1 /* 2 * Copyright 2008 Advanced Micro Devices, Inc. 3 * Copyright 2008 Red Hat Inc. 4 * Copyright 2009 Jerome Glisse. 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 * Authors: Dave Airlie 25 * Alex Deucher 26 * Jerome Glisse 27 */ 28 #include <drm/drmP.h> 29 #include <drm/radeon_drm.h> 30 #include "radeon_reg.h" 31 #include "radeon.h" 32 #include "radeon_asic.h" 33 34 #include "r100d.h" 35 #include "r200_reg_safe.h" 36 37 #include "r100_track.h" 38 39 static int r200_get_vtx_size_0(uint32_t vtx_fmt_0) 40 { 41 int vtx_size, i; 42 vtx_size = 2; 43 44 if (vtx_fmt_0 & R200_VTX_Z0) 45 vtx_size++; 46 if (vtx_fmt_0 & R200_VTX_W0) 47 vtx_size++; 48 /* blend weight */ 49 if (vtx_fmt_0 & (0x7 << R200_VTX_WEIGHT_COUNT_SHIFT)) 50 vtx_size += (vtx_fmt_0 >> R200_VTX_WEIGHT_COUNT_SHIFT) & 0x7; 51 if (vtx_fmt_0 & R200_VTX_PV_MATRIX_SEL) 52 vtx_size++; 53 if (vtx_fmt_0 & R200_VTX_N0) 54 vtx_size += 3; 55 if (vtx_fmt_0 & R200_VTX_POINT_SIZE) 56 vtx_size++; 57 if (vtx_fmt_0 & R200_VTX_DISCRETE_FOG) 58 vtx_size++; 59 if (vtx_fmt_0 & R200_VTX_SHININESS_0) 60 vtx_size++; 61 if (vtx_fmt_0 & R200_VTX_SHININESS_1) 62 vtx_size++; 63 for (i = 0; i < 8; i++) { 64 int color_size = (vtx_fmt_0 >> (11 + 2*i)) & 0x3; 65 switch (color_size) { 66 case 0: break; 67 case 1: vtx_size++; break; 68 case 2: vtx_size += 3; break; 69 case 3: vtx_size += 4; break; 70 } 71 } 72 if (vtx_fmt_0 & R200_VTX_XY1) 73 vtx_size += 2; 74 if (vtx_fmt_0 & R200_VTX_Z1) 75 vtx_size++; 76 if (vtx_fmt_0 & R200_VTX_W1) 77 vtx_size++; 78 if (vtx_fmt_0 & R200_VTX_N1) 79 vtx_size += 3; 80 return vtx_size; 81 } 82 83 int r200_copy_dma(struct radeon_device *rdev, 84 uint64_t src_offset, 85 uint64_t dst_offset, 86 unsigned num_gpu_pages, 87 struct radeon_fence **fence) 88 { 89 struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; 90 uint32_t size; 91 uint32_t cur_size; 92 int i, num_loops; 93 int r = 0; 94 95 /* radeon pitch is /64 */ 96 size = num_gpu_pages << RADEON_GPU_PAGE_SHIFT; 97 num_loops = DIV_ROUND_UP(size, 0x1FFFFF); 98 r = radeon_ring_lock(rdev, ring, num_loops * 4 + 64); 99 if (r) { 100 DRM_ERROR("radeon: moving bo (%d).\n", r); 101 return r; 102 } 103 /* Must wait for 2D idle & clean before DMA or hangs might happen */ 104 radeon_ring_write(ring, PACKET0(RADEON_WAIT_UNTIL, 0)); 105 radeon_ring_write(ring, (1 << 16)); 106 for (i = 0; i < num_loops; i++) { 107 cur_size = size; 108 if (cur_size > 0x1FFFFF) { 109 cur_size = 0x1FFFFF; 110 } 111 size -= cur_size; 112 radeon_ring_write(ring, PACKET0(0x720, 2)); 113 radeon_ring_write(ring, src_offset); 114 radeon_ring_write(ring, dst_offset); 115 radeon_ring_write(ring, cur_size | (1 << 31) | (1 << 30)); 116 src_offset += cur_size; 117 dst_offset += cur_size; 118 } 119 radeon_ring_write(ring, PACKET0(RADEON_WAIT_UNTIL, 0)); 120 radeon_ring_write(ring, RADEON_WAIT_DMA_GUI_IDLE); 121 if (fence) { 122 r = radeon_fence_emit(rdev, fence, RADEON_RING_TYPE_GFX_INDEX); 123 } 124 radeon_ring_unlock_commit(rdev, ring); 125 return r; 126 } 127 128 129 static int r200_get_vtx_size_1(uint32_t vtx_fmt_1) 130 { 131 int vtx_size, i, tex_size; 132 vtx_size = 0; 133 for (i = 0; i < 6; i++) { 134 tex_size = (vtx_fmt_1 >> (i * 3)) & 0x7; 135 if (tex_size > 4) 136 continue; 137 vtx_size += tex_size; 138 } 139 return vtx_size; 140 } 141 142 int r200_packet0_check(struct radeon_cs_parser *p, 143 struct radeon_cs_packet *pkt, 144 unsigned idx, unsigned reg) 145 { 146 struct radeon_cs_reloc *reloc; 147 struct r100_cs_track *track; 148 volatile uint32_t *ib; 149 uint32_t tmp; 150 int r; 151 int i; 152 int face; 153 u32 tile_flags = 0; 154 u32 idx_value; 155 156 ib = p->ib.ptr; 157 track = (struct r100_cs_track *)p->track; 158 idx_value = radeon_get_ib_value(p, idx); 159 switch (reg) { 160 case RADEON_CRTC_GUI_TRIG_VLINE: 161 r = r100_cs_packet_parse_vline(p); 162 if (r) { 163 DRM_ERROR("No reloc for ib[%d]=0x%04X\n", 164 idx, reg); 165 radeon_cs_dump_packet(p, pkt); 166 return r; 167 } 168 break; 169 /* FIXME: only allow PACKET3 blit? easier to check for out of 170 * range access */ 171 case RADEON_DST_PITCH_OFFSET: 172 case RADEON_SRC_PITCH_OFFSET: 173 r = r100_reloc_pitch_offset(p, pkt, idx, reg); 174 if (r) 175 return r; 176 break; 177 case RADEON_RB3D_DEPTHOFFSET: 178 r = radeon_cs_packet_next_reloc(p, &reloc, 0); 179 if (r) { 180 DRM_ERROR("No reloc for ib[%d]=0x%04X\n", 181 idx, reg); 182 radeon_cs_dump_packet(p, pkt); 183 return r; 184 } 185 track->zb.robj = reloc->robj; 186 track->zb.offset = idx_value; 187 track->zb_dirty = true; 188 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset); 189 break; 190 case RADEON_RB3D_COLOROFFSET: 191 r = radeon_cs_packet_next_reloc(p, &reloc, 0); 192 if (r) { 193 DRM_ERROR("No reloc for ib[%d]=0x%04X\n", 194 idx, reg); 195 radeon_cs_dump_packet(p, pkt); 196 return r; 197 } 198 track->cb[0].robj = reloc->robj; 199 track->cb[0].offset = idx_value; 200 track->cb_dirty = true; 201 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset); 202 break; 203 case R200_PP_TXOFFSET_0: 204 case R200_PP_TXOFFSET_1: 205 case R200_PP_TXOFFSET_2: 206 case R200_PP_TXOFFSET_3: 207 case R200_PP_TXOFFSET_4: 208 case R200_PP_TXOFFSET_5: 209 i = (reg - R200_PP_TXOFFSET_0) / 24; 210 r = radeon_cs_packet_next_reloc(p, &reloc, 0); 211 if (r) { 212 DRM_ERROR("No reloc for ib[%d]=0x%04X\n", 213 idx, reg); 214 radeon_cs_dump_packet(p, pkt); 215 return r; 216 } 217 if (!(p->cs_flags & RADEON_CS_KEEP_TILING_FLAGS)) { 218 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) 219 tile_flags |= R200_TXO_MACRO_TILE; 220 if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) 221 tile_flags |= R200_TXO_MICRO_TILE; 222 223 tmp = idx_value & ~(0x7 << 2); 224 tmp |= tile_flags; 225 ib[idx] = tmp + ((u32)reloc->lobj.gpu_offset); 226 } else 227 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset); 228 track->textures[i].robj = reloc->robj; 229 track->tex_dirty = true; 230 break; 231 case R200_PP_CUBIC_OFFSET_F1_0: 232 case R200_PP_CUBIC_OFFSET_F2_0: 233 case R200_PP_CUBIC_OFFSET_F3_0: 234 case R200_PP_CUBIC_OFFSET_F4_0: 235 case R200_PP_CUBIC_OFFSET_F5_0: 236 case R200_PP_CUBIC_OFFSET_F1_1: 237 case R200_PP_CUBIC_OFFSET_F2_1: 238 case R200_PP_CUBIC_OFFSET_F3_1: 239 case R200_PP_CUBIC_OFFSET_F4_1: 240 case