| 1 | /* |
| 2 | * Copyright 2013 Advanced Micro Devices, Inc. |
| 3 | * All Rights Reserved. |
| 4 | * |
| 5 | * Permission is hereby granted, free of charge, to any person obtaining a |
| 6 | * copy of this software and associated documentation files (the |
| 7 | * "Software"), to deal in the Software without restriction, including |
| 8 | * without limitation the rights to use, copy, modify, merge, publish, |
| 9 | * distribute, sub license, and/or sell copies of the Software, and to |
| 10 | * permit persons to whom the Software is furnished to do so, subject to |
| 11 | * the following conditions: |
| 12 | * |
| 13 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
| 14 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
| 15 | * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL |
| 16 | * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, |
| 17 | * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR |
| 18 | * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE |
| 19 | * USE OR OTHER DEALINGS IN THE SOFTWARE. |
| 20 | * |
| 21 | * The above copyright notice and this permission notice (including the |
| 22 | * next paragraph) shall be included in all copies or substantial portions |
| 23 | * of the Software. |
| 24 | * |
| 25 | * Authors: Christian König <christian.koenig@amd.com> |
| 26 | */ |
| 27 | |
| 28 | #include <linux/firmware.h> |
| 29 | |
| 30 | #include "amdgpu.h" |
| 31 | #include "amdgpu_vce.h" |
| 32 | #include "cikd.h" |
| 33 | #include "vce/vce_2_0_d.h" |
| 34 | #include "vce/vce_2_0_sh_mask.h" |
| 35 | #include "smu/smu_7_0_1_d.h" |
| 36 | #include "smu/smu_7_0_1_sh_mask.h" |
| 37 | #include "oss/oss_2_0_d.h" |
| 38 | #include "oss/oss_2_0_sh_mask.h" |
| 39 | |
| 40 | #define VCE_V2_0_FW_SIZE (256 * 1024) |
| 41 | #define VCE_V2_0_STACK_SIZE (64 * 1024) |
| 42 | #define VCE_V2_0_DATA_SIZE (23552 * AMDGPU_MAX_VCE_HANDLES) |
| 43 | #define VCE_STATUS_VCPU_REPORT_FW_LOADED_MASK 0x02 |
| 44 | |
| 45 | static void vce_v2_0_set_ring_funcs(struct amdgpu_device *adev); |
| 46 | static void vce_v2_0_set_irq_funcs(struct amdgpu_device *adev); |
| 47 | |
| 48 | /** |
| 49 | * vce_v2_0_ring_get_rptr - get read pointer |
| 50 | * |
| 51 | * @ring: amdgpu_ring pointer |
| 52 | * |
| 53 | * Returns the current hardware read pointer |
| 54 | */ |
| 55 | static uint64_t vce_v2_0_ring_get_rptr(struct amdgpu_ring *ring) |
| 56 | { |
| 57 | struct amdgpu_device *adev = ring->adev; |
| 58 | |
| 59 | if (ring->me == 0) |
| 60 | return RREG32(mmVCE_RB_RPTR); |
| 61 | else |
| 62 | return RREG32(mmVCE_RB_RPTR2); |
| 63 | } |
| 64 | |
| 65 | /** |
| 66 | * vce_v2_0_ring_get_wptr - get write pointer |
| 67 | * |
| 68 | * @ring: amdgpu_ring pointer |
| 69 | * |
| 70 | * Returns the current hardware write pointer |
| 71 | */ |
| 72 | static uint64_t vce_v2_0_ring_get_wptr(struct amdgpu_ring *ring) |
