Commit | Line | Data |
---|---|---|
457c8996 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
1da177e4 LT |
2 | /* |
3 | * fs/libfs.c | |
4 | * Library for filesystems writers. | |
5 | */ | |
6 | ||
ac13a829 | 7 | #include <linux/blkdev.h> |
630d9c47 | 8 | #include <linux/export.h> |
1da177e4 | 9 | #include <linux/pagemap.h> |
5a0e3ad6 | 10 | #include <linux/slab.h> |
5b825c3a | 11 | #include <linux/cred.h> |
1da177e4 LT |
12 | #include <linux/mount.h> |
13 | #include <linux/vfs.h> | |
7bb46a67 | 14 | #include <linux/quotaops.h> |
7cf34c76 | 15 | #include <linux/mutex.h> |
87dc800b | 16 | #include <linux/namei.h> |
2596110a | 17 | #include <linux/exportfs.h> |
d5aacad5 | 18 | #include <linux/writeback.h> |
ff01bb48 | 19 | #include <linux/buffer_head.h> /* sync_mapping_buffers */ |
31d6d5ce DH |
20 | #include <linux/fs_context.h> |
21 | #include <linux/pseudo_fs.h> | |
a3d1e7eb | 22 | #include <linux/fsnotify.h> |
c843843e DR |
23 | #include <linux/unicode.h> |
24 | #include <linux/fscrypt.h> | |
7cf34c76 | 25 | |
7c0f6ba6 | 26 | #include <linux/uaccess.h> |
1da177e4 | 27 | |
a4464dbc AV |
28 | #include "internal.h" |
29 | ||
549c7297 CB |
30 | int simple_getattr(struct user_namespace *mnt_userns, const struct path *path, |
31 | struct kstat *stat, u32 request_mask, | |
32 | unsigned int query_flags) | |
1da177e4 | 33 | { |
a528d35e | 34 | struct inode *inode = d_inode(path->dentry); |
0d56a451 | 35 | generic_fillattr(&init_user_ns, inode, stat); |
09cbfeaf | 36 | stat->blocks = inode->i_mapping->nrpages << (PAGE_SHIFT - 9); |
1da177e4 LT |
37 | return 0; |
38 | } | |
12f38872 | 39 | EXPORT_SYMBOL(simple_getattr); |
1da177e4 | 40 | |
726c3342 | 41 | int simple_statfs(struct dentry *dentry, struct kstatfs *buf) |
1da177e4 | 42 | { |
726c3342 | 43 | buf->f_type = dentry->d_sb->s_magic; |
09cbfeaf | 44 | buf->f_bsize = PAGE_SIZE; |
1da177e4 LT |
45 | buf->f_namelen = NAME_MAX; |
46 | return 0; | |
47 | } | |
12f38872 | 48 | EXPORT_SYMBOL(simple_statfs); |
1da177e4 LT |
49 | |
50 | /* | |
51 | * Retaining negative dentries for an in-memory filesystem just wastes | |
52 | * memory and lookup time: arrange for them to be deleted immediately. | |
53 | */ | |
b26d4cd3 | 54 | int always_delete_dentry(const struct dentry *dentry) |
1da177e4 LT |
55 | { |
56 | return 1; | |
57 | } | |
b26d4cd3 AV |
58 | EXPORT_SYMBOL(always_delete_dentry); |
59 | ||
60 | const struct dentry_operations simple_dentry_operations = { | |
61 | .d_delete = always_delete_dentry, | |
62 | }; | |
63 | EXPORT_SYMBOL(simple_dentry_operations); | |
1da177e4 LT |
64 | |
65 | /* | |
66 | * Lookup the data. This is trivial - if the dentry didn't already | |
67 | * exist, we know it is negative. Set d_op to delete negative dentries. | |
68 | */ | |
00cd8dd3 | 69 | struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
1da177e4 | 70 | { |
1da177e4 LT |
71 | if (dentry->d_name.len > NAME_MAX) |
72 | return ERR_PTR(-ENAMETOOLONG); | |
74931da7 AV |
73 | if (!dentry->d_sb->s_d_op) |
74 | d_set_d_op(dentry, &simple_dentry_operations); | |
1da177e4 LT |
75 | d_add(dentry, NULL); |
76 | return NULL; | |
77 | } | |
12f38872 | 78 | EXPORT_SYMBOL(simple_lookup); |
1da177e4 | 79 | |
1da177e4 LT |
80 | int dcache_dir_open(struct inode *inode, struct file *file) |
81 | { | |
ba65dc5e | 82 | file->private_data = d_alloc_cursor(file->f_path.dentry); |
1da177e4 LT |
83 | |
84 | return file->private_data ? 0 : -ENOMEM; | |
85 | } | |
12f38872 | 86 | EXPORT_SYMBOL(dcache_dir_open); |
1da177e4 LT |
87 | |
88 | int dcache_dir_close(struct inode *inode, struct file *file) | |
89 | { | |
90 | dput(file->private_data); | |
91 | return 0; | |
92 | } | |
12f38872 | 93 | EXPORT_SYMBOL(dcache_dir_close); |
1da177e4 | 94 | |
4f42c1b5 | 95 | /* parent is locked at least shared */ |
d4f4de5e AV |
96 | /* |
97 | * Returns an element of siblings' list. | |
98 | * We are looking for <count>th positive after <p>; if | |
26b6c984 AV |
99 | * found, dentry is grabbed and returned to caller. |
100 | * If no such element exists, NULL is returned. | |
d4f4de5e | 101 | */ |
26b6c984 | 102 | static struct dentry *scan_positives(struct dentry *cursor, |
d4f4de5e AV |
103 | struct list_head *p, |
104 | loff_t count, | |
26b6c984 | 105 | struct dentry *last) |
4f42c1b5 | 106 | { |
d4f4de5e AV |
107 | struct dentry *dentry = cursor->d_parent, *found = NULL; |
108 | ||
109 | spin_lock(&dentry->d_lock); | |
110 | while ((p = p->next) != &dentry->d_subdirs) { | |
4f42c1b5 | 111 | struct dentry *d = list_entry(p, struct dentry, d_child); |
d4f4de5e AV |
112 | // we must at least skip cursors, to avoid livelocks |
113 | if (d->d_flags & DCACHE_DENTRY_CURSOR) | |
114 | continue; | |
115 | if (simple_positive(d) && !--count) { | |
116 | spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED); | |
117 | if (simple_positive(d)) | |
118 | found = dget_dlock(d); | |
119 | spin_unlock(&d->d_lock); | |
120 | if (likely(found)) | |
121 | break; | |
122 | count = 1; | |
123 | } | |
124 | if (need_resched()) { | |
125 | list_move(&cursor->d_child, p); | |
126 | p = &cursor->d_child; | |
127 | spin_unlock(&dentry->d_lock); | |
128 | cond_resched(); | |
129 | spin_lock(&dentry->d_lock); | |
4f42c1b5 AV |
130 | } |
131 | } | |
d4f4de5e | 132 | spin_unlock(&dentry->d_lock); |
26b6c984 AV |
133 | dput(last); |
134 | return found; | |
4f42c1b5 AV |
135 | } |
136 | ||
965c8e59 | 137 | loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence) |
1da177e4 | 138 | { |
2fd6b7f5 | 139 | struct dentry *dentry = file->f_path.dentry; |
965c8e59 | 140 | switch (whence) { |
1da177e4 LT |
141 | case 1: |
142 | offset += file->f_pos; | |
df561f66 | 143 | fallthrough; |
1da177e4 LT |
144 | case 0: |
145 | if (offset >= 0) | |
146 | break; | |
df561f66 | 147 | fallthrough; |
1da177e4 | 148 | default: |
1da177e4 LT |
149 | return -EINVAL; |
150 | } | |
151 | if (offset != file->f_pos) { | |
d4f4de5e AV |
152 | struct dentry *cursor = file->private_data; |
153 | struct dentry *to = NULL; | |
d4f4de5e | 154 | |
d4f4de5e AV |
155 | inode_lock_shared(dentry->d_inode); |
156 | ||
26b6c984 AV |
157 | if (offset > 2) |
158 | to = scan_positives(cursor, &dentry->d_subdirs, | |
159 | offset - 2, NULL); | |
160 | spin_lock(&dentry->d_lock); | |
161 | if (to) | |
162 | list_move(&cursor->d_child, &to->d_child); | |
163 | else | |
d4f4de5e | 164 | list_del_init(&cursor->d_child); |
26b6c984 | 165 | spin_unlock(&dentry->d_lock); |
d4f4de5e AV |
166 | dput(to); |
167 | ||
26b6c984 AV |
168 | file->f_pos = offset; |
169 | ||
d4f4de5e | 170 | inode_unlock_shared(dentry->d_inode); |
1da177e4 | 171 | } |
1da177e4 LT |
172 | return offset; |
173 | } | |
12f38872 | 174 | EXPORT_SYMBOL(dcache_dir_lseek); |
1da177e4 LT |
175 | |
176 | /* Relationship between i_mode and the DT_xxx types */ | |
177 | static inline unsigned char dt_type(struct inode *inode) | |
178 | { | |
179 | return (inode->i_mode >> 12) & 15; | |
180 | } | |
181 | ||
182 | /* | |
183 | * Directory is locked and all positive dentries in it are safe, since | |
184 | * for ramfs-type trees they can't go away without unlink() or rmdir(), | |
185 | * both impossible due to the lock on directory. | |
186 | */ | |
187 | ||
5f99f4e7 | 188 | int dcache_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 189 | { |
5f99f4e7 AV |
190 | struct dentry *dentry = file->f_path.dentry; |
191 | struct dentry *cursor = file->private_data; | |
