huge_pages.c 38.6 KB
Newer Older
M
Matthew Auld 已提交
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
/*
 * Copyright © 2017 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.
 *
 */

#include "../i915_selftest.h"

#include <linux/prime_numbers.h>

#include "mock_drm.h"
30
#include "i915_random.h"
M
Matthew Auld 已提交
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

static const unsigned int page_sizes[] = {
	I915_GTT_PAGE_SIZE_2M,
	I915_GTT_PAGE_SIZE_64K,
	I915_GTT_PAGE_SIZE_4K,
};

static unsigned int get_largest_page_size(struct drm_i915_private *i915,
					  u64 rem)
{
	int i;

	for (i = 0; i < ARRAY_SIZE(page_sizes); ++i) {
		unsigned int page_size = page_sizes[i];

		if (HAS_PAGE_SIZES(i915, page_size) && rem >= page_size)
			return page_size;
	}

	return 0;
}

static void huge_pages_free_pages(struct sg_table *st)
{
	struct scatterlist *sg;

	for (sg = st->sgl; sg; sg = __sg_next(sg)) {
		if (sg_page(sg))
			__free_pages(sg_page(sg), get_order(sg->length));
	}

	sg_free_table(st);
	kfree(st);
}

static int get_huge_pages(struct drm_i915_gem_object *obj)
{
#define GFP (GFP_KERNEL | __GFP_NOWARN | __GFP_NORETRY)
	unsigned int page_mask = obj->mm.page_mask;
	struct sg_table *st;
	struct scatterlist *sg;
M
Matthew Auld 已提交
72
	unsigned int sg_page_sizes;
M
Matthew Auld 已提交
73 74 75 76 77 78 79 80 81 82 83 84 85 86
	u64 rem;

	st = kmalloc(sizeof(*st), GFP);
	if (!st)
		return -ENOMEM;

	if (sg_alloc_table(st, obj->base.size >> PAGE_SHIFT, GFP)) {
		kfree(st);
		return -ENOMEM;
	}

	rem = obj->base.size;
	sg = st->sgl;
	st->nents = 0;
M
Matthew Auld 已提交
87
	sg_page_sizes = 0;
M
Matthew Auld 已提交
88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107

	/*
	 * Our goal here is simple, we want to greedily fill the object from
	 * largest to smallest page-size, while ensuring that we use *every*
	 * page-size as per the given page-mask.
	 */
	do {
		unsigned int bit = ilog2(page_mask);
		unsigned int page_size = BIT(bit);
		int order = get_order(page_size);

		do {
			struct page *page;

			GEM_BUG_ON(order >= MAX_ORDER);
			page = alloc_pages(GFP | __GFP_ZERO, order);
			if (!page)
				goto err;

			sg_set_page(sg, page, page_size, 0);
M
Matthew Auld 已提交
108
			sg_page_sizes |= page_size;
M
Matthew Auld 已提交
109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127
			st->nents++;

			rem -= page_size;
			if (!rem) {
				sg_mark_end(sg);
				break;
			}

			sg = __sg_next(sg);
		} while ((rem - ((page_size-1) & page_mask)) >= page_size);

		page_mask &= (page_size-1);
	} while (page_mask);

	if (i915_gem_gtt_prepare_pages(obj, st))
		goto err;

	obj->mm.madv = I915_MADV_DONTNEED;

M
Matthew Auld 已提交
128 129
	GEM_BUG_ON(sg_page_sizes != obj->mm.page_mask);
	__i915_gem_object_set_pages(obj, st, sg_page_sizes);
M
Matthew Auld 已提交
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

	return 0;

err:
	sg_set_page(sg, NULL, 0, 0);
	sg_mark_end(sg);
	huge_pages_free_pages(st);

	return -ENOMEM;
}

static void put_huge_pages(struct drm_i915_gem_object *obj,
			   struct sg_table *pages)
{
	i915_gem_gtt_finish_pages(obj, pages);
	huge_pages_free_pages(pages);

	obj->mm.dirty = false;
	obj->mm.madv = I915_MADV_WILLNEED;
}

static const struct drm_i915_gem_object_ops huge_page_ops = {
	.flags = I915_GEM_OBJECT_HAS_STRUCT_PAGE |
		 I915_GEM_OBJECT_IS_SHRINKABLE,
	.get_pages = get_huge_pages,
	.put_pages = put_huge_pages,
};

static struct drm_i915_gem_object *
huge_pages_object(struct drm_i915_private *i915,
		  u64 size,
		  unsigned int page_mask)
{
	struct drm_i915_gem_object *obj;

	GEM_BUG_ON(!size);
	GEM_BUG_ON(!IS_ALIGNED(size, BIT(__ffs(page_mask))));

	if (size >> PAGE_SHIFT > INT_MAX)
		return ERR_PTR(-E2BIG);

	if (overflows_type(size, obj->base.size))
		return ERR_PTR(-E2BIG);

	obj = i915_gem_object_alloc(i915);
	if (!obj)
		return ERR_PTR(-ENOMEM);

	drm_gem_private_object_init(&i915->drm, &obj->base, size);
	i915_gem_object_init(obj, &huge_page_ops);

181 182
	obj->write_domain = I915_GEM_DOMAIN_CPU;
	obj->read_domains = I915_GEM_DOMAIN_CPU;
M
Matthew Auld 已提交
183 184 185 186 187 188 189 190 191 192 193 194 195
	obj->cache_level = I915_CACHE_NONE;

	obj->mm.page_mask = page_mask;

	return obj;
}

static int fake_get_huge_pages(struct drm_i915_gem_object *obj)
{
	struct drm_i915_private *i915 = to_i915(obj->base.dev);
	const u64 max_len = rounddown_pow_of_two(UINT_MAX);
	struct sg_table *st;
	struct scatterlist *sg;
M
Matthew Auld 已提交
196
	unsigned int sg_page_sizes;
M
Matthew Auld 已提交
197 198 199 200 201 202 203 204 205 206 207 208 209 210 211
	u64 rem;

	st = kmalloc(sizeof(*st), GFP);
	if (!st)
		return -ENOMEM;

	if (sg_alloc_table(st, obj->base.size >> PAGE_SHIFT, GFP)) {
		kfree(st);
		return -ENOMEM;
	}

	/* Use optimal page sized chunks to fill in the sg table */
	rem = obj->base.size;
	sg = st->sgl;
	st->nents = 0;
M
Matthew Auld 已提交
212
	sg_page_sizes = 0;
M
Matthew Auld 已提交
213 214 215 216 217 218 219 220 221 222 223 224
	do {
		unsigned int page_size = get_largest_page_size(i915, rem);
		unsigned int len = min(page_size * div_u64(rem, page_size),
				       max_len);

		GEM_BUG_ON(!page_size);

		sg->offset = 0;
		sg->length = len;
		sg_dma_len(sg) = len;
		sg_dma_address(sg) = page_size;

