2 * Copyright(c) 2011-2016 Intel Corporation. All rights reserved.
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:
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
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
25 * Kevin Tian <kevin.tian@intel.com>
26 * Zhiyuan Lv <zhiyuan.lv@intel.com>
29 * Min He <min.he@intel.com>
30 * Ping Gao <ping.a.gao@intel.com>
31 * Tina Zhang <tina.zhang@intel.com>
32 * Yulei Zhang <yulei.zhang@intel.com>
33 * Zhi Wang <zhi.a.wang@intel.com>
37 #include <linux/slab.h>
40 #include "i915_pvinfo.h"
43 #define INVALID_OP (~0U)
47 #define OP_LEN_3D_MEDIA 16
48 #define OP_LEN_MFX_VC 16
49 #define OP_LEN_VEBOX 16
51 #define CMD_TYPE(cmd) (((cmd) >> 29) & 7)
61 struct sub_op_bits *sub_op;
64 #define MAX_CMD_BUDGET 0x7fffffff
65 #define MI_WAIT_FOR_PLANE_C_FLIP_PENDING (1<<15)
66 #define MI_WAIT_FOR_PLANE_B_FLIP_PENDING (1<<9)
67 #define MI_WAIT_FOR_PLANE_A_FLIP_PENDING (1<<1)
69 #define MI_WAIT_FOR_SPRITE_C_FLIP_PENDING (1<<20)
70 #define MI_WAIT_FOR_SPRITE_B_FLIP_PENDING (1<<10)
71 #define MI_WAIT_FOR_SPRITE_A_FLIP_PENDING (1<<2)
73 /* Render Command Map */
75 /* MI_* command Opcode (28:23) */
76 #define OP_MI_NOOP 0x0
77 #define OP_MI_SET_PREDICATE 0x1 /* HSW+ */
78 #define OP_MI_USER_INTERRUPT 0x2
79 #define OP_MI_WAIT_FOR_EVENT 0x3
80 #define OP_MI_FLUSH 0x4
81 #define OP_MI_ARB_CHECK 0x5
82 #define OP_MI_RS_CONTROL 0x6 /* HSW+ */
83 #define OP_MI_REPORT_HEAD 0x7
84 #define OP_MI_ARB_ON_OFF 0x8
85 #define OP_MI_URB_ATOMIC_ALLOC 0x9 /* HSW+ */
86 #define OP_MI_BATCH_BUFFER_END 0xA
87 #define OP_MI_SUSPEND_FLUSH 0xB
88 #define OP_MI_PREDICATE 0xC /* IVB+ */
89 #define OP_MI_TOPOLOGY_FILTER 0xD /* IVB+ */
90 #define OP_MI_SET_APPID 0xE /* IVB+ */
91 #define OP_MI_RS_CONTEXT 0xF /* HSW+ */
92 #define OP_MI_LOAD_SCAN_LINES_INCL 0x12 /* HSW+ */
93 #define OP_MI_DISPLAY_FLIP 0x14
94 #define OP_MI_SEMAPHORE_MBOX 0x16
95 #define OP_MI_SET_CONTEXT 0x18
96 #define OP_MI_MATH 0x1A
97 #define OP_MI_URB_CLEAR 0x19
98 #define OP_MI_SEMAPHORE_SIGNAL 0x1B /* BDW+ */
99 #define OP_MI_SEMAPHORE_WAIT 0x1C /* BDW+ */
101 #define OP_MI_STORE_DATA_IMM 0x20
102 #define OP_MI_STORE_DATA_INDEX 0x21
103 #define OP_MI_LOAD_REGISTER_IMM 0x22
104 #define OP_MI_UPDATE_GTT 0x23
105 #define OP_MI_STORE_REGISTER_MEM 0x24
106 #define OP_MI_FLUSH_DW 0x26
107 #define OP_MI_CLFLUSH 0x27
108 #define OP_MI_REPORT_PERF_COUNT 0x28
109 #define OP_MI_LOAD_REGISTER_MEM 0x29 /* HSW+ */
110 #define OP_MI_LOAD_REGISTER_REG 0x2A /* HSW+ */
111 #define OP_MI_RS_STORE_DATA_IMM 0x2B /* HSW+ */
112 #define OP_MI_LOAD_URB_MEM 0x2C /* HSW+ */
113 #define OP_MI_STORE_URM_MEM 0x2D /* HSW+ */
114 #define OP_MI_2E 0x2E /* BDW+ */
115 #define OP_MI_2F 0x2F /* BDW+ */
116 #define OP_MI_BATCH_BUFFER_START 0x31
118 /* Bit definition for dword 0 */
119 #define _CMDBIT_BB_START_IN_PPGTT (1UL << 8)
121 #define OP_MI_CONDITIONAL_BATCH_BUFFER_END 0x36
123 #define BATCH_BUFFER_ADDR_MASK ((1UL << 32) - (1U << 2))
124 #define BATCH_BUFFER_ADDR_HIGH_MASK ((1UL << 16) - (1U))
125 #define BATCH_BUFFER_ADR_SPACE_BIT(x) (((x) >> 8) & 1U)
126 #define BATCH_BUFFER_2ND_LEVEL_BIT(x) ((x) >> 22 & 1U)
128 /* 2D command: Opcode (28:22) */
129 #define OP_2D(x) ((2<<7) | x)
131 #define OP_XY_SETUP_BLT OP_2D(0x1)
132 #define OP_XY_SETUP_CLIP_BLT OP_2D(0x3)
133 #define OP_XY_SETUP_MONO_PATTERN_SL_BLT OP_2D(0x11)
134 #define OP_XY_PIXEL_BLT OP_2D(0x24)
135 #define OP_XY_SCANLINES_BLT OP_2D(0x25)
136 #define OP_XY_TEXT_BLT OP_2D(0x26)
137 #define OP_XY_TEXT_IMMEDIATE_BLT OP_2D(0x31)
138 #define OP_XY_COLOR_BLT OP_2D(0x50)
139 #define OP_XY_PAT_BLT OP_2D(0x51)
140 #define OP_XY_MONO_PAT_BLT OP_2D(0x52)
141 #define OP_XY_SRC_COPY_BLT OP_2D(0x53)
142 #define OP_XY_MONO_SRC_COPY_BLT OP_2D(0x54)
143 #define OP_XY_FULL_BLT OP_2D(0x55)
144 #define OP_XY_FULL_MONO_SRC_BLT OP_2D(0x56)
145 #define OP_XY_FULL_MONO_PATTERN_BLT OP_2D(0x57)
146 #define OP_XY_FULL_MONO_PATTERN_MONO_SRC_BLT OP_2D(0x58)
147 #define OP_XY_MONO_PAT_FIXED_BLT OP_2D(0x59)
148 #define OP_XY_MONO_SRC_COPY_IMMEDIATE_BLT OP_2D(0x71)
149 #define OP_XY_PAT_BLT_IMMEDIATE OP_2D(0x72)
150 #define OP_XY_SRC_COPY_CHROMA_BLT OP_2D(0x73)
151 #define OP_XY_FULL_IMMEDIATE_PATTERN_BLT OP_2D(0x74)
152 #define OP_XY_FULL_MONO_SRC_IMMEDIATE_PATTERN_BLT OP_2D(0x75)
153 #define OP_XY_PAT_CHROMA_BLT OP_2D(0x76)
154 #define OP_XY_PAT_CHROMA_BLT_IMMEDIATE OP_2D(0x77)
156 /* 3D/Media Command: Pipeline Type(28:27) Opcode(26:24) Sub Opcode(23:16) */
157 #define OP_3D_MEDIA(sub_type, opcode, sub_opcode) \
158 ((3 << 13) | ((sub_type) << 11) | ((opcode) << 8) | (sub_opcode))
160 #define OP_STATE_PREFETCH OP_3D_MEDIA(0x0, 0x0, 0x03)
162 #define OP_STATE_BASE_ADDRESS OP_3D_MEDIA(0x0, 0x1, 0x01)
163 #define OP_STATE_SIP OP_3D_MEDIA(0x0, 0x1, 0x02)
164 #define OP_3D_MEDIA_0_1_4 OP_3D_MEDIA(0x0, 0x1, 0x04)
166 #define OP_3DSTATE_VF_STATISTICS_GM45 OP_3D_MEDIA(0x1, 0x0, 0x0B)
168 #define OP_PIPELINE_SELECT OP_3D_MEDIA(0x1, 0x1, 0x04)
170 #define OP_MEDIA_VFE_STATE OP_3D_MEDIA(0x2, 0x0, 0x0)
171 #define OP_MEDIA_CURBE_LOAD OP_3D_MEDIA(0x2, 0x0, 0x1)
172 #define OP_MEDIA_INTERFACE_DESCRIPTOR_LOAD OP_3D_MEDIA(0x2, 0x0, 0x2)
173 #define OP_MEDIA_GATEWAY_STATE OP_3D_MEDIA(0x2, 0x0, 0x3)
174 #define OP_MEDIA_STATE_FLUSH OP_3D_MEDIA(0x2, 0x0, 0x4)
176 #define OP_MEDIA_OBJECT OP_3D_MEDIA(0x2, 0x1, 0x0)
177 #define OP_MEDIA_OBJECT_PRT OP_3D_MEDIA(0x2, 0x1, 0x2)
178 #define OP_MEDIA_OBJECT_WALKER OP_3D_MEDIA(0x2, 0x1, 0x3)
179 #define OP_GPGPU_WALKER OP_3D_MEDIA(0x2, 0x1, 0x5)
181 #define OP_3DSTATE_CLEAR_PARAMS OP_3D_MEDIA(0x3, 0x0, 0x04) /* IVB+ */
182 #define OP_3DSTATE_DEPTH_BUFFER OP_3D_MEDIA(0x3, 0x0, 0x05) /* IVB+ */
183 #define OP_3DSTATE_STENCIL_BUFFER OP_3D_MEDIA(0x3, 0x0, 0x06) /* IVB+ */
184 #define OP_3DSTATE_HIER_DEPTH_BUFFER OP_3D_MEDIA(0x3, 0x0, 0x07) /* IVB+ */
185 #define OP_3DSTATE_VERTEX_BUFFERS OP_3D_MEDIA(0x3, 0x0, 0x08)
186 #define OP_3DSTATE_VERTEX_ELEMENTS OP_3D_MEDIA(0x3, 0x0, 0x09)
187 #define OP_3DSTATE_INDEX_BUFFER OP_3D_MEDIA(0x3, 0x0, 0x0A)
188 #define OP_3DSTATE_VF_STATISTICS OP_3D_MEDIA(0x3, 0x0, 0x0B)
189 #define OP_3DSTATE_VF OP_3D_MEDIA(0x3, 0x0, 0x0C) /* HSW+ */
190 #define OP_3DSTATE_CC_STATE_POINTERS OP_3D_MEDIA(0x3, 0x0, 0x0E)
191 #define OP_3DSTATE_SCISSOR_STATE_POINTERS OP_3D_MEDIA(0x3, 0x0, 0x0F)
192 #define OP_3DSTATE_VS OP_3D_MEDIA(0x3, 0x0, 0x10)
193 #define OP_3DSTATE_GS OP_3D_MEDIA(0x3, 0x0, 0x11)
194 #define OP_3DSTATE_CLIP OP_3D_MEDIA(0x3, 0x0, 0x12)
195 #define OP_3DSTATE_SF OP_3D_MEDIA(0x3, 0x0, 0x13)
196 #define OP_3DSTATE_WM OP_3D_MEDIA(0x3, 0x0, 0x14)
197 #define OP_3DSTATE_CONSTANT_VS OP_3D_MEDIA(0x3, 0x0, 0x15)
198 #define OP_3DSTATE_CONSTANT_GS OP_3D_MEDIA(0x3, 0x0, 0x16)
199 #define OP_3DSTATE_CONSTANT_PS OP_3D_MEDIA(0x3, 0x0, 0x17)
200 #define OP_3DSTATE_SAMPLE_MASK OP_3D_MEDIA(0x3, 0x0, 0x18)
201 #define OP_3DSTATE_CONSTANT_HS OP_3D_MEDIA(0x3, 0x0, 0x19) /* IVB+ */
202 #define OP_3DSTATE_CONSTANT_DS OP_3D_MEDIA(0x3, 0x0, 0x1A) /* IVB+ */
203 #define OP_3DSTATE_HS OP_3D_MEDIA(0x3, 0x0, 0x1B) /* IVB+ */
204 #define OP_3DSTATE_TE OP_3D_MEDIA(0x3, 0x0, 0x1C) /* IVB+ */
205 #define OP_3DSTATE_DS OP_3D_MEDIA(0x3, 0x0, 0x1D) /* IVB+ */
206 #define OP_3DSTATE_STREAMOUT OP_3D_MEDIA(0x3, 0x0, 0x1E) /* IVB+ */
207 #define OP_3DSTATE_SBE OP_3D_MEDIA(0x3, 0x0, 0x1F) /* IVB+ */
208 #define OP_3DSTATE_PS OP_3D_MEDIA(0x3, 0x0, 0x20) /* IVB+ */
209 #define OP_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP OP_3D_MEDIA(0x3, 0x0, 0x21) /* IVB+ */
210 #define OP_3DSTATE_VIEWPORT_STATE_POINTERS_CC OP_3D_MEDIA(0x3, 0x0, 0x23) /* IVB+ */
211 #define OP_3DSTATE_BLEND_STATE_POINTERS OP_3D_MEDIA(0x3, 0x0, 0x24) /* IVB+ */
212 #define OP_3DSTATE_DEPTH_STENCIL_STATE_POINTERS OP_3D_MEDIA(0x3, 0x0, 0x25) /* IVB+ */
213 #define OP_3DSTATE_BINDING_TABLE_POINTERS_VS OP_3D_MEDIA(0x3, 0x0, 0x26) /* IVB+ */
214 #define OP_3DSTATE_BINDING_TABLE_POINTERS_HS OP_3D_MEDIA(0x3, 0x0, 0x27) /* IVB+ */
215 #define OP_3DSTATE_BINDING_TABLE_POINTERS_DS OP_3D_MEDIA(0x3, 0x0, 0x28) /* IVB+ */
216 #define OP_3DSTATE_BINDING_TABLE_POINTERS_GS OP_3D_MEDIA(0x3, 0x0, 0x29) /* IVB+ */
217 #define OP_3DSTATE_BINDING_TABLE_POINTERS_PS OP_3D_MEDIA(0x3, 0x0, 0x2A) /* IVB+ */
218 #define OP_3DSTATE_SAMPLER_STATE_POINTERS_VS OP_3D_MEDIA(0x3, 0x0, 0x2B) /* IVB+ */
219 #define OP_3DSTATE_SAMPLER_STATE_POINTERS_HS OP_3D_MEDIA(0x3, 0x0, 0x2C) /* IVB+ */
220 #define OP_3DSTATE_SAMPLER_STATE_POINTERS_DS OP_3D_MEDIA(0x3, 0x0, 0x2D) /* IVB+ */
221 #define OP_3DSTATE_SAMPLER_STATE_POINTERS_GS OP_3D_MEDIA(0x3, 0x0, 0x2E) /* IVB+ */
222 #define OP_3DSTATE_SAMPLER_STATE_POINTERS_PS OP_3D_MEDIA(0x3, 0x0, 0x2F) /* IVB+ */
223 #define OP_3DSTATE_URB_VS OP_3D_MEDIA(0x3, 0x0, 0x30) /* IVB+ */
224 #define OP_3DSTATE_URB_HS OP_3D_MEDIA(0x3, 0x0, 0x31) /* IVB+ */
225 #define OP_3DSTATE_URB_DS OP_3D_MEDIA(0x3, 0x0, 0x32) /* IVB+ */
226 #define OP_3DSTATE_URB_GS OP_3D_MEDIA(0x3, 0x0, 0x33) /* IVB+ */
