inode.c 26.2 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 7 8 9 10 11 12 13 14 15
 *
 * Copyright (C) 2002 Linus Torvalds.
 */

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

#include <asm/uaccess.h>

38
static const struct super_operations hugetlbfs_ops;
39
static const struct address_space_operations hugetlbfs_aops;
40
const struct file_operations hugetlbfs_file_operations;
41 42
static const struct inode_operations hugetlbfs_dir_inode_operations;
static const struct inode_operations hugetlbfs_inode_operations;
L
Linus Torvalds 已提交
43

D
David Gibson 已提交
44
struct hugetlbfs_config {
45 46
	kuid_t   uid;
	kgid_t   gid;
D
David Gibson 已提交
47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62
	umode_t mode;
	long	nr_blocks;
	long	nr_inodes;
	struct hstate *hstate;
};

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
63
static struct backing_dev_info hugetlbfs_backing_dev_info = {
64
	.name		= "hugetlbfs",
L
Linus Torvalds 已提交
65
	.ra_pages	= 0,	/* No readahead */
66
	.capabilities	= BDI_CAP_NO_ACCT_AND_WRITEBACK,
L
Linus Torvalds 已提交
67 68 69 70
};

int sysctl_hugetlb_shm_group;

71 72 73
enum {
	Opt_size, Opt_nr_inodes,
	Opt_mode, Opt_uid, Opt_gid,
74
	Opt_pagesize,
75 76 77
	Opt_err,
};

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

88 89 90 91 92 93 94 95 96 97
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
98 99
static int hugetlbfs_file_mmap(struct file *file, struct vm_area_struct *vma)
{
A
Al Viro 已提交
100
	struct inode *inode = file_inode(file);
L
Linus Torvalds 已提交
101 102
	loff_t len, vma_len;
	int ret;
103
	struct hstate *h = hstate_file(file);
L
Linus Torvalds 已提交
104

105
	/*
106 107 108 109 110 111
	 * 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).
112
	 */
113
	vma->vm_flags |= VM_HUGETLB | VM_DONTEXPAND;
114
	vma->vm_ops = &hugetlb_vm_ops;
L
Linus Torvalds 已提交
115

116
	if (vma->vm_pgoff & (~huge_page_mask(h) >> PAGE_SHIFT))
117 118
		return -EINVAL;

L
Linus Torvalds 已提交
119 120
	vma_len = (loff_t)(vma->vm_end - vma->vm_start);

121
	mutex_lock(&inode->i_mutex);
L
Linus Torvalds 已提交
122 123 124 125 126
	file_accessed(file);

	ret = -ENOMEM;
	len = vma_len + ((loff_t)vma->vm_pgoff << PAGE_SHIFT);

127
	if (hugetlb_reserve_pages(inode,
128
				vma->vm_pgoff >> huge_page_order(h),
129 130
				len >> huge_page_shift(h), vma,
				vma->vm_flags))
131
		goto out;
132

A
Adam Litke 已提交
133 134
	ret = 0;
	hugetlb_prefault_arch_hook(vma->vm_mm);
135
	if (vma->vm_flags & VM_WRITE && inode->i_size < len)
L
Linus Torvalds 已提交
136 137
		inode->i_size = len;
out:
138
	mutex_unlock(&inode->i_mutex);
L
Linus Torvalds 已提交
139 140 141 142 143

	return ret;
}

/*
144
 * Called under down_write(mmap_sem).
L
Linus Torvalds 已提交
145 146
 */

147
#ifndef HAVE_ARCH_HUGETLB_UNMAPPED_AREA
L
Linus Torvalds 已提交
148 149 150 151 152 153
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;
154
	struct hstate *h = hstate_file(file);
155
	struct vm_unmapped_area_info info;
L
Linus Torvalds 已提交
156

157
	if (len & ~huge_page_mask(h))
L
Linus Torvalds 已提交
158 159 160 161
		return -EINVAL;
	if (len > TASK_SIZE)
		return -ENOMEM;

162
	if (flags & MAP_FIXED) {
163
		if (prepare_hugepage_range(file, addr, len))
164 165 166 167
			return -EINVAL;
		return addr;
	}

L
Linus Torvalds 已提交
168
	if (addr) {
169
		addr = ALIGN(addr, huge_page_size(h));
L
Linus Torvalds 已提交
170 171 172 173 174 175
		vma = find_vma(mm, addr);
		if (TASK_SIZE - len >= addr &&
		    (!vma || addr + len <= vma->vm_start))
			return addr;
	}

176 177 178 179 180 181 182
	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
183 184 185
}
#endif

B
Badari Pulavarty 已提交
186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231
static int
hugetlbfs_read_actor(struct page *page, unsigned long offset,
			char __user *buf, unsigned long count,
			unsigned long size)
{
	char *kaddr;
	unsigned long left, copied = 0;
	int i, chunksize;

	if (size > count)
		size = count;

