]>
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> | |
2d4e2157 | 22 | #include <linux/io.h> |
7cf34c76 | 23 | |
7c0f6ba6 | 24 | #include <linux/uaccess.h> |
1da177e4 | 25 | |
a4464dbc AV |
26 | #include "internal.h" |
27 | ||
a528d35e DH |
28 | int simple_getattr(const struct path *path, struct kstat *stat, |
29 | u32 request_mask, unsigned int query_flags) | |
1da177e4 | 30 | { |
a528d35e | 31 | struct inode *inode = d_inode(path->dentry); |
1da177e4 | 32 | generic_fillattr(inode, stat); |
09cbfeaf | 33 | stat->blocks = inode->i_mapping->nrpages << (PAGE_SHIFT - 9); |
1da177e4 LT |
34 | return 0; |
35 | } | |
12f38872 | 36 | EXPORT_SYMBOL(simple_getattr); |
1da177e4 | 37 | |
726c3342 | 38 | int simple_statfs(struct dentry *dentry, struct kstatfs *buf) |
1da177e4 | 39 | { |
726c3342 | 40 | buf->f_type = dentry->d_sb->s_magic; |
09cbfeaf | 41 | buf->f_bsize = PAGE_SIZE; |
1da177e4 LT |
42 | buf->f_namelen = NAME_MAX; |
43 | return 0; | |
44 | } | |
12f38872 | 45 | EXPORT_SYMBOL(simple_statfs); |
1da177e4 LT |
46 | |
47 | /* | |
48 | * Retaining negative dentries for an in-memory filesystem just wastes | |
49 | * memory and lookup time: arrange for them to be deleted immediately. | |
50 | */ | |
b26d4cd3 | 51 | int always_delete_dentry(const struct dentry *dentry) |
1da177e4 LT |
52 | { |
53 | return 1; | |
54 | } | |
b26d4cd3 AV |
55 | EXPORT_SYMBOL(always_delete_dentry); |
56 | ||
57 | const struct dentry_operations simple_dentry_operations = { | |
58 | .d_delete = always_delete_dentry, | |
59 | }; | |
60 | EXPORT_SYMBOL(simple_dentry_operations); | |
1da177e4 LT |
61 | |
62 | /* | |
63 | * Lookup the data. This is trivial - if the dentry didn't already | |
64 | * exist, we know it is negative. Set d_op to delete negative dentries. | |
65 | */ | |
00cd8dd3 | 66 | struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
1da177e4 | 67 | { |
1da177e4 LT |
68 | if (dentry->d_name.len > NAME_MAX) |
69 | return ERR_PTR(-ENAMETOOLONG); | |
74931da7 AV |
70 | if (!dentry->d_sb->s_d_op) |
71 | d_set_d_op(dentry, &simple_dentry_operations); | |
1da177e4 LT |
72 | d_add(dentry, NULL); |
73 | return NULL; | |
74 | } | |
12f38872 | 75 | EXPORT_SYMBOL(simple_lookup); |
1da177e4 | 76 | |
1da177e4 LT |
77 | int dcache_dir_open(struct inode *inode, struct file *file) |
78 | { | |
ba65dc5e | 79 | file->private_data = d_alloc_cursor(file->f_path.dentry); |
1da177e4 LT |
80 | |
81 | return file->private_data ? 0 : -ENOMEM; | |
82 | } | |
12f38872 | 83 | EXPORT_SYMBOL(dcache_dir_open); |
1da177e4 LT |
84 | |
85 | int dcache_dir_close(struct inode *inode, struct file *file) | |
86 | { | |
87 | dput(file->private_data); | |
88 | return 0; | |
89 | } | |
12f38872 | 90 | EXPORT_SYMBOL(dcache_dir_close); |
1da177e4 | 91 | |
4f42c1b5 | 92 | /* parent is locked at least shared */ |
d4f4de5e AV |
93 | /* |
94 | * Returns an element of siblings' list. | |
95 | * We are looking for <count>th positive after <p>; if | |
26b6c984 AV |
96 | * found, dentry is grabbed and returned to caller. |
97 | * If no such element exists, NULL is returned. | |
d4f4de5e | 98 | */ |
26b6c984 | 99 | static struct dentry *scan_positives(struct dentry *cursor, |
d4f4de5e AV |
100 | struct list_head *p, |
101 | loff_t count, | |
26b6c984 | 102 | struct dentry *last) |
4f42c1b5 | 103 | { |
d4f4de5e AV |
104 | struct dentry *dentry = cursor->d_parent, *found = NULL; |
105 | ||
106 | spin_lock(&dentry->d_lock); | |
107 | while ((p = p->next) != &dentry->d_subdirs) { | |
4f42c1b5 | 108 | struct dentry *d = list_entry(p, struct dentry, d_child); |
d4f4de5e AV |
109 | // we must at least skip cursors, to avoid livelocks |
110 | if (d->d_flags & DCACHE_DENTRY_CURSOR) | |
111 | continue; | |
112 | if (simple_positive(d) && !--count) { | |
113 | spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED); | |
114 | if (simple_positive(d)) | |
115 | found = dget_dlock(d); | |
116 | spin_unlock(&d->d_lock); | |
117 | if (likely(found)) | |
118 | break; | |
119 | count = 1; | |
120 | } | |
121 | if (need_resched()) { | |
122 | list_move(&cursor->d_child, p); | |
123 | p = &cursor->d_child; | |
124 | spin_unlock(&dentry->d_lock); | |
125 | cond_resched(); | |
126 | spin_lock(&dentry->d_lock); | |
4f42c1b5 AV |
127 | } |
128 | } | |
d4f4de5e | 129 | spin_unlock(&dentry->d_lock); |
26b6c984 AV |
130 | dput(last); |
131 | return found; | |
4f42c1b5 AV |
132 | } |
133 | ||
965c8e59 | 134 | loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence) |
1da177e4 | 135 | { |
2fd6b7f5 | 136 | struct dentry *dentry = file->f_path.dentry; |
965c8e59 | 137 | switch (whence) { |
1da177e4 LT |
138 | case 1: |
139 | offset += file->f_pos; | |
0a4c9265 | 140 | /* fall through */ |
1da177e4 LT |
141 | case 0: |
142 | if (offset >= 0) | |
143 | break; | |
0a4c9265 | 144 | /* fall through */ |
1da177e4 | 145 | default: |
1da177e4 LT |
146 | return -EINVAL; |
147 | } | |
148 | if (offset != file->f_pos) { | |
d4f4de5e AV |
149 | struct dentry *cursor = file->private_data; |
150 | struct dentry *to = NULL; | |
d4f4de5e | 151 | |
d4f4de5e AV |
152 | inode_lock_shared(dentry->d_inode); |
153 | ||
26b6c984 AV |
154 | if (offset > 2) |
155 | to = scan_positives(cursor, &dentry->d_subdirs, | |
156 | offset - 2, NULL); | |
157 | spin_lock(&dentry->d_lock); | |
158 | if (to) | |
159 | list_move(&cursor->d_child, &to->d_child); | |
160 | else | |
d4f4de5e | 161 | list_del_init(&cursor->d_child); |
26b6c984 | 162 | spin_unlock(&dentry->d_lock); |
d4f4de5e AV |
163 | dput(to); |
164 | ||
26b6c984 AV |
165 | file->f_pos = offset; |
166 | ||
d4f4de5e | 167 | inode_unlock_shared(dentry->d_inode); |
1da177e4 | 168 | } |
1da177e4 LT |
169 | return offset; |
170 | } | |
12f38872 | 171 | EXPORT_SYMBOL(dcache_dir_lseek); |
1da177e4 LT |
172 | |
173 | /* Relationship between i_mode and the DT_xxx types */ | |
174 | static inline unsigned char dt_type(struct inode *inode) | |
175 | { | |
176 | return (inode->i_mode >> 12) & 15; | |
177 | } | |
178 | ||
179 | /* | |
180 | * Directory is locked and all positive dentries in it are safe, since | |
181 | * for ramfs-type trees they can't go away without unlink() or rmdir(), | |
182 | * both impossible due to the lock on directory. | |
183 | */ | |
184 | ||
5f99f4e7 | 185 | int dcache_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 186 | { |
