2 * Copyright 2012 Red Hat Inc.
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:
11 * The above copyright notice and this permission notice shall be included in
12 * all copies or substantial portions of the Software.
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.
28 #include <core/client.h>
29 #include <core/handle.h>
30 #include <core/option.h>
31 #include <engine/fifo.h>
32 #include <subdev/fb.h>
33 #include <subdev/mc.h>
34 #include <subdev/timer.h>
36 #include <nvif/class.h>
37 #include <nvif/unpack.h>
39 /*******************************************************************************
40 * Zero Bandwidth Clear
41 ******************************************************************************/
44 gf100_gr_zbc_clear_color(struct gf100_gr
*gr
, int zbc
)
46 struct nvkm_device
*device
= gr
->base
.engine
.subdev
.device
;
47 if (gr
->zbc_color
[zbc
].format
) {
48 nvkm_wr32(device
, 0x405804, gr
->zbc_color
[zbc
].ds
[0]);
49 nvkm_wr32(device
, 0x405808, gr
->zbc_color
[zbc
].ds
[1]);
50 nvkm_wr32(device
, 0x40580c, gr
->zbc_color
[zbc
].ds
[2]);
51 nvkm_wr32(device
, 0x405810, gr
->zbc_color
[zbc
].ds
[3]);
53 nvkm_wr32(device
, 0x405814, gr
->zbc_color
[zbc
].format
);
54 nvkm_wr32(device
, 0x405820, zbc
);
55 nvkm_wr32(device
, 0x405824, 0x00000004); /* TRIGGER | WRITE | COLOR */
59 gf100_gr_zbc_color_get(struct gf100_gr
*gr
, int format
,
60 const u32 ds
[4], const u32 l2
[4])
62 struct nvkm_ltc
*ltc
= nvkm_ltc(gr
);
65 for (i
= ltc
->zbc_min
; i
<= ltc
->zbc_max
; i
++) {
66 if (gr
->zbc_color
[i
].format
) {
67 if (gr
->zbc_color
[i
].format
!= format
)
69 if (memcmp(gr
->zbc_color
[i
].ds
, ds
, sizeof(
70 gr
->zbc_color
[i
].ds
)))
72 if (memcmp(gr
->zbc_color
[i
].l2
, l2
, sizeof(
73 gr
->zbc_color
[i
].l2
))) {
79 zbc
= (zbc
< 0) ? i
: zbc
;
86 memcpy(gr
->zbc_color
[zbc
].ds
, ds
, sizeof(gr
->zbc_color
[zbc
].ds
));
87 memcpy(gr
->zbc_color
[zbc
].l2
, l2
, sizeof(gr
->zbc_color
[zbc
].l2
));
88 gr
->zbc_color
[zbc
].format
= format
;
89 ltc
->zbc_color_get(ltc
, zbc
, l2
);
90 gf100_gr_zbc_clear_color(gr
, zbc
);
95 gf100_gr_zbc_clear_depth(struct gf100_gr
*gr
, int zbc
)
97 struct nvkm_device
*device
= gr
->base
.engine
.subdev
.device
;
98 if (gr
->zbc_depth
[zbc
].format
)
99 nvkm_wr32(device
, 0x405818, gr
->zbc_depth
[zbc
].ds
);
100 nvkm_wr32(device
, 0x40581c, gr
->zbc_depth
[zbc
].format
);
101 nvkm_wr32(device
, 0x405820, zbc
);
102 nvkm_wr32(device
, 0x405824, 0x00000005); /* TRIGGER | WRITE | DEPTH */
106 gf100_gr_zbc_depth_get(struct gf100_gr
*gr
, int format
,
107 const u32 ds
, const u32 l2
)
109 struct nvkm_ltc
*ltc
= nvkm_ltc(gr
);
110 int zbc
= -ENOSPC
, i
;
112 for (i
= ltc
->zbc_min
; i
<= ltc
->zbc_max
; i
++) {
113 if (gr
->zbc_depth
[i
].format
) {
114 if (gr
->zbc_depth
[i
].format
!= format
)
116 if (gr
->zbc_depth
[i
].ds
!= ds
)
118 if (gr
->zbc_depth
[i
].l2
!= l2
) {
124 zbc
= (zbc
< 0) ? i
: zbc
;
131 gr
->zbc_depth
[zbc
].format
= format
;
132 gr
->zbc_depth
[zbc
].ds
= ds
;
133 gr
->zbc_depth
[zbc
].l2
= l2
;
134 ltc
->zbc_depth_get(ltc
, zbc
, l2
);
135 gf100_gr_zbc_clear_depth(gr
, zbc
);
139 /*******************************************************************************
140 * Graphics object classes
141 ******************************************************************************/
144 gf100_fermi_mthd_zbc_color(struct nvkm_object
*object
, void *data
, u32 size
)
146 struct gf100_gr
*gr
= (void *)object
->engine
;
148 struct fermi_a_zbc_color_v0 v0
;
152 if (nvif_unpack(args
->v0
, 0, 0, false)) {
153 switch (args
->v0
.format
) {
154 case FERMI_A_ZBC_COLOR_V0_FMT_ZERO
:
155 case FERMI_A_ZBC_COLOR_V0_FMT_UNORM_ONE
:
156 case FERMI_A_ZBC_COLOR_V0_FMT_RF32_GF32_BF32_AF32
:
157 case FERMI_A_ZBC_COLOR_V0_FMT_R16_G16_B16_A16
:
158 case FERMI_A_ZBC_COLOR_V0_FMT_RN16_GN16_BN16_AN16
:
159 case FERMI_A_ZBC_COLOR_V0_FMT_RS16_GS16_BS16_AS16
:
160 case FERMI_A_ZBC_COLOR_V0_FMT_RU16_GU16_BU16_AU16
:
161 case FERMI_A_ZBC_COLOR_V0_FMT_RF16_GF16_BF16_AF16
:
162 case FERMI_A_ZBC_COLOR_V0_FMT_A8R8G8B8
:
163 case FERMI_A_ZBC_COLOR_V0_FMT_A8RL8GL8BL8
:
164 case FERMI_A_ZBC_COLOR_V0_FMT_A2B10G10R10
:
165 case FERMI_A_ZBC_COLOR_V0_FMT_AU2BU10GU10RU10
:
166 case FERMI_A_ZBC_COLOR_V0_FMT_A8B8G8R8
:
167 case FERMI_A_ZBC_COLOR_V0_FMT_A8BL8GL8RL8
:
168 case FERMI_A_ZBC_COLOR_V0_FMT_AN8BN8GN8RN8
:
169 case FERMI_A_ZBC_COLOR_V0_FMT_AS8BS8GS8RS8
:
170 case FERMI_A_ZBC_COLOR_V0_FMT_AU8BU8GU8RU8
:
171 case FERMI_A_ZBC_COLOR_V0_FMT_A2R10G10B10
:
172 case FERMI_A_ZBC_COLOR_V0_FMT_BF10GF11RF11
:
173 ret
= gf100_gr_zbc_color_get(gr
, args
->v0
.format
,
177 args
->v0
.index
= ret
;
190 gf100_fermi_mthd_zbc_depth(struct nvkm_object
*object
, void *data
, u32 size
)
192 struct gf100_gr
*gr
= (void *)object
->engine
;
194 struct fermi_a_zbc_depth_v0 v0
;
198 if (nvif_unpack(args
->v0
, 0, 0, false)) {
199 switch (args
->v0
.format
) {
200 case FERMI_A_ZBC_DEPTH_V0_FMT_FP32
:
201 ret
= gf100_gr_zbc_depth_get(gr
, args
->v0
.format
,
204 return (ret
>= 0) ? 0 : -ENOSPC
;
214 gf100_fermi_mthd(struct nvkm_object
*object
, u32 mthd
, void *data
, u32 size
)
217 case FERMI_A_ZBC_COLOR
:
218 return gf100_fermi_mthd_zbc_color(object
, data
, size
);
219 case FERMI_A_ZBC_DEPTH
:
220 return gf100_fermi_mthd_zbc_depth(object
, data
, size
);
228 gf100_fermi_ofuncs
= {
229 .ctor
= _nvkm_object_ctor
,
230 .dtor
= nvkm_object_destroy
,
231 .init
= nvkm_object_init
,
232 .fini
= nvkm_object_fini
,
233 .mthd
= gf100_fermi_mthd
,
237 gf100_gr_set_shader_exceptions(struct nvkm_object
*object
, u32 mthd
,
238 void *pdata
, u32 size
)
240 struct gf100_gr
*gr
= (void *)object
->engine
;
241 struct nvkm_device
*device
= gr
->base
.engine
.subdev
.device
;
242 if (size
>= sizeof(u32
)) {
243 u32 data
= *(u32
*)pdata
? 0xffffffff : 0x00000000;
244 nvkm_wr32(device
, 0x419e44, data
);
245 nvkm_wr32(device
, 0x419e4c, data
);
252 gf100_gr_9097_omthds
[] = {