| 73 | { |
| 74 | struct amdgpu_device *adev = ring->adev; |
| 75 | |
| 76 | if (ring->me == 0) |
| 77 | return RREG32(mmVCE_RB_WPTR); |
| 78 | else |
| 79 | return RREG32(mmVCE_RB_WPTR2); |
| 80 | } |
| 81 | |
| 82 | /** |
| 83 | * vce_v2_0_ring_set_wptr - set write pointer |
| 84 | * |
| 85 | * @ring: amdgpu_ring pointer |
| 86 | * |
| 87 | * Commits the write pointer to the hardware |
| 88 | */ |
| 89 | static void vce_v2_0_ring_set_wptr(struct amdgpu_ring *ring) |
| 90 | { |
| 91 | struct amdgpu_device *adev = ring->adev; |
| 92 | |
| 93 | if (ring->me == 0) |
| 94 | WREG32(mmVCE_RB_WPTR, lower_32_bits(ring->wptr)); |
| 95 | else |
| 96 | WREG32(mmVCE_RB_WPTR2, lower_32_bits(ring->wptr)); |
| 97 | } |
| 98 | |
| 99 | static int vce_v2_0_lmi_clean(struct amdgpu_device *adev) |
| 100 | { |
| 101 | int i, j; |
| 102 | |
| 103 | for (i = 0; i < 10; ++i) { |
| 104 | for (j = 0; j < 100; ++j) { |
| 105 | uint32_t status = RREG32(mmVCE_LMI_STATUS); |
| 106 | |
| 107 | if (status & 0x337f) |
| 108 | return 0; |
| 109 | mdelay(10); |
| 110 | } |
| 111 | } |
| 112 | |
| 113 | return -ETIMEDOUT; |
| 114 | } |
| 115 | |
| 116 | static int vce_v2_0_firmware_loaded(struct amdgpu_device *adev) |
| 117 | { |
| 118 | int i, j; |
| 119 | |
| 120 | for (i = 0; i < 10; ++i) { |
| 121 | for (j = 0; j < 100; ++j) { |
| 122 | uint32_t status = RREG32(mmVCE_STATUS); |
| 123 | |
| 124 | if (status & VCE_STATUS_VCPU_REPORT_FW_LOADED_MASK) |
| 125 | return 0; |
| 126 | mdelay(10); |
| 127 | } |
| 128 | |
| 129 | DRM_ERROR("VCE not responding, trying to reset the ECPU!!!\n" ); |
| 130 | WREG32_P(mmVCE_SOFT_RESET, |
| 131 | VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK, |
| 132 | ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); |
| 133 | mdelay(10); |
| 134 | WREG32_P(mmVCE_SOFT_RESET, 0, |
| 135 | ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK); |
| 136 | mdelay(10); |
| 137 | } |
| 138 | |
| 139 | return -ETIMEDOUT; |
| 140 | } |
| 141 | |
| 142 | static void vce_v2_0_disable_cg(struct amdgpu_device *adev) |
| 143 | { |
| 144 | WREG32(mmVCE_CGTT_CLK_OVERRIDE, 7); |
| 145 | } |
| 146 | |
| 147 | static void vce_v2_0_init_cg(struct amdgpu_device *adev) |
| 148 | { |
| 149 | u32 tmp; |
| 150 | |
| 151 | tmp = RREG32(mmVCE_CLOCK_GATING_A); |
| 152 | tmp &= ~0xfff; |
| 153 | tmp |= ((0 << 0) | (4 << 4)); |
| 154 | tmp |= 0x40000; |
| 155 | WREG32(mmVCE_CLOCK_GATING_A, tmp); |
| 156 | |
| 157 | tmp = RREG32(mmVCE_UENC_CLOCK_GATING); |
| 158 | tmp &= ~0xfff; |
| 159 | tmp |= ((0 << 0) | (4 << 4)); |
| 160 | WREG32(mmVCE_UENC_CLOCK_GATING, tmp); |
| 161 | |
| 162 | tmp = RREG32(mmVCE_CLOCK_GATING_B); |
| 163 | tmp |= 0x10; |
| 164 | tmp &= ~0x100000; |
| 165 | WREG32(mmVCE_CLOCK_GATING_B, tmp); |
| 166 | } |
| 167 | |
| 168 | static void vce_v2_0_mc_resume(struct amdgpu_device *adev) |
