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, |
da549bdd | 107 | struct hlist_node **p, |
d4f4de5e | 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); | |
da549bdd AV |
114 | while (*p) { |
115 | struct dentry *d = hlist_entry(*p, struct dentry, d_sib); | |
116 | p = &d->d_sib.next; | |
d4f4de5e AV |
117 | // we must at least skip cursors, to avoid livelocks |
118 | if (d->d_flags & DCACHE_DENTRY_CURSOR) | |
119 | continue; | |
120 | if (simple_positive(d) && !--count) { | |
121 | spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED); | |
122 | if (simple_positive(d)) | |
123 | found = dget_dlock(d); | |
124 | spin_unlock(&d->d_lock); | |
125 | if (likely(found)) | |
126 | break; | |
127 | count = 1; | |
128 | } | |
129 | if (need_resched()) { | |
da549bdd AV |
130 | if (!hlist_unhashed(&cursor->d_sib)) |
131 | __hlist_del(&cursor->d_sib); | |
132 | hlist_add_behind(&cursor->d_sib, &d->d_sib); | |
133 | p = &cursor->d_sib.next; | |
d4f4de5e AV |
134 | spin_unlock(&dentry->d_lock); |
135 | cond_resched(); | |
136 | spin_lock(&dentry->d_lock); | |
4f42c1b5 AV |
137 | } |
138 | } | |
d4f4de5e | 139 | spin_unlock(&dentry->d_lock); |
26b6c984 AV |
140 | dput(last); |
141 | return found; | |
4f42c1b5 AV |
142 | } |
143 | ||
965c8e59 | 144 | loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence) |
1da177e4 | 145 | { |
2fd6b7f5 | 146 | struct dentry *dentry = file->f_path.dentry; |
965c8e59 | 147 | switch (whence) { |
1da177e4 LT |
148 | case 1: |
149 | offset += file->f_pos; | |
df561f66 | 150 | fallthrough; |
1da177e4 LT |
151 | case 0: |
152 | if (offset >= 0) | |
153 | break; | |
df561f66 | 154 | fallthrough; |
1da177e4 | 155 | default: |
1da177e4 LT |
156 | return -EINVAL; |
157 | } | |
158 | if (offset != file->f_pos) { | |
d4f4de5e AV |
159 | struct dentry *cursor = file->private_data; |
160 | struct dentry *to = NULL; | |
d4f4de5e | 161 | |
d4f4de5e AV |
162 | inode_lock_shared(dentry->d_inode); |
163 | ||
26b6c984 | 164 | if (offset > 2) |
da549bdd | 165 | to = scan_positives(cursor, &dentry->d_children.first, |
26b6c984 AV |
166 | offset - 2, NULL); |
167 | spin_lock(&dentry->d_lock); | |
da549bdd | 168 | hlist_del_init(&cursor->d_sib); |
26b6c984 | 169 | if (to) |
da549bdd | 170 | hlist_add_behind(&cursor->d_sib, &to->d_sib); |
26b6c984 | 171 | spin_unlock(&dentry->d_lock); |
d4f4de5e AV |
172 | dput(to); |
173 | ||
26b6c984 AV |
174 | file->f_pos = offset; |
175 | ||
d4f4de5e | 176 | inode_unlock_shared(dentry->d_inode); |
1da177e4 | 177 | } |
1da177e4 LT |
178 | return offset; |
179 | } | |
12f38872 | 180 | EXPORT_SYMBOL(dcache_dir_lseek); |
1da177e4 | 181 | |
1da177e4 LT |
182 | /* |
183 | * Directory is locked and all positive dentries in it are safe, since | |
184 | * for ramfs-type trees they can't go away without unlink() or rmdir(), | |
185 | * both impossible due to the lock on directory. | |
186 | */ | |
187 | ||
5f99f4e7 | 188 | int dcache_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 189 | { |
5f99f4e7 AV |
190 | struct dentry *dentry = file->f_path.dentry; |
191 | struct dentry *cursor = file->private_data; | |
d4f4de5e | 192 | struct dentry *next = NULL; |
da549bdd | 193 | struct hlist_node **p; |
1da177e4 | 194 | |
5f99f4e7 AV |
195 | if (!dir_emit_dots(file, ctx)) |
196 | return 0; | |
5f99f4e7 | 197 | |
4f42c1b5 | 198 | if (ctx->pos == 2) |
da549bdd | 199 | p = &dentry->d_children.first; |
26b6c984 | 200 | else |
da549bdd | 201 | p = &cursor->d_sib.next; |
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, |
364595a6 JL |
205 | d_inode(next)->i_ino, |
206 | fs_umode_to_dtype(d_inode(next)->i_mode))) | |
4f42c1b5 | 207 | break; |
5f99f4e7 | 208 | ctx->pos++; |
da549bdd | 209 | p = &next->d_sib.next; |
1da177e4 | 210 | } |
d4f4de5e | 211 | spin_lock(&dentry->d_lock); |
da549bdd | 212 | hlist_del_init(&cursor->d_sib); |
26b6c984 | 213 | if (next) |
da549bdd | 214 | hlist_add_before(&cursor->d_sib, &next->d_sib); |
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 | |
6faddda6 CL |
243 | static void offset_set(struct dentry *dentry, u32 offset) |
244 | { | |
245 | dentry->d_fsdata = (void *)((uintptr_t)(offset)); | |
246 | } | |
247 | ||
248 | static u32 dentry2offset(struct dentry *dentry) | |
249 | { | |
250 | return (u32)((uintptr_t)(dentry->d_fsdata)); | |
251 | } | |
252 | ||
bbaef797 CL |
253 | static struct lock_class_key simple_offset_xa_lock; |
254 | ||
6faddda6 CL |
255 | /** |
256 | * simple_offset_init - initialize an offset_ctx | |
257 | * @octx: directory offset map to be initialized | |
258 | * | |
259 | */ | |
260 | void simple_offset_init(struct offset_ctx *octx) | |
261 | { | |
262 | xa_init_flags(&octx->xa, XA_FLAGS_ALLOC1); | |
bbaef797 | 263 | lockdep_set_class(&octx->xa.xa_lock, &simple_offset_xa_lock); |
6faddda6 CL |
264 | |
265 | /* 0 is '.', 1 is '..', so always start with offset 2 */ | |
266 | octx->next_offset = 2; | |
267 | } | |
268 | ||
269 | /** | |
270 | * simple_offset_add - Add an entry to a directory's offset map | |
271 | * @octx: directory offset ctx to be updated | |
272 | * @dentry: new dentry being added | |
273 | * | |
274 | * Returns zero on success. @so_ctx and the dentry offset are updated. | |
275 | * Otherwise, a negative errno value is returned. | |
276 | */ | |
277 | int simple_offset_add(struct offset_ctx *octx, struct dentry *dentry) | |
278 | { | |
279 | static const struct xa_limit limit = XA_LIMIT(2, U32_MAX); | |
280 | u32 offset; | |
281 | int ret; | |
282 | ||
283 | if (dentry2offset(dentry) != 0) | |
284 | return -EBUSY; | |
285 | ||
286 | ret = xa_alloc_cyclic(&octx->xa, &offset, dentry, limit, | |
287 | &octx->next_offset, GFP_KERNEL); | |
288 | if (ret < 0) | |
289 | return ret; | |
290 | ||
291 | offset_set(dentry, offset); | |
292 | return 0; | |
293 | } | |
294 | ||
295 | /** | |
296 | * simple_offset_remove - Remove an entry to a directory's offset map | |
297 | * @octx: directory offset ctx to be updated | |
298 | * @dentry: dentry being removed | |
299 | * | |
300 | */ | |
301 | void simple_offset_remove(struct offset_ctx *octx, struct dentry *dentry) | |
302 | { | |
303 | u32 offset; | |
304 | ||
305 | offset = dentry2offset(dentry); | |
306 | if (offset == 0) | |
307 | return; | |
308 | ||
309 | xa_erase(&octx->xa, offset); | |
310 | offset_set(dentry, 0); | |
311 | } | |
312 | ||
313 | /** | |
314 | * simple_offset_rename_exchange - exchange rename with directory offsets | |
315 | * @old_dir: parent of dentry being moved | |
316 | * @old_dentry: dentry being moved | |
317 | * @new_dir: destination parent | |
318 | * @new_dentry: destination dentry | |
319 | * | |
320 | * Returns zero on success. Otherwise a negative errno is returned and the | |
321 | * rename is rolled back. | |
322 | */ | |
323 | int simple_offset_rename_exchange(struct inode *old_dir, | |
324 | struct dentry *old_dentry, | |
325 | struct inode *new_dir, | |
326 | struct dentry *new_dentry) | |
327 | { | |
328 | struct offset_ctx *old_ctx = old_dir->i_op->get_offset_ctx(old_dir); | |
329 | struct offset_ctx *new_ctx = new_dir->i_op->get_offset_ctx(new_dir); | |
330 | u32 old_index = dentry2offset(old_dentry); | |
331 | u32 new_index = dentry2offset(new_dentry); | |
332 | int ret; | |
333 | ||
334 | simple_offset_remove(old_ctx, old_dentry); | |
335 | simple_offset_remove(new_ctx, new_dentry); | |
336 | ||
337 | ret = simple_offset_add(new_ctx, old_dentry); | |
338 | if (ret) | |
339 | goto out_restore; | |
340 | ||
341 | ret = simple_offset_add(old_ctx, new_dentry); | |
342 | if (ret) { | |
343 | simple_offset_remove(new_ctx, old_dentry); | |
344 | goto out_restore; | |
345 | } | |
346 | ||
347 | ret = simple_rename_exchange(old_dir, old_dentry, new_dir, new_dentry); | |
348 | if (ret) { | |
349 | simple_offset_remove(new_ctx, old_dentry); | |
350 | simple_offset_remove(old_ctx, new_dentry); | |
351 | goto out_restore; | |
352 | } | |
353 | return 0; | |
354 | ||
355 | out_restore: | |
356 | offset_set(old_dentry, old_index); | |
357 | xa_store(&old_ctx->xa, old_index, old_dentry, GFP_KERNEL); | |
358 | offset_set(new_dentry, new_index); | |
359 | xa_store(&new_ctx->xa, new_index, new_dentry, GFP_KERNEL); | |
360 | return ret; | |
361 | } | |
362 | ||
363 | /** | |
364 | * simple_offset_destroy - Release offset map | |
365 | * @octx: directory offset ctx that is about to be destroyed | |
366 | * | |
367 | * During fs teardown (eg. umount), a directory's offset map might still | |
368 | * contain entries. xa_destroy() cleans out anything that remains. | |
369 | */ | |
370 | void simple_offset_destroy(struct offset_ctx *octx) | |
371 | { | |
372 | xa_destroy(&octx->xa); | |
373 | } | |
374 | ||
375 | /** | |
376 | * offset_dir_llseek - Advance the read position of a directory descriptor | |
377 | * @file: an open directory whose position is to be updated | |
