summaryrefslogtreecommitdiff
path: root/drivers/gpu/drm/i915/gvt/execlist.c
blob: f21b8fb5b37e11a4cfb071d3adff9fe19245422f (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
/*
 * Copyright(c) 2011-2016 Intel Corporation. All rights reserved.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice (including the next
 * paragraph) shall be included in all copies or substantial portions of the
 * Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 *
 * Authors:
 *    Zhiyuan Lv <zhiyuan.lv@intel.com>
 *    Zhi Wang <zhi.a.wang@intel.com>
 *
 * Contributors:
 *    Min He <min.he@intel.com>
 *    Bing Niu <bing.niu@intel.com>
 *    Ping Gao <ping.a.gao@intel.com>
 *    Tina Zhang <tina.zhang@intel.com>
 *
 */

#include "i915_drv.h"
#include "gvt.h"

#define _EL_OFFSET_STATUS       0x234
#define _EL_OFFSET_STATUS_BUF   0x370
#define _EL_OFFSET_STATUS_PTR   0x3A0

#define execlist_ring_mmio(gvt, ring_id, offset) \
	(gvt->dev_priv->engine[ring_id]->mmio_base + (offset))

#define valid_context(ctx) ((ctx)->valid)
#define same_context(a, b) (((a)->context_id == (b)->context_id) && \
		((a)->lrca == (b)->lrca))

static int context_switch_events[] = {
	[RCS0]  = RCS_AS_CONTEXT_SWITCH,
	[BCS0]  = BCS_AS_CONTEXT_SWITCH,
	[VCS0]  = VCS_AS_CONTEXT_SWITCH,
	[VCS1]  = VCS2_AS_CONTEXT_SWITCH,
	[VECS0] = VECS_AS_CONTEXT_SWITCH,
};

static int ring_id_to_context_switch_event(unsigned int ring_id)
{
	if (WARN_ON(ring_id >= ARRAY_SIZE(context_switch_events)))
		return -EINVAL;

	return context_switch_events[ring_id];
}

static void switch_virtual_execlist_slot(struct intel_vgpu_execlist *execlist)
{
	gvt_dbg_el("[before] running slot %d/context %x pending slot %d\n",
			execlist->running_slot ?
			execlist->running_slot->index : -1,
			execlist->running_context ?
			execlist->running_context->context_id : 0,
			execlist->pending_slot ?
			execlist->pending_slot->index : -1);

	execlist->running_slot = execlist->pending_slot;
	execlist->pending_slot = NULL;
	execlist->running_context = execlist->running_context ?
		&execlist->running_slot->ctx[0] : NULL;

	gvt_dbg_el("[after] running slot %d/context %x pending slot %d\n",
			execlist->running_slot ?
			execlist->running_slot->index : -1,
			execlist->running_context ?
			execlist->running_context->context_id : 0,
			execlist->pending_slot ?
			execlist->pending_slot->index : -1);
}

static void emulate_execlist_status(struct intel_vgpu_execlist *execlist)
{
	struct intel_vgpu_execlist_slot *running = execlist->running_slot;
	struct intel_vgpu_execlist_slot *pending = execlist->pending_slot;
	struct execlist_ctx_descriptor_format *desc = execlist->running_context;
	struct intel_vgpu *vgpu = execlist->vgpu;
	struct execlist_status_format status;
	int ring_id = execlist->ring_id;
	u32 status_reg = execlist_ring_mmio(vgpu->gvt,
			ring_id, _EL_OFFSET_STATUS);

	status.ldw = vgpu_vreg(vgpu, status_reg);
	status.udw = vgpu_vreg(vgpu, status_reg + 4);

	if (running) {
		status.current_execlist_pointer = !!running->index;
		status.execlist_write_pointer = !!!running->index;
		status.execlist_0_active = status.execlist_0_valid =
			!!!(running->index);
		status.execlist_1_active = status.execlist_1_valid =
			!!(running->index);
	} else {
		status.context_id = 0;
		status.execlist_0_active = status.execlist_0_valid = 0;
		status.execlist_1_active = status.execlist_1_valid = 0;
	}

