module/zfs/vdev_indirect_births.c

   1 /*
   2  * CDDL HEADER START
   3  *
   4  * This file and its contents are supplied under the terms of the
   5  * Common Development and Distribution License ("CDDL"), version 1.0.
   6  * You may only use this file in accordance with the terms of version
   7  * 1.0 of the CDDL.
   8  *
   9  * A full copy of the text of the CDDL should have accompanied this
  10  * source.  A copy of the CDDL is also available via the Internet at
  11  * http://www.illumos.org/license/CDDL.
  12  *
  13  * CDDL HEADER END
  14  */
  15
  16 /*
  17  * Copyright (c) 2015 by Delphix. All rights reserved.
  18  */
  19
  20 #include <sys/dmu_tx.h>
  21 #include <sys/spa.h>
  22 #include <sys/dmu.h>
  23 #include <sys/dsl_pool.h>
  24 #include <sys/vdev_indirect_births.h>
  25
  26 #ifdef ZFS_DEBUG
  27 static boolean_t
  28 vdev_indirect_births_verify(vdev_indirect_births_t *vib)
  29 {
  30         ASSERT(vib != NULL);
  31
  32         ASSERT(vib->vib_object != 0);
  33         ASSERT(vib->vib_objset != NULL);
  34         ASSERT(vib->vib_phys != NULL);
  35         ASSERT(vib->vib_dbuf != NULL);
  36
  37         EQUIV(vib->vib_phys->vib_count > 0, vib->vib_entries != NULL);
  38
  39         return (B_TRUE);
  40 }
  41 #endif
  42
  43 uint64_t
  44 vdev_indirect_births_count(vdev_indirect_births_t *vib)
  45 {
  46         ASSERT(vdev_indirect_births_verify(vib));
  47
  48         return (vib->vib_phys->vib_count);
  49 }
  50
  51 uint64_t
  52 vdev_indirect_births_object(vdev_indirect_births_t *vib)
  53 {
  54         ASSERT(vdev_indirect_births_verify(vib));
  55
  56         return (vib->vib_object);
  57 }
  58
  59 static uint64_t
  60 vdev_indirect_births_size_impl(vdev_indirect_births_t *vib)
  61 {
  62         return (vib->vib_phys->vib_count * sizeof (*vib->vib_entries));
  63 }
  64
  65 void
  66 vdev_indirect_births_close(vdev_indirect_births_t *vib)
  67 {
  68         ASSERT(vdev_indirect_births_verify(vib));
  69
  70         if (vib->vib_phys->vib_count > 0) {
  71                 uint64_t births_size = vdev_indirect_births_size_impl(vib);
  72
  73                 kmem_free(vib->vib_entries, births_size);
  74                 vib->vib_entries = NULL;
  75         }
  76
  77         dmu_buf_rele(vib->vib_dbuf, vib);
  78
  79         vib->vib_objset = NULL;
  80         vib->vib_object = 0;
  81         vib->vib_dbuf = NULL;
  82         vib->vib_phys = NULL;
  83
  84         kmem_free(vib, sizeof (*vib));
  85 }
  86
  87 uint64_t
  88 vdev_indirect_births_alloc(objset_t *os, dmu_tx_t *tx)
  89 {
  90         ASSERT(dmu_tx_is_syncing(tx));
  91
  92         return (dmu_object_alloc(os,
  93             DMU_OTN_UINT64_METADATA, SPA_OLD_MAXBLOCKSIZE,
  94             DMU_OTN_UINT64_METADATA, sizeof (vdev_indirect_birth_phys_t),
  95             tx));
  96 }
  97
  98 vdev_indirect_births_t *
  99 vdev_indirect_births_open(objset_t *os, uint64_t births_object)
 100 {
 101         vdev_indirect_births_t *vib = kmem_zalloc(sizeof (*vib), KM_SLEEP);
 102
 103         vib->vib_objset = os;
 104         vib->vib_object = births_object;
 105
 106         VERIFY0(dmu_bonus_hold(os, vib->vib_object, vib, &vib->vib_dbuf));
 107         vib->vib_phys = vib->vib_dbuf->db_data;
 108
 109         if (vib->vib_phys->vib_count > 0) {
 110                 uint64_t births_size = vdev_indirect_births_size_impl(vib);
 111                 vib->vib_entries = kmem_alloc(births_size, KM_SLEEP);
 112                 VERIFY0(dmu_read(vib->vib_objset, vib->vib_object, 0,
 113                     births_size, vib->vib_entries, DMU_READ_PREFETCH));
 114         }
 115
 116         ASSERT(vdev_indirect_births_verify(vib));
 117
 118         return (vib);
 119 }
 120
 121 void
 122 vdev_indirect_births_free(objset_t *os, uint64_t object, dmu_tx_t *tx)
 123 {
 124         VERIFY0(dmu_object_free(os, object, tx));
 125 }
 126
 127 void
 128 vdev_indirect_births_add_entry(vdev_indirect_births_t *vib,
