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 | 43 | |
a1d776ee DG |
44 | struct hugetlbfs_config { |
45 | uid_t uid; | |
46 | gid_t gid; | |
47 | umode_t mode; | |
48 | long nr_blocks; | |
49 | long nr_inodes; | |
50 | struct hstate *hstate; | |
51 | }; | |
52 | ||
53 | struct hugetlbfs_inode_info { | |
54 | struct shared_policy policy; | |
55 | struct inode vfs_inode; | |
56 | }; | |
57 | ||
58 | static inline struct hugetlbfs_inode_info *HUGETLBFS_I(struct inode *inode) | |
59 | { | |
60 | return container_of(inode, struct hugetlbfs_inode_info, vfs_inode); | |
61 | } | |
62 | ||
1da177e4 | 63 | static struct backing_dev_info hugetlbfs_backing_dev_info = { |
d993831f | 64 | .name = "hugetlbfs", |
1da177e4 | 65 | .ra_pages = 0, /* No readahead */ |
e4ad08fe | 66 | .capabilities = BDI_CAP_NO_ACCT_AND_WRITEBACK, |
1da177e4 LT |
67 | }; |
68 | ||
69 | int sysctl_hugetlb_shm_group; | |
70 | ||
e73a75fa RD |
71 | enum { |
72 | Opt_size, Opt_nr_inodes, | |
73 | Opt_mode, Opt_uid, Opt_gid, | |
a137e1cc | 74 | Opt_pagesize, |
e73a75fa RD |
75 | Opt_err, |
76 | }; | |
77 | ||
a447c093 | 78 | static const match_table_t tokens = { |
e73a75fa RD |
79 | {Opt_size, "size=%s"}, |
80 | {Opt_nr_inodes, "nr_inodes=%s"}, | |
81 | {Opt_mode, "mode=%o"}, | |
82 | {Opt_uid, "uid=%u"}, | |
83 | {Opt_gid, "gid=%u"}, | |
a137e1cc | 84 | {Opt_pagesize, "pagesize=%s"}, |
e73a75fa RD |
85 | {Opt_err, NULL}, |
86 | }; | |
87 | ||
2e9b367c AL |
88 | static void huge_pagevec_release(struct pagevec *pvec) |
89 | { | |
90 | int i; | |
91 | ||
92 | for (i = 0; i < pagevec_count(pvec); ++i) | |
93 | put_page(pvec->pages[i]); | |
94 | ||
95 | pagevec_reinit(pvec); | |
96 | } | |
97 | ||
1da177e4 LT |
98 | static int hugetlbfs_file_mmap(struct file *file, struct vm_area_struct *vma) |
99 | { | |
b39424e2 | 100 | struct inode *inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
101 | loff_t len, vma_len; |
102 | int ret; | |
a5516438 | 103 | struct hstate *h = hstate_file(file); |
1da177e4 | 104 | |
68589bc3 | 105 | /* |
dec4ad86 DG |
106 | * vma address alignment (but not the pgoff alignment) has |
107 | * already been checked by prepare_hugepage_range. If you add | |
108 | * any error returns here, do so after setting VM_HUGETLB, so | |
109 | * is_vm_hugetlb_page tests below unmap_region go the right | |
110 | * way when do_mmap_pgoff unwinds (may be important on powerpc | |
111 | * and ia64). | |
68589bc3 HD |
112 | */ |
113 | vma->vm_flags |= VM_HUGETLB | VM_RESERVED; | |
114 | vma->vm_ops = &hugetlb_vm_ops; | |
1da177e4 | 115 | |
2b37c35e | 116 | if (vma->vm_pgoff & (~huge_page_mask(h) >> PAGE_SHIFT)) |
dec4ad86 DG |
117 | return -EINVAL; |
118 | ||
1da177e4 LT |
119 | vma_len = (loff_t)(vma->vm_end - vma->vm_start); |
120 | ||
1b1dcc1b | 121 | mutex_lock(&inode->i_mutex); |
1da177e4 | 122 | file_accessed(file); |
1da177e4 LT |
123 | |
124 | ret = -ENOMEM; | |
125 | len = vma_len + ((loff_t)vma->vm_pgoff << PAGE_SHIFT); | |
1da177e4 | 126 | |
a1e78772 | 127 | if (hugetlb_reserve_pages(inode, |
a5516438 | 128 | vma->vm_pgoff >> huge_page_order(h), |
5a6fe125 MG |
129 | len >> huge_page_shift(h), vma, |
130 | vma->vm_flags)) | |
a43a8c39 | 131 | goto out; |
b45b5bd6 | 132 | |
4c887265 AL |
133 | ret = 0; |
134 | hugetlb_prefault_arch_hook(vma->vm_mm); | |
b6174df5 | 135 | if (vma->vm_flags & VM_WRITE && inode->i_size < len) |
1da177e4 LT |
136 | inode->i_size = len; |
137 | out: | |
1b1dcc1b | 138 | mutex_unlock(&inode->i_mutex); |
1da177e4 LT |
139 | |
140 | return ret; | |
141 | } | |
142 | ||
143 | /* | |
508034a3 | 144 | * Called under down_write(mmap_sem). |
1da177e4 LT |
145 | */ |
146 | ||
d2ba27e8 | 147 | #ifndef HAVE_ARCH_HUGETLB_UNMAPPED_AREA |
1da177e4 LT |
148 | static unsigned long |
149 | hugetlb_get_unmapped_area(struct file *file, unsigned long addr, | |
150 | unsigned long len, unsigned long pgoff, unsigned long flags) | |
151 | { | |
152 | struct mm_struct *mm = current->mm; | |
153 | struct vm_area_struct *vma; | |
154 | unsigned long start_addr; | |
a5516438 | 155 | struct hstate *h = hstate_file(file); |
1da177e4 | 156 | |
a5516438 | 157 | if (len & ~huge_page_mask(h)) |
1da177e4 LT |
158 | return -EINVAL; |
159 | if (len > TASK_SIZE) | |
160 | return -ENOMEM; | |
161 | ||
036e0856 | 162 | if (flags & MAP_FIXED) { |
a5516438 | 163 | if (prepare_hugepage_range(file, addr, len)) |
036e0856 BH |
164 | return -EINVAL; |
165 | return addr; | |
166 | } | |
167 | ||
1da177e4 | 168 | if (addr) { |
a5516438 | 169 | addr = ALIGN(addr, huge_page_size(h)); |
