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 #include "amdgpu.h" 24 #include "amdgpu_atombios.h" 25 #include "nbio_v6_1.h" 26 27 #include "nbio/nbio_6_1_default.h" 28 #include "nbio/nbio_6_1_offset.h" 29 #include "nbio/nbio_6_1_sh_mask.h" 30 #include "nbio/nbio_6_1_smn.h" 31 #include "vega10_enum.h" 32 33 static u32 nbio_v6_1_get_rev_id(struct amdgpu_device *adev) 34 { 35 u32 tmp = RREG32_SOC15(NBIO, 0, mmRCC_DEV0_EPF0_STRAP0); 36 37 tmp &= RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0_MASK; 38 tmp >>= RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0__SHIFT; 39 40 return tmp; 41 } 42 43 static void nbio_v6_1_mc_access_enable(struct amdgpu_device *adev, bool enable) 44 { 45 if (enable) 46 WREG32_SOC15(NBIO, 0, mmBIF_FB_EN, 47 BIF_FB_EN__FB_READ_EN_MASK | 48 BIF_FB_EN__FB_WRITE_EN_MASK); 49 else 50 WREG32_SOC15(NBIO, 0, mmBIF_FB_EN, 0); 51 } 52 53 static void nbio_v6_1_hdp_flush(struct amdgpu_device *adev, 54 struct amdgpu_ring *ring) 55 { 56 if (!ring || !ring->funcs->emit_wreg) 57 WREG32_SOC15_NO_KIQ(NBIO, 0, 58 mmBIF_BX_PF0_HDP_MEM_COHERENCY_FLUSH_CNTL, 59 0); 60 else 61 amdgpu_ring_emit_wreg(ring, SOC15_REG_OFFSET( 62 NBIO, 0, mmBIF_BX_PF0_HDP_MEM_COHERENCY_FLUSH_CNTL), 0); 63 } 64 65 static u32 nbio_v6_1_get_memsize(struct amdgpu_device *adev) 66 { 67 return RREG32_SOC15(NBIO, 0, mmRCC_PF_0_0_RCC_CONFIG_MEMSIZE); 68 } 69 70 static void nbio_v6_1_sdma_doorbell_range(struct amdgpu_device *adev, int instance, 71 bool use_doorbell, int doorbell_index, int doorbell_size) 72 { 73 u32 reg = instance == 0 ? SOC15_REG_OFFSET(NBIO, 0, mmBIF_SDMA0_DOORBELL_RANGE) : 74 SOC15_REG_OFFSET(NBIO, 0, mmBIF_SDMA1_DOORBELL_RANGE); 75 76 u32 doorbell_range = RREG32(reg); 77 78 if (use_doorbell) { 79 doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, OFFSET, doorbell_index); 80 doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, SIZE, doorbell_size); 81 } else 82 doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, SIZE, 0); 83 84 WREG32(reg, doorbell_range); 85 86 } 87 88 static void nbio_v6_1_enable_doorbell_aperture(struct amdgpu_device *adev, 89 bool enable) 90 { 91 WREG32_FIELD15(NBIO, 0, RCC_PF_0_0_RCC_DOORBELL_APER_EN, BIF_DOORBELL_APER_EN, enable ? 1 : 0); 92 } 93 94 static void nbio_v6_1_enable_doorbell_selfring_aperture(struct amdgpu_device *adev, 95 bool enable) 96 { 97 u32 tmp = 0; 98 99 if (enable) { 100 tmp = REG_SET_FIELD(tmp, BIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL, DOORBELL_SELFRING_GPA_APER_EN, 1) | 101 REG_SET_FIELD(tmp, BIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL, DOORBELL_SELFRING_GPA_APER_MODE, 1) | 102 REG_SET_FIELD(tmp, BIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL, DOORBELL_SELFRING_GPA_APER_SIZE, 0); 103 104 WREG32_SOC15(NBIO, 0, mmBIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_BASE_LOW, 105 lower_32_bits(adev->doorbell.base)); 106 WREG32_SOC15(NBIO, 0, mmBIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_BASE_HIGH, 107 upper_32_bits(adev->doorbell.base)); 108 } 109 110 WREG32_SOC15(NBIO, 0, mmBIF_BX_PF0_DOORBELL_SELFRING_GPA_APER_CNTL, tmp); 111 } 112 113 114 static void nbio_v6_1_ih_doorbell_range(struct amdgpu_device *adev, 115 bool use_doorbell, int doorbell_index) 116 { 117 u32 ih_doorbell_range = RREG32_SOC15(NBIO, 0 , mmBIF_IH_DOORBELL_RANGE); 118 119 if (use_doorbell) { 120 ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, BIF_IH_DOORBELL_RANGE, OFFSET, doorbell_index); 121 ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, BIF_IH_DOORBELL_RANGE, SIZE, 2); 122 } else 