]>
Commit | Line | Data |
---|---|---|
90586523 EP |
1 | /* |
2 | * Copyright (C) 2008 Red Hat, Inc., Eric Paris <eparis@redhat.com> | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or modify | |
5 | * it under the terms of the GNU General Public License as published by | |
6 | * the Free Software Foundation; either version 2, or (at your option) | |
7 | * any later version. | |
8 | * | |
9 | * This program is distributed in the hope that it will be useful, | |
10 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
11 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
12 | * GNU General Public License for more details. | |
13 | * | |
14 | * You should have received a copy of the GNU General Public License | |
15 | * along with this program; see the file COPYING. If not, write to | |
16 | * the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. | |
17 | */ | |
18 | ||
19 | #include <linux/dcache.h> | |
20 | #include <linux/fs.h> | |
5a0e3ad6 | 21 | #include <linux/gfp.h> |
90586523 EP |
22 | #include <linux/init.h> |
23 | #include <linux/module.h> | |
7131485a | 24 | #include <linux/mount.h> |
90586523 EP |
25 | #include <linux/srcu.h> |
26 | ||
27 | #include <linux/fsnotify_backend.h> | |
28 | #include "fsnotify.h" | |
29 | ||
3be25f49 EP |
30 | /* |
31 | * Clear all of the marks on an inode when it is being evicted from core | |
32 | */ | |
33 | void __fsnotify_inode_delete(struct inode *inode) | |
34 | { | |
35 | fsnotify_clear_marks_by_inode(inode); | |
36 | } | |
37 | EXPORT_SYMBOL_GPL(__fsnotify_inode_delete); | |
38 | ||
ca9c726e AG |
39 | void __fsnotify_vfsmount_delete(struct vfsmount *mnt) |
40 | { | |
41 | fsnotify_clear_marks_by_mount(mnt); | |
42 | } | |
43 | ||
ebb3b47e JK |
44 | /** |
45 | * fsnotify_unmount_inodes - an sb is unmounting. handle any watched inodes. | |
46 | * @sb: superblock being unmounted. | |
47 | * | |
48 | * Called during unmount with no locks held, so needs to be safe against | |
49 | * concurrent modifiers. We temporarily drop sb->s_inode_list_lock and CAN block. | |
50 | */ | |
1e6cb723 | 51 | static void fsnotify_unmount_inodes(struct super_block *sb) |
ebb3b47e JK |
52 | { |
53 | struct inode *inode, *iput_inode = NULL; | |
54 | ||
55 | spin_lock(&sb->s_inode_list_lock); | |
56 | list_for_each_entry(inode, &sb->s_inodes, i_sb_list) { | |
57 | /* | |
58 | * We cannot __iget() an inode in state I_FREEING, | |
59 | * I_WILL_FREE, or I_NEW which is fine because by that point | |
60 | * the inode cannot have any associated watches. | |
61 | */ | |
62 | spin_lock(&inode->i_lock); | |
63 | if (inode->i_state & (I_FREEING|I_WILL_FREE|I_NEW)) { | |
64 | spin_unlock(&inode->i_lock); | |
65 | continue; | |
66 | } | |
67 | ||
68 | /* | |
69 | * If i_count is zero, the inode cannot have any watches and | |
1751e8a6 | 70 | * doing an __iget/iput with SB_ACTIVE clear would actually |
ebb3b47e JK |
71 | * evict all inodes with zero i_count from icache which is |
72 | * unnecessarily violent and may in fact be illegal to do. | |
73 | */ | |
74 | if (!atomic_read(&inode->i_count)) { | |
75 | spin_unlock(&inode->i_lock); | |
76 | continue; | |
77 | } | |
78 | ||
79 | __iget(inode); | |
80 | spin_unlock(&inode->i_lock); | |
81 | spin_unlock(&sb->s_inode_list_lock); | |
82 | ||
83 | if (iput_inode) | |
