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_ENC_CMD_NO_OP 0x00000000 48 #define VCN_ENC_CMD_END 0x00000001 49 #define VCN_ENC_CMD_IB 0x00000002 50 #define VCN_ENC_CMD_FENCE 0x00000003 51 #define VCN_ENC_CMD_TRAP 0x00000004 52 #define VCN_ENC_CMD_REG_WRITE 0x0000000b 53 #define VCN_ENC_CMD_REG_WAIT 0x0000000c 54 55 #define VCN_VID_SOC_ADDRESS_2_0 0x1fa00 56 #define VCN1_VID_SOC_ADDRESS_3_0 0x48200 57 #define VCN_AON_SOC_ADDRESS_2_0 0x1f800 58 #define VCN1_AON_SOC_ADDRESS_3_0 0x48000 59 #define VCN_VID_IP_ADDRESS_2_0 0x0 60 #define VCN_AON_IP_ADDRESS_2_0 0x30000 61 62 #define mmUVD_RBC_XX_IB_REG_CHECK 0x026b 63 #define mmUVD_RBC_XX_IB_REG_CHECK_BASE_IDX 1 64 #define mmUVD_REG_XX_MASK 0x026c 65 #define mmUVD_REG_XX_MASK_BASE_IDX 1 66 67 /* 1 second timeout */ 68 #define VCN_IDLE_TIMEOUT msecs_to_jiffies(1000) 69 70 #define RREG32_SOC15_DPG_MODE_1_0(ip, inst_idx, reg, mask, sram_sel) \ 71 ({ WREG32_SOC15(ip, inst_idx, mmUVD_DPG_LMA_MASK, mask); \ 72 WREG32_SOC15(ip, inst_idx, mmUVD_DPG_LMA_CTL, \ 73 UVD_DPG_LMA_CTL__MASK_EN_MASK | \ 74 ((adev->reg_offset[ip##_HWIP][inst_idx][reg##_BASE_IDX] + reg) \ 75 << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT) | \ 76 (sram_sel << UVD_DPG_LMA_CTL__SRAM_SEL__SHIFT)); \ 77 RREG32_SOC15(ip, inst_idx, mmUVD_DPG_LMA_DATA); \ 78 }) 79 80 #define WREG32_SOC15_DPG_MODE_1_0(ip, inst_idx, reg, value, mask, sram_sel) \ 81 do { \ 82 WREG32_SOC15(ip, inst_idx, mmUVD_DPG_LMA_DATA, value); \ 83 WREG32_SOC15(ip, inst_idx, mmUVD_DPG_LMA_MASK, mask); \ 84 WREG32_SOC15(ip, inst_idx, mmUVD_DPG_LMA_CTL, \ 85 UVD_DPG_LMA_CTL__READ_WRITE_MASK | \ 86 ((adev->reg_offset[ip##_HWIP][inst_idx][reg##_BASE_IDX] + reg) \ 87 << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT) | \ 88 (sram_sel << UVD_DPG_LMA_CTL__SRAM_SEL__SHIFT)); \ 89 } while (0) 90 91 #define SOC15_DPG_MODE_OFFSET(ip, inst_idx, reg) \ 92 ({ \ 93 uint32_t internal_reg_offset, addr; \ 94 bool video_range, video1_range, aon_range, aon1_range; \ 95 \ 96 addr = (adev->reg_offset[ip##_HWIP][inst_idx][reg##_BASE_IDX] + reg); \ 97 addr <<= 2; \ 98 video_range = ((((0xFFFFF & addr) >= (VCN_VID_SOC_ADDRESS_2_0)) && \ 99 ((0xFFFFF & addr) < ((VCN_VID_SOC_ADDRESS_2_0 + 0x2600))))); \ 100 video1_range = ((((0xFFFFF & addr) >= (VCN1_VID_SOC_ADDRESS_3_0)) && \ 101 ((0xFFFFF & addr) < ((VCN1_VID_SOC_ADDRESS_3_0 + 0x2600))))); \ 102 aon_range = ((((0xFFFFF & addr) >= (VCN_AON_SOC_ADDRESS_2_0)) && \ 103 ((0xFFFFF & addr) < ((VCN_AON_SOC_ADDRESS_2_0 + 0x600))))); \ 104 aon1_range = ((((0xFFFFF & addr) >= (VCN1_AON_SOC_ADDRESS_3_0)) && \ 105 ((0xFFFFF & addr) < ((VCN1_AON_SOC_ADDRESS_3_0 + 0x600))))); \ 106 if (video_range) \ 107 internal_reg_offset = ((0xFFFFF & addr) - (VCN_VID_SOC_ADDRESS_2_0) + \ 108 (VCN_VID_IP_ADDRESS_2_0)); \ 109 else if (aon_range) \ 110 internal_reg_offset = ((0xFFFFF & addr) - (VCN_AON_SOC_ADDRESS_2_0) + \ 111 (VCN_AON_IP_ADDRESS_2_0)); \ 112 else if (video1_range) \ 113 internal_reg_offset = ((0xFFFFF & addr) - (VCN1_VID_SOC_ADDRESS_3_0) + \ 114 (VCN_VID_IP_ADDRESS_2_0)); \ 115 else if (aon1_range) \ 116 internal_reg_offset = ((0xFFFFF & addr) - (VCN1_AON_SOC_ADDRESS_3_0) + \ 117 (VCN_AON_IP_ADDRESS_2_0)); \ 118 else \ 119 internal_reg_offset = (0xFFFFF & addr); \ 120 \ 121 internal_reg_offset >>= 2; \ 122 }) 123 124 #define RREG32_SOC15_DPG_MODE(inst_idx, offset, mask_en) \ 125 ({ \ 126 WREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_CTL, \ 127 (0x0 << UVD_DPG_LMA_CTL__READ_WRITE__SHIFT | \ 128 mask_en << UVD_DPG_LMA_CTL__MASK_EN__SHIFT | \ 129 offset << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT)); \ 130 RREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_DATA); \ 131 }) 132 133 #define WREG32_SOC15_DPG_MODE(inst_idx, offset, value, mask_en, indirect) \ 134 do { \ 135 if (!indirect) { \ 136 WREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_DATA, value); \ 137 WREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_CTL, \ 138 (0x1 << 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 } else { \ 142 *adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ = offset; \ 143 *adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ = value; \ 144 } \ 145 } while (0) 146 147 #define AMDGPU_VCN_MULTI_QUEUE_FLAG (1 << 8) 148 149 enum fw_queue_mode { 150 FW_QUEUE_RING_RESET = 1, 151 FW_QUEUE_DPG_HOLD_OFF = 2, 152 }; 153 154 enum engine_status_constants { 155 UVD_PGFSM_STATUS__UVDM_UVDU_PWR_ON = 0x2AAAA0, 156 UVD_PGFSM_STATUS__UVDM_UVDU_PWR_ON_2_0 = 0xAAAA0, 157 UVD_PGFSM_STATUS__UVDM_UVDU_UVDLM_PWR_ON_3_0 = 0x2A2A8AA0, 158 UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON = 0x00000002, 159 UVD_STATUS__UVD_BUSY = 0x00000004, 160 GB_ADDR_CONFIG_DEFAULT = 0x26010011, 161 UVD_STATUS__IDLE = 0x2, 162 UVD_STATUS__BUSY = 0x5, 163 UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF = 0x1, 164 UVD_STATUS__RBC_BUSY = 0x1, 165 UVD_PGFSM_STATUS_UVDJ_PWR_ON = 0, 166 }; 167 168 enum internal_dpg_state { 169 VCN_DPG_STATE__UNPAUSE = 0, 170 VCN_DPG_STATE__PAUSE, 171 }; 172 173 struct dpg_pause_state { 174 enum internal_dpg_state fw_based; 175 enum internal_dpg_state jpeg; 176 }; 177 178 struct amdgpu_vcn_reg{ 179 unsigned data0; 180 unsigned data1; 181 unsigned cmd; 182 unsigned nop; 183 unsigned context_id; 184 unsigned ib_vmid; 185 unsigned ib_bar_low; 186 unsigned ib_bar_high; 187 unsigned ib_size; 188 unsigned gp_scratch8; 189 unsigned scratch9; 190 }; 191 192 struct amdgpu_vcn_inst { 193 struct amdgpu_bo *vcpu_bo; 194 void *cpu_addr; 195 uint64_t gpu_addr; 196 void *saved_bo; 197 struct amdgpu_ring ring_dec; 198 struct amdgpu_ring ring_enc[AMDGPU_VCN_MAX_ENC_RINGS]; 199 struct amdgpu_irq_src irq; 200 struct amdgpu_vcn_reg external; 201 struct amdgpu_bo *dpg_sram_bo; 202 struct dpg_pause_state pause_state; 203 void *dpg_sram_cpu_addr; 204 uint64_t dpg_sram_gpu_addr; 205 uint32_t *dpg_sram_curr_addr; 206 atomic_t dpg_enc_submission_cnt; 207 void *fw_shared_cpu_addr; 208 uint64_t fw_shared_gpu_addr; 209 }; 210 211 struct amdgpu_vcn { 212 unsigned fw_version; 213 struct delayed_work idle_work; 214 const struct firmware *fw; /* VCN firmware */ 215 unsigned num_enc_rings; 216 enum amd_powergating_state cur_state; 217 bool indirect_sram; 218 219 uint8_t num_vcn_inst; 220 struct amdgpu_vcn_inst inst[AMDGPU_MAX_VCN_INSTANCES]; 221 struct amdgpu_vcn_reg internal; 222 struct mutex vcn_pg_lock; 223 struct mutex vcn1_jpeg1_workaround; 224 atomic_t total_submission_cnt; 225 226 unsigned harvest_config; 227 int (*pause_dpg_mode)(struct amdgpu_device *adev, 228 int inst_idx, struct dpg_pause_state *new_state); 229 }; 230 231 struct amdgpu_fw_shared_multi_queue { 232 uint8_t decode_queue_mode; 233 uint8_t encode_generalpurpose_queue_mode; 234 uint8_t encode_lowlatency_queue_mode; 235 uint8_t encode_realtime_queue_mode; 236 uint8_t padding[4]; 237 }; 238 239 struct amdgpu_fw_shared { 240 uint32_t present_flag_0; 241 uint8_t pad[53]; 242 struct amdgpu_fw_shared_multi_queue multi_queue; 243 } __attribute__((__packed__)); 244 245 int amdgpu_vcn_sw_init(struct amdgpu_device *adev); 246 int amdgpu_vcn_sw_fini(struct amdgpu_device *adev); 247 int amdgpu_vcn_suspend(struct amdgpu_device *adev); 248 int amdgpu_vcn_resume(struct amdgpu_device *adev); 249 void amdgpu_vcn_ring_begin_use(struct amdgpu_ring *ring); 250 void amdgpu_vcn_ring_end_use(struct amdgpu_ring *ring); 251 252 int amdgpu_vcn_dec_ring_test_ring(struct amdgpu_ring *ring); 253 int amdgpu_vcn_dec_ring_test_ib(struct amdgpu_ring *ring, long timeout); 254 255 int amdgpu_vcn_enc_ring_test_ring(struct amdgpu_ring *ring); 256 int amdgpu_vcn_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout); 257 258 #endif 259
