]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - fs/afs/volume.c
perf report: Drop cycles 0 for LBR print
[mirror_ubuntu-artful-kernel.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 108
91b467e0 109 volume->bdi.ra_pages = VM_MAX_READAHEAD*1024/PAGE_SIZE;
b4caecd4 110 ret = bdi_setup_and_register(&volume->bdi, "afs");
e1da0222
JA
111 if (ret)
112 goto error_bdi;
113
1da177e4
LT
114 init_rwsem(&volume->server_sem);
115
116 /* look up all the applicable server records */
117 for (loop = 0; loop < 8; loop++) {
118 if (vlocation->vldb.srvtmask[loop] & (1 << volume->type)) {
08e0e7c8
DH
119 server = afs_lookup_server(
120 volume->cell, &vlocation->vldb.servers[loop]);
121 if (IS_ERR(server)) {
122 ret = PTR_ERR(server);
1da177e4 123 goto error_discard;
08e0e7c8 124 }
1da177e4 125
08e0e7c8 126 volume->servers[volume->nservers] = server;
1da177e4
LT
127 volume->nservers++;
128 }
129 }
130
131 /* attach the cache and volume location */
9b3f26c9
DH
132#ifdef CONFIG_AFS_FSCACHE
133 volume->cache = fscache_acquire_cookie(vlocation->cache,
134 &afs_volume_cache_index_def,
94d30ae9 135 volume, true);
1da177e4 136#endif
1da177e4
LT
137 afs_get_vlocation(vlocation);
138 volume->vlocation = vlocation;
139
00d3b7a4 140 vlocation->vols[volume->type] = volume;
1da177e4 141
ec26815a 142success:
1da177e4
LT
143 _debug("kAFS selected %s volume %08x",
144 afs_voltypes[volume->type], volume->vid);
00d3b7a4 145 up_write(&params->cell->vl_sem);
08e0e7c8 146 afs_put_vlocation(vlocation);
08e0e7c8
DH
147 _leave(" = %p", volume);
148 return volume;
1da177e4
LT
149
150 /* clean up */
ec26815a 151error_up:
00d3b7a4 152 up_write(&params->cell->vl_sem);
ec26815a 153error:
1da177e4 154 afs_put_vlocation(vlocation);
08e0e7c8
DH
155 _leave(" = %d", ret);
156 return ERR_PTR(ret);
1da177e4 157
ec26815a 158error_discard:
e1da0222
JA
159 bdi_destroy(&volume->bdi);
160error_bdi:
00d3b7a4 161 up_write(&params->cell->vl_sem);
1da177e4
LT
162
163 for (loop = volume->nservers - 1; loop >= 0; loop--)
164 afs_put_server(volume->servers[loop]);
165
166 kfree(volume);
167 goto error;
ec26815a 168}
1da177e4 169
1da177e4
LT
170/*
171 * destroy a volume record
172 */
173void afs_put_volume(struct afs_volume *volume)
174{
175 struct afs_vlocation *vlocation;
176 int loop;
177
178 if (!volume)
179 return;
180
181 _enter("%p", volume);
182
08e0e7c8 183 ASSERTCMP(atomic_read(&volume->usage), >, 0);
1da177e4 184
08e0e7c8 185 vlocation = volume->vlocation;
1da177e4
LT
186
187 /* to prevent a race, the decrement and the dequeue must be effectively
188 * atomic */
189 down_write(&vlocation->cell->vl_sem);
190
191 if (likely(!atomic_dec_and_test(&volume->usage))) {
192 up_write(&vlocation->cell->vl_sem);
193 _leave("");
194 return;
195 }
196
197 vlocation->vols[volume->type] = NULL;
198
199 up_write(&vlocation->cell->vl_sem);
200
201 /* finish cleaning up the volume */
9b3f26c9
DH
202#ifdef CONFIG_AFS_FSCACHE
203 fscache_relinquish_cookie(volume->cache, 0);
1da177e4
LT
204#endif
205 afs_put_vlocation(vlocation);
206
207 for (loop = volume->nservers - 1; loop >= 0; loop--)
208 afs_put_server(volume->servers[loop]);
209
e1da0222 210 bdi_destroy(&volume->bdi);
1da177e4
LT
211 kfree(volume);
212
213 _leave(" [destroyed]");
ec26815a 214}
1da177e4 215
1da177e4
LT
216/*
217 * pick a server to use to try accessing this volume
218 * - returns with an elevated usage count on the server chosen
219 */
08e0e7c8 220struct afs_server *afs_volume_pick_fileserver(struct afs_vnode *vnode)
1da177e4 221{
08e0e7c8 222 struct afs_volume *volume = vnode->volume;
1da177e4
LT
223 struct afs_server *server;
224 int ret, state, loop;
225
226 _enter("%s", volume->vlocation->vldb.name);
227
08e0e7c8
DH
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;
233 }
234
1da177e4
LT
235 down_read(&volume->server_sem);
236
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);
08e0e7c8 242 return ERR_PTR(ret);
