1 /* 2 * Copyright 2012 Red Hat 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 * Authors: Ben Skeggs 23 */ 24 #include "gf100.h" 25 #include "ctxgf100.h" 26 #include "fuc/os.h" 27 28 #include <core/client.h> 29 #include <core/firmware.h> 30 #include <core/option.h> 31 #include <subdev/acr.h> 32 #include <subdev/fb.h> 33 #include <subdev/mc.h> 34 #include <subdev/pmu.h> 35 #include <subdev/therm.h> 36 #include <subdev/timer.h> 37 #include <engine/fifo.h> 38 39 #include <nvif/class.h> 40 #include <nvif/cl9097.h> 41 #include <nvif/if900d.h> 42 #include <nvif/unpack.h> 43 44 /******************************************************************************* 45 * Zero Bandwidth Clear 46 ******************************************************************************/ 47 48 static void 49 gf100_gr_zbc_clear_color(struct gf100_gr *gr, int zbc) 50 { 51 struct nvkm_device *device = gr->base.engine.subdev.device; 52 if (gr->zbc_color[zbc].format) { 53 nvkm_wr32(device, 0x405804, gr->zbc_color[zbc].ds[0]); 54 nvkm_wr32(device, 0x405808, gr->zbc_color[zbc].ds[1]); 55 nvkm_wr32(device, 0x40580c, gr->zbc_color[zbc].ds[2]); 56 nvkm_wr32(device, 0x405810, gr->zbc_color[zbc].ds[3]); 57 } 58 nvkm_wr32(device, 0x405814, gr->zbc_color[zbc].format); 59 nvkm_wr32(device, 0x405820, zbc); 60 nvkm_wr32(device, 0x405824, 0x00000004); /* TRIGGER | WRITE | COLOR */ 61 } 62 63 static int 64 gf100_gr_zbc_color_get(struct gf100_gr *gr, int format, 65 const u32 ds[4], const u32 l2[4]) 66 { 67 struct nvkm_ltc *ltc = gr->base.engine.subdev.device->ltc; 68 int zbc = -ENOSPC, i; 69 70 for (i = ltc->zbc_min; i <= ltc->zbc_max; i++) { 71 if (gr->zbc_color[i].format) { 72 if (gr->zbc_color[i].format != format) 73 continue; 74 if (memcmp(gr->zbc_color[i].ds, ds, sizeof( 75 gr->zbc_color[i].ds))) 76 continue; 77 if (memcmp(gr->zbc_color[i].l2, l2, sizeof( 78 gr->zbc_color[i].l2))) { 79 WARN_ON(1); 80 return -EINVAL; 81 } 82 return i; 83 } else { 84 zbc = (zbc < 0) ? i : zbc; 85 } 86 } 87 88 if (zbc < 0) 89 return zbc; 90 91 memcpy(gr->zbc_color[zbc].ds, ds, sizeof(gr->zbc_color[zbc].ds)); 92 memcpy(gr->zbc_color[zbc].l2, l2, sizeof(gr->zbc_color[zbc].l2)); 93 gr->zbc_color[zbc].format = format; 94 nvkm_ltc_zbc_color_get(ltc, zbc, l2); 95 gr->func->zbc->clear_color(gr, zbc); 96 return zbc; 97 } 98 99 static void 100 gf100_gr_zbc_clear_depth(struct gf100_gr *gr, int zbc) 101 { 102 struct nvkm_device *device = gr->base.engine.subdev.device; 103 if (gr->zbc_depth[zbc].format) 104 nvkm_wr32(device, 0x405818, gr->zbc_depth[zbc].ds); 105 nvkm_wr32(device, 0x40581c, gr->zbc_depth[zbc].format); 106 nvkm_wr32(device, 0x405820, zbc); 107 nvkm_wr32(device, 0x405824, 0x00000005); /* TRIGGER | WRITE | DEPTH */ 108 } 109 110 static int 111 gf100_gr_zbc_depth_get(struct gf100_gr *gr, int format, 112 const u32 ds, const u32 l2) 113 { 114 struct nvkm_ltc *ltc = gr->base.engine.subdev.device->ltc; 115 int zbc = -ENOSPC, i; 116 117 for (i = ltc->zbc_min; i <= ltc->zbc_max; i++) { 118 if (gr->zbc_depth[i].format) { 119 if (gr->zbc_depth[i].format != format) 120 continue; 121 if (gr->zbc_depth[i].ds != ds) 122 continue; 123 if (gr->zbc_depth[i].l2 != l2) { 124 WARN_ON(1); 125 return -EINVAL; 126 } 127 return i; 128 } else { 129 zbc = (zbc < 0) ? i : zbc; 130 } 131 } 132 133 if (zbc < 0) 134 return zbc; 135 136 gr->zbc_depth[zbc].format = format; 137 gr->zbc_depth[zbc].ds = ds; 138 gr->zbc_depth[zbc].l2 = l2; 139 nvkm_ltc_zbc_depth_get(ltc, zbc, l2); 140 gr->func->zbc->clear_depth(gr, zbc); 141 return zbc; 142 } 143 144 const struct gf100_gr_func_zbc 145 gf100_gr_zbc = { 146 .clear_color = gf100_gr_zbc_clear_color, 147 .clear_depth = gf100_gr_zbc_clear_depth, 148 }; 149 150 /******************************************************************************* 151 * Graphics object classes 152 ******************************************************************************/ 153 #define gf100_gr_object(p) container_of((p), struct gf100_gr_object, object) 154 155 struct gf100_gr_object { 156 struct nvkm_object object; 157 struct gf100_gr_chan *chan; 158 }; 159 160 static int 161 gf100_fermi_mthd_zbc_color(struct nvkm_object *object, void *data, u32 size) 162 { 163 struct gf100_gr *gr = gf100_gr(nvkm_gr(object->engine)); 164 union { 165 struct fermi_a_zbc_color_v0 v0; 166 } *args = data; 167 int ret = -ENOSYS; 168 169 if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, false))) { 170 switch (args->v0.format) { 171 case FERMI_A_ZBC_COLOR_V0_FMT_ZERO: 172 case FERMI_A_ZBC_COLOR_V0_FMT_UNORM_ONE: 173 case FERMI_A_ZBC_COLOR_V0_FMT_RF32_GF32_BF32_AF32: 174 case FERMI_A_ZBC_COLOR_V0_FMT_R16_G16_B16_A16: 175 case FERMI_A_ZBC_COLOR_V0_FMT_RN16_GN16_BN16_AN16: 176 case FERMI_A_ZBC_COLOR_V0_FMT_RS16_GS16_BS16_AS16: 177 case FERMI_A_ZBC_COLOR_V0_FMT_RU16_GU16_BU16_AU16: 178 case FERMI_A_ZBC_COLOR_V0_FMT_RF16_GF16_BF16_AF16: 179 case FERMI_A_ZBC_COLOR_V0_FMT_A8R8G8B8: 180 case FERMI_A_ZBC_COLOR_V0_FMT_A8RL8GL8BL8: 181 case FERMI_A_ZBC_COLOR_V0_FMT_A2B10G10R10: 182 case FERMI_A_ZBC_COLOR_V0_FMT_AU2BU10GU10RU10: 183 case FERMI_A_ZBC_COLOR_V0_FMT_A8B8G8R8: 184 case FERMI_A_ZBC_COLOR_V0_FMT_A8BL8GL8RL8: 185 case FERMI_A_ZBC_COLOR_V0_FMT_AN8BN8GN8RN8: 186 case FERMI_A_ZBC_COLOR_V0_FMT_AS8BS8GS8RS8: 187 case FERMI_A_ZBC_COLOR_V0_FMT_AU8BU8GU8RU8: 188 case FERMI_A_ZBC_COLOR_V0_FMT_A2R10G10B10: 189 case FERMI_A_ZBC_COLOR_V0_FMT_BF10GF11RF11: 190 ret = gf100_gr_zbc_color_get(gr, args->v0.format, 191 args->v0.ds, 192 args->v0.l2); 193 if (ret >= 0) { 194 args->v0.index = ret; 195 return 0; 196 } 197 break; 198 default: 199 return -EINVAL; 200 } 201 } 202 203 return ret; 204 } 205 206 static int 207 gf100_fermi_mthd_zbc_depth(struct nvkm_object *object, void *data, u32 size) 208 { 209 struct gf100_gr *gr = gf100_gr(nvkm_gr(object->engine)); 210 union { 211 struct fermi_a_zbc_depth_v0 v0; 212 } *args = data; 213 int ret = -ENOSYS; 214 215 if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, false))) { 216 switch (args->v0.format) { 217 case FERMI_A_ZBC_DEPTH_V0_FMT_FP32: 218 ret = gf100_gr_zbc_depth_get(gr, args->v0.format, 219 args->v0.ds, 220 args->v0.l2); 221 return (ret >= 0) ? 0 : -ENOSPC; 222 default: 223 return -EINVAL; 224 } 225 } 226 227 return ret; 228 } 229 230 static int 231 gf100_fermi_mthd(struct nvkm_object *object, u32 mthd, void *data, u32 size) 232 { 233 nvif_ioctl(object, "fermi mthd %08x\n", mthd); 234 switch (mthd) { 235 case FERMI_A_ZBC_COLOR: 236 return gf100_fermi_mthd_zbc_color(object, data, size); 237 case FERMI_A_ZBC_DEPTH: 238 return gf100_fermi_mthd_zbc_depth(object, data, size); 239 default: 240 break; 241 } 242 return -EINVAL; 243 } 244 245 const struct nvkm_object_func 246 gf100_fermi = { 247 .mthd = gf100_fermi_mthd, 248 }; 249 250 static void 251 gf100_gr_mthd_set_shader_exceptions(struct nvkm_device *device, u32 data) 252 { 253 nvkm_wr32(device, 0x419e44, data ? 0xffffffff : 0x00000000); 254 nvkm_wr32(device, 0x419e4c, data ? 0xffffffff : 0x00000000); 255 } 256 257 static bool 258 gf100_gr_mthd_sw(struct nvkm_device *device, u16 class, u32 mthd, u32 data) 259 { 260 switch (class & 0x00ff) { 261 case 0x97: 262 case 0xc0: 263 switch (mthd) { 264 case 0x1528: 265 gf100_gr_mthd_set_shader_exceptions(device, data); 266 return true; 267 default: 268 break; 269 } 270 break; 271 default: 272 break; 273 } 274 return false; 275 } 276 277 static const struct nvkm_object_func 278 gf100_gr_object_func = { 279 }; 280 281 static int 282 gf100_gr_object_new(const struct nvkm_oclass *oclass, void *data, u32 size, 283 struct nvkm_object **pobject) 284 { 285 struct gf100_gr_chan *chan = gf100_gr_chan(oclass->parent); 286 struct gf100_gr_object *object; 287 288 if (!(object = kzalloc(sizeof(*object), GFP_KERNEL))) 289 return -ENOMEM; 290 *pobject = &object->object; 291 292 nvkm_object_ctor(oclass->base.func ? oclass->base.func : 293 &gf100_gr_object_func, oclass, &object->object); 294 object->chan = chan; 295 return 0; 296 } 297 298 static int 299 gf100_gr_object_get(struct nvkm_gr *base, int index, struct nvkm_sclass *sclass) 300 { 301 struct gf100_gr *gr = gf100_gr(base); 302 int c = 0; 303 304 while (gr->func->sclass[c].oclass) { 305 if (c++ == index) { 306 *sclass = gr->func->sclass[index]; 307 sclass->ctor = gf100_gr_object_new; 308 return index; 309 } 310 } 311 312 return c; 313 } 314 315 /******************************************************************************* 316 * PGRAPH context 317 ******************************************************************************/ 318 319 static int 320 gf100_gr_chan_bind(struct nvkm_object *object, struct nvkm_gpuobj *parent, 321 int align, struct nvkm_gpuobj **pgpuobj) 322 { 323 struct gf100_gr_chan *chan = gf100_gr_chan(object); 324 struct gf100_gr *gr = chan->gr; 325 int ret, i; 326 327 ret = nvkm_gpuobj_new(gr->base.engine.subdev.device, gr->size, 328 align, false, parent, pgpuobj); 329 if (ret) 330 return ret; 331 332 nvkm_kmap(*pgpuobj); 333 for (i = 0; i < gr->size; i += 4) 334 nvkm_wo32(*pgpuobj, i, gr->data[i / 4]); 335 336 if (!gr->firmware) { 337 nvkm_wo32(*pgpuobj, 0x00, chan->mmio_nr / 2); 338 nvkm_wo32(*pgpuobj, 0x04, chan->mmio_vma->addr >> 8); 339 } else { 340 nvkm_wo32(*pgpuobj, 0xf4, 0); 341 nvkm_wo32(*pgpuobj, 0xf8, 0); 342 nvkm_wo32(*pgpuobj, 