fs: no games with DCACHE_UNHASHED
[linux-2.6-block.git] / fs / afs / volume.c
CommitLineData
ec26815a 1/* AFS volume management
1da177e4 2 *
08e0e7c8 3 * Copyright (C) 2002, 2007 Red Hat, Inc. All Rights Reserved.
1da177e4
LT
4 * 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 */
11
12#include <linux/kernel.h>
13#include <linux/module.h>
14#include <linux/init.h>
15#include <linux/slab.h>
16#include <linux/fs.h>
17#include <linux/pagemap.h>
e8edc6e0 18#include <linux/sched.h>
1da177e4
LT
19#include "internal.h"
20
1da177e4 21static const char *afs_voltypes[] = { "R/W", "R/O", "BAK" };
1da177e4 22
1da177e4
LT
23/*
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
33 *
34 * The cell name is optional, and defaults to the current cell.
35 *
36 * See "The Rules of Mount Point Traversal" in Chapter 5 of the AFS SysAdmin
37 * Guide
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
41 * if not available
42 * - Rule 3: If parent volume is R/W, then only mount R/W volume unless
43 * explicitly told otherwise
44 */
00d3b7a4 45struct afs_volume *afs_volume_lookup(struct afs_mount_params *params)
1da177e4
LT
46{
47 struct afs_vlocation *vlocation = NULL;
48 struct afs_volume *volume = NULL;
08e0e7c8 49 struct afs_server *server = NULL;
1da177e4 50 char srvtmask;
00d3b7a4 51 int ret, loop;
1da177e4 52
00d3b7a4
DH
53 _enter("{%*.*s,%d}",
54 params->volnamesz, params->volnamesz, params->volname, params->rwpath);
1da177e4
LT
55
56 /* lookup the volume location record */
00d3b7a4
DH
57 vlocation = afs_vlocation_lookup(params->cell, params->key,
58 params->volname, params->volnamesz);
08e0e7c8
DH
59 if (IS_ERR(vlocation)) {
60 ret = PTR_ERR(vlocation);
61 vlocation = NULL;
1da177e4 62 goto error;
08e0e7c8 63 }
1da177e4
LT
64
65 /* make the final decision on the type we want */
66 ret = -ENOMEDIUM;
00d3b7a4 67 if (params->force && !(vlocation->vldb.vidmask & (1 << params->type)))
1da177e4
LT
68 goto error;
69
70 srvtmask = 0;
71 for (loop = 0; loop < vlocation->vldb.nservers; loop++)
72 srvtmask |= vlocation->vldb.srvtmask[loop];
73
00d3b7a4
DH
74 if (params->force) {
75 if (!(srvtmask & (1 << params->type)))
1da177e4 76 goto error;
ec26815a 77 } else if (srvtmask & AFS_VOL_VTM_RO) {
00d3b7a4 78 params->type = AFSVL_ROVOL;
ec26815a 79 } else if (srvtmask & AFS_VOL_VTM_RW) {
00d3b7a4 80 params->type = AFSVL_RWVOL;
ec26815a 81 } else {
1da177e4
LT
82 goto error;
83 }
84
00d3b7a4 85 down_write(&params->cell->vl_sem);
1da177e4
LT
86
87 /* is the volume already active? */
00d3b7a4 88 if (vlocation->vols[params->type]) {
1da177e4 89 /* yes - re-use it */
00d3b7a4 90 volume = vlocation->vols[params->type];
1da177e4
LT
91 afs_get_volume(volume);
92 goto success;
93 }
94
95 /* create a new volume record */
96 _debug("creating new volume record");
97
98 ret = -ENOMEM;
f8314dc6 99 volume = kzalloc(sizeof(struct afs_volume), GFP_KERNEL);
1da177e4
LT
100 if (!volume)
101 goto error_up;
102
1da177e4 103 atomic_set(&volume->usage, 1);
00d3b7a4
DH
104 volume->type = params->type;
105 volume->type_force = params->force;
106 volume->cell = params->cell;
107 volume->vid = vlocation->vldb.vid[params->type];
1da177e4
LT
108
109 init_rwsem(&volume->server_sem);
110
111 /* look up all the applicable server records */
112 for (loop = 0; loop < 8; loop++) {
113 if (vlocation->vldb.srvtmask[loop] & (1 << volume->type)) {
08e0e7c8
DH
114 server = afs_lookup_server(
115 volume->cell, &vlocation->vldb.servers[loop]);
116 if (IS_ERR(server)) {
117 ret = PTR_ERR(server);
1da177e4 118 goto error_discard;
08e0e7c8 119 }
1da177e4 120
08e0e7c8 121 volume->servers[volume->nservers] = server;
1da177e4
LT
122 volume->nservers++;
123 }
124 }
125
126 /* attach the cache and volume location */
9b3f26c9
DH
127#ifdef CONFIG_AFS_FSCACHE
