ib->gpu_addr = parser->ib->sa_bo.offset;
+	r = radeon_ib_schedule(rdev, parser->ib);
+out:
+	if (!r) {
+		if (vm->fence) {
+			radeon_fence_unref(&vm->fence);
+		}
+		vm->fence = radeon_fence_ref(parser->ib->fence);
+	}
+	mutex_unlock(&fpriv->vm.mutex);
+	return r;
+}
+
 int radeon_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 {
 	struct radeon_device *rdev = dev->dev_private;
 	struct radeon_cs_parser parser;
-	struct radeon_cs_chunk *ib_chunk;
 	int r;
 
 	radeon_mutex_lock(&rdev->cs_mutex);
@@ -246,15 +421,6 @@
 		radeon_mutex_unlock(&rdev->cs_mutex);
 		return r;
 	}
-	ib_chunk = &parser.chunks[parser.chunk_ib_idx];
-	r =  radeon_ib_get(rdev, RADEON_RING_TYPE_GFX_INDEX, &parser.ib,
-			   ib_chunk->length_dw * 4);
-	if (r) {
-		DRM_ERROR("Failed to get ib !\n");
-		radeon_cs_parser_fini(&parser, r);
-		radeon_mutex_unlock(&rdev->cs_mutex);
-		return r;
-	}
 	r = radeon_cs_parser_relocs(&parser);
 	if (r) {
 		if (r != -ERESTARTSYS)
@@ -263,28 +429,15 @@
 		radeon_mutex_unlock(&rdev->cs_mutex);
 		return r;
 	}
-	/* Copy the packet into the IB, the parser will read from the
-	 * input memory (cached) and write to the IB (which can be
-	 * uncached). */
-	parser.ib->length_dw = ib_chunk->length_dw;
-	r = radeon_cs_parse(&parser);
-	if (r || parser.parser_error) {
-		DRM_ERROR("Invalid command stream !\n");
-		radeon_cs_parser_fini(&parser, r);
-		radeon_mutex_unlock(&rdev->cs_mutex);
-		return r;
-	}
-	r = radeon_cs_finish_pages(&parser);
+	r = radeon_cs_ib_chunk(rdev, &parser);
 	if (r) {
-		DRM_ERROR("Invalid command stream !\n");
-		radeon_cs_parser_fini(&parser, r);
-		radeon_mutex_unlock(&rdev->cs_mutex);
-		return r;
+		goto out;
 	}
-	r = radeon_ib_schedule(rdev, parser.ib);
+	r = radeon_cs_ib_vm_chunk(rdev, &parser);
 	if (r) {
-		DRM_ERROR("Failed to schedule IB !\n");
+		goto out;
 	}
+out:
 	radeon_cs_parser_fini(&parser, r);
 	radeon_mutex_unlock(&rdev->cs_mutex);
 	return r;
diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c
index 79b08b4..0afb13b 100644
--- a/drivers/gpu/drm/radeon/radeon_device.c
+++ b/drivers/gpu/drm/radeon/radeon_device.c
@@ -735,6 +735,10 @@
 	init_waitqueue_head(&rdev->irq.vblank_queue);
 	init_waitqueue_head(&rdev->irq.idle_queue);
 	INIT_LIST_HEAD(&rdev->semaphore_drv.bo);
+	/* initialize vm here */
+	rdev->vm_manager.use_bitmap = 1;
+	rdev->vm_manager.max_pfn = 1 << 20;
+	INIT_LIST_HEAD(&rdev->vm_manager.lru_vm);
 
 	/* Set asic functions */
 	r = radeon_asic_init(rdev);
diff --git a/drivers/gpu/drm/radeon/radeon_drv.c b/drivers/gpu/drm/radeon/radeon_drv.c
index c3ef1d2..31da622 100644
--- a/drivers/gpu/drm/radeon/radeon_drv.c
+++ b/drivers/gpu/drm/radeon/radeon_drv.c
@@ -54,9 +54,10 @@
  *   2.10.0 - fusion 2D tiling
  *   2.11.0 - backend map, initial compute support for the CS checker
  *   2.12.0 - RADEON_CS_KEEP_TILING_FLAGS
+ *   2.13.0 - virtual memory support
  */
 #define KMS_DRIVER_MAJOR	2
-#define KMS_DRIVER_MINOR	12
+#define KMS_DRIVER_MINOR	13
