]>
Commit | Line | Data |
---|---|---|
4b223eef | 1 | /* |
ebb945a9 | 2 | * Copyright 2012 Red Hat Inc. |
4b223eef BS |
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 | */ | |
e3c71eb2 BS |
24 | #include "gf100.h" |
25 | #include "ctxgf100.h" | |
26 | #include "fuc/os.h" | |
27 | ||
28 | #include <core/client.h> | |
e3c71eb2 | 29 | #include <core/option.h> |
33bcb4c3 | 30 | #include <core/firmware.h> |
c9469aae | 31 | #include <subdev/secboot.h> |
e3c71eb2 BS |
32 | #include <subdev/fb.h> |
33 | #include <subdev/mc.h> | |
c85ee6ca | 34 | #include <subdev/pmu.h> |
e3c71eb2 | 35 | #include <subdev/timer.h> |
a65955e1 | 36 | #include <engine/fifo.h> |
e3c71eb2 BS |
37 | |
38 | #include <nvif/class.h> | |
53a6df77 | 39 | #include <nvif/cl9097.h> |
019e4d76 | 40 | #include <nvif/if900d.h> |
e3c71eb2 | 41 | #include <nvif/unpack.h> |
0411de85 | 42 | |
ac9738bb BS |
43 | /******************************************************************************* |
44 | * Zero Bandwidth Clear | |
45 | ******************************************************************************/ | |
46 | ||
47 | static void | |
bfee3f3d | 48 | gf100_gr_zbc_clear_color(struct gf100_gr *gr, int zbc) |
ac9738bb | 49 | { |
276836d4 | 50 | struct nvkm_device *device = gr->base.engine.subdev.device; |
bfee3f3d | 51 | if (gr->zbc_color[zbc].format) { |
276836d4 BS |
52 | nvkm_wr32(device, 0x405804, gr->zbc_color[zbc].ds[0]); |
53 | nvkm_wr32(device, 0x405808, gr->zbc_color[zbc].ds[1]); | |
54 | nvkm_wr32(device, 0x40580c, gr->zbc_color[zbc].ds[2]); | |
55 | nvkm_wr32(device, 0x405810, gr->zbc_color[zbc].ds[3]); | |
56 | } | |
57 | nvkm_wr32(device, 0x405814, gr->zbc_color[zbc].format); | |
58 | nvkm_wr32(device, 0x405820, zbc); | |
59 | nvkm_wr32(device, 0x405824, 0x00000004); /* TRIGGER | WRITE | COLOR */ | |
ac9738bb BS |
60 | } |
61 | ||
62 | static int | |
bfee3f3d | 63 | gf100_gr_zbc_color_get(struct gf100_gr *gr, int format, |
e3c71eb2 | 64 | const u32 ds[4], const u32 l2[4]) |
ac9738bb | 65 | { |
70bc7182 | 66 | struct nvkm_ltc *ltc = gr->base.engine.subdev.device->ltc; |
ac9738bb BS |
67 | int zbc = -ENOSPC, i; |
68 | ||
69 | for (i = ltc->zbc_min; i <= ltc->zbc_max; i++) { | |
bfee3f3d BS |
70 | if (gr->zbc_color[i].format) { |
71 | if (gr->zbc_color[i].format != format) | |
ac9738bb | 72 | continue; |
bfee3f3d BS |
73 | if (memcmp(gr->zbc_color[i].ds, ds, sizeof( |
74 | gr->zbc_color[i].ds))) | |
ac9738bb | 75 | continue; |
bfee3f3d BS |
76 | if (memcmp(gr->zbc_color[i].l2, l2, sizeof( |
77 | gr->zbc_color[i].l2))) { | |
ac9738bb BS |
78 | WARN_ON(1); |
79 | return -EINVAL; | |
80 | } | |
81 | return i; | |
82 | } else { | |
83 | zbc = (zbc < 0) ? i : zbc; | |
84 | } | |
85 | } | |
86 | ||
da7c74ea BS |
87 | if (zbc < 0) |
88 | return zbc; | |
89 | ||
bfee3f3d BS |
90 | memcpy(gr->zbc_color[zbc].ds, ds, sizeof(gr->zbc_color[zbc].ds)); |
91 | memcpy(gr->zbc_color[zbc].l2, l2, sizeof(gr->zbc_color[zbc].l2)); | |
92 | gr->zbc_color[zbc].format = format; | |
70bc7182 | 93 | nvkm_ltc_zbc_color_get(ltc, zbc, l2); |
bfee3f3d | 94 | gf100_gr_zbc_clear_color(gr, zbc); |
ac9738bb BS |
95 | return zbc; |
96 | } | |
97 | ||
98 | static void | |
bfee3f3d | 99 | gf100_gr_zbc_clear_depth(struct gf100_gr *gr, int zbc) |
ac9738bb | 100 | { |
276836d4 | 101 | struct nvkm_device *device = gr->base.engine.subdev.device; |
bfee3f3d | 102 | if (gr->zbc_depth[zbc].format) |
276836d4 BS |
103 | nvkm_wr32(device, 0x405818, gr->zbc_depth[zbc].ds); |
104 | nvkm_wr32(device, 0x40581c, gr->zbc_depth[zbc].format); | |
105 | nvkm_wr32(device, 0x405820, zbc); | |
106 | nvkm_wr32(device, 0x405824, 0x00000005); /* TRIGGER | WRITE | DEPTH */ | |
ac9738bb BS |
107 | } |
108 | ||
109 | static int | |
bfee3f3d | 110 | gf100_gr_zbc_depth_get(struct gf100_gr *gr, int format, |
e3c71eb2 | 111 | const u32 ds, const u32 l2) |
ac9738bb | 112 | { |
70bc7182 | 113 | struct nvkm_ltc *ltc = gr->base.engine.subdev.device->ltc; |
ac9738bb BS |
114 | int zbc = -ENOSPC, i; |
115 | ||
116 | for (i = ltc->zbc_min; i <= ltc->zbc_max; i++) { | |
bfee3f3d BS |
117 | if (gr->zbc_depth[i].format) { |
118 | if (gr->zbc_depth[i].format != format) | |
ac9738bb | 119 | continue; |
bfee3f3d | 120 | if (gr->zbc_depth[i].ds != ds) |
ac9738bb | 121 | continue; |
bfee3f3d | 122 | if (gr->zbc_depth[i].l2 != l2) { |
ac9738bb BS |
123 | WARN_ON(1); |
124 | return -EINVAL; | |
125 | } | |
126 | return i; | |
127 | } else { | |
128 | zbc = (zbc < 0) ? i : zbc; | |
129 | } | |
130 | } | |
131 | ||
da7c74ea BS |
132 | if (zbc < 0) |
133 | return zbc; | |
134 | ||
bfee3f3d BS |
135 | gr->zbc_depth[zbc].format = format; |
136 | gr->zbc_depth[zbc].ds = ds; | |
137 | gr->zbc_depth[zbc].l2 = l2; | |
70bc7182 | 138 | nvkm_ltc_zbc_depth_get(ltc, zbc, l2); |
bfee3f3d | 139 | gf100_gr_zbc_clear_depth(gr, zbc); |
ac9738bb BS |
140 | return zbc; |
141 | } | |
142 | ||
ebb945a9 BS |
143 | /******************************************************************************* |
144 | * Graphics object classes | |
145 | ******************************************************************************/ | |
5bf561ee BS |
146 | #define gf100_gr_object(p) container_of((p), struct gf100_gr_object, object) |
147 | ||
148 | struct gf100_gr_object { | |
149 | struct nvkm_object object; | |
150 | struct gf100_gr_chan *chan; | |
151 | }; | |
ebb945a9 | 152 | |
ac9738bb | 153 | static int |
e3c71eb2 | 154 | gf100_fermi_mthd_zbc_color(struct nvkm_object *object, void *data, u32 size) |
ac9738bb | 155 | { |
0d7fc246 | 156 | struct gf100_gr *gr = gf100_gr(nvkm_gr(object->engine)); |
ac9738bb BS |
157 | union { |
158 | struct fermi_a_zbc_color_v0 v0; | |
159 | } *args = data; | |
f01c4e68 | 160 | int ret = -ENOSYS; |
ac9738bb | 161 | |
f01c4e68 | 162 | if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, false))) { |
ac9738bb BS |
163 | switch (args->v0.format) { |
164 | case FERMI_A_ZBC_COLOR_V0_FMT_ZERO: | |
165 | case FERMI_A_ZBC_COLOR_V0_FMT_UNORM_ONE: | |
166 | case FERMI_A_ZBC_COLOR_V0_FMT_RF32_GF32_BF32_AF32: | |
167 | case FERMI_A_ZBC_COLOR_V0_FMT_R16_G16_B16_A16: | |
168 | case FERMI_A_ZBC_COLOR_V0_FMT_RN16_GN16_BN16_AN16: | |
169 | case FERMI_A_ZBC_COLOR_V0_FMT_RS16_GS16_BS16_AS16: | |
170 | case FERMI_A_ZBC_COLOR_V0_FMT_RU16_GU16_BU16_AU16: | |
171 | case FERMI_A_ZBC_COLOR_V0_FMT_RF16_GF16_BF16_AF16: | |
172 | case FERMI_A_ZBC_COLOR_V0_FMT_A8R8G8B8: | |
173 | case FERMI_A_ZBC_COLOR_V0_FMT_A8RL8GL8BL8: | |
174 | case FERMI_A_ZBC_COLOR_V0_FMT_A2B10G10R10: | |
175 | case FERMI_A_ZBC_COLOR_V0_FMT_AU2BU10GU10RU10: | |
176 | case FERMI_A_ZBC_COLOR_V0_FMT_A8B8G8R8: | |
177 | case FERMI_A_ZBC_COLOR_V0_FMT_A8BL8GL8RL8: | |
178 | case FERMI_A_ZBC_COLOR_V0_FMT_AN8BN8GN8RN8: | |
179 | case FERMI_A_ZBC_COLOR_V0_FMT_AS8BS8GS8RS8: | |
180 | case FERMI_A_ZBC_COLOR_V0_FMT_AU8BU8GU8RU8: | |
181 | case FERMI_A_ZBC_COLOR_V0_FMT_A2R10G10B10: | |
182 | case FERMI_A_ZBC_COLOR_V0_FMT_BF10GF11RF11: | |
bfee3f3d | 183 | ret = gf100_gr_zbc_color_get(gr, args->v0.format, |
e3c71eb2 BS |
184 | args->v0.ds, |
185 | args->v0.l2); | |
ac9738bb BS |
186 | if (ret >= 0) { |
187 | args->v0.index = ret; | |
188 | return 0; | |
189 | } | |
190 | break; | |
191 | default: | |
192 | return -EINVAL; | |
193 | } | |
194 | } | |
195 | ||
196 | return ret; | |
197 | } | |
198 | ||
199 | static int | |
e3c71eb2 | 200 | gf100_fermi_mthd_zbc_depth(struct nvkm_object *object, void *data, u32 size) |
ac9738bb | 201 | { |
0d7fc246 | 202 | struct gf100_gr *gr = gf100_gr(nvkm_gr(object->engine)); |
ac9738bb BS |
203 | union { |
204 | struct fermi_a_zbc_depth_v0 v0; | |
205 | } *args = data; | |
f01c4e68 | 206 | int ret = -ENOSYS; |
ac9738bb | 207 | |
f01c4e68 | 208 | if (!(ret = nvif_unpack(ret, &data, &size, args->v0, 0, 0, false))) { |
ac9738bb BS |
209 | switch (args->v0.format) { |
210 | case FERMI_A_ZBC_DEPTH_V0_FMT_FP32: | |
bfee3f3d | 211 | ret = gf100_gr_zbc_depth_get(gr, args->v0.format, |
e3c71eb2 BS |
212 | args->v0.ds, |
213 | args->v0.l2); | |
ac9738bb BS |
214 | return (ret >= 0) ? 0 : -ENOSPC; |
215 | default: | |
216 | return -EINVAL; | |
217 | } | |
218 | } | |
219 | ||
220 | return ret; | |
221 | } | |
222 | ||
223 | static int | |
e3c71eb2 | 224 | gf100_fermi_mthd(struct nvkm_object *object, u32 mthd, void *data, u32 size) |
ac9738bb | 225 | { |
f01c4e68 | 226 | nvif_ioctl(object, "fermi mthd %08x\n", mthd); |
ac9738bb BS |
227 | switch (mthd) { |
228 | case FERMI_A_ZBC_COLOR: | |
e3c71eb2 | 229 | return gf100_fermi_mthd_zbc_color(object, data, size); |
ac9738bb | 230 | case FERMI_A_ZBC_DEPTH: |
e3c71eb2 | 231 | return gf100_fermi_mthd_zbc_depth(object, data, size); |
ac9738bb BS |
232 | default: |
233 | break; | |
234 | } | |
235 | return -EINVAL; | |
236 | } | |
237 | ||
27f3d6cf BS |
238 | const struct nvkm_object_func |
239 | gf100_fermi = { | |
e3c71eb2 | 240 | .mthd = gf100_fermi_mthd, |
ac9738bb BS |
241 | }; |
242 | ||
a65955e1 BS |
243 | static void |
244 | gf100_gr_mthd_set_shader_exceptions(struct nvkm_device *device, u32 data) | |
d6bd3803 | 245 | { |
a65955e1 BS |
246 | nvkm_wr32(device, 0x419e44, data ? 0xffffffff : 0x00000000); |
247 | nvkm_wr32(device, 0x419e4c, data ? 0xffffffff : 0x00000000); | |
d6bd3803 BS |
248 | } |
249 | ||
a65955e1 BS |
250 | static bool |
251 | gf100_gr_mthd_sw(struct nvkm_device *device, u16 class, u32 mthd, u32 data) | |
252 | { | |
253 | switch (class & 0x00ff) { | |
254 | case 0x97: | |
255 | case 0xc0: | |
256 | switch (mthd) { | |
257 | case 0x1528: | |
258 | gf100_gr_mthd_set_shader_exceptions(device, data); | |
259 | return true; | |
260 | default: | |
261 | break; | |
262 | } | |
263 | break; | |
264 | default: | |
265 | break; | |
266 | } | |
267 | return false; | |
268 | } | |
d6bd3803 | 269 | |
5bf561ee BS |
270 | static const struct nvkm_object_func |
271 | gf100_gr_object_func = { | |
272 | }; | |
273 | ||
274 | static int | |
275 | gf100_gr_object_new(const struct nvkm_oclass *oclass, void *data, u32 size, | |
276 | struct nvkm_object **pobject) | |
277 | { | |
278 | struct gf100_gr_chan *chan = gf100_gr_chan(oclass->parent); | |
279 | struct gf100_gr_object *object; | |
280 | ||
281 | if (!(object = kzalloc(sizeof(*object), GFP_KERNEL))) | |
282 | return -ENOMEM; | |
283 | *pobject = &object->object; | |
284 | ||
285 | nvkm_object_ctor(oclass->base.func ? oclass->base.func : | |
