1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* AFS cell and server record management
4 * Copyright (C) 2002, 2017 Red Hat, Inc. All Rights Reserved.
8 #include <linux/slab.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>
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;
22 static void afs_manage_cell(struct work_struct *);
24 static void afs_dec_cells_outstanding(struct afs_net *net)
26 if (atomic_dec_and_test(&net->cells_outstanding))
27 wake_up_var(&net->cells_outstanding);
31 * Set the cell timer to fire after a given delay, assuming it's not already
32 * set for an earlier time.
34 static void afs_set_cell_timer(struct afs_net *net, time64_t delay)
37 atomic_inc(&net->cells_outstanding);
38 if (timer_reduce(&net->cells_timer, jiffies + delay * HZ))
39 afs_dec_cells_outstanding(net);
44 * Look up and get an activation reference on a cell record under RCU
45 * conditions. The caller must hold the RCU read lock.
47 struct afs_cell *afs_lookup_cell_rcu(struct afs_net *net,
48 const char *name, unsigned int namesz)
50 struct afs_cell *cell = NULL;
52 int n, seq = 0, ret = 0;
54 _enter("%*.*s", namesz, namesz, name);
56 if (name && namesz == 0)
57 return ERR_PTR(-EINVAL);
58 if (namesz > AFS_MAXCELLNAME)
59 return ERR_PTR(-ENAMETOOLONG);
62 /* Unfortunately, rbtree walking doesn't give reliable results
63 * under just the RCU read lock, so we have to check for
67 afs_put_cell(net, cell);
71 read_seqbegin_or_lock(&net->cells_lock, &seq);
74 cell = rcu_dereference_raw(net->ws_cell);
84 p = rcu_dereference_raw(net->cells.rb_node);
86 cell = rb_entry(p, struct afs_cell, net_node);
88 n = strncasecmp(cell->name, name,
89 min_t(size_t, cell->name_len, namesz));
91 n = cell->name_len - namesz;
93 p = rcu_dereference_raw(p->rb_left);
95 p = rcu_dereference_raw(p->rb_right);
97 if (atomic_inc_not_zero(&cell->usage)) {
101 /* We want to repeat the search, this time with
102 * the lock properly locked.
108 } while (need_seqretry(&net->cells_lock, seq));
110 done_seqretry(&net->cells_lock, seq);
112 if (ret != 0 && cell)
113 afs_put_cell(net, cell);
115 return ret == 0 ? cell : ERR_PTR(ret);
119 * Set up a cell record and fill in its name, VL server address list and
120 * allocate an anonymous key
122 static struct afs_cell *afs_alloc_cell(struct afs_net *net,
123 const char *name, unsigned int namelen,
124 const char *addresses)
126 struct afs_vlserver_list *vllist;
127 struct afs_cell *cell;
132 return ERR_PTR(-EINVAL);
133 if (namelen > AFS_MAXCELLNAME) {
134 _leave(" = -ENAMETOOLONG");
135 return ERR_PTR(-ENAMETOOLONG);
138 /* Prohibit cell names that contain unprintable chars, '/' and '@' or
139 * that begin with a dot. This also precludes "@cell".
142 return ERR_PTR(-EINVAL);
143 for (i = 0; i < namelen; i++) {
145 if (!isprint(ch) || ch == '/' || ch == '@')
146 return ERR_PTR(-EINVAL);
149 _enter("%*.*s,%s", namelen, namelen, name, addresses);
151 cell = kzalloc(sizeof(struct afs_cell), GFP_KERNEL);
153 _leave(" = -ENOMEM");
154 return ERR_PTR(-ENOMEM);
158 cell->name_len = namelen;
159 for (i = 0; i < namelen; i++)
160 cell->name[i] = tolower(name[i]);
162 atomic_set(&cell->usage, 2);
163 INIT_WORK(&cell->manager, afs_manage_cell);
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 rwlock_init(&cell->vl_servers_lock);
170 cell->flags = (1 << AFS_CELL_FL_CHECK_ALIAS);
172 /* Provide a VL server list, filling it in if we were given a list of
176 vllist = afs_parse_text_addrs(net,
177 addresses, strlen(addresses), ':',
178 VL_SERVICE, AFS_VL_PORT);
179 if (IS_ERR(vllist)) {
180 ret = PTR_ERR(vllist);
184 vllist->source = DNS_RECORD_FROM_CONFIG;
185 vllist->status = DNS_LOOKUP_NOT_DONE;
186 cell->dns_expiry = TIME64_MAX;
189 vllist = afs_alloc_vlserver_list(0);
192 vllist->source = DNS_RECORD_UNAVAILABLE;
193 vllist->status = DNS_LOOKUP_NOT_DONE;
194 cell->dns_expiry = ktime_get_real_seconds();
197 rcu_assign_pointer(cell->vl_servers, vllist);
199 cell->dns_source = vllist->source;
200 cell->dns_status = vllist->status;
201 smp_store_release(&cell->dns_lookup_count, 1); /* vs source/status */
203 _leave(" = %p", cell);
208 printk(KERN_ERR "kAFS: bad VL server IP address\n");
211 _leave(" = %d", ret);
216 * afs_lookup_cell - Look up or create a cell record.