0x10, chan->mmio_nr / 2); 343 nvkm_wo32(*pgpuobj, 0x14, lower_32_bits(chan->mmio_vma->addr)); 344 nvkm_wo32(*pgpuobj, 0x18, upper_32_bits(chan->mmio_vma->addr)); 345 nvkm_wo32(*pgpuobj, 0x1c, 1); 346 nvkm_wo32(*pgpuobj, 0x20, 0); 347 nvkm_wo32(*pgpuobj, 0x28, 0); 348 nvkm_wo32(*pgpuobj, 0x2c, 0); 349 } 350 nvkm_done(*pgpuobj); 351 return 0; 352 } 353 354 static void * 355 gf100_gr_chan_dtor(struct nvkm_object *object) 356 { 357 struct gf100_gr_chan *chan = gf100_gr_chan(object); 358 int i; 359 360 for (i = 0; i < ARRAY_SIZE(chan->data); i++) { 361 nvkm_vmm_put(chan->vmm, &chan->data[i].vma); 362 nvkm_memory_unref(&chan->data[i].mem); 363 } 364 365 nvkm_vmm_put(chan->vmm, &chan->mmio_vma); 366 nvkm_memory_unref(&chan->mmio); 367 nvkm_vmm_unref(&chan->vmm); 368 return chan; 369 } 370 371 static const struct nvkm_object_func 372 gf100_gr_chan = { 373 .dtor = gf100_gr_chan_dtor, 374 .bind = gf100_gr_chan_bind, 375 }; 376 377 static int 378 gf100_gr_chan_new(struct nvkm_gr *base, struct nvkm_fifo_chan *fifoch, 379 const struct nvkm_oclass *oclass, 380 struct nvkm_object **pobject) 381 { 382 struct gf100_gr *gr = gf100_gr(base); 383 struct gf100_gr_data *data = gr->mmio_data; 384 struct gf100_gr_mmio *mmio = gr->mmio_list; 385 struct gf100_gr_chan *chan; 386 struct gf100_vmm_map_v0 args = { .priv = 1 }; 387 struct nvkm_device *device = gr->base.engine.subdev.device; 388 int ret, i; 389 390 if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) 391 return -ENOMEM; 392 nvkm_object_ctor(&gf100_gr_chan, oclass, &chan->object); 393 chan->gr = gr; 394 chan->vmm = nvkm_vmm_ref(fifoch->vmm); 395 *pobject = &chan->object; 396 397 /* allocate memory for a "mmio list" buffer that's used by the HUB 398 * fuc to modify some per-context register settings on first load 399 * of the context. 400 */ 401 ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, 0x1000, 0x100, 402 false, &chan->mmio); 403 if (ret) 404 return ret; 405 406 ret = nvkm_vmm_get(fifoch->vmm, 12, 0x1000, &chan->mmio_vma); 407 if (ret) 408 return ret; 409 410 ret = nvkm_memory_map(chan->mmio, 0, fifoch->vmm, 411 chan->mmio_vma, &args, sizeof(args)); 412 if (ret) 413 return ret; 414 415 /* allocate buffers referenced by mmio list */ 416 for (i = 0; data->size && i < ARRAY_SIZE(gr->mmio_data); i++) { 417 ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, 418 data->size, data->align, false, 419 &chan->data[i].mem); 420 if (ret) 421 return ret; 422 423 ret = nvkm_vmm_get(fifoch->vmm, 12, 424 nvkm_memory_size(chan->data[i].mem), 425 &chan->data[i].vma); 426 if (ret) 427 return ret; 428 429 args.priv = data->priv; 430 431 ret = nvkm_memory_map(chan->data[i].mem, 0, chan->vmm, 432 chan->data[i].vma, &args, sizeof(args)); 433 if (ret) 434 return ret; 435 436 data++; 437 } 438 439 /* finally, fill in the mmio list and point the context at it */ 440 nvkm_kmap(chan->mmio); 441 for (i = 0; mmio->addr && i < ARRAY_SIZE(gr->mmio_list); i++) { 442 u32 addr = mmio->addr; 443 u32 data = mmio->data; 444 445 if (mmio->buffer >= 0) { 446 u64 info = chan->data[mmio->buffer].vma->addr; 447 data |= info >> mmio->shift; 448 } 449 450 nvkm_wo32(chan->mmio, chan->mmio_nr++ * 4, addr); 451 nvkm_wo32(chan->mmio, chan->mmio_nr++ * 4, data); 452 mmio++; 453 } 454 nvkm_done(chan->mmio); 455 return 0; 456 } 457 458 /******************************************************************************* 459 * PGRAPH register lists 460 ******************************************************************************/ 461 462 const struct gf100_gr_init 463 gf100_gr_init_main_0[] = { 464 { 0x400080, 1, 0x04, 0x003083c2 }, 465 { 0x400088, 1, 0x04, 0x00006fe7 }, 466 { 0x40008c, 1, 0x04, 0x00000000 }, 467 { 0x400090, 1, 0x04, 0x00000030 }, 468 { 0x40013c, 1, 0x04, 0x013901f7 }, 469 { 0x400140, 1, 0x04, 0x00000100 }, 470 { 0x400144, 1, 0x04, 0x00000000 }, 471 { 0x400148, 1, 0x04, 0x00000110 }, 472 { 0x400138, 1, 0x04, 0x00000000 }, 473 { 0x400130, 2, 0x04, 0x00000000 }, 474 { 0x400124, 1, 0x04, 0x00000002 }, 475 {} 476 }; 477 478 const struct gf100_gr_init 479 gf100_gr_init_fe_0[] = { 480 { 0x40415c, 1, 0x04, 0x00000000 }, 481 { 0x404170, 1, 0x04, 0x00000000 }, 482 {} 483 }; 484 485 const struct gf100_gr_init 486 gf100_gr_init_pri_0[] = { 487 { 0x404488, 2, 0x04, 0x00000000 }, 488 {} 489 }; 490 491 const struct gf100_gr_init 492 gf100_gr_init_rstr2d_0[] = { 493 { 0x407808, 1, 0x04, 0x00000000 }, 494 {} 495 }; 496 497 const struct gf100_gr_init 498 gf100_gr_init_pd_0[] = { 499 { 0x406024, 1, 0x04, 0x00000000 }, 500 {} 501 }; 502 503 const struct gf100_gr_init 504 gf100_gr_init_ds_0[] = { 505 { 0x405844, 1, 0x04, 0x00ffffff }, 506 { 0x405850, 1, 0x04, 0x00000000 }, 507 { 0x405908, 1, 0x04, 0x00000000 }, 508 {} 509 }; 510 511 const struct gf100_gr_init 512 gf100_gr_init_scc_0[] = { 513 { 0x40803c, 1, 0x04, 0x00000000 }, 514 {} 515 }; 516 517 const struct gf100_gr_init 518 gf100_gr_init_prop_0[] = { 519 { 0x4184a0, 1, 0x04, 0x00000000 }, 520 {} 521 }; 522 523 const struct gf100_gr_init 524 gf100_gr_init_gpc_unk_0[] = { 525 { 0x418604, 1, 0x04, 0x00000000 }, 526 { 0x418680, 1, 0x04, 0x00000000 }, 527 { 0x418714, 1, 0x04, 0x80000000 }, 528 { 0x418384, 1, 0x04, 0x00000000 }, 529 {} 530 }; 531 532 const struct gf100_gr_init 533 gf100_gr_init_setup_0[] = { 534 { 0x418814, 3, 0x04, 0x00000000 }, 535 {} 536 }; 537 538 const struct gf100_gr_init 539 gf100_gr_init_crstr_0[] = { 540 { 0x418b04, 1, 0x04, 0x00000000 }, 541 {} 542 }; 543 544 const struct gf100_gr_init 545 gf100_gr_init_setup_1[] = { 546 { 0x4188c8, 1, 0x04, 0x80000000 }, 547 { 0x4188cc, 1, 0x04, 0x00000000 }, 548 { 0x4188d0, 1, 0x04, 0x00010000 }, 549 { 0x4188d4, 1, 0x04, 0x00000001 }, 550 {} 551 }; 552 553 const struct gf100_gr_init 554 gf100_gr_init_zcull_0[] = { 555 { 0x418910, 1, 0x04, 0x00010001 }, 556 { 0x418914, 1, 0x04, 0x00000301 }, 557 { 0x418918, 1, 0x04, 0x00800000 }, 558 { 0x418980, 1, 0x04, 0x77777770 }, 559 { 0x418984, 3, 0x04, 0x77777777 }, 560 {} 561 }; 562 563 const struct gf100_gr_init 564 gf100_gr_init_gpm_0[] = { 565 { 0x418c04, 1, 0x04, 0x00000000 }, 566 { 0x418c88, 1, 0x04, 0x00000000 }, 567 {} 568 }; 569 570 const struct gf100_gr_init 571 gf100_gr_init_gpc_unk_1[] = { 572 { 0x418d00, 1, 0x04, 0x00000000 }, 573 { 0x418f08, 1, 0x04, 0x00000000 }, 574 { 0x418e00, 1, 0x04, 0x00000050 }, 575 { 0x418e08, 1, 0x04, 0x00000000 }, 576 {} 577 }; 578 579 const struct gf100_gr_init 580 gf100_gr_init_gcc_0[] = { 581 { 0x41900c, 1, 0x04, 0x00000000 }, 582 { 0x419018, 1, 0x04, 0x00000000 }, 583 {} 584 }; 585 586 const struct gf100_gr_init 587 gf100_gr_init_tpccs_0[] = { 588 { 0x419d08, 2, 0x04, 0x00000000 }, 589 { 0x419d10, 1, 0x04, 0x00000014 }, 590 {} 591 }; 592 593 const struct gf100_gr_init 594 gf100_gr_init_tex_0[] = { 595 { 0x419ab0, 1, 0x04, 0x00000000 }, 596 { 0x419ab8, 1, 0x04, 0x000000e7 }, 597 { 0x419abc, 2, 0x04, 0x00000000 }, 598 {} 599 }; 600 601 const struct gf100_gr_init 602 gf100_gr_init_pe_0[] = { 603 { 0x41980c, 3, 0x04, 0x00000000 }, 604 { 0x419844, 1, 0x04, 0x00000000 }, 605 { 0x41984c, 1, 0x04, 0x00005bc5 }, 606 { 0x419850, 4, 0x04, 0x00000000 }, 607 {} 608 }; 609 610 const struct gf100_gr_init 611 gf100_gr_init_l1c_0[] = { 612 { 0x419c98, 1, 0x04, 0x00000000 }, 613 { 0x419ca8, 1, 0x04, 0x80000000 }, 614 { 0x419cb4, 1, 0x04, 0x00000000 }, 615 { 0x419cb8, 1, 0x04, 0x00008bf4 }, 616 { 0x419cbc, 1, 0x04, 0x28137606 }, 617 { 0x419cc0, 2, 0x04, 0x00000000 }, 618 {} 619 }; 620 621 const struct gf100_gr_init 622 gf100_gr_init_wwdx_0[] = { 623 { 0x419bd4, 1, 0x04, 0x00800000 }, 624 { 0x419bdc, 1, 0x04, 0x00000000 }, 625 {} 626 }; 627 628 const struct gf100_gr_init 629 gf100_gr_init_tpccs_1[] = { 630 { 0x419d2c, 1, 0x04, 0x00000000 }, 631 {} 632 }; 633 634 const struct gf100_gr_init 635 gf100_gr_init_mpc_0[] = { 636 { 0x419c0c, 1, 0x04, 0x00000000 }, 637 {} 638 }; 639 640 static const struct gf100_gr_init 641 gf100_gr_init_sm_0[] = { 642 { 0x419e00, 1, 0x04, 0x00000000 }, 643 { 0x419ea0, 1, 0x04, 0x00000000 }, 644 { 0x419ea4, 1, 0x04, 0x00000100 }, 645 { 0x419ea8, 1, 0x04, 0x00001100 }, 646 { 0x419eac, 1, 0x04, 0x11100702 }, 647 { 0x419eb0, 1, 0x04, 0x00000003 }, 648 { 0x419eb4, 4, 0x04, 0x00000000 }, 649 { 0x419ec8, 1, 0x04, 0x06060618 }, 650 { 0x419ed0, 1, 0x04, 0x0eff0e38 }, 651 { 0x419ed4, 1, 0x04, 0x011104f1 }, 652 { 0x419edc, 1, 0x04, 0x00000000 }, 653 { 0x419f00, 1, 0x04, 0x00000000 }, 654 { 0x419f2c, 1, 0x04, 0x00000000 }, 655 {} 656 }; 657 658 const struct gf100_gr_init 659 gf100_gr_init_be_0[] = { 660 { 0x40880c, 1, 0x04, 0x00000000 }, 661 { 0x408910, 9, 0x04, 0x00000000 }, 662 { 0x408950, 1, 0x04, 0x00000000 }, 663 { 0x408954, 1, 0x04, 0x0000ffff }, 664 { 0x408984, 1, 0x04, 0x00000000 }, 665 { 0x408988, 1, 0x04, 0x08040201 }, 666 { 0x40898c, 1, 0x04, 0x80402010 }, 667 {} 668 }; 669 670 const struct gf100_gr_init 671 gf100_gr_init_fe_1[] = { 672 { 0x4040f0, 1, 0x04, 0x00000000 }, 673 {} 674 }; 675 676 const struct gf100_gr_init 677 gf100_gr_init_pe_1[] = { 678 { 0x419880, 1, 0x04, 0x00000002 }, 679 {} 680 }; 681 682 static const struct gf100_gr_pack 683 gf100_gr_pack_mmio[] = { 684 { gf100_gr_init_main_0 }, 685 { gf100_gr_init_fe_0 }, 686 { gf100_gr_init_pri_0 }, 687 { gf100_gr_init_rstr2d_0 }, 688 { gf100_gr_init_pd_0 }, 689 { gf100_gr_init_ds_0 }, 690 { gf100_gr_init_scc_0 }, 691 { gf100_gr_init_prop_0 }, 692 { gf100_gr_init_gpc_unk_0 }, 693 { gf100_gr_init_setup_0 }, 694 { gf100_gr_init_crstr_0 }, 695 { gf100_gr_init_setup_1 }, 696 { gf100_gr_init_zcull_0 }, 697 { gf100_gr_init_gpm_0 }, 698 { gf100_gr_init_gpc_unk_1 }, 699 { gf100_gr_init_gcc_0 }, 700 { gf100_gr_init_tpccs_0 }, 701 { gf100_gr_init_tex_0 }, 702 { gf100_gr_init_pe_0 }, 703 { gf100_gr_init_l1c_0 }, 704 { gf100_gr_init_wwdx_0 }, 705 { gf100_gr_init_tpccs_1 }, 706 { gf100_gr_init_mpc_0 }, 707 { gf100_gr_init_sm_0 }, 708 { gf100_gr_init_be_0 }, 709 { gf100_gr_init_fe_1 }, 710 { gf100_gr_init_pe_1 }, 711 {} 712 }; 713 714 /******************************************************************************* 715 * PGRAPH engine/subdev functions 716 ******************************************************************************/ 717 718 static u32 719 gf100_gr_ctxsw_inst(struct nvkm_gr *gr) 720 { 721 return nvkm_rd32(gr->engine.subdev.device, 0x409b00); 722 } 723 724 static int 725 gf100_gr_fecs_ctrl_ctxsw(struct gf100_gr *gr, u32 mthd) 726 { 727 struct nvkm_device *device = gr->base.engine.subdev.device; 728 729 nvkm_wr32(device, 0x409804, 0xffffffff); 730 nvkm_wr32(device, 0x409840, 0xffffffff); 731 nvkm_wr32(device, 0x409500, 0xffffffff); 732 nvkm_wr32(device, 0x409504, mthd); 733 nvkm_msec(device, 2000, 734 u32 stat = nvkm_rd32(device, 0x409804); 735 if (stat == 0x00000002) 736 return -EIO; 737 if (stat == 0x00000001) 738 return 0; 739 ); 740 741 return -ETIMEDOUT; 742 } 743 744 int 745 gf100_gr_fecs_start_ctxsw(struct nvkm_gr *base) 746 { 747 struct gf100_gr *gr = gf100_gr(base); 748 int ret = 0; 749 750 mutex_lock(&gr->fecs.mutex); 751 if (!--gr->fecs.disable) { 752 if (WARN_ON(ret = gf100_gr_fecs_ctrl_ctxsw(gr, 0x39))) 753 gr->fecs.disable++; 754 } 755 mutex_unlock(&gr->fecs.mutex); 756 return ret; 757 } 758 759 int 760 gf100_gr_fecs_stop_ctxsw(struct nvkm_gr *base) 761 { 762 struct gf100_gr *gr = gf100_gr(base); 763 int ret = 0; 764 765 mutex_lock(&gr->fecs.mutex); 766 if (!gr->fecs.disable++) { 767 if (WARN_ON(ret = gf100_gr_fecs_ctrl_ctxsw(gr, 0x38))) 768 gr->fecs.disable--; 769 } 770 mutex_unlock(&gr->fecs.mutex); 771 return ret; 772 } 773 774 int 775 gf100_gr_fecs_bind_pointer(struct gf100_gr *gr, u32 inst) 776 { 777 struct nvkm_device *device = gr->base.engine.subdev.device; 778 779 nvkm_wr32(device, 0x409840, 0x00000030); 780 nvkm_wr32(device, 0x409500, inst); 781 nvkm_wr32(device, 0x409504, 0x00000003); 782 nvkm_msec(device, 2000, 783 u32 stat = nvkm_rd32(device, 0x409800); 784 if (stat & 0x00000020) 785 return -EIO; 786 if (stat & 0x00000010) 787 return 0; 788 ); 789 790 return -ETIMEDOUT; 791 } 792 793 static int 794 gf100_gr_fecs_set_reglist_virtual_address(struct gf100_gr *gr, u64 addr) 795 { 796 struct nvkm_device *device = gr->base.engine.subdev.device; 797 798 nvkm_wr32(device, 0x409810, addr >> 8); 799 nvkm_wr32(device, 0x409800, 0x00000000); 800 nvkm_wr32(device, 0x409500, 0x00000001); 801 nvkm_wr32(device, 0x409504, 0x00000032); 802 nvkm_msec(device, 2000, 803 if (nvkm_rd32(device, 0x409800) == 0x00000001) 804 return 0; 805 ); 806 807 return -ETIMEDOUT; 808 } 809 810 static int 811 gf100_gr_fecs_set_reglist_bind_instance(struct gf100_gr *gr, u32 inst) 812 { 813 struct nvkm_device *device = gr->base.engine.subdev.device; 814 815 nvkm_wr32(device, 0x409810, inst); 816 nvkm_wr32(device, 0x409800, 0x00000000); 817 nvkm_wr32(device, 0x409500, 0x00000001); 818 nvkm_wr32(device, 0x409504, 0x00000031); 819 nvkm_msec(device, 2000, 820 if (nvkm_rd32(device, 0x409800) == 0x00000001) 821 return 0; 822 ); 823 824 return -ETIMEDOUT; 825 } 826 827 static int 828 gf100_gr_fecs_discover_reglist_image_size(struct gf100_gr *gr, u32 *psize) 829 { 830 struct nvkm_device *device = gr->base.engine.subdev.device; 831 832 nvkm_wr32(device, 0x409800, 0x00000000); 833 nvkm_wr32(device, 0x409500, 0x00000001); 834 nvkm_wr32(device, 0x409504, 0x00000030); 835 nvkm_msec(device, 2000, 836 if ((*psize = nvkm_rd32(device, 0x409800))) 837 return 0; 838 ); 839 840 return -ETIMEDOUT; 841 } 842 843 static int 844 gf100_gr_fecs_elpg_bind(struct gf100_gr *gr) 845 { 846 u32 size; 847 int ret; 848 849 ret = gf100_gr_fecs_discover_reglist_image_size(gr, &size); 850 if (ret) 851 return ret; 852 853 /*XXX: We need to allocate + map the above into PMU's inst block, 854 * which which means we probably need a proper PMU before we 855 * even bother. 856 */ 857 858 ret = gf100_gr_fecs_set_reglist_bind_instance(gr, 0); 859 if (ret) 860 return ret; 861 862 return gf100_gr_fecs_set_reglist_virtual_address(gr, 0); 863 } 864 865 static int 866 gf100_gr_fecs_discover_pm_image_size(struct gf100_gr *gr, u32 *psize) 867 { 868 struct nvkm_device *device = gr->base.engine.subdev.device; 869 870 nvkm_wr32(device, 0x409840, 0xffffffff); 871 nvkm_wr32(device, 0x409500, 0x00000000); 872 nvkm_wr32(device, 0x409504, 0x00000025); 873 nvkm_msec(device, 2000, 874 if ((*psize = nvkm_rd32(device, 0x409800))) 875 return 0; 876 ); 877 878 return -ETIMEDOUT; 879 } 880 881 static int 882 gf100_gr_fecs_discover_zcull_image_size(struct gf100_gr *gr, u32 *psize) 883 { 884 struct nvkm_device *device = gr->base.engine.subdev.device; 885 886 nvkm_wr32(device, 0x409840, 0xffffffff); 887 nvkm_wr32(device, 0x409500, 0x00000000); 888 nvkm_wr32(device, 0x409504, 0x00000016); 889 nvkm_msec(device, 2000, 890 if ((*psize = nvkm_rd32(device, 0x409800))) 891 return 0; 892 ); 893 894 return -ETIMEDOUT; 895 } 896 897 static int 898 gf100_gr_fecs_discover_image_size(struct gf100_gr *gr, u32 *psize) 899 { 900 struct nvkm_device *device = gr->base.engine.subdev.device; 901 902 nvkm_wr32(device, 0x409840, 0xffffffff); 903 nvkm_wr32(device, 0x409500, 0x00000000); 904 nvkm_wr32(device, 0x409504, 0x00000010); 905 nvkm_msec(device, 2000, 906 if ((*psize = nvkm_rd32(device, 0x409800))) 907 return 0; 908 ); 909 910 return -ETIMEDOUT; 911 } 912 913 static void 914 gf100_gr_fecs_set_watchdog_timeout(struct gf100_gr *gr, u32 timeout) 915 { 916 struct nvkm_device *device = gr->base.engine.subdev.device; 917 918 nvkm_wr32(device, 0x409840, 0xffffffff); 919 nvkm_wr32(device, 0x409500, timeout); 920 nvkm_wr32(device, 0x409504, 0x00000021); 921 } 922 923 static bool 924 gf100_gr_chsw_load(struct nvkm_gr *base) 925 { 926 struct gf100_gr *gr = gf100_gr(base); 927 if (!gr->firmware) { 928 u32 trace = nvkm_rd32(gr->base.engine.subdev.device, 0x40981c); 929 if (trace & 0x00000040) 930 return true; 931 } else { 932 u32 mthd = nvkm_rd32(gr->base.engine.subdev.device, 0x409808); 933 if (mthd & 0x00080000) 934 return true; 935 } 936 return false; 937 } 938 939 int 940 gf100_gr_rops(struct gf100_gr *gr) 941 { 942 struct nvkm_device *device = gr->base.engine.subdev.device; 943 return (nvkm_rd32(device, 0x409604) & 0x001f0000) >> 16; 944 } 945 946 void 947 gf100_gr_zbc_init(struct gf100_gr *gr) 948 { 949 const u32 zero[] = { 0x00000000, 0x00000000, 0x00000000, 0x00000000, 950 0x00000000, 0x00000000, 0x00000000, 0x00000000 }; 951 const u32 one[] = { 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 952 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }; 953 const u32 f32_0[] = { 0x00000000, 0x00000000, 0x00000000, 0x00000000, 954 0x00000000, 0x00000000, 0x00000000, 0x00000000 }; 955 const u32 f32_1[] = { 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 956 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000 }; 957 struct nvkm_ltc *ltc = gr->base.engine.subdev.device->ltc; 958 int index, c = ltc->zbc_min, d = ltc->zbc_min, s = ltc->zbc_min; 959 960 if (!gr->zbc_color[0].format) { 961 gf100_gr_zbc_color_get(gr, 1, & zero[0], &zero[4]); c++; 962 gf100_gr_zbc_color_get(gr, 2, & one[0], &one[4]); c++; 963 gf100_gr_zbc_color_get(gr, 4, &f32_0[0], &f32_0[4]); c++; 964 gf100_gr_zbc_color_get(gr, 4, &f32_1[0], &f32_1[4]); c++; 965 gf100_gr_zbc_depth_get(gr, 1, 0x00000000, 0x00000000); d++; 966 gf100_gr_zbc_depth_get(gr, 1, 0x3f800000, 0x3f800000); d++; 967 if (gr->func->zbc->stencil_get) { 968 gr->func->zbc->stencil_get(gr, 1, 0x00, 0x00); s++; 969 gr->func->zbc->stencil_get(gr, 1, 0x01, 0x01); s++; 970 gr->func->zbc->stencil_get(gr, 1, 0xff, 0xff); s++; 971 } 972 } 973 974 for (index = c; index <= ltc->zbc_max; index++) 975 gr->func->zbc->clear_color(gr, index); 976 for (index = d; index <= ltc->zbc_max; index++) 977 gr->func->zbc->clear_depth(gr, index); 978 979 if (gr->func->zbc->clear_stencil) { 980 for (index = s; index <= ltc->zbc_max; index++) 981 gr->func->zbc->clear_stencil(gr, index); 982 } 983 } 984 985 /** 986 * Wait until GR goes idle. GR is considered idle if it is disabled by the 987 * MC (0x200) register, or GR is not busy and a context switch is not in 988 * progress. 