M
Matthew Auld 已提交
225
		sg_page_sizes |= len;
M
Matthew Auld 已提交
226 227 228 229 230 231 232 233 234 235 236 237 238 239

		st->nents++;

		rem -= len;
		if (!rem) {
			sg_mark_end(sg);
			break;
		}

		sg = sg_next(sg);
	} while (1);

	obj->mm.madv = I915_MADV_DONTNEED;

M
Matthew Auld 已提交
240
	__i915_gem_object_set_pages(obj, st, sg_page_sizes);
M
Matthew Auld 已提交
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

	return 0;
}

static int fake_get_huge_pages_single(struct drm_i915_gem_object *obj)
{
	struct drm_i915_private *i915 = to_i915(obj->base.dev);
	struct sg_table *st;
	struct scatterlist *sg;
	unsigned int page_size;

	st = kmalloc(sizeof(*st), GFP);
	if (!st)
		return -ENOMEM;

	if (sg_alloc_table(st, 1, GFP)) {
		kfree(st);
		return -ENOMEM;
	}

	sg = st->sgl;
	st->nents = 1;

	page_size = get_largest_page_size(i915, obj->base.size);
	GEM_BUG_ON(!page_size);

	sg->offset = 0;
	sg->length = obj->base.size;
	sg_dma_len(sg) = obj->base.size;
	sg_dma_address(sg) = page_size;

	obj->mm.madv = I915_MADV_DONTNEED;

	__i915_gem_object_set_pages(obj, st, sg->length);

	return 0;
#undef GFP
}

static void fake_free_huge_pages(struct drm_i915_gem_object *obj,
				 struct sg_table *pages)
{
	sg_free_table(pages);
	kfree(pages);
}

static void fake_put_huge_pages(struct drm_i915_gem_object *obj,
				struct sg_table *pages)
{
	fake_free_huge_pages(obj, pages);
	obj->mm.dirty = false;
	obj->mm.madv = I915_MADV_WILLNEED;
}

static const struct drm_i915_gem_object_ops fake_ops = {
	.flags = I915_GEM_OBJECT_IS_SHRINKABLE,
	.get_pages = fake_get_huge_pages,
	.put_pages = fake_put_huge_pages,
};

static const struct drm_i915_gem_object_ops fake_ops_single = {
	.flags = I915_GEM_OBJECT_IS_SHRINKABLE,
	.get_pages = fake_get_huge_pages_single,
	.put_pages = fake_put_huge_pages,
};

static struct drm_i915_gem_object *
fake_huge_pages_object(struct drm_i915_private *i915, u64 size, bool single)
{
	struct drm_i915_gem_object *obj;

	GEM_BUG_ON(!size);
	GEM_BUG_ON(!IS_ALIGNED(size, I915_GTT_PAGE_SIZE));

	if (size >> PAGE_SHIFT > UINT_MAX)
		return ERR_PTR(-E2BIG);

	if (overflows_type(size, obj->base.size))
		return ERR_PTR(-E2BIG);

	obj = i915_gem_object_alloc(i915);
	if (!obj)
		return ERR_PTR(-ENOMEM);

	drm_gem_private_object_init(&i915->drm, &obj->base, size);

	if (single)
		i915_gem_object_init(obj, &fake_ops_single);
	else
		i915_gem_object_init(obj, &fake_ops);

332 333
	obj->write_domain = I915_GEM_DOMAIN_CPU;
	obj->read_domains = I915_GEM_DOMAIN_CPU;
M
Matthew Auld 已提交
334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612
	obj->cache_level = I915_CACHE_NONE;

	return obj;
}

static int igt_check_page_sizes(struct i915_vma *vma)
{
	struct drm_i915_private *i915 = to_i915(vma->obj->base.dev);
	unsigned int supported = INTEL_INFO(i915)->page_sizes;
	struct drm_i915_gem_object *obj = vma->obj;
	int err = 0;

	if (!HAS_PAGE_SIZES(i915, vma->page_sizes.sg)) {
		pr_err("unsupported page_sizes.sg=%u, supported=%u\n",
		       vma->page_sizes.sg & ~supported, supported);
		err = -EINVAL;
	}

	if (!HAS_PAGE_SIZES(i915, vma->page_sizes.gtt)) {
		pr_err("unsupported page_sizes.gtt=%u, supported=%u\n",
		       vma->page_sizes.gtt & ~supported, supported);
		err = -EINVAL;
	}

	if (vma->page_sizes.phys != obj->mm.page_sizes.phys) {
		pr_err("vma->page_sizes.phys(%u) != obj->mm.page_sizes.phys(%u)\n",
		       vma->page_sizes.phys, obj->mm.page_sizes.phys);
		err = -EINVAL;
	}

	if (vma->page_sizes.sg != obj->mm.page_sizes.sg) {
		pr_err("vma->page_sizes.sg(%u) != obj->mm.page_sizes.sg(%u)\n",
		       vma->page_sizes.sg, obj->mm.page_sizes.sg);
		err = -EINVAL;
	}

	if (obj->mm.page_sizes.gtt) {
		pr_err("obj->page_sizes.gtt(%u) should never be set\n",
		       obj->mm.page_sizes.gtt);
		err = -EINVAL;
	}

	return err;
}

static int igt_mock_exhaust_device_supported_pages(void *arg)
{
	struct i915_hw_ppgtt *ppgtt = arg;
	struct drm_i915_private *i915 = ppgtt->base.i915;
	unsigned int saved_mask = INTEL_INFO(i915)->page_sizes;
	struct drm_i915_gem_object *obj;
	struct i915_vma *vma;
	int i, j, single;
	int err;

	/*
	 * Sanity check creating objects with every valid page support
	 * combination for our mock device.
	 */

	for (i = 1; i < BIT(ARRAY_SIZE(page_sizes)); i++) {
		unsigned int combination = 0;

		for (j = 0; j < ARRAY_SIZE(page_sizes); j++) {
			if (i & BIT(j))
				combination |= page_sizes[j];
		}

		mkwrite_device_info(i915)->page_sizes = combination;

		for (single = 0; single <= 1; ++single) {
			obj = fake_huge_pages_object(i915, combination, !!single);
			if (IS_ERR(obj)) {
				err = PTR_ERR(obj);
				goto out_device;
			}

			if (obj->base.size != combination) {
				pr_err("obj->base.size=%zu, expected=%u\n",
				       obj->base.size, combination);
				err = -EINVAL;
				goto out_put;
			}

			vma = i915_vma_instance(obj, &ppgtt->base, NULL);
			if (IS_ERR(vma)) {
				err = PTR_ERR(vma);
				goto out_put;
			}

			err = i915_vma_pin(vma, 0, 0, PIN_USER);
			if (err)
				goto out_close;

			err = igt_check_page_sizes(vma);

			if (vma->page_sizes.sg != combination) {
				pr_err("page_sizes.sg=%u, expected=%u\n",
				       vma->page_sizes.sg, combination);
				err = -EINVAL;
			}

			i915_vma_unpin(vma);
			i915_vma_close(vma);

			i915_gem_object_put(obj);

			if (err)
				goto out_device;
		}
	}

	goto out_device;

out_close:
	i915_vma_close(vma);
out_put:
	i915_gem_object_put(obj);
out_device:
	mkwrite_device_info(i915)->page_sizes = saved_mask;

	return err;
}

static int igt_mock_ppgtt_misaligned_dma(void *arg)
{
	struct i915_hw_ppgtt *ppgtt = arg;
	struct drm_i915_private *i915 = ppgtt->base.i915;
	unsigned long supported = INTEL_INFO(i915)->page_sizes;
	struct drm_i915_gem_object *obj;
	int bit;
	int err;

