drivers/gpu/drm/i915/gvt/cmd_parser.c

   1 /*
   2  * Copyright(c) 2011-2016 Intel Corporation. All rights reserved.
   3  *
   4  * Permission is hereby granted, free of charge, to any person obtaining a
   5  * copy of this software and associated documentation files (the "Software"),
   6  * to deal in the Software without restriction, including without limitation
   7  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
   8  * and/or sell copies of the Software, and to permit persons to whom the
   9  * Software is furnished to do so, subject to the following conditions:
  10  *
  11  * The above copyright notice and this permission notice (including the next
  12  * paragraph) shall be included in all copies or substantial portions of the
  13  * Software.
  14  *
  15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  16  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  17  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  18  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  19  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  20  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  21  * SOFTWARE.
  22  *
  23  * Authors:
  24  *    Ke Yu
  25  *    Kevin Tian <[email protected]>
  26  *    Zhiyuan Lv <[email protected]>
  27  *
  28  * Contributors:
  29  *    Min He <[email protected]>
  30  *    Ping Gao <[email protected]>
  31  *    Tina Zhang <[email protected]>
  32  *    Yulei Zhang <[email protected]>
  33  *    Zhi Wang <[email protected]>
  34  *
  35  */
  36
  37 #include <linux/slab.h>
  38
  39 #include "i915_drv.h"
  40 #include "i915_reg.h"
  41 #include "gt/intel_engine_regs.h"
  42 #include "gt/intel_gpu_commands.h"
  43 #include "gt/intel_gt_regs.h"
  44 #include "gt/intel_lrc.h"
  45 #include "gt/intel_ring.h"
  46 #include "gt/intel_gt_requests.h"
  47 #include "gt/shmem_utils.h"
  48 #include "gvt.h"
  49 #include "i915_pvinfo.h"
  50 #include "trace.h"
  51
  52 #include "display/i9xx_plane_regs.h"
  53 #include "display/intel_sprite_regs.h"
  54 #include "gem/i915_gem_context.h"
  55 #include "gem/i915_gem_pm.h"
  56 #include "gt/intel_context.h"
  57
  58 #define INVALID_OP    (~0U)
  59
  60 #define OP_LEN_MI           9
  61 #define OP_LEN_2D           10
  62 #define OP_LEN_3D_MEDIA     16
  63 #define OP_LEN_MFX_VC       16
  64 #define OP_LEN_VEBOX        16
  65
  66 #define CMD_TYPE(cmd)   (((cmd) >> 29) & 7)
  67
  68 struct sub_op_bits {
  69         int hi;
  70         int low;
  71 };
  72 struct decode_info {
  73         const char *name;
  74         int op_len;
  75         int nr_sub_op;
  76         const struct sub_op_bits *sub_op;
  77 };
  78
  79 #define   MAX_CMD_BUDGET                        0x7fffffff
  80 #define   MI_WAIT_FOR_PLANE_C_FLIP_PENDING      (1<<15)
  81 #define   MI_WAIT_FOR_PLANE_B_FLIP_PENDING      (1<<9)
  82 #define   MI_WAIT_FOR_PLANE_A_FLIP_PENDING      (1<<1)
  83
  84 #define   MI_WAIT_FOR_SPRITE_C_FLIP_PENDING      (1<<20)
  85 #define   MI_WAIT_FOR_SPRITE_B_FLIP_PENDING      (1<<10)
  86 #define   MI_WAIT_FOR_SPRITE_A_FLIP_PENDING      (1<<2)
  87
  88 /* Render Command Map */
  89
  90 /* MI_* command Opcode (28:23) */
  91 #define OP_MI_NOOP                          0x0
  92 #define OP_MI_SET_PREDICATE                 0x1  /* HSW+ */
  93 #define OP_MI_USER_INTERRUPT                0x2
  94 #define OP_MI_WAIT_FOR_EVENT                0x3
  95 #define OP_MI_FLUSH                         0x4
  96 #define OP_MI_ARB_CHECK                     0x5
  97 #define OP_MI_RS_CONTROL                    0x6  /* HSW+ */
  98 #define OP_MI_REPORT_HEAD                   0x7
  99 #define OP_MI_ARB_ON_OFF                    0x8
 100 #define OP_MI_URB_ATOMIC_ALLOC              0x9  /* HSW+ */
 101 #define OP_MI_BATCH_BUFFER_END              0xA
 102 #define OP_MI_SUSPEND_FLUSH                 0xB
 103 #define OP_MI_PREDICATE                     0xC  /* IVB+ */
 104 #define OP_MI_TOPOLOGY_FILTER               0xD  /* IVB+ */
 105 #define OP_MI_SET_APPID                     0xE  /* IVB+ */
 106 #define OP_MI_RS_CONTEXT                    0xF  /* HSW+ */
 107 #define OP_MI_LOAD_SCAN_LINES_INCL          0x12 /* HSW+ */
 108 #define OP_MI_DISPLAY_FLIP                  0x14
 109 #define OP_MI_SEMAPHORE_MBOX                0x16
 110 #define OP_MI_SET_CONTEXT                   0x18
 111 #define OP_MI_MATH                          0x1A
 112 #define OP_MI_URB_CLEAR                     0x19
 113 #define OP_MI_SEMAPHORE_SIGNAL              0x1B  /* BDW+ */
 114 #define OP_MI_SEMAPHORE_WAIT                0x1C  /* BDW+ */
 115
 116 #define OP_MI_STORE_DATA_IMM                0x20
 117 #define OP_MI_STORE_DATA_INDEX              0x21
 118 #define OP_MI_LOAD_REGISTER_IMM             0x22
 119 #define OP_MI_UPDATE_GTT                    0x23
 120 #define OP_MI_STORE_REGISTER_MEM            0x24
 121 #define OP_MI_FLUSH_DW                      0x26
 122 #define OP_MI_CLFLUSH                       0x27
 123 #define OP_MI_REPORT_PERF_COUNT             0x28
 124 #define OP_MI_LOAD_REGISTER_MEM             0x29  /* HSW+ */
 125 #define OP_MI_LOAD_REGISTER_REG             0x2A  /* HSW+ */
 126 #define OP_MI_RS_STORE_DATA_IMM             0x2B  /* HSW+ */
 127 #define OP_MI_LOAD_URB_MEM                  0x2C  /* HSW+ */
 128 #define OP_MI_STORE_URM_MEM                 0x2D  /* HSW+ */
 129 #define OP_MI_2E                            0x2E  /* BDW+ */
 130 #define OP_MI_2F                            0x2F  /* BDW+ */
 131 #define OP_MI_BATCH_BUFFER_START            0x31
 132
 133 /* Bit definition for dword 0 */
 134 #define _CMDBIT_BB_START_IN_PPGTT       (1UL << 8)
 135
 136 #define OP_MI_CONDITIONAL_BATCH_BUFFER_END  0x36
 137
 138 #define BATCH_BUFFER_ADDR_MASK ((1UL << 32) - (1U << 2))
 139 #define BATCH_BUFFER_ADDR_HIGH_MASK ((1UL << 16) - (1U))
 140 #define BATCH_BUFFER_ADR_SPACE_BIT(x)   (((x) >> 8) & 1U)
 141 #define BATCH_BUFFER_2ND_LEVEL_BIT(x)   ((x) >> 22 & 1U)
 142
 143 /* 2D command: Opcode (28:22) */
 144 #define OP_2D(x)    ((2<<7) | x)
 145
 146 #define OP_XY_SETUP_BLT                             OP_2D(0x1)
 147 #define OP_XY_SETUP_CLIP_BLT                        OP_2D(0x3)
 148 #define OP_XY_SETUP_MONO_PATTERN_SL_BLT             OP_2D(0x11)
 149 #define OP_XY_PIXEL_BLT                             OP_2D(0x24)
 150 #define OP_XY_SCANLINES_BLT                         OP_2D(0x25)
 151 #define OP_XY_TEXT_BLT                              OP_2D(0x26)
 152 #define OP_XY_TEXT_IMMEDIATE_BLT                    OP_2D(0x31)
 153 #define OP_XY_COLOR_BLT                             OP_2D(0x50)
 154 #define OP_XY_PAT_BLT                               OP_2D(0x51)
 155 #define OP_XY_MONO_PAT_BLT                          OP_2D(0x52)
 156 #define OP_XY_SRC_COPY_BLT                          OP_2D(0x53)
 157 #define OP_XY_MONO_SRC_COPY_BLT                     OP_2D(0x54)
 158 #define OP_XY_FULL_BLT                              OP_2D(0x55)
 159 #define OP_XY_FULL_MONO_SRC_BLT                     OP_2D(0x56)
 160 #define OP_XY_FULL_MONO_PATTERN_BLT                 OP_2D(0x57)
 161 #define OP_XY_FULL_MONO_PATTERN_MONO_SRC_BLT        OP_2D(0x58)
 162 #define OP_XY_MONO_PAT_FIXED_BLT                    OP_2D(0x59)
 163 #define OP_XY_MONO_SRC_COPY_IMMEDIATE_BLT           OP_2D(0x71)
 164 #define OP_XY_PAT_BLT_IMMEDIATE                     OP_2D(0x72)
 165 #define OP_XY_SRC_COPY_CHROMA_BLT                   OP_2D(0x73)
 166 #define OP_XY_FULL_IMMEDIATE_PATTERN_BLT            OP_2D(0x74)
 167 #define OP_XY_FULL_MONO_SRC_IMMEDIATE_PATTERN_BLT   OP_2D(0x75)
 168 #define OP_XY_PAT_CHROMA_BLT                        OP_2D(0x76)
 169 #define OP_XY_PAT_CHROMA_BLT_IMMEDIATE              OP_2D(0x77)
 170
 171 /* 3D/Media Command: Pipeline Type(28:27) Opcode(26:24) Sub Opcode(23:16) */
 172 #define OP_3D_MEDIA(sub_type, opcode, sub_opcode) \
 173         ((3 << 13) | ((sub_type) << 11) | ((opcode) << 8) | (sub_opcode))
 174
 175 #define OP_STATE_PREFETCH                       OP_3D_MEDIA(0x0, 0x0, 0x03)
 176
 177 #define OP_STATE_BASE_ADDRESS                   OP_3D_MEDIA(0x0, 0x1, 0x01)
 178 #define OP_STATE_SIP                            OP_3D_MEDIA(0x0, 0x1, 0x02)
 179 #define OP_3D_MEDIA_0_1_4                       OP_3D_MEDIA(0x0, 0x1, 0x04)
 180 #define OP_SWTESS_BASE_ADDRESS                  OP_3D_MEDIA(0x0, 0x1, 0x03)
 181
 182 #define OP_3DSTATE_VF_STATISTICS_GM45           OP_3D_MEDIA(0x1, 0x0, 0x0B)
 183
 184 #define OP_PIPELINE_SELECT                      OP_3D_MEDIA(0x1, 0x1, 0x04)
 185
 186 #define OP_MEDIA_VFE_STATE                      OP_3D_MEDIA(0x2, 0x0, 0x0)
 187 #define OP_MEDIA_CURBE_LOAD                     OP_3D_MEDIA(0x2, 0x0, 0x1)
 188 #define OP_MEDIA_INTERFACE_DESCRIPTOR_LOAD      OP_3D_MEDIA(0x2, 0x0, 0x2)
 189 #define OP_MEDIA_GATEWAY_STATE                  OP_3D_MEDIA(0x2, 0x0, 0x3)
 190 #define OP_MEDIA_STATE_FLUSH                    OP_3D_MEDIA(0x2, 0x0, 0x4)
 191 #define OP_MEDIA_POOL_STATE                     OP_3D_MEDIA(0x2, 0x0, 0x5)
 192
 193 #define OP_MEDIA_OBJECT                         OP_3D_MEDIA(0x2, 0x1, 0x0)
 194 #define OP_MEDIA_OBJECT_PRT                     OP_3D_MEDIA(0x2, 0x1, 0x2)
 195 #define OP_MEDIA_OBJECT_WALKER                  OP_3D_MEDIA(0x2, 0x1, 0x3)
 196 #define OP_GPGPU_WALKER                         OP_3D_MEDIA(0x2, 0x1, 0x5)
 197
 198 #define OP_3DSTATE_CLEAR_PARAMS                 OP_3D_MEDIA(0x3, 0x0, 0x04) /* IVB+ */
 199 #define OP_3DSTATE_DEPTH_BUFFER                 OP_3D_MEDIA(0x3, 0x0, 0x05) /* IVB+ */
 200 #define OP_3DSTATE_STENCIL_BUFFER               OP_3D_MEDIA(0x3, 0x0, 0x06) /* IVB+ */
 201 #define OP_3DSTATE_HIER_DEPTH_BUFFER            OP_3D_MEDIA(0x3, 0x0, 0x07) /* IVB+ */
 202 #define OP_3DSTATE_VERTEX_BUFFERS               OP_3D_MEDIA(0x3, 0x0, 0x08)
 203 #define OP_3DSTATE_VERTEX_ELEMENTS              OP_3D_MEDIA(0x3, 0x0, 0x09)
 204 #define OP_3DSTATE_INDEX_BUFFER                 OP_3D_MEDIA(0x3, 0x0, 0x0A)
 205 #define OP_3DSTATE_VF_STATISTICS                OP_3D_MEDIA(0x3, 0x0, 0x0B)
 206 #define OP_3DSTATE_VF                           OP_3D_MEDIA(0x3, 0x0, 0x0C)  /* HSW+ */
 207 #define OP_3DSTATE_CC_STATE_POINTERS            OP_3D_MEDIA(0x3, 0x0, 0x0E)
 208 #define OP_3DSTATE_SCISSOR_STATE_POINTERS       OP_3D_MEDIA(0x3, 0x0, 0x0F)
 209 #define OP_3DSTATE_VS                           OP_3D_MEDIA(0x3, 0x0, 0x10)
 210 #define OP_3DSTATE_GS                           OP_3D_MEDIA(0x3, 0x0, 0x11)
 211 #define OP_3DSTATE_CLIP                         OP_3D_MEDIA(0x3, 0x0, 0x12)
 212 #define OP_3DSTATE_SF                           OP_3D_MEDIA(0x3, 0x0, 0x13)
 213 #define OP_3DSTATE_WM                           OP_3D_MEDIA(0x3, 0x0, 0x14)
 214 #define OP_3DSTATE_CONSTANT_VS                  OP_3D_MEDIA(0x3, 0x0, 0x15)
 215 #define OP_3DSTATE_CONSTANT_GS                  OP_3D_MEDIA(0x3, 0x0, 0x16)
 216 #define OP_3DSTATE_CONSTANT_PS                  OP_3D_MEDIA(0x3, 0x0, 0x17)
 217 #define OP_3DSTATE_SAMPLE_MASK                  OP_3D_MEDIA(0x3, 0x0, 0x18)
 218 #define OP_3DSTATE_CONSTANT_HS                  OP_3D_MEDIA(0x3, 0x0, 0x19) /* IVB+ */
 219 #define OP_3DSTATE_CONSTANT_DS                  OP_3D_MEDIA(0x3, 0x0, 0x1A) /* IVB+ */
 220 #define OP_3DSTATE_HS                           OP_3D_MEDIA(0x3, 0x0, 0x1B) /* IVB+ */
 221 #define OP_3DSTATE_TE                           OP_3D_MEDIA(0x3, 0x0, 0x1C) /* IVB+ */
 222 #define OP_3DSTATE_DS                           OP_3D_MEDIA(0x3, 0x0, 0x1D) /* IVB+ */
 223 #define OP_3DSTATE_STREAMOUT                    OP_3D_MEDIA(0x3, 0x0, 0x1E) /* IVB+ */
 224 #define OP_3DSTATE_SBE                          OP_3D_MEDIA(0x3, 0x0, 0x1F) /* IVB+ */
 225 #define OP_3DSTATE_PS                           OP_3D_MEDIA(0x3, 0x0, 0x20) /* IVB+ */
 226 #define OP_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP OP_3D_MEDIA(0x3, 0x0, 0x21) /* IVB+ */
 227 #define OP_3DSTATE_VIEWPORT_STATE_POINTERS_CC   OP_3D_MEDIA(0x3, 0x0, 0x23) /* IVB+ */
 228 #define OP_3DSTATE_BLEND_STATE_POINTERS         OP_3D_MEDIA(0x3, 0x0, 0x24) /* IVB+ */
 229 #define OP_3DSTATE_DEPTH_STENCIL_STATE_POINTERS OP_3D_MEDIA(0x3, 0x0, 0x25) /* IVB+ */
 230 #define OP_3DSTATE_BINDING_TABLE_POINTERS_VS    OP_3D_MEDIA(0x3, 0x0, 0x26) /* IVB+ */
 231 #define OP_3DSTATE_BINDING_TABLE_POINTERS_HS    OP_3D_MEDIA(0x3, 0x0, 0x27) /* IVB+ */
 232 #define OP_3DSTATE_BINDING_TABLE_POINTERS_DS    OP_3D_MEDIA(0x3, 0x0, 0x28) /* IVB+ */
 233 #define OP_3DSTATE_BINDING_TABLE_POINTERS_GS    OP_3D_MEDIA(0x3, 0x0, 0x29) /* IVB+ */
 234 #define OP_3DSTATE_BINDING_TABLE_POINTERS_PS    OP_3D_MEDIA(0x3, 0x0, 0x2A) /* IVB+ */
 235 #define OP_3DSTATE_SAMPLER_STATE_POINTERS_VS    OP_3D_MEDIA(0x3, 0x0, 0x2B) /* IVB+ */
 236 #define OP_3DSTATE_SAMPLER_STATE_POINTERS_HS    OP_3D_MEDIA(0x3, 0x0, 0x2C) /* IVB+ */
 237 #define OP_3DSTATE_SAMPLER_STATE_POINTERS_DS    OP_3D_MEDIA(0x3, 0x0, 0x2D) /* IVB+ */
 238 #define OP_3DSTATE_SAMPLER_STATE_POINTERS_GS    OP_3D_MEDIA(0x3, 0x0, 0x2E) /* IVB+ */
 239 #define OP_3DSTATE_SAMPLER_STATE_POINTERS_PS    OP_3D_MEDIA(0x3, 0x0, 0x2F) /* IVB+ */
 240 #define OP_3DSTATE_URB_VS                       OP_3D_MEDIA(0x3, 0x0, 0x30) /* IVB+ */
 241 #define OP_3DSTATE_URB_HS                       OP_3D_MEDIA(0x3, 0x0, 0x31) /* IVB+ */
 242 #define OP_3DSTATE_URB_DS                       OP_3D_MEDIA(0x3, 0x0, 0x32) /* IVB+ */
 243 #define OP_3DSTATE_URB_GS                       OP_3D_MEDIA(0x3, 0x0, 0x33) /* IVB+ */
 244 #define OP_3DSTATE_GATHER_CONSTANT_VS           OP_3D_MEDIA(0x3, 0x0, 0x34) /* HSW+ */
 245 #define OP_3DSTATE_GATHER_CONSTANT_GS           OP_3D_MEDIA(0x3, 0x0, 0x35) /* HSW+ */
 246 #define OP_3DSTATE_GATHER_CONSTANT_HS           OP_3D_MEDIA(0x3, 0x0, 0x36) /* HSW+ */
 247 #define OP_3DSTATE_GATHER_CONSTANT_DS           OP_3D_MEDIA(0x3, 0x0, 0x37) /* HSW+ */
 248 #define OP_3DSTATE_GATHER_CONSTANT_PS           OP_3D_MEDIA(0x3, 0x0, 0x38) /* HSW+ */
 249 #define OP_3DSTATE_DX9_CONSTANTF_VS             OP_3D_MEDIA(0x3, 0x0, 0x39) /* HSW+ */
 250 #define OP_3DSTATE_DX9_CONSTANTF_PS             OP_3D_MEDIA(0x3, 0x0, 0x3A) /* HSW+ */
 251 #define OP_3DSTATE_DX9_CONSTANTI_VS             OP_3D_MEDIA(0x3, 0x0, 0x3B) /* HSW+ */
 252 #define OP_3DSTATE_DX9_CONSTANTI_PS             OP_3D_MEDIA(0x3, 0x0, 0x3C) /* HSW+ */
 253 #define OP_3DSTATE_DX9_CONSTANTB_VS             OP_3D_MEDIA(0x3, 0x0, 0x3D) /* HSW+ */
 254 #define OP_3DSTATE_DX9_CONSTANTB_PS             OP_3D_MEDIA(0x3, 0x0, 0x3E) /* HSW+ */
 255 #define OP_3DSTATE_DX9_LOCAL_VALID_VS           OP_3D_MEDIA(0x3, 0x0, 0x3F) /* HSW+ */
 256 #define OP_3DSTATE_DX9_LOCAL_VALID_PS           OP_3D_MEDIA(0x3, 0x0, 0x40) /* HSW+ */
 257 #define OP_3DSTATE_DX9_GENERATE_ACTIVE_VS       OP_3D_MEDIA(0x3, 0x0, 0x41) /* HSW+ */
 258 #define OP_3DSTATE_DX9_GENERATE_ACTIVE_PS       OP_3D_MEDIA(0x3, 0x0, 0x42) /* HSW+ */
 259 #define OP_3DSTATE_BINDING_TABLE_EDIT_VS        OP_3D_MEDIA(0x3, 0x0, 0x43) /* HSW+ */
 260 #define OP_3DSTATE_BINDING_TABLE_EDIT_GS        OP_3D_MEDIA(0x3, 0x0, 0x44) /* HSW+ */
 261 #define OP_3DSTATE_BINDING_TABLE_EDIT_HS        OP_3D_MEDIA(0x3, 0x0, 0x45) /* HSW+ */
 262 #define OP_3DSTATE_BINDING_TABLE_EDIT_DS        OP_3D_MEDIA(0x3, 0x0, 0x46) /* HSW+ */
 263 #define OP_3DSTATE_BINDING_TABLE_EDIT_PS        OP_3D_MEDIA(0x3, 0x0, 0x47) /* HSW+ */
 264
 265 #define OP_3DSTATE_VF_INSTANCING                OP_3D_MEDIA(0x3, 0x0, 0x49) /* BDW+ */
 266 #define OP_3DSTATE_VF_SGVS                      OP_3D_MEDIA(0x3, 0x0, 0x4A) /* BDW+ */
 267 #define OP_3DSTATE_VF_TOPOLOGY                  OP_3D_MEDIA(0x3, 0x0, 0x4B) /* BDW+ */
 268 #define OP_3DSTATE_WM_CHROMAKEY                 OP_3D_MEDIA(0x3, 0x0, 0x4C) /* BDW+ */
 269 #define OP_3DSTATE_PS_BLEND                     OP_3D_MEDIA(0x3, 0x0, 0x4D) /* BDW+ */
 270 #define OP_3DSTATE_WM_DEPTH_STENCIL             OP_3D_MEDIA(0x3, 0x0, 0x4E) /* BDW+ */
 271 #define OP_3DSTATE_PS_EXTRA                     OP_3D_MEDIA(0x3, 0x0, 0x4F) /* BDW+ */
 272 #define OP_3DSTATE_RASTER                       OP_3D_MEDIA(0x3, 0x0, 0x50) /* BDW+ */
 273 #define OP_3DSTATE_SBE_SWIZ                     OP_3D_MEDIA(0x3, 0x0, 0x51) /* BDW+ */
 274 #define OP_3DSTATE_WM_HZ_OP                     OP_3D_MEDIA(0x3, 0x0, 0x52) /* BDW+ */
 275 #define OP_3DSTATE_COMPONENT_PACKING            OP_3D_MEDIA(0x3, 0x0, 0x55) /* SKL+ */
 276
 277 #define OP_3DSTATE_DRAWING_RECTANGLE            OP_3D_MEDIA(0x3, 0x1, 0x00)
 278 #define OP_3DSTATE_SAMPLER_PALETTE_LOAD0        OP_3D_MEDIA(0x3, 0x1, 0x02)
 279 #define OP_3DSTATE_CHROMA_KEY                   OP_3D_MEDIA(0x3, 0x1, 0x04)
 280 #define OP_SNB_3DSTATE_DEPTH_BUFFER             OP_3D_MEDIA(0x3, 0x1, 0x05)
 281 #define OP_3DSTATE_POLY_STIPPLE_OFFSET          OP_3D_MEDIA(0x3, 0x1, 0x06)
 282 #define OP_3DSTATE_POLY_STIPPLE_PATTERN         OP_3D_MEDIA(0x3, 0x1, 0x07)
 283 #define OP_3DSTATE_LINE_STIPPLE                 OP_3D_MEDIA(0x3, 0x1, 0x08)
 284 #define OP_3DSTATE_AA_LINE_PARAMS               OP_3D_MEDIA(0x3, 0x1, 0x0A)
 285 #define OP_3DSTATE_GS_SVB_INDEX                 OP_3D_MEDIA(0x3, 0x1, 0x0B)
