summaryrefslogtreecommitdiffstats
path: root/drivers/gpu/drm/msm/adreno/a3xx_gpu.c
blob: 5273dc8498381ce09e878894f4eb56263900be39 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
// SPDX-License-Identifier: GPL-2.0-only
/*
 * Copyright (C) 2013 Red Hat
 * Author: Rob Clark <robdclark@gmail.com>
 *
 * Copyright (c) 2014 The Linux Foundation. All rights reserved.
 */

#include "a3xx_gpu.h"

#define A3XX_INT0_MASK \
	(A3XX_INT0_RBBM_AHB_ERROR |        \
	 A3XX_INT0_RBBM_ATB_BUS_OVERFLOW | \
	 A3XX_INT0_CP_T0_PACKET_IN_IB |    \
	 A3XX_INT0_CP_OPCODE_ERROR |       \
	 A3XX_INT0_CP_RESERVED_BIT_ERROR | \
	 A3XX_INT0_CP_HW_FAULT |           \
	 A3XX_INT0_CP_IB1_INT |            \
	 A3XX_INT0_CP_IB2_INT |            \
	 A3XX_INT0_CP_RB_INT |             \
	 A3XX_INT0_CP_REG_PROTECT_FAULT |  \
	 A3XX_INT0_CP_AHB_ERROR_HALT |     \
	 A3XX_INT0_CACHE_FLUSH_TS |        \
	 A3XX_INT0_UCHE_OOB_ACCESS)

extern bool hang_debug;

static void a3xx_dump(struct msm_gpu *gpu);
static bool a3xx_idle(struct msm_gpu *gpu);

static void a3xx_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit)
{
	struct msm_ringbuffer *ring = submit->ring;
	unsigned int i;

	for (i = 0; i < submit->nr_cmds; i++) {
		switch (submit->cmd[i].type) {
		case MSM_SUBMIT_CMD_IB_TARGET_BUF:
			/* ignore IB-targets */
			break;
		case MSM_SUBMIT_CMD_CTX_RESTORE_BUF:
			/* ignore if there has not been a ctx switch: */
			if (gpu->cur_ctx_seqno == submit->queue->ctx->seqno)
				break;
			fallthrough;
		case MSM_SUBMIT_CMD_BUF:
			OUT_PKT3(ring, CP_INDIRECT_BUFFER_PFD, 2);
			OUT_RING(ring, lower_32_bits(submit->cmd[i].iova));
			OUT_RING(ring, submit->cmd[i].size);
			OUT_PKT2(ring);
			break;
		}
	}

	OUT_PKT0(ring, REG_AXXX_CP_SCRATCH_REG2, 1);
	OUT_RING(ring, submit->seqno);

	/* Flush HLSQ lazy updates to make sure there is nothing
	 * pending for indirect loads after the timestamp has
	 * passed:
	 */
	OUT_PKT3(ring, CP_EVENT_WRITE, 1);
	OUT_RING(ring, HLSQ_FLUSH);

	/* wait for idle before cache flush/interrupt */
	OUT_PKT3(ring, CP_WAIT_FOR_IDLE, 1);
	OUT_RING(ring, 0x00000000);

	/* BIT(31) of CACHE_FLUSH_TS triggers CACHE_FLUSH_TS IRQ from GPU */
	OUT_PKT3(ring, CP_EVENT_WRITE, 3);
	OUT_RING(ring, CACHE_FLUSH_TS | CP_EVENT_WRITE_0_IRQ);
	OUT_RING(ring, rbmemptr(ring, fence));
	OUT_RING(ring, submit->seqno);

#if 0
	/* Dummy set-constant to trigger context rollover */
	OUT_PKT3(ring, CP_SET_CONSTANT, 2);
	OUT_RING(ring, CP_REG(REG_A3XX_HLSQ_CL_KERNEL_GROUP_X_REG));
	OUT_RING(ring, 0x00000000);
#endif

	adreno_flush(gpu, ring, REG_AXXX_CP_RB_WPTR);
}

static bool a3xx_me_init(struct msm_gpu *gpu)
{
	struct msm_ringbuffer *ring = gpu->rb[0];

