summaryrefslogtreecommitdiffstats
path: root/drivers/gpu/drm/i915/gt/sysfs_engines.c
blob: f2d9858d827c23bd3a5bb16a4feaf8d1c9c7b4b8 (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
// SPDX-License-Identifier: MIT
/*
 * Copyright © 2019 Intel Corporation
 */

#include <linux/kobject.h>
#include <linux/sysfs.h>

#include "i915_drv.h"
#include "intel_engine.h"
#include "intel_engine_heartbeat.h"
#include "sysfs_engines.h"

struct kobj_engine {
	struct kobject base;
	struct intel_engine_cs *engine;
};

static struct intel_engine_cs *kobj_to_engine(struct kobject *kobj)
{
	return container_of(kobj, struct kobj_engine, base)->engine;
}

static ssize_t
name_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
{
	return sprintf(buf, "%s\n", kobj_to_engine(kobj)->name);
}

static struct kobj_attribute name_attr =
__ATTR(name, 0444, name_show, NULL);

static ssize_t
class_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
{
	return sprintf(buf, "%d\n", kobj_to_engine(kobj)->uabi_class);
}

static struct kobj_attribute class_attr =
__ATTR(class, 0444, class_show, NULL);

static ssize_t
inst_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
{
	return sprintf(buf, "%d\n", kobj_to_engine(kobj)->uabi_instance);
}

static struct kobj_attribute inst_attr =
__ATTR(instance, 0444, inst_show, NULL);

static ssize_t
mmio_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
{
	return sprintf(buf, "0x%x\n", kobj_to_engine(kobj)->mmio_base);
}

static struct kobj_attribute mmio_attr =
__ATTR(mmio_base, 0444, mmio_show, NULL);

static const char * const vcs_caps[] = {
	[ilog2(I915_VIDEO_CLASS_CAPABILITY_HEVC)] = "hevc",
	[ilog2(I915_VIDEO_AND_ENHANCE_CLASS_CAPABILITY_SFC)] = "sfc",
};

static const char * const vecs_caps[] = {
	[ilog2(I915_VIDEO_AND_ENHANCE_CLASS_CAPABILITY_SFC)] = "sfc",
};

static ssize_t repr_trim(char *buf, ssize_t len)
{
	/* Trim off the trailing space and replace with a newline */
	if (len > PAGE_SIZE)
		len = PAGE_SIZE;
	if (len > 0)
		buf[len - 1] = '\n';

	return len;
}

static ssize_t
__caps_show(struct intel_engine_cs *engine,
	    unsigned long caps, char *buf, bool show_unknown)
{
	const char * const *repr;
	int count, n;
	ssize_t len;

	switch (engine->class) {
	case VIDEO_DECODE_CLASS:
		repr = vcs_caps;
		count = ARRAY_SIZE(vcs_caps);
		break;

	case VIDEO_ENHANCEMENT_CLASS:
		repr = vecs_caps;
		count = ARRAY_SIZE(vecs_caps);
		break;

	default:
		repr = NULL;
		count = 0;
		break;
	}
	GEM_BUG_ON(count > BITS_PER_LONG);

	len = 0;
	for_each_set_bit(n, &caps, show_unknown ? BITS_PER_LONG : count) {
		if (n >= count || !repr[n]) {
			if (GEM_WARN_ON(show_unknown))
				len += snprintf(buf + len, PAGE_SIZE - len,
						"[%x] ", n);
		} else {
			len += snprintf(buf + len, PAGE_SIZE - len,
					"%s ", repr[n]);
		}
		if (GEM_WARN_ON(len >= PAGE_SIZE))
			break;
	}
	return repr_trim(buf, len);
}

static ssize_t
caps_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
{
	struct intel_engine_cs *engine = kobj_to_engine(kobj);

	return __caps_show(engine, engine->uabi_capabilities, buf, true);
}

static struct kobj_attribute caps_attr =
__ATTR(capabilities, 0444, caps_show, NULL);

static ssize_t
all_caps_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
{
	return __caps_show(kobj_to_engine(kobj), -1, buf, false);
}

static struct kobj_attribute all_caps_attr =
__ATTR(known_capabilities, 0444, all_caps_show, NULL);

static ssize_t
max_spin_store(struct kobject *kobj, struct kobj_attribute *attr,
	       const char *buf, size_t count)
{
	struct intel_engine_cs *engine = kobj_to_engine(kobj);
	unsigned long long duration, clamped;
	int err;

