inode.c 25.3 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 | VM_DONTDUMP;
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
	truncate_hugepages(inode, 0);
370
	clear_inode(inode);
371 372
}

L
Linus Torvalds 已提交
373
static inline void
374
hugetlb_vmtruncate_list(struct rb_root *root, pgoff_t pgoff)
L
Linus Torvalds 已提交
375 376 377
{
	struct vm_area_struct *vma;

378
	vma_interval_tree_foreach(vma, root, pgoff, ULONG_MAX) {
L
Linus Torvalds 已提交
379 380 381
		unsigned long v_offset;

		/*
H
Hugh Dickins 已提交
382
		 * Can the expression below overflow on 32-bit arches?
383
		 * No, because the interval tree returns us only those vmas
H
Hugh Dickins 已提交
384 385
		 * 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
386
		 */
H
Hugh Dickins 已提交
387 388 389
		if (vma->vm_pgoff < pgoff)
			v_offset = (pgoff - vma->vm_pgoff) << PAGE_SHIFT;
		else
L
Linus Torvalds 已提交
390 391
			v_offset = 0;

392 393
		unmap_hugepage_range(vma, vma->vm_start + v_offset,
				     vma->vm_end, NULL);
L
Linus Torvalds 已提交
394 395 396 397 398
	}
}

static int hugetlb_vmtruncate(struct inode *inode, loff_t offset)
{
H
Hugh Dickins 已提交
399
	pgoff_t pgoff;
L
Linus Torvalds 已提交
400
	struct address_space *mapping = inode->i_mapping;
401
	struct hstate *h = hstate_inode(inode);
L
Linus Torvalds 已提交
402

403
	BUG_ON(offset & ~huge_page_mask(h));
H
Hugh Dickins 已提交
404
	pgoff = offset >> PAGE_SHIFT;
L
Linus Torvalds 已提交
405

406
	i_size_write(inode, offset);
407
	mutex_lock(&mapping->i_mmap_mutex);
408
	if (!RB_EMPTY_ROOT(&mapping->i_mmap))
L
Linus Torvalds 已提交
409
		hugetlb_vmtruncate_list(&mapping->i_mmap, pgoff);
410
	mutex_unlock(&mapping->i_mmap_mutex);
411
	truncate_hugepages(inode, offset);
L
Linus Torvalds 已提交
412 413 414 415 416 417
	return 0;
}

static int hugetlbfs_setattr(struct dentry *dentry, struct iattr *attr)
{
	struct inode *inode = dentry->d_inode;
418
	struct hstate *h = hstate_inode(inode);
L
Linus Torvalds 已提交
419 420 421 422 423 424 425
	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
426
		return error;
L
Linus Torvalds 已提交
427 428 429

	if (ia_valid & ATTR_SIZE) {
		error = -EINVAL;
C
Christoph Hellwig 已提交
430 431 432
		if (attr->ia_size & ~huge_page_mask(h))
			return -EINVAL;
		error = hugetlb_vmtruncate(inode, attr->ia_size);
L
Linus Torvalds 已提交
433
		if (error)
C
Christoph Hellwig 已提交
434
			return error;
L
Linus Torvalds 已提交
435
	}
C
Christoph Hellwig 已提交
436 437 438 439

	setattr_copy(inode, attr);
	mark_inode_dirty(inode);
	return 0;
L
Linus Torvalds 已提交
440 441
}

