1 // SPDX-License-Identifier: GPL-2.0-only 1 // SPDX-License-Identifier: GPL-2.0-only
2 /* 2 /*
3 * fs/libfs.c 3 * fs/libfs.c
4 * Library for filesystems writers. 4 * Library for filesystems writers.
5 */ 5 */
6 6
7 #include <linux/blkdev.h> 7 #include <linux/blkdev.h>
8 #include <linux/export.h> 8 #include <linux/export.h>
9 #include <linux/pagemap.h> 9 #include <linux/pagemap.h>
10 #include <linux/slab.h> 10 #include <linux/slab.h>
11 #include <linux/cred.h> 11 #include <linux/cred.h>
12 #include <linux/mount.h> 12 #include <linux/mount.h>
13 #include <linux/vfs.h> 13 #include <linux/vfs.h>
14 #include <linux/quotaops.h> 14 #include <linux/quotaops.h>
15 #include <linux/mutex.h> 15 #include <linux/mutex.h>
16 #include <linux/namei.h> 16 #include <linux/namei.h>
17 #include <linux/exportfs.h> 17 #include <linux/exportfs.h>
18 #include <linux/iversion.h> <<
19 #include <linux/writeback.h> 18 #include <linux/writeback.h>
20 #include <linux/buffer_head.h> /* sync_mapping 19 #include <linux/buffer_head.h> /* sync_mapping_buffers */
21 #include <linux/fs_context.h> 20 #include <linux/fs_context.h>
22 #include <linux/pseudo_fs.h> 21 #include <linux/pseudo_fs.h>
23 #include <linux/fsnotify.h> 22 #include <linux/fsnotify.h>
24 #include <linux/unicode.h> 23 #include <linux/unicode.h>
25 #include <linux/fscrypt.h> 24 #include <linux/fscrypt.h>
26 #include <linux/pidfs.h> <<
27 25
28 #include <linux/uaccess.h> 26 #include <linux/uaccess.h>
29 27
30 #include "internal.h" 28 #include "internal.h"
31 29
32 int simple_getattr(struct mnt_idmap *idmap, co !! 30 int simple_getattr(struct user_namespace *mnt_userns, const struct path *path,
33 struct kstat *stat, u32 req 31 struct kstat *stat, u32 request_mask,
34 unsigned int query_flags) 32 unsigned int query_flags)
35 { 33 {
36 struct inode *inode = d_inode(path->de 34 struct inode *inode = d_inode(path->dentry);
37 generic_fillattr(&nop_mnt_idmap, reque !! 35 generic_fillattr(&init_user_ns, inode, stat);
38 stat->blocks = inode->i_mapping->nrpag 36 stat->blocks = inode->i_mapping->nrpages << (PAGE_SHIFT - 9);
39 return 0; 37 return 0;
40 } 38 }
41 EXPORT_SYMBOL(simple_getattr); 39 EXPORT_SYMBOL(simple_getattr);
42 40
43 int simple_statfs(struct dentry *dentry, struc 41 int simple_statfs(struct dentry *dentry, struct kstatfs *buf)
44 { 42 {
45 u64 id = huge_encode_dev(dentry->d_sb- <<
46 <<
47 buf->f_fsid = u64_to_fsid(id); <<
48 buf->f_type = dentry->d_sb->s_magic; 43 buf->f_type = dentry->d_sb->s_magic;
49 buf->f_bsize = PAGE_SIZE; 44 buf->f_bsize = PAGE_SIZE;
50 buf->f_namelen = NAME_MAX; 45 buf->f_namelen = NAME_MAX;
51 return 0; 46 return 0;
52 } 47 }
53 EXPORT_SYMBOL(simple_statfs); 48 EXPORT_SYMBOL(simple_statfs);
54 49
55 /* 50 /*
56 * Retaining negative dentries for an in-memor 51 * Retaining negative dentries for an in-memory filesystem just wastes
57 * memory and lookup time: arrange for them to 52 * memory and lookup time: arrange for them to be deleted immediately.
58 */ 53 */
59 int always_delete_dentry(const struct dentry * 54 int always_delete_dentry(const struct dentry *dentry)
60 { 55 {
61 return 1; 56 return 1;
62 } 57 }
63 EXPORT_SYMBOL(always_delete_dentry); 58 EXPORT_SYMBOL(always_delete_dentry);
64 59
65 const struct dentry_operations simple_dentry_o 60 const struct dentry_operations simple_dentry_operations = {
66 .d_delete = always_delete_dentry, 61 .d_delete = always_delete_dentry,
67 }; 62 };
68 EXPORT_SYMBOL(simple_dentry_operations); 63 EXPORT_SYMBOL(simple_dentry_operations);
69 64
70 /* 65 /*
71 * Lookup the data. This is trivial - if the d 66 * Lookup the data. This is trivial - if the dentry didn't already
72 * exist, we know it is negative. Set d_op to 67 * exist, we know it is negative. Set d_op to delete negative dentries.
73 */ 68 */
74 struct dentry *simple_lookup(struct inode *dir 69 struct dentry *simple_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
75 { 70 {
76 if (dentry->d_name.len > NAME_MAX) 71 if (dentry->d_name.len > NAME_MAX)
77 return ERR_PTR(-ENAMETOOLONG); 72 return ERR_PTR(-ENAMETOOLONG);
78 if (!dentry->d_sb->s_d_op) 73 if (!dentry->d_sb->s_d_op)
79 d_set_d_op(dentry, &simple_den 74 d_set_d_op(dentry, &simple_dentry_operations);
80 d_add(dentry, NULL); 75 d_add(dentry, NULL);
81 return NULL; 76 return NULL;
82 } 77 }
83 EXPORT_SYMBOL(simple_lookup); 78 EXPORT_SYMBOL(simple_lookup);
84 79
85 int dcache_dir_open(struct inode *inode, struc 80 int dcache_dir_open(struct inode *inode, struct file *file)
86 { 81 {
87 file->private_data = d_alloc_cursor(fi 82 file->private_data = d_alloc_cursor(file->f_path.dentry);
88 83
89 return file->private_data ? 0 : -ENOME 84 return file->private_data ? 0 : -ENOMEM;
90 } 85 }
91 EXPORT_SYMBOL(dcache_dir_open); 86 EXPORT_SYMBOL(dcache_dir_open);
92 87
93 int dcache_dir_close(struct inode *inode, stru 88 int dcache_dir_close(struct inode *inode, struct file *file)
94 { 89 {
95 dput(file->private_data); 90 dput(file->private_data);
96 return 0; 91 return 0;
97 } 92 }
98 EXPORT_SYMBOL(dcache_dir_close); 93 EXPORT_SYMBOL(dcache_dir_close);
99 94
100 /* parent is locked at least shared */ 95 /* parent is locked at least shared */
101 /* 96 /*
102 * Returns an element of siblings' list. 97 * Returns an element of siblings' list.
103 * We are looking for <count>th positive after 98 * We are looking for <count>th positive after <p>; if
104 * found, dentry is grabbed and returned to ca 99 * found, dentry is grabbed and returned to caller.
105 * If no such element exists, NULL is returned 100 * If no such element exists, NULL is returned.
106 */ 101 */
107 static struct dentry *scan_positives(struct de 102 static struct dentry *scan_positives(struct dentry *cursor,
108 struct !! 103 struct list_head *p,
109 loff_t 104 loff_t count,
110 struct 105 struct dentry *last)
111 { 106 {
112 struct dentry *dentry = cursor->d_pare 107 struct dentry *dentry = cursor->d_parent, *found = NULL;
113 108
114 spin_lock(&dentry->d_lock); 109 spin_lock(&dentry->d_lock);
115 while (*p) { !! 110 while ((p = p->next) != &dentry->d_subdirs) {
116 struct dentry *d = hlist_entry !! 111 struct dentry *d = list_entry(p, struct dentry, d_child);
117 p = &d->d_sib.next; <<
118 // we must at least skip curso 112 // we must at least skip cursors, to avoid livelocks
119 if (d->d_flags & DCACHE_DENTRY 113 if (d->d_flags & DCACHE_DENTRY_CURSOR)
120 continue; 114 continue;
121 if (simple_positive(d) && !--c 115 if (simple_positive(d) && !--count) {
122 spin_lock_nested(&d->d 116 spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
123 if (simple_positive(d) 117 if (simple_positive(d))
124 found = dget_d 118 found = dget_dlock(d);
125 spin_unlock(&d->d_lock 119 spin_unlock(&d->d_lock);
126 if (likely(found)) 120 if (likely(found))
127 break; 121 break;
128 count = 1; 122 count = 1;
129 } 123 }
130 if (need_resched()) { 124 if (need_resched()) {
131 if (!hlist_unhashed(&c !! 125 list_move(&cursor->d_child, p);
132 __hlist_del(&c !! 126 p = &cursor->d_child;
133 hlist_add_behind(&curs <<
134 p = &cursor->d_sib.nex <<
135 spin_unlock(&dentry->d 127 spin_unlock(&dentry->d_lock);
136 cond_resched(); 128 cond_resched();
137 spin_lock(&dentry->d_l 129 spin_lock(&dentry->d_lock);
138 } 130 }
139 } 131 }
140 spin_unlock(&dentry->d_lock); 132 spin_unlock(&dentry->d_lock);
141 dput(last); 133 dput(last);
142 return found; 134 return found;
143 } 135 }
144 136
145 loff_t dcache_dir_lseek(struct file *file, lof 137 loff_t dcache_dir_lseek(struct file *file, loff_t offset, int whence)
146 { 138 {
147 struct dentry *dentry = file->f_path.d 139 struct dentry *dentry = file->f_path.dentry;
148 switch (whence) { 140 switch (whence) {
149 case 1: 141 case 1:
150 offset += file->f_pos; 142 offset += file->f_pos;
151 fallthrough; 143 fallthrough;
152 case 0: 144 case 0:
153 if (offset >= 0) 145 if (offset >= 0)
154 break; 146 break;
155 fallthrough; 147 fallthrough;
156 default: 148 default:
157 return -EINVAL; 149 return -EINVAL;
158 } 150 }
159 if (offset != file->f_pos) { 151 if (offset != file->f_pos) {
160 struct dentry *cursor = file-> 152 struct dentry *cursor = file->private_data;
161 struct dentry *to = NULL; 153 struct dentry *to = NULL;
162 154
163 inode_lock_shared(dentry->d_in 155 inode_lock_shared(dentry->d_inode);
164 156
165 if (offset > 2) 157 if (offset > 2)
166 to = scan_positives(cu !! 158 to = scan_positives(cursor, &dentry->d_subdirs,
167 of 159 offset - 2, NULL);
168 spin_lock(&dentry->d_lock); 160 spin_lock(&dentry->d_lock);
169 hlist_del_init(&cursor->d_sib) <<
170 if (to) 161 if (to)
171 hlist_add_behind(&curs !! 162 list_move(&cursor->d_child, &to->d_child);
>> 163 else
>> 164 list_del_init(&cursor->d_child);
172 spin_unlock(&dentry->d_lock); 165 spin_unlock(&dentry->d_lock);
173 dput(to); 166 dput(to);
174 167
175 file->f_pos = offset; 168 file->f_pos = offset;
176 169
177 inode_unlock_shared(dentry->d_ 170 inode_unlock_shared(dentry->d_inode);
178 } 171 }
179 return offset; 172 return offset;
180 } 173 }
181 EXPORT_SYMBOL(dcache_dir_lseek); 174 EXPORT_SYMBOL(dcache_dir_lseek);
182 175
>> 176 /* Relationship between i_mode and the DT_xxx types */
>> 177 static inline unsigned char dt_type(struct inode *inode)
>> 178 {
>> 179 return (inode->i_mode >> 12) & 15;
>> 180 }
>> 181
183 /* 182 /*
184 * Directory is locked and all positive dentri 183 * Directory is locked and all positive dentries in it are safe, since
185 * for ramfs-type trees they can't go away wit 184 * for ramfs-type trees they can't go away without unlink() or rmdir(),
186 * both impossible due to the lock on director 185 * both impossible due to the lock on directory.
