.. | .. |
---|
| 1 | +// SPDX-License-Identifier: GPL-2.0-or-later |
---|
1 | 2 | /* AFS volume management |
---|
2 | 3 | * |
---|
3 | 4 | * Copyright (C) 2002, 2007 Red Hat, Inc. All Rights Reserved. |
---|
4 | 5 | * Written by David Howells (dhowells@redhat.com) |
---|
5 | | - * |
---|
6 | | - * This program is free software; you can redistribute it and/or |
---|
7 | | - * modify it under the terms of the GNU General Public License |
---|
8 | | - * as published by the Free Software Foundation; either version |
---|
9 | | - * 2 of the License, or (at your option) any later version. |
---|
10 | 6 | */ |
---|
11 | 7 | |
---|
12 | 8 | #include <linux/kernel.h> |
---|
.. | .. |
---|
16 | 12 | unsigned __read_mostly afs_volume_gc_delay = 10; |
---|
17 | 13 | unsigned __read_mostly afs_volume_record_life = 60 * 60; |
---|
18 | 14 | |
---|
19 | | -static const char *const afs_voltypes[] = { "R/W", "R/O", "BAK" }; |
---|
| 15 | +/* |
---|
| 16 | + * Insert a volume into a cell. If there's an existing volume record, that is |
---|
| 17 | + * returned instead with a ref held. |
---|
| 18 | + */ |
---|
| 19 | +static struct afs_volume *afs_insert_volume_into_cell(struct afs_cell *cell, |
---|
| 20 | + struct afs_volume *volume) |
---|
| 21 | +{ |
---|
| 22 | + struct afs_volume *p; |
---|
| 23 | + struct rb_node *parent = NULL, **pp; |
---|
| 24 | + |
---|
| 25 | + write_seqlock(&cell->volume_lock); |
---|
| 26 | + |
---|
| 27 | + pp = &cell->volumes.rb_node; |
---|
| 28 | + while (*pp) { |
---|
| 29 | + parent = *pp; |
---|
| 30 | + p = rb_entry(parent, struct afs_volume, cell_node); |
---|
| 31 | + if (p->vid < volume->vid) { |
---|
| 32 | + pp = &(*pp)->rb_left; |
---|
| 33 | + } else if (p->vid > volume->vid) { |
---|
| 34 | + pp = &(*pp)->rb_right; |
---|
| 35 | + } else { |
---|
| 36 | + volume = afs_get_volume(p, afs_volume_trace_get_cell_insert); |
---|
| 37 | + goto found; |
---|
| 38 | + } |
---|
| 39 | + } |
---|
| 40 | + |
---|
| 41 | + rb_link_node_rcu(&volume->cell_node, parent, pp); |
---|
| 42 | + rb_insert_color(&volume->cell_node, &cell->volumes); |
---|
| 43 | + hlist_add_head_rcu(&volume->proc_link, &cell->proc_volumes); |
---|
| 44 | + |
---|
| 45 | +found: |
---|
| 46 | + write_sequnlock(&cell->volume_lock); |
---|
| 47 | + return volume; |
---|
| 48 | + |
---|
| 49 | +} |
---|
| 50 | + |
---|
| 51 | +static void afs_remove_volume_from_cell(struct afs_volume *volume) |
---|
| 52 | +{ |
---|
| 53 | + struct afs_cell *cell = volume->cell; |
---|
| 54 | + |
---|
| 55 | + if (!hlist_unhashed(&volume->proc_link)) { |
---|
| 56 | + trace_afs_volume(volume->vid, atomic_read(&volume->usage), |
---|
| 57 | + afs_volume_trace_remove); |
---|
| 58 | + write_seqlock(&cell->volume_lock); |
---|
| 59 | + hlist_del_rcu(&volume->proc_link); |
---|
| 60 | + rb_erase(&volume->cell_node, &cell->volumes); |
---|
| 61 | + write_sequnlock(&cell->volume_lock); |
---|
| 62 | + } |
---|
| 63 | +} |
---|
20 | 64 | |
---|
21 | 65 | /* |
---|
22 | 66 | * Allocate a volume record and load it up from a vldb record. |
---|
23 | 67 | */ |
---|