5f99f4e7 AV |
187 | struct dentry *dentry = file->f_path.dentry; |
188 | struct dentry *cursor = file->private_data; | |
d4f4de5e AV |
189 | struct list_head *anchor = &dentry->d_subdirs; |
190 | struct dentry *next = NULL; | |
191 | struct list_head *p; | |
1da177e4 | 192 | |
5f99f4e7 AV |
193 | if (!dir_emit_dots(file, ctx)) |
194 | return 0; | |
5f99f4e7 | 195 | |
4f42c1b5 | 196 | if (ctx->pos == 2) |
d4f4de5e | 197 | p = anchor; |
26b6c984 | 198 | else if (!list_empty(&cursor->d_child)) |
d4f4de5e | 199 | p = &cursor->d_child; |
26b6c984 AV |
200 | else |
201 | return 0; | |
d4f4de5e | 202 | |
26b6c984 | 203 | while ((next = scan_positives(cursor, p, 1, next)) != NULL) { |
5f99f4e7 | 204 | if (!dir_emit(ctx, next->d_name.name, next->d_name.len, |
dea655c2 | 205 | d_inode(next)->i_ino, dt_type(d_inode(next)))) |
4f42c1b5 | 206 | break; |
5f99f4e7 | 207 | ctx->pos++; |
26b6c984 | 208 | p = &next->d_child; |
1da177e4 | 209 | } |
d4f4de5e | 210 | spin_lock(&dentry->d_lock); |
26b6c984 AV |
211 | if (next) |
212 | list_move_tail(&cursor->d_child, &next->d_child); | |
213 | else | |
214 | list_del_init(&cursor->d_child); | |
d4f4de5e AV |
215 | spin_unlock(&dentry->d_lock); |
216 | dput(next); | |
217 | ||
1da177e4 LT |
218 | return 0; |
219 | } | |
12f38872 | 220 | EXPORT_SYMBOL(dcache_readdir); |
1da177e4 LT |
221 | |
222 | ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos) | |
223 | { | |
224 | return -EISDIR; | |
225 | } | |
12f38872 | 226 | EXPORT_SYMBOL(generic_read_dir); |
1da177e4 | 227 | |
4b6f5d20 | 228 | const struct file_operations simple_dir_operations = { |
1da177e4 LT |
229 | .open = dcache_dir_open, |
230 | .release = dcache_dir_close, | |
231 | .llseek = dcache_dir_lseek, | |
232 | .read = generic_read_dir, | |
4e82901c | 233 | .iterate_shared = dcache_readdir, |
1b061d92 | 234 | .fsync = noop_fsync, |
1da177e4 | 235 | }; |
12f38872 | 236 | EXPORT_SYMBOL(simple_dir_operations); |
1da177e4 | 237 | |
92e1d5be | 238 | const struct inode_operations simple_dir_inode_operations = { |
1da177e4 LT |
239 | .lookup = simple_lookup, |
240 | }; | |
12f38872 | 241 | EXPORT_SYMBOL(simple_dir_inode_operations); |
1da177e4 | 242 | |
759b9775 HD |
243 | static const struct super_operations simple_super_operations = { |
244 | .statfs = simple_statfs, | |
245 | }; | |
246 | ||
db2c246a | 247 | static int pseudo_fs_fill_super(struct super_block *s, struct fs_context *fc) |
1da177e4 | 248 | { |
31d6d5ce | 249 | struct pseudo_fs_context *ctx = fc->fs_private; |
1da177e4 | 250 | struct inode *root; |
1da177e4 | 251 | |
89a4eb4b | 252 | s->s_maxbytes = MAX_LFS_FILESIZE; |
3971e1a9 AN |
253 | s->s_blocksize = PAGE_SIZE; |
254 | s->s_blocksize_bits = PAGE_SHIFT; | |
8d9e46d8 AV |
255 | s->s_magic = ctx->magic; |
256 | s->s_op = ctx->ops ?: &simple_super_operations; | |
257 | s->s_xattr = ctx->xattr; | |
1da177e4 LT |
258 | s->s_time_gran = 1; |
259 | root = new_inode(s); | |
260 | if (!root) | |
db2c246a DH |
261 | return -ENOMEM; |
262 | ||
1a1c9bb4 JL |
263 | /* |
264 | * since this is the first inode, make it number 1. New inodes created | |
265 | * after this must take care not to collide with it (by passing | |
266 | * max_reserved of 1 to iunique). | |
267 | */ | |
268 | root->i_ino = 1; | |
1da177e4 | 269 | root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR; |
078cd827 | 270 | root->i_atime = root->i_mtime = root->i_ctime = current_time(root); |
8d9e46d8 AV |
271 | s->s_root = d_make_root(root); |
272 | if (!s->s_root) | |
db2c246a | 273 | return -ENOMEM; |
8d9e46d8 | 274 | s->s_d_op = ctx->dops; |
31d6d5ce | 275 | return 0; |
db2c246a | 276 | } |
8d9e46d8 | 277 | |
db2c246a DH |
278 | static int pseudo_fs_get_tree(struct fs_context *fc) |
279 | { | |
2ac295d4 | 280 | return get_tree_nodev(fc, pseudo_fs_fill_super); |
31d6d5ce DH |
281 | } |
282 | ||
283 | static void pseudo_fs_free(struct fs_context *fc) | |
284 | { | |
285 | kfree(fc->fs_private); | |
286 | } | |
287 | ||
288 | static const struct fs_context_operations pseudo_fs_context_ops = { | |
289 | .free = pseudo_fs_free, | |
290 | .get_tree = pseudo_fs_get_tree, | |
291 | }; | |
292 | ||
293 | /* | |
294 | * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that | |
295 | * will never be mountable) | |
296 | */ | |
297 | struct pseudo_fs_context *init_pseudo(struct fs_context *fc, | |
298 | unsigned long magic) | |
299 | { | |
300 | struct pseudo_fs_context *ctx; | |
301 | ||
302 | ctx = kzalloc(sizeof(struct pseudo_fs_context), GFP_KERNEL); | |
303 | if (likely(ctx)) { | |
304 | ctx->magic = magic; | |
305 | fc->fs_private = ctx; | |
306 | fc->ops = &pseudo_fs_context_ops; | |
db2c246a DH |
307 | fc->sb_flags |= SB_NOUSER; |
308 | fc->global = true; | |
1da177e4 | 309 | } |
31d6d5ce | 310 | return ctx; |
1da177e4 | 311 | } |
31d6d5ce | 312 | EXPORT_SYMBOL(init_pseudo); |
1da177e4 | 313 | |
20955e89 SB |
314 | int simple_open(struct inode *inode, struct file *file) |
315 | { | |
316 | if (inode->i_private) | |
317 | file->private_data = inode->i_private; | |
318 | return 0; | |
319 | } | |
12f38872 | 320 | EXPORT_SYMBOL(simple_open); |
20955e89 | 321 | |
1da177e4 LT |
322 | int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry) |
323 | { | |
dea655c2 | 324 | struct inode *inode = d_inode(old_dentry); |
1da177e4 | 325 | |
078cd827 | 326 | inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode); |
d8c76e6f | 327 | inc_nlink(inode); |
7de9c6ee | 328 | ihold(inode); |
1da177e4 LT |
329 | dget(dentry); |
330 | d_instantiate(dentry, inode); | |
331 | return 0; | |
332 | } | |
12f38872 | 333 | EXPORT_SYMBOL(simple_link); |
1da177e4 | 334 | |
1da177e4 LT |
335 | int simple_empty(struct dentry *dentry) |
336 | { | |
337 | struct dentry *child; | |
338 | int ret = 0; | |
339 | ||
2fd6b7f5 | 340 | spin_lock(&dentry->d_lock); |
946e51f2 | 341 | list_for_each_entry(child, &dentry->d_subdirs, d_child) { |
da502956 NP |
342 | spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED); |
343 | if (simple_positive(child)) { | |
344 | spin_unlock(&child->d_lock); | |
1da177e4 | 345 | goto out; |
da502956 NP |
346 | } |
347 | spin_unlock(&child->d_lock); | |
348 | } | |
1da177e4 LT |
349 | ret = 1; |
350 | out: | |
2fd6b7f5 | 351 | spin_unlock(&dentry->d_lock); |
1da177e4 LT |