286 | &gf100_gr_object_func, oclass, &object->object); | |
287 | object->chan = chan; | |
288 | return 0; | |
289 | } | |
290 | ||
27f3d6cf BS |
291 | static int |
292 | gf100_gr_object_get(struct nvkm_gr *base, int index, struct nvkm_sclass *sclass) | |
293 | { | |
294 | struct gf100_gr *gr = gf100_gr(base); | |
295 | int c = 0; | |
296 | ||
297 | while (gr->func->sclass[c].oclass) { | |
298 | if (c++ == index) { | |
299 | *sclass = gr->func->sclass[index]; | |
5bf561ee | 300 | sclass->ctor = gf100_gr_object_new; |
27f3d6cf BS |
301 | return index; |
302 | } | |
303 | } | |
304 | ||
305 | return c; | |
306 | } | |
ebb945a9 | 307 | |
ebb945a9 BS |
308 | /******************************************************************************* |
309 | * PGRAPH context | |
310 | ******************************************************************************/ | |
966a5b7d | 311 | |
27f3d6cf BS |
312 | static int |
313 | gf100_gr_chan_bind(struct nvkm_object *object, struct nvkm_gpuobj *parent, | |
314 | int align, struct nvkm_gpuobj **pgpuobj) | |
966a5b7d | 315 | { |
27f3d6cf BS |
316 | struct gf100_gr_chan *chan = gf100_gr_chan(object); |
317 | struct gf100_gr *gr = chan->gr; | |
318 | int ret, i; | |
319 | ||
320 | ret = nvkm_gpuobj_new(gr->base.engine.subdev.device, gr->size, | |
321 | align, false, parent, pgpuobj); | |
322 | if (ret) | |
323 | return ret; | |
324 | ||
325 | nvkm_kmap(*pgpuobj); | |
326 | for (i = 0; i < gr->size; i += 4) | |
327 | nvkm_wo32(*pgpuobj, i, gr->data[i / 4]); | |
328 | ||
329 | if (!gr->firmware) { | |
330 | nvkm_wo32(*pgpuobj, 0x00, chan->mmio_nr / 2); | |
019e4d76 | 331 | nvkm_wo32(*pgpuobj, 0x04, chan->mmio_vma->addr >> 8); |
27f3d6cf BS |
332 | } else { |
333 | nvkm_wo32(*pgpuobj, 0xf4, 0); | |
334 | nvkm_wo32(*pgpuobj, 0xf8, 0); | |
335 | nvkm_wo32(*pgpuobj, 0x10, chan->mmio_nr / 2); | |
019e4d76 BS |
336 | nvkm_wo32(*pgpuobj, 0x14, lower_32_bits(chan->mmio_vma->addr)); |
337 | nvkm_wo32(*pgpuobj, 0x18, upper_32_bits(chan->mmio_vma->addr)); | |
27f3d6cf BS |
338 | nvkm_wo32(*pgpuobj, 0x1c, 1); |
339 | nvkm_wo32(*pgpuobj, 0x20, 0); | |
340 | nvkm_wo32(*pgpuobj, 0x28, 0); | |
341 | nvkm_wo32(*pgpuobj, 0x2c, 0); | |
342 | } | |
343 | nvkm_done(*pgpuobj); | |
344 | return 0; | |
345 | } | |
346 | ||
347 | static void * | |
348 | gf100_gr_chan_dtor(struct nvkm_object *object) | |
349 | { | |
350 | struct gf100_gr_chan *chan = gf100_gr_chan(object); | |
351 | int i; | |
352 | ||
353 | for (i = 0; i < ARRAY_SIZE(chan->data); i++) { | |
019e4d76 | 354 | nvkm_vmm_put(chan->vmm, &chan->data[i].vma); |
997a8900 | 355 | nvkm_memory_unref(&chan->data[i].mem); |
27f3d6cf BS |
356 | } |
357 | ||
019e4d76 | 358 | nvkm_vmm_put(chan->vmm, &chan->mmio_vma); |
997a8900 | 359 | nvkm_memory_unref(&chan->mmio); |
019e4d76 | 360 | nvkm_vmm_unref(&chan->vmm); |
27f3d6cf BS |
361 | return chan; |
362 | } | |
363 | ||
364 | static const struct nvkm_object_func | |
365 | gf100_gr_chan = { | |
366 | .dtor = gf100_gr_chan_dtor, | |
367 | .bind = gf100_gr_chan_bind, | |
368 | }; | |
369 | ||
370 | static int | |
371 | gf100_gr_chan_new(struct nvkm_gr *base, struct nvkm_fifo_chan *fifoch, | |
372 | const struct nvkm_oclass *oclass, | |
373 | struct nvkm_object **pobject) | |
374 | { | |
375 | struct gf100_gr *gr = gf100_gr(base); | |
bfee3f3d BS |
376 | struct gf100_gr_data *data = gr->mmio_data; |
377 | struct gf100_gr_mmio *mmio = gr->mmio_list; | |
e3c71eb2 | 378 | struct gf100_gr_chan *chan; |
019e4d76 | 379 | struct gf100_vmm_map_v0 args = { .priv = 1 }; |
227c95d9 | 380 | struct nvkm_device *device = gr->base.engine.subdev.device; |
966a5b7d | 381 | int ret, i; |
966a5b7d | 382 | |
27f3d6cf BS |
383 | if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) |
384 | return -ENOMEM; | |
385 | nvkm_object_ctor(&gf100_gr_chan, oclass, &chan->object); | |
386 | chan->gr = gr; | |
019e4d76 | 387 | chan->vmm = nvkm_vmm_ref(fifoch->vmm); |
27f3d6cf | 388 | *pobject = &chan->object; |
966a5b7d | 389 | |
ac1499d9 BS |
390 | /* allocate memory for a "mmio list" buffer that's used by the HUB |
391 | * fuc to modify some per-context register settings on first load | |
392 | * of the context. | |
393 | */ | |
227c95d9 BS |
394 | ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, 0x1000, 0x100, |
395 | false, &chan->mmio); | |
73a60c0d BS |
396 | if (ret) |
397 | return ret; | |
398 | ||
019e4d76 | 399 | ret = nvkm_vmm_get(fifoch->vmm, 12, 0x1000, &chan->mmio_vma); |
73a60c0d BS |
400 | if (ret) |
401 | return ret; | |
402 | ||
f66c57d9 | 403 | ret = nvkm_memory_map(chan->mmio, 0, fifoch->vmm, |
019e4d76 | 404 | chan->mmio_vma, &args, sizeof(args)); |
19a82e49 BS |
405 | if (ret) |
406 | return ret; | |
227c95d9 | 407 | |
ac1499d9 | 408 | /* allocate buffers referenced by mmio list */ |
bfee3f3d | 409 | for (i = 0; data->size && i < ARRAY_SIZE(gr->mmio_data); i++) { |
227c95d9 BS |
410 | ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, |
411 | data->size, data->align, false, | |
412 | &chan->data[i].mem); | |
ac1499d9 BS |
413 | if (ret) |
414 | return ret; | |
73a60c0d | 415 | |
019e4d76 BS |
416 | ret = nvkm_vmm_get(fifoch->vmm, 12, |
417 | nvkm_memory_size(chan->data[i].mem), | |
418 | &chan->data[i].vma); | |
ac1499d9 BS |
419 | if (ret) |
420 | return ret; | |
966a5b7d | 421 | |
019e4d76 BS |
422 | args.priv = data->priv; |
423 | ||
424 | ret = nvkm_memory_map(chan->data[i].mem, 0, chan->vmm, | |
425 | chan->data[i].vma, &args, sizeof(args)); | |
19a82e49 BS |
426 | if (ret) |
427 | return ret; | |
428 | ||
ac1499d9 | 429 | data++; |
966a5b7d BS |
430 | } |
431 | ||
ac1499d9 | 432 | /* finally, fill in the mmio list and point the context at it */ |
142ea05f | 433 | nvkm_kmap(chan->mmio); |
bfee3f3d | 434 | for (i = 0; mmio->addr && i < ARRAY_SIZE(gr->mmio_list); i++) { |
ac1499d9 BS |
435 | u32 addr = mmio->addr; |
436 | u32 data = mmio->data; | |
966a5b7d | 437 | |
694c6caf | 438 | if (mmio->buffer >= 0) { |
019e4d76 | 439 | u64 info = chan->data[mmio->buffer].vma->addr; |
ac1499d9 BS |
440 | data |= info >> mmio->shift; |
441 | } | |
73a60c0d | 442 | |
142ea05f BS |
443 | nvkm_wo32(chan->mmio, chan->mmio_nr++ * 4, addr); |
444 | nvkm_wo32(chan->mmio, chan->mmio_nr++ * 4, data); | |
ac1499d9 BS |
445 | mmio++; |
446 | } | |
142ea05f | 447 | nvkm_done(chan->mmio); |
ebb945a9 | 448 | return 0; |
4b223eef BS |
449 | } |
450 | ||
ebb945a9 | 451 | /******************************************************************************* |
c33b1e8c | 452 | * PGRAPH register lists |
ebb945a9 BS |
453 | ******************************************************************************/ |
454 | ||
e3c71eb2 BS |
455 | const struct gf100_gr_init |
456 | gf100_gr_init_main_0[] = { | |
30f4e087 BS |
457 | { 0x400080, 1, 0x04, 0x003083c2 }, |
458 | { 0x400088, 1, 0x04, 0x00006fe7 }, | |
459 | { 0x40008c, 1, 0x04, 0x00000000 }, | |
460 | { 0x400090, 1, 0x04, 0x00000030 }, | |
461 | { 0x40013c, 1, 0x04, 0x013901f7 }, | |
462 | { 0x400140, 1, 0x04, 0x00000100 }, | |
463 | { 0x400144, 1, 0x04, 0x00000000 }, | |
464 | { 0x400148, 1, 0x04, 0x00000110 }, | |
465 | { 0x400138, 1, 0x04, 0x00000000 }, | |
466 | { 0x400130, 2, 0x04, 0x00000000 }, | |
467 | { 0x400124, 1, 0x04, 0x00000002 }, | |
468 | {} | |
469 | }; | |
470 | ||
e3c71eb2 BS |
471 | const struct gf100_gr_init |
472 | gf100_gr_init_fe_0[] = { | |
30f4e087 BS |
473 | { 0x40415c, 1, 0x04, 0x00000000 }, |
474 | { 0x404170, 1, 0x04, 0x00000000 }, | |
475 | {} | |
476 | }; | |
477 | ||
e3c71eb2 BS |
478 | const struct gf100_gr_init |
479 | gf100_gr_init_pri_0[] = { | |
30f4e087 BS |
480 | { 0x404488, 2, 0x04, 0x00000000 }, |
481 | {} | |
482 | }; | |
483 | ||
e3c71eb2 BS |
484 | const struct gf100_gr_init |
485 | gf100_gr_init_rstr2d_0[] = { | |
30f4e087 BS |
486 | { 0x407808, 1, 0x04, 0x00000000 }, |
487 | {} | |
488 | }; | |
489 | ||
e3c71eb2 BS |
490 | const struct gf100_gr_init |
491 | gf100_gr_init_pd_0[] = { | |
30f4e087 BS |
492 | { 0x406024, 1, 0x04, 0x00000000 }, |
493 | {} | |
494 | }; | |
495 | ||
e3c71eb2 BS |
496 | const struct gf100_gr_init |
497 | gf100_gr_init_ds_0[] = { | |
30f4e087 BS |
498 | { 0x405844, 1, 0x04, 0x00ffffff }, |
499 | { 0x405850, 1, 0x04, 0x00000000 }, | |
500 | { 0x405908, 1, 0x04, 0x00000000 }, | |
501 | {} | |
502 | }; | |
503 | ||
e3c71eb2 BS |
504 | const struct gf100_gr_init |
505 | gf100_gr_init_scc_0[] = { | |
30f4e087 BS |
506 | { 0x40803c, 1, 0x04, 0x00000000 }, |
507 | {} | |
508 | }; | |
509 | ||
e3c71eb2 BS |
510 | const struct gf100_gr_init |
511 | gf100_gr_init_prop_0[] = { | |
30f4e087 | 512 | { 0x4184a0, 1, 0x04, 0x00000000 }, |
97af71fa BS |
513 | {} |
514 | }; | |
515 | ||
e3c71eb2 BS |
516 | const struct gf100_gr_init |
517 | gf100_gr_init_gpc_unk_0[] = { | |
30f4e087 BS |
518 | { 0x418604, 1, 0x04, 0x00000000 }, |
519 | { 0x418680, 1, 0x04, 0x00000000 }, | |
520 | { 0x418714, 1, 0x04, 0x80000000 }, | |
521 | { 0x418384, 1, 0x04, 0x00000000 }, | |
97af71fa BS |
522 | {} |
523 | }; | |
524 | ||
e3c71eb2 BS |
525 | const struct gf100_gr_init |
526 | gf100_gr_init_setup_0[] = { | |
30f4e087 | 527 | { 0x418814, 3, 0x04, 0x00000000 }, |
97af71fa BS |
528 | {} |
529 | }; | |
530 | ||
e3c71eb2 BS |
531 | const struct gf100_gr_init |
532 | gf100_gr_init_crstr_0[] = { | |
30f4e087 | 533 | { 0x418b04, 1, 0x04, 0x00000000 }, |
97af71fa BS |
534 | {} |
535 | }; | |
536 | ||
e3c71eb2 BS |
537 | const struct gf100_gr_init |
538 | gf100_gr_init_setup_1[] = { | |
30f4e087 BS |
539 | { 0x4188c8, 1, 0x04, 0x80000000 }, |
540 | { 0x4188cc, 1, 0x04, 0x00000000 }, | |
541 | { 0x4188d0, 1, 0x04, 0x00010000 }, | |
542 | { 0x4188d4, 1, 0x04, 0x00000001 }, | |
97af71fa BS |
543 | {} |
544 | }; | |
545 | ||
e3c71eb2 BS |
546 | const struct gf100_gr_init |
547 | gf100_gr_init_zcull_0[] = { | |
30f4e087 BS |
548 | { 0x418910, 1, 0x04, 0x00010001 }, |
549 | { 0x418914, 1, 0x04, 0x00000301 }, | |
550 | { 0x418918, 1, 0x04, 0x00800000 }, | |
551 | { 0x418980, 1, 0x04, 0x77777770 }, | |
552 | { 0x418984, 3, 0x04, 0x77777777 }, | |
97af71fa BS |
553 | {} |
554 | }; | |
555 | ||
e3c71eb2 BS |
556 | const struct gf100_gr_init |
557 | gf100_gr_init_gpm_0[] = { | |
30f4e087 BS |
558 | { 0x418c04, 1, 0x04, 0x00000000 }, |
559 | { 0x418c88, 1, 0x04, 0x00000000 }, | |
97af71fa BS |
560 | {} |
561 | }; | |
562 | ||
