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> |
5ca14835 | 18 | #include <linux/iversion.h> |
d5aacad5 | 19 | #include <linux/writeback.h> |
ff01bb48 | 20 | #include <linux/buffer_head.h> /* sync_mapping_buffers */ |
31d6d5ce DH |
21 | #include <linux/fs_context.h> |
22 | #include <linux/pseudo_fs.h> | |
a3d1e7eb | 23 | #include <linux/fsnotify.h> |
c843843e DR |
24 | #include <linux/unicode.h> |
25 | #include <linux/fscrypt.h> | |
7cf34c76 | 26 | |
7c0f6ba6 | 27 | #include <linux/uaccess.h> |
1da177e4 | 28 | |
a4464dbc AV |
29 | #include "internal.h" |
30 | ||
b74d24f7 | 31 | int simple_getattr(struct mnt_idmap *idmap, const struct path *path, |
549c7297 CB |
32 | struct kstat *stat, u32 request_mask, |
33 | unsigned int query_flags) | |
1da177e4 | 34 | { |
a528d35e | 35 | struct inode *inode = d_inode(path->dentry); |
0d72b928 | 36 | generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat); |
09cbfeaf | 37 | stat->blocks = inode->i_mapping->nrpages << (PAGE_SHIFT - 9); |
1da177e4 LT |
38 | return 0; |
39 | } | |
12f38872 | 40 | EXPORT_SYMBOL(simple_getattr); |
1da177e4 | 41 | |
726c3342 | 42 | int simple_statfs(struct dentry *dentry, struct kstatfs *buf) |
1da177e4 | 43 | { |
ae62bcb5 AG |
44 | u64 id = huge_encode_dev(dentry->d_sb->s_dev); |
45 | ||
46 | buf->f_fsid = u64_to_fsid(id); | |
726c3342 | 47 | buf->f_type = dentry->d_sb->s_magic; |
09cbfeaf | 48 | buf->f_bsize = PAGE_SIZE; |
1da177e4 LT |
49 | buf->f_namelen = NAME_MAX; |
50 | return 0; | |
51 | } | |
12f38872 | 52 | EXPORT_SYMBOL(simple_statfs); |
1da177e4 LT |
53 | |
54 | /* | |
55 | * Retaining negative dentries for an in-memory filesystem just wastes | |
56 | * memory and lookup time: arrange for them to be deleted immediately. | |
57 | */ | |
b26d4cd3 | 58 | int always_delete_dentry(const struct dentry *dentry) |
1da177e4 LT |
59 | { |
60 | return 1; | |
61 | } | |
b26d4cd3 AV |
62 | EXPORT_SYMBOL(always_delete_dentry); |
63 | ||
64 | const struct dentry_operations simple_dentry_operations = { | |
65 | .d_delete = always_delete_dentry, | |
66 | }; | |
67 | EXPORT_SYMBOL(simple_dentry_operations); | |
1da177e4 LT |
68 | |
69 | /* | |
70 | * Lookup the data. This is trivial - if the dentry didn't already | |
71 | * exist, we know it is negative. Set d_op to delete negative dentries. | |
72 | */ | |
00cd8dd3 | 73 | struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
1da177e4 | 74 | { |
1da177e4 LT |
75 | if (dentry->d_name.len > NAME_MAX) |
76 | return ERR_PTR(-ENAMETOOLONG); | |
74931da7 AV |
77 | if (!dentry->d_sb->s_d_op) |
78 | d_set_d_op(dentry, &simple_dentry_operations); | |
1da177e4 LT |
79 | d_add(dentry, NULL); |
80 | return NULL; | |
81 | } | |
12f38872 | 82 | EXPORT_SYMBOL(simple_lookup); |
1da177e4 | 83 | |
1da177e4 LT |
84 | int dcache_dir_open(struct inode *inode, struct file *file) |
85 | { | |
ba65dc5e | 86 | file->private_data = d_alloc_cursor(file->f_path.dentry); |
1da177e4 LT |
87 | |
88 | return file->private_data ? 0 : -ENOMEM; | |
89 | } | |
12f38872 | 90 | EXPORT_SYMBOL(dcache_dir_open); |
1da177e4 LT |
91 | |
92 | int dcache_dir_close(struct inode *inode, struct file *file) | |
93 | { | |
94 | dput(file->private_data); | |
95 | return 0; | |
96 | } | |
12f38872 | 97 | EXPORT_SYMBOL(dcache_dir_close); |
1da177e4 | 98 | |
4f42c1b5 | 99 | /* parent is locked at least shared */ |
d4f4de5e AV |
100 | /* |
101 | * Returns an element of siblings' list. | |
102 | * We are looking for <count>th positive after <p>; if | |
26b6c984 AV |
103 | * found, dentry is grabbed and returned to caller. |
104 | * If no such element exists, NULL is returned. | |
d4f4de5e | 105 | */ |
26b6c984 | 106 | static struct dentry *scan_positives(struct dentry *cursor, |
d4f4de5e AV |
107 | struct list_head *p, |
108 | loff_t count, | |
26b6c984 | 109 | struct dentry *last) |
4f42c1b5 | 110 | { |
d4f4de5e AV |
111 | struct dentry *dentry = cursor->d_parent, *found = NULL; |
112 | ||
113 | spin_lock(&dentry->d_lock); | |
114 | while ((p = p->next) != &dentry->d_subdirs) { | |
4f42c1b5 | 115 | struct dentry *d = list_entry(p, struct dentry, d_child); |
d4f4de5e AV |
116 | // we must at least skip cursors, to avoid livelocks |
117 | if (d->d_flags & DCACHE_DENTRY_CURSOR) | |
118 | continue; | |
119 | if (simple_positive(d) && !--count) { | |
120 | spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED); | |
121 | if (simple_positive(d)) | |
122 | found = dget_dlock(d); | |
123 | spin_unlock(&d->d_lock); | |
124 | if (likely(found)) | |
125 | break; | |
126 | count = 1; | |
127 | } | |
128 | if (need_resched()) { | |
129 | list_move(&cursor->d_child, p); | |
130 | p = &cursor->d_child; | |
131 | spin_unlock(&dentry->d_lock); | |
132 | cond_resched(); | |
133 | spin_lock(&dentry->d_lock); | |
4f42c1b5 AV |
134 | } |
135 | } | |
d4f4de5e | 136 | spin_unlock(&dentry->d_lock); |
26b6c984 AV |
137 | dput(last); |
138 | return found; | |
4f42c1b5 AV |
139 | } |
140 | ||
965c8e59 | 141 | loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence) |
1da177e4 | 142 | { |
2fd6b7f5 | 143 | struct dentry *dentry = file->f_path.dentry; |
965c8e59 | 144 | switch (whence) { |
1da177e4 LT |
145 | case 1: |
146 | offset += file->f_pos; | |
df561f66 | 147 | fallthrough; |
1da177e4 LT |
148 | case 0: |
149 | if (offset >= 0) | |
150 | break; | |
df561f66 | 151 | fallthrough; |
1da177e4 | 152 | default: |
1da177e4 LT |
153 | return -EINVAL; |
154 | } | |
155 | if (offset != file->f_pos) { | |
d4f4de5e AV |
156 | struct dentry *cursor = file->private_data; |
157 | struct dentry *to = NULL; | |
d4f4de5e | 158 | |
d4f4de5e AV |
159 | inode_lock_shared(dentry->d_inode); |
160 | ||
26b6c984 AV |
161 | if (offset > 2) |
162 | to = scan_positives(cursor, &dentry->d_subdirs, | |
163 | offset - 2, NULL); | |
164 | spin_lock(&dentry->d_lock); | |
165 | if (to) | |
166 | list_move(&cursor->d_child, &to->d_child); | |
167 | else | |
d4f4de5e | 168 | list_del_init(&cursor->d_child); |
26b6c984 | 169 | spin_unlock(&dentry->d_lock); |
d4f4de5e AV |
170 | dput(to); |
171 | ||
26b6c984 AV |
172 | file->f_pos = offset; |
173 | ||
d4f4de5e | 174 | inode_unlock_shared(dentry->d_inode); |
1da177e4 | 175 | } |
1da177e4 LT |
176 | return offset; |
177 | } | |
12f38872 | 178 | EXPORT_SYMBOL(dcache_dir_lseek); |
1da177e4 | 179 | |
1da177e4 LT |
180 | /* |
181 | * Directory is locked and all positive dentries in it are safe, since | |
182 | * for ramfs-type trees they can't go away without unlink() or rmdir(), | |
183 | * both impossible due to the lock on directory. | |
184 | */ | |
185 | ||
5f99f4e7 | 186 | int dcache_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 187 | { |
5f99f4e7 AV |
188 | struct dentry *dentry = file->f_path.dentry; |
189 | struct dentry *cursor = file->private_data; | |
d4f4de5e AV |
190 | struct list_head *anchor = &dentry->d_subdirs; |
191 | struct dentry *next = NULL; | |
192 | struct list_head *p; | |
1da177e4 | 193 | |
5f99f4e7 AV |
194 | if (!dir_emit_dots(file, ctx)) |
195 | return 0; | |
5f99f4e7 | 196 | |
4f42c1b5 | 197 | if (ctx->pos == 2) |
d4f4de5e | 198 | p = anchor; |
26b6c984 | 199 | else if (!list_empty(&cursor->d_child)) |
d4f4de5e | 200 | p = &cursor->d_child; |
26b6c984 AV |
201 | else |
202 | return 0; | |
d4f4de5e | 203 | |
26b6c984 | 204 | while ((next = scan_positives(cursor, p, 1, next)) != NULL) { |
5f99f4e7 | 205 | if (!dir_emit(ctx, next->d_name.name, next->d_name.len, |
364595a6 JL |
206 | d_inode(next)->i_ino, |
207 | fs_umode_to_dtype(d_inode(next)->i_mode))) | |
4f42c1b5 | 208 | break; |
5f99f4e7 | 209 | ctx->pos++; |
26b6c984 | 210 | p = &next->d_child; |
1da177e4 | 211 | } |
d4f4de5e | 212 | spin_lock(&dentry->d_lock); |
26b6c984 AV |
213 | if (next) |
214 | list_move_tail(&cursor->d_child, &next->d_child); | |
215 | else | |
216 | list_del_init(&cursor->d_child); | |
d4f4de5e AV |
217 | spin_unlock(&dentry->d_lock); |
218 | dput(next); | |
219 | ||
1da177e4 LT |
220 | return 0; |
221 | } | |
12f38872 | 222 | EXPORT_SYMBOL(dcache_readdir); |
1da177e4 LT |
223 | |
224 | ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos) | |
225 | { | |
226 | return -EISDIR; | |
227 | } | |
12f38872 | 228 | EXPORT_SYMBOL(generic_read_dir); |
1da177e4 | 229 | |
4b6f5d20 | 230 | const struct file_operations simple_dir_operations = { |
1da177e4 LT |
231 | .open = dcache_dir_open, |
232 | .release = dcache_dir_close, | |
233 | .llseek = dcache_dir_lseek, | |
234 | .read = generic_read_dir, | |
4e82901c | 235 | .iterate_shared = dcache_readdir, |
1b061d92 | 236 | .fsync = noop_fsync, |
1da177e4 | 237 | }; |
12f38872 | 238 | EXPORT_SYMBOL(simple_dir_operations); |
1da177e4 | 239 | |
92e1d5be | 240 | const struct inode_operations simple_dir_inode_operations = { |
1da177e4 LT |
241 | .lookup = simple_lookup, |
242 | }; | |
12f38872 | 243 | EXPORT_SYMBOL(simple_dir_inode_operations); |
1da177e4 | 244 | |
6faddda6 CL |
245 | static void offset_set(struct dentry *dentry, u32 offset) |
246 | { | |
247 | dentry->d_fsdata = (void *)((uintptr_t)(offset)); | |
248 | } | |
249 | ||
250 | static u32 dentry2offset(struct dentry *dentry) | |
251 | { | |
252 | return (u32)((uintptr_t)(dentry->d_fsdata)); | |
253 | } | |
254 | ||
bbaef797 CL |
255 | static struct lock_class_key simple_offset_xa_lock; |
256 | ||
6faddda6 CL |
257 | /** |
258 | * simple_offset_init - initialize an offset_ctx | |
259 | * @octx: directory offset map to be initialized | |
260 | * | |
261 | */ | |
262 | void simple_offset_init(struct offset_ctx *octx) | |
263 | { | |
264 | xa_init_flags(&octx->xa, XA_FLAGS_ALLOC1); | |
