Pull motherboard into test branch
[linux-2.6] / drivers / char / drm / r128_state.c
1 /* r128_state.c -- State support for r128 -*- linux-c -*-
2  * Created: Thu Jan 27 02:53:43 2000 by gareth@valinux.com
3  */
4 /*
5  * Copyright 2000 VA Linux Systems, Inc., Sunnyvale, California.
6  * All Rights Reserved.
7  *
8  * Permission is hereby granted, free of charge, to any person obtaining a
9  * copy of this software and associated documentation files (the "Software"),
10  * to deal in the Software without restriction, including without limitation
11  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
12  * and/or sell copies of the Software, and to permit persons to whom the
13  * Software is furnished to do so, subject to the following conditions:
14  *
15  * The above copyright notice and this permission notice (including the next
16  * paragraph) shall be included in all copies or substantial portions of the
17  * Software.
18  *
19  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
20  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
21  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
22  * PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
23  * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
24  * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
25  * DEALINGS IN THE SOFTWARE.
26  *
27  * Authors:
28  *    Gareth Hughes <gareth@valinux.com>
29  */
30
31 #include "drmP.h"
32 #include "drm.h"
33 #include "r128_drm.h"
34 #include "r128_drv.h"
35
36 /* ================================================================
37  * CCE hardware state programming functions
38  */
39
40 static void r128_emit_clip_rects(drm_r128_private_t * dev_priv,
41                                  drm_clip_rect_t * boxes, int count)
42 {
43         u32 aux_sc_cntl = 0x00000000;
44         RING_LOCALS;
45         DRM_DEBUG("    %s\n", __FUNCTION__);
46
47         BEGIN_RING((count < 3 ? count : 3) * 5 + 2);
48
49         if (count >= 1) {
50                 OUT_RING(CCE_PACKET0(R128_AUX1_SC_LEFT, 3));
51                 OUT_RING(boxes[0].x1);
52                 OUT_RING(boxes[0].x2 - 1);
53                 OUT_RING(boxes[0].y1);
54                 OUT_RING(boxes[0].y2 - 1);
55
56                 aux_sc_cntl |= (R128_AUX1_SC_EN | R128_AUX1_SC_MODE_OR);
57         }
58         if (count >= 2) {
59                 OUT_RING(CCE_PACKET0(R128_AUX2_SC_LEFT, 3));
60                 OUT_RING(boxes[1].x1);
61                 OUT_RING(boxes[1].x2 - 1);
62                 OUT_RING(boxes[1].y1);
63                 OUT_RING(boxes[1].y2 - 1);
64
65                 aux_sc_cntl |= (R128_AUX2_SC_EN | R128_AUX2_SC_MODE_OR);
66         }
67         if (count >= 3) {
68                 OUT_RING(CCE_PACKET0(R128_AUX3_SC_LEFT, 3));
69                 OUT_RING(boxes[2].x1);
70                 OUT_RING(boxes[2].x2 - 1);
71                 OUT_RING(boxes[2].y1);
72                 OUT_RING(boxes[2].y2 - 1);
73
74                 aux_sc_cntl |= (R128_AUX3_SC_EN | R128_AUX3_SC_MODE_OR);
75         }
76
77         OUT_RING(CCE_PACKET0(R128_AUX_SC_CNTL, 0));
78         OUT_RING(aux_sc_cntl);
79
80         ADVANCE_RING();
81 }
82
83 static __inline__ void r128_emit_core(drm_r128_private_t * dev_priv)
84 {
85         drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
86         drm_r128_context_regs_t *ctx = &sarea_priv->context_state;
87         RING_LOCALS;
88         DRM_DEBUG("    %s\n", __FUNCTION__);
89
90         BEGIN_RING(2);
91
92         OUT_RING(CCE_PACKET0(R128_SCALE_3D_CNTL, 0));
93         OUT_RING(ctx->scale_3d_cntl);
94
95         ADVANCE_RING();
96 }
97
98 static __inline__ void r128_emit_context(drm_r128_private_t * dev_priv)
99 {
100         drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
101         drm_r128_context_regs_t *ctx = &sarea_priv->context_state;
102         RING_LOCALS;
103         DRM_DEBUG("    %s\n", __FUNCTION__);
104
105         BEGIN_RING(13);
106
107         OUT_RING(CCE_PACKET0(R128_DST_PITCH_OFFSET_C, 11));
108         OUT_RING(ctx->dst_pitch_offset_c);
109         OUT_RING(ctx->dp_gui_master_cntl_c);
110         OUT_RING(ctx->sc_top_left_c);
111         OUT_RING(ctx->sc_bottom_right_c);
112         OUT_RING(ctx->z_offset_c);
113         OUT_RING(ctx->z_pitch_c);
114         OUT_RING(ctx->z_sten_cntl_c);
115         OUT_RING(ctx->tex_cntl_c);
116         OUT_RING(ctx->misc_3d_state_cntl_reg);
117         OUT_RING(ctx->texture_clr_cmp_clr_c);
118         OUT_RING(ctx->texture_clr_cmp_msk_c);
119         OUT_RING(ctx->fog_color_c);
120
121         ADVANCE_RING();
122 }
123
124 static __inline__ void r128_emit_setup(drm_r128_private_t * dev_priv)
125 {
126         drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
127         drm_r128_context_regs_t *ctx = &sarea_priv->context_state;
128         RING_LOCALS;
129         DRM_DEBUG("    %s\n", __FUNCTION__);
130
131         BEGIN_RING(3);
132
133         OUT_RING(CCE_PACKET1(R128_SETUP_CNTL, R128_PM4_VC_FPU_SETUP));
134         OUT_RING(ctx->setup_cntl);
135         OUT_RING(ctx->pm4_vc_fpu_setup);
136
137         ADVANCE_RING();
138 }
139
140 static __inline__ void r128_emit_masks(drm_r128_private_t * dev_priv)
141 {
142         drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
143         drm_r128_context_regs_t *ctx = &sarea_priv->context_state;
144         RING_LOCALS;
145         DRM_DEBUG("    %s\n", __FUNCTION__);
146
147         BEGIN_RING(5);
148
149         OUT_RING(CCE_PACKET0(R128_DP_WRITE_MASK, 0));
150         OUT_RING(ctx->dp_write_mask);
151
152         OUT_RING(CCE_PACKET0(R128_STEN_REF_MASK_C, 1));
153         OUT_RING(ctx->sten_ref_mask_c);
154         OUT_RING(ctx->plane_3d_mask_c);
155
156         ADVANCE_RING();
157 }
158
159 static __inline__ void r128_emit_window(drm_r128_private_t * dev_priv)
160 {
161         drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
162         drm_r128_context_regs_t *ctx = &sarea_priv->context_state;
163         RING_LOCALS;
164         DRM_DEBUG("    %s\n", __FUNCTION__);
165
166         BEGIN_RING(2);
167
168         OUT_RING(CCE_PACKET0(R128_WINDOW_XY_OFFSET, 0));
169         OUT_RING(ctx->window_xy_offset);
170
171         ADVANCE_RING();
172 }
173
174 static __inline__ void r128_emit_tex0(drm_r128_private_t * dev_priv)
175 {
176         drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
177         drm_r128_context_regs_t *ctx = &sarea_priv->context_state;
178         drm_r128_texture_regs_t *tex = &sarea_priv->tex_state[0];
179         int i;
180         RING_LOCALS;
181         DRM_DEBUG("    %s\n", __FUNCTION__);
182
183         BEGIN_RING(7 + R128_MAX_TEXTURE_LEVELS);
184
185         OUT_RING(CCE_PACKET0(R128_PRIM_TEX_CNTL_C,
186                              2 + R128_MAX_TEXTURE_LEVELS));
187         OUT_RING(tex->tex_cntl);
188         OUT_RING(tex->tex_combine_cntl);
189         OUT_RING(ctx->tex_size_pitch_c);
190         for (i = 0; i < R128_MAX_TEXTURE_LEVELS; i++) {
191                 OUT_RING(tex->tex_offset[i]);
192         }
193
194         OUT_RING(CCE_PACKET0(R128_CONSTANT_COLOR_C, 1));
195         OUT_RING(ctx->constant_color_c);
196         OUT_RING(tex->tex_border_color);
197
198         ADVANCE_RING();
199 }
200
201 static __inline__ void r128_emit_tex1(drm_r128_private_t * dev_priv)
202 {
203         drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
204         drm_r128_texture_regs_t *tex = &sarea_priv->tex_state[1];
205         int i;
206         RING_LOCALS;
207         DRM_DEBUG("    %s\n", __FUNCTION__);
208
209         BEGIN_RING(5 + R128_MAX_TEXTURE_LEVELS);
210
211         OUT_RING(CCE_PACKET0(R128_SEC_TEX_CNTL_C, 1 + R128_MAX_TEXTURE_LEVELS));
212         OUT_RING(tex->tex_cntl);
213         OUT_RING(tex->tex_combine_cntl);
214         for (i = 0; i < R128_MAX_TEXTURE_LEVELS; i++) {
215                 OUT_RING(tex->tex_offset[i]);
216         }
217
218         OUT_RING(CCE_PACKET0(R128_SEC_TEXTURE_BORDER_COLOR_C, 0));
219         OUT_RING(tex->tex_border_color);
220
221         ADVANCE_RING();
222 }
223
224 static void r128_emit_state(drm_r128_private_t * dev_priv)
225 {
226         drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
227         unsigned int dirty = sarea_priv->dirty;
228
229         DRM_DEBUG("%s: dirty=0x%08x\n", __FUNCTION__, dirty);
230
231         if (dirty & R128_UPLOAD_CORE) {
232                 r128_emit_core(dev_priv);
233                 sarea_priv->dirty &= ~R128_UPLOAD_CORE;
234         }
235
236         if (dirty & R128_UPLOAD_CONTEXT) {
237                 r128_emit_context(dev_priv);
238                 sarea_priv->dirty &= ~R128_UPLOAD_CONTEXT;
239         }
240
241         if (dirty & R128_UPLOAD_SETUP) {
242                 r128_emit_setup(dev_priv);
243                 sarea_priv->dirty &= ~R128_UPLOAD_SETUP;