	/*
	 * When waiting for a request, if is it currently being executed
	 * on the GPU, we busywait for a short while before sleeping. The
	 * premise is that most requests are short, and if it is already
	 * executing then there is a good chance that it will complete
	 * before we can setup the interrupt handler and go to sleep.
	 * We try to offset the cost of going to sleep, by first spinning
	 * on the request -- if it completed in less time than it would take
	 * to go sleep, process the interrupt and return back to the client,
	 * then we have saved the client some latency, albeit at the cost
	 * of spinning on an expensive CPU core.
	 *
	 * While we try to avoid waiting at all for a request that is unlikely
	 * to complete, deciding how long it is worth spinning is for is an
	 * arbitrary decision: trading off power vs latency.
	 */

	err = kstrtoull(buf, 0, &duration);
	if (err)
		return err;

	clamped = intel_clamp_max_busywait_duration_ns(engine, duration);
	if (duration != clamped)
		return -EINVAL;

	WRITE_ONCE(engine->props.max_busywait_duration_ns, duration);

	return count;
}

static ssize_t
max_spin_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
{
	struct intel_engine_cs *engine = kobj_to_engine(kobj);

	return sprintf(buf, "%lu\n", engine->props.max_busywait_duration_ns);
}

static struct kobj_attribute max_spin_attr =
__ATTR(max_busywait_duration_ns, 0644, max_spin_show, max_spin_store);

static ssize_t
max_spin_default(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
{
	struct intel_engine_cs *engine = kobj_to_engine(kobj);

	return sprintf(buf, "%lu\n", engine->defaults.max_busywait_duration_ns);
}

static struct kobj_attribute max_spin_def =
__ATTR(max_busywait_duration_ns, 0444, max_spin_default, NULL);

static ssize_t
timeslice_store(struct kobject *kobj, struct kobj_attribute *attr,
		const char *buf, size_t count)
{
	struct intel_engine_cs *engine = kobj_to_engine(kobj);
	unsigned long long duration, clamped;
	int err;

	/*
	 * Execlists uses a scheduling quantum (a timeslice) to alternate
	 * execution between ready-to-run contexts of equal priority. This
	 * ensures that all users (though only if they of equal importance)
	 * have the opportunity to run and prevents livelocks where contexts
	 * may have implicit ordering due to userspace semaphores.
	 */

	err = kstrtoull(buf, 0, &duration);
	if (err)
		return err;

	clamped = intel_clamp_timeslice_duration_ms(engine, duration);
	if (duration != clamped)
		return -EINVAL;

	WRITE_ONCE(engine->props.timeslice_duration_ms, duration);

	if (execlists_active(&engine->execlists))
		set_timer_ms(&engine->execlists.timer, duration);

	return count;
}

static ssize_t
timeslice_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
{
	struct intel_engine_cs *engine = kobj_to_engine(kobj);

	return sprintf(buf, "%lu\n", engine->props.timeslice_duration_ms);
}

static struct kobj_attribute timeslice_duration_attr =
__ATTR(timeslice_duration_ms, 0644, timeslice_show, timeslice_store);

static ssize_t
timeslice_default(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
{
	struct intel_engine_cs *engine = kobj_to_engine(kobj);

	return sprintf(buf, "%lu\n", engine->defaults.timeslice_duration_ms);
}

static struct kobj_attribute timeslice_duration_def =
__ATTR(timeslice_duration_ms, 0444, timeslice_default, NULL);

static ssize_t
stop_store(struct kobject *kobj, struct kobj_attribute *attr,
	   const char *buf, size_t count)
{
	struct intel_engine_cs *engine = kobj_to_engine(kobj);
	unsigned long long duration, clamped;
	int err;

