drm/nv20: Add Z compression support.
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / drivers / gpu / drm / nouveau / nv20_graph.c
1 #include "drmP.h"
2 #include "drm.h"
3 #include "nouveau_drv.h"
4 #include "nouveau_drm.h"
5
6 /*
7 * NV20
8 * -----
9 * There are 3 families :
10 * NV20 is 0x10de:0x020*
11 * NV25/28 is 0x10de:0x025* / 0x10de:0x028*
12 * NV2A is 0x10de:0x02A0
13 *
14 * NV30
15 * -----
16 * There are 3 families :
17 * NV30/31 is 0x10de:0x030* / 0x10de:0x031*
18 * NV34 is 0x10de:0x032*
19 * NV35/36 is 0x10de:0x033* / 0x10de:0x034*
20 *
21 * Not seen in the wild, no dumps (probably NV35) :
22 * NV37 is 0x10de:0x00fc, 0x10de:0x00fd
23 * NV38 is 0x10de:0x0333, 0x10de:0x00fe
24 *
25 */
26
27 #define NV20_GRCTX_SIZE (3580*4)
28 #define NV25_GRCTX_SIZE (3529*4)
29 #define NV2A_GRCTX_SIZE (3500*4)
30
31 #define NV30_31_GRCTX_SIZE (24392)
32 #define NV34_GRCTX_SIZE (18140)
33 #define NV35_36_GRCTX_SIZE (22396)
34
35 static int nv20_graph_register(struct drm_device *);
36 static int nv30_graph_register(struct drm_device *);
37
38 static void
39 nv20_graph_context_init(struct drm_device *dev, struct nouveau_gpuobj *ctx)
40 {
41 int i;
42
43 nv_wo32(ctx, 0x033c, 0xffff0000);
44 nv_wo32(ctx, 0x03a0, 0x0fff0000);
45 nv_wo32(ctx, 0x03a4, 0x0fff0000);
46 nv_wo32(ctx, 0x047c, 0x00000101);
47 nv_wo32(ctx, 0x0490, 0x00000111);
48 nv_wo32(ctx, 0x04a8, 0x44400000);
49 for (i = 0x04d4; i <= 0x04e0; i += 4)
50 nv_wo32(ctx, i, 0x00030303);
51 for (i = 0x04f4; i <= 0x0500; i += 4)
52 nv_wo32(ctx, i, 0x00080000);
53 for (i = 0x050c; i <= 0x0518; i += 4)
54 nv_wo32(ctx, i, 0x01012000);
55 for (i = 0x051c; i <= 0x0528; i += 4)
56 nv_wo32(ctx, i, 0x000105b8);
57 for (i = 0x052c; i <= 0x0538; i += 4)
58 nv_wo32(ctx, i, 0x00080008);
59 for (i = 0x055c; i <= 0x0598; i += 4)
60 nv_wo32(ctx, i, 0x07ff0000);
61 nv_wo32(ctx, 0x05a4, 0x4b7fffff);
62 nv_wo32(ctx, 0x05fc, 0x00000001);
63 nv_wo32(ctx, 0x0604, 0x00004000);
64 nv_wo32(ctx, 0x0610, 0x00000001);
65 nv_wo32(ctx, 0x0618, 0x00040000);
66 nv_wo32(ctx, 0x061c, 0x00010000);
67 for (i = 0x1c1c; i <= 0x248c; i += 16) {
68 nv_wo32(ctx, (i + 0), 0x10700ff9);
69 nv_wo32(ctx, (i + 4), 0x0436086c);
70 nv_wo32(ctx, (i + 8), 0x000c001b);
71 }
72 nv_wo32(ctx, 0x281c, 0x3f800000);
73 nv_wo32(ctx, 0x2830, 0x3f800000);
74 nv_wo32(ctx, 0x285c, 0x40000000);
75 nv_wo32(ctx, 0x2860, 0x3f800000);
76 nv_wo32(ctx, 0x2864, 0x3f000000);
77 nv_wo32(ctx, 0x286c, 0x40000000);
78 nv_wo32(ctx, 0x2870, 0x3f800000);
79 nv_wo32(ctx, 0x2878, 0xbf800000);
80 nv_wo32(ctx, 0x2880, 0xbf800000);
81 nv_wo32(ctx, 0x34a4, 0x000fe000);
82 nv_wo32(ctx, 0x3530, 0x000003f8);
83 nv_wo32(ctx, 0x3540, 0x002fe000);
84 for (i = 0x355c; i <= 0x3578; i += 4)
85 nv_wo32(ctx, i, 0x001c527c);
86 }
87
88 static void
89 nv25_graph_context_init(struct drm_device *dev, struct nouveau_gpuobj *ctx)
90 {
91 int i;
92
93 nv_wo32(ctx, 0x035c, 0xffff0000);
94 nv_wo32(ctx, 0x03c0, 0x0fff0000);
95 nv_wo32(ctx, 0x03c4, 0x0fff0000);
96 nv_wo32(ctx, 0x049c, 0x00000101);
97 nv_wo32(ctx, 0x04b0, 0x00000111);
98 nv_wo32(ctx, 0x04c8, 0x00000080);
99 nv_wo32(ctx, 0x04cc, 0xffff0000);
100 nv_wo32(ctx, 0x04d0, 0x00000001);
101 nv_wo32(ctx, 0x04e4, 0x44400000);
102 nv_wo32(ctx, 0x04fc, 0x4b800000);
103 for (i = 0x0510; i <= 0x051c; i += 4)
104 nv_wo32(ctx, i, 0x00030303);
105 for (i = 0x0530; i <= 0x053c; i += 4)
106 nv_wo32(ctx, i, 0x00080000);
107 for (i = 0x0548; i <= 0x0554; i += 4)
108 nv_wo32(ctx, i, 0x01012000);
109 for (i = 0x0558; i <= 0x0564; i += 4)
110 nv_wo32(ctx, i, 0x000105b8);
