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, 2003 Cluster File Systems, Inc.
   5  *   Author: Peter Braam <braam@clusterfs.com>
   6  *   Author: Phil Schwan <phil@clusterfs.com>
   7  *
   8  *   This file is part of Lustre, http://www.lustre.org.
   9  *
  10  *   Lustre is free software; you can redistribute it and/or
  11  *   modify it under the terms of version 2 of the GNU General Public
  12  *   License as published by the Free Software Foundation.
  13  *
  14  *   Lustre is distributed in the hope that it will be useful,
  15  *   but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  17  *   GNU General Public License for more details.
  18  *
  19  *   You should have received a copy of the GNU General Public License
  20  *   along with Lustre; if not, write to the Free Software
  21  *   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  22  */
  23
  24 #ifndef EXPORT_SYMTAB
  25 # define EXPORT_SYMTAB
  26 #endif
  27 #define DEBUG_SUBSYSTEM S_LDLM
  28
  29 #ifdef __KERNEL__
  30 # include <linux/module.h>
  31 # include <linux/slab.h>
  32 # include <linux/init.h>
  33 # include <linux/wait.h>
  34 #else
  35 # include <liblustre.h>
  36 #endif
  37
  38 #include <linux/lustre_dlm.h>
  39 #include <linux/obd_class.h>
  40 #include <portals/list.h>
  41 #include "ldlm_internal.h"
  42
  43 extern kmem_cache_t *ldlm_resource_slab;
  44 extern kmem_cache_t *ldlm_lock_slab;
  45 extern struct lustre_lock ldlm_handle_lock;
  46 extern struct list_head ldlm_namespace_list;
  47
  48 static DECLARE_MUTEX(ldlm_ref_sem);
  49 static int ldlm_refcount;
  50
  51 /* LDLM state */
  52
  53 static struct ldlm_state *ldlm_state;
  54
  55 inline unsigned long round_timeout(unsigned long timeout)
  56 {
  57         return ((timeout / HZ) + 1) * HZ;
  58 }
  59
  60 #ifdef __KERNEL__
  61 /* XXX should this be per-ldlm? */
  62 static struct list_head waiting_locks_list;
  63 static spinlock_t waiting_locks_spinlock;
  64 static struct timer_list waiting_locks_timer;
  65
  66 static struct expired_lock_thread {
  67         wait_queue_head_t         elt_waitq;
  68         int                       elt_state;
  69         struct list_head          elt_expired_locks;
  70         spinlock_t                elt_lock;
  71 } expired_lock_thread;
  72 #endif
  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         struct list_head        blp_list;
  81         wait_queue_head_t       blp_waitq;
  82         atomic_t                blp_num_threads;
  83         struct completion       blp_comp;
  84 };
  85
  86 struct ldlm_bl_work_item {
  87         struct list_head        blwi_entry;
  88         struct ldlm_namespace   *blwi_ns;
  89         struct ldlm_lock_desc   blwi_ld;
  90         struct ldlm_lock        *blwi_lock;
  91 };
  92
  93 #ifdef __KERNEL__
  94
  95 static inline int have_expired_locks(void)
  96 {
  97         int need_to_run;
  98
  99         spin_lock_bh(&expired_lock_thread.elt_lock);
 100         need_to_run = !list_empty(&expired_lock_thread.elt_expired_locks);
 101         spin_unlock_bh(&expired_lock_thread.elt_lock);
 102
 103         RETURN(need_to_run);
 104 }
 105
 106 static int expired_lock_main(void *arg)
 107 {
 108         struct list_head *expired = &expired_lock_thread.elt_expired_locks;
 109         struct l_wait_info lwi = { 0 };
 110         unsigned long flags;
 111
 112         ENTRY;
 113         lock_kernel();
 114         kportal_daemonize("ldlm_elt");
 115
 116         SIGNAL_MASK_LOCK(current, flags);
 117         sigfillset(&current->blocked);
 118         RECALC_SIGPENDING;
 119         SIGNAL_MASK_UNLOCK(current, flags);
 120
 121         unlock_kernel();
 122
 123         expired_lock_thread.elt_state = ELT_READY;
 124         wake_up(&expired_lock_thread.elt_waitq);
 125
 126         while (1) {
 127                 l_wait_event(expired_lock_thread.elt_waitq,
 128                              have_expired_locks() ||
 129                              expired_lock_thread.elt_state == ELT_TERMINATE,
 130                              &lwi);
 131
 132                 spin_lock_bh(&expired_lock_thread.elt_lock);
 133                 while (!list_empty(expired)) {
 134                         struct obd_export *export;
 135                         struct ldlm_lock *lock;
 136
 137                         lock = list_entry(expired->next, struct ldlm_lock,
 138                                           l_pending_chain);
 139                         if ((void *)lock < LP_POISON + PAGE_SIZE &&
 140                             (void *)lock >= LP_POISON) {
 141                                 CERROR("free lock on elt list %p\n", lock);
 142                                 LBUG();
 143                         }
 144                         list_del_init(&lock->l_pending_chain);
 145                         if ((void *)lock->l_export < LP_POISON + PAGE_SIZE &&
 146                             (void *)lock->l_export >= LP_POISON + PAGE_SIZE) {
 147                                 CERROR("lock with free export on elt list %p\n",
 148                                        export);
 149                                 lock->l_export = NULL;
 150                                 LDLM_ERROR(lock, "free export\n");
 151                                 continue;
 152                         }
 153                         export = class_export_get(lock->l_export);
 154                         spin_unlock_bh(&expired_lock_thread.elt_lock);
 155
 156                         ptlrpc_fail_export(export);
 157                         class_export_put(export);
 158                         spin_lock_bh(&expired_lock_thread.elt_lock);
 159                 }
 160                 spin_unlock_bh(&expired_lock_thread.elt_lock);
 161
 162                 if (expired_lock_thread.elt_state == ELT_TERMINATE)
 163                         break;
 164         }
 165
 166         expired_lock_thread.elt_state = ELT_STOPPED;
 167         wake_up(&expired_lock_thread.elt_waitq);
 168         RETURN(0);
 169 }
 170
 171 static void waiting_locks_callback(unsigned long unused)
 172 {
 173         struct ldlm_lock *lock;
 174         char str[PTL_NALFMT_SIZE];
 175
 176         spin_lock_bh(&waiting_locks_spinlock);
 177         while (!list_empty(&waiting_locks_list)) {
 178                 lock = list_entry(waiting_locks_list.next, struct ldlm_lock,
 179                                   l_pending_chain);
 180
 181                 if (lock->l_callback_timeout > jiffies)
 182                         break;
 183
 184                 LDLM_ERROR(lock, "lock callback timer expired: evicting client "
 185                            "%s@%s nid "LPX64" (%s) ",
 186                            lock->l_export->exp_client_uuid.uuid,
 187                            lock->l_export->exp_connection->c_remote_uuid.uuid,
 188                            lock->l_export->exp_connection->c_peer.peer_nid,
 189                            portals_nid2str(lock->l_export->exp_connection->c_peer.peer_ni->pni_number,
 190                                            lock->l_export->exp_connection->c_peer.peer_nid,
 191                                            str));
 192
 193                 spin_lock_bh(&expired_lock_thread.elt_lock);
 194                 list_del(&lock->l_pending_chain);
 195                 list_add(&lock->l_pending_chain,
 196                          &expired_lock_thread.elt_expired_locks);
 197                 spin_unlock_bh(&expired_lock_thread.elt_lock);
 198                 wake_up(&expired_lock_thread.elt_waitq);
 199         }
 200
 201         /*
 202          * Make sure the timer will fire again if we have any locks
 203          * left.
 204          */
 205         if (!list_empty(&waiting_locks_list)) {
 206                 unsigned long timeout_rounded;
 207                 lock = list_entry(waiting_locks_list.next, struct ldlm_lock,
 208                                   l_pending_chain);
 209                 timeout_rounded = round_timeout(lock->l_callback_timeout);
 210                 mod_timer(&waiting_locks_timer, timeout_rounded);
 211         }
 212         spin_unlock_bh(&waiting_locks_spinlock);
 213 }
 214
 215 /*
 216  * Indicate that we're waiting for a client to call us back cancelling a given
 217  * lock.  We add it to the pending-callback chain, and schedule the lock-timeout
 218  * timer to fire appropriately.  (We round up to the next second, to avoid
 219  * floods of timer firings during periods of high lock contention and traffic).