1da177e4 LT |
170 | vma = find_vma(mm, addr); |
171 | if (TASK_SIZE - len >= addr && | |
172 | (!vma || addr + len <= vma->vm_start)) | |
173 | return addr; | |
174 | } | |
175 | ||
4bfc130d XG |
176 | if (len > mm->cached_hole_size) |
177 | start_addr = mm->free_area_cache; | |
178 | else { | |
1363c3cd | 179 | start_addr = TASK_UNMAPPED_BASE; |
4bfc130d XG |
180 | mm->cached_hole_size = 0; |
181 | } | |
1363c3cd | 182 | |
1da177e4 | 183 | full_search: |
a5516438 | 184 | addr = ALIGN(start_addr, huge_page_size(h)); |
1da177e4 LT |
185 | |
186 | for (vma = find_vma(mm, addr); ; vma = vma->vm_next) { | |
187 | /* At this point: (!vma || addr < vma->vm_end). */ | |
188 | if (TASK_SIZE - len < addr) { | |
189 | /* | |
190 | * Start a new search - just in case we missed | |
191 | * some holes. | |
192 | */ | |
193 | if (start_addr != TASK_UNMAPPED_BASE) { | |
194 | start_addr = TASK_UNMAPPED_BASE; | |
4bfc130d | 195 | mm->cached_hole_size = 0; |
1da177e4 LT |
196 | goto full_search; |
197 | } | |
198 | return -ENOMEM; | |
199 | } | |
200 | ||
4bfc130d XG |
201 | if (!vma || addr + len <= vma->vm_start) { |
202 | mm->free_area_cache = addr + len; | |
1da177e4 | 203 | return addr; |
4bfc130d XG |
204 | } |
205 | if (addr + mm->cached_hole_size < vma->vm_start) | |
206 | mm->cached_hole_size = vma->vm_start - addr; | |
a5516438 | 207 | addr = ALIGN(vma->vm_end, huge_page_size(h)); |
1da177e4 LT |
208 | } |
209 | } | |
210 | #endif | |
211 | ||
e63e1e5a BP |
212 | static int |
213 | hugetlbfs_read_actor(struct page *page, unsigned long offset, | |
214 | char __user *buf, unsigned long count, | |
215 | unsigned long size) | |
216 | { | |
217 | char *kaddr; | |
218 | unsigned long left, copied = 0; | |
219 | int i, chunksize; | |
220 | ||
221 | if (size > count) | |
222 | size = count; | |
223 | ||
224 | /* Find which 4k chunk and offset with in that chunk */ | |
225 | i = offset >> PAGE_CACHE_SHIFT; | |
226 | offset = offset & ~PAGE_CACHE_MASK; | |
227 | ||
228 | while (size) { | |
229 | chunksize = PAGE_CACHE_SIZE; | |
230 | if (offset) | |
231 | chunksize -= offset; | |
232 | if (chunksize > size) | |
233 | chunksize = size; | |
234 | kaddr = kmap(&page[i]); | |
235 | left = __copy_to_user(buf, kaddr + offset, chunksize); | |
236 | kunmap(&page[i]); | |
237 | if (left) { | |
238 | copied += (chunksize - left); | |
239 | break; | |
240 | } | |
241 | offset = 0; | |
242 | size -= chunksize; | |
243 | buf += chunksize; | |
244 | copied += chunksize; | |
245 | i++; | |
246 | } | |
247 | return copied ? copied : -EFAULT; | |
248 | } | |
249 | ||
250 | /* | |
251 | * Support for read() - Find the page attached to f_mapping and copy out the | |
252 | * data. Its *very* similar to do_generic_mapping_read(), we can't use that | |
253 | * since it has PAGE_CACHE_SIZE assumptions. | |
254 | */ | |
255 | static ssize_t hugetlbfs_read(struct file *filp, char __user *buf, | |
256 | size_t len, loff_t *ppos) | |
257 | { | |
a5516438 | 258 | struct hstate *h = hstate_file(filp); |
e63e1e5a BP |
259 | struct address_space *mapping = filp->f_mapping; |
260 | struct inode *inode = mapping->host; | |
a5516438 AK |
261 | unsigned long index = *ppos >> huge_page_shift(h); |
262 | unsigned long offset = *ppos & ~huge_page_mask(h); | |
e63e1e5a BP |
263 | unsigned long end_index; |
264 | loff_t isize; | |
265 | ssize_t retval = 0; | |
266 | ||
e63e1e5a BP |
267 | /* validate length */ |
268 | if (len == 0) | |
269 | goto out; | |
270 | ||
e63e1e5a BP |
271 | for (;;) { |
272 | struct page *page; | |
a5516438 | 273 | unsigned long nr, ret; |
91bf189c | 274 | int ra; |
e63e1e5a BP |
275 | |
276 | /* nr is the maximum number of bytes to copy from this page */ | |
a5516438 | 277 | nr = huge_page_size(h); |
a05b0855 AK |
278 | isize = i_size_read(inode); |
279 | if (!isize) | |
280 | goto out; | |
281 | end_index = (isize - 1) >> huge_page_shift(h); | |
e63e1e5a BP |
282 | if (index >= end_index) { |
283 | if (index > end_index) | |
284 | goto out; | |
a5516438 | 285 | nr = ((isize - 1) & ~huge_page_mask(h)) + 1; |
a05b0855 | 286 | if (nr <= offset) |
e63e1e5a | 287 | goto out; |
e63e1e5a BP |
288 | } |
289 | nr = nr - offset; | |
290 | ||
291 | /* Find the page */ | |
a05b0855 | 292 | page = find_lock_page(mapping, index); |
e63e1e5a BP |
293 | if (unlikely(page == NULL)) { |
294 | /* | |
295 | * We have a HOLE, zero out the user-buffer for the | |
296 | * length of the hole or request. | |
297 | */ | |
298 | ret = len < nr ? len : nr; | |