e3c71eb2 BS |
563 | const struct gf100_gr_init |
564 | gf100_gr_init_gpc_unk_1[] = { | |
30f4e087 BS |
565 | { 0x418d00, 1, 0x04, 0x00000000 }, |
566 | { 0x418f08, 1, 0x04, 0x00000000 }, | |
567 | { 0x418e00, 1, 0x04, 0x00000050 }, | |
568 | { 0x418e08, 1, 0x04, 0x00000000 }, | |
97af71fa BS |
569 | {} |
570 | }; | |
571 | ||
e3c71eb2 BS |
572 | const struct gf100_gr_init |
573 | gf100_gr_init_gcc_0[] = { | |
30f4e087 BS |
574 | { 0x41900c, 1, 0x04, 0x00000000 }, |
575 | { 0x419018, 1, 0x04, 0x00000000 }, | |
576 | {} | |
577 | }; | |
578 | ||
e3c71eb2 BS |
579 | const struct gf100_gr_init |
580 | gf100_gr_init_tpccs_0[] = { | |
30f4e087 BS |
581 | { 0x419d08, 2, 0x04, 0x00000000 }, |
582 | { 0x419d10, 1, 0x04, 0x00000014 }, | |
7e194533 BS |
583 | {} |
584 | }; | |
585 | ||
e3c71eb2 BS |
586 | const struct gf100_gr_init |
587 | gf100_gr_init_tex_0[] = { | |
30f4e087 BS |
588 | { 0x419ab0, 1, 0x04, 0x00000000 }, |
589 | { 0x419ab8, 1, 0x04, 0x000000e7 }, | |
590 | { 0x419abc, 2, 0x04, 0x00000000 }, | |
7e194533 BS |
591 | {} |
592 | }; | |
593 | ||
e3c71eb2 BS |
594 | const struct gf100_gr_init |
595 | gf100_gr_init_pe_0[] = { | |
30f4e087 BS |
596 | { 0x41980c, 3, 0x04, 0x00000000 }, |
597 | { 0x419844, 1, 0x04, 0x00000000 }, | |
598 | { 0x41984c, 1, 0x04, 0x00005bc5 }, | |
599 | { 0x419850, 4, 0x04, 0x00000000 }, | |
7e194533 BS |
600 | {} |
601 | }; | |
602 | ||
e3c71eb2 BS |
603 | const struct gf100_gr_init |
604 | gf100_gr_init_l1c_0[] = { | |
30f4e087 BS |
605 | { 0x419c98, 1, 0x04, 0x00000000 }, |
606 | { 0x419ca8, 1, 0x04, 0x80000000 }, | |
607 | { 0x419cb4, 1, 0x04, 0x00000000 }, | |
608 | { 0x419cb8, 1, 0x04, 0x00008bf4 }, | |
609 | { 0x419cbc, 1, 0x04, 0x28137606 }, | |
610 | { 0x419cc0, 2, 0x04, 0x00000000 }, | |
7e194533 BS |
611 | {} |
612 | }; | |
613 | ||
e3c71eb2 BS |
614 | const struct gf100_gr_init |
615 | gf100_gr_init_wwdx_0[] = { | |
30f4e087 BS |
616 | { 0x419bd4, 1, 0x04, 0x00800000 }, |
617 | { 0x419bdc, 1, 0x04, 0x00000000 }, | |
7e194533 BS |
618 | {} |
619 | }; | |
620 | ||
e3c71eb2 BS |
621 | const struct gf100_gr_init |
622 | gf100_gr_init_tpccs_1[] = { | |
30f4e087 | 623 | { 0x419d2c, 1, 0x04, 0x00000000 }, |
7e194533 BS |
624 | {} |
625 | }; | |
626 | ||
e3c71eb2 BS |
627 | const struct gf100_gr_init |
628 | gf100_gr_init_mpc_0[] = { | |
30f4e087 | 629 | { 0x419c0c, 1, 0x04, 0x00000000 }, |
7e194533 BS |
630 | {} |
631 | }; | |
632 | ||
e3c71eb2 BS |
633 | static const struct gf100_gr_init |
634 | gf100_gr_init_sm_0[] = { | |
30f4e087 BS |
635 | { 0x419e00, 1, 0x04, 0x00000000 }, |
636 | { 0x419ea0, 1, 0x04, 0x00000000 }, | |
637 | { 0x419ea4, 1, 0x04, 0x00000100 }, | |
638 | { 0x419ea8, 1, 0x04, 0x00001100 }, | |
639 | { 0x419eac, 1, 0x04, 0x11100702 }, | |
640 | { 0x419eb0, 1, 0x04, 0x00000003 }, | |
641 | { 0x419eb4, 4, 0x04, 0x00000000 }, | |
642 | { 0x419ec8, 1, 0x04, 0x06060618 }, | |
643 | { 0x419ed0, 1, 0x04, 0x0eff0e38 }, | |
644 | { 0x419ed4, 1, 0x04, 0x011104f1 }, | |
645 | { 0x419edc, 1, 0x04, 0x00000000 }, | |
646 | { 0x419f00, 1, 0x04, 0x00000000 }, | |
647 | { 0x419f2c, 1, 0x04, 0x00000000 }, | |
648 | {} | |
649 | }; | |
650 | ||
e3c71eb2 BS |
651 | const struct gf100_gr_init |
652 | gf100_gr_init_be_0[] = { | |
30f4e087 BS |
653 | { 0x40880c, 1, 0x04, 0x00000000 }, |
654 | { 0x408910, 9, 0x04, 0x00000000 }, | |
655 | { 0x408950, 1, 0x04, 0x00000000 }, | |
656 | { 0x408954, 1, 0x04, 0x0000ffff }, | |
657 | { 0x408984, 1, 0x04, 0x00000000 }, | |
658 | { 0x408988, 1, 0x04, 0x08040201 }, | |
659 | { 0x40898c, 1, 0x04, 0x80402010 }, | |
660 | {} | |
661 | }; | |
662 | ||
e3c71eb2 BS |
663 | const struct gf100_gr_init |
664 | gf100_gr_init_fe_1[] = { | |
c33b1e8c BS |
665 | { 0x4040f0, 1, 0x04, 0x00000000 }, |
666 | {} | |
667 | }; | |
668 | ||
e3c71eb2 BS |
669 | const struct gf100_gr_init |
670 | gf100_gr_init_pe_1[] = { | |
c33b1e8c BS |
671 | { 0x419880, 1, 0x04, 0x00000002 }, |
672 | {} | |
673 | }; | |
674 | ||
e3c71eb2 BS |
675 | static const struct gf100_gr_pack |
676 | gf100_gr_pack_mmio[] = { | |
677 | { gf100_gr_init_main_0 }, | |
678 | { gf100_gr_init_fe_0 }, | |
679 | { gf100_gr_init_pri_0 }, | |
680 | { gf100_gr_init_rstr2d_0 }, | |
681 | { gf100_gr_init_pd_0 }, | |
682 | { gf100_gr_init_ds_0 }, | |
683 | { gf100_gr_init_scc_0 }, | |
684 | { gf100_gr_init_prop_0 }, | |
685 | { gf100_gr_init_gpc_unk_0 }, | |
686 | { gf100_gr_init_setup_0 }, | |
687 | { gf100_gr_init_crstr_0 }, | |
688 | { gf100_gr_init_setup_1 }, | |
689 | { gf100_gr_init_zcull_0 }, | |
690 | { gf100_gr_init_gpm_0 }, | |
691 | { gf100_gr_init_gpc_unk_1 }, | |
692 | { gf100_gr_init_gcc_0 }, | |
693 | { gf100_gr_init_tpccs_0 }, | |
694 | { gf100_gr_init_tex_0 }, | |
695 | { gf100_gr_init_pe_0 }, | |
696 | { gf100_gr_init_l1c_0 }, | |
697 | { gf100_gr_init_wwdx_0 }, | |
698 | { gf100_gr_init_tpccs_1 }, | |
699 | { gf100_gr_init_mpc_0 }, | |
700 | { gf100_gr_init_sm_0 }, | |
701 | { gf100_gr_init_be_0 }, | |
702 | { gf100_gr_init_fe_1 }, | |
703 | { gf100_gr_init_pe_1 }, | |
26410c67 ML |
704 | {} |
705 | }; | |
706 | ||
c33b1e8c BS |
707 | /******************************************************************************* |
708 | * PGRAPH engine/subdev functions | |
709 | ******************************************************************************/ | |
710 | ||
6845c313 BS |
711 | static bool |
712 | gf100_gr_chsw_load(struct nvkm_gr *base) | |
713 | { | |
714 | struct gf100_gr *gr = gf100_gr(base); | |
715 | if (!gr->firmware) { | |
716 | u32 trace = nvkm_rd32(gr->base.engine.subdev.device, 0x40981c); | |
717 | if (trace & 0x00000040) | |
718 | return true; | |
719 | } else { | |
720 | u32 mthd = nvkm_rd32(gr->base.engine.subdev.device, 0x409808); | |
721 | if (mthd & 0x00080000) | |
722 | return true; | |
723 | } | |
724 | return false; | |
725 | } | |
726 | ||
64cb5a31 BS |
727 | int |
728 | gf100_gr_rops(struct gf100_gr *gr) | |
729 | { | |
730 | struct nvkm_device *device = gr->base.engine.subdev.device; | |
731 | return (nvkm_rd32(device, 0x409604) & 0x001f0000) >> 16; | |
732 | } | |
733 | ||
ac9738bb | 734 | void |
bfee3f3d | 735 | gf100_gr_zbc_init(struct gf100_gr *gr) |
ac9738bb BS |
736 | { |
737 | const u32 zero[] = { 0x00000000, 0x00000000, 0x00000000, 0x00000000, | |
738 | 0x00000000, 0x00000000, 0x00000000, 0x00000000 }; | |
739 | const u32 one[] = { 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, | |
740 | 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }; | |
741 | const u32 f32_0[] = { 0x00000000, 0x00000000, 0x00000000, 0x00000000, | |
742 | 0x00000000, 0x00000000, 0x00000000, 0x00000000 }; | |
743 | const u32 f32_1[] = { 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, | |
744 | 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000 }; | |
70bc7182 | 745 | struct nvkm_ltc *ltc = gr->base.engine.subdev.device->ltc; |
ac9738bb BS |
746 | int index; |
747 | ||
bfee3f3d BS |
748 | if (!gr->zbc_color[0].format) { |
749 | gf100_gr_zbc_color_get(gr, 1, & zero[0], &zero[4]); | |
750 | gf100_gr_zbc_color_get(gr, 2, & one[0], &one[4]); | |
751 | gf100_gr_zbc_color_get(gr, 4, &f32_0[0], &f32_0[4]); | |
752 | gf100_gr_zbc_color_get(gr, 4, &f32_1[0], &f32_1[4]); | |
753 | gf100_gr_zbc_depth_get(gr, 1, 0x00000000, 0x00000000); | |
754 | gf100_gr_zbc_depth_get(gr, 1, 0x3f800000, 0x3f800000); | |
ac9738bb BS |
755 | } |
756 | ||
757 | for (index = ltc->zbc_min; index <= ltc->zbc_max; index++) | |
bfee3f3d | 758 | gf100_gr_zbc_clear_color(gr, index); |
ac9738bb | 759 | for (index = ltc->zbc_min; index <= ltc->zbc_max; index++) |
bfee3f3d | 760 | gf100_gr_zbc_clear_depth(gr, index); |
ac9738bb BS |
761 | } |
762 | ||
4a8cf451 AC |
763 | /** |
764 | * Wait until GR goes idle. GR is considered idle if it is disabled by the | |
765 | * MC (0x200) register, or GR is not busy and a context switch is not in | |
766 | * progress. | |
767 | */ | |
768 | int | |
bfee3f3d | 769 | gf100_gr_wait_idle(struct gf100_gr *gr) |
4a8cf451 | 770 | { |
109c2f2f BS |
771 | struct nvkm_subdev *subdev = &gr->base.engine.subdev; |
772 | struct nvkm_device *device = subdev->device; | |
4a8cf451 AC |
773 | unsigned long end_jiffies = jiffies + msecs_to_jiffies(2000); |
774 | bool gr_enabled, ctxsw_active, gr_busy; | |
775 | ||
776 | do { | |
777 | /* | |
778 | * required to make sure FIFO_ENGINE_STATUS (0x2640) is | |
779 | * up-to-date | |
780 | */ | |
276836d4 | 781 | nvkm_rd32(device, 0x400700); |
4a8cf451 | 782 | |
276836d4 BS |
783 | gr_enabled = nvkm_rd32(device, 0x200) & 0x1000; |
784 | ctxsw_active = nvkm_rd32(device, 0x2640) & 0x8000; | |
785 | gr_busy = nvkm_rd32(device, 0x40060c) & 0x1; | |
4a8cf451 AC |
786 | |
787 | if (!gr_enabled || (!gr_busy && !ctxsw_active)) | |
788 | return 0; | |
789 | } while (time_before(jiffies, end_jiffies)); | |
790 | ||
109c2f2f BS |
791 | nvkm_error(subdev, |
792 | "wait for idle timeout (en: %d, ctxsw: %d, busy: %d)\n", | |
793 | gr_enabled, ctxsw_active, gr_busy); | |
4a8cf451 AC |
794 | return -EAGAIN; |
795 | } | |
796 | ||
30f4e087 | 797 | void |
bfee3f3d | 798 | gf100_gr_mmio(struct gf100_gr *gr, const struct gf100_gr_pack *p) |
4b223eef | 799 | { |
276836d4 | 800 | struct nvkm_device *device = gr->base.engine.subdev.device; |
e3c71eb2 BS |
801 | const struct gf100_gr_pack *pack; |
802 | const struct gf100_gr_init *init; | |
c33b1e8c BS |
803 | |
804 | pack_for_each_init(init, pack, p) { | |
805 | u32 next = init->addr + init->count * init->pitch; | |
806 | u32 addr = init->addr; | |
807 | while (addr < next) { | |
276836d4 | 808 | nvkm_wr32(device, addr, init->data); |
30f4e087 BS |
809 | addr += init->pitch; |
810 | } | |
811 | } | |
ebb945a9 BS |
812 | } |
813 | ||
814 | void | |
bfee3f3d | 815 | gf100_gr_icmd(struct gf100_gr *gr, const struct gf100_gr_pack *p) |
ebb945a9 | 816 | { |
276836d4 | 817 | struct nvkm_device *device = gr->base.engine.subdev.device; |
e3c71eb2 BS |
818 | const struct gf100_gr_pack *pack; |
819 | const struct gf100_gr_init *init; | |
c33b1e8c | 820 | u32 data = 0; |
30f4e087 | 821 | |
276836d4 | 822 | nvkm_wr32(device, 0x400208, 0x80000000); |
c33b1e8c BS |
823 | |
824 | pack_for_each_init(init, pack, p) { | |
825 | u32 next = init->addr + init->count * init->pitch; | |
826 | u32 addr = init->addr; | |
827 | ||
828 | if ((pack == p && init == p->init) || data != init->data) { | |