bbaef797 | 265 | lockdep_set_class(&octx->xa.xa_lock, &simple_offset_xa_lock); |
6faddda6 CL |
266 | |
267 | /* 0 is '.', 1 is '..', so always start with offset 2 */ | |
268 | octx->next_offset = 2; | |
269 | } | |
270 | ||
271 | /** | |
272 | * simple_offset_add - Add an entry to a directory's offset map | |
273 | * @octx: directory offset ctx to be updated | |
274 | * @dentry: new dentry being added | |
275 | * | |
276 | * Returns zero on success. @so_ctx and the dentry offset are updated. | |
277 | * Otherwise, a negative errno value is returned. | |
278 | */ | |
279 | int simple_offset_add(struct offset_ctx *octx, struct dentry *dentry) | |
280 | { | |
281 | static const struct xa_limit limit = XA_LIMIT(2, U32_MAX); | |
282 | u32 offset; | |
283 | int ret; | |
284 | ||
285 | if (dentry2offset(dentry) != 0) | |
286 | return -EBUSY; | |
287 | ||
288 | ret = xa_alloc_cyclic(&octx->xa, &offset, dentry, limit, | |
289 | &octx->next_offset, GFP_KERNEL); | |
290 | if (ret < 0) | |
291 | return ret; | |
292 | ||
293 | offset_set(dentry, offset); | |
294 | return 0; | |
295 | } | |
296 | ||
297 | /** | |
298 | * simple_offset_remove - Remove an entry to a directory's offset map | |
299 | * @octx: directory offset ctx to be updated | |
300 | * @dentry: dentry being removed | |
301 | * | |
302 | */ | |
303 | void simple_offset_remove(struct offset_ctx *octx, struct dentry *dentry) | |
304 | { | |
305 | u32 offset; | |
306 | ||
307 | offset = dentry2offset(dentry); | |
308 | if (offset == 0) | |
309 | return; | |
310 | ||
311 | xa_erase(&octx->xa, offset); | |
312 | offset_set(dentry, 0); | |
313 | } | |
314 | ||
315 | /** | |
316 | * simple_offset_rename_exchange - exchange rename with directory offsets | |
317 | * @old_dir: parent of dentry being moved | |
318 | * @old_dentry: dentry being moved | |
319 | * @new_dir: destination parent | |
320 | * @new_dentry: destination dentry | |
321 | * | |
322 | * Returns zero on success. Otherwise a negative errno is returned and the | |
323 | * rename is rolled back. | |
324 | */ | |
325 | int simple_offset_rename_exchange(struct inode *old_dir, | |
326 | struct dentry *old_dentry, | |
327 | struct inode *new_dir, | |
328 | struct dentry *new_dentry) | |
329 | { | |
330 | struct offset_ctx *old_ctx = old_dir->i_op->get_offset_ctx(old_dir); | |
331 | struct offset_ctx *new_ctx = new_dir->i_op->get_offset_ctx(new_dir); | |
332 | u32 old_index = dentry2offset(old_dentry); | |
333 | u32 new_index = dentry2offset(new_dentry); | |
334 | int ret; | |
335 | ||
336 | simple_offset_remove(old_ctx, old_dentry); | |
337 | simple_offset_remove(new_ctx, new_dentry); | |
338 | ||
339 | ret = simple_offset_add(new_ctx, old_dentry); | |
340 | if (ret) | |
341 | goto out_restore; | |
342 | ||
343 | ret = simple_offset_add(old_ctx, new_dentry); | |
344 | if (ret) { | |
345 | simple_offset_remove(new_ctx, old_dentry); | |
346 | goto out_restore; | |
347 | } | |
348 | ||
349 | ret = simple_rename_exchange(old_dir, old_dentry, new_dir, new_dentry); | |
350 | if (ret) { | |
351 | simple_offset_remove(new_ctx, old_dentry); | |
352 | simple_offset_remove(old_ctx, new_dentry); | |
353 | goto out_restore; | |
354 | } | |
355 | return 0; | |
356 | ||
357 | out_restore: | |
358 | offset_set(old_dentry, old_index); | |
359 | xa_store(&old_ctx->xa, old_index, old_dentry, GFP_KERNEL); | |
360 | offset_set(new_dentry, new_index); | |
361 | xa_store(&new_ctx->xa, new_index, new_dentry, GFP_KERNEL); | |
362 | return ret; | |
363 | } | |
364 | ||
365 | /** | |
366 | * simple_offset_destroy - Release offset map | |
367 | * @octx: directory offset ctx that is about to be destroyed | |
368 | * | |
369 | * During fs teardown (eg. umount), a directory's offset map might still | |
370 | * contain entries. xa_destroy() cleans out anything that remains. | |
371 | */ | |
372 | void simple_offset_destroy(struct offset_ctx *octx) | |
373 | { | |
374 | xa_destroy(&octx->xa); | |
375 | } | |
376 | ||
377 | /** | |
378 | * offset_dir_llseek - Advance the read position of a directory descriptor | |
379 | * @file: an open directory whose position is to be updated | |
380 | * @offset: a byte offset | |
381 | * @whence: enumerator describing the starting position for this update | |
382 | * | |
383 | * SEEK_END, SEEK_DATA, and SEEK_HOLE are not supported for directories. | |
384 | * | |
385 | * Returns the updated read position if successful; otherwise a | |
386 | * negative errno is returned and the read position remains unchanged. | |
387 | */ | |
388 | static loff_t offset_dir_llseek(struct file *file, loff_t offset, int whence) | |
389 | { | |
390 | switch (whence) { | |
391 | case SEEK_CUR: | |
392 | offset += file->f_pos; | |
393 | fallthrough; | |
394 | case SEEK_SET: | |
395 | if (offset >= 0) | |
396 | break; | |
397 | fallthrough; | |
398 | default: | |
399 | return -EINVAL; | |
400 | } | |
401 | ||
402 | return vfs_setpos(file, offset, U32_MAX); | |
403 | } | |
404 | ||
405 | static struct dentry *offset_find_next(struct xa_state *xas) | |
406 | { | |
407 | struct dentry *child, *found = NULL; | |
408 | ||
409 | rcu_read_lock(); | |
410 | child = xas_next_entry(xas, U32_MAX); | |
411 | if (!child) | |
412 | goto out; | |
2be4f05a | 413 | spin_lock(&child->d_lock); |
6faddda6 CL |
414 | if (simple_positive(child)) |
415 | found = dget_dlock(child); | |
416 | spin_unlock(&child->d_lock); | |
417 | out: | |
418 | rcu_read_unlock(); | |
419 | return found; | |
420 | } | |
421 | ||
422 | static bool offset_dir_emit(struct dir_context *ctx, struct dentry *dentry) | |
423 | { | |
424 | u32 offset = dentry2offset(dentry); | |
425 | struct inode *inode = d_inode(dentry); | |
426 | ||
427 | return ctx->actor(ctx, dentry->d_name.name, dentry->d_name.len, offset, | |
428 | inode->i_ino, fs_umode_to_dtype(inode->i_mode)); | |
429 | } | |
430 | ||
2be4f05a | 431 | static void offset_iterate_dir(struct inode *inode, struct dir_context *ctx) |
6faddda6 | 432 | { |
6faddda6 CL |
433 | struct offset_ctx *so_ctx = inode->i_op->get_offset_ctx(inode); |
434 | XA_STATE(xas, &so_ctx->xa, ctx->pos); | |
435 | struct dentry *dentry; | |
436 | ||
437 | while (true) { | |
6faddda6 | 438 | dentry = offset_find_next(&xas); |
6faddda6 CL |
439 | if (!dentry) |
440 | break; | |
441 | ||
442 | if (!offset_dir_emit(ctx, dentry)) { | |
443 | dput(dentry); | |
444 | break; | |
445 | } | |
446 | ||
447 | dput(dentry); | |
448 | ctx->pos = xas.xa_index + 1; | |
449 | } | |
450 | } | |
451 | ||
452 | /** | |
453 | * offset_readdir - Emit entries starting at offset @ctx->pos | |
454 | * @file: an open directory to iterate over | |
455 | * @ctx: directory iteration context | |
456 | * | |
457 | * Caller must hold @file's i_rwsem to prevent insertion or removal of | |
458 | * entries during this call. | |
459 | * | |
460 | * On entry, @ctx->pos contains an offset that represents the first entry | |
461 | * to be read from the directory. | |
462 | * | |
463 | * The operation continues until there are no more entries to read, or | |
464 | * until the ctx->actor indicates there is no more space in the caller's | |
465 | * output buffer. | |
466 | * | |
467 | * On return, @ctx->pos contains an offset that will read the next entry | |
2be4f05a | 468 | * in this directory when offset_readdir() is called again with @ctx. |
6faddda6 CL |
469 | * |
470 | * Return values: | |
471 | * %0 - Complete | |
472 | */ | |
473 | static int offset_readdir(struct file *file, struct dir_context *ctx) | |
474 | { | |
475 | struct dentry *dir = file->f_path.dentry; | |
476 | ||
477 | lockdep_assert_held(&d_inode(dir)->i_rwsem); | |
478 | ||
479 | if (!dir_emit_dots(file, ctx)) | |
480 | return 0; | |
481 | ||
2be4f05a | 482 | offset_iterate_dir(d_inode(dir), ctx); |
6faddda6 CL |
483 | return 0; |
484 | } | |
485 | ||
486 | const struct file_operations simple_offset_dir_operations = { | |
487 | .llseek = offset_dir_llseek, | |
488 | .iterate_shared = offset_readdir, | |
489 | .read = generic_read_dir, | |
490 | .fsync = noop_fsync, | |
491 | }; | |
492 | ||
a3d1e7eb AV |
493 | static struct dentry *find_next_child(struct dentry *parent, struct dentry *prev) |
494 | { | |
495 | struct dentry *child = NULL; | |
496 | struct list_head *p = prev ? &prev->d_child : &parent->d_subdirs; | |
497 | ||
498 | spin_lock(&parent->d_lock); | |
499 | while ((p = p->next) != &parent->d_subdirs) { | |
500 | struct dentry *d = container_of(p, struct dentry, d_child); | |
501 | if (simple_positive(d)) { | |
502 | spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED); | |
503 | if (simple_positive(d)) | |
504 | child = dget_dlock(d); | |
505 | spin_unlock(&d->d_lock); | |
506 | if (likely(child)) | |
507 | break; | |
508 | } | |
509 | } | |
510 | spin_unlock(&parent->d_lock); | |
511 | dput(prev); | |
512 | return child; | |
513 | } | |
514 | ||
515 | void simple_recursive_removal(struct dentry *dentry, | |
516 | void (*callback)(struct dentry *)) | |
517 | { | |
518 | struct dentry *this = dget(dentry); | |
519 | while (true) { | |
520 | struct dentry *victim = NULL, *child; | |
521 | struct inode *inode = this->d_inode; | |
522 | ||
523 | inode_lock(inode); | |
524 | if (d_is_dir(this)) | |
525 | inode->i_flags |= S_DEAD; | |
526 | while ((child = find_next_child(this, victim)) == NULL) { | |
527 | // kill and ascend | |
528 | // update metadata while it's still locked | |
f7f43858 | 529 | inode_set_ctime_current(inode); |
a3d1e7eb AV |
530 | clear_nlink(inode); |
531 | inode_unlock(inode); | |
532 | victim = this; | |
533 | this = this->d_parent; | |
534 | inode = this->d_inode; | |