	/*
	 * When we allow ourselves to sleep before a GPU reset after disabling
	 * submission, even for a few milliseconds, gives an innocent context
	 * the opportunity to clear the GPU before the reset occurs. However,
	 * how long to sleep depends on the typical non-preemptible duration
	 * (a similar problem to determining the ideal preempt-reset timeout
	 * or even the heartbeat interval).
	 */

	err = kstrtoull(buf, 0, &duration);
	if (err)
		return err;

	clamped = intel_clamp_stop_timeout_ms(engine, duration);
	if (duration != clamped)
		return -EINVAL;

	WRITE_ONCE(engine->props.stop_timeout_ms, duration);
	return count;
}

static ssize_t
stop_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
{
	struct intel_engine_cs *engine = kobj_to_engine(kobj);

	return sprintf(buf, "%lu\n", engine->props.stop_timeout_ms);
}

static struct kobj_attribute stop_timeout_attr =
__ATTR(stop_timeout_ms, 0644, stop_show, stop_store);

static ssize_t
stop_default(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
{
	struct intel_engine_cs *engine = kobj_to_engine(kobj);

	return sprintf(buf, "%lu\n", engine->defaults.stop_timeout_ms);
}

static struct kobj_attribute stop_timeout_def =
__ATTR(stop_timeout_ms, 0444, stop_default, NULL);

static ssize_t
preempt_timeout_store(struct kobject *kobj, struct kobj_attribute *attr,
		      const char *buf, size_t count)
{
	struct intel_engine_cs *engine = kobj_to_engine(kobj);
	unsigned long long timeout, clamped;
	int err;

	/*
	 * After initialising a preemption request, we give the current
	 * resident a small amount of time to vacate the GPU. The preemption
	 * request is for a higher priority context and should be immediate to
	 * maintain high quality of service (and avoid priority inversion).
	 * However, the preemption granularity of the GPU can be quite coarse
	 * and so we need a compromise.
	 */

	err = kstrtoull(buf, 0, &timeout);
	if (err)
		return err;

	clamped = intel_clamp_preempt_timeout_ms(engine, timeout);
	if (timeout != clamped)
		return -EINVAL;

	WRITE_ONCE(engine->props.preempt_timeout_ms, timeout);

	if (READ_ONCE(engine->execlists.pending[0]))
		set_timer_ms(&engine->execlists.preempt, timeout);

	return count;
}

static ssize_t
preempt_timeout_show(struct kobject *kobj, struct kobj_attribute *attr,
		     char *buf)
{
	struct intel_engine_cs *engine = kobj_to_engine(kobj);

	return sprintf(buf, "%lu\n", engine->props.preempt_timeout_ms);
}

static struct kobj_attribute preempt_timeout_attr =
__ATTR(preempt_timeout_ms, 0644, preempt_timeout_show, preempt_timeout_store);

static ssize_t
preempt_timeout_default(struct kobject *kobj, struct kobj_attribute *attr,
			char *buf)
{
	struct intel_engine_cs *engine = kobj_to_engine(kobj);

	return sprintf(buf, "%lu\n", engine->defaults.preempt_timeout_ms);
}

static struct kobj_attribute preempt_timeout_def =
__ATTR(preempt_timeout_ms, 0444, preempt_timeout_default, NULL);

static ssize_t
heartbeat_store(struct kobject *kobj, struct kobj_attribute *attr,
		const char *buf, size_t count)
{
	struct intel_engine_cs *engine = kobj_to_engine(kobj);
	unsigned long long delay, clamped;
	int err;

	/*
	 * We monitor the health of the system via periodic heartbeat pulses.
	 * The pulses also provide the opportunity to perform garbage
	 * collection.  However, we interpret an incomplete pulse (a missed
	 * heartbeat) as an indication that the system is no longer responsive,
	 * i.e. hung, and perform an engine or full GPU reset. Given that the
	 * preemption granularity can be very coarse on a system, the optimal
	 * value for any workload is unknowable!
	 */

	err = kstrtoull(buf, 0, &delay);
	if (err)
		return err;

	clamped = intel_clamp_heartbeat_interval_ms(engine, delay);
	if (delay != clamped)
		return -EINVAL;

	err = intel_engine_set_heartbeat(engine, delay);
	if (err)
		return err;

	return count;
}

static ssize_t
heartbeat_show(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
{
	struct intel_engine_cs *engine = kobj_to_engine(kobj);

