granicus.if.org Git - zfs/blob - include/sys/vdev_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 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) 2011, 2019 by Delphix. All rights reserved.
  24  * Copyright (c) 2017, Intel Corporation.
  25  */
  26
  27 #ifndef _SYS_VDEV_IMPL_H
  28 #define _SYS_VDEV_IMPL_H
  29
  30 #include <sys/avl.h>
  31 #include <sys/bpobj.h>
  32 #include <sys/dmu.h>
  33 #include <sys/metaslab.h>
  34 #include <sys/nvpair.h>
  35 #include <sys/space_map.h>
  36 #include <sys/vdev.h>
  37 #include <sys/dkio.h>
  38 #include <sys/uberblock_impl.h>
  39 #include <sys/vdev_indirect_mapping.h>
  40 #include <sys/vdev_indirect_births.h>
  41 #include <sys/vdev_removal.h>
  42 #include <sys/zfs_ratelimit.h>
  43
  44 #ifdef  __cplusplus
  45 extern "C" {
  46 #endif
  47
  48 /*
  49  * Virtual device descriptors.
  50  *
  51  * All storage pool operations go through the virtual device framework,
  52  * which provides data replication and I/O scheduling.
  53  */
  54
  55 /*
  56  * Forward declarations that lots of things need.
  57  */
  58 typedef struct vdev_queue vdev_queue_t;
  59 typedef struct vdev_cache vdev_cache_t;
  60 typedef struct vdev_cache_entry vdev_cache_entry_t;
  61 struct abd;
  62
  63 extern int zfs_vdev_queue_depth_pct;
  64 extern int zfs_vdev_def_queue_depth;
  65 extern uint32_t zfs_vdev_async_write_max_active;
  66
  67 /*
  68  * Virtual device operations
  69  */
  70 typedef int     vdev_open_func_t(vdev_t *vd, uint64_t *size, uint64_t *max_size,
  71     uint64_t *ashift);
  72 typedef void    vdev_close_func_t(vdev_t *vd);
  73 typedef uint64_t vdev_asize_func_t(vdev_t *vd, uint64_t psize);
  74 typedef void    vdev_io_start_func_t(zio_t *zio);
  75 typedef void    vdev_io_done_func_t(zio_t *zio);
  76 typedef void    vdev_state_change_func_t(vdev_t *vd, int, int);
  77 typedef boolean_t vdev_need_resilver_func_t(vdev_t *vd, uint64_t, size_t);
  78 typedef void    vdev_hold_func_t(vdev_t *vd);
  79 typedef void    vdev_rele_func_t(vdev_t *vd);
  80
  81 typedef void    vdev_remap_cb_t(uint64_t inner_offset, vdev_t *vd,
  82     uint64_t offset, uint64_t size, void *arg);
  83 typedef void    vdev_remap_func_t(vdev_t *vd, uint64_t offset, uint64_t size,
  84     vdev_remap_cb_t callback, void *arg);
  85 /*
  86  * Given a target vdev, translates the logical range "in" to the physical
  87  * range "res"
  88  */
  89 typedef void vdev_xlation_func_t(vdev_t *cvd, const range_seg_t *in,
  90     range_seg_t *res);
  91
  92 typedef const struct vdev_ops {
  93         vdev_open_func_t                *vdev_op_open;
  94         vdev_close_func_t               *vdev_op_close;
  95         vdev_asize_func_t               *vdev_op_asize;
  96         vdev_io_start_func_t            *vdev_op_io_start;
  97         vdev_io_done_func_t             *vdev_op_io_done;
  98         vdev_state_change_func_t        *vdev_op_state_change;
  99         vdev_need_resilver_func_t       *vdev_op_need_resilver;
 100         vdev_hold_func_t                *vdev_op_hold;
 101         vdev_rele_func_t                *vdev_op_rele;
 102         vdev_remap_func_t               *vdev_op_remap;
 103         /*
 104          * For translating ranges from non-leaf vdevs (e.g. raidz) to leaves.
 105          * Used when initializing vdevs. Isn't used by leaf ops.
 106          */
 107         vdev_xlation_func_t             *vdev_op_xlate;
 108         char                            vdev_op_type[16];
 109         boolean_t                       vdev_op_leaf;
 110 } vdev_ops_t;
 111
 112 /*
 113  * Virtual device properties
 114  */
 115 struct vdev_cache_entry {
 116         struct abd      *ve_abd;
 117         uint64_t        ve_offset;
 118         clock_t         ve_lastused;
 119         avl_node_t      ve_offset_node;
 120         avl_node_t      ve_lastused_node;
 121         uint32_t        ve_hits;
 122         uint16_t        ve_missed_update;
 123         zio_t           *ve_fill_io;
 124 };
 125
 126 struct vdev_cache {
 127         avl_tree_t      vc_offset_tree;
 128         avl_tree_t      vc_lastused_tree;
 129         kmutex_t        vc_lock;
 130 };
 131
 132 typedef struct vdev_queue_class {
 133         uint32_t        vqc_active;
 134
 135         /*
 136          * Sorted by offset or timestamp, depending on if the queue is
 137          * LBA-ordered vs FIFO.