 129     uint64_t max_offset, uint64_t txg, dmu_tx_t *tx)
 130 {
 131         vdev_indirect_birth_entry_phys_t vibe;
 132         uint64_t old_size;
 133         uint64_t new_size;
 134         vdev_indirect_birth_entry_phys_t *new_entries;
 135
 136         ASSERT(dmu_tx_is_syncing(tx));
 137         ASSERT(dsl_pool_sync_context(dmu_tx_pool(tx)));
 138         ASSERT(vdev_indirect_births_verify(vib));
 139
 140         dmu_buf_will_dirty(vib->vib_dbuf, tx);
 141
 142         vibe.vibe_offset = max_offset;
 143         vibe.vibe_phys_birth_txg = txg;
 144
 145         old_size = vdev_indirect_births_size_impl(vib);
 146         dmu_write(vib->vib_objset, vib->vib_object, old_size, sizeof (vibe),
 147             &vibe, tx);
 148         vib->vib_phys->vib_count++;
 149         new_size = vdev_indirect_births_size_impl(vib);
 150
 151         new_entries = kmem_alloc(new_size, KM_SLEEP);
 152         if (old_size > 0) {
 153                 bcopy(vib->vib_entries, new_entries, old_size);
 154                 kmem_free(vib->vib_entries, old_size);
 155         }
 156         new_entries[vib->vib_phys->vib_count - 1] = vibe;
 157         vib->vib_entries = new_entries;
 158 }
 159
 160 uint64_t
 161 vdev_indirect_births_last_entry_txg(vdev_indirect_births_t *vib)
 162 {
 163         ASSERT(vdev_indirect_births_verify(vib));
 164         ASSERT(vib->vib_phys->vib_count > 0);
 165
 166         vdev_indirect_birth_entry_phys_t *last =
 167             &vib->vib_entries[vib->vib_phys->vib_count - 1];
 168         return (last->vibe_phys_birth_txg);
 169 }
 170
 171 /*
 172  * Return the txg in which the given range was copied (i.e. its physical
 173  * birth txg).  The specified offset+asize must be contiguously mapped
 174  * (i.e. not a split block).
 175  *
 176  * The entries are sorted by increasing phys_birth, and also by increasing
 177  * offset.  We find the specified offset by binary search.  Note that we
 178  * can not use bsearch() because looking at each entry independently is
 179  * insufficient to find the correct entry.  Each entry implicitly relies
 180  * on the previous entry: an entry indicates that the offsets from the
 181  * end of the previous entry to the end of this entry were written in the
 182  * specified txg.
 183  */
 184 uint64_t
 185 vdev_indirect_births_physbirth(vdev_indirect_births_t *vib, uint64_t offset,
 186     uint64_t asize)
 187 {
 188         vdev_indirect_birth_entry_phys_t *base;
 189         vdev_indirect_birth_entry_phys_t *last;
 190
 191         ASSERT(vdev_indirect_births_verify(vib));
 192         ASSERT(vib->vib_phys->vib_count > 0);
 193
 194         base = vib->vib_entries;
 195         last = base + vib->vib_phys->vib_count - 1;
 196
 197         ASSERT3U(offset, <, last->vibe_offset);
 198
 199         while (last >= base) {
 200                 vdev_indirect_birth_entry_phys_t *p =
 201                     base + ((last - base) / 2);
 202                 if (offset >= p->vibe_offset) {
 203                         base = p + 1;
 204                 } else if (p == vib->vib_entries ||
 205                     offset >= (p - 1)->vibe_offset) {
 206                         ASSERT3U(offset + asize, <=, p->vibe_offset);
 207                         return (p->vibe_phys_birth_txg);
 208                 } else {
 209                         last = p - 1;
 210                 }
 211         }
 212         ASSERT(!"offset not found");
 213         return (-1);
 214 }
 215
 216 #if defined(_KERNEL)
 217 EXPORT_SYMBOL(vdev_indirect_births_add_entry);
 218 EXPORT_SYMBOL(vdev_indirect_births_alloc);
 219 EXPORT_SYMBOL(vdev_indirect_births_close);
 220 EXPORT_SYMBOL(vdev_indirect_births_count);
 221 EXPORT_SYMBOL(vdev_indirect_births_free);
 222 EXPORT_SYMBOL(vdev_indirect_births_last_entry_txg);
 223 EXPORT_SYMBOL(vdev_indirect_births_object);
 224 EXPORT_SYMBOL(vdev_indirect_births_open);
 225 EXPORT_SYMBOL(vdev_indirect_births_physbirth);
 226 #endif