378 | * @offset: a byte offset | |
379 | * @whence: enumerator describing the starting position for this update | |
380 | * | |
381 | * SEEK_END, SEEK_DATA, and SEEK_HOLE are not supported for directories. | |
382 | * | |
383 | * Returns the updated read position if successful; otherwise a | |
384 | * negative errno is returned and the read position remains unchanged. | |
385 | */ | |
386 | static loff_t offset_dir_llseek(struct file *file, loff_t offset, int whence) | |
387 | { | |
388 | switch (whence) { | |
389 | case SEEK_CUR: | |
390 | offset += file->f_pos; | |
391 | fallthrough; | |
392 | case SEEK_SET: | |
393 | if (offset >= 0) | |
394 | break; | |
395 | fallthrough; | |
396 | default: | |
397 | return -EINVAL; | |
398 | } | |
399 | ||
796432ef CL |
400 | /* In this case, ->private_data is protected by f_pos_lock */ |
401 | file->private_data = NULL; | |
6faddda6 CL |
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 | ||
796432ef | 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 | 439 | if (!dentry) |
796432ef | 440 | return ERR_PTR(-ENOENT); |
6faddda6 CL |
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 | } | |
796432ef | 450 | return NULL; |
6faddda6 CL |
451 | } |
452 | ||
453 | /** | |
454 | * offset_readdir - Emit entries starting at offset @ctx->pos | |
455 | * @file: an open directory to iterate over | |
456 | * @ctx: directory iteration context | |
457 | * | |
458 | * Caller must hold @file's i_rwsem to prevent insertion or removal of | |
459 | * entries during this call. | |
460 | * | |
461 | * On entry, @ctx->pos contains an offset that represents the first entry | |
462 | * to be read from the directory. | |
463 | * | |
464 | * The operation continues until there are no more entries to read, or | |
465 | * until the ctx->actor indicates there is no more space in the caller's | |
466 | * output buffer. | |
467 | * | |
468 | * On return, @ctx->pos contains an offset that will read the next entry | |
2be4f05a | 469 | * in this directory when offset_readdir() is called again with @ctx. |
6faddda6 CL |
470 | * |
471 | * Return values: | |
472 | * %0 - Complete | |
473 | */ | |
474 | static int offset_readdir(struct file *file, struct dir_context *ctx) | |
475 | { | |
476 | struct dentry *dir = file->f_path.dentry; | |
477 | ||
478 | lockdep_assert_held(&d_inode(dir)->i_rwsem); | |
479 | ||
480 | if (!dir_emit_dots(file, ctx)) | |
481 | return 0; | |
482 | ||
796432ef CL |
483 | /* In this case, ->private_data is protected by f_pos_lock */ |
484 | if (ctx->pos == 2) | |
485 | file->private_data = NULL; | |
486 | else if (file->private_data == ERR_PTR(-ENOENT)) | |
487 | return 0; | |
488 | file->private_data = offset_iterate_dir(d_inode(dir), ctx); | |
6faddda6 CL |
489 | return 0; |
490 | } | |
491 | ||
492 | const struct file_operations simple_offset_dir_operations = { | |
493 | .llseek = offset_dir_llseek, | |
494 | .iterate_shared = offset_readdir, | |
495 | .read = generic_read_dir, | |
496 | .fsync = noop_fsync, | |
497 | }; | |
498 | ||
a3d1e7eb AV |
499 | static struct dentry *find_next_child(struct dentry *parent, struct dentry *prev) |
500 | { | |
da549bdd | 501 | struct dentry *child = NULL, *d; |
a3d1e7eb AV |
502 | |
503 | spin_lock(&parent->d_lock); | |
da549bdd AV |
504 | d = prev ? d_next_sibling(prev) : d_first_child(parent); |
505 | hlist_for_each_entry_from(d, d_sib) { | |
a3d1e7eb AV |
506 | if (simple_positive(d)) { |
507 | spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED); | |
508 | if (simple_positive(d)) | |
509 | child = dget_dlock(d); | |
510 | spin_unlock(&d->d_lock); | |
511 | if (likely(child)) | |
512 | break; | |
513 | } | |
514 | } | |
515 | spin_unlock(&parent->d_lock); | |
516 | dput(prev); | |
517 | return child; | |
518 | } | |
519 | ||
520 | void simple_recursive_removal(struct dentry *dentry, | |
521 | void (*callback)(struct dentry *)) | |
522 | { | |
523 | struct dentry *this = dget(dentry); | |
524 | while (true) { | |
525 | struct dentry *victim = NULL, *child; | |
526 | struct inode *inode = this->d_inode; | |
527 | ||
528 | inode_lock(inode); | |
529 | if (d_is_dir(this)) | |
530 | inode->i_flags |= S_DEAD; | |
531 | while ((child = find_next_child(this, victim)) == NULL) { | |
532 | // kill and ascend | |
533 | // update metadata while it's still locked | |
f7f43858 | 534 | inode_set_ctime_current(inode); |
a3d1e7eb AV |
535 | clear_nlink(inode); |
536 | inode_unlock(inode); | |
537 | victim = this; | |
538 | this = this->d_parent; | |
539 | inode = this->d_inode; | |
540 | inode_lock(inode); | |
541 | if (simple_positive(victim)) { | |
542 | d_invalidate(victim); // avoid lost mounts | |
543 | if (d_is_dir(victim)) | |
544 | fsnotify_rmdir(inode, victim); | |
545 | else | |
546 | fsnotify_unlink(inode, victim); | |
547 | if (callback) | |
548 | callback(victim); | |
549 | dput(victim); // unpin it | |
550 | } | |
551 | if (victim == dentry) { | |
077c212f JL |
552 | inode_set_mtime_to_ts(inode, |
553 | inode_set_ctime_current(inode)); | |
a3d1e7eb AV |
554 | if (d_is_dir(dentry)) |
555 | drop_nlink(inode); | |
556 | inode_unlock(inode); | |
557 | dput(dentry); | |
558 | return; | |
559 | } | |
560 | } | |
561 | inode_unlock(inode); | |
562 | this = child; | |
563 | } | |
564 | } | |
565 | EXPORT_SYMBOL(simple_recursive_removal); | |
566 | ||
759b9775 HD |
567 | static const struct super_operations simple_super_operations = { |
568 | .statfs = simple_statfs, | |
569 | }; | |
570 | ||
db2c246a | 571 | static int pseudo_fs_fill_super(struct super_block *s, struct fs_context *fc) |
1da177e4 | 572 | { |
31d6d5ce | 573 | struct pseudo_fs_context *ctx = fc->fs_private; |
1da177e4 | 574 | struct inode *root; |
1da177e4 | 575 | |
89a4eb4b | 576 | s->s_maxbytes = MAX_LFS_FILESIZE; |
3971e1a9 AN |
577 | s->s_blocksize = PAGE_SIZE; |
578 | s->s_blocksize_bits = PAGE_SHIFT; | |
8d9e46d8 AV |
579 | s->s_magic = ctx->magic; |
580 | s->s_op = ctx->ops ?: &simple_super_operations; | |
581 | s->s_xattr = ctx->xattr; | |
1da177e4 LT |
582 | s->s_time_gran = 1; |
583 | root = new_inode(s); | |
584 | if (!root) | |
db2c246a DH |
585 | return -ENOMEM; |
586 | ||
1a1c9bb4 JL |
587 | /* |
588 | * since this is the first inode, make it number 1. New inodes created | |
589 | * after this must take care not to collide with it (by passing | |
590 | * max_reserved of 1 to iunique). | |
591 | */ | |
592 | root->i_ino = 1; | |
1da177e4 | 593 | root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR; |
077c212f | 594 | simple_inode_init_ts(root); |
8d9e46d8 AV |
595 | s->s_root = d_make_root(root); |
596 | if (!s->s_root) | |
db2c246a | 597 | return -ENOMEM; |
8d9e46d8 | 598 | s->s_d_op = ctx->dops; |
31d6d5ce | 599 | return 0; |
db2c246a | 600 | } |
8d9e46d8 | 601 | |
db2c246a DH |
602 | static int pseudo_fs_get_tree(struct fs_context *fc) |
603 | { | |
2ac295d4 | 604 | return get_tree_nodev(fc, pseudo_fs_fill_super); |
31d6d5ce DH |
605 | } |
606 | ||
607 | static void pseudo_fs_free(struct fs_context *fc) | |
608 | { | |
609 | kfree(fc->fs_private); | |
610 | } | |
611 | ||
612 | static const struct fs_context_operations pseudo_fs_context_ops = { | |
613 | .free = pseudo_fs_free, | |
614 | .get_tree = pseudo_fs_get_tree, | |
615 | }; | |
616 | ||
617 | /* | |
618 | * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that | |
619 | * will never be mountable) | |
620 | */ | |
621 | struct pseudo_fs_context *init_pseudo(struct fs_context *fc, | |
622 | unsigned long magic) | |
623 | { | |
624 | struct pseudo_fs_context *ctx; | |
625 | ||
626 | ctx = kzalloc(sizeof(struct pseudo_fs_context), GFP_KERNEL); | |
627 | if (likely(ctx)) { | |
628 | ctx->magic = magic; | |
629 | fc->fs_private = ctx; | |
630 | fc->ops = &pseudo_fs_context_ops; | |
db2c246a DH |
631 | fc->sb_flags |= SB_NOUSER; |
632 | fc->global = true; | |
1da177e4 | 633 | } |
31d6d5ce | 634 | return ctx; |
1da177e4 | 635 | } |
31d6d5ce | 636 | EXPORT_SYMBOL(init_pseudo); |
1da177e4 | 637 | |
20955e89 SB |
638 | int simple_open(struct inode *inode, struct file *file) |
639 | { | |
640 | if (inode->i_private) | |
641 | file->private_data = inode->i_private; | |
642 | return 0; | |
643 | } | |
12f38872 | 644 | EXPORT_SYMBOL(simple_open); |
20955e89 | 645 | |
1da177e4 LT |
646 | int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry) |
647 | { | |
dea655c2 | 648 | struct inode *inode = d_inode(old_dentry); |
1da177e4 | 649 | |
077c212f JL |
650 | inode_set_mtime_to_ts(dir, |
651 | inode_set_ctime_to_ts(dir, inode_set_ctime_current(inode))); | |
d8c76e6f | 652 | inc_nlink(inode); |
7de9c6ee | 653 | ihold(inode); |
1da177e4 LT |
654 | dget(dentry); |
655 | d_instantiate(dentry, inode); | |
656 | return 0; | |
657 | } | |
12f38872 | 658 | EXPORT_SYMBOL(simple_link); |
1da177e4 | 659 | |
1da177e4 LT |
660 | int simple_empty(struct dentry *dentry) |
661 | { | |