442 443
static struct inode *hugetlbfs_get_root(struct super_block *sb,
					struct hugetlbfs_config *config)
L
Linus Torvalds 已提交
444 445 446 447 448 449
{
	struct inode *inode;

	inode = new_inode(sb);
	if (inode) {
		struct hugetlbfs_inode_info *info;
450
		inode->i_ino = get_next_ino();
451 452 453 454 455 456 457 458 459 460
		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);
461
		lockdep_annotate_inode_mutex_key(inode);
462 463 464 465 466 467
	}
	return inode;
}

static struct inode *hugetlbfs_get_inode(struct super_block *sb,
					struct inode *dir,
A
Al Viro 已提交
468
					umode_t mode, dev_t dev)
469 470 471 472 473 474 475 476
{
	struct inode *inode;

	inode = new_inode(sb);
	if (inode) {
		struct hugetlbfs_inode_info *info;
		inode->i_ino = get_next_ino();
		inode_init_owner(inode, dir, mode);
L
Linus Torvalds 已提交
477 478 479
		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;
480
		INIT_LIST_HEAD(&inode->i_mapping->private_list);
L
Linus Torvalds 已提交
481
		info = HUGETLBFS_I(inode);
482 483 484 485 486 487 488
		/*
		 * 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.
		 */
489
		mpol_shared_policy_init(&info->policy, NULL);
L
Linus Torvalds 已提交
490 491 492 493 494 495 496 497 498 499 500 501 502
		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) */
503
			inc_nlink(inode);
L
Linus Torvalds 已提交
504 505 506 507 508
			break;
		case S_IFLNK:
			inode->i_op = &page_symlink_inode_operations;
			break;
		}
509
		lockdep_annotate_inode_mutex_key(inode);
L
Linus Torvalds 已提交
510 511 512 513 514 515 516 517
	}
	return inode;
}

/*
 * File creation. Allocate an inode, and we're done..
 */
static int hugetlbfs_mknod(struct inode *dir,
A
Al Viro 已提交
518
			struct dentry *dentry, umode_t mode, dev_t dev)
L
Linus Torvalds 已提交
519 520 521
{
	struct inode *inode;
	int error = -ENOSPC;
522 523

	inode = hugetlbfs_get_inode(dir->i_sb, dir, mode, dev);
L
Linus Torvalds 已提交
524 525 526 527 528 529 530 531 532
	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;
}

533
static int hugetlbfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
L
Linus Torvalds 已提交
534 535 536
{
	int retval = hugetlbfs_mknod(dir, dentry, mode | S_IFDIR, 0);
	if (!retval)
537
		inc_nlink(dir);
L
Linus Torvalds 已提交
538 539 540
	return retval;
}

A
Al Viro 已提交
541
static int hugetlbfs_create(struct inode *dir, struct dentry *dentry, umode_t mode, bool excl)
L
Linus Torvalds 已提交
542 543 544 545 546 547 548 549 550 551
{
	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;

552
	inode = hugetlbfs_get_inode(dir->i_sb, dir, S_IFLNK|S_IRWXUGO, 0);
L
Linus Torvalds 已提交
553 554 555 556 557 558 559 560 561 562 563 564 565 566 567
	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;
}

/*
568
 * mark the head page dirty
L
Linus Torvalds 已提交
569 570 571
 */
static int hugetlbfs_set_page_dirty(struct page *page)
{
572
	struct page *head = compound_head(page);
573 574

	SetPageDirty(head);
L
Linus Torvalds 已提交
575 576 577
	return 0;
}

N
Naoya Horiguchi 已提交
578
static int hugetlbfs_migrate_page(struct address_space *mapping,
579
				struct page *newpage, struct page *page,
580
				enum migrate_mode mode)
N
Naoya Horiguchi 已提交
581 582 583 584
{
	int rc;

	rc = migrate_huge_page_move_mapping(mapping, newpage, page);
585
	if (rc != MIGRATEPAGE_SUCCESS)
N
Naoya Horiguchi 已提交
586 587 588
		return rc;
	migrate_page_copy(newpage, page);

