]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * hugetlbpage-backed filesystem. Based on ramfs. | |
3 | * | |
4 | * William Irwin, 2002 | |
5 | * | |
6 | * Copyright (C) 2002 Linus Torvalds. | |
7 | */ | |
8 | ||
9 | #include <linux/module.h> | |
10 | #include <linux/thread_info.h> | |
11 | #include <asm/current.h> | |
12 | #include <linux/sched.h> /* remove ASAP */ | |
13 | #include <linux/fs.h> | |
14 | #include <linux/mount.h> | |
15 | #include <linux/file.h> | |
e73a75fa | 16 | #include <linux/kernel.h> |
1da177e4 LT |
17 | #include <linux/writeback.h> |
18 | #include <linux/pagemap.h> | |
19 | #include <linux/highmem.h> | |
20 | #include <linux/init.h> | |
21 | #include <linux/string.h> | |
16f7e0fe | 22 | #include <linux/capability.h> |
e73a75fa | 23 | #include <linux/ctype.h> |
1da177e4 LT |
24 | #include <linux/backing-dev.h> |
25 | #include <linux/hugetlb.h> | |
26 | #include <linux/pagevec.h> | |
e73a75fa | 27 | #include <linux/parser.h> |
036e0856 | 28 | #include <linux/mman.h> |
1da177e4 LT |
29 | #include <linux/slab.h> |
30 | #include <linux/dnotify.h> | |
31 | #include <linux/statfs.h> | |
32 | #include <linux/security.h> | |
1fd7317d | 33 | #include <linux/magic.h> |
290408d4 | 34 | #include <linux/migrate.h> |
1da177e4 LT |
35 | |
36 | #include <asm/uaccess.h> | |
37 | ||
ee9b6d61 | 38 | static const struct super_operations hugetlbfs_ops; |
f5e54d6e | 39 | static const struct address_space_operations hugetlbfs_aops; |
4b6f5d20 | 40 | const struct file_operations hugetlbfs_file_operations; |
92e1d5be AV |
41 | static const struct inode_operations hugetlbfs_dir_inode_operations; |
42 | static const struct inode_operations hugetlbfs_inode_operations; | |
1da177e4 LT |
43 | |
44 | static struct backing_dev_info hugetlbfs_backing_dev_info = { | |
d993831f | 45 | .name = "hugetlbfs", |
1da177e4 | 46 | .ra_pages = 0, /* No readahead */ |
e4ad08fe | 47 | .capabilities = BDI_CAP_NO_ACCT_AND_WRITEBACK, |
1da177e4 LT |
48 | }; |
49 | ||
50 | int sysctl_hugetlb_shm_group; | |
51 | ||
e73a75fa RD |
52 | enum { |
53 | Opt_size, Opt_nr_inodes, | |
54 | Opt_mode, Opt_uid, Opt_gid, | |
a137e1cc | 55 | Opt_pagesize, |
e73a75fa RD |
56 | Opt_err, |
57 | }; | |
58 | ||
a447c093 | 59 | static const match_table_t tokens = { |
e73a75fa RD |
60 | {Opt_size, "size=%s"}, |
61 | {Opt_nr_inodes, "nr_inodes=%s"}, | |
62 | {Opt_mode, "mode=%o"}, | |
63 | {Opt_uid, "uid=%u"}, | |
64 | {Opt_gid, "gid=%u"}, | |
a137e1cc | 65 | {Opt_pagesize, "pagesize=%s"}, |
e73a75fa RD |
66 | {Opt_err, NULL}, |
67 | }; | |
68 | ||
2e9b367c AL |
69 | static void huge_pagevec_release(struct pagevec *pvec) |
70 | { | |
71 | int i; | |
72 | ||
73 | for (i = 0; i < pagevec_count(pvec); ++i) | |
74 | put_page(pvec->pages[i]); | |
75 | ||
76 | pagevec_reinit(pvec); | |
77 | } | |
78 | ||
1da177e4 LT |
79 | static int hugetlbfs_file_mmap(struct file *file, struct vm_area_struct *vma) |
80 | { | |
b39424e2 | 81 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
82 | loff_t len, vma_len; |
83 | int ret; | |
a5516438 | 84 | struct hstate *h = hstate_file(file); |
1da177e4 | 85 | |
68589bc3 | 86 | /* |
dec4ad86 DG |
87 | * vma address alignment (but not the pgoff alignment) has |
88 | * already been checked by prepare_hugepage_range. If you add | |
89 | * any error returns here, do so after setting VM_HUGETLB, so | |
90 | * is_vm_hugetlb_page tests below unmap_region go the right | |
91 | * way when do_mmap_pgoff unwinds (may be important on powerpc | |
92 | * and ia64). | |
68589bc3 HD |
93 | */ |
94 | vma->vm_flags |= VM_HUGETLB | VM_RESERVED; | |
95 | vma->vm_ops = &hugetlb_vm_ops; | |
1da177e4 | 96 | |
2b37c35e | 97 | if (vma->vm_pgoff & (~huge_page_mask(h) >> PAGE_SHIFT)) |
dec4ad86 DG |
98 | return -EINVAL; |
99 | ||
1da177e4 LT |
100 | vma_len = (loff_t)(vma->vm_end - vma->vm_start); |
101 | ||
1b1dcc1b | 102 | mutex_lock(&inode->i_mutex); |
1da177e4 | 103 | file_accessed(file); |
1da177e4 LT |
104 | |
105 | ret = -ENOMEM; | |
106 | len = vma_len + ((loff_t)vma->vm_pgoff << PAGE_SHIFT); | |
1da177e4 | 107 | |
a1e78772 | 108 | if (hugetlb_reserve_pages(inode, |
a5516438 | 109 | vma->vm_pgoff >> huge_page_order(h), |
5a6fe125 MG |
110 | len >> huge_page_shift(h), vma, |
111 | vma->vm_flags)) | |
a43a8c39 | 112 | goto out; |
b45b5bd6 | 113 | |
4c887265 AL |
114 | ret = 0; |
115 | hugetlb_prefault_arch_hook(vma->vm_mm); | |
b6174df5 | 116 | if (vma->vm_flags & VM_WRITE && inode->i_size < len) |
1da177e4 LT |
117 | inode->i_size = len; |
118 | out: | |
1b1dcc1b | 119 | mutex_unlock(&inode->i_mutex); |
1da177e4 LT |
120 | |
121 | return ret; | |
122 | } | |
123 | ||
124 | /* | |
508034a3 | 125 | * Called under down_write(mmap_sem). |
1da177e4 LT |
126 | */ |
127 | ||
