summaryrefslogtreecommitdiffstats
path: root/drivers/gpu/drm/nouveau/nouveau_channel.c
blob: 6dfb425cbae9432f4718ae33c77b754c4ac9a8dc (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
/*
 * Copyright 2005-2006 Stephane Marchesin
 * 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
 * PRECISION INSIGHT AND/OR ITS SUPPLIERS 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.
 */

#include "drmP.h"
#include "drm.h"
#include "nouveau_drv.h"
#include "nouveau_drm.h"
#include "nouveau_dma.h"

static int
nouveau_channel_pushbuf_ctxdma_init(struct nouveau_channel *chan)
{
	struct drm_device *dev = chan->dev;
	struct drm_nouveau_private *dev_priv = dev->dev_private;
	struct nouveau_bo *pb = chan->pushbuf_bo;
	struct nouveau_gpuobj *pushbuf = NULL;
	int ret;

	if (dev_priv->card_type >= NV_50) {
		ret = nouveau_gpuobj_dma_new(chan, NV_CLASS_DMA_IN_MEMORY, 0,
					     dev_priv->vm_end, NV_DMA_ACCESS_RO,
					     NV_DMA_TARGET_AGP, &pushbuf);
		chan->pushbuf_base = pb->bo.offset;
	} else
	if (pb->bo.mem.mem_type == TTM_PL_TT) {
		ret = nouveau_gpuobj_gart_dma_new(chan, 0,
						  dev_priv->gart_info.aper_size,
						  NV_DMA_ACCESS_RO, &pushbuf,
						  NULL);
		chan->pushbuf_base = pb->bo.mem.mm_node->start << PAGE_SHIFT;
	} else
	if (dev_priv->card_type != NV_04) {
		ret = nouveau_gpuobj_dma_new(chan, NV_CLASS_DMA_IN_MEMORY, 0,
					     dev_priv->fb_available_size,
					     NV_DMA_ACCESS_RO,
					     NV_DMA_TARGET_VIDMEM, &pushbuf);
		chan->pushbuf_base = pb->bo.mem.mm_node->start << PAGE_SHIFT;
	} else {
		/* NV04 cmdbuf hack, from original ddx.. not sure of it's
		 * exact reason for existing :)  PCI access to cmdbuf in
		 * VRAM.
		 */
		ret = nouveau_gpuobj_dma_new(chan, NV_CLASS_DMA_IN_MEMORY,
					     drm_get_resource_start(dev, 1),
					     dev_priv->fb_available_size,
					     NV_DMA_ACCESS_RO,
					     NV_DMA_TARGET_PCI, &pushbuf);
		chan->pushbuf_base = pb->bo.mem.mm_node->start << PAGE_SHIFT;
	}

	ret = nouveau_gpuobj_ref_add(dev, chan, 0, pushbuf, &chan->pushbuf);
	if (ret) {
		NV_ERROR(dev, "Error referencing pushbuf ctxdma: %d\n", ret);
		if (pushbuf != dev_priv->gart_info.sg_ctxdma)
			nouveau_gpuobj_del(dev, &pushbuf);
		return ret;
	}

	return 0;
}

static struct nouveau_bo *
nouveau_channel_user_pushbuf_alloc(struct drm_device *dev)
{
	struct nouveau_bo *pushbuf = NULL;
	int location, ret;

	if (nouveau_vram_pushbuf)
		location = TTM_PL_FLAG_VRAM;
	else
		location = TTM_PL_FLAG_TT;

	ret = nouveau_bo_new(dev, NULL, 65536, 0, location, 0, 0x0000, false,
			     true, &pushbuf);
	if (ret) {
		NV_ERROR(dev, "error allocating DMA push buffer: %d\n", ret);
		return NULL;
	}

	ret = nouveau_bo_pin(pushbuf, location);
	if (ret) {
		NV_ERROR(dev, "error pinning DMA push buffer: %d\n", ret);
		nouveau_bo_ref(NULL, &pushbuf);
		return NULL;
	}

	return pushbuf;
}

/* allocates and initializes a fifo for user space consumption */
int
nouveau_channel_alloc(struct drm_device *dev, struct nouveau_channel **chan_ret,
		      struct drm_file *file_priv,
		      uint32_t vram_handle, uint32_t tt_handle)
{
	struct drm_nouveau_private *dev_priv = dev->dev_private;
	struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
	struct nouveau_fifo_engine *pfifo = &dev_priv->engine.fifo;
	struct nouveau_channel *chan;
	int channel, user;
	int ret;

