1 /* $NetBSD: amdgpu_vcn.h,v 1.2 2021/12/18 23:44:58 riastradh Exp $ */ 2 3 /* 4 * Copyright 2016 Advanced Micro Devices, Inc. 5 * 6 * Permission is hereby granted, free of charge, to any person obtaining a 7 * copy of this software and associated documentation files (the "Software"), 8 * to deal in the Software without restriction, including without limitation 9 * the rights to use, copy, modify, merge, publish, distribute, sublicense, 10 * and/or sell copies of the Software, and to permit persons to whom the 11 * Software is furnished to do so, subject to the following conditions: 12 * 13 * The above copyright notice and this permission notice shall be included in 14 * all copies or substantial portions of the Software. 15 * 16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 19 * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR 20 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, 21 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR 22 * OTHER DEALINGS IN THE SOFTWARE. 23 * 24 */ 25 26 #ifndef __AMDGPU_VCN_H__ 27 #define __AMDGPU_VCN_H__ 28 29 #define AMDGPU_VCN_STACK_SIZE (128*1024) 30 #define AMDGPU_VCN_CONTEXT_SIZE (512*1024) 31 32 #define AMDGPU_VCN_FIRMWARE_OFFSET 256 33 #define AMDGPU_VCN_MAX_ENC_RINGS 3 34 35 #define AMDGPU_MAX_VCN_INSTANCES 2 36 #define AMDGPU_MAX_VCN_ENC_RINGS AMDGPU_VCN_MAX_ENC_RINGS * AMDGPU_MAX_VCN_INSTANCES 37 38 #define AMDGPU_VCN_HARVEST_VCN0 (1 << 0) 39 #define AMDGPU_VCN_HARVEST_VCN1 (1 << 1) 40 41 #define VCN_DEC_KMD_CMD 0x80000000 42 #define VCN_DEC_CMD_FENCE 0x00000000 43 #define VCN_DEC_CMD_TRAP 0x00000001 44 #define VCN_DEC_CMD_WRITE_REG 0x00000004 45 #define VCN_DEC_CMD_REG_READ_COND_WAIT 0x00000006 46 #define VCN_DEC_CMD_PACKET_START 0x0000000a 47 #define VCN_DEC_CMD_PACKET_END 0x0000000b 48 49 #define VCN_ENC_CMD_NO_OP 0x00000000 50 #define VCN_ENC_CMD_END 0x00000001 51 #define VCN_ENC_CMD_IB 0x00000002 52 #define VCN_ENC_CMD_FENCE 0x00000003 53 #define VCN_ENC_CMD_TRAP 0x00000004 54 #define VCN_ENC_CMD_REG_WRITE 0x0000000b 55 #define VCN_ENC_CMD_REG_WAIT 0x0000000c 56 57 #define VCN_VID_SOC_ADDRESS_2_0 0x1fa00 58 #define VCN_AON_SOC_ADDRESS_2_0 0x1f800 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(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(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_2_0(ip, inst_idx, reg) \ 92 ({ \ 93 uint32_t internal_reg_offset, addr; \ 94 bool video_range, aon_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 aon_range = ((((0xFFFFF & addr) >= (VCN_AON_SOC_ADDRESS_2_0)) && \ 101 ((0xFFFFF & addr) < ((VCN_AON_SOC_ADDRESS_2_0 + 0x600))))); \ 102 if (video_range) \ 103 internal_reg_offset = ((0xFFFFF & addr) - (VCN_VID_SOC_ADDRESS_2_0) + \ 104 (VCN_VID_IP_ADDRESS_2_0)); \ 105 else if (aon_range) \ 106 internal_reg_offset = ((0xFFFFF & addr) - (VCN_AON_SOC_ADDRESS_2_0) + \ 107 (VCN_AON_IP_ADDRESS_2_0)); \ 108 else \ 109 internal_reg_offset = (0xFFFFF & addr); \ 110 \ 111 internal_reg_offset >>= 2; \ 112 }) 113 114 #define RREG32_SOC15_DPG_MODE_2_0(inst_idx, offset, mask_en) \ 115 ({ \ 116 WREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_CTL, \ 117 (0x0 << UVD_DPG_LMA_CTL__READ_WRITE__SHIFT | \ 118 mask_en << UVD_DPG_LMA_CTL__MASK_EN__SHIFT | \ 119 offset << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT)); \ 120 RREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_DATA); \ 121 }) 122 123 #define