187 */ 186 */
188 187
189 int dcache_readdir(struct file *file, struct d 188 int dcache_readdir(struct file *file, struct dir_context *ctx)
190 { 189 {
191 struct dentry *dentry = file->f_path.d 190 struct dentry *dentry = file->f_path.dentry;
192 struct dentry *cursor = file->private_ 191 struct dentry *cursor = file->private_data;
>> 192 struct list_head *anchor = &dentry->d_subdirs;
193 struct dentry *next = NULL; 193 struct dentry *next = NULL;
194 struct hlist_node **p; !! 194 struct list_head *p;
195 195
196 if (!dir_emit_dots(file, ctx)) 196 if (!dir_emit_dots(file, ctx))
197 return 0; 197 return 0;
198 198
199 if (ctx->pos == 2) 199 if (ctx->pos == 2)
200 p = &dentry->d_children.first; !! 200 p = anchor;
>> 201 else if (!list_empty(&cursor->d_child))
>> 202 p = &cursor->d_child;
201 else 203 else
202 p = &cursor->d_sib.next; !! 204 return 0;
203 205
204 while ((next = scan_positives(cursor, 206 while ((next = scan_positives(cursor, p, 1, next)) != NULL) {
205 if (!dir_emit(ctx, next->d_nam 207 if (!dir_emit(ctx, next->d_name.name, next->d_name.len,
206 d_inode(next)->i !! 208 d_inode(next)->i_ino, dt_type(d_inode(next))))
207 fs_umode_to_dtyp <<
208 break; 209 break;
209 ctx->pos++; 210 ctx->pos++;
210 p = &next->d_sib.next; !! 211 p = &next->d_child;
211 } 212 }
212 spin_lock(&dentry->d_lock); 213 spin_lock(&dentry->d_lock);
213 hlist_del_init(&cursor->d_sib); <<
214 if (next) 214 if (next)
215 hlist_add_before(&cursor->d_si !! 215 list_move_tail(&cursor->d_child, &next->d_child);
>> 216 else
>> 217 list_del_init(&cursor->d_child);
216 spin_unlock(&dentry->d_lock); 218 spin_unlock(&dentry->d_lock);
217 dput(next); 219 dput(next);
218 220
219 return 0; 221 return 0;
220 } 222 }
221 EXPORT_SYMBOL(dcache_readdir); 223 EXPORT_SYMBOL(dcache_readdir);
222 224
223 ssize_t generic_read_dir(struct file *filp, ch 225 ssize_t generic_read_dir(struct file *filp, char __user *buf, size_t siz, loff_t *ppos)
224 { 226 {
225 return -EISDIR; 227 return -EISDIR;
226 } 228 }
227 EXPORT_SYMBOL(generic_read_dir); 229 EXPORT_SYMBOL(generic_read_dir);
228 230
229 const struct file_operations simple_dir_operat 231 const struct file_operations simple_dir_operations = {
230 .open = dcache_dir_open, 232 .open = dcache_dir_open,
231 .release = dcache_dir_close, 233 .release = dcache_dir_close,
232 .llseek = dcache_dir_lseek, 234 .llseek = dcache_dir_lseek,
233 .read = generic_read_dir, 235 .read = generic_read_dir,
234 .iterate_shared = dcache_readdir, 236 .iterate_shared = dcache_readdir,
235 .fsync = noop_fsync, 237 .fsync = noop_fsync,
236 }; 238 };
237 EXPORT_SYMBOL(simple_dir_operations); 239 EXPORT_SYMBOL(simple_dir_operations);
238 240
239 const struct inode_operations simple_dir_inode 241 const struct inode_operations simple_dir_inode_operations = {
240 .lookup = simple_lookup, 242 .lookup = simple_lookup,
241 }; 243 };
242 EXPORT_SYMBOL(simple_dir_inode_operations); 244 EXPORT_SYMBOL(simple_dir_inode_operations);
243 245
244 /* 0 is '.', 1 is '..', so always start with o <<
245 enum { <<
246 DIR_OFFSET_MIN = 2, <<
247 }; <<
248 <<
249 static void offset_set(struct dentry *dentry, <<
250 { <<
251 dentry->d_fsdata = (void *)offset; <<
252 } <<
253 <<
254 static long dentry2offset(struct dentry *dentr <<
255 { <<
256 return (long)dentry->d_fsdata; <<
257 } <<
258 <<
259 static struct lock_class_key simple_offset_loc <<
260 <<
261 /** <<
262 * simple_offset_init - initialize an offset_c <<
263 * @octx: directory offset map to be initializ <<
264 * <<
265 */ <<
266 void simple_offset_init(struct offset_ctx *oct <<
267 { <<
268 mt_init_flags(&octx->mt, MT_FLAGS_ALLO <<
269 lockdep_set_class(&octx->mt.ma_lock, & <<
270 octx->next_offset = DIR_OFFSET_MIN; <<
271 } <<
272 <<
273 /** <<
274 * simple_offset_add - Add an entry to a direc <<
275 * @octx: directory offset ctx to be updated <<
276 * @dentry: new dentry being added <<
277 * <<
278 * Returns zero on success. @octx and the dent <<
279 * Otherwise, a negative errno value is return <<
280 */ <<
281 int simple_offset_add(struct offset_ctx *octx, <<
282 { <<
283 unsigned long offset; <<
284 int ret; <<
285 <<
286 if (dentry2offset(dentry) != 0) <<
287 return -EBUSY; <<
288 <<
289 ret = mtree_alloc_cyclic(&octx->mt, &o <<
290 LONG_MAX, &oc <<
291 if (ret < 0) <<
292 return ret; <<
293 <<
294 offset_set(dentry, offset); <<
295 return 0; <<
296 } <<
297 <<
298 static int simple_offset_replace(struct offset <<
299 long offset) <<
300 { <<
301 int ret; <<
302 <<
303 ret = mtree_store(&octx->mt, offset, d <<
304 if (ret) <<
305 return ret; <<
306 offset_set(dentry, offset); <<
307 return 0; <<
308 } <<
309 <<
310 /** <<
311 * simple_offset_remove - Remove an entry to a <<
312 * @octx: directory offset ctx to be updated <<
313 * @dentry: dentry being removed <<
314 * <<
315 */ <<
316 void simple_offset_remove(struct offset_ctx *o <<
317 { <<
318 long offset; <<
319 <<
320 offset = dentry2offset(dentry); <<
321 if (offset == 0) <<
322 return; <<
323 <<
324 mtree_erase(&octx->mt, offset); <<
325 offset_set(dentry, 0); <<
326 } <<
327 <<
328 /** <<
329 * simple_offset_empty - Check if a dentry can <<
330 * @dentry: dentry to be tested <<
331 * <<
332 * Returns 0 if @dentry is a non-empty directo <<
333 */ <<
334 int simple_offset_empty(struct dentry *dentry) <<
335 { <<
336 struct inode *inode = d_inode(dentry); <<
337 struct offset_ctx *octx; <<
338 struct dentry *child; <<
339 unsigned long index; <<
340 int ret = 1; <<
341 <<
342 if (!inode || !S_ISDIR(inode->i_mode)) <<
343 return ret; <<
344 <<
345 index = DIR_OFFSET_MIN; <<
346 octx = inode->i_op->get_offset_ctx(ino <<
347 mt_for_each(&octx->mt, child, index, L <<
348 spin_lock(&child->d_lock); <<
349 if (simple_positive(child)) { <<
350 spin_unlock(&child->d_ <<
351 ret = 0; <<
352 break; <<
353 } <<
354 spin_unlock(&child->d_lock); <<
355 } <<
356 <<
357 return ret; <<
358 } <<
359 <<
360 /** <<
361 * simple_offset_rename - handle directory off <<
362 * @old_dir: parent directory of source entry <<
363 * @old_dentry: dentry of source entry <<
364 * @new_dir: parent_directory of destination e <<
365 * @new_dentry: dentry of destination <<
366 * <<
367 * Caller provides appropriate serialization. <<
368 * <<
369 * User space expects the directory offset val <<
370 * (new) directory entry to be unchanged after <<
371 * <<
372 * Returns zero on success, a negative errno v <<
373 */ <<
374 int simple_offset_rename(struct inode *old_dir <<
375 struct inode *new_dir <<
376 { <<
377 struct offset_ctx *old_ctx = old_dir-> <<
378 struct offset_ctx *new_ctx = new_dir-> <<
379 long new_offset = dentry2offset(new_de <<
380 <<
381 simple_offset_remove(old_ctx, old_dent <<
382 <<
383 if (new_offset) { <<
384 offset_set(new_dentry, 0); <<
385 return simple_offset_replace(n <<
386 } <<
387 return simple_offset_add(new_ctx, old_ <<
388 } <<
389 <<
390 /** <<
391 * simple_offset_rename_exchange - exchange re <<
392 * @old_dir: parent of dentry being moved <<
393 * @old_dentry: dentry being moved <<
394 * @new_dir: destination parent <<
395 * @new_dentry: destination dentry <<
396 * <<
397 * This API preserves the directory offset val <<
398 * appropriate serialization. <<
399 * <<
400 * Returns zero on success. Otherwise a negati <<
401 * rename is rolled back. <<
402 */ <<
403 int simple_offset_rename_exchange(struct inode <<
404 struct dentr <<
405 struct inode <<
406 struct dentr <<
407 { <<
408 struct offset_ctx *old_ctx = old_dir-> <<
409 struct offset_ctx *new_ctx = new_dir-> <<
410 long old_index = dentry2offset(old_den <<
411 long new_index = dentry2offset(new_den <<
412 int ret; <<
413 <<
414 simple_offset_remove(old_ctx, old_dent <<
415 simple_offset_remove(new_ctx, new_dent <<
416 <<
417 ret = simple_offset_replace(new_ctx, o <<
418 if (ret) <<
419 goto out_restore; <<
420 <<
421 ret = simple_offset_replace(old_ctx, n <<
422 if (ret) { <<
423 simple_offset_remove(new_ctx, <<
424 goto out_restore; <<
425 } <<
426 <<
427 ret = simple_rename_exchange(old_dir, <<
428 if (ret) { <<
429 simple_offset_remove(new_ctx, <<
430 simple_offset_remove(old_ctx, <<
431 goto out_restore; <<
432 } <<
433 return 0; <<
434 <<
435 out_restore: <<
436 (void)simple_offset_replace(old_ctx, o <<
437 (void)simple_offset_replace(new_ctx, n <<
438 return ret; <<
439 } <<
440 <<
441 /** <<
442 * simple_offset_destroy - Release offset map <<
443 * @octx: directory offset ctx that is about t <<
444 * <<
445 * During fs teardown (eg. umount), a director <<
446 * contain entries. xa_destroy() cleans out an <<
447 */ <<
448 void simple_offset_destroy(struct offset_ctx * <<
449 { <<
450 mtree_destroy(&octx->mt); <<
451 } <<
452 <<
453 static int offset_dir_open(struct inode *inode <<
454 { <<
455 struct offset_ctx *ctx = inode->i_op-> <<
456 <<
457 file->private_data = (void *)ctx->next <<
458 return 0; <<
459 } <<
460 <<
461 /** <<
462 * offset_dir_llseek - Advance the read positi <<
463 * @file: an open directory whose position is <<
464 * @offset: a byte offset <<
465 * @whence: enumerator describing the starting <<
466 * <<
467 * SEEK_END, SEEK_DATA, and SEEK_HOLE are not <<
468 * <<
469 * Returns the updated read position if succes <<
470 * negative errno is returned and the read pos <<
471 */ <<
472 static loff_t offset_dir_llseek(struct file *f <<
473 { <<
474 struct inode *inode = file->f_inode; <<
475 struct offset_ctx *ctx = inode->i_op-> <<
476 <<
477 switch (whence) { <<
478 case SEEK_CUR: <<
479 offset += file->f_pos; <<
480 fallthrough; <<
481 case SEEK_SET: <<
482 if (offset >= 0) <<
483 break; <<
484 fallthrough; <<
485 default: <<
486 return -EINVAL; <<
487 } <<
488 <<
489 /* In this case, ->private_data is pro <<
490 if (!offset) <<
491 file->private_data = (void *)c <<
492 return vfs_setpos(file, offset, LONG_M <<
493 } <<
494 <<
495 static struct dentry *offset_find_next(struct <<
496 { <<
497 MA_STATE(mas, &octx->mt, offset, offse <<
498 struct dentry *child, *found = NULL; <<
499 <<
500 rcu_read_lock(); <<
501 child = mas_find(&mas, LONG_MAX); <<
502 if (!child) <<
503 goto out; <<
504 spin_lock(&child->d_lock); <<
505 if (simple_positive(child)) <<
506 found = dget_dlock(child); <<
507 spin_unlock(&child->d_lock); <<
508 out: <<
509 rcu_read_unlock(); <<
510 return found; <<
511 } <<
512 <<
513 static bool offset_dir_emit(struct dir_context <<
514 { <<
515 struct inode *inode = d_inode(dentry); <<
516 long offset = dentry2offset(dentry); <<
517 <<
518 return ctx->actor(ctx, dentry->d_name. <<
519 inode->i_ino, fs_umo <<
520 } <<
521 <<
522 static void offset_iterate_dir(struct inode *i <<
523 { <<
524 struct offset_ctx *octx = inode->i_op- <<
525 struct dentry *dentry; <<
526 <<
527 while (true) { <<
528 dentry = offset_find_next(octx <<
529 if (!dentry) <<
530 return; <<
531 <<
532 if (dentry2offset(dentry) >= l <<
533 dput(dentry); <<
534 return; <<
535 } <<
536 <<
537 if (!offset_dir_emit(ctx, dent <<
538 dput(dentry); <<
539 return; <<
540 } <<
541 <<
542 ctx->pos = dentry2offset(dentr <<
543 dput(dentry); <<
544 } <<
545 } <<
546 <<
547 /** <<
548 * offset_readdir - Emit entries starting at o <<
549 * @file: an open directory to iterate over <<
550 * @ctx: directory iteration context <<
551 * <<
552 * Caller must hold @file's i_rwsem to prevent <<
553 * entries during this call. <<
554 * <<
555 * On entry, @ctx->pos contains an offset that <<
556 * to be read from the directory. <<
557 * <<
558 * The operation continues until there are no <<
559 * until the ctx->actor indicates there is no <<
560 * output buffer. <<
561 * <<
562 * On return, @ctx->pos contains an offset tha <<
563 * in this directory when offset_readdir() is <<
564 * <<
565 * Return values: <<
566 * %0 - Complete <<
567 */ <<
568 static int offset_readdir(struct file *file, s <<
569 { <<
570 struct dentry *dir = file->f_path.dent <<
571 long last_index = (long)file->private_ <<
572 <<
573 lockdep_assert_held(&d_inode(dir)->i_r <<
574 <<
575 if (!dir_emit_dots(file, ctx)) <<
576 return 0; <<
577 <<
578 offset_iterate_dir(d_inode(dir), ctx, <<
579 return 0; <<
580 } <<
581 <<
582 const struct file_operations simple_offset_dir <<
583 .open = offset_dir_open, <<
584 .llseek = offset_dir_llseek, <<
585 .iterate_shared = offset_readdir, <<
586 .read = generic_read_dir, <<
587 .fsync = noop_fsync, <<
588 }; <<
589 <<
590 static struct dentry *find_next_child(struct d 246 static struct dentry *find_next_child(struct dentry *parent, struct dentry *prev)
591 { 247 {
592 struct dentry *child = NULL, *d; !! 248 struct dentry *child = NULL;
>> 249 struct list_head *p = prev ? &prev->d_child : &parent->d_subdirs;
593 250
594 spin_lock(&parent->d_lock); 251 spin_lock(&parent->d_lock);
595 d = prev ? d_next_sibling(prev) : d_fi !! 252 while ((p = p->next) != &parent->d_subdirs) {
596 hlist_for_each_entry_from(d, d_sib) { !! 253 struct dentry *d = container_of(p, struct dentry, d_child);
597 if (simple_positive(d)) { 254 if (simple_positive(d)) {
598 spin_lock_nested(&d->d 255 spin_lock_nested(&d->d_lock, DENTRY_D_LOCK_NESTED);
599 if (simple_positive(d) 256 if (simple_positive(d))
600 child = dget_d 257 child = dget_dlock(d);
601 spin_unlock(&d->d_lock 258 spin_unlock(&d->d_lock);
602 if (likely(child)) 259 if (likely(child))
603 break; 260 break;
604 } 261 }
605 } 262 }
606 spin_unlock(&parent->d_lock); 263 spin_unlock(&parent->d_lock);
607 dput(prev); 264 dput(prev);
608 return child; 265 return child;
609 } 266 }
610 267
611 void simple_recursive_removal(struct dentry *d 268 void simple_recursive_removal(struct dentry *dentry,
612 void (*callback) 269 void (*callback)(struct dentry *))
613 { 270 {
614 struct dentry *this = dget(dentry); 271 struct dentry *this = dget(dentry);
615 while (true) { 272 while (true) {
616 struct dentry *victim = NULL, 273 struct dentry *victim = NULL, *child;
617 struct inode *inode = this->d_ 274 struct inode *inode = this->d_inode;
618 275
619 inode_lock(inode); 276 inode_lock(inode);
620 if (d_is_dir(this)) 277 if (d_is_dir(this))
621 inode->i_flags |= S_DE 278 inode->i_flags |= S_DEAD;
622 while ((child = find_next_chil 279 while ((child = find_next_child(this, victim)) == NULL) {
623 // kill and ascend 280 // kill and ascend
624 // update metadata whi 281 // update metadata while it's still locked
625 inode_set_ctime_curren !! 282 inode->i_ctime = current_time(inode);
626 clear_nlink(inode); 283 clear_nlink(inode);
627 inode_unlock(inode); 284 inode_unlock(inode);
628 victim = this; 285 victim = this;
629 this = this->d_parent; 286 this = this->d_parent;
630 inode = this->d_inode; 287 inode = this->d_inode;
631 inode_lock(inode); 288 inode_lock(inode);
632 if (simple_positive(vi 289 if (simple_positive(victim)) {
633 d_invalidate(v 290 d_invalidate(victim); // avoid lost mounts
634 if (d_is_dir(v 291 if (d_is_dir(victim))
635 fsnoti 292 fsnotify_rmdir(inode, victim);
636 else 293 else
637 fsnoti 294 fsnotify_unlink(inode, victim);
638 if (callback) 295 if (callback)
639 callba 296 callback(victim);
640 dput(victim); 297 dput(victim); // unpin it
641 } 298 }
642 if (victim == dentry) 299 if (victim == dentry) {
643 inode_set_mtim !! 300 inode->i_ctime = inode->i_mtime =
644 !! 301 current_time(inode);
645 if (d_is_dir(d 302 if (d_is_dir(dentry))
646 drop_n 303 drop_nlink(inode);
647 inode_unlock(i 304 inode_unlock(inode);
648 dput(dentry); 305 dput(dentry);
649 return; 306 return;
650 } 307 }
651 } 308 }
652 inode_unlock(inode); 309 inode_unlock(inode);
653 this = child; 310 this = child;
654 } 311 }
655 } 312 }
656 EXPORT_SYMBOL(simple_recursive_removal); 313 EXPORT_SYMBOL(simple_recursive_removal);
657 314
658 static const struct super_operations simple_su 315 static const struct super_operations simple_super_operations = {
659 .statfs = simple_statfs, 316 .statfs = simple_statfs,
660 }; 317 };
661 318
662 static int pseudo_fs_fill_super(struct super_b 319 static int pseudo_fs_fill_super(struct super_block *s, struct fs_context *fc)
663 { 320 {
664 struct pseudo_fs_context *ctx = fc->fs 321 struct pseudo_fs_context *ctx = fc->fs_private;
665 struct inode *root; 322 struct inode *root;
666 323
667 s->s_maxbytes = MAX_LFS_FILESIZE; 324 s->s_maxbytes = MAX_LFS_FILESIZE;
668 s->s_blocksize = PAGE_SIZE; 325 s->s_blocksize = PAGE_SIZE;
669 s->s_blocksize_bits = PAGE_SHIFT; 326 s->s_blocksize_bits = PAGE_SHIFT;
670 s->s_magic = ctx->magic; 327 s->s_magic = ctx->magic;
671 s->s_op = ctx->ops ?: &simple_super_op 328 s->s_op = ctx->ops ?: &simple_super_operations;
672 s->s_xattr = ctx->xattr; 329 s->s_xattr = ctx->xattr;
673 s->s_time_gran = 1; 330 s->s_time_gran = 1;
674 root = new_inode(s); 331 root = new_inode(s);
675 if (!root) 332 if (!root)
676 return -ENOMEM; 333 return -ENOMEM;
677 334
678 /* 335 /*
679 * since this is the first inode, make 336 * since this is the first inode, make it number 1. New inodes created
680 * after this must take care not to co 337 * after this must take care not to collide with it (by passing
681 * max_reserved of 1 to iunique). 338 * max_reserved of 1 to iunique).
