hugetlbpage.c 26.7 KB
Newer Older
L
Linus Torvalds 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
/*
 * PPC64 (POWER4) Huge TLB Page Support for Kernel.
 *
 * Copyright (C) 2003 David Gibson, IBM Corporation.
 *
 * Based on the IA-32 version:
 * Copyright (C) 2002, Rohit Seth <rohit.seth@intel.com>
 */

#include <linux/init.h>
#include <linux/fs.h>
#include <linux/mm.h>
#include <linux/hugetlb.h>
#include <linux/pagemap.h>
#include <linux/smp_lock.h>
#include <linux/slab.h>
#include <linux/err.h>
#include <linux/sysctl.h>
#include <asm/mman.h>
#include <asm/pgalloc.h>
#include <asm/tlb.h>
#include <asm/tlbflush.h>
#include <asm/mmu_context.h>
#include <asm/machdep.h>
#include <asm/cputable.h>
#include <asm/tlb.h>

#include <linux/sysctl.h>

30 31 32
#define NUM_LOW_AREAS	(0x100000000UL >> SID_SHIFT)
#define NUM_HIGH_AREAS	(PGTABLE_RANGE >> HTLB_AREA_SHIFT)

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
#ifdef CONFIG_PPC_64K_PAGES
#define HUGEPTE_INDEX_SIZE	(PMD_SHIFT-HPAGE_SHIFT)
#else
#define HUGEPTE_INDEX_SIZE	(PUD_SHIFT-HPAGE_SHIFT)
#endif
#define PTRS_PER_HUGEPTE	(1 << HUGEPTE_INDEX_SIZE)
#define HUGEPTE_TABLE_SIZE	(sizeof(pte_t) << HUGEPTE_INDEX_SIZE)

#define HUGEPD_SHIFT		(HPAGE_SHIFT + HUGEPTE_INDEX_SIZE)
#define HUGEPD_SIZE		(1UL << HUGEPD_SHIFT)
#define HUGEPD_MASK		(~(HUGEPD_SIZE-1))

#define huge_pgtable_cache	(pgtable_cache[HUGEPTE_CACHE_NUM])

/* Flag to mark huge PD pointers.  This means pmd_bad() and pud_bad()
 * will choke on pointers to hugepte tables, which is handy for
 * catching screwups early. */
#define HUGEPD_OK	0x1

typedef struct { unsigned long pd; } hugepd_t;

#define hugepd_none(hpd)	((hpd).pd == 0)

static inline pte_t *hugepd_page(hugepd_t hpd)
{
	BUG_ON(!(hpd.pd & HUGEPD_OK));
	return (pte_t *)(hpd.pd & ~HUGEPD_OK);
}

static inline pte_t *hugepte_offset(hugepd_t *hpdp, unsigned long addr)
{
	unsigned long idx = ((addr >> HPAGE_SHIFT) & (PTRS_PER_HUGEPTE-1));
	pte_t *dir = hugepd_page(*hpdp);

	return dir + idx;
}

static int __hugepte_alloc(struct mm_struct *mm, hugepd_t *hpdp,
			   unsigned long address)
{
	pte_t *new = kmem_cache_alloc(huge_pgtable_cache,
				      GFP_KERNEL|__GFP_REPEAT);

	if (! new)
		return -ENOMEM;

	spin_lock(&mm->page_table_lock);
	if (!hugepd_none(*hpdp))
		kmem_cache_free(huge_pgtable_cache, new);
	else
		hpdp->pd = (unsigned long)new | HUGEPD_OK;
	spin_unlock(&mm->page_table_lock);
	return 0;
}

88 89
/* Modelled after find_linux_pte() */
pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr)
L
Linus Torvalds 已提交
90
{
91 92
	pgd_t *pg;
	pud_t *pu;
L
Linus Torvalds 已提交
93

94
	BUG_ON(! in_hugepage_area(mm->context, addr));
L
Linus Torvalds 已提交
95

96 97 98 99 100 101
	addr &= HPAGE_MASK;

	pg = pgd_offset(mm, addr);
	if (!pgd_none(*pg)) {
		pu = pud_offset(pg, addr);
		if (!pud_none(*pu)) {
102
#ifdef CONFIG_PPC_64K_PAGES
103 104 105 106 107 108 109
			pmd_t *pm;
			pm = pmd_offset(pu, addr);
			if (!pmd_none(*pm))
				return hugepte_offset((hugepd_t *)pm, addr);
#else
			return hugepte_offset((hugepd_t *)pu, addr);
#endif
110 111
		}
	}
L
Linus Torvalds 已提交
112

113
	return NULL;
L
Linus Torvalds 已提交
114 115
}

116
pte_t *huge_pte_alloc(struct mm_struct *mm, unsigned long addr)
L
Linus Torvalds 已提交
117
{
118 119
	pgd_t *pg;
	pud_t *pu;
120
	hugepd_t *hpdp = NULL;
L
Linus Torvalds 已提交
121 122 123

	BUG_ON(! in_hugepage_area(mm->context, addr));

124
	addr &= HPAGE_MASK;
L
Linus Torvalds 已提交
125

126 127
	pg = pgd_offset(mm, addr);
	pu = pud_alloc(mm, pg, addr);
L
Linus Torvalds 已提交
128

