inode.c 34.7 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3
/*
 * hugetlbpage-backed filesystem.  Based on ramfs.
 *
4
 * Nadia Yvette Chambers, 2002
L
Linus Torvalds 已提交
5 6
 *
 * Copyright (C) 2002 Linus Torvalds.
7
 * License: GPL
L
Linus Torvalds 已提交
8 9
 */

10 11
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

L
Linus Torvalds 已提交
12 13
#include <linux/thread_info.h>
#include <asm/current.h>
14
#include <linux/sched/signal.h>		/* remove ASAP */
15
#include <linux/falloc.h>
L
Linus Torvalds 已提交
16 17 18
#include <linux/fs.h>
#include <linux/mount.h>
#include <linux/file.h>
19
#include <linux/kernel.h>
L
Linus Torvalds 已提交
20 21 22 23 24
#include <linux/writeback.h>
#include <linux/pagemap.h>
#include <linux/highmem.h>
#include <linux/init.h>
#include <linux/string.h>
25
#include <linux/capability.h>
26
#include <linux/ctype.h>
L
Linus Torvalds 已提交
27 28 29
#include <linux/backing-dev.h>
#include <linux/hugetlb.h>
#include <linux/pagevec.h>
30
#include <linux/parser.h>
31
#include <linux/mman.h>
L
Linus Torvalds 已提交
32 33 34 35
#include <linux/slab.h>
#include <linux/dnotify.h>
#include <linux/statfs.h>
#include <linux/security.h>
N
Nick Black 已提交
36
#include <linux/magic.h>
N
Naoya Horiguchi 已提交
37
#include <linux/migrate.h>
A
Al Viro 已提交
38
#include <linux/uio.h>
L
Linus Torvalds 已提交
39

40
#include <linux/uaccess.h>
L
Linus Torvalds 已提交
41

42
static const struct super_operations hugetlbfs_ops;
43
static const struct address_space_operations hugetlbfs_aops;
44
const struct file_operations hugetlbfs_file_operations;
45 46
static const struct inode_operations hugetlbfs_dir_inode_operations;
static const struct inode_operations hugetlbfs_inode_operations;
L
Linus Torvalds 已提交
47

D
David Gibson 已提交
48
struct hugetlbfs_config {
49 50
	kuid_t   uid;
	kgid_t   gid;
D
David Gibson 已提交
51
	umode_t mode;
52
	long	max_hpages;
D
David Gibson 已提交
53 54
	long	nr_inodes;
	struct hstate *hstate;
55
	long    min_hpages;
D
David Gibson 已提交
56 57 58 59 60 61 62 63 64 65 66 67
};

struct hugetlbfs_inode_info {
	struct shared_policy policy;
	struct inode vfs_inode;
};

static inline struct hugetlbfs_inode_info *HUGETLBFS_I(struct inode *inode)
{
	return container_of(inode, struct hugetlbfs_inode_info, vfs_inode);
}

L
Linus Torvalds 已提交
68 69
int sysctl_hugetlb_shm_group;

70 71 72
enum {
	Opt_size, Opt_nr_inodes,
	Opt_mode, Opt_uid, Opt_gid,
73
	Opt_pagesize, Opt_min_size,
74 75 76
	Opt_err,
};

77
static const match_table_t tokens = {
78 79 80 81 82
	{Opt_size,	"size=%s"},
	{Opt_nr_inodes,	"nr_inodes=%s"},
	{Opt_mode,	"mode=%o"},
	{Opt_uid,	"uid=%u"},
	{Opt_gid,	"gid=%u"},
83
	{Opt_pagesize,	"pagesize=%s"},
84
	{Opt_min_size,	"min_size=%s"},
85 86 87
	{Opt_err,	NULL},
};

88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110
#ifdef CONFIG_NUMA
static inline void hugetlb_set_vma_policy(struct vm_area_struct *vma,
					struct inode *inode, pgoff_t index)
{
	vma->vm_policy = mpol_shared_policy_lookup(&HUGETLBFS_I(inode)->policy,
							index);
}

static inline void hugetlb_drop_vma_policy(struct vm_area_struct *vma)
{
	mpol_cond_put(vma->vm_policy);
}
#else
static inline void hugetlb_set_vma_policy(struct vm_area_struct *vma,
					struct inode *inode, pgoff_t index)
{
}

static inline void hugetlb_drop_vma_policy(struct vm_area_struct *vma)
{
}
#endif

111 112 113 114 115 116 117 118 119 120
static void huge_pagevec_release(struct pagevec *pvec)
{
	int i;

	for (i = 0; i < pagevec_count(pvec); ++i)
		put_page(pvec->pages[i]);

	pagevec_reinit(pvec);
}

L
Linus Torvalds 已提交
121 122
static int hugetlbfs_file_mmap(struct file *file, struct vm_area_struct *vma)
{
A
Al Viro 已提交
123
	struct inode *inode = file_inode(file);
L
Linus Torvalds 已提交
124 125
	loff_t len, vma_len;
	int ret;
126
	struct hstate *h = hstate_file(file);
L
Linus Torvalds 已提交
127

128
	/*
129 130 131 132 133 134
	 * vma address alignment (but not the pgoff alignment) has
	 * already been checked by prepare_hugepage_range.  If you add
	 * any error returns here, do so after setting VM_HUGETLB, so
	 * is_vm_hugetlb_page tests below unmap_region go the right
	 * way when do_mmap_pgoff unwinds (may be important on powerpc
	 * and ia64).
135
	 */
136
	vma->vm_flags |= VM_HUGETLB | VM_DONTEXPAND;
137
	vma->vm_ops = &hugetlb_vm_ops;
L
Linus Torvalds 已提交
138

139 140 141 142 143 144 145
	/*
	 * Offset passed to mmap (before page shift) could have been
	 * negative when represented as a (l)off_t.
	 */
	if (((loff_t)vma->vm_pgoff << PAGE_SHIFT) < 0)
		return -EINVAL;

146
	if (vma->vm_pgoff & (~huge_page_mask(h) >> PAGE_SHIFT))
147 148
		return -EINVAL;

L
Linus Torvalds 已提交
149
	vma_len = (loff_t)(vma->vm_end - vma->vm_start);
150 151 152 153
	len = vma_len + ((loff_t)vma->vm_pgoff << PAGE_SHIFT);
	/* check for overflow */
	if (len < vma_len)
		return -EINVAL;
L
Linus Torvalds 已提交
154

A
Al Viro 已提交
155
	inode_lock(inode);
L
Linus Torvalds 已提交
156 157 158
	file_accessed(file);

	ret = -ENOMEM;
159
	if (hugetlb_reserve_pages(inode,
160
				vma->vm_pgoff >> huge_page_order(h),
161 162
				len >> huge_page_shift(h), vma,
				vma->vm_flags))
163
		goto out;
164

A
Adam Litke 已提交
165
	ret = 0;
166
	if (vma->vm_flags & VM_WRITE && inode->i_size < len)
167
		i_size_write(inode, len);
L
Linus Torvalds 已提交
168
out:
A
Al Viro 已提交
169
	inode_unlock(inode);
L
Linus Torvalds 已提交
170 171 172 173 174

	return ret;
}

/*
175
 * Called under down_write(mmap_sem).
L
Linus Torvalds 已提交
176 177
 */

178
#ifndef HAVE_ARCH_HUGETLB_UNMAPPED_AREA
L
Linus Torvalds 已提交
179 180 181 182 183 184
static unsigned long
hugetlb_get_unmapped_area(struct file *file, unsigned long addr,
		unsigned long len, unsigned long pgoff, unsigned long flags)
{
	struct mm_struct *mm = current->mm;
	struct vm_area_struct *vma;
185
	struct hstate *h = hstate_file(file);
186
	struct vm_unmapped_area_info info;
L
Linus Torvalds 已提交
187

