Free Electrons

Embedded Linux Experts

  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
/*
 * Copyright © 2008 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:
 *    Eric Anholt <eric@anholt.net>
 *
 */

#include "linux/string.h"
#include "linux/bitops.h"
#include "drmP.h"
#include "drm.h"
#include "i915_drm.h"
#include "i915_drv.h"

/** @file i915_gem_tiling.c
 *
 * Support for managing tiling state of buffer objects.
 *
 * The idea behind tiling is to increase cache hit rates by rearranging
 * pixel data so that a group of pixel accesses are in the same cacheline.
 * Performance improvement from doing this on the back/depth buffer are on
 * the order of 30%.
 *
 * Intel architectures make this somewhat more complicated, though, by
 * adjustments made to addressing of data when the memory is in interleaved
 * mode (matched pairs of DIMMS) to improve memory bandwidth.
 * For interleaved memory, the CPU sends every sequential 64 bytes
 * to an alternate memory channel so it can get the bandwidth from both.
 *
 * The GPU also rearranges its accesses for increased bandwidth to interleaved
 * memory, and it matches what the CPU does for non-tiled.  However, when tiled
 * it does it a little differently, since one walks addresses not just in the
 * X direction but also Y.  So, along with alternating channels when bit
 * 6 of the address flips, it also alternates when other bits flip --  Bits 9
 * (every 512 bytes, an X tile scanline) and 10 (every two X tile scanlines)
 * are common to both the 915 and 965-class hardware.
 *
 * The CPU also sometimes XORs in higher bits as well, to improve
 * bandwidth doing strided access like we do so frequently in graphics.  This
 * is called "Channel XOR Randomization" in the MCH documentation.  The result
 * is that the CPU is XORing in either bit 11 or bit 17 to bit 6 of its address
 * decode.
 *
 * All of this bit 6 XORing has an effect on our memory management,
 * as we need to make sure that the 3d driver can correctly address object
 * contents.
 *
 * If we don't have interleaved memory, all tiling is safe and no swizzling is
 * required.
 *
 * When bit 17 is XORed in, we simply refuse to tile at all.  Bit
 * 17 is not just a page offset, so as we page an objet out and back in,
 * individual pages in it will have different bit 17 addresses, resulting in
 * each 64 bytes being swapped with its neighbor!
 *
 * Otherwise, if interleaved, we have to tell the 3d driver what the address
 * swizzling it needs to do is, since it's writing with the CPU to the pages
 * (bit 6 and potentially bit 11 XORed in), and the GPU is reading from the
 * pages (bit 6, 9, and 10 XORed in), resulting in a cumulative bit swizzling
 * required by the CPU of XORing in bit 6, 9, 10, and potentially 11, in order
 * to match what the GPU expects.
 */

/**
 * Detects bit 6 swizzling of address lookup between IGD access and CPU
 * access through main memory.
 */
void
i915_gem_detect_bit_6_swizzle(struct drm_device *dev)
{
	drm_i915_private_t *dev_priv = dev->dev_private;
	uint32_t swizzle_x = I915_BIT_6_SWIZZLE_UNKNOWN;
	uint32_t swizzle_y = I915_BIT_6_SWIZZLE_UNKNOWN;

	if (INTEL_INFO(dev)->gen >= 6) {
		uint32_t dimm_c0, dimm_c1;
		dimm_c0 = I915_READ(MAD_DIMM_C0);
		dimm_c1 = I915_READ(MAD_DIMM_C1);
		dimm_c0 &= MAD_DIMM_A_SIZE_MASK | MAD_DIMM_B_SIZE_MASK;
		dimm_c1 &= MAD_DIMM_A_SIZE_MASK | MAD_DIMM_B_SIZE_MASK;
		/* Enable swizzling when the channels are populated with
		 * identically sized dimms. We don't need to check the 3rd
		 * channel because no cpu with gpu attached ships in that
		 * configuration. Also, swizzling only makes sense for 2
		 * channels anyway. */
		if (dimm_c0 == dimm_c1) {
			swizzle_x = I915_BIT_6_SWIZZLE_9_10;
			swizzle_y = I915_BIT_6_SWIZZLE_9;
		} else {
			swizzle_x = I915_BIT_6_SWIZZLE_NONE;
			swizzle_y = I915_BIT_6_SWIZZLE_NONE;
		}
	} else if (IS_GEN5(dev)) {
		/* On Ironlake whatever DRAM config, GPU always do
		 * same swizzling setup.
		 */
		swizzle_x = I915_BIT_6_SWIZZLE_9_10;
		swizzle_y = I915_BIT_6_SWIZZLE_9;
	} else if (IS_GEN2(dev)) {
		/* As far as we know, the 865 doesn't have these bit 6
		 * swizzling issues.
		 */
		swizzle_x = I915_BIT_6_SWIZZLE_NONE;
		swizzle_y = I915_BIT_6_SWIZZLE_NONE;
	} else if (IS_MOBILE(dev) || (IS_GEN3(dev) && !IS_G33(dev))) {
		uint32_t dcc;