d4f4de5e AV |
192 | struct list_head *anchor = &dentry->d_subdirs; |
193 | struct dentry *next = NULL; | |
194 | struct list_head *p; | |
1da177e4 | 195 | |
5f99f4e7 AV |
196 | if (!dir_emit_dots(file, ctx)) |
197 | return 0; | |
5f99f4e7 | 198 | |
4f42c1b5 | 199 | if (ctx->pos == 2) |
d4f4de5e | 200 | p = anchor; |
26b6c984 | 201 | else if (!list_empty(&cursor->d_child)) |
d4f4de5e | 202 | p = &cursor->d_child; |
26b6c984 AV |
203 | else |
204 | return 0; | |
d4f4de5e | 205 | |
26b6c984 | 206 | while ((next = scan_positives(cursor, p, 1, next)) != NULL) { |
5f99f4e7 | 207 | if (!dir_emit(ctx, next->d_name.name, next->d_name.len, |
dea655c2 | 208 | d_inode(next)->i_ino, dt_type(d_inode(next)))) |
4f42c1b5 | 209 | break; |
5f99f4e7 | 210 | ctx->pos++; |
26b6c984 | 211 | p = &next->d_child; |
1da177e4 | 212 | } |
d4f4de5e | 213 | spin_lock(&dentry->d_lock); |
26b6c984 AV |
214 | if (next) |
215 | list_move_tail(&cursor->d_child, &next->d_child); | |
216 | else | |
217 | list_del_init(&cursor->d_child); | |
d4f4de5e AV |
218 | spin_unlock(&dentry->d_lock); |
219 | dput(next); | |
220 | ||
1da177e4 LT |
221 | return 0; |
222 | } | |
12f38872 | 223 | EXPORT_SYMBOL(dcache_readdir); |
1da177e4 LT |
224 | |
225 | ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos) | |
226 | { | |
227 | return -EISDIR; | |
228 | } | |
12f38872 | 229 | EXPORT_SYMBOL(generic_read_dir); |
1da177e4 | 230 | |
4b6f5d20 | 231 | const struct file_operations simple_dir_operations = { |
1da177e4 LT |
232 | .open = dcache_dir_open, |
233 | .release = dcache_dir_close, | |
234 | .llseek = dcache_dir_lseek, | |
235 | .read = generic_read_dir, | |
4e82901c | 236 | .iterate_shared = dcache_readdir, |
1b061d92 | 237 | .fsync = noop_fsync, |
1da177e4 | 238 | }; |
12f38872 | 239 | EXPORT_SYMBOL(simple_dir_operations); |
1da177e4 | 240 | |
92e1d5be | 241 | const struct inode_operations simple_dir_inode_operations = { |
1da177e4 LT |
242 | .lookup = simple_lookup, |
243 | }; | |
12f38872 | 244 | EXPORT_SYMBOL(simple_dir_inode_operations); |
1da177e4 | 245 | |
a3d1e7eb AV |
246 | static struct dentry *find_next_child(struct dentry *parent, struct dentry *prev) |
247 | { | |
248 | struct dentry *child = NULL; | |
249 | struct list_head *p = prev ? &prev->d_child : &parent->d_subdirs; | |
250 | ||
251 | spin_lock(&parent->d_lock); | |
252 | while ((p = p->next) != &parent->d_subdirs) { | |
253 | struct dentry *d = container_of(p, struct dentry, d_child); | |
254 | if (simple_positive(d)) { | |
255 | spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED); | |
256 | if (simple_positive(d)) | |
257 | child = dget_dlock(d); | |
258 | spin_unlock(&d->d_lock); | |
259 | if (likely(child)) | |
260 | break; | |
261 | } | |
262 | } | |
263 | spin_unlock(&parent->d_lock); | |
264 | dput(prev); | |
265 | return child; | |
266 | } | |
267 | ||
268 | void simple_recursive_removal(struct dentry *dentry, | |
269 | void (*callback)(struct dentry *)) | |
270 | { | |
271 | struct dentry *this = dget(dentry); | |
272 | while (true) { | |
273 | struct dentry *victim = NULL, *child; | |
274 | struct inode *inode = this->d_inode; | |
275 | ||
276 | inode_lock(inode); | |
277 | if (d_is_dir(this)) | |
278 | inode->i_flags |= S_DEAD; | |
279 | while ((child = find_next_child(this, victim)) == NULL) { | |
280 | // kill and ascend | |
281 | // update metadata while it's still locked | |
282 | inode->i_ctime = current_time(inode); | |
283 | clear_nlink(inode); | |
284 | inode_unlock(inode); | |
285 | victim = this; | |
286 | this = this->d_parent; | |
287 | inode = this->d_inode; | |
288 | inode_lock(inode); | |
289 | if (simple_positive(victim)) { | |
290 | d_invalidate(victim); // avoid lost mounts | |
291 | if (d_is_dir(victim)) | |
292 | fsnotify_rmdir(inode, victim); | |
293 | else | |
294 | fsnotify_unlink(inode, victim); | |
295 | if (callback) | |
296 | callback(victim); | |
297 | dput(victim); // unpin it | |
298 | } | |
299 | if (victim == dentry) { | |
300 | inode->i_ctime = inode->i_mtime = | |
301 | current_time(inode); | |
302 | if (d_is_dir(dentry)) | |
303 | drop_nlink(inode); | |
304 | inode_unlock(inode); | |
305 | dput(dentry); | |
306 | return; | |
307 | } | |
308 | } | |
309 | inode_unlock(inode); | |
310 | this = child; | |
311 | } | |
312 | } | |
313 | EXPORT_SYMBOL(simple_recursive_removal); | |
314 | ||
759b9775 HD |
315 | static const struct super_operations simple_super_operations = { |
316 | .statfs = simple_statfs, | |
317 | }; | |
318 | ||
db2c246a | 319 | static int pseudo_fs_fill_super(struct super_block *s, struct fs_context *fc) |
1da177e4 | 320 | { |
31d6d5ce | 321 | struct pseudo_fs_context *ctx = fc->fs_private; |
1da177e4 | 322 | struct inode *root; |
1da177e4 | 323 | |
89a4eb4b | 324 | s->s_maxbytes = MAX_LFS_FILESIZE; |
3971e1a9 AN |
325 | s->s_blocksize = PAGE_SIZE; |
326 | s->s_blocksize_bits = PAGE_SHIFT; | |
8d9e46d8 AV |
327 | s->s_magic = ctx->magic; |
328 | s->s_op = ctx->ops ?: &simple_super_operations; | |
329 | s->s_xattr = ctx->xattr; | |
1da177e4 LT |
330 | s->s_time_gran = 1; |
331 | root = new_inode(s); | |
332 | if (!root) | |
db2c246a DH |
333 | return -ENOMEM; |
334 | ||
1a1c9bb4 JL |
335 | /* |
336 | * since this is the first inode, make it number 1. New inodes created | |
337 | * after this must take care not to collide with it (by passing | |
338 | * max_reserved of 1 to iunique). | |
339 | */ | |
340 | root->i_ino = 1; | |
1da177e4 | 341 | root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR; |
078cd827 | 342 | root->i_atime = root->i_mtime = root->i_ctime = current_time(root); |
8d9e46d8 AV |
343 | s->s_root = d_make_root(root); |
344 | if (!s->s_root) | |
db2c246a | 345 | return -ENOMEM; |
8d9e46d8 | 346 | s->s_d_op = ctx->dops; |
31d6d5ce | 347 | return 0; |
db2c246a | 348 | } |
8d9e46d8 | 349 | |
db2c246a DH |
350 | static int pseudo_fs_get_tree(struct fs_context *fc) |
351 | { | |
2ac295d4 | 352 | return get_tree_nodev(fc, pseudo_fs_fill_super); |
31d6d5ce DH |
353 | } |
354 | ||
355 | static void pseudo_fs_free(struct fs_context *fc) | |
356 | { | |
357 | kfree(fc->fs_private); | |
358 | } | |
359 | ||
360 | static const struct fs_context_operations pseudo_fs_context_ops = { | |
361 | .free = pseudo_fs_free, | |
362 | .get_tree = pseudo_fs_get_tree, | |
363 | }; | |
364 | ||
365 | /* | |
366 | * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that | |
367 | * will never be mountable) | |
368 | */ | |
369 | struct pseudo_fs_context *init_pseudo(struct fs_context *fc, | |
370 | unsigned long magic) | |
371 | { | |
372 | struct pseudo_fs_context *ctx; | |
373 | ||
374 | ctx = kzalloc(sizeof(struct pseudo_fs_context), GFP_KERNEL); | |
375 | if (likely(ctx)) { | |
376 | ctx->magic = magic; | |
377 | fc->fs_private = ctx; | |
378 | fc->ops = &pseudo_fs_context_ops; | |
db2c246a DH |
379 | fc->sb_flags |= SB_NOUSER; |
380 | fc->global = true; | |
1da177e4 | 381 | } |
31d6d5ce | 382 | return ctx; |
1da177e4 | 383 | } |
31d6d5ce | 384 | EXPORT_SYMBOL(init_pseudo); |
1da177e4 | 385 | |
20955e89 SB |
386 | int simple_open(struct inode *inode, struct file *file) |
387 | { | |
388 | if (inode->i_private) | |
389 | file->private_data = inode->i_private; | |
390 | return 0; | |
391 | } | |
12f38872 | 392 | EXPORT_SYMBOL(simple_open); |
20955e89 | 393 | |
1da177e4 LT |
394 | int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry) |
