include/sys/dmu_impl.h

   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 https://opensource.org/licenses/CDDL-1.0.
  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 2010 Sun Microsystems, Inc.  All rights reserved.
  23  * Use is subject to license terms.
  24  */
  25 /*
  26  * Copyright (c) 2012, Joyent, Inc. All rights reserved.
  27  * Copyright (c) 2013, 2018 by Delphix. All rights reserved.
  28  */
  29
  30 #ifndef _SYS_DMU_IMPL_H
  31 #define _SYS_DMU_IMPL_H
  32
  33 #include <sys/txg_impl.h>
  34 #include <sys/zio.h>
  35 #include <sys/dnode.h>
  36 #include <sys/zfs_context.h>
  37 #include <sys/zfs_ioctl.h>
  38 #include <sys/uio.h>
  39 #include <sys/abd.h>
  40 #include <sys/arc.h>
  41 #include <sys/dbuf.h>
  42
  43 #ifdef  __cplusplus
  44 extern "C" {
  45 #endif
  46
  47 /*
  48  * This is the locking strategy for the DMU.  Numbers in parenthesis are
  49  * cases that use that lock order, referenced below:
  50  *
  51  * ARC is self-contained
  52  * bplist is self-contained
  53  * refcount is self-contained
  54  * txg is self-contained (hopefully!)
  55  * zst_lock
  56  * zf_rwlock
  57  *
  58  * XXX try to improve evicting path?
  59  *
  60  * dp_config_rwlock > os_obj_lock > dn_struct_rwlock >
  61  *      dn_dbufs_mtx > hash_mutexes > db_mtx > dd_lock > leafs
  62  *
  63  * dp_config_rwlock
  64  *    must be held before: everything
  65  *    protects dd namespace changes
  66  *    protects property changes globally
  67  *    held from:
  68  *      dsl_dir_open/r:
  69  *      dsl_dir_create_sync/w:
  70  *      dsl_dir_sync_destroy/w:
  71  *      dsl_dir_rename_sync/w:
  72  *      dsl_prop_changed_notify/r:
  73  *
  74  * os_obj_lock
  75  *   must be held before:
  76  *      everything except dp_config_rwlock
  77  *   protects os_obj_next
  78  *   held from:
  79  *      dmu_object_alloc: dn_dbufs_mtx, db_mtx, hash_mutexes, dn_struct_rwlock
  80  *
  81  * dn_struct_rwlock
  82  *   must be held before:
  83  *      everything except dp_config_rwlock and os_obj_lock
  84  *   protects structure of dnode (eg. nlevels)
  85  *      db_blkptr can change when syncing out change to nlevels
  86  *      dn_maxblkid
  87  *      dn_nlevels
  88  *      dn_*blksz*
  89  *      phys nlevels, maxblkid, physical blkptr_t's (?)
  90  *   held from:
  91  *      callers of dbuf_read_impl, dbuf_hold[_impl], dbuf_prefetch
  92  *      dmu_object_info_from_dnode: dn_dirty_mtx (dn_datablksz)
  93  *      dbuf_read_impl: db_mtx, dmu_zfetch()
  94  *      dmu_zfetch: zf_rwlock/r, zst_lock, dbuf_prefetch()
  95  *      dbuf_new_size: db_mtx
  96  *      dbuf_dirty: db_mtx
  97  *      dbuf_findbp: (callers, phys? - the real need)
  98  *      dbuf_create: dn_dbufs_mtx, hash_mutexes, db_mtx (phys?)