129
	if (pu) {
130 131
#ifdef CONFIG_PPC_64K_PAGES
		pmd_t *pm;
132
		pm = pmd_alloc(mm, pu, addr);
133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155
		if (pm)
			hpdp = (hugepd_t *)pm;
#else
		hpdp = (hugepd_t *)pu;
#endif
	}

	if (! hpdp)
		return NULL;

	if (hugepd_none(*hpdp) && __hugepte_alloc(mm, hpdp, addr))
		return NULL;

	return hugepte_offset(hpdp, addr);
}

static void free_hugepte_range(struct mmu_gather *tlb, hugepd_t *hpdp)
{
	pte_t *hugepte = hugepd_page(*hpdp);

	hpdp->pd = 0;
	tlb->need_flush = 1;
	pgtable_free_tlb(tlb, pgtable_free_cache(hugepte, HUGEPTE_CACHE_NUM,
A
Adam Litke 已提交
156
						 PGF_CACHENUM_MASK));
157 158
}

159
#ifdef CONFIG_PPC_64K_PAGES
160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183
static void hugetlb_free_pmd_range(struct mmu_gather *tlb, pud_t *pud,
				   unsigned long addr, unsigned long end,
				   unsigned long floor, unsigned long ceiling)
{
	pmd_t *pmd;
	unsigned long next;
	unsigned long start;

	start = addr;
	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
		if (pmd_none(*pmd))
			continue;
		free_hugepte_range(tlb, (hugepd_t *)pmd);
	} while (pmd++, addr = next, addr != end);

	start &= PUD_MASK;
	if (start < floor)
		return;
	if (ceiling) {
		ceiling &= PUD_MASK;
		if (!ceiling)
			return;
L
Linus Torvalds 已提交
184
	}
185 186
	if (end - 1 > ceiling - 1)
		return;
L
Linus Torvalds 已提交
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
	pmd = pmd_offset(pud, start);
	pud_clear(pud);
	pmd_free_tlb(tlb, pmd);
}
#endif

static void hugetlb_free_pud_range(struct mmu_gather *tlb, pgd_t *pgd,
				   unsigned long addr, unsigned long end,
				   unsigned long floor, unsigned long ceiling)
{
	pud_t *pud;
	unsigned long next;
	unsigned long start;

	start = addr;
	pud = pud_offset(pgd, addr);
	do {
		next = pud_addr_end(addr, end);
#ifdef CONFIG_PPC_64K_PAGES
		if (pud_none_or_clear_bad(pud))
			continue;
		hugetlb_free_pmd_range(tlb, pud, addr, next, floor, ceiling);
#else
		if (pud_none(*pud))
			continue;
		free_hugepte_range(tlb, (hugepd_t *)pud);
#endif
	} while (pud++, addr = next, addr != end);

	start &= PGDIR_MASK;
	if (start < floor)
		return;
	if (ceiling) {
		ceiling &= PGDIR_MASK;
		if (!ceiling)
			return;
	}
	if (end - 1 > ceiling - 1)
		return;

	pud = pud_offset(pgd, start);
	pgd_clear(pgd);
	pud_free_tlb(tlb, pud);
}

/*
 * This function frees user-level page tables of a process.
 *
 * Must be called with pagetable lock held.
 */
void hugetlb_free_pgd_range(struct mmu_gather **tlb,
			    unsigned long addr, unsigned long end,
			    unsigned long floor, unsigned long ceiling)
{
	pgd_t *pgd;
	unsigned long next;
	unsigned long start;

	/*
	 * Comments below take from the normal free_pgd_range().  They
	 * apply here too.  The tests against HUGEPD_MASK below are
	 * essential, because we *don't* test for this at the bottom
	 * level.  Without them we'll attempt to free a hugepte table
	 * when we unmap just part of it, even if there are other
	 * active mappings using it.
	 *
	 * The next few lines have given us lots of grief...
	 *
	 * Why are we testing HUGEPD* at this top level?  Because
	 * often there will be no work to do at all, and we'd prefer
	 * not to go all the way down to the bottom just to discover
	 * that.
	 *
	 * Why all these "- 1"s?  Because 0 represents both the bottom
	 * of the address space and the top of it (using -1 for the
	 * top wouldn't help much: the masks would do the wrong thing).
	 * The rule is that addr 0 and floor 0 refer to the bottom of
	 * the address space, but end 0 and ceiling 0 refer to the top
	 * Comparisons need to use "end - 1" and "ceiling - 1" (though
	 * that end 0 case should be mythical).
	 *
	 * Wherever addr is brought up or ceiling brought down, we
	 * must be careful to reject "the opposite 0" before it
	 * confuses the subsequent tests.  But what about where end is
	 * brought down by HUGEPD_SIZE below? no, end can't go down to
	 * 0 there.
	 *
	 * Whereas we round start (addr) and ceiling down, by different
	 * masks at different levels, in order to test whether a table
	 * now has no other vmas using it, so can be freed, we don't
	 * bother to round floor or end up - the tests don't need that.
	 */

	addr &= HUGEPD_MASK;
	if (addr < floor) {
		addr += HUGEPD_SIZE;
		if (!addr)
			return;
	}
	if (ceiling) {
		ceiling &= HUGEPD_MASK;
		if (!ceiling)
			return;
	}
	if (end - 1 > ceiling - 1)
		end -= HUGEPD_SIZE;
	if (addr > end - 1)
		return;

	start = addr;
	pgd = pgd_offset((*tlb)->mm, addr);
	do {
		BUG_ON(! in_hugepage_area((*tlb)->mm->context, addr));
		next = pgd_addr_end(addr, end);
		if (pgd_none_or_clear_bad(pgd))
			continue;
		hugetlb_free_pud_range(*tlb, pgd, addr, next, floor, ceiling);
	} while (pgd++, addr = next, addr != end);
L
Linus Torvalds 已提交
306 307
}