395 | { | |
dea655c2 | 396 | struct inode *inode = d_inode(old_dentry); |
1da177e4 | 397 | |
078cd827 | 398 | inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode); |
d8c76e6f | 399 | inc_nlink(inode); |
7de9c6ee | 400 | ihold(inode); |
1da177e4 LT |
401 | dget(dentry); |
402 | d_instantiate(dentry, inode); | |
403 | return 0; | |
404 | } | |
12f38872 | 405 | EXPORT_SYMBOL(simple_link); |
1da177e4 | 406 | |
1da177e4 LT |
407 | int simple_empty(struct dentry *dentry) |
408 | { | |
409 | struct dentry *child; | |
410 | int ret = 0; | |
411 | ||
2fd6b7f5 | 412 | spin_lock(&dentry->d_lock); |
946e51f2 | 413 | list_for_each_entry(child, &dentry->d_subdirs, d_child) { |
da502956 NP |
414 | spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED); |
415 | if (simple_positive(child)) { | |
416 | spin_unlock(&child->d_lock); | |
1da177e4 | 417 | goto out; |
da502956 NP |
418 | } |
419 | spin_unlock(&child->d_lock); | |
420 | } | |
1da177e4 LT |
421 | ret = 1; |
422 | out: | |
2fd6b7f5 | 423 | spin_unlock(&dentry->d_lock); |
1da177e4 LT |
424 | return ret; |
425 | } | |
12f38872 | 426 | EXPORT_SYMBOL(simple_empty); |
1da177e4 LT |
427 | |
428 | int simple_unlink(struct inode *dir, struct dentry *dentry) | |
429 | { | |
dea655c2 | 430 | struct inode *inode = d_inode(dentry); |
1da177e4 | 431 | |
078cd827 | 432 | inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode); |
9a53c3a7 | 433 | drop_nlink(inode); |
1da177e4 LT |
434 | dput(dentry); |
435 | return 0; | |
436 | } | |
12f38872 | 437 | EXPORT_SYMBOL(simple_unlink); |
1da177e4 LT |
438 | |
439 | int simple_rmdir(struct inode *dir, struct dentry *dentry) | |
440 | { | |
441 | if (!simple_empty(dentry)) | |
442 | return -ENOTEMPTY; | |
443 | ||
dea655c2 | 444 | drop_nlink(d_inode(dentry)); |
1da177e4 | 445 | simple_unlink(dir, dentry); |
9a53c3a7 | 446 | drop_nlink(dir); |
1da177e4 LT |
447 | return 0; |
448 | } | |
12f38872 | 449 | EXPORT_SYMBOL(simple_rmdir); |
1da177e4 | 450 | |
549c7297 CB |
451 | int simple_rename(struct user_namespace *mnt_userns, struct inode *old_dir, |
452 | struct dentry *old_dentry, struct inode *new_dir, | |
453 | struct dentry *new_dentry, unsigned int flags) | |
1da177e4 | 454 | { |
dea655c2 | 455 | struct inode *inode = d_inode(old_dentry); |
e36cb0b8 | 456 | int they_are_dirs = d_is_dir(old_dentry); |
1da177e4 | 457 | |
e0e0be8a MS |
458 | if (flags & ~RENAME_NOREPLACE) |
459 | return -EINVAL; | |
460 | ||
1da177e4 LT |
461 | if (!simple_empty(new_dentry)) |
462 | return -ENOTEMPTY; | |
463 | ||
dea655c2 | 464 | if (d_really_is_positive(new_dentry)) { |
1da177e4 | 465 | simple_unlink(new_dir, new_dentry); |
841590ce | 466 | if (they_are_dirs) { |
dea655c2 | 467 | drop_nlink(d_inode(new_dentry)); |
9a53c3a7 | 468 | drop_nlink(old_dir); |
841590ce | 469 | } |
1da177e4 | 470 | } else if (they_are_dirs) { |
9a53c3a7 | 471 | drop_nlink(old_dir); |
d8c76e6f | 472 | inc_nlink(new_dir); |
1da177e4 LT |
473 | } |
474 | ||
475 | old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime = | |
078cd827 | 476 | new_dir->i_mtime = inode->i_ctime = current_time(old_dir); |
1da177e4 LT |
477 | |
478 | return 0; | |
479 | } | |
12f38872 | 480 | EXPORT_SYMBOL(simple_rename); |
1da177e4 | 481 | |
7bb46a67 | 482 | /** |
eef2380c | 483 | * simple_setattr - setattr for simple filesystem |
59347d99 | 484 | * @mnt_userns: user namespace of the target mount |
7bb46a67 | 485 | * @dentry: dentry |
486 | * @iattr: iattr structure | |
487 | * | |
488 | * Returns 0 on success, -error on failure. | |
489 | * | |
eef2380c CH |
490 | * simple_setattr is a simple ->setattr implementation without a proper |
491 | * implementation of size changes. | |
492 | * | |
493 | * It can either be used for in-memory filesystems or special files | |
494 | * on simple regular filesystems. Anything that needs to change on-disk | |
495 | * or wire state on size changes needs its own setattr method. | |
7bb46a67 | 496 | */ |
549c7297 CB |
497 | int simple_setattr(struct user_namespace *mnt_userns, struct dentry *dentry, |
498 | struct iattr *iattr) | |
7bb46a67 | 499 | { |
dea655c2 | 500 | struct inode *inode = d_inode(dentry); |
7bb46a67 | 501 | int error; |
502 | ||
549c7297 | 503 | error = setattr_prepare(mnt_userns, dentry, iattr); |
7bb46a67 | 504 | if (error) |
505 | return error; | |
506 | ||
2c27c65e CH |
507 | if (iattr->ia_valid & ATTR_SIZE) |
508 | truncate_setsize(inode, iattr->ia_size); | |
549c7297 | 509 | setattr_copy(mnt_userns, inode, iattr); |
eef2380c CH |
510 | mark_inode_dirty(inode); |
511 | return 0; | |
7bb46a67 | 512 | } |
513 | EXPORT_SYMBOL(simple_setattr); | |
514 | ||
1da177e4 LT |
515 | int simple_readpage(struct file *file, struct page *page) |
516 | { | |
c0d92cbc | 517 | clear_highpage(page); |
1da177e4 LT |
518 | flush_dcache_page(page); |
519 | SetPageUptodate(page); | |
1da177e4 LT |
520 | unlock_page(page); |
521 | return 0; | |
522 | } | |
12f38872 | 523 | EXPORT_SYMBOL(simple_readpage); |
1da177e4 | 524 | |
afddba49 NP |
525 | int simple_write_begin(struct file *file, struct address_space *mapping, |
526 | loff_t pos, unsigned len, unsigned flags, | |
527 | struct page **pagep, void **fsdata) | |
528 | { | |
529 | struct page *page; | |
530 | pgoff_t index; | |
afddba49 | 531 | |
09cbfeaf | 532 | index = pos >> PAGE_SHIFT; |
afddba49 | 533 | |
54566b2c | 534 | page = grab_cache_page_write_begin(mapping, index, flags); |
afddba49 NP |
535 | if (!page) |
536 | return -ENOMEM; | |
537 | ||
538 | *pagep = page; | |
539 | ||
09cbfeaf KS |
540 | if (!PageUptodate(page) && (len != PAGE_SIZE)) { |
541 | unsigned from = pos & (PAGE_SIZE - 1); | |
193cf4b9 | 542 | |
09cbfeaf | 543 | zero_user_segments(page, 0, from, from + len, PAGE_SIZE); |
193cf4b9 BH |
544 | } |
545 | return 0; | |
afddba49 | 546 | } |
12f38872 | 547 | EXPORT_SYMBOL(simple_write_begin); |
afddba49 | 548 | |
ad2a722f BH |
549 | /** |
550 | * simple_write_end - .write_end helper for non-block-device FSes | |
8e88bfba | 551 | * @file: See .write_end of address_space_operations |
ad2a722f BH |
552 | * @mapping: " |
553 | * @pos: " | |
554 | * @len: " | |
555 | * @copied: " | |
556 | * @page: " | |
557 | * @fsdata: " | |
558 | * | |
559 | * simple_write_end does the minimum needed for updating a page after writing is | |
560 | * done. It has the same API signature as the .write_end of | |
561 | * address_space_operations vector. So it can just be set onto .write_end for | |
562 | * FSes that don't need any other processing. i_mutex is assumed to be held. | |
563 | * Block based filesystems should use generic_write_end(). | |
564 | * NOTE: Even though i_size might get updated by this function, mark_inode_dirty | |
565 | * is not called, so a filesystem that actually does store data in .write_inode | |
566 | * should extend on what's done here with a call to mark_inode_dirty() in the | |
567 | * case that i_size has changed. | |
04fff641 AV |
568 | * |
569 | * Use *ONLY* with simple_readpage() | |
ad2a722f | 570 | */ |
afddba49 NP |
571 | int simple_write_end(struct file *file, struct address_space *mapping, |
572 | loff_t pos, unsigned len, unsigned copied, | |
573 | struct page *page, void *fsdata) | |
574 | { | |
ad2a722f BH |
575 | struct inode *inode = page->mapping->host; |
576 | loff_t last_pos = pos + copied; | |
afddba49 NP |
577 | |
578 | /* zero the stale part of the page if we did a short copy */ | |