188
	if (len & ~huge_page_mask(h))
L
Linus Torvalds 已提交
189 190 191 192
		return -EINVAL;
	if (len > TASK_SIZE)
		return -ENOMEM;

193
	if (flags & MAP_FIXED) {
194
		if (prepare_hugepage_range(file, addr, len))
195 196 197 198
			return -EINVAL;
		return addr;
	}

L
Linus Torvalds 已提交
199
	if (addr) {
200
		addr = ALIGN(addr, huge_page_size(h));
L
Linus Torvalds 已提交
201 202 203 204 205 206
		vma = find_vma(mm, addr);
		if (TASK_SIZE - len >= addr &&
		    (!vma || addr + len <= vma->vm_start))
			return addr;
	}

207 208 209 210 211 212 213
	info.flags = 0;
	info.length = len;
	info.low_limit = TASK_UNMAPPED_BASE;
	info.high_limit = TASK_SIZE;
	info.align_mask = PAGE_MASK & ~huge_page_mask(h);
	info.align_offset = 0;
	return vm_unmapped_area(&info);
L
Linus Torvalds 已提交
214 215 216
}
#endif

A
Al Viro 已提交
217
static size_t
B
Badari Pulavarty 已提交
218
hugetlbfs_read_actor(struct page *page, unsigned long offset,
A
Al Viro 已提交
219
			struct iov_iter *to, unsigned long size)
B
Badari Pulavarty 已提交
220
{
A
Al Viro 已提交
221
	size_t copied = 0;
B
Badari Pulavarty 已提交
222 223 224
	int i, chunksize;

	/* Find which 4k chunk and offset with in that chunk */
225 226
	i = offset >> PAGE_SHIFT;
	offset = offset & ~PAGE_MASK;
B
Badari Pulavarty 已提交
227 228

	while (size) {
A
Al Viro 已提交
229
		size_t n;
230
		chunksize = PAGE_SIZE;
B
Badari Pulavarty 已提交
231 232 233 234
		if (offset)
			chunksize -= offset;
		if (chunksize > size)
			chunksize = size;
A
Al Viro 已提交
235 236 237 238
		n = copy_page_to_iter(&page[i], offset, chunksize, to);
		copied += n;
		if (n != chunksize)
			return copied;
B
Badari Pulavarty 已提交
239 240 241 242
		offset = 0;
		size -= chunksize;
		i++;
	}
A
Al Viro 已提交
243
	return copied;
B
Badari Pulavarty 已提交
244 245 246 247 248
}

/*
 * Support for read() - Find the page attached to f_mapping and copy out the
 * data. Its *very* similar to do_generic_mapping_read(), we can't use that
249
 * since it has PAGE_SIZE assumptions.
B
Badari Pulavarty 已提交
250
 */
A
Al Viro 已提交
251
static ssize_t hugetlbfs_read_iter(struct kiocb *iocb, struct iov_iter *to)
B
Badari Pulavarty 已提交
252
{
A
Al Viro 已提交
253 254 255
	struct file *file = iocb->ki_filp;
	struct hstate *h = hstate_file(file);
	struct address_space *mapping = file->f_mapping;
B
Badari Pulavarty 已提交
256
	struct inode *inode = mapping->host;
A
Al Viro 已提交
257 258
	unsigned long index = iocb->ki_pos >> huge_page_shift(h);
	unsigned long offset = iocb->ki_pos & ~huge_page_mask(h);
B
Badari Pulavarty 已提交
259 260 261 262
	unsigned long end_index;
	loff_t isize;
	ssize_t retval = 0;

A
Al Viro 已提交
263
	while (iov_iter_count(to)) {
B
Badari Pulavarty 已提交
264
		struct page *page;
A
Al Viro 已提交
265
		size_t nr, copied;
B
Badari Pulavarty 已提交
266 267

		/* nr is the maximum number of bytes to copy from this page */
268
		nr = huge_page_size(h);
269 270
		isize = i_size_read(inode);
		if (!isize)
A
Al Viro 已提交
271
			break;
272
		end_index = (isize - 1) >> huge_page_shift(h);
A
Al Viro 已提交
273 274 275
		if (index > end_index)
			break;
		if (index == end_index) {
276
			nr = ((isize - 1) & ~huge_page_mask(h)) + 1;
277
			if (nr <= offset)
A
Al Viro 已提交
278
				break;
B
Badari Pulavarty 已提交
279 280 281 282
		}
		nr = nr - offset;

		/* Find the page */
283
		page = find_lock_page(mapping, index);
B
Badari Pulavarty 已提交
284 285 286 287 288
		if (unlikely(page == NULL)) {
			/*
			 * We have a HOLE, zero out the user-buffer for the
			 * length of the hole or request.
			 */
A
Al Viro 已提交
289
			copied = iov_iter_zero(nr, to);
B
Badari Pulavarty 已提交
290
		} else {
291 292
			unlock_page(page);

B
Badari Pulavarty 已提交
293 294 295
			/*
			 * We have the page, copy it to user space buffer.
			 */
A
Al Viro 已提交
296
			copied = hugetlbfs_read_actor(page, offset, to, nr);
297
			put_page(page);
B
Badari Pulavarty 已提交
298
		}
A
Al Viro 已提交
299 300 301 302 303 304
		offset += copied;
		retval += copied;
		if (copied != nr && iov_iter_count(to)) {
			if (!retval)
				retval = -EFAULT;
			break;
B
Badari Pulavarty 已提交
305
		}
306 307
		index += offset >> huge_page_shift(h);
		offset &= ~huge_page_mask(h);
B
Badari Pulavarty 已提交
308
	}
A
Al Viro 已提交
309
	iocb->ki_pos = ((loff_t)index << huge_page_shift(h)) + offset;
B
Badari Pulavarty 已提交
310 311 312
	return retval;
}

N
Nick Piggin 已提交
313 314 315 316
static int hugetlbfs_write_begin(struct file *file,
			struct address_space *mapping,
			loff_t pos, unsigned len, unsigned flags,
			struct page **pagep, void **fsdata)
L
Linus Torvalds 已提交
317 318 319 320
{
	return -EINVAL;
}

N
Nick Piggin 已提交
321 322 323
static int hugetlbfs_write_end(struct file *file, struct address_space *mapping,
			loff_t pos, unsigned len, unsigned copied,
			struct page *page, void *fsdata)
L
Linus Torvalds 已提交
324
{
N
Nick Piggin 已提交
325
	BUG();
L
Linus Torvalds 已提交
326 327 328
	return -EINVAL;
}

329
static void remove_huge_page(struct page *page)
L
Linus Torvalds 已提交
330
{
331
	ClearPageDirty(page);
L
Linus Torvalds 已提交
332
	ClearPageUptodate(page);
333
	delete_from_page_cache(page);
L
Linus Torvalds 已提交
334 335
}

336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372
static void
hugetlb_vmdelete_list(struct rb_root *root, pgoff_t start, pgoff_t end)
{
	struct vm_area_struct *vma;

