vmalloc.c 70.3 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7
/*
 *  linux/mm/vmalloc.c
 *
 *  Copyright (C) 1993  Linus Torvalds
 *  Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999
 *  SMP-safe vmalloc/vfree/ioremap, Tigran Aivazian <tigran@veritas.com>, May 2000
 *  Major rework to support vmap/vunmap, Christoph Hellwig, SGI, August 2002
C
Christoph Lameter 已提交
8
 *  Numa awareness, Christoph Lameter, SGI, June 2005
L
Linus Torvalds 已提交
9 10
 */

N
Nick Piggin 已提交
11
#include <linux/vmalloc.h>
L
Linus Torvalds 已提交
12 13 14
#include <linux/mm.h>
#include <linux/module.h>
#include <linux/highmem.h>
15
#include <linux/sched/signal.h>
L
Linus Torvalds 已提交
16 17 18
#include <linux/slab.h>
#include <linux/spinlock.h>
#include <linux/interrupt.h>
19
#include <linux/proc_fs.h>
20
#include <linux/seq_file.h>
21
#include <linux/debugobjects.h>
22
#include <linux/kallsyms.h>
N
Nick Piggin 已提交
23
#include <linux/list.h>
24
#include <linux/notifier.h>
N
Nick Piggin 已提交
25 26 27
#include <linux/rbtree.h>
#include <linux/radix-tree.h>
#include <linux/rcupdate.h>
28
#include <linux/pfn.h>
29
#include <linux/kmemleak.h>
A
Arun Sharma 已提交
30
#include <linux/atomic.h>
31
#include <linux/compiler.h>
32
#include <linux/llist.h>
33
#include <linux/bitops.h>
34

35
#include <linux/uaccess.h>
L
Linus Torvalds 已提交
36
#include <asm/tlbflush.h>
37
#include <asm/shmparam.h>
L
Linus Torvalds 已提交
38

39 40
#include "internal.h"

41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
struct vfree_deferred {
	struct llist_head list;
	struct work_struct wq;
};
static DEFINE_PER_CPU(struct vfree_deferred, vfree_deferred);

static void __vunmap(const void *, int);

static void free_work(struct work_struct *w)
{
	struct vfree_deferred *p = container_of(w, struct vfree_deferred, wq);
	struct llist_node *llnode = llist_del_all(&p->list);
	while (llnode) {
		void *p = llnode;
		llnode = llist_next(llnode);
		__vunmap(p, 1);
	}
}

N
Nick Piggin 已提交
60
/*** Page table manipulation functions ***/
A
Adrian Bunk 已提交
61

L
Linus Torvalds 已提交
62 63 64 65 66 67 68 69 70 71 72
static void vunmap_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end)
{
	pte_t *pte;

	pte = pte_offset_kernel(pmd, addr);
	do {
		pte_t ptent = ptep_get_and_clear(&init_mm, addr, pte);
		WARN_ON(!pte_none(ptent) && !pte_present(ptent));
	} while (pte++, addr += PAGE_SIZE, addr != end);
}

N
Nick Piggin 已提交
73
static void vunmap_pmd_range(pud_t *pud, unsigned long addr, unsigned long end)
L
Linus Torvalds 已提交
74 75 76 77 78 79 80
{
	pmd_t *pmd;
	unsigned long next;

	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
81 82
		if (pmd_clear_huge(pmd))
			continue;
L
Linus Torvalds 已提交
83 84 85 86 87 88
		if (pmd_none_or_clear_bad(pmd))
			continue;
		vunmap_pte_range(pmd, addr, next);
	} while (pmd++, addr = next, addr != end);
}

89
static void vunmap_pud_range(p4d_t *p4d, unsigned long addr, unsigned long end)
L
Linus Torvalds 已提交
90 91 92 93
{
	pud_t *pud;
	unsigned long next;

94
	pud = pud_offset(p4d, addr);
L
Linus Torvalds 已提交
95 96
	do {
		next = pud_addr_end(addr, end);
97 98
		if (pud_clear_huge(pud))
			continue;
L
Linus Torvalds 已提交
99 100 101 102 103 104
		if (pud_none_or_clear_bad(pud))
			continue;
		vunmap_pmd_range(pud, addr, next);
	} while (pud++, addr = next, addr != end);
}

105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120
static void vunmap_p4d_range(pgd_t *pgd, unsigned long addr, unsigned long end)
{
	p4d_t *p4d;
	unsigned long next;

	p4d = p4d_offset(pgd, addr);
	do {
		next = p4d_addr_end(addr, end);
		if (p4d_clear_huge(p4d))
			continue;
		if (p4d_none_or_clear_bad(p4d))
			continue;
		vunmap_pud_range(p4d, addr, next);
	} while (p4d++, addr = next, addr != end);
}

N
Nick Piggin 已提交
121
static void vunmap_page_range(unsigned long addr, unsigned long end)
L
Linus Torvalds 已提交
122 123 124 125 126 127 128 129 130 131
{
	pgd_t *pgd;
	unsigned long next;

	BUG_ON(addr >= end);
	pgd = pgd_offset_k(addr);
	do {
		next = pgd_addr_end(addr, end);
		if (pgd_none_or_clear_bad(pgd))
			continue;
132
		vunmap_p4d_range(pgd, addr, next);
L
Linus Torvalds 已提交
133 134 135 136
	} while (pgd++, addr = next, addr != end);
}

static int vmap_pte_range(pmd_t *pmd, unsigned long addr,
N
Nick Piggin 已提交
137
		unsigned long end, pgprot_t prot, struct page **pages, int *nr)
L
Linus Torvalds 已提交
138 139 140
{
	pte_t *pte;

N
Nick Piggin 已提交
141 142 143 144 145
	/*
	 * nr is a running index into the array which helps higher level
	 * callers keep track of where we're up to.
	 */

H
Hugh Dickins 已提交
146
	pte = pte_alloc_kernel(pmd, addr);
L
Linus Torvalds 已提交
147 148 149
	if (!pte)
		return -ENOMEM;
	do {
N
Nick Piggin 已提交
150 151 152 153 154
		struct page *page = pages[*nr];

		if (WARN_ON(!pte_none(*pte)))
			return -EBUSY;
		if (WARN_ON(!page))
L
Linus Torvalds 已提交
155 156
			return -ENOMEM;
		set_pte_at(&init_mm, addr, pte, mk_pte(page, prot));
N
Nick Piggin 已提交
157
		(*nr)++;
L
Linus Torvalds 已提交
158 159 160 161
	} while (pte++, addr += PAGE_SIZE, addr != end);
	return 0;
}

N
Nick Piggin 已提交
162 163
static int vmap_pmd_range(pud_t *pud, unsigned long addr,
		unsigned long end, pgprot_t prot, struct page **pages, int *nr)
L
Linus Torvalds 已提交
164 165 166 167 168 169 170 171 172
{
	pmd_t *pmd;
	unsigned long next;

	pmd = pmd_alloc(&init_mm, pud, addr);
	if (!pmd)
		return -ENOMEM;
	do {
		next = pmd_addr_end(addr, end);
N
Nick Piggin 已提交
173
		if (vmap_pte_range(pmd, addr, next, prot, pages, nr))
L
Linus Torvalds 已提交
174 175 176 177 178
			return -ENOMEM;
	} while (pmd++, addr = next, addr != end);
	return 0;
}

179
static int vmap_pud_range(p4d_t *p4d, unsigned long addr,
N
Nick Piggin 已提交
180
		unsigned long end, pgprot_t prot, struct page **pages, int *nr)
L
Linus Torvalds 已提交
181 182 183 184
{
	pud_t *pud;
	unsigned long next;

185
	pud = pud_alloc(&init_mm, p4d, addr);
L
Linus Torvalds 已提交
186 187 188 189
	if (!pud)
		return -ENOMEM;
	do {
		next = pud_addr_end(addr, end);
N
Nick Piggin 已提交
190
		if (vmap_pmd_range(pud, addr, next, prot, pages, nr))
L
Linus Torvalds 已提交
191 192 193 194 195
			return -ENOMEM;
	} while (pud++, addr = next, addr != end);
	return 0;
}

196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212
static int vmap_p4d_range(pgd_t *pgd, unsigned long addr,
		unsigned long end, pgprot_t prot, struct page **pages, int *nr)
{
	p4d_t *p4d;
	unsigned long next;

	p4d = p4d_alloc(&init_mm, pgd, addr);
	if (!p4d)
		return -ENOMEM;
	do {
		next = p4d_addr_end(addr, end);
		if (vmap_pud_range(p4d, addr, next, prot, pages, nr))
			return -ENOMEM;
	} while (p4d++, addr = next, addr != end);
	return 0;
}

N
Nick Piggin 已提交
213 214 215 216 217 218
/*
 * Set up page tables in kva (addr, end). The ptes shall have prot "prot", and
 * will have pfns corresponding to the "pages" array.
 *
 * Ie. pte at addr+N*PAGE_SIZE shall point to pfn corresponding to pages[N]
 */
219 220
static int vmap_page_range_noflush(unsigned long start, unsigned long end,
				   pgprot_t prot, struct page **pages)
L
Linus Torvalds 已提交
221 222 223
{
	pgd_t *pgd;
	unsigned long next;
224
	unsigned long addr = start;
N
Nick Piggin 已提交
225 226
	int err = 0;
	int nr = 0;
L
Linus Torvalds 已提交
227 228 229 230 231

	BUG_ON(addr >= end);
	pgd = pgd_offset_k(addr);
	do {
		next = pgd_addr_end(addr, end);
232
		err = vmap_p4d_range(pgd, addr, next, prot, pages, &nr);
L
Linus Torvalds 已提交
233
		if (err)
234
			return err;
L
Linus Torvalds 已提交
235
	} while (pgd++, addr = next, addr != end);
N
Nick Piggin 已提交
236 237

	return nr;
L
Linus Torvalds 已提交
238 239
}

240 241 242 243 244 245 246 247 248 249
static int vmap_page_range(unsigned long start, unsigned long end,
			   pgprot_t prot, struct page **pages)
{
	int ret;

	ret = vmap_page_range_noflush(start, end, prot, pages);
	flush_cache_vmap(start, end);
	return ret;
}

250
int is_vmalloc_or_module_addr(const void *x)
251 252
{
	/*
253
	 * ARM, x86-64 and sparc64 put modules in a special place,
254 255 256 257 258 259 260 261 262 263 264
	 * and fall back on vmalloc() if that fails. Others
	 * just put it in the vmalloc space.
	 */
#if defined(CONFIG_MODULES) && defined(MODULES_VADDR)
	unsigned long addr = (unsigned long)x;
	if (addr >= MODULES_VADDR && addr < MODULES_END)
		return 1;
#endif
	return is_vmalloc_addr(x);
}

265
/*
266
 * Walk a vmap address to the struct page it maps.
267
 */
268
struct page *vmalloc_to_page(const void *vmalloc_addr)
269 270
{
	unsigned long addr = (unsigned long) vmalloc_addr;
271
	struct page *page = NULL;
272
	pgd_t *pgd = pgd_offset_k(addr);
273 274 275 276
	p4d_t *p4d;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *ptep, pte;
277

278 279 280 281
	/*
	 * XXX we might need to change this if we add VIRTUAL_BUG_ON for
	 * architectures that do not vmalloc module space
	 */
282
	VIRTUAL_BUG_ON(!is_vmalloc_or_module_addr(vmalloc_addr));
J
Jiri Slaby 已提交
283

284 285 286 287 288 289
	if (pgd_none(*pgd))
		return NULL;
	p4d = p4d_offset(pgd, addr);
	if (p4d_none(*p4d))
		return NULL;
	pud = pud_offset(p4d, addr);
290 291 292 293 294 295 296 297 298 299 300

	/*
	 * Don't dereference bad PUD or PMD (below) entries. This will also
	 * identify huge mappings, which we may encounter on architectures
	 * that define CONFIG_HAVE_ARCH_HUGE_VMAP=y. Such regions will be
	 * identified as vmalloc addresses by is_vmalloc_addr(), but are
	 * not [unambiguously] associated with a struct page, so there is
	 * no correct value to return for them.
	 */
	WARN_ON_ONCE(pud_bad(*pud));
	if (pud_none(*pud) || pud_bad(*pud))
301 302
		return NULL;
	pmd = pmd_offset(pud, addr);
303 304
	WARN_ON_ONCE(pmd_bad(*pmd));
	if (pmd_none(*pmd) || pmd_bad(*pmd))
305 306 307 308 309 310 311
		return NULL;

	ptep = pte_offset_map(pmd, addr);
	pte = *ptep;
	if (pte_present(pte))
		page = pte_page(pte);
	pte_unmap(ptep);
312
	return page;
313
}
314
EXPORT_SYMBOL(vmalloc_to_page);
315 316

/*
317
 * Map a vmalloc()-space virtual address to the physical page frame number.
318
 */
319
unsigned long vmalloc_to_pfn(const void *vmalloc_addr)
320
{
321
	return page_to_pfn(vmalloc_to_page(vmalloc_addr));
322
}
323
EXPORT_SYMBOL(vmalloc_to_pfn);
324

N
Nick Piggin 已提交
325 326 327

/*** Global kva allocator ***/

328
#define VM_LAZY_FREE	0x02
N
Nick Piggin 已提交
329 330 331
#define VM_VM_AREA	0x04

static DEFINE_SPINLOCK(vmap_area_lock);
332 333
/* Export for kexec only */
LIST_HEAD(vmap_area_list);
334
static LLIST_HEAD(vmap_purge_list);
N
Nick Piggin 已提交
335 336 337 338 339 340 341 342
static struct rb_root vmap_area_root = RB_ROOT;

/* The vmap cache globals are protected by vmap_area_lock */
static struct rb_node *free_vmap_cache;
static unsigned long cached_hole_size;
static unsigned long cached_vstart;
static unsigned long cached_align;