	OUT_PKT3(ring, CP_ME_INIT, 17);
	OUT_RING(ring, 0x000003f7);
	OUT_RING(ring, 0x00000000);
	OUT_RING(ring, 0x00000000);
	OUT_RING(ring, 0x00000000);
	OUT_RING(ring, 0x00000080);
	OUT_RING(ring, 0x00000100);
	OUT_RING(ring, 0x00000180);
	OUT_RING(ring, 0x00006600);
	OUT_RING(ring, 0x00000150);
	OUT_RING(ring, 0x0000014e);
	OUT_RING(ring, 0x00000154);
	OUT_RING(ring, 0x00000001);
	OUT_RING(ring, 0x00000000);
	OUT_RING(ring, 0x00000000);
	OUT_RING(ring, 0x00000000);
	OUT_RING(ring, 0x00000000);
	OUT_RING(ring, 0x00000000);

	adreno_flush(gpu, ring, REG_AXXX_CP_RB_WPTR);
	return a3xx_idle(gpu);
}

static int a3xx_hw_init(struct msm_gpu *gpu)
{
	struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu);
	struct a3xx_gpu *a3xx_gpu = to_a3xx_gpu(adreno_gpu);
	uint32_t *ptr, len;
	int i, ret;

	DBG("%s", gpu->name);

	if (adreno_is_a305(adreno_gpu)) {
		/* Set up 16 deep read/write request queues: */
		gpu_write(gpu, REG_A3XX_VBIF_IN_RD_LIM_CONF0, 0x10101010);
		gpu_write(gpu, REG_A3XX_VBIF_IN_RD_LIM_CONF1, 0x10101010);
		gpu_write(gpu, REG_A3XX_VBIF_OUT_RD_LIM_CONF0, 0x10101010);
		gpu_write(gpu, REG_A3XX_VBIF_OUT_WR_LIM_CONF0, 0x10101010);
		gpu_write(gpu, REG_A3XX_VBIF_DDR_OUT_MAX_BURST, 0x0000303);
		gpu_write(gpu, REG_A3XX_VBIF_IN_WR_LIM_CONF0, 0x10101010);
		gpu_write(gpu, REG_A3XX_VBIF_IN_WR_LIM_CONF1, 0x10101010);
		/* Enable WR-REQ: */
		gpu_write(gpu, REG_A3XX_VBIF_GATE_OFF_WRREQ_EN, 0x0000ff);
		/* Set up round robin arbitration between both AXI ports: */
		gpu_write(gpu, REG_A3XX_VBIF_ARB_CTL, 0x00000030);
		/* Set up AOOO: */
		gpu_write(gpu, REG_A3XX_VBIF_OUT_AXI_AOOO_EN, 0x0000003c);
		gpu_write(gpu, REG_A3XX_VBIF_OUT_AXI_AOOO, 0x003c003c);
	} else if (adreno_is_a305b(adreno_gpu)) {
		gpu_write(gpu, REG_A3XX_VBIF_IN_RD_LIM_CONF0, 0x00181818);
		gpu_write(gpu, REG_A3XX_VBIF_IN_WR_LIM_CONF0, 0x00181818);
		gpu_write(gpu, REG_A3XX_VBIF_OUT_RD_LIM_CONF0, 0x00000018);
		gpu_write(gpu, REG_A3XX_VBIF_OUT_WR_LIM_CONF0, 0x00000018);
		gpu_write(gpu, REG_A3XX_VBIF_DDR_OUT_MAX_BURST, 0x00000303);
		gpu_write(gpu, REG_A3XX_VBIF_ROUND_ROBIN_QOS_ARB, 0x0003);
	} else if (adreno_is_a306(adreno_gpu)) {
		gpu_write(gpu, REG_A3XX_VBIF_ROUND_ROBIN_QOS_ARB, 0x0003);
		gpu_write(gpu, REG_A3XX_VBIF_OUT_RD_LIM_CONF0, 0x0000000a);
		gpu_write(gpu, REG_A3XX_VBIF_OUT_WR_LIM_CONF0, 0x0000000a);
	} else if (adreno_is_a320(adreno_gpu)) {
		/* Set up 16 deep read/write request queues: */
		gpu_write(gpu, REG_A3XX_VBIF_IN_RD_LIM_CONF0, 0x10101010);
		gpu_write(gpu, REG_A3XX_VBIF_IN_RD_LIM_CONF1, 0x10101010);
		gpu_write(gpu, REG_A3XX_VBIF_OUT_RD_LIM_CONF0, 0x10101010);
		gpu_write(gpu, REG_A3XX_VBIF_OUT_WR_LIM_CONF0, 0x10101010);
		gpu_write(gpu, REG_A3XX_VBIF_DDR_OUT_MAX_BURST, 0x0000303);
		gpu_write(gpu, REG_A3XX_VBIF_IN_WR_LIM_CONF0, 0x10101010);
		gpu_write(gpu, REG_A3XX_VBIF_IN_WR_LIM_CONF1, 0x10101010);
		/* Enable WR-REQ: */
		gpu_write(gpu, REG_A3XX_VBIF_GATE_OFF_WRREQ_EN, 0x0000ff);
		/* Set up round robin arbitration between both AXI ports: */
		gpu_write(gpu, REG_A3XX_VBIF_ARB_CTL, 0x00000030);
		/* Set up AOOO: */
		gpu_write(gpu, REG_A3XX_VBIF_OUT_AXI_AOOO_EN, 0x0000003c);
		gpu_write(gpu, REG_A3XX_VBIF_OUT_AXI_AOOO, 0x003c003c);
		/* Enable 1K sort: */
		gpu_write(gpu, REG_A3XX_VBIF_ABIT_SORT, 0x000000ff);
		gpu_write(gpu, REG_A3XX_VBIF_ABIT_SORT_CONF, 0x000000a4);