	/*
	 * end == 0 indicates that the entire range after
	 * start should be unmapped.
	 */
	vma_interval_tree_foreach(vma, root, start, end ? end : ULONG_MAX) {
		unsigned long v_offset;
		unsigned long v_end;

		/*
		 * Can the expression below overflow on 32-bit arches?
		 * No, because the interval tree returns us only those vmas
		 * which overlap the truncated area starting at pgoff,
		 * and no vma on a 32-bit arch can span beyond the 4GB.
		 */
		if (vma->vm_pgoff < start)
			v_offset = (start - vma->vm_pgoff) << PAGE_SHIFT;
		else
			v_offset = 0;

		if (!end)
			v_end = vma->vm_end;
		else {
			v_end = ((end - vma->vm_pgoff) << PAGE_SHIFT)
							+ vma->vm_start;
			if (v_end > vma->vm_end)
				v_end = vma->vm_end;
		}

		unmap_hugepage_range(vma, vma->vm_start + v_offset, v_end,
									NULL);
	}
}
373 374 375 376

/*
 * remove_inode_hugepages handles two distinct cases: truncation and hole
 * punch.  There are subtle differences in operation for each case.
377
 *
378 379 380
 * truncation is indicated by end of range being LLONG_MAX
 *	In this case, we first scan the range and release found pages.
 *	After releasing pages, hugetlb_unreserve_pages cleans up region/reserv
381 382 383 384 385
 *	maps and global counts.  Page faults can not race with truncation
 *	in this routine.  hugetlb_no_page() prevents page faults in the
 *	truncated range.  It checks i_size before allocation, and again after
 *	with the page table lock for the page held.  The same lock must be
 *	acquired to unmap a page.
386 387 388 389
 * hole punch is indicated if end is not LLONG_MAX
 *	In the hole punch case we scan the range and release found pages.
 *	Only when releasing a page is the associated region/reserv map
 *	deleted.  The region/reserv map for ranges without associated
390 391
 *	pages are not modified.  Page faults can race with hole punch.
 *	This is indicated if we find a mapped page.
392 393 394 395 396
 * Note: If the passed end of range value is beyond the end of file, but
 * not LLONG_MAX this routine still performs a hole punch operation.
 */
static void remove_inode_hugepages(struct inode *inode, loff_t lstart,
				   loff_t lend)
L
Linus Torvalds 已提交
397
{
398
	struct hstate *h = hstate_inode(inode);
399
	struct address_space *mapping = &inode->i_data;
400
	const pgoff_t start = lstart >> huge_page_shift(h);
401 402
	const pgoff_t end = lend >> huge_page_shift(h);
	struct vm_area_struct pseudo_vma;
L
Linus Torvalds 已提交
403 404
	struct pagevec pvec;
	pgoff_t next;
405
	int i, freed = 0;
406 407
	long lookup_nr = PAGEVEC_SIZE;
	bool truncate_op = (lend == LLONG_MAX);
L
Linus Torvalds 已提交
408

409 410
	memset(&pseudo_vma, 0, sizeof(struct vm_area_struct));
	pseudo_vma.vm_flags = (VM_HUGETLB | VM_MAYSHARE | VM_SHARED);
L
Linus Torvalds 已提交
411 412
	pagevec_init(&pvec, 0);
	next = start;
413 414
	while (next < end) {
		/*
415
		 * Don't grab more pages than the number left in the range.
416 417 418 419 420
		 */
		if (end - next < lookup_nr)
			lookup_nr = end - next;

		/*
421
		 * When no more pages are found, we are done.
422
		 */
423 424
		if (!pagevec_lookup(&pvec, mapping, next, lookup_nr))
			break;
L
Linus Torvalds 已提交
425 426 427

		for (i = 0; i < pagevec_count(&pvec); ++i) {
			struct page *page = pvec.pages[i];
428 429
			u32 hash;

430 431 432 433 434 435 436 437 438
			/*
			 * The page (index) could be beyond end.  This is
			 * only possible in the punch hole case as end is
			 * max page offset in the truncate case.
			 */
			next = page->index;
			if (next >= end)
				break;

439 440 441 442
			hash = hugetlb_fault_mutex_hash(h, current->mm,
							&pseudo_vma,
							mapping, next, 0);
			mutex_lock(&hugetlb_fault_mutex_table[hash]);
L
Linus Torvalds 已提交
443

444 445 446 447 448 449 450 451 452 453
			/*
			 * If page is mapped, it was faulted in after being
			 * unmapped in caller.  Unmap (again) now after taking
			 * the fault mutex.  The mutex will prevent faults
			 * until we finish removing the page.
			 *
			 * This race can only happen in the hole punch case.
			 * Getting here in a truncate operation is a bug.
			 */
			if (unlikely(page_mapped(page))) {
454
				BUG_ON(truncate_op);
455 456 457 458 459 460 461 462 463 464 465 466 467 468

				i_mmap_lock_write(mapping);
				hugetlb_vmdelete_list(&mapping->i_mmap,
					next * pages_per_huge_page(h),
					(next + 1) * pages_per_huge_page(h));
				i_mmap_unlock_write(mapping);
			}

			lock_page(page);
			/*
			 * We must free the huge page and remove from page
			 * cache (remove_huge_page) BEFORE removing the
			 * region/reserve map (hugetlb_unreserve_pages).  In
			 * rare out of memory conditions, removal of the
469 470 471
			 * region/reserve map could fail. Correspondingly,
			 * the subpool and global reserve usage count can need
			 * to be adjusted.
472
			 */
473
			VM_BUG_ON(PagePrivate(page));
474 475 476 477 478
			remove_huge_page(page);
			freed++;
			if (!truncate_op) {
				if (unlikely(hugetlb_unreserve_pages(inode,
							next, next + 1, 1)))
479
					hugetlb_fix_reserve_counts(inode);
480 481
			}

L
Linus Torvalds 已提交
482
			unlock_page(page);
483
			mutex_unlock(&hugetlb_fault_mutex_table[hash]);
L
Linus Torvalds 已提交
484
		}
485
		++next;
L
Linus Torvalds 已提交
486
		huge_pagevec_release(&pvec);
487
		cond_resched();
L
Linus Torvalds 已提交
488
	}
489 490 491

	if (truncate_op)
		(void)hugetlb_unreserve_pages(inode, start, LONG_MAX, freed);
L
Linus Torvalds 已提交
492 493
}

A
Al Viro 已提交
494
static void hugetlbfs_evict_inode(struct inode *inode)
L
Linus Torvalds 已提交
495
{
496 497
	struct resv_map *resv_map;

498
	remove_inode_hugepages(inode, 0, LLONG_MAX);
499 500 501 502
	resv_map = (struct resv_map *)inode->i_mapping->private_data;
	/* root inode doesn't have the resv_map, so we should check it */
	if (resv_map)
		resv_map_release(&resv_map->refs);
503
	clear_inode(inode);
504 505
}

L
Linus Torvalds 已提交
506 507
static int hugetlb_vmtruncate(struct inode *inode, loff_t offset)
{
H
Hugh Dickins 已提交
508
	pgoff_t pgoff;
L
Linus Torvalds 已提交
509
	struct address_space *mapping = inode->i_mapping;
510
	struct hstate *h = hstate_inode(inode);
L
Linus Torvalds 已提交
511

