1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* AFS cell and server record management
3 *
4 * Copyright (C) 2002, 2017 Red Hat, Inc. All Rights Reserved.
5 * Written by David Howells (dhowells@redhat.com)
6 */
7
8 #include <linux/slab.h>
9 #include <linux/key.h>
10 #include <linux/ctype.h>
11 #include <linux/dns_resolver.h>
12 #include <linux/sched.h>
13 #include <linux/inet.h>
14 #include <linux/namei.h>
15 #include <keys/rxrpc-type.h>
16 #include "internal.h"
17
18 static unsigned __read_mostly afs_cell_gc_delay = 10;
19 static unsigned __read_mostly afs_cell_min_ttl = 10 * 60;
20 static unsigned __read_mostly afs_cell_max_ttl = 24 * 60 * 60;
21 static atomic_t cell_debug_id;
22
23 static void afs_queue_cell_manager(struct afs_net *);
24 static void afs_manage_cell_work(struct work_struct *);
25
afs_dec_cells_outstanding(struct afs_net * net)26 static void afs_dec_cells_outstanding(struct afs_net *net)
27 {
28 if (atomic_dec_and_test(&net->cells_outstanding))
29 wake_up_var(&net->cells_outstanding);
30 }
31
32 /*
33 * Set the cell timer to fire after a given delay, assuming it's not already
34 * set for an earlier time.
35 */
afs_set_cell_timer(struct afs_net * net,time64_t delay)36 static void afs_set_cell_timer(struct afs_net *net, time64_t delay)
37 {
38 if (net->live) {
39 atomic_inc(&net->cells_outstanding);
40 if (timer_reduce(&net->cells_timer, jiffies + delay * HZ))
41 afs_dec_cells_outstanding(net);
42 } else {
43 afs_queue_cell_manager(net);
44 }
45 }
46
47 /*
48 * Look up and get an activation reference on a cell record. The caller must
49 * hold net->cells_lock at least read-locked.
50 */
afs_find_cell_locked(struct afs_net * net,const char * name,unsigned int namesz,enum afs_cell_trace reason)51 static struct afs_cell *afs_find_cell_locked(struct afs_net *net,
52 const char *name, unsigned int namesz,
53 enum afs_cell_trace reason)
54 {
55 struct afs_cell *cell = NULL;
56 struct rb_node *p;
57 int n;
58
59 _enter("%*.*s", namesz, namesz, name);
60
61 if (name && namesz == 0)
62 return ERR_PTR(-EINVAL);
63 if (namesz > AFS_MAXCELLNAME)
64 return ERR_PTR(-ENAMETOOLONG);
65
66 if (!name) {
67 cell = net->ws_cell;
68 if (!cell)
69 return ERR_PTR(-EDESTADDRREQ);
70 goto found;
71 }
72
73 p = net->cells.rb_node;
74 while (p) {
75 cell = rb_entry(p, struct afs_cell, net_node);
76
77 n = strncasecmp(cell->name, name,
78 min_t(size_t, cell->name_len, namesz));
79 if (n == 0)
80 n = cell->name_len - namesz;
81 if (n < 0)
82 p = p->rb_left;
83 else if (n > 0)
84 p = p->rb_right;
85 else
86 goto found;
87 }
88
89 return ERR_PTR(-ENOENT);
90
91 found:
92 return afs_use_cell(cell, reason);
93 }
94
95 /*
96 * Look up and get an activation reference on a cell record.
97 */
afs_find_cell(struct afs_net * net,const char * name,unsigned int namesz,enum afs_cell_trace reason)98 struct afs_cell *afs_find_cell(struct afs_net *net,
99 const char *name, unsigned int namesz,
100 enum afs_cell_trace reason)
101 {
102 struct afs_cell *cell;
103
104 down_read(&net->cells_lock);
105 cell = afs_find_cell_locked(net, name, namesz, reason);
106 up_read(&net->cells_lock);
107 return cell;
108 }
109
110 /*
111 * Set up a cell record and fill in its name, VL server address list and
112 * allocate an anonymous key
113 */
afs_alloc_cell(struct afs_net * net,const char * name,unsigned int namelen,const char * addresses)114 static struct afs_cell *afs_alloc_cell(struct afs_net *net,
115 const char *name, unsigned int namelen,
116 const char *addresses)
117 {
118 struct afs_vlserver_list *vllist;
119 struct afs_cell *cell;
120 int i, ret;
121
122 ASSERT(name);
123 if (namelen == 0)
124 return ERR_PTR(-EINVAL);
125 if (namelen > AFS_MAXCELLNAME) {
126 _leave(" = -ENAMETOOLONG");
127 return ERR_PTR(-ENAMETOOLONG);
128 }
129
130 /* Prohibit cell names that contain unprintable chars, '/' and '@' or
131 * that begin with a dot. This also precludes "@cell".
