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.sun.com/software/products/lustre/docs/GPLv2.pdf
  19  *
  20  * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
  21  * CA 95054 USA or visit www.sun.com if you need additional information or
  22  * have any questions.
  23  *
  24  * GPL HEADER END
  25  */
  26 /*
  27  * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved.
  28  * Use is subject to license terms.
  29  *
  30  * Copyright (c) 2010, 2015, Intel Corporation.
  31  */
  32 /*
  33  * This file is part of Lustre, http://www.lustre.org/
  34  * Lustre is a trademark of Sun Microsystems, Inc.
  35  *
  36  * lustre/ldlm/ldlm_lockd.c
  37  *
  38  * Author: Peter Braam <braam@clusterfs.com>
  39  * Author: Phil Schwan <phil@clusterfs.com>
  40  */
  41
  42 #define DEBUG_SUBSYSTEM S_LDLM
  43
  44 #include "../../include/linux/libcfs/libcfs.h"
  45 #include "../include/lustre_dlm.h"
  46 #include "../include/obd_class.h"
  47 #include <linux/list.h>
  48 #include "ldlm_internal.h"
  49
  50 static int ldlm_num_threads;
  51 module_param(ldlm_num_threads, int, 0444);
  52 MODULE_PARM_DESC(ldlm_num_threads, "number of DLM service threads to start");
  53
  54 static char *ldlm_cpts;
  55 module_param(ldlm_cpts, charp, 0444);
  56 MODULE_PARM_DESC(ldlm_cpts, "CPU partitions ldlm threads should run on");
  57
  58 static struct mutex     ldlm_ref_mutex;
  59 static int ldlm_refcount;
  60
  61 static struct kobject *ldlm_kobj;
  62 struct kset *ldlm_ns_kset;
  63 static struct kset *ldlm_svc_kset;
  64
  65 struct ldlm_cb_async_args {
  66         struct ldlm_cb_set_arg *ca_set_arg;
  67         struct ldlm_lock       *ca_lock;
  68 };
  69
  70 /* LDLM state */
  71
  72 static struct ldlm_state *ldlm_state;
  73
  74 #define ELT_STOPPED   0
  75 #define ELT_READY     1
  76 #define ELT_TERMINATE 2
  77
  78 struct ldlm_bl_pool {
  79         spinlock_t              blp_lock;
  80
  81         /*
  82          * blp_prio_list is used for callbacks that should be handled
  83          * as a priority. It is used for LDLM_FL_DISCARD_DATA requests.
  84          * see bug 13843
  85          */
  86         struct list_head              blp_prio_list;
  87
  88         /*
  89          * blp_list is used for all other callbacks which are likely
  90          * to take longer to process.
  91          */
  92         struct list_head              blp_list;
  93
  94         wait_queue_head_t            blp_waitq;
  95         struct completion       blp_comp;
  96         atomic_t            blp_num_threads;
  97         atomic_t            blp_busy_threads;
  98         int                  blp_min_threads;
  99         int                  blp_max_threads;
 100 };
 101
 102 struct ldlm_bl_work_item {
 103         struct list_head              blwi_entry;
 104         struct ldlm_namespace  *blwi_ns;
 105         struct ldlm_lock_desc   blwi_ld;
 106         struct ldlm_lock       *blwi_lock;
 107         struct list_head              blwi_head;
 108         int                  blwi_count;
 109         struct completion       blwi_comp;
 110         enum ldlm_cancel_flags  blwi_flags;
 111         int                  blwi_mem_pressure;
 112 };
 113
 114 /**
 115  * Callback handler for receiving incoming blocking ASTs.
 116  *
 117  * This can only happen on client side.
 118  */
 119 void ldlm_handle_bl_callback(struct ldlm_namespace *ns,
 120                              struct ldlm_lock_desc *ld, struct ldlm_lock *lock)
 121 {
 122         int do_ast;
 123
 124         LDLM_DEBUG(lock, "client blocking AST callback handler");
 125
 126         lock_res_and_lock(lock);
 127         ldlm_set_cbpending(lock);
 128
 129         if (ldlm_is_cancel_on_block(lock))
 130                 ldlm_set_cancel(lock);
 131
 132         do_ast = !lock->l_readers && !lock->l_writers;
 133         unlock_res_and_lock(lock);
 134
 135         if (do_ast) {
 136                 CDEBUG(D_DLMTRACE,
 137                        "Lock %p already unused, calling callback (%p)\n", lock,
 138                        lock->l_blocking_ast);
 139                 if (lock->l_blocking_ast)
 140                         lock->l_blocking_ast(lock, ld, lock->l_ast_data,
 141                                              LDLM_CB_BLOCKING);
 142         } else {
 143                 CDEBUG(D_DLMTRACE,
 144                        "Lock %p is referenced, will be cancelled later\n",
 145                        lock);
 146         }
 147
 148         LDLM_DEBUG(lock, "client blocking callback handler END");
 149         LDLM_LOCK_RELEASE(lock);
 150 }
 151
 152 /**
 153  * Callback handler for receiving incoming completion ASTs.
 154  *
 155  * This only can happen on client side.