	/* Find which 4k chunk and offset with in that chunk */
	i = offset >> PAGE_CACHE_SHIFT;
	offset = offset & ~PAGE_CACHE_MASK;

	while (size) {
		chunksize = PAGE_CACHE_SIZE;
		if (offset)
			chunksize -= offset;
		if (chunksize > size)
			chunksize = size;
		kaddr = kmap(&page[i]);
		left = __copy_to_user(buf, kaddr + offset, chunksize);
		kunmap(&page[i]);
		if (left) {
			copied += (chunksize - left);
			break;
		}
		offset = 0;
		size -= chunksize;
		buf += chunksize;
		copied += chunksize;
		i++;
	}
	return copied ? copied : -EFAULT;
}

/*
 * 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
 * since it has PAGE_CACHE_SIZE assumptions.
 */
static ssize_t hugetlbfs_read(struct file *filp, char __user *buf,
			      size_t len, loff_t *ppos)
{
232
	struct hstate *h = hstate_file(filp);
B
Badari Pulavarty 已提交
233 234
	struct address_space *mapping = filp->f_mapping;
	struct inode *inode = mapping->host;
235 236
	unsigned long index = *ppos >> huge_page_shift(h);
	unsigned long offset = *ppos & ~huge_page_mask(h);
B
Badari Pulavarty 已提交
237 238 239 240 241 242 243 244 245 246
	unsigned long end_index;
	loff_t isize;
	ssize_t retval = 0;

	/* validate length */
	if (len == 0)
		goto out;

	for (;;) {
		struct page *page;
247
		unsigned long nr, ret;
248
		int ra;
B
Badari Pulavarty 已提交
249 250

		/* nr is the maximum number of bytes to copy from this page */
251
		nr = huge_page_size(h);
252 253 254 255
		isize = i_size_read(inode);
		if (!isize)
			goto out;
		end_index = (isize - 1) >> huge_page_shift(h);
B
Badari Pulavarty 已提交
256 257 258
		if (index >= end_index) {
			if (index > end_index)
				goto out;
259
			nr = ((isize - 1) & ~huge_page_mask(h)) + 1;
260
			if (nr <= offset)
B
Badari Pulavarty 已提交
261 262 263 264 265
				goto out;
		}
		nr = nr - offset;

		/* Find the page */
266
		page = find_lock_page(mapping, index);
B
Badari Pulavarty 已提交
267 268 269 270 271 272 273
		if (unlikely(page == NULL)) {
			/*
			 * We have a HOLE, zero out the user-buffer for the
			 * length of the hole or request.
			 */
			ret = len < nr ? len : nr;
			if (clear_user(buf, ret))
274 275 276
				ra = -EFAULT;
			else
				ra = 0;
B
Badari Pulavarty 已提交
277
		} else {
278 279
			unlock_page(page);

B
Badari Pulavarty 已提交
280 281 282
			/*
			 * We have the page, copy it to user space buffer.
			 */
283 284
			ra = hugetlbfs_read_actor(page, offset, buf, len, nr);
			ret = ra;
285
			page_cache_release(page);
B
Badari Pulavarty 已提交
286
		}
287
		if (ra < 0) {
B
Badari Pulavarty 已提交
288
			if (retval == 0)
289
				retval = ra;
B
Badari Pulavarty 已提交
290 291 292 293 294 295
			goto out;
		}

		offset += ret;
		retval += ret;
		len -= ret;
296 297
		index += offset >> huge_page_shift(h);
		offset &= ~huge_page_mask(h);
B
Badari Pulavarty 已提交
298 299 300 301 302 303

		/* short read or no more work */
		if ((ret != nr) || (len == 0))
			break;
	}
out:
304
	*ppos = ((loff_t)index << huge_page_shift(h)) + offset;
B
Badari Pulavarty 已提交
305 306 307
	return retval;
}

N
Nick Piggin 已提交
308 309 310 311
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
312 313 314 315
{
	return -EINVAL;
}

N
Nick Piggin 已提交
316 317 318
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
319
{
N
Nick Piggin 已提交
320
	BUG();
L
Linus Torvalds 已提交
321 322 323 324 325
	return -EINVAL;
}

static void truncate_huge_page(struct page *page)
{
326
	cancel_dirty_page(page, /* No IO accounting for huge pages? */0);
L
Linus Torvalds 已提交
327
	ClearPageUptodate(page);
328
	delete_from_page_cache(page);
L
Linus Torvalds 已提交
329 330
}