535 | inode_lock(inode); | |
536 | if (simple_positive(victim)) { | |
537 | d_invalidate(victim); // avoid lost mounts | |
538 | if (d_is_dir(victim)) | |
539 | fsnotify_rmdir(inode, victim); | |
540 | else | |
541 | fsnotify_unlink(inode, victim); | |
542 | if (callback) | |
543 | callback(victim); | |
544 | dput(victim); // unpin it | |
545 | } | |
546 | if (victim == dentry) { | |
077c212f JL |
547 | inode_set_mtime_to_ts(inode, |
548 | inode_set_ctime_current(inode)); | |
a3d1e7eb AV |
549 | if (d_is_dir(dentry)) |
550 | drop_nlink(inode); | |
551 | inode_unlock(inode); | |
552 | dput(dentry); | |
553 | return; | |
554 | } | |
555 | } | |
556 | inode_unlock(inode); | |
557 | this = child; | |
558 | } | |
559 | } | |
560 | EXPORT_SYMBOL(simple_recursive_removal); | |
561 | ||
759b9775 HD |
562 | static const struct super_operations simple_super_operations = { |
563 | .statfs = simple_statfs, | |
564 | }; | |
565 | ||
db2c246a | 566 | static int pseudo_fs_fill_super(struct super_block *s, struct fs_context *fc) |
1da177e4 | 567 | { |
31d6d5ce | 568 | struct pseudo_fs_context *ctx = fc->fs_private; |
1da177e4 | 569 | struct inode *root; |
1da177e4 | 570 | |
89a4eb4b | 571 | s->s_maxbytes = MAX_LFS_FILESIZE; |
3971e1a9 AN |
572 | s->s_blocksize = PAGE_SIZE; |
573 | s->s_blocksize_bits = PAGE_SHIFT; | |
8d9e46d8 AV |
574 | s->s_magic = ctx->magic; |
575 | s->s_op = ctx->ops ?: &simple_super_operations; | |
576 | s->s_xattr = ctx->xattr; | |
1da177e4 LT |
577 | s->s_time_gran = 1; |
578 | root = new_inode(s); | |
579 | if (!root) | |
db2c246a DH |
580 | return -ENOMEM; |
581 | ||
1a1c9bb4 JL |
582 | /* |
583 | * since this is the first inode, make it number 1. New inodes created | |
584 | * after this must take care not to collide with it (by passing | |
585 | * max_reserved of 1 to iunique). | |
586 | */ | |
587 | root->i_ino = 1; | |
1da177e4 | 588 | root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR; |
077c212f | 589 | simple_inode_init_ts(root); |
8d9e46d8 AV |
590 | s->s_root = d_make_root(root); |
591 | if (!s->s_root) | |
db2c246a | 592 | return -ENOMEM; |
8d9e46d8 | 593 | s->s_d_op = ctx->dops; |
31d6d5ce | 594 | return 0; |
db2c246a | 595 | } |
8d9e46d8 | 596 | |
db2c246a DH |
597 | static int pseudo_fs_get_tree(struct fs_context *fc) |
598 | { | |
2ac295d4 | 599 | return get_tree_nodev(fc, pseudo_fs_fill_super); |
31d6d5ce DH |
600 | } |
601 | ||
602 | static void pseudo_fs_free(struct fs_context *fc) | |
603 | { | |
604 | kfree(fc->fs_private); | |
605 | } | |
606 | ||
607 | static const struct fs_context_operations pseudo_fs_context_ops = { | |
608 | .free = pseudo_fs_free, | |
609 | .get_tree = pseudo_fs_get_tree, | |
610 | }; | |
611 | ||
612 | /* | |
613 | * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that | |
614 | * will never be mountable) | |
615 | */ | |
616 | struct pseudo_fs_context *init_pseudo(struct fs_context *fc, | |
617 | unsigned long magic) | |
618 | { | |
619 | struct pseudo_fs_context *ctx; | |
620 | ||
621 | ctx = kzalloc(sizeof(struct pseudo_fs_context), GFP_KERNEL); | |
622 | if (likely(ctx)) { | |
623 | ctx->magic = magic; | |
624 | fc->fs_private = ctx; | |
625 | fc->ops = &pseudo_fs_context_ops; | |
db2c246a DH |
626 | fc->sb_flags |= SB_NOUSER; |
627 | fc->global = true; | |
1da177e4 | 628 | } |
31d6d5ce | 629 | return ctx; |
1da177e4 | 630 | } |
31d6d5ce | 631 | EXPORT_SYMBOL(init_pseudo); |
1da177e4 | 632 | |
20955e89 SB |
633 | int simple_open(struct inode *inode, struct file *file) |
634 | { | |
635 | if (inode->i_private) | |
636 | file->private_data = inode->i_private; | |
637 | return 0; | |
638 | } | |
12f38872 | 639 | EXPORT_SYMBOL(simple_open); |
20955e89 | 640 | |
1da177e4 LT |
641 | int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry) |
642 | { | |
dea655c2 | 643 | struct inode *inode = d_inode(old_dentry); |
1da177e4 | 644 | |
077c212f JL |
645 | inode_set_mtime_to_ts(dir, |
646 | inode_set_ctime_to_ts(dir, inode_set_ctime_current(inode))); | |
d8c76e6f | 647 | inc_nlink(inode); |
7de9c6ee | 648 | ihold(inode); |
1da177e4 LT |
649 | dget(dentry); |
650 | d_instantiate(dentry, inode); | |
651 | return 0; | |
652 | } | |
12f38872 | 653 | EXPORT_SYMBOL(simple_link); |
1da177e4 | 654 | |
1da177e4 LT |
655 | int simple_empty(struct dentry *dentry) |
656 | { | |
657 | struct dentry *child; | |
658 | int ret = 0; | |
659 | ||
2fd6b7f5 | 660 | spin_lock(&dentry->d_lock); |
946e51f2 | 661 | list_for_each_entry(child, &dentry->d_subdirs, d_child) { |
da502956 NP |
662 | spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED); |
663 | if (simple_positive(child)) { | |
664 | spin_unlock(&child->d_lock); | |
1da177e4 | 665 | goto out; |
da502956 NP |
666 | } |
667 | spin_unlock(&child->d_lock); | |
668 | } | |
1da177e4 LT |
669 | ret = 1; |
670 | out: | |
2fd6b7f5 | 671 | spin_unlock(&dentry->d_lock); |
1da177e4 LT |
672 | return ret; |
673 | } | |
12f38872 | 674 | EXPORT_SYMBOL(simple_empty); |
1da177e4 LT |
675 | |
676 | int simple_unlink(struct inode *dir, struct dentry *dentry) | |
677 | { | |
dea655c2 | 678 | struct inode *inode = d_inode(dentry); |
1da177e4 | 679 | |
077c212f JL |
680 | inode_set_mtime_to_ts(dir, |
681 | inode_set_ctime_to_ts(dir, inode_set_ctime_current(inode))); | |
9a53c3a7 | 682 | drop_nlink(inode); |
1da177e4 LT |
683 | dput(dentry); |
684 | return 0; | |
685 | } | |
12f38872 | 686 | EXPORT_SYMBOL(simple_unlink); |
1da177e4 LT |
687 | |
688 | int simple_rmdir(struct inode *dir, struct dentry *dentry) | |
689 | { | |
690 | if (!simple_empty(dentry)) | |
691 | return -ENOTEMPTY; | |
692 | ||
dea655c2 | 693 | drop_nlink(d_inode(dentry)); |
1da177e4 | 694 | simple_unlink(dir, dentry); |
9a53c3a7 | 695 | drop_nlink(dir); |
1da177e4 LT |
696 | return 0; |
697 | } | |
12f38872 | 698 | EXPORT_SYMBOL(simple_rmdir); |
1da177e4 | 699 | |
0c476792 JL |
700 | /** |
701 | * simple_rename_timestamp - update the various inode timestamps for rename | |
702 | * @old_dir: old parent directory | |
703 | * @old_dentry: dentry that is being renamed | |
704 | * @new_dir: new parent directory | |
705 | * @new_dentry: target for rename | |
706 | * | |
707 | * POSIX mandates that the old and new parent directories have their ctime and | |
708 | * mtime updated, and that inodes of @old_dentry and @new_dentry (if any), have | |
709 | * their ctime updated. | |
710 | */ | |
711 | void simple_rename_timestamp(struct inode *old_dir, struct dentry *old_dentry, | |
712 | struct inode *new_dir, struct dentry *new_dentry) | |
713 | { | |
714 | struct inode *newino = d_inode(new_dentry); | |
715 | ||
077c212f | 716 | inode_set_mtime_to_ts(old_dir, inode_set_ctime_current(old_dir)); |
0c476792 | 717 | if (new_dir != old_dir) |
077c212f JL |
718 | inode_set_mtime_to_ts(new_dir, |
719 | inode_set_ctime_current(new_dir)); | |
0c476792 JL |
720 | inode_set_ctime_current(d_inode(old_dentry)); |
721 | if (newino) | |
722 | inode_set_ctime_current(newino); | |
723 | } | |
724 | EXPORT_SYMBOL_GPL(simple_rename_timestamp); | |
725 | ||
6429e463 LB |
726 | int simple_rename_exchange(struct inode *old_dir, struct dentry *old_dentry, |
727 | struct inode *new_dir, struct dentry *new_dentry) | |
728 | { | |
729 | bool old_is_dir = d_is_dir(old_dentry); | |
730 | bool new_is_dir = d_is_dir(new_dentry); | |
731 | ||
732 | if (old_dir != new_dir && old_is_dir != new_is_dir) { | |
733 | if (old_is_dir) { | |
734 | drop_nlink(old_dir); | |
735 | inc_nlink(new_dir); | |
736 | } else { | |
737 | drop_nlink(new_dir); | |
738 | inc_nlink(old_dir); | |
739 | } | |
740 | } | |
0c476792 | 741 | simple_rename_timestamp(old_dir, old_dentry, new_dir, new_dentry); |
6429e463 LB |
742 | return 0; |
743 | } | |
744 | EXPORT_SYMBOL_GPL(simple_rename_exchange); | |
745 | ||
e18275ae | 746 | int simple_rename(struct mnt_idmap *idmap, struct inode *old_dir, |
549c7297 CB |
747 | struct dentry *old_dentry, struct inode *new_dir, |
748 | struct dentry *new_dentry, unsigned int flags) | |
1da177e4 | 749 | { |
e36cb0b8 | 750 | int they_are_dirs = d_is_dir(old_dentry); |
1da177e4 | 751 | |
3871cb8c | 752 | if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE)) |
e0e0be8a MS |
753 | return -EINVAL; |
754 | ||
3871cb8c LB |
755 | if (flags & RENAME_EXCHANGE) |
756 | return simple_rename_exchange(old_dir, old_dentry, new_dir, new_dentry); | |
757 | ||
1da177e4 LT |
758 | if (!simple_empty(new_dentry)) |
759 | return -ENOTEMPTY; | |
760 | ||
dea655c2 | 761 | if (d_really_is_positive(new_dentry)) { |
1da177e4 | 762 | simple_unlink(new_dir, new_dentry); |
841590ce | 763 | if (they_are_dirs) { |
dea655c2 | 764 | drop_nlink(d_inode(new_dentry)); |
9a53c3a7 | 765 | drop_nlink(old_dir); |
841590ce | 766 | } |
1da177e4 | 767 | } else if (they_are_dirs) { |
9a53c3a7 | 768 | drop_nlink(old_dir); |
d8c76e6f | 769 | inc_nlink(new_dir); |
1da177e4 LT |
770 | } |
771 | ||
0c476792 | 772 | simple_rename_timestamp(old_dir, old_dentry, new_dir, new_dentry); |
1da177e4 LT |
773 | return 0; |
774 | } | |
12f38872 | 775 | EXPORT_SYMBOL(simple_rename); |
1da177e4 | 776 | |
7bb46a67 | 777 | /** |
eef2380c | 778 | * simple_setattr - setattr for simple filesystem |
c1632a0f | 779 | * @idmap: idmap of the target mount |
7bb46a67 | 780 | * @dentry: dentry |
781 | * @iattr: iattr structure | |
782 | * | |
783 | * Returns 0 on success, -error on failure. | |
784 | * | |
eef2380c CH |
785 | * simple_setattr is a simple ->setattr implementation without a proper |
786 | * implementation of size changes. | |
787 | * | |
788 | * It can either be used for in-memory filesystems or special files | |