132 */
133 if (name[0] == '.')
134 return ERR_PTR(-EINVAL);
135 for (i = 0; i < namelen; i++) {
136 char ch = name[i];
137 if (!isprint(ch) || ch == '/' || ch == '@')
138 return ERR_PTR(-EINVAL);
139 }
140
141 _enter("%*.*s,%s", namelen, namelen, name, addresses);
142
143 cell = kzalloc(sizeof(struct afs_cell), GFP_KERNEL);
144 if (!cell) {
145 _leave(" = -ENOMEM");
146 return ERR_PTR(-ENOMEM);
147 }
148
149 cell->name = kmalloc(namelen + 1, GFP_KERNEL);
150 if (!cell->name) {
151 kfree(cell);
152 return ERR_PTR(-ENOMEM);
153 }
154
155 cell->net = net;
156 cell->name_len = namelen;
157 for (i = 0; i < namelen; i++)
158 cell->name[i] = tolower(name[i]);
159 cell->name[i] = 0;
160
161 refcount_set(&cell->ref, 1);
162 atomic_set(&cell->active, 0);
163 INIT_WORK(&cell->manager, afs_manage_cell_work);
164 cell->volumes = RB_ROOT;
165 INIT_HLIST_HEAD(&cell->proc_volumes);
166 seqlock_init(&cell->volume_lock);
167 cell->fs_servers = RB_ROOT;
168 seqlock_init(&cell->fs_lock);
169 INIT_LIST_HEAD(&cell->fs_open_mmaps);
170 init_rwsem(&cell->fs_open_mmaps_lock);
171 rwlock_init(&cell->vl_servers_lock);
172 cell->flags = (1 << AFS_CELL_FL_CHECK_ALIAS);
173
174 /* Provide a VL server list, filling it in if we were given a list of
175 * addresses to use.
176 */
177 if (addresses) {
178 vllist = afs_parse_text_addrs(net,
179 addresses, strlen(addresses), ':',
180 VL_SERVICE, AFS_VL_PORT);
181 if (IS_ERR(vllist)) {
182 ret = PTR_ERR(vllist);
183 goto parse_failed;
184 }
185
186 vllist->source = DNS_RECORD_FROM_CONFIG;
187 vllist->status = DNS_LOOKUP_NOT_DONE;
188 cell->dns_expiry = TIME64_MAX;
189 } else {
190 ret = -ENOMEM;
191 vllist = afs_alloc_vlserver_list(0);
192 if (!vllist)
193 goto error;
194 vllist->source = DNS_RECORD_UNAVAILABLE;
195 vllist->status = DNS_LOOKUP_NOT_DONE;
196 cell->dns_expiry = ktime_get_real_seconds();
197 }
198
199 rcu_assign_pointer(cell->vl_servers, vllist);
200
201 cell->dns_source = vllist->source;
202 cell->dns_status = vllist->status;
203 smp_store_release(&cell->dns_lookup_count, 1); /* vs source/status */
204 atomic_inc(&net->cells_outstanding);
205 cell->debug_id = atomic_inc_return(&cell_debug_id);
206 trace_afs_cell(cell->debug_id, 1, 0, afs_cell_trace_alloc);
207
208 _leave(" = %p", cell);
209 return cell;
210
211 parse_failed:
212 if (ret == -EINVAL)
213 printk(KERN_ERR "kAFS: bad VL server IP address\n");
214 error:
215 kfree(cell->name);
216 kfree(cell);
217 _leave(" = %d", ret);
218 return ERR_PTR(ret);
219 }
220
221 /*
222 * afs_lookup_cell - Look up or create a cell record.
223 * @net: The network namespace
224 * @name: The name of the cell.
225 * @namesz: The strlen of the cell name.
226 * @vllist: A colon/comma separated list of numeric IP addresses or NULL.
227 * @excl: T if an error should be given if the cell name already exists.
228 *
229 * Look up a cell record by name and query the DNS for VL server addresses if
230 * needed. Note that that actual DNS query is punted off to the manager thread
231 * so that this function can return immediately if interrupted whilst allowing
232 * cell records to be shared even if not yet fully constructed.
