lustre/osd-zfs/osd_object.c

   1 /*
   2  * GPL HEADER START
   3  *
   4  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   5  *
   6  * This program is free software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License version 2 only,
   8  * as published by the Free Software Foundation.
   9  *
  10  * This program is distributed in the hope that it will be useful, but
  11  * WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  13  * General Public License version 2 for more details (a copy is included
  14  * in the LICENSE file that accompanied this code).
  15  *
  16  * You should have received a copy of the GNU General Public License
  17  * version 2 along with this program; If not, see
  18  * http://www.gnu.org/licenses/gpl-2.0.html
  19  *
  20  * GPL HEADER END
  21  */
  22 /*
  23  * Copyright (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved.
  24  * Use is subject to license terms.
  25  *
  26  * Copyright (c) 2012, 2016, Intel Corporation.
  27  */
  28 /*
  29  * This file is part of Lustre, http://www.lustre.org/
  30  * Lustre is a trademark of Sun Microsystems, Inc.
  31  *
  32  * lustre/osd-zfs/osd_object.c
  33  *
  34  * Author: Alex Zhuravlev <bzzz@whamcloud.com>
  35  * Author: Mike Pershin <tappro@whamcloud.com>
  36  * Author: Johann Lombardi <johann@whamcloud.com>
  37  */
  38
  39 #define DEBUG_SUBSYSTEM S_OSD
  40
  41 #include <libcfs/libcfs.h>
  42 #include <obd_support.h>
  43 #include <lustre_net.h>
  44 #include <obd.h>
  45 #include <obd_class.h>
  46 #include <lustre_disk.h>
  47 #include <lustre_fid.h>
  48
  49 #include "osd_internal.h"
  50
  51 #include <sys/dnode.h>
  52 #include <sys/dbuf.h>
  53 #include <sys/spa.h>
  54 #include <sys/stat.h>
  55 #include <sys/zap.h>
  56 #include <sys/spa_impl.h>
  57 #include <sys/zfs_znode.h>
  58 #include <sys/dmu_tx.h>
  59 #include <sys/dmu_objset.h>
  60 #include <sys/dsl_prop.h>
  61 #include <sys/sa_impl.h>
  62 #include <sys/txg.h>
  63
  64 char *osd_obj_tag = "osd_object";
  65
  66 static struct dt_object_operations osd_obj_ops;
  67 static struct lu_object_operations osd_lu_obj_ops;
  68 extern struct dt_body_operations osd_body_ops;
  69 static struct dt_object_operations osd_obj_otable_it_ops;
  70
  71 extern struct kmem_cache *osd_object_kmem;
  72
  73 static void
  74 osd_object_sa_fini(struct osd_object *obj)
  75 {
  76         if (obj->oo_sa_hdl) {
  77                 sa_handle_destroy(obj->oo_sa_hdl);
  78                 obj->oo_sa_hdl = NULL;
  79         }
  80 }
  81
  82 static int
  83 osd_object_sa_init(struct osd_object *obj, struct osd_device *o)
  84 {
  85         int rc;
  86
  87         LASSERT(obj->oo_sa_hdl == NULL);
  88         LASSERT(obj->oo_dn != NULL);
  89
  90         rc = osd_sa_handle_get(obj);
  91         if (rc)
  92                 return rc;
  93
  94         /* Cache the xattr object id, valid for the life of the object */
  95         rc = -sa_lookup(obj->oo_sa_hdl, SA_ZPL_XATTR(o), &obj->oo_xattr, 8);
  96         if (rc == -ENOENT) {
  97                 obj->oo_xattr = ZFS_NO_OBJECT;
  98                 rc = 0;
  99         } else if (rc) {
 100                 osd_object_sa_fini(obj);
 101         }
 102
 103         return rc;
 104 }
 105
 106 /*
 107  * Add object to list of dirty objects in tx handle.
 108  */
 109 void osd_object_sa_dirty_add(struct osd_object *obj, struct osd_thandle *oh)
 110 {
 111         if (!list_empty(&obj->oo_sa_linkage))
 112                 return;
 113
 114         write_lock(&obj->oo_attr_lock);
 115         if (likely(list_empty(&obj->oo_sa_linkage)))
 116                 list_add(&obj->oo_sa_linkage, &oh->ot_sa_list);
 117         write_unlock(&obj->oo_attr_lock);
 118 }
 119
 120 /*
 121  * Release spill block dbuf hold for all dirty SAs.
 122  */
 123 void osd_object_sa_dirty_rele(const struct lu_env *env, struct osd_thandle *oh)
 124 {
 125         struct osd_object *obj;
 126
 127         while (!list_empty(&oh->ot_sa_list)) {
 128                 obj = list_entry(oh->ot_sa_list.next,
 129                                  struct osd_object, oo_sa_linkage);
 130                 write_lock(&obj->oo_attr_lock);
 131                 list_del_init(&obj->oo_sa_linkage);
 132                 write_unlock(&obj->oo_attr_lock);
 133                 if (obj->oo_late_xattr) {
 134                         /*
 135                          * take oo_guard to protect oo_sa_xattr buffer
 136                          * from concurrent update by osd_xattr_set()
 137                          */
 138                         LASSERT(oh->ot_assigned != 0);
 139                         down_write(&obj->oo_guard);
 140                         if (obj->oo_late_attr_set)
 141                                 __osd_sa_attr_init(env, obj, oh);
 142                         else if (obj->oo_late_xattr)
 143                                 __osd_sa_xattr_update(env, obj, oh);
 144                         up_write(&obj->oo_guard);
 145                 }
 146                 sa_spill_rele(obj->oo_sa_hdl);
 147         }
 148 }
 149
 150 /*
 151  * Update the SA and add the object to the dirty list.
 152  */
 153 int osd_object_sa_update(struct osd_object *obj, sa_attr_type_t type,
 154                          void *buf, uint32_t buflen, struct osd_thandle *oh)
 155 {
 156         int rc;
 157
 158         LASSERT(obj->oo_sa_hdl != NULL);
 159         LASSERT(oh->ot_tx != NULL);
 160
 161         rc = -sa_update(obj->oo_sa_hdl, type, buf, buflen, oh->ot_tx);
 162         osd_object_sa_dirty_add(obj, oh);
 163
 164         return rc;
 165 }
 166
 167 /*
 168  * Bulk update the SA and add the object to the dirty list.
 169  */
 170 static int
 171 osd_object_sa_bulk_update(struct osd_object *obj, sa_bulk_attr_t *attrs,
 172                           int count, struct osd_thandle *oh)
 173 {
 174         int rc;
 175
 176         LASSERT(obj->oo_sa_hdl != NULL);
 177         LASSERT(oh->ot_tx != NULL);
 178
 179         rc = -sa_bulk_update(obj->oo_sa_hdl, attrs, count, oh->ot_tx);
 180         osd_object_sa_dirty_add(obj, oh);
 181
 182         return rc;
 183 }
 184
 185 /*
 186  * Retrieve the attributes of a DMU object
 187  */
 188 int __osd_object_attr_get(const struct lu_env *env, struct osd_device *o,
 189                           struct osd_object *obj, struct lu_attr *la)
 190 {
 191         struct osa_attr *osa = &osd_oti_get(env)->oti_osa;
 192         sa_bulk_attr_t  *bulk = osd_oti_get(env)->oti_attr_bulk;
 193         int              cnt = 0;
 194         int              rc;
 195         ENTRY;
 196
 197         LASSERT(obj->oo_dn != NULL);
 198
 199         la->la_valid |= LA_ATIME | LA_MTIME | LA_CTIME | LA_MODE | LA_TYPE |
 200                         LA_SIZE | LA_UID | LA_GID | LA_FLAGS | LA_NLINK;
 201
 202         SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(o), NULL, osa->atime, 16);
 203         SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(o), NULL, osa->mtime, 16);
 204         SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(o), NULL, osa->ctime, 16);
 205         SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(o), NULL, &osa->mode, 8);
 206         SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_SIZE(o), NULL, &osa->size, 8);
 207         SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_LINKS(o), NULL, &osa->nlink, 8);
 208         SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_UID(o), NULL, &osa->uid, 8);
 209         SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_GID(o), NULL, &osa->gid, 8);
 210         SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(o), NULL, &osa->flags, 8);
 211         LASSERT(cnt <= ARRAY_SIZE(osd_oti_get(env)->oti_attr_bulk));
 212
 213         rc = -sa_bulk_lookup(obj->oo_sa_hdl, bulk, cnt);
 214         if (rc)
 215                 GOTO(out_sa, rc);
 216
 217 #ifdef ZFS_PROJINHERIT
 218         if (o->od_projectused_dn && osa->flags & ZFS_PROJID) {
 219                 rc = -sa_lookup(obj->oo_sa_hdl, SA_ZPL_PROJID(o),
 220                                 &osa->projid, 8);
 221                 if (rc)
 222                         GOTO(out_sa, rc);
 223
 224                 la->la_projid = osa->projid;
 225                 la->la_valid |= LA_PROJID;
 226                 obj->oo_with_projid = 1;
 227         } else {
 228                 la->la_projid = ZFS_DEFAULT_PROJID;
 229                 la->la_valid &= ~LA_PROJID;
 230         }
 231 #else
 232         la->la_projid = 0;
 233         la->la_valid &= ~LA_PROJID;
 234 #endif
 235
 236         la->la_atime = osa->atime[0];
 237         la->la_mtime = osa->mtime[0];
 238         la->la_ctime = osa->ctime[0];
 239         la->la_mode = osa->mode;
 240         la->la_uid = osa->uid;
 241         la->la_gid = osa->gid;
 242         la->la_nlink = osa->nlink;
 243         la->la_flags = attrs_zfs2fs(osa->flags);
 244         la->la_size = osa->size;
 245
 246         /* Try to get extra flag from LMA. Right now, only LMAI_ORPHAN
 247          * flags is stored in LMA, and it is only for orphan directory */
 248         if (S_ISDIR(la->la_mode) && dt_object_exists(&obj->oo_dt)) {
 249                 struct osd_thread_info *info = osd_oti_get(env);
 250                 struct lustre_mdt_attrs *lma;
 251                 struct lu_buf buf;
 252
 253                 lma = (struct lustre_mdt_attrs *)info->oti_buf;
 254                 buf.lb_buf = lma;
 255                 buf.lb_len = sizeof(info->oti_buf);
 256                 rc = osd_xattr_get(env, &obj->oo_dt, &buf, XATTR_NAME_LMA);
 257                 if (rc > 0) {
 258                         rc = 0;
 259                         lma->lma_incompat = le32_to_cpu(lma->lma_incompat);
 260                         obj->oo_lma_flags =
 261                                 lma_to_lustre_flags(lma->lma_incompat);
 262
 263                 } else if (rc == -ENODATA) {
 264                         rc = 0;
 265                 }
 266         }
 267
 268         if (S_ISCHR(la->la_mode) || S_ISBLK(la->la_mode)) {
 269                 rc = -sa_lookup(obj->oo_sa_hdl, SA_ZPL_RDEV(o), &osa->rdev, 8);
 270                 if (rc)
 271                         GOTO(out_sa, rc);
 272                 la->la_rdev = osa->rdev;
 273                 la->la_valid |= LA_RDEV;
 274         }
 275 out_sa:
 276
 277         RETURN(rc);
 278 }
 279
 280 int __osd_obj2dnode(objset_t *os, uint64_t oid, dnode_t **dnp)
 281 {
 282         dmu_buf_t *db;
 283         dmu_buf_impl_t *dbi;
 284         int rc;
 285
 286         rc = -dmu_bonus_hold(os, oid, osd_obj_tag, &db);
 287         if (rc)
 288                 return rc;
 289
 290         dbi = (dmu_buf_impl_t *)db;
 291         DB_DNODE_ENTER(dbi);
 292         *dnp = DB_DNODE(dbi);
 293         LASSERT(*dnp != NULL);
 294
 295         return 0;
 296 }
 297
 298 /*
 299  * Concurrency: no concurrent access is possible that early in object
 300  * life-cycle.