 156  */
 157 static void ldlm_handle_cp_callback(struct ptlrpc_request *req,
 158                                     struct ldlm_namespace *ns,
 159                                     struct ldlm_request *dlm_req,
 160                                     struct ldlm_lock *lock)
 161 {
 162         int lvb_len;
 163         LIST_HEAD(ast_list);
 164         int rc = 0;
 165
 166         LDLM_DEBUG(lock, "client completion callback handler START");
 167
 168         if (OBD_FAIL_CHECK(OBD_FAIL_LDLM_CANCEL_BL_CB_RACE)) {
 169                 int to = cfs_time_seconds(1);
 170
 171                 while (to > 0) {
 172                         set_current_state(TASK_INTERRUPTIBLE);
 173                         schedule_timeout(to);
 174                         if (lock->l_granted_mode == lock->l_req_mode ||
 175                             ldlm_is_destroyed(lock))
 176                                 break;
 177                 }
 178         }
 179
 180         lvb_len = req_capsule_get_size(&req->rq_pill, &RMF_DLM_LVB, RCL_CLIENT);
 181         if (lvb_len < 0) {
 182                 LDLM_ERROR(lock, "Fail to get lvb_len, rc = %d", lvb_len);
 183                 rc = lvb_len;
 184                 goto out;
 185         } else if (lvb_len > 0) {
 186                 if (lock->l_lvb_len > 0) {
 187                         /* for extent lock, lvb contains ost_lvb{}. */
 188                         LASSERT(lock->l_lvb_data);
 189
 190                         if (unlikely(lock->l_lvb_len < lvb_len)) {
 191                                 LDLM_ERROR(lock, "Replied LVB is larger than expectation, expected = %d, replied = %d",
 192                                            lock->l_lvb_len, lvb_len);
 193                                 rc = -EINVAL;
 194                                 goto out;
 195                         }
 196                 } else if (ldlm_has_layout(lock)) { /* for layout lock, lvb has
 197                                                      * variable length
 198                                                      */
 199                         void *lvb_data;
 200
 201                         lvb_data = kzalloc(lvb_len, GFP_NOFS);
 202                         if (!lvb_data) {
 203                                 LDLM_ERROR(lock, "No memory: %d.\n", lvb_len);
 204                                 rc = -ENOMEM;
 205                                 goto out;
 206                         }
 207
 208                         lock_res_and_lock(lock);
 209                         LASSERT(!lock->l_lvb_data);
 210                         lock->l_lvb_type = LVB_T_LAYOUT;
 211                         lock->l_lvb_data = lvb_data;
 212                         lock->l_lvb_len = lvb_len;
 213                         unlock_res_and_lock(lock);
 214                 }
 215         }
 216
 217         lock_res_and_lock(lock);
 218         if (ldlm_is_destroyed(lock) ||
 219             lock->l_granted_mode == lock->l_req_mode) {
 220                 /* bug 11300: the lock has already been granted */
 221                 unlock_res_and_lock(lock);
 222                 LDLM_DEBUG(lock, "Double grant race happened");
 223                 rc = 0;
 224                 goto out;
 225         }
 226
 227         /* If we receive the completion AST before the actual enqueue returned,
 228          * then we might need to switch lock modes, resources, or extents.
 229          */
 230         if (dlm_req->lock_desc.l_granted_mode != lock->l_req_mode) {
 231                 lock->l_req_mode = dlm_req->lock_desc.l_granted_mode;
 232                 LDLM_DEBUG(lock, "completion AST, new lock mode");
 233         }
 234
 235         if (lock->l_resource->lr_type != LDLM_PLAIN) {
 236                 ldlm_convert_policy_to_local(req->rq_export,
 237                                           dlm_req->lock_desc.l_resource.lr_type,
 238                                           &dlm_req->lock_desc.l_policy_data,
 239                                           &lock->l_policy_data);
 240                 LDLM_DEBUG(lock, "completion AST, new policy data");
 241         }
 242
 243         ldlm_resource_unlink_lock(lock);
 244         if (memcmp(&dlm_req->lock_desc.l_resource.lr_name,
 245                    &lock->l_resource->lr_name,
 246                    sizeof(lock->l_resource->lr_name)) != 0) {
 247                 unlock_res_and_lock(lock);
 248                 rc = ldlm_lock_change_resource(ns, lock,
 249                                 &dlm_req->lock_desc.l_resource.lr_name);
 250                 if (rc < 0) {
 251                         LDLM_ERROR(lock, "Failed to allocate resource");
 252                         goto out;
 253                 }
 254                 LDLM_DEBUG(lock, "completion AST, new resource");
 255                 CERROR("change resource!\n");
 256                 lock_res_and_lock(lock);
 257         }
 258
 259         if (dlm_req->lock_flags & LDLM_FL_AST_SENT) {
 260                 /* BL_AST locks are not needed in LRU.
 261                  * Let ldlm_cancel_lru() be fast.
 262                  */
 263                 ldlm_lock_remove_from_lru(lock);
 264                 lock->l_flags |= LDLM_FL_CBPENDING | LDLM_FL_BL_AST;
 265                 LDLM_DEBUG(lock, "completion AST includes blocking AST");
 266         }
 267
 268         if (lock->l_lvb_len > 0) {
 269                 rc = ldlm_fill_lvb(lock, &req->rq_pill, RCL_CLIENT,
 270                                    lock->l_lvb_data, lvb_len);
 271                 if (rc < 0) {
 272                         unlock_res_and_lock(lock);
 273                         goto out;
 274                 }
 275         }
 276
 277         ldlm_grant_lock(lock, &ast_list);
 278         unlock_res_and_lock(lock);
 279
 280         LDLM_DEBUG(lock, "callback handler finished, about to run_ast_work");
 281
 282         /* Let Enqueue to call osc_lock_upcall() and initialize l_ast_data */
 283         OBD_FAIL_TIMEOUT(OBD_FAIL_OSC_CP_ENQ_RACE, 2);
 284
 285         ldlm_run_ast_work(ns, &ast_list, LDLM_WORK_CP_AST);
 286
 287         LDLM_DEBUG_NOLOCK("client completion callback handler END (lock %p)",
 288                           lock);
 289         goto out;
 290
 291 out:
 292         if (rc < 0) {
 293                 lock_res_and_lock(lock);
 294                 ldlm_set_failed(lock);
 295                 unlock_res_and_lock(lock);
 296                 wake_up(&lock->l_waitq);
 297         }
 298         LDLM_LOCK_RELEASE(lock);
 299 }
 300
 301 /**
 302  * Callback handler for receiving incoming glimpse ASTs.
 303  *
 304  * This only can happen on client side.  After handling the glimpse AST
 305  * we also consider dropping the lock here if it is unused locally for a
 306  * long time.
