]>
Commit | Line | Data |
---|---|---|
b6450630 SF |
1 | /* |
2 | * Copyright (C) 2005-2017 Junjiro R. Okajima | |
3 | * | |
4 | * This program, aufs 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 of the License, or | |
7 | * (at your option) 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. If not, see <http://www.gnu.org/licenses/>. | |
16 | */ | |
17 | ||
18 | /* | |
19 | * pseudo-link | |
20 | */ | |
21 | ||
22 | #include "aufs.h" | |
23 | ||
24 | /* | |
25 | * the pseudo-link maintenance mode. | |
26 | * during a user process maintains the pseudo-links, | |
27 | * prohibit adding a new plink and branch manipulation. | |
28 | * | |
29 | * Flags | |
30 | * NOPLM: | |
31 | * For entry functions which will handle plink, and i_mutex is already held | |
32 | * in VFS. | |
33 | * They cannot wait and should return an error at once. | |
34 | * Callers has to check the error. | |
35 | * NOPLMW: | |
36 | * For entry functions which will handle plink, but i_mutex is not held | |
37 | * in VFS. | |
38 | * They can wait the plink maintenance mode to finish. | |
39 | * | |
40 | * They behave like F_SETLK and F_SETLKW. | |
41 | * If the caller never handle plink, then both flags are unnecessary. | |
42 | */ | |
43 | ||
44 | int au_plink_maint(struct super_block *sb, int flags) | |
45 | { | |
46 | int err; | |
47 | pid_t pid, ppid; | |
48 | struct task_struct *parent, *prev; | |
49 | struct au_sbinfo *sbi; | |
50 | ||
51 | SiMustAnyLock(sb); | |
52 | ||
53 | err = 0; | |
54 | if (!au_opt_test(au_mntflags(sb), PLINK)) | |
55 | goto out; | |
56 | ||
57 | sbi = au_sbi(sb); | |
58 | pid = sbi->si_plink_maint_pid; | |
59 | if (!pid || pid == current->pid) | |
60 | goto out; | |
61 | ||
62 | /* todo: it highly depends upon /sbin/mount.aufs */ | |
63 | prev = NULL; | |
64 | parent = current; | |
65 | ppid = 0; | |
66 | rcu_read_lock(); | |
67 | while (1) { | |
68 | parent = rcu_dereference(parent->real_parent); | |
69 | if (parent == prev) | |
70 | break; | |
71 | ppid = task_pid_vnr(parent); | |
72 | if (pid == ppid) { | |
73 | rcu_read_unlock(); | |
74 | goto out; | |
75 | } | |
76 | prev = parent; | |
77 | } | |
78 | rcu_read_unlock(); | |
79 | ||
80 | if (au_ftest_lock(flags, NOPLMW)) { | |
81 | /* if there is no i_mutex lock in VFS, we don't need to wait */ | |
82 | /* AuDebugOn(!lockdep_depth(current)); */ | |
83 | while (sbi->si_plink_maint_pid) { | |
84 | si_read_unlock(sb); | |
85 | /* gave up wake_up_bit() */ | |
86 | wait_event(sbi->si_plink_wq, !sbi->si_plink_maint_pid); | |
87 | ||
88 | if (au_ftest_lock(flags, FLUSH)) | |
89 | au_nwt_flush(&sbi->si_nowait); | |
90 | si_noflush_read_lock(sb); | |
91 | } | |
92 | } else if (au_ftest_lock(flags, NOPLM)) { | |
93 | AuDbg("ppid %d, pid %d\n", ppid, pid); | |
94 | err = -EAGAIN; | |
95 | } | |
96 | ||
97 | out: | |
98 | return err; | |
99 | } | |
100 | ||
101 | void au_plink_maint_leave(struct au_sbinfo *sbinfo) | |
102 | { | |
103 | spin_lock(&sbinfo->si_plink_maint_lock); | |
104 | sbinfo->si_plink_maint_pid = 0; | |
105 | spin_unlock(&sbinfo->si_plink_maint_lock); | |
106 | wake_up_all(&sbinfo->si_plink_wq); | |
107 | } | |
108 | ||
109 | int au_plink_maint_enter(struct super_block *sb) | |
110 | { | |
111 | int err; | |
112 | struct au_sbinfo *sbinfo; | |
113 | ||
114 | err = 0; | |
