lustre/osp/osp_precreate.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) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
  28  * Use is subject to license terms.
  29  *
  30  * Copyright (c) 2012, 2013, 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/osp/osp_sync.c
  37  *
  38  * Lustre OST Proxy Device
  39  *
  40  * Author: Alex Zhuravlev <alexey.zhuravlev@intel.com>
  41  * Author: Mikhail Pershin <mike.pershin@intel.com>
  42  * Author: Di Wang <di.wang@intel.com>
  43  */
  44
  45 #define DEBUG_SUBSYSTEM S_MDS
  46
  47 #include "osp_internal.h"
  48
  49 /*
  50  * there are two specific states to take care about:
  51  *
  52  * = import is disconnected =
  53  *
  54  * = import is inactive =
  55  *   in this case osp_declare_object_create() returns an error
  56  *
  57  */
  58
  59 /*
  60  * statfs
  61  */
  62 static inline int osp_statfs_need_update(struct osp_device *d)
  63 {
  64         return !cfs_time_before(cfs_time_current(),
  65                                 d->opd_statfs_fresh_till);
  66 }
  67
  68 /*
  69  * OSP tries to maintain pool of available objects so that calls to create
  70  * objects don't block most of time
  71  *
  72  * each time OSP gets connected to OST, we should start from precreation cleanup
  73  */
  74 static inline bool osp_precreate_running(struct osp_device *d)
  75 {
  76         return !!(d->opd_pre_thread.t_flags & SVC_RUNNING);
  77 }
  78
  79 static inline bool osp_precreate_stopped(struct osp_device *d)
  80 {
  81         return !!(d->opd_pre_thread.t_flags & SVC_STOPPED);
  82 }
  83
  84 static void osp_statfs_timer_cb(unsigned long _d)
  85 {
  86         struct osp_device *d = (struct osp_device *) _d;
  87
  88         LASSERT(d);
  89         if (d->opd_pre != NULL && osp_precreate_running(d))
  90                 wake_up(&d->opd_pre_waitq);
  91 }
  92
  93 static int osp_statfs_interpret(const struct lu_env *env,
  94                                 struct ptlrpc_request *req,
  95                                 union ptlrpc_async_args *aa, int rc)
  96 {
  97         struct obd_import       *imp = req->rq_import;
  98         struct obd_statfs       *msfs;
  99         struct osp_device       *d;
 100
 101         ENTRY;
 102
 103         aa = ptlrpc_req_async_args(req);
 104         d = aa->pointer_arg[0];
 105         LASSERT(d);
 106
 107         if (rc != 0)
 108                 GOTO(out, rc);
 109
 110         msfs = req_capsule_server_get(&req->rq_pill, &RMF_OBD_STATFS);
 111         if (msfs == NULL)
 112                 GOTO(out, rc = -EPROTO);
 113
 114         d->opd_statfs = *msfs;
 115
 116         osp_pre_update_status(d, rc);
 117
 118         /* schedule next update */
 119         d->opd_statfs_fresh_till = cfs_time_shift(d->opd_statfs_maxage);
 120         cfs_timer_arm(&d->opd_statfs_timer, d->opd_statfs_fresh_till);
 121         d->opd_statfs_update_in_progress = 0;
 122
 123         CDEBUG(D_CACHE, "updated statfs %p\n", d);
 124
 125         RETURN(0);
 126 out:
 127         /* couldn't update statfs, try again as soon as possible */
 128         if (d->opd_pre != NULL && osp_precreate_running(d))
 129                 wake_up(&d->opd_pre_waitq);
 130
 131         if (req->rq_import_generation == imp->imp_generation)
 132                 CDEBUG(D_CACHE, "%s: couldn't update statfs: rc = %d\n",
 133                        d->opd_obd->obd_name, rc);
 134         RETURN(rc);
 135 }
 136
 137 static int osp_statfs_update(struct osp_device *d)
 138 {
 139         struct ptlrpc_request   *req;
 140         struct obd_import       *imp;
 141         union ptlrpc_async_args *aa;
 142         int                      rc;
 143
 144         ENTRY;
 145
 146         CDEBUG(D_CACHE, "going to update statfs\n");
 147
 148         imp = d->opd_obd->u.cli.cl_import;
 149         LASSERT(imp);
 150
 151         req = ptlrpc_request_alloc(imp, &RQF_OST_STATFS);
 152         if (req == NULL)
 153                 RETURN(-ENOMEM);
 154
 155         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_STATFS);
 156         if (rc) {
 157                 ptlrpc_request_free(req);
 158                 RETURN(rc);
 159         }
 160         ptlrpc_request_set_replen(req);
 161         req->rq_request_portal = OST_CREATE_PORTAL;
 162         ptlrpc_at_set_req_timeout(req);
 163
 164         req->rq_interpret_reply = (ptlrpc_interpterer_t)osp_statfs_interpret;
 165         aa = ptlrpc_req_async_args(req);
 166         aa->pointer_arg[0] = d;
 167
 168         /*
 169          * no updates till reply
 170          */
 171         cfs_timer_disarm(&d->opd_statfs_timer);
 172         d->opd_statfs_fresh_till = cfs_time_shift(obd_timeout * 1000);
 173         d->opd_statfs_update_in_progress = 1;
 174
 175         ptlrpcd_add_req(req, PDL_POLICY_ROUND, -1);
 176
 177         RETURN(0);
 178 }
 179
 180 /*
 181  * XXX: there might be a case where removed object(s) do not add free
 182  * space (empty object). if the number of such deletions is high, then
 183  * we can start to update statfs too often - a rpc storm
 184  * TODO: some throttling is needed
 185  */
 186 void osp_statfs_need_now(struct osp_device *d)
 187 {
 188         if (!d->opd_statfs_update_in_progress) {
 189                 /*
 190                  * if current status is -ENOSPC (lack of free space on OST)
 191                  * then we should poll OST immediately once object destroy
 192                  * is replied
 193                  */
 194                 d->opd_statfs_fresh_till = cfs_time_shift(-1);
 195                 cfs_timer_disarm(&d->opd_statfs_timer);
 196                 wake_up(&d->opd_pre_waitq);
 197         }
 198 }
 199
 200 static inline int osp_objs_precreated(const struct lu_env *env,
 201                                       struct osp_device *osp)
 202 {
 203         struct lu_fid *fid1 = &osp->opd_pre_last_created_fid;
 204         struct lu_fid *fid2 = &osp->opd_pre_used_fid;
 205
 206         LASSERTF(fid_seq(fid1) == fid_seq(fid2),
 207                  "Created fid"DFID" Next fid "DFID"\n", PFID(fid1), PFID(fid2));
 208
 209         if (fid_is_idif(fid1)) {
 210                 struct ost_id *oi1 = &osp_env_info(env)->osi_oi;
 211                 struct ost_id *oi2 = &osp_env_info(env)->osi_oi2;
 212
 213                 LASSERT(fid_is_idif(fid1) && fid_is_idif(fid2));
 214                 fid_to_ostid(fid1, oi1);
 215                 fid_to_ostid(fid2, oi2);
 216                 LASSERT(ostid_id(oi1) >= ostid_id(oi2));
 217
 218                 return ostid_id(oi1) - ostid_id(oi2);
 219         }
 220
 221         return fid_oid(fid1) - fid_oid(fid2);
 222 }
 223
 224 static inline int osp_precreate_near_empty_nolock(const struct lu_env *env,
 225                                                   struct osp_device *d)
 226 {
 227         int window = osp_objs_precreated(env, d);
 228
 229         /* don't consider new precreation till OST is healty and
 230          * has free space */
 231         return ((window - d->opd_pre_reserved < d->opd_pre_grow_count / 2) &&
 232                 (d->opd_pre_status == 0));
 233 }
 234
 235 static inline int osp_precreate_near_empty(const struct lu_env *env,
 236                                            struct osp_device *d)
 237 {
 238         int rc;
 239
 240         /* XXX: do we really need locking here? */
 241         spin_lock(&d->opd_pre_lock);
 242         rc = osp_precreate_near_empty_nolock(env, d);
 243         spin_unlock(&d->opd_pre_lock);
 244         return rc;
 245 }
 246
 247 static inline int osp_create_end_seq(const struct lu_env *env,
 248                                      struct osp_device *osp)
 249 {
 250         struct lu_fid *fid = &osp->opd_pre_used_fid;
 251         int rc;
 252
 253         spin_lock(&osp->opd_pre_lock);
 254         rc = osp_fid_end_seq(env, fid);
 255         spin_unlock(&osp->opd_pre_lock);
 256         return rc;
 257 }
 258
 259 /**
 260  * Write fid into last_oid/last_seq file.