343
static unsigned long vmap_area_pcpu_hole;
N
Nick Piggin 已提交
344 345

static struct vmap_area *__find_vmap_area(unsigned long addr)
L
Linus Torvalds 已提交
346
{
N
Nick Piggin 已提交
347 348 349 350 351 352 353 354
	struct rb_node *n = vmap_area_root.rb_node;

	while (n) {
		struct vmap_area *va;

		va = rb_entry(n, struct vmap_area, rb_node);
		if (addr < va->va_start)
			n = n->rb_left;
355
		else if (addr >= va->va_end)
N
Nick Piggin 已提交
356 357 358 359 360 361 362 363 364 365 366 367 368 369 370
			n = n->rb_right;
		else
			return va;
	}

	return NULL;
}

static void __insert_vmap_area(struct vmap_area *va)
{
	struct rb_node **p = &vmap_area_root.rb_node;
	struct rb_node *parent = NULL;
	struct rb_node *tmp;

	while (*p) {
371
		struct vmap_area *tmp_va;
N
Nick Piggin 已提交
372 373

		parent = *p;
374 375
		tmp_va = rb_entry(parent, struct vmap_area, rb_node);
		if (va->va_start < tmp_va->va_end)
N
Nick Piggin 已提交
376
			p = &(*p)->rb_left;
377
		else if (va->va_end > tmp_va->va_start)
N
Nick Piggin 已提交
378 379 380 381 382 383 384 385
			p = &(*p)->rb_right;
		else
			BUG();
	}

	rb_link_node(&va->rb_node, parent, p);
	rb_insert_color(&va->rb_node, &vmap_area_root);

386
	/* address-sort this list */
N
Nick Piggin 已提交
387 388 389 390 391 392 393 394 395 396 397
	tmp = rb_prev(&va->rb_node);
	if (tmp) {
		struct vmap_area *prev;
		prev = rb_entry(tmp, struct vmap_area, rb_node);
		list_add_rcu(&va->list, &prev->list);
	} else
		list_add_rcu(&va->list, &vmap_area_list);
}

static void purge_vmap_area_lazy(void);

398 399
static BLOCKING_NOTIFIER_HEAD(vmap_notify_list);

N
Nick Piggin 已提交
400 401 402 403 404 405 406 407 408 409 410
/*
 * Allocate a region of KVA of the specified size and alignment, within the
 * vstart and vend.
 */
static struct vmap_area *alloc_vmap_area(unsigned long size,
				unsigned long align,
				unsigned long vstart, unsigned long vend,
				int node, gfp_t gfp_mask)
{
	struct vmap_area *va;
	struct rb_node *n;
L
Linus Torvalds 已提交
411
	unsigned long addr;
N
Nick Piggin 已提交
412
	int purged = 0;
N
Nick Piggin 已提交
413
	struct vmap_area *first;
N
Nick Piggin 已提交
414

N
Nick Piggin 已提交
415
	BUG_ON(!size);
416
	BUG_ON(offset_in_page(size));
N
Nick Piggin 已提交
417
	BUG_ON(!is_power_of_2(align));
N
Nick Piggin 已提交
418

419
	might_sleep();
420

N
Nick Piggin 已提交
421 422 423 424 425
	va = kmalloc_node(sizeof(struct vmap_area),
			gfp_mask & GFP_RECLAIM_MASK, node);
	if (unlikely(!va))
		return ERR_PTR(-ENOMEM);

426 427 428 429 430 431
	/*
	 * Only scan the relevant parts containing pointers to other objects
	 * to avoid false negatives.
	 */
	kmemleak_scan_area(&va->rb_node, SIZE_MAX, gfp_mask & GFP_RECLAIM_MASK);

N
Nick Piggin 已提交
432 433
retry:
	spin_lock(&vmap_area_lock);
N
Nick Piggin 已提交
434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457
	/*
	 * Invalidate cache if we have more permissive parameters.
	 * cached_hole_size notes the largest hole noticed _below_
	 * the vmap_area cached in free_vmap_cache: if size fits
	 * into that hole, we want to scan from vstart to reuse
	 * the hole instead of allocating above free_vmap_cache.
	 * Note that __free_vmap_area may update free_vmap_cache
	 * without updating cached_hole_size or cached_align.
	 */
	if (!free_vmap_cache ||
			size < cached_hole_size ||
			vstart < cached_vstart ||
			align < cached_align) {
nocache:
		cached_hole_size = 0;
		free_vmap_cache = NULL;
	}
	/* record if we encounter less permissive parameters */
	cached_vstart = vstart;
	cached_align = align;

	/* find starting point for our search */
	if (free_vmap_cache) {
		first = rb_entry(free_vmap_cache, struct vmap_area, rb_node);
458
		addr = ALIGN(first->va_end, align);
N
Nick Piggin 已提交
459 460
		if (addr < vstart)
			goto nocache;
461
		if (addr + size < addr)
N
Nick Piggin 已提交
462 463 464 465
			goto overflow;

	} else {
		addr = ALIGN(vstart, align);
466
		if (addr + size < addr)
N
Nick Piggin 已提交
467 468 469 470 471 472
			goto overflow;

		n = vmap_area_root.rb_node;
		first = NULL;

		while (n) {
N
Nick Piggin 已提交
473 474 475 476
			struct vmap_area *tmp;
			tmp = rb_entry(n, struct vmap_area, rb_node);
			if (tmp->va_end >= addr) {
				first = tmp;
N
Nick Piggin 已提交
477 478 479 480
				if (tmp->va_start <= addr)
					break;
				n = n->rb_left;
			} else
N
Nick Piggin 已提交
481
				n = n->rb_right;
N
Nick Piggin 已提交
482
		}
N
Nick Piggin 已提交
483 484 485 486

		if (!first)
			goto found;
	}
N
Nick Piggin 已提交
487 488

	/* from the starting point, walk areas until a suitable hole is found */
489
	while (addr + size > first->va_start && addr + size <= vend) {
N
Nick Piggin 已提交
490 491
		if (addr + cached_hole_size < first->va_start)
			cached_hole_size = first->va_start - addr;
492
		addr = ALIGN(first->va_end, align);
493
		if (addr + size < addr)
N
Nick Piggin 已提交
494 495
			goto overflow;

496
		if (list_is_last(&first->list, &vmap_area_list))
N
Nick Piggin 已提交
497
			goto found;
498

499
		first = list_next_entry(first, list);
N
Nick Piggin 已提交
500 501
	}

N
Nick Piggin 已提交
502 503 504
found:
	if (addr + size > vend)
		goto overflow;
N
Nick Piggin 已提交
505 506 507 508 509

	va->va_start = addr;
	va->va_end = addr + size;
	va->flags = 0;
	__insert_vmap_area(va);
N
Nick Piggin 已提交
510
	free_vmap_cache = &va->rb_node;
N
Nick Piggin 已提交
511 512
	spin_unlock(&vmap_area_lock);

513
	BUG_ON(!IS_ALIGNED(va->va_start, align));
N
Nick Piggin 已提交
514 515 516
	BUG_ON(va->va_start < vstart);
	BUG_ON(va->va_end > vend);

N
Nick Piggin 已提交
517
	return va;
N
Nick Piggin 已提交
518 519 520 521 522 523 524 525

overflow:
	spin_unlock(&vmap_area_lock);
	if (!purged) {
		purge_vmap_area_lazy();
		purged = 1;
		goto retry;
	}
526 527 528 529 530 531 532 533 534 535

	if (gfpflags_allow_blocking(gfp_mask)) {
		unsigned long freed = 0;
		blocking_notifier_call_chain(&vmap_notify_list, 0, &freed);
		if (freed > 0) {
			purged = 0;
			goto retry;
		}
	}

536
	if (!(gfp_mask & __GFP_NOWARN) && printk_ratelimit())
J
Joe Perches 已提交
537 538
		pr_warn("vmap allocation for size %lu failed: use vmalloc=<size> to increase size\n",
			size);
N
Nick Piggin 已提交
539 540
	kfree(va);
	return ERR_PTR(-EBUSY);
N
Nick Piggin 已提交
541 542
}

543 544 545 546 547 548 549 550 551 552 553 554
int register_vmap_purge_notifier(struct notifier_block *nb)
{
	return blocking_notifier_chain_register(&vmap_notify_list, nb);
}
EXPORT_SYMBOL_GPL(register_vmap_purge_notifier);

int unregister_vmap_purge_notifier(struct notifier_block *nb)
{
	return blocking_notifier_chain_unregister(&vmap_notify_list, nb);
}
EXPORT_SYMBOL_GPL(unregister_vmap_purge_notifier);

N
Nick Piggin 已提交
555 556 557
static void __free_vmap_area(struct vmap_area *va)
{
	BUG_ON(RB_EMPTY_NODE(&va->rb_node));
N
Nick Piggin 已提交
558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573

	if (free_vmap_cache) {
		if (va->va_end < cached_vstart) {
			free_vmap_cache = NULL;
		} else {
			struct vmap_area *cache;
			cache = rb_entry(free_vmap_cache, struct vmap_area, rb_node);
			if (va->va_start <= cache->va_start) {
				free_vmap_cache = rb_prev(&va->rb_node);
				/*
				 * We don't try to update cached_hole_size or
				 * cached_align, but it won't go very wrong.
				 */
			}
		}
	}
N
Nick Piggin 已提交
574 575 576 577
	rb_erase(&va->rb_node, &vmap_area_root);
	RB_CLEAR_NODE(&va->rb_node);
	list_del_rcu(&va->list);

578 579 580 581 582 583 584 585 586
	/*
	 * Track the highest possible candidate for pcpu area
	 * allocation.  Areas outside of vmalloc area can be returned
	 * here too, consider only end addresses which fall inside
	 * vmalloc area proper.
	 */
	if (va->va_end > VMALLOC_START && va->va_end <= VMALLOC_END)
		vmap_area_pcpu_hole = max(vmap_area_pcpu_hole, va->va_end);

587
	kfree_rcu(va, rcu_head);
N
Nick Piggin 已提交
588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607
}

/*
 * Free a region of KVA allocated by alloc_vmap_area
 */
static void free_vmap_area(struct vmap_area *va)
{
	spin_lock(&vmap_area_lock);
	__free_vmap_area(va);
	spin_unlock(&vmap_area_lock);
}

/*
 * Clear the pagetable entries of a given vmap_area
 */
static void unmap_vmap_area(struct vmap_area *va)
{
	vunmap_page_range(va->va_start, va->va_end);
}

608 609 610
static void vmap_debug_free_range(unsigned long start, unsigned long end)
{
	/*
611 612 613 614
	 * Unmap page tables and force a TLB flush immediately if pagealloc
	 * debugging is enabled.  This catches use after free bugs similarly to
	 * those in linear kernel virtual address space after a page has been
	 * freed.
615
	 *
616 617
	 * All the lazy freeing logic is still retained, in order to minimise
	 * intrusiveness of this debugging feature.
618
	 *
619 620
	 * This is going to be *slow* (linear kernel virtual address debugging
	 * doesn't do a broadcast TLB flush so it is a lot faster).
621
	 */
622 623 624 625
	if (debug_pagealloc_enabled()) {
		vunmap_page_range(start, end);
		flush_tlb_kernel_range(start, end);
	}
626 627
}

N
Nick Piggin 已提交
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
/*
 * lazy_max_pages is the maximum amount of virtual address space we gather up
 * before attempting to purge with a TLB flush.
 *
 * There is a tradeoff here: a larger number will cover more kernel page tables
 * and take slightly longer to purge, but it will linearly reduce the number of
 * global TLB flushes that must be performed. It would seem natural to scale
 * this number up linearly with the number of CPUs (because vmapping activity
 * could also scale linearly with the number of CPUs), however it is likely
 * that in practice, workloads might be constrained in other ways that mean
 * vmap activity will not scale linearly with CPUs. Also, I want to be
 * conservative and not introduce a big latency on huge systems, so go with
 * a less aggressive log scale. It will still be an improvement over the old
 * code, and it will be simple to change the scale factor if we find that it
 * becomes a problem on bigger systems.
 */
static unsigned long lazy_max_pages(void)
{
	unsigned int log;

	log = fls(num_online_cpus());

	return log * (32UL * 1024 * 1024 / PAGE_SIZE);
}

static atomic_t vmap_lazy_nr = ATOMIC_INIT(0);

655 656 657 658 659
/*
 * Serialize vmap purging.  There is no actual criticial section protected
 * by this look, but we want to avoid concurrent calls for performance
 * reasons and to make the pcpu_get_vm_areas more deterministic.
 */
660
static DEFINE_MUTEX(vmap_purge_lock);
661

662 663 664
/* for per-CPU blocks */
static void purge_fragmented_blocks_allcpus(void);

665 666 667 668 669 670 671 672 673
/*
 * called before a call to iounmap() if the caller wants vm_area_struct's
 * immediately freed.
 */
void set_iounmap_nonlazy(void)
{
	atomic_set(&vmap_lazy_nr, lazy_max_pages()+1);
}

N
Nick Piggin 已提交
674 675 676
/*
 * Purges all lazily-freed vmap areas.
 */
677
static bool __purge_vmap_area_lazy(unsigned long start, unsigned long end)
N
Nick Piggin 已提交
678
{
679
	struct llist_node *valist;
N
Nick Piggin 已提交
680
	struct vmap_area *va;
681
	struct vmap_area *n_va;
682
	bool do_free = false;
N
Nick Piggin 已提交
683

684
	lockdep_assert_held(&vmap_purge_lock);
685

686 687
	valist = llist_del_all(&vmap_purge_list);
	llist_for_each_entry(va, valist, purge_list) {
688 689 690 691
		if (va->va_start < start)
			start = va->va_start;
		if (va->va_end > end)
			end = va->va_end;
692
		do_free = true;
N
Nick Piggin 已提交
693 694
	}