115 | sbinfo = au_sbi(sb); | |
116 | /* make sure i am the only one in this fs */ | |
117 | si_write_lock(sb, AuLock_FLUSH); | |
118 | if (au_opt_test(au_mntflags(sb), PLINK)) { | |
119 | spin_lock(&sbinfo->si_plink_maint_lock); | |
120 | if (!sbinfo->si_plink_maint_pid) | |
121 | sbinfo->si_plink_maint_pid = current->pid; | |
122 | else | |
123 | err = -EBUSY; | |
124 | spin_unlock(&sbinfo->si_plink_maint_lock); | |
125 | } | |
126 | si_write_unlock(sb); | |
127 | ||
128 | return err; | |
129 | } | |
130 | ||
131 | /* ---------------------------------------------------------------------- */ | |
132 | ||
133 | #ifdef CONFIG_AUFS_DEBUG | |
134 | void au_plink_list(struct super_block *sb) | |
135 | { | |
136 | int i; | |
137 | struct au_sbinfo *sbinfo; | |
138 | struct hlist_head *plink_hlist; | |
139 | struct au_icntnr *icntnr; | |
140 | ||
141 | SiMustAnyLock(sb); | |
142 | ||
143 | sbinfo = au_sbi(sb); | |
144 | AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK)); | |
145 | AuDebugOn(au_plink_maint(sb, AuLock_NOPLM)); | |
146 | ||
147 | for (i = 0; i < AuPlink_NHASH; i++) { | |
148 | plink_hlist = &sbinfo->si_plink[i].head; | |
149 | rcu_read_lock(); | |
150 | hlist_for_each_entry_rcu(icntnr, plink_hlist, plink) | |
151 | AuDbg("%lu\n", icntnr->vfs_inode.i_ino); | |
152 | rcu_read_unlock(); | |
153 | } | |
154 | } | |
155 | #endif | |
156 | ||
157 | /* is the inode pseudo-linked? */ | |
158 | int au_plink_test(struct inode *inode) | |
159 | { | |
160 | int found, i; | |
161 | struct au_sbinfo *sbinfo; | |
162 | struct hlist_head *plink_hlist; | |
163 | struct au_icntnr *icntnr; | |
164 | ||
165 | sbinfo = au_sbi(inode->i_sb); | |
166 | AuRwMustAnyLock(&sbinfo->si_rwsem); | |
167 | AuDebugOn(!au_opt_test(au_mntflags(inode->i_sb), PLINK)); | |
168 | AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM)); | |
169 | ||
170 | found = 0; | |
171 | i = au_plink_hash(inode->i_ino); | |
172 | plink_hlist = &sbinfo->si_plink[i].head; | |
173 | rcu_read_lock(); | |
174 | hlist_for_each_entry_rcu(icntnr, plink_hlist, plink) | |
175 | if (&icntnr->vfs_inode == inode) { | |
176 | found = 1; | |
177 | break; | |
178 | } | |
179 | rcu_read_unlock(); | |
180 | return found; | |
181 | } | |
182 | ||
183 | /* ---------------------------------------------------------------------- */ | |
184 | ||
185 | /* | |
186 | * generate a name for plink. | |
187 | * the file will be stored under AUFS_WH_PLINKDIR. | |
188 | */ | |
189 | /* 20 is max digits length of ulong 64 */ | |
190 | #define PLINK_NAME_LEN ((20 + 1) * 2) | |
191 | ||
192 | static int plink_name(char *name, int len, struct inode *inode, | |
193 | aufs_bindex_t bindex) | |
194 | { | |
195 | int rlen; | |
196 | struct inode *h_inode; | |
197 | ||
198 | h_inode = au_h_iptr(inode, bindex); | |
199 | rlen = snprintf(name, len, "%lu.%lu", inode->i_ino, h_inode->i_ino); | |
200 | return rlen; | |
201 | } | |
202 | ||
203 | struct au_do_plink_lkup_args { | |
204 | struct dentry **errp; | |
205 | struct qstr *tgtname; | |
206 | struct dentry *h_parent; | |
207 | struct au_branch *br; | |
208 | }; | |
209 | ||
210 | static struct dentry *au_do_plink_lkup(struct qstr *tgtname, | |
211 | struct dentry *h_parent, | |
212 | struct au_branch *br) | |
213 | { | |
214 | struct dentry *h_dentry; | |