331
static void truncate_hugepages(struct inode *inode, loff_t lstart)
L
Linus Torvalds 已提交
332
{
333
	struct hstate *h = hstate_inode(inode);
334
	struct address_space *mapping = &inode->i_data;
335
	const pgoff_t start = lstart >> huge_page_shift(h);
L
Linus Torvalds 已提交
336 337
	struct pagevec pvec;
	pgoff_t next;
338
	int i, freed = 0;
L
Linus Torvalds 已提交
339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358

	pagevec_init(&pvec, 0);
	next = start;
	while (1) {
		if (!pagevec_lookup(&pvec, mapping, next, PAGEVEC_SIZE)) {
			if (next == start)
				break;
			next = start;
			continue;
		}

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

			lock_page(page);
			if (page->index > next)
				next = page->index;
			++next;
			truncate_huge_page(page);
			unlock_page(page);
359
			freed++;
L
Linus Torvalds 已提交
360 361 362 363
		}
		huge_pagevec_release(&pvec);
	}
	BUG_ON(!lstart && mapping->nrpages);
364
	hugetlb_unreserve_pages(inode, start, freed);
L
Linus Torvalds 已提交
365 366
}

A
Al Viro 已提交
367
static void hugetlbfs_evict_inode(struct inode *inode)
L
Linus Torvalds 已提交
368
{
369 370
	struct resv_map *resv_map;

371
	truncate_hugepages(inode, 0);
372 373 374 375
	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);
376
	clear_inode(inode);
377 378
}

L
Linus Torvalds 已提交
379
static inline void
380
hugetlb_vmtruncate_list(struct rb_root *root, pgoff_t pgoff)
L
Linus Torvalds 已提交
381 382 383
{
	struct vm_area_struct *vma;

384
	vma_interval_tree_foreach(vma, root, pgoff, ULONG_MAX) {
L
Linus Torvalds 已提交
385 386 387
		unsigned long v_offset;

		/*
H
Hugh Dickins 已提交
388
		 * Can the expression below overflow on 32-bit arches?
389
		 * No, because the interval tree returns us only those vmas
H
Hugh Dickins 已提交
390 391
		 * which overlap the truncated area starting at pgoff,
		 * and no vma on a 32-bit arch can span beyond the 4GB.
L
Linus Torvalds 已提交
392
		 */
H
Hugh Dickins 已提交
393 394 395
		if (vma->vm_pgoff < pgoff)
			v_offset = (pgoff - vma->vm_pgoff) << PAGE_SHIFT;
		else
L
Linus Torvalds 已提交
396 397
			v_offset = 0;

398 399
		unmap_hugepage_range(vma, vma->vm_start + v_offset,
				     vma->vm_end, NULL);
L
Linus Torvalds 已提交
400 401 402 403 404
	}
}

static int hugetlb_vmtruncate(struct inode *inode, loff_t offset)
{
H
Hugh Dickins 已提交
405
	pgoff_t pgoff;
L
Linus Torvalds 已提交
406
	struct address_space *mapping = inode->i_mapping;
407
	struct hstate *h = hstate_inode(inode);
L
Linus Torvalds 已提交
408

409
	BUG_ON(offset & ~huge_page_mask(h));
H
Hugh Dickins 已提交
410
	pgoff = offset >> PAGE_SHIFT;
L
Linus Torvalds 已提交
411

412
	i_size_write(inode, offset);
413
	mutex_lock(&mapping->i_mmap_mutex);
414
	if (!RB_EMPTY_ROOT(&mapping->i_mmap))
L
Linus Torvalds 已提交
415
		hugetlb_vmtruncate_list(&mapping->i_mmap, pgoff);
416
	mutex_unlock(&mapping->i_mmap_mutex);
417
	truncate_hugepages(inode, offset);
L
Linus Torvalds 已提交
418 419 420 421 422 423
	return 0;
}

static int hugetlbfs_setattr(struct dentry *dentry, struct iattr *attr)
{
	struct inode *inode = dentry->d_inode;
424
	struct hstate *h = hstate_inode(inode);
L
Linus Torvalds 已提交
425 426 427 428 429 430 431
	int error;
	unsigned int ia_valid = attr->ia_valid;

	BUG_ON(!inode);

	error = inode_change_ok(inode, attr);
	if (error)
C
Christoph Hellwig 已提交
432
		return error;
L
Linus Torvalds 已提交
433 434 435

	if (ia_valid & ATTR_SIZE) {
		error = -EINVAL;
C
Christoph Hellwig 已提交
436 437 438
		if (attr->ia_size & ~huge_page_mask(h))
			return -EINVAL;
		error = hugetlb_vmtruncate(inode, attr->ia_size);
L
Linus Torvalds 已提交
439
		if (error)
C
Christoph Hellwig 已提交
440
			return error;
L
Linus Torvalds 已提交
441
	}
C
Christoph Hellwig 已提交
442 443 444 445

	setattr_copy(inode, attr);
	mark_inode_dirty(inode);
	return 0;
L
Linus Torvalds 已提交
446 447
}