299 | if (clear_user(buf, ret)) | |
91bf189c RK |
300 | ra = -EFAULT; |
301 | else | |
302 | ra = 0; | |
e63e1e5a | 303 | } else { |
a05b0855 AK |
304 | unlock_page(page); |
305 | ||
e63e1e5a BP |
306 | /* |
307 | * We have the page, copy it to user space buffer. | |
308 | */ | |
91bf189c RK |
309 | ra = hugetlbfs_read_actor(page, offset, buf, len, nr); |
310 | ret = ra; | |
a05b0855 | 311 | page_cache_release(page); |
e63e1e5a | 312 | } |
91bf189c | 313 | if (ra < 0) { |
e63e1e5a | 314 | if (retval == 0) |
91bf189c | 315 | retval = ra; |
e63e1e5a BP |
316 | goto out; |
317 | } | |
318 | ||
319 | offset += ret; | |
320 | retval += ret; | |
321 | len -= ret; | |
a5516438 AK |
322 | index += offset >> huge_page_shift(h); |
323 | offset &= ~huge_page_mask(h); | |
e63e1e5a | 324 | |
e63e1e5a BP |
325 | /* short read or no more work */ |
326 | if ((ret != nr) || (len == 0)) | |
327 | break; | |
328 | } | |
329 | out: | |
a5516438 | 330 | *ppos = ((loff_t)index << huge_page_shift(h)) + offset; |
e63e1e5a BP |
331 | return retval; |
332 | } | |
333 | ||
800d15a5 NP |
334 | static int hugetlbfs_write_begin(struct file *file, |
335 | struct address_space *mapping, | |
336 | loff_t pos, unsigned len, unsigned flags, | |
337 | struct page **pagep, void **fsdata) | |
1da177e4 LT |
338 | { |
339 | return -EINVAL; | |
340 | } | |
341 | ||
800d15a5 NP |
342 | static int hugetlbfs_write_end(struct file *file, struct address_space *mapping, |
343 | loff_t pos, unsigned len, unsigned copied, | |
344 | struct page *page, void *fsdata) | |
1da177e4 | 345 | { |
800d15a5 | 346 | BUG(); |
1da177e4 LT |
347 | return -EINVAL; |
348 | } | |
349 | ||
1da177e4 LT |
350 | static void truncate_huge_page(struct page *page) |
351 | { | |
fba2591b | 352 | cancel_dirty_page(page, /* No IO accounting for huge pages? */0); |
1da177e4 | 353 | ClearPageUptodate(page); |
bd65cb86 | 354 | delete_from_page_cache(page); |
1da177e4 LT |
355 | } |
356 | ||
b45b5bd6 | 357 | static void truncate_hugepages(struct inode *inode, loff_t lstart) |
1da177e4 | 358 | { |
a5516438 | 359 | struct hstate *h = hstate_inode(inode); |
b45b5bd6 | 360 | struct address_space *mapping = &inode->i_data; |
a5516438 | 361 | const pgoff_t start = lstart >> huge_page_shift(h); |
1da177e4 LT |
362 | struct pagevec pvec; |
363 | pgoff_t next; | |
a43a8c39 | 364 | int i, freed = 0; |
1da177e4 LT |
365 | |
366 | pagevec_init(&pvec, 0); | |
367 | next = start; | |
368 | while (1) { | |
369 | if (!pagevec_lookup(&pvec, mapping, next, PAGEVEC_SIZE)) { | |
370 | if (next == start) | |
371 | break; | |
372 | next = start; | |
373 | continue; | |
374 | } | |
375 | ||
376 | for (i = 0; i < pagevec_count(&pvec); ++i) { | |
377 | struct page *page = pvec.pages[i]; | |
378 | ||
379 | lock_page(page); | |
380 | if (page->index > next) | |
381 | next = page->index; | |
382 | ++next; | |
383 | truncate_huge_page(page); | |
384 | unlock_page(page); | |
a43a8c39 | 385 | freed++; |
1da177e4 LT |
386 | } |
387 | huge_pagevec_release(&pvec); | |
388 | } | |
389 | BUG_ON(!lstart && mapping->nrpages); | |
a43a8c39 | 390 | hugetlb_unreserve_pages(inode, start, freed); |
1da177e4 LT |
391 | } |
392 | ||
2bbbda30 | 393 | static void hugetlbfs_evict_inode(struct inode *inode) |
1da177e4 | 394 | { |
b45b5bd6 | 395 | truncate_hugepages(inode, 0); |
b0683aa6 | 396 | end_writeback(inode); |
149f4211 CH |
397 | } |
398 | ||
1da177e4 | 399 | static inline void |
856fc295 | 400 | hugetlb_vmtruncate_list(struct prio_tree_root *root, pgoff_t pgoff) |
1da177e4 LT |
401 | { |
402 | struct vm_area_struct *vma; | |
403 | struct prio_tree_iter iter; | |
404 | ||
856fc295 | 405 | vma_prio_tree_foreach(vma, &iter, root, pgoff, ULONG_MAX) { |
1da177e4 LT |
406 | unsigned long v_offset; |
407 | ||
1da177e4 | 408 | /* |
856fc295 HD |
409 | * Can the expression below overflow on 32-bit arches? |
410 | * No, because the prio_tree returns us only those vmas | |
411 | * which overlap the truncated area starting at pgoff, | |
412 | * and no vma on a 32-bit arch can span beyond the 4GB. | |
1da177e4 | 413 | */ |
856fc295 HD |
414 | if (vma->vm_pgoff < pgoff) |
415 | v_offset = (pgoff - vma->vm_pgoff) << PAGE_SHIFT; | |
416 | else | |
1da177e4 LT |
417 | v_offset = 0; |
418 | ||
502717f4 | 419 | __unmap_hugepage_range(vma, |
04f2cbe3 | 420 | vma->vm_start + v_offset, vma->vm_end, NULL); |
1da177e4 LT |
421 | } |
422 | } | |
423 | ||
1da177e4 LT |
424 | static int hugetlb_vmtruncate(struct inode *inode, loff_t offset) |
425 | { | |
