lustre/ldlm/ldlm_lockd.c

   1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
   2  * vim:expandtab:shiftwidth=8:tabstop=8:
   3  *
   4  * Copyright (C) 2002 Cluster File Systems, Inc.
   5  *
   6  * This code is issued under the GNU General Public License.
   7  * See the file COPYING in this distribution
   8  *
   9  * by Cluster File Systems, Inc.
  10  */
  11
  12 #define EXPORT_SYMTAB
  13 #define DEBUG_SUBSYSTEM S_LDLM
  14
  15 #include <linux/module.h>
  16 #include <linux/slab.h>
  17 #include <linux/lustre_dlm.h>
  18
  19 extern kmem_cache_t *ldlm_resource_slab;
  20 extern kmem_cache_t *ldlm_lock_slab;
  21 extern int (*mds_reint_p)(int offset, struct ptlrpc_request *req);
  22 extern int (*mds_getattr_name_p)(int offset, struct ptlrpc_request *req);
  23
  24 static int ldlm_handle_enqueue(struct ptlrpc_request *req)
  25 {
  26         struct obd_device *obddev = req->rq_export->export_obd;
  27         struct ldlm_reply *dlm_rep;
  28         struct ldlm_request *dlm_req;
  29         int rc, size = sizeof(*dlm_rep), cookielen = 0;
  30         __u32 flags;
  31         ldlm_error_t err;
  32         struct ldlm_lock *lock = NULL;
  33         void *cookie = NULL;
  34         ENTRY;
  35
  36         LDLM_DEBUG_NOLOCK("server-side enqueue handler START");
  37
  38         dlm_req = lustre_msg_buf(req->rq_reqmsg, 0);
  39         if (dlm_req->lock_desc.l_resource.lr_type == LDLM_MDSINTENT) {
  40                 /* In this case, the reply buffer is allocated deep in
  41                  * local_lock_enqueue by the policy function. */
  42                 cookie = req;
  43                 cookielen = sizeof(*req);
  44         } else {
  45                 rc = lustre_pack_msg(1, &size, NULL, &req->rq_replen,
  46                                      &req->rq_repmsg);
  47                 if (rc) {
  48                         CERROR("out of memory\n");
  49                         RETURN(-ENOMEM);
  50                 }
  51                 if (dlm_req->lock_desc.l_resource.lr_type == LDLM_EXTENT) {
  52                         cookie = &dlm_req->lock_desc.l_extent;
  53                         cookielen = sizeof(struct ldlm_extent);
  54                 }
  55         }
  56
  57         lock = ldlm_lock_create(obddev->obd_namespace,
  58                                 &dlm_req->lock_handle2,
  59                                 dlm_req->lock_desc.l_resource.lr_name,
  60                                 dlm_req->lock_desc.l_resource.lr_type,
  61                                 dlm_req->lock_desc.l_req_mode, NULL, 0);
  62         if (!lock)
  63                 GOTO(out, err = -ENOMEM);
  64
  65         memcpy(&lock->l_remote_handle, &dlm_req->lock_handle1,
  66                sizeof(lock->l_remote_handle));
  67         LDLM_DEBUG(lock, "server-side enqueue handler, new lock created");
  68
  69         flags = dlm_req->lock_flags;
  70         err = ldlm_lock_enqueue(lock, cookie, cookielen, &flags,
  71                                 ldlm_server_ast, ldlm_server_ast);
  72         if (err != ELDLM_OK)
  73                 GOTO(out, err);
  74
  75         dlm_rep = lustre_msg_buf(req->rq_repmsg, 0);
  76         dlm_rep->lock_flags = flags;
  77
  78         ldlm_lock2handle(lock, &dlm_rep->lock_handle);
  79         if (dlm_req->lock_desc.l_resource.lr_type == LDLM_EXTENT)
  80                 memcpy(&dlm_rep->lock_extent, &lock->l_extent,
  81                        sizeof(lock->l_extent));
  82         if (dlm_rep->lock_flags & LDLM_FL_LOCK_CHANGED)
  83                 memcpy(dlm_rep->lock_resource_name, lock->l_resource->lr_name,
  84                        sizeof(dlm_rep->lock_resource_name));
  85
  86         lock->l_connection = ptlrpc_connection_addref(req->rq_connection);
