Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-2.6
[pandora-kernel.git] / drivers / gpu / drm / radeon / r200.c
1 /*
2  * Copyright 2008 Advanced Micro Devices, Inc.
3  * Copyright 2008 Red Hat Inc.
4  * Copyright 2009 Jerome Glisse.
5  *
6  * Permission is hereby granted, free of charge, to any person obtaining a
7  * copy of this software and associated documentation files (the "Software"),
8  * to deal in the Software without restriction, including without limitation
9  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
10  * and/or sell copies of the Software, and to permit persons to whom the
11  * Software is furnished to do so, subject to the following conditions:
12  *
13  * The above copyright notice and this permission notice shall be included in
14  * all copies or substantial portions of the Software.
15  *
16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
19  * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
20  * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
21  * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
22  * OTHER DEALINGS IN THE SOFTWARE.
23  *
24  * Authors: Dave Airlie
25  *          Alex Deucher
26  *          Jerome Glisse
27  */
28 #include "drmP.h"
29 #include "drm.h"
30 #include "radeon_drm.h"
31 #include "radeon_reg.h"
32 #include "radeon.h"
33 #include "radeon_asic.h"
34
35 #include "r100d.h"
36 #include "r200_reg_safe.h"
37
38 #include "r100_track.h"
39
40 static int r200_get_vtx_size_0(uint32_t vtx_fmt_0)
41 {
42         int vtx_size, i;
43         vtx_size = 2;
44
45         if (vtx_fmt_0 & R200_VTX_Z0)
46                 vtx_size++;
47         if (vtx_fmt_0 & R200_VTX_W0)
48                 vtx_size++;
49         /* blend weight */
50         if (vtx_fmt_0 & (0x7 << R200_VTX_WEIGHT_COUNT_SHIFT))
51                 vtx_size += (vtx_fmt_0 >> R200_VTX_WEIGHT_COUNT_SHIFT) & 0x7;
52         if (vtx_fmt_0 & R200_VTX_PV_MATRIX_SEL)
53                 vtx_size++;
54         if (vtx_fmt_0 & R200_VTX_N0)
55                 vtx_size += 3;
56         if (vtx_fmt_0 & R200_VTX_POINT_SIZE)
57                 vtx_size++;
58         if (vtx_fmt_0 & R200_VTX_DISCRETE_FOG)
59                 vtx_size++;
60         if (vtx_fmt_0 & R200_VTX_SHININESS_0)
61                 vtx_size++;
62         if (vtx_fmt_0 & R200_VTX_SHININESS_1)
63                 vtx_size++;
64         for (i = 0; i < 8; i++) {
65                 int color_size = (vtx_fmt_0 >> (11 + 2*i)) & 0x3;
66                 switch (color_size) {
67                 case 0: break;
68                 case 1: vtx_size++; break;
69                 case 2: vtx_size += 3; break;
70                 case 3: vtx_size += 4; break;
71                 }
72         }
73         if (vtx_fmt_0 & R200_VTX_XY1)
74                 vtx_size += 2;
75         if (vtx_fmt_0 & R200_VTX_Z1)
76                 vtx_size++;
77         if (vtx_fmt_0 & R200_VTX_W1)
78                 vtx_size++;
79         if (vtx_fmt_0 & R200_VTX_N1)
80                 vtx_size += 3;
81         return vtx_size;
82 }
83
84 int r200_copy_dma(struct radeon_device *rdev,
85                   uint64_t src_offset,
86                   uint64_t dst_offset,
87                   unsigned num_pages,
88                   struct radeon_fence *fence)
89 {
90         uint32_t size;
91         uint32_t cur_size;
92         int i, num_loops;
93         int r = 0;
94
95         /* radeon pitch is /64 */
96         size = num_pages << PAGE_SHIFT;
97         num_loops = DIV_ROUND_UP(size, 0x1FFFFF);
98         r = radeon_ring_lock(rdev, num_loops * 4 + 64);
99         if (r) {
100                 DRM_ERROR("radeon: moving bo (%d).\n", r);
101                 return r;
102         }
103         /* Must wait for 2D idle & clean before DMA or hangs might happen */
104         radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0));
105         radeon_ring_write(rdev, (1 << 16));
106         for (i = 0; i < num_loops; i++) {
107                 cur_size = size;
108                 if (cur_size > 0x1FFFFF) {
109                         cur_size = 0x1FFFFF;
110                 }
111                 size -= cur_size;
112                 radeon_ring_write(rdev, PACKET0(0x720, 2));
113                 radeon_ring_write(rdev, src_offset);
114                 radeon_ring_write(rdev, dst_offset);
115                 radeon_ring_write(rdev, cur_size | (1 << 31) | (1 << 30));
116                 src_offset += cur_size;
