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 <linux/pci.h>
32 #include <linux/slab.h>
33 #include <linux/uaccess.h>
34 
35 #include <drm/drm_device.h>
36 #include <drm/drm_file.h>
37 #include <drm/drm_print.h>
38 #include <drm/r128_drm.h>
39 
40 #include "r128_drv.h"
41 
42 /* ================================================================
43  * CCE hardware state programming functions
44  */
45 
r128_emit_clip_rects(drm_r128_private_t * dev_priv,struct drm_clip_rect * boxes,int count)46 static void r128_emit_clip_rects(drm_r128_private_t *dev_priv,
47 				 struct drm_clip_rect *boxes, int count)
48 {
49 	u32 aux_sc_cntl = 0x00000000;
50 	RING_LOCALS;
51 	DRM_DEBUG("\n");
52 
53 	BEGIN_RING((count < 3 ? count : 3) * 5 + 2);
54 
55 	if (count >= 1) {
56 		OUT_RING(CCE_PACKET0(R128_AUX1_SC_LEFT, 3));
57 		OUT_RING(boxes[0].x1);
58 		OUT_RING(boxes[0].x2 - 1);
59 		OUT_RING(boxes[0].y1);
60 		OUT_RING(boxes[0].y2 - 1);
61 
62 		aux_sc_cntl |= (R128_AUX1_SC_EN | R128_AUX1_SC_MODE_OR);
63 	}
64 	if (count >= 2) {
65 		OUT_RING(CCE_PACKET0(R128_AUX2_SC_LEFT, 3));
66 		OUT_RING(boxes[1].x1);
67 		OUT_RING(boxes[1].x2 - 1);
68 		OUT_RING(boxes[1].y1);
69 		OUT_RING(boxes[1].y2 - 1);
70 
71 		aux_sc_cntl |= (R128_AUX2_SC_EN | R128_AUX2_SC_MODE_OR);
72 	}
73 	if (count >= 3) {
74 		OUT_RING(CCE_PACKET0(R128_AUX3_SC_LEFT, 3));
75 		OUT_RING(boxes[2].x1);
76 		OUT_RING(boxes[2].x2 - 1);
77 		OUT_RING(boxes[2].y1);
78 		OUT_RING(boxes[2].y2 - 1);
79 
80 		aux_sc_cntl |= (R128_AUX3_SC_EN | R128_AUX3_SC_MODE_OR);
81 	}
82 
83 	OUT_RING(CCE_PACKET0(R128_AUX_SC_CNTL, 0));
84 	OUT_RING(aux_sc_cntl);
85 
86 	ADVANCE_RING();
87 }
88 
r128_emit_core(drm_r128_private_t * dev_priv)89 static __inline__ void r128_emit_core(drm_r128_private_t *dev_priv)
90 {
91 	drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
92 	drm_r128_context_regs_t *ctx = &sarea_priv->context_state;
93 	RING_LOCALS;
94 	DRM_DEBUG("\n");
95 
96 	BEGIN_RING(2);
97 
98 	OUT_RING(CCE_PACKET0(R128_SCALE_3D_CNTL, 0));
99 	OUT_RING(ctx->scale_3d_cntl);
100 
101 	ADVANCE_RING();
102 }
103 
r128_emit_context(drm_r128_private_t * dev_priv)104 static __inline__ void r128_emit_context(drm_r128_private_t *dev_priv)
105 {
106 	drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
107 	drm_r128_context_regs_t *ctx = &sarea_priv->context_state;
108 	RING_LOCALS;
109 	DRM_DEBUG("\n");
110 
111 	BEGIN_RING(13);
112 
113 	OUT_RING(CCE_PACKET0(R128_DST_PITCH_OFFSET_C, 11));
114 	OUT_RING(ctx->dst_pitch_offset_c);
115 	OUT_RING(ctx->dp_gui_master_cntl_c);
116 	OUT_RING(ctx->sc_top_left_c);
117 	OUT_RING(ctx->sc_bottom_right_c);
118 	OUT_RING(ctx->z_offset_c);
119 	OUT_RING(ctx->z_pitch_c);
120 	OUT_RING(ctx->z_sten_cntl_c);
121 	OUT_RING(ctx->tex_cntl_c);
122 	OUT_RING(ctx->misc_3d_state_cntl_reg);
123 	OUT_RING(ctx->texture_clr_cmp_clr_c);
124 	OUT_RING(ctx->texture_clr_cmp_msk_c);
125 	OUT_RING(ctx->fog_color_c);
126 
127 	ADVANCE_RING();
128 }
129 
r128_emit_setup(drm_r128_private_t * dev_priv)130 static __inline__ void r128_emit_setup(drm_r128_private_t *dev_priv)
131 {
132 	drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
133 	drm_r128_context_regs_t *ctx = &sarea_priv->context_state;
134 	RING_LOCALS;
135 	DRM_DEBUG("\n");
136 
137 	BEGIN_RING(3);
138 
139 	OUT_RING(CCE_PACKET1(R128_SETUP_CNTL, R128_PM4_VC_FPU_SETUP));
140 	OUT_RING(ctx->setup_cntl);
141 	OUT_RING(ctx->pm4_vc_fpu_setup);
142 
143 	ADVANCE_RING();
144 }
145 
r128_emit_masks(drm_r128_private_t * dev_priv)146 static __inline__ void r128_emit_masks(drm_r128_private_t *dev_priv)
147 {
148 	drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
149 	drm_r128_context_regs_t *ctx = &sarea_priv->context_state;
150 	RING_LOCALS;
151 	DRM_DEBUG("\n");
152 
153 	BEGIN_RING(5);
154 
155 	OUT_RING(CCE_PACKET0(R128_DP_WRITE_MASK, 0));
156 	OUT_RING(ctx->dp_write_mask);
157 
158 	OUT_RING(CCE_PACKET0(R128_STEN_REF_MASK_C, 1));
159 	OUT_RING(ctx->sten_ref_mask_c);
160 	OUT_RING(ctx->plane_3d_mask_c);
161 
162 	ADVANCE_RING();
163 }
164 
r128_emit_window(drm_r128_private_t * dev_priv)165 static __inline__ void r128_emit_window(drm_r128_private_t *dev_priv)
166 {
167 	drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
168 	drm_r128_context_regs_t *ctx = &sarea_priv->context_state;
169 	RING_LOCALS;
170 	DRM_DEBUG("\n");
171 
172 	BEGIN_RING(2);
173 
174 	OUT_RING(CCE_PACKET0(R128_WINDOW_XY_OFFSET, 0));
175 	OUT_RING(ctx->window_xy_offset);
176 
177 	ADVANCE_RING();
178 }
179 
r128_emit_tex0(drm_r128_private_t * dev_priv)180 static __inline__ void r128_emit_tex0(drm_r128_private_t *dev_priv)
181 {
182 	drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
183 	drm_r128_context_regs_t *ctx = &sarea_priv->context_state;
184 	drm_r128_texture_regs_t *tex = &sarea_priv->tex_state[0];
185 	int i;
186 	RING_LOCALS;
187 	DRM_DEBUG("\n");
188 
189 	BEGIN_RING(7 + R128_MAX_TEXTURE_LEVELS);
190 
191 	OUT_RING(CCE_PACKET0(R128_PRIM_TEX_CNTL_C,
192 			     2 + R128_MAX_TEXTURE_LEVELS));
193 	OUT_RING(tex->tex_cntl);
194 	OUT_RING(tex->tex_combine_cntl);
195 	OUT_RING(ctx->tex_size_pitch_c);
196 	for (i = 0; i < R128_MAX_TEXTURE_LEVELS; i++)
197 		OUT_RING(tex->tex_offset[i]);
198 
199 	OUT_RING(CCE_PACKET0(R128_CONSTANT_COLOR_C, 1));
200 	OUT_RING(ctx->constant_color_c);
201 	OUT_RING(tex->tex_border_color);
