summaryrefslogtreecommitdiffstats
path: root/drivers/gpu/drm/i915/i915_gem_context.c
blob: a9d58d72bb4da2215592bb8ba6554e9835fb48cf (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
/*
 * Copyright © 2011-2012 Intel Corporation
 *
 * 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:
 *    Ben Widawsky <ben@bwidawsk.net>
 *
 */

/*
 * This file implements HW context support. On gen5+ a HW context consists of an
 * opaque GPU object which is referenced at times of context saves and restores.
 * With RC6 enabled, the context is also referenced as the GPU enters and exists
 * from RC6 (GPU has it's own internal power context, except on gen5). Though
 * something like a context does exist for the media ring, the code only
 * supports contexts for the render ring.
 *
 * In software, there is a distinction between contexts created by the user,
 * and the default HW context. The default HW context is used by GPU clients
 * that do not request setup of their own hardware context. The default
 * context's state is never restored to help prevent programming errors. This
 * would happen if a client ran and piggy-backed off another clients GPU state.
 * The default context only exists to give the GPU some offset to load as the
 * current to invoke a save of the context we actually care about. In fact, the
 * code could likely be constructed, albeit in a more complicated fashion, to
 * never use the default context, though that limits the driver's ability to
 * swap out, and/or destroy other contexts.
 *
 * All other contexts are created as a request by the GPU client. These contexts
 * store GPU state, and thus allow GPU clients to not re-emit state (and
 * potentially query certain state) at any time. The kernel driver makes
 * certain that the appropriate commands are inserted.
 *
 * The context life cycle is semi-complicated in that context BOs may live
 * longer than the context itself because of the way the hardware, and object
 * tracking works. Below is a very crude representation of the state machine
 * describing the context life.
 *                                         refcount     pincount     active
 * S0: initial state                          0            0           0
 * S1: context created                        1            0           0
 * S2: context is currently running           2            1           X
 * S3: GPU referenced, but not current        2            0           1
 * S4: context is current, but destroyed      1            1           0
 * S5: like S3, but destroyed                 1            0           1
 *
 * The most common (but not all) transitions:
 * S0->S1: client creates a context
 * S1->S2: client submits execbuf with context
 * S2->S3: other clients submits execbuf with context
 * S3->S1: context object was retired
 * S3->S2: clients submits another execbuf
 * S2->S4: context destroy called with current context
 * S3->S5->S0: destroy path
 * S4->S5->S0: destroy path on current context
 *
 * There are two confusing terms used above:
 *  The "current context" means the context which is currently running on the
 *  GPU. The GPU has loaded it's state already and has stored away the gtt
 *  offset of the BO. The GPU is not actively referencing the data at this
 *  offset, but it will on the next context switch. The only way to avoid this
 *  is to do a GPU reset.
 *
 *  An "active context' is one which was previously the "current context" and is
 *  on the active list waiting for the next context switch to occur. Until this
 *  happens, the object must remain at the same gtt offset. It is therefore
 *  possible to destroy a context, but it is still active.
 *
 */

#include "drmP.h"
#include "i915_drm.h"
#include "i915_drv.h"

/* This is a HW constraint. The value below is the largest known requirement
 * I've seen in a spec to date, and that was a workaround for a non-shipping
 * part. It should be safe to decrease this, but it's more future proof as is.
 */
#define CONTEXT_ALIGN (64<<10)

static struct i915_hw_context *
i915_gem_context_get(struct drm_i915_file_private *file_priv, u32 id);
static int do_switch(struct drm_i915_gem_object *from_obj,
		     struct i915_hw_context *to, u32 seqno);

static int get_context_size(struct drm_device *dev)
{
	struct drm_i915_private *dev_priv = dev->dev_private;
	int ret;
	u32 reg;

	switch (INTEL_INFO(dev)->gen) {
	case 6:
		reg = I915_READ(CXT_SIZE);
		ret = GEN6_CXT_TOTAL_SIZE(reg) * 64;
		break;
	case 7:
		reg = I915_READ(GEN7_CXT_SIZE);
		ret = GEN7_CXT_TOTAL_SIZE(reg) * 64;
		break;
	default:
		BUG();
	}

	return ret;
}

static void do_destroy(struct i915_hw_context *ctx)
{
	struct drm_device *dev = ctx->obj->base.dev;
	struct drm_i915_private *dev_priv = dev->dev_private;