117                 dst_offset += cur_size;
118         }
119         radeon_ring_write(rdev, PACKET0(RADEON_WAIT_UNTIL, 0));
120         radeon_ring_write(rdev, RADEON_WAIT_DMA_GUI_IDLE);
121         if (fence) {
122                 r = radeon_fence_emit(rdev, fence);
123         }
124         radeon_ring_unlock_commit(rdev);
125         return r;
126 }
127
128
129 static int r200_get_vtx_size_1(uint32_t vtx_fmt_1)
130 {
131         int vtx_size, i, tex_size;
132         vtx_size = 0;
133         for (i = 0; i < 6; i++) {
134                 tex_size = (vtx_fmt_1 >> (i * 3)) & 0x7;
135                 if (tex_size > 4)
136                         continue;
137                 vtx_size += tex_size;
138         }
139         return vtx_size;
140 }
141
142 int r200_packet0_check(struct radeon_cs_parser *p,
143                        struct radeon_cs_packet *pkt,
144                        unsigned idx, unsigned reg)
145 {
146         struct radeon_cs_reloc *reloc;
147         struct r100_cs_track *track;
148         volatile uint32_t *ib;
149         uint32_t tmp;
150         int r;
151         int i;
152         int face;
153         u32 tile_flags = 0;
154         u32 idx_value;
155
156         ib = p->ib->ptr;
157         track = (struct r100_cs_track *)p->track;
158         idx_value = radeon_get_ib_value(p, idx);
159         switch (reg) {
160         case RADEON_CRTC_GUI_TRIG_VLINE:
161                 r = r100_cs_packet_parse_vline(p);
162                 if (r) {
163                         DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
164                                   idx, reg);
165                         r100_cs_dump_packet(p, pkt);
166                         return r;
167                 }
168                 break;
169                 /* FIXME: only allow PACKET3 blit? easier to check for out of
170                  * range access */
171         case RADEON_DST_PITCH_OFFSET:
172         case RADEON_SRC_PITCH_OFFSET:
173                 r = r100_reloc_pitch_offset(p, pkt, idx, reg);
174                 if (r)
175                         return r;
176                 break;
177         case RADEON_RB3D_DEPTHOFFSET:
178                 r = r100_cs_packet_next_reloc(p, &reloc);
179                 if (r) {
180                         DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
181                                   idx, reg);
182                         r100_cs_dump_packet(p, pkt);
183                         return r;
184                 }
185                 track->zb.robj = reloc->robj;
186                 track->zb.offset = idx_value;
187                 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
188                 break;
189         case RADEON_RB3D_COLOROFFSET:
190                 r = r100_cs_packet_next_reloc(p, &reloc);
191                 if (r) {
192                         DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
193                                   idx, reg);
194                         r100_cs_dump_packet(p, pkt);
195                         return r;
196                 }
197                 track->cb[0].robj = reloc->robj;
198                 track->cb[0].offset = idx_value;
199                 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
200                 break;
201         case R200_PP_TXOFFSET_0:
202         case R200_PP_TXOFFSET_1:
203         case R200_PP_TXOFFSET_2:
204         case R200_PP_TXOFFSET_3:
205         case R200_PP_TXOFFSET_4:
206         case R200_PP_TXOFFSET_5:
207                 i = (reg - R200_PP_TXOFFSET_0) / 24;
208                 r = r100_cs_packet_next_reloc(p, &reloc);
209                 if (r) {
210                         DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
211                                   idx, reg);
212                         r100_cs_dump_packet(p, pkt);
213                         return r;
214                 }
215                 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
216                 track->textures[i].robj = reloc->robj;
217                 break;
218         case R200_PP_CUBIC_OFFSET_F1_0:
219         case R200_PP_CUBIC_OFFSET_F2_0:
220         case R200_PP_CUBIC_OFFSET_F3_0:
221         case R200_PP_CUBIC_OFFSET_F4_0:
222         case R200_PP_CUBIC_OFFSET_F5_0:
223         case R200_PP_CUBIC_OFFSET_F1_1:
224         case R200_PP_CUBIC_OFFSET_F2_1:
225         case R200_PP_CUBIC_OFFSET_F3_1:
226         case R200_PP_CUBIC_OFFSET_F4_1:
227         case R200_PP_CUBIC_OFFSET_F5_1:
228         case R200_PP_CUBIC_OFFSET_F1_2:
229         case R200_PP_CUBIC_OFFSET_F2_2:
230         case R200_PP_CUBIC_OFFSET_F3_2:
231         case R200_PP_CUBIC_OFFSET_F4_2:
232         case R200_PP_CUBIC_OFFSET_F5_2:
233         case R200_PP_CUBIC_OFFSET_F1_3:
234         case R200_PP_CUBIC_OFFSET_F2_3:
235         case R200_PP_CUBIC_OFFSET_F3_3:
236         case R200_PP_CUBIC_OFFSET_F4_3:
237         case R200_PP_CUBIC_OFFSET_F5_3:
238         case R200_PP_CUBIC_OFFSET_F1_4:
239         case R200_PP_CUBIC_OFFSET_F2_4:
240         case R200_PP_CUBIC_OFFSET_F3_4:
241         case R200_PP_CUBIC_OFFSET_F4_4:
242         case R200_PP_CUBIC_OFFSET_F5_4:
243         case R200_PP_CUBIC_OFFSET_F1_5:
244         case R200_PP_CUBIC_OFFSET_F2_5:
245         case R200_PP_CUBIC_OFFSET_F3_5:
246         case R200_PP_CUBIC_OFFSET_F4_5:
247         case R200_PP_CUBIC_OFFSET_F5_5:
248                 i = (reg - R200_PP_TXOFFSET_0) / 24;
249                 face = (reg - ((i * 24) + R200_PP_TXOFFSET_0)) / 4;
250                 r = r100_cs_packet_next_reloc(p, &reloc);
251                 if (r) {
252                         DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
253                                   idx, reg);
254                         r100_cs_dump_packet(p, pkt);
255                         return r;
256                 }
257                 track->textures[i].cube_info[face - 1].offset = idx_value;
258                 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
259                 track->textures[i].cube_info[face - 1].robj = reloc->robj;
260                 break;
261         case RADEON_RE_WIDTH_HEIGHT:
262                 track->maxy = ((idx_value >> 16) & 0x7FF);
263                 break;
264         case RADEON_RB3D_COLORPITCH:
265                 r = r100_cs_packet_next_reloc(p, &reloc);
266                 if (r) {
267                         DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
268                                   idx, reg);
269                         r100_cs_dump_packet(p, pkt);
270                         return r;
271                 }
272
273                 if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO)
274                         tile_flags |= RADEON_COLOR_TILE_ENABLE;
275                 if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO)
276                         tile_flags |= RADEON_COLOR_MICROTILE_ENABLE;
277
278                 tmp = idx_value & ~(0x7 << 16);
279                 tmp |= tile_flags;
280                 ib[idx] = tmp;
281
282                 track->cb[0].pitch = idx_value & RADEON_COLORPITCH_MASK;
283                 break;
284         case RADEON_RB3D_DEPTHPITCH:
285                 track->zb.pitch = idx_value & RADEON_DEPTHPITCH_MASK;
286                 break;
287         case RADEON_RB3D_CNTL:
288                 switch ((idx_value >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f) {
289                 case 7:
290                 case 8:
291                 case 9:
292                 case 11:
293                 case 12:
294                         track->cb[0].cpp = 1;
295                         break;
296                 case 3:
297                 case 4:
298                 case 15:
299                         track->cb[0].cpp = 2;
300                         break;
301                 case 6:
302                         track->cb[0].cpp = 4;
303                         break;
304                 default:
305                         DRM_ERROR("Invalid color buffer format (%d) !\n",
306                                   ((idx_value >> RADEON_RB3D_COLOR_FORMAT_SHIFT) & 0x1f));
307                         return -EINVAL;
308                 }
309                 if (idx_value & RADEON_DEPTHXY_OFFSET_ENABLE) {
310                         DRM_ERROR("No support for depth xy offset in kms\n");
311                         return -EINVAL;
312                 }
313
314                 track->z_enabled = !!(idx_value & RADEON_Z_ENABLE);
315                 break;
316         case RADEON_RB3D_ZSTENCILCNTL:
317                 switch (idx_value & 0xf) {
318                 case 0:
319                         track->zb.cpp = 2;
320                         break;
321                 case 2:
322                 case 3:
323                 case 4:
324                 case 5:
325                 case 9:
326                 case 11:
327                         track->zb.cpp = 4;
328                         break;
329                 default:
330                         break;
331                 }
332                 break;
333         case RADEON_RB3D_ZPASS_ADDR:
334                 r = r100_cs_packet_next_reloc(p, &reloc);
335                 if (r) {
336                         DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
337                                   idx, reg);
338                         r100_cs_dump_packet(p, pkt);
339                         return r;
340                 }
341                 ib[idx] = idx_value + ((u32)reloc->lobj.gpu_offset);
342                 break;
343         case RADEON_PP_CNTL:
344                 {
345                         uint32_t temp = idx_value >> 4;
346                         for (i = 0; i < track->num_texture; i++)
347                                 track->textures[i].enabled = !!(temp & (1 << i));
348                 }
349                 break;
350         case RADEON_SE_VF_CNTL:
351                 track->vap_vf_cntl = idx_value;
352                 break;
353         case 0x210c:
354                 /* VAP_VF_MAX_VTX_INDX */
355                 track->max_indx = idx_value & 0x00FFFFFFUL;
356                 break;
357         case R200_SE_VTX_FMT_0:
358                 track->vtx_size = r200_get_vtx_size_0(idx_value);
359                 break;
360         case R200_SE_VTX_FMT_1:
361                 track->vtx_size += r200_get_vtx_size_1(idx_value);
362                 break;
363         case R200_PP_TXSIZE_0:
364         case R200_PP_TXSIZE_1:
365         case R200_PP_TXSIZE_2:
366         case R200_PP_TXSIZE_3:
367         case R200_PP_TXSIZE_4:
368         case R200_PP_TXSIZE_5:
369                 i = (reg - R200_PP_TXSIZE_0) / 32;
370                 track->textures[i].width = (idx_value & RADEON_TEX_USIZE_MASK) + 1;
371                 track->textures[i].height = ((idx_value & RADEON_TEX_VSIZE_MASK) >> RADEON_TEX_VSIZE_SHIFT) + 1;
372                 break;
373         case R200_PP_TXPITCH_0:
374         case R200_PP_TXPITCH_1:
375         case R200_PP_TXPITCH_2:
376         case R200_PP_TXPITCH_3:
377         case R200_PP_TXPITCH_4:
378         case R200_PP_TXPITCH_5:
379                 i = (reg - R200_PP_TXPITCH_0) / 32;
380                 track->textures[i].pitch = idx_value + 32;
381                 break;
382         case R200_PP_TXFILTER_0:
383         case R200_PP_TXFILTER_1:
384         case R200_PP_TXFILTER_2:
385         case R200_PP_TXFILTER_3:
386         case R200_PP_TXFILTER_4:
387         case R200_PP_TXFILTER_5:
388                 i = (reg - R200_PP_TXFILTER_0) / 32;
389                 track->textures[i].num_levels = ((idx_value & R200_MAX_MIP_LEVEL_MASK)
390                                                  >> R200_MAX_MIP_LEVEL_SHIFT);
391                 tmp = (idx_value >> 23) & 0x7;
392                 if (tmp == 2 || tmp == 6)
393                         track->textures[i].roundup_w = false;
394                 tmp = (idx_value >> 27) & 0x7;
395                 if (tmp == 2 || tmp == 6)
396                         track->textures[i].roundup_h = false;
397                 break;
398         case R200_PP_TXMULTI_CTL_0:
399         case R200_PP_TXMULTI_CTL_1:
400         case R200_PP_TXMULTI_CTL_2:
401         case R200_PP_TXMULTI_CTL_3:
402         case R200_PP_TXMULTI_CTL_4:
403         case R200_PP_TXMULTI_CTL_5:
404                 i = (reg - R200_PP_TXMULTI_CTL_0) / 32;
405                 break;
406         case R200_PP_TXFORMAT_X_0:
407         case R200_PP_TXFORMAT_X_1:
408         case R200_PP_TXFORMAT_X_2:
409         case R200_PP_TXFORMAT_X_3:
410         case R200_PP_TXFORMAT_X_4:
411         case R200_PP_TXFORMAT_X_5:
412                 i = (reg - R200_PP_TXFORMAT_X_0) / 32;
413                 track->textures[i].txdepth = idx_value & 0x7;
414                 tmp = (idx_value >> 16) & 0x3;
415                 /* 2D, 3D, CUBE */
416                 switch (tmp) {
417                 case 0:
418                 case 3:
419                 case 4:
420                 case 5:
421                 case 6:
422                 case 7:
423                         /* 1D/2D */
424                         track->textures[i].tex_coord_type = 0;
425                         break;
426                 case 1:
427                         /* CUBE */
428                         track->textures[i].tex_coord_type = 2;
