memory.c 36.4 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47
/*
 *  linux/mm/memory.c
 *
 *  Copyright (C) 1991, 1992, 1993, 1994  Linus Torvalds
 */

/*
 * demand-loading started 01.12.91 - seems it is high on the list of
 * things wanted, and it should be easy to implement. - Linus
 */

/*
 * Ok, demand-loading was easy, shared pages a little bit tricker. Shared
 * pages started 02.12.91, seems to work. - Linus.
 *
 * Tested sharing by executing about 30 /bin/sh: under the old kernel it
 * would have taken more than the 6M I have free, but it worked well as
 * far as I could see.
 *
 * Also corrected some "invalidate()"s - I wasn't doing enough of them.
 */

/*
 * Real VM (paging to/from disk) started 18.12.91. Much more work and
 * thought has to go into this. Oh, well..
 * 19.12.91  -  works, somewhat. Sometimes I get faults, don't know why.
 *		Found it. Everything seems to work now.
 * 20.12.91  -  Ok, making the swap-device changeable like the root.
 */

/*
 * 05.04.94  -  Multi-page memory management added for v1.1.
 * 		Idea by Alex Bligh (alex@cconcepts.co.uk)
 *
 * 16.07.99  -  Support of BIGMEM added by Gerhard Wichert, Siemens AG
 *		(Gerhard.Wichert@pdb.siemens.de)
 */

#include <linux/mm.h>
#include <linux/mman.h>
#include <linux/swap.h>
#include <linux/smp_lock.h>
#include <linux/swapctl.h>
#include <linux/iobuf.h>
#include <linux/highmem.h>
#include <linux/pagemap.h>

Linus Torvalds's avatar
Linus Torvalds committed
48 49
#include <asm/pgalloc.h>
#include <asm/uaccess.h>
Linus Torvalds's avatar
Linus Torvalds committed
50
#include <asm/tlb.h>
51
#include <asm/tlbflush.h>
Linus Torvalds's avatar
Linus Torvalds committed
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

unsigned long max_mapnr;
unsigned long num_physpages;
void * high_memory;
struct page *highmem_start_page;

/*
 * We special-case the C-O-W ZERO_PAGE, because it's such
 * a common occurrence (no need to read the page to know
 * that it's zero - better for the cache and memory subsystem).
 */
static inline void copy_cow_page(struct page * from, struct page * to, unsigned long address)
{
	if (from == ZERO_PAGE(address)) {
		clear_user_highpage(to, address);
		return;
	}
	copy_user_highpage(to, from, address);
}

mem_map_t * mem_map;

/*
 * Note: this doesn't free the actual pages themselves. That
 * has been handled earlier when unmapping all the memory regions.
 */
78
static inline void free_one_pmd(mmu_gather_t *tlb, pmd_t * dir)
Linus Torvalds's avatar
Linus Torvalds committed
79
{
80
	struct page *pte;
Linus Torvalds's avatar
Linus Torvalds committed
81 82 83 84 85 86 87 88

	if (pmd_none(*dir))
		return;
	if (pmd_bad(*dir)) {
		pmd_ERROR(*dir);
		pmd_clear(dir);
		return;
	}
89
	pte = pmd_page(*dir);
Linus Torvalds's avatar
Linus Torvalds committed
90
	pmd_clear(dir);
91
	pte_free_tlb(tlb, pte);
Linus Torvalds's avatar
Linus Torvalds committed
92 93
}

94
static inline void free_one_pgd(mmu_gather_t *tlb, pgd_t * dir)
Linus Torvalds's avatar
Linus Torvalds committed
95 96 97 98 99 100 101 102 103 104 105 106 107
{
	int j;
	pmd_t * pmd;

	if (pgd_none(*dir))
		return;
	if (pgd_bad(*dir)) {
		pgd_ERROR(*dir);
		pgd_clear(dir);
		return;
	}
	pmd = pmd_offset(dir, 0);
	pgd_clear(dir);
Linus Torvalds's avatar
Linus Torvalds committed
108 109
	for (j = 0; j < PTRS_PER_PMD ; j++) {
		prefetchw(pmd+j+(PREFETCH_STRIDE/16));
110
		free_one_pmd(tlb, pmd+j);
Linus Torvalds's avatar
Linus Torvalds committed
111
	}
112
	pmd_free_tlb(tlb, pmd);
Linus Torvalds's avatar
Linus Torvalds committed
113 114 115 116 117
}

/*
 * This function clears all user-level page tables of a process - this
 * is needed by execve(), so that old pages aren't in the way.
118 119
 *
 * Must be called with pagetable lock held.
Linus Torvalds's avatar
Linus Torvalds committed
120
 */
121
void clear_page_tables(mmu_gather_t *tlb, unsigned long first, int nr)
Linus Torvalds's avatar
Linus Torvalds committed
122
{
123
	pgd_t * page_dir = tlb->mm->pgd;
Linus Torvalds's avatar
Linus Torvalds committed
124 125 126

	page_dir += first;
	do {
127
		free_one_pgd(tlb, page_dir);
Linus Torvalds's avatar
Linus Torvalds committed
128 129
		page_dir++;
	} while (--nr);
130 131 132 133 134 135
}

pte_t * pte_alloc_map(struct mm_struct *mm, pmd_t *pmd, unsigned long address)
{
	if (!pmd_present(*pmd)) {
		struct page *new;
Linus Torvalds's avatar
Linus Torvalds committed
136

137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156
		spin_unlock(&mm->page_table_lock);
		new = pte_alloc_one(mm, address);
		spin_lock(&mm->page_table_lock);
		if (!new)
			return NULL;

		/*
		 * Because we dropped the lock, we should re-check the
		 * entry, as somebody else could have populated it..
		 */
		if (pmd_present(*pmd)) {
			pte_free(new);
			goto out;
		}
		pmd_populate(mm, pmd, new);
	}
out:
	if (pmd_present(*pmd))
		return pte_offset_map(pmd, address);
	return NULL;
Linus Torvalds's avatar
Linus Torvalds committed
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
pte_t * pte_alloc_kernel(struct mm_struct *mm, pmd_t *pmd, unsigned long address)
{
	if (!pmd_present(*pmd)) {
		pte_t *new;

		spin_unlock(&mm->page_table_lock);
		new = pte_alloc_one_kernel(mm, address);
		spin_lock(&mm->page_table_lock);
		if (!new)
			return NULL;

		/*
		 * Because we dropped the lock, we should re-check the
		 * entry, as somebody else could have populated it..
		 */
		if (pmd_present(*pmd)) {
			pte_free_kernel(new);
			goto out;
		}
		pmd_populate_kernel(mm, pmd, new);
	}
out:
	return pte_offset_kernel(pmd, address);
}
Linus Torvalds's avatar
Linus Torvalds committed
183 184 185 186 187 188 189 190 191 192
#define PTE_TABLE_MASK	((PTRS_PER_PTE-1) * sizeof(pte_t))
#define PMD_TABLE_MASK	((PTRS_PER_PMD-1) * sizeof(pmd_t))

/*
 * copy one vm_area from one task to the other. Assumes the page tables
 * already present in the new task to be cleared in the whole range
 * covered by this vma.
 *
 * 08Jan98 Merged into one routine from several inline routines to reduce
 *         variable count and make things faster. -jj
Linus Torvalds's avatar
Linus Torvalds committed
193 194
 *
 * dst->page_table_lock is held on entry and exit,
195
 * but may be dropped within pmd_alloc() and pte_alloc_map().
Linus Torvalds's avatar
Linus Torvalds committed
196 197 198 199 200 201 202
 */
int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
			struct vm_area_struct *vma)
{
	pgd_t * src_pgd, * dst_pgd;
	unsigned long address = vma->vm_start;
	unsigned long end = vma->vm_end;
203
	unsigned long cow = (vma->vm_flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
Linus Torvalds's avatar
Linus Torvalds committed
204 205 206

	src_pgd = pgd_offset(src, address)-1;
	dst_pgd = pgd_offset(dst, address)-1;
Linus Torvalds's avatar
Linus Torvalds committed
207

Linus Torvalds's avatar
Linus Torvalds committed
208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224
	for (;;) {
		pmd_t * src_pmd, * dst_pmd;

		src_pgd++; dst_pgd++;
		
		/* copy_pmd_range */
		
		if (pgd_none(*src_pgd))
			goto skip_copy_pmd_range;
		if (pgd_bad(*src_pgd)) {
			pgd_ERROR(*src_pgd);
			pgd_clear(src_pgd);
skip_copy_pmd_range:	address = (address + PGDIR_SIZE) & PGDIR_MASK;
			if (!address || (address >= end))
				goto out;
			continue;
		}
Linus Torvalds's avatar
Linus Torvalds committed
225

Linus Torvalds's avatar
Linus Torvalds committed
226
		src_pmd = pmd_offset(src_pgd, address);
Linus Torvalds's avatar
Linus Torvalds committed
227 228 229
		dst_pmd = pmd_alloc(dst, dst_pgd, address);
		if (!dst_pmd)
			goto nomem;
Linus Torvalds's avatar
Linus Torvalds committed
230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245

		do {
			pte_t * src_pte, * dst_pte;
		
