drivers/staging/lustre/lustre/ldlm/ldlm_lockd.c

   1 /*
   2  * GPL HEADER START
   3  *
   4  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   5  *
   6  * This program is free software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License version 2 only,
   8  * as published by the Free Software Foundation.
   9  *
  10  * This program is distributed in the hope that it will be useful, but
  11  * WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  13  * General Public License version 2 for more details (a copy is included
  14  * in the LICENSE file that accompanied this code).
  15  *
  16  * You should have received a copy of the GNU General Public License
  17  * version 2 along with this program; If not, see
  18  * http://www.gnu.org/licenses/gpl-2.0.html
  19  *
  20  * GPL HEADER END
  21  */
  22 /*
  23  * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved.
  24  * Use is subject to license terms.
  25  *
  26  * Copyright (c) 2010, 2015, Intel Corporation.
  27  */
  28 /*
  29  * This file is part of Lustre, http://www.lustre.org/
  30  * Lustre is a trademark of Sun Microsystems, Inc.
  31  *
  32  * lustre/ldlm/ldlm_lockd.c
  33  *
  34  * Author: Peter Braam <braam@clusterfs.com>
  35  * Author: Phil Schwan <phil@clusterfs.com>
  36  */
  37
  38 #define DEBUG_SUBSYSTEM S_LDLM
  39
  40 #include "../../include/linux/libcfs/libcfs.h"
  41 #include "../include/lustre_dlm.h"
  42 #include "../include/obd_class.h"
  43 #include <linux/list.h>
  44 #include "ldlm_internal.h"
  45
  46 static int ldlm_num_threads;
  47 module_param(ldlm_num_threads, int, 0444);
  48 MODULE_PARM_DESC(ldlm_num_threads, "number of DLM service threads to start");
  49
  50 static char *ldlm_cpts;
  51 module_param(ldlm_cpts, charp, 0444);
  52 MODULE_PARM_DESC(ldlm_cpts, "CPU partitions ldlm threads should run on");
  53
  54 static struct mutex     ldlm_ref_mutex;
  55 static int ldlm_refcount;
  56
  57 static struct kobject *ldlm_kobj;
  58 struct kset *ldlm_ns_kset;
  59 static struct kset *ldlm_svc_kset;
  60
  61 struct ldlm_cb_async_args {
  62         struct ldlm_cb_set_arg *ca_set_arg;
  63         struct ldlm_lock       *ca_lock;
  64 };
  65
  66 /* LDLM state */
  67
  68 static struct ldlm_state *ldlm_state;
  69
  70 #define ELT_STOPPED   0
  71 #define ELT_READY     1
  72 #define ELT_TERMINATE 2
  73
  74 struct ldlm_bl_pool {
  75         spinlock_t              blp_lock;
  76
  77         /*
  78          * blp_prio_list is used for callbacks that should be handled
  79          * as a priority. It is used for LDLM_FL_DISCARD_DATA requests.
  80          * see bug 13843
  81          */
  82         struct list_head              blp_prio_list;
  83
  84         /*
  85          * blp_list is used for all other callbacks which are likely
  86          * to take longer to process.
  87          */
  88         struct list_head              blp_list;
  89
  90         wait_queue_head_t            blp_waitq;
  91         struct completion       blp_comp;
  92         atomic_t            blp_num_threads;
  93         atomic_t            blp_busy_threads;
  94         int                  blp_min_threads;
  95         int                  blp_max_threads;
  96 };
  97
  98 struct ldlm_bl_work_item {
  99         struct list_head              blwi_entry;
 100         struct ldlm_namespace  *blwi_ns;
 101         struct ldlm_lock_desc   blwi_ld;
 102         struct ldlm_lock       *blwi_lock;
 103         struct list_head              blwi_head;
 104         int                  blwi_count;
 105         struct completion       blwi_comp;
 106         enum ldlm_cancel_flags  blwi_flags;
 107         int                  blwi_mem_pressure;
 108 };
 109
 110 /**
 111  * Callback handler for receiving incoming blocking ASTs.
 112  *
 113  * This can only happen on client side.
 114  */
 115 void ldlm_handle_bl_callback(struct ldlm_namespace *ns,
 116                              struct ldlm_lock_desc *ld, struct ldlm_lock *lock)
 117 {
 118         int do_ast;
 119
 120         LDLM_DEBUG(lock, "client blocking AST callback handler");
 121
 122         lock_res_and_lock(lock);
 123         ldlm_set_cbpending(lock);
 124
 125         if (ldlm_is_cancel_on_block(lock))
 126                 ldlm_set_cancel(lock);
 127
 128         do_ast = !lock->l_readers && !lock->l_writers;
 129         unlock_res_and_lock(lock);
 130
 131         if (do_ast) {
 132                 CDEBUG(D_DLMTRACE,
 133                        "Lock %p already unused, calling callback (%p)\n", lock,
 134                        lock->l_blocking_ast);
 135                 if (lock->l_blocking_ast)
 136                         lock->l_blocking_ast(lock, ld, lock->l_ast_data,
 137                                              LDLM_CB_BLOCKING);
 138         } else {
 139                 CDEBUG(D_DLMTRACE,
 140                        "Lock %p is referenced, will be cancelled later\n",
 141                        lock);
 142         }
 143
 144         LDLM_DEBUG(lock, "client blocking callback handler END");
 145         LDLM_LOCK_RELEASE(lock);
 146 }
 147
 148 /**
 149  * Callback handler for receiving incoming completion ASTs.
 150  *
 151  * This only can happen on client side.