 286 #define OP_3DSTATE_SAMPLER_PALETTE_LOAD1        OP_3D_MEDIA(0x3, 0x1, 0x0C)
 287 #define OP_3DSTATE_MULTISAMPLE_BDW              OP_3D_MEDIA(0x3, 0x0, 0x0D)
 288 #define OP_SNB_3DSTATE_STENCIL_BUFFER           OP_3D_MEDIA(0x3, 0x1, 0x0E)
 289 #define OP_SNB_3DSTATE_HIER_DEPTH_BUFFER        OP_3D_MEDIA(0x3, 0x1, 0x0F)
 290 #define OP_SNB_3DSTATE_CLEAR_PARAMS             OP_3D_MEDIA(0x3, 0x1, 0x10)
 291 #define OP_3DSTATE_MONOFILTER_SIZE              OP_3D_MEDIA(0x3, 0x1, 0x11)
 292 #define OP_3DSTATE_PUSH_CONSTANT_ALLOC_VS       OP_3D_MEDIA(0x3, 0x1, 0x12) /* IVB+ */
 293 #define OP_3DSTATE_PUSH_CONSTANT_ALLOC_HS       OP_3D_MEDIA(0x3, 0x1, 0x13) /* IVB+ */
 294 #define OP_3DSTATE_PUSH_CONSTANT_ALLOC_DS       OP_3D_MEDIA(0x3, 0x1, 0x14) /* IVB+ */
 295 #define OP_3DSTATE_PUSH_CONSTANT_ALLOC_GS       OP_3D_MEDIA(0x3, 0x1, 0x15) /* IVB+ */
 296 #define OP_3DSTATE_PUSH_CONSTANT_ALLOC_PS       OP_3D_MEDIA(0x3, 0x1, 0x16) /* IVB+ */
 297 #define OP_3DSTATE_SO_DECL_LIST                 OP_3D_MEDIA(0x3, 0x1, 0x17)
 298 #define OP_3DSTATE_SO_BUFFER                    OP_3D_MEDIA(0x3, 0x1, 0x18)
 299 #define OP_3DSTATE_BINDING_TABLE_POOL_ALLOC     OP_3D_MEDIA(0x3, 0x1, 0x19) /* HSW+ */
 300 #define OP_3DSTATE_GATHER_POOL_ALLOC            OP_3D_MEDIA(0x3, 0x1, 0x1A) /* HSW+ */
 301 #define OP_3DSTATE_DX9_CONSTANT_BUFFER_POOL_ALLOC OP_3D_MEDIA(0x3, 0x1, 0x1B) /* HSW+ */
 302 #define OP_3DSTATE_SAMPLE_PATTERN               OP_3D_MEDIA(0x3, 0x1, 0x1C)
 303 #define OP_PIPE_CONTROL                         OP_3D_MEDIA(0x3, 0x2, 0x00)
 304 #define OP_3DPRIMITIVE                          OP_3D_MEDIA(0x3, 0x3, 0x00)
 305
 306 /* VCCP Command Parser */
 307
 308 /*
 309  * Below MFX and VBE cmd definition is from vaapi intel driver project (BSD License)
 310  * git://anongit.freedesktop.org/vaapi/intel-driver
 311  * src/i965_defines.h
 312  *
 313  */
 314
 315 #define OP_MFX(pipeline, op, sub_opa, sub_opb)     \
 316         (3 << 13 | \
 317          (pipeline) << 11 | \
 318          (op) << 8 | \
 319          (sub_opa) << 5 | \
 320          (sub_opb))
 321
 322 #define OP_MFX_PIPE_MODE_SELECT                    OP_MFX(2, 0, 0, 0)  /* ALL */
 323 #define OP_MFX_SURFACE_STATE                       OP_MFX(2, 0, 0, 1)  /* ALL */
 324 #define OP_MFX_PIPE_BUF_ADDR_STATE                 OP_MFX(2, 0, 0, 2)  /* ALL */
 325 #define OP_MFX_IND_OBJ_BASE_ADDR_STATE             OP_MFX(2, 0, 0, 3)  /* ALL */
 326 #define OP_MFX_BSP_BUF_BASE_ADDR_STATE             OP_MFX(2, 0, 0, 4)  /* ALL */
 327 #define OP_2_0_0_5                                 OP_MFX(2, 0, 0, 5)  /* ALL */
 328 #define OP_MFX_STATE_POINTER                       OP_MFX(2, 0, 0, 6)  /* ALL */
 329 #define OP_MFX_QM_STATE                            OP_MFX(2, 0, 0, 7)  /* IVB+ */
 330 #define OP_MFX_FQM_STATE                           OP_MFX(2, 0, 0, 8)  /* IVB+ */
 331 #define OP_MFX_PAK_INSERT_OBJECT                   OP_MFX(2, 0, 2, 8)  /* IVB+ */
 332 #define OP_MFX_STITCH_OBJECT                       OP_MFX(2, 0, 2, 0xA)  /* IVB+ */
 333
 334 #define OP_MFD_IT_OBJECT                           OP_MFX(2, 0, 1, 9) /* ALL */
 335
 336 #define OP_MFX_WAIT                                OP_MFX(1, 0, 0, 0) /* IVB+ */
 337 #define OP_MFX_AVC_IMG_STATE                       OP_MFX(2, 1, 0, 0) /* ALL */
 338 #define OP_MFX_AVC_QM_STATE                        OP_MFX(2, 1, 0, 1) /* ALL */
 339 #define OP_MFX_AVC_DIRECTMODE_STATE                OP_MFX(2, 1, 0, 2) /* ALL */
 340 #define OP_MFX_AVC_SLICE_STATE                     OP_MFX(2, 1, 0, 3) /* ALL */
 341 #define OP_MFX_AVC_REF_IDX_STATE                   OP_MFX(2, 1, 0, 4) /* ALL */
 342 #define OP_MFX_AVC_WEIGHTOFFSET_STATE              OP_MFX(2, 1, 0, 5) /* ALL */
 343 #define OP_MFD_AVC_PICID_STATE                     OP_MFX(2, 1, 1, 5) /* HSW+ */
 344 #define OP_MFD_AVC_DPB_STATE                       OP_MFX(2, 1, 1, 6) /* IVB+ */
 345 #define OP_MFD_AVC_SLICEADDR                       OP_MFX(2, 1, 1, 7) /* IVB+ */
 346 #define OP_MFD_AVC_BSD_OBJECT                      OP_MFX(2, 1, 1, 8) /* ALL */
 347 #define OP_MFC_AVC_PAK_OBJECT                      OP_MFX(2, 1, 2, 9) /* ALL */
 348
 349 #define OP_MFX_VC1_PRED_PIPE_STATE                 OP_MFX(2, 2, 0, 1) /* ALL */
 350 #define OP_MFX_VC1_DIRECTMODE_STATE                OP_MFX(2, 2, 0, 2) /* ALL */
 351 #define OP_MFD_VC1_SHORT_PIC_STATE                 OP_MFX(2, 2, 1, 0) /* IVB+ */
 352 #define OP_MFD_VC1_LONG_PIC_STATE                  OP_MFX(2, 2, 1, 1) /* IVB+ */
 353 #define OP_MFD_VC1_BSD_OBJECT                      OP_MFX(2, 2, 1, 8) /* ALL */
 354
 355 #define OP_MFX_MPEG2_PIC_STATE                     OP_MFX(2, 3, 0, 0) /* ALL */
 356 #define OP_MFX_MPEG2_QM_STATE                      OP_MFX(2, 3, 0, 1) /* ALL */
 357 #define OP_MFD_MPEG2_BSD_OBJECT                    OP_MFX(2, 3, 1, 8) /* ALL */
 358 #define OP_MFC_MPEG2_SLICEGROUP_STATE              OP_MFX(2, 3, 2, 3) /* ALL */
 359 #define OP_MFC_MPEG2_PAK_OBJECT                    OP_MFX(2, 3, 2, 9) /* ALL */
 360
 361 #define OP_MFX_2_6_0_0                             OP_MFX(2, 6, 0, 0) /* IVB+ */
 362 #define OP_MFX_2_6_0_8                             OP_MFX(2, 6, 0, 8) /* IVB+ */
 363 #define OP_MFX_2_6_0_9                             OP_MFX(2, 6, 0, 9) /* IVB+ */
 364
 365 #define OP_MFX_JPEG_PIC_STATE                      OP_MFX(2, 7, 0, 0)
 366 #define OP_MFX_JPEG_HUFF_TABLE_STATE               OP_MFX(2, 7, 0, 2)
 367 #define OP_MFD_JPEG_BSD_OBJECT                     OP_MFX(2, 7, 1, 8)
 368
 369 #define OP_VEB(pipeline, op, sub_opa, sub_opb) \
 370         (3 << 13 | \
 371          (pipeline) << 11 | \
 372          (op) << 8 | \
 373          (sub_opa) << 5 | \
 374          (sub_opb))
 375
 376 #define OP_VEB_SURFACE_STATE                       OP_VEB(2, 4, 0, 0)
 377 #define OP_VEB_STATE                               OP_VEB(2, 4, 0, 2)
 378 #define OP_VEB_DNDI_IECP_STATE                     OP_VEB(2, 4, 0, 3)
 379
 380 struct parser_exec_state;
 381
 382 typedef int (*parser_cmd_handler)(struct parser_exec_state *s);
 383
 384 #define GVT_CMD_HASH_BITS   7
 385
 386 /* which DWords need address fix */
 387 #define ADDR_FIX_1(x1)                  (1 << (x1))
 388 #define ADDR_FIX_2(x1, x2)              (ADDR_FIX_1(x1) | ADDR_FIX_1(x2))
 389 #define ADDR_FIX_3(x1, x2, x3)          (ADDR_FIX_1(x1) | ADDR_FIX_2(x2, x3))
 390 #define ADDR_FIX_4(x1, x2, x3, x4)      (ADDR_FIX_1(x1) | ADDR_FIX_3(x2, x3, x4))
 391 #define ADDR_FIX_5(x1, x2, x3, x4, x5)  (ADDR_FIX_1(x1) | ADDR_FIX_4(x2, x3, x4, x5))
 392
 393 #define DWORD_FIELD(dword, end, start) \
 394         FIELD_GET(GENMASK(end, start), cmd_val(s, dword))
 395
 396 #define OP_LENGTH_BIAS 2
 397 #define CMD_LEN(value)  (value + OP_LENGTH_BIAS)
 398
 399 static int gvt_check_valid_cmd_length(int len, int valid_len)
 400 {
 401         if (valid_len != len) {
 402                 gvt_err("len is not valid:  len=%u  valid_len=%u\n",
 403                         len, valid_len);
 404                 return -EFAULT;
 405         }
 406         return 0;
 407 }
 408
 409 struct cmd_info {
 410         const char *name;
 411         u32 opcode;
 412
 413 #define F_LEN_MASK      3U
 414 #define F_LEN_CONST  1U
 415 #define F_LEN_VAR    0U
 416 /* value is const although LEN maybe variable */
 417 #define F_LEN_VAR_FIXED    (1<<1)
 418
 419 /*
 420  * command has its own ip advance logic
 421  * e.g. MI_BATCH_START, MI_BATCH_END
 422  */
 423 #define F_IP_ADVANCE_CUSTOM (1<<2)
 424         u32 flag;
 425
 426 #define R_RCS   BIT(RCS0)
 427 #define R_VCS1  BIT(VCS0)
 428 #define R_VCS2  BIT(VCS1)
 429 #define R_VCS   (R_VCS1 | R_VCS2)
 430 #define R_BCS   BIT(BCS0)
 431 #define R_VECS  BIT(VECS0)
 432 #define R_ALL (R_RCS | R_VCS | R_BCS | R_VECS)
 433         /* rings that support this cmd: BLT/RCS/VCS/VECS */
 434         intel_engine_mask_t rings;
 435
 436         /* devices that support this cmd: SNB/IVB/HSW/... */
 437         u16 devices;
 438
 439         /* which DWords are address that need fix up.
 440          * bit 0 means a 32-bit non address operand in command
 441          * bit 1 means address operand, which could be 32-bit
 442          * or 64-bit depending on different architectures.(
 443          * defined by "gmadr_bytes_in_cmd" in intel_gvt.
 444          * No matter the address length, each address only takes
 445          * one bit in the bitmap.
 446          */
 447         u16 addr_bitmap;
 448
 449         /* flag == F_LEN_CONST : command length
 450          * flag == F_LEN_VAR : length bias bits
 451          * Note: length is in DWord
 452          */
 453         u32 len;
 454
 455         parser_cmd_handler handler;
 456
 457         /* valid length in DWord */
 458         u32 valid_len;
 459 };
 460
 461 struct cmd_entry {
 462         struct hlist_node hlist;
 463         const struct cmd_info *info;
 464 };
 465
 466 enum {
 467         RING_BUFFER_INSTRUCTION,
 468         BATCH_BUFFER_INSTRUCTION,
 469         BATCH_BUFFER_2ND_LEVEL,
 470         RING_BUFFER_CTX,
 471 };
 472
 473 enum {
 474         GTT_BUFFER,
 475         PPGTT_BUFFER
 476 };
 477
 478 struct parser_exec_state {
 479         struct intel_vgpu *vgpu;
 480         const struct intel_engine_cs *engine;
 481
 482         int buf_type;
 483
 484         /* batch buffer address type */
 485         int buf_addr_type;
 486
 487         /* graphics memory address of ring buffer start */
 488         unsigned long ring_start;
 489         unsigned long ring_size;
 490         unsigned long ring_head;
 491         unsigned long ring_tail;
 492
 493         /* instruction graphics memory address */
 494         unsigned long ip_gma;
 495
 496         /* mapped va of the instr_gma */
 497         void *ip_va;
 498         void *rb_va;
 499
 500         void *ret_bb_va;
 501         /* next instruction when return from  batch buffer to ring buffer */
 502         unsigned long ret_ip_gma_ring;
 503
 504         /* next instruction when return from 2nd batch buffer to batch buffer */
 505         unsigned long ret_ip_gma_bb;
 506
 507         /* batch buffer address type (GTT or PPGTT)
 508          * used when ret from 2nd level batch buffer
 509          */
 510         int saved_buf_addr_type;
 511         bool is_ctx_wa;
 512         bool is_init_ctx;
 513
 514         const struct cmd_info *info;
 515
 516         struct intel_vgpu_workload *workload;
 517 };
 518
 519 #define gmadr_dw_number(s)      \
 520         (s->vgpu->gvt->device_info.gmadr_bytes_in_cmd >> 2)
 521
 522 static unsigned long bypass_scan_mask = 0;
 523
 524 /* ring ALL, type = 0 */
 525 static const struct sub_op_bits sub_op_mi[] = {
 526         {31, 29},
 527         {28, 23},
 528 };
 529
 530 static const struct decode_info decode_info_mi = {
 531         "MI",
 532         OP_LEN_MI,
 533         ARRAY_SIZE(sub_op_mi),
 534         sub_op_mi,
 535 };
 536
 537 /* ring RCS, command type 2 */
 538 static const struct sub_op_bits sub_op_2d[] = {
 539         {31, 29},
 540         {28, 22},
 541 };
 542
 543 static const struct decode_info decode_info_2d = {
 544         "2D",
 545         OP_LEN_2D,
 546         ARRAY_SIZE(sub_op_2d),
 547         sub_op_2d,
 548 };
 549
 550 /* ring RCS, command type 3 */
 551 static const struct sub_op_bits sub_op_3d_media[] = {
 552         {31, 29},
 553         {28, 27},
 554         {26, 24},
 555         {23, 16},
 556 };
 557
 558 static const struct decode_info decode_info_3d_media = {
 559         "3D_Media",
 560         OP_LEN_3D_MEDIA,
 561         ARRAY_SIZE(sub_op_3d_media),
 562         sub_op_3d_media,
 563 };
 564
 565 /* ring VCS, command type 3 */
 566 static const struct sub_op_bits sub_op_mfx_vc[] = {
 567         {31, 29},
 568         {28, 27},
 569         {26, 24},
 570         {23, 21},
 571         {20, 16},
 572 };
 573
 574 static const struct decode_info decode_info_mfx_vc = {
 575         "MFX_VC",
 576         OP_LEN_MFX_VC,
 577         ARRAY_SIZE(sub_op_mfx_vc),
 578         sub_op_mfx_vc,
 579 };
 580
 581 /* ring VECS, command type 3 */
 582 static const struct sub_op_bits sub_op_vebox[] = {
 583         {31, 29},
 584         {28, 27},
 585         {26, 24},
 586         {23, 21},
 587         {20, 16},
 588 };
 589
 590 static const struct decode_info decode_info_vebox = {
 591         "VEBOX",
 592         OP_LEN_VEBOX,
 593         ARRAY_SIZE(sub_op_vebox),
 594         sub_op_vebox,
 595 };
 596
 597 static const struct decode_info *ring_decode_info[I915_NUM_ENGINES][8] = {
 598         [RCS0] = {
 599                 &decode_info_mi,
 600                 NULL,
 601                 NULL,
 602                 &decode_info_3d_media,
 603                 NULL,
 604                 NULL,
 605                 NULL,
 606                 NULL,
 607         },
 608
 609         [VCS0] = {
 610                 &decode_info_mi,
 611                 NULL,
 612                 NULL,
 613                 &decode_info_mfx_vc,
 614                 NULL,
 615                 NULL,
 616                 NULL,
 617                 NULL,
 618         },
 619
 620         [BCS0] = {
 621                 &decode_info_mi,
 622                 NULL,
 623                 &decode_info_2d,
 624                 NULL,
 625                 NULL,
 626                 NULL,
 627                 NULL,
 628                 NULL,
 629         },
 630
 631         [VECS0] = {
 632                 &decode_info_mi,
 633                 NULL,
 634                 NULL,
 635                 &decode_info_vebox,
 636                 NULL,
 637                 NULL,
 638                 NULL,
 639                 NULL,
 640         },
 641
 642         [VCS1] = {
 643                 &decode_info_mi,
 644                 NULL,
 645                 NULL,
 646                 &decode_info_mfx_vc,
 647                 NULL,
 648                 NULL,
 649                 NULL,
 650                 NULL,
 651         },
 652 };
 653
 654 static inline u32 get_opcode(u32 cmd, const struct intel_engine_cs *engine)
 655 {
 656         const struct decode_info *d_info;
 657
 658         d_info = ring_decode_info[engine->id][CMD_TYPE(cmd)];
 659         if (d_info == NULL)
 660                 return INVALID_OP;
 661
 662         return cmd >> (32 - d_info->op_len);
 663 }
 664
 665 static inline const struct cmd_info *
 666 find_cmd_entry(struct intel_gvt *gvt, unsigned int opcode,
 667                const struct intel_engine_cs *engine)
 668 {
 669         struct cmd_entry *e;
 670
 671         hash_for_each_possible(gvt->cmd_table, e, hlist, opcode) {
 672                 if (opcode == e->info->opcode &&
 673                     e->info->rings & engine->mask)
 674                         return e->info;
 675         }
 676         return NULL;
 677 }
 678
 679 static inline const struct cmd_info *
 680 get_cmd_info(struct intel_gvt *gvt, u32 cmd,
 681              const struct intel_engine_cs *engine)
 682 {
 683         u32 opcode;
 684
 685         opcode = get_opcode(cmd, engine);
 686         if (opcode == INVALID_OP)
 687                 return NULL;
 688
 689         return find_cmd_entry(gvt, opcode, engine);
 690 }
 691
 692 static inline u32 sub_op_val(u32 cmd, u32 hi, u32 low)
 693 {
 694         return (cmd >> low) & ((1U << (hi - low + 1)) - 1);
 695 }
 696
 697 static inline void print_opcode(u32 cmd, const struct intel_engine_cs *engine)
 698 {
 699         const struct decode_info *d_info;
 700         int i;
 701
 702         d_info = ring_decode_info[engine->id][CMD_TYPE(cmd)];
 703         if (d_info == NULL)
 704                 return;
 705
 706         gvt_dbg_cmd("opcode=0x%x %s sub_ops:",
 707                         cmd >> (32 - d_info->op_len), d_info->name);
 708
 709         for (i = 0; i < d_info->nr_sub_op; i++)
 710                 pr_err("0x%x ", sub_op_val(cmd, d_info->sub_op[i].hi,
 711                                         d_info->sub_op[i].low));
 712
 713         pr_err("\n");
 714 }
 715
 716 static inline u32 *cmd_ptr(struct parser_exec_state *s, int index)
 717 {
 718         return s->ip_va + (index << 2);
 719 }
 720
 721 static inline u32 cmd_val(struct parser_exec_state *s, int index)
 722 {
 723         return *cmd_ptr(s, index);
 724 }
 725
 726 static inline bool is_init_ctx(struct parser_exec_state *s)
 727 {
 728         return (s->buf_type == RING_BUFFER_CTX && s->is_init_ctx);
 729 }
 730
 731 static void parser_exec_state_dump(struct parser_exec_state *s)
 732 {
 733         int cnt = 0;
 734         int i;
 735
 736         gvt_dbg_cmd("  vgpu%d RING%s: ring_start(%08lx) ring_end(%08lx)"
 737                     " ring_head(%08lx) ring_tail(%08lx)\n",
 738                     s->vgpu->id, s->engine->name,
 739                     s->ring_start, s->ring_start + s->ring_size,
 740                     s->ring_head, s->ring_tail);
 741
 742         gvt_dbg_cmd("  %s %s ip_gma(%08lx) ",
 743                         s->buf_type == RING_BUFFER_INSTRUCTION ?