253 { 0x1528, 0x1528, gf100_gr_set_shader_exceptions
},
258 gf100_gr_90c0_omthds
[] = {
259 { 0x1528, 0x1528, gf100_gr_set_shader_exceptions
},
264 gf100_gr_sclass
[] = {
265 { FERMI_TWOD_A
, &nvkm_object_ofuncs
},
266 { FERMI_MEMORY_TO_MEMORY_FORMAT_A
, &nvkm_object_ofuncs
},
267 { FERMI_A
, &gf100_fermi_ofuncs
, gf100_gr_9097_omthds
},
268 { FERMI_COMPUTE_A
, &nvkm_object_ofuncs
, gf100_gr_90c0_omthds
},
272 /*******************************************************************************
274 ******************************************************************************/
277 gf100_gr_context_ctor(struct nvkm_object
*parent
, struct nvkm_object
*engine
,
278 struct nvkm_oclass
*oclass
, void *args
, u32 size
,
279 struct nvkm_object
**pobject
)
281 struct nvkm_vm
*vm
= nvkm_client(parent
)->vm
;
282 struct gf100_gr
*gr
= (void *)engine
;
283 struct gf100_gr_data
*data
= gr
->mmio_data
;
284 struct gf100_gr_mmio
*mmio
= gr
->mmio_list
;
285 struct gf100_gr_chan
*chan
;
288 /* allocate memory for context, and fill with default values */
289 ret
= nvkm_gr_context_create(parent
, engine
, oclass
, NULL
,
291 NVOBJ_FLAG_ZERO_ALLOC
, &chan
);
292 *pobject
= nv_object(chan
);
296 /* allocate memory for a "mmio list" buffer that's used by the HUB
297 * fuc to modify some per-context register settings on first load
300 ret
= nvkm_gpuobj_new(nv_object(chan
), NULL
, 0x1000, 0x100, 0,
305 ret
= nvkm_gpuobj_map_vm(nv_gpuobj(chan
->mmio
), vm
,
306 NV_MEM_ACCESS_RW
| NV_MEM_ACCESS_SYS
,
311 /* allocate buffers referenced by mmio list */
312 for (i
= 0; data
->size
&& i
< ARRAY_SIZE(gr
->mmio_data
); i
++) {
313 ret
= nvkm_gpuobj_new(nv_object(chan
), NULL
, data
->size
,
314 data
->align
, 0, &chan
->data
[i
].mem
);
318 ret
= nvkm_gpuobj_map_vm(chan
->data
[i
].mem
, vm
, data
->access
,
326 /* finally, fill in the mmio list and point the context at it */
327 for (i
= 0; mmio
->addr
&& i
< ARRAY_SIZE(gr
->mmio_list
); i
++) {
328 u32 addr
= mmio
->addr
;
329 u32 data
= mmio
->data
;
331 if (mmio
->buffer
>= 0) {
332 u64 info
= chan
->data
[mmio
->buffer
].vma
.offset
;
333 data
|= info
>> mmio
->shift
;
336 nv_wo32(chan
->mmio
, chan
->mmio_nr
++ * 4, addr
);
337 nv_wo32(chan
->mmio
, chan
->mmio_nr
++ * 4, data
);
341 for (i
= 0; i
< gr
->size
; i
+= 4)
342 nv_wo32(chan
, i
, gr
->data
[i
/ 4]);
345 nv_wo32(chan
, 0x00, chan
->mmio_nr
/ 2);
346 nv_wo32(chan
, 0x04, chan
->mmio_vma
.offset
>> 8);
348 nv_wo32(chan
, 0xf4, 0);
349 nv_wo32(chan
, 0xf8, 0);
350 nv_wo32(chan
, 0x10, chan
->mmio_nr
/ 2);
351 nv_wo32(chan
, 0x14, lower_32_bits(chan
->mmio_vma
.offset
));
352 nv_wo32(chan
, 0x18, upper_32_bits(chan
->mmio_vma
.offset
));
353 nv_wo32(chan
, 0x1c, 1);
354 nv_wo32(chan
, 0x20, 0);
355 nv_wo32(chan
, 0x28, 0);
356 nv_wo32(chan
, 0x2c, 0);
363 gf100_gr_context_dtor(struct nvkm_object
*object
)
365 struct gf100_gr_chan
*chan
= (void *)object
;
368 for (i
= 0; i
< ARRAY_SIZE(chan
->data
); i
++) {
369 nvkm_gpuobj_unmap(&chan
->data
[i
].vma
);
370 nvkm_gpuobj_ref(NULL
, &chan
->data
[i
].mem
);
373 nvkm_gpuobj_unmap(&chan
->mmio_vma
);
374 nvkm_gpuobj_ref(NULL
, &chan
->mmio
);
376 nvkm_gr_context_destroy(&chan
->base
);
379 /*******************************************************************************
380 * PGRAPH register lists
381 ******************************************************************************/
383 const struct gf100_gr_init
384 gf100_gr_init_main_0
[] = {
385 { 0x400080, 1, 0x04, 0x003083c2 },
386 { 0x400088, 1, 0x04, 0x00006fe7 },
387 { 0x40008c, 1, 0x04, 0x00000000 },
388 { 0x400090, 1, 0x04, 0x00000030 },
389 { 0x40013c, 1, 0x04, 0x013901f7 },
390 { 0x400140, 1, 0x04, 0x00000100 },
391 { 0x400144, 1, 0x04, 0x00000000 },
392 { 0x400148, 1, 0x04, 0x00000110 },
393 { 0x400138, 1, 0x04, 0x00000000 },
394 { 0x400130, 2, 0x04, 0x00000000 },
395 { 0x400124, 1, 0x04, 0x00000002 },
399 const struct gf100_gr_init
400 gf100_gr_init_fe_0
[] = {
401 { 0x40415c, 1, 0x04, 0x00000000 },
402 { 0x404170, 1, 0x04, 0x00000000 },
406 const struct gf100_gr_init
407 gf100_gr_init_pri_0
[] = {
408 { 0x404488, 2, 0x04, 0x00000000 },
412 const struct gf100_gr_init
413 gf100_gr_init_rstr2d_0
[] = {
414 { 0x407808, 1, 0x04, 0x00000000 },
418 const struct gf100_gr_init
419 gf100_gr_init_pd_0
[] = {
420 { 0x406024, 1, 0x04, 0x00000000 },
424 const struct gf100_gr_init
425 gf100_gr_init_ds_0
[] = {
426 { 0x405844, 1, 0x04, 0x00ffffff },
427 { 0x405850, 1, 0x04, 0x00000000 },
428 { 0x405908, 1, 0x04, 0x00000000 },
432 const struct gf100_gr_init
433 gf100_gr_init_scc_0
[] = {
434 { 0x40803c, 1, 0x04, 0x00000000 },
438 const struct gf100_gr_init
439 gf100_gr_init_prop_0
[] = {
440 { 0x4184a0, 1, 0x04, 0x00000000 },
444 const struct gf100_gr_init
445 gf100_gr_init_gpc_unk_0
[] = {
446 { 0x418604, 1, 0x04, 0x00000000 },
447 { 0x418680, 1, 0x04, 0x00000000 },
448 { 0x418714, 1, 0x04, 0x80000000 },
449 { 0x418384, 1, 0x04, 0x00000000 },
453 const struct gf100_gr_init
454 gf100_gr_init_setup_0
[] = {
455 { 0x418814, 3, 0x04, 0x00000000 },
459 const struct gf100_gr_init
460 gf100_gr_init_crstr_0
[] = {
461 { 0x418b04, 1, 0x04, 0x00000000 },
465 const struct gf100_gr_init
466 gf100_gr_init_setup_1
[] = {
467 { 0x4188c8, 1, 0x04, 0x80000000 },
468 { 0x4188cc, 1, 0x04, 0x00000000 },
469 { 0x4188d0, 1, 0x04, 0x00010000 },
470 { 0x4188d4, 1, 0x04, 0x00000001 },
474 const struct gf100_gr_init
475 gf100_gr_init_zcull_0
[] = {
476 { 0x418910, 1, 0x04, 0x00010001 },
477 { 0x418914, 1, 0x04, 0x00000301 },
478 { 0x418918, 1, 0x04, 0x00800000 },
479 { 0x418980, 1, 0x04, 0x77777770 },
480 { 0x418984, 3, 0x04, 0x77777777 },
484 const struct gf100_gr_init
485 gf100_gr_init_gpm_0
[] = {
486 { 0x418c04, 1, 0x04, 0x00000000 },
487 { 0x418c88, 1, 0x04, 0x00000000 },
491 const struct gf100_gr_init
492 gf100_gr_init_gpc_unk_1
[] = {
493 { 0x418d00, 1, 0x04, 0x00000000 },
494 { 0x418f08, 1, 0x04, 0x00000000 },
495 { 0x418e00, 1, 0x04, 0x00000050 },