989 */ 990 int 991 gf100_gr_wait_idle(struct gf100_gr *gr) 992 { 993 struct nvkm_subdev *subdev = &gr->base.engine.subdev; 994 struct nvkm_device *device = subdev->device; 995 unsigned long end_jiffies = jiffies + msecs_to_jiffies(2000); 996 bool gr_enabled, ctxsw_active, gr_busy; 997 998 do { 999 /* 1000 * required to make sure FIFO_ENGINE_STATUS (0x2640) is 1001 * up-to-date 1002 */ 1003 nvkm_rd32(device, 0x400700); 1004 1005 gr_enabled = nvkm_rd32(device, 0x200) & 0x1000; 1006 ctxsw_active = nvkm_rd32(device, 0x2640) & 0x8000; 1007 gr_busy = nvkm_rd32(device, 0x40060c) & 0x1; 1008 1009 if (!gr_enabled || (!gr_busy && !ctxsw_active)) 1010 return 0; 1011 } while (time_before(jiffies, end_jiffies)); 1012 1013 nvkm_error(subdev, 1014 "wait for idle timeout (en: %d, ctxsw: %d, busy: %d)\n", 1015 gr_enabled, ctxsw_active, gr_busy); 1016 return -EAGAIN; 1017 } 1018 1019 void 1020 gf100_gr_mmio(struct gf100_gr *gr, const struct gf100_gr_pack *p) 1021 { 1022 struct nvkm_device *device = gr->base.engine.subdev.device; 1023 const struct gf100_gr_pack *pack; 1024 const struct gf100_gr_init *init; 1025 1026 pack_for_each_init(init, pack, p) { 1027 u32 next = init->addr + init->count * init->pitch; 1028 u32 addr = init->addr; 1029 while (addr < next) { 1030 nvkm_wr32(device, addr, init->data); 1031 addr += init->pitch; 1032 } 1033 } 1034 } 1035 1036 void 1037 gf100_gr_icmd(struct gf100_gr *gr, const struct gf100_gr_pack *p) 1038 { 1039 struct nvkm_device *device = gr->base.engine.subdev.device; 1040 const struct gf100_gr_pack *pack; 1041 const struct gf100_gr_init *init; 1042 u32 data = 0; 1043 1044 nvkm_wr32(device, 0x400208, 0x80000000); 1045 1046 pack_for_each_init(init, pack, p) { 1047 u32 next = init->addr + init->count * init->pitch; 1048 u32 addr = init->addr; 1049 1050 if ((pack == p && init == p->init) || data != init->data) { 1051 nvkm_wr32(device, 0x400204, init->data); 1052 data = init->data; 1053 } 1054 1055 while (addr < next) { 1056 nvkm_wr32(device, 0x400200, addr); 1057 /** 1058 * Wait for GR to go idle after submitting a 1059 * GO_IDLE bundle 1060 */ 1061 if ((addr & 0xffff) == 0xe100) 1062 gf100_gr_wait_idle(gr); 1063 nvkm_msec(device, 2000, 1064 if (!(nvkm_rd32(device, 0x400700) & 0x00000004)) 1065 break; 1066 ); 1067 addr += init->pitch; 1068 } 1069 } 1070 1071 nvkm_wr32(device, 0x400208, 0x00000000); 1072 } 1073 1074 void 1075 gf100_gr_mthd(struct gf100_gr *gr, const struct gf100_gr_pack *p) 1076 { 1077 struct nvkm_device *device = gr->base.engine.subdev.device; 1078 const struct gf100_gr_pack *pack; 1079 const struct gf100_gr_init *init; 1080 u32 data = 0; 1081 1082 pack_for_each_init(init, pack, p) { 1083 u32 ctrl = 0x80000000 | pack->type; 1084 u32 next = init->addr + init->count * init->pitch; 1085 u32 addr = init->addr; 1086 1087 if ((pack == p && init == p->init) || data != init->data) { 1088 nvkm_wr32(device, 0x40448c, init->data); 1089 data = init->data; 1090 } 1091 1092 while (addr < next) { 1093 nvkm_wr32(device, 0x404488, ctrl | (addr << 14)); 1094 addr += init->pitch; 1095 } 1096 } 1097 } 1098 1099 u64 1100 gf100_gr_units(struct nvkm_gr *base) 1101 { 1102 struct gf100_gr *gr = gf100_gr(base); 1103 u64 cfg; 1104 1105 cfg = (u32)gr->gpc_nr; 1106 cfg |= (u32)gr->tpc_total << 8; 1107 cfg |= (u64)gr->rop_nr << 32; 1108 1109 return cfg; 1110 } 1111 1112 static const struct nvkm_bitfield gf100_dispatch_error[] = { 1113 { 0x00000001, "INJECTED_BUNDLE_ERROR" }, 1114 { 0x00000002, "CLASS_SUBCH_MISMATCH" }, 1115 { 0x00000004, "SUBCHSW_DURING_NOTIFY" }, 1116 {} 1117 }; 1118 1119 static const struct nvkm_bitfield gf100_m2mf_error[] = { 1120 { 0x00000001, "PUSH_TOO_MUCH_DATA" }, 1121 { 0x00000002, "PUSH_NOT_ENOUGH_DATA" }, 1122 {} 1123 }; 1124 1125 static const struct nvkm_bitfield gf100_unk6_error[] = { 1126 { 0x00000001, "TEMP_TOO_SMALL" }, 1127 {} 1128 }; 1129 1130 static const struct nvkm_bitfield gf100_ccache_error[] = { 1131 { 0x00000001, "INTR" }, 1132 { 0x00000002, "LDCONST_OOB" }, 1133 {} 1134 }; 1135 1136 static const struct nvkm_bitfield gf100_macro_error[] = { 1137 { 0x00000001, "TOO_FEW_PARAMS" }, 1138 { 0x00000002, "TOO_MANY_PARAMS" }, 1139 { 0x00000004, "ILLEGAL_OPCODE" }, 1140 { 0x00000008, "DOUBLE_BRANCH" }, 1141 { 0x00000010, "WATCHDOG" }, 1142 {} 1143 }; 1144 1145 static const struct nvkm_bitfield gk104_sked_error[] = { 1146 { 0x00000040, "CTA_RESUME" }, 1147 { 0x00000080, "CONSTANT_BUFFER_SIZE" }, 1148 { 0x00000200, "LOCAL_MEMORY_SIZE_POS" }, 1149 { 0x00000400, "LOCAL_MEMORY_SIZE_NEG" }, 1150 { 0x00000800, "WARP_CSTACK_SIZE" }, 1151 { 0x00001000, "TOTAL_TEMP_SIZE" }, 1152 { 0x00002000, "REGISTER_COUNT" }, 1153 { 0x00040000, "TOTAL_THREADS" }, 1154 { 0x00100000, "PROGRAM_OFFSET" }, 1155 { 0x00200000, "SHARED_MEMORY_SIZE" }, 1156 { 0x00800000, "CTA_THREAD_DIMENSION_ZERO" }, 1157 { 0x01000000, "MEMORY_WINDOW_OVERLAP" }, 1158 { 0x02000000, "SHARED_CONFIG_TOO_SMALL" }, 1159 { 0x04000000, "TOTAL_REGISTER_COUNT" }, 1160 {} 1161 }; 1162 1163 static const struct nvkm_bitfield gf100_gpc_rop_error[] = { 1164 { 0x00000002, "RT_PITCH_OVERRUN" }, 1165 { 0x00000010, "RT_WIDTH_OVERRUN" }, 1166 { 0x00000020, "RT_HEIGHT_OVERRUN" }, 1167 { 0x00000080, "ZETA_STORAGE_TYPE_MISMATCH" }, 1168 { 0x00000100, "RT_STORAGE_TYPE_MISMATCH" }, 1169 { 0x00000400, "RT_LINEAR_MISMATCH" }, 1170 {} 1171 }; 1172 1173 static void 1174 gf100_gr_trap_gpc_rop(struct gf100_gr *gr, int gpc) 1175 { 1176 struct nvkm_subdev *subdev = &gr->base.engine.subdev; 1177 struct nvkm_device *device = subdev->device; 1178 char error[128]; 1179 u32 trap[4]; 1180 1181 trap[0] = nvkm_rd32(device, GPC_UNIT(gpc, 0x0420)) & 0x3fffffff; 1182 trap[1] = nvkm_rd32(device, GPC_UNIT(gpc, 0x0434)); 1183 trap[2] = nvkm_rd32(device, GPC_UNIT(gpc, 0x0438)); 1184 trap[3] = nvkm_rd32(device, GPC_UNIT(gpc, 0x043c)); 1185 1186 nvkm_snprintbf(error, sizeof(error), gf100_gpc_rop_error, trap[0]); 1187 1188 nvkm_error(subdev, "GPC%d/PROP trap: %08x [%s] x = %u, y = %u, " 1189 "format = %x, storage type = %x\n", 1190 gpc, trap[0], error, trap[1] & 0xffff, trap[1] >> 16, 1191 (trap[2] >> 8) & 0x3f, trap[3] & 0xff); 1192 nvkm_wr32(device, GPC_UNIT(gpc, 0x0420), 0xc0000000); 1193 } 1194 1195 const struct nvkm_enum gf100_mp_warp_error[] = { 1196 { 0x01, "STACK_ERROR" }, 1197 { 0x02, "API_STACK_ERROR" }, 1198 { 0x03, "RET_EMPTY_STACK_ERROR" }, 1199 { 0x04, "PC_WRAP" }, 1200 { 0x05, "MISALIGNED_PC" }, 1201 { 0x06, "PC_OVERFLOW" }, 1202 { 0x07, "MISALIGNED_IMMC_ADDR" }, 1203 { 0x08, "MISALIGNED_REG" }, 1204 { 0x09, "ILLEGAL_INSTR_ENCODING" }, 1205 { 0x0a, "ILLEGAL_SPH_INSTR_COMBO" }, 1206 { 0x0b, "ILLEGAL_INSTR_PARAM" }, 1207 { 0x0c, "INVALID_CONST_ADDR" }, 1208 { 0x0d, "OOR_REG" }, 1209 { 0x0e, "OOR_ADDR" }, 1210 { 0x0f, "MISALIGNED_ADDR" }, 1211 { 0x10, "INVALID_ADDR_SPACE" }, 1212 { 0x11, "ILLEGAL_INSTR_PARAM2" }, 1213 { 0x12, "INVALID_CONST_ADDR_LDC" }, 1214 { 0x13, "GEOMETRY_SM_ERROR" }, 1215 { 0x14, "DIVERGENT" }, 1216 { 0x15, "WARP_EXIT" }, 1217 {} 1218 }; 1219 1220 const struct nvkm_bitfield gf100_mp_global_error[] = { 1221 { 0x00000001, "SM_TO_SM_FAULT" }, 1222 { 0x00000002, "L1_ERROR" }, 1223 { 0x00000004, "MULTIPLE_WARP_ERRORS" }, 1224 { 0x00000008, "PHYSICAL_STACK_OVERFLOW" }, 1225 { 0x00000010, "BPT_INT" }, 1226 { 0x00000020, "BPT_PAUSE" }, 1227 { 0x00000040, "SINGLE_STEP_COMPLETE" }, 1228 { 0x20000000, "ECC_SEC_ERROR" }, 1229 { 0x40000000, "ECC_DED_ERROR" }, 1230 { 0x80000000, "TIMEOUT" }, 1231 {} 1232 }; 1233 1234 void 1235 gf100_gr_trap_mp(struct gf100_gr *gr, int gpc, int tpc) 1236 { 1237 struct nvkm_subdev *subdev = &gr->base.engine.subdev; 1238 struct nvkm_device *device = subdev->device; 1239 u32 werr = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x648)); 1240 u32 gerr = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x650)); 1241 const struct nvkm_enum *warp; 1242 char glob[128]; 1243 1244 nvkm_snprintbf(glob, sizeof(glob), gf100_mp_global_error, gerr); 1245 warp = nvkm_enum_find(gf100_mp_warp_error, werr & 0xffff); 1246 1247 nvkm_error(subdev, "GPC%i/TPC%i/MP trap: " 1248 "global %08x [%s] warp %04x [%s]\n", 1249 gpc, tpc, gerr, glob, werr, warp ? warp->name : ""); 1250 1251 nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x648), 0x00000000); 1252 nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x650), gerr); 1253 } 1254 1255 static void 1256 gf100_gr_trap_tpc(struct gf100_gr *gr, int gpc, int tpc) 1257 { 1258 struct nvkm_subdev *subdev = &gr->base.engine.subdev; 1259 struct nvkm_device *device = subdev->device; 1260 u32 stat = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x0508)); 1261 1262 if (stat & 0x00000001) { 1263 u32 trap = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x0224)); 1264 nvkm_error(subdev, "GPC%d/TPC%d/TEX: %08x\n", gpc, tpc, trap); 1265 nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x0224), 0xc0000000); 1266 stat &= ~0x00000001; 1267 } 1268 1269 if (stat & 0x00000002) { 1270 gr->func->trap_mp(gr, gpc, tpc); 1271 stat &= ~0x00000002; 1272 } 1273 1274 if (stat & 0x00000004) { 1275 u32 trap = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x0084)); 1276 nvkm_error(subdev, "GPC%d/TPC%d/POLY: %08x\n", gpc, tpc, trap); 1277 nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x0084), 0xc0000000); 1278 stat &= ~0x00000004; 1279 } 1280 1281 if (stat & 0x00000008) { 1282 u32 trap = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x048c)); 1283 nvkm_error(subdev, "GPC%d/TPC%d/L1C: %08x\n", gpc, tpc, trap); 1284 nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x048c), 0xc0000000); 1285 stat &= ~0x00000008; 1286 } 1287 1288 if (stat & 0x00000010) { 1289 u32 trap = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x0430)); 1290 nvkm_error(subdev, "GPC%d/TPC%d/MPC: %08x\n", gpc, tpc, trap); 1291 nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x0430), 