1 /* 2 * Copyright 2016 Advanced Micro Devices, Inc. 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 shall be included in 12 * all copies or substantial portions of the Software. 13 * 14 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 15 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 16 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 17 * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR 18 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, 19 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR 20 * OTHER DEALINGS IN THE SOFTWARE. 21 * 22 */ 23 24 #ifndef __AMDGPU_VCN_H__ 25 #define __AMDGPU_VCN_H__ 26 27 #define AMDGPU_VCN_STACK_SIZE (128*1024) 28 #define AMDGPU_VCN_CONTEXT_SIZE (512*1024) 29 30 #define AMDGPU_VCN_FIRMWARE_OFFSET 256 31 #define AMDGPU_VCN_MAX_ENC_RINGS 3 32 33 #define AMDGPU_MAX_VCN_INSTANCES 2 34 #define AMDGPU_MAX_VCN_ENC_RINGS AMDGPU_VCN_MAX_ENC_RINGS * AMDGPU_MAX_VCN_INSTANCES 35 36 #define AMDGPU_VCN_HARVEST_VCN0 (1 << 0) 37 #define AMDGPU_VCN_HARVEST_VCN1 (1 << 1) 38 39 #define VCN_DEC_KMD_CMD 0x80000000 40 #define VCN_DEC_CMD_FENCE 0x00000000 41 #define VCN_DEC_CMD_TRAP 0x00000001 42 #define VCN_DEC_CMD_WRITE_REG 0x00000004 43 #define VCN_DEC_CMD_REG_READ_COND_WAIT 0x00000006 44 #define VCN_DEC_CMD_PACKET_START 0x0000000a 45 #define VCN_DEC_CMD_PACKET_END 0x0000000b 46 47 #define VCN_DEC_SW_CMD_NO_OP 0x00000000 48 #define VCN_DEC_SW_CMD_END 0x00000001 49 #define VCN_DEC_SW_CMD_IB 0x00000002 50 #define VCN_DEC_SW_CMD_FENCE 0x00000003 51 #define VCN_DEC_SW_CMD_TRAP 0x00000004 52 #define VCN_DEC_SW_CMD_IB_AUTO 0x00000005 53 #define VCN_DEC_SW_CMD_SEMAPHORE 0x00000006 54 #define VCN_DEC_SW_CMD_PREEMPT_FENCE 0x00000009 55 #define VCN_DEC_SW_CMD_REG_WRITE 0x0000000b 56 #define VCN_DEC_SW_CMD_REG_WAIT 0x0000000c 57 58 #define VCN_ENC_CMD_NO_OP 0x00000000 59 #define VCN_ENC_CMD_END 0x00000001 60 #define VCN_ENC_CMD_IB 0x00000002 61 #define VCN_ENC_CMD_FENCE 0x00000003 62 #define VCN_ENC_CMD_TRAP 0x00000004 63 #define VCN_ENC_CMD_REG_WRITE 0x0000000b 64 #define VCN_ENC_CMD_REG_WAIT 0x0000000c 65 66 #define VCN_VID_SOC_ADDRESS_2_0 0x1fa00 67 #define VCN1_VID_SOC_ADDRESS_3_0 0x48200 68 #define VCN_AON_SOC_ADDRESS_2_0 0x1f800 69 #define VCN1_AON_SOC_ADDRESS_3_0 0x48000 70 #define VCN_VID_IP_ADDRESS_2_0 0x0 71 #define VCN_AON_IP_ADDRESS_2_0 0x30000 72 73 #define mmUVD_RBC_XX_IB_REG_CHECK 0x026b 74 #define mmUVD_RBC_XX_IB_REG_CHECK_BASE_IDX 1 75 #define mmUVD_REG_XX_MASK 0x026c 76 #define mmUVD_REG_XX_MASK_BASE_IDX 1 77 78 /* 1 second timeout */ 79 #define VCN_IDLE_TIMEOUT msecs_to_jiffies(1000) 80 81 #define RREG32_SOC15_DPG_MODE_1_0(ip, inst_idx, reg, mask, sram_sel) \ 82 ({ WREG32_SOC15(ip, inst_idx, mmUVD_DPG_LMA_MASK, mask); \ 83 WREG32_SOC15(ip, inst_idx, mmUVD_DPG_LMA_CTL, \ 84 UVD_DPG_LMA_CTL__MASK_EN_MASK | \ 85 ((adev->reg_offset[ip##_HWIP][inst_idx][reg##_BASE_IDX] + reg) \ 86 << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT) | \ 87 (sram_sel << UVD_DPG_LMA_CTL__SRAM_SEL__SHIFT)); \ 88 RREG32_SOC15(ip, inst_idx, mmUVD_DPG_LMA_DATA); \ 89 }) 90 91 #define WREG32_SOC15_DPG_MODE_1_0(ip, inst_idx, reg, value, mask, sram_sel) \ 92 do { \ 93 WREG32_SOC15(ip, inst_idx, mmUVD_DPG_LMA_DATA, value); \ 94 WREG32_SOC15(ip, inst_idx, mmUVD_DPG_LMA_MASK, mask); \ 95 WREG32_SOC15(ip, inst_idx, mmUVD_DPG_LMA_CTL, \ 96 UVD_DPG_LMA_CTL__READ_WRITE_MASK | \ 97 ((adev->reg_offset[ip##_HWIP][inst_idx][reg##_BASE_IDX] + reg) \ 98 << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT) | \ 99 (sram_sel << UVD_DPG_LMA_CTL__SRAM_SEL__SHIFT)); \ 100 } while (0) 101 102 #define SOC15_DPG_MODE_OFFSET(ip, inst_idx, reg) \ 103 ({ \ 104 uint32_t internal_reg_offset, addr; \ 105 bool video_range, video1_range, aon_range, aon1_range; \ 106 \ 107 addr = (adev->reg_offset[ip##_HWIP][inst_idx][reg##_BASE_IDX] + reg); \ 108 addr <<= 2; \ 109 video_range = ((((0xFFFFF & addr) >= (VCN_VID_SOC_ADDRESS_2_0)) && \ 110 ((0xFFFFF & addr) < ((VCN_VID_SOC_ADDRESS_2_0 + 0x2600))))); \ 111 video1_range = ((((0xFFFFF & addr) >= (VCN1_VID_SOC_ADDRESS_3_0)) && \ 112 ((0xFFFFF & addr) < ((VCN1_VID_SOC_ADDRESS_3_0 + 0x2600))))); \ 113 aon_range = ((((0xFFFFF & addr) >= (VCN_AON_SOC_ADDRESS_2_0)) && \ 114 ((0xFFFFF & addr) < ((VCN_AON_SOC_ADDRESS_2_0 + 0x600))))); \ 115 aon1_range = ((((0xFFFFF & addr) >= (VCN1_AON_SOC_ADDRESS_3_0)) && \ 116 ((0xFFFFF & addr) < ((VCN1_AON_SOC_ADDRESS_3_0 + 0x600))))); \ 117 if (video_range) \ 118 internal_reg_offset = ((0xFFFFF & addr) - (VCN_VID_SOC_ADDRESS_2_0) + \ 119 (VCN_VID_IP_ADDRESS_2_0)); \ 120 else if (aon_range) \ 121 internal_reg_offset = ((0xFFFFF & addr) - (VCN_AON_SOC_ADDRESS_2_0) + \ 122 (VCN_AON_IP_ADDRESS_2_0)); \ 123 else if (video1_range) \ 124 internal_reg_offset = ((0xFFFFF & addr) - (VCN1_VID_SOC_ADDRESS_3_0) + \ 125 (VCN_VID_IP_ADDRESS_2_0)); \ 126 else if (aon1_range) \ 127 internal_reg_offset = ((0xFFFFF & addr) - (VCN1_AON_SOC_ADDRESS_3_0) + \ 128 (VCN_AON_IP_ADDRESS_2_0)); \ 129 else \ 130 internal_reg_offset = (0xFFFFF & addr); \ 131 \ 132 internal_reg_offset >>= 2; \ 133 }) 134 135 #define RREG32_SOC15_DPG_MODE(inst_idx, offset, mask_en) \ 136 ({ \ 137 WREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_CTL, \ 138 (0x0 << UVD_DPG_LMA_CTL__READ_WRITE__SHIFT | \ 139 mask_en << UVD_DPG_LMA_CTL__MASK_EN__SHIFT | \ 140 offset << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT)); \ 141 RREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_DATA); \ 142 }) 143 144 #define WREG32_SOC15_DPG_MODE(inst_idx, offset, value, mask_en, indirect) \ 145 do { \ 146 if (!indirect) { \ 147 WREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_DATA, value); \ 148 WREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_CTL, \ 149 (0x1 << UVD_DPG_LMA_CTL__READ_WRITE__SHIFT | \ 150 mask_en << UVD_DPG_LMA_CTL__MASK_EN__SHIFT | \ 151 offset << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT)); \ 152 } else { \ 153 *adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ = offset; \ 154 *adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ = value; \ 155 } \ 156 } while (0) 157 158 #define AMDGPU_VCN_MULTI_QUEUE_FLAG (1 << 8) 159 #define AMDGPU_VCN_SW_RING_FLAG (1 << 9) 160 161 #define AMDGPU_VCN_IB_FLAG_DECODE_BUFFER 0x00000001 162 #define AMDGPU_VCN_CMD_FLAG_MSG_BUFFER 0x00000001 163 164 enum fw_queue_mode { 165 FW_QUEUE_RING_RESET = 1, 166 FW_QUEUE_DPG_HOLD_OFF = 2, 167 }; 168 169 enum engine_status_constants { 170 UVD_PGFSM_STATUS__UVDM_UVDU_PWR_ON = 0x2AAAA0, 171 UVD_PGFSM_STATUS__UVDM_UVDU_PWR_ON_2_0 = 0xAAAA0, 172 UVD_PGFSM_STATUS__UVDM_UVDU_UVDLM_PWR_ON_3_0 = 0x2A2A8AA0, 173 UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON = 0x00000002, 174 UVD_STATUS__UVD_BUSY = 0x00000004, 175 GB_ADDR_CONFIG_DEFAULT = 0x26010011, 176 UVD_STATUS__IDLE = 0x2, 177 UVD_STATUS__BUSY = 0x5, 178 UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF = 0x1, 179 UVD_STATUS__RBC_BUSY = 0x1, 180 UVD_PGFSM_STATUS_UVDJ_PWR_ON = 0, 181 }; 182 183 enum internal_dpg_state { 184 VCN_DPG_STATE__UNPAUSE = 0, 185 VCN_DPG_STATE__PAUSE, 186 }; 187 188 struct dpg_pause_state { 189 enum internal_dpg_state fw_based; 190 enum internal_dpg_state jpeg; 191 }; 192 193 struct amdgpu_vcn_reg{ 194 unsigned data0; 195 unsigned data1; 196 unsigned cmd; 197 unsigned nop; 198 unsigned context_id; 199 unsigned ib_vmid; 200 unsigned ib_bar_low; 201 unsigned ib_bar_high; 202 unsigned ib_size; 203 unsigned gp_scratch8; 204 unsigned scratch9; 205 }; 206 207 struct amdgpu_vcn_inst { 208 struct amdgpu_bo *vcpu_bo; 209 void *cpu_addr; 210 uint64_t gpu_addr; 211 void *saved_bo; 212 struct amdgpu_ring ring_dec; 213 struct amdgpu_ring ring_enc[AMDGPU_VCN_MAX_ENC_RINGS]; 214 struct amdgpu_irq_src irq; 215 struct amdgpu_vcn_reg external; 216 struct amdgpu_bo *dpg_sram_bo; 217 struct dpg_pause_state pause_state; 218 void *dpg_sram_cpu_addr; 219 uint64_t dpg_sram_gpu_addr; 220 uint32_t *dpg_sram_curr_addr; 221 atomic_t dpg_enc_submission_cnt; 222 void *fw_shared_cpu_addr; 223 uint64_t fw_shared_gpu_addr; 224 }; 225 226 struct amdgpu_vcn { 227 unsigned fw_version; 228 struct delayed_work idle_work; 229 const struct firmware *fw; /* VCN firmware */ 230 unsigned num_enc_rings; 231 enum amd_powergating_state cur_state; 232 bool indirect_sram; 233 234 uint8_t num_vcn_inst; 235 struct amdgpu_vcn_inst inst[AMDGPU_MAX_VCN_INSTANCES]; 236 struct amdgpu_vcn_reg internal; 237 struct mutex vcn_pg_lock; 238 struct mutex vcn1_jpeg1_workaround; 239 atomic_t total_submission_cnt; 240 241 unsigned harvest_config; 242 int (*pause_dpg_mode)(struct amdgpu_device *adev, 243 int inst_idx, struct dpg_pause_state *new_state); 244 }; 245 246 struct amdgpu_fw_shared_multi_queue { 247 uint8_t decode_queue_mode; 248 uint8_t encode_generalpurpose_queue_mode; 249 uint8_t encode_lowlatency_queue_mode; 250 uint8_t encode_realtime_queue_mode; 251 uint8_t padding[4]; 252 }; 253 254 struct amdgpu_fw_shared_sw_ring { 255 uint8_t is_enabled; 256 uint8_t padding[3]; 257 }; 258 259 struct amdgpu_fw_shared { 260 uint32_t present_flag_0; 261 uint8_t pad[53]; 262 struct amdgpu_fw_shared_multi_queue multi_queue; 263 struct amdgpu_fw_shared_sw_ring sw_ring; 264 } __attribute__((__packed__)); 265 266 struct amdgpu_vcn_decode_buffer { 267 uint32_t valid_buf_flag; 268 uint32_t msg_buffer_address_hi; 269 uint32_t msg_buffer_address_lo; 270 uint32_t pad[30]; 271 }; 272 273 int amdgpu_vcn_sw_init(struct amdgpu_device *adev); 274 int amdgpu_vcn_sw_fini(struct amdgpu_device *adev); 275 int amdgpu_vcn_suspend(struct amdgpu_device *adev); 276 int amdgpu_vcn_resume(struct amdgpu_device *adev); 277 void amdgpu_vcn_ring_begin_use(struct amdgpu_ring *ring); 278 void amdgpu_vcn_ring_end_use(struct amdgpu_ring *ring); 279 280 int amdgpu_vcn_dec_ring_test_ring(struct amdgpu_ring *ring); 281 int amdgpu_vcn_dec_ring_test_ib(struct amdgpu_ring *ring, long timeout); 282 int amdgpu_vcn_dec_sw_ring_test_ring(struct amdgpu_ring *ring); 283 int amdgpu_vcn_dec_sw_ring_test_ib(struct amdgpu_ring *ring, long timeout); 284 285 int amdgpu_vcn_enc_ring_test_ring(struct amdgpu_ring *ring); 286 int amdgpu_vcn_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout); 287 288 #endif 289