| 169 | { |
| 170 | uint32_t size, offset; |
| 171 | |
| 172 | WREG32_P(mmVCE_CLOCK_GATING_A, 0, ~(1 << 16)); |
| 173 | WREG32_P(mmVCE_UENC_CLOCK_GATING, 0x1FF000, ~0xFF9FF000); |
| 174 | WREG32_P(mmVCE_UENC_REG_CLOCK_GATING, 0x3F, ~0x3F); |
| 175 | WREG32(mmVCE_CLOCK_GATING_B, 0xf7); |
| 176 | |
| 177 | WREG32(mmVCE_LMI_CTRL, 0x00398000); |
| 178 | WREG32_P(mmVCE_LMI_CACHE_CTRL, 0x0, ~0x1); |
| 179 | WREG32(mmVCE_LMI_SWAP_CNTL, 0); |
| 180 | WREG32(mmVCE_LMI_SWAP_CNTL1, 0); |
| 181 | WREG32(mmVCE_LMI_VM_CTRL, 0); |
| 182 | |
| 183 | WREG32(mmVCE_LMI_VCPU_CACHE_40BIT_BAR, (adev->vce.gpu_addr >> 8)); |
| 184 | |
| 185 | offset = AMDGPU_VCE_FIRMWARE_OFFSET; |
| 186 | size = VCE_V2_0_FW_SIZE; |
| 187 | WREG32(mmVCE_VCPU_CACHE_OFFSET0, offset & 0x7fffffff); |
| 188 | WREG32(mmVCE_VCPU_CACHE_SIZE0, size); |
| 189 | |
| 190 | offset += size; |
| 191 | size = VCE_V2_0_STACK_SIZE; |
| 192 | WREG32(mmVCE_VCPU_CACHE_OFFSET1, offset & 0x7fffffff); |
| 193 | WREG32(mmVCE_VCPU_CACHE_SIZE1, size); |
| 194 | |
| 195 | offset += size; |
| 196 | size = VCE_V2_0_DATA_SIZE; |
| 197 | WREG32(mmVCE_VCPU_CACHE_OFFSET2, offset & 0x7fffffff); |
| 198 | WREG32(mmVCE_VCPU_CACHE_SIZE2, size); |
| 199 | |
| 200 | WREG32_P(mmVCE_LMI_CTRL2, 0x0, ~0x100); |
| 201 | WREG32_FIELD(VCE_SYS_INT_EN, VCE_SYS_INT_TRAP_INTERRUPT_EN, 1); |
| 202 | } |
| 203 | |
| 204 | static bool vce_v2_0_is_idle(struct amdgpu_ip_block *ip_block) |
| 205 | { |
| 206 | struct amdgpu_device *adev = ip_block->adev; |
| 207 | |
| 208 | return !(RREG32(mmSRBM_STATUS2) & SRBM_STATUS2__VCE_BUSY_MASK); |
| 209 | } |
| 210 | |
| 211 | static int vce_v2_0_wait_for_idle(struct amdgpu_ip_block *ip_block) |
| 212 | { |
| 213 | struct amdgpu_device *adev = ip_block->adev; |
| 214 | unsigned i; |
| 215 | |
| 216 | for (i = 0; i < adev->usec_timeout; i++) { |
| 217 | if (vce_v2_0_is_idle(ip_block)) |
| 218 | return 0; |
| 219 | } |
| 220 | return -ETIMEDOUT; |
| 221 | } |
| 222 | |
| 223 | /** |
| 224 | * vce_v2_0_start - start VCE block |
| 225 | * |
| 226 | * @adev: amdgpu_device pointer |
| 227 | * |
| 228 | * Setup and start the VCE block |
| 229 | */ |
| 230 | static int vce_v2_0_start(struct amdgpu_device *adev) |
| 231 | { |
| 232 | struct amdgpu_ring *ring; |
| 233 | int r; |
| 234 | |
| 235 | /* set BUSY flag */ |
| 236 | WREG32_P(mmVCE_STATUS, 1, ~1); |
| 237 | |
| 238 | vce_v2_0_init_cg(adev); |
| 239 | vce_v2_0_disable_cg(adev); |
| 240 | |
| 241 | vce_v2_0_mc_resume(adev); |
| 242 | |
| 243 | ring = &adev->vce.ring[0]; |
| 244 | WREG32(mmVCE_RB_RPTR, lower_32_bits(ring->wptr)); |
| 245 | WREG32(mmVCE_RB_WPTR, lower_32_bits(ring->wptr)); |
| 246 | WREG32(mmVCE_RB_BASE_LO, ring->gpu_addr); |
| 247 | WREG32(mmVCE_RB_BASE_HI, upper_32_bits(ring->gpu_addr)); |
| 248 | WREG32(mmVCE_RB_SIZE, ring->ring_size / 4); |