 744                         "RING_BUFFER" : ((s->buf_type == RING_BUFFER_CTX) ?
 745                                 "CTX_BUFFER" : "BATCH_BUFFER"),
 746                         s->buf_addr_type == GTT_BUFFER ?
 747                         "GTT" : "PPGTT", s->ip_gma);
 748
 749         if (s->ip_va == NULL) {
 750                 gvt_dbg_cmd(" ip_va(NULL)");
 751                 return;
 752         }
 753
 754         gvt_dbg_cmd("  ip_va=%p: %08x %08x %08x %08x\n",
 755                         s->ip_va, cmd_val(s, 0), cmd_val(s, 1),
 756                         cmd_val(s, 2), cmd_val(s, 3));
 757
 758         print_opcode(cmd_val(s, 0), s->engine);
 759
 760         s->ip_va = (u32 *)((((u64)s->ip_va) >> 12) << 12);
 761
 762         while (cnt < 1024) {
 763                 gvt_dbg_cmd("ip_va=%p: ", s->ip_va);
 764                 for (i = 0; i < 8; i++)
 765                         gvt_dbg_cmd("%08x ", cmd_val(s, i));
 766                 gvt_dbg_cmd("\n");
 767
 768                 s->ip_va += 8 * sizeof(u32);
 769                 cnt += 8;
 770         }
 771 }
 772
 773 static inline void update_ip_va(struct parser_exec_state *s)
 774 {
 775         unsigned long len = 0;
 776
 777         if (WARN_ON(s->ring_head == s->ring_tail))
 778                 return;
 779
 780         if (s->buf_type == RING_BUFFER_INSTRUCTION ||
 781                         s->buf_type == RING_BUFFER_CTX) {
 782                 unsigned long ring_top = s->ring_start + s->ring_size;
 783
 784                 if (s->ring_head > s->ring_tail) {
 785                         if (s->ip_gma >= s->ring_head && s->ip_gma < ring_top)
 786                                 len = (s->ip_gma - s->ring_head);
 787                         else if (s->ip_gma >= s->ring_start &&
 788                                         s->ip_gma <= s->ring_tail)
 789                                 len = (ring_top - s->ring_head) +
 790                                         (s->ip_gma - s->ring_start);
 791                 } else
 792                         len = (s->ip_gma - s->ring_head);
 793
 794                 s->ip_va = s->rb_va + len;
 795         } else {/* shadow batch buffer */
 796                 s->ip_va = s->ret_bb_va;
 797         }
 798 }
 799
 800 static inline int ip_gma_set(struct parser_exec_state *s,
 801                 unsigned long ip_gma)
 802 {
 803         WARN_ON(!IS_ALIGNED(ip_gma, 4));
 804
 805         s->ip_gma = ip_gma;
 806         update_ip_va(s);
 807         return 0;
 808 }
 809
 810 static inline int ip_gma_advance(struct parser_exec_state *s,
 811                 unsigned int dw_len)
 812 {
 813         s->ip_gma += (dw_len << 2);
 814
 815         if (s->buf_type == RING_BUFFER_INSTRUCTION) {
 816                 if (s->ip_gma >= s->ring_start + s->ring_size)
 817                         s->ip_gma -= s->ring_size;
 818                 update_ip_va(s);
 819         } else {
 820                 s->ip_va += (dw_len << 2);
 821         }
 822
 823         return 0;
 824 }
 825
 826 static inline int get_cmd_length(const struct cmd_info *info, u32 cmd)
 827 {
 828         if ((info->flag & F_LEN_MASK) == F_LEN_CONST)
 829                 return info->len;
 830         else
 831                 return (cmd & ((1U << info->len) - 1)) + 2;
 832         return 0;
 833 }
 834
 835 static inline int cmd_length(struct parser_exec_state *s)
 836 {
 837         return get_cmd_length(s->info, cmd_val(s, 0));
 838 }
 839
 840 /* do not remove this, some platform may need clflush here */
 841 #define patch_value(s, addr, val) do { \
 842         *addr = val; \
 843 } while (0)
 844
 845 static inline bool is_mocs_mmio(unsigned int offset)
 846 {
 847         return ((offset >= 0xc800) && (offset <= 0xcff8)) ||
 848                 ((offset >= 0xb020) && (offset <= 0xb0a0));
 849 }
 850
 851 static int is_cmd_update_pdps(unsigned int offset,
 852                               struct parser_exec_state *s)
 853 {
 854         u32 base = s->workload->engine->mmio_base;
 855         return i915_mmio_reg_equal(_MMIO(offset), GEN8_RING_PDP_UDW(base, 0));
 856 }
 857
 858 static int cmd_pdp_mmio_update_handler(struct parser_exec_state *s,
 859                                        unsigned int offset, unsigned int index)
 860 {
 861         struct intel_vgpu *vgpu = s->vgpu;
 862         struct intel_vgpu_mm *shadow_mm = s->workload->shadow_mm;
 863         struct intel_vgpu_mm *mm;
 864         u64 pdps[GEN8_3LVL_PDPES];
 865
 866         if (shadow_mm->ppgtt_mm.root_entry_type ==
 867             GTT_TYPE_PPGTT_ROOT_L4_ENTRY) {
 868                 pdps[0] = (u64)cmd_val(s, 2) << 32;
 869                 pdps[0] |= cmd_val(s, 4);
 870
 871                 mm = intel_vgpu_find_ppgtt_mm(vgpu, pdps);
 872                 if (!mm) {
 873                         gvt_vgpu_err("failed to get the 4-level shadow vm\n");
 874                         return -EINVAL;
 875                 }
 876                 intel_vgpu_mm_get(mm);
 877                 list_add_tail(&mm->ppgtt_mm.link,
 878                               &s->workload->lri_shadow_mm);
 879                 *cmd_ptr(s, 2) = upper_32_bits(mm->ppgtt_mm.shadow_pdps[0]);
 880                 *cmd_ptr(s, 4) = lower_32_bits(mm->ppgtt_mm.shadow_pdps[0]);
 881         } else {
 882                 /* Currently all guests use PML4 table and now can't
 883                  * have a guest with 3-level table but uses LRI for
 884                  * PPGTT update. So this is simply un-testable. */
 885                 GEM_BUG_ON(1);
 886                 gvt_vgpu_err("invalid shared shadow vm type\n");
 887                 return -EINVAL;
 888         }
 889         return 0;
 890 }
 891
 892 static int cmd_reg_handler(struct parser_exec_state *s,
 893         unsigned int offset, unsigned int index, char *cmd)
 894 {
 895         struct intel_vgpu *vgpu = s->vgpu;
 896         struct intel_gvt *gvt = vgpu->gvt;
 897         u32 ctx_sr_ctl;
 898         u32 *vreg, vreg_old;
 899
 900         if (offset + 4 > gvt->device_info.mmio_size) {
 901                 gvt_vgpu_err("%s access to (%x) outside of MMIO range\n",
 902                                 cmd, offset);
 903                 return -EFAULT;
 904         }
 905
 906         if (is_init_ctx(s)) {
 907                 struct intel_gvt_mmio_info *mmio_info;
 908
 909                 intel_gvt_mmio_set_cmd_accessible(gvt, offset);
 910                 mmio_info = intel_gvt_find_mmio_info(gvt, offset);
 911                 if (mmio_info && mmio_info->write)
 912                         intel_gvt_mmio_set_cmd_write_patch(gvt, offset);
 913                 return 0;
 914         }
 915
 916         if (!intel_gvt_mmio_is_cmd_accessible(gvt, offset)) {
 917                 gvt_vgpu_err("%s access to non-render register (%x)\n",
 918                                 cmd, offset);
 919                 return -EBADRQC;
 920         }
 921
 922         if (!strncmp(cmd, "srm", 3) ||
 923                         !strncmp(cmd, "lrm", 3)) {
 924                 if (offset == i915_mmio_reg_offset(GEN8_L3SQCREG4) ||
 925                     offset == 0x21f0 ||
 926                     (IS_BROADWELL(gvt->gt->i915) &&
 927                      offset == i915_mmio_reg_offset(INSTPM)))
 928                         return 0;
 929                 else {
 930                         gvt_vgpu_err("%s access to register (%x)\n",
 931                                         cmd, offset);
 932                         return -EPERM;
 933                 }
 934         }
 935
 936         if (!strncmp(cmd, "lrr-src", 7) ||
 937                         !strncmp(cmd, "lrr-dst", 7)) {
 938                 if (IS_BROADWELL(gvt->gt->i915) && offset == 0x215c)
 939                         return 0;
 940                 else {
 941                         gvt_vgpu_err("not allowed cmd %s reg (%x)\n", cmd, offset);
 942                         return -EPERM;
 943                 }
 944         }
 945
 946         if (!strncmp(cmd, "pipe_ctrl", 9)) {
 947                 /* TODO: add LRI POST logic here */
 948                 return 0;
 949         }
 950
 951         if (strncmp(cmd, "lri", 3))
 952                 return -EPERM;
 953
 954         /* below are all lri handlers */
 955         vreg = &vgpu_vreg(s->vgpu, offset);
 956
 957         if (is_cmd_update_pdps(offset, s) &&
 958             cmd_pdp_mmio_update_handler(s, offset, index))
 959                 return -EINVAL;
 960
 961         if (offset == i915_mmio_reg_offset(DERRMR) ||
 962                 offset == i915_mmio_reg_offset(FORCEWAKE_MT)) {
 963                 /* Writing to HW VGT_PVINFO_PAGE offset will be discarded */
 964                 patch_value(s, cmd_ptr(s, index), VGT_PVINFO_PAGE);
 965         }
 966
 967         if (is_mocs_mmio(offset))
 968                 *vreg = cmd_val(s, index + 1);
 969
 970         vreg_old = *vreg;
 971
 972         if (intel_gvt_mmio_is_cmd_write_patch(gvt, offset)) {
 973                 u32 cmdval_new, cmdval;
 974                 struct intel_gvt_mmio_info *mmio_info;
 975
 976                 cmdval = cmd_val(s, index + 1);
 977
 978                 mmio_info = intel_gvt_find_mmio_info(gvt, offset);
 979                 if (!mmio_info) {
 980                         cmdval_new = cmdval;
 981                 } else {
 982                         u64 ro_mask = mmio_info->ro_mask;
 983                         int ret;
 984
 985                         if (likely(!ro_mask))
 986                                 ret = mmio_info->write(s->vgpu, offset,
 987                                                 &cmdval, 4);
 988                         else {
 989                                 gvt_vgpu_err("try to write RO reg %x\n",
 990                                                 offset);
 991                                 ret = -EBADRQC;
 992                         }
 993                         if (ret)
 994                                 return ret;
 995                         cmdval_new = *vreg;
 996                 }
 997                 if (cmdval_new != cmdval)
 998                         patch_value(s, cmd_ptr(s, index+1), cmdval_new);
 999         }
1000
1001         /* only patch cmd. restore vreg value if changed in mmio write handler*/
1002         *vreg = vreg_old;
1003
1004         /* TODO
1005          * In order to let workload with inhibit context to generate
1006          * correct image data into memory, vregs values will be loaded to
1007          * hw via LRIs in the workload with inhibit context. But as
1008          * indirect context is loaded prior to LRIs in workload, we don't
1009          * want reg values specified in indirect context overwritten by
1010          * LRIs in workloads. So, when scanning an indirect context, we
1011          * update reg values in it into vregs, so LRIs in workload with
1012          * inhibit context will restore with correct values
1013          */
1014         if (GRAPHICS_VER(s->engine->i915) == 9 &&
1015             intel_gvt_mmio_is_sr_in_ctx(gvt, offset) &&
1016             !strncmp(cmd, "lri", 3)) {
1017                 intel_gvt_read_gpa(s->vgpu,
1018                         s->workload->ring_context_gpa + 12, &ctx_sr_ctl, 4);
1019                 /* check inhibit context */
1020                 if (ctx_sr_ctl & 1) {
1021                         u32 data = cmd_val(s, index + 1);
1022
1023                         if (intel_gvt_mmio_has_mode_mask(s->vgpu->gvt, offset))
1024                                 intel_vgpu_mask_mmio_write(vgpu,
1025                                                         offset, &data, 4);
1026                         else
1027                                 vgpu_vreg(vgpu, offset) = data;
1028                 }
1029         }
1030
1031         return 0;
1032 }
1033
1034 #define cmd_reg(s, i) \
1035         (cmd_val(s, i) & GENMASK(22, 2))
1036
1037 #define cmd_reg_inhibit(s, i) \
1038         (cmd_val(s, i) & GENMASK(22, 18))
1039
1040 #define cmd_gma(s, i) \
1041         (cmd_val(s, i) & GENMASK(31, 2))
1042
1043 #define cmd_gma_hi(s, i) \
1044         (cmd_val(s, i) & GENMASK(15, 0))
1045
1046 static int cmd_handler_lri(struct parser_exec_state *s)
1047 {
1048         int i, ret = 0;
1049         int cmd_len = cmd_length(s);
1050
1051         for (i = 1; i < cmd_len; i += 2) {
1052                 if (IS_BROADWELL(s->engine->i915) && s->engine->id != RCS0) {
1053                         if (s->engine->id == BCS0 &&
1054                             cmd_reg(s, i) == i915_mmio_reg_offset(DERRMR))
1055                                 ret |= 0;
1056                         else
1057                                 ret |= cmd_reg_inhibit(s, i) ? -EBADRQC : 0;
1058                 }
1059                 if (ret)
1060                         break;
1061                 ret |= cmd_reg_handler(s, cmd_reg(s, i), i, "lri");
1062                 if (ret)
1063                         break;
1064         }
1065         return ret;
1066 }
1067
1068 static int cmd_handler_lrr(struct parser_exec_state *s)
1069 {
1070         int i, ret = 0;
1071         int cmd_len = cmd_length(s);
1072
1073         for (i = 1; i < cmd_len; i += 2) {
1074                 if (IS_BROADWELL(s->engine->i915))
1075                         ret |= ((cmd_reg_inhibit(s, i) ||
1076                                  (cmd_reg_inhibit(s, i + 1)))) ?
