fs/afs/volume.c

   1 /* AFS volume management
   2  *
   3  * Copyright (C) 2002, 2007 Red Hat, Inc. All Rights Reserved.
   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>
  18 #include "internal.h"
  19
  20 static const char *afs_voltypes[] = { "R/W", "R/O", "BAK" };
  21
  22 /*
  23  * lookup a volume by name
  24  * - this can be one of the following:
  25  *      "%[cell:]volume[.]"             R/W volume
  26  *      "#[cell:]volume[.]"             R/O or R/W volume (rwparent=0),
  27  *                                       or R/W (rwparent=1) volume
  28  *      "%[cell:]volume.readonly"       R/O volume
  29  *      "#[cell:]volume.readonly"       R/O volume
  30  *      "%[cell:]volume.backup"         Backup volume
  31  *      "#[cell:]volume.backup"         Backup volume
  32  *
  33  * The cell name is optional, and defaults to the current cell.
  34  *
  35  * See "The Rules of Mount Point Traversal" in Chapter 5 of the AFS SysAdmin
  36  * Guide
  37  * - Rule 1: Explicit type suffix forces access of that type or nothing
  38  *           (no suffix, then use Rule 2 & 3)
  39  * - Rule 2: If parent volume is R/O, then mount R/O volume by preference, R/W
  40  *           if not available
  41  * - Rule 3: If parent volume is R/W, then only mount R/W volume unless
  42  *           explicitly told otherwise
  43  */
  44 struct afs_volume *afs_volume_lookup(struct afs_mount_params *params)
  45 {
  46         struct afs_vlocation *vlocation = NULL;
  47         struct afs_volume *volume = NULL;
  48         struct afs_server *server = NULL;
  49         char srvtmask;
  50         int ret, loop;
  51
  52         _enter("{%*.*s,%d}",
  53                params->volnamesz, params->volnamesz, params->volname, params->rwpath);
  54
  55         /* lookup the volume location record */
  56         vlocation = afs_vlocation_lookup(params->cell, params->key,
  57                                          params->volname, params->volnamesz);
  58         if (IS_ERR(vlocation)) {
  59                 ret = PTR_ERR(vlocation);
  60                 vlocation = NULL;
  61                 goto error;
  62         }
  63
  64         /* make the final decision on the type we want */
  65         ret = -ENOMEDIUM;
  66         if (params->force && !(vlocation->vldb.vidmask & (1 << params->type)))
  67                 goto error;
  68
  69         srvtmask = 0;
  70         for (loop = 0; loop < vlocation->vldb.nservers; loop++)
  71                 srvtmask |= vlocation->vldb.srvtmask[loop];
  72
  73         if (params->force) {
  74                 if (!(srvtmask & (1 << params->type)))
  75                         goto error;
  76         } else if (srvtmask & AFS_VOL_VTM_RO) {
  77                 params->type = AFSVL_ROVOL;
  78         } else if (srvtmask & AFS_VOL_VTM_RW) {
  79                 params->type = AFSVL_RWVOL;
  80         } else {
  81                 goto error;
  82         }
  83
  84         down_write(&params->cell->vl_sem);
  85
  86         /* is the volume already active? */
  87         if (vlocation->vols[params->type]) {
  88                 /* yes - re-use it */
  89                 volume = vlocation->vols[params->type];
  90                 afs_get_volume(volume);
  91                 goto success;
  92         }
  93
  94         /* create a new volume record */
  95         _debug("creating new volume record");
  96
  97         ret = -ENOMEM;
  98         volume = kzalloc(sizeof(struct afs_volume), GFP_KERNEL);
  99         if (!volume)
 100                 goto error_up;
 101
 102         atomic_set(&volume->usage, 1);
 103         volume->type            = params->type;
 104         volume->type_force      = params->force;
 105         volume->cell            = params->cell;
 106         volume->vid             = vlocation->vldb.vid[params->type];
 107
 108         init_rwsem(&volume->server_sem);
 109
 110         /* look up all the applicable server records */
 111         for (loop = 0; loop < 8; loop++) {
 112                 if (vlocation->vldb.srvtmask[loop] & (1 << volume->type)) {
 113                         server = afs_lookup_server(
 114                                volume->cell, &vlocation->vldb.servers[loop]);
 115                         if (IS_ERR(server)) {
 116                                 ret = PTR_ERR(server);
 117                                 goto error_discard;
 118                         }
 119
 120                         volume->servers[volume->nservers] = server;
 121                         volume->nservers++;
 122                 }
 123         }
 124
 125         /* attach the cache and volume location */
 126 #ifdef AFS_CACHING_SUPPORT
 127         cachefs_acquire_cookie(vlocation->cache,