695
	if (!do_free)
696
		return false;
N
Nick Piggin 已提交
697

698
	flush_tlb_kernel_range(start, end);
N
Nick Piggin 已提交
699

700
	spin_lock(&vmap_area_lock);
701 702 703
	llist_for_each_entry_safe(va, n_va, valist, purge_list) {
		int nr = (va->va_end - va->va_start) >> PAGE_SHIFT;

704
		__free_vmap_area(va);
705 706 707
		atomic_sub(nr, &vmap_lazy_nr);
		cond_resched_lock(&vmap_area_lock);
	}
708 709
	spin_unlock(&vmap_area_lock);
	return true;
N
Nick Piggin 已提交
710 711
}

N
Nick Piggin 已提交
712 713 714 715 716 717
/*
 * Kick off a purge of the outstanding lazy areas. Don't bother if somebody
 * is already purging.
 */
static void try_purge_vmap_area_lazy(void)
{
718
	if (mutex_trylock(&vmap_purge_lock)) {
719
		__purge_vmap_area_lazy(ULONG_MAX, 0);
720
		mutex_unlock(&vmap_purge_lock);
721
	}
N
Nick Piggin 已提交
722 723
}

N
Nick Piggin 已提交
724 725 726 727 728
/*
 * Kick off a purge of the outstanding lazy areas.
 */
static void purge_vmap_area_lazy(void)
{
729
	mutex_lock(&vmap_purge_lock);
730 731
	purge_fragmented_blocks_allcpus();
	__purge_vmap_area_lazy(ULONG_MAX, 0);
732
	mutex_unlock(&vmap_purge_lock);
N
Nick Piggin 已提交
733 734 735
}

/*
736 737 738
 * Free a vmap area, caller ensuring that the area has been unmapped
 * and flush_cache_vunmap had been called for the correct range
 * previously.
N
Nick Piggin 已提交
739
 */
740
static void free_vmap_area_noflush(struct vmap_area *va)
N
Nick Piggin 已提交
741
{
742 743 744 745 746 747 748 749 750
	int nr_lazy;

	nr_lazy = atomic_add_return((va->va_end - va->va_start) >> PAGE_SHIFT,
				    &vmap_lazy_nr);

	/* After this point, we may free va at any time */
	llist_add(&va->purge_list, &vmap_purge_list);

	if (unlikely(nr_lazy > lazy_max_pages()))
N
Nick Piggin 已提交
751
		try_purge_vmap_area_lazy();
N
Nick Piggin 已提交
752 753
}

754 755 756 757 758 759
/*
 * Free and unmap a vmap area
 */
static void free_unmap_vmap_area(struct vmap_area *va)
{
	flush_cache_vunmap(va->va_start, va->va_end);
760 761
	unmap_vmap_area(va);
	free_vmap_area_noflush(va);
762 763
}

N
Nick Piggin 已提交
764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797
static struct vmap_area *find_vmap_area(unsigned long addr)
{
	struct vmap_area *va;

	spin_lock(&vmap_area_lock);
	va = __find_vmap_area(addr);
	spin_unlock(&vmap_area_lock);

	return va;
}

/*** Per cpu kva allocator ***/

/*
 * vmap space is limited especially on 32 bit architectures. Ensure there is
 * room for at least 16 percpu vmap blocks per CPU.
 */
/*
 * If we had a constant VMALLOC_START and VMALLOC_END, we'd like to be able
 * to #define VMALLOC_SPACE		(VMALLOC_END-VMALLOC_START). Guess
 * instead (we just need a rough idea)
 */
#if BITS_PER_LONG == 32
#define VMALLOC_SPACE		(128UL*1024*1024)
#else
#define VMALLOC_SPACE		(128UL*1024*1024*1024)
#endif

#define VMALLOC_PAGES		(VMALLOC_SPACE / PAGE_SIZE)
#define VMAP_MAX_ALLOC		BITS_PER_LONG	/* 256K with 4K pages */
#define VMAP_BBMAP_BITS_MAX	1024	/* 4MB with 4K pages */
#define VMAP_BBMAP_BITS_MIN	(VMAP_MAX_ALLOC*2)
#define VMAP_MIN(x, y)		((x) < (y) ? (x) : (y)) /* can't use min() */
#define VMAP_MAX(x, y)		((x) > (y) ? (x) : (y)) /* can't use max() */
798 799 800 801
#define VMAP_BBMAP_BITS		\
		VMAP_MIN(VMAP_BBMAP_BITS_MAX,	\
		VMAP_MAX(VMAP_BBMAP_BITS_MIN,	\
			VMALLOC_PAGES / roundup_pow_of_two(NR_CPUS) / 16))
N
Nick Piggin 已提交
802 803 804

#define VMAP_BLOCK_SIZE		(VMAP_BBMAP_BITS * PAGE_SIZE)

805 806
static bool vmap_initialized __read_mostly = false;

N
Nick Piggin 已提交
807 808 809 810 811 812 813 814 815
struct vmap_block_queue {
	spinlock_t lock;
	struct list_head free;
};

struct vmap_block {
	spinlock_t lock;
	struct vmap_area *va;
	unsigned long free, dirty;
816
	unsigned long dirty_min, dirty_max; /*< dirty range */
817 818
	struct list_head free_list;
	struct rcu_head rcu_head;
819
	struct list_head purge;
N
Nick Piggin 已提交
820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846
};

/* Queue of free and dirty vmap blocks, for allocation and flushing purposes */
static DEFINE_PER_CPU(struct vmap_block_queue, vmap_block_queue);

/*
 * Radix tree of vmap blocks, indexed by address, to quickly find a vmap block
 * in the free path. Could get rid of this if we change the API to return a
 * "cookie" from alloc, to be passed to free. But no big deal yet.
 */
static DEFINE_SPINLOCK(vmap_block_tree_lock);
static RADIX_TREE(vmap_block_tree, GFP_ATOMIC);

/*
 * We should probably have a fallback mechanism to allocate virtual memory
 * out of partially filled vmap blocks. However vmap block sizing should be
 * fairly reasonable according to the vmalloc size, so it shouldn't be a
 * big problem.
 */

static unsigned long addr_to_vb_idx(unsigned long addr)
{
	addr -= VMALLOC_START & ~(VMAP_BLOCK_SIZE-1);
	addr /= VMAP_BLOCK_SIZE;
	return addr;
}

847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864
static void *vmap_block_vaddr(unsigned long va_start, unsigned long pages_off)
{
	unsigned long addr;

	addr = va_start + (pages_off << PAGE_SHIFT);
	BUG_ON(addr_to_vb_idx(addr) != addr_to_vb_idx(va_start));
	return (void *)addr;
}

/**
 * new_vmap_block - allocates new vmap_block and occupies 2^order pages in this
 *                  block. Of course pages number can't exceed VMAP_BBMAP_BITS
 * @order:    how many 2^order pages should be occupied in newly allocated block
 * @gfp_mask: flags for the page level allocator
 *
 * Returns: virtual address in a newly allocated block or ERR_PTR(-errno)
 */
static void *new_vmap_block(unsigned int order, gfp_t gfp_mask)
N
Nick Piggin 已提交
865 866 867 868 869 870
{
	struct vmap_block_queue *vbq;
	struct vmap_block *vb;
	struct vmap_area *va;
	unsigned long vb_idx;
	int node, err;
871
	void *vaddr;
N
Nick Piggin 已提交
872 873 874 875 876 877 878 879 880 881 882

	node = numa_node_id();

	vb = kmalloc_node(sizeof(struct vmap_block),
			gfp_mask & GFP_RECLAIM_MASK, node);
	if (unlikely(!vb))
		return ERR_PTR(-ENOMEM);

	va = alloc_vmap_area(VMAP_BLOCK_SIZE, VMAP_BLOCK_SIZE,
					VMALLOC_START, VMALLOC_END,
					node, gfp_mask);
883
	if (IS_ERR(va)) {
N
Nick Piggin 已提交
884
		kfree(vb);
J
Julia Lawall 已提交
885
		return ERR_CAST(va);
N
Nick Piggin 已提交
886 887 888 889 890 891 892 893 894
	}

	err = radix_tree_preload(gfp_mask);
	if (unlikely(err)) {
		kfree(vb);
		free_vmap_area(va);
		return ERR_PTR(err);
	}

895
	vaddr = vmap_block_vaddr(va->va_start, 0);
N
Nick Piggin 已提交
896 897
	spin_lock_init(&vb->lock);
	vb->va = va;
898 899 900
	/* At least something should be left free */
	BUG_ON(VMAP_BBMAP_BITS <= (1UL << order));
	vb->free = VMAP_BBMAP_BITS - (1UL << order);
N
Nick Piggin 已提交
901
	vb->dirty = 0;
902 903
	vb->dirty_min = VMAP_BBMAP_BITS;
	vb->dirty_max = 0;
N
Nick Piggin 已提交
904 905 906 907 908 909 910 911 912 913 914
	INIT_LIST_HEAD(&vb->free_list);

	vb_idx = addr_to_vb_idx(va->va_start);
	spin_lock(&vmap_block_tree_lock);
	err = radix_tree_insert(&vmap_block_tree, vb_idx, vb);
	spin_unlock(&vmap_block_tree_lock);
	BUG_ON(err);
	radix_tree_preload_end();

	vbq = &get_cpu_var(vmap_block_queue);
	spin_lock(&vbq->lock);
915
	list_add_tail_rcu(&vb->free_list, &vbq->free);
N
Nick Piggin 已提交
916
	spin_unlock(&vbq->lock);
917
	put_cpu_var(vmap_block_queue);
N
Nick Piggin 已提交
918

919
	return vaddr;
N
Nick Piggin 已提交
920 921 922 923 924 925 926 927 928 929 930 931 932
}

static void free_vmap_block(struct vmap_block *vb)
{
	struct vmap_block *tmp;
	unsigned long vb_idx;

	vb_idx = addr_to_vb_idx(vb->va->va_start);
	spin_lock(&vmap_block_tree_lock);
	tmp = radix_tree_delete(&vmap_block_tree, vb_idx);
	spin_unlock(&vmap_block_tree_lock);
	BUG_ON(tmp != vb);

933
	free_vmap_area_noflush(vb->va);
934
	kfree_rcu(vb, rcu_head);
N
Nick Piggin 已提交
935 936
}

937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953
static void purge_fragmented_blocks(int cpu)
{
	LIST_HEAD(purge);
	struct vmap_block *vb;
	struct vmap_block *n_vb;
	struct vmap_block_queue *vbq = &per_cpu(vmap_block_queue, cpu);

	rcu_read_lock();
	list_for_each_entry_rcu(vb, &vbq->free, free_list) {

		if (!(vb->free + vb->dirty == VMAP_BBMAP_BITS && vb->dirty != VMAP_BBMAP_BITS))
			continue;

		spin_lock(&vb->lock);
		if (vb->free + vb->dirty == VMAP_BBMAP_BITS && vb->dirty != VMAP_BBMAP_BITS) {
			vb->free = 0; /* prevent further allocs after releasing lock */
			vb->dirty = VMAP_BBMAP_BITS; /* prevent purging it again */
954 955
			vb->dirty_min = 0;
			vb->dirty_max = VMAP_BBMAP_BITS;
956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979
			spin_lock(&vbq->lock);
			list_del_rcu(&vb->free_list);
			spin_unlock(&vbq->lock);
			spin_unlock(&vb->lock);
			list_add_tail(&vb->purge, &purge);
		} else
			spin_unlock(&vb->lock);
	}
	rcu_read_unlock();

	list_for_each_entry_safe(vb, n_vb, &purge, purge) {
		list_del(&vb->purge);
		free_vmap_block(vb);
	}
}

static void purge_fragmented_blocks_allcpus(void)
{
	int cpu;

	for_each_possible_cpu(cpu)
		purge_fragmented_blocks(cpu);
}

N
Nick Piggin 已提交
980 981 982 983
static void *vb_alloc(unsigned long size, gfp_t gfp_mask)
{
	struct vmap_block_queue *vbq;
	struct vmap_block *vb;
984
	void *vaddr = NULL;
N
Nick Piggin 已提交
985 986
	unsigned int order;

987
	BUG_ON(offset_in_page(size));
N
Nick Piggin 已提交
988
	BUG_ON(size > PAGE_SIZE*VMAP_MAX_ALLOC);
J
Jan Kara 已提交
989 990 991 992 993 994 995 996
	if (WARN_ON(size == 0)) {
		/*
		 * Allocating 0 bytes isn't what caller wants since
		 * get_order(0) returns funny result. Just warn and terminate
		 * early.
		 */
		return NULL;
	}
N
Nick Piggin 已提交
997 998 999 1000 1001
	order = get_order(size);

	rcu_read_lock();
	vbq = &get_cpu_var(vmap_block_queue);
	list_for_each_entry_rcu(vb, &vbq->free, free_list) {
1002
		unsigned long pages_off;
N
Nick Piggin 已提交
1003 1004

		spin_lock(&vb->lock);
1005 1006 1007 1008
		if (vb->free < (1UL << order)) {
			spin_unlock(&vb->lock);
			continue;
		}
1009

1010 1011
		pages_off = VMAP_BBMAP_BITS - vb->free;
		vaddr = vmap_block_vaddr(vb->va->va_start, pages_off);
1012 1013 1014 1015 1016 1017
		vb->free -= 1UL << order;
		if (vb->free == 0) {
			spin_lock(&vbq->lock);
			list_del_rcu(&vb->free_list);
			spin_unlock(&vbq->lock);
		}
1018

1019 1020
		spin_unlock(&vb->lock);
		break;
N
Nick Piggin 已提交
1021
	}
1022

1023
	put_cpu_var(vmap_block_queue);
N
Nick Piggin 已提交
1024 1025
	rcu_read_unlock();

1026 1027 1028
	/* Allocate new block if nothing was found */
	if (!vaddr)
		vaddr = new_vmap_block(order, gfp_mask);
N
Nick Piggin 已提交
1029