789 | * on simple regular filesystems. Anything that needs to change on-disk | |
790 | * or wire state on size changes needs its own setattr method. | |
7bb46a67 | 791 | */ |
c1632a0f | 792 | int simple_setattr(struct mnt_idmap *idmap, struct dentry *dentry, |
549c7297 | 793 | struct iattr *iattr) |
7bb46a67 | 794 | { |
dea655c2 | 795 | struct inode *inode = d_inode(dentry); |
7bb46a67 | 796 | int error; |
797 | ||
c1632a0f | 798 | error = setattr_prepare(idmap, dentry, iattr); |
7bb46a67 | 799 | if (error) |
800 | return error; | |
801 | ||
2c27c65e CH |
802 | if (iattr->ia_valid & ATTR_SIZE) |
803 | truncate_setsize(inode, iattr->ia_size); | |
c1632a0f | 804 | setattr_copy(idmap, inode, iattr); |
eef2380c CH |
805 | mark_inode_dirty(inode); |
806 | return 0; | |
7bb46a67 | 807 | } |
808 | EXPORT_SYMBOL(simple_setattr); | |
809 | ||
a77f580a | 810 | static int simple_read_folio(struct file *file, struct folio *folio) |
1da177e4 | 811 | { |
a77f580a MWO |
812 | folio_zero_range(folio, 0, folio_size(folio)); |
813 | flush_dcache_folio(folio); | |
814 | folio_mark_uptodate(folio); | |
815 | folio_unlock(folio); | |
1da177e4 LT |
816 | return 0; |
817 | } | |
818 | ||
afddba49 | 819 | int simple_write_begin(struct file *file, struct address_space *mapping, |
9d6b0cd7 | 820 | loff_t pos, unsigned len, |
afddba49 NP |
821 | struct page **pagep, void **fsdata) |
822 | { | |
5522d9f7 | 823 | struct folio *folio; |
afddba49 | 824 | |
5522d9f7 MWO |
825 | folio = __filemap_get_folio(mapping, pos / PAGE_SIZE, FGP_WRITEBEGIN, |
826 | mapping_gfp_mask(mapping)); | |
827 | if (IS_ERR(folio)) | |
828 | return PTR_ERR(folio); | |
afddba49 | 829 | |
5522d9f7 | 830 | *pagep = &folio->page; |
afddba49 | 831 | |
5522d9f7 MWO |
832 | if (!folio_test_uptodate(folio) && (len != folio_size(folio))) { |
833 | size_t from = offset_in_folio(folio, pos); | |
193cf4b9 | 834 | |
5522d9f7 MWO |
835 | folio_zero_segments(folio, 0, from, |
836 | from + len, folio_size(folio)); | |
193cf4b9 BH |
837 | } |
838 | return 0; | |
afddba49 | 839 | } |
12f38872 | 840 | EXPORT_SYMBOL(simple_write_begin); |
afddba49 | 841 | |
ad2a722f BH |
842 | /** |
843 | * simple_write_end - .write_end helper for non-block-device FSes | |
8e88bfba | 844 | * @file: See .write_end of address_space_operations |
ad2a722f BH |
845 | * @mapping: " |
846 | * @pos: " | |
847 | * @len: " | |
848 | * @copied: " | |
849 | * @page: " | |
850 | * @fsdata: " | |
851 | * | |
852 | * simple_write_end does the minimum needed for updating a page after writing is | |
853 | * done. It has the same API signature as the .write_end of | |
854 | * address_space_operations vector. So it can just be set onto .write_end for | |
855 | * FSes that don't need any other processing. i_mutex is assumed to be held. | |
856 | * Block based filesystems should use generic_write_end(). | |
857 | * NOTE: Even though i_size might get updated by this function, mark_inode_dirty | |
858 | * is not called, so a filesystem that actually does store data in .write_inode | |
859 | * should extend on what's done here with a call to mark_inode_dirty() in the | |
860 | * case that i_size has changed. | |
04fff641 | 861 | * |
a77f580a | 862 | * Use *ONLY* with simple_read_folio() |
ad2a722f | 863 | */ |
c1e3dbe9 | 864 | static int simple_write_end(struct file *file, struct address_space *mapping, |
afddba49 NP |
865 | loff_t pos, unsigned len, unsigned copied, |
866 | struct page *page, void *fsdata) | |
867 | { | |
5522d9f7 MWO |
868 | struct folio *folio = page_folio(page); |
869 | struct inode *inode = folio->mapping->host; | |
ad2a722f | 870 | loff_t last_pos = pos + copied; |
afddba49 | 871 | |
5522d9f7 MWO |
872 | /* zero the stale part of the folio if we did a short copy */ |
873 | if (!folio_test_uptodate(folio)) { | |
04fff641 | 874 | if (copied < len) { |
5522d9f7 | 875 | size_t from = offset_in_folio(folio, pos); |
afddba49 | 876 | |
5522d9f7 | 877 | folio_zero_range(folio, from + copied, len - copied); |
04fff641 | 878 | } |
5522d9f7 | 879 | folio_mark_uptodate(folio); |
04fff641 | 880 | } |
ad2a722f BH |
881 | /* |
882 | * No need to use i_size_read() here, the i_size | |
883 | * cannot change under us because we hold the i_mutex. | |
884 | */ | |
885 | if (last_pos > inode->i_size) | |
886 | i_size_write(inode, last_pos); | |
afddba49 | 887 | |
5522d9f7 MWO |
888 | folio_mark_dirty(folio); |
889 | folio_unlock(folio); | |
890 | folio_put(folio); | |
afddba49 NP |
891 | |
892 | return copied; | |
893 | } | |
c1e3dbe9 CH |
894 | |
895 | /* | |
896 | * Provides ramfs-style behavior: data in the pagecache, but no writeback. | |
897 | */ | |
898 | const struct address_space_operations ram_aops = { | |
a77f580a | 899 | .read_folio = simple_read_folio, |
c1e3dbe9 CH |
900 | .write_begin = simple_write_begin, |
901 | .write_end = simple_write_end, | |
46de8b97 | 902 | .dirty_folio = noop_dirty_folio, |
c1e3dbe9 CH |
903 | }; |
904 | EXPORT_SYMBOL(ram_aops); | |
afddba49 | 905 | |
1a1c9bb4 JL |
906 | /* |
907 | * the inodes created here are not hashed. If you use iunique to generate | |
908 | * unique inode values later for this filesystem, then you must take care | |
909 | * to pass it an appropriate max_reserved value to avoid collisions. | |
910 | */ | |
7d683a09 | 911 | int simple_fill_super(struct super_block *s, unsigned long magic, |
cda37124 | 912 | const struct tree_descr *files) |
1da177e4 | 913 | { |
1da177e4 LT |
914 | struct inode *inode; |
915 | struct dentry *root; | |
916 | struct dentry *dentry; | |
917 | int i; | |
918 | ||
09cbfeaf KS |
919 | s->s_blocksize = PAGE_SIZE; |
920 | s->s_blocksize_bits = PAGE_SHIFT; | |
1da177e4 | 921 | s->s_magic = magic; |
759b9775 | 922 | s->s_op = &simple_super_operations; |
1da177e4 LT |
923 | s->s_time_gran = 1; |
924 | ||
925 | inode = new_inode(s); | |
926 | if (!inode) | |
927 | return -ENOMEM; | |
1a1c9bb4 JL |
928 | /* |
929 | * because the root inode is 1, the files array must not contain an | |
930 | * entry at index 1 | |
931 | */ | |
932 | inode->i_ino = 1; | |
1da177e4 | 933 | inode->i_mode = S_IFDIR | 0755; |
077c212f | 934 | simple_inode_init_ts(inode); |
1da177e4 LT |
935 | inode->i_op = &simple_dir_inode_operations; |
936 | inode->i_fop = &simple_dir_operations; | |
bfe86848 | 937 | set_nlink(inode, 2); |
48fde701 AV |
938 | root = d_make_root(inode); |
939 | if (!root) | |
1da177e4 | 940 | return -ENOMEM; |
1da177e4 LT |
941 | for (i = 0; !files->name || files->name[0]; i++, files++) { |
942 | if (!files->name) | |
943 | continue; | |
1a1c9bb4 JL |
944 | |
945 | /* warn if it tries to conflict with the root inode */ | |
946 | if (unlikely(i == 1)) | |
947 | printk(KERN_WARNING "%s: %s passed in a files array" | |
948 | "with an index of 1!\n", __func__, | |
949 | s->s_type->name); | |
950 | ||
1da177e4 LT |
951 | dentry = d_alloc_name(root, files->name); |
952 | if (!dentry) | |
953 | goto out; | |
954 | inode = new_inode(s); | |
32096ea1 KK |
955 | if (!inode) { |
956 | dput(dentry); | |
1da177e4 | 957 | goto out; |
32096ea1 | 958 | } |
1da177e4 | 959 | inode->i_mode = S_IFREG | files->mode; |
077c212f | 960 | simple_inode_init_ts(inode); |
1da177e4 LT |
961 | inode->i_fop = files->ops; |
962 | inode->i_ino = i; | |
963 | d_add(dentry, inode); | |
964 | } | |
965 | s->s_root = root; | |
966 | return 0; | |
967 | out: | |
968 | d_genocide(root); | |
640946f2 | 969 | shrink_dcache_parent(root); |
1da177e4 LT |
970 | dput(root); |
971 | return -ENOMEM; | |
972 | } | |
12f38872 | 973 | EXPORT_SYMBOL(simple_fill_super); |
1da177e4 LT |
974 | |
975 | static DEFINE_SPINLOCK(pin_fs_lock); | |
976 | ||
1f5ce9e9 | 977 | int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count) |
1da177e4 LT |
978 | { |
979 | struct vfsmount *mnt = NULL; | |
980 | spin_lock(&pin_fs_lock); | |
981 | if (unlikely(!*mount)) { | |
982 | spin_unlock(&pin_fs_lock); | |
1751e8a6 | 983 | mnt = vfs_kern_mount(type, SB_KERNMOUNT, type->name, NULL); |
1da177e4 LT |
984 | if (IS_ERR(mnt)) |
985 | return PTR_ERR(mnt); | |
986 | spin_lock(&pin_fs_lock); | |
987 | if (!*mount) | |
988 | *mount = mnt; | |
989 | } | |
990 | mntget(*mount); | |
991 | ++*count; | |
992 | spin_unlock(&pin_fs_lock); | |
993 | mntput(mnt); | |
994 | return 0; | |
995 | } | |
12f38872 | 996 | EXPORT_SYMBOL(simple_pin_fs); |
1da177e4 LT |
997 | |
998 | void simple_release_fs(struct vfsmount **mount, int *count) | |
999 | { | |
1000 | struct vfsmount *mnt; | |
1001 | spin_lock(&pin_fs_lock); | |
1002 | mnt = *mount; | |
1003 | if (!--*count) | |
1004 | *mount = NULL; | |
1005 | spin_unlock(&pin_fs_lock); | |
1006 | mntput(mnt); | |
1007 | } | |
12f38872 | 1008 | EXPORT_SYMBOL(simple_release_fs); |
1da177e4 | 1009 | |
6d1029b5 AM |
1010 | /** |
1011 | * simple_read_from_buffer - copy data from the buffer to user space | |
1012 | * @to: the user space buffer to read to | |
1013 | * @count: the maximum number of bytes to read | |
1014 | * @ppos: the current position in the buffer | |
1015 | * @from: the buffer to read from | |
1016 | * @available: the size of the buffer | |
1017 | * | |
1018 | * The simple_read_from_buffer() function reads up to @count bytes from the | |
1019 | * buffer @from at offset @ppos into the user space address starting at @to. | |
1020 | * | |
1021 | * On success, the number of bytes read is returned and the offset @ppos is | |
1022 | * advanced by this number, or negative value is returned on error. | |
1023 | **/ | |
1da177e4 LT |
1024 | ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos, |
1025 | const void *from, size_t available) | |
1026 | { | |
1027 | loff_t pos = *ppos; | |
