pgtable.c 26.8 KB
Newer Older
1
/*
2
 *    Copyright IBM Corp. 2007, 2011
3 4 5 6 7 8
 *    Author(s): Martin Schwidefsky <schwidefsky@de.ibm.com>
 */

#include <linux/sched.h>
#include <linux/kernel.h>
#include <linux/errno.h>
9
#include <linux/gfp.h>
10 11 12 13 14 15 16 17
#include <linux/mm.h>
#include <linux/swap.h>
#include <linux/smp.h>
#include <linux/highmem.h>
#include <linux/pagemap.h>
#include <linux/spinlock.h>
#include <linux/module.h>
#include <linux/quicklist.h>
18
#include <linux/rcupdate.h>
19
#include <linux/slab.h>
20 21 22 23 24

#include <asm/pgtable.h>
#include <asm/pgalloc.h>
#include <asm/tlb.h>
#include <asm/tlbflush.h>
25
#include <asm/mmu_context.h>
26 27 28

#ifndef CONFIG_64BIT
#define ALLOC_ORDER	1
29
#define FRAG_MASK	0x0f
30 31
#else
#define ALLOC_ORDER	2
32
#define FRAG_MASK	0x03
33 34
#endif

35

36
unsigned long *crst_table_alloc(struct mm_struct *mm)
37 38 39 40 41 42 43 44
{
	struct page *page = alloc_pages(GFP_KERNEL, ALLOC_ORDER);

	if (!page)
		return NULL;
	return (unsigned long *) page_to_phys(page);
}

45 46
void crst_table_free(struct mm_struct *mm, unsigned long *table)
{
47
	free_pages((unsigned long) table, ALLOC_ORDER);
48 49
}

50 51 52 53 54 55 56 57
#ifdef CONFIG_64BIT
int crst_table_upgrade(struct mm_struct *mm, unsigned long limit)
{
	unsigned long *table, *pgd;
	unsigned long entry;

	BUG_ON(limit > (1UL << 53));
repeat:
58
	table = crst_table_alloc(mm);
59 60
	if (!table)
		return -ENOMEM;
61
	spin_lock_bh(&mm->page_table_lock);
62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79
	if (mm->context.asce_limit < limit) {
		pgd = (unsigned long *) mm->pgd;
		if (mm->context.asce_limit <= (1UL << 31)) {
			entry = _REGION3_ENTRY_EMPTY;
			mm->context.asce_limit = 1UL << 42;
			mm->context.asce_bits = _ASCE_TABLE_LENGTH |
						_ASCE_USER_BITS |
						_ASCE_TYPE_REGION3;
		} else {
			entry = _REGION2_ENTRY_EMPTY;
			mm->context.asce_limit = 1UL << 53;
			mm->context.asce_bits = _ASCE_TABLE_LENGTH |
						_ASCE_USER_BITS |
						_ASCE_TYPE_REGION2;
		}
		crst_table_init(table, entry);
		pgd_populate(mm, (pgd_t *) table, (pud_t *) pgd);
		mm->pgd = (pgd_t *) table;
80
		mm->task_size = mm->context.asce_limit;
81 82
		table = NULL;
	}
83
	spin_unlock_bh(&mm->page_table_lock);
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
	if (table)
		crst_table_free(mm, table);
	if (mm->context.asce_limit < limit)
		goto repeat;
	return 0;
}

void crst_table_downgrade(struct mm_struct *mm, unsigned long limit)
{
	pgd_t *pgd;

	while (mm->context.asce_limit > limit) {
		pgd = mm->pgd;
		switch (pgd_val(*pgd) & _REGION_ENTRY_TYPE_MASK) {
		case _REGION_ENTRY_TYPE_R2:
			mm->context.asce_limit = 1UL << 42;
			mm->context.asce_bits = _ASCE_TABLE_LENGTH |
						_ASCE_USER_BITS |
						_ASCE_TYPE_REGION3;
			break;
		case _REGION_ENTRY_TYPE_R3:
			mm->context.asce_limit = 1UL << 31;
			mm->context.asce_bits = _ASCE_TABLE_LENGTH |
						_ASCE_USER_BITS |
						_ASCE_TYPE_SEGMENT;
			break;
		default:
			BUG();
		}
		mm->pgd = (pgd_t *) (pgd_val(*pgd) & _REGION_ENTRY_ORIGIN);
114
		mm->task_size = mm->context.asce_limit;
115 116 117 118 119
		crst_table_free(mm, (unsigned long *) pgd);
	}
}
#endif

120 121 122 123 124 125 126 127 128
#ifdef CONFIG_PGSTE

/**
 * gmap_alloc - allocate a guest address space
 * @mm: pointer to the parent mm_struct
 *
 * Returns a guest address space structure.
 */
struct gmap *gmap_alloc(struct mm_struct *mm)
129
{
130 131 132
	struct gmap *gmap;
	struct page *page;
	unsigned long *table;
133