 138          */
 139         avl_tree_t      vqc_queued_tree;
 140 } vdev_queue_class_t;
 141
 142 struct vdev_queue {
 143         vdev_t          *vq_vdev;
 144         vdev_queue_class_t vq_class[ZIO_PRIORITY_NUM_QUEUEABLE];
 145         avl_tree_t      vq_active_tree;
 146         avl_tree_t      vq_read_offset_tree;
 147         avl_tree_t      vq_write_offset_tree;
 148         avl_tree_t      vq_trim_offset_tree;
 149         uint64_t        vq_last_offset;
 150         hrtime_t        vq_io_complete_ts; /* time last i/o completed */
 151         hrtime_t        vq_io_delta_ts;
 152         zio_t           vq_io_search; /* used as local for stack reduction */
 153         kmutex_t        vq_lock;
 154 };
 155
 156 typedef enum vdev_alloc_bias {
 157         VDEV_BIAS_NONE,
 158         VDEV_BIAS_LOG,          /* dedicated to ZIL data (SLOG) */
 159         VDEV_BIAS_SPECIAL,      /* dedicated to ddt, metadata, and small blks */
 160         VDEV_BIAS_DEDUP         /* dedicated to dedup metadata */
 161 } vdev_alloc_bias_t;
 162
 163
 164 /*
 165  * On-disk indirect vdev state.
 166  *
 167  * An indirect vdev is described exclusively in the MOS config of a pool.
 168  * The config for an indirect vdev includes several fields, which are
 169  * accessed in memory by a vdev_indirect_config_t.
 170  */
 171 typedef struct vdev_indirect_config {
 172         /*
 173          * Object (in MOS) which contains the indirect mapping. This object
 174          * contains an array of vdev_indirect_mapping_entry_phys_t ordered by
 175          * vimep_src. The bonus buffer for this object is a
 176          * vdev_indirect_mapping_phys_t. This object is allocated when a vdev
 177          * removal is initiated.
 178          *
 179          * Note that this object can be empty if none of the data on the vdev
 180          * has been copied yet.
 181          */
 182         uint64_t        vic_mapping_object;
 183
 184         /*
 185          * Object (in MOS) which contains the birth times for the mapping
 186          * entries. This object contains an array of
 187          * vdev_indirect_birth_entry_phys_t sorted by vibe_offset. The bonus
 188          * buffer for this object is a vdev_indirect_birth_phys_t. This object
 189          * is allocated when a vdev removal is initiated.
 190          *
 191          * Note that this object can be empty if none of the vdev has yet been
 192          * copied.
 193          */
 194         uint64_t        vic_births_object;
 195
 196         /*
 197          * This is the vdev ID which was removed previous to this vdev, or
 198          * UINT64_MAX if there are no previously removed vdevs.
 199          */
 200         uint64_t        vic_prev_indirect_vdev;
 201 } vdev_indirect_config_t;
 202
 203 /*
 204  * Virtual device descriptor
 205  */
 206 struct vdev {
 207         /*
 208          * Common to all vdev types.
 209          */
 210         uint64_t        vdev_id;        /* child number in vdev parent  */
 211         uint64_t        vdev_guid;      /* unique ID for this vdev      */
 212         uint64_t        vdev_guid_sum;  /* self guid + all child guids  */
 213         uint64_t        vdev_orig_guid; /* orig. guid prior to remove   */
 214         uint64_t        vdev_asize;     /* allocatable device capacity  */
 215         uint64_t        vdev_min_asize; /* min acceptable asize         */
 216         uint64_t        vdev_max_asize; /* max acceptable asize         */
 217         uint64_t        vdev_ashift;    /* block alignment shift        */
 218         uint64_t        vdev_state;     /* see VDEV_STATE_* #defines    */
 219         uint64_t        vdev_prevstate; /* used when reopening a vdev   */
 220         vdev_ops_t      *vdev_ops;      /* vdev operations              */
 221         spa_t           *vdev_spa;      /* spa for this vdev            */
 222         void            *vdev_tsd;      /* type-specific data           */
 223         vnode_t         *vdev_name_vp;  /* vnode for pathname           */
 224         vnode_t         *vdev_devid_vp; /* vnode for devid              */
 225         vdev_t          *vdev_top;      /* top-level vdev               */
 226         vdev_t          *vdev_parent;   /* parent vdev                  */
 227         vdev_t          **vdev_child;   /* array of children            */
 228         uint64_t        vdev_children;  /* number of children           */
 229         vdev_stat_t     vdev_stat;      /* virtual device statistics    */
 230         vdev_stat_ex_t  vdev_stat_ex;   /* extended statistics          */
 231         boolean_t       vdev_expanding; /* expand the vdev?             */
 232         boolean_t       vdev_reopening; /* reopen in progress?          */
 233         boolean_t       vdev_nonrot;    /* true if solid state          */
 234         int             vdev_open_error; /* error on last open          */
 235         kthread_t       *vdev_open_thread; /* thread opening children   */
 236         uint64_t        vdev_crtxg;     /* txg when top-level was added */
 237
 238         /*
 239          * Top-level vdev state.