589
	return MIGRATEPAGE_SUCCESS;
N
Naoya Horiguchi 已提交
590 591
}

592
static int hugetlbfs_statfs(struct dentry *dentry, struct kstatfs *buf)
L
Linus Torvalds 已提交
593
{
594
	struct hugetlbfs_sb_info *sbinfo = HUGETLBFS_SB(dentry->d_sb);
595
	struct hstate *h = hstate_inode(dentry->d_inode);
L
Linus Torvalds 已提交
596 597

	buf->f_type = HUGETLBFS_MAGIC;
598
	buf->f_bsize = huge_page_size(h);
L
Linus Torvalds 已提交
599 600
	if (sbinfo) {
		spin_lock(&sbinfo->stat_lock);
601 602
		/* If no limits set, just report 0 for max/free/used
		 * blocks, like simple_statfs() */
603 604 605 606 607 608 609 610 611
		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);
612 613 614
			buf->f_files = sbinfo->max_inodes;
			buf->f_ffree = sbinfo->free_inodes;
		}
L
Linus Torvalds 已提交
615 616 617 618 619 620 621 622 623 624 625 626
		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;
627 628 629 630

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

L
Linus Torvalds 已提交
631 632 633 634
		kfree(sbi);
	}
}

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
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);
	}
}


660
static struct kmem_cache *hugetlbfs_inode_cachep;
L
Linus Torvalds 已提交
661 662 663

static struct inode *hugetlbfs_alloc_inode(struct super_block *sb)
{
664
	struct hugetlbfs_sb_info *sbinfo = HUGETLBFS_SB(sb);
L
Linus Torvalds 已提交
665 666
	struct hugetlbfs_inode_info *p;

667 668
	if (unlikely(!hugetlbfs_dec_free_inodes(sbinfo)))
		return NULL;
669
	p = kmem_cache_alloc(hugetlbfs_inode_cachep, GFP_KERNEL);
670 671
	if (unlikely(!p)) {
		hugetlbfs_inc_free_inodes(sbinfo);
L
Linus Torvalds 已提交
672
		return NULL;
673
	}
L
Linus Torvalds 已提交
674 675 676
	return &p->vfs_inode;
}

N
Nick Piggin 已提交
677 678 679 680 681 682
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
683 684
static void hugetlbfs_destroy_inode(struct inode *inode)
{
685
	hugetlbfs_inc_free_inodes(HUGETLBFS_SB(inode->i_sb));
L
Linus Torvalds 已提交
686
	mpol_free_shared_policy(&HUGETLBFS_I(inode)->policy);
N
Nick Piggin 已提交
687
	call_rcu(&inode->i_rcu, hugetlbfs_i_callback);
L
Linus Torvalds 已提交
688 689
}

690
static const struct address_space_operations hugetlbfs_aops = {
N
Nick Piggin 已提交
691 692
	.write_begin	= hugetlbfs_write_begin,
	.write_end	= hugetlbfs_write_end,
L
Linus Torvalds 已提交
693
	.set_page_dirty	= hugetlbfs_set_page_dirty,
N
Naoya Horiguchi 已提交
694
	.migratepage    = hugetlbfs_migrate_page,
L
Linus Torvalds 已提交
695 696
};

697

698
static void init_once(void *foo)
699 700 701
{
	struct hugetlbfs_inode_info *ei = (struct hugetlbfs_inode_info *)foo;

C
Christoph Lameter 已提交
702
	inode_init_once(&ei->vfs_inode);
703 704
}

705
const struct file_operations hugetlbfs_file_operations = {
B
Badari Pulavarty 已提交
706
	.read			= hugetlbfs_read,
L
Linus Torvalds 已提交
707
	.mmap			= hugetlbfs_file_mmap,
708
	.fsync			= noop_fsync,
L
Linus Torvalds 已提交
709
	.get_unmapped_area	= hugetlb_get_unmapped_area,
710
	.llseek		= default_llseek,
L
Linus Torvalds 已提交
711 712
};

713
static const struct inode_operations hugetlbfs_dir_inode_operations = {
L
Linus Torvalds 已提交
714 715 716 717 718 719 720 721 722 723 724 725
	.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,
};

726
static const struct inode_operations hugetlbfs_inode_operations = {
L
Linus Torvalds 已提交
727 728 729
	.setattr	= hugetlbfs_setattr,
};