0xc0000000); 1292 stat &= ~0x00000010; 1293 } 1294 1295 if (stat) { 1296 nvkm_error(subdev, "GPC%d/TPC%d/%08x: unknown\n", gpc, tpc, stat); 1297 } 1298 } 1299 1300 static void 1301 gf100_gr_trap_gpc(struct gf100_gr *gr, int gpc) 1302 { 1303 struct nvkm_subdev *subdev = &gr->base.engine.subdev; 1304 struct nvkm_device *device = subdev->device; 1305 u32 stat = nvkm_rd32(device, GPC_UNIT(gpc, 0x2c90)); 1306 int tpc; 1307 1308 if (stat & 0x00000001) { 1309 gf100_gr_trap_gpc_rop(gr, gpc); 1310 stat &= ~0x00000001; 1311 } 1312 1313 if (stat & 0x00000002) { 1314 u32 trap = nvkm_rd32(device, GPC_UNIT(gpc, 0x0900)); 1315 nvkm_error(subdev, "GPC%d/ZCULL: %08x\n", gpc, trap); 1316 nvkm_wr32(device, GPC_UNIT(gpc, 0x0900), 0xc0000000); 1317 stat &= ~0x00000002; 1318 } 1319 1320 if (stat & 0x00000004) { 1321 u32 trap = nvkm_rd32(device, GPC_UNIT(gpc, 0x1028)); 1322 nvkm_error(subdev, "GPC%d/CCACHE: %08x\n", gpc, trap); 1323 nvkm_wr32(device, GPC_UNIT(gpc, 0x1028), 0xc0000000); 1324 stat &= ~0x00000004; 1325 } 1326 1327 if (stat & 0x00000008) { 1328 u32 trap = nvkm_rd32(device, GPC_UNIT(gpc, 0x0824)); 1329 nvkm_error(subdev, "GPC%d/ESETUP: %08x\n", gpc, trap); 1330 nvkm_wr32(device, GPC_UNIT(gpc, 0x0824), 0xc0000000); 1331 stat &= ~0x00000009; 1332 } 1333 1334 for (tpc = 0; tpc < gr->tpc_nr[gpc]; tpc++) { 1335 u32 mask = 0x00010000 << tpc; 1336 if (stat & mask) { 1337 gf100_gr_trap_tpc(gr, gpc, tpc); 1338 nvkm_wr32(device, GPC_UNIT(gpc, 0x2c90), mask); 1339 stat &= ~mask; 1340 } 1341 } 1342 1343 if (stat) { 1344 nvkm_error(subdev, "GPC%d/%08x: unknown\n", gpc, stat); 1345 } 1346 } 1347 1348 static void 1349 gf100_gr_trap_intr(struct gf100_gr *gr) 1350 { 1351 struct nvkm_subdev *subdev = &gr->base.engine.subdev; 1352 struct nvkm_device *device = subdev->device; 1353 char error[128]; 1354 u32 trap = nvkm_rd32(device, 0x400108); 1355 int rop, gpc; 1356 1357 if (trap & 0x00000001) { 1358 u32 stat = nvkm_rd32(device, 0x404000); 1359 1360 nvkm_snprintbf(error, sizeof(error), gf100_dispatch_error, 1361 stat & 0x3fffffff); 1362 nvkm_error(subdev, "DISPATCH %08x [%s]\n", stat, error); 1363 nvkm_wr32(device, 0x404000, 0xc0000000); 1364 nvkm_wr32(device, 0x400108, 0x00000001); 1365 trap &= ~0x00000001; 1366 } 1367 1368 if (trap & 0x00000002) { 1369 u32 stat = nvkm_rd32(device, 0x404600); 1370 1371 nvkm_snprintbf(error, sizeof(error), gf100_m2mf_error, 1372 stat & 0x3fffffff); 1373 nvkm_error(subdev, "M2MF %08x [%s]\n", stat, error); 1374 1375 nvkm_wr32(device, 0x404600, 0xc0000000); 1376 nvkm_wr32(device, 0x400108, 0x00000002); 1377 trap &= ~0x00000002; 1378 } 1379 1380 if (trap & 0x00000008) { 1381 u32 stat = nvkm_rd32(device, 0x408030); 1382 1383 nvkm_snprintbf(error, sizeof(error), gf100_ccache_error, 1384 stat & 0x3fffffff); 1385 nvkm_error(subdev, "CCACHE %08x [%s]\n", stat, error); 1386 nvkm_wr32(device, 0x408030, 0xc0000000); 1387 nvkm_wr32(device, 0x400108, 0x00000008); 1388 trap &= ~0x00000008; 1389 } 1390 1391 if (trap & 0x00000010) { 1392 u32 stat = nvkm_rd32(device, 0x405840); 1393 nvkm_error(subdev, "SHADER %08x, sph: 0x%06x, stage: 0x%02x\n", 1394 stat, stat & 0xffffff, (stat >> 24) & 0x3f); 1395 nvkm_wr32(device, 0x405840, 0xc0000000); 1396 nvkm_wr32(device, 0x400108, 0x00000010); 1397 trap &= ~0x00000010; 1398 } 1399 1400 if (trap & 0x00000040) { 1401 u32 stat = nvkm_rd32(device, 0x40601c); 1402 1403 nvkm_snprintbf(error, sizeof(error), gf100_unk6_error, 1404 stat & 0x3fffffff); 1405 nvkm_error(subdev, "UNK6 %08x [%s]\n", stat, error); 1406 1407 nvkm_wr32(device, 0x40601c, 0xc0000000); 1408 nvkm_wr32(device, 0x400108, 0x00000040); 1409 trap &= ~0x00000040; 1410 } 1411 1412 if (trap & 0x00000080) { 1413 u32 stat = nvkm_rd32(device, 0x404490); 1414 u32 pc = nvkm_rd32(device, 0x404494); 1415 u32 op = nvkm_rd32(device, 0x40449c); 1416 1417 nvkm_snprintbf(error, sizeof(error), gf100_macro_error, 1418 stat & 0x1fffffff); 1419 nvkm_error(subdev, "MACRO %08x [%s], pc: 0x%03x%s, op: 0x%08x\n", 1420 stat, error, pc & 0x7ff, 1421 (pc & 0x10000000) ? "" : " (invalid)", 1422 op); 1423 1424 nvkm_wr32(device, 0x404490, 0xc0000000); 1425 nvkm_wr32(device, 0x400108, 0x00000080); 1426 trap &= ~0x00000080; 1427 } 1428 1429 if (trap & 0x00000100) { 1430 u32 stat = nvkm_rd32(device, 0x407020) & 0x3fffffff; 1431 1432 nvkm_snprintbf(error, sizeof(error), gk104_sked_error, stat); 1433 nvkm_error(subdev, "SKED: %08x [%s]\n", stat, error); 1434 1435 if (stat) 1436 nvkm_wr32(device, 0x407020, 0x40000000); 1437 nvkm_wr32(device, 0x400108, 0x00000100); 1438 trap &= ~0x00000100; 1439 } 1440 1441 if (trap & 0x01000000) { 1442 u32 stat = nvkm_rd32(device, 0x400118); 1443 for (gpc = 0; stat && gpc < gr->gpc_nr; gpc++) { 1444 u32 mask = 0x00000001 << gpc; 1445 if (stat & mask) { 1446 gf100_gr_trap_gpc(gr, gpc); 1447 nvkm_wr32(device, 0x400118, mask); 1448 stat &= ~mask; 1449 } 1450 } 1451 nvkm_wr32(device, 0x400108, 0x01000000); 1452 trap &= ~0x01000000; 1453 } 1454 1455 if (trap & 0x02000000) { 1456 for (rop = 0; rop < gr->rop_nr; rop++) { 1457 u32 statz = nvkm_rd32(device, ROP_UNIT(rop, 0x070)); 1458 u32 statc = nvkm_rd32(device, ROP_UNIT(rop, 0x144)); 1459 nvkm_error(subdev, "ROP%d %08x %08x\n", 1460 rop, statz, statc); 1461 nvkm_wr32(device, ROP_UNIT(rop, 0x070), 0xc0000000); 1462 nvkm_wr32(device, ROP_UNIT(rop, 0x144), 0xc0000000); 1463 } 1464 nvkm_wr32(device, 0x400108, 0x02000000); 1465 trap &= ~0x02000000; 1466 } 1467 1468 if (trap) { 1469 nvkm_error(subdev, "TRAP UNHANDLED %08x\n", trap); 1470 nvkm_wr32(device, 0x400108, trap); 1471 } 1472 } 1473 1474 static void 1475 gf100_gr_ctxctl_debug_unit(struct gf100_gr *gr, u32 base) 1476 { 1477 struct nvkm_subdev *subdev = &gr->base.engine.subdev; 1478 struct nvkm_device *device = subdev->device; 1479 nvkm_error(subdev, "%06x - done %08x\n", base, 1480 nvkm_rd32(device, base + 0x400)); 1481 nvkm_error(subdev, "%06x - stat %08x %08x %08x %08x\n", base, 1482 nvkm_rd32(device, base + 0x800), 1483 nvkm_rd32(device, base + 0x804), 1484 nvkm_rd32(device, base + 0x808), 1485 nvkm_rd32(device, base + 0x80c)); 1486 nvkm_error(subdev, "%06x - stat %08x %08x %08x %08x\n", base, 1487 nvkm_rd32(device, base + 0x810), 1488 nvkm_rd32(device, base + 0x814), 1489 nvkm_rd32(device, base + 0x818), 1490 nvkm_rd32(device, base + 0x81c)); 1491 } 1492 1493 void 1494 gf100_gr_ctxctl_debug(struct gf100_gr *gr) 1495 { 1496 struct nvkm_device *device = gr->base.engine.subdev.device; 1497 u32 gpcnr = nvkm_rd32(device, 0x409604) & 0xffff; 1498 u32 gpc; 1499 1500 gf100_gr_ctxctl_debug_unit(gr, 0x409000); 1501 for (gpc = 0; gpc < gpcnr; gpc++) 1502 gf100_gr_ctxctl_debug_unit(gr, 0x502000 + (gpc * 0x8000)); 1503 } 1504 1505 static void 1506 gf100_gr_ctxctl_isr(struct gf100_gr *gr) 1507 { 1508 struct nvkm_subdev *subdev = &gr->base.engine.subdev; 1509 struct nvkm_device *device = subdev->device; 1510 u32 stat = nvkm_rd32(device, 0x409c18); 1511 1512 if (!gr->firmware && (stat & 0x00000001)) { 1513 u32 code = nvkm_rd32(device, 0x409814); 1514 if (code == E_BAD_FWMTHD) { 1515 u32 class = nvkm_rd32(device, 0x409808); 1516 u32 addr = nvkm_rd32(device, 0x40980c); 1517 u32 subc = (addr & 0x00070000) >> 16; 1518 u32 mthd = (addr & 0x00003ffc); 1519 u32 data = nvkm_rd32(device, 0x409810); 1520 1521 nvkm_error(subdev, "FECS MTHD subc %d class %04x " 1522 "mthd %04x data %08x\n", 1523 subc, class, mthd, data); 1524 } else { 1525 nvkm_error(subdev, "FECS ucode error %d\n", code); 1526 } 1527 nvkm_wr32(device, 0x409c20, 0x00000001); 1528 stat &= ~0x00000001; 1529 } 1530 1531 if (!gr->firmware && (stat & 0x00080000)) { 1532 nvkm_error(subdev, "FECS watchdog timeout\n"); 1533 gf100_gr_ctxctl_debug(gr); 1534 nvkm_wr32(device, 0x409c20, 0x00080000); 1535 stat &= ~0x00080000; 1536 } 1537 1538 if (stat) { 1539 nvkm_error(subdev, "FECS %08x\n", stat); 1540 gf100_gr_ctxctl_debug(gr); 1541 nvkm_wr32(device, 0x409c20, stat); 1542 } 1543 } 1544 1545 static void 1546 gf100_gr_intr(struct nvkm_gr *base) 1547 { 1548 struct gf100_gr *gr = gf100_gr(base); 1549 struct nvkm_subdev *subdev = &gr->base.engine.subdev; 1550 struct nvkm_device *device = subdev->device; 1551 struct nvkm_fifo_chan *chan; 1552 unsigned long flags; 1553 u64 inst = nvkm_rd32(device, 0x409b00) & 0x0fffffff; 1554 u32 stat = nvkm_rd32(device, 0x400100); 1555 u32 addr = nvkm_rd32(device, 0x400704); 1556 u32 mthd = (addr & 0x00003ffc); 1557 u32 subc = (addr & 0x00070000) >> 16; 1558 u32 data = nvkm_rd32(device, 0x400708); 1559 u32 code = nvkm_rd32(device, 0x400110); 1560 u32 class; 1561 const char *name = "unknown"; 1562 int chid = -1; 1563 1564 chan = nvkm_fifo_chan_inst(device->fifo, (u64)inst << 12, &flags); 1565 if (chan) { 1566 name = chan->object.client->name; 1567 chid = chan->chid; 1568 } 1569 1570 if (device->card_type < NV_E0 || subc < 4) 1571 class = nvkm_rd32(device, 0x404200 + (subc * 4)); 1572 else 1573 class = 0x0000; 1574 1575 if (stat & 0x00000001) { 1576 /* 1577 * notifier interrupt, only needed for cyclestats 1578 * can be safely ignored 1579 */ 1580 nvkm_wr32(device, 0x400100, 0x00000001); 1581 stat &= ~0x00000001; 1582 } 1583 1584 if (stat & 0x00000010) { 1585 if (!gf100_gr_mthd_sw(device, class, mthd, data)) { 1586 nvkm_error(subdev, "ILLEGAL_MTHD ch %d [%010llx %s] " 1587 "subc %d class %04x mthd %04x data %08x\n", 1588 chid, inst << 12, name, subc, 1589 class, mthd, data); 1590 } 1591 nvkm_wr32(device, 0x400100, 0x00000010); 1592 stat &= ~0x00000010; 1593 } 1594 1595 if (stat & 0x00000020) { 1596 nvkm_error(subdev, "ILLEGAL_CLASS ch %d [%010llx %s] " 1597 "subc %d class %04x mthd %04x data %08x\n", 1598 chid, inst << 12, name, subc, class, mthd, data); 1599 nvkm_wr32(device, 0x400100, 0x00000020); 1600 stat &= ~0x00000020; 1601 } 1602 1603 if (stat & 0x00100000) { 1604 const struct nvkm_enum *en = 1605 nvkm_enum_find(nv50_data_error_names, code); 1606 nvkm_error(subdev, "DATA_ERROR %08x [%s] ch %d [%010llx %s] " 1607 "subc %d class %04x mthd %04x data %08x\n", 1608 code, en ? en->name : "", chid, inst << 12, 1609 name, subc, class, mthd, data); 1610 nvkm_wr32(device, 0x400100, 0x00100000); 1611 stat &= ~0x00100000; 1612 } 1613 1614 if (stat & 0x00200000) { 1615 nvkm_error(subdev, "TRAP ch %d [%010llx %s]\n", 1616 chid, inst << 12, name); 1617 gf100_gr_trap_intr(gr); 1618 nvkm_wr32(device, 0x400100, 0x00200000); 1619 stat &= ~0x00200000; 1620 } 1621 1622 if (stat & 0x00080000) { 1623 gf100_gr_ctxctl_isr(gr); 1624 nvkm_wr32(device, 0x400100, 0x00080000); 1625 stat &= ~0x00080000; 1626 } 1627 1628 if (stat) { 1629 nvkm_error(subdev, "intr %08x\n", stat); 1630 nvkm_wr32(device, 0x400100, stat); 1631 } 1632 1633 nvkm_wr32(device, 0x400500, 0x00010001); 1634 nvkm_fifo_chan_put(device->fifo, flags, &chan); 1635 } 1636 1637 static void 1638 gf100_gr_init_fw(struct nvkm_falcon *falcon, 1639 struct nvkm_blob *code, struct nvkm_blob *data) 1640 { 1641 nvkm_falcon_load_dmem(falcon, data->data, 0x0, data->size, 0); 1642 nvkm_falcon_load_imem(falcon, code->data, 0x0, code->size, 0, 0, false); 1643 } 1644 1645 static void 1646 gf100_gr_init_csdata(struct gf100_gr *gr, 1647 const struct gf100_gr_pack *pack, 1648 u32 falcon, u32 starstar, u32 base) 1649 { 1650 struct nvkm_device *device = gr->base.engine.subdev.device; 1651 const struct gf100_gr_pack *iter; 1652 const struct gf100_gr_init *init; 1653 u32 addr = ~0, prev = ~0, xfer = 0; 1654 u32 star, temp; 1655 1656 nvkm_wr32(device, falcon + 0x01c0, 0x02000000 + starstar); 1657 star = nvkm_rd32(device, falcon + 0x01c4); 1658 temp = nvkm_rd32(device, falcon + 0x01c4); 1659 if (temp > star) 1660 star = temp; 1661 nvkm_wr32(device, falcon + 0x01c0, 0x01000000 + star); 1662 1663 pack_for_each_init(init, iter, pack) { 1664 u32 head = init->addr - base; 1665 u32 tail = head + init->count * init->pitch; 1666 while (head < tail) { 1667 if (head != prev + 4 || xfer >= 32) { 1668 if (xfer) { 1669 u32 data = ((--xfer << 26) | addr); 1670 nvkm_wr32(device, falcon + 0x01c4, data); 1671 star += 4; 1672 } 1673 addr = head; 1674 xfer = 0; 1675 } 1676 prev = head; 1677 xfer = xfer + 1; 1678 head = head + init->pitch; 1679 } 1680 } 1681 1682 nvkm_wr32(device, falcon + 0x01c4, (--xfer << 26) | addr); 1683 nvkm_wr32(device, falcon + 0x01c0, 0x01000004 + starstar); 1684 nvkm_wr32(device, falcon + 0x01c4, star + 4); 1685 } 1686 1687 /* Initialize context from an external (secure or not) firmware */ 1688 static int 1689 gf100_gr_init_ctxctl_ext(struct gf100_gr *gr) 1690 { 1691 struct nvkm_subdev *subdev = &gr->base.engine.subdev; 1692 struct nvkm_device *device = subdev->device; 1693 u32 lsf_mask = 0; 1694 int ret; 1695 1696 /* load fuc microcode */ 1697 nvkm_mc_unk260(device, 0); 1698 1699 /* securely-managed falcons must be reset using secure boot */ 1700 1701 if (!nvkm_acr_managed_falcon(device, NVKM_ACR_LSF_FECS)) { 1702 gf100_gr_init_fw(&gr->fecs.falcon, &gr->fecs.inst, 1703 &gr->fecs.data); 1704 } else { 1705 lsf_mask |= BIT(NVKM_ACR_LSF_FECS); 1706 } 1707 1708 if (!nvkm_acr_managed_falcon(device, NVKM_ACR_LSF_GPCCS)) { 1709 gf100_gr_init_fw(&gr->gpccs.falcon, &gr->gpccs.inst, 1710 &gr->gpccs.data); 1711 } else { 1712 lsf_mask |= BIT(NVKM_ACR_LSF_GPCCS); 1713 } 1714 1715 if (lsf_mask) { 1716 ret = nvkm_acr_bootstrap_falcons(device, lsf_mask); 1717 if (ret) 1718 return ret; 1719 } 1720 1721 nvkm_mc_unk260(device, 1); 1722 1723 /* start both of them running */ 1724 nvkm_wr32(device, 0x409840, 0xffffffff); 1725 nvkm_wr32(device, 0x41a10c, 0x00000000); 1726 nvkm_wr32(device, 0x40910c, 0x00000000); 1727 1728 nvkm_falcon_start(&gr->gpccs.falcon); 1729 nvkm_falcon_start(&gr->fecs.falcon); 1730 1731 if (nvkm_msec(device, 2000, 1732 if (nvkm_rd32(device, 0x409800) & 0x00000001) 1733 break; 1734 ) < 0) 1735 return -EBUSY; 1736 1737 gf100_gr_fecs_set_watchdog_timeout(gr, 0x7fffffff); 1738 1739 /* Determine how much memory is required to store main context image. */ 1740 ret = gf100_gr_fecs_discover_image_size(gr, &gr->size); 1741 if (ret) 1742 return ret; 1743 1744 /* Determine how much memory is required to store ZCULL image. */ 1745 ret = gf100_gr_fecs_discover_zcull_image_size(gr, &gr->size_zcull); 1746 if (ret) 1747 return ret; 1748 1749 /* Determine how much memory is required to store PerfMon image. */ 1750 ret = gf100_gr_fecs_discover_pm_image_size(gr, &gr->size_pm); 1751 if (ret) 1752 return ret; 1753 1754 /*XXX: We (likely) require PMU support to even bother with this. 1755 * 1756 * Also, it seems like not all GPUs support ELPG. Traces I 1757 * have here show RM enabling it on Kepler/Turing, but none 1758 * of the GPUs between those. NVGPU decides this by PCIID. 1759 */ 1760 if (0) { 1761 ret = gf100_gr_fecs_elpg_bind(gr); 1762 if (ret) 1763 return ret; 1764 } 1765 1766 /* Generate golden context image. */ 1767 if (gr->data == NULL) { 1768 int ret = gf100_grctx_generate(gr); 1769 if (ret) { 1770 nvkm_error(subdev, "failed to construct context\n"); 1771 return ret; 1772 } 1773 } 1774 1775 return 0; 1776 } 1777 1778 static int 1779 gf100_gr_init_ctxctl_int(struct gf100_gr *gr) 1780 { 1781 const struct gf100_grctx_func *grctx = gr->func->grctx; 1782 struct nvkm_subdev *subdev = &gr->base.engine.subdev; 1783 struct nvkm_device *device = subdev->device; 1784 1785 if (!gr->func->fecs.ucode) { 1786 return -ENOSYS; 1787 } 1788 1789 /* load HUB microcode */ 1790 nvkm_mc_unk260(device, 0); 1791 nvkm_falcon_load_dmem(&gr->fecs.falcon, 1792 gr->func->fecs.ucode->data.data, 0x0, 1793 gr->func->fecs.ucode->data.size, 0); 1794 nvkm_falcon_load_imem(&gr->fecs.falcon, 1795 gr->func->fecs.ucode->code.data, 0x0, 1796 gr->func->fecs.ucode->code.size, 0, 0, false); 1797 1798 /* load GPC microcode */ 1799 nvkm_falcon_load_dmem(&gr->gpccs.falcon, 1800 gr->func->gpccs.ucode->data.data, 0x0, 1801 gr->func->gpccs.ucode->data.size, 0); 1802 nvkm_falcon_load_imem(&gr->gpccs.falcon, 1803 gr->func->gpccs.ucode->code.data, 0x0, 1804 gr->func->gpccs.ucode->code.size, 0, 0, false); 1805 nvkm_mc_unk260(device, 1); 1806 1807 /* load register lists */ 1808 gf100_gr_init_csdata(gr, grctx->hub, 0x409000, 0x000, 0x000000); 1809 gf100_gr_init_csdata(gr, grctx->gpc_0, 0x41a000, 0x000, 0x418000); 1810 gf100_gr_init_csdata(gr, grctx->gpc_1, 0x41a000, 0x000, 0x418000); 1811 gf100_gr_init_csdata(gr, grctx->tpc, 0x41a000, 0x004, 0x419800); 1812 gf100_gr_init_csdata(gr, grctx->ppc, 0x41a000, 0x008, 0x41be00); 1813 1814 /* start HUB ucode running, it'll init the GPCs */ 1815 nvkm_wr32(device, 0x40910c, 0x00000000); 1816 nvkm_wr32(device, 0x409100, 0x00000002); 1817 if (nvkm_msec(device, 2000, 1818 if (nvkm_rd32(device, 0x409800) & 0x80000000) 1819 break; 1820 ) < 0) { 1821 gf100_gr_ctxctl_debug(gr); 1822 return -EBUSY; 1823 } 1824 1825 gr->size = nvkm_rd32(device, 0x409804); 1826 if (gr->data == NULL) { 1827 int ret = gf100_grctx_generate(gr); 1828 if (ret) { 1829 nvkm_error(subdev, "failed to construct context\n"); 1830 return ret; 1831 } 1832 } 1833 1834 return 0; 1835 } 1836 1837 int 1838 gf100_gr_init_ctxctl(struct gf100_gr *gr) 1839 { 1840 int ret; 1841 1842 if (gr->firmware) 1843 ret = gf100_gr_init_ctxctl_ext(gr); 1844 else 1845 ret = gf100_gr_init_ctxctl_int(gr); 1846 1847 return ret; 1848 } 1849 1850 void 1851 gf100_gr_oneinit_sm_id(struct gf100_gr *gr) 1852 { 1853 int tpc, gpc; 1854 for (tpc = 0; tpc < gr->tpc_max; tpc++) { 1855 for (gpc = 0; gpc < gr->gpc_nr; gpc++) { 1856 if (tpc < gr->tpc_nr[gpc]) { 1857 gr->sm[gr->sm_nr].gpc = gpc; 1858 gr->sm[gr->sm_nr].tpc = tpc; 1859 gr->sm_nr++; 1860 } 1861 } 1862 } 1863 } 1864 1865 void 1866 gf100_gr_oneinit_tiles(struct gf100_gr *gr) 1867 { 1868 static const u8 primes[] = { 1869 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47, 53, 59, 61 1870 }; 1871 int init_frac[GPC_MAX], init_err[GPC_MAX], run_err[GPC_MAX], i, j; 1872 u32 mul_factor, comm_denom; 1873 u8 gpc_map[GPC_MAX]; 1874 bool sorted; 1875 1876 switch (gr->tpc_total) { 1877 case 15: gr->screen_tile_row_offset = 0x06; break; 1878 case 14: gr->screen_tile_row_offset = 0x05; break; 1879 case 13: gr->screen_tile_row_offset = 0x02; break; 1880 case 11: gr->screen_tile_row_offset = 0x07; break; 1881 case 10: gr->screen_tile_row_offset = 0x06; break; 1882 case 7: 1883 case 5: gr->screen_tile_row_offset = 0x01; break; 1884 case 3: gr->screen_tile_row_offset = 0x02; break; 1885 case 2: 1886 case 1: gr->screen_tile_row_offset = 0x01; break; 1887 default: gr->screen_tile_row_offset = 0x03; 1888 for (i = 0; i < ARRAY_SIZE(primes); i++) { 1889 if (gr->tpc_total % primes[i]) { 1890 gr->screen_tile_row_offset = primes[i]; 1891 break; 1892 } 1893 } 1894 break; 1895 } 1896 1897 /* Sort GPCs by TPC count, highest-to-lowest. */ 1898 for (i = 0; i < gr->gpc_nr; i++) 1899 gpc_map[i] = i; 1900 sorted = false; 1901 1902 while (!sorted) { 1903 for (sorted = true, i = 0; i < gr->gpc_nr - 1; i++) { 1904 if (gr->tpc_nr[gpc_map[i + 1]] > 1905 gr->tpc_nr[gpc_map[i + 0]]) { 1906 u8 swap = gpc_map[i]; 1907 gpc_map[i + 0] = gpc_map[i + 1]; 1908 gpc_map[i + 1] = swap; 1909 sorted = false; 1910 } 1911 } 1912 } 1913 1914 /* Determine tile->GPC mapping */ 1915 mul_factor = gr->gpc_nr * gr->tpc_max; 1916 if (mul_factor & 1) 1917 mul_factor = 2; 1918 else 1919 mul_factor = 1; 1920 1921 comm_denom = gr->gpc_nr * gr->tpc_max * mul_factor; 1922 1923 for (i = 0; i < gr->gpc_nr; i++) { 1924 init_frac[i] = gr->tpc_nr[gpc_map[i]] * gr->gpc_nr * mul_factor; 1925 init_err[i] = i * gr->tpc_max * mul_factor - comm_denom/2; 1926 run_err[i] = init_frac[i] + init_err[i]; 1927 } 1928 1929 for (i = 0; i < gr->tpc_total;) { 1930 for (j = 0; j < gr->gpc_nr; j++) { 1931 if ((run_err[j] * 2) >= comm_denom) { 1932 gr->tile[i++] = gpc_map[j]; 1933 run_err[j] += init_frac[j] - comm_denom; 1934 } else { 1935 run_err[j] += init_frac[j]; 1936 } 1937 } 1938 } 1939 } 1940 1941 static int 1942 gf100_gr_oneinit(struct nvkm_gr *base) 1943 { 1944 struct gf100_gr *gr = gf100_gr(base); 1945 struct nvkm_subdev *subdev = &gr->base.engine.subdev; 1946 struct nvkm_device *device = subdev->device; 1947 int i, j; 1948 1949 nvkm_pmu_pgob(device->pmu, false); 1950 1951 gr->rop_nr = gr->func->rops(gr); 1952 gr->gpc_nr = nvkm_rd32(device, 0x409604) & 0x0000001f; 1953 for (i = 0; i < gr->gpc_nr; i++) { 1954 gr->tpc_nr[i] = nvkm_rd32(device, GPC_UNIT(i, 0x2608)); 1955 gr->tpc_max = max(gr->tpc_max, gr->tpc_nr[i]); 1956 gr->tpc_total += gr->tpc_nr[i]; 1957 gr->ppc_nr[i] = gr->func->ppc_nr; 1958 for (j = 0; j < gr->ppc_nr[i]; j++) { 1959 gr->ppc_tpc_mask[i][j] = 1960 nvkm_rd32(device, GPC_UNIT(i, 0x0c30 + (j * 4))); 1961 if (gr->ppc_tpc_mask[i][j] == 0) 1962 continue; 1963 gr->ppc_mask[i] |= (1 << j); 1964 gr->ppc_tpc_nr[i][j] = hweight8(gr->ppc_tpc_mask[i][j]); 1965 if (gr->ppc_tpc_min == 0 || 1966 gr->ppc_tpc_min > gr->ppc_tpc_nr[i][j]) 1967 gr->ppc_tpc_min = gr->ppc_tpc_nr[i][j]; 1968 if (gr->ppc_tpc_max < gr->ppc_tpc_nr[i][j]) 1969 gr->ppc_tpc_max = gr->ppc_tpc_nr[i][j]; 1970 } 1971 } 1972 1973 memset(gr->tile, 0xff, sizeof(gr->tile)); 1974 gr->func->oneinit_tiles(gr); 1975 gr->func->oneinit_sm_id(gr); 1976 return 0; 1977 } 1978 1979 static int 1980 gf100_gr_init_(struct nvkm_gr *base) 1981 { 1982 struct gf100_gr *gr = gf100_gr(base); 1983 struct nvkm_subdev *subdev = &base->engine.subdev; 1984 struct nvkm_device *device = subdev->device; 1985 bool reset = device->chipset == 0x137 || device->chipset == 0x138; 1986 u32 ret; 1987 1988 /* On certain GP107/GP108 boards, we trigger a weird issue where 1989 * GR will stop responding to PRI accesses after we've asked the 1990 * SEC2 RTOS to boot the GR falcons. This happens with far more 1991 * frequency when cold-booting a board (ie. returning from D3). 1992 * 1993 * The root cause for this is not known and has proven difficult 1994 * to isolate, with many avenues being dead-ends. 1995 * 1996 * A workaround was discovered by Karol, whereby putting GR into 1997 * reset for an extended period right before initialisation 1998 * prevents the problem from occuring. 1999 * 2000 * XXX: As RM does not require any such workaround, this is more 2001 * of a hack than a true fix. 2002 */ 2003 reset = nvkm_boolopt(device->cfgopt, "NvGrResetWar", reset); 2004 if (reset) { 2005 nvkm_mask(device, 0x000200, 0x00001000, 0x00000000); 2006 nvkm_rd32(device, 0x000200); 2007 msleep(50); 2008 nvkm_mask(device, 0x000200, 0x00001000, 0x00001000); 2009 nvkm_rd32(device, 0x000200); 2010 } 2011 2012 nvkm_pmu_pgob(gr->base.engine.subdev.device->pmu, false); 2013 2014 ret = nvkm_falcon_get(&gr->fecs.falcon, subdev); 2015 if (ret) 2016 return ret; 2017 2018 ret = nvkm_falcon_get(&gr->gpccs.falcon, subdev); 2019 if (ret) 2020 return ret; 2021 2022 return gr->func->init(gr); 2023 } 2024 2025 static int 2026 gf100_gr_fini(struct nvkm_gr *base, bool suspend) 2027 { 2028 struct gf100_gr *gr = gf100_gr(base); 2029 struct nvkm_subdev *subdev = &gr->base.engine.subdev; 2030 nvkm_falcon_put(&gr->gpccs.falcon, subdev); 2031 nvkm_falcon_put(&gr->fecs.falcon, subdev); 2032 return 0; 2033 } 2034 2035 void * 2036 gf100_gr_dtor(struct nvkm_gr *base) 2037 { 2038 struct gf100_gr *gr = gf100_gr(base); 2039 2040 kfree(gr->data); 2041 2042 nvkm_falcon_dtor(&gr->gpccs.falcon); 2043 nvkm_falcon_dtor(&gr->fecs.falcon); 2044 2045 nvkm_blob_dtor(&gr->fecs.inst); 2046 nvkm_blob_dtor(&gr->fecs.data); 2047 nvkm_blob_dtor(&gr->gpccs.inst); 2048 nvkm_blob_dtor(&gr->gpccs.data); 2049 2050 vfree(gr->bundle); 2051 vfree(gr->method); 2052 vfree(gr->sw_ctx); 2053 vfree(gr->sw_nonctx); 2054 2055 return gr; 2056 } 2057 2058 static const struct nvkm_gr_func 2059 gf100_gr_ = { 2060 .dtor = gf100_gr_dtor, 2061 .oneinit = gf100_gr_oneinit, 2062 .init = gf100_gr_init_, 2063 .fini = gf100_gr_fini, 2064 .intr = gf100_gr_intr, 2065 .units = gf100_gr_units, 2066 .chan_new = gf100_gr_chan_new, 2067 .object_get = gf100_gr_object_get, 2068 .chsw_load = gf100_gr_chsw_load, 2069 .ctxsw.pause = gf100_gr_fecs_stop_ctxsw, 2070 .ctxsw.resume = gf100_gr_fecs_start_ctxsw, 2071 .ctxsw.inst = gf100_gr_ctxsw_inst, 2072 }; 2073 2074 static const struct nvkm_falcon_func 2075 gf100_gr_flcn = { 2076 .fbif = 0x600, 2077 .load_imem = nvkm_falcon_v1_load_imem, 2078 .load_dmem = nvkm_falcon_v1_load_dmem, 2079 .read_dmem = nvkm_falcon_v1_read_dmem, 2080 .bind_context = nvkm_falcon_v1_bind_context, 2081 .wait_for_halt = nvkm_falcon_v1_wait_for_halt, 2082 .clear_interrupt = nvkm_falcon_v1_clear_interrupt, 2083 .set_start_addr = nvkm_falcon_v1_set_start_addr, 2084 .start = nvkm_falcon_v1_start, 2085 .enable = nvkm_falcon_v1_enable, 2086 .disable = nvkm_falcon_v1_disable, 2087 }; 2088 2089 int 2090 gf100_gr_new_(const struct gf100_gr_fwif *fwif, 2091 struct nvkm_device *device, int index, struct nvkm_gr **pgr) 2092 { 2093 struct gf100_gr *gr; 2094 int ret; 2095 2096 if (!(gr = kzalloc(sizeof(*gr), GFP_KERNEL))) 2097 return -ENOMEM; 2098 *pgr = &gr->base; 2099 2100 ret = nvkm_gr_ctor(&gf100_gr_, device, index, true, &gr->base); 2101 if (ret) 2102 return ret; 2103 2104 fwif = nvkm_firmware_load(&gr->base.engine.subdev, fwif, "Gr", gr); 2105 if (IS_ERR(fwif)) 2106 return -ENODEV; 2107 2108 gr->func = fwif->func; 2109 2110 ret = nvkm_falcon_ctor(&gf100_gr_flcn, &gr->base.engine.subdev, 2111 "fecs", 0x409000, &gr->fecs.falcon); 2112 if (ret) 2113 return ret; 2114 2115 mutex_init(&gr->fecs.mutex); 2116 2117 ret = nvkm_falcon_ctor(&gf100_gr_flcn, &gr->base.engine.subdev, 2118 "gpccs", 0x41a000, &gr->gpccs.falcon); 2119 if (ret) 2120 return ret; 2121 2122 return 0; 2123 } 2124 2125 void 2126 gf100_gr_init_num_tpc_per_gpc(struct gf100_gr *gr, bool pd, bool ds) 2127 { 2128 struct nvkm_device *device = gr->base.engine.subdev.device; 2129 int gpc, i, j; 2130 u32 data; 2131 2132 for (gpc = 0, i = 0; i < 4; i++) { 2133 for (data = 0, j = 0; j < 8 && gpc < gr->gpc_nr; j++, gpc++) 2134 data |= gr->tpc_nr[gpc] << (j * 4); 2135 if (pd) 2136 nvkm_wr32(device, 0x406028 + (i * 4), data); 2137 if (ds) 2138 nvkm_wr32(device, 0x405870 + (i * 4), data); 2139 } 2140 } 2141 2142 void 2143 gf100_gr_init_400054(struct gf100_gr *gr) 2144 { 2145 nvkm_wr32(gr->base.engine.subdev.device, 0x400054, 0x34ce3464); 2146 } 2147 2148 void 2149 gf100_gr_init_shader_exceptions(struct gf100_gr *gr, int gpc, int tpc) 2150 { 2151 struct nvkm_device *device = gr->base.engine.subdev.device; 2152 nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x644), 0x001ffffe); 2153 nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x64c), 0x0000000f); 2154 } 2155 2156 void 2157 gf100_gr_init_tex_hww_esr(struct gf100_gr *gr, int gpc, int tpc) 2158 { 2159 struct nvkm_device *device = gr->base.engine.subdev.device; 2160 nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x224), 0xc0000000); 2161 } 2162 2163 void 2164 gf100_gr_init_419eb4(struct gf100_gr *gr) 2165 { 2166 struct nvkm_device *device = gr->base.engine.subdev.device; 2167 nvkm_mask(device, 0x419eb4, 0x00001000, 0x00001000); 2168 } 2169 2170 void 2171 gf100_gr_init_419cc0(struct gf100_gr *gr) 2172 { 2173 struct nvkm_device *device = gr->base.engine.subdev.device; 2174 int gpc, tpc; 2175 2176 nvkm_mask(device, 0x419cc0, 0x00000008, 0x00000008); 2177 2178 for (gpc = 0; gpc < gr->gpc_nr; gpc++) { 2179 for (tpc = 0; tpc < gr->tpc_nr[gpc]; tpc++) 2180 nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x48c), 0xc0000000); 2181 } 2182 } 2183 2184 void 2185 gf100_gr_init_40601c(struct gf100_gr *gr) 2186 { 2187 nvkm_wr32(gr->base.engine.subdev.device, 0x40601c, 0xc0000000); 2188 } 2189 2190 void 2191 gf100_gr_init_fecs_exceptions(struct gf100_gr *gr) 2192 { 2193 const u32 data = gr->firmware ? 