217 * @net: The network namespace
218 * @name: The name of the cell.
219 * @namesz: The strlen of the cell name.
220 * @vllist: A colon/comma separated list of numeric IP addresses or NULL.
221 * @excl: T if an error should be given if the cell name already exists.
223 * Look up a cell record by name and query the DNS for VL server addresses if
224 * needed. Note that that actual DNS query is punted off to the manager thread
225 * so that this function can return immediately if interrupted whilst allowing
226 * cell records to be shared even if not yet fully constructed.
228 struct afs_cell *afs_lookup_cell(struct afs_net *net,
229 const char *name, unsigned int namesz,
230 const char *vllist, bool excl)
232 struct afs_cell *cell, *candidate, *cursor;
233 struct rb_node *parent, **pp;
234 enum afs_cell_state state;
237 _enter("%s,%s", name, vllist);
241 cell = afs_lookup_cell_rcu(net, name, namesz);
247 /* Assume we're probably going to create a cell and preallocate and
248 * mostly set up a candidate record. We can then use this to stash the
249 * name, the net namespace and VL server addresses.
251 * We also want to do this before we hold any locks as it may involve
252 * upcalling to userspace to make DNS queries.
254 candidate = afs_alloc_cell(net, name, namesz, vllist);
255 if (IS_ERR(candidate)) {
256 _leave(" = %ld", PTR_ERR(candidate));
260 /* Find the insertion point and check to see if someone else added a
261 * cell whilst we were allocating.
263 write_seqlock(&net->cells_lock);
265 pp = &net->cells.rb_node;
269 cursor = rb_entry(parent, struct afs_cell, net_node);
271 n = strncasecmp(cursor->name, name,
272 min_t(size_t, cursor->name_len, namesz));
274 n = cursor->name_len - namesz;
276 pp = &(*pp)->rb_left;
278 pp = &(*pp)->rb_right;
280 goto cell_already_exists;
285 rb_link_node_rcu(&cell->net_node, parent, pp);
286 rb_insert_color(&cell->net_node, &net->cells);
287 atomic_inc(&net->cells_outstanding);
288 write_sequnlock(&net->cells_lock);
290 queue_work(afs_wq, &cell->manager);
293 _debug("wait_for_cell");
294 wait_var_event(&cell->state,
296 state = smp_load_acquire(&cell->state); /* vs error */
297 state == AFS_CELL_ACTIVE || state == AFS_CELL_FAILED;
300 /* Check the state obtained from the wait check. */
301 if (state == AFS_CELL_FAILED) {
306 _leave(" = %p [cell]", cell);
310 _debug("cell exists");
315 afs_get_cell(cursor);
318 write_sequnlock(&net->cells_lock);
324 afs_put_cell(net, cell);
326 _leave(" = %d [error]", ret);
331 * set the root cell information
332 * - can be called with a module parameter string
333 * - can be called from a write to /proc/fs/afs/rootcell
335 int afs_cell_init(struct afs_net *net, const char *rootcell)
337 struct afs_cell *old_root, *new_root;
338 const char *cp, *vllist;
344 /* module is loaded with no parameters, or built statically.
345 * - in the future we might initialize cell DB here.
347 _leave(" = 0 [no root]");
351 cp = strchr(rootcell, ':');
353 _debug("kAFS: no VL server IP addresses specified");
355 len = strlen(rootcell);
361 /* allocate a cell record for the root cell */
362 new_root = afs_lookup_cell(net, rootcell, len, vllist, false);
363 if (IS_ERR(new_root)) {
364 _leave(" = %ld", PTR_ERR(new_root));
365 return PTR_ERR(new_root);
368 if (!test_and_set_bit(AFS_CELL_FL_NO_GC, &new_root->flags))
369 afs_get_cell(new_root);
371 /* install the new cell */
372 write_seqlock(&net->cells_lock);
373 old_root = rcu_access_pointer(net->ws_cell);
374 rcu_assign_pointer(net->ws_cell, new_root);
375 write_sequnlock(&net->cells_lock);
377 afs_put_cell(net, old_root);
383 * Update a cell's VL server address list from the DNS.