 301  */
 302 struct lu_object *osd_object_alloc(const struct lu_env *env,
 303                                    const struct lu_object_header *hdr,
 304                                    struct lu_device *d)
 305 {
 306         struct osd_object *mo;
 307
 308         OBD_SLAB_ALLOC_PTR_GFP(mo, osd_object_kmem, GFP_NOFS);
 309         if (mo != NULL) {
 310                 struct lu_object *l;
 311
 312                 l = &mo->oo_dt.do_lu;
 313                 dt_object_init(&mo->oo_dt, NULL, d);
 314                 mo->oo_dt.do_ops = &osd_obj_ops;
 315                 l->lo_ops = &osd_lu_obj_ops;
 316                 INIT_LIST_HEAD(&mo->oo_sa_linkage);
 317                 INIT_LIST_HEAD(&mo->oo_unlinked_linkage);
 318                 init_rwsem(&mo->oo_sem);
 319                 init_rwsem(&mo->oo_guard);
 320                 rwlock_init(&mo->oo_attr_lock);
 321                 mo->oo_destroy = OSD_DESTROY_NONE;
 322                 return l;
 323         } else {
 324                 return NULL;
 325         }
 326 }
 327
 328 /*
 329  * Concurrency: shouldn't matter.
 330  */
 331 int osd_object_init0(const struct lu_env *env, struct osd_object *obj)
 332 {
 333         struct osd_device       *osd = osd_obj2dev(obj);
 334         const struct lu_fid     *fid = lu_object_fid(&obj->oo_dt.do_lu);
 335         int                      rc = 0;
 336         ENTRY;
 337
 338         if (obj->oo_dn == NULL)
 339                 RETURN(0);
 340
 341         /* object exist */
 342
 343         rc = osd_object_sa_init(obj, osd);
 344         if (rc)
 345                 RETURN(rc);
 346
 347         /* cache attrs in object */
 348         rc = __osd_object_attr_get(env, osd, obj, &obj->oo_attr);
 349         if (rc)
 350                 RETURN(rc);
 351
 352         if (likely(!fid_is_acct(fid)))
 353                 /* no body operations for accounting objects */
 354                 obj->oo_dt.do_body_ops = &osd_body_ops;
 355
 356         /*
 357          * initialize object before marking it existing
 358          */
 359         obj->oo_dt.do_lu.lo_header->loh_attr |= obj->oo_attr.la_mode & S_IFMT;
 360
 361         smp_mb();
 362         obj->oo_dt.do_lu.lo_header->loh_attr |= LOHA_EXISTS;
 363
 364         RETURN(0);
 365 }
 366
 367 static int osd_check_lma(const struct lu_env *env, struct osd_object *obj)
 368 {
 369         struct osd_thread_info  *info = osd_oti_get(env);
 370         struct lu_buf           buf;
 371         int                     rc;
 372         struct lustre_mdt_attrs *lma;
 373         ENTRY;
 374
 375         CLASSERT(sizeof(info->oti_buf) >= sizeof(*lma));
 376         lma = (struct lustre_mdt_attrs *)info->oti_buf;
 377         buf.lb_buf = lma;
 378         buf.lb_len = sizeof(info->oti_buf);
 379
 380         rc = osd_xattr_get(env, &obj->oo_dt, &buf, XATTR_NAME_LMA);
 381         if (rc > 0) {
 382                 rc = 0;
 383                 lustre_lma_swab(lma);
 384                 if (unlikely((lma->lma_incompat & ~LMA_INCOMPAT_SUPP) ||
 385                              CFS_FAIL_CHECK(OBD_FAIL_OSD_LMA_INCOMPAT))) {
 386                         CWARN("%s: unsupported incompat LMA feature(s) %#x for "
 387                               "fid = "DFID"\n", osd_obj2dev(obj)->od_svname,
 388                               lma->lma_incompat & ~LMA_INCOMPAT_SUPP,
 389                               PFID(lu_object_fid(&obj->oo_dt.do_lu)));
 390                         rc = -EOPNOTSUPP;
 391                 }
 392         } else if (rc == -ENODATA) {
 393                 /* haven't initialize LMA xattr */
 394                 rc = 0;
 395         }
 396
 397         RETURN(rc);
 398 }
 399
 400 /**
 401  * Helper function to retrieve DMU object id from fid for accounting object
 402  */
 403 static dnode_t *osd_quota_fid2dmu(const struct osd_device *osd,
 404                                   const struct lu_fid *fid)
 405 {
 406         dnode_t *dn = NULL;
 407
 408         LASSERT(fid_is_acct(fid));
 409
 410         switch (fid_oid(fid)) {
 411         case ACCT_USER_OID:
 412                 dn = osd->od_userused_dn;
 413                 break;
 414         case ACCT_GROUP_OID:
 415                 dn = osd->od_groupused_dn;
 416                 break;
 417 #ifdef ZFS_PROJINHERIT
 418         case ACCT_PROJECT_OID:
 419                 dn = osd->od_projectused_dn;
 420                 break;
 421 #endif
 422         default:
 423                 break;
 424         }
 425
 426         return dn;
 427 }
 428
 429 /*
 430  * Concurrency: no concurrent access is possible that early in object
 431  * life-cycle.
 432  */
 433 static int osd_object_init(const struct lu_env *env, struct lu_object *l,
 434                            const struct lu_object_conf *conf)
 435 {
 436         struct osd_object *obj = osd_obj(l);
 437         struct osd_device *osd = osd_obj2dev(obj);
 438         const struct lu_fid *fid = lu_object_fid(l);
 439         uint64_t oid;
 440         int rc = 0;
 441         ENTRY;
 442
 443         LASSERT(osd_invariant(obj));
 444
 445         if (fid_is_otable_it(&l->lo_header->loh_fid)) {
 446                 obj->oo_dt.do_ops = &osd_obj_otable_it_ops;
 447                 l->lo_header->loh_attr |= LOHA_EXISTS;
 448                 RETURN(0);
 449         }
 450
 451         if (conf != NULL && conf->loc_flags & LOC_F_NEW)
 452                 GOTO(out, rc = 0);
 453
 454         if (unlikely(fid_is_acct(fid))) {
 455                 obj->oo_dn = osd_quota_fid2dmu(osd, fid);
 456                 if (obj->oo_dn) {
 457                         obj->oo_dt.do_index_ops = &osd_acct_index_ops;
 458                         l->lo_header->loh_attr |= LOHA_EXISTS;
 459                 }
 460
 461                 GOTO(out, rc = 0);
 462         }
 463
 464         rc = osd_fid_lookup(env, osd, fid, &oid);
 465         if (rc == 0) {
 466                 LASSERT(obj->oo_dn == NULL);
 467                 rc = __osd_obj2dnode(osd->od_os, oid, &obj->oo_dn);
 468                 /* EEXIST will be returned if object is being deleted in ZFS */
 469                 if (rc == -EEXIST) {
 470                         rc = 0;
 471                         GOTO(out, rc);
 472                 }
 473                 if (rc != 0) {
 474                         CERROR("%s: lookup "DFID"/%#llx failed: rc = %d\n",
 475                                osd->od_svname, PFID(lu_object_fid(l)), oid, rc);
 476                         GOTO(out, rc);
 477                 }
 478                 LASSERT(obj->oo_dn);
 479                 rc = osd_object_init0(env, obj);
 480                 if (rc != 0)
 481                         GOTO(out, rc);
 482
 483                 rc = osd_check_lma(env, obj);
 484                 if (rc != 0)
 485                         GOTO(out, rc);
 486         } else if (rc == -ENOENT) {
 487                 rc = 0;
 488         }
 489         LASSERT(osd_invariant(obj));
 490 out:
 491         RETURN(rc);
 492 }
 493
 494 /*
 495  * Concurrency: no concurrent access is possible that late in object
 496  * life-cycle.