	/*
	 * Sanity check dma misalignment for huge pages -- the dma addresses we
	 * insert into the paging structures need to always respect the page
	 * size alignment.
	 */

	bit = ilog2(I915_GTT_PAGE_SIZE_64K);

	for_each_set_bit_from(bit, &supported,
			      ilog2(I915_GTT_MAX_PAGE_SIZE) + 1) {
		IGT_TIMEOUT(end_time);
		unsigned int page_size = BIT(bit);
		unsigned int flags = PIN_USER | PIN_OFFSET_FIXED;
		unsigned int offset;
		unsigned int size =
			round_up(page_size, I915_GTT_PAGE_SIZE_2M) << 1;
		struct i915_vma *vma;

		obj = fake_huge_pages_object(i915, size, true);
		if (IS_ERR(obj))
			return PTR_ERR(obj);

		if (obj->base.size != size) {
			pr_err("obj->base.size=%zu, expected=%u\n",
			       obj->base.size, size);
			err = -EINVAL;
			goto out_put;
		}

		err = i915_gem_object_pin_pages(obj);
		if (err)
			goto out_put;

		/* Force the page size for this object */
		obj->mm.page_sizes.sg = page_size;

		vma = i915_vma_instance(obj, &ppgtt->base, NULL);
		if (IS_ERR(vma)) {
			err = PTR_ERR(vma);
			goto out_unpin;
		}

		err = i915_vma_pin(vma, 0, 0, flags);
		if (err) {
			i915_vma_close(vma);
			goto out_unpin;
		}


		err = igt_check_page_sizes(vma);

		if (vma->page_sizes.gtt != page_size) {
			pr_err("page_sizes.gtt=%u, expected %u\n",
			       vma->page_sizes.gtt, page_size);
			err = -EINVAL;
		}

		i915_vma_unpin(vma);

		if (err) {
			i915_vma_close(vma);
			goto out_unpin;
		}

		/*
		 * Try all the other valid offsets until the next
		 * boundary -- should always fall back to using 4K
		 * pages.
		 */
		for (offset = 4096; offset < page_size; offset += 4096) {
			err = i915_vma_unbind(vma);
			if (err) {
				i915_vma_close(vma);
				goto out_unpin;
			}

			err = i915_vma_pin(vma, 0, 0, flags | offset);
			if (err) {
				i915_vma_close(vma);
				goto out_unpin;
			}

			err = igt_check_page_sizes(vma);

			if (vma->page_sizes.gtt != I915_GTT_PAGE_SIZE_4K) {
				pr_err("page_sizes.gtt=%u, expected %lu\n",
				       vma->page_sizes.gtt, I915_GTT_PAGE_SIZE_4K);
				err = -EINVAL;
			}

			i915_vma_unpin(vma);

			if (err) {
				i915_vma_close(vma);
				goto out_unpin;
			}

			if (igt_timeout(end_time,
					"%s timed out at offset %x with page-size %x\n",
					__func__, offset, page_size))
				break;
		}

		i915_vma_close(vma);

		i915_gem_object_unpin_pages(obj);
		i915_gem_object_put(obj);
	}

	return 0;

out_unpin:
	i915_gem_object_unpin_pages(obj);
out_put:
	i915_gem_object_put(obj);

	return err;
}

static void close_object_list(struct list_head *objects,
			      struct i915_hw_ppgtt *ppgtt)
{
	struct drm_i915_gem_object *obj, *on;

	list_for_each_entry_safe(obj, on, objects, st_link) {
		struct i915_vma *vma;

		vma = i915_vma_instance(obj, &ppgtt->base, NULL);
		if (!IS_ERR(vma))
			i915_vma_close(vma);

		list_del(&obj->st_link);
		i915_gem_object_unpin_pages(obj);
		i915_gem_object_put(obj);
	}
}

static int igt_mock_ppgtt_huge_fill(void *arg)
{
	struct i915_hw_ppgtt *ppgtt = arg;
	struct drm_i915_private *i915 = ppgtt->base.i915;
	unsigned long max_pages = ppgtt->base.total >> PAGE_SHIFT;
	unsigned long page_num;
	bool single = false;
	LIST_HEAD(objects);
	IGT_TIMEOUT(end_time);
613
	int err = -ENODEV;
M
Matthew Auld 已提交
614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966

	for_each_prime_number_from(page_num, 1, max_pages) {
		struct drm_i915_gem_object *obj;
		u64 size = page_num << PAGE_SHIFT;
		struct i915_vma *vma;
		unsigned int expected_gtt = 0;
		int i;

		obj = fake_huge_pages_object(i915, size, single);
		if (IS_ERR(obj)) {
			err = PTR_ERR(obj);
			break;
		}

		if (obj->base.size != size) {
			pr_err("obj->base.size=%zd, expected=%llu\n",
			       obj->base.size, size);
			i915_gem_object_put(obj);
			err = -EINVAL;
			break;
		}

		err = i915_gem_object_pin_pages(obj);
		if (err) {
			i915_gem_object_put(obj);
			break;
		}

		list_add(&obj->st_link, &objects);

		vma = i915_vma_instance(obj, &ppgtt->base, NULL);
		if (IS_ERR(vma)) {
			err = PTR_ERR(vma);
			break;
		}

		err = i915_vma_pin(vma, 0, 0, PIN_USER);
		if (err)
			break;

		err = igt_check_page_sizes(vma);
		if (err) {
			i915_vma_unpin(vma);
			break;
		}

		/*
		 * Figure out the expected gtt page size knowing that we go from
		 * largest to smallest page size sg chunks, and that we align to
		 * the largest page size.
		 */
		for (i = 0; i < ARRAY_SIZE(page_sizes); ++i) {
			unsigned int page_size = page_sizes[i];

			if (HAS_PAGE_SIZES(i915, page_size) &&
			    size >= page_size) {
				expected_gtt |= page_size;
				size &= page_size-1;
			}
		}