385 static int afs_update_cell(struct afs_cell *cell)
387 struct afs_vlserver_list *vllist, *old = NULL, *p;
388 unsigned int min_ttl = READ_ONCE(afs_cell_min_ttl);
389 unsigned int max_ttl = READ_ONCE(afs_cell_max_ttl);
390 time64_t now, expiry = 0;
393 _enter("%s", cell->name);
395 vllist = afs_dns_query(cell, &expiry);
396 if (IS_ERR(vllist)) {
397 ret = PTR_ERR(vllist);
399 _debug("%s: fail %d", cell->name, ret);
404 vllist = afs_alloc_vlserver_list(0);
411 vllist->status = DNS_LOOKUP_GOT_NOT_FOUND;
415 vllist->status = DNS_LOOKUP_GOT_TEMP_FAILURE;
418 vllist->status = DNS_LOOKUP_GOT_LOCAL_FAILURE;
423 _debug("%s: got list %d %d", cell->name, vllist->source, vllist->status);
424 cell->dns_status = vllist->status;
426 now = ktime_get_real_seconds();
427 if (min_ttl > max_ttl)
429 if (expiry < now + min_ttl)
430 expiry = now + min_ttl;
431 else if (expiry > now + max_ttl)
432 expiry = now + max_ttl;
434 _debug("%s: status %d", cell->name, vllist->status);
435 if (vllist->source == DNS_RECORD_UNAVAILABLE) {
436 switch (vllist->status) {
437 case DNS_LOOKUP_GOT_NOT_FOUND:
438 /* The DNS said that the cell does not exist or there
439 * weren't any addresses to be had.
441 cell->dns_expiry = expiry;
445 case DNS_LOOKUP_GOT_LOCAL_FAILURE:
446 case DNS_LOOKUP_GOT_TEMP_FAILURE:
447 case DNS_LOOKUP_GOT_NS_FAILURE:
449 cell->dns_expiry = now + 10;
453 cell->dns_expiry = expiry;
456 /* Replace the VL server list if the new record has servers or the old
459 write_lock(&cell->vl_servers_lock);
460 p = rcu_dereference_protected(cell->vl_servers, true);
461 if (vllist->nr_servers > 0 || p->nr_servers == 0) {
462 rcu_assign_pointer(cell->vl_servers, vllist);
463 cell->dns_source = vllist->source;
466 write_unlock(&cell->vl_servers_lock);
467 afs_put_vlserverlist(cell->net, old);
470 smp_store_release(&cell->dns_lookup_count,
471 cell->dns_lookup_count + 1); /* vs source/status */
472 wake_up_var(&cell->dns_lookup_count);
473 _leave(" = %d", ret);
478 * Destroy a cell record
480 static void afs_cell_destroy(struct rcu_head *rcu)
482 struct afs_cell *cell = container_of(rcu, struct afs_cell, rcu);
484 _enter("%p{%s}", cell, cell->name);
486 ASSERTCMP(atomic_read(&cell->usage), ==, 0);
488 afs_put_volume(cell->net, cell->root_volume, afs_volume_trace_put_cell_root);
489 afs_put_vlserverlist(cell->net, rcu_access_pointer(cell->vl_servers));
490 afs_put_cell(cell->net, cell->alias_of);
491 key_put(cell->anonymous_key);
494 _leave(" [destroyed]");
498 * Queue the cell manager.
500 static void afs_queue_cell_manager(struct afs_net *net)
502 int outstanding = atomic_inc_return(&net->cells_outstanding);
504 _enter("%d", outstanding);
506 if (!queue_work(afs_wq, &net->cells_manager))
507 afs_dec_cells_outstanding(net);
511 * Cell management timer. We have an increment on cells_outstanding that we
512 * need to pass along to the work item.
514 void afs_cells_timer(struct timer_list *timer)
516 struct afs_net *net = container_of(timer, struct afs_net, cells_timer);
519 if (!queue_work(afs_wq, &net->cells_manager))
520 afs_dec_cells_outstanding(net);
524 * Get a reference on a cell record.
526 struct afs_cell *afs_get_cell(struct afs_cell *cell)
528 atomic_inc(&cell->usage);
533 * Drop a reference on a cell record.