		/* On 9xx chipsets, channel interleave by the CPU is
		 * determined by DCC.  For single-channel, neither the CPU
		 * nor the GPU do swizzling.  For dual channel interleaved,
		 * the GPU's interleave is bit 9 and 10 for X tiled, and bit
		 * 9 for Y tiled.  The CPU's interleave is independent, and
		 * can be based on either bit 11 (haven't seen this yet) or
		 * bit 17 (common).
		 */
		dcc = I915_READ(DCC);
		switch (dcc & DCC_ADDRESSING_MODE_MASK) {
		case DCC_ADDRESSING_MODE_SINGLE_CHANNEL:
		case DCC_ADDRESSING_MODE_DUAL_CHANNEL_ASYMMETRIC:
			swizzle_x = I915_BIT_6_SWIZZLE_NONE;
			swizzle_y = I915_BIT_6_SWIZZLE_NONE;
			break;
		case DCC_ADDRESSING_MODE_DUAL_CHANNEL_INTERLEAVED:
			if (dcc & DCC_CHANNEL_XOR_DISABLE) {
				/* This is the base swizzling by the GPU for
				 * tiled buffers.
				 */
				swizzle_x = I915_BIT_6_SWIZZLE_9_10;
				swizzle_y = I915_BIT_6_SWIZZLE_9;
			} else if ((dcc & DCC_CHANNEL_XOR_BIT_17) == 0) {
				/* Bit 11 swizzling by the CPU in addition. */
				swizzle_x = I915_BIT_6_SWIZZLE_9_10_11;
				swizzle_y = I915_BIT_6_SWIZZLE_9_11;
			} else {
				/* Bit 17 swizzling by the CPU in addition. */
				swizzle_x = I915_BIT_6_SWIZZLE_9_10_17;
				swizzle_y = I915_BIT_6_SWIZZLE_9_17;
			}
			break;
		}
		if (dcc == 0xffffffff) {
			DRM_ERROR("Couldn't read from MCHBAR.  "
				  "Disabling tiling.\n");
			swizzle_x = I915_BIT_6_SWIZZLE_UNKNOWN;
			swizzle_y = I915_BIT_6_SWIZZLE_UNKNOWN;
		}
	} else {
		/* The 965, G33, and newer, have a very flexible memory
		 * configuration.  It will enable dual-channel mode
		 * (interleaving) on as much memory as it can, and the GPU
		 * will additionally sometimes enable different bit 6
		 * swizzling for tiled objects from the CPU.
		 *
		 * Here's what I found on the G965:
		 *    slot fill         memory size  swizzling
		 * 0A   0B   1A   1B    1-ch   2-ch
		 * 512  0    0    0     512    0     O
		 * 512  0    512  0     16     1008  X
		 * 512  0    0    512   16     1008  X
		 * 0    512  0    512   16     1008  X
		 * 1024 1024 1024 0     2048   1024  O
		 *
		 * We could probably detect this based on either the DRB
		 * matching, which was the case for the swizzling required in
		 * the table above, or from the 1-ch value being less than
		 * the minimum size of a rank.
		 */
		if (I915_READ16(C0DRB3) != I915_READ16(C1DRB3)) {
			swizzle_x = I915_BIT_6_SWIZZLE_NONE;
			swizzle_y = I915_BIT_6_SWIZZLE_NONE;
		} else {
			swizzle_x = I915_BIT_6_SWIZZLE_9_10;
			swizzle_y = I915_BIT_6_SWIZZLE_9;
		}
	}

	dev_priv->mm.bit_6_swizzle_x = swizzle_x;
	dev_priv->mm.bit_6_swizzle_y = swizzle_y;
}

/* Check pitch constriants for all chips & tiling formats */
static bool
i915_tiling_ok(struct drm_device *dev, int stride, int size, int tiling_mode)
{
	int tile_width;

	/* Linear is always fine */
	if (tiling_mode == I915_TILING_NONE)
		return true;

	if (IS_GEN2(dev) ||
	    (tiling_mode == I915_TILING_Y && HAS_128_BYTE_Y_TILING(dev)))
		tile_width = 128;
	else
		tile_width = 512;