111 for (i = 0x0568; i <= 0x0574; i += 4)
112 nv_wo32(ctx, i, 0x00080008);
113 for (i = 0x0598; i <= 0x05d4; i += 4)
114 nv_wo32(ctx, i, 0x07ff0000);
115 nv_wo32(ctx, 0x05e0, 0x4b7fffff);
116 nv_wo32(ctx, 0x0620, 0x00000080);
117 nv_wo32(ctx, 0x0624, 0x30201000);
118 nv_wo32(ctx, 0x0628, 0x70605040);
119 nv_wo32(ctx, 0x062c, 0xb0a09080);
120 nv_wo32(ctx, 0x0630, 0xf0e0d0c0);
121 nv_wo32(ctx, 0x0664, 0x00000001);
122 nv_wo32(ctx, 0x066c, 0x00004000);
123 nv_wo32(ctx, 0x0678, 0x00000001);
124 nv_wo32(ctx, 0x0680, 0x00040000);
125 nv_wo32(ctx, 0x0684, 0x00010000);
126 for (i = 0x1b04; i <= 0x2374; i += 16) {
127 nv_wo32(ctx, (i + 0), 0x10700ff9);
128 nv_wo32(ctx, (i + 4), 0x0436086c);
129 nv_wo32(ctx, (i + 8), 0x000c001b);
130 }
131 nv_wo32(ctx, 0x2704, 0x3f800000);
132 nv_wo32(ctx, 0x2718, 0x3f800000);
133 nv_wo32(ctx, 0x2744, 0x40000000);
134 nv_wo32(ctx, 0x2748, 0x3f800000);
135 nv_wo32(ctx, 0x274c, 0x3f000000);
136 nv_wo32(ctx, 0x2754, 0x40000000);
137 nv_wo32(ctx, 0x2758, 0x3f800000);
138 nv_wo32(ctx, 0x2760, 0xbf800000);
139 nv_wo32(ctx, 0x2768, 0xbf800000);
140 nv_wo32(ctx, 0x308c, 0x000fe000);
141 nv_wo32(ctx, 0x3108, 0x000003f8);
142 nv_wo32(ctx, 0x3468, 0x002fe000);
143 for (i = 0x3484; i <= 0x34a0; i += 4)
144 nv_wo32(ctx, i, 0x001c527c);
145 }
146
147 static void
148 nv2a_graph_context_init(struct drm_device *dev, struct nouveau_gpuobj *ctx)
149 {
150 int i;
151
152 nv_wo32(ctx, 0x033c, 0xffff0000);
153 nv_wo32(ctx, 0x03a0, 0x0fff0000);
154 nv_wo32(ctx, 0x03a4, 0x0fff0000);
155 nv_wo32(ctx, 0x047c, 0x00000101);
156 nv_wo32(ctx, 0x0490, 0x00000111);
157 nv_wo32(ctx, 0x04a8, 0x44400000);
158 for (i = 0x04d4; i <= 0x04e0; i += 4)
159 nv_wo32(ctx, i, 0x00030303);
160 for (i = 0x04f4; i <= 0x0500; i += 4)
161 nv_wo32(ctx, i, 0x00080000);
162 for (i = 0x050c; i <= 0x0518; i += 4)
163 nv_wo32(ctx, i, 0x01012000);
164 for (i = 0x051c; i <= 0x0528; i += 4)
165 nv_wo32(ctx, i, 0x000105b8);
166 for (i = 0x052c; i <= 0x0538; i += 4)
167 nv_wo32(ctx, i, 0x00080008);
168 for (i = 0x055c; i <= 0x0598; i += 4)
169 nv_wo32(ctx, i, 0x07ff0000);
170 nv_wo32(ctx, 0x05a4, 0x4b7fffff);
171 nv_wo32(ctx, 0x05fc, 0x00000001);
172 nv_wo32(ctx, 0x0604, 0x00004000);
173 nv_wo32(ctx, 0x0610, 0x00000001);
174 nv_wo32(ctx, 0x0618, 0x00040000);
175 nv_wo32(ctx, 0x061c, 0x00010000);
176 for (i = 0x1a9c; i <= 0x22fc; i += 16) { /*XXX: check!! */
177 nv_wo32(ctx, (i + 0), 0x10700ff9);
178 nv_wo32(ctx, (i + 4), 0x0436086c);
179 nv_wo32(ctx, (i + 8), 0x000c001b);
180 }
181 nv_wo32(ctx, 0x269c, 0x3f800000);
182 nv_wo32(ctx, 0x26b0, 0x3f800000);
183 nv_wo32(ctx, 0x26dc, 0x40000000);
184 nv_wo32(ctx, 0x26e0, 0x3f800000);
185 nv_wo32(ctx, 0x26e4, 0x3f000000);
186 nv_wo32(ctx, 0x26ec, 0x40000000);
187 nv_wo32(ctx, 0x26f0, 0x3f800000);
188 nv_wo32(ctx, 0x26f8, 0xbf800000);
189 nv_wo32(ctx, 0x2700, 0xbf800000);
190 nv_wo32(ctx, 0x3024, 0x000fe000);
191 nv_wo32(ctx, 0x30a0, 0x000003f8);
192 nv_wo32(ctx, 0x33fc, 0x002fe000);
193 for (i = 0x341c; i <= 0x3438; i += 4)
194 nv_wo32(ctx, i, 0x001c527c);
195 }
196
197 static void
198 nv30_31_graph_context_init(struct drm_device *dev, struct nouveau_gpuobj *ctx)
199 {
200 int i;
201
202 nv_wo32(ctx, 0x0410, 0x00000101);
203 nv_wo32(ctx, 0x0424, 0x00000111);
204 nv_wo32(ctx, 0x0428, 0x00000060);
205 nv_wo32(ctx, 0x0444, 0x00000080);
206 nv_wo32(ctx, 0x0448, 0xffff0000);
207 nv_wo32(ctx, 0x044c, 0x00000001);
208 nv_wo32(ctx, 0x0460, 0x44400000);
209 nv_wo32(ctx, 0x048c, 0xffff0000);
210 for (i = 0x04e0; i < 0x04e8; i += 4)