			/* copy_pte_range */
		
			if (pmd_none(*src_pmd))
				goto skip_copy_pte_range;
			if (pmd_bad(*src_pmd)) {
				pmd_ERROR(*src_pmd);
				pmd_clear(src_pmd);
skip_copy_pte_range:		address = (address + PMD_SIZE) & PMD_MASK;
				if (address >= end)
					goto out;
				goto cont_copy_pmd_range;
			}
Linus Torvalds's avatar
Linus Torvalds committed
246

247
			dst_pte = pte_alloc_map(dst, dst_pmd, address);
Linus Torvalds's avatar
Linus Torvalds committed
248 249
			if (!dst_pte)
				goto nomem;
Linus Torvalds's avatar
Linus Torvalds committed
250
			spin_lock(&src->page_table_lock);			
251
			src_pte = pte_offset_map_nested(src_pmd, address);
Linus Torvalds's avatar
Linus Torvalds committed
252 253 254
			do {
				pte_t pte = *src_pte;
				struct page *ptepage;
255
				unsigned long pfn;
Linus Torvalds's avatar
Linus Torvalds committed
256 257 258 259 260 261 262 263 264
				
				/* copy_one_pte */

				if (pte_none(pte))
					goto cont_copy_pte_range_noset;
				if (!pte_present(pte)) {
					swap_duplicate(pte_to_swp_entry(pte));
					goto cont_copy_pte_range;
				}
265 266 267 268 269
				pfn = pte_pfn(pte);
				if (!pfn_valid(pfn))
					goto cont_copy_pte_range;
				ptepage = pfn_to_page(pfn);
				if (PageReserved(ptepage))
Linus Torvalds's avatar
Linus Torvalds committed
270 271 272
					goto cont_copy_pte_range;

				/* If it's a COW mapping, write protect it both in the parent and the child */
273
				if (cow && pte_write(pte)) {
Linus Torvalds's avatar
Linus Torvalds committed
274 275 276 277 278 279 280 281 282
					ptep_set_wrprotect(src_pte);
					pte = *src_pte;
				}

				/* If it's a shared mapping, mark it clean in the child */
				if (vma->vm_flags & VM_SHARED)
					pte = pte_mkclean(pte);
				pte = pte_mkold(pte);
				get_page(ptepage);
Linus Torvalds's avatar
Linus Torvalds committed
283
				dst->rss++;
Linus Torvalds's avatar
Linus Torvalds committed
284 285 286

cont_copy_pte_range:		set_pte(dst_pte, pte);
cont_copy_pte_range_noset:	address += PAGE_SIZE;
287
				if (address >= end) {
288
					pte_unmap_nested(src_pte);
289
					pte_unmap(dst_pte);
Linus Torvalds's avatar
Linus Torvalds committed
290
					goto out_unlock;
291
				}
Linus Torvalds's avatar
Linus Torvalds committed
292 293 294
				src_pte++;
				dst_pte++;
			} while ((unsigned long)src_pte & PTE_TABLE_MASK);
295
			pte_unmap_nested(src_pte-1);
296
			pte_unmap(dst_pte-1);
Linus Torvalds's avatar
Linus Torvalds committed
297
			spin_unlock(&src->page_table_lock);
Linus Torvalds's avatar
Linus Torvalds committed
298 299 300 301 302
		
cont_copy_pmd_range:	src_pmd++;
			dst_pmd++;
		} while ((unsigned long)src_pmd & PMD_TABLE_MASK);
	}
Linus Torvalds's avatar
Linus Torvalds committed
303 304
out_unlock:
	spin_unlock(&src->page_table_lock);
Linus Torvalds's avatar
Linus Torvalds committed
305
out:
Linus Torvalds's avatar
Linus Torvalds committed
306
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
307 308 309 310 311 312 313 314 315 316 317
nomem:
	return -ENOMEM;
}

/*
 * Return indicates whether a page was freed so caller can adjust rss
 */
static inline void forget_pte(pte_t page)
{
	if (!pte_none(page)) {
		printk("forget_pte: old mapping existed!\n");
Linus Torvalds's avatar
Linus Torvalds committed
318
		BUG();
Linus Torvalds's avatar
Linus Torvalds committed
319 320 321
	}
}

322
static void zap_pte_range(mmu_gather_t *tlb, pmd_t * pmd, unsigned long address, unsigned long size)
Linus Torvalds's avatar
Linus Torvalds committed
323
{
Linus Torvalds's avatar
Linus Torvalds committed
324
	unsigned long offset;
325
	pte_t *ptep;
Linus Torvalds's avatar
Linus Torvalds committed
326 327

	if (pmd_none(*pmd))
328
		return;
Linus Torvalds's avatar
Linus Torvalds committed
329 330 331
	if (pmd_bad(*pmd)) {
		pmd_ERROR(*pmd);
		pmd_clear(pmd);
332
		return;
Linus Torvalds's avatar
Linus Torvalds committed
333
	}
334
	ptep = pte_offset_map(pmd, address);
Linus Torvalds's avatar
Linus Torvalds committed
335 336 337 338 339 340 341
	offset = address & ~PMD_MASK;
	if (offset + size > PMD_SIZE)
		size = PMD_SIZE - offset;
	size &= PAGE_MASK;
	for (offset=0; offset < size; ptep++, offset += PAGE_SIZE) {
		pte_t pte = *ptep;
		if (pte_none(pte))
Linus Torvalds's avatar
Linus Torvalds committed
342
			continue;
Linus Torvalds's avatar
Linus Torvalds committed
343
		if (pte_present(pte)) {
344 345
			unsigned long pfn = pte_pfn(pte);

346
			pte = ptep_get_and_clear(ptep);
347
			tlb_remove_tlb_entry(tlb, pte, address+offset);
348 349 350 351 352
			if (pfn_valid(pfn)) {
				struct page *page = pfn_to_page(pfn);
				if (!PageReserved(page)) {
					if (pte_dirty(pte))
						set_page_dirty(page);
353
					tlb->freed++;
354 355 356
					tlb_remove_page(tlb, page);
				}
			}
Linus Torvalds's avatar
Linus Torvalds committed
357
		} else {
Linus Torvalds's avatar
Linus Torvalds committed
358
			free_swap_and_cache(pte_to_swp_entry(pte));
Linus Torvalds's avatar
Linus Torvalds committed
359 360
			pte_clear(ptep);
		}
Linus Torvalds's avatar
Linus Torvalds committed
361
	}
362
	pte_unmap(ptep-1);
Linus Torvalds's avatar
Linus Torvalds committed
363 364
}

365
static void zap_pmd_range(mmu_gather_t *tlb, pgd_t * dir, unsigned long address, unsigned long size)
Linus Torvalds's avatar
Linus Torvalds committed
366 367 368 369 370
{
	pmd_t * pmd;
	unsigned long end;

	if (pgd_none(*dir))
371
		return;
Linus Torvalds's avatar
Linus Torvalds committed
372 373 374
	if (pgd_bad(*dir)) {
		pgd_ERROR(*dir);
		pgd_clear(dir);
375
		return;
Linus Torvalds's avatar
Linus Torvalds committed
376 377 378
	}
	pmd = pmd_offset(dir, address);
	end = address + size;
Linus Torvalds's avatar
Linus Torvalds committed
379 380
	if (end > ((address + PGDIR_SIZE) & PGDIR_MASK))
		end = ((address + PGDIR_SIZE) & PGDIR_MASK);
Linus Torvalds's avatar
Linus Torvalds committed
381
	do {
382
		zap_pte_range(tlb, pmd, address, end - address);
Linus Torvalds's avatar
Linus Torvalds committed
383 384 385
		address = (address + PMD_SIZE) & PMD_MASK; 
		pmd++;
	} while (address < end);
386 387 388 389 390 391 392 393 394
}

void unmap_page_range(mmu_gather_t *tlb, struct vm_area_struct *vma, unsigned long address, unsigned long end)
{
	pgd_t * dir;

	if (address >= end)
		BUG();
	dir = pgd_offset(vma->vm_mm, address);
Linus Torvalds's avatar
Linus Torvalds committed
395
	tlb_start_vma(tlb, vma);
396 397 398 399 400
	do {
		zap_pmd_range(tlb, dir, address, end - address);
		address = (address + PGDIR_SIZE) & PGDIR_MASK;
		dir++;
	} while (address && (address < end));
Linus Torvalds's avatar
Linus Torvalds committed
401
	tlb_end_vma(tlb, vma);
Linus Torvalds's avatar
Linus Torvalds committed
402 403 404 405 406
}

