1 /*
2  * Copyright (c) 2014-2015, NVIDIA CORPORATION. All rights reserved.
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 AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
18  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
19  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
20  * DEALINGS IN THE SOFTWARE.
21  */
22 #include "gf100.h"
23 #include "ctxgf100.h"
24 
25 #include <core/firmware.h>
26 #include <subdev/timer.h>
27 
28 #include <nvif/class.h>
29 
30 struct gk20a_fw_av
31 {
32 	u32 addr;
33 	u32 data;
34 };
35 
36 static int
gk20a_gr_av_to_init(struct gf100_gr * gr,const char * path,const char * name,int ver,struct gf100_gr_pack ** ppack)37 gk20a_gr_av_to_init(struct gf100_gr *gr, const char *path, const char *name,
38 		    int ver, struct gf100_gr_pack **ppack)
39 {
40 	struct nvkm_subdev *subdev = &gr->base.engine.subdev;
41 	struct nvkm_blob blob;
42 	struct gf100_gr_init *init;
43 	struct gf100_gr_pack *pack;
44 	int nent;
45 	int ret;
46 	int i;
47 
48 	ret = nvkm_firmware_load_blob(subdev, path, name, ver, &blob);
49 	if (ret)
50 		return ret;
51 
52 	nent = (blob.size / sizeof(struct gk20a_fw_av));
53 	pack = vzalloc((sizeof(*pack) * 2) + (sizeof(*init) * (nent + 1)));
54 	if (!pack) {
55 		ret = -ENOMEM;
56 		goto end;
57 	}
58 
59 	init = (void *)(pack + 2);
60 	pack[0].init = init;
61 
62 	for (i = 0; i < nent; i++) {
63 		struct gf100_gr_init *ent = &init[i];
64 		struct gk20a_fw_av *av = &((struct gk20a_fw_av *)blob.data)[i];
65 
66 		ent->addr = av->addr;
67 		ent->data = av->data;
68 		ent->count = 1;
69 		ent->pitch = 1;
70 	}
71 
72 	*ppack = pack;
73 
74 end:
75 	nvkm_blob_dtor(&blob);
76 	return ret;
77 }
78 
79 struct gk20a_fw_aiv
80 {
81 	u32 addr;
82 	u32 index;
83 	u32 data;
84 };
85 
86 static int
gk20a_gr_aiv_to_init(struct gf100_gr * gr,const char * path,const char * name,int ver,struct gf100_gr_pack ** ppack)87 gk20a_gr_aiv_to_init(struct gf100_gr *gr, const char *path, const char *name,
88 		     int ver, struct gf100_gr_pack **ppack)
89 {
90 	struct nvkm_subdev *subdev = &gr->base.engine.subdev;
91 	struct nvkm_blob blob;
92 	struct gf100_gr_init *init;
93 	struct gf100_gr_pack *pack;
94 	int nent;
95 	int ret;
96 	int i;
97 
98 	ret = nvkm_firmware_load_blob(subdev, path, name, ver, &blob);
99 	if (ret)
100 		return ret;
101 
102 	nent = (blob.size / sizeof(struct gk20a_fw_aiv));
103 	pack = vzalloc((sizeof(*pack) * 2) + (sizeof(*init) * (nent + 1)));
104 	if (!pack) {
105 		ret = -ENOMEM;
106 		goto end;
107 	}
108 
109 	init = (void *)(pack + 2);
110 	pack[0].init = init;
111 
112 	for (i = 0; i < nent; i++) {
113 		struct gf100_gr_init *ent = &init[i];
114 		struct gk20a_fw_aiv *av = &((struct gk20a_fw_aiv *)blob.data)[i];
115 
116 		ent->addr = av->addr;
117 		ent->data = av->data;
118 		ent->count = 1;
119 		ent->pitch = 1;
120 	}
121 
122 	*ppack = pack;
123 
124 end:
125 	nvkm_blob_dtor(&blob);
126 	return ret;
127 }
128 
129 static int
gk20a_gr_av_to_method(struct gf100_gr * gr,const char * path,const char * name,int ver,struct gf100_gr_pack ** ppack)130 gk20a_gr_av_to_method(struct gf100_gr *gr, const char *path, const char *name,
131 		      int ver, struct gf100_gr_pack **ppack)
132 {
133 	struct nvkm_subdev *subdev = &gr->base.engine.subdev;
134 	struct nvkm_blob blob;
135 	struct gf100_gr_init *init;
136 	struct gf100_gr_pack *pack;
137 	/* We don't suppose we will initialize more than 16 classes here... */