128 volume->cache = fscache_acquire_cookie(vlocation->cache,
129 &afs_volume_cache_index_def,
130 volume);
1da177e4 131#endif
1da177e4
LT
132 afs_get_vlocation(vlocation);
133 volume->vlocation = vlocation;
134
00d3b7a4 135 vlocation->vols[volume->type] = volume;
1da177e4 136
ec26815a 137success:
1da177e4
LT
138 _debug("kAFS selected %s volume %08x",
139 afs_voltypes[volume->type], volume->vid);
00d3b7a4 140 up_write(&params->cell->vl_sem);
08e0e7c8 141 afs_put_vlocation(vlocation);
08e0e7c8
DH
142 _leave(" = %p", volume);
143 return volume;
1da177e4
LT
144
145 /* clean up */
ec26815a 146error_up:
00d3b7a4 147 up_write(&params->cell->vl_sem);
ec26815a 148error:
1da177e4 149 afs_put_vlocation(vlocation);
08e0e7c8
DH
150 _leave(" = %d", ret);
151 return ERR_PTR(ret);
1da177e4 152
ec26815a 153error_discard:
00d3b7a4 154 up_write(&params->cell->vl_sem);
1da177e4
LT
155
156 for (loop = volume->nservers - 1; loop >= 0; loop--)
157 afs_put_server(volume->servers[loop]);
158
159 kfree(volume);
160 goto error;
ec26815a 161}
1da177e4 162
1da177e4
LT
163/*
164 * destroy a volume record
165 */
166void afs_put_volume(struct afs_volume *volume)
167{
168 struct afs_vlocation *vlocation;
169 int loop;
170
171 if (!volume)
172 return;
173
174 _enter("%p", volume);
175
08e0e7c8 176 ASSERTCMP(atomic_read(&volume->usage), >, 0);
1da177e4 177
08e0e7c8 178 vlocation = volume->vlocation;
1da177e4
LT
179
180 /* to prevent a race, the decrement and the dequeue must be effectively
181 * atomic */
182 down_write(&vlocation->cell->vl_sem);
183
184 if (likely(!atomic_dec_and_test(&volume->usage))) {
185 up_write(&vlocation->cell->vl_sem);
186 _leave("");
187 return;
188 }
189
190 vlocation->vols[volume->type] = NULL;
191
192 up_write(&vlocation->cell->vl_sem);
193
194 /* finish cleaning up the volume */
9b3f26c9
DH
195#ifdef CONFIG_AFS_FSCACHE
196 fscache_relinquish_cookie(volume->cache, 0);
1da177e4
LT
197#endif
198 afs_put_vlocation(vlocation);
199
200 for (loop = volume->nservers - 1; loop >= 0; loop--)
201 afs_put_server(volume->servers[loop]);
202
203 kfree(volume);
204
205 _leave(" [destroyed]");
ec26815a 206}
1da177e4 207
1da177e4
LT
208/*
209 * pick a server to use to try accessing this volume
210 * - returns with an elevated usage count on the server chosen
211 */
08e0e7c8 212struct afs_server *afs_volume_pick_fileserver(struct afs_vnode *vnode)
1da177e4 213{
08e0e7c8 214 struct afs_volume *volume = vnode->volume;
1da177e4
LT
215 struct afs_server *server;
216 int ret, state, loop;
217
218 _enter("%s", volume->vlocation->vldb.name);
219
08e0e7c8
DH
220 /* stick with the server we're already using if we can */
221 if (vnode->server && vnode->server->fs_state == 0) {
222 afs_get_server(vnode->server);
223 _leave(" = %p [current]", vnode->server);
224 return vnode->server;
225 }
226
1da177e4
LT
227 down_read(&volume->server_sem);
228
229 /* handle the no-server case */
230 if (volume->nservers == 0) {
231 ret = volume->rjservers ? -ENOMEDIUM : -ESTALE;
232 up_read(&volume->server_sem);
233 _leave(" = %d [no servers]", ret);
08e0e7c8 234 return ERR_PTR(ret);
1da177e4
LT
235 }
236
237 /* basically, just search the list for the first live server and use
238 * that */
239 ret = 0;
240 for (loop = 0; loop < volume->nservers; loop++) {
241 server = volume->servers[loop];
242 state = server->fs_state;
243
08e0e7c8
DH
244 _debug("consider %d [%d]", loop, state);
245
1da177e4
LT
246 switch (state) {
247 /* found an apparently healthy server */
248 case 0:
249 afs_get_server(server);
250 up_read(&volume->server_sem);
08e0e7c8
DH
251 _leave(" = %p (picked %08x)",
252 server, ntohl(server->addr.s_addr));
253 return server;
1da177e4
LT
254
255 case -ENETUNREACH:
256 if (ret == 0)
257 ret = state;
258 break;
259
260 case -EHOSTUNREACH:
261 if (ret == 0 ||
262 ret == -ENETUNREACH)
263 ret = state;
264 break;
265
266 case -ECONNREFUSED:
267 if (ret == 0 ||
268 ret == -ENETUNREACH ||
269 ret == -EHOSTUNREACH)
270 ret = state;
271 break;
272
273 default:
274 case -EREMOTEIO:
275 if (ret == 0 ||
276 ret == -ENETUNREACH ||
277 ret == -EHOSTUNREACH ||
278 ret == -ECONNREFUSED)
279 ret = state;
280 break;
281 }
282 }
283
284 /* no available servers
285 * - TODO: handle the no active servers case better
286 */
287 up_read(&volume->server_sem);
288 _leave(" = %d", ret);
08e0e7c8 289 return ERR_PTR(ret);
ec26815a 290}
1da177e4 291
1da177e4
LT
292/*
293 * release a server after use
294 * - releases the ref on the server struct that was acquired by picking
295 * - records result of using a particular server to access a volume
296 * - return 0 to try again, 1 if okay or to issue error
260a9803 297 * - the caller must release the server struct if result was 0
1da177e4 298 */
08e0e7c8 299int afs_volume_release_fileserver(struct afs_vnode *vnode,
1da177e4
LT
300 struct afs_server *server,
301 int result)
302{
08e0e7c8 303 struct afs_volume *volume = vnode->volume;
1da177e4
LT
304 unsigned loop;
305
306 _enter("%s,%08x,%d",
307 volume->vlocation->vldb.name, ntohl(server->addr.s_addr),
308 result);
309
310 switch (result) {
311 /* success */
312 case 0:
313 server->fs_act_jif = jiffies;
08e0e7c8 314 server->fs_state = 0;
260a9803
DH
315 _leave("");
316 return 1;
1da177e4
LT
317
318 /* the fileserver denied all knowledge of the volume */
319 case -ENOMEDIUM:
320 server->fs_act_jif = jiffies;
321 down_write(&volume->server_sem);
322
08e0e7c8 323 /* firstly, find where the server is in the active list (if it
1da177e4
LT
324 * is) */
325 for (loop = 0; loop < volume->nservers; loop++)
326 if (volume->servers[loop] == server)
327 goto present;
328
329 /* no longer there - may have been discarded by another op */
330 goto try_next_server_upw;
331
332 present:
333 volume->nservers--;
334 memmove(&volume->servers[loop],
335 &volume->servers[loop + 1],
336 sizeof(volume->servers[loop]) *
337 (volume->nservers - loop));
338 volume->servers[volume->nservers] = NULL;
339 afs_put_server(server);
340 volume->rjservers++;
341
342 if (volume->nservers > 0)
343 /* another server might acknowledge its existence */
344 goto try_next_server_upw;
345
346 /* handle the case where all the fileservers have rejected the
347 * volume
348 * - TODO: try asking the fileservers for volume information
349 * - TODO: contact the VL server again to see if the volume is
350 * no longer registered
351 */
352 up_write(&volume->server_sem);
353 afs_put_server(server);
354 _leave(" [completely rejected]");
355 return 1;
356
357 /* problem reaching the server */
358 case -ENETUNREACH:
359 case -EHOSTUNREACH:
360 case -ECONNREFUSED:
08e0e7c8 361 case -ETIME:
1da177e4
LT
362 case -ETIMEDOUT:
363 case -EREMOTEIO:
364 /* mark the server as dead
365 * TODO: vary dead timeout depending on error
366 */
367 spin_lock(&server->fs_lock);
368 if (!server->fs_state) {
369 server->fs_dead_jif = jiffies + HZ * 10;
370 server->fs_state = result;
371 printk("kAFS: SERVER DEAD state=%d\n", result);
372 }
373 spin_unlock(&server->fs_lock);
374 goto try_next_server;
375
376 /* miscellaneous error */
377 default:
378 server->fs_act_jif = jiffies;
379 case -ENOMEM:
380 case -ENONET:
260a9803
DH
381 /* tell the caller to accept the result */
382 afs_put_server(server);
383 _leave(" [local failure]");
384 return 1;
1da177e4
LT
385 }
386
1da177e4 387 /* tell the caller to loop around and try the next server */
ec26815a 388try_next_server_upw:
1da177e4 389 up_write(&volume->server_sem);
ec26815a 390try_next_server:
1da177e4
LT
391 afs_put_server(server);
392 _leave(" [try next server]");
393 return 0;
ec26815a 394}