	/* check maximum stride & object size */
	if (INTEL_INFO(dev)->gen >= 4) {
		/* i965 stores the end address of the gtt mapping in the fence
		 * reg, so dont bother to check the size */
		if (stride / 128 > I965_FENCE_MAX_PITCH_VAL)
			return false;
	} else {
		if (stride > 8192)
			return false;

		if (IS_GEN3(dev)) {
			if (size > I830_FENCE_MAX_SIZE_VAL << 20)
				return false;
		} else {
			if (size > I830_FENCE_MAX_SIZE_VAL << 19)
				return false;
		}
	}

	/* 965+ just needs multiples of tile width */
	if (INTEL_INFO(dev)->gen >= 4) {
		if (stride & (tile_width - 1))
			return false;
		return true;
	}

	/* Pre-965 needs power of two tile widths */
	if (stride < tile_width)
		return false;

	if (stride & (stride - 1))
		return false;

	return true;
}

/* Is the current GTT allocation valid for the change in tiling? */
static bool
i915_gem_object_fence_ok(struct drm_i915_gem_object *obj, int tiling_mode)
{
	u32 size;

	if (tiling_mode == I915_TILING_NONE)
		return true;

	if (INTEL_INFO(obj->base.dev)->gen >= 4)
		return true;

	if (INTEL_INFO(obj->base.dev)->gen == 3) {
		if (obj->gtt_offset & ~I915_FENCE_START_MASK)
			return false;
	} else {
		if (obj->gtt_offset & ~I830_FENCE_START_MASK)
			return false;
	}

	/*
	 * Previous chips need to be aligned to the size of the smallest
	 * fence register that can contain the object.
	 */
	if (INTEL_INFO(obj->base.dev)->gen == 3)
		size = 1024*1024;
	else
		size = 512*1024;

	while (size < obj->base.size)
		size <<= 1;

	if (obj->gtt_space->size != size)
		return false;

	if (obj->gtt_offset & (size - 1))
		return false;

	return true;
}

/**
 * Sets the tiling mode of an object, returning the required swizzling of
 * bit 6 of addresses in the object.
 */
int
i915_gem_set_tiling(struct drm_device *dev, void *data,
		   struct drm_file *file)
{
	struct drm_i915_gem_set_tiling *args = data;
	drm_i915_private_t *dev_priv = dev->dev_private;
	struct drm_i915_gem_object *obj;
	int ret = 0;

	obj = to_intel_bo(drm_gem_object_lookup(dev, file, args->handle));
	if (&obj->base == NULL)
		return -ENOENT;

	if (!i915_tiling_ok(dev,
			    args->stride, obj->base.size, args->tiling_mode)) {
		drm_gem_object_unreference_unlocked(&obj->base);
		return -EINVAL;
	}

	if (obj->pin_count) {
		drm_gem_object_unreference_unlocked(&obj->base);
		return -EBUSY;
	}

	if (args->tiling_mode == I915_TILING_NONE) {
		args->swizzle_mode = I915_BIT_6_SWIZZLE_NONE;
		args->stride = 0;
	} else {
		if (args->tiling_mode == I915_TILING_X)
			args->swizzle_mode = dev_priv->mm.bit_6_swizzle_x;
		else
			args->swizzle_mode = dev_priv->mm.bit_6_swizzle_y;

		/* Hide bit 17 swizzling from the user.  This prevents old Mesa
		 * from aborting the application on sw fallbacks to bit 17,
		 * and we use the pread/pwrite bit17 paths to swizzle for it.
		 * If there was a user that was relying on the swizzle
		 * information for drm_intel_bo_map()ed reads/writes this would
		 * break it, but we don't have any of those.
		 */
		if (args->swizzle_mode == I915_BIT_6_SWIZZLE_9_17)
			args->swizzle_mode = I915_BIT_6_SWIZZLE_9;
		if (args->swizzle_mode == I915_BIT_6_SWIZZLE_9_10_17)
			args->swizzle_mode = I915_BIT_6_SWIZZLE_9_10;

		/* If we can't handle the swizzling, make it untiled. */
		if (args->swizzle_mode == I915_BIT_6_SWIZZLE_UNKNOWN) {
			args->tiling_mode = I915_TILING_NONE;
			args->swizzle_mode = I915_BIT_6_SWIZZLE_NONE;
			args->stride = 0;
		}
	}

	mutex_lock(&dev->struct_mutex);
	if (args->tiling_mode != obj->tiling_mode ||
	    args->stride != obj->stride) {
		/* We need to rebind the object if its current allocation
		 * no longer meets the alignment restrictions for its new
		 * tiling mode. Otherwise we can just leave it alone, but
		 * need to ensure that any fence register is cleared.
		 */
		i915_gem_release_mmap(obj);

		obj->map_and_fenceable =
			obj->gtt_space == NULL ||
			(obj->gtt_offset + obj->base.size <= dev_priv->mm.gtt_mappable_end &&
			 i915_gem_object_fence_ok(obj, args->tiling_mode));