512
	BUG_ON(offset & ~huge_page_mask(h));
H
Hugh Dickins 已提交
513
	pgoff = offset >> PAGE_SHIFT;
L
Linus Torvalds 已提交
514

515
	i_size_write(inode, offset);
516
	i_mmap_lock_write(mapping);
517
	if (!RB_EMPTY_ROOT(&mapping->i_mmap))
518
		hugetlb_vmdelete_list(&mapping->i_mmap, pgoff, 0);
519
	i_mmap_unlock_write(mapping);
520
	remove_inode_hugepages(inode, offset, LLONG_MAX);
L
Linus Torvalds 已提交
521 522 523
	return 0;
}

524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539
static long hugetlbfs_punch_hole(struct inode *inode, loff_t offset, loff_t len)
{
	struct hstate *h = hstate_inode(inode);
	loff_t hpage_size = huge_page_size(h);
	loff_t hole_start, hole_end;

	/*
	 * For hole punch round up the beginning offset of the hole and
	 * round down the end.
	 */
	hole_start = round_up(offset, hpage_size);
	hole_end = round_down(offset + len, hpage_size);

	if (hole_end > hole_start) {
		struct address_space *mapping = inode->i_mapping;

A
Al Viro 已提交
540
		inode_lock(inode);
541 542 543 544 545 546 547
		i_mmap_lock_write(mapping);
		if (!RB_EMPTY_ROOT(&mapping->i_mmap))
			hugetlb_vmdelete_list(&mapping->i_mmap,
						hole_start >> PAGE_SHIFT,
						hole_end  >> PAGE_SHIFT);
		i_mmap_unlock_write(mapping);
		remove_inode_hugepages(inode, hole_start, hole_end);
A
Al Viro 已提交
548
		inode_unlock(inode);
549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581
	}

	return 0;
}

static long hugetlbfs_fallocate(struct file *file, int mode, loff_t offset,
				loff_t len)
{
	struct inode *inode = file_inode(file);
	struct address_space *mapping = inode->i_mapping;
	struct hstate *h = hstate_inode(inode);
	struct vm_area_struct pseudo_vma;
	struct mm_struct *mm = current->mm;
	loff_t hpage_size = huge_page_size(h);
	unsigned long hpage_shift = huge_page_shift(h);
	pgoff_t start, index, end;
	int error;
	u32 hash;

	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
		return -EOPNOTSUPP;

	if (mode & FALLOC_FL_PUNCH_HOLE)
		return hugetlbfs_punch_hole(inode, offset, len);

	/*
	 * Default preallocate case.
	 * For this range, start is rounded down and end is rounded up
	 * as well as being converted to page offsets.
	 */
	start = offset >> hpage_shift;
	end = (offset + len + hpage_size - 1) >> hpage_shift;

A
Al Viro 已提交
582
	inode_lock(inode);
583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666

	/* We need to check rlimit even when FALLOC_FL_KEEP_SIZE */
	error = inode_newsize_ok(inode, offset + len);
	if (error)
		goto out;

	/*
	 * Initialize a pseudo vma as this is required by the huge page
	 * allocation routines.  If NUMA is configured, use page index
	 * as input to create an allocation policy.
	 */
	memset(&pseudo_vma, 0, sizeof(struct vm_area_struct));
	pseudo_vma.vm_flags = (VM_HUGETLB | VM_MAYSHARE | VM_SHARED);
	pseudo_vma.vm_file = file;

	for (index = start; index < end; index++) {
		/*
		 * This is supposed to be the vaddr where the page is being
		 * faulted in, but we have no vaddr here.
		 */
		struct page *page;
		unsigned long addr;
		int avoid_reserve = 0;

		cond_resched();

		/*
		 * fallocate(2) manpage permits EINTR; we may have been
		 * interrupted because we are using up too much memory.
		 */
		if (signal_pending(current)) {
			error = -EINTR;
			break;
		}

		/* Set numa allocation policy based on index */
		hugetlb_set_vma_policy(&pseudo_vma, inode, index);

		/* addr is the offset within the file (zero based) */
		addr = index * hpage_size;

		/* mutex taken here, fault path and hole punch */
		hash = hugetlb_fault_mutex_hash(h, mm, &pseudo_vma, mapping,
						index, addr);
		mutex_lock(&hugetlb_fault_mutex_table[hash]);

		/* See if already present in mapping to avoid alloc/free */
		page = find_get_page(mapping, index);
		if (page) {
			put_page(page);
			mutex_unlock(&hugetlb_fault_mutex_table[hash]);
			hugetlb_drop_vma_policy(&pseudo_vma);
			continue;
		}

		/* Allocate page and add to page cache */
		page = alloc_huge_page(&pseudo_vma, addr, avoid_reserve);
		hugetlb_drop_vma_policy(&pseudo_vma);
		if (IS_ERR(page)) {
			mutex_unlock(&hugetlb_fault_mutex_table[hash]);
			error = PTR_ERR(page);
			goto out;
		}
		clear_huge_page(page, addr, pages_per_huge_page(h));
		__SetPageUptodate(page);
		error = huge_add_to_page_cache(page, mapping, index);
		if (unlikely(error)) {
			put_page(page);
			mutex_unlock(&hugetlb_fault_mutex_table[hash]);
			goto out;
		}

		mutex_unlock(&hugetlb_fault_mutex_table[hash]);

		/*
		 * page_put due to reference from alloc_huge_page()
		 * unlock_page because locked by add_to_page_cache()
		 */
		put_page(page);
		unlock_page(page);
	}

	if (!(mode & FALLOC_FL_KEEP_SIZE) && offset + len > inode->i_size)
		i_size_write(inode, offset + len);
667
	inode->i_ctime = current_time(inode);
668
out:
A
Al Viro 已提交
669
	inode_unlock(inode);
670 671 672
	return error;
}

L
Linus Torvalds 已提交
673 674
static int hugetlbfs_setattr(struct dentry *dentry, struct iattr *attr)
{
675
	struct inode *inode = d_inode(dentry);
676
	struct hstate *h = hstate_inode(inode);
L
Linus Torvalds 已提交
677 678 679 680 681
	int error;
	unsigned int ia_valid = attr->ia_valid;

	BUG_ON(!inode);

682
	error = setattr_prepare(dentry, attr);
L
Linus Torvalds 已提交
683
	if (error)
C
Christoph Hellwig 已提交
684
		return error;
L
Linus Torvalds 已提交
685 686 687

	if (ia_valid & ATTR_SIZE) {
		error = -EINVAL;
C
Christoph Hellwig 已提交
688 689 690
		if (attr->ia_size & ~huge_page_mask(h))
			return -EINVAL;
		error = hugetlb_vmtruncate(inode, attr->ia_size);
L
Linus Torvalds 已提交
691
		if (error)
C
Christoph Hellwig 已提交
692
			return error;
L
Linus Torvalds 已提交
693
	}
C
Christoph Hellwig 已提交
694 695 696 697

	setattr_copy(inode, attr);
	mark_inode_dirty(inode);
	return 0;
L
Linus Torvalds 已提交
698 699
}