211 nv_wo32(ctx, i, 0x0fff0000);
212 nv_wo32(ctx, 0x04ec, 0x00011100);
213 for (i = 0x0508; i < 0x0548; i += 4)
214 nv_wo32(ctx, i, 0x07ff0000);
215 nv_wo32(ctx, 0x0550, 0x4b7fffff);
216 nv_wo32(ctx, 0x058c, 0x00000080);
217 nv_wo32(ctx, 0x0590, 0x30201000);
218 nv_wo32(ctx, 0x0594, 0x70605040);
219 nv_wo32(ctx, 0x0598, 0xb8a89888);
220 nv_wo32(ctx, 0x059c, 0xf8e8d8c8);
221 nv_wo32(ctx, 0x05b0, 0xb0000000);
222 for (i = 0x0600; i < 0x0640; i += 4)
223 nv_wo32(ctx, i, 0x00010588);
224 for (i = 0x0640; i < 0x0680; i += 4)
225 nv_wo32(ctx, i, 0x00030303);
226 for (i = 0x06c0; i < 0x0700; i += 4)
227 nv_wo32(ctx, i, 0x0008aae4);
228 for (i = 0x0700; i < 0x0740; i += 4)
229 nv_wo32(ctx, i, 0x01012000);
230 for (i = 0x0740; i < 0x0780; i += 4)
231 nv_wo32(ctx, i, 0x00080008);
232 nv_wo32(ctx, 0x085c, 0x00040000);
233 nv_wo32(ctx, 0x0860, 0x00010000);
234 for (i = 0x0864; i < 0x0874; i += 4)
235 nv_wo32(ctx, i, 0x00040004);
236 for (i = 0x1f18; i <= 0x3088 ; i += 16) {
237 nv_wo32(ctx, i + 0, 0x10700ff9);
238 nv_wo32(ctx, i + 1, 0x0436086c);
239 nv_wo32(ctx, i + 2, 0x000c001b);
240 }
241 for (i = 0x30b8; i < 0x30c8; i += 4)
242 nv_wo32(ctx, i, 0x0000ffff);
243 nv_wo32(ctx, 0x344c, 0x3f800000);
244 nv_wo32(ctx, 0x3808, 0x3f800000);
245 nv_wo32(ctx, 0x381c, 0x3f800000);
246 nv_wo32(ctx, 0x3848, 0x40000000);
247 nv_wo32(ctx, 0x384c, 0x3f800000);
248 nv_wo32(ctx, 0x3850, 0x3f000000);
249 nv_wo32(ctx, 0x3858, 0x40000000);
250 nv_wo32(ctx, 0x385c, 0x3f800000);
251 nv_wo32(ctx, 0x3864, 0xbf800000);
252 nv_wo32(ctx, 0x386c, 0xbf800000);
253 }
254
255 static void
256 nv34_graph_context_init(struct drm_device *dev, struct nouveau_gpuobj *ctx)
257 {
258 int i;
259
260 nv_wo32(ctx, 0x040c, 0x01000101);
261 nv_wo32(ctx, 0x0420, 0x00000111);
262 nv_wo32(ctx, 0x0424, 0x00000060);
263 nv_wo32(ctx, 0x0440, 0x00000080);
264 nv_wo32(ctx, 0x0444, 0xffff0000);
265 nv_wo32(ctx, 0x0448, 0x00000001);
266 nv_wo32(ctx, 0x045c, 0x44400000);
267 nv_wo32(ctx, 0x0480, 0xffff0000);
268 for (i = 0x04d4; i < 0x04dc; i += 4)
269 nv_wo32(ctx, i, 0x0fff0000);
270 nv_wo32(ctx, 0x04e0, 0x00011100);
271 for (i = 0x04fc; i < 0x053c; i += 4)
272 nv_wo32(ctx, i, 0x07ff0000);
273 nv_wo32(ctx, 0x0544, 0x4b7fffff);
274 nv_wo32(ctx, 0x057c, 0x00000080);
275 nv_wo32(ctx, 0x0580, 0x30201000);
276 nv_wo32(ctx, 0x0584, 0x70605040);
277 nv_wo32(ctx, 0x0588, 0xb8a89888);
278 nv_wo32(ctx, 0x058c, 0xf8e8d8c8);
279 nv_wo32(ctx, 0x05a0, 0xb0000000);
280 for (i = 0x05f0; i < 0x0630; i += 4)
281 nv_wo32(ctx, i, 0x00010588);
282 for (i = 0x0630; i < 0x0670; i += 4)
283 nv_wo32(ctx, i, 0x00030303);
284 for (i = 0x06b0; i < 0x06f0; i += 4)
285 nv_wo32(ctx, i, 0x0008aae4);
286 for (i = 0x06f0; i < 0x0730; i += 4)
287 nv_wo32(ctx, i, 0x01012000);
288 for (i = 0x0730; i < 0x0770; i += 4)
289 nv_wo32(ctx, i, 0x00080008);
290 nv_wo32(ctx, 0x0850, 0x00040000);
291 nv_wo32(ctx, 0x0854, 0x00010000);
292 for (i = 0x0858; i < 0x0868; i += 4)
293 nv_wo32(ctx, i, 0x00040004);
294 for (i = 0x15ac; i <= 0x271c ; i += 16) {
295 nv_wo32(ctx, i + 0, 0x10700ff9);
296 nv_wo32(ctx, i + 1, 0x0436086c);
297 nv_wo32(ctx, i + 2, 0x000c001b);
298 }
299 for (i = 0x274c; i < 0x275c; i += 4)
300 nv_wo32(ctx, i, 0x0000ffff);
301 nv_wo32(ctx, 0x2ae0, 0x3f800000);
302 nv_wo32(ctx, 0x2e9c, 0x3f800000);
303 nv_wo32(ctx, 0x2eb0, 0x3f800000);
304 nv_wo32(ctx, 0x2edc, 0x40000000);
305 nv_wo32(ctx, 0x2ee0, 0x3f800000);
306 nv_wo32(ctx, 0x2ee4, 0x3f000000);
307 nv_wo32(ctx, 0x2eec, 0x40000000);
308 nv_wo32(ctx, 0x2ef0, 0x3f800000);
309 nv_wo32(ctx, 0x2ef8, 0xbf800000);