| 249 | |
| 250 | ring = &adev->vce.ring[1]; |
| 251 | WREG32(mmVCE_RB_RPTR2, lower_32_bits(ring->wptr)); |
| 252 | WREG32(mmVCE_RB_WPTR2, lower_32_bits(ring->wptr)); |
| 253 | WREG32(mmVCE_RB_BASE_LO2, ring->gpu_addr); |
| 254 | WREG32(mmVCE_RB_BASE_HI2, upper_32_bits(ring->gpu_addr)); |
| 255 | WREG32(mmVCE_RB_SIZE2, ring->ring_size / 4); |
| 256 | |
| 257 | WREG32_FIELD(VCE_VCPU_CNTL, CLK_EN, 1); |
| 258 | WREG32_FIELD(VCE_SOFT_RESET, ECPU_SOFT_RESET, 1); |
| 259 | mdelay(100); |
| 260 | WREG32_FIELD(VCE_SOFT_RESET, ECPU_SOFT_RESET, 0); |
| 261 | |
| 262 | r = vce_v2_0_firmware_loaded(adev); |
| 263 | |
| 264 | /* clear BUSY flag */ |
| 265 | WREG32_P(mmVCE_STATUS, 0, ~1); |
| 266 | |
| 267 | if (r) { |
| 268 | DRM_ERROR("VCE not responding, giving up!!!\n" ); |
| 269 | return r; |
| 270 | } |
| 271 | |
| 272 | return 0; |
| 273 | } |
| 274 | |
| 275 | static int vce_v2_0_stop(struct amdgpu_device *adev) |
| 276 | { |
| 277 | struct amdgpu_ip_block *ip_block; |
| 278 | int i; |
| 279 | int status; |
| 280 | |
| 281 | |
| 282 | if (vce_v2_0_lmi_clean(adev)) { |
| 283 | DRM_INFO("VCE is not idle \n" ); |
| 284 | return 0; |
| 285 | } |
| 286 | |
| 287 | ip_block = amdgpu_device_ip_get_ip_block(adev, type: AMD_IP_BLOCK_TYPE_VCE); |
| 288 | if (!ip_block) |
| 289 | return -EINVAL; |
| 290 | |
| 291 | if (vce_v2_0_wait_for_idle(ip_block)) { |
| 292 | DRM_INFO("VCE is busy, Can't set clock gating" ); |
| 293 | return 0; |
| 294 | } |
| 295 | |
| 296 | /* Stall UMC and register bus before resetting VCPU */ |
| 297 | WREG32_P(mmVCE_LMI_CTRL2, 1 << 8, ~(1 << 8)); |
| 298 | |
| 299 | for (i = 0; i < 100; ++i) { |
| 300 | status = RREG32(mmVCE_LMI_STATUS); |
| 301 | if (status & 0x240) |
| 302 | break; |
| 303 | mdelay(1); |
| 304 | } |
| 305 | |
| 306 | WREG32_P(mmVCE_VCPU_CNTL, 0, ~0x80001); |
| 307 | |
| 308 | /* put LMI, VCPU, RBC etc... into reset */ |
| 309 | WREG32_P(mmVCE_SOFT_RESET, 1, ~0x1); |
| 310 | |
| 311 | WREG32(mmVCE_STATUS, 0); |
| 312 | |
| 313 | return 0; |
| 314 | } |
| 315 | |
| 316 | static void vce_v2_0_set_sw_cg(struct amdgpu_device *adev, bool gated) |
| 317 | { |
| 318 | u32 tmp; |
| 319 | |
| 320 | if (gated) { |
| 321 | tmp = RREG32(mmVCE_CLOCK_GATING_B); |
| 322 | tmp |= 0xe70000; |
| 323 | WREG32(mmVCE_CLOCK_GATING_B, tmp); |
| 324 | |
| 325 | tmp = RREG32(mmVCE_UENC_CLOCK_GATING); |
| 326 | tmp |= 0xff000000; |
| 327 | WREG32(mmVCE_UENC_CLOCK_GATING, tmp); |
| 328 | |
| 329 | tmp = RREG32(mmVCE_UENC_REG_CLOCK_GATING); |
| 330 | tmp &= ~0x3fc; |
| 331 | WREG32(mmVCE_UENC_REG_CLOCK_GATING, tmp); |
| 332 | |
| 333 | WREG32(mmVCE_CGTT_CLK_OVERRIDE, 0); |
| 334 | } else { |
| 335 | tmp = RREG32(mmVCE_CLOCK_GATING_B); |
| 336 | tmp |= 0xe7; |
| 337 | tmp &= ~0xe70000; |
| 338 | WREG32(mmVCE_CLOCK_GATING_B, tmp); |
| 339 | |
| 340 | tmp = RREG32(mmVCE_UENC_CLOCK_GATING); |