276836d4 | 829 | nvkm_wr32(device, 0x400204, init->data); |
30f4e087 BS |
830 | data = init->data; |
831 | } | |
ebb945a9 | 832 | |
c33b1e8c | 833 | while (addr < next) { |
276836d4 | 834 | nvkm_wr32(device, 0x400200, addr); |
4a8cf451 AC |
835 | /** |
836 | * Wait for GR to go idle after submitting a | |
837 | * GO_IDLE bundle | |
838 | */ | |
839 | if ((addr & 0xffff) == 0xe100) | |
bfee3f3d | 840 | gf100_gr_wait_idle(gr); |
c4584adc BS |
841 | nvkm_msec(device, 2000, |
842 | if (!(nvkm_rd32(device, 0x400700) & 0x00000004)) | |
843 | break; | |
844 | ); | |
30f4e087 | 845 | addr += init->pitch; |
30f4e087 BS |
846 | } |
847 | } | |
c33b1e8c | 848 | |
276836d4 | 849 | nvkm_wr32(device, 0x400208, 0x00000000); |
30f4e087 BS |
850 | } |
851 | ||
852 | void | |
bfee3f3d | 853 | gf100_gr_mthd(struct gf100_gr *gr, const struct gf100_gr_pack *p) |
30f4e087 | 854 | { |
276836d4 | 855 | struct nvkm_device *device = gr->base.engine.subdev.device; |
e3c71eb2 BS |
856 | const struct gf100_gr_pack *pack; |
857 | const struct gf100_gr_init *init; | |
c33b1e8c | 858 | u32 data = 0; |
30f4e087 | 859 | |
c33b1e8c BS |
860 | pack_for_each_init(init, pack, p) { |
861 | u32 ctrl = 0x80000000 | pack->type; | |
862 | u32 next = init->addr + init->count * init->pitch; | |
863 | u32 addr = init->addr; | |
864 | ||
865 | if ((pack == p && init == p->init) || data != init->data) { | |
276836d4 | 866 | nvkm_wr32(device, 0x40448c, init->data); |
c33b1e8c BS |
867 | data = init->data; |
868 | } | |
869 | ||
870 | while (addr < next) { | |
276836d4 | 871 | nvkm_wr32(device, 0x404488, ctrl | (addr << 14)); |
c33b1e8c | 872 | addr += init->pitch; |
30f4e087 BS |
873 | } |
874 | } | |
875 | } | |
876 | ||
877 | u64 | |
c85ee6ca | 878 | gf100_gr_units(struct nvkm_gr *base) |
30f4e087 | 879 | { |
c85ee6ca | 880 | struct gf100_gr *gr = gf100_gr(base); |
30f4e087 BS |
881 | u64 cfg; |
882 | ||
bfee3f3d BS |
883 | cfg = (u32)gr->gpc_nr; |
884 | cfg |= (u32)gr->tpc_total << 8; | |
885 | cfg |= (u64)gr->rop_nr << 32; | |
30f4e087 BS |
886 | |
887 | return cfg; | |
ebb945a9 BS |
888 | } |
889 | ||
caf2be8a IM |
890 | static const struct nvkm_bitfield gf100_dispatch_error[] = { |
891 | { 0x00000001, "INJECTED_BUNDLE_ERROR" }, | |
892 | { 0x00000002, "CLASS_SUBCH_MISMATCH" }, | |
893 | { 0x00000004, "SUBCHSW_DURING_NOTIFY" }, | |
894 | {} | |
895 | }; | |
896 | ||
897 | static const struct nvkm_bitfield gf100_m2mf_error[] = { | |
898 | { 0x00000001, "PUSH_TOO_MUCH_DATA" }, | |
899 | { 0x00000002, "PUSH_NOT_ENOUGH_DATA" }, | |
900 | {} | |
901 | }; | |
902 | ||
903 | static const struct nvkm_bitfield gf100_unk6_error[] = { | |
904 | { 0x00000001, "TEMP_TOO_SMALL" }, | |
905 | {} | |
906 | }; | |
907 | ||
908 | static const struct nvkm_bitfield gf100_ccache_error[] = { | |
909 | { 0x00000001, "INTR" }, | |
910 | { 0x00000002, "LDCONST_OOB" }, | |
911 | {} | |
912 | }; | |
913 | ||
914 | static const struct nvkm_bitfield gf100_macro_error[] = { | |
915 | { 0x00000001, "TOO_FEW_PARAMS" }, | |
916 | { 0x00000002, "TOO_MANY_PARAMS" }, | |
917 | { 0x00000004, "ILLEGAL_OPCODE" }, | |
918 | { 0x00000008, "DOUBLE_BRANCH" }, | |
919 | { 0x00000010, "WATCHDOG" }, | |
920 | {} | |
921 | }; | |
922 | ||
109c2f2f | 923 | static const struct nvkm_bitfield gk104_sked_error[] = { |
caf2be8a | 924 | { 0x00000040, "CTA_RESUME" }, |
109c2f2f BS |
925 | { 0x00000080, "CONSTANT_BUFFER_SIZE" }, |
926 | { 0x00000200, "LOCAL_MEMORY_SIZE_POS" }, | |
927 | { 0x00000400, "LOCAL_MEMORY_SIZE_NEG" }, | |
928 | { 0x00000800, "WARP_CSTACK_SIZE" }, | |
929 | { 0x00001000, "TOTAL_TEMP_SIZE" }, | |
930 | { 0x00002000, "REGISTER_COUNT" }, | |
931 | { 0x00040000, "TOTAL_THREADS" }, | |
932 | { 0x00100000, "PROGRAM_OFFSET" }, | |
933 | { 0x00200000, "SHARED_MEMORY_SIZE" }, | |
caf2be8a IM |
934 | { 0x00800000, "CTA_THREAD_DIMENSION_ZERO" }, |
935 | { 0x01000000, "MEMORY_WINDOW_OVERLAP" }, | |
109c2f2f BS |
936 | { 0x02000000, "SHARED_CONFIG_TOO_SMALL" }, |
937 | { 0x04000000, "TOTAL_REGISTER_COUNT" }, | |
30f4e087 BS |
938 | {} |
939 | }; | |
940 | ||
109c2f2f BS |
941 | static const struct nvkm_bitfield gf100_gpc_rop_error[] = { |
942 | { 0x00000002, "RT_PITCH_OVERRUN" }, | |
943 | { 0x00000010, "RT_WIDTH_OVERRUN" }, | |
944 | { 0x00000020, "RT_HEIGHT_OVERRUN" }, | |
945 | { 0x00000080, "ZETA_STORAGE_TYPE_MISMATCH" }, | |
946 | { 0x00000100, "RT_STORAGE_TYPE_MISMATCH" }, | |
947 | { 0x00000400, "RT_LINEAR_MISMATCH" }, | |
30f4e087 BS |
948 | {} |
949 | }; | |
950 | ||
ebb945a9 | 951 | static void |
bfee3f3d | 952 | gf100_gr_trap_gpc_rop(struct gf100_gr *gr, int gpc) |
ebb945a9 | 953 | { |
109c2f2f BS |
954 | struct nvkm_subdev *subdev = &gr->base.engine.subdev; |
955 | struct nvkm_device *device = subdev->device; | |
956 | char error[128]; | |
30f4e087 | 957 | u32 trap[4]; |
ebb945a9 | 958 | |
109c2f2f | 959 | trap[0] = nvkm_rd32(device, GPC_UNIT(gpc, 0x0420)) & 0x3fffffff; |
276836d4 BS |
960 | trap[1] = nvkm_rd32(device, GPC_UNIT(gpc, 0x0434)); |
961 | trap[2] = nvkm_rd32(device, GPC_UNIT(gpc, 0x0438)); | |
962 | trap[3] = nvkm_rd32(device, GPC_UNIT(gpc, 0x043c)); | |
ebb945a9 | 963 | |
109c2f2f | 964 | nvkm_snprintbf(error, sizeof(error), gf100_gpc_rop_error, trap[0]); |
30f4e087 | 965 | |
109c2f2f BS |
966 | nvkm_error(subdev, "GPC%d/PROP trap: %08x [%s] x = %u, y = %u, " |
967 | "format = %x, storage type = %x\n", | |
968 | gpc, trap[0], error, trap[1] & 0xffff, trap[1] >> 16, | |
969 | (trap[2] >> 8) & 0x3f, trap[3] & 0xff); | |
276836d4 | 970 | nvkm_wr32(device, GPC_UNIT(gpc, 0x0420), 0xc0000000); |
ebb945a9 BS |
971 | } |
972 | ||
e3c71eb2 | 973 | static const struct nvkm_enum gf100_mp_warp_error[] = { |
383d0a41 BS |
974 | { 0x01, "STACK_ERROR" }, |
975 | { 0x02, "API_STACK_ERROR" }, | |
976 | { 0x03, "RET_EMPTY_STACK_ERROR" }, | |
977 | { 0x04, "PC_WRAP" }, | |
fec43a72 | 978 | { 0x05, "MISALIGNED_PC" }, |
383d0a41 BS |
979 | { 0x06, "PC_OVERFLOW" }, |
980 | { 0x07, "MISALIGNED_IMMC_ADDR" }, | |
981 | { 0x08, "MISALIGNED_REG" }, | |
982 | { 0x09, "ILLEGAL_INSTR_ENCODING" }, | |
983 | { 0x0a, "ILLEGAL_SPH_INSTR_COMBO" }, | |
984 | { 0x0b, "ILLEGAL_INSTR_PARAM" }, | |
985 | { 0x0c, "INVALID_CONST_ADDR" }, | |
986 | { 0x0d, "OOR_REG" }, | |
987 | { 0x0e, "OOR_ADDR" }, | |
988 | { 0x0f, "MISALIGNED_ADDR" }, | |
3988f645 | 989 | { 0x10, "INVALID_ADDR_SPACE" }, |
383d0a41 BS |
990 | { 0x11, "ILLEGAL_INSTR_PARAM2" }, |
991 | { 0x12, "INVALID_CONST_ADDR_LDC" }, | |
992 | { 0x13, "GEOMETRY_SM_ERROR" }, | |
993 | { 0x14, "DIVERGENT" }, | |
994 | { 0x15, "WARP_EXIT" }, | |
fec43a72 BS |
995 | {} |
996 | }; | |
997 | ||
e3c71eb2 | 998 | static const struct nvkm_bitfield gf100_mp_global_error[] = { |
383d0a41 BS |
999 | { 0x00000001, "SM_TO_SM_FAULT" }, |
1000 | { 0x00000002, "L1_ERROR" }, | |
fec43a72 | 1001 | { 0x00000004, "MULTIPLE_WARP_ERRORS" }, |
383d0a41 BS |
1002 | { 0x00000008, "PHYSICAL_STACK_OVERFLOW" }, |
1003 | { 0x00000010, "BPT_INT" }, | |
1004 | { 0x00000020, "BPT_PAUSE" }, | |
1005 | { 0x00000040, "SINGLE_STEP_COMPLETE" }, | |
1006 | { 0x20000000, "ECC_SEC_ERROR" }, | |
1007 | { 0x40000000, "ECC_DED_ERROR" }, | |
1008 | { 0x80000000, "TIMEOUT" }, | |
fec43a72 BS |
1009 | {} |
1010 | }; | |
1011 | ||
1012 | static void | |
bfee3f3d | 1013 | gf100_gr_trap_mp(struct gf100_gr *gr, int gpc, int tpc) |
fec43a72 | 1014 | { |
109c2f2f BS |
1015 | struct nvkm_subdev *subdev = &gr->base.engine.subdev; |
1016 | struct nvkm_device *device = subdev->device; | |
276836d4 BS |
1017 | u32 werr = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x648)); |
1018 | u32 gerr = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x650)); | |
109c2f2f BS |
1019 | const struct nvkm_enum *warp; |
1020 | char glob[128]; | |
fec43a72 | 1021 | |
109c2f2f BS |
1022 | nvkm_snprintbf(glob, sizeof(glob), gf100_mp_global_error, gerr); |
1023 | warp = nvkm_enum_find(gf100_mp_warp_error, werr & 0xffff); | |
1024 | ||
1025 | nvkm_error(subdev, "GPC%i/TPC%i/MP trap: " | |
1026 | "global %08x [%s] warp %04x [%s]\n", | |
1027 | gpc, tpc, gerr, glob, werr, warp ? warp->name : ""); | |
fec43a72 | 1028 | |
276836d4 BS |
1029 | nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x648), 0x00000000); |
1030 | nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x650), gerr); | |
fec43a72 BS |
1031 | } |
1032 | ||
f73221e4 | 1033 | static void |
bfee3f3d | 1034 | gf100_gr_trap_tpc(struct gf100_gr *gr, int gpc, int tpc) |
f73221e4 | 1035 | { |
109c2f2f BS |
1036 | struct nvkm_subdev *subdev = &gr->base.engine.subdev; |
1037 | struct nvkm_device *device = subdev->device; | |
276836d4 | 1038 | u32 stat = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x0508)); |
f73221e4 BS |
1039 | |
1040 | if (stat & 0x00000001) { | |
276836d4 | 1041 | u32 trap = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x0224)); |
109c2f2f | 1042 | nvkm_error(subdev, "GPC%d/TPC%d/TEX: %08x\n", gpc, tpc, trap); |
276836d4 | 1043 | nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x0224), 0xc0000000); |
f73221e4 BS |
1044 | stat &= ~0x00000001; |
1045 | } | |
1046 | ||
1047 | if (stat & 0x00000002) { | |
bfee3f3d | 1048 | gf100_gr_trap_mp(gr, gpc, tpc); |
f73221e4 BS |
1049 | stat &= ~0x00000002; |
1050 | } | |
1051 | ||
1052 | if (stat & 0x00000004) { | |
276836d4 | 1053 | u32 trap = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x0084)); |
109c2f2f | 1054 | nvkm_error(subdev, "GPC%d/TPC%d/POLY: %08x\n", gpc, tpc, trap); |
276836d4 | 1055 | nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x0084), 0xc0000000); |
f73221e4 BS |
1056 | stat &= ~0x00000004; |
1057 | } | |
1058 | ||
1059 | if (stat & 0x00000008) { | |
276836d4 | 1060 | u32 trap = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x048c)); |
109c2f2f | 1061 | nvkm_error(subdev, "GPC%d/TPC%d/L1C: %08x\n", gpc, tpc, trap); |
276836d4 | 1062 | nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x048c), 0xc0000000); |
f73221e4 BS |
1063 | stat &= ~0x00000008; |
1064 | } | |
1065 | ||
64373e4b BS |
1066 | if (stat & 0x00000010) { |
1067 | u32 trap = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x0430)); | |
1068 | nvkm_error(subdev, "GPC%d/TPC%d/MPC: %08x\n", gpc, tpc, trap); | |
1069 | nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x0430), 0xc0000000); | |
1070 | stat &= ~0x00000010; | |
1071 | } | |
1072 | ||