310 nv_wo32(ctx, 0x2f00, 0xbf800000);
311 }
312
313 static void
314 nv35_36_graph_context_init(struct drm_device *dev, struct nouveau_gpuobj *ctx)
315 {
316 int i;
317
318 nv_wo32(ctx, 0x040c, 0x00000101);
319 nv_wo32(ctx, 0x0420, 0x00000111);
320 nv_wo32(ctx, 0x0424, 0x00000060);
321 nv_wo32(ctx, 0x0440, 0x00000080);
322 nv_wo32(ctx, 0x0444, 0xffff0000);
323 nv_wo32(ctx, 0x0448, 0x00000001);
324 nv_wo32(ctx, 0x045c, 0x44400000);
325 nv_wo32(ctx, 0x0488, 0xffff0000);
326 for (i = 0x04dc; i < 0x04e4; i += 4)
327 nv_wo32(ctx, i, 0x0fff0000);
328 nv_wo32(ctx, 0x04e8, 0x00011100);
329 for (i = 0x0504; i < 0x0544; i += 4)
330 nv_wo32(ctx, i, 0x07ff0000);
331 nv_wo32(ctx, 0x054c, 0x4b7fffff);
332 nv_wo32(ctx, 0x0588, 0x00000080);
333 nv_wo32(ctx, 0x058c, 0x30201000);
334 nv_wo32(ctx, 0x0590, 0x70605040);
335 nv_wo32(ctx, 0x0594, 0xb8a89888);
336 nv_wo32(ctx, 0x0598, 0xf8e8d8c8);
337 nv_wo32(ctx, 0x05ac, 0xb0000000);
338 for (i = 0x0604; i < 0x0644; i += 4)
339 nv_wo32(ctx, i, 0x00010588);
340 for (i = 0x0644; i < 0x0684; i += 4)
341 nv_wo32(ctx, i, 0x00030303);
342 for (i = 0x06c4; i < 0x0704; i += 4)
343 nv_wo32(ctx, i, 0x0008aae4);
344 for (i = 0x0704; i < 0x0744; i += 4)
345 nv_wo32(ctx, i, 0x01012000);
346 for (i = 0x0744; i < 0x0784; i += 4)
347 nv_wo32(ctx, i, 0x00080008);
348 nv_wo32(ctx, 0x0860, 0x00040000);
349 nv_wo32(ctx, 0x0864, 0x00010000);
350 for (i = 0x0868; i < 0x0878; i += 4)
351 nv_wo32(ctx, i, 0x00040004);
352 for (i = 0x1f1c; i <= 0x308c ; i += 16) {
353 nv_wo32(ctx, i + 0, 0x10700ff9);
354 nv_wo32(ctx, i + 4, 0x0436086c);
355 nv_wo32(ctx, i + 8, 0x000c001b);
356 }
357 for (i = 0x30bc; i < 0x30cc; i += 4)
358 nv_wo32(ctx, i, 0x0000ffff);
359 nv_wo32(ctx, 0x3450, 0x3f800000);
360 nv_wo32(ctx, 0x380c, 0x3f800000);
361 nv_wo32(ctx, 0x3820, 0x3f800000);
362 nv_wo32(ctx, 0x384c, 0x40000000);
363 nv_wo32(ctx, 0x3850, 0x3f800000);
364 nv_wo32(ctx, 0x3854, 0x3f000000);
365 nv_wo32(ctx, 0x385c, 0x40000000);
366 nv_wo32(ctx, 0x3860, 0x3f800000);
367 nv_wo32(ctx, 0x3868, 0xbf800000);
368 nv_wo32(ctx, 0x3870, 0xbf800000);
369 }
370
371 int
372 nv20_graph_create_context(struct nouveau_channel *chan)
373 {
374 struct drm_device *dev = chan->dev;
375 struct drm_nouveau_private *dev_priv = dev->dev_private;
376 struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
377 void (*ctx_init)(struct drm_device *, struct nouveau_gpuobj *);
378 unsigned int idoffs = 0x28;
379 int ret;
380
381 switch (dev_priv->chipset) {
382 case 0x20:
383 ctx_init = nv20_graph_context_init;
384 idoffs = 0;
385 break;
386 case 0x25:
387 case 0x28:
388 ctx_init = nv25_graph_context_init;
389 break;
390 case 0x2a:
391 ctx_init = nv2a_graph_context_init;
392 idoffs = 0;
393 break;
394 case 0x30:
395 case 0x31:
396 ctx_init = nv30_31_graph_context_init;
397 break;
398 case 0x34:
399 ctx_init = nv34_graph_context_init;
400 break;
401 case 0x35:
402 case 0x36:
403 ctx_init = nv35_36_graph_context_init;
404 break;
405 default:
406 BUG_ON(1);
407 }
408
409 ret = nouveau_gpuobj_new(dev, chan, pgraph->grctx_size, 16,
410 NVOBJ_FLAG_ZERO_ALLOC, &chan->ramin_grctx);
411 if (ret)
412 return ret;
413
414 /* Initialise default context values */
415 ctx_init(dev, chan->ramin_grctx);
416
417 /* nv20: nv_wo32(dev, chan->ramin_grctx->gpuobj, 10, chan->id<<24); */
418 nv_wo32(chan->ramin_grctx, idoffs,
419 (chan->id << 24) | 0x1); /* CTX_USER */
420
421 nv_wo32(pgraph->ctx_table, chan->id * 4, chan->ramin_grctx->pinst >> 4);
422 return 0;
423 }
424
425 void