308 309 310 311
void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
		     pte_t *ptep, pte_t pte)
{
	if (pte_present(*ptep)) {
312 313 314 315 316 317
		/* We open-code pte_clear because we need to pass the right
		 * argument to hpte_update (huge / !huge)
		 */
		unsigned long old = pte_update(ptep, ~0UL);
		if (old & _PAGE_HASHPTE)
			hpte_update(mm, addr & HPAGE_MASK, ptep, old, 1);
318 319
		flush_tlb_pending();
	}
320
	*ptep = __pte(pte_val(pte) & ~_PAGE_HPTEFLAGS);
L
Linus Torvalds 已提交
321 322
}

323 324
pte_t huge_ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
			      pte_t *ptep)
L
Linus Torvalds 已提交
325
{
326
	unsigned long old = pte_update(ptep, ~0UL);
L
Linus Torvalds 已提交
327

328
	if (old & _PAGE_HASHPTE)
329 330
		hpte_update(mm, addr & HPAGE_MASK, ptep, old, 1);
	*ptep = __pte(0);
L
Linus Torvalds 已提交
331

332
	return __pte(old);
L
Linus Torvalds 已提交
333 334
}

335 336 337 338 339
struct slb_flush_info {
	struct mm_struct *mm;
	u16 newareas;
};

340
static void flush_low_segments(void *parm)
L
Linus Torvalds 已提交
341
{
342
	struct slb_flush_info *fi = parm;
L
Linus Torvalds 已提交
343 344
	unsigned long i;

345 346 347 348
	BUILD_BUG_ON((sizeof(fi->newareas)*8) != NUM_LOW_AREAS);

	if (current->active_mm != fi->mm)
		return;
L
Linus Torvalds 已提交
349

350 351 352 353 354
	/* Only need to do anything if this CPU is working in the same
	 * mm as the one which has changed */

	/* update the paca copy of the context struct */
	get_paca()->context = current->active_mm->context;
355

356
	asm volatile("isync" : : : "memory");
357
	for (i = 0; i < NUM_LOW_AREAS; i++) {
358
		if (! (fi->newareas & (1U << i)))
L
Linus Torvalds 已提交
359
			continue;
360 361
		asm volatile("slbie %0"
			     : : "r" ((i << SID_SHIFT) | SLBIE_C));
L
Linus Torvalds 已提交
362 363 364 365
	}
	asm volatile("isync" : : : "memory");
}

366 367
static void flush_high_segments(void *parm)
{
368
	struct slb_flush_info *fi = parm;
369 370 371
	unsigned long i, j;


372 373 374 375 376 377 378
	BUILD_BUG_ON((sizeof(fi->newareas)*8) != NUM_HIGH_AREAS);

	if (current->active_mm != fi->mm)
		return;

	/* Only need to do anything if this CPU is working in the same
	 * mm as the one which has changed */
379

380 381 382 383
	/* update the paca copy of the context struct */
	get_paca()->context = current->active_mm->context;

	asm volatile("isync" : : : "memory");
384
	for (i = 0; i < NUM_HIGH_AREAS; i++) {
385
		if (! (fi->newareas & (1U << i)))
386 387 388
			continue;
		for (j = 0; j < (1UL << (HTLB_AREA_SHIFT-SID_SHIFT)); j++)
			asm volatile("slbie %0"
389
				     :: "r" (((i << HTLB_AREA_SHIFT)
390
					      + (j << SID_SHIFT)) | SLBIE_C));
391 392 393 394 395
	}
	asm volatile("isync" : : : "memory");
}

static int prepare_low_area_for_htlb(struct mm_struct *mm, unsigned long area)
L
Linus Torvalds 已提交
396
{
397 398
	unsigned long start = area << SID_SHIFT;
	unsigned long end = (area+1) << SID_SHIFT;
L
Linus Torvalds 已提交
399 400
	struct vm_area_struct *vma;

401
	BUG_ON(area >= NUM_LOW_AREAS);
L
Linus Torvalds 已提交
402 403 404 405 406 407 408 409 410

	/* Check no VMAs are in the region */
	vma = find_vma(mm, start);
	if (vma && (vma->vm_start < end))
		return -EBUSY;

	return 0;
}

411 412 413 414 415 416 417 418
static int prepare_high_area_for_htlb(struct mm_struct *mm, unsigned long area)
{
	unsigned long start = area << HTLB_AREA_SHIFT;
	unsigned long end = (area+1) << HTLB_AREA_SHIFT;
	struct vm_area_struct *vma;

	BUG_ON(area >= NUM_HIGH_AREAS);