/*
 * remove user pages in a given range.
 */
Linus Torvalds's avatar
Linus Torvalds committed
407
void zap_page_range(struct vm_area_struct *vma, unsigned long address, unsigned long size)
Linus Torvalds's avatar
Linus Torvalds committed
408
{
Linus Torvalds's avatar
Linus Torvalds committed
409
	struct mm_struct *mm = vma->vm_mm;
Linus Torvalds's avatar
Linus Torvalds committed
410
	mmu_gather_t *tlb;
Linus Torvalds's avatar
Linus Torvalds committed
411
	pgd_t * dir;
Linus Torvalds's avatar
Linus Torvalds committed
412
	unsigned long start = address, end = address + size;
Linus Torvalds's avatar
Linus Torvalds committed
413 414 415 416 417 418 419 420 421 422 423 424 425

	dir = pgd_offset(mm, address);

	/*
	 * This is a long-lived spinlock. That's fine.
	 * There's no contention, because the page table
	 * lock only protects against kswapd anyway, and
	 * even if kswapd happened to be looking at this
	 * process we _want_ it to get stuck.
	 */
	if (address >= end)
		BUG();
	spin_lock(&mm->page_table_lock);
Linus Torvalds's avatar
Linus Torvalds committed
426
	flush_cache_range(vma, address, end);
Linus Torvalds's avatar
Linus Torvalds committed
427

428 429
	tlb = tlb_gather_mmu(mm);
	unmap_page_range(tlb, vma, address, end);
Linus Torvalds's avatar
Linus Torvalds committed
430
	tlb_finish_mmu(tlb, start, end);
Linus Torvalds's avatar
Linus Torvalds committed
431
	spin_unlock(&mm->page_table_lock);
Linus Torvalds's avatar
Linus Torvalds committed
432 433 434 435 436
}

/*
 * Do a quick page-table lookup for a single page. 
 */
Linus Torvalds's avatar
Linus Torvalds committed
437
static struct page * follow_page(struct mm_struct *mm, unsigned long address, int write) 
Linus Torvalds's avatar
Linus Torvalds committed
438 439 440
{
	pgd_t *pgd;
	pmd_t *pmd;
Linus Torvalds's avatar
Linus Torvalds committed
441
	pte_t *ptep, pte;
442
	unsigned long pfn;
Linus Torvalds's avatar
Linus Torvalds committed
443

Linus Torvalds's avatar
Linus Torvalds committed
444
	pgd = pgd_offset(mm, address);
Linus Torvalds's avatar
Linus Torvalds committed
445 446 447
	if (pgd_none(*pgd) || pgd_bad(*pgd))
		goto out;

Linus Torvalds's avatar
Linus Torvalds committed
448
	pmd = pmd_offset(pgd, address);
Linus Torvalds's avatar
Linus Torvalds committed
449 450 451
	if (pmd_none(*pmd) || pmd_bad(*pmd))
		goto out;

452 453 454 455
	preempt_disable();
	ptep = pte_offset_map(pmd, address);
	if (!ptep) {
		preempt_enable();
Linus Torvalds's avatar
Linus Torvalds committed
456
		goto out;
457
	}
Linus Torvalds's avatar
Linus Torvalds committed
458 459

	pte = *ptep;
460 461
	pte_unmap(ptep);
	preempt_enable();
Linus Torvalds's avatar
Linus Torvalds committed
462 463
	if (pte_present(pte)) {
		if (!write ||
464 465 466 467 468
		    (pte_write(pte) && pte_dirty(pte))) {
			pfn = pte_pfn(pte);
			if (pfn_valid(pfn))
				return pfn_to_page(pfn);
		}
Linus Torvalds's avatar
Linus Torvalds committed
469
	}
Linus Torvalds's avatar
Linus Torvalds committed
470 471 472

out:
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
473 474 475 476 477 478 479 480 481 482 483 484 485
}

/* 
 * Given a physical address, is there a useful struct page pointing to
 * it?  This may become more complex in the future if we start dealing
 * with IO-aperture pages in kiobufs.
 */

static inline struct page * get_page_map(struct page *page)
{
	return page;
}

Linus Torvalds's avatar
Linus Torvalds committed
486 487 488
int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
		int len, int write, int force, struct page **pages, struct vm_area_struct **vmas)
{
489 490 491 492 493 494 495 496 497 498 499
	int i;
	unsigned int flags;

	/* 
	 * Require read or write permissions.
	 * If 'force' is set, we only require the "MAY" flags.
	 */
	flags = write ? (VM_WRITE | VM_MAYWRITE) : (VM_READ | VM_MAYREAD);
	flags &= force ? (VM_MAYREAD | VM_MAYWRITE) : (VM_READ | VM_WRITE);
	i = 0;
	
Linus Torvalds's avatar
Linus Torvalds committed
500 501 502 503 504 505

	do {
		struct vm_area_struct *	vma;

		vma = find_extend_vma(mm, start);

506 507
		if ( !vma || !(flags & vma->vm_flags) )
			return i ? : -EFAULT;
Linus Torvalds's avatar
Linus Torvalds committed
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

		spin_lock(&mm->page_table_lock);
		do {
			struct page *map;
			while (!(map = follow_page(mm, start, write))) {
				spin_unlock(&mm->page_table_lock);
				switch (handle_mm_fault(mm, vma, start, write)) {
				case 1:
					tsk->min_flt++;
					break;
				case 2:
					tsk->maj_flt++;
					break;
				case 0:
					if (i) return i;
					return -EFAULT;
				default:
					if (i) return i;
					return -ENOMEM;
				}
				spin_lock(&mm->page_table_lock);
			}
			if (pages) {
				pages[i] = get_page_map(map);
				/* FIXME: call the correct function,
				 * depending on the type of the found page
				 */
				if (pages[i])
					page_cache_get(pages[i]);
			}
			if (vmas)
				vmas[i] = vma;
			i++;
			start += PAGE_SIZE;
			len--;
		} while(len && start < vma->vm_end);
		spin_unlock(&mm->page_table_lock);
	} while(len);
	return i;
}

Linus Torvalds's avatar
Linus Torvalds committed
549 550 551 552 553
/*
 * Force in an entire range of pages from the current process's user VA,
 * and pin them in physical memory.  
 */
#define dprintk(x...)
Linus Torvalds's avatar
Linus Torvalds committed
554

Linus Torvalds's avatar
Linus Torvalds committed
555 556
int map_user_kiobuf(int rw, struct kiobuf *iobuf, unsigned long va, size_t len)
{
Linus Torvalds's avatar
Linus Torvalds committed
557
	int pgcount, err;
Linus Torvalds's avatar
Linus Torvalds committed
558 559 560 561 562 563 564 565 566
	struct mm_struct *	mm;
	
	/* Make sure the iobuf is not already mapped somewhere. */
	if (iobuf->nr_pages)
		return -EINVAL;

	mm = current->mm;
	dprintk ("map_user_kiobuf: begin\n");
	
Linus Torvalds's avatar
Linus Torvalds committed
567 568 569 570
	pgcount = (va + len + PAGE_SIZE - 1)/PAGE_SIZE - va/PAGE_SIZE;
	/* mapping 0 bytes is not permitted */
	if (!pgcount) BUG();
	err = expand_kiobuf(iobuf, pgcount);
Linus Torvalds's avatar
Linus Torvalds committed
571 572 573 574
	if (err)
		return err;

	iobuf->locked = 0;
Linus Torvalds's avatar
Linus Torvalds committed
575
	iobuf->offset = va & (PAGE_SIZE-1);
Linus Torvalds's avatar
Linus Torvalds committed
576 577
	iobuf->length = len;
	
Linus Torvalds's avatar
Linus Torvalds committed
578 579 580 581 582
	/* Try to fault in all of the necessary pages */
	down_read(&mm->mmap_sem);
	/* rw==READ means read from disk, write into memory area */
	err = get_user_pages(current, mm, va, pgcount,
			(rw==READ), 0, iobuf->maplist, NULL);
Linus Torvalds's avatar
Linus Torvalds committed
583
	up_read(&mm->mmap_sem);
Linus Torvalds's avatar
Linus Torvalds committed
584 585 586 587 588 589 590 591 592 593 594 595
	if (err < 0) {
		unmap_kiobuf(iobuf);
		dprintk ("map_user_kiobuf: end %d\n", err);
		return err;
	}
	iobuf->nr_pages = err;
	while (pgcount--) {
		/* FIXME: flush superflous for rw==READ,
		 * probably wrong function for rw==WRITE
		 */
		flush_dcache_page(iobuf->maplist[pgcount]);
	}
Linus Torvalds's avatar
Linus Torvalds committed
596 597 598 599
	dprintk ("map_user_kiobuf: end OK\n");
	return 0;
}

Linus Torvalds's avatar
Linus Torvalds committed
600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630
/*
 * Mark all of the pages in a kiobuf as dirty 
 *
 * We need to be able to deal with short reads from disk: if an IO error
 * occurs, the number of bytes read into memory may be less than the
 * size of the kiobuf, so we have to stop marking pages dirty once the
 * requested byte count has been reached.
 */

void mark_dirty_kiobuf(struct kiobuf *iobuf, int bytes)
{
	int index, offset, remaining;
	struct page *page;
	
	index = iobuf->offset >> PAGE_SHIFT;
	offset = iobuf->offset & ~PAGE_MASK;
	remaining = bytes;
	if (remaining > iobuf->length)
		remaining = iobuf->length;
	
	while (remaining > 0 && index < iobuf->nr_pages) {
		page = iobuf->maplist[index];
		
		if (!PageReserved(page))
			SetPageDirty(page);

		remaining -= (PAGE_SIZE - offset);
		offset = 0;
		index++;
	}
}
Linus Torvalds's avatar
Linus Torvalds committed
631 632 633 634 635 636 637 638 639 640 641 642 643 644 645