682 */ 339 */
683 root->i_ino = 1; 340 root->i_ino = 1;
684 root->i_mode = S_IFDIR | S_IRUSR | S_I 341 root->i_mode = S_IFDIR | S_IRUSR | S_IWUSR;
685 simple_inode_init_ts(root); !! 342 root->i_atime = root->i_mtime = root->i_ctime = current_time(root);
686 s->s_root = d_make_root(root); 343 s->s_root = d_make_root(root);
687 if (!s->s_root) 344 if (!s->s_root)
688 return -ENOMEM; 345 return -ENOMEM;
689 s->s_d_op = ctx->dops; 346 s->s_d_op = ctx->dops;
690 return 0; 347 return 0;
691 } 348 }
692 349
693 static int pseudo_fs_get_tree(struct fs_contex 350 static int pseudo_fs_get_tree(struct fs_context *fc)
694 { 351 {
695 return get_tree_nodev(fc, pseudo_fs_fi 352 return get_tree_nodev(fc, pseudo_fs_fill_super);
696 } 353 }
697 354
698 static void pseudo_fs_free(struct fs_context * 355 static void pseudo_fs_free(struct fs_context *fc)
699 { 356 {
700 kfree(fc->fs_private); 357 kfree(fc->fs_private);
701 } 358 }
702 359
703 static const struct fs_context_operations pseu 360 static const struct fs_context_operations pseudo_fs_context_ops = {
704 .free = pseudo_fs_free, 361 .free = pseudo_fs_free,
705 .get_tree = pseudo_fs_get_tree, 362 .get_tree = pseudo_fs_get_tree,
706 }; 363 };
707 364
708 /* 365 /*
709 * Common helper for pseudo-filesystems (sockf 366 * Common helper for pseudo-filesystems (sockfs, pipefs, bdev - stuff that
710 * will never be mountable) 367 * will never be mountable)
711 */ 368 */
712 struct pseudo_fs_context *init_pseudo(struct f 369 struct pseudo_fs_context *init_pseudo(struct fs_context *fc,
713 unsign 370 unsigned long magic)
714 { 371 {
715 struct pseudo_fs_context *ctx; 372 struct pseudo_fs_context *ctx;
716 373
717 ctx = kzalloc(sizeof(struct pseudo_fs_ 374 ctx = kzalloc(sizeof(struct pseudo_fs_context), GFP_KERNEL);
718 if (likely(ctx)) { 375 if (likely(ctx)) {
719 ctx->magic = magic; 376 ctx->magic = magic;
720 fc->fs_private = ctx; 377 fc->fs_private = ctx;
721 fc->ops = &pseudo_fs_context_o 378 fc->ops = &pseudo_fs_context_ops;
722 fc->sb_flags |= SB_NOUSER; 379 fc->sb_flags |= SB_NOUSER;
723 fc->global = true; 380 fc->global = true;
724 } 381 }
725 return ctx; 382 return ctx;
726 } 383 }
727 EXPORT_SYMBOL(init_pseudo); 384 EXPORT_SYMBOL(init_pseudo);
728 385
729 int simple_open(struct inode *inode, struct fi 386 int simple_open(struct inode *inode, struct file *file)
730 { 387 {
731 if (inode->i_private) 388 if (inode->i_private)
732 file->private_data = inode->i_ 389 file->private_data = inode->i_private;
733 return 0; 390 return 0;
734 } 391 }
735 EXPORT_SYMBOL(simple_open); 392 EXPORT_SYMBOL(simple_open);
736 393
737 int simple_link(struct dentry *old_dentry, str 394 int simple_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry)
738 { 395 {
739 struct inode *inode = d_inode(old_dent 396 struct inode *inode = d_inode(old_dentry);
740 397
741 inode_set_mtime_to_ts(dir, !! 398 inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
742 inode_set_ctime_ <<
743 inc_nlink(inode); 399 inc_nlink(inode);
744 ihold(inode); 400 ihold(inode);
745 dget(dentry); 401 dget(dentry);
746 d_instantiate(dentry, inode); 402 d_instantiate(dentry, inode);
747 return 0; 403 return 0;
748 } 404 }
749 EXPORT_SYMBOL(simple_link); 405 EXPORT_SYMBOL(simple_link);
750 406
751 int simple_empty(struct dentry *dentry) 407 int simple_empty(struct dentry *dentry)
752 { 408 {
753 struct dentry *child; 409 struct dentry *child;
754 int ret = 0; 410 int ret = 0;
755 411
756 spin_lock(&dentry->d_lock); 412 spin_lock(&dentry->d_lock);
757 hlist_for_each_entry(child, &dentry->d !! 413 list_for_each_entry(child, &dentry->d_subdirs, d_child) {
758 spin_lock_nested(&child->d_loc 414 spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED);
759 if (simple_positive(child)) { 415 if (simple_positive(child)) {
760 spin_unlock(&child->d_ 416 spin_unlock(&child->d_lock);
761 goto out; 417 goto out;
762 } 418 }
763 spin_unlock(&child->d_lock); 419 spin_unlock(&child->d_lock);
764 } 420 }
765 ret = 1; 421 ret = 1;
766 out: 422 out:
767 spin_unlock(&dentry->d_lock); 423 spin_unlock(&dentry->d_lock);
768 return ret; 424 return ret;
769 } 425 }
770 EXPORT_SYMBOL(simple_empty); 426 EXPORT_SYMBOL(simple_empty);
771 427
772 int simple_unlink(struct inode *dir, struct de 428 int simple_unlink(struct inode *dir, struct dentry *dentry)
773 { 429 {
774 struct inode *inode = d_inode(dentry); 430 struct inode *inode = d_inode(dentry);
775 431
776 inode_set_mtime_to_ts(dir, !! 432 inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode);
777 inode_set_ctime_ <<
778 drop_nlink(inode); 433 drop_nlink(inode);
779 dput(dentry); 434 dput(dentry);
780 return 0; 435 return 0;
781 } 436 }
782 EXPORT_SYMBOL(simple_unlink); 437 EXPORT_SYMBOL(simple_unlink);
783 438
784 int simple_rmdir(struct inode *dir, struct den 439 int simple_rmdir(struct inode *dir, struct dentry *dentry)
785 { 440 {
786 if (!simple_empty(dentry)) 441 if (!simple_empty(dentry))
787 return -ENOTEMPTY; 442 return -ENOTEMPTY;
788 443
789 drop_nlink(d_inode(dentry)); 444 drop_nlink(d_inode(dentry));
790 simple_unlink(dir, dentry); 445 simple_unlink(dir, dentry);
791 drop_nlink(dir); 446 drop_nlink(dir);
792 return 0; 447 return 0;
793 } 448 }
794 EXPORT_SYMBOL(simple_rmdir); 449 EXPORT_SYMBOL(simple_rmdir);
795 450
796 /** <<
797 * simple_rename_timestamp - update the variou <<
798 * @old_dir: old parent directory <<
799 * @old_dentry: dentry that is being renamed <<
800 * @new_dir: new parent directory <<
801 * @new_dentry: target for rename <<
802 * <<
803 * POSIX mandates that the old and new parent <<
804 * mtime updated, and that inodes of @old_dent <<
805 * their ctime updated. <<
806 */ <<
807 void simple_rename_timestamp(struct inode *old <<
808 struct inode *new <<
809 { <<
810 struct inode *newino = d_inode(new_den <<
811 <<
812 inode_set_mtime_to_ts(old_dir, inode_s <<
813 if (new_dir != old_dir) <<
814 inode_set_mtime_to_ts(new_dir, <<
815 inode_se <<
816 inode_set_ctime_current(d_inode(old_de <<
817 if (newino) <<
818 inode_set_ctime_current(newino <<
819 } <<
820 EXPORT_SYMBOL_GPL(simple_rename_timestamp); <<
821 <<
822 int simple_rename_exchange(struct inode *old_d 451 int simple_rename_exchange(struct inode *old_dir, struct dentry *old_dentry,
823 struct inode *new_d 452 struct inode *new_dir, struct dentry *new_dentry)
824 { 453 {
825 bool old_is_dir = d_is_dir(old_dentry) 454 bool old_is_dir = d_is_dir(old_dentry);
826 bool new_is_dir = d_is_dir(new_dentry) 455 bool new_is_dir = d_is_dir(new_dentry);
827 456
828 if (old_dir != new_dir && old_is_dir ! 457 if (old_dir != new_dir && old_is_dir != new_is_dir) {
829 if (old_is_dir) { 458 if (old_is_dir) {
830 drop_nlink(old_dir); 459 drop_nlink(old_dir);
831 inc_nlink(new_dir); 460 inc_nlink(new_dir);
832 } else { 461 } else {
833 drop_nlink(new_dir); 462 drop_nlink(new_dir);
834 inc_nlink(old_dir); 463 inc_nlink(old_dir);
835 } 464 }
836 } 465 }
837 simple_rename_timestamp(old_dir, old_d !! 466 old_dir->i_ctime = old_dir->i_mtime =
>> 467 new_dir->i_ctime = new_dir->i_mtime =
>> 468 d_inode(old_dentry)->i_ctime =
>> 469 d_inode(new_dentry)->i_ctime = current_time(old_dir);
>> 470
838 return 0; 471 return 0;
839 } 472 }
840 EXPORT_SYMBOL_GPL(simple_rename_exchange); 473 EXPORT_SYMBOL_GPL(simple_rename_exchange);
841 474
842 int simple_rename(struct mnt_idmap *idmap, str !! 475 int simple_rename(struct user_namespace *mnt_userns, struct inode *old_dir,
843 struct dentry *old_dentry, s 476 struct dentry *old_dentry, struct inode *new_dir,
844 struct dentry *new_dentry, u 477 struct dentry *new_dentry, unsigned int flags)
845 { 478 {
>> 479 struct inode *inode = d_inode(old_dentry);
846 int they_are_dirs = d_is_dir(old_dentr 480 int they_are_dirs = d_is_dir(old_dentry);
847 481
848 if (flags & ~(RENAME_NOREPLACE | RENAM 482 if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE))
849 return -EINVAL; 483 return -EINVAL;
850 484
851 if (flags & RENAME_EXCHANGE) 485 if (flags & RENAME_EXCHANGE)
852 return simple_rename_exchange( 486 return simple_rename_exchange(old_dir, old_dentry, new_dir, new_dentry);
853 487
854 if (!simple_empty(new_dentry)) 488 if (!simple_empty(new_dentry))
855 return -ENOTEMPTY; 489 return -ENOTEMPTY;
856 490
857 if (d_really_is_positive(new_dentry)) 491 if (d_really_is_positive(new_dentry)) {
858 simple_unlink(new_dir, new_den 492 simple_unlink(new_dir, new_dentry);
859 if (they_are_dirs) { 493 if (they_are_dirs) {
860 drop_nlink(d_inode(new 494 drop_nlink(d_inode(new_dentry));
861 drop_nlink(old_dir); 495 drop_nlink(old_dir);
862 } 496 }
863 } else if (they_are_dirs) { 497 } else if (they_are_dirs) {
864 drop_nlink(old_dir); 498 drop_nlink(old_dir);
865 inc_nlink(new_dir); 499 inc_nlink(new_dir);
866 } 500 }
867 501
868 simple_rename_timestamp(old_dir, old_d !! 502 old_dir->i_ctime = old_dir->i_mtime = new_dir->i_ctime =
>> 503 new_dir->i_mtime = inode->i_ctime = current_time(old_dir);
>> 504
869 return 0; 505 return 0;
870 } 506 }
871 EXPORT_SYMBOL(simple_rename); 507 EXPORT_SYMBOL(simple_rename);
872 508
873 /** 509 /**
874 * simple_setattr - setattr for simple filesys 510 * simple_setattr - setattr for simple filesystem
875 * @idmap: idmap of the target mount !! 511 * @mnt_userns: user namespace of the target mount
876 * @dentry: dentry 512 * @dentry: dentry
877 * @iattr: iattr structure 513 * @iattr: iattr structure
878 * 514 *
879 * Returns 0 on success, -error on failure. 515 * Returns 0 on success, -error on failure.
880 * 516 *
881 * simple_setattr is a simple ->setattr implem 517 * simple_setattr is a simple ->setattr implementation without a proper
882 * implementation of size changes. 518 * implementation of size changes.
883 * 519 *
884 * It can either be used for in-memory filesys 520 * It can either be used for in-memory filesystems or special files
885 * on simple regular filesystems. Anything th 521 * on simple regular filesystems. Anything that needs to change on-disk
886 * or wire state on size changes needs its own 522 * or wire state on size changes needs its own setattr method.
887 */ 523 */
888 int simple_setattr(struct mnt_idmap *idmap, st !! 524 int simple_setattr(struct user_namespace *mnt_userns, struct dentry *dentry,
889 struct iattr *iattr) 525 struct iattr *iattr)
890 { 526 {
891 struct inode *inode = d_inode(dentry); 527 struct inode *inode = d_inode(dentry);
892 int error; 528 int error;
893 529
894 error = setattr_prepare(idmap, dentry, !! 530 error = setattr_prepare(mnt_userns, dentry, iattr);
895 if (error) 531 if (error)
896 return error; 532 return error;
897 533
898 if (iattr->ia_valid & ATTR_SIZE) 534 if (iattr->ia_valid & ATTR_SIZE)
899 truncate_setsize(inode, iattr- 535 truncate_setsize(inode, iattr->ia_size);
900 setattr_copy(idmap, inode, iattr); !! 536 setattr_copy(mnt_userns, inode, iattr);
901 mark_inode_dirty(inode); 537 mark_inode_dirty(inode);
902 return 0; 538 return 0;
903 } 539 }
904 EXPORT_SYMBOL(simple_setattr); 540 EXPORT_SYMBOL(simple_setattr);
905 541
906 static int simple_read_folio(struct file *file !! 542 static int simple_readpage(struct file *file, struct page *page)
907 { 543 {
908 folio_zero_range(folio, 0, folio_size( !! 544 clear_highpage(page);
909 flush_dcache_folio(folio); !! 545 flush_dcache_page(page);
910 folio_mark_uptodate(folio); !! 546 SetPageUptodate(page);
911 folio_unlock(folio); !! 547 unlock_page(page);
912 return 0; 548 return 0;
913 } 549 }
914 550
915 int simple_write_begin(struct file *file, stru 551 int simple_write_begin(struct file *file, struct address_space *mapping,
916 loff_t pos, unsigned l !! 552 loff_t pos, unsigned len, unsigned flags,
917 struct folio **foliop, !! 553 struct page **pagep, void **fsdata)
918 { 554 {
919 struct folio *folio; !! 555 struct page *page;
>> 556 pgoff_t index;
>> 557
>> 558 index = pos >> PAGE_SHIFT;
920 559
921 folio = __filemap_get_folio(mapping, p !! 560 page = grab_cache_page_write_begin(mapping, index, flags);
922 mapping_gfp_mask(mappi !! 561 if (!page)
923 if (IS_ERR(folio)) !! 562 return -ENOMEM;
924 return PTR_ERR(folio); <<
925 563
926 *foliop = folio; !! 564 *pagep = page;
927 565
928 if (!folio_test_uptodate(folio) && (le !! 566 if (!PageUptodate(page) && (len != PAGE_SIZE)) {
929 size_t from = offset_in_folio( !! 567 unsigned from = pos & (PAGE_SIZE - 1);
930 568
931 folio_zero_segments(folio, 0, !! 569 zero_user_segments(page, 0, from, from + len, PAGE_SIZE);
932 from + len, fo <<
933 } 570 }
934 return 0; 571 return 0;
935 } 572 }
936 EXPORT_SYMBOL(simple_write_begin); 573 EXPORT_SYMBOL(simple_write_begin);
937 574
938 /** 575 /**
939 * simple_write_end - .write_end helper for no 576 * simple_write_end - .write_end helper for non-block-device FSes
940 * @file: See .write_end of address_space_oper 577 * @file: See .write_end of address_space_operations
941 * @mapping: " 578 * @mapping: "
942 * @pos: " 579 * @pos: "
943 * @len: " 580 * @len: "
944 * @copied: " 581 * @copied: "
945 * @folio: " !! 582 * @page: "
946 * @fsdata: " 583 * @fsdata: "
947 * 584 *
948 * simple_write_end does the minimum needed fo !! 585 * simple_write_end does the minimum needed for updating a page after writing is
949 * writing is done. It has the same API signat !! 586 * done. It has the same API signature as the .write_end of
950 * address_space_operations vector. So it can 587 * address_space_operations vector. So it can just be set onto .write_end for
951 * FSes that don't need any other processing. 588 * FSes that don't need any other processing. i_mutex is assumed to be held.