	} else if (adreno_is_a330v2(adreno_gpu)) {
		/*
		 * Most of the VBIF registers on 8974v2 have the correct
		 * values at power on, so we won't modify those if we don't
		 * need to
		 */
		/* Enable 1k sort: */
		gpu_write(gpu, REG_A3XX_VBIF_ABIT_SORT, 0x0001003f);
		gpu_write(gpu, REG_A3XX_VBIF_ABIT_SORT_CONF, 0x000000a4);
		/* Enable WR-REQ: */
		gpu_write(gpu, REG_A3XX_VBIF_GATE_OFF_WRREQ_EN, 0x00003f);
		gpu_write(gpu, REG_A3XX_VBIF_DDR_OUT_MAX_BURST, 0x0000303);
		/* Set up VBIF_ROUND_ROBIN_QOS_ARB: */
		gpu_write(gpu, REG_A3XX_VBIF_ROUND_ROBIN_QOS_ARB, 0x0003);

	} else if (adreno_is_a330(adreno_gpu)) {
		/* Set up 16 deep read/write request queues: */
		gpu_write(gpu, REG_A3XX_VBIF_IN_RD_LIM_CONF0, 0x18181818);
		gpu_write(gpu, REG_A3XX_VBIF_IN_RD_LIM_CONF1, 0x18181818);
		gpu_write(gpu, REG_A3XX_VBIF_OUT_RD_LIM_CONF0, 0x18181818);
		gpu_write(gpu, REG_A3XX_VBIF_OUT_WR_LIM_CONF0, 0x18181818);
		gpu_write(gpu, REG_A3XX_VBIF_DDR_OUT_MAX_BURST, 0x0000303);
		gpu_write(gpu, REG_A3XX_VBIF_IN_WR_LIM_CONF0, 0x18181818);
		gpu_write(gpu, REG_A3XX_VBIF_IN_WR_LIM_CONF1, 0x18181818);
		/* Enable WR-REQ: */
		gpu_write(gpu, REG_A3XX_VBIF_GATE_OFF_WRREQ_EN, 0x00003f);
		/* Set up round robin arbitration between both AXI ports: */
		gpu_write(gpu, REG_A3XX_VBIF_ARB_CTL, 0x00000030);
		/* Set up VBIF_ROUND_ROBIN_QOS_ARB: */
		gpu_write(gpu, REG_A3XX_VBIF_ROUND_ROBIN_QOS_ARB, 0x0001);
		/* Set up AOOO: */
		gpu_write(gpu, REG_A3XX_VBIF_OUT_AXI_AOOO_EN, 0x0000003f);
		gpu_write(gpu, REG_A3XX_VBIF_OUT_AXI_AOOO, 0x003f003f);
		/* Enable 1K sort: */
		gpu_write(gpu, REG_A3XX_VBIF_ABIT_SORT, 0x0001003f);
		gpu_write(gpu, REG_A3XX_VBIF_ABIT_SORT_CONF, 0x000000a4);
		/* Disable VBIF clock gating. This is to enable AXI running
		 * higher frequency than GPU:
		 */
		gpu_write(gpu, REG_A3XX_VBIF_CLKON, 0x00000001);