/*
 * Unmap all of the pages referenced by a kiobuf.  We release the pages,
 * and unlock them if they were locked. 
 */

void unmap_kiobuf (struct kiobuf *iobuf) 
{
	int i;
	struct page *map;
	
	for (i = 0; i < iobuf->nr_pages; i++) {
		map = iobuf->maplist[i];
		if (map) {
			if (iobuf->locked)
Andrew Morton's avatar
Andrew Morton committed
646
				unlock_page(map);
Linus Torvalds's avatar
Linus Torvalds committed
647 648 649
			/* FIXME: cache flush missing for rw==READ
			 * FIXME: call the correct reference counting function
			 */
Linus Torvalds's avatar
Linus Torvalds committed
650
			page_cache_release(map);
Linus Torvalds's avatar
Linus Torvalds committed
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
		}
	}
	
	iobuf->nr_pages = 0;
	iobuf->locked = 0;
}


/*
 * Lock down all of the pages of a kiovec for IO.
 *
 * If any page is mapped twice in the kiovec, we return the error -EINVAL.
 *
 * The optional wait parameter causes the lock call to block until all
 * pages can be locked if set.  If wait==0, the lock operation is
 * aborted if any locked pages are found and -EAGAIN is returned.
 */

int lock_kiovec(int nr, struct kiobuf *iovec[], int wait)
{
	struct kiobuf *iobuf;
	int i, j;
	struct page *page, **ppage;
	int doublepage = 0;
	int repeat = 0;
	
 repeat:
	
	for (i = 0; i < nr; i++) {
		iobuf = iovec[i];

		if (iobuf->locked)
			continue;

		ppage = iobuf->maplist;
		for (j = 0; j < iobuf->nr_pages; ppage++, j++) {
			page = *ppage;
			if (!page)
				continue;
			
Andrew Morton's avatar
Andrew Morton committed
691
			if (TestSetPageLocked(page)) {
Linus Torvalds's avatar
Linus Torvalds committed
692
				while (j--) {
Linus Torvalds's avatar
Linus Torvalds committed
693 694
					struct page *tmp = *--ppage;
					if (tmp)
Andrew Morton's avatar
Andrew Morton committed
695
						unlock_page(tmp);
Linus Torvalds's avatar
Linus Torvalds committed
696
				}
Linus Torvalds's avatar
Linus Torvalds committed
697
				goto retry;
Linus Torvalds's avatar
Linus Torvalds committed
698
			}
Linus Torvalds's avatar
Linus Torvalds committed
699
		}
Linus Torvalds's avatar
Linus Torvalds committed
700
		iobuf->locked = 1;
Linus Torvalds's avatar
Linus Torvalds committed
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
	}

	return 0;
	
 retry:
	
	/* 
	 * We couldn't lock one of the pages.  Undo the locking so far,
	 * wait on the page we got to, and try again.  
	 */
	
	unlock_kiovec(nr, iovec);
	if (!wait)
		return -EAGAIN;
	
	/* 
	 * Did the release also unlock the page we got stuck on?
	 */
	if (!PageLocked(page)) {
		/* 
		 * If so, we may well have the page mapped twice
		 * in the IO address range.  Bad news.  Of
		 * course, it _might_ just be a coincidence,
		 * but if it happens more than once, chances
		 * are we have a double-mapped page. 
		 */
		if (++doublepage >= 3) 
			return -EINVAL;
		
		/* Try again...  */
731
		wait_on_page_locked(page);
Linus Torvalds's avatar
Linus Torvalds committed
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
	}
	
	if (++repeat < 16)
		goto repeat;
	return -EAGAIN;
}

/*
 * Unlock all of the pages of a kiovec after IO.
 */

int unlock_kiovec(int nr, struct kiobuf *iovec[])
{
	struct kiobuf *iobuf;
	int i, j;
	struct page *page, **ppage;
	
	for (i = 0; i < nr; i++) {
		iobuf = iovec[i];

		if (!iobuf->locked)
			continue;
		iobuf->locked = 0;
		
		ppage = iobuf->maplist;
		for (j = 0; j < iobuf->nr_pages; ppage++, j++) {
			page = *ppage;
			if (!page)
				continue;
Andrew Morton's avatar
Andrew Morton committed
761
			unlock_page(page);
Linus Torvalds's avatar
Linus Torvalds committed
762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785
		}
	}
	return 0;
}

static inline void zeromap_pte_range(pte_t * pte, unsigned long address,
                                     unsigned long size, pgprot_t prot)
{
	unsigned long end;

	address &= ~PMD_MASK;
	end = address + size;
	if (end > PMD_SIZE)
		end = PMD_SIZE;
	do {
		pte_t zero_pte = pte_wrprotect(mk_pte(ZERO_PAGE(address), prot));
		pte_t oldpage = ptep_get_and_clear(pte);
		set_pte(pte, zero_pte);
		forget_pte(oldpage);
		address += PAGE_SIZE;
		pte++;
	} while (address && (address < end));
}

Linus Torvalds's avatar
Linus Torvalds committed
786
static inline int zeromap_pmd_range(struct mm_struct *mm, pmd_t * pmd, unsigned long address,
Linus Torvalds's avatar
Linus Torvalds committed
787 788 789 790 791 792 793 794 795
                                    unsigned long size, pgprot_t prot)
{
	unsigned long end;

	address &= ~PGDIR_MASK;
	end = address + size;
	if (end > PGDIR_SIZE)
		end = PGDIR_SIZE;
	do {
796
		pte_t * pte = pte_alloc_map(mm, pmd, address);
Linus Torvalds's avatar
Linus Torvalds committed
797 798 799
		if (!pte)
			return -ENOMEM;
		zeromap_pte_range(pte, address, end - address, prot);
800
		pte_unmap(pte);
Linus Torvalds's avatar
Linus Torvalds committed
801 802 803 804 805 806
		address = (address + PMD_SIZE) & PMD_MASK;
		pmd++;
	} while (address && (address < end));
	return 0;
}

Linus Torvalds's avatar
Linus Torvalds committed
807
int zeromap_page_range(struct vm_area_struct *vma, unsigned long address, unsigned long size, pgprot_t prot)
Linus Torvalds's avatar
Linus Torvalds committed
808 809 810 811 812
{
	int error = 0;
	pgd_t * dir;
	unsigned long beg = address;
	unsigned long end = address + size;
Linus Torvalds's avatar
Linus Torvalds committed
813
	struct mm_struct *mm = vma->vm_mm;
Linus Torvalds's avatar
Linus Torvalds committed
814

Linus Torvalds's avatar
Linus Torvalds committed
815
	dir = pgd_offset(mm, address);
Linus Torvalds's avatar
Linus Torvalds committed
816
	flush_cache_range(vma, beg, end);
Linus Torvalds's avatar
Linus Torvalds committed
817 818
	if (address >= end)
		BUG();
Linus Torvalds's avatar
Linus Torvalds committed
819 820

	spin_lock(&mm->page_table_lock);
Linus Torvalds's avatar
Linus Torvalds committed
821
	do {
Linus Torvalds's avatar
Linus Torvalds committed
822
		pmd_t *pmd = pmd_alloc(mm, dir, address);
Linus Torvalds's avatar
Linus Torvalds committed
823 824 825
		error = -ENOMEM;
		if (!pmd)
			break;
Linus Torvalds's avatar
Linus Torvalds committed
826
		error = zeromap_pmd_range(mm, pmd, address, end - address, prot);
Linus Torvalds's avatar
Linus Torvalds committed
827 828 829 830 831
		if (error)
			break;
		address = (address + PGDIR_SIZE) & PGDIR_MASK;
		dir++;
	} while (address && (address < end));
Linus Torvalds's avatar
Linus Torvalds committed
832
	flush_tlb_range(vma, beg, end);
833
	spin_unlock(&mm->page_table_lock);
Linus Torvalds's avatar
Linus Torvalds committed
834 835 836 837 838 839 840 841 842 843 844 845
	return error;
}

/*
 * maps a range of physical memory into the requested pages. the old
 * mappings are removed. any references to nonexistent pages results
 * in null mappings (currently treated as "copy-on-access")
 */
static inline void remap_pte_range(pte_t * pte, unsigned long address, unsigned long size,
	unsigned long phys_addr, pgprot_t prot)
{
	unsigned long end;
846
	unsigned long pfn;
Linus Torvalds's avatar
Linus Torvalds committed
847 848 849 850 851

	address &= ~PMD_MASK;
	end = address + size;
	if (end > PMD_SIZE)
		end = PMD_SIZE;
852
	pfn = phys_addr >> PAGE_SHIFT;
Linus Torvalds's avatar
Linus Torvalds committed
853
	do {
854
		pte_t oldpage = ptep_get_and_clear(pte);
Linus Torvalds's avatar
Linus Torvalds committed
855