0x000e0000 : 0x000e0001; 2194 nvkm_wr32(gr->base.engine.subdev.device, 0x409c24, data); 2195 } 2196 2197 void 2198 gf100_gr_init_gpc_mmu(struct gf100_gr *gr) 2199 { 2200 struct nvkm_device *device = gr->base.engine.subdev.device; 2201 struct nvkm_fb *fb = device->fb; 2202 2203 nvkm_wr32(device, 0x418880, nvkm_rd32(device, 0x100c80) & 0x00000001); 2204 nvkm_wr32(device, 0x4188a4, 0x03000000); 2205 nvkm_wr32(device, 0x418888, 0x00000000); 2206 nvkm_wr32(device, 0x41888c, 0x00000000); 2207 nvkm_wr32(device, 0x418890, 0x00000000); 2208 nvkm_wr32(device, 0x418894, 0x00000000); 2209 nvkm_wr32(device, 0x4188b4, nvkm_memory_addr(fb->mmu_wr) >> 8); 2210 nvkm_wr32(device, 0x4188b8, nvkm_memory_addr(fb->mmu_rd) >> 8); 2211 } 2212 2213 void 2214 gf100_gr_init_num_active_ltcs(struct gf100_gr *gr) 2215 { 2216 struct nvkm_device *device = gr->base.engine.subdev.device; 2217 nvkm_wr32(device, GPC_BCAST(0x08ac), nvkm_rd32(device, 0x100800)); 2218 } 2219 2220 void 2221 gf100_gr_init_zcull(struct gf100_gr *gr) 2222 { 2223 struct nvkm_device *device = gr->base.engine.subdev.device; 2224 const u32 magicgpc918 = DIV_ROUND_UP(0x00800000, gr->tpc_total); 2225 const u8 tile_nr = ALIGN(gr->tpc_total, 32); 2226 u8 bank[GPC_MAX] = {}, gpc, i, j; 2227 u32 data; 2228 2229 for (i = 0; i < tile_nr; i += 8) { 2230 for (data = 0, j = 0; j < 8 && i + j < gr->tpc_total; j++) { 2231 data |= bank[gr->tile[i + j]] << (j * 4); 2232 bank[gr->tile[i + j]]++; 2233 } 2234 nvkm_wr32(device, GPC_BCAST(0x0980 + ((i / 8) * 4)), data); 2235 } 2236 2237 for (gpc = 0; gpc < gr->gpc_nr; gpc++) { 2238 nvkm_wr32(device, GPC_UNIT(gpc, 0x0914), 2239 gr->screen_tile_row_offset << 8 | gr->tpc_nr[gpc]); 2240 nvkm_wr32(device, GPC_UNIT(gpc, 0x0910), 0x00040000 | 2241 gr->tpc_total); 2242 nvkm_wr32(device, GPC_UNIT(gpc, 0x0918), magicgpc918); 2243 } 2244 2245 nvkm_wr32(device, GPC_BCAST(0x1bd4), magicgpc918); 2246 } 2247 2248 void 2249 gf100_gr_init_vsc_stream_master(struct gf100_gr *gr) 2250 { 2251 struct nvkm_device *device = gr->base.engine.subdev.device; 2252 nvkm_mask(device, TPC_UNIT(0, 0, 0x05c), 0x00000001, 0x00000001); 2253 } 2254 2255 int 2256 gf100_gr_init(struct gf100_gr *gr) 2257 { 2258 struct nvkm_device *device = gr->base.engine.subdev.device; 2259 int gpc, tpc, rop; 2260 2261 if (gr->func->init_419bd8) 2262 gr->func->init_419bd8(gr); 2263 2264 gr->func->init_gpc_mmu(gr); 2265 2266 if (gr->sw_nonctx) 2267 gf100_gr_mmio(gr, gr->sw_nonctx); 2268 else 2269 gf100_gr_mmio(gr, gr->func->mmio); 2270 2271 gf100_gr_wait_idle(gr); 2272 2273 if (gr->func->init_r405a14) 2274 gr->func->init_r405a14(gr); 2275 2276 if (gr->func->clkgate_pack) 2277 nvkm_therm_clkgate_init(device->therm, gr->func->clkgate_pack); 2278 2279 if (gr->func->init_bios) 2280 gr->func->init_bios(gr); 2281 2282 gr->func->init_vsc_stream_master(gr); 2283 gr->func->init_zcull(gr); 2284 gr->func->init_num_active_ltcs(gr); 2285 if (gr->func->init_rop_active_fbps) 2286 gr->func->init_rop_active_fbps(gr); 2287 if (gr->func->init_bios_2) 2288 gr->func->init_bios_2(gr); 2289 if (gr->func->init_swdx_pes_mask) 2290 gr->func->init_swdx_pes_mask(gr); 2291 if (gr->func->init_fs) 2292 gr->func->init_fs(gr); 2293 2294 nvkm_wr32(device, 0x400500, 0x00010001); 2295 2296 nvkm_wr32(device, 0x400100, 0xffffffff); 2297 nvkm_wr32(device, 0x40013c, 0xffffffff); 2298 nvkm_wr32(device, 0x400124, 0x00000002); 2299 2300 gr->func->init_fecs_exceptions(gr); 2301 if (gr->func->init_ds_hww_esr_2) 2302 gr->func->init_ds_hww_esr_2(gr); 2303 2304 nvkm_wr32(device, 0x404000, 0xc0000000); 2305 nvkm_wr32(device, 0x404600, 0xc0000000); 2306 nvkm_wr32(device, 0x408030, 0xc0000000); 2307 2308 if (gr->func->init_40601c) 2309 gr->func->init_40601c(gr); 2310 2311 nvkm_wr32(device, 0x406018, 0xc0000000); 2312 nvkm_wr32(device, 0x404490, 0xc0000000); 2313 2314 if (gr->func->init_sked_hww_esr) 2315 gr->func->init_sked_hww_esr(gr); 2316 2317 nvkm_wr32(device, 0x405840, 0xc0000000); 2318 nvkm_wr32(device, 0x405844, 0x00ffffff); 2319 2320 if (gr->func->init_419cc0) 2321 gr->func->init_419cc0(gr); 2322 if (gr->func->init_419eb4) 2323 gr->func->init_419eb4(gr); 2324 if (gr->func->init_419c9c) 2325 gr->func->init_419c9c(gr); 2326 2327 if (gr->func->init_ppc_exceptions) 2328 gr->func->init_ppc_exceptions(gr); 2329 2330 for (gpc = 0; gpc < gr->gpc_nr; gpc++) { 2331 nvkm_wr32(device, GPC_UNIT(gpc, 0x0420), 0xc0000000); 2332 nvkm_wr32(device, GPC_UNIT(gpc, 0x0900), 0xc0000000); 2333 nvkm_wr32(device, GPC_UNIT(gpc, 0x1028), 0xc0000000); 2334 nvkm_wr32(device, GPC_UNIT(gpc, 0x0824), 0xc0000000); 2335 for (tpc = 0; tpc < gr->tpc_nr[gpc]; tpc++) { 2336 nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x508), 0xffffffff); 2337 nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x50c), 0xffffffff); 2338 if (gr->func->init_tex_hww_esr) 2339 gr->func->init_tex_hww_esr(gr, gpc, tpc); 2340 nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x084), 0xc0000000); 2341 if (gr->func->init_504430) 2342 gr->func->init_504430(gr, gpc, tpc); 2343 gr->func->init_shader_exceptions(gr, gpc, tpc); 2344 } 2345 nvkm_wr32(device, GPC_UNIT(gpc, 0x2c90), 0xffffffff); 2346 nvkm_wr32(device, GPC_UNIT(gpc, 0x2c94), 0xffffffff); 2347 } 2348 2349 for (rop = 0; rop < gr->rop_nr; rop++) { 2350 nvkm_wr32(device, ROP_UNIT(rop, 0x144), 0x40000000); 2351 nvkm_wr32(device, ROP_UNIT(rop, 0x070), 0x40000000); 2352 nvkm_wr32(device, ROP_UNIT(rop, 0x204), 0xffffffff); 2353 nvkm_wr32(device, ROP_UNIT(rop, 0x208), 0xffffffff); 2354 } 2355 2356 nvkm_wr32(device, 0x400108, 0xffffffff); 2357 nvkm_wr32(device, 0x400138, 0xffffffff); 2358 nvkm_wr32(device, 0x400118, 0xffffffff); 2359 nvkm_wr32(device, 0x400130, 0xffffffff); 2360 nvkm_wr32(device, 0x40011c, 0xffffffff); 2361 nvkm_wr32(device, 0x400134, 0xffffffff); 2362 2363 if (gr->func->init_400054) 2364 gr->func->init_400054(gr); 2365 2366 gf100_gr_zbc_init(gr); 2367 2368 if (gr->func->init_4188a4) 2369 gr->func->init_4188a4(gr); 2370 2371 return gf100_gr_init_ctxctl(gr); 2372 } 2373 2374 #include "fuc/hubgf100.fuc3.h" 2375 2376 struct gf100_gr_ucode 2377 gf100_gr_fecs_ucode = { 2378 .code.data = gf100_grhub_code, 2379 .code.size = sizeof(gf100_grhub_code), 2380 .data.data = gf100_grhub_data, 2381 .data.size = sizeof(gf100_grhub_data), 2382 }; 2383 2384 #include "fuc/gpcgf100.fuc3.h" 2385 2386 struct gf100_gr_ucode 2387 gf100_gr_gpccs_ucode = { 2388 .code.data = gf100_grgpc_code, 2389 .code.size = sizeof(gf100_grgpc_code), 2390 .data.data = gf100_grgpc_data, 2391 .data.size = sizeof(gf100_grgpc_data), 2392 }; 2393 2394 static const struct gf100_gr_func 2395 gf100_gr = { 2396 .oneinit_tiles = gf100_gr_oneinit_tiles, 2397 .oneinit_sm_id = gf100_gr_oneinit_sm_id, 2398 .init = gf100_gr_init, 2399 .init_gpc_mmu = gf100_gr_init_gpc_mmu, 2400 .init_vsc_stream_master = gf100_gr_init_vsc_stream_master, 2401 .init_zcull = gf100_gr_init_zcull, 2402 .init_num_active_ltcs = gf100_gr_init_num_active_ltcs, 2403 .init_fecs_exceptions = gf100_gr_init_fecs_exceptions, 2404 .init_40601c = gf100_gr_init_40601c, 2405 .init_419cc0 = gf100_gr_init_419cc0, 2406 .init_419eb4 = gf100_gr_init_419eb4, 2407 .init_tex_hww_esr = gf100_gr_init_tex_hww_esr, 2408 .init_shader_exceptions = gf100_gr_init_shader_exceptions, 2409 .init_400054 = gf100_gr_init_400054, 2410 .trap_mp = gf100_gr_trap_mp, 2411 .mmio = gf100_gr_pack_mmio, 2412 .fecs.ucode = &gf100_gr_fecs_ucode, 2413 .gpccs.ucode = &gf100_gr_gpccs_ucode, 2414 .rops = gf100_gr_rops, 2415 .grctx = &gf100_grctx, 2416 .zbc = &gf100_gr_zbc, 2417 .sclass = { 2418 { -1, -1, FERMI_TWOD_A }, 2419 { -1, -1, FERMI_MEMORY_TO_MEMORY_FORMAT_A }, 2420 { -1, -1, FERMI_A, &gf100_fermi }, 2421 { -1, -1, FERMI_COMPUTE_A }, 2422 {} 2423 } 2424 }; 2425 2426 int 2427 gf100_gr_nofw(struct gf100_gr *gr, int ver, const struct gf100_gr_fwif *fwif) 2428 { 2429 gr->firmware = false; 2430 return 0; 2431 } 2432 2433 static int 2434 gf100_gr_load_fw(struct gf100_gr *gr, const char *name, 2435 struct nvkm_blob *blob) 2436 { 2437 struct nvkm_subdev *subdev = &gr->base.engine.subdev; 2438 struct nvkm_device *device = subdev->device; 2439 const struct firmware *fw; 2440 char f[32]; 2441 int ret; 2442 2443 snprintf(f, sizeof(f), "nouveau/nv%02x_%s", device->chipset, name); 2444 ret = request_firmware(&fw, f, device->dev); 2445 if (ret) { 2446 snprintf(f, sizeof(f), "nouveau/%s", name); 2447 ret = request_firmware(&fw, f, device->dev); 2448 if (ret) { 2449 nvkm_error(subdev, "failed to load %s\n", name); 2450 return ret; 2451 } 2452 } 2453 2454 blob->size = fw->size; 2455 blob->data = kmemdup(fw->data, blob->size, GFP_KERNEL); 2456 release_firmware(fw); 2457 return (blob->data != NULL) ? 0 : -ENOMEM; 2458 } 2459 2460 int 2461 gf100_gr_load(struct gf100_gr *gr, int ver, const struct gf100_gr_fwif *fwif) 2462 { 2463 struct nvkm_device *device = gr->base.engine.subdev.device; 2464 2465 if (!nvkm_boolopt(device->cfgopt, "NvGrUseFW", false)) 2466 return -EINVAL; 2467 2468 if (gf100_gr_load_fw(gr, "fuc409c", &gr->fecs.inst) || 2469 gf100_gr_load_fw(gr, "fuc409d", &gr->fecs.data) || 2470 gf100_gr_load_fw(gr, "fuc41ac", &gr->gpccs.inst) || 2471 gf100_gr_load_fw(gr, "fuc41ad", &gr->gpccs.data)) 2472 return -ENOENT; 2473 2474 gr->firmware = true; 2475 return 0; 2476 } 2477 2478 static const struct gf100_gr_fwif 2479 gf100_gr_fwif[] = { 2480 { -1, gf100_gr_load, &gf100_gr }, 2481 { -1, gf100_gr_nofw, &gf100_gr }, 2482 {} 2483 }; 2484 2485 int 2486 gf100_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) 2487 { 2488 return gf100_gr_new_(gf100_gr_fwif, device, index, pgr); 2489 } 2490