• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Copyright © 2009-2011 Intel Corporation
3  *
4  * Permission is hereby granted, free of charge, to any person obtaining a
5  * copy of this software and associated documentation files (the "Software"),
6  * to deal in the Software without restriction, including without limitation
7  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8  * and/or sell copies of the Software, and to permit persons to whom the
9  * Software is furnished to do so, subject to the following conditions:
10  *
11  * The above copyright notice and this permission notice (including the next
12  * paragraph) shall be included in all copies or substantial portions of the
13  * Software.
14  *
15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
18  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
20  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
21  * IN THE SOFTWARE.
22  */
23 
24 #ifdef HAVE_CONFIG_H
25 #include "config.h"
26 #endif
27 
28 #include <assert.h>
29 #include <stdint.h>
30 #include <stdlib.h>
31 #include <stdio.h>
32 #include <stdbool.h>
33 #include <stdarg.h>
34 #include <string.h>
35 
36 #include "libdrm_macros.h"
37 #include "xf86drm.h"
38 #include "intel_chipset.h"
39 #include "intel_bufmgr.h"
40 
41 
42 /* Struct for tracking drm_intel_decode state. */
43 struct drm_intel_decode {
44 	/** stdio file where the output should land.  Defaults to stdout. */
45 	FILE *out;
46 
47 	/** PCI device ID. */
48 	uint32_t devid;
49 
50 	/**
51 	 * Shorthand device identifier: 3 is 915, 4 is 965, 5 is
52 	 * Ironlake, etc.
53 	 */
54 	int gen;
55 
56 	/** GPU address of the start of the current packet. */
57 	uint32_t hw_offset;
58 	/** CPU virtual address of the start of the current packet. */
59 	uint32_t *data;
60 	/** DWORDs of remaining batchbuffer data starting from the packet. */
61 	uint32_t count;
62 
63 	/** GPU address of the start of the batchbuffer data. */
64 	uint32_t base_hw_offset;
65 	/** CPU Virtual address of the start of the batchbuffer data. */
66 	uint32_t *base_data;
67 	/** Number of DWORDs of batchbuffer data. */
68 	uint32_t base_count;
69 
70 	/** @{
71 	 * GPU head and tail pointers, which will be noted in the dump, or ~0.
72 	 */
73 	uint32_t head, tail;
74 	/** @} */
75 
76 	/**
77 	 * Whether to dump the dwords after MI_BATCHBUFFER_END.
78 	 *
79 	 * This sometimes provides clues in corrupted batchbuffers,
80 	 * and is used by the intel-gpu-tools.
81 	 */
82 	bool dump_past_end;
83 
84 	bool overflowed;
85 };
86 
87 static FILE *out;
88 static uint32_t saved_s2 = 0, saved_s4 = 0;
89 static char saved_s2_set = 0, saved_s4_set = 0;
90 static uint32_t head_offset = 0xffffffff;	/* undefined */
91 static uint32_t tail_offset = 0xffffffff;	/* undefined */
92 
93 #ifndef ARRAY_SIZE
94 #define ARRAY_SIZE(A) (sizeof(A)/sizeof(A[0]))
95 #endif
96 
97 #define BUFFER_FAIL(_count, _len, _name) do {			\
98     fprintf(out, "Buffer size too small in %s (%d < %d)\n",	\
99 	    (_name), (_count), (_len));				\
100     return _count;						\
101 } while (0)
102 
int_as_float(uint32_t intval)103 static float int_as_float(uint32_t intval)
104 {
105 	union intfloat {
106 		uint32_t i;
107 		float f;
108 	} uval;
109 
110 	uval.i = intval;
111 	return uval.f;
112 }
113 
114 static void DRM_PRINTFLIKE(3, 4)
instr_out(struct drm_intel_decode * ctx,unsigned int index,const char * fmt,...)115 instr_out(struct drm_intel_decode *ctx, unsigned int index,
116 	  const char *fmt, ...)
117 {
118 	va_list va;
119 	const char *parseinfo;
120 	uint32_t offset = ctx->hw_offset + index * 4;
121 
122 	if (index > ctx->count) {
123 		if (!ctx->overflowed) {
124 			fprintf(out, "ERROR: Decode attempted to continue beyond end of batchbuffer\n");
125 			ctx->overflowed = true;
126 		}
127 		return;
128 	}
129 
130 	if (offset == head_offset)
131 		parseinfo = "HEAD";
132 	else if (offset == tail_offset)
133 		parseinfo = "TAIL";
134 	else
135 		parseinfo = "    ";
136 
137 	fprintf(out, "0x%08x: %s 0x%08x: %s", offset, parseinfo,
138 		ctx->data[index], index == 0 ? "" : "   ");
139 	va_start(va, fmt);
140 	vfprintf(out, fmt, va);
141 	va_end(va);
142 }
143 
144 static int
decode_MI_SET_CONTEXT(struct drm_intel_decode * ctx)145 decode_MI_SET_CONTEXT(struct drm_intel_decode *ctx)
146 {
147 	uint32_t data = ctx->data[1];
148 	if (ctx->gen > 7)
149 		return 1;
150 
151 	instr_out(ctx, 0, "MI_SET_CONTEXT\n");
152 	instr_out(ctx, 1, "gtt offset = 0x%x%s%s\n",
153 		  data & ~0xfff,
154 		  data & (1<<1)? ", Force Restore": "",
155 		  data & (1<<0)? ", Restore Inhibit": "");
156 
157 	return 2;
158 }
159 
160 static int
decode_MI_WAIT_FOR_EVENT(struct drm_intel_decode * ctx)161 decode_MI_WAIT_FOR_EVENT(struct drm_intel_decode *ctx)
162 {
163 	const char *cc_wait;
164 	int cc_shift = 0;
165 	uint32_t data = ctx->data[0];
166 
167 	if (ctx->gen <= 5)
168 		cc_shift = 9;
169 	else
170 		cc_shift = 16;
171 
172 	switch ((data >> cc_shift) & 0x1f) {
173 	case 1:
174 		cc_wait = ", cc wait 1";
175 		break;
176 	case 2:
177 		cc_wait = ", cc wait 2";
178 		break;
179 	case 3:
180 		cc_wait = ", cc wait 3";
181 		break;
182 	case 4:
183 		cc_wait = ", cc wait 4";
184 		break;
185 	case 5:
186 		cc_wait = ", cc wait 4";
187 		break;
188 	default:
189 		cc_wait = "";
190 		break;
191 	}
192 
193 	if (ctx->gen <= 5) {
194 		instr_out(ctx, 0, "MI_WAIT_FOR_EVENT%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
195 			  data & (1<<18)? ", pipe B start vblank wait": "",
196 			  data & (1<<17)? ", pipe A start vblank wait": "",
197 			  data & (1<<16)? ", overlay flip pending wait": "",
198 			  data & (1<<14)? ", pipe B hblank wait": "",
199 			  data & (1<<13)? ", pipe A hblank wait": "",
200 			  cc_wait,
201 			  data & (1<<8)? ", plane C pending flip wait": "",
202 			  data & (1<<7)? ", pipe B vblank wait": "",
203 			  data & (1<<6)? ", plane B pending flip wait": "",
204 			  data & (1<<5)? ", pipe B scan line wait": "",
205 			  data & (1<<4)? ", fbc idle wait": "",
206 			  data & (1<<3)? ", pipe A vblank wait": "",
207 			  data & (1<<2)? ", plane A pending flip wait": "",
208 			  data & (1<<1)? ", plane A scan line wait": "");
209 	} else {
210 		instr_out(ctx, 0, "MI_WAIT_FOR_EVENT%s%s%s%s%s%s%s%s%s%s%s%s\n",
211 			  data & (1<<20)? ", sprite C pending flip wait": "", /* ivb */
212 			  cc_wait,
213 			  data & (1<<13)? ", pipe B hblank wait": "",
214 			  data & (1<<11)? ", pipe B vblank wait": "",
215 			  data & (1<<10)? ", sprite B pending flip wait": "",
216 			  data & (1<<9)? ", plane B pending flip wait": "",
217 			  data & (1<<8)? ", plane B scan line wait": "",
218 			  data & (1<<5)? ", pipe A hblank wait": "",
219 			  data & (1<<3)? ", pipe A vblank wait": "",
220 			  data & (1<<2)? ", sprite A pending flip wait": "",
221 			  data & (1<<1)? ", plane A pending flip wait": "",
222 			  data & (1<<0)? ", plane A scan line wait": "");
223 	}
224 
225 	return 1;
226 }
227 
228 static int
decode_mi(struct drm_intel_decode * ctx)229 decode_mi(struct drm_intel_decode *ctx)
230 {
231 	unsigned int opcode, len = -1;
232 	const char *post_sync_op = "";
233 	uint32_t *data = ctx->data;
234 
235 	struct {
236 		uint32_t opcode;
237 		int len_mask;
238 		unsigned int min_len;
239 		unsigned int max_len;
240 		const char *name;
241 		int (*func)(struct drm_intel_decode *ctx);
242 	} opcodes_mi[] = {
243 		{ 0x08, 0, 1, 1, "MI_ARB_ON_OFF" },
244 		{ 0x0a, 0, 1, 1, "MI_BATCH_BUFFER_END" },
245 		{ 0x30, 0x3f, 3, 3, "MI_BATCH_BUFFER" },
246 		{ 0x31, 0x3f, 2, 2, "MI_BATCH_BUFFER_START" },
247 		{ 0x14, 0x3f, 3, 3, "MI_DISPLAY_BUFFER_INFO" },
248 		{ 0x04, 0, 1, 1, "MI_FLUSH" },
249 		{ 0x22, 0x1f, 3, 3, "MI_LOAD_REGISTER_IMM" },
250 		{ 0x13, 0x3f, 2, 2, "MI_LOAD_SCAN_LINES_EXCL" },
251 		{ 0x12, 0x3f, 2, 2, "MI_LOAD_SCAN_LINES_INCL" },
252 		{ 0x00, 0, 1, 1, "MI_NOOP" },
253 		{ 0x11, 0x3f, 2, 2, "MI_OVERLAY_FLIP" },
254 		{ 0x07, 0, 1, 1, "MI_REPORT_HEAD" },
255 		{ 0x18, 0x3f, 2, 2, "MI_SET_CONTEXT", decode_MI_SET_CONTEXT },
256 		{ 0x20, 0x3f, 3, 4, "MI_STORE_DATA_IMM" },
257 		{ 0x21, 0x3f, 3, 4, "MI_STORE_DATA_INDEX" },
258 		{ 0x24, 0x3f, 3, 3, "MI_STORE_REGISTER_MEM" },
259 		{ 0x02, 0, 1, 1, "MI_USER_INTERRUPT" },
260 		{ 0x03, 0, 1, 1, "MI_WAIT_FOR_EVENT", decode_MI_WAIT_FOR_EVENT },
261 		{ 0x16, 0x7f, 3, 3, "MI_SEMAPHORE_MBOX" },
262 		{ 0x26, 0x1f, 3, 4, "MI_FLUSH_DW" },
263 		{ 0x28, 0x3f, 3, 3, "MI_REPORT_PERF_COUNT" },
264 		{ 0x29, 0xff, 3, 3, "MI_LOAD_REGISTER_MEM" },
265 		{ 0x0b, 0, 1, 1, "MI_SUSPEND_FLUSH"},
266 	}, *opcode_mi = NULL;
267 
268 	/* check instruction length */
269 	for (opcode = 0; opcode < sizeof(opcodes_mi) / sizeof(opcodes_mi[0]);
270 	     opcode++) {
271 		if ((data[0] & 0x1f800000) >> 23 == opcodes_mi[opcode].opcode) {
272 			len = 1;
273 			if (opcodes_mi[opcode].max_len > 1) {
274 				len =
275 				    (data[0] & opcodes_mi[opcode].len_mask) + 2;
276 				if (len < opcodes_mi[opcode].min_len
277 				    || len > opcodes_mi[opcode].max_len) {
278 					fprintf(out,
279 						"Bad length (%d) in %s, [%d, %d]\n",
280 						len, opcodes_mi[opcode].name,
281 						opcodes_mi[opcode].min_len,
282 						opcodes_mi[opcode].max_len);
283 				}
284 			}
285 			opcode_mi = &opcodes_mi[opcode];
286 			break;
287 		}
288 	}
289 
290 	if (opcode_mi && opcode_mi->func)
291 		return opcode_mi->func(ctx);
292 
293 	switch ((data[0] & 0x1f800000) >> 23) {
294 	case 0x0a:
295 		instr_out(ctx, 0, "MI_BATCH_BUFFER_END\n");
296 		return -1;
297 	case 0x16:
298 		instr_out(ctx, 0, "MI_SEMAPHORE_MBOX%s%s%s%s %u\n",
299 			  data[0] & (1 << 22) ? " global gtt," : "",
300 			  data[0] & (1 << 21) ? " update semaphore," : "",
301 			  data[0] & (1 << 20) ? " compare semaphore," : "",
302 			  data[0] & (1 << 18) ? " use compare reg" : "",
303 			  (data[0] & (0x3 << 16)) >> 16);
304 		instr_out(ctx, 1, "value\n");
305 		instr_out(ctx, 2, "address\n");
306 		return len;
307 	case 0x21:
308 		instr_out(ctx, 0, "MI_STORE_DATA_INDEX%s\n",
309 			  data[0] & (1 << 21) ? " use per-process HWS," : "");
310 		instr_out(ctx, 1, "index\n");
311 		instr_out(ctx, 2, "dword\n");
312 		if (len == 4)
313 			instr_out(ctx, 3, "upper dword\n");
314 		return len;
315 	case 0x00:
316 		if (data[0] & (1 << 22))
317 			instr_out(ctx, 0,
318 				  "MI_NOOP write NOPID reg, val=0x%x\n",
319 				  data[0] & ((1 << 22) - 1));
320 		else
321 			instr_out(ctx, 0, "MI_NOOP\n");
322 		return len;
323 	case 0x26:
324 		switch (data[0] & (0x3 << 14)) {
325 		case (0 << 14):
326 			post_sync_op = "no write";
327 			break;
328 		case (1 << 14):
329 			post_sync_op = "write data";
330 			break;
331 		case (2 << 14):
332 			post_sync_op = "reserved";
333 			break;
334 		case (3 << 14):
335 			post_sync_op = "write TIMESTAMP";
336 			break;
337 		}
338 		instr_out(ctx, 0,
339 			  "MI_FLUSH_DW%s%s%s%s post_sync_op='%s' %s%s\n",
340 			  data[0] & (1 << 22) ?
341 			  " enable protected mem (BCS-only)," : "",
342 			  data[0] & (1 << 21) ? " store in hws," : "",
343 			  data[0] & (1 << 18) ? " invalidate tlb," : "",
344 			  data[0] & (1 << 17) ? " flush gfdt," : "",
345 			  post_sync_op,
346 			  data[0] & (1 << 8) ? " enable notify interrupt," : "",
347 			  data[0] & (1 << 7) ?
348 			  " invalidate video state (BCS-only)," : "");
349 		if (data[0] & (1 << 21))
350 			instr_out(ctx, 1, "hws index\n");
351 		else
352 			instr_out(ctx, 1, "address\n");
353 		instr_out(ctx, 2, "dword\n");
354 		if (len == 4)
355 			instr_out(ctx, 3, "upper dword\n");
356 		return len;
357 	}
358 
359 	for (opcode = 0; opcode < sizeof(opcodes_mi) / sizeof(opcodes_mi[0]);
360 	     opcode++) {
361 		if ((data[0] & 0x1f800000) >> 23 == opcodes_mi[opcode].opcode) {
362 			unsigned int i;
363 
364 			instr_out(ctx, 0, "%s\n",
365 				  opcodes_mi[opcode].name);
366 			for (i = 1; i < len; i++) {
367 				instr_out(ctx, i, "dword %d\n", i);
368 			}
369 
370 			return len;
371 		}
372 	}
373 
374 	instr_out(ctx, 0, "MI UNKNOWN\n");
375 	return 1;
376 }
377 
378 static void
decode_2d_br00(struct drm_intel_decode * ctx,const char * cmd)379 decode_2d_br00(struct drm_intel_decode *ctx, const char *cmd)
380 {
381 	instr_out(ctx, 0,
382 		  "%s (rgb %sabled, alpha %sabled, src tile %d, dst tile %d)\n",
383 		  cmd,
384 		  (ctx->data[0] & (1 << 20)) ? "en" : "dis",
385 		  (ctx->data[0] & (1 << 21)) ? "en" : "dis",
386 		  (ctx->data[0] >> 15) & 1,
387 		  (ctx->data[0] >> 11) & 1);
388 }
389 
390 static void
decode_2d_br01(struct drm_intel_decode * ctx)391 decode_2d_br01(struct drm_intel_decode *ctx)
392 {
393 	const char *format;
394 	switch ((ctx->data[1] >> 24) & 0x3) {
395 	case 0:
396 		format = "8";
397 		break;
398 	case 1:
399 		format = "565";
400 		break;
401 	case 2:
402 		format = "1555";
403 		break;
404 	case 3:
405 		format = "8888";
406 		break;
407 	}
408 
409 	instr_out(ctx, 1,
410 		  "format %s, pitch %d, rop 0x%02x, "
411 		  "clipping %sabled, %s%s \n",
412 		  format,
413 		  (short)(ctx->data[1] & 0xffff),
414 		  (ctx->data[1] >> 16) & 0xff,
415 		  ctx->data[1] & (1 << 30) ? "en" : "dis",
416 		  ctx->data[1] & (1 << 31) ? "solid pattern enabled, " : "",
417 		  ctx->data[1] & (1 << 31) ?