227 #define OP_3DSTATE_GATHER_CONSTANT_VS OP_3D_MEDIA(0x3, 0x0, 0x34) /* HSW+ */
228 #define OP_3DSTATE_GATHER_CONSTANT_GS OP_3D_MEDIA(0x3, 0x0, 0x35) /* HSW+ */
229 #define OP_3DSTATE_GATHER_CONSTANT_HS OP_3D_MEDIA(0x3, 0x0, 0x36) /* HSW+ */
230 #define OP_3DSTATE_GATHER_CONSTANT_DS OP_3D_MEDIA(0x3, 0x0, 0x37) /* HSW+ */
231 #define OP_3DSTATE_GATHER_CONSTANT_PS OP_3D_MEDIA(0x3, 0x0, 0x38) /* HSW+ */
232 #define OP_3DSTATE_DX9_CONSTANTF_VS OP_3D_MEDIA(0x3, 0x0, 0x39) /* HSW+ */
233 #define OP_3DSTATE_DX9_CONSTANTF_PS OP_3D_MEDIA(0x3, 0x0, 0x3A) /* HSW+ */
234 #define OP_3DSTATE_DX9_CONSTANTI_VS OP_3D_MEDIA(0x3, 0x0, 0x3B) /* HSW+ */
235 #define OP_3DSTATE_DX9_CONSTANTI_PS OP_3D_MEDIA(0x3, 0x0, 0x3C) /* HSW+ */
236 #define OP_3DSTATE_DX9_CONSTANTB_VS OP_3D_MEDIA(0x3, 0x0, 0x3D) /* HSW+ */
237 #define OP_3DSTATE_DX9_CONSTANTB_PS OP_3D_MEDIA(0x3, 0x0, 0x3E) /* HSW+ */
238 #define OP_3DSTATE_DX9_LOCAL_VALID_VS OP_3D_MEDIA(0x3, 0x0, 0x3F) /* HSW+ */
239 #define OP_3DSTATE_DX9_LOCAL_VALID_PS OP_3D_MEDIA(0x3, 0x0, 0x40) /* HSW+ */
240 #define OP_3DSTATE_DX9_GENERATE_ACTIVE_VS OP_3D_MEDIA(0x3, 0x0, 0x41) /* HSW+ */
241 #define OP_3DSTATE_DX9_GENERATE_ACTIVE_PS OP_3D_MEDIA(0x3, 0x0, 0x42) /* HSW+ */
242 #define OP_3DSTATE_BINDING_TABLE_EDIT_VS OP_3D_MEDIA(0x3, 0x0, 0x43) /* HSW+ */
243 #define OP_3DSTATE_BINDING_TABLE_EDIT_GS OP_3D_MEDIA(0x3, 0x0, 0x44) /* HSW+ */
244 #define OP_3DSTATE_BINDING_TABLE_EDIT_HS OP_3D_MEDIA(0x3, 0x0, 0x45) /* HSW+ */
245 #define OP_3DSTATE_BINDING_TABLE_EDIT_DS OP_3D_MEDIA(0x3, 0x0, 0x46) /* HSW+ */
246 #define OP_3DSTATE_BINDING_TABLE_EDIT_PS OP_3D_MEDIA(0x3, 0x0, 0x47) /* HSW+ */
248 #define OP_3DSTATE_VF_INSTANCING OP_3D_MEDIA(0x3, 0x0, 0x49) /* BDW+ */
249 #define OP_3DSTATE_VF_SGVS OP_3D_MEDIA(0x3, 0x0, 0x4A) /* BDW+ */
250 #define OP_3DSTATE_VF_TOPOLOGY OP_3D_MEDIA(0x3, 0x0, 0x4B) /* BDW+ */
251 #define OP_3DSTATE_WM_CHROMAKEY OP_3D_MEDIA(0x3, 0x0, 0x4C) /* BDW+ */
252 #define OP_3DSTATE_PS_BLEND OP_3D_MEDIA(0x3, 0x0, 0x4D) /* BDW+ */
253 #define OP_3DSTATE_WM_DEPTH_STENCIL OP_3D_MEDIA(0x3, 0x0, 0x4E) /* BDW+ */
254 #define OP_3DSTATE_PS_EXTRA OP_3D_MEDIA(0x3, 0x0, 0x4F) /* BDW+ */
255 #define OP_3DSTATE_RASTER OP_3D_MEDIA(0x3, 0x0, 0x50) /* BDW+ */
256 #define OP_3DSTATE_SBE_SWIZ OP_3D_MEDIA(0x3, 0x0, 0x51) /* BDW+ */
257 #define OP_3DSTATE_WM_HZ_OP OP_3D_MEDIA(0x3, 0x0, 0x52) /* BDW+ */
258 #define OP_3DSTATE_COMPONENT_PACKING OP_3D_MEDIA(0x3, 0x0, 0x55) /* SKL+ */
260 #define OP_3DSTATE_DRAWING_RECTANGLE OP_3D_MEDIA(0x3, 0x1, 0x00)
261 #define OP_3DSTATE_SAMPLER_PALETTE_LOAD0 OP_3D_MEDIA(0x3, 0x1, 0x02)
262 #define OP_3DSTATE_CHROMA_KEY OP_3D_MEDIA(0x3, 0x1, 0x04)
263 #define OP_SNB_3DSTATE_DEPTH_BUFFER OP_3D_MEDIA(0x3, 0x1, 0x05)
264 #define OP_3DSTATE_POLY_STIPPLE_OFFSET OP_3D_MEDIA(0x3, 0x1, 0x06)
265 #define OP_3DSTATE_POLY_STIPPLE_PATTERN OP_3D_MEDIA(0x3, 0x1, 0x07)
266 #define OP_3DSTATE_LINE_STIPPLE OP_3D_MEDIA(0x3, 0x1, 0x08)
267 #define OP_3DSTATE_AA_LINE_PARAMS OP_3D_MEDIA(0x3, 0x1, 0x0A)
268 #define OP_3DSTATE_GS_SVB_INDEX OP_3D_MEDIA(0x3, 0x1, 0x0B)
269 #define OP_3DSTATE_SAMPLER_PALETTE_LOAD1 OP_3D_MEDIA(0x3, 0x1, 0x0C)
270 #define OP_3DSTATE_MULTISAMPLE_BDW OP_3D_MEDIA(0x3, 0x0, 0x0D)
271 #define OP_SNB_3DSTATE_STENCIL_BUFFER OP_3D_MEDIA(0x3, 0x1, 0x0E)
272 #define OP_SNB_3DSTATE_HIER_DEPTH_BUFFER OP_3D_MEDIA(0x3, 0x1, 0x0F)
273 #define OP_SNB_3DSTATE_CLEAR_PARAMS OP_3D_MEDIA(0x3, 0x1, 0x10)
274 #define OP_3DSTATE_MONOFILTER_SIZE OP_3D_MEDIA(0x3, 0x1, 0x11)
275 #define OP_3DSTATE_PUSH_CONSTANT_ALLOC_VS OP_3D_MEDIA(0x3, 0x1, 0x12) /* IVB+ */
276 #define OP_3DSTATE_PUSH_CONSTANT_ALLOC_HS OP_3D_MEDIA(0x3, 0x1, 0x13) /* IVB+ */
277 #define OP_3DSTATE_PUSH_CONSTANT_ALLOC_DS OP_3D_MEDIA(0x3, 0x1, 0x14) /* IVB+ */
278 #define OP_3DSTATE_PUSH_CONSTANT_ALLOC_GS OP_3D_MEDIA(0x3, 0x1, 0x15) /* IVB+ */
279 #define OP_3DSTATE_PUSH_CONSTANT_ALLOC_PS OP_3D_MEDIA(0x3, 0x1, 0x16) /* IVB+ */
280 #define OP_3DSTATE_SO_DECL_LIST OP_3D_MEDIA(0x3, 0x1, 0x17)
281 #define OP_3DSTATE_SO_BUFFER OP_3D_MEDIA(0x3, 0x1, 0x18)
282 #define OP_3DSTATE_BINDING_TABLE_POOL_ALLOC OP_3D_MEDIA(0x3, 0x1, 0x19) /* HSW+ */
283 #define OP_3DSTATE_GATHER_POOL_ALLOC OP_3D_MEDIA(0x3, 0x1, 0x1A) /* HSW+ */
284 #define OP_3DSTATE_DX9_CONSTANT_BUFFER_POOL_ALLOC OP_3D_MEDIA(0x3, 0x1, 0x1B) /* HSW+ */
285 #define OP_3DSTATE_SAMPLE_PATTERN OP_3D_MEDIA(0x3, 0x1, 0x1C)
286 #define OP_PIPE_CONTROL OP_3D_MEDIA(0x3, 0x2, 0x00)
287 #define OP_3DPRIMITIVE OP_3D_MEDIA(0x3, 0x3, 0x00)
289 /* VCCP Command Parser */
292 * Below MFX and VBE cmd definition is from vaapi intel driver project (BSD License)
293 * git://anongit.freedesktop.org/vaapi/intel-driver
298 #define OP_MFX(pipeline, op, sub_opa, sub_opb) \
305 #define OP_MFX_PIPE_MODE_SELECT OP_MFX(2, 0, 0, 0) /* ALL */
306 #define OP_MFX_SURFACE_STATE OP_MFX(2, 0, 0, 1) /* ALL */
307 #define OP_MFX_PIPE_BUF_ADDR_STATE OP_MFX(2, 0, 0, 2) /* ALL */
308 #define OP_MFX_IND_OBJ_BASE_ADDR_STATE OP_MFX(2, 0, 0, 3) /* ALL */
309 #define OP_MFX_BSP_BUF_BASE_ADDR_STATE OP_MFX(2, 0, 0, 4) /* ALL */
310 #define OP_2_0_0_5 OP_MFX(2, 0, 0, 5) /* ALL */
311 #define OP_MFX_STATE_POINTER OP_MFX(2, 0, 0, 6) /* ALL */
312 #define OP_MFX_QM_STATE OP_MFX(2, 0, 0, 7) /* IVB+ */
313 #define OP_MFX_FQM_STATE OP_MFX(2, 0, 0, 8) /* IVB+ */
314 #define OP_MFX_PAK_INSERT_OBJECT OP_MFX(2, 0, 2, 8) /* IVB+ */
315 #define OP_MFX_STITCH_OBJECT OP_MFX(2, 0, 2, 0xA) /* IVB+ */
317 #define OP_MFD_IT_OBJECT OP_MFX(2, 0, 1, 9) /* ALL */
319 #define OP_MFX_WAIT OP_MFX(1, 0, 0, 0) /* IVB+ */
320 #define OP_MFX_AVC_IMG_STATE OP_MFX(2, 1, 0, 0) /* ALL */
321 #define OP_MFX_AVC_QM_STATE OP_MFX(2, 1, 0, 1) /* ALL */
322 #define OP_MFX_AVC_DIRECTMODE_STATE OP_MFX(2, 1, 0, 2) /* ALL */
323 #define OP_MFX_AVC_SLICE_STATE OP_MFX(2, 1, 0, 3) /* ALL */
324 #define OP_MFX_AVC_REF_IDX_STATE OP_MFX(2, 1, 0, 4) /* ALL */
325 #define OP_MFX_AVC_WEIGHTOFFSET_STATE OP_MFX(2, 1, 0, 5) /* ALL */
326 #define OP_MFD_AVC_PICID_STATE OP_MFX(2, 1, 1, 5) /* HSW+ */
327 #define OP_MFD_AVC_DPB_STATE OP_MFX(2, 1, 1, 6) /* IVB+ */
328 #define OP_MFD_AVC_SLICEADDR OP_MFX(2, 1, 1, 7) /* IVB+ */
329 #define OP_MFD_AVC_BSD_OBJECT OP_MFX(2, 1, 1, 8) /* ALL */
330 #define OP_MFC_AVC_PAK_OBJECT OP_MFX(2, 1, 2, 9) /* ALL */
332 #define OP_MFX_VC1_PRED_PIPE_STATE OP_MFX(2, 2, 0, 1) /* ALL */
333 #define OP_MFX_VC1_DIRECTMODE_STATE OP_MFX(2, 2, 0, 2) /* ALL */
334 #define OP_MFD_VC1_SHORT_PIC_STATE OP_MFX(2, 2, 1, 0) /* IVB+ */
335 #define OP_MFD_VC1_LONG_PIC_STATE OP_MFX(2, 2, 1, 1) /* IVB+ */
336 #define OP_MFD_VC1_BSD_OBJECT OP_MFX(2, 2, 1, 8) /* ALL */
338 #define OP_MFX_MPEG2_PIC_STATE OP_MFX(2, 3, 0, 0) /* ALL */
339 #define OP_MFX_MPEG2_QM_STATE OP_MFX(2, 3, 0, 1) /* ALL */
340 #define OP_MFD_MPEG2_BSD_OBJECT OP_MFX(2, 3, 1, 8) /* ALL */
341 #define OP_MFC_MPEG2_SLICEGROUP_STATE OP_MFX(2, 3, 2, 3) /* ALL */
342 #define OP_MFC_MPEG2_PAK_OBJECT OP_MFX(2, 3, 2, 9) /* ALL */
344 #define OP_MFX_2_6_0_0 OP_MFX(2, 6, 0, 0) /* IVB+ */
345 #define OP_MFX_2_6_0_8 OP_MFX(2, 6, 0, 8) /* IVB+ */
346 #define OP_MFX_2_6_0_9 OP_MFX(2, 6, 0, 9) /* IVB+ */
348 #define OP_MFX_JPEG_PIC_STATE OP_MFX(2, 7, 0, 0)
349 #define OP_MFX_JPEG_HUFF_TABLE_STATE OP_MFX(2, 7, 0, 2)
350 #define OP_MFD_JPEG_BSD_OBJECT OP_MFX(2, 7, 1, 8)
352 #define OP_VEB(pipeline, op, sub_opa, sub_opb) \
359 #define OP_VEB_SURFACE_STATE OP_VEB(2, 4, 0, 0)
360 #define OP_VEB_STATE OP_VEB(2, 4, 0, 2)
361 #define OP_VEB_DNDI_IECP_STATE OP_VEB(2, 4, 0, 3)
363 struct parser_exec_state;
365 typedef int (*parser_cmd_handler)(struct parser_exec_state *s);
367 #define GVT_CMD_HASH_BITS 7
369 /* which DWords need address fix */
370 #define ADDR_FIX_1(x1) (1 << (x1))
371 #define ADDR_FIX_2(x1, x2) (ADDR_FIX_1(x1) | ADDR_FIX_1(x2))
372 #define ADDR_FIX_3(x1, x2, x3) (ADDR_FIX_1(x1) | ADDR_FIX_2(x2, x3))
373 #define ADDR_FIX_4(x1, x2, x3, x4) (ADDR_FIX_1(x1) | ADDR_FIX_3(x2, x3, x4))
374 #define ADDR_FIX_5(x1, x2, x3, x4, x5) (ADDR_FIX_1(x1) | ADDR_FIX_4(x2, x3, x4, x5))
380 #define F_LEN_MASK (1U<<0)
381 #define F_LEN_CONST 1U
385 * command has its own ip advance logic
386 * e.g. MI_BATCH_START, MI_BATCH_END
388 #define F_IP_ADVANCE_CUSTOM (1<<1)
390 #define F_POST_HANDLE (1<<2)
393 #define R_RCS (1 << RCS)
394 #define R_VCS1 (1 << VCS)
395 #define R_VCS2 (1 << VCS2)
396 #define R_VCS (R_VCS1 | R_VCS2)
397 #define R_BCS (1 << BCS)
398 #define R_VECS (1 << VECS)
399 #define R_ALL (R_RCS | R_VCS | R_BCS | R_VECS)
400 /* rings that support this cmd: BLT/RCS/VCS/VECS */
403 /* devices that support this cmd: SNB/IVB/HSW/... */
406 /* which DWords are address that need fix up.
407 * bit 0 means a 32-bit non address operand in command
408 * bit 1 means address operand, which could be 32-bit
409 * or 64-bit depending on different architectures.(
410 * defined by "gmadr_bytes_in_cmd" in intel_gvt.
411 * No matter the address length, each address only takes
412 * one bit in the bitmap.