d2ba27e8 | 128 | #ifndef HAVE_ARCH_HUGETLB_UNMAPPED_AREA |
1da177e4 LT |
129 | static unsigned long |
130 | hugetlb_get_unmapped_area(struct file *file, unsigned long addr, | |
131 | unsigned long len, unsigned long pgoff, unsigned long flags) | |
132 | { | |
133 | struct mm_struct *mm = current->mm; | |
134 | struct vm_area_struct *vma; | |
135 | unsigned long start_addr; | |
a5516438 | 136 | struct hstate *h = hstate_file(file); |
1da177e4 | 137 | |
a5516438 | 138 | if (len & ~huge_page_mask(h)) |
1da177e4 LT |
139 | return -EINVAL; |
140 | if (len > TASK_SIZE) | |
141 | return -ENOMEM; | |
142 | ||
036e0856 | 143 | if (flags & MAP_FIXED) { |
a5516438 | 144 | if (prepare_hugepage_range(file, addr, len)) |
036e0856 BH |
145 | return -EINVAL; |
146 | return addr; | |
147 | } | |
148 | ||
1da177e4 | 149 | if (addr) { |
a5516438 | 150 | addr = ALIGN(addr, huge_page_size(h)); |
1da177e4 LT |
151 | vma = find_vma(mm, addr); |
152 | if (TASK_SIZE - len >= addr && | |
153 | (!vma || addr + len <= vma->vm_start)) | |
154 | return addr; | |
155 | } | |
156 | ||
4bfc130d XG |
157 | if (len > mm->cached_hole_size) |
158 | start_addr = mm->free_area_cache; | |
159 | else { | |
1363c3cd | 160 | start_addr = TASK_UNMAPPED_BASE; |
4bfc130d XG |
161 | mm->cached_hole_size = 0; |
162 | } | |
1363c3cd | 163 | |
1da177e4 | 164 | full_search: |
a5516438 | 165 | addr = ALIGN(start_addr, huge_page_size(h)); |
1da177e4 LT |
166 | |
167 | for (vma = find_vma(mm, addr); ; vma = vma->vm_next) { | |
168 | /* At this point: (!vma || addr < vma->vm_end). */ | |
169 | if (TASK_SIZE - len < addr) { | |
170 | /* | |
171 | * Start a new search - just in case we missed | |
172 | * some holes. | |
173 | */ | |
174 | if (start_addr != TASK_UNMAPPED_BASE) { | |
175 | start_addr = TASK_UNMAPPED_BASE; | |
4bfc130d | 176 | mm->cached_hole_size = 0; |
1da177e4 LT |
177 | goto full_search; |
178 | } | |
179 | return -ENOMEM; | |
180 | } | |
181 | ||
4bfc130d XG |
182 | if (!vma || addr + len <= vma->vm_start) { |
183 | mm->free_area_cache = addr + len; | |
1da177e4 | 184 | return addr; |
4bfc130d XG |
185 | } |
186 | if (addr + mm->cached_hole_size < vma->vm_start) | |
187 | mm->cached_hole_size = vma->vm_start - addr; | |
a5516438 | 188 | addr = ALIGN(vma->vm_end, huge_page_size(h)); |
1da177e4 LT |
189 | } |
190 | } | |
191 | #endif | |
192 | ||
e63e1e5a BP |
193 | static int |
194 | hugetlbfs_read_actor(struct page *page, unsigned long offset, | |
195 | char __user *buf, unsigned long count, | |
196 | unsigned long size) | |
197 | { | |
198 | char *kaddr; | |
199 | unsigned long left, copied = 0; | |
200 | int i, chunksize; | |
201 | ||
202 | if (size > count) | |
203 | size = count; | |
204 | ||
205 | /* Find which 4k chunk and offset with in that chunk */ | |
206 | i = offset >> PAGE_CACHE_SHIFT; | |
207 | offset = offset & ~PAGE_CACHE_MASK; | |
208 | ||
209 | while (size) { | |
210 | chunksize = PAGE_CACHE_SIZE; | |
211 | if (offset) | |
212 | chunksize -= offset; | |
213 | if (chunksize > size) | |
214 | chunksize = size; | |
215 | kaddr = kmap(&page[i]); | |
216 | left = __copy_to_user(buf, kaddr + offset, chunksize); | |
217 | kunmap(&page[i]); | |
218 | if (left) { | |
219 | copied += (chunksize - left); | |
220 | break; | |
221 | } | |
222 | offset = 0; | |
223 | size -= chunksize; | |
224 | buf += chunksize; | |
225 | copied += chunksize; | |
226 | i++; | |
227 | } | |
228 | return copied ? copied : -EFAULT; | |
229 | } | |
230 | ||
231 | /* | |
232 | * Support for read() - Find the page attached to f_mapping and copy out the | |
233 | * data. Its *very* similar to do_generic_mapping_read(), we can't use that | |
234 | * since it has PAGE_CACHE_SIZE assumptions. | |
235 | */ | |
236 | static ssize_t hugetlbfs_read(struct file *filp, char __user *buf, | |
237 | size_t len, loff_t *ppos) | |
238 | { | |
a5516438 | 239 | struct hstate *h = hstate_file(filp); |
e63e1e5a BP |
240 | struct address_space *mapping = filp->f_mapping; |
241 | struct inode *inode = mapping->host; | |
a5516438 AK |
242 | unsigned long index = *ppos >> huge_page_shift(h); |
243 | unsigned long offset = *ppos & ~huge_page_mask(h); | |
e63e1e5a BP |
244 | unsigned long end_index; |
245 | loff_t isize; | |
246 | ssize_t retval = 0; | |
247 | ||
248 | mutex_lock(&inode->i_mutex); | |
249 | ||
250 | /* validate length */ | |
251 | if (len == 0) | |
252 | goto out; | |
253 | ||
254 | isize = i_size_read(inode); | |
255 | if (!isize) | |
256 | goto out; | |
257 | ||
a5516438 | 258 | end_index = (isize - 1) >> huge_page_shift(h); |
e63e1e5a BP |
259 | for (;;) { |
260 | struct page *page; | |
a5516438 | 261 | unsigned long nr, ret; |
91bf189c | 262 | int ra; |
e63e1e5a BP |
263 | |
264 | /* nr is the maximum number of bytes to copy from this page */ | |