 261  **/
 262 int osp_write_last_oid_seq_files(struct lu_env *env, struct osp_device *osp,
 263                                  struct lu_fid *fid, int sync)
 264 {
 265         struct osp_thread_info  *oti = osp_env_info(env);
 266         struct lu_buf      *lb_oid = &oti->osi_lb;
 267         struct lu_buf      *lb_oseq = &oti->osi_lb2;
 268         loff_t             oid_off;
 269         loff_t             oseq_off;
 270         struct thandle    *th;
 271         int                   rc;
 272         ENTRY;
 273
 274         /* Note: through f_oid is only 32bits, it will also write
 275          * 64 bits for oid to keep compatiblity with the previous
 276          * version. */
 277         lb_oid->lb_buf = &fid->f_oid;
 278         lb_oid->lb_len = sizeof(obd_id);
 279         oid_off = sizeof(obd_id) * osp->opd_index;
 280
 281         lb_oseq->lb_buf = &fid->f_seq;
 282         lb_oseq->lb_len = sizeof(obd_id);
 283         oseq_off = sizeof(obd_id) * osp->opd_index;
 284
 285         th = dt_trans_create(env, osp->opd_storage);
 286         if (IS_ERR(th))
 287                 RETURN(PTR_ERR(th));
 288
 289         th->th_sync |= sync;
 290         rc = dt_declare_record_write(env, osp->opd_last_used_oid_file,
 291                                      lb_oid->lb_len, oid_off, th);
 292         if (rc != 0)
 293                 GOTO(out, rc);
 294
 295         rc = dt_declare_record_write(env, osp->opd_last_used_seq_file,
 296                                      lb_oseq->lb_len, oseq_off, th);
 297         if (rc != 0)
 298                 GOTO(out, rc);
 299
 300         rc = dt_trans_start_local(env, osp->opd_storage, th);
 301         if (rc != 0)
 302                 GOTO(out, rc);
 303
 304         rc = dt_record_write(env, osp->opd_last_used_oid_file, lb_oid,
 305                              &oid_off, th);
 306         if (rc != 0) {
 307                 CERROR("%s: can not write to last seq file: rc = %d\n",
 308                         osp->opd_obd->obd_name, rc);
 309                 GOTO(out, rc);
 310         }
 311         rc = dt_record_write(env, osp->opd_last_used_seq_file, lb_oseq,
 312                              &oseq_off, th);
 313         if (rc) {
 314                 CERROR("%s: can not write to last seq file: rc = %d\n",
 315                         osp->opd_obd->obd_name, rc);
 316                 GOTO(out, rc);
 317         }
 318 out:
 319         dt_trans_stop(env, osp->opd_storage, th);
 320         RETURN(rc);
 321 }
 322
 323 int osp_precreate_rollover_new_seq(struct lu_env *env, struct osp_device *osp)
 324 {
 325         struct lu_fid   *fid = &osp_env_info(env)->osi_fid;
 326         struct lu_fid   *last_fid = &osp->opd_last_used_fid;
 327         int             rc;
 328         ENTRY;
 329
 330         rc = seq_client_get_seq(env, osp->opd_obd->u.cli.cl_seq, &fid->f_seq);
 331         if (rc != 0) {
 332                 CERROR("%s: alloc fid error: rc = %d\n",
 333                        osp->opd_obd->obd_name, rc);
 334                 RETURN(rc);
 335         }
 336
 337         fid->f_oid = 1;
 338         fid->f_ver = 0;
 339         LASSERTF(fid_seq(fid) != fid_seq(last_fid),
 340                  "fid "DFID", last_fid "DFID"\n", PFID(fid),
 341                  PFID(last_fid));
 342
 343         rc = osp_write_last_oid_seq_files(env, osp, fid, 1);
 344         if (rc != 0) {
 345                 CERROR("%s: Can not update oid/seq file: rc = %d\n",
 346                        osp->opd_obd->obd_name, rc);
 347                 RETURN(rc);
 348         }
 349
 350         LCONSOLE_INFO("%s: update sequence from "LPX64" to "LPX64"\n",
 351                       osp->opd_obd->obd_name, fid_seq(last_fid),
 352                       fid_seq(fid));
 353         /* Update last_xxx to the new seq */
 354         spin_lock(&osp->opd_pre_lock);
 355         osp->opd_last_used_fid = *fid;
 356         osp->opd_gap_start_fid = *fid;
 357         osp->opd_pre_used_fid = *fid;
 358         osp->opd_pre_last_created_fid = *fid;
 359         spin_unlock(&osp->opd_pre_lock);
 360
 361         RETURN(rc);
 362 }
 363
 364 /**
 365  * alloc fids for precreation.
 366  * rc = 0 Success, @grow is the count of real allocation.
 367  * rc = 1 Current seq is used up.
