| 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> |
| 21 | #include <linux/gfp.h> |
| 22 | #include <linux/init.h> |
| 23 | #include <linux/module.h> |
| 24 | #include <linux/mount.h> |
| 25 | #include <linux/srcu.h> |
| 26 | |
| 27 | #include <linux/fsnotify_backend.h> |
| 28 | #include "fsnotify.h" |
| 29 | |
| 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 | |
| 39 | void __fsnotify_vfsmount_delete(struct vfsmount *mnt) |
| 40 | { |
| 41 | fsnotify_clear_marks_by_mount(mnt); |
| 42 | } |
| 43 | |
| 44 | /* |
| 45 | * Given an inode, first check if we care what happens to our children. Inotify |
| 46 | * and dnotify both tell their parents about events. If we care about any event |
| 47 | * on a child we run all of our children and set a dentry flag saying that the |
| 48 | * parent cares. Thus when an event happens on a child it can quickly tell if |
| 49 | * if there is a need to find a parent and send the event to the parent. |
| 50 | */ |
| 51 | void __fsnotify_update_child_dentry_flags(struct inode *inode) |
| 52 | { |
| 53 | struct dentry *alias; |
| 54 | int watched; |
| 55 | |
| 56 | if (!S_ISDIR(inode->i_mode)) |
| 57 | return; |
| 58 | |
| 59 | /* determine if the children should tell inode about their events */ |
| 60 | watched = fsnotify_inode_watches_children(inode); |
| 61 | |
| 62 | spin_lock(&inode->i_lock); |
| 63 | /* run all of the dentries associated with this inode. Since this is a |
| 64 | * directory, there damn well better only be one item on this list */ |
| 65 | hlist_for_each_entry(alias, &inode->i_dentry, d_u.d_alias) { |
| 66 | struct dentry *child; |
| 67 | |
| 68 | /* run all of the children of the original inode and fix their |
| 69 | * d_flags to indicate parental interest (their parent is the |
| 70 | * original inode) */ |
| 71 | spin_lock(&alias->d_lock); |
| 72 | list_for_each_entry(child, &alias->d_subdirs, d_child) { |
| 73 | if (!child->d_inode) |
| 74 | continue; |
| 75 | |
| 76 | spin_lock_nested(&child->d_lock, DENTRY_D_LOCK_NESTED); |
| 77 | if (watched) |
| 78 | child->d_flags |= DCACHE_FSNOTIFY_PARENT_WATCHED; |
| 79 | else |
| 80 | child->d_flags &= ~DCACHE_FSNOTIFY_PARENT_WATCHED; |
| 81 | spin_unlock(&child->d_lock); |
| 82 | } |
| 83 | spin_unlock(&alias->d_lock); |
| 84 | } |
| 85 | spin_unlock(&inode->i_lock); |
| 86 | } |
| 87 | |
| 88 | /* Notify this dentry's parent about a child's events. */ |
| 89 | int __fsnotify_parent(struct path *path, struct dentry *dentry, __u32 mask) |
| 90 | { |
| 91 | struct dentry *parent; |
| 92 | struct inode *p_inode; |
| 93 | int ret = 0; |
| 94 | |
| 95 | if (!dentry) |
| 96 | dentry = path->dentry; |
| 97 | |
| 98 | if (!(dentry->d_flags & DCACHE_FSNOTIFY_PARENT_WATCHED)) |
| 99 | return 0; |
| 100 | |
| 101 | parent = dget_parent(dentry); |
| 102 | p_inode = parent->d_inode; |
| 103 | |
| 104 | if (unlikely(!fsnotify_inode_watches_children(p_inode))) |
| 105 | __fsnotify_update_child_dentry_flags(p_inode); |
| 106 | else if (p_inode->i_fsnotify_mask & mask) { |
| 107 | /* we are notifying a parent so come up with the new mask which |
| 108 | * specifies these are events which came from a child. */ |
| 109 | mask |= FS_EVENT_ON_CHILD; |
| 110 | |
| 111 | if (path) |
| 112 | ret = fsnotify(p_inode, mask, path, FSNOTIFY_EVENT_PATH, |
| 113 | dentry->d_name.name, 0); |
| 114 | else |
| 115 | ret = fsnotify(p_inode, mask, dentry->d_inode, FSNOTIFY_EVENT_INODE, |
| 116 | dentry->d_name.name, 0); |
| 117 | } |
| 118 | |
| 119 | dput(parent); |
| 120 | |
| 121 | return ret; |
| 122 | } |
