1 /* volume.c: AFS volume management
3 * Copyright (C) 2002 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
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>
22 #include "cmservice.h"
28 static const char *afs_voltypes[] = { "R/W", "R/O", "BAK" };
31 #ifdef AFS_CACHING_SUPPORT
32 static cachefs_match_val_t afs_volume_cache_match(void *target,
34 static void afs_volume_cache_update(void *source, void *entry);
36 struct cachefs_index_def afs_volume_cache_index_def = {
38 .data_size = sizeof(struct afs_cache_vhash),
39 .keys[0] = { CACHEFS_INDEX_KEYS_BIN, 1 },
40 .keys[1] = { CACHEFS_INDEX_KEYS_BIN, 1 },
41 .match = afs_volume_cache_match,
42 .update = afs_volume_cache_update,
46 /*****************************************************************************/
48 * lookup a volume by name
49 * - this can be one of the following:
50 * "%[cell:]volume[.]" R/W volume
51 * "#[cell:]volume[.]" R/O or R/W volume (rwparent=0),
52 * or R/W (rwparent=1) volume
53 * "%[cell:]volume.readonly" R/O volume
54 * "#[cell:]volume.readonly" R/O volume
55 * "%[cell:]volume.backup" Backup volume
56 * "#[cell:]volume.backup" Backup volume
58 * The cell name is optional, and defaults to the current cell.
60 * See "The Rules of Mount Point Traversal" in Chapter 5 of the AFS SysAdmin
62 * - Rule 1: Explicit type suffix forces access of that type or nothing
63 * (no suffix, then use Rule 2 & 3)
64 * - Rule 2: If parent volume is R/O, then mount R/O volume by preference, R/W
66 * - Rule 3: If parent volume is R/W, then only mount R/W volume unless
67 * explicitly told otherwise
69 int afs_volume_lookup(const char *name, struct afs_cell *cell, int rwpath,
70 struct afs_volume **_volume)
72 struct afs_vlocation *vlocation = NULL;
73 struct afs_volume *volume = NULL;
75 const char *cellname, *volname, *suffix;
77 int force, ret, loop, cellnamesz, volnamesz;
79 _enter("%s,,%d,", name, rwpath);
81 if (!name || (name[0] != '%' && name[0] != '#') || !name[1]) {
82 printk("kAFS: unparsable volume name\n");
86 /* determine the type of volume we're looking for */
90 if (rwpath || name[0] == '%') {
95 suffix = strrchr(name, '.');
97 if (strcmp(suffix, ".readonly") == 0) {
101 else if (strcmp(suffix, ".backup") == 0) {
102 type = AFSVL_BACKVOL;
105 else if (suffix[1] == 0) {
112 /* split the cell and volume names */
114 volname = strchr(name, ':');
117 cellnamesz = volname - name;
126 volnamesz = suffix ? suffix - volname : strlen(volname);
128 _debug("CELL:%*.*s [%p] VOLUME:%*.*s SUFFIX:%s TYPE:%d%s",
129 cellnamesz, cellnamesz, cellname ?: "", cell,
130 volnamesz, volnamesz, volname, suffix ?: "-",
132 force ? " FORCE" : "");
134 /* lookup the cell record */
135 if (cellname || !cell) {
136 ret = afs_cell_lookup(cellname, cellnamesz, &cell);
138 printk("kAFS: unable to lookup cell '%s'\n",
147 /* lookup the volume location record */
148 ret = afs_vlocation_lookup(cell, volname, volnamesz, &vlocation);
152 /* make the final decision on the type we want */
154 if (force && !(vlocation->vldb.vidmask & (1 << type)))
158 for (loop = 0; loop < vlocation->vldb.nservers; loop++)
159 srvtmask |= vlocation->vldb.srvtmask[loop];
162 if (!(srvtmask & (1 << type)))
165 else if (srvtmask & AFS_VOL_VTM_RO) {
168 else if (srvtmask & AFS_VOL_VTM_RW) {
175 down_write(&cell->vl_sem);
177 /* is the volume already active? */
178 if (vlocation->vols[type]) {
179 /* yes - re-use it */
180 volume = vlocation->vols[type];
181 afs_get_volume(volume);
185 /* create a new volume record */
186 _debug("creating new volume record");
189 volume = kzalloc(sizeof(struct afs_volume), GFP_KERNEL);
193 atomic_set(&volume->usage, 1);
195 volume->type_force = force;
197 volume->vid = vlocation->vldb.vid[type];
199 init_rwsem(&volume->server_sem);
201 /* look up all the applicable server records */
202 for (loop = 0; loop < 8; loop++) {
203 if (vlocation->vldb.srvtmask[loop] & (1 << volume->type)) {
204 ret = afs_server_lookup(
206 &vlocation->vldb.servers[loop],
207 &volume->servers[volume->nservers]);
215 /* attach the cache and volume location */
216 #ifdef AFS_CACHING_SUPPORT
217 cachefs_acquire_cookie(vlocation->cache,
218 &afs_vnode_cache_index_def,
223 afs_get_vlocation(vlocation);
224 volume->vlocation = vlocation;
226 vlocation->vols[type] = volume;
229 _debug("kAFS selected %s volume %08x",
230 afs_voltypes[volume->type], volume->vid);
236 up_write(&cell->vl_sem);
238 afs_put_vlocation(vlocation);
241 _leave(" = %d (%p)", ret, volume);
245 up_write(&cell->vl_sem);
247 for (loop = volume->nservers - 1; loop >= 0; loop--)
248 afs_put_server(volume->servers[loop]);
252 } /* end afs_volume_lookup() */
254 /*****************************************************************************/
256 * destroy a volume record
258 void afs_put_volume(struct afs_volume *volume)
260 struct afs_vlocation *vlocation;
266 _enter("%p", volume);
268 vlocation = volume->vlocation;
271 BUG_ON(atomic_read(&volume->usage) <= 0);
273 /* to prevent a race, the decrement and the dequeue must be effectively
275 down_write(&vlocation->cell->vl_sem);