	status.context_id = desc ? desc->context_id : 0;
	status.execlist_queue_full = !!(pending);

	vgpu_vreg(vgpu, status_reg) = status.ldw;
	vgpu_vreg(vgpu, status_reg + 4) = status.udw;

	gvt_dbg_el("vgpu%d: status reg offset %x ldw %x udw %x\n",
		vgpu->id, status_reg, status.ldw, status.udw);
}

static void emulate_csb_update(struct intel_vgpu_execlist *execlist,
		struct execlist_context_status_format *status,
		bool trigger_interrupt_later)
{
	struct intel_vgpu *vgpu = execlist->vgpu;
	int ring_id = execlist->ring_id;
	struct execlist_context_status_pointer_format ctx_status_ptr;
	u32 write_pointer;
	u32 ctx_status_ptr_reg, ctx_status_buf_reg, offset;
	unsigned long hwsp_gpa;
	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;

	ctx_status_ptr_reg = execlist_ring_mmio(vgpu->gvt, ring_id,
			_EL_OFFSET_STATUS_PTR);
	ctx_status_buf_reg = execlist_ring_mmio(vgpu->gvt, ring_id,
			_EL_OFFSET_STATUS_BUF);

	ctx_status_ptr.dw = vgpu_vreg(vgpu, ctx_status_ptr_reg);

	write_pointer = ctx_status_ptr.write_ptr;

	if (write_pointer == 0x7)
		write_pointer = 0;
	else {
		++write_pointer;
		write_pointer %= 0x6;
	}

	offset = ctx_status_buf_reg + write_pointer * 8;

	vgpu_vreg(vgpu, offset) = status->ldw;
	vgpu_vreg(vgpu, offset + 4) = status->udw;

	ctx_status_ptr.write_ptr = write_pointer;
	vgpu_vreg(vgpu, ctx_status_ptr_reg) = ctx_status_ptr.dw;

	/* Update the CSB and CSB write pointer in HWSP */
	hwsp_gpa = intel_vgpu_gma_to_gpa(vgpu->gtt.ggtt_mm,
					 vgpu->hws_pga[ring_id]);
	if (hwsp_gpa != INTEL_GVT_INVALID_ADDR) {
		intel_gvt_hypervisor_write_gpa(vgpu,
			hwsp_gpa + I915_HWS_CSB_BUF0_INDEX * 4 +
			write_pointer * 8,
			status, 8);
		intel_gvt_hypervisor_write_gpa(vgpu,
			hwsp_gpa +
			intel_hws_csb_write_index(dev_priv) * 4,
			&write_pointer, 4);
	}

	gvt_dbg_el("vgpu%d: w pointer %u reg %x csb l %x csb h %x\n",
		vgpu->id, write_pointer, offset, status->ldw, status->udw);

	if (trigger_interrupt_later)
		return;

	intel_vgpu_trigger_virtual_event(vgpu,
			ring_id_to_context_switch_event(execlist->ring_id));
}

static int emulate_execlist_ctx_schedule_out(
		struct intel_vgpu_execlist *execlist,
		struct execlist_ctx_descriptor_format *ctx)
{
	struct intel_vgpu *vgpu = execlist->vgpu;
	struct intel_vgpu_execlist_slot *running = execlist->running_slot;
	struct intel_vgpu_execlist_slot *pending = execlist->pending_slot;
	struct execlist_ctx_descriptor_format *ctx0 = &running->ctx[0];
	struct execlist_ctx_descriptor_format *ctx1 = &running->ctx[1];
	struct execlist_context_status_format status;

	memset(&status, 0, sizeof(status));

	gvt_dbg_el("schedule out context id %x\n", ctx->context_id);

	if (WARN_ON(!same_context(ctx, execlist->running_context))) {
		gvt_vgpu_err("schedule out context is not running context,"
				"ctx id %x running ctx id %x\n",
				ctx->context_id,
				execlist->running_context->context_id);
		return -EINVAL;
	}