426 nv20_graph_destroy_context(struct nouveau_channel *chan)
427 {
428 struct drm_device *dev = chan->dev;
429 struct drm_nouveau_private *dev_priv = dev->dev_private;
430 struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
431 unsigned long flags;
432
433 spin_lock_irqsave(&dev_priv->context_switch_lock, flags);
434 pgraph->fifo_access(dev, false);
435
436 /* Unload the context if it's the currently active one */
437 if (pgraph->channel(dev) == chan)
438 pgraph->unload_context(dev);
439
440 pgraph->fifo_access(dev, true);
441 spin_unlock_irqrestore(&dev_priv->context_switch_lock, flags);
442
443 /* Free the context resources */
444 nv_wo32(pgraph->ctx_table, chan->id * 4, 0);
445 nouveau_gpuobj_ref(NULL, &chan->ramin_grctx);
446 }
447
448 int
449 nv20_graph_load_context(struct nouveau_channel *chan)
450 {
451 struct drm_device *dev = chan->dev;
452 uint32_t inst;
453
454 if (!chan->ramin_grctx)
455 return -EINVAL;
456 inst = chan->ramin_grctx->pinst >> 4;
457
458 nv_wr32(dev, NV20_PGRAPH_CHANNEL_CTX_POINTER, inst);
459 nv_wr32(dev, NV20_PGRAPH_CHANNEL_CTX_XFER,
460 NV20_PGRAPH_CHANNEL_CTX_XFER_LOAD);
461 nv_wr32(dev, NV10_PGRAPH_CTX_CONTROL, 0x10010100);
462
463 nouveau_wait_for_idle(dev);
464 return 0;
465 }
466
467 int
468 nv20_graph_unload_context(struct drm_device *dev)
469 {
470 struct drm_nouveau_private *dev_priv = dev->dev_private;
471 struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
472 struct nouveau_fifo_engine *pfifo = &dev_priv->engine.fifo;
473 struct nouveau_channel *chan;
474 uint32_t inst, tmp;
475
476 chan = pgraph->channel(dev);
477 if (!chan)
478 return 0;
479 inst = chan->ramin_grctx->pinst >> 4;
480
481 nv_wr32(dev, NV20_PGRAPH_CHANNEL_CTX_POINTER, inst);
482 nv_wr32(dev, NV20_PGRAPH_CHANNEL_CTX_XFER,
483 NV20_PGRAPH_CHANNEL_CTX_XFER_SAVE);
484
485 nouveau_wait_for_idle(dev);
486
487 nv_wr32(dev, NV10_PGRAPH_CTX_CONTROL, 0x10000000);
488 tmp = nv_rd32(dev, NV10_PGRAPH_CTX_USER) & 0x00ffffff;
489 tmp |= (pfifo->channels - 1) << 24;
490 nv_wr32(dev, NV10_PGRAPH_CTX_USER, tmp);
491 return 0;
492 }
493
494 static void
495 nv20_graph_rdi(struct drm_device *dev)
496 {
497 struct drm_nouveau_private *dev_priv = dev->dev_private;
498 int i, writecount = 32;
499 uint32_t rdi_index = 0x2c80000;
500
501 if (dev_priv->chipset == 0x20) {
502 rdi_index = 0x3d0000;
503 writecount = 15;
504 }
505
506 nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, rdi_index);
507 for (i = 0; i < writecount; i++)
508 nv_wr32(dev, NV10_PGRAPH_RDI_DATA, 0);
509
510 nouveau_wait_for_idle(dev);
511 }
512
513 void
514 nv20_graph_set_tile_region(struct drm_device *dev, int i)
515 {
516 struct drm_nouveau_private *dev_priv = dev->dev_private;
517 struct nouveau_tile_reg *tile = &dev_priv->tile.reg[i];
518
519 nv_wr32(dev, NV20_PGRAPH_TLIMIT(i), tile->limit);
520 nv_wr32(dev, NV20_PGRAPH_TSIZE(i), tile->pitch);
521 nv_wr32(dev, NV20_PGRAPH_TILE(i), tile->addr);
522
523 nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA0030 + 4 * i);
524 nv_wr32(dev, NV10_PGRAPH_RDI_DATA, tile->limit);
525 nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA0050 + 4 * i);
526 nv_wr32(dev, NV10_PGRAPH_RDI_DATA, tile->pitch);
527 nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA0010 + 4 * i);
528 nv_wr32(dev, NV10_PGRAPH_RDI_DATA, tile->addr);
529
530 if (dev_priv->card_type == NV_20) {
531 nv_wr32(dev, NV20_PGRAPH_ZCOMP(i), tile->zcomp);
532 nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00ea0090 + 4 * i);
533 nv_wr32(dev, NV10_PGRAPH_RDI_DATA, tile->zcomp);
534 }
535 }