a5516438 | 265 | nr = huge_page_size(h); |
e63e1e5a BP |
266 | if (index >= end_index) { |
267 | if (index > end_index) | |
268 | goto out; | |
a5516438 | 269 | nr = ((isize - 1) & ~huge_page_mask(h)) + 1; |
e63e1e5a BP |
270 | if (nr <= offset) { |
271 | goto out; | |
272 | } | |
273 | } | |
274 | nr = nr - offset; | |
275 | ||
276 | /* Find the page */ | |
277 | page = find_get_page(mapping, index); | |
278 | if (unlikely(page == NULL)) { | |
279 | /* | |
280 | * We have a HOLE, zero out the user-buffer for the | |
281 | * length of the hole or request. | |
282 | */ | |
283 | ret = len < nr ? len : nr; | |
284 | if (clear_user(buf, ret)) | |
91bf189c RK |
285 | ra = -EFAULT; |
286 | else | |
287 | ra = 0; | |
e63e1e5a BP |
288 | } else { |
289 | /* | |
290 | * We have the page, copy it to user space buffer. | |
291 | */ | |
91bf189c RK |
292 | ra = hugetlbfs_read_actor(page, offset, buf, len, nr); |
293 | ret = ra; | |
e63e1e5a | 294 | } |
91bf189c | 295 | if (ra < 0) { |
e63e1e5a | 296 | if (retval == 0) |
91bf189c | 297 | retval = ra; |
e63e1e5a BP |
298 | if (page) |
299 | page_cache_release(page); | |
300 | goto out; | |
301 | } | |
302 | ||
303 | offset += ret; | |
304 | retval += ret; | |
305 | len -= ret; | |
a5516438 AK |
306 | index += offset >> huge_page_shift(h); |
307 | offset &= ~huge_page_mask(h); | |
e63e1e5a BP |
308 | |
309 | if (page) | |
310 | page_cache_release(page); | |
311 | ||
312 | /* short read or no more work */ | |
313 | if ((ret != nr) || (len == 0)) | |
314 | break; | |
315 | } | |
316 | out: | |
a5516438 | 317 | *ppos = ((loff_t)index << huge_page_shift(h)) + offset; |
e63e1e5a BP |
318 | mutex_unlock(&inode->i_mutex); |
319 | return retval; | |
320 | } | |
321 | ||
800d15a5 NP |
322 | static int hugetlbfs_write_begin(struct file *file, |
323 | struct address_space *mapping, | |
324 | loff_t pos, unsigned len, unsigned flags, | |
325 | struct page **pagep, void **fsdata) | |
1da177e4 LT |
326 | { |
327 | return -EINVAL; | |
328 | } | |
329 | ||
800d15a5 NP |
330 | static int hugetlbfs_write_end(struct file *file, struct address_space *mapping, |
331 | loff_t pos, unsigned len, unsigned copied, | |
332 | struct page *page, void *fsdata) | |
1da177e4 | 333 | { |
800d15a5 | 334 | BUG(); |
1da177e4 LT |
335 | return -EINVAL; |
336 | } | |
337 | ||
1da177e4 LT |
338 | static void truncate_huge_page(struct page *page) |
339 | { | |
fba2591b | 340 | cancel_dirty_page(page, /* No IO accounting for huge pages? */0); |
1da177e4 | 341 | ClearPageUptodate(page); |
bd65cb86 | 342 | delete_from_page_cache(page); |
1da177e4 LT |
343 | } |
344 | ||
b45b5bd6 | 345 | static void truncate_hugepages(struct inode *inode, loff_t lstart) |
1da177e4 | 346 | { |
a5516438 | 347 | struct hstate *h = hstate_inode(inode); |
b45b5bd6 | 348 | struct address_space *mapping = &inode->i_data; |
a5516438 | 349 | const pgoff_t start = lstart >> huge_page_shift(h); |
1da177e4 LT |
350 | struct pagevec pvec; |
351 | pgoff_t next; | |
a43a8c39 | 352 | int i, freed = 0; |
1da177e4 LT |
353 | |
354 | pagevec_init(&pvec, 0); | |
355 | next = start; | |
356 | while (1) { | |
357 | if (!pagevec_lookup(&pvec, mapping, next, PAGEVEC_SIZE)) { | |
358 | if (next == start) | |
359 | break; | |
360 | next = start; | |
361 | continue; | |
362 | } | |
363 | ||
364 | for (i = 0; i < pagevec_count(&pvec); ++i) { | |
365 | struct page *page = pvec.pages[i]; | |
366 | ||
367 | lock_page(page); | |
368 | if (page->index > next) | |
369 | next = page->index; | |
370 | ++next; | |
371 | truncate_huge_page(page); | |
372 | unlock_page(page); | |
a43a8c39 | 373 | freed++; |
1da177e4 LT |
374 | } |
375 | huge_pagevec_release(&pvec); | |
376 | } | |
377 | BUG_ON(!lstart && mapping->nrpages); | |
a43a8c39 | 378 | hugetlb_unreserve_pages(inode, start, freed); |
1da177e4 LT |
379 | } |
380 | ||
2bbbda30 | 381 | static void hugetlbfs_evict_inode(struct inode *inode) |
1da177e4 | 382 | { |
b45b5bd6 | 383 | truncate_hugepages(inode, 0); |
b0683aa6 | 384 | end_writeback(inode); |
149f4211 CH |
385 | } |
386 | ||
1da177e4 | 387 | static inline void |
856fc295 | 388 | hugetlb_vmtruncate_list(struct prio_tree_root *root, pgoff_t pgoff) |
1da177e4 LT |
389 | { |
390 | struct vm_area_struct *vma; | |
391 | struct prio_tree_iter iter; | |
392 | ||
856fc295 | 393 | vma_prio_tree_foreach(vma, &iter, root, pgoff, ULONG_MAX) { |
1da177e4 LT |
394 | unsigned long v_offset; |
395 | ||
1da177e4 | 396 | /* |
856fc295 HD |
397 | * Can the expression below overflow on 32-bit arches? |
398 | * No, because the prio_tree returns us only those vmas | |
399 | * which overlap the truncated area starting at pgoff, | |
400 | * and no vma on a 32-bit arch can span beyond the 4GB. | |
1da177e4 | 401 | */ |
856fc295 HD |