134 135 136 137 138 139 140 141 142 143 144 145
	gmap = kzalloc(sizeof(struct gmap), GFP_KERNEL);
	if (!gmap)
		goto out;
	INIT_LIST_HEAD(&gmap->crst_list);
	gmap->mm = mm;
	page = alloc_pages(GFP_KERNEL, ALLOC_ORDER);
	if (!page)
		goto out_free;
	list_add(&page->lru, &gmap->crst_list);
	table = (unsigned long *) page_to_phys(page);
	crst_table_init(table, _REGION1_ENTRY_EMPTY);
	gmap->table = table;
146 147
	gmap->asce = _ASCE_TYPE_REGION1 | _ASCE_TABLE_LENGTH |
		     _ASCE_USER_BITS | __pa(table);
148 149 150 151 152 153 154
	list_add(&gmap->list, &mm->context.gmap_list);
	return gmap;

out_free:
	kfree(gmap);
out:
	return NULL;
155
}
156
EXPORT_SYMBOL_GPL(gmap_alloc);
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
static int gmap_unlink_segment(struct gmap *gmap, unsigned long *table)
{
	struct gmap_pgtable *mp;
	struct gmap_rmap *rmap;
	struct page *page;

	if (*table & _SEGMENT_ENTRY_INV)
		return 0;
	page = pfn_to_page(*table >> PAGE_SHIFT);
	mp = (struct gmap_pgtable *) page->index;
	list_for_each_entry(rmap, &mp->mapper, list) {
		if (rmap->entry != table)
			continue;
		list_del(&rmap->list);
		kfree(rmap);
		break;
	}
	*table = _SEGMENT_ENTRY_INV | _SEGMENT_ENTRY_RO | mp->vmaddr;
	return 1;
}

static void gmap_flush_tlb(struct gmap *gmap)
{
	if (MACHINE_HAS_IDTE)
		__tlb_flush_idte((unsigned long) gmap->table |
				 _ASCE_TYPE_REGION1);
	else
		__tlb_flush_global();
}

/**
 * gmap_free - free a guest address space
 * @gmap: pointer to the guest address space structure
191
 */
192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207
void gmap_free(struct gmap *gmap)
{
	struct page *page, *next;
	unsigned long *table;
	int i;


	/* Flush tlb. */
	if (MACHINE_HAS_IDTE)
		__tlb_flush_idte((unsigned long) gmap->table |
				 _ASCE_TYPE_REGION1);
	else
		__tlb_flush_global();

	/* Free all segment & region tables. */
	down_read(&gmap->mm->mmap_sem);
208
	spin_lock(&gmap->mm->page_table_lock);
209 210 211 212 213 214 215 216
	list_for_each_entry_safe(page, next, &gmap->crst_list, lru) {
		table = (unsigned long *) page_to_phys(page);
		if ((*table & _REGION_ENTRY_TYPE_MASK) == 0)
			/* Remove gmap rmap structures for segment table. */
			for (i = 0; i < PTRS_PER_PMD; i++, table++)
				gmap_unlink_segment(gmap, table);
		__free_pages(page, ALLOC_ORDER);
	}
217
	spin_unlock(&gmap->mm->page_table_lock);
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
	up_read(&gmap->mm->mmap_sem);
	list_del(&gmap->list);
	kfree(gmap);
}
EXPORT_SYMBOL_GPL(gmap_free);

/**
 * gmap_enable - switch primary space to the guest address space
 * @gmap: pointer to the guest address space structure
 */
void gmap_enable(struct gmap *gmap)
{
	S390_lowcore.gmap = (unsigned long) gmap;
}
EXPORT_SYMBOL_GPL(gmap_enable);

/**
 * gmap_disable - switch back to the standard primary address space
 * @gmap: pointer to the guest address space structure
 */
void gmap_disable(struct gmap *gmap)
{
	S390_lowcore.gmap = 0UL;
}
EXPORT_SYMBOL_GPL(gmap_disable);

244 245 246
/*
 * gmap_alloc_table is assumed to be called with mmap_sem held
 */
247 248 249 250 251 252
static int gmap_alloc_table(struct gmap *gmap,
			       unsigned long *table, unsigned long init)
{
	struct page *page;
	unsigned long *new;

253 254
	/* since we dont free the gmap table until gmap_free we can unlock */
	spin_unlock(&gmap->mm->page_table_lock);
255
	page = alloc_pages(GFP_KERNEL, ALLOC_ORDER);
256
	spin_lock(&gmap->mm->page_table_lock);
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
	if (!page)
		return -ENOMEM;
	new = (unsigned long *) page_to_phys(page);
	crst_table_init(new, init);
	if (*table & _REGION_ENTRY_INV) {
		list_add(&page->lru, &gmap->crst_list);
		*table = (unsigned long) new | _REGION_ENTRY_LENGTH |
			(*table & _REGION_ENTRY_TYPE_MASK);
	} else
		__free_pages(page, ALLOC_ORDER);
	return 0;
}