 128                                &afs_vnode_cache_index_def,
 129                                volume,
 130                                &volume->cache);
 131 #endif
 132
 133         afs_get_vlocation(vlocation);
 134         volume->vlocation = vlocation;
 135
 136         vlocation->vols[volume->type] = volume;
 137
 138 success:
 139         _debug("kAFS selected %s volume %08x",
 140                afs_voltypes[volume->type], volume->vid);
 141         up_write(&params->cell->vl_sem);
 142         afs_put_vlocation(vlocation);
 143         _leave(" = %p", volume);
 144         return volume;
 145
 146         /* clean up */
 147 error_up:
 148         up_write(&params->cell->vl_sem);
 149 error:
 150         afs_put_vlocation(vlocation);
 151         _leave(" = %d", ret);
 152         return ERR_PTR(ret);
 153
 154 error_discard:
 155         up_write(&params->cell->vl_sem);
 156
 157         for (loop = volume->nservers - 1; loop >= 0; loop--)
 158                 afs_put_server(volume->servers[loop]);
 159
 160         kfree(volume);
 161         goto error;
 162 }
 163
 164 /*
 165  * destroy a volume record
 166  */
 167 void afs_put_volume(struct afs_volume *volume)
 168 {
 169         struct afs_vlocation *vlocation;
 170         int loop;
 171
 172         if (!volume)
 173                 return;
 174
 175         _enter("%p", volume);
 176
 177         ASSERTCMP(atomic_read(&volume->usage), >, 0);
 178
 179         vlocation = volume->vlocation;
 180
 181         /* to prevent a race, the decrement and the dequeue must be effectively
 182          * atomic */
 183         down_write(&vlocation->cell->vl_sem);
 184
 185         if (likely(!atomic_dec_and_test(&volume->usage))) {
 186                 up_write(&vlocation->cell->vl_sem);
 187                 _leave("");
 188                 return;
 189         }
 190
 191         vlocation->vols[volume->type] = NULL;
 192
 193         up_write(&vlocation->cell->vl_sem);
 194
 195         /* finish cleaning up the volume */
 196 #ifdef AFS_CACHING_SUPPORT
 197         cachefs_relinquish_cookie(volume->cache, 0);
 198 #endif
 199         afs_put_vlocation(vlocation);
 200
 201         for (loop = volume->nservers - 1; loop >= 0; loop--)
 202                 afs_put_server(volume->servers[loop]);
 203
 204         kfree(volume);
 205
 206         _leave(" [destroyed]");
 207 }
 208
 209 /*
 210  * pick a server to use to try accessing this volume
 211  * - returns with an elevated usage count on the server chosen
 212  */
 213 struct afs_server *afs_volume_pick_fileserver(struct afs_vnode *vnode)
 214 {
 215         struct afs_volume *volume = vnode->volume;
 216         struct afs_server *server;
 217         int ret, state, loop;
 218
 219         _enter("%s", volume->vlocation->vldb.name);
 220
 221         /* stick with the server we're already using if we can */
 222         if (vnode->server && vnode->server->fs_state == 0) {
 223                 afs_get_server(vnode->server);
 224                 _leave(" = %p [current]", vnode->server);
 225                 return vnode->server;
 226         }
 227
 228         down_read(&volume->server_sem);
 229
 230         /* handle the no-server case */
 231         if (volume->nservers == 0) {
 232                 ret = volume->rjservers ? -ENOMEDIUM : -ESTALE;
 233                 up_read(&volume->server_sem);
 234                 _leave(" = %d [no servers]", ret);
 235                 return ERR_PTR(ret);
 236         }
 237
 238         /* basically, just search the list for the first live server and use
 239          * that */
 240         ret = 0;
 241         for (loop = 0; loop < volume->nservers; loop++) {
 242                 server = volume->servers[loop];
 243                 state = server->fs_state;
 244
 245                 _debug("consider %d [%d]", loop, state);
 246
 247                 switch (state) {
 248                         /* found an apparently healthy server */
 249                 case 0:
 250                         afs_get_server(server);
 251                         up_read(&volume->server_sem);
 252                         _leave(" = %p (picked %08x)",
 253                                server, ntohl(server->addr.s_addr));
 254                         return server;
 255
 256                 case -ENETUNREACH:
 257                         if (ret == 0)
 258                                 ret = state;
 259                         break;
 260
 261                 case -EHOSTUNREACH:
 262                         if (ret == 0 ||
 263                             ret == -ENETUNREACH)
 264                                 ret = state;
 265                         break;
 266
 267                 case -ECONNREFUSED:
 268                         if (ret == 0 ||