418 		  "mono pattern transparency enabled, " : "");
419 
420 }
421 
422 static int
decode_2d(struct drm_intel_decode * ctx)423 decode_2d(struct drm_intel_decode *ctx)
424 {
425 	unsigned int opcode, len;
426 	uint32_t *data = ctx->data;
427 
428 	struct {
429 		uint32_t opcode;
430 		unsigned int min_len;
431 		unsigned int max_len;
432 		const char *name;
433 	} opcodes_2d[] = {
434 		{ 0x40, 5, 5, "COLOR_BLT" },
435 		{ 0x43, 6, 6, "SRC_COPY_BLT" },
436 		{ 0x01, 8, 8, "XY_SETUP_BLT" },
437 		{ 0x11, 9, 9, "XY_SETUP_MONO_PATTERN_SL_BLT" },
438 		{ 0x03, 3, 3, "XY_SETUP_CLIP_BLT" },
439 		{ 0x24, 2, 2, "XY_PIXEL_BLT" },
440 		{ 0x25, 3, 3, "XY_SCANLINES_BLT" },
441 		{ 0x26, 4, 4, "Y_TEXT_BLT" },
442 		{ 0x31, 5, 134, "XY_TEXT_IMMEDIATE_BLT" },
443 		{ 0x50, 6, 6, "XY_COLOR_BLT" },
444 		{ 0x51, 6, 6, "XY_PAT_BLT" },
445 		{ 0x76, 8, 8, "XY_PAT_CHROMA_BLT" },
446 		{ 0x72, 7, 135, "XY_PAT_BLT_IMMEDIATE" },
447 		{ 0x77, 9, 137, "XY_PAT_CHROMA_BLT_IMMEDIATE" },
448 		{ 0x52, 9, 9, "XY_MONO_PAT_BLT" },
449 		{ 0x59, 7, 7, "XY_MONO_PAT_FIXED_BLT" },
450 		{ 0x53, 8, 8, "XY_SRC_COPY_BLT" },
451 		{ 0x54, 8, 8, "XY_MONO_SRC_COPY_BLT" },
452 		{ 0x71, 9, 137, "XY_MONO_SRC_COPY_IMMEDIATE_BLT" },
453 		{ 0x55, 9, 9, "XY_FULL_BLT" },
454 		{ 0x55, 9, 137, "XY_FULL_IMMEDIATE_PATTERN_BLT" },
455 		{ 0x56, 9, 9, "XY_FULL_MONO_SRC_BLT" },
456 		{ 0x75, 10, 138, "XY_FULL_MONO_SRC_IMMEDIATE_PATTERN_BLT" },
457 		{ 0x57, 12, 12, "XY_FULL_MONO_PATTERN_BLT" },
458 		{ 0x58, 12, 12, "XY_FULL_MONO_PATTERN_MONO_SRC_BLT"},
459 	};
460 
461 	switch ((data[0] & 0x1fc00000) >> 22) {
462 	case 0x25:
463 		instr_out(ctx, 0,
464 			  "XY_SCANLINES_BLT (pattern seed (%d, %d), dst tile %d)\n",
465 			  (data[0] >> 12) & 0x8,
466 			  (data[0] >> 8) & 0x8, (data[0] >> 11) & 1);
467 
468 		len = (data[0] & 0x000000ff) + 2;
469 		if (len != 3)
470 			fprintf(out, "Bad count in XY_SCANLINES_BLT\n");
471 
472 		instr_out(ctx, 1, "dest (%d,%d)\n",
473 			  data[1] & 0xffff, data[1] >> 16);
474 		instr_out(ctx, 2, "dest (%d,%d)\n",
475 			  data[2] & 0xffff, data[2] >> 16);
476 		return len;
477 	case 0x01:
478 		decode_2d_br00(ctx, "XY_SETUP_BLT");
479 
480 		len = (data[0] & 0x000000ff) + 2;
481 		if (len != 8)
482 			fprintf(out, "Bad count in XY_SETUP_BLT\n");
483 
484 		decode_2d_br01(ctx);
485 		instr_out(ctx, 2, "cliprect (%d,%d)\n",
486 			  data[2] & 0xffff, data[2] >> 16);
487 		instr_out(ctx, 3, "cliprect (%d,%d)\n",
488 			  data[3] & 0xffff, data[3] >> 16);
489 		instr_out(ctx, 4, "setup dst offset 0x%08x\n",
490 			  data[4]);
491 		instr_out(ctx, 5, "setup background color\n");
492 		instr_out(ctx, 6, "setup foreground color\n");
493 		instr_out(ctx, 7, "color pattern offset\n");
494 		return len;
495 	case 0x03:
496 		decode_2d_br00(ctx, "XY_SETUP_CLIP_BLT");
497 
498 		len = (data[0] & 0x000000ff) + 2;
499 		if (len != 3)
500 			fprintf(out, "Bad count in XY_SETUP_CLIP_BLT\n");
501 
502 		instr_out(ctx, 1, "cliprect (%d,%d)\n",
503 			  data[1] & 0xffff, data[2] >> 16);
504 		instr_out(ctx, 2, "cliprect (%d,%d)\n",
505 			  data[2] & 0xffff, data[3] >> 16);
506 		return len;
507 	case 0x11:
508 		decode_2d_br00(ctx, "XY_SETUP_MONO_PATTERN_SL_BLT");
509 
510 		len = (data[0] & 0x000000ff) + 2;
511 		if (len != 9)
512 			fprintf(out,
513 				"Bad count in XY_SETUP_MONO_PATTERN_SL_BLT\n");
514 
515 		decode_2d_br01(ctx);
516 		instr_out(ctx, 2, "cliprect (%d,%d)\n",
517 			  data[2] & 0xffff, data[2] >> 16);
518 		instr_out(ctx, 3, "cliprect (%d,%d)\n",
519 			  data[3] & 0xffff, data[3] >> 16);
520 		instr_out(ctx, 4, "setup dst offset 0x%08x\n",
521 			  data[4]);
522 		instr_out(ctx, 5, "setup background color\n");
523 		instr_out(ctx, 6, "setup foreground color\n");
524 		instr_out(ctx, 7, "mono pattern dw0\n");
525 		instr_out(ctx, 8, "mono pattern dw1\n");
526 		return len;
527 	case 0x50:
528 		decode_2d_br00(ctx, "XY_COLOR_BLT");
529 
530 		len = (data[0] & 0x000000ff) + 2;
531 		if (len != 6)
532 			fprintf(out, "Bad count in XY_COLOR_BLT\n");
533 
534 		decode_2d_br01(ctx);
535 		instr_out(ctx, 2, "(%d,%d)\n",
536 			  data[2] & 0xffff, data[2] >> 16);
537 		instr_out(ctx, 3, "(%d,%d)\n",
538 			  data[3] & 0xffff, data[3] >> 16);
539 		instr_out(ctx, 4, "offset 0x%08x\n", data[4]);
540 		instr_out(ctx, 5, "color\n");
541 		return len;
542 	case 0x53:
543 		decode_2d_br00(ctx, "XY_SRC_COPY_BLT");
544 
545 		len = (data[0] & 0x000000ff) + 2;
546 		if (len != 8)
547 			fprintf(out, "Bad count in XY_SRC_COPY_BLT\n");
548 
549 		decode_2d_br01(ctx);
550 		instr_out(ctx, 2, "dst (%d,%d)\n",
551 			  data[2] & 0xffff, data[2] >> 16);
552 		instr_out(ctx, 3, "dst (%d,%d)\n",
553 			  data[3] & 0xffff, data[3] >> 16);
554 		instr_out(ctx, 4, "dst offset 0x%08x\n", data[4]);
555 		instr_out(ctx, 5, "src (%d,%d)\n",
556 			  data[5] & 0xffff, data[5] >> 16);
557 		instr_out(ctx, 6, "src pitch %d\n",
558 			  (short)(data[6] & 0xffff));
559 		instr_out(ctx, 7, "src offset 0x%08x\n", data[7]);
560 		return len;
561 	}
562 
563 	for (opcode = 0; opcode < sizeof(opcodes_2d) / sizeof(opcodes_2d[0]);
564 	     opcode++) {
565 		if ((data[0] & 0x1fc00000) >> 22 == opcodes_2d[opcode].opcode) {
566 			unsigned int i;
567 
568 			len = 1;
569 			instr_out(ctx, 0, "%s\n",
570 				  opcodes_2d[opcode].name);
571 			if (opcodes_2d[opcode].max_len > 1) {
572 				len = (data[0] & 0x000000ff) + 2;
573 				if (len < opcodes_2d[opcode].min_len ||
574 				    len > opcodes_2d[opcode].max_len) {
575 					fprintf(out, "Bad count in %s\n",
576 						opcodes_2d[opcode].name);
577 				}
578 			}
579 
580 			for (i = 1; i < len; i++) {
581 				instr_out(ctx, i, "dword %d\n", i);
582 			}
583 
584 			return len;
585 		}
586 	}
587 
588 	instr_out(ctx, 0, "2D UNKNOWN\n");
589 	return 1;
590 }
591 
592 static int
decode_3d_1c(struct drm_intel_decode * ctx)593 decode_3d_1c(struct drm_intel_decode *ctx)
594 {
595 	uint32_t *data = ctx->data;
596 	uint32_t opcode;
597 
598 	opcode = (data[0] & 0x00f80000) >> 19;
599 
600 	switch (opcode) {
601 	case 0x11:
602 		instr_out(ctx, 0,
603 			  "3DSTATE_DEPTH_SUBRECTANGLE_DISABLE\n");
604 		return 1;
605 	case 0x10:
606 		instr_out(ctx, 0, "3DSTATE_SCISSOR_ENABLE %s\n",
607 			  data[0] & 1 ? "enabled" : "disabled");
608 		return 1;
609 	case 0x01:
610 		instr_out(ctx, 0, "3DSTATE_MAP_COORD_SET_I830\n");
611 		return 1;
612 	case 0x0a:
613 		instr_out(ctx, 0, "3DSTATE_MAP_CUBE_I830\n");
614 		return 1;
615 	case 0x05:
616 		instr_out(ctx, 0, "3DSTATE_MAP_TEX_STREAM_I830\n");
617 		return 1;
618 	}
619 
620 	instr_out(ctx, 0, "3D UNKNOWN: 3d_1c opcode = 0x%x\n",
621 		  opcode);
622 	return 1;
623 }
624 
625 /** Sets the string dstname to describe the destination of the PS instruction */
626 static void
i915_get_instruction_dst(uint32_t * data,int i,char * dstname,int do_mask)627 i915_get_instruction_dst(uint32_t *data, int i, char *dstname, int do_mask)
628 {
629 	uint32_t a0 = data[i];
630 	int dst_nr = (a0 >> 14) & 0xf;
631 	char dstmask[8];
632 	const char *sat;
633 
634 	if (do_mask) {
635 		if (((a0 >> 10) & 0xf) == 0xf) {
636 			dstmask[0] = 0;
637 		} else {
638 			int dstmask_index = 0;
639 
640 			dstmask[dstmask_index++] = '.';
641 			if (a0 & (1 << 10))
642 				dstmask[dstmask_index++] = 'x';
643 			if (a0 & (1 << 11))
644 				dstmask[dstmask_index++] = 'y';
645 			if (a0 & (1 << 12))
646 				dstmask[dstmask_index++] = 'z';
647 			if (a0 & (1 << 13))
648 				dstmask[dstmask_index++] = 'w';
649 			dstmask[dstmask_index++] = 0;
650 		}
651 
652 		if (a0 & (1 << 22))
653 			sat = ".sat";
654 		else
655 			sat = "";
656 	} else {
657 		dstmask[0] = 0;
658 		sat = "";
659 	}
660 
661 	switch ((a0 >> 19) & 0x7) {
662 	case 0:
663 		if (dst_nr > 15)
664 			fprintf(out, "bad destination reg R%d\n", dst_nr);
665 		sprintf(dstname, "R%d%s%s", dst_nr, dstmask, sat);
666 		break;
667 	case 4:
668 		if (dst_nr > 0)
669 			fprintf(out, "bad destination reg oC%d\n", dst_nr);
670 		sprintf(dstname, "oC%s%s", dstmask, sat);
671 		break;
672 	case 5:
673 		if (dst_nr > 0)
674 			fprintf(out, "bad destination reg oD%d\n", dst_nr);
675 		sprintf(dstname, "oD%s%s", dstmask, sat);
676 		break;
677 	case 6:
678 		if (dst_nr > 3)
679 			fprintf(out, "bad destination reg U%d\n", dst_nr);
680 		sprintf(dstname, "U%d%s%s", dst_nr, dstmask, sat);
681 		break;
682 	default:
683 		sprintf(dstname, "RESERVED");
684 		break;
685 	}
686 }
687 
688 static const char *
i915_get_channel_swizzle(uint32_t select)689 i915_get_channel_swizzle(uint32_t select)
690 {
691 	switch (select & 0x7) {
692 	case 0:
693 		return (select & 8) ? "-x" : "x";
694 	case 1:
695 		return (select & 8) ? "-y" : "y";
696 	case 2:
697 		return (select & 8) ? "-z" : "z";
698 	case 3:
699 		return (select & 8) ? "-w" : "w";
700 	case 4:
701 		return (select & 8) ? "-0" : "0";
702 	case 5:
703 		return (select & 8) ? "-1" : "1";
704 	default:
705 		return (select & 8) ? "-bad" : "bad";
706 	}
707 }
708 
709 static void
i915_get_instruction_src_name(uint32_t src_type,uint32_t src_nr,char * name)710 i915_get_instruction_src_name(uint32_t src_type, uint32_t src_nr, char *name)
711 {
712 	switch (src_type) {
713 	case 0:
714 		sprintf(name, "R%d", src_nr);
715 		if (src_nr > 15)
716 			fprintf(out, "bad src reg %s\n", name);
717 		break;
718 	case 1:
719 		if (src_nr < 8)
720 			sprintf(name, "T%d", src_nr);
721 		else if (src_nr == 8)
722 			sprintf(name, "DIFFUSE");
723 		else if (src_nr == 9)
724 			sprintf(name, "SPECULAR");
725 		else if (src_nr == 10)
726 			sprintf(name, "FOG");
727 		else {
728 			fprintf(out, "bad src reg T%d\n", src_nr);
729 			sprintf(name, "RESERVED");
730 		}
731 		break;
732 	case 2:
733 		sprintf(name, "C%d", src_nr);
734 		if (src_nr > 31)
735 			fprintf(out, "bad src reg %s\n", name);
736 		break;
737 	case 4:
738 		sprintf(name, "oC");
739 		if (src_nr > 0)
740 			fprintf(out, "bad src reg oC%d\n", src_nr);
741 		break;
742 	case 5:
743 		sprintf(name, "oD");
744 		if (src_nr > 0)
745 			fprintf(out, "bad src reg oD%d\n", src_nr);
746 		break;
747 	case 6:
748 		sprintf(name, "U%d", src_nr);
749 		if (src_nr > 3)
750 			fprintf(out, "bad src reg %s\n", name);
751 		break;
752 	default:
753 		fprintf(out, "bad src reg type %d\n", src_type);
754 		sprintf(name, "RESERVED");
755 		break;
756 	}
757 }
758 
i915_get_instruction_src0(uint32_t * data,int i,char * srcname)759 static void i915_get_instruction_src0(uint32_t *data, int i, char *srcname)
760 {
761 	uint32_t a0 = data[i];
762 	uint32_t a1 = data[i + 1];
763 	int src_nr = (a0 >> 2) & 0x1f;
764 	const char *swizzle_x = i915_get_channel_swizzle((a1 >> 28) & 0xf);
765 	const char *swizzle_y = i915_get_channel_swizzle((a1 >> 24) & 0xf);
766 	const char *swizzle_z = i915_get_channel_swizzle((a1 >> 20) & 0xf);
767 	const char *swizzle_w = i915_get_channel_swizzle((a1 >> 16) & 0xf);
768 	char swizzle[100];
769 
770 	i915_get_instruction_src_name((a0 >> 7) & 0x7, src_nr, srcname);
771 	sprintf(swizzle, ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z,
772 		swizzle_w);
773 	if (strcmp(swizzle, ".xyzw") != 0)
774 		strcat(srcname, swizzle);
775 }
776 
i915_get_instruction_src1(uint32_t * data,int i,char * srcname)777 static void i915_get_instruction_src1(uint32_t *data, int i, char *srcname)
778 {
779 	uint32_t a1 = data[i + 1];
780 	uint32_t a2 = data[i + 2];
781 	int src_nr = (a1 >> 8) & 0x1f;
782 	const char *swizzle_x = i915_get_channel_swizzle((a1 >> 4) & 0xf);
783 	const char *swizzle_y = i915_get_channel_swizzle((a1 >> 0) & 0xf);
784 	const char *swizzle_z = i915_get_channel_swizzle((a2 >> 28) & 0xf);
785 	const char *swizzle_w = i915_get_channel_swizzle((a2 >> 24) & 0xf);
786 	char swizzle[100];
787 
788 	i915_get_instruction_src_name((a1 >> 13) & 0x7, src_nr, srcname);
789 	sprintf(swizzle, ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z,
790 		swizzle_w);
791 	if (strcmp(swizzle, ".xyzw") != 0)
792 		strcat(srcname, swizzle);
793 }
794 
i915_get_instruction_src2(uint32_t * data,int i,char * srcname)795 static void i915_get_instruction_src2(uint32_t *data, int i, char *srcname)
796 {
797 	uint32_t a2 = data[i + 2];
798 	int src_nr = (a2 >> 16) & 0x1f;
799 	const char *swizzle_x = i915_get_channel_swizzle((a2 >> 12) & 0xf);
800 	const char *swizzle_y = i915_get_channel_swizzle((a2 >> 8) & 0xf);
801 	const char *swizzle_z = i915_get_channel_swizzle((a2 >> 4) & 0xf);
802 	const char *swizzle_w = i915_get_channel_swizzle((a2 >> 0) & 0xf);
803 	char swizzle[100];
804 
805 	i915_get_instruction_src_name((a2 >> 21) & 0x7, src_nr, srcname);
806 	sprintf(swizzle, ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z,
807 		swizzle_w);
808 	if (strcmp(swizzle, ".xyzw") != 0)
809 		strcat(srcname, swizzle);
810 }
811 
812 static void
i915_get_instruction_addr(uint32_t src_type,uint32_t src_nr,char * name)813 i915_get_instruction_addr(uint32_t src_type, uint32_t src_nr, char *name)
814 {
815 	switch (src_type) {
816 	case 0:
817 		sprintf(name, "R%d", src_nr);
818 		if (src_nr > 15)
819 			fprintf(out, "bad src reg %s\n", name);
820 		break;
821 	case 1:
822 		if (src_nr < 8)
823 			sprintf(name, "T%d", src_nr);
824 		else if (src_nr == 8)
825 			sprintf(name, "DIFFUSE");
826 		else if (src_nr == 9)
827 			sprintf(name, "SPECULAR");
828 		else if (src_nr == 10)
829 			sprintf(name, "FOG");
830 		else {
831 			fprintf(out, "bad src reg T%d\n", src_nr);
832 			sprintf(name, "RESERVED");
833 		}
834 		break;
835 	case 4:
836 		sprintf(name, "oC");
837 		if (src_nr > 0)
838 			fprintf(out, "bad src reg oC%d\n", src_nr);
839 		break;
840 	case 5:
841 		sprintf(name, "oD");
842 		if (src_nr > 0)
843 			fprintf(out, "bad src reg oD%d\n", src_nr);
844 		break;
845 	default:
846 		fprintf(out, "bad src reg type %d\n", src_type);
847 		sprintf(name, "RESERVED");
848 		break;
849 	}
850 }
851 
852 static void
i915_decode_alu1(struct drm_intel_decode * ctx,int i,char * instr_prefix,const char * op_name)853 i915_decode_alu1(struct drm_intel_decode *ctx,
854 		 int i, char *instr_prefix, const char *op_name)
855 {
856 	char dst[100], src0[100];
857 
858 	i915_get_instruction_dst(ctx->data, i, dst, 1);
859 	i915_get_instruction_src0(ctx->data, i, src0);
860 
861 	instr_out(ctx, i++, "%s: %s %s, %s\n", instr_prefix,
862 		  op_name, dst, src0);
863 	instr_out(ctx, i++, "%s\n", instr_prefix);
864 	instr_out(ctx, i++, "%s\n", instr_prefix);
865 }
866 
867 static void
i915_decode_alu2(struct drm_intel_decode * ctx,int i,char * instr_prefix,const char * op_name)868 i915_decode_alu2(struct drm_intel_decode *ctx,
869 		 int i, char *instr_prefix, const char *op_name)
870 {
871 	char dst[100], src0[100], src1[100];
872 
873 	i915_get_instruction_dst(ctx->data, i, dst, 1);
874 	i915_get_instruction_src0(ctx->data, i, src0);
875 	i915_get_instruction_src1(ctx->data, i, src1);
876 
877 	instr_out(ctx, i++, "%s: %s %s, %s, %s\n", instr_prefix,
878 		  op_name, dst, src0, src1);
879 	instr_out(ctx, i++, "%s\n", instr_prefix);
880 	instr_out(ctx, i++, "%s\n", instr_prefix);
881 }
882 
883 static void
i915_decode_alu3(struct drm_intel_decode * ctx,int i,char * instr_prefix,const char * op_name)884 i915_decode_alu3(struct drm_intel_decode *ctx,
885 		 int i, char *instr_prefix, const char *op_name)
886 {
887 	char dst[100], src0[100], src1[100], src2[100];
888 
889 	i915_get_instruction_dst(ctx->data, i, dst, 1);
890 	i915_get_instruction_src0(ctx->data, i, src0);
891 	i915_get_instruction_src1(ctx->data, i, src1);
892 	i915_get_instruction_src2(ctx->data, i, src2);
893 
894 	instr_out(ctx, i++, "%s: %s %s, %s, %s, %s\n", instr_prefix,
895 		  op_name, dst, src0, src1, src2);
896 	instr_out(ctx, i++, "%s\n", instr_prefix);
897 	instr_out(ctx, i++, "%s\n", instr_prefix);
898 }
899 
900 static void
i915_decode_tex(struct drm_intel_decode * ctx,int i,const char * instr_prefix,const char * tex_name)901 i915_decode_tex(struct drm_intel_decode *ctx, int i,
902 		const char *instr_prefix, const char *tex_name)
903 {
904 	uint32_t t0 = ctx->data[i];
905 	uint32_t t1 = ctx->data[i + 1];
906 	char dst_name[100];
907 	char addr_name[100];
908 	int sampler_nr;
909 
910 	i915_get_instruction_dst(ctx->data, i, dst_name, 0);
911 	i915_get_instruction_addr((t1 >> 24) & 0x7,
912 				  (t1 >> 17) & 0xf, addr_name);
913 	sampler_nr = t0 & 0xf;
914 
915 	instr_out(ctx, i++, "%s: %s %s, S%d, %s\n", instr_prefix,
916 		  tex_name, dst_name, sampler_nr, addr_name);
917 	instr_out(ctx, i++, "%s\n", instr_prefix);
918 	instr_out(ctx, i++, "%s\n", instr_prefix);
919 }
920 
921 static void
i915_decode_dcl(struct drm_intel_decode * ctx,int i,char * instr_prefix)922 i915_decode_dcl(struct drm_intel_decode *ctx, int i, char *instr_prefix)
923 {
924 	uint32_t d0 = ctx->data[i];
925 	const char *sampletype;
926 	int dcl_nr = (d0 >> 14) & 0xf;
927 	const char *dcl_x = d0 & (1 << 10) ? "x" : "";
928 	const char *dcl_y = d0 & (1 << 11) ? "y" : "";
929 	const char *dcl_z = d0 & (1 << 12) ? "z" : "";
930 	const char *dcl_w = d0 & (1 << 13) ? "w" : "";
931 	char dcl_mask[10];
932 
933 	switch ((d0 >> 19) & 0x3) {
934 	case 1:
935 		sprintf(dcl_mask, ".%s%s%s%s", dcl_x, dcl_y, dcl_z, dcl_w);
936 		if (strcmp(dcl_mask, ".") == 0)
937 			fprintf(out, "bad (empty) dcl mask\n");
938 
939 		if (dcl_nr > 10)
940 			fprintf(out, "bad T%d dcl register number\n", dcl_nr);
941 		if (dcl_nr < 8) {
942 			if (strcmp(dcl_mask, ".x") != 0 &&
943 			    strcmp(dcl_mask, ".xy") != 0 &&
944 			    strcmp(dcl_mask, ".xz") != 0 &&
945 			    strcmp(dcl_mask, ".w") != 0 &&
946 			    strcmp(dcl_mask, ".xyzw") != 0) {
947 				fprintf(out, "bad T%d.%s dcl mask\n", dcl_nr,
948 					dcl_mask);
949 			}
950 			instr_out(ctx, i++, "%s: DCL T%d%s\n",
951 				  instr_prefix, dcl_nr, dcl_mask);
952 		} else {
953 			if (strcmp(dcl_mask, ".xz") == 0)
954 				fprintf(out, "errataed bad dcl mask %s\n",
955 					dcl_mask);
956 			else if (strcmp(dcl_mask, ".xw") == 0)
957 				fprintf(out, "errataed bad dcl mask %s\n",
958 					dcl_mask);
959 			else if (strcmp(dcl_mask, ".xzw") == 0)
960 				fprintf(out, "errataed bad dcl mask %s\n",
961 					dcl_mask);
962 
963 			if (dcl_nr == 8) {
964 				instr_out(ctx, i++,
965 					  "%s: DCL DIFFUSE%s\n", instr_prefix,
966 					  dcl_mask);
967 			} else if (dcl_nr == 9) {
968 				instr_out(ctx, i++,
969 					  "%s: DCL SPECULAR%s\n", instr_prefix,
970 					  dcl_mask);
971 			} else if (dcl_nr == 10) {
972 				instr_out(ctx, i++,
973 					  "%s: DCL FOG%s\n", instr_prefix,
974 					  dcl_mask);
975 			}
976 		}
977 		instr_out(ctx, i++, "%s\n", instr_prefix);
978 		instr_out(ctx, i++, "%s\n", instr_prefix);
979 		break;
980 	case 3:
981 		switch ((d0 >> 22) & 0x3) {
982 		case 0:
983 			sampletype = "2D";
984 			break;
985 		case 1:
986 			sampletype = "CUBE";
987 			break;
988 		case 2:
989 			sampletype = "3D";
990 			break;
991 		default:
992 			sampletype = "RESERVED";
993 			break;
994 		}
995 		if (dcl_nr > 15)
996 			fprintf(out, "bad S%d dcl register number\n", dcl_nr);
997 		instr_out(ctx, i++, "%s: DCL S%d %s\n",
998 			  instr_prefix, dcl_nr, sampletype);
999 		instr_out(ctx, i++, "%s\n", instr_prefix);
1000 		instr_out(ctx, i++, "%s\n", instr_prefix);
1001 		break;
1002 	default:
1003 		instr_out(ctx, i++, "%s: DCL RESERVED%d\n",
1004 			  instr_prefix, dcl_nr);
1005 		instr_out(ctx, i++, "%s\n", instr_prefix);
1006 		instr_out(ctx, i++, "%s\n", instr_prefix);
1007 	}
1008 }
1009 
1010 static void
i915_decode_instruction(struct drm_intel_decode * ctx,int i,char * instr_prefix)1011 i915_decode_instruction(struct drm_intel_decode *ctx,
1012 			int i, char *instr_prefix)
1013 {
1014 	switch ((ctx->data[i] >> 24) & 0x1f) {
1015 	case 0x0:
1016 		instr_out(ctx, i++, "%s: NOP\n", instr_prefix);
1017 		instr_out(ctx, i++, "%s\n", instr_prefix);
1018 		instr_out(ctx, i++, "%s\n", instr_prefix);
1019 		break;
1020 	case 0x01:
1021 		i915_decode_alu2(ctx, i, instr_prefix, "ADD");
1022 		break;
1023 	case 0x02:
1024 		i915_decode_alu1(ctx, i, instr_prefix, "MOV");
1025 		break;
1026 	case 0x03:
1027 		i915_decode_alu2(ctx, i, instr_prefix, "MUL");
1028 		break;
1029 	case 0x04:
1030 		i915_decode_alu3(ctx, i, instr_prefix, "MAD");
1031 		break;
1032 	case 0x05:
1033 		i915_decode_alu3(ctx, i, instr_prefix, "DP2ADD");
1034 		break;
1035 	case 0x06:
1036 		i915_decode_alu2(ctx, i, instr_prefix, "DP3");
1037 		break;
1038 	case 0x07:
1039 		i915_decode_alu2(ctx, i, instr_prefix, "DP4");
1040 		break;
1041 	case 0x08:
1042 		i915_decode_alu1(ctx, i, instr_prefix, "FRC");
1043 		break;
1044 	case 0x09:
1045 		i915_decode_alu1(ctx, i, instr_prefix, "RCP");
1046 		break;
1047 	case 0x0a:
1048 		i915_decode_alu1(ctx, i, instr_prefix, "RSQ");
1049 		break;
1050 	case 0x0b:
1051 		i915_decode_alu1(ctx, i, instr_prefix, "EXP");
1052 		break;
1053 	case 0x0c:
1054 		i915_decode_alu1(ctx, i, instr_prefix, "LOG");
1055 		break;
1056 	case 0x0d:
1057 		i915_decode_alu2(ctx, i, instr_prefix, "CMP");
1058 		break;
1059 	case 0x0e:
1060 		i915_decode_alu2(ctx, i, instr_prefix, "MIN");
1061 		break;
1062 	case 0x0f:
1063 		i915_decode_alu2(ctx, i, instr_prefix, "MAX");
1064 		break;
1065 	case 0x10:
1066 		i915_decode_alu1(ctx, i, instr_prefix, "FLR");
1067 		break;
1068 	case 0x11:
1069 		i915_decode_alu1(ctx, i, instr_prefix, "MOD");
1070 		break;
1071 	case 0x12:
1072 		i915_decode_alu1(ctx, i, instr_prefix, "TRC");
1073 		break;
1074 	case 0x13:
1075 		i915_decode_alu2(ctx, i, instr_prefix, "SGE");
1076 		break;
1077 	case 0x14:
1078 		i915_decode_alu2(ctx, i, instr_prefix, "SLT");
1079 		break;
1080 	case 0x15:
1081 		i915_decode_tex(ctx, i, instr_prefix, "TEXLD");
1082 		break;
1083 	case 0x16:
1084 		i915_decode_tex(ctx, i, instr_prefix, "TEXLDP");
1085 		break;
1086 	case 0x17:
1087 		i915_decode_tex(ctx, i, instr_prefix, "TEXLDB");
1088 		break;
1089 	case 0x19:
1090 		i915_decode_dcl(ctx, i, instr_prefix);
1091 		break;
1092 	default:
1093 		instr_out(ctx, i++, "%s: unknown\n", instr_prefix);
1094 		instr_out(ctx, i++, "%s\n", instr_prefix);
1095 		instr_out(ctx, i++, "%s\n", instr_prefix);
1096 		break;
1097 	}
1098 }
1099 
1100 static const char *
decode_compare_func(uint32_t op)1101 decode_compare_func(uint32_t op)
1102 {
1103 	switch (op & 0x7) {
1104 	case 0:
1105 		return "always";
1106 	case 1:
1107 		return "never";
1108 	case 2:
1109 		return "less";
1110 	case 3:
1111 		return "equal";
1112 	case 4:
1113 		return "lequal";
1114 	case 5:
1115 		return "greater";
1116 	case 6:
1117 		return "notequal";
1118 	case 7:
1119 		return "gequal";
1120 	}
1121 	return "";
1122 }
1123 
1124 static const char *
decode_stencil_op(uint32_t op)1125 decode_stencil_op(uint32_t op)
1126 {
1127 	switch (op & 0x7) {
1128 	case 0:
1129 		return "keep";
1130 	case 1:
1131 		return "zero";
1132 	case 2:
1133 		return "replace";
1134 	case 3:
1135 		return "incr_sat";
1136 	case 4:
1137 		return "decr_sat";
1138 	case 5:
1139 		return "greater";
1140 	case 6:
1141 		return "incr";
1142 	case 7:
1143 		return "decr";
1144 	}
1145 	return "";
1146 }
1147 
1148 #if 0
1149 static const char *
1150 decode_logic_op(uint32_t op)
1151 {
1152 	switch (op & 0xf) {
1153 	case 0:
1154 		return "clear";
1155 	case 1:
1156 		return "nor";
1157 	case 2:
1158 		return "and_inv";
1159 	case 3:
1160 		return "copy_inv";
1161 	case 4:
1162 		return "and_rvrse";
1163 	case 5:
1164 		return "inv";
1165 	case 6:
1166 		return "xor";
1167 	case 7:
1168 		return "nand";
1169 	case 8:
1170 		return "and";
1171 	case 9:
1172 		return "equiv";
1173 	case 10:
1174 		return "noop";
1175 	case 11:
1176 		return "or_inv";
1177 	case 12:
1178 		return "copy";
1179 	case 13:
1180 		return "or_rvrse";
1181 	case 14:
1182 		return "or";
1183 	case 15:
1184 		return "set";
1185 	}
1186 	return "";
1187 }
1188 #endif
1189 
1190 static const char *
decode_blend_fact(uint32_t op)1191 decode_blend_fact(uint32_t op)
1192 {
1193 	switch (op & 0xf) {
1194 	case 1:
1195 		return "zero";
1196 	case 2:
1197 		return "one";
1198 	case 3:
1199 		return "src_colr";
1200 	case 4:
1201 		return "inv_src_colr";
1202 	case 5:
1203 		return "src_alpha";
1204 	case 6:
1205 		return "inv_src_alpha";
1206 	case 7:
1207 		return "dst_alpha";
1208 	case 8:
1209 		return "inv_dst_alpha";
1210 	case 9:
1211 		return "dst_colr";
1212 	case 10:
1213 		return "inv_dst_colr";
1214 	case 11:
1215 		return "src_alpha_sat";
1216 	case 12:
1217 		return "cnst_colr";
1218 	case 13:
1219 		return "inv_cnst_colr";
1220 	case 14:
1221 		return "cnst_alpha";
1222 	case 15:
1223 		return "inv_const_alpha";
1224 	}
1225 	return "";
1226 }
1227 
1228 static const char *
decode_tex_coord_mode(uint32_t mode)1229 decode_tex_coord_mode(uint32_t mode)
1230 {
1231 	switch (mode & 0x7) {
1232 	case 0:
1233 		return "wrap";
1234 	case 1:
1235 		return "mirror";
1236 	case 2:
1237 		return "clamp_edge";
1238 	case 3:
1239 		return "cube";
1240 	case 4:
1241 		return "clamp_border";
1242 	case 5:
1243 		return "mirror_once";
1244 	}
1245 	return "";
1246 }
1247 
1248 static const char *
decode_sample_filter(uint32_t mode)1249 decode_sample_filter(uint32_t mode)
1250 {
1251 	switch (mode & 0x7) {
1252 	case 0:
1253 		return "nearest";
1254 	case 1:
1255 		return "linear";
1256 	case 2:
1257 		return "anisotropic";
1258 	case 3:
1259 		return "4x4_1";
1260 	case 4:
1261 		return "4x4_2";
1262 	case 5:
1263 		return "4x4_flat";
1264 	case 6:
1265 		return "6x5_mono";
1266 	}
1267 	return "";
1268 }
1269 
1270 static int
decode_3d_1d(struct drm_intel_decode * ctx)1271 decode_3d_1d(struct drm_intel_decode *ctx)
1272 {
1273 	unsigned int len, i, c, idx, word, map, sampler, instr;
1274 	const char *format, *zformat, *type;
1275 	uint32_t opcode;
1276 	uint32_t *data = ctx->data;
1277 	uint32_t devid = ctx->devid;
1278 
1279 	struct {
1280 		uint32_t opcode;
1281 		int i830_only;
1282 		unsigned int min_len;
1283 		unsigned int max_len;
1284 		const char *name;
1285 	} opcodes_3d_1d[] = {
1286 		{ 0x86, 0, 4, 4, "3DSTATE_CHROMA_KEY" },
1287 		{ 0x88, 0, 2, 2, "3DSTATE_CONSTANT_BLEND_COLOR" },
1288 		{ 0x99, 0, 2, 2, "3DSTATE_DEFAULT_DIFFUSE" },
1289 		{ 0x9a, 0, 2, 2, "3DSTATE_DEFAULT_SPECULAR" },
1290 		{ 0x98, 0, 2, 2, "3DSTATE_DEFAULT_Z" },
1291 		{ 0x97, 0, 2, 2, "3DSTATE_DEPTH_OFFSET_SCALE" },
1292 		{ 0x9d, 0, 65, 65, "3DSTATE_FILTER_COEFFICIENTS_4X4" },
1293 		{ 0x9e, 0, 4, 4, "3DSTATE_MONO_FILTER" },
1294 		{ 0x89, 0, 4, 4, "3DSTATE_FOG_MODE" },
1295 		{ 0x8f, 0, 2, 16, "3DSTATE_MAP_PALLETE_LOAD_32" },
1296 		{ 0x83, 0, 2, 2, "3DSTATE_SPAN_STIPPLE" },
1297 		{ 0x8c, 1, 2, 2, "3DSTATE_MAP_COORD_TRANSFORM_I830" },
1298 		{ 0x8b, 1, 2, 2, "3DSTATE_MAP_VERTEX_TRANSFORM_I830" },
1299 		{ 0x8d, 1, 3, 3, "3DSTATE_W_STATE_I830" },
1300 		{ 0x01, 1, 2, 2, "3DSTATE_COLOR_FACTOR_I830" },
1301 		{ 0x02, 1, 2, 2, "3DSTATE_MAP_COORD_SETBIND_I830"},
1302 	}, *opcode_3d_1d;
1303 
1304 	opcode = (data[0] & 0x00ff0000) >> 16;
1305 
1306 	switch (opcode) {
1307 	case 0x07:
1308 		/* This instruction is unusual.  A 0 length means just
1309 		 * 1 DWORD instead of 2.  The 0 length is specified in
1310 		 * one place to be unsupported, but stated to be
1311 		 * required in another, and 0 length LOAD_INDIRECTs
1312 		 * appear to cause no harm at least.
1313 		 */
1314 		instr_out(ctx, 0, "3DSTATE_LOAD_INDIRECT\n");
1315 		len = (data[0] & 0x000000ff) + 1;
1316 		i = 1;
1317 		if (data[0] & (0x01 << 8)) {
1318 			instr_out(ctx, i++, "SIS.0\n");
1319 			instr_out(ctx, i++, "SIS.1\n");
1320 		}
1321 		if (data[0] & (0x02 << 8)) {
1322 			instr_out(ctx, i++, "DIS.0\n");
1323 		}
1324 		if (data[0] & (0x04 << 8)) {
1325 			instr_out(ctx, i++, "SSB.0\n");
1326 			instr_out(ctx, i++, "SSB.1\n");
1327 		}
1328 		if (data[0] & (0x08 << 8)) {
1329 			instr_out(ctx, i++, "MSB.0\n");
1330 			instr_out(ctx, i++, "MSB.1\n");
1331 		}
1332 		if (data[0] & (0x10 << 8)) {
1333 			instr_out(ctx, i++, "PSP.0\n");
1334 			instr_out(ctx, i++, "PSP.1\n");
1335 		}
1336 		if (data[0] & (0x20 << 8)) {
1337 			instr_out(ctx, i++, "PSC.0\n");
1338 			instr_out(ctx, i++, "PSC.1\n");
1339 		}
1340 		if (len != i) {
1341 			fprintf(out, "Bad count in 3DSTATE_LOAD_INDIRECT\n");
1342 			return len;
1343 		}
1344 		return len;
1345 	case 0x04:
1346 		instr_out(ctx, 0,
1347 			  "3DSTATE_LOAD_STATE_IMMEDIATE_1\n");
1348 		len = (data[0] & 0x0000000f) + 2;
1349 		i = 1;
1350 		for (word = 0; word <= 8; word++) {
1351 			if (data[0] & (1 << (4 + word))) {
1352 				/* save vertex state for decode */
1353 				if (!IS_GEN2(devid)) {
1354 					int tex_num;
1355 
1356 					if (word == 2) {
1357 						saved_s2_set = 1;
1358 						saved_s2 = data[i];
1359 					}
1360 					if (word == 4) {
1361 						saved_s4_set = 1;
1362 						saved_s4 = data[i];
1363 					}
1364 
1365 					switch (word) {
1366 					case 0:
1367 						instr_out(ctx, i,
1368 							  "S0: vbo offset: 0x%08x%s\n",
1369 							  data[i] & (~1),
1370 							  data[i] & 1 ?
1371 							  ", auto cache invalidate disabled"
1372 							  : "");
1373 						break;
1374 					case 1:
1375 						instr_out(ctx, i,
1376 							  "S1: vertex width: %i, vertex pitch: %i\n",
1377 							  (data[i] >> 24) &
1378 							  0x3f,
1379 							  (data[i] >> 16) &
1380 							  0x3f);
1381 						break;
1382 					case 2:
1383 						instr_out(ctx, i,
1384 							  "S2: texcoord formats: ");
1385 						for (tex_num = 0;
1386 						     tex_num < 8; tex_num++) {
1387 							switch ((data[i] >>
1388 								 tex_num *
1389 								 4) & 0xf) {
1390 							case 0:
1391 								fprintf(out,
1392 									"%i=2D ",
1393 									tex_num);
1394 								break;
1395 							case 1:
1396 								fprintf(out,
1397 									"%i=3D ",
1398 									tex_num);
1399 								break;
1400 							case 2:
1401 								fprintf(out,
1402 									"%i=4D ",
1403 									tex_num);
1404 								break;
1405 							case 3:
1406 								fprintf(out,
1407 									"%i=1D ",
1408 									tex_num);
1409 								break;
1410 							case 4:
1411 								fprintf(out,
1412 									"%i=2D_16 ",
1413 									tex_num);
1414 								break;
1415 							case 5:
1416 								fprintf(out,
1417 									"%i=4D_16 ",
1418 									tex_num);
1419 								break;
1420 							case 0xf:
1421 								fprintf(out,
1422 									"%i=NP ",
1423 									tex_num);
1424 								break;
1425 							}
1426 						}
1427 						fprintf(out, "\n");
1428 
1429 						break;
1430 					case 3:
1431 						instr_out(ctx, i,
1432 							  "S3: not documented\n");
1433 						break;
1434 					case 4:
1435 						{
1436 							const char *cullmode = "";
1437 							const char *vfmt_xyzw = "";
1438 							switch ((data[i] >> 13)
1439 								& 0x3) {
1440 							case 0:
1441 								cullmode =
1442 								    "both";
1443 								break;
1444 							case 1:
1445 								cullmode =
1446 								    "none";
1447 								break;
1448 							case 2:
1449 								cullmode = "cw";
1450 								break;
1451 							case 3:
1452 								cullmode =
1453 								    "ccw";
1454 								break;
1455 							}
1456 							switch (data[i] &
1457 								(7 << 6 | 1 <<
1458 								 2)) {
1459 							case 1 << 6:
1460 								vfmt_xyzw =
1461 								    "XYZ,";
1462 								break;
1463 							case 2 << 6:
1464 								vfmt_xyzw =
1465 								    "XYZW,";
1466 								break;
1467 							case 3 << 6:
1468 								vfmt_xyzw =
1469 								    "XY,";
1470 								break;
1471 							case 4 << 6:
1472 								vfmt_xyzw =
1473 								    "XYW,";
1474 								break;
1475 							case 1 << 6 | 1 << 2:
1476 								vfmt_xyzw =
1477 								    "XYZF,";
1478 								break;
1479 							case 2 << 6 | 1 << 2:
1480 								vfmt_xyzw =
1481 								    "XYZWF,";
1482 								break;
1483 							case 3 << 6 | 1 << 2:
1484 								vfmt_xyzw =
1485 								    "XYF,";
1486 								break;
1487 							case 4 << 6 | 1 << 2:
1488 								vfmt_xyzw =
1489 								    "XYWF,";
1490 								break;
1491 							}
1492 							instr_out(ctx, i,
1493 								  "S4: point_width=%i, line_width=%.1f,"
1494 								  "%s%s%s%s%s cullmode=%s, vfmt=%s%s%s%s%s%s "
1495 								  "%s%s%s%s%s\n",
1496 								  (data[i] >>
1497 								   23) & 0x1ff,
1498 								  ((data[i] >>
1499 								    19) & 0xf) /
1500 								  2.0,
1501 								  data[i] & (0xf
1502 									     <<
1503 									     15)
1504 								  ?
1505 								  " flatshade="
1506 								  : "",
1507 								  data[i] & (1
1508 									     <<
1509 									     18)
1510 								  ? "Alpha," :
1511 								  "",
1512 								  data[i] & (1
1513 									     <<
1514 									     17)
1515 								  ? "Fog," : "",
1516 								  data[i] & (1
1517 									     <<
1518 									     16)
1519 								  ? "Specular,"
1520 								  : "",
1521 								  data[i] & (1
1522 									     <<
1523 									     15)
1524 								  ? "Color," :
1525 								  "", cullmode,
1526 								  data[i] & (1
1527 									     <<
1528 									     12)
1529 								  ?
1530 								  "PointWidth,"
1531 								  : "",
1532 								  data[i] & (1
1533 									     <<
1534 									     11)
1535 								  ? "SpecFog," :
1536 								  "",
1537 								  data[i] & (1
1538 									     <<
1539 									     10)
1540 								  ? "Color," :
1541 								  "",
1542 								  data[i] & (1
1543 									     <<
1544 									     9)
1545 								  ? "DepthOfs,"
1546 								  : "",
1547 								  vfmt_xyzw,
1548 								  data[i] & (1
1549 									     <<
1550 									     9)
1551 								  ? "FogParam,"
1552 								  : "",
1553 								  data[i] & (1
1554 									     <<
1555 									     5)
1556 								  ?
1557 								  "force default diffuse, "
1558 								  : "",
1559 								  data[i] & (1
1560 									     <<
1561 									     4)
1562 								  ?
1563 								  "force default specular, "
1564 								  : "",
1565 								  data[i] & (1
1566 									     <<
1567 									     3)
1568 								  ?
1569 								  "local depth ofs enable, "
1570 								  : "",
1571 								  data[i] & (1
1572 									     <<
1573 									     1)
1574 								  ?
1575 								  "point sprite enable, "
1576 								  : "",
1577 								  data[i] & (1
1578 									     <<
1579 									     0)
1580 								  ?