244         }
245
246         if (dirty & R128_UPLOAD_MASKS) {
247                 r128_emit_masks(dev_priv);
248                 sarea_priv->dirty &= ~R128_UPLOAD_MASKS;
249         }
250
251         if (dirty & R128_UPLOAD_WINDOW) {
252                 r128_emit_window(dev_priv);
253                 sarea_priv->dirty &= ~R128_UPLOAD_WINDOW;
254         }
255
256         if (dirty & R128_UPLOAD_TEX0) {
257                 r128_emit_tex0(dev_priv);
258                 sarea_priv->dirty &= ~R128_UPLOAD_TEX0;
259         }
260
261         if (dirty & R128_UPLOAD_TEX1) {
262                 r128_emit_tex1(dev_priv);
263                 sarea_priv->dirty &= ~R128_UPLOAD_TEX1;
264         }
265
266         /* Turn off the texture cache flushing */
267         sarea_priv->context_state.tex_cntl_c &= ~R128_TEX_CACHE_FLUSH;
268
269         sarea_priv->dirty &= ~R128_REQUIRE_QUIESCENCE;
270 }
271
272 #if R128_PERFORMANCE_BOXES
273 /* ================================================================
274  * Performance monitoring functions
275  */
276
277 static void r128_clear_box(drm_r128_private_t * dev_priv,
278                            int x, int y, int w, int h, int r, int g, int b)
279 {
280         u32 pitch, offset;
281         u32 fb_bpp, color;
282         RING_LOCALS;
283
284         switch (dev_priv->fb_bpp) {
285         case 16:
286                 fb_bpp = R128_GMC_DST_16BPP;
287                 color = (((r & 0xf8) << 8) |
288                          ((g & 0xfc) << 3) | ((b & 0xf8) >> 3));
289                 break;
290         case 24:
291                 fb_bpp = R128_GMC_DST_24BPP;
292                 color = ((r << 16) | (g << 8) | b);
293                 break;
294         case 32:
295                 fb_bpp = R128_GMC_DST_32BPP;
296                 color = (((0xff) << 24) | (r << 16) | (g << 8) | b);
297                 break;
298         default:
299                 return;
300         }
301
302         offset = dev_priv->back_offset;
303         pitch = dev_priv->back_pitch >> 3;
304
305         BEGIN_RING(6);
306
307         OUT_RING(CCE_PACKET3(R128_CNTL_PAINT_MULTI, 4));
308         OUT_RING(R128_GMC_DST_PITCH_OFFSET_CNTL |
309                  R128_GMC_BRUSH_SOLID_COLOR |
310                  fb_bpp |
311                  R128_GMC_SRC_DATATYPE_COLOR |
312                  R128_ROP3_P |
313                  R128_GMC_CLR_CMP_CNTL_DIS | R128_GMC_AUX_CLIP_DIS);
314
315         OUT_RING((pitch << 21) | (offset >> 5));
316         OUT_RING(color);
317
318         OUT_RING((x << 16) | y);
319         OUT_RING((w << 16) | h);
320
321         ADVANCE_RING();
322 }
323
324 static void r128_cce_performance_boxes(drm_r128_private_t * dev_priv)
325 {
326         if (atomic_read(&dev_priv->idle_count) == 0) {
327                 r128_clear_box(dev_priv, 64, 4, 8, 8, 0, 255, 0);
328         } else {
329                 atomic_set(&dev_priv->idle_count, 0);
330         }
331 }
332
333 #endif
334
335 /* ================================================================
336  * CCE command dispatch functions
337  */
338
339 static void r128_print_dirty(const char *msg, unsigned int flags)
340 {
341         DRM_INFO("%s: (0x%x) %s%s%s%s%s%s%s%s%s\n",
342                  msg,
343                  flags,
344                  (flags & R128_UPLOAD_CORE) ? "core, " : "",
345                  (flags & R128_UPLOAD_CONTEXT) ? "context, " : "",
346                  (flags & R128_UPLOAD_SETUP) ? "setup, " : "",
347                  (flags & R128_UPLOAD_TEX0) ? "tex0, " : "",
348                  (flags & R128_UPLOAD_TEX1) ? "tex1, " : "",
349                  (flags & R128_UPLOAD_MASKS) ? "masks, " : "",
350                  (flags & R128_UPLOAD_WINDOW) ? "window, " : "",
351                  (flags & R128_UPLOAD_CLIPRECTS) ? "cliprects, " : "",
352                  (flags & R128_REQUIRE_QUIESCENCE) ? "quiescence, " : "");
353 }
354
355 static void r128_cce_dispatch_clear(drm_device_t * dev,
356                                     drm_r128_clear_t * clear)
357 {
358         drm_r128_private_t *dev_priv = dev->dev_private;
359         drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
360         int nbox = sarea_priv->nbox;
361         drm_clip_rect_t *pbox = sarea_priv->boxes;
362         unsigned int flags = clear->flags;
363         int i;
364         RING_LOCALS;
365         DRM_DEBUG("%s\n", __FUNCTION__);
366
367         if (dev_priv->page_flipping && dev_priv->current_page == 1) {
368                 unsigned int tmp = flags;
369
370                 flags &= ~(R128_FRONT | R128_BACK);
371                 if (tmp & R128_FRONT)
372                         flags |= R128_BACK;
373                 if (tmp & R128_BACK)
374                         flags |= R128_FRONT;
375         }
376
377         for (i = 0; i < nbox; i++) {
378                 int x = pbox[i].x1;
379                 int y = pbox[i].y1;
380                 int w = pbox[i].x2 - x;
381                 int h = pbox[i].y2 - y;
382
383                 DRM_DEBUG("dispatch clear %d,%d-%d,%d flags 0x%x\n",
384                           pbox[i].x1, pbox[i].y1, pbox[i].x2,
385                           pbox[i].y2, flags);
386
387                 if (flags & (R128_FRONT | R128_BACK)) {
388                         BEGIN_RING(2);
389
390                         OUT_RING(CCE_PACKET0(R128_DP_WRITE_MASK, 0));
391                         OUT_RING(clear->color_mask);
392
393                         ADVANCE_RING();
394                 }
395
396                 if (flags & R128_FRONT) {
397                         BEGIN_RING(6);
398
399                         OUT_RING(CCE_PACKET3(R128_CNTL_PAINT_MULTI, 4));
400                         OUT_RING(R128_GMC_DST_PITCH_OFFSET_CNTL |
401                                  R128_GMC_BRUSH_SOLID_COLOR |
402                                  (dev_priv->color_fmt << 8) |
403                                  R128_GMC_SRC_DATATYPE_COLOR |
404                                  R128_ROP3_P |
405                                  R128_GMC_CLR_CMP_CNTL_DIS |
406                                  R128_GMC_AUX_CLIP_DIS);
407
408                         OUT_RING(dev_priv->front_pitch_offset_c);
409                         OUT_RING(clear->clear_color);
410
411                         OUT_RING((x << 16) | y);
412                         OUT_RING((w << 16) | h);
413
414                         ADVANCE_RING();
415                 }
416
417                 if (flags & R128_BACK) {
418                         BEGIN_RING(6);
419
420                         OUT_RING(CCE_PACKET3(R128_CNTL_PAINT_MULTI, 4));
421                         OUT_RING(R128_GMC_DST_PITCH_OFFSET_CNTL |
422                                  R128_GMC_BRUSH_SOLID_COLOR |
423                                  (dev_priv->color_fmt << 8) |
424                                  R128_GMC_SRC_DATATYPE_COLOR |
425                                  R128_ROP3_P |
426                                  R128_GMC_CLR_CMP_CNTL_DIS |
427                                  R128_GMC_AUX_CLIP_DIS);
428
429                         OUT_RING(dev_priv->back_pitch_offset_c);
430                         OUT_RING(clear->clear_color);
431
432                         OUT_RING((x << 16) | y);
433                         OUT_RING((w << 16) | h);
434
435                         ADVANCE_RING();
436                 }
437
438                 if (flags & R128_DEPTH) {
439                         BEGIN_RING(6);
440
441                         OUT_RING(CCE_PACKET3(R128_CNTL_PAINT_MULTI, 4));
442                         OUT_RING(R128_GMC_DST_PITCH_OFFSET_CNTL |
443                                  R128_GMC_BRUSH_SOLID_COLOR |
444                                  (dev_priv->depth_fmt << 8) |
445                                  R128_GMC_SRC_DATATYPE_COLOR |
446                                  R128_ROP3_P |
447                                  R128_GMC_CLR_CMP_CNTL_DIS |
448                                  R128_GMC_AUX_CLIP_DIS | R128_GMC_WR_MSK_DIS);
449
450                         OUT_RING(dev_priv->depth_pitch_offset_c);
451                         OUT_RING(clear->clear_depth);
452
453                         OUT_RING((x << 16) | y);
454                         OUT_RING((w << 16) | h);
455
456                         ADVANCE_RING();
457                 }
458         }
459 }
460
461 static void r128_cce_dispatch_swap(drm_device_t * dev)
462 {
463         drm_r128_private_t *dev_priv = dev->dev_private;
464         drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
465         int nbox = sarea_priv->nbox;
466         drm_clip_rect_t *pbox = sarea_priv->boxes;
467         int i;
468         RING_LOCALS;
469         DRM_DEBUG("%s\n", __FUNCTION__);
470
471 #if R128_PERFORMANCE_BOXES
472         /* Do some trivial performance monitoring...