233 */
afs_lookup_cell(struct afs_net * net,const char * name,unsigned int namesz,const char * vllist,bool excl)234 struct afs_cell *afs_lookup_cell(struct afs_net *net,
235 const char *name, unsigned int namesz,
236 const char *vllist, bool excl)
237 {
238 struct afs_cell *cell, *candidate, *cursor;
239 struct rb_node *parent, **pp;
240 enum afs_cell_state state;
241 int ret, n;
242
243 _enter("%s,%s", name, vllist);
244
245 if (!excl) {
246 cell = afs_find_cell(net, name, namesz, afs_cell_trace_use_lookup);
247 if (!IS_ERR(cell))
248 goto wait_for_cell;
249 }
250
251 /* Assume we're probably going to create a cell and preallocate and
252 * mostly set up a candidate record. We can then use this to stash the
253 * name, the net namespace and VL server addresses.
254 *
255 * We also want to do this before we hold any locks as it may involve
256 * upcalling to userspace to make DNS queries.
257 */
258 candidate = afs_alloc_cell(net, name, namesz, vllist);
259 if (IS_ERR(candidate)) {
260 _leave(" = %ld", PTR_ERR(candidate));
261 return candidate;
262 }
263
264 /* Find the insertion point and check to see if someone else added a
265 * cell whilst we were allocating.
266 */
267 down_write(&net->cells_lock);
268
269 pp = &net->cells.rb_node;
270 parent = NULL;
271 while (*pp) {
272 parent = *pp;
273 cursor = rb_entry(parent, struct afs_cell, net_node);
274
275 n = strncasecmp(cursor->name, name,
276 min_t(size_t, cursor->name_len, namesz));
277 if (n == 0)
278 n = cursor->name_len - namesz;
279 if (n < 0)
280 pp = &(*pp)->rb_left;
281 else if (n > 0)
282 pp = &(*pp)->rb_right;
283 else
284 goto cell_already_exists;
285 }
286
287 cell = candidate;
288 candidate = NULL;
289 atomic_set(&cell->active, 2);
290 trace_afs_cell(cell->debug_id, refcount_read(&cell->ref), 2, afs_cell_trace_insert);
291 rb_link_node_rcu(&cell->net_node, parent, pp);
292 rb_insert_color(&cell->net_node, &net->cells);
293 up_write(&net->cells_lock);
294
295 afs_queue_cell(cell, afs_cell_trace_get_queue_new);
296
297 wait_for_cell:
298 trace_afs_cell(cell->debug_id, refcount_read(&cell->ref), atomic_read(&cell->active),
299 afs_cell_trace_wait);
300 _debug("wait_for_cell");
301 wait_var_event(&cell->state,
302 ({
303 state = smp_load_acquire(&cell->state); /* vs error */
304 state == AFS_CELL_ACTIVE || state == AFS_CELL_REMOVED;
305 }));
306
307 /* Check the state obtained from the wait check. */
308 if (state == AFS_CELL_REMOVED) {
309 ret = cell->error;
310 goto error;
311 }
312
313 _leave(" = %p [cell]", cell);
314 return cell;
315
316 cell_already_exists:
317 _debug("cell exists");
318 cell = cursor;
319 if (excl) {
320 ret = -EEXIST;
321 } else {
322 afs_use_cell(cursor, afs_cell_trace_use_lookup);
323 ret = 0;
324 }
325 up_write(&net->cells_lock);
326 if (candidate)
327 afs_put_cell(candidate, afs_cell_trace_put_candidate);
328 if (ret == 0)
329 goto wait_for_cell;
330 goto error_noput;
331 error:
332 afs_unuse_cell(net, cell, afs_cell_trace_unuse_lookup);
333 error_noput:
334 _leave(" = %d [error]", ret);
335 return ERR_PTR(ret);
336 }
337
338 /*
339 * set the root cell information
340 * - can be called with a module parameter string
341 * - can be called from a write to /proc/fs/afs/rootcell
342 */
afs_cell_init(struct afs_net * net,const char * rootcell)343 int afs_cell_init(struct afs_net *net, const char *rootcell)
344 {
345 struct afs_cell *old_root, *new_root;
346 const char *cp, *vllist;
347 size_t len;
348
349 _enter("");
350
351 if (!rootcell) {
352 /* module is loaded with no parameters, or built statically.
353 * - in the future we might initialize cell DB here.