414 uint16_t addr_bitmap;
416 /* flag == F_LEN_CONST : command length
417 * flag == F_LEN_VAR : length bias bits
418 * Note: length is in DWord
422 parser_cmd_handler handler;
426 struct hlist_node hlist;
427 struct cmd_info *info;
431 RING_BUFFER_INSTRUCTION,
432 BATCH_BUFFER_INSTRUCTION,
433 BATCH_BUFFER_2ND_LEVEL,
441 struct parser_exec_state {
442 struct intel_vgpu *vgpu;
447 /* batch buffer address type */
450 /* graphics memory address of ring buffer start */
451 unsigned long ring_start;
452 unsigned long ring_size;
453 unsigned long ring_head;
454 unsigned long ring_tail;
456 /* instruction graphics memory address */
457 unsigned long ip_gma;
459 /* mapped va of the instr_gma */
464 /* next instruction when return from batch buffer to ring buffer */
465 unsigned long ret_ip_gma_ring;
467 /* next instruction when return from 2nd batch buffer to batch buffer */
468 unsigned long ret_ip_gma_bb;
470 /* batch buffer address type (GTT or PPGTT)
471 * used when ret from 2nd level batch buffer
473 int saved_buf_addr_type;
476 struct cmd_info *info;
478 struct intel_vgpu_workload *workload;
481 #define gmadr_dw_number(s) \
482 (s->vgpu->gvt->device_info.gmadr_bytes_in_cmd >> 2)
484 static unsigned long bypass_scan_mask = 0;
486 /* ring ALL, type = 0 */
487 static struct sub_op_bits sub_op_mi[] = {
492 static struct decode_info decode_info_mi = {
495 ARRAY_SIZE(sub_op_mi),
499 /* ring RCS, command type 2 */
500 static struct sub_op_bits sub_op_2d[] = {
505 static struct decode_info decode_info_2d = {
508 ARRAY_SIZE(sub_op_2d),
512 /* ring RCS, command type 3 */
513 static struct sub_op_bits sub_op_3d_media[] = {
520 static struct decode_info decode_info_3d_media = {
523 ARRAY_SIZE(sub_op_3d_media),
527 /* ring VCS, command type 3 */
528 static struct sub_op_bits sub_op_mfx_vc[] = {
536 static struct decode_info decode_info_mfx_vc = {
539 ARRAY_SIZE(sub_op_mfx_vc),
543 /* ring VECS, command type 3 */
544 static struct sub_op_bits sub_op_vebox[] = {
552 static struct decode_info decode_info_vebox = {
555 ARRAY_SIZE(sub_op_vebox),
559 static struct decode_info *ring_decode_info[I915_NUM_ENGINES][8] = {
564 &decode_info_3d_media,
616 static inline u32 get_opcode(u32 cmd, int ring_id)
618 struct decode_info *d_info;
620 d_info = ring_decode_info[ring_id][CMD_TYPE(cmd)];
624 return cmd >> (32 - d_info->op_len);
627 static inline struct cmd_info *find_cmd_entry(struct intel_gvt *gvt,
628 unsigned int opcode, int ring_id)
632 hash_for_each_possible(gvt->cmd_table, e, hlist, opcode) {
633 if ((opcode == e->info->opcode) &&
634 (e->info->rings & (1 << ring_id)))
640 static inline struct cmd_info *get_cmd_info(struct intel_gvt *gvt,
641 u32 cmd, int ring_id)
645 opcode = get_opcode(cmd, ring_id);
646 if (opcode == INVALID_OP)
649 return find_cmd_entry(gvt, opcode, ring_id);
652 static inline u32 sub_op_val(u32 cmd, u32 hi, u32 low)
654 return (cmd >> low) & ((1U << (hi - low + 1)) - 1);
657 static inline void print_opcode(u32 cmd, int ring_id)
659 struct decode_info *d_info;
662 d_info = ring_decode_info[ring_id][CMD_TYPE(cmd)];
666 gvt_dbg_cmd("opcode=0x%x %s sub_ops:",
667 cmd >> (32 - d_info->op_len), d_info->name);
669 for (i = 0; i < d_info->nr_sub_op; i++)
670 pr_err("0x%x ", sub_op_val(cmd, d_info->sub_op[i].hi,
671 d_info->sub_op[i].low));
676 static inline u32 *cmd_ptr(struct parser_exec_state *s, int index)
678 return s->ip_va + (index << 2);
681 static inline u32 cmd_val(struct parser_exec_state *s, int index)
683 return *cmd_ptr(s, index);
686 static void parser_exec_state_dump(struct parser_exec_state *s)
691 gvt_dbg_cmd(" vgpu%d RING%d: ring_start(%08lx) ring_end(%08lx)"
692 " ring_head(%08lx) ring_tail(%08lx)\n", s->vgpu->id,
693 s->ring_id, s->ring_start, s->ring_start + s->ring_size,
694 s->ring_head, s->ring_tail);
696 gvt_dbg_cmd(" %s %s ip_gma(%08lx) ",
697 s->buf_type == RING_BUFFER_INSTRUCTION ?
698 "RING_BUFFER" : "BATCH_BUFFER",
699 s->buf_addr_type == GTT_BUFFER ?
700 "GTT" : "PPGTT", s->ip_gma);
702 if (s->ip_va == NULL) {
703 gvt_dbg_cmd(" ip_va(NULL)");
707 gvt_dbg_cmd(" ip_va=%p: %08x %08x %08x %08x\n",
708 s->ip_va, cmd_val(s, 0), cmd_val(s, 1),
709 cmd_val(s, 2), cmd_val(s, 3));
711 print_opcode(cmd_val(s, 0), s->ring_id);
713 s->ip_va = (u32 *)((((u64)s->ip_va) >> 12) << 12);
716 gvt_dbg_cmd("ip_va=%p: ", s->ip_va);
717 for (i = 0; i < 8; i++)
718 gvt_dbg_cmd("%08x ", cmd_val(s, i));
721 s->ip_va += 8 * sizeof(u32);
726 static inline void update_ip_va(struct parser_exec_state *s)
728 unsigned long len = 0;
730 if (WARN_ON(s->ring_head == s->ring_tail))
733 if (s->buf_type == RING_BUFFER_INSTRUCTION) {
734 unsigned long ring_top = s->ring_start + s->ring_size;
736 if (s->ring_head > s->ring_tail) {
737 if (s->ip_gma >= s->ring_head && s->ip_gma < ring_top)
738 len = (s->ip_gma - s->ring_head);
739 else if (s->ip_gma >= s->ring_start &&
740 s->ip_gma <= s->ring_tail)
741 len = (ring_top - s->ring_head) +
742 (s->ip_gma - s->ring_start);
744 len = (s->ip_gma - s->ring_head);
746 s->ip_va = s->rb_va + len;
747 } else {/* shadow batch buffer */
748 s->ip_va = s->ret_bb_va;
752 static inline int ip_gma_set(struct parser_exec_state *s,
753 unsigned long ip_gma)
755 WARN_ON(!IS_ALIGNED(ip_gma, 4));
762 static inline int ip_gma_advance(struct parser_exec_state *s,
765 s->ip_gma += (dw_len << 2);
767 if (s->buf_type == RING_BUFFER_INSTRUCTION) {
768 if (s->ip_gma >= s->ring_start + s->ring_size)
769 s->ip_gma -= s->ring_size;
772 s->ip_va += (dw_len << 2);
778 static inline int get_cmd_length(struct cmd_info *info, u32 cmd)
780 if ((info->flag & F_LEN_MASK) == F_LEN_CONST)
783 return (cmd & ((1U << info->len) - 1)) + 2;
787 static inline int cmd_length(struct parser_exec_state *s)
789 return get_cmd_length(s->info, cmd_val(s, 0));
792 /* do not remove this, some platform may need clflush here */
793 #define patch_value(s, addr, val) do { \
797 static bool is_shadowed_mmio(unsigned int offset)
801 if ((offset == 0x2168) || /*BB current head register UDW */
802 (offset == 0x2140) || /*BB current header register */
803 (offset == 0x211c) || /*second BB header register UDW */
804 (offset == 0x2114)) { /*second BB header register UDW */
810 static inline bool is_force_nonpriv_mmio(unsigned int offset)
812 return (offset >= 0x24d0 && offset < 0x2500);
815 static int force_nonpriv_reg_handler(struct parser_exec_state *s,
816 unsigned int offset, unsigned int index)
818 struct intel_gvt *gvt = s->vgpu->gvt;
819 unsigned int data = cmd_val(s, index + 1);
821 if (!intel_gvt_in_force_nonpriv_whitelist(gvt, data)) {
822 gvt_err("Unexpected forcenonpriv 0x%x LRI write, value=0x%x\n",
829 static inline bool is_mocs_mmio(unsigned int offset)
831 return ((offset >= 0xc800) && (offset <= 0xcff8)) ||
832 ((offset >= 0xb020) && (offset <= 0xb0a0));
835 static int mocs_cmd_reg_handler(struct parser_exec_state *s,
836 unsigned int offset, unsigned int index)
838 if (!is_mocs_mmio(offset))
840 vgpu_vreg(s->vgpu, offset) = cmd_val(s, index + 1);
844 static int cmd_reg_handler(struct parser_exec_state *s,
845 unsigned int offset, unsigned int index, char *cmd)
847 struct intel_vgpu *vgpu = s->vgpu;
848 struct intel_gvt *gvt = vgpu->gvt;
850 if (offset + 4 > gvt->device_info.mmio_size) {
851 gvt_vgpu_err("%s access to (%x) outside of MMIO range\n",
856 if (!intel_gvt_mmio_is_cmd_access(gvt, offset)) {
857 gvt_vgpu_err("%s access to non-render register (%x)\n",
862 if (is_shadowed_mmio(offset)) {
863 gvt_vgpu_err("found access of shadowed MMIO %x\n", offset);
867 if (is_mocs_mmio(offset) &&
868 mocs_cmd_reg_handler(s, offset, index))
871 if (is_force_nonpriv_mmio(offset) &&
872 force_nonpriv_reg_handler(s, offset, index))
875 if (offset == i915_mmio_reg_offset(DERRMR) ||
876 offset == i915_mmio_reg_offset(FORCEWAKE_MT)) {
877 /* Writing to HW VGT_PVINFO_PAGE offset will be discarded */
878 patch_value(s, cmd_ptr(s, index), VGT_PVINFO_PAGE);
881 /* TODO: Update the global mask if this MMIO is a masked-MMIO */
882 intel_gvt_mmio_set_cmd_accessed(gvt, offset);
886 #define cmd_reg(s, i) \
887 (cmd_val(s, i) & GENMASK(22, 2))
889 #define cmd_reg_inhibit(s, i) \
890 (cmd_val(s, i) & GENMASK(22, 18))
892 #define cmd_gma(s, i) \
893 (cmd_val(s, i) & GENMASK(31, 2))
895 #define cmd_gma_hi(s, i) \
896 (cmd_val(s, i) & GENMASK(15, 0))
898 static int cmd_handler_lri(struct parser_exec_state *s)
901 int cmd_len = cmd_length(s);
902 struct intel_gvt *gvt = s->vgpu->gvt;
904 for (i = 1; i < cmd_len; i += 2) {
905 if (IS_BROADWELL(gvt->dev_priv) &&
906 (s->ring_id != RCS)) {
907 if (s->ring_id == BCS &&
909 i915_mmio_reg_offset(DERRMR))
912 ret |= (cmd_reg_inhibit(s, i)) ?
917 ret |= cmd_reg_handler(s, cmd_reg(s, i), i, "lri");
924 static int cmd_handler_lrr(struct parser_exec_state *s)
927 int cmd_len = cmd_length(s);
929 for (i = 1; i < cmd_len; i += 2) {
930 if (IS_BROADWELL(s->vgpu->gvt->dev_priv))
931 ret |= ((cmd_reg_inhibit(s, i) ||
932 (cmd_reg_inhibit(s, i + 1)))) ?
936 ret |= cmd_reg_handler(s, cmd_reg(s, i), i, "lrr-src");
939 ret |= cmd_reg_handler(s, cmd_reg(s, i + 1), i, "lrr-dst");
946 static inline int cmd_address_audit(struct parser_exec_state *s,
947 unsigned long guest_gma, int op_size, bool index_mode);
949 static int cmd_handler_lrm(struct parser_exec_state *s)
951 struct intel_gvt *gvt = s->vgpu->gvt;
952 int gmadr_bytes = gvt->device_info.gmadr_bytes_in_cmd;
955 int cmd_len = cmd_length(s);
957 for (i = 1; i < cmd_len;) {
958 if (IS_BROADWELL(gvt->dev_priv))
959 ret |= (cmd_reg_inhibit(s, i)) ? -EBADRQC : 0;
962 ret |= cmd_reg_handler(s, cmd_reg(s, i), i, "lrm");
965 if (cmd_val(s, 0) & (1 << 22)) {
966 gma = cmd_gma(s, i + 1);
967 if (gmadr_bytes == 8)
968 gma |= (cmd_gma_hi(s, i + 2)) << 32;
969 ret |= cmd_address_audit(s, gma, sizeof(u32), false);
973 i += gmadr_dw_number(s) + 1;
978 static int cmd_handler_srm(struct parser_exec_state *s)
980 int gmadr_bytes = s->vgpu->gvt->device_info.gmadr_bytes_in_cmd;
983 int cmd_len = cmd_length(s);
985 for (i = 1; i < cmd_len;) {
986 ret |= cmd_reg_handler(s, cmd_reg(s, i), i, "srm");
989 if (cmd_val(s, 0) & (1 << 22)) {
990 gma = cmd_gma(s, i + 1);
991 if (gmadr_bytes == 8)
992 gma |= (cmd_gma_hi(s, i + 2)) << 32;
993 ret |= cmd_address_audit(s, gma, sizeof(u32), false);
997 i += gmadr_dw_number(s) + 1;
1002 struct cmd_interrupt_event {
1003 int pipe_control_notify;
1005 int mi_user_interrupt;
1008 static struct cmd_interrupt_event cmd_interrupt_events[] = {
1010 .pipe_control_notify = RCS_PIPE_CONTROL,
1011 .mi_flush_dw = INTEL_GVT_EVENT_RESERVED,
1012 .mi_user_interrupt = RCS_MI_USER_INTERRUPT,
1015 .pipe_control_notify = INTEL_GVT_EVENT_RESERVED,
1016 .mi_flush_dw = BCS_MI_FLUSH_DW,
1017 .mi_user_interrupt = BCS_MI_USER_INTERRUPT,
1020 .pipe_control_notify = INTEL_GVT_EVENT_RESERVED,
1021 .mi_flush_dw = VCS_MI_FLUSH_DW,
1022 .mi_user_interrupt = VCS_MI_USER_INTERRUPT,
1025 .pipe_control_notify = INTEL_GVT_EVENT_RESERVED,
1026 .mi_flush_dw = VCS2_MI_FLUSH_DW,
1027 .mi_user_interrupt = VCS2_MI_USER_INTERRUPT,
1030 .pipe_control_notify = INTEL_GVT_EVENT_RESERVED,
1031 .mi_flush_dw = VECS_MI_FLUSH_DW,
1032 .mi_user_interrupt = VECS_MI_USER_INTERRUPT,
1036 static int cmd_handler_pipe_control(struct parser_exec_state *s)
1038 int gmadr_bytes = s->vgpu->gvt->device_info.gmadr_bytes_in_cmd;
1040 bool index_mode = false;
1041 unsigned int post_sync;
1044 post_sync = (cmd_val(s, 1) & PIPE_CONTROL_POST_SYNC_OP_MASK) >> 14;
1047 if (cmd_val(s, 1) & PIPE_CONTROL_MMIO_WRITE)
1048 ret = cmd_reg_handler(s, cmd_reg(s, 2), 1, "pipe_ctrl");
1050 else if (post_sync) {
1052 ret = cmd_reg_handler(s, 0x2350, 1, "pipe_ctrl");
1053 else if (post_sync == 3)
1054 ret = cmd_reg_handler(s, 0x2358, 1, "pipe_ctrl");
1055 else if (post_sync == 1) {
1057 if ((cmd_val(s, 1) & PIPE_CONTROL_GLOBAL_GTT_IVB)) {
1058 gma = cmd_val(s, 2) & GENMASK(31, 3);
1059 if (gmadr_bytes == 8)
1060 gma |= (cmd_gma_hi(s, 3)) << 32;
1061 /* Store Data Index */
1062 if (cmd_val(s, 1) & (1 << 21))
1064 ret |= cmd_address_audit(s, gma, sizeof(u64),
1073 if (cmd_val(s, 1) & PIPE_CONTROL_NOTIFY)
1074 set_bit(cmd_interrupt_events[s->ring_id].pipe_control_notify,
1075 s->workload->pending_events);
1079 static int cmd_handler_mi_user_interrupt(struct parser_exec_state *s)
1081 set_bit(cmd_interrupt_events[s->ring_id].mi_user_interrupt,
1082 s->workload->pending_events);
1083 patch_value(s, cmd_ptr(s, 0), MI_NOOP);
1087 static int cmd_advance_default(struct parser_exec_state *s)
1089 return ip_gma_advance(s, cmd_length(s));