24 | | -static struct afs_volume *afs_alloc_volume(struct afs_mount_params *params, |
---|
| 68 | +static struct afs_volume *afs_alloc_volume(struct afs_fs_context *params, |
---|
25 | 69 | struct afs_vldb_entry *vldb, |
---|
26 | 70 | unsigned long type_mask) |
---|
27 | 71 | { |
---|
.. | .. |
---|
39 | 83 | |
---|
40 | 84 | volume->vid = vldb->vid[params->type]; |
---|
41 | 85 | volume->update_at = ktime_get_real_seconds() + afs_volume_record_life; |
---|
42 | | - volume->cell = afs_get_cell(params->cell); |
---|
| 86 | + volume->cell = afs_get_cell(params->cell, afs_cell_trace_get_vol); |
---|
43 | 87 | volume->type = params->type; |
---|
44 | 88 | volume->type_force = params->force; |
---|
45 | 89 | volume->name_len = vldb->name_len; |
---|
46 | 90 | |
---|
47 | 91 | atomic_set(&volume->usage, 1); |
---|
48 | | - INIT_LIST_HEAD(&volume->proc_link); |
---|
| 92 | + INIT_HLIST_NODE(&volume->proc_link); |
---|
49 | 93 | rwlock_init(&volume->servers_lock); |
---|
50 | 94 | rwlock_init(&volume->cb_v_break_lock); |
---|
51 | 95 | memcpy(volume->name, vldb->name, vldb->name_len + 1); |
---|
.. | .. |
---|
57 | 101 | } |
---|
58 | 102 | |
---|
59 | 103 | refcount_set(&slist->usage, 1); |
---|
60 | | - volume->servers = slist; |
---|
| 104 | + rcu_assign_pointer(volume->servers, slist); |
---|
| 105 | + trace_afs_volume(volume->vid, 1, afs_volume_trace_alloc); |
---|
61 | 106 | return volume; |
---|
62 | 107 | |
---|
63 | 108 | error_1: |
---|
64 | | - afs_put_cell(params->net, volume->cell); |
---|
| 109 | + afs_put_cell(volume->cell, afs_cell_trace_put_vol); |
---|
65 | 110 | kfree(volume); |
---|
66 | 111 | error_0: |
---|
67 | 112 | return ERR_PTR(ret); |
---|
| 113 | +} |
---|
| 114 | + |
---|
| 115 | +/* |
---|
| 116 | + * Look up or allocate a volume record. |
---|
| 117 | + */ |
---|
| 118 | +static struct afs_volume *afs_lookup_volume(struct afs_fs_context *params, |
---|
| 119 | + struct afs_vldb_entry *vldb, |
---|
| 120 | + unsigned long type_mask) |
---|
| 121 | +{ |
---|
| 122 | + struct afs_volume *candidate, *volume; |
---|
| 123 | + |
---|
| 124 | + candidate = afs_alloc_volume(params, vldb, type_mask); |
---|
| 125 | + if (IS_ERR(candidate)) |
---|
| 126 | + return candidate; |
---|
| 127 | + |
---|
| 128 | + volume = afs_insert_volume_into_cell(params->cell, candidate); |
---|
| 129 | + if (volume != candidate) |
---|
| 130 | + afs_put_volume(params->net, candidate, afs_volume_trace_put_cell_dup); |
---|
| 131 | + return volume; |
---|
68 | 132 | } |
---|
69 | 133 | |
---|
70 | 134 | /* |
---|
.. | .. |
---|
75 | 139 | const char *volname, |
---|
76 | 140 | size_t volnamesz) |
---|
77 | 141 | { |
---|
78 | | - struct afs_addr_cursor ac; |
---|
79 | | - struct afs_vldb_entry *vldb; |
---|
| 142 | + struct afs_vldb_entry *vldb = ERR_PTR(-EDESTADDRREQ); |
---|
| 143 | + struct afs_vl_cursor vc; |
---|
80 | 144 | int ret; |
---|
81 | 145 | |
---|
82 | | - ret = afs_set_vl_cursor(&ac, cell); |
---|
83 | | - if (ret < 0) |
---|
84 | | - return ERR_PTR(ret); |
---|
| 146 | + if (!afs_begin_vlserver_operation(&vc, cell, key)) |
---|
| 147 | + return ERR_PTR(-ERESTARTSYS); |
---|
85 | 148 | |
---|