 307  */
 308 static void ldlm_handle_gl_callback(struct ptlrpc_request *req,
 309                                     struct ldlm_namespace *ns,
 310                                     struct ldlm_request *dlm_req,
 311                                     struct ldlm_lock *lock)
 312 {
 313         int rc = -ENOSYS;
 314
 315         LDLM_DEBUG(lock, "client glimpse AST callback handler");
 316
 317         if (lock->l_glimpse_ast)
 318                 rc = lock->l_glimpse_ast(lock, req);
 319
 320         if (req->rq_repmsg) {
 321                 ptlrpc_reply(req);
 322         } else {
 323                 req->rq_status = rc;
 324                 ptlrpc_error(req);
 325         }
 326
 327         lock_res_and_lock(lock);
 328         if (lock->l_granted_mode == LCK_PW &&
 329             !lock->l_readers && !lock->l_writers &&
 330             cfs_time_after(cfs_time_current(),
 331                            cfs_time_add(lock->l_last_used,
 332                                         cfs_time_seconds(10)))) {
 333                 unlock_res_and_lock(lock);
 334                 if (ldlm_bl_to_thread_lock(ns, NULL, lock))
 335                         ldlm_handle_bl_callback(ns, NULL, lock);
 336
 337                 return;
 338         }
 339         unlock_res_and_lock(lock);
 340         LDLM_LOCK_RELEASE(lock);
 341 }
 342
 343 static int ldlm_callback_reply(struct ptlrpc_request *req, int rc)
 344 {
 345         if (req->rq_no_reply)
 346                 return 0;
 347
 348         req->rq_status = rc;
 349         if (!req->rq_packed_final) {
 350                 rc = lustre_pack_reply(req, 1, NULL, NULL);
 351                 if (rc)
 352                         return rc;
 353         }
 354         return ptlrpc_reply(req);
 355 }
 356
 357 static int __ldlm_bl_to_thread(struct ldlm_bl_work_item *blwi,
 358                                enum ldlm_cancel_flags cancel_flags)
 359 {
 360         struct ldlm_bl_pool *blp = ldlm_state->ldlm_bl_pool;
 361
 362         spin_lock(&blp->blp_lock);
 363         if (blwi->blwi_lock && ldlm_is_discard_data(blwi->blwi_lock)) {
 364                 /* add LDLM_FL_DISCARD_DATA requests to the priority list */
 365                 list_add_tail(&blwi->blwi_entry, &blp->blp_prio_list);
 366         } else {
 367                 /* other blocking callbacks are added to the regular list */
 368                 list_add_tail(&blwi->blwi_entry, &blp->blp_list);
 369         }
 370         spin_unlock(&blp->blp_lock);
 371
 372         wake_up(&blp->blp_waitq);
 373
 374         /* can not check blwi->blwi_flags as blwi could be already freed in
 375          * LCF_ASYNC mode
 376          */
 377         if (!(cancel_flags & LCF_ASYNC))
 378                 wait_for_completion(&blwi->blwi_comp);
 379
 380         return 0;
 381 }
 382
 383 static inline void init_blwi(struct ldlm_bl_work_item *blwi,
 384                              struct ldlm_namespace *ns,
 385                              struct ldlm_lock_desc *ld,
 386                              struct list_head *cancels, int count,
 387                              struct ldlm_lock *lock,
 388                              enum ldlm_cancel_flags cancel_flags)
 389 {
 390         init_completion(&blwi->blwi_comp);
 391         INIT_LIST_HEAD(&blwi->blwi_head);
 392
 393         if (memory_pressure_get())
 394                 blwi->blwi_mem_pressure = 1;
 395
 396         blwi->blwi_ns = ns;
 397         blwi->blwi_flags = cancel_flags;
 398         if (ld)
 399                 blwi->blwi_ld = *ld;
 400         if (count) {
 401                 list_add(&blwi->blwi_head, cancels);
 402                 list_del_init(cancels);
 403                 blwi->blwi_count = count;
 404         } else {
 405                 blwi->blwi_lock = lock;
 406         }
 407 }
 408
 409 /**
 410  * Queues a list of locks \a cancels containing \a count locks
 411  * for later processing by a blocking thread.  If \a count is zero,
 412  * then the lock referenced as \a lock is queued instead.
 413  *
 414  * The blocking thread would then call ->l_blocking_ast callback in the lock.
 415  * If list addition fails an error is returned and caller is supposed to
 416  * call ->l_blocking_ast itself.
