1 /* AFS volume management
3 * Copyright (C) 2002, 2007 Red Hat, Inc. All Rights Reserved.
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.
12 #include <linux/kernel.h>
13 #include <linux/module.h>
14 #include <linux/init.h>
15 #include <linux/slab.h>
17 #include <linux/pagemap.h>
18 #include <linux/sched.h>
21 static const char *afs_voltypes[] = { "R/W", "R/O", "BAK" };
24 * lookup a volume by name
25 * - this can be one of the following:
26 * "%[cell:]volume[.]" R/W volume
27 * "#[cell:]volume[.]" R/O or R/W volume (rwparent=0),
28 * or R/W (rwparent=1) volume
29 * "%[cell:]volume.readonly" R/O volume
30 * "#[cell:]volume.readonly" R/O volume
31 * "%[cell:]volume.backup" Backup volume
32 * "#[cell:]volume.backup" Backup volume
34 * The cell name is optional, and defaults to the current cell.
36 * See "The Rules of Mount Point Traversal" in Chapter 5 of the AFS SysAdmin
38 * - Rule 1: Explicit type suffix forces access of that type or nothing
39 * (no suffix, then use Rule 2 & 3)
40 * - Rule 2: If parent volume is R/O, then mount R/O volume by preference, R/W
42 * - Rule 3: If parent volume is R/W, then only mount R/W volume unless
43 * explicitly told otherwise
45 struct afs_volume *afs_volume_lookup(struct afs_mount_params *params)
47 struct afs_vlocation *vlocation = NULL;
48 struct afs_volume *volume = NULL;
49 struct afs_server *server = NULL;
54 params->volnamesz, params->volnamesz, params->volname, params->rwpath);
56 /* lookup the volume location record */
57 vlocation = afs_vlocation_lookup(params->cell, params->key,
58 params->volname, params->volnamesz);
59 if (IS_ERR(vlocation)) {
60 ret = PTR_ERR(vlocation);
65 /* make the final decision on the type we want */
67 if (params->force && !(vlocation->vldb.vidmask & (1 << params->type)))
71 for (loop = 0; loop < vlocation->vldb.nservers; loop++)
72 srvtmask |= vlocation->vldb.srvtmask[loop];
75 if (!(srvtmask & (1 << params->type)))
77 } else if (srvtmask & AFS_VOL_VTM_RO) {
78 params->type = AFSVL_ROVOL;
79 } else if (srvtmask & AFS_VOL_VTM_RW) {
80 params->type = AFSVL_RWVOL;
85 down_write(¶ms->cell->vl_sem);
87 /* is the volume already active? */
88 if (vlocation->vols[params->type]) {
90 volume = vlocation->vols[params->type];
91 afs_get_volume(volume);
95 /* create a new volume record */
96 _debug("creating new volume record");
99 volume = kzalloc(sizeof(struct afs_volume), GFP_KERNEL);
103 atomic_set(&volume->usage, 1);
104 volume->type = params->type;
105 volume->type_force = params->force;
106 volume->cell = params->cell;
107 volume->vid = vlocation->vldb.vid[params->type];
109 volume->bdi.ra_pages = VM_MAX_READAHEAD*1024/PAGE_SIZE;
110 ret = bdi_setup_and_register(&volume->bdi, "afs");
114 init_rwsem(&volume->server_sem);
116 /* look up all the applicable server records */
117 for (loop = 0; loop < 8; loop++) {
118 if (vlocation->vldb.srvtmask[loop] & (1 << volume->type)) {
119 server = afs_lookup_server(
120 volume->cell, &vlocation->vldb.servers[loop]);
121 if (IS_ERR(server)) {
122 ret = PTR_ERR(server);
126 volume->servers[volume->nservers] = server;
131 /* attach the cache and volume location */
132 #ifdef CONFIG_AFS_FSCACHE
133 volume->cache = fscache_acquire_cookie(vlocation->cache,
134 &afs_volume_cache_index_def,
137 afs_get_vlocation(vlocation);
138 volume->vlocation = vlocation;
140 vlocation->vols[volume->type] = volume;
143 _debug("kAFS selected %s volume %08x",
144 afs_voltypes[volume->type], volume->vid);
145 up_write(¶ms->cell->vl_sem);
146 afs_put_vlocation(vlocation);
147 _leave(" = %p", volume);
152 up_write(¶ms->cell->vl_sem);
154 afs_put_vlocation(vlocation);
155 _leave(" = %d", ret);
159 bdi_destroy(&volume->bdi);
161 up_write(¶ms->cell->vl_sem);
163 for (loop = volume->nservers - 1; loop >= 0; loop--)
164 afs_put_server(volume->servers[loop]);
171 * destroy a volume record
173 void afs_put_volume(struct afs_volume *volume)
175 struct afs_vlocation *vlocation;
181 _enter("%p", volume);
183 ASSERTCMP(atomic_read(&volume->usage), >, 0);
185 vlocation = volume->vlocation;
187 /* to prevent a race, the decrement and the dequeue must be effectively
189 down_write(&vlocation->cell->vl_sem);
191 if (likely(!atomic_dec_and_test(&volume->usage))) {