1581 								  "line AA enable, "
1582 								  : "");
1583 							break;
1584 						}
1585 					case 5:
1586 						{
1587 							instr_out(ctx, i,
1588 								  "S5:%s%s%s%s%s"
1589 								  "%s%s%s%s stencil_ref=0x%x, stencil_test=%s, "
1590 								  "stencil_fail=%s, stencil_pass_z_fail=%s, "
1591 								  "stencil_pass_z_pass=%s, %s%s%s%s\n",
1592 								  data[i] & (0xf
1593 									     <<
1594 									     28)
1595 								  ?
1596 								  " write_disable="
1597 								  : "",
1598 								  data[i] & (1
1599 									     <<
1600 									     31)
1601 								  ? "Alpha," :
1602 								  "",
1603 								  data[i] & (1
1604 									     <<
1605 									     30)
1606 								  ? "Red," : "",
1607 								  data[i] & (1
1608 									     <<
1609 									     29)
1610 								  ? "Green," :
1611 								  "",
1612 								  data[i] & (1
1613 									     <<
1614 									     28)
1615 								  ? "Blue," :
1616 								  "",
1617 								  data[i] & (1
1618 									     <<
1619 									     27)
1620 								  ?
1621 								  " force default point size,"
1622 								  : "",
1623 								  data[i] & (1
1624 									     <<
1625 									     26)
1626 								  ?
1627 								  " last pixel enable,"
1628 								  : "",
1629 								  data[i] & (1
1630 									     <<
1631 									     25)
1632 								  ?
1633 								  " global depth ofs enable,"
1634 								  : "",
1635 								  data[i] & (1
1636 									     <<
1637 									     24)
1638 								  ?
1639 								  " fog enable,"
1640 								  : "",
1641 								  (data[i] >>
1642 								   16) & 0xff,
1643 								  decode_compare_func
1644 								  (data[i] >>
1645 								   13),
1646 								  decode_stencil_op
1647 								  (data[i] >>
1648 								   10),
1649 								  decode_stencil_op
1650 								  (data[i] >>
1651 								   7),
1652 								  decode_stencil_op
1653 								  (data[i] >>
1654 								   4),
1655 								  data[i] & (1
1656 									     <<
1657 									     3)
1658 								  ?
1659 								  "stencil write enable, "
1660 								  : "",
1661 								  data[i] & (1
1662 									     <<
1663 									     2)
1664 								  ?
1665 								  "stencil test enable, "
1666 								  : "",
1667 								  data[i] & (1
1668 									     <<
1669 									     1)
1670 								  ?
1671 								  "color dither enable, "
1672 								  : "",
1673 								  data[i] & (1
1674 									     <<
1675 									     0)
1676 								  ?
1677 								  "logicop enable, "
1678 								  : "");
1679 						}
1680 						break;
1681 					case 6:
1682 						instr_out(ctx, i,
1683 							  "S6: %salpha_test=%s, alpha_ref=0x%x, "
1684 							  "depth_test=%s, %ssrc_blnd_fct=%s, dst_blnd_fct=%s, "
1685 							  "%s%stristrip_provoking_vertex=%i\n",
1686 							  data[i] & (1 << 31) ?
1687 							  "alpha test enable, "
1688 							  : "",
1689 							  decode_compare_func
1690 							  (data[i] >> 28),
1691 							  data[i] & (0xff <<
1692 								     20),
1693 							  decode_compare_func
1694 							  (data[i] >> 16),
1695 							  data[i] & (1 << 15) ?
1696 							  "cbuf blend enable, "
1697 							  : "",
1698 							  decode_blend_fact(data
1699 									    [i]
1700 									    >>
1701 									    8),
1702 							  decode_blend_fact(data
1703 									    [i]
1704 									    >>
1705 									    4),
1706 							  data[i] & (1 << 3) ?
1707 							  "depth write enable, "
1708 							  : "",
1709 							  data[i] & (1 << 2) ?
1710 							  "cbuf write enable, "
1711 							  : "",
1712 							  data[i] & (0x3));
1713 						break;
1714 					case 7:
1715 						instr_out(ctx, i,
1716 							  "S7: depth offset constant: 0x%08x\n",
1717 							  data[i]);
1718 						break;
1719 					}
1720 				} else {
1721 					instr_out(ctx, i,
1722 						  "S%d: 0x%08x\n", word, data[i]);
1723 				}
1724 				i++;
1725 			}
1726 		}
1727 		if (len != i) {
1728 			fprintf(out,
1729 				"Bad count in 3DSTATE_LOAD_STATE_IMMEDIATE_1\n");
1730 		}
1731 		return len;
1732 	case 0x03:
1733 		instr_out(ctx, 0,
1734 			  "3DSTATE_LOAD_STATE_IMMEDIATE_2\n");
1735 		len = (data[0] & 0x0000000f) + 2;
1736 		i = 1;
1737 		for (word = 6; word <= 14; word++) {
1738 			if (data[0] & (1 << word)) {
1739 				if (word == 6)
1740 					instr_out(ctx, i++,
1741 						  "TBCF\n");
1742 				else if (word >= 7 && word <= 10) {
1743 					instr_out(ctx, i++,
1744 						  "TB%dC\n", word - 7);
1745 					instr_out(ctx, i++,
1746 						  "TB%dA\n", word - 7);
1747 				} else if (word >= 11 && word <= 14) {
1748 					instr_out(ctx, i,
1749 						  "TM%dS0: offset=0x%08x, %s\n",
1750 						  word - 11,
1751 						  data[i] & 0xfffffffe,
1752 						  data[i] & 1 ? "use fence" :
1753 						  "");
1754 					i++;
1755 					instr_out(ctx, i,
1756 						  "TM%dS1: height=%i, width=%i, %s\n",
1757 						  word - 11, data[i] >> 21,
1758 						  (data[i] >> 10) & 0x3ff,
1759 						  data[i] & 2 ? (data[i] & 1 ?
1760 								 "y-tiled" :
1761 								 "x-tiled") :
1762 						  "");
1763 					i++;
1764 					instr_out(ctx, i,
1765 						  "TM%dS2: pitch=%i, \n",
1766 						  word - 11,
1767 						  ((data[i] >> 21) + 1) * 4);
1768 					i++;
1769 					instr_out(ctx, i++,
1770 						  "TM%dS3\n", word - 11);
1771 					instr_out(ctx, i++,
1772 						  "TM%dS4: dflt color\n",
1773 						  word - 11);
1774 				}
1775 			}
1776 		}
1777 		if (len != i) {
1778 			fprintf(out,
1779 				"Bad count in 3DSTATE_LOAD_STATE_IMMEDIATE_2\n");
1780 		}
1781 		return len;
1782 	case 0x00:
1783 		instr_out(ctx, 0, "3DSTATE_MAP_STATE\n");
1784 		len = (data[0] & 0x0000003f) + 2;
1785 		instr_out(ctx, 1, "mask\n");
1786 
1787 		i = 2;
1788 		for (map = 0; map <= 15; map++) {
1789 			if (data[1] & (1 << map)) {
1790 				int width, height, pitch, dword;
1791 				const char *tiling;
1792 
1793 				dword = data[i];
1794 				instr_out(ctx, i++,
1795 					  "map %d MS2 %s%s%s\n", map,
1796 					  dword & (1 << 31) ?
1797 					  "untrusted surface, " : "",
1798 					  dword & (1 << 1) ?
1799 					  "vertical line stride enable, " : "",
1800 					  dword & (1 << 0) ?
1801 					  "vertical ofs enable, " : "");
1802 
1803 				dword = data[i];
1804 				width = ((dword >> 10) & ((1 << 11) - 1)) + 1;
1805 				height = ((dword >> 21) & ((1 << 11) - 1)) + 1;
1806 
1807 				tiling = "none";
1808 				if (dword & (1 << 2))
1809 					tiling = "fenced";
1810 				else if (dword & (1 << 1))
1811 					tiling = dword & (1 << 0) ? "Y" : "X";
1812 				type = " BAD";
1813 				format = "BAD";
1814 				switch ((dword >> 7) & 0x7) {
1815 				case 1:
1816 					type = "8b";
1817 					switch ((dword >> 3) & 0xf) {
1818 					case 0:
1819 						format = "I";
1820 						break;
1821 					case 1:
1822 						format = "L";
1823 						break;
1824 					case 4:
1825 						format = "A";
1826 						break;
1827 					case 5:
1828 						format = " mono";
1829 						break;
1830 					}
1831 					break;
1832 				case 2:
1833 					type = "16b";
1834 					switch ((dword >> 3) & 0xf) {
1835 					case 0:
1836 						format = " rgb565";
1837 						break;
1838 					case 1:
1839 						format = " argb1555";
1840 						break;
1841 					case 2:
1842 						format = " argb4444";
1843 						break;
1844 					case 5:
1845 						format = " ay88";
1846 						break;
1847 					case 6:
1848 						format = " bump655";
1849 						break;
1850 					case 7:
1851 						format = "I";
1852 						break;
1853 					case 8:
1854 						format = "L";
1855 						break;
1856 					case 9:
1857 						format = "A";
1858 						break;
1859 					}
1860 					break;
1861 				case 3:
1862 					type = "32b";
1863 					switch ((dword >> 3) & 0xf) {
1864 					case 0:
1865 						format = " argb8888";
1866 						break;
1867 					case 1:
1868 						format = " abgr8888";
1869 						break;
1870 					case 2:
1871 						format = " xrgb8888";
1872 						break;
1873 					case 3:
1874 						format = " xbgr8888";
1875 						break;
1876 					case 4:
1877 						format = " qwvu8888";
1878 						break;
1879 					case 5:
1880 						format = " axvu8888";
1881 						break;
1882 					case 6:
1883 						format = " lxvu8888";
1884 						break;
1885 					case 7:
1886 						format = " xlvu8888";
1887 						break;
1888 					case 8:
1889 						format = " argb2101010";
1890 						break;
1891 					case 9:
1892 						format = " abgr2101010";
1893 						break;
1894 					case 10:
1895 						format = " awvu2101010";
1896 						break;
1897 					case 11:
1898 						format = " gr1616";
1899 						break;
1900 					case 12:
1901 						format = " vu1616";
1902 						break;
1903 					case 13:
1904 						format = " xI824";
1905 						break;
1906 					case 14:
1907 						format = " xA824";
1908 						break;
1909 					case 15:
1910 						format = " xL824";
1911 						break;
1912 					}
1913 					break;
1914 				case 5:
1915 					type = "422";
1916 					switch ((dword >> 3) & 0xf) {
1917 					case 0:
1918 						format = " yuv_swapy";
1919 						break;
1920 					case 1:
1921 						format = " yuv";
1922 						break;
1923 					case 2:
1924 						format = " yuv_swapuv";
1925 						break;
1926 					case 3:
1927 						format = " yuv_swapuvy";
1928 						break;
1929 					}
1930 					break;
1931 				case 6:
1932 					type = "compressed";
1933 					switch ((dword >> 3) & 0x7) {
1934 					case 0:
1935 						format = " dxt1";
1936 						break;
1937 					case 1:
1938 						format = " dxt2_3";
1939 						break;
1940 					case 2:
1941 						format = " dxt4_5";
1942 						break;
1943 					case 3:
1944 						format = " fxt1";
1945 						break;
1946 					case 4:
1947 						format = " dxt1_rb";
1948 						break;
1949 					}
1950 					break;
1951 				case 7:
1952 					type = "4b indexed";
1953 					switch ((dword >> 3) & 0xf) {
1954 					case 7:
1955 						format = " argb8888";
1956 						break;
1957 					}
1958 					break;
1959 				}
1960 				dword = data[i];
1961 				instr_out(ctx, i++,
1962 					  "map %d MS3 [width=%d, height=%d, format=%s%s, tiling=%s%s]\n",
1963 					  map, width, height, type, format,
1964 					  tiling,
1965 					  dword & (1 << 9) ? " palette select" :
1966 					  "");
1967 
1968 				dword = data[i];
1969 				pitch =
1970 				    4 * (((dword >> 21) & ((1 << 11) - 1)) + 1);
1971 				instr_out(ctx, i++,
1972 					  "map %d MS4 [pitch=%d, max_lod=%i, vol_depth=%i, cube_face_ena=%x, %s]\n",
1973 					  map, pitch, (dword >> 9) & 0x3f,
1974 					  dword & 0xff, (dword >> 15) & 0x3f,
1975 					  dword & (1 << 8) ? "miplayout legacy"
1976 					  : "miplayout right");
1977 			}
1978 		}
1979 		if (len != i) {
1980 			fprintf(out, "Bad count in 3DSTATE_MAP_STATE\n");
1981 			return len;
1982 		}
1983 		return len;
1984 	case 0x06:
1985 		instr_out(ctx, 0,
1986 			  "3DSTATE_PIXEL_SHADER_CONSTANTS\n");
1987 		len = (data[0] & 0x000000ff) + 2;
1988 
1989 		i = 2;
1990 		for (c = 0; c <= 31; c++) {
1991 			if (data[1] & (1 << c)) {
1992 				instr_out(ctx, i, "C%d.X = %f\n", c,
1993 					  int_as_float(data[i]));
1994 				i++;
1995 				instr_out(ctx, i, "C%d.Y = %f\n",
1996 					  c, int_as_float(data[i]));
1997 				i++;
1998 				instr_out(ctx, i, "C%d.Z = %f\n",
1999 					  c, int_as_float(data[i]));
2000 				i++;
2001 				instr_out(ctx, i, "C%d.W = %f\n",
2002 					  c, int_as_float(data[i]));
2003 				i++;
2004 			}
2005 		}
2006 		if (len != i) {
2007 			fprintf(out,
2008 				"Bad count in 3DSTATE_PIXEL_SHADER_CONSTANTS\n");
2009 		}
2010 		return len;
2011 	case 0x05:
2012 		instr_out(ctx, 0, "3DSTATE_PIXEL_SHADER_PROGRAM\n");
2013 		len = (data[0] & 0x000000ff) + 2;
2014 		if ((len - 1) % 3 != 0 || len > 370) {
2015 			fprintf(out,
2016 				"Bad count in 3DSTATE_PIXEL_SHADER_PROGRAM\n");
2017 		}
2018 		i = 1;
2019 		for (instr = 0; instr < (len - 1) / 3; instr++) {
2020 			char instr_prefix[10];
2021 
2022 			sprintf(instr_prefix, "PS%03d", instr);
2023 			i915_decode_instruction(ctx, i,
2024 						instr_prefix);
2025 			i += 3;
2026 		}
2027 		return len;
2028 	case 0x01:
2029 		if (IS_GEN2(devid))
2030 			break;
2031 		instr_out(ctx, 0, "3DSTATE_SAMPLER_STATE\n");
2032 		instr_out(ctx, 1, "mask\n");
2033 		len = (data[0] & 0x0000003f) + 2;
2034 		i = 2;
2035 		for (sampler = 0; sampler <= 15; sampler++) {
2036 			if (data[1] & (1 << sampler)) {
2037 				uint32_t dword;
2038 				const char *mip_filter = "";
2039 
2040 				dword = data[i];
2041 				switch ((dword >> 20) & 0x3) {
2042 				case 0:
2043 					mip_filter = "none";
2044 					break;
2045 				case 1:
2046 					mip_filter = "nearest";
2047 					break;
2048 				case 3:
2049 					mip_filter = "linear";
2050 					break;
2051 				}
2052 				instr_out(ctx, i++,
2053 					  "sampler %d SS2:%s%s%s "
2054 					  "base_mip_level=%i, mip_filter=%s, mag_filter=%s, min_filter=%s "
2055 					  "lod_bias=%.2f,%s max_aniso=%i, shadow_func=%s\n",
2056 					  sampler,
2057 					  dword & (1 << 31) ? " reverse gamma,"
2058 					  : "",
2059 					  dword & (1 << 30) ? " packed2planar,"
2060 					  : "",
2061 					  dword & (1 << 29) ?
2062 					  " colorspace conversion," : "",
2063 					  (dword >> 22) & 0x1f, mip_filter,
2064 					  decode_sample_filter(dword >> 17),
2065 					  decode_sample_filter(dword >> 14),
2066 					  ((dword >> 5) & 0x1ff) / (0x10 * 1.0),
2067 					  dword & (1 << 4) ? " shadow," : "",
2068 					  dword & (1 << 3) ? 4 : 2,
2069 					  decode_compare_func(dword));
2070 				dword = data[i];
2071 				instr_out(ctx, i++,
2072 					  "sampler %d SS3: min_lod=%.2f,%s "
2073 					  "tcmode_x=%s, tcmode_y=%s, tcmode_z=%s,%s texmap_idx=%i,%s\n",
2074 					  sampler,
2075 					  ((dword >> 24) & 0xff) / (0x10 * 1.0),
2076 					  dword & (1 << 17) ?
2077 					  " kill pixel enable," : "",
2078 					  decode_tex_coord_mode(dword >> 12),
2079 					  decode_tex_coord_mode(dword >> 9),
2080 					  decode_tex_coord_mode(dword >> 6),
2081 					  dword & (1 << 5) ?