730
static const struct super_operations hugetlbfs_ops = {
L
Linus Torvalds 已提交
731 732
	.alloc_inode    = hugetlbfs_alloc_inode,
	.destroy_inode  = hugetlbfs_destroy_inode,
A
Al Viro 已提交
733
	.evict_inode	= hugetlbfs_evict_inode,
L
Linus Torvalds 已提交
734 735
	.statfs		= hugetlbfs_statfs,
	.put_super	= hugetlbfs_put_super,
M
Miklos Szeredi 已提交
736
	.show_options	= generic_show_options,
L
Linus Torvalds 已提交
737 738 739 740 741
};

static int
hugetlbfs_parse_options(char *options, struct hugetlbfs_config *pconfig)
{
742 743 744
	char *p, *rest;
	substring_t args[MAX_OPT_ARGS];
	int option;
745 746
	unsigned long long size = 0;
	enum { NO_SIZE, SIZE_STD, SIZE_PERCENT } setsize = NO_SIZE;
L
Linus Torvalds 已提交
747 748 749 750

	if (!options)
		return 0;

751 752
	while ((p = strsep(&options, ",")) != NULL) {
		int token;
753 754
		if (!*p)
			continue;
755 756 757 758 759 760

		token = match_token(p, tokens, args);
		switch (token) {
		case Opt_uid:
			if (match_int(&args[0], &option))
 				goto bad_val;
761 762 763
			pconfig->uid = make_kuid(current_user_ns(), option);
			if (!uid_valid(pconfig->uid))
				goto bad_val;
764 765 766 767 768
			break;

		case Opt_gid:
			if (match_int(&args[0], &option))
 				goto bad_val;
769 770 771
			pconfig->gid = make_kgid(current_user_ns(), option);
			if (!gid_valid(pconfig->gid))
				goto bad_val;
772 773 774 775 776
			break;

		case Opt_mode:
			if (match_octal(&args[0], &option))
 				goto bad_val;
777
			pconfig->mode = option & 01777U;
778 779 780 781 782 783 784
			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);
785 786 787
			setsize = SIZE_STD;
			if (*rest == '%')
				setsize = SIZE_PERCENT;
788 789
			break;
		}
L
Linus Torvalds 已提交
790

791 792 793 794 795 796 797
		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;

798 799 800 801 802 803 804 805 806 807 808 809 810
		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;
		}

811
		default:
812 813 814
			printk(KERN_ERR "hugetlbfs: Bad mount option: \"%s\"\n",
				 p);
			return -EINVAL;
815 816
			break;
		}
L
Linus Torvalds 已提交
817
	}
818 819 820 821 822 823 824 825 826 827 828 829

	/* 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
830
	return 0;
831 832 833 834

bad_val:
 	printk(KERN_ERR "hugetlbfs: Bad value '%s' for mount option '%s'\n",
	       args[0].from, p);
835
 	return -EINVAL;
L
Linus Torvalds 已提交
836 837 838 839 840 841 842 843 844
}

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
845 846
	save_mount_options(sb, data);

L
Linus Torvalds 已提交
847 848
	config.nr_blocks = -1; /* No limit on size by default */
	config.nr_inodes = -1; /* No limit on number of inodes by default */
849 850
	config.uid = current_fsuid();
	config.gid = current_fsgid();
L
Linus Torvalds 已提交
851
	config.mode = 0755;
852
	config.hstate = &default_hstate;
L
Linus Torvalds 已提交
853 854 855 856 857 858 859 860
	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;
861
	sbinfo->hstate = config.hstate;
L
Linus Torvalds 已提交
862 863 864
	spin_lock_init(&sbinfo->stat_lock);
	sbinfo->max_inodes = config.nr_inodes;
	sbinfo->free_inodes = config.nr_inodes;
865 866 867 868 869 870
	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
871
	sb->s_maxbytes = MAX_LFS_FILESIZE;
872 873
	sb->s_blocksize = huge_page_size(config.hstate);
	sb->s_blocksize_bits = huge_page_shift(config.hstate);
L
Linus Torvalds 已提交
874 875 876
	sb->s_magic = HUGETLBFS_MAGIC;
	sb->s_op = &hugetlbfs_ops;
	sb->s_time_gran = 1;
877 878
	sb->s_root = d_make_root(hugetlbfs_get_root(sb, &config));
	if (!sb->s_root)
L
Linus Torvalds 已提交
879 880 881
		goto out_free;
	return 0;
out_free:
882 883
	if (sbinfo->spool)
		kfree(sbinfo->spool);
L
Linus Torvalds 已提交
884 885 886 887
	kfree(sbinfo);
	return -ENOMEM;
}