	if (ctx->file_priv)
		idr_remove(&ctx->file_priv->context_idr, ctx->id);
	else
		BUG_ON(ctx != dev_priv->ring[RCS].default_context);

	drm_gem_object_unreference(&ctx->obj->base);
	kfree(ctx);
}

static struct i915_hw_context *
create_hw_context(struct drm_device *dev,
		  struct drm_i915_file_private *file_priv)
{
	struct drm_i915_private *dev_priv = dev->dev_private;
	struct i915_hw_context *ctx;
	int ret, id;

	ctx = kzalloc(sizeof(struct drm_i915_file_private), GFP_KERNEL);
	if (ctx == NULL)
		return ERR_PTR(-ENOMEM);

	ctx->obj = i915_gem_alloc_object(dev, dev_priv->hw_context_size);
	if (ctx->obj == NULL) {
		kfree(ctx);
		DRM_DEBUG_DRIVER("Context object allocated failed\n");
		return ERR_PTR(-ENOMEM);
	}

	/* The ring associated with the context object is handled by the normal
	 * object tracking code. We give an initial ring value simple to pass an
	 * assertion in the context switch code.
	 */
	ctx->ring = &dev_priv->ring[RCS];

	/* Default context will never have a file_priv */
	if (file_priv == NULL)
		return ctx;

	ctx->file_priv = file_priv;

again:
	if (idr_pre_get(&file_priv->context_idr, GFP_KERNEL) == 0) {
		ret = -ENOMEM;
		DRM_DEBUG_DRIVER("idr allocation failed\n");
		goto err_out;
	}

	ret = idr_get_new_above(&file_priv->context_idr, ctx,
				DEFAULT_CONTEXT_ID + 1, &id);
	if (ret == 0)
		ctx->id = id;

	if (ret == -EAGAIN)
		goto again;
	else if (ret)
		goto err_out;

	return ctx;

err_out:
	do_destroy(ctx);
	return ERR_PTR(ret);
}

static inline bool is_default_context(struct i915_hw_context *ctx)
{
	return (ctx == ctx->ring->default_context);
}

/**
 * The default context needs to exist per ring that uses contexts. It stores the
 * context state of the GPU for applications that don't utilize HW contexts, as
 * well as an idle case.
 */
static int create_default_context(struct drm_i915_private *dev_priv)
{
	struct i915_hw_context *ctx;
	int ret;

	BUG_ON(!mutex_is_locked(&dev_priv->dev->struct_mutex));

	ctx = create_hw_context(dev_priv->dev, NULL);
	if (IS_ERR(ctx))
		return PTR_ERR(ctx);

	/* We may need to do things with the shrinker which require us to
	 * immediately switch back to the default context. This can cause a
	 * problem as pinning the default context also requires GTT space which
	 * may not be available. To avoid this we always pin the
	 * default context.
	 */
	dev_priv->ring[RCS].default_context = ctx;
	ret = i915_gem_object_pin(ctx->obj, CONTEXT_ALIGN, false);
	if (ret) {
		do_destroy(ctx);
		return ret;
	}

	ret = do_switch(NULL, ctx, 0);
	if (ret) {
		i915_gem_object_unpin(ctx->obj);
		do_destroy(ctx);
	} else {
		DRM_DEBUG_DRIVER("Default HW context loaded\n");
	}

	return ret;
}

void i915_gem_context_init(struct drm_device *dev)
{
	struct drm_i915_private *dev_priv = dev->dev_private;
	uint32_t ctx_size;

	if (!HAS_HW_CONTEXTS(dev)) {
		dev_priv->hw_contexts_disabled = true;
		return;
	}

	/* If called from reset, or thaw... we've been here already */
	if (dev_priv->hw_contexts_disabled ||
	    dev_priv->ring[RCS].default_context)
		return;

	ctx_size = get_context_size(dev);
	dev_priv->hw_context_size = get_context_size(dev);
	dev_priv->hw_context_size = round_up(dev_priv->hw_context_size, 4096);

	if (ctx_size <= 0 || ctx_size > (1<<20)) {
		dev_priv->hw_contexts_disabled = true;
		return;
	}

	if (create_default_context(dev_priv)) {
		dev_priv->hw_contexts_disabled = true;
		return;
	}

	DRM_DEBUG_DRIVER("HW context support initialized\n");
}

void i915_gem_context_fini(struct drm_device *dev)
{
	struct drm_i915_private *dev_priv = dev->dev_private;

	if (dev_priv->hw_contexts_disabled)
		return;