| 123 | EXPORT_SYMBOL_GPL(__fsnotify_parent); |
| 124 | |
| 125 | static int send_to_group(struct inode *to_tell, |
| 126 | struct fsnotify_mark *inode_mark, |
| 127 | struct fsnotify_mark *vfsmount_mark, |
| 128 | __u32 mask, void *data, |
| 129 | int data_is, u32 cookie, |
| 130 | const unsigned char *file_name) |
| 131 | { |
| 132 | struct fsnotify_group *group = NULL; |
| 133 | __u32 inode_test_mask = 0; |
| 134 | __u32 vfsmount_test_mask = 0; |
| 135 | |
| 136 | if (unlikely(!inode_mark && !vfsmount_mark)) { |
| 137 | BUG(); |
| 138 | return 0; |
| 139 | } |
| 140 | |
| 141 | /* clear ignored on inode modification */ |
| 142 | if (mask & FS_MODIFY) { |
| 143 | if (inode_mark && |
| 144 | !(inode_mark->flags & FSNOTIFY_MARK_FLAG_IGNORED_SURV_MODIFY)) |
| 145 | inode_mark->ignored_mask = 0; |
| 146 | if (vfsmount_mark && |
| 147 | !(vfsmount_mark->flags & FSNOTIFY_MARK_FLAG_IGNORED_SURV_MODIFY)) |
| 148 | vfsmount_mark->ignored_mask = 0; |
| 149 | } |
| 150 | |
| 151 | /* does the inode mark tell us to do something? */ |
| 152 | if (inode_mark) { |
| 153 | group = inode_mark->group; |
| 154 | inode_test_mask = (mask & ~FS_EVENT_ON_CHILD); |
| 155 | inode_test_mask &= inode_mark->mask; |
| 156 | inode_test_mask &= ~inode_mark->ignored_mask; |
| 157 | } |
| 158 | |
| 159 | /* does the vfsmount_mark tell us to do something? */ |
| 160 | if (vfsmount_mark) { |
| 161 | vfsmount_test_mask = (mask & ~FS_EVENT_ON_CHILD); |
| 162 | group = vfsmount_mark->group; |
| 163 | vfsmount_test_mask &= vfsmount_mark->mask; |
| 164 | vfsmount_test_mask &= ~vfsmount_mark->ignored_mask; |
| 165 | if (inode_mark) |
| 166 | vfsmount_test_mask &= ~inode_mark->ignored_mask; |
| 167 | } |
| 168 | |
| 169 | pr_debug("%s: group=%p to_tell=%p mask=%x inode_mark=%p" |
| 170 | " inode_test_mask=%x vfsmount_mark=%p vfsmount_test_mask=%x" |
| 171 | " data=%p data_is=%d cookie=%d\n", |
| 172 | __func__, group, to_tell, mask, inode_mark, |
| 173 | inode_test_mask, vfsmount_mark, vfsmount_test_mask, data, |
| 174 | data_is, cookie); |
| 175 | |
| 176 | if (!inode_test_mask && !vfsmount_test_mask) |
| 177 | return 0; |
| 178 | |
| 179 | return group->ops->handle_event(group, to_tell, inode_mark, |
| 180 | vfsmount_mark, mask, data, data_is, |
| 181 | file_name, cookie); |
| 182 | } |
| 183 | |
| 184 | /* |
| 185 | * This is the main call to fsnotify. The VFS calls into hook specific functions |
| 186 | * in linux/fsnotify.h. Those functions then in turn call here. Here will call |
| 187 | * out to all of the registered fsnotify_group. Those groups can then use the |
| 188 | * notification event in whatever means they feel necessary. |
| 189 | */ |
| 190 | int fsnotify(struct inode *to_tell, __u32 mask, void *data, int data_is, |
| 191 | const unsigned char *file_name, u32 cookie) |
| 192 | { |
| 193 | struct hlist_node *inode_node = NULL, *vfsmount_node = NULL; |
| 194 | struct fsnotify_mark *inode_mark = NULL, *vfsmount_mark = NULL; |
| 195 | struct fsnotify_group *inode_group, *vfsmount_group; |
| 196 | struct mount *mnt; |
| 197 | int idx, ret = 0; |
| 198 | /* global tests shouldn't care about events on child only the specific event */ |
| 199 | __u32 test_mask = (mask & ~FS_EVENT_ON_CHILD); |
| 200 | |
| 201 | if (data_is == FSNOTIFY_EVENT_PATH) |
| 202 | mnt = real_mount(((struct path *)data)->mnt); |
| 203 | else |
| 204 | mnt = NULL; |
| 205 | |
| 206 | /* |
| 207 | * Optimization: srcu_read_lock() has a memory barrier which can |
| 208 | * be expensive. It protects walking the *_fsnotify_marks lists. |