		GEM_BUG_ON(!expected_gtt);
		GEM_BUG_ON(size);

		if (expected_gtt & I915_GTT_PAGE_SIZE_4K)
			expected_gtt &= ~I915_GTT_PAGE_SIZE_64K;

		i915_vma_unpin(vma);

		if (vma->page_sizes.sg & I915_GTT_PAGE_SIZE_64K) {
			if (!IS_ALIGNED(vma->node.start,
					I915_GTT_PAGE_SIZE_2M)) {
				pr_err("node.start(%llx) not aligned to 2M\n",
				       vma->node.start);
				err = -EINVAL;
				break;
			}

			if (!IS_ALIGNED(vma->node.size,
					I915_GTT_PAGE_SIZE_2M)) {
				pr_err("node.size(%llx) not aligned to 2M\n",
				       vma->node.size);
				err = -EINVAL;
				break;
			}
		}

		if (vma->page_sizes.gtt != expected_gtt) {
			pr_err("gtt=%u, expected=%u, size=%zd, single=%s\n",
			       vma->page_sizes.gtt, expected_gtt,
			       obj->base.size, yesno(!!single));
			err = -EINVAL;
			break;
		}

		if (igt_timeout(end_time,
				"%s timed out at size %zd\n",
				__func__, obj->base.size))
			break;

		single = !single;
	}

	close_object_list(&objects, ppgtt);

	if (err == -ENOMEM || err == -ENOSPC)
		err = 0;

	return err;
}

static int igt_mock_ppgtt_64K(void *arg)
{
	struct i915_hw_ppgtt *ppgtt = arg;
	struct drm_i915_private *i915 = ppgtt->base.i915;
	struct drm_i915_gem_object *obj;
	const struct object_info {
		unsigned int size;
		unsigned int gtt;
		unsigned int offset;
	} objects[] = {
		/* Cases with forced padding/alignment */
		{
			.size = SZ_64K,
			.gtt = I915_GTT_PAGE_SIZE_64K,
			.offset = 0,
		},
		{
			.size = SZ_64K + SZ_4K,
			.gtt = I915_GTT_PAGE_SIZE_4K,
			.offset = 0,
		},
		{
			.size = SZ_64K - SZ_4K,
			.gtt = I915_GTT_PAGE_SIZE_4K,
			.offset = 0,
		},
		{
			.size = SZ_2M,
			.gtt = I915_GTT_PAGE_SIZE_64K,
			.offset = 0,
		},
		{
			.size = SZ_2M - SZ_4K,
			.gtt = I915_GTT_PAGE_SIZE_4K,
			.offset = 0,
		},
		{
			.size = SZ_2M + SZ_4K,
			.gtt = I915_GTT_PAGE_SIZE_64K | I915_GTT_PAGE_SIZE_4K,
			.offset = 0,
		},
		{
			.size = SZ_2M + SZ_64K,
			.gtt = I915_GTT_PAGE_SIZE_64K,
			.offset = 0,
		},
		{
			.size = SZ_2M - SZ_64K,
			.gtt = I915_GTT_PAGE_SIZE_64K,
			.offset = 0,
		},
		/* Try without any forced padding/alignment */
		{
			.size = SZ_64K,
			.offset = SZ_2M,
			.gtt = I915_GTT_PAGE_SIZE_4K,
		},
		{
			.size = SZ_128K,
			.offset = SZ_2M - SZ_64K,
			.gtt = I915_GTT_PAGE_SIZE_4K,
		},
	};
	struct i915_vma *vma;
	int i, single;
	int err;

	/*
	 * Sanity check some of the trickiness with 64K pages -- either we can
	 * safely mark the whole page-table(2M block) as 64K, or we have to
	 * always fallback to 4K.
	 */

	if (!HAS_PAGE_SIZES(i915, I915_GTT_PAGE_SIZE_64K))
		return 0;

	for (i = 0; i < ARRAY_SIZE(objects); ++i) {
		unsigned int size = objects[i].size;
		unsigned int expected_gtt = objects[i].gtt;
		unsigned int offset = objects[i].offset;
		unsigned int flags = PIN_USER;

		for (single = 0; single <= 1; single++) {
			obj = fake_huge_pages_object(i915, size, !!single);
			if (IS_ERR(obj))
				return PTR_ERR(obj);

			err = i915_gem_object_pin_pages(obj);
			if (err)
				goto out_object_put;

			/*
			 * Disable 2M pages -- We only want to use 64K/4K pages
			 * for this test.
			 */
			obj->mm.page_sizes.sg &= ~I915_GTT_PAGE_SIZE_2M;

			vma = i915_vma_instance(obj, &ppgtt->base, NULL);
			if (IS_ERR(vma)) {
				err = PTR_ERR(vma);
				goto out_object_unpin;
			}

			if (offset)
				flags |= PIN_OFFSET_FIXED | offset;

			err = i915_vma_pin(vma, 0, 0, flags);
			if (err)
				goto out_vma_close;

			err = igt_check_page_sizes(vma);
			if (err)
				goto out_vma_unpin;

			if (!offset && vma->page_sizes.sg & I915_GTT_PAGE_SIZE_64K) {
				if (!IS_ALIGNED(vma->node.start,
						I915_GTT_PAGE_SIZE_2M)) {
					pr_err("node.start(%llx) not aligned to 2M\n",
					       vma->node.start);
					err = -EINVAL;
					goto out_vma_unpin;
				}

				if (!IS_ALIGNED(vma->node.size,
						I915_GTT_PAGE_SIZE_2M)) {
					pr_err("node.size(%llx) not aligned to 2M\n",
					       vma->node.size);
					err = -EINVAL;
					goto out_vma_unpin;
				}
			}

			if (vma->page_sizes.gtt != expected_gtt) {
				pr_err("gtt=%u, expected=%u, i=%d, single=%s\n",
				       vma->page_sizes.gtt, expected_gtt, i,
				       yesno(!!single));
				err = -EINVAL;
				goto out_vma_unpin;
			}

			i915_vma_unpin(vma);
			i915_vma_close(vma);

			i915_gem_object_unpin_pages(obj);
			i915_gem_object_put(obj);
		}
	}

	return 0;

out_vma_unpin:
	i915_vma_unpin(vma);
out_vma_close:
	i915_vma_close(vma);
out_object_unpin:
	i915_gem_object_unpin_pages(obj);
out_object_put:
	i915_gem_object_put(obj);

	return err;
}

static struct i915_vma *
gpu_write_dw(struct i915_vma *vma, u64 offset, u32 val)
{
	struct drm_i915_private *i915 = to_i915(vma->obj->base.dev);
	const int gen = INTEL_GEN(vma->vm->i915);
	unsigned int count = vma->size >> PAGE_SHIFT;
	struct drm_i915_gem_object *obj;
	struct i915_vma *batch;
	unsigned int size;
	u32 *cmd;
	int n;
	int err;

	size = (1 + 4 * count) * sizeof(u32);
	size = round_up(size, PAGE_SIZE);
	obj = i915_gem_object_create_internal(i915, size);
	if (IS_ERR(obj))
		return ERR_CAST(obj);

	cmd = i915_gem_object_pin_map(obj, I915_MAP_WB);
	if (IS_ERR(cmd)) {
		err = PTR_ERR(cmd);
		goto err;
	}

	offset += vma->node.start;