856fc295 | 426 | pgoff_t pgoff; |
1da177e4 | 427 | struct address_space *mapping = inode->i_mapping; |
a5516438 | 428 | struct hstate *h = hstate_inode(inode); |
1da177e4 | 429 | |
a5516438 | 430 | BUG_ON(offset & ~huge_page_mask(h)); |
856fc295 | 431 | pgoff = offset >> PAGE_SHIFT; |
1da177e4 | 432 | |
7aa91e10 | 433 | i_size_write(inode, offset); |
3d48ae45 | 434 | mutex_lock(&mapping->i_mmap_mutex); |
1da177e4 LT |
435 | if (!prio_tree_empty(&mapping->i_mmap)) |
436 | hugetlb_vmtruncate_list(&mapping->i_mmap, pgoff); | |
3d48ae45 | 437 | mutex_unlock(&mapping->i_mmap_mutex); |
b45b5bd6 | 438 | truncate_hugepages(inode, offset); |
1da177e4 LT |
439 | return 0; |
440 | } | |
441 | ||
442 | static int hugetlbfs_setattr(struct dentry *dentry, struct iattr *attr) | |
443 | { | |
444 | struct inode *inode = dentry->d_inode; | |
a5516438 | 445 | struct hstate *h = hstate_inode(inode); |
1da177e4 LT |
446 | int error; |
447 | unsigned int ia_valid = attr->ia_valid; | |
448 | ||
449 | BUG_ON(!inode); | |
450 | ||
451 | error = inode_change_ok(inode, attr); | |
452 | if (error) | |
1025774c | 453 | return error; |
1da177e4 LT |
454 | |
455 | if (ia_valid & ATTR_SIZE) { | |
456 | error = -EINVAL; | |
1025774c CH |
457 | if (attr->ia_size & ~huge_page_mask(h)) |
458 | return -EINVAL; | |
459 | error = hugetlb_vmtruncate(inode, attr->ia_size); | |
1da177e4 | 460 | if (error) |
1025774c | 461 | return error; |
1da177e4 | 462 | } |
1025774c CH |
463 | |
464 | setattr_copy(inode, attr); | |
465 | mark_inode_dirty(inode); | |
466 | return 0; | |
1da177e4 LT |
467 | } |
468 | ||
7d54fa64 AV |
469 | static struct inode *hugetlbfs_get_root(struct super_block *sb, |
470 | struct hugetlbfs_config *config) | |
1da177e4 LT |
471 | { |
472 | struct inode *inode; | |
1da177e4 LT |
473 | |
474 | inode = new_inode(sb); | |
475 | if (inode) { | |
476 | struct hugetlbfs_inode_info *info; | |
85fe4025 | 477 | inode->i_ino = get_next_ino(); |
7d54fa64 AV |
478 | inode->i_mode = S_IFDIR | config->mode; |
479 | inode->i_uid = config->uid; | |
480 | inode->i_gid = config->gid; | |
481 | inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME; | |
482 | info = HUGETLBFS_I(inode); | |
483 | mpol_shared_policy_init(&info->policy, NULL); | |
484 | inode->i_op = &hugetlbfs_dir_inode_operations; | |
485 | inode->i_fop = &simple_dir_operations; | |
486 | /* directory inodes start off with i_nlink == 2 (for "." entry) */ | |
487 | inc_nlink(inode); | |
488 | } | |
489 | return inode; | |
490 | } | |
491 | ||
492 | static struct inode *hugetlbfs_get_inode(struct super_block *sb, | |
493 | struct inode *dir, | |
18df2252 | 494 | umode_t mode, dev_t dev) |
7d54fa64 AV |
495 | { |
496 | struct inode *inode; | |
497 | ||
498 | inode = new_inode(sb); | |
499 | if (inode) { | |
500 | struct hugetlbfs_inode_info *info; | |
501 | inode->i_ino = get_next_ino(); | |
502 | inode_init_owner(inode, dir, mode); | |
1da177e4 LT |
503 | inode->i_mapping->a_ops = &hugetlbfs_aops; |
504 | inode->i_mapping->backing_dev_info =&hugetlbfs_backing_dev_info; | |
505 | inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME; | |
a43a8c39 | 506 | INIT_LIST_HEAD(&inode->i_mapping->private_list); |
1da177e4 | 507 | info = HUGETLBFS_I(inode); |
6bfde05b EM |
508 | /* |
509 | * The policy is initialized here even if we are creating a | |
510 | * private inode because initialization simply creates an | |
511 | * an empty rb tree and calls spin_lock_init(), later when we | |
512 | * call mpol_free_shared_policy() it will just return because | |
513 | * the rb tree will still be empty. | |
514 | */ | |
71fe804b | 515 | mpol_shared_policy_init(&info->policy, NULL); |
1da177e4 LT |
516 | switch (mode & S_IFMT) { |
517 | default: | |
518 | init_special_inode(inode, mode, dev); | |
519 | break; | |
520 | case S_IFREG: | |
521 | inode->i_op = &hugetlbfs_inode_operations; | |
522 | inode->i_fop = &hugetlbfs_file_operations; | |
523 | break; | |
524 | case S_IFDIR: | |
525 | inode->i_op = &hugetlbfs_dir_inode_operations; | |
526 | inode->i_fop = &simple_dir_operations; | |
527 | ||
528 | /* directory inodes start off with i_nlink == 2 (for "." entry) */ | |
d8c76e6f | 529 | inc_nlink(inode); |
1da177e4 LT |
530 | break; |
531 | case S_IFLNK: | |
532 | inode->i_op = &page_symlink_inode_operations; | |
533 | break; | |
534 | } | |
e096d0c7 | 535 | lockdep_annotate_inode_mutex_key(inode); |
1da177e4 LT |
536 | } |
537 | return inode; | |
538 | } | |
539 | ||
540 | /* | |
541 | * File creation. Allocate an inode, and we're done.. | |