402 | if (vma->vm_pgoff < pgoff) |
403 | v_offset = (pgoff - vma->vm_pgoff) << PAGE_SHIFT; | |
404 | else | |
1da177e4 LT |
405 | v_offset = 0; |
406 | ||
502717f4 | 407 | __unmap_hugepage_range(vma, |
04f2cbe3 | 408 | vma->vm_start + v_offset, vma->vm_end, NULL); |
1da177e4 LT |
409 | } |
410 | } | |
411 | ||
1da177e4 LT |
412 | static int hugetlb_vmtruncate(struct inode *inode, loff_t offset) |
413 | { | |
856fc295 | 414 | pgoff_t pgoff; |
1da177e4 | 415 | struct address_space *mapping = inode->i_mapping; |
a5516438 | 416 | struct hstate *h = hstate_inode(inode); |
1da177e4 | 417 | |
a5516438 | 418 | BUG_ON(offset & ~huge_page_mask(h)); |
856fc295 | 419 | pgoff = offset >> PAGE_SHIFT; |
1da177e4 | 420 | |
7aa91e10 | 421 | i_size_write(inode, offset); |
3d48ae45 | 422 | mutex_lock(&mapping->i_mmap_mutex); |
1da177e4 LT |
423 | if (!prio_tree_empty(&mapping->i_mmap)) |
424 | hugetlb_vmtruncate_list(&mapping->i_mmap, pgoff); | |
3d48ae45 | 425 | mutex_unlock(&mapping->i_mmap_mutex); |
b45b5bd6 | 426 | truncate_hugepages(inode, offset); |
1da177e4 LT |
427 | return 0; |
428 | } | |
429 | ||
430 | static int hugetlbfs_setattr(struct dentry *dentry, struct iattr *attr) | |
431 | { | |
432 | struct inode *inode = dentry->d_inode; | |
a5516438 | 433 | struct hstate *h = hstate_inode(inode); |
1da177e4 LT |
434 | int error; |
435 | unsigned int ia_valid = attr->ia_valid; | |
436 | ||
437 | BUG_ON(!inode); | |
438 | ||
439 | error = inode_change_ok(inode, attr); | |
440 | if (error) | |
1025774c | 441 | return error; |
1da177e4 LT |
442 | |
443 | if (ia_valid & ATTR_SIZE) { | |
444 | error = -EINVAL; | |
1025774c CH |
445 | if (attr->ia_size & ~huge_page_mask(h)) |
446 | return -EINVAL; | |
447 | error = hugetlb_vmtruncate(inode, attr->ia_size); | |
1da177e4 | 448 | if (error) |
1025774c | 449 | return error; |
1da177e4 | 450 | } |
1025774c CH |
451 | |
452 | setattr_copy(inode, attr); | |
453 | mark_inode_dirty(inode); | |
454 | return 0; | |
1da177e4 LT |
455 | } |
456 | ||
7d54fa64 AV |
457 | static struct inode *hugetlbfs_get_root(struct super_block *sb, |
458 | struct hugetlbfs_config *config) | |
1da177e4 LT |
459 | { |
460 | struct inode *inode; | |
1da177e4 LT |
461 | |
462 | inode = new_inode(sb); | |
463 | if (inode) { | |
464 | struct hugetlbfs_inode_info *info; | |
85fe4025 | 465 | inode->i_ino = get_next_ino(); |
7d54fa64 AV |
466 | inode->i_mode = S_IFDIR | config->mode; |
467 | inode->i_uid = config->uid; | |
468 | inode->i_gid = config->gid; | |
469 | inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME; | |
470 | info = HUGETLBFS_I(inode); | |
471 | mpol_shared_policy_init(&info->policy, NULL); | |
472 | inode->i_op = &hugetlbfs_dir_inode_operations; | |
473 | inode->i_fop = &simple_dir_operations; | |
474 | /* directory inodes start off with i_nlink == 2 (for "." entry) */ | |
475 | inc_nlink(inode); | |
476 | } | |
477 | return inode; | |
478 | } | |
479 | ||
480 | static struct inode *hugetlbfs_get_inode(struct super_block *sb, | |
481 | struct inode *dir, | |
18df2252 | 482 | umode_t mode, dev_t dev) |
7d54fa64 AV |
483 | { |
484 | struct inode *inode; | |
485 | ||
486 | inode = new_inode(sb); | |
487 | if (inode) { | |
488 | struct hugetlbfs_inode_info *info; | |
489 | inode->i_ino = get_next_ino(); | |
490 | inode_init_owner(inode, dir, mode); | |
1da177e4 LT |
491 | inode->i_mapping->a_ops = &hugetlbfs_aops; |
492 | inode->i_mapping->backing_dev_info =&hugetlbfs_backing_dev_info; | |
493 | inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME; | |
a43a8c39 | 494 | INIT_LIST_HEAD(&inode->i_mapping->private_list); |
1da177e4 | 495 | info = HUGETLBFS_I(inode); |
6bfde05b EM |
496 | /* |
497 | * The policy is initialized here even if we are creating a | |
498 | * private inode because initialization simply creates an | |
499 | * an empty rb tree and calls spin_lock_init(), later when we | |
500 | * call mpol_free_shared_policy() it will just return because | |
501 | * the rb tree will still be empty. | |
502 | */ | |
71fe804b | 503 | mpol_shared_policy_init(&info->policy, NULL); |
1da177e4 LT |
504 | switch (mode & S_IFMT) { |
505 | default: | |
506 | init_special_inode(inode, mode, dev); | |
507 | break; | |
508 | case S_IFREG: | |
509 | inode->i_op = &hugetlbfs_inode_operations; | |
510 | inode->i_fop = &hugetlbfs_file_operations; | |
511 | break; | |
512 | case S_IFDIR: | |
513 | inode->i_op = &hugetlbfs_dir_inode_operations; | |
514 | inode->i_fop = &simple_dir_operations; | |
515 | ||
516 | /* directory inodes start off with i_nlink == 2 (for "." entry) */ | |
d8c76e6f | 517 | inc_nlink(inode); |
1da177e4 LT |
518 | break; |
519 | case S_IFLNK: | |
520 | inode->i_op = &page_symlink_inode_operations; | |
521 | break; | |
522 | } | |