	} else {
		BUG();
	}

	/* Make all blocks contribute to the GPU BUSY perf counter: */
	gpu_write(gpu, REG_A3XX_RBBM_GPU_BUSY_MASKED, 0xffffffff);

	/* Tune the hystersis counters for SP and CP idle detection: */
	gpu_write(gpu, REG_A3XX_RBBM_SP_HYST_CNT, 0x10);
	gpu_write(gpu, REG_A3XX_RBBM_WAIT_IDLE_CLOCKS_CTL, 0x10);

	/* Enable the RBBM error reporting bits.  This lets us get
	 * useful information on failure:
	 */
	gpu_write(gpu, REG_A3XX_RBBM_AHB_CTL0, 0x00000001);

	/* Enable AHB error reporting: */
	gpu_write(gpu, REG_A3XX_RBBM_AHB_CTL1, 0xa6ffffff);

	/* Turn on the power counters: */
	gpu_write(gpu, REG_A3XX_RBBM_RBBM_CTL, 0x00030000);

	/* Turn on hang detection - this spews a lot of useful information
	 * into the RBBM registers on a hang:
	 */
	gpu_write(gpu, REG_A3XX_RBBM_INTERFACE_HANG_INT_CTL, 0x00010fff);

	/* Enable 64-byte cacheline size. HW Default is 32-byte (0x000000E0): */
	gpu_write(gpu, REG_A3XX_UCHE_CACHE_MODE_CONTROL_REG, 0x00000001);

	/* Enable Clock gating: */
	if (adreno_is_a305b(adreno_gpu) || adreno_is_a306(adreno_gpu))
		gpu_write(gpu, REG_A3XX_RBBM_CLOCK_CTL, 0xaaaaaaaa);
	else if (adreno_is_a320(adreno_gpu))
		gpu_write(gpu, REG_A3XX_RBBM_CLOCK_CTL, 0xbfffffff);
	else if (adreno_is_a330v2(adreno_gpu))
		gpu_write(gpu, REG_A3XX_RBBM_CLOCK_CTL, 0xaaaaaaaa);
	else if (adreno_is_a330(adreno_gpu))
		gpu_write(gpu, REG_A3XX_RBBM_CLOCK_CTL, 0xbffcffff);

	if (adreno_is_a330v2(adreno_gpu))
		gpu_write(gpu, REG_A3XX_RBBM_GPR0_CTL, 0x05515455);
	else if (adreno_is_a330(adreno_gpu))
		gpu_write(gpu, REG_A3XX_RBBM_GPR0_CTL, 0x00000000);

	/* Set the OCMEM base address for A330, etc */
	if (a3xx_gpu->ocmem.hdl) {
		gpu_write(gpu, REG_A3XX_RB_GMEM_BASE_ADDR,
			(unsigned int)(a3xx_gpu->ocmem.base >> 14));
	}

	/* Turn on performance counters: */
	gpu_write(gpu, REG_A3XX_RBBM_PERFCTR_CTL, 0x01);

	/* Enable the perfcntrs that we use.. */
	for (i = 0; i < gpu->num_perfcntrs; i++) {
		const struct msm_gpu_perfcntr *perfcntr = &gpu->perfcntrs[i];
		gpu_write(gpu, perfcntr->select_reg, perfcntr->select_val);
	}

	gpu_write(gpu, REG_A3XX_RBBM_INT_0_MASK, A3XX_INT0_MASK);

	ret = adreno_hw_init(gpu);
	if (ret)
		return ret;

	/*
	 * Use the default ringbuffer size and block size but disable the RPTR
	 * shadow
	 */
	gpu_write(gpu, REG_AXXX_CP_RB_CNTL,
		MSM_GPU_RB_CNTL_DEFAULT | AXXX_CP_RB_CNTL_NO_UPDATE);

	/* Set the ringbuffer address */
	gpu_write(gpu, REG_AXXX_CP_RB_BASE, lower_32_bits(gpu->rb[0]->iova));

	/* setup access protection: */
	gpu_write(gpu, REG_A3XX_CP_PROTECT_CTRL, 0x00000007);