/**
 * gmap_unmap_segment - unmap segment from the guest address space
 * @gmap: pointer to the guest address space structure
 * @addr: address in the guest address space
 * @len: length of the memory area to unmap
 *
 * Returns 0 if the unmap succeded, -EINVAL if not.
 */
int gmap_unmap_segment(struct gmap *gmap, unsigned long to, unsigned long len)
{
	unsigned long *table;
	unsigned long off;
	int flush;

	if ((to | len) & (PMD_SIZE - 1))
		return -EINVAL;
	if (len == 0 || to + len < to)
		return -EINVAL;

	flush = 0;
	down_read(&gmap->mm->mmap_sem);
291
	spin_lock(&gmap->mm->page_table_lock);
292 293 294 295
	for (off = 0; off < len; off += PMD_SIZE) {
		/* Walk the guest addr space page table */
		table = gmap->table + (((to + off) >> 53) & 0x7ff);
		if (*table & _REGION_ENTRY_INV)
296
			goto out;
297 298 299
		table = (unsigned long *)(*table & _REGION_ENTRY_ORIGIN);
		table = table + (((to + off) >> 42) & 0x7ff);
		if (*table & _REGION_ENTRY_INV)
300
			goto out;
301 302 303
		table = (unsigned long *)(*table & _REGION_ENTRY_ORIGIN);
		table = table + (((to + off) >> 31) & 0x7ff);
		if (*table & _REGION_ENTRY_INV)
304
			goto out;
305 306 307 308 309 310 311
		table = (unsigned long *)(*table & _REGION_ENTRY_ORIGIN);
		table = table + (((to + off) >> 20) & 0x7ff);

		/* Clear segment table entry in guest address space. */
		flush |= gmap_unlink_segment(gmap, table);
		*table = _SEGMENT_ENTRY_INV;
	}
312
out:
313
	spin_unlock(&gmap->mm->page_table_lock);
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
	up_read(&gmap->mm->mmap_sem);
	if (flush)
		gmap_flush_tlb(gmap);
	return 0;
}
EXPORT_SYMBOL_GPL(gmap_unmap_segment);

/**
 * gmap_mmap_segment - map a segment to the guest address space
 * @gmap: pointer to the guest address space structure
 * @from: source address in the parent address space
 * @to: target address in the guest address space
 *
 * Returns 0 if the mmap succeded, -EINVAL or -ENOMEM if not.
 */
int gmap_map_segment(struct gmap *gmap, unsigned long from,
		     unsigned long to, unsigned long len)
{
	unsigned long *table;
	unsigned long off;
	int flush;

	if ((from | to | len) & (PMD_SIZE - 1))
		return -EINVAL;
	if (len == 0 || from + len > PGDIR_SIZE ||
	    from + len < from || to + len < to)
		return -EINVAL;

	flush = 0;
	down_read(&gmap->mm->mmap_sem);
344
	spin_lock(&gmap->mm->page_table_lock);
345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367
	for (off = 0; off < len; off += PMD_SIZE) {
		/* Walk the gmap address space page table */
		table = gmap->table + (((to + off) >> 53) & 0x7ff);
		if ((*table & _REGION_ENTRY_INV) &&
		    gmap_alloc_table(gmap, table, _REGION2_ENTRY_EMPTY))
			goto out_unmap;
		table = (unsigned long *)(*table & _REGION_ENTRY_ORIGIN);
		table = table + (((to + off) >> 42) & 0x7ff);
		if ((*table & _REGION_ENTRY_INV) &&
		    gmap_alloc_table(gmap, table, _REGION3_ENTRY_EMPTY))
			goto out_unmap;
		table = (unsigned long *)(*table & _REGION_ENTRY_ORIGIN);
		table = table + (((to + off) >> 31) & 0x7ff);
		if ((*table & _REGION_ENTRY_INV) &&
		    gmap_alloc_table(gmap, table, _SEGMENT_ENTRY_EMPTY))
			goto out_unmap;
		table = (unsigned long *) (*table & _REGION_ENTRY_ORIGIN);
		table = table + (((to + off) >> 20) & 0x7ff);

		/* Store 'from' address in an invalid segment table entry. */
		flush |= gmap_unlink_segment(gmap, table);
		*table = _SEGMENT_ENTRY_INV | _SEGMENT_ENTRY_RO | (from + off);
	}
368
	spin_unlock(&gmap->mm->page_table_lock);
369 370 371 372 373 374
	up_read(&gmap->mm->mmap_sem);
	if (flush)
		gmap_flush_tlb(gmap);
	return 0;

out_unmap:
375
	spin_unlock(&gmap->mm->page_table_lock);
376 377 378 379 380 381
	up_read(&gmap->mm->mmap_sem);
	gmap_unmap_segment(gmap, to, len);
	return -ENOMEM;
}
EXPORT_SYMBOL_GPL(gmap_map_segment);

