]>
Commit | Line | Data |
---|---|---|
c088e31d 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_bl_head *hbl; | |
139 | struct hlist_bl_node *pos; | |
140 | struct au_icntnr *icntnr; | |
141 | ||
142 | SiMustAnyLock(sb); | |
143 | ||
144 | sbinfo = au_sbi(sb); | |
145 | AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK)); | |
146 | AuDebugOn(au_plink_maint(sb, AuLock_NOPLM)); | |
147 | ||
148 | for (i = 0; i < AuPlink_NHASH; i++) { | |
149 | hbl = sbinfo->si_plink + i; | |
150 | hlist_bl_lock(hbl); | |
151 | hlist_bl_for_each_entry(icntnr, pos, hbl, plink) | |
152 | AuDbg("%lu\n", icntnr->vfs_inode.i_ino); | |
153 | hlist_bl_unlock(hbl); | |
154 | } | |
155 | } | |
156 | #endif | |
157 | ||
158 | /* is the inode pseudo-linked? */ | |
159 | int au_plink_test(struct inode *inode) | |
160 | { | |
161 | int found, i; | |
162 | struct au_sbinfo *sbinfo; | |
163 | struct hlist_bl_head *hbl; | |
164 | struct hlist_bl_node *pos; | |
165 | struct au_icntnr *icntnr; | |
166 | ||
167 | sbinfo = au_sbi(inode->i_sb); | |
168 | AuRwMustAnyLock(&sbinfo->si_rwsem); | |
169 | AuDebugOn(!au_opt_test(au_mntflags(inode->i_sb), PLINK)); | |
170 | AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM)); | |
171 | ||
172 | found = 0; | |
173 | i = au_plink_hash(inode->i_ino); | |
174 | hbl = sbinfo->si_plink + i; | |
175 | hlist_bl_lock(hbl); | |
176 | hlist_bl_for_each_entry(icntnr, pos, hbl, plink) | |
177 | if (&icntnr->vfs_inode == inode) { | |
178 | found = 1; | |
179 | break; | |
180 | } | |
181 | hlist_bl_unlock(hbl); | |
182 | return found; | |
183 | } | |
184 | ||
185 | /* ---------------------------------------------------------------------- */ | |
186 | ||
187 | /* | |
188 | * generate a name for plink. | |
189 | * the file will be stored under AUFS_WH_PLINKDIR. | |
190 | */ | |
191 | /* 20 is max digits length of ulong 64 */ | |
192 | #define PLINK_NAME_LEN ((20 + 1) * 2) | |
193 | ||
194 | static int plink_name(char *name, int len, struct inode *inode, | |
195 | aufs_bindex_t bindex) | |
196 | { | |
197 | int rlen; | |
198 | struct inode *h_inode; | |
199 | ||
200 | h_inode = au_h_iptr(inode, bindex); | |
201 | rlen = snprintf(name, len, "%lu.%lu", inode->i_ino, h_inode->i_ino); | |
202 | return rlen; | |
203 | } | |
204 | ||
205 | struct au_do_plink_lkup_args { | |
206 | struct dentry **errp; | |
207 | struct qstr *tgtname; | |
208 | struct dentry *h_parent; | |
209 | struct au_branch *br; | |
210 | }; | |
211 | ||
212 | static struct dentry *au_do_plink_lkup(struct qstr *tgtname, | |
213 | struct dentry *h_parent, | |
214 | struct au_branch *br) | |
215 | { | |
216 | struct dentry *h_dentry; | |
217 | struct inode *h_inode; | |
218 | ||
219 | h_inode = d_inode(h_parent); | |
220 | vfsub_inode_lock_shared_nested(h_inode, AuLsc_I_CHILD2); | |
221 | h_dentry = vfsub_lkup_one(tgtname, h_parent); | |
222 | inode_unlock_shared(h_inode); | |
223 | return h_dentry; | |
224 | } | |
225 | ||
226 | static void au_call_do_plink_lkup(void *args) | |
227 | { | |
228 | struct au_do_plink_lkup_args *a = args; | |
229 | *a->errp = au_do_plink_lkup(a->tgtname, a->h_parent, a->br); | |