1092 static int cmd_handler_mi_batch_buffer_end(struct parser_exec_state *s)
1096 if (s->buf_type == BATCH_BUFFER_2ND_LEVEL) {
1097 s->buf_type = BATCH_BUFFER_INSTRUCTION;
1098 ret = ip_gma_set(s, s->ret_ip_gma_bb);
1099 s->buf_addr_type = s->saved_buf_addr_type;
1101 s->buf_type = RING_BUFFER_INSTRUCTION;
1102 s->buf_addr_type = GTT_BUFFER;
1103 if (s->ret_ip_gma_ring >= s->ring_start + s->ring_size)
1104 s->ret_ip_gma_ring -= s->ring_size;
1105 ret = ip_gma_set(s, s->ret_ip_gma_ring);
1110 struct mi_display_flip_command_info {
1114 i915_reg_t stride_reg;
1115 i915_reg_t ctrl_reg;
1116 i915_reg_t surf_reg;
1123 struct plane_code_mapping {
1129 static int gen8_decode_mi_display_flip(struct parser_exec_state *s,
1130 struct mi_display_flip_command_info *info)
1132 struct drm_i915_private *dev_priv = s->vgpu->gvt->dev_priv;
1133 struct plane_code_mapping gen8_plane_code[] = {
1134 [0] = {PIPE_A, PLANE_A, PRIMARY_A_FLIP_DONE},
1135 [1] = {PIPE_B, PLANE_A, PRIMARY_B_FLIP_DONE},
1136 [2] = {PIPE_A, PLANE_B, SPRITE_A_FLIP_DONE},
1137 [3] = {PIPE_B, PLANE_B, SPRITE_B_FLIP_DONE},
1138 [4] = {PIPE_C, PLANE_A, PRIMARY_C_FLIP_DONE},
1139 [5] = {PIPE_C, PLANE_B, SPRITE_C_FLIP_DONE},
1141 u32 dword0, dword1, dword2;
1144 dword0 = cmd_val(s, 0);
1145 dword1 = cmd_val(s, 1);
1146 dword2 = cmd_val(s, 2);
1148 v = (dword0 & GENMASK(21, 19)) >> 19;
1149 if (WARN_ON(v >= ARRAY_SIZE(gen8_plane_code)))
1152 info->pipe = gen8_plane_code[v].pipe;
1153 info->plane = gen8_plane_code[v].plane;
1154 info->event = gen8_plane_code[v].event;
1155 info->stride_val = (dword1 & GENMASK(15, 6)) >> 6;
1156 info->tile_val = (dword1 & 0x1);
1157 info->surf_val = (dword2 & GENMASK(31, 12)) >> 12;
1158 info->async_flip = ((dword2 & GENMASK(1, 0)) == 0x1);
1160 if (info->plane == PLANE_A) {
1161 info->ctrl_reg = DSPCNTR(info->pipe);
1162 info->stride_reg = DSPSTRIDE(info->pipe);
1163 info->surf_reg = DSPSURF(info->pipe);
1164 } else if (info->plane == PLANE_B) {
1165 info->ctrl_reg = SPRCTL(info->pipe);
1166 info->stride_reg = SPRSTRIDE(info->pipe);
1167 info->surf_reg = SPRSURF(info->pipe);
1175 static int skl_decode_mi_display_flip(struct parser_exec_state *s,
1176 struct mi_display_flip_command_info *info)
1178 struct drm_i915_private *dev_priv = s->vgpu->gvt->dev_priv;
1179 struct intel_vgpu *vgpu = s->vgpu;
1180 u32 dword0 = cmd_val(s, 0);
1181 u32 dword1 = cmd_val(s, 1);
1182 u32 dword2 = cmd_val(s, 2);
1183 u32 plane = (dword0 & GENMASK(12, 8)) >> 8;
1185 info->plane = PRIMARY_PLANE;
1188 case MI_DISPLAY_FLIP_SKL_PLANE_1_A:
1189 info->pipe = PIPE_A;
1190 info->event = PRIMARY_A_FLIP_DONE;
1192 case MI_DISPLAY_FLIP_SKL_PLANE_1_B:
1193 info->pipe = PIPE_B;
1194 info->event = PRIMARY_B_FLIP_DONE;
1196 case MI_DISPLAY_FLIP_SKL_PLANE_1_C:
1197 info->pipe = PIPE_C;
1198 info->event = PRIMARY_C_FLIP_DONE;
1201 case MI_DISPLAY_FLIP_SKL_PLANE_2_A:
1202 info->pipe = PIPE_A;
1203 info->event = SPRITE_A_FLIP_DONE;
1204 info->plane = SPRITE_PLANE;
1206 case MI_DISPLAY_FLIP_SKL_PLANE_2_B:
1207 info->pipe = PIPE_B;
1208 info->event = SPRITE_B_FLIP_DONE;
1209 info->plane = SPRITE_PLANE;
1211 case MI_DISPLAY_FLIP_SKL_PLANE_2_C:
1212 info->pipe = PIPE_C;
1213 info->event = SPRITE_C_FLIP_DONE;
1214 info->plane = SPRITE_PLANE;
1218 gvt_vgpu_err("unknown plane code %d\n", plane);
1222 info->stride_val = (dword1 & GENMASK(15, 6)) >> 6;
1223 info->tile_val = (dword1 & GENMASK(2, 0));
1224 info->surf_val = (dword2 & GENMASK(31, 12)) >> 12;
1225 info->async_flip = ((dword2 & GENMASK(1, 0)) == 0x1);
1227 info->ctrl_reg = DSPCNTR(info->pipe);
1228 info->stride_reg = DSPSTRIDE(info->pipe);
1229 info->surf_reg = DSPSURF(info->pipe);
1234 static int gen8_check_mi_display_flip(struct parser_exec_state *s,
1235 struct mi_display_flip_command_info *info)
1237 struct drm_i915_private *dev_priv = s->vgpu->gvt->dev_priv;
1240 if (!info->async_flip)
1243 if (IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv)) {
1244 stride = vgpu_vreg_t(s->vgpu, info->stride_reg) & GENMASK(9, 0);
1245 tile = (vgpu_vreg_t(s->vgpu, info->ctrl_reg) &
1246 GENMASK(12, 10)) >> 10;
1248 stride = (vgpu_vreg_t(s->vgpu, info->stride_reg) &
1249 GENMASK(15, 6)) >> 6;
1250 tile = (vgpu_vreg_t(s->vgpu, info->ctrl_reg) & (1 << 10)) >> 10;
1253 if (stride != info->stride_val)
1254 gvt_dbg_cmd("cannot change stride during async flip\n");
1256 if (tile != info->tile_val)
1257 gvt_dbg_cmd("cannot change tile during async flip\n");
1262 static int gen8_update_plane_mmio_from_mi_display_flip(
1263 struct parser_exec_state *s,
1264 struct mi_display_flip_command_info *info)
1266 struct drm_i915_private *dev_priv = s->vgpu->gvt->dev_priv;
1267 struct intel_vgpu *vgpu = s->vgpu;
1269 set_mask_bits(&vgpu_vreg_t(vgpu, info->surf_reg), GENMASK(31, 12),
1270 info->surf_val << 12);
1271 if (IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv)) {
1272 set_mask_bits(&vgpu_vreg_t(vgpu, info->stride_reg), GENMASK(9, 0),
1274 set_mask_bits(&vgpu_vreg_t(vgpu, info->ctrl_reg), GENMASK(12, 10),
1275 info->tile_val << 10);
1277 set_mask_bits(&vgpu_vreg_t(vgpu, info->stride_reg), GENMASK(15, 6),
1278 info->stride_val << 6);
1279 set_mask_bits(&vgpu_vreg_t(vgpu, info->ctrl_reg), GENMASK(10, 10),
1280 info->tile_val << 10);
1283 vgpu_vreg_t(vgpu, PIPE_FRMCOUNT_G4X(info->pipe))++;
1284 intel_vgpu_trigger_virtual_event(vgpu, info->event);
1288 static int decode_mi_display_flip(struct parser_exec_state *s,
1289 struct mi_display_flip_command_info *info)
1291 struct drm_i915_private *dev_priv = s->vgpu->gvt->dev_priv;
1293 if (IS_BROADWELL(dev_priv))
1294 return gen8_decode_mi_display_flip(s, info);
1295 if (IS_SKYLAKE(dev_priv) || IS_KABYLAKE(dev_priv))
1296 return skl_decode_mi_display_flip(s, info);
1301 static int check_mi_display_flip(struct parser_exec_state *s,
1302 struct mi_display_flip_command_info *info)
1304 struct drm_i915_private *dev_priv = s->vgpu->gvt->dev_priv;
1306 if (IS_BROADWELL(dev_priv)
1307 || IS_SKYLAKE(dev_priv)
1308 || IS_KABYLAKE(dev_priv))
1309 return gen8_check_mi_display_flip(s, info);
1313 static int update_plane_mmio_from_mi_display_flip(
1314 struct parser_exec_state *s,
1315 struct mi_display_flip_command_info *info)
1317 struct drm_i915_private *dev_priv = s->vgpu->gvt->dev_priv;
1319 if (IS_BROADWELL(dev_priv)
1320 || IS_SKYLAKE(dev_priv)
1321 || IS_KABYLAKE(dev_priv))
1322 return gen8_update_plane_mmio_from_mi_display_flip(s, info);
1326 static int cmd_handler_mi_display_flip(struct parser_exec_state *s)
1328 struct mi_display_flip_command_info info;
1329 struct intel_vgpu *vgpu = s->vgpu;
1332 int len = cmd_length(s);
1334 ret = decode_mi_display_flip(s, &info);
1336 gvt_vgpu_err("fail to decode MI display flip command\n");
1340 ret = check_mi_display_flip(s, &info);
1342 gvt_vgpu_err("invalid MI display flip command\n");
1346 ret = update_plane_mmio_from_mi_display_flip(s, &info);
1348 gvt_vgpu_err("fail to update plane mmio\n");
1352 for (i = 0; i < len; i++)
1353 patch_value(s, cmd_ptr(s, i), MI_NOOP);
1357 static bool is_wait_for_flip_pending(u32 cmd)
1359 return cmd & (MI_WAIT_FOR_PLANE_A_FLIP_PENDING |
1360 MI_WAIT_FOR_PLANE_B_FLIP_PENDING |
1361 MI_WAIT_FOR_PLANE_C_FLIP_PENDING |
1362 MI_WAIT_FOR_SPRITE_A_FLIP_PENDING |
1363 MI_WAIT_FOR_SPRITE_B_FLIP_PENDING |
1364 MI_WAIT_FOR_SPRITE_C_FLIP_PENDING);
1367 static int cmd_handler_mi_wait_for_event(struct parser_exec_state *s)
1369 u32 cmd = cmd_val(s, 0);
1371 if (!is_wait_for_flip_pending(cmd))
1374 patch_value(s, cmd_ptr(s, 0), MI_NOOP);
1378 static unsigned long get_gma_bb_from_cmd(struct parser_exec_state *s, int index)
1381 unsigned long gma_high, gma_low;
1382 struct intel_vgpu *vgpu = s->vgpu;
1383 int gmadr_bytes = vgpu->gvt->device_info.gmadr_bytes_in_cmd;
1385 if (WARN_ON(gmadr_bytes != 4 && gmadr_bytes != 8)) {
1386 gvt_vgpu_err("invalid gma bytes %d\n", gmadr_bytes);
1387 return INTEL_GVT_INVALID_ADDR;
1390 gma_low = cmd_val(s, index) & BATCH_BUFFER_ADDR_MASK;
1391 if (gmadr_bytes == 4) {
1394 gma_high = cmd_val(s, index + 1) & BATCH_BUFFER_ADDR_HIGH_MASK;
1395 addr = (((unsigned long)gma_high) << 32) | gma_low;
1400 static inline int cmd_address_audit(struct parser_exec_state *s,
1401 unsigned long guest_gma, int op_size, bool index_mode)
1403 struct intel_vgpu *vgpu = s->vgpu;
1404 u32 max_surface_size = vgpu->gvt->device_info.max_surface_size;
1408 if (op_size > max_surface_size) {
1409 gvt_vgpu_err("command address audit fail name %s\n",
1415 if (guest_gma >= I915_GTT_PAGE_SIZE / sizeof(u64)) {
1419 } else if (!intel_gvt_ggtt_validate_range(vgpu, guest_gma, op_size)) {
1427 gvt_vgpu_err("cmd_parser: Malicious %s detected, addr=0x%lx, len=%d!\n",
1428 s->info->name, guest_gma, op_size);
1430 pr_err("cmd dump: ");
1431 for (i = 0; i < cmd_length(s); i++) {
1433 pr_err("\n%08x ", cmd_val(s, i));
1435 pr_err("%08x ", cmd_val(s, i));
1437 pr_err("\nvgpu%d: aperture 0x%llx - 0x%llx, hidden 0x%llx - 0x%llx\n",
1439 vgpu_aperture_gmadr_base(vgpu),
1440 vgpu_aperture_gmadr_end(vgpu),
1441 vgpu_hidden_gmadr_base(vgpu),
1442 vgpu_hidden_gmadr_end(vgpu));
1446 static int cmd_handler_mi_store_data_imm(struct parser_exec_state *s)
1448 int gmadr_bytes = s->vgpu->gvt->device_info.gmadr_bytes_in_cmd;
1449 int op_size = (cmd_length(s) - 3) * sizeof(u32);
1450 int core_id = (cmd_val(s, 2) & (1 << 0)) ? 1 : 0;
1451 unsigned long gma, gma_low, gma_high;
1455 if (!(cmd_val(s, 0) & (1 << 22)))
1458 gma = cmd_val(s, 2) & GENMASK(31, 2);
1460 if (gmadr_bytes == 8) {
1461 gma_low = cmd_val(s, 1) & GENMASK(31, 2);
1462 gma_high = cmd_val(s, 2) & GENMASK(15, 0);
1463 gma = (gma_high << 32) | gma_low;
1464 core_id = (cmd_val(s, 1) & (1 << 0)) ? 1 : 0;
1466 ret = cmd_address_audit(s, gma + op_size * core_id, op_size, false);
1470 static inline int unexpected_cmd(struct parser_exec_state *s)
1472 struct intel_vgpu *vgpu = s->vgpu;
1474 gvt_vgpu_err("Unexpected %s in command buffer!\n", s->info->name);
1479 static int cmd_handler_mi_semaphore_wait(struct parser_exec_state *s)
1481 return unexpected_cmd(s);
1484 static int cmd_handler_mi_report_perf_count(struct parser_exec_state *s)
1486 return unexpected_cmd(s);
1489 static int cmd_handler_mi_op_2e(struct parser_exec_state *s)
1491 return unexpected_cmd(s);
1494 static int cmd_handler_mi_op_2f(struct parser_exec_state *s)
1496 int gmadr_bytes = s->vgpu->gvt->device_info.gmadr_bytes_in_cmd;
1497 int op_size = (1 << ((cmd_val(s, 0) & GENMASK(20, 19)) >> 19)) *
1499 unsigned long gma, gma_high;
1502 if (!(cmd_val(s, 0) & (1 << 22)))
1505 gma = cmd_val(s, 1) & GENMASK(31, 2);
1506 if (gmadr_bytes == 8) {
1507 gma_high = cmd_val(s, 2) & GENMASK(15, 0);
1508 gma = (gma_high << 32) | gma;
1510 ret = cmd_address_audit(s, gma, op_size, false);
1514 static int cmd_handler_mi_store_data_index(struct parser_exec_state *s)
1516 return unexpected_cmd(s);
1519 static int cmd_handler_mi_clflush(struct parser_exec_state *s)
1521 return unexpected_cmd(s);
1524 static int cmd_handler_mi_conditional_batch_buffer_end(
1525 struct parser_exec_state *s)
1527 return unexpected_cmd(s);
1530 static int cmd_handler_mi_update_gtt(struct parser_exec_state *s)
1532 return unexpected_cmd(s);
1535 static int cmd_handler_mi_flush_dw(struct parser_exec_state *s)
1537 int gmadr_bytes = s->vgpu->gvt->device_info.gmadr_bytes_in_cmd;
1539 bool index_mode = false;
1542 /* Check post-sync and ppgtt bit */
1543 if (((cmd_val(s, 0) >> 14) & 0x3) && (cmd_val(s, 1) & (1 << 2))) {
1544 gma = cmd_val(s, 1) & GENMASK(31, 3);
1545 if (gmadr_bytes == 8)
1546 gma |= (cmd_val(s, 2) & GENMASK(15, 0)) << 32;
1547 /* Store Data Index */
1548 if (cmd_val(s, 0) & (1 << 21))
1550 ret = cmd_address_audit(s, gma, sizeof(u64), index_mode);
1552 /* Check notify bit */
1553 if ((cmd_val(s, 0) & (1 << 8)))
1554 set_bit(cmd_interrupt_events[s->ring_id].mi_flush_dw,
1555 s->workload->pending_events);
1559 static void addr_type_update_snb(struct parser_exec_state *s)
1561 if ((s->buf_type == RING_BUFFER_INSTRUCTION) &&
1562 (BATCH_BUFFER_ADR_SPACE_BIT(cmd_val(s, 0)) == 1)) {
1563 s->buf_addr_type = PPGTT_BUFFER;
1568 static int copy_gma_to_hva(struct intel_vgpu *vgpu, struct intel_vgpu_mm *mm,
1569 unsigned long gma, unsigned long end_gma, void *va)
1571 unsigned long copy_len, offset;
1572 unsigned long len = 0;
1575 while (gma != end_gma) {
1576 gpa = intel_vgpu_gma_to_gpa(mm, gma);
1577 if (gpa == INTEL_GVT_INVALID_ADDR) {
1578 gvt_vgpu_err("invalid gma address: %lx\n", gma);
1582 offset = gma & (I915_GTT_PAGE_SIZE - 1);
1584 copy_len = (end_gma - gma) >= (I915_GTT_PAGE_SIZE - offset) ?