	/* RBBM registers */
	gpu_write(gpu, REG_A3XX_CP_PROTECT(0), 0x63000040);
	gpu_write(gpu, REG_A3XX_CP_PROTECT(1), 0x62000080);
	gpu_write(gpu, REG_A3XX_CP_PROTECT(2), 0x600000cc);
	gpu_write(gpu, REG_A3XX_CP_PROTECT(3), 0x60000108);
	gpu_write(gpu, REG_A3XX_CP_PROTECT(4), 0x64000140);
	gpu_write(gpu, REG_A3XX_CP_PROTECT(5), 0x66000400);

	/* CP registers */
	gpu_write(gpu, REG_A3XX_CP_PROTECT(6), 0x65000700);
	gpu_write(gpu, REG_A3XX_CP_PROTECT(7), 0x610007d8);
	gpu_write(gpu, REG_A3XX_CP_PROTECT(8), 0x620007e0);
	gpu_write(gpu, REG_A3XX_CP_PROTECT(9), 0x61001178);
	gpu_write(gpu, REG_A3XX_CP_PROTECT(10), 0x64001180);

	/* RB registers */
	gpu_write(gpu, REG_A3XX_CP_PROTECT(11), 0x60003300);

	/* VBIF registers */
	gpu_write(gpu, REG_A3XX_CP_PROTECT(12), 0x6b00c000);

	/* NOTE: PM4/micro-engine firmware registers look to be the same
	 * for a2xx and a3xx.. we could possibly push that part down to
	 * adreno_gpu base class.  Or push both PM4 and PFP but
	 * parameterize the pfp ucode addr/data registers..
	 */

	/* Load PM4: */
	ptr = (uint32_t *)(adreno_gpu->fw[ADRENO_FW_PM4]->data);
	len = adreno_gpu->fw[ADRENO_FW_PM4]->size / 4;
	DBG("loading PM4 ucode version: %x", ptr[1]);

	gpu_write(gpu, REG_AXXX_CP_DEBUG,
			AXXX_CP_DEBUG_DYNAMIC_CLK_DISABLE |
			AXXX_CP_DEBUG_MIU_128BIT_WRITE_ENABLE);
	gpu_write(gpu, REG_AXXX_CP_ME_RAM_WADDR, 0);
	for (i = 1; i < len; i++)
		gpu_write(gpu, REG_AXXX_CP_ME_RAM_DATA, ptr[i]);

	/* Load PFP: */
	ptr = (uint32_t *)(adreno_gpu->fw[ADRENO_FW_PFP]->data);
	len = adreno_gpu->fw[ADRENO_FW_PFP]->size / 4;
	DBG("loading PFP ucode version: %x", ptr[5]);

	gpu_write(gpu, REG_A3XX_CP_PFP_UCODE_ADDR, 0);
	for (i = 1; i < len; i++)
		gpu_write(gpu, REG_A3XX_CP_PFP_UCODE_DATA, ptr[i]);

	/* CP ROQ queue sizes (bytes) - RB:16, ST:16, IB1:32, IB2:64 */
	if (adreno_is_a305(adreno_gpu) || adreno_is_a306(adreno_gpu) ||
			adreno_is_a320(adreno_gpu)) {
		gpu_write(gpu, REG_AXXX_CP_QUEUE_THRESHOLDS,
				AXXX_CP_QUEUE_THRESHOLDS_CSQ_IB1_START(2) |
				AXXX_CP_QUEUE_THRESHOLDS_CSQ_IB2_START(6) |
				AXXX_CP_QUEUE_THRESHOLDS_CSQ_ST_START(14));
	} else if (adreno_is_a330(adreno_gpu) || adreno_is_a305b(adreno_gpu)) {
		/* NOTE: this (value take from downstream android driver)
		 * includes some bits outside of the known bitfields.  But
		 * A330 has this "MERCIU queue" thing too, which might
		 * explain a new bitfield or reshuffling:
		 */
		gpu_write(gpu, REG_AXXX_CP_QUEUE_THRESHOLDS, 0x003e2008);
	}