448 449
static struct inode *hugetlbfs_get_root(struct super_block *sb,
					struct hugetlbfs_config *config)
L
Linus Torvalds 已提交
450 451 452 453 454 455
{
	struct inode *inode;

	inode = new_inode(sb);
	if (inode) {
		struct hugetlbfs_inode_info *info;
456
		inode->i_ino = get_next_ino();
457 458 459 460 461 462 463 464 465 466
		inode->i_mode = S_IFDIR | config->mode;
		inode->i_uid = config->uid;
		inode->i_gid = config->gid;
		inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
		info = HUGETLBFS_I(inode);
		mpol_shared_policy_init(&info->policy, NULL);
		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);
467
		lockdep_annotate_inode_mutex_key(inode);
468 469 470 471
	}
	return inode;
}

472 473 474 475 476 477 478 479
/*
 * Hugetlbfs is not reclaimable; therefore its i_mmap_mutex will never
 * be taken from reclaim -- unlike regular filesystems. This needs an
 * annotation because huge_pmd_share() does an allocation under
 * i_mmap_mutex.
 */
struct lock_class_key hugetlbfs_i_mmap_mutex_key;

480 481
static struct inode *hugetlbfs_get_inode(struct super_block *sb,
					struct inode *dir,
A
Al Viro 已提交
482
					umode_t mode, dev_t dev)
483 484
{
	struct inode *inode;
485 486 487 488 489
	struct resv_map *resv_map;

	resv_map = resv_map_alloc();
	if (!resv_map)
		return NULL;
490 491 492 493 494 495

	inode = new_inode(sb);
	if (inode) {
		struct hugetlbfs_inode_info *info;
		inode->i_ino = get_next_ino();
		inode_init_owner(inode, dir, mode);
496 497
		lockdep_set_class(&inode->i_mapping->i_mmap_mutex,
				&hugetlbfs_i_mmap_mutex_key);
L
Linus Torvalds 已提交
498 499 500
		inode->i_mapping->a_ops = &hugetlbfs_aops;
		inode->i_mapping->backing_dev_info =&hugetlbfs_backing_dev_info;
		inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
501
		inode->i_mapping->private_data = resv_map;
L
Linus Torvalds 已提交
502
		info = HUGETLBFS_I(inode);
503 504 505 506 507 508 509
		/*
		 * The policy is initialized here even if we are creating a
		 * private inode because initialization simply creates an
		 * an empty rb tree and calls spin_lock_init(), later when we
		 * call mpol_free_shared_policy() it will just return because
		 * the rb tree will still be empty.
		 */
510
		mpol_shared_policy_init(&info->policy, NULL);
L
Linus Torvalds 已提交
511 512 513 514 515 516 517 518 519 520 521 522 523
		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) */
524
			inc_nlink(inode);
L
Linus Torvalds 已提交
525 526 527 528 529
			break;
		case S_IFLNK:
			inode->i_op = &page_symlink_inode_operations;
			break;
		}
530
		lockdep_annotate_inode_mutex_key(inode);
531 532 533
	} else
		kref_put(&resv_map->refs, resv_map_release);

L
Linus Torvalds 已提交
534 535 536 537 538 539 540
	return inode;
}

/*
 * File creation. Allocate an inode, and we're done..
 */
static int hugetlbfs_mknod(struct inode *dir,
A
Al Viro 已提交
541
			struct dentry *dentry, umode_t mode, dev_t dev)
L
Linus Torvalds 已提交
542 543 544
{
	struct inode *inode;
	int error = -ENOSPC;
545 546

	inode = hugetlbfs_get_inode(dir->i_sb, dir, mode, dev);
L
Linus Torvalds 已提交
547 548 549 550 551 552 553 554 555
	if (inode) {
		dir->i_ctime = dir->i_mtime = CURRENT_TIME;
		d_instantiate(dentry, inode);
		dget(dentry);	/* Extra count - pin the dentry in core */
		error = 0;
	}
	return error;
}

556
static int hugetlbfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
L
Linus Torvalds 已提交
557 558 559
{
	int retval = hugetlbfs_mknod(dir, dentry, mode | S_IFDIR, 0);
	if (!retval)
560
		inc_nlink(dir);
L
Linus Torvalds 已提交
561 562 563
	return retval;
}

A
Al Viro 已提交
564
static int hugetlbfs_create(struct inode *dir, struct dentry *dentry, umode_t mode, bool excl)
L
Linus Torvalds 已提交
565 566 567 568 569 570 571 572 573 574
{
	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;