  87         EXIT;
  88  out:
  89         if (lock) {
  90                 LDLM_DEBUG(lock, "server-side enqueue handler, sending reply");
  91                 ldlm_lock_put(lock);
  92         }
  93         req->rq_status = err;
  94         CDEBUG(D_INFO, "err = %d\n", err);
  95
  96         if (ptlrpc_reply(req->rq_svc, req))
  97                 LBUG();
  98
  99         if (!err)
 100                 ldlm_reprocess_all(lock->l_resource);
 101         LDLM_DEBUG_NOLOCK("server-side enqueue handler END");
 102
 103         return 0;
 104 }
 105
 106 static int ldlm_handle_convert(struct ptlrpc_request *req)
 107 {
 108         struct ldlm_request *dlm_req;
 109         struct ldlm_reply *dlm_rep;
 110         struct ldlm_lock *lock;
 111         int rc, size = sizeof(*dlm_rep);
 112         ENTRY;
 113
 114         rc = lustre_pack_msg(1, &size, NULL, &req->rq_replen, &req->rq_repmsg);
 115         if (rc) {
 116                 CERROR("out of memory\n");
 117                 RETURN(-ENOMEM);
 118         }
 119         dlm_req = lustre_msg_buf(req->rq_reqmsg, 0);
 120         dlm_rep = lustre_msg_buf(req->rq_repmsg, 0);
 121         dlm_rep->lock_flags = dlm_req->lock_flags;
 122
 123         lock = ldlm_handle2lock(&dlm_req->lock_handle1);
 124         if (!lock) {
 125                 req->rq_status = EINVAL;
 126         } else {
 127                 LDLM_DEBUG(lock, "server-side convert handler START");
 128                 ldlm_lock_convert(lock, dlm_req->lock_desc.l_req_mode,
 129                                   &dlm_rep->lock_flags);
 130                 req->rq_status = 0;
 131         }
 132         if (ptlrpc_reply(req->rq_svc, req) != 0)
 133                 LBUG();
 134
 135         ldlm_reprocess_all(lock->l_resource);
 136         ldlm_lock_put(lock);
 137         LDLM_DEBUG(lock, "server-side convert handler END");
 138
 139         RETURN(0);
 140 }
 141
 142 static int ldlm_handle_cancel(struct ptlrpc_request *req)
 143 {
 144         struct ldlm_request *dlm_req;
 145         struct ldlm_lock *lock;
 146         int rc;
 147         ENTRY;
 148
 149         rc = lustre_pack_msg(0, NULL, NULL, &req->rq_replen, &req->rq_repmsg);
 150         if (rc) {
 151                 CERROR("out of memory\n");
 152                 RETURN(-ENOMEM);
 153         }
 154         dlm_req = lustre_msg_buf(req->rq_reqmsg, 0);
 155
 156         lock = ldlm_handle2lock(&dlm_req->lock_handle1);
 157         if (!lock) {
 158                 req->rq_status = ESTALE;
 159         } else {
 160                 LDLM_DEBUG(lock, "server-side cancel handler START");
 161                 ldlm_lock_cancel(lock);
 162                 req->rq_status = 0;
 163         }
 164
 165         if (ptlrpc_reply(req->rq_svc, req) != 0)
 166                 LBUG();
 167
 168         ldlm_reprocess_all(lock->l_resource);
 169         ldlm_lock_put(lock);
 170         LDLM_DEBUG_NOLOCK("server-side cancel handler END");
 171
 172         RETURN(0);
 173 }
 174
 175 static int ldlm_handle_callback(struct ptlrpc_request *req)
 176 {
 177         struct ldlm_request *dlm_req;
 178         struct ldlm_lock_desc *descp = NULL;
 179         struct ldlm_lock *lock;
 180         __u64 is_blocking_ast = 0;
 181         int rc;
 182         ENTRY;
 183
 184         rc = lustre_pack_msg(0, NULL, NULL, &req->rq_replen, &req->rq_repmsg);
 185         if (rc) {
 186                 CERROR("out of memory\n");
 187                 RETURN(-ENOMEM);
 188         }
 189         dlm_req = lustre_msg_buf(req->rq_reqmsg, 0);
 190
 191         /* We must send the reply first, so that the thread is free to handle
 192          * any requests made in common_callback() */
 193         rc = ptlrpc_reply(req->rq_svc, req);