1585 I915_GTT_PAGE_SIZE - offset : end_gma - gma;
1587 intel_gvt_hypervisor_read_gpa(vgpu, gpa, va + len, copy_len);
1597 * Check whether a batch buffer needs to be scanned. Currently
1598 * the only criteria is based on privilege.
1600 static int batch_buffer_needs_scan(struct parser_exec_state *s)
1602 struct intel_gvt *gvt = s->vgpu->gvt;
1604 if (IS_BROADWELL(gvt->dev_priv) || IS_SKYLAKE(gvt->dev_priv)
1605 || IS_KABYLAKE(gvt->dev_priv)) {
1606 /* BDW decides privilege based on address space */
1607 if (cmd_val(s, 0) & (1 << 8) &&
1608 !(s->vgpu->scan_nonprivbb & (1 << s->ring_id)))
1614 static int find_bb_size(struct parser_exec_state *s, unsigned long *bb_size)
1616 unsigned long gma = 0;
1617 struct cmd_info *info;
1618 uint32_t cmd_len = 0;
1619 bool bb_end = false;
1620 struct intel_vgpu *vgpu = s->vgpu;
1622 struct intel_vgpu_mm *mm = (s->buf_addr_type == GTT_BUFFER) ?
1623 s->vgpu->gtt.ggtt_mm : s->workload->shadow_mm;
1627 /* get the start gm address of the batch buffer */
1628 gma = get_gma_bb_from_cmd(s, 1);
1629 if (gma == INTEL_GVT_INVALID_ADDR)
1632 cmd = cmd_val(s, 0);
1633 info = get_cmd_info(s->vgpu->gvt, cmd, s->ring_id);
1635 gvt_vgpu_err("unknown cmd 0x%x, opcode=0x%x, addr_type=%s, ring %d, workload=%p\n",
1636 cmd, get_opcode(cmd, s->ring_id),
1637 (s->buf_addr_type == PPGTT_BUFFER) ?
1638 "ppgtt" : "ggtt", s->ring_id, s->workload);
1642 if (copy_gma_to_hva(s->vgpu, mm,
1643 gma, gma + 4, &cmd) < 0)
1645 info = get_cmd_info(s->vgpu->gvt, cmd, s->ring_id);
1647 gvt_vgpu_err("unknown cmd 0x%x, opcode=0x%x, addr_type=%s, ring %d, workload=%p\n",
1648 cmd, get_opcode(cmd, s->ring_id),
1649 (s->buf_addr_type == PPGTT_BUFFER) ?
1650 "ppgtt" : "ggtt", s->ring_id, s->workload);
1654 if (info->opcode == OP_MI_BATCH_BUFFER_END) {
1656 } else if (info->opcode == OP_MI_BATCH_BUFFER_START) {
1657 if (BATCH_BUFFER_2ND_LEVEL_BIT(cmd) == 0)
1658 /* chained batch buffer */
1661 cmd_len = get_cmd_length(info, cmd) << 2;
1662 *bb_size += cmd_len;
1669 static int perform_bb_shadow(struct parser_exec_state *s)
1671 struct intel_vgpu *vgpu = s->vgpu;
1672 struct intel_vgpu_shadow_bb *bb;
1673 unsigned long gma = 0;
1674 unsigned long bb_size;
1676 struct intel_vgpu_mm *mm = (s->buf_addr_type == GTT_BUFFER) ?
1677 s->vgpu->gtt.ggtt_mm : s->workload->shadow_mm;
1678 unsigned long gma_start_offset = 0;
1680 /* get the start gm address of the batch buffer */
1681 gma = get_gma_bb_from_cmd(s, 1);
1682 if (gma == INTEL_GVT_INVALID_ADDR)
1685 ret = find_bb_size(s, &bb_size);
1689 bb = kzalloc(sizeof(*bb), GFP_KERNEL);
1693 bb->ppgtt = (s->buf_addr_type == GTT_BUFFER) ? false : true;
1695 /* the gma_start_offset stores the batch buffer's start gma's
1696 * offset relative to page boundary. so for non-privileged batch
1697 * buffer, the shadowed gem object holds exactly the same page
1698 * layout as original gem object. This is for the convience of
1699 * replacing the whole non-privilged batch buffer page to this
1700 * shadowed one in PPGTT at the same gma address. (this replacing
1701 * action is not implemented yet now, but may be necessary in
1703 * for prileged batch buffer, we just change start gma address to
1704 * that of shadowed page.
1707 gma_start_offset = gma & ~I915_GTT_PAGE_MASK;
1709 bb->obj = i915_gem_object_create(s->vgpu->gvt->dev_priv,
1710 roundup(bb_size + gma_start_offset, PAGE_SIZE));
1711 if (IS_ERR(bb->obj)) {
1712 ret = PTR_ERR(bb->obj);
1716 ret = i915_gem_obj_prepare_shmem_write(bb->obj, &bb->clflush);
1720 bb->va = i915_gem_object_pin_map(bb->obj, I915_MAP_WB);
1721 if (IS_ERR(bb->va)) {
1722 ret = PTR_ERR(bb->va);
1723 goto err_finish_shmem_access;
1726 if (bb->clflush & CLFLUSH_BEFORE) {
1727 drm_clflush_virt_range(bb->va, bb->obj->base.size);
1728 bb->clflush &= ~CLFLUSH_BEFORE;
1731 ret = copy_gma_to_hva(s->vgpu, mm,
1733 bb->va + gma_start_offset);
1735 gvt_vgpu_err("fail to copy guest ring buffer\n");
1740 INIT_LIST_HEAD(&bb->list);
1741 list_add(&bb->list, &s->workload->shadow_bb);
1743 bb->accessing = true;
1744 bb->bb_start_cmd_va = s->ip_va;
1746 if ((s->buf_type == BATCH_BUFFER_INSTRUCTION) && (!s->is_ctx_wa))
1747 bb->bb_offset = s->ip_va - s->rb_va;
1752 * ip_va saves the virtual address of the shadow batch buffer, while
1753 * ip_gma saves the graphics address of the original batch buffer.
1754 * As the shadow batch buffer is just a copy from the originial one,
1755 * it should be right to use shadow batch buffer'va and original batch
1756 * buffer's gma in pair. After all, we don't want to pin the shadow
1757 * buffer here (too early).
1759 s->ip_va = bb->va + gma_start_offset;
1763 i915_gem_object_unpin_map(bb->obj);
1764 err_finish_shmem_access:
1765 i915_gem_obj_finish_shmem_access(bb->obj);
1767 i915_gem_object_put(bb->obj);
1773 static int cmd_handler_mi_batch_buffer_start(struct parser_exec_state *s)
1777 struct intel_vgpu *vgpu = s->vgpu;
1779 if (s->buf_type == BATCH_BUFFER_2ND_LEVEL) {
1780 gvt_vgpu_err("Found MI_BATCH_BUFFER_START in 2nd level BB\n");
1784 second_level = BATCH_BUFFER_2ND_LEVEL_BIT(cmd_val(s, 0)) == 1;
1785 if (second_level && (s->buf_type != BATCH_BUFFER_INSTRUCTION)) {
1786 gvt_vgpu_err("Jumping to 2nd level BB from RB is not allowed\n");
1790 s->saved_buf_addr_type = s->buf_addr_type;
1791 addr_type_update_snb(s);
1792 if (s->buf_type == RING_BUFFER_INSTRUCTION) {
1793 s->ret_ip_gma_ring = s->ip_gma + cmd_length(s) * sizeof(u32);
1794 s->buf_type = BATCH_BUFFER_INSTRUCTION;
1795 } else if (second_level) {
1796 s->buf_type = BATCH_BUFFER_2ND_LEVEL;
1797 s->ret_ip_gma_bb = s->ip_gma + cmd_length(s) * sizeof(u32);
1798 s->ret_bb_va = s->ip_va + cmd_length(s) * sizeof(u32);
1801 if (batch_buffer_needs_scan(s)) {
1802 ret = perform_bb_shadow(s);
1804 gvt_vgpu_err("invalid shadow batch buffer\n");
1806 /* emulate a batch buffer end to do return right */
1807 ret = cmd_handler_mi_batch_buffer_end(s);
1814 static struct cmd_info cmd_info[] = {
1815 {"MI_NOOP", OP_MI_NOOP, F_LEN_CONST, R_ALL, D_ALL, 0, 1, NULL},
1817 {"MI_SET_PREDICATE", OP_MI_SET_PREDICATE, F_LEN_CONST, R_ALL, D_ALL,
1820 {"MI_USER_INTERRUPT", OP_MI_USER_INTERRUPT, F_LEN_CONST, R_ALL, D_ALL,
1821 0, 1, cmd_handler_mi_user_interrupt},
1823 {"MI_WAIT_FOR_EVENT", OP_MI_WAIT_FOR_EVENT, F_LEN_CONST, R_RCS | R_BCS,
1824 D_ALL, 0, 1, cmd_handler_mi_wait_for_event},
1826 {"MI_FLUSH", OP_MI_FLUSH, F_LEN_CONST, R_ALL, D_ALL, 0, 1, NULL},
1828 {"MI_ARB_CHECK", OP_MI_ARB_CHECK, F_LEN_CONST, R_ALL, D_ALL, 0, 1,
1831 {"MI_RS_CONTROL", OP_MI_RS_CONTROL, F_LEN_CONST, R_RCS, D_ALL, 0, 1,
1834 {"MI_REPORT_HEAD", OP_MI_REPORT_HEAD, F_LEN_CONST, R_ALL, D_ALL, 0, 1,
1837 {"MI_ARB_ON_OFF", OP_MI_ARB_ON_OFF, F_LEN_CONST, R_ALL, D_ALL, 0, 1,
1840 {"MI_URB_ATOMIC_ALLOC", OP_MI_URB_ATOMIC_ALLOC, F_LEN_CONST, R_RCS,
1843 {"MI_BATCH_BUFFER_END", OP_MI_BATCH_BUFFER_END,
1844 F_IP_ADVANCE_CUSTOM | F_LEN_CONST, R_ALL, D_ALL, 0, 1,
1845 cmd_handler_mi_batch_buffer_end},
1847 {"MI_SUSPEND_FLUSH", OP_MI_SUSPEND_FLUSH, F_LEN_CONST, R_ALL, D_ALL,
1850 {"MI_PREDICATE", OP_MI_PREDICATE, F_LEN_CONST, R_RCS, D_ALL, 0, 1,
1853 {"MI_TOPOLOGY_FILTER", OP_MI_TOPOLOGY_FILTER, F_LEN_CONST, R_ALL,
1856 {"MI_SET_APPID", OP_MI_SET_APPID, F_LEN_CONST, R_ALL, D_ALL, 0, 1,
1859 {"MI_RS_CONTEXT", OP_MI_RS_CONTEXT, F_LEN_CONST, R_RCS, D_ALL, 0, 1,
1862 {"MI_DISPLAY_FLIP", OP_MI_DISPLAY_FLIP, F_LEN_VAR | F_POST_HANDLE,
1863 R_RCS | R_BCS, D_ALL, 0, 8, cmd_handler_mi_display_flip},
1865 {"MI_SEMAPHORE_MBOX", OP_MI_SEMAPHORE_MBOX, F_LEN_VAR, R_ALL, D_ALL,
1868 {"MI_MATH", OP_MI_MATH, F_LEN_VAR, R_ALL, D_ALL, 0, 8, NULL},
1870 {"MI_URB_CLEAR", OP_MI_URB_CLEAR, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
1872 {"ME_SEMAPHORE_SIGNAL", OP_MI_SEMAPHORE_SIGNAL, F_LEN_VAR, R_ALL,
1873 D_BDW_PLUS, 0, 8, NULL},
1875 {"ME_SEMAPHORE_WAIT", OP_MI_SEMAPHORE_WAIT, F_LEN_VAR, R_ALL, D_BDW_PLUS,
1876 ADDR_FIX_1(2), 8, cmd_handler_mi_semaphore_wait},
1878 {"MI_STORE_DATA_IMM", OP_MI_STORE_DATA_IMM, F_LEN_VAR, R_ALL, D_BDW_PLUS,
1879 ADDR_FIX_1(1), 10, cmd_handler_mi_store_data_imm},
1881 {"MI_STORE_DATA_INDEX", OP_MI_STORE_DATA_INDEX, F_LEN_VAR, R_ALL, D_ALL,
1882 0, 8, cmd_handler_mi_store_data_index},
1884 {"MI_LOAD_REGISTER_IMM", OP_MI_LOAD_REGISTER_IMM, F_LEN_VAR, R_ALL,
1885 D_ALL, 0, 8, cmd_handler_lri},
1887 {"MI_UPDATE_GTT", OP_MI_UPDATE_GTT, F_LEN_VAR, R_ALL, D_BDW_PLUS, 0, 10,
1888 cmd_handler_mi_update_gtt},
1890 {"MI_STORE_REGISTER_MEM", OP_MI_STORE_REGISTER_MEM, F_LEN_VAR, R_ALL,
1891 D_ALL, ADDR_FIX_1(2), 8, cmd_handler_srm},
1893 {"MI_FLUSH_DW", OP_MI_FLUSH_DW, F_LEN_VAR, R_ALL, D_ALL, 0, 6,
1894 cmd_handler_mi_flush_dw},
1896 {"MI_CLFLUSH", OP_MI_CLFLUSH, F_LEN_VAR, R_ALL, D_ALL, ADDR_FIX_1(1),
1897 10, cmd_handler_mi_clflush},
1899 {"MI_REPORT_PERF_COUNT", OP_MI_REPORT_PERF_COUNT, F_LEN_VAR, R_ALL,
1900 D_ALL, ADDR_FIX_1(1), 6, cmd_handler_mi_report_perf_count},
1902 {"MI_LOAD_REGISTER_MEM", OP_MI_LOAD_REGISTER_MEM, F_LEN_VAR, R_ALL,
1903 D_ALL, ADDR_FIX_1(2), 8, cmd_handler_lrm},
1905 {"MI_LOAD_REGISTER_REG", OP_MI_LOAD_REGISTER_REG, F_LEN_VAR, R_ALL,
1906 D_ALL, 0, 8, cmd_handler_lrr},
1908 {"MI_RS_STORE_DATA_IMM", OP_MI_RS_STORE_DATA_IMM, F_LEN_VAR, R_RCS,
1911 {"MI_LOAD_URB_MEM", OP_MI_LOAD_URB_MEM, F_LEN_VAR, R_RCS, D_ALL,
1912 ADDR_FIX_1(2), 8, NULL},
1914 {"MI_STORE_URM_MEM", OP_MI_STORE_URM_MEM, F_LEN_VAR, R_RCS, D_ALL,
1915 ADDR_FIX_1(2), 8, NULL},
1917 {"MI_OP_2E", OP_MI_2E, F_LEN_VAR, R_ALL, D_BDW_PLUS, ADDR_FIX_2(1, 2),
1918 8, cmd_handler_mi_op_2e},
1920 {"MI_OP_2F", OP_MI_2F, F_LEN_VAR, R_ALL, D_BDW_PLUS, ADDR_FIX_1(1),
1921 8, cmd_handler_mi_op_2f},
1923 {"MI_BATCH_BUFFER_START", OP_MI_BATCH_BUFFER_START,
1924 F_IP_ADVANCE_CUSTOM, R_ALL, D_ALL, 0, 8,
1925 cmd_handler_mi_batch_buffer_start},
1927 {"MI_CONDITIONAL_BATCH_BUFFER_END", OP_MI_CONDITIONAL_BATCH_BUFFER_END,
1928 F_LEN_VAR, R_ALL, D_ALL, ADDR_FIX_1(2), 8,
1929 cmd_handler_mi_conditional_batch_buffer_end},
1931 {"MI_LOAD_SCAN_LINES_INCL", OP_MI_LOAD_SCAN_LINES_INCL, F_LEN_CONST,