419 420 421 422 423 424
	/* Hack, so that each addresses is controlled by exactly one
	 * of the high or low area bitmaps, the first high area starts
	 * at 4GB, not 0 */
	if (start == 0)
		start = 0x100000000UL;

425 426 427 428 429 430 431 432 433
	/* Check no VMAs are in the region */
	vma = find_vma(mm, start);
	if (vma && (vma->vm_start < end))
		return -EBUSY;

	return 0;
}

static int open_low_hpage_areas(struct mm_struct *mm, u16 newareas)
L
Linus Torvalds 已提交
434 435
{
	unsigned long i;
436
	struct slb_flush_info fi;
L
Linus Torvalds 已提交
437

438 439 440 441 442
	BUILD_BUG_ON((sizeof(newareas)*8) != NUM_LOW_AREAS);
	BUILD_BUG_ON((sizeof(mm->context.low_htlb_areas)*8) != NUM_LOW_AREAS);

	newareas &= ~(mm->context.low_htlb_areas);
	if (! newareas)
L
Linus Torvalds 已提交
443 444
		return 0; /* The segments we want are already open */

445 446 447 448 449 450 451 452 453 454
	for (i = 0; i < NUM_LOW_AREAS; i++)
		if ((1 << i) & newareas)
			if (prepare_low_area_for_htlb(mm, i) != 0)
				return -EBUSY;

	mm->context.low_htlb_areas |= newareas;

	/* the context change must make it to memory before the flush,
	 * so that further SLB misses do the right thing. */
	mb();
455 456 457 458

	fi.mm = mm;
	fi.newareas = newareas;
	on_each_cpu(flush_low_segments, &fi, 0, 1);
459 460 461 462 463 464

	return 0;
}

static int open_high_hpage_areas(struct mm_struct *mm, u16 newareas)
{
465
	struct slb_flush_info fi;
466 467 468 469 470 471 472 473 474 475 476 477 478
	unsigned long i;

	BUILD_BUG_ON((sizeof(newareas)*8) != NUM_HIGH_AREAS);
	BUILD_BUG_ON((sizeof(mm->context.high_htlb_areas)*8)
		     != NUM_HIGH_AREAS);

	newareas &= ~(mm->context.high_htlb_areas);
	if (! newareas)
		return 0; /* The areas we want are already open */

	for (i = 0; i < NUM_HIGH_AREAS; i++)
		if ((1 << i) & newareas)
			if (prepare_high_area_for_htlb(mm, i) != 0)
L
Linus Torvalds 已提交
479 480
				return -EBUSY;

481
	mm->context.high_htlb_areas |= newareas;
L
Linus Torvalds 已提交
482 483 484 485

	/* the context change must make it to memory before the flush,
	 * so that further SLB misses do the right thing. */
	mb();
486 487 488 489

	fi.mm = mm;
	fi.newareas = newareas;
	on_each_cpu(flush_high_segments, &fi, 0, 1);
L
Linus Torvalds 已提交
490 491 492 493

	return 0;
}

494
int prepare_hugepage_range(unsigned long addr, unsigned long len, pgoff_t pgoff)
L
Linus Torvalds 已提交
495
{
496
	int err = 0;
497

498 499 500 501 502
	if (pgoff & (~HPAGE_MASK >> PAGE_SHIFT))
		return -EINVAL;
	if (len & ~HPAGE_MASK)
		return -EINVAL;
	if (addr & ~HPAGE_MASK)
503 504
		return -EINVAL;

505
	if (addr < 0x100000000UL)
506
		err = open_low_hpage_areas(current->mm,
L
Linus Torvalds 已提交
507
					  LOW_ESID_MASK(addr, len));
508
	if ((addr + len) > 0x100000000UL)
509 510 511 512 513 514 515
		err = open_high_hpage_areas(current->mm,
					    HTLB_AREA_MASK(addr, len));
	if (err) {
		printk(KERN_DEBUG "prepare_hugepage_range(%lx, %lx)"
		       " failed (lowmask: 0x%04hx, highmask: 0x%04hx)\n",
		       addr, len,
		       LOW_ESID_MASK(addr, len), HTLB_AREA_MASK(addr, len));
L
Linus Torvalds 已提交
516 517 518
		return err;
	}

519
	return 0;
L
Linus Torvalds 已提交
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
}

struct page *
follow_huge_addr(struct mm_struct *mm, unsigned long address, int write)
{
	pte_t *ptep;
	struct page *page;

	if (! in_hugepage_area(mm->context, address))
		return ERR_PTR(-EINVAL);

	ptep = huge_pte_offset(mm, address);
	page = pte_page(*ptep);
	if (page)
		page += (address % HPAGE_SIZE) / PAGE_SIZE;

	return page;
}

int pmd_huge(pmd_t pmd)
{
	return 0;
}

struct page *
follow_huge_pmd(struct mm_struct *mm, unsigned long address,
		pmd_t *pmd, int write)
{
	BUG();
	return NULL;
}