 417  */
 418 static int ldlm_bl_to_thread(struct ldlm_namespace *ns,
 419                              struct ldlm_lock_desc *ld,
 420                              struct ldlm_lock *lock,
 421                              struct list_head *cancels, int count,
 422                              enum ldlm_cancel_flags cancel_flags)
 423 {
 424         if (cancels && count == 0)
 425                 return 0;
 426
 427         if (cancel_flags & LCF_ASYNC) {
 428                 struct ldlm_bl_work_item *blwi;
 429
 430                 blwi = kzalloc(sizeof(*blwi), GFP_NOFS);
 431                 if (!blwi)
 432                         return -ENOMEM;
 433                 init_blwi(blwi, ns, ld, cancels, count, lock, cancel_flags);
 434
 435                 return __ldlm_bl_to_thread(blwi, cancel_flags);
 436         } else {
 437                 /* if it is synchronous call do minimum mem alloc, as it could
 438                  * be triggered from kernel shrinker
 439                  */
 440                 struct ldlm_bl_work_item blwi;
 441
 442                 memset(&blwi, 0, sizeof(blwi));
 443                 init_blwi(&blwi, ns, ld, cancels, count, lock, cancel_flags);
 444                 return __ldlm_bl_to_thread(&blwi, cancel_flags);
 445         }
 446 }
 447
 448 int ldlm_bl_to_thread_lock(struct ldlm_namespace *ns, struct ldlm_lock_desc *ld,
 449                            struct ldlm_lock *lock)
 450 {
 451         return ldlm_bl_to_thread(ns, ld, lock, NULL, 0, LCF_ASYNC);
 452 }
 453
 454 int ldlm_bl_to_thread_list(struct ldlm_namespace *ns, struct ldlm_lock_desc *ld,
 455                            struct list_head *cancels, int count,
 456                            enum ldlm_cancel_flags cancel_flags)
 457 {
 458         return ldlm_bl_to_thread(ns, ld, NULL, cancels, count, cancel_flags);
 459 }
 460
 461 /* Setinfo coming from Server (eg MDT) to Client (eg MDC)! */
 462 static int ldlm_handle_setinfo(struct ptlrpc_request *req)
 463 {
 464         struct obd_device *obd = req->rq_export->exp_obd;
 465         char *key;
 466         void *val;
 467         int keylen, vallen;
 468         int rc = -ENOSYS;
 469
 470         DEBUG_REQ(D_HSM, req, "%s: handle setinfo\n", obd->obd_name);
 471
 472         req_capsule_set(&req->rq_pill, &RQF_OBD_SET_INFO);
 473
 474         key = req_capsule_client_get(&req->rq_pill, &RMF_SETINFO_KEY);
 475         if (!key) {
 476                 DEBUG_REQ(D_IOCTL, req, "no set_info key");
 477                 return -EFAULT;
 478         }
 479         keylen = req_capsule_get_size(&req->rq_pill, &RMF_SETINFO_KEY,
 480                                       RCL_CLIENT);
 481         val = req_capsule_client_get(&req->rq_pill, &RMF_SETINFO_VAL);
 482         if (!val) {
 483                 DEBUG_REQ(D_IOCTL, req, "no set_info val");
 484                 return -EFAULT;
 485         }
 486         vallen = req_capsule_get_size(&req->rq_pill, &RMF_SETINFO_VAL,
 487                                       RCL_CLIENT);
 488
 489         /* We are responsible for swabbing contents of val */
 490
 491         if (KEY_IS(KEY_HSM_COPYTOOL_SEND))
 492                 /* Pass it on to mdc (the "export" in this case) */
 493                 rc = obd_set_info_async(req->rq_svc_thread->t_env,
 494                                         req->rq_export,
 495                                         sizeof(KEY_HSM_COPYTOOL_SEND),
 496                                         KEY_HSM_COPYTOOL_SEND,
 497                                         vallen, val, NULL);
 498         else
 499                 DEBUG_REQ(D_WARNING, req, "ignoring unknown key %s", key);
 500
 501         return rc;
 502 }
 503
 504 static inline void ldlm_callback_errmsg(struct ptlrpc_request *req,
 505                                         const char *msg, int rc,
 506                                         struct lustre_handle *handle)
 507 {
 508         DEBUG_REQ((req->rq_no_reply || rc) ? D_WARNING : D_DLMTRACE, req,
 509                   "%s: [nid %s] [rc %d] [lock %#llx]",
 510                   msg, libcfs_id2str(req->rq_peer), rc,
 511                   handle ? handle->cookie : 0);
 512         if (req->rq_no_reply)
 513                 CWARN("No reply was sent, maybe cause bug 21636.\n");
 514         else if (rc)
 515                 CWARN("Send reply failed, maybe cause bug 21636.\n");
 516 }
 517
 518 static int ldlm_handle_qc_callback(struct ptlrpc_request *req)
 519 {
 520         struct obd_quotactl *oqctl;
 521         struct client_obd *cli = &req->rq_export->exp_obd->u.cli;
 522
 523         oqctl = req_capsule_client_get(&req->rq_pill, &RMF_OBD_QUOTACTL);
 524         if (!oqctl) {
 525                 CERROR("Can't unpack obd_quotactl\n");
 526                 return -EPROTO;
 527         }
 528
 529         oqctl->qc_stat = ptlrpc_status_ntoh(oqctl->qc_stat);
 530
 531         cli->cl_qchk_stat = oqctl->qc_stat;
 532         return 0;
 533 }
 534
 535 /* TODO: handle requests in a similar way as MDT: see mdt_handle_common() */
 536 static int ldlm_callback_handler(struct ptlrpc_request *req)
 537 {
 538         struct ldlm_namespace *ns;
 539         struct ldlm_request *dlm_req;
 540         struct ldlm_lock *lock;
 541         int rc;
 542
 543         /* Requests arrive in sender's byte order.  The ptlrpc service
 544          * handler has already checked and, if necessary, byte-swapped the
 545          * incoming request message body, but I am responsible for the
 546          * message buffers.
 547          */
 548
 549         /* do nothing for sec context finalize */
 550         if (lustre_msg_get_opc(req->rq_reqmsg) == SEC_CTX_FINI)
 551                 return 0;
 552
 553         req_capsule_init(&req->rq_pill, req, RCL_SERVER);
 554
 555         if (!req->rq_export) {
 556                 rc = ldlm_callback_reply(req, -ENOTCONN);
 557                 ldlm_callback_errmsg(req, "Operate on unconnected server",
 558                                      rc, NULL);
 559                 return 0;
 560         }
 561
 562         LASSERT(req->rq_export->exp_obd);
 563
 564         switch (lustre_msg_get_opc(req->rq_reqmsg)) {
 565         case LDLM_BL_CALLBACK:
 566                 if (OBD_FAIL_CHECK(OBD_FAIL_LDLM_BL_CALLBACK_NET))
 567                         return 0;
 568                 break;
 569         case LDLM_CP_CALLBACK:
 570                 if (OBD_FAIL_CHECK(OBD_FAIL_LDLM_CP_CALLBACK_NET))
 571                         return 0;
 572                 break;
 573         case LDLM_GL_CALLBACK:
 574                 if (OBD_FAIL_CHECK(OBD_FAIL_LDLM_GL_CALLBACK_NET))
 575                         return 0;
 576                 break;
 577         case LDLM_SET_INFO:
 578                 rc = ldlm_handle_setinfo(req);
 579                 ldlm_callback_reply(req, rc);
 580                 return 0;
 581         case OBD_QC_CALLBACK:
 582                 req_capsule_set(&req->rq_pill, &RQF_QC_CALLBACK);
 583                 if (OBD_FAIL_CHECK(OBD_FAIL_OBD_QC_CALLBACK_NET))
 584                         return 0;
 585                 rc = ldlm_handle_qc_callback(req);
 586                 ldlm_callback_reply(req, rc);
 587                 return 0;
 588         default:
 589                 CERROR("unknown opcode %u\n",
 590                        lustre_msg_get_opc(req->rq_reqmsg));
 591                 ldlm_callback_reply(req, -EPROTO);
 592                 return 0;
 593         }
 594
 595         ns = req->rq_export->exp_obd->obd_namespace;
 596         LASSERT(ns);
 597
 598         req_capsule_set(&req->rq_pill, &RQF_LDLM_CALLBACK);
 599
 600         dlm_req = req_capsule_client_get(&req->rq_pill, &RMF_DLM_REQ);
 601         if (!dlm_req) {
 602                 rc = ldlm_callback_reply(req, -EPROTO);
 603                 ldlm_callback_errmsg(req, "Operate without parameter", rc,
 604                                      NULL);
 605                 return 0;
 606         }
 607
 608         /* Force a known safe race, send a cancel to the server for a lock
 609          * which the server has already started a blocking callback on.