352 | return ret; |
353 | } | |
12f38872 | 354 | EXPORT_SYMBOL(simple_empty); |
1da177e4 LT |
355 | |
356 | int simple_unlink(struct inode *dir, struct dentry *dentry) | |
357 | { | |
dea655c2 | 358 | struct inode *inode = d_inode(dentry); |
1da177e4 | 359 | |
078cd827 | 360 | inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode); |
9a53c3a7 | 361 | drop_nlink(inode); |
1da177e4 LT |
362 | dput(dentry); |
363 | return 0; | |
364 | } | |
12f38872 | 365 | EXPORT_SYMBOL(simple_unlink); |
1da177e4 LT |
366 | |
367 | int simple_rmdir(struct inode *dir, struct dentry *dentry) | |
368 | { | |
369 | if (!simple_empty(dentry)) | |
370 | return -ENOTEMPTY; | |
371 | ||
dea655c2 | 372 | drop_nlink(d_inode(dentry)); |
1da177e4 | 373 | simple_unlink(dir, dentry); |
9a53c3a7 | 374 | drop_nlink(dir); |
1da177e4 LT |
375 | return 0; |
376 | } | |
12f38872 | 377 | EXPORT_SYMBOL(simple_rmdir); |
1da177e4 LT |
378 | |
379 | int simple_rename(struct inode *old_dir, struct dentry *old_dentry, | |
e0e0be8a MS |
380 | struct inode *new_dir, struct dentry *new_dentry, |
381 | unsigned int flags) | |
1da177e4 | 382 | { |
dea655c2 | 383 | struct inode *inode = d_inode(old_dentry); |
e36cb0b8 | 384 | int they_are_dirs = d_is_dir(old_dentry); |
1da177e4 | 385 | |
e0e0be8a MS |
386 | if (flags & ~RENAME_NOREPLACE) |
387 | return -EINVAL; | |
388 | ||
1da177e4 LT |
389 | if (!simple_empty(new_dentry)) |
390 | return -ENOTEMPTY; | |
391 | ||
dea655c2 | 392 | if (d_really_is_positive(new_dentry)) { |
1da177e4 | 393 | simple_unlink(new_dir, new_dentry); |
841590ce | 394 | if (they_are_dirs) { |
dea655c2 | 395 | drop_nlink(d_inode(new_dentry)); |
9a53c3a7 | 396 | drop_nlink(old_dir); |
841590ce | 397 | } |
1da177e4 | 398 | } else if (they_are_dirs) { |
9a53c3a7 | 399 | drop_nlink(old_dir); |
d8c76e6f | 400 | inc_nlink(new_dir); |
1da177e4 LT |
401 | } |
402 | ||
403 | old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime = | |
078cd827 | 404 | new_dir->i_mtime = inode->i_ctime = current_time(old_dir); |
1da177e4 LT |
405 | |
406 | return 0; | |
407 | } | |
12f38872 | 408 | EXPORT_SYMBOL(simple_rename); |
1da177e4 | 409 | |
7bb46a67 | 410 | /** |
eef2380c | 411 | * simple_setattr - setattr for simple filesystem |
7bb46a67 | 412 | * @dentry: dentry |
413 | * @iattr: iattr structure | |
414 | * | |
415 | * Returns 0 on success, -error on failure. | |
416 | * | |
eef2380c CH |
417 | * simple_setattr is a simple ->setattr implementation without a proper |
418 | * implementation of size changes. | |
419 | * | |
420 | * It can either be used for in-memory filesystems or special files | |
421 | * on simple regular filesystems. Anything that needs to change on-disk | |
422 | * or wire state on size changes needs its own setattr method. | |
7bb46a67 | 423 | */ |
424 | int simple_setattr(struct dentry *dentry, struct iattr *iattr) | |
425 | { | |
dea655c2 | 426 | struct inode *inode = d_inode(dentry); |
7bb46a67 | 427 | int error; |
428 | ||
31051c85 | 429 | error = setattr_prepare(dentry, iattr); |
7bb46a67 | 430 | if (error) |
431 | return error; | |
432 | ||
2c27c65e CH |
433 | if (iattr->ia_valid & ATTR_SIZE) |
434 | truncate_setsize(inode, iattr->ia_size); | |
6a1a90ad | 435 | setattr_copy(inode, iattr); |
eef2380c CH |
436 | mark_inode_dirty(inode); |
437 | return 0; | |
7bb46a67 | 438 | } |
439 | EXPORT_SYMBOL(simple_setattr); | |
440 | ||
1da177e4 LT |
441 | int simple_readpage(struct file *file, struct page *page) |
442 | { | |
c0d92cbc | 443 | clear_highpage(page); |
1da177e4 LT |
444 | flush_dcache_page(page); |
445 | SetPageUptodate(page); | |
1da177e4 LT |
446 | unlock_page(page); |
447 | return 0; | |
448 | } | |
12f38872 | 449 | EXPORT_SYMBOL(simple_readpage); |
1da177e4 | 450 | |
afddba49 NP |
451 | int simple_write_begin(struct file *file, struct address_space *mapping, |
452 | loff_t pos, unsigned len, unsigned flags, | |
453 | struct page **pagep, void **fsdata) | |
454 | { | |
455 | struct page *page; | |
456 | pgoff_t index; | |
afddba49 | 457 | |
09cbfeaf | 458 | index = pos >> PAGE_SHIFT; |
afddba49 | 459 | |
54566b2c | 460 | page = grab_cache_page_write_begin(mapping, index, flags); |
afddba49 NP |
461 | if (!page) |
462 | return -ENOMEM; | |
463 | ||
464 | *pagep = page; | |
465 | ||
09cbfeaf KS |
466 | if (!PageUptodate(page) && (len != PAGE_SIZE)) { |
467 | unsigned from = pos & (PAGE_SIZE - 1); | |
193cf4b9 | 468 | |
09cbfeaf | 469 | zero_user_segments(page, 0, from, from + len, PAGE_SIZE); |
193cf4b9 BH |
470 | } |
471 | return 0; | |
afddba49 | 472 | } |
12f38872 | 473 | EXPORT_SYMBOL(simple_write_begin); |
afddba49 | 474 | |
ad2a722f BH |
475 | /** |
476 | * simple_write_end - .write_end helper for non-block-device FSes | |
8e88bfba | 477 | * @file: See .write_end of address_space_operations |
ad2a722f BH |
478 | * @mapping: " |
479 | * @pos: " | |
480 | * @len: " | |
481 | * @copied: " | |
482 | * @page: " | |
483 | * @fsdata: " | |
484 | * | |
485 | * simple_write_end does the minimum needed for updating a page after writing is | |
486 | * done. It has the same API signature as the .write_end of | |
487 | * address_space_operations vector. So it can just be set onto .write_end for | |
488 | * FSes that don't need any other processing. i_mutex is assumed to be held. | |
489 | * Block based filesystems should use generic_write_end(). | |
490 | * NOTE: Even though i_size might get updated by this function, mark_inode_dirty | |
491 | * is not called, so a filesystem that actually does store data in .write_inode | |
492 | * should extend on what's done here with a call to mark_inode_dirty() in the | |
493 | * case that i_size has changed. | |
04fff641 AV |
494 | * |
495 | * Use *ONLY* with simple_readpage() | |
ad2a722f | 496 | */ |
afddba49 NP |
497 | int simple_write_end(struct file *file, struct address_space *mapping, |
498 | loff_t pos, unsigned len, unsigned copied, | |
499 | struct page *page, void *fsdata) | |
500 | { | |
ad2a722f BH |
501 | struct inode *inode = page->mapping->host; |
502 | loff_t last_pos = pos + copied; | |
afddba49 NP |
503 | |
504 | /* zero the stale part of the page if we did a short copy */ | |
04fff641 AV |
505 | if (!PageUptodate(page)) { |
506 | if (copied < len) { | |
507 | unsigned from = pos & (PAGE_SIZE - 1); | |
afddba49 | 508 | |
04fff641 AV |