 #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);
@@ -84,6 +85,10 @@
 			 struct drm_file *file_priv);
 int radeon_gem_object_init(struct drm_gem_object *obj);
 void radeon_gem_object_free(struct drm_gem_object *obj);
+int radeon_gem_object_open(struct drm_gem_object *obj,
+				struct drm_file *file_priv);
+void radeon_gem_object_close(struct drm_gem_object *obj,
+				struct drm_file *file_priv);
 extern int radeon_get_crtc_scanoutpos(struct drm_device *dev, int crtc,
 				      int *vpos, int *hpos);
 extern struct drm_ioctl_desc radeon_ioctls_kms[];
@@ -350,6 +355,8 @@
 	.ioctls = radeon_ioctls_kms,
 	.gem_init_object = radeon_gem_object_init,
 	.gem_free_object = radeon_gem_object_free,
+	.gem_open_object = radeon_gem_object_open,
+	.gem_close_object = radeon_gem_object_close,
 	.dma_ioctl = radeon_dma_ioctl_kms,
 	.dumb_create = radeon_mode_dumb_create,
 	.dumb_map_offset = radeon_mode_dumb_mmap,
diff --git a/drivers/gpu/drm/radeon/radeon_gart.c b/drivers/gpu/drm/radeon/radeon_gart.c
index a4d9816..3ef58ca 100644
--- a/drivers/gpu/drm/radeon/radeon_gart.c
+++ b/drivers/gpu/drm/radeon/radeon_gart.c
@@ -276,3 +276,391 @@
 
 	radeon_dummy_page_fini(rdev);
 }
+
+/*
+ * vm helpers
+ *
+ * TODO bind a default page at vm initialization for default address
+ */
+int radeon_vm_manager_init(struct radeon_device *rdev)
+{
+	int r;
+
+	/* mark first vm as always in use, it's the system one */
+	r = radeon_sa_bo_manager_init(rdev, &rdev->vm_manager.sa_manager,
+				      rdev->vm_manager.max_pfn * 8,
+				      RADEON_GEM_DOMAIN_VRAM);
+	if (r) {
+		dev_err(rdev->dev, "failed to allocate vm bo (%dKB)\n",
+			(rdev->vm_manager.max_pfn * 8) >> 10);
+		return r;
+	}
+	return rdev->vm_manager.funcs->init(rdev);
+}
+
+/* cs mutex must be lock */
+static void radeon_vm_unbind_locked(struct radeon_device *rdev,
+				    struct radeon_vm *vm)
+{
+	struct radeon_bo_va *bo_va;
+
+	if (vm->id == -1) {
+		return;
+	}
+
+	/* wait for vm use to end */
+	if (vm->fence) {
+		radeon_fence_wait(vm->fence, false);
+		radeon_fence_unref(&vm->fence);
+	}
+
+	/* hw unbind */
+	rdev->vm_manager.funcs->unbind(rdev, vm);
+	rdev->vm_manager.use_bitmap &= ~(1 << vm->id);
+	list_del_init(&vm->list);
+	vm->id = -1;
+	radeon_sa_bo_free(rdev, &vm->sa_bo);
+	vm->pt = NULL;
+
+	list_for_each_entry(bo_va, &vm->va, vm_list) {
+		bo_va->valid = false;
+	}
+}
+
+void radeon_vm_manager_fini(struct radeon_device *rdev)
+{
+	if (rdev->vm_manager.sa_manager.bo == NULL)
+		return;
+	radeon_vm_manager_suspend(rdev);
+	rdev->vm_manager.funcs->fini(rdev);
+	radeon_sa_bo_manager_fini(rdev, &rdev->vm_manager.sa_manager);
+}
+
+int radeon_vm_manager_start(struct radeon_device *rdev)
+{
+	if (rdev->vm_manager.sa_manager.bo == NULL) {
+		return -EINVAL;
+	}
+	return radeon_sa_bo_manager_start(rdev, &rdev->vm_manager.sa_manager);
+}
+
+int radeon_vm_manager_suspend(struct radeon_device *rdev)
+{
+	struct radeon_vm *vm, *tmp;
+