 368  * rc < 0 Other error.
 369  **/
 370 static int osp_precreate_fids(const struct lu_env *env, struct osp_device *osp,
 371                               struct lu_fid *fid, int *grow)
 372 {
 373         struct osp_thread_info  *osi = osp_env_info(env);
 374         __u64                   end;
 375         int                     i = 0;
 376
 377         if (fid_is_idif(fid)) {
 378                 struct lu_fid   *last_fid;
 379                 struct ost_id   *oi = &osi->osi_oi;
 380
 381                 spin_lock(&osp->opd_pre_lock);
 382                 last_fid = &osp->opd_pre_last_created_fid;
 383                 fid_to_ostid(last_fid, oi);
 384                 end = min(ostid_id(oi) + *grow, IDIF_MAX_OID);
 385                 *grow = end - ostid_id(oi);
 386                 ostid_set_id(oi, ostid_id(oi) + *grow);
 387                 spin_unlock(&osp->opd_pre_lock);
 388
 389                 if (*grow == 0)
 390                         return 1;
 391
 392                 ostid_to_fid(fid, oi, osp->opd_index);
 393                 return 0;
 394         }
 395
 396         spin_lock(&osp->opd_pre_lock);
 397         *fid = osp->opd_pre_last_created_fid;
 398         end = fid->f_oid;
 399         end = min((end + *grow), (__u64)LUSTRE_DATA_SEQ_MAX_WIDTH);
 400         *grow = end - fid->f_oid;
 401         fid->f_oid += end - fid->f_oid;
 402         spin_unlock(&osp->opd_pre_lock);
 403
 404         CDEBUG(D_INFO, "Expect %d, actual %d ["DFID" -- "DFID"]\n",
 405                *grow, i, PFID(fid), PFID(&osp->opd_pre_last_created_fid));
 406
 407         return *grow > 0 ? 0 : 1;
 408 }
 409
 410 static int osp_precreate_send(const struct lu_env *env, struct osp_device *d)
 411 {
 412         struct osp_thread_info  *oti = osp_env_info(env);
 413         struct ptlrpc_request   *req;
 414         struct obd_import       *imp;
 415         struct ost_body         *body;
 416         int                      rc, grow, diff;
 417         struct lu_fid           *fid = &oti->osi_fid;
 418         ENTRY;
 419
 420         /* don't precreate new objects till OST healthy and has free space */
 421         if (unlikely(d->opd_pre_status)) {
 422                 CDEBUG(D_INFO, "%s: don't send new precreate: rc = %d\n",
 423                        d->opd_obd->obd_name, d->opd_pre_status);
 424                 RETURN(0);
 425         }
 426
 427         /*
 428          * if not connection/initialization is compeleted, ignore
 429          */
 430         imp = d->opd_obd->u.cli.cl_import;
 431         LASSERT(imp);
 432
 433         req = ptlrpc_request_alloc(imp, &RQF_OST_CREATE);
 434         if (req == NULL)
 435                 RETURN(-ENOMEM);
 436         req->rq_request_portal = OST_CREATE_PORTAL;
 437         /* we should not resend create request - anyway we will have delorphan
 438          * and kill these objects */
 439         req->rq_no_delay = req->rq_no_resend = 1;
 440
 441         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_CREATE);
 442         if (rc) {
 443                 ptlrpc_request_free(req);
 444                 RETURN(rc);
 445         }
 446
 447         spin_lock(&d->opd_pre_lock);
 448         if (d->opd_pre_grow_count > d->opd_pre_max_grow_count / 2)
 449                 d->opd_pre_grow_count = d->opd_pre_max_grow_count / 2;
 450         grow = d->opd_pre_grow_count;
 451         spin_unlock(&d->opd_pre_lock);
 452
 453         body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
 454         LASSERT(body);
 455
 456         *fid = d->opd_pre_last_created_fid;
 457         rc = osp_precreate_fids(env, d, fid, &grow);
 458         if (rc == 1) {
 459                 /* Current seq has been used up*/
 460                 if (!osp_is_fid_client(d)) {
 461                         osp_pre_update_status(d, -ENOSPC);
 462                         rc = -ENOSPC;
 463                 }
 464                 wake_up(&d->opd_pre_waitq);
 465                 GOTO(out_req, rc);
 466         }
 467
 468         if (!osp_is_fid_client(d)) {
 469                 /* Non-FID client will always send seq 0 because of
 470                  * compatiblity */
 471                 LASSERTF(fid_is_idif(fid), "Invalid fid "DFID"\n", PFID(fid));
 472                 fid->f_seq = 0;
 473         }
 474
 475         fid_to_ostid(fid, &body->oa.o_oi);
 476         body->oa.o_valid = OBD_MD_FLGROUP;
 477
 478         ptlrpc_request_set_replen(req);
 479
 480         rc = ptlrpc_queue_wait(req);
 481         if (rc) {
 482                 CERROR("%s: can't precreate: rc = %d\n", d->opd_obd->obd_name,
 483                        rc);
 484                 GOTO(out_req, rc);
 485         }
 486         LASSERT(req->rq_transno == 0);
 487
 488         body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
 489         if (body == NULL)
 490                 GOTO(out_req, rc = -EPROTO);
 491
 492         ostid_to_fid(fid, &body->oa.o_oi, d->opd_index);
 493         LASSERTF(lu_fid_diff(fid, &d->opd_pre_used_fid) > 0,
 494                  "reply fid "DFID" pre used fid "DFID"\n", PFID(fid),
 495                  PFID(&d->opd_pre_used_fid));
 496
 497         diff = lu_fid_diff(fid, &d->opd_pre_last_created_fid);
 498
 499         spin_lock(&d->opd_pre_lock);
 500         if (diff < grow) {
 501                 /* the OST has not managed to create all the
 502                  * objects we asked for */
 503                 d->opd_pre_grow_count = max(diff, OST_MIN_PRECREATE);
 504                 d->opd_pre_grow_slow = 1;
 505         } else {
 506                 /* the OST is able to keep up with the work,
 507                  * we could consider increasing grow_count
 508                  * next time if needed */
 509                 d->opd_pre_grow_slow = 0;
 510         }
 511
 512         d->opd_pre_last_created_fid = *fid;
 513         spin_unlock(&d->opd_pre_lock);
 514
 515         CDEBUG(D_HA, "%s: current precreated pool: "DFID"-"DFID"\n",
 516                d->opd_obd->obd_name, PFID(&d->opd_pre_used_fid),
 517                PFID(&d->opd_pre_last_created_fid));
 518 out_req:
 519         /* now we can wakeup all users awaiting for objects */
 520         osp_pre_update_status(d, rc);
 521         wake_up(&d->opd_pre_user_waitq);
 522
 523         ptlrpc_req_finished(req);
 524         RETURN(rc);
 525 }
 526
 527 static int osp_get_lastfid_from_ost(const struct lu_env *env,
 528                                     struct osp_device *d)
 529 {
 530         struct ptlrpc_request   *req = NULL;
 531         struct obd_import       *imp;
 532         struct lu_fid           *last_fid;
 533         char                    *tmp;
 534         int                     rc;
 535         ENTRY;
 536
 537         imp = d->opd_obd->u.cli.cl_import;
 538         LASSERT(imp);
 539
 540         req = ptlrpc_request_alloc(imp, &RQF_OST_GET_INFO_LAST_FID);