1932 R_RCS | R_BCS, D_ALL, 0, 2, NULL},
1934 {"XY_SETUP_BLT", OP_XY_SETUP_BLT, F_LEN_VAR, R_BCS, D_ALL,
1935 ADDR_FIX_2(4, 7), 8, NULL},
1937 {"XY_SETUP_CLIP_BLT", OP_XY_SETUP_CLIP_BLT, F_LEN_VAR, R_BCS, D_ALL,
1940 {"XY_SETUP_MONO_PATTERN_SL_BLT", OP_XY_SETUP_MONO_PATTERN_SL_BLT,
1941 F_LEN_VAR, R_BCS, D_ALL, ADDR_FIX_1(4), 8, NULL},
1943 {"XY_PIXEL_BLT", OP_XY_PIXEL_BLT, F_LEN_VAR, R_BCS, D_ALL, 0, 8, NULL},
1945 {"XY_SCANLINES_BLT", OP_XY_SCANLINES_BLT, F_LEN_VAR, R_BCS, D_ALL,
1948 {"XY_TEXT_BLT", OP_XY_TEXT_BLT, F_LEN_VAR, R_BCS, D_ALL,
1949 ADDR_FIX_1(3), 8, NULL},
1951 {"XY_TEXT_IMMEDIATE_BLT", OP_XY_TEXT_IMMEDIATE_BLT, F_LEN_VAR, R_BCS,
1954 {"XY_COLOR_BLT", OP_XY_COLOR_BLT, F_LEN_VAR, R_BCS, D_ALL,
1955 ADDR_FIX_1(4), 8, NULL},
1957 {"XY_PAT_BLT", OP_XY_PAT_BLT, F_LEN_VAR, R_BCS, D_ALL,
1958 ADDR_FIX_2(4, 5), 8, NULL},
1960 {"XY_MONO_PAT_BLT", OP_XY_MONO_PAT_BLT, F_LEN_VAR, R_BCS, D_ALL,
1961 ADDR_FIX_1(4), 8, NULL},
1963 {"XY_SRC_COPY_BLT", OP_XY_SRC_COPY_BLT, F_LEN_VAR, R_BCS, D_ALL,
1964 ADDR_FIX_2(4, 7), 8, NULL},
1966 {"XY_MONO_SRC_COPY_BLT", OP_XY_MONO_SRC_COPY_BLT, F_LEN_VAR, R_BCS,
1967 D_ALL, ADDR_FIX_2(4, 5), 8, NULL},
1969 {"XY_FULL_BLT", OP_XY_FULL_BLT, F_LEN_VAR, R_BCS, D_ALL, 0, 8, NULL},
1971 {"XY_FULL_MONO_SRC_BLT", OP_XY_FULL_MONO_SRC_BLT, F_LEN_VAR, R_BCS,
1972 D_ALL, ADDR_FIX_3(4, 5, 8), 8, NULL},
1974 {"XY_FULL_MONO_PATTERN_BLT", OP_XY_FULL_MONO_PATTERN_BLT, F_LEN_VAR,
1975 R_BCS, D_ALL, ADDR_FIX_2(4, 7), 8, NULL},
1977 {"XY_FULL_MONO_PATTERN_MONO_SRC_BLT",
1978 OP_XY_FULL_MONO_PATTERN_MONO_SRC_BLT,
1979 F_LEN_VAR, R_BCS, D_ALL, ADDR_FIX_2(4, 5), 8, NULL},
1981 {"XY_MONO_PAT_FIXED_BLT", OP_XY_MONO_PAT_FIXED_BLT, F_LEN_VAR, R_BCS,
1982 D_ALL, ADDR_FIX_1(4), 8, NULL},
1984 {"XY_MONO_SRC_COPY_IMMEDIATE_BLT", OP_XY_MONO_SRC_COPY_IMMEDIATE_BLT,
1985 F_LEN_VAR, R_BCS, D_ALL, ADDR_FIX_1(4), 8, NULL},
1987 {"XY_PAT_BLT_IMMEDIATE", OP_XY_PAT_BLT_IMMEDIATE, F_LEN_VAR, R_BCS,
1988 D_ALL, ADDR_FIX_1(4), 8, NULL},
1990 {"XY_SRC_COPY_CHROMA_BLT", OP_XY_SRC_COPY_CHROMA_BLT, F_LEN_VAR, R_BCS,
1991 D_ALL, ADDR_FIX_2(4, 7), 8, NULL},
1993 {"XY_FULL_IMMEDIATE_PATTERN_BLT", OP_XY_FULL_IMMEDIATE_PATTERN_BLT,
1994 F_LEN_VAR, R_BCS, D_ALL, ADDR_FIX_2(4, 7), 8, NULL},
1996 {"XY_FULL_MONO_SRC_IMMEDIATE_PATTERN_BLT",
1997 OP_XY_FULL_MONO_SRC_IMMEDIATE_PATTERN_BLT,
1998 F_LEN_VAR, R_BCS, D_ALL, ADDR_FIX_2(4, 5), 8, NULL},
2000 {"XY_PAT_CHROMA_BLT", OP_XY_PAT_CHROMA_BLT, F_LEN_VAR, R_BCS, D_ALL,
2001 ADDR_FIX_2(4, 5), 8, NULL},
2003 {"XY_PAT_CHROMA_BLT_IMMEDIATE", OP_XY_PAT_CHROMA_BLT_IMMEDIATE,
2004 F_LEN_VAR, R_BCS, D_ALL, ADDR_FIX_1(4), 8, NULL},
2006 {"3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP",
2007 OP_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP,
2008 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2010 {"3DSTATE_VIEWPORT_STATE_POINTERS_CC",
2011 OP_3DSTATE_VIEWPORT_STATE_POINTERS_CC,
2012 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2014 {"3DSTATE_BLEND_STATE_POINTERS",
2015 OP_3DSTATE_BLEND_STATE_POINTERS,
2016 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2018 {"3DSTATE_DEPTH_STENCIL_STATE_POINTERS",
2019 OP_3DSTATE_DEPTH_STENCIL_STATE_POINTERS,
2020 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2022 {"3DSTATE_BINDING_TABLE_POINTERS_VS",
2023 OP_3DSTATE_BINDING_TABLE_POINTERS_VS,
2024 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2026 {"3DSTATE_BINDING_TABLE_POINTERS_HS",
2027 OP_3DSTATE_BINDING_TABLE_POINTERS_HS,
2028 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2030 {"3DSTATE_BINDING_TABLE_POINTERS_DS",
2031 OP_3DSTATE_BINDING_TABLE_POINTERS_DS,
2032 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2034 {"3DSTATE_BINDING_TABLE_POINTERS_GS",
2035 OP_3DSTATE_BINDING_TABLE_POINTERS_GS,
2036 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2038 {"3DSTATE_BINDING_TABLE_POINTERS_PS",
2039 OP_3DSTATE_BINDING_TABLE_POINTERS_PS,
2040 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2042 {"3DSTATE_SAMPLER_STATE_POINTERS_VS",
2043 OP_3DSTATE_SAMPLER_STATE_POINTERS_VS,
2044 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2046 {"3DSTATE_SAMPLER_STATE_POINTERS_HS",
2047 OP_3DSTATE_SAMPLER_STATE_POINTERS_HS,
2048 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2050 {"3DSTATE_SAMPLER_STATE_POINTERS_DS",
2051 OP_3DSTATE_SAMPLER_STATE_POINTERS_DS,
2052 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2054 {"3DSTATE_SAMPLER_STATE_POINTERS_GS",
2055 OP_3DSTATE_SAMPLER_STATE_POINTERS_GS,
2056 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2058 {"3DSTATE_SAMPLER_STATE_POINTERS_PS",
2059 OP_3DSTATE_SAMPLER_STATE_POINTERS_PS,
2060 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2062 {"3DSTATE_URB_VS", OP_3DSTATE_URB_VS, F_LEN_VAR, R_RCS, D_ALL,
2065 {"3DSTATE_URB_HS", OP_3DSTATE_URB_HS, F_LEN_VAR, R_RCS, D_ALL,
2068 {"3DSTATE_URB_DS", OP_3DSTATE_URB_DS, F_LEN_VAR, R_RCS, D_ALL,
2071 {"3DSTATE_URB_GS", OP_3DSTATE_URB_GS, F_LEN_VAR, R_RCS, D_ALL,
2074 {"3DSTATE_GATHER_CONSTANT_VS", OP_3DSTATE_GATHER_CONSTANT_VS,
2075 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2077 {"3DSTATE_GATHER_CONSTANT_GS", OP_3DSTATE_GATHER_CONSTANT_GS,
2078 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2080 {"3DSTATE_GATHER_CONSTANT_HS", OP_3DSTATE_GATHER_CONSTANT_HS,
2081 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2083 {"3DSTATE_GATHER_CONSTANT_DS", OP_3DSTATE_GATHER_CONSTANT_DS,
2084 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2086 {"3DSTATE_GATHER_CONSTANT_PS", OP_3DSTATE_GATHER_CONSTANT_PS,
2087 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2089 {"3DSTATE_DX9_CONSTANTF_VS", OP_3DSTATE_DX9_CONSTANTF_VS,
2090 F_LEN_VAR, R_RCS, D_ALL, 0, 11, NULL},
2092 {"3DSTATE_DX9_CONSTANTF_PS", OP_3DSTATE_DX9_CONSTANTF_PS,
2093 F_LEN_VAR, R_RCS, D_ALL, 0, 11, NULL},
2095 {"3DSTATE_DX9_CONSTANTI_VS", OP_3DSTATE_DX9_CONSTANTI_VS,
2096 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2098 {"3DSTATE_DX9_CONSTANTI_PS", OP_3DSTATE_DX9_CONSTANTI_PS,
2099 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2101 {"3DSTATE_DX9_CONSTANTB_VS", OP_3DSTATE_DX9_CONSTANTB_VS,
2102 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2104 {"3DSTATE_DX9_CONSTANTB_PS", OP_3DSTATE_DX9_CONSTANTB_PS,
2105 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2107 {"3DSTATE_DX9_LOCAL_VALID_VS", OP_3DSTATE_DX9_LOCAL_VALID_VS,
2108 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2110 {"3DSTATE_DX9_LOCAL_VALID_PS", OP_3DSTATE_DX9_LOCAL_VALID_PS,
2111 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2113 {"3DSTATE_DX9_GENERATE_ACTIVE_VS", OP_3DSTATE_DX9_GENERATE_ACTIVE_VS,
2114 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2116 {"3DSTATE_DX9_GENERATE_ACTIVE_PS", OP_3DSTATE_DX9_GENERATE_ACTIVE_PS,
2117 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2119 {"3DSTATE_BINDING_TABLE_EDIT_VS", OP_3DSTATE_BINDING_TABLE_EDIT_VS,
2120 F_LEN_VAR, R_RCS, D_ALL, 0, 9, NULL},
2122 {"3DSTATE_BINDING_TABLE_EDIT_GS", OP_3DSTATE_BINDING_TABLE_EDIT_GS,
2123 F_LEN_VAR, R_RCS, D_ALL, 0, 9, NULL},
2125 {"3DSTATE_BINDING_TABLE_EDIT_HS", OP_3DSTATE_BINDING_TABLE_EDIT_HS,
2126 F_LEN_VAR, R_RCS, D_ALL, 0, 9, NULL},
2128 {"3DSTATE_BINDING_TABLE_EDIT_DS", OP_3DSTATE_BINDING_TABLE_EDIT_DS,
2129 F_LEN_VAR, R_RCS, D_ALL, 0, 9, NULL},
2131 {"3DSTATE_BINDING_TABLE_EDIT_PS", OP_3DSTATE_BINDING_TABLE_EDIT_PS,
2132 F_LEN_VAR, R_RCS, D_ALL, 0, 9, NULL},
2134 {"3DSTATE_VF_INSTANCING", OP_3DSTATE_VF_INSTANCING, F_LEN_VAR, R_RCS,
2135 D_BDW_PLUS, 0, 8, NULL},
2137 {"3DSTATE_VF_SGVS", OP_3DSTATE_VF_SGVS, F_LEN_VAR, R_RCS, D_BDW_PLUS, 0, 8,
2140 {"3DSTATE_VF_TOPOLOGY", OP_3DSTATE_VF_TOPOLOGY, F_LEN_VAR, R_RCS,
2141 D_BDW_PLUS, 0, 8, NULL},
2143 {"3DSTATE_WM_CHROMAKEY", OP_3DSTATE_WM_CHROMAKEY, F_LEN_VAR, R_RCS,
2144 D_BDW_PLUS, 0, 8, NULL},
2146 {"3DSTATE_PS_BLEND", OP_3DSTATE_PS_BLEND, F_LEN_VAR, R_RCS, D_BDW_PLUS, 0,
2149 {"3DSTATE_WM_DEPTH_STENCIL", OP_3DSTATE_WM_DEPTH_STENCIL, F_LEN_VAR,
2150 R_RCS, D_BDW_PLUS, 0, 8, NULL},
2152 {"3DSTATE_PS_EXTRA", OP_3DSTATE_PS_EXTRA, F_LEN_VAR, R_RCS, D_BDW_PLUS, 0,
2155 {"3DSTATE_RASTER", OP_3DSTATE_RASTER, F_LEN_VAR, R_RCS, D_BDW_PLUS, 0, 8,
2158 {"3DSTATE_SBE_SWIZ", OP_3DSTATE_SBE_SWIZ, F_LEN_VAR, R_RCS, D_BDW_PLUS, 0, 8,
2161 {"3DSTATE_WM_HZ_OP", OP_3DSTATE_WM_HZ_OP, F_LEN_VAR, R_RCS, D_BDW_PLUS, 0, 8,
2164 {"3DSTATE_VERTEX_BUFFERS", OP_3DSTATE_VERTEX_BUFFERS, F_LEN_VAR, R_RCS,
2165 D_BDW_PLUS, 0, 8, NULL},
2167 {"3DSTATE_VERTEX_ELEMENTS", OP_3DSTATE_VERTEX_ELEMENTS, F_LEN_VAR,
2168 R_RCS, D_ALL, 0, 8, NULL},
2170 {"3DSTATE_INDEX_BUFFER", OP_3DSTATE_INDEX_BUFFER, F_LEN_VAR, R_RCS,
2171 D_BDW_PLUS, ADDR_FIX_1(2), 8, NULL},
2173 {"3DSTATE_VF_STATISTICS", OP_3DSTATE_VF_STATISTICS, F_LEN_CONST,
2174 R_RCS, D_ALL, 0, 1, NULL},
2176 {"3DSTATE_VF", OP_3DSTATE_VF, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2178 {"3DSTATE_CC_STATE_POINTERS", OP_3DSTATE_CC_STATE_POINTERS, F_LEN_VAR,
2179 R_RCS, D_ALL, 0, 8, NULL},
2181 {"3DSTATE_SCISSOR_STATE_POINTERS", OP_3DSTATE_SCISSOR_STATE_POINTERS,
2182 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2184 {"3DSTATE_GS", OP_3DSTATE_GS, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2186 {"3DSTATE_CLIP", OP_3DSTATE_CLIP, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2188 {"3DSTATE_WM", OP_3DSTATE_WM, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2190 {"3DSTATE_CONSTANT_GS", OP_3DSTATE_CONSTANT_GS, F_LEN_VAR, R_RCS,
2191 D_BDW_PLUS, 0, 8, NULL},
2193 {"3DSTATE_CONSTANT_PS", OP_3DSTATE_CONSTANT_PS, F_LEN_VAR, R_RCS,
2194 D_BDW_PLUS, 0, 8, NULL},
2196 {"3DSTATE_SAMPLE_MASK", OP_3DSTATE_SAMPLE_MASK, F_LEN_VAR, R_RCS,
2199 {"3DSTATE_CONSTANT_HS", OP_3DSTATE_CONSTANT_HS, F_LEN_VAR, R_RCS,
2200 D_BDW_PLUS, 0, 8, NULL},
2202 {"3DSTATE_CONSTANT_DS", OP_3DSTATE_CONSTANT_DS, F_LEN_VAR, R_RCS,
2203 D_BDW_PLUS, 0, 8, NULL},