	/* clear ME_HALT to start micro engine */
	gpu_write(gpu, REG_AXXX_CP_ME_CNTL, 0);

	return a3xx_me_init(gpu) ? 0 : -EINVAL;
}

static void a3xx_recover(struct msm_gpu *gpu)
{
	int i;

	adreno_dump_info(gpu);

	for (i = 0; i < 8; i++) {
		printk("CP_SCRATCH_REG%d: %u\n", i,
			gpu_read(gpu, REG_AXXX_CP_SCRATCH_REG0 + i));
	}

	/* dump registers before resetting gpu, if enabled: */
	if (hang_debug)
		a3xx_dump(gpu);

	gpu_write(gpu, REG_A3XX_RBBM_SW_RESET_CMD, 1);
	gpu_read(gpu, REG_A3XX_RBBM_SW_RESET_CMD);
	gpu_write(gpu, REG_A3XX_RBBM_SW_RESET_CMD, 0);
	adreno_recover(gpu);
}

static void a3xx_destroy(struct msm_gpu *gpu)
{
	struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu);
	struct a3xx_gpu *a3xx_gpu = to_a3xx_gpu(adreno_gpu);

	DBG("%s", gpu->name);

	adreno_gpu_cleanup(adreno_gpu);

	adreno_gpu_ocmem_cleanup(&a3xx_gpu->ocmem);

	kfree(a3xx_gpu);
}

static bool a3xx_idle(struct msm_gpu *gpu)
{
	/* wait for ringbuffer to drain: */
	if (!adreno_idle(gpu, gpu->rb[0]))
		return false;

	/* then wait for GPU to finish: */
	if (spin_until(!(gpu_read(gpu, REG_A3XX_RBBM_STATUS) &
			A3XX_RBBM_STATUS_GPU_BUSY))) {
		DRM_ERROR("%s: timeout waiting for GPU to idle!\n", gpu->name);

		/* TODO maybe we need to reset GPU here to recover from hang? */
		return false;
	}

	return true;
}

static irqreturn_t a3xx_irq(struct msm_gpu *gpu)
{
	uint32_t status;

	status = gpu_read(gpu, REG_A3XX_RBBM_INT_0_STATUS);
	DBG("%s: %08x", gpu->name, status);

	// TODO

	gpu_write(gpu, REG_A3XX_RBBM_INT_CLEAR_CMD, status);

	msm_gpu_retire(gpu);