 194         if (rc != 0)
 195                 RETURN(rc);
 196
 197         lock = ldlm_handle2lock(&dlm_req->lock_handle1);
 198         if (!lock) {
 199                 CERROR("callback on lock %Lx - lock disappeared\n",
 200                        dlm_req->lock_handle1.addr);
 201                 RETURN(0);
 202         }
 203
 204         /* check if this is a blocking AST */
 205         if (dlm_req->lock_desc.l_req_mode !=
 206             dlm_req->lock_desc.l_granted_mode) {
 207                 descp = &dlm_req->lock_desc;
 208                 is_blocking_ast = 1;
 209         }
 210
 211         LDLM_DEBUG(lock, "client %s callback handler START",
 212                    is_blocking_ast ? "blocked" : "completion");
 213
 214         if (descp) {
 215                 int do_ast;
 216                 l_lock(&lock->l_resource->lr_namespace->ns_lock);
 217                 lock->l_flags |= LDLM_FL_CBPENDING;
 218                 do_ast = (!lock->l_readers && !lock->l_writers);
 219                 l_unlock(&lock->l_resource->lr_namespace->ns_lock);
 220
 221                 if (do_ast) {
 222                         LDLM_DEBUG(lock, "already unused, calling "
 223                                    "callback (%p)", lock->l_blocking_ast);
 224                         if (lock->l_blocking_ast != NULL) {
 225                                 struct lustre_handle lockh;
 226                                 ldlm_lock2handle(lock, &lockh);
 227                                 lock->l_blocking_ast(&lockh, descp,
 228                                                      lock->l_data,
 229                                                      lock->l_data_len);
 230                         }
 231                 } else {
 232                         LDLM_DEBUG(lock, "Lock still has references, will be"
 233                                " cancelled later");
 234                 }
 235                 ldlm_lock_put(lock);
 236         } else {
 237                 struct list_head rpc_list = LIST_HEAD_INIT(rpc_list);
 238
 239                 l_lock(&lock->l_resource->lr_namespace->ns_lock);
 240                 lock->l_req_mode = dlm_req->lock_desc.l_granted_mode;
 241
 242                 /* If we receive the completion AST before the actual enqueue
 243                  * returned, then we might need to switch resources. */
 244                 if (memcmp(dlm_req->lock_desc.l_resource.lr_name,
 245                            lock->l_resource->lr_name,
 246                            sizeof(__u64) * RES_NAME_SIZE) != 0) {
 247                         ldlm_lock_change_resource(lock, dlm_req->lock_desc.l_resource.lr_name);
 248                         LDLM_DEBUG(lock, "completion AST, new resource");
 249                 }
 250                 lock->l_resource->lr_tmp = &rpc_list;
 251                 ldlm_resource_unlink_lock(lock);
 252                 ldlm_grant_lock(lock);
 253                 /*  FIXME: we want any completion function, not just wake_up */
 254                 wake_up(&lock->l_waitq);
 255                 lock->l_resource->lr_tmp = NULL;
 256                 l_unlock(&lock->l_resource->lr_namespace->ns_lock);
 257                 ldlm_lock_put(lock);
 258
 259                 ldlm_run_ast_work(&rpc_list);
 260         }
 261
 262         LDLM_DEBUG_NOLOCK("client %s callback handler END (lock: %p)",
 263                           is_blocking_ast ? "blocked" : "completion", lock);
 264         RETURN(0);
 265 }
 266
 267 static int ldlm_handle(struct ptlrpc_request *req)
 268 {
 269         int rc;
 270         ENTRY;
 271
 272         rc = lustre_unpack_msg(req->rq_reqmsg, req->rq_reqlen);
 273         if (rc) {
 274                 CERROR("lustre_ldlm: Invalid request\n");
 275                 GOTO(out, rc);