1077                                 -EBADRQC : 0;
1078                 if (ret)
1079                         break;
1080                 ret |= cmd_reg_handler(s, cmd_reg(s, i), i, "lrr-src");
1081                 if (ret)
1082                         break;
1083                 ret |= cmd_reg_handler(s, cmd_reg(s, i + 1), i, "lrr-dst");
1084                 if (ret)
1085                         break;
1086         }
1087         return ret;
1088 }
1089
1090 static inline int cmd_address_audit(struct parser_exec_state *s,
1091                 unsigned long guest_gma, int op_size, bool index_mode);
1092
1093 static int cmd_handler_lrm(struct parser_exec_state *s)
1094 {
1095         struct intel_gvt *gvt = s->vgpu->gvt;
1096         int gmadr_bytes = gvt->device_info.gmadr_bytes_in_cmd;
1097         unsigned long gma;
1098         int i, ret = 0;
1099         int cmd_len = cmd_length(s);
1100
1101         for (i = 1; i < cmd_len;) {
1102                 if (IS_BROADWELL(s->engine->i915))
1103                         ret |= (cmd_reg_inhibit(s, i)) ? -EBADRQC : 0;
1104                 if (ret)
1105                         break;
1106                 ret |= cmd_reg_handler(s, cmd_reg(s, i), i, "lrm");
1107                 if (ret)
1108                         break;
1109                 if (cmd_val(s, 0) & (1 << 22)) {
1110                         gma = cmd_gma(s, i + 1);
1111                         if (gmadr_bytes == 8)
1112                                 gma |= (cmd_gma_hi(s, i + 2)) << 32;
1113                         ret |= cmd_address_audit(s, gma, sizeof(u32), false);
1114                         if (ret)
1115                                 break;
1116                 }
1117                 i += gmadr_dw_number(s) + 1;
1118         }
1119         return ret;
1120 }
1121
1122 static int cmd_handler_srm(struct parser_exec_state *s)
1123 {
1124         int gmadr_bytes = s->vgpu->gvt->device_info.gmadr_bytes_in_cmd;
1125         unsigned long gma;
1126         int i, ret = 0;
1127         int cmd_len = cmd_length(s);
1128
1129         for (i = 1; i < cmd_len;) {
1130                 ret |= cmd_reg_handler(s, cmd_reg(s, i), i, "srm");
1131                 if (ret)
1132                         break;
1133                 if (cmd_val(s, 0) & (1 << 22)) {
1134                         gma = cmd_gma(s, i + 1);
1135                         if (gmadr_bytes == 8)
1136                                 gma |= (cmd_gma_hi(s, i + 2)) << 32;
1137                         ret |= cmd_address_audit(s, gma, sizeof(u32), false);
1138                         if (ret)
1139                                 break;
1140                 }
1141                 i += gmadr_dw_number(s) + 1;
1142         }
1143         return ret;
1144 }
1145
1146 struct cmd_interrupt_event {
1147         int pipe_control_notify;
1148         int mi_flush_dw;
1149         int mi_user_interrupt;
1150 };
1151
1152 static const struct cmd_interrupt_event cmd_interrupt_events[] = {
1153         [RCS0] = {
1154                 .pipe_control_notify = RCS_PIPE_CONTROL,
1155                 .mi_flush_dw = INTEL_GVT_EVENT_RESERVED,
1156                 .mi_user_interrupt = RCS_MI_USER_INTERRUPT,
1157         },
1158         [BCS0] = {
1159                 .pipe_control_notify = INTEL_GVT_EVENT_RESERVED,
1160                 .mi_flush_dw = BCS_MI_FLUSH_DW,
1161                 .mi_user_interrupt = BCS_MI_USER_INTERRUPT,
1162         },
1163         [VCS0] = {
1164                 .pipe_control_notify = INTEL_GVT_EVENT_RESERVED,
1165                 .mi_flush_dw = VCS_MI_FLUSH_DW,
1166                 .mi_user_interrupt = VCS_MI_USER_INTERRUPT,
1167         },
1168         [VCS1] = {
1169                 .pipe_control_notify = INTEL_GVT_EVENT_RESERVED,
1170                 .mi_flush_dw = VCS2_MI_FLUSH_DW,
1171                 .mi_user_interrupt = VCS2_MI_USER_INTERRUPT,
1172         },
1173         [VECS0] = {
1174                 .pipe_control_notify = INTEL_GVT_EVENT_RESERVED,
1175                 .mi_flush_dw = VECS_MI_FLUSH_DW,
1176                 .mi_user_interrupt = VECS_MI_USER_INTERRUPT,
1177         },
1178 };
1179
1180 static int cmd_handler_pipe_control(struct parser_exec_state *s)
1181 {
1182         int gmadr_bytes = s->vgpu->gvt->device_info.gmadr_bytes_in_cmd;
1183         unsigned long gma;
1184         bool index_mode = false;
1185         unsigned int post_sync;
1186         int ret = 0;
1187         u32 hws_pga, val;
1188
1189         post_sync = (cmd_val(s, 1) & PIPE_CONTROL_POST_SYNC_OP_MASK) >> 14;
1190
1191         /* LRI post sync */
1192         if (cmd_val(s, 1) & PIPE_CONTROL_MMIO_WRITE)
1193                 ret = cmd_reg_handler(s, cmd_reg(s, 2), 1, "pipe_ctrl");
1194         /* post sync */
1195         else if (post_sync) {
1196                 if (post_sync == 2)
1197                         ret = cmd_reg_handler(s, 0x2350, 1, "pipe_ctrl");
1198                 else if (post_sync == 3)
1199                         ret = cmd_reg_handler(s, 0x2358, 1, "pipe_ctrl");
1200                 else if (post_sync == 1) {
1201                         /* check ggtt*/
1202                         if ((cmd_val(s, 1) & PIPE_CONTROL_GLOBAL_GTT_IVB)) {
1203                                 gma = cmd_val(s, 2) & GENMASK(31, 3);
1204                                 if (gmadr_bytes == 8)
1205                                         gma |= (cmd_gma_hi(s, 3)) << 32;
1206                                 /* Store Data Index */
1207                                 if (cmd_val(s, 1) & (1 << 21))
1208                                         index_mode = true;
1209                                 ret |= cmd_address_audit(s, gma, sizeof(u64),
1210                                                 index_mode);
1211                                 if (ret)
1212                                         return ret;
1213                                 if (index_mode) {
1214                                         hws_pga = s->vgpu->hws_pga[s->engine->id];
1215                                         gma = hws_pga + gma;
1216                                         patch_value(s, cmd_ptr(s, 2), gma);
1217                                         val = cmd_val(s, 1) & (~(1 << 21));
1218                                         patch_value(s, cmd_ptr(s, 1), val);
1219                                 }
1220                         }
1221                 }
1222         }
1223
1224         if (ret)
1225                 return ret;
1226
1227         if (cmd_val(s, 1) & PIPE_CONTROL_NOTIFY)
1228                 set_bit(cmd_interrupt_events[s->engine->id].pipe_control_notify,
1229                         s->workload->pending_events);
1230         return 0;
1231 }
1232
1233 static int cmd_handler_mi_user_interrupt(struct parser_exec_state *s)
1234 {
1235         set_bit(cmd_interrupt_events[s->engine->id].mi_user_interrupt,
1236                 s->workload->pending_events);
1237         patch_value(s, cmd_ptr(s, 0), MI_NOOP);
1238         return 0;
1239 }
1240
1241 static int cmd_advance_default(struct parser_exec_state *s)
1242 {
1243         return ip_gma_advance(s, cmd_length(s));
1244 }
1245
1246 static int cmd_handler_mi_batch_buffer_end(struct parser_exec_state *s)
1247 {
1248         int ret;
1249
1250         if (s->buf_type == BATCH_BUFFER_2ND_LEVEL) {
1251                 s->buf_type = BATCH_BUFFER_INSTRUCTION;
1252                 ret = ip_gma_set(s, s->ret_ip_gma_bb);
1253                 s->buf_addr_type = s->saved_buf_addr_type;
1254         } else if (s->buf_type == RING_BUFFER_CTX) {
1255                 ret = ip_gma_set(s, s->ring_tail);
1256         } else {
1257                 s->buf_type = RING_BUFFER_INSTRUCTION;
1258                 s->buf_addr_type = GTT_BUFFER;
1259                 if (s->ret_ip_gma_ring >= s->ring_start + s->ring_size)
1260                         s->ret_ip_gma_ring -= s->ring_size;
1261                 ret = ip_gma_set(s, s->ret_ip_gma_ring);
1262         }
1263         return ret;
1264 }
1265
1266 struct mi_display_flip_command_info {
1267         int pipe;
1268         int plane;
1269         int event;
1270         i915_reg_t stride_reg;
1271         i915_reg_t ctrl_reg;
1272         i915_reg_t surf_reg;
1273         u64 stride_val;
1274         u64 tile_val;
1275         u64 surf_val;
1276         bool async_flip;
1277 };
1278
1279 struct plane_code_mapping {
1280         int pipe;
1281         int plane;
1282         int event;
1283 };
1284
1285 static int gen8_decode_mi_display_flip(struct parser_exec_state *s,
1286                 struct mi_display_flip_command_info *info)
1287 {
1288         struct drm_i915_private *dev_priv = s->engine->i915;
1289         struct intel_display *display = &dev_priv->display;
1290         struct plane_code_mapping gen8_plane_code[] = {
1291                 [0] = {PIPE_A, PLANE_A, PRIMARY_A_FLIP_DONE},
1292                 [1] = {PIPE_B, PLANE_A, PRIMARY_B_FLIP_DONE},
1293                 [2] = {PIPE_A, PLANE_B, SPRITE_A_FLIP_DONE},
1294                 [3] = {PIPE_B, PLANE_B, SPRITE_B_FLIP_DONE},
1295                 [4] = {PIPE_C, PLANE_A, PRIMARY_C_FLIP_DONE},
1296                 [5] = {PIPE_C, PLANE_B, SPRITE_C_FLIP_DONE},
1297         };
1298         u32 dword0, dword1, dword2;
1299         u32 v;
1300
1301         dword0 = cmd_val(s, 0);
1302         dword1 = cmd_val(s, 1);
1303         dword2 = cmd_val(s, 2);
1304
1305         v = (dword0 & GENMASK(21, 19)) >> 19;
1306         if (drm_WARN_ON(&dev_priv->drm, v >= ARRAY_SIZE(gen8_plane_code)))
1307                 return -EBADRQC;
1308
1309         info->pipe = gen8_plane_code[v].pipe;
1310         info->plane = gen8_plane_code[v].plane;
1311         info->event = gen8_plane_code[v].event;
1312         info->stride_val = (dword1 & GENMASK(15, 6)) >> 6;
1313         info->tile_val = (dword1 & 0x1);
1314         info->surf_val = (dword2 & GENMASK(31, 12)) >> 12;
1315         info->async_flip = ((dword2 & GENMASK(1, 0)) == 0x1);
1316
1317         if (info->plane == PLANE_A) {
1318                 info->ctrl_reg = DSPCNTR(display, info->pipe);
1319                 info->stride_reg = DSPSTRIDE(display, info->pipe);
1320                 info->surf_reg = DSPSURF(display, info->pipe);
1321         } else if (info->plane == PLANE_B) {
1322                 info->ctrl_reg = SPRCTL(info->pipe);
1323                 info->stride_reg = SPRSTRIDE(info->pipe);
1324                 info->surf_reg = SPRSURF(info->pipe);
1325         } else {
1326                 drm_WARN_ON(&dev_priv->drm, 1);
1327                 return -EBADRQC;
1328         }
1329         return 0;
1330 }
1331
1332 static int skl_decode_mi_display_flip(struct parser_exec_state *s,
1333                 struct mi_display_flip_command_info *info)
1334 {
1335         struct drm_i915_private *dev_priv = s->engine->i915;
1336         struct intel_display *display = &dev_priv->display;
1337         struct intel_vgpu *vgpu = s->vgpu;
1338         u32 dword0 = cmd_val(s, 0);
1339         u32 dword1 = cmd_val(s, 1);
1340         u32 dword2 = cmd_val(s, 2);
1341         u32 plane = (dword0 & GENMASK(12, 8)) >> 8;
1342
1343         info->plane = PRIMARY_PLANE;
1344
1345         switch (plane) {
1346         case MI_DISPLAY_FLIP_SKL_PLANE_1_A:
1347                 info->pipe = PIPE_A;
1348                 info->event = PRIMARY_A_FLIP_DONE;
1349                 break;
1350         case MI_DISPLAY_FLIP_SKL_PLANE_1_B:
1351                 info->pipe = PIPE_B;
1352                 info->event = PRIMARY_B_FLIP_DONE;
1353                 break;
1354         case MI_DISPLAY_FLIP_SKL_PLANE_1_C:
1355                 info->pipe = PIPE_C;
1356                 info->event = PRIMARY_C_FLIP_DONE;
1357                 break;
1358
1359         case MI_DISPLAY_FLIP_SKL_PLANE_2_A:
1360                 info->pipe = PIPE_A;
1361                 info->event = SPRITE_A_FLIP_DONE;
1362                 info->plane = SPRITE_PLANE;
1363                 break;
1364         case MI_DISPLAY_FLIP_SKL_PLANE_2_B:
1365                 info->pipe = PIPE_B;
1366                 info->event = SPRITE_B_FLIP_DONE;
1367                 info->plane = SPRITE_PLANE;
1368                 break;
1369         case MI_DISPLAY_FLIP_SKL_PLANE_2_C:
1370                 info->pipe = PIPE_C;
1371                 info->event = SPRITE_C_FLIP_DONE;
1372                 info->plane = SPRITE_PLANE;
1373                 break;
1374
1375         default:
1376                 gvt_vgpu_err("unknown plane code %d\n", plane);
1377                 return -EBADRQC;
1378         }
1379
1380         info->stride_val = (dword1 & GENMASK(15, 6)) >> 6;
1381         info->tile_val = (dword1 & GENMASK(2, 0));
1382         info->surf_val = (dword2 & GENMASK(31, 12)) >> 12;
1383         info->async_flip = ((dword2 & GENMASK(1, 0)) == 0x1);
1384
1385         info->ctrl_reg = DSPCNTR(display, info->pipe);
1386         info->stride_reg = DSPSTRIDE(display, info->pipe);
1387         info->surf_reg = DSPSURF(display, info->pipe);
1388
1389         return 0;
1390 }
1391
1392 static int gen8_check_mi_display_flip(struct parser_exec_state *s,
1393                 struct mi_display_flip_command_info *info)
1394 {
1395         u32 stride, tile;
1396
1397         if (!info->async_flip)
1398                 return 0;
1399
1400         if (GRAPHICS_VER(s->engine->i915) >= 9) {
1401                 stride = vgpu_vreg_t(s->vgpu, info->stride_reg) & GENMASK(9, 0);
1402                 tile = (vgpu_vreg_t(s->vgpu, info->ctrl_reg) &
1403                                 GENMASK(12, 10)) >> 10;
1404         } else {
1405                 stride = (vgpu_vreg_t(s->vgpu, info->stride_reg) &
1406                                 GENMASK(15, 6)) >> 6;
1407                 tile = (vgpu_vreg_t(s->vgpu, info->ctrl_reg) & (1 << 10)) >> 10;
1408         }
1409
1410         if (stride != info->stride_val)
1411                 gvt_dbg_cmd("cannot change stride during async flip\n");
1412
1413         if (tile != info->tile_val)
1414                 gvt_dbg_cmd("cannot change tile during async flip\n");
1415
1416         return 0;
1417 }
1418
1419 static int gen8_update_plane_mmio_from_mi_display_flip(
1420                 struct parser_exec_state *s,
1421                 struct mi_display_flip_command_info *info)
1422 {
1423         struct drm_i915_private *dev_priv = s->engine->i915;
1424         struct intel_display *display = &dev_priv->display;
1425         struct intel_vgpu *vgpu = s->vgpu;
1426
1427         set_mask_bits(&vgpu_vreg_t(vgpu, info->surf_reg), GENMASK(31, 12),
1428                       info->surf_val << 12);
1429         if (GRAPHICS_VER(dev_priv) >= 9) {
1430                 set_mask_bits(&vgpu_vreg_t(vgpu, info->stride_reg), GENMASK(9, 0),
1431                               info->stride_val);
1432                 set_mask_bits(&vgpu_vreg_t(vgpu, info->ctrl_reg), GENMASK(12, 10),
1433                               info->tile_val << 10);
1434         } else {
1435                 set_mask_bits(&vgpu_vreg_t(vgpu, info->stride_reg), GENMASK(15, 6),
1436                               info->stride_val << 6);
1437                 set_mask_bits(&vgpu_vreg_t(vgpu, info->ctrl_reg), GENMASK(10, 10),
1438                               info->tile_val << 10);
1439         }
1440
1441         if (info->plane == PLANE_PRIMARY)
1442                 vgpu_vreg_t(vgpu, PIPE_FLIPCOUNT_G4X(display, info->pipe))++;
1443
1444         if (info->async_flip)
1445                 intel_vgpu_trigger_virtual_event(vgpu, info->event);
1446         else
1447                 set_bit(info->event, vgpu->irq.flip_done_event[info->pipe]);
1448
1449         return 0;
1450 }
1451
1452 static int decode_mi_display_flip(struct parser_exec_state *s,
1453                 struct mi_display_flip_command_info *info)
1454 {
1455         if (IS_BROADWELL(s->engine->i915))
1456                 return gen8_decode_mi_display_flip(s, info);
1457         if (GRAPHICS_VER(s->engine->i915) >= 9)
1458                 return skl_decode_mi_display_flip(s, info);
1459
1460         return -ENODEV;
1461 }
1462
1463 static int check_mi_display_flip(struct parser_exec_state *s,
1464                 struct mi_display_flip_command_info *info)
1465 {
1466         return gen8_check_mi_display_flip(s, info);
1467 }
1468
1469 static int update_plane_mmio_from_mi_display_flip(
1470                 struct parser_exec_state *s,
1471                 struct mi_display_flip_command_info *info)
1472 {
1473         return gen8_update_plane_mmio_from_mi_display_flip(s, info);
1474 }
1475
1476 static int cmd_handler_mi_display_flip(struct parser_exec_state *s)
1477 {
1478         struct mi_display_flip_command_info info;
1479         struct intel_vgpu *vgpu = s->vgpu;
1480         int ret;
1481         int i;
1482         int len = cmd_length(s);
1483         u32 valid_len = CMD_LEN(1);
1484
1485         /* Flip Type == Stereo 3D Flip */
1486         if (DWORD_FIELD(2, 1, 0) == 2)
1487                 valid_len++;
1488         ret = gvt_check_valid_cmd_length(cmd_length(s),
1489                         valid_len);
1490         if (ret)
1491                 return ret;
1492
1493         ret = decode_mi_display_flip(s, &info);
1494         if (ret) {
1495                 gvt_vgpu_err("fail to decode MI display flip command\n");
1496                 return ret;
1497         }
1498
1499         ret = check_mi_display_flip(s, &info);
1500         if (ret) {
1501                 gvt_vgpu_err("invalid MI display flip command\n");
1502                 return ret;
1503         }
1504
1505         ret = update_plane_mmio_from_mi_display_flip(s, &info);
1506         if (ret) {
1507                 gvt_vgpu_err("fail to update plane mmio\n");
1508                 return ret;
1509         }
1510
1511         for (i = 0; i < len; i++)
1512                 patch_value(s, cmd_ptr(s, i), MI_NOOP);
1513         return 0;
1514 }
1515
1516 static bool is_wait_for_flip_pending(u32 cmd)
1517 {
1518         return cmd & (MI_WAIT_FOR_PLANE_A_FLIP_PENDING |