| 341 | tmp |= 0x1fe000; |
| 342 | tmp &= ~0xff000000; |
| 343 | WREG32(mmVCE_UENC_CLOCK_GATING, tmp); |
| 344 | |
| 345 | tmp = RREG32(mmVCE_UENC_REG_CLOCK_GATING); |
| 346 | tmp |= 0x3fc; |
| 347 | WREG32(mmVCE_UENC_REG_CLOCK_GATING, tmp); |
| 348 | } |
| 349 | } |
| 350 | |
| 351 | static void vce_v2_0_set_dyn_cg(struct amdgpu_device *adev, bool gated) |
| 352 | { |
| 353 | u32 orig, tmp; |
| 354 | |
| 355 | /* LMI_MC/LMI_UMC always set in dynamic, |
| 356 | * set {CGC_*_GATE_MODE, CGC_*_SW_GATE} = {0, 0} |
| 357 | */ |
| 358 | tmp = RREG32(mmVCE_CLOCK_GATING_B); |
| 359 | tmp &= ~0x00060006; |
| 360 | |
| 361 | /* Exception for ECPU, IH, SEM, SYS blocks needs to be turned on/off by SW */ |
| 362 | if (gated) { |
| 363 | tmp |= 0xe10000; |
| 364 | WREG32(mmVCE_CLOCK_GATING_B, tmp); |
| 365 | } else { |
| 366 | tmp |= 0xe1; |
| 367 | tmp &= ~0xe10000; |
| 368 | WREG32(mmVCE_CLOCK_GATING_B, tmp); |
| 369 | } |
| 370 | |
| 371 | orig = tmp = RREG32(mmVCE_UENC_CLOCK_GATING); |
| 372 | tmp &= ~0x1fe000; |
| 373 | tmp &= ~0xff000000; |
| 374 | if (tmp != orig) |
| 375 | WREG32(mmVCE_UENC_CLOCK_GATING, tmp); |
| 376 | |
| 377 | orig = tmp = RREG32(mmVCE_UENC_REG_CLOCK_GATING); |
| 378 | tmp &= ~0x3fc; |
| 379 | if (tmp != orig) |
| 380 | WREG32(mmVCE_UENC_REG_CLOCK_GATING, tmp); |
| 381 | |
| 382 | /* set VCE_UENC_REG_CLOCK_GATING always in dynamic mode */ |
| 383 | WREG32(mmVCE_UENC_REG_CLOCK_GATING, 0x00); |
| 384 | |
| 385 | if(gated) |
| 386 | WREG32(mmVCE_CGTT_CLK_OVERRIDE, 0); |
| 387 | } |
| 388 | |
| 389 | static void vce_v2_0_enable_mgcg(struct amdgpu_device *adev, bool enable, |
| 390 | bool sw_cg) |
| 391 | { |
| 392 | if (enable && (adev->cg_flags & AMD_CG_SUPPORT_VCE_MGCG)) { |
| 393 | if (sw_cg) |
| 394 | vce_v2_0_set_sw_cg(adev, gated: true); |
| 395 | else |
| 396 | vce_v2_0_set_dyn_cg(adev, gated: true); |
| 397 | } else { |
| 398 | vce_v2_0_disable_cg(adev); |
| 399 | |
| 400 | if (sw_cg) |
| 401 | vce_v2_0_set_sw_cg(adev, gated: false); |
| 402 | else |
| 403 | vce_v2_0_set_dyn_cg(adev, gated: false); |
| 404 | } |
| 405 | } |
| 406 | |
| 407 | static int vce_v2_0_early_init(struct amdgpu_ip_block *ip_block) |
| 408 | { |
| 409 | struct amdgpu_device *adev = ip_block->adev; |
| 410 | |
| 411 | adev->vce.num_rings = 2; |
| 412 | |
| 413 | vce_v2_0_set_ring_funcs(adev); |
| 414 | vce_v2_0_set_irq_funcs(adev); |
| 415 | |
| 416 | return 0; |
| 417 | } |
| 418 | |
| 419 | static int vce_v2_0_sw_init(struct amdgpu_ip_block *ip_block) |
| 420 | { |
| 421 | struct amdgpu_ring *ring; |
| 422 | int r, i; |
| 423 | struct amdgpu_device *adev = ip_block->adev; |
| 424 | |
| 425 | /* VCE */ |
| 426 | r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, src_id: 167, source: &adev->vce.irq); |
| 427 | if (r) |
| 428 | return r; |
| 429 | |
| 430 | r = amdgpu_vce_sw_init(adev, VCE_V2_0_FW_SIZE + |