 276         }
 277
 278         if (req->rq_reqmsg->type != PTL_RPC_MSG_REQUEST) {
 279                 CERROR("lustre_ldlm: wrong packet type sent %d\n",
 280                        req->rq_reqmsg->type);
 281                 GOTO(out, rc = -EINVAL);
 282         }
 283
 284         if (!req->rq_export &&
 285             req->rq_reqmsg->opc == LDLM_ENQUEUE) {
 286                 CERROR("No export handle for enqueue request.\n");
 287                 GOTO(out, rc = -ENOTCONN);
 288         }
 289         switch (req->rq_reqmsg->opc) {
 290         case LDLM_ENQUEUE:
 291                 CDEBUG(D_INODE, "enqueue\n");
 292                 OBD_FAIL_RETURN(OBD_FAIL_LDLM_ENQUEUE, 0);
 293                 rc = ldlm_handle_enqueue(req);
 294                 break;
 295
 296         case LDLM_CONVERT:
 297                 CDEBUG(D_INODE, "convert\n");
 298                 OBD_FAIL_RETURN(OBD_FAIL_LDLM_CONVERT, 0);
 299                 rc = ldlm_handle_convert(req);
 300                 break;
 301
 302         case LDLM_CANCEL:
 303                 CDEBUG(D_INODE, "cancel\n");
 304                 OBD_FAIL_RETURN(OBD_FAIL_LDLM_CANCEL, 0);
 305                 rc = ldlm_handle_cancel(req);
 306                 break;
 307
 308         case LDLM_CALLBACK:
 309                 CDEBUG(D_INODE, "callback\n");
 310                 OBD_FAIL_RETURN(OBD_FAIL_LDLM_CALLBACK, 0);
 311                 rc = ldlm_handle_callback(req);
 312                 break;
 313
 314         default:
 315                 rc = ptlrpc_error(req->rq_svc, req);
 316                 RETURN(rc);
 317         }
 318
 319         EXIT;
 320 out:
 321         if (rc)
 322                 RETURN(ptlrpc_error(req->rq_svc, req));
 323         return 0;
 324 }
 325
 326 static int ldlm_iocontrol(long cmd, struct lustre_handle *conn, int len, void *karg,
 327                           void *uarg)
 328 {
 329         struct obd_device *obddev = class_conn2obd(conn);
 330         struct ptlrpc_connection *connection;
 331         int err;
 332         ENTRY;
 333
 334         if (_IOC_TYPE(cmd) != IOC_LDLM_TYPE || _IOC_NR(cmd) < IOC_LDLM_MIN_NR ||
 335             _IOC_NR(cmd) > IOC_LDLM_MAX_NR) {
 336                 CDEBUG(D_IOCTL, "invalid ioctl (type %ld, nr %ld, size %ld)\n",
 337                        _IOC_TYPE(cmd), _IOC_NR(cmd), _IOC_SIZE(cmd));
 338                 RETURN(-EINVAL);
 339         }
 340
 341         OBD_ALLOC(obddev->u.ldlm.ldlm_client,
 342                   sizeof(*obddev->u.ldlm.ldlm_client));
 343         ptlrpc_init_client(NULL, NULL,
 344                            LDLM_REQUEST_PORTAL, LDLM_REPLY_PORTAL,
 345                            obddev->u.ldlm.ldlm_client);
 346         connection = ptlrpc_uuid_to_connection("ldlm");
 347         if (!connection)
 348                 CERROR("No LDLM UUID found: assuming ldlm is local.\n");
 349
 350         switch (cmd) {
 351         case IOC_LDLM_TEST: {
 352                 err = ldlm_test(obddev, connection);
 353                 CERROR("-- done err %d\n", err);
 354                 GOTO(out, err);
 355         }
 356         default:
 357                 GOTO(out, err = -EINVAL);
 358         }
 359
 360  out:
 361         if (connection)
 362                 ptlrpc_put_connection(connection);
 363         OBD_FREE(obddev->u.ldlm.ldlm_client,
 364                  sizeof(*obddev->u.ldlm.ldlm_client));
 365         return err;
 366 }
 367
 368 #define LDLM_NUM_THREADS        8
 369
 370 static int ldlm_setup(struct obd_device *obddev, obd_count len, void *buf)
 371 {
 372         struct ldlm_obd *ldlm = &obddev->u.ldlm;
 373         int rc;
 374         int i;
 375         ENTRY;
 376
 377         MOD_INC_USE_COUNT;
 378         ldlm->ldlm_service =