354 */
355 _leave(" = 0 [no root]");
356 return 0;
357 }
358
359 cp = strchr(rootcell, ':');
360 if (!cp) {
361 _debug("kAFS: no VL server IP addresses specified");
362 vllist = NULL;
363 len = strlen(rootcell);
364 } else {
365 vllist = cp + 1;
366 len = cp - rootcell;
367 }
368
369 /* allocate a cell record for the root cell */
370 new_root = afs_lookup_cell(net, rootcell, len, vllist, false);
371 if (IS_ERR(new_root)) {
372 _leave(" = %ld", PTR_ERR(new_root));
373 return PTR_ERR(new_root);
374 }
375
376 if (!test_and_set_bit(AFS_CELL_FL_NO_GC, &new_root->flags))
377 afs_use_cell(new_root, afs_cell_trace_use_pin);
378
379 /* install the new cell */
380 down_write(&net->cells_lock);
381 afs_see_cell(new_root, afs_cell_trace_see_ws);
382 old_root = net->ws_cell;
383 net->ws_cell = new_root;
384 up_write(&net->cells_lock);
385
386 afs_unuse_cell(net, old_root, afs_cell_trace_unuse_ws);
387 _leave(" = 0");
388 return 0;
389 }
390
391 /*
392 * Update a cell's VL server address list from the DNS.
393 */
afs_update_cell(struct afs_cell * cell)394 static int afs_update_cell(struct afs_cell *cell)
395 {
396 struct afs_vlserver_list *vllist, *old = NULL, *p;
397 unsigned int min_ttl = READ_ONCE(afs_cell_min_ttl);
398 unsigned int max_ttl = READ_ONCE(afs_cell_max_ttl);
399 time64_t now, expiry = 0;
400 int ret = 0;
401
402 _enter("%s", cell->name);
403
404 vllist = afs_dns_query(cell, &expiry);
405 if (IS_ERR(vllist)) {
406 ret = PTR_ERR(vllist);
407
408 _debug("%s: fail %d", cell->name, ret);
409 if (ret == -ENOMEM)
410 goto out_wake;
411
412 vllist = afs_alloc_vlserver_list(0);
413 if (!vllist) {
414 if (ret >= 0)
415 ret = -ENOMEM;
416 goto out_wake;
417 }
418
419 switch (ret) {
420 case -ENODATA:
421 case -EDESTADDRREQ:
422 vllist->status = DNS_LOOKUP_GOT_NOT_FOUND;
423 break;
424 case -EAGAIN:
425 case -ECONNREFUSED:
426 vllist->status = DNS_LOOKUP_GOT_TEMP_FAILURE;
427 break;
428 default:
429 vllist->status = DNS_LOOKUP_GOT_LOCAL_FAILURE;
430 break;
431 }
432 }
433
434 _debug("%s: got list %d %d", cell->name, vllist->source, vllist->status);
435 cell->dns_status = vllist->status;
436
437 now = ktime_get_real_seconds();
438 if (min_ttl > max_ttl)
439 max_ttl = min_ttl;
440 if (expiry < now + min_ttl)
441 expiry = now + min_ttl;
442 else if (expiry > now + max_ttl)
443 expiry = now + max_ttl;
444
445 _debug("%s: status %d", cell->name, vllist->status);
446 if (vllist->source == DNS_RECORD_UNAVAILABLE) {
447 switch (vllist->status) {
448 case DNS_LOOKUP_GOT_NOT_FOUND:
449 /* The DNS said that the cell does not exist or there
450 * weren't any addresses to be had.
451 */
452 cell->dns_expiry = expiry;
453 break;
454
455 case DNS_LOOKUP_BAD:
456 case DNS_LOOKUP_GOT_LOCAL_FAILURE:
457 case DNS_LOOKUP_GOT_TEMP_FAILURE:
458 case DNS_LOOKUP_GOT_NS_FAILURE:
459 default:
460 cell->dns_expiry = now + 10;
461 break;
462 }
463 } else {
464 cell->dns_expiry = expiry;
465 }
466
467 /* Replace the VL server list if the new record has servers or the old
468 * record doesn't.