509 | zero_user(page, from + copied, len - copied); |
510 | } | |
ad2a722f | 511 | SetPageUptodate(page); |
04fff641 | 512 | } |
ad2a722f BH |
513 | /* |
514 | * No need to use i_size_read() here, the i_size | |
515 | * cannot change under us because we hold the i_mutex. | |
516 | */ | |
517 | if (last_pos > inode->i_size) | |
518 | i_size_write(inode, last_pos); | |
afddba49 | 519 | |
ad2a722f | 520 | set_page_dirty(page); |
afddba49 | 521 | unlock_page(page); |
09cbfeaf | 522 | put_page(page); |
afddba49 NP |
523 | |
524 | return copied; | |
525 | } | |
12f38872 | 526 | EXPORT_SYMBOL(simple_write_end); |
afddba49 | 527 | |
1a1c9bb4 JL |
528 | /* |
529 | * the inodes created here are not hashed. If you use iunique to generate | |
530 | * unique inode values later for this filesystem, then you must take care | |
531 | * to pass it an appropriate max_reserved value to avoid collisions. | |
532 | */ | |
7d683a09 | 533 | int simple_fill_super(struct super_block *s, unsigned long magic, |
cda37124 | 534 | const struct tree_descr *files) |
1da177e4 | 535 | { |
1da177e4 LT |
536 | struct inode *inode; |
537 | struct dentry *root; | |
538 | struct dentry *dentry; | |
539 | int i; | |
540 | ||
09cbfeaf KS |
541 | s->s_blocksize = PAGE_SIZE; |
542 | s->s_blocksize_bits = PAGE_SHIFT; | |
1da177e4 | 543 | s->s_magic = magic; |
759b9775 | 544 | s->s_op = &simple_super_operations; |
1da177e4 LT |
545 | s->s_time_gran = 1; |
546 | ||
547 | inode = new_inode(s); | |
548 | if (!inode) | |
549 | return -ENOMEM; | |
1a1c9bb4 JL |
550 | /* |
551 | * because the root inode is 1, the files array must not contain an | |
552 | * entry at index 1 | |
553 | */ | |
554 | inode->i_ino = 1; | |
1da177e4 | 555 | inode->i_mode = S_IFDIR | 0755; |
078cd827 | 556 | inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode); |
1da177e4 LT |
557 | inode->i_op = &simple_dir_inode_operations; |
558 | inode->i_fop = &simple_dir_operations; | |
bfe86848 | 559 | set_nlink(inode, 2); |
48fde701 AV |
560 | root = d_make_root(inode); |
561 | if (!root) | |
1da177e4 | 562 | return -ENOMEM; |
1da177e4 LT |
563 | for (i = 0; !files->name || files->name[0]; i++, files++) { |
564 | if (!files->name) | |
565 | continue; | |
1a1c9bb4 JL |
566 | |
567 | /* warn if it tries to conflict with the root inode */ | |
568 | if (unlikely(i == 1)) | |
569 | printk(KERN_WARNING "%s: %s passed in a files array" | |
570 | "with an index of 1!\n", __func__, | |
571 | s->s_type->name); | |
572 | ||
1da177e4 LT |
573 | dentry = d_alloc_name(root, files->name); |
574 | if (!dentry) | |
575 | goto out; | |
576 | inode = new_inode(s); | |
32096ea1 KK |
577 | if (!inode) { |
578 | dput(dentry); | |
1da177e4 | 579 | goto out; |
32096ea1 | 580 | } |
1da177e4 | 581 | inode->i_mode = S_IFREG | files->mode; |
078cd827 | 582 | inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode); |
1da177e4 LT |
583 | inode->i_fop = files->ops; |
584 | inode->i_ino = i; | |
585 | d_add(dentry, inode); | |
586 | } | |
587 | s->s_root = root; | |
588 | return 0; | |
589 | out: | |
590 | d_genocide(root); | |
640946f2 | 591 | shrink_dcache_parent(root); |
1da177e4 LT |
592 | dput(root); |
593 | return -ENOMEM; | |
594 | } | |
12f38872 | 595 | EXPORT_SYMBOL(simple_fill_super); |
1da177e4 LT |
596 | |
597 | static DEFINE_SPINLOCK(pin_fs_lock); | |
598 | ||
1f5ce9e9 | 599 | int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count) |
1da177e4 LT |
600 | { |
601 | struct vfsmount *mnt = NULL; | |
602 | spin_lock(&pin_fs_lock); | |
603 | if (unlikely(!*mount)) { | |
604 | spin_unlock(&pin_fs_lock); | |
1751e8a6 | 605 | mnt = vfs_kern_mount(type, SB_KERNMOUNT, type->name, NULL); |
1da177e4 LT |
606 | if (IS_ERR(mnt)) |
607 | return PTR_ERR(mnt); | |
608 | spin_lock(&pin_fs_lock); | |
609 | if (!*mount) | |
610 | *mount = mnt; | |
611 | } | |
612 | mntget(*mount); | |
613 | ++*count; | |
614 | spin_unlock(&pin_fs_lock); | |
615 | mntput(mnt); | |
616 | return 0; | |
617 | } | |
12f38872 | 618 | EXPORT_SYMBOL(simple_pin_fs); |
1da177e4 LT |
619 | |
620 | void simple_release_fs(struct vfsmount **mount, int *count) | |
621 | { | |
622 | struct vfsmount *mnt; | |
623 | spin_lock(&pin_fs_lock); | |
624 | mnt = *mount; | |
625 | if (!--*count) | |
626 | *mount = NULL; | |
627 | spin_unlock(&pin_fs_lock); | |
628 | mntput(mnt); | |
629 | } | |
12f38872 | 630 | EXPORT_SYMBOL(simple_release_fs); |
1da177e4 | 631 | |
6d1029b5 AM |
632 | /** |
633 | * simple_read_from_buffer - copy data from the buffer to user space | |
634 | * @to: the user space buffer to read to | |
635 | * @count: the maximum number of bytes to read | |
636 | * @ppos: the current position in the buffer | |
637 | * @from: the buffer to read from | |
638 | * @available: the size of the buffer | |
639 | * | |
640 | * The simple_read_from_buffer() function reads up to @count bytes from the | |
641 | * buffer @from at offset @ppos into the user space address starting at @to. | |
642 | * | |
643 | * On success, the number of bytes read is returned and the offset @ppos is | |
644 | * advanced by this number, or negative value is returned on error. | |
645 | **/ | |
1da177e4 LT |
646 | ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos, |
647 | const void *from, size_t available) | |
648 | { | |
649 | loff_t pos = *ppos; | |
14be2746 SR |
650 | size_t ret; |
651 | ||
1da177e4 LT |
652 | if (pos < 0) |
653 | return -EINVAL; | |
14be2746 | 654 | if (pos >= available || !count) |
1da177e4 LT |
655 | return 0; |
656 | if (count > available - pos) | |
657 | count = available - pos; | |
14be2746 SR |
658 | ret = copy_to_user(to, from + pos, count); |
659 | if (ret == count) | |
1da177e4 | 660 | return -EFAULT; |
14be2746 | 661 | count -= ret; |
1da177e4 LT |
662 | *ppos = pos + count; |
663 | return count; | |
664 | } | |
12f38872 | 665 | EXPORT_SYMBOL(simple_read_from_buffer); |
1da177e4 | 666 | |
6a727b43 JS |
667 | /** |
668 | * simple_write_to_buffer - copy data from user space to the buffer | |
669 | * @to: the buffer to write to | |
670 | * @available: the size of the buffer | |
671 | * @ppos: the current position in the buffer | |
672 | * @from: the user space buffer to read from | |
673 | * @count: the maximum number of bytes to read | |
674 | * | |
675 | * The simple_write_to_buffer() function reads up to @count bytes from the user | |