2082 					  " normalized coords," : "",
2083 					  (dword >> 1) & 0xf,
2084 					  dword & (1 << 0) ? " deinterlacer," :
2085 					  "");
2086 				dword = data[i];
2087 				instr_out(ctx, i++,
2088 					  "sampler %d SS4: border color\n",
2089 					  sampler);
2090 			}
2091 		}
2092 		if (len != i) {
2093 			fprintf(out, "Bad count in 3DSTATE_SAMPLER_STATE\n");
2094 		}
2095 		return len;
2096 	case 0x85:
2097 		len = (data[0] & 0x0000000f) + 2;
2098 
2099 		if (len != 2)
2100 			fprintf(out,
2101 				"Bad count in 3DSTATE_DEST_BUFFER_VARIABLES\n");
2102 
2103 		instr_out(ctx, 0,
2104 			  "3DSTATE_DEST_BUFFER_VARIABLES\n");
2105 
2106 		switch ((data[1] >> 8) & 0xf) {
2107 		case 0x0:
2108 			format = "g8";
2109 			break;
2110 		case 0x1:
2111 			format = "x1r5g5b5";
2112 			break;
2113 		case 0x2:
2114 			format = "r5g6b5";
2115 			break;
2116 		case 0x3:
2117 			format = "a8r8g8b8";
2118 			break;
2119 		case 0x4:
2120 			format = "ycrcb_swapy";
2121 			break;
2122 		case 0x5:
2123 			format = "ycrcb_normal";
2124 			break;
2125 		case 0x6:
2126 			format = "ycrcb_swapuv";
2127 			break;
2128 		case 0x7:
2129 			format = "ycrcb_swapuvy";
2130 			break;
2131 		case 0x8:
2132 			format = "a4r4g4b4";
2133 			break;
2134 		case 0x9:
2135 			format = "a1r5g5b5";
2136 			break;
2137 		case 0xa:
2138 			format = "a2r10g10b10";
2139 			break;
2140 		default:
2141 			format = "BAD";
2142 			break;
2143 		}
2144 		switch ((data[1] >> 2) & 0x3) {
2145 		case 0x0:
2146 			zformat = "u16";
2147 			break;
2148 		case 0x1:
2149 			zformat = "f16";
2150 			break;
2151 		case 0x2:
2152 			zformat = "u24x8";
2153 			break;
2154 		default:
2155 			zformat = "BAD";
2156 			break;
2157 		}
2158 		instr_out(ctx, 1,
2159 			  "%s format, %s depth format, early Z %sabled\n",
2160 			  format, zformat,
2161 			  (data[1] & (1 << 31)) ? "en" : "dis");
2162 		return len;
2163 
2164 	case 0x8e:
2165 		{
2166 			const char *name, *tiling;
2167 
2168 			len = (data[0] & 0x0000000f) + 2;
2169 			if (len != 3)
2170 				fprintf(out,
2171 					"Bad count in 3DSTATE_BUFFER_INFO\n");
2172 
2173 			switch ((data[1] >> 24) & 0x7) {
2174 			case 0x3:
2175 				name = "color";
2176 				break;
2177 			case 0x7:
2178 				name = "depth";
2179 				break;
2180 			default:
2181 				name = "unknown";
2182 				break;
2183 			}
2184 
2185 			tiling = "none";
2186 			if (data[1] & (1 << 23))
2187 				tiling = "fenced";
2188 			else if (data[1] & (1 << 22))
2189 				tiling = data[1] & (1 << 21) ? "Y" : "X";
2190 
2191 			instr_out(ctx, 0, "3DSTATE_BUFFER_INFO\n");
2192 			instr_out(ctx, 1,
2193 				  "%s, tiling = %s, pitch=%d\n", name, tiling,
2194 				  data[1] & 0xffff);
2195 
2196 			instr_out(ctx, 2, "address\n");
2197 			return len;
2198 		}
2199 	case 0x81:
2200 		len = (data[0] & 0x0000000f) + 2;
2201 
2202 		if (len != 3)
2203 			fprintf(out,
2204 				"Bad count in 3DSTATE_SCISSOR_RECTANGLE\n");
2205 
2206 		instr_out(ctx, 0, "3DSTATE_SCISSOR_RECTANGLE\n");
2207 		instr_out(ctx, 1, "(%d,%d)\n",
2208 			  data[1] & 0xffff, data[1] >> 16);
2209 		instr_out(ctx, 2, "(%d,%d)\n",
2210 			  data[2] & 0xffff, data[2] >> 16);
2211 
2212 		return len;
2213 	case 0x80:
2214 		len = (data[0] & 0x0000000f) + 2;
2215 
2216 		if (len != 5)
2217 			fprintf(out,
2218 				"Bad count in 3DSTATE_DRAWING_RECTANGLE\n");
2219 
2220 		instr_out(ctx, 0, "3DSTATE_DRAWING_RECTANGLE\n");
2221 		instr_out(ctx, 1, "%s\n",
2222 			  data[1] & (1 << 30) ? "depth ofs disabled " : "");
2223 		instr_out(ctx, 2, "(%d,%d)\n",
2224 			  data[2] & 0xffff, data[2] >> 16);
2225 		instr_out(ctx, 3, "(%d,%d)\n",
2226 			  data[3] & 0xffff, data[3] >> 16);
2227 		instr_out(ctx, 4, "(%d,%d)\n",
2228 			  data[4] & 0xffff, data[4] >> 16);
2229 
2230 		return len;
2231 	case 0x9c:
2232 		len = (data[0] & 0x0000000f) + 2;
2233 
2234 		if (len != 7)
2235 			fprintf(out, "Bad count in 3DSTATE_CLEAR_PARAMETERS\n");
2236 
2237 		instr_out(ctx, 0, "3DSTATE_CLEAR_PARAMETERS\n");
2238 		instr_out(ctx, 1, "prim_type=%s, clear=%s%s%s\n",
2239 			  data[1] & (1 << 16) ? "CLEAR_RECT" : "ZONE_INIT",
2240 			  data[1] & (1 << 2) ? "color," : "",
2241 			  data[1] & (1 << 1) ? "depth," : "",
2242 			  data[1] & (1 << 0) ? "stencil," : "");
2243 		instr_out(ctx, 2, "clear color\n");
2244 		instr_out(ctx, 3, "clear depth/stencil\n");
2245 		instr_out(ctx, 4, "color value (rgba8888)\n");
2246 		instr_out(ctx, 5, "depth value %f\n",
2247 			  int_as_float(data[5]));
2248 		instr_out(ctx, 6, "clear stencil\n");
2249 		return len;
2250 	}
2251 
2252 	for (idx = 0; idx < ARRAY_SIZE(opcodes_3d_1d); idx++) {
2253 		opcode_3d_1d = &opcodes_3d_1d[idx];
2254 		if (opcode_3d_1d->i830_only && !IS_GEN2(devid))
2255 			continue;
2256 
2257 		if (((data[0] & 0x00ff0000) >> 16) == opcode_3d_1d->opcode) {
2258 			len = 1;
2259 
2260 			instr_out(ctx, 0, "%s\n",
2261 				  opcode_3d_1d->name);
2262 			if (opcode_3d_1d->max_len > 1) {
2263 				len = (data[0] & 0x0000ffff) + 2;
2264 				if (len < opcode_3d_1d->min_len ||
2265 				    len > opcode_3d_1d->max_len) {
2266 					fprintf(out, "Bad count in %s\n",
2267 						opcode_3d_1d->name);
2268 				}
2269 			}
2270 
2271 			for (i = 1; i < len; i++) {
2272 				instr_out(ctx, i, "dword %d\n", i);
2273 			}
2274 
2275 			return len;
2276 		}
2277 	}
2278 
2279 	instr_out(ctx, 0, "3D UNKNOWN: 3d_1d opcode = 0x%x\n",
2280 		  opcode);
2281 	return 1;
2282 }
2283 
2284 static int
decode_3d_primitive(struct drm_intel_decode * ctx)2285 decode_3d_primitive(struct drm_intel_decode *ctx)
2286 {
2287 	uint32_t *data = ctx->data;
2288 	uint32_t count = ctx->count;
2289 	char immediate = (data[0] & (1 << 23)) == 0;
2290 	unsigned int len, i, j, ret;
2291 	const char *primtype;
2292 	int original_s2 = saved_s2;
2293 	int original_s4 = saved_s4;
2294 
2295 	switch ((data[0] >> 18) & 0xf) {
2296 	case 0x0:
2297 		primtype = "TRILIST";
2298 		break;
2299 	case 0x1:
2300 		primtype = "TRISTRIP";
2301 		break;
2302 	case 0x2:
2303 		primtype = "TRISTRIP_REVERSE";
2304 		break;
2305 	case 0x3:
2306 		primtype = "TRIFAN";
2307 		break;
2308 	case 0x4:
2309 		primtype = "POLYGON";
2310 		break;
2311 	case 0x5:
2312 		primtype = "LINELIST";
2313 		break;
2314 	case 0x6:
2315 		primtype = "LINESTRIP";
2316 		break;
2317 	case 0x7:
2318 		primtype = "RECTLIST";
2319 		break;
2320 	case 0x8:
2321 		primtype = "POINTLIST";
2322 		break;
2323 	case 0x9:
2324 		primtype = "DIB";
2325 		break;
2326 	case 0xa:
2327 		primtype = "CLEAR_RECT";
2328 		saved_s4 = 3 << 6;
2329 		saved_s2 = ~0;
2330 		break;
2331 	default:
2332 		primtype = "unknown";
2333 		break;
2334 	}
2335 
2336 	/* XXX: 3DPRIM_DIB not supported */
2337 	if (immediate) {
2338 		len = (data[0] & 0x0003ffff) + 2;
2339 		instr_out(ctx, 0, "3DPRIMITIVE inline %s\n",
2340 			  primtype);
2341 		if (count < len)
2342 			BUFFER_FAIL(count, len, "3DPRIMITIVE inline");
2343 		if (!saved_s2_set || !saved_s4_set) {
2344 			fprintf(out, "unknown vertex format\n");
2345 			for (i = 1; i < len; i++) {
2346 				instr_out(ctx, i,
2347 					  "           vertex data (%f float)\n",
2348 					  int_as_float(data[i]));
2349 			}
2350 		} else {
2351 			unsigned int vertex = 0;
2352 			for (i = 1; i < len;) {
2353 				unsigned int tc;
2354 
2355 #define VERTEX_OUT(fmt, ...) do {					\
2356     if (i < len)							\
2357 	instr_out(ctx, i, " V%d."fmt"\n", vertex, __VA_ARGS__); \
2358     else								\
2359 	fprintf(out, " missing data in V%d\n", vertex);			\
2360     i++;								\
2361 } while (0)
2362 
2363 				VERTEX_OUT("X = %f", int_as_float(data[i]));
2364 				VERTEX_OUT("Y = %f", int_as_float(data[i]));
2365 				switch (saved_s4 >> 6 & 0x7) {
2366 				case 0x1:
2367 					VERTEX_OUT("Z = %f",
2368 						   int_as_float(data[i]));
2369 					break;
2370 				case 0x2:
2371 					VERTEX_OUT("Z = %f",
2372 						   int_as_float(data[i]));
2373 					VERTEX_OUT("W = %f",
2374 						   int_as_float(data[i]));
2375 					break;
2376 				case 0x3:
2377 					break;
2378 				case 0x4:
2379 					VERTEX_OUT("W = %f",
2380 						   int_as_float(data[i]));
2381 					break;
2382 				default:
2383 					fprintf(out, "bad S4 position mask\n");
2384 				}
2385 
2386 				if (saved_s4 & (1 << 10)) {
2387 					VERTEX_OUT
2388 					    ("color = (A=0x%02x, R=0x%02x, G=0x%02x, "
2389 					     "B=0x%02x)", data[i] >> 24,
2390 					     (data[i] >> 16) & 0xff,
2391 					     (data[i] >> 8) & 0xff,
2392 					     data[i] & 0xff);
2393 				}
2394 				if (saved_s4 & (1 << 11)) {
2395 					VERTEX_OUT
2396 					    ("spec = (A=0x%02x, R=0x%02x, G=0x%02x, "
2397 					     "B=0x%02x)", data[i] >> 24,
2398 					     (data[i] >> 16) & 0xff,
2399 					     (data[i] >> 8) & 0xff,
2400 					     data[i] & 0xff);
2401 				}
2402 				if (saved_s4 & (1 << 12))
2403 					VERTEX_OUT("width = 0x%08x)", data[i]);
2404 
2405 				for (tc = 0; tc <= 7; tc++) {
2406 					switch ((saved_s2 >> (tc * 4)) & 0xf) {
2407 					case 0x0:
2408 						VERTEX_OUT("T%d.X = %f", tc,
2409 							   int_as_float(data
2410 									[i]));
2411 						VERTEX_OUT("T%d.Y = %f", tc,
2412 							   int_as_float(data
2413 									[i]));
2414 						break;
2415 					case 0x1:
2416 						VERTEX_OUT("T%d.X = %f", tc,
2417 							   int_as_float(data
2418 									[i]));
2419 						VERTEX_OUT("T%d.Y = %f", tc,
2420 							   int_as_float(data
2421 									[i]));
2422 						VERTEX_OUT("T%d.Z = %f", tc,
2423 							   int_as_float(data
2424 									[i]));
2425 						break;
2426 					case 0x2:
2427 						VERTEX_OUT("T%d.X = %f", tc,
2428 							   int_as_float(data
2429 									[i]));
2430 						VERTEX_OUT("T%d.Y = %f", tc,
2431 							   int_as_float(data
2432 									[i]));
2433 						VERTEX_OUT("T%d.Z = %f", tc,
2434 							   int_as_float(data
2435 									[i]));
2436 						VERTEX_OUT("T%d.W = %f", tc,
2437 							   int_as_float(data
2438 									[i]));
2439 						break;
2440 					case 0x3:
2441 						VERTEX_OUT("T%d.X = %f", tc,
2442 							   int_as_float(data
2443 									[i]));
2444 						break;
2445 					case 0x4:
2446 						VERTEX_OUT
2447 						    ("T%d.XY = 0x%08x half-float",
2448 						     tc, data[i]);
2449 						break;
2450 					case 0x5:
2451 						VERTEX_OUT
2452 						    ("T%d.XY = 0x%08x half-float",
2453 						     tc, data[i]);
2454 						VERTEX_OUT
2455 						    ("T%d.ZW = 0x%08x half-float",
2456 						     tc, data[i]);
2457 						break;
2458 					case 0xf:
2459 						break;
2460 					default:
2461 						fprintf(out,
2462 							"bad S2.T%d format\n",
2463 							tc);
2464 					}
2465 				}
2466 				vertex++;
2467 			}
2468 		}
2469 
2470 		ret = len;
2471 	} else {
2472 		/* indirect vertices */
2473 		len = data[0] & 0x0000ffff;	/* index count */
2474 		if (data[0] & (1 << 17)) {
2475 			/* random vertex access */
2476 			if (count < (len + 1) / 2 + 1) {
2477 				BUFFER_FAIL(count, (len + 1) / 2 + 1,
2478 					    "3DPRIMITIVE random indirect");
2479 			}
2480 			instr_out(ctx, 0,
2481 				  "3DPRIMITIVE random indirect %s (%d)\n",
2482 				  primtype, len);
2483 			if (len == 0) {
2484 				/* vertex indices continue until 0xffff is
2485 				 * found
2486 				 */
2487 				for (i = 1; i < count; i++) {
2488 					if ((data[i] & 0xffff) == 0xffff) {
2489 						instr_out(ctx, i,
2490 							  "    indices: (terminator)\n");
2491 						ret = i;
2492 						goto out;
2493 					} else if ((data[i] >> 16) == 0xffff) {
2494 						instr_out(ctx, i,
2495 							  "    indices: 0x%04x, (terminator)\n",
2496 							  data[i] & 0xffff);
2497 						ret = i;
2498 						goto out;
2499 					} else {
2500 						instr_out(ctx, i,
2501 							  "    indices: 0x%04x, 0x%04x\n",
2502 							  data[i] & 0xffff,
2503 							  data[i] >> 16);
2504 					}
2505 				}
2506 				fprintf(out,
2507 					"3DPRIMITIVE: no terminator found in index buffer\n");
2508 				ret = count;
2509 				goto out;
2510 			} else {
2511 				/* fixed size vertex index buffer */
2512 				for (j = 1, i = 0; i < len; i += 2, j++) {
2513 					if (i * 2 == len - 1) {
2514 						instr_out(ctx, j,
2515 							  "    indices: 0x%04x\n",
2516 							  data[j] & 0xffff);
2517 					} else {
2518 						instr_out(ctx, j,
2519 							  "    indices: 0x%04x, 0x%04x\n",
2520 							  data[j] & 0xffff,
2521 							  data[j] >> 16);
2522 					}
2523 				}
2524 			}
2525 			ret = (len + 1) / 2 + 1;
2526 			goto out;
2527 		} else {
2528 			/* sequential vertex access */
2529 			instr_out(ctx, 0,
2530 				  "3DPRIMITIVE sequential indirect %s, %d starting from "
2531 				  "%d\n", primtype, len, data[1] & 0xffff);
2532 			instr_out(ctx, 1, "           start\n");
2533 			ret = 2;
2534 			goto out;
2535 		}
2536 	}
2537 
2538 out:
2539 	saved_s2 = original_s2;
2540 	saved_s4 = original_s4;
2541 	return ret;
2542 }
2543 
2544 static int
decode_3d(struct drm_intel_decode * ctx)2545 decode_3d(struct drm_intel_decode *ctx)
2546 {
2547 	uint32_t opcode;
2548 	unsigned int idx;
2549 	uint32_t *data = ctx->data;
2550 
2551 	struct {
2552 		uint32_t opcode;
2553 		unsigned int min_len;
2554 		unsigned int max_len;
2555 		const char *name;
2556 	} opcodes_3d[] = {
2557 		{ 0x06, 1, 1, "3DSTATE_ANTI_ALIASING" },
2558 		{ 0x08, 1, 1, "3DSTATE_BACKFACE_STENCIL_OPS" },
2559 		{ 0x09, 1, 1, "3DSTATE_BACKFACE_STENCIL_MASKS" },
2560 		{ 0x16, 1, 1, "3DSTATE_COORD_SET_BINDINGS" },
2561 		{ 0x15, 1, 1, "3DSTATE_FOG_COLOR" },
2562 		{ 0x0b, 1, 1, "3DSTATE_INDEPENDENT_ALPHA_BLEND" },
2563 		{ 0x0d, 1, 1, "3DSTATE_MODES_4" },
2564 		{ 0x0c, 1, 1, "3DSTATE_MODES_5" },
2565 		{ 0x07, 1, 1, "3DSTATE_RASTERIZATION_RULES"},
2566 	}, *opcode_3d;
2567 
2568 	opcode = (data[0] & 0x1f000000) >> 24;
2569 
2570 	switch (opcode) {
2571 	case 0x1f:
2572 		return decode_3d_primitive(ctx);
2573 	case 0x1d:
2574 		return decode_3d_1d(ctx);
2575 	case 0x1c:
2576 		return decode_3d_1c(ctx);
2577 	}
2578 
2579 	for (idx = 0; idx < ARRAY_SIZE(opcodes_3d); idx++) {
2580 		opcode_3d = &opcodes_3d[idx];
2581 		if (opcode == opcode_3d->opcode) {
2582 			unsigned int len = 1, i;
2583 
2584 			instr_out(ctx, 0, "%s\n", opcode_3d->name);
2585 			if (opcode_3d->max_len > 1) {
2586 				len = (data[0] & 0xff) + 2;
2587 				if (len < opcode_3d->min_len ||
2588 				    len > opcode_3d->max_len) {
2589 					fprintf(out, "Bad count in %s\n",
2590 						opcode_3d->name);
2591 				}
2592 			}
2593 
2594 			for (i = 1; i < len; i++) {
2595 				instr_out(ctx, i, "dword %d\n", i);
2596 			}
2597 			return len;
2598 		}
2599 	}
2600 
2601 	instr_out(ctx, 0, "3D UNKNOWN: 3d opcode = 0x%x\n", opcode);
2602 	return 1;
2603 }
2604 
get_965_surfacetype(unsigned int surfacetype)2605 static const char *get_965_surfacetype(unsigned int surfacetype)