/* Because we have an exclusive hugepage region which lies within the
 * normal user address space, we have to take special measures to make
 * non-huge mmap()s evade the hugepage reserved regions. */
unsigned long arch_get_unmapped_area(struct file *filp, unsigned long addr,
				     unsigned long len, unsigned long pgoff,
				     unsigned long flags)
{
	struct mm_struct *mm = current->mm;
	struct vm_area_struct *vma;
	unsigned long start_addr;

	if (len > TASK_SIZE)
		return -ENOMEM;

	if (addr) {
		addr = PAGE_ALIGN(addr);
		vma = find_vma(mm, addr);
		if (((TASK_SIZE - len) >= addr)
		    && (!vma || (addr+len) <= vma->vm_start)
		    && !is_hugepage_only_range(mm, addr,len))
			return addr;
	}
574 575 576 577 578 579
	if (len > mm->cached_hole_size) {
	        start_addr = addr = mm->free_area_cache;
	} else {
	        start_addr = addr = TASK_UNMAPPED_BASE;
	        mm->cached_hole_size = 0;
	}
L
Linus Torvalds 已提交
580 581 582 583 584 585 586 587 588 589 590

full_search:
	vma = find_vma(mm, addr);
	while (TASK_SIZE - len >= addr) {
		BUG_ON(vma && (addr >= vma->vm_end));

		if (touches_hugepage_low_range(mm, addr, len)) {
			addr = ALIGN(addr+1, 1<<SID_SHIFT);
			vma = find_vma(mm, addr);
			continue;
		}
591 592
		if (touches_hugepage_high_range(mm, addr, len)) {
			addr = ALIGN(addr+1, 1UL<<HTLB_AREA_SHIFT);
L
Linus Torvalds 已提交
593 594 595 596 597 598 599 600 601 602
			vma = find_vma(mm, addr);
			continue;
		}
		if (!vma || addr + len <= vma->vm_start) {
			/*
			 * Remember the place where we stopped the search:
			 */
			mm->free_area_cache = addr + len;
			return addr;
		}
603 604
		if (addr + mm->cached_hole_size < vma->vm_start)
		        mm->cached_hole_size = vma->vm_start - addr;
L
Linus Torvalds 已提交
605 606 607 608 609 610 611
		addr = vma->vm_end;
		vma = vma->vm_next;
	}

	/* Make sure we didn't miss any holes */
	if (start_addr != TASK_UNMAPPED_BASE) {
		start_addr = addr = TASK_UNMAPPED_BASE;
612
		mm->cached_hole_size = 0;
L
Linus Torvalds 已提交
613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633
		goto full_search;
	}
	return -ENOMEM;
}

/*
 * This mmap-allocator allocates new areas top-down from below the
 * stack's low limit (the base):
 *
 * Because we have an exclusive hugepage region which lies within the
 * normal user address space, we have to take special measures to make
 * non-huge mmap()s evade the hugepage reserved regions.
 */
unsigned long
arch_get_unmapped_area_topdown(struct file *filp, const unsigned long addr0,
			  const unsigned long len, const unsigned long pgoff,
			  const unsigned long flags)
{
	struct vm_area_struct *vma, *prev_vma;
	struct mm_struct *mm = current->mm;
	unsigned long base = mm->mmap_base, addr = addr0;
634
	unsigned long largest_hole = mm->cached_hole_size;
L
Linus Torvalds 已提交
635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654
	int first_time = 1;

	/* requested length too big for entire address space */
	if (len > TASK_SIZE)
		return -ENOMEM;

	/* dont allow allocations above current base */
	if (mm->free_area_cache > base)
		mm->free_area_cache = base;

	/* requesting a specific address */
	if (addr) {
		addr = PAGE_ALIGN(addr);
		vma = find_vma(mm, addr);
		if (TASK_SIZE - len >= addr &&
				(!vma || addr + len <= vma->vm_start)
				&& !is_hugepage_only_range(mm, addr,len))
			return addr;
	}

655 656 657 658
	if (len <= largest_hole) {
	        largest_hole = 0;
		mm->free_area_cache = base;
	}
L
Linus Torvalds 已提交
659 660 661 662 663 664 665 666 667 668 669 670
try_again:
	/* make sure it can fit in the remaining address space */
	if (mm->free_area_cache < len)
		goto fail;

	/* either no address requested or cant fit in requested address hole */
	addr = (mm->free_area_cache - len) & PAGE_MASK;
	do {
hugepage_recheck:
		if (touches_hugepage_low_range(mm, addr, len)) {
			addr = (addr & ((~0) << SID_SHIFT)) - len;
			goto hugepage_recheck;
671 672 673
		} else if (touches_hugepage_high_range(mm, addr, len)) {
			addr = (addr & ((~0UL) << HTLB_AREA_SHIFT)) - len;
			goto hugepage_recheck;
L
Linus Torvalds 已提交
674 675 676 677 678 679 680 681 682 683 684 685 686 687
		}

		/*
		 * Lookup failure means no vma is above this address,
		 * i.e. return with success:
		 */
 	 	if (!(vma = find_vma_prev(mm, addr, &prev_vma)))
			return addr;

		/*
		 * new region fits between prev_vma->vm_end and
		 * vma->vm_start, use it:
		 */
		if (addr+len <= vma->vm_start &&
688
		          (!prev_vma || (addr >= prev_vma->vm_end))) {
L
Linus Torvalds 已提交
689
			/* remember the address as a hint for next time */
690 691 692
		        mm->cached_hole_size = largest_hole;
		        return (mm->free_area_cache = addr);
		} else {
L
Linus Torvalds 已提交
693
			/* pull free_area_cache down to the first hole */
694
		        if (mm->free_area_cache == vma->vm_end) {
L
Linus Torvalds 已提交
695
				mm->free_area_cache = vma->vm_start;
696 697 698 699 700 701 702
				mm->cached_hole_size = largest_hole;
			}
		}