86 | | - while (afs_iterate_addresses(&ac)) { |
---|
87 | | - if (!test_bit(ac.index, &ac.alist->probed)) { |
---|
88 | | - ret = afs_vl_get_capabilities(cell->net, &ac, key); |
---|
89 | | - switch (ret) { |
---|
90 | | - case VL_SERVICE: |
---|
91 | | - clear_bit(ac.index, &ac.alist->yfs); |
---|
92 | | - set_bit(ac.index, &ac.alist->probed); |
---|
93 | | - ac.addr->srx_service = ret; |
---|
94 | | - break; |
---|
95 | | - case YFS_VL_SERVICE: |
---|
96 | | - set_bit(ac.index, &ac.alist->yfs); |
---|
97 | | - set_bit(ac.index, &ac.alist->probed); |
---|
98 | | - ac.addr->srx_service = ret; |
---|
99 | | - break; |
---|
100 | | - } |
---|
101 | | - } |
---|
102 | | - |
---|
103 | | - vldb = afs_vl_get_entry_by_name_u(cell->net, &ac, key, |
---|
104 | | - volname, volnamesz); |
---|
105 | | - switch (ac.error) { |
---|
106 | | - case 0: |
---|
107 | | - afs_end_cursor(&ac); |
---|
108 | | - return vldb; |
---|
109 | | - case -ECONNABORTED: |
---|
110 | | - ac.error = afs_abort_to_error(ac.abort_code); |
---|
111 | | - goto error; |
---|
112 | | - case -ENOMEM: |
---|
113 | | - case -ENONET: |
---|
114 | | - goto error; |
---|
115 | | - case -ENETUNREACH: |
---|
116 | | - case -EHOSTUNREACH: |
---|
117 | | - case -ECONNREFUSED: |
---|
118 | | - break; |
---|
119 | | - default: |
---|
120 | | - ac.error = -EIO; |
---|
121 | | - goto error; |
---|
122 | | - } |
---|
| 149 | + while (afs_select_vlserver(&vc)) { |
---|
| 150 | + vldb = afs_vl_get_entry_by_name_u(&vc, volname, volnamesz); |
---|
123 | 151 | } |
---|
124 | 152 | |
---|
125 | | -error: |
---|
126 | | - return ERR_PTR(afs_end_cursor(&ac)); |
---|
| 153 | + ret = afs_end_vlserver_operation(&vc); |
---|
| 154 | + return ret < 0 ? ERR_PTR(ret) : vldb; |
---|
127 | 155 | } |
---|
128 | 156 | |
---|
129 | 157 | /* |
---|
.. | .. |
---|
150 | 178 | * - Rule 3: If parent volume is R/W, then only mount R/W volume unless |
---|
151 | 179 | * explicitly told otherwise |
---|
152 | 180 | */ |
---|
153 | | -struct afs_volume *afs_create_volume(struct afs_mount_params *params) |
---|
| 181 | +struct afs_volume *afs_create_volume(struct afs_fs_context *params) |
---|
154 | 182 | { |
---|
155 | 183 | struct afs_vldb_entry *vldb; |
---|
156 | 184 | struct afs_volume *volume; |
---|
.. | .. |
---|
180 | 208 | } |
---|
181 | 209 | |
---|
182 | 210 | type_mask = 1UL << params->type; |
---|
183 | | - volume = afs_alloc_volume(params, vldb, type_mask); |
---|
| 211 | + volume = afs_lookup_volume(params, vldb, type_mask); |
---|
184 | 212 | |
---|
185 | 213 | error: |
---|
186 | 214 | kfree(vldb); |
---|
.. | .. |
---|
198 | 226 | ASSERTCMP(volume->cache, ==, NULL); |
---|
199 | 227 | #endif |
---|
200 | 228 | |
---|
201 | | - afs_put_serverlist(net, volume->servers); |
---|
202 | | - afs_put_cell(net, volume->cell); |
---|
203 | | - kfree(volume); |
---|
| 229 | + afs_remove_volume_from_cell(volume); |
---|
| 230 | + afs_put_serverlist(net, rcu_access_pointer(volume->servers)); |
---|
| 231 | + afs_put_cell(volume->cell, afs_cell_trace_put_vol); |
---|
| 232 | + trace_afs_volume(volume->vid, atomic_read(&volume->usage), |
---|