04fff641 AV |
579 | if (!PageUptodate(page)) { |
580 | if (copied < len) { | |
581 | unsigned from = pos & (PAGE_SIZE - 1); | |
afddba49 | 582 | |
04fff641 AV |
583 | zero_user(page, from + copied, len - copied); |
584 | } | |
ad2a722f | 585 | SetPageUptodate(page); |
04fff641 | 586 | } |
ad2a722f BH |
587 | /* |
588 | * No need to use i_size_read() here, the i_size | |
589 | * cannot change under us because we hold the i_mutex. | |
590 | */ | |
591 | if (last_pos > inode->i_size) | |
592 | i_size_write(inode, last_pos); | |
afddba49 | 593 | |
ad2a722f | 594 | set_page_dirty(page); |
afddba49 | 595 | unlock_page(page); |
09cbfeaf | 596 | put_page(page); |
afddba49 NP |
597 | |
598 | return copied; | |
599 | } | |
12f38872 | 600 | EXPORT_SYMBOL(simple_write_end); |
afddba49 | 601 | |
1a1c9bb4 JL |
602 | /* |
603 | * the inodes created here are not hashed. If you use iunique to generate | |
604 | * unique inode values later for this filesystem, then you must take care | |
605 | * to pass it an appropriate max_reserved value to avoid collisions. | |
606 | */ | |
7d683a09 | 607 | int simple_fill_super(struct super_block *s, unsigned long magic, |
cda37124 | 608 | const struct tree_descr *files) |
1da177e4 | 609 | { |
1da177e4 LT |
610 | struct inode *inode; |
611 | struct dentry *root; | |
612 | struct dentry *dentry; | |
613 | int i; | |
614 | ||
09cbfeaf KS |
615 | s->s_blocksize = PAGE_SIZE; |
616 | s->s_blocksize_bits = PAGE_SHIFT; | |
1da177e4 | 617 | s->s_magic = magic; |
759b9775 | 618 | s->s_op = &simple_super_operations; |
1da177e4 LT |
619 | s->s_time_gran = 1; |
620 | ||
621 | inode = new_inode(s); | |
622 | if (!inode) | |
623 | return -ENOMEM; | |
1a1c9bb4 JL |
624 | /* |
625 | * because the root inode is 1, the files array must not contain an | |
626 | * entry at index 1 | |
627 | */ | |
628 | inode->i_ino = 1; | |
1da177e4 | 629 | inode->i_mode = S_IFDIR | 0755; |
078cd827 | 630 | inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode); |
1da177e4 LT |
631 | inode->i_op = &simple_dir_inode_operations; |
632 | inode->i_fop = &simple_dir_operations; | |
bfe86848 | 633 | set_nlink(inode, 2); |
48fde701 AV |
634 | root = d_make_root(inode); |
635 | if (!root) | |
1da177e4 | 636 | return -ENOMEM; |
1da177e4 LT |
637 | for (i = 0; !files->name || files->name[0]; i++, files++) { |
638 | if (!files->name) | |
639 | continue; | |
1a1c9bb4 JL |
640 | |
641 | /* warn if it tries to conflict with the root inode */ | |
642 | if (unlikely(i == 1)) | |
643 | printk(KERN_WARNING "%s: %s passed in a files array" | |
644 | "with an index of 1!\n", __func__, | |
645 | s->s_type->name); | |
646 | ||
1da177e4 LT |
647 | dentry = d_alloc_name(root, files->name); |
648 | if (!dentry) | |
649 | goto out; | |
650 | inode = new_inode(s); | |
32096ea1 KK |
651 | if (!inode) { |
652 | dput(dentry); | |
1da177e4 | 653 | goto out; |
32096ea1 | 654 | } |
1da177e4 | 655 | inode->i_mode = S_IFREG | files->mode; |
078cd827 | 656 | inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode); |
1da177e4 LT |
657 | inode->i_fop = files->ops; |
658 | inode->i_ino = i; | |
659 | d_add(dentry, inode); | |
660 | } | |
661 | s->s_root = root; | |
662 | return 0; | |
663 | out: | |
664 | d_genocide(root); | |
640946f2 | 665 | shrink_dcache_parent(root); |
1da177e4 LT |
666 | dput(root); |
667 | return -ENOMEM; | |
668 | } | |
12f38872 | 669 | EXPORT_SYMBOL(simple_fill_super); |
1da177e4 LT |
670 | |
671 | static DEFINE_SPINLOCK(pin_fs_lock); | |
672 | ||
1f5ce9e9 | 673 | int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count) |
1da177e4 LT |
674 | { |
675 | struct vfsmount *mnt = NULL; | |
676 | spin_lock(&pin_fs_lock); | |
677 | if (unlikely(!*mount)) { | |
678 | spin_unlock(&pin_fs_lock); | |
1751e8a6 | 679 | mnt = vfs_kern_mount(type, SB_KERNMOUNT, type->name, NULL); |
1da177e4 LT |
680 | if (IS_ERR(mnt)) |
681 | return PTR_ERR(mnt); | |
682 | spin_lock(&pin_fs_lock); | |
683 | if (!*mount) | |
684 | *mount = mnt; | |
685 | } | |
686 | mntget(*mount); | |
687 | ++*count; | |
688 | spin_unlock(&pin_fs_lock); | |
689 | mntput(mnt); | |
690 | return 0; | |
691 | } | |
12f38872 | 692 | EXPORT_SYMBOL(simple_pin_fs); |
1da177e4 LT |
693 | |
694 | void simple_release_fs(struct vfsmount **mount, int *count) | |
695 | { | |
696 | struct vfsmount *mnt; | |
697 | spin_lock(&pin_fs_lock); | |
698 | mnt = *mount; | |
699 | if (!--*count) | |
700 | *mount = NULL; | |
701 | spin_unlock(&pin_fs_lock); | |
702 | mntput(mnt); | |
703 | } | |
12f38872 | 704 | EXPORT_SYMBOL(simple_release_fs); |
1da177e4 | 705 | |
6d1029b5 AM |
706 | /** |
707 | * simple_read_from_buffer - copy data from the buffer to user space | |
708 | * @to: the user space buffer to read to | |
709 | * @count: the maximum number of bytes to read | |
710 | * @ppos: the current position in the buffer | |
711 | * @from: the buffer to read from | |
712 | * @available: the size of the buffer | |
713 | * | |
714 | * The simple_read_from_buffer() function reads up to @count bytes from the | |
715 | * buffer @from at offset @ppos into the user space address starting at @to. | |
716 | * | |
717 | * On success, the number of bytes read is returned and the offset @ppos is | |
718 | * advanced by this number, or negative value is returned on error. | |
719 | **/ | |
1da177e4 LT |
720 | ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos, |
721 | const void *from, size_t available) | |
722 | { | |
723 | loff_t pos = *ppos; | |
14be2746 SR |
724 | size_t ret; |
725 | ||
1da177e4 LT |
726 | if (pos < 0) |
727 | return -EINVAL; | |
14be2746 | 728 | if (pos >= available || !count) |
1da177e4 LT |
729 | return 0; |
730 | if (count > available - pos) | |
731 | count = available - pos; | |
14be2746 SR |
732 | ret = copy_to_user(to, from + pos, count); |
733 | if (ret == count) | |
1da177e4 | 734 | return -EFAULT; |
14be2746 | 735 | count -= ret; |
1da177e4 LT |
736 | *ppos = pos + count; |
737 | return count; | |
738 | } | |
12f38872 | 739 | EXPORT_SYMBOL(simple_read_from_buffer); |
1da177e4 | 740 | |
6a727b43 JS |
741 | /** |
742 | * simple_write_to_buffer - copy data from user space to the buffer | |
743 | * @to: the buffer to write to | |
744 | * @available: the size of the buffer | |
745 | * @ppos: the current position in the buffer | |
746 | * @from: the user space buffer to read from | |
747 | * @count: the maximum number of bytes to read | |
748 | * | |
749 | * The simple_write_to_buffer() function reads up to @count bytes from the user | |
750 | * space address starting at @from into the buffer @to at offset @ppos. | |
751 | * | |
752 | * On success, the number of bytes written is returned and the offset @ppos is | |
753 | * advanced by this number, or negative value is returned on error. | |
754 | **/ | |
755 | ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos, | |
756 | const void __user *from, size_t count) | |
757 | { | |
758 | loff_t pos = *ppos; | |
759 | size_t res; | |
760 | ||
761 | if (pos < 0) | |
762 | return -EINVAL; | |
763 | if (pos >= available || !count) | |
764 | return 0; | |
765 | if (count > available - pos) | |
766 | count = available - pos; | |
767 | res = copy_from_user(to + pos, from, count); | |
768 | if (res == count) | |
769 | return -EFAULT; | |
770 | count -= res; | |
771 | *ppos = pos + count; | |
772 | return count; | |
773 | } | |