	for (n = 0; n < count; n++) {
		if (gen >= 8) {
			*cmd++ = MI_STORE_DWORD_IMM_GEN4;
			*cmd++ = lower_32_bits(offset);
			*cmd++ = upper_32_bits(offset);
			*cmd++ = val;
		} else if (gen >= 4) {
			*cmd++ = MI_STORE_DWORD_IMM_GEN4 |
				(gen < 6 ? 1 << 22 : 0);
			*cmd++ = 0;
			*cmd++ = offset;
			*cmd++ = val;
		} else {
			*cmd++ = MI_STORE_DWORD_IMM | 1 << 22;
			*cmd++ = offset;
			*cmd++ = val;
		}

		offset += PAGE_SIZE;
	}

	*cmd = MI_BATCH_BUFFER_END;

	i915_gem_object_unpin_map(obj);

	err = i915_gem_object_set_to_gtt_domain(obj, false);
	if (err)
		goto err;

	batch = i915_vma_instance(obj, vma->vm, NULL);
	if (IS_ERR(batch)) {
		err = PTR_ERR(batch);
		goto err;
	}

	err = i915_vma_pin(batch, 0, 0, PIN_USER);
	if (err)
		goto err;

	return batch;

err:
	i915_gem_object_put(obj);

	return ERR_PTR(err);
}

static int gpu_write(struct i915_vma *vma,
		     struct i915_gem_context *ctx,
		     struct intel_engine_cs *engine,
		     u32 dword,
		     u32 value)
{
967
	struct i915_request *rq;
M
Matthew Auld 已提交
968 969 970 971 972 973 974 975 976 977
	struct i915_vma *batch;
	int flags = 0;
	int err;

	GEM_BUG_ON(!intel_engine_can_store_dword(engine));

	err = i915_gem_object_set_to_gtt_domain(vma->obj, true);
	if (err)
		return err;

978
	rq = i915_request_alloc(engine, ctx);
M
Matthew Auld 已提交
979 980 981 982 983 984 985 986 987 988 989 990 991 992
	if (IS_ERR(rq))
		return PTR_ERR(rq);

	batch = gpu_write_dw(vma, dword * sizeof(u32), value);
	if (IS_ERR(batch)) {
		err = PTR_ERR(batch);
		goto err_request;
	}

	i915_vma_move_to_active(batch, rq, 0);
	i915_gem_object_set_active_reference(batch->obj);
	i915_vma_unpin(batch);
	i915_vma_close(batch);

993 994 995
	err = engine->emit_bb_start(rq,
				    batch->node.start, batch->node.size,
				    flags);
M
Matthew Auld 已提交
996 997 998 999 1000 1001 1002 1003 1004 1005
	if (err)
		goto err_request;

	i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);

	reservation_object_lock(vma->resv, NULL);
	reservation_object_add_excl_fence(vma->resv, &rq->fence);
	reservation_object_unlock(vma->resv);

err_request:
1006
	__i915_request_add(rq, err == 0);
M
Matthew Auld 已提交
1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042

	return err;
}

static int cpu_check(struct drm_i915_gem_object *obj, u32 dword, u32 val)
{
	unsigned int needs_flush;
	unsigned long n;
	int err;

	err = i915_gem_obj_prepare_shmem_read(obj, &needs_flush);
	if (err)
		return err;

	for (n = 0; n < obj->base.size >> PAGE_SHIFT; ++n) {
		u32 *ptr = kmap_atomic(i915_gem_object_get_page(obj, n));

		if (needs_flush & CLFLUSH_BEFORE)
			drm_clflush_virt_range(ptr, PAGE_SIZE);

		if (ptr[dword] != val) {
			pr_err("n=%lu ptr[%u]=%u, val=%u\n",
			       n, dword, ptr[dword], val);
			kunmap_atomic(ptr);
			err = -EINVAL;
			break;
		}

		kunmap_atomic(ptr);
	}

	i915_gem_obj_finish_shmem_access(obj);

	return err;
}

1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093
static int __igt_write_huge(struct i915_gem_context *ctx,
			    struct intel_engine_cs *engine,
			    struct drm_i915_gem_object *obj,
			    u64 size, u64 offset,
			    u32 dword, u32 val)
{
	struct drm_i915_private *i915 = to_i915(obj->base.dev);
	struct i915_address_space *vm = ctx->ppgtt ? &ctx->ppgtt->base : &i915->ggtt.base;
	unsigned int flags = PIN_USER | PIN_OFFSET_FIXED;
	struct i915_vma *vma;
	int err;

	vma = i915_vma_instance(obj, vm, NULL);
	if (IS_ERR(vma))
		return PTR_ERR(vma);

	err = i915_vma_unbind(vma);
	if (err)
		goto out_vma_close;

	err = i915_vma_pin(vma, size, 0, flags | offset);
	if (err) {
		/*
		 * The ggtt may have some pages reserved so
		 * refrain from erroring out.
		 */
		if (err == -ENOSPC && i915_is_ggtt(vm))
			err = 0;

		goto out_vma_close;
	}

	err = igt_check_page_sizes(vma);
	if (err)
		goto out_vma_unpin;

	err = gpu_write(vma, ctx, engine, dword, val);
	if (err) {
		pr_err("gpu-write failed at offset=%llx\n", offset);
		goto out_vma_unpin;
	}

	err = cpu_check(obj, dword, val);
	if (err) {
		pr_err("cpu-check failed at offset=%llx\n", offset);
		goto out_vma_unpin;
	}

out_vma_unpin:
	i915_vma_unpin(vma);
out_vma_close:
1094
	i915_vma_destroy(vma);
1095 1096 1097 1098

	return err;
}

1099 1100
static int igt_write_huge(struct i915_gem_context *ctx,
			  struct drm_i915_gem_object *obj)
M
Matthew Auld 已提交
1101 1102 1103
{
	struct drm_i915_private *i915 = to_i915(obj->base.dev);
	struct i915_address_space *vm = ctx->ppgtt ? &ctx->ppgtt->base : &i915->ggtt.base;
1104
	static struct intel_engine_cs *engines[I915_NUM_ENGINES];
M
Matthew Auld 已提交
1105
	struct intel_engine_cs *engine;
1106 1107
	I915_RND_STATE(prng);
	IGT_TIMEOUT(end_time);
M
Matthew Auld 已提交
1108 1109 1110 1111 1112
	unsigned int max_page_size;
	unsigned int id;
	u64 max;
	u64 num;
	u64 size;
1113 1114
	int *order;
	int i, n;
M
Matthew Auld 已提交
1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125
	int err = 0;

	GEM_BUG_ON(!i915_gem_object_has_pinned_pages(obj));

	size = obj->base.size;
	if (obj->mm.page_sizes.sg & I915_GTT_PAGE_SIZE_64K)
		size = round_up(size, I915_GTT_PAGE_SIZE_2M);

	max_page_size = rounddown_pow_of_two(obj->mm.page_sizes.sg);
	max = div_u64((vm->total - size), max_page_size);