| 209 | * However, if we do not walk the lists, we do not have to do |
| 210 | * SRCU because we have no references to any objects and do not |
| 211 | * need SRCU to keep them "alive". |
| 212 | */ |
| 213 | if (hlist_empty(&to_tell->i_fsnotify_marks) && |
| 214 | (!mnt || hlist_empty(&mnt->mnt_fsnotify_marks))) |
| 215 | return 0; |
| 216 | /* |
| 217 | * if this is a modify event we may need to clear the ignored masks |
| 218 | * otherwise return if neither the inode nor the vfsmount care about |
| 219 | * this type of event. |
| 220 | */ |
| 221 | if (!(mask & FS_MODIFY) && |
| 222 | !(test_mask & to_tell->i_fsnotify_mask) && |
| 223 | !(mnt && test_mask & mnt->mnt_fsnotify_mask)) |
| 224 | return 0; |
| 225 | |
| 226 | idx = srcu_read_lock(&fsnotify_mark_srcu); |
| 227 | |
| 228 | if ((mask & FS_MODIFY) || |
| 229 | (test_mask & to_tell->i_fsnotify_mask)) |
| 230 | inode_node = srcu_dereference(to_tell->i_fsnotify_marks.first, |
| 231 | &fsnotify_mark_srcu); |
| 232 | |
| 233 | if (mnt && ((mask & FS_MODIFY) || |
| 234 | (test_mask & mnt->mnt_fsnotify_mask))) { |
| 235 | vfsmount_node = srcu_dereference(mnt->mnt_fsnotify_marks.first, |
| 236 | &fsnotify_mark_srcu); |
| 237 | inode_node = srcu_dereference(to_tell->i_fsnotify_marks.first, |
| 238 | &fsnotify_mark_srcu); |
| 239 | } |
| 240 | |
| 241 | /* |
| 242 | * We need to merge inode & vfsmount mark lists so that inode mark |
| 243 | * ignore masks are properly reflected for mount mark notifications. |
| 244 | * That's why this traversal is so complicated... |
| 245 | */ |
| 246 | while (inode_node || vfsmount_node) { |
| 247 | inode_group = NULL; |
| 248 | inode_mark = NULL; |
| 249 | vfsmount_group = NULL; |
| 250 | vfsmount_mark = NULL; |
| 251 | |
| 252 | if (inode_node) { |
| 253 | inode_mark = hlist_entry(srcu_dereference(inode_node, &fsnotify_mark_srcu), |
| 254 | struct fsnotify_mark, obj_list); |
| 255 | inode_group = inode_mark->group; |
| 256 | } |
| 257 | |
| 258 | if (vfsmount_node) { |
| 259 | vfsmount_mark = hlist_entry(srcu_dereference(vfsmount_node, &fsnotify_mark_srcu), |
| 260 | struct fsnotify_mark, obj_list); |
| 261 | vfsmount_group = vfsmount_mark->group; |
| 262 | } |
| 263 | |
| 264 | if (inode_group && vfsmount_group) { |
| 265 | int cmp = fsnotify_compare_groups(inode_group, |
| 266 | vfsmount_group); |
| 267 | if (cmp > 0) { |
| 268 | inode_group = NULL; |
| 269 | inode_mark = NULL; |
| 270 | } else if (cmp < 0) { |
| 271 | vfsmount_group = NULL; |
| 272 | vfsmount_mark = NULL; |
| 273 | } |
| 274 | } |
| 275 | ret = send_to_group(to_tell, inode_mark, vfsmount_mark, mask, |
| 276 | data, data_is, cookie, file_name); |
| 277 | |
| 278 | if (ret && (mask & ALL_FSNOTIFY_PERM_EVENTS)) |
| 279 | goto out; |
| 280 | |
| 281 | if (inode_group) |
| 282 | inode_node = srcu_dereference(inode_node->next, |
| 283 | &fsnotify_mark_srcu); |
| 284 | if (vfsmount_group) |
| 285 | vfsmount_node = srcu_dereference(vfsmount_node->next, |
| 286 | &fsnotify_mark_srcu); |
| 287 | } |
| 288 | ret = 0; |
| 289 | out: |
| 290 | srcu_read_unlock(&fsnotify_mark_srcu, idx); |
| 291 | |
| 292 | return ret; |
| 293 | } |
| 294 | EXPORT_SYMBOL_GPL(fsnotify); |
| 295 | |
| 296 | static __init int fsnotify_init(void) |
| 297 | { |
| 298 | int ret; |
| 299 | |
| 300 | BUG_ON(hweight32(ALL_FSNOTIFY_EVENTS) != 23); |
| 301 | |
| 302 | ret = init_srcu_struct(&fsnotify_mark_srcu); |
| 303 | if (ret) |
| 304 | panic("initializing fsnotify_mark_srcu"); |
| 305 | |
| 306 | return 0; |
| 307 | } |
| 308 | core_initcall(fsnotify_init); |