542 | */ | |
543 | static int hugetlbfs_mknod(struct inode *dir, | |
1a67aafb | 544 | struct dentry *dentry, umode_t mode, dev_t dev) |
1da177e4 LT |
545 | { |
546 | struct inode *inode; | |
547 | int error = -ENOSPC; | |
7d54fa64 AV |
548 | |
549 | inode = hugetlbfs_get_inode(dir->i_sb, dir, mode, dev); | |
1da177e4 LT |
550 | if (inode) { |
551 | dir->i_ctime = dir->i_mtime = CURRENT_TIME; | |
552 | d_instantiate(dentry, inode); | |
553 | dget(dentry); /* Extra count - pin the dentry in core */ | |
554 | error = 0; | |
555 | } | |
556 | return error; | |
557 | } | |
558 | ||
18bb1db3 | 559 | static int hugetlbfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode) |
1da177e4 LT |
560 | { |
561 | int retval = hugetlbfs_mknod(dir, dentry, mode | S_IFDIR, 0); | |
562 | if (!retval) | |
d8c76e6f | 563 | inc_nlink(dir); |
1da177e4 LT |
564 | return retval; |
565 | } | |
566 | ||
4acdaf27 | 567 | static int hugetlbfs_create(struct inode *dir, struct dentry *dentry, umode_t mode, struct nameidata *nd) |
1da177e4 LT |
568 | { |
569 | return hugetlbfs_mknod(dir, dentry, mode | S_IFREG, 0); | |
570 | } | |
571 | ||
572 | static int hugetlbfs_symlink(struct inode *dir, | |
573 | struct dentry *dentry, const char *symname) | |
574 | { | |
575 | struct inode *inode; | |
576 | int error = -ENOSPC; | |
1da177e4 | 577 | |
7d54fa64 | 578 | inode = hugetlbfs_get_inode(dir->i_sb, dir, S_IFLNK|S_IRWXUGO, 0); |
1da177e4 LT |
579 | if (inode) { |
580 | int l = strlen(symname)+1; | |
581 | error = page_symlink(inode, symname, l); | |
582 | if (!error) { | |
583 | d_instantiate(dentry, inode); | |
584 | dget(dentry); | |
585 | } else | |
586 | iput(inode); | |
587 | } | |
588 | dir->i_ctime = dir->i_mtime = CURRENT_TIME; | |
589 | ||
590 | return error; | |
591 | } | |
592 | ||
593 | /* | |
6649a386 | 594 | * mark the head page dirty |
1da177e4 LT |
595 | */ |
596 | static int hugetlbfs_set_page_dirty(struct page *page) | |
597 | { | |
d85f3385 | 598 | struct page *head = compound_head(page); |
6649a386 KC |
599 | |
600 | SetPageDirty(head); | |
1da177e4 LT |
601 | return 0; |
602 | } | |
603 | ||
290408d4 | 604 | static int hugetlbfs_migrate_page(struct address_space *mapping, |
b969c4ab | 605 | struct page *newpage, struct page *page, |
a6bc32b8 | 606 | enum migrate_mode mode) |
290408d4 NH |
607 | { |
608 | int rc; | |
609 | ||
610 | rc = migrate_huge_page_move_mapping(mapping, newpage, page); | |
611 | if (rc) | |
612 | return rc; | |
613 | migrate_page_copy(newpage, page); | |
614 | ||
615 | return 0; | |
616 | } | |
617 | ||
726c3342 | 618 | static int hugetlbfs_statfs(struct dentry *dentry, struct kstatfs *buf) |
1da177e4 | 619 | { |
726c3342 | 620 | struct hugetlbfs_sb_info *sbinfo = HUGETLBFS_SB(dentry->d_sb); |
a5516438 | 621 | struct hstate *h = hstate_inode(dentry->d_inode); |
1da177e4 LT |
622 | |
623 | buf->f_type = HUGETLBFS_MAGIC; | |
a5516438 | 624 | buf->f_bsize = huge_page_size(h); |
1da177e4 LT |
625 | if (sbinfo) { |
626 | spin_lock(&sbinfo->stat_lock); | |
74a8a65c DG |
627 | /* If no limits set, just report 0 for max/free/used |
628 | * blocks, like simple_statfs() */ | |
90481622 DG |
629 | if (sbinfo->spool) { |
630 | long free_pages; | |
631 | ||
632 | spin_lock(&sbinfo->spool->lock); | |
633 | buf->f_blocks = sbinfo->spool->max_hpages; | |
634 | free_pages = sbinfo->spool->max_hpages | |
635 | - sbinfo->spool->used_hpages; | |
636 | buf->f_bavail = buf->f_bfree = free_pages; | |
637 | spin_unlock(&sbinfo->spool->lock); | |
74a8a65c DG |
638 | buf->f_files = sbinfo->max_inodes; |
639 | buf->f_ffree = sbinfo->free_inodes; | |
640 | } | |
1da177e4 LT |
641 | spin_unlock(&sbinfo->stat_lock); |
642 | } | |
643 | buf->f_namelen = NAME_MAX; | |
644 | return 0; | |
645 | } | |
646 | ||
647 | static void hugetlbfs_put_super(struct super_block *sb) | |
648 | { | |
649 | struct hugetlbfs_sb_info *sbi = HUGETLBFS_SB(sb); | |
650 | ||
651 | if (sbi) { | |
652 | sb->s_fs_info = NULL; | |
90481622 DG |
653 | |
654 | if (sbi->spool) | |
655 | hugepage_put_subpool(sbi->spool); | |
656 | ||
1da177e4 LT |
657 | kfree(sbi); |
658 | } | |
659 | } | |
660 | ||
96527980 CH |
661 | static inline int hugetlbfs_dec_free_inodes(struct hugetlbfs_sb_info *sbinfo) |
662 | { | |
663 | if (sbinfo->free_inodes >= 0) { | |
664 | spin_lock(&sbinfo->stat_lock); | |
665 | if (unlikely(!sbinfo->free_inodes)) { | |
666 | spin_unlock(&sbinfo->stat_lock); | |
667 | return 0; | |
668 | } | |
669 | sbinfo->free_inodes--; | |