473          */
474         r128_cce_performance_boxes(dev_priv);
475 #endif
476
477         for (i = 0; i < nbox; i++) {
478                 int x = pbox[i].x1;
479                 int y = pbox[i].y1;
480                 int w = pbox[i].x2 - x;
481                 int h = pbox[i].y2 - y;
482
483                 BEGIN_RING(7);
484
485                 OUT_RING(CCE_PACKET3(R128_CNTL_BITBLT_MULTI, 5));
486                 OUT_RING(R128_GMC_SRC_PITCH_OFFSET_CNTL |
487                          R128_GMC_DST_PITCH_OFFSET_CNTL |
488                          R128_GMC_BRUSH_NONE |
489                          (dev_priv->color_fmt << 8) |
490                          R128_GMC_SRC_DATATYPE_COLOR |
491                          R128_ROP3_S |
492                          R128_DP_SRC_SOURCE_MEMORY |
493                          R128_GMC_CLR_CMP_CNTL_DIS |
494                          R128_GMC_AUX_CLIP_DIS | R128_GMC_WR_MSK_DIS);
495
496                 /* Make this work even if front & back are flipped:
497                  */
498                 if (dev_priv->current_page == 0) {
499                         OUT_RING(dev_priv->back_pitch_offset_c);
500                         OUT_RING(dev_priv->front_pitch_offset_c);
501                 } else {
502                         OUT_RING(dev_priv->front_pitch_offset_c);
503                         OUT_RING(dev_priv->back_pitch_offset_c);
504                 }
505
506                 OUT_RING((x << 16) | y);
507                 OUT_RING((x << 16) | y);
508                 OUT_RING((w << 16) | h);
509
510                 ADVANCE_RING();
511         }
512
513         /* Increment the frame counter.  The client-side 3D driver must
514          * throttle the framerate by waiting for this value before
515          * performing the swapbuffer ioctl.
516          */
517         dev_priv->sarea_priv->last_frame++;
518
519         BEGIN_RING(2);
520
521         OUT_RING(CCE_PACKET0(R128_LAST_FRAME_REG, 0));
522         OUT_RING(dev_priv->sarea_priv->last_frame);
523
524         ADVANCE_RING();
525 }
526
527 static void r128_cce_dispatch_flip(drm_device_t * dev)
528 {
529         drm_r128_private_t *dev_priv = dev->dev_private;
530         RING_LOCALS;
531         DRM_DEBUG("%s: page=%d pfCurrentPage=%d\n",
532                   __FUNCTION__,
533                   dev_priv->current_page, dev_priv->sarea_priv->pfCurrentPage);
534
535 #if R128_PERFORMANCE_BOXES
536         /* Do some trivial performance monitoring...
537          */
538         r128_cce_performance_boxes(dev_priv);
539 #endif
540
541         BEGIN_RING(4);
542
543         R128_WAIT_UNTIL_PAGE_FLIPPED();
544         OUT_RING(CCE_PACKET0(R128_CRTC_OFFSET, 0));
545
546         if (dev_priv->current_page == 0) {
547                 OUT_RING(dev_priv->back_offset);
548         } else {
549                 OUT_RING(dev_priv->front_offset);
550         }
551
552         ADVANCE_RING();
553
554         /* Increment the frame counter.  The client-side 3D driver must
555          * throttle the framerate by waiting for this value before
556          * performing the swapbuffer ioctl.
557          */
558         dev_priv->sarea_priv->last_frame++;
559         dev_priv->sarea_priv->pfCurrentPage = dev_priv->current_page =
560             1 - dev_priv->current_page;
561
562         BEGIN_RING(2);
563
564         OUT_RING(CCE_PACKET0(R128_LAST_FRAME_REG, 0));
565         OUT_RING(dev_priv->sarea_priv->last_frame);
566
567         ADVANCE_RING();
568 }
569
570 static void r128_cce_dispatch_vertex(drm_device_t * dev, drm_buf_t * buf)
571 {
572         drm_r128_private_t *dev_priv = dev->dev_private;
573         drm_r128_buf_priv_t *buf_priv = buf->dev_private;
574         drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
575         int format = sarea_priv->vc_format;
576         int offset = buf->bus_address;
577         int size = buf->used;
578         int prim = buf_priv->prim;
579         int i = 0;
580         RING_LOCALS;
581         DRM_DEBUG("buf=%d nbox=%d\n", buf->idx, sarea_priv->nbox);
582
583         if (0)
584                 r128_print_dirty("dispatch_vertex", sarea_priv->dirty);
585
586         if (buf->used) {
587                 buf_priv->dispatched = 1;
588
589                 if (sarea_priv->dirty & ~R128_UPLOAD_CLIPRECTS) {
590                         r128_emit_state(dev_priv);
591                 }
592
593                 do {
594                         /* Emit the next set of up to three cliprects */
595                         if (i < sarea_priv->nbox) {
596                                 r128_emit_clip_rects(dev_priv,
597                                                      &sarea_priv->boxes[i],
598                                                      sarea_priv->nbox - i);
599                         }
600
601                         /* Emit the vertex buffer rendering commands */
602                         BEGIN_RING(5);
603
604                         OUT_RING(CCE_PACKET3(R128_3D_RNDR_GEN_INDX_PRIM, 3));
605                         OUT_RING(offset);
606                         OUT_RING(size);
607                         OUT_RING(format);
608                         OUT_RING(prim | R128_CCE_VC_CNTL_PRIM_WALK_LIST |
609                                  (size << R128_CCE_VC_CNTL_NUM_SHIFT));
610
611                         ADVANCE_RING();
612
613                         i += 3;
614                 } while (i < sarea_priv->nbox);
615         }
616
617         if (buf_priv->discard) {
618                 buf_priv->age = dev_priv->sarea_priv->last_dispatch;
619
620                 /* Emit the vertex buffer age */
621                 BEGIN_RING(2);
622
623                 OUT_RING(CCE_PACKET0(R128_LAST_DISPATCH_REG, 0));
624                 OUT_RING(buf_priv->age);
625
626                 ADVANCE_RING();
627
628                 buf->pending = 1;
629                 buf->used = 0;
630                 /* FIXME: Check dispatched field */
631                 buf_priv->dispatched = 0;
632         }
633
634         dev_priv->sarea_priv->last_dispatch++;
635
636         sarea_priv->dirty &= ~R128_UPLOAD_CLIPRECTS;
637         sarea_priv->nbox = 0;
638 }
639
640 static void r128_cce_dispatch_indirect(drm_device_t * dev,
641                                        drm_buf_t * buf, int start, int end)
642 {
643         drm_r128_private_t *dev_priv = dev->dev_private;
644         drm_r128_buf_priv_t *buf_priv = buf->dev_private;
645         RING_LOCALS;
646         DRM_DEBUG("indirect: buf=%d s=0x%x e=0x%x\n", buf->idx, start, end);
647
648         if (start != end) {
649                 int offset = buf->bus_address + start;
650                 int dwords = (end - start + 3) / sizeof(u32);
651
652                 /* Indirect buffer data must be an even number of
653                  * dwords, so if we've been given an odd number we must
654                  * pad the data with a Type-2 CCE packet.