1519                         MI_WAIT_FOR_PLANE_B_FLIP_PENDING |
1520                         MI_WAIT_FOR_PLANE_C_FLIP_PENDING |
1521                         MI_WAIT_FOR_SPRITE_A_FLIP_PENDING |
1522                         MI_WAIT_FOR_SPRITE_B_FLIP_PENDING |
1523                         MI_WAIT_FOR_SPRITE_C_FLIP_PENDING);
1524 }
1525
1526 static int cmd_handler_mi_wait_for_event(struct parser_exec_state *s)
1527 {
1528         u32 cmd = cmd_val(s, 0);
1529
1530         if (!is_wait_for_flip_pending(cmd))
1531                 return 0;
1532
1533         patch_value(s, cmd_ptr(s, 0), MI_NOOP);
1534         return 0;
1535 }
1536
1537 static unsigned long get_gma_bb_from_cmd(struct parser_exec_state *s, int index)
1538 {
1539         unsigned long addr;
1540         unsigned long gma_high, gma_low;
1541         struct intel_vgpu *vgpu = s->vgpu;
1542         int gmadr_bytes = vgpu->gvt->device_info.gmadr_bytes_in_cmd;
1543
1544         if (WARN_ON(gmadr_bytes != 4 && gmadr_bytes != 8)) {
1545                 gvt_vgpu_err("invalid gma bytes %d\n", gmadr_bytes);
1546                 return INTEL_GVT_INVALID_ADDR;
1547         }
1548
1549         gma_low = cmd_val(s, index) & BATCH_BUFFER_ADDR_MASK;
1550         if (gmadr_bytes == 4) {
1551                 addr = gma_low;
1552         } else {
1553                 gma_high = cmd_val(s, index + 1) & BATCH_BUFFER_ADDR_HIGH_MASK;
1554                 addr = (((unsigned long)gma_high) << 32) | gma_low;
1555         }
1556         return addr;
1557 }
1558
1559 static inline int cmd_address_audit(struct parser_exec_state *s,
1560                 unsigned long guest_gma, int op_size, bool index_mode)
1561 {
1562         struct intel_vgpu *vgpu = s->vgpu;
1563         u32 max_surface_size = vgpu->gvt->device_info.max_surface_size;
1564         int i;
1565         int ret;
1566
1567         if (op_size > max_surface_size) {
1568                 gvt_vgpu_err("command address audit fail name %s\n",
1569                         s->info->name);
1570                 return -EFAULT;
1571         }
1572
1573         if (index_mode) {
1574                 if (guest_gma >= I915_GTT_PAGE_SIZE) {
1575                         ret = -EFAULT;
1576                         goto err;
1577                 }
1578         } else if (!intel_gvt_ggtt_validate_range(vgpu, guest_gma, op_size)) {
1579                 ret = -EFAULT;
1580                 goto err;
1581         }
1582
1583         return 0;
1584
1585 err:
1586         gvt_vgpu_err("cmd_parser: Malicious %s detected, addr=0x%lx, len=%d!\n",
1587                         s->info->name, guest_gma, op_size);
1588
1589         pr_err("cmd dump: ");
1590         for (i = 0; i < cmd_length(s); i++) {
1591                 if (!(i % 4))
1592                         pr_err("\n%08x ", cmd_val(s, i));
1593                 else
1594                         pr_err("%08x ", cmd_val(s, i));
1595         }
1596         pr_err("\nvgpu%d: aperture 0x%llx - 0x%llx, hidden 0x%llx - 0x%llx\n",
1597                         vgpu->id,
1598                         vgpu_aperture_gmadr_base(vgpu),
1599                         vgpu_aperture_gmadr_end(vgpu),
1600                         vgpu_hidden_gmadr_base(vgpu),
1601                         vgpu_hidden_gmadr_end(vgpu));
1602         return ret;
1603 }
1604
1605 static int cmd_handler_mi_store_data_imm(struct parser_exec_state *s)
1606 {
1607         int gmadr_bytes = s->vgpu->gvt->device_info.gmadr_bytes_in_cmd;
1608         int op_size = (cmd_length(s) - 3) * sizeof(u32);
1609         int core_id = (cmd_val(s, 2) & (1 << 0)) ? 1 : 0;
1610         unsigned long gma, gma_low, gma_high;
1611         u32 valid_len = CMD_LEN(2);
1612         int ret = 0;
1613
1614         /* check ppggt */
1615         if (!(cmd_val(s, 0) & (1 << 22)))
1616                 return 0;
1617
1618         /* check if QWORD */
1619         if (DWORD_FIELD(0, 21, 21))
1620                 valid_len++;
1621         ret = gvt_check_valid_cmd_length(cmd_length(s),
1622                         valid_len);
1623         if (ret)
1624                 return ret;
1625
1626         gma = cmd_val(s, 2) & GENMASK(31, 2);
1627
1628         if (gmadr_bytes == 8) {
1629                 gma_low = cmd_val(s, 1) & GENMASK(31, 2);
1630                 gma_high = cmd_val(s, 2) & GENMASK(15, 0);
1631                 gma = (gma_high << 32) | gma_low;
1632                 core_id = (cmd_val(s, 1) & (1 << 0)) ? 1 : 0;
1633         }
1634         ret = cmd_address_audit(s, gma + op_size * core_id, op_size, false);
1635         return ret;
1636 }
1637
1638 static inline int unexpected_cmd(struct parser_exec_state *s)
1639 {
1640         struct intel_vgpu *vgpu = s->vgpu;
1641
1642         gvt_vgpu_err("Unexpected %s in command buffer!\n", s->info->name);
1643
1644         return -EBADRQC;
1645 }
1646
1647 static int cmd_handler_mi_semaphore_wait(struct parser_exec_state *s)
1648 {
1649         return unexpected_cmd(s);
1650 }
1651
1652 static int cmd_handler_mi_report_perf_count(struct parser_exec_state *s)
1653 {
1654         return unexpected_cmd(s);
1655 }
1656
1657 static int cmd_handler_mi_op_2e(struct parser_exec_state *s)
1658 {
1659         return unexpected_cmd(s);
1660 }
1661
1662 static int cmd_handler_mi_op_2f(struct parser_exec_state *s)
1663 {
1664         int gmadr_bytes = s->vgpu->gvt->device_info.gmadr_bytes_in_cmd;
1665         int op_size = (1 << ((cmd_val(s, 0) & GENMASK(20, 19)) >> 19)) *
1666                         sizeof(u32);
1667         unsigned long gma, gma_high;
1668         u32 valid_len = CMD_LEN(1);
1669         int ret = 0;
1670
1671         if (!(cmd_val(s, 0) & (1 << 22)))
1672                 return ret;
1673
1674         /* check inline data */
1675         if (cmd_val(s, 0) & BIT(18))
1676                 valid_len = CMD_LEN(9);
1677         ret = gvt_check_valid_cmd_length(cmd_length(s),
1678                         valid_len);
1679         if (ret)
1680                 return ret;
1681
1682         gma = cmd_val(s, 1) & GENMASK(31, 2);
1683         if (gmadr_bytes == 8) {
1684                 gma_high = cmd_val(s, 2) & GENMASK(15, 0);
1685                 gma = (gma_high << 32) | gma;
1686         }
1687         ret = cmd_address_audit(s, gma, op_size, false);
1688         return ret;
1689 }
1690
1691 static int cmd_handler_mi_store_data_index(struct parser_exec_state *s)
1692 {
1693         return unexpected_cmd(s);
1694 }
1695
1696 static int cmd_handler_mi_clflush(struct parser_exec_state *s)
1697 {
1698         return unexpected_cmd(s);
1699 }
1700
1701 static int cmd_handler_mi_conditional_batch_buffer_end(
1702                 struct parser_exec_state *s)
1703 {
1704         return unexpected_cmd(s);
1705 }
1706
1707 static int cmd_handler_mi_update_gtt(struct parser_exec_state *s)
1708 {
1709         return unexpected_cmd(s);
1710 }
1711
1712 static int cmd_handler_mi_flush_dw(struct parser_exec_state *s)
1713 {
1714         int gmadr_bytes = s->vgpu->gvt->device_info.gmadr_bytes_in_cmd;
1715         unsigned long gma;
1716         bool index_mode = false;
1717         int ret = 0;
1718         u32 hws_pga, val;
1719         u32 valid_len = CMD_LEN(2);
1720
1721         ret = gvt_check_valid_cmd_length(cmd_length(s),
1722                         valid_len);
1723         if (ret) {
1724                 /* Check again for Qword */
1725                 ret = gvt_check_valid_cmd_length(cmd_length(s),
1726                         ++valid_len);
1727                 return ret;
1728         }
1729
1730         /* Check post-sync and ppgtt bit */
1731         if (((cmd_val(s, 0) >> 14) & 0x3) && (cmd_val(s, 1) & (1 << 2))) {
1732                 gma = cmd_val(s, 1) & GENMASK(31, 3);
1733                 if (gmadr_bytes == 8)
1734                         gma |= (cmd_val(s, 2) & GENMASK(15, 0)) << 32;
1735                 /* Store Data Index */
1736                 if (cmd_val(s, 0) & (1 << 21))
1737                         index_mode = true;
1738                 ret = cmd_address_audit(s, gma, sizeof(u64), index_mode);
1739                 if (ret)
1740                         return ret;
1741                 if (index_mode) {
1742                         hws_pga = s->vgpu->hws_pga[s->engine->id];
1743                         gma = hws_pga + gma;
1744                         patch_value(s, cmd_ptr(s, 1), gma);
1745                         val = cmd_val(s, 0) & (~(1 << 21));
1746                         patch_value(s, cmd_ptr(s, 0), val);
1747                 }
1748         }
1749         /* Check notify bit */
1750         if ((cmd_val(s, 0) & (1 << 8)))
1751                 set_bit(cmd_interrupt_events[s->engine->id].mi_flush_dw,
1752                         s->workload->pending_events);
1753         return ret;
1754 }
1755
1756 static void addr_type_update_snb(struct parser_exec_state *s)
1757 {
1758         if ((s->buf_type == RING_BUFFER_INSTRUCTION) &&
1759                         (BATCH_BUFFER_ADR_SPACE_BIT(cmd_val(s, 0)) == 1)) {
1760                 s->buf_addr_type = PPGTT_BUFFER;
1761         }
1762 }
1763
1764
1765 static int copy_gma_to_hva(struct intel_vgpu *vgpu, struct intel_vgpu_mm *mm,
1766                 unsigned long gma, unsigned long end_gma, void *va)
1767 {
1768         unsigned long copy_len, offset;
1769         unsigned long len = 0;
1770         unsigned long gpa;
1771
1772         while (gma != end_gma) {
1773                 gpa = intel_vgpu_gma_to_gpa(mm, gma);
1774                 if (gpa == INTEL_GVT_INVALID_ADDR) {
1775                         gvt_vgpu_err("invalid gma address: %lx\n", gma);
1776                         return -EFAULT;
1777                 }
1778
1779                 offset = gma & (I915_GTT_PAGE_SIZE - 1);
1780
1781                 copy_len = (end_gma - gma) >= (I915_GTT_PAGE_SIZE - offset) ?
1782                         I915_GTT_PAGE_SIZE - offset : end_gma - gma;
1783
1784                 intel_gvt_read_gpa(vgpu, gpa, va + len, copy_len);
1785
1786                 len += copy_len;
1787                 gma += copy_len;
1788         }
1789         return len;
1790 }
1791
1792
1793 /*
1794  * Check whether a batch buffer needs to be scanned. Currently
1795  * the only criteria is based on privilege.
1796  */
1797 static int batch_buffer_needs_scan(struct parser_exec_state *s)
1798 {
1799         /* Decide privilege based on address space */
1800         if (cmd_val(s, 0) & BIT(8) &&
1801             !(s->vgpu->scan_nonprivbb & s->engine->mask))
1802                 return 0;
1803
1804         return 1;
1805 }
1806
1807 static const char *repr_addr_type(unsigned int type)
1808 {
1809         return type == PPGTT_BUFFER ? "ppgtt" : "ggtt";
1810 }
1811
1812 static int find_bb_size(struct parser_exec_state *s,
1813                         unsigned long *bb_size,
1814                         unsigned long *bb_end_cmd_offset)
1815 {
1816         unsigned long gma = 0;
1817         const struct cmd_info *info;
1818         u32 cmd_len = 0;
1819         bool bb_end = false;
1820         struct intel_vgpu *vgpu = s->vgpu;
1821         u32 cmd;
1822         struct intel_vgpu_mm *mm = (s->buf_addr_type == GTT_BUFFER) ?
1823                 s->vgpu->gtt.ggtt_mm : s->workload->shadow_mm;
1824
1825         *bb_size = 0;
1826         *bb_end_cmd_offset = 0;
1827
1828         /* get the start gm address of the batch buffer */
1829         gma = get_gma_bb_from_cmd(s, 1);
1830         if (gma == INTEL_GVT_INVALID_ADDR)
1831                 return -EFAULT;
1832
1833         cmd = cmd_val(s, 0);
1834         info = get_cmd_info(s->vgpu->gvt, cmd, s->engine);
1835         if (info == NULL) {
1836                 gvt_vgpu_err("unknown cmd 0x%x, opcode=0x%x, addr_type=%s, ring %s, workload=%p\n",
1837                              cmd, get_opcode(cmd, s->engine),
1838                              repr_addr_type(s->buf_addr_type),
1839                              s->engine->name, s->workload);
1840                 return -EBADRQC;
1841         }
1842         do {
1843                 if (copy_gma_to_hva(s->vgpu, mm,
1844                                     gma, gma + 4, &cmd) < 0)
1845                         return -EFAULT;
1846                 info = get_cmd_info(s->vgpu->gvt, cmd, s->engine);
1847                 if (info == NULL) {
1848                         gvt_vgpu_err("unknown cmd 0x%x, opcode=0x%x, addr_type=%s, ring %s, workload=%p\n",
1849                                      cmd, get_opcode(cmd, s->engine),
1850                                      repr_addr_type(s->buf_addr_type),
1851                                      s->engine->name, s->workload);
1852                         return -EBADRQC;
1853                 }
1854
1855                 if (info->opcode == OP_MI_BATCH_BUFFER_END) {
1856                         bb_end = true;
1857                 } else if (info->opcode == OP_MI_BATCH_BUFFER_START) {
1858                         if (BATCH_BUFFER_2ND_LEVEL_BIT(cmd) == 0)
1859                                 /* chained batch buffer */
1860                                 bb_end = true;
1861                 }
1862
1863                 if (bb_end)
1864                         *bb_end_cmd_offset = *bb_size;
1865
1866                 cmd_len = get_cmd_length(info, cmd) << 2;
1867                 *bb_size += cmd_len;
1868                 gma += cmd_len;
1869         } while (!bb_end);
1870
1871         return 0;
1872 }
1873
1874 static int audit_bb_end(struct parser_exec_state *s, void *va)
1875 {
1876         struct intel_vgpu *vgpu = s->vgpu;
1877         u32 cmd = *(u32 *)va;
1878         const struct cmd_info *info;
1879
1880         info = get_cmd_info(s->vgpu->gvt, cmd, s->engine);
1881         if (info == NULL) {
1882                 gvt_vgpu_err("unknown cmd 0x%x, opcode=0x%x, addr_type=%s, ring %s, workload=%p\n",
1883                              cmd, get_opcode(cmd, s->engine),
1884                              repr_addr_type(s->buf_addr_type),
1885                              s->engine->name, s->workload);
1886                 return -EBADRQC;
1887         }
1888
1889         if ((info->opcode == OP_MI_BATCH_BUFFER_END) ||
1890             ((info->opcode == OP_MI_BATCH_BUFFER_START) &&
1891              (BATCH_BUFFER_2ND_LEVEL_BIT(cmd) == 0)))
1892                 return 0;
1893
1894         return -EBADRQC;
1895 }
1896
1897 static int perform_bb_shadow(struct parser_exec_state *s)
1898 {
1899         struct intel_vgpu *vgpu = s->vgpu;
1900         struct intel_vgpu_shadow_bb *bb;
1901         unsigned long gma = 0;
1902         unsigned long bb_size;
1903         unsigned long bb_end_cmd_offset;
1904         int ret = 0;
1905         struct intel_vgpu_mm *mm = (s->buf_addr_type == GTT_BUFFER) ?
1906                 s->vgpu->gtt.ggtt_mm : s->workload->shadow_mm;
1907         unsigned long start_offset = 0;
1908
1909         /* get the start gm address of the batch buffer */
1910         gma = get_gma_bb_from_cmd(s, 1);
1911         if (gma == INTEL_GVT_INVALID_ADDR)
1912                 return -EFAULT;
1913
1914         ret = find_bb_size(s, &bb_size, &bb_end_cmd_offset);
1915         if (ret)
1916                 return ret;
1917
1918         bb = kzalloc(sizeof(*bb), GFP_KERNEL);
1919         if (!bb)
1920                 return -ENOMEM;
1921
1922         bb->ppgtt = (s->buf_addr_type == GTT_BUFFER) ? false : true;
1923
1924         /* the start_offset stores the batch buffer's start gma's
1925          * offset relative to page boundary. so for non-privileged batch
1926          * buffer, the shadowed gem object holds exactly the same page
1927          * layout as original gem object. This is for the convience of
1928          * replacing the whole non-privilged batch buffer page to this
1929          * shadowed one in PPGTT at the same gma address. (this replacing
1930          * action is not implemented yet now, but may be necessary in
1931          * future).
1932          * for prileged batch buffer, we just change start gma address to
1933          * that of shadowed page.
1934          */
1935         if (bb->ppgtt)
1936                 start_offset = gma & ~I915_GTT_PAGE_MASK;
1937
1938         bb->obj = i915_gem_object_create_shmem(s->engine->i915,
1939                                                round_up(bb_size + start_offset,
1940                                                         PAGE_SIZE));
1941         if (IS_ERR(bb->obj)) {
1942                 ret = PTR_ERR(bb->obj);
1943                 goto err_free_bb;
1944         }
1945
1946         bb->va = i915_gem_object_pin_map(bb->obj, I915_MAP_WB);
1947         if (IS_ERR(bb->va)) {
1948                 ret = PTR_ERR(bb->va);
1949                 goto err_free_obj;
1950         }
1951
1952         ret = copy_gma_to_hva(s->vgpu, mm,
1953                               gma, gma + bb_size,
1954                               bb->va + start_offset);
1955         if (ret < 0) {
1956                 gvt_vgpu_err("fail to copy guest ring buffer\n");
1957                 ret = -EFAULT;
1958                 goto err_unmap;
1959         }
1960
1961         ret = audit_bb_end(s, bb->va + start_offset + bb_end_cmd_offset);
1962         if (ret)
1963                 goto err_unmap;
1964
1965         i915_gem_object_unlock(bb->obj);
1966         INIT_LIST_HEAD(&bb->list);
1967         list_add(&bb->list, &s->workload->shadow_bb);
1968
1969         bb->bb_start_cmd_va = s->ip_va;
1970
1971         if ((s->buf_type == BATCH_BUFFER_INSTRUCTION) && (!s->is_ctx_wa))
1972                 bb->bb_offset = s->ip_va - s->rb_va;
1973         else
1974                 bb->bb_offset = 0;
1975
1976         /*
1977          * ip_va saves the virtual address of the shadow batch buffer, while
1978          * ip_gma saves the graphics address of the original batch buffer.
1979          * As the shadow batch buffer is just a copy from the originial one,
1980          * it should be right to use shadow batch buffer'va and original batch
1981          * buffer's gma in pair. After all, we don't want to pin the shadow
1982          * buffer here (too early).