 220  */
 221 static int ldlm_add_waiting_lock(struct ldlm_lock *lock)
 222 {
 223         unsigned long timeout_rounded;
 224
 225         spin_lock_bh(&waiting_locks_spinlock);
 226         if (!list_empty(&lock->l_pending_chain)) {
 227                 LDLM_DEBUG(lock, "not re-adding to wait list");
 228                 spin_unlock_bh(&waiting_locks_spinlock);
 229                 return 0;
 230         }
 231         LDLM_DEBUG(lock, "adding to wait list");
 232
 233         lock->l_callback_timeout = jiffies + (obd_timeout * HZ / 2);
 234
 235         timeout_rounded = round_timeout(lock->l_callback_timeout);
 236
 237         if (timeout_rounded < waiting_locks_timer.expires ||
 238             !timer_pending(&waiting_locks_timer)) {
 239                 mod_timer(&waiting_locks_timer, timeout_rounded);
 240         }
 241         list_add_tail(&lock->l_pending_chain, &waiting_locks_list); /* FIFO */
 242         spin_unlock_bh(&waiting_locks_spinlock);
 243         return 1;
 244 }
 245
 246 /*
 247  * Remove a lock from the pending list, likely because it had its cancellation
 248  * callback arrive without incident.  This adjusts the lock-timeout timer if
 249  * needed.  Returns 0 if the lock wasn't pending after all, 1 if it was.
 250  */
 251 int ldlm_del_waiting_lock(struct ldlm_lock *lock)
 252 {
 253         struct list_head *list_next;
 254
 255         if (lock->l_export == NULL) {
 256                 /* We don't have a "waiting locks list" on clients. */
 257                 LDLM_DEBUG(lock, "client lock: no-op");
 258                 return 0;
 259         }
 260
 261         spin_lock_bh(&waiting_locks_spinlock);
 262
 263         if (list_empty(&lock->l_pending_chain)) {
 264                 spin_unlock_bh(&waiting_locks_spinlock);
 265                 LDLM_DEBUG(lock, "wasn't waiting");
 266                 return 0;
 267         }
 268
 269         list_next = lock->l_pending_chain.next;
 270         if (lock->l_pending_chain.prev == &waiting_locks_list) {
 271                 /* Removing the head of the list, adjust timer. */
 272                 if (list_next == &waiting_locks_list) {
 273                         /* No more, just cancel. */
 274                         del_timer(&waiting_locks_timer);
 275                 } else {
 276                         struct ldlm_lock *next;
 277                         next = list_entry(list_next, struct ldlm_lock,
 278                                           l_pending_chain);
 279                         mod_timer(&waiting_locks_timer,
 280                                   round_timeout(next->l_callback_timeout));
 281                 }
 282         }
 283
 284         spin_lock_bh(&expired_lock_thread.elt_lock);
 285         list_del_init(&lock->l_pending_chain);
 286         spin_unlock_bh(&expired_lock_thread.elt_lock);
 287
 288         spin_unlock_bh(&waiting_locks_spinlock);
 289         LDLM_DEBUG(lock, "removed");
 290         return 1;
 291 }
 292
 293 #else /* !__KERNEL__ */
 294
 295 static int ldlm_add_waiting_lock(struct ldlm_lock *lock)
 296 {
 297         RETURN(1);
 298 }
 299
 300 int ldlm_del_waiting_lock(struct ldlm_lock *lock)
 301 {
 302         RETURN(0);
 303 }
 304
 305 #endif /* __KERNEL__ */
 306
 307 static void ldlm_failed_ast(struct ldlm_lock *lock, int rc,const char *ast_type)
 308 {
 309         const struct ptlrpc_connection *conn = lock->l_export->exp_connection;
 310         char str[PTL_NALFMT_SIZE];
 311
 312         LDLM_ERROR(lock, "%s AST failed (%d): evicting client %s@%s NID "LPX64
 313                    " (%s)", ast_type, rc, lock->l_export->exp_client_uuid.uuid,
 314                    conn->c_remote_uuid.uuid, conn->c_peer.peer_nid,
 315                    portals_nid2str(conn->c_peer.peer_ni->pni_number,
 316                                    conn->c_peer.peer_nid, str));
 317         ptlrpc_fail_export(lock->l_export);
 318 }
 319
 320 static int ldlm_handle_ast_error(struct ldlm_lock *lock,
 321                                  struct ptlrpc_request *req, int rc,
 322                                  const char *ast_type)
 323 {
 324         if (rc == -ETIMEDOUT || rc == -EINTR || rc == -ENOTCONN) {
 325                 LASSERT(lock->l_export);
 326                 if (lock->l_export->exp_libclient) {
 327                         LDLM_DEBUG(lock, "%s AST to liblustre client (nid "
 328                                    LPU64") timeout, just cancelling lock",
 329                                    ast_type, req->rq_peer.peer_nid);
 330                         ldlm_lock_cancel(lock);
 331                         rc = -ERESTART;
 332                 } else {
 333                         ldlm_del_waiting_lock(lock);
 334                         ldlm_failed_ast(lock, rc, ast_type);
 335                 }
 336         } else if (rc) {
 337                 if (rc == -EINVAL)
 338                         LDLM_DEBUG(lock, "client (nid "LPU64") returned %d"
 339                                    " from %s AST - normal race",
 340                                    req->rq_peer.peer_nid,
 341                                    req->rq_repmsg->status, ast_type);
 342                 else
 343                         LDLM_ERROR(lock, "client (nid "LPU64") returned %d "
 344                                    "from %s AST", req->rq_peer.peer_nid,
 345                                    (req->rq_repmsg != NULL) ?
 346                                    req->rq_repmsg->status : 0, ast_type);
 347                 ldlm_lock_cancel(lock);
 348                 /* Server-side AST functions are called from ldlm_reprocess_all,
 349                  * which needs to be told to please restart its reprocessing. */
 350                 rc = -ERESTART;
 351         }
 352
 353         return rc;
 354 }
 355
 356 int ldlm_server_blocking_ast(struct ldlm_lock *lock,
 357                              struct ldlm_lock_desc *desc,
 358                              void *data, int flag)
 359 {
 360         struct ldlm_request *body;
 361         struct ptlrpc_request *req;
 362         int rc = 0, size = sizeof(*body);
 363         ENTRY;
 364
 365         if (flag == LDLM_CB_CANCELING) {
 366                 /* Don't need to do anything here. */
 367                 RETURN(0);
 368         }
 369
 370         LASSERT(lock);
 371
 372         l_lock(&lock->l_resource->lr_namespace->ns_lock);
 373         if (lock->l_granted_mode != lock->l_req_mode) {
 374                 /* this blocking AST will be communicated as part of the
 375                  * completion AST instead */
 376                 l_unlock(&lock->l_resource->lr_namespace->ns_lock);
 377                 LDLM_DEBUG(lock, "lock not granted, not sending blocking AST");
 378                 RETURN(0);
 379         }
 380
 381         if (lock->l_destroyed) {
 382                 /* What's the point? */
 383                 l_unlock(&lock->l_resource->lr_namespace->ns_lock);
 384                 RETURN(0);
 385         }
 386
 387 #if 0
 388         if (LTIME_S(CURRENT_TIME) - lock->l_export->exp_last_request_time > 30){
 389                 ldlm_failed_ast(lock, -ETIMEDOUT, "Not-attempted blocking");
 390                 l_unlock(&lock->l_resource->lr_namespace->ns_lock);
 391                 RETURN(-ETIMEDOUT);
 392         }
 393 #endif
 394
 395         req = ptlrpc_prep_req(lock->l_export->exp_imp_reverse,
 396                               LDLM_BL_CALLBACK, 1, &size, NULL);
 397         if (req == NULL) {
 398                 l_unlock(&lock->l_resource->lr_namespace->ns_lock);