655                  */
656                 if (dwords & 1) {
657                         u32 *data = (u32 *)
658                             ((char *)dev->agp_buffer_map->handle
659                              + buf->offset + start);
660                         data[dwords++] = cpu_to_le32(R128_CCE_PACKET2);
661                 }
662
663                 buf_priv->dispatched = 1;
664
665                 /* Fire off the indirect buffer */
666                 BEGIN_RING(3);
667
668                 OUT_RING(CCE_PACKET0(R128_PM4_IW_INDOFF, 1));
669                 OUT_RING(offset);
670                 OUT_RING(dwords);
671
672                 ADVANCE_RING();
673         }
674
675         if (buf_priv->discard) {
676                 buf_priv->age = dev_priv->sarea_priv->last_dispatch;
677
678                 /* Emit the indirect buffer age */
679                 BEGIN_RING(2);
680
681                 OUT_RING(CCE_PACKET0(R128_LAST_DISPATCH_REG, 0));
682                 OUT_RING(buf_priv->age);
683
684                 ADVANCE_RING();
685
686                 buf->pending = 1;
687                 buf->used = 0;
688                 /* FIXME: Check dispatched field */
689                 buf_priv->dispatched = 0;
690         }
691
692         dev_priv->sarea_priv->last_dispatch++;
693 }
694
695 static void r128_cce_dispatch_indices(drm_device_t * dev,
696                                       drm_buf_t * buf,
697                                       int start, int end, int count)
698 {
699         drm_r128_private_t *dev_priv = dev->dev_private;
700         drm_r128_buf_priv_t *buf_priv = buf->dev_private;
701         drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
702         int format = sarea_priv->vc_format;
703         int offset = dev->agp_buffer_map->offset - dev_priv->cce_buffers_offset;
704         int prim = buf_priv->prim;
705         u32 *data;
706         int dwords;
707         int i = 0;
708         RING_LOCALS;
709         DRM_DEBUG("indices: s=%d e=%d c=%d\n", start, end, count);
710
711         if (0)
712                 r128_print_dirty("dispatch_indices", sarea_priv->dirty);
713
714         if (start != end) {
715                 buf_priv->dispatched = 1;
716
717                 if (sarea_priv->dirty & ~R128_UPLOAD_CLIPRECTS) {
718                         r128_emit_state(dev_priv);
719                 }
720
721                 dwords = (end - start + 3) / sizeof(u32);
722
723                 data = (u32 *) ((char *)dev->agp_buffer_map->handle
724                                 + buf->offset + start);
725
726                 data[0] = cpu_to_le32(CCE_PACKET3(R128_3D_RNDR_GEN_INDX_PRIM,
727                                                   dwords - 2));
728
729                 data[1] = cpu_to_le32(offset);
730                 data[2] = cpu_to_le32(R128_MAX_VB_VERTS);
731                 data[3] = cpu_to_le32(format);
732                 data[4] = cpu_to_le32((prim | R128_CCE_VC_CNTL_PRIM_WALK_IND |
733                                        (count << 16)));
734
735                 if (count & 0x1) {
736 #ifdef __LITTLE_ENDIAN
737                         data[dwords - 1] &= 0x0000ffff;
738 #else
739                         data[dwords - 1] &= 0xffff0000;
740 #endif
741                 }
742
743                 do {
744                         /* Emit the next set of up to three cliprects */
745                         if (i < sarea_priv->nbox) {
746                                 r128_emit_clip_rects(dev_priv,
747                                                      &sarea_priv->boxes[i],
748                                                      sarea_priv->nbox - i);
749                         }
750
751                         r128_cce_dispatch_indirect(dev, buf, start, end);
752
753                         i += 3;
754                 } while (i < sarea_priv->nbox);
755         }
756
757         if (buf_priv->discard) {
758                 buf_priv->age = dev_priv->sarea_priv->last_dispatch;
759
760                 /* Emit the vertex buffer age */
761                 BEGIN_RING(2);
762
763                 OUT_RING(CCE_PACKET0(R128_LAST_DISPATCH_REG, 0));
764                 OUT_RING(buf_priv->age);
765
766                 ADVANCE_RING();
767
768                 buf->pending = 1;
769                 /* FIXME: Check dispatched field */
770                 buf_priv->dispatched = 0;
771         }
772
773         dev_priv->sarea_priv->last_dispatch++;
774
775         sarea_priv->dirty &= ~R128_UPLOAD_CLIPRECTS;
776         sarea_priv->nbox = 0;
777 }
778
779 static int r128_cce_dispatch_blit(DRMFILE filp,
780                                   drm_device_t * dev, drm_r128_blit_t * blit)
781 {
782         drm_r128_private_t *dev_priv = dev->dev_private;
783         drm_device_dma_t *dma = dev->dma;
784         drm_buf_t *buf;
785         drm_r128_buf_priv_t *buf_priv;
786         u32 *data;
787         int dword_shift, dwords;
788         RING_LOCALS;
789         DRM_DEBUG("\n");
790
791         /* The compiler won't optimize away a division by a variable,
792          * even if the only legal values are powers of two.  Thus, we'll
793          * use a shift instead.
794          */
795         switch (blit->format) {
796         case R128_DATATYPE_ARGB8888:
797                 dword_shift = 0;
798                 break;
799         case R128_DATATYPE_ARGB1555:
800         case R128_DATATYPE_RGB565:
801         case R128_DATATYPE_ARGB4444:
802         case R128_DATATYPE_YVYU422:
803         case R128_DATATYPE_VYUY422:
804                 dword_shift = 1;
805                 break;
806         case R128_DATATYPE_CI8:
807         case R128_DATATYPE_RGB8:
808                 dword_shift = 2;
809                 break;
810         default:
811                 DRM_ERROR("invalid blit format %d\n", blit->format);
812                 return DRM_ERR(EINVAL);
813         }
814
815         /* Flush the pixel cache, and mark the contents as Read Invalid.
816          * This ensures no pixel data gets mixed up with the texture
817          * data from the host data blit, otherwise part of the texture
818          * image may be corrupted.
819          */
820         BEGIN_RING(2);
821
822         OUT_RING(CCE_PACKET0(R128_PC_GUI_CTLSTAT, 0));
823         OUT_RING(R128_PC_RI_GUI | R128_PC_FLUSH_GUI);
824
825         ADVANCE_RING();
826
827         /* Dispatch the indirect buffer.
828          */
829         buf = dma->buflist[blit->idx];
830         buf_priv = buf->dev_private;
831
832         if (buf->filp != filp) {
833                 DRM_ERROR("process %d using buffer owned by %p\n",
834                           DRM_CURRENTPID, buf->filp);
835                 return DRM_ERR(EINVAL);
836         }
837         if (buf->pending) {
838                 DRM_ERROR("sending pending buffer %d\n", blit->idx);
839                 return DRM_ERR(EINVAL);
840         }
841
842         buf_priv->discard = 1;
843
844         dwords = (blit->width * blit->height) >> dword_shift;
845
846         data = (u32 *) ((char *)dev->agp_buffer_map->handle + buf->offset);
847
848         data[0] = cpu_to_le32(CCE_PACKET3(R128_CNTL_HOSTDATA_BLT, dwords + 6));
849         data[1] = cpu_to_le32((R128_GMC_DST_PITCH_OFFSET_CNTL |
850                                R128_GMC_BRUSH_NONE |
851                                (blit->format << 8) |
852                                R128_GMC_SRC_DATATYPE_COLOR |
853                                R128_ROP3_S |
854                                R128_DP_SRC_SOURCE_HOST_DATA |
855                                R128_GMC_CLR_CMP_CNTL_DIS |
856                                R128_GMC_AUX_CLIP_DIS | R128_GMC_WR_MSK_DIS));
857
858         data[2] = cpu_to_le32((blit->pitch << 21) | (blit->offset >> 5));
859         data[3] = cpu_to_le32(0xffffffff);
860         data[4] = cpu_to_le32(0xffffffff);
861         data[5] = cpu_to_le32((blit->y << 16) | blit->x);
862         data[6] = cpu_to_le32((blit->height << 16) | blit->width);
863         data[7] = cpu_to_le32(dwords);
864
865         buf->used = (dwords + 8) * sizeof(u32);
866
867         r128_cce_dispatch_indirect(dev, buf, 0, buf->used);
868
869         /* Flush the pixel cache after the blit completes.  This ensures
870          * the texture data is written out to memory before rendering
871          * continues.
872          */
873         BEGIN_RING(2);
874
875         OUT_RING(CCE_PACKET0(R128_PC_GUI_CTLSTAT, 0));
876         OUT_RING(R128_PC_FLUSH_GUI);
877
878         ADVANCE_RING();
879
880         return 0;
881 }
882
883 /* ================================================================
884  * Tiled depth buffer management
885  *
886  * FIXME: These should all set the destination write mask for when we
887  * have hardware stencil support.