700 701
static struct inode *hugetlbfs_get_root(struct super_block *sb,
					struct hugetlbfs_config *config)
L
Linus Torvalds 已提交
702 703 704 705 706
{
	struct inode *inode;

	inode = new_inode(sb);
	if (inode) {
707
		inode->i_ino = get_next_ino();
708 709 710
		inode->i_mode = S_IFDIR | config->mode;
		inode->i_uid = config->uid;
		inode->i_gid = config->gid;
711
		inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
712 713 714 715
		inode->i_op = &hugetlbfs_dir_inode_operations;
		inode->i_fop = &simple_dir_operations;
		/* directory inodes start off with i_nlink == 2 (for "." entry) */
		inc_nlink(inode);
716
		lockdep_annotate_inode_mutex_key(inode);
717 718 719 720
	}
	return inode;
}

721
/*
722
 * Hugetlbfs is not reclaimable; therefore its i_mmap_rwsem will never
723
 * be taken from reclaim -- unlike regular filesystems. This needs an
724
 * annotation because huge_pmd_share() does an allocation under hugetlb's
725
 * i_mmap_rwsem.
726
 */
727
static struct lock_class_key hugetlbfs_i_mmap_rwsem_key;
728

729 730
static struct inode *hugetlbfs_get_inode(struct super_block *sb,
					struct inode *dir,
A
Al Viro 已提交
731
					umode_t mode, dev_t dev)
732 733
{
	struct inode *inode;
734 735 736 737 738
	struct resv_map *resv_map;

	resv_map = resv_map_alloc();
	if (!resv_map)
		return NULL;
739 740 741 742 743

	inode = new_inode(sb);
	if (inode) {
		inode->i_ino = get_next_ino();
		inode_init_owner(inode, dir, mode);
744 745
		lockdep_set_class(&inode->i_mapping->i_mmap_rwsem,
				&hugetlbfs_i_mmap_rwsem_key);
L
Linus Torvalds 已提交
746
		inode->i_mapping->a_ops = &hugetlbfs_aops;
747
		inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
748
		inode->i_mapping->private_data = resv_map;
L
Linus Torvalds 已提交
749 750 751 752 753 754 755 756 757 758 759 760 761
		switch (mode & S_IFMT) {
		default:
			init_special_inode(inode, mode, dev);
			break;
		case S_IFREG:
			inode->i_op = &hugetlbfs_inode_operations;
			inode->i_fop = &hugetlbfs_file_operations;
			break;
		case S_IFDIR:
			inode->i_op = &hugetlbfs_dir_inode_operations;
			inode->i_fop = &simple_dir_operations;

			/* directory inodes start off with i_nlink == 2 (for "." entry) */
762
			inc_nlink(inode);
L
Linus Torvalds 已提交
763 764 765
			break;
		case S_IFLNK:
			inode->i_op = &page_symlink_inode_operations;
766
			inode_nohighmem(inode);
L
Linus Torvalds 已提交
767 768
			break;
		}
769
		lockdep_annotate_inode_mutex_key(inode);
770 771 772
	} else
		kref_put(&resv_map->refs, resv_map_release);

L
Linus Torvalds 已提交
773 774 775 776 777 778 779
	return inode;
}

/*
 * File creation. Allocate an inode, and we're done..
 */
static int hugetlbfs_mknod(struct inode *dir,
A
Al Viro 已提交
780
			struct dentry *dentry, umode_t mode, dev_t dev)
L
Linus Torvalds 已提交
781 782 783
{
	struct inode *inode;
	int error = -ENOSPC;
784 785

	inode = hugetlbfs_get_inode(dir->i_sb, dir, mode, dev);
L
Linus Torvalds 已提交
786
	if (inode) {
787
		dir->i_ctime = dir->i_mtime = current_time(dir);
L
Linus Torvalds 已提交
788 789 790 791 792 793 794
		d_instantiate(dentry, inode);
		dget(dentry);	/* Extra count - pin the dentry in core */
		error = 0;
	}
	return error;
}

795
static int hugetlbfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
L
Linus Torvalds 已提交
796 797 798
{
	int retval = hugetlbfs_mknod(dir, dentry, mode | S_IFDIR, 0);
	if (!retval)
799
		inc_nlink(dir);
L
Linus Torvalds 已提交
800 801 802
	return retval;
}

A
Al Viro 已提交
803
static int hugetlbfs_create(struct inode *dir, struct dentry *dentry, umode_t mode, bool excl)
L
Linus Torvalds 已提交
804 805 806 807 808 809 810 811 812 813
{
	return hugetlbfs_mknod(dir, dentry, mode | S_IFREG, 0);
}

static int hugetlbfs_symlink(struct inode *dir,
			struct dentry *dentry, const char *symname)
{
	struct inode *inode;
	int error = -ENOSPC;

814
	inode = hugetlbfs_get_inode(dir->i_sb, dir, S_IFLNK|S_IRWXUGO, 0);
L
Linus Torvalds 已提交
815 816 817 818 819 820 821 822 823
	if (inode) {
		int l = strlen(symname)+1;
		error = page_symlink(inode, symname, l);
		if (!error) {
			d_instantiate(dentry, inode);
			dget(dentry);
		} else
			iput(inode);
	}
824
	dir->i_ctime = dir->i_mtime = current_time(dir);
L
Linus Torvalds 已提交
825 826 827 828 829

	return error;
}

/*
830
 * mark the head page dirty
L
Linus Torvalds 已提交
831 832 833
 */
static int hugetlbfs_set_page_dirty(struct page *page)
{
834
	struct page *head = compound_head(page);
835 836

	SetPageDirty(head);
L
Linus Torvalds 已提交
837 838 839
	return 0;
}

N
Naoya Horiguchi 已提交
840
static int hugetlbfs_migrate_page(struct address_space *mapping,
841
				struct page *newpage, struct page *page,
842
				enum migrate_mode mode)
N
Naoya Horiguchi 已提交
843 844 845 846
{
	int rc;

	rc = migrate_huge_page_move_mapping(mapping, newpage, page);
847
	if (rc != MIGRATEPAGE_SUCCESS)
N
Naoya Horiguchi 已提交
848 849 850
		return rc;
	migrate_page_copy(newpage, page);

851
	return MIGRATEPAGE_SUCCESS;
N
Naoya Horiguchi 已提交
852 853
}

854
static int hugetlbfs_statfs(struct dentry *dentry, struct kstatfs *buf)
L
Linus Torvalds 已提交
855
{
856
	struct hugetlbfs_sb_info *sbinfo = HUGETLBFS_SB(dentry->d_sb);
857
	struct hstate *h = hstate_inode(d_inode(dentry));
L
Linus Torvalds 已提交
858 859

	buf->f_type = HUGETLBFS_MAGIC;
860
	buf->f_bsize = huge_page_size(h);
L
Linus Torvalds 已提交
861 862
	if (sbinfo) {
		spin_lock(&sbinfo->stat_lock);
863 864
		/* If no limits set, just report 0 for max/free/used
		 * blocks, like simple_statfs() */
865 866 867 868 869 870 871 872 873
		if (sbinfo->spool) {
			long free_pages;

			spin_lock(&sbinfo->spool->lock);
			buf->f_blocks = sbinfo->spool->max_hpages;
			free_pages = sbinfo->spool->max_hpages
				- sbinfo->spool->used_hpages;
			buf->f_bavail = buf->f_bfree = free_pages;
			spin_unlock(&sbinfo->spool->lock);
874 875 876
			buf->f_files = sbinfo->max_inodes;
			buf->f_ffree = sbinfo->free_inodes;
		}
L
Linus Torvalds 已提交
877 878 879 880 881 882 883 884 885 886 887 888
		spin_unlock(&sbinfo->stat_lock);
	}
	buf->f_namelen = NAME_MAX;
	return 0;
}

static void hugetlbfs_put_super(struct super_block *sb)
{
	struct hugetlbfs_sb_info *sbi = HUGETLBFS_SB(sb);

	if (sbi) {
		sb->s_fs_info = NULL;
889 890 891 892

		if (sbi->spool)
			hugepage_put_subpool(sbi->spool);