496 { 0x418e08, 1, 0x04, 0x00000000 },
500 const struct gf100_gr_init
501 gf100_gr_init_gcc_0
[] = {
502 { 0x41900c, 1, 0x04, 0x00000000 },
503 { 0x419018, 1, 0x04, 0x00000000 },
507 const struct gf100_gr_init
508 gf100_gr_init_tpccs_0
[] = {
509 { 0x419d08, 2, 0x04, 0x00000000 },
510 { 0x419d10, 1, 0x04, 0x00000014 },
514 const struct gf100_gr_init
515 gf100_gr_init_tex_0
[] = {
516 { 0x419ab0, 1, 0x04, 0x00000000 },
517 { 0x419ab8, 1, 0x04, 0x000000e7 },
518 { 0x419abc, 2, 0x04, 0x00000000 },
522 const struct gf100_gr_init
523 gf100_gr_init_pe_0
[] = {
524 { 0x41980c, 3, 0x04, 0x00000000 },
525 { 0x419844, 1, 0x04, 0x00000000 },
526 { 0x41984c, 1, 0x04, 0x00005bc5 },
527 { 0x419850, 4, 0x04, 0x00000000 },
531 const struct gf100_gr_init
532 gf100_gr_init_l1c_0
[] = {
533 { 0x419c98, 1, 0x04, 0x00000000 },
534 { 0x419ca8, 1, 0x04, 0x80000000 },
535 { 0x419cb4, 1, 0x04, 0x00000000 },
536 { 0x419cb8, 1, 0x04, 0x00008bf4 },
537 { 0x419cbc, 1, 0x04, 0x28137606 },
538 { 0x419cc0, 2, 0x04, 0x00000000 },
542 const struct gf100_gr_init
543 gf100_gr_init_wwdx_0
[] = {
544 { 0x419bd4, 1, 0x04, 0x00800000 },
545 { 0x419bdc, 1, 0x04, 0x00000000 },
549 const struct gf100_gr_init
550 gf100_gr_init_tpccs_1
[] = {
551 { 0x419d2c, 1, 0x04, 0x00000000 },
555 const struct gf100_gr_init
556 gf100_gr_init_mpc_0
[] = {
557 { 0x419c0c, 1, 0x04, 0x00000000 },
561 static const struct gf100_gr_init
562 gf100_gr_init_sm_0
[] = {
563 { 0x419e00, 1, 0x04, 0x00000000 },
564 { 0x419ea0, 1, 0x04, 0x00000000 },
565 { 0x419ea4, 1, 0x04, 0x00000100 },
566 { 0x419ea8, 1, 0x04, 0x00001100 },
567 { 0x419eac, 1, 0x04, 0x11100702 },
568 { 0x419eb0, 1, 0x04, 0x00000003 },
569 { 0x419eb4, 4, 0x04, 0x00000000 },
570 { 0x419ec8, 1, 0x04, 0x06060618 },
571 { 0x419ed0, 1, 0x04, 0x0eff0e38 },
572 { 0x419ed4, 1, 0x04, 0x011104f1 },
573 { 0x419edc, 1, 0x04, 0x00000000 },
574 { 0x419f00, 1, 0x04, 0x00000000 },
575 { 0x419f2c, 1, 0x04, 0x00000000 },
579 const struct gf100_gr_init
580 gf100_gr_init_be_0
[] = {
581 { 0x40880c, 1, 0x04, 0x00000000 },
582 { 0x408910, 9, 0x04, 0x00000000 },
583 { 0x408950, 1, 0x04, 0x00000000 },
584 { 0x408954, 1, 0x04, 0x0000ffff },
585 { 0x408984, 1, 0x04, 0x00000000 },
586 { 0x408988, 1, 0x04, 0x08040201 },
587 { 0x40898c, 1, 0x04, 0x80402010 },
591 const struct gf100_gr_init
592 gf100_gr_init_fe_1
[] = {
593 { 0x4040f0, 1, 0x04, 0x00000000 },
597 const struct gf100_gr_init
598 gf100_gr_init_pe_1
[] = {
599 { 0x419880, 1, 0x04, 0x00000002 },
603 static const struct gf100_gr_pack
604 gf100_gr_pack_mmio
[] = {
605 { gf100_gr_init_main_0
},
606 { gf100_gr_init_fe_0
},
607 { gf100_gr_init_pri_0
},
608 { gf100_gr_init_rstr2d_0
},
609 { gf100_gr_init_pd_0
},
610 { gf100_gr_init_ds_0
},
611 { gf100_gr_init_scc_0
},
612 { gf100_gr_init_prop_0
},
613 { gf100_gr_init_gpc_unk_0
},
614 { gf100_gr_init_setup_0
},
615 { gf100_gr_init_crstr_0
},
616 { gf100_gr_init_setup_1
},
617 { gf100_gr_init_zcull_0
},
618 { gf100_gr_init_gpm_0
},
619 { gf100_gr_init_gpc_unk_1
},
620 { gf100_gr_init_gcc_0
},
621 { gf100_gr_init_tpccs_0
},
622 { gf100_gr_init_tex_0
},
623 { gf100_gr_init_pe_0
},
624 { gf100_gr_init_l1c_0
},
625 { gf100_gr_init_wwdx_0
},
626 { gf100_gr_init_tpccs_1
},
627 { gf100_gr_init_mpc_0
},
628 { gf100_gr_init_sm_0
},
629 { gf100_gr_init_be_0
},
630 { gf100_gr_init_fe_1
},
631 { gf100_gr_init_pe_1
},
635 /*******************************************************************************
636 * PGRAPH engine/subdev functions
637 ******************************************************************************/
640 gf100_gr_zbc_init(struct gf100_gr
*gr
)
642 const u32 zero
[] = { 0x00000000, 0x00000000, 0x00000000, 0x00000000,
643 0x00000000, 0x00000000, 0x00000000, 0x00000000 };
644 const u32 one
[] = { 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000,
645 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff };
646 const u32 f32_0
[] = { 0x00000000, 0x00000000, 0x00000000, 0x00000000,
647 0x00000000, 0x00000000, 0x00000000, 0x00000000 };
648 const u32 f32_1
[] = { 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000,
649 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000 };
650 struct nvkm_ltc
*ltc
= nvkm_ltc(gr
);
653 if (!gr
->zbc_color
[0].format
) {
654 gf100_gr_zbc_color_get(gr
, 1, & zero
[0], &zero
[4]);
655 gf100_gr_zbc_color_get(gr
, 2, & one
[0], &one
[4]);
656 gf100_gr_zbc_color_get(gr
, 4, &f32_0
[0], &f32_0
[4]);
657 gf100_gr_zbc_color_get(gr
, 4, &f32_1
[0], &f32_1
[4]);
658 gf100_gr_zbc_depth_get(gr
, 1, 0x00000000, 0x00000000);
659 gf100_gr_zbc_depth_get(gr
, 1, 0x3f800000, 0x3f800000);
662 for (index
= ltc
->zbc_min
; index
<= ltc
->zbc_max
; index
++)
663 gf100_gr_zbc_clear_color(gr
, index
);
664 for (index
= ltc
->zbc_min
; index
<= ltc
->zbc_max
; index
++)
665 gf100_gr_zbc_clear_depth(gr
, index
);
669 * Wait until GR goes idle. GR is considered idle if it is disabled by the
670 * MC (0x200) register, or GR is not busy and a context switch is not in
674 gf100_gr_wait_idle(struct gf100_gr
*gr
)
676 struct nvkm_device
*device
= gr
->base
.engine
.subdev
.device
;
677 unsigned long end_jiffies
= jiffies
+ msecs_to_jiffies(2000);
678 bool gr_enabled
, ctxsw_active
, gr_busy
;
682 * required to make sure FIFO_ENGINE_STATUS (0x2640) is
685 nvkm_rd32(device
, 0x400700);
687 gr_enabled
= nvkm_rd32(device
, 0x200) & 0x1000;
688 ctxsw_active
= nvkm_rd32(device
, 0x2640) & 0x8000;
689 gr_busy
= nvkm_rd32(device
, 0x40060c) & 0x1;
691 if (!gr_enabled
|| (!gr_busy
&& !ctxsw_active
))
693 } while (time_before(jiffies
, end_jiffies
));
695 nv_error(gr
, "wait for idle timeout (en: %d, ctxsw: %d, busy: %d)\n",
696 gr_enabled
, ctxsw_active
, gr_busy
);
701 gf100_gr_mmio(struct gf100_gr
*gr
, const struct gf100_gr_pack
*p
)
703 struct nvkm_device
*device
= gr
->base
.engine
.subdev
.device
;
704 const struct gf100_gr_pack
*pack
;
705 const struct gf100_gr_init
*init
;
707 pack_for_each_init(init