+	radeon_mutex_lock(&rdev->cs_mutex);
+	/* unbind all active vm */
+	list_for_each_entry_safe(vm, tmp, &rdev->vm_manager.lru_vm, list) {
+		radeon_vm_unbind_locked(rdev, vm);
+	}
+	rdev->vm_manager.funcs->fini(rdev);
+	radeon_mutex_unlock(&rdev->cs_mutex);
+	return radeon_sa_bo_manager_suspend(rdev, &rdev->vm_manager.sa_manager);
+}
+
+/* cs mutex must be lock */
+void radeon_vm_unbind(struct radeon_device *rdev, struct radeon_vm *vm)
+{
+	mutex_lock(&vm->mutex);
+	radeon_vm_unbind_locked(rdev, vm);
+	mutex_unlock(&vm->mutex);
+}
+
+/* cs mutex must be lock & vm mutex must be lock */
+int radeon_vm_bind(struct radeon_device *rdev, struct radeon_vm *vm)
+{
+	struct radeon_vm *vm_evict;
+	unsigned i;
+	int id = -1, r;
+
+	if (vm == NULL) {
+		return -EINVAL;
+	}
+
+	if (vm->id != -1) {
+		/* update lru */
+		list_del_init(&vm->list);
+		list_add_tail(&vm->list, &rdev->vm_manager.lru_vm);
+		return 0;
+	}
+
+retry:
+	r = radeon_sa_bo_new(rdev, &rdev->vm_manager.sa_manager, &vm->sa_bo,
+			     RADEON_GPU_PAGE_ALIGN(vm->last_pfn * 8),
+			     RADEON_GPU_PAGE_SIZE);
+	if (r) {
+		if (list_empty(&rdev->vm_manager.lru_vm)) {
+			return r;
+		}
+		vm_evict = list_first_entry(&rdev->vm_manager.lru_vm, struct radeon_vm, list);
+		radeon_vm_unbind(rdev, vm_evict);
+		goto retry;
+	}
+	vm->pt = rdev->vm_manager.sa_manager.cpu_ptr;
+	vm->pt += (vm->sa_bo.offset >> 3);
+	vm->pt_gpu_addr = rdev->vm_manager.sa_manager.gpu_addr;
+	vm->pt_gpu_addr += vm->sa_bo.offset;
+	memset(vm->pt, 0, RADEON_GPU_PAGE_ALIGN(vm->last_pfn * 8));
+
+retry_id:
+	/* search for free vm */
+	for (i = 0; i < rdev->vm_manager.nvm; i++) {
+		if (!(rdev->vm_manager.use_bitmap & (1 << i))) {
+			id = i;
+			break;
+		}
+	}
+	/* evict vm if necessary */
+	if (id == -1) {
+		vm_evict = list_first_entry(&rdev->vm_manager.lru_vm, struct radeon_vm, list);
+		radeon_vm_unbind(rdev, vm_evict);
+		goto retry_id;
+	}
+
+	/* do hw bind */
+	r = rdev->vm_manager.funcs->bind(rdev, vm, id);
+	if (r) {
+		radeon_sa_bo_free(rdev, &vm->sa_bo);
+		return r;
+	}
+	rdev->vm_manager.use_bitmap |= 1 << id;
+	vm->id = id;
+	list_add_tail(&vm->list, &rdev->vm_manager.lru_vm);
+	return radeon_vm_bo_update_pte(rdev, vm, rdev->ib_pool.sa_manager.bo,
+				       &rdev->ib_pool.sa_manager.bo->tbo.mem);
+}
+
+/* object have to be reserved */
+int radeon_vm_bo_add(struct radeon_device *rdev,
+		     struct radeon_vm *vm,
+		     struct radeon_bo *bo,
+		     uint64_t offset,
+		     uint32_t flags)
+{
+	struct radeon_bo_va *bo_va, *tmp;
+	struct list_head *head;
+	uint64_t size = radeon_bo_size(bo), last_offset = 0;
+	unsigned last_pfn;
+
+	bo_va = kzalloc(sizeof(struct radeon_bo_va), GFP_KERNEL);
+	if (bo_va == NULL) {
+		return -ENOMEM;
+	}
+	bo_va->vm = vm;
+	bo_va->bo = bo;
+	bo_va->soffset = offset;
+	bo_va->eoffset = offset + size;
+	bo_va->flags = flags;
+	bo_va->valid = false;
+	INIT_LIST_HEAD(&bo_va->bo_list);
+	INIT_LIST_HEAD(&bo_va->vm_list);
+	/* make sure object fit at this offset */