	/*
	 * Alright, here is the full story
	 * Nvidia cards have multiple hw fifo contexts (praise them for that,
	 * no complicated crash-prone context switches)
	 * We allocate a new context for each app and let it write to it
	 * directly (woo, full userspace command submission !)
	 * When there are no more contexts, you lost
	 */
	for (channel = 0; channel < pfifo->channels; channel++) {
		if (dev_priv->fifos[channel] == NULL)
			break;
	}

	/* no more fifos. you lost. */
	if (channel == pfifo->channels)
		return -EINVAL;

	dev_priv->fifos[channel] = kzalloc(sizeof(struct nouveau_channel),
					   GFP_KERNEL);
	if (!dev_priv->fifos[channel])
		return -ENOMEM;
	dev_priv->fifo_alloc_count++;
	chan = dev_priv->fifos[channel];
	INIT_LIST_HEAD(&chan->nvsw.vbl_wait);
	INIT_LIST_HEAD(&chan->fence.pending);
	chan->dev = dev;
	chan->id = channel;
	chan->file_priv = file_priv;
	chan->vram_handle = vram_handle;
	chan->gart_handle = tt_handle;

	NV_INFO(dev, "Allocating FIFO number %d\n", channel);

	/* Allocate DMA push buffer */
	chan->pushbuf_bo = nouveau_channel_user_pushbuf_alloc(dev);
	if (!chan->pushbuf_bo) {
		ret = -ENOMEM;
		NV_ERROR(dev, "pushbuf %d\n", ret);
		nouveau_channel_free(chan);
		return ret;
	}

	nouveau_dma_pre_init(chan);

	/* Locate channel's user control regs */
	if (dev_priv->card_type < NV_40)
		user = NV03_USER(channel);
	else
	if (dev_priv->card_type < NV_50)
		user = NV40_USER(channel);
	else
		user = NV50_USER(channel);

	chan->user = ioremap(pci_resource_start(dev->pdev, 0) + user,
								PAGE_SIZE);
	if (!chan->user) {
		NV_ERROR(dev, "ioremap of regs failed.\n");
		nouveau_channel_free(chan);
		return -ENOMEM;
	}
	chan->user_put = 0x40;
	chan->user_get = 0x44;

	/* Allocate space for per-channel fixed notifier memory */
	ret = nouveau_notifier_init_channel(chan);
	if (ret) {
		NV_ERROR(dev, "ntfy %d\n", ret);
		nouveau_channel_free(chan);
		return ret;
	}

	/* Setup channel's default objects */
	ret = nouveau_gpuobj_channel_init(chan, vram_handle, tt_handle);
	if (ret) {
		NV_ERROR(dev, "gpuobj %d\n", ret);
		nouveau_channel_free(chan);
		return ret;
	}

	/* Create a dma object for the push buffer */
	ret = nouveau_channel_pushbuf_ctxdma_init(chan);
	if (ret) {
		NV_ERROR(dev, "pbctxdma %d\n", ret);
		nouveau_channel_free(chan);
		return ret;
	}

	/* disable the fifo caches */
	pfifo->reassign(dev, false);

	/* Create a graphics context for new channel */
	ret = pgraph->create_context(chan);
	if (ret) {
		nouveau_channel_free(chan);
		return ret;
	}

	/* Construct inital RAMFC for new channel */
	ret = pfifo->create_context(chan);
	if (ret) {
		nouveau_channel_free(chan);
		return ret;
	}

	pfifo->reassign(dev, true);

	ret = nouveau_dma_init(chan);
	if (!ret)
		ret = nouveau_fence_init(chan);
	if (ret) {
		nouveau_channel_free(chan);
		return ret;
	}

	nouveau_debugfs_channel_init(chan);

	NV_INFO(dev, "%s: initialised FIFO %d\n", __func__, channel);
	*chan_ret = chan;
	return 0;
}

/* stops a fifo */
void
nouveau_channel_free(struct nouveau_channel *chan)
{
	struct drm_device *dev = chan->dev;
	struct drm_nouveau_private *dev_priv = dev->dev_private;
	struct nouveau_pgraph_engine *pgraph = &dev_priv->engine.graph;
	struct nouveau_fifo_engine *pfifo = &dev_priv->engine.fifo;
	unsigned long flags;
	int ret;