		/* remember the largest hole we saw so far */
		if (addr + largest_hole < vma->vm_start)
		        largest_hole = vma->vm_start - addr;
L
Linus Torvalds 已提交
703 704 705 706 707 708 709 710 711 712 713 714

		/* try just below the current vma->vm_start */
		addr = vma->vm_start-len;
	} while (len <= vma->vm_start);

fail:
	/*
	 * if hint left us with no space for the requested
	 * mapping then try again:
	 */
	if (first_time) {
		mm->free_area_cache = base;
715
		largest_hole = 0;
L
Linus Torvalds 已提交
716 717 718 719 720 721 722 723 724 725
		first_time = 0;
		goto try_again;
	}
	/*
	 * A failed mmap() very likely causes application failure,
	 * so fall back to the bottom-up function here. This scenario
	 * can happen with large stack limits and large mmap()
	 * allocations.
	 */
	mm->free_area_cache = TASK_UNMAPPED_BASE;
726
	mm->cached_hole_size = ~0UL;
L
Linus Torvalds 已提交
727 728 729 730 731
	addr = arch_get_unmapped_area(filp, addr0, len, pgoff, flags);
	/*
	 * Restore the topdown base:
	 */
	mm->free_area_cache = base;
732
	mm->cached_hole_size = ~0UL;
L
Linus Torvalds 已提交
733 734 735 736

	return addr;
}

737 738 739 740 741 742 743 744 745 746 747
static int htlb_check_hinted_area(unsigned long addr, unsigned long len)
{
	struct vm_area_struct *vma;

	vma = find_vma(current->mm, addr);
	if (!vma || ((addr + len) <= vma->vm_start))
		return 0;

	return -ENOMEM;
}

L
Linus Torvalds 已提交
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
static unsigned long htlb_get_low_area(unsigned long len, u16 segmask)
{
	unsigned long addr = 0;
	struct vm_area_struct *vma;

	vma = find_vma(current->mm, addr);
	while (addr + len <= 0x100000000UL) {
		BUG_ON(vma && (addr >= vma->vm_end)); /* invariant */

		if (! __within_hugepage_low_range(addr, len, segmask)) {
			addr = ALIGN(addr+1, 1<<SID_SHIFT);
			vma = find_vma(current->mm, addr);
			continue;
		}

		if (!vma || (addr + len) <= vma->vm_start)
			return addr;
		addr = ALIGN(vma->vm_end, HPAGE_SIZE);
		/* Depending on segmask this might not be a confirmed
		 * hugepage region, so the ALIGN could have skipped
		 * some VMAs */
		vma = find_vma(current->mm, addr);
	}

	return -ENOMEM;
}

775
static unsigned long htlb_get_high_area(unsigned long len, u16 areamask)
L
Linus Torvalds 已提交
776
{
777
	unsigned long addr = 0x100000000UL;
L
Linus Torvalds 已提交
778 779 780
	struct vm_area_struct *vma;

	vma = find_vma(current->mm, addr);
781
	while (addr + len <= TASK_SIZE_USER64) {
L
Linus Torvalds 已提交
782
		BUG_ON(vma && (addr >= vma->vm_end)); /* invariant */
783 784 785 786 787 788

		if (! __within_hugepage_high_range(addr, len, areamask)) {
			addr = ALIGN(addr+1, 1UL<<HTLB_AREA_SHIFT);
			vma = find_vma(current->mm, addr);
			continue;
		}
L
Linus Torvalds 已提交
789 790 791 792

		if (!vma || (addr + len) <= vma->vm_start)
			return addr;
		addr = ALIGN(vma->vm_end, HPAGE_SIZE);
793 794 795 796
		/* Depending on segmask this might not be a confirmed
		 * hugepage region, so the ALIGN could have skipped
		 * some VMAs */
		vma = find_vma(current->mm, addr);
L
Linus Torvalds 已提交
797 798 799 800 801 802 803 804 805
	}

	return -ENOMEM;
}

unsigned long hugetlb_get_unmapped_area(struct file *file, unsigned long addr,
					unsigned long len, unsigned long pgoff,
					unsigned long flags)
{
806 807 808
	int lastshift;
	u16 areamask, curareas;

809 810
	if (HPAGE_SHIFT == 0)
		return -EINVAL;
L
Linus Torvalds 已提交
811 812 813 814 815 816
	if (len & ~HPAGE_MASK)
		return -EINVAL;

	if (!cpu_has_feature(CPU_FTR_16M_PAGE))
		return -EINVAL;

817 818 819
	/* Paranoia, caller should have dealt with this */
	BUG_ON((addr + len)  < addr);

L
Linus Torvalds 已提交
820
	if (test_thread_flag(TIF_32BIT)) {
821 822 823
		/* Paranoia, caller should have dealt with this */
		BUG_ON((addr + len) > 0x100000000UL);