536
537 int
538 nv20_graph_init(struct drm_device *dev)
539 {
540 struct drm_nouveau_private *dev_priv = dev->dev_private;
541 struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
542 uint32_t tmp, vramsz;
543 int ret, i;
544
545 switch (dev_priv->chipset) {
546 case 0x20:
547 pgraph->grctx_size = NV20_GRCTX_SIZE;
548 break;
549 case 0x25:
550 case 0x28:
551 pgraph->grctx_size = NV25_GRCTX_SIZE;
552 break;
553 case 0x2a:
554 pgraph->grctx_size = NV2A_GRCTX_SIZE;
555 break;
556 default:
557 NV_ERROR(dev, "unknown chipset, disabling acceleration\n");
558 pgraph->accel_blocked = true;
559 return 0;
560 }
561
562 nv_wr32(dev, NV03_PMC_ENABLE,
563 nv_rd32(dev, NV03_PMC_ENABLE) & ~NV_PMC_ENABLE_PGRAPH);
564 nv_wr32(dev, NV03_PMC_ENABLE,
565 nv_rd32(dev, NV03_PMC_ENABLE) | NV_PMC_ENABLE_PGRAPH);
566
567 if (!pgraph->ctx_table) {
568 /* Create Context Pointer Table */
569 ret = nouveau_gpuobj_new(dev, NULL, 32 * 4, 16,
570 NVOBJ_FLAG_ZERO_ALLOC,
571 &pgraph->ctx_table);
572 if (ret)
573 return ret;
574 }
575
576 nv_wr32(dev, NV20_PGRAPH_CHANNEL_CTX_TABLE,
577 pgraph->ctx_table->pinst >> 4);
578
579 nv20_graph_rdi(dev);
580
581 ret = nv20_graph_register(dev);
582 if (ret) {
583 nouveau_gpuobj_ref(NULL, &pgraph->ctx_table);
584 return ret;
585 }
586
587 nv_wr32(dev, NV03_PGRAPH_INTR , 0xFFFFFFFF);
588 nv_wr32(dev, NV03_PGRAPH_INTR_EN, 0xFFFFFFFF);
589
590 nv_wr32(dev, NV04_PGRAPH_DEBUG_0, 0xFFFFFFFF);
591 nv_wr32(dev, NV04_PGRAPH_DEBUG_0, 0x00000000);
592 nv_wr32(dev, NV04_PGRAPH_DEBUG_1, 0x00118700);
593 nv_wr32(dev, NV04_PGRAPH_DEBUG_3, 0xF3CE0475); /* 0x4 = auto ctx switch */
594 nv_wr32(dev, NV10_PGRAPH_DEBUG_4, 0x00000000);
595 nv_wr32(dev, 0x40009C , 0x00000040);
596
597 if (dev_priv->chipset >= 0x25) {
598 nv_wr32(dev, 0x400890, 0x00a8cfff);
599 nv_wr32(dev, 0x400610, 0x304B1FB6);
600 nv_wr32(dev, 0x400B80, 0x1cbd3883);
601 nv_wr32(dev, 0x400B84, 0x44000000);
602 nv_wr32(dev, 0x400098, 0x40000080);
603 nv_wr32(dev, 0x400B88, 0x000000ff);
604
605 } else {
606 nv_wr32(dev, 0x400880, 0x0008c7df);
607 nv_wr32(dev, 0x400094, 0x00000005);
608 nv_wr32(dev, 0x400B80, 0x45eae20e);
609 nv_wr32(dev, 0x400B84, 0x24000000);
610 nv_wr32(dev, 0x400098, 0x00000040);
611 nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00E00038);
612 nv_wr32(dev, NV10_PGRAPH_RDI_DATA , 0x00000030);
613 nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00E10038);
614 nv_wr32(dev, NV10_PGRAPH_RDI_DATA , 0x00000030);
615 }
616
617 /* Turn all the tiling regions off. */
618 for (i = 0; i < NV10_PFB_TILE__SIZE; i++)
619 nv20_graph_set_tile_region(dev, i);
620
621 nv_wr32(dev, 0x4009a0, nv_rd32(dev, 0x100324));
622 nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA000C);
623 nv_wr32(dev, NV10_PGRAPH_RDI_DATA, nv_rd32(dev, 0x100324));
624
625 nv_wr32(dev, NV10_PGRAPH_CTX_CONTROL, 0x10000100);
626 nv_wr32(dev, NV10_PGRAPH_STATE , 0xFFFFFFFF);
627
628 tmp = nv_rd32(dev, NV10_PGRAPH_SURFACE) & 0x0007ff00;
629 nv_wr32(dev, NV10_PGRAPH_SURFACE, tmp);
630 tmp = nv_rd32(dev, NV10_PGRAPH_SURFACE) | 0x00020100;
631 nv_wr32(dev, NV10_PGRAPH_SURFACE, tmp);
632
633 /* begin RAM config */
634 vramsz = pci_resource_len(dev->pdev, 0) - 1;
635 nv_wr32(dev, 0x4009A4, nv_rd32(dev, NV04_PFB_CFG0));
636 nv_wr32(dev, 0x4009A8, nv_rd32(dev, NV04_PFB_CFG1));
637 nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA0000);
638 nv_wr32(dev, NV10_PGRAPH_RDI_DATA , nv_rd32(dev, NV04_PFB_CFG0));
639 nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA0004);
640 nv_wr32(dev, NV10_PGRAPH_RDI_DATA , nv_rd32(dev, NV04_PFB_CFG1));