L
Linus Torvalds 已提交
893 894 895 896
		kfree(sbi);
	}
}

897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921
static inline int hugetlbfs_dec_free_inodes(struct hugetlbfs_sb_info *sbinfo)
{
	if (sbinfo->free_inodes >= 0) {
		spin_lock(&sbinfo->stat_lock);
		if (unlikely(!sbinfo->free_inodes)) {
			spin_unlock(&sbinfo->stat_lock);
			return 0;
		}
		sbinfo->free_inodes--;
		spin_unlock(&sbinfo->stat_lock);
	}

	return 1;
}

static void hugetlbfs_inc_free_inodes(struct hugetlbfs_sb_info *sbinfo)
{
	if (sbinfo->free_inodes >= 0) {
		spin_lock(&sbinfo->stat_lock);
		sbinfo->free_inodes++;
		spin_unlock(&sbinfo->stat_lock);
	}
}


922
static struct kmem_cache *hugetlbfs_inode_cachep;
L
Linus Torvalds 已提交
923 924 925

static struct inode *hugetlbfs_alloc_inode(struct super_block *sb)
{
926
	struct hugetlbfs_sb_info *sbinfo = HUGETLBFS_SB(sb);
L
Linus Torvalds 已提交
927 928
	struct hugetlbfs_inode_info *p;

929 930
	if (unlikely(!hugetlbfs_dec_free_inodes(sbinfo)))
		return NULL;
931
	p = kmem_cache_alloc(hugetlbfs_inode_cachep, GFP_KERNEL);
932 933
	if (unlikely(!p)) {
		hugetlbfs_inc_free_inodes(sbinfo);
L
Linus Torvalds 已提交
934
		return NULL;
935
	}
936 937 938 939 940 941 942 943 944 945 946 947

	/*
	 * Any time after allocation, hugetlbfs_destroy_inode can be called
	 * for the inode.  mpol_free_shared_policy is unconditionally called
	 * as part of hugetlbfs_destroy_inode.  So, initialize policy here
	 * in case of a quick call to destroy.
	 *
	 * Note that the policy is initialized even if we are creating a
	 * private inode.  This simplifies hugetlbfs_destroy_inode.
	 */
	mpol_shared_policy_init(&p->policy, NULL);

L
Linus Torvalds 已提交
948 949 950
	return &p->vfs_inode;
}

N
Nick Piggin 已提交
951 952 953 954 955 956
static void hugetlbfs_i_callback(struct rcu_head *head)
{
	struct inode *inode = container_of(head, struct inode, i_rcu);
	kmem_cache_free(hugetlbfs_inode_cachep, HUGETLBFS_I(inode));
}

L
Linus Torvalds 已提交
957 958
static void hugetlbfs_destroy_inode(struct inode *inode)
{
959
	hugetlbfs_inc_free_inodes(HUGETLBFS_SB(inode->i_sb));
L
Linus Torvalds 已提交
960
	mpol_free_shared_policy(&HUGETLBFS_I(inode)->policy);
N
Nick Piggin 已提交
961
	call_rcu(&inode->i_rcu, hugetlbfs_i_callback);
L
Linus Torvalds 已提交
962 963
}

964
static const struct address_space_operations hugetlbfs_aops = {
N
Nick Piggin 已提交
965 966
	.write_begin	= hugetlbfs_write_begin,
	.write_end	= hugetlbfs_write_end,
L
Linus Torvalds 已提交
967
	.set_page_dirty	= hugetlbfs_set_page_dirty,
N
Naoya Horiguchi 已提交
968
	.migratepage    = hugetlbfs_migrate_page,
L
Linus Torvalds 已提交
969 970
};

971

972
static void init_once(void *foo)
973 974 975
{
	struct hugetlbfs_inode_info *ei = (struct hugetlbfs_inode_info *)foo;

C
Christoph Lameter 已提交
976
	inode_init_once(&ei->vfs_inode);
977 978
}

979
const struct file_operations hugetlbfs_file_operations = {
A
Al Viro 已提交
980
	.read_iter		= hugetlbfs_read_iter,
L
Linus Torvalds 已提交
981
	.mmap			= hugetlbfs_file_mmap,
982
	.fsync			= noop_fsync,
L
Linus Torvalds 已提交
983
	.get_unmapped_area	= hugetlb_get_unmapped_area,
984 985
	.llseek			= default_llseek,
	.fallocate		= hugetlbfs_fallocate,
L
Linus Torvalds 已提交
986 987
};

988
static const struct inode_operations hugetlbfs_dir_inode_operations = {
L
Linus Torvalds 已提交
989 990 991 992 993 994 995 996 997 998 999 1000
	.create		= hugetlbfs_create,
	.lookup		= simple_lookup,
	.link		= simple_link,
	.unlink		= simple_unlink,
	.symlink	= hugetlbfs_symlink,
	.mkdir		= hugetlbfs_mkdir,
	.rmdir		= simple_rmdir,
	.mknod		= hugetlbfs_mknod,
	.rename		= simple_rename,
	.setattr	= hugetlbfs_setattr,
};

1001
static const struct inode_operations hugetlbfs_inode_operations = {
L
Linus Torvalds 已提交
1002 1003 1004
	.setattr	= hugetlbfs_setattr,
};

1005
static const struct super_operations hugetlbfs_ops = {
L
Linus Torvalds 已提交
1006 1007
	.alloc_inode    = hugetlbfs_alloc_inode,
	.destroy_inode  = hugetlbfs_destroy_inode,
A
Al Viro 已提交
1008
	.evict_inode	= hugetlbfs_evict_inode,
L
Linus Torvalds 已提交
1009 1010
	.statfs		= hugetlbfs_statfs,
	.put_super	= hugetlbfs_put_super,
M
Miklos Szeredi 已提交
1011
	.show_options	= generic_show_options,
L
Linus Torvalds 已提交
1012 1013
};

1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037
enum { NO_SIZE, SIZE_STD, SIZE_PERCENT };

/*
 * Convert size option passed from command line to number of huge pages
 * in the pool specified by hstate.  Size option could be in bytes
 * (val_type == SIZE_STD) or percentage of the pool (val_type == SIZE_PERCENT).
 */
static long long
hugetlbfs_size_to_hpages(struct hstate *h, unsigned long long size_opt,
								int val_type)
{
	if (val_type == NO_SIZE)
		return -1;

	if (val_type == SIZE_PERCENT) {
		size_opt <<= huge_page_shift(h);
		size_opt *= h->max_huge_pages;
		do_div(size_opt, 100);
	}

	size_opt >>= huge_page_shift(h);
	return size_opt;
}

L
Linus Torvalds 已提交
1038 1039 1040
static int
hugetlbfs_parse_options(char *options, struct hugetlbfs_config *pconfig)
{
1041 1042 1043
	char *p, *rest;
	substring_t args[MAX_OPT_ARGS];
	int option;
1044 1045
	unsigned long long max_size_opt = 0, min_size_opt = 0;
	int max_val_type = NO_SIZE, min_val_type = NO_SIZE;
L
Linus Torvalds 已提交
1046 1047 1048 1049

	if (!options)
		return 0;