575
	inode = hugetlbfs_get_inode(dir->i_sb, dir, S_IFLNK|S_IRWXUGO, 0);
L
Linus Torvalds 已提交
576 577 578 579 580 581 582 583 584 585 586 587 588 589 590
	if (inode) {
		int l = strlen(symname)+1;
		error = page_symlink(inode, symname, l);
		if (!error) {
			d_instantiate(dentry, inode);
			dget(dentry);
		} else
			iput(inode);
	}
	dir->i_ctime = dir->i_mtime = CURRENT_TIME;

	return error;
}

/*
591
 * mark the head page dirty
L
Linus Torvalds 已提交
592 593 594
 */
static int hugetlbfs_set_page_dirty(struct page *page)
{
595
	struct page *head = compound_head(page);
596 597

	SetPageDirty(head);
L
Linus Torvalds 已提交
598 599 600
	return 0;
}

N
Naoya Horiguchi 已提交
601
static int hugetlbfs_migrate_page(struct address_space *mapping,
602
				struct page *newpage, struct page *page,
603
				enum migrate_mode mode)
N
Naoya Horiguchi 已提交
604 605 606 607
{
	int rc;

	rc = migrate_huge_page_move_mapping(mapping, newpage, page);
608
	if (rc != MIGRATEPAGE_SUCCESS)
N
Naoya Horiguchi 已提交
609 610 611
		return rc;
	migrate_page_copy(newpage, page);

612
	return MIGRATEPAGE_SUCCESS;
N
Naoya Horiguchi 已提交
613 614
}

615
static int hugetlbfs_statfs(struct dentry *dentry, struct kstatfs *buf)
L
Linus Torvalds 已提交
616
{
617
	struct hugetlbfs_sb_info *sbinfo = HUGETLBFS_SB(dentry->d_sb);
618
	struct hstate *h = hstate_inode(dentry->d_inode);
L
Linus Torvalds 已提交
619 620

	buf->f_type = HUGETLBFS_MAGIC;
621
	buf->f_bsize = huge_page_size(h);
L
Linus Torvalds 已提交
622 623
	if (sbinfo) {
		spin_lock(&sbinfo->stat_lock);
624 625
		/* If no limits set, just report 0 for max/free/used
		 * blocks, like simple_statfs() */
626 627 628 629 630 631 632 633 634
		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);
635 636 637
			buf->f_files = sbinfo->max_inodes;
			buf->f_ffree = sbinfo->free_inodes;
		}
L
Linus Torvalds 已提交
638 639 640 641 642 643 644 645 646 647 648 649
		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;
650 651 652 653

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

L
Linus Torvalds 已提交
654 655 656 657
		kfree(sbi);
	}
}

658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682
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);
	}
}


683
static struct kmem_cache *hugetlbfs_inode_cachep;
L
Linus Torvalds 已提交
684 685 686

static struct inode *hugetlbfs_alloc_inode(struct super_block *sb)
{
687
	struct hugetlbfs_sb_info *sbinfo = HUGETLBFS_SB(sb);
L
Linus Torvalds 已提交
688 689
	struct hugetlbfs_inode_info *p;

690 691
	if (unlikely(!hugetlbfs_dec_free_inodes(sbinfo)))
		return NULL;
692
	p = kmem_cache_alloc(hugetlbfs_inode_cachep, GFP_KERNEL);
693 694
	if (unlikely(!p)) {
		hugetlbfs_inc_free_inodes(sbinfo);
L
Linus Torvalds 已提交
695
		return NULL;
696
	}
L
Linus Torvalds 已提交
697 698 699
	return &p->vfs_inode;
}

N
Nick Piggin 已提交
700 701 702 703 704 705
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
706 707
static void hugetlbfs_destroy_inode(struct inode *inode)
{
708
	hugetlbfs_inc_free_inodes(HUGETLBFS_SB(inode->i_sb));
L
Linus Torvalds 已提交
709
	mpol_free_shared_policy(&HUGETLBFS_I(inode)->policy);
N
Nick Piggin 已提交
710
	call_rcu(&inode->i_rcu, hugetlbfs_i_callback);
L
Linus Torvalds 已提交
711 712
}

713
static const struct address_space_operations hugetlbfs_aops = {
N
Nick Piggin 已提交
714 715
	.write_begin	= hugetlbfs_write_begin,
	.write_end	= hugetlbfs_write_end,
L
Linus Torvalds 已提交
716
	.set_page_dirty	= hugetlbfs_set_page_dirty,
N
Naoya Horiguchi 已提交
717
	.migratepage    = hugetlbfs_migrate_page,
L
Linus Torvalds 已提交
718 719
};

720

721
static void init_once(void *foo)
722 723 724
{
	struct hugetlbfs_inode_info *ei = (struct hugetlbfs_inode_info *)foo;