| 233 | + afs_volume_trace_free); |
---|
| 234 | + kfree_rcu(volume, rcu); |
---|
204 | 235 | |
---|
205 | 236 | _leave(" [destroyed]"); |
---|
206 | 237 | } |
---|
207 | 238 | |
---|
208 | 239 | /* |
---|
209 | | - * Drop a reference on a volume record. |
---|
| 240 | + * Get a reference on a volume record. |
---|
210 | 241 | */ |
---|
211 | | -void afs_put_volume(struct afs_cell *cell, struct afs_volume *volume) |
---|
| 242 | +struct afs_volume *afs_get_volume(struct afs_volume *volume, |
---|
| 243 | + enum afs_volume_trace reason) |
---|
212 | 244 | { |
---|
213 | 245 | if (volume) { |
---|
214 | | - _enter("%s", volume->name); |
---|
| 246 | + int u = atomic_inc_return(&volume->usage); |
---|
| 247 | + trace_afs_volume(volume->vid, u, reason); |
---|
| 248 | + } |
---|
| 249 | + return volume; |
---|
| 250 | +} |
---|
215 | 251 | |
---|
216 | | - if (atomic_dec_and_test(&volume->usage)) |
---|
217 | | - afs_destroy_volume(cell->net, volume); |
---|
| 252 | + |
---|
| 253 | +/* |
---|
| 254 | + * Drop a reference on a volume record. |
---|
| 255 | + */ |
---|
| 256 | +void afs_put_volume(struct afs_net *net, struct afs_volume *volume, |
---|
| 257 | + enum afs_volume_trace reason) |
---|
| 258 | +{ |
---|
| 259 | + if (volume) { |
---|
| 260 | + afs_volid_t vid = volume->vid; |
---|
| 261 | + int u = atomic_dec_return(&volume->usage); |
---|
| 262 | + trace_afs_volume(vid, u, reason); |
---|
| 263 | + if (u == 0) |
---|
| 264 | + afs_destroy_volume(net, volume); |
---|
218 | 265 | } |
---|
219 | 266 | } |
---|
220 | 267 | |
---|
.. | .. |
---|
230 | 277 | NULL, 0, |
---|
231 | 278 | volume, 0, true); |
---|
232 | 279 | #endif |
---|
233 | | - |
---|
234 | | - write_lock(&volume->cell->proc_lock); |
---|
235 | | - list_add_tail(&volume->proc_link, &volume->cell->proc_volumes); |
---|
236 | | - write_unlock(&volume->cell->proc_lock); |
---|
237 | 280 | } |
---|
238 | 281 | |
---|
239 | 282 | /* |
---|
.. | .. |
---|
242 | 285 | void afs_deactivate_volume(struct afs_volume *volume) |
---|
243 | 286 | { |
---|
244 | 287 | _enter("%s", volume->name); |
---|
245 | | - |
---|
246 | | - write_lock(&volume->cell->proc_lock); |
---|
247 | | - list_del_init(&volume->proc_link); |
---|
248 | | - write_unlock(&volume->cell->proc_lock); |
---|
249 | 288 | |
---|
250 | 289 | #ifdef CONFIG_AFS_FSCACHE |
---|
251 | 290 | fscache_relinquish_cookie(volume->cache, NULL, |
---|
.. | .. |
---|
271 | 310 | /* We look up an ID by passing it as a decimal string in the |
---|
272 | 311 | * operation's name parameter. |
---|
273 | 312 | */ |
---|
274 | | - idsz = sprintf(idbuf, "%u", volume->vid); |
---|
| 313 | + idsz = sprintf(idbuf, "%llu", volume->vid); |
---|
275 | 314 | |
---|
276 | 315 | vldb = afs_vl_lookup_vldb(volume->cell, key, idbuf, idsz); |
---|
277 | 316 | if (IS_ERR(vldb)) { |
---|
.. | .. |
---|
298 | 337 | write_lock(&volume->servers_lock); |
---|
299 | 338 | |
---|
300 | 339 | discard = new; |
---|
301 | | - old = volume->servers; |
---|
| 340 | + old = rcu_dereference_protected(volume->servers, |
---|
| 341 | + lockdep_is_held(&volume->servers_lock)); |
---|