1050 1051
	while ((p = strsep(&options, ",")) != NULL) {
		int token;
1052 1053
		if (!*p)
			continue;
1054 1055 1056 1057 1058 1059

		token = match_token(p, tokens, args);
		switch (token) {
		case Opt_uid:
			if (match_int(&args[0], &option))
 				goto bad_val;
1060 1061 1062
			pconfig->uid = make_kuid(current_user_ns(), option);
			if (!uid_valid(pconfig->uid))
				goto bad_val;
1063 1064 1065 1066 1067
			break;

		case Opt_gid:
			if (match_int(&args[0], &option))
 				goto bad_val;
1068 1069 1070
			pconfig->gid = make_kgid(current_user_ns(), option);
			if (!gid_valid(pconfig->gid))
				goto bad_val;
1071 1072 1073 1074 1075
			break;

		case Opt_mode:
			if (match_octal(&args[0], &option))
 				goto bad_val;
1076
			pconfig->mode = option & 01777U;
1077 1078 1079 1080 1081 1082
			break;

		case Opt_size: {
			/* memparse() will accept a K/M/G without a digit */
			if (!isdigit(*args[0].from))
				goto bad_val;
1083 1084
			max_size_opt = memparse(args[0].from, &rest);
			max_val_type = SIZE_STD;
1085
			if (*rest == '%')
1086
				max_val_type = SIZE_PERCENT;
1087 1088
			break;
		}
L
Linus Torvalds 已提交
1089

1090 1091 1092 1093 1094 1095 1096
		case Opt_nr_inodes:
			/* memparse() will accept a K/M/G without a digit */
			if (!isdigit(*args[0].from))
				goto bad_val;
			pconfig->nr_inodes = memparse(args[0].from, &rest);
			break;

1097 1098 1099 1100 1101
		case Opt_pagesize: {
			unsigned long ps;
			ps = memparse(args[0].from, &rest);
			pconfig->hstate = size_to_hstate(ps);
			if (!pconfig->hstate) {
1102
				pr_err("Unsupported page size %lu MB\n",
1103 1104 1105 1106 1107 1108
					ps >> 20);
				return -EINVAL;
			}
			break;
		}

1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119
		case Opt_min_size: {
			/* memparse() will accept a K/M/G without a digit */
			if (!isdigit(*args[0].from))
				goto bad_val;
			min_size_opt = memparse(args[0].from, &rest);
			min_val_type = SIZE_STD;
			if (*rest == '%')
				min_val_type = SIZE_PERCENT;
			break;
		}

1120
		default:
1121
			pr_err("Bad mount option: \"%s\"\n", p);
1122
			return -EINVAL;
1123 1124
			break;
		}
L
Linus Torvalds 已提交
1125
	}
1126

1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142
	/*
	 * Use huge page pool size (in hstate) to convert the size
	 * options to number of huge pages.  If NO_SIZE, -1 is returned.
	 */
	pconfig->max_hpages = hugetlbfs_size_to_hpages(pconfig->hstate,
						max_size_opt, max_val_type);
	pconfig->min_hpages = hugetlbfs_size_to_hpages(pconfig->hstate,
						min_size_opt, min_val_type);

	/*
	 * If max_size was specified, then min_size must be smaller
	 */
	if (max_val_type > NO_SIZE &&
	    pconfig->min_hpages > pconfig->max_hpages) {
		pr_err("minimum size can not be greater than maximum size\n");
		return -EINVAL;
1143 1144
	}

L
Linus Torvalds 已提交
1145
	return 0;
1146 1147

bad_val:
1148
	pr_err("Bad value '%s' for mount option '%s'\n", args[0].from, p);
1149
 	return -EINVAL;
L
Linus Torvalds 已提交
1150 1151 1152 1153 1154 1155 1156 1157 1158
}

static int
hugetlbfs_fill_super(struct super_block *sb, void *data, int silent)
{
	int ret;
	struct hugetlbfs_config config;
	struct hugetlbfs_sb_info *sbinfo;

M
Miklos Szeredi 已提交
1159 1160
	save_mount_options(sb, data);

1161
	config.max_hpages = -1; /* No limit on size by default */
L
Linus Torvalds 已提交
1162
	config.nr_inodes = -1; /* No limit on number of inodes by default */
1163 1164
	config.uid = current_fsuid();
	config.gid = current_fsgid();
L
Linus Torvalds 已提交
1165
	config.mode = 0755;
1166
	config.hstate = &default_hstate;
1167
	config.min_hpages = -1; /* No default minimum size */
L
Linus Torvalds 已提交
1168 1169 1170 1171 1172 1173 1174 1175
	ret = hugetlbfs_parse_options(data, &config);
	if (ret)
		return ret;

	sbinfo = kmalloc(sizeof(struct hugetlbfs_sb_info), GFP_KERNEL);
	if (!sbinfo)
		return -ENOMEM;
	sb->s_fs_info = sbinfo;
1176
	sbinfo->hstate = config.hstate;
L
Linus Torvalds 已提交
1177 1178 1179
	spin_lock_init(&sbinfo->stat_lock);
	sbinfo->max_inodes = config.nr_inodes;
	sbinfo->free_inodes = config.nr_inodes;
1180
	sbinfo->spool = NULL;
1181 1182 1183 1184 1185 1186 1187 1188 1189
	/*
	 * Allocate and initialize subpool if maximum or minimum size is
	 * specified.  Any needed reservations (for minimim size) are taken
	 * taken when the subpool is created.
	 */
	if (config.max_hpages != -1 || config.min_hpages != -1) {
		sbinfo->spool = hugepage_new_subpool(config.hstate,
							config.max_hpages,
							config.min_hpages);
1190 1191 1192
		if (!sbinfo->spool)
			goto out_free;
	}
L
Linus Torvalds 已提交
1193
	sb->s_maxbytes = MAX_LFS_FILESIZE;
1194 1195
	sb->s_blocksize = huge_page_size(config.hstate);
	sb->s_blocksize_bits = huge_page_shift(config.hstate);
L
Linus Torvalds 已提交
1196 1197 1198
	sb->s_magic = HUGETLBFS_MAGIC;
	sb->s_op = &hugetlbfs_ops;
	sb->s_time_gran = 1;
1199 1200
	sb->s_root = d_make_root(hugetlbfs_get_root(sb, &config));
	if (!sb->s_root)
L
Linus Torvalds 已提交
1201 1202 1203
		goto out_free;
	return 0;
out_free:
1204
	kfree(sbinfo->spool);
L
Linus Torvalds 已提交
1205 1206 1207 1208
	kfree(sbinfo);
	return -ENOMEM;
}

A
Al Viro 已提交
1209 1210
static struct dentry *hugetlbfs_mount(struct file_system_type *fs_type,
	int flags, const char *dev_name, void *data)
L
Linus Torvalds 已提交
1211
{
A
Al Viro 已提交
1212
	return mount_nodev(fs_type, flags, data, hugetlbfs_fill_super);
L
Linus Torvalds 已提交
1213 1214 1215 1216
}

static struct file_system_type hugetlbfs_fs_type = {
	.name		= "hugetlbfs",
A
Al Viro 已提交
1217
	.mount		= hugetlbfs_mount,
L
Linus Torvalds 已提交
1218 1219 1220
	.kill_sb	= kill_litter_super,
};

