hugetlbpage.c 18.1 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
/* Modelled after find_linux_pte() */
pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr)
L
Linus Torvalds 已提交
35
{
36 37 38 39
	pgd_t *pg;
	pud_t *pu;
	pmd_t *pm;
	pte_t *pt;
L
Linus Torvalds 已提交
40

41
	BUG_ON(! in_hugepage_area(mm->context, addr));
L
Linus Torvalds 已提交
42

43 44 45 46 47 48 49 50 51 52 53 54 55
	addr &= HPAGE_MASK;

	pg = pgd_offset(mm, addr);
	if (!pgd_none(*pg)) {
		pu = pud_offset(pg, addr);
		if (!pud_none(*pu)) {
			pm = pmd_offset(pu, addr);
			pt = (pte_t *)pm;
			BUG_ON(!pmd_none(*pm)
			       && !(pte_present(*pt) && pte_huge(*pt)));
			return pt;
		}
	}
L
Linus Torvalds 已提交
56

57
	return NULL;
L
Linus Torvalds 已提交
58 59
}

60
pte_t *huge_pte_alloc(struct mm_struct *mm, unsigned long addr)
L
Linus Torvalds 已提交
61
{
62 63 64 65
	pgd_t *pg;
	pud_t *pu;
	pmd_t *pm;
	pte_t *pt;
L
Linus Torvalds 已提交
66 67 68

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

69
	addr &= HPAGE_MASK;
L
Linus Torvalds 已提交
70

71 72
	pg = pgd_offset(mm, addr);
	pu = pud_alloc(mm, pg, addr);
L
Linus Torvalds 已提交
73

74 75 76 77 78 79 80
	if (pu) {
		pm = pmd_alloc(mm, pu, addr);
		if (pm) {
			pt = (pte_t *)pm;
			BUG_ON(!pmd_none(*pm)
			       && !(pte_present(*pt) && pte_huge(*pt)));
			return pt;
L
Linus Torvalds 已提交
81 82 83
		}
	}

84
	return NULL;
L
Linus Torvalds 已提交
85 86
}

87
#define HUGEPTE_BATCH_SIZE	(HPAGE_SIZE / PMD_SIZE)
L
Linus Torvalds 已提交
88

89 90 91 92
void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
		     pte_t *ptep, pte_t pte)
{
	int i;
L
Linus Torvalds 已提交
93

94 95 96 97
	if (pte_present(*ptep)) {
		pte_clear(mm, addr, ptep);
		flush_tlb_pending();
	}
L
Linus Torvalds 已提交
98

99 100 101 102
	for (i = 0; i < HUGEPTE_BATCH_SIZE; i++) {
		*ptep = __pte(pte_val(pte) & ~_PAGE_HPTEFLAGS);
		ptep++;
	}
L
Linus Torvalds 已提交
103 104
}

105 106
pte_t huge_ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
			      pte_t *ptep)
L
Linus Torvalds 已提交
107
{
108 109
	unsigned long old = pte_update(ptep, ~0UL);
	int i;
L
Linus Torvalds 已提交
110

111 112
	if (old & _PAGE_HASHPTE)
		hpte_update(mm, addr, old, 0);
L
Linus Torvalds 已提交
113

114 115
	for (i = 1; i < HUGEPTE_BATCH_SIZE; i++)
		ptep[i] = __pte(0);
L
Linus Torvalds 已提交
116

117
	return __pte(old);
L
Linus Torvalds 已提交
118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134
}

/*
 * This function checks for proper alignment of input addr and len parameters.
 */
int is_aligned_hugepage_range(unsigned long addr, unsigned long len)
{
	if (len & ~HPAGE_MASK)
		return -EINVAL;
	if (addr & ~HPAGE_MASK)
		return -EINVAL;
	if (! (within_hugepage_low_range(addr, len)
	       || within_hugepage_high_range(addr, len)) )
		return -EINVAL;
	return 0;
}

135
static void flush_low_segments(void *parm)
L
Linus Torvalds 已提交
136
{
137
	u16 areas = (unsigned long) parm;
L
Linus Torvalds 已提交
138 139 140 141
	unsigned long i;

	asm volatile("isync" : : : "memory");