215 | struct inode *h_inode; | |
216 | ||
217 | h_inode = d_inode(h_parent); | |
218 | vfsub_inode_lock_shared_nested(h_inode, AuLsc_I_CHILD2); | |
219 | h_dentry = vfsub_lkup_one(tgtname, h_parent); | |
220 | inode_unlock_shared(h_inode); | |
221 | return h_dentry; | |
222 | } | |
223 | ||
224 | static void au_call_do_plink_lkup(void *args) | |
225 | { | |
226 | struct au_do_plink_lkup_args *a = args; | |
227 | *a->errp = au_do_plink_lkup(a->tgtname, a->h_parent, a->br); | |
228 | } | |
229 | ||
230 | /* lookup the plink-ed @inode under the branch at @bindex */ | |
231 | struct dentry *au_plink_lkup(struct inode *inode, aufs_bindex_t bindex) | |
232 | { | |
233 | struct dentry *h_dentry, *h_parent; | |
234 | struct au_branch *br; | |
235 | int wkq_err; | |
236 | char a[PLINK_NAME_LEN]; | |
237 | struct qstr tgtname = QSTR_INIT(a, 0); | |
238 | ||
239 | AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM)); | |
240 | ||
241 | br = au_sbr(inode->i_sb, bindex); | |
242 | h_parent = br->br_wbr->wbr_plink; | |
243 | tgtname.len = plink_name(a, sizeof(a), inode, bindex); | |
244 | ||
245 | if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID)) { | |
246 | struct au_do_plink_lkup_args args = { | |
247 | .errp = &h_dentry, | |
248 | .tgtname = &tgtname, | |
249 | .h_parent = h_parent, | |
250 | .br = br | |
251 | }; | |
252 | ||
253 | wkq_err = au_wkq_wait(au_call_do_plink_lkup, &args); | |
254 | if (unlikely(wkq_err)) | |
255 | h_dentry = ERR_PTR(wkq_err); | |
256 | } else | |
257 | h_dentry = au_do_plink_lkup(&tgtname, h_parent, br); | |
258 | ||
259 | return h_dentry; | |
260 | } | |
261 | ||
262 | /* create a pseudo-link */ | |
263 | static int do_whplink(struct qstr *tgt, struct dentry *h_parent, | |
264 | struct dentry *h_dentry, struct au_branch *br) | |
265 | { | |
266 | int err; | |
267 | struct path h_path = { | |
268 | .mnt = au_br_mnt(br) | |
269 | }; | |
270 | struct inode *h_dir, *delegated; | |
271 | ||
272 | h_dir = d_inode(h_parent); | |
273 | inode_lock_nested(h_dir, AuLsc_I_CHILD2); | |
274 | again: | |
275 | h_path.dentry = vfsub_lkup_one(tgt, h_parent); | |
276 | err = PTR_ERR(h_path.dentry); | |
277 | if (IS_ERR(h_path.dentry)) | |
278 | goto out; | |
279 | ||
280 | err = 0; | |
281 | /* wh.plink dir is not monitored */ | |
282 | /* todo: is it really safe? */ | |
283 | if (d_is_positive(h_path.dentry) | |
284 | && d_inode(h_path.dentry) != d_inode(h_dentry)) { | |
285 | delegated = NULL; | |
286 | err = vfsub_unlink(h_dir, &h_path, &delegated, /*force*/0); | |
287 | if (unlikely(err == -EWOULDBLOCK)) { | |
288 | pr_warn("cannot retry for NFSv4 delegation" | |
289 | " for an internal unlink\n"); | |
290 | iput(delegated); | |
291 | } | |
292 | dput(h_path.dentry); | |
293 | h_path.dentry = NULL; | |
294 | if (!err) | |
295 | goto again; | |
296 | } | |
297 | if (!err && d_is_negative(h_path.dentry)) { | |
298 | delegated = NULL; | |
299 | err = vfsub_link(h_dentry, h_dir, &h_path, &delegated); | |
300 | if (unlikely(err == -EWOULDBLOCK)) { | |
301 | pr_warn("cannot retry for NFSv4 delegation" | |
302 | " for an internal link\n"); | |
303 | iput(delegated); | |
304 | } | |
305 | } | |
306 | dput(h_path.dentry); | |
307 | ||
308 | out: | |
309 | inode_unlock(h_dir); | |
310 | return err; | |
311 | } | |
312 | ||