 610          */
 611         if (OBD_FAIL_CHECK(OBD_FAIL_LDLM_CANCEL_BL_CB_RACE) &&
 612             lustre_msg_get_opc(req->rq_reqmsg) == LDLM_BL_CALLBACK) {
 613                 rc = ldlm_cli_cancel(&dlm_req->lock_handle[0], 0);
 614                 if (rc < 0)
 615                         CERROR("ldlm_cli_cancel: %d\n", rc);
 616         }
 617
 618         lock = ldlm_handle2lock_long(&dlm_req->lock_handle[0], 0);
 619         if (!lock) {
 620                 CDEBUG(D_DLMTRACE, "callback on lock %#llx - lock disappeared\n",
 621                        dlm_req->lock_handle[0].cookie);
 622                 rc = ldlm_callback_reply(req, -EINVAL);
 623                 ldlm_callback_errmsg(req, "Operate with invalid parameter", rc,
 624                                      &dlm_req->lock_handle[0]);
 625                 return 0;
 626         }
 627
 628         if (ldlm_is_fail_loc(lock) &&
 629             lustre_msg_get_opc(req->rq_reqmsg) == LDLM_BL_CALLBACK)
 630                 OBD_RACE(OBD_FAIL_LDLM_CP_BL_RACE);
 631
 632         /* Copy hints/flags (e.g. LDLM_FL_DISCARD_DATA) from AST. */
 633         lock_res_and_lock(lock);
 634         lock->l_flags |= ldlm_flags_from_wire(dlm_req->lock_flags &
 635                                               LDLM_FL_AST_MASK);
 636         if (lustre_msg_get_opc(req->rq_reqmsg) == LDLM_BL_CALLBACK) {
 637                 /* If somebody cancels lock and cache is already dropped,
 638                  * or lock is failed before cp_ast received on client,
 639                  * we can tell the server we have no lock. Otherwise, we
 640                  * should send cancel after dropping the cache.
 641                  */
 642                 if ((ldlm_is_canceling(lock) && ldlm_is_bl_done(lock)) ||
 643                     ldlm_is_failed(lock)) {
 644                         LDLM_DEBUG(lock, "callback on lock %#llx - lock disappeared\n",
 645                                    dlm_req->lock_handle[0].cookie);
 646                         unlock_res_and_lock(lock);
 647                         LDLM_LOCK_RELEASE(lock);
 648                         rc = ldlm_callback_reply(req, -EINVAL);
 649                         ldlm_callback_errmsg(req, "Operate on stale lock", rc,
 650                                              &dlm_req->lock_handle[0]);
 651                         return 0;
 652                 }
 653                 /* BL_AST locks are not needed in LRU.
 654                  * Let ldlm_cancel_lru() be fast.
 655                  */
 656                 ldlm_lock_remove_from_lru(lock);
 657                 ldlm_set_bl_ast(lock);
 658         }
 659         unlock_res_and_lock(lock);
 660
 661         /* We want the ost thread to get this reply so that it can respond
 662          * to ost requests (write cache writeback) that might be triggered
 663          * in the callback.
 664          *
 665          * But we'd also like to be able to indicate in the reply that we're
 666          * cancelling right now, because it's unused, or have an intent result
 667          * in the reply, so we might have to push the responsibility for sending
 668          * the reply down into the AST handlers, alas.