 497  */
 498 static void osd_object_free(const struct lu_env *env, struct lu_object *l)
 499 {
 500         struct osd_object *obj = osd_obj(l);
 501
 502         LASSERT(osd_invariant(obj));
 503
 504         dt_object_fini(&obj->oo_dt);
 505         OBD_SLAB_FREE_PTR(obj, osd_object_kmem);
 506 }
 507
 508 static int
 509 osd_object_unlinked_add(struct osd_object *obj, struct osd_thandle *oh)
 510 {
 511         int rc = -EBUSY;
 512
 513         LASSERT(obj->oo_destroy == OSD_DESTROY_ASYNC);
 514
 515         /* the object is supposed to be exclusively locked by
 516          * the caller (osd_destroy()), while the transaction
 517          * (oh) is per-thread and not shared */
 518         if (likely(list_empty(&obj->oo_unlinked_linkage))) {
 519                 list_add(&obj->oo_unlinked_linkage, &oh->ot_unlinked_list);
 520                 rc = 0;
 521         }
 522
 523         return rc;
 524 }
 525
 526 /* Default to max data size covered by a level-1 indirect block */
 527 static unsigned long osd_sync_destroy_max_size =
 528         1UL << (DN_MAX_INDBLKSHIFT - SPA_BLKPTRSHIFT + SPA_MAXBLOCKSHIFT);
 529 module_param(osd_sync_destroy_max_size, ulong, 0444);
 530 MODULE_PARM_DESC(osd_sync_destroy_max_size, "Maximum object size to use synchronous destroy.");
 531
 532 static inline void
 533 osd_object_set_destroy_type(struct osd_object *obj)
 534 {
 535         /*
 536          * Lock-less OST_WRITE can race with OST_DESTROY, so set destroy type
 537          * only once and use it consistently thereafter.
 538          */
 539         down_write(&obj->oo_guard);
 540         if (obj->oo_destroy == OSD_DESTROY_NONE) {
 541                 if (obj->oo_attr.la_size <= osd_sync_destroy_max_size)
 542                         obj->oo_destroy = OSD_DESTROY_SYNC;
 543                 else /* Larger objects are destroyed asynchronously */
 544                         obj->oo_destroy = OSD_DESTROY_ASYNC;
 545         }
 546         up_write(&obj->oo_guard);
 547 }
 548
 549 static int osd_declare_destroy(const struct lu_env *env, struct dt_object *dt,
 550                                struct thandle *th)
 551 {
 552         const struct lu_fid     *fid = lu_object_fid(&dt->do_lu);
 553         struct osd_object       *obj = osd_dt_obj(dt);
 554         struct osd_device       *osd = osd_obj2dev(obj);
 555         struct osd_thandle      *oh;
 556         dnode_t *dn;
 557         int                      rc;
 558         uint64_t                 zapid;
 559         ENTRY;
 560
 561         LASSERT(th != NULL);
 562         LASSERT(dt_object_exists(dt));
 563
 564         oh = container_of0(th, struct osd_thandle, ot_super);
 565         LASSERT(oh->ot_tx != NULL);
 566
 567         /* declare that we'll remove object from fid-dnode mapping */
 568         zapid = osd_get_name_n_idx(env, osd, fid, NULL, 0, &dn);
 569         osd_tx_hold_zap(oh->ot_tx, zapid, dn, FALSE, NULL);
 570
 571         osd_declare_xattrs_destroy(env, obj, oh);
 572
 573         /* one less inode */
 574         rc = osd_declare_quota(env, osd, obj->oo_attr.la_uid,
 575                                obj->oo_attr.la_gid, obj->oo_attr.la_projid,
 576                                -1, oh, NULL, OSD_QID_INODE);
 577         if (rc)
 578                 RETURN(rc);
 579
 580         /* data to be truncated */
 581         rc = osd_declare_quota(env, osd, obj->oo_attr.la_uid,
 582                                obj->oo_attr.la_gid, obj->oo_attr.la_projid,
 583                                0, oh, NULL, OSD_QID_BLK);
 584         if (rc)
 585                 RETURN(rc);
 586
 587         osd_object_set_destroy_type(obj);
 588         if (obj->oo_destroy == OSD_DESTROY_SYNC)
 589                 dmu_tx_hold_free(oh->ot_tx, obj->oo_dn->dn_object,
 590                                  0, DMU_OBJECT_END);
 591         else
 592                 osd_tx_hold_zap(oh->ot_tx, osd->od_unlinked->dn_object,
 593                                 osd->od_unlinked, TRUE, NULL);
 594
 595         /* will help to find FID->ino when this object is being
 596          * added to PENDING/ */
 597         osd_idc_find_and_init(env, osd, obj);
 598
 599         RETURN(0);
 600 }
 601
 602 static int osd_destroy(const struct lu_env *env, struct dt_object *dt,
 603                        struct thandle *th)
 604 {
 605         struct osd_thread_info  *info = osd_oti_get(env);
 606         char                    *buf = info->oti_str;
 607         struct osd_object       *obj = osd_dt_obj(dt);
 608         struct osd_device       *osd = osd_obj2dev(obj);
 609         const struct lu_fid     *fid = lu_object_fid(&dt->do_lu);
 610         struct osd_thandle      *oh;
 611         int                      rc;
 612         uint64_t                 oid, zapid;
 613         dnode_t *zdn;
 614         ENTRY;
 615
 616         down_write(&obj->oo_guard);
 617
 618         if (unlikely(!dt_object_exists(dt) || obj->oo_destroyed))
 619                 GOTO(out, rc = -ENOENT);
 620
 621         LASSERT(obj->oo_dn != NULL);
 622
 623         oh = container_of0(th, struct osd_thandle, ot_super);
 624         LASSERT(oh != NULL);
 625         LASSERT(oh->ot_tx != NULL);
 626
 627         /* remove obj ref from index dir (it depends) */
 628         zapid = osd_get_name_n_idx(env, osd, fid, buf,
 629                                    sizeof(info->oti_str), &zdn);
 630         rc = osd_zap_remove(osd, zapid, zdn, buf, oh->ot_tx);
 631         if (rc) {
 632                 CERROR("%s: zap_remove(%s) failed: rc = %d\n",
 633                        osd->od_svname, buf, rc);
 634                 GOTO(out, rc);
 635         }
 636
 637         rc = osd_xattrs_destroy(env, obj, oh);
 638         if (rc) {
 639                 CERROR("%s: cannot destroy xattrs for %s: rc = %d\n",
 640                        osd->od_svname, buf, rc);
 641                 GOTO(out, rc);
 642         }
 643
 644         oid = obj->oo_dn->dn_object;
 645         if (unlikely(obj->oo_destroy == OSD_DESTROY_NONE)) {
 646                 /* this may happen if the destroy wasn't declared
 647                  * e.g. when the object is created and then destroyed
 648                  * in the same transaction - we don't need additional
 649                  * space for destroy specifically */
 650                 LASSERT(obj->oo_attr.la_size <= osd_sync_destroy_max_size);
 651                 rc = -dmu_object_free(osd->od_os, oid, oh->ot_tx);
 652                 if (rc)
 653                         CERROR("%s: failed to free %s %llu: rc = %d\n",
 654                                osd->od_svname, buf, oid, rc);
 655         } else if (obj->oo_destroy == OSD_DESTROY_SYNC) {
 656                 rc = -dmu_object_free(osd->od_os, oid, oh->ot_tx);
 657                 if (rc)
 658                         CERROR("%s: failed to free %s %llu: rc = %d\n",
 659                                osd->od_svname, buf, oid, rc);
 660         } else { /* asynchronous destroy */
 661                 char *key = info->oti_key;
 662
 663                 rc = osd_object_unlinked_add(obj, oh);
 664                 if (rc)
 665                         GOTO(out, rc);
 666
 667                 snprintf(key, sizeof(info->oti_key), "%llx", oid);
 668                 rc = osd_zap_add(osd, osd->od_unlinked->dn_object,
 669                                  osd->od_unlinked, key, 8, 1, &oid, oh->ot_tx);
 670                 if (rc)
 671                         CERROR("%s: zap_add_int() failed %s %llu: rc = %d\n",
 672                                osd->od_svname, buf, oid, rc);
 673         }
 674
 675 out:
 676         /* not needed in the cache anymore */
 677         set_bit(LU_OBJECT_HEARD_BANSHEE, &dt->do_lu.lo_header->loh_flags);
 678         if (rc == 0)
 679                 obj->oo_destroyed = 1;
 680         up_write(&obj->oo_guard);
 681         RETURN (0);
 682 }
 683
 684 static void osd_object_delete(const struct lu_env *env, struct lu_object *l)
 685 {
 686         struct osd_object *obj = osd_obj(l);
 687         const struct lu_fid *fid = lu_object_fid(l);
 688
 689         if (obj->oo_dn) {
 690                 if (likely(!fid_is_acct(fid))) {
 691                         osd_object_sa_fini(obj);
 692                         if (obj->oo_sa_xattr) {
 693                                 nvlist_free(obj->oo_sa_xattr);
 694                                 obj->oo_sa_xattr = NULL;
 695                         }
 696                         osd_dnode_rele(obj->oo_dn);
 697                         list_del(&obj->oo_sa_linkage);
 698                 }
 699                 obj->oo_dn = NULL;
 700         }
 701 }
 702
 703 /*
 704  * Concurrency: ->loo_object_release() is called under site spin-lock.
 705  */
 706 static void osd_object_release(const struct lu_env *env,
 707                                struct lu_object *l)
 708 {
 709 }
 710
 711 /*
 712  * Concurrency: shouldn't matter.