+	if (bo_va->soffset >= bo_va->eoffset) {
+		kfree(bo_va);
+		return -EINVAL;
+	}
+
+	last_pfn = bo_va->eoffset / RADEON_GPU_PAGE_SIZE;
+	if (last_pfn > rdev->vm_manager.max_pfn) {
+		kfree(bo_va);
+		dev_err(rdev->dev, "va above limit (0x%08X > 0x%08X)\n",
+			last_pfn, rdev->vm_manager.max_pfn);
+		return -EINVAL;
+	}
+
+	mutex_lock(&vm->mutex);
+	if (last_pfn > vm->last_pfn) {
+		/* grow va space 32M by 32M */
+		unsigned align = ((32 << 20) >> 12) - 1;
+		radeon_mutex_lock(&rdev->cs_mutex);
+		radeon_vm_unbind_locked(rdev, vm);
+		radeon_mutex_unlock(&rdev->cs_mutex);
+		vm->last_pfn = (last_pfn + align) & ~align;
+	}
+	head = &vm->va;
+	last_offset = 0;
+	list_for_each_entry(tmp, &vm->va, vm_list) {
+		if (bo_va->soffset >= last_offset && bo_va->eoffset < tmp->soffset) {
+			/* bo can be added before this one */
+			break;
+		}
+		if (bo_va->soffset >= tmp->soffset && bo_va->soffset < tmp->eoffset) {
+			/* bo and tmp overlap, invalid offset */
+			kfree(bo_va);
+			dev_err(rdev->dev, "bo %p va 0x%08X conflict with (bo %p 0x%08X 0x%08X)\n",
+				bo, (unsigned)bo_va->soffset, tmp->bo,
+				(unsigned)tmp->soffset, (unsigned)tmp->eoffset);
+			mutex_unlock(&vm->mutex);
+			return -EINVAL;
+		}
+		last_offset = tmp->eoffset;
+		head = &tmp->vm_list;
+	}
+	list_add(&bo_va->vm_list, head);
+	list_add_tail(&bo_va->bo_list, &bo->va);
+	mutex_unlock(&vm->mutex);
+	return 0;
+}
+
+static u64 radeon_vm_get_addr(struct radeon_device *rdev,
+			      struct ttm_mem_reg *mem,
+			      unsigned pfn)
+{
+	u64 addr = 0;
+
+	switch (mem->mem_type) {
+	case TTM_PL_VRAM:
+		addr = (mem->start << PAGE_SHIFT);
+		addr += pfn * RADEON_GPU_PAGE_SIZE;
+		addr += rdev->vm_manager.vram_base_offset;
+		break;
+	case TTM_PL_TT:
+		/* offset inside page table */
+		addr = mem->start << PAGE_SHIFT;
+		addr += pfn * RADEON_GPU_PAGE_SIZE;
+		addr = addr >> PAGE_SHIFT;
+		/* page table offset */
+		addr = rdev->gart.pages_addr[addr];
+		/* in case cpu page size != gpu page size*/
+		addr += (pfn * RADEON_GPU_PAGE_SIZE) & (~PAGE_MASK);
+		break;
+	default:
+		break;
+	}
+	return addr;
+}
+
+/* object have to be reserved & cs mutex took & vm mutex took */
+int radeon_vm_bo_update_pte(struct radeon_device *rdev,
+			    struct radeon_vm *vm,
+			    struct radeon_bo *bo,
+			    struct ttm_mem_reg *mem)
+{
+	struct radeon_bo_va *bo_va;
+	unsigned ngpu_pages, i;
+	uint64_t addr = 0, pfn;
+	uint32_t flags;
+
+	/* nothing to do if vm isn't bound */
+	if (vm->id == -1)
+		return 0;;
+
+	bo_va = radeon_bo_va(bo, vm);
+	if (bo_va == NULL) {
+		dev_err(rdev->dev, "bo %p not in vm %p\n", bo, vm);
+		return -EINVAL;
+	}
+
+	if (bo_va->valid)
+		return 0;
+
+	ngpu_pages = radeon_bo_ngpu_pages(bo);
+	bo_va->flags &= ~RADEON_VM_PAGE_VALID;
+	bo_va->flags &= ~RADEON_VM_PAGE_SYSTEM;
+	if (mem) {
+		if (mem->mem_type != TTM_PL_SYSTEM) {
+			bo_va->flags |= RADEON_VM_PAGE_VALID;
+			bo_va->valid = true;
+		}
+		if (mem->mem_type == TTM_PL_TT) {