R200_PP_CUBIC_OFFSET_F5_1: 241 case R200_PP_CUBIC_OFFSET_F1_2: 242 case R200_PP_CUBIC_OFFSET_F2_2: 243 case R200_PP_CUBIC_OFFSET_F3_2: 244 case R200_PP_CUBIC_OFFSET_F4_2: 245 case R200_PP_CUBIC_OFFSET_F5_2: 246 case R200_PP_CUBIC_OFFSET_F1_3: 247 case R200_PP_CUBIC_OFFSET_F2_3: 248 case R200_PP_CUBIC_OFFSET_F3_3: 249 case R200_PP_CUBIC_OFFSET_F4_3: 250 case R200_PP_CUBIC_OFFSET_F5_3: 251 case R200_PP_CUBIC_OFFSET_F1_4: 252 case R200_PP_CUBIC_OFFSET_F2_4: 253 case R200_PP_CUBIC_OFFSET_F3_4: 254 case R200_PP_CUBIC_OFFSET_F4_4: 255 case R200_PP_CUBIC_OFFSET_F5_4: 256 case R200_PP_CUBIC_OFFSET_F1_5: 257 case R200_PP_CUBIC_OFFSET_F2_5: 258 case R200_PP_CUBIC_OFFSET_F3_5: 259 case R200_PP_CUBIC_OFFSET_F4_5: 260 case R200_PP_CUBIC_OFFSET_F5_5: 261 i = (reg - R200_PP_TXOFFSET_0) / 24; 262 face = (reg - ((i * 24) + R200_PP_TXOFFSET_0)) / 4; 263 r = radeon_cs_packet_next_reloc(p, &reloc, 0); 264 if (r) { 265 DRM_ERROR("No reloc for ib[%d]=0x%04X\n", 266 idx, reg); 267 radeon_cs_dump_packet(p, pkt); 268 return r; 269 } 270 track->textures[i].cube_info[face - 1].offset = idx_value; 271 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset); 272 track->textures[i].cube_info[face - 1].robj = reloc->robj; 273 track->tex_dirty = true; 274 break; 275 case RADEON_RE_WIDTH_HEIGHT: 276 track->maxy = ((idx_value >> 16) & 0x7FF); 277 track->cb_dirty = true; 278 track->zb_dirty = true; 279 break; 280 case RADEON_RB3D_COLORPITCH: 281 r = radeon_cs_packet_next_reloc(p, &reloc, 0); 282 if (r) { 283 DRM_ERROR("No reloc for ib[%d]=0x%04X\n", 284 idx, reg); 285 radeon_cs_dump_packet(p, pkt); 286 return r; 287 } 288 289 if (!(p->cs_flags & RADEON_CS_KEEP_TILING_FLAGS)) { 290 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) 291 tile_flags |= RADEON_COLOR_TILE_ENABLE; 292 if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) 293 tile_flags |= RADEON_COLOR_MICROTILE_ENABLE; 294 295 tmp = idx_value & ~(0x7 << 16); 296 tmp |= tile_flags; 297 ib[idx] = tmp; 298 } else 299 ib[idx] = idx_value; 300 301 track->cb[0].pitch = idx_value & RADEON_COLORPITCH_MASK; 302 track->cb_dirty = true; 303 break; 304 case RADEON_RB3D_DEPTHPITCH: 305 track->zb.pitch = idx_value & RADEON_DEPTHPITCH_MASK; 306 track->zb_dirty = true; 307 break; 308 case RADEON_RB3D_CNTL: 309 switch ((idx_value >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f) { 310 case 7: 311 case 8: 312 case 9: 313 case 11: 314 case 12: 315 track->cb[0].cpp = 1; 316 break; 317 case 3: 318 case 4: 319 case 15: 320 track->cb[0].cpp = 2; 321 break; 322 case 6: 323 track->cb[0].cpp = 4; 324 break; 325 default: 326 DRM_ERROR("Invalid color buffer format (%d) !\n", 327 ((idx_value >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f)); 328 return -EINVAL; 329 } 330 if (idx_value & RADEON_DEPTHXY_OFFSET_ENABLE) { 331 DRM_ERROR("No support for depth xy offset in kms\n"); 332 return -EINVAL; 333 } 334 335 track->z_enabled = !!