 713  */
 714 static int osd_object_print(const struct lu_env *env, void *cookie,
 715                             lu_printer_t p, const struct lu_object *l)
 716 {
 717         struct osd_object *o = osd_obj(l);
 718
 719         return (*p)(env, cookie, LUSTRE_OSD_ZFS_NAME"-object@%p", o);
 720 }
 721
 722 static void osd_read_lock(const struct lu_env *env, struct dt_object *dt,
 723                           unsigned role)
 724 {
 725         struct osd_object *obj = osd_dt_obj(dt);
 726
 727         LASSERT(osd_invariant(obj));
 728
 729         down_read_nested(&obj->oo_sem, role);
 730 }
 731
 732 static void osd_write_lock(const struct lu_env *env, struct dt_object *dt,
 733                            unsigned role)
 734 {
 735         struct osd_object *obj = osd_dt_obj(dt);
 736
 737         LASSERT(osd_invariant(obj));
 738
 739         down_write_nested(&obj->oo_sem, role);
 740 }
 741
 742 static void osd_read_unlock(const struct lu_env *env, struct dt_object *dt)
 743 {
 744         struct osd_object *obj = osd_dt_obj(dt);
 745
 746         LASSERT(osd_invariant(obj));
 747         up_read(&obj->oo_sem);
 748 }
 749
 750 static void osd_write_unlock(const struct lu_env *env, struct dt_object *dt)
 751 {
 752         struct osd_object *obj = osd_dt_obj(dt);
 753
 754         LASSERT(osd_invariant(obj));
 755         up_write(&obj->oo_sem);
 756 }
 757
 758 static int osd_write_locked(const struct lu_env *env, struct dt_object *dt)
 759 {
 760         struct osd_object *obj = osd_dt_obj(dt);
 761         int rc = 1;
 762
 763         LASSERT(osd_invariant(obj));
 764
 765         if (down_write_trylock(&obj->oo_sem)) {
 766                 rc = 0;
 767                 up_write(&obj->oo_sem);
 768         }
 769         return rc;
 770 }
 771
 772 static int osd_attr_get(const struct lu_env *env,
 773                         struct dt_object *dt,
 774                         struct lu_attr *attr)
 775 {
 776         struct osd_object       *obj = osd_dt_obj(dt);
 777         uint64_t                 blocks;
 778         uint32_t                 blksize;
 779         int                      rc = 0;
 780
 781         down_read(&obj->oo_guard);
 782
 783         if (unlikely(!dt_object_exists(dt) || obj->oo_destroyed))
 784                 GOTO(out, rc = -ENOENT);
 785
 786         if (unlikely(fid_is_acct(lu_object_fid(&dt->do_lu))))
 787                 GOTO(out, rc = 0);
 788
 789         LASSERT(osd_invariant(obj));
 790         LASSERT(obj->oo_dn);
 791
 792         read_lock(&obj->oo_attr_lock);
 793         *attr = obj->oo_attr;
 794         if (obj->oo_lma_flags & LUSTRE_ORPHAN_FL)
 795                 attr->la_flags |= LUSTRE_ORPHAN_FL;
 796         read_unlock(&obj->oo_attr_lock);
 797
 798         /* with ZFS_DEBUG zrl_add_debug() called by DB_DNODE_ENTER()
 799          * from within sa_object_size() can block on a mutex, so
 800          * we can't call sa_object_size() holding rwlock */
 801         sa_object_size(obj->oo_sa_hdl, &blksize, &blocks);
 802         /* we do not control size of indices, so always calculate
 803          * it from number of blocks reported by DMU */
 804         if (S_ISDIR(attr->la_mode))
 805                 attr->la_size = 512 * blocks;
 806         /* Block size may be not set; suggest maximal I/O transfers. */
 807         if (blksize == 0)
 808                 blksize = osd_spa_maxblocksize(
 809                         dmu_objset_spa(osd_obj2dev(obj)->od_os));
 810
 811         attr->la_blksize = blksize;
 812         attr->la_blocks = blocks;
 813         attr->la_valid |= LA_BLOCKS | LA_BLKSIZE;
 814
 815 out:
 816         up_read(&obj->oo_guard);
 817         return rc;
 818 }
 819
 820 /* Simple wrapper on top of qsd API which implement quota transfer for osd
 821  * setattr needs. As a reminder, only the root user can change ownership of
 822  * a file, that's why EDQUOT & EINPROGRESS errors are discarded */
 823 static inline int qsd_transfer(const struct lu_env *env,
 824                                struct qsd_instance *qsd,
 825                                struct lquota_trans *trans, int qtype,
 826                                __u64 orig_id, __u64 new_id, __u64 bspace,
 827                                struct lquota_id_info *qi)
 828 {
 829         int     rc;
 830
 831         if (unlikely(qsd == NULL))
 832                 return 0;
 833
 834         LASSERT(qtype >= 0 && qtype < LL_MAXQUOTAS);
 835         qi->lqi_type = qtype;
 836
 837         /* inode accounting */
 838         qi->lqi_is_blk = false;
 839
 840         /* one more inode for the new owner ... */
 841         qi->lqi_id.qid_uid = new_id;
 842         qi->lqi_space      = 1;
 843         rc = qsd_op_begin(env, qsd, trans, qi, NULL);
 844         if (rc == -EDQUOT || rc == -EINPROGRESS)
 845                 rc = 0;
 846         if (rc)
 847                 return rc;
 848
 849         /* and one less inode for the current id */
 850         qi->lqi_id.qid_uid = orig_id;;
 851         qi->lqi_space      = -1;
 852         /* can't get EDQUOT when reducing usage */
 853         rc = qsd_op_begin(env, qsd, trans, qi, NULL);
 854         if (rc == -EINPROGRESS)
 855                 rc = 0;
 856         if (rc)
 857                 return rc;
 858
 859         /* block accounting */
 860         qi->lqi_is_blk = true;
 861
 862         /* more blocks for the new owner ... */
 863         qi->lqi_id.qid_uid = new_id;
 864         qi->lqi_space      = bspace;
 865         rc = qsd_op_begin(env, qsd, trans, qi, NULL);
 866         if (rc == -EDQUOT || rc == -EINPROGRESS)
 867                 rc = 0;
 868         if (rc)
 869                 return rc;
 870
 871         /* and finally less blocks for the current owner */
 872         qi->lqi_id.qid_uid = orig_id;
 873         qi->lqi_space      = -bspace;
 874         rc = qsd_op_begin(env, qsd, trans, qi, NULL);
 875         /* can't get EDQUOT when reducing usage */
 876         if (rc == -EINPROGRESS)
 877                 rc = 0;
 878         return rc;
 879 }
 880
 881 static int osd_declare_attr_set(const struct lu_env *env,
 882                                 struct dt_object *dt,
 883                                 const struct lu_attr *attr,
 884                                 struct thandle *handle)
 885 {
 886         struct osd_thread_info  *info = osd_oti_get(env);
 887         struct osd_object       *obj = osd_dt_obj(dt);
 888         struct osd_device       *osd = osd_obj2dev(obj);
 889         dmu_tx_hold_t           *txh;
 890         struct osd_thandle      *oh;
 891         uint64_t                 bspace;
 892         uint32_t                 blksize;
 893         int                      rc = 0;
 894         bool                     found;
 895         ENTRY;
 896
 897
 898         LASSERT(handle != NULL);
 899         LASSERT(osd_invariant(obj));
 900
 901         oh = container_of0(handle, struct osd_thandle, ot_super);
 902
 903         down_read(&obj->oo_guard);
 904         if (unlikely(!dt_object_exists(dt) || obj->oo_destroyed))
 905                 GOTO(out, rc = 0);
 906
 907         LASSERT(obj->oo_sa_hdl != NULL);
 908         LASSERT(oh->ot_tx != NULL);
 909         /* regular attributes are part of the bonus buffer */
 910         /* let's check whether this object is already part of
 911          * transaction.. */
 912         found = false;
 913         for (txh = list_head(&oh->ot_tx->tx_holds); txh;
 914              txh = list_next(&oh->ot_tx->tx_holds, txh)) {
 915                 if (txh->txh_dnode == NULL)
 916                         continue;
 917                 if (txh->txh_dnode->dn_object != obj->oo_dn->dn_object)
 918                         continue;
 919                 /* this object is part of the transaction already
 920                  * we don't need to declare bonus again */
 921                 found = true;
 922                 break;
 923         }
 924         if (!found)
 925                 dmu_tx_hold_bonus(oh->ot_tx, obj->oo_dn->dn_object);
 926         if (oh->ot_tx->tx_err != 0)
 927                 GOTO(out, rc = -oh->ot_tx->tx_err);
 928
 929         if (attr && attr->la_valid & LA_FLAGS) {
 930                 /* LMA is usually a part of bonus, no need to declare
 931                  * anything else */
 932         }
 933
 934         if (attr && (attr->la_valid & (LA_UID | LA_GID | LA_PROJID))) {
 935                 sa_object_size(obj->oo_sa_hdl, &blksize, &bspace);
 936                 bspace = toqb(bspace * blksize);
 937         }
 938
 939         if (attr && attr->la_valid & LA_UID) {
 940                 /* quota enforcement for user */
 941                 if (attr->la_uid != obj->oo_attr.la_uid) {
 942                         rc = qsd_transfer(env, osd->od_quota_slave,
 943                                           &oh->ot_quota_trans, USRQUOTA,
 944                                           obj->oo_attr.la_uid, attr->la_uid,
 945                                           bspace, &info->oti_qi);
 946                         if (rc)
 947                                 GOTO(out, rc);
 948                 }
 949         }
 950         if (attr && attr->la_valid & LA_GID) {
 951                 /* quota enforcement for group */
 952                 if (attr->la_gid != obj->oo_attr.la_gid) {
 953                         rc = qsd_transfer(env, osd->od_quota_slave,
 954                                           &oh->ot_quota_trans, GRPQUOTA,
 955                                           obj->oo_attr.la_gid, attr->la_gid,
 956                                           bspace, &info->oti_qi);
 957                         if (rc)
 958                                 GOTO(out, rc);
 959                 }
 960         }
 961 #ifdef ZFS_PROJINHERIT
 962         if (attr && attr->la_valid & LA_PROJID) {
 963                 if (!osd->od_projectused_dn)
 964                         GOTO(out, rc = -EOPNOTSUPP);
 965
 966                 /* Usually, if project quota is upgradable for the device,
 967                  * then the upgrade will be done before or when mount the
 968                  * device. So when we come here, this project should have
 969                  * project ID attribute already (that is zero by default).