 669          */
 670
 671         switch (lustre_msg_get_opc(req->rq_reqmsg)) {
 672         case LDLM_BL_CALLBACK:
 673                 CDEBUG(D_INODE, "blocking ast\n");
 674                 req_capsule_extend(&req->rq_pill, &RQF_LDLM_BL_CALLBACK);
 675                 if (!ldlm_is_cancel_on_block(lock)) {
 676                         rc = ldlm_callback_reply(req, 0);
 677                         if (req->rq_no_reply || rc)
 678                                 ldlm_callback_errmsg(req, "Normal process", rc,
 679                                                      &dlm_req->lock_handle[0]);
 680                 }
 681                 if (ldlm_bl_to_thread_lock(ns, &dlm_req->lock_desc, lock))
 682                         ldlm_handle_bl_callback(ns, &dlm_req->lock_desc, lock);
 683                 break;
 684         case LDLM_CP_CALLBACK:
 685                 CDEBUG(D_INODE, "completion ast\n");
 686                 req_capsule_extend(&req->rq_pill, &RQF_LDLM_CP_CALLBACK);
 687                 ldlm_callback_reply(req, 0);
 688                 ldlm_handle_cp_callback(req, ns, dlm_req, lock);
 689                 break;
 690         case LDLM_GL_CALLBACK:
 691                 CDEBUG(D_INODE, "glimpse ast\n");
 692                 req_capsule_extend(&req->rq_pill, &RQF_LDLM_GL_CALLBACK);
 693                 ldlm_handle_gl_callback(req, ns, dlm_req, lock);
 694                 break;
 695         default:
 696                 LBUG();                  /* checked above */
 697         }
 698
 699         return 0;
 700 }
 701
 702 static struct ldlm_bl_work_item *ldlm_bl_get_work(struct ldlm_bl_pool *blp)
 703 {
 704         struct ldlm_bl_work_item *blwi = NULL;
 705         static unsigned int num_bl;
 706
 707         spin_lock(&blp->blp_lock);
 708         /* process a request from the blp_list at least every blp_num_threads */
 709         if (!list_empty(&blp->blp_list) &&
 710             (list_empty(&blp->blp_prio_list) || num_bl == 0))
 711                 blwi = list_entry(blp->blp_list.next,
 712                                       struct ldlm_bl_work_item, blwi_entry);
 713         else
 714                 if (!list_empty(&blp->blp_prio_list))
 715                         blwi = list_entry(blp->blp_prio_list.next,
 716                                               struct ldlm_bl_work_item,
 717                                               blwi_entry);
 718
 719         if (blwi) {
 720                 if (++num_bl >= atomic_read(&blp->blp_num_threads))
 721                         num_bl = 0;
 722                 list_del(&blwi->blwi_entry);
 723         }
 724         spin_unlock(&blp->blp_lock);
 725
 726         return blwi;
 727 }
 728
 729 /* This only contains temporary data until the thread starts */
 730 struct ldlm_bl_thread_data {
 731         char                    bltd_name[CFS_CURPROC_COMM_MAX];
 732         struct ldlm_bl_pool     *bltd_blp;
 733         struct completion       bltd_comp;
 734         int                     bltd_num;
 735 };
 736
 737 static int ldlm_bl_thread_main(void *arg);
 738
 739 static int ldlm_bl_thread_start(struct ldlm_bl_pool *blp)
 740 {
 741         struct ldlm_bl_thread_data bltd = { .bltd_blp = blp };
 742         struct task_struct *task;
 743
 744         init_completion(&bltd.bltd_comp);
 745         bltd.bltd_num = atomic_read(&blp->blp_num_threads);
 746         snprintf(bltd.bltd_name, sizeof(bltd.bltd_name),
 747                 "ldlm_bl_%02d", bltd.bltd_num);
 748         task = kthread_run(ldlm_bl_thread_main, &bltd, "%s", bltd.bltd_name);
 749         if (IS_ERR(task)) {
 750                 CERROR("cannot start LDLM thread ldlm_bl_%02d: rc %ld\n",
 751                        atomic_read(&blp->blp_num_threads), PTR_ERR(task));
 752                 return PTR_ERR(task);
 753         }
 754         wait_for_completion(&bltd.bltd_comp);
 755
 756         return 0;
 757 }
 758
 759 /**
 760  * Main blocking requests processing thread.
 761  *
 762  * Callers put locks into its queue by calling ldlm_bl_to_thread.
 763  * This thread in the end ends up doing actual call to ->l_blocking_ast
 764  * for queued locks.
 765  */
 766 static int ldlm_bl_thread_main(void *arg)
 767 {
 768         struct ldlm_bl_pool *blp;
 769
 770         {
 771                 struct ldlm_bl_thread_data *bltd = arg;
 772
 773                 blp = bltd->bltd_blp;
 774
 775                 atomic_inc(&blp->blp_num_threads);
 776                 atomic_inc(&blp->blp_busy_threads);
 777
 778                 complete(&bltd->bltd_comp);
 779                 /* cannot use bltd after this, it is only on caller's stack */
 780         }
 781
 782         while (1) {
 783                 struct l_wait_info lwi = { 0 };
 784                 struct ldlm_bl_work_item *blwi = NULL;
 785                 int busy;
 786
 787                 blwi = ldlm_bl_get_work(blp);
 788
 789                 if (!blwi) {
 790                         atomic_dec(&blp->blp_busy_threads);
 791                         l_wait_event_exclusive(blp->blp_waitq,
 792                                          (blwi = ldlm_bl_get_work(blp)),
 793                                          &lwi);
 794                         busy = atomic_inc_return(&blp->blp_busy_threads);
 795                 } else {
 796                         busy = atomic_read(&blp->blp_busy_threads);
 797                 }
 798
 799                 if (!blwi->blwi_ns)
 800                         /* added by ldlm_cleanup() */
 801                         break;
 802
 803                 /* Not fatal if racy and have a few too many threads */
 804                 if (unlikely(busy < blp->blp_max_threads &&
 805                              busy >= atomic_read(&blp->blp_num_threads) &&
 806                              !blwi->blwi_mem_pressure))
 807                         /* discard the return value, we tried */
 808                         ldlm_bl_thread_start(blp);
 809
 810                 if (blwi->blwi_mem_pressure)
 811                         memory_pressure_set();
 812
 813                 if (blwi->blwi_count) {
 814                         int count;
 815                         /* The special case when we cancel locks in LRU
 816                          * asynchronously, we pass the list of locks here.
 817                          * Thus locks are marked LDLM_FL_CANCELING, but NOT
 818                          * canceled locally yet.