(idx_value & RADEON_Z_ENABLE); 336 track->cb_dirty = true; 337 track->zb_dirty = true; 338 break; 339 case RADEON_RB3D_ZSTENCILCNTL: 340 switch (idx_value & 0xf) { 341 case 0: 342 track->zb.cpp = 2; 343 break; 344 case 2: 345 case 3: 346 case 4: 347 case 5: 348 case 9: 349 case 11: 350 track->zb.cpp = 4; 351 break; 352 default: 353 break; 354 } 355 track->zb_dirty = true; 356 break; 357 case RADEON_RB3D_ZPASS_ADDR: 358 r = radeon_cs_packet_next_reloc(p, &reloc, 0); 359 if (r) { 360 DRM_ERROR("No reloc for ib[%d]=0x%04X\n", 361 idx, reg); 362 radeon_cs_dump_packet(p, pkt); 363 return r; 364 } 365 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset); 366 break; 367 case RADEON_PP_CNTL: 368 { 369 uint32_t temp = idx_value >> 4; 370 for (i = 0; i < track->num_texture; i++) 371 track->textures[i].enabled = !!(temp & (1 << i)); 372 track->tex_dirty = true; 373 } 374 break; 375 case RADEON_SE_VF_CNTL: 376 track->vap_vf_cntl = idx_value; 377 break; 378 case 0x210c: 379 /* VAP_VF_MAX_VTX_INDX */ 380 track->max_indx = idx_value & 0x00FFFFFFUL; 381 break; 382 case R200_SE_VTX_FMT_0: 383 track->vtx_size = r200_get_vtx_size_0(idx_value); 384 break; 385 case R200_SE_VTX_FMT_1: 386 track->vtx_size += r200_get_vtx_size_1(idx_value); 387 break; 388 case R200_PP_TXSIZE_0: 389 case R200_PP_TXSIZE_1: 390 case R200_PP_TXSIZE_2: 391 case R200_PP_TXSIZE_3: 392 case R200_PP_TXSIZE_4: 393 case R200_PP_TXSIZE_5: 394 i = (reg - R200_PP_TXSIZE_0) / 32; 395 track->textures[i].width = (idx_value & RADEON_TEX_USIZE_MASK) + 1; 396 track->textures[i].height = ((idx_value & RADEON_TEX_VSIZE_MASK) >> RADEON_TEX_VSIZE_SHIFT) + 1; 397 track->tex_dirty = true; 398 break; 399 case R200_PP_TXPITCH_0: 400 case R200_PP_TXPITCH_1: 401 case R200_PP_TXPITCH_2: 402 case R200_PP_TXPITCH_3: 403 case R200_PP_TXPITCH_4: 404 case R200_PP_TXPITCH_5: 405 i = (reg - R200_PP_TXPITCH_0) / 32; 406 track->textures[i].pitch = idx_value + 32; 407 track->tex_dirty = true; 408 break; 409 case R200_PP_TXFILTER_0: 410 case R200_PP_TXFILTER_1: 411 case R200_PP_TXFILTER_2: 412 case R200_PP_TXFILTER_3: 413 case R200_PP_TXFILTER_4: 414 case R200_PP_TXFILTER_5: 415 i = (reg - R200_PP_TXFILTER_0) / 32; 416 track->textures[i].num_levels = ((idx_value & R200_MAX_MIP_LEVEL_MASK) 417 >> R200_MAX_MIP_LEVEL_SHIFT); 418 tmp = (idx_value >> 23) & 0x7; 419 if (tmp == 2 || tmp == 6) 420 track->textures[i].roundup_w = false; 421 tmp = (idx_value >> 27) & 0x7; 422 if (tmp == 2 || tmp == 6) 423 track->textures[i].roundup_h = false; 424 track->tex_dirty = true; 425 break; 426 case R200_PP_TXMULTI_CTL_0: 427 case R200_PP_TXMULTI_CTL_1: 428 case R200_PP_TXMULTI_CTL_2: 429 case R200_PP_TXMULTI_CTL_3: 430 case R200_PP_TXMULTI_CTL_4: 431 case R200_PP_TXMULTI_CTL_5: 432 i = (reg - R200_PP_TXMULTI_CTL_0) / 32; 433 break; 434 case R200_PP_TXFORMAT_X_0: 435 case R200_PP_TXFORMAT_X_1: 436 case R200_PP_TXFORMAT_X_2: 437 case R200_PP_TXFORMAT_X_3: 438 case R200_PP_TXFORMAT_X_4: 439 case R200_PP_TXFORMAT_X_5: 440 i = (reg - R200_PP_TXFORMAT_X_0) / 32; 441 track->textures[i].txdepth = idx_value & 0x7; 442 tmp = (idx_value >> 16) & 0x3; 443 /* 2D, 3D, CUBE */ 444 switch (tmp) { 445 case 0: 446 case 3: 447 case 4: 448 case 5: 449 case 6: 450 case 7: 451 /* 1D/2D */ 452 track->textures[i].tex_coord_type = 0; 453 break; 454 case 1: 455 /* CUBE */ 456 