2606 {
2607 	switch (surfacetype) {
2608 	case 0:
2609 		return "1D";
2610 	case 1:
2611 		return "2D";
2612 	case 2:
2613 		return "3D";
2614 	case 3:
2615 		return "CUBE";
2616 	case 4:
2617 		return "BUFFER";
2618 	case 7:
2619 		return "NULL";
2620 	default:
2621 		return "unknown";
2622 	}
2623 }
2624 
get_965_depthformat(unsigned int depthformat)2625 static const char *get_965_depthformat(unsigned int depthformat)
2626 {
2627 	switch (depthformat) {
2628 	case 0:
2629 		return "s8_z24float";
2630 	case 1:
2631 		return "z32float";
2632 	case 2:
2633 		return "z24s8";
2634 	case 5:
2635 		return "z16";
2636 	default:
2637 		return "unknown";
2638 	}
2639 }
2640 
get_965_element_component(uint32_t data,int component)2641 static const char *get_965_element_component(uint32_t data, int component)
2642 {
2643 	uint32_t component_control = (data >> (16 + (3 - component) * 4)) & 0x7;
2644 
2645 	switch (component_control) {
2646 	case 0:
2647 		return "nostore";
2648 	case 1:
2649 		switch (component) {
2650 		case 0:
2651 			return "X";
2652 		case 1:
2653 			return "Y";
2654 		case 2:
2655 			return "Z";
2656 		case 3:
2657 			return "W";
2658 		default:
2659 			return "fail";
2660 		}
2661 	case 2:
2662 		return "0.0";
2663 	case 3:
2664 		return "1.0";
2665 	case 4:
2666 		return "0x1";
2667 	case 5:
2668 		return "VID";
2669 	default:
2670 		return "fail";
2671 	}
2672 }
2673 
get_965_prim_type(uint32_t primtype)2674 static const char *get_965_prim_type(uint32_t primtype)
2675 {
2676 	switch (primtype) {
2677 	case 0x01:
2678 		return "point list";
2679 	case 0x02:
2680 		return "line list";
2681 	case 0x03:
2682 		return "line strip";
2683 	case 0x04:
2684 		return "tri list";
2685 	case 0x05:
2686 		return "tri strip";
2687 	case 0x06:
2688 		return "tri fan";
2689 	case 0x07:
2690 		return "quad list";
2691 	case 0x08:
2692 		return "quad strip";
2693 	case 0x09:
2694 		return "line list adj";
2695 	case 0x0a:
2696 		return "line strip adj";
2697 	case 0x0b:
2698 		return "tri list adj";
2699 	case 0x0c:
2700 		return "tri strip adj";
2701 	case 0x0d:
2702 		return "tri strip reverse";
2703 	case 0x0e:
2704 		return "polygon";
2705 	case 0x0f:
2706 		return "rect list";
2707 	case 0x10:
2708 		return "line loop";
2709 	case 0x11:
2710 		return "point list bf";
2711 	case 0x12:
2712 		return "line strip cont";
2713 	case 0x13:
2714 		return "line strip bf";
2715 	case 0x14:
2716 		return "line strip cont bf";
2717 	case 0x15:
2718 		return "tri fan no stipple";
2719 	default:
2720 		return "fail";
2721 	}
2722 }
2723 
2724 static int
i965_decode_urb_fence(struct drm_intel_decode * ctx,int len)2725 i965_decode_urb_fence(struct drm_intel_decode *ctx, int len)
2726 {
2727 	uint32_t vs_fence, clip_fence, gs_fence, sf_fence, vfe_fence, cs_fence;
2728 	uint32_t *data = ctx->data;
2729 
2730 	if (len != 3)
2731 		fprintf(out, "Bad count in URB_FENCE\n");
2732 
2733 	vs_fence = data[1] & 0x3ff;
2734 	gs_fence = (data[1] >> 10) & 0x3ff;
2735 	clip_fence = (data[1] >> 20) & 0x3ff;
2736 	sf_fence = data[2] & 0x3ff;
2737 	vfe_fence = (data[2] >> 10) & 0x3ff;
2738 	cs_fence = (data[2] >> 20) & 0x7ff;
2739 
2740 	instr_out(ctx, 0, "URB_FENCE: %s%s%s%s%s%s\n",
2741 		  (data[0] >> 13) & 1 ? "cs " : "",
2742 		  (data[0] >> 12) & 1 ? "vfe " : "",
2743 		  (data[0] >> 11) & 1 ? "sf " : "",
2744 		  (data[0] >> 10) & 1 ? "clip " : "",
2745 		  (data[0] >> 9) & 1 ? "gs " : "",
2746 		  (data[0] >> 8) & 1 ? "vs " : "");
2747 	instr_out(ctx, 1,
2748 		  "vs fence: %d, clip_fence: %d, gs_fence: %d\n",
2749 		  vs_fence, clip_fence, gs_fence);
2750 	instr_out(ctx, 2,
2751 		  "sf fence: %d, vfe_fence: %d, cs_fence: %d\n",
2752 		  sf_fence, vfe_fence, cs_fence);
2753 	if (gs_fence < vs_fence)
2754 		fprintf(out, "gs fence < vs fence!\n");
2755 	if (clip_fence < gs_fence)
2756 		fprintf(out, "clip fence < gs fence!\n");
2757 	if (sf_fence < clip_fence)
2758 		fprintf(out, "sf fence < clip fence!\n");
2759 	if (cs_fence < sf_fence)
2760 		fprintf(out, "cs fence < sf fence!\n");
2761 
2762 	return len;
2763 }
2764 
2765 static void
state_base_out(struct drm_intel_decode * ctx,unsigned int index,const char * name)2766 state_base_out(struct drm_intel_decode *ctx, unsigned int index,
2767 	       const char *name)
2768 {
2769 	if (ctx->data[index] & 1) {
2770 		instr_out(ctx, index,
2771 			  "%s state base address 0x%08x\n", name,
2772 			  ctx->data[index] & ~1);
2773 	} else {
2774 		instr_out(ctx, index, "%s state base not updated\n",
2775 			  name);
2776 	}
2777 }
2778 
2779 static void
state_max_out(struct drm_intel_decode * ctx,unsigned int index,const char * name)2780 state_max_out(struct drm_intel_decode *ctx, unsigned int index,
2781 	      const char *name)
2782 {
2783 	if (ctx->data[index] & 1) {
2784 		if (ctx->data[index] == 1) {
2785 			instr_out(ctx, index,
2786 				  "%s state upper bound disabled\n", name);
2787 		} else {
2788 			instr_out(ctx, index,
2789 				  "%s state upper bound 0x%08x\n", name,
2790 				  ctx->data[index] & ~1);
2791 		}
2792 	} else {
2793 		instr_out(ctx, index,
2794 			  "%s state upper bound not updated\n", name);
2795 	}
2796 }
2797 
2798 static int
gen7_3DSTATE_VIEWPORT_STATE_POINTERS_CC(struct drm_intel_decode * ctx)2799 gen7_3DSTATE_VIEWPORT_STATE_POINTERS_CC(struct drm_intel_decode *ctx)
2800 {
2801 	instr_out(ctx, 0, "3DSTATE_VIEWPORT_STATE_POINTERS_CC\n");
2802 	instr_out(ctx, 1, "pointer to CC viewport\n");
2803 
2804 	return 2;
2805 }
2806 
2807 static int
gen7_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP(struct drm_intel_decode * ctx)2808 gen7_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP(struct drm_intel_decode *ctx)
2809 {
2810 	instr_out(ctx, 0, "3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP\n");
2811 	instr_out(ctx, 1, "pointer to SF_CLIP viewport\n");
2812 
2813 	return 2;
2814 }
2815 
2816 static int
gen7_3DSTATE_BLEND_STATE_POINTERS(struct drm_intel_decode * ctx)2817 gen7_3DSTATE_BLEND_STATE_POINTERS(struct drm_intel_decode *ctx)
2818 {
2819 	instr_out(ctx, 0, "3DSTATE_BLEND_STATE_POINTERS\n");
2820 	instr_out(ctx, 1, "pointer to BLEND_STATE at 0x%08x (%s)\n",
2821 		  ctx->data[1] & ~1,
2822 		  (ctx->data[1] & 1) ? "changed" : "unchanged");
2823 
2824 	return 2;
2825 }
2826 
2827 static int
gen7_3DSTATE_DEPTH_STENCIL_STATE_POINTERS(struct drm_intel_decode * ctx)2828 gen7_3DSTATE_DEPTH_STENCIL_STATE_POINTERS(struct drm_intel_decode *ctx)
2829 {
2830 	instr_out(ctx, 0, "3DSTATE_DEPTH_STENCIL_STATE_POINTERS\n");
2831 	instr_out(ctx, 1,
2832 		  "pointer to DEPTH_STENCIL_STATE at 0x%08x (%s)\n",
2833 		  ctx->data[1] & ~1,
2834 		  (ctx->data[1] & 1) ? "changed" : "unchanged");
2835 
2836 	return 2;
2837 }
2838 
2839 static int
gen7_3DSTATE_HIER_DEPTH_BUFFER(struct drm_intel_decode * ctx)2840 gen7_3DSTATE_HIER_DEPTH_BUFFER(struct drm_intel_decode *ctx)
2841 {
2842 	instr_out(ctx, 0, "3DSTATE_HIER_DEPTH_BUFFER\n");
2843 	instr_out(ctx, 1, "pitch %db\n",
2844 		  (ctx->data[1] & 0x1ffff) + 1);
2845 	instr_out(ctx, 2, "pointer to HiZ buffer\n");
2846 
2847 	return 3;
2848 }
2849 
2850 static int
gen6_3DSTATE_CC_STATE_POINTERS(struct drm_intel_decode * ctx)2851 gen6_3DSTATE_CC_STATE_POINTERS(struct drm_intel_decode *ctx)
2852 {
2853 	instr_out(ctx, 0, "3DSTATE_CC_STATE_POINTERS\n");
2854 	instr_out(ctx, 1, "blend change %d\n", ctx->data[1] & 1);
2855 	instr_out(ctx, 2, "depth stencil change %d\n",
2856 		  ctx->data[2] & 1);
2857 	instr_out(ctx, 3, "cc change %d\n", ctx->data[3] & 1);
2858 
2859 	return 4;
2860 }
2861 
2862 static int
gen7_3DSTATE_CC_STATE_POINTERS(struct drm_intel_decode * ctx)2863 gen7_3DSTATE_CC_STATE_POINTERS(struct drm_intel_decode *ctx)
2864 {
2865 	instr_out(ctx, 0, "3DSTATE_CC_STATE_POINTERS\n");
2866 	instr_out(ctx, 1, "pointer to COLOR_CALC_STATE at 0x%08x "
2867 		  "(%s)\n",
2868 		  ctx->data[1] & ~1,
2869 		  (ctx->data[1] & 1) ? "changed" : "unchanged");
2870 
2871 	return 2;
2872 }
2873 
2874 static int
gen7_3DSTATE_URB_unit(struct drm_intel_decode * ctx,const char * unit)2875 gen7_3DSTATE_URB_unit(struct drm_intel_decode *ctx, const char *unit)
2876 {
2877     int start_kb = ((ctx->data[1] >> 25) & 0x3f) * 8;
2878     /* the field is # of 512-bit rows - 1, we print bytes */
2879     int entry_size = (((ctx->data[1] >> 16) & 0x1ff) + 1);
2880     int nr_entries = ctx->data[1] & 0xffff;
2881 
2882     instr_out(ctx, 0, "3DSTATE_URB_%s\n", unit);
2883     instr_out(ctx, 1,
2884 	      "%dKB start, size=%d 64B rows, nr_entries=%d, total size %dB\n",
2885 	      start_kb, entry_size, nr_entries, nr_entries * 64 * entry_size);
2886 
2887     return 2;
2888 }
2889 
2890 static int
gen7_3DSTATE_URB_VS(struct drm_intel_decode * ctx)2891 gen7_3DSTATE_URB_VS(struct drm_intel_decode *ctx)
2892 {
2893 	return gen7_3DSTATE_URB_unit(ctx, "VS");
2894 }
2895 
2896 static int
gen7_3DSTATE_URB_HS(struct drm_intel_decode * ctx)2897 gen7_3DSTATE_URB_HS(struct drm_intel_decode *ctx)
2898 {
2899 	return gen7_3DSTATE_URB_unit(ctx, "HS");
2900 }
2901 
2902 static int
gen7_3DSTATE_URB_DS(struct drm_intel_decode * ctx)2903 gen7_3DSTATE_URB_DS(struct drm_intel_decode *ctx)
2904 {
2905 	return gen7_3DSTATE_URB_unit(ctx, "DS");
2906 }
2907 
2908 static int
gen7_3DSTATE_URB_GS(struct drm_intel_decode * ctx)2909 gen7_3DSTATE_URB_GS(struct drm_intel_decode *ctx)
2910 {
2911 	return gen7_3DSTATE_URB_unit(ctx, "GS");
2912 }
2913 
2914 static int
gen7_3DSTATE_CONSTANT(struct drm_intel_decode * ctx,const char * unit)2915 gen7_3DSTATE_CONSTANT(struct drm_intel_decode *ctx, const char *unit)
2916 {
2917 	int rlen[4];
2918 
2919 	rlen[0] = (ctx->data[1] >> 0) & 0xffff;
2920 	rlen[1] = (ctx->data[1] >> 16) & 0xffff;
2921 	rlen[2] = (ctx->data[2] >> 0) & 0xffff;
2922 	rlen[3] = (ctx->data[2] >> 16) & 0xffff;
2923 
2924 	instr_out(ctx, 0, "3DSTATE_CONSTANT_%s\n", unit);
2925 	instr_out(ctx, 1, "len 0 = %d, len 1 = %d\n", rlen[0], rlen[1]);
2926 	instr_out(ctx, 2, "len 2 = %d, len 3 = %d\n", rlen[2], rlen[3]);
2927 	instr_out(ctx, 3, "pointer to constbuf 0\n");
2928 	instr_out(ctx, 4, "pointer to constbuf 1\n");
2929 	instr_out(ctx, 5, "pointer to constbuf 2\n");
2930 	instr_out(ctx, 6, "pointer to constbuf 3\n");
2931 
2932 	return 7;
2933 }
2934 
2935 static int
gen7_3DSTATE_CONSTANT_VS(struct drm_intel_decode * ctx)2936 gen7_3DSTATE_CONSTANT_VS(struct drm_intel_decode *ctx)
2937 {
2938 	return gen7_3DSTATE_CONSTANT(ctx, "VS");
2939 }
2940 
2941 static int
gen7_3DSTATE_CONSTANT_GS(struct drm_intel_decode * ctx)2942 gen7_3DSTATE_CONSTANT_GS(struct drm_intel_decode *ctx)
2943 {
2944 	return gen7_3DSTATE_CONSTANT(ctx, "GS");
2945 }
2946 
2947 static int
gen7_3DSTATE_CONSTANT_PS(struct drm_intel_decode * ctx)2948 gen7_3DSTATE_CONSTANT_PS(struct drm_intel_decode *ctx)
2949 {
2950 	return gen7_3DSTATE_CONSTANT(ctx, "PS");
2951 }
2952 
2953 static int
gen7_3DSTATE_CONSTANT_DS(struct drm_intel_decode * ctx)2954 gen7_3DSTATE_CONSTANT_DS(struct drm_intel_decode *ctx)
2955 {
2956 	return gen7_3DSTATE_CONSTANT(ctx, "DS");
2957 }
2958 
2959 static int
gen7_3DSTATE_CONSTANT_HS(struct drm_intel_decode * ctx)2960 gen7_3DSTATE_CONSTANT_HS(struct drm_intel_decode *ctx)
2961 {
2962 	return gen7_3DSTATE_CONSTANT(ctx, "HS");
2963 }
2964 
2965 
2966 static int
gen6_3DSTATE_WM(struct drm_intel_decode * ctx)2967 gen6_3DSTATE_WM(struct drm_intel_decode *ctx)
2968 {
2969 	instr_out(ctx, 0, "3DSTATE_WM\n");
2970 	instr_out(ctx, 1, "kernel start pointer 0\n");
2971 	instr_out(ctx, 2,
2972 		  "SPF=%d, VME=%d, Sampler Count %d, "
2973 		  "Binding table count %d\n",
2974 		  (ctx->data[2] >> 31) & 1,
2975 		  (ctx->data[2] >> 30) & 1,
2976 		  (ctx->data[2] >> 27) & 7,
2977 		  (ctx->data[2] >> 18) & 0xff);
2978 	instr_out(ctx, 3, "scratch offset\n");
2979 	instr_out(ctx, 4,
2980 		  "Depth Clear %d, Depth Resolve %d, HiZ Resolve %d, "
2981 		  "Dispatch GRF start[0] %d, start[1] %d, start[2] %d\n",
2982 		  (ctx->data[4] & (1 << 30)) != 0,
2983 		  (ctx->data[4] & (1 << 28)) != 0,
2984 		  (ctx->data[4] & (1 << 27)) != 0,
2985 		  (ctx->data[4] >> 16) & 0x7f,
2986 		  (ctx->data[4] >> 8) & 0x7f,
2987 		  (ctx->data[4] & 0x7f));
2988 	instr_out(ctx, 5,
2989 		  "MaxThreads %d, PS KillPixel %d, PS computed Z %d, "
2990 		  "PS use sourceZ %d, Thread Dispatch %d, PS use sourceW %d, "
2991 		  "Dispatch32 %d, Dispatch16 %d, Dispatch8 %d\n",
2992 		  ((ctx->data[5] >> 25) & 0x7f) + 1,
2993 		  (ctx->data[5] & (1 << 22)) != 0,
2994 		  (ctx->data[5] & (1 << 21)) != 0,
2995 		  (ctx->data[5] & (1 << 20)) != 0,
2996 		  (ctx->data[5] & (1 << 19)) != 0,
2997 		  (ctx->data[5] & (1 << 8)) != 0,
2998 		  (ctx->data[5] & (1 << 2)) != 0,
2999 		  (ctx->data[5] & (1 << 1)) != 0,
3000 		  (ctx->data[5] & (1 << 0)) != 0);
3001 	instr_out(ctx, 6,
3002 		  "Num SF output %d, Pos XY offset %d, ZW interp mode %d , "
3003 		  "Barycentric interp mode 0x%x, Point raster rule %d, "
3004 		  "Multisample mode %d, "
3005 		  "Multisample Dispatch mode %d\n",
3006 		  (ctx->data[6] >> 20) & 0x3f,
3007 		  (ctx->data[6] >> 18) & 3,
3008 		  (ctx->data[6] >> 16) & 3,
3009 		  (ctx->data[6] >> 10) & 0x3f,
3010 		  (ctx->data[6] & (1 << 9)) != 0,
3011 		  (ctx->data[6] >> 1) & 3,
3012 		  (ctx->data[6] & 1));
3013 	instr_out(ctx, 7, "kernel start pointer 1\n");
3014 	instr_out(ctx, 8, "kernel start pointer 2\n");
3015 
3016 	return 9;
3017 }
3018 
3019 static int
gen7_3DSTATE_WM(struct drm_intel_decode * ctx)3020 gen7_3DSTATE_WM(struct drm_intel_decode *ctx)
3021 {
3022 	const char *computed_depth = "";
3023 	const char *early_depth = "";
3024 	const char *zw_interp = "";
3025 
3026 	switch ((ctx->data[1] >> 23) & 0x3) {
3027 	case 0:
3028 		computed_depth = "";
3029 		break;
3030 	case 1:
3031 		computed_depth = "computed depth";
3032 		break;
3033 	case 2:
3034 		computed_depth = "computed depth >=";
3035 		break;
3036 	case 3:
3037 		computed_depth = "computed depth <=";
3038 		break;
3039 	}
3040 
3041 	switch ((ctx->data[1] >> 21) & 0x3) {
3042 	case 0:
3043 		early_depth = "";
3044 		break;
3045 	case 1:
3046 		early_depth = ", EDSC_PSEXEC";
3047 		break;
3048 	case 2:
3049 		early_depth = ", EDSC_PREPS";
3050 		break;
3051 	case 3:
3052 		early_depth = ", BAD EDSC";
3053 		break;
3054 	}
3055 
3056 	switch ((ctx->data[1] >> 17) & 0x3) {
3057 	case 0:
3058 		early_depth = "";
3059 		break;
3060 	case 1:
3061 		early_depth = ", BAD ZW interp";
3062 		break;
3063 	case 2:
3064 		early_depth = ", ZW centroid";
3065 		break;
3066 	case 3:
3067 		early_depth = ", ZW sample";
3068 		break;
3069 	}
3070 
3071 	instr_out(ctx, 0, "3DSTATE_WM\n");