 819                          */
 820                         count = ldlm_cli_cancel_list_local(&blwi->blwi_head,
 821                                                            blwi->blwi_count,
 822                                                            LCF_BL_AST);
 823                         ldlm_cli_cancel_list(&blwi->blwi_head, count, NULL,
 824                                              blwi->blwi_flags);
 825                 } else {
 826                         ldlm_handle_bl_callback(blwi->blwi_ns, &blwi->blwi_ld,
 827                                                 blwi->blwi_lock);
 828                 }
 829                 if (blwi->blwi_mem_pressure)
 830                         memory_pressure_clr();
 831
 832                 if (blwi->blwi_flags & LCF_ASYNC)
 833                         kfree(blwi);
 834                 else
 835                         complete(&blwi->blwi_comp);
 836         }
 837
 838         atomic_dec(&blp->blp_busy_threads);
 839         atomic_dec(&blp->blp_num_threads);
 840         complete(&blp->blp_comp);
 841         return 0;
 842 }
 843
 844 static int ldlm_setup(void);
 845 static int ldlm_cleanup(void);
 846
 847 int ldlm_get_ref(void)
 848 {
 849         int rc = 0;
 850
 851         mutex_lock(&ldlm_ref_mutex);
 852         if (++ldlm_refcount == 1) {
 853                 rc = ldlm_setup();
 854                 if (rc)
 855                         ldlm_refcount--;
 856         }
 857         mutex_unlock(&ldlm_ref_mutex);
 858
 859         return rc;
 860 }
 861 EXPORT_SYMBOL(ldlm_get_ref);
 862
 863 void ldlm_put_ref(void)
 864 {
 865         mutex_lock(&ldlm_ref_mutex);
 866         if (ldlm_refcount == 1) {
 867                 int rc = ldlm_cleanup();
 868
 869                 if (rc)
 870                         CERROR("ldlm_cleanup failed: %d\n", rc);
 871                 else
 872                         ldlm_refcount--;
 873         } else {
 874                 ldlm_refcount--;
 875         }
 876         mutex_unlock(&ldlm_ref_mutex);
 877 }
 878 EXPORT_SYMBOL(ldlm_put_ref);
 879
 880 extern unsigned int ldlm_cancel_unused_locks_before_replay;
 881
 882 static ssize_t cancel_unused_locks_before_replay_show(struct kobject *kobj,
 883                                                       struct attribute *attr,
 884                                                       char *buf)
 885 {
 886         return sprintf(buf, "%d\n", ldlm_cancel_unused_locks_before_replay);
 887 }
 888
 889 static ssize_t cancel_unused_locks_before_replay_store(struct kobject *kobj,
 890                                                        struct attribute *attr,
 891                                                        const char *buffer,
 892                                                        size_t count)
 893 {
 894         int rc;
 895         unsigned long val;
 896
 897         rc = kstrtoul(buffer, 10, &val);
 898         if (rc)
 899                 return rc;
 900
 901         ldlm_cancel_unused_locks_before_replay = val;
 902
 903         return count;
 904 }
 905 LUSTRE_RW_ATTR(cancel_unused_locks_before_replay);
 906
 907 /* These are for root of /sys/fs/lustre/ldlm */
 908 static struct attribute *ldlm_attrs[] = {
 909         &lustre_attr_cancel_unused_locks_before_replay.attr,
 910         NULL,
 911 };
 912
 913 static struct attribute_group ldlm_attr_group = {
 914         .attrs = ldlm_attrs,
 915 };
 916
 917 static int ldlm_setup(void)
 918 {
 919         static struct ptlrpc_service_conf       conf;
 920         struct ldlm_bl_pool                     *blp = NULL;
 921         int rc = 0;
 922         int i;
 923
 924         if (ldlm_state)
 925                 return -EALREADY;
 926
 927         ldlm_state = kzalloc(sizeof(*ldlm_state), GFP_NOFS);
 928         if (!ldlm_state)
 929                 return -ENOMEM;
 930
 931         ldlm_kobj = kobject_create_and_add("ldlm", lustre_kobj);
 932         if (!ldlm_kobj) {
 933                 rc = -ENOMEM;
 934                 goto out;
 935         }
 936
 937         rc = sysfs_create_group(ldlm_kobj, &ldlm_attr_group);
 938         if (rc)
 939                 goto out;
 940
 941         ldlm_ns_kset = kset_create_and_add("namespaces", NULL, ldlm_kobj);
 942         if (!ldlm_ns_kset) {
 943                 rc = -ENOMEM;
 944                 goto out;
 945         }
 946
 947         ldlm_svc_kset = kset_create_and_add("services", NULL, ldlm_kobj);
 948         if (!ldlm_svc_kset) {
 949                 rc = -ENOMEM;
 950                 goto out;
 951         }
 952
 953         rc = ldlm_debugfs_setup();
 954         if (rc != 0)
 955                 goto out;
 956
 957         memset(&conf, 0, sizeof(conf));
 958         conf = (typeof(conf)) {
 959                 .psc_name               = "ldlm_cbd",
 960                 .psc_watchdog_factor    = 2,
 961                 .psc_buf                = {
 962                         .bc_nbufs               = LDLM_CLIENT_NBUFS,
 963                         .bc_buf_size            = LDLM_BUFSIZE,
 964                         .bc_req_max_size        = LDLM_MAXREQSIZE,
 965                         .bc_rep_max_size        = LDLM_MAXREPSIZE,
 966                         .bc_req_portal          = LDLM_CB_REQUEST_PORTAL,
 967                         .bc_rep_portal          = LDLM_CB_REPLY_PORTAL,
 968                 },
 969                 .psc_thr                = {
 970                         .tc_thr_name            = "ldlm_cb",
 971                         .tc_thr_factor          = LDLM_THR_FACTOR,
 972                         .tc_nthrs_init          = LDLM_NTHRS_INIT,
 973                         .tc_nthrs_base          = LDLM_NTHRS_BASE,
 974                         .tc_nthrs_max           = LDLM_NTHRS_MAX,
 975                         .tc_nthrs_user          = ldlm_num_threads,
 976                         .tc_cpu_affinity        = 1,
 977                         .tc_ctx_tags            = LCT_MD_THREAD | LCT_DT_THREAD,
 978                 },