	NV_INFO(dev, "%s: freeing fifo %d\n", __func__, chan->id);

	nouveau_debugfs_channel_fini(chan);

	/* Give outstanding push buffers a chance to complete */
	spin_lock_irqsave(&chan->fence.lock, flags);
	nouveau_fence_update(chan);
	spin_unlock_irqrestore(&chan->fence.lock, flags);
	if (chan->fence.sequence != chan->fence.sequence_ack) {
		struct nouveau_fence *fence = NULL;

		ret = nouveau_fence_new(chan, &fence, true);
		if (ret == 0) {
			ret = nouveau_fence_wait(fence, NULL, false, false);
			nouveau_fence_unref((void *)&fence);
		}

		if (ret)
			NV_ERROR(dev, "Failed to idle channel %d.\n", chan->id);
	}

	/* Ensure all outstanding fences are signaled.  They should be if the
	 * above attempts at idling were OK, but if we failed this'll tell TTM
	 * we're done with the buffers.
	 */
	nouveau_fence_fini(chan);

	/* This will prevent pfifo from switching channels. */
	pfifo->reassign(dev, false);

	/* We want to give pgraph a chance to idle and get rid of all potential
	 * errors. We need to do this before the lock, otherwise the irq handler
	 * is unable to process them.
	 */
	if (pgraph->channel(dev) == chan)
		nouveau_wait_for_idle(dev);

	spin_lock_irqsave(&dev_priv->context_switch_lock, flags);

	pgraph->fifo_access(dev, false);
	if (pgraph->channel(dev) == chan)
		pgraph->unload_context(dev);
	pgraph->destroy_context(chan);
	pgraph->fifo_access(dev, true);

	if (pfifo->channel_id(dev) == chan->id) {
		pfifo->disable(dev);
		pfifo->unload_context(dev);
		pfifo->enable(dev);
	}
	pfifo->destroy_context(chan);

	pfifo->reassign(dev, true);

	spin_unlock_irqrestore(&dev_priv->context_switch_lock, flags);

	/* Release the channel's resources */
	nouveau_gpuobj_ref_del(dev, &chan->pushbuf);
	if (chan->pushbuf_bo) {
		nouveau_bo_unpin(chan->pushbuf_bo);
		nouveau_bo_ref(NULL, &chan->pushbuf_bo);
	}
	nouveau_gpuobj_channel_takedown(chan);
	nouveau_notifier_takedown_channel(chan);
	if (chan->user)
		iounmap(chan->user);

	dev_priv->fifos[chan->id] = NULL;
	dev_priv->fifo_alloc_count--;
	kfree(chan);
}

/* cleans up all the fifos from file_priv */
void
nouveau_channel_cleanup(struct drm_device *dev, struct drm_file *file_priv)
{
	struct drm_nouveau_private *dev_priv = dev->dev_private;
	struct nouveau_engine *engine = &dev_priv->engine;
	int i;

	NV_DEBUG(dev, "clearing FIFO enables from file_priv\n");
	for (i = 0; i < engine->fifo.channels; i++) {
		struct nouveau_channel *chan = dev_priv->fifos[i];

		if (chan && chan->file_priv == file_priv)
			nouveau_channel_free(chan);
	}
}

int
nouveau_channel_owner(struct drm_device *dev, struct drm_file *file_priv,
		      int channel)
{
	struct drm_nouveau_private *dev_priv = dev->dev_private;
	struct nouveau_engine *engine = &dev_priv->engine;

	if (channel >= engine->fifo.channels)
		return 0;
	if (dev_priv->fifos[channel] == NULL)
		return 0;

	return (dev_priv->fifos[channel]->file_priv == file_priv);
}

/***********************************
 * ioctls wrapping the functions
 ***********************************/

static int
nouveau_ioctl_fifo_alloc(struct drm_device *dev, void *data,
			 struct drm_file *file_priv)
{
	struct drm_nouveau_private *dev_priv = dev->dev_private;
	struct drm_nouveau_channel_alloc *init = data;
	struct nouveau_channel *chan;
	int ret;