12f38872 | 774 | EXPORT_SYMBOL(simple_write_to_buffer); |
6a727b43 | 775 | |
6d1029b5 AM |
776 | /** |
777 | * memory_read_from_buffer - copy data from the buffer | |
778 | * @to: the kernel space buffer to read to | |
779 | * @count: the maximum number of bytes to read | |
780 | * @ppos: the current position in the buffer | |
781 | * @from: the buffer to read from | |
782 | * @available: the size of the buffer | |
783 | * | |
784 | * The memory_read_from_buffer() function reads up to @count bytes from the | |
785 | * buffer @from at offset @ppos into the kernel space address starting at @to. | |
786 | * | |
787 | * On success, the number of bytes read is returned and the offset @ppos is | |
788 | * advanced by this number, or negative value is returned on error. | |
789 | **/ | |
93b07113 AM |
790 | ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos, |
791 | const void *from, size_t available) | |
792 | { | |
793 | loff_t pos = *ppos; | |
794 | ||
795 | if (pos < 0) | |
796 | return -EINVAL; | |
797 | if (pos >= available) | |
798 | return 0; | |
799 | if (count > available - pos) | |
800 | count = available - pos; | |
801 | memcpy(to, from + pos, count); | |
802 | *ppos = pos + count; | |
803 | ||
804 | return count; | |
805 | } | |
12f38872 | 806 | EXPORT_SYMBOL(memory_read_from_buffer); |
93b07113 | 807 | |
1da177e4 LT |
808 | /* |
809 | * Transaction based IO. | |
810 | * The file expects a single write which triggers the transaction, and then | |
811 | * possibly a read which collects the result - which is stored in a | |
812 | * file-local buffer. | |
813 | */ | |
76791ab2 IM |
814 | |
815 | void simple_transaction_set(struct file *file, size_t n) | |
816 | { | |
817 | struct simple_transaction_argresp *ar = file->private_data; | |
818 | ||
819 | BUG_ON(n > SIMPLE_TRANSACTION_LIMIT); | |
820 | ||
821 | /* | |
822 | * The barrier ensures that ar->size will really remain zero until | |
823 | * ar->data is ready for reading. | |
824 | */ | |
825 | smp_mb(); | |
826 | ar->size = n; | |
827 | } | |
12f38872 | 828 | EXPORT_SYMBOL(simple_transaction_set); |
76791ab2 | 829 | |
1da177e4 LT |
830 | char *simple_transaction_get(struct file *file, const char __user *buf, size_t size) |
831 | { | |
832 | struct simple_transaction_argresp *ar; | |
833 | static DEFINE_SPINLOCK(simple_transaction_lock); | |
834 | ||
835 | if (size > SIMPLE_TRANSACTION_LIMIT - 1) | |
836 | return ERR_PTR(-EFBIG); | |
837 | ||
838 | ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL); | |
839 | if (!ar) | |
840 | return ERR_PTR(-ENOMEM); | |
841 | ||
842 | spin_lock(&simple_transaction_lock); | |
843 | ||
844 | /* only one write allowed per open */ | |
845 | if (file->private_data) { | |
846 | spin_unlock(&simple_transaction_lock); | |
847 | free_page((unsigned long)ar); | |
848 | return ERR_PTR(-EBUSY); | |
849 | } | |
850 | ||
851 | file->private_data = ar; | |
852 | ||
853 | spin_unlock(&simple_transaction_lock); | |
854 | ||
855 | if (copy_from_user(ar->data, buf, size)) | |
856 | return ERR_PTR(-EFAULT); | |
857 | ||
858 | return ar->data; | |
859 | } | |
12f38872 | 860 | EXPORT_SYMBOL(simple_transaction_get); |
1da177e4 LT |
861 | |
862 | ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos) | |
863 | { | |
864 | struct simple_transaction_argresp *ar = file->private_data; | |
865 | ||
866 | if (!ar) | |
867 | return 0; | |
868 | return simple_read_from_buffer(buf, size, pos, ar->data, ar->size); | |
869 | } | |
12f38872 | 870 | EXPORT_SYMBOL(simple_transaction_read); |
1da177e4 LT |
871 | |
872 | int simple_transaction_release(struct inode *inode, struct file *file) | |
873 | { | |
874 | free_page((unsigned long)file->private_data); | |
875 | return 0; | |
876 | } | |
12f38872 | 877 | EXPORT_SYMBOL(simple_transaction_release); |
1da177e4 | 878 | |
acaefc25 AB |
879 | /* Simple attribute files */ |
880 | ||
881 | struct simple_attr { | |
8b88b099 CH |
882 | int (*get)(void *, u64 *); |
883 | int (*set)(void *, u64); | |
acaefc25 AB |
884 | char get_buf[24]; /* enough to store a u64 and "\n\0" */ |
885 | char set_buf[24]; | |
886 | void *data; | |
887 | const char *fmt; /* format for read operation */ | |
7cf34c76 | 888 | struct mutex mutex; /* protects access to these buffers */ |
acaefc25 AB |
889 | }; |
890 | ||
891 | /* simple_attr_open is called by an actual attribute open file operation | |
892 | * to set the attribute specific access operations. */ | |
893 | int simple_attr_open(struct inode *inode, struct file *file, | |
8b88b099 | 894 | int (*get)(void *, u64 *), int (*set)(void *, u64), |
acaefc25 AB |
895 | const char *fmt) |
896 | { | |
897 | struct simple_attr *attr; | |
898 | ||
a65cab7d | 899 | attr = kzalloc(sizeof(*attr), GFP_KERNEL); |
acaefc25 AB |
900 | if (!attr) |
901 | return -ENOMEM; | |
902 | ||
903 | attr->get = get; | |
904 | attr->set = set; | |
8e18e294 | 905 | attr->data = inode->i_private; |
acaefc25 | 906 | attr->fmt = fmt; |
7cf34c76 | 907 | mutex_init(&attr->mutex); |
acaefc25 AB |
908 | |
909 | file->private_data = attr; | |
910 | ||
911 | return nonseekable_open(inode, file); | |
912 | } | |
12f38872 | 913 | EXPORT_SYMBOL_GPL(simple_attr_open); |
acaefc25 | 914 | |
74bedc4d | 915 | int simple_attr_release(struct inode *inode, struct file *file) |
acaefc25 AB |
916 | { |
917 | kfree(file->private_data); | |
918 | return 0; | |
919 | } | |
12f38872 | 920 | EXPORT_SYMBOL_GPL(simple_attr_release); /* GPL-only? This? Really? */ |
acaefc25 AB |
921 | |
922 | /* read from the buffer that is filled with the get function */ | |
923 | ssize_t simple_attr_read(struct file *file, char __user *buf, | |
924 | size_t len, loff_t *ppos) | |
925 | { | |
926 | struct simple_attr *attr; | |
927 | size_t size; | |
928 | ssize_t ret; | |
929 | ||
930 | attr = file->private_data; | |
931 | ||
932 | if (!attr->get) | |
933 | return -EACCES; | |
934 | ||
9261303a CH |
935 | ret = mutex_lock_interruptible(&attr->mutex); |
936 | if (ret) | |
937 | return ret; | |
938 | ||
a65cab7d EB |
939 | if (*ppos && attr->get_buf[0]) { |
940 | /* continued read */ | |
acaefc25 | 941 | size = strlen(attr->get_buf); |
a65cab7d EB |
942 | } else { |
943 | /* first read */ | |
8b88b099 CH |
944 | u64 val; |
945 | ret = attr->get(attr->data, &val); | |
946 | if (ret) | |
947 | goto out; | |
948 | ||
acaefc25 | 949 | size = scnprintf(attr->get_buf, sizeof(attr->get_buf), |
8b88b099 CH |
950 | attr->fmt, (unsigned long long)val); |
951 | } | |
acaefc25 AB |
952 | |
953 | ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size); | |
8b88b099 | 954 | out: |
7cf34c76 | 955 | mutex_unlock(&attr->mutex); |
acaefc25 AB |
956 | return ret; |
957 | } | |
12f38872 | 958 | EXPORT_SYMBOL_GPL(simple_attr_read); |
acaefc25 AB |
959 | |
960 | /* interpret the buffer as a number to call the set function with */ | |
961 | ssize_t simple_attr_write(struct file *file, const char __user *buf, | |
962 | size_t len, loff_t *ppos) | |
963 | { | |
964 | struct simple_attr *attr; | |
488dac0c | 965 | unsigned long long val; |
acaefc25 AB |
966 | size_t size; |
967 | ssize_t ret; | |
968 | ||
969 | attr = file->private_data; | |
acaefc25 AB |
970 | if (!attr->set) |
971 | return -EACCES; | |
972 | ||
9261303a CH |
973 | ret = mutex_lock_interruptible(&attr->mutex); |
974 | if (ret) | |
975 | return ret; | |
976 | ||
acaefc25 AB |
977 | ret = -EFAULT; |
978 | size = min(sizeof(attr->set_buf) - 1, len); | |