 979                 .psc_cpt                = {
 980                         .cc_pattern             = ldlm_cpts,
 981                 },
 982                 .psc_ops                = {
 983                         .so_req_handler         = ldlm_callback_handler,
 984                 },
 985         };
 986         ldlm_state->ldlm_cb_service =
 987                         ptlrpc_register_service(&conf, ldlm_svc_kset,
 988                                                 ldlm_svc_debugfs_dir);
 989         if (IS_ERR(ldlm_state->ldlm_cb_service)) {
 990                 CERROR("failed to start service\n");
 991                 rc = PTR_ERR(ldlm_state->ldlm_cb_service);
 992                 ldlm_state->ldlm_cb_service = NULL;
 993                 goto out;
 994         }
 995
 996         blp = kzalloc(sizeof(*blp), GFP_NOFS);
 997         if (!blp) {
 998                 rc = -ENOMEM;
 999                 goto out;
1000         }
1001         ldlm_state->ldlm_bl_pool = blp;
1002
1003         spin_lock_init(&blp->blp_lock);
1004         INIT_LIST_HEAD(&blp->blp_list);
1005         INIT_LIST_HEAD(&blp->blp_prio_list);
1006         init_waitqueue_head(&blp->blp_waitq);
1007         atomic_set(&blp->blp_num_threads, 0);
1008         atomic_set(&blp->blp_busy_threads, 0);
1009
1010         if (ldlm_num_threads == 0) {
1011                 blp->blp_min_threads = LDLM_NTHRS_INIT;
1012                 blp->blp_max_threads = LDLM_NTHRS_MAX;
1013         } else {
1014                 blp->blp_min_threads = blp->blp_max_threads =
1015                         min_t(int, LDLM_NTHRS_MAX, max_t(int, LDLM_NTHRS_INIT,
1016                                                          ldlm_num_threads));
1017         }
1018
1019         for (i = 0; i < blp->blp_min_threads; i++) {
1020                 rc = ldlm_bl_thread_start(blp);
1021                 if (rc < 0)
1022                         goto out;
1023         }
1024
1025         rc = ldlm_pools_init();
1026         if (rc) {
1027                 CERROR("Failed to initialize LDLM pools: %d\n", rc);
1028                 goto out;
1029         }
1030         return 0;
1031
1032  out:
1033         ldlm_cleanup();
1034         return rc;
1035 }
1036
1037 static int ldlm_cleanup(void)
1038 {
1039         if (!list_empty(ldlm_namespace_list(LDLM_NAMESPACE_SERVER)) ||
1040             !list_empty(ldlm_namespace_list(LDLM_NAMESPACE_CLIENT))) {
1041                 CERROR("ldlm still has namespaces; clean these up first.\n");
1042                 ldlm_dump_all_namespaces(LDLM_NAMESPACE_SERVER, D_DLMTRACE);
1043                 ldlm_dump_all_namespaces(LDLM_NAMESPACE_CLIENT, D_DLMTRACE);
1044                 return -EBUSY;
1045         }
1046
1047         ldlm_pools_fini();
1048
1049         if (ldlm_state->ldlm_bl_pool) {
1050                 struct ldlm_bl_pool *blp = ldlm_state->ldlm_bl_pool;
1051
1052                 while (atomic_read(&blp->blp_num_threads) > 0) {
1053                         struct ldlm_bl_work_item blwi = { .blwi_ns = NULL };
1054
1055                         init_completion(&blp->blp_comp);
1056
1057                         spin_lock(&blp->blp_lock);
1058                         list_add_tail(&blwi.blwi_entry, &blp->blp_list);
1059                         wake_up(&blp->blp_waitq);
1060                         spin_unlock(&blp->blp_lock);
1061
1062                         wait_for_completion(&blp->blp_comp);
1063                 }
1064
1065                 kfree(blp);
1066         }
1067
1068         if (ldlm_state->ldlm_cb_service)
1069                 ptlrpc_unregister_service(ldlm_state->ldlm_cb_service);
1070
1071         if (ldlm_ns_kset)
1072                 kset_unregister(ldlm_ns_kset);
1073         if (ldlm_svc_kset)
1074                 kset_unregister(ldlm_svc_kset);
1075         if (ldlm_kobj)
1076                 kobject_put(ldlm_kobj);
1077
1078         ldlm_debugfs_cleanup();
1079
1080         kfree(ldlm_state);
1081         ldlm_state = NULL;
1082
1083         return 0;
1084 }
1085
1086 int ldlm_init(void)
1087 {
1088         mutex_init(&ldlm_ref_mutex);
1089         mutex_init(ldlm_namespace_lock(LDLM_NAMESPACE_SERVER));
1090         mutex_init(ldlm_namespace_lock(LDLM_NAMESPACE_CLIENT));
1091         ldlm_resource_slab = kmem_cache_create("ldlm_resources",
1092                                                sizeof(struct ldlm_resource), 0,
1093                                                SLAB_HWCACHE_ALIGN, NULL);
1094         if (!ldlm_resource_slab)
1095                 return -ENOMEM;
1096
1097         ldlm_lock_slab = kmem_cache_create("ldlm_locks",
1098                               sizeof(struct ldlm_lock), 0,
1099                               SLAB_HWCACHE_ALIGN | SLAB_DESTROY_BY_RCU, NULL);
1100         if (!ldlm_lock_slab) {
1101                 kmem_cache_destroy(ldlm_resource_slab);
1102                 return -ENOMEM;
1103         }
1104
1105         ldlm_interval_slab = kmem_cache_create("interval_node",
1106                                         sizeof(struct ldlm_interval),
1107                                         0, SLAB_HWCACHE_ALIGN, NULL);
1108         if (!ldlm_interval_slab) {
1109                 kmem_cache_destroy(ldlm_resource_slab);
1110                 kmem_cache_destroy(ldlm_lock_slab);
1111                 return -ENOMEM;
1112         }
1113 #if LUSTRE_TRACKS_LOCK_EXP_REFS
1114         class_export_dump_hook = ldlm_dump_export_locks;
1115 #endif
1116         return 0;
1117 }
1118
1119 void ldlm_exit(void)
1120 {
1121         if (ldlm_refcount)
1122                 CERROR("ldlm_refcount is %d in ldlm_exit!\n", ldlm_refcount);
1123         kmem_cache_destroy(ldlm_resource_slab);
1124         /* ldlm_lock_put() use RCU to call ldlm_lock_free, so need call
1125          * synchronize_rcu() to wait a grace period elapsed, so that
1126          * ldlm_lock_free() get a chance to be called.
1127          */
1128         synchronize_rcu();
1129         kmem_cache_destroy(ldlm_lock_slab);
1130         kmem_cache_destroy(ldlm_interval_slab);
1131 }