f73221e4 | 1073 | if (stat) { |
109c2f2f | 1074 | nvkm_error(subdev, "GPC%d/TPC%d/%08x: unknown\n", gpc, tpc, stat); |
f73221e4 BS |
1075 | } |
1076 | } | |
1077 | ||
1078 | static void | |
bfee3f3d | 1079 | gf100_gr_trap_gpc(struct gf100_gr *gr, int gpc) |
f73221e4 | 1080 | { |
109c2f2f BS |
1081 | struct nvkm_subdev *subdev = &gr->base.engine.subdev; |
1082 | struct nvkm_device *device = subdev->device; | |
276836d4 | 1083 | u32 stat = nvkm_rd32(device, GPC_UNIT(gpc, 0x2c90)); |
f73221e4 BS |
1084 | int tpc; |
1085 | ||
1086 | if (stat & 0x00000001) { | |
bfee3f3d | 1087 | gf100_gr_trap_gpc_rop(gr, gpc); |
f73221e4 BS |
1088 | stat &= ~0x00000001; |
1089 | } | |
1090 | ||
1091 | if (stat & 0x00000002) { | |
276836d4 | 1092 | u32 trap = nvkm_rd32(device, GPC_UNIT(gpc, 0x0900)); |
109c2f2f | 1093 | nvkm_error(subdev, "GPC%d/ZCULL: %08x\n", gpc, trap); |
276836d4 | 1094 | nvkm_wr32(device, GPC_UNIT(gpc, 0x0900), 0xc0000000); |
f73221e4 BS |
1095 | stat &= ~0x00000002; |
1096 | } | |
1097 | ||
1098 | if (stat & 0x00000004) { | |
276836d4 | 1099 | u32 trap = nvkm_rd32(device, GPC_UNIT(gpc, 0x1028)); |
109c2f2f | 1100 | nvkm_error(subdev, "GPC%d/CCACHE: %08x\n", gpc, trap); |
276836d4 | 1101 | nvkm_wr32(device, GPC_UNIT(gpc, 0x1028), 0xc0000000); |
f73221e4 BS |
1102 | stat &= ~0x00000004; |
1103 | } | |
1104 | ||
1105 | if (stat & 0x00000008) { | |
276836d4 | 1106 | u32 trap = nvkm_rd32(device, GPC_UNIT(gpc, 0x0824)); |
109c2f2f | 1107 | nvkm_error(subdev, "GPC%d/ESETUP: %08x\n", gpc, trap); |
276836d4 | 1108 | nvkm_wr32(device, GPC_UNIT(gpc, 0x0824), 0xc0000000); |
f73221e4 BS |
1109 | stat &= ~0x00000009; |
1110 | } | |
1111 | ||
bfee3f3d | 1112 | for (tpc = 0; tpc < gr->tpc_nr[gpc]; tpc++) { |
f73221e4 BS |
1113 | u32 mask = 0x00010000 << tpc; |
1114 | if (stat & mask) { | |
bfee3f3d | 1115 | gf100_gr_trap_tpc(gr, gpc, tpc); |
276836d4 | 1116 | nvkm_wr32(device, GPC_UNIT(gpc, 0x2c90), mask); |
f73221e4 BS |
1117 | stat &= ~mask; |
1118 | } | |
1119 | } | |
1120 | ||
1121 | if (stat) { | |
109c2f2f | 1122 | nvkm_error(subdev, "GPC%d/%08x: unknown\n", gpc, stat); |
f73221e4 BS |
1123 | } |
1124 | } | |
1125 | ||
1126 | static void | |
bfee3f3d | 1127 | gf100_gr_trap_intr(struct gf100_gr *gr) |
f73221e4 | 1128 | { |
109c2f2f BS |
1129 | struct nvkm_subdev *subdev = &gr->base.engine.subdev; |
1130 | struct nvkm_device *device = subdev->device; | |
caf2be8a | 1131 | char error[128]; |
276836d4 | 1132 | u32 trap = nvkm_rd32(device, 0x400108); |
109c2f2f | 1133 | int rop, gpc; |
f73221e4 BS |
1134 | |
1135 | if (trap & 0x00000001) { | |
276836d4 | 1136 | u32 stat = nvkm_rd32(device, 0x404000); |
caf2be8a IM |
1137 | |
1138 | nvkm_snprintbf(error, sizeof(error), gf100_dispatch_error, | |
1139 | stat & 0x3fffffff); | |
1140 | nvkm_error(subdev, "DISPATCH %08x [%s]\n", stat, error); | |
276836d4 BS |
1141 | nvkm_wr32(device, 0x404000, 0xc0000000); |
1142 | nvkm_wr32(device, 0x400108, 0x00000001); | |
f73221e4 BS |
1143 | trap &= ~0x00000001; |
1144 | } | |
1145 | ||
1146 | if (trap & 0x00000002) { | |
276836d4 | 1147 | u32 stat = nvkm_rd32(device, 0x404600); |
caf2be8a IM |
1148 | |
1149 | nvkm_snprintbf(error, sizeof(error), gf100_m2mf_error, | |
1150 | stat & 0x3fffffff); | |
1151 | nvkm_error(subdev, "M2MF %08x [%s]\n", stat, error); | |
1152 | ||
276836d4 BS |
1153 | nvkm_wr32(device, 0x404600, 0xc0000000); |
1154 | nvkm_wr32(device, 0x400108, 0x00000002); | |
f73221e4 BS |
1155 | trap &= ~0x00000002; |
1156 | } | |
1157 | ||
1158 | if (trap & 0x00000008) { | |
276836d4 | 1159 | u32 stat = nvkm_rd32(device, 0x408030); |
caf2be8a | 1160 | |
1894054d | 1161 | nvkm_snprintbf(error, sizeof(error), gf100_ccache_error, |
caf2be8a IM |
1162 | stat & 0x3fffffff); |
1163 | nvkm_error(subdev, "CCACHE %08x [%s]\n", stat, error); | |
276836d4 BS |
1164 | nvkm_wr32(device, 0x408030, 0xc0000000); |
1165 | nvkm_wr32(device, 0x400108, 0x00000008); | |
f73221e4 BS |
1166 | trap &= ~0x00000008; |
1167 | } | |
1168 | ||
1169 | if (trap & 0x00000010) { | |
276836d4 | 1170 | u32 stat = nvkm_rd32(device, 0x405840); |
caf2be8a IM |
1171 | nvkm_error(subdev, "SHADER %08x, sph: 0x%06x, stage: 0x%02x\n", |
1172 | stat, stat & 0xffffff, (stat >> 24) & 0x3f); | |
276836d4 BS |
1173 | nvkm_wr32(device, 0x405840, 0xc0000000); |
1174 | nvkm_wr32(device, 0x400108, 0x00000010); | |
f73221e4 BS |
1175 | trap &= ~0x00000010; |
1176 | } | |
1177 | ||
1178 | if (trap & 0x00000040) { | |
276836d4 | 1179 | u32 stat = nvkm_rd32(device, 0x40601c); |
caf2be8a IM |
1180 | |
1181 | nvkm_snprintbf(error, sizeof(error), gf100_unk6_error, | |
1182 | stat & 0x3fffffff); | |
1183 | nvkm_error(subdev, "UNK6 %08x [%s]\n", stat, error); | |
1184 | ||
276836d4 BS |
1185 | nvkm_wr32(device, 0x40601c, 0xc0000000); |
1186 | nvkm_wr32(device, 0x400108, 0x00000040); | |
f73221e4 BS |
1187 | trap &= ~0x00000040; |
1188 | } | |
1189 | ||
1190 | if (trap & 0x00000080) { | |
276836d4 | 1191 | u32 stat = nvkm_rd32(device, 0x404490); |
caf2be8a IM |
1192 | u32 pc = nvkm_rd32(device, 0x404494); |
1193 | u32 op = nvkm_rd32(device, 0x40449c); | |
1194 | ||
1195 | nvkm_snprintbf(error, sizeof(error), gf100_macro_error, | |
1196 | stat & 0x1fffffff); | |
1197 | nvkm_error(subdev, "MACRO %08x [%s], pc: 0x%03x%s, op: 0x%08x\n", | |
1198 | stat, error, pc & 0x7ff, | |
1199 | (pc & 0x10000000) ? "" : " (invalid)", | |
1200 | op); | |
1201 | ||
276836d4 BS |
1202 | nvkm_wr32(device, 0x404490, 0xc0000000); |
1203 | nvkm_wr32(device, 0x400108, 0x00000080); | |
f73221e4 BS |
1204 | trap &= ~0x00000080; |
1205 | } | |
1206 | ||
30f4e087 | 1207 | if (trap & 0x00000100) { |
109c2f2f | 1208 | u32 stat = nvkm_rd32(device, 0x407020) & 0x3fffffff; |
30f4e087 | 1209 | |
caf2be8a IM |
1210 | nvkm_snprintbf(error, sizeof(error), gk104_sked_error, stat); |
1211 | nvkm_error(subdev, "SKED: %08x [%s]\n", stat, error); | |
30f4e087 | 1212 | |
109c2f2f | 1213 | if (stat) |
276836d4 BS |
1214 | nvkm_wr32(device, 0x407020, 0x40000000); |
1215 | nvkm_wr32(device, 0x400108, 0x00000100); | |
30f4e087 BS |
1216 | trap &= ~0x00000100; |
1217 | } | |
1218 | ||
f73221e4 | 1219 | if (trap & 0x01000000) { |
276836d4 | 1220 | u32 stat = nvkm_rd32(device, 0x400118); |
bfee3f3d | 1221 | for (gpc = 0; stat && gpc < gr->gpc_nr; gpc++) { |
f73221e4 BS |
1222 | u32 mask = 0x00000001 << gpc; |
1223 | if (stat & mask) { | |
bfee3f3d | 1224 | gf100_gr_trap_gpc(gr, gpc); |
276836d4 | 1225 | nvkm_wr32(device, 0x400118, mask); |
f73221e4 BS |
1226 | stat &= ~mask; |
1227 | } | |
1228 | } | |
276836d4 | 1229 | nvkm_wr32(device, 0x400108, 0x01000000); |
f73221e4 BS |
1230 | trap &= ~0x01000000; |
1231 | } | |
1232 | ||
1233 | if (trap & 0x02000000) { | |
bfee3f3d | 1234 | for (rop = 0; rop < gr->rop_nr; rop++) { |
276836d4 BS |
1235 | u32 statz = nvkm_rd32(device, ROP_UNIT(rop, 0x070)); |
1236 | u32 statc = nvkm_rd32(device, ROP_UNIT(rop, 0x144)); | |
109c2f2f | 1237 | nvkm_error(subdev, "ROP%d %08x %08x\n", |
f73221e4 | 1238 | rop, statz, statc); |
276836d4 BS |
1239 | nvkm_wr32(device, ROP_UNIT(rop, 0x070), 0xc0000000); |
1240 | nvkm_wr32(device, ROP_UNIT(rop, 0x144), 0xc0000000); | |
f73221e4 | 1241 | } |
276836d4 | 1242 | nvkm_wr32(device, 0x400108, 0x02000000); |
f73221e4 BS |
1243 | trap &= ~0x02000000; |
1244 | } | |
1245 | ||
1246 | if (trap) { | |
109c2f2f | 1247 | nvkm_error(subdev, "TRAP UNHANDLED %08x\n", trap); |
276836d4 | 1248 | nvkm_wr32(device, 0x400108, trap); |
f73221e4 BS |
1249 | } |
1250 | } | |
1251 | ||
30f4e087 | 1252 | static void |
bfee3f3d | 1253 | gf100_gr_ctxctl_debug_unit(struct gf100_gr *gr, u32 base) |
30f4e087 | 1254 | { |
109c2f2f BS |
1255 | struct nvkm_subdev *subdev = &gr->base.engine.subdev; |
1256 | struct nvkm_device *device = subdev->device; | |
1257 | nvkm_error(subdev, "%06x - done %08x\n", base, | |
1258 | nvkm_rd32(device, base + 0x400)); | |
1259 | nvkm_error(subdev, "%06x - stat %08x %08x %08x %08x\n", base, | |
1260 | nvkm_rd32(device, base + 0x800), | |
1261 | nvkm_rd32(device, base + 0x804), | |
1262 | nvkm_rd32(device, base + 0x808), | |
1263 | nvkm_rd32(device, base + 0x80c)); | |
1264 | nvkm_error(subdev, "%06x - stat %08x %08x %08x %08x\n", base, | |
1265 | nvkm_rd32(device, base + 0x810), | |
1266 | nvkm_rd32(device, base + 0x814), | |
1267 | nvkm_rd32(device, base + 0x818), | |
1268 | nvkm_rd32(device, base + 0x81c)); | |
30f4e087 BS |
1269 | } |
1270 | ||
1271 | void | |
bfee3f3d | 1272 | gf100_gr_ctxctl_debug(struct gf100_gr *gr) |
30f4e087 | 1273 | { |
276836d4 BS |
1274 | struct nvkm_device *device = gr->base.engine.subdev.device; |
1275 | u32 gpcnr = nvkm_rd32(device, 0x409604) & 0xffff; | |
30f4e087 BS |
1276 | u32 gpc; |
1277 | ||
bfee3f3d | 1278 | gf100_gr_ctxctl_debug_unit(gr, 0x409000); |
30f4e087 | 1279 | for (gpc = 0; gpc < gpcnr; gpc++) |
bfee3f3d | 1280 | gf100_gr_ctxctl_debug_unit(gr, 0x502000 + (gpc * 0x8000)); |
30f4e087 BS |
1281 | } |
1282 | ||
1283 | static void | |
bfee3f3d | 1284 | gf100_gr_ctxctl_isr(struct gf100_gr *gr) |
30f4e087 | 1285 | { |
109c2f2f BS |
1286 | struct nvkm_subdev *subdev = &gr->base.engine.subdev; |
1287 | struct nvkm_device *device = subdev->device; | |
276836d4 | 1288 | u32 stat = nvkm_rd32(device, 0x409c18); |
30f4e087 | 1289 | |
732be807 | 1290 | if (!gr->firmware && (stat & 0x00000001)) { |
276836d4 | 1291 | u32 code = nvkm_rd32(device, 0x409814); |
23f67841 | 1292 | if (code == E_BAD_FWMTHD) { |
276836d4 BS |
1293 | u32 class = nvkm_rd32(device, 0x409808); |
1294 | u32 addr = nvkm_rd32(device, 0x40980c); | |
23f67841 BS |
1295 | u32 subc = (addr & 0x00070000) >> 16; |
1296 | u32 mthd = (addr & 0x00003ffc); | |
276836d4 | 1297 | u32 data = nvkm_rd32(device, 0x409810); |
23f67841 | 1298 | |
109c2f2f BS |
1299 | nvkm_error(subdev, "FECS MTHD subc %d class %04x " |
1300 | "mthd %04x data %08x\n", | |
1301 | subc, class, mthd, data); | |
23f67841 | 1302 | } else { |
109c2f2f | 1303 | nvkm_error(subdev, "FECS ucode error %d\n", code); |
23f67841 | 1304 | } |
48dac935 BS |
1305 | nvkm_wr32(device, 0x409c20, 0x00000001); |
1306 | stat &= ~0x00000001; | |
23f67841 | 1307 | } |
30f4e087 | 1308 | |
732be807 | 1309 | if (!gr->firmware && (stat & 0x00080000)) { |
109c2f2f | 1310 | nvkm_error(subdev, "FECS watchdog timeout\n"); |
bfee3f3d | 1311 | gf100_gr_ctxctl_debug(gr); |
276836d4 | 1312 | nvkm_wr32(device, 0x409c20, 0x00080000); |
