summaryrefslogtreecommitdiffstats
path: root/drivers/staging/gma500/gtt.c
blob: e770bd190a5c8a2efda446339805869944528a81 (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
/*
 * Copyright (c) 2007, Intel Corporation.
 * All Rights Reserved.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
 * version 2, as published by the Free Software Foundation.
 *
 * This program is distributed in the hope it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
 * more details.
 *
 * You should have received a copy of the GNU General Public License along with
 * this program; if not, write to the Free Software Foundation, Inc.,
 * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
 *
 * Authors: Thomas Hellstrom <thomas-at-tungstengraphics.com>
 *	    Alan Cox <alan@linux.intel.com>
 */

#include <drm/drmP.h>
#include "psb_drv.h"


/*
 *	GTT resource allocator - manage page mappings in GTT space
 */

/**
 *	psb_gtt_mask_pte	-	generate GTT pte entry
 *	@pfn: page number to encode
 *	@type: type of memory in the GTT
 *
 *	Set the GTT entry for the appropriate memory type.
 */
static inline uint32_t psb_gtt_mask_pte(uint32_t pfn, int type)
{
	uint32_t mask = PSB_PTE_VALID;

	if (type & PSB_MMU_CACHED_MEMORY)
		mask |= PSB_PTE_CACHED;
	if (type & PSB_MMU_RO_MEMORY)
		mask |= PSB_PTE_RO;
	if (type & PSB_MMU_WO_MEMORY)
		mask |= PSB_PTE_WO;

	return (pfn << PAGE_SHIFT) | mask;
}

/**
 *	psb_gtt_entry		-	find the GTT entries for a gtt_range
 *	@dev: our DRM device
 *	@r: our GTT range
 *
 *	Given a gtt_range object return the GTT offset of the page table
 *	entries for this gtt_range
 */
u32 *psb_gtt_entry(struct drm_device *dev, struct gtt_range *r)
{
	struct drm_psb_private *dev_priv = dev->dev_private;
	unsigned long offset;

	offset = r->resource.start - dev_priv->gtt_mem->start;

	return dev_priv->gtt_map + (offset >> PAGE_SHIFT);
}

/**
 *	psb_gtt_insert	-	put an object into the GTT
 *	@dev: our DRM device
 *	@r: our GTT range
 *
 *	Take our preallocated GTT range and insert the GEM object into
 *	the GTT. This is protected via the gtt mutex which the caller
 *	must hold.
 */
static int psb_gtt_insert(struct drm_device *dev, struct gtt_range *r)
{
	u32 *gtt_slot, pte;
	struct page **pages;
	int i;

	if (r->pages == NULL) {
		WARN_ON(1);
		return -EINVAL;
	}

	WARN_ON(r->stolen);	/* refcount these maybe ? */

	gtt_slot = psb_gtt_entry(dev, r);
	pages = r->pages;

	/* Make sure changes are visible to the GPU */
	set_pages_array_uc(pages, r->npage);

	/* Write our page entries into the GTT itself */
	for (i = r->roll; i < r->npage; i++) {
		pte = psb_gtt_mask_pte(page_to_pfn(r->pages[i]), 0);
		iowrite32(pte, gtt_slot++);
	}
	for (i = 0; i < r->roll; i++) {
		pte = psb_gtt_mask_pte(page_to_pfn(r->pages[i]), 0);
		iowrite32(pte, gtt_slot++);
	}
	/* Make sure all the entries are set before we return */
	ioread32(gtt_slot - 1);

	return 0;
}

/**
 *	psb_gtt_remove	-	remove an object from the GTT
 *	@dev: our DRM device
 *	@r: our GTT range
 *
 *	Remove a preallocated GTT range from the GTT. Overwrite all the
 *	page table entries with the dummy page. This is protected via the gtt
 *	mutex which the caller must hold.
 */
static void psb_gtt_remove(struct drm_device *dev, struct gtt_range *r)
{
	struct drm_psb_private *dev_priv = dev->dev_private;
	u32 *gtt_slot, pte;
	int i;