	return IRQ_HANDLED;
}

static const unsigned int a3xx_registers[] = {
	0x0000, 0x0002, 0x0010, 0x0012, 0x0018, 0x0018, 0x0020, 0x0027,
	0x0029, 0x002b, 0x002e, 0x0033, 0x0040, 0x0042, 0x0050, 0x005c,
	0x0060, 0x006c, 0x0080, 0x0082, 0x0084, 0x0088, 0x0090, 0x00e5,
	0x00ea, 0x00ed, 0x0100, 0x0100, 0x0110, 0x0123, 0x01c0, 0x01c1,
	0x01c3, 0x01c5, 0x01c7, 0x01c7, 0x01d5, 0x01d9, 0x01dc, 0x01dd,
	0x01ea, 0x01ea, 0x01ee, 0x01f1, 0x01f5, 0x01f5, 0x01fc, 0x01ff,
	0x0440, 0x0440, 0x0443, 0x0443, 0x0445, 0x0445, 0x044d, 0x044f,
	0x0452, 0x0452, 0x0454, 0x046f, 0x047c, 0x047c, 0x047f, 0x047f,
	0x0578, 0x057f, 0x0600, 0x0602, 0x0605, 0x0607, 0x060a, 0x060e,
	0x0612, 0x0614, 0x0c01, 0x0c02, 0x0c06, 0x0c1d, 0x0c3d, 0x0c3f,
	0x0c48, 0x0c4b, 0x0c80, 0x0c80, 0x0c88, 0x0c8b, 0x0ca0, 0x0cb7,
	0x0cc0, 0x0cc1, 0x0cc6, 0x0cc7, 0x0ce4, 0x0ce5, 0x0e00, 0x0e05,
	0x0e0c, 0x0e0c, 0x0e22, 0x0e23, 0x0e41, 0x0e45, 0x0e64, 0x0e65,
	0x0e80, 0x0e82, 0x0e84, 0x0e89, 0x0ea0, 0x0ea1, 0x0ea4, 0x0ea7,
	0x0ec4, 0x0ecb, 0x0ee0, 0x0ee0, 0x0f00, 0x0f01, 0x0f03, 0x0f09,
	0x2040, 0x2040, 0x2044, 0x2044, 0x2048, 0x204d, 0x2068, 0x2069,
	0x206c, 0x206d, 0x2070, 0x2070, 0x2072, 0x2072, 0x2074, 0x2075,
	0x2079, 0x207a, 0x20c0, 0x20d3, 0x20e4, 0x20ef, 0x2100, 0x2109,
	0x210c, 0x210c, 0x210e, 0x210e, 0x2110, 0x2111, 0x2114, 0x2115,
	0x21e4, 0x21e4, 0x21ea, 0x21ea, 0x21ec, 0x21ed, 0x21f0, 0x21f0,
	0x2200, 0x2212, 0x2214, 0x2217, 0x221a, 0x221a, 0x2240, 0x227e,
	0x2280, 0x228b, 0x22c0, 0x22c0, 0x22c4, 0x22ce, 0x22d0, 0x22d8,
	0x22df, 0x22e6, 0x22e8, 0x22e9, 0x22ec, 0x22ec, 0x22f0, 0x22f7,
	0x22ff, 0x22ff, 0x2340, 0x2343, 0x2440, 0x2440, 0x2444, 0x2444,
	0x2448, 0x244d, 0x2468, 0x2469, 0x246c, 0x246d, 0x2470, 0x2470,
	0x2472, 0x2472, 0x2474, 0x2475, 0x2479, 0x247a, 0x24c0, 0x24d3,
	0x24e4, 0x24ef, 0x2500, 0x2509, 0x250c, 0x250c, 0x250e, 0x250e,
	0x2510, 0x2511, 0x2514, 0x2515, 0x25e4, 0x25e4, 0x25ea, 0x25ea,
	0x25ec, 0x25ed, 0x25f0, 0x25f0, 0x2600, 0x2612, 0x2614, 0x2617,
	0x261a, 0x261a, 0x2640, 0x267e, 0x2680, 0x268b, 0x26c0, 0x26c0,
	0x26c4, 0x26ce, 0x26d0, 0x26d8, 0x26df, 0x26e6, 0x26e8, 0x26e9,
	0x26ec, 0x26ec, 0x26f0, 0x26f7, 0x26ff, 0x26ff, 0x2740, 0x2743,
	0x300c, 0x300e, 0x301c, 0x301d, 0x302a, 0x302a, 0x302c, 0x302d,
	0x3030, 0x3031, 0x3034, 0x3036, 0x303c, 0x303c, 0x305e, 0x305f,
	~0   /* sentinel */
};

/* would be nice to not have to duplicate the _show() stuff with printk(): */
static void a3xx_dump(struct msm_gpu *gpu)
{
	printk("status:   %08x\n",
			gpu_read(gpu, REG_A3XX_RBBM_STATUS));
	adreno_dump(gpu);
}

static struct msm_gpu_state *a3xx_gpu_state_get(struct msm_gpu *gpu)
{
	struct msm_gpu_state *state = kzalloc(sizeof(*state), GFP_KERNEL);

	if (!state)
		return ERR_PTR(-ENOMEM);

	adreno_gpu_state_get(gpu, state);

	state->rbbm_status = gpu_read(gpu, REG_A3XX_RBBM_STATUS);

	return state;
}

static u64 a3xx_gpu_busy(struct msm_gpu *gpu, unsigned long *out_sample_rate)
{
	u64 busy_cycles;

	busy_cycles = gpu_read64(gpu, REG_A3XX_RBBM_PERFCTR_RBBM_1_LO);
	*out_sample_rate = clk_get_rate(gpu->core_clk);

	return busy_cycles;
}

static u32 a3xx_get_rptr(struct msm_gpu *gpu, struct msm_ringbuffer *ring)
{
	ring->memptrs->rptr = gpu_read(gpu, REG_AXXX_CP_RB_RPTR);
	return ring->memptrs->rptr;
}