277 if (likely(!atomic_dec_and_test(&volume->usage))) {
278 up_write(&vlocation->cell->vl_sem);
283 vlocation->vols[volume->type] = NULL;
285 up_write(&vlocation->cell->vl_sem);
287 /* finish cleaning up the volume */
288 #ifdef AFS_CACHING_SUPPORT
289 cachefs_relinquish_cookie(volume->cache, 0);
291 afs_put_vlocation(vlocation);
293 for (loop = volume->nservers - 1; loop >= 0; loop--)
294 afs_put_server(volume->servers[loop]);
298 _leave(" [destroyed]");
299 } /* end afs_put_volume() */
301 /*****************************************************************************/
303 * pick a server to use to try accessing this volume
304 * - returns with an elevated usage count on the server chosen
306 int afs_volume_pick_fileserver(struct afs_volume *volume,
307 struct afs_server **_server)
309 struct afs_server *server;
310 int ret, state, loop;
312 _enter("%s", volume->vlocation->vldb.name);
314 down_read(&volume->server_sem);
316 /* handle the no-server case */
317 if (volume->nservers == 0) {
318 ret = volume->rjservers ? -ENOMEDIUM : -ESTALE;
319 up_read(&volume->server_sem);
320 _leave(" = %d [no servers]", ret);
324 /* basically, just search the list for the first live server and use
327 for (loop = 0; loop < volume->nservers; loop++) {
328 server = volume->servers[loop];
329 state = server->fs_state;
332 /* found an apparently healthy server */
334 afs_get_server(server);
335 up_read(&volume->server_sem);
337 _leave(" = 0 (picked %08x)",
338 ntohl(server->addr.s_addr));
354 ret == -ENETUNREACH ||
355 ret == -EHOSTUNREACH)
362 ret == -ENETUNREACH ||
363 ret == -EHOSTUNREACH ||
364 ret == -ECONNREFUSED)
370 /* no available servers
371 * - TODO: handle the no active servers case better
373 up_read(&volume->server_sem);
374 _leave(" = %d", ret);
376 } /* end afs_volume_pick_fileserver() */
378 /*****************************************************************************/
380 * release a server after use
381 * - releases the ref on the server struct that was acquired by picking
382 * - records result of using a particular server to access a volume
383 * - return 0 to try again, 1 if okay or to issue error
385 int afs_volume_release_fileserver(struct afs_volume *volume,
386 struct afs_server *server,
392 volume->vlocation->vldb.name, ntohl(server->addr.s_addr),
398 server->fs_act_jif = jiffies;
401 /* the fileserver denied all knowledge of the volume */
403 server->fs_act_jif = jiffies;
404 down_write(&volume->server_sem);
406 /* first, find where the server is in the active list (if it
408 for (loop = 0; loop < volume->nservers; loop++)
409 if (volume->servers[loop] == server)
412 /* no longer there - may have been discarded by another op */
413 goto try_next_server_upw;
417 memmove(&volume->servers[loop],
418 &volume->servers[loop + 1],
419 sizeof(volume->servers[loop]) *
420 (volume->nservers - loop));
421 volume->servers[volume->nservers] = NULL;
422 afs_put_server(server);
425 if (volume->nservers > 0)
426 /* another server might acknowledge its existence */
427 goto try_next_server_upw;
429 /* handle the case where all the fileservers have rejected the
431 * - TODO: try asking the fileservers for volume information
432 * - TODO: contact the VL server again to see if the volume is
433 * no longer registered
435 up_write(&volume->server_sem);
436 afs_put_server(server);
437 _leave(" [completely rejected]");
440 /* problem reaching the server */
446 /* mark the server as dead
447 * TODO: vary dead timeout depending on error
449 spin_lock(&server->fs_lock);
450 if (!server->fs_state) {
451 server->fs_dead_jif = jiffies + HZ * 10;
452 server->fs_state = result;
453 printk("kAFS: SERVER DEAD state=%d\n", result);
455 spin_unlock(&server->fs_lock);
456 goto try_next_server;
458 /* miscellaneous error */
460 server->fs_act_jif = jiffies;
466 /* tell the caller to accept the result */
467 afs_put_server(server);
471 /* tell the caller to loop around and try the next server */
473 up_write(&volume->server_sem);
475 afs_put_server(server);
476 _leave(" [try next server]");
479 } /* end afs_volume_release_fileserver() */
481 /*****************************************************************************/
483 * match a volume hash record stored in the cache
485 #ifdef AFS_CACHING_SUPPORT
486 static cachefs_match_val_t afs_volume_cache_match(void *target,
489 const struct afs_cache_vhash *vhash = entry;
490 struct afs_volume *volume = target;
492 _enter("{%u},{%u}", volume->type, vhash->vtype);
494 if (volume->type == vhash->vtype) {
495 _leave(" = SUCCESS");
496 return CACHEFS_MATCH_SUCCESS;
500 return CACHEFS_MATCH_FAILED;
501 } /* end afs_volume_cache_match() */
504 /*****************************************************************************/
506 * update a volume hash record stored in the cache
508 #ifdef AFS_CACHING_SUPPORT
509 static void afs_volume_cache_update(void *source, void *entry)
511 struct afs_cache_vhash *vhash = entry;
512 struct afs_volume *volume = source;
516 vhash->vtype = volume->type;
518 } /* end afs_volume_cache_update() */