662 | struct dentry *child; | |
663 | int ret = 0; | |
664 | ||
2fd6b7f5 | 665 | spin_lock(&dentry->d_lock); |
da549bdd | 666 | hlist_for_each_entry(child, &dentry->d_children, d_sib) { |
da502956 NP |
667 | spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED); |
668 | if (simple_positive(child)) { | |
669 | spin_unlock(&child->d_lock); | |
1da177e4 | 670 | goto out; |
da502956 NP |
671 | } |
672 | spin_unlock(&child->d_lock); | |
673 | } | |
1da177e4 LT |
674 | ret = 1; |
675 | out: | |
2fd6b7f5 | 676 | spin_unlock(&dentry->d_lock); |
1da177e4 LT |
677 | return ret; |
678 | } | |
12f38872 | 679 | EXPORT_SYMBOL(simple_empty); |
1da177e4 LT |
680 | |
681 | int simple_unlink(struct inode *dir, struct dentry *dentry) | |
682 | { | |
dea655c2 | 683 | struct inode *inode = d_inode(dentry); |
1da177e4 | 684 | |
077c212f JL |
685 | inode_set_mtime_to_ts(dir, |
686 | inode_set_ctime_to_ts(dir, inode_set_ctime_current(inode))); | |
9a53c3a7 | 687 | drop_nlink(inode); |
1da177e4 LT |
688 | dput(dentry); |
689 | return 0; | |
690 | } | |
12f38872 | 691 | EXPORT_SYMBOL(simple_unlink); |
1da177e4 LT |
692 | |
693 | int simple_rmdir(struct inode *dir, struct dentry *dentry) | |
694 | { | |
695 | if (!simple_empty(dentry)) | |
696 | return -ENOTEMPTY; | |
697 | ||
dea655c2 | 698 | drop_nlink(d_inode(dentry)); |
1da177e4 | 699 | simple_unlink(dir, dentry); |
9a53c3a7 | 700 | drop_nlink(dir); |
1da177e4 LT |
701 | return 0; |
702 | } | |
12f38872 | 703 | EXPORT_SYMBOL(simple_rmdir); |
1da177e4 | 704 | |
0c476792 JL |
705 | /** |
706 | * simple_rename_timestamp - update the various inode timestamps for rename | |
707 | * @old_dir: old parent directory | |
708 | * @old_dentry: dentry that is being renamed | |
709 | * @new_dir: new parent directory | |
710 | * @new_dentry: target for rename | |
711 | * | |
712 | * POSIX mandates that the old and new parent directories have their ctime and | |
713 | * mtime updated, and that inodes of @old_dentry and @new_dentry (if any), have | |
714 | * their ctime updated. | |
715 | */ | |
716 | void simple_rename_timestamp(struct inode *old_dir, struct dentry *old_dentry, | |
717 | struct inode *new_dir, struct dentry *new_dentry) | |
718 | { | |
719 | struct inode *newino = d_inode(new_dentry); | |
720 | ||
077c212f | 721 | inode_set_mtime_to_ts(old_dir, inode_set_ctime_current(old_dir)); |
0c476792 | 722 | if (new_dir != old_dir) |
077c212f JL |
723 | inode_set_mtime_to_ts(new_dir, |
724 | inode_set_ctime_current(new_dir)); | |
0c476792 JL |
725 | inode_set_ctime_current(d_inode(old_dentry)); |
726 | if (newino) | |
727 | inode_set_ctime_current(newino); | |
728 | } | |
729 | EXPORT_SYMBOL_GPL(simple_rename_timestamp); | |
730 | ||
6429e463 LB |
731 | int simple_rename_exchange(struct inode *old_dir, struct dentry *old_dentry, |
732 | struct inode *new_dir, struct dentry *new_dentry) | |
733 | { | |
734 | bool old_is_dir = d_is_dir(old_dentry); | |
735 | bool new_is_dir = d_is_dir(new_dentry); | |
736 | ||
737 | if (old_dir != new_dir && old_is_dir != new_is_dir) { | |
738 | if (old_is_dir) { | |
739 | drop_nlink(old_dir); | |
740 | inc_nlink(new_dir); | |
741 | } else { | |
742 | drop_nlink(new_dir); | |
743 | inc_nlink(old_dir); | |
744 | } | |
745 | } | |
0c476792 | 746 | simple_rename_timestamp(old_dir, old_dentry, new_dir, new_dentry); |
6429e463 LB |
747 | return 0; |
748 | } | |
749 | EXPORT_SYMBOL_GPL(simple_rename_exchange); | |
750 | ||
e18275ae | 751 | int simple_rename(struct mnt_idmap *idmap, struct inode *old_dir, |
549c7297 CB |
752 | struct dentry *old_dentry, struct inode *new_dir, |
753 | struct dentry *new_dentry, unsigned int flags) | |
1da177e4 | 754 | { |
e36cb0b8 | 755 | int they_are_dirs = d_is_dir(old_dentry); |
1da177e4 | 756 | |
3871cb8c | 757 | if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE)) |
e0e0be8a MS |
758 | return -EINVAL; |
759 | ||
3871cb8c LB |
760 | if (flags & RENAME_EXCHANGE) |
761 | return simple_rename_exchange(old_dir, old_dentry, new_dir, new_dentry); | |
762 | ||
1da177e4 LT |
763 | if (!simple_empty(new_dentry)) |
764 | return -ENOTEMPTY; | |
765 | ||
dea655c2 | 766 | if (d_really_is_positive(new_dentry)) { |
1da177e4 | 767 | simple_unlink(new_dir, new_dentry); |
841590ce | 768 | if (they_are_dirs) { |
dea655c2 | 769 | drop_nlink(d_inode(new_dentry)); |
9a53c3a7 | 770 | drop_nlink(old_dir); |
841590ce | 771 | } |
1da177e4 | 772 | } else if (they_are_dirs) { |
9a53c3a7 | 773 | drop_nlink(old_dir); |
d8c76e6f | 774 | inc_nlink(new_dir); |
1da177e4 LT |
775 | } |
776 | ||
0c476792 | 777 | simple_rename_timestamp(old_dir, old_dentry, new_dir, new_dentry); |
1da177e4 LT |
778 | return 0; |
779 | } | |
12f38872 | 780 | EXPORT_SYMBOL(simple_rename); |
1da177e4 | 781 | |
7bb46a67 | 782 | /** |
eef2380c | 783 | * simple_setattr - setattr for simple filesystem |
c1632a0f | 784 | * @idmap: idmap of the target mount |
7bb46a67 | 785 | * @dentry: dentry |
786 | * @iattr: iattr structure | |
787 | * | |
788 | * Returns 0 on success, -error on failure. | |
789 | * | |
eef2380c CH |
790 | * simple_setattr is a simple ->setattr implementation without a proper |
791 | * implementation of size changes. | |
792 | * | |
793 | * It can either be used for in-memory filesystems or special files | |
794 | * on simple regular filesystems. Anything that needs to change on-disk | |
795 | * or wire state on size changes needs its own setattr method. | |
7bb46a67 | 796 | */ |
c1632a0f | 797 | int simple_setattr(struct mnt_idmap *idmap, struct dentry *dentry, |
549c7297 | 798 | struct iattr *iattr) |
7bb46a67 | 799 | { |
dea655c2 | 800 | struct inode *inode = d_inode(dentry); |
7bb46a67 | 801 | int error; |
802 | ||
c1632a0f | 803 | error = setattr_prepare(idmap, dentry, iattr); |
7bb46a67 | 804 | if (error) |
805 | return error; | |
806 | ||
2c27c65e CH |
807 | if (iattr->ia_valid & ATTR_SIZE) |
808 | truncate_setsize(inode, iattr->ia_size); | |
c1632a0f | 809 | setattr_copy(idmap, inode, iattr); |
eef2380c CH |
810 | mark_inode_dirty(inode); |
811 | return 0; | |
7bb46a67 | 812 | } |
813 | EXPORT_SYMBOL(simple_setattr); | |
814 | ||
a77f580a | 815 | static int simple_read_folio(struct file *file, struct folio *folio) |
1da177e4 | 816 | { |
a77f580a MWO |
817 | folio_zero_range(folio, 0, folio_size(folio)); |
818 | flush_dcache_folio(folio); | |
819 | folio_mark_uptodate(folio); | |
820 | folio_unlock(folio); | |
1da177e4 LT |
821 | return 0; |
822 | } | |
823 | ||
afddba49 | 824 | int simple_write_begin(struct file *file, struct address_space *mapping, |
9d6b0cd7 | 825 | loff_t pos, unsigned len, |
afddba49 NP |
826 | struct page **pagep, void **fsdata) |
827 | { | |
5522d9f7 | 828 | struct folio *folio; |
afddba49 | 829 | |
5522d9f7 MWO |
830 | folio = __filemap_get_folio(mapping, pos / PAGE_SIZE, FGP_WRITEBEGIN, |
831 | mapping_gfp_mask(mapping)); | |
832 | if (IS_ERR(folio)) | |
833 | return PTR_ERR(folio); | |
afddba49 | 834 | |
5522d9f7 | 835 | *pagep = &folio->page; |
afddba49 | 836 | |
5522d9f7 MWO |
837 | if (!folio_test_uptodate(folio) && (len != folio_size(folio))) { |
838 | size_t from = offset_in_folio(folio, pos); | |
193cf4b9 | 839 | |
5522d9f7 MWO |
840 | folio_zero_segments(folio, 0, from, |
841 | from + len, folio_size(folio)); | |
193cf4b9 BH |
842 | } |
843 | return 0; | |
afddba49 | 844 | } |
12f38872 | 845 | EXPORT_SYMBOL(simple_write_begin); |
afddba49 | 846 | |
ad2a722f BH |
847 | /** |
848 | * simple_write_end - .write_end helper for non-block-device FSes | |
8e88bfba | 849 | * @file: See .write_end of address_space_operations |
ad2a722f BH |
850 | * @mapping: " |
851 | * @pos: " | |
852 | * @len: " | |
853 | * @copied: " | |
854 | * @page: " | |
855 | * @fsdata: " | |
856 | * | |
857 | * simple_write_end does the minimum needed for updating a page after writing is | |
858 | * done. It has the same API signature as the .write_end of | |
859 | * address_space_operations vector. So it can just be set onto .write_end for | |
860 | * FSes that don't need any other processing. i_mutex is assumed to be held. | |
861 | * Block based filesystems should use generic_write_end(). | |
862 | * NOTE: Even though i_size might get updated by this function, mark_inode_dirty | |
863 | * is not called, so a filesystem that actually does store data in .write_inode | |
864 | * should extend on what's done here with a call to mark_inode_dirty() in the | |
865 | * case that i_size has changed. | |
04fff641 | 866 | * |
a77f580a | 867 | * Use *ONLY* with simple_read_folio() |
ad2a722f | 868 | */ |
c1e3dbe9 | 869 | static int simple_write_end(struct file *file, struct address_space *mapping, |
afddba49 NP |
870 | loff_t pos, unsigned len, unsigned copied, |
871 | struct page *page, void *fsdata) | |
872 | { | |
5522d9f7 MWO |
873 | struct folio *folio = page_folio(page); |
874 | struct inode *inode = folio->mapping->host; | |