1126
	n = 0;
M
Matthew Auld 已提交
1127 1128
	for_each_engine(engine, i915, id) {
		if (!intel_engine_can_store_dword(engine)) {
1129
			pr_info("store-dword-imm not supported on engine=%u\n", id);
M
Matthew Auld 已提交
1130 1131
			continue;
		}
1132 1133
		engines[n++] = engine;
	}
M
Matthew Auld 已提交
1134

1135 1136
	if (!n)
		return 0;
M
Matthew Auld 已提交
1137

1138 1139 1140 1141 1142 1143 1144 1145
	/*
	 * To keep things interesting when alternating between engines in our
	 * randomized order, lets also make feeding to the same engine a few
	 * times in succession a possibility by enlarging the permutation array.
	 */
	order = i915_random_order(n * I915_NUM_ENGINES, &prng);
	if (!order)
		return -ENOMEM;
M
Matthew Auld 已提交
1146

1147
	/*
1148 1149 1150
	 * Try various offsets in an ascending/descending fashion until we
	 * timeout -- we want to avoid issues hidden by effectively always using
	 * offset = 0.
1151 1152 1153
	 */
	i = 0;
	for_each_prime_number_from(num, 0, max) {
1154 1155 1156
		u64 offset_low = num * max_page_size;
		u64 offset_high = (max - num) * max_page_size;
		u32 dword = offset_in_page(num) / 4;
M
Matthew Auld 已提交
1157

1158 1159
		engine = engines[order[i] % n];
		i = (i + 1) % (n * I915_NUM_ENGINES);
M
Matthew Auld 已提交
1160

1161 1162 1163
		err = __igt_write_huge(ctx, engine, obj, size, offset_low, dword, num + 1);
		if (err)
			break;
1164

1165 1166 1167
		err = __igt_write_huge(ctx, engine, obj, size, offset_high, dword, num + 1);
		if (err)
			break;
1168 1169

		if (igt_timeout(end_time,
1170 1171
				"%s timed out on engine=%u, offset_low=%llx offset_high=%llx, max_page_size=%x\n",
				__func__, engine->id, offset_low, offset_high, max_page_size))
1172
			break;
M
Matthew Auld 已提交
1173 1174
	}

1175
	kfree(order);
M
Matthew Auld 已提交
1176 1177 1178 1179 1180 1181

	return err;
}

static int igt_ppgtt_exhaust_huge(void *arg)
{
1182 1183
	struct i915_gem_context *ctx = arg;
	struct drm_i915_private *i915 = ctx->i915;
M
Matthew Auld 已提交
1184 1185 1186 1187 1188 1189
	unsigned long supported = INTEL_INFO(i915)->page_sizes;
	static unsigned int pages[ARRAY_SIZE(page_sizes)];
	struct drm_i915_gem_object *obj;
	unsigned int size_mask;
	unsigned int page_mask;
	int n, i;
1190
	int err = -ENODEV;
M
Matthew Auld 已提交
1191

1192 1193 1194
	if (supported == I915_GTT_PAGE_SIZE_4K)
		return 0;

M
Matthew Auld 已提交
1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256
	/*
	 * Sanity check creating objects with a varying mix of page sizes --
	 * ensuring that our writes lands in the right place.
	 */

	n = 0;
	for_each_set_bit(i, &supported, ilog2(I915_GTT_MAX_PAGE_SIZE) + 1)
		pages[n++] = BIT(i);

	for (size_mask = 2; size_mask < BIT(n); size_mask++) {
		unsigned int size = 0;

		for (i = 0; i < n; i++) {
			if (size_mask & BIT(i))
				size |= pages[i];
		}

		/*
		 * For our page mask we want to enumerate all the page-size
		 * combinations which will fit into our chosen object size.
		 */
		for (page_mask = 2; page_mask <= size_mask; page_mask++) {
			unsigned int page_sizes = 0;

			for (i = 0; i < n; i++) {
				if (page_mask & BIT(i))
					page_sizes |= pages[i];
			}

			/*
			 * Ensure that we can actually fill the given object
			 * with our chosen page mask.
			 */
			if (!IS_ALIGNED(size, BIT(__ffs(page_sizes))))
				continue;

			obj = huge_pages_object(i915, size, page_sizes);
			if (IS_ERR(obj)) {
				err = PTR_ERR(obj);
				goto out_device;
			}

			err = i915_gem_object_pin_pages(obj);
			if (err) {
				i915_gem_object_put(obj);

				if (err == -ENOMEM) {
					pr_info("unable to get pages, size=%u, pages=%u\n",
						size, page_sizes);
					err = 0;
					break;
				}

				pr_err("pin_pages failed, size=%u, pages=%u\n",
				       size_mask, page_mask);

				goto out_device;
			}

			/* Force the page-size for the gtt insertion */
			obj->mm.page_sizes.sg = page_sizes;

1257
			err = igt_write_huge(ctx, obj);
M
Matthew Auld 已提交
1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281
			if (err) {
				pr_err("exhaust write-huge failed with size=%u\n",
				       size);
				goto out_unpin;
			}

			i915_gem_object_unpin_pages(obj);
			i915_gem_object_put(obj);
		}
	}

	goto out_device;

out_unpin:
	i915_gem_object_unpin_pages(obj);
	i915_gem_object_put(obj);
out_device:
	mkwrite_device_info(i915)->page_sizes = supported;

	return err;
}

static int igt_ppgtt_internal_huge(void *arg)
{
1282 1283
	struct i915_gem_context *ctx = arg;
	struct drm_i915_private *i915 = ctx->i915;
M
Matthew Auld 已提交
1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317
	struct drm_i915_gem_object *obj;
	static const unsigned int sizes[] = {
		SZ_64K,
		SZ_128K,
		SZ_256K,
		SZ_512K,
		SZ_1M,
		SZ_2M,
	};
	int i;
	int err;

	/*
	 * Sanity check that the HW uses huge pages correctly through internal
	 * -- ensure that our writes land in the right place.
	 */

	for (i = 0; i < ARRAY_SIZE(sizes); ++i) {
		unsigned int size = sizes[i];

		obj = i915_gem_object_create_internal(i915, size);
		if (IS_ERR(obj))
			return PTR_ERR(obj);

		err = i915_gem_object_pin_pages(obj);
		if (err)
			goto out_put;

		if (obj->mm.page_sizes.phys < I915_GTT_PAGE_SIZE_64K) {
			pr_info("internal unable to allocate huge-page(s) with size=%u\n",
				size);
			goto out_unpin;
		}