979 | if (copy_from_user(attr->set_buf, buf, size)) | |
980 | goto out; | |
981 | ||
acaefc25 | 982 | attr->set_buf[size] = '\0'; |
488dac0c YY |
983 | ret = kstrtoull(attr->set_buf, 0, &val); |
984 | if (ret) | |
985 | goto out; | |
05cc0cee WF |
986 | ret = attr->set(attr->data, val); |
987 | if (ret == 0) | |
988 | ret = len; /* on success, claim we got the whole input */ | |
acaefc25 | 989 | out: |
7cf34c76 | 990 | mutex_unlock(&attr->mutex); |
acaefc25 AB |
991 | return ret; |
992 | } | |
12f38872 | 993 | EXPORT_SYMBOL_GPL(simple_attr_write); |
acaefc25 | 994 | |
2596110a CH |
995 | /** |
996 | * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation | |
997 | * @sb: filesystem to do the file handle conversion on | |
998 | * @fid: file handle to convert | |
999 | * @fh_len: length of the file handle in bytes | |
1000 | * @fh_type: type of file handle | |
1001 | * @get_inode: filesystem callback to retrieve inode | |
1002 | * | |
1003 | * This function decodes @fid as long as it has one of the well-known | |
1004 | * Linux filehandle types and calls @get_inode on it to retrieve the | |
1005 | * inode for the object specified in the file handle. | |
1006 | */ | |
1007 | struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid, | |
1008 | int fh_len, int fh_type, struct inode *(*get_inode) | |
1009 | (struct super_block *sb, u64 ino, u32 gen)) | |
1010 | { | |
1011 | struct inode *inode = NULL; | |
1012 | ||
1013 | if (fh_len < 2) | |
1014 | return NULL; | |
1015 | ||
1016 | switch (fh_type) { | |
1017 | case FILEID_INO32_GEN: | |
1018 | case FILEID_INO32_GEN_PARENT: | |
1019 | inode = get_inode(sb, fid->i32.ino, fid->i32.gen); | |
1020 | break; | |
1021 | } | |
1022 | ||
4ea3ada2 | 1023 | return d_obtain_alias(inode); |
2596110a CH |
1024 | } |
1025 | EXPORT_SYMBOL_GPL(generic_fh_to_dentry); | |
1026 | ||
1027 | /** | |
ca186830 | 1028 | * generic_fh_to_parent - generic helper for the fh_to_parent export operation |
2596110a CH |
1029 | * @sb: filesystem to do the file handle conversion on |
1030 | * @fid: file handle to convert | |
1031 | * @fh_len: length of the file handle in bytes | |
1032 | * @fh_type: type of file handle | |
1033 | * @get_inode: filesystem callback to retrieve inode | |
1034 | * | |
1035 | * This function decodes @fid as long as it has one of the well-known | |
1036 | * Linux filehandle types and calls @get_inode on it to retrieve the | |
1037 | * inode for the _parent_ object specified in the file handle if it | |
1038 | * is specified in the file handle, or NULL otherwise. | |
1039 | */ | |
1040 | struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid, | |
1041 | int fh_len, int fh_type, struct inode *(*get_inode) | |
1042 | (struct super_block *sb, u64 ino, u32 gen)) | |
1043 | { | |
1044 | struct inode *inode = NULL; | |
1045 | ||
1046 | if (fh_len <= 2) | |
1047 | return NULL; | |
1048 | ||
1049 | switch (fh_type) { | |
1050 | case FILEID_INO32_GEN_PARENT: | |
1051 | inode = get_inode(sb, fid->i32.parent_ino, | |
1052 | (fh_len > 3 ? fid->i32.parent_gen : 0)); | |
1053 | break; | |
1054 | } | |
1055 | ||
4ea3ada2 | 1056 | return d_obtain_alias(inode); |
2596110a CH |
1057 | } |
1058 | EXPORT_SYMBOL_GPL(generic_fh_to_parent); | |
1059 | ||
1b061d92 | 1060 | /** |
ac13a829 FF |
1061 | * __generic_file_fsync - generic fsync implementation for simple filesystems |
1062 | * | |
1b061d92 | 1063 | * @file: file to synchronize |
ac13a829 FF |
1064 | * @start: start offset in bytes |
1065 | * @end: end offset in bytes (inclusive) | |
1b061d92 CH |
1066 | * @datasync: only synchronize essential metadata if true |
1067 | * | |
1068 | * This is a generic implementation of the fsync method for simple | |
1069 | * filesystems which track all non-inode metadata in the buffers list | |
1070 | * hanging off the address_space structure. | |
1071 | */ | |
ac13a829 FF |
1072 | int __generic_file_fsync(struct file *file, loff_t start, loff_t end, |
1073 | int datasync) | |
d5aacad5 | 1074 | { |
7ea80859 | 1075 | struct inode *inode = file->f_mapping->host; |
d5aacad5 AV |
1076 | int err; |
1077 | int ret; | |
1078 | ||
383aa543 | 1079 | err = file_write_and_wait_range(file, start, end); |
02c24a82 JB |
1080 | if (err) |
1081 | return err; | |
1082 | ||
5955102c | 1083 | inode_lock(inode); |
d5aacad5 | 1084 | ret = sync_mapping_buffers(inode->i_mapping); |
0ae45f63 | 1085 | if (!(inode->i_state & I_DIRTY_ALL)) |
02c24a82 | 1086 | goto out; |
d5aacad5 | 1087 | if (datasync && !(inode->i_state & I_DIRTY_DATASYNC)) |
02c24a82 | 1088 | goto out; |
d5aacad5 | 1089 | |
c3765016 | 1090 | err = sync_inode_metadata(inode, 1); |
d5aacad5 AV |
1091 | if (ret == 0) |
1092 | ret = err; | |
ac13a829 | 1093 | |
02c24a82 | 1094 | out: |
5955102c | 1095 | inode_unlock(inode); |
383aa543 JL |
1096 | /* check and advance again to catch errors after syncing out buffers */ |
1097 | err = file_check_and_advance_wb_err(file); | |
1098 | if (ret == 0) | |
1099 | ret = err; | |
1100 | return ret; | |
d5aacad5 | 1101 | } |
ac13a829 FF |
1102 | EXPORT_SYMBOL(__generic_file_fsync); |
1103 | ||
1104 | /** | |
1105 | * generic_file_fsync - generic fsync implementation for simple filesystems | |
1106 | * with flush | |
1107 | * @file: file to synchronize | |
1108 | * @start: start offset in bytes | |
1109 | * @end: end offset in bytes (inclusive) | |
1110 | * @datasync: only synchronize essential metadata if true | |
1111 | * | |
1112 | */ | |
1113 | ||
1114 | int generic_file_fsync(struct file *file, loff_t start, loff_t end, | |
1115 | int datasync) | |
1116 | { | |
1117 | struct inode *inode = file->f_mapping->host; | |
1118 | int err; | |
1119 | ||
1120 | err = __generic_file_fsync(file, start, end, datasync); | |
1121 | if (err) | |
1122 | return err; | |
c6bf3f0e | 1123 | return blkdev_issue_flush(inode->i_sb->s_bdev); |
ac13a829 | 1124 | } |
1b061d92 CH |
1125 | EXPORT_SYMBOL(generic_file_fsync); |
1126 | ||
30ca22c7 PL |
1127 | /** |
1128 | * generic_check_addressable - Check addressability of file system | |
1129 | * @blocksize_bits: log of file system block size | |
1130 | * @num_blocks: number of blocks in file system | |
1131 | * | |
1132 | * Determine whether a file system with @num_blocks blocks (and a | |
1133 | * block size of 2**@blocksize_bits) is addressable by the sector_t | |
1134 | * and page cache of the system. Return 0 if so and -EFBIG otherwise. | |
1135 | */ | |
1136 | int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks) | |
1137 | { | |
1138 | u64 last_fs_block = num_blocks - 1; | |
a33f13ef | 1139 | u64 last_fs_page = |
09cbfeaf | 1140 | last_fs_block >> (PAGE_SHIFT - blocksize_bits); |
30ca22c7 PL |
1141 | |
1142 | if (unlikely(num_blocks == 0)) | |
1143 | return 0; | |
1144 | ||
09cbfeaf | 1145 | if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT)) |
30ca22c7 PL |
1146 | return -EINVAL; |
1147 | ||
a33f13ef JB |
1148 | if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) || |
1149 | (last_fs_page > (pgoff_t)(~0ULL))) { | |
30ca22c7 PL |
1150 | return -EFBIG; |
1151 | } | |
1152 | return 0; | |
1153 | } | |
1154 | EXPORT_SYMBOL(generic_check_addressable); | |
1155 | ||
1b061d92 CH |
1156 | /* |
1157 | * No-op implementation of ->fsync for in-memory filesystems. | |
1158 | */ | |
02c24a82 | 1159 | int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync) |
1b061d92 CH |
1160 | { |