ad2a722f | 875 | loff_t last_pos = pos + copied; |
afddba49 | 876 | |
5522d9f7 MWO |
877 | /* zero the stale part of the folio if we did a short copy */ |
878 | if (!folio_test_uptodate(folio)) { | |
04fff641 | 879 | if (copied < len) { |
5522d9f7 | 880 | size_t from = offset_in_folio(folio, pos); |
afddba49 | 881 | |
5522d9f7 | 882 | folio_zero_range(folio, from + copied, len - copied); |
04fff641 | 883 | } |
5522d9f7 | 884 | folio_mark_uptodate(folio); |
04fff641 | 885 | } |
ad2a722f BH |
886 | /* |
887 | * No need to use i_size_read() here, the i_size | |
888 | * cannot change under us because we hold the i_mutex. | |
889 | */ | |
890 | if (last_pos > inode->i_size) | |
891 | i_size_write(inode, last_pos); | |
afddba49 | 892 | |
5522d9f7 MWO |
893 | folio_mark_dirty(folio); |
894 | folio_unlock(folio); | |
895 | folio_put(folio); | |
afddba49 NP |
896 | |
897 | return copied; | |
898 | } | |
c1e3dbe9 CH |
899 | |
900 | /* | |
901 | * Provides ramfs-style behavior: data in the pagecache, but no writeback. | |
902 | */ | |
903 | const struct address_space_operations ram_aops = { | |
a77f580a | 904 | .read_folio = simple_read_folio, |
c1e3dbe9 CH |
905 | .write_begin = simple_write_begin, |
906 | .write_end = simple_write_end, | |
46de8b97 | 907 | .dirty_folio = noop_dirty_folio, |
c1e3dbe9 CH |
908 | }; |
909 | EXPORT_SYMBOL(ram_aops); | |
afddba49 | 910 | |
1a1c9bb4 JL |
911 | /* |
912 | * the inodes created here are not hashed. If you use iunique to generate | |
913 | * unique inode values later for this filesystem, then you must take care | |
914 | * to pass it an appropriate max_reserved value to avoid collisions. | |
915 | */ | |
7d683a09 | 916 | int simple_fill_super(struct super_block *s, unsigned long magic, |
cda37124 | 917 | const struct tree_descr *files) |
1da177e4 | 918 | { |
1da177e4 | 919 | struct inode *inode; |
1da177e4 LT |
920 | struct dentry *dentry; |
921 | int i; | |
922 | ||
09cbfeaf KS |
923 | s->s_blocksize = PAGE_SIZE; |
924 | s->s_blocksize_bits = PAGE_SHIFT; | |
1da177e4 | 925 | s->s_magic = magic; |
759b9775 | 926 | s->s_op = &simple_super_operations; |
1da177e4 LT |
927 | s->s_time_gran = 1; |
928 | ||
929 | inode = new_inode(s); | |
930 | if (!inode) | |
931 | return -ENOMEM; | |
1a1c9bb4 JL |
932 | /* |
933 | * because the root inode is 1, the files array must not contain an | |
934 | * entry at index 1 | |
935 | */ | |
936 | inode->i_ino = 1; | |
1da177e4 | 937 | inode->i_mode = S_IFDIR | 0755; |
077c212f | 938 | simple_inode_init_ts(inode); |
1da177e4 LT |
939 | inode->i_op = &simple_dir_inode_operations; |
940 | inode->i_fop = &simple_dir_operations; | |
bfe86848 | 941 | set_nlink(inode, 2); |
715cd66a AV |
942 | s->s_root = d_make_root(inode); |
943 | if (!s->s_root) | |
1da177e4 | 944 | return -ENOMEM; |
1da177e4 LT |
945 | for (i = 0; !files->name || files->name[0]; i++, files++) { |
946 | if (!files->name) | |
947 | continue; | |
1a1c9bb4 JL |
948 | |
949 | /* warn if it tries to conflict with the root inode */ | |
950 | if (unlikely(i == 1)) | |
951 | printk(KERN_WARNING "%s: %s passed in a files array" | |
952 | "with an index of 1!\n", __func__, | |
953 | s->s_type->name); | |
954 | ||
715cd66a | 955 | dentry = d_alloc_name(s->s_root, files->name); |
1da177e4 | 956 | if (!dentry) |
715cd66a | 957 | return -ENOMEM; |
1da177e4 | 958 | inode = new_inode(s); |
32096ea1 KK |
959 | if (!inode) { |
960 | dput(dentry); | |
715cd66a | 961 | return -ENOMEM; |
32096ea1 | 962 | } |
1da177e4 | 963 | inode->i_mode = S_IFREG | files->mode; |
077c212f | 964 | simple_inode_init_ts(inode); |
1da177e4 LT |
965 | inode->i_fop = files->ops; |
966 | inode->i_ino = i; | |
967 | d_add(dentry, inode); | |
968 | } | |
1da177e4 | 969 | return 0; |
1da177e4 | 970 | } |
12f38872 | 971 | EXPORT_SYMBOL(simple_fill_super); |
1da177e4 LT |
972 | |
973 | static DEFINE_SPINLOCK(pin_fs_lock); | |
974 | ||
1f5ce9e9 | 975 | int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count) |
1da177e4 LT |
976 | { |
977 | struct vfsmount *mnt = NULL; | |
978 | spin_lock(&pin_fs_lock); | |
979 | if (unlikely(!*mount)) { | |
980 | spin_unlock(&pin_fs_lock); | |
1751e8a6 | 981 | mnt = vfs_kern_mount(type, SB_KERNMOUNT, type->name, NULL); |
1da177e4 LT |
982 | if (IS_ERR(mnt)) |
983 | return PTR_ERR(mnt); | |
984 | spin_lock(&pin_fs_lock); | |
985 | if (!*mount) | |
986 | *mount = mnt; | |
987 | } | |
988 | mntget(*mount); | |
989 | ++*count; | |
990 | spin_unlock(&pin_fs_lock); | |
991 | mntput(mnt); | |
992 | return 0; | |
993 | } | |
12f38872 | 994 | EXPORT_SYMBOL(simple_pin_fs); |
1da177e4 LT |
995 | |
996 | void simple_release_fs(struct vfsmount **mount, int *count) | |
997 | { | |
998 | struct vfsmount *mnt; | |
999 | spin_lock(&pin_fs_lock); | |
1000 | mnt = *mount; | |
1001 | if (!--*count) | |
1002 | *mount = NULL; | |
1003 | spin_unlock(&pin_fs_lock); | |
1004 | mntput(mnt); | |
1005 | } | |
12f38872 | 1006 | EXPORT_SYMBOL(simple_release_fs); |
1da177e4 | 1007 | |
6d1029b5 AM |
1008 | /** |
1009 | * simple_read_from_buffer - copy data from the buffer to user space | |
1010 | * @to: the user space buffer to read to | |
1011 | * @count: the maximum number of bytes to read | |
1012 | * @ppos: the current position in the buffer | |
1013 | * @from: the buffer to read from | |
1014 | * @available: the size of the buffer | |
1015 | * | |
1016 | * The simple_read_from_buffer() function reads up to @count bytes from the | |
1017 | * buffer @from at offset @ppos into the user space address starting at @to. | |
1018 | * | |
1019 | * On success, the number of bytes read is returned and the offset @ppos is | |
1020 | * advanced by this number, or negative value is returned on error. | |
1021 | **/ | |
1da177e4 LT |
1022 | ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos, |
1023 | const void *from, size_t available) | |
1024 | { | |
1025 | loff_t pos = *ppos; | |
14be2746 SR |
1026 | size_t ret; |
1027 | ||
1da177e4 LT |
1028 | if (pos < 0) |
1029 | return -EINVAL; | |
14be2746 | 1030 | if (pos >= available || !count) |
1da177e4 LT |
1031 | return 0; |
1032 | if (count > available - pos) | |
1033 | count = available - pos; | |
14be2746 SR |
1034 | ret = copy_to_user(to, from + pos, count); |
1035 | if (ret == count) | |
1da177e4 | 1036 | return -EFAULT; |
14be2746 | 1037 | count -= ret; |
1da177e4 LT |
1038 | *ppos = pos + count; |
1039 | return count; | |
1040 | } | |
12f38872 | 1041 | EXPORT_SYMBOL(simple_read_from_buffer); |
1da177e4 | 1042 | |
6a727b43 JS |
1043 | /** |
1044 | * simple_write_to_buffer - copy data from user space to the buffer | |
1045 | * @to: the buffer to write to | |
1046 | * @available: the size of the buffer | |
1047 | * @ppos: the current position in the buffer | |
1048 | * @from: the user space buffer to read from | |
1049 | * @count: the maximum number of bytes to read | |
1050 | * | |
1051 | * The simple_write_to_buffer() function reads up to @count bytes from the user | |
1052 | * space address starting at @from into the buffer @to at offset @ppos. | |
1053 | * | |
1054 | * On success, the number of bytes written is returned and the offset @ppos is | |
1055 | * advanced by this number, or negative value is returned on error. | |
1056 | **/ | |
1057 | ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos, | |
1058 | const void __user *from, size_t count) | |
1059 | { | |
1060 | loff_t pos = *ppos; | |
1061 | size_t res; | |
1062 | ||
1063 | if (pos < 0) | |
1064 | return -EINVAL; | |
1065 | if (pos >= available || !count) | |
1066 | return 0; | |
1067 | if (count > available - pos) | |
1068 | count = available - pos; | |
1069 | res = copy_from_user(to + pos, from, count); | |
1070 | if (res == count) | |
1071 | return -EFAULT; | |
1072 | count -= res; | |
1073 | *ppos = pos + count; | |
1074 | return count; | |
1075 | } | |
12f38872 | 1076 | EXPORT_SYMBOL(simple_write_to_buffer); |
6a727b43 | 1077 | |
6d1029b5 AM |
1078 | /** |
1079 | * memory_read_from_buffer - copy data from the buffer | |
1080 | * @to: the kernel space buffer to read to | |
1081 | * @count: the maximum number of bytes to read | |
1082 | * @ppos: the current position in the buffer | |
1083 | * @from: the buffer to read from | |
1084 | * @available: the size of the buffer | |
1085 | * | |
1086 | * The memory_read_from_buffer() function reads up to @count bytes from the | |
1087 | * buffer @from at offset @ppos into the kernel space address starting at @to. | |
1088 | * | |
1089 | * On success, the number of bytes read is returned and the offset @ppos is | |
1090 | * advanced by this number, or negative value is returned on error. | |
1091 | **/ | |
93b07113 AM |
1092 | ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos, |
1093 | const void *from, size_t available) | |
1094 | { | |
1095 | loff_t pos = *ppos; | |
1096 | ||
1097 | if (pos < 0) | |