	NOUVEAU_CHECK_INITIALISED_WITH_RETURN;

	if (dev_priv->engine.graph.accel_blocked)
		return -ENODEV;

	if (init->fb_ctxdma_handle == ~0 || init->tt_ctxdma_handle == ~0)
		return -EINVAL;

	ret = nouveau_channel_alloc(dev, &chan, file_priv,
				    init->fb_ctxdma_handle,
				    init->tt_ctxdma_handle);
	if (ret)
		return ret;
	init->channel  = chan->id;

	if (chan->dma.ib_max)
		init->pushbuf_domains = NOUVEAU_GEM_DOMAIN_VRAM |
					NOUVEAU_GEM_DOMAIN_GART;
	else if (chan->pushbuf_bo->bo.mem.mem_type == TTM_PL_VRAM)
		init->pushbuf_domains = NOUVEAU_GEM_DOMAIN_VRAM;
	else
		init->pushbuf_domains = NOUVEAU_GEM_DOMAIN_GART;

	init->subchan[0].handle = NvM2MF;
	if (dev_priv->card_type < NV_50)
		init->subchan[0].grclass = 0x0039;
	else
		init->subchan[0].grclass = 0x5039;
	init->subchan[1].handle = NvSw;
	init->subchan[1].grclass = NV_SW;
	init->nr_subchan = 2;

	/* Named memory object area */
	ret = drm_gem_handle_create(file_priv, chan->notifier_bo->gem,
				    &init->notifier_handle);
	if (ret) {
		nouveau_channel_free(chan);
		return ret;
	}

	return 0;
}

static int
nouveau_ioctl_fifo_free(struct drm_device *dev, void *data,
			struct drm_file *file_priv)
{
	struct drm_nouveau_channel_free *cfree = data;
	struct nouveau_channel *chan;

	NOUVEAU_CHECK_INITIALISED_WITH_RETURN;
	NOUVEAU_GET_USER_CHANNEL_WITH_RETURN(cfree->channel, file_priv, chan);

	nouveau_channel_free(chan);
	return 0;
}

/***********************************
 * finally, the ioctl table
 ***********************************/

struct drm_ioctl_desc nouveau_ioctls[] = {
	DRM_IOCTL_DEF(DRM_NOUVEAU_GETPARAM, nouveau_ioctl_getparam, DRM_AUTH),
	DRM_IOCTL_DEF(DRM_NOUVEAU_SETPARAM, nouveau_ioctl_setparam, DRM_AUTH|DRM_MASTER|DRM_ROOT_ONLY),
	DRM_IOCTL_DEF(DRM_NOUVEAU_CHANNEL_ALLOC, nouveau_ioctl_fifo_alloc, DRM_AUTH),
	DRM_IOCTL_DEF(DRM_NOUVEAU_CHANNEL_FREE, nouveau_ioctl_fifo_free, DRM_AUTH),
	DRM_IOCTL_DEF(DRM_NOUVEAU_GROBJ_ALLOC, nouveau_ioctl_grobj_alloc, DRM_AUTH),
	DRM_IOCTL_DEF(DRM_NOUVEAU_NOTIFIEROBJ_ALLOC, nouveau_ioctl_notifier_alloc, DRM_AUTH),
	DRM_IOCTL_DEF(DRM_NOUVEAU_GPUOBJ_FREE, nouveau_ioctl_gpuobj_free, DRM_AUTH),
	DRM_IOCTL_DEF(DRM_NOUVEAU_GEM_NEW, nouveau_gem_ioctl_new, DRM_AUTH),
	DRM_IOCTL_DEF(DRM_NOUVEAU_GEM_PUSHBUF, nouveau_gem_ioctl_pushbuf, DRM_AUTH),
	DRM_IOCTL_DEF(DRM_NOUVEAU_GEM_CPU_PREP, nouveau_gem_ioctl_cpu_prep, DRM_AUTH),
	DRM_IOCTL_DEF(DRM_NOUVEAU_GEM_CPU_FINI, nouveau_gem_ioctl_cpu_fini, DRM_AUTH),
	DRM_IOCTL_DEF(DRM_NOUVEAU_GEM_INFO, nouveau_gem_ioctl_info, DRM_AUTH),
};

int nouveau_max_ioctl = DRM_ARRAY_SIZE(nouveau_ioctls);
OpenPOWER on IntegriCloud