 541         if (req == NULL)
 542                 RETURN(-ENOMEM);
 543
 544         req_capsule_set_size(&req->rq_pill, &RMF_GETINFO_KEY, RCL_CLIENT,
 545                              sizeof(KEY_LAST_FID));
 546
 547         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_GET_INFO);
 548         if (rc) {
 549                 ptlrpc_request_free(req);
 550                 RETURN(rc);
 551         }
 552
 553         tmp = req_capsule_client_get(&req->rq_pill, &RMF_GETINFO_KEY);
 554         memcpy(tmp, KEY_LAST_FID, sizeof(KEY_LAST_FID));
 555
 556         req->rq_no_delay = req->rq_no_resend = 1;
 557         last_fid = req_capsule_client_get(&req->rq_pill, &RMF_FID);
 558         fid_cpu_to_le(last_fid, &d->opd_last_used_fid);
 559
 560         ptlrpc_request_set_replen(req);
 561
 562         rc = ptlrpc_queue_wait(req);
 563         if (rc) {
 564                 /* bad-bad OST.. let sysadm sort this out */
 565                 if (rc == -ENOTSUPP) {
 566                         CERROR("%s: server does not support FID: rc = %d\n",
 567                                d->opd_obd->obd_name, -ENOTSUPP);
 568                 }
 569                 ptlrpc_set_import_active(imp, 0);
 570                 GOTO(out, rc);
 571         }
 572
 573         last_fid = req_capsule_server_get(&req->rq_pill, &RMF_FID);
 574         if (last_fid == NULL) {
 575                 CERROR("%s: Got last_fid failed.\n", d->opd_obd->obd_name);
 576                 GOTO(out, rc = -EPROTO);
 577         }
 578
 579         if (!fid_is_sane(last_fid)) {
 580                 CERROR("%s: Got insane last_fid "DFID"\n",
 581                        d->opd_obd->obd_name, PFID(last_fid));
 582                 GOTO(out, rc = -EPROTO);
 583         }
 584
 585         /* Only update the last used fid, if the OST has objects for
 586          * this sequence, i.e. fid_oid > 0 */
 587         if (fid_oid(last_fid) > 0)
 588                 d->opd_last_used_fid = *last_fid;
 589
 590         CDEBUG(D_HA, "%s: Got last_fid "DFID"\n", d->opd_obd->obd_name,
 591                PFID(last_fid));
 592
 593 out:
 594         ptlrpc_req_finished(req);
 595         RETURN(rc);
 596 }
 597
 598 /**
 599  * asks OST to clean precreate orphans
 600  * and gets next id for new objects
 601  */
 602 static int osp_precreate_cleanup_orphans(struct lu_env *env,
 603                                          struct osp_device *d)
 604 {
 605         struct osp_thread_info  *osi = osp_env_info(env);
 606         struct lu_fid           *last_fid = &osi->osi_fid;
 607         struct ptlrpc_request   *req = NULL;
 608         struct obd_import       *imp;
 609         struct ost_body         *body;
 610         struct l_wait_info       lwi = { 0 };
 611         int                      update_status = 0;
 612         int                      rc;
 613         int                      diff;
 614
 615         ENTRY;
 616
 617         /*
 618          * wait for local recovery to finish, so we can cleanup orphans
 619          * orphans are all objects since "last used" (assigned), but
 620          * there might be objects reserved and in some cases they won't
 621          * be used. we can't cleanup them till we're sure they won't be
 622          * used. also can't we allow new reservations because they may
 623          * end up getting orphans being cleaned up below. so we block
 624          * new reservations and wait till all reserved objects either
 625          * user or released.
 626          */
 627         spin_lock(&d->opd_pre_lock);
 628         d->opd_pre_recovering = 1;
 629         spin_unlock(&d->opd_pre_lock);
 630         /*
 631          * The locking above makes sure the opd_pre_reserved check below will
 632          * catch all osp_precreate_reserve() calls who find
 633          * "!opd_pre_recovering".
 634          */
 635         l_wait_event(d->opd_pre_waitq,
 636                      (!d->opd_pre_reserved && d->opd_recovery_completed) ||
 637                      !osp_precreate_running(d) || d->opd_got_disconnected,
 638                      &lwi);
 639         if (!osp_precreate_running(d) || d->opd_got_disconnected)
 640                 GOTO(out, rc = -EAGAIN);
 641
 642         CDEBUG(D_HA, "%s: going to cleanup orphans since "DFID"\n",
 643                d->opd_obd->obd_name, PFID(&d->opd_last_used_fid));
 644
 645         *last_fid = d->opd_last_used_fid;
 646         /* The OSP should already get the valid seq now */
 647         LASSERT(!fid_is_zero(last_fid));
 648         if (fid_oid(&d->opd_last_used_fid) < 2) {
 649                 /* lastfid looks strange... ask OST */
 650                 rc = osp_get_lastfid_from_ost(env, d);
 651                 if (rc)
 652                         GOTO(out, rc);
 653         }
 654
 655         imp = d->opd_obd->u.cli.cl_import;
 656         LASSERT(imp);
 657
 658         req = ptlrpc_request_alloc(imp, &RQF_OST_CREATE);
 659         if (req == NULL)
 660                 GOTO(out, rc = -ENOMEM);
 661
 662         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_CREATE);
 663         if (rc) {
 664                 ptlrpc_request_free(req);
 665                 req = NULL;
 666                 GOTO(out, rc);
 667         }
 668
 669         body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
 670         if (body == NULL)
 671                 GOTO(out, rc = -EPROTO);
 672
 673         body->oa.o_flags = OBD_FL_DELORPHAN;
 674         body->oa.o_valid = OBD_MD_FLFLAGS | OBD_MD_FLGROUP;
 675
 676         fid_to_ostid(&d->opd_last_used_fid, &body->oa.o_oi);
 677
 678         ptlrpc_request_set_replen(req);
 679
 680         /* Don't resend the delorphan req */
 681         req->rq_no_resend = req->rq_no_delay = 1;
 682
 683         rc = ptlrpc_queue_wait(req);
 684         if (rc) {
 685                 update_status = 1;
 686                 GOTO(out, rc);
 687         }
 688
 689         body = req_capsule_server_get(&req->rq_pill, &RMF_OST_BODY);
 690         if (body == NULL)
 691                 GOTO(out, rc = -EPROTO);
 692
 693         /*
 694          * OST provides us with id new pool starts from in body->oa.o_id
 695          */
 696         ostid_to_fid(last_fid, &body->oa.o_oi, d->opd_index);
 697
 698         spin_lock(&d->opd_pre_lock);
 699         diff = lu_fid_diff(&d->opd_last_used_fid, last_fid);
 700         if (diff > 0) {
 701                 d->opd_pre_grow_count = OST_MIN_PRECREATE + diff;
 702                 d->opd_pre_last_created_fid = d->opd_last_used_fid;
 703         } else {
 704                 d->opd_pre_grow_count = OST_MIN_PRECREATE;
 705                 d->opd_pre_last_created_fid = *last_fid;
 706         }
 707         /*
 708          * This empties the pre-creation pool and effectively blocks any new
 709          * reservations.