+			bo_va->flags |= RADEON_VM_PAGE_SYSTEM;
+		}
+	}
+	pfn = bo_va->soffset / RADEON_GPU_PAGE_SIZE;
+	flags = rdev->vm_manager.funcs->page_flags(rdev, bo_va->vm, bo_va->flags);
+	for (i = 0, addr = 0; i < ngpu_pages; i++) {
+		if (mem && bo_va->valid) {
+			addr = radeon_vm_get_addr(rdev, mem, i);
+		}
+		rdev->vm_manager.funcs->set_page(rdev, bo_va->vm, i + pfn, addr, flags);
+	}
+	rdev->vm_manager.funcs->tlb_flush(rdev, bo_va->vm);
+	return 0;
+}
+
+/* object have to be reserved */
+int radeon_vm_bo_rmv(struct radeon_device *rdev,
+		     struct radeon_vm *vm,
+		     struct radeon_bo *bo)
+{
+	struct radeon_bo_va *bo_va;
+
+	bo_va = radeon_bo_va(bo, vm);
+	if (bo_va == NULL)
+		return 0;
+
+	list_del(&bo_va->bo_list);
+	mutex_lock(&vm->mutex);
+	radeon_mutex_lock(&rdev->cs_mutex);
+	radeon_vm_bo_update_pte(rdev, vm, bo, NULL);
+	radeon_mutex_unlock(&rdev->cs_mutex);
+	list_del(&bo_va->vm_list);
+	mutex_lock(&vm->mutex);
+
+	kfree(bo_va);
+	return 0;
+}
+
+void radeon_vm_bo_invalidate(struct radeon_device *rdev,
+			     struct radeon_bo *bo)
+{
+	struct radeon_bo_va *bo_va;
+
+	BUG_ON(!atomic_read(&bo->tbo.reserved));
+	list_for_each_entry(bo_va, &bo->va, bo_list) {
+		bo_va->valid = false;
+	}
+}
+
+int radeon_vm_init(struct radeon_device *rdev, struct radeon_vm *vm)
+{
+	int r;
+
+	vm->id = -1;
+	vm->fence = NULL;
+	mutex_init(&vm->mutex);
+	INIT_LIST_HEAD(&vm->list);
+	INIT_LIST_HEAD(&vm->va);
+	vm->last_pfn = 0;
+	/* map the ib pool buffer at 0 in virtual address space, set
+	 * read only
+	 */
+	r = radeon_vm_bo_add(rdev, vm, rdev->ib_pool.sa_manager.bo, 0,
+			     RADEON_VM_PAGE_READABLE | RADEON_VM_PAGE_SNOOPED);
+	return r;
+}
+
+void radeon_vm_fini(struct radeon_device *rdev, struct radeon_vm *vm)
+{
+	struct radeon_bo_va *bo_va, *tmp;
+	int r;
+
+	mutex_lock(&vm->mutex);
+
+	radeon_mutex_lock(&rdev->cs_mutex);
+	radeon_vm_unbind_locked(rdev, vm);
+	radeon_mutex_unlock(&rdev->cs_mutex);
+
+	/* remove all bo */
+	r = radeon_bo_reserve(rdev->ib_pool.sa_manager.bo, false);
+	if (!r) {
+		bo_va = radeon_bo_va(rdev->ib_pool.sa_manager.bo, vm);
+		list_del_init(&bo_va->bo_list);
+		list_del_init(&bo_va->vm_list);
+		radeon_bo_unreserve(rdev->ib_pool.sa_manager.bo);
+		kfree(bo_va);
+	}
+	if (!list_empty(&vm->va)) {
+		dev_err(rdev->dev, "still active bo inside vm\n");
+	}
+	list_for_each_entry_safe(bo_va, tmp, &vm->va, vm_list) {
+		list_del_init(&bo_va->vm_list);
+		r = radeon_bo_reserve(bo_va->bo, false);
+		if (!r) {
+			list_del_init(&bo_va->bo_list);
+			radeon_bo_unreserve(bo_va->bo);
+			kfree(bo_va);
+		}
+	}
+	mutex_unlock(&vm->mutex);
+}
diff --git a/drivers/gpu/drm/radeon/radeon_gem.c b/drivers/gpu/drm/radeon/radeon_gem.c
index ae32197..003eeec 100644
--- a/drivers/gpu/drm/radeon/radeon_gem.c
+++ b/drivers/gpu/drm/radeon/radeon_gem.c
@@ -142,6 +142,44 @@
 	radeon_bo_force_delete(rdev);
 }
 
+/*
+ * Call from drm_gem_handle_create which appear in both new and open ioctl
+ * case.