676 | * space address starting at @from into the buffer @to at offset @ppos. | |
677 | * | |
678 | * On success, the number of bytes written is returned and the offset @ppos is | |
679 | * advanced by this number, or negative value is returned on error. | |
680 | **/ | |
681 | ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos, | |
682 | const void __user *from, size_t count) | |
683 | { | |
684 | loff_t pos = *ppos; | |
685 | size_t res; | |
686 | ||
687 | if (pos < 0) | |
688 | return -EINVAL; | |
689 | if (pos >= available || !count) | |
690 | return 0; | |
691 | if (count > available - pos) | |
692 | count = available - pos; | |
693 | res = copy_from_user(to + pos, from, count); | |
694 | if (res == count) | |
695 | return -EFAULT; | |
696 | count -= res; | |
697 | *ppos = pos + count; | |
698 | return count; | |
699 | } | |
12f38872 | 700 | EXPORT_SYMBOL(simple_write_to_buffer); |
6a727b43 | 701 | |
6d1029b5 AM |
702 | /** |
703 | * memory_read_from_buffer - copy data from the buffer | |
704 | * @to: the kernel space buffer to read to | |
705 | * @count: the maximum number of bytes to read | |
706 | * @ppos: the current position in the buffer | |
707 | * @from: the buffer to read from | |
708 | * @available: the size of the buffer | |
709 | * | |
710 | * The memory_read_from_buffer() function reads up to @count bytes from the | |
711 | * buffer @from at offset @ppos into the kernel space address starting at @to. | |
712 | * | |
713 | * On success, the number of bytes read is returned and the offset @ppos is | |
714 | * advanced by this number, or negative value is returned on error. | |
715 | **/ | |
93b07113 AM |
716 | ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos, |
717 | const void *from, size_t available) | |
718 | { | |
719 | loff_t pos = *ppos; | |
720 | ||
721 | if (pos < 0) | |
722 | return -EINVAL; | |
723 | if (pos >= available) | |
724 | return 0; | |
725 | if (count > available - pos) | |
726 | count = available - pos; | |
727 | memcpy(to, from + pos, count); | |
728 | *ppos = pos + count; | |
729 | ||
730 | return count; | |
731 | } | |
12f38872 | 732 | EXPORT_SYMBOL(memory_read_from_buffer); |
93b07113 | 733 | |
2d4e2157 CIK |
734 | /** |
735 | * memory_read_from_io_buffer - copy data from a io memory mapped buffer | |
736 | * @to: the kernel space buffer to read to | |
737 | * @count: the maximum number of bytes to read | |
738 | * @ppos: the current position in the buffer | |
739 | * @from: the buffer to read from | |
740 | * @available: the size of the buffer | |
741 | * | |
742 | * The memory_read_from_buffer() function reads up to @count bytes from the | |
743 | * io memory mappy buffer @from at offset @ppos into the kernel space address | |
744 | * starting at @to. | |
745 | * | |
746 | * On success, the number of bytes read is returned and the offset @ppos is | |
747 | * advanced by this number, or negative value is returned on error. | |
748 | **/ | |
749 | ssize_t memory_read_from_io_buffer(void *to, size_t count, loff_t *ppos, | |
750 | const void *from, size_t available) | |
751 | { | |
752 | loff_t pos = *ppos; | |
753 | ||
754 | if (pos < 0) | |
755 | return -EINVAL; | |
756 | if (pos >= available) | |
757 | return 0; | |
758 | if (count > available - pos) | |
759 | count = available - pos; | |
760 | memcpy_fromio(to, from + pos, count); | |
761 | *ppos = pos + count; | |
762 | ||
763 | return count; | |
764 | } | |
765 | EXPORT_SYMBOL(memory_read_from_io_buffer); | |
766 | ||
1da177e4 LT |
767 | /* |
768 | * Transaction based IO. | |
769 | * The file expects a single write which triggers the transaction, and then | |
770 | * possibly a read which collects the result - which is stored in a | |
771 | * file-local buffer. | |
772 | */ | |
76791ab2 IM |
773 | |
774 | void simple_transaction_set(struct file *file, size_t n) | |
775 | { | |
776 | struct simple_transaction_argresp *ar = file->private_data; | |
777 | ||
778 | BUG_ON(n > SIMPLE_TRANSACTION_LIMIT); | |
779 | ||
780 | /* | |
781 | * The barrier ensures that ar->size will really remain zero until | |
782 | * ar->data is ready for reading. | |
783 | */ | |
784 | smp_mb(); | |
785 | ar->size = n; | |
786 | } | |
12f38872 | 787 | EXPORT_SYMBOL(simple_transaction_set); |
76791ab2 | 788 | |
1da177e4 LT |
789 | char *simple_transaction_get(struct file *file, const char __user *buf, size_t size) |
790 | { | |
791 | struct simple_transaction_argresp *ar; | |
792 | static DEFINE_SPINLOCK(simple_transaction_lock); | |
793 | ||
794 | if (size > SIMPLE_TRANSACTION_LIMIT - 1) | |
795 | return ERR_PTR(-EFBIG); | |
796 | ||
797 | ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL); | |
798 | if (!ar) | |
799 | return ERR_PTR(-ENOMEM); | |
800 | ||
801 | spin_lock(&simple_transaction_lock); | |
802 | ||
803 | /* only one write allowed per open */ | |
804 | if (file->private_data) { | |
805 | spin_unlock(&simple_transaction_lock); | |
806 | free_page((unsigned long)ar); | |
807 | return ERR_PTR(-EBUSY); | |
808 | } | |
809 | ||
810 | file->private_data = ar; | |
811 | ||
812 | spin_unlock(&simple_transaction_lock); | |
813 | ||
814 | if (copy_from_user(ar->data, buf, size)) | |
815 | return ERR_PTR(-EFAULT); | |
816 | ||
817 | return ar->data; | |
818 | } | |
12f38872 | 819 | EXPORT_SYMBOL(simple_transaction_get); |
1da177e4 LT |
820 | |
821 | ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos) | |
822 | { | |
823 | struct simple_transaction_argresp *ar = file->private_data; | |
824 | ||
825 | if (!ar) | |
826 | return 0; | |
827 | return simple_read_from_buffer(buf, size, pos, ar->data, ar->size); | |
828 | } | |
12f38872 | 829 | EXPORT_SYMBOL(simple_transaction_read); |
1da177e4 LT |
830 | |
831 | int simple_transaction_release(struct inode *inode, struct file *file) | |
832 | { | |
833 | free_page((unsigned long)file->private_data); | |
834 | return 0; | |
835 | } | |
12f38872 | 836 | EXPORT_SYMBOL(simple_transaction_release); |
1da177e4 | 837 | |
acaefc25 AB |
838 | /* Simple attribute files */ |
839 | ||
840 | struct simple_attr { | |
8b88b099 CH |
841 | int (*get)(void *, u64 *); |
842 | int (*set)(void *, u64); | |
acaefc25 AB |
843 | char get_buf[24]; /* enough to store a u64 and "\n\0" */ |
844 | char set_buf[24]; | |