856
		if (!pfn_valid(pfn) || PageReserved(pfn_to_page(pfn)))
857
 			set_pte(pte, pfn_pte(pfn, prot));
Linus Torvalds's avatar
Linus Torvalds committed
858 859
		forget_pte(oldpage);
		address += PAGE_SIZE;
860
		pfn++;
Linus Torvalds's avatar
Linus Torvalds committed
861 862 863 864
		pte++;
	} while (address && (address < end));
}

Linus Torvalds's avatar
Linus Torvalds committed
865
static inline int remap_pmd_range(struct mm_struct *mm, pmd_t * pmd, unsigned long address, unsigned long size,
Linus Torvalds's avatar
Linus Torvalds committed
866 867 868 869 870 871 872 873 874 875
	unsigned long phys_addr, pgprot_t prot)
{
	unsigned long end;

	address &= ~PGDIR_MASK;
	end = address + size;
	if (end > PGDIR_SIZE)
		end = PGDIR_SIZE;
	phys_addr -= address;
	do {
876
		pte_t * pte = pte_alloc_map(mm, pmd, address);
Linus Torvalds's avatar
Linus Torvalds committed
877 878 879
		if (!pte)
			return -ENOMEM;
		remap_pte_range(pte, address, end - address, address + phys_addr, prot);
880
		pte_unmap(pte);
Linus Torvalds's avatar
Linus Torvalds committed
881 882 883 884 885 886 887
		address = (address + PMD_SIZE) & PMD_MASK;
		pmd++;
	} while (address && (address < end));
	return 0;
}

/*  Note: this is only safe if the mm semaphore is held when called. */
Linus Torvalds's avatar
Linus Torvalds committed
888
int remap_page_range(struct vm_area_struct *vma, unsigned long from, unsigned long phys_addr, unsigned long size, pgprot_t prot)
Linus Torvalds's avatar
Linus Torvalds committed
889 890 891 892 893
{
	int error = 0;
	pgd_t * dir;
	unsigned long beg = from;
	unsigned long end = from + size;
Linus Torvalds's avatar
Linus Torvalds committed
894
	struct mm_struct *mm = vma->vm_mm;
Linus Torvalds's avatar
Linus Torvalds committed
895 896

	phys_addr -= from;
Linus Torvalds's avatar
Linus Torvalds committed
897
	dir = pgd_offset(mm, from);
Linus Torvalds's avatar
Linus Torvalds committed
898
	flush_cache_range(vma, beg, end);
Linus Torvalds's avatar
Linus Torvalds committed
899 900
	if (from >= end)
		BUG();
Linus Torvalds's avatar
Linus Torvalds committed
901 902

	spin_lock(&mm->page_table_lock);
Linus Torvalds's avatar
Linus Torvalds committed
903
	do {
Linus Torvalds's avatar
Linus Torvalds committed
904
		pmd_t *pmd = pmd_alloc(mm, dir, from);
Linus Torvalds's avatar
Linus Torvalds committed
905 906 907
		error = -ENOMEM;
		if (!pmd)
			break;
Linus Torvalds's avatar
Linus Torvalds committed
908
		error = remap_pmd_range(mm, pmd, from, end - from, phys_addr + from, prot);
Linus Torvalds's avatar
Linus Torvalds committed
909 910 911 912 913
		if (error)
			break;
		from = (from + PGDIR_SIZE) & PGDIR_MASK;
		dir++;
	} while (from && (from < end));
Linus Torvalds's avatar
Linus Torvalds committed
914
	flush_tlb_range(vma, beg, end);
915
	spin_unlock(&mm->page_table_lock);
Linus Torvalds's avatar
Linus Torvalds committed
916 917 918 919 920 921 922 923
	return error;
}

/*
 * Establish a new mapping:
 *  - flush the old one
 *  - update the page tables
 *  - inform the TLB about the new one
Linus Torvalds's avatar
Linus Torvalds committed
924 925
 *
 * We hold the mm semaphore for reading and vma->vm_mm->page_table_lock
Linus Torvalds's avatar
Linus Torvalds committed
926 927 928 929 930 931 932 933
 */
static inline void establish_pte(struct vm_area_struct * vma, unsigned long address, pte_t *page_table, pte_t entry)
{
	set_pte(page_table, entry);
	flush_tlb_page(vma, address);
	update_mmu_cache(vma, address, entry);
}

Linus Torvalds's avatar
Linus Torvalds committed
934 935 936
/*
 * We hold the mm semaphore for reading and vma->vm_mm->page_table_lock
 */
Linus Torvalds's avatar
Linus Torvalds committed
937
static inline void break_cow(struct vm_area_struct * vma, struct page * new_page, unsigned long address, 
Linus Torvalds's avatar
Linus Torvalds committed
938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961
		pte_t *page_table)
{
	flush_page_to_ram(new_page);
	flush_cache_page(vma, address);
	establish_pte(vma, address, page_table, pte_mkwrite(pte_mkdirty(mk_pte(new_page, vma->vm_page_prot))));
}

/*
 * This routine handles present pages, when users try to write
 * to a shared page. It is done by copying the page to a new address
 * and decrementing the shared-page counter for the old page.
 *
 * Goto-purists beware: the only reason for goto's here is that it results
 * in better assembly code.. The "default" path will see no jumps at all.
 *
 * Note that this routine assumes that the protection checks have been
 * done by the caller (the low-level page fault routine in most cases).
 * Thus we can safely just mark it writable once we've done any necessary
 * COW.
 *
 * We also mark the page dirty at this point even though the page will
 * change only once the write actually happens. This avoids a few races,
 * and potentially makes it more efficient.
 *
Linus Torvalds's avatar
Linus Torvalds committed
962 963
 * We hold the mm semaphore and the page_table_lock on entry and exit
 * with the page_table_lock released.
Linus Torvalds's avatar
Linus Torvalds committed
964 965
 */
static int do_wp_page(struct mm_struct *mm, struct vm_area_struct * vma,
966
	unsigned long address, pte_t *page_table, pmd_t *pmd, pte_t pte)
Linus Torvalds's avatar
Linus Torvalds committed
967 968
{
	struct page *old_page, *new_page;
969
	unsigned long pfn = pte_pfn(pte);
Linus Torvalds's avatar
Linus Torvalds committed
970

971
	if (!pfn_valid(pfn))
Linus Torvalds's avatar
Linus Torvalds committed
972
		goto bad_wp_page;
973
	old_page = pfn_to_page(pfn);
Linus Torvalds's avatar
Linus Torvalds committed
974

Andrew Morton's avatar
Andrew Morton committed
975
	if (!TestSetPageLocked(old_page)) {
Linus Torvalds's avatar
Linus Torvalds committed
976 977 978 979 980
		int reuse = can_share_swap_page(old_page);
		unlock_page(old_page);
		if (reuse) {
			flush_cache_page(vma, address);
			establish_pte(vma, address, page_table, pte_mkyoung(pte_mkdirty(pte_mkwrite(pte))));
981
			pte_unmap(page_table);
Linus Torvalds's avatar
Linus Torvalds committed
982 983 984
			spin_unlock(&mm->page_table_lock);
			return 1;	/* Minor fault */
		}
Linus Torvalds's avatar
Linus Torvalds committed
985
	}
986
	pte_unmap(page_table);
Linus Torvalds's avatar
Linus Torvalds committed
987 988 989 990

	/*
	 * Ok, we need to copy. Oh, well..
	 */
Linus Torvalds's avatar
Linus Torvalds committed
991
	page_cache_get(old_page);
Linus Torvalds's avatar
Linus Torvalds committed
992
	spin_unlock(&mm->page_table_lock);
Linus Torvalds's avatar
Linus Torvalds committed
993

Linus Torvalds's avatar
Linus Torvalds committed
994
	new_page = alloc_page(GFP_HIGHUSER);
Linus Torvalds's avatar
Linus Torvalds committed
995
	if (!new_page)
Linus Torvalds's avatar
Linus Torvalds committed
996 997
		goto no_mem;
	copy_cow_page(old_page,new_page,address);
Linus Torvalds's avatar
Linus Torvalds committed
998 999 1000 1001

	/*
	 * Re-check the pte - we dropped the lock
	 */
Linus Torvalds's avatar
Linus Torvalds committed
1002
	spin_lock(&mm->page_table_lock);
1003
	page_table = pte_offset_map(pmd, address);
Linus Torvalds's avatar
Linus Torvalds committed
1004 1005 1006
	if (pte_same(*page_table, pte)) {
		if (PageReserved(old_page))
			++mm->rss;
Linus Torvalds's avatar
Linus Torvalds committed
1007
		break_cow(vma, new_page, address, page_table);
Linus Torvalds's avatar
Linus Torvalds committed
1008
		lru_cache_add(new_page);
Linus Torvalds's avatar
Linus Torvalds committed
1009 1010 1011 1012