 710          */
 711         LASSERT(fid_oid(&d->opd_pre_last_created_fid) <=
 712                 LUSTRE_DATA_SEQ_MAX_WIDTH);
 713         d->opd_pre_used_fid = d->opd_pre_last_created_fid;
 714         d->opd_pre_grow_slow = 0;
 715         spin_unlock(&d->opd_pre_lock);
 716
 717         CDEBUG(D_HA, "%s: Got last_id "DFID" from OST, last_created "DFID
 718                "last_used is "DFID"\n", d->opd_obd->obd_name, PFID(last_fid),
 719                PFID(&d->opd_pre_last_created_fid), PFID(&d->opd_last_used_fid));
 720 out:
 721         if (req)
 722                 ptlrpc_req_finished(req);
 723
 724         spin_lock(&d->opd_pre_lock);
 725         d->opd_pre_recovering = 0;
 726         spin_unlock(&d->opd_pre_lock);
 727
 728         /*
 729          * If rc is zero, the pre-creation window should have been emptied.
 730          * Since waking up the herd would be useless without pre-created
 731          * objects, we defer the signal to osp_precreate_send() in that case.
 732          */
 733         if (rc != 0) {
 734                 if (update_status) {
 735                         CERROR("%s: cannot cleanup orphans: rc = %d\n",
 736                                d->opd_obd->obd_name, rc);
 737                         /* we can't proceed from here, OST seem to
 738                          * be in a bad shape, better to wait for
 739                          * a new instance of the server and repeat
 740                          * from the beginning. notify possible waiters
 741                          * this OSP isn't quite functional yet */
 742                         osp_pre_update_status(d, rc);
 743                 } else {
 744                         wake_up(&d->opd_pre_user_waitq);
 745                 }
 746         }
 747
 748         RETURN(rc);
 749 }
 750
 751 /*
 752  * the function updates current precreation status used: functional or not
 753  *
 754  * rc is a last code from the transport, rc == 0 meaning transport works
 755  * well and users of lod can use objects from this OSP
 756  *
 757  * the status depends on current usage of OST
 758  */
 759 void osp_pre_update_status(struct osp_device *d, int rc)
 760 {
 761         struct obd_statfs       *msfs = &d->opd_statfs;
 762         int                      old = d->opd_pre_status;
 763         __u64                    used;
 764
 765         d->opd_pre_status = rc;
 766         if (rc)
 767                 goto out;
 768
 769         /* Add a bit of hysteresis so this flag isn't continually flapping,
 770          * and ensure that new files don't get extremely fragmented due to
 771          * only a small amount of available space in the filesystem.
 772          * We want to set the NOSPC flag when there is less than ~0.1% free
 773          * and clear it when there is at least ~0.2% free space, so:
 774          *                   avail < ~0.1% max          max = avail + used
 775          *            1025 * avail < avail + used       used = blocks - free
 776          *            1024 * avail < used
 777          *            1024 * avail < blocks - free
 778          *                   avail < ((blocks - free) >> 10)
 779          *
 780          * On very large disk, say 16TB 0.1% will be 16 GB. We don't want to
 781          * lose that amount of space so in those cases we report no space left
 782          * if their is less than 1 GB left.                             */
 783         if (likely(msfs->os_type)) {
 784                 used = min_t(__u64, (msfs->os_blocks - msfs->os_bfree) >> 10,
 785                                     1 << 30);
 786                 if ((msfs->os_ffree < 32) || (msfs->os_bavail < used)) {
 787                         d->opd_pre_status = -ENOSPC;
 788                         if (old != -ENOSPC)
 789                                 CDEBUG(D_INFO, "%s: status: "LPU64" blocks, "
 790                                        LPU64" free, "LPU64" used, "LPU64" "
 791                                        "avail -> %d: rc = %d\n",
 792                                        d->opd_obd->obd_name, msfs->os_blocks,
 793                                        msfs->os_bfree, used, msfs->os_bavail,
 794                                        d->opd_pre_status, rc);
 795                         CDEBUG(D_INFO,
 796                                "non-commited changes: %lu, in progress: %u\n",
 797                                d->opd_syn_changes, d->opd_syn_rpc_in_progress);
 798                 } else if (old == -ENOSPC) {
 799                         d->opd_pre_status = 0;
 800                         spin_lock(&d->opd_pre_lock);
 801                         d->opd_pre_grow_slow = 0;
 802                         d->opd_pre_grow_count = OST_MIN_PRECREATE;
 803                         spin_unlock(&d->opd_pre_lock);
 804                         wake_up(&d->opd_pre_waitq);
 805                         CDEBUG(D_INFO, "%s: no space: "LPU64" blocks, "LPU64
 806                                " free, "LPU64" used, "LPU64" avail -> %d: "
 807                                "rc = %d\n", d->opd_obd->obd_name,
 808                                msfs->os_blocks, msfs->os_bfree, used,
 809                                msfs->os_bavail, d->opd_pre_status, rc);
 810                 }
 811         }
 812
 813 out:
 814         wake_up(&d->opd_pre_user_waitq);
 815 }
 816
 817 int osp_init_pre_fid(struct osp_device *osp)
 818 {
 819         struct lu_env           env;
 820         struct osp_thread_info  *osi;
 821         struct lu_client_seq    *cli_seq;
 822         struct lu_fid           *last_fid;
 823         int                     rc;
 824         ENTRY;
 825
 826         LASSERT(osp->opd_pre != NULL);
 827
 828         /* Return if last_used fid has been initialized */