845 | void *data; | |
846 | const char *fmt; /* format for read operation */ | |
7cf34c76 | 847 | struct mutex mutex; /* protects access to these buffers */ |
acaefc25 AB |
848 | }; |
849 | ||
850 | /* simple_attr_open is called by an actual attribute open file operation | |
851 | * to set the attribute specific access operations. */ | |
852 | int simple_attr_open(struct inode *inode, struct file *file, | |
8b88b099 | 853 | int (*get)(void *, u64 *), int (*set)(void *, u64), |
acaefc25 AB |
854 | const char *fmt) |
855 | { | |
856 | struct simple_attr *attr; | |
857 | ||
ef9fd780 | 858 | attr = kzalloc(sizeof(*attr), GFP_KERNEL); |
acaefc25 AB |
859 | if (!attr) |
860 | return -ENOMEM; | |
861 | ||
862 | attr->get = get; | |
863 | attr->set = set; | |
8e18e294 | 864 | attr->data = inode->i_private; |
acaefc25 | 865 | attr->fmt = fmt; |
7cf34c76 | 866 | mutex_init(&attr->mutex); |
acaefc25 AB |
867 | |
868 | file->private_data = attr; | |
869 | ||
870 | return nonseekable_open(inode, file); | |
871 | } | |
12f38872 | 872 | EXPORT_SYMBOL_GPL(simple_attr_open); |
acaefc25 | 873 | |
74bedc4d | 874 | int simple_attr_release(struct inode *inode, struct file *file) |
acaefc25 AB |
875 | { |
876 | kfree(file->private_data); | |
877 | return 0; | |
878 | } | |
12f38872 | 879 | EXPORT_SYMBOL_GPL(simple_attr_release); /* GPL-only? This? Really? */ |
acaefc25 AB |
880 | |
881 | /* read from the buffer that is filled with the get function */ | |
882 | ssize_t simple_attr_read(struct file *file, char __user *buf, | |
883 | size_t len, loff_t *ppos) | |
884 | { | |
885 | struct simple_attr *attr; | |
886 | size_t size; | |
887 | ssize_t ret; | |
888 | ||
889 | attr = file->private_data; | |
890 | ||
891 | if (!attr->get) | |
892 | return -EACCES; | |
893 | ||
9261303a CH |
894 | ret = mutex_lock_interruptible(&attr->mutex); |
895 | if (ret) | |
896 | return ret; | |
897 | ||
ef9fd780 EB |
898 | if (*ppos && attr->get_buf[0]) { |
899 | /* continued read */ | |
acaefc25 | 900 | size = strlen(attr->get_buf); |
ef9fd780 EB |
901 | } else { |
902 | /* first read */ | |
8b88b099 CH |
903 | u64 val; |
904 | ret = attr->get(attr->data, &val); | |
905 | if (ret) | |
906 | goto out; | |
907 | ||
acaefc25 | 908 | size = scnprintf(attr->get_buf, sizeof(attr->get_buf), |
8b88b099 CH |
909 | attr->fmt, (unsigned long long)val); |
910 | } | |
acaefc25 AB |
911 | |
912 | ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size); | |
8b88b099 | 913 | out: |
7cf34c76 | 914 | mutex_unlock(&attr->mutex); |
acaefc25 AB |
915 | return ret; |
916 | } | |
12f38872 | 917 | EXPORT_SYMBOL_GPL(simple_attr_read); |
acaefc25 AB |
918 | |
919 | /* interpret the buffer as a number to call the set function with */ | |
920 | ssize_t simple_attr_write(struct file *file, const char __user *buf, | |
921 | size_t len, loff_t *ppos) | |
922 | { | |
923 | struct simple_attr *attr; | |
f7a3c54a | 924 | unsigned long long val; |
acaefc25 AB |
925 | size_t size; |
926 | ssize_t ret; | |
927 | ||
928 | attr = file->private_data; | |
acaefc25 AB |
929 | if (!attr->set) |
930 | return -EACCES; | |
931 | ||
9261303a CH |
932 | ret = mutex_lock_interruptible(&attr->mutex); |
933 | if (ret) | |
934 | return ret; | |
935 | ||
acaefc25 AB |
936 | ret = -EFAULT; |
937 | size = min(sizeof(attr->set_buf) - 1, len); | |
938 | if (copy_from_user(attr->set_buf, buf, size)) | |
939 | goto out; | |
940 | ||
acaefc25 | 941 | attr->set_buf[size] = '\0'; |
f7a3c54a YY |
942 | ret = kstrtoull(attr->set_buf, 0, &val); |
943 | if (ret) | |
944 | goto out; | |
05cc0cee WF |
945 | ret = attr->set(attr->data, val); |
946 | if (ret == 0) | |
947 | ret = len; /* on success, claim we got the whole input */ | |
acaefc25 | 948 | out: |
7cf34c76 | 949 | mutex_unlock(&attr->mutex); |
acaefc25 AB |
950 | return ret; |
951 | } | |
12f38872 | 952 | EXPORT_SYMBOL_GPL(simple_attr_write); |
acaefc25 | 953 | |
2596110a CH |
954 | /** |
955 | * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation | |
956 | * @sb: filesystem to do the file handle conversion on | |
957 | * @fid: file handle to convert | |
958 | * @fh_len: length of the file handle in bytes | |
959 | * @fh_type: type of file handle | |
960 | * @get_inode: filesystem callback to retrieve inode | |
961 | * | |
962 | * This function decodes @fid as long as it has one of the well-known | |
963 | * Linux filehandle types and calls @get_inode on it to retrieve the | |
964 | * inode for the object specified in the file handle. | |
965 | */ | |
966 | struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid, | |
967 | int fh_len, int fh_type, struct inode *(*get_inode) | |
968 | (struct super_block *sb, u64 ino, u32 gen)) | |
969 | { | |
970 | struct inode *inode = NULL; | |
971 | ||
972 | if (fh_len < 2) | |
973 | return NULL; | |
974 | ||
975 | switch (fh_type) { | |
976 | case FILEID_INO32_GEN: | |
977 | case FILEID_INO32_GEN_PARENT: | |
978 | inode = get_inode(sb, fid->i32.ino, fid->i32.gen); | |
979 | break; | |
980 | } | |
981 | ||
4ea3ada2 | 982 | return d_obtain_alias(inode); |
2596110a CH |
983 | } |
984 | EXPORT_SYMBOL_GPL(generic_fh_to_dentry); | |
985 | ||
986 | /** | |
ca186830 | 987 | * generic_fh_to_parent - generic helper for the fh_to_parent export operation |
2596110a CH |
988 | * @sb: filesystem to do the file handle conversion on |
989 | * @fid: file handle to convert | |
990 | * @fh_len: length of the file handle in bytes | |
991 | * @fh_type: type of file handle | |
992 | * @get_inode: filesystem callback to retrieve inode | |
993 | * | |
994 | * This function decodes @fid as long as it has one of the well-known | |
995 | * Linux filehandle types and calls @get_inode on it to retrieve the | |
996 | * inode for the _parent_ object specified in the file handle if it | |
997 | * is specified in the file handle, or NULL otherwise. | |
998 | */ | |
999 | struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid, | |
1000 | int fh_len, int fh_type, struct inode *(*get_inode) | |
1001 | (struct super_block *sb, u64 ino, u32 gen)) | |
1002 | { | |
1003 | struct inode *inode = NULL; | |
1004 | ||
1005 | if (fh_len <= 2) | |
1006 | return NULL; | |
1007 | ||
1008 | switch (fh_type) { | |
1009 | case FILEID_INO32_GEN_PARENT: | |
1010 | inode = get_inode(sb, fid->i32.parent_ino, | |