 152  */
 153 static void ldlm_handle_cp_callback(struct ptlrpc_request *req,
 154                                     struct ldlm_namespace *ns,
 155                                     struct ldlm_request *dlm_req,
 156                                     struct ldlm_lock *lock)
 157 {
 158         int lvb_len;
 159         LIST_HEAD(ast_list);
 160         int rc = 0;
 161
 162         LDLM_DEBUG(lock, "client completion callback handler START");
 163
 164         if (OBD_FAIL_CHECK(OBD_FAIL_LDLM_CANCEL_BL_CB_RACE)) {
 165                 int to = cfs_time_seconds(1);
 166
 167                 while (to > 0) {
 168                         set_current_state(TASK_INTERRUPTIBLE);
 169                         schedule_timeout(to);
 170                         if (lock->l_granted_mode == lock->l_req_mode ||
 171                             ldlm_is_destroyed(lock))
 172                                 break;
 173                 }
 174         }
 175
 176         lvb_len = req_capsule_get_size(&req->rq_pill, &RMF_DLM_LVB, RCL_CLIENT);
 177         if (lvb_len < 0) {
 178                 LDLM_ERROR(lock, "Fail to get lvb_len, rc = %d", lvb_len);
 179                 rc = lvb_len;
 180                 goto out;
 181         } else if (lvb_len > 0) {
 182                 if (lock->l_lvb_len > 0) {
 183                         /* for extent lock, lvb contains ost_lvb{}. */
 184                         LASSERT(lock->l_lvb_data);
 185
 186                         if (unlikely(lock->l_lvb_len < lvb_len)) {
 187                                 LDLM_ERROR(lock, "Replied LVB is larger than expectation, expected = %d, replied = %d",
 188                                            lock->l_lvb_len, lvb_len);
 189                                 rc = -EINVAL;
 190                                 goto out;
 191                         }
 192                 } else if (ldlm_has_layout(lock)) { /* for layout lock, lvb has
 193                                                      * variable length
 194                                                      */
 195                         void *lvb_data;
 196
 197                         lvb_data = kzalloc(lvb_len, GFP_NOFS);
 198                         if (!lvb_data) {
 199                                 LDLM_ERROR(lock, "No memory: %d.\n", lvb_len);
 200                                 rc = -ENOMEM;
 201                                 goto out;
 202                         }
 203
 204                         lock_res_and_lock(lock);
 205                         LASSERT(!lock->l_lvb_data);
 206                         lock->l_lvb_type = LVB_T_LAYOUT;
 207                         lock->l_lvb_data = lvb_data;
 208                         lock->l_lvb_len = lvb_len;
 209                         unlock_res_and_lock(lock);
 210                 }
 211         }
 212
 213         lock_res_and_lock(lock);
 214         if (ldlm_is_destroyed(lock) ||
 215             lock->l_granted_mode == lock->l_req_mode) {
 216                 /* bug 11300: the lock has already been granted */
 217                 unlock_res_and_lock(lock);
 218                 LDLM_DEBUG(lock, "Double grant race happened");
 219                 rc = 0;
 220                 goto out;
 221         }
 222
 223         /* If we receive the completion AST before the actual enqueue returned,
 224          * then we might need to switch lock modes, resources, or extents.
 225          */
 226         if (dlm_req->lock_desc.l_granted_mode != lock->l_req_mode) {
 227                 lock->l_req_mode = dlm_req->lock_desc.l_granted_mode;
 228                 LDLM_DEBUG(lock, "completion AST, new lock mode");
 229         }
 230
 231         if (lock->l_resource->lr_type != LDLM_PLAIN) {
 232                 ldlm_convert_policy_to_local(req->rq_export,
 233                                           dlm_req->lock_desc.l_resource.lr_type,
 234                                           &dlm_req->lock_desc.l_policy_data,
 235                                           &lock->l_policy_data);
 236                 LDLM_DEBUG(lock, "completion AST, new policy data");
 237         }
 238
 239         ldlm_resource_unlink_lock(lock);
 240         if (memcmp(&dlm_req->lock_desc.l_resource.lr_name,
 241                    &lock->l_resource->lr_name,
 242                    sizeof(lock->l_resource->lr_name)) != 0) {
 243                 unlock_res_and_lock(lock);
 244                 rc = ldlm_lock_change_resource(ns, lock,
 245                                 &dlm_req->lock_desc.l_resource.lr_name);
 246                 if (rc < 0) {
 247                         LDLM_ERROR(lock, "Failed to allocate resource");
 248                         goto out;
 249                 }
 250                 LDLM_DEBUG(lock, "completion AST, new resource");
 251                 CERROR("change resource!\n");
 252                 lock_res_and_lock(lock);
 253         }
 254
 255         if (dlm_req->lock_flags & LDLM_FL_AST_SENT) {
 256                 /* BL_AST locks are not needed in LRU.
 257                  * Let ldlm_cancel_lru() be fast.
 258                  */
 259                 ldlm_lock_remove_from_lru(lock);
 260                 lock->l_flags |= LDLM_FL_CBPENDING | LDLM_FL_BL_AST;
 261                 LDLM_DEBUG(lock, "completion AST includes blocking AST");
 262         }
 263
 264         if (lock->l_lvb_len > 0) {
 265                 rc = ldlm_fill_lvb(lock, &req->rq_pill, RCL_CLIENT,
 266                                    lock->l_lvb_data, lvb_len);
 267                 if (rc < 0) {
 268                         unlock_res_and_lock(lock);
 269                         goto out;
 270                 }
 271         }
 272
 273         ldlm_grant_lock(lock, &ast_list);
 274         unlock_res_and_lock(lock);
 275
 276         LDLM_DEBUG(lock, "callback handler finished, about to run_ast_work");
 277
 278         /* Let Enqueue to call osc_lock_upcall() and initialize l_ast_data */
 279         OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_CP_ENQ_RACE, 2);
 280
 281         ldlm_run_ast_work(ns, &ast_list, LDLM_WORK_CP_AST);
 282
 283         LDLM_DEBUG_NOLOCK("client completion callback handler END (lock %p)",
 284                           lock);
 285         goto out;
 286
 287 out:
 288         if (rc < 0) {
 289                 lock_res_and_lock(lock);
 290                 ldlm_set_failed(lock);
 291                 unlock_res_and_lock(lock);
 292                 wake_up(&lock->l_waitq);
 293         }
 294         LDLM_LOCK_RELEASE(lock);
 295 }
 296
 297 /**
 298  * Callback handler for receiving incoming glimpse ASTs.