670 | spin_unlock(&sbinfo->stat_lock); | |
671 | } | |
672 | ||
673 | return 1; | |
674 | } | |
675 | ||
676 | static void hugetlbfs_inc_free_inodes(struct hugetlbfs_sb_info *sbinfo) | |
677 | { | |
678 | if (sbinfo->free_inodes >= 0) { | |
679 | spin_lock(&sbinfo->stat_lock); | |
680 | sbinfo->free_inodes++; | |
681 | spin_unlock(&sbinfo->stat_lock); | |
682 | } | |
683 | } | |
684 | ||
685 | ||
e18b890b | 686 | static struct kmem_cache *hugetlbfs_inode_cachep; |
1da177e4 LT |
687 | |
688 | static struct inode *hugetlbfs_alloc_inode(struct super_block *sb) | |
689 | { | |
96527980 | 690 | struct hugetlbfs_sb_info *sbinfo = HUGETLBFS_SB(sb); |
1da177e4 LT |
691 | struct hugetlbfs_inode_info *p; |
692 | ||
96527980 CH |
693 | if (unlikely(!hugetlbfs_dec_free_inodes(sbinfo))) |
694 | return NULL; | |
e94b1766 | 695 | p = kmem_cache_alloc(hugetlbfs_inode_cachep, GFP_KERNEL); |
96527980 CH |
696 | if (unlikely(!p)) { |
697 | hugetlbfs_inc_free_inodes(sbinfo); | |
1da177e4 | 698 | return NULL; |
96527980 | 699 | } |
1da177e4 LT |
700 | return &p->vfs_inode; |
701 | } | |
702 | ||
fa0d7e3d NP |
703 | static void hugetlbfs_i_callback(struct rcu_head *head) |
704 | { | |
705 | struct inode *inode = container_of(head, struct inode, i_rcu); | |
fa0d7e3d NP |
706 | kmem_cache_free(hugetlbfs_inode_cachep, HUGETLBFS_I(inode)); |
707 | } | |
708 | ||
1da177e4 LT |
709 | static void hugetlbfs_destroy_inode(struct inode *inode) |
710 | { | |
96527980 | 711 | hugetlbfs_inc_free_inodes(HUGETLBFS_SB(inode->i_sb)); |
1da177e4 | 712 | mpol_free_shared_policy(&HUGETLBFS_I(inode)->policy); |
fa0d7e3d | 713 | call_rcu(&inode->i_rcu, hugetlbfs_i_callback); |
1da177e4 LT |
714 | } |
715 | ||
f5e54d6e | 716 | static const struct address_space_operations hugetlbfs_aops = { |
800d15a5 NP |
717 | .write_begin = hugetlbfs_write_begin, |
718 | .write_end = hugetlbfs_write_end, | |
1da177e4 | 719 | .set_page_dirty = hugetlbfs_set_page_dirty, |
290408d4 | 720 | .migratepage = hugetlbfs_migrate_page, |
1da177e4 LT |
721 | }; |
722 | ||
96527980 | 723 | |
51cc5068 | 724 | static void init_once(void *foo) |
96527980 CH |
725 | { |
726 | struct hugetlbfs_inode_info *ei = (struct hugetlbfs_inode_info *)foo; | |
727 | ||
a35afb83 | 728 | inode_init_once(&ei->vfs_inode); |
96527980 CH |
729 | } |
730 | ||
4b6f5d20 | 731 | const struct file_operations hugetlbfs_file_operations = { |
e63e1e5a | 732 | .read = hugetlbfs_read, |
1da177e4 | 733 | .mmap = hugetlbfs_file_mmap, |
1b061d92 | 734 | .fsync = noop_fsync, |
1da177e4 | 735 | .get_unmapped_area = hugetlb_get_unmapped_area, |
6038f373 | 736 | .llseek = default_llseek, |
1da177e4 LT |
737 | }; |
738 | ||
92e1d5be | 739 | static const struct inode_operations hugetlbfs_dir_inode_operations = { |
1da177e4 LT |
740 | .create = hugetlbfs_create, |
741 | .lookup = simple_lookup, | |
742 | .link = simple_link, | |
743 | .unlink = simple_unlink, | |
744 | .symlink = hugetlbfs_symlink, | |
745 | .mkdir = hugetlbfs_mkdir, | |
746 | .rmdir = simple_rmdir, | |
747 | .mknod = hugetlbfs_mknod, | |
748 | .rename = simple_rename, | |
749 | .setattr = hugetlbfs_setattr, | |
750 | }; | |
751 | ||
92e1d5be | 752 | static const struct inode_operations hugetlbfs_inode_operations = { |
1da177e4 LT |
753 | .setattr = hugetlbfs_setattr, |
754 | }; | |
755 | ||
ee9b6d61 | 756 | static const struct super_operations hugetlbfs_ops = { |
1da177e4 LT |
757 | .alloc_inode = hugetlbfs_alloc_inode, |
758 | .destroy_inode = hugetlbfs_destroy_inode, | |
2bbbda30 | 759 | .evict_inode = hugetlbfs_evict_inode, |
1da177e4 | 760 | .statfs = hugetlbfs_statfs, |
1da177e4 | 761 | .put_super = hugetlbfs_put_super, |
10f19a86 | 762 | .show_options = generic_show_options, |
1da177e4 LT |
763 | }; |
764 | ||
765 | static int | |
766 | hugetlbfs_parse_options(char *options, struct hugetlbfs_config *pconfig) | |
767 | { | |
e73a75fa RD |
768 | char *p, *rest; |
769 | substring_t args[MAX_OPT_ARGS]; | |
770 | int option; | |
a137e1cc AK |
771 | unsigned long long size = 0; |
772 | enum { NO_SIZE, SIZE_STD, SIZE_PERCENT } setsize = NO_SIZE; | |
1da177e4 LT |
773 | |
774 | if (!options) | |
775 | return 0; | |
1da177e4 | 776 | |
e73a75fa RD |
777 | while ((p = strsep(&options, ",")) != NULL) { |
778 | int token; | |
b4c07bce LS |
779 | if (!*p) |
780 | continue; | |
e73a75fa RD |
781 | |
782 | token = match_token(p, tokens, args); | |
783 | switch (token) { | |
784 | case Opt_uid: | |
785 | if (match_int(&args[0], &option)) | |
786 | goto bad_val; | |
787 | pconfig->uid = option; | |