3072 	instr_out(ctx, 1, "(%s%s%s%s%s%s)%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
3073 		  (ctx->data[1] & (1 << 11)) ? "PP " : "",
3074 		  (ctx->data[1] & (1 << 12)) ? "PC " : "",
3075 		  (ctx->data[1] & (1 << 13)) ? "PS " : "",
3076 		  (ctx->data[1] & (1 << 14)) ? "NPP " : "",
3077 		  (ctx->data[1] & (1 << 15)) ? "NPC " : "",
3078 		  (ctx->data[1] & (1 << 16)) ? "NPS " : "",
3079 		  (ctx->data[1] & (1 << 30)) ? ", depth clear" : "",
3080 		  (ctx->data[1] & (1 << 29)) ? "" : ", disabled",
3081 		  (ctx->data[1] & (1 << 28)) ? ", depth resolve" : "",
3082 		  (ctx->data[1] & (1 << 27)) ? ", hiz resolve" : "",
3083 		  (ctx->data[1] & (1 << 25)) ? ", kill" : "",
3084 		  computed_depth,
3085 		  early_depth,
3086 		  zw_interp,
3087 		  (ctx->data[1] & (1 << 20)) ? ", source depth" : "",
3088 		  (ctx->data[1] & (1 << 19)) ? ", source W" : "",
3089 		  (ctx->data[1] & (1 << 10)) ? ", coverage" : "",
3090 		  (ctx->data[1] & (1 << 4)) ? ", poly stipple" : "",
3091 		  (ctx->data[1] & (1 << 3)) ? ", line stipple" : "",
3092 		  (ctx->data[1] & (1 << 2)) ? ", point UL" : ", point UR"
3093 		  );
3094 	instr_out(ctx, 2, "MS\n");
3095 
3096 	return 3;
3097 }
3098 
3099 static int
gen4_3DPRIMITIVE(struct drm_intel_decode * ctx)3100 gen4_3DPRIMITIVE(struct drm_intel_decode *ctx)
3101 {
3102 	instr_out(ctx, 0,
3103 		  "3DPRIMITIVE: %s %s\n",
3104 		  get_965_prim_type((ctx->data[0] >> 10) & 0x1f),
3105 		  (ctx->data[0] & (1 << 15)) ? "random" : "sequential");
3106 	instr_out(ctx, 1, "vertex count\n");
3107 	instr_out(ctx, 2, "start vertex\n");
3108 	instr_out(ctx, 3, "instance count\n");
3109 	instr_out(ctx, 4, "start instance\n");
3110 	instr_out(ctx, 5, "index bias\n");
3111 
3112 	return 6;
3113 }
3114 
3115 static int
gen7_3DPRIMITIVE(struct drm_intel_decode * ctx)3116 gen7_3DPRIMITIVE(struct drm_intel_decode *ctx)
3117 {
3118 	bool indirect = !!(ctx->data[0] & (1 << 10));
3119 
3120 	instr_out(ctx, 0,
3121 		  "3DPRIMITIVE: %s%s\n",
3122 		  indirect ? " indirect" : "",
3123 		  (ctx->data[0] & (1 << 8)) ? " predicated" : "");
3124 	instr_out(ctx, 1, "%s %s\n",
3125 		  get_965_prim_type(ctx->data[1] & 0x3f),
3126 		  (ctx->data[1] & (1 << 8)) ? "random" : "sequential");
3127 	instr_out(ctx, 2, indirect ? "ignored" : "vertex count\n");
3128 	instr_out(ctx, 3, indirect ? "ignored" : "start vertex\n");
3129 	instr_out(ctx, 4, indirect ? "ignored" : "instance count\n");
3130 	instr_out(ctx, 5, indirect ? "ignored" : "start instance\n");
3131 	instr_out(ctx, 6, indirect ? "ignored" : "index bias\n");
3132 
3133 	return 7;
3134 }
3135 
3136 static int
decode_3d_965(struct drm_intel_decode * ctx)3137 decode_3d_965(struct drm_intel_decode *ctx)
3138 {
3139 	uint32_t opcode;
3140 	unsigned int len;
3141 	unsigned int i, j, sba_len;
3142 	const char *desc1 = NULL;
3143 	uint32_t *data = ctx->data;
3144 	uint32_t devid = ctx->devid;
3145 
3146 	struct {
3147 		uint32_t opcode;
3148 		uint32_t len_mask;
3149 		int unsigned min_len;
3150 		int unsigned max_len;
3151 		const char *name;
3152 		int gen;
3153 		int (*func)(struct drm_intel_decode *ctx);
3154 	} opcodes_3d[] = {
3155 		{ 0x6000, 0x00ff, 3, 3, "URB_FENCE" },
3156 		{ 0x6001, 0xffff, 2, 2, "CS_URB_STATE" },
3157 		{ 0x6002, 0x00ff, 2, 2, "CONSTANT_BUFFER" },
3158 		{ 0x6101, 0xffff, 6, 10, "STATE_BASE_ADDRESS" },
3159 		{ 0x6102, 0xffff, 2, 2, "STATE_SIP" },
3160 		{ 0x6104, 0xffff, 1, 1, "3DSTATE_PIPELINE_SELECT" },
3161 		{ 0x680b, 0xffff, 1, 1, "3DSTATE_VF_STATISTICS" },
3162 		{ 0x6904, 0xffff, 1, 1, "3DSTATE_PIPELINE_SELECT" },
3163 		{ 0x7800, 0xffff, 7, 7, "3DSTATE_PIPELINED_POINTERS" },
3164 		{ 0x7801, 0x00ff, 4, 6, "3DSTATE_BINDING_TABLE_POINTERS" },
3165 		{ 0x7802, 0x00ff, 4, 4, "3DSTATE_SAMPLER_STATE_POINTERS" },
3166 		{ 0x7805, 0x00ff, 7, 7, "3DSTATE_DEPTH_BUFFER", 7 },
3167 		{ 0x7805, 0x00ff, 3, 3, "3DSTATE_URB" },
3168 		{ 0x7804, 0x00ff, 3, 3, "3DSTATE_CLEAR_PARAMS" },
3169 		{ 0x7806, 0x00ff, 3, 3, "3DSTATE_STENCIL_BUFFER" },
3170 		{ 0x790f, 0x00ff, 3, 3, "3DSTATE_HIER_DEPTH_BUFFER", 6 },
3171 		{ 0x7807, 0x00ff, 3, 3, "3DSTATE_HIER_DEPTH_BUFFER", 7, gen7_3DSTATE_HIER_DEPTH_BUFFER },
3172 		{ 0x7808, 0x00ff, 5, 257, "3DSTATE_VERTEX_BUFFERS" },
3173 		{ 0x7809, 0x00ff, 3, 256, "3DSTATE_VERTEX_ELEMENTS" },
3174 		{ 0x780a, 0x00ff, 3, 3, "3DSTATE_INDEX_BUFFER" },
3175 		{ 0x780b, 0xffff, 1, 1, "3DSTATE_VF_STATISTICS" },
3176 		{ 0x780d, 0x00ff, 4, 4, "3DSTATE_VIEWPORT_STATE_POINTERS" },
3177 		{ 0x780e, 0xffff, 4, 4, NULL, 6, gen6_3DSTATE_CC_STATE_POINTERS },
3178 		{ 0x780e, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_CC_STATE_POINTERS },
3179 		{ 0x780f, 0x00ff, 2, 2, "3DSTATE_SCISSOR_POINTERS" },
3180 		{ 0x7810, 0x00ff, 6, 6, "3DSTATE_VS" },
3181 		{ 0x7811, 0x00ff, 7, 7, "3DSTATE_GS" },
3182 		{ 0x7812, 0x00ff, 4, 4, "3DSTATE_CLIP" },
3183 		{ 0x7813, 0x00ff, 20, 20, "3DSTATE_SF", 6 },
3184 		{ 0x7813, 0x00ff, 7, 7, "3DSTATE_SF", 7 },
3185 		{ 0x7814, 0x00ff, 3, 3, "3DSTATE_WM", 7, gen7_3DSTATE_WM },
3186 		{ 0x7814, 0x00ff, 9, 9, "3DSTATE_WM", 6, gen6_3DSTATE_WM },
3187 		{ 0x7815, 0x00ff, 5, 5, "3DSTATE_CONSTANT_VS_STATE", 6 },
3188 		{ 0x7815, 0x00ff, 7, 7, "3DSTATE_CONSTANT_VS", 7, gen7_3DSTATE_CONSTANT_VS },
3189 		{ 0x7816, 0x00ff, 5, 5, "3DSTATE_CONSTANT_GS_STATE", 6 },
3190 		{ 0x7816, 0x00ff, 7, 7, "3DSTATE_CONSTANT_GS", 7, gen7_3DSTATE_CONSTANT_GS },
3191 		{ 0x7817, 0x00ff, 5, 5, "3DSTATE_CONSTANT_PS_STATE", 6 },
3192 		{ 0x7817, 0x00ff, 7, 7, "3DSTATE_CONSTANT_PS", 7, gen7_3DSTATE_CONSTANT_PS },
3193 		{ 0x7818, 0xffff, 2, 2, "3DSTATE_SAMPLE_MASK" },
3194 		{ 0x7819, 0x00ff, 7, 7, "3DSTATE_CONSTANT_HS", 7, gen7_3DSTATE_CONSTANT_HS },
3195 		{ 0x781a, 0x00ff, 7, 7, "3DSTATE_CONSTANT_DS", 7, gen7_3DSTATE_CONSTANT_DS },
3196 		{ 0x781b, 0x00ff, 7, 7, "3DSTATE_HS" },
3197 		{ 0x781c, 0x00ff, 4, 4, "3DSTATE_TE" },
3198 		{ 0x781d, 0x00ff, 6, 6, "3DSTATE_DS" },
3199 		{ 0x781e, 0x00ff, 3, 3, "3DSTATE_STREAMOUT" },
3200 		{ 0x781f, 0x00ff, 14, 14, "3DSTATE_SBE" },
3201 		{ 0x7820, 0x00ff, 8, 8, "3DSTATE_PS" },
3202 		{ 0x7821, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP },
3203 		{ 0x7823, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_VIEWPORT_STATE_POINTERS_CC },
3204 		{ 0x7824, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_BLEND_STATE_POINTERS },
3205 		{ 0x7825, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_DEPTH_STENCIL_STATE_POINTERS },
3206 		{ 0x7826, 0x00ff, 2, 2, "3DSTATE_BINDING_TABLE_POINTERS_VS" },
3207 		{ 0x7827, 0x00ff, 2, 2, "3DSTATE_BINDING_TABLE_POINTERS_HS" },
3208 		{ 0x7828, 0x00ff, 2, 2, "3DSTATE_BINDING_TABLE_POINTERS_DS" },
3209 		{ 0x7829, 0x00ff, 2, 2, "3DSTATE_BINDING_TABLE_POINTERS_GS" },
3210 		{ 0x782a, 0x00ff, 2, 2, "3DSTATE_BINDING_TABLE_POINTERS_PS" },
3211 		{ 0x782b, 0x00ff, 2, 2, "3DSTATE_SAMPLER_STATE_POINTERS_VS" },
3212 		{ 0x782c, 0x00ff, 2, 2, "3DSTATE_SAMPLER_STATE_POINTERS_HS" },
3213 		{ 0x782d, 0x00ff, 2, 2, "3DSTATE_SAMPLER_STATE_POINTERS_DS" },
3214 		{ 0x782e, 0x00ff, 2, 2, "3DSTATE_SAMPLER_STATE_POINTERS_GS" },
3215 		{ 0x782f, 0x00ff, 2, 2, "3DSTATE_SAMPLER_STATE_POINTERS_PS" },
3216 		{ 0x7830, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_URB_VS },
3217 		{ 0x7831, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_URB_HS },
3218 		{ 0x7832, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_URB_DS },
3219 		{ 0x7833, 0x00ff, 2, 2, NULL, 7, gen7_3DSTATE_URB_GS },
3220 		{ 0x7900, 0xffff, 4, 4, "3DSTATE_DRAWING_RECTANGLE" },
3221 		{ 0x7901, 0xffff, 5, 5, "3DSTATE_CONSTANT_COLOR" },
3222 		{ 0x7905, 0xffff, 5, 7, "3DSTATE_DEPTH_BUFFER" },
3223 		{ 0x7906, 0xffff, 2, 2, "3DSTATE_POLY_STIPPLE_OFFSET" },
3224 		{ 0x7907, 0xffff, 33, 33, "3DSTATE_POLY_STIPPLE_PATTERN" },
3225 		{ 0x7908, 0xffff, 3, 3, "3DSTATE_LINE_STIPPLE" },
3226 		{ 0x7909, 0xffff, 2, 2, "3DSTATE_GLOBAL_DEPTH_OFFSET_CLAMP" },
3227 		{ 0x7909, 0xffff, 2, 2, "3DSTATE_CLEAR_PARAMS" },
3228 		{ 0x790a, 0xffff, 3, 3, "3DSTATE_AA_LINE_PARAMETERS" },
3229 		{ 0x790b, 0xffff, 4, 4, "3DSTATE_GS_SVB_INDEX" },
3230 		{ 0x790d, 0xffff, 3, 3, "3DSTATE_MULTISAMPLE", 6 },
3231 		{ 0x790d, 0xffff, 4, 4, "3DSTATE_MULTISAMPLE", 7 },
3232 		{ 0x7910, 0x00ff, 2, 2, "3DSTATE_CLEAR_PARAMS" },
3233 		{ 0x7912, 0x00ff, 2, 2, "3DSTATE_PUSH_CONSTANT_ALLOC_VS" },
3234 		{ 0x7913, 0x00ff, 2, 2, "3DSTATE_PUSH_CONSTANT_ALLOC_HS" },
3235 		{ 0x7914, 0x00ff, 2, 2, "3DSTATE_PUSH_CONSTANT_ALLOC_DS" },
3236 		{ 0x7915, 0x00ff, 2, 2, "3DSTATE_PUSH_CONSTANT_ALLOC_GS" },
3237 		{ 0x7916, 0x00ff, 2, 2, "3DSTATE_PUSH_CONSTANT_ALLOC_PS" },
3238 		{ 0x7917, 0x00ff, 2, 2+128*2, "3DSTATE_SO_DECL_LIST" },
3239 		{ 0x7918, 0x00ff, 4, 4, "3DSTATE_SO_BUFFER" },
3240 		{ 0x7a00, 0x00ff, 4, 6, "PIPE_CONTROL" },
3241 		{ 0x7b00, 0x00ff, 7, 7, NULL, 7, gen7_3DPRIMITIVE },
3242 		{ 0x7b00, 0x00ff, 6, 6, NULL, 0, gen4_3DPRIMITIVE },
3243 	}, *opcode_3d = NULL;
3244 
3245 	opcode = (data[0] & 0xffff0000) >> 16;
3246 
3247 	for (i = 0; i < ARRAY_SIZE(opcodes_3d); i++) {
3248 		if (opcode != opcodes_3d[i].opcode)
3249 			continue;
3250 
3251 		/* If it's marked as not our gen, skip. */
3252 		if (opcodes_3d[i].gen && opcodes_3d[i].gen != ctx->gen)
3253 			continue;
3254 
3255 		opcode_3d = &opcodes_3d[i];
3256 		break;
3257 	}
3258 
3259 	if (opcode_3d) {
3260 		if (opcode_3d->max_len == 1)
3261 			len = 1;
3262 		else
3263 			len = (data[0] & opcode_3d->len_mask) + 2;
3264 
3265 		if (len < opcode_3d->min_len ||
3266 		    len > opcode_3d->max_len) {
3267 			fprintf(out, "Bad length %d in %s, expected %d-%d\n",
3268 				len, opcode_3d->name,
3269 				opcode_3d->min_len, opcode_3d->max_len);
3270 		}
3271 	} else {
3272 		len = (data[0] & 0x0000ffff) + 2;
3273 	}
3274 
3275 	switch (opcode) {
3276 	case 0x6000:
3277 		return i965_decode_urb_fence(ctx, len);
3278 	case 0x6001:
3279 		instr_out(ctx, 0, "CS_URB_STATE\n");
3280 		instr_out(ctx, 1,
3281 			  "entry_size: %d [%d bytes], n_entries: %d\n",
3282 			  (data[1] >> 4) & 0x1f,
3283 			  (((data[1] >> 4) & 0x1f) + 1) * 64, data[1] & 0x7);
3284 		return len;
3285 	case 0x6002:
3286 		instr_out(ctx, 0, "CONSTANT_BUFFER: %s\n",
3287 			  (data[0] >> 8) & 1 ? "valid" : "invalid");
3288 		instr_out(ctx, 1,
3289 			  "offset: 0x%08x, length: %d bytes\n", data[1] & ~0x3f,
3290 			  ((data[1] & 0x3f) + 1) * 64);
3291 		return len;
3292 	case 0x6101:
3293 		i = 0;
3294 		instr_out(ctx, 0, "STATE_BASE_ADDRESS\n");
3295 		i++;
3296 
3297 		if (IS_GEN6(devid) || IS_GEN7(devid))
3298 			sba_len = 10;
3299 		else if (IS_GEN5(devid))
3300 			sba_len = 8;
3301 		else
3302 			sba_len = 6;
3303 		if (len != sba_len)
3304 			fprintf(out, "Bad count in STATE_BASE_ADDRESS\n");
3305 
3306 		state_base_out(ctx, i++, "general");
3307 		state_base_out(ctx, i++, "surface");
3308 		if (IS_GEN6(devid) || IS_GEN7(devid))
3309 			state_base_out(ctx, i++, "dynamic");
3310 		state_base_out(ctx, i++, "indirect");
3311 		if (IS_GEN5(devid) || IS_GEN6(devid) || IS_GEN7(devid))
3312 			state_base_out(ctx, i++, "instruction");
3313 
3314 		state_max_out(ctx, i++, "general");
3315 		if (IS_GEN6(devid) || IS_GEN7(devid))
3316 			state_max_out(ctx, i++, "dynamic");
3317 		state_max_out(ctx, i++, "indirect");
3318 		if (IS_GEN5(devid) || IS_GEN6(devid) || IS_GEN7(devid))
3319 			state_max_out(ctx, i++, "instruction");
3320 
3321 		return len;
3322 	case 0x7800:
3323 		instr_out(ctx, 0, "3DSTATE_PIPELINED_POINTERS\n");
3324 		instr_out(ctx, 1, "VS state\n");
3325 		instr_out(ctx, 2, "GS state\n");
3326 		instr_out(ctx, 3, "Clip state\n");
3327 		instr_out(ctx, 4, "SF state\n");
3328 		instr_out(ctx, 5, "WM state\n");
3329 		instr_out(ctx, 6, "CC state\n");
3330 		return len;
3331 	case 0x7801:
3332 		if (len != 6 && len != 4)
3333 			fprintf(out,
3334 				"Bad count in 3DSTATE_BINDING_TABLE_POINTERS\n");
3335 		if (len == 6) {
3336 			instr_out(ctx, 0,
3337 				  "3DSTATE_BINDING_TABLE_POINTERS\n");
3338 			instr_out(ctx, 1, "VS binding table\n");
3339 			instr_out(ctx, 2, "GS binding table\n");
3340 			instr_out(ctx, 3, "Clip binding table\n");
3341 			instr_out(ctx, 4, "SF binding table\n");
3342 			instr_out(ctx, 5, "WM binding table\n");
3343 		} else {
3344 			instr_out(ctx, 0,
3345 				  "3DSTATE_BINDING_TABLE_POINTERS: VS mod %d, "
3346 				  "GS mod %d, PS mod %d\n",
3347 				  (data[0] & (1 << 8)) != 0,
3348 				  (data[0] & (1 << 9)) != 0,
3349 				  (data[0] & (1 << 12)) != 0);
3350 			instr_out(ctx, 1, "VS binding table\n");
3351 			instr_out(ctx, 2, "GS binding table\n");
3352 			instr_out(ctx, 3, "WM binding table\n");
3353 		}
3354 
3355 		return len;
3356 	case 0x7802:
3357 		instr_out(ctx, 0,
3358 			  "3DSTATE_SAMPLER_STATE_POINTERS: VS mod %d, "
3359 			  "GS mod %d, PS mod %d\n", (data[0] & (1 << 8)) != 0,
3360 			  (data[0] & (1 << 9)) != 0,
3361 			  (data[0] & (1 << 12)) != 0);
3362 		instr_out(ctx, 1, "VS sampler state\n");
3363 		instr_out(ctx, 2, "GS sampler state\n");
3364 		instr_out(ctx, 3, "WM sampler state\n");
3365 		return len;
3366 	case 0x7805:
3367 		/* Actually 3DSTATE_DEPTH_BUFFER on gen7. */
3368 		if (ctx->gen == 7)
3369 			break;
3370 
3371 		instr_out(ctx, 0, "3DSTATE_URB\n");
3372 		instr_out(ctx, 1,
3373 			  "VS entries %d, alloc size %d (1024bit row)\n",
3374 			  data[1] & 0xffff, ((data[1] >> 16) & 0x07f) + 1);
3375 		instr_out(ctx, 2,
3376 			  "GS entries %d, alloc size %d (1024bit row)\n",
3377 			  (data[2] >> 8) & 0x3ff, (data[2] & 7) + 1);
3378 		return len;
3379 
3380 	case 0x7808:
3381 		if ((len - 1) % 4 != 0)
3382 			fprintf(out, "Bad count in 3DSTATE_VERTEX_BUFFERS\n");
3383 		instr_out(ctx, 0, "3DSTATE_VERTEX_BUFFERS\n");
3384 
3385 		for (i = 1; i < len;) {
3386 			int idx, access;
3387 			if (IS_GEN6(devid)) {
3388 				idx = 26;
3389 				access = 20;
3390 			} else {
3391 				idx = 27;
3392 				access = 26;
3393 			}
3394 			instr_out(ctx, i,
3395 				  "buffer %d: %s, pitch %db\n", data[i] >> idx,
3396 				  data[i] & (1 << access) ? "random" :
3397 				  "sequential", data[i] & 0x07ff);
3398 			i++;
3399 			instr_out(ctx, i++, "buffer address\n");
3400 			instr_out(ctx, i++, "max index\n");
3401 			instr_out(ctx, i++, "mbz\n");
3402 		}
3403 		return len;
3404 
3405 	case 0x7809:
3406 		if ((len + 1) % 2 != 0)
3407 			fprintf(out, "Bad count in 3DSTATE_VERTEX_ELEMENTS\n");
3408 		instr_out(ctx, 0, "3DSTATE_VERTEX_ELEMENTS\n");
3409 
3410 		for (i = 1; i < len;) {
3411 			instr_out(ctx, i,
3412 				  "buffer %d: %svalid, type 0x%04x, "
3413 				  "src offset 0x%04x bytes\n",
3414 				  data[i] >> ((IS_GEN6(devid) || IS_GEN7(devid)) ? 26 : 27),
3415 				  data[i] & (1 << ((IS_GEN6(devid) || IS_GEN7(devid)) ? 25 : 26)) ?