e096d0c7 | 523 | lockdep_annotate_inode_mutex_key(inode); |
1da177e4 LT |
524 | } |
525 | return inode; | |
526 | } | |
527 | ||
528 | /* | |
529 | * File creation. Allocate an inode, and we're done.. | |
530 | */ | |
531 | static int hugetlbfs_mknod(struct inode *dir, | |
1a67aafb | 532 | struct dentry *dentry, umode_t mode, dev_t dev) |
1da177e4 LT |
533 | { |
534 | struct inode *inode; | |
535 | int error = -ENOSPC; | |
7d54fa64 AV |
536 | |
537 | inode = hugetlbfs_get_inode(dir->i_sb, dir, mode, dev); | |
1da177e4 LT |
538 | if (inode) { |
539 | dir->i_ctime = dir->i_mtime = CURRENT_TIME; | |
540 | d_instantiate(dentry, inode); | |
541 | dget(dentry); /* Extra count - pin the dentry in core */ | |
542 | error = 0; | |
543 | } | |
544 | return error; | |
545 | } | |
546 | ||
18bb1db3 | 547 | static int hugetlbfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode) |
1da177e4 LT |
548 | { |
549 | int retval = hugetlbfs_mknod(dir, dentry, mode | S_IFDIR, 0); | |
550 | if (!retval) | |
d8c76e6f | 551 | inc_nlink(dir); |
1da177e4 LT |
552 | return retval; |
553 | } | |
554 | ||
4acdaf27 | 555 | static int hugetlbfs_create(struct inode *dir, struct dentry *dentry, umode_t mode, struct nameidata *nd) |
1da177e4 LT |
556 | { |
557 | return hugetlbfs_mknod(dir, dentry, mode | S_IFREG, 0); | |
558 | } | |
559 | ||
560 | static int hugetlbfs_symlink(struct inode *dir, | |
561 | struct dentry *dentry, const char *symname) | |
562 | { | |
563 | struct inode *inode; | |
564 | int error = -ENOSPC; | |
1da177e4 | 565 | |
7d54fa64 | 566 | inode = hugetlbfs_get_inode(dir->i_sb, dir, S_IFLNK|S_IRWXUGO, 0); |
1da177e4 LT |
567 | if (inode) { |
568 | int l = strlen(symname)+1; | |
569 | error = page_symlink(inode, symname, l); | |
570 | if (!error) { | |
571 | d_instantiate(dentry, inode); | |
572 | dget(dentry); | |
573 | } else | |
574 | iput(inode); | |
575 | } | |
576 | dir->i_ctime = dir->i_mtime = CURRENT_TIME; | |
577 | ||
578 | return error; | |
579 | } | |
580 | ||
581 | /* | |
6649a386 | 582 | * mark the head page dirty |
1da177e4 LT |
583 | */ |
584 | static int hugetlbfs_set_page_dirty(struct page *page) | |
585 | { | |
d85f3385 | 586 | struct page *head = compound_head(page); |
6649a386 KC |
587 | |
588 | SetPageDirty(head); | |
1da177e4 LT |
589 | return 0; |
590 | } | |
591 | ||
290408d4 | 592 | static int hugetlbfs_migrate_page(struct address_space *mapping, |
b969c4ab | 593 | struct page *newpage, struct page *page, |
a6bc32b8 | 594 | enum migrate_mode mode) |
290408d4 NH |
595 | { |
596 | int rc; | |
597 | ||
598 | rc = migrate_huge_page_move_mapping(mapping, newpage, page); | |
599 | if (rc) | |
600 | return rc; | |
601 | migrate_page_copy(newpage, page); | |
602 | ||
603 | return 0; | |
604 | } | |
605 | ||
726c3342 | 606 | static int hugetlbfs_statfs(struct dentry *dentry, struct kstatfs *buf) |
1da177e4 | 607 | { |
726c3342 | 608 | struct hugetlbfs_sb_info *sbinfo = HUGETLBFS_SB(dentry->d_sb); |
a5516438 | 609 | struct hstate *h = hstate_inode(dentry->d_inode); |
1da177e4 LT |
610 | |
611 | buf->f_type = HUGETLBFS_MAGIC; | |
a5516438 | 612 | buf->f_bsize = huge_page_size(h); |
1da177e4 LT |
613 | if (sbinfo) { |
614 | spin_lock(&sbinfo->stat_lock); | |
74a8a65c DG |
615 | /* If no limits set, just report 0 for max/free/used |
616 | * blocks, like simple_statfs() */ | |
617 | if (sbinfo->max_blocks >= 0) { | |
618 | buf->f_blocks = sbinfo->max_blocks; | |
619 | buf->f_bavail = buf->f_bfree = sbinfo->free_blocks; | |
620 | buf->f_files = sbinfo->max_inodes; | |
621 | buf->f_ffree = sbinfo->free_inodes; | |
622 | } | |
1da177e4 LT |
623 | spin_unlock(&sbinfo->stat_lock); |
624 | } | |
625 | buf->f_namelen = NAME_MAX; | |
626 | return 0; | |
627 | } | |
628 | ||
629 | static void hugetlbfs_put_super(struct super_block *sb) | |
630 | { | |
631 | struct hugetlbfs_sb_info *sbi = HUGETLBFS_SB(sb); | |
632 | ||
633 | if (sbi) { | |
634 | sb->s_fs_info = NULL; | |
635 | kfree(sbi); | |
636 | } | |
637 | } | |
638 | ||
96527980 CH |
639 | static inline int hugetlbfs_dec_free_inodes(struct hugetlbfs_sb_info *sbinfo) |
640 | { | |
641 | if (sbinfo->free_inodes >= 0) { | |
642 | spin_lock(&sbinfo->stat_lock); | |
643 | if (unlikely(!sbinfo->free_inodes)) { | |
644 | spin_unlock(&sbinfo->stat_lock); | |
645 | return 0; | |
646 | } | |
647 | sbinfo->free_inodes--; | |
648 | spin_unlock(&sbinfo->stat_lock); | |
649 | } | |
650 | ||
651 | return 1; | |
652 | } | |
653 | ||
654 | static void hugetlbfs_inc_free_inodes(struct hugetlbfs_sb_info *sbinfo) | |
655 | { | |
656 | if (sbinfo->free_inodes >= 0) { | |
657 | spin_lock(&sbinfo->stat_lock); | |
658 | sbinfo->free_inodes++; | |
659 | spin_unlock(&sbinfo->stat_lock); | |
660 | } | |
661 | } | |
662 | ||
663 | ||