84 | iput(iput_inode); | |
85 | ||
86 | /* for each watch, send FS_UNMOUNT and then remove it */ | |
87 | fsnotify(inode, FS_UNMOUNT, inode, FSNOTIFY_EVENT_INODE, NULL, 0); | |
88 | ||
89 | fsnotify_inode_delete(inode); | |
90 | ||
91 | iput_inode = inode; | |
92 | ||
93 | spin_lock(&sb->s_inode_list_lock); | |
94 | } | |
95 | spin_unlock(&sb->s_inode_list_lock); | |
96 | ||
97 | if (iput_inode) | |
98 | iput(iput_inode); | |
99 | } | |
100 | ||
1e6cb723 AG |
101 | void fsnotify_sb_delete(struct super_block *sb) |
102 | { | |
103 | fsnotify_unmount_inodes(sb); | |
104 | fsnotify_clear_marks_by_sb(sb); | |
105 | } | |
106 | ||
c28f7e56 EP |
107 | /* |
108 | * Given an inode, first check if we care what happens to our children. Inotify | |
109 | * and dnotify both tell their parents about events. If we care about any event | |
110 | * on a child we run all of our children and set a dentry flag saying that the | |
111 | * parent cares. Thus when an event happens on a child it can quickly tell if | |
112 | * if there is a need to find a parent and send the event to the parent. | |
113 | */ | |
114 | void __fsnotify_update_child_dentry_flags(struct inode *inode) | |
115 | { | |
116 | struct dentry *alias; | |
117 | int watched; | |
118 | ||
119 | if (!S_ISDIR(inode->i_mode)) | |
120 | return; | |
121 | ||
122 | /* determine if the children should tell inode about their events */ | |
123 | watched = fsnotify_inode_watches_children(inode); | |
124 | ||
873feea0 | 125 | spin_lock(&inode->i_lock); |
c28f7e56 EP |
126 | /* run all of the dentries associated with this inode. Since this is a |
127 | * directory, there damn well better only be one item on this list */ | |
946e51f2 | 128 | hlist_for_each_entry(alias, &inode->i_dentry, d_u.d_alias) { |
c28f7e56 EP |
129 | struct dentry *child; |
130 | ||
131 | /* run all of the children of the original inode and fix their | |
132 | * d_flags to indicate parental interest (their parent is the | |
133 | * original inode) */ | |
2fd6b7f5 | 134 | spin_lock(&alias->d_lock); |
946e51f2 | 135 | list_for_each_entry(child, &alias->d_subdirs, d_child) { |
c28f7e56 EP |
136 | if (!child->d_inode) |
137 | continue; | |
138 | ||
2fd6b7f5 | 139 | spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED); |
c28f7e56 EP |
140 | if (watched) |
141 | child->d_flags |= DCACHE_FSNOTIFY_PARENT_WATCHED; | |
142 | else | |
143 | child->d_flags &= ~DCACHE_FSNOTIFY_PARENT_WATCHED; | |
144 | spin_unlock(&child->d_lock); | |
145 | } | |
2fd6b7f5 | 146 | spin_unlock(&alias->d_lock); |
c28f7e56 | 147 | } |
873feea0 | 148 | spin_unlock(&inode->i_lock); |
c28f7e56 EP |
149 | } |
150 | ||
151 | /* Notify this dentry's parent about a child's events. */ | |
12c7f9dc | 152 | int __fsnotify_parent(const struct path *path, struct dentry *dentry, __u32 mask) |
c28f7e56 EP |
153 | { |
154 | struct dentry *parent; | |
155 | struct inode *p_inode; | |
52420392 | 156 | int ret = 0; |
c28f7e56 | 157 | |
72acc854 | 158 | if (!dentry) |
2069601b | 159 | dentry = path->dentry; |
28c60e37 | 160 | |
c28f7e56 | 161 | if (!(dentry->d_flags & DCACHE_FSNOTIFY_PARENT_WATCHED)) |
52420392 | 162 | return 0; |
c28f7e56 | 163 | |
4d4eb366 | 164 | parent = dget_parent(dentry); |
c28f7e56 EP |
165 | p_inode = parent->d_inode; |
166 | ||