952 * Block based filesystems should use generic_ 589 * Block based filesystems should use generic_write_end().
953 * NOTE: Even though i_size might get updated 590 * NOTE: Even though i_size might get updated by this function, mark_inode_dirty
954 * is not called, so a filesystem that actuall 591 * is not called, so a filesystem that actually does store data in .write_inode
955 * should extend on what's done here with a ca 592 * should extend on what's done here with a call to mark_inode_dirty() in the
956 * case that i_size has changed. 593 * case that i_size has changed.
957 * 594 *
958 * Use *ONLY* with simple_read_folio() !! 595 * Use *ONLY* with simple_readpage()
959 */ 596 */
960 static int simple_write_end(struct file *file, 597 static int simple_write_end(struct file *file, struct address_space *mapping,
961 loff_t pos, unsigned l 598 loff_t pos, unsigned len, unsigned copied,
962 struct folio *folio, v !! 599 struct page *page, void *fsdata)
963 { 600 {
964 struct inode *inode = folio->mapping-> !! 601 struct inode *inode = page->mapping->host;
965 loff_t last_pos = pos + copied; 602 loff_t last_pos = pos + copied;
966 603
967 /* zero the stale part of the folio if !! 604 /* zero the stale part of the page if we did a short copy */
968 if (!folio_test_uptodate(folio)) { !! 605 if (!PageUptodate(page)) {
969 if (copied < len) { 606 if (copied < len) {
970 size_t from = offset_i !! 607 unsigned from = pos & (PAGE_SIZE - 1);
971 608
972 folio_zero_range(folio !! 609 zero_user(page, from + copied, len - copied);
973 } 610 }
974 folio_mark_uptodate(folio); !! 611 SetPageUptodate(page);
975 } 612 }
976 /* 613 /*
977 * No need to use i_size_read() here, 614 * No need to use i_size_read() here, the i_size
978 * cannot change under us because we h 615 * cannot change under us because we hold the i_mutex.
979 */ 616 */
980 if (last_pos > inode->i_size) 617 if (last_pos > inode->i_size)
981 i_size_write(inode, last_pos); 618 i_size_write(inode, last_pos);
982 619
983 folio_mark_dirty(folio); !! 620 set_page_dirty(page);
984 folio_unlock(folio); !! 621 unlock_page(page);
985 folio_put(folio); !! 622 put_page(page);
986 623
987 return copied; 624 return copied;
988 } 625 }
989 626
990 /* 627 /*
991 * Provides ramfs-style behavior: data in the 628 * Provides ramfs-style behavior: data in the pagecache, but no writeback.
992 */ 629 */
993 const struct address_space_operations ram_aops 630 const struct address_space_operations ram_aops = {
994 .read_folio = simple_read_folio, !! 631 .readpage = simple_readpage,
995 .write_begin = simple_write_begin, 632 .write_begin = simple_write_begin,
996 .write_end = simple_write_end, 633 .write_end = simple_write_end,
997 .dirty_folio = noop_dirty_folio, !! 634 .set_page_dirty = __set_page_dirty_no_writeback,
998 }; 635 };
999 EXPORT_SYMBOL(ram_aops); 636 EXPORT_SYMBOL(ram_aops);
1000 637
1001 /* 638 /*
1002 * the inodes created here are not hashed. If 639 * the inodes created here are not hashed. If you use iunique to generate
1003 * unique inode values later for this filesys 640 * unique inode values later for this filesystem, then you must take care
1004 * to pass it an appropriate max_reserved val 641 * to pass it an appropriate max_reserved value to avoid collisions.
1005 */ 642 */
1006 int simple_fill_super(struct super_block *s, 643 int simple_fill_super(struct super_block *s, unsigned long magic,
1007 const struct tree_descr 644 const struct tree_descr *files)
1008 { 645 {
1009 struct inode *inode; 646 struct inode *inode;
>> 647 struct dentry *root;
1010 struct dentry *dentry; 648 struct dentry *dentry;
1011 int i; 649 int i;
1012 650
1013 s->s_blocksize = PAGE_SIZE; 651 s->s_blocksize = PAGE_SIZE;
1014 s->s_blocksize_bits = PAGE_SHIFT; 652 s->s_blocksize_bits = PAGE_SHIFT;
1015 s->s_magic = magic; 653 s->s_magic = magic;
1016 s->s_op = &simple_super_operations; 654 s->s_op = &simple_super_operations;
1017 s->s_time_gran = 1; 655 s->s_time_gran = 1;
1018 656
1019 inode = new_inode(s); 657 inode = new_inode(s);
1020 if (!inode) 658 if (!inode)
1021 return -ENOMEM; 659 return -ENOMEM;
1022 /* 660 /*
1023 * because the root inode is 1, the f 661 * because the root inode is 1, the files array must not contain an
1024 * entry at index 1 662 * entry at index 1
1025 */ 663 */
1026 inode->i_ino = 1; 664 inode->i_ino = 1;
1027 inode->i_mode = S_IFDIR | 0755; 665 inode->i_mode = S_IFDIR | 0755;
1028 simple_inode_init_ts(inode); !! 666 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1029 inode->i_op = &simple_dir_inode_opera 667 inode->i_op = &simple_dir_inode_operations;
1030 inode->i_fop = &simple_dir_operations 668 inode->i_fop = &simple_dir_operations;
1031 set_nlink(inode, 2); 669 set_nlink(inode, 2);
1032 s->s_root = d_make_root(inode); !! 670 root = d_make_root(inode);
1033 if (!s->s_root) !! 671 if (!root)
1034 return -ENOMEM; 672 return -ENOMEM;
1035 for (i = 0; !files->name || files->na 673 for (i = 0; !files->name || files->name[0]; i++, files++) {
1036 if (!files->name) 674 if (!files->name)
1037 continue; 675 continue;
1038 676
1039 /* warn if it tries to confli 677 /* warn if it tries to conflict with the root inode */
1040 if (unlikely(i == 1)) 678 if (unlikely(i == 1))
1041 printk(KERN_WARNING " 679 printk(KERN_WARNING "%s: %s passed in a files array"
1042 "with an inde 680 "with an index of 1!\n", __func__,
1043 s->s_type->na 681 s->s_type->name);
1044 682
1045 dentry = d_alloc_name(s->s_ro !! 683 dentry = d_alloc_name(root, files->name);
1046 if (!dentry) 684 if (!dentry)
1047 return -ENOMEM; !! 685 goto out;
1048 inode = new_inode(s); 686 inode = new_inode(s);
1049 if (!inode) { 687 if (!inode) {
1050 dput(dentry); 688 dput(dentry);
1051 return -ENOMEM; !! 689 goto out;
1052 } 690 }
1053 inode->i_mode = S_IFREG | fil 691 inode->i_mode = S_IFREG | files->mode;
1054 simple_inode_init_ts(inode); !! 692 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1055 inode->i_fop = files->ops; 693 inode->i_fop = files->ops;
1056 inode->i_ino = i; 694 inode->i_ino = i;
1057 d_add(dentry, inode); 695 d_add(dentry, inode);
1058 } 696 }
>> 697 s->s_root = root;
1059 return 0; 698 return 0;
>> 699 out:
>> 700 d_genocide(root);
>> 701 shrink_dcache_parent(root);
>> 702 dput(root);
>> 703 return -ENOMEM;
1060 } 704 }
1061 EXPORT_SYMBOL(simple_fill_super); 705 EXPORT_SYMBOL(simple_fill_super);
1062 706
1063 static DEFINE_SPINLOCK(pin_fs_lock); 707 static DEFINE_SPINLOCK(pin_fs_lock);
1064 708
1065 int simple_pin_fs(struct file_system_type *ty 709 int simple_pin_fs(struct file_system_type *type, struct vfsmount **mount, int *count)
1066 { 710 {
1067 struct vfsmount *mnt = NULL; 711 struct vfsmount *mnt = NULL;
1068 spin_lock(&pin_fs_lock); 712 spin_lock(&pin_fs_lock);
1069 if (unlikely(!*mount)) { 713 if (unlikely(!*mount)) {
1070 spin_unlock(&pin_fs_lock); 714 spin_unlock(&pin_fs_lock);
1071 mnt = vfs_kern_mount(type, SB 715 mnt = vfs_kern_mount(type, SB_KERNMOUNT, type->name, NULL);
1072 if (IS_ERR(mnt)) 716 if (IS_ERR(mnt))
1073 return PTR_ERR(mnt); 717 return PTR_ERR(mnt);
1074 spin_lock(&pin_fs_lock); 718 spin_lock(&pin_fs_lock);
1075 if (!*mount) 719 if (!*mount)
1076 *mount = mnt; 720 *mount = mnt;
1077 } 721 }
1078 mntget(*mount); 722 mntget(*mount);
1079 ++*count; 723 ++*count;
1080 spin_unlock(&pin_fs_lock); 724 spin_unlock(&pin_fs_lock);
1081 mntput(mnt); 725 mntput(mnt);
1082 return 0; 726 return 0;
1083 } 727 }
1084 EXPORT_SYMBOL(simple_pin_fs); 728 EXPORT_SYMBOL(simple_pin_fs);
1085 729
1086 void simple_release_fs(struct vfsmount **moun 730 void simple_release_fs(struct vfsmount **mount, int *count)
1087 { 731 {
1088 struct vfsmount *mnt; 732 struct vfsmount *mnt;
1089 spin_lock(&pin_fs_lock); 733 spin_lock(&pin_fs_lock);
1090 mnt = *mount; 734 mnt = *mount;
1091 if (!--*count) 735 if (!--*count)
1092 *mount = NULL; 736 *mount = NULL;
1093 spin_unlock(&pin_fs_lock); 737 spin_unlock(&pin_fs_lock);
1094 mntput(mnt); 738 mntput(mnt);
1095 } 739 }
1096 EXPORT_SYMBOL(simple_release_fs); 740 EXPORT_SYMBOL(simple_release_fs);
1097 741
1098 /** 742 /**
1099 * simple_read_from_buffer - copy data from t 743 * simple_read_from_buffer - copy data from the buffer to user space
1100 * @to: the user space buffer to read to 744 * @to: the user space buffer to read to
1101 * @count: the maximum number of bytes to rea 745 * @count: the maximum number of bytes to read
1102 * @ppos: the current position in the buffer 746 * @ppos: the current position in the buffer
1103 * @from: the buffer to read from 747 * @from: the buffer to read from
1104 * @available: the size of the buffer 748 * @available: the size of the buffer
1105 * 749 *
1106 * The simple_read_from_buffer() function rea 750 * The simple_read_from_buffer() function reads up to @count bytes from the
1107 * buffer @from at offset @ppos into the user 751 * buffer @from at offset @ppos into the user space address starting at @to.
1108 * 752 *
1109 * On success, the number of bytes read is re 753 * On success, the number of bytes read is returned and the offset @ppos is
1110 * advanced by this number, or negative value 754 * advanced by this number, or negative value is returned on error.
1111 **/ 755 **/
1112 ssize_t simple_read_from_buffer(void __user * 756 ssize_t simple_read_from_buffer(void __user *to, size_t count, loff_t *ppos,
1113 const void *f 757 const void *from, size_t available)
1114 { 758 {
1115 loff_t pos = *ppos; 759 loff_t pos = *ppos;
1116 size_t ret; 760 size_t ret;
1117 761
1118 if (pos < 0) 762 if (pos < 0)
1119 return -EINVAL; 763 return -EINVAL;
1120 if (pos >= available || !count) 764 if (pos >= available || !count)
1121 return 0; 765 return 0;
1122 if (count > available - pos) 766 if (count > available - pos)
1123 count = available - pos; 767 count = available - pos;
1124 ret = copy_to_user(to, from + pos, co 768 ret = copy_to_user(to, from + pos, count);
1125 if (ret == count) 769 if (ret == count)
1126 return -EFAULT; 770 return -EFAULT;
1127 count -= ret; 771 count -= ret;
1128 *ppos = pos + count; 772 *ppos = pos + count;
1129 return count; 773 return count;
1130 } 774 }
1131 EXPORT_SYMBOL(simple_read_from_buffer); 775 EXPORT_SYMBOL(simple_read_from_buffer);
1132 776
1133 /** 777 /**
1134 * simple_write_to_buffer - copy data from us 778 * simple_write_to_buffer - copy data from user space to the buffer
1135 * @to: the buffer to write to 779 * @to: the buffer to write to
1136 * @available: the size of the buffer 780 * @available: the size of the buffer
1137 * @ppos: the current position in the buffer 781 * @ppos: the current position in the buffer
1138 * @from: the user space buffer to read from 782 * @from: the user space buffer to read from
1139 * @count: the maximum number of bytes to rea 783 * @count: the maximum number of bytes to read
1140 * 784 *
1141 * The simple_write_to_buffer() function read 785 * The simple_write_to_buffer() function reads up to @count bytes from the user
1142 * space address starting at @from into the b 786 * space address starting at @from into the buffer @to at offset @ppos.
1143 * 787 *
1144 * On success, the number of bytes written is 788 * On success, the number of bytes written is returned and the offset @ppos is
1145 * advanced by this number, or negative value 789 * advanced by this number, or negative value is returned on error.
1146 **/ 790 **/
1147 ssize_t simple_write_to_buffer(void *to, size 791 ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
1148 const void __user *from, size 792 const void __user *from, size_t count)
1149 { 793 {
1150 loff_t pos = *ppos; 794 loff_t pos = *ppos;
1151 size_t res; 795 size_t res;
1152 796
1153 if (pos < 0) 797 if (pos < 0)
1154 return -EINVAL; 798 return -EINVAL;
1155 if (pos >= available || !count) 799 if (pos >= available || !count)
1156 return 0; 800 return 0;
1157 if (count > available - pos) 801 if (count > available - pos)
1158 count = available - pos; 802 count = available - pos;
1159 res = copy_from_user(to + pos, from, 803 res = copy_from_user(to + pos, from, count);
1160 if (res == count) 804 if (res == count)
1161 return -EFAULT; 805 return -EFAULT;
1162 count -= res; 806 count -= res;
1163 *ppos = pos + count; 807 *ppos = pos + count;
1164 return count; 808 return count;
1165 } 809 }
1166 EXPORT_SYMBOL(simple_write_to_buffer); 810 EXPORT_SYMBOL(simple_write_to_buffer);
1167 811
1168 /** 812 /**
1169 * memory_read_from_buffer - copy data from t 813 * memory_read_from_buffer - copy data from the buffer
1170 * @to: the kernel space buffer to read to 814 * @to: the kernel space buffer to read to
1171 * @count: the maximum number of bytes to rea 815 * @count: the maximum number of bytes to read
1172 * @ppos: the current position in the buffer 816 * @ppos: the current position in the buffer
1173 * @from: the buffer to read from 817 * @from: the buffer to read from
1174 * @available: the size of the buffer 818 * @available: the size of the buffer
1175 * 819 *
1176 * The memory_read_from_buffer() function rea 820 * The memory_read_from_buffer() function reads up to @count bytes from the
1177 * buffer @from at offset @ppos into the kern 821 * buffer @from at offset @ppos into the kernel space address starting at @to.