| 431 | VCE_V2_0_STACK_SIZE + VCE_V2_0_DATA_SIZE); |
| 432 | if (r) |
| 433 | return r; |
| 434 | |
| 435 | r = amdgpu_vce_resume(adev); |
| 436 | if (r) |
| 437 | return r; |
| 438 | |
| 439 | for (i = 0; i < adev->vce.num_rings; i++) { |
| 440 | enum amdgpu_ring_priority_level hw_prio = amdgpu_vce_get_ring_prio(ring: i); |
| 441 | |
| 442 | ring = &adev->vce.ring[i]; |
| 443 | sprintf(buf: ring->name, fmt: "vce%d" , i); |
| 444 | r = amdgpu_ring_init(adev, ring, max_dw: 512, irq_src: &adev->vce.irq, irq_type: 0, |
| 445 | hw_prio, NULL); |
| 446 | if (r) |
| 447 | return r; |
| 448 | } |
| 449 | |
| 450 | return r; |
| 451 | } |
| 452 | |
| 453 | static int vce_v2_0_sw_fini(struct amdgpu_ip_block *ip_block) |
| 454 | { |
| 455 | int r; |
| 456 | struct amdgpu_device *adev = ip_block->adev; |
| 457 | |
| 458 | r = amdgpu_vce_suspend(adev); |
| 459 | if (r) |
| 460 | return r; |
| 461 | |
| 462 | return amdgpu_vce_sw_fini(adev); |
| 463 | } |
| 464 | |
| 465 | static int vce_v2_0_hw_init(struct amdgpu_ip_block *ip_block) |
| 466 | { |
| 467 | int r, i; |
| 468 | struct amdgpu_device *adev = ip_block->adev; |
| 469 | |
| 470 | amdgpu_asic_set_vce_clocks(adev, 10000, 10000); |
| 471 | vce_v2_0_enable_mgcg(adev, enable: true, sw_cg: false); |
| 472 | |
| 473 | for (i = 0; i < adev->vce.num_rings; i++) { |
| 474 | r = amdgpu_ring_test_helper(ring: &adev->vce.ring[i]); |
| 475 | if (r) |
| 476 | return r; |
| 477 | } |
| 478 | |
| 479 | DRM_INFO("VCE initialized successfully.\n" ); |
| 480 | |
| 481 | return 0; |
| 482 | } |
| 483 | |
| 484 | static int vce_v2_0_hw_fini(struct amdgpu_ip_block *ip_block) |
| 485 | { |
| 486 | cancel_delayed_work_sync(dwork: &ip_block->adev->vce.idle_work); |
| 487 | |
| 488 | return 0; |
| 489 | } |
| 490 | |
| 491 | static int vce_v2_0_suspend(struct amdgpu_ip_block *ip_block) |
| 492 | { |
| 493 | int r; |
| 494 | struct amdgpu_device *adev = ip_block->adev; |
| 495 | |
| 496 | |
| 497 | /* |
| 498 | * Proper cleanups before halting the HW engine: |
| 499 | * - cancel the delayed idle work |
| 500 | * - enable powergating |
| 501 | * - enable clockgating |
| 502 | * - disable dpm |
| 503 | * |
| 504 | * TODO: to align with the VCN implementation, move the |
| 505 | * jobs for clockgating/powergating/dpm setting to |
| 506 | * ->set_powergating_state(). |
| 507 | */ |
| 508 | cancel_delayed_work_sync(dwork: &adev->vce.idle_work); |
| 509 | |
| 510 | if (adev->pm.dpm_enabled) { |
| 511 | amdgpu_dpm_enable_vce(adev, enable: false); |
| 512 | } else { |
| 513 | amdgpu_asic_set_vce_clocks(adev, 0, 0); |
| 514 | amdgpu_device_ip_set_powergating_state(dev: adev, block_type: AMD_IP_BLOCK_TYPE_VCE, |
| 515 | state: AMD_PG_STATE_GATE); |
| 516 | amdgpu_device_ip_set_clockgating_state(dev: adev, block_type: AMD_IP_BLOCK_TYPE_VCE, |