888  */
889
890 static int r128_cce_dispatch_write_span(drm_device_t * dev,
891                                         drm_r128_depth_t * depth)
892 {
893         drm_r128_private_t *dev_priv = dev->dev_private;
894         int count, x, y;
895         u32 *buffer;
896         u8 *mask;
897         int i, buffer_size, mask_size;
898         RING_LOCALS;
899         DRM_DEBUG("\n");
900
901         count = depth->n;
902         if (count > 4096 || count <= 0)
903                 return DRM_ERR(EMSGSIZE);
904
905         if (DRM_COPY_FROM_USER(&x, depth->x, sizeof(x))) {
906                 return DRM_ERR(EFAULT);
907         }
908         if (DRM_COPY_FROM_USER(&y, depth->y, sizeof(y))) {
909                 return DRM_ERR(EFAULT);
910         }
911
912         buffer_size = depth->n * sizeof(u32);
913         buffer = drm_alloc(buffer_size, DRM_MEM_BUFS);
914         if (buffer == NULL)
915                 return DRM_ERR(ENOMEM);
916         if (DRM_COPY_FROM_USER(buffer, depth->buffer, buffer_size)) {
917                 drm_free(buffer, buffer_size, DRM_MEM_BUFS);
918                 return DRM_ERR(EFAULT);
919         }
920
921         mask_size = depth->n * sizeof(u8);
922         if (depth->mask) {
923                 mask = drm_alloc(mask_size, DRM_MEM_BUFS);
924                 if (mask == NULL) {
925                         drm_free(buffer, buffer_size, DRM_MEM_BUFS);
926                         return DRM_ERR(ENOMEM);
927                 }
928                 if (DRM_COPY_FROM_USER(mask, depth->mask, mask_size)) {
929                         drm_free(buffer, buffer_size, DRM_MEM_BUFS);
930                         drm_free(mask, mask_size, DRM_MEM_BUFS);
931                         return DRM_ERR(EFAULT);
932                 }
933
934                 for (i = 0; i < count; i++, x++) {
935                         if (mask[i]) {
936                                 BEGIN_RING(6);
937
938                                 OUT_RING(CCE_PACKET3(R128_CNTL_PAINT_MULTI, 4));
939                                 OUT_RING(R128_GMC_DST_PITCH_OFFSET_CNTL |
940                                          R128_GMC_BRUSH_SOLID_COLOR |
941                                          (dev_priv->depth_fmt << 8) |
942                                          R128_GMC_SRC_DATATYPE_COLOR |
943                                          R128_ROP3_P |
944                                          R128_GMC_CLR_CMP_CNTL_DIS |
945                                          R128_GMC_WR_MSK_DIS);
946
947                                 OUT_RING(dev_priv->depth_pitch_offset_c);
948                                 OUT_RING(buffer[i]);
949
950                                 OUT_RING((x << 16) | y);
951                                 OUT_RING((1 << 16) | 1);
952
953                                 ADVANCE_RING();
954                         }
955                 }
956
957                 drm_free(mask, mask_size, DRM_MEM_BUFS);
958         } else {
959                 for (i = 0; i < count; i++, x++) {
960                         BEGIN_RING(6);
961
962                         OUT_RING(CCE_PACKET3(R128_CNTL_PAINT_MULTI, 4));
963                         OUT_RING(R128_GMC_DST_PITCH_OFFSET_CNTL |
964                                  R128_GMC_BRUSH_SOLID_COLOR |
965                                  (dev_priv->depth_fmt << 8) |
966                                  R128_GMC_SRC_DATATYPE_COLOR |
967                                  R128_ROP3_P |
968                                  R128_GMC_CLR_CMP_CNTL_DIS |
969                                  R128_GMC_WR_MSK_DIS);
970
971                         OUT_RING(dev_priv->depth_pitch_offset_c);
972                         OUT_RING(buffer[i]);
973
974                         OUT_RING((x << 16) | y);
975                         OUT_RING((1 << 16) | 1);
976
977                         ADVANCE_RING();
978                 }
979         }
980
981         drm_free(buffer, buffer_size, DRM_MEM_BUFS);
982
983         return 0;
984 }
985
986 static int r128_cce_dispatch_write_pixels(drm_device_t * dev,
987                                           drm_r128_depth_t * depth)
988 {
989         drm_r128_private_t *dev_priv = dev->dev_private;
990         int count, *x, *y;
991         u32 *buffer;
992         u8 *mask;
993         int i, xbuf_size, ybuf_size, buffer_size, mask_size;
994         RING_LOCALS;
995         DRM_DEBUG("\n");
996
997         count = depth->n;
998         if (count > 4096 || count <= 0)
999                 return DRM_ERR(EMSGSIZE);
1000
1001         xbuf_size = count * sizeof(*x);
1002         ybuf_size = count * sizeof(*y);
1003         x = drm_alloc(xbuf_size, DRM_MEM_BUFS);
1004         if (x == NULL) {
1005                 return DRM_ERR(ENOMEM);
1006         }
1007         y = drm_alloc(ybuf_size, DRM_MEM_BUFS);
1008         if (y == NULL) {
1009                 drm_free(x, xbuf_size, DRM_MEM_BUFS);
1010                 return DRM_ERR(ENOMEM);
1011         }
1012         if (DRM_COPY_FROM_USER(x, depth->x, xbuf_size)) {
1013                 drm_free(x, xbuf_size, DRM_MEM_BUFS);
1014                 drm_free(y, ybuf_size, DRM_MEM_BUFS);
1015                 return DRM_ERR(EFAULT);
1016         }
1017         if (DRM_COPY_FROM_USER(y, depth->y, xbuf_size)) {
1018                 drm_free(x, xbuf_size, DRM_MEM_BUFS);
1019                 drm_free(y, ybuf_size, DRM_MEM_BUFS);
1020                 return DRM_ERR(EFAULT);
1021         }
1022
1023         buffer_size = depth->n * sizeof(u32);
1024         buffer = drm_alloc(buffer_size, DRM_MEM_BUFS);
1025         if (buffer == NULL) {
1026                 drm_free(x, xbuf_size, DRM_MEM_BUFS);
1027                 drm_free(y, ybuf_size, DRM_MEM_BUFS);
1028                 return DRM_ERR(ENOMEM);
1029         }
1030         if (DRM_COPY_FROM_USER(buffer, depth->buffer, buffer_size)) {
1031                 drm_free(x, xbuf_size, DRM_MEM_BUFS);
1032                 drm_free(y, ybuf_size, DRM_MEM_BUFS);
1033                 drm_free(buffer, buffer_size, DRM_MEM_BUFS);
1034                 return DRM_ERR(EFAULT);
1035         }
1036
1037         if (depth->mask) {
1038                 mask_size = depth->n * sizeof(u8);
1039                 mask = drm_alloc(mask_size, DRM_MEM_BUFS);
1040                 if (mask == NULL) {
1041                         drm_free(x, xbuf_size, DRM_MEM_BUFS);
1042                         drm_free(y, ybuf_size, DRM_MEM_BUFS);
1043                         drm_free(buffer, buffer_size, DRM_MEM_BUFS);
1044                         return DRM_ERR(ENOMEM);
1045                 }
1046                 if (DRM_COPY_FROM_USER(mask, depth->mask, mask_size)) {
1047                         drm_free(x, xbuf_size, DRM_MEM_BUFS);
1048                         drm_free(y, ybuf_size, DRM_MEM_BUFS);
1049                         drm_free(buffer, buffer_size, DRM_MEM_BUFS);
1050                         drm_free(mask, mask_size, DRM_MEM_BUFS);
1051                         return DRM_ERR(EFAULT);
1052                 }
1053
1054                 for (i = 0; i < count; i++) {
1055                         if (mask[i]) {
1056                                 BEGIN_RING(6);
1057
1058                                 OUT_RING(CCE_PACKET3(R128_CNTL_PAINT_MULTI, 4));
1059                                 OUT_RING(R128_GMC_DST_PITCH_OFFSET_CNTL |
1060                                          R128_GMC_BRUSH_SOLID_COLOR |
1061                                          (dev_priv->depth_fmt << 8) |
1062                                          R128_GMC_SRC_DATATYPE_COLOR |
1063                                          R128_ROP3_P |
1064                                          R128_GMC_CLR_CMP_CNTL_DIS |
1065                                          R128_GMC_WR_MSK_DIS);
1066
1067                                 OUT_RING(dev_priv->depth_pitch_offset_c);
1068                                 OUT_RING(buffer[i]);
1069
1070                                 OUT_RING((x[i] << 16) | y[i]);
1071                                 OUT_RING((1 << 16) | 1);
1072
1073                                 ADVANCE_RING();
1074                         }
1075                 }
1076
1077                 drm_free(mask, mask_size, DRM_MEM_BUFS);
1078         } else {