	return sprintf(buf, "%lu\n", engine->props.heartbeat_interval_ms);
}

static struct kobj_attribute heartbeat_interval_attr =
__ATTR(heartbeat_interval_ms, 0644, heartbeat_show, heartbeat_store);

static ssize_t
heartbeat_default(struct kobject *kobj, struct kobj_attribute *attr, char *buf)
{
	struct intel_engine_cs *engine = kobj_to_engine(kobj);

	return sprintf(buf, "%lu\n", engine->defaults.heartbeat_interval_ms);
}

static struct kobj_attribute heartbeat_interval_def =
__ATTR(heartbeat_interval_ms, 0444, heartbeat_default, NULL);

static void kobj_engine_release(struct kobject *kobj)
{
	kfree(kobj);
}

static struct kobj_type kobj_engine_type = {
	.release = kobj_engine_release,
	.sysfs_ops = &kobj_sysfs_ops
};

static struct kobject *
kobj_engine(struct kobject *dir, struct intel_engine_cs *engine)
{
	struct kobj_engine *ke;

	ke = kzalloc(sizeof(*ke), GFP_KERNEL);
	if (!ke)
		return NULL;

	kobject_init(&ke->base, &kobj_engine_type);
	ke->engine = engine;

	if (kobject_add(&ke->base, dir, "%s", engine->name)) {
		kobject_put(&ke->base);
		return NULL;
	}

	/* xfer ownership to sysfs tree */
	return &ke->base;
}

static void add_defaults(struct kobj_engine *parent)
{
	static const struct attribute *files[] = {
		&max_spin_def.attr,
		&stop_timeout_def.attr,
#if CONFIG_DRM_I915_HEARTBEAT_INTERVAL
		&heartbeat_interval_def.attr,
#endif
		NULL
	};
	struct kobj_engine *ke;

	ke = kzalloc(sizeof(*ke), GFP_KERNEL);
	if (!ke)
		return;

	kobject_init(&ke->base, &kobj_engine_type);
	ke->engine = parent->engine;

	if (kobject_add(&ke->base, &parent->base, "%s", ".defaults")) {
		kobject_put(&ke->base);
		return;
	}

	if (sysfs_create_files(&ke->base, files))
		return;

	if (intel_engine_has_timeslices(ke->engine) &&
	    sysfs_create_file(&ke->base, &timeslice_duration_def.attr))
		return;

	if (intel_engine_has_preempt_reset(ke->engine) &&
	    sysfs_create_file(&ke->base, &preempt_timeout_def.attr))
		return;
}

void intel_engines_add_sysfs(struct drm_i915_private *i915)
{
	static const struct attribute *files[] = {
		&name_attr.attr,
		&class_attr.attr,
		&inst_attr.attr,
		&mmio_attr.attr,
		&caps_attr.attr,
		&all_caps_attr.attr,
		&max_spin_attr.attr,
		&stop_timeout_attr.attr,
#if CONFIG_DRM_I915_HEARTBEAT_INTERVAL
		&heartbeat_interval_attr.attr,
#endif
		NULL
	};

	struct device *kdev = i915->drm.primary->kdev;
	struct intel_engine_cs *engine;
	struct kobject *dir;

	dir = kobject_create_and_add("engine", &kdev->kobj);
	if (!dir)
		return;

	for_each_uabi_engine(engine, i915) {
		struct kobject *kobj;

		kobj = kobj_engine(dir, engine);
		if (!kobj)
			goto err_engine;

		if (sysfs_create_files(kobj, files))
			goto err_object;

		if (intel_engine_has_timeslices(engine) &&
		    sysfs_create_file(kobj, &timeslice_duration_attr.attr))
			goto err_engine;

		if (intel_engine_has_preempt_reset(engine) &&
		    sysfs_create_file(kobj, &preempt_timeout_attr.attr))
			goto err_engine;

		add_defaults(container_of(kobj, struct kobj_engine, base));

		if (0) {
err_object:
			kobject_put(kobj);
err_engine:
			dev_err(kdev, "Failed to add sysfs engine '%s'\n",
				engine->name);
			break;
		}
	}
}