 299  *
 300  * This only can happen on client side.  After handling the glimpse AST
 301  * we also consider dropping the lock here if it is unused locally for a
 302  * long time.
 303  */
 304 static void ldlm_handle_gl_callback(struct ptlrpc_request *req,
 305                                     struct ldlm_namespace *ns,
 306                                     struct ldlm_request *dlm_req,
 307                                     struct ldlm_lock *lock)
 308 {
 309         int rc = -ENOSYS;
 310
 311         LDLM_DEBUG(lock, "client glimpse AST callback handler");
 312
 313         if (lock->l_glimpse_ast)
 314                 rc = lock->l_glimpse_ast(lock, req);
 315
 316         if (req->rq_repmsg) {
 317                 ptlrpc_reply(req);
 318         } else {
 319                 req->rq_status = rc;
 320                 ptlrpc_error(req);
 321         }
 322
 323         lock_res_and_lock(lock);
 324         if (lock->l_granted_mode == LCK_PW &&
 325             !lock->l_readers && !lock->l_writers &&
 326             cfs_time_after(cfs_time_current(),
 327                            cfs_time_add(lock->l_last_used,
 328                                         cfs_time_seconds(10)))) {
 329                 unlock_res_and_lock(lock);
 330                 if (ldlm_bl_to_thread_lock(ns, NULL, lock))
 331                         ldlm_handle_bl_callback(ns, NULL, lock);
 332
 333                 return;
 334         }
 335         unlock_res_and_lock(lock);
 336         LDLM_LOCK_RELEASE(lock);
 337 }
 338
 339 static int ldlm_callback_reply(struct ptlrpc_request *req, int rc)
 340 {
 341         if (req->rq_no_reply)
 342                 return 0;
 343
 344         req->rq_status = rc;
 345         if (!req->rq_packed_final) {
 346                 rc = lustre_pack_reply(req, 1, NULL, NULL);
 347                 if (rc)
 348                         return rc;
 349         }
 350         return ptlrpc_reply(req);
 351 }
 352
 353 static int __ldlm_bl_to_thread(struct ldlm_bl_work_item *blwi,
 354                                enum ldlm_cancel_flags cancel_flags)
 355 {
 356         struct ldlm_bl_pool *blp = ldlm_state->ldlm_bl_pool;
 357
 358         spin_lock(&blp->blp_lock);
 359         if (blwi->blwi_lock && ldlm_is_discard_data(blwi->blwi_lock)) {
 360                 /* add LDLM_FL_DISCARD_DATA requests to the priority list */
 361                 list_add_tail(&blwi->blwi_entry, &blp->blp_prio_list);
 362         } else {
 363                 /* other blocking callbacks are added to the regular list */
 364                 list_add_tail(&blwi->blwi_entry, &blp->blp_list);
 365         }
 366         spin_unlock(&blp->blp_lock);
 367
 368         wake_up(&blp->blp_waitq);
 369
 370         /* can not check blwi->blwi_flags as blwi could be already freed in
 371          * LCF_ASYNC mode
 372          */
 373         if (!(cancel_flags & LCF_ASYNC))
 374                 wait_for_completion(&blwi->blwi_comp);
 375
 376         return 0;
 377 }
 378
 379 static inline void init_blwi(struct ldlm_bl_work_item *blwi,
 380                              struct ldlm_namespace *ns,
 381                              struct ldlm_lock_desc *ld,
 382                              struct list_head *cancels, int count,
 383                              struct ldlm_lock *lock,
 384                              enum ldlm_cancel_flags cancel_flags)
 385 {
 386         init_completion(&blwi->blwi_comp);
 387         INIT_LIST_HEAD(&blwi->blwi_head);
 388
 389         if (memory_pressure_get())
 390                 blwi->blwi_mem_pressure = 1;
 391
 392         blwi->blwi_ns = ns;
 393         blwi->blwi_flags = cancel_flags;
 394         if (ld)
 395                 blwi->blwi_ld = *ld;
 396         if (count) {
 397                 list_add(&blwi->blwi_head, cancels);
 398                 list_del_init(cancels);
 399                 blwi->blwi_count = count;
 400         } else {
 401                 blwi->blwi_lock = lock;
 402         }
 403 }
 404
 405 /**
 406  * Queues a list of locks \a cancels containing \a count locks
 407  * for later processing by a blocking thread.  If \a count is zero,
 408  * then the lock referenced as \a lock is queued instead.
 409  *
 410  * The blocking thread would then call ->l_blocking_ast callback in the lock.
 411  * If list addition fails an error is returned and caller is supposed to
 412  * call ->l_blocking_ast itself.