1079                 for (i = 0; i < count; i++) {
1080                         BEGIN_RING(6);
1081
1082                         OUT_RING(CCE_PACKET3(R128_CNTL_PAINT_MULTI, 4));
1083                         OUT_RING(R128_GMC_DST_PITCH_OFFSET_CNTL |
1084                                  R128_GMC_BRUSH_SOLID_COLOR |
1085                                  (dev_priv->depth_fmt << 8) |
1086                                  R128_GMC_SRC_DATATYPE_COLOR |
1087                                  R128_ROP3_P |
1088                                  R128_GMC_CLR_CMP_CNTL_DIS |
1089                                  R128_GMC_WR_MSK_DIS);
1090
1091                         OUT_RING(dev_priv->depth_pitch_offset_c);
1092                         OUT_RING(buffer[i]);
1093
1094                         OUT_RING((x[i] << 16) | y[i]);
1095                         OUT_RING((1 << 16) | 1);
1096
1097                         ADVANCE_RING();
1098                 }
1099         }
1100
1101         drm_free(x, xbuf_size, DRM_MEM_BUFS);
1102         drm_free(y, ybuf_size, DRM_MEM_BUFS);
1103         drm_free(buffer, buffer_size, DRM_MEM_BUFS);
1104
1105         return 0;
1106 }
1107
1108 static int r128_cce_dispatch_read_span(drm_device_t * dev,
1109                                        drm_r128_depth_t * depth)
1110 {
1111         drm_r128_private_t *dev_priv = dev->dev_private;
1112         int count, x, y;
1113         RING_LOCALS;
1114         DRM_DEBUG("\n");
1115
1116         count = depth->n;
1117         if (count > 4096 || count <= 0)
1118                 return DRM_ERR(EMSGSIZE);
1119
1120         if (DRM_COPY_FROM_USER(&x, depth->x, sizeof(x))) {
1121                 return DRM_ERR(EFAULT);
1122         }
1123         if (DRM_COPY_FROM_USER(&y, depth->y, sizeof(y))) {
1124                 return DRM_ERR(EFAULT);
1125         }
1126
1127         BEGIN_RING(7);
1128
1129         OUT_RING(CCE_PACKET3(R128_CNTL_BITBLT_MULTI, 5));
1130         OUT_RING(R128_GMC_SRC_PITCH_OFFSET_CNTL |
1131                  R128_GMC_DST_PITCH_OFFSET_CNTL |
1132                  R128_GMC_BRUSH_NONE |
1133                  (dev_priv->depth_fmt << 8) |
1134                  R128_GMC_SRC_DATATYPE_COLOR |
1135                  R128_ROP3_S |
1136                  R128_DP_SRC_SOURCE_MEMORY |
1137                  R128_GMC_CLR_CMP_CNTL_DIS | R128_GMC_WR_MSK_DIS);
1138
1139         OUT_RING(dev_priv->depth_pitch_offset_c);
1140         OUT_RING(dev_priv->span_pitch_offset_c);
1141
1142         OUT_RING((x << 16) | y);
1143         OUT_RING((0 << 16) | 0);
1144         OUT_RING((count << 16) | 1);
1145
1146         ADVANCE_RING();
1147
1148         return 0;
1149 }
1150
1151 static int r128_cce_dispatch_read_pixels(drm_device_t * dev,
1152                                          drm_r128_depth_t * depth)
1153 {
1154         drm_r128_private_t *dev_priv = dev->dev_private;
1155         int count, *x, *y;
1156         int i, xbuf_size, ybuf_size;
1157         RING_LOCALS;
1158         DRM_DEBUG("%s\n", __FUNCTION__);
1159
1160         count = depth->n;
1161         if (count > 4096 || count <= 0)
1162                 return DRM_ERR(EMSGSIZE);
1163
1164         if (count > dev_priv->depth_pitch) {
1165                 count = dev_priv->depth_pitch;
1166         }
1167
1168         xbuf_size = count * sizeof(*x);
1169         ybuf_size = count * sizeof(*y);
1170         x = drm_alloc(xbuf_size, DRM_MEM_BUFS);
1171         if (x == NULL) {
1172                 return DRM_ERR(ENOMEM);
1173         }
1174         y = drm_alloc(ybuf_size, DRM_MEM_BUFS);
1175         if (y == NULL) {
1176                 drm_free(x, xbuf_size, DRM_MEM_BUFS);
1177                 return DRM_ERR(ENOMEM);
1178         }
1179         if (DRM_COPY_FROM_USER(x, depth->x, xbuf_size)) {
1180                 drm_free(x, xbuf_size, DRM_MEM_BUFS);
1181                 drm_free(y, ybuf_size, DRM_MEM_BUFS);
1182                 return DRM_ERR(EFAULT);
1183         }
1184         if (DRM_COPY_FROM_USER(y, depth->y, ybuf_size)) {
1185                 drm_free(x, xbuf_size, DRM_MEM_BUFS);
1186                 drm_free(y, ybuf_size, DRM_MEM_BUFS);
1187                 return DRM_ERR(EFAULT);
1188         }
1189
1190         for (i = 0; i < count; i++) {
1191                 BEGIN_RING(7);
1192
1193                 OUT_RING(CCE_PACKET3(R128_CNTL_BITBLT_MULTI, 5));
1194                 OUT_RING(R128_GMC_SRC_PITCH_OFFSET_CNTL |
1195                          R128_GMC_DST_PITCH_OFFSET_CNTL |
1196                          R128_GMC_BRUSH_NONE |
1197                          (dev_priv->depth_fmt << 8) |
1198                          R128_GMC_SRC_DATATYPE_COLOR |
1199                          R128_ROP3_S |
1200                          R128_DP_SRC_SOURCE_MEMORY |
1201                          R128_GMC_CLR_CMP_CNTL_DIS | R128_GMC_WR_MSK_DIS);
1202
1203                 OUT_RING(dev_priv->depth_pitch_offset_c);
1204                 OUT_RING(dev_priv->span_pitch_offset_c);
1205
1206                 OUT_RING((x[i] << 16) | y[i]);
1207                 OUT_RING((i << 16) | 0);
1208                 OUT_RING((1 << 16) | 1);
1209
1210                 ADVANCE_RING();
1211         }
1212
1213         drm_free(x, xbuf_size, DRM_MEM_BUFS);
1214         drm_free(y, ybuf_size, DRM_MEM_BUFS);
1215
1216         return 0;
1217 }
1218
1219 /* ================================================================
1220  * Polygon stipple
1221  */
1222
1223 static void r128_cce_dispatch_stipple(drm_device_t * dev, u32 * stipple)
1224 {
1225         drm_r128_private_t *dev_priv = dev->dev_private;
1226         int i;
1227         RING_LOCALS;
1228         DRM_DEBUG("%s\n", __FUNCTION__);
1229
1230         BEGIN_RING(33);
1231
1232         OUT_RING(CCE_PACKET0(R128_BRUSH_DATA0, 31));
1233         for (i = 0; i < 32; i++) {
1234                 OUT_RING(stipple[i]);
1235         }
1236
1237         ADVANCE_RING();
1238 }
1239
1240 /* ================================================================
1241  * IOCTL functions
1242  */
1243
1244 static int r128_cce_clear(DRM_IOCTL_ARGS)
1245 {
1246         DRM_DEVICE;
1247         drm_r128_private_t *dev_priv = dev->dev_private;
1248         drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
1249         drm_r128_clear_t clear;
1250         DRM_DEBUG("\n");
1251
1252         LOCK_TEST_WITH_RETURN(dev, filp);
1253
1254         DRM_COPY_FROM_USER_IOCTL(clear, (drm_r128_clear_t __user *) data,
1255                                  sizeof(clear));
1256
1257         RING_SPACE_TEST_WITH_RETURN(dev_priv);
1258
1259         if (sarea_priv->nbox > R128_NR_SAREA_CLIPRECTS)
1260                 sarea_priv->nbox = R128_NR_SAREA_CLIPRECTS;
1261
1262         r128_cce_dispatch_clear(dev, &clear);
1263         COMMIT_RING();
1264
1265         /* Make sure we restore the 3D state next time.
1266          */
1267         dev_priv->sarea_priv->dirty |= R128_UPLOAD_CONTEXT | R128_UPLOAD_MASKS;
1268
1269         return 0;
1270 }
1271
1272 static int r128_do_init_pageflip(drm_device_t * dev)
1273 {
1274         drm_r128_private_t *dev_priv = dev->dev_private;
1275         DRM_DEBUG("\n");
1276
1277         dev_priv->crtc_offset = R128_READ(R128_CRTC_OFFSET);
1278         dev_priv->crtc_offset_cntl = R128_READ(R128_CRTC_OFFSET_CNTL);
1279
1280         R128_WRITE(R128_CRTC_OFFSET, dev_priv->front_offset);
1281         R128_WRITE(R128_CRTC_OFFSET_CNTL,
1282                    dev_priv->crtc_offset_cntl | R128_CRTC_OFFSET_FLIP_CNTL);
1283
1284         dev_priv->page_flipping = 1;
1285         dev_priv->current_page = 0;
1286         dev_priv->sarea_priv->pfCurrentPage = dev_priv->current_page;
1287
1288         return 0;
1289 }
1290
1291 static int r128_do_cleanup_pageflip(drm_device_t * dev)
1292 {
1293         drm_r128_private_t *dev_priv = dev->dev_private;
1294         DRM_DEBUG("\n");
1295
1296         R128_WRITE(R128_CRTC_OFFSET, dev_priv->crtc_offset);
1297         R128_WRITE(R128_CRTC_OFFSET_CNTL, dev_priv->crtc_offset_cntl);
1298
1299         if (dev_priv->current_page != 0) {
1300                 r128_cce_dispatch_flip(dev);
1301                 COMMIT_RING();
1302         }
1303
1304         dev_priv->page_flipping = 0;
1305         return 0;
1306 }
1307
1308 /* Swapping and flipping are different operations, need different ioctls.