469 */
470 write_lock(&cell->vl_servers_lock);
471 p = rcu_dereference_protected(cell->vl_servers, true);
472 if (vllist->nr_servers > 0 || p->nr_servers == 0) {
473 rcu_assign_pointer(cell->vl_servers, vllist);
474 cell->dns_source = vllist->source;
475 old = p;
476 }
477 write_unlock(&cell->vl_servers_lock);
478 afs_put_vlserverlist(cell->net, old);
479
480 out_wake:
481 smp_store_release(&cell->dns_lookup_count,
482 cell->dns_lookup_count + 1); /* vs source/status */
483 wake_up_var(&cell->dns_lookup_count);
484 _leave(" = %d", ret);
485 return ret;
486 }
487
488 /*
489 * Destroy a cell record
490 */
afs_cell_destroy(struct rcu_head * rcu)491 static void afs_cell_destroy(struct rcu_head *rcu)
492 {
493 struct afs_cell *cell = container_of(rcu, struct afs_cell, rcu);
494 struct afs_net *net = cell->net;
495 int r;
496
497 _enter("%p{%s}", cell, cell->name);
498
499 r = refcount_read(&cell->ref);
500 ASSERTCMP(r, ==, 0);
501 trace_afs_cell(cell->debug_id, r, atomic_read(&cell->active), afs_cell_trace_free);
502
503 afs_put_vlserverlist(net, rcu_access_pointer(cell->vl_servers));
504 afs_unuse_cell(net, cell->alias_of, afs_cell_trace_unuse_alias);
505 key_put(cell->anonymous_key);
506 kfree(cell->name);
507 kfree(cell);
508
509 afs_dec_cells_outstanding(net);
510 _leave(" [destroyed]");
511 }
512
513 /*
514 * Queue the cell manager.
515 */
afs_queue_cell_manager(struct afs_net * net)516 static void afs_queue_cell_manager(struct afs_net *net)
517 {
518 int outstanding = atomic_inc_return(&net->cells_outstanding);
519
520 _enter("%d", outstanding);
521
522 if (!queue_work(afs_wq, &net->cells_manager))
523 afs_dec_cells_outstanding(net);
524 }
525
526 /*
527 * Cell management timer. We have an increment on cells_outstanding that we
528 * need to pass along to the work item.
529 */
afs_cells_timer(struct timer_list * timer)530 void afs_cells_timer(struct timer_list *timer)
531 {
532 struct afs_net *net = container_of(timer, struct afs_net, cells_timer);
533
534 _enter("");
535 if (!queue_work(afs_wq, &net->cells_manager))
536 afs_dec_cells_outstanding(net);
537 }
538
539 /*
540 * Get a reference on a cell record.
541 */
afs_get_cell(struct afs_cell * cell,enum afs_cell_trace reason)542 struct afs_cell *afs_get_cell(struct afs_cell *cell, enum afs_cell_trace reason)
543 {
544 int r;
545
546 __refcount_inc(&cell->ref, &r);
547 trace_afs_cell(cell->debug_id, r + 1, atomic_read(&cell->active), reason);
548 return cell;
549 }
550
551 /*
552 * Drop a reference on a cell record.
553 */
afs_put_cell(struct afs_cell * cell,enum afs_cell_trace reason)554 void afs_put_cell(struct afs_cell *cell, enum afs_cell_trace reason)
555 {
556 if (cell) {
557 unsigned int debug_id = cell->debug_id;
558 unsigned int a;
559 bool zero;
560 int r;
561
562 a = atomic_read(&cell->active);
563 zero = __refcount_dec_and_test(&cell->ref, &r);
564 trace_afs_cell(debug_id, r - 1, a, reason);
565 if (zero) {
566 a = atomic_read(&cell->active);
567 WARN(a != 0, "Cell active count %u > 0\n", a);
568 call_rcu(&cell->rcu, afs_cell_destroy);
569 }
570 }
571 }
572
573 /*
574 * Note a cell becoming more active.
575 */
afs_use_cell(struct afs_cell * cell,enum afs_cell_trace reason)576 struct afs_cell *afs_use_cell(struct afs_cell *cell, enum afs_cell_trace reason)
577 {
578 int r, a;
579
580 r = refcount_read(&cell->ref);
581 WARN_ON(r == 0);
582 a = atomic_inc_return(&cell->active);
583 trace_afs_cell(cell->debug_id, r, a, reason);
584 return cell;
585 }
586
587 /*
588 * Record a cell becoming less active. When the active counter reaches 1, it
589 * is scheduled for destruction, but may get reactivated.