1161 | return 0; | |
1162 | } | |
1b061d92 | 1163 | EXPORT_SYMBOL(noop_fsync); |
87dc800b | 1164 | |
f44c7763 DW |
1165 | int noop_set_page_dirty(struct page *page) |
1166 | { | |
1167 | /* | |
1168 | * Unlike __set_page_dirty_no_writeback that handles dirty page | |
1169 | * tracking in the page object, dax does all dirty tracking in | |
1170 | * the inode address_space in response to mkwrite faults. In the | |
1171 | * dax case we only need to worry about potentially dirty CPU | |
1172 | * caches, not dirty page cache pages to write back. | |
1173 | * | |
1174 | * This callback is defined to prevent fallback to | |
1175 | * __set_page_dirty_buffers() in set_page_dirty(). | |
1176 | */ | |
1177 | return 0; | |
1178 | } | |
1179 | EXPORT_SYMBOL_GPL(noop_set_page_dirty); | |
1180 | ||
1181 | void noop_invalidatepage(struct page *page, unsigned int offset, | |
1182 | unsigned int length) | |
1183 | { | |
1184 | /* | |
1185 | * There is no page cache to invalidate in the dax case, however | |
1186 | * we need this callback defined to prevent falling back to | |
1187 | * block_invalidatepage() in do_invalidatepage(). | |
1188 | */ | |
1189 | } | |
1190 | EXPORT_SYMBOL_GPL(noop_invalidatepage); | |
1191 | ||
1192 | ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter) | |
1193 | { | |
1194 | /* | |
1195 | * iomap based filesystems support direct I/O without need for | |
1196 | * this callback. However, it still needs to be set in | |
1197 | * inode->a_ops so that open/fcntl know that direct I/O is | |
1198 | * generally supported. | |
1199 | */ | |
1200 | return -EINVAL; | |
1201 | } | |
1202 | EXPORT_SYMBOL_GPL(noop_direct_IO); | |
1203 | ||
fceef393 AV |
1204 | /* Because kfree isn't assignment-compatible with void(void*) ;-/ */ |
1205 | void kfree_link(void *p) | |
87dc800b | 1206 | { |
fceef393 | 1207 | kfree(p); |
87dc800b | 1208 | } |
fceef393 | 1209 | EXPORT_SYMBOL(kfree_link); |
6987843f AV |
1210 | |
1211 | /* | |
1212 | * nop .set_page_dirty method so that people can use .page_mkwrite on | |
1213 | * anon inodes. | |
1214 | */ | |
1215 | static int anon_set_page_dirty(struct page *page) | |
1216 | { | |
1217 | return 0; | |
1218 | }; | |
1219 | ||
6987843f AV |
1220 | struct inode *alloc_anon_inode(struct super_block *s) |
1221 | { | |
1222 | static const struct address_space_operations anon_aops = { | |
1223 | .set_page_dirty = anon_set_page_dirty, | |
1224 | }; | |
1225 | struct inode *inode = new_inode_pseudo(s); | |
1226 | ||
1227 | if (!inode) | |
1228 | return ERR_PTR(-ENOMEM); | |
1229 | ||
1230 | inode->i_ino = get_next_ino(); | |
1231 | inode->i_mapping->a_ops = &anon_aops; | |
1232 | ||
1233 | /* | |
1234 | * Mark the inode dirty from the very beginning, | |
1235 | * that way it will never be moved to the dirty | |
1236 | * list because mark_inode_dirty() will think | |
1237 | * that it already _is_ on the dirty list. | |
1238 | */ | |
1239 | inode->i_state = I_DIRTY; | |
1240 | inode->i_mode = S_IRUSR | S_IWUSR; | |
1241 | inode->i_uid = current_fsuid(); | |
1242 | inode->i_gid = current_fsgid(); | |
1243 | inode->i_flags |= S_PRIVATE; | |
078cd827 | 1244 | inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode); |
6987843f AV |
1245 | return inode; |
1246 | } | |
1247 | EXPORT_SYMBOL(alloc_anon_inode); | |
1c994a09 JL |
1248 | |
1249 | /** | |
1250 | * simple_nosetlease - generic helper for prohibiting leases | |
1251 | * @filp: file pointer | |
1252 | * @arg: type of lease to obtain | |
1253 | * @flp: new lease supplied for insertion | |
e6f5c789 | 1254 | * @priv: private data for lm_setup operation |
1c994a09 JL |
1255 | * |
1256 | * Generic helper for filesystems that do not wish to allow leases to be set. | |
1257 | * All arguments are ignored and it just returns -EINVAL. | |
1258 | */ | |
1259 | int | |
e6f5c789 JL |
1260 | simple_nosetlease(struct file *filp, long arg, struct file_lock **flp, |
1261 | void **priv) | |
1c994a09 JL |
1262 | { |
1263 | return -EINVAL; | |
1264 | } | |
1265 | EXPORT_SYMBOL(simple_nosetlease); | |
61ba64fc | 1266 | |
6ee9706a EB |
1267 | /** |
1268 | * simple_get_link - generic helper to get the target of "fast" symlinks | |
1269 | * @dentry: not used here | |
1270 | * @inode: the symlink inode | |
1271 | * @done: not used here | |
1272 | * | |
1273 | * Generic helper for filesystems to use for symlink inodes where a pointer to | |
1274 | * the symlink target is stored in ->i_link. NOTE: this isn't normally called, | |
1275 | * since as an optimization the path lookup code uses any non-NULL ->i_link | |
1276 | * directly, without calling ->get_link(). But ->get_link() still must be set, | |
1277 | * to mark the inode_operations as being for a symlink. | |
1278 | * | |
1279 | * Return: the symlink target | |
1280 | */ | |
6b255391 | 1281 | const char *simple_get_link(struct dentry *dentry, struct inode *inode, |
fceef393 | 1282 | struct delayed_call *done) |
61ba64fc | 1283 | { |
6b255391 | 1284 | return inode->i_link; |
61ba64fc | 1285 | } |
6b255391 | 1286 | EXPORT_SYMBOL(simple_get_link); |
61ba64fc AV |
1287 | |
1288 | const struct inode_operations simple_symlink_inode_operations = { | |
6b255391 | 1289 | .get_link = simple_get_link, |
61ba64fc AV |
1290 | }; |
1291 | EXPORT_SYMBOL(simple_symlink_inode_operations); | |
fbabfd0f EB |
1292 | |
1293 | /* | |
1294 | * Operations for a permanently empty directory. | |
1295 | */ | |
1296 | static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) | |
1297 | { | |
1298 | return ERR_PTR(-ENOENT); | |
1299 | } | |
1300 | ||
549c7297 CB |
1301 | static int empty_dir_getattr(struct user_namespace *mnt_userns, |
1302 | const struct path *path, struct kstat *stat, | |
a528d35e | 1303 | u32 request_mask, unsigned int query_flags) |
fbabfd0f | 1304 | { |
a528d35e | 1305 | struct inode *inode = d_inode(path->dentry); |
0d56a451 | 1306 | generic_fillattr(&init_user_ns, inode, stat); |
fbabfd0f EB |
1307 | return 0; |
1308 | } | |
1309 | ||
549c7297 CB |
1310 | static int empty_dir_setattr(struct user_namespace *mnt_userns, |
1311 | struct dentry *dentry, struct iattr *attr) | |
fbabfd0f EB |
1312 | { |
1313 | return -EPERM; | |
1314 | } | |
1315 | ||
fbabfd0f EB |
1316 | static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size) |
1317 | { | |
1318 | return -EOPNOTSUPP; | |
1319 | } | |
1320 | ||
1321 | static const struct inode_operations empty_dir_inode_operations = { | |
1322 | .lookup = empty_dir_lookup, | |
1323 | .permission = generic_permission, | |
1324 | .setattr = empty_dir_setattr, | |
1325 | .getattr = empty_dir_getattr, | |
fbabfd0f EB |
1326 | .listxattr = empty_dir_listxattr, |
1327 | }; | |
1328 | ||
1329 | static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence) | |
1330 | { | |
1331 | /* An empty directory has two entries . and .. at offsets 0 and 1 */ | |
1332 | return generic_file_llseek_size(file, offset, whence, 2, 2); | |
1333 | } | |
1334 | ||
1335 | static int empty_dir_readdir(struct file *file, struct dir_context *ctx) | |
1336 | { | |
1337 | dir_emit_dots(file, ctx); | |
1338 | return 0; | |
1339 | } | |
1340 | ||
1341 | static const struct file_operations empty_dir_operations = { | |
1342 | .llseek = empty_dir_llseek, | |
1343 | .read = generic_read_dir, | |
c51da20c | 1344 | .iterate_shared = empty_dir_readdir, |
fbabfd0f EB |
1345 | .fsync = noop_fsync, |
1346 | }; | |
1347 | ||
1348 | ||