, pack
, p
) {
708 u32 next
= init
->addr
+ init
->count
* init
->pitch
;
709 u32 addr
= init
->addr
;
710 while (addr
< next
) {
711 nvkm_wr32(device
, addr
, init
->data
);
718 gf100_gr_icmd(struct gf100_gr
*gr
, const struct gf100_gr_pack
*p
)
720 struct nvkm_device
*device
= gr
->base
.engine
.subdev
.device
;
721 const struct gf100_gr_pack
*pack
;
722 const struct gf100_gr_init
*init
;
725 nvkm_wr32(device
, 0x400208, 0x80000000);
727 pack_for_each_init(init
, pack
, p
) {
728 u32 next
= init
->addr
+ init
->count
* init
->pitch
;
729 u32 addr
= init
->addr
;
731 if ((pack
== p
&& init
== p
->init
) || data
!= init
->data
) {
732 nvkm_wr32(device
, 0x400204, init
->data
);
736 while (addr
< next
) {
737 nvkm_wr32(device
, 0x400200, addr
);
739 * Wait for GR to go idle after submitting a
742 if ((addr
& 0xffff) == 0xe100)
743 gf100_gr_wait_idle(gr
);
744 nv_wait(gr
, 0x400700, 0x00000004, 0x00000000);
749 nvkm_wr32(device
, 0x400208, 0x00000000);
753 gf100_gr_mthd(struct gf100_gr
*gr
, const struct gf100_gr_pack
*p
)
755 struct nvkm_device
*device
= gr
->base
.engine
.subdev
.device
;
756 const struct gf100_gr_pack
*pack
;
757 const struct gf100_gr_init
*init
;
760 pack_for_each_init(init
, pack
, p
) {
761 u32 ctrl
= 0x80000000 | pack
->type
;
762 u32 next
= init
->addr
+ init
->count
* init
->pitch
;
763 u32 addr
= init
->addr
;
765 if ((pack
== p
&& init
== p
->init
) || data
!= init
->data
) {
766 nvkm_wr32(device
, 0x40448c, init
->data
);
770 while (addr
< next
) {
771 nvkm_wr32(device
, 0x404488, ctrl
| (addr
<< 14));
778 gf100_gr_units(struct nvkm_gr
*obj
)
780 struct gf100_gr
*gr
= container_of(obj
, typeof(*gr
), base
);
783 cfg
= (u32
)gr
->gpc_nr
;
784 cfg
|= (u32
)gr
->tpc_total
<< 8;
785 cfg
|= (u64
)gr
->rop_nr
<< 32;
790 static const struct nvkm_enum gk104_sked_error
[] = {
791 { 7, "CONSTANT_BUFFER_SIZE" },
792 { 9, "LOCAL_MEMORY_SIZE_POS" },
793 { 10, "LOCAL_MEMORY_SIZE_NEG" },
794 { 11, "WARP_CSTACK_SIZE" },
795 { 12, "TOTAL_TEMP_SIZE" },
796 { 13, "REGISTER_COUNT" },
797 { 18, "TOTAL_THREADS" },
798 { 20, "PROGRAM_OFFSET" },
799 { 21, "SHARED_MEMORY_SIZE" },
800 { 25, "SHARED_CONFIG_TOO_SMALL" },
801 { 26, "TOTAL_REGISTER_COUNT" },
805 static const struct nvkm_enum gf100_gpc_rop_error
[] = {
806 { 1, "RT_PITCH_OVERRUN" },
807 { 4, "RT_WIDTH_OVERRUN" },
808 { 5, "RT_HEIGHT_OVERRUN" },
809 { 7, "ZETA_STORAGE_TYPE_MISMATCH" },
810 { 8, "RT_STORAGE_TYPE_MISMATCH" },
811 { 10, "RT_LINEAR_MISMATCH" },
816 gf100_gr_trap_gpc_rop(struct gf100_gr
*gr
, int gpc
)
818 struct nvkm_device
*device
= gr
->base
.engine
.subdev
.device
;
822 trap
[0] = nvkm_rd32(device
, GPC_UNIT(gpc
, 0x0420));
823 trap
[1] = nvkm_rd32(device
, GPC_UNIT(gpc
, 0x0434));
824 trap
[2] = nvkm_rd32(device
, GPC_UNIT(gpc
, 0x0438));
825 trap
[3] = nvkm_rd32(device
, GPC_UNIT(gpc
, 0x043c));
827 nv_error(gr
, "GPC%d/PROP trap:", gpc
);
828 for (i
= 0; i
<= 29; ++i
) {
829 if (!(trap
[0] & (1 << i
)))
832 nvkm_enum_print(gf100_gpc_rop_error
, i
);
836 nv_error(gr
, "x = %u, y = %u, format = %x, storage type = %x\n",
837 trap
[1] & 0xffff, trap
[1] >> 16, (trap
[2] >> 8) & 0x3f,
839 nvkm_wr32(device
, GPC_UNIT(gpc
, 0x0420), 0xc0000000);
842 static const struct nvkm_enum gf100_mp_warp_error
[] = {
843 { 0x00, "NO_ERROR" },
844 { 0x01, "STACK_MISMATCH" },
845 { 0x05, "MISALIGNED_PC" },
846 { 0x08, "MISALIGNED_GPR" },
847 { 0x09, "INVALID_OPCODE" },
848 { 0x0d, "GPR_OUT_OF_BOUNDS" },
849 { 0x0e, "MEM_OUT_OF_BOUNDS" },
850 { 0x0f, "UNALIGNED_MEM_ACCESS" },
851 { 0x11, "INVALID_PARAM" },
855 static const struct nvkm_bitfield gf100_mp_global_error
[] = {
856 { 0x00000004, "MULTIPLE_WARP_ERRORS" },
857 { 0x00000008, "OUT_OF_STACK_SPACE" },
862 gf100_gr_trap_mp(struct gf100_gr
*gr
, int gpc
, int tpc
)
864 struct nvkm_device
*device
= gr
->base
.engine
.subdev
.device
;
865 u32 werr
= nvkm_rd32(device
, TPC_UNIT(gpc
, tpc
, 0x648));
866 u32 gerr
= nvkm_rd32(device
, TPC_UNIT(gpc
, tpc
, 0x650));
868 nv_error(gr
, "GPC%i/TPC%i/MP trap:", gpc
, tpc
);
869 nvkm_bitfield_print(gf100_mp_global_error
, gerr
);
872 nvkm_enum_print(gf100_mp_warp_error
, werr
& 0xffff);
876 nvkm_wr32(device
, TPC_UNIT(gpc
, tpc
, 0x648), 0x00000000);
877 nvkm_wr32(device
, TPC_UNIT(gpc
, tpc
, 0x650), gerr
);
881 gf100_gr_trap_tpc(struct gf100_gr
*gr
, int gpc
, int tpc
)
883 struct nvkm_device
*device
= gr
->base
.engine
.subdev
.device
;
884 u32 stat
= nvkm_rd32(device
, TPC_UNIT(gpc
, tpc
, 0x0508));
886 if (stat
& 0x00000001) {
887 u32 trap
= nvkm_rd32(device
, TPC_UNIT(gpc
, tpc
, 0x0224));
888 nv_error(gr
, "GPC%d/TPC%d/TEX: 0x%08x\n", gpc
, tpc
, trap
);
889 nvkm_wr32(device
, TPC_UNIT(gpc
, tpc
, 0x0224), 0xc0000000);
893 if (stat
& 0x00000002) {
894 gf100_gr_trap_mp(gr
, gpc
, tpc
);
898 if (stat
& 0x00000004) {
899 u32 trap
= nvkm_rd32(device
, TPC_UNIT(gpc
, tpc
, 0x0084));
900 nv_error(gr
, "GPC%d/TPC%d/POLY: 0x%08x\n", gpc
, tpc
, trap
);
901 nvkm_wr32(device
, TPC_UNIT(gpc
, tpc
, 0x0084), 0xc0000000);
905 if (stat
& 0x00000008) {
906 u32 trap
= nvkm_rd32(device
, TPC_UNIT(gpc
, tpc
, 0x048c));
907 nv_error(gr
, "GPC%d/TPC%d/L1C: 0x%08x\n", gpc
, tpc
, trap
);
908 nvkm_wr32(device
, TPC_UNIT(gpc
, tpc
, 0x048c), 0xc0000000);
913 nv_error(gr
, "GPC%d/TPC%d/0x%08x: unknown\n", gpc
, tpc
, stat
);
918 gf100_gr_trap_gpc(struct gf100_gr
*gr
, int gpc
)
920 struct nvkm_device
*device
= gr
->base
.engine
.subdev
.device
;
921 u32 stat
= nvkm_rd32(device
, GPC_UNIT(gpc
, 0x2c90));
924 if (stat
& 0x00000001) {
925 gf100_gr_trap_gpc_rop(gr
, gpc
);
929 if (stat
& 0x00000002) {
930 u32 trap
= nvkm_rd32(device
, GPC_UNIT(gpc
, 0x0900));
931 nv_error(gr
, "GPC%d/ZCULL: 0x%08x\n", gpc
, trap
);
932 nvkm_wr32(device
, GPC_UNIT(gpc
, 0x0900), 0xc0000000);
936 if (stat
& 0x00000004) {
937 u32 trap
= nvkm_rd32(device
, GPC_UNIT(gpc
, 0x1028));
938 nv_error(gr
, "GPC%d/CCACHE: 0x%08x\n", gpc
, trap
);
939 nvkm_wr32(device
, GPC_UNIT(gpc