138 	static const unsigned int max_classes = 16;
139 	u32 classidx = 0, prevclass = 0;
140 	int nent;
141 	int ret;
142 	int i;
143 
144 	ret = nvkm_firmware_load_blob(subdev, path, name, ver, &blob);
145 	if (ret)
146 		return ret;
147 
148 	nent = (blob.size / sizeof(struct gk20a_fw_av));
149 
150 	pack = vzalloc((sizeof(*pack) * (max_classes + 1)) +
151 		       (sizeof(*init) * (nent + max_classes + 1)));
152 	if (!pack) {
153 		ret = -ENOMEM;
154 		goto end;
155 	}
156 
157 	init = (void *)(pack + max_classes + 1);
158 
159 	for (i = 0; i < nent; i++, init++) {
160 		struct gk20a_fw_av *av = &((struct gk20a_fw_av *)blob.data)[i];
161 		u32 class = av->addr & 0xffff;
162 		u32 addr = (av->addr & 0xffff0000) >> 14;
163 
164 		if (prevclass != class) {
165 			if (prevclass) /* Add terminator to the method list. */
166 				init++;
167 			pack[classidx].init = init;
168 			pack[classidx].type = class;
169 			prevclass = class;
170 			if (++classidx >= max_classes) {
171 				vfree(pack);
172 				ret = -ENOSPC;
173 				goto end;
174 			}
175 		}
176 
177 		init->addr = addr;
178 		init->data = av->data;
179 		init->count = 1;
180 		init->pitch = 1;
181 	}
182 
183 	*ppack = pack;
184 
185 end:
186 	nvkm_blob_dtor(&blob);
187 	return ret;
188 }
189 
190 static int
gk20a_gr_wait_mem_scrubbing(struct gf100_gr * gr)191 gk20a_gr_wait_mem_scrubbing(struct gf100_gr *gr)
192 {
193 	struct nvkm_subdev *subdev = &gr->base.engine.subdev;
194 	struct nvkm_device *device = subdev->device;
195 
196 	if (nvkm_msec(device, 2000,
197 		if (!(nvkm_rd32(device, 0x40910c) & 0x00000006))
198 			break;
199 	) < 0) {
200 		nvkm_error(subdev, "FECS mem scrubbing timeout\n");
201 		return -ETIMEDOUT;
202 	}
203 
204 	if (nvkm_msec(device, 2000,
205 		if (!(nvkm_rd32(device, 0x41a10c) & 0x00000006))
206 			break;
207 	) < 0) {
208 		nvkm_error(subdev, "GPCCS mem scrubbing timeout\n");
209 		return -ETIMEDOUT;
210 	}
211 
212 	return 0;
213 }
214 
215 static void
gk20a_gr_set_hww_esr_report_mask(struct gf100_gr * gr)216 gk20a_gr_set_hww_esr_report_mask(struct gf100_gr *gr)
217 {
218 	struct nvkm_device *device = gr->base.engine.subdev.device;
219 	nvkm_wr32(device, 0x419e44, 0x1ffffe);
220 	nvkm_wr32(device, 0x419e4c, 0x7f);
221 }
222 
223 int
gk20a_gr_init(struct gf100_gr * gr)224 gk20a_gr_init(struct gf100_gr *gr)
225 {
226 	struct nvkm_device *device = gr->base.engine.subdev.device;
227 	int ret;
228 
229 	/* Clear SCC RAM */
230 	nvkm_wr32(device, 0x40802c, 0x1);
231 
232 	gf100_gr_mmio(gr, gr->sw_nonctx);
233 
234 	ret = gk20a_gr_wait_mem_scrubbing(gr);
235 	if (ret)
236 		return ret;
237 
238 	ret = gf100_gr_wait_idle(gr);
239 	if (ret)
240 		return ret;
241 
242 	/* MMU debug buffer */
243 	if (gr->func->init_gpc_mmu)
244 		gr->func->init_gpc_mmu(gr);
245 
246 	/* Set the PE as stream master */
247 	nvkm_mask(device, 0x503018, 0x1, 0x1);
248 
249 	/* Zcull init */
250 	gr->func->init_zcull(gr);
251 
252 	gr->func->init_rop_active_fbps(gr);
253 
254 	/* Enable FIFO access */
255 	nvkm_wr32(device, 0x400500, 0x00010001);
256 
257 	/* Enable interrupts */
258 	nvkm_wr32(device, 0x400100, 0xffffffff);
259 	nvkm_wr32(device, 0x40013c, 0xffffffff);
260 
261 	/* Enable FECS error interrupts */
262 	nvkm_wr32(device, 0x409c24, 0x000f0000);
263 
264 	/* Enable hardware warning exceptions */
265 	nvkm_wr32(device, 0x404000, 0xc0000000);
266 	nvkm_wr32(device, 0x404600, 0xc0000000);
267 
268 	if (gr->func->set_hww_esr_report_mask)