1178 * 822 *
1179 * On success, the number of bytes read is re 823 * On success, the number of bytes read is returned and the offset @ppos is
1180 * advanced by this number, or negative value 824 * advanced by this number, or negative value is returned on error.
1181 **/ 825 **/
1182 ssize_t memory_read_from_buffer(void *to, siz 826 ssize_t memory_read_from_buffer(void *to, size_t count, loff_t *ppos,
1183 const void *f 827 const void *from, size_t available)
1184 { 828 {
1185 loff_t pos = *ppos; 829 loff_t pos = *ppos;
1186 830
1187 if (pos < 0) 831 if (pos < 0)
1188 return -EINVAL; 832 return -EINVAL;
1189 if (pos >= available) 833 if (pos >= available)
1190 return 0; 834 return 0;
1191 if (count > available - pos) 835 if (count > available - pos)
1192 count = available - pos; 836 count = available - pos;
1193 memcpy(to, from + pos, count); 837 memcpy(to, from + pos, count);
1194 *ppos = pos + count; 838 *ppos = pos + count;
1195 839
1196 return count; 840 return count;
1197 } 841 }
1198 EXPORT_SYMBOL(memory_read_from_buffer); 842 EXPORT_SYMBOL(memory_read_from_buffer);
1199 843
1200 /* 844 /*
1201 * Transaction based IO. 845 * Transaction based IO.
1202 * The file expects a single write which trig 846 * The file expects a single write which triggers the transaction, and then
1203 * possibly a read which collects the result 847 * possibly a read which collects the result - which is stored in a
1204 * file-local buffer. 848 * file-local buffer.
1205 */ 849 */
1206 850
1207 void simple_transaction_set(struct file *file 851 void simple_transaction_set(struct file *file, size_t n)
1208 { 852 {
1209 struct simple_transaction_argresp *ar 853 struct simple_transaction_argresp *ar = file->private_data;
1210 854
1211 BUG_ON(n > SIMPLE_TRANSACTION_LIMIT); 855 BUG_ON(n > SIMPLE_TRANSACTION_LIMIT);
1212 856
1213 /* 857 /*
1214 * The barrier ensures that ar->size 858 * The barrier ensures that ar->size will really remain zero until
1215 * ar->data is ready for reading. 859 * ar->data is ready for reading.
1216 */ 860 */
1217 smp_mb(); 861 smp_mb();
1218 ar->size = n; 862 ar->size = n;
1219 } 863 }
1220 EXPORT_SYMBOL(simple_transaction_set); 864 EXPORT_SYMBOL(simple_transaction_set);
1221 865
1222 char *simple_transaction_get(struct file *fil 866 char *simple_transaction_get(struct file *file, const char __user *buf, size_t size)
1223 { 867 {
1224 struct simple_transaction_argresp *ar 868 struct simple_transaction_argresp *ar;
1225 static DEFINE_SPINLOCK(simple_transac 869 static DEFINE_SPINLOCK(simple_transaction_lock);
1226 870
1227 if (size > SIMPLE_TRANSACTION_LIMIT - 871 if (size > SIMPLE_TRANSACTION_LIMIT - 1)
1228 return ERR_PTR(-EFBIG); 872 return ERR_PTR(-EFBIG);
1229 873
1230 ar = (struct simple_transaction_argre 874 ar = (struct simple_transaction_argresp *)get_zeroed_page(GFP_KERNEL);
1231 if (!ar) 875 if (!ar)
1232 return ERR_PTR(-ENOMEM); 876 return ERR_PTR(-ENOMEM);
1233 877
1234 spin_lock(&simple_transaction_lock); 878 spin_lock(&simple_transaction_lock);
1235 879
1236 /* only one write allowed per open */ 880 /* only one write allowed per open */
1237 if (file->private_data) { 881 if (file->private_data) {
1238 spin_unlock(&simple_transacti 882 spin_unlock(&simple_transaction_lock);
1239 free_page((unsigned long)ar); 883 free_page((unsigned long)ar);
1240 return ERR_PTR(-EBUSY); 884 return ERR_PTR(-EBUSY);
1241 } 885 }
1242 886
1243 file->private_data = ar; 887 file->private_data = ar;
1244 888
1245 spin_unlock(&simple_transaction_lock) 889 spin_unlock(&simple_transaction_lock);
1246 890
1247 if (copy_from_user(ar->data, buf, siz 891 if (copy_from_user(ar->data, buf, size))
1248 return ERR_PTR(-EFAULT); 892 return ERR_PTR(-EFAULT);
1249 893
1250 return ar->data; 894 return ar->data;
1251 } 895 }
1252 EXPORT_SYMBOL(simple_transaction_get); 896 EXPORT_SYMBOL(simple_transaction_get);
1253 897
1254 ssize_t simple_transaction_read(struct file * 898 ssize_t simple_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
1255 { 899 {
1256 struct simple_transaction_argresp *ar 900 struct simple_transaction_argresp *ar = file->private_data;
1257 901
1258 if (!ar) 902 if (!ar)
1259 return 0; 903 return 0;
1260 return simple_read_from_buffer(buf, s 904 return simple_read_from_buffer(buf, size, pos, ar->data, ar->size);
1261 } 905 }
1262 EXPORT_SYMBOL(simple_transaction_read); 906 EXPORT_SYMBOL(simple_transaction_read);
1263 907
1264 int simple_transaction_release(struct inode * 908 int simple_transaction_release(struct inode *inode, struct file *file)
1265 { 909 {
1266 free_page((unsigned long)file->privat 910 free_page((unsigned long)file->private_data);
1267 return 0; 911 return 0;
1268 } 912 }
1269 EXPORT_SYMBOL(simple_transaction_release); 913 EXPORT_SYMBOL(simple_transaction_release);
1270 914
1271 /* Simple attribute files */ 915 /* Simple attribute files */
1272 916
1273 struct simple_attr { 917 struct simple_attr {
1274 int (*get)(void *, u64 *); 918 int (*get)(void *, u64 *);
1275 int (*set)(void *, u64); 919 int (*set)(void *, u64);
1276 char get_buf[24]; /* enough to 920 char get_buf[24]; /* enough to store a u64 and "\n\0" */
1277 char set_buf[24]; 921 char set_buf[24];
1278 void *data; 922 void *data;
1279 const char *fmt; /* format for 923 const char *fmt; /* format for read operation */
1280 struct mutex mutex; /* protects a 924 struct mutex mutex; /* protects access to these buffers */
1281 }; 925 };
1282 926
1283 /* simple_attr_open is called by an actual at 927 /* simple_attr_open is called by an actual attribute open file operation
1284 * to set the attribute specific access opera 928 * to set the attribute specific access operations. */
1285 int simple_attr_open(struct inode *inode, str 929 int simple_attr_open(struct inode *inode, struct file *file,
1286 int (*get)(void *, u64 * 930 int (*get)(void *, u64 *), int (*set)(void *, u64),
1287 const char *fmt) 931 const char *fmt)
1288 { 932 {
1289 struct simple_attr *attr; 933 struct simple_attr *attr;
1290 934
1291 attr = kzalloc(sizeof(*attr), GFP_KER 935 attr = kzalloc(sizeof(*attr), GFP_KERNEL);
1292 if (!attr) 936 if (!attr)
1293 return -ENOMEM; 937 return -ENOMEM;
1294 938
1295 attr->get = get; 939 attr->get = get;
1296 attr->set = set; 940 attr->set = set;
1297 attr->data = inode->i_private; 941 attr->data = inode->i_private;
1298 attr->fmt = fmt; 942 attr->fmt = fmt;
1299 mutex_init(&attr->mutex); 943 mutex_init(&attr->mutex);
1300 944
1301 file->private_data = attr; 945 file->private_data = attr;
1302 946
1303 return nonseekable_open(inode, file); 947 return nonseekable_open(inode, file);
1304 } 948 }
1305 EXPORT_SYMBOL_GPL(simple_attr_open); 949 EXPORT_SYMBOL_GPL(simple_attr_open);
1306 950
1307 int simple_attr_release(struct inode *inode, 951 int simple_attr_release(struct inode *inode, struct file *file)
1308 { 952 {
1309 kfree(file->private_data); 953 kfree(file->private_data);
1310 return 0; 954 return 0;
1311 } 955 }
1312 EXPORT_SYMBOL_GPL(simple_attr_release); /* GP 956 EXPORT_SYMBOL_GPL(simple_attr_release); /* GPL-only? This? Really? */
1313 957
1314 /* read from the buffer that is filled with t 958 /* read from the buffer that is filled with the get function */
1315 ssize_t simple_attr_read(struct file *file, c 959 ssize_t simple_attr_read(struct file *file, char __user *buf,
1316 size_t len, loff_t * 960 size_t len, loff_t *ppos)
1317 { 961 {
1318 struct simple_attr *attr; 962 struct simple_attr *attr;
1319 size_t size; 963 size_t size;
1320 ssize_t ret; 964 ssize_t ret;
1321 965
1322 attr = file->private_data; 966 attr = file->private_data;
1323 967
1324 if (!attr->get) 968 if (!attr->get)
1325 return -EACCES; 969 return -EACCES;
1326 970
1327 ret = mutex_lock_interruptible(&attr- 971 ret = mutex_lock_interruptible(&attr->mutex);
1328 if (ret) 972 if (ret)
1329 return ret; 973 return ret;
1330 974
1331 if (*ppos && attr->get_buf[0]) { 975 if (*ppos && attr->get_buf[0]) {
1332 /* continued read */ 976 /* continued read */
1333 size = strlen(attr->get_buf); 977 size = strlen(attr->get_buf);
1334 } else { 978 } else {
1335 /* first read */ 979 /* first read */
1336 u64 val; 980 u64 val;
1337 ret = attr->get(attr->data, & 981 ret = attr->get(attr->data, &val);
1338 if (ret) 982 if (ret)
1339 goto out; 983 goto out;
1340 984
1341 size = scnprintf(attr->get_bu 985 size = scnprintf(attr->get_buf, sizeof(attr->get_buf),
1342 attr->fmt, ( 986 attr->fmt, (unsigned long long)val);
1343 } 987 }
1344 988
1345 ret = simple_read_from_buffer(buf, le 989 ret = simple_read_from_buffer(buf, len, ppos, attr->get_buf, size);
1346 out: 990 out:
1347 mutex_unlock(&attr->mutex); 991 mutex_unlock(&attr->mutex);
1348 return ret; 992 return ret;
1349 } 993 }
1350 EXPORT_SYMBOL_GPL(simple_attr_read); 994 EXPORT_SYMBOL_GPL(simple_attr_read);
1351 995
1352 /* interpret the buffer as a number to call t 996 /* interpret the buffer as a number to call the set function with */
1353 static ssize_t simple_attr_write_xsigned(stru !! 997 ssize_t simple_attr_write(struct file *file, const char __user *buf,
1354 size_t len, loff_t !! 998 size_t len, loff_t *ppos)
1355 { 999 {
1356 struct simple_attr *attr; 1000 struct simple_attr *attr;
1357 unsigned long long val; 1001 unsigned long long val;
1358 size_t size; 1002 size_t size;
1359 ssize_t ret; 1003 ssize_t ret;
1360 1004
1361 attr = file->private_data; 1005 attr = file->private_data;
1362 if (!attr->set) 1006 if (!attr->set)
1363 return -EACCES; 1007 return -EACCES;
1364 1008
1365 ret = mutex_lock_interruptible(&attr- 1009 ret = mutex_lock_interruptible(&attr->mutex);
1366 if (ret) 1010 if (ret)
1367 return ret; 1011 return ret;
1368 1012
1369 ret = -EFAULT; 1013 ret = -EFAULT;
1370 size = min(sizeof(attr->set_buf) - 1, 1014 size = min(sizeof(attr->set_buf) - 1, len);
1371 if (copy_from_user(attr->set_buf, buf 1015 if (copy_from_user(attr->set_buf, buf, size))
1372 goto out; 1016 goto out;
1373 1017
1374 attr->set_buf[size] = '\0'; 1018 attr->set_buf[size] = '\0';
1375 if (is_signed) !! 1019 ret = kstrtoull(attr->set_buf, 0, &val);
1376 ret = kstrtoll(attr->set_buf, <<
1377 else <<
1378 ret = kstrtoull(attr->set_buf <<
1379 if (ret) 1020 if (ret)
1380 goto out; 1021 goto out;
1381 ret = attr->set(attr->data, val); 1022 ret = attr->set(attr->data, val);
1382 if (ret == 0) 1023 if (ret == 0)
1383 ret = len; /* on success, cla 1024 ret = len; /* on success, claim we got the whole input */
1384 out: 1025 out:
1385 mutex_unlock(&attr->mutex); 1026 mutex_unlock(&attr->mutex);
1386 return ret; 1027 return ret;
1387 } 1028 }
1388 <<
1389 ssize_t simple_attr_write(struct file *file, <<
1390 size_t len, loff_t <<
1391 { <<
1392 return simple_attr_write_xsigned(file <<
1393 } <<
1394 EXPORT_SYMBOL_GPL(simple_attr_write); 1029 EXPORT_SYMBOL_GPL(simple_attr_write);
1395 1030
1396 ssize_t simple_attr_write_signed(struct file <<
1397 size_t len, loff_t <<
1398 { <<
1399 return simple_attr_write_xsigned(file <<
1400 } <<
1401 EXPORT_SYMBOL_GPL(simple_attr_write_signed); <<
1402 <<
1403 /** <<
1404 * generic_encode_ino32_fh - generic export_o <<
1405 * @inode: the object to encode <<
1406 * @fh: where to store the file handle f <<
1407 * @max_len: maximum length to store there (i <<
1408 * @parent: parent directory inode, if wante <<
1409 * <<
1410 * This generic encode_fh function assumes th <<
1411 * is suitable for locating an inode, and tha <<
1412 * can be used to check that it is still vali <<
1413 * filehandle fragment where export_decode_fh <<
1414 */ <<
1415 int generic_encode_ino32_fh(struct inode *ino <<
1416 struct inode *par <<
1417 { <<
1418 struct fid *fid = (void *)fh; <<
1419 int len = *max_len; <<
1420 int type = FILEID_INO32_GEN; <<
1421 <<
1422 if (parent && (len < 4)) { <<
1423 *max_len = 4; <<
1424 return FILEID_INVALID; <<
1425 } else if (len < 2) { <<
1426 *max_len = 2; <<
1427 return FILEID_INVALID; <<
1428 } <<
1429 <<
1430 len = 2; <<
1431 fid->i32.ino = inode->i_ino; <<
1432 fid->i32.gen = inode->i_generation; <<
1433 if (parent) { <<
1434 fid->i32.parent_ino = parent- <<
1435 fid->i32.parent_gen = parent- <<
1436 len = 4; <<
1437 type = FILEID_INO32_GEN_PAREN <<
1438 } <<
1439 *max_len = len; <<
1440 return type; <<
1441 } <<
1442 EXPORT_SYMBOL_GPL(generic_encode_ino32_fh); <<
1443 <<
1444 /** 1031 /**
1445 * generic_fh_to_dentry - generic helper for 1032 * generic_fh_to_dentry - generic helper for the fh_to_dentry export operation
1446 * @sb: filesystem to do the file han 1033 * @sb: filesystem to do the file handle conversion on
1447 * @fid: file handle to convert 1034 * @fid: file handle to convert
1448 * @fh_len: length of the file handle in 1035 * @fh_len: length of the file handle in bytes
1449 * @fh_type: type of file handle 1036 * @fh_type: type of file handle
1450 * @get_inode: filesystem callback to retrie 1037 * @get_inode: filesystem callback to retrieve inode
1451 * 1038 *
1452 * This function decodes @fid as long as it h 1039 * This function decodes @fid as long as it has one of the well-known
1453 * Linux filehandle types and calls @get_inod 1040 * Linux filehandle types and calls @get_inode on it to retrieve the
1454 * inode for the object specified in the file 1041 * inode for the object specified in the file handle.