1349 | void make_empty_dir_inode(struct inode *inode) | |
1350 | { | |
1351 | set_nlink(inode, 2); | |
1352 | inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO; | |
1353 | inode->i_uid = GLOBAL_ROOT_UID; | |
1354 | inode->i_gid = GLOBAL_ROOT_GID; | |
1355 | inode->i_rdev = 0; | |
4b75de86 | 1356 | inode->i_size = 0; |
fbabfd0f EB |
1357 | inode->i_blkbits = PAGE_SHIFT; |
1358 | inode->i_blocks = 0; | |
1359 | ||
1360 | inode->i_op = &empty_dir_inode_operations; | |
f5c24438 | 1361 | inode->i_opflags &= ~IOP_XATTR; |
fbabfd0f EB |
1362 | inode->i_fop = &empty_dir_operations; |
1363 | } | |
1364 | ||
1365 | bool is_empty_dir_inode(struct inode *inode) | |
1366 | { | |
1367 | return (inode->i_fop == &empty_dir_operations) && | |
1368 | (inode->i_op == &empty_dir_inode_operations); | |
1369 | } | |
c843843e DR |
1370 | |
1371 | #ifdef CONFIG_UNICODE | |
1372 | /* | |
1373 | * Determine if the name of a dentry should be casefolded. | |
1374 | * | |
1375 | * Return: if names will need casefolding | |
1376 | */ | |
1377 | static bool needs_casefold(const struct inode *dir) | |
1378 | { | |
1379 | return IS_CASEFOLDED(dir) && dir->i_sb->s_encoding; | |
1380 | } | |
1381 | ||
1382 | /** | |
1383 | * generic_ci_d_compare - generic d_compare implementation for casefolding filesystems | |
1384 | * @dentry: dentry whose name we are checking against | |
1385 | * @len: len of name of dentry | |
1386 | * @str: str pointer to name of dentry | |
1387 | * @name: Name to compare against | |
1388 | * | |
1389 | * Return: 0 if names match, 1 if mismatch, or -ERRNO | |
1390 | */ | |
794c43f7 EB |
1391 | static int generic_ci_d_compare(const struct dentry *dentry, unsigned int len, |
1392 | const char *str, const struct qstr *name) | |
c843843e DR |
1393 | { |
1394 | const struct dentry *parent = READ_ONCE(dentry->d_parent); | |
1395 | const struct inode *dir = READ_ONCE(parent->d_inode); | |
1396 | const struct super_block *sb = dentry->d_sb; | |
1397 | const struct unicode_map *um = sb->s_encoding; | |
1398 | struct qstr qstr = QSTR_INIT(str, len); | |
1399 | char strbuf[DNAME_INLINE_LEN]; | |
1400 | int ret; | |
1401 | ||
1402 | if (!dir || !needs_casefold(dir)) | |
1403 | goto fallback; | |
1404 | /* | |
1405 | * If the dentry name is stored in-line, then it may be concurrently | |
1406 | * modified by a rename. If this happens, the VFS will eventually retry | |
1407 | * the lookup, so it doesn't matter what ->d_compare() returns. | |
1408 | * However, it's unsafe to call utf8_strncasecmp() with an unstable | |
1409 | * string. Therefore, we have to copy the name into a temporary buffer. | |
1410 | */ | |
1411 | if (len <= DNAME_INLINE_LEN - 1) { | |
1412 | memcpy(strbuf, str, len); | |
1413 | strbuf[len] = 0; | |
1414 | qstr.name = strbuf; | |
1415 | /* prevent compiler from optimizing out the temporary buffer */ | |
1416 | barrier(); | |
1417 | } | |
1418 | ret = utf8_strncasecmp(um, name, &qstr); | |
1419 | if (ret >= 0) | |
1420 | return ret; | |
1421 | ||
1422 | if (sb_has_strict_encoding(sb)) | |
1423 | return -EINVAL; | |
1424 | fallback: | |
1425 | if (len != name->len) | |
1426 | return 1; | |
1427 | return !!memcmp(str, name->name, len); | |
1428 | } | |
c843843e DR |
1429 | |
1430 | /** | |
1431 | * generic_ci_d_hash - generic d_hash implementation for casefolding filesystems | |
1432 | * @dentry: dentry of the parent directory | |
1433 | * @str: qstr of name whose hash we should fill in | |
1434 | * | |
1435 | * Return: 0 if hash was successful or unchanged, and -EINVAL on error | |
1436 | */ | |
794c43f7 | 1437 | static int generic_ci_d_hash(const struct dentry *dentry, struct qstr *str) |
c843843e DR |
1438 | { |
1439 | const struct inode *dir = READ_ONCE(dentry->d_inode); | |
1440 | struct super_block *sb = dentry->d_sb; | |
1441 | const struct unicode_map *um = sb->s_encoding; | |
1442 | int ret = 0; | |
1443 | ||
1444 | if (!dir || !needs_casefold(dir)) | |
1445 | return 0; | |
1446 | ||
1447 | ret = utf8_casefold_hash(um, dentry, str); | |
1448 | if (ret < 0 && sb_has_strict_encoding(sb)) | |
1449 | return -EINVAL; | |
1450 | return 0; | |
1451 | } | |
608af703 DR |
1452 | |
1453 | static const struct dentry_operations generic_ci_dentry_ops = { | |
1454 | .d_hash = generic_ci_d_hash, | |
1455 | .d_compare = generic_ci_d_compare, | |
1456 | }; | |
1457 | #endif | |
1458 | ||
1459 | #ifdef CONFIG_FS_ENCRYPTION | |
1460 | static const struct dentry_operations generic_encrypted_dentry_ops = { | |
1461 | .d_revalidate = fscrypt_d_revalidate, | |
1462 | }; | |
1463 | #endif | |
1464 | ||
1465 | #if defined(CONFIG_FS_ENCRYPTION) && defined(CONFIG_UNICODE) | |
1466 | static const struct dentry_operations generic_encrypted_ci_dentry_ops = { | |
1467 | .d_hash = generic_ci_d_hash, | |
1468 | .d_compare = generic_ci_d_compare, | |
1469 | .d_revalidate = fscrypt_d_revalidate, | |
1470 | }; | |
1471 | #endif | |
1472 | ||
1473 | /** | |
1474 | * generic_set_encrypted_ci_d_ops - helper for setting d_ops for given dentry | |
1475 | * @dentry: dentry to set ops on | |
1476 | * | |
1477 | * Casefolded directories need d_hash and d_compare set, so that the dentries | |
1478 | * contained in them are handled case-insensitively. Note that these operations | |
1479 | * are needed on the parent directory rather than on the dentries in it, and | |
1480 | * while the casefolding flag can be toggled on and off on an empty directory, | |
1481 | * dentry_operations can't be changed later. As a result, if the filesystem has | |
1482 | * casefolding support enabled at all, we have to give all dentries the | |
1483 | * casefolding operations even if their inode doesn't have the casefolding flag | |
1484 | * currently (and thus the casefolding ops would be no-ops for now). | |
1485 | * | |
1486 | * Encryption works differently in that the only dentry operation it needs is | |
1487 | * d_revalidate, which it only needs on dentries that have the no-key name flag. | |
1488 | * The no-key flag can't be set "later", so we don't have to worry about that. | |
1489 | * | |
1490 | * Finally, to maximize compatibility with overlayfs (which isn't compatible | |
1491 | * with certain dentry operations) and to avoid taking an unnecessary | |
1492 | * performance hit, we use custom dentry_operations for each possible | |
1493 | * combination rather than always installing all operations. | |
1494 | */ | |
1495 | void generic_set_encrypted_ci_d_ops(struct dentry *dentry) | |
1496 | { | |
1497 | #ifdef CONFIG_FS_ENCRYPTION | |
1498 | bool needs_encrypt_ops = dentry->d_flags & DCACHE_NOKEY_NAME; | |
1499 | #endif | |
1500 | #ifdef CONFIG_UNICODE | |
1501 | bool needs_ci_ops = dentry->d_sb->s_encoding; | |
1502 | #endif | |
1503 | #if defined(CONFIG_FS_ENCRYPTION) && defined(CONFIG_UNICODE) | |
1504 | if (needs_encrypt_ops && needs_ci_ops) { | |
1505 | d_set_d_op(dentry, &generic_encrypted_ci_dentry_ops); | |
1506 | return; | |
1507 | } | |
c843843e | 1508 | #endif |
608af703 DR |
1509 | #ifdef CONFIG_FS_ENCRYPTION |
1510 | if (needs_encrypt_ops) { | |
1511 | d_set_d_op(dentry, &generic_encrypted_dentry_ops); | |
1512 | return; | |
1513 | } | |
1514 | #endif | |
1515 | #ifdef CONFIG_UNICODE | |
1516 | if (needs_ci_ops) { | |
1517 | d_set_d_op(dentry, &generic_ci_dentry_ops); | |
1518 | return; | |
1519 | } | |
1520 | #endif | |
1521 | } | |
1522 | EXPORT_SYMBOL(generic_set_encrypted_ci_d_ops); |