e18b890b | 664 | static struct kmem_cache *hugetlbfs_inode_cachep; |
1da177e4 LT |
665 | |
666 | static struct inode *hugetlbfs_alloc_inode(struct super_block *sb) | |
667 | { | |
96527980 | 668 | struct hugetlbfs_sb_info *sbinfo = HUGETLBFS_SB(sb); |
1da177e4 LT |
669 | struct hugetlbfs_inode_info *p; |
670 | ||
96527980 CH |
671 | if (unlikely(!hugetlbfs_dec_free_inodes(sbinfo))) |
672 | return NULL; | |
e94b1766 | 673 | p = kmem_cache_alloc(hugetlbfs_inode_cachep, GFP_KERNEL); |
96527980 CH |
674 | if (unlikely(!p)) { |
675 | hugetlbfs_inc_free_inodes(sbinfo); | |
1da177e4 | 676 | return NULL; |
96527980 | 677 | } |
1da177e4 LT |
678 | return &p->vfs_inode; |
679 | } | |
680 | ||
fa0d7e3d NP |
681 | static void hugetlbfs_i_callback(struct rcu_head *head) |
682 | { | |
683 | struct inode *inode = container_of(head, struct inode, i_rcu); | |
fa0d7e3d NP |
684 | kmem_cache_free(hugetlbfs_inode_cachep, HUGETLBFS_I(inode)); |
685 | } | |
686 | ||
1da177e4 LT |
687 | static void hugetlbfs_destroy_inode(struct inode *inode) |
688 | { | |
96527980 | 689 | hugetlbfs_inc_free_inodes(HUGETLBFS_SB(inode->i_sb)); |
1da177e4 | 690 | mpol_free_shared_policy(&HUGETLBFS_I(inode)->policy); |
fa0d7e3d | 691 | call_rcu(&inode->i_rcu, hugetlbfs_i_callback); |
1da177e4 LT |
692 | } |
693 | ||
f5e54d6e | 694 | static const struct address_space_operations hugetlbfs_aops = { |
800d15a5 NP |
695 | .write_begin = hugetlbfs_write_begin, |
696 | .write_end = hugetlbfs_write_end, | |
1da177e4 | 697 | .set_page_dirty = hugetlbfs_set_page_dirty, |
290408d4 | 698 | .migratepage = hugetlbfs_migrate_page, |
1da177e4 LT |
699 | }; |
700 | ||
96527980 | 701 | |
51cc5068 | 702 | static void init_once(void *foo) |
96527980 CH |
703 | { |
704 | struct hugetlbfs_inode_info *ei = (struct hugetlbfs_inode_info *)foo; | |
705 | ||
a35afb83 | 706 | inode_init_once(&ei->vfs_inode); |
96527980 CH |
707 | } |
708 | ||
4b6f5d20 | 709 | const struct file_operations hugetlbfs_file_operations = { |
e63e1e5a | 710 | .read = hugetlbfs_read, |
1da177e4 | 711 | .mmap = hugetlbfs_file_mmap, |
1b061d92 | 712 | .fsync = noop_fsync, |
1da177e4 | 713 | .get_unmapped_area = hugetlb_get_unmapped_area, |
6038f373 | 714 | .llseek = default_llseek, |
1da177e4 LT |
715 | }; |
716 | ||
92e1d5be | 717 | static const struct inode_operations hugetlbfs_dir_inode_operations = { |
1da177e4 LT |
718 | .create = hugetlbfs_create, |
719 | .lookup = simple_lookup, | |
720 | .link = simple_link, | |
721 | .unlink = simple_unlink, | |
722 | .symlink = hugetlbfs_symlink, | |
723 | .mkdir = hugetlbfs_mkdir, | |
724 | .rmdir = simple_rmdir, | |
725 | .mknod = hugetlbfs_mknod, | |
726 | .rename = simple_rename, | |
727 | .setattr = hugetlbfs_setattr, | |
728 | }; | |
729 | ||
92e1d5be | 730 | static const struct inode_operations hugetlbfs_inode_operations = { |
1da177e4 LT |
731 | .setattr = hugetlbfs_setattr, |
732 | }; | |
733 | ||
ee9b6d61 | 734 | static const struct super_operations hugetlbfs_ops = { |
1da177e4 LT |
735 | .alloc_inode = hugetlbfs_alloc_inode, |
736 | .destroy_inode = hugetlbfs_destroy_inode, | |
2bbbda30 | 737 | .evict_inode = hugetlbfs_evict_inode, |
1da177e4 | 738 | .statfs = hugetlbfs_statfs, |
1da177e4 | 739 | .put_super = hugetlbfs_put_super, |
10f19a86 | 740 | .show_options = generic_show_options, |
1da177e4 LT |
741 | }; |
742 | ||
743 | static int | |
744 | hugetlbfs_parse_options(char *options, struct hugetlbfs_config *pconfig) | |
745 | { | |
e73a75fa RD |
746 | char *p, *rest; |
747 | substring_t args[MAX_OPT_ARGS]; | |
748 | int option; | |
a137e1cc AK |
749 | unsigned long long size = 0; |
750 | enum { NO_SIZE, SIZE_STD, SIZE_PERCENT } setsize = NO_SIZE; | |
1da177e4 LT |
751 | |
752 | if (!options) | |
753 | return 0; | |
1da177e4 | 754 | |
e73a75fa RD |
755 | while ((p = strsep(&options, ",")) != NULL) { |
756 | int token; | |
b4c07bce LS |
757 | if (!*p) |
758 | continue; | |
e73a75fa RD |
759 | |
760 | token = match_token(p, tokens, args); | |
761 | switch (token) { | |
762 | case Opt_uid: | |
763 | if (match_int(&args[0], &option)) | |
764 | goto bad_val; | |
765 | pconfig->uid = option; | |
766 | break; | |
767 | ||
768 | case Opt_gid: | |
769 | if (match_int(&args[0], &option)) | |
770 | goto bad_val; | |
771 | pconfig->gid = option; | |
772 | break; | |
773 | ||
774 | case Opt_mode: | |
775 | if (match_octal(&args[0], &option)) | |
776 | goto bad_val; | |
75897d60 | 777 | pconfig->mode = option & 01777U; |
e73a75fa RD |
778 | break; |
779 | ||
780 | case Opt_size: { | |
e73a75fa RD |
781 | /* memparse() will accept a K/M/G without a digit */ |
782 | if (!isdigit(*args[0].from)) | |
783 | goto bad_val; | |
784 | size = memparse(args[0].from, &rest); | |
a137e1cc AK |
785 | setsize = SIZE_STD; |
786 | if (*rest == '%') | |