14be2746 SR |
1028 | size_t ret; |
1029 | ||
1da177e4 LT |
1030 | if (pos < 0) |
1031 | return -EINVAL; | |
14be2746 | 1032 | if (pos >= available || !count) |
1da177e4 LT |
1033 | return 0; |
1034 | if (count > available - pos) | |
1035 | count = available - pos; | |
14be2746 SR |
1036 | ret = copy_to_user(to, from + pos, count); |
1037 | if (ret == count) | |
1da177e4 | 1038 | return -EFAULT; |
14be2746 | 1039 | count -= ret; |
1da177e4 LT |
1040 | *ppos = pos + count; |
1041 | return count; | |
1042 | } | |
12f38872 | 1043 | EXPORT_SYMBOL(simple_read_from_buffer); |
1da177e4 | 1044 | |
6a727b43 JS |
1045 | /** |
1046 | * simple_write_to_buffer - copy data from user space to the buffer | |
1047 | * @to: the buffer to write to | |
1048 | * @available: the size of the buffer | |
1049 | * @ppos: the current position in the buffer | |
1050 | * @from: the user space buffer to read from | |
1051 | * @count: the maximum number of bytes to read | |
1052 | * | |
1053 | * The simple_write_to_buffer() function reads up to @count bytes from the user | |
1054 | * space address starting at @from into the buffer @to at offset @ppos. | |
1055 | * | |
1056 | * On success, the number of bytes written is returned and the offset @ppos is | |
1057 | * advanced by this number, or negative value is returned on error. | |
1058 | **/ | |
1059 | ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos, | |
1060 | const void __user *from, size_t count) | |
1061 | { | |
1062 | loff_t pos = *ppos; | |
1063 | size_t res; | |
1064 | ||
1065 | if (pos < 0) | |
1066 | return -EINVAL; | |
1067 | if (pos >= available || !count) | |
1068 | return 0; | |
1069 | if (count > available - pos) | |
1070 | count = available - pos; | |
1071 | res = copy_from_user(to + pos, from, count); | |
1072 | if (res == count) | |
1073 | return -EFAULT; | |
1074 | count -= res; | |
1075 | *ppos = pos + count; | |
1076 | return count; | |
1077 | } | |
12f38872 | 1078 | EXPORT_SYMBOL(simple_write_to_buffer); |
6a727b43 | 1079 | |
6d1029b5 AM |
1080 | /** |
1081 | * memory_read_from_buffer - copy data from the buffer | |
1082 | * @to: the kernel space buffer to read to | |
1083 | * @count: the maximum number of bytes to read | |
1084 | * @ppos: the current position in the buffer | |
1085 | * @from: the buffer to read from | |
1086 | * @available: the size of the buffer | |
1087 | * | |
1088 | * The memory_read_from_buffer() function reads up to @count bytes from the | |
1089 | * buffer @from at offset @ppos into the kernel space address starting at @to. | |
1090 | * | |
1091 | * On success, the number of bytes read is returned and the offset @ppos is | |
1092 | * advanced by this number, or negative value is returned on error. | |
1093 | **/ | |
93b07113 AM |
1094 | ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos, |
1095 | const void *from, size_t available) | |
1096 | { | |
1097 | loff_t pos = *ppos; | |
1098 | ||
1099 | if (pos < 0) | |
1100 | return -EINVAL; | |
1101 | if (pos >= available) | |
1102 | return 0; | |
1103 | if (count > available - pos) | |
1104 | count = available - pos; | |
1105 | memcpy(to, from + pos, count); | |
1106 | *ppos = pos + count; | |
1107 | ||
1108 | return count; | |
1109 | } | |
12f38872 | 1110 | EXPORT_SYMBOL(memory_read_from_buffer); |
93b07113 | 1111 | |
1da177e4 LT |
1112 | /* |
1113 | * Transaction based IO. | |
1114 | * The file expects a single write which triggers the transaction, and then | |
1115 | * possibly a read which collects the result - which is stored in a | |
1116 | * file-local buffer. | |
1117 | */ | |
76791ab2 IM |
1118 | |
1119 | void simple_transaction_set(struct file *file, size_t n) | |
1120 | { | |
1121 | struct simple_transaction_argresp *ar = file->private_data; | |
1122 | ||
1123 | BUG_ON(n > SIMPLE_TRANSACTION_LIMIT); | |
1124 | ||
1125 | /* | |
1126 | * The barrier ensures that ar->size will really remain zero until | |
1127 | * ar->data is ready for reading. | |
1128 | */ | |
1129 | smp_mb(); | |
1130 | ar->size = n; | |
1131 | } | |
12f38872 | 1132 | EXPORT_SYMBOL(simple_transaction_set); |
76791ab2 | 1133 | |
1da177e4 LT |
1134 | char *simple_transaction_get(struct file *file, const char __user *buf, size_t size) |
1135 | { | |
1136 | struct simple_transaction_argresp *ar; | |
1137 | static DEFINE_SPINLOCK(simple_transaction_lock); | |
1138 | ||
1139 | if (size > SIMPLE_TRANSACTION_LIMIT - 1) | |
1140 | return ERR_PTR(-EFBIG); | |
1141 | ||
1142 | ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL); | |
1143 | if (!ar) | |
1144 | return ERR_PTR(-ENOMEM); | |
1145 | ||
1146 | spin_lock(&simple_transaction_lock); | |
1147 | ||
1148 | /* only one write allowed per open */ | |
1149 | if (file->private_data) { | |
1150 | spin_unlock(&simple_transaction_lock); | |
1151 | free_page((unsigned long)ar); | |
1152 | return ERR_PTR(-EBUSY); | |
1153 | } | |
1154 | ||
1155 | file->private_data = ar; | |
1156 | ||
1157 | spin_unlock(&simple_transaction_lock); | |
1158 | ||
1159 | if (copy_from_user(ar->data, buf, size)) | |
1160 | return ERR_PTR(-EFAULT); | |
1161 | ||
1162 | return ar->data; | |
1163 | } | |
12f38872 | 1164 | EXPORT_SYMBOL(simple_transaction_get); |
1da177e4 LT |
1165 | |
1166 | ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos) | |
1167 | { | |
1168 | struct simple_transaction_argresp *ar = file->private_data; | |
1169 | ||
1170 | if (!ar) | |
1171 | return 0; | |
1172 | return simple_read_from_buffer(buf, size, pos, ar->data, ar->size); | |
1173 | } | |
12f38872 | 1174 | EXPORT_SYMBOL(simple_transaction_read); |
1da177e4 LT |
1175 | |
1176 | int simple_transaction_release(struct inode *inode, struct file *file) | |
1177 | { | |
1178 | free_page((unsigned long)file->private_data); | |
1179 | return 0; | |
1180 | } | |
12f38872 | 1181 | EXPORT_SYMBOL(simple_transaction_release); |
1da177e4 | 1182 | |
acaefc25 AB |
1183 | /* Simple attribute files */ |
1184 | ||
1185 | struct simple_attr { | |
8b88b099 CH |
1186 | int (*get)(void *, u64 *); |
1187 | int (*set)(void *, u64); | |
acaefc25 AB |
1188 | char get_buf[24]; /* enough to store a u64 and "\n\0" */ |
1189 | char set_buf[24]; | |
1190 | void *data; | |
1191 | const char *fmt; /* format for read operation */ | |
7cf34c76 | 1192 | struct mutex mutex; /* protects access to these buffers */ |
acaefc25 AB |
1193 | }; |
1194 | ||
1195 | /* simple_attr_open is called by an actual attribute open file operation | |
1196 | * to set the attribute specific access operations. */ | |
1197 | int simple_attr_open(struct inode *inode, struct file *file, | |
8b88b099 | 1198 | int (*get)(void *, u64 *), int (*set)(void *, u64), |
acaefc25 AB |
1199 | const char *fmt) |
1200 | { | |
1201 | struct simple_attr *attr; | |
1202 | ||
a65cab7d | 1203 | attr = kzalloc(sizeof(*attr), GFP_KERNEL); |
acaefc25 AB |
1204 | if (!attr) |
1205 | return -ENOMEM; | |
1206 | ||
1207 | attr->get = get; | |
1208 | attr->set = set; | |
8e18e294 | 1209 | attr->data = inode->i_private; |
acaefc25 | 1210 | attr->fmt = fmt; |
7cf34c76 | 1211 | mutex_init(&attr->mutex); |
acaefc25 AB |
1212 | |
1213 | file->private_data = attr; | |
1214 | ||
1215 | return nonseekable_open(inode, file); | |
1216 | } | |
12f38872 | 1217 | EXPORT_SYMBOL_GPL(simple_attr_open); |
acaefc25 | 1218 | |
74bedc4d | 1219 | int simple_attr_release(struct inode *inode, struct file *file) |
acaefc25 AB |
1220 | { |
1221 | kfree(file->private_data); | |
1222 | return 0; | |
1223 | } | |
12f38872 | 1224 | EXPORT_SYMBOL_GPL(simple_attr_release); /* GPL-only? This? Really? */ |
acaefc25 AB |
1225 | |
1226 | /* read from the buffer that is filled with the get function */ | |
1227 | ssize_t simple_attr_read(struct file *file, char __user *buf, | |
1228 | size_t len, loff_t *ppos) | |
1229 | { | |
1230 | struct simple_attr *attr; | |
1231 | size_t size; | |
1232 | ssize_t ret; | |
1233 | ||
1234 | attr = file->private_data; | |
1235 | ||
1236 | if (!attr->get) | |
1237 | return -EACCES; | |
1238 | ||
9261303a CH |
1239 | ret = mutex_lock_interruptible(&attr->mutex); |
1240 | if (ret) | |
1241 | return ret; | |
1242 | ||
a65cab7d EB |
1243 | if (*ppos && attr->get_buf[0]) { |
1244 | /* continued read */ | |
acaefc25 | 1245 | size = strlen(attr->get_buf); |
a65cab7d EB |
1246 | } else { |
1247 | /* first read */ | |
8b88b099 CH |
1248 | u64 val; |
1249 | ret = attr->get(attr->data, &val); | |
1250 | if (ret) | |
1251 | goto out; | |
1252 | ||
acaefc25 | 1253 | size = scnprintf(attr->get_buf, sizeof(attr->get_buf), |
8b88b099 CH |
1254 | attr->fmt, (unsigned long long)val); |
1255 | } | |
acaefc25 AB |
1256 | |
1257 | ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size); | |
8b88b099 | 1258 | out: |
7cf34c76 | 1259 | mutex_unlock(&attr->mutex); |
acaefc25 AB |
1260 | return ret; |
1261 | } | |
12f38872 | 1262 | EXPORT_SYMBOL_GPL(simple_attr_read); |
acaefc25 AB |
1263 | |
1264 | /* interpret the buffer as a number to call the set function with */ | |
2e41f274 AM |
1265 | static ssize_t simple_attr_write_xsigned(struct file *file, const char __user *buf, |
1266 | size_t len, loff_t *ppos, bool is_signed) | |
acaefc25 AB |
1267 | { |
1268 | struct simple_attr *attr; | |
488dac0c | 1269 | unsigned long long val; |
acaefc25 AB |
1270 | size_t size; |
1271 | ssize_t ret; | |
1272 | ||
1273 | attr = file->private_data; | |
acaefc25 AB |
1274 | if (!attr->set) |
1275 | return -EACCES; | |
1276 | ||
9261303a CH |
1277 | ret = mutex_lock_interruptible(&attr->mutex); |
1278 | if (ret) | |
1279 | return ret; | |
1280 | ||
acaefc25 AB |
1281 | ret = -EFAULT; |