1309  * They can & should be intermixed to support multiple 3d windows.
1310  */
1311
1312 static int r128_cce_flip(DRM_IOCTL_ARGS)
1313 {
1314         DRM_DEVICE;
1315         drm_r128_private_t *dev_priv = dev->dev_private;
1316         DRM_DEBUG("%s\n", __FUNCTION__);
1317
1318         LOCK_TEST_WITH_RETURN(dev, filp);
1319
1320         RING_SPACE_TEST_WITH_RETURN(dev_priv);
1321
1322         if (!dev_priv->page_flipping)
1323                 r128_do_init_pageflip(dev);
1324
1325         r128_cce_dispatch_flip(dev);
1326
1327         COMMIT_RING();
1328         return 0;
1329 }
1330
1331 static int r128_cce_swap(DRM_IOCTL_ARGS)
1332 {
1333         DRM_DEVICE;
1334         drm_r128_private_t *dev_priv = dev->dev_private;
1335         drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
1336         DRM_DEBUG("%s\n", __FUNCTION__);
1337
1338         LOCK_TEST_WITH_RETURN(dev, filp);
1339
1340         RING_SPACE_TEST_WITH_RETURN(dev_priv);
1341
1342         if (sarea_priv->nbox > R128_NR_SAREA_CLIPRECTS)
1343                 sarea_priv->nbox = R128_NR_SAREA_CLIPRECTS;
1344
1345         r128_cce_dispatch_swap(dev);
1346         dev_priv->sarea_priv->dirty |= (R128_UPLOAD_CONTEXT |
1347                                         R128_UPLOAD_MASKS);
1348
1349         COMMIT_RING();
1350         return 0;
1351 }
1352
1353 static int r128_cce_vertex(DRM_IOCTL_ARGS)
1354 {
1355         DRM_DEVICE;
1356         drm_r128_private_t *dev_priv = dev->dev_private;
1357         drm_device_dma_t *dma = dev->dma;
1358         drm_buf_t *buf;
1359         drm_r128_buf_priv_t *buf_priv;
1360         drm_r128_vertex_t vertex;
1361
1362         LOCK_TEST_WITH_RETURN(dev, filp);
1363
1364         if (!dev_priv) {
1365                 DRM_ERROR("%s called with no initialization\n", __FUNCTION__);
1366                 return DRM_ERR(EINVAL);
1367         }
1368
1369         DRM_COPY_FROM_USER_IOCTL(vertex, (drm_r128_vertex_t __user *) data,
1370                                  sizeof(vertex));
1371
1372         DRM_DEBUG("pid=%d index=%d count=%d discard=%d\n",
1373                   DRM_CURRENTPID, vertex.idx, vertex.count, vertex.discard);
1374
1375         if (vertex.idx < 0 || vertex.idx >= dma->buf_count) {
1376                 DRM_ERROR("buffer index %d (of %d max)\n",
1377                           vertex.idx, dma->buf_count - 1);
1378                 return DRM_ERR(EINVAL);
1379         }
1380         if (vertex.prim < 0 ||
1381             vertex.prim > R128_CCE_VC_CNTL_PRIM_TYPE_TRI_TYPE2) {
1382                 DRM_ERROR("buffer prim %d\n", vertex.prim);
1383                 return DRM_ERR(EINVAL);
1384         }
1385
1386         RING_SPACE_TEST_WITH_RETURN(dev_priv);
1387         VB_AGE_TEST_WITH_RETURN(dev_priv);
1388
1389         buf = dma->buflist[vertex.idx];
1390         buf_priv = buf->dev_private;
1391
1392         if (buf->filp != filp) {
1393                 DRM_ERROR("process %d using buffer owned by %p\n",
1394                           DRM_CURRENTPID, buf->filp);
1395                 return DRM_ERR(EINVAL);
1396         }
1397         if (buf->pending) {
1398                 DRM_ERROR("sending pending buffer %d\n", vertex.idx);
1399                 return DRM_ERR(EINVAL);
1400         }
1401
1402         buf->used = vertex.count;
1403         buf_priv->prim = vertex.prim;
1404         buf_priv->discard = vertex.discard;
1405
1406         r128_cce_dispatch_vertex(dev, buf);
1407
1408         COMMIT_RING();
1409         return 0;
1410 }
1411
1412 static int r128_cce_indices(DRM_IOCTL_ARGS)
1413 {
1414         DRM_DEVICE;
1415         drm_r128_private_t *dev_priv = dev->dev_private;
1416         drm_device_dma_t *dma = dev->dma;
1417         drm_buf_t *buf;
1418         drm_r128_buf_priv_t *buf_priv;
1419         drm_r128_indices_t elts;
1420         int count;
1421
1422         LOCK_TEST_WITH_RETURN(dev, filp);
1423
1424         if (!dev_priv) {
1425                 DRM_ERROR("%s called with no initialization\n", __FUNCTION__);
1426                 return DRM_ERR(EINVAL);
1427         }
1428
1429         DRM_COPY_FROM_USER_IOCTL(elts, (drm_r128_indices_t __user *) data,
1430                                  sizeof(elts));
1431
1432         DRM_DEBUG("pid=%d buf=%d s=%d e=%d d=%d\n", DRM_CURRENTPID,
1433                   elts.idx, elts.start, elts.end, elts.discard);
1434
1435         if (elts.idx < 0 || elts.idx >= dma->buf_count) {
1436                 DRM_ERROR("buffer index %d (of %d max)\n",
1437                           elts.idx, dma->buf_count - 1);
1438                 return DRM_ERR(EINVAL);
1439         }
1440         if (elts.prim < 0 || elts.prim > R128_CCE_VC_CNTL_PRIM_TYPE_TRI_TYPE2) {
1441                 DRM_ERROR("buffer prim %d\n", elts.prim);
1442                 return DRM_ERR(EINVAL);
1443         }
1444
1445         RING_SPACE_TEST_WITH_RETURN(dev_priv);
1446         VB_AGE_TEST_WITH_RETURN(dev_priv);
1447
1448         buf = dma->buflist[elts.idx];
1449         buf_priv = buf->dev_private;
1450
1451         if (buf->filp != filp) {
1452                 DRM_ERROR("process %d using buffer owned by %p\n",
1453                           DRM_CURRENTPID, buf->filp);
1454                 return DRM_ERR(EINVAL);
1455         }
1456         if (buf->pending) {
1457                 DRM_ERROR("sending pending buffer %d\n", elts.idx);
1458                 return DRM_ERR(EINVAL);
1459         }
1460
1461         count = (elts.end - elts.start) / sizeof(u16);
1462         elts.start -= R128_INDEX_PRIM_OFFSET;
1463
1464         if (elts.start & 0x7) {
1465                 DRM_ERROR("misaligned buffer 0x%x\n", elts.start);
1466                 return DRM_ERR(EINVAL);
1467         }
1468         if (elts.start < buf->used) {
1469                 DRM_ERROR("no header 0x%x - 0x%x\n", elts.start, buf->used);
1470                 return DRM_ERR(EINVAL);
1471         }
1472
1473         buf->used = elts.end;
1474         buf_priv->prim = elts.prim;
1475         buf_priv->discard = elts.discard;
1476
1477         r128_cce_dispatch_indices(dev, buf, elts.start, elts.end, count);
1478
1479         COMMIT_RING();
1480         return 0;
1481 }
1482
1483 static int r128_cce_blit(DRM_IOCTL_ARGS)
1484 {
1485         DRM_DEVICE;
1486         drm_device_dma_t *dma = dev->dma;
1487         drm_r128_private_t *dev_priv = dev->dev_private;
1488         drm_r128_blit_t blit;
1489         int ret;
1490
1491         LOCK_TEST_WITH_RETURN(dev, filp);
1492
1493         DRM_COPY_FROM_USER_IOCTL(blit, (drm_r128_blit_t __user *) data,
1494                                  sizeof(blit));
1495
1496         DRM_DEBUG("pid=%d index=%d\n", DRM_CURRENTPID, blit.idx);
1497
1498         if (blit.idx < 0 || blit.idx >= dma->buf_count) {
1499                 DRM_ERROR("buffer index %d (of %d max)\n",
1500                           blit.idx, dma->buf_count - 1);
1501                 return DRM_ERR(EINVAL);
1502         }
1503
1504         RING_SPACE_TEST_WITH_RETURN(dev_priv);
1505         VB_AGE_TEST_WITH_RETURN(dev_priv);
1506
1507         ret = r128_cce_dispatch_blit(filp, dev, &blit);
1508
1509         COMMIT_RING();
1510         return ret;
1511 }
1512
1513 static int r128_cce_depth(DRM_IOCTL_ARGS)
1514 {
1515         DRM_DEVICE;
1516         drm_r128_private_t *dev_priv = dev->dev_private;
1517         drm_r128_depth_t depth;
1518         int ret;
1519
1520         LOCK_TEST_WITH_RETURN(dev, filp);
1521