302 | 342 | if (afs_annotate_server_list(new, old)) { |
---|
303 | 343 | new->seq = volume->servers_seq + 1; |
---|
304 | | - volume->servers = new; |
---|
| 344 | + rcu_assign_pointer(volume->servers, new); |
---|
305 | 345 | smp_wmb(); |
---|
306 | 346 | volume->servers_seq++; |
---|
307 | 347 | discard = old; |
---|
308 | 348 | } |
---|
309 | 349 | |
---|
310 | 350 | volume->update_at = ktime_get_real_seconds() + afs_volume_record_life; |
---|
311 | | - clear_bit(AFS_VOLUME_NEEDS_UPDATE, &volume->flags); |
---|
312 | 351 | write_unlock(&volume->servers_lock); |
---|
313 | 352 | ret = 0; |
---|
314 | 353 | |
---|
.. | .. |
---|
323 | 362 | /* |
---|
324 | 363 | * Make sure the volume record is up to date. |
---|
325 | 364 | */ |
---|
326 | | -int afs_check_volume_status(struct afs_volume *volume, struct key *key) |
---|
| 365 | +int afs_check_volume_status(struct afs_volume *volume, struct afs_operation *op) |
---|
327 | 366 | { |
---|
328 | | - time64_t now = ktime_get_real_seconds(); |
---|
329 | 367 | int ret, retries = 0; |
---|
330 | 368 | |
---|
331 | 369 | _enter(""); |
---|
332 | 370 | |
---|
333 | | - if (volume->update_at <= now) |
---|
334 | | - set_bit(AFS_VOLUME_NEEDS_UPDATE, &volume->flags); |
---|
335 | | - |
---|
336 | 371 | retry: |
---|
337 | | - if (!test_bit(AFS_VOLUME_NEEDS_UPDATE, &volume->flags) && |
---|
338 | | - !test_bit(AFS_VOLUME_WAIT, &volume->flags)) { |
---|
339 | | - _leave(" = 0"); |
---|
340 | | - return 0; |
---|
341 | | - } |
---|
| 372 | + if (test_bit(AFS_VOLUME_WAIT, &volume->flags)) |
---|
| 373 | + goto wait; |
---|
| 374 | + if (volume->update_at <= ktime_get_real_seconds() || |
---|
| 375 | + test_bit(AFS_VOLUME_NEEDS_UPDATE, &volume->flags)) |
---|
| 376 | + goto update; |
---|
| 377 | + _leave(" = 0"); |
---|
| 378 | + return 0; |
---|
342 | 379 | |
---|
| 380 | +update: |
---|
343 | 381 | if (!test_and_set_bit_lock(AFS_VOLUME_UPDATING, &volume->flags)) { |
---|
344 | | - ret = afs_update_volume_status(volume, key); |
---|
| 382 | + clear_bit(AFS_VOLUME_NEEDS_UPDATE, &volume->flags); |
---|
| 383 | + ret = afs_update_volume_status(volume, op->key); |
---|
| 384 | + if (ret < 0) |
---|
| 385 | + set_bit(AFS_VOLUME_NEEDS_UPDATE, &volume->flags); |
---|
345 | 386 | clear_bit_unlock(AFS_VOLUME_WAIT, &volume->flags); |
---|
346 | 387 | clear_bit_unlock(AFS_VOLUME_UPDATING, &volume->flags); |
---|
347 | 388 | wake_up_bit(&volume->flags, AFS_VOLUME_WAIT); |
---|
.. | .. |
---|
349 | 390 | return ret; |
---|
350 | 391 | } |
---|
351 | 392 | |
---|
| 393 | +wait: |
---|
352 | 394 | if (!test_bit(AFS_VOLUME_WAIT, &volume->flags)) { |
---|
353 | 395 | _leave(" = 0 [no wait]"); |
---|
354 | 396 | return 0; |
---|
355 | 397 | } |
---|
356 | 398 | |
---|
357 | | - ret = wait_on_bit(&volume->flags, AFS_VOLUME_WAIT, TASK_INTERRUPTIBLE); |
---|
| 399 | + ret = wait_on_bit(&volume->flags, AFS_VOLUME_WAIT, |
---|
| 400 | + (op->flags & AFS_OPERATION_UNINTR) ? |
---|
| 401 | + TASK_UNINTERRUPTIBLE : TASK_INTERRUPTIBLE); |
---|
358 | 402 | if (ret == -ERESTARTSYS) { |
---|
359 | 403 | _leave(" = %d", ret); |
---|
360 | 404 | return ret; |
---|