1282 | size = min(sizeof(attr->set_buf) - 1, len); | |
1283 | if (copy_from_user(attr->set_buf, buf, size)) | |
1284 | goto out; | |
1285 | ||
acaefc25 | 1286 | attr->set_buf[size] = '\0'; |
2e41f274 AM |
1287 | if (is_signed) |
1288 | ret = kstrtoll(attr->set_buf, 0, &val); | |
1289 | else | |
1290 | ret = kstrtoull(attr->set_buf, 0, &val); | |
488dac0c YY |
1291 | if (ret) |
1292 | goto out; | |
05cc0cee WF |
1293 | ret = attr->set(attr->data, val); |
1294 | if (ret == 0) | |
1295 | ret = len; /* on success, claim we got the whole input */ | |
acaefc25 | 1296 | out: |
7cf34c76 | 1297 | mutex_unlock(&attr->mutex); |
acaefc25 AB |
1298 | return ret; |
1299 | } | |
2e41f274 AM |
1300 | |
1301 | ssize_t simple_attr_write(struct file *file, const char __user *buf, | |
1302 | size_t len, loff_t *ppos) | |
1303 | { | |
1304 | return simple_attr_write_xsigned(file, buf, len, ppos, false); | |
1305 | } | |
12f38872 | 1306 | EXPORT_SYMBOL_GPL(simple_attr_write); |
acaefc25 | 1307 | |
2e41f274 AM |
1308 | ssize_t simple_attr_write_signed(struct file *file, const char __user *buf, |
1309 | size_t len, loff_t *ppos) | |
1310 | { | |
1311 | return simple_attr_write_xsigned(file, buf, len, ppos, true); | |
1312 | } | |
1313 | EXPORT_SYMBOL_GPL(simple_attr_write_signed); | |
1314 | ||
d9e5d922 AG |
1315 | /** |
1316 | * generic_encode_ino32_fh - generic export_operations->encode_fh function | |
1317 | * @inode: the object to encode | |
1318 | * @fh: where to store the file handle fragment | |
1319 | * @max_len: maximum length to store there (in 4 byte units) | |
1320 | * @parent: parent directory inode, if wanted | |
1321 | * | |
1322 | * This generic encode_fh function assumes that the 32 inode number | |
1323 | * is suitable for locating an inode, and that the generation number | |
1324 | * can be used to check that it is still valid. It places them in the | |
1325 | * filehandle fragment where export_decode_fh expects to find them. | |
1326 | */ | |
1327 | int generic_encode_ino32_fh(struct inode *inode, __u32 *fh, int *max_len, | |
1328 | struct inode *parent) | |
1329 | { | |
1330 | struct fid *fid = (void *)fh; | |
1331 | int len = *max_len; | |
1332 | int type = FILEID_INO32_GEN; | |
1333 | ||
1334 | if (parent && (len < 4)) { | |
1335 | *max_len = 4; | |
1336 | return FILEID_INVALID; | |
1337 | } else if (len < 2) { | |
1338 | *max_len = 2; | |
1339 | return FILEID_INVALID; | |
1340 | } | |
1341 | ||
1342 | len = 2; | |
1343 | fid->i32.ino = inode->i_ino; | |
1344 | fid->i32.gen = inode->i_generation; | |
1345 | if (parent) { | |
1346 | fid->i32.parent_ino = parent->i_ino; | |
1347 | fid->i32.parent_gen = parent->i_generation; | |
1348 | len = 4; | |
1349 | type = FILEID_INO32_GEN_PARENT; | |
1350 | } | |
1351 | *max_len = len; | |
1352 | return type; | |
1353 | } | |
1354 | EXPORT_SYMBOL_GPL(generic_encode_ino32_fh); | |
1355 | ||
2596110a CH |
1356 | /** |
1357 | * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation | |
1358 | * @sb: filesystem to do the file handle conversion on | |
1359 | * @fid: file handle to convert | |
1360 | * @fh_len: length of the file handle in bytes | |
1361 | * @fh_type: type of file handle | |
1362 | * @get_inode: filesystem callback to retrieve inode | |
1363 | * | |
1364 | * This function decodes @fid as long as it has one of the well-known | |
1365 | * Linux filehandle types and calls @get_inode on it to retrieve the | |
1366 | * inode for the object specified in the file handle. | |
1367 | */ | |
1368 | struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid, | |
1369 | int fh_len, int fh_type, struct inode *(*get_inode) | |
1370 | (struct super_block *sb, u64 ino, u32 gen)) | |
1371 | { | |
1372 | struct inode *inode = NULL; | |
1373 | ||
1374 | if (fh_len < 2) | |
1375 | return NULL; | |
1376 | ||
1377 | switch (fh_type) { | |
1378 | case FILEID_INO32_GEN: | |
1379 | case FILEID_INO32_GEN_PARENT: | |
1380 | inode = get_inode(sb, fid->i32.ino, fid->i32.gen); | |
1381 | break; | |
1382 | } | |
1383 | ||
4ea3ada2 | 1384 | return d_obtain_alias(inode); |
2596110a CH |
1385 | } |
1386 | EXPORT_SYMBOL_GPL(generic_fh_to_dentry); | |
1387 | ||
1388 | /** | |
ca186830 | 1389 | * generic_fh_to_parent - generic helper for the fh_to_parent export operation |
2596110a CH |
1390 | * @sb: filesystem to do the file handle conversion on |
1391 | * @fid: file handle to convert | |
1392 | * @fh_len: length of the file handle in bytes | |
1393 | * @fh_type: type of file handle | |
1394 | * @get_inode: filesystem callback to retrieve inode | |
1395 | * | |
1396 | * This function decodes @fid as long as it has one of the well-known | |
1397 | * Linux filehandle types and calls @get_inode on it to retrieve the | |
1398 | * inode for the _parent_ object specified in the file handle if it | |
1399 | * is specified in the file handle, or NULL otherwise. | |
1400 | */ | |
1401 | struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid, | |
1402 | int fh_len, int fh_type, struct inode *(*get_inode) | |
1403 | (struct super_block *sb, u64 ino, u32 gen)) | |
1404 | { | |
1405 | struct inode *inode = NULL; | |
1406 | ||
1407 | if (fh_len <= 2) | |
1408 | return NULL; | |
1409 | ||
1410 | switch (fh_type) { | |
1411 | case FILEID_INO32_GEN_PARENT: | |
1412 | inode = get_inode(sb, fid->i32.parent_ino, | |
1413 | (fh_len > 3 ? fid->i32.parent_gen : 0)); | |
1414 | break; | |
1415 | } | |
1416 | ||
4ea3ada2 | 1417 | return d_obtain_alias(inode); |
2596110a CH |
1418 | } |
1419 | EXPORT_SYMBOL_GPL(generic_fh_to_parent); | |
1420 | ||
1b061d92 | 1421 | /** |
ac13a829 FF |
1422 | * __generic_file_fsync - generic fsync implementation for simple filesystems |
1423 | * | |
1b061d92 | 1424 | * @file: file to synchronize |
ac13a829 FF |
1425 | * @start: start offset in bytes |
1426 | * @end: end offset in bytes (inclusive) | |
1b061d92 CH |
1427 | * @datasync: only synchronize essential metadata if true |
1428 | * | |
1429 | * This is a generic implementation of the fsync method for simple | |
1430 | * filesystems which track all non-inode metadata in the buffers list | |
1431 | * hanging off the address_space structure. | |
1432 | */ | |
ac13a829 FF |
1433 | int __generic_file_fsync(struct file *file, loff_t start, loff_t end, |
1434 | int datasync) | |
d5aacad5 | 1435 | { |
7ea80859 | 1436 | struct inode *inode = file->f_mapping->host; |
d5aacad5 AV |
1437 | int err; |
1438 | int ret; | |
1439 | ||
383aa543 | 1440 | err = file_write_and_wait_range(file, start, end); |
02c24a82 JB |
1441 | if (err) |
1442 | return err; | |
1443 | ||
5955102c | 1444 | inode_lock(inode); |
d5aacad5 | 1445 | ret = sync_mapping_buffers(inode->i_mapping); |
0ae45f63 | 1446 | if (!(inode->i_state & I_DIRTY_ALL)) |
02c24a82 | 1447 | goto out; |
d5aacad5 | 1448 | if (datasync && !(inode->i_state & I_DIRTY_DATASYNC)) |
02c24a82 | 1449 | goto out; |
d5aacad5 | 1450 | |
c3765016 | 1451 | err = sync_inode_metadata(inode, 1); |
d5aacad5 AV |
1452 | if (ret == 0) |
1453 | ret = err; | |
ac13a829 | 1454 | |
02c24a82 | 1455 | out: |
5955102c | 1456 | inode_unlock(inode); |
383aa543 JL |
1457 | /* check and advance again to catch errors after syncing out buffers */ |
1458 | err = file_check_and_advance_wb_err(file); | |
1459 | if (ret == 0) | |
1460 | ret = err; | |
1461 | return ret; | |
d5aacad5 | 1462 | } |
ac13a829 FF |
1463 | EXPORT_SYMBOL(__generic_file_fsync); |
1464 | ||
1465 | /** | |
1466 | * generic_file_fsync - generic fsync implementation for simple filesystems | |
1467 | * with flush | |
1468 | * @file: file to synchronize | |
1469 | * @start: start offset in bytes | |
1470 | * @end: end offset in bytes (inclusive) | |
1471 | * @datasync: only synchronize essential metadata if true | |
1472 | * | |
1473 | */ | |
1474 | ||
1475 | int generic_file_fsync(struct file *file, loff_t start, loff_t end, | |
1476 | int datasync) | |
1477 | { | |
1478 | struct inode *inode = file->f_mapping->host; | |
1479 | int err; | |
1480 | ||
1481 | err = __generic_file_fsync(file, start, end, datasync); | |
1482 | if (err) | |
1483 | return err; | |
c6bf3f0e | 1484 | return blkdev_issue_flush(inode->i_sb->s_bdev); |
ac13a829 | 1485 | } |
1b061d92 CH |
1486 | EXPORT_SYMBOL(generic_file_fsync); |
1487 | ||
30ca22c7 PL |
1488 | /** |
1489 | * generic_check_addressable - Check addressability of file system | |
1490 | * @blocksize_bits: log of file system block size | |
1491 | * @num_blocks: number of blocks in file system | |
1492 | * | |
1493 | * Determine whether a file system with @num_blocks blocks (and a | |
1494 | * block size of 2**@blocksize_bits) is addressable by the sector_t | |
1495 | * and page cache of the system. Return 0 if so and -EFBIG otherwise. | |
1496 | */ | |
1497 | int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks) | |
1498 | { | |
1499 | u64 last_fs_block = num_blocks - 1; | |
a33f13ef | 1500 | u64 last_fs_page = |
09cbfeaf | 1501 | last_fs_block >> (PAGE_SHIFT - blocksize_bits); |
30ca22c7 PL |
1502 | |
1503 | if (unlikely(num_blocks == 0)) | |
1504 | return 0; | |
1505 | ||
09cbfeaf | 1506 | if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT)) |
30ca22c7 PL |
1507 | return -EINVAL; |
1508 | ||
a33f13ef JB |
1509 | if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) || |
1510 | (last_fs_page > (pgoff_t)(~0ULL))) { | |
30ca22c7 PL |
1511 | return -EFBIG; |
1512 | } | |
1513 | return 0; | |
1514 | } | |
1515 | EXPORT_SYMBOL(generic_check_addressable); | |
1516 | ||
1b061d92 CH |
1517 | /* |
1518 | * No-op implementation of ->fsync for in-memory filesystems. | |
1519 | */ | |