202 
203 	ADVANCE_RING();
204 }
205 
r128_emit_tex1(drm_r128_private_t * dev_priv)206 static __inline__ void r128_emit_tex1(drm_r128_private_t *dev_priv)
207 {
208 	drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
209 	drm_r128_texture_regs_t *tex = &sarea_priv->tex_state[1];
210 	int i;
211 	RING_LOCALS;
212 	DRM_DEBUG("\n");
213 
214 	BEGIN_RING(5 + R128_MAX_TEXTURE_LEVELS);
215 
216 	OUT_RING(CCE_PACKET0(R128_SEC_TEX_CNTL_C, 1 + R128_MAX_TEXTURE_LEVELS));
217 	OUT_RING(tex->tex_cntl);
218 	OUT_RING(tex->tex_combine_cntl);
219 	for (i = 0; i < R128_MAX_TEXTURE_LEVELS; i++)
220 		OUT_RING(tex->tex_offset[i]);
221 
222 	OUT_RING(CCE_PACKET0(R128_SEC_TEXTURE_BORDER_COLOR_C, 0));
223 	OUT_RING(tex->tex_border_color);
224 
225 	ADVANCE_RING();
226 }
227 
r128_emit_state(drm_r128_private_t * dev_priv)228 static void r128_emit_state(drm_r128_private_t *dev_priv)
229 {
230 	drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
231 	unsigned int dirty = sarea_priv->dirty;
232 
233 	DRM_DEBUG("dirty=0x%08x\n", dirty);
234 
235 	if (dirty & R128_UPLOAD_CORE) {
236 		r128_emit_core(dev_priv);
237 		sarea_priv->dirty &= ~R128_UPLOAD_CORE;
238 	}
239 
240 	if (dirty & R128_UPLOAD_CONTEXT) {
241 		r128_emit_context(dev_priv);
242 		sarea_priv->dirty &= ~R128_UPLOAD_CONTEXT;
243 	}
244 
245 	if (dirty & R128_UPLOAD_SETUP) {
246 		r128_emit_setup(dev_priv);
247 		sarea_priv->dirty &= ~R128_UPLOAD_SETUP;
248 	}
249 
250 	if (dirty & R128_UPLOAD_MASKS) {
251 		r128_emit_masks(dev_priv);
252 		sarea_priv->dirty &= ~R128_UPLOAD_MASKS;
253 	}
254 
255 	if (dirty & R128_UPLOAD_WINDOW) {
256 		r128_emit_window(dev_priv);
257 		sarea_priv->dirty &= ~R128_UPLOAD_WINDOW;
258 	}
259 
260 	if (dirty & R128_UPLOAD_TEX0) {
261 		r128_emit_tex0(dev_priv);
262 		sarea_priv->dirty &= ~R128_UPLOAD_TEX0;
263 	}
264 
265 	if (dirty & R128_UPLOAD_TEX1) {
266 		r128_emit_tex1(dev_priv);
267 		sarea_priv->dirty &= ~R128_UPLOAD_TEX1;
268 	}
269 
270 	/* Turn off the texture cache flushing */
271 	sarea_priv->context_state.tex_cntl_c &= ~R128_TEX_CACHE_FLUSH;
272 
273 	sarea_priv->dirty &= ~R128_REQUIRE_QUIESCENCE;
274 }
275 
276 #if R128_PERFORMANCE_BOXES
277 /* ================================================================
278  * Performance monitoring functions
279  */
280 
r128_clear_box(drm_r128_private_t * dev_priv,int x,int y,int w,int h,int r,int g,int b)281 static void r128_clear_box(drm_r128_private_t *dev_priv,
282 			   int x, int y, int w, int h, int r, int g, int b)
283 {
284 	u32 pitch, offset;
285 	u32 fb_bpp, color;
286 	RING_LOCALS;
287 
288 	switch (dev_priv->fb_bpp) {
289 	case 16:
290 		fb_bpp = R128_GMC_DST_16BPP;
291 		color = (((r & 0xf8) << 8) |
292 			 ((g & 0xfc) << 3) | ((b & 0xf8) >> 3));
293 		break;
294 	case 24:
295 		fb_bpp = R128_GMC_DST_24BPP;
296 		color = ((r << 16) | (g << 8) | b);
297 		break;
298 	case 32:
299 		fb_bpp = R128_GMC_DST_32BPP;
300 		color = (((0xff) << 24) | (r << 16) | (g << 8) | b);
301 		break;
302 	default:
303 		return;
304 	}
305 
306 	offset = dev_priv->back_offset;
307 	pitch = dev_priv->back_pitch >> 3;
308 
309 	BEGIN_RING(6);
310 
311 	OUT_RING(CCE_PACKET3(R128_CNTL_PAINT_MULTI, 4));
312 	OUT_RING(R128_GMC_DST_PITCH_OFFSET_CNTL |
313 		 R128_GMC_BRUSH_SOLID_COLOR |
314 		 fb_bpp |
315 		 R128_GMC_SRC_DATATYPE_COLOR |
316 		 R128_ROP3_P |
317 		 R128_GMC_CLR_CMP_CNTL_DIS | R128_GMC_AUX_CLIP_DIS);
318 
319 	OUT_RING((pitch << 21) | (offset >> 5));
320 	OUT_RING(color);
321 
322 	OUT_RING((x << 16) | y);
323 	OUT_RING((w << 16) | h);
324 
325 	ADVANCE_RING();
326 }
327 
r128_cce_performance_boxes(drm_r128_private_t * dev_priv)328 static void r128_cce_performance_boxes(drm_r128_private_t *dev_priv)
329 {
330 	if (atomic_read(&dev_priv->idle_count) == 0)
331 		r128_clear_box(dev_priv, 64, 4, 8, 8, 0, 255, 0);
332 	else
333 		atomic_set(&dev_priv->idle_count, 0);
334 }
335 
336 #endif
337 
338 /* ================================================================
339  * CCE command dispatch functions
340  */
341 
r128_print_dirty(const char * msg,unsigned int flags)342 static void r128_print_dirty(const char *msg, unsigned int flags)
343 {
344 	DRM_INFO("%s: (0x%x) %s%s%s%s%s%s%s%s%s\n",
345 		 msg,
346 		 flags,
347 		 (flags & R128_UPLOAD_CORE) ? "core, " : "",
348 		 (flags & R128_UPLOAD_CONTEXT) ? "context, " : "",
349 		 (flags & R128_UPLOAD_SETUP) ? "setup, " : "",
350 		 (flags & R128_UPLOAD_TEX0) ? "tex0, " : "",
351 		 (flags & R128_UPLOAD_TEX1) ? "tex1, " : "",
352 		 (flags & R128_UPLOAD_MASKS) ? "masks, " : "",
353 		 (flags & R128_UPLOAD_WINDOW) ? "window, " : "",
354 		 (flags & R128_UPLOAD_CLIPRECTS) ? "cliprects, " : "",
355 		 (flags & R128_REQUIRE_QUIESCENCE) ? "quiescence, " : "");
356 }
357 
r128_cce_dispatch_clear(struct drm_device * dev,drm_r128_clear_t * clear)358 static void r128_cce_dispatch_clear(struct drm_device *dev,
359 				    drm_r128_clear_t *clear)
360 {
361 	drm_r128_private_t *dev_priv = dev->dev_private;
362 	drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
363 	int nbox = sarea_priv->nbox;
364 	struct drm_clip_rect *pbox = sarea_priv->boxes;
365 	unsigned int flags = clear->flags;
366 	int i;
367 	RING_LOCALS;
368 	DRM_DEBUG("\n");
369 
370 	if (dev_priv->page_flipping && dev_priv->current_page == 1) {
