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 = kmalloc(sizeof(struct afs_volume), GFP_KERNEL);
193 memset(volume, 0, sizeof(struct afs_volume));
194 atomic_set(&volume->usage, 1);
196 volume->type_force = force;
198 volume->vid = vlocation->vldb.vid[type];
200 init_rwsem(&volume->server_sem);
202 /* look up all the applicable server records */
203 for (loop = 0; loop < 8; loop++) {
204 if (vlocation->vldb.srvtmask[loop] & (1 << volume->type)) {
205 ret = afs_server_lookup(
207 &vlocation->vldb.servers[loop],
208 &volume->servers[volume->nservers]);
216 /* attach the cache and volume location */
217 #ifdef AFS_CACHING_SUPPORT
218 cachefs_acquire_cookie(vlocation->cache,
219 &afs_vnode_cache_index_def,
224 afs_get_vlocation(vlocation);
225 volume->vlocation = vlocation;
227 vlocation->vols[type] = volume;
230 _debug("kAFS selected %s volume %08x",
231 afs_voltypes[volume->type], volume->vid);
237 up_write(&cell->vl_sem);
239 afs_put_vlocation(vlocation);
242 _leave(" = %d (%p)", ret, volume);
246 up_write(&cell->vl_sem);
248 for (loop = volume->nservers - 1; loop >= 0; loop--)
249 afs_put_server(volume->servers[loop]);
253 } /* end afs_volume_lookup() */
255 /*****************************************************************************/
257 * destroy a volume record
259 void afs_put_volume(struct afs_volume *volume)
261 struct afs_vlocation *vlocation;
267 _enter("%p", volume);
269 vlocation = volume->vlocation;
272 BUG_ON(atomic_read(&volume->usage) <= 0);
274 /* to prevent a race, the decrement and the dequeue must be effectively
276 down_write(&vlocation->cell->vl_sem);
278 if (likely(!atomic_dec_and_test(&volume->usage))) {
279 up_write(&vlocation->cell->vl_sem);
284 vlocation->vols[volume->type] = NULL;
286 up_write(&vlocation->cell->vl_sem);
288 /* finish cleaning up the volume */
289 #ifdef AFS_CACHING_SUPPORT
290 cachefs_relinquish_cookie(volume->cache, 0);
292 afs_put_vlocation(vlocation);
294 for (loop = volume->nservers - 1; loop >= 0; loop--)
295 afs_put_server(volume->servers[loop]);
299 _leave(" [destroyed]");
300 } /* end afs_put_volume() */
302 /*****************************************************************************/
304 * pick a server to use to try accessing this volume
305 * - returns with an elevated usage count on the server chosen
307 int afs_volume_pick_fileserver(struct afs_volume *volume,
308 struct afs_server **_server)
310 struct afs_server *server;
311 int ret, state, loop;
313 _enter("%s", volume->vlocation->vldb.name);
315 down_read(&volume->server_sem);
317 /* handle the no-server case */
318 if (volume->nservers == 0) {
319 ret = volume->rjservers ? -ENOMEDIUM : -ESTALE;
320 up_read(&volume->server_sem);
321 _leave(" = %d [no servers]", ret);
325 /* basically, just search the list for the first live server and use
328 for (loop = 0; loop < volume->nservers; loop++) {
329 server = volume->servers[loop];
330 state = server->fs_state;
333 /* found an apparently healthy server */
335 afs_get_server(server);
336 up_read(&volume->server_sem);
338 _leave(" = 0 (picked %08x)",
339 ntohl(server->addr.s_addr));
355 ret == -ENETUNREACH ||
356 ret == -EHOSTUNREACH)
363 ret == -ENETUNREACH ||
364 ret == -EHOSTUNREACH ||
365 ret == -ECONNREFUSED)
371 /* no available servers
372 * - TODO: handle the no active servers case better
374 up_read(&volume->server_sem);
375 _leave(" = %d", ret);
377 } /* end afs_volume_pick_fileserver() */
379 /*****************************************************************************/
381 * release a server after use
382 * - releases the ref on the server struct that was acquired by picking
383 * - records result of using a particular server to access a volume
384 * - return 0 to try again, 1 if okay or to issue error
386 int afs_volume_release_fileserver(struct afs_volume *volume,
387 struct afs_server *server,
393 volume->vlocation->vldb.name, ntohl(server->addr.s_addr),
399 server->fs_act_jif = jiffies;
402 /* the fileserver denied all knowledge of the volume */
404 server->fs_act_jif = jiffies;
405 down_write(&volume->server_sem);
407 /* first, find where the server is in the active list (if it
409 for (loop = 0; loop < volume->nservers; loop++)
410 if (volume->servers[loop] == server)
413 /* no longer there - may have been discarded by another op */
414 goto try_next_server_upw;
418 memmove(&volume->servers[loop],
419 &volume->servers[loop + 1],
420 sizeof(volume->servers[loop]) *
421 (volume->nservers - loop));
422 volume->servers[volume->nservers] = NULL;
423 afs_put_server(server);
426 if (volume->nservers > 0)
427 /* another server might acknowledge its existence */
428 goto try_next_server_upw;
430 /* handle the case where all the fileservers have rejected the
432 * - TODO: try asking the fileservers for volume information
433 * - TODO: contact the VL server again to see if the volume is
434 * no longer registered
436 up_write(&volume->server_sem);
437 afs_put_server(server);
438 _leave(" [completely rejected]");
441 /* problem reaching the server */
447 /* mark the server as dead
448 * TODO: vary dead timeout depending on error
450 spin_lock(&server->fs_lock);
451 if (!server->fs_state) {
452 server->fs_dead_jif = jiffies + HZ * 10;
453 server->fs_state = result;
454 printk("kAFS: SERVER DEAD state=%d\n", result);
456 spin_unlock(&server->fs_lock);
457 goto try_next_server;
459 /* miscellaneous error */
461 server->fs_act_jif = jiffies;
467 /* tell the caller to accept the result */
468 afs_put_server(server);
472 /* tell the caller to loop around and try the next server */
474 up_write(&volume->server_sem);
476 afs_put_server(server);
477 _leave(" [try next server]");
480 } /* end afs_volume_release_fileserver() */
482 /*****************************************************************************/
484 * match a volume hash record stored in the cache
486 #ifdef AFS_CACHING_SUPPORT
487 static cachefs_match_val_t afs_volume_cache_match(void *target,
490 const struct afs_cache_vhash *vhash = entry;
491 struct afs_volume *volume = target;
493 _enter("{%u},{%u}", volume->type, vhash->vtype);
495 if (volume->type == vhash->vtype) {
496 _leave(" = SUCCESS");
497 return CACHEFS_MATCH_SUCCESS;
501 return CACHEFS_MATCH_FAILED;
502 } /* end afs_volume_cache_match() */
505 /*****************************************************************************/
507 * update a volume hash record stored in the cache
509 #ifdef AFS_CACHING_SUPPORT
510 static void afs_volume_cache_update(void *source, void *entry)
512 struct afs_cache_vhash *vhash = entry;
513 struct afs_volume *volume = source;
517 vhash->vtype = volume->type;
519 } /* end afs_volume_cache_update() */