 829         if (!fid_is_zero(&osp->opd_last_used_fid))
 830                 RETURN(0);
 831
 832         rc = lu_env_init(&env, osp->opd_dt_dev.dd_lu_dev.ld_type->ldt_ctx_tags);
 833         if (rc) {
 834                 CERROR("%s: init env error: rc = %d\n",
 835                        osp->opd_obd->obd_name, rc);
 836                 RETURN(rc);
 837         }
 838
 839         osi = osp_env_info(&env);
 840         last_fid = &osi->osi_fid;
 841         fid_zero(last_fid);
 842         /* For a freshed fs, it will allocate a new sequence first */
 843         if (osp_is_fid_client(osp) && osp->opd_group != 0) {
 844                 cli_seq = osp->opd_obd->u.cli.cl_seq;
 845                 rc = seq_client_get_seq(&env, cli_seq, &last_fid->f_seq);
 846                 if (rc != 0) {
 847                         CERROR("%s: alloc fid error: rc = %d\n",
 848                                osp->opd_obd->obd_name, rc);
 849                         GOTO(out, rc);
 850                 }
 851         } else {
 852                 last_fid->f_seq = fid_idif_seq(0, osp->opd_index);
 853         }
 854         last_fid->f_oid = 1;
 855         last_fid->f_ver = 0;
 856
 857         spin_lock(&osp->opd_pre_lock);
 858         osp->opd_last_used_fid = *last_fid;
 859         osp->opd_pre_used_fid = *last_fid;
 860         osp->opd_pre_last_created_fid = *last_fid;
 861         spin_unlock(&osp->opd_pre_lock);
 862         rc = osp_write_last_oid_seq_files(&env, osp, last_fid, 1);
 863         if (rc != 0) {
 864                 CERROR("%s: write fid error: rc = %d\n",
 865                        osp->opd_obd->obd_name, rc);
 866                 GOTO(out, rc);
 867         }
 868 out:
 869         lu_env_fini(&env);
 870         RETURN(rc);
 871 }
 872
 873 static int osp_precreate_thread(void *_arg)
 874 {
 875         struct osp_device       *d = _arg;
 876         struct ptlrpc_thread    *thread = &d->opd_pre_thread;
 877         struct l_wait_info       lwi = { 0 };
 878         struct lu_env            env;
 879         int                      rc;
 880
 881         ENTRY;
 882
 883         rc = lu_env_init(&env, d->opd_dt_dev.dd_lu_dev.ld_type->ldt_ctx_tags);
 884         if (rc) {
 885                 CERROR("%s: init env error: rc = %d\n", d->opd_obd->obd_name,
 886                        rc);
 887                 RETURN(rc);
 888         }
 889
 890         spin_lock(&d->opd_pre_lock);
 891         thread->t_flags = SVC_RUNNING;
 892         spin_unlock(&d->opd_pre_lock);
 893         wake_up(&thread->t_ctl_waitq);
 894
 895         while (osp_precreate_running(d)) {
 896                 /*
 897                  * need to be connected to OST
 898                  */
 899                 while (osp_precreate_running(d)) {
 900                         l_wait_event(d->opd_pre_waitq,
 901                                      !osp_precreate_running(d) ||
 902                                      d->opd_new_connection,
 903                                      &lwi);
 904
 905                         if (!d->opd_new_connection)
 906                                 continue;
 907
 908                         d->opd_new_connection = 0;
 909                         d->opd_got_disconnected = 0;
 910                         break;
 911                 }
 912
 913                 if (!osp_precreate_running(d))
 914                         break;
 915
 916                 LASSERT(d->opd_obd->u.cli.cl_seq != NULL);
 917                 /* Sigh, fid client is not ready yet */
 918                 if (d->opd_obd->u.cli.cl_seq->lcs_exp == NULL)
 919                         continue;
 920
 921                 /* Init fid for osp_precreate if necessary */
 922                 rc = osp_init_pre_fid(d);
 923                 if (rc != 0) {
 924                         class_export_put(d->opd_exp);
 925                         d->opd_obd->u.cli.cl_seq->lcs_exp = NULL;
 926                         CERROR("%s: init pre fid error: rc = %d\n",
 927                                d->opd_obd->obd_name, rc);
 928                         continue;
 929                 }
 930
 931                 osp_statfs_update(d);
 932
 933                 /*
 934                  * Clean up orphans or recreate missing objects.
 935                  */
 936                 rc = osp_precreate_cleanup_orphans(&env, d);
 937                 if (rc != 0)
 938                         continue;
 939                 /*
 940                  * connected, can handle precreates now
 941                  */
 942                 while (osp_precreate_running(d)) {
 943                         l_wait_event(d->opd_pre_waitq,
 944                                      !osp_precreate_running(d) ||
 945                                      osp_precreate_near_empty(&env, d) ||
 946                                      osp_statfs_need_update(d) ||
 947                                      d->opd_got_disconnected, &lwi);
 948
 949                         if (!osp_precreate_running(d))
 950                                 break;
 951
 952                         /* something happened to the connection
 953                          * have to start from the beginning */
 954                         if (d->opd_got_disconnected)
 955                                 break;
 956
 957                         if (osp_statfs_need_update(d))
 958                                 osp_statfs_update(d);
 959
 960                         /* To avoid handling different seq in precreate/orphan
 961                          * cleanup, it will hold precreate until current seq is
 962                          * used up. */
 963                         if (unlikely(osp_precreate_end_seq(&env, d) &&
 964                             !osp_create_end_seq(&env, d)))
 965                                 continue;