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
static unsigned long *gmap_table_walk(unsigned long address, struct gmap *gmap)
{
	unsigned long *table;

	table = gmap->table + ((address >> 53) & 0x7ff);
	if (unlikely(*table & _REGION_ENTRY_INV))
		return ERR_PTR(-EFAULT);
	table = (unsigned long *)(*table & _REGION_ENTRY_ORIGIN);
	table = table + ((address >> 42) & 0x7ff);
	if (unlikely(*table & _REGION_ENTRY_INV))
		return ERR_PTR(-EFAULT);
	table = (unsigned long *)(*table & _REGION_ENTRY_ORIGIN);
	table = table + ((address >> 31) & 0x7ff);
	if (unlikely(*table & _REGION_ENTRY_INV))
		return ERR_PTR(-EFAULT);
	table = (unsigned long *)(*table & _REGION_ENTRY_ORIGIN);
	table = table + ((address >> 20) & 0x7ff);
	return table;
}

/**
 * __gmap_translate - translate a guest address to a user space address
 * @address: guest address
 * @gmap: pointer to guest mapping meta data structure
 *
 * Returns user space address which corresponds to the guest address or
 * -EFAULT if no such mapping exists.
 * This function does not establish potentially missing page table entries.
 * The mmap_sem of the mm that belongs to the address space must be held
 * when this function gets called.
 */
unsigned long __gmap_translate(unsigned long address, struct gmap *gmap)
{
	unsigned long *segment_ptr, vmaddr, segment;
	struct gmap_pgtable *mp;
	struct page *page;

	current->thread.gmap_addr = address;
	segment_ptr = gmap_table_walk(address, gmap);
	if (IS_ERR(segment_ptr))
		return PTR_ERR(segment_ptr);
	/* Convert the gmap address to an mm address. */
	segment = *segment_ptr;
	if (!(segment & _SEGMENT_ENTRY_INV)) {
		page = pfn_to_page(segment >> PAGE_SHIFT);
		mp = (struct gmap_pgtable *) page->index;
		return mp->vmaddr | (address & ~PMD_MASK);
	} else if (segment & _SEGMENT_ENTRY_RO) {
		vmaddr = segment & _SEGMENT_ENTRY_ORIGIN;
		return vmaddr | (address & ~PMD_MASK);
	}
	return -EFAULT;
}
EXPORT_SYMBOL_GPL(__gmap_translate);

/**
 * gmap_translate - translate a guest address to a user space address
 * @address: guest address
 * @gmap: pointer to guest mapping meta data structure
 *
 * Returns user space address which corresponds to the guest address or
 * -EFAULT if no such mapping exists.
 * This function does not establish potentially missing page table entries.
 */
unsigned long gmap_translate(unsigned long address, struct gmap *gmap)
{
	unsigned long rc;

	down_read(&gmap->mm->mmap_sem);
	rc = __gmap_translate(address, gmap);
	up_read(&gmap->mm->mmap_sem);
	return rc;
}
EXPORT_SYMBOL_GPL(gmap_translate);

457 458 459 460
/*
 * this function is assumed to be called with mmap_sem held
 */
unsigned long __gmap_fault(unsigned long address, struct gmap *gmap)
461
{
462 463
	unsigned long *segment_ptr, vmaddr, segment;
	struct vm_area_struct *vma;
464 465
	struct gmap_pgtable *mp;
	struct gmap_rmap *rmap;
466
	struct mm_struct *mm;
467 468 469 470 471 472
	struct page *page;
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;

	current->thread.gmap_addr = address;
473 474
	segment_ptr = gmap_table_walk(address, gmap);
	if (IS_ERR(segment_ptr))
475 476
		return -EFAULT;
	/* Convert the gmap address to an mm address. */
477 478
	segment = *segment_ptr;
	if (!(segment & _SEGMENT_ENTRY_INV)) {
479 480 481 482
		page = pfn_to_page(segment >> PAGE_SHIFT);
		mp = (struct gmap_pgtable *) page->index;
		return mp->vmaddr | (address & ~PMD_MASK);
	} else if (segment & _SEGMENT_ENTRY_RO) {
483
		mm = gmap->mm;
484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506
		vmaddr = segment & _SEGMENT_ENTRY_ORIGIN;
		vma = find_vma(mm, vmaddr);
		if (!vma || vma->vm_start > vmaddr)
			return -EFAULT;