1011 | (fh_len > 3 ? fid->i32.parent_gen : 0)); | |
1012 | break; | |
1013 | } | |
1014 | ||
4ea3ada2 | 1015 | return d_obtain_alias(inode); |
2596110a CH |
1016 | } |
1017 | EXPORT_SYMBOL_GPL(generic_fh_to_parent); | |
1018 | ||
1b061d92 | 1019 | /** |
ac13a829 FF |
1020 | * __generic_file_fsync - generic fsync implementation for simple filesystems |
1021 | * | |
1b061d92 | 1022 | * @file: file to synchronize |
ac13a829 FF |
1023 | * @start: start offset in bytes |
1024 | * @end: end offset in bytes (inclusive) | |
1b061d92 CH |
1025 | * @datasync: only synchronize essential metadata if true |
1026 | * | |
1027 | * This is a generic implementation of the fsync method for simple | |
1028 | * filesystems which track all non-inode metadata in the buffers list | |
1029 | * hanging off the address_space structure. | |
1030 | */ | |
ac13a829 FF |
1031 | int __generic_file_fsync(struct file *file, loff_t start, loff_t end, |
1032 | int datasync) | |
d5aacad5 | 1033 | { |
7ea80859 | 1034 | struct inode *inode = file->f_mapping->host; |
d5aacad5 AV |
1035 | int err; |
1036 | int ret; | |
1037 | ||
383aa543 | 1038 | err = file_write_and_wait_range(file, start, end); |
02c24a82 JB |
1039 | if (err) |
1040 | return err; | |
1041 | ||
5955102c | 1042 | inode_lock(inode); |
d5aacad5 | 1043 | ret = sync_mapping_buffers(inode->i_mapping); |
0ae45f63 | 1044 | if (!(inode->i_state & I_DIRTY_ALL)) |
02c24a82 | 1045 | goto out; |
d5aacad5 | 1046 | if (datasync && !(inode->i_state & I_DIRTY_DATASYNC)) |
02c24a82 | 1047 | goto out; |
d5aacad5 | 1048 | |
c3765016 | 1049 | err = sync_inode_metadata(inode, 1); |
d5aacad5 AV |
1050 | if (ret == 0) |
1051 | ret = err; | |
ac13a829 | 1052 | |
02c24a82 | 1053 | out: |
5955102c | 1054 | inode_unlock(inode); |
383aa543 JL |
1055 | /* check and advance again to catch errors after syncing out buffers */ |
1056 | err = file_check_and_advance_wb_err(file); | |
1057 | if (ret == 0) | |
1058 | ret = err; | |
1059 | return ret; | |
d5aacad5 | 1060 | } |
ac13a829 FF |
1061 | EXPORT_SYMBOL(__generic_file_fsync); |
1062 | ||
1063 | /** | |
1064 | * generic_file_fsync - generic fsync implementation for simple filesystems | |
1065 | * with flush | |
1066 | * @file: file to synchronize | |
1067 | * @start: start offset in bytes | |
1068 | * @end: end offset in bytes (inclusive) | |
1069 | * @datasync: only synchronize essential metadata if true | |
1070 | * | |
1071 | */ | |
1072 | ||
1073 | int generic_file_fsync(struct file *file, loff_t start, loff_t end, | |
1074 | int datasync) | |
1075 | { | |
1076 | struct inode *inode = file->f_mapping->host; | |
1077 | int err; | |
1078 | ||
1079 | err = __generic_file_fsync(file, start, end, datasync); | |
1080 | if (err) | |
1081 | return err; | |
1082 | return blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL); | |
1083 | } | |
1b061d92 CH |
1084 | EXPORT_SYMBOL(generic_file_fsync); |
1085 | ||
30ca22c7 PL |
1086 | /** |
1087 | * generic_check_addressable - Check addressability of file system | |
1088 | * @blocksize_bits: log of file system block size | |
1089 | * @num_blocks: number of blocks in file system | |
1090 | * | |
1091 | * Determine whether a file system with @num_blocks blocks (and a | |
1092 | * block size of 2**@blocksize_bits) is addressable by the sector_t | |
1093 | * and page cache of the system. Return 0 if so and -EFBIG otherwise. | |
1094 | */ | |
1095 | int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks) | |
1096 | { | |
1097 | u64 last_fs_block = num_blocks - 1; | |
a33f13ef | 1098 | u64 last_fs_page = |
09cbfeaf | 1099 | last_fs_block >> (PAGE_SHIFT - blocksize_bits); |
30ca22c7 PL |
1100 | |
1101 | if (unlikely(num_blocks == 0)) | |
1102 | return 0; | |
1103 | ||
09cbfeaf | 1104 | if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT)) |
30ca22c7 PL |
1105 | return -EINVAL; |
1106 | ||
a33f13ef JB |
1107 | if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) || |
1108 | (last_fs_page > (pgoff_t)(~0ULL))) { | |
30ca22c7 PL |
1109 | return -EFBIG; |
1110 | } | |
1111 | return 0; | |
1112 | } | |
1113 | EXPORT_SYMBOL(generic_check_addressable); | |
1114 | ||
1b061d92 CH |
1115 | /* |
1116 | * No-op implementation of ->fsync for in-memory filesystems. | |
1117 | */ | |
02c24a82 | 1118 | int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync) |
1b061d92 CH |
1119 | { |
1120 | return 0; | |
1121 | } | |
1b061d92 | 1122 | EXPORT_SYMBOL(noop_fsync); |
87dc800b | 1123 | |
f44c7763 DW |
1124 | int noop_set_page_dirty(struct page *page) |
1125 | { | |
1126 | /* | |
1127 | * Unlike __set_page_dirty_no_writeback that handles dirty page | |
1128 | * tracking in the page object, dax does all dirty tracking in | |
1129 | * the inode address_space in response to mkwrite faults. In the | |
1130 | * dax case we only need to worry about potentially dirty CPU | |
1131 | * caches, not dirty page cache pages to write back. | |
1132 | * | |
1133 | * This callback is defined to prevent fallback to | |
1134 | * __set_page_dirty_buffers() in set_page_dirty(). | |
1135 | */ | |
1136 | return 0; | |
1137 | } | |
1138 | EXPORT_SYMBOL_GPL(noop_set_page_dirty); | |
1139 | ||
1140 | void noop_invalidatepage(struct page *page, unsigned int offset, | |
1141 | unsigned int length) | |
1142 | { | |
1143 | /* | |
1144 | * There is no page cache to invalidate in the dax case, however | |
1145 | * we need this callback defined to prevent falling back to | |
1146 | * block_invalidatepage() in do_invalidatepage(). | |
1147 | */ | |
1148 | } | |
1149 | EXPORT_SYMBOL_GPL(noop_invalidatepage); | |
1150 | ||
1151 | ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter) | |
1152 | { | |
1153 | /* | |
1154 | * iomap based filesystems support direct I/O without need for | |
1155 | * this callback. However, it still needs to be set in | |
1156 | * inode->a_ops so that open/fcntl know that direct I/O is | |
1157 | * generally supported. | |
1158 | */ | |
1159 | return -EINVAL; | |
1160 | } | |
1161 | EXPORT_SYMBOL_GPL(noop_direct_IO); | |
1162 | ||
fceef393 AV |
1163 | /* Because kfree isn't assignment-compatible with void(void*) ;-/ */ |
1164 | void kfree_link(void *p) | |
87dc800b | 1165 | { |
fceef393 | 1166 | kfree(p); |
87dc800b | 1167 | } |
fceef393 | 1168 | EXPORT_SYMBOL(kfree_link); |
6987843f AV |
1169 | |
1170 | /* | |