429                         break;
430                 case 2:
431                         /* 3D */
432                         track->textures[i].tex_coord_type = 1;
433                         break;
434                 }
435                 break;
436         case R200_PP_TXFORMAT_0:
437         case R200_PP_TXFORMAT_1:
438         case R200_PP_TXFORMAT_2:
439         case R200_PP_TXFORMAT_3:
440         case R200_PP_TXFORMAT_4:
441         case R200_PP_TXFORMAT_5:
442                 i = (reg - R200_PP_TXFORMAT_0) / 32;
443                 if (idx_value & R200_TXFORMAT_NON_POWER2) {
444                         track->textures[i].use_pitch = 1;
445                 } else {
446                         track->textures[i].use_pitch = 0;
447                         track->textures[i].width = 1 << ((idx_value >> RADEON_TXFORMAT_WIDTH_SHIFT) & RADEON_TXFORMAT_WIDTH_MASK);
448                         track->textures[i].height = 1 << ((idx_value >> RADEON_TXFORMAT_HEIGHT_SHIFT) & RADEON_TXFORMAT_HEIGHT_MASK);
449                 }
450                 if (idx_value & R200_TXFORMAT_LOOKUP_DISABLE)
451                         track->textures[i].lookup_disable = true;
452                 switch ((idx_value & RADEON_TXFORMAT_FORMAT_MASK)) {
453                 case R200_TXFORMAT_I8:
454                 case R200_TXFORMAT_RGB332:
455                 case R200_TXFORMAT_Y8:
456                         track->textures[i].cpp = 1;
457                         track->textures[i].compress_format = R100_TRACK_COMP_NONE;
458                         break;
459                 case R200_TXFORMAT_AI88:
460                 case R200_TXFORMAT_ARGB1555:
461                 case R200_TXFORMAT_RGB565:
462                 case R200_TXFORMAT_ARGB4444:
463                 case R200_TXFORMAT_VYUY422:
464                 case R200_TXFORMAT_YVYU422:
465                 case R200_TXFORMAT_LDVDU655:
466                 case R200_TXFORMAT_DVDU88:
467                 case R200_TXFORMAT_AVYU4444:
468                         track->textures[i].cpp = 2;
469                         track->textures[i].compress_format = R100_TRACK_COMP_NONE;
470                         break;
471                 case R200_TXFORMAT_ARGB8888:
472                 case R200_TXFORMAT_RGBA8888:
473                 case R200_TXFORMAT_ABGR8888:
474                 case R200_TXFORMAT_BGR111110:
475                 case R200_TXFORMAT_LDVDU8888:
476                         track->textures[i].cpp = 4;
477                         track->textures[i].compress_format = R100_TRACK_COMP_NONE;
478                         break;
479                 case R200_TXFORMAT_DXT1:
480                         track->textures[i].cpp = 1;
481                         track->textures[i].compress_format = R100_TRACK_COMP_DXT1;
482                         break;
483                 case R200_TXFORMAT_DXT23:
484                 case R200_TXFORMAT_DXT45:
485                         track->textures[i].cpp = 1;
486                         track->textures[i].compress_format = R100_TRACK_COMP_DXT1;
487                         break;
488                 }
489                 track->textures[i].cube_info[4].width = 1 << ((idx_value >> 16) & 0xf);
490                 track->textures[i].cube_info[4].height = 1 << ((idx_value >> 20) & 0xf);
491                 break;
492         case R200_PP_CUBIC_FACES_0:
493         case R200_PP_CUBIC_FACES_1:
494         case R200_PP_CUBIC_FACES_2:
495         case R200_PP_CUBIC_FACES_3:
496         case R200_PP_CUBIC_FACES_4:
497         case R200_PP_CUBIC_FACES_5:
498                 tmp = idx_value;
499                 i = (reg - R200_PP_CUBIC_FACES_0) / 32;
500                 for (face = 0; face < 4; face++) {
501                         track->textures[i].cube_info[face].width = 1 << ((tmp >> (face * 8)) & 0xf);
502                         track->textures[i].cube_info[face].height = 1 << ((tmp >> ((face * 8) + 4)) & 0xf);
503                 }
504                 break;
505         default:
506                 printk(KERN_ERR "Forbidden register 0x%04X in cs at %d\n",
507                        reg, idx);
508                 return -EINVAL;
509         }
510         return 0;
511 }
512
513 void r200_set_safe_registers(struct radeon_device *rdev)
514 {
515         rdev->config.r100.reg_safe_bm = r200_reg_safe_bm;
516         rdev->config.r100.reg_safe_bm_size = ARRAY_SIZE(r200_reg_safe_bm);
517 }