1030
	return vaddr;
N
Nick Piggin 已提交
1031 1032 1033 1034 1035 1036 1037 1038 1039
}

static void vb_free(const void *addr, unsigned long size)
{
	unsigned long offset;
	unsigned long vb_idx;
	unsigned int order;
	struct vmap_block *vb;

1040
	BUG_ON(offset_in_page(size));
N
Nick Piggin 已提交
1041
	BUG_ON(size > PAGE_SIZE*VMAP_MAX_ALLOC);
1042 1043 1044

	flush_cache_vunmap((unsigned long)addr, (unsigned long)addr + size);

N
Nick Piggin 已提交
1045 1046 1047
	order = get_order(size);

	offset = (unsigned long)addr & (VMAP_BLOCK_SIZE - 1);
1048
	offset >>= PAGE_SHIFT;
N
Nick Piggin 已提交
1049 1050 1051 1052 1053 1054 1055

	vb_idx = addr_to_vb_idx((unsigned long)addr);
	rcu_read_lock();
	vb = radix_tree_lookup(&vmap_block_tree, vb_idx);
	rcu_read_unlock();
	BUG_ON(!vb);

1056 1057
	vunmap_page_range((unsigned long)addr, (unsigned long)addr + size);

N
Nick Piggin 已提交
1058
	spin_lock(&vb->lock);
1059 1060 1061 1062

	/* Expand dirty range */
	vb->dirty_min = min(vb->dirty_min, offset);
	vb->dirty_max = max(vb->dirty_max, offset + (1UL << order));
1063

N
Nick Piggin 已提交
1064 1065
	vb->dirty += 1UL << order;
	if (vb->dirty == VMAP_BBMAP_BITS) {
1066
		BUG_ON(vb->free);
N
Nick Piggin 已提交
1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091
		spin_unlock(&vb->lock);
		free_vmap_block(vb);
	} else
		spin_unlock(&vb->lock);
}

/**
 * vm_unmap_aliases - unmap outstanding lazy aliases in the vmap layer
 *
 * The vmap/vmalloc layer lazily flushes kernel virtual mappings primarily
 * to amortize TLB flushing overheads. What this means is that any page you
 * have now, may, in a former life, have been mapped into kernel virtual
 * address by the vmap layer and so there might be some CPUs with TLB entries
 * still referencing that page (additional to the regular 1:1 kernel mapping).
 *
 * vm_unmap_aliases flushes all such lazy mappings. After it returns, we can
 * be sure that none of the pages we have control over will have any aliases
 * from the vmap layer.
 */
void vm_unmap_aliases(void)
{
	unsigned long start = ULONG_MAX, end = 0;
	int cpu;
	int flush = 0;

1092 1093 1094
	if (unlikely(!vmap_initialized))
		return;

1095 1096
	might_sleep();

N
Nick Piggin 已提交
1097 1098 1099 1100 1101 1102 1103
	for_each_possible_cpu(cpu) {
		struct vmap_block_queue *vbq = &per_cpu(vmap_block_queue, cpu);
		struct vmap_block *vb;

		rcu_read_lock();
		list_for_each_entry_rcu(vb, &vbq->free, free_list) {
			spin_lock(&vb->lock);
1104 1105
			if (vb->dirty) {
				unsigned long va_start = vb->va->va_start;
N
Nick Piggin 已提交
1106
				unsigned long s, e;
1107

1108 1109
				s = va_start + (vb->dirty_min << PAGE_SHIFT);
				e = va_start + (vb->dirty_max << PAGE_SHIFT);
N
Nick Piggin 已提交
1110

1111 1112
				start = min(s, start);
				end   = max(e, end);
N
Nick Piggin 已提交
1113

1114
				flush = 1;
N
Nick Piggin 已提交
1115 1116 1117 1118 1119 1120
			}
			spin_unlock(&vb->lock);
		}
		rcu_read_unlock();
	}

1121
	mutex_lock(&vmap_purge_lock);
1122 1123 1124
	purge_fragmented_blocks_allcpus();
	if (!__purge_vmap_area_lazy(start, end) && flush)
		flush_tlb_kernel_range(start, end);
1125
	mutex_unlock(&vmap_purge_lock);
N
Nick Piggin 已提交
1126 1127 1128 1129 1130 1131 1132 1133 1134 1135
}
EXPORT_SYMBOL_GPL(vm_unmap_aliases);

/**
 * vm_unmap_ram - unmap linear kernel address space set up by vm_map_ram
 * @mem: the pointer returned by vm_map_ram
 * @count: the count passed to that vm_map_ram call (cannot unmap partial)
 */
void vm_unmap_ram(const void *mem, unsigned int count)
{
1136
	unsigned long size = (unsigned long)count << PAGE_SHIFT;
N
Nick Piggin 已提交
1137
	unsigned long addr = (unsigned long)mem;
1138
	struct vmap_area *va;
N
Nick Piggin 已提交
1139

1140
	might_sleep();
N
Nick Piggin 已提交
1141 1142 1143
	BUG_ON(!addr);
	BUG_ON(addr < VMALLOC_START);
	BUG_ON(addr > VMALLOC_END);
1144
	BUG_ON(!PAGE_ALIGNED(addr));
N
Nick Piggin 已提交
1145 1146

	debug_check_no_locks_freed(mem, size);
1147
	vmap_debug_free_range(addr, addr+size);
N
Nick Piggin 已提交
1148

1149
	if (likely(count <= VMAP_MAX_ALLOC)) {
N
Nick Piggin 已提交
1150
		vb_free(mem, size);
1151 1152 1153 1154 1155 1156
		return;
	}

	va = find_vmap_area(addr);
	BUG_ON(!va);
	free_unmap_vmap_area(va);
N
Nick Piggin 已提交
1157 1158 1159 1160 1161 1162 1163 1164 1165
}
EXPORT_SYMBOL(vm_unmap_ram);

/**
 * vm_map_ram - map pages linearly into kernel virtual address (vmalloc space)
 * @pages: an array of pointers to the pages to be mapped
 * @count: number of pages
 * @node: prefer to allocate data structures on this node
 * @prot: memory protection to use. PAGE_KERNEL for regular RAM
1166
 *
1167 1168 1169 1170 1171 1172
 * If you use this function for less than VMAP_MAX_ALLOC pages, it could be
 * faster than vmap so it's good.  But if you mix long-life and short-life
 * objects with vm_map_ram(), it could consume lots of address space through
 * fragmentation (especially on a 32bit machine).  You could see failures in
 * the end.  Please use this function for short-lived objects.
 *
1173
 * Returns: a pointer to the address that has been mapped, or %NULL on failure
N
Nick Piggin 已提交
1174 1175 1176
 */
void *vm_map_ram(struct page **pages, unsigned int count, int node, pgprot_t prot)
{
1177
	unsigned long size = (unsigned long)count << PAGE_SHIFT;
N
Nick Piggin 已提交
1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203
	unsigned long addr;
	void *mem;

	if (likely(count <= VMAP_MAX_ALLOC)) {
		mem = vb_alloc(size, GFP_KERNEL);
		if (IS_ERR(mem))
			return NULL;
		addr = (unsigned long)mem;
	} else {
		struct vmap_area *va;
		va = alloc_vmap_area(size, PAGE_SIZE,
				VMALLOC_START, VMALLOC_END, node, GFP_KERNEL);
		if (IS_ERR(va))
			return NULL;

		addr = va->va_start;
		mem = (void *)addr;
	}
	if (vmap_page_range(addr, addr + size, prot, pages) < 0) {
		vm_unmap_ram(mem, count);
		return NULL;
	}
	return mem;
}
EXPORT_SYMBOL(vm_map_ram);

1204
static struct vm_struct *vmlist __initdata;
N
Nicolas Pitre 已提交
1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230
/**
 * vm_area_add_early - add vmap area early during boot
 * @vm: vm_struct to add
 *
 * This function is used to add fixed kernel vm area to vmlist before
 * vmalloc_init() is called.  @vm->addr, @vm->size, and @vm->flags
 * should contain proper values and the other fields should be zero.
 *
 * DO NOT USE THIS FUNCTION UNLESS YOU KNOW WHAT YOU'RE DOING.
 */
void __init vm_area_add_early(struct vm_struct *vm)
{
	struct vm_struct *tmp, **p;

	BUG_ON(vmap_initialized);
	for (p = &vmlist; (tmp = *p) != NULL; p = &tmp->next) {
		if (tmp->addr >= vm->addr) {
			BUG_ON(tmp->addr < vm->addr + vm->size);
			break;
		} else
			BUG_ON(tmp->addr + tmp->size > vm->addr);
	}
	vm->next = *p;
	*p = vm;
}

1231 1232 1233
/**
 * vm_area_register_early - register vmap area early during boot
 * @vm: vm_struct to register
1234
 * @align: requested alignment
1235 1236 1237 1238 1239 1240 1241 1242
 *
 * This function is used to register kernel vm area before
 * vmalloc_init() is called.  @vm->size and @vm->flags should contain
 * proper values on entry and other fields should be zero.  On return,
 * vm->addr contains the allocated address.
 *
 * DO NOT USE THIS FUNCTION UNLESS YOU KNOW WHAT YOU'RE DOING.
 */
1243
void __init vm_area_register_early(struct vm_struct *vm, size_t align)
1244 1245
{
	static size_t vm_init_off __initdata;
1246 1247 1248 1249
	unsigned long addr;

	addr = ALIGN(VMALLOC_START + vm_init_off, align);
	vm_init_off = PFN_ALIGN(addr + vm->size) - VMALLOC_START;
1250

1251
	vm->addr = (void *)addr;
1252

N
Nicolas Pitre 已提交
1253
	vm_area_add_early(vm);
1254 1255
}

N
Nick Piggin 已提交
1256 1257
void __init vmalloc_init(void)
{
I
Ivan Kokshaysky 已提交
1258 1259
	struct vmap_area *va;
	struct vm_struct *tmp;
N
Nick Piggin 已提交
1260 1261 1262 1263
	int i;

	for_each_possible_cpu(i) {
		struct vmap_block_queue *vbq;
1264
		struct vfree_deferred *p;
N
Nick Piggin 已提交
1265 1266 1267 1268

		vbq = &per_cpu(vmap_block_queue, i);
		spin_lock_init(&vbq->lock);
		INIT_LIST_HEAD(&vbq->free);
1269 1270 1271
		p = &per_cpu(vfree_deferred, i);
		init_llist_head(&p->list);
		INIT_WORK(&p->wq, free_work);
N
Nick Piggin 已提交
1272
	}
1273

I
Ivan Kokshaysky 已提交
1274 1275
	/* Import existing vmlist entries. */
	for (tmp = vmlist; tmp; tmp = tmp->next) {
1276
		va = kzalloc(sizeof(struct vmap_area), GFP_NOWAIT);
1277
		va->flags = VM_VM_AREA;
I
Ivan Kokshaysky 已提交
1278 1279
		va->va_start = (unsigned long)tmp->addr;
		va->va_end = va->va_start + tmp->size;
1280
		va->vm = tmp;
I
Ivan Kokshaysky 已提交
1281 1282
		__insert_vmap_area(va);
	}
1283 1284 1285

	vmap_area_pcpu_hole = VMALLOC_END;

1286
	vmap_initialized = true;
N
Nick Piggin 已提交
1287 1288
}

1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331
/**
 * map_kernel_range_noflush - map kernel VM area with the specified pages
 * @addr: start of the VM area to map
 * @size: size of the VM area to map
 * @prot: page protection flags to use
 * @pages: pages to map
 *
 * Map PFN_UP(@size) pages at @addr.  The VM area @addr and @size
 * specify should have been allocated using get_vm_area() and its
 * friends.
 *
 * NOTE:
 * This function does NOT do any cache flushing.  The caller is
 * responsible for calling flush_cache_vmap() on to-be-mapped areas
 * before calling this function.
 *
 * RETURNS:
 * The number of pages mapped on success, -errno on failure.
 */
int map_kernel_range_noflush(unsigned long addr, unsigned long size,
			     pgprot_t prot, struct page **pages)
{
	return vmap_page_range_noflush(addr, addr + size, prot, pages);
}

/**
 * unmap_kernel_range_noflush - unmap kernel VM area
 * @addr: start of the VM area to unmap
 * @size: size of the VM area to unmap
 *
 * Unmap PFN_UP(@size) pages at @addr.  The VM area @addr and @size
 * specify should have been allocated using get_vm_area() and its
 * friends.
 *
 * NOTE:
 * This function does NOT do any cache flushing.  The caller is
 * responsible for calling flush_cache_vunmap() on to-be-mapped areas
 * before calling this function and flush_tlb_kernel_range() after.
 */
void unmap_kernel_range_noflush(unsigned long addr, unsigned long size)
{
	vunmap_page_range(addr, addr + size);
}
1332
EXPORT_SYMBOL_GPL(unmap_kernel_range_noflush);
1333 1334 1335 1336 1337 1338 1339 1340 1341

/**
 * unmap_kernel_range - unmap kernel VM area and flush cache and TLB
 * @addr: start of the VM area to unmap
 * @size: size of the VM area to unmap
 *
 * Similar to unmap_kernel_range_noflush() but flushes vcache before
 * the unmapping and tlb after.
 */
N
Nick Piggin 已提交
1342 1343 1344
void unmap_kernel_range(unsigned long addr, unsigned long size)
{
	unsigned long end = addr + size;
1345 1346

	flush_cache_vunmap(addr, end);
N
Nick Piggin 已提交
1347 1348 1349
	vunmap_page_range(addr, end);
	flush_tlb_kernel_range(addr, end);
}
1350
EXPORT_SYMBOL_GPL(unmap_kernel_range);
N
Nick Piggin 已提交
1351

1352
int map_vm_area(struct vm_struct *area, pgprot_t prot, struct page **pages)
N
Nick Piggin 已提交
1353 1354
{
	unsigned long addr = (unsigned long)area->addr;
1355
	unsigned long end = addr + get_vm_area_size(area);
N
Nick Piggin 已提交
1356 1357
	int err;