590 */
afs_unuse_cell(struct afs_net * net,struct afs_cell * cell,enum afs_cell_trace reason)591 void afs_unuse_cell(struct afs_net *net, struct afs_cell *cell, enum afs_cell_trace reason)
592 {
593 unsigned int debug_id;
594 time64_t now, expire_delay;
595 int r, a;
596
597 if (!cell)
598 return;
599
600 _enter("%s", cell->name);
601
602 now = ktime_get_real_seconds();
603 cell->last_inactive = now;
604 expire_delay = 0;
605 if (cell->vl_servers->nr_servers)
606 expire_delay = afs_cell_gc_delay;
607
608 debug_id = cell->debug_id;
609 r = refcount_read(&cell->ref);
610 a = atomic_dec_return(&cell->active);
611 trace_afs_cell(debug_id, r, a, reason);
612 WARN_ON(a == 0);
613 if (a == 1)
614 /* 'cell' may now be garbage collected. */
615 afs_set_cell_timer(net, expire_delay);
616 }
617
618 /*
619 * Note that a cell has been seen.
620 */
afs_see_cell(struct afs_cell * cell,enum afs_cell_trace reason)621 void afs_see_cell(struct afs_cell *cell, enum afs_cell_trace reason)
622 {
623 int r, a;
624
625 r = refcount_read(&cell->ref);
626 a = atomic_read(&cell->active);
627 trace_afs_cell(cell->debug_id, r, a, reason);
628 }
629
630 /*
631 * Queue a cell for management, giving the workqueue a ref to hold.
632 */
afs_queue_cell(struct afs_cell * cell,enum afs_cell_trace reason)633 void afs_queue_cell(struct afs_cell *cell, enum afs_cell_trace reason)
634 {
635 afs_get_cell(cell, reason);
636 if (!queue_work(afs_wq, &cell->manager))
637 afs_put_cell(cell, afs_cell_trace_put_queue_fail);
638 }
639
640 /*
641 * Allocate a key to use as a placeholder for anonymous user security.
642 */
afs_alloc_anon_key(struct afs_cell * cell)643 static int afs_alloc_anon_key(struct afs_cell *cell)
644 {
645 struct key *key;
646 char keyname[4 + AFS_MAXCELLNAME + 1], *cp, *dp;
647
648 /* Create a key to represent an anonymous user. */
649 memcpy(keyname, "afs@", 4);
650 dp = keyname + 4;
651 cp = cell->name;
652 do {
653 *dp++ = tolower(*cp);
654 } while (*cp++);
655
656 key = rxrpc_get_null_key(keyname);
657 if (IS_ERR(key))
658 return PTR_ERR(key);
659
660 cell->anonymous_key = key;
661
662 _debug("anon key %p{%x}",
663 cell->anonymous_key, key_serial(cell->anonymous_key));
664 return 0;
665 }
666
667 /*
668 * Activate a cell.
669 */
afs_activate_cell(struct afs_net * net,struct afs_cell * cell)670 static int afs_activate_cell(struct afs_net *net, struct afs_cell *cell)
671 {
672 struct hlist_node **p;
673 struct afs_cell *pcell;
674 int ret;
675
676 if (!cell->anonymous_key) {
677 ret = afs_alloc_anon_key(cell);
678 if (ret < 0)
679 return ret;
680 }
681
682 #ifdef CONFIG_AFS_FSCACHE
683 cell->cache = fscache_acquire_cookie(afs_cache_netfs.primary_index,
684 &afs_cell_cache_index_def,
685 cell->name, strlen(cell->name),
686 NULL, 0,
687 cell, 0, true);
688 #endif
689 ret = afs_proc_cell_setup(cell);
690 if (ret < 0)
691 return ret;
692
693 mutex_lock(&net->proc_cells_lock);
694 for (p = &net->proc_cells.first; *p; p = &(*p)->next) {
695 pcell = hlist_entry(*p, struct afs_cell, proc_link);
696 if (strcmp(cell->name, pcell->name) < 0)
697 break;
698 }
699
700 cell->proc_link.pprev = p;
701 cell->proc_link.next = *p;
702 rcu_assign_pointer(*p, &cell->proc_link.next);
703 if (cell->proc_link.next)
704 cell->proc_link.next->pprev = &cell->proc_link.next;
705
706 afs_dynroot_mkdir(net, cell);
707 mutex_unlock(&net->proc_cells_lock);
708 return 0;
709 }
710
711 /*
712 * Deactivate a cell.