+ */
+int radeon_gem_object_open(struct drm_gem_object *obj, struct drm_file *file_priv)
+{
+	return 0;
+}
+
+void radeon_gem_object_close(struct drm_gem_object *obj,
+			     struct drm_file *file_priv)
+{
+	struct radeon_bo *rbo = gem_to_radeon_bo(obj);
+	struct radeon_device *rdev = rbo->rdev;
+	struct radeon_fpriv *fpriv = file_priv->driver_priv;
+	struct radeon_vm *vm = &fpriv->vm;
+	struct radeon_bo_va *bo_va, *tmp;
+
+	if (rdev->family < CHIP_CAYMAN) {
+		return;
+	}
+
+	if (radeon_bo_reserve(rbo, false)) {
+		return;
+	}
+	list_for_each_entry_safe(bo_va, tmp, &rbo->va, bo_list) {
+		if (bo_va->vm == vm) {
+			/* remove from this vm address space */
+			mutex_lock(&vm->mutex);
+			list_del(&bo_va->vm_list);
+			mutex_unlock(&vm->mutex);
+			list_del(&bo_va->bo_list);
+			kfree(bo_va);
+		}
+	}
+	radeon_bo_unreserve(rbo);
+}
+
 
 /*
  * GEM ioctls.
@@ -354,6 +392,104 @@
 	return r;
 }
 
+int radeon_gem_va_ioctl(struct drm_device *dev, void *data,
+			  struct drm_file *filp)
+{
+	struct drm_radeon_gem_va *args = data;
+	struct drm_gem_object *gobj;
+	struct radeon_device *rdev = dev->dev_private;
+	struct radeon_fpriv *fpriv = filp->driver_priv;
+	struct radeon_bo *rbo;
+	struct radeon_bo_va *bo_va;
+	u32 invalid_flags;
+	int r = 0;
+
+	/* !! DONT REMOVE !!
+	 * We don't support vm_id yet, to be sure we don't have have broken
+	 * userspace, reject anyone trying to use non 0 value thus moving
+	 * forward we can use those fields without breaking existant userspace
+	 */
+	if (args->vm_id) {
+		args->operation = RADEON_VA_RESULT_ERROR;
+		return -EINVAL;
+	}
+
+	if (args->offset < RADEON_VA_RESERVED_SIZE) {
+		dev_err(&dev->pdev->dev,
+			"offset 0x%lX is in reserved area 0x%X\n",
+			(unsigned long)args->offset,
+			RADEON_VA_RESERVED_SIZE);
+		args->operation = RADEON_VA_RESULT_ERROR;
+		return -EINVAL;
+	}
+
+	/* don't remove, we need to enforce userspace to set the snooped flag
+	 * otherwise we will endup with broken userspace and we won't be able
+	 * to enable this feature without adding new interface
+	 */
+	invalid_flags = RADEON_VM_PAGE_VALID | RADEON_VM_PAGE_SYSTEM;
+	if ((args->flags & invalid_flags)) {
+		dev_err(&dev->pdev->dev, "invalid flags 0x%08X vs 0x%08X\n",
+			args->flags, invalid_flags);
+		args->operation = RADEON_VA_RESULT_ERROR;
+		return -EINVAL;
+	}
+	if (!(args->flags & RADEON_VM_PAGE_SNOOPED)) {
+		dev_err(&dev->pdev->dev, "only supported snooped mapping for now\n");
+		args->operation = RADEON_VA_RESULT_ERROR;
+		return -EINVAL;
+	}
+
+	switch (args->operation) {
+	case RADEON_VA_MAP:
+	case RADEON_VA_UNMAP:
+		break;
+	default:
+		dev_err(&dev->pdev->dev, "unsupported operation %d\n",
+			args->operation);
+		args->operation = RADEON_VA_RESULT_ERROR;
+		return -EINVAL;
+	}
+
+	gobj = drm_gem_object_lookup(dev, filp, args->handle);
+	if (gobj == NULL) {
+		args->operation = RADEON_VA_RESULT_ERROR;
+		return -ENOENT;
+	}
+	rbo = gem_to_radeon_bo(gobj);
+	r = radeon_bo_reserve(rbo, false);
+	if (r) {
+		args->operation = RADEON_VA_RESULT_ERROR;
+		drm_gem_object_unreference_unlocked(gobj);
+		return r;
+	}
+	switch (args->operation) {
+	case RADEON_VA_MAP:
+		bo_va = radeon_bo_va(rbo, &fpriv->vm);
+		if (bo_va) {