787 | setsize = SIZE_PERCENT; | |
e73a75fa RD |
788 | break; |
789 | } | |
1da177e4 | 790 | |
e73a75fa RD |
791 | case Opt_nr_inodes: |
792 | /* memparse() will accept a K/M/G without a digit */ | |
793 | if (!isdigit(*args[0].from)) | |
794 | goto bad_val; | |
795 | pconfig->nr_inodes = memparse(args[0].from, &rest); | |
796 | break; | |
797 | ||
a137e1cc AK |
798 | case Opt_pagesize: { |
799 | unsigned long ps; | |
800 | ps = memparse(args[0].from, &rest); | |
801 | pconfig->hstate = size_to_hstate(ps); | |
802 | if (!pconfig->hstate) { | |
803 | printk(KERN_ERR | |
804 | "hugetlbfs: Unsupported page size %lu MB\n", | |
805 | ps >> 20); | |
806 | return -EINVAL; | |
807 | } | |
808 | break; | |
809 | } | |
810 | ||
e73a75fa | 811 | default: |
b4c07bce LS |
812 | printk(KERN_ERR "hugetlbfs: Bad mount option: \"%s\"\n", |
813 | p); | |
814 | return -EINVAL; | |
e73a75fa RD |
815 | break; |
816 | } | |
1da177e4 | 817 | } |
a137e1cc AK |
818 | |
819 | /* Do size after hstate is set up */ | |
820 | if (setsize > NO_SIZE) { | |
821 | struct hstate *h = pconfig->hstate; | |
822 | if (setsize == SIZE_PERCENT) { | |
823 | size <<= huge_page_shift(h); | |
824 | size *= h->max_huge_pages; | |
825 | do_div(size, 100); | |
826 | } | |
827 | pconfig->nr_blocks = (size >> huge_page_shift(h)); | |
828 | } | |
829 | ||
1da177e4 | 830 | return 0; |
e73a75fa RD |
831 | |
832 | bad_val: | |
833 | printk(KERN_ERR "hugetlbfs: Bad value '%s' for mount option '%s'\n", | |
834 | args[0].from, p); | |
c12ddba0 | 835 | return -EINVAL; |
1da177e4 LT |
836 | } |
837 | ||
838 | static int | |
839 | hugetlbfs_fill_super(struct super_block *sb, void *data, int silent) | |
840 | { | |
841 | struct inode * inode; | |
842 | struct dentry * root; | |
843 | int ret; | |
844 | struct hugetlbfs_config config; | |
845 | struct hugetlbfs_sb_info *sbinfo; | |
846 | ||
10f19a86 MS |
847 | save_mount_options(sb, data); |
848 | ||
1da177e4 LT |
849 | config.nr_blocks = -1; /* No limit on size by default */ |
850 | config.nr_inodes = -1; /* No limit on number of inodes by default */ | |
77c70de1 DH |
851 | config.uid = current_fsuid(); |
852 | config.gid = current_fsgid(); | |
1da177e4 | 853 | config.mode = 0755; |
a137e1cc | 854 | config.hstate = &default_hstate; |
1da177e4 | 855 | ret = hugetlbfs_parse_options(data, &config); |
1da177e4 LT |
856 | if (ret) |
857 | return ret; | |
858 | ||
859 | sbinfo = kmalloc(sizeof(struct hugetlbfs_sb_info), GFP_KERNEL); | |
860 | if (!sbinfo) | |
861 | return -ENOMEM; | |
862 | sb->s_fs_info = sbinfo; | |
a137e1cc | 863 | sbinfo->hstate = config.hstate; |
1da177e4 LT |
864 | spin_lock_init(&sbinfo->stat_lock); |
865 | sbinfo->max_blocks = config.nr_blocks; | |
866 | sbinfo->free_blocks = config.nr_blocks; | |
867 | sbinfo->max_inodes = config.nr_inodes; | |
868 | sbinfo->free_inodes = config.nr_inodes; | |
869 | sb->s_maxbytes = MAX_LFS_FILESIZE; | |
a137e1cc AK |
870 | sb->s_blocksize = huge_page_size(config.hstate); |
871 | sb->s_blocksize_bits = huge_page_shift(config.hstate); | |
1da177e4 LT |
872 | sb->s_magic = HUGETLBFS_MAGIC; |
873 | sb->s_op = &hugetlbfs_ops; | |
874 | sb->s_time_gran = 1; | |
7d54fa64 | 875 | inode = hugetlbfs_get_root(sb, &config); |
1da177e4 LT |
876 | if (!inode) |
877 | goto out_free; | |
878 | ||
879 | root = d_alloc_root(inode); | |
880 | if (!root) { | |
881 | iput(inode); | |
882 | goto out_free; | |
883 | } | |
884 | sb->s_root = root; | |
885 | return 0; | |
886 | out_free: | |
887 | kfree(sbinfo); | |
888 | return -ENOMEM; | |
889 | } | |
890 | ||
9a119c05 | 891 | int hugetlb_get_quota(struct address_space *mapping, long delta) |
1da177e4 LT |
892 | { |
893 | int ret = 0; | |
894 | struct hugetlbfs_sb_info *sbinfo = HUGETLBFS_SB(mapping->host->i_sb); | |
895 | ||
896 | if (sbinfo->free_blocks > -1) { | |
897 | spin_lock(&sbinfo->stat_lock); | |
9a119c05 AL |
898 | if (sbinfo->free_blocks - delta >= 0) |
899 | sbinfo->free_blocks -= delta; | |
1da177e4 LT |
900 | else |
901 | ret = -ENOMEM; | |
902 | spin_unlock(&sbinfo->stat_lock); | |
903 | } | |
904 | ||
905 | return ret; | |
906 | } | |
907 | ||
9a119c05 | 908 | void hugetlb_put_quota(struct address_space *mapping, long delta) |
1da177e4 LT |
909 | { |
910 | struct hugetlbfs_sb_info *sbinfo = HUGETLBFS_SB(mapping->host->i_sb); | |
911 | ||
912 | if (sbinfo->free_blocks > -1) { | |
913 | spin_lock(&sbinfo->stat_lock); | |
9a119c05 | 914 | sbinfo->free_blocks += delta; |
1da177e4 LT |
915 | spin_unlock(&sbinfo->stat_lock); |
916 | } | |
917 | } | |
918 | ||
3c26ff6e AV |
919 | static struct dentry *hugetlbfs_mount(struct file_system_type *fs_type, |
920 | int flags, const char *dev_name, void *data) | |
1da177e4 | 921 | { |
3c26ff6e | 922 | return mount_nodev(fs_type, flags, data, hugetlbfs_fill_super); |