824
		curareas = current->mm->context.low_htlb_areas;
L
Linus Torvalds 已提交
825

826 827 828 829 830 831 832 833
		/* First see if we can use the hint address */
		if (addr && (htlb_check_hinted_area(addr, len) == 0)) {
			areamask = LOW_ESID_MASK(addr, len);
			if (open_low_hpage_areas(current->mm, areamask) == 0)
				return addr;
		}

		/* Next see if we can map in the existing low areas */
834
		addr = htlb_get_low_area(len, curareas);
L
Linus Torvalds 已提交
835 836 837
		if (addr != -ENOMEM)
			return addr;

838
		/* Finally go looking for areas to open */
839 840 841 842
		lastshift = 0;
		for (areamask = LOW_ESID_MASK(0x100000000UL-len, len);
		     ! lastshift; areamask >>=1) {
			if (areamask & 1)
L
Linus Torvalds 已提交
843 844
				lastshift = 1;

845
			addr = htlb_get_low_area(len, curareas | areamask);
L
Linus Torvalds 已提交
846
			if ((addr != -ENOMEM)
847
			    && open_low_hpage_areas(current->mm, areamask) == 0)
L
Linus Torvalds 已提交
848 849 850
				return addr;
		}
	} else {
851 852
		curareas = current->mm->context.high_htlb_areas;

853 854 855 856 857 858 859 860 861 862 863
		/* First see if we can use the hint address */
		/* We discourage 64-bit processes from doing hugepage
		 * mappings below 4GB (must use MAP_FIXED) */
		if ((addr >= 0x100000000UL)
		    && (htlb_check_hinted_area(addr, len) == 0)) {
			areamask = HTLB_AREA_MASK(addr, len);
			if (open_high_hpage_areas(current->mm, areamask) == 0)
				return addr;
		}

		/* Next see if we can map in the existing high areas */
864 865 866 867
		addr = htlb_get_high_area(len, curareas);
		if (addr != -ENOMEM)
			return addr;

868
		/* Finally go looking for areas to open */
869 870 871 872 873 874 875 876 877 878 879
		lastshift = 0;
		for (areamask = HTLB_AREA_MASK(TASK_SIZE_USER64-len, len);
		     ! lastshift; areamask >>=1) {
			if (areamask & 1)
				lastshift = 1;

			addr = htlb_get_high_area(len, curareas | areamask);
			if ((addr != -ENOMEM)
			    && open_high_hpage_areas(current->mm, areamask) == 0)
				return addr;
		}
L
Linus Torvalds 已提交
880
	}
881 882 883
	printk(KERN_DEBUG "hugetlb_get_unmapped_area() unable to open"
	       " enough areas\n");
	return -ENOMEM;
L
Linus Torvalds 已提交
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
/*
 * Called by asm hashtable.S for doing lazy icache flush
 */
static unsigned int hash_huge_page_do_lazy_icache(unsigned long rflags,
						  pte_t pte, int trap)
{
	struct page *page;
	int i;

	if (!pfn_valid(pte_pfn(pte)))
		return rflags;

	page = pte_page(pte);

	/* page is dirty */
	if (!test_bit(PG_arch_1, &page->flags) && !PageReserved(page)) {
		if (trap == 0x400) {
			for (i = 0; i < (HPAGE_SIZE / PAGE_SIZE); i++)
				__flush_dcache_icache(page_address(page+i));
			set_bit(PG_arch_1, &page->flags);
		} else {
			rflags |= HPTE_R_N;
		}
	}
	return rflags;
}

L
Linus Torvalds 已提交
913
int hash_huge_page(struct mm_struct *mm, unsigned long access,
914 915
		   unsigned long ea, unsigned long vsid, int local,
		   unsigned long trap)
L
Linus Torvalds 已提交
916 917
{
	pte_t *ptep;
918 919
	unsigned long old_pte, new_pte;
	unsigned long va, rflags, pa;
L
Linus Torvalds 已提交
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
	long slot;
	int err = 1;

	ptep = huge_pte_offset(mm, ea);

	/* Search the Linux page table for a match with va */
	va = (vsid << 28) | (ea & 0x0fffffff);

	/*
	 * If no pte found or not present, send the problem up to
	 * do_page_fault
	 */
	if (unlikely(!ptep || pte_none(*ptep)))
		goto out;

	/* 
	 * Check the user's access rights to the page.  If access should be
	 * prevented then send the problem up to do_page_fault.
	 */
	if (unlikely(access & ~pte_val(*ptep)))
		goto out;
	/*
	 * At this point, we have a pte (old_pte) which can be used to build
	 * or update an HPTE. There are 2 cases:
	 *
	 * 1. There is a valid (present) pte with no associated HPTE (this is 
	 *	the most common case)
	 * 2. There is a valid (present) pte with an associated HPTE. The
	 *	current values of the pp bits in the HPTE prevent access
	 *	because we are doing software DIRTY bit management and the
	 *	page is currently not DIRTY. 
	 */