2205 {"3DSTATE_HS", OP_3DSTATE_HS, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2207 {"3DSTATE_TE", OP_3DSTATE_TE, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2209 {"3DSTATE_DS", OP_3DSTATE_DS, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2211 {"3DSTATE_STREAMOUT", OP_3DSTATE_STREAMOUT, F_LEN_VAR, R_RCS,
2214 {"3DSTATE_SBE", OP_3DSTATE_SBE, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2216 {"3DSTATE_PS", OP_3DSTATE_PS, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2218 {"3DSTATE_DRAWING_RECTANGLE", OP_3DSTATE_DRAWING_RECTANGLE, F_LEN_VAR,
2219 R_RCS, D_ALL, 0, 8, NULL},
2221 {"3DSTATE_SAMPLER_PALETTE_LOAD0", OP_3DSTATE_SAMPLER_PALETTE_LOAD0,
2222 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2224 {"3DSTATE_CHROMA_KEY", OP_3DSTATE_CHROMA_KEY, F_LEN_VAR, R_RCS, D_ALL,
2227 {"3DSTATE_DEPTH_BUFFER", OP_3DSTATE_DEPTH_BUFFER, F_LEN_VAR, R_RCS,
2228 D_ALL, ADDR_FIX_1(2), 8, NULL},
2230 {"3DSTATE_POLY_STIPPLE_OFFSET", OP_3DSTATE_POLY_STIPPLE_OFFSET,
2231 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2233 {"3DSTATE_POLY_STIPPLE_PATTERN", OP_3DSTATE_POLY_STIPPLE_PATTERN,
2234 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2236 {"3DSTATE_LINE_STIPPLE", OP_3DSTATE_LINE_STIPPLE, F_LEN_VAR, R_RCS,
2239 {"3DSTATE_AA_LINE_PARAMS", OP_3DSTATE_AA_LINE_PARAMS, F_LEN_VAR, R_RCS,
2242 {"3DSTATE_GS_SVB_INDEX", OP_3DSTATE_GS_SVB_INDEX, F_LEN_VAR, R_RCS,
2245 {"3DSTATE_SAMPLER_PALETTE_LOAD1", OP_3DSTATE_SAMPLER_PALETTE_LOAD1,
2246 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2248 {"3DSTATE_MULTISAMPLE", OP_3DSTATE_MULTISAMPLE_BDW, F_LEN_VAR, R_RCS,
2249 D_BDW_PLUS, 0, 8, NULL},
2251 {"3DSTATE_STENCIL_BUFFER", OP_3DSTATE_STENCIL_BUFFER, F_LEN_VAR, R_RCS,
2252 D_ALL, ADDR_FIX_1(2), 8, NULL},
2254 {"3DSTATE_HIER_DEPTH_BUFFER", OP_3DSTATE_HIER_DEPTH_BUFFER, F_LEN_VAR,
2255 R_RCS, D_ALL, ADDR_FIX_1(2), 8, NULL},
2257 {"3DSTATE_CLEAR_PARAMS", OP_3DSTATE_CLEAR_PARAMS, F_LEN_VAR,
2258 R_RCS, D_ALL, 0, 8, NULL},
2260 {"3DSTATE_PUSH_CONSTANT_ALLOC_VS", OP_3DSTATE_PUSH_CONSTANT_ALLOC_VS,
2261 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2263 {"3DSTATE_PUSH_CONSTANT_ALLOC_HS", OP_3DSTATE_PUSH_CONSTANT_ALLOC_HS,
2264 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2266 {"3DSTATE_PUSH_CONSTANT_ALLOC_DS", OP_3DSTATE_PUSH_CONSTANT_ALLOC_DS,
2267 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2269 {"3DSTATE_PUSH_CONSTANT_ALLOC_GS", OP_3DSTATE_PUSH_CONSTANT_ALLOC_GS,
2270 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2272 {"3DSTATE_PUSH_CONSTANT_ALLOC_PS", OP_3DSTATE_PUSH_CONSTANT_ALLOC_PS,
2273 F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2275 {"3DSTATE_MONOFILTER_SIZE", OP_3DSTATE_MONOFILTER_SIZE, F_LEN_VAR,
2276 R_RCS, D_ALL, 0, 8, NULL},
2278 {"3DSTATE_SO_DECL_LIST", OP_3DSTATE_SO_DECL_LIST, F_LEN_VAR, R_RCS,
2281 {"3DSTATE_SO_BUFFER", OP_3DSTATE_SO_BUFFER, F_LEN_VAR, R_RCS, D_BDW_PLUS,
2282 ADDR_FIX_2(2, 4), 8, NULL},
2284 {"3DSTATE_BINDING_TABLE_POOL_ALLOC",
2285 OP_3DSTATE_BINDING_TABLE_POOL_ALLOC,
2286 F_LEN_VAR, R_RCS, D_BDW_PLUS, ADDR_FIX_1(1), 8, NULL},
2288 {"3DSTATE_GATHER_POOL_ALLOC", OP_3DSTATE_GATHER_POOL_ALLOC,
2289 F_LEN_VAR, R_RCS, D_BDW_PLUS, ADDR_FIX_1(1), 8, NULL},
2291 {"3DSTATE_DX9_CONSTANT_BUFFER_POOL_ALLOC",
2292 OP_3DSTATE_DX9_CONSTANT_BUFFER_POOL_ALLOC,
2293 F_LEN_VAR, R_RCS, D_BDW_PLUS, ADDR_FIX_1(1), 8, NULL},
2295 {"3DSTATE_SAMPLE_PATTERN", OP_3DSTATE_SAMPLE_PATTERN, F_LEN_VAR, R_RCS,
2296 D_BDW_PLUS, 0, 8, NULL},
2298 {"PIPE_CONTROL", OP_PIPE_CONTROL, F_LEN_VAR, R_RCS, D_ALL,
2299 ADDR_FIX_1(2), 8, cmd_handler_pipe_control},
2301 {"3DPRIMITIVE", OP_3DPRIMITIVE, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2303 {"PIPELINE_SELECT", OP_PIPELINE_SELECT, F_LEN_CONST, R_RCS, D_ALL, 0,
2306 {"STATE_PREFETCH", OP_STATE_PREFETCH, F_LEN_VAR, R_RCS, D_ALL,
2307 ADDR_FIX_1(1), 8, NULL},
2309 {"STATE_SIP", OP_STATE_SIP, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2311 {"STATE_BASE_ADDRESS", OP_STATE_BASE_ADDRESS, F_LEN_VAR, R_RCS, D_BDW_PLUS,
2312 ADDR_FIX_5(1, 3, 4, 5, 6), 8, NULL},
2314 {"OP_3D_MEDIA_0_1_4", OP_3D_MEDIA_0_1_4, F_LEN_VAR, R_RCS, D_ALL,
2315 ADDR_FIX_1(1), 8, NULL},
2317 {"3DSTATE_VS", OP_3DSTATE_VS, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2319 {"3DSTATE_SF", OP_3DSTATE_SF, F_LEN_VAR, R_RCS, D_ALL, 0, 8, NULL},
2321 {"3DSTATE_CONSTANT_VS", OP_3DSTATE_CONSTANT_VS, F_LEN_VAR, R_RCS, D_BDW_PLUS,
2324 {"3DSTATE_COMPONENT_PACKING", OP_3DSTATE_COMPONENT_PACKING, F_LEN_VAR, R_RCS,
2325 D_SKL_PLUS, 0, 8, NULL},
2327 {"MEDIA_INTERFACE_DESCRIPTOR_LOAD", OP_MEDIA_INTERFACE_DESCRIPTOR_LOAD,
2328 F_LEN_VAR, R_RCS, D_ALL, 0, 16, NULL},
2330 {"MEDIA_GATEWAY_STATE", OP_MEDIA_GATEWAY_STATE, F_LEN_VAR, R_RCS, D_ALL,
2333 {"MEDIA_STATE_FLUSH", OP_MEDIA_STATE_FLUSH, F_LEN_VAR, R_RCS, D_ALL,
2336 {"MEDIA_OBJECT", OP_MEDIA_OBJECT, F_LEN_VAR, R_RCS, D_ALL, 0, 16, NULL},
2338 {"MEDIA_CURBE_LOAD", OP_MEDIA_CURBE_LOAD, F_LEN_VAR, R_RCS, D_ALL,
2341 {"MEDIA_OBJECT_PRT", OP_MEDIA_OBJECT_PRT, F_LEN_VAR, R_RCS, D_ALL,
2344 {"MEDIA_OBJECT_WALKER", OP_MEDIA_OBJECT_WALKER, F_LEN_VAR, R_RCS, D_ALL,
2347 {"GPGPU_WALKER", OP_GPGPU_WALKER, F_LEN_VAR, R_RCS, D_ALL,
2350 {"MEDIA_VFE_STATE", OP_MEDIA_VFE_STATE, F_LEN_VAR, R_RCS, D_ALL, 0, 16,
2353 {"3DSTATE_VF_STATISTICS_GM45", OP_3DSTATE_VF_STATISTICS_GM45,
2354 F_LEN_CONST, R_ALL, D_ALL, 0, 1, NULL},
2356 {"MFX_PIPE_MODE_SELECT", OP_MFX_PIPE_MODE_SELECT, F_LEN_VAR,
2357 R_VCS, D_ALL, 0, 12, NULL},
2359 {"MFX_SURFACE_STATE", OP_MFX_SURFACE_STATE, F_LEN_VAR,
2360 R_VCS, D_ALL, 0, 12, NULL},
2362 {"MFX_PIPE_BUF_ADDR_STATE", OP_MFX_PIPE_BUF_ADDR_STATE, F_LEN_VAR,
2363 R_VCS, D_BDW_PLUS, 0, 12, NULL},
2365 {"MFX_IND_OBJ_BASE_ADDR_STATE", OP_MFX_IND_OBJ_BASE_ADDR_STATE,
2366 F_LEN_VAR, R_VCS, D_BDW_PLUS, 0, 12, NULL},
2368 {"MFX_BSP_BUF_BASE_ADDR_STATE", OP_MFX_BSP_BUF_BASE_ADDR_STATE,
2369 F_LEN_VAR, R_VCS, D_BDW_PLUS, ADDR_FIX_3(1, 3, 5), 12, NULL},
2371 {"OP_2_0_0_5", OP_2_0_0_5, F_LEN_VAR, R_VCS, D_BDW_PLUS, 0, 12, NULL},
2373 {"MFX_STATE_POINTER", OP_MFX_STATE_POINTER, F_LEN_VAR,
2374 R_VCS, D_ALL, 0, 12, NULL},
2376 {"MFX_QM_STATE", OP_MFX_QM_STATE, F_LEN_VAR,
2377 R_VCS, D_ALL, 0, 12, NULL},
2379 {"MFX_FQM_STATE", OP_MFX_FQM_STATE, F_LEN_VAR,
2380 R_VCS, D_ALL, 0, 12, NULL},
2382 {"MFX_PAK_INSERT_OBJECT", OP_MFX_PAK_INSERT_OBJECT, F_LEN_VAR,
2383 R_VCS, D_ALL, 0, 12, NULL},
2385 {"MFX_STITCH_OBJECT", OP_MFX_STITCH_OBJECT, F_LEN_VAR,
2386 R_VCS, D_ALL, 0, 12, NULL},
2388 {"MFD_IT_OBJECT", OP_MFD_IT_OBJECT, F_LEN_VAR,
2389 R_VCS, D_ALL, 0, 12, NULL},
2391 {"MFX_WAIT", OP_MFX_WAIT, F_LEN_VAR,
2392 R_VCS, D_ALL, 0, 6, NULL},
2394 {"MFX_AVC_IMG_STATE", OP_MFX_AVC_IMG_STATE, F_LEN_VAR,
2395 R_VCS, D_ALL, 0, 12, NULL},
2397 {"MFX_AVC_QM_STATE", OP_MFX_AVC_QM_STATE, F_LEN_VAR,
2398 R_VCS, D_ALL, 0, 12, NULL},
2400 {"MFX_AVC_DIRECTMODE_STATE", OP_MFX_AVC_DIRECTMODE_STATE, F_LEN_VAR,
2401 R_VCS, D_ALL, 0, 12, NULL},
2403 {"MFX_AVC_SLICE_STATE", OP_MFX_AVC_SLICE_STATE, F_LEN_VAR,
2404 R_VCS, D_ALL, 0, 12, NULL},
2406 {"MFX_AVC_REF_IDX_STATE", OP_MFX_AVC_REF_IDX_STATE, F_LEN_VAR,
2407 R_VCS, D_ALL, 0, 12, NULL},
2409 {"MFX_AVC_WEIGHTOFFSET_STATE", OP_MFX_AVC_WEIGHTOFFSET_STATE, F_LEN_VAR,
2410 R_VCS, D_ALL, 0, 12, NULL},
2412 {"MFD_AVC_PICID_STATE", OP_MFD_AVC_PICID_STATE, F_LEN_VAR,
2413 R_VCS, D_ALL, 0, 12, NULL},
2414 {"MFD_AVC_DPB_STATE", OP_MFD_AVC_DPB_STATE, F_LEN_VAR,
2415 R_VCS, D_ALL, 0, 12, NULL},
2417 {"MFD_AVC_BSD_OBJECT", OP_MFD_AVC_BSD_OBJECT, F_LEN_VAR,
2418 R_VCS, D_ALL, 0, 12, NULL},
2420 {"MFD_AVC_SLICEADDR", OP_MFD_AVC_SLICEADDR, F_LEN_VAR,
2421 R_VCS, D_ALL, ADDR_FIX_1(2), 12, NULL},
2423 {"MFC_AVC_PAK_OBJECT", OP_MFC_AVC_PAK_OBJECT, F_LEN_VAR,
2424 R_VCS, D_ALL, 0, 12, NULL},
2426 {"MFX_VC1_PRED_PIPE_STATE", OP_MFX_VC1_PRED_PIPE_STATE, F_LEN_VAR,
2427 R_VCS, D_ALL, 0, 12, NULL},
2429 {"MFX_VC1_DIRECTMODE_STATE", OP_MFX_VC1_DIRECTMODE_STATE, F_LEN_VAR,
2430 R_VCS, D_ALL, 0, 12, NULL},
2432 {"MFD_VC1_SHORT_PIC_STATE", OP_MFD_VC1_SHORT_PIC_STATE, F_LEN_VAR,
2433 R_VCS, D_ALL, 0, 12, NULL},
2435 {"MFD_VC1_LONG_PIC_STATE", OP_MFD_VC1_LONG_PIC_STATE, F_LEN_VAR,
2436 R_VCS, D_ALL, 0, 12, NULL},
2438 {"MFD_VC1_BSD_OBJECT", OP_MFD_VC1_BSD_OBJECT, F_LEN_VAR,
2439 R_VCS, D_ALL, 0, 12, NULL},
2441 {"MFC_MPEG2_SLICEGROUP_STATE", OP_MFC_MPEG2_SLICEGROUP_STATE, F_LEN_VAR,
2442 R_VCS, D_ALL, 0, 12, NULL},
2444 {"MFC_MPEG2_PAK_OBJECT", OP_MFC_MPEG2_PAK_OBJECT, F_LEN_VAR,
2445 R_VCS, D_ALL, 0, 12, NULL},
2447 {"MFX_MPEG2_PIC_STATE", OP_MFX_MPEG2_PIC_STATE, F_LEN_VAR,
2448 R_VCS, D_ALL, 0, 12, NULL},
2450 {"MFX_MPEG2_QM_STATE", OP_MFX_MPEG2_QM_STATE, F_LEN_VAR,
2451 R_VCS, D_ALL, 0, 12, NULL},
2453 {"MFD_MPEG2_BSD_OBJECT", OP_MFD_MPEG2_BSD_OBJECT, F_LEN_VAR,
2454 R_VCS, D_ALL, 0, 12, NULL},
2456 {"MFX_2_6_0_0", OP_MFX_2_6_0_0, F_LEN_VAR, R_VCS, D_ALL,
2459 {"MFX_2_6_0_9", OP_MFX_2_6_0_9, F_LEN_VAR, R_VCS, D_ALL, 0, 16, NULL},
2461 {"MFX_2_6_0_8", OP_MFX_2_6_0_8, F_LEN_VAR, R_VCS, D_ALL, 0, 16, NULL},
2463 {"MFX_JPEG_PIC_STATE", OP_MFX_JPEG_PIC_STATE, F_LEN_VAR,
2464 R_VCS, D_ALL, 0, 12, NULL},
2466 {"MFX_JPEG_HUFF_TABLE_STATE", OP_MFX_JPEG_HUFF_TABLE_STATE, F_LEN_VAR,
2467 R_VCS, D_ALL, 0, 12, NULL},
2469 {"MFD_JPEG_BSD_OBJECT", OP_MFD_JPEG_BSD_OBJECT, F_LEN_VAR,
2470 R_VCS, D_ALL, 0, 12, NULL},
2472 {"VEBOX_STATE", OP_VEB_STATE, F_LEN_VAR, R_VECS, D_ALL, 0, 12, NULL},
2474 {"VEBOX_SURFACE_STATE", OP_VEB_SURFACE_STATE, F_LEN_VAR, R_VECS, D_ALL,
2477 {"VEB_DI_IECP", OP_VEB_DNDI_IECP_STATE, F_LEN_VAR, R_VECS, D_BDW_PLUS,
2481 static void add_cmd_entry(struct intel_gvt *gvt, struct cmd_entry *e)
2483 hash_add(gvt->cmd_table, &e->hlist, e->info->opcode);
2486 /* call the cmd handler, and advance ip */
2487 static int cmd_parser_exec(struct parser_exec_state *s)
2489 struct intel_vgpu *vgpu = s->vgpu;
2490 struct cmd_info *info;
2494 cmd = cmd_val(s, 0);
2496 info = get_cmd_info(s->vgpu->gvt, cmd, s->ring_id);
2498 gvt_vgpu_err("unknown cmd 0x%x, opcode=0x%x, addr_type=%s, ring %d, workload=%p\n",
2499 cmd, get_opcode(cmd, s->ring_id),
2500 (s->buf_addr_type == PPGTT_BUFFER) ?