788 | break; | |
789 | ||
790 | case Opt_gid: | |
791 | if (match_int(&args[0], &option)) | |
792 | goto bad_val; | |
793 | pconfig->gid = option; | |
794 | break; | |
795 | ||
796 | case Opt_mode: | |
797 | if (match_octal(&args[0], &option)) | |
798 | goto bad_val; | |
75897d60 | 799 | pconfig->mode = option & 01777U; |
e73a75fa RD |
800 | break; |
801 | ||
802 | case Opt_size: { | |
e73a75fa RD |
803 | /* memparse() will accept a K/M/G without a digit */ |
804 | if (!isdigit(*args[0].from)) | |
805 | goto bad_val; | |
806 | size = memparse(args[0].from, &rest); | |
a137e1cc AK |
807 | setsize = SIZE_STD; |
808 | if (*rest == '%') | |
809 | setsize = SIZE_PERCENT; | |
e73a75fa RD |
810 | break; |
811 | } | |
1da177e4 | 812 | |
e73a75fa RD |
813 | case Opt_nr_inodes: |
814 | /* memparse() will accept a K/M/G without a digit */ | |
815 | if (!isdigit(*args[0].from)) | |
816 | goto bad_val; | |
817 | pconfig->nr_inodes = memparse(args[0].from, &rest); | |
818 | break; | |
819 | ||
a137e1cc AK |
820 | case Opt_pagesize: { |
821 | unsigned long ps; | |
822 | ps = memparse(args[0].from, &rest); | |
823 | pconfig->hstate = size_to_hstate(ps); | |
824 | if (!pconfig->hstate) { | |
825 | printk(KERN_ERR | |
826 | "hugetlbfs: Unsupported page size %lu MB\n", | |
827 | ps >> 20); | |
828 | return -EINVAL; | |
829 | } | |
830 | break; | |
831 | } | |
832 | ||
e73a75fa | 833 | default: |
b4c07bce LS |
834 | printk(KERN_ERR "hugetlbfs: Bad mount option: \"%s\"\n", |
835 | p); | |
836 | return -EINVAL; | |
e73a75fa RD |
837 | break; |
838 | } | |
1da177e4 | 839 | } |
a137e1cc AK |
840 | |
841 | /* Do size after hstate is set up */ | |
842 | if (setsize > NO_SIZE) { | |
843 | struct hstate *h = pconfig->hstate; | |
844 | if (setsize == SIZE_PERCENT) { | |
845 | size <<= huge_page_shift(h); | |
846 | size *= h->max_huge_pages; | |
847 | do_div(size, 100); | |
848 | } | |
849 | pconfig->nr_blocks = (size >> huge_page_shift(h)); | |
850 | } | |
851 | ||
1da177e4 | 852 | return 0; |
e73a75fa RD |
853 | |
854 | bad_val: | |
855 | printk(KERN_ERR "hugetlbfs: Bad value '%s' for mount option '%s'\n", | |
856 | args[0].from, p); | |
c12ddba0 | 857 | return -EINVAL; |
1da177e4 LT |
858 | } |
859 | ||
860 | static int | |
861 | hugetlbfs_fill_super(struct super_block *sb, void *data, int silent) | |
862 | { | |
1da177e4 LT |
863 | int ret; |
864 | struct hugetlbfs_config config; | |
865 | struct hugetlbfs_sb_info *sbinfo; | |
866 | ||
10f19a86 MS |
867 | save_mount_options(sb, data); |
868 | ||
1da177e4 LT |
869 | config.nr_blocks = -1; /* No limit on size by default */ |
870 | config.nr_inodes = -1; /* No limit on number of inodes by default */ | |
77c70de1 DH |
871 | config.uid = current_fsuid(); |
872 | config.gid = current_fsgid(); | |
1da177e4 | 873 | config.mode = 0755; |
a137e1cc | 874 | config.hstate = &default_hstate; |
1da177e4 | 875 | ret = hugetlbfs_parse_options(data, &config); |
1da177e4 LT |
876 | if (ret) |
877 | return ret; | |
878 | ||
879 | sbinfo = kmalloc(sizeof(struct hugetlbfs_sb_info), GFP_KERNEL); | |
880 | if (!sbinfo) | |
881 | return -ENOMEM; | |
882 | sb->s_fs_info = sbinfo; | |
a137e1cc | 883 | sbinfo->hstate = config.hstate; |
1da177e4 | 884 | spin_lock_init(&sbinfo->stat_lock); |
1da177e4 LT |
885 | sbinfo->max_inodes = config.nr_inodes; |
886 | sbinfo->free_inodes = config.nr_inodes; | |
90481622 DG |
887 | sbinfo->spool = NULL; |
888 | if (config.nr_blocks != -1) { | |
889 | sbinfo->spool = hugepage_new_subpool(config.nr_blocks); | |
890 | if (!sbinfo->spool) | |
891 | goto out_free; | |
892 | } | |
1da177e4 | 893 | sb->s_maxbytes = MAX_LFS_FILESIZE; |
a137e1cc AK |
894 | sb->s_blocksize = huge_page_size(config.hstate); |
895 | sb->s_blocksize_bits = huge_page_shift(config.hstate); | |
1da177e4 LT |
896 | sb->s_magic = HUGETLBFS_MAGIC; |
897 | sb->s_op = &hugetlbfs_ops; | |
898 | sb->s_time_gran = 1; | |
48fde701 AV |
899 | sb->s_root = d_make_root(hugetlbfs_get_root(sb, &config)); |
900 | if (!sb->s_root) | |
1da177e4 | 901 | goto out_free; |
1da177e4 LT |
902 | return 0; |
903 | out_free: | |
90481622 DG |
904 | if (sbinfo->spool) |
905 | kfree(sbinfo->spool); | |
1da177e4 LT |
906 | kfree(sbinfo); |
907 | return -ENOMEM; | |
908 | } | |
909 | ||
3c26ff6e AV |
910 | static struct dentry *hugetlbfs_mount(struct file_system_type *fs_type, |
911 | int flags, const char *dev_name, void *data) | |
1da177e4 | 912 | { |
3c26ff6e | 913 | return mount_nodev(fs_type, flags, data, hugetlbfs_fill_super); |
1da177e4 LT |