A
Al Viro 已提交
888 889
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
890
{
A
Al Viro 已提交
891
	return mount_nodev(fs_type, flags, data, hugetlbfs_fill_super);
L
Linus Torvalds 已提交
892 893 894 895
}

static struct file_system_type hugetlbfs_fs_type = {
	.name		= "hugetlbfs",
A
Al Viro 已提交
896
	.mount		= hugetlbfs_mount,
L
Linus Torvalds 已提交
897 898 899
	.kill_sb	= kill_litter_super,
};

900
static struct vfsmount *hugetlbfs_vfsmount[HUGE_MAX_HSTATE];
L
Linus Torvalds 已提交
901

902
static int can_do_hugetlb_shm(void)
L
Linus Torvalds 已提交
903
{
904 905 906
	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
907 908
}

909 910 911 912 913 914 915 916 917 918 919 920
static int get_hstate_idx(int page_size_log)
{
	struct hstate *h;

	if (!page_size_log)
		return default_hstate_idx;
	h = size_to_hstate(1 << page_size_log);
	if (!h)
		return -1;
	return h - hstates;
}

921 922
struct file *hugetlb_file_setup(const char *name, unsigned long addr,
				size_t size, vm_flags_t acctflag,
923 924
				struct user_struct **user,
				int creat_flags, int page_size_log)
L
Linus Torvalds 已提交
925 926 927 928
{
	int error = -ENOMEM;
	struct file *file;
	struct inode *inode;
929 930
	struct path path;
	struct dentry *root;
L
Linus Torvalds 已提交
931
	struct qstr quick_string;
932 933
	struct hstate *hstate;
	unsigned long num_pages;
934 935 936 937 938
	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
939

940
	*user = NULL;
941
	if (!hugetlbfs_vfsmount[hstate_idx])
942 943
		return ERR_PTR(-ENOENT);

944
	if (creat_flags == HUGETLB_SHMFS_INODE && !can_do_hugetlb_shm()) {
945 946
		*user = current_user();
		if (user_shm_lock(size, *user)) {
947 948 949 950 951
			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);
952 953
		} else {
			*user = NULL;
954
			return ERR_PTR(-EPERM);
955
		}
956
	}
L
Linus Torvalds 已提交
957

958
	root = hugetlbfs_vfsmount[hstate_idx]->mnt_root;
959
	quick_string.name = name;
L
Linus Torvalds 已提交
960 961
	quick_string.len = strlen(quick_string.name);
	quick_string.hash = 0;
962 963
	path.dentry = d_alloc(root, &quick_string);
	if (!path.dentry)
L
Linus Torvalds 已提交
964 965
		goto out_shm_unlock;

966
	path.mnt = mntget(hugetlbfs_vfsmount[hstate_idx]);
L
Linus Torvalds 已提交
967
	error = -ENOSPC;
968
	inode = hugetlbfs_get_inode(root->d_sb, NULL, S_IFREG | S_IRWXUGO, 0);
L
Linus Torvalds 已提交
969
	if (!inode)
970
		goto out_dentry;
L
Linus Torvalds 已提交
971