 970                  * Otherwise, there was something wrong during the former
 971                  * upgrade, let's return failure to report that.
 972                  *
 973                  * Please note that, different from other attributes, you
 974                  * can NOT simply set the project ID attribute under such
 975                  * case, because adding (NOT change) project ID attribute
 976                  * needs to change the object's attribute layout to match
 977                  * zfs backend quota accounting requirement. */
 978                 if (unlikely(!obj->oo_with_projid))
 979                         GOTO(out, rc = -ENXIO);
 980
 981                 /* quota enforcement for project */
 982                 if (attr->la_projid != obj->oo_attr.la_projid) {
 983                         rc = qsd_transfer(env, osd->od_quota_slave,
 984                                           &oh->ot_quota_trans, PRJQUOTA,
 985                                           obj->oo_attr.la_projid,
 986                                           attr->la_projid, bspace,
 987                                           &info->oti_qi);
 988                         if (rc)
 989                                 GOTO(out, rc);
 990                 }
 991         }
 992 #endif
 993 out:
 994         up_read(&obj->oo_guard);
 995         RETURN(rc);
 996 }
 997
 998 /*
 999  * Set the attributes of an object
1000  *
1001  * The transaction passed to this routine must have
1002  * dmu_tx_hold_bonus(tx, oid) called and then assigned
1003  * to a transaction group.
1004  */
1005 static int osd_attr_set(const struct lu_env *env, struct dt_object *dt,
1006                         const struct lu_attr *la, struct thandle *handle)
1007 {
1008         struct osd_thread_info  *info = osd_oti_get(env);
1009         sa_bulk_attr_t          *bulk = osd_oti_get(env)->oti_attr_bulk;
1010         struct osd_object       *obj = osd_dt_obj(dt);
1011         struct osd_device       *osd = osd_obj2dev(obj);
1012         struct osd_thandle      *oh;
1013         struct osa_attr         *osa = &info->oti_osa;
1014         __u64                    valid = la->la_valid;
1015         int                      cnt;
1016         int                      rc = 0;
1017
1018         ENTRY;
1019
1020         down_read(&obj->oo_guard);
1021         if (unlikely(!dt_object_exists(dt) || obj->oo_destroyed))
1022                 GOTO(out, rc = -ENOENT);
1023
1024         LASSERT(handle != NULL);
1025         LASSERT(osd_invariant(obj));
1026         LASSERT(obj->oo_sa_hdl);
1027
1028         oh = container_of0(handle, struct osd_thandle, ot_super);
1029         /* Assert that the transaction has been assigned to a
1030            transaction group. */
1031         LASSERT(oh->ot_tx->tx_txg != 0);
1032
1033         /* Only allow set size for regular file */
1034         if (!S_ISREG(dt->do_lu.lo_header->loh_attr))
1035                 valid &= ~(LA_SIZE | LA_BLOCKS);
1036
1037         if (valid & LA_CTIME && la->la_ctime == obj->oo_attr.la_ctime)
1038                 valid &= ~LA_CTIME;
1039
1040         if (valid & LA_MTIME && la->la_mtime == obj->oo_attr.la_mtime)
1041                 valid &= ~LA_MTIME;
1042
1043         if (valid & LA_ATIME && la->la_atime == obj->oo_attr.la_atime)
1044                 valid &= ~LA_ATIME;
1045
1046         if (valid == 0)
1047                 GOTO(out, rc = 0);
1048
1049         if (valid & LA_FLAGS) {
1050                 struct lustre_mdt_attrs *lma;
1051                 struct lu_buf buf;
1052
1053                 if (la->la_flags & LUSTRE_LMA_FL_MASKS) {
1054                         CLASSERT(sizeof(info->oti_buf) >= sizeof(*lma));
1055                         lma = (struct lustre_mdt_attrs *)&info->oti_buf;
1056                         buf.lb_buf = lma;
1057                         buf.lb_len = sizeof(info->oti_buf);
1058                         rc = osd_xattr_get(env, &obj->oo_dt, &buf,
1059                                            XATTR_NAME_LMA);
1060                         if (rc > 0) {
1061                                 lma->lma_incompat =
1062                                         le32_to_cpu(lma->lma_incompat);
1063                                 lma->lma_incompat |=
1064                                         lustre_to_lma_flags(la->la_flags);
1065                                 lma->lma_incompat =
1066                                         cpu_to_le32(lma->lma_incompat);
1067                                 buf.lb_buf = lma;
1068                                 buf.lb_len = sizeof(*lma);
1069                                 rc = osd_xattr_set_internal(env, obj, &buf,
1070                                                             XATTR_NAME_LMA,
1071                                                             LU_XATTR_REPLACE,
1072                                                             oh);
1073                         }
1074                         if (rc < 0) {
1075                                 CWARN("%s: failed to set LMA flags: rc = %d\n",
1076                                        osd->od_svname, rc);
1077                                 GOTO(out, rc);
1078                         }
1079                 }
1080         }
1081
1082         write_lock(&obj->oo_attr_lock);
1083         cnt = 0;
1084
1085         if (valid & LA_PROJID) {
1086 #ifdef ZFS_PROJINHERIT
1087                 /* osd_declare_attr_set() must be called firstly.
1088                  * If osd::od_projectused_dn is not set, then we
1089                  * can not arrive at here. */
1090                 LASSERT(osd->od_projectused_dn);
1091                 LASSERT(obj->oo_with_projid);
1092
1093                 osa->projid = obj->oo_attr.la_projid = la->la_projid;
1094                 SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_PROJID(osd), NULL,
1095                                  &osa->projid, 8);
1096 #else
1097                 valid &= ~LA_PROJID;
1098 #endif
1099         }
1100
1101         if (valid & LA_ATIME) {
1102                 osa->atime[0] = obj->oo_attr.la_atime = la->la_atime;
1103                 SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(osd), NULL,
1104                                  osa->atime, 16);
1105         }
1106         if (valid & LA_MTIME) {
1107                 osa->mtime[0] = obj->oo_attr.la_mtime = la->la_mtime;
1108                 SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(osd), NULL,
1109                                  osa->mtime, 16);
1110         }
1111         if (valid & LA_CTIME) {
1112                 osa->ctime[0] = obj->oo_attr.la_ctime = la->la_ctime;
1113                 SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(osd), NULL,
1114                                  osa->ctime, 16);
1115         }
1116         if (valid & LA_MODE) {
1117                 /* mode is stored along with type, so read it first */
1118                 obj->oo_attr.la_mode = (obj->oo_attr.la_mode & S_IFMT) |
1119                         (la->la_mode & ~S_IFMT);
1120                 osa->mode = obj->oo_attr.la_mode;
1121                 SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(osd), NULL,
1122                                  &osa->mode, 8);
1123         }
1124         if (valid & LA_SIZE) {
1125                 osa->size = obj->oo_attr.la_size = la->la_size;
1126                 SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_SIZE(osd), NULL,
1127                                  &osa->size, 8);
1128         }
1129         if (valid & LA_NLINK) {
1130                 osa->nlink = obj->oo_attr.la_nlink = la->la_nlink;
1131                 SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_LINKS(osd), NULL,
1132                                  &osa->nlink, 8);
1133         }
1134         if (valid & LA_RDEV) {
1135                 osa->rdev = obj->oo_attr.la_rdev = la->la_rdev;
1136                 SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_RDEV(osd), NULL,
1137                                  &osa->rdev, 8);
1138         }
1139         if (valid & LA_FLAGS) {
1140                 osa->flags = attrs_fs2zfs(la->la_flags);
1141                 /* many flags are not supported by zfs, so ensure a good cached
1142                  * copy */
1143                 obj->oo_attr.la_flags = attrs_zfs2fs(osa->flags);
1144 #ifdef ZFS_PROJINHERIT
1145                 if (obj->oo_with_projid)
1146                         osa->flags |= ZFS_PROJID;
1147 #endif
1148                 SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(osd), NULL,
1149                                  &osa->flags, 8);
1150         }
1151         if (valid & LA_UID) {
1152                 osa->uid = obj->oo_attr.la_uid = la->la_uid;
1153                 SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_UID(osd), NULL,
1154                                  &osa->uid, 8);
1155         }
1156         if (valid & LA_GID) {
1157                 osa->gid = obj->oo_attr.la_gid = la->la_gid;
1158                 SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_GID(osd), NULL,
1159                                  &osa->gid, 8);
1160         }
1161         obj->oo_attr.la_valid |= valid;
1162         write_unlock(&obj->oo_attr_lock);
1163
1164         LASSERT(cnt <= ARRAY_SIZE(osd_oti_get(env)->oti_attr_bulk));
1165         rc = osd_object_sa_bulk_update(obj, bulk, cnt, oh);
1166
1167 out:
1168         up_read(&obj->oo_guard);
1169         RETURN(rc);
1170 }
1171
1172 /*
1173  * Object creation.
1174  *
1175  * XXX temporary solution.