713 */
afs_deactivate_cell(struct afs_net * net,struct afs_cell * cell)714 static void afs_deactivate_cell(struct afs_net *net, struct afs_cell *cell)
715 {
716 _enter("%s", cell->name);
717
718 afs_proc_cell_remove(cell);
719
720 mutex_lock(&net->proc_cells_lock);
721 hlist_del_rcu(&cell->proc_link);
722 afs_dynroot_rmdir(net, cell);
723 mutex_unlock(&net->proc_cells_lock);
724
725 #ifdef CONFIG_AFS_FSCACHE
726 fscache_relinquish_cookie(cell->cache, NULL, false);
727 cell->cache = NULL;
728 #endif
729
730 _leave("");
731 }
732
733 /*
734 * Manage a cell record, initialising and destroying it, maintaining its DNS
735 * records.
736 */
afs_manage_cell(struct afs_cell * cell)737 static void afs_manage_cell(struct afs_cell *cell)
738 {
739 struct afs_net *net = cell->net;
740 int ret, active;
741
742 _enter("%s", cell->name);
743
744 again:
745 _debug("state %u", cell->state);
746 switch (cell->state) {
747 case AFS_CELL_INACTIVE:
748 case AFS_CELL_FAILED:
749 down_write(&net->cells_lock);
750 active = 1;
751 if (atomic_try_cmpxchg_relaxed(&cell->active, &active, 0)) {
752 rb_erase(&cell->net_node, &net->cells);
753 trace_afs_cell(cell->debug_id, refcount_read(&cell->ref), 0,
754 afs_cell_trace_unuse_delete);
755 smp_store_release(&cell->state, AFS_CELL_REMOVED);
756 }
757 up_write(&net->cells_lock);
758 if (cell->state == AFS_CELL_REMOVED) {
759 wake_up_var(&cell->state);
760 goto final_destruction;
761 }
762 if (cell->state == AFS_CELL_FAILED)
763 goto done;
764 smp_store_release(&cell->state, AFS_CELL_UNSET);
765 wake_up_var(&cell->state);
766 goto again;
767
768 case AFS_CELL_UNSET:
769 smp_store_release(&cell->state, AFS_CELL_ACTIVATING);
770 wake_up_var(&cell->state);
771 goto again;
772
773 case AFS_CELL_ACTIVATING:
774 ret = afs_activate_cell(net, cell);
775 if (ret < 0)
776 goto activation_failed;
777
778 smp_store_release(&cell->state, AFS_CELL_ACTIVE);
779 wake_up_var(&cell->state);
780 goto again;
781
782 case AFS_CELL_ACTIVE:
783 if (atomic_read(&cell->active) > 1) {
784 if (test_and_clear_bit(AFS_CELL_FL_DO_LOOKUP, &cell->flags)) {
785 ret = afs_update_cell(cell);
786 if (ret < 0)
787 cell->error = ret;
788 }
789 goto done;
790 }
791 smp_store_release(&cell->state, AFS_CELL_DEACTIVATING);
792 wake_up_var(&cell->state);
793 goto again;
794
795 case AFS_CELL_DEACTIVATING:
796 if (atomic_read(&cell->active) > 1)
797 goto reverse_deactivation;
798 afs_deactivate_cell(net, cell);
799 smp_store_release(&cell->state, AFS_CELL_INACTIVE);
800 wake_up_var(&cell->state);
801 goto again;
802
803 case AFS_CELL_REMOVED:
804 goto done;
805
806 default:
807 break;
808 }
809 _debug("bad state %u", cell->state);
810 BUG(); /* Unhandled state */
811
812 activation_failed:
813 cell->error = ret;
814 afs_deactivate_cell(net, cell);
815
816 smp_store_release(&cell->state, AFS_CELL_FAILED); /* vs error */
817 wake_up_var(&cell->state);
818 goto again;
819
820 reverse_deactivation:
821 smp_store_release(&cell->state, AFS_CELL_ACTIVE);
822 wake_up_var(&cell->state);
823 _leave(" [deact->act]");
824 return;
825
826 done:
827 _leave(" [done %u]", cell->state);
828 return;
829
830 final_destruction:
831 /* The root volume is pinning the cell */
832 afs_put_volume(cell->net, cell->root_volume, afs_volume_trace_put_cell_root);
833 cell->root_volume = NULL;
834 afs_put_cell(cell, afs_cell_trace_put_destroy);
835 }
836
afs_manage_cell_work(struct work_struct * work)837 static void afs_manage_cell_work(struct work_struct *work)
838 {
839 struct afs_cell *cell = container_of(work, struct afs_cell, manager);
840
841 afs_manage_cell(cell);
842 afs_put_cell(cell, afs_cell_trace_put_queue_work);
843 }
844
845 /*
846 * Manage the records of cells known to a network namespace. This includes
847 * updating the DNS records and garbage collecting unused cells that were
848 * automatically added.