	/* The only known way to stop the gpu from accessing the hw context is
	 * to reset it. Do this as the very last operation to avoid confusing
	 * other code, leading to spurious errors. */
	intel_gpu_reset(dev);

	i915_gem_object_unpin(dev_priv->ring[RCS].default_context->obj);

	do_destroy(dev_priv->ring[RCS].default_context);
}

static int context_idr_cleanup(int id, void *p, void *data)
{
	struct i915_hw_context *ctx = p;

	BUG_ON(id == DEFAULT_CONTEXT_ID);

	do_destroy(ctx);

	return 0;
}

void i915_gem_context_close(struct drm_device *dev, struct drm_file *file)
{
	struct drm_i915_file_private *file_priv = file->driver_priv;

	mutex_lock(&dev->struct_mutex);
	idr_for_each(&file_priv->context_idr, context_idr_cleanup, NULL);
	idr_destroy(&file_priv->context_idr);
	mutex_unlock(&dev->struct_mutex);
}

static struct i915_hw_context *
i915_gem_context_get(struct drm_i915_file_private *file_priv, u32 id)
{
	return (struct i915_hw_context *)idr_find(&file_priv->context_idr, id);
}

static inline int
mi_set_context(struct intel_ring_buffer *ring,
	       struct i915_hw_context *new_context,
	       u32 hw_flags)
{
	int ret;

	/* w/a: If Flush TLB Invalidation Mode is enabled, driver must do a TLB
	 * invalidation prior to MI_SET_CONTEXT. On GEN6 we don't set the value
	 * explicitly, so we rely on the value at ring init, stored in
	 * itlb_before_ctx_switch.
	 */
	if (IS_GEN6(ring->dev) && ring->itlb_before_ctx_switch) {
		ret = ring->flush(ring, 0, 0);
		if (ret)
			return ret;
	}

	ret = intel_ring_begin(ring, 6);
	if (ret)
		return ret;

	if (IS_GEN7(ring->dev))
		intel_ring_emit(ring, MI_ARB_ON_OFF | MI_ARB_DISABLE);
	else
		intel_ring_emit(ring, MI_NOOP);

	intel_ring_emit(ring, MI_NOOP);
	intel_ring_emit(ring, MI_SET_CONTEXT);
	intel_ring_emit(ring, new_context->obj->gtt_offset |
			MI_MM_SPACE_GTT |
			MI_SAVE_EXT_STATE_EN |
			MI_RESTORE_EXT_STATE_EN |
			hw_flags);
	/* w/a: MI_SET_CONTEXT must always be followed by MI_NOOP */
	intel_ring_emit(ring, MI_NOOP);

	if (IS_GEN7(ring->dev))
		intel_ring_emit(ring, MI_ARB_ON_OFF | MI_ARB_ENABLE);
	else
		intel_ring_emit(ring, MI_NOOP);

	intel_ring_advance(ring);

	return ret;
}

static int do_switch(struct drm_i915_gem_object *from_obj,
		     struct i915_hw_context *to,
		     u32 seqno)
{
	struct intel_ring_buffer *ring = NULL;
	u32 hw_flags = 0;
	int ret;

	BUG_ON(to == NULL);
	BUG_ON(from_obj != NULL && from_obj->pin_count == 0);

	ret = i915_gem_object_pin(to->obj, CONTEXT_ALIGN, false);
	if (ret)
		return ret;

	/* Clear this page out of any CPU caches for coherent swap-in/out. Note
	 * that thanks to write = false in this call and us not setting any gpu
	 * write domains when putting a context object onto the active list
	 * (when switching away from it), this won't block.
	 * XXX: We need a real interface to do this instead of trickery. */
	ret = i915_gem_object_set_to_gtt_domain(to->obj, false);
	if (ret) {
		i915_gem_object_unpin(to->obj);
		return ret;
	}

	if (!to->obj->has_global_gtt_mapping)
		i915_gem_gtt_bind_object(to->obj, to->obj->cache_level);

	if (!to->is_initialized || is_default_context(to))
		hw_flags |= MI_RESTORE_INHIBIT;
	else if (WARN_ON_ONCE(from_obj == to->obj)) /* not yet expected */
		hw_flags |= MI_FORCE_RESTORE;

	ring = to->ring;
	ret = mi_set_context(ring, to, hw_flags);
	if (ret) {
		i915_gem_object_unpin(to->obj);
		return ret;
	}