 399                 RETURN(-ENOMEM);
 400         }
 401
 402         body = lustre_msg_buf(req->rq_reqmsg, 0, sizeof (*body));
 403         memcpy(&body->lock_handle1, &lock->l_remote_handle,
 404                sizeof(body->lock_handle1));
 405         memcpy(&body->lock_desc, desc, sizeof(*desc));
 406         body->lock_flags |= (lock->l_flags & LDLM_AST_FLAGS);
 407
 408         LDLM_DEBUG(lock, "server preparing blocking AST");
 409         req->rq_replen = lustre_msg_size(0, NULL);
 410
 411         if (lock->l_granted_mode == lock->l_req_mode)
 412                 ldlm_add_waiting_lock(lock);
 413         l_unlock(&lock->l_resource->lr_namespace->ns_lock);
 414
 415         req->rq_send_state = LUSTRE_IMP_FULL;
 416         req->rq_timeout = ldlm_timeout; /* timeout for initial AST reply */
 417         rc = ptlrpc_queue_wait(req);
 418         if (rc != 0)
 419                 rc = ldlm_handle_ast_error(lock, req, rc, "blocking");
 420
 421         ptlrpc_req_finished(req);
 422
 423         RETURN(rc);
 424 }
 425
 426 /* XXX copied from ptlrpc/service.c */
 427 static long timeval_sub(struct timeval *large, struct timeval *small)
 428 {
 429         return (large->tv_sec - small->tv_sec) * 1000000 +
 430                 (large->tv_usec - small->tv_usec);
 431 }
 432
 433 int ldlm_server_completion_ast(struct ldlm_lock *lock, int flags, void *data)
 434 {
 435         struct ldlm_request *body;
 436         struct ptlrpc_request *req;
 437         struct timeval granted_time;
 438         long total_enqueue_wait;
 439         int rc = 0, size[2] = {sizeof(*body)}, buffers = 1;
 440         ENTRY;
 441
 442         LASSERT(lock != NULL);
 443
 444         do_gettimeofday(&granted_time);
 445         total_enqueue_wait = timeval_sub(&granted_time, &lock->l_enqueued_time);
 446
 447         if (total_enqueue_wait / 1000000 > obd_timeout)
 448                 LDLM_ERROR(lock, "enqueue wait took %luus", total_enqueue_wait);
 449
 450         down(&lock->l_resource->lr_lvb_sem);
 451         if (lock->l_resource->lr_lvb_len) {
 452                 buffers = 2;
 453                 size[1] = lock->l_resource->lr_lvb_len;
 454         }
 455         up(&lock->l_resource->lr_lvb_sem);
 456
 457         req = ptlrpc_prep_req(lock->l_export->exp_imp_reverse,
 458                               LDLM_CP_CALLBACK, buffers, size, NULL);
 459         if (req == NULL)
 460                 RETURN(-ENOMEM);
 461
 462         body = lustre_msg_buf(req->rq_reqmsg, 0, sizeof (*body));
 463         memcpy(&body->lock_handle1, &lock->l_remote_handle,
 464                sizeof(body->lock_handle1));
 465         body->lock_flags = flags;
 466         ldlm_lock2desc(lock, &body->lock_desc);
 467
 468         if (buffers == 2) {
 469                 void *lvb;
 470
 471                 down(&lock->l_resource->lr_lvb_sem);
 472                 lvb = lustre_msg_buf(req->rq_reqmsg, 1,
 473                                      lock->l_resource->lr_lvb_len);
 474                 memcpy(lvb, lock->l_resource->lr_lvb_data,
 475                        lock->l_resource->lr_lvb_len);
 476                 up(&lock->l_resource->lr_lvb_sem);
 477         }
 478
 479         LDLM_DEBUG(lock, "server preparing completion AST (after %ldus wait)",
 480                    total_enqueue_wait);
 481         req->rq_replen = lustre_msg_size(0, NULL);
 482
 483         req->rq_send_state = LUSTRE_IMP_FULL;
 484         req->rq_timeout = ldlm_timeout; /* timeout for initial AST reply */
 485
 486         /* We only send real blocking ASTs after the lock is granted */
 487         l_lock(&lock->l_resource->lr_namespace->ns_lock);
 488         if (lock->l_flags & LDLM_FL_AST_SENT) {
 489                 body->lock_flags |= LDLM_FL_AST_SENT;
 490                 ldlm_add_waiting_lock(lock); /* start the lock-timeout clock */
 491         }
 492         l_unlock(&lock->l_resource->lr_namespace->ns_lock);
 493
 494         rc = ptlrpc_queue_wait(req);
 495         if (rc != 0)
 496                 rc = ldlm_handle_ast_error(lock, req, rc, "completion");
 497
 498         ptlrpc_req_finished(req);
 499
 500         RETURN(rc);
 501 }
 502
 503 int ldlm_server_glimpse_ast(struct ldlm_lock *lock, void *data)
 504 {
 505         struct ldlm_resource *res = lock->l_resource;
 506         struct ldlm_request *body;
 507         struct ptlrpc_request *req;
 508         int rc = 0, size = sizeof(*body);
 509         ENTRY;
 510
 511         LASSERT(lock != NULL);
 512
 513         req = ptlrpc_prep_req(lock->l_export->exp_imp_reverse,
 514                               LDLM_GL_CALLBACK, 1, &size, NULL);
 515         if (req == NULL)
 516                 RETURN(-ENOMEM);
 517
 518         body = lustre_msg_buf(req->rq_reqmsg, 0, sizeof(*body));
 519         memcpy(&body->lock_handle1, &lock->l_remote_handle,
 520                sizeof(body->lock_handle1));
 521         ldlm_lock2desc(lock, &body->lock_desc);
 522
 523         down(&lock->l_resource->lr_lvb_sem);
 524         size = lock->l_resource->lr_lvb_len;
 525         up(&lock->l_resource->lr_lvb_sem);
 526         req->rq_replen = lustre_msg_size(1, &size);
 527
 528         req->rq_send_state = LUSTRE_IMP_FULL;
 529         req->rq_timeout = ldlm_timeout; /* timeout for initial AST reply */
 530
 531         rc = ptlrpc_queue_wait(req);
 532         if (rc == -ELDLM_NO_LOCK_DATA)
 533                 LDLM_DEBUG(lock, "lost race - client has a lock but no inode");
 534         else if (rc != 0)
 535                 rc = ldlm_handle_ast_error(lock, req, rc, "glimpse");
 536         else
 537                 rc = res->lr_namespace->ns_lvbo->lvbo_update
 538                         (res, req->rq_repmsg, 0, 1);
 539         ptlrpc_req_finished(req);
 540         RETURN(rc);
 541 }
 542
 543 int ldlm_handle_enqueue(struct ptlrpc_request *req,
 544                         ldlm_completion_callback completion_callback,
 545                         ldlm_blocking_callback blocking_callback,
 546                         ldlm_glimpse_callback glimpse_callback)
 547 {
 548         struct obd_device *obddev = req->rq_export->exp_obd;
 549         struct ldlm_reply *dlm_rep;
 550         struct ldlm_request *dlm_req;
 551         int rc = 0, size[2] = {sizeof(*dlm_rep)};
 552         __u32 flags;
 553         ldlm_error_t err = ELDLM_OK;
 554         struct ldlm_lock *lock = NULL;
 555         void *cookie = NULL;
 556         ENTRY;
 557
 558         LDLM_DEBUG_NOLOCK("server-side enqueue handler START");
 559
 560         dlm_req = lustre_swab_reqbuf (req, 0, sizeof (*dlm_req),
 561                                       lustre_swab_ldlm_request);
 562         if (dlm_req == NULL) {
 563                 CERROR ("Can't unpack dlm_req\n");
 564                 GOTO(out, rc = -EFAULT);
 565         }
 566
 567         flags = dlm_req->lock_flags;
 568
 569         /* The lock's callback data might be set in the policy function */
 570         lock = ldlm_lock_create(obddev->obd_namespace, &dlm_req->lock_handle2,
 571                                 dlm_req->lock_desc.l_resource.lr_name,
 572                                 dlm_req->lock_desc.l_resource.lr_type,
 573                                 dlm_req->lock_desc.l_req_mode,
 574                                 blocking_callback, completion_callback,
 575                                 glimpse_callback, NULL, 0);
 576         if (!lock)
 577                 GOTO(out, rc = -ENOMEM);
 578
 579         do_gettimeofday(&lock->l_enqueued_time);
 580         memcpy(&lock->l_remote_handle, &dlm_req->lock_handle1,