		/* Walk the parent mm page table */
		pgd = pgd_offset(mm, vmaddr);
		pud = pud_alloc(mm, pgd, vmaddr);
		if (!pud)
			return -ENOMEM;
		pmd = pmd_alloc(mm, pud, vmaddr);
		if (!pmd)
			return -ENOMEM;
		if (!pmd_present(*pmd) &&
		    __pte_alloc(mm, vma, pmd, vmaddr))
			return -ENOMEM;
		/* pmd now points to a valid segment table entry. */
		rmap = kmalloc(sizeof(*rmap), GFP_KERNEL|__GFP_REPEAT);
		if (!rmap)
			return -ENOMEM;
		/* Link gmap segment table entry location to page table. */
		page = pmd_page(*pmd);
		mp = (struct gmap_pgtable *) page->index;
507
		rmap->entry = segment_ptr;
508
		spin_lock(&mm->page_table_lock);
509
		list_add(&rmap->list, &mp->mapper);
510
		spin_unlock(&mm->page_table_lock);
511
		/* Set gmap segment table entry to page table. */
512
		*segment_ptr = pmd_val(*pmd) & PAGE_MASK;
513 514 515
		return vmaddr | (address & ~PMD_MASK);
	}
	return -EFAULT;
516 517 518 519 520 521 522 523 524
}

unsigned long gmap_fault(unsigned long address, struct gmap *gmap)
{
	unsigned long rc;

	down_read(&gmap->mm->mmap_sem);
	rc = __gmap_fault(address, gmap);
	up_read(&gmap->mm->mmap_sem);
525

526
	return rc;
527 528 529
}
EXPORT_SYMBOL_GPL(gmap_fault);

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
void gmap_discard(unsigned long from, unsigned long to, struct gmap *gmap)
{

	unsigned long *table, address, size;
	struct vm_area_struct *vma;
	struct gmap_pgtable *mp;
	struct page *page;

	down_read(&gmap->mm->mmap_sem);
	address = from;
	while (address < to) {
		/* Walk the gmap address space page table */
		table = gmap->table + ((address >> 53) & 0x7ff);
		if (unlikely(*table & _REGION_ENTRY_INV)) {
			address = (address + PMD_SIZE) & PMD_MASK;
			continue;
		}
		table = (unsigned long *)(*table & _REGION_ENTRY_ORIGIN);
		table = table + ((address >> 42) & 0x7ff);
		if (unlikely(*table & _REGION_ENTRY_INV)) {
			address = (address + PMD_SIZE) & PMD_MASK;
			continue;
		}
		table = (unsigned long *)(*table & _REGION_ENTRY_ORIGIN);
		table = table + ((address >> 31) & 0x7ff);
		if (unlikely(*table & _REGION_ENTRY_INV)) {
			address = (address + PMD_SIZE) & PMD_MASK;
			continue;
		}
		table = (unsigned long *)(*table & _REGION_ENTRY_ORIGIN);
		table = table + ((address >> 20) & 0x7ff);
		if (unlikely(*table & _SEGMENT_ENTRY_INV)) {
			address = (address + PMD_SIZE) & PMD_MASK;
			continue;
		}
		page = pfn_to_page(*table >> PAGE_SHIFT);
		mp = (struct gmap_pgtable *) page->index;
		vma = find_vma(gmap->mm, mp->vmaddr);
		size = min(to - address, PMD_SIZE - (address & ~PMD_MASK));
		zap_page_range(vma, mp->vmaddr | (address & ~PMD_MASK),
			       size, NULL);
		address = (address + PMD_SIZE) & PMD_MASK;
	}
	up_read(&gmap->mm->mmap_sem);
}
EXPORT_SYMBOL_GPL(gmap_discard);

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
void gmap_unmap_notifier(struct mm_struct *mm, unsigned long *table)
{
	struct gmap_rmap *rmap, *next;
	struct gmap_pgtable *mp;
	struct page *page;
	int flush;

	flush = 0;
	spin_lock(&mm->page_table_lock);
	page = pfn_to_page(__pa(table) >> PAGE_SHIFT);
	mp = (struct gmap_pgtable *) page->index;
	list_for_each_entry_safe(rmap, next, &mp->mapper, list) {
		*rmap->entry =
			_SEGMENT_ENTRY_INV | _SEGMENT_ENTRY_RO | mp->vmaddr;
		list_del(&rmap->list);
		kfree(rmap);
		flush = 1;
	}
	spin_unlock(&mm->page_table_lock);
	if (flush)
		__tlb_flush_global();
}

static inline unsigned long *page_table_alloc_pgste(struct mm_struct *mm,
						    unsigned long vmaddr)
602 603 604
{
	struct page *page;
	unsigned long *table;
605
	struct gmap_pgtable *mp;
606 607 608 609

	page = alloc_page(GFP_KERNEL|__GFP_REPEAT);
	if (!page)
		return NULL;
610 611 612 613 614
	mp = kmalloc(sizeof(*mp), GFP_KERNEL|__GFP_REPEAT);
	if (!mp) {
		__free_page(page);
		return NULL;
	}
615
	pgtable_page_ctor(page);
616 617 618
	mp->vmaddr = vmaddr & PMD_MASK;
	INIT_LIST_HEAD(&mp->mapper);
	page->index = (unsigned long) mp;
619 620 621 622 623 624 625 626 627 628
	atomic_set(&page->_mapcount, 3);
	table = (unsigned long *) page_to_phys(page);
	clear_table(table, _PAGE_TYPE_EMPTY, PAGE_SIZE/2);
	clear_table(table + PTRS_PER_PTE, 0, PAGE_SIZE/2);
	return table;
}

static inline void page_table_free_pgste(unsigned long *table)
{
	struct page *page;
629
	struct gmap_pgtable *mp;
630 631

	page = pfn_to_page(__pa(table) >> PAGE_SHIFT);
632 633
	mp = (struct gmap_pgtable *) page->index;
	BUG_ON(!list_empty(&mp->mapper));
634
	pgtable_page_dtor(page);
635
	atomic_set(&page->_mapcount, -1);
636
	kfree(mp);
637 638 639
	__free_page(page);
}