, 0x1028), 0xc0000000);
943 if (stat
& 0x00000008) {
944 u32 trap
= nvkm_rd32(device
, GPC_UNIT(gpc
, 0x0824));
945 nv_error(gr
, "GPC%d/ESETUP: 0x%08x\n", gpc
, trap
);
946 nvkm_wr32(device
, GPC_UNIT(gpc
, 0x0824), 0xc0000000);
950 for (tpc
= 0; tpc
< gr
->tpc_nr
[gpc
]; tpc
++) {
951 u32 mask
= 0x00010000 << tpc
;
953 gf100_gr_trap_tpc(gr
, gpc
, tpc
);
954 nvkm_wr32(device
, GPC_UNIT(gpc
, 0x2c90), mask
);
960 nv_error(gr
, "GPC%d/0x%08x: unknown\n", gpc
, stat
);
965 gf100_gr_trap_intr(struct gf100_gr
*gr
)
967 struct nvkm_device
*device
= gr
->base
.engine
.subdev
.device
;
968 u32 trap
= nvkm_rd32(device
, 0x400108);
971 if (trap
& 0x00000001) {
972 u32 stat
= nvkm_rd32(device
, 0x404000);
973 nv_error(gr
, "DISPATCH 0x%08x\n", stat
);
974 nvkm_wr32(device
, 0x404000, 0xc0000000);
975 nvkm_wr32(device
, 0x400108, 0x00000001);
979 if (trap
& 0x00000002) {
980 u32 stat
= nvkm_rd32(device
, 0x404600);
981 nv_error(gr
, "M2MF 0x%08x\n", stat
);
982 nvkm_wr32(device
, 0x404600, 0xc0000000);
983 nvkm_wr32(device
, 0x400108, 0x00000002);
987 if (trap
& 0x00000008) {
988 u32 stat
= nvkm_rd32(device
, 0x408030);
989 nv_error(gr
, "CCACHE 0x%08x\n", stat
);
990 nvkm_wr32(device
, 0x408030, 0xc0000000);
991 nvkm_wr32(device
, 0x400108, 0x00000008);
995 if (trap
& 0x00000010) {
996 u32 stat
= nvkm_rd32(device
, 0x405840);
997 nv_error(gr
, "SHADER 0x%08x\n", stat
);
998 nvkm_wr32(device
, 0x405840, 0xc0000000);
999 nvkm_wr32(device
, 0x400108, 0x00000010);
1000 trap
&= ~0x00000010;
1003 if (trap
& 0x00000040) {
1004 u32 stat
= nvkm_rd32(device
, 0x40601c);
1005 nv_error(gr
, "UNK6 0x%08x\n", stat
);
1006 nvkm_wr32(device
, 0x40601c, 0xc0000000);
1007 nvkm_wr32(device
, 0x400108, 0x00000040);
1008 trap
&= ~0x00000040;
1011 if (trap
& 0x00000080) {
1012 u32 stat
= nvkm_rd32(device
, 0x404490);
1013 nv_error(gr
, "MACRO 0x%08x\n", stat
);
1014 nvkm_wr32(device
, 0x404490, 0xc0000000);
1015 nvkm_wr32(device
, 0x400108, 0x00000080);
1016 trap
&= ~0x00000080;
1019 if (trap
& 0x00000100) {
1020 u32 stat
= nvkm_rd32(device
, 0x407020);
1022 nv_error(gr
, "SKED:");
1023 for (i
= 0; i
<= 29; ++i
) {
1024 if (!(stat
& (1 << i
)))
1027 nvkm_enum_print(gk104_sked_error
, i
);
1031 if (stat
& 0x3fffffff)
1032 nvkm_wr32(device
, 0x407020, 0x40000000);
1033 nvkm_wr32(device
, 0x400108, 0x00000100);
1034 trap
&= ~0x00000100;
1037 if (trap
& 0x01000000) {
1038 u32 stat
= nvkm_rd32(device
, 0x400118);
1039 for (gpc
= 0; stat
&& gpc
< gr
->gpc_nr
; gpc
++) {
1040 u32 mask
= 0x00000001 << gpc
;
1042 gf100_gr_trap_gpc(gr
, gpc
);
1043 nvkm_wr32(device
, 0x400118, mask
);
1047 nvkm_wr32(device
, 0x400108, 0x01000000);
1048 trap
&= ~0x01000000;
1051 if (trap
& 0x02000000) {
1052 for (rop
= 0; rop
< gr
->rop_nr
; rop
++) {
1053 u32 statz
= nvkm_rd32(device
, ROP_UNIT(rop
, 0x070));
1054 u32 statc
= nvkm_rd32(device
, ROP_UNIT(rop
, 0x144));
1055 nv_error(gr
, "ROP%d 0x%08x 0x%08x\n",
1057 nvkm_wr32(device
, ROP_UNIT(rop
, 0x070), 0xc0000000);
1058 nvkm_wr32(device
, ROP_UNIT(rop
, 0x144), 0xc0000000);
1060 nvkm_wr32(device
, 0x400108, 0x02000000);
1061 trap
&= ~0x02000000;
1065 nv_error(gr
, "TRAP UNHANDLED 0x%08x\n", trap
);
1066 nvkm_wr32(device
, 0x400108, trap
);
1071 gf100_gr_ctxctl_debug_unit(struct gf100_gr
*gr
, u32 base
)
1073 struct nvkm_device
*device
= gr
->base
.engine
.subdev
.device
;
1074 nv_error(gr
, "%06x - done 0x%08x\n", base
,
1075 nvkm_rd32(device
, base
+ 0x400));
1076 nv_error(gr
, "%06x - stat 0x%08x 0x%08x 0x%08x 0x%08x\n", base
,
1077 nvkm_rd32(device
, base
+ 0x800), nvkm_rd32(device
, base
+ 0x804),
1078 nvkm_rd32(device
, base
+ 0x808), nvkm_rd32(device
, base
+ 0x80c));
1079 nv_error(gr
, "%06x - stat 0x%08x 0x%08x 0x%08x 0x%08x\n", base
,
1080 nvkm_rd32(device
, base
+ 0x810), nvkm_rd32(device
, base
+ 0x814),
1081 nvkm_rd32(device
, base
+ 0x818), nvkm_rd32(device
, base
+ 0x81c));
1085 gf100_gr_ctxctl_debug(struct gf100_gr
*gr
)
1087 struct nvkm_device
*device
= gr
->base
.engine
.subdev
.device
;
1088 u32 gpcnr
= nvkm_rd32(device
, 0x409604) & 0xffff;
1091 gf100_gr_ctxctl_debug_unit(gr
, 0x409000);
1092 for (gpc
= 0; gpc
< gpcnr
; gpc
++)
1093 gf100_gr_ctxctl_debug_unit(gr
, 0x502000 + (gpc
* 0x8000));
1097 gf100_gr_ctxctl_isr(struct gf100_gr
*gr
)
1099 struct nvkm_device
*device
= gr
->base
.engine
.subdev
.device
;
1100 u32 stat
= nvkm_rd32(device
, 0x409c18);
1102 if (stat
& 0x00000001) {
1103 u32 code
= nvkm_rd32(device
, 0x409814);
1104 if (code
== E_BAD_FWMTHD
) {
1105 u32
class = nvkm_rd32(device
, 0x409808);
1106 u32 addr
= nvkm_rd32(device
, 0x40980c);
1107 u32 subc
= (addr
& 0x00070000) >> 16;
1108 u32 mthd
= (addr
& 0x00003ffc);
1109 u32 data
= nvkm_rd32(device
, 0x409810);
1111 nv_error(gr
, "FECS MTHD subc %d class 0x%04x "
1112 "mthd 0x%04x data 0x%08x\n",
1113 subc
, class, mthd
, data
);
1115 nvkm_wr32(device
, 0x409c20, 0x00000001);
1116 stat
&= ~0x00000001;
1118 nv_error(gr
, "FECS ucode error %d\n", code
);
1122 if (stat
& 0x00080000) {
1123 nv_error(gr
, "FECS watchdog timeout\n");
1124 gf100_gr_ctxctl_debug(gr
);
1125 nvkm_wr32(device
, 0x409c20, 0x00080000);
1126 stat
&= ~0x00080000;
1130 nv_error(gr
, "FECS 0x%08x\n", stat
);
1131 gf100_gr_ctxctl_debug(gr
);
1132 nvkm_wr32(device
, 0x409c20, stat
);
1137 gf100_gr_intr(struct nvkm_subdev
*subdev
)
1139 struct gf100_gr
*gr
= (void *)subdev
;
1140 struct nvkm_device
*device
= gr
->base
.engine
.subdev
.device
;
1141 struct nvkm_fifo
*fifo
= device
->fifo
;
1142 struct nvkm_engine
*engine
= nv_engine(subdev
);
1143 struct nvkm_object
*engctx
;
1144 struct nvkm_handle
*handle
;
1145 u64 inst
= nvkm_rd32(device
, 0x409b00) & 0x0fffffff;
1146 u32 stat
= nvkm_rd32(device
, 0x400100);
1147 u32 addr
= nvkm_rd32(device
, 0x400704);
1148 u32 mthd
= (addr
& 0x00003ffc);
1149 u32 subc
= (addr
& 0x00070000) >> 16;
1150 u32 data
= nvkm_rd32(device
, 0x400708);
1151 u32 code
= nvkm_rd32(device
, 0x400110);
1155 if (nv_device(gr
