| 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 <linux/sched.h> |
| 19 | #include "internal.h" |
| 20 | |
| 21 | static const char *afs_voltypes[] = { "R/W", "R/O", "BAK" }; |
| 22 | |
| 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 | */ |
| 45 | struct afs_volume *afs_volume_lookup(struct afs_mount_params *params) |
| 46 | { |
| 47 | struct afs_vlocation *vlocation = NULL; |
| 48 | struct afs_volume *volume = NULL; |
| 49 | struct afs_server *server = NULL; |
| 50 | char srvtmask; |
| 51 | int ret, loop; |
| 52 | |
| 53 | _enter("{%*.*s,%d}", |
| 54 | params->volnamesz, params->volnamesz, params->volname, params->rwpath); |
| 55 | |
| 56 | /* lookup the volume location record */ |
| 57 | vlocation = afs_vlocation_lookup(params->cell, params->key, |
| 58 | params->volname, params->volnamesz); |
| 59 | if (IS_ERR(vlocation)) { |
| 60 | ret = PTR_ERR(vlocation); |
| 61 | vlocation = NULL; |
| 62 | goto error; |
| 63 | } |
| 64 | |
| 65 | /* make the final decision on the type we want */ |
| 66 | ret = -ENOMEDIUM; |
| 67 | if (params->force && !(vlocation->vldb.vidmask & (1 << params->type))) |
| 68 | goto error; |
| 69 | |
| 70 | srvtmask = 0; |
| 71 | for (loop = 0; loop < vlocation->vldb.nservers; loop++) |
| 72 | srvtmask |= vlocation->vldb.srvtmask[loop]; |
| 73 | |
| 74 | if (params->force) { |
| 75 | if (!(srvtmask & (1 << params->type))) |
| 76 | goto error; |
| 77 | } else if (srvtmask & AFS_VOL_VTM_RO) { |
| 78 | params->type = AFSVL_ROVOL; |
| 79 | } else if (srvtmask & AFS_VOL_VTM_RW) { |
| 80 | params->type = AFSVL_RWVOL; |
| 81 | } else { |
| 82 | goto error; |
| 83 | } |
| 84 | |
| 85 | down_write(¶ms->cell->vl_sem); |
| 86 | |
| 87 | /* is the volume already active? */ |
| 88 | if (vlocation->vols[params->type]) { |
| 89 | /* yes - re-use it */ |
| 90 | volume = vlocation->vols[params->type]; |
| 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; |
| 99 | volume = kzalloc(sizeof(struct afs_volume), GFP_KERNEL); |
| 100 | if (!volume) |
| 101 | goto error_up; |
| 102 | |
| 103 | atomic_set(&volume->usage, 1); |
| 104 | volume->type = params->type; |
| 105 | volume->type_force = params->force; |
| 106 | volume->cell = params->cell; |
| 107 | volume->vid = vlocation->vldb.vid[params->type]; |
| 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)) { |
| 114 | server = afs_lookup_server( |
| 115 | volume->cell, &vlocation->vldb.servers[loop]); |
| 116 | if (IS_ERR(server)) { |
| 117 | ret = PTR_ERR(server); |
| 118 | goto error_discard; |
| 119 | } |
| 120 | |
| 121 | volume->servers[volume->nservers] = server; |
| 122 | volume->nservers++; |
| 123 | } |
| 124 | } |
| 125 | |
| 126 | /* attach the cache and volume location */ |
| 127 | #ifdef CONFIG_AFS_FSCACHE |
| 128 | volume->cache = fscache_acquire_cookie(vlocation->cache, |
| 129 | &afs_volume_cache_index_def, |
| 130 | volume, true); |
| 131 | #endif |
| 132 | afs_get_vlocation(vlocation); |
| 133 | volume->vlocation = vlocation; |
| 134 | |
| 135 | vlocation->vols[volume->type] = volume; |
| 136 | |
| 137 | success: |
| 138 | _debug("kAFS selected %s volume %08x", |
| 139 | afs_voltypes[volume->type], volume->vid); |
| 140 | up_write(¶ms->cell->vl_sem); |
| 141 | afs_put_vlocation(vlocation); |
| 142 | _leave(" = %p", volume); |
| 143 | return volume; |
| 144 | |
| 145 | /* clean up */ |
| 146 | error_up: |
| 147 | up_write(¶ms->cell->vl_sem); |
| 148 | error: |
| 149 | afs_put_vlocation(vlocation); |
| 150 | _leave(" = %d", ret); |
| 151 | return ERR_PTR(ret); |
| 152 | |
| 153 | error_discard: |
| 154 | up_write(¶ms->cell->vl_sem); |
| 155 | |
| 156 | for (loop = volume->nservers - 1; loop >= 0; loop--) |
| 157 | afs_put_server(volume->servers[loop]); |
| 158 | |
| 159 | kfree(volume); |
| 160 | goto error; |
| 161 | } |
| 162 | |
| 163 | /* |
| 164 | * destroy a volume record |
| 165 | */ |
| 166 | void 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 | |
| 176 | ASSERTCMP(atomic_read(&volume->usage), >, 0); |
| 177 | |
| 178 | vlocation = volume->vlocation; |
| 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 */ |
| 195 | #ifdef CONFIG_AFS_FSCACHE |
| 196 | fscache_relinquish_cookie(volume->cache, 0); |
| 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]"); |
| 206 | } |
| 207 | |
| 208 | /* |
| 209 | * pick a server to use to try accessing this volume |
| 210 | * - returns with an elevated usage count on the server chosen |
| 211 | */ |
| 212 | struct afs_server *afs_volume_pick_fileserver(struct afs_vnode *vnode) |
| 213 | { |
| 214 | struct afs_volume *volume = vnode->volume; |
| 215 | struct afs_server *server; |
| 216 | int ret, state, loop; |
| 217 | |
| 218 | _enter("%s", volume->vlocation->vldb.name); |
| 219 | |
| 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 | |
| 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); |
| 234 | return ERR_PTR(ret); |
| 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 | |
| 244 | _debug("consider %d [%d]", loop, state); |
| 245 | |
| 246 | switch (state) { |
| 247 | /* found an apparently healthy server */ |
| 248 | case 0: |
| 249 | afs_get_server(server); |
| 250 | up_read(&volume->server_sem); |
| 251 | _leave(" = %p (picked %08x)", |
| 252 | server, ntohl(server->addr.s_addr)); |
| 253 | return server; |
| 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); |
| 289 | return ERR_PTR(ret); |
| 290 | } |
| 291 | |
| 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 |
| 297 | * - the caller must release the server struct if result was 0 |
| 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 | _leave(""); |
| 316 | return 1; |
| 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 | |
| 323 | /* firstly, find where the server is in the active list (if it |
| 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: |
| 361 | case -ETIME: |
| 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: |
| 381 | /* tell the caller to accept the result */ |
| 382 | afs_put_server(server); |
| 383 | _leave(" [local failure]"); |
| 384 | return 1; |
| 385 | } |
| 386 | |
| 387 | /* tell the caller to loop around and try the next server */ |
| 388 | try_next_server_upw: |
| 389 | up_write(&volume->server_sem); |
| 390 | try_next_server: |
| 391 | afs_put_server(server); |
| 392 | _leave(" [try next server]"); |
| 393 | return 0; |
| 394 | } |