1098 | return -EINVAL; | |
1099 | if (pos >= available) | |
1100 | return 0; | |
1101 | if (count > available - pos) | |
1102 | count = available - pos; | |
1103 | memcpy(to, from + pos, count); | |
1104 | *ppos = pos + count; | |
1105 | ||
1106 | return count; | |
1107 | } | |
12f38872 | 1108 | EXPORT_SYMBOL(memory_read_from_buffer); |
93b07113 | 1109 | |
1da177e4 LT |
1110 | /* |
1111 | * Transaction based IO. | |
1112 | * The file expects a single write which triggers the transaction, and then | |
1113 | * possibly a read which collects the result - which is stored in a | |
1114 | * file-local buffer. | |
1115 | */ | |
76791ab2 IM |
1116 | |
1117 | void simple_transaction_set(struct file *file, size_t n) | |
1118 | { | |
1119 | struct simple_transaction_argresp *ar = file->private_data; | |
1120 | ||
1121 | BUG_ON(n > SIMPLE_TRANSACTION_LIMIT); | |
1122 | ||
1123 | /* | |
1124 | * The barrier ensures that ar->size will really remain zero until | |
1125 | * ar->data is ready for reading. | |
1126 | */ | |
1127 | smp_mb(); | |
1128 | ar->size = n; | |
1129 | } | |
12f38872 | 1130 | EXPORT_SYMBOL(simple_transaction_set); |
76791ab2 | 1131 | |
1da177e4 LT |
1132 | char *simple_transaction_get(struct file *file, const char __user *buf, size_t size) |
1133 | { | |
1134 | struct simple_transaction_argresp *ar; | |
1135 | static DEFINE_SPINLOCK(simple_transaction_lock); | |
1136 | ||
1137 | if (size > SIMPLE_TRANSACTION_LIMIT - 1) | |
1138 | return ERR_PTR(-EFBIG); | |
1139 | ||
1140 | ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL); | |
1141 | if (!ar) | |
1142 | return ERR_PTR(-ENOMEM); | |
1143 | ||
1144 | spin_lock(&simple_transaction_lock); | |
1145 | ||
1146 | /* only one write allowed per open */ | |
1147 | if (file->private_data) { | |
1148 | spin_unlock(&simple_transaction_lock); | |
1149 | free_page((unsigned long)ar); | |
1150 | return ERR_PTR(-EBUSY); | |
1151 | } | |
1152 | ||
1153 | file->private_data = ar; | |
1154 | ||
1155 | spin_unlock(&simple_transaction_lock); | |
1156 | ||
1157 | if (copy_from_user(ar->data, buf, size)) | |
1158 | return ERR_PTR(-EFAULT); | |
1159 | ||
1160 | return ar->data; | |
1161 | } | |
12f38872 | 1162 | EXPORT_SYMBOL(simple_transaction_get); |
1da177e4 LT |
1163 | |
1164 | ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos) | |
1165 | { | |
1166 | struct simple_transaction_argresp *ar = file->private_data; | |
1167 | ||
1168 | if (!ar) | |
1169 | return 0; | |
1170 | return simple_read_from_buffer(buf, size, pos, ar->data, ar->size); | |
1171 | } | |
12f38872 | 1172 | EXPORT_SYMBOL(simple_transaction_read); |
1da177e4 LT |
1173 | |
1174 | int simple_transaction_release(struct inode *inode, struct file *file) | |
1175 | { | |
1176 | free_page((unsigned long)file->private_data); | |
1177 | return 0; | |
1178 | } | |
12f38872 | 1179 | EXPORT_SYMBOL(simple_transaction_release); |
1da177e4 | 1180 | |
acaefc25 AB |
1181 | /* Simple attribute files */ |
1182 | ||
1183 | struct simple_attr { | |
8b88b099 CH |
1184 | int (*get)(void *, u64 *); |
1185 | int (*set)(void *, u64); | |
acaefc25 AB |
1186 | char get_buf[24]; /* enough to store a u64 and "\n\0" */ |
1187 | char set_buf[24]; | |
1188 | void *data; | |
1189 | const char *fmt; /* format for read operation */ | |
7cf34c76 | 1190 | struct mutex mutex; /* protects access to these buffers */ |
acaefc25 AB |
1191 | }; |
1192 | ||
1193 | /* simple_attr_open is called by an actual attribute open file operation | |
1194 | * to set the attribute specific access operations. */ | |
1195 | int simple_attr_open(struct inode *inode, struct file *file, | |
8b88b099 | 1196 | int (*get)(void *, u64 *), int (*set)(void *, u64), |
acaefc25 AB |
1197 | const char *fmt) |
1198 | { | |
1199 | struct simple_attr *attr; | |
1200 | ||
a65cab7d | 1201 | attr = kzalloc(sizeof(*attr), GFP_KERNEL); |
acaefc25 AB |
1202 | if (!attr) |
1203 | return -ENOMEM; | |
1204 | ||
1205 | attr->get = get; | |
1206 | attr->set = set; | |
8e18e294 | 1207 | attr->data = inode->i_private; |
acaefc25 | 1208 | attr->fmt = fmt; |
7cf34c76 | 1209 | mutex_init(&attr->mutex); |
acaefc25 AB |
1210 | |
1211 | file->private_data = attr; | |
1212 | ||
1213 | return nonseekable_open(inode, file); | |
1214 | } | |
12f38872 | 1215 | EXPORT_SYMBOL_GPL(simple_attr_open); |
acaefc25 | 1216 | |
74bedc4d | 1217 | int simple_attr_release(struct inode *inode, struct file *file) |
acaefc25 AB |
1218 | { |
1219 | kfree(file->private_data); | |
1220 | return 0; | |
1221 | } | |
12f38872 | 1222 | EXPORT_SYMBOL_GPL(simple_attr_release); /* GPL-only? This? Really? */ |
acaefc25 AB |
1223 | |
1224 | /* read from the buffer that is filled with the get function */ | |
1225 | ssize_t simple_attr_read(struct file *file, char __user *buf, | |
1226 | size_t len, loff_t *ppos) | |
1227 | { | |
1228 | struct simple_attr *attr; | |
1229 | size_t size; | |
1230 | ssize_t ret; | |
1231 | ||
1232 | attr = file->private_data; | |
1233 | ||
1234 | if (!attr->get) | |
1235 | return -EACCES; | |
1236 | ||
9261303a CH |
1237 | ret = mutex_lock_interruptible(&attr->mutex); |
1238 | if (ret) | |
1239 | return ret; | |
1240 | ||
a65cab7d EB |
1241 | if (*ppos && attr->get_buf[0]) { |
1242 | /* continued read */ | |
acaefc25 | 1243 | size = strlen(attr->get_buf); |
a65cab7d EB |
1244 | } else { |
1245 | /* first read */ | |
8b88b099 CH |
1246 | u64 val; |
1247 | ret = attr->get(attr->data, &val); | |
1248 | if (ret) | |
1249 | goto out; | |
1250 | ||
acaefc25 | 1251 | size = scnprintf(attr->get_buf, sizeof(attr->get_buf), |
8b88b099 CH |
1252 | attr->fmt, (unsigned long long)val); |
1253 | } | |
acaefc25 AB |
1254 | |
1255 | ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size); | |
8b88b099 | 1256 | out: |
7cf34c76 | 1257 | mutex_unlock(&attr->mutex); |
acaefc25 AB |
1258 | return ret; |
1259 | } | |
12f38872 | 1260 | EXPORT_SYMBOL_GPL(simple_attr_read); |
acaefc25 AB |
1261 | |
1262 | /* interpret the buffer as a number to call the set function with */ | |
2e41f274 AM |
1263 | static ssize_t simple_attr_write_xsigned(struct file *file, const char __user *buf, |
1264 | size_t len, loff_t *ppos, bool is_signed) | |
acaefc25 AB |
1265 | { |
1266 | struct simple_attr *attr; | |
488dac0c | 1267 | unsigned long long val; |
acaefc25 AB |
1268 | size_t size; |
1269 | ssize_t ret; | |
1270 | ||
1271 | attr = file->private_data; | |
acaefc25 AB |
1272 | if (!attr->set) |
1273 | return -EACCES; | |
1274 | ||
9261303a CH |
1275 | ret = mutex_lock_interruptible(&attr->mutex); |
1276 | if (ret) | |
1277 | return ret; | |
1278 | ||
acaefc25 AB |
1279 | ret = -EFAULT; |
1280 | size = min(sizeof(attr->set_buf) - 1, len); | |
1281 | if (copy_from_user(attr->set_buf, buf, size)) | |
1282 | goto out; | |
1283 | ||
acaefc25 | 1284 | attr->set_buf[size] = '\0'; |
2e41f274 AM |
1285 | if (is_signed) |
1286 | ret = kstrtoll(attr->set_buf, 0, &val); | |
1287 | else | |
1288 | ret = kstrtoull(attr->set_buf, 0, &val); | |
488dac0c YY |
1289 | if (ret) |
1290 | goto out; | |
05cc0cee WF |
1291 | ret = attr->set(attr->data, val); |
1292 | if (ret == 0) | |
1293 | ret = len; /* on success, claim we got the whole input */ | |
acaefc25 | 1294 | out: |
7cf34c76 | 1295 | mutex_unlock(&attr->mutex); |
acaefc25 AB |
1296 | return ret; |
1297 | } | |
2e41f274 AM |
1298 | |
1299 | ssize_t simple_attr_write(struct file *file, const char __user *buf, | |
1300 | size_t len, loff_t *ppos) | |
1301 | { | |
1302 | return simple_attr_write_xsigned(file, buf, len, ppos, false); | |
1303 | } | |
12f38872 | 1304 | EXPORT_SYMBOL_GPL(simple_attr_write); |
acaefc25 | 1305 | |
2e41f274 AM |
1306 | ssize_t simple_attr_write_signed(struct file *file, const char __user *buf, |
1307 | size_t len, loff_t *ppos) | |
1308 | { | |
1309 | return simple_attr_write_xsigned(file, buf, len, ppos, true); | |
1310 | } | |
1311 | EXPORT_SYMBOL_GPL(simple_attr_write_signed); | |
1312 | ||
d9e5d922 AG |
1313 | /** |
1314 | * generic_encode_ino32_fh - generic export_operations->encode_fh function | |
1315 | * @inode: the object to encode | |
1316 | * @fh: where to store the file handle fragment | |
1317 | * @max_len: maximum length to store there (in 4 byte units) | |
1318 | * @parent: parent directory inode, if wanted | |
1319 | * | |
1320 | * This generic encode_fh function assumes that the 32 inode number | |
1321 | * is suitable for locating an inode, and that the generation number | |
1322 | * can be used to check that it is still valid. It places them in the | |
1323 | * filehandle fragment where export_decode_fh expects to find them. | |
1324 | */ | |
1325 | int generic_encode_ino32_fh(struct inode *inode, __u32 *fh, int *max_len, | |
1326 | struct inode *parent) | |
1327 | { | |
1328 | struct fid *fid = (void *)fh; | |
1329 | int len = *max_len; | |
1330 | int type = FILEID_INO32_GEN; | |
1331 | ||