371 		unsigned int tmp = flags;
372 
373 		flags &= ~(R128_FRONT | R128_BACK);
374 		if (tmp & R128_FRONT)
375 			flags |= R128_BACK;
376 		if (tmp & R128_BACK)
377 			flags |= R128_FRONT;
378 	}
379 
380 	for (i = 0; i < nbox; i++) {
381 		int x = pbox[i].x1;
382 		int y = pbox[i].y1;
383 		int w = pbox[i].x2 - x;
384 		int h = pbox[i].y2 - y;
385 
386 		DRM_DEBUG("dispatch clear %d,%d-%d,%d flags 0x%x\n",
387 			  pbox[i].x1, pbox[i].y1, pbox[i].x2,
388 			  pbox[i].y2, flags);
389 
390 		if (flags & (R128_FRONT | R128_BACK)) {
391 			BEGIN_RING(2);
392 
393 			OUT_RING(CCE_PACKET0(R128_DP_WRITE_MASK, 0));
394 			OUT_RING(clear->color_mask);
395 
396 			ADVANCE_RING();
397 		}
398 
399 		if (flags & R128_FRONT) {
400 			BEGIN_RING(6);
401 
402 			OUT_RING(CCE_PACKET3(R128_CNTL_PAINT_MULTI, 4));
403 			OUT_RING(R128_GMC_DST_PITCH_OFFSET_CNTL |
404 				 R128_GMC_BRUSH_SOLID_COLOR |
405 				 (dev_priv->color_fmt << 8) |
406 				 R128_GMC_SRC_DATATYPE_COLOR |
407 				 R128_ROP3_P |
408 				 R128_GMC_CLR_CMP_CNTL_DIS |
409 				 R128_GMC_AUX_CLIP_DIS);
410 
411 			OUT_RING(dev_priv->front_pitch_offset_c);
412 			OUT_RING(clear->clear_color);
413 
414 			OUT_RING((x << 16) | y);
415 			OUT_RING((w << 16) | h);
416 
417 			ADVANCE_RING();
418 		}
419 
420 		if (flags & R128_BACK) {
421 			BEGIN_RING(6);
422 
423 			OUT_RING(CCE_PACKET3(R128_CNTL_PAINT_MULTI, 4));
424 			OUT_RING(R128_GMC_DST_PITCH_OFFSET_CNTL |
425 				 R128_GMC_BRUSH_SOLID_COLOR |
426 				 (dev_priv->color_fmt << 8) |
427 				 R128_GMC_SRC_DATATYPE_COLOR |
428 				 R128_ROP3_P |
429 				 R128_GMC_CLR_CMP_CNTL_DIS |
430 				 R128_GMC_AUX_CLIP_DIS);
431 
432 			OUT_RING(dev_priv->back_pitch_offset_c);
433 			OUT_RING(clear->clear_color);
434 
435 			OUT_RING((x << 16) | y);
436 			OUT_RING((w << 16) | h);
437 
438 			ADVANCE_RING();
439 		}
440 
441 		if (flags & R128_DEPTH) {
442 			BEGIN_RING(6);
443 
444 			OUT_RING(CCE_PACKET3(R128_CNTL_PAINT_MULTI, 4));
445 			OUT_RING(R128_GMC_DST_PITCH_OFFSET_CNTL |
446 				 R128_GMC_BRUSH_SOLID_COLOR |
447 				 (dev_priv->depth_fmt << 8) |
448 				 R128_GMC_SRC_DATATYPE_COLOR |
449 				 R128_ROP3_P |
450 				 R128_GMC_CLR_CMP_CNTL_DIS |
451 				 R128_GMC_AUX_CLIP_DIS | R128_GMC_WR_MSK_DIS);
452 
453 			OUT_RING(dev_priv->depth_pitch_offset_c);
454 			OUT_RING(clear->clear_depth);
455 
456 			OUT_RING((x << 16) | y);
457 			OUT_RING((w << 16) | h);
458 
459 			ADVANCE_RING();
460 		}
461 	}
462 }
463 
r128_cce_dispatch_swap(struct drm_device * dev)464 static void r128_cce_dispatch_swap(struct drm_device *dev)
465 {
466 	drm_r128_private_t *dev_priv = dev->dev_private;
467 	drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
468 	int nbox = sarea_priv->nbox;
469 	struct drm_clip_rect *pbox = sarea_priv->boxes;
470 	int i;
471 	RING_LOCALS;
472 	DRM_DEBUG("\n");
473 
474 #if R128_PERFORMANCE_BOXES
475 	/* Do some trivial performance monitoring...
476 	 */
477 	r128_cce_performance_boxes(dev_priv);
478 #endif
479 
480 	for (i = 0; i < nbox; i++) {
481 		int x = pbox[i].x1;
482 		int y = pbox[i].y1;
483 		int w = pbox[i].x2 - x;
484 		int h = pbox[i].y2 - y;
485 
486 		BEGIN_RING(7);
487 
488 		OUT_RING(CCE_PACKET3(R128_CNTL_BITBLT_MULTI, 5));
489 		OUT_RING(R128_GMC_SRC_PITCH_OFFSET_CNTL |
490 			 R128_GMC_DST_PITCH_OFFSET_CNTL |
491 			 R128_GMC_BRUSH_NONE |
492 			 (dev_priv->color_fmt << 8) |
493 			 R128_GMC_SRC_DATATYPE_COLOR |
494 			 R128_ROP3_S |
495 			 R128_DP_SRC_SOURCE_MEMORY |
496 			 R128_GMC_CLR_CMP_CNTL_DIS |
497 			 R128_GMC_AUX_CLIP_DIS | R128_GMC_WR_MSK_DIS);
498 
499 		/* Make this work even if front & back are flipped:
500 		 */
501 		if (dev_priv->current_page == 0) {
502 			OUT_RING(dev_priv->back_pitch_offset_c);
503 			OUT_RING(dev_priv->front_pitch_offset_c);
504 		} else {
505 			OUT_RING(dev_priv->front_pitch_offset_c);
506 			OUT_RING(dev_priv->back_pitch_offset_c);
507 		}
508 
509 		OUT_RING((x << 16) | y);
510 		OUT_RING((x << 16) | y);
511 		OUT_RING((w << 16) | h);
512 
513 		ADVANCE_RING();
514 	}
515 
516 	/* Increment the frame counter.  The client-side 3D driver must
517 	 * throttle the framerate by waiting for this value before
518 	 * performing the swapbuffer ioctl.
519 	 */
520 	dev_priv->sarea_priv->last_frame++;
521 
522 	BEGIN_RING(2);
523 
524 	OUT_RING(CCE_PACKET0(R128_LAST_FRAME_REG, 0));
525 	OUT_RING(dev_priv->sarea_priv->last_frame);
526 
527 	ADVANCE_RING();
528 }
529 
r128_cce_dispatch_flip(struct drm_device * dev)530 static void r128_cce_dispatch_flip(struct drm_device *dev)
531 {
532 	drm_r128_private_t *dev_priv = dev->dev_private;
533 	RING_LOCALS;
534 	DRM_DEBUG("page=%d pfCurrentPage=%d\n",
535 		  dev_priv->current_page, dev_priv->sarea_priv->pfCurrentPage);
536 
537 #if R128_PERFORMANCE_BOXES
538 	/* Do some trivial performance monitoring...
539 	 */
540 	r128_cce_performance_boxes(dev_priv);
541 #endif
542 
543 	BEGIN_RING(4);
544 
545 	R128_WAIT_UNTIL_PAGE_FLIPPED();
546 	OUT_RING(CCE_PACKET0(R128_CRTC_OFFSET, 0));
547 
548 	if (dev_priv->current_page == 0)
549 		OUT_RING(dev_priv->back_offset);
550 	else
551 		OUT_RING(dev_priv->front_offset);
552 
553 	ADVANCE_RING();
554 
555 	/* Increment the frame counter.  The client-side 3D driver must
556 	 * throttle the framerate by waiting for this value before
557 	 * performing the swapbuffer ioctl.