1318
		err = igt_write_huge(ctx, obj);
M
Matthew Auld 已提交
1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345
		if (err) {
			pr_err("internal write-huge failed with size=%u\n",
			       size);
			goto out_unpin;
		}

		i915_gem_object_unpin_pages(obj);
		i915_gem_object_put(obj);
	}

	return 0;

out_unpin:
	i915_gem_object_unpin_pages(obj);
out_put:
	i915_gem_object_put(obj);

	return err;
}

static inline bool igt_can_allocate_thp(struct drm_i915_private *i915)
{
	return i915->mm.gemfs && has_transparent_hugepage();
}

static int igt_ppgtt_gemfs_huge(void *arg)
{
1346 1347
	struct i915_gem_context *ctx = arg;
	struct drm_i915_private *i915 = ctx->i915;
M
Matthew Auld 已提交
1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385
	struct drm_i915_gem_object *obj;
	static const unsigned int sizes[] = {
		SZ_2M,
		SZ_4M,
		SZ_8M,
		SZ_16M,
		SZ_32M,
	};
	int i;
	int err;

	/*
	 * Sanity check that the HW uses huge pages correctly through gemfs --
	 * ensure that our writes land in the right place.
	 */

	if (!igt_can_allocate_thp(i915)) {
		pr_info("missing THP support, skipping\n");
		return 0;
	}

	for (i = 0; i < ARRAY_SIZE(sizes); ++i) {
		unsigned int size = sizes[i];

		obj = i915_gem_object_create(i915, size);
		if (IS_ERR(obj))
			return PTR_ERR(obj);

		err = i915_gem_object_pin_pages(obj);
		if (err)
			goto out_put;

		if (obj->mm.page_sizes.phys < I915_GTT_PAGE_SIZE_2M) {
			pr_info("finishing test early, gemfs unable to allocate huge-page(s) with size=%u\n",
				size);
			goto out_unpin;
		}

1386
		err = igt_write_huge(ctx, obj);
M
Matthew Auld 已提交
1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408
		if (err) {
			pr_err("gemfs write-huge failed with size=%u\n",
			       size);
			goto out_unpin;
		}

		i915_gem_object_unpin_pages(obj);
		i915_gem_object_put(obj);
	}

	return 0;

out_unpin:
	i915_gem_object_unpin_pages(obj);
out_put:
	i915_gem_object_put(obj);

	return err;
}

static int igt_ppgtt_pin_update(void *arg)
{
1409 1410
	struct i915_gem_context *ctx = arg;
	struct drm_i915_private *dev_priv = ctx->i915;
M
Matthew Auld 已提交
1411
	unsigned long supported = INTEL_INFO(dev_priv)->page_sizes;
1412
	struct i915_hw_ppgtt *ppgtt = ctx->ppgtt;
M
Matthew Auld 已提交
1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512
	struct drm_i915_gem_object *obj;
	struct i915_vma *vma;
	unsigned int flags = PIN_USER | PIN_OFFSET_FIXED;
	int first, last;
	int err;

	/*
	 * Make sure there's no funny business when doing a PIN_UPDATE -- in the
	 * past we had a subtle issue with being able to incorrectly do multiple
	 * alloc va ranges on the same object when doing a PIN_UPDATE, which
	 * resulted in some pretty nasty bugs, though only when using
	 * huge-gtt-pages.
	 */

	if (!USES_FULL_48BIT_PPGTT(dev_priv)) {
		pr_info("48b PPGTT not supported, skipping\n");
		return 0;
	}

	first = ilog2(I915_GTT_PAGE_SIZE_64K);
	last = ilog2(I915_GTT_PAGE_SIZE_2M);

	for_each_set_bit_from(first, &supported, last + 1) {
		unsigned int page_size = BIT(first);

		obj = i915_gem_object_create_internal(dev_priv, page_size);
		if (IS_ERR(obj))
			return PTR_ERR(obj);

		vma = i915_vma_instance(obj, &ppgtt->base, NULL);
		if (IS_ERR(vma)) {
			err = PTR_ERR(vma);
			goto out_put;
		}

		err = i915_vma_pin(vma, SZ_2M, 0, flags);
		if (err)
			goto out_close;

		if (vma->page_sizes.sg < page_size) {
			pr_info("Unable to allocate page-size %x, finishing test early\n",
				page_size);
			goto out_unpin;
		}

		err = igt_check_page_sizes(vma);
		if (err)
			goto out_unpin;

		if (vma->page_sizes.gtt != page_size) {
			dma_addr_t addr = i915_gem_object_get_dma_address(obj, 0);

			/*
			 * The only valid reason for this to ever fail would be
			 * if the dma-mapper screwed us over when we did the
			 * dma_map_sg(), since it has the final say over the dma
			 * address.
			 */
			if (IS_ALIGNED(addr, page_size)) {
				pr_err("page_sizes.gtt=%u, expected=%u\n",
				       vma->page_sizes.gtt, page_size);
				err = -EINVAL;
			} else {
				pr_info("dma address misaligned, finishing test early\n");
			}

			goto out_unpin;
		}

		err = i915_vma_bind(vma, I915_CACHE_NONE, PIN_UPDATE);
		if (err)
			goto out_unpin;

		i915_vma_unpin(vma);
		i915_vma_close(vma);

		i915_gem_object_put(obj);
	}

	obj = i915_gem_object_create_internal(dev_priv, PAGE_SIZE);
	if (IS_ERR(obj))
		return PTR_ERR(obj);

	vma = i915_vma_instance(obj, &ppgtt->base, NULL);
	if (IS_ERR(vma)) {
		err = PTR_ERR(vma);
		goto out_put;
	}

	err = i915_vma_pin(vma, 0, 0, flags);
	if (err)
		goto out_close;

	/*
	 * Make sure we don't end up with something like where the pde is still
	 * pointing to the 2M page, and the pt we just filled-in is dangling --
	 * we can check this by writing to the first page where it would then
	 * land in the now stale 2M page.
	 */

1513
	err = gpu_write(vma, ctx, dev_priv->engine[RCS], 0, 0xdeadbeaf);
M
Matthew Auld 已提交
1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530
	if (err)
		goto out_unpin;

	err = cpu_check(obj, 0, 0xdeadbeaf);

out_unpin:
	i915_vma_unpin(vma);
out_close:
	i915_vma_close(vma);
out_put:
	i915_gem_object_put(obj);

	return err;
}

static int igt_tmpfs_fallback(void *arg)
{
1531 1532
	struct i915_gem_context *ctx = arg;
	struct drm_i915_private *i915 = ctx->i915;
M
Matthew Auld 已提交
1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587
	struct vfsmount *gemfs = i915->mm.gemfs;
	struct i915_address_space *vm = ctx->ppgtt ? &ctx->ppgtt->base : &i915->ggtt.base;
	struct drm_i915_gem_object *obj;
	struct i915_vma *vma;
	u32 *vaddr;
	int err = 0;