	WARN_ON(r->stolen);

	gtt_slot = psb_gtt_entry(dev, r);
	pte = psb_gtt_mask_pte(page_to_pfn(dev_priv->scratch_page), 0);

	for (i = 0; i < r->npage; i++)
		iowrite32(pte, gtt_slot++);
	ioread32(gtt_slot - 1);
	set_pages_array_wb(r->pages, r->npage);
}

/**
 *	psb_gtt_roll	-	set scrolling position
 *	@dev: our DRM device
 *	@r: the gtt mapping we are using
 *	@roll: roll offset
 *
 *	Roll an existing pinned mapping by moving the pages through the GTT.
 *	This allows us to implement hardware scrolling on the consoles without
 *	a 2D engine
 */
void psb_gtt_roll(struct drm_device *dev, struct gtt_range *r, int roll)
{
	u32 *gtt_slot, pte;
	int i;

	if (roll >= r->npage) {
		WARN_ON(1);
		return;
	}

	r->roll = roll;

	/* Not currently in the GTT - no worry we will write the mapping at
	   the right position when it gets pinned */
	if (!r->stolen && !r->in_gart)
		return;

	gtt_slot = psb_gtt_entry(dev, r);

	for (i = r->roll; i < r->npage; i++) {
		pte = psb_gtt_mask_pte(page_to_pfn(r->pages[i]), 0);
		iowrite32(pte, gtt_slot++);
	}
	for (i = 0; i < r->roll; i++) {
		pte = psb_gtt_mask_pte(page_to_pfn(r->pages[i]), 0);
		iowrite32(pte, gtt_slot++);
	}
	ioread32(gtt_slot - 1);
}

/**
 *	psb_gtt_attach_pages	-	attach and pin GEM pages
 *	@gt: the gtt range
 *
 *	Pin and build an in kernel list of the pages that back our GEM object.
 *	While we hold this the pages cannot be swapped out. This is protected
 *	via the gtt mutex which the caller must hold.
 */
static int psb_gtt_attach_pages(struct gtt_range *gt)
{
	struct inode *inode;
	struct address_space *mapping;
	int i;
	struct page *p;
	int pages = gt->gem.size / PAGE_SIZE;

	WARN_ON(gt->pages);

	/* This is the shared memory object that backs the GEM resource */
	inode = gt->gem.filp->f_path.dentry->d_inode;
	mapping = inode->i_mapping;

	gt->pages = kmalloc(pages * sizeof(struct page *), GFP_KERNEL);
	if (gt->pages == NULL)
		return -ENOMEM;
	gt->npage = pages;

	for (i = 0; i < pages; i++) {
		/* FIXME: needs updating as per mail from Hugh Dickins */
		p = read_cache_page_gfp(mapping, i,
					__GFP_COLD | GFP_KERNEL);
		if (IS_ERR(p))
			goto err;
		gt->pages[i] = p;
	}
	return 0;

err:
	while (i--)
		page_cache_release(gt->pages[i]);
	kfree(gt->pages);
	gt->pages = NULL;
	return PTR_ERR(p);
}

/**
 *	psb_gtt_detach_pages	-	attach and pin GEM pages
 *	@gt: the gtt range
 *
 *	Undo the effect of psb_gtt_attach_pages. At this point the pages
 *	must have been removed from the GTT as they could now be paged out
 *	and move bus address. This is protected via the gtt mutex which the
 *	caller must hold.
 */
static void psb_gtt_detach_pages(struct gtt_range *gt)
{
	int i;
	for (i = 0; i < gt->npage; i++) {
		/* FIXME: do we need to force dirty */
		set_page_dirty(gt->pages[i]);
		page_cache_release(gt->pages[i]);
	}
	kfree(gt->pages);
	gt->pages = NULL;
}