1176  */
1177
1178 static void osd_ah_init(const struct lu_env *env, struct dt_allocation_hint *ah,
1179                         struct dt_object *parent, struct dt_object *child,
1180                         umode_t child_mode)
1181 {
1182         LASSERT(ah);
1183
1184         ah->dah_parent = parent;
1185         ah->dah_mode = child_mode;
1186
1187         if (parent != NULL && !dt_object_remote(parent)) {
1188                 /* will help to find FID->ino at dt_insert("..") */
1189                 struct osd_object *pobj = osd_dt_obj(parent);
1190
1191                 osd_idc_find_and_init(env, osd_obj2dev(pobj), pobj);
1192         }
1193 }
1194
1195 static int osd_declare_create(const struct lu_env *env, struct dt_object *dt,
1196                               struct lu_attr *attr,
1197                               struct dt_allocation_hint *hint,
1198                               struct dt_object_format *dof,
1199                               struct thandle *handle)
1200 {
1201         const struct lu_fid     *fid = lu_object_fid(&dt->do_lu);
1202         struct osd_object       *obj = osd_dt_obj(dt);
1203         struct osd_device       *osd = osd_obj2dev(obj);
1204         struct osd_thandle      *oh;
1205         uint64_t                 zapid;
1206         dnode_t                 *dn;
1207         int                      rc, dnode_size;
1208         ENTRY;
1209
1210         LASSERT(dof);
1211
1212         switch (dof->dof_type) {
1213                 case DFT_REGULAR:
1214                 case DFT_SYM:
1215                 case DFT_NODE:
1216                         if (obj->oo_dt.do_body_ops == NULL)
1217                                 obj->oo_dt.do_body_ops = &osd_body_ops;
1218                         break;
1219                 default:
1220                         break;
1221         }
1222
1223         LASSERT(handle != NULL);
1224         oh = container_of0(handle, struct osd_thandle, ot_super);
1225         LASSERT(oh->ot_tx != NULL);
1226
1227         /* this is the minimum set of EAs on every Lustre object */
1228         obj->oo_ea_in_bonus = ZFS_SA_BASE_ATTR_SIZE +
1229                                 sizeof(__u64) + /* VBR VERSION */
1230                                 sizeof(struct lustre_mdt_attrs); /* LMA */
1231         /* reserve 32 bytes for extra stuff like ACLs */
1232         dnode_size = size_roundup_power2(obj->oo_ea_in_bonus + 32);
1233
1234         switch (dof->dof_type) {
1235                 case DFT_DIR:
1236                         dt->do_index_ops = &osd_dir_ops;
1237                 case DFT_INDEX:
1238                         /* for zap create */
1239                         dmu_tx_hold_zap(oh->ot_tx, DMU_NEW_OBJECT, FALSE, NULL);
1240                         dmu_tx_hold_sa_create(oh->ot_tx, dnode_size);
1241                         break;
1242                 case DFT_REGULAR:
1243                 case DFT_SYM:
1244                 case DFT_NODE:
1245                         /* first, we'll create new object */
1246                         dmu_tx_hold_sa_create(oh->ot_tx, dnode_size);
1247                         break;
1248
1249                 default:
1250                         LBUG();
1251                         break;
1252         }
1253
1254         /* and we'll add it to some mapping */
1255         zapid = osd_get_name_n_idx(env, osd, fid, NULL, 0, &dn);
1256         osd_tx_hold_zap(oh->ot_tx, zapid, dn, TRUE, NULL);
1257
1258         /* will help to find FID->ino mapping at dt_insert() */
1259         osd_idc_find_and_init(env, osd, obj);
1260
1261         rc = osd_declare_quota(env, osd, attr->la_uid, attr->la_gid,
1262                                attr->la_projid, 1, oh, NULL, OSD_QID_INODE);
1263
1264         RETURN(rc);
1265 }
1266
1267 int __osd_attr_init(const struct lu_env *env, struct osd_device *osd,
1268                     struct osd_object *obj, sa_handle_t *sa_hdl, dmu_tx_t *tx,
1269                     struct lu_attr *la, uint64_t parent,
1270                     nvlist_t *xattr)
1271 {
1272         sa_bulk_attr_t  *bulk = osd_oti_get(env)->oti_attr_bulk;
1273         struct osa_attr *osa = &osd_oti_get(env)->oti_osa;
1274         uint64_t         gen;
1275         uint64_t         crtime[2];
1276         timestruc_t      now;
1277         int              cnt;
1278         int              rc;
1279         char *dxattr = NULL;
1280         size_t sa_size;
1281
1282
1283         LASSERT(sa_hdl);
1284
1285         gen = dmu_tx_get_txg(tx);
1286         gethrestime(&now);
1287         ZFS_TIME_ENCODE(&now, crtime);
1288
1289         osa->atime[0] = la->la_atime;
1290         osa->ctime[0] = la->la_ctime;
1291         osa->mtime[0] = la->la_mtime;
1292         osa->mode = la->la_mode;
1293         osa->uid = la->la_uid;
1294         osa->gid = la->la_gid;
1295         osa->rdev = la->la_rdev;
1296         osa->nlink = la->la_nlink;
1297         if (la->la_valid & LA_FLAGS)
1298                 osa->flags = attrs_fs2zfs(la->la_flags);
1299         else
1300                 osa->flags = 0;
1301         osa->size  = la->la_size;
1302 #ifdef ZFS_PROJINHERIT
1303         if (osd->od_projectused_dn) {
1304                 if (la->la_valid & LA_PROJID)
1305                         osa->projid = la->la_projid;
1306                 else
1307                         osa->projid = ZFS_DEFAULT_PROJID;
1308                 osa->flags |= ZFS_PROJID;
1309                 if (obj)
1310                         obj->oo_with_projid = 1;
1311         } else {
1312                 osa->flags &= ~ZFS_PROJID;
1313         }
1314 #endif
1315
1316         /*
1317          * we need to create all SA below upon object create.
1318          *
1319          * XXX The attribute order matters since the accounting callback relies
1320          * on static offsets (i.e. SA_*_OFFSET, see zfs_space_delta_cb()) to
1321          * look up the UID/GID/PROJID attributes. Moreover, the callback does
1322          * not seem to support the spill block.
1323          * We define attributes in the same order as SA_*_OFFSET in order to
1324          * work around the problem. See ORI-610.
1325          */
1326         cnt = 0;
1327         SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(osd), NULL, &osa->mode, 8);
1328         SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_SIZE(osd), NULL, &osa->size, 8);
1329         SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_GEN(osd), NULL, &gen, 8);
1330         SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_UID(osd), NULL, &osa->uid, 8);
1331         SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_GID(osd), NULL, &osa->gid, 8);
1332         SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_PARENT(osd), NULL, &parent, 8);
1333         SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(osd), NULL, &osa->flags, 8);
1334         SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(osd), NULL, osa->atime, 16);
1335         SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(osd), NULL, osa->mtime, 16);
1336         SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(osd), NULL, osa->ctime, 16);
1337         SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CRTIME(osd), NULL, crtime, 16);
1338         SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_LINKS(osd), NULL, &osa->nlink, 8);
1339 #ifdef ZFS_PROJINHERIT
1340         if (osd->od_projectused_dn)
1341                 SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_PROJID(osd), NULL,
1342                                  &osa->projid, 8);
1343 #endif
1344         SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_RDEV(osd), NULL, &osa->rdev, 8);
1345         LASSERT(cnt <= ARRAY_SIZE(osd_oti_get(env)->oti_attr_bulk));
1346
1347         if (xattr) {
1348                 rc = -nvlist_size(xattr, &sa_size, NV_ENCODE_XDR);
1349                 LASSERT(rc == 0);
1350
1351                 dxattr = osd_zio_buf_alloc(sa_size);
1352                 LASSERT(dxattr);
1353
1354                 rc = -nvlist_pack(xattr, &dxattr, &sa_size,
1355                                 NV_ENCODE_XDR, KM_SLEEP);
1356                 LASSERT(rc == 0);
1357
1358                 SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_DXATTR(osd),
1359                                 NULL, dxattr, sa_size);
1360         }
1361
1362         rc = -sa_replace_all_by_template(sa_hdl, bulk, cnt, tx);
1363         if (dxattr)
1364                 osd_zio_buf_free(dxattr, sa_size);
1365
1366         return rc;
1367 }
1368
1369 static int osd_find_new_dnode(const struct lu_env *env, dmu_tx_t *tx,
1370                               uint64_t oid, dnode_t **dnp)
1371 {
1372         dmu_tx_hold_t *txh;
1373         int rc = 0;
1374
1375         /* take dnode_t from tx to save on dnode#->dnode_t lookup */
1376         for (txh = list_tail(&tx->tx_holds); txh;
1377              txh = list_prev(&tx->tx_holds, txh)) {
1378                 dnode_t *dn = txh->txh_dnode;
1379                 dmu_buf_impl_t *db;
1380
1381                 if (dn == NULL)
1382                         continue;
1383                 if (dn->dn_object != oid)
1384                         continue;
1385                 db = dn->dn_bonus;
1386                 if (db == NULL) {
1387                         rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
1388                         if (dn->dn_bonus == NULL)
1389                                 dbuf_create_bonus(dn);
1390                         rw_exit(&dn->dn_struct_rwlock);
1391                 }
1392                 db = dn->dn_bonus;
1393                 LASSERT(db);
1394                 LASSERT(dn->dn_handle);
1395                 DB_DNODE_ENTER(db);
1396                 if (refcount_add(&db->db_holds, osd_obj_tag) == 1) {
1397                         refcount_add(&dn->dn_holds, tag);
1398                         atomic_inc_32(&dn->dn_dbufs_count);
1399                 }
1400                 *dnp = dn;
1401                 dbuf_read(db, NULL, DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH);
1402                 break;
1403         }
1404
1405         if (unlikely(*dnp == NULL))
1406                 rc = __osd_obj2dnode(tx->tx_objset, oid, dnp);
1407
1408         return rc;
1409 }
1410
1411 #ifdef HAVE_DMU_OBJECT_ALLOC_DNSIZE
1412 static int osd_find_dnsize(struct osd_object *obj)
1413 {
1414         struct osd_device *osd = osd_obj2dev(obj);
1415         int dnsize;
1416
1417         if (osd->od_dnsize == ZFS_DNSIZE_AUTO) {
1418                 dnsize = DNODE_MIN_SIZE;
1419                 do {
1420                         if (DN_BONUS_SIZE(dnsize) >= obj->oo_ea_in_bonus + 32)
1421                                 break;
1422                         dnsize <<= 1;
1423                 } while (dnsize < DNODE_MAX_SIZE);
1424                 if (dnsize > DNODE_MAX_SIZE)
1425                         dnsize = DNODE_MAX_SIZE;
1426         } else if (osd->od_dnsize == ZFS_DNSIZE_1K) {
1427                 dnsize = 1024;
1428         } else if (osd->od_dnsize == ZFS_DNSIZE_2K) {
1429                 dnsize = 2048;
1430         } else if (osd->od_dnsize == ZFS_DNSIZE_4K) {
1431                 dnsize = 4096;
1432         } else if (osd->od_dnsize == ZFS_DNSIZE_8K) {
1433                 dnsize = 8192;
1434         } else if (osd->od_dnsize == ZFS_DNSIZE_16K) {
1435                 dnsize = 16384;
1436         } else {
1437                 dnsize = DNODE_MIN_SIZE;
1438         }
1439         return dnsize;
1440 }
1441 #else
1442 static int inline osd_find_dnsize(struct osd_object *obj)
1443 {
1444         return DN_MAX_BONUSLEN;
1445 }
1446 #endif
1447
1448 /*
1449  * The transaction passed to this routine must have
1450  * dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT) called and then assigned
1451  * to a transaction group.