| 517 | state: AMD_CG_STATE_GATE); |
| 518 | } |
| 519 | |
| 520 | r = vce_v2_0_hw_fini(ip_block); |
| 521 | if (r) |
| 522 | return r; |
| 523 | |
| 524 | return amdgpu_vce_suspend(adev); |
| 525 | } |
| 526 | |
| 527 | static int vce_v2_0_resume(struct amdgpu_ip_block *ip_block) |
| 528 | { |
| 529 | int r; |
| 530 | |
| 531 | r = amdgpu_vce_resume(adev: ip_block->adev); |
| 532 | if (r) |
| 533 | return r; |
| 534 | |
| 535 | return vce_v2_0_hw_init(ip_block); |
| 536 | } |
| 537 | |
| 538 | static int vce_v2_0_soft_reset(struct amdgpu_ip_block *ip_block) |
| 539 | { |
| 540 | struct amdgpu_device *adev = ip_block->adev; |
| 541 | |
| 542 | WREG32_FIELD(SRBM_SOFT_RESET, SOFT_RESET_VCE, 1); |
| 543 | mdelay(5); |
| 544 | |
| 545 | return vce_v2_0_start(adev); |
| 546 | } |
| 547 | |
| 548 | static int vce_v2_0_set_interrupt_state(struct amdgpu_device *adev, |
| 549 | struct amdgpu_irq_src *source, |
| 550 | unsigned type, |
| 551 | enum amdgpu_interrupt_state state) |
| 552 | { |
| 553 | uint32_t val = 0; |
| 554 | |
| 555 | if (state == AMDGPU_IRQ_STATE_ENABLE) |
| 556 | val |= VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK; |
| 557 | |
| 558 | WREG32_P(mmVCE_SYS_INT_EN, val, ~VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK); |
| 559 | return 0; |
| 560 | } |
| 561 | |
| 562 | static int vce_v2_0_process_interrupt(struct amdgpu_device *adev, |
| 563 | struct amdgpu_irq_src *source, |
| 564 | struct amdgpu_iv_entry *entry) |
| 565 | { |
| 566 | DRM_DEBUG("IH: VCE\n" ); |
| 567 | switch (entry->src_data[0]) { |
| 568 | case 0: |
| 569 | case 1: |
| 570 | amdgpu_fence_process(ring: &adev->vce.ring[entry->src_data[0]]); |
| 571 | break; |
| 572 | default: |
| 573 | DRM_ERROR("Unhandled interrupt: %d %d\n" , |
| 574 | entry->src_id, entry->src_data[0]); |
| 575 | break; |
| 576 | } |
| 577 | |
| 578 | return 0; |
| 579 | } |
| 580 | |
| 581 | static int vce_v2_0_set_clockgating_state(struct amdgpu_ip_block *ip_block, |
| 582 | enum amd_clockgating_state state) |
| 583 | { |
| 584 | bool gate = false; |
| 585 | bool sw_cg = false; |
| 586 | |
| 587 | struct amdgpu_device *adev = ip_block->adev; |
| 588 | |
| 589 | if (state == AMD_CG_STATE_GATE) { |
| 590 | gate = true; |
| 591 | sw_cg = true; |
| 592 | } |
| 593 | |
| 594 | vce_v2_0_enable_mgcg(adev, enable: gate, sw_cg); |
| 595 | |
| 596 | return 0; |
| 597 | } |
| 598 | |
| 599 | static int vce_v2_0_set_powergating_state(struct amdgpu_ip_block *ip_block, |
| 600 | enum amd_powergating_state state) |
| 601 | { |
| 602 | /* This doesn't actually powergate the VCE block. |
| 603 | * That's done in the dpm code via the SMC. This |
| 604 | * just re-inits the block as necessary. The actual |
| 605 | * gating still happens in the dpm code. We should |
| 606 | * revisit this when there is a cleaner line between |