/**
 *	psb_gtt_pin		-	pin pages into the GTT
 *	@gt: range to pin
 *
 *	Pin a set of pages into the GTT. The pins are refcounted so that
 *	multiple pins need multiple unpins to undo.
 *
 *	Non GEM backed objects treat this as a no-op as they are always GTT
 *	backed objects.
 */
int psb_gtt_pin(struct gtt_range *gt)
{
	int ret = 0;
	struct drm_device *dev = gt->gem.dev;
	struct drm_psb_private *dev_priv = dev->dev_private;

	mutex_lock(&dev_priv->gtt_mutex);

	if (gt->in_gart == 0 && gt->stolen == 0) {
		ret = psb_gtt_attach_pages(gt);
		if (ret < 0)
			goto out;
		ret = psb_gtt_insert(dev, gt);
		if (ret < 0) {
			psb_gtt_detach_pages(gt);
			goto out;
		}
	}
	gt->in_gart++;
out:
	mutex_unlock(&dev_priv->gtt_mutex);
	return ret;
}

/**
 *	psb_gtt_unpin		-	Drop a GTT pin requirement
 *	@gt: range to pin
 *
 *	Undoes the effect of psb_gtt_pin. On the last drop the GEM object
 *	will be removed from the GTT which will also drop the page references
 *	and allow the VM to clean up or page stuff.
 *
 *	Non GEM backed objects treat this as a no-op as they are always GTT
 *	backed objects.
 */
void psb_gtt_unpin(struct gtt_range *gt)
{
	struct drm_device *dev = gt->gem.dev;
	struct drm_psb_private *dev_priv = dev->dev_private;

	mutex_lock(&dev_priv->gtt_mutex);

	WARN_ON(!gt->in_gart);

	gt->in_gart--;
	if (gt->in_gart == 0 && gt->stolen == 0) {
		psb_gtt_remove(dev, gt);
		psb_gtt_detach_pages(gt);
	}
	mutex_unlock(&dev_priv->gtt_mutex);
}

/*
 *	GTT resource allocator - allocate and manage GTT address space
 */

/**
 *	psb_gtt_alloc_range	-	allocate GTT address space
 *	@dev: Our DRM device
 *	@len: length (bytes) of address space required
 *	@name: resource name
 *	@backed: resource should be backed by stolen pages
 *
 *	Ask the kernel core to find us a suitable range of addresses
 *	to use for a GTT mapping.
 *
 *	Returns a gtt_range structure describing the object, or NULL on
 *	error. On successful return the resource is both allocated and marked
 *	as in use.
 */
struct gtt_range *psb_gtt_alloc_range(struct drm_device *dev, int len,
						const char *name, int backed)
{
	struct drm_psb_private *dev_priv = dev->dev_private;
	struct gtt_range *gt;
	struct resource *r = dev_priv->gtt_mem;
	int ret;
	unsigned long start, end;

	if (backed) {
		/* The start of the GTT is the stolen pages */
		start = r->start;
		end = r->start + dev_priv->gtt.stolen_size - 1;
	} else {
		/* The rest we will use for GEM backed objects */
		start = r->start + dev_priv->gtt.stolen_size;
		end = r->end;
	}

	gt = kzalloc(sizeof(struct gtt_range), GFP_KERNEL);
	if (gt == NULL)
		return NULL;
	gt->resource.name = name;
	gt->stolen = backed;
	gt->in_gart = backed;
	gt->roll = 0;
	/* Ensure this is set for non GEM objects */
	gt->gem.dev = dev;
	ret = allocate_resource(dev_priv->gtt_mem, &gt->resource,
				len, start, end, PAGE_SIZE, NULL, NULL);
	if (ret == 0) {
		gt->offset = gt->resource.start - r->start;
		return gt;
	}
	kfree(gt);
	return NULL;
}