914 | } |
915 | ||
916 | static struct file_system_type hugetlbfs_fs_type = { | |
917 | .name = "hugetlbfs", | |
3c26ff6e | 918 | .mount = hugetlbfs_mount, |
1da177e4 LT |
919 | .kill_sb = kill_litter_super, |
920 | }; | |
921 | ||
922 | static struct vfsmount *hugetlbfs_vfsmount; | |
923 | ||
ef1ff6b8 | 924 | static int can_do_hugetlb_shm(void) |
1da177e4 | 925 | { |
ef1ff6b8 | 926 | return capable(CAP_IPC_LOCK) || in_group_p(sysctl_hugetlb_shm_group); |
1da177e4 LT |
927 | } |
928 | ||
40716e29 ST |
929 | struct file *hugetlb_file_setup(const char *name, unsigned long addr, |
930 | size_t size, vm_flags_t acctflag, | |
6bfde05b | 931 | struct user_struct **user, int creat_flags) |
1da177e4 LT |
932 | { |
933 | int error = -ENOMEM; | |
934 | struct file *file; | |
935 | struct inode *inode; | |
2c48b9c4 AV |
936 | struct path path; |
937 | struct dentry *root; | |
1da177e4 | 938 | struct qstr quick_string; |
40716e29 ST |
939 | struct hstate *hstate; |
940 | unsigned long num_pages; | |
1da177e4 | 941 | |
353d5c30 | 942 | *user = NULL; |
5bc98594 AM |
943 | if (!hugetlbfs_vfsmount) |
944 | return ERR_PTR(-ENOENT); | |
945 | ||
ef1ff6b8 | 946 | if (creat_flags == HUGETLB_SHMFS_INODE && !can_do_hugetlb_shm()) { |
353d5c30 HD |
947 | *user = current_user(); |
948 | if (user_shm_lock(size, *user)) { | |
21a3c273 DR |
949 | task_lock(current); |
950 | printk_once(KERN_WARNING | |
951 | "%s (%d): Using mlock ulimits for SHM_HUGETLB is deprecated\n", | |
952 | current->comm, current->pid); | |
953 | task_unlock(current); | |
353d5c30 HD |
954 | } else { |
955 | *user = NULL; | |
2584e517 | 956 | return ERR_PTR(-EPERM); |
353d5c30 | 957 | } |
2584e517 | 958 | } |
1da177e4 | 959 | |
1da177e4 | 960 | root = hugetlbfs_vfsmount->mnt_root; |
9d66586f | 961 | quick_string.name = name; |
1da177e4 LT |
962 | quick_string.len = strlen(quick_string.name); |
963 | quick_string.hash = 0; | |
2c48b9c4 AV |
964 | path.dentry = d_alloc(root, &quick_string); |
965 | if (!path.dentry) | |
1da177e4 LT |
966 | goto out_shm_unlock; |
967 | ||
2c48b9c4 | 968 | path.mnt = mntget(hugetlbfs_vfsmount); |
1da177e4 | 969 | error = -ENOSPC; |
7d54fa64 | 970 | inode = hugetlbfs_get_inode(root->d_sb, NULL, S_IFREG | S_IRWXUGO, 0); |
1da177e4 | 971 | if (!inode) |
ce8d2cdf | 972 | goto out_dentry; |
1da177e4 | 973 | |
40716e29 ST |
974 | hstate = hstate_inode(inode); |
975 | size += addr & ~huge_page_mask(hstate); | |
976 | num_pages = ALIGN(size, huge_page_size(hstate)) >> | |
977 | huge_page_shift(hstate); | |
b45b5bd6 | 978 | error = -ENOMEM; |
40716e29 | 979 | if (hugetlb_reserve_pages(inode, 0, num_pages, NULL, acctflag)) |
b45b5bd6 DG |
980 | goto out_inode; |
981 | ||
2c48b9c4 | 982 | d_instantiate(path.dentry, inode); |
1da177e4 | 983 | inode->i_size = size; |
6d6b77f1 | 984 | clear_nlink(inode); |
ce8d2cdf DH |
985 | |
986 | error = -ENFILE; | |
2c48b9c4 | 987 | file = alloc_file(&path, FMODE_WRITE | FMODE_READ, |
ce8d2cdf DH |
988 | &hugetlbfs_file_operations); |
989 | if (!file) | |
b4d232e6 | 990 | goto out_dentry; /* inode is already attached */ |
ce8d2cdf | 991 | |
1da177e4 LT |
992 | return file; |
993 | ||
b45b5bd6 DG |
994 | out_inode: |
995 | iput(inode); | |
1da177e4 | 996 | out_dentry: |
2c48b9c4 | 997 | path_put(&path); |
1da177e4 | 998 | out_shm_unlock: |
353d5c30 HD |
999 | if (*user) { |
1000 | user_shm_unlock(size, *user); | |
1001 | *user = NULL; | |
1002 | } | |
1da177e4 LT |
1003 | return ERR_PTR(error); |
1004 | } | |
1005 | ||
1006 | static int __init init_hugetlbfs_fs(void) | |
1007 | { | |
1008 | int error; | |
1009 | struct vfsmount *vfsmount; | |
1010 | ||
e0bf68dd PZ |
1011 | error = bdi_init(&hugetlbfs_backing_dev_info); |
1012 | if (error) | |
1013 | return error; | |
1014 | ||
d1d5e05f | 1015 | error = -ENOMEM; |
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: | |
d1d5e05f | 1036 | kmem_cache_destroy(hugetlbfs_inode_cachep); |
e0bf68dd PZ |
1037 | out2: |
1038 | bdi_destroy(&hugetlbfs_backing_dev_info); | |
1da177e4 LT |
1039 | return error; |
1040 | } | |
1041 | ||
1042 | static void __exit exit_hugetlbfs_fs(void) | |
1043 | { | |
1044 | kmem_cache_destroy(hugetlbfs_inode_cachep); | |
423e0ab0 | 1045 | kern_unmount(hugetlbfs_vfsmount); |
1da177e4 | 1046 | unregister_filesystem(&hugetlbfs_fs_type); |
e0bf68dd | 1047 | bdi_destroy(&hugetlbfs_backing_dev_info); |
1da177e4 LT |
1048 | } |
1049 | ||
1050 | module_init(init_hugetlbfs_fs) | |
1051 | module_exit(exit_hugetlbfs_fs) | |
1052 | ||
1053 | MODULE_LICENSE("GPL"); |