| 607 | * the smc and the hw blocks |
| 608 | */ |
| 609 | struct amdgpu_device *adev = ip_block->adev; |
| 610 | |
| 611 | if (state == AMD_PG_STATE_GATE) |
| 612 | return vce_v2_0_stop(adev); |
| 613 | else |
| 614 | return vce_v2_0_start(adev); |
| 615 | } |
| 616 | |
| 617 | static const struct amd_ip_funcs vce_v2_0_ip_funcs = { |
| 618 | .name = "vce_v2_0" , |
| 619 | .early_init = vce_v2_0_early_init, |
| 620 | .sw_init = vce_v2_0_sw_init, |
| 621 | .sw_fini = vce_v2_0_sw_fini, |
| 622 | .hw_init = vce_v2_0_hw_init, |
| 623 | .hw_fini = vce_v2_0_hw_fini, |
| 624 | .suspend = vce_v2_0_suspend, |
| 625 | .resume = vce_v2_0_resume, |
| 626 | .is_idle = vce_v2_0_is_idle, |
| 627 | .wait_for_idle = vce_v2_0_wait_for_idle, |
| 628 | .soft_reset = vce_v2_0_soft_reset, |
| 629 | .set_clockgating_state = vce_v2_0_set_clockgating_state, |
| 630 | .set_powergating_state = vce_v2_0_set_powergating_state, |
| 631 | }; |
| 632 | |
| 633 | static const struct amdgpu_ring_funcs vce_v2_0_ring_funcs = { |
| 634 | .type = AMDGPU_RING_TYPE_VCE, |
| 635 | .align_mask = 0xf, |
| 636 | .nop = VCE_CMD_NO_OP, |
| 637 | .support_64bit_ptrs = false, |
| 638 | .no_user_fence = true, |
| 639 | .get_rptr = vce_v2_0_ring_get_rptr, |
| 640 | .get_wptr = vce_v2_0_ring_get_wptr, |
| 641 | .set_wptr = vce_v2_0_ring_set_wptr, |
| 642 | .parse_cs = amdgpu_vce_ring_parse_cs, |
| 643 | .emit_frame_size = 6, /* amdgpu_vce_ring_emit_fence x1 no user fence */ |
| 644 | .emit_ib_size = 4, /* amdgpu_vce_ring_emit_ib */ |
| 645 | .emit_ib = amdgpu_vce_ring_emit_ib, |
| 646 | .emit_fence = amdgpu_vce_ring_emit_fence, |
| 647 | .test_ring = amdgpu_vce_ring_test_ring, |
| 648 | .test_ib = amdgpu_vce_ring_test_ib, |
| 649 | .insert_nop = amdgpu_ring_insert_nop, |
| 650 | .pad_ib = amdgpu_ring_generic_pad_ib, |
| 651 | .begin_use = amdgpu_vce_ring_begin_use, |
| 652 | .end_use = amdgpu_vce_ring_end_use, |
| 653 | }; |
| 654 | |
| 655 | static void vce_v2_0_set_ring_funcs(struct amdgpu_device *adev) |
| 656 | { |
| 657 | int i; |
| 658 | |
| 659 | for (i = 0; i < adev->vce.num_rings; i++) { |
| 660 | adev->vce.ring[i].funcs = &vce_v2_0_ring_funcs; |
| 661 | adev->vce.ring[i].me = i; |
| 662 | } |
| 663 | } |
| 664 | |
| 665 | static const struct amdgpu_irq_src_funcs vce_v2_0_irq_funcs = { |
| 666 | .set = vce_v2_0_set_interrupt_state, |
| 667 | .process = vce_v2_0_process_interrupt, |
| 668 | }; |
| 669 | |
| 670 | static void vce_v2_0_set_irq_funcs(struct amdgpu_device *adev) |
| 671 | { |
| 672 | adev->vce.irq.num_types = 1; |
| 673 | adev->vce.irq.funcs = &vce_v2_0_irq_funcs; |
| 674 | }; |
| 675 | |
| 676 | const struct amdgpu_ip_block_version vce_v2_0_ip_block = |
| 677 | { |
| 678 | .type = AMD_IP_BLOCK_TYPE_VCE, |
| 679 | .major = 2, |
| 680 | .minor = 0, |
| 681 | .rev = 0, |
| 682 | .funcs = &vce_v2_0_ip_funcs, |
| 683 | }; |
| 684 | |