641 nv_wr32(dev, 0x400820, 0);
642 nv_wr32(dev, 0x400824, 0);
643 nv_wr32(dev, 0x400864, vramsz - 1);
644 nv_wr32(dev, 0x400868, vramsz - 1);
645
646 /* interesting.. the below overwrites some of the tile setup above.. */
647 nv_wr32(dev, 0x400B20, 0x00000000);
648 nv_wr32(dev, 0x400B04, 0xFFFFFFFF);
649
650 nv_wr32(dev, NV03_PGRAPH_ABS_UCLIP_XMIN, 0);
651 nv_wr32(dev, NV03_PGRAPH_ABS_UCLIP_YMIN, 0);
652 nv_wr32(dev, NV03_PGRAPH_ABS_UCLIP_XMAX, 0x7fff);
653 nv_wr32(dev, NV03_PGRAPH_ABS_UCLIP_YMAX, 0x7fff);
654
655 return 0;
656 }
657
658 void
659 nv20_graph_takedown(struct drm_device *dev)
660 {
661 struct drm_nouveau_private *dev_priv = dev->dev_private;
662 struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
663
664 nouveau_gpuobj_ref(NULL, &pgraph->ctx_table);
665 }
666
667 int
668 nv30_graph_init(struct drm_device *dev)
669 {
670 struct drm_nouveau_private *dev_priv = dev->dev_private;
671 struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
672 int ret, i;
673
674 switch (dev_priv->chipset) {
675 case 0x30:
676 case 0x31:
677 pgraph->grctx_size = NV30_31_GRCTX_SIZE;
678 break;
679 case 0x34:
680 pgraph->grctx_size = NV34_GRCTX_SIZE;
681 break;
682 case 0x35:
683 case 0x36:
684 pgraph->grctx_size = NV35_36_GRCTX_SIZE;
685 break;
686 default:
687 NV_ERROR(dev, "unknown chipset, disabling acceleration\n");
688 pgraph->accel_blocked = true;
689 return 0;
690 }
691
692 nv_wr32(dev, NV03_PMC_ENABLE,
693 nv_rd32(dev, NV03_PMC_ENABLE) & ~NV_PMC_ENABLE_PGRAPH);
694 nv_wr32(dev, NV03_PMC_ENABLE,
695 nv_rd32(dev, NV03_PMC_ENABLE) | NV_PMC_ENABLE_PGRAPH);
696
697 if (!pgraph->ctx_table) {
698 /* Create Context Pointer Table */
699 ret = nouveau_gpuobj_new(dev, NULL, 32 * 4, 16,
700 NVOBJ_FLAG_ZERO_ALLOC,
701 &pgraph->ctx_table);
702 if (ret)
703 return ret;
704 }
705
706 ret = nv30_graph_register(dev);
707 if (ret) {
708 nouveau_gpuobj_ref(NULL, &pgraph->ctx_table);
709 return ret;
710 }
711
712 nv_wr32(dev, NV20_PGRAPH_CHANNEL_CTX_TABLE,
713 pgraph->ctx_table->pinst >> 4);
714
715 nv_wr32(dev, NV03_PGRAPH_INTR , 0xFFFFFFFF);
716 nv_wr32(dev, NV03_PGRAPH_INTR_EN, 0xFFFFFFFF);
717
718 nv_wr32(dev, NV04_PGRAPH_DEBUG_0, 0xFFFFFFFF);
719 nv_wr32(dev, NV04_PGRAPH_DEBUG_0, 0x00000000);
720 nv_wr32(dev, NV04_PGRAPH_DEBUG_1, 0x401287c0);
721 nv_wr32(dev, 0x400890, 0x01b463ff);
722 nv_wr32(dev, NV04_PGRAPH_DEBUG_3, 0xf2de0475);
723 nv_wr32(dev, NV10_PGRAPH_DEBUG_4, 0x00008000);
724 nv_wr32(dev, NV04_PGRAPH_LIMIT_VIOL_PIX, 0xf04bdff6);
725 nv_wr32(dev, 0x400B80, 0x1003d888);
726 nv_wr32(dev, 0x400B84, 0x0c000000);
727 nv_wr32(dev, 0x400098, 0x00000000);
728 nv_wr32(dev, 0x40009C, 0x0005ad00);
729 nv_wr32(dev, 0x400B88, 0x62ff00ff); /* suspiciously like PGRAPH_DEBUG_2 */
730 nv_wr32(dev, 0x4000a0, 0x00000000);
731 nv_wr32(dev, 0x4000a4, 0x00000008);
732 nv_wr32(dev, 0x4008a8, 0xb784a400);
733 nv_wr32(dev, 0x400ba0, 0x002f8685);
734 nv_wr32(dev, 0x400ba4, 0x00231f3f);
735 nv_wr32(dev, 0x4008a4, 0x40000020);
736
737 if (dev_priv->chipset == 0x34) {
738 nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA0004);
739 nv_wr32(dev, NV10_PGRAPH_RDI_DATA , 0x00200201);
740 nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA0008);
741 nv_wr32(dev, NV10_PGRAPH_RDI_DATA , 0x00000008);
742 nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00EA0000);
743 nv_wr32(dev, NV10_PGRAPH_RDI_DATA , 0x00000032);
744 nv_wr32(dev, NV10_PGRAPH_RDI_INDEX, 0x00E00004);
745 nv_wr32(dev, NV10_PGRAPH_RDI_DATA , 0x00000002);
746 }
747