1452  */
1453 int __osd_object_create(const struct lu_env *env, struct osd_object *obj,
1454                         dnode_t **dnp, dmu_tx_t *tx, struct lu_attr *la)
1455 {
1456         struct osd_device   *osd = osd_obj2dev(obj);
1457         const struct lu_fid *fid = lu_object_fid(&obj->oo_dt.do_lu);
1458         dmu_object_type_t    type = DMU_OT_PLAIN_FILE_CONTENTS;
1459         uint64_t oid;
1460
1461         /* Use DMU_OTN_UINT8_METADATA for local objects so their data blocks
1462          * would get an additional ditto copy */
1463         if (unlikely(S_ISREG(la->la_mode) &&
1464                      fid_seq_is_local_file(fid_seq(fid))))
1465                 type = DMU_OTN_UINT8_METADATA;
1466
1467         /* Create a new DMU object using the default dnode size. */
1468         oid = osd_dmu_object_alloc(osd->od_os, type, 0,
1469                                    osd_find_dnsize(obj), tx);
1470
1471         LASSERT(la->la_valid & LA_MODE);
1472         la->la_size = 0;
1473         la->la_nlink = 1;
1474
1475         return osd_find_new_dnode(env, tx, oid, dnp);
1476 }
1477
1478 /*
1479  * The transaction passed to this routine must have
1480  * dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, ...) called and then assigned
1481  * to a transaction group.
1482  *
1483  * Using ZAP_FLAG_HASH64 will force the ZAP to always be a FAT ZAP.
1484  * This is fine for directories today, because storing the FID in the dirent
1485  * will also require a FAT ZAP.  If there is a new type of micro ZAP created
1486  * then we might need to re-evaluate the use of this flag and instead do
1487  * a conversion from the different internal ZAP hash formats being used. */
1488 int __osd_zap_create(const struct lu_env *env, struct osd_device *osd,
1489                      dnode_t **dnp, dmu_tx_t *tx, struct lu_attr *la,
1490                      unsigned dnsize, zap_flags_t flags)
1491 {
1492         uint64_t oid;
1493
1494         /* Assert that the transaction has been assigned to a
1495            transaction group. */
1496         LASSERT(tx->tx_txg != 0);
1497         *dnp = NULL;
1498
1499         oid = osd_zap_create_flags(osd->od_os, 0, flags | ZAP_FLAG_HASH64,
1500                                    DMU_OT_DIRECTORY_CONTENTS,
1501                                    14, /* == ZFS fzap_default_blockshift */
1502                                    DN_MAX_INDBLKSHIFT, /* indirect blockshift */
1503                                    dnsize, tx);
1504
1505         la->la_size = 2;
1506         la->la_nlink = 1;
1507
1508         return osd_find_new_dnode(env, tx, oid, dnp);
1509 }
1510
1511 static dnode_t *osd_mkidx(const struct lu_env *env, struct osd_object *obj,
1512                           struct lu_attr *la, struct osd_thandle *oh)
1513 {
1514         dnode_t *dn;
1515         int rc;
1516
1517         /* Index file should be created as regular file in order not to confuse
1518          * ZPL which could interpret them as directory.
1519          * We set ZAP_FLAG_UINT64_KEY to let ZFS know than we are going to use
1520          * binary keys */
1521         LASSERT(S_ISREG(la->la_mode));
1522         rc = __osd_zap_create(env, osd_obj2dev(obj), &dn, oh->ot_tx, la,
1523                               osd_find_dnsize(obj), ZAP_FLAG_UINT64_KEY);
1524         if (rc)
1525                 return ERR_PTR(rc);
1526         return dn;
1527 }
1528
1529 static dnode_t *osd_mkdir(const struct lu_env *env, struct osd_object *obj,
1530                           struct lu_attr *la, struct osd_thandle *oh)
1531 {
1532         dnode_t *dn;
1533         int rc;
1534
1535         LASSERT(S_ISDIR(la->la_mode));
1536         rc = __osd_zap_create(env, osd_obj2dev(obj), &dn, oh->ot_tx, la,
1537                               osd_find_dnsize(obj), 0);
1538         if (rc)
1539                 return ERR_PTR(rc);
1540         return dn;
1541 }
1542
1543 static dnode_t *osd_mkreg(const struct lu_env *env, struct osd_object *obj,
1544                           struct lu_attr *la, struct osd_thandle *oh)
1545 {
1546         const struct lu_fid *fid = lu_object_fid(&obj->oo_dt.do_lu);
1547         struct osd_device *osd = osd_obj2dev(obj);
1548         dnode_t *dn;
1549         int rc;
1550
1551         LASSERT(S_ISREG(la->la_mode));
1552         rc = __osd_object_create(env, obj, &dn, oh->ot_tx, la);
1553         if (rc)
1554                 return ERR_PTR(rc);
1555
1556         if ((fid_is_idif(fid) || fid_is_norm(fid) || fid_is_echo(fid)) &&
1557             osd->od_is_ost) {
1558                 /* The minimum block size must be at least page size otherwise
1559                  * it will break the assumption in tgt_thread_big_cache where
1560                  * the array size is PTLRPC_MAX_BRW_PAGES. It will also affect
1561                  * RDMA due to subpage transfer size */
1562                 rc = -dmu_object_set_blocksize(osd->od_os, dn->dn_object,
1563                                                PAGE_SIZE, 0, oh->ot_tx);
1564                 if (unlikely(rc)) {
1565                         CERROR("%s: can't change blocksize: %d\n",
1566                                osd->od_svname, rc);
1567                         return ERR_PTR(rc);
1568                 }
1569         }
1570
1571         return dn;
1572 }
1573
1574 static dnode_t *osd_mksym(const struct lu_env *env, struct osd_object *obj,
1575                           struct lu_attr *la, struct osd_thandle *oh)
1576 {
1577         dnode_t *dn;
1578         int rc;
1579
1580         LASSERT(S_ISLNK(la->la_mode));
1581         rc = __osd_object_create(env, obj, &dn, oh->ot_tx, la);
1582         if (rc)
1583                 return ERR_PTR(rc);
1584         return dn;
1585 }
1586
1587 static dnode_t *osd_mknod(const struct lu_env *env, struct osd_object *obj,
1588                           struct lu_attr *la, struct osd_thandle *oh)
1589 {
1590         dnode_t *dn;
1591         int rc;
1592
1593         if (S_ISCHR(la->la_mode) || S_ISBLK(la->la_mode))
1594                 la->la_valid |= LA_RDEV;
1595
1596         rc = __osd_object_create(env, obj, &dn, oh->ot_tx, la);
1597         if (rc)
1598                 return ERR_PTR(rc);
1599         return dn;
1600 }
1601
1602 typedef dnode_t *(*osd_obj_type_f)(const struct lu_env *env,
1603                                    struct osd_object *obj,
1604                                    struct lu_attr *la,
1605                                    struct osd_thandle *oh);
1606
1607 static osd_obj_type_f osd_create_type_f(enum dt_format_type type)
1608 {
1609         osd_obj_type_f result;
1610
1611         switch (type) {
1612         case DFT_DIR:
1613                 result = osd_mkdir;
1614                 break;
1615         case DFT_INDEX:
1616                 result = osd_mkidx;
1617                 break;
1618         case DFT_REGULAR:
1619                 result = osd_mkreg;
1620                 break;
1621         case DFT_SYM:
1622                 result = osd_mksym;
1623                 break;
1624         case DFT_NODE:
1625                 result = osd_mknod;
1626                 break;
1627         default:
1628                 LBUG();
1629                 break;
1630         }
1631         return result;
1632 }
1633
1634 /*
1635  * Concurrency: @dt is write locked.
1636  */
1637 static int osd_create(const struct lu_env *env, struct dt_object *dt,
1638                       struct lu_attr *attr, struct dt_allocation_hint *hint,
1639                       struct dt_object_format *dof, struct thandle *th)
1640 {
1641         struct osd_thread_info  *info = osd_oti_get(env);
1642         struct lustre_mdt_attrs *lma = &info->oti_mdt_attrs;
1643         struct zpl_direntry     *zde = &info->oti_zde.lzd_reg;
1644         const struct lu_fid     *fid = lu_object_fid(&dt->do_lu);
1645         struct osd_object       *obj = osd_dt_obj(dt);
1646         struct osd_device       *osd = osd_obj2dev(obj);
1647         char                    *buf = info->oti_str;
1648         struct osd_thandle      *oh;
1649         dnode_t *dn = NULL, *zdn = NULL;
1650         uint64_t                 zapid, parent = 0;
1651         int                      rc;
1652
1653         ENTRY;
1654
1655         LASSERT(!fid_is_acct(fid));
1656
1657         /* concurrent create declarations should not see
1658          * the object inconsistent (db, attr, etc).