/**
 *	psb_gtt_free_range	-	release GTT address space
 *	@dev: our DRM device
 *	@gt: a mapping created with psb_gtt_alloc_range
 *
 *	Release a resource that was allocated with psb_gtt_alloc_range. If the
 *	object has been pinned by mmap users we clean this up here currently.
 */
void psb_gtt_free_range(struct drm_device *dev, struct gtt_range *gt)
{
	/* Undo the mmap pin if we are destroying the object */
	if (gt->mmapping) {
		psb_gtt_unpin(gt);
		gt->mmapping = 0;
	}
	WARN_ON(gt->in_gart && !gt->stolen);
	release_resource(&gt->resource);
	kfree(gt);
}

void psb_gtt_alloc(struct drm_device *dev)
{
	struct drm_psb_private *dev_priv = dev->dev_private;
	init_rwsem(&dev_priv->gtt.sem);
}

void psb_gtt_takedown(struct drm_device *dev)
{
	struct drm_psb_private *dev_priv = dev->dev_private;

	if (dev_priv->gtt_map) {
		iounmap(dev_priv->gtt_map);
		dev_priv->gtt_map = NULL;
	}
	if (dev_priv->gtt_initialized) {
		pci_write_config_word(dev->pdev, PSB_GMCH_CTRL,
				      dev_priv->gmch_ctrl);
		PSB_WVDC32(dev_priv->pge_ctl, PSB_PGETBL_CTL);
		(void) PSB_RVDC32(PSB_PGETBL_CTL);
	}
	if (dev_priv->vram_addr)
		iounmap(dev_priv->gtt_map);
}

int psb_gtt_init(struct drm_device *dev, int resume)
{
	struct drm_psb_private *dev_priv = dev->dev_private;
	unsigned gtt_pages;
	unsigned long stolen_size, vram_stolen_size;
	unsigned i, num_pages;
	unsigned pfn_base;
	uint32_t vram_pages;
	uint32_t dvmt_mode = 0;
	struct psb_gtt *pg;

	int ret = 0;
	uint32_t pte;

	mutex_init(&dev_priv->gtt_mutex);

	psb_gtt_alloc(dev);
	pg = &dev_priv->gtt;

	/* Enable the GTT */
	pci_read_config_word(dev->pdev, PSB_GMCH_CTRL, &dev_priv->gmch_ctrl);
	pci_write_config_word(dev->pdev, PSB_GMCH_CTRL,
			      dev_priv->gmch_ctrl | _PSB_GMCH_ENABLED);

	dev_priv->pge_ctl = PSB_RVDC32(PSB_PGETBL_CTL);
	PSB_WVDC32(dev_priv->pge_ctl | _PSB_PGETBL_ENABLED, PSB_PGETBL_CTL);
	(void) PSB_RVDC32(PSB_PGETBL_CTL);

	/* The root resource we allocate address space from */
	dev_priv->gtt_initialized = 1;

	pg->gtt_phys_start = dev_priv->pge_ctl & PAGE_MASK;

	/*
	 *	The video mmu has a hw bug when accessing 0x0D0000000.
	 *	Make gatt start at 0x0e000,0000. This doesn't actually
	 *	matter for us but may do if the video acceleration ever
	 *	gets opened up.
	 */
	pg->mmu_gatt_start = 0xE0000000;

	pg->gtt_start = pci_resource_start(dev->pdev, PSB_GTT_RESOURCE);
	gtt_pages = pci_resource_len(dev->pdev, PSB_GTT_RESOURCE)
								>> PAGE_SHIFT;
	/* Some CDV firmware doesn't report this currently. In which case the
	   system has 64 gtt pages */
	if (pg->gtt_start == 0 || gtt_pages == 0) {
		dev_err(dev->dev, "GTT PCI BAR not initialized.\n");
		gtt_pages = 64;
		pg->gtt_start = dev_priv->pge_ctl;
	}

	pg->gatt_start = pci_resource_start(dev->pdev, PSB_GATT_RESOURCE);
	pg->gatt_pages = pci_resource_len(dev->pdev, PSB_GATT_RESOURCE)
								>> PAGE_SHIFT;
	dev_priv->gtt_mem = &dev->pdev->resource[PSB_GATT_RESOURCE];