+			args->operation = RADEON_VA_RESULT_VA_EXIST;
+			args->offset = bo_va->soffset;
+			goto out;
+		}
+		r = radeon_vm_bo_add(rdev, &fpriv->vm, rbo,
+				     args->offset, args->flags);
+		break;
+	case RADEON_VA_UNMAP:
+		r = radeon_vm_bo_rmv(rdev, &fpriv->vm, rbo);
+		break;
+	default:
+		break;
+	}
+	args->operation = RADEON_VA_RESULT_OK;
+	if (r) {
+		args->operation = RADEON_VA_RESULT_ERROR;
+	}
+out:
+	radeon_bo_unreserve(rbo);
+	drm_gem_object_unreference_unlocked(gobj);
+	return r;
+}
+
 int radeon_mode_dumb_create(struct drm_file *file_priv,
 			    struct drm_device *dev,
 			    struct drm_mode_create_dumb *args)
diff --git a/drivers/gpu/drm/radeon/radeon_kms.c b/drivers/gpu/drm/radeon/radeon_kms.c
index be2c122..d335288 100644
--- a/drivers/gpu/drm/radeon/radeon_kms.c
+++ b/drivers/gpu/drm/radeon/radeon_kms.c
@@ -250,6 +250,18 @@
 			return -EINVAL;
 		}
 		break;
+	case RADEON_INFO_VA_START:
+		/* this is where we report if vm is supported or not */
+		if (rdev->family < CHIP_CAYMAN)
+			return -EINVAL;
+		value = RADEON_VA_RESERVED_SIZE;
+		break;
+	case RADEON_INFO_IB_VM_MAX_SIZE:
+		/* this is where we report if vm is supported or not */
+		if (rdev->family < CHIP_CAYMAN)
+			return -EINVAL;
+		value = RADEON_IB_VM_MAX_SIZE;
+		break;
 	default:
 		DRM_DEBUG_KMS("Invalid request %d\n", info->request);
 		return -EINVAL;
@@ -270,7 +282,6 @@
 	return 0;
 }
 
-
 void radeon_driver_lastclose_kms(struct drm_device *dev)
 {
 	vga_switcheroo_process_delayed_switch();
@@ -278,12 +289,45 @@
 
 int radeon_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv)
 {
+	struct radeon_device *rdev = dev->dev_private;
+
+	file_priv->driver_priv = NULL;
+
+	/* new gpu have virtual address space support */
+	if (rdev->family >= CHIP_CAYMAN) {
+		struct radeon_fpriv *fpriv;
+		int r;
+
+		fpriv = kzalloc(sizeof(*fpriv), GFP_KERNEL);
+		if (unlikely(!fpriv)) {
+			return -ENOMEM;
+		}
+
+		r = radeon_vm_init(rdev, &fpriv->vm);
+		if (r) {
+			radeon_vm_fini(rdev, &fpriv->vm);
+			kfree(fpriv);
+			return r;
+		}
+
+		file_priv->driver_priv = fpriv;
+	}
 	return 0;
 }
 
 void radeon_driver_postclose_kms(struct drm_device *dev,
 				 struct drm_file *file_priv)
 {
+	struct radeon_device *rdev = dev->dev_private;
+
+	/* new gpu have virtual address space support */
+	if (rdev->family >= CHIP_CAYMAN && file_priv->driver_priv) {
+		struct radeon_fpriv *fpriv = file_priv->driver_priv;
+
+		radeon_vm_fini(rdev, &fpriv->vm);
+		kfree(fpriv);
+		file_priv->driver_priv = NULL;
+	}
 }
 
 void radeon_driver_preclose_kms(struct drm_device *dev,
@@ -451,5 +495,6 @@
 	DRM_IOCTL_DEF_DRV(RADEON_GEM_SET_TILING, radeon_gem_set_tiling_ioctl, DRM_AUTH|DRM_UNLOCKED),
 	DRM_IOCTL_DEF_DRV(RADEON_GEM_GET_TILING, radeon_gem_get_tiling_ioctl, DRM_AUTH|DRM_UNLOCKED),
 	DRM_IOCTL_DEF_DRV(RADEON_GEM_BUSY, radeon_gem_busy_ioctl, DRM_AUTH|DRM_UNLOCKED),
+	DRM_IOCTL_DEF_DRV(RADEON_GEM_VA, radeon_gem_va_ioctl, DRM_AUTH|DRM_UNLOCKED),
 };
 int radeon_max_kms_ioctl = DRM_ARRAY_SIZE(radeon_ioctls_kms);
diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c
index 695b480..d45df17 100644
--- a/drivers/gpu/drm/radeon/radeon_object.c
+++ b/drivers/gpu/drm/radeon/radeon_object.c
@@ -46,6 +46,20 @@
  * function are calling it.