640 641 642 643 644
#else /* CONFIG_PGSTE */

static inline unsigned long *page_table_alloc_pgste(struct mm_struct *mm,
						    unsigned long vmaddr)
{
645
	return NULL;
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
}

static inline void page_table_free_pgste(unsigned long *table)
{
}

static inline void gmap_unmap_notifier(struct mm_struct *mm,
					  unsigned long *table)
{
}

#endif /* CONFIG_PGSTE */

static inline unsigned int atomic_xor_bits(atomic_t *v, unsigned int bits)
{
	unsigned int old, new;

	do {
		old = atomic_read(v);
		new = old ^ bits;
	} while (atomic_cmpxchg(v, old, new) != old);
	return new;
}

/*
 * page table entry allocation/free routines.
 */
unsigned long *page_table_alloc(struct mm_struct *mm, unsigned long vmaddr)
674
{
675 676
	unsigned long *uninitialized_var(table);
	struct page *uninitialized_var(page);
677
	unsigned int mask, bit;
678

679
	if (mm_has_pgste(mm))
680
		return page_table_alloc_pgste(mm, vmaddr);
681
	/* Allocate fragments of a 4K page as 1K/2K page table */
682
	spin_lock_bh(&mm->context.list_lock);
683
	mask = FRAG_MASK;
684 685 686
	if (!list_empty(&mm->context.pgtable_list)) {
		page = list_first_entry(&mm->context.pgtable_list,
					struct page, lru);
687 688 689
		table = (unsigned long *) page_to_phys(page);
		mask = atomic_read(&page->_mapcount);
		mask = mask | (mask >> 4);
690
	}
691
	if ((mask & FRAG_MASK) == FRAG_MASK) {
692
		spin_unlock_bh(&mm->context.list_lock);
693 694
		page = alloc_page(GFP_KERNEL|__GFP_REPEAT);
		if (!page)
695
			return NULL;
696
		pgtable_page_ctor(page);
697
		atomic_set(&page->_mapcount, 1);
698
		table = (unsigned long *) page_to_phys(page);
699
		clear_table(table, _PAGE_TYPE_EMPTY, PAGE_SIZE);
700
		spin_lock_bh(&mm->context.list_lock);
701
		list_add(&page->lru, &mm->context.pgtable_list);
702 703 704 705 706 707
	} else {
		for (bit = 1; mask & bit; bit <<= 1)
			table += PTRS_PER_PTE;
		mask = atomic_xor_bits(&page->_mapcount, bit);
		if ((mask & FRAG_MASK) == FRAG_MASK)
			list_del(&page->lru);
708
	}
709
	spin_unlock_bh(&mm->context.list_lock);
710 711 712
	return table;
}

713
void page_table_free(struct mm_struct *mm, unsigned long *table)
714 715
{
	struct page *page;
716
	unsigned int bit, mask;
717

718 719
	if (mm_has_pgste(mm)) {
		gmap_unmap_notifier(mm, table);
720
		return page_table_free_pgste(table);
721
	}
722
	/* Free 1K/2K page table fragment of a 4K page */
723
	page = pfn_to_page(__pa(table) >> PAGE_SHIFT);
724 725 726 727 728 729 730 731 732
	bit = 1 << ((__pa(table) & ~PAGE_MASK)/(PTRS_PER_PTE*sizeof(pte_t)));
	spin_lock_bh(&mm->context.list_lock);
	if ((atomic_read(&page->_mapcount) & FRAG_MASK) != FRAG_MASK)
		list_del(&page->lru);
	mask = atomic_xor_bits(&page->_mapcount, bit);
	if (mask & FRAG_MASK)
		list_add(&page->lru, &mm->context.pgtable_list);
	spin_unlock_bh(&mm->context.list_lock);
	if (mask == 0) {
733
		pgtable_page_dtor(page);
734
		atomic_set(&page->_mapcount, -1);
735 736 737 738
		__free_page(page);
	}
}

739
static void __page_table_free_rcu(void *table, unsigned bit)
740
{
741
	struct page *page;
742

743 744 745
	if (bit == FRAG_MASK)
		return page_table_free_pgste(table);
	/* Free 1K/2K page table fragment of a 4K page */
746
	page = pfn_to_page(__pa(table) >> PAGE_SHIFT);
747
	if (atomic_xor_bits(&page->_mapcount, bit) == 0) {
748
		pgtable_page_dtor(page);
749
		atomic_set(&page->_mapcount, -1);
750 751 752
		__free_page(page);
	}
}
753