748 nv_wr32(dev, 0x4000c0, 0x00000016);
749
750 /* Turn all the tiling regions off. */
751 for (i = 0; i < NV10_PFB_TILE__SIZE; i++)
752 nv20_graph_set_tile_region(dev, i);
753
754 nv_wr32(dev, NV10_PGRAPH_CTX_CONTROL, 0x10000100);
755 nv_wr32(dev, NV10_PGRAPH_STATE , 0xFFFFFFFF);
756 nv_wr32(dev, 0x0040075c , 0x00000001);
757
758 /* begin RAM config */
759 /* vramsz = pci_resource_len(dev->pdev, 0) - 1; */
760 nv_wr32(dev, 0x4009A4, nv_rd32(dev, NV04_PFB_CFG0));
761 nv_wr32(dev, 0x4009A8, nv_rd32(dev, NV04_PFB_CFG1));
762 if (dev_priv->chipset != 0x34) {
763 nv_wr32(dev, 0x400750, 0x00EA0000);
764 nv_wr32(dev, 0x400754, nv_rd32(dev, NV04_PFB_CFG0));
765 nv_wr32(dev, 0x400750, 0x00EA0004);
766 nv_wr32(dev, 0x400754, nv_rd32(dev, NV04_PFB_CFG1));
767 }
768
769 return 0;
770 }
771
772 static int
773 nv20_graph_register(struct drm_device *dev)
774 {
775 struct drm_nouveau_private *dev_priv = dev->dev_private;
776
777 if (dev_priv->engine.graph.registered)
778 return 0;
779
780 NVOBJ_CLASS(dev, 0x506e, SW); /* nvsw */
781 NVOBJ_CLASS(dev, 0x0030, GR); /* null */
782 NVOBJ_CLASS(dev, 0x0039, GR); /* m2mf */
783 NVOBJ_CLASS(dev, 0x004a, GR); /* gdirect */
784 NVOBJ_CLASS(dev, 0x009f, GR); /* imageblit (nv12) */
785 NVOBJ_CLASS(dev, 0x008a, GR); /* ifc */
786 NVOBJ_CLASS(dev, 0x0089, GR); /* sifm */
787 NVOBJ_CLASS(dev, 0x0062, GR); /* surf2d */
788 NVOBJ_CLASS(dev, 0x0043, GR); /* rop */
789 NVOBJ_CLASS(dev, 0x0012, GR); /* beta1 */
790 NVOBJ_CLASS(dev, 0x0072, GR); /* beta4 */
791 NVOBJ_CLASS(dev, 0x0019, GR); /* cliprect */
792 NVOBJ_CLASS(dev, 0x0044, GR); /* pattern */
793 NVOBJ_CLASS(dev, 0x009e, GR); /* swzsurf */
794 NVOBJ_CLASS(dev, 0x0096, GR); /* celcius */
795
796 /* kelvin */
797 if (dev_priv->chipset < 0x25)
798 NVOBJ_CLASS(dev, 0x0097, GR);
799 else
800 NVOBJ_CLASS(dev, 0x0597, GR);
801
802 /* nvsw */
803 NVOBJ_CLASS(dev, 0x506e, SW);
804 NVOBJ_MTHD (dev, 0x506e, 0x0500, nv04_graph_mthd_page_flip);
805
806 dev_priv->engine.graph.registered = true;
807 return 0;
808 }
809
810 static int
811 nv30_graph_register(struct drm_device *dev)
812 {
813 struct drm_nouveau_private *dev_priv = dev->dev_private;
814
815 if (dev_priv->engine.graph.registered)
816 return 0;
817
818 NVOBJ_CLASS(dev, 0x506e, SW); /* nvsw */
819 NVOBJ_CLASS(dev, 0x0030, GR); /* null */
820 NVOBJ_CLASS(dev, 0x0039, GR); /* m2mf */
821 NVOBJ_CLASS(dev, 0x004a, GR); /* gdirect */
822 NVOBJ_CLASS(dev, 0x009f, GR); /* imageblit (nv12) */
823 NVOBJ_CLASS(dev, 0x008a, GR); /* ifc */
824 NVOBJ_CLASS(dev, 0x038a, GR); /* ifc (nv30) */
825 NVOBJ_CLASS(dev, 0x0089, GR); /* sifm */
826 NVOBJ_CLASS(dev, 0x0389, GR); /* sifm (nv30) */
827 NVOBJ_CLASS(dev, 0x0062, GR); /* surf2d */
828 NVOBJ_CLASS(dev, 0x0362, GR); /* surf2d (nv30) */
829 NVOBJ_CLASS(dev, 0x0043, GR); /* rop */
830 NVOBJ_CLASS(dev, 0x0012, GR); /* beta1 */
831 NVOBJ_CLASS(dev, 0x0072, GR); /* beta4 */
832 NVOBJ_CLASS(dev, 0x0019, GR); /* cliprect */
833 NVOBJ_CLASS(dev, 0x0044, GR); /* pattern */
834 NVOBJ_CLASS(dev, 0x039e, GR); /* swzsurf */
835
836 /* rankine */
837 if (0x00000003 & (1 << (dev_priv->chipset & 0x0f)))
838 NVOBJ_CLASS(dev, 0x0397, GR);
839 else
840 if (0x00000010 & (1 << (dev_priv->chipset & 0x0f)))
841 NVOBJ_CLASS(dev, 0x0697, GR);
842 else
843 if (0x000001e0 & (1 << (dev_priv->chipset & 0x0f)))
844 NVOBJ_CLASS(dev, 0x0497, GR);
845
846 /* nvsw */
847 NVOBJ_CLASS(dev, 0x506e, SW);
848 NVOBJ_MTHD (dev, 0x506e, 0x0500, nv04_graph_mthd_page_flip);
849
850 dev_priv->engine.graph.registered = true;
851 return 0;
852 }