558 	 */
559 	dev_priv->sarea_priv->last_frame++;
560 	dev_priv->sarea_priv->pfCurrentPage = dev_priv->current_page =
561 	    1 - dev_priv->current_page;
562 
563 	BEGIN_RING(2);
564 
565 	OUT_RING(CCE_PACKET0(R128_LAST_FRAME_REG, 0));
566 	OUT_RING(dev_priv->sarea_priv->last_frame);
567 
568 	ADVANCE_RING();
569 }
570 
r128_cce_dispatch_vertex(struct drm_device * dev,struct drm_buf * buf)571 static void r128_cce_dispatch_vertex(struct drm_device *dev, struct drm_buf *buf)
572 {
573 	drm_r128_private_t *dev_priv = dev->dev_private;
574 	drm_r128_buf_priv_t *buf_priv = buf->dev_private;
575 	drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
576 	int format = sarea_priv->vc_format;
577 	int offset = buf->bus_address;
578 	int size = buf->used;
579 	int prim = buf_priv->prim;
580 	int i = 0;
581 	RING_LOCALS;
582 	DRM_DEBUG("buf=%d nbox=%d\n", buf->idx, sarea_priv->nbox);
583 
584 	if (0)
585 		r128_print_dirty("dispatch_vertex", sarea_priv->dirty);
586 
587 	if (buf->used) {
588 		buf_priv->dispatched = 1;
589 
590 		if (sarea_priv->dirty & ~R128_UPLOAD_CLIPRECTS)
591 			r128_emit_state(dev_priv);
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 
r128_cce_dispatch_indirect(struct drm_device * dev,struct drm_buf * buf,int start,int end)640 static void r128_cce_dispatch_indirect(struct drm_device *dev,
641 				       struct drm_buf *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 
r128_cce_dispatch_indices(struct drm_device * dev,struct drm_buf * buf,int start,int end,int count)695 static void r128_cce_dispatch_indices(struct drm_device *dev,
696 				      struct drm_buf *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 		dwords = (end - start + 3) / sizeof(u32);
721 
722 		data = (u32 *) ((char *)dev->agp_buffer_map->handle
723 				+ buf->offset + start);
724 
725 		data[0] = cpu_to_le32(CCE_PACKET3(R128_3D_RNDR_GEN_INDX_PRIM,
726 						  dwords - 2));
727 
728 		data[1] = cpu_to_le32(offset);
729 		data[2] = cpu_to_le32(R128_MAX_VB_VERTS);
730 		data[3] = cpu_to_le32(format);
731 		data[4] = cpu_to_le32((prim | R128_CCE_VC_CNTL_PRIM_WALK_IND |
732 				       (count << 16)));
733 
734 		if (count & 0x1) {
735 #ifdef __LITTLE_ENDIAN
736 			data[dwords - 1] &= 0x0000ffff;
737 #else
738 			data[dwords - 1] &= 0xffff0000;
739 #endif
740 		}
741 
742 		do {
743 			/* Emit the next set of up to three cliprects */
744 			if (i < sarea_priv->nbox) {
745 				r128_emit_clip_rects(dev_priv,
746 						     &sarea_priv->boxes[i],
747 						     sarea_priv->nbox - i);
748 			}
749 
750 			r128_cce_dispatch_indirect(dev, buf, start, end);
751 
752 			i += 3;
753 		} while (i < sarea_priv->nbox);
754 	}
755 
756 	if (buf_priv->discard) {
757 		buf_priv->age = dev_priv->sarea_priv->last_dispatch;
758 
759 		/* Emit the vertex buffer age */
760 		BEGIN_RING(2);
761 
762 		OUT_RING(CCE_PACKET0(R128_LAST_DISPATCH_REG, 0));
763 		OUT_RING(buf_priv->age);
764 
765 		ADVANCE_RING();
766 
767 		buf->pending = 1;
768 		/* FIXME: Check dispatched field */
769 		buf_priv->dispatched = 0;
770 	}
771 
772 	dev_priv->sarea_priv->last_dispatch++;
773 
774 	sarea_priv->dirty &= ~R128_UPLOAD_CLIPRECTS;
775 	sarea_priv->nbox = 0;
776 }
777 
r128_cce_dispatch_blit(struct drm_device * dev,struct drm_file * file_priv,drm_r128_blit_t * blit)778 static int r128_cce_dispatch_blit(struct drm_device *dev,
779 				  struct drm_file *file_priv,
780 				  drm_r128_blit_t *blit)
781 {
782 	drm_r128_private_t *dev_priv = dev->dev_private;
783 	struct drm_device_dma *dma = dev->dma;
784 	struct drm_buf *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 -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->file_priv != file_priv) {
833 		DRM_ERROR("process %d using buffer owned by %p\n",
834 			  task_pid_nr(current), buf->file_priv);
835 		return -EINVAL;
836 	}
837 	if (buf->pending) {
838 		DRM_ERROR("sending pending buffer %d\n", blit->idx);
839 		return -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 
r128_cce_dispatch_write_span(struct drm_device * dev,drm_r128_depth_t * depth)890 static int r128_cce_dispatch_write_span(struct drm_device *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 -EMSGSIZE;
904 
905 	if (copy_from_user(&x, depth->x, sizeof(x)))
906 		return -EFAULT;
907 	if (copy_from_user(&y, depth->y, sizeof(y)))
908 		return -EFAULT;
909 
910 	buffer_size = depth->n * sizeof(u32);
911 	buffer = memdup_user(depth->buffer, buffer_size);
912 	if (IS_ERR(buffer))
913 		return PTR_ERR(buffer);
914 
915 	mask_size = depth->n;
916 	if (depth->mask) {
917 		mask = memdup_user(depth->mask, mask_size);
918 		if (IS_ERR(mask)) {
919 			kfree(buffer);
920 			return PTR_ERR(mask);
921 		}
922 
923 		for (i = 0; i < count; i++, x++) {
924 			if (mask[i]) {
925 				BEGIN_RING(6);
926 
927 				OUT_RING(CCE_PACKET3(R128_CNTL_PAINT_MULTI, 4));