static const struct adreno_gpu_funcs funcs = {
	.base = {
		.get_param = adreno_get_param,
		.set_param = adreno_set_param,
		.hw_init = a3xx_hw_init,
		.pm_suspend = msm_gpu_pm_suspend,
		.pm_resume = msm_gpu_pm_resume,
		.recover = a3xx_recover,
		.submit = a3xx_submit,
		.active_ring = adreno_active_ring,
		.irq = a3xx_irq,
		.destroy = a3xx_destroy,
#if defined(CONFIG_DEBUG_FS) || defined(CONFIG_DEV_COREDUMP)
		.show = adreno_show,
#endif
		.gpu_busy = a3xx_gpu_busy,
		.gpu_state_get = a3xx_gpu_state_get,
		.gpu_state_put = adreno_gpu_state_put,
		.create_address_space = adreno_create_address_space,
		.get_rptr = a3xx_get_rptr,
	},
};

static const struct msm_gpu_perfcntr perfcntrs[] = {
	{ REG_A3XX_SP_PERFCOUNTER6_SELECT, REG_A3XX_RBBM_PERFCTR_SP_6_LO,
			SP_ALU_ACTIVE_CYCLES, "ALUACTIVE" },
	{ REG_A3XX_SP_PERFCOUNTER7_SELECT, REG_A3XX_RBBM_PERFCTR_SP_7_LO,
			SP_FS_FULL_ALU_INSTRUCTIONS, "ALUFULL" },
};

struct msm_gpu *a3xx_gpu_init(struct drm_device *dev)
{
	struct a3xx_gpu *a3xx_gpu = NULL;
	struct adreno_gpu *adreno_gpu;
	struct msm_gpu *gpu;
	struct msm_drm_private *priv = dev->dev_private;
	struct platform_device *pdev = priv->gpu_pdev;
	struct icc_path *ocmem_icc_path;
	struct icc_path *icc_path;
	int ret;

	if (!pdev) {
		DRM_DEV_ERROR(dev->dev, "no a3xx device\n");
		ret = -ENXIO;
		goto fail;
	}

	a3xx_gpu = kzalloc(sizeof(*a3xx_gpu), GFP_KERNEL);
	if (!a3xx_gpu) {
		ret = -ENOMEM;
		goto fail;
	}

	adreno_gpu = &a3xx_gpu->base;
	gpu = &adreno_gpu->base;

	gpu->perfcntrs = perfcntrs;
	gpu->num_perfcntrs = ARRAY_SIZE(perfcntrs);

	adreno_gpu->registers = a3xx_registers;

	ret = adreno_gpu_init(dev, pdev, adreno_gpu, &funcs, 1);
	if (ret)
		goto fail;

	/* if needed, allocate gmem: */
	if (adreno_is_a330(adreno_gpu) || adreno_is_a305b(adreno_gpu)) {
		ret = adreno_gpu_ocmem_init(&adreno_gpu->base.pdev->dev,
					    adreno_gpu, &a3xx_gpu->ocmem);
		if (ret)
			goto fail;
	}

	if (!gpu->aspace) {
		/* TODO we think it is possible to configure the GPU to
		 * restrict access to VRAM carveout.  But the required
		 * registers are unknown.  For now just bail out and
		 * limp along with just modesetting.  If it turns out
		 * to not be possible to restrict access, then we must
		 * implement a cmdstream validator.
		 */
		DRM_DEV_ERROR(dev->dev, "No memory protection without IOMMU\n");
		if (!allow_vram_carveout) {
			ret = -ENXIO;
			goto fail;
		}
	}

	icc_path = devm_of_icc_get(&pdev->dev, "gfx-mem");
	if (IS_ERR(icc_path)) {
		ret = PTR_ERR(icc_path);
		goto fail;
	}

	ocmem_icc_path = devm_of_icc_get(&pdev->dev, "ocmem");
	if (IS_ERR(ocmem_icc_path)) {
		ret = PTR_ERR(ocmem_icc_path);
		/* allow -ENODATA, ocmem icc is optional */
		if (ret != -ENODATA)
			goto fail;
		ocmem_icc_path = NULL;
	}


	/*
	 * Set the ICC path to maximum speed for now by multiplying the fastest
	 * frequency by the bus width (8). We'll want to scale this later on to
	 * improve battery life.
	 */
	icc_set_bw(icc_path, 0, Bps_to_icc(gpu->fast_rate) * 8);
	icc_set_bw(ocmem_icc_path, 0, Bps_to_icc(gpu->fast_rate) * 8);

	return gpu;

fail:
	if (a3xx_gpu)
		a3xx_destroy(&a3xx_gpu->base.base);

	return ERR_PTR(ret);
}