1da177e4
LT
243 }
244
245 /* basically, just search the list for the first live server and use
246 * that */
247 ret = 0;
248 for (loop = 0; loop < volume->nservers; loop++) {
249 server = volume->servers[loop];
250 state = server->fs_state;
251
08e0e7c8
DH
252 _debug("consider %d [%d]", loop, state);
253
1da177e4
LT
254 switch (state) {
255 /* found an apparently healthy server */
256 case 0:
257 afs_get_server(server);
258 up_read(&volume->server_sem);
08e0e7c8
DH
259 _leave(" = %p (picked %08x)",
260 server, ntohl(server->addr.s_addr));
261 return server;
1da177e4
LT
262
263 case -ENETUNREACH:
264 if (ret == 0)
265 ret = state;
266 break;
267
268 case -EHOSTUNREACH:
269 if (ret == 0 ||
270 ret == -ENETUNREACH)
271 ret = state;
272 break;
273
274 case -ECONNREFUSED:
275 if (ret == 0 ||
276 ret == -ENETUNREACH ||
277 ret == -EHOSTUNREACH)
278 ret = state;
279 break;
280
281 default:
282 case -EREMOTEIO:
283 if (ret == 0 ||
284 ret == -ENETUNREACH ||
285 ret == -EHOSTUNREACH ||
286 ret == -ECONNREFUSED)
287 ret = state;
288 break;
289 }
290 }
291
292 /* no available servers
293 * - TODO: handle the no active servers case better
294 */
295 up_read(&volume->server_sem);
296 _leave(" = %d", ret);
08e0e7c8 297 return ERR_PTR(ret);
ec26815a 298}
1da177e4 299
1da177e4
LT
300/*
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
260a9803 305 * - the caller must release the server struct if result was 0
1da177e4 306 */
08e0e7c8 307int afs_volume_release_fileserver(struct afs_vnode *vnode,
1da177e4
LT
308 struct afs_server *server,
309 int result)
310{
08e0e7c8 311 struct afs_volume *volume = vnode->volume;
1da177e4
LT
312 unsigned loop;
313
314 _enter("%s,%08x,%d",
315 volume->vlocation->vldb.name, ntohl(server->addr.s_addr),
316 result);
317
318 switch (result) {
319 /* success */
320 case 0:
321 server->fs_act_jif = jiffies;
08e0e7c8 322 server->fs_state = 0;
260a9803
DH
323 _leave("");
324 return 1;
1da177e4
LT
325
326 /* the fileserver denied all knowledge of the volume */
327 case -ENOMEDIUM:
328 server->fs_act_jif = jiffies;
329 down_write(&volume->server_sem);
330
08e0e7c8 331 /* firstly, find where the server is in the active list (if it
1da177e4
LT
332 * is) */
333 for (loop = 0; loop < volume->nservers; loop++)
334 if (volume->servers[loop] == server)
335 goto present;
336
337 /* no longer there - may have been discarded by another op */
338 goto try_next_server_upw;
339
340 present:
341 volume->nservers--;
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);
348 volume->rjservers++;
349
350 if (volume->nservers > 0)
351 /* another server might acknowledge its existence */
352 goto try_next_server_upw;
353
354 /* handle the case where all the fileservers have rejected the
355 * volume
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
359 */
360 up_write(&volume->server_sem);
361 afs_put_server(server);
362 _leave(" [completely rejected]");
363 return 1;
364
365 /* problem reaching the server */
366 case -ENETUNREACH:
367 case -EHOSTUNREACH:
368 case -ECONNREFUSED:
08e0e7c8 369 case -ETIME:
1da177e4
LT
370 case -ETIMEDOUT:
371 case -EREMOTEIO:
372 /* mark the server as dead
373 * TODO: vary dead timeout depending on error
374 */
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);
380 }
381 spin_unlock(&server->fs_lock);
382 goto try_next_server;
383
384 /* miscellaneous error */
385 default:
386 server->fs_act_jif = jiffies;
387 case -ENOMEM:
388 case -ENONET:
260a9803
DH
389 /* tell the caller to accept the result */
390 afs_put_server(server);
391 _leave(" [local failure]");
392 return 1;
1da177e4
LT
393 }
394
1da177e4 395 /* tell the caller to loop around and try the next server */
ec26815a 396try_next_server_upw:
1da177e4 397 up_write(&volume->server_sem);
ec26815a 398try_next_server:
1da177e4
LT
399 afs_put_server(server);
400 _leave(" [try next server]");
401 return 0;
ec26815a 402}