928 				OUT_RING(R128_GMC_DST_PITCH_OFFSET_CNTL |
929 					 R128_GMC_BRUSH_SOLID_COLOR |
930 					 (dev_priv->depth_fmt << 8) |
931 					 R128_GMC_SRC_DATATYPE_COLOR |
932 					 R128_ROP3_P |
933 					 R128_GMC_CLR_CMP_CNTL_DIS |
934 					 R128_GMC_WR_MSK_DIS);
935 
936 				OUT_RING(dev_priv->depth_pitch_offset_c);
937 				OUT_RING(buffer[i]);
938 
939 				OUT_RING((x << 16) | y);
940 				OUT_RING((1 << 16) | 1);
941 
942 				ADVANCE_RING();
943 			}
944 		}
945 
946 		kfree(mask);
947 	} else {
948 		for (i = 0; i < count; i++, x++) {
949 			BEGIN_RING(6);
950 
951 			OUT_RING(CCE_PACKET3(R128_CNTL_PAINT_MULTI, 4));
952 			OUT_RING(R128_GMC_DST_PITCH_OFFSET_CNTL |
953 				 R128_GMC_BRUSH_SOLID_COLOR |
954 				 (dev_priv->depth_fmt << 8) |
955 				 R128_GMC_SRC_DATATYPE_COLOR |
956 				 R128_ROP3_P |
957 				 R128_GMC_CLR_CMP_CNTL_DIS |
958 				 R128_GMC_WR_MSK_DIS);
959 
960 			OUT_RING(dev_priv->depth_pitch_offset_c);
961 			OUT_RING(buffer[i]);
962 
963 			OUT_RING((x << 16) | y);
964 			OUT_RING((1 << 16) | 1);
965 
966 			ADVANCE_RING();
967 		}
968 	}
969 
970 	kfree(buffer);
971 
972 	return 0;
973 }
974 
r128_cce_dispatch_write_pixels(struct drm_device * dev,drm_r128_depth_t * depth)975 static int r128_cce_dispatch_write_pixels(struct drm_device *dev,
976 					  drm_r128_depth_t *depth)
977 {
978 	drm_r128_private_t *dev_priv = dev->dev_private;
979 	int count, *x, *y;
980 	u32 *buffer;
981 	u8 *mask;
982 	int i, xbuf_size, ybuf_size, buffer_size, mask_size;
983 	RING_LOCALS;
984 	DRM_DEBUG("\n");
985 
986 	count = depth->n;
987 	if (count > 4096 || count <= 0)
988 		return -EMSGSIZE;
989 
990 	xbuf_size = count * sizeof(*x);
991 	ybuf_size = count * sizeof(*y);
992 	x = memdup_user(depth->x, xbuf_size);
993 	if (IS_ERR(x))
994 		return PTR_ERR(x);
995 	y = memdup_user(depth->y, ybuf_size);
996 	if (IS_ERR(y)) {
997 		kfree(x);
998 		return PTR_ERR(y);
999 	}
1000 	buffer_size = depth->n * sizeof(u32);
1001 	buffer = memdup_user(depth->buffer, buffer_size);
1002 	if (IS_ERR(buffer)) {
1003 		kfree(x);
1004 		kfree(y);
1005 		return PTR_ERR(buffer);
1006 	}
1007 
1008 	if (depth->mask) {
1009 		mask_size = depth->n;
1010 		mask = memdup_user(depth->mask, mask_size);
1011 		if (IS_ERR(mask)) {
1012 			kfree(x);
1013 			kfree(y);
1014 			kfree(buffer);
1015 			return PTR_ERR(mask);
1016 		}
1017 
1018 		for (i = 0; i < count; i++) {
1019 			if (mask[i]) {
1020 				BEGIN_RING(6);
1021 
1022 				OUT_RING(CCE_PACKET3(R128_CNTL_PAINT_MULTI, 4));
1023 				OUT_RING(R128_GMC_DST_PITCH_OFFSET_CNTL |
1024 					 R128_GMC_BRUSH_SOLID_COLOR |
1025 					 (dev_priv->depth_fmt << 8) |
1026 					 R128_GMC_SRC_DATATYPE_COLOR |
1027 					 R128_ROP3_P |
1028 					 R128_GMC_CLR_CMP_CNTL_DIS |
1029 					 R128_GMC_WR_MSK_DIS);
1030 
1031 				OUT_RING(dev_priv->depth_pitch_offset_c);
1032 				OUT_RING(buffer[i]);
1033 
1034 				OUT_RING((x[i] << 16) | y[i]);
1035 				OUT_RING((1 << 16) | 1);
1036 
1037 				ADVANCE_RING();
1038 			}
1039 		}
1040 
1041 		kfree(mask);
1042 	} else {
1043 		for (i = 0; i < count; i++) {
1044 			BEGIN_RING(6);
1045 
1046 			OUT_RING(CCE_PACKET3(R128_CNTL_PAINT_MULTI, 4));
1047 			OUT_RING(R128_GMC_DST_PITCH_OFFSET_CNTL |
1048 				 R128_GMC_BRUSH_SOLID_COLOR |
1049 				 (dev_priv->depth_fmt << 8) |
1050 				 R128_GMC_SRC_DATATYPE_COLOR |
1051 				 R128_ROP3_P |
1052 				 R128_GMC_CLR_CMP_CNTL_DIS |
1053 				 R128_GMC_WR_MSK_DIS);
1054 
1055 			OUT_RING(dev_priv->depth_pitch_offset_c);
1056 			OUT_RING(buffer[i]);
1057 
1058 			OUT_RING((x[i] << 16) | y[i]);
1059 			OUT_RING((1 << 16) | 1);
1060 
1061 			ADVANCE_RING();
1062 		}
1063 	}
1064 
1065 	kfree(x);
1066 	kfree(y);
1067 	kfree(buffer);
1068 
1069 	return 0;
1070 }
1071 
r128_cce_dispatch_read_span(struct drm_device * dev,drm_r128_depth_t * depth)1072 static int r128_cce_dispatch_read_span(struct drm_device *dev,
1073 				       drm_r128_depth_t *depth)
1074 {
1075 	drm_r128_private_t *dev_priv = dev->dev_private;
1076 	int count, x, y;
1077 	RING_LOCALS;
1078 	DRM_DEBUG("\n");
1079 
1080 	count = depth->n;
1081 	if (count > 4096 || count <= 0)
1082 		return -EMSGSIZE;
1083 
1084 	if (copy_from_user(&x, depth->x, sizeof(x)))
1085 		return -EFAULT;
1086 	if (copy_from_user(&y, depth->y, sizeof(y)))
1087 		return -EFAULT;
1088 
1089 	BEGIN_RING(7);
1090 
1091 	OUT_RING(CCE_PACKET3(R128_CNTL_BITBLT_MULTI, 5));
1092 	OUT_RING(R128_GMC_SRC_PITCH_OFFSET_CNTL |
1093 		 R128_GMC_DST_PITCH_OFFSET_CNTL |
1094 		 R128_GMC_BRUSH_NONE |
1095 		 (dev_priv->depth_fmt << 8) |
1096 		 R128_GMC_SRC_DATATYPE_COLOR |
1097 		 R128_ROP3_S |
1098 		 R128_DP_SRC_SOURCE_MEMORY |
1099 		 R128_GMC_CLR_CMP_CNTL_DIS | R128_GMC_WR_MSK_DIS);
1100 
1101 	OUT_RING(dev_priv->depth_pitch_offset_c);
1102 	OUT_RING(dev_priv->span_pitch_offset_c);
1103 
1104 	OUT_RING((x << 16) | y);
1105 	OUT_RING((0 << 16) | 0);