	/* ctx1 is valid, ctx0/ctx is scheduled-out -> element switch */
	if (valid_context(ctx1) && same_context(ctx0, ctx)) {
		gvt_dbg_el("ctx 1 valid, ctx/ctx 0 is scheduled-out\n");

		execlist->running_context = ctx1;

		emulate_execlist_status(execlist);

		status.context_complete = status.element_switch = 1;
		status.context_id = ctx->context_id;

		emulate_csb_update(execlist, &status, false);
		/*
		 * ctx1 is not valid, ctx == ctx0
		 * ctx1 is valid, ctx1 == ctx
		 *	--> last element is finished
		 * emulate:
		 *	active-to-idle if there is *no* pending execlist
		 *	context-complete if there *is* pending execlist
		 */
	} else if ((!valid_context(ctx1) && same_context(ctx0, ctx))
			|| (valid_context(ctx1) && same_context(ctx1, ctx))) {
		gvt_dbg_el("need to switch virtual execlist slot\n");

		switch_virtual_execlist_slot(execlist);

		emulate_execlist_status(execlist);

		status.context_complete = status.active_to_idle = 1;
		status.context_id = ctx->context_id;

		if (!pending) {
			emulate_csb_update(execlist, &status, false);
		} else {
			emulate_csb_update(execlist, &status, true);

			memset(&status, 0, sizeof(status));

			status.idle_to_active = 1;
			status.context_id = 0;

			emulate_csb_update(execlist, &status, false);
		}
	} else {
		WARN_ON(1);
		return -EINVAL;
	}

	return 0;
}

static struct intel_vgpu_execlist_slot *get_next_execlist_slot(
		struct intel_vgpu_execlist *execlist)
{
	struct intel_vgpu *vgpu = execlist->vgpu;
	int ring_id = execlist->ring_id;
	u32 status_reg = execlist_ring_mmio(vgpu->gvt, ring_id,
			_EL_OFFSET_STATUS);
	struct execlist_status_format status;

	status.ldw = vgpu_vreg(vgpu, status_reg);
	status.udw = vgpu_vreg(vgpu, status_reg + 4);

	if (status.execlist_queue_full) {
		gvt_vgpu_err("virtual execlist slots are full\n");
		return NULL;
	}

	return &execlist->slot[status.execlist_write_pointer];
}

static int emulate_execlist_schedule_in(struct intel_vgpu_execlist *execlist,
		struct execlist_ctx_descriptor_format ctx[2])
{
	struct intel_vgpu_execlist_slot *running = execlist->running_slot;
	struct intel_vgpu_execlist_slot *slot =
		get_next_execlist_slot(execlist);

	struct execlist_ctx_descriptor_format *ctx0, *ctx1;
	struct execlist_context_status_format status;
	struct intel_vgpu *vgpu = execlist->vgpu;

	gvt_dbg_el("emulate schedule-in\n");

	if (!slot) {
		gvt_vgpu_err("no available execlist slot\n");
		return -EINVAL;
	}

	memset(&status, 0, sizeof(status));
	memset(slot->ctx, 0, sizeof(slot->ctx));

	slot->ctx[0] = ctx[0];
	slot->ctx[1] = ctx[1];

	gvt_dbg_el("alloc slot index %d ctx 0 %x ctx 1 %x\n",
			slot->index, ctx[0].context_id,
			ctx[1].context_id);

	/*
	 * no running execlist, make this write bundle as running execlist
	 * -> idle-to-active
	 */
	if (!running) {
		gvt_dbg_el("no current running execlist\n");

		execlist->running_slot = slot;
		execlist->pending_slot = NULL;
		execlist->running_context = &slot->ctx[0];

		gvt_dbg_el("running slot index %d running context %x\n",
				execlist->running_slot->index,
				execlist->running_context->context_id);

		emulate_execlist_status(execlist);

		status.idle_to_active = 1;
		status.context_id = 0;

		emulate_csb_update(execlist, &status, false);
		return 0;
	}

	ctx0 = &running->ctx[0];
	ctx1 = &running->ctx[1];

	gvt_dbg_el("current running slot index %d ctx 0 %x ctx 1 %x\n",
		running->index, ctx0->context_id, ctx1->context_id);