142 143 144 145
	BUILD_BUG_ON((sizeof(areas)*8) != NUM_LOW_AREAS);

	for (i = 0; i < NUM_LOW_AREAS; i++) {
		if (! (areas & (1U << i)))
L
Linus Torvalds 已提交
146
			continue;
147 148
		asm volatile("slbie %0"
			     : : "r" ((i << SID_SHIFT) | SLBIE_C));
L
Linus Torvalds 已提交
149 150 151 152 153
	}

	asm volatile("isync" : : : "memory");
}

154 155 156 157 158 159 160 161 162 163 164 165 166 167
static void flush_high_segments(void *parm)
{
	u16 areas = (unsigned long) parm;
	unsigned long i, j;

	asm volatile("isync" : : : "memory");

	BUILD_BUG_ON((sizeof(areas)*8) != NUM_HIGH_AREAS);

	for (i = 0; i < NUM_HIGH_AREAS; i++) {
		if (! (areas & (1U << i)))
			continue;
		for (j = 0; j < (1UL << (HTLB_AREA_SHIFT-SID_SHIFT)); j++)
			asm volatile("slbie %0"
168 169
				     :: "r" (((i << HTLB_AREA_SHIFT)
					     + (j << SID_SHIFT)) | SLBIE_C));
170 171 172 173 174 175
	}

	asm volatile("isync" : : : "memory");
}

static int prepare_low_area_for_htlb(struct mm_struct *mm, unsigned long area)
L
Linus Torvalds 已提交
176
{
177 178
	unsigned long start = area << SID_SHIFT;
	unsigned long end = (area+1) << SID_SHIFT;
L
Linus Torvalds 已提交
179 180
	struct vm_area_struct *vma;

181
	BUG_ON(area >= NUM_LOW_AREAS);
L
Linus Torvalds 已提交
182 183 184 185 186 187 188 189 190

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

	return 0;
}

191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207
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);

	/* 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
208 209 210
{
	unsigned long i;

211 212 213 214 215
	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
216 217
		return 0; /* The segments we want are already open */

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
	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;

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

	/* the context change must make it to memory before the flush,
	 * so that further SLB misses do the right thing. */
	mb();
	on_each_cpu(flush_low_segments, (void *)(unsigned long)newareas, 0, 1);

	return 0;
}

static int open_high_hpage_areas(struct mm_struct *mm, u16 newareas)
{
	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
251 252
				return -EBUSY;

253
	mm->context.high_htlb_areas |= newareas;
L
Linus Torvalds 已提交
254 255 256 257 258 259 260

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

	/* the context change must make it to memory before the flush,
	 * so that further SLB misses do the right thing. */
	mb();
261
	on_each_cpu(flush_high_segments, (void *)(unsigned long)newareas, 0, 1);
L
Linus Torvalds 已提交
262 263 264 265 266 267

	return 0;
}

int prepare_hugepage_range(unsigned long addr, unsigned long len)
{
268 269 270 271 272 273 274
	int err;

	if ( (addr+len) < addr )
		return -EINVAL;

	if ((addr + len) < 0x100000000UL)
		err = open_low_hpage_areas(current->mm,
L
Linus Torvalds 已提交
275
					  LOW_ESID_MASK(addr, len));
276 277 278 279 280 281 282 283
	else
		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
284 285 286
		return err;
	}

287
	return 0;
L
Linus Torvalds 已提交
288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341
}

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;
	}
342 343 344 345 346 347
	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
348 349 350 351 352 353 354 355 356 357 358

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;
		}
359 360
		if (touches_hugepage_high_range(mm, addr, len)) {
			addr = ALIGN(addr+1, 1UL<<HTLB_AREA_SHIFT);
L
Linus Torvalds 已提交
361 362 363 364 365 366 367 368 369 370
			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;
		}
371 372
		if (addr + mm->cached_hole_size < vma->vm_start)
		        mm->cached_hole_size = vma->vm_start - addr;
L
Linus Torvalds 已提交
373 374 375 376 377 378 379
		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;
380
		mm->cached_hole_size = 0;
L
Linus Torvalds 已提交
381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401
		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;
402
	unsigned long largest_hole = mm->cached_hole_size;
L
Linus Torvalds 已提交
403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422
	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;
	}