1358
	err = vmap_page_range(addr, end, prot, pages);
N
Nick Piggin 已提交
1359

1360
	return err > 0 ? 0 : err;
N
Nick Piggin 已提交
1361 1362 1363
}
EXPORT_SYMBOL_GPL(map_vm_area);

1364
static void setup_vmalloc_vm(struct vm_struct *vm, struct vmap_area *va,
1365
			      unsigned long flags, const void *caller)
1366
{
1367
	spin_lock(&vmap_area_lock);
1368 1369 1370 1371
	vm->flags = flags;
	vm->addr = (void *)va->va_start;
	vm->size = va->va_end - va->va_start;
	vm->caller = caller;
1372
	va->vm = vm;
1373
	va->flags |= VM_VM_AREA;
1374
	spin_unlock(&vmap_area_lock);
1375
}
1376

1377
static void clear_vm_uninitialized_flag(struct vm_struct *vm)
1378
{
1379
	/*
1380
	 * Before removing VM_UNINITIALIZED,
1381 1382 1383 1384
	 * we should make sure that vm has proper values.
	 * Pair with smp_rmb() in show_numa_info().
	 */
	smp_wmb();
1385
	vm->flags &= ~VM_UNINITIALIZED;
1386 1387
}

N
Nick Piggin 已提交
1388
static struct vm_struct *__get_vm_area_node(unsigned long size,
1389
		unsigned long align, unsigned long flags, unsigned long start,
1390
		unsigned long end, int node, gfp_t gfp_mask, const void *caller)
N
Nick Piggin 已提交
1391
{
1392
	struct vmap_area *va;
N
Nick Piggin 已提交
1393
	struct vm_struct *area;
L
Linus Torvalds 已提交
1394

1395
	BUG_ON(in_interrupt());
L
Linus Torvalds 已提交
1396
	size = PAGE_ALIGN(size);
1397 1398
	if (unlikely(!size))
		return NULL;
L
Linus Torvalds 已提交
1399

1400 1401 1402 1403
	if (flags & VM_IOREMAP)
		align = 1ul << clamp_t(int, get_count_order_long(size),
				       PAGE_SHIFT, IOREMAP_MAX_ORDER);

1404
	area = kzalloc_node(sizeof(*area), gfp_mask & GFP_RECLAIM_MASK, node);
L
Linus Torvalds 已提交
1405 1406 1407
	if (unlikely(!area))
		return NULL;

1408 1409
	if (!(flags & VM_NO_GUARD))
		size += PAGE_SIZE;
L
Linus Torvalds 已提交
1410

N
Nick Piggin 已提交
1411 1412 1413 1414
	va = alloc_vmap_area(size, align, start, end, node, gfp_mask);
	if (IS_ERR(va)) {
		kfree(area);
		return NULL;
L
Linus Torvalds 已提交
1415 1416
	}

1417
	setup_vmalloc_vm(area, va, flags, caller);
1418

L
Linus Torvalds 已提交
1419 1420 1421
	return area;
}

C
Christoph Lameter 已提交
1422 1423 1424
struct vm_struct *__get_vm_area(unsigned long size, unsigned long flags,
				unsigned long start, unsigned long end)
{
D
David Rientjes 已提交
1425 1426
	return __get_vm_area_node(size, 1, flags, start, end, NUMA_NO_NODE,
				  GFP_KERNEL, __builtin_return_address(0));
C
Christoph Lameter 已提交
1427
}
1428
EXPORT_SYMBOL_GPL(__get_vm_area);
C
Christoph Lameter 已提交
1429

1430 1431
struct vm_struct *__get_vm_area_caller(unsigned long size, unsigned long flags,
				       unsigned long start, unsigned long end,
1432
				       const void *caller)
1433
{
D
David Rientjes 已提交
1434 1435
	return __get_vm_area_node(size, 1, flags, start, end, NUMA_NO_NODE,
				  GFP_KERNEL, caller);
1436 1437
}

L
Linus Torvalds 已提交
1438
/**
S
Simon Arlott 已提交
1439
 *	get_vm_area  -  reserve a contiguous kernel virtual area
L
Linus Torvalds 已提交
1440 1441 1442 1443 1444 1445 1446 1447 1448
 *	@size:		size of the area
 *	@flags:		%VM_IOREMAP for I/O mappings or VM_ALLOC
 *
 *	Search an area of @size in the kernel virtual mapping area,
 *	and reserved it for out purposes.  Returns the area descriptor
 *	on success or %NULL on failure.
 */
struct vm_struct *get_vm_area(unsigned long size, unsigned long flags)
{
1449
	return __get_vm_area_node(size, 1, flags, VMALLOC_START, VMALLOC_END,
D
David Rientjes 已提交
1450 1451
				  NUMA_NO_NODE, GFP_KERNEL,
				  __builtin_return_address(0));
1452 1453 1454
}

struct vm_struct *get_vm_area_caller(unsigned long size, unsigned long flags,
1455
				const void *caller)
1456
{
1457
	return __get_vm_area_node(size, 1, flags, VMALLOC_START, VMALLOC_END,
D
David Rientjes 已提交
1458
				  NUMA_NO_NODE, GFP_KERNEL, caller);
L
Linus Torvalds 已提交
1459 1460
}

1461 1462 1463 1464 1465 1466 1467 1468 1469
/**
 *	find_vm_area  -  find a continuous kernel virtual area
 *	@addr:		base address
 *
 *	Search for the kernel VM area starting at @addr, and return it.
 *	It is up to the caller to do all required locking to keep the returned
 *	pointer valid.
 */
struct vm_struct *find_vm_area(const void *addr)
1470
{
N
Nick Piggin 已提交
1471
	struct vmap_area *va;
1472

N
Nick Piggin 已提交
1473 1474
	va = find_vmap_area((unsigned long)addr);
	if (va && va->flags & VM_VM_AREA)
1475
		return va->vm;
L
Linus Torvalds 已提交
1476 1477 1478 1479

	return NULL;
}

1480
/**
S
Simon Arlott 已提交
1481
 *	remove_vm_area  -  find and remove a continuous kernel virtual area
1482 1483 1484 1485 1486 1487
 *	@addr:		base address
 *
 *	Search for the kernel VM area starting at @addr, and remove it.
 *	This function returns the found VM area, but using it is NOT safe
 *	on SMP machines, except for its size or flags.
 */
1488
struct vm_struct *remove_vm_area(const void *addr)
1489
{
N
Nick Piggin 已提交
1490 1491
	struct vmap_area *va;

1492 1493
	might_sleep();

N
Nick Piggin 已提交
1494 1495
	va = find_vmap_area((unsigned long)addr);
	if (va && va->flags & VM_VM_AREA) {
1496
		struct vm_struct *vm = va->vm;
1497

1498 1499 1500
		spin_lock(&vmap_area_lock);
		va->vm = NULL;
		va->flags &= ~VM_VM_AREA;
1501
		va->flags |= VM_LAZY_FREE;
1502 1503
		spin_unlock(&vmap_area_lock);

1504
		vmap_debug_free_range(va->va_start, va->va_end);
1505
		kasan_free_shadow(vm);
1506 1507
		free_unmap_vmap_area(va);

N
Nick Piggin 已提交
1508 1509 1510
		return vm;
	}
	return NULL;
1511 1512
}

1513
static void __vunmap(const void *addr, int deallocate_pages)
L
Linus Torvalds 已提交
1514 1515 1516 1517 1518 1519
{
	struct vm_struct *area;

	if (!addr)
		return;

1520
	if (WARN(!PAGE_ALIGNED(addr), "Trying to vfree() bad address (%p)\n",
D
Dan Carpenter 已提交
1521
			addr))
L
Linus Torvalds 已提交
1522 1523 1524 1525
		return;

	area = remove_vm_area(addr);
	if (unlikely(!area)) {
A
Arjan van de Ven 已提交
1526
		WARN(1, KERN_ERR "Trying to vfree() nonexistent vm area (%p)\n",
L
Linus Torvalds 已提交
1527 1528 1529 1530
				addr);
		return;
	}

1531 1532
	debug_check_no_locks_freed(addr, get_vm_area_size(area));
	debug_check_no_obj_freed(addr, get_vm_area_size(area));
1533

L
Linus Torvalds 已提交
1534 1535 1536 1537
	if (deallocate_pages) {
		int i;

		for (i = 0; i < area->nr_pages; i++) {
1538 1539 1540
			struct page *page = area->pages[i];

			BUG_ON(!page);
1541
			__free_pages(page, 0);
L
Linus Torvalds 已提交
1542 1543
		}

D
David Rientjes 已提交
1544
		kvfree(area->pages);
L
Linus Torvalds 已提交
1545 1546 1547 1548 1549
	}

	kfree(area);
	return;
}
A
Andrey Ryabinin 已提交
1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582

static inline void __vfree_deferred(const void *addr)
{
	/*
	 * Use raw_cpu_ptr() because this can be called from preemptible
	 * context. Preemption is absolutely fine here, because the llist_add()
	 * implementation is lockless, so it works even if we are adding to
	 * nother cpu's list.  schedule_work() should be fine with this too.
	 */
	struct vfree_deferred *p = raw_cpu_ptr(&vfree_deferred);

	if (llist_add((struct llist_node *)addr, &p->list))
		schedule_work(&p->wq);
}

/**
 *	vfree_atomic  -  release memory allocated by vmalloc()
 *	@addr:		memory base address
 *
 *	This one is just like vfree() but can be called in any atomic context
 *	except NMIs.
 */
void vfree_atomic(const void *addr)
{
	BUG_ON(in_nmi());

	kmemleak_free(addr);

	if (!addr)
		return;
	__vfree_deferred(addr);
}

L
Linus Torvalds 已提交
1583 1584 1585 1586
/**
 *	vfree  -  release memory allocated by vmalloc()
 *	@addr:		memory base address
 *
S
Simon Arlott 已提交
1587
 *	Free the virtually continuous memory area starting at @addr, as
1588 1589
 *	obtained from vmalloc(), vmalloc_32() or __vmalloc(). If @addr is
 *	NULL, no operation is performed.
L
Linus Torvalds 已提交
1590
 *
1591 1592 1593
 *	Must not be called in NMI context (strictly speaking, only if we don't
 *	have CONFIG_ARCH_HAVE_NMI_SAFE_CMPXCHG, but making the calling
 *	conventions for vfree() arch-depenedent would be a really bad idea)
A
Andrew Morton 已提交
1594
 *
1595
 *	NOTE: assumes that the object at @addr has a size >= sizeof(llist_node)
L
Linus Torvalds 已提交
1596
 */
1597
void vfree(const void *addr)
L
Linus Torvalds 已提交
1598
{
1599
	BUG_ON(in_nmi());
1600 1601 1602

	kmemleak_free(addr);

1603 1604
	if (!addr)
		return;
A
Andrey Ryabinin 已提交
1605 1606 1607
	if (unlikely(in_interrupt()))
		__vfree_deferred(addr);
	else
1608
		__vunmap(addr, 1);
L
Linus Torvalds 已提交
1609 1610 1611 1612 1613 1614 1615 1616 1617 1618
}
EXPORT_SYMBOL(vfree);

/**
 *	vunmap  -  release virtual mapping obtained by vmap()
 *	@addr:		memory base address
 *
 *	Free the virtually contiguous memory area starting at @addr,
 *	which was created from the page array passed to vmap().
 *
1619
 *	Must not be called in interrupt context.
L
Linus Torvalds 已提交
1620
 */
1621
void vunmap(const void *addr)
L
Linus Torvalds 已提交
1622 1623
{
	BUG_ON(in_interrupt());
1624
	might_sleep();
1625 1626
	if (addr)
		__vunmap(addr, 0);
L
Linus Torvalds 已提交
1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643
}
EXPORT_SYMBOL(vunmap);

/**
 *	vmap  -  map an array of pages into virtually contiguous space
 *	@pages:		array of page pointers
 *	@count:		number of pages to map
 *	@flags:		vm_area->flags
 *	@prot:		page protection for the mapping
 *
 *	Maps @count pages from @pages into contiguous kernel virtual
 *	space.
 */
void *vmap(struct page **pages, unsigned int count,
		unsigned long flags, pgprot_t prot)
{
	struct vm_struct *area;
1644
	unsigned long size;		/* In bytes */
L
Linus Torvalds 已提交
1645

1646 1647
	might_sleep();

1648
	if (count > totalram_pages)
L
Linus Torvalds 已提交
1649 1650
		return NULL;

1651 1652
	size = (unsigned long)count << PAGE_SHIFT;
	area = get_vm_area_caller(size, flags, __builtin_return_address(0));
L
Linus Torvalds 已提交
1653 1654
	if (!area)
		return NULL;
1655

1656
	if (map_vm_area(area, prot, pages)) {
L
Linus Torvalds 已提交
1657 1658 1659 1660 1661 1662 1663 1664
		vunmap(area->addr);
		return NULL;
	}

	return area->addr;
}
EXPORT_SYMBOL(vmap);

1665 1666 1667
static void *__vmalloc_node(unsigned long size, unsigned long align,
			    gfp_t gfp_mask, pgprot_t prot,
			    int node, const void *caller);
A
Adrian Bunk 已提交
1668
static void *__vmalloc_area_node(struct vm_struct *area, gfp_t gfp_mask,
1669
				 pgprot_t prot, int node)
L
Linus Torvalds 已提交
1670 1671 1672
{
	struct page **pages;
	unsigned int nr_pages, array_size, i;
1673
	const gfp_t nested_gfp = (gfp_mask & GFP_RECLAIM_MASK) | __GFP_ZERO;
1674
	const gfp_t alloc_mask = gfp_mask | __GFP_HIGHMEM | __GFP_NOWARN;
L
Linus Torvalds 已提交
1675