4d4eb366 CH |
167 | if (unlikely(!fsnotify_inode_watches_children(p_inode))) |
168 | __fsnotify_update_child_dentry_flags(p_inode); | |
169 | else if (p_inode->i_fsnotify_mask & mask) { | |
49d31c2f AV |
170 | struct name_snapshot name; |
171 | ||
c28f7e56 EP |
172 | /* we are notifying a parent so come up with the new mask which |
173 | * specifies these are events which came from a child. */ | |
174 | mask |= FS_EVENT_ON_CHILD; | |
175 | ||
49d31c2f | 176 | take_dentry_name_snapshot(&name, dentry); |
2069601b | 177 | if (path) |
52420392 | 178 | ret = fsnotify(p_inode, mask, path, FSNOTIFY_EVENT_PATH, |
49d31c2f | 179 | name.name, 0); |
28c60e37 | 180 | else |
52420392 | 181 | ret = fsnotify(p_inode, mask, dentry->d_inode, FSNOTIFY_EVENT_INODE, |
49d31c2f AV |
182 | name.name, 0); |
183 | release_dentry_name_snapshot(&name); | |
c28f7e56 EP |
184 | } |
185 | ||
4d4eb366 | 186 | dput(parent); |
52420392 EP |
187 | |
188 | return ret; | |
c28f7e56 EP |
189 | } |
190 | EXPORT_SYMBOL_GPL(__fsnotify_parent); | |
191 | ||
fd657170 | 192 | static int send_to_group(struct inode *to_tell, |
e637835e | 193 | __u32 mask, const void *data, |
613a807f | 194 | int data_is, u32 cookie, |
9385a84d JK |
195 | const unsigned char *file_name, |
196 | struct fsnotify_iter_info *iter_info) | |
7131485a | 197 | { |
faa9560a | 198 | struct fsnotify_group *group = NULL; |
92183a42 AG |
199 | __u32 test_mask = (mask & ~FS_EVENT_ON_CHILD); |
200 | __u32 marks_mask = 0; | |
201 | __u32 marks_ignored_mask = 0; | |
3dca1a74 AG |
202 | struct fsnotify_mark *mark; |
203 | int type; | |
613a807f | 204 | |
5b0457ad | 205 | if (WARN_ON(!iter_info->report_mask)) |
faa9560a | 206 | return 0; |
ce8f76fb EP |
207 | |
208 | /* clear ignored on inode modification */ | |
209 | if (mask & FS_MODIFY) { | |
3dca1a74 AG |
210 | fsnotify_foreach_obj_type(type) { |
211 | if (!fsnotify_iter_should_report_type(iter_info, type)) | |
212 | continue; | |
213 | mark = iter_info->marks[type]; | |
214 | if (mark && | |
215 | !(mark->flags & FSNOTIFY_MARK_FLAG_IGNORED_SURV_MODIFY)) | |
216 | mark->ignored_mask = 0; | |
217 | } | |
ce8f76fb | 218 | } |
613a807f | 219 | |
3dca1a74 AG |
220 | fsnotify_foreach_obj_type(type) { |
221 | if (!fsnotify_iter_should_report_type(iter_info, type)) | |
222 | continue; | |
223 | mark = iter_info->marks[type]; | |
224 | /* does the object mark tell us to do something? */ | |
225 | if (mark) { | |
226 | group = mark->group; | |
227 | marks_mask |= mark->mask; | |
228 | marks_ignored_mask |= mark->ignored_mask; | |
229 | } | |
ce8f76fb EP |
230 | } |
231 | ||
3dca1a74 | 232 | pr_debug("%s: group=%p to_tell=%p mask=%x marks_mask=%x marks_ignored_mask=%x" |
7053aee2 | 233 | " data=%p data_is=%d cookie=%d\n", |
3dca1a74 AG |
234 | __func__, group, to_tell, mask, marks_mask, marks_ignored_mask, |
235 | data, data_is, cookie); | |
faa9560a | 236 | |
92183a42 | 237 | if (!(test_mask & marks_mask & ~marks_ignored_mask)) |
613a807f EP |
238 | return 0; |
239 | ||
5b0457ad | 240 | return group->ops->handle_event(group, to_tell, mask, data, data_is, |
9385a84d | 241 | file_name, cookie, iter_info); |
7131485a EP |
242 | } |
243 | ||
3427ce71 MS |
244 | static struct fsnotify_mark *fsnotify_first_mark(struct fsnotify_mark_connector **connp) |
245 | { | |