230 | } | |
231 | ||
232 | /* lookup the plink-ed @inode under the branch at @bindex */ | |
233 | struct dentry *au_plink_lkup(struct inode *inode, aufs_bindex_t bindex) | |
234 | { | |
235 | struct dentry *h_dentry, *h_parent; | |
236 | struct au_branch *br; | |
237 | int wkq_err; | |
238 | char a[PLINK_NAME_LEN]; | |
239 | struct qstr tgtname = QSTR_INIT(a, 0); | |
240 | ||
241 | AuDebugOn(au_plink_maint(inode->i_sb, AuLock_NOPLM)); | |
242 | ||
243 | br = au_sbr(inode->i_sb, bindex); | |
244 | h_parent = br->br_wbr->wbr_plink; | |
245 | tgtname.len = plink_name(a, sizeof(a), inode, bindex); | |
246 | ||
247 | if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID)) { | |
248 | struct au_do_plink_lkup_args args = { | |
249 | .errp = &h_dentry, | |
250 | .tgtname = &tgtname, | |
251 | .h_parent = h_parent, | |
252 | .br = br | |
253 | }; | |
254 | ||
255 | wkq_err = au_wkq_wait(au_call_do_plink_lkup, &args); | |
256 | if (unlikely(wkq_err)) | |
257 | h_dentry = ERR_PTR(wkq_err); | |
258 | } else | |
259 | h_dentry = au_do_plink_lkup(&tgtname, h_parent, br); | |
260 | ||
261 | return h_dentry; | |
262 | } | |
263 | ||
264 | /* create a pseudo-link */ | |
265 | static int do_whplink(struct qstr *tgt, struct dentry *h_parent, | |
266 | struct dentry *h_dentry, struct au_branch *br) | |
267 | { | |
268 | int err; | |
269 | struct path h_path = { | |
270 | .mnt = au_br_mnt(br) | |
271 | }; | |
272 | struct inode *h_dir, *delegated; | |
273 | ||
274 | h_dir = d_inode(h_parent); | |
275 | inode_lock_nested(h_dir, AuLsc_I_CHILD2); | |
276 | again: | |
277 | h_path.dentry = vfsub_lkup_one(tgt, h_parent); | |
278 | err = PTR_ERR(h_path.dentry); | |
279 | if (IS_ERR(h_path.dentry)) | |
280 | goto out; | |
281 | ||
282 | err = 0; | |
283 | /* wh.plink dir is not monitored */ | |
284 | /* todo: is it really safe? */ | |
285 | if (d_is_positive(h_path.dentry) | |
286 | && d_inode(h_path.dentry) != d_inode(h_dentry)) { | |
287 | delegated = NULL; | |
288 | err = vfsub_unlink(h_dir, &h_path, &delegated, /*force*/0); | |
289 | if (unlikely(err == -EWOULDBLOCK)) { | |
290 | pr_warn("cannot retry for NFSv4 delegation" | |
291 | " for an internal unlink\n"); | |
292 | iput(delegated); | |
293 | } | |
294 | dput(h_path.dentry); | |
295 | h_path.dentry = NULL; | |
296 | if (!err) | |
297 | goto again; | |
298 | } | |
299 | if (!err && d_is_negative(h_path.dentry)) { | |
300 | delegated = NULL; | |
301 | err = vfsub_link(h_dentry, h_dir, &h_path, &delegated); | |
302 | if (unlikely(err == -EWOULDBLOCK)) { | |
303 | pr_warn("cannot retry for NFSv4 delegation" | |
304 | " for an internal link\n"); | |
305 | iput(delegated); | |
306 | } | |
307 | } | |
308 | dput(h_path.dentry); | |
309 | ||
310 | out: | |
311 | inode_unlock(h_dir); | |
312 | return err; | |
313 | } | |
314 | ||
315 | struct do_whplink_args { | |
316 | int *errp; | |
317 | struct qstr *tgt; | |
318 | struct dentry *h_parent; | |
319 | struct dentry *h_dentry; | |
320 | struct au_branch *br; | |
321 | }; | |
322 | ||
323 | static void call_do_whplink(void *args) | |
324 | { | |