23f67841 BS |
1313 | stat &= ~0x00080000; |
1314 | } | |
1315 | ||
1316 | if (stat) { | |
109c2f2f | 1317 | nvkm_error(subdev, "FECS %08x\n", stat); |
bfee3f3d | 1318 | gf100_gr_ctxctl_debug(gr); |
276836d4 | 1319 | nvkm_wr32(device, 0x409c20, stat); |
23f67841 | 1320 | } |
30f4e087 BS |
1321 | } |
1322 | ||
ebb945a9 | 1323 | static void |
c85ee6ca | 1324 | gf100_gr_intr(struct nvkm_gr *base) |
ebb945a9 | 1325 | { |
c85ee6ca BS |
1326 | struct gf100_gr *gr = gf100_gr(base); |
1327 | struct nvkm_subdev *subdev = &gr->base.engine.subdev; | |
1328 | struct nvkm_device *device = subdev->device; | |
a65955e1 BS |
1329 | struct nvkm_fifo_chan *chan; |
1330 | unsigned long flags; | |
276836d4 BS |
1331 | u64 inst = nvkm_rd32(device, 0x409b00) & 0x0fffffff; |
1332 | u32 stat = nvkm_rd32(device, 0x400100); | |
1333 | u32 addr = nvkm_rd32(device, 0x400704); | |
ebb945a9 BS |
1334 | u32 mthd = (addr & 0x00003ffc); |
1335 | u32 subc = (addr & 0x00070000) >> 16; | |
276836d4 BS |
1336 | u32 data = nvkm_rd32(device, 0x400708); |
1337 | u32 code = nvkm_rd32(device, 0x400110); | |
91c772ec | 1338 | u32 class; |
8f0649b5 BS |
1339 | const char *name = "unknown"; |
1340 | int chid = -1; | |
72a14827 | 1341 | |
a65955e1 | 1342 | chan = nvkm_fifo_chan_inst(device->fifo, (u64)inst << 12, &flags); |
8f0649b5 BS |
1343 | if (chan) { |
1344 | name = chan->object.client->name; | |
1345 | chid = chan->chid; | |
1346 | } | |
a65955e1 | 1347 | |
c85ee6ca | 1348 | if (device->card_type < NV_E0 || subc < 4) |
276836d4 | 1349 | class = nvkm_rd32(device, 0x404200 + (subc * 4)); |
91c772ec BS |
1350 | else |
1351 | class = 0x0000; | |
1352 | ||
c6a7b026 LP |
1353 | if (stat & 0x00000001) { |
1354 | /* | |
1355 | * notifier interrupt, only needed for cyclestats | |
1356 | * can be safely ignored | |
1357 | */ | |
276836d4 | 1358 | nvkm_wr32(device, 0x400100, 0x00000001); |
c6a7b026 LP |
1359 | stat &= ~0x00000001; |
1360 | } | |
1361 | ||
ebb945a9 | 1362 | if (stat & 0x00000010) { |
a65955e1 | 1363 | if (!gf100_gr_mthd_sw(device, class, mthd, data)) { |
109c2f2f BS |
1364 | nvkm_error(subdev, "ILLEGAL_MTHD ch %d [%010llx %s] " |
1365 | "subc %d class %04x mthd %04x data %08x\n", | |
8f0649b5 BS |
1366 | chid, inst << 12, name, subc, |
1367 | class, mthd, data); | |
ebb945a9 | 1368 | } |
276836d4 | 1369 | nvkm_wr32(device, 0x400100, 0x00000010); |
ebb945a9 BS |
1370 | stat &= ~0x00000010; |
1371 | } | |
1372 | ||
1373 | if (stat & 0x00000020) { | |
109c2f2f BS |
1374 | nvkm_error(subdev, "ILLEGAL_CLASS ch %d [%010llx %s] " |
1375 | "subc %d class %04x mthd %04x data %08x\n", | |
8f0649b5 | 1376 | chid, inst << 12, name, subc, class, mthd, data); |
276836d4 | 1377 | nvkm_wr32(device, 0x400100, 0x00000020); |
ebb945a9 BS |
1378 | stat &= ~0x00000020; |
1379 | } | |
1380 | ||
1381 | if (stat & 0x00100000) { | |
109c2f2f BS |
1382 | const struct nvkm_enum *en = |
1383 | nvkm_enum_find(nv50_data_error_names, code); | |
1384 | nvkm_error(subdev, "DATA_ERROR %08x [%s] ch %d [%010llx %s] " | |
1385 | "subc %d class %04x mthd %04x data %08x\n", | |
1386 | code, en ? en->name : "", chid, inst << 12, | |
8f0649b5 | 1387 | name, subc, class, mthd, data); |
276836d4 | 1388 | nvkm_wr32(device, 0x400100, 0x00100000); |
ebb945a9 BS |
1389 | stat &= ~0x00100000; |
1390 | } | |
1391 | ||
1392 | if (stat & 0x00200000) { | |
109c2f2f | 1393 | nvkm_error(subdev, "TRAP ch %d [%010llx %s]\n", |
8f0649b5 | 1394 | chid, inst << 12, name); |
bfee3f3d | 1395 | gf100_gr_trap_intr(gr); |
276836d4 | 1396 | nvkm_wr32(device, 0x400100, 0x00200000); |
ebb945a9 BS |
1397 | stat &= ~0x00200000; |
1398 | } | |
1399 | ||
1400 | if (stat & 0x00080000) { | |
bfee3f3d | 1401 | gf100_gr_ctxctl_isr(gr); |
276836d4 | 1402 | nvkm_wr32(device, 0x400100, 0x00080000); |
ebb945a9 BS |
1403 | stat &= ~0x00080000; |
1404 | } | |
1405 | ||
1406 | if (stat) { | |
109c2f2f | 1407 | nvkm_error(subdev, "intr %08x\n", stat); |
276836d4 | 1408 | nvkm_wr32(device, 0x400100, stat); |
ebb945a9 BS |
1409 | } |
1410 | ||
276836d4 | 1411 | nvkm_wr32(device, 0x400500, 0x00010001); |
a65955e1 | 1412 | nvkm_fifo_chan_put(device->fifo, flags, &chan); |
ebb945a9 BS |
1413 | } |
1414 | ||
e08a1d97 | 1415 | static void |
89cd6e20 | 1416 | gf100_gr_init_fw(struct nvkm_falcon *falcon, |
e3c71eb2 | 1417 | struct gf100_gr_fuc *code, struct gf100_gr_fuc *data) |
ebb945a9 | 1418 | { |
89cd6e20 AC |
1419 | nvkm_falcon_load_dmem(falcon, data->data, 0x0, data->size, 0); |
1420 | nvkm_falcon_load_imem(falcon, code->data, 0x0, code->size, 0, 0, false); | |
966a5b7d BS |
1421 | } |
1422 | ||
a32b2ffb | 1423 | static void |
bfee3f3d | 1424 | gf100_gr_init_csdata(struct gf100_gr *gr, |
e3c71eb2 BS |
1425 | const struct gf100_gr_pack *pack, |
1426 | u32 falcon, u32 starstar, u32 base) | |
a32b2ffb | 1427 | { |
276836d4 | 1428 | struct nvkm_device *device = gr->base.engine.subdev.device; |
e3c71eb2 BS |
1429 | const struct gf100_gr_pack *iter; |
1430 | const struct gf100_gr_init *init; | |
c33b1e8c | 1431 | u32 addr = ~0, prev = ~0, xfer = 0; |
a32b2ffb BS |
1432 | u32 star, temp; |
1433 | ||
276836d4 BS |
1434 | nvkm_wr32(device, falcon + 0x01c0, 0x02000000 + starstar); |
1435 | star = nvkm_rd32(device, falcon + 0x01c4); | |
1436 | temp = nvkm_rd32(device, falcon + 0x01c4); | |
a32b2ffb BS |
1437 | if (temp > star) |
1438 | star = temp; | |
276836d4 | 1439 | nvkm_wr32(device, falcon + 0x01c0, 0x01000000 + star); |
a32b2ffb | 1440 | |
c33b1e8c BS |
1441 | pack_for_each_init(init, iter, pack) { |
1442 | u32 head = init->addr - base; | |
1443 | u32 tail = head + init->count * init->pitch; | |
1444 | while (head < tail) { | |
1445 | if (head != prev + 4 || xfer >= 32) { | |
1446 | if (xfer) { | |
1447 | u32 data = ((--xfer << 26) | addr); | |
276836d4 | 1448 | nvkm_wr32(device, falcon + 0x01c4, data); |
c33b1e8c BS |
1449 | star += 4; |
1450 | } | |
1451 | addr = head; | |
1452 | xfer = 0; | |
a32b2ffb | 1453 | } |
c33b1e8c BS |
1454 | prev = head; |
1455 | xfer = xfer + 1; | |
1456 | head = head + init->pitch; | |
a32b2ffb | 1457 | } |
c33b1e8c | 1458 | } |
a32b2ffb | 1459 | |
276836d4 BS |
1460 | nvkm_wr32(device, falcon + 0x01c4, (--xfer << 26) | addr); |
1461 | nvkm_wr32(device, falcon + 0x01c0, 0x01000004 + starstar); | |
1462 | nvkm_wr32(device, falcon + 0x01c4, star + 4); | |
a32b2ffb BS |
1463 | } |
1464 | ||
0296b5d9 AC |
1465 | /* Initialize context from an external (secure or not) firmware */ |
1466 | static int | |
1467 | gf100_gr_init_ctxctl_ext(struct gf100_gr *gr) | |
966a5b7d | 1468 | { |
109c2f2f BS |
1469 | struct nvkm_subdev *subdev = &gr->base.engine.subdev; |
1470 | struct nvkm_device *device = subdev->device; | |
c9469aae | 1471 | struct nvkm_secboot *sb = device->secboot; |
598a8148 | 1472 | u32 secboot_mask = 0; |
ebb945a9 | 1473 | |
0296b5d9 AC |
1474 | /* load fuc microcode */ |
1475 | nvkm_mc_unk260(device, 0); | |
c9469aae | 1476 | |
0296b5d9 AC |
1477 | /* securely-managed falcons must be reset using secure boot */ |
1478 | if (nvkm_secboot_is_managed(sb, NVKM_SECBOOT_FALCON_FECS)) | |
598a8148 | 1479 | secboot_mask |= BIT(NVKM_SECBOOT_FALCON_FECS); |
0296b5d9 | 1480 | else |
89cd6e20 | 1481 | gf100_gr_init_fw(gr->fecs, &gr->fuc409c, &gr->fuc409d); |
7fcab839 | 1482 | |
0296b5d9 | 1483 | if (nvkm_secboot_is_managed(sb, NVKM_SECBOOT_FALCON_GPCCS)) |
598a8148 | 1484 | secboot_mask |= BIT(NVKM_SECBOOT_FALCON_GPCCS); |
0296b5d9 | 1485 | else |
89cd6e20 | 1486 | gf100_gr_init_fw(gr->gpccs, &gr->fuc41ac, &gr->fuc41ad); |
598a8148 AC |
1487 | |
1488 | if (secboot_mask != 0) { | |
1489 | int ret = nvkm_secboot_reset(sb, secboot_mask); | |
1490 | if (ret) | |
1491 | return ret; | |
1492 | } | |
c9469aae | 1493 | |
0296b5d9 AC |
1494 | nvkm_mc_unk260(device, 1); |
1495 | ||
1496 | /* start both of them running */ | |
1497 | nvkm_wr32(device, 0x409840, 0xffffffff); | |
1498 | nvkm_wr32(device, 0x41a10c, 0x00000000); | |
1499 | nvkm_wr32(device, 0x40910c, 0x00000000); | |
1500 | ||
89cd6e20 AC |
1501 | nvkm_falcon_start(gr->gpccs); |
1502 | nvkm_falcon_start(gr->fecs); | |
1503 | ||
0296b5d9 AC |
1504 | if (nvkm_msec(device, 2000, |
1505 | if (nvkm_rd32(device, 0x409800) & 0x00000001) | |
1506 | break; | |
1507 | ) < 0) | |
1508 | return -EBUSY; | |
1509 | ||
1510 | nvkm_wr32(device, 0x409840, 0xffffffff); | |
1511 | nvkm_wr32(device, 0x409500, 0x7fffffff); | |
1512 | nvkm_wr32(device, 0x409504, 0x00000021); | |
1513 | ||
1514 | nvkm_wr32(device, 0x409840, 0xffffffff); | |
1515 | nvkm_wr32(device, 0x409500, 0x00000000); | |
1516 | nvkm_wr32(device, 0x409504, 0x00000010); | |
1517 | if (nvkm_msec(device, 2000, | |
1518 | if ((gr->size = nvkm_rd32(device, 0x409800))) | |
1519 | break; | |
1520 | ) < 0) | |
1521 | return -EBUSY; | |
1522 | ||
1523 | nvkm_wr32(device, 0x409840, 0xffffffff); | |
1524 | nvkm_wr32(device, 0x409500, 0x00000000); | |
1525 | nvkm_wr32(device, 0x409504, 0x00000016); | |
1526 | if (nvkm_msec(device, 2000, | |
1527 | if (nvkm_rd32(device, 0x409800)) | |
1528 | break; | |
1529 | ) < 0) | |
1530 | return -EBUSY; | |
bfee3f3d | 1531 | |
0296b5d9 AC |
1532 | nvkm_wr32(device, 0x409840, 0xffffffff); |
1533 | nvkm_wr32(device, 0x409500, 0x00000000); | |
1534 | nvkm_wr32(device, 0x409504, 0x00000025); | |
1535 | if (nvkm_msec(device, 2000, | |
1536 | if (nvkm_rd32(device, 0x409800)) | |
1537 | break; | |
1538 | ) < 0) | |
1539 | return -EBUSY; | |
bfee3f3d | 1540 | |
0296b5d9 AC |
1541 | if (device->chipset >= 0xe0) { |
1542 | nvkm_wr32(device, 0x409800, 0x00000000); | |
1543 | nvkm_wr32(device, 0x409500, 0x00000001); | |
1544 | nvkm_wr32(device, 0x409504, 0x00000030); | |
c4584adc | 1545 | if (nvkm_msec(device, 2000, |
0296b5d9 | 1546 | if (nvkm_rd32(device, 0x409800)) |
c4584adc BS |
1547 | break; |
1548 | ) < 0) | |
30f4e087 | 1549 | return -EBUSY; |
0411de85 | 1550 | |
0296b5d9 AC |
1551 | nvkm_wr32(device, 0x409810, 0xb00095c8); |
1552 | nvkm_wr32(device, 0x409800, 0x00000000); | |
1553 | nvkm_wr32(device, 0x409500, 0x00000001); | |
1554 | nvkm_wr32(device, 0x409504, 0x00000031); | |
c4584adc BS |
1555 | if (nvkm_msec(device, 2000, |
1556 | if (nvkm_rd32(device, 0x409800)) | |
1557 | break; | |
1558 | ) < 0) | |
ac1499d9 | 1559 | return -EBUSY; |
ac1499d9 | 1560 | |
0296b5d9 AC |
1561 | nvkm_wr32(device, 0x409810, 0x00080420); |