C
Christoph Lameter 已提交
725
	inode_init_once(&ei->vfs_inode);
726 727
}

728
const struct file_operations hugetlbfs_file_operations = {
B
Badari Pulavarty 已提交
729
	.read			= hugetlbfs_read,
L
Linus Torvalds 已提交
730
	.mmap			= hugetlbfs_file_mmap,
731
	.fsync			= noop_fsync,
L
Linus Torvalds 已提交
732
	.get_unmapped_area	= hugetlb_get_unmapped_area,
733
	.llseek		= default_llseek,
L
Linus Torvalds 已提交
734 735
};

736
static const struct inode_operations hugetlbfs_dir_inode_operations = {
L
Linus Torvalds 已提交
737 738 739 740 741 742 743 744 745 746 747 748
	.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,
};

749
static const struct inode_operations hugetlbfs_inode_operations = {
L
Linus Torvalds 已提交
750 751 752
	.setattr	= hugetlbfs_setattr,
};

753
static const struct super_operations hugetlbfs_ops = {
L
Linus Torvalds 已提交
754 755
	.alloc_inode    = hugetlbfs_alloc_inode,
	.destroy_inode  = hugetlbfs_destroy_inode,
A
Al Viro 已提交
756
	.evict_inode	= hugetlbfs_evict_inode,
L
Linus Torvalds 已提交
757 758
	.statfs		= hugetlbfs_statfs,
	.put_super	= hugetlbfs_put_super,
M
Miklos Szeredi 已提交
759
	.show_options	= generic_show_options,
L
Linus Torvalds 已提交
760 761 762 763 764
};

static int
hugetlbfs_parse_options(char *options, struct hugetlbfs_config *pconfig)
{
765 766 767
	char *p, *rest;
	substring_t args[MAX_OPT_ARGS];
	int option;
768 769
	unsigned long long size = 0;
	enum { NO_SIZE, SIZE_STD, SIZE_PERCENT } setsize = NO_SIZE;
L
Linus Torvalds 已提交
770 771 772 773

	if (!options)
		return 0;

774 775
	while ((p = strsep(&options, ",")) != NULL) {
		int token;
776 777
		if (!*p)
			continue;
778 779 780 781 782 783

		token = match_token(p, tokens, args);
		switch (token) {
		case Opt_uid:
			if (match_int(&args[0], &option))
 				goto bad_val;
784 785 786
			pconfig->uid = make_kuid(current_user_ns(), option);
			if (!uid_valid(pconfig->uid))
				goto bad_val;
787 788 789 790 791
			break;

		case Opt_gid:
			if (match_int(&args[0], &option))
 				goto bad_val;
792 793 794
			pconfig->gid = make_kgid(current_user_ns(), option);
			if (!gid_valid(pconfig->gid))
				goto bad_val;
795 796 797 798 799
			break;

		case Opt_mode:
			if (match_octal(&args[0], &option))
 				goto bad_val;
800
			pconfig->mode = option & 01777U;
801 802 803 804 805 806 807
			break;

		case Opt_size: {
			/* memparse() will accept a K/M/G without a digit */
			if (!isdigit(*args[0].from))
				goto bad_val;
			size = memparse(args[0].from, &rest);
808 809 810
			setsize = SIZE_STD;
			if (*rest == '%')
				setsize = SIZE_PERCENT;
811 812
			break;
		}
L
Linus Torvalds 已提交
813

814 815 816 817 818 819 820
		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;

821 822 823 824 825 826 827 828 829 830 831 832 833
		case Opt_pagesize: {
			unsigned long ps;
			ps = memparse(args[0].from, &rest);
			pconfig->hstate = size_to_hstate(ps);
			if (!pconfig->hstate) {
				printk(KERN_ERR
				"hugetlbfs: Unsupported page size %lu MB\n",
					ps >> 20);
				return -EINVAL;
			}
			break;
		}

834
		default:
835 836 837
			printk(KERN_ERR "hugetlbfs: Bad mount option: \"%s\"\n",
				 p);
			return -EINVAL;
838 839
			break;
		}
L
Linus Torvalds 已提交
840
	}
841 842 843 844 845 846 847 848 849 850 851 852

	/* Do size after hstate is set up */
	if (setsize > NO_SIZE) {
		struct hstate *h = pconfig->hstate;
		if (setsize == SIZE_PERCENT) {
			size <<= huge_page_shift(h);
			size *= h->max_huge_pages;
			do_div(size, 100);
		}
		pconfig->nr_blocks = (size >> huge_page_shift(h));
	}

L
Linus Torvalds 已提交
853
	return 0;
854 855 856 857

bad_val:
 	printk(KERN_ERR "hugetlbfs: Bad value '%s' for mount option '%s'\n",
	       args[0].from, p);
858
 	return -EINVAL;
L
Linus Torvalds 已提交
859 860 861 862 863 864 865 866 867
}

