module/zfs/dmu_object.c

   1 /*
   2  * CDDL HEADER START
   3  *
   4  * The contents of this file are subject to the terms of the
   5  * Common Development and Distribution License (the "License").
   6  * You may not use this file except in compliance with the License.
   7  *
   8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
   9  * or http://www.opensolaris.org/os/licensing.
  10  * See the License for the specific language governing permissions
  11  * and limitations under the License.
  12  *
  13  * When distributing Covered Code, include this CDDL HEADER in each
  14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
  15  * If applicable, add the following below this CDDL HEADER, with the
  16  * fields enclosed by brackets "[]" replaced with your own identifying
  17  * information: Portions Copyright [yyyy] [name of copyright owner]
  18  *
  19  * CDDL HEADER END
  20  */
  21 /*
  22  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
  23  * Copyright (c) 2013, 2014 by Delphix. All rights reserved.
  24  * Copyright 2014 HybridCluster. All rights reserved.
  25  */
  26
  27 #include <sys/dmu.h>
  28 #include <sys/dmu_objset.h>
  29 #include <sys/dmu_tx.h>
  30 #include <sys/dnode.h>
  31 #include <sys/zap.h>
  32 #include <sys/zfeature.h>
  33
  34 uint64_t
  35 dmu_object_alloc(objset_t *os, dmu_object_type_t ot, int blocksize,
  36     dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx)
  37 {
  38         uint64_t object;
  39         uint64_t L2_dnode_count = DNODES_PER_BLOCK <<
  40             (DMU_META_DNODE(os)->dn_indblkshift - SPA_BLKPTRSHIFT);
  41         dnode_t *dn = NULL;
  42         int restarted = B_FALSE;
  43
  44         mutex_enter(&os->os_obj_lock);
  45         for (;;) {
  46                 object = os->os_obj_next;
  47                 /*
  48                  * Each time we polish off an L2 bp worth of dnodes
  49                  * (2^13 objects), move to another L2 bp that's still
  50                  * reasonably sparse (at most 1/4 full).  Look from the
  51                  * beginning once, but after that keep looking from here.
  52                  * If we can't find one, just keep going from here.
  53                  */
  54                 if (P2PHASE(object, L2_dnode_count) == 0) {
  55                         uint64_t offset = restarted ? object << DNODE_SHIFT : 0;
  56                         int error = dnode_next_offset(DMU_META_DNODE(os),
  57                             DNODE_FIND_HOLE,
  58                             &offset, 2, DNODES_PER_BLOCK >> 2, 0);
  59                         restarted = B_TRUE;
  60                         if (error == 0)
  61                                 object = offset >> DNODE_SHIFT;
  62                 }
  63                 os->os_obj_next = ++object;
  64
  65                 /*
  66                  * XXX We should check for an i/o error here and return
  67                  * up to our caller.  Actually we should pre-read it in
  68                  * dmu_tx_assign(), but there is currently no mechanism
  69                  * to do so.
  70                  */
  71                 (void) dnode_hold_impl(os, object, DNODE_MUST_BE_FREE,
  72                     FTAG, &dn);
  73                 if (dn)
  74                         break;
  75
  76                 if (dmu_object_next(os, &object, B_TRUE, 0) == 0)
  77                         os->os_obj_next = object - 1;
  78         }
  79
  80         dnode_allocate(dn, ot, blocksize, 0, bonustype, bonuslen, tx);
  81         dnode_rele(dn, FTAG);
  82
  83         mutex_exit(&os->os_obj_lock);
  84
  85         dmu_tx_add_new_object(tx, os, object);
  86         return (object);
  87 }
  88
  89 int
  90 dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot,
  91     int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx)
  92 {
  93         dnode_t *dn;
  94         int err;
  95
  96         if (object == DMU_META_DNODE_OBJECT && !dmu_tx_private_ok(tx))