1983          */
1984         s->ip_va = bb->va + start_offset;
1985         s->ip_gma = gma;
1986         return 0;
1987 err_unmap:
1988         i915_gem_object_unpin_map(bb->obj);
1989 err_free_obj:
1990         i915_gem_object_put(bb->obj);
1991 err_free_bb:
1992         kfree(bb);
1993         return ret;
1994 }
1995
1996 static int cmd_handler_mi_batch_buffer_start(struct parser_exec_state *s)
1997 {
1998         bool second_level;
1999         int ret = 0;
2000         struct intel_vgpu *vgpu = s->vgpu;
2001
2002         if (s->buf_type == BATCH_BUFFER_2ND_LEVEL) {
2003                 gvt_vgpu_err("Found MI_BATCH_BUFFER_START in 2nd level BB\n");
2004                 return -EFAULT;
2005         }
2006
2007         second_level = BATCH_BUFFER_2ND_LEVEL_BIT(cmd_val(s, 0)) == 1;
2008         if (second_level && (s->buf_type != BATCH_BUFFER_INSTRUCTION)) {
2009                 gvt_vgpu_err("Jumping to 2nd level BB from RB is not allowed\n");
2010                 return -EFAULT;
2011         }
2012
2013         s->saved_buf_addr_type = s->buf_addr_type;
2014         addr_type_update_snb(s);
2015         if (s->buf_type == RING_BUFFER_INSTRUCTION) {
2016                 s->ret_ip_gma_ring = s->ip_gma + cmd_length(s) * sizeof(u32);
2017                 s->buf_type = BATCH_BUFFER_INSTRUCTION;
2018         } else if (second_level) {
2019                 s->buf_type = BATCH_BUFFER_2ND_LEVEL;
2020                 s->ret_ip_gma_bb = s->ip_gma + cmd_length(s) * sizeof(u32);
2021                 s->ret_bb_va = s->ip_va + cmd_length(s) * sizeof(u32);
2022         }
2023
2024         if (batch_buffer_needs_scan(s)) {
2025                 ret = perform_bb_shadow(s);
2026                 if (ret < 0)
2027                         gvt_vgpu_err("invalid shadow batch buffer\n");
2028         } else {
2029                 /* emulate a batch buffer end to do return right */
2030                 ret = cmd_handler_mi_batch_buffer_end(s);
2031                 if (ret < 0)
2032                         return ret;
2033         }
2034         return ret;
2035 }
2036
2037 static int mi_noop_index;
2038
2039 static const struct cmd_info cmd_info[] = {
2040         {"MI_NOOP", OP_MI_NOOP, F_LEN_CONST, R_ALL, D_ALL, 0, 1, NULL},
2041
2042         {"MI_SET_PREDICATE", OP_MI_SET_PREDICATE, F_LEN_CONST, R_ALL, D_ALL,
2043                 0, 1, NULL},
2044
2045         {"MI_USER_INTERRUPT", OP_MI_USER_INTERRUPT, F_LEN_CONST, R_ALL, D_ALL,
2046                 0, 1, cmd_handler_mi_user_interrupt},
2047
2048         {"MI_WAIT_FOR_EVENT", OP_MI_WAIT_FOR_EVENT, F_LEN_CONST, R_RCS | R_BCS,
2049                 D_ALL, 0, 1, cmd_handler_mi_wait_for_event},
2050
2051         {"MI_FLUSH", OP_MI_FLUSH, F_LEN_CONST, R_ALL, D_ALL, 0, 1, NULL},
2052
2053         {"MI_ARB_CHECK", OP_MI_ARB_CHECK, F_LEN_CONST, R_ALL, D_ALL, 0, 1,
2054                 NULL},
2055
2056         {"MI_RS_CONTROL", OP_MI_RS_CONTROL, F_LEN_CONST, R_RCS, D_ALL, 0, 1,
2057                 NULL},
2058
2059         {"MI_REPORT_HEAD", OP_MI_REPORT_HEAD, F_LEN_CONST, R_ALL, D_ALL, 0, 1,
2060                 NULL},
2061
2062         {"MI_ARB_ON_OFF", OP_MI_ARB_ON_OFF, F_LEN_CONST, R_ALL, D_ALL, 0, 1,
2063                 NULL},
2064
2065         {"MI_URB_ATOMIC_ALLOC", OP_MI_URB_ATOMIC_ALLOC, F_LEN_CONST, R_RCS,
2066                 D_ALL, 0, 1, NULL},
2067
2068         {"MI_BATCH_BUFFER_END", OP_MI_BATCH_BUFFER_END,
2069                 F_IP_ADVANCE_CUSTOM | F_LEN_CONST, R_ALL, D_ALL, 0, 1,
2070                 cmd_handler_mi_batch_buffer_end},
2071
2072         {"MI_SUSPEND_FLUSH", OP_MI_SUSPEND_FLUSH, F_LEN_CONST, R_ALL, D_ALL,
2073                 0, 1, NULL},
2074
2075         {"MI_PREDICATE", OP_MI_PREDICATE, F_LEN_CONST, R_RCS, D_ALL, 0, 1,
2076                 NULL},
2077
2078         {"MI_TOPOLOGY_FILTER", OP_MI_TOPOLOGY_FILTER, F_LEN_CONST, R_ALL,
2079                 D_ALL, 0, 1, NULL},
2080
2081         {"MI_SET_APPID", OP_MI_SET_APPID, F_LEN_CONST, R_ALL, D_ALL, 0, 1,
2082                 NULL},
2083
2084         {"MI_RS_CONTEXT", OP_MI_RS_CONTEXT, F_LEN_CONST, R_RCS, D_ALL, 0, 1,
2085                 NULL},
2086
2087         {"MI_DISPLAY_FLIP", OP_MI_DISPLAY_FLIP, F_LEN_VAR,
2088                 R_RCS | R_BCS, D_ALL, 0, 8, cmd_handler_mi_display_flip},
2089
2090         {"MI_SEMAPHORE_MBOX", OP_MI_SEMAPHORE_MBOX, F_LEN_VAR | F_LEN_VAR_FIXED,
2091                 R_ALL, D_ALL, 0, 8, NULL, CMD_LEN(1)},
2092
2093         {"MI_MATH", OP_MI_MATH, F_LEN_VAR, R_ALL, D_ALL, 0, 8, NULL},
2094
2095         {"MI_URB_CLEAR", OP_MI_URB_CLEAR, F_LEN_VAR | F_LEN_VAR_FIXED, R_RCS,
2096                 D_ALL, 0, 8, NULL, CMD_LEN(0)},
2097
2098         {"MI_SEMAPHORE_SIGNAL", OP_MI_SEMAPHORE_SIGNAL,
2099                 F_LEN_VAR | F_LEN_VAR_FIXED, R_ALL, D_BDW_PLUS, 0, 8,
2100                 NULL, CMD_LEN(0)},
2101
2102         {"MI_SEMAPHORE_WAIT", OP_MI_SEMAPHORE_WAIT,
2103                 F_LEN_VAR | F_LEN_VAR_FIXED, R_ALL, D_BDW_PLUS, ADDR_FIX_1(2),
2104                 8, cmd_handler_mi_semaphore_wait, CMD_LEN(2)},
2105
2106         {"MI_STORE_DATA_IMM", OP_MI_STORE_DATA_IMM, F_LEN_VAR, R_ALL, D_BDW_PLUS,
2107                 ADDR_FIX_1(1), 10, cmd_handler_mi_store_data_imm},
2108
2109         {"MI_STORE_DATA_INDEX", OP_MI_STORE_DATA_INDEX, F_LEN_VAR, R_ALL, D_ALL,
2110                 0, 8, cmd_handler_mi_store_data_index},
2111
2112         {"MI_LOAD_REGISTER_IMM", OP_MI_LOAD_REGISTER_IMM, F_LEN_VAR, R_ALL,
2113                 D_ALL, 0, 8, cmd_handler_lri},
2114
2115         {"MI_UPDATE_GTT", OP_MI_UPDATE_GTT, F_LEN_VAR, R_ALL, D_BDW_PLUS, 0, 10,
2116                 cmd_handler_mi_update_gtt},
2117
2118         {"MI_STORE_REGISTER_MEM", OP_MI_STORE_REGISTER_MEM,
2119                 F_LEN_VAR | F_LEN_VAR_FIXED, R_ALL, D_ALL, ADDR_FIX_1(2), 8,
2120                 cmd_handler_srm, CMD_LEN(2)},
2121
2122         {"MI_FLUSH_DW", OP_MI_FLUSH_DW, F_LEN_VAR, R_ALL, D_ALL, 0, 6,
2123                 cmd_handler_mi_flush_dw},
2124
2125         {"MI_CLFLUSH", OP_MI_CLFLUSH, F_LEN_VAR, R_ALL, D_ALL, ADDR_FIX_1(1),
2126                 10, cmd_handler_mi_clflush},
2127
2128         {"MI_REPORT_PERF_COUNT", OP_MI_REPORT_PERF_COUNT,
2129                 F_LEN_VAR | F_LEN_VAR_FIXED, R_ALL, D_ALL, ADDR_FIX_1(1), 6,
2130                 cmd_handler_mi_report_perf_count, CMD_LEN(2)},
2131
2132         {"MI_LOAD_REGISTER_MEM", OP_MI_LOAD_REGISTER_MEM,
2133                 F_LEN_VAR | F_LEN_VAR_FIXED, R_ALL, D_ALL, ADDR_FIX_1(2), 8,
2134                 cmd_handler_lrm, CMD_LEN(2)},
2135
2136         {"MI_LOAD_REGISTER_REG", OP_MI_LOAD_REGISTER_REG,
2137                 F_LEN_VAR | F_LEN_VAR_FIXED, R_ALL, D_ALL, 0, 8,
2138                 cmd_handler_lrr, CMD_LEN(1)},
2139
2140         {"MI_RS_STORE_DATA_IMM", OP_MI_RS_STORE_DATA_IMM,
2141                 F_LEN_VAR | F_LEN_VAR_FIXED, R_RCS, D_ALL, 0,
2142                 8, NULL, CMD_LEN(2)},
2143
2144         {"MI_LOAD_URB_MEM", OP_MI_LOAD_URB_MEM, F_LEN_VAR | F_LEN_VAR_FIXED,
2145                 R_RCS, D_ALL, ADDR_FIX_1(2), 8, NULL, CMD_LEN(2)},
2146
2147         {"MI_STORE_URM_MEM", OP_MI_STORE_URM_MEM, F_LEN_VAR, R_RCS, D_ALL,
2148                 ADDR_FIX_1(2), 8, NULL},
2149
2150         {"MI_OP_2E", OP_MI_2E, F_LEN_VAR | F_LEN_VAR_FIXED, R_ALL, D_BDW_PLUS,
2151                 ADDR_FIX_2(1, 2), 8, cmd_handler_mi_op_2e, CMD_LEN(3)},
2152
2153         {"MI_OP_2F", OP_MI_2F, F_LEN_VAR, R_ALL, D_BDW_PLUS, ADDR_FIX_1(1),
2154                 8, cmd_handler_mi_op_2f},
2155
2156         {"MI_BATCH_BUFFER_START", OP_MI_BATCH_BUFFER_START,
2157                 F_IP_ADVANCE_CUSTOM, R_ALL, D_ALL, 0, 8,
2158                 cmd_handler_mi_batch_buffer_start},
2159
2160         {"MI_CONDITIONAL_BATCH_BUFFER_END", OP_MI_CONDITIONAL_BATCH_BUFFER_END,
2161                 F_LEN_VAR | F_LEN_VAR_FIXED, R_ALL, D_ALL, ADDR_FIX_1(2), 8,
2162                 cmd_handler_mi_conditional_batch_buffer_end, CMD_LEN(2)},
2163
2164         {"MI_LOAD_SCAN_LINES_INCL", OP_MI_LOAD_SCAN_LINES_INCL, F_LEN_CONST,
2165                 R_RCS | R_BCS, D_ALL, 0, 2, NULL},
2166
2167         {"XY_SETUP_BLT", OP_XY_SETUP_BLT, F_LEN_VAR, R_BCS, D_ALL,
2168                 ADDR_FIX_2(4, 7), 8, NULL},
2169
2170         {"XY_SETUP_CLIP_BLT", OP_XY_SETUP_CLIP_BLT, F_LEN_VAR, R_BCS, D_ALL,
2171                 0, 8, NULL},
2172
2173         {"XY_SETUP_MONO_PATTERN_SL_BLT", OP_XY_SETUP_MONO_PATTERN_SL_BLT,
2174                 F_LEN_VAR, R_BCS, D_ALL, ADDR_FIX_1(4), 8, NULL},
2175
2176         {"XY_PIXEL_BLT", OP_XY_PIXEL_BLT, F_LEN_VAR, R_BCS, D_ALL, 0, 8, NULL},
2177
2178         {"XY_SCANLINES_BLT", OP_XY_SCANLINES_BLT, F_LEN_VAR, R_BCS, D_ALL,
2179                 0, 8, NULL},
2180
2181         {"XY_TEXT_BLT", OP_XY_TEXT_BLT, F_LEN_VAR, R_BCS, D_ALL,
2182                 ADDR_FIX_1(3), 8, NULL},
2183
2184         {"XY_TEXT_IMMEDIATE_BLT", OP_XY_TEXT_IMMEDIATE_BLT, F_LEN_VAR, R_BCS,
2185                 D_ALL, 0, 8, NULL},
2186
2187         {"XY_COLOR_BLT", OP_XY_COLOR_BLT, F_LEN_VAR, R_BCS, D_ALL,
2188                 ADDR_FIX_1(4), 8, NULL},
2189
2190         {"XY_PAT_BLT", OP_XY_PAT_BLT, F_LEN_VAR, R_BCS, D_ALL,
2191                 ADDR_FIX_2(4, 5), 8, NULL},
2192
2193         {"XY_MONO_PAT_BLT", OP_XY_MONO_PAT_BLT, F_LEN_VAR, R_BCS, D_ALL,
2194                 ADDR_FIX_1(4), 8, NULL},
2195
2196         {"XY_SRC_COPY_BLT", OP_XY_SRC_COPY_BLT, F_LEN_VAR, R_BCS, D_ALL,
2197                 ADDR_FIX_2(4, 7), 8, NULL},
2198
2199         {"XY_MONO_SRC_COPY_BLT", OP_XY_MONO_SRC_COPY_BLT, F_LEN_VAR, R_BCS,
2200                 D_ALL, ADDR_FIX_2(4, 5), 8, NULL},
2201
2202         {"XY_FULL_BLT", OP_XY_FULL_BLT, F_LEN_VAR, R_BCS, D_ALL, 0, 8, NULL},
2203
2204         {"XY_FULL_MONO_SRC_BLT", OP_XY_FULL_MONO_SRC_BLT, F_LEN_VAR, R_BCS,
2205                 D_ALL, ADDR_FIX_3(4, 5, 8), 8, NULL},
2206
2207         {"XY_FULL_MONO_PATTERN_BLT", OP_XY_FULL_MONO_PATTERN_BLT, F_LEN_VAR,
2208                 R_BCS, D_ALL, ADDR_FIX_2(4, 7), 8, NULL},
2209
2210         {"XY_FULL_MONO_PATTERN_MONO_SRC_BLT",
2211                 OP_XY_FULL_MONO_PATTERN_MONO_SRC_BLT,
2212                 F_LEN_VAR, R_BCS, D_ALL, ADDR_FIX_2(4, 5), 8, NULL},
2213
2214         {"XY_MONO_PAT_FIXED_BLT", OP_XY_MONO_PAT_FIXED_BLT, F_LEN_VAR, R_BCS,
2215                 D_ALL, ADDR_FIX_1(4), 8, NULL},
2216
2217         {"XY_MONO_SRC_COPY_IMMEDIATE_BLT", OP_XY_MONO_SRC_COPY_IMMEDIATE_BLT,
2218                 F_LEN_VAR, R_BCS, D_ALL, ADDR_FIX_1(4), 8, NULL},
2219
2220         {"XY_PAT_BLT_IMMEDIATE", OP_XY_PAT_BLT_IMMEDIATE, F_LEN_VAR, R_BCS,
2221                 D_ALL, ADDR_FIX_1(4), 8, NULL},
2222
2223         {"XY_SRC_COPY_CHROMA_BLT", OP_XY_SRC_COPY_CHROMA_BLT, F_LEN_VAR, R_BCS,
2224                 D_ALL, ADDR_FIX_2(4, 7), 8, NULL},
2225
2226         {"XY_FULL_IMMEDIATE_PATTERN_BLT", OP_XY_FULL_IMMEDIATE_PATTERN_BLT,
2227                 F_LEN_VAR, R_BCS, D_ALL, ADDR_FIX_2(4, 7), 8, NULL},
2228
2229         {"XY_FULL_MONO_SRC_IMMEDIATE_PATTERN_BLT",
2230                 OP_XY_FULL_MONO_SRC_IMMEDIATE_PATTERN_BLT,
2231                 F_LEN_VAR, R_BCS, D_ALL, ADDR_FIX_2(4, 5), 8, NULL},
2232
2233         {"XY_PAT_CHROMA_BLT", OP_XY_PAT_CHROMA_BLT, F_LEN_VAR, R_BCS, D_ALL,
2234                 ADDR_FIX_2(4, 5), 8, NULL},
2235
2236         {"XY_PAT_CHROMA_BLT_IMMEDIATE", OP_XY_PAT_CHROMA_BLT_IMMEDIATE,
2237                 F_LEN_VAR, R_BCS, D_ALL, ADDR_FIX_1(4), 8, NULL},
2238
2239         {"3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP",
2240                 OP_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP,
2241                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2242
2243         {"3DSTATE_VIEWPORT_STATE_POINTERS_CC",
2244                 OP_3DSTATE_VIEWPORT_STATE_POINTERS_CC,
2245                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2246
2247         {"3DSTATE_BLEND_STATE_POINTERS",
2248                 OP_3DSTATE_BLEND_STATE_POINTERS,
2249                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2250
2251         {"3DSTATE_DEPTH_STENCIL_STATE_POINTERS",
2252                 OP_3DSTATE_DEPTH_STENCIL_STATE_POINTERS,
2253                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2254
2255         {"3DSTATE_BINDING_TABLE_POINTERS_VS",
2256                 OP_3DSTATE_BINDING_TABLE_POINTERS_VS,
2257                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2258
2259         {"3DSTATE_BINDING_TABLE_POINTERS_HS",
2260                 OP_3DSTATE_BINDING_TABLE_POINTERS_HS,
2261                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2262
2263         {"3DSTATE_BINDING_TABLE_POINTERS_DS",
2264                 OP_3DSTATE_BINDING_TABLE_POINTERS_DS,
2265                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2266
2267         {"3DSTATE_BINDING_TABLE_POINTERS_GS",
2268                 OP_3DSTATE_BINDING_TABLE_POINTERS_GS,
2269                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2270
2271         {"3DSTATE_BINDING_TABLE_POINTERS_PS",
2272                 OP_3DSTATE_BINDING_TABLE_POINTERS_PS,
2273                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2274
2275         {"3DSTATE_SAMPLER_STATE_POINTERS_VS",
2276                 OP_3DSTATE_SAMPLER_STATE_POINTERS_VS,
2277                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2278
2279         {"3DSTATE_SAMPLER_STATE_POINTERS_HS",
2280                 OP_3DSTATE_SAMPLER_STATE_POINTERS_HS,
2281                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2282
2283         {"3DSTATE_SAMPLER_STATE_POINTERS_DS",
2284                 OP_3DSTATE_SAMPLER_STATE_POINTERS_DS,
2285                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2286
2287         {"3DSTATE_SAMPLER_STATE_POINTERS_GS",
2288                 OP_3DSTATE_SAMPLER_STATE_POINTERS_GS,
2289                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2290
2291         {"3DSTATE_SAMPLER_STATE_POINTERS_PS",
2292                 OP_3DSTATE_SAMPLER_STATE_POINTERS_PS,
2293                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2294
2295         {"3DSTATE_URB_VS", OP_3DSTATE_URB_VS, F_LEN_VAR, R_RCS, D_ALL,
2296                 0, 8, NULL},
2297
2298         {"3DSTATE_URB_HS", OP_3DSTATE_URB_HS, F_LEN_VAR, R_RCS, D_ALL,
2299                 0, 8, NULL},
2300
2301         {"3DSTATE_URB_DS", OP_3DSTATE_URB_DS, F_LEN_VAR, R_RCS, D_ALL,
2302                 0, 8, NULL},
2303
2304         {"3DSTATE_URB_GS", OP_3DSTATE_URB_GS, F_LEN_VAR, R_RCS, D_ALL,
2305                 0, 8, NULL},
2306
2307         {"3DSTATE_GATHER_CONSTANT_VS", OP_3DSTATE_GATHER_CONSTANT_VS,
2308                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2309
2310         {"3DSTATE_GATHER_CONSTANT_GS", OP_3DSTATE_GATHER_CONSTANT_GS,
2311                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2312
2313         {"3DSTATE_GATHER_CONSTANT_HS", OP_3DSTATE_GATHER_CONSTANT_HS,
2314                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2315
2316         {"3DSTATE_GATHER_CONSTANT_DS", OP_3DSTATE_GATHER_CONSTANT_DS,
2317                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2318
2319         {"3DSTATE_GATHER_CONSTANT_PS", OP_3DSTATE_GATHER_CONSTANT_PS,
2320                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2321
2322         {"3DSTATE_DX9_CONSTANTF_VS", OP_3DSTATE_DX9_CONSTANTF_VS,
2323                 F_LEN_VAR, R_RCS, D_ALL, 0, 11, NULL},
2324
2325         {"3DSTATE_DX9_CONSTANTF_PS", OP_3DSTATE_DX9_CONSTANTF_PS,
2326                 F_LEN_VAR, R_RCS, D_ALL, 0, 11, NULL},
2327
2328         {"3DSTATE_DX9_CONSTANTI_VS", OP_3DSTATE_DX9_CONSTANTI_VS,
2329                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2330
2331         {"3DSTATE_DX9_CONSTANTI_PS", OP_3DSTATE_DX9_CONSTANTI_PS,
2332                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2333
2334         {"3DSTATE_DX9_CONSTANTB_VS", OP_3DSTATE_DX9_CONSTANTB_VS,
2335                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2336
2337         {"3DSTATE_DX9_CONSTANTB_PS", OP_3DSTATE_DX9_CONSTANTB_PS,
2338                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2339
2340         {"3DSTATE_DX9_LOCAL_VALID_VS", OP_3DSTATE_DX9_LOCAL_VALID_VS,
2341                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2342
2343         {"3DSTATE_DX9_LOCAL_VALID_PS", OP_3DSTATE_DX9_LOCAL_VALID_PS,
2344                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2345
2346         {"3DSTATE_DX9_GENERATE_ACTIVE_VS", OP_3DSTATE_DX9_GENERATE_ACTIVE_VS,
2347                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2348
2349         {"3DSTATE_DX9_GENERATE_ACTIVE_PS", OP_3DSTATE_DX9_GENERATE_ACTIVE_PS,
2350                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2351
2352         {"3DSTATE_BINDING_TABLE_EDIT_VS", OP_3DSTATE_BINDING_TABLE_EDIT_VS,
2353                 F_LEN_VAR, R_RCS, D_ALL, 0, 9, NULL},
2354
2355         {"3DSTATE_BINDING_TABLE_EDIT_GS", OP_3DSTATE_BINDING_TABLE_EDIT_GS,
2356                 F_LEN_VAR, R_RCS, D_ALL, 0, 9, NULL},
2357
2358         {"3DSTATE_BINDING_TABLE_EDIT_HS", OP_3DSTATE_BINDING_TABLE_EDIT_HS,
2359                 F_LEN_VAR, R_RCS, D_ALL, 0, 9, NULL},
2360
2361         {"3DSTATE_BINDING_TABLE_EDIT_DS", OP_3DSTATE_BINDING_TABLE_EDIT_DS,
2362                 F_LEN_VAR, R_RCS, D_ALL, 0, 9, NULL},
2363
2364         {"3DSTATE_BINDING_TABLE_EDIT_PS", OP_3DSTATE_BINDING_TABLE_EDIT_PS,
2365                 F_LEN_VAR, R_RCS, D_ALL, 0, 9, NULL},
2366
2367         {"3DSTATE_VF_INSTANCING", OP_3DSTATE_VF_INSTANCING, F_LEN_VAR, R_RCS,
2368                 D_BDW_PLUS, 0, 8, NULL},
2369
2370         {"3DSTATE_VF_SGVS", OP_3DSTATE_VF_SGVS, F_LEN_VAR, R_RCS, D_BDW_PLUS, 0, 8,
2371                 NULL},
2372
2373         {"3DSTATE_VF_TOPOLOGY", OP_3DSTATE_VF_TOPOLOGY, F_LEN_VAR, R_RCS,
2374                 D_BDW_PLUS, 0, 8, NULL},
2375
2376         {"3DSTATE_WM_CHROMAKEY", OP_3DSTATE_WM_CHROMAKEY, F_LEN_VAR, R_RCS,
2377                 D_BDW_PLUS, 0, 8, NULL},
2378