535 void afs_put_cell(struct afs_net *net, struct afs_cell *cell)
537 time64_t now, expire_delay;
542 _enter("%s", cell->name);
544 now = ktime_get_real_seconds();
545 cell->last_inactive = now;
547 if (cell->vl_servers->nr_servers)
548 expire_delay = afs_cell_gc_delay;
550 if (atomic_dec_return(&cell->usage) > 1)
553 /* 'cell' may now be garbage collected. */
554 afs_set_cell_timer(net, expire_delay);
558 * Allocate a key to use as a placeholder for anonymous user security.
560 static int afs_alloc_anon_key(struct afs_cell *cell)
563 char keyname[4 + AFS_MAXCELLNAME + 1], *cp, *dp;
565 /* Create a key to represent an anonymous user. */
566 memcpy(keyname, "afs@", 4);
570 *dp++ = tolower(*cp);
573 key = rxrpc_get_null_key(keyname);
577 cell->anonymous_key = key;
579 _debug("anon key %p{%x}",
580 cell->anonymous_key, key_serial(cell->anonymous_key));
587 static int afs_activate_cell(struct afs_net *net, struct afs_cell *cell)
589 struct hlist_node **p;
590 struct afs_cell *pcell;
593 if (!cell->anonymous_key) {
594 ret = afs_alloc_anon_key(cell);
599 #ifdef CONFIG_AFS_FSCACHE
600 cell->cache = fscache_acquire_cookie(afs_cache_netfs.primary_index,
601 &afs_cell_cache_index_def,
602 cell->name, strlen(cell->name),
606 ret = afs_proc_cell_setup(cell);
610 mutex_lock(&net->proc_cells_lock);
611 for (p = &net->proc_cells.first; *p; p = &(*p)->next) {
612 pcell = hlist_entry(*p, struct afs_cell, proc_link);
613 if (strcmp(cell->name, pcell->name) < 0)
617 cell->proc_link.pprev = p;
618 cell->proc_link.next = *p;
619 rcu_assign_pointer(*p, &cell->proc_link.next);
620 if (cell->proc_link.next)
621 cell->proc_link.next->pprev = &cell->proc_link.next;
623 afs_dynroot_mkdir(net, cell);
624 mutex_unlock(&net->proc_cells_lock);
631 static void afs_deactivate_cell(struct afs_net *net, struct afs_cell *cell)
633 _enter("%s", cell->name);
635 afs_proc_cell_remove(cell);
637 mutex_lock(&net->proc_cells_lock);
638 hlist_del_rcu(&cell->proc_link);
639 afs_dynroot_rmdir(net, cell);
640 mutex_unlock(&net->proc_cells_lock);
642 #ifdef CONFIG_AFS_FSCACHE
643 fscache_relinquish_cookie(cell->cache, NULL, false);
651 * Manage a cell record, initialising and destroying it, maintaining its DNS
654 static void afs_manage_cell(struct work_struct *work)
656 struct afs_cell *cell = container_of(work, struct afs_cell, manager);
657 struct afs_net *net = cell->net;
661 _enter("%s", cell->name);
664 _debug("state %u", cell->state);
665 switch (cell->state) {
666 case AFS_CELL_INACTIVE:
667 case AFS_CELL_FAILED:
668 write_seqlock(&net->cells_lock);
670 deleted = atomic_try_cmpxchg_relaxed(&cell->usage, &usage, 0);
672 rb_erase(&cell->net_node, &net->cells);
673 write_sequnlock(&net->cells_lock);
675 goto final_destruction;
676 if (cell->state == AFS_CELL_FAILED)
678 smp_store_release(&cell->state, AFS_CELL_UNSET);
679 wake_up_var(&cell->state);
683 smp_store_release(&cell->state, AFS_CELL_ACTIVATING);
684 wake_up_var(&cell->state);
687 case AFS_CELL_ACTIVATING:
688 ret = afs_activate_cell(net, cell);
690 goto activation_failed;
692 smp_store_release(&cell->state, AFS_CELL_ACTIVE);
693 wake_up_var(&cell->state);
696 case AFS_CELL_ACTIVE:
697 if (atomic_read(&cell->usage) > 1) {
698 if (test_and_clear_bit(AFS_CELL_FL_DO_LOOKUP, &cell->flags)) {
699 ret = afs_update_cell(cell);
705 smp_store_release(&cell->state, AFS_CELL_DEACTIVATING);
706 wake_up_var(&cell->state);
709 case AFS_CELL_DEACTIVATING:
710 if (atomic_read(&cell->usage) > 1)
711 goto reverse_deactivation;
712 afs_deactivate_cell(net, cell);
713 smp_store_release(&cell->state, AFS_CELL_INACTIVE);
714 wake_up_var(&cell->state);
720 _debug("bad state %u", cell->state);
721 BUG(); /* Unhandled state */
725 afs_deactivate_cell(net, cell);
727 smp_store_release(&cell->state, AFS_CELL_FAILED); /* vs error */
728 wake_up_var(&cell->state);
731 reverse_deactivation:
732 smp_store_release(&cell->state, AFS_CELL_ACTIVE);
733 wake_up_var(&cell->state);
734 _leave(" [deact->act]");
738 _leave(" [done %u]", cell->state);
742 call_rcu(&cell->rcu, afs_cell_destroy);
743 afs_dec_cells_outstanding(net);
744 _leave(" [destruct %d]", atomic_read(&net->cells_outstanding));
748 * Manage the records of cells known to a network namespace. This includes
749 * updating the DNS records and garbage collecting unused cells that were
750 * automatically added.