	/*
	 * Make sure that we don't burst into a ball of flames upon falling back
	 * to tmpfs, which we rely on if on the off-chance we encouter a failure
	 * when setting up gemfs.
	 */

	i915->mm.gemfs = NULL;

	obj = i915_gem_object_create(i915, PAGE_SIZE);
	if (IS_ERR(obj)) {
		err = PTR_ERR(obj);
		goto out_restore;
	}

	vaddr = i915_gem_object_pin_map(obj, I915_MAP_WB);
	if (IS_ERR(vaddr)) {
		err = PTR_ERR(vaddr);
		goto out_put;
	}
	*vaddr = 0xdeadbeaf;

	i915_gem_object_unpin_map(obj);

	vma = i915_vma_instance(obj, vm, NULL);
	if (IS_ERR(vma)) {
		err = PTR_ERR(vma);
		goto out_put;
	}

	err = i915_vma_pin(vma, 0, 0, PIN_USER);
	if (err)
		goto out_close;

	err = igt_check_page_sizes(vma);

	i915_vma_unpin(vma);
out_close:
	i915_vma_close(vma);
out_put:
	i915_gem_object_put(obj);
out_restore:
	i915->mm.gemfs = gemfs;

	return err;
}

static int igt_shrink_thp(void *arg)
{
1588 1589
	struct i915_gem_context *ctx = arg;
	struct drm_i915_private *i915 = ctx->i915;
M
Matthew Auld 已提交
1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628
	struct i915_address_space *vm = ctx->ppgtt ? &ctx->ppgtt->base : &i915->ggtt.base;
	struct drm_i915_gem_object *obj;
	struct i915_vma *vma;
	unsigned int flags = PIN_USER;
	int err;

	/*
	 * Sanity check shrinking huge-paged object -- make sure nothing blows
	 * up.
	 */

	if (!igt_can_allocate_thp(i915)) {
		pr_info("missing THP support, skipping\n");
		return 0;
	}

	obj = i915_gem_object_create(i915, SZ_2M);
	if (IS_ERR(obj))
		return PTR_ERR(obj);

	vma = i915_vma_instance(obj, vm, NULL);
	if (IS_ERR(vma)) {
		err = PTR_ERR(vma);
		goto out_put;
	}

	err = i915_vma_pin(vma, 0, 0, flags);
	if (err)
		goto out_close;

	if (obj->mm.page_sizes.phys < I915_GTT_PAGE_SIZE_2M) {
		pr_info("failed to allocate THP, finishing test early\n");
		goto out_unpin;
	}

	err = igt_check_page_sizes(vma);
	if (err)
		goto out_unpin;

1629
	err = gpu_write(vma, ctx, i915->engine[RCS], 0, 0xdeadbeaf);
M
Matthew Auld 已提交
1630 1631 1632 1633 1634 1635 1636 1637 1638 1639
	if (err)
		goto out_unpin;

	i915_vma_unpin(vma);

	/*
	 * Now that the pages are *unpinned* shrink-all should invoke
	 * shmem to truncate our pages.
	 */
	i915_gem_shrink_all(i915);
1640
	if (i915_gem_object_has_pages(obj)) {
M
Matthew Auld 已提交
1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737
		pr_err("shrink-all didn't truncate the pages\n");
		err = -EINVAL;
		goto out_close;
	}

	if (obj->mm.page_sizes.sg || obj->mm.page_sizes.phys) {
		pr_err("residual page-size bits left\n");
		err = -EINVAL;
		goto out_close;
	}

	err = i915_vma_pin(vma, 0, 0, flags);
	if (err)
		goto out_close;

	err = cpu_check(obj, 0, 0xdeadbeaf);

out_unpin:
	i915_vma_unpin(vma);
out_close:
	i915_vma_close(vma);
out_put:
	i915_gem_object_put(obj);

	return err;
}

int i915_gem_huge_page_mock_selftests(void)
{
	static const struct i915_subtest tests[] = {
		SUBTEST(igt_mock_exhaust_device_supported_pages),
		SUBTEST(igt_mock_ppgtt_misaligned_dma),
		SUBTEST(igt_mock_ppgtt_huge_fill),
		SUBTEST(igt_mock_ppgtt_64K),
	};
	int saved_ppgtt = i915_modparams.enable_ppgtt;
	struct drm_i915_private *dev_priv;
	struct pci_dev *pdev;
	struct i915_hw_ppgtt *ppgtt;
	int err;

	dev_priv = mock_gem_device();
	if (!dev_priv)
		return -ENOMEM;

	/* Pretend to be a device which supports the 48b PPGTT */
	i915_modparams.enable_ppgtt = 3;

	pdev = dev_priv->drm.pdev;
	dma_coerce_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(39));

	mutex_lock(&dev_priv->drm.struct_mutex);
	ppgtt = i915_ppgtt_create(dev_priv, ERR_PTR(-ENODEV), "mock");
	if (IS_ERR(ppgtt)) {
		err = PTR_ERR(ppgtt);
		goto out_unlock;
	}

	if (!i915_vm_is_48bit(&ppgtt->base)) {
		pr_err("failed to create 48b PPGTT\n");
		err = -EINVAL;
		goto out_close;
	}

	/* If we were ever hit this then it's time to mock the 64K scratch */
	if (!i915_vm_has_scratch_64K(&ppgtt->base)) {
		pr_err("PPGTT missing 64K scratch page\n");
		err = -EINVAL;
		goto out_close;
	}

	err = i915_subtests(tests, ppgtt);

out_close:
	i915_ppgtt_close(&ppgtt->base);
	i915_ppgtt_put(ppgtt);

out_unlock:
	mutex_unlock(&dev_priv->drm.struct_mutex);

	i915_modparams.enable_ppgtt = saved_ppgtt;

	drm_dev_unref(&dev_priv->drm);

	return err;
}

int i915_gem_huge_page_live_selftests(struct drm_i915_private *dev_priv)
{
	static const struct i915_subtest tests[] = {
		SUBTEST(igt_shrink_thp),
		SUBTEST(igt_ppgtt_pin_update),
		SUBTEST(igt_tmpfs_fallback),
		SUBTEST(igt_ppgtt_exhaust_huge),
		SUBTEST(igt_ppgtt_gemfs_huge),
		SUBTEST(igt_ppgtt_internal_huge),
	};
1738 1739
	struct drm_file *file;
	struct i915_gem_context *ctx;
M
Matthew Auld 已提交
1740 1741 1742 1743 1744 1745 1746
	int err;

	if (!USES_PPGTT(dev_priv)) {
		pr_info("PPGTT not supported, skipping live-selftests\n");
		return 0;
	}

1747 1748 1749 1750
	file = mock_file(dev_priv);
	if (IS_ERR(file))
		return PTR_ERR(file);

M
Matthew Auld 已提交
1751
	mutex_lock(&dev_priv->drm.struct_mutex);
1752
	intel_runtime_pm_get(dev_priv);
1753 1754 1755 1756 1757 1758 1759 1760 1761 1762

	ctx = live_context(dev_priv, file);
	if (IS_ERR(ctx)) {
		err = PTR_ERR(ctx);
		goto out_unlock;
	}

	err = i915_subtests(tests, ctx);

out_unlock:
1763
	intel_runtime_pm_put(dev_priv);
M
Matthew Auld 已提交
1764 1765
	mutex_unlock(&dev_priv->drm.struct_mutex);

1766 1767
	mock_file_free(dev_priv, file);

M
Matthew Auld 已提交
1768 1769
	return err;
}