1455 */ 1042 */
1456 struct dentry *generic_fh_to_dentry(struct su 1043 struct dentry *generic_fh_to_dentry(struct super_block *sb, struct fid *fid,
1457 int fh_len, int fh_type, stru 1044 int fh_len, int fh_type, struct inode *(*get_inode)
1458 (struct super_block * 1045 (struct super_block *sb, u64 ino, u32 gen))
1459 { 1046 {
1460 struct inode *inode = NULL; 1047 struct inode *inode = NULL;
1461 1048
1462 if (fh_len < 2) 1049 if (fh_len < 2)
1463 return NULL; 1050 return NULL;
1464 1051
1465 switch (fh_type) { 1052 switch (fh_type) {
1466 case FILEID_INO32_GEN: 1053 case FILEID_INO32_GEN:
1467 case FILEID_INO32_GEN_PARENT: 1054 case FILEID_INO32_GEN_PARENT:
1468 inode = get_inode(sb, fid->i3 1055 inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
1469 break; 1056 break;
1470 } 1057 }
1471 1058
1472 return d_obtain_alias(inode); 1059 return d_obtain_alias(inode);
1473 } 1060 }
1474 EXPORT_SYMBOL_GPL(generic_fh_to_dentry); 1061 EXPORT_SYMBOL_GPL(generic_fh_to_dentry);
1475 1062
1476 /** 1063 /**
1477 * generic_fh_to_parent - generic helper for 1064 * generic_fh_to_parent - generic helper for the fh_to_parent export operation
1478 * @sb: filesystem to do the file han 1065 * @sb: filesystem to do the file handle conversion on
1479 * @fid: file handle to convert 1066 * @fid: file handle to convert
1480 * @fh_len: length of the file handle in 1067 * @fh_len: length of the file handle in bytes
1481 * @fh_type: type of file handle 1068 * @fh_type: type of file handle
1482 * @get_inode: filesystem callback to retrie 1069 * @get_inode: filesystem callback to retrieve inode
1483 * 1070 *
1484 * This function decodes @fid as long as it h 1071 * This function decodes @fid as long as it has one of the well-known
1485 * Linux filehandle types and calls @get_inod 1072 * Linux filehandle types and calls @get_inode on it to retrieve the
1486 * inode for the _parent_ object specified in 1073 * inode for the _parent_ object specified in the file handle if it
1487 * is specified in the file handle, or NULL o 1074 * is specified in the file handle, or NULL otherwise.
1488 */ 1075 */
1489 struct dentry *generic_fh_to_parent(struct su 1076 struct dentry *generic_fh_to_parent(struct super_block *sb, struct fid *fid,
1490 int fh_len, int fh_type, stru 1077 int fh_len, int fh_type, struct inode *(*get_inode)
1491 (struct super_block * 1078 (struct super_block *sb, u64 ino, u32 gen))
1492 { 1079 {
1493 struct inode *inode = NULL; 1080 struct inode *inode = NULL;
1494 1081
1495 if (fh_len <= 2) 1082 if (fh_len <= 2)
1496 return NULL; 1083 return NULL;
1497 1084
1498 switch (fh_type) { 1085 switch (fh_type) {
1499 case FILEID_INO32_GEN_PARENT: 1086 case FILEID_INO32_GEN_PARENT:
1500 inode = get_inode(sb, fid->i3 1087 inode = get_inode(sb, fid->i32.parent_ino,
1501 (fh_len > 3 1088 (fh_len > 3 ? fid->i32.parent_gen : 0));
1502 break; 1089 break;
1503 } 1090 }
1504 1091
1505 return d_obtain_alias(inode); 1092 return d_obtain_alias(inode);
1506 } 1093 }
1507 EXPORT_SYMBOL_GPL(generic_fh_to_parent); 1094 EXPORT_SYMBOL_GPL(generic_fh_to_parent);
1508 1095
1509 /** 1096 /**
1510 * __generic_file_fsync - generic fsync imple 1097 * __generic_file_fsync - generic fsync implementation for simple filesystems
1511 * 1098 *
1512 * @file: file to synchronize 1099 * @file: file to synchronize
1513 * @start: start offset in bytes 1100 * @start: start offset in bytes
1514 * @end: end offset in bytes (inclusiv 1101 * @end: end offset in bytes (inclusive)
1515 * @datasync: only synchronize essential me 1102 * @datasync: only synchronize essential metadata if true
1516 * 1103 *
1517 * This is a generic implementation of the fs 1104 * This is a generic implementation of the fsync method for simple
1518 * filesystems which track all non-inode meta 1105 * filesystems which track all non-inode metadata in the buffers list
1519 * hanging off the address_space structure. 1106 * hanging off the address_space structure.
1520 */ 1107 */
1521 int __generic_file_fsync(struct file *file, l 1108 int __generic_file_fsync(struct file *file, loff_t start, loff_t end,
1522 int datasync 1109 int datasync)
1523 { 1110 {
1524 struct inode *inode = file->f_mapping 1111 struct inode *inode = file->f_mapping->host;
1525 int err; 1112 int err;
1526 int ret; 1113 int ret;
1527 1114
1528 err = file_write_and_wait_range(file, 1115 err = file_write_and_wait_range(file, start, end);
1529 if (err) 1116 if (err)
1530 return err; 1117 return err;
1531 1118
1532 inode_lock(inode); 1119 inode_lock(inode);
1533 ret = sync_mapping_buffers(inode->i_m 1120 ret = sync_mapping_buffers(inode->i_mapping);
1534 if (!(inode->i_state & I_DIRTY_ALL)) 1121 if (!(inode->i_state & I_DIRTY_ALL))
1535 goto out; 1122 goto out;
1536 if (datasync && !(inode->i_state & I_ 1123 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
1537 goto out; 1124 goto out;
1538 1125
1539 err = sync_inode_metadata(inode, 1); 1126 err = sync_inode_metadata(inode, 1);
1540 if (ret == 0) 1127 if (ret == 0)
1541 ret = err; 1128 ret = err;
1542 1129
1543 out: 1130 out:
1544 inode_unlock(inode); 1131 inode_unlock(inode);
1545 /* check and advance again to catch e 1132 /* check and advance again to catch errors after syncing out buffers */
1546 err = file_check_and_advance_wb_err(f 1133 err = file_check_and_advance_wb_err(file);
1547 if (ret == 0) 1134 if (ret == 0)
1548 ret = err; 1135 ret = err;
1549 return ret; 1136 return ret;
1550 } 1137 }
1551 EXPORT_SYMBOL(__generic_file_fsync); 1138 EXPORT_SYMBOL(__generic_file_fsync);
1552 1139
1553 /** 1140 /**
1554 * generic_file_fsync - generic fsync impleme 1141 * generic_file_fsync - generic fsync implementation for simple filesystems
1555 * with flush 1142 * with flush
1556 * @file: file to synchronize 1143 * @file: file to synchronize
1557 * @start: start offset in bytes 1144 * @start: start offset in bytes
1558 * @end: end offset in bytes (inclusiv 1145 * @end: end offset in bytes (inclusive)
1559 * @datasync: only synchronize essential me 1146 * @datasync: only synchronize essential metadata if true
1560 * 1147 *
1561 */ 1148 */
1562 1149
1563 int generic_file_fsync(struct file *file, lof 1150 int generic_file_fsync(struct file *file, loff_t start, loff_t end,
1564 int datasync) 1151 int datasync)
1565 { 1152 {
1566 struct inode *inode = file->f_mapping 1153 struct inode *inode = file->f_mapping->host;
1567 int err; 1154 int err;
1568 1155
1569 err = __generic_file_fsync(file, star 1156 err = __generic_file_fsync(file, start, end, datasync);
1570 if (err) 1157 if (err)
1571 return err; 1158 return err;
1572 return blkdev_issue_flush(inode->i_sb 1159 return blkdev_issue_flush(inode->i_sb->s_bdev);
1573 } 1160 }
1574 EXPORT_SYMBOL(generic_file_fsync); 1161 EXPORT_SYMBOL(generic_file_fsync);
1575 1162
1576 /** 1163 /**
1577 * generic_check_addressable - Check addressa 1164 * generic_check_addressable - Check addressability of file system
1578 * @blocksize_bits: log of file system bl 1165 * @blocksize_bits: log of file system block size
1579 * @num_blocks: number of blocks in f 1166 * @num_blocks: number of blocks in file system
1580 * 1167 *
1581 * Determine whether a file system with @num_ 1168 * Determine whether a file system with @num_blocks blocks (and a
1582 * block size of 2**@blocksize_bits) is addre 1169 * block size of 2**@blocksize_bits) is addressable by the sector_t
1583 * and page cache of the system. Return 0 if 1170 * and page cache of the system. Return 0 if so and -EFBIG otherwise.
1584 */ 1171 */
1585 int generic_check_addressable(unsigned blocks 1172 int generic_check_addressable(unsigned blocksize_bits, u64 num_blocks)
1586 { 1173 {
1587 u64 last_fs_block = num_blocks - 1; 1174 u64 last_fs_block = num_blocks - 1;
1588 u64 last_fs_page = 1175 u64 last_fs_page =
1589 last_fs_block >> (PAGE_SHIFT 1176 last_fs_block >> (PAGE_SHIFT - blocksize_bits);
1590 1177
1591 if (unlikely(num_blocks == 0)) 1178 if (unlikely(num_blocks == 0))
1592 return 0; 1179 return 0;
1593 1180
1594 if ((blocksize_bits < 9) || (blocksiz 1181 if ((blocksize_bits < 9) || (blocksize_bits > PAGE_SHIFT))
1595 return -EINVAL; 1182 return -EINVAL;
1596 1183
1597 if ((last_fs_block > (sector_t)(~0ULL 1184 if ((last_fs_block > (sector_t)(~0ULL) >> (blocksize_bits - 9)) ||
1598 (last_fs_page > (pgoff_t)(~0ULL)) 1185 (last_fs_page > (pgoff_t)(~0ULL))) {
1599 return -EFBIG; 1186 return -EFBIG;
1600 } 1187 }
1601 return 0; 1188 return 0;
1602 } 1189 }
1603 EXPORT_SYMBOL(generic_check_addressable); 1190 EXPORT_SYMBOL(generic_check_addressable);
1604 1191
1605 /* 1192 /*
1606 * No-op implementation of ->fsync for in-mem 1193 * No-op implementation of ->fsync for in-memory filesystems.
1607 */ 1194 */
1608 int noop_fsync(struct file *file, loff_t star 1195 int noop_fsync(struct file *file, loff_t start, loff_t end, int datasync)
1609 { 1196 {
1610 return 0; 1197 return 0;
1611 } 1198 }
1612 EXPORT_SYMBOL(noop_fsync); 1199 EXPORT_SYMBOL(noop_fsync);
1613 1200
>> 1201 void noop_invalidatepage(struct page *page, unsigned int offset,
>> 1202 unsigned int length)
>> 1203 {
>> 1204 /*
>> 1205 * There is no page cache to invalidate in the dax case, however
>> 1206 * we need this callback defined to prevent falling back to
>> 1207 * block_invalidatepage() in do_invalidatepage().
>> 1208 */
>> 1209 }
>> 1210 EXPORT_SYMBOL_GPL(noop_invalidatepage);
>> 1211
1614 ssize_t noop_direct_IO(struct kiocb *iocb, st 1212 ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
1615 { 1213 {
1616 /* 1214 /*
1617 * iomap based filesystems support di 1215 * iomap based filesystems support direct I/O without need for
1618 * this callback. However, it still n 1216 * this callback. However, it still needs to be set in
1619 * inode->a_ops so that open/fcntl kn 1217 * inode->a_ops so that open/fcntl know that direct I/O is
1620 * generally supported. 1218 * generally supported.
1621 */ 1219 */
1622 return -EINVAL; 1220 return -EINVAL;
1623 } 1221 }
1624 EXPORT_SYMBOL_GPL(noop_direct_IO); 1222 EXPORT_SYMBOL_GPL(noop_direct_IO);
1625 1223
1626 /* Because kfree isn't assignment-compatible 1224 /* Because kfree isn't assignment-compatible with void(void*) ;-/ */
1627 void kfree_link(void *p) 1225 void kfree_link(void *p)
1628 { 1226 {
1629 kfree(p); 1227 kfree(p);
1630 } 1228 }
1631 EXPORT_SYMBOL(kfree_link); 1229 EXPORT_SYMBOL(kfree_link);
1632 1230
1633 struct inode *alloc_anon_inode(struct super_b 1231 struct inode *alloc_anon_inode(struct super_block *s)
1634 { 1232 {
1635 static const struct address_space_ope 1233 static const struct address_space_operations anon_aops = {
1636 .dirty_folio = noop_dirty_ !! 1234 .set_page_dirty = __set_page_dirty_no_writeback,
1637 }; 1235 };
1638 struct inode *inode = new_inode_pseud 1236 struct inode *inode = new_inode_pseudo(s);
1639 1237
1640 if (!inode) 1238 if (!inode)
1641 return ERR_PTR(-ENOMEM); 1239 return ERR_PTR(-ENOMEM);
1642 1240
1643 inode->i_ino = get_next_ino(); 1241 inode->i_ino = get_next_ino();
1644 inode->i_mapping->a_ops = &anon_aops; 1242 inode->i_mapping->a_ops = &anon_aops;
1645 1243
1646 /* 1244 /*
1647 * Mark the inode dirty from the very 1245 * Mark the inode dirty from the very beginning,
1648 * that way it will never be moved to 1246 * that way it will never be moved to the dirty
1649 * list because mark_inode_dirty() wi 1247 * list because mark_inode_dirty() will think
1650 * that it already _is_ on the dirty 1248 * that it already _is_ on the dirty list.
1651 */ 1249 */
1652 inode->i_state = I_DIRTY; 1250 inode->i_state = I_DIRTY;
1653 inode->i_mode = S_IRUSR | S_IWUSR; 1251 inode->i_mode = S_IRUSR | S_IWUSR;
1654 inode->i_uid = current_fsuid(); 1252 inode->i_uid = current_fsuid();
1655 inode->i_gid = current_fsgid(); 1253 inode->i_gid = current_fsgid();
1656 inode->i_flags |= S_PRIVATE; 1254 inode->i_flags |= S_PRIVATE;
1657 simple_inode_init_ts(inode); !! 1255 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1658 return inode; 1256 return inode;
1659 } 1257 }
1660 EXPORT_SYMBOL(alloc_anon_inode); 1258 EXPORT_SYMBOL(alloc_anon_inode);
1661 1259
1662 /** 1260 /**
1663 * simple_nosetlease - generic helper for pro 1261 * simple_nosetlease - generic helper for prohibiting leases
1664 * @filp: file pointer 1262 * @filp: file pointer
1665 * @arg: type of lease to obtain 1263 * @arg: type of lease to obtain
1666 * @flp: new lease supplied for insertion 1264 * @flp: new lease supplied for insertion
1667 * @priv: private data for lm_setup operation 1265 * @priv: private data for lm_setup operation
1668 * 1266 *
1669 * Generic helper for filesystems that do not 1267 * Generic helper for filesystems that do not wish to allow leases to be set.
1670 * All arguments are ignored and it just retu 1268 * All arguments are ignored and it just returns -EINVAL.