313 | struct do_whplink_args { | |
314 | int *errp; | |
315 | struct qstr *tgt; | |
316 | struct dentry *h_parent; | |
317 | struct dentry *h_dentry; | |
318 | struct au_branch *br; | |
319 | }; | |
320 | ||
321 | static void call_do_whplink(void *args) | |
322 | { | |
323 | struct do_whplink_args *a = args; | |
324 | *a->errp = do_whplink(a->tgt, a->h_parent, a->h_dentry, a->br); | |
325 | } | |
326 | ||
327 | static int whplink(struct dentry *h_dentry, struct inode *inode, | |
328 | aufs_bindex_t bindex, struct au_branch *br) | |
329 | { | |
330 | int err, wkq_err; | |
331 | struct au_wbr *wbr; | |
332 | struct dentry *h_parent; | |
333 | char a[PLINK_NAME_LEN]; | |
334 | struct qstr tgtname = QSTR_INIT(a, 0); | |
335 | ||
336 | wbr = au_sbr(inode->i_sb, bindex)->br_wbr; | |
337 | h_parent = wbr->wbr_plink; | |
338 | tgtname.len = plink_name(a, sizeof(a), inode, bindex); | |
339 | ||
340 | /* always superio. */ | |
341 | if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID)) { | |
342 | struct do_whplink_args args = { | |
343 | .errp = &err, | |
344 | .tgt = &tgtname, | |
345 | .h_parent = h_parent, | |
346 | .h_dentry = h_dentry, | |
347 | .br = br | |
348 | }; | |
349 | wkq_err = au_wkq_wait(call_do_whplink, &args); | |
350 | if (unlikely(wkq_err)) | |
351 | err = wkq_err; | |
352 | } else | |
353 | err = do_whplink(&tgtname, h_parent, h_dentry, br); | |
354 | ||
355 | return err; | |
356 | } | |
357 | ||
358 | /* | |
359 | * create a new pseudo-link for @h_dentry on @bindex. | |
360 | * the linked inode is held in aufs @inode. | |
361 | */ | |
362 | void au_plink_append(struct inode *inode, aufs_bindex_t bindex, | |
363 | struct dentry *h_dentry) | |
364 | { | |
365 | struct super_block *sb; | |
366 | struct au_sbinfo *sbinfo; | |
367 | struct hlist_head *plink_hlist; | |
368 | struct au_icntnr *icntnr; | |
369 | struct au_sphlhead *sphl; | |
370 | int found, err, cnt, i; | |
371 | ||
372 | sb = inode->i_sb; | |
373 | sbinfo = au_sbi(sb); | |
374 | AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK)); | |
375 | AuDebugOn(au_plink_maint(sb, AuLock_NOPLM)); | |
376 | ||
377 | found = au_plink_test(inode); | |
378 | if (found) | |
379 | return; | |
380 | ||
381 | i = au_plink_hash(inode->i_ino); | |
382 | sphl = sbinfo->si_plink + i; | |
383 | plink_hlist = &sphl->head; | |
384 | au_igrab(inode); | |
385 | ||
386 | spin_lock(&sphl->spin); | |
387 | hlist_for_each_entry(icntnr, plink_hlist, plink) { | |
388 | if (&icntnr->vfs_inode == inode) { | |
389 | found = 1; | |
390 | break; | |
391 | } | |
392 | } | |
393 | if (!found) { | |
394 | icntnr = container_of(inode, struct au_icntnr, vfs_inode); | |
395 | hlist_add_head_rcu(&icntnr->plink, plink_hlist); | |
396 | } | |
397 | spin_unlock(&sphl->spin); | |
398 | if (!found) { | |
399 | cnt = au_sphl_count(sphl); | |
400 | #define msg "unexpectedly unblanced or too many pseudo-links" | |
401 | if (cnt > AUFS_PLINK_WARN) | |
402 | AuWarn1(msg ", %d\n", cnt); | |
403 | #undef msg | |
404 | err = whplink(h_dentry, inode, bindex, au_sbr(sb, bindex)); | |
405 | if (unlikely(err)) { | |
406 | pr_warn("err %d, damaged pseudo link.\n", err); | |
407 | au_sphl_del_rcu(&icntnr->plink, sphl); | |
408 | iput(&icntnr->vfs_inode); | |
409 | } | |
410 | } else | |
411 | iput(&icntnr->vfs_inode); | |
412 | } | |