2501 "ppgtt" : "ggtt", s->ring_id, s->workload);
2507 trace_gvt_command(vgpu->id, s->ring_id, s->ip_gma, s->ip_va,
2508 cmd_length(s), s->buf_type, s->buf_addr_type,
2509 s->workload, info->name);
2511 if (info->handler) {
2512 ret = info->handler(s);
2514 gvt_vgpu_err("%s handler error\n", info->name);
2519 if (!(info->flag & F_IP_ADVANCE_CUSTOM)) {
2520 ret = cmd_advance_default(s);
2522 gvt_vgpu_err("%s IP advance error\n", info->name);
2529 static inline bool gma_out_of_range(unsigned long gma,
2530 unsigned long gma_head, unsigned int gma_tail)
2532 if (gma_tail >= gma_head)
2533 return (gma < gma_head) || (gma > gma_tail);
2535 return (gma > gma_tail) && (gma < gma_head);
2538 /* Keep the consistent return type, e.g EBADRQC for unknown
2539 * cmd, EFAULT for invalid address, EPERM for nonpriv. later
2540 * works as the input of VM healthy status.
2542 static int command_scan(struct parser_exec_state *s,
2543 unsigned long rb_head, unsigned long rb_tail,
2544 unsigned long rb_start, unsigned long rb_len)
2547 unsigned long gma_head, gma_tail, gma_bottom;
2549 struct intel_vgpu *vgpu = s->vgpu;
2551 gma_head = rb_start + rb_head;
2552 gma_tail = rb_start + rb_tail;
2553 gma_bottom = rb_start + rb_len;
2555 while (s->ip_gma != gma_tail) {
2556 if (s->buf_type == RING_BUFFER_INSTRUCTION) {
2557 if (!(s->ip_gma >= rb_start) ||
2558 !(s->ip_gma < gma_bottom)) {
2559 gvt_vgpu_err("ip_gma %lx out of ring scope."
2560 "(base:0x%lx, bottom: 0x%lx)\n",
2561 s->ip_gma, rb_start,
2563 parser_exec_state_dump(s);
2566 if (gma_out_of_range(s->ip_gma, gma_head, gma_tail)) {
2567 gvt_vgpu_err("ip_gma %lx out of range."
2568 "base 0x%lx head 0x%lx tail 0x%lx\n",
2569 s->ip_gma, rb_start,
2571 parser_exec_state_dump(s);
2575 ret = cmd_parser_exec(s);
2577 gvt_vgpu_err("cmd parser error\n");
2578 parser_exec_state_dump(s);
2586 static int scan_workload(struct intel_vgpu_workload *workload)
2588 unsigned long gma_head, gma_tail, gma_bottom;
2589 struct parser_exec_state s;
2592 /* ring base is page aligned */
2593 if (WARN_ON(!IS_ALIGNED(workload->rb_start, I915_GTT_PAGE_SIZE)))
2596 gma_head = workload->rb_start + workload->rb_head;
2597 gma_tail = workload->rb_start + workload->rb_tail;
2598 gma_bottom = workload->rb_start + _RING_CTL_BUF_SIZE(workload->rb_ctl);
2600 s.buf_type = RING_BUFFER_INSTRUCTION;
2601 s.buf_addr_type = GTT_BUFFER;
2602 s.vgpu = workload->vgpu;
2603 s.ring_id = workload->ring_id;
2604 s.ring_start = workload->rb_start;
2605 s.ring_size = _RING_CTL_BUF_SIZE(workload->rb_ctl);
2606 s.ring_head = gma_head;
2607 s.ring_tail = gma_tail;
2608 s.rb_va = workload->shadow_ring_buffer_va;
2609 s.workload = workload;
2610 s.is_ctx_wa = false;
2612 if ((bypass_scan_mask & (1 << workload->ring_id)) ||
2613 gma_head == gma_tail)
2616 if (!intel_gvt_ggtt_validate_range(s.vgpu, s.ring_start, s.ring_size)) {
2621 ret = ip_gma_set(&s, gma_head);
2625 ret = command_scan(&s, workload->rb_head, workload->rb_tail,
2626 workload->rb_start, _RING_CTL_BUF_SIZE(workload->rb_ctl));
2632 static int scan_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
2635 unsigned long gma_head, gma_tail, gma_bottom, ring_size, ring_tail;
2636 struct parser_exec_state s;
2638 struct intel_vgpu_workload *workload = container_of(wa_ctx,
2639 struct intel_vgpu_workload,
2642 /* ring base is page aligned */
2643 if (WARN_ON(!IS_ALIGNED(wa_ctx->indirect_ctx.guest_gma,
2644 I915_GTT_PAGE_SIZE)))
2647 ring_tail = wa_ctx->indirect_ctx.size + 3 * sizeof(uint32_t);
2648 ring_size = round_up(wa_ctx->indirect_ctx.size + CACHELINE_BYTES,
2650 gma_head = wa_ctx->indirect_ctx.guest_gma;
2651 gma_tail = wa_ctx->indirect_ctx.guest_gma + ring_tail;
2652 gma_bottom = wa_ctx->indirect_ctx.guest_gma + ring_size;
2654 s.buf_type = RING_BUFFER_INSTRUCTION;
2655 s.buf_addr_type = GTT_BUFFER;
2656 s.vgpu = workload->vgpu;
2657 s.ring_id = workload->ring_id;
2658 s.ring_start = wa_ctx->indirect_ctx.guest_gma;
2659 s.ring_size = ring_size;
2660 s.ring_head = gma_head;
2661 s.ring_tail = gma_tail;
2662 s.rb_va = wa_ctx->indirect_ctx.shadow_va;
2663 s.workload = workload;
2666 if (!intel_gvt_ggtt_validate_range(s.vgpu, s.ring_start, s.ring_size)) {
2671 ret = ip_gma_set(&s, gma_head);
2675 ret = command_scan(&s, 0, ring_tail,
2676 wa_ctx->indirect_ctx.guest_gma, ring_size);
2681 static int shadow_workload_ring_buffer(struct intel_vgpu_workload *workload)
2683 struct intel_vgpu *vgpu = workload->vgpu;
2684 struct intel_vgpu_submission *s = &vgpu->submission;
2685 unsigned long gma_head, gma_tail, gma_top, guest_rb_size;
2686 void *shadow_ring_buffer_va;
2687 int ring_id = workload->ring_id;
2690 guest_rb_size = _RING_CTL_BUF_SIZE(workload->rb_ctl);
2692 /* calculate workload ring buffer size */
2693 workload->rb_len = (workload->rb_tail + guest_rb_size -
2694 workload->rb_head) % guest_rb_size;
2696 gma_head = workload->rb_start + workload->rb_head;
2697 gma_tail = workload->rb_start + workload->rb_tail;
2698 gma_top = workload->rb_start + guest_rb_size;
2700 if (workload->rb_len > s->ring_scan_buffer_size[ring_id]) {
2703 /* realloc the new ring buffer if needed */
2704 p = krealloc(s->ring_scan_buffer[ring_id], workload->rb_len,
2707 gvt_vgpu_err("fail to re-alloc ring scan buffer\n");
2710 s->ring_scan_buffer[ring_id] = p;
2711 s->ring_scan_buffer_size[ring_id] = workload->rb_len;
2714 shadow_ring_buffer_va = s->ring_scan_buffer[ring_id];
2716 /* get shadow ring buffer va */
2717 workload->shadow_ring_buffer_va = shadow_ring_buffer_va;
2719 /* head > tail --> copy head <-> top */
2720 if (gma_head > gma_tail) {
2721 ret = copy_gma_to_hva(vgpu, vgpu->gtt.ggtt_mm,
2722 gma_head, gma_top, shadow_ring_buffer_va);
2724 gvt_vgpu_err("fail to copy guest ring buffer\n");
2727 shadow_ring_buffer_va += ret;
2728 gma_head = workload->rb_start;
2731 /* copy head or start <-> tail */
2732 ret = copy_gma_to_hva(vgpu, vgpu->gtt.ggtt_mm, gma_head, gma_tail,
2733 shadow_ring_buffer_va);
2735 gvt_vgpu_err("fail to copy guest ring buffer\n");
2741 int intel_gvt_scan_and_shadow_ringbuffer(struct intel_vgpu_workload *workload)
2744 struct intel_vgpu *vgpu = workload->vgpu;
2746 ret = shadow_workload_ring_buffer(workload);
2748 gvt_vgpu_err("fail to shadow workload ring_buffer\n");
2752 ret = scan_workload(workload);
2754 gvt_vgpu_err("scan workload error\n");
2760 static int shadow_indirect_ctx(struct intel_shadow_wa_ctx *wa_ctx)
2762 int ctx_size = wa_ctx->indirect_ctx.size;
2763 unsigned long guest_gma = wa_ctx->indirect_ctx.guest_gma;
2764 struct intel_vgpu_workload *workload = container_of(wa_ctx,
2765 struct intel_vgpu_workload,
2767 struct intel_vgpu *vgpu = workload->vgpu;
2768 struct drm_i915_gem_object *obj;
2772 obj = i915_gem_object_create(workload->vgpu->gvt->dev_priv,
2773 roundup(ctx_size + CACHELINE_BYTES,
2776 return PTR_ERR(obj);
2778 /* get the va of the shadow batch buffer */
2779 map = i915_gem_object_pin_map(obj, I915_MAP_WB);
2781 gvt_vgpu_err("failed to vmap shadow indirect ctx\n");
2786 ret = i915_gem_object_set_to_cpu_domain(obj, false);
2788 gvt_vgpu_err("failed to set shadow indirect ctx to CPU\n");
2792 ret = copy_gma_to_hva(workload->vgpu,
2793 workload->vgpu->gtt.ggtt_mm,
2794 guest_gma, guest_gma + ctx_size,
2797 gvt_vgpu_err("fail to copy guest indirect ctx\n");
2801 wa_ctx->indirect_ctx.obj = obj;
2802 wa_ctx->indirect_ctx.shadow_va = map;
2806 i915_gem_object_unpin_map(obj);
2808 i915_gem_object_put(obj);
2812 static int combine_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
2814 uint32_t per_ctx_start[CACHELINE_DWORDS] = {0};
2815 unsigned char *bb_start_sva;
2817 if (!wa_ctx->per_ctx.valid)
2820 per_ctx_start[0] = 0x18800001;
2821 per_ctx_start[1] = wa_ctx->per_ctx.guest_gma;
2823 bb_start_sva = (unsigned char *)wa_ctx->indirect_ctx.shadow_va +
2824 wa_ctx->indirect_ctx.size;
2826 memcpy(bb_start_sva, per_ctx_start, CACHELINE_BYTES);
2831 int intel_gvt_scan_and_shadow_wa_ctx(struct intel_shadow_wa_ctx *wa_ctx)
2834 struct intel_vgpu_workload *workload = container_of(wa_ctx,
2835 struct intel_vgpu_workload,
2837 struct intel_vgpu *vgpu = workload->vgpu;
2839 if (wa_ctx->indirect_ctx.size == 0)
2842 ret = shadow_indirect_ctx(wa_ctx);
2844 gvt_vgpu_err("fail to shadow indirect ctx\n");
2848 combine_wa_ctx(wa_ctx);
2850 ret = scan_wa_ctx(wa_ctx);
2852 gvt_vgpu_err("scan wa ctx error\n");
2859 static struct cmd_info *find_cmd_entry_any_ring(struct intel_gvt *gvt,
2860 unsigned int opcode, unsigned long rings)
2862 struct cmd_info *info = NULL;
2865 for_each_set_bit(ring, &rings, I915_NUM_ENGINES) {
2866 info = find_cmd_entry(gvt, opcode, ring);
2873 static int init_cmd_table(struct intel_gvt *gvt)
2876 struct cmd_entry *e;
2877 struct cmd_info *info;
2878 unsigned int gen_type;
2880 gen_type = intel_gvt_get_device_type(gvt);
2882 for (i = 0; i < ARRAY_SIZE(cmd_info); i++) {
2883 if (!(cmd_info[i].devices & gen_type))
2886 e = kzalloc(sizeof(*e), GFP_KERNEL);
2890 e->info = &cmd_info[i];
2891 info = find_cmd_entry_any_ring(gvt,
2892 e->info->opcode, e->info->rings);
2894 gvt_err("%s %s duplicated\n", e->info->name,
2899 INIT_HLIST_NODE(&e->hlist);
2900 add_cmd_entry(gvt, e);
2901 gvt_dbg_cmd("add %-30s op %04x flag %x devs %02x rings %02x\n",
2902 e->info->name, e->info->opcode, e->info->flag,
2903 e->info->devices, e->info->rings);
2908 static void clean_cmd_table(struct intel_gvt *gvt)
2910 struct hlist_node *tmp;
2911 struct cmd_entry *e;
2914 hash_for_each_safe(gvt->cmd_table, i, tmp, e, hlist)
2917 hash_init(gvt->cmd_table);
2920 void intel_gvt_clean_cmd_parser(struct intel_gvt *gvt)
2922 clean_cmd_table(gvt);
2925 int intel_gvt_init_cmd_parser(struct intel_gvt *gvt)
2929 ret = init_cmd_table(gvt);
2931 intel_gvt_clean_cmd_parser(gvt);