1562 | nvkm_wr32(device, 0x409800, 0x00000000); | |
1563 | nvkm_wr32(device, 0x409500, 0x00000001); | |
1564 | nvkm_wr32(device, 0x409504, 0x00000032); | |
c4584adc BS |
1565 | if (nvkm_msec(device, 2000, |
1566 | if (nvkm_rd32(device, 0x409800)) | |
1567 | break; | |
1568 | ) < 0) | |
ac1499d9 | 1569 | return -EBUSY; |
ac1499d9 | 1570 | |
0296b5d9 AC |
1571 | nvkm_wr32(device, 0x409614, 0x00000070); |
1572 | nvkm_wr32(device, 0x409614, 0x00000770); | |
1573 | nvkm_wr32(device, 0x40802c, 0x00000001); | |
1574 | } | |
30f4e087 | 1575 | |
0296b5d9 AC |
1576 | if (gr->data == NULL) { |
1577 | int ret = gf100_grctx_generate(gr); | |
1578 | if (ret) { | |
1579 | nvkm_error(subdev, "failed to construct context\n"); | |
1580 | return ret; | |
30f4e087 | 1581 | } |
0296b5d9 | 1582 | } |
30f4e087 | 1583 | |
0296b5d9 AC |
1584 | return 0; |
1585 | } | |
1586 | ||
1587 | static int | |
1588 | gf100_gr_init_ctxctl_int(struct gf100_gr *gr) | |
1589 | { | |
1590 | const struct gf100_grctx_func *grctx = gr->func->grctx; | |
1591 | struct nvkm_subdev *subdev = &gr->base.engine.subdev; | |
1592 | struct nvkm_device *device = subdev->device; | |
ebb945a9 | 1593 | |
c85ee6ca | 1594 | if (!gr->func->fecs.ucode) { |
96616b4c | 1595 | return -ENOSYS; |
0411de85 | 1596 | } |
966a5b7d | 1597 | |
ac1499d9 | 1598 | /* load HUB microcode */ |
d3981190 | 1599 | nvkm_mc_unk260(device, 0); |
89cd6e20 AC |
1600 | nvkm_falcon_load_dmem(gr->fecs, gr->func->fecs.ucode->data.data, 0x0, |
1601 | gr->func->fecs.ucode->data.size, 0); | |
1602 | nvkm_falcon_load_imem(gr->fecs, gr->func->fecs.ucode->code.data, 0x0, | |
1603 | gr->func->fecs.ucode->code.size, 0, 0, false); | |
ac1499d9 BS |
1604 | |
1605 | /* load GPC microcode */ | |
89cd6e20 AC |
1606 | nvkm_falcon_load_dmem(gr->gpccs, gr->func->gpccs.ucode->data.data, 0x0, |
1607 | gr->func->gpccs.ucode->data.size, 0); | |
1608 | nvkm_falcon_load_imem(gr->gpccs, gr->func->gpccs.ucode->code.data, 0x0, | |
1609 | gr->func->gpccs.ucode->code.size, 0, 0, false); | |
d3981190 | 1610 | nvkm_mc_unk260(device, 1); |
966a5b7d | 1611 | |
c33b1e8c | 1612 | /* load register lists */ |
27f3d6cf BS |
1613 | gf100_gr_init_csdata(gr, grctx->hub, 0x409000, 0x000, 0x000000); |
1614 | gf100_gr_init_csdata(gr, grctx->gpc, 0x41a000, 0x000, 0x418000); | |
1615 | gf100_gr_init_csdata(gr, grctx->tpc, 0x41a000, 0x004, 0x419800); | |
1616 | gf100_gr_init_csdata(gr, grctx->ppc, 0x41a000, 0x008, 0x41be00); | |
a32b2ffb | 1617 | |
ac1499d9 | 1618 | /* start HUB ucode running, it'll init the GPCs */ |
276836d4 BS |
1619 | nvkm_wr32(device, 0x40910c, 0x00000000); |
1620 | nvkm_wr32(device, 0x409100, 0x00000002); | |
c4584adc BS |
1621 | if (nvkm_msec(device, 2000, |
1622 | if (nvkm_rd32(device, 0x409800) & 0x80000000) | |
1623 | break; | |
1624 | ) < 0) { | |
bfee3f3d | 1625 | gf100_gr_ctxctl_debug(gr); |
966a5b7d BS |
1626 | return -EBUSY; |
1627 | } | |
966a5b7d | 1628 | |
276836d4 | 1629 | gr->size = nvkm_rd32(device, 0x409804); |
bfee3f3d BS |
1630 | if (gr->data == NULL) { |
1631 | int ret = gf100_grctx_generate(gr); | |
ac1499d9 | 1632 | if (ret) { |
109c2f2f | 1633 | nvkm_error(subdev, "failed to construct context\n"); |
ac1499d9 BS |
1634 | return ret; |
1635 | } | |
966a5b7d BS |
1636 | } |
1637 | ||
1638 | return 0; | |
4b223eef BS |
1639 | } |
1640 | ||
0296b5d9 AC |
1641 | int |
1642 | gf100_gr_init_ctxctl(struct gf100_gr *gr) | |
1643 | { | |
1644 | int ret; | |
1645 | ||
1646 | if (gr->firmware) | |
1647 | ret = gf100_gr_init_ctxctl_ext(gr); | |
1648 | else | |
1649 | ret = gf100_gr_init_ctxctl_int(gr); | |
1650 | ||
1651 | return ret; | |
1652 | } | |
1653 | ||
c85ee6ca BS |
1654 | static int |
1655 | gf100_gr_oneinit(struct nvkm_gr *base) | |
1656 | { | |
1657 | struct gf100_gr *gr = gf100_gr(base); | |
9e439757 AC |
1658 | struct nvkm_subdev *subdev = &gr->base.engine.subdev; |
1659 | struct nvkm_device *device = subdev->device; | |
99c59172 | 1660 | int i, j; |
9e439757 AC |
1661 | int ret; |
1662 | ||
1663 | ret = nvkm_falcon_v1_new(subdev, "FECS", 0x409000, &gr->fecs); | |
1664 | if (ret) | |
1665 | return ret; | |
1666 | ||
1667 | ret = nvkm_falcon_v1_new(subdev, "GPCCS", 0x41a000, &gr->gpccs); | |
1668 | if (ret) | |
1669 | return ret; | |
c85ee6ca BS |
1670 | |
1671 | nvkm_pmu_pgob(device->pmu, false); | |
1672 | ||
64cb5a31 BS |
1673 | gr->rop_nr = gr->func->rops(gr); |
1674 | gr->gpc_nr = nvkm_rd32(device, 0x409604) & 0x0000001f; | |
c85ee6ca BS |
1675 | for (i = 0; i < gr->gpc_nr; i++) { |
1676 | gr->tpc_nr[i] = nvkm_rd32(device, GPC_UNIT(i, 0x2608)); | |
1677 | gr->tpc_total += gr->tpc_nr[i]; | |
1678 | gr->ppc_nr[i] = gr->func->ppc_nr; | |
1679 | for (j = 0; j < gr->ppc_nr[i]; j++) { | |
1680 | u8 mask = nvkm_rd32(device, GPC_UNIT(i, 0x0c30 + (j * 4))); | |
2fb2b3c6 BS |
1681 | if (mask) |
1682 | gr->ppc_mask[i] |= (1 << j); | |
c85ee6ca BS |
1683 | gr->ppc_tpc_nr[i][j] = hweight8(mask); |
1684 | } | |
1685 | } | |
1686 | ||
1687 | /*XXX: these need figuring out... though it might not even matter */ | |
1688 | switch (device->chipset) { | |
1689 | case 0xc0: | |
1690 | if (gr->tpc_total == 11) { /* 465, 3/4/4/0, 4 */ | |
5ec3def7 | 1691 | gr->screen_tile_row_offset = 0x07; |
c85ee6ca BS |
1692 | } else |
1693 | if (gr->tpc_total == 14) { /* 470, 3/3/4/4, 5 */ | |
5ec3def7 | 1694 | gr->screen_tile_row_offset = 0x05; |
c85ee6ca BS |
1695 | } else |
1696 | if (gr->tpc_total == 15) { /* 480, 3/4/4/4, 6 */ | |
5ec3def7 | 1697 | gr->screen_tile_row_offset = 0x06; |
c85ee6ca BS |
1698 | } |
1699 | break; | |
1700 | case 0xc3: /* 450, 4/0/0/0, 2 */ | |
5ec3def7 | 1701 | gr->screen_tile_row_offset = 0x03; |
c85ee6ca BS |
1702 | break; |
1703 | case 0xc4: /* 460, 3/4/0/0, 4 */ | |
5ec3def7 | 1704 | gr->screen_tile_row_offset = 0x01; |
c85ee6ca BS |
1705 | break; |
1706 | case 0xc1: /* 2/0/0/0, 1 */ | |
5ec3def7 | 1707 | gr->screen_tile_row_offset = 0x01; |
c85ee6ca BS |
1708 | break; |
1709 | case 0xc8: /* 4/4/3/4, 5 */ | |
5ec3def7 | 1710 | gr->screen_tile_row_offset = 0x06; |
c85ee6ca BS |
1711 | break; |
1712 | case 0xce: /* 4/4/0/0, 4 */ | |
5ec3def7 | 1713 | gr->screen_tile_row_offset = 0x03; |
c85ee6ca BS |
1714 | break; |
1715 | case 0xcf: /* 4/0/0/0, 3 */ | |
5ec3def7 | 1716 | gr->screen_tile_row_offset = 0x03; |
c85ee6ca BS |
1717 | break; |
1718 | case 0xd7: | |
1719 | case 0xd9: /* 1/0/0/0, 1 */ | |
1720 | case 0xea: /* gk20a */ | |
1721 | case 0x12b: /* gm20b */ | |
5ec3def7 | 1722 | gr->screen_tile_row_offset = 0x01; |
c85ee6ca BS |
1723 | break; |
1724 | } | |
1725 | ||
1726 | return 0; | |
1727 | } | |
1728 | ||
e08a1d97 | 1729 | static int |
c85ee6ca BS |
1730 | gf100_gr_init_(struct nvkm_gr *base) |
1731 | { | |
1732 | struct gf100_gr *gr = gf100_gr(base); | |
89cd6e20 AC |
1733 | struct nvkm_subdev *subdev = &base->engine.subdev; |
1734 | u32 ret; | |
1735 | ||
c85ee6ca | 1736 | nvkm_pmu_pgob(gr->base.engine.subdev.device->pmu, false); |
89cd6e20 AC |
1737 | |
1738 | ret = nvkm_falcon_get(gr->fecs, subdev); | |
1739 | if (ret) | |
1740 | return ret; | |
1741 | ||
1742 | ret = nvkm_falcon_get(gr->gpccs, subdev); | |
1743 | if (ret) | |
1744 | return ret; | |
1745 | ||
c85ee6ca BS |
1746 | return gr->func->init(gr); |
1747 | } | |
1748 | ||
89cd6e20 AC |
1749 | static int |
1750 | gf100_gr_fini_(struct nvkm_gr *base, bool suspend) | |
1751 | { | |
1752 | struct gf100_gr *gr = gf100_gr(base); | |
1753 | struct nvkm_subdev *subdev = &gr->base.engine.subdev; | |
1754 | nvkm_falcon_put(gr->gpccs, subdev); | |
1755 | nvkm_falcon_put(gr->fecs, subdev); | |
1756 | return 0; | |
1757 | } | |
1758 | ||
c85ee6ca BS |
1759 | void |
1760 | gf100_gr_dtor_fw(struct gf100_gr_fuc *fuc) | |
1761 | { | |
1762 | kfree(fuc->data); | |
1763 | fuc->data = NULL; | |
1764 | } | |
1765 | ||
336c4652 AC |
1766 | static void |
1767 | gf100_gr_dtor_init(struct gf100_gr_pack *pack) | |
1768 | { | |
1769 | vfree(pack); | |
1770 | } | |
1771 | ||
c85ee6ca BS |
1772 | void * |
1773 | gf100_gr_dtor(struct nvkm_gr *base) | |
1774 | { | |
1775 | struct gf100_gr *gr = gf100_gr(base); | |
1776 | ||
1777 | if (gr->func->dtor) | |
1778 | gr->func->dtor(gr); | |
1779 | kfree(gr->data); | |
1780 | ||
89cd6e20 AC |
1781 | nvkm_falcon_del(&gr->gpccs); |
1782 | nvkm_falcon_del(&gr->fecs); | |
1783 | ||
c85ee6ca BS |
1784 | gf100_gr_dtor_fw(&gr->fuc409c); |
1785 | gf100_gr_dtor_fw(&gr->fuc409d); | |
1786 | gf100_gr_dtor_fw(&gr->fuc41ac); | |
1787 | gf100_gr_dtor_fw(&gr->fuc41ad); | |
1788 | ||
336c4652 AC |
1789 | gf100_gr_dtor_init(gr->fuc_bundle); |
1790 | gf100_gr_dtor_init(gr->fuc_method); | |
1791 | gf100_gr_dtor_init(gr->fuc_sw_ctx); | |
1792 | gf100_gr_dtor_init(gr->fuc_sw_nonctx); | |
1793 | ||
c85ee6ca BS |
1794 | return gr; |
1795 | } | |
1796 | ||
1797 | static const struct nvkm_gr_func | |
1798 | gf100_gr_ = { | |
1799 | .dtor = gf100_gr_dtor, | |
1800 | .oneinit = gf100_gr_oneinit, | |
1801 | .init = gf100_gr_init_, | |
89cd6e20 | 1802 | .fini = gf100_gr_fini_, |
c85ee6ca BS |
1803 | .intr = gf100_gr_intr, |
1804 | .units = gf100_gr_units, | |
1805 | .chan_new = gf100_gr_chan_new, | |
1806 | .object_get = gf100_gr_object_get, | |
6845c313 | 1807 | .chsw_load = gf100_gr_chsw_load, |
c85ee6ca BS |
1808 | }; |
1809 | ||
e137040e AC |
1810 | int |
1811 | gf100_gr_ctor_fw_legacy(struct gf100_gr *gr, const char *fwname, | |
1812 | struct gf100_gr_fuc *fuc, int ret) | |
1813 | { | |
1814 | struct nvkm_subdev *subdev = &gr->base.engine.subdev; | |
1815 | struct nvkm_device *device = subdev->device; | |
1816 | const struct firmware *fw; | |
1817 | char f[32]; | |
1818 | ||
1819 | /* see if this firmware has a legacy path */ | |
1820 | if (!strcmp(fwname, "fecs_inst")) | |
1821 | fwname = "fuc409c"; | |
1822 | else if (!strcmp(fwname, "fecs_data")) | |
1823 | fwname = "fuc409d"; | |
1824 | else if (!strcmp(fwname, "gpccs_inst")) | |
1825 | fwname = "fuc41ac"; | |
1826 | else if (!strcmp(fwname, "gpccs_data")) | |
1827 | fwname = "fuc41ad"; | |
1828 | else { | |
1829 | /* nope, let's just return the error we got */ | |
1830 | nvkm_error(subdev, "failed to load %s\n", fwname); | |
1831 | return ret; | |
1832 | } | |
1833 | ||
1834 | /* yes, try to load from the legacy path */ | |
1835 | nvkm_debug(subdev, "%s: falling back to legacy path\n", fwname); | |
1836 | ||