413 | ||
414 | /* free all plinks */ | |
415 | void au_plink_put(struct super_block *sb, int verbose) | |
416 | { | |
417 | int i, warned; | |
418 | struct au_sbinfo *sbinfo; | |
419 | struct hlist_head *plink_hlist; | |
420 | struct hlist_node *tmp; | |
421 | struct au_icntnr *icntnr; | |
422 | ||
423 | SiMustWriteLock(sb); | |
424 | ||
425 | sbinfo = au_sbi(sb); | |
426 | AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK)); | |
427 | AuDebugOn(au_plink_maint(sb, AuLock_NOPLM)); | |
428 | ||
429 | /* no spin_lock since sbinfo is write-locked */ | |
430 | warned = 0; | |
431 | for (i = 0; i < AuPlink_NHASH; i++) { | |
432 | plink_hlist = &sbinfo->si_plink[i].head; | |
433 | if (!warned && verbose && !hlist_empty(plink_hlist)) { | |
434 | pr_warn("pseudo-link is not flushed"); | |
435 | warned = 1; | |
436 | } | |
437 | hlist_for_each_entry_safe(icntnr, tmp, plink_hlist, plink) | |
438 | iput(&icntnr->vfs_inode); | |
439 | INIT_HLIST_HEAD(plink_hlist); | |
440 | } | |
441 | } | |
442 | ||
443 | void au_plink_clean(struct super_block *sb, int verbose) | |
444 | { | |
445 | struct dentry *root; | |
446 | ||
447 | root = sb->s_root; | |
448 | aufs_write_lock(root); | |
449 | if (au_opt_test(au_mntflags(sb), PLINK)) | |
450 | au_plink_put(sb, verbose); | |
451 | aufs_write_unlock(root); | |
452 | } | |
453 | ||
454 | static int au_plink_do_half_refresh(struct inode *inode, aufs_bindex_t br_id) | |
455 | { | |
456 | int do_put; | |
457 | aufs_bindex_t btop, bbot, bindex; | |
458 | ||
459 | do_put = 0; | |
460 | btop = au_ibtop(inode); | |
461 | bbot = au_ibbot(inode); | |
462 | if (btop >= 0) { | |
463 | for (bindex = btop; bindex <= bbot; bindex++) { | |
464 | if (!au_h_iptr(inode, bindex) | |
465 | || au_ii_br_id(inode, bindex) != br_id) | |
466 | continue; | |
467 | au_set_h_iptr(inode, bindex, NULL, 0); | |
468 | do_put = 1; | |
469 | break; | |
470 | } | |
471 | if (do_put) | |
472 | for (bindex = btop; bindex <= bbot; bindex++) | |
473 | if (au_h_iptr(inode, bindex)) { | |
474 | do_put = 0; | |
475 | break; | |
476 | } | |
477 | } else | |
478 | do_put = 1; | |
479 | ||
480 | return do_put; | |
481 | } | |
482 | ||
483 | /* free the plinks on a branch specified by @br_id */ | |
484 | void au_plink_half_refresh(struct super_block *sb, aufs_bindex_t br_id) | |
485 | { | |
486 | struct au_sbinfo *sbinfo; | |
487 | struct hlist_head *plink_hlist; | |
488 | struct hlist_node *tmp; | |
489 | struct au_icntnr *icntnr; | |
490 | struct inode *inode; | |
491 | int i, do_put; | |
492 | ||
493 | SiMustWriteLock(sb); | |
494 | ||
495 | sbinfo = au_sbi(sb); | |
496 | AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK)); | |
497 | AuDebugOn(au_plink_maint(sb, AuLock_NOPLM)); | |
498 | ||
499 | /* no spin_lock since sbinfo is write-locked */ | |
500 | for (i = 0; i < AuPlink_NHASH; i++) { | |
501 | plink_hlist = &sbinfo->si_plink[i].head; | |
502 | hlist_for_each_entry_safe(icntnr, tmp, plink_hlist, plink) { | |
503 | inode = au_igrab(&icntnr->vfs_inode); | |
504 | ii_write_lock_child(inode); | |
505 | do_put = au_plink_do_half_refresh(inode, br_id); | |
506 | if (do_put) { | |
507 | hlist_del(&icntnr->plink); | |
508 | iput(inode); | |
509 | } | |
510 | ii_write_unlock(inode); | |
511 | iput(inode); | |
512 | } | |
513 | } | |
514 | } |