1da177e4 LT |
923 | } |
924 | ||
925 | static struct file_system_type hugetlbfs_fs_type = { | |
926 | .name = "hugetlbfs", | |
3c26ff6e | 927 | .mount = hugetlbfs_mount, |
1da177e4 LT |
928 | .kill_sb = kill_litter_super, |
929 | }; | |
930 | ||
931 | static struct vfsmount *hugetlbfs_vfsmount; | |
932 | ||
ef1ff6b8 | 933 | static int can_do_hugetlb_shm(void) |
1da177e4 | 934 | { |
ef1ff6b8 | 935 | return capable(CAP_IPC_LOCK) || in_group_p(sysctl_hugetlb_shm_group); |
1da177e4 LT |
936 | } |
937 | ||
ca16d140 KM |
938 | struct file *hugetlb_file_setup(const char *name, size_t size, |
939 | vm_flags_t acctflag, | |
6bfde05b | 940 | struct user_struct **user, int creat_flags) |
1da177e4 LT |
941 | { |
942 | int error = -ENOMEM; | |
943 | struct file *file; | |
944 | struct inode *inode; | |
2c48b9c4 AV |
945 | struct path path; |
946 | struct dentry *root; | |
1da177e4 | 947 | struct qstr quick_string; |
1da177e4 | 948 | |
353d5c30 | 949 | *user = NULL; |
5bc98594 AM |
950 | if (!hugetlbfs_vfsmount) |
951 | return ERR_PTR(-ENOENT); | |
952 | ||
ef1ff6b8 | 953 | if (creat_flags == HUGETLB_SHMFS_INODE && !can_do_hugetlb_shm()) { |
353d5c30 HD |
954 | *user = current_user(); |
955 | if (user_shm_lock(size, *user)) { | |
52ca0e84 | 956 | printk_once(KERN_WARNING "Using mlock ulimits for SHM_HUGETLB is deprecated\n"); |
353d5c30 HD |
957 | } else { |
958 | *user = NULL; | |
2584e517 | 959 | return ERR_PTR(-EPERM); |
353d5c30 | 960 | } |
2584e517 | 961 | } |
1da177e4 | 962 | |
1da177e4 | 963 | root = hugetlbfs_vfsmount->mnt_root; |
9d66586f | 964 | quick_string.name = name; |
1da177e4 LT |
965 | quick_string.len = strlen(quick_string.name); |
966 | quick_string.hash = 0; | |
2c48b9c4 AV |
967 | path.dentry = d_alloc(root, &quick_string); |
968 | if (!path.dentry) | |
1da177e4 LT |
969 | goto out_shm_unlock; |
970 | ||
2c48b9c4 | 971 | path.mnt = mntget(hugetlbfs_vfsmount); |
1da177e4 | 972 | error = -ENOSPC; |
7d54fa64 | 973 | inode = hugetlbfs_get_inode(root->d_sb, NULL, S_IFREG | S_IRWXUGO, 0); |
1da177e4 | 974 | if (!inode) |
ce8d2cdf | 975 | goto out_dentry; |
1da177e4 | 976 | |
b45b5bd6 | 977 | error = -ENOMEM; |
a5516438 | 978 | if (hugetlb_reserve_pages(inode, 0, |
5a6fe125 MG |
979 | size >> huge_page_shift(hstate_inode(inode)), NULL, |
980 | acctflag)) | |
b45b5bd6 DG |
981 | goto out_inode; |
982 | ||
2c48b9c4 | 983 | d_instantiate(path.dentry, inode); |
1da177e4 | 984 | inode->i_size = size; |
6d6b77f1 | 985 | clear_nlink(inode); |
ce8d2cdf DH |
986 | |
987 | error = -ENFILE; | |
2c48b9c4 | 988 | file = alloc_file(&path, FMODE_WRITE | FMODE_READ, |
ce8d2cdf DH |
989 | &hugetlbfs_file_operations); |
990 | if (!file) | |
b4d232e6 | 991 | goto out_dentry; /* inode is already attached */ |
ce8d2cdf | 992 | |
1da177e4 LT |
993 | return file; |
994 | ||
b45b5bd6 DG |
995 | out_inode: |
996 | iput(inode); | |
1da177e4 | 997 | out_dentry: |
2c48b9c4 | 998 | path_put(&path); |
1da177e4 | 999 | out_shm_unlock: |
353d5c30 HD |
1000 | if (*user) { |
1001 | user_shm_unlock(size, *user); | |
1002 | *user = NULL; | |
1003 | } | |
1da177e4 LT |
1004 | return ERR_PTR(error); |
1005 | } | |
1006 | ||
1007 | static int __init init_hugetlbfs_fs(void) | |
1008 | { | |
1009 | int error; | |
1010 | struct vfsmount *vfsmount; | |
1011 | ||
e0bf68dd PZ |
1012 | error = bdi_init(&hugetlbfs_backing_dev_info); |
1013 | if (error) | |
1014 | return error; | |
1015 | ||
1da177e4 LT |
1016 | hugetlbfs_inode_cachep = kmem_cache_create("hugetlbfs_inode_cache", |
1017 | sizeof(struct hugetlbfs_inode_info), | |
20c2df83 | 1018 | 0, 0, init_once); |
1da177e4 | 1019 | if (hugetlbfs_inode_cachep == NULL) |
e0bf68dd | 1020 | goto out2; |
1da177e4 LT |
1021 | |
1022 | error = register_filesystem(&hugetlbfs_fs_type); | |
1023 | if (error) | |
1024 | goto out; | |
1025 | ||
1026 | vfsmount = kern_mount(&hugetlbfs_fs_type); | |
1027 | ||
1028 | if (!IS_ERR(vfsmount)) { | |
1029 | hugetlbfs_vfsmount = vfsmount; | |
1030 | return 0; | |
1031 | } | |
1032 | ||
1033 | error = PTR_ERR(vfsmount); | |
1034 | ||
1035 | out: | |
1036 | if (error) | |
1037 | kmem_cache_destroy(hugetlbfs_inode_cachep); | |
e0bf68dd PZ |
1038 | out2: |
1039 | bdi_destroy(&hugetlbfs_backing_dev_info); | |
1da177e4 LT |
1040 | return error; |
1041 | } | |
1042 | ||
1043 | static void __exit exit_hugetlbfs_fs(void) | |
1044 | { | |
1045 | kmem_cache_destroy(hugetlbfs_inode_cachep); | |
423e0ab0 | 1046 | kern_unmount(hugetlbfs_vfsmount); |
1da177e4 | 1047 | unregister_filesystem(&hugetlbfs_fs_type); |
e0bf68dd | 1048 | bdi_destroy(&hugetlbfs_backing_dev_info); |
1da177e4 LT |
1049 | } |
1050 | ||
1051 | module_init(init_hugetlbfs_fs) | |
1052 | module_exit(exit_hugetlbfs_fs) | |
1053 | ||
1054 | MODULE_LICENSE("GPL"); |