2379         {"3DSTATE_PS_BLEND", OP_3DSTATE_PS_BLEND, F_LEN_VAR, R_RCS, D_BDW_PLUS, 0,
2380                 8, NULL},
2381
2382         {"3DSTATE_WM_DEPTH_STENCIL", OP_3DSTATE_WM_DEPTH_STENCIL, F_LEN_VAR,
2383                 R_RCS, D_BDW_PLUS, 0, 8, NULL},
2384
2385         {"3DSTATE_PS_EXTRA", OP_3DSTATE_PS_EXTRA, F_LEN_VAR, R_RCS, D_BDW_PLUS, 0,
2386                 8, NULL},
2387
2388         {"3DSTATE_RASTER", OP_3DSTATE_RASTER, F_LEN_VAR, R_RCS, D_BDW_PLUS, 0, 8,
2389                 NULL},
2390
2391         {"3DSTATE_SBE_SWIZ", OP_3DSTATE_SBE_SWIZ, F_LEN_VAR, R_RCS, D_BDW_PLUS, 0, 8,
2392                 NULL},
2393
2394         {"3DSTATE_WM_HZ_OP", OP_3DSTATE_WM_HZ_OP, F_LEN_VAR, R_RCS, D_BDW_PLUS, 0, 8,
2395                 NULL},
2396
2397         {"3DSTATE_VERTEX_BUFFERS", OP_3DSTATE_VERTEX_BUFFERS, F_LEN_VAR, R_RCS,
2398                 D_BDW_PLUS, 0, 8, NULL},
2399
2400         {"3DSTATE_VERTEX_ELEMENTS", OP_3DSTATE_VERTEX_ELEMENTS, F_LEN_VAR,
2401                 R_RCS, D_ALL, 0, 8, NULL},
2402
2403         {"3DSTATE_INDEX_BUFFER", OP_3DSTATE_INDEX_BUFFER, F_LEN_VAR, R_RCS,
2404                 D_BDW_PLUS, ADDR_FIX_1(2), 8, NULL},
2405
2406         {"3DSTATE_VF_STATISTICS", OP_3DSTATE_VF_STATISTICS, F_LEN_CONST,
2407                 R_RCS, D_ALL, 0, 1, NULL},
2408
2409         {"3DSTATE_VF", OP_3DSTATE_VF, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2410
2411         {"3DSTATE_CC_STATE_POINTERS", OP_3DSTATE_CC_STATE_POINTERS, F_LEN_VAR,
2412                 R_RCS, D_ALL, 0, 8, NULL},
2413
2414         {"3DSTATE_SCISSOR_STATE_POINTERS", OP_3DSTATE_SCISSOR_STATE_POINTERS,
2415                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2416
2417         {"3DSTATE_GS", OP_3DSTATE_GS, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2418
2419         {"3DSTATE_CLIP", OP_3DSTATE_CLIP, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2420
2421         {"3DSTATE_WM", OP_3DSTATE_WM, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2422
2423         {"3DSTATE_CONSTANT_GS", OP_3DSTATE_CONSTANT_GS, F_LEN_VAR, R_RCS,
2424                 D_BDW_PLUS, 0, 8, NULL},
2425
2426         {"3DSTATE_CONSTANT_PS", OP_3DSTATE_CONSTANT_PS, F_LEN_VAR, R_RCS,
2427                 D_BDW_PLUS, 0, 8, NULL},
2428
2429         {"3DSTATE_SAMPLE_MASK", OP_3DSTATE_SAMPLE_MASK, F_LEN_VAR, R_RCS,
2430                 D_ALL, 0, 8, NULL},
2431
2432         {"3DSTATE_CONSTANT_HS", OP_3DSTATE_CONSTANT_HS, F_LEN_VAR, R_RCS,
2433                 D_BDW_PLUS, 0, 8, NULL},
2434
2435         {"3DSTATE_CONSTANT_DS", OP_3DSTATE_CONSTANT_DS, F_LEN_VAR, R_RCS,
2436                 D_BDW_PLUS, 0, 8, NULL},
2437
2438         {"3DSTATE_HS", OP_3DSTATE_HS, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2439
2440         {"3DSTATE_TE", OP_3DSTATE_TE, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2441
2442         {"3DSTATE_DS", OP_3DSTATE_DS, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2443
2444         {"3DSTATE_STREAMOUT", OP_3DSTATE_STREAMOUT, F_LEN_VAR, R_RCS,
2445                 D_ALL, 0, 8, NULL},
2446
2447         {"3DSTATE_SBE", OP_3DSTATE_SBE, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2448
2449         {"3DSTATE_PS", OP_3DSTATE_PS, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2450
2451         {"3DSTATE_DRAWING_RECTANGLE", OP_3DSTATE_DRAWING_RECTANGLE, F_LEN_VAR,
2452                 R_RCS, D_ALL, 0, 8, NULL},
2453
2454         {"3DSTATE_SAMPLER_PALETTE_LOAD0", OP_3DSTATE_SAMPLER_PALETTE_LOAD0,
2455                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2456
2457         {"3DSTATE_CHROMA_KEY", OP_3DSTATE_CHROMA_KEY, F_LEN_VAR, R_RCS, D_ALL,
2458                 0, 8, NULL},
2459
2460         {"3DSTATE_DEPTH_BUFFER", OP_3DSTATE_DEPTH_BUFFER, F_LEN_VAR, R_RCS,
2461                 D_ALL, ADDR_FIX_1(2), 8, NULL},
2462
2463         {"3DSTATE_POLY_STIPPLE_OFFSET", OP_3DSTATE_POLY_STIPPLE_OFFSET,
2464                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2465
2466         {"3DSTATE_POLY_STIPPLE_PATTERN", OP_3DSTATE_POLY_STIPPLE_PATTERN,
2467                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2468
2469         {"3DSTATE_LINE_STIPPLE", OP_3DSTATE_LINE_STIPPLE, F_LEN_VAR, R_RCS,
2470                 D_ALL, 0, 8, NULL},
2471
2472         {"3DSTATE_AA_LINE_PARAMS", OP_3DSTATE_AA_LINE_PARAMS, F_LEN_VAR, R_RCS,
2473                 D_ALL, 0, 8, NULL},
2474
2475         {"3DSTATE_GS_SVB_INDEX", OP_3DSTATE_GS_SVB_INDEX, F_LEN_VAR, R_RCS,
2476                 D_ALL, 0, 8, NULL},
2477
2478         {"3DSTATE_SAMPLER_PALETTE_LOAD1", OP_3DSTATE_SAMPLER_PALETTE_LOAD1,
2479                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2480
2481         {"3DSTATE_MULTISAMPLE", OP_3DSTATE_MULTISAMPLE_BDW, F_LEN_VAR, R_RCS,
2482                 D_BDW_PLUS, 0, 8, NULL},
2483
2484         {"3DSTATE_STENCIL_BUFFER", OP_3DSTATE_STENCIL_BUFFER, F_LEN_VAR, R_RCS,
2485                 D_ALL, ADDR_FIX_1(2), 8, NULL},
2486
2487         {"3DSTATE_HIER_DEPTH_BUFFER", OP_3DSTATE_HIER_DEPTH_BUFFER, F_LEN_VAR,
2488                 R_RCS, D_ALL, ADDR_FIX_1(2), 8, NULL},
2489
2490         {"3DSTATE_CLEAR_PARAMS", OP_3DSTATE_CLEAR_PARAMS, F_LEN_VAR,
2491                 R_RCS, D_ALL, 0, 8, NULL},
2492
2493         {"3DSTATE_PUSH_CONSTANT_ALLOC_VS", OP_3DSTATE_PUSH_CONSTANT_ALLOC_VS,
2494                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2495
2496         {"3DSTATE_PUSH_CONSTANT_ALLOC_HS", OP_3DSTATE_PUSH_CONSTANT_ALLOC_HS,
2497                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2498
2499         {"3DSTATE_PUSH_CONSTANT_ALLOC_DS", OP_3DSTATE_PUSH_CONSTANT_ALLOC_DS,
2500                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2501
2502         {"3DSTATE_PUSH_CONSTANT_ALLOC_GS", OP_3DSTATE_PUSH_CONSTANT_ALLOC_GS,
2503                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2504
2505         {"3DSTATE_PUSH_CONSTANT_ALLOC_PS", OP_3DSTATE_PUSH_CONSTANT_ALLOC_PS,
2506                 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2507
2508         {"3DSTATE_MONOFILTER_SIZE", OP_3DSTATE_MONOFILTER_SIZE, F_LEN_VAR,
2509                 R_RCS, D_ALL, 0, 8, NULL},
2510
2511         {"3DSTATE_SO_DECL_LIST", OP_3DSTATE_SO_DECL_LIST, F_LEN_VAR, R_RCS,
2512                 D_ALL, 0, 9, NULL},
2513
2514         {"3DSTATE_SO_BUFFER", OP_3DSTATE_SO_BUFFER, F_LEN_VAR, R_RCS, D_BDW_PLUS,
2515                 ADDR_FIX_2(2, 4), 8, NULL},
2516
2517         {"3DSTATE_BINDING_TABLE_POOL_ALLOC",
2518                 OP_3DSTATE_BINDING_TABLE_POOL_ALLOC,
2519                 F_LEN_VAR, R_RCS, D_BDW_PLUS, ADDR_FIX_1(1), 8, NULL},
2520
2521         {"3DSTATE_GATHER_POOL_ALLOC", OP_3DSTATE_GATHER_POOL_ALLOC,
2522                 F_LEN_VAR, R_RCS, D_BDW_PLUS, ADDR_FIX_1(1), 8, NULL},
2523
2524         {"3DSTATE_DX9_CONSTANT_BUFFER_POOL_ALLOC",
2525                 OP_3DSTATE_DX9_CONSTANT_BUFFER_POOL_ALLOC,
2526                 F_LEN_VAR, R_RCS, D_BDW_PLUS, ADDR_FIX_1(1), 8, NULL},
2527
2528         {"3DSTATE_SAMPLE_PATTERN", OP_3DSTATE_SAMPLE_PATTERN, F_LEN_VAR, R_RCS,
2529                 D_BDW_PLUS, 0, 8, NULL},
2530
2531         {"PIPE_CONTROL", OP_PIPE_CONTROL, F_LEN_VAR, R_RCS, D_ALL,
2532                 ADDR_FIX_1(2), 8, cmd_handler_pipe_control},
2533
2534         {"3DPRIMITIVE", OP_3DPRIMITIVE, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2535
2536         {"PIPELINE_SELECT", OP_PIPELINE_SELECT, F_LEN_CONST, R_RCS, D_ALL, 0,
2537                 1, NULL},
2538
2539         {"STATE_PREFETCH", OP_STATE_PREFETCH, F_LEN_VAR, R_RCS, D_ALL,
2540                 ADDR_FIX_1(1), 8, NULL},
2541
2542         {"STATE_SIP", OP_STATE_SIP, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2543
2544         {"STATE_BASE_ADDRESS", OP_STATE_BASE_ADDRESS, F_LEN_VAR, R_RCS, D_BDW_PLUS,
2545                 ADDR_FIX_5(1, 3, 4, 5, 6), 8, NULL},
2546
2547         {"OP_3D_MEDIA_0_1_4", OP_3D_MEDIA_0_1_4, F_LEN_VAR, R_RCS, D_ALL,
2548                 ADDR_FIX_1(1), 8, NULL},
2549
2550         {"OP_SWTESS_BASE_ADDRESS", OP_SWTESS_BASE_ADDRESS,
2551                 F_LEN_VAR, R_RCS, D_ALL, ADDR_FIX_2(1, 2), 3, NULL},
2552
2553         {"3DSTATE_VS", OP_3DSTATE_VS, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2554
2555         {"3DSTATE_SF", OP_3DSTATE_SF, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2556
2557         {"3DSTATE_CONSTANT_VS", OP_3DSTATE_CONSTANT_VS, F_LEN_VAR, R_RCS, D_BDW_PLUS,
2558                 0, 8, NULL},
2559
2560         {"3DSTATE_COMPONENT_PACKING", OP_3DSTATE_COMPONENT_PACKING, F_LEN_VAR, R_RCS,
2561                 D_SKL_PLUS, 0, 8, NULL},
2562
2563         {"MEDIA_INTERFACE_DESCRIPTOR_LOAD", OP_MEDIA_INTERFACE_DESCRIPTOR_LOAD,
2564                 F_LEN_VAR, R_RCS, D_ALL, 0, 16, NULL},
2565
2566         {"MEDIA_GATEWAY_STATE", OP_MEDIA_GATEWAY_STATE, F_LEN_VAR, R_RCS, D_ALL,
2567                 0, 16, NULL},
2568
2569         {"MEDIA_STATE_FLUSH", OP_MEDIA_STATE_FLUSH, F_LEN_VAR, R_RCS, D_ALL,
2570                 0, 16, NULL},
2571
2572         {"MEDIA_POOL_STATE", OP_MEDIA_POOL_STATE, F_LEN_VAR, R_RCS, D_ALL,
2573                 0, 16, NULL},
2574
2575         {"MEDIA_OBJECT", OP_MEDIA_OBJECT, F_LEN_VAR, R_RCS, D_ALL, 0, 16, NULL},
2576
2577         {"MEDIA_CURBE_LOAD", OP_MEDIA_CURBE_LOAD, F_LEN_VAR, R_RCS, D_ALL,
2578                 0, 16, NULL},
2579
2580         {"MEDIA_OBJECT_PRT", OP_MEDIA_OBJECT_PRT, F_LEN_VAR, R_RCS, D_ALL,
2581                 0, 16, NULL},
2582
2583         {"MEDIA_OBJECT_WALKER", OP_MEDIA_OBJECT_WALKER, F_LEN_VAR, R_RCS, D_ALL,
2584                 0, 16, NULL},
2585
2586         {"GPGPU_WALKER", OP_GPGPU_WALKER, F_LEN_VAR, R_RCS, D_ALL,
2587                 0, 8, NULL},
2588
2589         {"MEDIA_VFE_STATE", OP_MEDIA_VFE_STATE, F_LEN_VAR, R_RCS, D_ALL, 0, 16,
2590                 NULL},
2591
2592         {"3DSTATE_VF_STATISTICS_GM45", OP_3DSTATE_VF_STATISTICS_GM45,
2593                 F_LEN_CONST, R_ALL, D_ALL, 0, 1, NULL},
2594
2595         {"MFX_PIPE_MODE_SELECT", OP_MFX_PIPE_MODE_SELECT, F_LEN_VAR,
2596                 R_VCS, D_ALL, 0, 12, NULL},
2597
2598         {"MFX_SURFACE_STATE", OP_MFX_SURFACE_STATE, F_LEN_VAR,
2599                 R_VCS, D_ALL, 0, 12, NULL},
2600
2601         {"MFX_PIPE_BUF_ADDR_STATE", OP_MFX_PIPE_BUF_ADDR_STATE, F_LEN_VAR,
2602                 R_VCS, D_BDW_PLUS, 0, 12, NULL},
2603
2604         {"MFX_IND_OBJ_BASE_ADDR_STATE", OP_MFX_IND_OBJ_BASE_ADDR_STATE,
2605                 F_LEN_VAR, R_VCS, D_BDW_PLUS, 0, 12, NULL},
2606
2607         {"MFX_BSP_BUF_BASE_ADDR_STATE", OP_MFX_BSP_BUF_BASE_ADDR_STATE,
2608                 F_LEN_VAR, R_VCS, D_BDW_PLUS, ADDR_FIX_3(1, 3, 5), 12, NULL},
2609
2610         {"OP_2_0_0_5", OP_2_0_0_5, F_LEN_VAR, R_VCS, D_BDW_PLUS, 0, 12, NULL},
2611
2612         {"MFX_STATE_POINTER", OP_MFX_STATE_POINTER, F_LEN_VAR,
2613                 R_VCS, D_ALL, 0, 12, NULL},
2614
2615         {"MFX_QM_STATE", OP_MFX_QM_STATE, F_LEN_VAR,
2616                 R_VCS, D_ALL, 0, 12, NULL},
2617
2618         {"MFX_FQM_STATE", OP_MFX_FQM_STATE, F_LEN_VAR,
2619                 R_VCS, D_ALL, 0, 12, NULL},
2620
2621         {"MFX_PAK_INSERT_OBJECT", OP_MFX_PAK_INSERT_OBJECT, F_LEN_VAR,
2622                 R_VCS, D_ALL, 0, 12, NULL},
2623
2624         {"MFX_STITCH_OBJECT", OP_MFX_STITCH_OBJECT, F_LEN_VAR,
2625                 R_VCS, D_ALL, 0, 12, NULL},
2626
2627         {"MFD_IT_OBJECT", OP_MFD_IT_OBJECT, F_LEN_VAR,
2628                 R_VCS, D_ALL, 0, 12, NULL},
2629
2630         {"MFX_WAIT", OP_MFX_WAIT, F_LEN_VAR,
2631                 R_VCS, D_ALL, 0, 6, NULL},
2632
2633         {"MFX_AVC_IMG_STATE", OP_MFX_AVC_IMG_STATE, F_LEN_VAR,
2634                 R_VCS, D_ALL, 0, 12, NULL},
2635
2636         {"MFX_AVC_QM_STATE", OP_MFX_AVC_QM_STATE, F_LEN_VAR,
2637                 R_VCS, D_ALL, 0, 12, NULL},
2638
2639         {"MFX_AVC_DIRECTMODE_STATE", OP_MFX_AVC_DIRECTMODE_STATE, F_LEN_VAR,
2640                 R_VCS, D_ALL, 0, 12, NULL},
2641
2642         {"MFX_AVC_SLICE_STATE", OP_MFX_AVC_SLICE_STATE, F_LEN_VAR,
2643                 R_VCS, D_ALL, 0, 12, NULL},
2644
2645         {"MFX_AVC_REF_IDX_STATE", OP_MFX_AVC_REF_IDX_STATE, F_LEN_VAR,
2646                 R_VCS, D_ALL, 0, 12, NULL},
2647
2648         {"MFX_AVC_WEIGHTOFFSET_STATE", OP_MFX_AVC_WEIGHTOFFSET_STATE, F_LEN_VAR,
2649                 R_VCS, D_ALL, 0, 12, NULL},
2650
2651         {"MFD_AVC_PICID_STATE", OP_MFD_AVC_PICID_STATE, F_LEN_VAR,
2652                 R_VCS, D_ALL, 0, 12, NULL},
2653         {"MFD_AVC_DPB_STATE", OP_MFD_AVC_DPB_STATE, F_LEN_VAR,
2654                 R_VCS, D_ALL, 0, 12, NULL},
2655
2656         {"MFD_AVC_BSD_OBJECT", OP_MFD_AVC_BSD_OBJECT, F_LEN_VAR,
2657                 R_VCS, D_ALL, 0, 12, NULL},
2658
2659         {"MFD_AVC_SLICEADDR", OP_MFD_AVC_SLICEADDR, F_LEN_VAR,
2660                 R_VCS, D_ALL, ADDR_FIX_1(2), 12, NULL},
2661
2662         {"MFC_AVC_PAK_OBJECT", OP_MFC_AVC_PAK_OBJECT, F_LEN_VAR,
2663                 R_VCS, D_ALL, 0, 12, NULL},
2664
2665         {"MFX_VC1_PRED_PIPE_STATE", OP_MFX_VC1_PRED_PIPE_STATE, F_LEN_VAR,
2666                 R_VCS, D_ALL, 0, 12, NULL},
2667
2668         {"MFX_VC1_DIRECTMODE_STATE", OP_MFX_VC1_DIRECTMODE_STATE, F_LEN_VAR,
2669                 R_VCS, D_ALL, 0, 12, NULL},
2670
2671         {"MFD_VC1_SHORT_PIC_STATE", OP_MFD_VC1_SHORT_PIC_STATE, F_LEN_VAR,
2672                 R_VCS, D_ALL, 0, 12, NULL},
2673
2674         {"MFD_VC1_LONG_PIC_STATE", OP_MFD_VC1_LONG_PIC_STATE, F_LEN_VAR,
2675                 R_VCS, D_ALL, 0, 12, NULL},
2676
2677         {"MFD_VC1_BSD_OBJECT", OP_MFD_VC1_BSD_OBJECT, F_LEN_VAR,
2678                 R_VCS, D_ALL, 0, 12, NULL},
2679
2680         {"MFC_MPEG2_SLICEGROUP_STATE", OP_MFC_MPEG2_SLICEGROUP_STATE, F_LEN_VAR,
2681                 R_VCS, D_ALL, 0, 12, NULL},
2682
2683         {"MFC_MPEG2_PAK_OBJECT", OP_MFC_MPEG2_PAK_OBJECT, F_LEN_VAR,
2684                 R_VCS, D_ALL, 0, 12, NULL},
2685
2686         {"MFX_MPEG2_PIC_STATE", OP_MFX_MPEG2_PIC_STATE, F_LEN_VAR,
2687                 R_VCS, D_ALL, 0, 12, NULL},
2688
2689         {"MFX_MPEG2_QM_STATE", OP_MFX_MPEG2_QM_STATE, F_LEN_VAR,
2690                 R_VCS, D_ALL, 0, 12, NULL},
2691
2692         {"MFD_MPEG2_BSD_OBJECT", OP_MFD_MPEG2_BSD_OBJECT, F_LEN_VAR,
2693                 R_VCS, D_ALL, 0, 12, NULL},
2694
2695         {"MFX_2_6_0_0", OP_MFX_2_6_0_0, F_LEN_VAR, R_VCS, D_ALL,
2696                 0, 16, NULL},
2697
2698         {"MFX_2_6_0_9", OP_MFX_2_6_0_9, F_LEN_VAR, R_VCS, D_ALL, 0, 16, NULL},
2699
2700         {"MFX_2_6_0_8", OP_MFX_2_6_0_8, F_LEN_VAR, R_VCS, D_ALL, 0, 16, NULL},
2701
2702         {"MFX_JPEG_PIC_STATE", OP_MFX_JPEG_PIC_STATE, F_LEN_VAR,
2703                 R_VCS, D_ALL, 0, 12, NULL},
2704
2705         {"MFX_JPEG_HUFF_TABLE_STATE", OP_MFX_JPEG_HUFF_TABLE_STATE, F_LEN_VAR,
2706                 R_VCS, D_ALL, 0, 12, NULL},
2707
2708         {"MFD_JPEG_BSD_OBJECT", OP_MFD_JPEG_BSD_OBJECT, F_LEN_VAR,
2709                 R_VCS, D_ALL, 0, 12, NULL},
2710
2711         {"VEBOX_STATE", OP_VEB_STATE, F_LEN_VAR, R_VECS, D_ALL, 0, 12, NULL},
2712
2713         {"VEBOX_SURFACE_STATE", OP_VEB_SURFACE_STATE, F_LEN_VAR, R_VECS, D_ALL,
2714                 0, 12, NULL},
2715
2716         {"VEB_DI_IECP", OP_VEB_DNDI_IECP_STATE, F_LEN_VAR, R_VECS, D_BDW_PLUS,
2717                 0, 12, NULL},
2718 };
2719
2720 static void add_cmd_entry(struct intel_gvt *gvt, struct cmd_entry *e)
2721 {
2722         hash_add(gvt->cmd_table, &e->hlist, e->info->opcode);
2723 }
2724
2725 /* call the cmd handler, and advance ip */
2726 static int cmd_parser_exec(struct parser_exec_state *s)
2727 {
2728         struct intel_vgpu *vgpu = s->vgpu;
2729         const struct cmd_info *info;
2730         u32 cmd;
2731         int ret = 0;
2732
2733         cmd = cmd_val(s, 0);
2734
2735         /* fastpath for MI_NOOP */
2736         if (cmd == MI_NOOP)
2737                 info = &cmd_info[mi_noop_index];
2738         else
2739                 info = get_cmd_info(s->vgpu->gvt, cmd, s->engine);
2740
2741         if (info == NULL) {
2742                 gvt_vgpu_err("unknown cmd 0x%x, opcode=0x%x, addr_type=%s, ring %s, workload=%p\n",
2743                              cmd, get_opcode(cmd, s->engine),
2744                              repr_addr_type(s->buf_addr_type),
2745                              s->engine->name, s->workload);
2746                 return -EBADRQC;
2747         }
2748
2749         s->info = info;
2750
2751         trace_gvt_command(vgpu->id, s->engine->id, s->ip_gma, s->ip_va,
2752                           cmd_length(s), s->buf_type, s->buf_addr_type,
2753                           s->workload, info->name);
2754
2755         if ((info->flag & F_LEN_MASK) == F_LEN_VAR_FIXED) {
2756                 ret = gvt_check_valid_cmd_length(cmd_length(s),
2757                                                  info->valid_len);
2758                 if (ret)
2759                         return ret;
2760         }
2761
2762         if (info->handler) {
2763                 ret = info->handler(s);
2764                 if (ret < 0) {
2765                         gvt_vgpu_err("%s handler error\n", info->name);
2766                         return ret;
2767                 }
2768         }
2769
2770         if (!(info->flag & F_IP_ADVANCE_CUSTOM)) {
2771                 ret = cmd_advance_default(s);
2772                 if (ret) {
2773                         gvt_vgpu_err("%s IP advance error\n", info->name);
2774                         return ret;
2775                 }
2776         }
2777         return 0;
2778 }
2779
2780 static inline bool gma_out_of_range(unsigned long gma,
2781                 unsigned long gma_head, unsigned int gma_tail)
2782 {
2783         if (gma_tail >= gma_head)
2784                 return (gma < gma_head) || (gma > gma_tail);
2785         else
2786                 return (gma > gma_tail) && (gma < gma_head);
2787 }
2788
2789 /* Keep the consistent return type, e.g EBADRQC for unknown
2790  * cmd, EFAULT for invalid address, EPERM for nonpriv. later
2791  * works as the input of VM healthy status.