1676
	nr_pages = get_vm_area_size(area) >> PAGE_SHIFT;
L
Linus Torvalds 已提交
1677 1678 1679 1680
	array_size = (nr_pages * sizeof(struct page *));

	area->nr_pages = nr_pages;
	/* Please note that the recursion is strictly bounded. */
1681
	if (array_size > PAGE_SIZE) {
1682
		pages = __vmalloc_node(array_size, 1, nested_gfp|__GFP_HIGHMEM,
1683
				PAGE_KERNEL, node, area->caller);
1684
	} else {
1685
		pages = kmalloc_node(array_size, nested_gfp, node);
1686
	}
L
Linus Torvalds 已提交
1687 1688 1689 1690 1691 1692 1693 1694
	area->pages = pages;
	if (!area->pages) {
		remove_vm_area(area->addr);
		kfree(area);
		return NULL;
	}

	for (i = 0; i < area->nr_pages; i++) {
1695 1696
		struct page *page;

1697 1698
		if (fatal_signal_pending(current)) {
			area->nr_pages = i;
1699
			goto fail_no_warn;
1700 1701
		}

J
Jianguo Wu 已提交
1702
		if (node == NUMA_NO_NODE)
1703
			page = alloc_page(alloc_mask);
C
Christoph Lameter 已提交
1704
		else
1705
			page = alloc_pages_node(node, alloc_mask, 0);
1706 1707

		if (unlikely(!page)) {
L
Linus Torvalds 已提交
1708 1709 1710 1711
			/* Successfully allocated i pages, free them in __vunmap() */
			area->nr_pages = i;
			goto fail;
		}
1712
		area->pages[i] = page;
1713
		if (gfpflags_allow_blocking(gfp_mask))
1714
			cond_resched();
L
Linus Torvalds 已提交
1715 1716
	}

1717
	if (map_vm_area(area, prot, pages))
L
Linus Torvalds 已提交
1718 1719 1720 1721
		goto fail;
	return area->addr;

fail:
1722
	warn_alloc(gfp_mask, NULL,
1723
			  "vmalloc: allocation failure, allocated %ld of %ld bytes",
1724
			  (area->nr_pages*PAGE_SIZE), area->size);
1725
fail_no_warn:
L
Linus Torvalds 已提交
1726 1727 1728 1729 1730
	vfree(area->addr);
	return NULL;
}

/**
1731
 *	__vmalloc_node_range  -  allocate virtually contiguous memory
L
Linus Torvalds 已提交
1732
 *	@size:		allocation size
1733
 *	@align:		desired alignment
1734 1735
 *	@start:		vm area range start
 *	@end:		vm area range end
L
Linus Torvalds 已提交
1736 1737
 *	@gfp_mask:	flags for the page level allocator
 *	@prot:		protection mask for the allocated pages
1738
 *	@vm_flags:	additional vm area flags (e.g. %VM_NO_GUARD)
D
David Rientjes 已提交
1739
 *	@node:		node to use for allocation or NUMA_NO_NODE
1740
 *	@caller:	caller's return address
L
Linus Torvalds 已提交
1741 1742 1743 1744 1745
 *
 *	Allocate enough pages to cover @size from the page level
 *	allocator with @gfp_mask flags.  Map them into contiguous
 *	kernel virtual space, using a pagetable protection of @prot.
 */
1746 1747
void *__vmalloc_node_range(unsigned long size, unsigned long align,
			unsigned long start, unsigned long end, gfp_t gfp_mask,
1748 1749
			pgprot_t prot, unsigned long vm_flags, int node,
			const void *caller)
L
Linus Torvalds 已提交
1750 1751
{
	struct vm_struct *area;
1752 1753
	void *addr;
	unsigned long real_size = size;
L
Linus Torvalds 已提交
1754 1755

	size = PAGE_ALIGN(size);
1756
	if (!size || (size >> PAGE_SHIFT) > totalram_pages)
1757
		goto fail;
L
Linus Torvalds 已提交
1758

1759 1760
	area = __get_vm_area_node(size, align, VM_ALLOC | VM_UNINITIALIZED |
				vm_flags, start, end, node, gfp_mask, caller);
L
Linus Torvalds 已提交
1761
	if (!area)
1762
		goto fail;
L
Linus Torvalds 已提交
1763

1764
	addr = __vmalloc_area_node(area, gfp_mask, prot, node);
1765
	if (!addr)
1766
		return NULL;
1767

1768
	/*
1769 1770
	 * In this function, newly allocated vm_struct has VM_UNINITIALIZED
	 * flag. It means that vm_struct is not fully initialized.
1771
	 * Now, it is fully initialized, so remove this flag here.
1772
	 */
1773
	clear_vm_uninitialized_flag(area);
1774

1775
	kmemleak_vmalloc(area, size, gfp_mask);
1776 1777

	return addr;
1778 1779

fail:
1780
	warn_alloc(gfp_mask, NULL,
1781
			  "vmalloc: allocation failure: %lu bytes", real_size);
1782
	return NULL;
L
Linus Torvalds 已提交
1783 1784
}

1785 1786 1787 1788 1789 1790
/**
 *	__vmalloc_node  -  allocate virtually contiguous memory
 *	@size:		allocation size
 *	@align:		desired alignment
 *	@gfp_mask:	flags for the page level allocator
 *	@prot:		protection mask for the allocated pages
D
David Rientjes 已提交
1791
 *	@node:		node to use for allocation or NUMA_NO_NODE
1792 1793 1794 1795 1796
 *	@caller:	caller's return address
 *
 *	Allocate enough pages to cover @size from the page level
 *	allocator with @gfp_mask flags.  Map them into contiguous
 *	kernel virtual space, using a pagetable protection of @prot.
M
Michal Hocko 已提交
1797
 *
1798
 *	Reclaim modifiers in @gfp_mask - __GFP_NORETRY, __GFP_RETRY_MAYFAIL
M
Michal Hocko 已提交
1799 1800 1801 1802 1803
 *	and __GFP_NOFAIL are not supported
 *
 *	Any use of gfp flags outside of GFP_KERNEL should be consulted
 *	with mm people.
 *
1804
 */
1805
static void *__vmalloc_node(unsigned long size, unsigned long align,
1806
			    gfp_t gfp_mask, pgprot_t prot,
1807
			    int node, const void *caller)
1808 1809
{
	return __vmalloc_node_range(size, align, VMALLOC_START, VMALLOC_END,
1810
				gfp_mask, prot, 0, node, caller);
1811 1812
}

C
Christoph Lameter 已提交
1813 1814
void *__vmalloc(unsigned long size, gfp_t gfp_mask, pgprot_t prot)
{
D
David Rientjes 已提交
1815
	return __vmalloc_node(size, 1, gfp_mask, prot, NUMA_NO_NODE,
1816
				__builtin_return_address(0));
C
Christoph Lameter 已提交
1817
}
L
Linus Torvalds 已提交
1818 1819
EXPORT_SYMBOL(__vmalloc);

1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833
static inline void *__vmalloc_node_flags(unsigned long size,
					int node, gfp_t flags)
{
	return __vmalloc_node(size, 1, flags, PAGE_KERNEL,
					node, __builtin_return_address(0));
}


void *__vmalloc_node_flags_caller(unsigned long size, int node, gfp_t flags,
				  void *caller)
{
	return __vmalloc_node(size, 1, flags, PAGE_KERNEL, node, caller);
}

L
Linus Torvalds 已提交
1834 1835 1836 1837 1838 1839
/**
 *	vmalloc  -  allocate virtually contiguous memory
 *	@size:		allocation size
 *	Allocate enough pages to cover @size from the page level
 *	allocator and map them into contiguous kernel virtual space.
 *
1840
 *	For tight control over page level allocator and protection flags
L
Linus Torvalds 已提交
1841 1842 1843 1844
 *	use __vmalloc() instead.
 */
void *vmalloc(unsigned long size)
{
D
David Rientjes 已提交
1845
	return __vmalloc_node_flags(size, NUMA_NO_NODE,
1846
				    GFP_KERNEL);
L
Linus Torvalds 已提交
1847 1848 1849
}
EXPORT_SYMBOL(vmalloc);

1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861
/**
 *	vzalloc - allocate virtually contiguous memory with zero fill
 *	@size:	allocation size
 *	Allocate enough pages to cover @size from the page level
 *	allocator and map them into contiguous kernel virtual space.
 *	The memory allocated is set to zero.
 *
 *	For tight control over page level allocator and protection flags
 *	use __vmalloc() instead.
 */
void *vzalloc(unsigned long size)
{
D
David Rientjes 已提交
1862
	return __vmalloc_node_flags(size, NUMA_NO_NODE,
1863
				GFP_KERNEL | __GFP_ZERO);
1864 1865 1866
}
EXPORT_SYMBOL(vzalloc);

1867
/**
1868 1869
 * vmalloc_user - allocate zeroed virtually contiguous memory for userspace
 * @size: allocation size
1870
 *
1871 1872
 * The resulting memory area is zeroed so it can be mapped to userspace
 * without leaking data.
1873 1874 1875 1876 1877 1878
 */
void *vmalloc_user(unsigned long size)
{
	struct vm_struct *area;
	void *ret;

1879
	ret = __vmalloc_node(size, SHMLBA,
1880
			     GFP_KERNEL | __GFP_ZERO,
D
David Rientjes 已提交
1881 1882
			     PAGE_KERNEL, NUMA_NO_NODE,
			     __builtin_return_address(0));
1883
	if (ret) {
N
Nick Piggin 已提交
1884
		area = find_vm_area(ret);
1885 1886
		area->flags |= VM_USERMAP;
	}
1887 1888 1889 1890
	return ret;
}
EXPORT_SYMBOL(vmalloc_user);

C
Christoph Lameter 已提交
1891 1892 1893
/**
 *	vmalloc_node  -  allocate memory on a specific node
 *	@size:		allocation size
1894
 *	@node:		numa node
C
Christoph Lameter 已提交
1895 1896 1897 1898
 *
 *	Allocate enough pages to cover @size from the page level
 *	allocator and map them into contiguous kernel virtual space.
 *
1899
 *	For tight control over page level allocator and protection flags
C
Christoph Lameter 已提交
1900 1901 1902 1903
 *	use __vmalloc() instead.
 */
void *vmalloc_node(unsigned long size, int node)
{
1904
	return __vmalloc_node(size, 1, GFP_KERNEL, PAGE_KERNEL,
1905
					node, __builtin_return_address(0));
C
Christoph Lameter 已提交
1906 1907 1908
}
EXPORT_SYMBOL(vmalloc_node);

1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923
/**
 * vzalloc_node - allocate memory on a specific node with zero fill
 * @size:	allocation size
 * @node:	numa node
 *
 * Allocate enough pages to cover @size from the page level
 * allocator and map them into contiguous kernel virtual space.
 * The memory allocated is set to zero.
 *
 * For tight control over page level allocator and protection flags
 * use __vmalloc_node() instead.
 */
void *vzalloc_node(unsigned long size, int node)
{
	return __vmalloc_node_flags(size, node,
1924
			 GFP_KERNEL | __GFP_ZERO);
1925 1926 1927
}
EXPORT_SYMBOL(vzalloc_node);

1928 1929 1930 1931
#ifndef PAGE_KERNEL_EXEC
# define PAGE_KERNEL_EXEC PAGE_KERNEL
#endif

L
Linus Torvalds 已提交
1932 1933 1934 1935 1936 1937 1938 1939
/**
 *	vmalloc_exec  -  allocate virtually contiguous, executable memory
 *	@size:		allocation size
 *
 *	Kernel-internal function to allocate enough pages to cover @size
 *	the page level allocator and map them into contiguous and
 *	executable kernel virtual space.
 *
1940
 *	For tight control over page level allocator and protection flags
L
Linus Torvalds 已提交
1941 1942 1943 1944 1945
 *	use __vmalloc() instead.
 */

void *vmalloc_exec(unsigned long size)
{
1946
	return __vmalloc_node(size, 1, GFP_KERNEL, PAGE_KERNEL_EXEC,
D
David Rientjes 已提交
1947
			      NUMA_NO_NODE, __builtin_return_address(0));
L
Linus Torvalds 已提交
1948 1949
}

1950
#if defined(CONFIG_64BIT) && defined(CONFIG_ZONE_DMA32)
1951
#define GFP_VMALLOC32 GFP_DMA32 | GFP_KERNEL
1952
#elif defined(CONFIG_64BIT) && defined(CONFIG_ZONE_DMA)
1953
#define GFP_VMALLOC32 GFP_DMA | GFP_KERNEL
1954 1955 1956 1957
#else
#define GFP_VMALLOC32 GFP_KERNEL
#endif

L
Linus Torvalds 已提交
1958 1959 1960 1961 1962 1963 1964 1965 1966
/**
 *	vmalloc_32  -  allocate virtually contiguous memory (32bit addressable)
 *	@size:		allocation size
 *
 *	Allocate enough 32bit PA addressable pages to cover @size from the
 *	page level allocator and map them into contiguous kernel virtual space.
 */
void *vmalloc_32(unsigned long size)
{
1967
	return __vmalloc_node(size, 1, GFP_VMALLOC32, PAGE_KERNEL,
D
David Rientjes 已提交
1968
			      NUMA_NO_NODE, __builtin_return_address(0));
L
Linus Torvalds 已提交
1969 1970 1971
}
EXPORT_SYMBOL(vmalloc_32);

1972
/**
1973
 * vmalloc_32_user - allocate zeroed virtually contiguous 32bit memory
1974
 *	@size:		allocation size
1975 1976 1977
 *
 * The resulting memory area is 32bit addressable and zeroed so it can be
 * mapped to userspace without leaking data.
1978 1979 1980 1981 1982 1983
 */
void *vmalloc_32_user(unsigned long size)
{
	struct vm_struct *area;
	void *ret;