 581                sizeof(lock->l_remote_handle));
 582         LDLM_DEBUG(lock, "server-side enqueue handler, new lock created");
 583
 584         LASSERT(req->rq_export);
 585
 586         OBD_FAIL_TIMEOUT(OBD_FAIL_LDLM_ENQUEUE_BLOCKED, obd_timeout * 2);
 587         l_lock(&lock->l_resource->lr_namespace->ns_lock);
 588         if (req->rq_export->exp_failed) {
 589                 LDLM_ERROR(lock,"lock on destroyed export %p\n",req->rq_export);
 590                 l_unlock(&lock->l_resource->lr_namespace->ns_lock);
 591                 GOTO(out, rc = -ENOTCONN);
 592         }
 593         lock->l_export = class_export_get(req->rq_export);
 594         list_add(&lock->l_export_chain,
 595                  &lock->l_export->exp_ldlm_data.led_held_locks);
 596         l_unlock(&lock->l_resource->lr_namespace->ns_lock);
 597
 598         if (flags & LDLM_FL_HAS_INTENT) {
 599                 /* In this case, the reply buffer is allocated deep in
 600                  * local_lock_enqueue by the policy function. */
 601                 cookie = req;
 602         } else {
 603                 int buffers = 1;
 604
 605                 down(&lock->l_resource->lr_lvb_sem);
 606                 if (lock->l_resource->lr_lvb_len) {
 607                         size[1] = lock->l_resource->lr_lvb_len;
 608                         buffers = 2;
 609                 }
 610                 up(&lock->l_resource->lr_lvb_sem);
 611
 612                 if (OBD_FAIL_CHECK_ONCE(OBD_FAIL_LDLM_ENQUEUE_EXTENT_ERR))
 613                         GOTO(out, rc = -ENOMEM);
 614
 615                 rc = lustre_pack_reply(req, buffers, size, NULL);
 616                 if (rc)
 617                         GOTO(out, rc);
 618         }
 619
 620         if (dlm_req->lock_desc.l_resource.lr_type != LDLM_PLAIN)
 621                 memcpy(&lock->l_policy_data, &dlm_req->lock_desc.l_policy_data,
 622                        sizeof(ldlm_policy_data_t));
 623         if (dlm_req->lock_desc.l_resource.lr_type == LDLM_EXTENT)
 624                 memcpy(&lock->l_req_extent, &lock->l_policy_data.l_extent,
 625                        sizeof(lock->l_req_extent));
 626
 627         err = ldlm_lock_enqueue(obddev->obd_namespace, &lock, cookie, &flags);
 628         if (err)
 629                 GOTO(out, err);
 630
 631         dlm_rep = lustre_msg_buf(req->rq_repmsg, 0, sizeof (*dlm_rep));
 632         dlm_rep->lock_flags = flags;
 633
 634         ldlm_lock2desc(lock, &dlm_rep->lock_desc);
 635         ldlm_lock2handle(lock, &dlm_rep->lock_handle);
 636
 637         /* We never send a blocking AST until the lock is granted, but
 638          * we can tell it right now */
 639         l_lock(&lock->l_resource->lr_namespace->ns_lock);
 640         if (lock->l_flags & LDLM_FL_AST_SENT) {
 641                 dlm_rep->lock_flags |= LDLM_FL_AST_SENT;
 642                 if (lock->l_granted_mode == lock->l_req_mode)
 643                         ldlm_add_waiting_lock(lock);
 644         }
 645         l_unlock(&lock->l_resource->lr_namespace->ns_lock);
 646
 647         EXIT;
 648  out:
 649         req->rq_status = err;
 650         if (req->rq_reply_state == NULL) {
 651                 err = lustre_pack_reply(req, 0, NULL, NULL);
 652                 if (rc == 0)
 653                         rc = err;
 654                 req->rq_status = rc;
 655         }
 656
 657         /* The LOCK_CHANGED code in ldlm_lock_enqueue depends on this
 658          * ldlm_reprocess_all.  If this moves, revisit that code. -phil */
 659         if (lock) {
 660                 LDLM_DEBUG(lock, "server-side enqueue handler, sending reply"
 661                            "(err=%d, rc=%d)", err, rc);
 662
 663                 if (rc == 0) {
 664                         down(&lock->l_resource->lr_lvb_sem);
 665                         size[1] = lock->l_resource->lr_lvb_len;
 666                         if (size[1] > 0) {
 667                                 void *lvb = lustre_msg_buf(req->rq_repmsg,
 668                                                            1, size[1]);
 669                                 LASSERTF(lvb != NULL, "req %p, lock %p\n",
 670                                          req, lock);
 671
 672                                 memcpy(lvb, lock->l_resource->lr_lvb_data,
 673                                        size[1]);
 674                         }
 675                         up(&lock->l_resource->lr_lvb_sem);
 676                 } else {
 677                         ldlm_lock_destroy(lock);
 678                 }
 679
 680                 if (!err && dlm_req->lock_desc.l_resource.lr_type != LDLM_FLOCK)
 681                         ldlm_reprocess_all(lock->l_resource);
 682                 LDLM_LOCK_PUT(lock);
 683         }
 684         LDLM_DEBUG_NOLOCK("server-side enqueue handler END (lock %p, rc %d)",
 685                           lock, rc);
 686
 687         return rc;
 688 }
 689
 690 int ldlm_handle_convert(struct ptlrpc_request *req)
 691 {
 692         struct ldlm_request *dlm_req;
 693         struct ldlm_reply *dlm_rep;
 694         struct ldlm_lock *lock;
 695         int rc, size = sizeof(*dlm_rep);
 696         ENTRY;
 697
 698         dlm_req = lustre_swab_reqbuf (req, 0, sizeof (*dlm_req),
 699                                       lustre_swab_ldlm_request);
 700         if (dlm_req == NULL) {
 701                 CERROR ("Can't unpack dlm_req\n");
 702                 RETURN (-EFAULT);
 703         }
 704
 705         rc = lustre_pack_reply(req, 1, &size, NULL);
 706         if (rc) {
 707                 CERROR("out of memory\n");
 708                 RETURN(-ENOMEM);
 709         }
 710         dlm_rep = lustre_msg_buf(req->rq_repmsg, 0, sizeof (*dlm_rep));
 711         dlm_rep->lock_flags = dlm_req->lock_flags;
 712
 713         lock = ldlm_handle2lock(&dlm_req->lock_handle1);
 714         if (!lock) {
 715                 req->rq_status = EINVAL;
 716         } else {
 717                 LDLM_DEBUG(lock, "server-side convert handler START");
 718                 ldlm_lock_convert(lock, dlm_req->lock_desc.l_req_mode,
 719                                   &dlm_rep->lock_flags);
 720                 if (ldlm_del_waiting_lock(lock))
 721                         CDEBUG(D_DLMTRACE, "converted waiting lock %p\n", lock);
 722                 req->rq_status = 0;
 723         }
 724
 725         if (lock) {
 726                 ldlm_reprocess_all(lock->l_resource);
 727                 LDLM_DEBUG(lock, "server-side convert handler END");
 728                 LDLM_LOCK_PUT(lock);
 729         } else
 730                 LDLM_DEBUG_NOLOCK("server-side convert handler END");
 731
 732         RETURN(0);
 733 }
 734
 735 int ldlm_handle_cancel(struct ptlrpc_request *req)
 736 {
 737         struct ldlm_request *dlm_req;
 738         struct ldlm_lock *lock;
 739         struct ldlm_resource *res;
 740         char str[PTL_NALFMT_SIZE];
 741         int rc;
 742         ENTRY;
 743
 744         dlm_req = lustre_swab_reqbuf(req, 0, sizeof (*dlm_req),
 745                                       lustre_swab_ldlm_request);
 746         if (dlm_req == NULL) {
 747                 CERROR("bad request buffer for cancel\n");
 748                 RETURN(-EFAULT);
 749         }
 750
 751         rc = lustre_pack_reply(req, 0, NULL, NULL);
 752         if (rc) {
 753                 CERROR("out of memory\n");
 754                 RETURN(-ENOMEM);
 755         }
 756
 757         lock = ldlm_handle2lock(&dlm_req->lock_handle1);
 758         if (!lock) {