849 *
850 * Note that constructed cell records may only be removed from net->cells by
851 * this work item, so it is safe for this work item to stash a cursor pointing
852 * into the tree and then return to caller (provided it skips cells that are
853 * still under construction).
854 *
855 * Note also that we were given an increment on net->cells_outstanding by
856 * whoever queued us that we need to deal with before returning.
857 */
afs_manage_cells(struct work_struct * work)858 void afs_manage_cells(struct work_struct *work)
859 {
860 struct afs_net *net = container_of(work, struct afs_net, cells_manager);
861 struct rb_node *cursor;
862 time64_t now = ktime_get_real_seconds(), next_manage = TIME64_MAX;
863 bool purging = !net->live;
864
865 _enter("");
866
867 /* Trawl the cell database looking for cells that have expired from
868 * lack of use and cells whose DNS results have expired and dispatch
869 * their managers.
870 */
871 down_read(&net->cells_lock);
872
873 for (cursor = rb_first(&net->cells); cursor; cursor = rb_next(cursor)) {
874 struct afs_cell *cell =
875 rb_entry(cursor, struct afs_cell, net_node);
876 unsigned active;
877 bool sched_cell = false;
878
879 active = atomic_read(&cell->active);
880 trace_afs_cell(cell->debug_id, refcount_read(&cell->ref),
881 active, afs_cell_trace_manage);
882
883 ASSERTCMP(active, >=, 1);
884
885 if (purging) {
886 if (test_and_clear_bit(AFS_CELL_FL_NO_GC, &cell->flags)) {
887 active = atomic_dec_return(&cell->active);
888 trace_afs_cell(cell->debug_id, refcount_read(&cell->ref),
889 active, afs_cell_trace_unuse_pin);
890 }
891 }
892
893 if (active == 1) {
894 struct afs_vlserver_list *vllist;
895 time64_t expire_at = cell->last_inactive;
896
897 read_lock(&cell->vl_servers_lock);
898 vllist = rcu_dereference_protected(
899 cell->vl_servers,
900 lockdep_is_held(&cell->vl_servers_lock));
901 if (vllist->nr_servers > 0)
902 expire_at += afs_cell_gc_delay;
903 read_unlock(&cell->vl_servers_lock);
904 if (purging || expire_at <= now)
905 sched_cell = true;
906 else if (expire_at < next_manage)
907 next_manage = expire_at;
908 }
909
910 if (!purging) {
911 if (test_bit(AFS_CELL_FL_DO_LOOKUP, &cell->flags))
912 sched_cell = true;
913 }
914
915 if (sched_cell)
916 afs_queue_cell(cell, afs_cell_trace_get_queue_manage);
917 }
918
919 up_read(&net->cells_lock);
920
921 /* Update the timer on the way out. We have to pass an increment on
922 * cells_outstanding in the namespace that we are in to the timer or
923 * the work scheduler.
924 */
925 if (!purging && next_manage < TIME64_MAX) {
926 now = ktime_get_real_seconds();
927
928 if (next_manage - now <= 0) {
929 if (queue_work(afs_wq, &net->cells_manager))
930 atomic_inc(&net->cells_outstanding);
931 } else {
932 afs_set_cell_timer(net, next_manage - now);
933 }
934 }
935
936 afs_dec_cells_outstanding(net);
937 _leave(" [%d]", atomic_read(&net->cells_outstanding));
938 }
939
940 /*
941 * Purge in-memory cell database.
942 */
afs_cell_purge(struct afs_net * net)943 void afs_cell_purge(struct afs_net *net)
944 {
945 struct afs_cell *ws;
946
947 _enter("");
948
949 down_write(&net->cells_lock);
950 ws = net->ws_cell;
951 net->ws_cell = NULL;
952 up_write(&net->cells_lock);
953 afs_unuse_cell(net, ws, afs_cell_trace_unuse_ws);
954
955 _debug("del timer");
956 if (del_timer_sync(&net->cells_timer))
957 atomic_dec(&net->cells_outstanding);
958
959 _debug("kick mgr");
960 afs_queue_cell_manager(net);
961
962 _debug("wait");
963 wait_var_event(&net->cells_outstanding,
964 !atomic_read(&net->cells_outstanding));
965 _leave("");
966 }
967