246 | struct fsnotify_mark_connector *conn; | |
247 | struct hlist_node *node = NULL; | |
248 | ||
249 | conn = srcu_dereference(*connp, &fsnotify_mark_srcu); | |
250 | if (conn) | |
251 | node = srcu_dereference(conn->list.first, &fsnotify_mark_srcu); | |
252 | ||
253 | return hlist_entry_safe(node, struct fsnotify_mark, obj_list); | |
254 | } | |
255 | ||
256 | static struct fsnotify_mark *fsnotify_next_mark(struct fsnotify_mark *mark) | |
257 | { | |
258 | struct hlist_node *node = NULL; | |
259 | ||
260 | if (mark) | |
261 | node = srcu_dereference(mark->obj_list.next, | |
262 | &fsnotify_mark_srcu); | |
263 | ||
264 | return hlist_entry_safe(node, struct fsnotify_mark, obj_list); | |
265 | } | |
266 | ||
d9a6f30b AG |
267 | /* |
268 | * iter_info is a multi head priority queue of marks. | |
269 | * Pick a subset of marks from queue heads, all with the | |
270 | * same group and set the report_mask for selected subset. | |
271 | * Returns the report_mask of the selected subset. | |
272 | */ | |
273 | static unsigned int fsnotify_iter_select_report_types( | |
274 | struct fsnotify_iter_info *iter_info) | |
275 | { | |
47d9c7cc AG |
276 | struct fsnotify_group *max_prio_group = NULL; |
277 | struct fsnotify_mark *mark; | |
278 | int type; | |
279 | ||
280 | /* Choose max prio group among groups of all queue heads */ | |
281 | fsnotify_foreach_obj_type(type) { | |
282 | mark = iter_info->marks[type]; | |
283 | if (mark && | |
284 | fsnotify_compare_groups(max_prio_group, mark->group) > 0) | |
285 | max_prio_group = mark->group; | |
286 | } | |
d9a6f30b | 287 | |
47d9c7cc | 288 | if (!max_prio_group) |
d9a6f30b AG |
289 | return 0; |
290 | ||
47d9c7cc | 291 | /* Set the report mask for marks from same group as max prio group */ |
d9a6f30b | 292 | iter_info->report_mask = 0; |
47d9c7cc AG |
293 | fsnotify_foreach_obj_type(type) { |
294 | mark = iter_info->marks[type]; | |
295 | if (mark && | |
296 | fsnotify_compare_groups(max_prio_group, mark->group) == 0) | |
297 | fsnotify_iter_set_report_type(iter_info, type); | |
298 | } | |
d9a6f30b AG |
299 | |
300 | return iter_info->report_mask; | |
301 | } | |
302 | ||
303 | /* | |
304 | * Pop from iter_info multi head queue, the marks that were iterated in the | |
305 | * current iteration step. | |
306 | */ | |
307 | static void fsnotify_iter_next(struct fsnotify_iter_info *iter_info) | |
308 | { | |
47d9c7cc | 309 | int type; |
d9a6f30b | 310 | |
47d9c7cc AG |
311 | fsnotify_foreach_obj_type(type) { |
312 | if (fsnotify_iter_should_report_type(iter_info, type)) | |
313 | iter_info->marks[type] = | |
314 | fsnotify_next_mark(iter_info->marks[type]); | |
315 | } | |
d9a6f30b AG |
316 | } |
317 | ||
90586523 EP |
318 | /* |
319 | * This is the main call to fsnotify. The VFS calls into hook specific functions | |
320 | * in linux/fsnotify.h. Those functions then in turn call here. Here will call | |
321 | * out to all of the registered fsnotify_group. Those groups can then use the | |
322 | * notification event in whatever means they feel necessary. | |
323 | */ | |
e637835e | 324 | int fsnotify(struct inode *to_tell, __u32 mask, const void *data, int data_is, |
c4ec54b4 | 325 | const unsigned char *file_name, u32 cookie) |
90586523 | 326 | { |
3427ce71 | 327 | struct fsnotify_iter_info iter_info = {}; |