1332 | if (parent && (len < 4)) { | |
1333 | *max_len = 4; | |
1334 | return FILEID_INVALID; | |
1335 | } else if (len < 2) { | |
1336 | *max_len = 2; | |
1337 | return FILEID_INVALID; | |
1338 | } | |
1339 | ||
1340 | len = 2; | |
1341 | fid->i32.ino = inode->i_ino; | |
1342 | fid->i32.gen = inode->i_generation; | |
1343 | if (parent) { | |
1344 | fid->i32.parent_ino = parent->i_ino; | |
1345 | fid->i32.parent_gen = parent->i_generation; | |
1346 | len = 4; | |
1347 | type = FILEID_INO32_GEN_PARENT; | |
1348 | } | |
1349 | *max_len = len; | |
1350 | return type; | |
1351 | } | |
1352 | EXPORT_SYMBOL_GPL(generic_encode_ino32_fh); | |
1353 | ||
2596110a CH |
1354 | /** |
1355 | * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation | |
1356 | * @sb: filesystem to do the file handle conversion on | |
1357 | * @fid: file handle to convert | |
1358 | * @fh_len: length of the file handle in bytes | |
1359 | * @fh_type: type of file handle | |
1360 | * @get_inode: filesystem callback to retrieve inode | |
1361 | * | |
1362 | * This function decodes @fid as long as it has one of the well-known | |
1363 | * Linux filehandle types and calls @get_inode on it to retrieve the | |
1364 | * inode for the object specified in the file handle. | |
1365 | */ | |
1366 | struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid, | |
1367 | int fh_len, int fh_type, struct inode *(*get_inode) | |
1368 | (struct super_block *sb, u64 ino, u32 gen)) | |
1369 | { | |
1370 | struct inode *inode = NULL; | |
1371 | ||
1372 | if (fh_len < 2) | |
1373 | return NULL; | |
1374 | ||
1375 | switch (fh_type) { | |
1376 | case FILEID_INO32_GEN: | |
1377 | case FILEID_INO32_GEN_PARENT: | |
1378 | inode = get_inode(sb, fid->i32.ino, fid->i32.gen); | |
1379 | break; | |
1380 | } | |
1381 | ||
4ea3ada2 | 1382 | return d_obtain_alias(inode); |
2596110a CH |
1383 | } |
1384 | EXPORT_SYMBOL_GPL(generic_fh_to_dentry); | |
1385 | ||
1386 | /** | |
ca186830 | 1387 | * generic_fh_to_parent - generic helper for the fh_to_parent export operation |
2596110a CH |
1388 | * @sb: filesystem to do the file handle conversion on |
1389 | * @fid: file handle to convert | |
1390 | * @fh_len: length of the file handle in bytes | |
1391 | * @fh_type: type of file handle | |
1392 | * @get_inode: filesystem callback to retrieve inode | |
1393 | * | |
1394 | * This function decodes @fid as long as it has one of the well-known | |
1395 | * Linux filehandle types and calls @get_inode on it to retrieve the | |
1396 | * inode for the _parent_ object specified in the file handle if it | |
1397 | * is specified in the file handle, or NULL otherwise. | |
1398 | */ | |
1399 | struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid, | |
1400 | int fh_len, int fh_type, struct inode *(*get_inode) | |
1401 | (struct super_block *sb, u64 ino, u32 gen)) | |
1402 | { | |
1403 | struct inode *inode = NULL; | |
1404 | ||
1405 | if (fh_len <= 2) | |
1406 | return NULL; | |
1407 | ||
1408 | switch (fh_type) { | |
1409 | case FILEID_INO32_GEN_PARENT: | |
1410 | inode = get_inode(sb, fid->i32.parent_ino, | |
1411 | (fh_len > 3 ? fid->i32.parent_gen : 0)); | |
1412 | break; | |
1413 | } | |
1414 | ||
4ea3ada2 | 1415 | return d_obtain_alias(inode); |
2596110a CH |
1416 | } |
1417 | EXPORT_SYMBOL_GPL(generic_fh_to_parent); | |
1418 | ||
1b061d92 | 1419 | /** |
ac13a829 FF |
1420 | * __generic_file_fsync - generic fsync implementation for simple filesystems |
1421 | * | |
1b061d92 | 1422 | * @file: file to synchronize |
ac13a829 FF |
1423 | * @start: start offset in bytes |
1424 | * @end: end offset in bytes (inclusive) | |
1b061d92 CH |
1425 | * @datasync: only synchronize essential metadata if true |
1426 | * | |
1427 | * This is a generic implementation of the fsync method for simple | |
1428 | * filesystems which track all non-inode metadata in the buffers list | |
1429 | * hanging off the address_space structure. | |
1430 | */ | |
ac13a829 FF |
1431 | int __generic_file_fsync(struct file *file, loff_t start, loff_t end, |
1432 | int datasync) | |
d5aacad5 | 1433 | { |
7ea80859 | 1434 | struct inode *inode = file->f_mapping->host; |
d5aacad5 AV |
1435 | int err; |
1436 | int ret; | |
1437 | ||
383aa543 | 1438 | err = file_write_and_wait_range(file, start, end); |
02c24a82 JB |
1439 | if (err) |
1440 | return err; | |
1441 | ||
5955102c | 1442 | inode_lock(inode); |
d5aacad5 | 1443 | ret = sync_mapping_buffers(inode->i_mapping); |
0ae45f63 | 1444 | if (!(inode->i_state & I_DIRTY_ALL)) |
02c24a82 | 1445 | goto out; |
d5aacad5 | 1446 | if (datasync && !(inode->i_state & I_DIRTY_DATASYNC)) |
02c24a82 | 1447 | goto out; |
d5aacad5 | 1448 | |
c3765016 | 1449 | err = sync_inode_metadata(inode, 1); |
d5aacad5 AV |
1450 | if (ret == 0) |
1451 | ret = err; | |
ac13a829 | 1452 | |
02c24a82 | 1453 | out: |
5955102c | 1454 | inode_unlock(inode); |
383aa543 JL |
1455 | /* check and advance again to catch errors after syncing out buffers */ |
1456 | err = file_check_and_advance_wb_err(file); | |
1457 | if (ret == 0) | |
1458 | ret = err; | |
1459 | return ret; | |
d5aacad5 | 1460 | } |
ac13a829 FF |
1461 | EXPORT_SYMBOL(__generic_file_fsync); |
1462 | ||
1463 | /** | |
1464 | * generic_file_fsync - generic fsync implementation for simple filesystems | |
1465 | * with flush | |
1466 | * @file: file to synchronize | |
1467 | * @start: start offset in bytes | |
1468 | * @end: end offset in bytes (inclusive) | |
1469 | * @datasync: only synchronize essential metadata if true | |
1470 | * | |
1471 | */ | |
1472 | ||
1473 | int generic_file_fsync(struct file *file, loff_t start, loff_t end, | |
1474 | int datasync) | |
1475 | { | |
1476 | struct inode *inode = file->f_mapping->host; | |
1477 | int err; | |
1478 | ||
1479 | err = __generic_file_fsync(file, start, end, datasync); | |
1480 | if (err) | |
1481 | return err; | |
c6bf3f0e | 1482 | return blkdev_issue_flush(inode->i_sb->s_bdev); |
ac13a829 | 1483 | } |
1b061d92 CH |
1484 | EXPORT_SYMBOL(generic_file_fsync); |
1485 | ||
30ca22c7 PL |
1486 | /** |
1487 | * generic_check_addressable - Check addressability of file system | |
1488 | * @blocksize_bits: log of file system block size | |
1489 | * @num_blocks: number of blocks in file system | |
1490 | * | |
1491 | * Determine whether a file system with @num_blocks blocks (and a | |
1492 | * block size of 2**@blocksize_bits) is addressable by the sector_t | |
1493 | * and page cache of the system. Return 0 if so and -EFBIG otherwise. | |
1494 | */ | |
1495 | int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks) | |
1496 | { | |
1497 | u64 last_fs_block = num_blocks - 1; | |
a33f13ef | 1498 | u64 last_fs_page = |
09cbfeaf | 1499 | last_fs_block >> (PAGE_SHIFT - blocksize_bits); |
30ca22c7 PL |
1500 | |
1501 | if (unlikely(num_blocks == 0)) | |
1502 | return 0; | |
1503 | ||
09cbfeaf | 1504 | if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT)) |
30ca22c7 PL |
1505 | return -EINVAL; |
1506 | ||
a33f13ef JB |
1507 | if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) || |
1508 | (last_fs_page > (pgoff_t)(~0ULL))) { | |
30ca22c7 PL |
1509 | return -EFBIG; |
1510 | } | |
1511 | return 0; | |
1512 | } | |
1513 | EXPORT_SYMBOL(generic_check_addressable); | |
1514 | ||
1b061d92 CH |
1515 | /* |
1516 | * No-op implementation of ->fsync for in-memory filesystems. | |
1517 | */ | |
02c24a82 | 1518 | int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync) |
1b061d92 CH |
1519 | { |
1520 | return 0; | |
1521 | } | |
1b061d92 | 1522 | EXPORT_SYMBOL(noop_fsync); |
87dc800b | 1523 | |
f44c7763 DW |
1524 | ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter) |
1525 | { | |
1526 | /* | |
1527 | * iomap based filesystems support direct I/O without need for | |
1528 | * this callback. However, it still needs to be set in | |
1529 | * inode->a_ops so that open/fcntl know that direct I/O is | |
1530 | * generally supported. | |
1531 | */ | |
1532 | return -EINVAL; | |
1533 | } | |
1534 | EXPORT_SYMBOL_GPL(noop_direct_IO); | |
1535 | ||
fceef393 AV |
1536 | /* Because kfree isn't assignment-compatible with void(void*) ;-/ */ |
1537 | void kfree_link(void *p) | |
87dc800b | 1538 | { |
fceef393 | 1539 | kfree(p); |
87dc800b | 1540 | } |
fceef393 | 1541 | EXPORT_SYMBOL(kfree_link); |
6987843f | 1542 | |
6987843f AV |
1543 | struct inode *alloc_anon_inode(struct super_block *s) |
1544 | { | |
1545 | static const struct address_space_operations anon_aops = { | |
46de8b97 | 1546 | .dirty_folio = noop_dirty_folio, |
6987843f AV |
1547 | }; |
1548 | struct inode *inode = new_inode_pseudo(s); | |
1549 | ||
1550 | if (!inode) | |
1551 | return ERR_PTR(-ENOMEM); | |
1552 | ||
1553 | inode->i_ino = get_next_ino(); | |
1554 | inode->i_mapping->a_ops = &anon_aops; | |
1555 | ||
1556 | /* | |
1557 | * Mark the inode dirty from the very beginning, | |