		/* Free the old page.. */
		new_page = old_page;
	}
1013
	pte_unmap(page_table);
Linus Torvalds's avatar
Linus Torvalds committed
1014
	spin_unlock(&mm->page_table_lock);
Linus Torvalds's avatar
Linus Torvalds committed
1015
	page_cache_release(new_page);
Linus Torvalds's avatar
Linus Torvalds committed
1016
	page_cache_release(old_page);
Linus Torvalds's avatar
Linus Torvalds committed
1017 1018 1019
	return 1;	/* Minor fault */

bad_wp_page:
1020
	pte_unmap(page_table);
Linus Torvalds's avatar
Linus Torvalds committed
1021
	spin_unlock(&mm->page_table_lock);
Anton Altaparmakov's avatar
Anton Altaparmakov committed
1022
	printk(KERN_ERR "do_wp_page: bogus page at address %08lx\n", address);
Linus Torvalds's avatar
Linus Torvalds committed
1023
	return -1;
Linus Torvalds's avatar
Linus Torvalds committed
1024
no_mem:
Linus Torvalds's avatar
Linus Torvalds committed
1025
	page_cache_release(old_page);
Linus Torvalds's avatar
Linus Torvalds committed
1026
	return -1;
Linus Torvalds's avatar
Linus Torvalds committed
1027 1028
}

1029
static void vmtruncate_list(list_t *head, unsigned long pgoff)
Linus Torvalds's avatar
Linus Torvalds committed
1030
{
1031 1032 1033 1034 1035 1036 1037 1038 1039
	unsigned long start, end, len, diff;
	struct vm_area_struct *vma;
	list_t *curr;

	list_for_each(curr, head) {
		vma = list_entry(curr, struct vm_area_struct, shared);
		start = vma->vm_start;
		end = vma->vm_end;
		len = end - start;
Linus Torvalds's avatar
Linus Torvalds committed
1040 1041

		/* mapping wholly truncated? */
1042 1043
		if (vma->vm_pgoff >= pgoff) {
			zap_page_range(vma, start, len);
Linus Torvalds's avatar
Linus Torvalds committed
1044 1045 1046 1047 1048
			continue;
		}

		/* mapping wholly unaffected? */
		len = len >> PAGE_SHIFT;
1049
		diff = pgoff - vma->vm_pgoff;
Linus Torvalds's avatar
Linus Torvalds committed
1050 1051 1052 1053 1054 1055
		if (diff >= len)
			continue;

		/* Ok, partially affected.. */
		start += diff << PAGE_SHIFT;
		len = (len - diff) << PAGE_SHIFT;
1056 1057
		zap_page_range(vma, start, len);
	}
Linus Torvalds's avatar
Linus Torvalds committed
1058 1059 1060 1061 1062 1063 1064 1065 1066 1067
}

/*
 * Handle all mappings that got truncated by a "truncate()"
 * system call.
 *
 * NOTE! We have to be ready to update the memory sharing
 * between the file and the memory map for a potential last
 * incomplete page.  Ugly, but necessary.
 */
Linus Torvalds's avatar
Linus Torvalds committed
1068
int vmtruncate(struct inode * inode, loff_t offset)
Linus Torvalds's avatar
Linus Torvalds committed
1069
{
Linus Torvalds's avatar
Linus Torvalds committed
1070
	unsigned long pgoff;
Linus Torvalds's avatar
Linus Torvalds committed
1071 1072 1073 1074 1075 1076 1077
	struct address_space *mapping = inode->i_mapping;
	unsigned long limit;

	if (inode->i_size < offset)
		goto do_expand;
	inode->i_size = offset;
	spin_lock(&mapping->i_shared_lock);
1078
	if (list_empty(&mapping->i_mmap) && list_empty(&mapping->i_mmap_shared))
Linus Torvalds's avatar
Linus Torvalds committed
1079 1080 1081
		goto out_unlock;

	pgoff = (offset + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1082 1083 1084 1085
	if (!list_empty(&mapping->i_mmap))
		vmtruncate_list(&mapping->i_mmap, pgoff);
	if (!list_empty(&mapping->i_mmap_shared))
		vmtruncate_list(&mapping->i_mmap_shared, pgoff);
Linus Torvalds's avatar
Linus Torvalds committed
1086 1087 1088

out_unlock:
	spin_unlock(&mapping->i_shared_lock);
Linus Torvalds's avatar
Linus Torvalds committed
1089
	truncate_inode_pages(mapping, offset);
Linus Torvalds's avatar
Linus Torvalds committed
1090
	goto out_truncate;
Linus Torvalds's avatar
Linus Torvalds committed
1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104

do_expand:
	limit = current->rlim[RLIMIT_FSIZE].rlim_cur;
	if (limit != RLIM_INFINITY) {
		if (inode->i_size >= limit) {
			send_sig(SIGXFSZ, current, 0);
			goto out;
		}
		if (offset > limit) {
			send_sig(SIGXFSZ, current, 0);
			offset = limit;
		}
	}
	inode->i_size = offset;
Linus Torvalds's avatar
Linus Torvalds committed
1105 1106

out_truncate:
1107
	if (inode->i_op && inode->i_op->truncate)
Linus Torvalds's avatar
Linus Torvalds committed
1108 1109
		inode->i_op->truncate(inode);
out:
Linus Torvalds's avatar
Linus Torvalds committed
1110
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125
}

/* 
 * Primitive swap readahead code. We simply read an aligned block of
 * (1 << page_cluster) entries in the swap area. This method is chosen
 * because it doesn't cost us any seek time.  We also make sure to queue
 * the 'original' request together with the readahead ones...  
 */
void swapin_readahead(swp_entry_t entry)
{
	int i, num;
	struct page *new_page;
	unsigned long offset;

	/*
Linus Torvalds's avatar
Linus Torvalds committed
1126
	 * Get the number of handles we should do readahead io to.
Linus Torvalds's avatar
Linus Torvalds committed
1127 1128 1129 1130
	 */
	num = valid_swaphandles(entry, &offset);
	for (i = 0; i < num; offset++, i++) {
		/* Ok, do the async read-ahead now */
Linus Torvalds's avatar
Linus Torvalds committed
1131
		new_page = read_swap_cache_async(SWP_ENTRY(SWP_TYPE(entry), offset));
Linus Torvalds's avatar
Linus Torvalds committed
1132 1133 1134
		if (!new_page)
			break;
		page_cache_release(new_page);
Linus Torvalds's avatar
Linus Torvalds committed
1135 1136 1137 1138
	}
	return;
}

Linus Torvalds's avatar
Linus Torvalds committed
1139
/*
Linus Torvalds's avatar
Linus Torvalds committed
1140 1141
 * We hold the mm semaphore and the page_table_lock on entry and
 * should release the pagetable lock on exit..
Linus Torvalds's avatar
Linus Torvalds committed
1142
 */
Linus Torvalds's avatar
Linus Torvalds committed
1143 1144
static int do_swap_page(struct mm_struct * mm,
	struct vm_area_struct * vma, unsigned long address,
1145
	pte_t *page_table, pmd_t *pmd, pte_t orig_pte, int write_access)
Linus Torvalds's avatar
Linus Torvalds committed
1146
{
Linus Torvalds's avatar
Linus Torvalds committed
1147
	struct page *page;
Linus Torvalds's avatar
Linus Torvalds committed
1148
	swp_entry_t entry = pte_to_swp_entry(orig_pte);
Linus Torvalds's avatar
Linus Torvalds committed
1149
	pte_t pte;
Linus Torvalds's avatar
Linus Torvalds committed
1150
	int ret = 1;
Linus Torvalds's avatar
Linus Torvalds committed
1151

1152
	pte_unmap(page_table);
Linus Torvalds's avatar
Linus Torvalds committed
1153 1154
	spin_unlock(&mm->page_table_lock);
	page = lookup_swap_cache(entry);
Linus Torvalds's avatar
Linus Torvalds committed
1155 1156
	if (!page) {
		swapin_readahead(entry);
Linus Torvalds's avatar
Linus Torvalds committed
1157
		page = read_swap_cache_async(entry);
Linus Torvalds's avatar
Linus Torvalds committed
1158
		if (!page) {
Linus Torvalds's avatar
Linus Torvalds committed
1159 1160 1161 1162
			/*
			 * Back out if somebody else faulted in this pte while
			 * we released the page table lock.
			 */
Linus Torvalds's avatar
Linus Torvalds committed
1163 1164
			int retval;
			spin_lock(&mm->page_table_lock);
1165
			page_table = pte_offset_map(pmd, address);
Linus Torvalds's avatar
Linus Torvalds committed
1166
			retval = pte_same(*page_table, orig_pte) ? -1 : 1;
1167
			pte_unmap(page_table);
Linus Torvalds's avatar
Linus Torvalds committed
1168 1169
			spin_unlock(&mm->page_table_lock);
			return retval;
Linus Torvalds's avatar
Linus Torvalds committed
1170
		}
Linus Torvalds's avatar
Linus Torvalds committed
1171 1172 1173

		/* Had to read the page from swap area: Major fault */
		ret = 2;
Linus Torvalds's avatar
Linus Torvalds committed
1174 1175
	}