	/*
	 * already has an running execlist
	 *	a. running ctx1 is valid,
	 *	   ctx0 is finished, and running ctx1 == new execlist ctx[0]
	 *	b. running ctx1 is not valid,
	 *	   ctx0 == new execlist ctx[0]
	 * ----> lite-restore + preempted
	 */
	if ((valid_context(ctx1) && same_context(ctx1, &slot->ctx[0]) &&
		/* condition a */
		(!same_context(ctx0, execlist->running_context))) ||
			(!valid_context(ctx1) &&
			 same_context(ctx0, &slot->ctx[0]))) { /* condition b */
		gvt_dbg_el("need to switch virtual execlist slot\n");

		execlist->pending_slot = slot;
		switch_virtual_execlist_slot(execlist);

		emulate_execlist_status(execlist);

		status.lite_restore = status.preempted = 1;
		status.context_id = ctx[0].context_id;

		emulate_csb_update(execlist, &status, false);
	} else {
		gvt_dbg_el("emulate as pending slot\n");
		/*
		 * otherwise
		 * --> emulate pending execlist exist + but no preemption case
		 */
		execlist->pending_slot = slot;
		emulate_execlist_status(execlist);
	}
	return 0;
}

#define get_desc_from_elsp_dwords(ed, i) \
	((struct execlist_ctx_descriptor_format *)&((ed)->data[i * 2]))

static int prepare_execlist_workload(struct intel_vgpu_workload *workload)
{
	struct intel_vgpu *vgpu = workload->vgpu;
	struct intel_vgpu_submission *s = &vgpu->submission;
	struct execlist_ctx_descriptor_format ctx[2];
	int ring_id = workload->ring_id;
	int ret;

	if (!workload->emulate_schedule_in)
		return 0;

	ctx[0] = *get_desc_from_elsp_dwords(&workload->elsp_dwords, 0);
	ctx[1] = *get_desc_from_elsp_dwords(&workload->elsp_dwords, 1);

	ret = emulate_execlist_schedule_in(&s->execlist[ring_id], ctx);
	if (ret) {
		gvt_vgpu_err("fail to emulate execlist schedule in\n");
		return ret;
	}
	return 0;
}

static int complete_execlist_workload(struct intel_vgpu_workload *workload)
{
	struct intel_vgpu *vgpu = workload->vgpu;
	int ring_id = workload->ring_id;
	struct intel_vgpu_submission *s = &vgpu->submission;
	struct intel_vgpu_execlist *execlist = &s->execlist[ring_id];
	struct intel_vgpu_workload *next_workload;
	struct list_head *next = workload_q_head(vgpu, ring_id)->next;
	bool lite_restore = false;
	int ret = 0;

	gvt_dbg_el("complete workload %p status %d\n", workload,
			workload->status);

	if (workload->status || (vgpu->resetting_eng & BIT(ring_id)))
		goto out;

	if (!list_empty(workload_q_head(vgpu, ring_id))) {
		struct execlist_ctx_descriptor_format *this_desc, *next_desc;

		next_workload = container_of(next,
				struct intel_vgpu_workload, list);
		this_desc = &workload->ctx_desc;
		next_desc = &next_workload->ctx_desc;

		lite_restore = same_context(this_desc, next_desc);
	}

	if (lite_restore) {
		gvt_dbg_el("next context == current - no schedule-out\n");
		goto out;
	}

	ret = emulate_execlist_ctx_schedule_out(execlist, &workload->ctx_desc);
out:
	intel_vgpu_unpin_mm(workload->shadow_mm);
	intel_vgpu_destroy_workload(workload);
	return ret;
}

static int submit_context(struct intel_vgpu *vgpu, int ring_id,
		struct execlist_ctx_descriptor_format *desc,
		bool emulate_schedule_in)
{
	struct intel_vgpu_submission *s = &vgpu->submission;
	struct intel_vgpu_workload *workload = NULL;

	workload = intel_vgpu_create_workload(vgpu, ring_id, desc);
	if (IS_ERR(workload))
		return PTR_ERR(workload);

	workload->prepare = prepare_execlist_workload;
	workload->complete = complete_execlist_workload;
	workload->emulate_schedule_in = emulate_schedule_in;

	if (emulate_schedule_in)
		workload->elsp_dwords = s->execlist[ring_id].elsp_dwords;