423 424 425 426
	if (len <= largest_hole) {
	        largest_hole = 0;
		mm->free_area_cache = base;
	}
L
Linus Torvalds 已提交
427 428 429 430 431 432 433 434 435 436 437 438
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;
439 440 441
		} 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
442 443 444 445 446 447 448 449 450 451 452 453 454 455
		}

		/*
		 * 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 &&
456
		          (!prev_vma || (addr >= prev_vma->vm_end))) {
L
Linus Torvalds 已提交
457
			/* remember the address as a hint for next time */
458 459 460
		        mm->cached_hole_size = largest_hole;
		        return (mm->free_area_cache = addr);
		} else {
L
Linus Torvalds 已提交
461
			/* pull free_area_cache down to the first hole */
462
		        if (mm->free_area_cache == vma->vm_end) {
L
Linus Torvalds 已提交
463
				mm->free_area_cache = vma->vm_start;
464 465 466 467 468 469 470
				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
471 472 473 474 475 476 477 478 479 480 481 482

		/* 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;
483
		largest_hole = 0;
L
Linus Torvalds 已提交
484 485 486 487 488 489 490 491 492 493
		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;
494
	mm->cached_hole_size = ~0UL;
L
Linus Torvalds 已提交
495 496 497 498 499
	addr = arch_get_unmapped_area(filp, addr0, len, pgoff, flags);
	/*
	 * Restore the topdown base:
	 */
	mm->free_area_cache = base;
500
	mm->cached_hole_size = ~0UL;
L
Linus Torvalds 已提交
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

	return addr;
}

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;
}

532
static unsigned long htlb_get_high_area(unsigned long len, u16 areamask)
L
Linus Torvalds 已提交
533
{
534
	unsigned long addr = 0x100000000UL;
L
Linus Torvalds 已提交
535 536 537
	struct vm_area_struct *vma;

	vma = find_vma(current->mm, addr);
538
	while (addr + len <= TASK_SIZE_USER64) {
L
Linus Torvalds 已提交
539
		BUG_ON(vma && (addr >= vma->vm_end)); /* invariant */
540 541 542 543 544 545

		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
546 547 548 549

		if (!vma || (addr + len) <= vma->vm_start)
			return addr;
		addr = ALIGN(vma->vm_end, HPAGE_SIZE);
550 551 552 553
		/* 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
554 555 556 557 558 559 560 561 562
	}

	return -ENOMEM;
}

unsigned long hugetlb_get_unmapped_area(struct file *file, unsigned long addr,
					unsigned long len, unsigned long pgoff,
					unsigned long flags)
{
563 564 565
	int lastshift;
	u16 areamask, curareas;

L
Linus Torvalds 已提交
566 567 568 569 570 571 572
	if (len & ~HPAGE_MASK)
		return -EINVAL;

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

	if (test_thread_flag(TIF_32BIT)) {
573
		curareas = current->mm->context.low_htlb_areas;
L
Linus Torvalds 已提交
574 575

		/* First see if we can do the mapping in the existing
576 577
		 * low areas */
		addr = htlb_get_low_area(len, curareas);
L
Linus Torvalds 已提交
578 579 580
		if (addr != -ENOMEM)
			return addr;

581 582 583 584
		lastshift = 0;
		for (areamask = LOW_ESID_MASK(0x100000000UL-len, len);
		     ! lastshift; areamask >>=1) {
			if (areamask & 1)
L
Linus Torvalds 已提交
585 586
				lastshift = 1;

587
			addr = htlb_get_low_area(len, curareas | areamask);
L
Linus Torvalds 已提交
588
			if ((addr != -ENOMEM)
589
			    && open_low_hpage_areas(current->mm, areamask) == 0)
L
Linus Torvalds 已提交
590 591 592
				return addr;
		}
	} else {
593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611
		curareas = current->mm->context.high_htlb_areas;

		/* First see if we can do the mapping in the existing
		 * high areas */
		addr = htlb_get_high_area(len, curareas);
		if (addr != -ENOMEM)
			return addr;