 413  */
 414 static int ldlm_bl_to_thread(struct ldlm_namespace *ns,
 415                              struct ldlm_lock_desc *ld,
 416                              struct ldlm_lock *lock,
 417                              struct list_head *cancels, int count,
 418                              enum ldlm_cancel_flags cancel_flags)
 419 {
 420         if (cancels && count == 0)
 421                 return 0;
 422
 423         if (cancel_flags & LCF_ASYNC) {
 424                 struct ldlm_bl_work_item *blwi;
 425
 426                 blwi = kzalloc(sizeof(*blwi), GFP_NOFS);
 427                 if (!blwi)
 428                         return -ENOMEM;
 429                 init_blwi(blwi, ns, ld, cancels, count, lock, cancel_flags);
 430
 431                 return __ldlm_bl_to_thread(blwi, cancel_flags);
 432         } else {
 433                 /* if it is synchronous call do minimum mem alloc, as it could
 434                  * be triggered from kernel shrinker
 435                  */
 436                 struct ldlm_bl_work_item blwi;
 437
 438                 memset(&blwi, 0, sizeof(blwi));
 439                 init_blwi(&blwi, ns, ld, cancels, count, lock, cancel_flags);
 440                 return __ldlm_bl_to_thread(&blwi, cancel_flags);
 441         }
 442 }
 443
 444 int ldlm_bl_to_thread_lock(struct ldlm_namespace *ns, struct ldlm_lock_desc *ld,
 445                            struct ldlm_lock *lock)
 446 {
 447         return ldlm_bl_to_thread(ns, ld, lock, NULL, 0, LCF_ASYNC);
 448 }
 449
 450 int ldlm_bl_to_thread_list(struct ldlm_namespace *ns, struct ldlm_lock_desc *ld,
 451                            struct list_head *cancels, int count,
 452                            enum ldlm_cancel_flags cancel_flags)
 453 {
 454         return ldlm_bl_to_thread(ns, ld, NULL, cancels, count, cancel_flags);
 455 }
 456
 457 /* Setinfo coming from Server (eg MDT) to Client (eg MDC)! */
 458 static int ldlm_handle_setinfo(struct ptlrpc_request *req)
 459 {
 460         struct obd_device *obd = req->rq_export->exp_obd;
 461         char *key;
 462         void *val;
 463         int keylen, vallen;
 464         int rc = -ENOSYS;
 465
 466         DEBUG_REQ(D_HSM, req, "%s: handle setinfo\n", obd->obd_name);
 467
 468         req_capsule_set(&req->rq_pill, &RQF_OBD_SET_INFO);
 469
 470         key = req_capsule_client_get(&req->rq_pill, &RMF_SETINFO_KEY);
 471         if (!key) {
 472                 DEBUG_REQ(D_IOCTL, req, "no set_info key");
 473                 return -EFAULT;
 474         }
 475         keylen = req_capsule_get_size(&req->rq_pill, &RMF_SETINFO_KEY,
 476                                       RCL_CLIENT);
 477         val = req_capsule_client_get(&req->rq_pill, &RMF_SETINFO_VAL);
 478         if (!val) {
 479                 DEBUG_REQ(D_IOCTL, req, "no set_info val");
 480                 return -EFAULT;
 481         }
 482         vallen = req_capsule_get_size(&req->rq_pill, &RMF_SETINFO_VAL,
 483                                       RCL_CLIENT);
 484
 485         /* We are responsible for swabbing contents of val */
 486
 487         if (KEY_IS(KEY_HSM_COPYTOOL_SEND))
 488                 /* Pass it on to mdc (the "export" in this case) */
 489                 rc = obd_set_info_async(req->rq_svc_thread->t_env,
 490                                         req->rq_export,
 491                                         sizeof(KEY_HSM_COPYTOOL_SEND),
 492                                         KEY_HSM_COPYTOOL_SEND,
 493                                         vallen, val, NULL);
 494         else
 495                 DEBUG_REQ(D_WARNING, req, "ignoring unknown key %s", key);
 496
 497         return rc;
 498 }
 499
 500 static inline void ldlm_callback_errmsg(struct ptlrpc_request *req,
 501                                         const char *msg, int rc,
 502                                         const struct lustre_handle *handle)
 503 {
 504         DEBUG_REQ((req->rq_no_reply || rc) ? D_WARNING : D_DLMTRACE, req,
 505                   "%s: [nid %s] [rc %d] [lock %#llx]",
 506                   msg, libcfs_id2str(req->rq_peer), rc,
 507                   handle ? handle->cookie : 0);
 508         if (req->rq_no_reply)
 509                 CWARN("No reply was sent, maybe cause bug 21636.\n");
 510         else if (rc)
 511                 CWARN("Send reply failed, maybe cause bug 21636.\n");
 512 }
 513
 514 /* TODO: handle requests in a similar way as MDT: see mdt_handle_common() */
 515 static int ldlm_callback_handler(struct ptlrpc_request *req)
 516 {
 517         struct ldlm_namespace *ns;
 518         struct ldlm_request *dlm_req;
 519         struct ldlm_lock *lock;
 520         int rc;
 521
 522         /* Requests arrive in sender's byte order.  The ptlrpc service
 523          * handler has already checked and, if necessary, byte-swapped the
 524          * incoming request message body, but I am responsible for the
 525          * message buffers.
 526          */
 527
 528         /* do nothing for sec context finalize */
 529         if (lustre_msg_get_opc(req->rq_reqmsg) == SEC_CTX_FINI)
 530                 return 0;
 531
 532         req_capsule_init(&req->rq_pill, req, RCL_SERVER);
 533
 534         if (!req->rq_export) {
 535                 rc = ldlm_callback_reply(req, -ENOTCONN);
 536                 ldlm_callback_errmsg(req, "Operate on unconnected server",
 537                                      rc, NULL);
 538                 return 0;
 539         }
 540
 541         LASSERT(req->rq_export->exp_obd);
 542
 543         switch (lustre_msg_get_opc(req->rq_reqmsg)) {
 544         case LDLM_BL_CALLBACK:
 545                 if (OBD_FAIL_CHECK(OBD_FAIL_LDLM_BL_CALLBACK_NET)) {
 546                         if (cfs_fail_err)
 547                                 ldlm_callback_reply(req, -(int)cfs_fail_err);
 548                         return 0;
 549                 }
 550                 break;
 551         case LDLM_CP_CALLBACK:
 552                 if (OBD_FAIL_CHECK(OBD_FAIL_LDLM_CP_CALLBACK_NET))
 553                         return 0;
 554                 break;
 555         case LDLM_GL_CALLBACK:
 556                 if (OBD_FAIL_CHECK(OBD_FAIL_LDLM_GL_CALLBACK_NET))
 557                         return 0;
 558                 break;
 559         case LDLM_SET_INFO:
 560                 rc = ldlm_handle_setinfo(req);
 561                 ldlm_callback_reply(req, rc);
 562                 return 0;
 563         default:
 564                 CERROR("unknown opcode %u\n",
 565                        lustre_msg_get_opc(req->rq_reqmsg));
 566                 ldlm_callback_reply(req, -EPROTO);
 567                 return 0;
 568         }
 569
 570         ns = req->rq_export->exp_obd->obd_namespace;
 571         LASSERT(ns);
 572
 573         req_capsule_set(&req->rq_pill, &RQF_LDLM_CALLBACK);
 574
 575         dlm_req = req_capsule_client_get(&req->rq_pill, &RMF_DLM_REQ);
 576         if (!dlm_req) {
 577                 rc = ldlm_callback_reply(req, -EPROTO);
 578                 ldlm_callback_errmsg(req, "Operate without parameter", rc,
 579                                      NULL);
 580                 return 0;
 581         }
 582
 583         /* Force a known safe race, send a cancel to the server for a lock
 584          * which the server has already started a blocking callback on.