1558 | * that way it will never be moved to the dirty | |
1559 | * list because mark_inode_dirty() will think | |
1560 | * that it already _is_ on the dirty list. | |
1561 | */ | |
1562 | inode->i_state = I_DIRTY; | |
1563 | inode->i_mode = S_IRUSR | S_IWUSR; | |
1564 | inode->i_uid = current_fsuid(); | |
1565 | inode->i_gid = current_fsgid(); | |
1566 | inode->i_flags |= S_PRIVATE; | |
077c212f | 1567 | simple_inode_init_ts(inode); |
6987843f AV |
1568 | return inode; |
1569 | } | |
1570 | EXPORT_SYMBOL(alloc_anon_inode); | |
1c994a09 JL |
1571 | |
1572 | /** | |
1573 | * simple_nosetlease - generic helper for prohibiting leases | |
1574 | * @filp: file pointer | |
1575 | * @arg: type of lease to obtain | |
1576 | * @flp: new lease supplied for insertion | |
e6f5c789 | 1577 | * @priv: private data for lm_setup operation |
1c994a09 JL |
1578 | * |
1579 | * Generic helper for filesystems that do not wish to allow leases to be set. | |
1580 | * All arguments are ignored and it just returns -EINVAL. | |
1581 | */ | |
1582 | int | |
ed5f17f6 | 1583 | simple_nosetlease(struct file *filp, int arg, struct file_lock **flp, |
e6f5c789 | 1584 | void **priv) |
1c994a09 JL |
1585 | { |
1586 | return -EINVAL; | |
1587 | } | |
1588 | EXPORT_SYMBOL(simple_nosetlease); | |
61ba64fc | 1589 | |
6ee9706a EB |
1590 | /** |
1591 | * simple_get_link - generic helper to get the target of "fast" symlinks | |
1592 | * @dentry: not used here | |
1593 | * @inode: the symlink inode | |
1594 | * @done: not used here | |
1595 | * | |
1596 | * Generic helper for filesystems to use for symlink inodes where a pointer to | |
1597 | * the symlink target is stored in ->i_link. NOTE: this isn't normally called, | |
1598 | * since as an optimization the path lookup code uses any non-NULL ->i_link | |
1599 | * directly, without calling ->get_link(). But ->get_link() still must be set, | |
1600 | * to mark the inode_operations as being for a symlink. | |
1601 | * | |
1602 | * Return: the symlink target | |
1603 | */ | |
6b255391 | 1604 | const char *simple_get_link(struct dentry *dentry, struct inode *inode, |
fceef393 | 1605 | struct delayed_call *done) |
61ba64fc | 1606 | { |
6b255391 | 1607 | return inode->i_link; |
61ba64fc | 1608 | } |
6b255391 | 1609 | EXPORT_SYMBOL(simple_get_link); |
61ba64fc AV |
1610 | |
1611 | const struct inode_operations simple_symlink_inode_operations = { | |
6b255391 | 1612 | .get_link = simple_get_link, |
61ba64fc AV |
1613 | }; |
1614 | EXPORT_SYMBOL(simple_symlink_inode_operations); | |
fbabfd0f EB |
1615 | |
1616 | /* | |
1617 | * Operations for a permanently empty directory. | |
1618 | */ | |
1619 | static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) | |
1620 | { | |
1621 | return ERR_PTR(-ENOENT); | |
1622 | } | |
1623 | ||
b74d24f7 | 1624 | static int empty_dir_getattr(struct mnt_idmap *idmap, |
549c7297 | 1625 | const struct path *path, struct kstat *stat, |
a528d35e | 1626 | u32 request_mask, unsigned int query_flags) |
fbabfd0f | 1627 | { |
a528d35e | 1628 | struct inode *inode = d_inode(path->dentry); |
0d72b928 | 1629 | generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat); |
fbabfd0f EB |
1630 | return 0; |
1631 | } | |
1632 | ||
c1632a0f | 1633 | static int empty_dir_setattr(struct mnt_idmap *idmap, |
549c7297 | 1634 | struct dentry *dentry, struct iattr *attr) |
fbabfd0f EB |
1635 | { |
1636 | return -EPERM; | |
1637 | } | |
1638 | ||
fbabfd0f EB |
1639 | static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size) |
1640 | { | |
1641 | return -EOPNOTSUPP; | |
1642 | } | |
1643 | ||
1644 | static const struct inode_operations empty_dir_inode_operations = { | |
1645 | .lookup = empty_dir_lookup, | |
1646 | .permission = generic_permission, | |
1647 | .setattr = empty_dir_setattr, | |
1648 | .getattr = empty_dir_getattr, | |
fbabfd0f EB |
1649 | .listxattr = empty_dir_listxattr, |
1650 | }; | |
1651 | ||
1652 | static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence) | |
1653 | { | |
1654 | /* An empty directory has two entries . and .. at offsets 0 and 1 */ | |
1655 | return generic_file_llseek_size(file, offset, whence, 2, 2); | |
1656 | } | |
1657 | ||
1658 | static int empty_dir_readdir(struct file *file, struct dir_context *ctx) | |
1659 | { | |
1660 | dir_emit_dots(file, ctx); | |
1661 | return 0; | |
1662 | } | |
1663 | ||
1664 | static const struct file_operations empty_dir_operations = { | |
1665 | .llseek = empty_dir_llseek, | |
1666 | .read = generic_read_dir, | |
c51da20c | 1667 | .iterate_shared = empty_dir_readdir, |
fbabfd0f EB |
1668 | .fsync = noop_fsync, |
1669 | }; | |
1670 | ||
1671 | ||
1672 | void make_empty_dir_inode(struct inode *inode) | |
1673 | { | |
1674 | set_nlink(inode, 2); | |
1675 | inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO; | |
1676 | inode->i_uid = GLOBAL_ROOT_UID; | |
1677 | inode->i_gid = GLOBAL_ROOT_GID; | |
1678 | inode->i_rdev = 0; | |
4b75de86 | 1679 | inode->i_size = 0; |
fbabfd0f EB |
1680 | inode->i_blkbits = PAGE_SHIFT; |
1681 | inode->i_blocks = 0; | |
1682 | ||
1683 | inode->i_op = &empty_dir_inode_operations; | |
f5c24438 | 1684 | inode->i_opflags &= ~IOP_XATTR; |
fbabfd0f EB |
1685 | inode->i_fop = &empty_dir_operations; |
1686 | } | |
1687 | ||
1688 | bool is_empty_dir_inode(struct inode *inode) | |
1689 | { | |
1690 | return (inode->i_fop == &empty_dir_operations) && | |
1691 | (inode->i_op == &empty_dir_inode_operations); | |
1692 | } | |
c843843e | 1693 | |
5298d4bf | 1694 | #if IS_ENABLED(CONFIG_UNICODE) |
c843843e DR |
1695 | /** |
1696 | * generic_ci_d_compare - generic d_compare implementation for casefolding filesystems | |
1697 | * @dentry: dentry whose name we are checking against | |
1698 | * @len: len of name of dentry | |
1699 | * @str: str pointer to name of dentry | |
1700 | * @name: Name to compare against | |
1701 | * | |
1702 | * Return: 0 if names match, 1 if mismatch, or -ERRNO | |
1703 | */ | |
794c43f7 EB |
1704 | static int generic_ci_d_compare(const struct dentry *dentry, unsigned int len, |
1705 | const char *str, const struct qstr *name) | |
c843843e DR |
1706 | { |
1707 | const struct dentry *parent = READ_ONCE(dentry->d_parent); | |
1708 | const struct inode *dir = READ_ONCE(parent->d_inode); | |
1709 | const struct super_block *sb = dentry->d_sb; | |
1710 | const struct unicode_map *um = sb->s_encoding; | |
1711 | struct qstr qstr = QSTR_INIT(str, len); | |
1712 | char strbuf[DNAME_INLINE_LEN]; | |
1713 | int ret; | |
1714 | ||
af494af3 | 1715 | if (!dir || !IS_CASEFOLDED(dir)) |
c843843e DR |
1716 | goto fallback; |
1717 | /* | |
1718 | * If the dentry name is stored in-line, then it may be concurrently | |
1719 | * modified by a rename. If this happens, the VFS will eventually retry | |
1720 | * the lookup, so it doesn't matter what ->d_compare() returns. | |
1721 | * However, it's unsafe to call utf8_strncasecmp() with an unstable | |
1722 | * string. Therefore, we have to copy the name into a temporary buffer. | |
1723 | */ | |
1724 | if (len <= DNAME_INLINE_LEN - 1) { | |
1725 | memcpy(strbuf, str, len); | |
1726 | strbuf[len] = 0; | |
1727 | qstr.name = strbuf; | |
1728 | /* prevent compiler from optimizing out the temporary buffer */ | |
1729 | barrier(); | |
1730 | } | |
1731 | ret = utf8_strncasecmp(um, name, &qstr); | |
1732 | if (ret >= 0) | |
1733 | return ret; | |
1734 | ||
1735 | if (sb_has_strict_encoding(sb)) | |
1736 | return -EINVAL; | |
1737 | fallback: | |
1738 | if (len != name->len) | |
1739 | return 1; | |
1740 | return !!memcmp(str, name->name, len); | |
1741 | } | |
c843843e DR |
1742 | |
1743 | /** | |
1744 | * generic_ci_d_hash - generic d_hash implementation for casefolding filesystems | |
1745 | * @dentry: dentry of the parent directory | |
1746 | * @str: qstr of name whose hash we should fill in | |
1747 | * | |
1748 | * Return: 0 if hash was successful or unchanged, and -EINVAL on error | |
1749 | */ | |
794c43f7 | 1750 | static int generic_ci_d_hash(const struct dentry *dentry, struct qstr *str) |
c843843e DR |
1751 | { |
1752 | const struct inode *dir = READ_ONCE(dentry->d_inode); | |
1753 | struct super_block *sb = dentry->d_sb; | |
1754 | const struct unicode_map *um = sb->s_encoding; | |
1755 | int ret = 0; | |
1756 | ||
af494af3 | 1757 | if (!dir || !IS_CASEFOLDED(dir)) |
c843843e DR |
1758 | return 0; |
1759 | ||
1760 | ret = utf8_casefold_hash(um, dentry, str); | |
1761 | if (ret < 0 && sb_has_strict_encoding(sb)) | |
1762 | return -EINVAL; | |
1763 | return 0; | |
1764 | } | |
608af703 DR |
1765 | |
1766 | static const struct dentry_operations generic_ci_dentry_ops = { | |
1767 | .d_hash = generic_ci_d_hash, | |
1768 | .d_compare = generic_ci_d_compare, | |
1769 | }; | |
1770 | #endif | |
1771 | ||
1772 | #ifdef CONFIG_FS_ENCRYPTION | |
1773 | static const struct dentry_operations generic_encrypted_dentry_ops = { | |