123 ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, BIF_IH_DOORBELL_RANGE, SIZE, 0); 124 125 WREG32_SOC15(NBIO, 0, mmBIF_IH_DOORBELL_RANGE, ih_doorbell_range); 126 } 127 128 static void nbio_v6_1_ih_control(struct amdgpu_device *adev) 129 { 130 u32 interrupt_cntl; 131 132 /* setup interrupt control */ 133 WREG32_SOC15(NBIO, 0, mmINTERRUPT_CNTL2, adev->dummy_page_addr >> 8); 134 interrupt_cntl = RREG32_SOC15(NBIO, 0, mmINTERRUPT_CNTL); 135 /* INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=0 - dummy read disabled with msi, enabled without msi 136 * INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=1 - dummy read controlled by IH_DUMMY_RD_EN 137 */ 138 interrupt_cntl = REG_SET_FIELD(interrupt_cntl, INTERRUPT_CNTL, IH_DUMMY_RD_OVERRIDE, 0); 139 /* INTERRUPT_CNTL__IH_REQ_NONSNOOP_EN_MASK=1 if ring is in non-cacheable memory, e.g., vram */ 140 interrupt_cntl = REG_SET_FIELD(interrupt_cntl, INTERRUPT_CNTL, IH_REQ_NONSNOOP_EN, 0); 141 WREG32_SOC15(NBIO, 0, mmINTERRUPT_CNTL, interrupt_cntl); 142 } 143 144 static void nbio_v6_1_update_medium_grain_clock_gating(struct amdgpu_device *adev, 145 bool enable) 146 { 147 uint32_t def, data; 148 149 def = data = RREG32_PCIE(smnCPM_CONTROL); 150 if (enable && (adev->cg_flags & AMD_CG_SUPPORT_BIF_MGCG)) { 151 data |= (CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK | 152 CPM_CONTROL__TXCLK_DYN_GATE_ENABLE_MASK | 153 CPM_CONTROL__TXCLK_PERM_GATE_ENABLE_MASK | 154 CPM_CONTROL__TXCLK_LCNT_GATE_ENABLE_MASK | 155 CPM_CONTROL__TXCLK_REGS_GATE_ENABLE_MASK | 156 CPM_CONTROL__TXCLK_PRBS_GATE_ENABLE_MASK | 157 CPM_CONTROL__REFCLK_REGS_GATE_ENABLE_MASK); 158 } else { 159 data &= ~(CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK | 160 CPM_CONTROL__TXCLK_DYN_GATE_ENABLE_MASK | 161 CPM_CONTROL__TXCLK_PERM_GATE_ENABLE_MASK | 162 CPM_CONTROL__TXCLK_LCNT_GATE_ENABLE_MASK | 163 CPM_CONTROL__TXCLK_REGS_GATE_ENABLE_MASK | 164 CPM_CONTROL__TXCLK_PRBS_GATE_ENABLE_MASK | 165 CPM_CONTROL__REFCLK_REGS_GATE_ENABLE_MASK); 166 } 167 168 if (def != data) 169 WREG32_PCIE(smnCPM_CONTROL, data); 170 } 171 172 static void nbio_v6_1_update_medium_grain_light_sleep(struct amdgpu_device *adev, 173 bool enable) 174 { 175 uint32_t def, data; 176 177 def = data = RREG32_PCIE(smnPCIE_CNTL2); 178 if (enable && (adev->cg_flags & AMD_CG_SUPPORT_BIF_LS)) { 179 data |= (PCIE_CNTL2__SLV_MEM_LS_EN_MASK | 180 PCIE_CNTL2__MST_MEM_LS_EN_MASK | 181 PCIE_CNTL2__REPLAY_MEM_LS_EN_MASK); 182 } else { 183 data &= ~(PCIE_CNTL2__SLV_MEM_LS_EN_MASK | 184 PCIE_CNTL2__MST_MEM_LS_EN_MASK | 185 PCIE_CNTL2__REPLAY_MEM_LS_EN_MASK); 186 } 187 188 if (def != data) 189 WREG32_PCIE(smnPCIE_CNTL2, data); 190 } 191 192 static void nbio_v6_1_get_clockgating_state(struct amdgpu_device *adev, 193 u32 *flags) 194 { 195 int data; 196 197 /* AMD_CG_SUPPORT_BIF_MGCG */ 198 data = RREG32_PCIE(smnCPM_CONTROL); 199 if (data & CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK) 200 *flags |= AMD_CG_SUPPORT_BIF_MGCG; 201 202 /* AMD_CG_SUPPORT_BIF_LS */ 203 data = RREG32_PCIE(smnPCIE_CNTL2); 204 if (data & PCIE_CNTL2__SLV_MEM_LS_EN_MASK) 205 *flags |= AMD_CG_SUPPORT_BIF_LS; 206 } 207 208 static u32 nbio_v6_1_get_hdp_flush_req_offset(struct amdgpu_device *adev) 209 { 210 return SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF0_GPU_HDP_FLUSH_REQ); 211 } 212 213 static u32 nbio_v6_1_get_hdp_flush_done_offset(struct amdgpu_device *adev) 214 { 215 return SOC15_REG_OFFSET(NBIO, 0, mmBIF_BX_PF0_GPU_HDP_FLUSH_DONE); 216 } 217 218 static u32 nbio_v6_1_get_pcie_index_offset(struct amdgpu_device *adev) 219 { 220 return SOC15_REG_OFFSET(NBIO, 0, mmPCIE_INDEX2); 221 } 222 223 static u32 nbio_v6_1_get_pcie_data_offset(struct amdgpu_device *adev) 224 { 225 return SOC15_REG_OFFSET(NBIO, 0, mmPCIE_DATA2); 226 } 227 228 static const struct nbio_hdp_flush_reg nbio_v6_1_hdp_flush_reg = { 229 .ref_and_mask_cp0 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP0_MASK, 230 .ref_and_mask_cp1 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP1_MASK, 231 .ref_and_mask_cp2 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP2_MASK, 232 .ref_and_mask_cp3 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP3_MASK, 233 .ref_and_mask_cp4 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP4_MASK, 234 .ref_and_mask_cp5 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP5_MASK, 235 .ref_and_mask_cp6 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP6_MASK, 236 .ref_and_mask_cp7 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP7_MASK, 237 .ref_and_mask_cp8 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP8_MASK, 238 .ref_and_mask_cp9 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__CP9_MASK, 239 .ref_and_mask_sdma0 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__SDMA0_MASK, 240 .ref_and_mask_sdma1 = BIF_BX_PF0_GPU_HDP_FLUSH_DONE__SDMA1_MASK 241 }; 242 243 static void nbio_v6_1_detect_hw_virt(struct amdgpu_device *adev) 244 { 245 uint32_t reg; 246 247 reg = RREG32_SOC15(NBIO, 0, mmRCC_PF_0_0_RCC_IOV_FUNC_IDENTIFIER); 248 if (reg & 1) 249 adev->virt.caps |= AMDGPU_SRIOV_CAPS_IS_VF; 250 251 if (reg & 0x80000000) 252 adev->virt.caps |= AMDGPU_SRIOV_CAPS_ENABLE_IOV; 253 254 if (!reg) { 255 if (is_virtual_machine()) /* passthrough mode exclus sriov mod */ 256 adev->virt.caps |= AMDGPU_PASSTHROUGH_MODE; 257 } 258 } 259 260 static void nbio_v6_1_init_registers(struct amdgpu_device *adev) 261 { 262 uint32_t def, data; 263 264 def = data = RREG32_PCIE(smnPCIE_CONFIG_CNTL); 265 data = REG_SET_FIELD(data, PCIE_CONFIG_CNTL, CI_SWUS_MAX_READ_REQUEST_SIZE_MODE, 1); 266 data = REG_SET_FIELD(data, PCIE_CONFIG_CNTL, CI_SWUS_MAX_READ_REQUEST_SIZE_PRIV, 1); 267 268 if (def != data) 269 WREG32_PCIE(smnPCIE_CONFIG_CNTL, data); 270 271 def = data = RREG32_PCIE(smnPCIE_CI_CNTL); 272 data = REG_SET_FIELD(data, PCIE_CI_CNTL, CI_SLV_ORDERING_DIS, 1); 273 274 if (def != data) 275 WREG32_PCIE(smnPCIE_CI_CNTL, data); 276 } 277 278 const struct amdgpu_nbio_funcs nbio_v6_1_funcs = { 279 .hdp_flush_reg = &nbio_v6_1_hdp_flush_reg, 280 .get_hdp_flush_req_offset = nbio_v6_1_get_hdp_flush_req_offset, 281 .get_hdp_flush_done_offset = nbio_v6_1_get_hdp_flush_done_offset, 282 .get_pcie_index_offset = nbio_v6_1_get_pcie_index_offset, 283 .get_pcie_data_offset = nbio_v6_1_get_pcie_data_offset, 284 .get_rev_id = nbio_v6_1_get_rev_id, 285 .mc_access_enable = nbio_v6_1_mc_access_enable, 286 .hdp_flush = nbio_v6_1_hdp_flush, 287 .get_memsize = nbio_v6_1_get_memsize, 288 .sdma_doorbell_range = nbio_v6_1_sdma_doorbell_range, 289 .enable_doorbell_aperture = nbio_v6_1_enable_doorbell_aperture, 290 .enable_doorbell_selfring_aperture = nbio_v6_1_enable_doorbell_selfring_aperture, 291 .ih_doorbell_range = nbio_v6_1_ih_doorbell_range, 292 .update_medium_grain_clock_gating = nbio_v6_1_update_medium_grain_clock_gating, 293 .update_medium_grain_light_sleep = nbio_v6_1_update_medium_grain_light_sleep, 294 .get_clockgating_state = nbio_v6_1_get_clockgating_state, 295 .ih_control = nbio_v6_1_ih_control, 296 .init_registers = nbio_v6_1_init_registers, 297 .detect_hw_virt = nbio_v6_1_detect_hw_virt, 298 }; 299