 585          */
 586         if (OBD_FAIL_CHECK(OBD_FAIL_LDLM_CANCEL_BL_CB_RACE) &&
 587             lustre_msg_get_opc(req->rq_reqmsg) == LDLM_BL_CALLBACK) {
 588                 rc = ldlm_cli_cancel(&dlm_req->lock_handle[0], 0);
 589                 if (rc < 0)
 590                         CERROR("ldlm_cli_cancel: %d\n", rc);
 591         }
 592
 593         lock = ldlm_handle2lock_long(&dlm_req->lock_handle[0], 0);
 594         if (!lock) {
 595                 CDEBUG(D_DLMTRACE, "callback on lock %#llx - lock disappeared\n",
 596                        dlm_req->lock_handle[0].cookie);
 597                 rc = ldlm_callback_reply(req, -EINVAL);
 598                 ldlm_callback_errmsg(req, "Operate with invalid parameter", rc,
 599                                      &dlm_req->lock_handle[0]);
 600                 return 0;
 601         }
 602
 603         if (ldlm_is_fail_loc(lock) &&
 604             lustre_msg_get_opc(req->rq_reqmsg) == LDLM_BL_CALLBACK)
 605                 OBD_RACE(OBD_FAIL_LDLM_CP_BL_RACE);
 606
 607         /* Copy hints/flags (e.g. LDLM_FL_DISCARD_DATA) from AST. */
 608         lock_res_and_lock(lock);
 609         lock->l_flags |= ldlm_flags_from_wire(dlm_req->lock_flags &
 610                                               LDLM_FL_AST_MASK);
 611         if (lustre_msg_get_opc(req->rq_reqmsg) == LDLM_BL_CALLBACK) {
 612                 /* If somebody cancels lock and cache is already dropped,
 613                  * or lock is failed before cp_ast received on client,
 614                  * we can tell the server we have no lock. Otherwise, we
 615                  * should send cancel after dropping the cache.
 616                  */
 617                 if ((ldlm_is_canceling(lock) && ldlm_is_bl_done(lock)) ||
 618                     ldlm_is_failed(lock)) {
 619                         LDLM_DEBUG(lock,
 620                                    "callback on lock %#llx - lock disappeared",
 621                                    dlm_req->lock_handle[0].cookie);
 622                         unlock_res_and_lock(lock);
 623                         LDLM_LOCK_RELEASE(lock);
 624                         rc = ldlm_callback_reply(req, -EINVAL);
 625                         ldlm_callback_errmsg(req, "Operate on stale lock", rc,
 626                                              &dlm_req->lock_handle[0]);
 627                         return 0;
 628                 }
 629                 /* BL_AST locks are not needed in LRU.
 630                  * Let ldlm_cancel_lru() be fast.
 631                  */
 632                 ldlm_lock_remove_from_lru(lock);
 633                 ldlm_set_bl_ast(lock);
 634         }
 635         unlock_res_and_lock(lock);
 636
 637         /* We want the ost thread to get this reply so that it can respond
 638          * to ost requests (write cache writeback) that might be triggered
 639          * in the callback.
 640          *
 641          * But we'd also like to be able to indicate in the reply that we're
 642          * cancelling right now, because it's unused, or have an intent result
 643          * in the reply, so we might have to push the responsibility for sending
 644          * the reply down into the AST handlers, alas.