752 * Note that constructed cell records may only be removed from net->cells by
753 * this work item, so it is safe for this work item to stash a cursor pointing
754 * into the tree and then return to caller (provided it skips cells that are
755 * still under construction).
757 * Note also that we were given an increment on net->cells_outstanding by
758 * whoever queued us that we need to deal with before returning.
760 void afs_manage_cells(struct work_struct *work)
762 struct afs_net *net = container_of(work, struct afs_net, cells_manager);
763 struct rb_node *cursor;
764 time64_t now = ktime_get_real_seconds(), next_manage = TIME64_MAX;
765 bool purging = !net->live;
769 /* Trawl the cell database looking for cells that have expired from
770 * lack of use and cells whose DNS results have expired and dispatch
773 read_seqlock_excl(&net->cells_lock);
775 for (cursor = rb_first(&net->cells); cursor; cursor = rb_next(cursor)) {
776 struct afs_cell *cell =
777 rb_entry(cursor, struct afs_cell, net_node);
779 bool sched_cell = false;
781 usage = atomic_read(&cell->usage);
782 _debug("manage %s %u", cell->name, usage);
784 ASSERTCMP(usage, >=, 1);
787 if (test_and_clear_bit(AFS_CELL_FL_NO_GC, &cell->flags))
788 usage = atomic_dec_return(&cell->usage);
789 ASSERTCMP(usage, ==, 1);
793 struct afs_vlserver_list *vllist;
794 time64_t expire_at = cell->last_inactive;
796 read_lock(&cell->vl_servers_lock);
797 vllist = rcu_dereference_protected(
799 lockdep_is_held(&cell->vl_servers_lock));
800 if (vllist->nr_servers > 0)
801 expire_at += afs_cell_gc_delay;
802 read_unlock(&cell->vl_servers_lock);
803 if (purging || expire_at <= now)
805 else if (expire_at < next_manage)
806 next_manage = expire_at;
810 if (test_bit(AFS_CELL_FL_DO_LOOKUP, &cell->flags))
815 queue_work(afs_wq, &cell->manager);
818 read_sequnlock_excl(&net->cells_lock);
820 /* Update the timer on the way out. We have to pass an increment on
821 * cells_outstanding in the namespace that we are in to the timer or
822 * the work scheduler.
824 if (!purging && next_manage < TIME64_MAX) {
825 now = ktime_get_real_seconds();
827 if (next_manage - now <= 0) {
828 if (queue_work(afs_wq, &net->cells_manager))
829 atomic_inc(&net->cells_outstanding);
831 afs_set_cell_timer(net, next_manage - now);
835 afs_dec_cells_outstanding(net);
836 _leave(" [%d]", atomic_read(&net->cells_outstanding));
840 * Purge in-memory cell database.
842 void afs_cell_purge(struct afs_net *net)
848 write_seqlock(&net->cells_lock);
849 ws = rcu_access_pointer(net->ws_cell);
850 RCU_INIT_POINTER(net->ws_cell, NULL);
851 write_sequnlock(&net->cells_lock);
852 afs_put_cell(net, ws);
855 if (del_timer_sync(&net->cells_timer))
856 atomic_dec(&net->cells_outstanding);
859 afs_queue_cell_manager(net);
862 wait_var_event(&net->cells_outstanding,
863 !atomic_read(&net->cells_outstanding));