02c24a82 | 1520 | int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync) |
1b061d92 CH |
1521 | { |
1522 | return 0; | |
1523 | } | |
1b061d92 | 1524 | EXPORT_SYMBOL(noop_fsync); |
87dc800b | 1525 | |
f44c7763 DW |
1526 | ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter) |
1527 | { | |
1528 | /* | |
1529 | * iomap based filesystems support direct I/O without need for | |
1530 | * this callback. However, it still needs to be set in | |
1531 | * inode->a_ops so that open/fcntl know that direct I/O is | |
1532 | * generally supported. | |
1533 | */ | |
1534 | return -EINVAL; | |
1535 | } | |
1536 | EXPORT_SYMBOL_GPL(noop_direct_IO); | |
1537 | ||
fceef393 AV |
1538 | /* Because kfree isn't assignment-compatible with void(void*) ;-/ */ |
1539 | void kfree_link(void *p) | |
87dc800b | 1540 | { |
fceef393 | 1541 | kfree(p); |
87dc800b | 1542 | } |
fceef393 | 1543 | EXPORT_SYMBOL(kfree_link); |
6987843f | 1544 | |
6987843f AV |
1545 | struct inode *alloc_anon_inode(struct super_block *s) |
1546 | { | |
1547 | static const struct address_space_operations anon_aops = { | |
46de8b97 | 1548 | .dirty_folio = noop_dirty_folio, |
6987843f AV |
1549 | }; |
1550 | struct inode *inode = new_inode_pseudo(s); | |
1551 | ||
1552 | if (!inode) | |
1553 | return ERR_PTR(-ENOMEM); | |
1554 | ||
1555 | inode->i_ino = get_next_ino(); | |
1556 | inode->i_mapping->a_ops = &anon_aops; | |
1557 | ||
1558 | /* | |
1559 | * Mark the inode dirty from the very beginning, | |
1560 | * that way it will never be moved to the dirty | |
1561 | * list because mark_inode_dirty() will think | |
1562 | * that it already _is_ on the dirty list. | |
1563 | */ | |
1564 | inode->i_state = I_DIRTY; | |
1565 | inode->i_mode = S_IRUSR | S_IWUSR; | |
1566 | inode->i_uid = current_fsuid(); | |
1567 | inode->i_gid = current_fsgid(); | |
1568 | inode->i_flags |= S_PRIVATE; | |
077c212f | 1569 | simple_inode_init_ts(inode); |
6987843f AV |
1570 | return inode; |
1571 | } | |
1572 | EXPORT_SYMBOL(alloc_anon_inode); | |
1c994a09 JL |
1573 | |
1574 | /** | |
1575 | * simple_nosetlease - generic helper for prohibiting leases | |
1576 | * @filp: file pointer | |
1577 | * @arg: type of lease to obtain | |
1578 | * @flp: new lease supplied for insertion | |
e6f5c789 | 1579 | * @priv: private data for lm_setup operation |
1c994a09 JL |
1580 | * |
1581 | * Generic helper for filesystems that do not wish to allow leases to be set. | |
1582 | * All arguments are ignored and it just returns -EINVAL. | |
1583 | */ | |
1584 | int | |
ed5f17f6 | 1585 | simple_nosetlease(struct file *filp, int arg, struct file_lock **flp, |
e6f5c789 | 1586 | void **priv) |
1c994a09 JL |
1587 | { |
1588 | return -EINVAL; | |
1589 | } | |
1590 | EXPORT_SYMBOL(simple_nosetlease); | |
61ba64fc | 1591 | |
6ee9706a EB |
1592 | /** |
1593 | * simple_get_link - generic helper to get the target of "fast" symlinks | |
1594 | * @dentry: not used here | |
1595 | * @inode: the symlink inode | |
1596 | * @done: not used here | |
1597 | * | |
1598 | * Generic helper for filesystems to use for symlink inodes where a pointer to | |
1599 | * the symlink target is stored in ->i_link. NOTE: this isn't normally called, | |
1600 | * since as an optimization the path lookup code uses any non-NULL ->i_link | |
1601 | * directly, without calling ->get_link(). But ->get_link() still must be set, | |
1602 | * to mark the inode_operations as being for a symlink. | |
1603 | * | |
1604 | * Return: the symlink target | |
1605 | */ | |
6b255391 | 1606 | const char *simple_get_link(struct dentry *dentry, struct inode *inode, |
fceef393 | 1607 | struct delayed_call *done) |
61ba64fc | 1608 | { |
6b255391 | 1609 | return inode->i_link; |
61ba64fc | 1610 | } |
6b255391 | 1611 | EXPORT_SYMBOL(simple_get_link); |
61ba64fc AV |
1612 | |
1613 | const struct inode_operations simple_symlink_inode_operations = { | |
6b255391 | 1614 | .get_link = simple_get_link, |
61ba64fc AV |
1615 | }; |
1616 | EXPORT_SYMBOL(simple_symlink_inode_operations); | |
fbabfd0f EB |
1617 | |
1618 | /* | |
1619 | * Operations for a permanently empty directory. | |
1620 | */ | |
1621 | static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) | |
1622 | { | |
1623 | return ERR_PTR(-ENOENT); | |
1624 | } | |
1625 | ||
b74d24f7 | 1626 | static int empty_dir_getattr(struct mnt_idmap *idmap, |
549c7297 | 1627 | const struct path *path, struct kstat *stat, |
a528d35e | 1628 | u32 request_mask, unsigned int query_flags) |
fbabfd0f | 1629 | { |
a528d35e | 1630 | struct inode *inode = d_inode(path->dentry); |
0d72b928 | 1631 | generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat); |
fbabfd0f EB |
1632 | return 0; |
1633 | } | |
1634 | ||
c1632a0f | 1635 | static int empty_dir_setattr(struct mnt_idmap *idmap, |
549c7297 | 1636 | struct dentry *dentry, struct iattr *attr) |
fbabfd0f EB |
1637 | { |
1638 | return -EPERM; | |
1639 | } | |
1640 | ||
fbabfd0f EB |
1641 | static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size) |
1642 | { | |
1643 | return -EOPNOTSUPP; | |
1644 | } | |
1645 | ||
1646 | static const struct inode_operations empty_dir_inode_operations = { | |
1647 | .lookup = empty_dir_lookup, | |
1648 | .permission = generic_permission, | |
1649 | .setattr = empty_dir_setattr, | |
1650 | .getattr = empty_dir_getattr, | |
fbabfd0f EB |
1651 | .listxattr = empty_dir_listxattr, |
1652 | }; | |
1653 | ||
1654 | static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence) | |
1655 | { | |
1656 | /* An empty directory has two entries . and .. at offsets 0 and 1 */ | |
1657 | return generic_file_llseek_size(file, offset, whence, 2, 2); | |
1658 | } | |
1659 | ||
1660 | static int empty_dir_readdir(struct file *file, struct dir_context *ctx) | |
1661 | { | |
1662 | dir_emit_dots(file, ctx); | |
1663 | return 0; | |
1664 | } | |
1665 | ||
1666 | static const struct file_operations empty_dir_operations = { | |
1667 | .llseek = empty_dir_llseek, | |
1668 | .read = generic_read_dir, | |
c51da20c | 1669 | .iterate_shared = empty_dir_readdir, |
fbabfd0f EB |
1670 | .fsync = noop_fsync, |
1671 | }; | |
1672 | ||
1673 | ||
1674 | void make_empty_dir_inode(struct inode *inode) | |
1675 | { | |
1676 | set_nlink(inode, 2); | |
1677 | inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO; | |
1678 | inode->i_uid = GLOBAL_ROOT_UID; | |
1679 | inode->i_gid = GLOBAL_ROOT_GID; | |
1680 | inode->i_rdev = 0; | |
4b75de86 | 1681 | inode->i_size = 0; |
fbabfd0f EB |
1682 | inode->i_blkbits = PAGE_SHIFT; |
1683 | inode->i_blocks = 0; | |
1684 | ||
1685 | inode->i_op = &empty_dir_inode_operations; | |
f5c24438 | 1686 | inode->i_opflags &= ~IOP_XATTR; |
fbabfd0f EB |
1687 | inode->i_fop = &empty_dir_operations; |
1688 | } | |
1689 | ||
1690 | bool is_empty_dir_inode(struct inode *inode) | |
1691 | { | |
1692 | return (inode->i_fop == &empty_dir_operations) && | |
1693 | (inode->i_op == &empty_dir_inode_operations); | |
1694 | } | |
c843843e | 1695 | |
5298d4bf | 1696 | #if IS_ENABLED(CONFIG_UNICODE) |
c843843e DR |
1697 | /** |
1698 | * generic_ci_d_compare - generic d_compare implementation for casefolding filesystems | |
1699 | * @dentry: dentry whose name we are checking against | |
1700 | * @len: len of name of dentry | |
1701 | * @str: str pointer to name of dentry | |
1702 | * @name: Name to compare against | |
1703 | * | |
1704 | * Return: 0 if names match, 1 if mismatch, or -ERRNO | |
1705 | */ | |
794c43f7 EB |
1706 | static int generic_ci_d_compare(const struct dentry *dentry, unsigned int len, |
1707 | const char *str, const struct qstr *name) | |
c843843e DR |
1708 | { |
1709 | const struct dentry *parent = READ_ONCE(dentry->d_parent); | |
1710 | const struct inode *dir = READ_ONCE(parent->d_inode); | |
1711 | const struct super_block *sb = dentry->d_sb; | |
1712 | const struct unicode_map *um = sb->s_encoding; | |
1713 | struct qstr qstr = QSTR_INIT(str, len); | |
1714 | char strbuf[DNAME_INLINE_LEN]; | |
1715 | int ret; | |
1716 | ||
af494af3 | 1717 | if (!dir || !IS_CASEFOLDED(dir)) |
c843843e DR |
1718 | goto fallback; |
1719 | /* | |
1720 | * If the dentry name is stored in-line, then it may be concurrently | |
1721 | * modified by a rename. If this happens, the VFS will eventually retry | |
1722 | * the lookup, so it doesn't matter what ->d_compare() returns. | |
1723 | * However, it's unsafe to call utf8_strncasecmp() with an unstable | |
1724 | * string. Therefore, we have to copy the name into a temporary buffer. | |
1725 | */ | |
1726 | if (len <= DNAME_INLINE_LEN - 1) { | |
1727 | memcpy(strbuf, str, len); | |
1728 | strbuf[len] = 0; | |
1729 | qstr.name = strbuf; | |
1730 | /* prevent compiler from optimizing out the temporary buffer */ | |
1731 | barrier(); | |
1732 | } | |
1733 | ret = utf8_strncasecmp(um, name, &qstr); | |
1734 | if (ret >= 0) | |
1735 | return ret; | |
1736 | ||
1737 | if (sb_has_strict_encoding(sb)) | |
1738 | return -EINVAL; | |
1739 | fallback: | |
1740 | if (len != name->len) | |
1741 | return 1; | |
1742 | return !!memcmp(str, name->name, len); | |
1743 | } | |
c843843e DR |
1744 | |
1745 | /** | |
1746 | * generic_ci_d_hash - generic d_hash implementation for casefolding filesystems | |
1747 | * @dentry: dentry of the parent directory | |
1748 | * @str: qstr of name whose hash we should fill in | |
1749 | * | |
1750 | * Return: 0 if hash was successful or unchanged, and -EINVAL on error | |
1751 | */ | |
794c43f7 | 1752 | static int generic_ci_d_hash(const struct dentry *dentry, struct qstr *str) |
c843843e DR |
1753 | { |
1754 | const struct inode *dir = READ_ONCE(dentry->d_inode); | |