		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
612
	}
613 614 615
	printk(KERN_DEBUG "hugetlb_get_unmapped_area() unable to open"
	       " enough areas\n");
	return -ENOMEM;
L
Linus Torvalds 已提交
616 617 618 619 620 621 622 623
}

int hash_huge_page(struct mm_struct *mm, unsigned long access,
		   unsigned long ea, unsigned long vsid, int local)
{
	pte_t *ptep;
	unsigned long va, vpn;
	pte_t old_pte, new_pte;
624
	unsigned long rflags, prpn;
L
Linus Torvalds 已提交
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
	long slot;
	int err = 1;

	spin_lock(&mm->page_table_lock);

	ptep = huge_pte_offset(mm, ea);

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

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

/* 	BUG_ON(pte_bad(*ptep)); */

	/* 
	 * 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. 
	 */


	old_pte = *ptep;
	new_pte = old_pte;

667
	rflags = 0x2 | (! (pte_val(new_pte) & _PAGE_RW));
L
Linus Torvalds 已提交
668
 	/* _PAGE_EXEC -> HW_NO_EXEC since it's inverted */
669
	rflags |= ((pte_val(new_pte) & _PAGE_EXEC) ? 0 : HW_NO_EXEC);
L
Linus Torvalds 已提交
670 671 672 673 674 675 676 677 678 679 680 681

	/* Check if pte already has an hpte (case 2) */
	if (unlikely(pte_val(old_pte) & _PAGE_HASHPTE)) {
		/* There MIGHT be an HPTE for this pte */
		unsigned long hash, slot;

		hash = hpt_hash(vpn, 1);
		if (pte_val(old_pte) & _PAGE_SECONDARY)
			hash = ~hash;
		slot = (hash & htab_hash_mask) * HPTES_PER_GROUP;
		slot += (pte_val(old_pte) & _PAGE_GROUP_IX) >> 12;

682
		if (ppc_md.hpte_updatepp(slot, rflags, va, 1, local) == -1)
L
Linus Torvalds 已提交
683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701
			pte_val(old_pte) &= ~_PAGE_HPTEFLAGS;
	}

	if (likely(!(pte_val(old_pte) & _PAGE_HASHPTE))) {
		unsigned long hash = hpt_hash(vpn, 1);
		unsigned long hpte_group;

		prpn = pte_pfn(old_pte);

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

		/* Update the linux pte with the HPTE slot */
		pte_val(new_pte) &= ~_PAGE_HPTEFLAGS;
		pte_val(new_pte) |= _PAGE_HASHPTE;

		/* Add in WIMG bits */
		/* XXX We should store these in the pte */
702
		rflags |= _PAGE_COHERENT;
L
Linus Torvalds 已提交
703

704 705
		slot = ppc_md.hpte_insert(hpte_group, va, prpn,
					  HPTE_V_LARGE, rflags);
L
Linus Torvalds 已提交
706 707 708 709 710 711 712

		/* Primary is full, try the secondary */
		if (unlikely(slot == -1)) {
			pte_val(new_pte) |= _PAGE_SECONDARY;
			hpte_group = ((~hash & htab_hash_mask) *
				      HPTES_PER_GROUP) & ~0x7UL; 
			slot = ppc_md.hpte_insert(hpte_group, va, prpn,
713 714 715
						  HPTE_V_LARGE |
						  HPTE_V_SECONDARY,
						  rflags);
L
Linus Torvalds 已提交
716 717
			if (slot == -1) {
				if (mftb() & 0x1)
718 719
					hpte_group = ((hash & htab_hash_mask) *
						      HPTES_PER_GROUP)&~0x7UL;
L
Linus Torvalds 已提交
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

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

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

		pte_val(new_pte) |= (slot<<12) & _PAGE_GROUP_IX;

		/* 
		 * No need to use ldarx/stdcx here because all who
		 * might be updating the pte will hold the
		 * page_table_lock
		 */
		*ptep = new_pte;
	}

	err = 0;

 out:
	spin_unlock(&mm->page_table_lock);

	return err;
}