c63181e6 | 328 | struct mount *mnt; |
9385a84d | 329 | int ret = 0; |
e42e2773 EP |
330 | /* global tests shouldn't care about events on child only the specific event */ |
331 | __u32 test_mask = (mask & ~FS_EVENT_ON_CHILD); | |
90586523 | 332 | |
2069601b | 333 | if (data_is == FSNOTIFY_EVENT_PATH) |
e637835e | 334 | mnt = real_mount(((const struct path *)data)->mnt); |
613a807f EP |
335 | else |
336 | mnt = NULL; | |
337 | ||
7c49b861 DH |
338 | /* |
339 | * Optimization: srcu_read_lock() has a memory barrier which can | |
340 | * be expensive. It protects walking the *_fsnotify_marks lists. | |
341 | * However, if we do not walk the lists, we do not have to do | |
342 | * SRCU because we have no references to any objects and do not | |
343 | * need SRCU to keep them "alive". | |
344 | */ | |
9dd813c1 JK |
345 | if (!to_tell->i_fsnotify_marks && |
346 | (!mnt || !mnt->mnt_fsnotify_marks)) | |
7c49b861 | 347 | return 0; |
613a807f EP |
348 | /* |
349 | * if this is a modify event we may need to clear the ignored masks | |
350 | * otherwise return if neither the inode nor the vfsmount care about | |
351 | * this type of event. | |
352 | */ | |
353 | if (!(mask & FS_MODIFY) && | |
354 | !(test_mask & to_tell->i_fsnotify_mask) && | |
355 | !(mnt && test_mask & mnt->mnt_fsnotify_mask)) | |
356 | return 0; | |
3a9fb89f | 357 | |
9385a84d | 358 | iter_info.srcu_idx = srcu_read_lock(&fsnotify_mark_srcu); |
7131485a | 359 | |
9bdda4e9 AG |
360 | iter_info.marks[FSNOTIFY_OBJ_TYPE_INODE] = |
361 | fsnotify_first_mark(&to_tell->i_fsnotify_marks); | |
362 | if (mnt) { | |
47d9c7cc | 363 | iter_info.marks[FSNOTIFY_OBJ_TYPE_VFSMOUNT] = |
3427ce71 | 364 | fsnotify_first_mark(&mnt->mnt_fsnotify_marks); |
90586523 | 365 | } |
75c1be48 | 366 | |
8edc6e16 JK |
367 | /* |
368 | * We need to merge inode & vfsmount mark lists so that inode mark | |
369 | * ignore masks are properly reflected for mount mark notifications. | |
370 | * That's why this traversal is so complicated... | |
371 | */ | |
d9a6f30b | 372 | while (fsnotify_iter_select_report_types(&iter_info)) { |
5b0457ad AG |
373 | ret = send_to_group(to_tell, mask, data, data_is, cookie, |
374 | file_name, &iter_info); | |
613a807f | 375 | |
ff8bcbd0 EP |
376 | if (ret && (mask & ALL_FSNOTIFY_PERM_EVENTS)) |
377 | goto out; | |
378 | ||
d9a6f30b | 379 | fsnotify_iter_next(&iter_info); |
7131485a | 380 | } |
ff8bcbd0 EP |
381 | ret = 0; |
382 | out: | |
9385a84d | 383 | srcu_read_unlock(&fsnotify_mark_srcu, iter_info.srcu_idx); |
c4ec54b4 | 384 | |
98b5c10d | 385 | return ret; |
90586523 EP |
386 | } |
387 | EXPORT_SYMBOL_GPL(fsnotify); | |
388 | ||
9dd813c1 JK |
389 | extern struct kmem_cache *fsnotify_mark_connector_cachep; |
390 | ||
90586523 EP |
391 | static __init int fsnotify_init(void) |
392 | { | |
75c1be48 EP |
393 | int ret; |
394 | ||
20dee624 EP |
395 | BUG_ON(hweight32(ALL_FSNOTIFY_EVENTS) != 23); |
396 | ||
75c1be48 EP |
397 | ret = init_srcu_struct(&fsnotify_mark_srcu); |
398 | if (ret) | |
399 | panic("initializing fsnotify_mark_srcu"); | |
400 | ||
9dd813c1 JK |
401 | fsnotify_mark_connector_cachep = KMEM_CACHE(fsnotify_mark_connector, |
402 | SLAB_PANIC); | |
403 | ||
75c1be48 | 404 | return 0; |
90586523 | 405 | } |
75c1be48 | 406 | core_initcall(fsnotify_init); |