3416 				  "" : "in", (data[i] >> 16) & 0x1ff,
3417 				  data[i] & 0x07ff);
3418 			i++;
3419 			instr_out(ctx, i, "(%s, %s, %s, %s), "
3420 				  "dst offset 0x%02x bytes\n",
3421 				  get_965_element_component(data[i], 0),
3422 				  get_965_element_component(data[i], 1),
3423 				  get_965_element_component(data[i], 2),
3424 				  get_965_element_component(data[i], 3),
3425 				  (data[i] & 0xff) * 4);
3426 			i++;
3427 		}
3428 		return len;
3429 
3430 	case 0x780d:
3431 		instr_out(ctx, 0,
3432 			  "3DSTATE_VIEWPORT_STATE_POINTERS\n");
3433 		instr_out(ctx, 1, "clip\n");
3434 		instr_out(ctx, 2, "sf\n");
3435 		instr_out(ctx, 3, "cc\n");
3436 		return len;
3437 
3438 	case 0x780a:
3439 		instr_out(ctx, 0, "3DSTATE_INDEX_BUFFER\n");
3440 		instr_out(ctx, 1, "beginning buffer address\n");
3441 		instr_out(ctx, 2, "ending buffer address\n");
3442 		return len;
3443 
3444 	case 0x780f:
3445 		instr_out(ctx, 0, "3DSTATE_SCISSOR_POINTERS\n");
3446 		instr_out(ctx, 1, "scissor rect offset\n");
3447 		return len;
3448 
3449 	case 0x7810:
3450 		instr_out(ctx, 0, "3DSTATE_VS\n");
3451 		instr_out(ctx, 1, "kernel pointer\n");
3452 		instr_out(ctx, 2,
3453 			  "SPF=%d, VME=%d, Sampler Count %d, "
3454 			  "Binding table count %d\n", (data[2] >> 31) & 1,
3455 			  (data[2] >> 30) & 1, (data[2] >> 27) & 7,
3456 			  (data[2] >> 18) & 0xff);
3457 		instr_out(ctx, 3, "scratch offset\n");
3458 		instr_out(ctx, 4,
3459 			  "Dispatch GRF start %d, VUE read length %d, "
3460 			  "VUE read offset %d\n", (data[4] >> 20) & 0x1f,
3461 			  (data[4] >> 11) & 0x3f, (data[4] >> 4) & 0x3f);
3462 		instr_out(ctx, 5,
3463 			  "Max Threads %d, Vertex Cache %sable, "
3464 			  "VS func %sable\n", ((data[5] >> 25) & 0x7f) + 1,
3465 			  (data[5] & (1 << 1)) != 0 ? "dis" : "en",
3466 			  (data[5] & 1) != 0 ? "en" : "dis");
3467 		return len;
3468 
3469 	case 0x7811:
3470 		instr_out(ctx, 0, "3DSTATE_GS\n");
3471 		instr_out(ctx, 1, "kernel pointer\n");
3472 		instr_out(ctx, 2,
3473 			  "SPF=%d, VME=%d, Sampler Count %d, "
3474 			  "Binding table count %d\n", (data[2] >> 31) & 1,
3475 			  (data[2] >> 30) & 1, (data[2] >> 27) & 7,
3476 			  (data[2] >> 18) & 0xff);
3477 		instr_out(ctx, 3, "scratch offset\n");
3478 		instr_out(ctx, 4,
3479 			  "Dispatch GRF start %d, VUE read length %d, "
3480 			  "VUE read offset %d\n", (data[4] & 0xf),
3481 			  (data[4] >> 11) & 0x3f, (data[4] >> 4) & 0x3f);
3482 		instr_out(ctx, 5,
3483 			  "Max Threads %d, Rendering %sable\n",
3484 			  ((data[5] >> 25) & 0x7f) + 1,
3485 			  (data[5] & (1 << 8)) != 0 ? "en" : "dis");
3486 		instr_out(ctx, 6,
3487 			  "Reorder %sable, Discard Adjaceny %sable, "
3488 			  "GS %sable\n",
3489 			  (data[6] & (1 << 30)) != 0 ? "en" : "dis",
3490 			  (data[6] & (1 << 29)) != 0 ? "en" : "dis",
3491 			  (data[6] & (1 << 15)) != 0 ? "en" : "dis");
3492 		return len;
3493 
3494 	case 0x7812:
3495 		instr_out(ctx, 0, "3DSTATE_CLIP\n");
3496 		instr_out(ctx, 1,
3497 			  "UserClip distance cull test mask 0x%x\n",
3498 			  data[1] & 0xff);
3499 		instr_out(ctx, 2,
3500 			  "Clip %sable, API mode %s, Viewport XY test %sable, "
3501 			  "Viewport Z test %sable, Guardband test %sable, Clip mode %d, "
3502 			  "Perspective Divide %sable, Non-Perspective Barycentric %sable, "
3503 			  "Tri Provoking %d, Line Provoking %d, Trifan Provoking %d\n",
3504 			  (data[2] & (1 << 31)) != 0 ? "en" : "dis",
3505 			  (data[2] & (1 << 30)) != 0 ? "D3D" : "OGL",
3506 			  (data[2] & (1 << 28)) != 0 ? "en" : "dis",
3507 			  (data[2] & (1 << 27)) != 0 ? "en" : "dis",
3508 			  (data[2] & (1 << 26)) != 0 ? "en" : "dis",
3509 			  (data[2] >> 13) & 7,
3510 			  (data[2] & (1 << 9)) != 0 ? "dis" : "en",
3511 			  (data[2] & (1 << 8)) != 0 ? "en" : "dis",
3512 			  (data[2] >> 4) & 3, (data[2] >> 2) & 3,
3513 			  (data[2] & 3));
3514 		instr_out(ctx, 3,
3515 			  "Min PointWidth %d, Max PointWidth %d, "
3516 			  "Force Zero RTAIndex %sable, Max VPIndex %d\n",
3517 			  (data[3] >> 17) & 0x7ff, (data[3] >> 6) & 0x7ff,
3518 			  (data[3] & (1 << 5)) != 0 ? "en" : "dis",
3519 			  (data[3] & 0xf));
3520 		return len;
3521 
3522 	case 0x7813:
3523 		if (ctx->gen == 7)
3524 			break;
3525 
3526 		instr_out(ctx, 0, "3DSTATE_SF\n");
3527 		instr_out(ctx, 1,
3528 			  "Attrib Out %d, Attrib Swizzle %sable, VUE read length %d, "
3529 			  "VUE read offset %d\n", (data[1] >> 22) & 0x3f,
3530 			  (data[1] & (1 << 21)) != 0 ? "en" : "dis",
3531 			  (data[1] >> 11) & 0x1f, (data[1] >> 4) & 0x3f);
3532 		instr_out(ctx, 2,
3533 			  "Legacy Global DepthBias %sable, FrontFace fill %d, BF fill %d, "
3534 			  "VP transform %sable, FrontWinding_%s\n",
3535 			  (data[2] & (1 << 11)) != 0 ? "en" : "dis",
3536 			  (data[2] >> 5) & 3, (data[2] >> 3) & 3,
3537 			  (data[2] & (1 << 1)) != 0 ? "en" : "dis",
3538 			  (data[2] & 1) != 0 ? "CCW" : "CW");
3539 		instr_out(ctx, 3,
3540 			  "AA %sable, CullMode %d, Scissor %sable, Multisample m ode %d\n",
3541 			  (data[3] & (1 << 31)) != 0 ? "en" : "dis",
3542 			  (data[3] >> 29) & 3,
3543 			  (data[3] & (1 << 11)) != 0 ? "en" : "dis",
3544 			  (data[3] >> 8) & 3);
3545 		instr_out(ctx, 4,
3546 			  "Last Pixel %sable, SubPixel Precision %d, Use PixelWidth %d\n",
3547 			  (data[4] & (1 << 31)) != 0 ? "en" : "dis",
3548 			  (data[4] & (1 << 12)) != 0 ? 4 : 8,
3549 			  (data[4] & (1 << 11)) != 0);
3550 		instr_out(ctx, 5,
3551 			  "Global Depth Offset Constant %f\n",
3552 			  *(float *)(&data[5]));
3553 		instr_out(ctx, 6, "Global Depth Offset Scale %f\n",
3554 			  *(float *)(&data[6]));
3555 		instr_out(ctx, 7, "Global Depth Offset Clamp %f\n",
3556 			  *(float *)(&data[7]));
3557 
3558 		for (i = 0, j = 0; i < 8; i++, j += 2)
3559 			instr_out(ctx, i + 8,
3560 				  "Attrib %d (Override %s%s%s%s, Const Source %d, Swizzle Select %d, "
3561 				  "Source %d); Attrib %d (Override %s%s%s%s, Const Source %d, Swizzle Select %d, Source %d)\n",
3562 				  j + 1,
3563 				  (data[8 + i] & (1 << 31)) != 0 ? "W" : "",
3564 				  (data[8 + i] & (1 << 30)) != 0 ? "Z" : "",
3565 				  (data[8 + i] & (1 << 29)) != 0 ? "Y" : "",
3566 				  (data[8 + i] & (1 << 28)) != 0 ? "X" : "",
3567 				  (data[8 + i] >> 25) & 3,
3568 				  (data[8 + i] >> 22) & 3,
3569 				  (data[8 + i] >> 16) & 0x1f, j,
3570 				  (data[8 + i] & (1 << 15)) != 0 ? "W" : "",
3571 				  (data[8 + i] & (1 << 14)) != 0 ? "Z" : "",
3572 				  (data[8 + i] & (1 << 13)) != 0 ? "Y" : "",
3573 				  (data[8 + i] & (1 << 12)) != 0 ? "X" : "",
3574 				  (data[8 + i] >> 9) & 3,
3575 				  (data[8 + i] >> 6) & 3, (data[8 + i] & 0x1f));
3576 		instr_out(ctx, 16,
3577 			  "Point Sprite TexCoord Enable\n");
3578 		instr_out(ctx, 17, "Const Interp Enable\n");
3579 		instr_out(ctx, 18,
3580 			  "Attrib 7-0 WrapShortest Enable\n");
3581 		instr_out(ctx, 19,
3582 			  "Attrib 15-8 WrapShortest Enable\n");
3583 
3584 		return len;
3585 
3586 	case 0x7900:
3587 		instr_out(ctx, 0, "3DSTATE_DRAWING_RECTANGLE\n");
3588 		instr_out(ctx, 1, "top left: %d,%d\n",
3589 			  data[1] & 0xffff, (data[1] >> 16) & 0xffff);
3590 		instr_out(ctx, 2, "bottom right: %d,%d\n",
3591 			  data[2] & 0xffff, (data[2] >> 16) & 0xffff);
3592 		instr_out(ctx, 3, "origin: %d,%d\n",
3593 			  (int)data[3] & 0xffff, ((int)data[3] >> 16) & 0xffff);
3594 
3595 		return len;
3596 
3597 	case 0x7905:
3598 		instr_out(ctx, 0, "3DSTATE_DEPTH_BUFFER\n");
3599 		if (IS_GEN5(devid) || IS_GEN6(devid))
3600 			instr_out(ctx, 1,
3601 				  "%s, %s, pitch = %d bytes, %stiled, HiZ %d, Separate Stencil %d\n",
3602 				  get_965_surfacetype(data[1] >> 29),
3603 				  get_965_depthformat((data[1] >> 18) & 0x7),
3604 				  (data[1] & 0x0001ffff) + 1,
3605 				  data[1] & (1 << 27) ? "" : "not ",
3606 				  (data[1] & (1 << 22)) != 0,
3607 				  (data[1] & (1 << 21)) != 0);
3608 		else
3609 			instr_out(ctx, 1,
3610 				  "%s, %s, pitch = %d bytes, %stiled\n",
3611 				  get_965_surfacetype(data[1] >> 29),
3612 				  get_965_depthformat((data[1] >> 18) & 0x7),
3613 				  (data[1] & 0x0001ffff) + 1,
3614 				  data[1] & (1 << 27) ? "" : "not ");
3615 		instr_out(ctx, 2, "depth offset\n");
3616 		instr_out(ctx, 3, "%dx%d\n",
3617 			  ((data[3] & 0x0007ffc0) >> 6) + 1,
3618 			  ((data[3] & 0xfff80000) >> 19) + 1);
3619 		instr_out(ctx, 4, "volume depth\n");
3620 		if (len >= 6)
3621 			instr_out(ctx, 5, "\n");
3622 		if (len >= 7) {
3623 			if (IS_GEN6(devid))
3624 				instr_out(ctx, 6, "\n");
3625 			else
3626 				instr_out(ctx, 6,
3627 					  "render target view extent\n");
3628 		}
3629 
3630 		return len;
3631 
3632 	case 0x7a00:
3633 		if (IS_GEN6(devid) || IS_GEN7(devid)) {
3634 			if (len != 4 && len != 5)
3635 				fprintf(out, "Bad count in PIPE_CONTROL\n");
3636 
3637 			switch ((data[1] >> 14) & 0x3) {
3638 			case 0:
3639 				desc1 = "no write";
3640 				break;
3641 			case 1:
3642 				desc1 = "qword write";
3643 				break;
3644 			case 2:
3645 				desc1 = "PS_DEPTH_COUNT write";
3646 				break;
3647 			case 3:
3648 				desc1 = "TIMESTAMP write";
3649 				break;
3650 			}
3651 			instr_out(ctx, 0, "PIPE_CONTROL\n");
3652 			instr_out(ctx, 1,
3653 				  "%s, %s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
3654 				  desc1,
3655 				  data[1] & (1 << 20) ? "cs stall, " : "",
3656 				  data[1] & (1 << 19) ?
3657 				  "global snapshot count reset, " : "",
3658 				  data[1] & (1 << 18) ? "tlb invalidate, " : "",
3659 				  data[1] & (1 << 17) ? "gfdt flush, " : "",
3660 				  data[1] & (1 << 17) ? "media state clear, " :
3661 				  "",
3662 				  data[1] & (1 << 13) ? "depth stall, " : "",
3663 				  data[1] & (1 << 12) ?
3664 				  "render target cache flush, " : "",
3665 				  data[1] & (1 << 11) ?
3666 				  "instruction cache invalidate, " : "",
3667 				  data[1] & (1 << 10) ?
3668 				  "texture cache invalidate, " : "",
3669 				  data[1] & (1 << 9) ?
3670 				  "indirect state invalidate, " : "",
3671 				  data[1] & (1 << 8) ? "notify irq, " : "",
3672 				  data[1] & (1 << 7) ? "PIPE_CONTROL flush, " :
3673 				  "",
3674 				  data[1] & (1 << 6) ? "protect mem app_id, " :
3675 				  "", data[1] & (1 << 5) ? "DC flush, " : "",
3676 				  data[1] & (1 << 4) ? "vf fetch invalidate, " :
3677 				  "",
3678 				  data[1] & (1 << 3) ?
3679 				  "constant cache invalidate, " : "",
3680 				  data[1] & (1 << 2) ?
3681 				  "state cache invalidate, " : "",
3682 				  data[1] & (1 << 1) ? "stall at scoreboard, " :
3683 				  "",
3684 				  data[1] & (1 << 0) ? "depth cache flush, " :
3685 				  "");
3686 			if (len == 5) {
3687 				instr_out(ctx, 2,
3688 					  "destination address\n");
3689 				instr_out(ctx, 3,
3690 					  "immediate dword low\n");
3691 				instr_out(ctx, 4,
3692 					  "immediate dword high\n");
3693 			} else {
3694 				for (i = 2; i < len; i++) {
3695 					instr_out(ctx, i, "\n");
3696 				}
3697 			}
3698 			return len;
3699 		} else {
3700 			if (len != 4)
3701 				fprintf(out, "Bad count in PIPE_CONTROL\n");
3702 
3703 			switch ((data[0] >> 14) & 0x3) {
3704 			case 0:
3705 				desc1 = "no write";
3706 				break;
3707 			case 1:
3708 				desc1 = "qword write";
3709 				break;
3710 			case 2:
3711 				desc1 = "PS_DEPTH_COUNT write";
3712 				break;
3713 			case 3:
3714 				desc1 = "TIMESTAMP write";
3715 				break;
3716 			}
3717 			instr_out(ctx, 0,
3718 				  "PIPE_CONTROL: %s, %sdepth stall, %sRC write flush, "
3719 				  "%sinst flush\n",
3720 				  desc1,
3721 				  data[0] & (1 << 13) ? "" : "no ",
3722 				  data[0] & (1 << 12) ? "" : "no ",
3723 				  data[0] & (1 << 11) ? "" : "no ");
3724 			instr_out(ctx, 1, "destination address\n");
3725 			instr_out(ctx, 2, "immediate dword low\n");
3726 			instr_out(ctx, 3, "immediate dword high\n");
3727 			return len;
3728 		}
3729 	}
3730 
3731 	if (opcode_3d) {
3732 		if (opcode_3d->func) {
3733 			return opcode_3d->func(ctx);
3734 		} else {
3735 			instr_out(ctx, 0, "%s\n", opcode_3d->name);
3736 
3737 			for (i = 1; i < len; i++) {
3738 				instr_out(ctx, i, "dword %d\n", i);
3739 			}
3740 			return len;
3741 		}
3742 	}
3743 
3744 	instr_out(ctx, 0, "3D UNKNOWN: 3d_965 opcode = 0x%x\n",
3745 		  opcode);
3746 	return 1;
3747 }
3748 
3749 static int
decode_3d_i830(struct drm_intel_decode * ctx)3750 decode_3d_i830(struct drm_intel_decode *ctx)
3751 {
3752 	unsigned int idx;
3753 	uint32_t opcode;
3754 	uint32_t *data = ctx->data;
3755 
3756 	struct {
3757 		uint32_t opcode;
3758 		unsigned int min_len;
3759 		unsigned int max_len;
3760 		const char *name;
3761 	} opcodes_3d[] = {
3762 		{ 0x02, 1, 1, "3DSTATE_MODES_3" },
3763 		{ 0x03, 1, 1, "3DSTATE_ENABLES_1" },
3764 		{ 0x04, 1, 1, "3DSTATE_ENABLES_2" },
3765 		{ 0x05, 1, 1, "3DSTATE_VFT0" },
3766 		{ 0x06, 1, 1, "3DSTATE_AA" },
3767 		{ 0x07, 1, 1, "3DSTATE_RASTERIZATION_RULES" },
3768 		{ 0x08, 1, 1, "3DSTATE_MODES_1" },
3769 		{ 0x09, 1, 1, "3DSTATE_STENCIL_TEST" },
3770 		{ 0x0a, 1, 1, "3DSTATE_VFT1" },
3771 		{ 0x0b, 1, 1, "3DSTATE_INDPT_ALPHA_BLEND" },
3772 		{ 0x0c, 1, 1, "3DSTATE_MODES_5" },
3773 		{ 0x0d, 1, 1, "3DSTATE_MAP_BLEND_OP" },
3774 		{ 0x0e, 1, 1, "3DSTATE_MAP_BLEND_ARG" },
3775 		{ 0x0f, 1, 1, "3DSTATE_MODES_2" },
3776 		{ 0x15, 1, 1, "3DSTATE_FOG_COLOR" },
3777 		{ 0x16, 1, 1, "3DSTATE_MODES_4"},
3778 	}, *opcode_3d;
3779 
3780 	opcode = (data[0] & 0x1f000000) >> 24;
3781 
3782 	switch (opcode) {
3783 	case 0x1f:
3784 		return decode_3d_primitive(ctx);
3785 	case 0x1d:
3786 		return decode_3d_1d(ctx);
3787 	case 0x1c:
3788 		return decode_3d_1c(ctx);
3789 	}
3790 
3791 	for (idx = 0; idx < ARRAY_SIZE(opcodes_3d); idx++) {
3792 		opcode_3d = &opcodes_3d[idx];
3793 		if ((data[0] & 0x1f000000) >> 24 == opcode_3d->opcode) {
3794 			unsigned int len = 1, i;
3795 
3796 			instr_out(ctx, 0, "%s\n", opcode_3d->name);
3797 			if (opcode_3d->max_len > 1) {
3798 				len = (data[0] & 0xff) + 2;
3799 				if (len < opcode_3d->min_len ||
3800 				    len > opcode_3d->max_len) {
3801 					fprintf(out, "Bad count in %s\n",
3802 						opcode_3d->name);
3803 				}
3804 			}
3805 
3806 			for (i = 1; i < len; i++) {
3807 				instr_out(ctx, i, "dword %d\n", i);
3808 			}
3809 			return len;
3810 		}
3811 	}
3812 
3813 	instr_out(ctx, 0, "3D UNKNOWN: 3d_i830 opcode = 0x%x\n",
3814 		  opcode);
3815 	return 1;
3816 }
3817 
3818 struct drm_intel_decode *
drm_intel_decode_context_alloc(uint32_t devid)3819 drm_intel_decode_context_alloc(uint32_t devid)
3820 {
3821 	struct drm_intel_decode *ctx;
3822 
3823 	ctx = calloc(1, sizeof(struct drm_intel_decode));
3824 	if (!ctx)
3825 		return NULL;
3826 
3827 	ctx->devid = devid;
3828 	ctx->out = stdout;
3829 
3830 	if (IS_GEN9(devid))
3831 		ctx->gen = 9;
3832 	else if (IS_GEN8(devid))
3833 		ctx->gen = 8;
3834 	else if (IS_GEN7(devid))
3835 		ctx->gen = 7;
3836 	else if (IS_GEN6(devid))
3837 		ctx->gen = 6;
3838 	else if (IS_GEN5(devid))
3839 		ctx->gen = 5;
3840 	else if (IS_GEN4(devid))
3841 		ctx->gen = 4;
3842 	else if (IS_9XX(devid))
3843 		ctx->gen = 3;
3844 	else {
3845 		assert(IS_GEN2(devid));
3846 		ctx->gen = 2;
3847 	}
3848 
3849 	return ctx;
3850 }
3851 
3852 void
drm_intel_decode_context_free(struct drm_intel_decode * ctx)3853 drm_intel_decode_context_free(struct drm_intel_decode *ctx)
3854 {
3855 	free(ctx);
3856 }
3857 
3858 void
drm_intel_decode_set_dump_past_end(struct drm_intel_decode * ctx,int dump_past_end)3859 drm_intel_decode_set_dump_past_end(struct drm_intel_decode *ctx,
3860 				   int dump_past_end)
3861 {
3862 	ctx->dump_past_end = !!dump_past_end;
3863 }
3864 
3865 void
drm_intel_decode_set_batch_pointer(struct drm_intel_decode * ctx,void * data,uint32_t hw_offset,int count)3866 drm_intel_decode_set_batch_pointer(struct drm_intel_decode *ctx,
3867 				   void *data, uint32_t hw_offset, int count)
3868 {
3869 	ctx->base_data = data;
3870 	ctx->base_hw_offset = hw_offset;
3871 	ctx->base_count = count;
3872 }
3873 
3874 void
drm_intel_decode_set_head_tail(struct drm_intel_decode * ctx,uint32_t head,uint32_t tail)3875 drm_intel_decode_set_head_tail(struct drm_intel_decode *ctx,
3876 			       uint32_t head, uint32_t tail)
3877 {
3878 	ctx->head = head;
3879 	ctx->tail = tail;
3880 }
3881 
3882 void
drm_intel_decode_set_output_file(struct drm_intel_decode * ctx,FILE * output)3883 drm_intel_decode_set_output_file(struct drm_intel_decode *ctx,
3884 				 FILE *output)
3885 {
3886 	ctx->out = output;
3887 }
3888 
3889 /**
3890  * Decodes an i830-i915 batch buffer, writing the output to stdout.
3891  *
3892  * \param data batch buffer contents
3893  * \param count number of DWORDs to decode in the batch buffer
3894  * \param hw_offset hardware address for the buffer
3895  */
3896 void
drm_intel_decode(struct drm_intel_decode * ctx)3897 drm_intel_decode(struct drm_intel_decode *ctx)
3898 {
3899 	int ret;
3900 	unsigned int index = 0;
3901 	uint32_t devid;
3902 	int size = ctx->base_count * 4;
3903 	void *temp;
3904 
3905 	if (!ctx)
3906 		return;
3907 
3908 	/* Put a scratch page full of obviously undefined data after
3909 	 * the batchbuffer.  This lets us avoid a bunch of length
3910 	 * checking in statically sized packets.
3911 	 */
3912 	temp = malloc(size + 4096);
3913 	memcpy(temp, ctx->base_data, size);
3914 	memset((char *)temp + size, 0xd0, 4096);
3915 	ctx->data = temp;
3916 
3917 	ctx->hw_offset = ctx->base_hw_offset;
3918 	ctx->count = ctx->base_count;
3919 
3920 	devid = ctx->devid;
3921 	head_offset = ctx->head;
3922 	tail_offset = ctx->tail;
3923 	out = ctx->out;
3924 
3925 	saved_s2_set = 0;
3926 	saved_s4_set = 1;
3927 
3928 	while (ctx->count > 0) {
3929 		index = 0;
3930 
3931 		switch ((ctx->data[index] & 0xe0000000) >> 29) {
3932 		case 0x0:
3933 			ret = decode_mi(ctx);
3934 
3935 			/* If MI_BATCHBUFFER_END happened, then dump
3936 			 * the rest of the output in case we some day
3937 			 * want it in debugging, but don't decode it
3938 			 * since it'll just confuse in the common
3939 			 * case.
3940 			 */
3941 			if (ret == -1) {
3942 				if (ctx->dump_past_end) {
3943 					index++;
3944 				} else {
3945 					for (index = index + 1; index < ctx->count;
3946 					     index++) {
3947 						instr_out(ctx, index, "\n");
3948 					}
3949 				}
3950 			} else
3951 				index += ret;
3952 			break;
3953 		case 0x2:
3954 			index += decode_2d(ctx);
3955 			break;
3956 		case 0x3:
3957 			if (IS_9XX(devid) && !IS_GEN3(devid)) {
3958 				index +=
3959 				    decode_3d_965(ctx);
3960 			} else if (IS_GEN3(devid)) {
3961 				index += decode_3d(ctx);
3962 			} else {
3963 				index +=
3964 				    decode_3d_i830(ctx);
3965 			}
3966 			break;
3967 		default:
3968 			instr_out(ctx, index, "UNKNOWN\n");
3969 			index++;
3970 			break;
3971 		}
3972 		fflush(out);
3973 
3974 		if (ctx->count < index)
3975 			break;
3976 
3977 		ctx->count -= index;
3978 		ctx->data += index;
3979 		ctx->hw_offset += 4 * index;
3980 	}
3981 
3982 	free(temp);
3983 }
3984