1837 | snprintf(f, sizeof(f), "nouveau/nv%02x_%s", device->chipset, fwname); | |
1838 | ret = request_firmware(&fw, f, device->dev); | |
1839 | if (ret) { | |
1840 | snprintf(f, sizeof(f), "nouveau/%s", fwname); | |
1841 | ret = request_firmware(&fw, f, device->dev); | |
1842 | if (ret) { | |
1843 | nvkm_error(subdev, "failed to load %s\n", fwname); | |
1844 | return ret; | |
1845 | } | |
1846 | } | |
1847 | ||
1848 | fuc->size = fw->size; | |
1849 | fuc->data = kmemdup(fw->data, fuc->size, GFP_KERNEL); | |
1850 | release_firmware(fw); | |
1851 | return (fuc->data != NULL) ? 0 : -ENOMEM; | |
1852 | } | |
1853 | ||
c85ee6ca BS |
1854 | int |
1855 | gf100_gr_ctor_fw(struct gf100_gr *gr, const char *fwname, | |
1856 | struct gf100_gr_fuc *fuc) | |
1857 | { | |
1858 | struct nvkm_subdev *subdev = &gr->base.engine.subdev; | |
1859 | struct nvkm_device *device = subdev->device; | |
1860 | const struct firmware *fw; | |
c85ee6ca | 1861 | int ret; |
c85ee6ca | 1862 | |
33bcb4c3 | 1863 | ret = nvkm_firmware_get(device, fwname, &fw); |
46bda4f4 BS |
1864 | if (ret) { |
1865 | ret = gf100_gr_ctor_fw_legacy(gr, fwname, fuc, ret); | |
1866 | if (ret) | |
1867 | return -ENODEV; | |
1868 | return 0; | |
1869 | } | |
c85ee6ca BS |
1870 | |
1871 | fuc->size = fw->size; | |
1872 | fuc->data = kmemdup(fw->data, fuc->size, GFP_KERNEL); | |
33bcb4c3 | 1873 | nvkm_firmware_put(fw); |
c85ee6ca BS |
1874 | return (fuc->data != NULL) ? 0 : -ENOMEM; |
1875 | } | |
1876 | ||
1877 | int | |
1878 | gf100_gr_ctor(const struct gf100_gr_func *func, struct nvkm_device *device, | |
1879 | int index, struct gf100_gr *gr) | |
1880 | { | |
c85ee6ca BS |
1881 | gr->func = func; |
1882 | gr->firmware = nvkm_boolopt(device->cfgopt, "NvGrUseFW", | |
1883 | func->fecs.ucode == NULL); | |
1884 | ||
9e439757 AC |
1885 | return nvkm_gr_ctor(&gf100_gr_, device, index, |
1886 | gr->firmware || func->fecs.ucode != NULL, | |
1887 | &gr->base); | |
c85ee6ca BS |
1888 | } |
1889 | ||
30f4e087 | 1890 | int |
c85ee6ca BS |
1891 | gf100_gr_new_(const struct gf100_gr_func *func, struct nvkm_device *device, |
1892 | int index, struct nvkm_gr **pgr) | |
1893 | { | |
1894 | struct gf100_gr *gr; | |
18cd5bc8 AC |
1895 | int ret; |
1896 | ||
c85ee6ca BS |
1897 | if (!(gr = kzalloc(sizeof(*gr), GFP_KERNEL))) |
1898 | return -ENOMEM; | |
1899 | *pgr = &gr->base; | |
18cd5bc8 AC |
1900 | |
1901 | ret = gf100_gr_ctor(func, device, index, gr); | |
1902 | if (ret) | |
1903 | return ret; | |
1904 | ||
1905 | if (gr->firmware) { | |
1906 | if (gf100_gr_ctor_fw(gr, "fecs_inst", &gr->fuc409c) || | |
1907 | gf100_gr_ctor_fw(gr, "fecs_data", &gr->fuc409d) || | |
1908 | gf100_gr_ctor_fw(gr, "gpccs_inst", &gr->fuc41ac) || | |
1909 | gf100_gr_ctor_fw(gr, "gpccs_data", &gr->fuc41ad)) | |
1910 | return -ENODEV; | |
1911 | } | |
1912 | ||
1913 | return 0; | |
c85ee6ca BS |
1914 | } |
1915 | ||
223eaf4b BS |
1916 | void |
1917 | gf100_gr_init_gpc_mmu(struct gf100_gr *gr) | |
1918 | { | |
1919 | struct nvkm_device *device = gr->base.engine.subdev.device; | |
1920 | struct nvkm_fb *fb = device->fb; | |
1921 | ||
b6838c14 | 1922 | nvkm_wr32(device, 0x418880, nvkm_rd32(device, 0x100c80) & 0x00000001); |
223eaf4b BS |
1923 | nvkm_wr32(device, 0x4188a4, 0x00000000); |
1924 | nvkm_wr32(device, 0x418888, 0x00000000); | |
1925 | nvkm_wr32(device, 0x41888c, 0x00000000); | |
1926 | nvkm_wr32(device, 0x418890, 0x00000000); | |
1927 | nvkm_wr32(device, 0x418894, 0x00000000); | |
1928 | nvkm_wr32(device, 0x4188b4, nvkm_memory_addr(fb->mmu_wr) >> 8); | |
1929 | nvkm_wr32(device, 0x4188b8, nvkm_memory_addr(fb->mmu_rd) >> 8); | |
1930 | } | |
1931 | ||
c85ee6ca BS |
1932 | int |
1933 | gf100_gr_init(struct gf100_gr *gr) | |
4b223eef | 1934 | { |
276836d4 | 1935 | struct nvkm_device *device = gr->base.engine.subdev.device; |
bfee3f3d | 1936 | const u32 magicgpc918 = DIV_ROUND_UP(0x00800000, gr->tpc_total); |
30f4e087 BS |
1937 | u32 data[TPC_MAX / 8] = {}; |
1938 | u8 tpcnr[GPC_MAX]; | |
1939 | int gpc, tpc, rop; | |
c85ee6ca | 1940 | int i; |
ebb945a9 | 1941 | |
223eaf4b | 1942 | gr->func->init_gpc_mmu(gr); |
30f4e087 | 1943 | |
c85ee6ca | 1944 | gf100_gr_mmio(gr, gr->func->mmio); |
30f4e087 | 1945 | |
28dca905 BS |
1946 | nvkm_mask(device, TPC_UNIT(0, 0, 0x05c), 0x00000001, 0x00000001); |
1947 | ||
bfee3f3d BS |
1948 | memcpy(tpcnr, gr->tpc_nr, sizeof(gr->tpc_nr)); |
1949 | for (i = 0, gpc = -1; i < gr->tpc_total; i++) { | |
30f4e087 | 1950 | do { |
bfee3f3d | 1951 | gpc = (gpc + 1) % gr->gpc_nr; |
30f4e087 | 1952 | } while (!tpcnr[gpc]); |
bfee3f3d | 1953 | tpc = gr->tpc_nr[gpc] - tpcnr[gpc]--; |
30f4e087 BS |
1954 | |
1955 | data[i / 8] |= tpc << ((i % 8) * 4); | |
1956 | } | |
1957 | ||
276836d4 BS |
1958 | nvkm_wr32(device, GPC_BCAST(0x0980), data[0]); |
1959 | nvkm_wr32(device, GPC_BCAST(0x0984), data[1]); | |
1960 | nvkm_wr32(device, GPC_BCAST(0x0988), data[2]); | |
1961 | nvkm_wr32(device, GPC_BCAST(0x098c), data[3]); | |
30f4e087 | 1962 | |
bfee3f3d | 1963 | for (gpc = 0; gpc < gr->gpc_nr; gpc++) { |
276836d4 | 1964 | nvkm_wr32(device, GPC_UNIT(gpc, 0x0914), |
5ec3def7 | 1965 | gr->screen_tile_row_offset << 8 | gr->tpc_nr[gpc]); |
276836d4 | 1966 | nvkm_wr32(device, GPC_UNIT(gpc, 0x0910), 0x00040000 | |
5ec3def7 | 1967 | gr->tpc_total); |
276836d4 | 1968 | nvkm_wr32(device, GPC_UNIT(gpc, 0x0918), magicgpc918); |
30f4e087 BS |
1969 | } |
1970 | ||
c85ee6ca | 1971 | if (device->chipset != 0xd7) |
276836d4 | 1972 | nvkm_wr32(device, GPC_BCAST(0x1bd4), magicgpc918); |
26410c67 | 1973 | else |
276836d4 | 1974 | nvkm_wr32(device, GPC_BCAST(0x3fd4), magicgpc918); |
bfee3f3d | 1975 | |
276836d4 | 1976 | nvkm_wr32(device, GPC_BCAST(0x08ac), nvkm_rd32(device, 0x100800)); |
bfee3f3d | 1977 | |
276836d4 | 1978 | nvkm_wr32(device, 0x400500, 0x00010001); |
bfee3f3d | 1979 | |
276836d4 BS |
1980 | nvkm_wr32(device, 0x400100, 0xffffffff); |
1981 | nvkm_wr32(device, 0x40013c, 0xffffffff); | |
bfee3f3d | 1982 | |
276836d4 BS |
1983 | nvkm_wr32(device, 0x409c24, 0x000f0000); |
1984 | nvkm_wr32(device, 0x404000, 0xc0000000); | |
1985 | nvkm_wr32(device, 0x404600, 0xc0000000); | |
1986 | nvkm_wr32(device, 0x408030, 0xc0000000); | |
1987 | nvkm_wr32(device, 0x40601c, 0xc0000000); | |
1988 | nvkm_wr32(device, 0x404490, 0xc0000000); | |
1989 | nvkm_wr32(device, 0x406018, 0xc0000000); | |
1990 | nvkm_wr32(device, 0x405840, 0xc0000000); | |
1991 | nvkm_wr32(device, 0x405844, 0x00ffffff); | |
1992 | nvkm_mask(device, 0x419cc0, 0x00000008, 0x00000008); | |
1993 | nvkm_mask(device, 0x419eb4, 0x00001000, 0x00001000); | |
bfee3f3d BS |
1994 | |
1995 | for (gpc = 0; gpc < gr->gpc_nr; gpc++) { | |
276836d4 BS |
1996 | nvkm_wr32(device, GPC_UNIT(gpc, 0x0420), 0xc0000000); |
1997 | nvkm_wr32(device, GPC_UNIT(gpc, 0x0900), 0xc0000000); | |
1998 | nvkm_wr32(device, GPC_UNIT(gpc, 0x1028), 0xc0000000); | |
1999 | nvkm_wr32(device, GPC_UNIT(gpc, 0x0824), 0xc0000000); | |
bfee3f3d | 2000 | for (tpc = 0; tpc < gr->tpc_nr[gpc]; tpc++) { |
276836d4 BS |
2001 | nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x508), 0xffffffff); |
2002 | nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x50c), 0xffffffff); | |
2003 | nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x224), 0xc0000000); | |
2004 | nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x48c), 0xc0000000); | |
2005 | nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x084), 0xc0000000); | |
2006 | nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x644), 0x001ffffe); | |
2007 | nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x64c), 0x0000000f); | |
30f4e087 | 2008 | } |
276836d4 BS |
2009 | nvkm_wr32(device, GPC_UNIT(gpc, 0x2c90), 0xffffffff); |
2010 | nvkm_wr32(device, GPC_UNIT(gpc, 0x2c94), 0xffffffff); | |
30f4e087 BS |
2011 | } |
2012 | ||
bfee3f3d | 2013 | for (rop = 0; rop < gr->rop_nr; rop++) { |
276836d4 BS |
2014 | nvkm_wr32(device, ROP_UNIT(rop, 0x144), 0xc0000000); |
2015 | nvkm_wr32(device, ROP_UNIT(rop, 0x070), 0xc0000000); | |
2016 | nvkm_wr32(device, ROP_UNIT(rop, 0x204), 0xffffffff); | |
2017 | nvkm_wr32(device, ROP_UNIT(rop, 0x208), 0xffffffff); | |
30f4e087 | 2018 | } |
ebb945a9 | 2019 | |
276836d4 BS |
2020 | nvkm_wr32(device, 0x400108, 0xffffffff); |
2021 | nvkm_wr32(device, 0x400138, 0xffffffff); | |
2022 | nvkm_wr32(device, 0x400118, 0xffffffff); | |
2023 | nvkm_wr32(device, 0x400130, 0xffffffff); | |
2024 | nvkm_wr32(device, 0x40011c, 0xffffffff); | |
2025 | nvkm_wr32(device, 0x400134, 0xffffffff); | |
30f4e087 | 2026 | |
276836d4 | 2027 | nvkm_wr32(device, 0x400054, 0x34ce3464); |
ac9738bb | 2028 | |
bfee3f3d | 2029 | gf100_gr_zbc_init(gr); |
ac9738bb | 2030 | |
bfee3f3d | 2031 | return gf100_gr_init_ctxctl(gr); |
30f4e087 BS |
2032 | } |
2033 | ||
e3c71eb2 | 2034 | #include "fuc/hubgf100.fuc3.h" |
30f4e087 | 2035 | |
e3c71eb2 BS |
2036 | struct gf100_gr_ucode |
2037 | gf100_gr_fecs_ucode = { | |
2038 | .code.data = gf100_grhub_code, | |
2039 | .code.size = sizeof(gf100_grhub_code), | |
2040 | .data.data = gf100_grhub_data, | |
2041 | .data.size = sizeof(gf100_grhub_data), | |
30f4e087 BS |
2042 | }; |
2043 | ||
e3c71eb2 | 2044 | #include "fuc/gpcgf100.fuc3.h" |
30f4e087 | 2045 | |
e3c71eb2 BS |
2046 | struct gf100_gr_ucode |
2047 | gf100_gr_gpccs_ucode = { | |
2048 | .code.data = gf100_grgpc_code, | |
2049 | .code.size = sizeof(gf100_grgpc_code), | |
2050 | .data.data = gf100_grgpc_data, | |
2051 | .data.size = sizeof(gf100_grgpc_data), | |
30f4e087 BS |
2052 | }; |
2053 | ||
27f3d6cf BS |
2054 | static const struct gf100_gr_func |
2055 | gf100_gr = { | |
c85ee6ca | 2056 | .init = gf100_gr_init, |
223eaf4b | 2057 | .init_gpc_mmu = gf100_gr_init_gpc_mmu, |
c85ee6ca BS |
2058 | .mmio = gf100_gr_pack_mmio, |
2059 | .fecs.ucode = &gf100_gr_fecs_ucode, | |
2060 | .gpccs.ucode = &gf100_gr_gpccs_ucode, | |
64cb5a31 | 2061 | .rops = gf100_gr_rops, |
27f3d6cf BS |
2062 | .grctx = &gf100_grctx, |
2063 | .sclass = { | |
2064 | { -1, -1, FERMI_TWOD_A }, | |
2065 | { -1, -1, FERMI_MEMORY_TO_MEMORY_FORMAT_A }, | |
2066 | { -1, -1, FERMI_A, &gf100_fermi }, | |
2067 | { -1, -1, FERMI_COMPUTE_A }, | |
2068 | {} | |
2069 | } | |
2070 | }; | |
2071 | ||
c85ee6ca BS |
2072 | int |
2073 | gf100_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) | |
2074 | { | |
2075 | return gf100_gr_new_(&gf100_gr, device, index, pgr); | |
2076 | } |