2792  */
2793 static int command_scan(struct parser_exec_state *s,
2794                 unsigned long rb_head, unsigned long rb_tail,
2795                 unsigned long rb_start, unsigned long rb_len)
2796 {
2797
2798         unsigned long gma_head, gma_tail, gma_bottom;
2799         int ret = 0;
2800         struct intel_vgpu *vgpu = s->vgpu;
2801
2802         gma_head = rb_start + rb_head;
2803         gma_tail = rb_start + rb_tail;
2804         gma_bottom = rb_start +  rb_len;
2805
2806         while (s->ip_gma != gma_tail) {
2807                 if (s->buf_type == RING_BUFFER_INSTRUCTION ||
2808                                 s->buf_type == RING_BUFFER_CTX) {
2809                         if (!(s->ip_gma >= rb_start) ||
2810                                 !(s->ip_gma < gma_bottom)) {
2811                                 gvt_vgpu_err("ip_gma %lx out of ring scope."
2812                                         "(base:0x%lx, bottom: 0x%lx)\n",
2813                                         s->ip_gma, rb_start,
2814                                         gma_bottom);
2815                                 parser_exec_state_dump(s);
2816                                 return -EFAULT;
2817                         }
2818                         if (gma_out_of_range(s->ip_gma, gma_head, gma_tail)) {
2819                                 gvt_vgpu_err("ip_gma %lx out of range."
2820                                         "base 0x%lx head 0x%lx tail 0x%lx\n",
2821                                         s->ip_gma, rb_start,
2822                                         rb_head, rb_tail);
2823                                 parser_exec_state_dump(s);
2824                                 break;
2825                         }
2826                 }
2827                 ret = cmd_parser_exec(s);
2828                 if (ret) {
2829                         gvt_vgpu_err("cmd parser error\n");
2830                         parser_exec_state_dump(s);
2831                         break;
2832                 }
2833         }
2834
2835         return ret;
2836 }
2837
2838 static int scan_workload(struct intel_vgpu_workload *workload)
2839 {
2840         unsigned long gma_head, gma_tail;
2841         struct parser_exec_state s;
2842         int ret = 0;
2843
2844         /* ring base is page aligned */
2845         if (WARN_ON(!IS_ALIGNED(workload->rb_start, I915_GTT_PAGE_SIZE)))
2846                 return -EINVAL;
2847
2848         gma_head = workload->rb_start + workload->rb_head;
2849         gma_tail = workload->rb_start + workload->rb_tail;
2850
2851         s.buf_type = RING_BUFFER_INSTRUCTION;
2852         s.buf_addr_type = GTT_BUFFER;
2853         s.vgpu = workload->vgpu;
2854         s.engine = workload->engine;
2855         s.ring_start = workload->rb_start;
2856         s.ring_size = _RING_CTL_BUF_SIZE(workload->rb_ctl);
2857         s.ring_head = gma_head;
2858         s.ring_tail = gma_tail;
2859         s.rb_va = workload->shadow_ring_buffer_va;
2860         s.workload = workload;
2861         s.is_ctx_wa = false;
2862
2863         if (bypass_scan_mask & workload->engine->mask || gma_head == gma_tail)
2864                 return 0;
2865
2866         ret = ip_gma_set(&s, gma_head);
2867         if (ret)
2868                 goto out;
2869
2870         ret = command_scan(&s, workload->rb_head, workload->rb_tail,
2871                 workload->rb_start, _RING_CTL_BUF_SIZE(workload->rb_ctl));
2872
2873 out:
2874         return ret;
2875 }
2876
2877 static int scan_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
2878 {
2879
2880         unsigned long gma_head, gma_tail, ring_size, ring_tail;
2881         struct parser_exec_state s;
2882         int ret = 0;
2883         struct intel_vgpu_workload *workload = container_of(wa_ctx,
2884                                 struct intel_vgpu_workload,
2885                                 wa_ctx);
2886
2887         /* ring base is page aligned */
2888         if (WARN_ON(!IS_ALIGNED(wa_ctx->indirect_ctx.guest_gma,
2889                                         I915_GTT_PAGE_SIZE)))
2890                 return -EINVAL;
2891
2892         ring_tail = wa_ctx->indirect_ctx.size + 3 * sizeof(u32);
2893         ring_size = round_up(wa_ctx->indirect_ctx.size + CACHELINE_BYTES,
2894                         PAGE_SIZE);
2895         gma_head = wa_ctx->indirect_ctx.guest_gma;
2896         gma_tail = wa_ctx->indirect_ctx.guest_gma + ring_tail;
2897
2898         s.buf_type = RING_BUFFER_INSTRUCTION;
2899         s.buf_addr_type = GTT_BUFFER;
2900         s.vgpu = workload->vgpu;
2901         s.engine = workload->engine;
2902         s.ring_start = wa_ctx->indirect_ctx.guest_gma;
2903         s.ring_size = ring_size;
2904         s.ring_head = gma_head;
2905         s.ring_tail = gma_tail;
2906         s.rb_va = wa_ctx->indirect_ctx.shadow_va;
2907         s.workload = workload;
2908         s.is_ctx_wa = true;
2909
2910         ret = ip_gma_set(&s, gma_head);
2911         if (ret)
2912                 goto out;
2913
2914         ret = command_scan(&s, 0, ring_tail,
2915                 wa_ctx->indirect_ctx.guest_gma, ring_size);
2916 out:
2917         return ret;
2918 }
2919
2920 static int shadow_workload_ring_buffer(struct intel_vgpu_workload *workload)
2921 {
2922         struct intel_vgpu *vgpu = workload->vgpu;
2923         struct intel_vgpu_submission *s = &vgpu->submission;
2924         unsigned long gma_head, gma_tail, gma_top, guest_rb_size;
2925         void *shadow_ring_buffer_va;
2926         int ret;
2927
2928         guest_rb_size = _RING_CTL_BUF_SIZE(workload->rb_ctl);
2929
2930         /* calculate workload ring buffer size */
2931         workload->rb_len = (workload->rb_tail + guest_rb_size -
2932                         workload->rb_head) % guest_rb_size;
2933
2934         gma_head = workload->rb_start + workload->rb_head;
2935         gma_tail = workload->rb_start + workload->rb_tail;
2936         gma_top = workload->rb_start + guest_rb_size;
2937
2938         if (workload->rb_len > s->ring_scan_buffer_size[workload->engine->id]) {
2939                 void *p;
2940
2941                 /* realloc the new ring buffer if needed */
2942                 p = krealloc(s->ring_scan_buffer[workload->engine->id],
2943                              workload->rb_len, GFP_KERNEL);
2944                 if (!p) {
2945                         gvt_vgpu_err("fail to re-alloc ring scan buffer\n");
2946                         return -ENOMEM;
2947                 }
2948                 s->ring_scan_buffer[workload->engine->id] = p;
2949                 s->ring_scan_buffer_size[workload->engine->id] = workload->rb_len;
2950         }
2951
2952         shadow_ring_buffer_va = s->ring_scan_buffer[workload->engine->id];
2953
2954         /* get shadow ring buffer va */
2955         workload->shadow_ring_buffer_va = shadow_ring_buffer_va;
2956
2957         /* head > tail --> copy head <-> top */
2958         if (gma_head > gma_tail) {
2959                 ret = copy_gma_to_hva(vgpu, vgpu->gtt.ggtt_mm,
2960                                       gma_head, gma_top, shadow_ring_buffer_va);
2961                 if (ret < 0) {
2962                         gvt_vgpu_err("fail to copy guest ring buffer\n");
2963                         return ret;
2964                 }
2965                 shadow_ring_buffer_va += ret;
2966                 gma_head = workload->rb_start;
2967         }
2968
2969         /* copy head or start <-> tail */
2970         ret = copy_gma_to_hva(vgpu, vgpu->gtt.ggtt_mm, gma_head, gma_tail,
2971                                 shadow_ring_buffer_va);
2972         if (ret < 0) {
2973                 gvt_vgpu_err("fail to copy guest ring buffer\n");
2974                 return ret;
2975         }
2976         return 0;
2977 }
2978
2979 int intel_gvt_scan_and_shadow_ringbuffer(struct intel_vgpu_workload *workload)
2980 {
2981         int ret;
2982         struct intel_vgpu *vgpu = workload->vgpu;
2983
2984         ret = shadow_workload_ring_buffer(workload);
2985         if (ret) {
2986                 gvt_vgpu_err("fail to shadow workload ring_buffer\n");
2987                 return ret;
2988         }
2989
2990         ret = scan_workload(workload);
2991         if (ret) {
2992                 gvt_vgpu_err("scan workload error\n");
2993                 return ret;
2994         }
2995         return 0;
2996 }
2997
2998 static int shadow_indirect_ctx(struct intel_shadow_wa_ctx *wa_ctx)
2999 {
3000         int ctx_size = wa_ctx->indirect_ctx.size;
3001         unsigned long guest_gma = wa_ctx->indirect_ctx.guest_gma;
3002         struct intel_vgpu_workload *workload = container_of(wa_ctx,
3003                                         struct intel_vgpu_workload,
3004                                         wa_ctx);
3005         struct intel_vgpu *vgpu = workload->vgpu;
3006         struct drm_i915_gem_object *obj;
3007         int ret = 0;
3008         void *map;
3009
3010         obj = i915_gem_object_create_shmem(workload->engine->i915,
3011                                            roundup(ctx_size + CACHELINE_BYTES,
3012                                                    PAGE_SIZE));
3013         if (IS_ERR(obj))
3014                 return PTR_ERR(obj);
3015
3016         /* get the va of the shadow batch buffer */
3017         map = i915_gem_object_pin_map(obj, I915_MAP_WB);
3018         if (IS_ERR(map)) {
3019                 gvt_vgpu_err("failed to vmap shadow indirect ctx\n");
3020                 ret = PTR_ERR(map);
3021                 goto put_obj;
3022         }
3023
3024         i915_gem_object_lock(obj, NULL);
3025         ret = i915_gem_object_set_to_cpu_domain(obj, false);
3026         i915_gem_object_unlock(obj);
3027         if (ret) {
3028                 gvt_vgpu_err("failed to set shadow indirect ctx to CPU\n");
3029                 goto unmap_src;
3030         }
3031
3032         ret = copy_gma_to_hva(workload->vgpu,
3033                                 workload->vgpu->gtt.ggtt_mm,
3034                                 guest_gma, guest_gma + ctx_size,
3035                                 map);
3036         if (ret < 0) {
3037                 gvt_vgpu_err("fail to copy guest indirect ctx\n");
3038                 goto unmap_src;
3039         }
3040
3041         wa_ctx->indirect_ctx.obj = obj;
3042         wa_ctx->indirect_ctx.shadow_va = map;
3043         return 0;
3044
3045 unmap_src:
3046         i915_gem_object_unpin_map(obj);
3047 put_obj:
3048         i915_gem_object_put(obj);
3049         return ret;
3050 }
3051
3052 static int combine_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
3053 {
3054         u32 per_ctx_start[CACHELINE_DWORDS] = {};
3055         unsigned char *bb_start_sva;
3056
3057         if (!wa_ctx->per_ctx.valid)
3058                 return 0;
3059
3060         per_ctx_start[0] = 0x18800001;
3061         per_ctx_start[1] = wa_ctx->per_ctx.guest_gma;
3062
3063         bb_start_sva = (unsigned char *)wa_ctx->indirect_ctx.shadow_va +
3064                                 wa_ctx->indirect_ctx.size;
3065
3066         memcpy(bb_start_sva, per_ctx_start, CACHELINE_BYTES);
3067
3068         return 0;
3069 }
3070
3071 int intel_gvt_scan_and_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
3072 {
3073         int ret;
3074         struct intel_vgpu_workload *workload = container_of(wa_ctx,
3075                                         struct intel_vgpu_workload,
3076                                         wa_ctx);
3077         struct intel_vgpu *vgpu = workload->vgpu;
3078
3079         if (wa_ctx->indirect_ctx.size == 0)
3080                 return 0;
3081
3082         ret = shadow_indirect_ctx(wa_ctx);
3083         if (ret) {
3084                 gvt_vgpu_err("fail to shadow indirect ctx\n");
3085                 return ret;
3086         }
3087
3088         combine_wa_ctx(wa_ctx);
3089
3090         ret = scan_wa_ctx(wa_ctx);
3091         if (ret) {
3092                 gvt_vgpu_err("scan wa ctx error\n");
3093                 return ret;
3094         }
3095
3096         return 0;
3097 }
3098
3099 /* generate dummy contexts by sending empty requests to HW, and let
3100  * the HW to fill Engine Contexts. This dummy contexts are used for
3101  * initialization purpose (update reg whitelist), so referred to as
3102  * init context here
3103  */
3104 void intel_gvt_update_reg_whitelist(struct intel_vgpu *vgpu)
3105 {
3106         const unsigned long start = LRC_STATE_PN * PAGE_SIZE;
3107         struct intel_gvt *gvt = vgpu->gvt;
3108         struct intel_engine_cs *engine;
3109         enum intel_engine_id id;
3110
3111         if (gvt->is_reg_whitelist_updated)
3112                 return;
3113
3114         /* scan init ctx to update cmd accessible list */
3115         for_each_engine(engine, gvt->gt, id) {
3116                 struct parser_exec_state s;
3117                 void *vaddr;
3118                 int ret;
3119
3120                 if (!engine->default_state)
3121                         continue;
3122
3123                 vaddr = shmem_pin_map(engine->default_state);
3124                 if (!vaddr) {
3125                         gvt_err("failed to map %s->default state\n",
3126                                 engine->name);
3127                         return;
3128                 }
3129
3130                 s.buf_type = RING_BUFFER_CTX;
3131                 s.buf_addr_type = GTT_BUFFER;
3132                 s.vgpu = vgpu;
3133                 s.engine = engine;
3134                 s.ring_start = 0;
3135                 s.ring_size = engine->context_size - start;
3136                 s.ring_head = 0;
3137                 s.ring_tail = s.ring_size;
3138                 s.rb_va = vaddr + start;
3139                 s.workload = NULL;
3140                 s.is_ctx_wa = false;
3141                 s.is_init_ctx = true;
3142
3143                 /* skipping the first RING_CTX_SIZE(0x50) dwords */
3144                 ret = ip_gma_set(&s, RING_CTX_SIZE);
3145                 if (ret == 0) {
3146                         ret = command_scan(&s, 0, s.ring_size, 0, s.ring_size);
3147                         if (ret)
3148                                 gvt_err("Scan init ctx error\n");
3149                 }
3150
3151                 shmem_unpin_map(engine->default_state, vaddr);
3152                 if (ret)
3153                         return;
3154         }
3155
3156         gvt->is_reg_whitelist_updated = true;
3157 }
3158
3159 int intel_gvt_scan_engine_context(struct intel_vgpu_workload *workload)
3160 {
3161         struct intel_vgpu *vgpu = workload->vgpu;
3162         unsigned long gma_head, gma_tail, gma_start, ctx_size;
3163         struct parser_exec_state s;
3164         int ring_id = workload->engine->id;
3165         struct intel_context *ce = vgpu->submission.shadow[ring_id];
3166         int ret;
3167
3168         GEM_BUG_ON(atomic_read(&ce->pin_count) < 0);
3169
3170         ctx_size = workload->engine->context_size - PAGE_SIZE;
3171
3172         /* Only ring contxt is loaded to HW for inhibit context, no need to
3173          * scan engine context
3174          */
3175         if (is_inhibit_context(ce))
3176                 return 0;
3177
3178         gma_start = i915_ggtt_offset(ce->state) + LRC_STATE_PN*PAGE_SIZE;
3179         gma_head = 0;
3180         gma_tail = ctx_size;
3181
3182         s.buf_type = RING_BUFFER_CTX;
3183         s.buf_addr_type = GTT_BUFFER;
3184         s.vgpu = workload->vgpu;
3185         s.engine = workload->engine;
3186         s.ring_start = gma_start;
3187         s.ring_size = ctx_size;
3188         s.ring_head = gma_start + gma_head;
3189         s.ring_tail = gma_start + gma_tail;
3190         s.rb_va = ce->lrc_reg_state;
3191         s.workload = workload;
3192         s.is_ctx_wa = false;
3193         s.is_init_ctx = false;
3194
3195         /* don't scan the first RING_CTX_SIZE(0x50) dwords, as it's ring
3196          * context
3197          */
3198         ret = ip_gma_set(&s, gma_start + gma_head + RING_CTX_SIZE);
3199         if (ret)
3200                 goto out;
3201
3202         ret = command_scan(&s, gma_head, gma_tail,
3203                 gma_start, ctx_size);
3204 out:
3205         if (ret)
3206                 gvt_vgpu_err("scan shadow ctx error\n");
3207
3208         return ret;
3209 }
3210
3211 static int init_cmd_table(struct intel_gvt *gvt)
3212 {
3213         unsigned int gen_type = intel_gvt_get_device_type(gvt);
3214         int i;
3215
3216         for (i = 0; i < ARRAY_SIZE(cmd_info); i++) {
3217                 struct cmd_entry *e;
3218
3219                 if (!(cmd_info[i].devices & gen_type))
3220                         continue;
3221
3222                 e = kzalloc(sizeof(*e), GFP_KERNEL);
3223                 if (!e)
3224                         return -ENOMEM;
3225
3226                 e->info = &cmd_info[i];
3227                 if (cmd_info[i].opcode == OP_MI_NOOP)
3228                         mi_noop_index = i;
3229
3230                 INIT_HLIST_NODE(&e->hlist);
3231                 add_cmd_entry(gvt, e);
3232                 gvt_dbg_cmd("add %-30s op %04x flag %x devs %02x rings %02x\n",
3233                             e->info->name, e->info->opcode, e->info->flag,
3234                             e->info->devices, e->info->rings);
3235         }
3236
3237         return 0;
3238 }
3239
3240 static void clean_cmd_table(struct intel_gvt *gvt)
3241 {
3242         struct hlist_node *tmp;
3243         struct cmd_entry *e;
3244         int i;
3245
3246         hash_for_each_safe(gvt->cmd_table, i, tmp, e, hlist)
3247                 kfree(e);
3248
3249         hash_init(gvt->cmd_table);
3250 }
3251
3252 void intel_gvt_clean_cmd_parser(struct intel_gvt *gvt)
3253 {
3254         clean_cmd_table(gvt);
3255 }
3256
3257 int intel_gvt_init_cmd_parser(struct intel_gvt *gvt)
3258 {
3259         int ret;
3260
3261         ret = init_cmd_table(gvt);
3262         if (ret) {
3263                 intel_gvt_clean_cmd_parser(gvt);
3264                 return ret;
3265         }
3266         return 0;
3267 }