	/* The backing object for the context is done after switching to the
	 * *next* context. Therefore we cannot retire the previous context until
	 * the next context has already started running. In fact, the below code
	 * is a bit suboptimal because the retiring can occur simply after the
	 * MI_SET_CONTEXT instead of when the next seqno has completed.
	 */
	if (from_obj != NULL) {
		from_obj->base.read_domains = I915_GEM_DOMAIN_INSTRUCTION;
		i915_gem_object_move_to_active(from_obj, ring, seqno);
		/* As long as MI_SET_CONTEXT is serializing, ie. it flushes the
		 * whole damn pipeline, we don't need to explicitly mark the
		 * object dirty. The only exception is that the context must be
		 * correct in case the object gets swapped out. Ideally we'd be
		 * able to defer doing this until we know the object would be
		 * swapped, but there is no way to do that yet.
		 */
		from_obj->dirty = 1;
		BUG_ON(from_obj->ring != to->ring);
		i915_gem_object_unpin(from_obj);

		drm_gem_object_unreference(&from_obj->base);
	}

	drm_gem_object_reference(&to->obj->base);
	ring->last_context_obj = to->obj;
	to->is_initialized = true;

	return 0;
}

/**
 * i915_switch_context() - perform a GPU context switch.
 * @ring: ring for which we'll execute the context switch
 * @file_priv: file_priv associated with the context, may be NULL
 * @id: context id number
 * @seqno: sequence number by which the new context will be switched to
 * @flags:
 *
 * The context life cycle is simple. The context refcount is incremented and
 * decremented by 1 and create and destroy. If the context is in use by the GPU,
 * it will have a refoucnt > 1. This allows us to destroy the context abstract
 * object while letting the normal object tracking destroy the backing BO.
 */
int i915_switch_context(struct intel_ring_buffer *ring,
			struct drm_file *file,
			int to_id)
{
	struct drm_i915_private *dev_priv = ring->dev->dev_private;
	struct drm_i915_file_private *file_priv = NULL;
	struct i915_hw_context *to;
	struct drm_i915_gem_object *from_obj = ring->last_context_obj;

	if (dev_priv->hw_contexts_disabled)
		return 0;

	if (ring != &dev_priv->ring[RCS])
		return 0;

	if (file)
		file_priv = file->driver_priv;

	if (to_id == DEFAULT_CONTEXT_ID) {
		to = ring->default_context;
	} else {
		to = i915_gem_context_get(file_priv, to_id);
		if (to == NULL)
			return -ENOENT;
	}

	if (from_obj == to->obj)
		return 0;

	return do_switch(from_obj, to, i915_gem_next_request_seqno(to->ring));
}

int i915_gem_context_create_ioctl(struct drm_device *dev, void *data,
				  struct drm_file *file)
{
	struct drm_i915_private *dev_priv = dev->dev_private;
	struct drm_i915_gem_context_create *args = data;
	struct drm_i915_file_private *file_priv = file->driver_priv;
	struct i915_hw_context *ctx;
	int ret;

	if (!(dev->driver->driver_features & DRIVER_GEM))
		return -ENODEV;

	if (dev_priv->hw_contexts_disabled)
		return -ENODEV;

	ret = i915_mutex_lock_interruptible(dev);
	if (ret)
		return ret;

	ctx = create_hw_context(dev, file_priv);
	mutex_unlock(&dev->struct_mutex);
	if (IS_ERR(ctx))
		return PTR_ERR(ctx);

	args->ctx_id = ctx->id;
	DRM_DEBUG_DRIVER("HW context %d created\n", args->ctx_id);

	return 0;
}

int i915_gem_context_destroy_ioctl(struct drm_device *dev, void *data,
				   struct drm_file *file)
{
	struct drm_i915_gem_context_destroy *args = data;
	struct drm_i915_file_private *file_priv = file->driver_priv;
	struct i915_hw_context *ctx;
	int ret;

	if (!(dev->driver->driver_features & DRIVER_GEM))
		return -ENODEV;

	ret = i915_mutex_lock_interruptible(dev);
	if (ret)
		return ret;

	ctx = i915_gem_context_get(file_priv, args->ctx_id);
	if (!ctx) {
		mutex_unlock(&dev->struct_mutex);
		return -ENOENT;
	}

	do_destroy(ctx);

	mutex_unlock(&dev->struct_mutex);

	DRM_DEBUG_DRIVER("HW context %d destroyed\n", args->ctx_id);
	return 0;
}
OpenPOWER on IntegriCloud