)->card_type
< NV_E0
|| subc
< 4)
1156 class = nvkm_rd32(device
, 0x404200 + (subc
* 4));
1160 engctx
= nvkm_engctx_get(engine
, inst
);
1161 chid
= fifo
->chid(fifo
, engctx
);
1163 if (stat
& 0x00000001) {
1165 * notifier interrupt, only needed for cyclestats
1166 * can be safely ignored
1168 nvkm_wr32(device
, 0x400100, 0x00000001);
1169 stat
&= ~0x00000001;
1172 if (stat
& 0x00000010) {
1173 handle
= nvkm_handle_get_class(engctx
, class);
1174 if (!handle
|| nv_call(handle
->object
, mthd
, data
)) {
1176 "ILLEGAL_MTHD ch %d [0x%010llx %s] subc %d class 0x%04x mthd 0x%04x data 0x%08x\n",
1177 chid
, inst
<< 12, nvkm_client_name(engctx
),
1178 subc
, class, mthd
, data
);
1180 nvkm_handle_put(handle
);
1181 nvkm_wr32(device
, 0x400100, 0x00000010);
1182 stat
&= ~0x00000010;
1185 if (stat
& 0x00000020) {
1187 "ILLEGAL_CLASS ch %d [0x%010llx %s] subc %d class 0x%04x mthd 0x%04x data 0x%08x\n",
1188 chid
, inst
<< 12, nvkm_client_name(engctx
), subc
,
1190 nvkm_wr32(device
, 0x400100, 0x00000020);
1191 stat
&= ~0x00000020;
1194 if (stat
& 0x00100000) {
1195 nv_error(gr
, "DATA_ERROR [");
1196 nvkm_enum_print(nv50_data_error_names
, code
);
1197 pr_cont("] ch %d [0x%010llx %s] subc %d class 0x%04x mthd 0x%04x data 0x%08x\n",
1198 chid
, inst
<< 12, nvkm_client_name(engctx
), subc
,
1200 nvkm_wr32(device
, 0x400100, 0x00100000);
1201 stat
&= ~0x00100000;
1204 if (stat
& 0x00200000) {
1205 nv_error(gr
, "TRAP ch %d [0x%010llx %s]\n", chid
, inst
<< 12,
1206 nvkm_client_name(engctx
));
1207 gf100_gr_trap_intr(gr
);
1208 nvkm_wr32(device
, 0x400100, 0x00200000);
1209 stat
&= ~0x00200000;
1212 if (stat
& 0x00080000) {
1213 gf100_gr_ctxctl_isr(gr
);
1214 nvkm_wr32(device
, 0x400100, 0x00080000);
1215 stat
&= ~0x00080000;
1219 nv_error(gr
, "unknown stat 0x%08x\n", stat
);
1220 nvkm_wr32(device
, 0x400100, stat
);
1223 nvkm_wr32(device
, 0x400500, 0x00010001);
1224 nvkm_engctx_put(engctx
);
1228 gf100_gr_init_fw(struct gf100_gr
*gr
, u32 fuc_base
,
1229 struct gf100_gr_fuc
*code
, struct gf100_gr_fuc
*data
)
1231 struct nvkm_device
*device
= gr
->base
.engine
.subdev
.device
;
1234 nvkm_wr32(device
, fuc_base
+ 0x01c0, 0x01000000);
1235 for (i
= 0; i
< data
->size
/ 4; i
++)
1236 nvkm_wr32(device
, fuc_base
+ 0x01c4, data
->data
[i
]);
1238 nvkm_wr32(device
, fuc_base
+ 0x0180, 0x01000000);
1239 for (i
= 0; i
< code
->size
/ 4; i
++) {
1240 if ((i
& 0x3f) == 0)
1241 nvkm_wr32(device
, fuc_base
+ 0x0188, i
>> 6);
1242 nvkm_wr32(device
, fuc_base
+ 0x0184, code
->data
[i
]);
1245 /* code must be padded to 0x40 words */
1246 for (; i
& 0x3f; i
++)
1247 nvkm_wr32(device
, fuc_base
+ 0x0184, 0);
1251 gf100_gr_init_csdata(struct gf100_gr
*gr
,
1252 const struct gf100_gr_pack
*pack
,
1253 u32 falcon
, u32 starstar
, u32 base
)
1255 struct nvkm_device
*device
= gr
->base
.engine
.subdev
.device
;
1256 const struct gf100_gr_pack
*iter
;
1257 const struct gf100_gr_init
*init
;
1258 u32 addr
= ~0, prev
= ~0, xfer
= 0;
1261 nvkm_wr32(device
, falcon
+ 0x01c0, 0x02000000 + starstar
);
1262 star
= nvkm_rd32(device
, falcon
+ 0x01c4);
1263 temp
= nvkm_rd32(device
, falcon
+ 0x01c4);
1266 nvkm_wr32(device
, falcon
+ 0x01c0, 0x01000000 + star
);
1268 pack_for_each_init(init
, iter
, pack
) {
1269 u32 head
= init
->addr
- base
;
1270 u32 tail
= head
+ init
->count
* init
->pitch
;
1271 while (head
< tail
) {
1272 if (head
!= prev
+ 4 || xfer
>= 32) {
1274 u32 data
= ((--xfer
<< 26) | addr
);
1275 nvkm_wr32(device
, falcon
+ 0x01c4, data
);
1283 head
= head
+ init
->pitch
;
1287 nvkm_wr32(device
, falcon
+ 0x01c4, (--xfer
<< 26) | addr
);
1288 nvkm_wr32(device
, falcon
+ 0x01c0, 0x01000004 + starstar
);
1289 nvkm_wr32(device
, falcon
+ 0x01c4, star
+ 4);
1293 gf100_gr_init_ctxctl(struct gf100_gr
*gr
)
1295 struct nvkm_device
*device
= gr
->base
.engine
.subdev
.device
;
1296 struct gf100_gr_oclass
*oclass
= (void *)nv_object(gr
)->oclass
;
1297 struct gf100_grctx_oclass
*cclass
= (void *)nv_engine(gr
)->cclass
;
1301 /* load fuc microcode */
1302 nvkm_mc(gr
)->unk260(nvkm_mc(gr
), 0);
1303 gf100_gr_init_fw(gr
, 0x409000, &gr
->fuc409c
,
1305 gf100_gr_init_fw(gr
, 0x41a000, &gr
->fuc41ac
,
1307 nvkm_mc(gr
)->unk260(nvkm_mc(gr
), 1);
1309 /* start both of them running */
1310 nvkm_wr32(device
, 0x409840, 0xffffffff);
1311 nvkm_wr32(device
, 0x41a10c, 0x00000000);
1312 nvkm_wr32(device
, 0x40910c, 0x00000000);
1313 nvkm_wr32(device
, 0x41a100, 0x00000002);
1314 nvkm_wr32(device
, 0x409100, 0x00000002);
1315 if (!nv_wait(gr
, 0x409800, 0x00000001, 0x00000001))
1316 nv_warn(gr
, "0x409800 wait failed\n");
1318 nvkm_wr32(device
, 0x409840, 0xffffffff);
1319 nvkm_wr32(device
, 0x409500, 0x7fffffff);
1320 nvkm_wr32(device
, 0x409504, 0x00000021);
1322 nvkm_wr32(device
, 0x409840, 0xffffffff);
1323 nvkm_wr32(device
, 0x409500, 0x00000000);
1324 nvkm_wr32(device
, 0x409504, 0x00000010);
1325 if (!nv_wait_ne(gr
, 0x409800, 0xffffffff, 0x00000000)) {
1326 nv_error(gr
, "fuc09 req 0x10 timeout\n");
1329 gr
->size
= nvkm_rd32(device
, 0x409800);
1331 nvkm_wr32(device
, 0x409840, 0xffffffff);
1332 nvkm_wr32(device
, 0x409500, 0x00000000);
1333 nvkm_wr32(device
, 0x409504, 0x00000016);
1334 if (!nv_wait_ne(gr
, 0x409800, 0xffffffff, 0x00000000)) {
1335 nv_error(gr
, "fuc09 req 0x16 timeout\n");
1339 nvkm_wr32(device
, 0x409840, 0xffffffff);
1340 nvkm_wr32(device
, 0x409500, 0x00000000);
1341 nvkm_wr32(device
, 0x409504, 0x00000025);
1342 if (!nv_wait_ne(gr
, 0x409800, 0xffffffff, 0x00000000)) {
1343 nv_error(gr
, "fuc09 req 0x25 timeout\n");
1347 if (nv_device(gr
)->chipset
>= 0xe0) {
1348 nvkm_wr32(device
, 0x409800, 0x00000000);
1349 nvkm_wr32(device
, 0x409500, 0x00000001);
1350 nvkm_wr32(device
, 0x409504, 0x00000030);
1351 if (!nv_wait_ne(gr
, 0x409800, 0xffffffff, 0x00000000)) {
1352 nv_error(gr
, "fuc09 req 0x30 timeout\n");
1356 nvkm_wr32(device
, 0x409810, 0xb00095c8);
1357 nvkm_wr32(device
, 0x409800, 0x00000000);