1659          * in regular cases acquisition should be cheap */
1660         down_write(&obj->oo_guard);
1661
1662         if (unlikely(dt_object_exists(dt)))
1663                 GOTO(out, rc = -EEXIST);
1664
1665         LASSERT(osd_invariant(obj));
1666         LASSERT(dof != NULL);
1667
1668         LASSERT(th != NULL);
1669         oh = container_of0(th, struct osd_thandle, ot_super);
1670
1671         LASSERT(obj->oo_dn == NULL);
1672
1673         /* to follow ZFS on-disk format we need
1674          * to initialize parent dnode properly */
1675         if (hint != NULL && hint->dah_parent != NULL &&
1676             !dt_object_remote(hint->dah_parent))
1677                 parent = osd_dt_obj(hint->dah_parent)->oo_dn->dn_object;
1678
1679         /* we may fix some attributes, better do not change the source */
1680         obj->oo_attr = *attr;
1681         obj->oo_attr.la_valid |= LA_SIZE | LA_NLINK | LA_TYPE;
1682
1683 #ifdef ZFS_PROJINHERIT
1684         if (osd->od_projectused_dn) {
1685                 if (!(obj->oo_attr.la_valid & LA_PROJID))
1686                         obj->oo_attr.la_projid = ZFS_DEFAULT_PROJID;
1687                 obj->oo_with_projid = 1;
1688         }
1689 #endif
1690
1691         dn = osd_create_type_f(dof->dof_type)(env, obj, &obj->oo_attr, oh);
1692         if (IS_ERR(dn)) {
1693                 rc = PTR_ERR(dn);
1694                 dn = NULL;
1695                 GOTO(out, rc);
1696         }
1697
1698         zde->zde_pad = 0;
1699         zde->zde_dnode = dn->dn_object;
1700         zde->zde_type = IFTODT(attr->la_mode & S_IFMT);
1701
1702         zapid = osd_get_name_n_idx(env, osd, fid, buf,
1703                                    sizeof(info->oti_str), &zdn);
1704         rc = osd_zap_add(osd, zapid, zdn, buf, 8, 1, zde, oh->ot_tx);
1705         if (rc)
1706                 GOTO(out, rc);
1707         obj->oo_dn = dn;
1708         /* Now add in all of the "SA" attributes */
1709         rc = osd_sa_handle_get(obj);
1710         if (rc)
1711                 GOTO(out, rc);
1712
1713         rc = -nvlist_alloc(&obj->oo_sa_xattr, NV_UNIQUE_NAME, KM_SLEEP);
1714         if (rc)
1715                 GOTO(out, rc);
1716
1717         /* initialize LMA */
1718         lustre_lma_init(lma, fid, 0, 0);
1719         lustre_lma_swab(lma);
1720         rc = -nvlist_add_byte_array(obj->oo_sa_xattr, XATTR_NAME_LMA,
1721                                     (uchar_t *)lma, sizeof(*lma));
1722         if (rc)
1723                 GOTO(out, rc);
1724
1725         /* configure new osd object */
1726         obj->oo_parent = parent != 0 ? parent : zapid;
1727         obj->oo_late_attr_set = 1;
1728         rc = __osd_sa_xattr_schedule_update(env, obj, oh);
1729         if (rc)
1730                 GOTO(out, rc);
1731
1732         /* XXX: oo_lma_flags */
1733         obj->oo_dt.do_lu.lo_header->loh_attr |= obj->oo_attr.la_mode & S_IFMT;
1734         if (likely(!fid_is_acct(lu_object_fid(&obj->oo_dt.do_lu))))
1735                 /* no body operations for accounting objects */
1736                 obj->oo_dt.do_body_ops = &osd_body_ops;
1737
1738         osd_idc_find_and_init(env, osd, obj);
1739
1740 out:
1741         if (unlikely(rc && dn)) {
1742                 dmu_object_free(osd->od_os, dn->dn_object, oh->ot_tx);
1743                 osd_dnode_rele(dn);
1744                 obj->oo_dn = NULL;
1745         } else if (!rc) {
1746                 obj->oo_dt.do_lu.lo_header->loh_attr |= LOHA_EXISTS;
1747         }
1748         up_write(&obj->oo_guard);
1749         RETURN(rc);
1750 }
1751
1752 static int osd_declare_ref_add(const struct lu_env *env, struct dt_object *dt,
1753                                struct thandle *th)
1754 {
1755         return osd_declare_attr_set(env, dt, NULL, th);
1756 }
1757
1758 /*
1759  * Concurrency: @dt is write locked.
1760  */
1761 static int osd_ref_add(const struct lu_env *env, struct dt_object *dt,
1762                        struct thandle *handle)
1763 {
1764         struct osd_object       *obj = osd_dt_obj(dt);
1765         struct osd_thandle      *oh;
1766         struct osd_device       *osd = osd_obj2dev(obj);
1767         uint64_t                 nlink;
1768         int rc;
1769
1770         ENTRY;
1771
1772         down_read(&obj->oo_guard);
1773         if (unlikely(!dt_object_exists(dt) || obj->oo_destroyed))
1774                 GOTO(out, rc = -ENOENT);
1775
1776         LASSERT(osd_invariant(obj));
1777         LASSERT(obj->oo_sa_hdl != NULL);
1778
1779         oh = container_of0(handle, struct osd_thandle, ot_super);
1780
1781         write_lock(&obj->oo_attr_lock);
1782         nlink = ++obj->oo_attr.la_nlink;
1783         write_unlock(&obj->oo_attr_lock);
1784
1785         rc = osd_object_sa_update(obj, SA_ZPL_LINKS(osd), &nlink, 8, oh);
1786
1787 out:
1788         up_read(&obj->oo_guard);
1789         RETURN(rc);
1790 }
1791
1792 static int osd_declare_ref_del(const struct lu_env *env, struct dt_object *dt,
1793                                struct thandle *handle)
1794 {
1795         return osd_declare_attr_set(env, dt, NULL, handle);
1796 }
1797
1798 /*
1799  * Concurrency: @dt is write locked.
1800  */
1801 static int osd_ref_del(const struct lu_env *env, struct dt_object *dt,
1802                        struct thandle *handle)
1803 {
1804         struct osd_object       *obj = osd_dt_obj(dt);
1805         struct osd_thandle      *oh;
1806         struct osd_device       *osd = osd_obj2dev(obj);
1807         uint64_t                 nlink;
1808         int                      rc;
1809
1810         ENTRY;
1811
1812         down_read(&obj->oo_guard);
1813
1814         if (unlikely(!dt_object_exists(dt) || obj->oo_destroyed))
1815                 GOTO(out, rc = -ENOENT);
1816
1817         LASSERT(osd_invariant(obj));
1818         LASSERT(obj->oo_sa_hdl != NULL);
1819
1820         oh = container_of0(handle, struct osd_thandle, ot_super);
1821         LASSERT(!lu_object_is_dying(dt->do_lu.lo_header));
1822
1823         write_lock(&obj->oo_attr_lock);
1824         nlink = --obj->oo_attr.la_nlink;
1825         write_unlock(&obj->oo_attr_lock);
1826
1827         rc = osd_object_sa_update(obj, SA_ZPL_LINKS(osd), &nlink, 8, oh);
1828
1829 out:
1830         up_read(&obj->oo_guard);
1831         RETURN(rc);
1832 }
1833
1834 static int osd_object_sync(const struct lu_env *env, struct dt_object *dt,
1835                            __u64 start, __u64 end)
1836 {
1837         struct osd_device *osd = osd_obj2dev(osd_dt_obj(dt));
1838         ENTRY;
1839
1840         /* XXX: no other option than syncing the whole filesystem until we
1841          * support ZIL.  If the object tracked the txg that it was last
1842          * modified in, it could pass that txg here instead of "0".  Maybe
1843          * the changes are already committed, so no wait is needed at all? */
1844         if (!osd->od_dt_dev.dd_rdonly)
1845                 txg_wait_synced(dmu_objset_pool(osd->od_os), 0ULL);
1846
1847         RETURN(0);
1848 }
1849
1850 static int osd_invalidate(const struct lu_env *env, struct dt_object *dt)
1851 {
1852         return 0;
1853 }
1854
1855 static struct dt_object_operations osd_obj_ops = {
1856         .do_read_lock           = osd_read_lock,
1857         .do_write_lock          = osd_write_lock,
1858         .do_read_unlock         = osd_read_unlock,
1859         .do_write_unlock        = osd_write_unlock,
1860         .do_write_locked        = osd_write_locked,
1861         .do_attr_get            = osd_attr_get,
1862         .do_declare_attr_set    = osd_declare_attr_set,
1863         .do_attr_set            = osd_attr_set,
1864         .do_ah_init             = osd_ah_init,
1865         .do_declare_create      = osd_declare_create,
1866         .do_create              = osd_create,
1867         .do_declare_destroy     = osd_declare_destroy,
1868         .do_destroy             = osd_destroy,
1869         .do_index_try           = osd_index_try,
1870         .do_declare_ref_add     = osd_declare_ref_add,
1871         .do_ref_add             = osd_ref_add,
1872         .do_declare_ref_del     = osd_declare_ref_del,
1873         .do_ref_del             = osd_ref_del,
1874         .do_xattr_get           = osd_xattr_get,
1875         .do_declare_xattr_set   = osd_declare_xattr_set,
1876         .do_xattr_set           = osd_xattr_set,
1877         .do_declare_xattr_del   = osd_declare_xattr_del,
1878         .do_xattr_del           = osd_xattr_del,
1879         .do_xattr_list          = osd_xattr_list,
1880         .do_object_sync         = osd_object_sync,
1881         .do_invalidate          = osd_invalidate,
1882 };
1883
1884 static struct lu_object_operations osd_lu_obj_ops = {
1885         .loo_object_init        = osd_object_init,
1886         .loo_object_delete      = osd_object_delete,
1887         .loo_object_release     = osd_object_release,
1888         .loo_object_free        = osd_object_free,
1889         .loo_object_print       = osd_object_print,
1890         .loo_object_invariant   = osd_object_invariant,
1891 };
1892
1893 static int osd_otable_it_attr_get(const struct lu_env *env,
1894                                 struct dt_object *dt,
1895                                 struct lu_attr *attr)
1896 {
1897         attr->la_valid = 0;
1898         return 0;
1899 }
1900
1901 static struct dt_object_operations osd_obj_otable_it_ops = {
1902         .do_attr_get            = osd_otable_it_attr_get,
1903         .do_index_try           = osd_index_try,
1904 };