Linus Torvalds's avatar
Linus Torvalds committed
1176
	lock_page(page);
Linus Torvalds's avatar
Linus Torvalds committed
1177 1178

	/*
Linus Torvalds's avatar
Linus Torvalds committed
1179 1180
	 * Back out if somebody else faulted in this pte while we
	 * released the page table lock.
Linus Torvalds's avatar
Linus Torvalds committed
1181
	 */
Linus Torvalds's avatar
Linus Torvalds committed
1182
	spin_lock(&mm->page_table_lock);
1183
	page_table = pte_offset_map(pmd, address);
Linus Torvalds's avatar
Linus Torvalds committed
1184
	if (!pte_same(*page_table, orig_pte)) {
1185
		pte_unmap(page_table);
Linus Torvalds's avatar
Linus Torvalds committed
1186
		spin_unlock(&mm->page_table_lock);
Linus Torvalds's avatar
Linus Torvalds committed
1187 1188
		unlock_page(page);
		page_cache_release(page);
Linus Torvalds's avatar
Linus Torvalds committed
1189 1190
		return 1;
	}
Linus Torvalds's avatar
Linus Torvalds committed
1191

Linus Torvalds's avatar
Linus Torvalds committed
1192
	/* The page isn't present yet, go ahead with the fault. */
Linus Torvalds's avatar
Linus Torvalds committed
1193 1194
		
	swap_free(entry);
Linus Torvalds's avatar
Linus Torvalds committed
1195
	if (vm_swap_full())
Linus Torvalds's avatar
Linus Torvalds committed
1196 1197
		remove_exclusive_swap_page(page);

Linus Torvalds's avatar
Linus Torvalds committed
1198 1199
	mm->rss++;
	pte = mk_pte(page, vma->vm_page_prot);
Linus Torvalds's avatar
Linus Torvalds committed
1200 1201
	if (write_access && can_share_swap_page(page))
		pte = pte_mkdirty(pte_mkwrite(pte));
Linus Torvalds's avatar
Linus Torvalds committed
1202
	unlock_page(page);
Linus Torvalds's avatar
Linus Torvalds committed
1203

Linus Torvalds's avatar
Linus Torvalds committed
1204 1205
	flush_page_to_ram(page);
	flush_icache_page(vma, page);
Linus Torvalds's avatar
Linus Torvalds committed
1206
	set_pte(page_table, pte);
Linus Torvalds's avatar
Linus Torvalds committed
1207

Linus Torvalds's avatar
Linus Torvalds committed
1208 1209
	/* No need to invalidate - it was non-present before */
	update_mmu_cache(vma, address, pte);
1210
	pte_unmap(page_table);
Linus Torvalds's avatar
Linus Torvalds committed
1211
	spin_unlock(&mm->page_table_lock);
Linus Torvalds's avatar
Linus Torvalds committed
1212
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
1213 1214 1215
}

/*
Linus Torvalds's avatar
Linus Torvalds committed
1216 1217 1218
 * We are called with the MM semaphore and page_table_lock
 * spinlock held to protect against concurrent faults in
 * multithreaded programs. 
Linus Torvalds's avatar
Linus Torvalds committed
1219
 */
1220
static int do_anonymous_page(struct mm_struct * mm, struct vm_area_struct * vma, pte_t *page_table, pmd_t *pmd, int write_access, unsigned long addr)
Linus Torvalds's avatar
Linus Torvalds committed
1221
{
Linus Torvalds's avatar
Linus Torvalds committed
1222 1223 1224 1225 1226 1227
	pte_t entry;

	/* Read-only mapping of ZERO_PAGE. */
	entry = pte_wrprotect(mk_pte(ZERO_PAGE(addr), vma->vm_page_prot));

	/* ..except if it's a write access */
Linus Torvalds's avatar
Linus Torvalds committed
1228
	if (write_access) {
Linus Torvalds's avatar
Linus Torvalds committed
1229 1230 1231
		struct page *page;

		/* Allocate our own private page. */
1232
		pte_unmap(page_table);
Linus Torvalds's avatar
Linus Torvalds committed
1233
		spin_unlock(&mm->page_table_lock);
Linus Torvalds's avatar
Linus Torvalds committed
1234

Linus Torvalds's avatar
Linus Torvalds committed
1235 1236
		page = alloc_page(GFP_HIGHUSER);
		if (!page)
Linus Torvalds's avatar
Linus Torvalds committed
1237 1238 1239 1240
			goto no_mem;
		clear_user_highpage(page, addr);

		spin_lock(&mm->page_table_lock);
1241 1242
		page_table = pte_offset_map(pmd, addr);

Linus Torvalds's avatar
Linus Torvalds committed
1243
		if (!pte_none(*page_table)) {
1244
			pte_unmap(page_table);
Linus Torvalds's avatar
Linus Torvalds committed
1245
			page_cache_release(page);
Linus Torvalds's avatar
Linus Torvalds committed
1246
			spin_unlock(&mm->page_table_lock);
Linus Torvalds's avatar
Linus Torvalds committed
1247 1248
			return 1;
		}
Linus Torvalds's avatar
Linus Torvalds committed
1249 1250
		mm->rss++;
		flush_page_to_ram(page);
Linus Torvalds's avatar
Linus Torvalds committed
1251
		entry = pte_mkwrite(pte_mkdirty(mk_pte(page, vma->vm_page_prot)));
Linus Torvalds's avatar
Linus Torvalds committed
1252
		lru_cache_add(page);
Linus Torvalds's avatar
Linus Torvalds committed
1253
	}
Linus Torvalds's avatar
Linus Torvalds committed
1254

Linus Torvalds's avatar
Linus Torvalds committed
1255
	set_pte(page_table, entry);
1256
	pte_unmap(page_table);
Linus Torvalds's avatar
Linus Torvalds committed
1257

Linus Torvalds's avatar
Linus Torvalds committed
1258 1259
	/* No need to invalidate - it was non-present before */
	update_mmu_cache(vma, addr, entry);
Linus Torvalds's avatar
Linus Torvalds committed
1260
	spin_unlock(&mm->page_table_lock);
Linus Torvalds's avatar
Linus Torvalds committed
1261
	return 1;	/* Minor fault */
Linus Torvalds's avatar
Linus Torvalds committed
1262 1263 1264

no_mem:
	return -1;
Linus Torvalds's avatar
Linus Torvalds committed
1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275
}

/*
 * do_no_page() tries to create a new page mapping. It aggressively
 * tries to share with existing pages, but makes a separate copy if
 * the "write_access" parameter is true in order to avoid the next
 * page fault.
 *
 * As this is called only for pages that do not currently exist, we
 * do not need to flush old virtual caches or the TLB.
 *
Linus Torvalds's avatar
Linus Torvalds committed
1276
 * This is called with the MM semaphore held and the page table
Linus Torvalds's avatar
Linus Torvalds committed
1277
 * spinlock held. Exit with the spinlock released.
Linus Torvalds's avatar
Linus Torvalds committed
1278 1279
 */
static int do_no_page(struct mm_struct * mm, struct vm_area_struct * vma,
1280
	unsigned long address, int write_access, pte_t *page_table, pmd_t *pmd)
Linus Torvalds's avatar
Linus Torvalds committed
1281 1282 1283 1284 1285
{
	struct page * new_page;
	pte_t entry;

	if (!vma->vm_ops || !vma->vm_ops->nopage)
1286 1287
		return do_anonymous_page(mm, vma, page_table, pmd, write_access, address);
	pte_unmap(page_table);
Linus Torvalds's avatar
Linus Torvalds committed
1288
	spin_unlock(&mm->page_table_lock);
Linus Torvalds's avatar
Linus Torvalds committed
1289

Linus Torvalds's avatar
Linus Torvalds committed
1290
	new_page = vma->vm_ops->nopage(vma, address & PAGE_MASK, 0);
Linus Torvalds's avatar
Linus Torvalds committed
1291

Linus Torvalds's avatar
Linus Torvalds committed
1292 1293 1294 1295
	if (new_page == NULL)	/* no page was available -- SIGBUS */
		return 0;
	if (new_page == NOPAGE_OOM)
		return -1;
Linus Torvalds's avatar
Linus Torvalds committed
1296 1297 1298 1299 1300 1301

	/*
	 * Should we do an early C-O-W break?
	 */
	if (write_access && !(vma->vm_flags & VM_SHARED)) {
		struct page * page = alloc_page(GFP_HIGHUSER);
Linus Torvalds's avatar
Linus Torvalds committed
1302 1303
		if (!page) {
			page_cache_release(new_page);
Linus Torvalds's avatar
Linus Torvalds committed
1304
			return -1;
Linus Torvalds's avatar
Linus Torvalds committed
1305
		}
Linus Torvalds's avatar
Linus Torvalds committed
1306
		copy_user_highpage(page, new_page, address);
Linus Torvalds's avatar
Linus Torvalds committed
1307 1308 1309 1310 1311 1312
		page_cache_release(new_page);
		lru_cache_add(page);
		new_page = page;
	}

	spin_lock(&mm->page_table_lock);
1313 1314
	page_table = pte_offset_map(pmd, address);