 966
 967                         if (unlikely(osp_precreate_end_seq(&env, d) &&
 968                                      osp_create_end_seq(&env, d))) {
 969                                 LCONSOLE_INFO("%s:"LPX64" is used up."
 970                                               " Update to new seq\n",
 971                                               d->opd_obd->obd_name,
 972                                          fid_seq(&d->opd_pre_last_created_fid));
 973                                 rc = osp_precreate_rollover_new_seq(&env, d);
 974                                 if (rc)
 975                                         continue;
 976                         }
 977
 978                         if (osp_precreate_near_empty(&env, d)) {
 979                                 rc = osp_precreate_send(&env, d);
 980                                 /* osp_precreate_send() sets opd_pre_status
 981                                  * in case of error, that prevent the using of
 982                                  * failed device. */
 983                                 if (rc < 0 && rc != -ENOSPC &&
 984                                     rc != -ETIMEDOUT && rc != -ENOTCONN)
 985                                         CERROR("%s: cannot precreate objects:"
 986                                                " rc = %d\n",
 987                                                d->opd_obd->obd_name, rc);
 988                         }
 989                 }
 990         }
 991
 992         thread->t_flags = SVC_STOPPED;
 993         lu_env_fini(&env);
 994         wake_up(&thread->t_ctl_waitq);
 995
 996         RETURN(0);
 997 }
 998
 999 static int osp_precreate_ready_condition(const struct lu_env *env,
1000                                          struct osp_device *d)
1001 {
1002         if (d->opd_pre_recovering)
1003                 return 0;
1004
1005         /* ready if got enough precreated objects */
1006         /* we need to wait for others (opd_pre_reserved) and our object (+1) */
1007         if (d->opd_pre_reserved + 1 < osp_objs_precreated(env, d))
1008                 return 1;
1009
1010         /* ready if OST reported no space and no destroys in progress */
1011         if (d->opd_syn_changes + d->opd_syn_rpc_in_progress == 0 &&
1012             d->opd_pre_status == -ENOSPC)
1013                 return 1;
1014
1015         /* Bail out I/O fails to OST */
1016         if (d->opd_pre_status != 0 &&
1017             d->opd_pre_status != -EAGAIN &&
1018             d->opd_pre_status != -ENODEV &&
1019             d->opd_pre_status != -ENOSPC) {
1020                 /* DEBUG LU-3230 */
1021                 if (d->opd_pre_status != -EIO)
1022                         CERROR("%s: precreate failed opd_pre_status %d\n",
1023                                d->opd_obd->obd_name, d->opd_pre_status);
1024                 return 1;
1025         }
1026
1027         return 0;
1028 }
1029
1030 static int osp_precreate_timeout_condition(void *data)
1031 {
1032         struct osp_device *d = data;
1033
1034         LCONSOLE_WARN("%s: slow creates, last="DFID", next="DFID", "
1035                       "reserved="LPU64", syn_changes=%lu, "
1036                       "syn_rpc_in_progress=%d, status=%d\n",
1037                       d->opd_obd->obd_name, PFID(&d->opd_pre_last_created_fid),
1038                       PFID(&d->opd_pre_used_fid), d->opd_pre_reserved,
1039                       d->opd_syn_changes, d->opd_syn_rpc_in_progress,
1040                       d->opd_pre_status);
1041
1042         return 1;
1043 }
1044
1045 /*
1046  * called to reserve object in the pool
1047  * return codes:
1048  *  ENOSPC - no space on corresponded OST
1049  *  EAGAIN - precreation is in progress, try later
1050  *  EIO    - no access to OST
1051  */
1052 int osp_precreate_reserve(const struct lu_env *env, struct osp_device *d)
1053 {
1054         struct l_wait_info       lwi;
1055         cfs_time_t               expire = cfs_time_shift(obd_timeout);
1056         int                      precreated, rc;
1057
1058         ENTRY;
1059
1060         LASSERTF(osp_objs_precreated(env, d) >= 0, "Last created FID "DFID
1061                  "Next FID "DFID"\n", PFID(&d->opd_pre_last_created_fid),
1062                  PFID(&d->opd_pre_used_fid));
1063
1064         /*
1065          * wait till:
1066          *  - preallocation is done
1067          *  - no free space expected soon
1068          *  - can't connect to OST for too long (obd_timeout)
1069          *  - OST can allocate fid sequence.
1070          */
1071         while ((rc = d->opd_pre_status) == 0 || rc == -ENOSPC ||
1072                 rc == -ENODEV || rc == -EAGAIN || rc == -ENOTCONN) {
1073
1074                 /*
1075                  * increase number of precreations
1076                  */
1077                 precreated = osp_objs_precreated(env, d);
1078                 if (d->opd_pre_grow_count < d->opd_pre_max_grow_count &&
1079                     d->opd_pre_grow_slow == 0 &&
1080                     precreated <= (d->opd_pre_grow_count / 4 + 1)) {
1081                         spin_lock(&d->opd_pre_lock);
1082                         d->opd_pre_grow_slow = 1;
1083                         d->opd_pre_grow_count *= 2;
1084                         spin_unlock(&d->opd_pre_lock);
1085                 }
1086
1087                 spin_lock(&d->opd_pre_lock);
1088                 precreated = osp_objs_precreated(env, d);
1089                 if (precreated > d->opd_pre_reserved &&
1090                     !d->opd_pre_recovering) {
1091                         d->opd_pre_reserved++;
1092                         spin_unlock(&d->opd_pre_lock);
1093                         rc = 0;
1094
1095                         /* XXX: don't wake up if precreation is in progress */
1096                         if (osp_precreate_near_empty_nolock(env, d) &&
1097                            !osp_precreate_end_seq_nolock(env, d))
1098                                 wake_up(&d->opd_pre_waitq);
1099
1100                         break;
1101                 }
1102                 spin_unlock(&d->opd_pre_lock);
1103
1104                 /*
1105                  * all precreated objects have been used and no-space
1106                  * status leave us no chance to succeed very soon
1107                  * but if there is destroy in progress, then we should
1108                  * wait till that is done - some space might be released
1109                  */
1110                 if (unlikely(rc == -ENOSPC)) {
1111                         if (d->opd_syn_changes) {
1112                                 /* force local commit to release space */
1113                                 dt_commit_async(env, d->opd_storage);
1114                         }
1115                         if (d->opd_syn_rpc_in_progress) {
1116                                 /* just wait till destroys are done */
1117                                 /* see l_wait_even() few lines below */
1118                         }
1119                         if (d->opd_syn_changes +
1120                             d->opd_syn_rpc_in_progress == 0) {
1121                                 /* no hope for free space */
1122                                 break;
1123                         }
1124                 }