972 973 974 975
	hstate = hstate_inode(inode);
	size += addr & ~huge_page_mask(hstate);
	num_pages = ALIGN(size, huge_page_size(hstate)) >>
			huge_page_shift(hstate);
976
	error = -ENOMEM;
977
	if (hugetlb_reserve_pages(inode, 0, num_pages, NULL, acctflag))
978 979
		goto out_inode;

980
	d_instantiate(path.dentry, inode);
L
Linus Torvalds 已提交
981
	inode->i_size = size;
982
	clear_nlink(inode);
983 984

	error = -ENFILE;
985
	file = alloc_file(&path, FMODE_WRITE | FMODE_READ,
986 987
			&hugetlbfs_file_operations);
	if (!file)
988
		goto out_dentry; /* inode is already attached */
989

L
Linus Torvalds 已提交
990 991
	return file;

992 993
out_inode:
	iput(inode);
L
Linus Torvalds 已提交
994
out_dentry:
995
	path_put(&path);
L
Linus Torvalds 已提交
996
out_shm_unlock:
997 998 999 1000
	if (*user) {
		user_shm_unlock(size, *user);
		*user = NULL;
	}
L
Linus Torvalds 已提交
1001 1002 1003 1004 1005
	return ERR_PTR(error);
}

static int __init init_hugetlbfs_fs(void)
{
1006
	struct hstate *h;
L
Linus Torvalds 已提交
1007
	int error;
1008
	int i;
L
Linus Torvalds 已提交
1009

P
Peter Zijlstra 已提交
1010 1011 1012 1013
	error = bdi_init(&hugetlbfs_backing_dev_info);
	if (error)
		return error;

1014
	error = -ENOMEM;
L
Linus Torvalds 已提交
1015 1016
	hugetlbfs_inode_cachep = kmem_cache_create("hugetlbfs_inode_cache",
					sizeof(struct hugetlbfs_inode_info),
1017
					0, 0, init_once);
L
Linus Torvalds 已提交
1018
	if (hugetlbfs_inode_cachep == NULL)
P
Peter Zijlstra 已提交
1019
		goto out2;
L
Linus Torvalds 已提交
1020 1021 1022 1023 1024

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

1025 1026 1027 1028
	i = 0;
	for_each_hstate(h) {
		char buf[50];
		unsigned ps_kb = 1U << (h->order + PAGE_SHIFT - 10);
L
Linus Torvalds 已提交
1029

1030 1031 1032
		snprintf(buf, sizeof(buf), "pagesize=%uK", ps_kb);
		hugetlbfs_vfsmount[i] = kern_mount_data(&hugetlbfs_fs_type,
							buf);
L
Linus Torvalds 已提交
1033

1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044
		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
1045 1046

 out:
1047
	kmem_cache_destroy(hugetlbfs_inode_cachep);
P
Peter Zijlstra 已提交
1048 1049
 out2:
	bdi_destroy(&hugetlbfs_backing_dev_info);
L
Linus Torvalds 已提交
1050 1051 1052 1053 1054
	return error;
}

static void __exit exit_hugetlbfs_fs(void)
{
1055 1056 1057 1058
	struct hstate *h;
	int i;


1059 1060 1061 1062 1063
	/*
	 * Make sure all delayed rcu free inodes are flushed before we
	 * destroy cache.
	 */
	rcu_barrier();
L
Linus Torvalds 已提交
1064
	kmem_cache_destroy(hugetlbfs_inode_cachep);
1065 1066 1067
	i = 0;
	for_each_hstate(h)
		kern_unmount(hugetlbfs_vfsmount[i++]);
L
Linus Torvalds 已提交
1068
	unregister_filesystem(&hugetlbfs_fs_type);
P
Peter Zijlstra 已提交
1069
	bdi_destroy(&hugetlbfs_backing_dev_info);
L
Linus Torvalds 已提交
1070 1071 1072 1073 1074 1075
}

module_init(init_hugetlbfs_fs)
module_exit(exit_hugetlbfs_fs)

MODULE_LICENSE("GPL");