 759                 CERROR("received cancel for unknown lock cookie "LPX64
 760                        " from client %s nid "LPX64" (%s)\n",
 761                        dlm_req->lock_handle1.cookie,
 762                        req->rq_export->exp_client_uuid.uuid,
 763                        req->rq_peer.peer_nid,
 764                        portals_nid2str(req->rq_peer.peer_ni->pni_number,
 765                                        req->rq_peer.peer_nid, str));
 766                 LDLM_DEBUG_NOLOCK("server-side cancel handler stale lock "
 767                                   "(cookie "LPU64")",
 768                                   dlm_req->lock_handle1.cookie);
 769                 req->rq_status = ESTALE;
 770         } else {
 771                 LDLM_DEBUG(lock, "server-side cancel handler START");
 772                 res = lock->l_resource;
 773                 if (res && res->lr_namespace->ns_lvbo &&
 774                     res->lr_namespace->ns_lvbo->lvbo_update) {
 775                         (void)res->lr_namespace->ns_lvbo->lvbo_update
 776                                 (res, NULL, 0, 0);
 777                                 //(res, req->rq_reqmsg, 1);
 778                 }
 779
 780                 ldlm_lock_cancel(lock);
 781                 if (ldlm_del_waiting_lock(lock))
 782                         CDEBUG(D_DLMTRACE, "cancelled waiting lock %p\n", lock);
 783                 req->rq_status = rc;
 784         }
 785
 786         if (ptlrpc_reply(req) != 0)
 787                 LBUG();
 788
 789         if (lock) {
 790                 ldlm_reprocess_all(lock->l_resource);
 791                 LDLM_DEBUG(lock, "server-side cancel handler END");
 792                 LDLM_LOCK_PUT(lock);
 793         }
 794
 795         RETURN(0);
 796 }
 797
 798 void ldlm_handle_bl_callback(struct ldlm_namespace *ns,
 799                              struct ldlm_lock_desc *ld, struct ldlm_lock *lock)
 800 {
 801         int do_ast;
 802         ENTRY;
 803
 804         l_lock(&ns->ns_lock);
 805         LDLM_DEBUG(lock, "client blocking AST callback handler START");
 806
 807         lock->l_flags |= LDLM_FL_CBPENDING;
 808         do_ast = (!lock->l_readers && !lock->l_writers);
 809
 810         if (do_ast) {
 811                 LDLM_DEBUG(lock, "already unused, calling "
 812                            "callback (%p)", lock->l_blocking_ast);
 813                 if (lock->l_blocking_ast != NULL) {
 814                         l_unlock(&ns->ns_lock);
 815                         l_check_no_ns_lock(ns);
 816                         lock->l_blocking_ast(lock, ld, lock->l_ast_data,
 817                                              LDLM_CB_BLOCKING);
 818                         l_lock(&ns->ns_lock);
 819                 }
 820         } else {
 821                 LDLM_DEBUG(lock, "Lock still has references, will be"
 822                            " cancelled later");
 823         }
 824
 825         LDLM_DEBUG(lock, "client blocking callback handler END");
 826         l_unlock(&ns->ns_lock);
 827         LDLM_LOCK_PUT(lock);
 828         EXIT;
 829 }
 830
 831 static void ldlm_handle_cp_callback(struct ptlrpc_request *req,
 832                                     struct ldlm_namespace *ns,
 833                                     struct ldlm_request *dlm_req,
 834                                     struct ldlm_lock *lock)
 835 {
 836         LIST_HEAD(ast_list);
 837         ENTRY;
 838
 839         l_lock(&ns->ns_lock);
 840         LDLM_DEBUG(lock, "client completion callback handler START");
 841
 842         /* If we receive the completion AST before the actual enqueue returned,
 843          * then we might need to switch lock modes, resources, or extents. */
 844         if (dlm_req->lock_desc.l_granted_mode != lock->l_req_mode) {
 845                 lock->l_req_mode = dlm_req->lock_desc.l_granted_mode;
 846                 LDLM_DEBUG(lock, "completion AST, new lock mode");
 847         }
 848
 849         if (lock->l_resource->lr_type != LDLM_PLAIN) {
 850                 memcpy(&lock->l_policy_data, &dlm_req->lock_desc.l_policy_data,
 851                        sizeof(lock->l_policy_data));
 852                 LDLM_DEBUG(lock, "completion AST, new policy data");
 853         }
 854
 855         ldlm_resource_unlink_lock(lock);
 856         if (memcmp(&dlm_req->lock_desc.l_resource.lr_name,
 857                    &lock->l_resource->lr_name,
 858                    sizeof(lock->l_resource->lr_name)) != 0) {
 859                 ldlm_lock_change_resource(ns, lock,
 860                                          dlm_req->lock_desc.l_resource.lr_name);
 861                 LDLM_DEBUG(lock, "completion AST, new resource");
 862         }
 863
 864         if (dlm_req->lock_flags & LDLM_FL_AST_SENT) {
 865                 lock->l_flags |= LDLM_FL_CBPENDING;
 866                 LDLM_DEBUG(lock, "completion AST includes blocking AST");
 867         }
 868
 869         if (lock->l_lvb_len) {
 870                 void *lvb;
 871                 lvb = lustre_swab_reqbuf(req, 1, lock->l_lvb_len,
 872                                          lock->l_lvb_swabber);
 873                 if (lvb == NULL) {
 874                         LDLM_ERROR(lock, "completion AST did not contain "
 875                                    "expected LVB!");
 876                 } else {
 877                         memcpy(lock->l_lvb_data, lvb, lock->l_lvb_len);
 878                 }
 879         }
 880
 881         lock->l_resource->lr_tmp = &ast_list;
 882         ldlm_grant_lock(lock, req, sizeof(*req), 1);
 883         lock->l_resource->lr_tmp = NULL;
 884         LDLM_DEBUG(lock, "callback handler finished, about to run_ast_work");
 885         l_unlock(&ns->ns_lock);
 886         LDLM_LOCK_PUT(lock);
 887
 888         ldlm_run_ast_work(ns, &ast_list);
 889
 890         LDLM_DEBUG_NOLOCK("client completion callback handler END (lock %p)",
 891                           lock);
 892         EXIT;
 893 }
 894
 895 static void ldlm_handle_gl_callback(struct ptlrpc_request *req,
 896                                     struct ldlm_namespace *ns,
 897                                     struct ldlm_request *dlm_req,
 898                                     struct ldlm_lock *lock)
 899 {
 900         int rc = -ENOSYS;
 901         ENTRY;
 902
 903         l_lock(&ns->ns_lock);
 904         LDLM_DEBUG(lock, "client glimpse AST callback handler");
 905
 906         if (lock->l_glimpse_ast != NULL) {
 907                 l_unlock(&ns->ns_lock);
 908                 l_check_no_ns_lock(ns);
 909                 rc = lock->l_glimpse_ast(lock, req);
 910                 l_lock(&ns->ns_lock);
 911         }
 912
 913         if (req->rq_repmsg != NULL) {
 914                 ptlrpc_reply(req);
 915         } else {
 916                 req->rq_status = rc;
 917                 ptlrpc_error(req);
 918         }
 919
 920         l_unlock(&ns->ns_lock);
 921         if (lock->l_granted_mode == LCK_PW &&
 922             !lock->l_readers && !lock->l_writers &&
 923             time_after(jiffies, lock->l_last_used + 10 * HZ)) {
 924                 if (ldlm_bl_to_thread(ns, NULL, lock))
 925                         ldlm_handle_bl_callback(ns, NULL, lock);
 926                 EXIT;
 927                 return;
 928         }
 929
 930         LDLM_LOCK_PUT(lock);
 931         EXIT;
 932 }
 933
 934 static int ldlm_callback_reply(struct ptlrpc_request *req, int rc)
 935 {
 936         req->rq_status = rc;
 937         if (req->rq_reply_state == NULL) {