1358 nvkm_wr32(device
, 0x409500, 0x00000001);
1359 nvkm_wr32(device
, 0x409504, 0x00000031);
1360 if (!nv_wait_ne(gr
, 0x409800, 0xffffffff, 0x00000000)) {
1361 nv_error(gr
, "fuc09 req 0x31 timeout\n");
1365 nvkm_wr32(device
, 0x409810, 0x00080420);
1366 nvkm_wr32(device
, 0x409800, 0x00000000);
1367 nvkm_wr32(device
, 0x409500, 0x00000001);
1368 nvkm_wr32(device
, 0x409504, 0x00000032);
1369 if (!nv_wait_ne(gr
, 0x409800, 0xffffffff, 0x00000000)) {
1370 nv_error(gr
, "fuc09 req 0x32 timeout\n");
1374 nvkm_wr32(device
, 0x409614, 0x00000070);
1375 nvkm_wr32(device
, 0x409614, 0x00000770);
1376 nvkm_wr32(device
, 0x40802c, 0x00000001);
1379 if (gr
->data
== NULL
) {
1380 int ret
= gf100_grctx_generate(gr
);
1382 nv_error(gr
, "failed to construct context\n");
1389 if (!oclass
->fecs
.ucode
) {
1393 /* load HUB microcode */
1394 nvkm_mc(gr
)->unk260(nvkm_mc(gr
), 0);
1395 nvkm_wr32(device
, 0x4091c0, 0x01000000);
1396 for (i
= 0; i
< oclass
->fecs
.ucode
->data
.size
/ 4; i
++)
1397 nvkm_wr32(device
, 0x4091c4, oclass
->fecs
.ucode
->data
.data
[i
]);
1399 nvkm_wr32(device
, 0x409180, 0x01000000);
1400 for (i
= 0; i
< oclass
->fecs
.ucode
->code
.size
/ 4; i
++) {
1401 if ((i
& 0x3f) == 0)
1402 nvkm_wr32(device
, 0x409188, i
>> 6);
1403 nvkm_wr32(device
, 0x409184, oclass
->fecs
.ucode
->code
.data
[i
]);
1406 /* load GPC microcode */
1407 nvkm_wr32(device
, 0x41a1c0, 0x01000000);
1408 for (i
= 0; i
< oclass
->gpccs
.ucode
->data
.size
/ 4; i
++)
1409 nvkm_wr32(device
, 0x41a1c4, oclass
->gpccs
.ucode
->data
.data
[i
]);
1411 nvkm_wr32(device
, 0x41a180, 0x01000000);
1412 for (i
= 0; i
< oclass
->gpccs
.ucode
->code
.size
/ 4; i
++) {
1413 if ((i
& 0x3f) == 0)
1414 nvkm_wr32(device
, 0x41a188, i
>> 6);
1415 nvkm_wr32(device
, 0x41a184, oclass
->gpccs
.ucode
->code
.data
[i
]);
1417 nvkm_mc(gr
)->unk260(nvkm_mc(gr
), 1);
1419 /* load register lists */
1420 gf100_gr_init_csdata(gr
, cclass
->hub
, 0x409000, 0x000, 0x000000);
1421 gf100_gr_init_csdata(gr
, cclass
->gpc
, 0x41a000, 0x000, 0x418000);
1422 gf100_gr_init_csdata(gr
, cclass
->tpc
, 0x41a000, 0x004, 0x419800);
1423 gf100_gr_init_csdata(gr
, cclass
->ppc
, 0x41a000, 0x008, 0x41be00);
1425 /* start HUB ucode running, it'll init the GPCs */
1426 nvkm_wr32(device
, 0x40910c, 0x00000000);
1427 nvkm_wr32(device
, 0x409100, 0x00000002);
1428 if (!nv_wait(gr
, 0x409800, 0x80000000, 0x80000000)) {
1429 nv_error(gr
, "HUB_INIT timed out\n");
1430 gf100_gr_ctxctl_debug(gr
);
1434 gr
->size
= nvkm_rd32(device
, 0x409804);
1435 if (gr
->data
== NULL
) {
1436 int ret
= gf100_grctx_generate(gr
);
1438 nv_error(gr
, "failed to construct context\n");
1447 gf100_gr_init(struct nvkm_object
*object
)
1449 struct gf100_gr
*gr
= (void *)object
;
1450 struct nvkm_device
*device
= gr
->base
.engine
.subdev
.device
;
1451 struct gf100_gr_oclass
*oclass
= (void *)object
->oclass
;
1452 const u32 magicgpc918
= DIV_ROUND_UP(0x00800000, gr
->tpc_total
);
1453 u32 data
[TPC_MAX
/ 8] = {};
1458 ret
= nvkm_gr_init(&gr
->base
);
1462 nvkm_wr32(device
, GPC_BCAST(0x0880), 0x00000000);
1463 nvkm_wr32(device
, GPC_BCAST(0x08a4), 0x00000000);
1464 nvkm_wr32(device
, GPC_BCAST(0x0888), 0x00000000);
1465 nvkm_wr32(device
, GPC_BCAST(0x088c), 0x00000000);
1466 nvkm_wr32(device
, GPC_BCAST(0x0890), 0x00000000);
1467 nvkm_wr32(device
, GPC_BCAST(0x0894), 0x00000000);
1468 nvkm_wr32(device
, GPC_BCAST(0x08b4), gr
->unk4188b4
->addr
>> 8);
1469 nvkm_wr32(device
, GPC_BCAST(0x08b8), gr
->unk4188b8
->addr
>> 8);
1471 gf100_gr_mmio(gr
, oclass
->mmio
);
1473 memcpy(tpcnr
, gr
->tpc_nr
, sizeof(gr
->tpc_nr
));
1474 for (i
= 0, gpc
= -1; i
< gr
->tpc_total
; i
++) {
1476 gpc
= (gpc
+ 1) % gr
->gpc_nr
;
1477 } while (!tpcnr
[gpc
]);
1478 tpc
= gr
->tpc_nr
[gpc
] - tpcnr
[gpc
]--;
1480 data
[i
/ 8] |= tpc
<< ((i
% 8) * 4);
1483 nvkm_wr32(device
, GPC_BCAST(0x0980), data
[0]);
1484 nvkm_wr32(device
, GPC_BCAST(0x0984), data
[1]);
1485 nvkm_wr32(device
, GPC_BCAST(0x0988), data
[2]);
1486 nvkm_wr32(device
, GPC_BCAST(0x098c), data
[3]);
1488 for (gpc
= 0; gpc
< gr
->gpc_nr
; gpc
++) {
1489 nvkm_wr32(device
, GPC_UNIT(gpc
, 0x0914),
1490 gr
->magic_not_rop_nr
<< 8 | gr
->tpc_nr
[gpc
]);
1491 nvkm_wr32(device
, GPC_UNIT(gpc
, 0x0910), 0x00040000 |
1493 nvkm_wr32(device
, GPC_UNIT(gpc
, 0x0918), magicgpc918
);
1496 if (nv_device(gr
)->chipset
!= 0xd7)
1497 nvkm_wr32(device
, GPC_BCAST(0x1bd4), magicgpc918
);
1499 nvkm_wr32(device
, GPC_BCAST(0x3fd4), magicgpc918
);
1501 nvkm_wr32(device
, GPC_BCAST(0x08ac), nvkm_rd32(device
, 0x100800));
1503 nvkm_wr32(device
, 0x400500, 0x00010001);
1505 nvkm_wr32(device
, 0x400100, 0xffffffff);
1506 nvkm_wr32(device
, 0x40013c, 0xffffffff);
1508 nvkm_wr32(device
, 0x409c24, 0x000f0000);
1509 nvkm_wr32(device
, 0x404000, 0xc0000000);
1510 nvkm_wr32(device
, 0x404600, 0xc0000000);
1511 nvkm_wr32(device
, 0x408030, 0xc0000000);
1512 nvkm_wr32(device
, 0x40601c, 0xc0000000);
1513 nvkm_wr32(device
, 0x404490, 0xc0000000);
1514 nvkm_wr32(device
, 0x406018, 0xc0000000);
1515 nvkm_wr32(device
, 0x405840, 0xc0000000);
1516 nvkm_wr32(device
, 0x405844, 0x00ffffff);
1517 nvkm_mask(device
, 0x419cc0, 0x00000008, 0x00000008);
1518 nvkm_mask(device
, 0x419eb4, 0x00001000, 0x00001000);
1520 for (gpc
= 0; gpc
< gr
->gpc_nr
; gpc
++) {
1521 nvkm_wr32(device
, GPC_UNIT(gpc
, 0x0420), 0xc0000000);
1522 nvkm_wr32(device
, GPC_UNIT(gpc
, 0x0900), 0xc0000000);
1523 nvkm_wr32(device
, GPC_UNIT(gpc
, 0x1028), 0xc0000000);
1524 nvkm_wr32(device
, GPC_UNIT(gpc
, 0x0824), 0xc0000000);
1525 for (tpc
= 0; tpc
< gr
->tpc_nr
[gpc
]; tpc
++) {
1526 nvkm_wr32(device
, TPC_UNIT(gpc
, tpc
, 0x508), 0xffffffff);
1527 nvkm_wr32(device
, TPC_UNIT(gpc
, tpc
, 0x50c), 0xffffffff);
1528 nvkm_wr32(device
, TPC_UNIT(gpc
, tpc
, 0x224), 0xc0000000);
1529 nvkm_wr32(device
, TPC_UNIT(gpc
, tpc
, 0x48c), 0xc0000000);
1530 nvkm_wr32(device
, TPC_UNIT(gpc
, tpc
, 0x084), 0xc0000000);