 379                 ptlrpc_init_svc(64 * 1024, LDLM_REQUEST_PORTAL,
 380                                 LDLM_REPLY_PORTAL, "self", ldlm_handle);
 381         if (!ldlm->ldlm_service) {
 382                 LBUG();
 383                 GOTO(out_dec, rc = -ENOMEM);
 384         }
 385
 386         for (i = 0; i < LDLM_NUM_THREADS; i++) {
 387                 rc = ptlrpc_start_thread(obddev, ldlm->ldlm_service,
 388                                          "lustre_dlm");
 389                 if (rc) {
 390                         CERROR("cannot start LDLM thread #%d: rc %d\n", i, rc);
 391                         LBUG();
 392                         GOTO(out_thread, rc);
 393                 }
 394         }
 395
 396         RETURN(0);
 397
 398 out_thread:
 399         ptlrpc_stop_all_threads(ldlm->ldlm_service);
 400         ptlrpc_unregister_service(ldlm->ldlm_service);
 401 out_dec:
 402         MOD_DEC_USE_COUNT;
 403         return rc;
 404 }
 405
 406 static int ldlm_cleanup(struct obd_device *obddev)
 407 {
 408         struct ldlm_obd *ldlm = &obddev->u.ldlm;
 409         ENTRY;
 410
 411         ptlrpc_stop_all_threads(ldlm->ldlm_service);
 412         ptlrpc_unregister_service(ldlm->ldlm_service);
 413
 414         if (mds_reint_p != NULL)
 415                 inter_module_put("mds_reint");
 416         if (mds_getattr_name_p != NULL)
 417                 inter_module_put("mds_getattr_name");
 418
 419         MOD_DEC_USE_COUNT;
 420         RETURN(0);
 421 }
 422
 423 struct obd_ops ldlm_obd_ops = {
 424         o_iocontrol:   ldlm_iocontrol,
 425         o_setup:       ldlm_setup,
 426         o_cleanup:     ldlm_cleanup,
 427         o_connect:     class_connect,
 428         o_disconnect:  class_disconnect
 429 };
 430
 431
 432 static int __init ldlm_init(void)
 433 {
 434         int rc = class_register_type(&ldlm_obd_ops, OBD_LDLM_DEVICENAME);
 435         if (rc != 0)
 436                 return rc;
 437
 438         ldlm_resource_slab = kmem_cache_create("ldlm_resources",
 439                                                sizeof(struct ldlm_resource), 0,
 440                                                SLAB_HWCACHE_ALIGN, NULL, NULL);
 441         if (ldlm_resource_slab == NULL)
 442                 return -ENOMEM;
 443
 444         ldlm_lock_slab = kmem_cache_create("ldlm_locks",
 445                                            sizeof(struct ldlm_lock), 0,
 446                                            SLAB_HWCACHE_ALIGN, NULL, NULL);
 447         if (ldlm_lock_slab == NULL) {
 448                 kmem_cache_destroy(ldlm_resource_slab);
 449                 return -ENOMEM;
 450         }
 451
 452         return 0;
 453 }
 454
 455 static void __exit ldlm_exit(void)
 456 {
 457         class_unregister_type(OBD_LDLM_DEVICENAME);
 458         if (kmem_cache_destroy(ldlm_resource_slab) != 0)
 459                 CERROR("couldn't free ldlm resource slab\n");
 460         if (kmem_cache_destroy(ldlm_lock_slab) != 0)
 461                 CERROR("couldn't free ldlm lock slab\n");
 462 }
 463
 464 EXPORT_SYMBOL(ldlm_lockname);
 465 EXPORT_SYMBOL(ldlm_typename);
 466 EXPORT_SYMBOL(ldlm_handle2lock);
 467 EXPORT_SYMBOL(ldlm_lock_match);
 468 EXPORT_SYMBOL(ldlm_lock_addref);
 469 EXPORT_SYMBOL(ldlm_lock_decref);
 470 EXPORT_SYMBOL(ldlm_cli_convert);
 471 EXPORT_SYMBOL(ldlm_cli_enqueue);
 472 EXPORT_SYMBOL(ldlm_cli_cancel);
 473 EXPORT_SYMBOL(ldlm_test);
 474 EXPORT_SYMBOL(ldlm_lock_dump);
 475 EXPORT_SYMBOL(ldlm_namespace_new);
 476 EXPORT_SYMBOL(ldlm_namespace_free);
 477
 478 MODULE_AUTHOR("Cluster File Systems, Inc. <info@clusterfs.com>");
 479 MODULE_DESCRIPTION("Lustre Lock Management Module v0.1");
 480 MODULE_LICENSE("GPL");
 481
 482 module_init(ldlm_init);
 483 module_exit(ldlm_exit);