269 		gr->func->set_hww_esr_report_mask(gr);
270 
271 	/* Enable TPC exceptions per GPC */
272 	nvkm_wr32(device, 0x419d0c, 0x2);
273 	nvkm_wr32(device, 0x41ac94, (((1 << gr->tpc_total) - 1) & 0xff) << 16);
274 
275 	/* Reset and enable all exceptions */
276 	nvkm_wr32(device, 0x400108, 0xffffffff);
277 	nvkm_wr32(device, 0x400138, 0xffffffff);
278 	nvkm_wr32(device, 0x400118, 0xffffffff);
279 	nvkm_wr32(device, 0x400130, 0xffffffff);
280 	nvkm_wr32(device, 0x40011c, 0xffffffff);
281 	nvkm_wr32(device, 0x400134, 0xffffffff);
282 
283 	gf100_gr_zbc_init(gr);
284 
285 	return gf100_gr_init_ctxctl(gr);
286 }
287 
288 static const struct gf100_gr_func
289 gk20a_gr = {
290 	.oneinit_tiles = gf100_gr_oneinit_tiles,
291 	.oneinit_sm_id = gf100_gr_oneinit_sm_id,
292 	.init = gk20a_gr_init,
293 	.init_zcull = gf117_gr_init_zcull,
294 	.init_rop_active_fbps = gk104_gr_init_rop_active_fbps,
295 	.trap_mp = gf100_gr_trap_mp,
296 	.set_hww_esr_report_mask = gk20a_gr_set_hww_esr_report_mask,
297 	.rops = gf100_gr_rops,
298 	.ppc_nr = 1,
299 	.grctx = &gk20a_grctx,
300 	.zbc = &gf100_gr_zbc,
301 	.sclass = {
302 		{ -1, -1, FERMI_TWOD_A },
303 		{ -1, -1, KEPLER_INLINE_TO_MEMORY_A },
304 		{ -1, -1, KEPLER_C, &gf100_fermi },
305 		{ -1, -1, KEPLER_COMPUTE_A },
306 		{}
307 	}
308 };
309 
310 int
gk20a_gr_load_sw(struct gf100_gr * gr,const char * path,int ver)311 gk20a_gr_load_sw(struct gf100_gr *gr, const char *path, int ver)
312 {
313 	if (gk20a_gr_av_to_init(gr, path, "sw_nonctx", ver, &gr->sw_nonctx) ||
314 	    gk20a_gr_aiv_to_init(gr, path, "sw_ctx", ver, &gr->sw_ctx) ||
315 	    gk20a_gr_av_to_init(gr, path, "sw_bundle_init", ver, &gr->bundle) ||
316 	    gk20a_gr_av_to_method(gr, path, "sw_method_init", ver, &gr->method))
317 		return -ENOENT;
318 
319 	return 0;
320 }
321 
322 #if IS_ENABLED(CONFIG_ARCH_TEGRA_124_SOC) || IS_ENABLED(CONFIG_ARCH_TEGRA_132_SOC)
323 MODULE_FIRMWARE("nvidia/gk20a/fecs_data.bin");
324 MODULE_FIRMWARE("nvidia/gk20a/fecs_inst.bin");
325 MODULE_FIRMWARE("nvidia/gk20a/gpccs_data.bin");
326 MODULE_FIRMWARE("nvidia/gk20a/gpccs_inst.bin");
327 MODULE_FIRMWARE("nvidia/gk20a/sw_bundle_init.bin");
328 MODULE_FIRMWARE("nvidia/gk20a/sw_ctx.bin");
329 MODULE_FIRMWARE("nvidia/gk20a/sw_method_init.bin");
330 MODULE_FIRMWARE("nvidia/gk20a/sw_nonctx.bin");
331 #endif
332 
333 static int
gk20a_gr_load(struct gf100_gr * gr,int ver,const struct gf100_gr_fwif * fwif)334 gk20a_gr_load(struct gf100_gr *gr, int ver, const struct gf100_gr_fwif *fwif)
335 {
336 	struct nvkm_subdev *subdev = &gr->base.engine.subdev;
337 
338 	if (nvkm_firmware_load_blob(subdev, "", "fecs_inst", ver,
339 				    &gr->fecs.inst) ||
340 	    nvkm_firmware_load_blob(subdev, "", "fecs_data", ver,
341 				    &gr->fecs.data) ||
342 	    nvkm_firmware_load_blob(subdev, "", "gpccs_inst", ver,
343 				    &gr->gpccs.inst) ||
344 	    nvkm_firmware_load_blob(subdev, "", "gpccs_data", ver,
345 				    &gr->gpccs.data))
346 		return -ENOENT;
347 
348 	gr->firmware = true;
349 
350 	return gk20a_gr_load_sw(gr, "", ver);
351 }
352 
353 static const struct gf100_gr_fwif
354 gk20a_gr_fwif[] = {
355 	{ 0, gk20a_gr_load, &gk20a_gr },
356 	{}
357 };
358 
359 int
gk20a_gr_new(struct nvkm_device * device,enum nvkm_subdev_type type,int inst,struct nvkm_gr ** pgr)360 gk20a_gr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst, struct nvkm_gr **pgr)
361 {
362 	return gf100_gr_new_(gk20a_gr_fwif, device, type, inst, pgr);
363 }
364