 938                 rc = lustre_pack_reply(req, 0, NULL, NULL);
 939                 if (rc)
 940                         return rc;
 941         }
 942         return ptlrpc_reply(req);
 943 }
 944
 945 int ldlm_bl_to_thread(struct ldlm_namespace *ns, struct ldlm_lock_desc *ld,
 946                       struct ldlm_lock *lock)
 947 {
 948 #ifdef __KERNEL__
 949         struct ldlm_bl_pool *blp = ldlm_state->ldlm_bl_pool;
 950         struct ldlm_bl_work_item *blwi;
 951         ENTRY;
 952
 953         OBD_ALLOC(blwi, sizeof(*blwi));
 954         if (blwi == NULL)
 955                 RETURN(-ENOMEM);
 956
 957         blwi->blwi_ns = ns;
 958         if (ld != NULL)
 959                 blwi->blwi_ld = *ld;
 960         blwi->blwi_lock = lock;
 961
 962         spin_lock(&blp->blp_lock);
 963         list_add_tail(&blwi->blwi_entry, &blp->blp_list);
 964         wake_up(&blp->blp_waitq);
 965         spin_unlock(&blp->blp_lock);
 966
 967         RETURN(0);
 968 #else
 969         RETURN(-ENOSYS);
 970 #endif
 971 }
 972
 973 static int ldlm_callback_handler(struct ptlrpc_request *req)
 974 {
 975         struct ldlm_namespace *ns;
 976         struct ldlm_request *dlm_req;
 977         struct ldlm_lock *lock;
 978         char str[PTL_NALFMT_SIZE];
 979         int rc;
 980         ENTRY;
 981
 982         /* Requests arrive in sender's byte order.  The ptlrpc service
 983          * handler has already checked and, if necessary, byte-swapped the
 984          * incoming request message body, but I am responsible for the
 985          * message buffers. */
 986
 987         if (req->rq_export == NULL) {
 988                 struct ldlm_request *dlm_req;
 989
 990                 CDEBUG(D_RPCTRACE, "operation %d from nid "LPX64" (%s) with bad "
 991                        "export cookie "LPX64" (ptl req %d/rep %d); this is "
 992                        "normal if this node rebooted with a lock held\n",
 993                        req->rq_reqmsg->opc, req->rq_peer.peer_nid,
 994                        portals_nid2str(req->rq_peer.peer_ni->pni_number,
 995                                        req->rq_peer.peer_nid, str),
 996                        req->rq_reqmsg->handle.cookie,
 997                        req->rq_request_portal, req->rq_reply_portal);
 998
 999                 dlm_req = lustre_swab_reqbuf(req, 0, sizeof (*dlm_req),
1000                                              lustre_swab_ldlm_request);
1001                 if (dlm_req != NULL)
1002                         CDEBUG(D_RPCTRACE, "--> lock cookie: "LPX64"\n",
1003                                dlm_req->lock_handle1.cookie);
1004
1005                 ldlm_callback_reply(req, -ENOTCONN);
1006                 RETURN(0);
1007         }
1008
1009         LASSERT(req->rq_export != NULL);
1010         LASSERT(req->rq_export->exp_obd != NULL);
1011
1012         switch(req->rq_reqmsg->opc) {
1013         case LDLM_BL_CALLBACK:
1014                 OBD_FAIL_RETURN(OBD_FAIL_LDLM_BL_CALLBACK, 0);
1015                 break;
1016         case LDLM_CP_CALLBACK:
1017                 OBD_FAIL_RETURN(OBD_FAIL_LDLM_CP_CALLBACK, 0);
1018                 break;
1019         case LDLM_GL_CALLBACK:
1020                 OBD_FAIL_RETURN(OBD_FAIL_LDLM_GL_CALLBACK, 0);
1021                 break;
1022         case OBD_LOG_CANCEL:
1023                 OBD_FAIL_RETURN(OBD_FAIL_OBD_LOG_CANCEL_NET, 0);
1024                 rc = llog_origin_handle_cancel(req);
1025                 ldlm_callback_reply(req, rc);
1026                 RETURN(0);
1027         case LLOG_ORIGIN_HANDLE_CREATE:
1028                 OBD_FAIL_RETURN(OBD_FAIL_OBD_LOGD_NET, 0);
1029                 rc = llog_origin_handle_create(req);
1030                 ldlm_callback_reply(req, rc);
1031                 RETURN(0);
1032         case LLOG_ORIGIN_HANDLE_NEXT_BLOCK:
1033                 OBD_FAIL_RETURN(OBD_FAIL_OBD_LOGD_NET, 0);
1034                 rc = llog_origin_handle_next_block(req);
1035                 ldlm_callback_reply(req, rc);
1036                 RETURN(0);
1037         case LLOG_ORIGIN_HANDLE_READ_HEADER:
1038                 OBD_FAIL_RETURN(OBD_FAIL_OBD_LOGD_NET, 0);
1039                 rc = llog_origin_handle_read_header(req);
1040                 ldlm_callback_reply(req, rc);
1041                 RETURN(0);
1042         case LLOG_ORIGIN_HANDLE_CLOSE:
1043                 OBD_FAIL_RETURN(OBD_FAIL_OBD_LOGD_NET, 0);
1044                 rc = llog_origin_handle_close(req);
1045                 ldlm_callback_reply(req, rc);
1046                 RETURN(0);
1047         default:
1048                 CERROR("unknown opcode %u\n", req->rq_reqmsg->opc);
1049                 ldlm_callback_reply(req, -EPROTO);
1050                 RETURN(0);
1051         }
1052
1053         ns = req->rq_export->exp_obd->obd_namespace;
1054         LASSERT(ns != NULL);
1055
1056         dlm_req = lustre_swab_reqbuf (req, 0, sizeof (*dlm_req),
1057                                       lustre_swab_ldlm_request);
1058         if (dlm_req == NULL) {
1059                 CERROR ("can't unpack dlm_req\n");
1060                 ldlm_callback_reply (req, -EPROTO);
1061                 RETURN (0);
1062         }
1063
1064         lock = ldlm_handle2lock_ns(ns, &dlm_req->lock_handle1);
1065         if (!lock) {
1066                 CDEBUG(D_INODE, "callback on lock "LPX64" - lock disappeared\n",
1067                        dlm_req->lock_handle1.cookie);
1068                 ldlm_callback_reply(req, -EINVAL);
1069                 RETURN(0);
1070         }
1071
1072         /* Copy hints/flags (e.g. LDLM_FL_DISCARD_DATA) from AST. */
1073         lock->l_flags |= (dlm_req->lock_flags & LDLM_AST_FLAGS);
1074
1075         /* We want the ost thread to get this reply so that it can respond
1076          * to ost requests (write cache writeback) that might be triggered
1077          * in the callback.
1078          *
1079          * But we'd also like to be able to indicate in the reply that we're
1080          * cancelling right now, because it's unused, or have an intent result
1081          * in the reply, so we might have to push the responsibility for sending
1082          * the reply down into the AST handlers, alas. */
1083
1084         switch (req->rq_reqmsg->opc) {
1085         case LDLM_BL_CALLBACK:
1086                 CDEBUG(D_INODE, "blocking ast\n");
1087                 ldlm_callback_reply(req, 0);
1088                 if (ldlm_bl_to_thread(ns, &dlm_req->lock_desc, lock))
1089                         ldlm_handle_bl_callback(ns, &dlm_req->lock_desc, lock);
1090                 break;
1091         case LDLM_CP_CALLBACK:
1092                 CDEBUG(D_INODE, "completion ast\n");
1093                 ldlm_callback_reply(req, 0);
1094                 ldlm_handle_cp_callback(req, ns, dlm_req, lock);
1095                 break;
1096         case LDLM_GL_CALLBACK:
1097                 CDEBUG(D_INODE, "glimpse ast\n");
1098                 ldlm_handle_gl_callback(req, ns, dlm_req, lock);
1099                 break;
1100         default:
1101                 LBUG();                         /* checked above */
1102         }
1103
1104         RETURN(0);
1105 }
1106
1107 static int ldlm_cancel_handler(struct ptlrpc_request *req)
1108 {
1109         int rc;
1110         ENTRY;
1111
1112         /* Requests arrive in sender's byte order.  The ptlrpc service
1113          * handler has already checked and, if necessary, byte-swapped the
1114          * incoming request message body, but I am responsible for the
1115          * message buffers. */