1671 */ 1269 */
1672 int 1270 int
1673 simple_nosetlease(struct file *filp, int arg, !! 1271 simple_nosetlease(struct file *filp, long arg, struct file_lock **flp,
1674 void **priv) 1272 void **priv)
1675 { 1273 {
1676 return -EINVAL; 1274 return -EINVAL;
1677 } 1275 }
1678 EXPORT_SYMBOL(simple_nosetlease); 1276 EXPORT_SYMBOL(simple_nosetlease);
1679 1277
1680 /** 1278 /**
1681 * simple_get_link - generic helper to get th 1279 * simple_get_link - generic helper to get the target of "fast" symlinks
1682 * @dentry: not used here 1280 * @dentry: not used here
1683 * @inode: the symlink inode 1281 * @inode: the symlink inode
1684 * @done: not used here 1282 * @done: not used here
1685 * 1283 *
1686 * Generic helper for filesystems to use for 1284 * Generic helper for filesystems to use for symlink inodes where a pointer to
1687 * the symlink target is stored in ->i_link. 1285 * the symlink target is stored in ->i_link. NOTE: this isn't normally called,
1688 * since as an optimization the path lookup c 1286 * since as an optimization the path lookup code uses any non-NULL ->i_link
1689 * directly, without calling ->get_link(). B 1287 * directly, without calling ->get_link(). But ->get_link() still must be set,
1690 * to mark the inode_operations as being for 1288 * to mark the inode_operations as being for a symlink.
1691 * 1289 *
1692 * Return: the symlink target 1290 * Return: the symlink target
1693 */ 1291 */
1694 const char *simple_get_link(struct dentry *de 1292 const char *simple_get_link(struct dentry *dentry, struct inode *inode,
1695 struct delayed_ca 1293 struct delayed_call *done)
1696 { 1294 {
1697 return inode->i_link; 1295 return inode->i_link;
1698 } 1296 }
1699 EXPORT_SYMBOL(simple_get_link); 1297 EXPORT_SYMBOL(simple_get_link);
1700 1298
1701 const struct inode_operations simple_symlink_ 1299 const struct inode_operations simple_symlink_inode_operations = {
1702 .get_link = simple_get_link, 1300 .get_link = simple_get_link,
1703 }; 1301 };
1704 EXPORT_SYMBOL(simple_symlink_inode_operations 1302 EXPORT_SYMBOL(simple_symlink_inode_operations);
1705 1303
1706 /* 1304 /*
1707 * Operations for a permanently empty directo 1305 * Operations for a permanently empty directory.
1708 */ 1306 */
1709 static struct dentry *empty_dir_lookup(struct 1307 static struct dentry *empty_dir_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
1710 { 1308 {
1711 return ERR_PTR(-ENOENT); 1309 return ERR_PTR(-ENOENT);
1712 } 1310 }
1713 1311
1714 static int empty_dir_getattr(struct mnt_idmap !! 1312 static int empty_dir_getattr(struct user_namespace *mnt_userns,
1715 const struct pat 1313 const struct path *path, struct kstat *stat,
1716 u32 request_mask 1314 u32 request_mask, unsigned int query_flags)
1717 { 1315 {
1718 struct inode *inode = d_inode(path->d 1316 struct inode *inode = d_inode(path->dentry);
1719 generic_fillattr(&nop_mnt_idmap, requ !! 1317 generic_fillattr(&init_user_ns, inode, stat);
1720 return 0; 1318 return 0;
1721 } 1319 }
1722 1320
1723 static int empty_dir_setattr(struct mnt_idmap !! 1321 static int empty_dir_setattr(struct user_namespace *mnt_userns,
1724 struct dentry *d 1322 struct dentry *dentry, struct iattr *attr)
1725 { 1323 {
1726 return -EPERM; 1324 return -EPERM;
1727 } 1325 }
1728 1326
1729 static ssize_t empty_dir_listxattr(struct den 1327 static ssize_t empty_dir_listxattr(struct dentry *dentry, char *list, size_t size)
1730 { 1328 {
1731 return -EOPNOTSUPP; 1329 return -EOPNOTSUPP;
1732 } 1330 }
1733 1331
1734 static const struct inode_operations empty_di 1332 static const struct inode_operations empty_dir_inode_operations = {
1735 .lookup = empty_dir_lookup, 1333 .lookup = empty_dir_lookup,
1736 .permission = generic_permission, 1334 .permission = generic_permission,
1737 .setattr = empty_dir_setattr, 1335 .setattr = empty_dir_setattr,
1738 .getattr = empty_dir_getattr, 1336 .getattr = empty_dir_getattr,
1739 .listxattr = empty_dir_listxattr 1337 .listxattr = empty_dir_listxattr,
1740 }; 1338 };
1741 1339
1742 static loff_t empty_dir_llseek(struct file *f 1340 static loff_t empty_dir_llseek(struct file *file, loff_t offset, int whence)
1743 { 1341 {
1744 /* An empty directory has two entries 1342 /* An empty directory has two entries . and .. at offsets 0 and 1 */
1745 return generic_file_llseek_size(file, 1343 return generic_file_llseek_size(file, offset, whence, 2, 2);
1746 } 1344 }
1747 1345
1748 static int empty_dir_readdir(struct file *fil 1346 static int empty_dir_readdir(struct file *file, struct dir_context *ctx)
1749 { 1347 {
1750 dir_emit_dots(file, ctx); 1348 dir_emit_dots(file, ctx);
1751 return 0; 1349 return 0;
1752 } 1350 }
1753 1351
1754 static const struct file_operations empty_dir 1352 static const struct file_operations empty_dir_operations = {
1755 .llseek = empty_dir_llseek, 1353 .llseek = empty_dir_llseek,
1756 .read = generic_read_dir, 1354 .read = generic_read_dir,
1757 .iterate_shared = empty_dir_readdir, 1355 .iterate_shared = empty_dir_readdir,
1758 .fsync = noop_fsync, 1356 .fsync = noop_fsync,
1759 }; 1357 };
1760 1358
1761 1359
1762 void make_empty_dir_inode(struct inode *inode 1360 void make_empty_dir_inode(struct inode *inode)
1763 { 1361 {
1764 set_nlink(inode, 2); 1362 set_nlink(inode, 2);
1765 inode->i_mode = S_IFDIR | S_IRUGO | S 1363 inode->i_mode = S_IFDIR | S_IRUGO | S_IXUGO;
1766 inode->i_uid = GLOBAL_ROOT_UID; 1364 inode->i_uid = GLOBAL_ROOT_UID;
1767 inode->i_gid = GLOBAL_ROOT_GID; 1365 inode->i_gid = GLOBAL_ROOT_GID;
1768 inode->i_rdev = 0; 1366 inode->i_rdev = 0;
1769 inode->i_size = 0; 1367 inode->i_size = 0;
1770 inode->i_blkbits = PAGE_SHIFT; 1368 inode->i_blkbits = PAGE_SHIFT;
1771 inode->i_blocks = 0; 1369 inode->i_blocks = 0;
1772 1370
1773 inode->i_op = &empty_dir_inode_operat 1371 inode->i_op = &empty_dir_inode_operations;
1774 inode->i_opflags &= ~IOP_XATTR; 1372 inode->i_opflags &= ~IOP_XATTR;
1775 inode->i_fop = &empty_dir_operations; 1373 inode->i_fop = &empty_dir_operations;
1776 } 1374 }
1777 1375
1778 bool is_empty_dir_inode(struct inode *inode) 1376 bool is_empty_dir_inode(struct inode *inode)
1779 { 1377 {
1780 return (inode->i_fop == &empty_dir_op 1378 return (inode->i_fop == &empty_dir_operations) &&
1781 (inode->i_op == &empty_dir_in 1379 (inode->i_op == &empty_dir_inode_operations);
1782 } 1380 }
1783 1381
1784 #if IS_ENABLED(CONFIG_UNICODE) 1382 #if IS_ENABLED(CONFIG_UNICODE)
>> 1383 /*
>> 1384 * Determine if the name of a dentry should be casefolded.
>> 1385 *
>> 1386 * Return: if names will need casefolding
>> 1387 */
>> 1388 static bool needs_casefold(const struct inode *dir)
>> 1389 {
>> 1390 return IS_CASEFOLDED(dir) && dir->i_sb->s_encoding;
>> 1391 }
>> 1392
1785 /** 1393 /**
1786 * generic_ci_d_compare - generic d_compare i 1394 * generic_ci_d_compare - generic d_compare implementation for casefolding filesystems
1787 * @dentry: dentry whose name we are chec 1395 * @dentry: dentry whose name we are checking against
1788 * @len: len of name of dentry 1396 * @len: len of name of dentry
1789 * @str: str pointer to name of dentry 1397 * @str: str pointer to name of dentry
1790 * @name: Name to compare against 1398 * @name: Name to compare against
1791 * 1399 *
1792 * Return: 0 if names match, 1 if mismatch, o 1400 * Return: 0 if names match, 1 if mismatch, or -ERRNO
1793 */ 1401 */
1794 static int generic_ci_d_compare(const struct 1402 static int generic_ci_d_compare(const struct dentry *dentry, unsigned int len,
1795 const char *s 1403 const char *str, const struct qstr *name)
1796 { 1404 {
1797 const struct dentry *parent; !! 1405 const struct dentry *parent = READ_ONCE(dentry->d_parent);
1798 const struct inode *dir; !! 1406 const struct inode *dir = READ_ONCE(parent->d_inode);
>> 1407 const struct super_block *sb = dentry->d_sb;
>> 1408 const struct unicode_map *um = sb->s_encoding;
>> 1409 struct qstr qstr = QSTR_INIT(str, len);
1799 char strbuf[DNAME_INLINE_LEN]; 1410 char strbuf[DNAME_INLINE_LEN];
1800 struct qstr qstr; !! 1411 int ret;
1801 <<
1802 /* <<
1803 * Attempt a case-sensitive match fir <<
1804 * should cover most lookups, includi <<
1805 * applications that expect a case-se <<
1806 * <<
1807 * This comparison is safe under RCU <<
1808 * guarantees the consistency between <<
1809 * __d_lookup_rcu_op_compare() for de <<
1810 */ <<
1811 if (len == name->len && !memcmp(str, <<
1812 return 0; <<
1813 <<
1814 parent = READ_ONCE(dentry->d_parent); <<
1815 dir = READ_ONCE(parent->d_inode); <<
1816 if (!dir || !IS_CASEFOLDED(dir)) <<
1817 return 1; <<
1818 1412
>> 1413 if (!dir || !needs_casefold(dir))
>> 1414 goto fallback;
1819 /* 1415 /*
1820 * If the dentry name is stored in-li 1416 * If the dentry name is stored in-line, then it may be concurrently
1821 * modified by a rename. If this hap 1417 * modified by a rename. If this happens, the VFS will eventually retry
1822 * the lookup, so it doesn't matter w 1418 * the lookup, so it doesn't matter what ->d_compare() returns.
1823 * However, it's unsafe to call utf8_ 1419 * However, it's unsafe to call utf8_strncasecmp() with an unstable
1824 * string. Therefore, we have to cop 1420 * string. Therefore, we have to copy the name into a temporary buffer.
1825 */ 1421 */
1826 if (len <= DNAME_INLINE_LEN - 1) { 1422 if (len <= DNAME_INLINE_LEN - 1) {
1827 memcpy(strbuf, str, len); 1423 memcpy(strbuf, str, len);
1828 strbuf[len] = 0; 1424 strbuf[len] = 0;
1829 str = strbuf; !! 1425 qstr.name = strbuf;
1830 /* prevent compiler from opti 1426 /* prevent compiler from optimizing out the temporary buffer */
1831 barrier(); 1427 barrier();
1832 } 1428 }
1833 qstr.len = len; !! 1429 ret = utf8_strncasecmp(um, name, &qstr);
1834 qstr.name = str; !! 1430 if (ret >= 0)
>> 1431 return ret;
1835 1432
1836 return utf8_strncasecmp(dentry->d_sb- !! 1433 if (sb_has_strict_encoding(sb))
>> 1434 return -EINVAL;
>> 1435 fallback:
>> 1436 if (len != name->len)
>> 1437 return 1;
>> 1438 return !!memcmp(str, name->name, len);
1837 } 1439 }
1838 1440
1839 /** 1441 /**
1840 * generic_ci_d_hash - generic d_hash impleme 1442 * generic_ci_d_hash - generic d_hash implementation for casefolding filesystems
1841 * @dentry: dentry of the parent director 1443 * @dentry: dentry of the parent directory
1842 * @str: qstr of name whose hash we sh 1444 * @str: qstr of name whose hash we should fill in
1843 * 1445 *
1844 * Return: 0 if hash was successful or unchan 1446 * Return: 0 if hash was successful or unchanged, and -EINVAL on error
1845 */ 1447 */
1846 static int generic_ci_d_hash(const struct den 1448 static int generic_ci_d_hash(const struct dentry *dentry, struct qstr *str)
1847 { 1449 {
1848 const struct inode *dir = READ_ONCE(d 1450 const struct inode *dir = READ_ONCE(dentry->d_inode);
1849 struct super_block *sb = dentry->d_sb 1451 struct super_block *sb = dentry->d_sb;
1850 const struct unicode_map *um = sb->s_ 1452 const struct unicode_map *um = sb->s_encoding;
1851 int ret; !! 1453 int ret = 0;
1852 1454
1853 if (!dir || !IS_CASEFOLDED(dir)) !! 1455 if (!dir || !needs_casefold(dir))
1854 return 0; 1456 return 0;
1855 1457
1856 ret = utf8_casefold_hash(um, dentry, 1458 ret = utf8_casefold_hash(um, dentry, str);
1857 if (ret < 0 && sb_has_strict_encoding 1459 if (ret < 0 && sb_has_strict_encoding(sb))
1858 return -EINVAL; 1460 return -EINVAL;
1859 return 0; 1461 return 0;
1860 } 1462 }
1861 1463
1862 static const struct dentry_operations generic 1464 static const struct dentry_operations generic_ci_dentry_ops = {
1863 .d_hash = generic_ci_d_hash, 1465 .d_hash = generic_ci_d_hash,
1864 .d_compare = generic_ci_d_compare, 1466 .d_compare = generic_ci_d_compare,
1865 #ifdef CONFIG_FS_ENCRYPTION <<
1866 .d_revalidate = fscrypt_d_revalidate, <<
1867 #endif <<
1868 }; 1467 };
1869 <<
1870 /** <<
1871 * generic_ci_match() - Match a name (case-in <<
1872 * This is a filesystem helper for comparison <<
1873 * generic_ci_d_compare should be used in VFS <<
1874 * <<
1875 * @parent: Inode of the parent of the dirent <<
1876 * @name: name under lookup. <<
1877 * @folded_name: Optional pre-folded name und <<
1878 * @de_name: Dirent name. <<
1879 * @de_name_len: dirent name length. <<
1880 * <<
1881 * Test whether a case-insensitive directory <<
1882 * being searched. If @folded_name is provid <<
1883 * recalculating the casefold of @name. <<
1884 * <<
1885 * Return: > 0 if the directory entry matches <<
1886 * < 0 on error. <<
1887 */ <<
1888 int generic_ci_match(const struct inode *pare <<
1889 const struct qstr *name, <<
1890 const struct qstr *folde <<
1891 const u8 *de_name, u32 d <<
1892 { <<
1893 const struct super_block *sb = parent <<
1894 const struct unicode_map *um = sb->s_ <<
1895 struct fscrypt_str decrypted_name = F <<
1896 struct qstr dirent = QSTR_INIT(de_nam <<
1897 int res = 0; <<
1898 <<
1899 if (IS_ENCRYPTED(parent)) { <<
1900 const struct fscrypt_str encr <<
1901 FSTR_INIT((u8 *) de_n <<
1902 <<
1903 if (WARN_ON_ONCE(!fscrypt_has <<
1904 return -EINVAL; <<
1905 <<
1906 decrypted_name.name = kmalloc <<
1907 if (!decrypted_name.name) <<
1908 return -ENOMEM; <<
1909 res = fscrypt_fname_disk_to_u <<
1910 <<
1911 if (res < 0) { <<
1912 kfree(decrypted_name. <<
1913 return res; <<
1914 } <<
1915 dirent.name = decrypted_name. <<
1916 dirent.len = decrypted_name.l <<
1917 } <<
1918 <<
1919 /* <<
1920 * Attempt a case-sensitive match fir <<
1921 * should cover most lookups, includi <<
1922 * applications that expect a case-se <<
1923 */ <<
1924 <<
1925 if (dirent.len == name->len && <<
1926 !memcmp(name->name, dirent.name, <<
1927 goto out; <<
1928 <<
1929 if (folded_name->name) <<
1930 res = utf8_strncasecmp_folded <<
1931 else <<
1932 res = utf8_strncasecmp(um, na <<
1933 <<
1934 out: <<
1935 kfree(decrypted_name.name); <<
1936 if (res < 0 && sb_has_strict_encoding <<
1937 pr_err_ratelimited("Directory <<
1938 return 0; <<
1939 } <<
1940 return !res; <<
1941 } <<
1942 EXPORT_SYMBOL(generic_ci_match); <<
1943 #endif 1468 #endif
1944 1469
1945 #ifdef CONFIG_FS_ENCRYPTION 1470 #ifdef CONFIG_FS_ENCRYPTION
1946 static const struct dentry_operations generic 1471 static const struct dentry_operations generic_encrypted_dentry_ops = {
1947 .d_revalidate = fscrypt_d_revalidate, 1472 .d_revalidate = fscrypt_d_revalidate,
1948 }; 1473 };
1949 #endif 1474 #endif
1950 1475
>> 1476 #if defined(CONFIG_FS_ENCRYPTION) && IS_ENABLED(CONFIG_UNICODE)
>> 1477 static const struct dentry_operations generic_encrypted_ci_dentry_ops = {
>> 1478 .d_hash = generic_ci_d_hash,
>> 1479 .d_compare = generic_ci_d_compare,
>> 1480 .d_revalidate = fscrypt_d_revalidate,
>> 1481 };
>> 1482 #endif
>> 1483
1951 /** 1484 /**
1952 * generic_set_sb_d_ops - helper for choosing !! 1485 * generic_set_encrypted_ci_d_ops - helper for setting d_ops for given dentry
1953 * filesystem-wide dentry operations for the !! 1486 * @dentry: dentry to set ops on
1954 * @sb: superblock to be configured !! 1487 *
1955 * !! 1488 * Casefolded directories need d_hash and d_compare set, so that the dentries
1956 * Filesystems supporting casefolding and/or !! 1489 * contained in them are handled case-insensitively. Note that these operations
1957 * helper at mount-time to configure sb->s_d_ !! 1490 * are needed on the parent directory rather than on the dentries in it, and
1958 * operations required for the enabled featur !! 1491 * while the casefolding flag can be toggled on and off on an empty directory,
1959 * called after these have been configured, b !! 1492 * dentry_operations can't be changed later. As a result, if the filesystem has
1960 * is created. !! 1493 * casefolding support enabled at all, we have to give all dentries the
>> 1494 * casefolding operations even if their inode doesn't have the casefolding flag
>> 1495 * currently (and thus the casefolding ops would be no-ops for now).