Linus Torvalds's avatar
Linus Torvalds committed
1315 1316 1317 1318 1319 1320 1321 1322 1323 1324
	/*
	 * This silly early PAGE_DIRTY setting removes a race
	 * due to the bad i386 page protection. But it's valid
	 * for other architectures too.
	 *
	 * Note that if write_access is true, we either now have
	 * an exclusive copy of the page, or this is a shared mapping,
	 * so we can make it writable and dirty to avoid having to
	 * handle that later.
	 */
Linus Torvalds's avatar
Linus Torvalds committed
1325 1326 1327 1328 1329 1330
	/* Only go through if we didn't race with anybody else... */
	if (pte_none(*page_table)) {
		++mm->rss;
		flush_page_to_ram(new_page);
		flush_icache_page(vma, new_page);
		entry = mk_pte(new_page, vma->vm_page_prot);
Linus Torvalds's avatar
Linus Torvalds committed
1331
		if (write_access)
Linus Torvalds's avatar
Linus Torvalds committed
1332 1333
			entry = pte_mkwrite(pte_mkdirty(entry));
		set_pte(page_table, entry);
1334
		pte_unmap(page_table);
Linus Torvalds's avatar
Linus Torvalds committed
1335 1336
	} else {
		/* One of our sibling threads was faster, back out. */
1337
		pte_unmap(page_table);
Linus Torvalds's avatar
Linus Torvalds committed
1338
		page_cache_release(new_page);
Linus Torvalds's avatar
Linus Torvalds committed
1339
		spin_unlock(&mm->page_table_lock);
Linus Torvalds's avatar
Linus Torvalds committed
1340
		return 1;
Linus Torvalds's avatar
Linus Torvalds committed
1341 1342
	}

Linus Torvalds's avatar
Linus Torvalds committed
1343 1344
	/* no need to invalidate: a not-present page shouldn't be cached */
	update_mmu_cache(vma, address, entry);
Linus Torvalds's avatar
Linus Torvalds committed
1345
	spin_unlock(&mm->page_table_lock);
Linus Torvalds's avatar
Linus Torvalds committed
1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365
	return 2;	/* Major fault */
}

/*
 * These routines also need to handle stuff like marking pages dirty
 * and/or accessed for architectures that don't do it in hardware (most
 * RISC architectures).  The early dirtying is also good on the i386.
 *
 * There is also a hook called "update_mmu_cache()" that architectures
 * with external mmu caches can use to update those (ie the Sparc or
 * PowerPC hashed page tables that act as extended TLBs).
 *
 * Note the "page_table_lock". It is to protect against kswapd removing
 * pages from under us. Note that kswapd only ever _removes_ pages, never
 * adds them. As such, once we have noticed that the page is not present,
 * we can drop the lock early.
 *
 * The adding of pages is protected by the MM semaphore (which we hold),
 * so we don't need to worry about a page being suddenly been added into
 * our VM.
Linus Torvalds's avatar
Linus Torvalds committed
1366 1367 1368
 *
 * We enter with the pagetable spinlock held, we are supposed to
 * release it when done.
Linus Torvalds's avatar
Linus Torvalds committed
1369 1370 1371
 */
static inline int handle_pte_fault(struct mm_struct *mm,
	struct vm_area_struct * vma, unsigned long address,
1372
	int write_access, pte_t *pte, pmd_t *pmd)
Linus Torvalds's avatar
Linus Torvalds committed
1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383
{
	pte_t entry;

	entry = *pte;
	if (!pte_present(entry)) {
		/*
		 * If it truly wasn't present, we know that kswapd
		 * and the PTE updates will not touch it later. So
		 * drop the lock.
		 */
		if (pte_none(entry))
1384 1385
			return do_no_page(mm, vma, address, write_access, pte, pmd);
		return do_swap_page(mm, vma, address, pte, pmd, entry, write_access);
Linus Torvalds's avatar
Linus Torvalds committed
1386 1387 1388 1389
	}

	if (write_access) {
		if (!pte_write(entry))
1390
			return do_wp_page(mm, vma, address, pte, pmd, entry);
Linus Torvalds's avatar
Linus Torvalds committed
1391 1392 1393 1394 1395

		entry = pte_mkdirty(entry);
	}
	entry = pte_mkyoung(entry);
	establish_pte(vma, address, pte, entry);
1396
	pte_unmap(pte);
Linus Torvalds's avatar
Linus Torvalds committed
1397
	spin_unlock(&mm->page_table_lock);
Linus Torvalds's avatar
Linus Torvalds committed
1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409
	return 1;
}

/*
 * By the time we get here, we already hold the mm semaphore
 */
int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct * vma,
	unsigned long address, int write_access)
{
	pgd_t *pgd;
	pmd_t *pmd;

Linus Torvalds's avatar
Linus Torvalds committed
1410
	current->state = TASK_RUNNING;
Linus Torvalds's avatar
Linus Torvalds committed
1411
	pgd = pgd_offset(mm, address);
Linus Torvalds's avatar
Linus Torvalds committed
1412 1413 1414 1415 1416 1417 1418

	/*
	 * We need the page table lock to synchronize with kswapd
	 * and the SMP-safe atomic PTE updates.
	 */
	spin_lock(&mm->page_table_lock);
	pmd = pmd_alloc(mm, pgd, address);
Linus Torvalds's avatar
Linus Torvalds committed
1419 1420

	if (pmd) {
1421
		pte_t * pte = pte_alloc_map(mm, pmd, address);
Linus Torvalds's avatar
Linus Torvalds committed
1422
		if (pte)
1423
			return handle_pte_fault(mm, vma, address, write_access, pte, pmd);
Linus Torvalds's avatar
Linus Torvalds committed
1424
	}
Linus Torvalds's avatar
Linus Torvalds committed
1425
	spin_unlock(&mm->page_table_lock);
Linus Torvalds's avatar
Linus Torvalds committed
1426
	return -1;
Linus Torvalds's avatar
Linus Torvalds committed
1427 1428
}

Linus Torvalds's avatar
Linus Torvalds committed
1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441
/*
 * Allocate page middle directory.
 *
 * We've already handled the fast-path in-line, and we own the
 * page table lock.
 *
 * On a two-level page table, this ends up actually being entirely
 * optimized away.
 */
pmd_t *__pmd_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
{
	pmd_t *new;

1442 1443 1444 1445 1446
	spin_unlock(&mm->page_table_lock);
	new = pmd_alloc_one(mm, address);
	spin_lock(&mm->page_table_lock);
	if (!new)
		return NULL;
Linus Torvalds's avatar
Linus Torvalds committed
1447

1448 1449 1450 1451 1452 1453 1454
	/*
	 * Because we dropped the lock, we should re-check the
	 * entry, as somebody else could have populated it..
	 */
	if (pgd_present(*pgd)) {
		pmd_free(new);
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
1455
	}
Linus Torvalds's avatar
Linus Torvalds committed
1456
	pgd_populate(mm, pgd, new);
Linus Torvalds's avatar
Linus Torvalds committed
1457 1458 1459 1460
out:
	return pmd_offset(pgd, address);
}

Linus Torvalds's avatar
Linus Torvalds committed
1461 1462
int make_pages_present(unsigned long addr, unsigned long end)
{
Linus Torvalds's avatar
Linus Torvalds committed
1463
	int ret, len, write;
Linus Torvalds's avatar
Linus Torvalds committed
1464 1465
	struct vm_area_struct * vma;

Linus Torvalds's avatar
Linus Torvalds committed
1466
	vma = find_vma(current->mm, addr);
Linus Torvalds's avatar
Linus Torvalds committed
1467 1468 1469
	write = (vma->vm_flags & VM_WRITE) != 0;
	if (addr >= end)
		BUG();
Linus Torvalds's avatar
Linus Torvalds committed
1470 1471 1472 1473 1474 1475
	if (end > vma->vm_end)
		BUG();
	len = (end+PAGE_SIZE-1)/PAGE_SIZE-addr/PAGE_SIZE;
	ret = get_user_pages(current, current->mm, addr,
			len, write, 0, NULL, NULL);
	return ret == len ? 0 : -1;
Linus Torvalds's avatar
Linus Torvalds committed
1476
}
1477 1478 1479 1480

/* 
 * Map a vmalloc()-space virtual address to the physical page.
 */
1481
struct page * vmalloc_to_page(void * vmalloc_addr)
1482
{
1483
	unsigned long addr = (unsigned long) vmalloc_addr;
1484
	struct page *page = NULL;
1485
	pgd_t *pgd = pgd_offset_k(addr);
1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502
	pmd_t *pmd;
	pte_t *ptep, pte;
  
	if (!pgd_none(*pgd)) {
		pmd = pmd_offset(pgd, addr);
		if (!pmd_none(*pmd)) {
			preempt_disable();
			ptep = pte_offset_map(pmd, addr);
			pte = *ptep;
			if (pte_present(pte))
				page = pte_page(pte);
			pte_unmap(ptep);
			preempt_enable();
		}
	}
	return page;
}