1755 | struct super_block *sb = dentry->d_sb; | |
1756 | const struct unicode_map *um = sb->s_encoding; | |
1757 | int ret = 0; | |
1758 | ||
af494af3 | 1759 | if (!dir || !IS_CASEFOLDED(dir)) |
c843843e DR |
1760 | return 0; |
1761 | ||
1762 | ret = utf8_casefold_hash(um, dentry, str); | |
1763 | if (ret < 0 && sb_has_strict_encoding(sb)) | |
1764 | return -EINVAL; | |
1765 | return 0; | |
1766 | } | |
608af703 DR |
1767 | |
1768 | static const struct dentry_operations generic_ci_dentry_ops = { | |
1769 | .d_hash = generic_ci_d_hash, | |
1770 | .d_compare = generic_ci_d_compare, | |
1771 | }; | |
1772 | #endif | |
1773 | ||
1774 | #ifdef CONFIG_FS_ENCRYPTION | |
1775 | static const struct dentry_operations generic_encrypted_dentry_ops = { | |
1776 | .d_revalidate = fscrypt_d_revalidate, | |
1777 | }; | |
1778 | #endif | |
1779 | ||
5298d4bf | 1780 | #if defined(CONFIG_FS_ENCRYPTION) && IS_ENABLED(CONFIG_UNICODE) |
608af703 DR |
1781 | static const struct dentry_operations generic_encrypted_ci_dentry_ops = { |
1782 | .d_hash = generic_ci_d_hash, | |
1783 | .d_compare = generic_ci_d_compare, | |
1784 | .d_revalidate = fscrypt_d_revalidate, | |
1785 | }; | |
1786 | #endif | |
1787 | ||
1788 | /** | |
1789 | * generic_set_encrypted_ci_d_ops - helper for setting d_ops for given dentry | |
1790 | * @dentry: dentry to set ops on | |
1791 | * | |
1792 | * Casefolded directories need d_hash and d_compare set, so that the dentries | |
1793 | * contained in them are handled case-insensitively. Note that these operations | |
1794 | * are needed on the parent directory rather than on the dentries in it, and | |
1795 | * while the casefolding flag can be toggled on and off on an empty directory, | |
1796 | * dentry_operations can't be changed later. As a result, if the filesystem has | |
1797 | * casefolding support enabled at all, we have to give all dentries the | |
1798 | * casefolding operations even if their inode doesn't have the casefolding flag | |
1799 | * currently (and thus the casefolding ops would be no-ops for now). | |
1800 | * | |
1801 | * Encryption works differently in that the only dentry operation it needs is | |
1802 | * d_revalidate, which it only needs on dentries that have the no-key name flag. | |
1803 | * The no-key flag can't be set "later", so we don't have to worry about that. | |
1804 | * | |
1805 | * Finally, to maximize compatibility with overlayfs (which isn't compatible | |
1806 | * with certain dentry operations) and to avoid taking an unnecessary | |
1807 | * performance hit, we use custom dentry_operations for each possible | |
1808 | * combination rather than always installing all operations. | |
1809 | */ | |
1810 | void generic_set_encrypted_ci_d_ops(struct dentry *dentry) | |
1811 | { | |
1812 | #ifdef CONFIG_FS_ENCRYPTION | |
1813 | bool needs_encrypt_ops = dentry->d_flags & DCACHE_NOKEY_NAME; | |
1814 | #endif | |
5298d4bf | 1815 | #if IS_ENABLED(CONFIG_UNICODE) |
608af703 DR |
1816 | bool needs_ci_ops = dentry->d_sb->s_encoding; |
1817 | #endif | |
5298d4bf | 1818 | #if defined(CONFIG_FS_ENCRYPTION) && IS_ENABLED(CONFIG_UNICODE) |
608af703 DR |
1819 | if (needs_encrypt_ops && needs_ci_ops) { |
1820 | d_set_d_op(dentry, &generic_encrypted_ci_dentry_ops); | |
1821 | return; | |
1822 | } | |
c843843e | 1823 | #endif |
608af703 DR |
1824 | #ifdef CONFIG_FS_ENCRYPTION |
1825 | if (needs_encrypt_ops) { | |
1826 | d_set_d_op(dentry, &generic_encrypted_dentry_ops); | |
1827 | return; | |
1828 | } | |
1829 | #endif | |
5298d4bf | 1830 | #if IS_ENABLED(CONFIG_UNICODE) |
608af703 DR |
1831 | if (needs_ci_ops) { |
1832 | d_set_d_op(dentry, &generic_ci_dentry_ops); | |
1833 | return; | |
1834 | } | |
1835 | #endif | |
1836 | } | |
1837 | EXPORT_SYMBOL(generic_set_encrypted_ci_d_ops); | |
5ca14835 AM |
1838 | |
1839 | /** | |
1840 | * inode_maybe_inc_iversion - increments i_version | |
1841 | * @inode: inode with the i_version that should be updated | |
1842 | * @force: increment the counter even if it's not necessary? | |
1843 | * | |
1844 | * Every time the inode is modified, the i_version field must be seen to have | |
1845 | * changed by any observer. | |
1846 | * | |
1847 | * If "force" is set or the QUERIED flag is set, then ensure that we increment | |
1848 | * the value, and clear the queried flag. | |
1849 | * | |
1850 | * In the common case where neither is set, then we can return "false" without | |
1851 | * updating i_version. | |
1852 | * | |
1853 | * If this function returns false, and no other metadata has changed, then we | |
1854 | * can avoid logging the metadata. | |
1855 | */ | |
1856 | bool inode_maybe_inc_iversion(struct inode *inode, bool force) | |
1857 | { | |
1858 | u64 cur, new; | |
1859 | ||
1860 | /* | |
1861 | * The i_version field is not strictly ordered with any other inode | |
1862 | * information, but the legacy inode_inc_iversion code used a spinlock | |
1863 | * to serialize increments. | |
1864 | * | |
1865 | * Here, we add full memory barriers to ensure that any de-facto | |
1866 | * ordering with other info is preserved. | |
1867 | * | |
1868 | * This barrier pairs with the barrier in inode_query_iversion() | |
1869 | */ | |
1870 | smp_mb(); | |
1871 | cur = inode_peek_iversion_raw(inode); | |
1872 | do { | |
1873 | /* If flag is clear then we needn't do anything */ | |
1874 | if (!force && !(cur & I_VERSION_QUERIED)) | |
1875 | return false; | |
1876 | ||
1877 | /* Since lowest bit is flag, add 2 to avoid it */ | |
1878 | new = (cur & ~I_VERSION_QUERIED) + I_VERSION_INCREMENT; | |
1879 | } while (!atomic64_try_cmpxchg(&inode->i_version, &cur, new)); | |
1880 | return true; | |
1881 | } | |
1882 | EXPORT_SYMBOL(inode_maybe_inc_iversion); | |
c5bc1b3f JL |
1883 | |
1884 | /** | |
1885 | * inode_query_iversion - read i_version for later use | |
1886 | * @inode: inode from which i_version should be read | |
1887 | * | |
1888 | * Read the inode i_version counter. This should be used by callers that wish | |
1889 | * to store the returned i_version for later comparison. This will guarantee | |
1890 | * that a later query of the i_version will result in a different value if | |
1891 | * anything has changed. | |
1892 | * | |
1893 | * In this implementation, we fetch the current value, set the QUERIED flag and | |
1894 | * then try to swap it into place with a cmpxchg, if it wasn't already set. If | |
1895 | * that fails, we try again with the newly fetched value from the cmpxchg. | |
1896 | */ | |
1897 | u64 inode_query_iversion(struct inode *inode) | |
1898 | { | |
1899 | u64 cur, new; | |
1900 | ||
1901 | cur = inode_peek_iversion_raw(inode); | |
1902 | do { | |
1903 | /* If flag is already set, then no need to swap */ | |
1904 | if (cur & I_VERSION_QUERIED) { | |
1905 | /* | |
1906 | * This barrier (and the implicit barrier in the | |
1907 | * cmpxchg below) pairs with the barrier in | |
1908 | * inode_maybe_inc_iversion(). | |
1909 | */ | |
1910 | smp_mb(); | |
1911 | break; | |
1912 | } | |
1913 | ||
1914 | new = cur | I_VERSION_QUERIED; | |
1915 | } while (!atomic64_try_cmpxchg(&inode->i_version, &cur, new)); | |
1916 | return cur >> I_VERSION_QUERIED_SHIFT; | |
1917 | } | |
1918 | EXPORT_SYMBOL(inode_query_iversion); | |
44fff0fa CH |
1919 | |
1920 | ssize_t direct_write_fallback(struct kiocb *iocb, struct iov_iter *iter, | |
1921 | ssize_t direct_written, ssize_t buffered_written) | |
1922 | { | |
1923 | struct address_space *mapping = iocb->ki_filp->f_mapping; | |
1924 | loff_t pos = iocb->ki_pos - buffered_written; | |
1925 | loff_t end = iocb->ki_pos - 1; | |
1926 | int err; | |
1927 | ||
1928 | /* | |
1929 | * If the buffered write fallback returned an error, we want to return | |
1930 | * the number of bytes which were written by direct I/O, or the error | |
1931 | * code if that was zero. | |
1932 | * | |
1933 | * Note that this differs from normal direct-io semantics, which will | |
1934 | * return -EFOO even if some bytes were written. | |
1935 | */ | |
1936 | if (unlikely(buffered_written < 0)) { | |
1937 | if (direct_written) | |
1938 | return direct_written; | |
1939 | return buffered_written; | |
1940 | } | |
1941 | ||
1942 | /* | |
1943 | * We need to ensure that the page cache pages are written to disk and | |
1944 | * invalidated to preserve the expected O_DIRECT semantics. | |
1945 | */ | |
1946 | err = filemap_write_and_wait_range(mapping, pos, end); | |
1947 | if (err < 0) { | |
1948 | /* | |
1949 | * We don't know how much we wrote, so just return the number of | |
1950 | * bytes which were direct-written | |
1951 | */ | |
8287474a | 1952 | iocb->ki_pos -= buffered_written; |
44fff0fa CH |
1953 | if (direct_written) |
1954 | return direct_written; | |
1955 | return err; | |
1956 | } | |
1957 | invalidate_mapping_pages(mapping, pos >> PAGE_SHIFT, end >> PAGE_SHIFT); | |
1958 | return direct_written + buffered_written; | |
1959 | } | |
1960 | EXPORT_SYMBOL_GPL(direct_write_fallback); | |
077c212f JL |
1961 | |
1962 | /** | |
1963 | * simple_inode_init_ts - initialize the timestamps for a new inode | |
1964 | * @inode: inode to be initialized | |
1965 | * | |
1966 | * When a new inode is created, most filesystems set the timestamps to the | |
1967 | * current time. Add a helper to do this. | |
1968 | */ | |
1969 | struct timespec64 simple_inode_init_ts(struct inode *inode) | |
1970 | { | |
1971 | struct timespec64 ts = inode_set_ctime_current(inode); | |
1972 | ||
1973 | inode_set_atime_to_ts(inode, ts); | |
1974 | inode_set_mtime_to_ts(inode, ts); | |
1975 | return ts; | |
1976 | } | |
1977 | EXPORT_SYMBOL(simple_inode_init_ts); |