1984
	ret = __vmalloc_node(size, 1, GFP_VMALLOC32 | __GFP_ZERO, PAGE_KERNEL,
D
David Rientjes 已提交
1985
			     NUMA_NO_NODE, __builtin_return_address(0));
1986
	if (ret) {
N
Nick Piggin 已提交
1987
		area = find_vm_area(ret);
1988 1989
		area->flags |= VM_USERMAP;
	}
1990 1991 1992 1993
	return ret;
}
EXPORT_SYMBOL(vmalloc_32_user);

1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
/*
 * small helper routine , copy contents to buf from addr.
 * If the page is not present, fill zero.
 */

static int aligned_vread(char *buf, char *addr, unsigned long count)
{
	struct page *p;
	int copied = 0;

	while (count) {
		unsigned long offset, length;

2007
		offset = offset_in_page(addr);
2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023
		length = PAGE_SIZE - offset;
		if (length > count)
			length = count;
		p = vmalloc_to_page(addr);
		/*
		 * To do safe access to this _mapped_ area, we need
		 * lock. But adding lock here means that we need to add
		 * overhead of vmalloc()/vfree() calles for this _debug_
		 * interface, rarely used. Instead of that, we'll use
		 * kmap() and get small overhead in this access function.
		 */
		if (p) {
			/*
			 * we can expect USER0 is not used (see vread/vwrite's
			 * function description)
			 */
2024
			void *map = kmap_atomic(p);
2025
			memcpy(buf, map + offset, length);
2026
			kunmap_atomic(map);
2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045
		} else
			memset(buf, 0, length);

		addr += length;
		buf += length;
		copied += length;
		count -= length;
	}
	return copied;
}

static int aligned_vwrite(char *buf, char *addr, unsigned long count)
{
	struct page *p;
	int copied = 0;

	while (count) {
		unsigned long offset, length;

2046
		offset = offset_in_page(addr);
2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062
		length = PAGE_SIZE - offset;
		if (length > count)
			length = count;
		p = vmalloc_to_page(addr);
		/*
		 * To do safe access to this _mapped_ area, we need
		 * lock. But adding lock here means that we need to add
		 * overhead of vmalloc()/vfree() calles for this _debug_
		 * interface, rarely used. Instead of that, we'll use
		 * kmap() and get small overhead in this access function.
		 */
		if (p) {
			/*
			 * we can expect USER0 is not used (see vread/vwrite's
			 * function description)
			 */
2063
			void *map = kmap_atomic(p);
2064
			memcpy(map + offset, buf, length);
2065
			kunmap_atomic(map);
2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091
		}
		addr += length;
		buf += length;
		copied += length;
		count -= length;
	}
	return copied;
}

/**
 *	vread() -  read vmalloc area in a safe way.
 *	@buf:		buffer for reading data
 *	@addr:		vm address.
 *	@count:		number of bytes to be read.
 *
 *	Returns # of bytes which addr and buf should be increased.
 *	(same number to @count). Returns 0 if [addr...addr+count) doesn't
 *	includes any intersect with alive vmalloc area.
 *
 *	This function checks that addr is a valid vmalloc'ed area, and
 *	copy data from that area to a given buffer. If the given memory range
 *	of [addr...addr+count) includes some valid address, data is copied to
 *	proper area of @buf. If there are memory holes, they'll be zero-filled.
 *	IOREMAP area is treated as memory hole and no copy is done.
 *
 *	If [addr...addr+count) doesn't includes any intersects with alive
2092
 *	vm_struct area, returns 0. @buf should be kernel's buffer.
2093 2094 2095 2096 2097 2098 2099 2100
 *
 *	Note: In usual ops, vread() is never necessary because the caller
 *	should know vmalloc() area is valid and can use memcpy().
 *	This is for routines which have to access vmalloc area without
 *	any informaion, as /dev/kmem.
 *
 */

L
Linus Torvalds 已提交
2101 2102
long vread(char *buf, char *addr, unsigned long count)
{
2103 2104
	struct vmap_area *va;
	struct vm_struct *vm;
L
Linus Torvalds 已提交
2105
	char *vaddr, *buf_start = buf;
2106
	unsigned long buflen = count;
L
Linus Torvalds 已提交
2107 2108 2109 2110 2111 2112
	unsigned long n;

	/* Don't allow overflow */
	if ((unsigned long) addr + count < count)
		count = -(unsigned long) addr;

2113 2114 2115 2116 2117 2118 2119 2120 2121 2122
	spin_lock(&vmap_area_lock);
	list_for_each_entry(va, &vmap_area_list, list) {
		if (!count)
			break;

		if (!(va->flags & VM_VM_AREA))
			continue;

		vm = va->vm;
		vaddr = (char *) vm->addr;
2123
		if (addr >= vaddr + get_vm_area_size(vm))
L
Linus Torvalds 已提交
2124 2125 2126 2127 2128 2129 2130 2131 2132
			continue;
		while (addr < vaddr) {
			if (count == 0)
				goto finished;
			*buf = '\0';
			buf++;
			addr++;
			count--;
		}
2133
		n = vaddr + get_vm_area_size(vm) - addr;
2134 2135
		if (n > count)
			n = count;
2136
		if (!(vm->flags & VM_IOREMAP))
2137 2138 2139 2140 2141 2142
			aligned_vread(buf, addr, n);
		else /* IOREMAP area is treated as memory hole */
			memset(buf, 0, n);
		buf += n;
		addr += n;
		count -= n;
L
Linus Torvalds 已提交
2143 2144
	}
finished:
2145
	spin_unlock(&vmap_area_lock);
2146 2147 2148 2149 2150 2151 2152 2153

	if (buf == buf_start)
		return 0;
	/* zero-fill memory holes */
	if (buf != buf_start + buflen)
		memset(buf, 0, buflen - (buf - buf_start));

	return buflen;
L
Linus Torvalds 已提交
2154 2155
}

2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173
/**
 *	vwrite() -  write vmalloc area in a safe way.
 *	@buf:		buffer for source data
 *	@addr:		vm address.
 *	@count:		number of bytes to be read.
 *
 *	Returns # of bytes which addr and buf should be incresed.
 *	(same number to @count).
 *	If [addr...addr+count) doesn't includes any intersect with valid
 *	vmalloc area, returns 0.
 *
 *	This function checks that addr is a valid vmalloc'ed area, and
 *	copy data from a buffer to the given addr. If specified range of
 *	[addr...addr+count) includes some valid address, data is copied from
 *	proper area of @buf. If there are memory holes, no copy to hole.
 *	IOREMAP area is treated as memory hole and no copy is done.
 *
 *	If [addr...addr+count) doesn't includes any intersects with alive
2174
 *	vm_struct area, returns 0. @buf should be kernel's buffer.
2175 2176 2177 2178 2179 2180 2181
 *
 *	Note: In usual ops, vwrite() is never necessary because the caller
 *	should know vmalloc() area is valid and can use memcpy().
 *	This is for routines which have to access vmalloc area without
 *	any informaion, as /dev/kmem.
 */

L
Linus Torvalds 已提交
2182 2183
long vwrite(char *buf, char *addr, unsigned long count)
{
2184 2185
	struct vmap_area *va;
	struct vm_struct *vm;
2186 2187 2188
	char *vaddr;
	unsigned long n, buflen;
	int copied = 0;
L
Linus Torvalds 已提交
2189 2190 2191 2192

	/* Don't allow overflow */
	if ((unsigned long) addr + count < count)
		count = -(unsigned long) addr;
2193
	buflen = count;
L
Linus Torvalds 已提交
2194

2195 2196 2197 2198 2199 2200 2201 2202 2203 2204
	spin_lock(&vmap_area_lock);
	list_for_each_entry(va, &vmap_area_list, list) {
		if (!count)
			break;

		if (!(va->flags & VM_VM_AREA))
			continue;

		vm = va->vm;
		vaddr = (char *) vm->addr;
2205
		if (addr >= vaddr + get_vm_area_size(vm))
L
Linus Torvalds 已提交
2206 2207 2208 2209 2210 2211 2212 2213
			continue;
		while (addr < vaddr) {
			if (count == 0)
				goto finished;
			buf++;
			addr++;
			count--;
		}
2214
		n = vaddr + get_vm_area_size(vm) - addr;
2215 2216
		if (n > count)
			n = count;
2217
		if (!(vm->flags & VM_IOREMAP)) {
2218 2219 2220 2221 2222 2223
			aligned_vwrite(buf, addr, n);
			copied++;
		}
		buf += n;
		addr += n;
		count -= n;
L
Linus Torvalds 已提交
2224 2225
	}
finished:
2226
	spin_unlock(&vmap_area_lock);
2227 2228 2229
	if (!copied)
		return 0;
	return buflen;
L
Linus Torvalds 已提交
2230
}
2231 2232

/**
2233 2234 2235 2236 2237
 *	remap_vmalloc_range_partial  -  map vmalloc pages to userspace
 *	@vma:		vma to cover
 *	@uaddr:		target user address to start at
 *	@kaddr:		virtual address of vmalloc kernel memory
 *	@size:		size of map area
2238 2239
 *
 *	Returns:	0 for success, -Exxx on failure
2240
 *
2241 2242 2243 2244
 *	This function checks that @kaddr is a valid vmalloc'ed area,
 *	and that it is big enough to cover the range starting at
 *	@uaddr in @vma. Will return failure if that criteria isn't
 *	met.
2245
 *
2246
 *	Similar to remap_pfn_range() (see mm/memory.c)
2247
 */
2248 2249
int remap_vmalloc_range_partial(struct vm_area_struct *vma, unsigned long uaddr,
				void *kaddr, unsigned long size)
2250 2251 2252
{
	struct vm_struct *area;

2253 2254 2255
	size = PAGE_ALIGN(size);

	if (!PAGE_ALIGNED(uaddr) || !PAGE_ALIGNED(kaddr))
2256 2257
		return -EINVAL;

2258
	area = find_vm_area(kaddr);
2259
	if (!area)
N
Nick Piggin 已提交
2260
		return -EINVAL;
2261 2262

	if (!(area->flags & VM_USERMAP))
N
Nick Piggin 已提交
2263
		return -EINVAL;
2264

2265
	if (kaddr + size > area->addr + area->size)
N
Nick Piggin 已提交
2266
		return -EINVAL;
2267 2268

	do {
2269
		struct page *page = vmalloc_to_page(kaddr);
N
Nick Piggin 已提交
2270 2271
		int ret;

2272 2273 2274 2275 2276
		ret = vm_insert_page(vma, uaddr, page);
		if (ret)
			return ret;

		uaddr += PAGE_SIZE;
2277 2278 2279
		kaddr += PAGE_SIZE;
		size -= PAGE_SIZE;
	} while (size > 0);
2280

2281
	vma->vm_flags |= VM_DONTEXPAND | VM_DONTDUMP;
2282

N
Nick Piggin 已提交
2283
	return 0;
2284
}
2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307
EXPORT_SYMBOL(remap_vmalloc_range_partial);

/**
 *	remap_vmalloc_range  -  map vmalloc pages to userspace
 *	@vma:		vma to cover (map full range of vma)
 *	@addr:		vmalloc memory
 *	@pgoff:		number of pages into addr before first page to map
 *
 *	Returns:	0 for success, -Exxx on failure
 *
 *	This function checks that addr is a valid vmalloc'ed area, and
 *	that it is big enough to cover the vma. Will return failure if
 *	that criteria isn't met.
 *
 *	Similar to remap_pfn_range() (see mm/memory.c)
 */
int remap_vmalloc_range(struct vm_area_struct *vma, void *addr,
						unsigned long pgoff)
{
	return remap_vmalloc_range_partial(vma, vma->vm_start,
					   addr + (pgoff << PAGE_SHIFT),
					   vma->vm_end - vma->vm_start);
}
2308 2309
EXPORT_SYMBOL(remap_vmalloc_range);

2310 2311 2312 2313
/*
 * Implement a stub for vmalloc_sync_all() if the architecture chose not to
 * have one.
 */
2314
void __weak vmalloc_sync_all(void)
2315 2316
{
}
2317 2318


2319
static int f(pte_t *pte, pgtable_t table, unsigned long addr, void *data)
2320
{
2321 2322 2323 2324 2325 2326
	pte_t ***p = data;

	if (p) {
		*(*p) = pte;
		(*p)++;
	}
2327 2328 2329 2330 2331 2332
	return 0;
}

/**
 *	alloc_vm_area - allocate a range of kernel address space
 *	@size:		size of the area
2333
 *	@ptes:		returns the PTEs for the address space
2334 2335
 *
 *	Returns:	NULL on failure, vm_struct on success
2336 2337 2338
 *
 *	This function reserves a range of kernel address space, and
 *	allocates pagetables to map that range.  No actual mappings
2339 2340 2341 2342
 *	are created.
 *
 *	If @ptes is non-NULL, pointers to the PTEs (in init_mm)
 *	allocated for the VM area are returned.
2343
 */
2344
struct vm_struct *alloc_vm_area(size_t size, pte_t **ptes)
2345 2346 2347
{
	struct vm_struct *area;

2348 2349
	area = get_vm_area_caller(size, VM_IOREMAP,
				__builtin_return_address(0));
2350 2351 2352 2353 2354 2355 2356 2357
	if (area == NULL)
		return NULL;

	/*
	 * This ensures that page tables are constructed for this region
	 * of kernel virtual address space and mapped into init_mm.
	 */
	if (apply_to_page_range(&init_mm, (unsigned long)area->addr,
2358
				size, f, ptes ? &ptes : NULL)) {
2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374
		free_vm_area(area);
		return NULL;
	}

	return area;
}
EXPORT_SYMBOL_GPL(alloc_vm_area);

void free_vm_area(struct vm_struct *area)
{
	struct vm_struct *ret;
	ret = remove_vm_area(area->addr);
	BUG_ON(ret != area);
	kfree(area);
}
EXPORT_SYMBOL_GPL(free_vm_area);
2375