track->textures[i].tex_coord_type = 2; 457 break; 458 case 2: 459 /* 3D */ 460 track->textures[i].tex_coord_type = 1; 461 break; 462 } 463 track->tex_dirty = true; 464 break; 465 case R200_PP_TXFORMAT_0: 466 case R200_PP_TXFORMAT_1: 467 case R200_PP_TXFORMAT_2: 468 case R200_PP_TXFORMAT_3: 469 case R200_PP_TXFORMAT_4: 470 case R200_PP_TXFORMAT_5: 471 i = (reg - R200_PP_TXFORMAT_0) / 32; 472 if (idx_value & R200_TXFORMAT_NON_POWER2) { 473 track->textures[i].use_pitch = 1; 474 } else { 475 track->textures[i].use_pitch = 0; 476 track->textures[i].width = 1 << ((idx_value >> RADEON_TXFORMAT_WIDTH_SHIFT) & RADEON_TXFORMAT_WIDTH_MASK); 477 track->textures[i].height = 1 << ((idx_value >> RADEON_TXFORMAT_HEIGHT_SHIFT) & RADEON_TXFORMAT_HEIGHT_MASK); 478 } 479 if (idx_value & R200_TXFORMAT_LOOKUP_DISABLE) 480 track->textures[i].lookup_disable = true; 481 switch ((idx_value & RADEON_TXFORMAT_FORMAT_MASK)) { 482 case R200_TXFORMAT_I8: 483 case R200_TXFORMAT_RGB332: 484 case R200_TXFORMAT_Y8: 485 track->textures[i].cpp = 1; 486 track->textures[i].compress_format = R100_TRACK_COMP_NONE; 487 break; 488 case R200_TXFORMAT_AI88: 489 case R200_TXFORMAT_ARGB1555: 490 case R200_TXFORMAT_RGB565: 491 case R200_TXFORMAT_ARGB4444: 492 case R200_TXFORMAT_VYUY422: 493 case R200_TXFORMAT_YVYU422: 494 case R200_TXFORMAT_LDVDU655: 495 case R200_TXFORMAT_DVDU88: 496 case R200_TXFORMAT_AVYU4444: 497 track->textures[i].cpp = 2; 498 track->textures[i].compress_format = R100_TRACK_COMP_NONE; 499 break; 500 case R200_TXFORMAT_ARGB8888: 501 case R200_TXFORMAT_RGBA8888: 502 case R200_TXFORMAT_ABGR8888: 503 case R200_TXFORMAT_BGR111110: 504 case R200_TXFORMAT_LDVDU8888: 505 track->textures[i].cpp = 4; 506 track->textures[i].compress_format = R100_TRACK_COMP_NONE; 507 break; 508 case R200_TXFORMAT_DXT1: 509 track->textures[i].cpp = 1; 510 track->textures[i].compress_format = R100_TRACK_COMP_DXT1; 511 break; 512 case R200_TXFORMAT_DXT23: 513 case R200_TXFORMAT_DXT45: 514 track->textures[i].cpp = 1; 515 track->textures[i].compress_format = R100_TRACK_COMP_DXT1; 516 break; 517 } 518 track->textures[i].cube_info[4].width = 1 << ((idx_value >> 16) & 0xf); 519 track->textures[i].cube_info[4].height = 1 << ((idx_value >> 20) & 0xf); 520 track->tex_dirty = true; 521 break; 522 case R200_PP_CUBIC_FACES_0: 523 case R200_PP_CUBIC_FACES_1: 524 case R200_PP_CUBIC_FACES_2: 525 case R200_PP_CUBIC_FACES_3: 526 case R200_PP_CUBIC_FACES_4: 527 case R200_PP_CUBIC_FACES_5: 528 tmp = idx_value; 529 i = (reg - R200_PP_CUBIC_FACES_0) / 32; 530 for (face = 0; face < 4; face++) { 531 track->textures[i].cube_info[face].width = 1 << ((tmp >> (face * 8)) & 0xf); 532 track->textures[i].cube_info[face].height = 1 << ((tmp >> ((face * 8) + 4)) & 0xf); 533 } 534 track->tex_dirty = true; 535 break; 536 default: 537 printk(KERN_ERR "Forbidden register 0x%04X in cs at %d\n", 538 reg, idx); 539 return -EINVAL; 540 } 541 return 0; 542 } 543 544 void r200_set_safe_registers(struct radeon_device *rdev) 545 { 546 rdev->config.r100.reg_safe_bm = r200_reg_safe_bm; 547 rdev->config.r100.reg_safe_bm_size = ARRAY_SIZE(r200_reg_safe_bm); 548 } 549