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