1106 	OUT_RING((count << 16) | 1);
1107 
1108 	ADVANCE_RING();
1109 
1110 	return 0;
1111 }
1112 
r128_cce_dispatch_read_pixels(struct drm_device * dev,drm_r128_depth_t * depth)1113 static int r128_cce_dispatch_read_pixels(struct drm_device *dev,
1114 					 drm_r128_depth_t *depth)
1115 {
1116 	drm_r128_private_t *dev_priv = dev->dev_private;
1117 	int count, *x, *y;
1118 	int i, xbuf_size, ybuf_size;
1119 	RING_LOCALS;
1120 	DRM_DEBUG("\n");
1121 
1122 	count = depth->n;
1123 	if (count > 4096 || count <= 0)
1124 		return -EMSGSIZE;
1125 
1126 	if (count > dev_priv->depth_pitch)
1127 		count = dev_priv->depth_pitch;
1128 
1129 	xbuf_size = count * sizeof(*x);
1130 	ybuf_size = count * sizeof(*y);
1131 	x = kmalloc(xbuf_size, GFP_KERNEL);
1132 	if (x == NULL)
1133 		return -ENOMEM;
1134 	y = kmalloc(ybuf_size, GFP_KERNEL);
1135 	if (y == NULL) {
1136 		kfree(x);
1137 		return -ENOMEM;
1138 	}
1139 	if (copy_from_user(x, depth->x, xbuf_size)) {
1140 		kfree(x);
1141 		kfree(y);
1142 		return -EFAULT;
1143 	}
1144 	if (copy_from_user(y, depth->y, ybuf_size)) {
1145 		kfree(x);
1146 		kfree(y);
1147 		return -EFAULT;
1148 	}
1149 
1150 	for (i = 0; i < count; i++) {
1151 		BEGIN_RING(7);
1152 
1153 		OUT_RING(CCE_PACKET3(R128_CNTL_BITBLT_MULTI, 5));
1154 		OUT_RING(R128_GMC_SRC_PITCH_OFFSET_CNTL |
1155 			 R128_GMC_DST_PITCH_OFFSET_CNTL |
1156 			 R128_GMC_BRUSH_NONE |
1157 			 (dev_priv->depth_fmt << 8) |
1158 			 R128_GMC_SRC_DATATYPE_COLOR |
1159 			 R128_ROP3_S |
1160 			 R128_DP_SRC_SOURCE_MEMORY |
1161 			 R128_GMC_CLR_CMP_CNTL_DIS | R128_GMC_WR_MSK_DIS);
1162 
1163 		OUT_RING(dev_priv->depth_pitch_offset_c);
1164 		OUT_RING(dev_priv->span_pitch_offset_c);
1165 
1166 		OUT_RING((x[i] << 16) | y[i]);
1167 		OUT_RING((i << 16) | 0);
1168 		OUT_RING((1 << 16) | 1);
1169 
1170 		ADVANCE_RING();
1171 	}
1172 
1173 	kfree(x);
1174 	kfree(y);
1175 
1176 	return 0;
1177 }
1178 
1179 /* ================================================================
1180  * Polygon stipple
1181  */
1182 
r128_cce_dispatch_stipple(struct drm_device * dev,u32 * stipple)1183 static void r128_cce_dispatch_stipple(struct drm_device *dev, u32 *stipple)
1184 {
1185 	drm_r128_private_t *dev_priv = dev->dev_private;
1186 	int i;
1187 	RING_LOCALS;
1188 	DRM_DEBUG("\n");
1189 
1190 	BEGIN_RING(33);
1191 
1192 	OUT_RING(CCE_PACKET0(R128_BRUSH_DATA0, 31));
1193 	for (i = 0; i < 32; i++)
1194 		OUT_RING(stipple[i]);
1195 
1196 	ADVANCE_RING();
1197 }
1198 
1199 /* ================================================================
1200  * IOCTL functions
1201  */
1202 
r128_cce_clear(struct drm_device * dev,void * data,struct drm_file * file_priv)1203 static int r128_cce_clear(struct drm_device *dev, void *data, struct drm_file *file_priv)
1204 {
1205 	drm_r128_private_t *dev_priv = dev->dev_private;
1206 	drm_r128_sarea_t *sarea_priv;
1207 	drm_r128_clear_t *clear = data;
1208 	DRM_DEBUG("\n");
1209 
1210 	LOCK_TEST_WITH_RETURN(dev, file_priv);
1211 
1212 	DEV_INIT_TEST_WITH_RETURN(dev_priv);
1213 
1214 	RING_SPACE_TEST_WITH_RETURN(dev_priv);
1215 
1216 	sarea_priv = dev_priv->sarea_priv;
1217 
1218 	if (sarea_priv->nbox > R128_NR_SAREA_CLIPRECTS)
1219 		sarea_priv->nbox = R128_NR_SAREA_CLIPRECTS;
1220 
1221 	r128_cce_dispatch_clear(dev, clear);
1222 	COMMIT_RING();
1223 
1224 	/* Make sure we restore the 3D state next time.
1225 	 */
1226 	dev_priv->sarea_priv->dirty |= R128_UPLOAD_CONTEXT | R128_UPLOAD_MASKS;
1227 
1228 	return 0;
1229 }
1230 
r128_do_init_pageflip(struct drm_device * dev)1231 static int r128_do_init_pageflip(struct drm_device *dev)
1232 {
1233 	drm_r128_private_t *dev_priv = dev->dev_private;
1234 	DRM_DEBUG("\n");
1235 
1236 	dev_priv->crtc_offset = R128_READ(R128_CRTC_OFFSET);
1237 	dev_priv->crtc_offset_cntl = R128_READ(R128_CRTC_OFFSET_CNTL);
1238 
1239 	R128_WRITE(R128_CRTC_OFFSET, dev_priv->front_offset);
1240 	R128_WRITE(R128_CRTC_OFFSET_CNTL,
1241 		   dev_priv->crtc_offset_cntl | R128_CRTC_OFFSET_FLIP_CNTL);
1242 
1243 	dev_priv->page_flipping = 1;
1244 	dev_priv->current_page = 0;
1245 	dev_priv->sarea_priv->pfCurrentPage = dev_priv->current_page;
1246 
1247 	return 0;
1248 }
1249 
r128_do_cleanup_pageflip(struct drm_device * dev)1250 static int r128_do_cleanup_pageflip(struct drm_device *dev)
1251 {
1252 	drm_r128_private_t *dev_priv = dev->dev_private;
1253 	DRM_DEBUG("\n");
1254 
1255 	R128_WRITE(R128_CRTC_OFFSET, dev_priv->crtc_offset);
1256 	R128_WRITE(R128_CRTC_OFFSET_CNTL, dev_priv->crtc_offset_cntl);
1257 
1258 	if (dev_priv->current_page != 0) {
1259 		r128_cce_dispatch_flip(dev);
1260 		COMMIT_RING();
1261 	}
1262 
1263 	dev_priv->page_flipping = 0;
1264 	return 0;
1265 }
1266 
1267 /* Swapping and flipping are different operations, need different ioctls.
1268  * They can & should be intermixed to support multiple 3d windows.