>> 1496 *
>> 1497 * Encryption works differently in that the only dentry operation it needs is
>> 1498 * d_revalidate, which it only needs on dentries that have the no-key name flag.
>> 1499 * The no-key flag can't be set "later", so we don't have to worry about that.
>> 1500 *
>> 1501 * Finally, to maximize compatibility with overlayfs (which isn't compatible
>> 1502 * with certain dentry operations) and to avoid taking an unnecessary
>> 1503 * performance hit, we use custom dentry_operations for each possible
>> 1504 * combination rather than always installing all operations.
1961 */ 1505 */
1962 void generic_set_sb_d_ops(struct super_block !! 1506 void generic_set_encrypted_ci_d_ops(struct dentry *dentry)
1963 { 1507 {
>> 1508 #ifdef CONFIG_FS_ENCRYPTION
>> 1509 bool needs_encrypt_ops = dentry->d_flags & DCACHE_NOKEY_NAME;
>> 1510 #endif
1964 #if IS_ENABLED(CONFIG_UNICODE) 1511 #if IS_ENABLED(CONFIG_UNICODE)
1965 if (sb->s_encoding) { !! 1512 bool needs_ci_ops = dentry->d_sb->s_encoding;
1966 sb->s_d_op = &generic_ci_dent !! 1513 #endif
>> 1514 #if defined(CONFIG_FS_ENCRYPTION) && IS_ENABLED(CONFIG_UNICODE)
>> 1515 if (needs_encrypt_ops && needs_ci_ops) {
>> 1516 d_set_d_op(dentry, &generic_encrypted_ci_dentry_ops);
1967 return; 1517 return;
1968 } 1518 }
1969 #endif 1519 #endif
1970 #ifdef CONFIG_FS_ENCRYPTION 1520 #ifdef CONFIG_FS_ENCRYPTION
1971 if (sb->s_cop) { !! 1521 if (needs_encrypt_ops) {
1972 sb->s_d_op = &generic_encrypt !! 1522 d_set_d_op(dentry, &generic_encrypted_dentry_ops);
1973 return; 1523 return;
1974 } 1524 }
1975 #endif 1525 #endif
1976 } !! 1526 #if IS_ENABLED(CONFIG_UNICODE)
1977 EXPORT_SYMBOL(generic_set_sb_d_ops); !! 1527 if (needs_ci_ops) {
1978 !! 1528 d_set_d_op(dentry, &generic_ci_dentry_ops);
1979 /** <<
1980 * inode_maybe_inc_iversion - increments i_ve <<
1981 * @inode: inode with the i_version that shou <<
1982 * @force: increment the counter even if it's <<
1983 * <<
1984 * Every time the inode is modified, the i_ve <<
1985 * changed by any observer. <<
1986 * <<
1987 * If "force" is set or the QUERIED flag is s <<
1988 * the value, and clear the queried flag. <<
1989 * <<
1990 * In the common case where neither is set, t <<
1991 * updating i_version. <<
1992 * <<
1993 * If this function returns false, and no oth <<
1994 * can avoid logging the metadata. <<
1995 */ <<
1996 bool inode_maybe_inc_iversion(struct inode *i <<
1997 { <<
1998 u64 cur, new; <<
1999 <<
2000 /* <<
2001 * The i_version field is not strictl <<
2002 * information, but the legacy inode_ <<
2003 * to serialize increments. <<
2004 * <<
2005 * We add a full memory barrier to en <<
2006 * with other state is preserved (eit <<
2007 * or explicitly from smp_mb if we do <<
2008 * the former). <<
2009 * <<
2010 * These barriers pair with inode_que <<
2011 */ <<
2012 cur = inode_peek_iversion_raw(inode); <<
2013 if (!force && !(cur & I_VERSION_QUERI <<
2014 smp_mb(); <<
2015 cur = inode_peek_iversion_raw <<
2016 } <<
2017 <<
2018 do { <<
2019 /* If flag is clear then we n <<
2020 if (!force && !(cur & I_VERSI <<
2021 return false; <<
2022 <<
2023 /* Since lowest bit is flag, <<
2024 new = (cur & ~I_VERSION_QUERI <<
2025 } while (!atomic64_try_cmpxchg(&inode <<
2026 return true; <<
2027 } <<
2028 EXPORT_SYMBOL(inode_maybe_inc_iversion); <<
2029 <<
2030 /** <<
2031 * inode_query_iversion - read i_version for <<
2032 * @inode: inode from which i_version should <<
2033 * <<
2034 * Read the inode i_version counter. This sho <<
2035 * to store the returned i_version for later <<
2036 * that a later query of the i_version will r <<
2037 * anything has changed. <<
2038 * <<
2039 * In this implementation, we fetch the curre <<
2040 * then try to swap it into place with a cmpx <<
2041 * that fails, we try again with the newly fe <<
2042 */ <<
2043 u64 inode_query_iversion(struct inode *inode) <<
2044 { <<
2045 u64 cur, new; <<
2046 bool fenced = false; <<
2047 <<
2048 /* <<
2049 * Memory barriers (implicit in cmpxc <<
2050 * inode_maybe_inc_iversion(), see th <<
2051 */ <<
2052 cur = inode_peek_iversion_raw(inode); <<
2053 do { <<
2054 /* If flag is already set, th <<
2055 if (cur & I_VERSION_QUERIED) <<
2056 if (!fenced) <<
2057 smp_mb(); <<
2058 break; <<
2059 } <<
2060 <<
2061 fenced = true; <<
2062 new = cur | I_VERSION_QUERIED <<
2063 } while (!atomic64_try_cmpxchg(&inode <<
2064 return cur >> I_VERSION_QUERIED_SHIFT <<
2065 } <<
2066 EXPORT_SYMBOL(inode_query_iversion); <<
2067 <<
2068 ssize_t direct_write_fallback(struct kiocb *i <<
2069 ssize_t direct_written, ssize <<
2070 { <<
2071 struct address_space *mapping = iocb- <<
2072 loff_t pos = iocb->ki_pos - buffered_ <<
2073 loff_t end = iocb->ki_pos - 1; <<
2074 int err; <<
2075 <<
2076 /* <<
2077 * If the buffered write fallback ret <<
2078 * the number of bytes which were wri <<
2079 * code if that was zero. <<
2080 * <<
2081 * Note that this differs from normal <<
2082 * return -EFOO even if some bytes we <<
2083 */ <<
2084 if (unlikely(buffered_written < 0)) { <<
2085 if (direct_written) <<
2086 return direct_written <<
2087 return buffered_written; <<
2088 } <<
2089 <<
2090 /* <<
2091 * We need to ensure that the page ca <<
2092 * invalidated to preserve the expect <<
2093 */ <<
2094 err = filemap_write_and_wait_range(ma <<
2095 if (err < 0) { <<
2096 /* <<
2097 * We don't know how much we <<
2098 * bytes which were direct-wr <<
2099 */ <<
2100 iocb->ki_pos -= buffered_writ <<
2101 if (direct_written) <<
2102 return direct_written <<
2103 return err; <<
2104 } <<
2105 invalidate_mapping_pages(mapping, pos <<
2106 return direct_written + buffered_writ <<
2107 } <<
2108 EXPORT_SYMBOL_GPL(direct_write_fallback); <<
2109 <<
2110 /** <<
2111 * simple_inode_init_ts - initialize the time <<
2112 * @inode: inode to be initialized <<
2113 * <<
2114 * When a new inode is created, most filesyst <<
2115 * current time. Add a helper to do this. <<
2116 */ <<
2117 struct timespec64 simple_inode_init_ts(struct <<
2118 { <<
2119 struct timespec64 ts = inode_set_ctim <<
2120 <<
2121 inode_set_atime_to_ts(inode, ts); <<
2122 inode_set_mtime_to_ts(inode, ts); <<
2123 return ts; <<
2124 } <<
2125 EXPORT_SYMBOL(simple_inode_init_ts); <<
2126 <<
2127 static inline struct dentry *get_stashed_dent <<
2128 { <<
2129 struct dentry *dentry; <<
2130 <<
2131 guard(rcu)(); <<
2132 dentry = rcu_dereference(*stashed); <<
2133 if (!dentry) <<
2134 return NULL; <<
2135 if (!lockref_get_not_dead(&dentry->d_ <<
2136 return NULL; <<
2137 return dentry; <<
2138 } <<
2139 <<
2140 static struct dentry *prepare_anon_dentry(str <<
2141 str <<
2142 voi <<
2143 { <<
2144 struct dentry *dentry; <<
2145 struct inode *inode; <<
2146 const struct stashed_operations *sops <<
2147 int ret; <<
2148 <<
2149 inode = new_inode_pseudo(sb); <<
2150 if (!inode) { <<
2151 sops->put_data(data); <<
2152 return ERR_PTR(-ENOMEM); <<
2153 } <<
2154 <<
2155 inode->i_flags |= S_IMMUTABLE; <<
2156 inode->i_mode = S_IFREG; <<
2157 simple_inode_init_ts(inode); <<
2158 <<
2159 ret = sops->init_inode(inode, data); <<
2160 if (ret < 0) { <<
2161 iput(inode); <<
2162 return ERR_PTR(ret); <<
2163 } <<
2164 <<
2165 /* Notice when this is changed. */ <<
2166 WARN_ON_ONCE(!S_ISREG(inode->i_mode)) <<
2167 WARN_ON_ONCE(!IS_IMMUTABLE(inode)); <<
2168 <<
2169 dentry = d_alloc_anon(sb); <<
2170 if (!dentry) { <<
2171 iput(inode); <<
2172 return ERR_PTR(-ENOMEM); <<
2173 } <<
2174 <<
2175 /* Store address of location where de <<
2176 dentry->d_fsdata = stashed; <<
2177 <<
2178 /* @data is now owned by the fs */ <<
2179 d_instantiate(dentry, inode); <<
2180 return dentry; <<
2181 } <<
2182 <<
2183 static struct dentry *stash_dentry(struct den <<
2184 struct den <<
2185 { <<
2186 guard(rcu)(); <<
2187 for (;;) { <<
2188 struct dentry *old; <<
2189 <<
2190 /* Assume any old dentry was <<
2191 old = cmpxchg(stashed, NULL, <<
2192 if (likely(!old)) <<
2193 return dentry; <<
2194 <<
2195 /* Check if somebody else ins <<
2196 if (lockref_get_not_dead(&old <<
2197 return old; <<
2198 <<
2199 /* There's an old dead dentry <<
2200 if (likely(try_cmpxchg(stashe <<
2201 return dentry; <<
2202 } <<
2203 } <<
2204 <<
2205 /** <<
2206 * path_from_stashed - create path from stash <<
2207 * @stashed: where to retrieve or stash de <<
2208 * @mnt: mnt of the filesystems to use <<
2209 * @data: data to store in inode->i_pri <<
2210 * @path: path to create <<
2211 * <<
2212 * The function tries to retrieve a stashed d <<
2213 * is still valid then it will be reused. If <<
2214 * will allocate a new dentry and inode. It w <<
2215 * can reuse an existing dentry in case one h <<
2216 * update @stashed with the newly added dentr <<
2217 * <<
2218 * Special-purpose helper for nsfs and pidfs. <<
2219 * <<
2220 * Return: On success zero and on failure a n <<
2221 */ <<
2222 int path_from_stashed(struct dentry **stashed <<
2223 struct path *path) <<
2224 { <<
2225 struct dentry *dentry; <<
2226 const struct stashed_operations *sops <<
2227 <<
2228 /* See if dentry can be reused. */ <<
2229 path->dentry = get_stashed_dentry(sta <<
2230 if (path->dentry) { <<
2231 sops->put_data(data); <<
2232 goto out_path; <<
2233 } <<
2234 <<
2235 /* Allocate a new dentry. */ <<
2236 dentry = prepare_anon_dentry(stashed, <<
2237 if (IS_ERR(dentry)) <<
2238 return PTR_ERR(dentry); <<
2239 <<
2240 /* Added a new dentry. @data is now o <<
2241 path->dentry = stash_dentry(stashed, <<
2242 if (path->dentry != dentry) <<
2243 dput(dentry); <<
2244 <<
2245 out_path: <<
2246 WARN_ON_ONCE(path->dentry->d_fsdata ! <<
2247 WARN_ON_ONCE(d_inode(path->dentry)->i <<
2248 path->mnt = mntget(mnt); <<
2249 return 0; <<
2250 } <<
2251 <<
2252 void stashed_dentry_prune(struct dentry *dent <<
2253 { <<
2254 struct dentry **stashed = dentry->d_f <<
2255 struct inode *inode = d_inode(dentry) <<
2256 <<
2257 if (WARN_ON_ONCE(!stashed)) <<
2258 return; <<
2259 <<
2260 if (!inode) <<
2261 return; 1529 return;
2262 !! 1530 }
2263 /* !! 1531 #endif
2264 * Only replace our own @dentry as so <<
2265 * already cleared out @dentry and st <<
2266 * dentry in there. <<
2267 */ <<
2268 cmpxchg(stashed, dentry, NULL); <<
2269 } 1532 }
>> 1533 EXPORT_SYMBOL(generic_set_encrypted_ci_d_ops);
2270 1534
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.