  99  *      dbuf_prefetch: dn_dirty_mtx, hash_mutexes, db_mtx, dn_dbufs_mtx
 100  *      dbuf_hold_impl: hash_mutexes, db_mtx, dn_dbufs_mtx, dbuf_findbp()
 101  *      dnode_sync/w (increase_indirection): db_mtx (phys)
 102  *      dnode_set_blksz/w: dn_dbufs_mtx (dn_*blksz*)
 103  *      dnode_new_blkid/w: (dn_maxblkid)
 104  *      dnode_free_range/w: dn_dirty_mtx (dn_maxblkid)
 105  *      dnode_next_offset: (phys)
 106  *
 107  * dn_dbufs_mtx
 108  *    must be held before:
 109  *      db_mtx, hash_mutexes
 110  *    protects:
 111  *      dn_dbufs
 112  *      dn_evicted
 113  *    held from:
 114  *      dmu_evict_user: db_mtx (dn_dbufs)
 115  *      dbuf_free_range: db_mtx (dn_dbufs)
 116  *      dbuf_remove_ref: db_mtx, callees:
 117  *              dbuf_hash_remove: hash_mutexes, db_mtx
 118  *      dbuf_create: hash_mutexes, db_mtx (dn_dbufs)
 119  *      dnode_set_blksz: (dn_dbufs)
 120  *
 121  * hash_mutexes (global)
 122  *   must be held before:
 123  *      db_mtx
 124  *   protects dbuf_hash_table (global) and db_hash_next
 125  *   held from:
 126  *      dbuf_find: db_mtx
 127  *      dbuf_hash_insert: db_mtx
 128  *      dbuf_hash_remove: db_mtx
 129  *
 130  * db_mtx (meta-leaf)
 131  *   must be held before:
 132  *      dn_mtx, dn_dirty_mtx, dd_lock (leaf mutexes)
 133  *   protects:
 134  *      db_state
 135  *      db_holds
 136  *      db_buf
 137  *      db_changed
 138  *      db_data_pending
 139  *      db_dirtied
 140  *      db_link
 141  *      db_dirty_records
 142  *      db_dirtycnt
 143  *      db_d.*
 144  *      db.*
 145  *   held from:
 146  *      dbuf_dirty: dn_mtx, dn_dirty_mtx
 147  *      dbuf_dirty->dsl_dir_willuse_space: dd_lock
 148  *      dbuf_dirty->dbuf_new_block->dsl_dataset_block_freeable: dd_lock
 149  *      dbuf_undirty: dn_dirty_mtx (db_d)
 150  *      dbuf_write_done: dn_dirty_mtx (db_state)
 151  *      dbuf_*
 152  *      dmu_buf_update_user: none (db_d)
 153  *      dmu_evict_user: none (db_d) (maybe can eliminate)
 154  *      dbuf_find: none (db_holds)
 155  *      dbuf_hash_insert: none (db_holds)
 156  *      dmu_buf_read_array_impl: none (db_state, db_changed)
 157  *      dmu_sync: none (db_dirty_records, db_d)
 158  *      dnode_reallocate: none (db)
 159  *      dmu_write_direct: none (db_dirty_records, db_d)
 160  *      dmu_write_direct_done: none (db_dirty_records, db_d)
 161  *
 162  * dn_mtx (leaf)
 163  *   protects:
 164  *      dn_dirty_dbufs
 165  *      dn_ranges
 166  *      phys accounting
 167  *      dn_allocated_txg
 168  *      dn_free_txg
 169  *      dn_assigned_txg
 170  *      dn_dirty_txg
 171  *      dd_assigned_tx
 172  *      dn_notxholds
 173  *      dn_nodnholds
 174  *      dn_dirtyctx
 175  *      dn_dirtyctx_firstset
 176  *      (dn_phys copy fields?)
 177  *      (dn_phys contents?)
 178  *   held from:
 179  *      dnode_*
 180  *      dbuf_dirty: none
 181  *      dbuf_sync: none (phys accounting)
 182  *      dbuf_undirty: none (dn_ranges, dn_dirty_dbufs)
 183  *      dbuf_write_done: none (phys accounting)
 184  *      dmu_object_info_from_dnode: none (accounting)
 185  *      dmu_tx_commit: none
 186  *      dmu_tx_hold_object_impl: none
 187  *      dmu_tx_try_assign: dn_notxholds(cv)
 188  *      dmu_tx_unassign: none
 189  *
 190  * dd_lock
 191  *    must be held before:
 192  *      ds_lock
 193  *      ancestors' dd_lock
 194  *    protects:
 195  *      dd_prop_cbs
 196  *      dd_sync_*
 197  *      dd_used_bytes
 198  *      dd_tempreserved
 199  *      dd_space_towrite
 200  *      dd_myname
 201  *      dd_phys accounting?