 269                             ret == -ENETUNREACH ||
 270                             ret == -EHOSTUNREACH)
 271                                 ret = state;
 272                         break;
 273
 274                 default:
 275                 case -EREMOTEIO:
 276                         if (ret == 0 ||
 277                             ret == -ENETUNREACH ||
 278                             ret == -EHOSTUNREACH ||
 279                             ret == -ECONNREFUSED)
 280                                 ret = state;
 281                         break;
 282                 }
 283         }
 284
 285         /* no available servers
 286          * - TODO: handle the no active servers case better
 287          */
 288         up_read(&volume->server_sem);
 289         _leave(" = %d", ret);
 290         return ERR_PTR(ret);
 291 }
 292
 293 /*
 294  * release a server after use
 295  * - releases the ref on the server struct that was acquired by picking
 296  * - records result of using a particular server to access a volume
 297  * - return 0 to try again, 1 if okay or to issue error
 298  */
 299 int afs_volume_release_fileserver(struct afs_vnode *vnode,
 300                                   struct afs_server *server,
 301                                   int result)
 302 {
 303         struct afs_volume *volume = vnode->volume;
 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;
 314                 server->fs_state = 0;
 315                 break;
 316
 317                 /* the fileserver denied all knowledge of the volume */
 318         case -ENOMEDIUM:
 319                 server->fs_act_jif = jiffies;
 320                 down_write(&volume->server_sem);
 321
 322                 /* firstly, find where the server is in the active list (if it
 323                  * is) */
 324                 for (loop = 0; loop < volume->nservers; loop++)
 325                         if (volume->servers[loop] == server)
 326                                 goto present;
 327
 328                 /* no longer there - may have been discarded by another op */
 329                 goto try_next_server_upw;
 330
 331         present:
 332                 volume->nservers--;
 333                 memmove(&volume->servers[loop],
 334                         &volume->servers[loop + 1],
 335                         sizeof(volume->servers[loop]) *
 336                         (volume->nservers - loop));
 337                 volume->servers[volume->nservers] = NULL;
 338                 afs_put_server(server);
 339                 volume->rjservers++;
 340
 341                 if (volume->nservers > 0)
 342                         /* another server might acknowledge its existence */
 343                         goto try_next_server_upw;
 344
 345                 /* handle the case where all the fileservers have rejected the
 346                  * volume
 347                  * - TODO: try asking the fileservers for volume information
 348                  * - TODO: contact the VL server again to see if the volume is
 349                  *         no longer registered
 350                  */
 351                 up_write(&volume->server_sem);
 352                 afs_put_server(server);
 353                 _leave(" [completely rejected]");
 354                 return 1;
 355
 356                 /* problem reaching the server */
 357         case -ENETUNREACH:
 358         case -EHOSTUNREACH:
 359         case -ECONNREFUSED:
 360         case -ETIME:
 361         case -ETIMEDOUT:
 362         case -EREMOTEIO:
 363                 /* mark the server as dead
 364                  * TODO: vary dead timeout depending on error
 365                  */
 366                 spin_lock(&server->fs_lock);
 367                 if (!server->fs_state) {
 368                         server->fs_dead_jif = jiffies + HZ * 10;
 369                         server->fs_state = result;
 370                         printk("kAFS: SERVER DEAD state=%d\n", result);
 371                 }
 372                 spin_unlock(&server->fs_lock);
 373                 goto try_next_server;
 374
 375                 /* miscellaneous error */
 376         default:
 377                 server->fs_act_jif = jiffies;
 378         case -ENOMEM:
 379         case -ENONET:
 380                 break;
 381         }
 382
 383         /* tell the caller to accept the result */
 384         afs_put_server(server);
 385         _leave("");
 386         return 1;
 387
 388         /* tell the caller to loop around and try the next server */
 389 try_next_server_upw:
 390         up_write(&volume->server_sem);
 391 try_next_server:
 392         afs_put_server(server);
 393         _leave(" [try next server]");
 394         return 0;
 395 }