1522         DRM_COPY_FROM_USER_IOCTL(depth, (drm_r128_depth_t __user *) data,
1523                                  sizeof(depth));
1524
1525         RING_SPACE_TEST_WITH_RETURN(dev_priv);
1526
1527         ret = DRM_ERR(EINVAL);
1528         switch (depth.func) {
1529         case R128_WRITE_SPAN:
1530                 ret = r128_cce_dispatch_write_span(dev, &depth);
1531                 break;
1532         case R128_WRITE_PIXELS:
1533                 ret = r128_cce_dispatch_write_pixels(dev, &depth);
1534                 break;
1535         case R128_READ_SPAN:
1536                 ret = r128_cce_dispatch_read_span(dev, &depth);
1537                 break;
1538         case R128_READ_PIXELS:
1539                 ret = r128_cce_dispatch_read_pixels(dev, &depth);
1540                 break;
1541         }
1542
1543         COMMIT_RING();
1544         return ret;
1545 }
1546
1547 static int r128_cce_stipple(DRM_IOCTL_ARGS)
1548 {
1549         DRM_DEVICE;
1550         drm_r128_private_t *dev_priv = dev->dev_private;
1551         drm_r128_stipple_t stipple;
1552         u32 mask[32];
1553
1554         LOCK_TEST_WITH_RETURN(dev, filp);
1555
1556         DRM_COPY_FROM_USER_IOCTL(stipple, (drm_r128_stipple_t __user *) data,
1557                                  sizeof(stipple));
1558
1559         if (DRM_COPY_FROM_USER(&mask, stipple.mask, 32 * sizeof(u32)))
1560                 return DRM_ERR(EFAULT);
1561
1562         RING_SPACE_TEST_WITH_RETURN(dev_priv);
1563
1564         r128_cce_dispatch_stipple(dev, mask);
1565
1566         COMMIT_RING();
1567         return 0;
1568 }
1569
1570 static int r128_cce_indirect(DRM_IOCTL_ARGS)
1571 {
1572         DRM_DEVICE;
1573         drm_r128_private_t *dev_priv = dev->dev_private;
1574         drm_device_dma_t *dma = dev->dma;
1575         drm_buf_t *buf;
1576         drm_r128_buf_priv_t *buf_priv;
1577         drm_r128_indirect_t indirect;
1578 #if 0
1579         RING_LOCALS;
1580 #endif
1581
1582         LOCK_TEST_WITH_RETURN(dev, filp);
1583
1584         if (!dev_priv) {
1585                 DRM_ERROR("%s called with no initialization\n", __FUNCTION__);
1586                 return DRM_ERR(EINVAL);
1587         }
1588
1589         DRM_COPY_FROM_USER_IOCTL(indirect, (drm_r128_indirect_t __user *) data,
1590                                  sizeof(indirect));
1591
1592         DRM_DEBUG("indirect: idx=%d s=%d e=%d d=%d\n",
1593                   indirect.idx, indirect.start, indirect.end, indirect.discard);
1594
1595         if (indirect.idx < 0 || indirect.idx >= dma->buf_count) {
1596                 DRM_ERROR("buffer index %d (of %d max)\n",
1597                           indirect.idx, dma->buf_count - 1);
1598                 return DRM_ERR(EINVAL);
1599         }
1600
1601         buf = dma->buflist[indirect.idx];
1602         buf_priv = buf->dev_private;
1603
1604         if (buf->filp != filp) {
1605                 DRM_ERROR("process %d using buffer owned by %p\n",
1606                           DRM_CURRENTPID, buf->filp);
1607                 return DRM_ERR(EINVAL);
1608         }
1609         if (buf->pending) {
1610                 DRM_ERROR("sending pending buffer %d\n", indirect.idx);
1611                 return DRM_ERR(EINVAL);
1612         }
1613
1614         if (indirect.start < buf->used) {
1615                 DRM_ERROR("reusing indirect: start=0x%x actual=0x%x\n",
1616                           indirect.start, buf->used);
1617                 return DRM_ERR(EINVAL);
1618         }
1619
1620         RING_SPACE_TEST_WITH_RETURN(dev_priv);
1621         VB_AGE_TEST_WITH_RETURN(dev_priv);
1622
1623         buf->used = indirect.end;
1624         buf_priv->discard = indirect.discard;
1625
1626 #if 0
1627         /* Wait for the 3D stream to idle before the indirect buffer
1628          * containing 2D acceleration commands is processed.
1629          */
1630         BEGIN_RING(2);
1631         RADEON_WAIT_UNTIL_3D_IDLE();
1632         ADVANCE_RING();
1633 #endif
1634
1635         /* Dispatch the indirect buffer full of commands from the
1636          * X server.  This is insecure and is thus only available to
1637          * privileged clients.
1638          */
1639         r128_cce_dispatch_indirect(dev, buf, indirect.start, indirect.end);
1640
1641         COMMIT_RING();
1642         return 0;
1643 }
1644
1645 static int r128_getparam(DRM_IOCTL_ARGS)
1646 {
1647         DRM_DEVICE;
1648         drm_r128_private_t *dev_priv = dev->dev_private;
1649         drm_r128_getparam_t param;
1650         int value;
1651
1652         if (!dev_priv) {
1653                 DRM_ERROR("%s called with no initialization\n", __FUNCTION__);
1654                 return DRM_ERR(EINVAL);
1655         }
1656
1657         DRM_COPY_FROM_USER_IOCTL(param, (drm_r128_getparam_t __user *) data,
1658                                  sizeof(param));
1659
1660         DRM_DEBUG("pid=%d\n", DRM_CURRENTPID);
1661
1662         switch (param.param) {
1663         case R128_PARAM_IRQ_NR:
1664                 value = dev->irq;
1665                 break;
1666         default:
1667                 return DRM_ERR(EINVAL);
1668         }
1669
1670         if (DRM_COPY_TO_USER(param.value, &value, sizeof(int))) {
1671                 DRM_ERROR("copy_to_user\n");
1672                 return DRM_ERR(EFAULT);
1673         }
1674
1675         return 0;
1676 }
1677
1678 void r128_driver_preclose(drm_device_t * dev, DRMFILE filp)
1679 {
1680         if (dev->dev_private) {
1681                 drm_r128_private_t *dev_priv = dev->dev_private;
1682                 if (dev_priv->page_flipping) {
1683                         r128_do_cleanup_pageflip(dev);
1684                 }
1685         }
1686 }
1687
1688 void r128_driver_lastclose(drm_device_t * dev)
1689 {
1690         r128_do_cleanup_cce(dev);
1691 }
1692
1693 drm_ioctl_desc_t r128_ioctls[] = {
1694         [DRM_IOCTL_NR(DRM_R128_INIT)] = {r128_cce_init, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY},
1695         [DRM_IOCTL_NR(DRM_R128_CCE_START)] = {r128_cce_start, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY},
1696         [DRM_IOCTL_NR(DRM_R128_CCE_STOP)] = {r128_cce_stop, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY},
1697         [DRM_IOCTL_NR(DRM_R128_CCE_RESET)] = {r128_cce_reset, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY},
1698         [DRM_IOCTL_NR(DRM_R128_CCE_IDLE)] = {r128_cce_idle, DRM_AUTH},
1699         [DRM_IOCTL_NR(DRM_R128_RESET)] = {r128_engine_reset, DRM_AUTH},
1700         [DRM_IOCTL_NR(DRM_R128_FULLSCREEN)] = {r128_fullscreen, DRM_AUTH},
1701         [DRM_IOCTL_NR(DRM_R128_SWAP)] = {r128_cce_swap, DRM_AUTH},
1702         [DRM_IOCTL_NR(DRM_R128_FLIP)] = {r128_cce_flip, DRM_AUTH},
1703         [DRM_IOCTL_NR(DRM_R128_CLEAR)] = {r128_cce_clear, DRM_AUTH},
1704         [DRM_IOCTL_NR(DRM_R128_VERTEX)] = {r128_cce_vertex, DRM_AUTH},
1705         [DRM_IOCTL_NR(DRM_R128_INDICES)] = {r128_cce_indices, DRM_AUTH},
1706         [DRM_IOCTL_NR(DRM_R128_BLIT)] = {r128_cce_blit, DRM_AUTH},
1707         [DRM_IOCTL_NR(DRM_R128_DEPTH)] = {r128_cce_depth, DRM_AUTH},
1708         [DRM_IOCTL_NR(DRM_R128_STIPPLE)] = {r128_cce_stipple, DRM_AUTH},
1709         [DRM_IOCTL_NR(DRM_R128_INDIRECT)] = {r128_cce_indirect, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY},
1710         [DRM_IOCTL_NR(DRM_R128_GETPARAM)] = {r128_getparam, DRM_AUTH},
1711 };
1712
1713 int r128_max_ioctl = DRM_ARRAY_SIZE(r128_ioctls);