754
void page_table_free_rcu(struct mmu_gather *tlb, unsigned long *table)
755
{
756
	struct mm_struct *mm;
757
	struct page *page;
758
	unsigned int bit, mask;
759

760 761
	mm = tlb->mm;
	if (mm_has_pgste(mm)) {
762
		gmap_unmap_notifier(mm, table);
763 764 765
		table = (unsigned long *) (__pa(table) | FRAG_MASK);
		tlb_remove_table(tlb, table);
		return;
766
	}
767
	bit = 1 << ((__pa(table) & ~PAGE_MASK) / (PTRS_PER_PTE*sizeof(pte_t)));
768 769
	page = pfn_to_page(__pa(table) >> PAGE_SHIFT);
	spin_lock_bh(&mm->context.list_lock);
770 771 772 773 774
	if ((atomic_read(&page->_mapcount) & FRAG_MASK) != FRAG_MASK)
		list_del(&page->lru);
	mask = atomic_xor_bits(&page->_mapcount, bit | (bit << 4));
	if (mask & FRAG_MASK)
		list_add_tail(&page->lru, &mm->context.pgtable_list);
775
	spin_unlock_bh(&mm->context.list_lock);
776 777 778 779 780 781
	table = (unsigned long *) (__pa(table) | (bit << 4));
	tlb_remove_table(tlb, table);
}

void __tlb_remove_table(void *_table)
{
782 783 784
	const unsigned long mask = (FRAG_MASK << 4) | FRAG_MASK;
	void *table = (void *)((unsigned long) _table & ~mask);
	unsigned type = (unsigned long) _table & mask;
785 786 787 788 789

	if (type)
		__page_table_free_rcu(table, type);
	else
		free_pages((unsigned long) table, ALLOC_ORDER);
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
static void tlb_remove_table_smp_sync(void *arg)
{
	/* Simply deliver the interrupt */
}

static void tlb_remove_table_one(void *table)
{
	/*
	 * This isn't an RCU grace period and hence the page-tables cannot be
	 * assumed to be actually RCU-freed.
	 *
	 * It is however sufficient for software page-table walkers that rely
	 * on IRQ disabling. See the comment near struct mmu_table_batch.
	 */
	smp_call_function(tlb_remove_table_smp_sync, NULL, 1);
	__tlb_remove_table(table);
}

static void tlb_remove_table_rcu(struct rcu_head *head)
{
	struct mmu_table_batch *batch;
	int i;

	batch = container_of(head, struct mmu_table_batch, rcu);

	for (i = 0; i < batch->nr; i++)
		__tlb_remove_table(batch->tables[i]);

	free_page((unsigned long)batch);
}

void tlb_table_flush(struct mmu_gather *tlb)
{
	struct mmu_table_batch **batch = &tlb->batch;

	if (*batch) {
		__tlb_flush_mm(tlb->mm);
		call_rcu_sched(&(*batch)->rcu, tlb_remove_table_rcu);
		*batch = NULL;
	}
}

void tlb_remove_table(struct mmu_gather *tlb, void *table)
{
	struct mmu_table_batch **batch = &tlb->batch;

	if (*batch == NULL) {
		*batch = (struct mmu_table_batch *)
			__get_free_page(GFP_NOWAIT | __GFP_NOWARN);
		if (*batch == NULL) {
			__tlb_flush_mm(tlb->mm);
			tlb_remove_table_one(table);
			return;
		}
		(*batch)->nr = 0;
	}
	(*batch)->tables[(*batch)->nr++] = table;
	if ((*batch)->nr == MAX_TABLE_BATCH)
		tlb_table_flush(tlb);
}
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
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
void thp_split_vma(struct vm_area_struct *vma)
{
	unsigned long addr;
	struct page *page;

	for (addr = vma->vm_start; addr < vma->vm_end; addr += PAGE_SIZE) {
		page = follow_page(vma, addr, FOLL_SPLIT);
	}
}

void thp_split_mm(struct mm_struct *mm)
{
	struct vm_area_struct *vma = mm->mmap;

	while (vma != NULL) {
		thp_split_vma(vma);
		vma->vm_flags &= ~VM_HUGEPAGE;
		vma->vm_flags |= VM_NOHUGEPAGE;
		vma = vma->vm_next;
	}
}
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */

877 878 879 880 881 882
/*
 * switch on pgstes for its userspace process (for kvm)
 */
int s390_enable_sie(void)
{
	struct task_struct *tsk = current;
883
	struct mm_struct *mm, *old_mm;
884

885
	/* Do we have switched amode? If no, we cannot do sie */
886
	if (s390_user_mode == HOME_SPACE_MODE)
887 888
		return -EINVAL;

889
	/* Do we have pgstes? if yes, we are done */
890
	if (mm_has_pgste(tsk->mm))
891
		return 0;
892