325 | struct do_whplink_args *a = args; | |
326 | *a->errp = do_whplink(a->tgt, a->h_parent, a->h_dentry, a->br); | |
327 | } | |
328 | ||
329 | static int whplink(struct dentry *h_dentry, struct inode *inode, | |
330 | aufs_bindex_t bindex, struct au_branch *br) | |
331 | { | |
332 | int err, wkq_err; | |
333 | struct au_wbr *wbr; | |
334 | struct dentry *h_parent; | |
335 | char a[PLINK_NAME_LEN]; | |
336 | struct qstr tgtname = QSTR_INIT(a, 0); | |
337 | ||
338 | wbr = au_sbr(inode->i_sb, bindex)->br_wbr; | |
339 | h_parent = wbr->wbr_plink; | |
340 | tgtname.len = plink_name(a, sizeof(a), inode, bindex); | |
341 | ||
342 | /* always superio. */ | |
343 | if (!uid_eq(current_fsuid(), GLOBAL_ROOT_UID)) { | |
344 | struct do_whplink_args args = { | |
345 | .errp = &err, | |
346 | .tgt = &tgtname, | |
347 | .h_parent = h_parent, | |
348 | .h_dentry = h_dentry, | |
349 | .br = br | |
350 | }; | |
351 | wkq_err = au_wkq_wait(call_do_whplink, &args); | |
352 | if (unlikely(wkq_err)) | |
353 | err = wkq_err; | |
354 | } else | |
355 | err = do_whplink(&tgtname, h_parent, h_dentry, br); | |
356 | ||
357 | return err; | |
358 | } | |
359 | ||
360 | /* | |
361 | * create a new pseudo-link for @h_dentry on @bindex. | |
362 | * the linked inode is held in aufs @inode. | |
363 | */ | |
364 | void au_plink_append(struct inode *inode, aufs_bindex_t bindex, | |
365 | struct dentry *h_dentry) | |
366 | { | |
367 | struct super_block *sb; | |
368 | struct au_sbinfo *sbinfo; | |
369 | struct hlist_bl_head *hbl; | |
370 | struct hlist_bl_node *pos; | |
371 | struct au_icntnr *icntnr; | |
372 | int found, err, cnt, i; | |
373 | ||
374 | sb = inode->i_sb; | |
375 | sbinfo = au_sbi(sb); | |
376 | AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK)); | |
377 | AuDebugOn(au_plink_maint(sb, AuLock_NOPLM)); | |
378 | ||
379 | found = au_plink_test(inode); | |
380 | if (found) | |
381 | return; | |
382 | ||
383 | i = au_plink_hash(inode->i_ino); | |
384 | hbl = sbinfo->si_plink + i; | |
385 | au_igrab(inode); | |
386 | ||
387 | hlist_bl_lock(hbl); | |
388 | hlist_bl_for_each_entry(icntnr, pos, hbl, plink) { | |
389 | if (&icntnr->vfs_inode == inode) { | |
390 | found = 1; | |
391 | break; | |
392 | } | |
393 | } | |
394 | if (!found) { | |
395 | icntnr = container_of(inode, struct au_icntnr, vfs_inode); | |
396 | hlist_bl_add_head(&icntnr->plink, hbl); | |
397 | } | |
398 | hlist_bl_unlock(hbl); | |
399 | if (!found) { | |
400 | cnt = au_hbl_count(hbl); | |
401 | #define msg "unexpectedly unblanced or too many pseudo-links" | |
402 | if (cnt > AUFS_PLINK_WARN) | |
403 | AuWarn1(msg ", %d\n", cnt); | |
404 | #undef msg | |
405 | err = whplink(h_dentry, inode, bindex, au_sbr(sb, bindex)); | |
406 | if (unlikely(err)) { | |
407 | pr_warn("err %d, damaged pseudo link.\n", err); | |
408 | au_hbl_del(&icntnr->plink, hbl); | |
409 | iput(&icntnr->vfs_inode); | |
410 | } | |
411 | } else | |
412 | iput(&icntnr->vfs_inode); | |
413 | } | |
414 | ||
415 | /* free all plinks */ | |
416 | void au_plink_put(struct super_block *sb, int verbose) | |
417 | { | |
418 | int i, warned; | |