	if (pg->gatt_pages == 0 || pg->gatt_start == 0) {
		static struct resource fudge;	/* Preferably peppermint */
		/* This can occur on CDV SDV systems. Fudge it in this case.
		   We really don't care what imaginary space is being allocated
		   at this point */
		dev_err(dev->dev, "GATT PCI BAR not initialized.\n");
		pg->gatt_start = 0x40000000;
		pg->gatt_pages = (128 * 1024 * 1024) >> PAGE_SHIFT;
		/* This is a little confusing but in fact the GTT is providing
		   a view from the GPU into memory and not vice versa. As such
		   this is really allocating space that is not the same as the
		   CPU address space on CDV */
		fudge.start = 0x40000000;
		fudge.end = 0x40000000 + 128 * 1024 * 1024 - 1;
		fudge.name = "fudge";
		fudge.flags = IORESOURCE_MEM;
		dev_priv->gtt_mem = &fudge;
	}

	pci_read_config_dword(dev->pdev, PSB_BSM, &dev_priv->stolen_base);
	vram_stolen_size = pg->gtt_phys_start - dev_priv->stolen_base
								- PAGE_SIZE;

	stolen_size = vram_stolen_size;

	printk(KERN_INFO "Stolen memory information\n");
	printk(KERN_INFO "       base in RAM: 0x%x\n", dev_priv->stolen_base);
	printk(KERN_INFO "       size: %luK, calculated by (GTT RAM base) - (Stolen base), seems wrong\n",
		vram_stolen_size/1024);
	dvmt_mode = (dev_priv->gmch_ctrl >> 4) & 0x7;
	printk(KERN_INFO "      the correct size should be: %dM(dvmt mode=%d)\n",
		(dvmt_mode == 1) ? 1 : (2 << (dvmt_mode - 1)), dvmt_mode);

	if (resume && (gtt_pages != pg->gtt_pages) &&
	    (stolen_size != pg->stolen_size)) {
		dev_err(dev->dev, "GTT resume error.\n");
		ret = -EINVAL;
		goto out_err;
	}

	pg->gtt_pages = gtt_pages;
	pg->stolen_size = stolen_size;
	dev_priv->vram_stolen_size = vram_stolen_size;

	/*
	 *	Map the GTT and the stolen memory area
	 */
	dev_priv->gtt_map = ioremap_nocache(pg->gtt_phys_start,
						gtt_pages << PAGE_SHIFT);
	if (!dev_priv->gtt_map) {
		dev_err(dev->dev, "Failure to map gtt.\n");
		ret = -ENOMEM;
		goto out_err;
	}

	dev_priv->vram_addr = ioremap_wc(dev_priv->stolen_base, stolen_size);
	if (!dev_priv->vram_addr) {
		dev_err(dev->dev, "Failure to map stolen base.\n");
		ret = -ENOMEM;
		goto out_err;
	}

	/*
	 * Insert vram stolen pages into the GTT
	 */

	pfn_base = dev_priv->stolen_base >> PAGE_SHIFT;
	vram_pages = num_pages = vram_stolen_size >> PAGE_SHIFT;
	printk(KERN_INFO"Set up %d stolen pages starting at 0x%08x, GTT offset %dK\n",
		num_pages, pfn_base << PAGE_SHIFT, 0);
	for (i = 0; i < num_pages; ++i) {
		pte = psb_gtt_mask_pte(pfn_base + i, 0);
		iowrite32(pte, dev_priv->gtt_map + i);
	}

	/*
	 * Init rest of GTT to the scratch page to avoid accidents or scribbles
	 */

	pfn_base = page_to_pfn(dev_priv->scratch_page);
	pte = psb_gtt_mask_pte(pfn_base, 0);
	for (; i < gtt_pages; ++i)
		iowrite32(pte, dev_priv->gtt_map + i);

	(void) ioread32(dev_priv->gtt_map + i - 1);
	return 0;

out_err:
	psb_gtt_takedown(dev);
	return ret;
}