893 894
	/* lets check if we are allowed to replace the mm */
	task_lock(tsk);
895
	if (!tsk->mm || atomic_read(&tsk->mm->mm_users) > 1 ||
896 897 898 899
#ifdef CONFIG_AIO
	    !hlist_empty(&tsk->mm->ioctx_list) ||
#endif
	    tsk->mm != tsk->active_mm) {
900 901 902 903
		task_unlock(tsk);
		return -EINVAL;
	}
	task_unlock(tsk);
904

905 906
	/* we copy the mm and let dup_mm create the page tables with_pgstes */
	tsk->mm->context.alloc_pgste = 1;
907 908
	/* make sure that both mms have a correct rss state */
	sync_mm_rss(tsk->mm);
909
	mm = dup_mm(tsk);
910
	tsk->mm->context.alloc_pgste = 0;
911
	if (!mm)
912 913
		return -ENOMEM;

914 915 916 917 918 919
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
	/* split thp mappings and disable thp for future mappings */
	thp_split_mm(mm);
	mm->def_flags |= VM_NOHUGEPAGE;
#endif

920
	/* Now lets check again if something happened */
921 922
	task_lock(tsk);
	if (!tsk->mm || atomic_read(&tsk->mm->mm_users) > 1 ||
923 924 925 926
#ifdef CONFIG_AIO
	    !hlist_empty(&tsk->mm->ioctx_list) ||
#endif
	    tsk->mm != tsk->active_mm) {
927 928 929 930 931 932 933
		mmput(mm);
		task_unlock(tsk);
		return -EINVAL;
	}

	/* ok, we are alone. No ptrace, no threads, etc. */
	old_mm = tsk->mm;
934 935 936
	tsk->mm = tsk->active_mm = mm;
	preempt_disable();
	update_mm(mm, tsk);
937 938
	atomic_inc(&mm->context.attach_count);
	atomic_dec(&old_mm->context.attach_count);
939
	cpumask_set_cpu(smp_processor_id(), mm_cpumask(mm));
940 941
	preempt_enable();
	task_unlock(tsk);
942 943
	mmput(old_mm);
	return 0;
944 945
}
EXPORT_SYMBOL_GPL(s390_enable_sie);
946

947
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969
int pmdp_clear_flush_young(struct vm_area_struct *vma, unsigned long address,
			   pmd_t *pmdp)
{
	VM_BUG_ON(address & ~HPAGE_PMD_MASK);
	/* No need to flush TLB
	 * On s390 reference bits are in storage key and never in TLB */
	return pmdp_test_and_clear_young(vma, address, pmdp);
}

int pmdp_set_access_flags(struct vm_area_struct *vma,
			  unsigned long address, pmd_t *pmdp,
			  pmd_t entry, int dirty)
{
	VM_BUG_ON(address & ~HPAGE_PMD_MASK);

	if (pmd_same(*pmdp, entry))
		return 0;
	pmdp_invalidate(vma, address, pmdp);
	set_pmd_at(vma->vm_mm, address, pmdp, entry);
	return 1;
}

970 971 972 973 974 975 976 977 978 979 980 981 982 983 984
static void pmdp_splitting_flush_sync(void *arg)
{
	/* Simply deliver the interrupt */
}

void pmdp_splitting_flush(struct vm_area_struct *vma, unsigned long address,
			  pmd_t *pmdp)
{
	VM_BUG_ON(address & ~HPAGE_PMD_MASK);
	if (!test_and_set_bit(_SEGMENT_ENTRY_SPLIT_BIT,
			      (unsigned long *) pmdp)) {
		/* need to serialize against gup-fast (IRQ disabled) */
		smp_call_function(pmdp_splitting_flush_sync, NULL, 1);
	}
}
985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022

void pgtable_trans_huge_deposit(struct mm_struct *mm, pgtable_t pgtable)
{
	struct list_head *lh = (struct list_head *) pgtable;

	assert_spin_locked(&mm->page_table_lock);

	/* FIFO */
	if (!mm->pmd_huge_pte)
		INIT_LIST_HEAD(lh);
	else
		list_add(lh, (struct list_head *) mm->pmd_huge_pte);
	mm->pmd_huge_pte = pgtable;
}

pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm)
{
	struct list_head *lh;
	pgtable_t pgtable;
	pte_t *ptep;

	assert_spin_locked(&mm->page_table_lock);

	/* FIFO */
	pgtable = mm->pmd_huge_pte;
	lh = (struct list_head *) pgtable;
	if (list_empty(lh))
		mm->pmd_huge_pte = NULL;
	else {
		mm->pmd_huge_pte = (pgtable_t) lh->next;
		list_del(lh);
	}
	ptep = (pte_t *) pgtable;
	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
	ptep++;
	pte_val(*ptep) = _PAGE_TYPE_EMPTY;
	return pgtable;
}
1023
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
新手
引导
客服 返回
顶部