1125
1126                 /* XXX: don't wake up if precreation is in progress */
1127                 wake_up(&d->opd_pre_waitq);
1128
1129                 lwi = LWI_TIMEOUT(expire - cfs_time_current(),
1130                                 osp_precreate_timeout_condition, d);
1131                 if (cfs_time_aftereq(cfs_time_current(), expire)) {
1132                         rc = -ETIMEDOUT;
1133                         break;
1134                 }
1135
1136                 l_wait_event(d->opd_pre_user_waitq,
1137                              osp_precreate_ready_condition(env, d), &lwi);
1138         }
1139
1140         RETURN(rc);
1141 }
1142
1143 /*
1144  * this function relies on reservation made before
1145  */
1146 int osp_precreate_get_fid(const struct lu_env *env, struct osp_device *d,
1147                           struct lu_fid *fid)
1148 {
1149         /* grab next id from the pool */
1150         spin_lock(&d->opd_pre_lock);
1151
1152         LASSERTF(lu_fid_diff(&d->opd_pre_used_fid,
1153                              &d->opd_pre_last_created_fid) < 0,
1154                  "next fid "DFID" last created fid "DFID"\n",
1155                  PFID(&d->opd_pre_used_fid),
1156                  PFID(&d->opd_pre_last_created_fid));
1157
1158         d->opd_pre_used_fid.f_oid++;
1159         memcpy(fid, &d->opd_pre_used_fid, sizeof(*fid));
1160         d->opd_pre_reserved--;
1161         /*
1162          * last_used_id must be changed along with getting new id otherwise
1163          * we might miscalculate gap causing object loss or leak
1164          */
1165         osp_update_last_fid(d, fid);
1166         spin_unlock(&d->opd_pre_lock);
1167
1168         /*
1169          * probably main thread suspended orphan cleanup till
1170          * all reservations are released, see comment in
1171          * osp_precreate_thread() just before orphan cleanup
1172          */
1173         if (unlikely(d->opd_pre_reserved == 0 && d->opd_pre_status))
1174                 wake_up(&d->opd_pre_waitq);
1175
1176         return 0;
1177 }
1178
1179 /*
1180  *
1181  */
1182 int osp_object_truncate(const struct lu_env *env, struct dt_object *dt,
1183                         __u64 size)
1184 {
1185         struct osp_device       *d = lu2osp_dev(dt->do_lu.lo_dev);
1186         struct ptlrpc_request   *req = NULL;
1187         struct obd_import       *imp;
1188         struct ost_body         *body;
1189         struct obdo             *oa = NULL;
1190         int                      rc;
1191
1192         ENTRY;
1193
1194         imp = d->opd_obd->u.cli.cl_import;
1195         LASSERT(imp);
1196
1197         req = ptlrpc_request_alloc(imp, &RQF_OST_PUNCH);
1198         if (req == NULL)
1199                 RETURN(-ENOMEM);
1200
1201         /* XXX: capa support? */
1202         /* osc_set_capa_size(req, &RMF_CAPA1, capa); */
1203         rc = ptlrpc_request_pack(req, LUSTRE_OST_VERSION, OST_PUNCH);
1204         if (rc) {
1205                 ptlrpc_request_free(req);
1206                 RETURN(rc);
1207         }
1208
1209         /*
1210          * XXX: decide how do we do here with resend
1211          * if we don't resend, then client may see wrong file size
1212          * if we do resend, then MDS thread can get stuck for quite long
1213          */
1214         req->rq_no_resend = req->rq_no_delay = 1;
1215
1216         req->rq_request_portal = OST_IO_PORTAL; /* bug 7198 */
1217         ptlrpc_at_set_req_timeout(req);
1218
1219         OBD_ALLOC_PTR(oa);
1220         if (oa == NULL)
1221                 GOTO(out, rc = -ENOMEM);
1222
1223         rc = fid_to_ostid(lu_object_fid(&dt->do_lu), &oa->o_oi);
1224         LASSERT(rc == 0);
1225         oa->o_size = size;
1226         oa->o_blocks = OBD_OBJECT_EOF;
1227         oa->o_valid = OBD_MD_FLSIZE | OBD_MD_FLBLOCKS |
1228                       OBD_MD_FLID | OBD_MD_FLGROUP;
1229
1230         body = req_capsule_client_get(&req->rq_pill, &RMF_OST_BODY);
1231         LASSERT(body);
1232         lustre_set_wire_obdo(&req->rq_import->imp_connect_data, &body->oa, oa);
1233
1234         /* XXX: capa support? */
1235         /* osc_pack_capa(req, body, capa); */
1236
1237         ptlrpc_request_set_replen(req);
1238
1239         rc = ptlrpc_queue_wait(req);
1240         if (rc)
1241                 CERROR("can't punch object: %d\n", rc);
1242 out:
1243         ptlrpc_req_finished(req);
1244         if (oa)
1245                 OBD_FREE_PTR(oa);
1246         RETURN(rc);
1247 }
1248
1249 int osp_init_precreate(struct osp_device *d)
1250 {
1251         struct l_wait_info       lwi = { 0 };
1252         struct task_struct              *task;
1253
1254         ENTRY;
1255
1256         OBD_ALLOC_PTR(d->opd_pre);
1257         if (d->opd_pre == NULL)
1258                 RETURN(-ENOMEM);
1259
1260         /* initially precreation isn't ready */
1261         d->opd_pre_status = -EAGAIN;
1262         fid_zero(&d->opd_pre_used_fid);
1263         d->opd_pre_used_fid.f_oid = 1;
1264         fid_zero(&d->opd_pre_last_created_fid);
1265         d->opd_pre_last_created_fid.f_oid = 1;
1266         d->opd_pre_reserved = 0;
1267         d->opd_got_disconnected = 1;
1268         d->opd_pre_grow_slow = 0;
1269         d->opd_pre_grow_count = OST_MIN_PRECREATE;
1270         d->opd_pre_min_grow_count = OST_MIN_PRECREATE;
1271         d->opd_pre_max_grow_count = OST_MAX_PRECREATE;
1272
1273         spin_lock_init(&d->opd_pre_lock);
1274         init_waitqueue_head(&d->opd_pre_waitq);
1275         init_waitqueue_head(&d->opd_pre_user_waitq);
1276         init_waitqueue_head(&d->opd_pre_thread.t_ctl_waitq);
1277
1278         /*
1279          * Initialize statfs-related things
1280          */
1281         d->opd_statfs_maxage = 5; /* default update interval */
1282         d->opd_statfs_fresh_till = cfs_time_shift(-1000);
1283         CDEBUG(D_OTHER, "current %llu, fresh till %llu\n",
1284                (unsigned long long)cfs_time_current(),
1285                (unsigned long long)d->opd_statfs_fresh_till);
1286         cfs_timer_init(&d->opd_statfs_timer, osp_statfs_timer_cb, d);
1287
1288         /*
1289          * start thread handling precreation and statfs updates
1290          */
1291         task = kthread_run(osp_precreate_thread, d,
1292                                "osp-pre-%u", d->opd_index);
1293         if (IS_ERR(task)) {
1294                 CERROR("can't start precreate thread %ld\n", PTR_ERR(task));
1295                 RETURN(PTR_ERR(task));
1296         }
1297
1298         l_wait_event(d->opd_pre_thread.t_ctl_waitq,
1299                      osp_precreate_running(d) || osp_precreate_stopped(d),
1300                      &lwi);
1301
1302         RETURN(0);
1303 }
1304
1305 void osp_precreate_fini(struct osp_device *d)
1306 {
1307         struct ptlrpc_thread *thread;
1308
1309         ENTRY;
1310
1311         cfs_timer_disarm(&d->opd_statfs_timer);
1312
1313         if (d->opd_pre == NULL)
1314                 RETURN_EXIT;
1315
1316         thread = &d->opd_pre_thread;
1317
1318         thread->t_flags = SVC_STOPPING;
1319         wake_up(&d->opd_pre_waitq);
1320
1321         wait_event(thread->t_ctl_waitq, thread->t_flags & SVC_STOPPED);
1322
1323         OBD_FREE_PTR(d->opd_pre);
1324         d->opd_pre = NULL;
1325
1326         EXIT;
1327 }
1328