1116
1117         if (req->rq_export == NULL) {
1118                 struct ldlm_request *dlm_req;
1119                 CERROR("operation %d with bad export (ptl req %d/rep %d)\n",
1120                        req->rq_reqmsg->opc, req->rq_request_portal,
1121                        req->rq_reply_portal);
1122                 CERROR("--> export cookie: "LPX64"\n",
1123                        req->rq_reqmsg->handle.cookie);
1124                 dlm_req = lustre_swab_reqbuf(req, 0, sizeof (*dlm_req),
1125                                              lustre_swab_ldlm_request);
1126                 if (dlm_req != NULL)
1127                         ldlm_lock_dump_handle(D_ERROR, &dlm_req->lock_handle1);
1128                 ldlm_callback_reply(req, -ENOTCONN);
1129                 RETURN(0);
1130         }
1131
1132         switch (req->rq_reqmsg->opc) {
1133
1134         /* XXX FIXME move this back to mds/handler.c, bug 249 */
1135         case LDLM_CANCEL:
1136                 CDEBUG(D_INODE, "cancel\n");
1137                 OBD_FAIL_RETURN(OBD_FAIL_LDLM_CANCEL, 0);
1138                 rc = ldlm_handle_cancel(req);
1139                 if (rc)
1140                         break;
1141                 RETURN(0);
1142
1143         default:
1144                 CERROR("invalid opcode %d\n", req->rq_reqmsg->opc);
1145                 ldlm_callback_reply(req, -EINVAL);
1146         }
1147
1148         RETURN(0);
1149 }
1150
1151 #ifdef __KERNEL__
1152 static struct ldlm_bl_work_item *ldlm_bl_get_work(struct ldlm_bl_pool *blp)
1153 {
1154         struct ldlm_bl_work_item *blwi = NULL;
1155
1156         spin_lock(&blp->blp_lock);
1157         if (!list_empty(&blp->blp_list)) {
1158                 blwi = list_entry(blp->blp_list.next, struct ldlm_bl_work_item,
1159                                   blwi_entry);
1160                 list_del(&blwi->blwi_entry);
1161         }
1162         spin_unlock(&blp->blp_lock);
1163
1164         return blwi;
1165 }
1166
1167 struct ldlm_bl_thread_data {
1168         int                     bltd_num;
1169         struct ldlm_bl_pool     *bltd_blp;
1170 };
1171
1172 static int ldlm_bl_thread_main(void *arg)
1173 {
1174         struct ldlm_bl_thread_data *bltd = arg;
1175         struct ldlm_bl_pool *blp = bltd->bltd_blp;
1176         unsigned long flags;
1177         ENTRY;
1178
1179         /* XXX boiler-plate */
1180         {
1181                 char name[sizeof(current->comm)];
1182                 snprintf(name, sizeof(name) - 1, "ldlm_bl_%02d",
1183                          bltd->bltd_num);
1184                 kportal_daemonize(name);
1185         }
1186         SIGNAL_MASK_LOCK(current, flags);
1187         sigfillset(&current->blocked);
1188         RECALC_SIGPENDING;
1189         SIGNAL_MASK_UNLOCK(current, flags);
1190
1191         atomic_inc(&blp->blp_num_threads);
1192         complete(&blp->blp_comp);
1193
1194         while(1) {
1195                 struct l_wait_info lwi = { 0 };
1196                 struct ldlm_bl_work_item *blwi = NULL;
1197
1198                 l_wait_event_exclusive(blp->blp_waitq,
1199                                        (blwi = ldlm_bl_get_work(blp)) != NULL,
1200                                        &lwi);
1201
1202                 if (blwi->blwi_ns == NULL)
1203                         break;
1204
1205                 ldlm_handle_bl_callback(blwi->blwi_ns, &blwi->blwi_ld,
1206                                         blwi->blwi_lock);
1207                 OBD_FREE(blwi, sizeof(*blwi));
1208         }
1209
1210         atomic_dec(&blp->blp_num_threads);
1211         complete(&blp->blp_comp);
1212         RETURN(0);
1213 }
1214
1215 #endif
1216
1217 static int ldlm_setup(void);
1218 static int ldlm_cleanup(int force);
1219
1220 int ldlm_get_ref(void)
1221 {
1222         int rc = 0;
1223         down(&ldlm_ref_sem);
1224         if (++ldlm_refcount == 1) {
1225                 rc = ldlm_setup();
1226                 if (rc)
1227                         ldlm_refcount--;
1228         }
1229         up(&ldlm_ref_sem);
1230
1231         RETURN(rc);
1232 }
1233
1234 void ldlm_put_ref(int force)
1235 {
1236         down(&ldlm_ref_sem);
1237         if (ldlm_refcount == 1) {
1238                 int rc = ldlm_cleanup(force);
1239                 if (rc)
1240                         CERROR("ldlm_cleanup failed: %d\n", rc);
1241                 else
1242                         ldlm_refcount--;
1243         } else {
1244                 ldlm_refcount--;
1245         }
1246         up(&ldlm_ref_sem);
1247
1248         EXIT;
1249 }
1250
1251 static int ldlm_setup(void)
1252 {
1253         struct ldlm_bl_pool *blp;
1254         int rc = 0;
1255 #ifdef __KERNEL__
1256         int i;
1257 #endif
1258         ENTRY;
1259
1260         if (ldlm_state != NULL)
1261                 RETURN(-EALREADY);
1262
1263         OBD_ALLOC(ldlm_state, sizeof(*ldlm_state));
1264         if (ldlm_state == NULL)
1265                 RETURN(-ENOMEM);
1266
1267 #ifdef __KERNEL__
1268         rc = ldlm_proc_setup();
1269         if (rc != 0)
1270                 GOTO(out_free, rc);
1271 #endif
1272
1273         ldlm_state->ldlm_cb_service =
1274                 ptlrpc_init_svc(LDLM_NBUFS, LDLM_BUFSIZE, LDLM_MAXREQSIZE,
1275                                 LDLM_CB_REQUEST_PORTAL, LDLM_CB_REPLY_PORTAL,
1276                                 ldlm_callback_handler, "ldlm_cbd",
1277                                 ldlm_svc_proc_dir);
1278
1279         if (!ldlm_state->ldlm_cb_service) {
1280                 CERROR("failed to start service\n");
1281                 GOTO(out_proc, rc = -ENOMEM);
1282         }
1283
1284         ldlm_state->ldlm_cancel_service =
1285                 ptlrpc_init_svc(LDLM_NBUFS, LDLM_BUFSIZE, LDLM_MAXREQSIZE,
1286                                 LDLM_CANCEL_REQUEST_PORTAL,
1287                                 LDLM_CANCEL_REPLY_PORTAL,
1288                                 ldlm_cancel_handler, "ldlm_canceld",
1289                                 ldlm_svc_proc_dir);
1290
1291         if (!ldlm_state->ldlm_cancel_service) {
1292                 CERROR("failed to start service\n");
1293                 GOTO(out_proc, rc = -ENOMEM);
1294         }
1295
1296         OBD_ALLOC(blp, sizeof(*blp));
1297         if (blp == NULL)
1298                 GOTO(out_proc, rc = -ENOMEM);
1299         ldlm_state->ldlm_bl_pool = blp;
1300
1301         atomic_set(&blp->blp_num_threads, 0);
1302         init_waitqueue_head(&blp->blp_waitq);
1303         spin_lock_init(&blp->blp_lock);
1304
1305         INIT_LIST_HEAD(&blp->blp_list);
1306
1307 #ifdef __KERNEL__
1308         for (i = 0; i < LDLM_NUM_THREADS; i++) {
1309                 struct ldlm_bl_thread_data bltd = {
1310                         .bltd_num = i,
1311                         .bltd_blp = blp,
1312                 };
1313                 init_completion(&blp->blp_comp);
1314                 rc = kernel_thread(ldlm_bl_thread_main, &bltd, 0);
1315                 if (rc < 0) {
1316                         CERROR("cannot start LDLM thread #%d: rc %d\n", i, rc);
1317                         GOTO(out_thread, rc);
1318                 }
1319                 wait_for_completion(&blp->blp_comp);
1320         }
1321
1322         rc = ptlrpc_start_n_threads(NULL, ldlm_state->ldlm_cancel_service,
1323                                     LDLM_NUM_THREADS, "ldlm_cn");
1324         if (rc)
1325                 GOTO(out_thread, rc);
1326
1327         rc = ptlrpc_start_n_threads(NULL, ldlm_state->ldlm_cb_service,