  97                 return (SET_ERROR(EBADF));
  98
  99         err = dnode_hold_impl(os, object, DNODE_MUST_BE_FREE, FTAG, &dn);
 100         if (err)
 101                 return (err);
 102         dnode_allocate(dn, ot, blocksize, 0, bonustype, bonuslen, tx);
 103         dnode_rele(dn, FTAG);
 104
 105         dmu_tx_add_new_object(tx, os, object);
 106         return (0);
 107 }
 108
 109 int
 110 dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot,
 111     int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx)
 112 {
 113         dnode_t *dn;
 114         int err;
 115
 116         if (object == DMU_META_DNODE_OBJECT)
 117                 return (SET_ERROR(EBADF));
 118
 119         err = dnode_hold_impl(os, object, DNODE_MUST_BE_ALLOCATED,
 120             FTAG, &dn);
 121         if (err)
 122                 return (err);
 123
 124         dnode_reallocate(dn, ot, blocksize, bonustype, bonuslen, tx);
 125
 126         dnode_rele(dn, FTAG);
 127         return (err);
 128 }
 129
 130 int
 131 dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx)
 132 {
 133         dnode_t *dn;
 134         int err;
 135
 136         ASSERT(object != DMU_META_DNODE_OBJECT || dmu_tx_private_ok(tx));
 137
 138         err = dnode_hold_impl(os, object, DNODE_MUST_BE_ALLOCATED,
 139             FTAG, &dn);
 140         if (err)
 141                 return (err);
 142
 143         ASSERT(dn->dn_type != DMU_OT_NONE);
 144         dnode_free_range(dn, 0, DMU_OBJECT_END, tx);
 145         dnode_free(dn, tx);
 146         dnode_rele(dn, FTAG);
 147
 148         return (0);
 149 }
 150
 151 int
 152 dmu_object_next(objset_t *os, uint64_t *objectp, boolean_t hole, uint64_t txg)
 153 {
 154         uint64_t offset = (*objectp + 1) << DNODE_SHIFT;
 155         int error;
 156
 157         error = dnode_next_offset(DMU_META_DNODE(os),
 158             (hole ? DNODE_FIND_HOLE : 0), &offset, 0, DNODES_PER_BLOCK, txg);
 159
 160         *objectp = offset >> DNODE_SHIFT;
 161
 162         return (error);
 163 }
 164
 165 /*
 166  * Turn this object from old_type into DMU_OTN_ZAP_METADATA, and bump the
 167  * refcount on SPA_FEATURE_EXTENSIBLE_DATASET.
 168  *
 169  * Only for use from syncing context, on MOS objects.
 170  */
 171 void
 172 dmu_object_zapify(objset_t *mos, uint64_t object, dmu_object_type_t old_type,
 173     dmu_tx_t *tx)
 174 {
 175         dnode_t *dn;
 176
 177         ASSERT(dmu_tx_is_syncing(tx));
 178
 179         VERIFY0(dnode_hold(mos, object, FTAG, &dn));
 180         if (dn->dn_type == DMU_OTN_ZAP_METADATA) {
 181                 dnode_rele(dn, FTAG);
 182                 return;
 183         }
 184         ASSERT3U(dn->dn_type, ==, old_type);
 185         ASSERT0(dn->dn_maxblkid);
 186         dn->dn_next_type[tx->tx_txg & TXG_MASK] = dn->dn_type =
 187             DMU_OTN_ZAP_METADATA;
 188         dnode_setdirty(dn, tx);
 189         dnode_rele(dn, FTAG);
 190
 191         mzap_create_impl(mos, object, 0, 0, tx);
 192
 193         spa_feature_incr(dmu_objset_spa(mos),
 194             SPA_FEATURE_EXTENSIBLE_DATASET, tx);
 195 }
 196
 197 void
 198 dmu_object_free_zapified(objset_t *mos, uint64_t object, dmu_tx_t *tx)
 199 {
 200         dnode_t *dn;
 201         dmu_object_type_t t;
 202
 203         ASSERT(dmu_tx_is_syncing(tx));
 204
 205         VERIFY0(dnode_hold(mos, object, FTAG, &dn));
 206         t = dn->dn_type;
 207         dnode_rele(dn, FTAG);
 208
 209         if (t == DMU_OTN_ZAP_METADATA) {
 210                 spa_feature_decr(dmu_objset_spa(mos),
 211                     SPA_FEATURE_EXTENSIBLE_DATASET, tx);
 212         }
 213         VERIFY0(dmu_object_free(mos, object, tx));
 214 }
 215
 216 #if defined(_KERNEL) && defined(HAVE_SPL)
 217 EXPORT_SYMBOL(dmu_object_alloc);
 218 EXPORT_SYMBOL(dmu_object_claim);
 219 EXPORT_SYMBOL(dmu_object_reclaim);
 220 EXPORT_SYMBOL(dmu_object_free);
 221 EXPORT_SYMBOL(dmu_object_next);
 222 EXPORT_SYMBOL(dmu_object_zapify);
 223 EXPORT_SYMBOL(dmu_object_free_zapified);
 224 #endif