 645          */
 646
 647         switch (lustre_msg_get_opc(req->rq_reqmsg)) {
 648         case LDLM_BL_CALLBACK:
 649                 CDEBUG(D_INODE, "blocking ast\n");
 650                 req_capsule_extend(&req->rq_pill, &RQF_LDLM_BL_CALLBACK);
 651                 if (!ldlm_is_cancel_on_block(lock)) {
 652                         rc = ldlm_callback_reply(req, 0);
 653                         if (req->rq_no_reply || rc)
 654                                 ldlm_callback_errmsg(req, "Normal process", rc,
 655                                                      &dlm_req->lock_handle[0]);
 656                 }
 657                 if (ldlm_bl_to_thread_lock(ns, &dlm_req->lock_desc, lock))
 658                         ldlm_handle_bl_callback(ns, &dlm_req->lock_desc, lock);
 659                 break;
 660         case LDLM_CP_CALLBACK:
 661                 CDEBUG(D_INODE, "completion ast\n");
 662                 req_capsule_extend(&req->rq_pill, &RQF_LDLM_CP_CALLBACK);
 663                 ldlm_callback_reply(req, 0);
 664                 ldlm_handle_cp_callback(req, ns, dlm_req, lock);
 665                 break;
 666         case LDLM_GL_CALLBACK:
 667                 CDEBUG(D_INODE, "glimpse ast\n");
 668                 req_capsule_extend(&req->rq_pill, &RQF_LDLM_GL_CALLBACK);
 669                 ldlm_handle_gl_callback(req, ns, dlm_req, lock);
 670                 break;
 671         default:
 672                 LBUG();                  /* checked above */
 673         }
 674
 675         return 0;
 676 }
 677
 678 static struct ldlm_bl_work_item *ldlm_bl_get_work(struct ldlm_bl_pool *blp)
 679 {
 680         struct ldlm_bl_work_item *blwi = NULL;
 681         static unsigned int num_bl;
 682
 683         spin_lock(&blp->blp_lock);
 684         /* process a request from the blp_list at least every blp_num_threads */
 685         if (!list_empty(&blp->blp_list) &&
 686             (list_empty(&blp->blp_prio_list) || num_bl == 0))
 687                 blwi = list_entry(blp->blp_list.next,
 688                                   struct ldlm_bl_work_item, blwi_entry);
 689         else
 690                 if (!list_empty(&blp->blp_prio_list))
 691                         blwi = list_entry(blp->blp_prio_list.next,
 692                                           struct ldlm_bl_work_item,
 693                                           blwi_entry);
 694
 695         if (blwi) {
 696                 if (++num_bl >= atomic_read(&blp->blp_num_threads))
 697                         num_bl = 0;
 698                 list_del(&blwi->blwi_entry);
 699         }
 700         spin_unlock(&blp->blp_lock);
 701
 702         return blwi;
 703 }
 704
 705 /* This only contains temporary data until the thread starts */
 706 struct ldlm_bl_thread_data {
 707         char                    bltd_name[CFS_CURPROC_COMM_MAX];
 708         struct ldlm_bl_pool     *bltd_blp;
 709         struct completion       bltd_comp;
 710         int                     bltd_num;
 711 };
 712
 713 static int ldlm_bl_thread_main(void *arg);
 714
 715 static int ldlm_bl_thread_start(struct ldlm_bl_pool *blp)
 716 {
 717         struct ldlm_bl_thread_data bltd = { .bltd_blp = blp };
 718         struct task_struct *task;
 719
 720         init_completion(&bltd.bltd_comp);
 721         bltd.bltd_num = atomic_read(&blp->blp_num_threads);
 722         snprintf(bltd.bltd_name, sizeof(bltd.bltd_name),
 723                  "ldlm_bl_%02d", bltd.bltd_num);
 724         task = kthread_run(ldlm_bl_thread_main, &bltd, "%s", bltd.bltd_name);
 725         if (IS_ERR(task)) {
 726                 CERROR("cannot start LDLM thread ldlm_bl_%02d: rc %ld\n",
 727                        atomic_read(&blp->blp_num_threads), PTR_ERR(task));
 728                 return PTR_ERR(task);
 729         }
 730         wait_for_completion(&bltd.bltd_comp);
 731
 732         return 0;
 733 }
 734
 735 /**
 736  * Main blocking requests processing thread.
 737  *
 738  * Callers put locks into its queue by calling ldlm_bl_to_thread.
 739  * This thread in the end ends up doing actual call to ->l_blocking_ast
 740  * for queued locks.
 741  */
 742 static int ldlm_bl_thread_main(void *arg)
 743 {
 744         struct ldlm_bl_pool *blp;
 745
 746         {
 747                 struct ldlm_bl_thread_data *bltd = arg;
 748
 749                 blp = bltd->bltd_blp;
 750
 751                 atomic_inc(&blp->blp_num_threads);
 752                 atomic_inc(&blp->blp_busy_threads);
 753
 754                 complete(&bltd->bltd_comp);
 755                 /* cannot use bltd after this, it is only on caller's stack */
 756         }
 757
 758         while (1) {
 759                 struct l_wait_info lwi = { 0 };
 760                 struct ldlm_bl_work_item *blwi = NULL;
 761                 int busy;
 762
 763                 blwi = ldlm_bl_get_work(blp);
 764
 765                 if (!blwi) {
 766                         atomic_dec(&blp->blp_busy_threads);
 767                         l_wait_event_exclusive(blp->blp_waitq,
 768                                                (blwi = ldlm_bl_get_work(blp)),
 769                                                &lwi);
 770                         busy = atomic_inc_return(&blp->blp_busy_threads);
 771                 } else {
 772                         busy = atomic_read(&blp->blp_busy_threads);
 773                 }
 774
 775                 if (!blwi->blwi_ns)
 776                         /* added by ldlm_cleanup() */
 777                         break;
 778
 779                 /* Not fatal if racy and have a few too many threads */
 780                 if (unlikely(busy < blp->blp_max_threads &&
 781                              busy >= atomic_read(&blp->blp_num_threads) &&
 782                              !blwi->blwi_mem_pressure))
 783                         /* discard the return value, we tried */
 784                         ldlm_bl_thread_start(blp);
 785
 786                 if (blwi->blwi_mem_pressure)
 787                         memory_pressure_set();
 788
 789                 if (blwi->blwi_count) {
 790                         int count;
 791                         /* The special case when we cancel locks in LRU
 792                          * asynchronously, we pass the list of locks here.
 793                          * Thus locks are marked LDLM_FL_CANCELING, but NOT
 794                          * canceled locally yet.