	gvt_dbg_el("workload %p emulate schedule_in %d\n", workload,
			emulate_schedule_in);

	intel_vgpu_queue_workload(workload);
	return 0;
}

int intel_vgpu_submit_execlist(struct intel_vgpu *vgpu, int ring_id)
{
	struct intel_vgpu_submission *s = &vgpu->submission;
	struct intel_vgpu_execlist *execlist = &s->execlist[ring_id];
	struct execlist_ctx_descriptor_format *desc[2];
	int i, ret;

	desc[0] = get_desc_from_elsp_dwords(&execlist->elsp_dwords, 0);
	desc[1] = get_desc_from_elsp_dwords(&execlist->elsp_dwords, 1);

	if (!desc[0]->valid) {
		gvt_vgpu_err("invalid elsp submission, desc0 is invalid\n");
		goto inv_desc;
	}

	for (i = 0; i < ARRAY_SIZE(desc); i++) {
		if (!desc[i]->valid)
			continue;
		if (!desc[i]->privilege_access) {
			gvt_vgpu_err("unexpected GGTT elsp submission\n");
			goto inv_desc;
		}
	}

	/* submit workload */
	for (i = 0; i < ARRAY_SIZE(desc); i++) {
		if (!desc[i]->valid)
			continue;
		ret = submit_context(vgpu, ring_id, desc[i], i == 0);
		if (ret) {
			gvt_vgpu_err("failed to submit desc %d\n", i);
			return ret;
		}
	}

	return 0;

inv_desc:
	gvt_vgpu_err("descriptors content: desc0 %08x %08x desc1 %08x %08x\n",
		     desc[0]->udw, desc[0]->ldw, desc[1]->udw, desc[1]->ldw);
	return -EINVAL;
}

static void init_vgpu_execlist(struct intel_vgpu *vgpu, int ring_id)
{
	struct intel_vgpu_submission *s = &vgpu->submission;
	struct intel_vgpu_execlist *execlist = &s->execlist[ring_id];
	struct execlist_context_status_pointer_format ctx_status_ptr;
	u32 ctx_status_ptr_reg;

	memset(execlist, 0, sizeof(*execlist));

	execlist->vgpu = vgpu;
	execlist->ring_id = ring_id;
	execlist->slot[0].index = 0;
	execlist->slot[1].index = 1;

	ctx_status_ptr_reg = execlist_ring_mmio(vgpu->gvt, ring_id,
			_EL_OFFSET_STATUS_PTR);
	ctx_status_ptr.dw = vgpu_vreg(vgpu, ctx_status_ptr_reg);
	ctx_status_ptr.read_ptr = 0;
	ctx_status_ptr.write_ptr = 0x7;
	vgpu_vreg(vgpu, ctx_status_ptr_reg) = ctx_status_ptr.dw;
}

static void clean_execlist(struct intel_vgpu *vgpu,
			   intel_engine_mask_t engine_mask)
{
	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
	struct intel_engine_cs *engine;
	struct intel_vgpu_submission *s = &vgpu->submission;
	intel_engine_mask_t tmp;

	for_each_engine_masked(engine, dev_priv, engine_mask, tmp) {
		kfree(s->ring_scan_buffer[engine->id]);
		s->ring_scan_buffer[engine->id] = NULL;
		s->ring_scan_buffer_size[engine->id] = 0;
	}
}

static void reset_execlist(struct intel_vgpu *vgpu,
			   intel_engine_mask_t engine_mask)
{
	struct drm_i915_private *dev_priv = vgpu->gvt->dev_priv;
	struct intel_engine_cs *engine;
	intel_engine_mask_t tmp;

	for_each_engine_masked(engine, dev_priv, engine_mask, tmp)
		init_vgpu_execlist(vgpu, engine->id);
}

static int init_execlist(struct intel_vgpu *vgpu,
			 intel_engine_mask_t engine_mask)
{
	reset_execlist(vgpu, engine_mask);
	return 0;
}

const struct intel_vgpu_submission_ops intel_vgpu_execlist_submission_ops = {
	.name = "execlist",
	.init = init_execlist,
	.reset = reset_execlist,
	.clean = clean_execlist,
};