WREG32_SOC15_DPG_MODE_2_0(inst_idx, offset, value, mask_en, indirect) \ 124 do { \ 125 if (!indirect) { \ 126 WREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_DATA, value); \ 127 WREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_CTL, \ 128 (0x1 << UVD_DPG_LMA_CTL__READ_WRITE__SHIFT | \ 129 mask_en << UVD_DPG_LMA_CTL__MASK_EN__SHIFT | \ 130 offset << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT)); \ 131 } else { \ 132 *adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ = offset; \ 133 *adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ = value; \ 134 } \ 135 } while (0) 136 137 enum engine_status_constants { 138 UVD_PGFSM_STATUS__UVDM_UVDU_PWR_ON = 0x2AAAA0, 139 UVD_PGFSM_STATUS__UVDM_UVDU_PWR_ON_2_0 = 0xAAAA0, 140 UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON = 0x00000002, 141 UVD_STATUS__UVD_BUSY = 0x00000004, 142 GB_ADDR_CONFIG_DEFAULT = 0x26010011, 143 UVD_STATUS__IDLE = 0x2, 144 UVD_STATUS__BUSY = 0x5, 145 UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF = 0x1, 146 UVD_STATUS__RBC_BUSY = 0x1, 147 UVD_PGFSM_STATUS_UVDJ_PWR_ON = 0, 148 }; 149 150 enum internal_dpg_state { 151 VCN_DPG_STATE__UNPAUSE = 0, 152 VCN_DPG_STATE__PAUSE, 153 }; 154 155 struct dpg_pause_state { 156 enum internal_dpg_state fw_based; 157 enum internal_dpg_state jpeg; 158 }; 159 160 struct amdgpu_vcn_reg{ 161 unsigned data0; 162 unsigned data1; 163 unsigned cmd; 164 unsigned nop; 165 unsigned context_id; 166 unsigned ib_vmid; 167 unsigned ib_bar_low; 168 unsigned ib_bar_high; 169 unsigned ib_size; 170 unsigned gp_scratch8; 171 unsigned scratch9; 172 }; 173 174 struct amdgpu_vcn_inst { 175 struct amdgpu_bo *vcpu_bo; 176 void *cpu_addr; 177 uint64_t gpu_addr; 178 void *saved_bo; 179 struct amdgpu_ring ring_dec; 180 struct amdgpu_ring ring_enc[AMDGPU_VCN_MAX_ENC_RINGS]; 181 struct amdgpu_irq_src irq; 182 struct amdgpu_vcn_reg external; 183 struct amdgpu_bo *dpg_sram_bo; 184 struct dpg_pause_state pause_state; 185 void *dpg_sram_cpu_addr; 186 uint64_t dpg_sram_gpu_addr; 187 uint32_t *dpg_sram_curr_addr; 188 }; 189 190 struct amdgpu_vcn { 191 unsigned fw_version; 192 struct delayed_work idle_work; 193 const struct firmware *fw; /* VCN firmware */ 194 unsigned num_enc_rings; 195 enum amd_powergating_state cur_state; 196 bool indirect_sram; 197 198 uint8_t num_vcn_inst; 199 struct amdgpu_vcn_inst inst[AMDGPU_MAX_VCN_INSTANCES]; 200 struct amdgpu_vcn_reg internal; 201 struct drm_gpu_scheduler *vcn_enc_sched[AMDGPU_MAX_VCN_ENC_RINGS]; 202 struct drm_gpu_scheduler *vcn_dec_sched[AMDGPU_MAX_VCN_INSTANCES]; 203 uint32_t num_vcn_enc_sched; 204 uint32_t num_vcn_dec_sched; 205 206 unsigned harvest_config; 207 int (*pause_dpg_mode)(struct amdgpu_device *adev, 208 int inst_idx, struct dpg_pause_state *new_state); 209 }; 210 211 int amdgpu_vcn_sw_init(struct amdgpu_device *adev); 212 int amdgpu_vcn_sw_fini(struct amdgpu_device *adev); 213 int amdgpu_vcn_suspend(struct amdgpu_device *adev); 214 int amdgpu_vcn_resume(struct amdgpu_device *adev); 215 void amdgpu_vcn_ring_begin_use(struct amdgpu_ring *ring); 216 void amdgpu_vcn_ring_end_use(struct amdgpu_ring *ring); 217 218 int amdgpu_vcn_dec_ring_test_ring(struct amdgpu_ring *ring); 219 int amdgpu_vcn_dec_ring_test_ib(struct amdgpu_ring *ring, long timeout); 220 221 int amdgpu_vcn_enc_ring_test_ring(struct amdgpu_ring *ring); 222 int amdgpu_vcn_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout); 223 224 #endif 225