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
868 869
	save_mount_options(sb, data);

L
Linus Torvalds 已提交
870 871
	config.nr_blocks = -1; /* No limit on size by default */
	config.nr_inodes = -1; /* No limit on number of inodes by default */
872 873
	config.uid = current_fsuid();
	config.gid = current_fsgid();
L
Linus Torvalds 已提交
874
	config.mode = 0755;
875
	config.hstate = &default_hstate;
L
Linus Torvalds 已提交
876 877 878 879 880 881 882 883
	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;
884
	sbinfo->hstate = config.hstate;
L
Linus Torvalds 已提交
885 886 887
	spin_lock_init(&sbinfo->stat_lock);
	sbinfo->max_inodes = config.nr_inodes;
	sbinfo->free_inodes = config.nr_inodes;
888 889 890 891 892 893
	sbinfo->spool = NULL;
	if (config.nr_blocks != -1) {
		sbinfo->spool = hugepage_new_subpool(config.nr_blocks);
		if (!sbinfo->spool)
			goto out_free;
	}
L
Linus Torvalds 已提交
894
	sb->s_maxbytes = MAX_LFS_FILESIZE;
895 896
	sb->s_blocksize = huge_page_size(config.hstate);
	sb->s_blocksize_bits = huge_page_shift(config.hstate);
L
Linus Torvalds 已提交
897 898 899
	sb->s_magic = HUGETLBFS_MAGIC;
	sb->s_op = &hugetlbfs_ops;
	sb->s_time_gran = 1;
900 901
	sb->s_root = d_make_root(hugetlbfs_get_root(sb, &config));
	if (!sb->s_root)
L
Linus Torvalds 已提交
902 903 904
		goto out_free;
	return 0;
out_free:
905 906
	if (sbinfo->spool)
		kfree(sbinfo->spool);
L
Linus Torvalds 已提交
907 908 909 910
	kfree(sbinfo);
	return -ENOMEM;
}

A
Al Viro 已提交
911 912
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
913
{
A
Al Viro 已提交
914
	return mount_nodev(fs_type, flags, data, hugetlbfs_fill_super);
L
Linus Torvalds 已提交
915 916 917 918
}

static struct file_system_type hugetlbfs_fs_type = {
	.name		= "hugetlbfs",
A
Al Viro 已提交
919
	.mount		= hugetlbfs_mount,
L
Linus Torvalds 已提交
920 921
	.kill_sb	= kill_litter_super,
};
922
MODULE_ALIAS_FS("hugetlbfs");
L
Linus Torvalds 已提交
923

924
static struct vfsmount *hugetlbfs_vfsmount[HUGE_MAX_HSTATE];
L
Linus Torvalds 已提交
925

926
static int can_do_hugetlb_shm(void)
L
Linus Torvalds 已提交
927
{
928 929 930
	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
931 932
}

933 934
static int get_hstate_idx(int page_size_log)
{
935
	struct hstate *h = hstate_sizelog(page_size_log);
936 937 938 939 940 941

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

942
static struct dentry_operations anon_ops = {
943
	.d_dname = simple_dname
944 945
};

946 947 948 949 950 951
/*
 * 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,
952
				int creat_flags, int page_size_log)
L
Linus Torvalds 已提交
953
{
954
	struct file *file = ERR_PTR(-ENOMEM);
L
Linus Torvalds 已提交
955
	struct inode *inode;
956
	struct path path;
957
	struct super_block *sb;
L
Linus Torvalds 已提交
958
	struct qstr quick_string;
959 960 961 962 963
	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
964

965
	*user = NULL;
966
	if (!hugetlbfs_vfsmount[hstate_idx])
967 968
		return ERR_PTR(-ENOENT);

969
	if (creat_flags == HUGETLB_SHMFS_INODE && !can_do_hugetlb_shm()) {
970 971
		*user = current_user();
		if (user_shm_lock(size, *user)) {
972 973 974 975 976
			task_lock(current);
			printk_once(KERN_WARNING
				"%s (%d): Using mlock ulimits for SHM_HUGETLB is deprecated\n",
				current->comm, current->pid);
			task_unlock(current);
977 978
		} else {
			*user = NULL;
979
			return ERR_PTR(-EPERM);
980
		}
981
	}
L
Linus Torvalds 已提交
982

983
	sb = hugetlbfs_vfsmount[hstate_idx]->mnt_sb;
984
	quick_string.name = name;
L
Linus Torvalds 已提交
985 986
	quick_string.len = strlen(quick_string.name);
	quick_string.hash = 0;
987
	path.dentry = d_alloc_pseudo(sb, &quick_string);
988
	if (!path.dentry)
L
Linus Torvalds 已提交
989 990
		goto out_shm_unlock;