1269  */
1270 
r128_cce_flip(struct drm_device * dev,void * data,struct drm_file * file_priv)1271 static int r128_cce_flip(struct drm_device *dev, void *data, struct drm_file *file_priv)
1272 {
1273 	drm_r128_private_t *dev_priv = dev->dev_private;
1274 	DRM_DEBUG("\n");
1275 
1276 	LOCK_TEST_WITH_RETURN(dev, file_priv);
1277 
1278 	DEV_INIT_TEST_WITH_RETURN(dev_priv);
1279 
1280 	RING_SPACE_TEST_WITH_RETURN(dev_priv);
1281 
1282 	if (!dev_priv->page_flipping)
1283 		r128_do_init_pageflip(dev);
1284 
1285 	r128_cce_dispatch_flip(dev);
1286 
1287 	COMMIT_RING();
1288 	return 0;
1289 }
1290 
r128_cce_swap(struct drm_device * dev,void * data,struct drm_file * file_priv)1291 static int r128_cce_swap(struct drm_device *dev, void *data, struct drm_file *file_priv)
1292 {
1293 	drm_r128_private_t *dev_priv = dev->dev_private;
1294 	drm_r128_sarea_t *sarea_priv = dev_priv->sarea_priv;
1295 	DRM_DEBUG("\n");
1296 
1297 	LOCK_TEST_WITH_RETURN(dev, file_priv);
1298 
1299 	DEV_INIT_TEST_WITH_RETURN(dev_priv);
1300 
1301 	RING_SPACE_TEST_WITH_RETURN(dev_priv);
1302 
1303 	if (sarea_priv->nbox > R128_NR_SAREA_CLIPRECTS)
1304 		sarea_priv->nbox = R128_NR_SAREA_CLIPRECTS;
1305 
1306 	r128_cce_dispatch_swap(dev);
1307 	dev_priv->sarea_priv->dirty |= (R128_UPLOAD_CONTEXT |
1308 					R128_UPLOAD_MASKS);
1309 
1310 	COMMIT_RING();
1311 	return 0;
1312 }
1313 
r128_cce_vertex(struct drm_device * dev,void * data,struct drm_file * file_priv)1314 static int r128_cce_vertex(struct drm_device *dev, void *data, struct drm_file *file_priv)
1315 {
1316 	drm_r128_private_t *dev_priv = dev->dev_private;
1317 	struct drm_device_dma *dma = dev->dma;
1318 	struct drm_buf *buf;
1319 	drm_r128_buf_priv_t *buf_priv;
1320 	drm_r128_vertex_t *vertex = data;
1321 
1322 	LOCK_TEST_WITH_RETURN(dev, file_priv);
1323 
1324 	DEV_INIT_TEST_WITH_RETURN(dev_priv);
1325 
1326 	DRM_DEBUG("pid=%d index=%d count=%d discard=%d\n",
1327 		  task_pid_nr(current), vertex->idx, vertex->count, vertex->discard);
1328 
1329 	if (vertex->idx < 0 || vertex->idx >= dma->buf_count) {
1330 		DRM_ERROR("buffer index %d (of %d max)\n",
1331 			  vertex->idx, dma->buf_count - 1);
1332 		return -EINVAL;
1333 	}
1334 	if (vertex->prim < 0 ||
1335 	    vertex->prim > R128_CCE_VC_CNTL_PRIM_TYPE_TRI_TYPE2) {
1336 		DRM_ERROR("buffer prim %d\n", vertex->prim);
1337 		return -EINVAL;
1338 	}
1339 
1340 	RING_SPACE_TEST_WITH_RETURN(dev_priv);
1341 	VB_AGE_TEST_WITH_RETURN(dev_priv);
1342 
1343 	buf = dma->buflist[vertex->idx];
1344 	buf_priv = buf->dev_private;
1345 
1346 	if (buf->file_priv != file_priv) {
1347 		DRM_ERROR("process %d using buffer owned by %p\n",
1348 			  task_pid_nr(current), buf->file_priv);
1349 		return -EINVAL;
1350 	}
1351 	if (buf->pending) {
1352 		DRM_ERROR("sending pending buffer %d\n", vertex->idx);
1353 		return -EINVAL;
1354 	}
1355 
1356 	buf->used = vertex->count;
1357 	buf_priv->prim = vertex->prim;
1358 	buf_priv->discard = vertex->discard;
1359 
1360 	r128_cce_dispatch_vertex(dev, buf);
1361 
1362 	COMMIT_RING();
1363 	return 0;
1364 }
1365 
r128_cce_indices(struct drm_device * dev,void * data,struct drm_file * file_priv)1366 static int r128_cce_indices(struct drm_device *dev, void *data, struct drm_file *file_priv)
1367 {
1368 	drm_r128_private_t *dev_priv = dev->dev_private;
1369 	struct drm_device_dma *dma = dev->dma;
1370 	struct drm_buf *buf;
1371 	drm_r128_buf_priv_t *buf_priv;
1372 	drm_r128_indices_t *elts = data;
1373 	int count;
1374 
1375 	LOCK_TEST_WITH_RETURN(dev, file_priv);
1376 
1377 	DEV_INIT_TEST_WITH_RETURN(dev_priv);
1378 
1379 	DRM_DEBUG("pid=%d buf=%d s=%d e=%d d=%d\n", task_pid_nr(current),
1380 		  elts->idx, elts->start, elts->end, elts->discard);
1381 
1382 	if (elts->idx < 0 || elts->idx >= dma->buf_count) {
1383 		DRM_ERROR("buffer index %d (of %d max)\n",
1384 			  elts->idx, dma->buf_count - 1);
1385 		return -EINVAL;
1386 	}
1387 	if (elts->prim < 0 ||
1388 	    elts->prim > R128_CCE_VC_CNTL_PRIM_TYPE_TRI_TYPE2) {
1389 		DRM_ERROR("buffer prim %d\n", elts->prim);
1390 		return -EINVAL;
1391 	}
1392 
1393 	RING_SPACE_TEST_WITH_RETURN(dev_priv);
1394 	VB_AGE_TEST_WITH_RETURN(dev_priv);
1395 
1396 	buf = dma->buflist[elts->idx];
1397 	buf_priv = buf->dev_private;
1398 
1399 	if (buf->file_priv != file_priv) {
1400 		DRM_ERROR("process %d using buffer owned by %p\n",
1401 			  task_pid_nr(current), buf->file_priv);
1402 		return -EINVAL;
1403 	}
1404 	if (buf->pending) {
1405 		DRM_ERROR("sending pending buffer %d\n", elts->idx);
1406 		return -EINVAL;
1407 	}
1408 
1409 	count = (elts->end - elts->start) / sizeof(u16);
1410 	elts->start -= R128_INDEX_PRIM_OFFSET;
1411 
1412 	if (elts->start & 0x7) {
1413 		DRM_ERROR("misaligned buffer 0x%x\n", elts->start);
1414 		return -EINVAL;
1415 	}
1416 	if (elts->start < buf->used) {
1417 		DRM_ERROR("no header 0x%x - 0x%x\n", elts->start, buf->used);
1418 		return -EINVAL;
1419 	}
1420 
1421 	buf->used = elts->end;
1422 	buf_priv->prim = elts->prim;
1423 	buf_priv->discard = elts->discard;
1424 
1425 	r128_cce_dispatch_indices(dev, buf, elts->start, elts->end, count);
1426 
1427 	COMMIT_RING();
1428 	return 0;
1429 }
1430 
r128_cce_blit(struct drm_device * dev,void * data,struct drm_file * file_priv)1431 static int r128_cce_blit(struct drm_device *dev, void *data, struct drm_file *file_priv)
1432 {
1433 	struct drm_device_dma *dma = dev->dma;
1434 	drm_r128_private_t *dev_priv = dev->dev_private;
1435 	drm_r128_blit_t *blit = data;
1436 	int ret;
1437 
1438 	LOCK_TEST_WITH_RETURN(dev, file_priv);
1439 
1440 	DEV_INIT_TEST_WITH_RETURN(dev_priv);
1441 
1442 	DRM_DEBUG("pid=%d index=%d\n", task_pid_nr(current), blit->idx);
1443 
1444 	if (blit->idx < 0 || blit->idx >= dma->buf_count) {
1445 		DRM_ERROR("buffer index %d (of %d max)\n",
1446 			  blit->idx, dma->buf_count - 1);
1447 		return -EINVAL;
1448 	}
1449 
1450 	RING_SPACE_TEST_WITH_RETURN(dev_priv);
1451 	VB_AGE_TEST_WITH_RETURN(dev_priv);
1452 
1453 	ret = r128_cce_dispatch_blit(dev, file_priv, blit);
1454 
1455 	COMMIT_RING();
1456 	return ret;
1457 }
1458 
r128_cce_depth(struct drm_device * dev,void * data,struct drm_file * file_priv)1459 int r128_cce_depth(struct drm_device *dev, void *data, struct drm_file *file_priv)
1460 {
1461 	drm_r128_private_t *dev_priv = dev->dev_private;