2376
#ifdef CONFIG_SMP
2377 2378
static struct vmap_area *node_to_va(struct rb_node *n)
{
G
Geliang Tang 已提交
2379
	return rb_entry_safe(n, struct vmap_area, rb_node);
2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471
}

/**
 * pvm_find_next_prev - find the next and prev vmap_area surrounding @end
 * @end: target address
 * @pnext: out arg for the next vmap_area
 * @pprev: out arg for the previous vmap_area
 *
 * Returns: %true if either or both of next and prev are found,
 *	    %false if no vmap_area exists
 *
 * Find vmap_areas end addresses of which enclose @end.  ie. if not
 * NULL, *pnext->va_end > @end and *pprev->va_end <= @end.
 */
static bool pvm_find_next_prev(unsigned long end,
			       struct vmap_area **pnext,
			       struct vmap_area **pprev)
{
	struct rb_node *n = vmap_area_root.rb_node;
	struct vmap_area *va = NULL;

	while (n) {
		va = rb_entry(n, struct vmap_area, rb_node);
		if (end < va->va_end)
			n = n->rb_left;
		else if (end > va->va_end)
			n = n->rb_right;
		else
			break;
	}

	if (!va)
		return false;

	if (va->va_end > end) {
		*pnext = va;
		*pprev = node_to_va(rb_prev(&(*pnext)->rb_node));
	} else {
		*pprev = va;
		*pnext = node_to_va(rb_next(&(*pprev)->rb_node));
	}
	return true;
}

/**
 * pvm_determine_end - find the highest aligned address between two vmap_areas
 * @pnext: in/out arg for the next vmap_area
 * @pprev: in/out arg for the previous vmap_area
 * @align: alignment
 *
 * Returns: determined end address
 *
 * Find the highest aligned address between *@pnext and *@pprev below
 * VMALLOC_END.  *@pnext and *@pprev are adjusted so that the aligned
 * down address is between the end addresses of the two vmap_areas.
 *
 * Please note that the address returned by this function may fall
 * inside *@pnext vmap_area.  The caller is responsible for checking
 * that.
 */
static unsigned long pvm_determine_end(struct vmap_area **pnext,
				       struct vmap_area **pprev,
				       unsigned long align)
{
	const unsigned long vmalloc_end = VMALLOC_END & ~(align - 1);
	unsigned long addr;

	if (*pnext)
		addr = min((*pnext)->va_start & ~(align - 1), vmalloc_end);
	else
		addr = vmalloc_end;

	while (*pprev && (*pprev)->va_end > addr) {
		*pnext = *pprev;
		*pprev = node_to_va(rb_prev(&(*pnext)->rb_node));
	}

	return addr;
}

/**
 * pcpu_get_vm_areas - allocate vmalloc areas for percpu allocator
 * @offsets: array containing offset of each area
 * @sizes: array containing size of each area
 * @nr_vms: the number of areas to allocate
 * @align: alignment, all entries in @offsets and @sizes must be aligned to this
 *
 * Returns: kmalloc'd vm_struct pointer array pointing to allocated
 *	    vm_structs on success, %NULL on failure
 *
 * Percpu allocator wants to use congruent vm areas so that it can
 * maintain the offsets among percpu areas.  This function allocates
2472 2473 2474 2475
 * congruent vmalloc areas for it with GFP_KERNEL.  These areas tend to
 * be scattered pretty far, distance between two areas easily going up
 * to gigabytes.  To avoid interacting with regular vmallocs, these
 * areas are allocated from top.
2476 2477 2478 2479 2480 2481 2482 2483 2484 2485
 *
 * Despite its complicated look, this allocator is rather simple.  It
 * does everything top-down and scans areas from the end looking for
 * matching slot.  While scanning, if any of the areas overlaps with
 * existing vmap_area, the base address is pulled down to fit the
 * area.  Scanning is repeated till all the areas fit and then all
 * necessary data structres are inserted and the result is returned.
 */
struct vm_struct **pcpu_get_vm_areas(const unsigned long *offsets,
				     const size_t *sizes, int nr_vms,
2486
				     size_t align)
2487 2488 2489 2490 2491 2492 2493 2494 2495 2496
{
	const unsigned long vmalloc_start = ALIGN(VMALLOC_START, align);
	const unsigned long vmalloc_end = VMALLOC_END & ~(align - 1);
	struct vmap_area **vas, *prev, *next;
	struct vm_struct **vms;
	int area, area2, last_area, term_area;
	unsigned long base, start, end, last_end;
	bool purged = false;

	/* verify parameters and allocate data structures */
2497
	BUG_ON(offset_in_page(align) || !is_power_of_2(align));
2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527
	for (last_area = 0, area = 0; area < nr_vms; area++) {
		start = offsets[area];
		end = start + sizes[area];

		/* is everything aligned properly? */
		BUG_ON(!IS_ALIGNED(offsets[area], align));
		BUG_ON(!IS_ALIGNED(sizes[area], align));

		/* detect the area with the highest address */
		if (start > offsets[last_area])
			last_area = area;

		for (area2 = 0; area2 < nr_vms; area2++) {
			unsigned long start2 = offsets[area2];
			unsigned long end2 = start2 + sizes[area2];

			if (area2 == area)
				continue;

			BUG_ON(start2 >= start && start2 < end);
			BUG_ON(end2 <= end && end2 > start);
		}
	}
	last_end = offsets[last_area] + sizes[last_area];

	if (vmalloc_end - vmalloc_start < last_end) {
		WARN_ON(true);
		return NULL;
	}

2528 2529
	vms = kcalloc(nr_vms, sizeof(vms[0]), GFP_KERNEL);
	vas = kcalloc(nr_vms, sizeof(vas[0]), GFP_KERNEL);
2530
	if (!vas || !vms)
2531
		goto err_free2;
2532 2533

	for (area = 0; area < nr_vms; area++) {
2534 2535
		vas[area] = kzalloc(sizeof(struct vmap_area), GFP_KERNEL);
		vms[area] = kzalloc(sizeof(struct vm_struct), GFP_KERNEL);
2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620
		if (!vas[area] || !vms[area])
			goto err_free;
	}
retry:
	spin_lock(&vmap_area_lock);

	/* start scanning - we scan from the top, begin with the last area */
	area = term_area = last_area;
	start = offsets[area];
	end = start + sizes[area];

	if (!pvm_find_next_prev(vmap_area_pcpu_hole, &next, &prev)) {
		base = vmalloc_end - last_end;
		goto found;
	}
	base = pvm_determine_end(&next, &prev, align) - end;

	while (true) {
		BUG_ON(next && next->va_end <= base + end);
		BUG_ON(prev && prev->va_end > base + end);

		/*
		 * base might have underflowed, add last_end before
		 * comparing.
		 */
		if (base + last_end < vmalloc_start + last_end) {
			spin_unlock(&vmap_area_lock);
			if (!purged) {
				purge_vmap_area_lazy();
				purged = true;
				goto retry;
			}
			goto err_free;
		}

		/*
		 * If next overlaps, move base downwards so that it's
		 * right below next and then recheck.
		 */
		if (next && next->va_start < base + end) {
			base = pvm_determine_end(&next, &prev, align) - end;
			term_area = area;
			continue;
		}

		/*
		 * If prev overlaps, shift down next and prev and move
		 * base so that it's right below new next and then
		 * recheck.
		 */
		if (prev && prev->va_end > base + start)  {
			next = prev;
			prev = node_to_va(rb_prev(&next->rb_node));
			base = pvm_determine_end(&next, &prev, align) - end;
			term_area = area;
			continue;
		}

		/*
		 * This area fits, move on to the previous one.  If
		 * the previous one is the terminal one, we're done.
		 */
		area = (area + nr_vms - 1) % nr_vms;
		if (area == term_area)
			break;
		start = offsets[area];
		end = start + sizes[area];
		pvm_find_next_prev(base + end, &next, &prev);
	}
found:
	/* we've found a fitting base, insert all va's */
	for (area = 0; area < nr_vms; area++) {
		struct vmap_area *va = vas[area];

		va->va_start = base + offsets[area];
		va->va_end = va->va_start + sizes[area];
		__insert_vmap_area(va);
	}

	vmap_area_pcpu_hole = base + offsets[last_area];

	spin_unlock(&vmap_area_lock);

	/* insert all vm's */
	for (area = 0; area < nr_vms; area++)
2621 2622
		setup_vmalloc_vm(vms[area], vas[area], VM_ALLOC,
				 pcpu_get_vm_areas);
2623 2624 2625 2626 2627 2628

	kfree(vas);
	return vms;

err_free:
	for (area = 0; area < nr_vms; area++) {
2629 2630
		kfree(vas[area]);
		kfree(vms[area]);
2631
	}
2632
err_free2:
2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652
	kfree(vas);
	kfree(vms);
	return NULL;
}

/**
 * pcpu_free_vm_areas - free vmalloc areas for percpu allocator
 * @vms: vm_struct pointer array returned by pcpu_get_vm_areas()
 * @nr_vms: the number of allocated areas
 *
 * Free vm_structs and the array allocated by pcpu_get_vm_areas().
 */
void pcpu_free_vm_areas(struct vm_struct **vms, int nr_vms)
{
	int i;

	for (i = 0; i < nr_vms; i++)
		free_vm_area(vms[i]);
	kfree(vms);
}
2653
#endif	/* CONFIG_SMP */
2654 2655 2656

#ifdef CONFIG_PROC_FS
static void *s_start(struct seq_file *m, loff_t *pos)
2657
	__acquires(&vmap_area_lock)
2658
{
2659
	spin_lock(&vmap_area_lock);
2660
	return seq_list_start(&vmap_area_list, *pos);
2661 2662 2663 2664
}

static void *s_next(struct seq_file *m, void *p, loff_t *pos)
{
2665
	return seq_list_next(p, &vmap_area_list, pos);
2666 2667 2668
}

static void s_stop(struct seq_file *m, void *p)
2669
	__releases(&vmap_area_lock)
2670
{
2671
	spin_unlock(&vmap_area_lock);
2672 2673
}

E
Eric Dumazet 已提交
2674 2675
static void show_numa_info(struct seq_file *m, struct vm_struct *v)
{
2676
	if (IS_ENABLED(CONFIG_NUMA)) {
E
Eric Dumazet 已提交
2677 2678 2679 2680 2681
		unsigned int nr, *counters = m->private;

		if (!counters)
			return;

2682 2683
		if (v->flags & VM_UNINITIALIZED)
			return;
2684 2685
		/* Pair with smp_wmb() in clear_vm_uninitialized_flag() */
		smp_rmb();
2686

E
Eric Dumazet 已提交
2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697
		memset(counters, 0, nr_node_ids * sizeof(unsigned int));

		for (nr = 0; nr < v->nr_pages; nr++)
			counters[page_to_nid(v->pages[nr])]++;

		for_each_node_state(nr, N_HIGH_MEMORY)
			if (counters[nr])
				seq_printf(m, " N%u=%u", nr, counters[nr]);
	}
}

2698 2699
static int s_show(struct seq_file *m, void *p)
{
2700
	struct vmap_area *va;
2701 2702
	struct vm_struct *v;

2703 2704
	va = list_entry(p, struct vmap_area, list);

2705 2706 2707 2708
	/*
	 * s_show can encounter race with remove_vm_area, !VM_VM_AREA on
	 * behalf of vmap area is being tear down or vm_map_ram allocation.
	 */
2709 2710 2711 2712 2713 2714
	if (!(va->flags & VM_VM_AREA)) {
		seq_printf(m, "0x%pK-0x%pK %7ld %s\n",
			(void *)va->va_start, (void *)va->va_end,
			va->va_end - va->va_start,
			va->flags & VM_LAZY_FREE ? "unpurged vm_area" : "vm_map_ram");

2715
		return 0;
2716
	}
2717 2718

	v = va->vm;
2719

K
Kees Cook 已提交
2720
	seq_printf(m, "0x%pK-0x%pK %7ld",
2721 2722
		v->addr, v->addr + v->size, v->size);

J
Joe Perches 已提交
2723 2724
	if (v->caller)
		seq_printf(m, " %pS", v->caller);
2725

2726 2727 2728 2729
	if (v->nr_pages)
		seq_printf(m, " pages=%d", v->nr_pages);

	if (v->phys_addr)
2730
		seq_printf(m, " phys=%pa", &v->phys_addr);
2731 2732

	if (v->flags & VM_IOREMAP)
2733
		seq_puts(m, " ioremap");
2734 2735

	if (v->flags & VM_ALLOC)
2736
		seq_puts(m, " vmalloc");
2737 2738

	if (v->flags & VM_MAP)
2739
		seq_puts(m, " vmap");
2740 2741

	if (v->flags & VM_USERMAP)
2742
		seq_puts(m, " user");
2743

D
David Rientjes 已提交
2744
	if (is_vmalloc_addr(v->pages))
2745
		seq_puts(m, " vpages");
2746

E
Eric Dumazet 已提交
2747
	show_numa_info(m, v);
2748 2749 2750 2751
	seq_putc(m, '\n');
	return 0;
}

2752
static const struct seq_operations vmalloc_op = {
2753 2754 2755 2756 2757
	.start = s_start,
	.next = s_next,
	.stop = s_stop,
	.show = s_show,
};
2758 2759 2760

static int vmalloc_open(struct inode *inode, struct file *file)
{
2761 2762 2763 2764 2765
	if (IS_ENABLED(CONFIG_NUMA))
		return seq_open_private(file, &vmalloc_op,
					nr_node_ids * sizeof(unsigned int));
	else
		return seq_open(file, &vmalloc_op);
2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780
}

static const struct file_operations proc_vmalloc_operations = {
	.open		= vmalloc_open,
	.read		= seq_read,
	.llseek		= seq_lseek,
	.release	= seq_release_private,
};

static int __init proc_vmalloc_init(void)
{
	proc_create("vmallocinfo", S_IRUSR, NULL, &proc_vmalloc_operations);
	return 0;
}
module_init(proc_vmalloc_init);
2781

2782 2783
#endif