 202  *    held from:
 203  *      dsl_dir_*
 204  *      dsl_prop_changed_notify: none (dd_prop_cbs)
 205  *      dsl_prop_register: none (dd_prop_cbs)
 206  *      dsl_prop_unregister: none (dd_prop_cbs)
 207  *
 208  * os_lock (leaf)
 209  *   protects:
 210  *      os_dirty_dnodes
 211  *      os_free_dnodes
 212  *      os_dnodes
 213  *      os_downgraded_dbufs
 214  *      dn_dirtyblksz
 215  *      dn_dirty_link
 216  *   held from:
 217  *      dnode_create: none (os_dnodes)
 218  *      dnode_destroy: none (os_dnodes)
 219  *      dnode_setdirty: none (dn_dirtyblksz, os_*_dnodes)
 220  *      dnode_free: none (dn_dirtyblksz, os_*_dnodes)
 221  *
 222  * ds_lock
 223  *    protects:
 224  *      ds_objset
 225  *      ds_open_refcount
 226  *      ds_snapname
 227  *      ds_phys accounting
 228  *      ds_phys userrefs zapobj
 229  *      ds_reserved
 230  *    held from:
 231  *      dsl_dataset_*
 232  *
 233  * dr_mtx (leaf)
 234  *    protects:
 235  *      dr_children
 236  *    held from:
 237  *      dbuf_dirty
 238  *      dbuf_undirty
 239  *      dbuf_sync_indirect
 240  *      dnode_new_blkid
 241  */
 242
 243 struct dmu_pool;
 244 struct dmu_buf;
 245 struct zgd;
 246
 247 typedef struct dmu_sendstatus {
 248         list_node_t dss_link;
 249         int dss_outfd;
 250         proc_t *dss_proc;
 251         offset_t *dss_off;
 252         uint64_t dss_blocks; /* blocks visited during the sending process */
 253 } dmu_sendstatus_t;
 254
 255 /*
 256  * dmu_sync_{ready/done} args
 257  */
 258 typedef struct {
 259         dbuf_dirty_record_t     *dsa_dr;
 260         void (*dsa_done)(struct zgd *, int);
 261         struct zgd              *dsa_zgd;
 262         dmu_tx_t                *dsa_tx;
 263 } dmu_sync_arg_t;
 264
 265 void dmu_sync_done(zio_t *, arc_buf_t *buf, void *varg);
 266 void dmu_sync_ready(zio_t *, arc_buf_t *buf, void *varg);
 267
 268 void dmu_object_zapify(objset_t *, uint64_t, dmu_object_type_t, dmu_tx_t *);
 269 void dmu_object_free_zapified(objset_t *, uint64_t, dmu_tx_t *);
 270
 271 int dmu_write_direct(zio_t *, dmu_buf_impl_t *, abd_t *, dmu_tx_t *);
 272 int dmu_read_abd(dnode_t *, uint64_t, uint64_t, abd_t *, uint32_t flags);
 273 int dmu_write_abd(dnode_t *, uint64_t, uint64_t, abd_t *, uint32_t, dmu_tx_t *);
 274 #if defined(_KERNEL)
 275 int dmu_read_uio_direct(dnode_t *, zfs_uio_t *, uint64_t);
 276 int dmu_write_uio_direct(dnode_t *, zfs_uio_t *, uint64_t, dmu_tx_t *);
 277 #endif
 278
 279 #ifdef  __cplusplus
 280 }
 281 #endif
 282
 283 #endif  /* _SYS_DMU_IMPL_H */