954 955 956 957 958 959 960 961 962 963
	do {
		old_pte = pte_val(*ptep);
		if (old_pte & _PAGE_BUSY)
			goto out;
		new_pte = old_pte | _PAGE_BUSY |
			_PAGE_ACCESSED | _PAGE_HASHPTE;
	} while(old_pte != __cmpxchg_u64((unsigned long *)ptep,
					 old_pte, new_pte));

	rflags = 0x2 | (!(new_pte & _PAGE_RW));
L
Linus Torvalds 已提交
964
 	/* _PAGE_EXEC -> HW_NO_EXEC since it's inverted */
965
	rflags |= ((new_pte & _PAGE_EXEC) ? 0 : HPTE_R_N);
966 967 968 969 970
	if (!cpu_has_feature(CPU_FTR_COHERENT_ICACHE))
		/* No CPU has hugepages but lacks no execute, so we
		 * don't need to worry about that case */
		rflags = hash_huge_page_do_lazy_icache(rflags, __pte(old_pte),
						       trap);
L
Linus Torvalds 已提交
971 972

	/* Check if pte already has an hpte (case 2) */
973
	if (unlikely(old_pte & _PAGE_HASHPTE)) {
L
Linus Torvalds 已提交
974 975 976
		/* There MIGHT be an HPTE for this pte */
		unsigned long hash, slot;

977 978
		hash = hpt_hash(va, HPAGE_SHIFT);
		if (old_pte & _PAGE_F_SECOND)
L
Linus Torvalds 已提交
979 980
			hash = ~hash;
		slot = (hash & htab_hash_mask) * HPTES_PER_GROUP;
981
		slot += (old_pte & _PAGE_F_GIX) >> 12;
L
Linus Torvalds 已提交
982

983 984
		if (ppc_md.hpte_updatepp(slot, rflags, va, mmu_huge_psize,
					 local) == -1)
985
			old_pte &= ~_PAGE_HPTEFLAGS;
L
Linus Torvalds 已提交
986 987
	}

988 989
	if (likely(!(old_pte & _PAGE_HASHPTE))) {
		unsigned long hash = hpt_hash(va, HPAGE_SHIFT);
L
Linus Torvalds 已提交
990 991
		unsigned long hpte_group;

992
		pa = pte_pfn(__pte(old_pte)) << PAGE_SHIFT;
L
Linus Torvalds 已提交
993 994 995 996 997

repeat:
		hpte_group = ((hash & htab_hash_mask) *
			      HPTES_PER_GROUP) & ~0x7UL;

998 999
		/* clear HPTE slot informations in new PTE */
		new_pte = (new_pte & ~_PAGE_HPTEFLAGS) | _PAGE_HASHPTE;
L
Linus Torvalds 已提交
1000 1001 1002

		/* Add in WIMG bits */
		/* XXX We should store these in the pte */
1003
		/* --BenH: I think they are ... */
1004
		rflags |= _PAGE_COHERENT;
L
Linus Torvalds 已提交
1005

1006 1007 1008
		/* Insert into the hash table, primary slot */
		slot = ppc_md.hpte_insert(hpte_group, va, pa, rflags, 0,
					  mmu_huge_psize);
L
Linus Torvalds 已提交
1009 1010 1011

		/* Primary is full, try the secondary */
		if (unlikely(slot == -1)) {
1012
			new_pte |= _PAGE_F_SECOND;
L
Linus Torvalds 已提交
1013 1014
			hpte_group = ((~hash & htab_hash_mask) *
				      HPTES_PER_GROUP) & ~0x7UL; 
1015
			slot = ppc_md.hpte_insert(hpte_group, va, pa, rflags,
1016
						  HPTE_V_SECONDARY,
1017
						  mmu_huge_psize);
L
Linus Torvalds 已提交
1018 1019
			if (slot == -1) {
				if (mftb() & 0x1)
1020 1021
					hpte_group = ((hash & htab_hash_mask) *
						      HPTES_PER_GROUP)&~0x7UL;
L
Linus Torvalds 已提交
1022 1023 1024 1025 1026 1027 1028 1029 1030

				ppc_md.hpte_remove(hpte_group);
				goto repeat;
                        }
		}

		if (unlikely(slot == -2))
			panic("hash_huge_page: pte_insert failed\n");

1031
		new_pte |= (slot << 12) & _PAGE_F_GIX;
L
Linus Torvalds 已提交
1032 1033
	}

1034
	/*
H
Hugh Dickins 已提交
1035
	 * No need to use ldarx/stdcx here
1036 1037 1038
	 */
	*ptep = __pte(new_pte & ~_PAGE_BUSY);

L
Linus Torvalds 已提交
1039 1040 1041 1042 1043
	err = 0;

 out:
	return err;
}
1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067

static void zero_ctor(void *addr, kmem_cache_t *cache, unsigned long flags)
{
	memset(addr, 0, kmem_cache_size(cache));
}

static int __init hugetlbpage_init(void)
{
	if (!cpu_has_feature(CPU_FTR_16M_PAGE))
		return -ENODEV;

	huge_pgtable_cache = kmem_cache_create("hugepte_cache",
					       HUGEPTE_TABLE_SIZE,
					       HUGEPTE_TABLE_SIZE,
					       SLAB_HWCACHE_ALIGN |
					       SLAB_MUST_HWCACHE_ALIGN,
					       zero_ctor, NULL);
	if (! huge_pgtable_cache)
		panic("hugetlbpage_init(): could not create hugepte cache\n");

	return 0;
}

module_init(hugetlbpage_init);