1462 	drm_r128_depth_t *depth = data;
1463 	int ret;
1464 
1465 	LOCK_TEST_WITH_RETURN(dev, file_priv);
1466 
1467 	DEV_INIT_TEST_WITH_RETURN(dev_priv);
1468 
1469 	RING_SPACE_TEST_WITH_RETURN(dev_priv);
1470 
1471 	ret = -EINVAL;
1472 	switch (depth->func) {
1473 	case R128_WRITE_SPAN:
1474 		ret = r128_cce_dispatch_write_span(dev, depth);
1475 		break;
1476 	case R128_WRITE_PIXELS:
1477 		ret = r128_cce_dispatch_write_pixels(dev, depth);
1478 		break;
1479 	case R128_READ_SPAN:
1480 		ret = r128_cce_dispatch_read_span(dev, depth);
1481 		break;
1482 	case R128_READ_PIXELS:
1483 		ret = r128_cce_dispatch_read_pixels(dev, depth);
1484 		break;
1485 	}
1486 
1487 	COMMIT_RING();
1488 	return ret;
1489 }
1490 
r128_cce_stipple(struct drm_device * dev,void * data,struct drm_file * file_priv)1491 int r128_cce_stipple(struct drm_device *dev, void *data, struct drm_file *file_priv)
1492 {
1493 	drm_r128_private_t *dev_priv = dev->dev_private;
1494 	drm_r128_stipple_t *stipple = data;
1495 	u32 mask[32];
1496 
1497 	LOCK_TEST_WITH_RETURN(dev, file_priv);
1498 
1499 	DEV_INIT_TEST_WITH_RETURN(dev_priv);
1500 
1501 	if (copy_from_user(&mask, stipple->mask, 32 * sizeof(u32)))
1502 		return -EFAULT;
1503 
1504 	RING_SPACE_TEST_WITH_RETURN(dev_priv);
1505 
1506 	r128_cce_dispatch_stipple(dev, mask);
1507 
1508 	COMMIT_RING();
1509 	return 0;
1510 }
1511 
r128_cce_indirect(struct drm_device * dev,void * data,struct drm_file * file_priv)1512 static int r128_cce_indirect(struct drm_device *dev, void *data, struct drm_file *file_priv)
1513 {
1514 	drm_r128_private_t *dev_priv = dev->dev_private;
1515 	struct drm_device_dma *dma = dev->dma;
1516 	struct drm_buf *buf;
1517 	drm_r128_buf_priv_t *buf_priv;
1518 	drm_r128_indirect_t *indirect = data;
1519 #if 0
1520 	RING_LOCALS;
1521 #endif
1522 
1523 	LOCK_TEST_WITH_RETURN(dev, file_priv);
1524 
1525 	DEV_INIT_TEST_WITH_RETURN(dev_priv);
1526 
1527 	DRM_DEBUG("idx=%d s=%d e=%d d=%d\n",
1528 		  indirect->idx, indirect->start, indirect->end,
1529 		  indirect->discard);
1530 
1531 	if (indirect->idx < 0 || indirect->idx >= dma->buf_count) {
1532 		DRM_ERROR("buffer index %d (of %d max)\n",
1533 			  indirect->idx, dma->buf_count - 1);
1534 		return -EINVAL;
1535 	}
1536 
1537 	buf = dma->buflist[indirect->idx];
1538 	buf_priv = buf->dev_private;
1539 
1540 	if (buf->file_priv != file_priv) {
1541 		DRM_ERROR("process %d using buffer owned by %p\n",
1542 			  task_pid_nr(current), buf->file_priv);
1543 		return -EINVAL;
1544 	}
1545 	if (buf->pending) {
1546 		DRM_ERROR("sending pending buffer %d\n", indirect->idx);
1547 		return -EINVAL;
1548 	}
1549 
1550 	if (indirect->start < buf->used) {
1551 		DRM_ERROR("reusing indirect: start=0x%x actual=0x%x\n",
1552 			  indirect->start, buf->used);
1553 		return -EINVAL;
1554 	}
1555 
1556 	RING_SPACE_TEST_WITH_RETURN(dev_priv);
1557 	VB_AGE_TEST_WITH_RETURN(dev_priv);
1558 
1559 	buf->used = indirect->end;
1560 	buf_priv->discard = indirect->discard;
1561 
1562 #if 0
1563 	/* Wait for the 3D stream to idle before the indirect buffer
1564 	 * containing 2D acceleration commands is processed.
1565 	 */
1566 	BEGIN_RING(2);
1567 	RADEON_WAIT_UNTIL_3D_IDLE();
1568 	ADVANCE_RING();
1569 #endif
1570 
1571 	/* Dispatch the indirect buffer full of commands from the
1572 	 * X server.  This is insecure and is thus only available to
1573 	 * privileged clients.
1574 	 */
1575 	r128_cce_dispatch_indirect(dev, buf, indirect->start, indirect->end);
1576 
1577 	COMMIT_RING();
1578 	return 0;
1579 }
1580 
r128_getparam(struct drm_device * dev,void * data,struct drm_file * file_priv)1581 int r128_getparam(struct drm_device *dev, void *data, struct drm_file *file_priv)
1582 {
1583 	drm_r128_private_t *dev_priv = dev->dev_private;
1584 	drm_r128_getparam_t *param = data;
1585 	struct pci_dev *pdev = to_pci_dev(dev->dev);
1586 	int value;
1587 
1588 	DEV_INIT_TEST_WITH_RETURN(dev_priv);
1589 
1590 	DRM_DEBUG("pid=%d\n", task_pid_nr(current));
1591 
1592 	switch (param->param) {
1593 	case R128_PARAM_IRQ_NR:
1594 		value = pdev->irq;
1595 		break;
1596 	default:
1597 		return -EINVAL;
1598 	}
1599 
1600 	if (copy_to_user(param->value, &value, sizeof(int))) {
1601 		DRM_ERROR("copy_to_user\n");
1602 		return -EFAULT;
1603 	}
1604 
1605 	return 0;
1606 }
1607 
r128_driver_preclose(struct drm_device * dev,struct drm_file * file_priv)1608 void r128_driver_preclose(struct drm_device *dev, struct drm_file *file_priv)
1609 {
1610 	if (dev->dev_private) {
1611 		drm_r128_private_t *dev_priv = dev->dev_private;
1612 		if (dev_priv->page_flipping)
1613 			r128_do_cleanup_pageflip(dev);
1614 	}
1615 }
r128_driver_lastclose(struct drm_device * dev)1616 void r128_driver_lastclose(struct drm_device *dev)
1617 {
1618 	r128_do_cleanup_cce(dev);
1619 }
1620 
1621 const struct drm_ioctl_desc r128_ioctls[] = {
1622 	DRM_IOCTL_DEF_DRV(R128_INIT, r128_cce_init, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY),
1623 	DRM_IOCTL_DEF_DRV(R128_CCE_START, r128_cce_start, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY),
1624 	DRM_IOCTL_DEF_DRV(R128_CCE_STOP, r128_cce_stop, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY),
1625 	DRM_IOCTL_DEF_DRV(R128_CCE_RESET, r128_cce_reset, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY),
1626 	DRM_IOCTL_DEF_DRV(R128_CCE_IDLE, r128_cce_idle, DRM_AUTH),
1627 	DRM_IOCTL_DEF_DRV(R128_RESET, r128_engine_reset, DRM_AUTH),
1628 	DRM_IOCTL_DEF_DRV(R128_FULLSCREEN, r128_fullscreen, DRM_AUTH),
1629 	DRM_IOCTL_DEF_DRV(R128_SWAP, r128_cce_swap, DRM_AUTH),
1630 	DRM_IOCTL_DEF_DRV(R128_FLIP, r128_cce_flip, DRM_AUTH),
1631 	DRM_IOCTL_DEF_DRV(R128_CLEAR, r128_cce_clear, DRM_AUTH),
1632 	DRM_IOCTL_DEF_DRV(R128_VERTEX, r128_cce_vertex, DRM_AUTH),
1633 	DRM_IOCTL_DEF_DRV(R128_INDICES, r128_cce_indices, DRM_AUTH),
1634 	DRM_IOCTL_DEF_DRV(R128_BLIT, r128_cce_blit, DRM_AUTH),
1635 	DRM_IOCTL_DEF_DRV(R128_DEPTH, r128_cce_depth, DRM_AUTH),
1636 	DRM_IOCTL_DEF_DRV(R128_STIPPLE, r128_cce_stipple, DRM_AUTH),
1637 	DRM_IOCTL_DEF_DRV(R128_INDIRECT, r128_cce_indirect, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY),
1638 	DRM_IOCTL_DEF_DRV(R128_GETPARAM, r128_getparam, DRM_AUTH),
1639 };
1640 
1641 int r128_max_ioctl = ARRAY_SIZE(r128_ioctls);
1642