1328                                     LDLM_NUM_THREADS, "ldlm_cb");
1329         if (rc)
1330                 GOTO(out_thread, rc);
1331
1332         INIT_LIST_HEAD(&expired_lock_thread.elt_expired_locks);
1333         spin_lock_init(&expired_lock_thread.elt_lock);
1334         expired_lock_thread.elt_state = ELT_STOPPED;
1335         init_waitqueue_head(&expired_lock_thread.elt_waitq);
1336
1337         rc = kernel_thread(expired_lock_main, NULL, CLONE_VM | CLONE_FS);
1338         if (rc < 0) {
1339                 CERROR("Cannot start ldlm expired-lock thread: %d\n", rc);
1340                 GOTO(out_thread, rc);
1341         }
1342
1343         wait_event(expired_lock_thread.elt_waitq,
1344                    expired_lock_thread.elt_state == ELT_READY);
1345
1346         INIT_LIST_HEAD(&waiting_locks_list);
1347         spin_lock_init(&waiting_locks_spinlock);
1348         waiting_locks_timer.function = waiting_locks_callback;
1349         waiting_locks_timer.data = 0;
1350         init_timer(&waiting_locks_timer);
1351 #endif
1352
1353         RETURN(0);
1354
1355 #ifdef __KERNEL__
1356  out_thread:
1357         ptlrpc_unregister_service(ldlm_state->ldlm_cancel_service);
1358         ptlrpc_unregister_service(ldlm_state->ldlm_cb_service);
1359 #endif
1360
1361  out_proc:
1362 #ifdef __KERNEL__
1363         ldlm_proc_cleanup();
1364  out_free:
1365 #endif
1366         OBD_FREE(ldlm_state, sizeof(*ldlm_state));
1367         ldlm_state = NULL;
1368         return rc;
1369 }
1370
1371 static int ldlm_cleanup(int force)
1372 {
1373 #ifdef __KERNEL__
1374         struct ldlm_bl_pool *blp = ldlm_state->ldlm_bl_pool;
1375 #endif
1376         ENTRY;
1377
1378         if (!list_empty(&ldlm_namespace_list)) {
1379                 CERROR("ldlm still has namespaces; clean these up first.\n");
1380                 ldlm_dump_all_namespaces();
1381                 RETURN(-EBUSY);
1382         }
1383
1384 #ifdef __KERNEL__
1385         while (atomic_read(&blp->blp_num_threads) > 0) {
1386                 struct ldlm_bl_work_item blwi = { .blwi_ns = NULL };
1387
1388                 init_completion(&blp->blp_comp);
1389
1390                 spin_lock(&blp->blp_lock);
1391                 list_add_tail(&blwi.blwi_entry, &blp->blp_list);
1392                 wake_up(&blp->blp_waitq);
1393                 spin_unlock(&blp->blp_lock);
1394
1395                 wait_for_completion(&blp->blp_comp);
1396         }
1397         OBD_FREE(blp, sizeof(*blp));
1398
1399         ptlrpc_stop_all_threads(ldlm_state->ldlm_cb_service);
1400         ptlrpc_unregister_service(ldlm_state->ldlm_cb_service);
1401         ptlrpc_stop_all_threads(ldlm_state->ldlm_cancel_service);
1402         ptlrpc_unregister_service(ldlm_state->ldlm_cancel_service);
1403         ldlm_proc_cleanup();
1404
1405         expired_lock_thread.elt_state = ELT_TERMINATE;
1406         wake_up(&expired_lock_thread.elt_waitq);
1407         wait_event(expired_lock_thread.elt_waitq,
1408                    expired_lock_thread.elt_state == ELT_STOPPED);
1409
1410 #endif
1411
1412         OBD_FREE(ldlm_state, sizeof(*ldlm_state));
1413         ldlm_state = NULL;
1414
1415         RETURN(0);
1416 }
1417
1418 int __init ldlm_init(void)
1419 {
1420         ldlm_resource_slab = kmem_cache_create("ldlm_resources",
1421                                                sizeof(struct ldlm_resource), 0,
1422                                                SLAB_HWCACHE_ALIGN, NULL, NULL);
1423         if (ldlm_resource_slab == NULL)
1424                 return -ENOMEM;
1425
1426         ldlm_lock_slab = kmem_cache_create("ldlm_locks",
1427                                            sizeof(struct ldlm_lock), 0,
1428                                            SLAB_HWCACHE_ALIGN, NULL, NULL);
1429         if (ldlm_lock_slab == NULL) {
1430                 kmem_cache_destroy(ldlm_resource_slab);
1431                 return -ENOMEM;
1432         }
1433
1434         l_lock_init(&ldlm_handle_lock);
1435
1436         return 0;
1437 }
1438
1439 void __exit ldlm_exit(void)
1440 {
1441         if ( ldlm_refcount )
1442                 CERROR("ldlm_refcount is %d in ldlm_exit!\n", ldlm_refcount);
1443         LASSERTF(kmem_cache_destroy(ldlm_resource_slab) == 0,
1444                  "couldn't free ldlm resource slab\n");
1445         LASSERTF(kmem_cache_destroy(ldlm_lock_slab) == 0,
1446                  "couldn't free ldlm lock slab\n");
1447 }
1448
1449 /* ldlm_flock.c */
1450 EXPORT_SYMBOL(ldlm_flock_completion_ast);
1451
1452 /* ldlm_extent.c */
1453 EXPORT_SYMBOL(ldlm_extent_shift_kms);
1454
1455 /* ldlm_lock.c */
1456 EXPORT_SYMBOL(ldlm_get_processing_policy);
1457 EXPORT_SYMBOL(ldlm_lock2desc);
1458 EXPORT_SYMBOL(ldlm_register_intent);
1459 EXPORT_SYMBOL(ldlm_lockname);
1460 EXPORT_SYMBOL(ldlm_typename);
1461 EXPORT_SYMBOL(ldlm_lock2handle);
1462 EXPORT_SYMBOL(__ldlm_handle2lock);
1463 EXPORT_SYMBOL(ldlm_lock_get);
1464 EXPORT_SYMBOL(ldlm_lock_put);
1465 EXPORT_SYMBOL(ldlm_lock_match);
1466 EXPORT_SYMBOL(ldlm_lock_cancel);
1467 EXPORT_SYMBOL(ldlm_lock_addref);
1468 EXPORT_SYMBOL(ldlm_lock_decref);
1469 EXPORT_SYMBOL(ldlm_lock_decref_and_cancel);
1470 EXPORT_SYMBOL(ldlm_lock_change_resource);
1471 EXPORT_SYMBOL(ldlm_lock_set_data);
1472 EXPORT_SYMBOL(ldlm_it2str);
1473 EXPORT_SYMBOL(ldlm_lock_dump);
1474 EXPORT_SYMBOL(ldlm_lock_dump_handle);
1475 EXPORT_SYMBOL(ldlm_cancel_locks_for_export);
1476 EXPORT_SYMBOL(ldlm_reprocess_all_ns);
1477 EXPORT_SYMBOL(ldlm_lock_allow_match);
1478
1479 /* ldlm_request.c */
1480 EXPORT_SYMBOL(ldlm_completion_ast);
1481 EXPORT_SYMBOL(ldlm_expired_completion_wait);
1482 EXPORT_SYMBOL(ldlm_cli_convert);
1483 EXPORT_SYMBOL(ldlm_cli_enqueue);
1484 EXPORT_SYMBOL(ldlm_cli_cancel);
1485 EXPORT_SYMBOL(ldlm_cli_cancel_unused);
1486 EXPORT_SYMBOL(ldlm_replay_locks);
1487 EXPORT_SYMBOL(ldlm_resource_foreach);
1488 EXPORT_SYMBOL(ldlm_namespace_foreach);
1489 EXPORT_SYMBOL(ldlm_namespace_foreach_res);
1490 EXPORT_SYMBOL(ldlm_change_cbdata);
1491
1492 /* ldlm_lockd.c */
1493 EXPORT_SYMBOL(ldlm_server_blocking_ast);
1494 EXPORT_SYMBOL(ldlm_server_completion_ast);
1495 EXPORT_SYMBOL(ldlm_server_glimpse_ast);
1496 EXPORT_SYMBOL(ldlm_handle_enqueue);
1497 EXPORT_SYMBOL(ldlm_handle_cancel);
1498 EXPORT_SYMBOL(ldlm_handle_convert);
1499 EXPORT_SYMBOL(ldlm_del_waiting_lock);
1500 EXPORT_SYMBOL(ldlm_get_ref);
1501 EXPORT_SYMBOL(ldlm_put_ref);
1502
1503 #if 0
1504 /* ldlm_test.c */
1505 EXPORT_SYMBOL(ldlm_test);
1506 EXPORT_SYMBOL(ldlm_regression_start);
1507 EXPORT_SYMBOL(ldlm_regression_stop);
1508 #endif
1509
1510 /* ldlm_resource.c */
1511 EXPORT_SYMBOL(ldlm_namespace_new);
1512 EXPORT_SYMBOL(ldlm_namespace_cleanup);
1513 EXPORT_SYMBOL(ldlm_namespace_free);
1514 EXPORT_SYMBOL(ldlm_namespace_dump);
1515 EXPORT_SYMBOL(ldlm_dump_all_namespaces);
1516 EXPORT_SYMBOL(ldlm_resource_get);
1517 EXPORT_SYMBOL(ldlm_resource_putref);
1518
1519 /* l_lock.c */
1520 EXPORT_SYMBOL(l_lock);
1521 EXPORT_SYMBOL(l_unlock);
1522
1523 /* ldlm_lib.c */
1524 EXPORT_SYMBOL(client_obd_setup);
1525 EXPORT_SYMBOL(client_obd_cleanup);
1526 EXPORT_SYMBOL(client_connect_import);
1527 EXPORT_SYMBOL(client_disconnect_export);
1528 EXPORT_SYMBOL(target_abort_recovery);
1529 EXPORT_SYMBOL(target_cleanup_recovery);
1530 EXPORT_SYMBOL(target_handle_connect);
1531 EXPORT_SYMBOL(target_destroy_export);
1532 EXPORT_SYMBOL(target_cancel_recovery_timer);
1533 EXPORT_SYMBOL(target_send_reply);
1534 EXPORT_SYMBOL(target_queue_recovery_request);
1535 EXPORT_SYMBOL(target_handle_ping);
1536 EXPORT_SYMBOL(target_handle_disconnect);
1537 EXPORT_SYMBOL(target_queue_final_reply);