1531 nvkm_wr32(device
, TPC_UNIT(gpc
, tpc
, 0x644), 0x001ffffe);
1532 nvkm_wr32(device
, TPC_UNIT(gpc
, tpc
, 0x64c), 0x0000000f);
1534 nvkm_wr32(device
, GPC_UNIT(gpc
, 0x2c90), 0xffffffff);
1535 nvkm_wr32(device
, GPC_UNIT(gpc
, 0x2c94), 0xffffffff);
1538 for (rop
= 0; rop
< gr
->rop_nr
; rop
++) {
1539 nvkm_wr32(device
, ROP_UNIT(rop
, 0x144), 0xc0000000);
1540 nvkm_wr32(device
, ROP_UNIT(rop
, 0x070), 0xc0000000);
1541 nvkm_wr32(device
, ROP_UNIT(rop
, 0x204), 0xffffffff);
1542 nvkm_wr32(device
, ROP_UNIT(rop
, 0x208), 0xffffffff);
1545 nvkm_wr32(device
, 0x400108, 0xffffffff);
1546 nvkm_wr32(device
, 0x400138, 0xffffffff);
1547 nvkm_wr32(device
, 0x400118, 0xffffffff);
1548 nvkm_wr32(device
, 0x400130, 0xffffffff);
1549 nvkm_wr32(device
, 0x40011c, 0xffffffff);
1550 nvkm_wr32(device
, 0x400134, 0xffffffff);
1552 nvkm_wr32(device
, 0x400054, 0x34ce3464);
1554 gf100_gr_zbc_init(gr
);
1556 return gf100_gr_init_ctxctl(gr
);
1560 gf100_gr_dtor_fw(struct gf100_gr_fuc
*fuc
)
1567 gf100_gr_ctor_fw(struct gf100_gr
*gr
, const char *fwname
,
1568 struct gf100_gr_fuc
*fuc
)
1570 struct nvkm_device
*device
= nv_device(gr
);
1571 const struct firmware
*fw
;
1577 /* Convert device name to lowercase */
1578 strncpy(cname
, device
->cname
, sizeof(cname
));
1579 cname
[sizeof(cname
) - 1] = '\0';
1583 cname
[i
] = tolower(cname
[i
]);
1586 snprintf(f
, sizeof(f
), "nvidia/%s/%s.bin", cname
, fwname
);
1587 ret
= request_firmware(&fw
, f
, nv_device_base(device
));
1589 nv_error(gr
, "failed to load %s\n", fwname
);
1593 fuc
->size
= fw
->size
;
1594 fuc
->data
= kmemdup(fw
->data
, fuc
->size
, GFP_KERNEL
);
1595 release_firmware(fw
);
1596 return (fuc
->data
!= NULL
) ? 0 : -ENOMEM
;
1600 gf100_gr_dtor(struct nvkm_object
*object
)
1602 struct gf100_gr
*gr
= (void *)object
;
1606 gf100_gr_dtor_fw(&gr
->fuc409c
);
1607 gf100_gr_dtor_fw(&gr
->fuc409d
);
1608 gf100_gr_dtor_fw(&gr
->fuc41ac
);
1609 gf100_gr_dtor_fw(&gr
->fuc41ad
);
1611 nvkm_gpuobj_ref(NULL
, &gr
->unk4188b8
);
1612 nvkm_gpuobj_ref(NULL
, &gr
->unk4188b4
);
1614 nvkm_gr_destroy(&gr
->base
);
1618 gf100_gr_ctor(struct nvkm_object
*parent
, struct nvkm_object
*engine
,
1619 struct nvkm_oclass
*bclass
, void *data
, u32 size
,
1620 struct nvkm_object
**pobject
)
1622 struct gf100_gr_oclass
*oclass
= (void *)bclass
;
1623 struct nvkm_device
*device
= nv_device(parent
);
1624 struct gf100_gr
*gr
;
1625 bool use_ext_fw
, enable
;
1628 use_ext_fw
= nvkm_boolopt(device
->cfgopt
, "NvGrUseFW",
1629 oclass
->fecs
.ucode
== NULL
);
1630 enable
= use_ext_fw
|| oclass
->fecs
.ucode
!= NULL
;
1632 ret
= nvkm_gr_create(parent
, engine
, bclass
, enable
, &gr
);
1633 *pobject
= nv_object(gr
);
1637 nv_subdev(gr
)->unit
= 0x08001000;
1638 nv_subdev(gr
)->intr
= gf100_gr_intr
;
1640 gr
->base
.units
= gf100_gr_units
;
1643 nv_info(gr
, "using external firmware\n");
1644 if (gf100_gr_ctor_fw(gr
, "fecs_inst", &gr
->fuc409c
) ||
1645 gf100_gr_ctor_fw(gr
, "fecs_data", &gr
->fuc409d
) ||
1646 gf100_gr_ctor_fw(gr
, "gpccs_inst", &gr
->fuc41ac
) ||
1647 gf100_gr_ctor_fw(gr
, "gpccs_data", &gr
->fuc41ad
))
1649 gr
->firmware
= true;
1652 ret
= nvkm_gpuobj_new(nv_object(gr
), NULL
, 0x1000, 256, 0,
1657 ret
= nvkm_gpuobj_new(nv_object(gr
), NULL
, 0x1000, 256, 0,
1662 for (i
= 0; i
< 0x1000; i
+= 4) {
1663 nv_wo32(gr
->unk4188b4
, i
, 0x00000010);
1664 nv_wo32(gr
->unk4188b8
, i
, 0x00000010);
1667 gr
->rop_nr
= (nvkm_rd32(device
, 0x409604) & 0x001f0000) >> 16;
1668 gr
->gpc_nr
= nvkm_rd32(device
, 0x409604) & 0x0000001f;
1669 for (i
= 0; i
< gr
->gpc_nr
; i
++) {
1670 gr
->tpc_nr
[i
] = nvkm_rd32(device
, GPC_UNIT(i
, 0x2608));
1671 gr
->tpc_total
+= gr
->tpc_nr
[i
];
1672 gr
->ppc_nr
[i
] = oclass
->ppc_nr
;
1673 for (j
= 0; j
< gr
->ppc_nr
[i
]; j
++) {
1674 u8 mask
= nvkm_rd32(device
, GPC_UNIT(i
, 0x0c30 + (j
* 4)));
1675 gr
->ppc_tpc_nr
[i
][j
] = hweight8(mask
);
1679 /*XXX: these need figuring out... though it might not even matter */
1680 switch (nv_device(gr
)->chipset
) {
1682 if (gr
->tpc_total
== 11) { /* 465, 3/4/4/0, 4 */
1683 gr
->magic_not_rop_nr
= 0x07;
1685 if (gr
->tpc_total
== 14) { /* 470, 3/3/4/4, 5 */
1686 gr
->magic_not_rop_nr
= 0x05;
1688 if (gr
->tpc_total
== 15) { /* 480, 3/4/4/4, 6 */
1689 gr
->magic_not_rop_nr
= 0x06;
1692 case 0xc3: /* 450, 4/0/0/0, 2 */
1693 gr
->magic_not_rop_nr
= 0x03;
1695 case 0xc4: /* 460, 3/4/0/0, 4 */
1696 gr
->magic_not_rop_nr
= 0x01;
1698 case 0xc1: /* 2/0/0/0, 1 */
1699 gr
->magic_not_rop_nr
= 0x01;
1701 case 0xc8: /* 4/4/3/4, 5 */
1702 gr
->magic_not_rop_nr
= 0x06;
1704 case 0xce: /* 4/4/0/0, 4 */
1705 gr
->magic_not_rop_nr
= 0x03;
1707 case 0xcf: /* 4/0/0/0, 3 */
1708 gr
->magic_not_rop_nr
= 0x03;
1711 case 0xd9: /* 1/0/0/0, 1 */
1712 case 0xea: /* gk20a */
1713 case 0x12b: /* gm20b */
1714 gr
->magic_not_rop_nr
= 0x01;
1718 nv_engine(gr
)->cclass
= *oclass
->cclass
;
1719 nv_engine(gr
)->sclass
= oclass
->sclass
;
1723 #include "fuc/hubgf100.fuc3.h"
1725 struct gf100_gr_ucode
1726 gf100_gr_fecs_ucode
= {
1727 .code
.data
= gf100_grhub_code
,
1728 .code
.size
= sizeof(gf100_grhub_code
),
1729 .data
.data
= gf100_grhub_data
,
1730 .data
.size
= sizeof(gf100_grhub_data
),
1733 #include "fuc/gpcgf100.fuc3.h"
1735 struct gf100_gr_ucode
1736 gf100_gr_gpccs_ucode
= {
1737 .code
.data
= gf100_grgpc_code
,
1738 .code
.size
= sizeof(gf100_grgpc_code
),
1739 .data
.data
= gf100_grgpc_data
,
1740 .data
.size
= sizeof(gf100_grgpc_data
),
1743 struct nvkm_oclass
*
1744 gf100_gr_oclass
= &(struct gf100_gr_oclass
) {
1745 .base
.handle
= NV_ENGINE(GR
, 0xc0),
1746 .base
.ofuncs
= &(struct nvkm_ofuncs
) {
1747 .ctor
= gf100_gr_ctor
,
1748 .dtor
= gf100_gr_dtor
,
1749 .init
= gf100_gr_init
,
1750 .fini
= _nvkm_gr_fini
,
1752 .cclass
= &gf100_grctx_oclass
,
1753 .sclass
= gf100_gr_sclass
,
1754 .mmio
= gf100_gr_pack_mmio
,
1755 .fecs
.ucode
= &gf100_gr_fecs_ucode
,
1756 .gpccs
.ucode
= &gf100_gr_gpccs_ucode
,