 795                          */
 796                         count = ldlm_cli_cancel_list_local(&blwi->blwi_head,
 797                                                            blwi->blwi_count,
 798                                                            LCF_BL_AST);
 799                         ldlm_cli_cancel_list(&blwi->blwi_head, count, NULL,
 800                                              blwi->blwi_flags);
 801                 } else {
 802                         ldlm_handle_bl_callback(blwi->blwi_ns, &blwi->blwi_ld,
 803                                                 blwi->blwi_lock);
 804                 }
 805                 if (blwi->blwi_mem_pressure)
 806                         memory_pressure_clr();
 807
 808                 if (blwi->blwi_flags & LCF_ASYNC)
 809                         kfree(blwi);
 810                 else
 811                         complete(&blwi->blwi_comp);
 812         }
 813
 814         atomic_dec(&blp->blp_busy_threads);
 815         atomic_dec(&blp->blp_num_threads);
 816         complete(&blp->blp_comp);
 817         return 0;
 818 }
 819
 820 static int ldlm_setup(void);
 821 static int ldlm_cleanup(void);
 822
 823 int ldlm_get_ref(void)
 824 {
 825         int rc = 0;
 826
 827         mutex_lock(&ldlm_ref_mutex);
 828         if (++ldlm_refcount == 1) {
 829                 rc = ldlm_setup();
 830                 if (rc)
 831                         ldlm_refcount--;
 832         }
 833         mutex_unlock(&ldlm_ref_mutex);
 834
 835         return rc;
 836 }
 837
 838 void ldlm_put_ref(void)
 839 {
 840         mutex_lock(&ldlm_ref_mutex);
 841         if (ldlm_refcount == 1) {
 842                 int rc = ldlm_cleanup();
 843
 844                 if (rc)
 845                         CERROR("ldlm_cleanup failed: %d\n", rc);
 846                 else
 847                         ldlm_refcount--;
 848         } else {
 849                 ldlm_refcount--;
 850         }
 851         mutex_unlock(&ldlm_ref_mutex);
 852 }
 853
 854 static ssize_t cancel_unused_locks_before_replay_show(struct kobject *kobj,
 855                                                       struct attribute *attr,
 856                                                       char *buf)
 857 {
 858         return sprintf(buf, "%d\n", ldlm_cancel_unused_locks_before_replay);
 859 }
 860
 861 static ssize_t cancel_unused_locks_before_replay_store(struct kobject *kobj,
 862                                                        struct attribute *attr,
 863                                                        const char *buffer,
 864                                                        size_t count)
 865 {
 866         int rc;
 867         unsigned long val;
 868
 869         rc = kstrtoul(buffer, 10, &val);
 870         if (rc)
 871                 return rc;
 872
 873         ldlm_cancel_unused_locks_before_replay = val;
 874
 875         return count;
 876 }
 877 LUSTRE_RW_ATTR(cancel_unused_locks_before_replay);
 878
 879 /* These are for root of /sys/fs/lustre/ldlm */
 880 static struct attribute *ldlm_attrs[] = {
 881         &lustre_attr_cancel_unused_locks_before_replay.attr,
 882         NULL,
 883 };
 884
 885 static struct attribute_group ldlm_attr_group = {
 886         .attrs = ldlm_attrs,
 887 };
 888
 889 static int ldlm_setup(void)
 890 {
 891         static struct ptlrpc_service_conf       conf;
 892         struct ldlm_bl_pool                     *blp = NULL;
 893         int rc = 0;
 894         int i;
 895
 896         if (ldlm_state)
 897                 return -EALREADY;
 898
 899         ldlm_state = kzalloc(sizeof(*ldlm_state), GFP_NOFS);
 900         if (!ldlm_state)
 901                 return -ENOMEM;
 902
 903         ldlm_kobj = kobject_create_and_add("ldlm", lustre_kobj);
 904         if (!ldlm_kobj) {
 905                 rc = -ENOMEM;
 906                 goto out;
 907         }
 908
 909         rc = sysfs_create_group(ldlm_kobj, &ldlm_attr_group);
 910         if (rc)
 911                 goto out;
 912
 913         ldlm_ns_kset = kset_create_and_add("namespaces", NULL, ldlm_kobj);
 914         if (!ldlm_ns_kset) {
 915                 rc = -ENOMEM;
 916                 goto out;
 917         }
 918
 919         ldlm_svc_kset = kset_create_and_add("services", NULL, ldlm_kobj);
 920         if (!ldlm_svc_kset) {
 921                 rc = -ENOMEM;
 922                 goto out;
 923         }
 924
 925         rc = ldlm_debugfs_setup();
 926         if (rc != 0)
 927                 goto out;
 928
 929         memset(&conf, 0, sizeof(conf));
 930         conf = (typeof(conf)) {
 931                 .psc_name               = "ldlm_cbd",
 932                 .psc_watchdog_factor    = 2,
 933                 .psc_buf                = {
 934                         .bc_nbufs               = LDLM_CLIENT_NBUFS,
 935                         .bc_buf_size            = LDLM_BUFSIZE,
 936                         .bc_req_max_size        = LDLM_MAXREQSIZE,
 937                         .bc_rep_max_size        = LDLM_MAXREPSIZE,
 938                         .bc_req_portal          = LDLM_CB_REQUEST_PORTAL,
 939                         .bc_rep_portal          = LDLM_CB_REPLY_PORTAL,
 940                 },
 941                 .psc_thr                = {
 942                         .tc_thr_name            = "ldlm_cb",
 943                         .tc_thr_factor          = LDLM_THR_FACTOR,
 944                         .tc_nthrs_init          = LDLM_NTHRS_INIT,
 945                         .tc_nthrs_base          = LDLM_NTHRS_BASE,
 946                         .tc_nthrs_max           = LDLM_NTHRS_MAX,
 947                         .tc_nthrs_user          = ldlm_num_threads,
 948                         .tc_cpu_affinity        = 1,
 949                         .tc_ctx_tags            = LCT_MD_THREAD | LCT_DT_THREAD,
 950                 },
 951                 .psc_cpt                = {
 952                         .cc_pattern             = ldlm_cpts,
 953                 },