1774 | .d_revalidate = fscrypt_d_revalidate, | |
1775 | }; | |
1776 | #endif | |
1777 | ||
5298d4bf | 1778 | #if defined(CONFIG_FS_ENCRYPTION) && IS_ENABLED(CONFIG_UNICODE) |
608af703 DR |
1779 | static const struct dentry_operations generic_encrypted_ci_dentry_ops = { |
1780 | .d_hash = generic_ci_d_hash, | |
1781 | .d_compare = generic_ci_d_compare, | |
1782 | .d_revalidate = fscrypt_d_revalidate, | |
1783 | }; | |
1784 | #endif | |
1785 | ||
1786 | /** | |
1787 | * generic_set_encrypted_ci_d_ops - helper for setting d_ops for given dentry | |
1788 | * @dentry: dentry to set ops on | |
1789 | * | |
1790 | * Casefolded directories need d_hash and d_compare set, so that the dentries | |
1791 | * contained in them are handled case-insensitively. Note that these operations | |
1792 | * are needed on the parent directory rather than on the dentries in it, and | |
1793 | * while the casefolding flag can be toggled on and off on an empty directory, | |
1794 | * dentry_operations can't be changed later. As a result, if the filesystem has | |
1795 | * casefolding support enabled at all, we have to give all dentries the | |
1796 | * casefolding operations even if their inode doesn't have the casefolding flag | |
1797 | * currently (and thus the casefolding ops would be no-ops for now). | |
1798 | * | |
1799 | * Encryption works differently in that the only dentry operation it needs is | |
1800 | * d_revalidate, which it only needs on dentries that have the no-key name flag. | |
1801 | * The no-key flag can't be set "later", so we don't have to worry about that. | |
1802 | * | |
1803 | * Finally, to maximize compatibility with overlayfs (which isn't compatible | |
1804 | * with certain dentry operations) and to avoid taking an unnecessary | |
1805 | * performance hit, we use custom dentry_operations for each possible | |
1806 | * combination rather than always installing all operations. | |
1807 | */ | |
1808 | void generic_set_encrypted_ci_d_ops(struct dentry *dentry) | |
1809 | { | |
1810 | #ifdef CONFIG_FS_ENCRYPTION | |
1811 | bool needs_encrypt_ops = dentry->d_flags & DCACHE_NOKEY_NAME; | |
1812 | #endif | |
5298d4bf | 1813 | #if IS_ENABLED(CONFIG_UNICODE) |
608af703 DR |
1814 | bool needs_ci_ops = dentry->d_sb->s_encoding; |
1815 | #endif | |
5298d4bf | 1816 | #if defined(CONFIG_FS_ENCRYPTION) && IS_ENABLED(CONFIG_UNICODE) |
608af703 DR |
1817 | if (needs_encrypt_ops && needs_ci_ops) { |
1818 | d_set_d_op(dentry, &generic_encrypted_ci_dentry_ops); | |
1819 | return; | |
1820 | } | |
c843843e | 1821 | #endif |
608af703 DR |
1822 | #ifdef CONFIG_FS_ENCRYPTION |
1823 | if (needs_encrypt_ops) { | |
1824 | d_set_d_op(dentry, &generic_encrypted_dentry_ops); | |
1825 | return; | |
1826 | } | |
1827 | #endif | |
5298d4bf | 1828 | #if IS_ENABLED(CONFIG_UNICODE) |
608af703 DR |
1829 | if (needs_ci_ops) { |
1830 | d_set_d_op(dentry, &generic_ci_dentry_ops); | |
1831 | return; | |
1832 | } | |
1833 | #endif | |
1834 | } | |
1835 | EXPORT_SYMBOL(generic_set_encrypted_ci_d_ops); | |
5ca14835 AM |
1836 | |
1837 | /** | |
1838 | * inode_maybe_inc_iversion - increments i_version | |
1839 | * @inode: inode with the i_version that should be updated | |
1840 | * @force: increment the counter even if it's not necessary? | |
1841 | * | |
1842 | * Every time the inode is modified, the i_version field must be seen to have | |
1843 | * changed by any observer. | |
1844 | * | |
1845 | * If "force" is set or the QUERIED flag is set, then ensure that we increment | |
1846 | * the value, and clear the queried flag. | |
1847 | * | |
1848 | * In the common case where neither is set, then we can return "false" without | |
1849 | * updating i_version. | |
1850 | * | |
1851 | * If this function returns false, and no other metadata has changed, then we | |
1852 | * can avoid logging the metadata. | |
1853 | */ | |
1854 | bool inode_maybe_inc_iversion(struct inode *inode, bool force) | |
1855 | { | |
1856 | u64 cur, new; | |
1857 | ||
1858 | /* | |
1859 | * The i_version field is not strictly ordered with any other inode | |
1860 | * information, but the legacy inode_inc_iversion code used a spinlock | |
1861 | * to serialize increments. | |
1862 | * | |
1863 | * Here, we add full memory barriers to ensure that any de-facto | |
1864 | * ordering with other info is preserved. | |
1865 | * | |
1866 | * This barrier pairs with the barrier in inode_query_iversion() | |
1867 | */ | |
1868 | smp_mb(); | |
1869 | cur = inode_peek_iversion_raw(inode); | |
1870 | do { | |
1871 | /* If flag is clear then we needn't do anything */ | |
1872 | if (!force && !(cur & I_VERSION_QUERIED)) | |
1873 | return false; | |
1874 | ||
1875 | /* Since lowest bit is flag, add 2 to avoid it */ | |
1876 | new = (cur & ~I_VERSION_QUERIED) + I_VERSION_INCREMENT; | |
1877 | } while (!atomic64_try_cmpxchg(&inode->i_version, &cur, new)); | |
1878 | return true; | |
1879 | } | |
1880 | EXPORT_SYMBOL(inode_maybe_inc_iversion); | |
c5bc1b3f JL |
1881 | |
1882 | /** | |
1883 | * inode_query_iversion - read i_version for later use | |
1884 | * @inode: inode from which i_version should be read | |
1885 | * | |
1886 | * Read the inode i_version counter. This should be used by callers that wish | |
1887 | * to store the returned i_version for later comparison. This will guarantee | |
1888 | * that a later query of the i_version will result in a different value if | |
1889 | * anything has changed. | |
1890 | * | |
1891 | * In this implementation, we fetch the current value, set the QUERIED flag and | |
1892 | * then try to swap it into place with a cmpxchg, if it wasn't already set. If | |
1893 | * that fails, we try again with the newly fetched value from the cmpxchg. | |
1894 | */ | |
1895 | u64 inode_query_iversion(struct inode *inode) | |
1896 | { | |
1897 | u64 cur, new; | |
1898 | ||
1899 | cur = inode_peek_iversion_raw(inode); | |
1900 | do { | |
1901 | /* If flag is already set, then no need to swap */ | |
1902 | if (cur & I_VERSION_QUERIED) { | |
1903 | /* | |
1904 | * This barrier (and the implicit barrier in the | |
1905 | * cmpxchg below) pairs with the barrier in | |
1906 | * inode_maybe_inc_iversion(). | |
1907 | */ | |
1908 | smp_mb(); | |
1909 | break; | |
1910 | } | |
1911 | ||
1912 | new = cur | I_VERSION_QUERIED; | |
1913 | } while (!atomic64_try_cmpxchg(&inode->i_version, &cur, new)); | |
1914 | return cur >> I_VERSION_QUERIED_SHIFT; | |
1915 | } | |
1916 | EXPORT_SYMBOL(inode_query_iversion); | |
44fff0fa CH |
1917 | |
1918 | ssize_t direct_write_fallback(struct kiocb *iocb, struct iov_iter *iter, | |
1919 | ssize_t direct_written, ssize_t buffered_written) | |
1920 | { | |
1921 | struct address_space *mapping = iocb->ki_filp->f_mapping; | |
1922 | loff_t pos = iocb->ki_pos - buffered_written; | |
1923 | loff_t end = iocb->ki_pos - 1; | |
1924 | int err; | |
1925 | ||
1926 | /* | |
1927 | * If the buffered write fallback returned an error, we want to return | |
1928 | * the number of bytes which were written by direct I/O, or the error | |
1929 | * code if that was zero. | |
1930 | * | |
1931 | * Note that this differs from normal direct-io semantics, which will | |
1932 | * return -EFOO even if some bytes were written. | |
1933 | */ | |
1934 | if (unlikely(buffered_written < 0)) { | |
1935 | if (direct_written) | |
1936 | return direct_written; | |
1937 | return buffered_written; | |
1938 | } | |
1939 | ||
1940 | /* | |
1941 | * We need to ensure that the page cache pages are written to disk and | |
1942 | * invalidated to preserve the expected O_DIRECT semantics. | |
1943 | */ | |
1944 | err = filemap_write_and_wait_range(mapping, pos, end); | |
1945 | if (err < 0) { | |
1946 | /* | |
1947 | * We don't know how much we wrote, so just return the number of | |
1948 | * bytes which were direct-written | |
1949 | */ | |
8287474a | 1950 | iocb->ki_pos -= buffered_written; |
44fff0fa CH |
1951 | if (direct_written) |
1952 | return direct_written; | |
1953 | return err; | |
1954 | } | |
1955 | invalidate_mapping_pages(mapping, pos >> PAGE_SHIFT, end >> PAGE_SHIFT); | |
1956 | return direct_written + buffered_written; | |
1957 | } | |
1958 | EXPORT_SYMBOL_GPL(direct_write_fallback); | |
077c212f JL |
1959 | |
1960 | /** | |
1961 | * simple_inode_init_ts - initialize the timestamps for a new inode | |
1962 | * @inode: inode to be initialized | |
1963 | * | |
1964 | * When a new inode is created, most filesystems set the timestamps to the | |
1965 | * current time. Add a helper to do this. | |
1966 | */ | |
1967 | struct timespec64 simple_inode_init_ts(struct inode *inode) | |
1968 | { | |
1969 | struct timespec64 ts = inode_set_ctime_current(inode); | |
1970 | ||
1971 | inode_set_atime_to_ts(inode, ts); | |
1972 | inode_set_mtime_to_ts(inode, ts); | |
1973 | return ts; | |
1974 | } | |
1975 | EXPORT_SYMBOL(simple_inode_init_ts); |