  */
 
+void radeon_bo_clear_va(struct radeon_bo *bo)
+{
+	struct radeon_bo_va *bo_va, *tmp;
+
+	list_for_each_entry_safe(bo_va, tmp, &bo->va, bo_list) {
+		/* remove from all vm address space */
+		mutex_lock(&bo_va->vm->mutex);
+		list_del(&bo_va->vm_list);
+		mutex_unlock(&bo_va->vm->mutex);
+		list_del(&bo_va->bo_list);
+		kfree(bo_va);
+	}
+}
+
 static void radeon_ttm_bo_destroy(struct ttm_buffer_object *tbo)
 {
 	struct radeon_bo *bo;
@@ -55,6 +69,7 @@
 	list_del_init(&bo->list);
 	mutex_unlock(&bo->rdev->gem.mutex);
 	radeon_bo_clear_surface_reg(bo);
+	radeon_bo_clear_va(bo);
 	drm_gem_object_release(&bo->gem_base);
 	kfree(bo);
 }
@@ -134,6 +149,7 @@
 	bo->gem_base.driver_private = NULL;
 	bo->surface_reg = -1;
 	INIT_LIST_HEAD(&bo->list);
+	INIT_LIST_HEAD(&bo->va);
 	radeon_ttm_placement_from_domain(bo, domain);
 	/* Kernel allocation are uninterruptible */
 	mutex_lock(&rdev->vram_mutex);
@@ -487,6 +503,7 @@
 		return;
 	rbo = container_of(bo, struct radeon_bo, tbo);
 	radeon_bo_check_tiling(rbo, 0, 1);
+	radeon_vm_bo_invalidate(rbo->rdev, rbo);
 }
 
 int radeon_bo_fault_reserve_notify(struct ttm_buffer_object *bo)
@@ -560,3 +577,16 @@
 	}
 	return 0;
 }
+
+/* object have to be reserved */
+struct radeon_bo_va *radeon_bo_va(struct radeon_bo *rbo, struct radeon_vm *vm)
+{
+	struct radeon_bo_va *bo_va;
+
+	list_for_each_entry(bo_va, &rbo->va, bo_list) {
+		if (bo_va->vm == vm) {
+			return bo_va;
+		}
+	}
+	return NULL;
+}
diff --git a/drivers/gpu/drm/radeon/radeon_object.h b/drivers/gpu/drm/radeon/radeon_object.h
index cc236fb..cde4303 100644
--- a/drivers/gpu/drm/radeon/radeon_object.h
+++ b/drivers/gpu/drm/radeon/radeon_object.h
@@ -83,6 +83,16 @@
 	return !!atomic_read(&bo->tbo.reserved);
 }
 
+static inline unsigned radeon_bo_ngpu_pages(struct radeon_bo *bo)
+{
+	return (bo->tbo.num_pages << PAGE_SHIFT) / RADEON_GPU_PAGE_SIZE;
+}
+
+static inline unsigned radeon_bo_gpu_page_alignment(struct radeon_bo *bo)
+{
+	return (bo->tbo.mem.page_alignment << PAGE_SHIFT) / RADEON_GPU_PAGE_SIZE;
+}
+
 /**
  * radeon_bo_mmap_offset - return mmap offset of bo
  * @bo:	radeon object for which we query the offset
@@ -128,6 +138,8 @@
 					struct ttm_mem_reg *mem);
 extern int radeon_bo_fault_reserve_notify(struct ttm_buffer_object *bo);
 extern int radeon_bo_get_surface_reg(struct radeon_bo *bo);
+extern struct radeon_bo_va *radeon_bo_va(struct radeon_bo *rbo,
+					 struct radeon_vm *vm);
 
 /*
  * sub allocation
diff --git a/drivers/gpu/drm/radeon/radeon_ring.c b/drivers/gpu/drm/radeon/radeon_ring.c
index 465fb34..e8bc709 100644
--- a/drivers/gpu/drm/radeon/radeon_ring.c
+++ b/drivers/gpu/drm/radeon/radeon_ring.c
@@ -132,6 +132,7 @@
 				(*ib)->gpu_addr = rdev->ib_pool.sa_manager.gpu_addr;
 				(*ib)->gpu_addr += (*ib)->sa_bo.offset;
 				(*ib)->fence = fence;
+				(*ib)->vm_id = 0;
 				/* ib are most likely to be allocated in a ring fashion
 				 * thus rdev->ib_pool.head_id should be the id of the
 				 * oldest ib