1171 | * nop .set_page_dirty method so that people can use .page_mkwrite on | |
1172 | * anon inodes. | |
1173 | */ | |
1174 | static int anon_set_page_dirty(struct page *page) | |
1175 | { | |
1176 | return 0; | |
1177 | }; | |
1178 | ||
1179 | /* | |
1180 | * A single inode exists for all anon_inode files. Contrary to pipes, | |
1181 | * anon_inode inodes have no associated per-instance data, so we need | |
1182 | * only allocate one of them. | |
1183 | */ | |
1184 | struct inode *alloc_anon_inode(struct super_block *s) | |
1185 | { | |
1186 | static const struct address_space_operations anon_aops = { | |
1187 | .set_page_dirty = anon_set_page_dirty, | |
1188 | }; | |
1189 | struct inode *inode = new_inode_pseudo(s); | |
1190 | ||
1191 | if (!inode) | |
1192 | return ERR_PTR(-ENOMEM); | |
1193 | ||
1194 | inode->i_ino = get_next_ino(); | |
1195 | inode->i_mapping->a_ops = &anon_aops; | |
1196 | ||
1197 | /* | |
1198 | * Mark the inode dirty from the very beginning, | |
1199 | * that way it will never be moved to the dirty | |
1200 | * list because mark_inode_dirty() will think | |
1201 | * that it already _is_ on the dirty list. | |
1202 | */ | |
1203 | inode->i_state = I_DIRTY; | |
1204 | inode->i_mode = S_IRUSR | S_IWUSR; | |
1205 | inode->i_uid = current_fsuid(); | |
1206 | inode->i_gid = current_fsgid(); | |
1207 | inode->i_flags |= S_PRIVATE; | |
078cd827 | 1208 | inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode); |
6987843f AV |
1209 | return inode; |
1210 | } | |
1211 | EXPORT_SYMBOL(alloc_anon_inode); | |
1c994a09 JL |
1212 | |
1213 | /** | |
1214 | * simple_nosetlease - generic helper for prohibiting leases | |
1215 | * @filp: file pointer | |
1216 | * @arg: type of lease to obtain | |
1217 | * @flp: new lease supplied for insertion | |
e6f5c789 | 1218 | * @priv: private data for lm_setup operation |
1c994a09 JL |
1219 | * |
1220 | * Generic helper for filesystems that do not wish to allow leases to be set. | |
1221 | * All arguments are ignored and it just returns -EINVAL. | |
1222 | */ | |
1223 | int | |
e6f5c789 JL |
1224 | simple_nosetlease(struct file *filp, long arg, struct file_lock **flp, |
1225 | void **priv) | |
1c994a09 JL |
1226 | { |
1227 | return -EINVAL; | |
1228 | } | |
1229 | EXPORT_SYMBOL(simple_nosetlease); | |
61ba64fc | 1230 | |
6ee9706a EB |
1231 | /** |
1232 | * simple_get_link - generic helper to get the target of "fast" symlinks | |
1233 | * @dentry: not used here | |
1234 | * @inode: the symlink inode | |
1235 | * @done: not used here | |
1236 | * | |
1237 | * Generic helper for filesystems to use for symlink inodes where a pointer to | |
1238 | * the symlink target is stored in ->i_link. NOTE: this isn't normally called, | |
1239 | * since as an optimization the path lookup code uses any non-NULL ->i_link | |
1240 | * directly, without calling ->get_link(). But ->get_link() still must be set, | |
1241 | * to mark the inode_operations as being for a symlink. | |
1242 | * | |
1243 | * Return: the symlink target | |
1244 | */ | |
6b255391 | 1245 | const char *simple_get_link(struct dentry *dentry, struct inode *inode, |
fceef393 | 1246 | struct delayed_call *done) |
61ba64fc | 1247 | { |
6b255391 | 1248 | return inode->i_link; |
61ba64fc | 1249 | } |
6b255391 | 1250 | EXPORT_SYMBOL(simple_get_link); |
61ba64fc AV |
1251 | |
1252 | const struct inode_operations simple_symlink_inode_operations = { | |
6b255391 | 1253 | .get_link = simple_get_link, |
61ba64fc AV |
1254 | }; |
1255 | EXPORT_SYMBOL(simple_symlink_inode_operations); | |
fbabfd0f EB |
1256 | |
1257 | /* | |
1258 | * Operations for a permanently empty directory. | |
1259 | */ | |
1260 | static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) | |
1261 | { | |
1262 | return ERR_PTR(-ENOENT); | |
1263 | } | |
1264 | ||
a528d35e DH |
1265 | static int empty_dir_getattr(const struct path *path, struct kstat *stat, |
1266 | u32 request_mask, unsigned int query_flags) | |
fbabfd0f | 1267 | { |
a528d35e | 1268 | struct inode *inode = d_inode(path->dentry); |
fbabfd0f EB |
1269 | generic_fillattr(inode, stat); |
1270 | return 0; | |
1271 | } | |
1272 | ||
1273 | static int empty_dir_setattr(struct dentry *dentry, struct iattr *attr) | |
1274 | { | |
1275 | return -EPERM; | |
1276 | } | |
1277 | ||
fbabfd0f EB |
1278 | static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size) |
1279 | { | |
1280 | return -EOPNOTSUPP; | |
1281 | } | |
1282 | ||
1283 | static const struct inode_operations empty_dir_inode_operations = { | |
1284 | .lookup = empty_dir_lookup, | |
1285 | .permission = generic_permission, | |
1286 | .setattr = empty_dir_setattr, | |
1287 | .getattr = empty_dir_getattr, | |
fbabfd0f EB |
1288 | .listxattr = empty_dir_listxattr, |
1289 | }; | |
1290 | ||
1291 | static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence) | |
1292 | { | |
1293 | /* An empty directory has two entries . and .. at offsets 0 and 1 */ | |
1294 | return generic_file_llseek_size(file, offset, whence, 2, 2); | |
1295 | } | |
1296 | ||
1297 | static int empty_dir_readdir(struct file *file, struct dir_context *ctx) | |
1298 | { | |
1299 | dir_emit_dots(file, ctx); | |
1300 | return 0; | |
1301 | } | |
1302 | ||
1303 | static const struct file_operations empty_dir_operations = { | |
1304 | .llseek = empty_dir_llseek, | |
1305 | .read = generic_read_dir, | |
c51da20c | 1306 | .iterate_shared = empty_dir_readdir, |
fbabfd0f EB |
1307 | .fsync = noop_fsync, |
1308 | }; | |
1309 | ||
1310 | ||
1311 | void make_empty_dir_inode(struct inode *inode) | |
1312 | { | |
1313 | set_nlink(inode, 2); | |
1314 | inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO; | |
1315 | inode->i_uid = GLOBAL_ROOT_UID; | |
1316 | inode->i_gid = GLOBAL_ROOT_GID; | |
1317 | inode->i_rdev = 0; | |
4b75de86 | 1318 | inode->i_size = 0; |
fbabfd0f EB |
1319 | inode->i_blkbits = PAGE_SHIFT; |
1320 | inode->i_blocks = 0; | |
1321 | ||
1322 | inode->i_op = &empty_dir_inode_operations; | |
f5c24438 | 1323 | inode->i_opflags &= ~IOP_XATTR; |
fbabfd0f EB |
1324 | inode->i_fop = &empty_dir_operations; |
1325 | } | |
1326 | ||
1327 | bool is_empty_dir_inode(struct inode *inode) | |
1328 | { | |
1329 | return (inode->i_fop == &empty_dir_operations) && | |
1330 | (inode->i_op == &empty_dir_inode_operations); | |
1331 | } |