419 | struct au_sbinfo *sbinfo; | |
420 | struct hlist_bl_head *hbl; | |
421 | struct hlist_bl_node *pos, *tmp; | |
422 | struct au_icntnr *icntnr; | |
423 | ||
424 | SiMustWriteLock(sb); | |
425 | ||
426 | sbinfo = au_sbi(sb); | |
427 | AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK)); | |
428 | AuDebugOn(au_plink_maint(sb, AuLock_NOPLM)); | |
429 | ||
430 | /* no spin_lock since sbinfo is write-locked */ | |
431 | warned = 0; | |
432 | for (i = 0; i < AuPlink_NHASH; i++) { | |
433 | hbl = sbinfo->si_plink + i; | |
434 | if (!warned && verbose && !hlist_bl_empty(hbl)) { | |
435 | pr_warn("pseudo-link is not flushed"); | |
436 | warned = 1; | |
437 | } | |
438 | hlist_bl_for_each_entry_safe(icntnr, pos, tmp, hbl, plink) | |
439 | iput(&icntnr->vfs_inode); | |
440 | INIT_HLIST_BL_HEAD(hbl); | |
441 | } | |
442 | } | |
443 | ||
444 | void au_plink_clean(struct super_block *sb, int verbose) | |
445 | { | |
446 | struct dentry *root; | |
447 | ||
448 | root = sb->s_root; | |
449 | aufs_write_lock(root); | |
450 | if (au_opt_test(au_mntflags(sb), PLINK)) | |
451 | au_plink_put(sb, verbose); | |
452 | aufs_write_unlock(root); | |
453 | } | |
454 | ||
455 | static int au_plink_do_half_refresh(struct inode *inode, aufs_bindex_t br_id) | |
456 | { | |
457 | int do_put; | |
458 | aufs_bindex_t btop, bbot, bindex; | |
459 | ||
460 | do_put = 0; | |
461 | btop = au_ibtop(inode); | |
462 | bbot = au_ibbot(inode); | |
463 | if (btop >= 0) { | |
464 | for (bindex = btop; bindex <= bbot; bindex++) { | |
465 | if (!au_h_iptr(inode, bindex) | |
466 | || au_ii_br_id(inode, bindex) != br_id) | |
467 | continue; | |
468 | au_set_h_iptr(inode, bindex, NULL, 0); | |
469 | do_put = 1; | |
470 | break; | |
471 | } | |
472 | if (do_put) | |
473 | for (bindex = btop; bindex <= bbot; bindex++) | |
474 | if (au_h_iptr(inode, bindex)) { | |
475 | do_put = 0; | |
476 | break; | |
477 | } | |
478 | } else | |
479 | do_put = 1; | |
480 | ||
481 | return do_put; | |
482 | } | |
483 | ||
484 | /* free the plinks on a branch specified by @br_id */ | |
485 | void au_plink_half_refresh(struct super_block *sb, aufs_bindex_t br_id) | |
486 | { | |
487 | struct au_sbinfo *sbinfo; | |
488 | struct hlist_bl_head *hbl; | |
489 | struct hlist_bl_node *pos, *tmp; | |
490 | struct au_icntnr *icntnr; | |
491 | struct inode *inode; | |
492 | int i, do_put; | |
493 | ||
494 | SiMustWriteLock(sb); | |
495 | ||
496 | sbinfo = au_sbi(sb); | |
497 | AuDebugOn(!au_opt_test(au_mntflags(sb), PLINK)); | |
498 | AuDebugOn(au_plink_maint(sb, AuLock_NOPLM)); | |
499 | ||
500 | /* no bit_lock since sbinfo is write-locked */ | |
501 | for (i = 0; i < AuPlink_NHASH; i++) { | |
502 | hbl = sbinfo->si_plink + i; | |
503 | hlist_bl_for_each_entry_safe(icntnr, pos, tmp, hbl, plink) { | |
504 | inode = au_igrab(&icntnr->vfs_inode); | |
505 | ii_write_lock_child(inode); | |
506 | do_put = au_plink_do_half_refresh(inode, br_id); | |
507 | if (do_put) { | |
508 | hlist_bl_del(&icntnr->plink); | |
509 | iput(inode); | |
510 | } | |
511 | ii_write_unlock(inode); | |
512 | iput(inode); | |
513 | } | |
514 | } | |
515 | } |