1221
static struct vfsmount *hugetlbfs_vfsmount[HUGE_MAX_HSTATE];
L
Linus Torvalds 已提交
1222

1223
static int can_do_hugetlb_shm(void)
L
Linus Torvalds 已提交
1224
{
1225 1226 1227
	kgid_t shm_group;
	shm_group = make_kgid(&init_user_ns, sysctl_hugetlb_shm_group);
	return capable(CAP_IPC_LOCK) || in_group_p(shm_group);
L
Linus Torvalds 已提交
1228 1229
}

1230 1231
static int get_hstate_idx(int page_size_log)
{
1232
	struct hstate *h = hstate_sizelog(page_size_log);
1233 1234 1235 1236 1237 1238

	if (!h)
		return -1;
	return h - hstates;
}

1239
static const struct dentry_operations anon_ops = {
1240
	.d_dname = simple_dname
1241 1242
};

1243 1244 1245 1246 1247 1248
/*
 * Note that size should be aligned to proper hugepage size in caller side,
 * otherwise hugetlb_reserve_pages reserves one less hugepages than intended.
 */
struct file *hugetlb_file_setup(const char *name, size_t size,
				vm_flags_t acctflag, struct user_struct **user,
1249
				int creat_flags, int page_size_log)
L
Linus Torvalds 已提交
1250
{
1251
	struct file *file = ERR_PTR(-ENOMEM);
L
Linus Torvalds 已提交
1252
	struct inode *inode;
1253
	struct path path;
1254
	struct super_block *sb;
L
Linus Torvalds 已提交
1255
	struct qstr quick_string;
1256 1257 1258 1259 1260
	int hstate_idx;

	hstate_idx = get_hstate_idx(page_size_log);
	if (hstate_idx < 0)
		return ERR_PTR(-ENODEV);
L
Linus Torvalds 已提交
1261

1262
	*user = NULL;
1263
	if (!hugetlbfs_vfsmount[hstate_idx])
1264 1265
		return ERR_PTR(-ENOENT);

1266
	if (creat_flags == HUGETLB_SHMFS_INODE && !can_do_hugetlb_shm()) {
1267 1268
		*user = current_user();
		if (user_shm_lock(size, *user)) {
1269
			task_lock(current);
1270
			pr_warn_once("%s (%d): Using mlock ulimits for SHM_HUGETLB is deprecated\n",
1271 1272
				current->comm, current->pid);
			task_unlock(current);
1273 1274
		} else {
			*user = NULL;
1275
			return ERR_PTR(-EPERM);
1276
		}
1277
	}
L
Linus Torvalds 已提交
1278

1279
	sb = hugetlbfs_vfsmount[hstate_idx]->mnt_sb;
1280
	quick_string.name = name;
L
Linus Torvalds 已提交
1281 1282
	quick_string.len = strlen(quick_string.name);
	quick_string.hash = 0;
1283
	path.dentry = d_alloc_pseudo(sb, &quick_string);
1284
	if (!path.dentry)
L
Linus Torvalds 已提交
1285 1286
		goto out_shm_unlock;

1287
	d_set_d_op(path.dentry, &anon_ops);
1288
	path.mnt = mntget(hugetlbfs_vfsmount[hstate_idx]);
1289
	file = ERR_PTR(-ENOSPC);
1290
	inode = hugetlbfs_get_inode(sb, NULL, S_IFREG | S_IRWXUGO, 0);
L
Linus Torvalds 已提交
1291
	if (!inode)
1292
		goto out_dentry;
1293 1294
	if (creat_flags == HUGETLB_SHMFS_INODE)
		inode->i_flags |= S_PRIVATE;
L
Linus Torvalds 已提交
1295

1296
	file = ERR_PTR(-ENOMEM);
1297 1298 1299
	if (hugetlb_reserve_pages(inode, 0,
			size >> huge_page_shift(hstate_inode(inode)), NULL,
			acctflag))
1300 1301
		goto out_inode;

1302
	d_instantiate(path.dentry, inode);
L
Linus Torvalds 已提交
1303
	inode->i_size = size;
1304
	clear_nlink(inode);
1305

1306
	file = alloc_file(&path, FMODE_WRITE | FMODE_READ,
1307
			&hugetlbfs_file_operations);
1308
	if (IS_ERR(file))
1309
		goto out_dentry; /* inode is already attached */
1310

L
Linus Torvalds 已提交
1311 1312
	return file;

1313 1314
out_inode:
	iput(inode);
L
Linus Torvalds 已提交
1315
out_dentry:
1316
	path_put(&path);
L
Linus Torvalds 已提交
1317
out_shm_unlock:
1318 1319 1320 1321
	if (*user) {
		user_shm_unlock(size, *user);
		*user = NULL;
	}
1322
	return file;
L
Linus Torvalds 已提交
1323 1324 1325 1326
}

static int __init init_hugetlbfs_fs(void)
{
1327
	struct hstate *h;
L
Linus Torvalds 已提交
1328
	int error;
1329
	int i;
L
Linus Torvalds 已提交
1330

1331
	if (!hugepages_supported()) {
1332
		pr_info("disabling because there are no supported hugepage sizes\n");
1333 1334 1335
		return -ENOTSUPP;
	}

1336
	error = -ENOMEM;
L
Linus Torvalds 已提交
1337 1338
	hugetlbfs_inode_cachep = kmem_cache_create("hugetlbfs_inode_cache",
					sizeof(struct hugetlbfs_inode_info),
1339
					0, SLAB_ACCOUNT, init_once);
L
Linus Torvalds 已提交
1340
	if (hugetlbfs_inode_cachep == NULL)
P
Peter Zijlstra 已提交
1341
		goto out2;
L
Linus Torvalds 已提交
1342 1343 1344 1345 1346

	error = register_filesystem(&hugetlbfs_fs_type);
	if (error)
		goto out;

1347 1348 1349 1350
	i = 0;
	for_each_hstate(h) {
		char buf[50];
		unsigned ps_kb = 1U << (h->order + PAGE_SHIFT - 10);
L
Linus Torvalds 已提交
1351

1352 1353 1354
		snprintf(buf, sizeof(buf), "pagesize=%uK", ps_kb);
		hugetlbfs_vfsmount[i] = kern_mount_data(&hugetlbfs_fs_type,
							buf);
L
Linus Torvalds 已提交
1355

1356
		if (IS_ERR(hugetlbfs_vfsmount[i])) {
1357
			pr_err("Cannot mount internal hugetlbfs for "
1358 1359 1360 1361 1362 1363 1364 1365 1366
				"page size %uK", ps_kb);
			error = PTR_ERR(hugetlbfs_vfsmount[i]);
			hugetlbfs_vfsmount[i] = NULL;
		}
		i++;
	}
	/* Non default hstates are optional */
	if (!IS_ERR_OR_NULL(hugetlbfs_vfsmount[default_hstate_idx]))
		return 0;
L
Linus Torvalds 已提交
1367 1368

 out:
1369
	kmem_cache_destroy(hugetlbfs_inode_cachep);
P
Peter Zijlstra 已提交
1370
 out2:
L
Linus Torvalds 已提交
1371 1372
	return error;
}
1373
fs_initcall(init_hugetlbfs_fs)