991
	d_set_d_op(path.dentry, &anon_ops);
992
	path.mnt = mntget(hugetlbfs_vfsmount[hstate_idx]);
993
	file = ERR_PTR(-ENOSPC);
994
	inode = hugetlbfs_get_inode(sb, NULL, S_IFREG | S_IRWXUGO, 0);
L
Linus Torvalds 已提交
995
	if (!inode)
996
		goto out_dentry;
L
Linus Torvalds 已提交
997

998
	file = ERR_PTR(-ENOMEM);
999 1000 1001
	if (hugetlb_reserve_pages(inode, 0,
			size >> huge_page_shift(hstate_inode(inode)), NULL,
			acctflag))
1002 1003
		goto out_inode;

1004
	d_instantiate(path.dentry, inode);
L
Linus Torvalds 已提交
1005
	inode->i_size = size;
1006
	clear_nlink(inode);
1007

1008
	file = alloc_file(&path, FMODE_WRITE | FMODE_READ,
1009
			&hugetlbfs_file_operations);
1010
	if (IS_ERR(file))
1011
		goto out_dentry; /* inode is already attached */
1012

L
Linus Torvalds 已提交
1013 1014
	return file;

1015 1016
out_inode:
	iput(inode);
L
Linus Torvalds 已提交
1017
out_dentry:
1018
	path_put(&path);
L
Linus Torvalds 已提交
1019
out_shm_unlock:
1020 1021 1022 1023
	if (*user) {
		user_shm_unlock(size, *user);
		*user = NULL;
	}
1024
	return file;
L
Linus Torvalds 已提交
1025 1026 1027 1028
}

static int __init init_hugetlbfs_fs(void)
{
1029
	struct hstate *h;
L
Linus Torvalds 已提交
1030
	int error;
1031
	int i;
L
Linus Torvalds 已提交
1032

1033 1034 1035 1036 1037
	if (!hugepages_supported()) {
		pr_info("hugetlbfs: disabling because there are no supported hugepage sizes\n");
		return -ENOTSUPP;
	}

P
Peter Zijlstra 已提交
1038 1039 1040 1041
	error = bdi_init(&hugetlbfs_backing_dev_info);
	if (error)
		return error;

1042
	error = -ENOMEM;
L
Linus Torvalds 已提交
1043 1044
	hugetlbfs_inode_cachep = kmem_cache_create("hugetlbfs_inode_cache",
					sizeof(struct hugetlbfs_inode_info),
1045
					0, 0, init_once);
L
Linus Torvalds 已提交
1046
	if (hugetlbfs_inode_cachep == NULL)
P
Peter Zijlstra 已提交
1047
		goto out2;
L
Linus Torvalds 已提交
1048 1049 1050 1051 1052

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

1053 1054 1055 1056
	i = 0;
	for_each_hstate(h) {
		char buf[50];
		unsigned ps_kb = 1U << (h->order + PAGE_SHIFT - 10);
L
Linus Torvalds 已提交
1057

1058 1059 1060
		snprintf(buf, sizeof(buf), "pagesize=%uK", ps_kb);
		hugetlbfs_vfsmount[i] = kern_mount_data(&hugetlbfs_fs_type,
							buf);
L
Linus Torvalds 已提交
1061

1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072
		if (IS_ERR(hugetlbfs_vfsmount[i])) {
			pr_err("hugetlb: Cannot mount internal hugetlbfs for "
				"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
1073 1074

 out:
1075
	kmem_cache_destroy(hugetlbfs_inode_cachep);
P
Peter Zijlstra 已提交
1076 1077
 out2:
	bdi_destroy(&hugetlbfs_backing_dev_info);
L
Linus Torvalds 已提交
1078 1079 1080 1081 1082
	return error;
}

static void __exit exit_hugetlbfs_fs(void)
{
1083 1084 1085 1086
	struct hstate *h;
	int i;


1087 1088 1089 1090 1091
	/*
	 * Make sure all delayed rcu free inodes are flushed before we
	 * destroy cache.
	 */
	rcu_barrier();
L
Linus Torvalds 已提交
1092
	kmem_cache_destroy(hugetlbfs_inode_cachep);
1093 1094 1095
	i = 0;
	for_each_hstate(h)
		kern_unmount(hugetlbfs_vfsmount[i++]);
L
Linus Torvalds 已提交
1096
	unregister_filesystem(&hugetlbfs_fs_type);
P
Peter Zijlstra 已提交
1097
	bdi_destroy(&hugetlbfs_backing_dev_info);
L
Linus Torvalds 已提交
1098 1099 1100 1101 1102 1103
}

module_init(init_hugetlbfs_fs)
module_exit(exit_hugetlbfs_fs)

MODULE_LICENSE("GPL");