 954                 .psc_ops                = {
 955                         .so_req_handler         = ldlm_callback_handler,
 956                 },
 957         };
 958         ldlm_state->ldlm_cb_service =
 959                         ptlrpc_register_service(&conf, ldlm_svc_kset,
 960                                                 ldlm_svc_debugfs_dir);
 961         if (IS_ERR(ldlm_state->ldlm_cb_service)) {
 962                 CERROR("failed to start service\n");
 963                 rc = PTR_ERR(ldlm_state->ldlm_cb_service);
 964                 ldlm_state->ldlm_cb_service = NULL;
 965                 goto out;
 966         }
 967
 968         blp = kzalloc(sizeof(*blp), GFP_NOFS);
 969         if (!blp) {
 970                 rc = -ENOMEM;
 971                 goto out;
 972         }
 973         ldlm_state->ldlm_bl_pool = blp;
 974
 975         spin_lock_init(&blp->blp_lock);
 976         INIT_LIST_HEAD(&blp->blp_list);
 977         INIT_LIST_HEAD(&blp->blp_prio_list);
 978         init_waitqueue_head(&blp->blp_waitq);
 979         atomic_set(&blp->blp_num_threads, 0);
 980         atomic_set(&blp->blp_busy_threads, 0);
 981
 982         if (ldlm_num_threads == 0) {
 983                 blp->blp_min_threads = LDLM_NTHRS_INIT;
 984                 blp->blp_max_threads = LDLM_NTHRS_MAX;
 985         } else {
 986                 blp->blp_min_threads = min_t(int, LDLM_NTHRS_MAX,
 987                                              max_t(int, LDLM_NTHRS_INIT,
 988                                                    ldlm_num_threads));
 989
 990                 blp->blp_max_threads = blp->blp_min_threads;
 991         }
 992
 993         for (i = 0; i < blp->blp_min_threads; i++) {
 994                 rc = ldlm_bl_thread_start(blp);
 995                 if (rc < 0)
 996                         goto out;
 997         }
 998
 999         rc = ldlm_pools_init();
1000         if (rc) {
1001                 CERROR("Failed to initialize LDLM pools: %d\n", rc);
1002                 goto out;
1003         }
1004         return 0;
1005
1006  out:
1007         ldlm_cleanup();
1008         return rc;
1009 }
1010
1011 static int ldlm_cleanup(void)
1012 {
1013         if (!list_empty(ldlm_namespace_list(LDLM_NAMESPACE_SERVER)) ||
1014             !list_empty(ldlm_namespace_list(LDLM_NAMESPACE_CLIENT))) {
1015                 CERROR("ldlm still has namespaces; clean these up first.\n");
1016                 ldlm_dump_all_namespaces(LDLM_NAMESPACE_SERVER, D_DLMTRACE);
1017                 ldlm_dump_all_namespaces(LDLM_NAMESPACE_CLIENT, D_DLMTRACE);
1018                 return -EBUSY;
1019         }
1020
1021         ldlm_pools_fini();
1022
1023         if (ldlm_state->ldlm_bl_pool) {
1024                 struct ldlm_bl_pool *blp = ldlm_state->ldlm_bl_pool;
1025
1026                 while (atomic_read(&blp->blp_num_threads) > 0) {
1027                         struct ldlm_bl_work_item blwi = { .blwi_ns = NULL };
1028
1029                         init_completion(&blp->blp_comp);
1030
1031                         spin_lock(&blp->blp_lock);
1032                         list_add_tail(&blwi.blwi_entry, &blp->blp_list);
1033                         wake_up(&blp->blp_waitq);
1034                         spin_unlock(&blp->blp_lock);
1035
1036                         wait_for_completion(&blp->blp_comp);
1037                 }
1038
1039                 kfree(blp);
1040         }
1041
1042         if (ldlm_state->ldlm_cb_service)
1043                 ptlrpc_unregister_service(ldlm_state->ldlm_cb_service);
1044
1045         if (ldlm_ns_kset)
1046                 kset_unregister(ldlm_ns_kset);
1047         if (ldlm_svc_kset)
1048                 kset_unregister(ldlm_svc_kset);
1049         if (ldlm_kobj)
1050                 kobject_put(ldlm_kobj);
1051
1052         ldlm_debugfs_cleanup();
1053
1054         kfree(ldlm_state);
1055         ldlm_state = NULL;
1056
1057         return 0;
1058 }
1059
1060 int ldlm_init(void)
1061 {
1062         mutex_init(&ldlm_ref_mutex);
1063         mutex_init(ldlm_namespace_lock(LDLM_NAMESPACE_SERVER));
1064         mutex_init(ldlm_namespace_lock(LDLM_NAMESPACE_CLIENT));
1065         ldlm_resource_slab = kmem_cache_create("ldlm_resources",
1066                                                sizeof(struct ldlm_resource), 0,
1067                                                SLAB_HWCACHE_ALIGN, NULL);
1068         if (!ldlm_resource_slab)
1069                 return -ENOMEM;
1070
1071         ldlm_lock_slab = kmem_cache_create("ldlm_locks",
1072                                            sizeof(struct ldlm_lock), 0,
1073                                            SLAB_HWCACHE_ALIGN |
1074                                            SLAB_DESTROY_BY_RCU, NULL);
1075         if (!ldlm_lock_slab) {
1076                 kmem_cache_destroy(ldlm_resource_slab);
1077                 return -ENOMEM;
1078         }
1079
1080         ldlm_interval_slab = kmem_cache_create("interval_node",
1081                                                sizeof(struct ldlm_interval),
1082                                                0, SLAB_HWCACHE_ALIGN, NULL);
1083         if (!ldlm_interval_slab) {
1084                 kmem_cache_destroy(ldlm_resource_slab);
1085                 kmem_cache_destroy(ldlm_lock_slab);
1086                 return -ENOMEM;
1087         }
1088 #if LUSTRE_TRACKS_LOCK_EXP_REFS
1089         class_export_dump_hook = ldlm_dump_export_locks;
1090 #endif
1091         return 0;
1092 }
1093
1094 void ldlm_exit(void)
1095 {
1096         if (ldlm_refcount)
1097                 CERROR("ldlm_refcount is %d in ldlm_exit!\n", ldlm_refcount);
1098         kmem_cache_destroy(ldlm_resource_slab);
1099         /* ldlm_lock_put() use RCU to call ldlm_lock_free, so need call
1100          * synchronize_rcu() to wait a grace period elapsed, so that
1101          * ldlm_lock_free() get a chance to be called.
1102          */
1103         synchronize_rcu();
1104         kmem_cache_destroy(ldlm_lock_slab);
1105         kmem_cache_destroy(ldlm_interval_slab);
1106 }