192 up_write(&vlocation->cell->vl_sem);
197 vlocation->vols[volume->type] = NULL;
199 up_write(&vlocation->cell->vl_sem);
201 /* finish cleaning up the volume */
202 #ifdef CONFIG_AFS_FSCACHE
203 fscache_relinquish_cookie(volume->cache, 0);
205 afs_put_vlocation(vlocation);
207 for (loop = volume->nservers - 1; loop >= 0; loop--)
208 afs_put_server(volume->servers[loop]);
210 bdi_destroy(&volume->bdi);
213 _leave(" [destroyed]");
217 * pick a server to use to try accessing this volume
218 * - returns with an elevated usage count on the server chosen
220 struct afs_server *afs_volume_pick_fileserver(struct afs_vnode *vnode)
222 struct afs_volume *volume = vnode->volume;
223 struct afs_server *server;
224 int ret, state, loop;
226 _enter("%s", volume->vlocation->vldb.name);
228 /* stick with the server we're already using if we can */
229 if (vnode->server && vnode->server->fs_state == 0) {
230 afs_get_server(vnode->server);
231 _leave(" = %p [current]", vnode->server);
232 return vnode->server;
235 down_read(&volume->server_sem);
237 /* handle the no-server case */
238 if (volume->nservers == 0) {
239 ret = volume->rjservers ? -ENOMEDIUM : -ESTALE;
240 up_read(&volume->server_sem);
241 _leave(" = %d [no servers]", ret);
245 /* basically, just search the list for the first live server and use
248 for (loop = 0; loop < volume->nservers; loop++) {
249 server = volume->servers[loop];
250 state = server->fs_state;
252 _debug("consider %d [%d]", loop, state);
255 /* found an apparently healthy server */
257 afs_get_server(server);
258 up_read(&volume->server_sem);
259 _leave(" = %p (picked %08x)",
260 server, ntohl(server->addr.s_addr));
276 ret == -ENETUNREACH ||
277 ret == -EHOSTUNREACH)
284 ret == -ENETUNREACH ||
285 ret == -EHOSTUNREACH ||
286 ret == -ECONNREFUSED)
292 /* no available servers
293 * - TODO: handle the no active servers case better
295 up_read(&volume->server_sem);
296 _leave(" = %d", ret);
301 * release a server after use
302 * - releases the ref on the server struct that was acquired by picking
303 * - records result of using a particular server to access a volume
304 * - return 0 to try again, 1 if okay or to issue error
305 * - the caller must release the server struct if result was 0
307 int afs_volume_release_fileserver(struct afs_vnode *vnode,
308 struct afs_server *server,
311 struct afs_volume *volume = vnode->volume;
315 volume->vlocation->vldb.name, ntohl(server->addr.s_addr),
321 server->fs_act_jif = jiffies;
322 server->fs_state = 0;
326 /* the fileserver denied all knowledge of the volume */
328 server->fs_act_jif = jiffies;
329 down_write(&volume->server_sem);
331 /* firstly, find where the server is in the active list (if it
333 for (loop = 0; loop < volume->nservers; loop++)
334 if (volume->servers[loop] == server)
337 /* no longer there - may have been discarded by another op */
338 goto try_next_server_upw;
342 memmove(&volume->servers[loop],
343 &volume->servers[loop + 1],
344 sizeof(volume->servers[loop]) *
345 (volume->nservers - loop));
346 volume->servers[volume->nservers] = NULL;
347 afs_put_server(server);
350 if (volume->nservers > 0)
351 /* another server might acknowledge its existence */
352 goto try_next_server_upw;
354 /* handle the case where all the fileservers have rejected the
356 * - TODO: try asking the fileservers for volume information
357 * - TODO: contact the VL server again to see if the volume is
358 * no longer registered
360 up_write(&volume->server_sem);
361 afs_put_server(server);
362 _leave(" [completely rejected]");
365 /* problem reaching the server */
372 /* mark the server as dead
373 * TODO: vary dead timeout depending on error
375 spin_lock(&server->fs_lock);
376 if (!server->fs_state) {
377 server->fs_dead_jif = jiffies + HZ * 10;
378 server->fs_state = result;
379 printk("kAFS: SERVER DEAD state=%d\n", result);
381 spin_unlock(&server->fs_lock);
382 goto try_next_server;
384 /* miscellaneous error */
386 server->fs_act_jif = jiffies;
389 /* tell the caller to accept the result */
390 afs_put_server(server);
391 _leave(" [local failure]");
395 /* tell the caller to loop around and try the next server */
397 up_write(&volume->server_sem);
399 afs_put_server(server);
400 _leave(" [try next server]");