		/* Rebind if we need a change of alignment */
		if (!obj->map_and_fenceable) {
			u32 unfenced_alignment =
				i915_gem_get_unfenced_gtt_alignment(dev,
								    obj->base.size,
								    args->tiling_mode);
			if (obj->gtt_offset & (unfenced_alignment - 1))
				ret = i915_gem_object_unbind(obj);
		}

		if (ret == 0) {
			obj->tiling_changed = true;
			obj->tiling_mode = args->tiling_mode;
			obj->stride = args->stride;
		}
	}
	/* we have to maintain this existing ABI... */
	args->stride = obj->stride;
	args->tiling_mode = obj->tiling_mode;
	drm_gem_object_unreference(&obj->base);
	mutex_unlock(&dev->struct_mutex);

	return ret;
}

/**
 * Returns the current tiling mode and required bit 6 swizzling for the object.
 */
int
i915_gem_get_tiling(struct drm_device *dev, void *data,
		   struct drm_file *file)
{
	struct drm_i915_gem_get_tiling *args = data;
	drm_i915_private_t *dev_priv = dev->dev_private;
	struct drm_i915_gem_object *obj;

	obj = to_intel_bo(drm_gem_object_lookup(dev, file, args->handle));
	if (&obj->base == NULL)
		return -ENOENT;

	mutex_lock(&dev->struct_mutex);

	args->tiling_mode = obj->tiling_mode;
	switch (obj->tiling_mode) {
	case I915_TILING_X:
		args->swizzle_mode = dev_priv->mm.bit_6_swizzle_x;
		break;
	case I915_TILING_Y:
		args->swizzle_mode = dev_priv->mm.bit_6_swizzle_y;
		break;
	case I915_TILING_NONE:
		args->swizzle_mode = I915_BIT_6_SWIZZLE_NONE;
		break;
	default:
		DRM_ERROR("unknown tiling mode\n");
	}

	/* Hide bit 17 from the user -- see comment in i915_gem_set_tiling */
	if (args->swizzle_mode == I915_BIT_6_SWIZZLE_9_17)
		args->swizzle_mode = I915_BIT_6_SWIZZLE_9;
	if (args->swizzle_mode == I915_BIT_6_SWIZZLE_9_10_17)
		args->swizzle_mode = I915_BIT_6_SWIZZLE_9_10;

	drm_gem_object_unreference(&obj->base);
	mutex_unlock(&dev->struct_mutex);

	return 0;
}

/**
 * Swap every 64 bytes of this page around, to account for it having a new
 * bit 17 of its physical address and therefore being interpreted differently
 * by the GPU.
 */
static void
i915_gem_swizzle_page(struct page *page)
{
	char temp[64];
	char *vaddr;
	int i;

	vaddr = kmap(page);

	for (i = 0; i < PAGE_SIZE; i += 128) {
		memcpy(temp, &vaddr[i], 64);
		memcpy(&vaddr[i], &vaddr[i + 64], 64);
		memcpy(&vaddr[i + 64], temp, 64);
	}

	kunmap(page);
}

void
i915_gem_object_do_bit_17_swizzle(struct drm_i915_gem_object *obj)
{
	int page_count = obj->base.size >> PAGE_SHIFT;
	int i;

	if (obj->bit_17 == NULL)
		return;

	for (i = 0; i < page_count; i++) {
		char new_bit_17 = page_to_phys(obj->pages[i]) >> 17;
		if ((new_bit_17 & 0x1) !=
		    (test_bit(i, obj->bit_17) != 0)) {
			i915_gem_swizzle_page(obj->pages[i]);
			set_page_dirty(obj->pages[i]);
		}
	}
}

void
i915_gem_object_save_bit_17_swizzle(struct drm_i915_gem_object *obj)
{
	int page_count = obj->base.size >> PAGE_SHIFT;
	int i;

	if (obj->bit_17 == NULL) {
		obj->bit_17 = kmalloc(BITS_TO_LONGS(page_count) *
					   sizeof(long), GFP_KERNEL);
		if (obj->bit_17 == NULL) {
			DRM_ERROR("Failed to allocate memory for bit 17 "
				  "record\n");
			return;
		}
	}

	for (i = 0; i < page_count; i++) {
		if (page_to_phys(obj->pages[i]) & (1 << 17))
			__set_bit(i, obj->bit_17);
		else
			__clear_bit(i, obj->bit_17);
	}
}