 240          */
 241         uint64_t        vdev_ms_array;  /* metaslab array object        */
 242         uint64_t        vdev_ms_shift;  /* metaslab size shift          */
 243         uint64_t        vdev_ms_count;  /* number of metaslabs          */
 244         metaslab_group_t *vdev_mg;      /* metaslab group               */
 245         metaslab_t      **vdev_ms;      /* metaslab array               */
 246         uint64_t        vdev_pending_fastwrite; /* allocated fastwrites */
 247         txg_list_t      vdev_ms_list;   /* per-txg dirty metaslab lists */
 248         txg_list_t      vdev_dtl_list;  /* per-txg dirty DTL lists      */
 249         txg_node_t      vdev_txg_node;  /* per-txg dirty vdev linkage   */
 250         boolean_t       vdev_remove_wanted; /* async remove wanted?     */
 251         boolean_t       vdev_probe_wanted; /* async probe wanted?       */
 252         list_node_t     vdev_config_dirty_node; /* config dirty list    */
 253         list_node_t     vdev_state_dirty_node; /* state dirty list      */
 254         uint64_t        vdev_deflate_ratio; /* deflation ratio (x512)   */
 255         uint64_t        vdev_islog;     /* is an intent log device      */
 256         uint64_t        vdev_removing;  /* device is being removed?     */
 257         boolean_t       vdev_ishole;    /* is a hole in the namespace   */
 258         uint64_t        vdev_top_zap;
 259         vdev_alloc_bias_t vdev_alloc_bias; /* metaslab allocation bias  */
 260
 261         /* pool checkpoint related */
 262         space_map_t     *vdev_checkpoint_sm;    /* contains reserved blocks */
 263
 264         /* Initialize related */
 265         boolean_t       vdev_initialize_exit_wanted;
 266         vdev_initializing_state_t       vdev_initialize_state;
 267         list_node_t     vdev_initialize_node;
 268         kthread_t       *vdev_initialize_thread;
 269         /* Protects vdev_initialize_thread and vdev_initialize_state. */
 270         kmutex_t        vdev_initialize_lock;
 271         kcondvar_t      vdev_initialize_cv;
 272         uint64_t        vdev_initialize_offset[TXG_SIZE];
 273         uint64_t        vdev_initialize_last_offset;
 274         range_tree_t    *vdev_initialize_tree;  /* valid while initializing */
 275         uint64_t        vdev_initialize_bytes_est;
 276         uint64_t        vdev_initialize_bytes_done;
 277         time_t          vdev_initialize_action_time;    /* start and end time */
 278
 279         /* TRIM related */
 280         boolean_t       vdev_trim_exit_wanted;
 281         boolean_t       vdev_autotrim_exit_wanted;
 282         vdev_trim_state_t       vdev_trim_state;
 283         list_node_t     vdev_trim_node;
 284         kmutex_t        vdev_autotrim_lock;
 285         kcondvar_t      vdev_autotrim_cv;
 286         kthread_t       *vdev_autotrim_thread;
 287         /* Protects vdev_trim_thread and vdev_trim_state. */
 288         kmutex_t        vdev_trim_lock;
 289         kcondvar_t      vdev_trim_cv;
 290         kthread_t       *vdev_trim_thread;
 291         uint64_t        vdev_trim_offset[TXG_SIZE];
 292         uint64_t        vdev_trim_last_offset;
 293         uint64_t        vdev_trim_bytes_est;
 294         uint64_t        vdev_trim_bytes_done;
 295         uint64_t        vdev_trim_rate;         /* requested rate (bytes/sec) */
 296         uint64_t        vdev_trim_partial;      /* requested partial TRIM */
 297         uint64_t        vdev_trim_secure;       /* requested secure TRIM */
 298         time_t          vdev_trim_action_time;  /* start and end time */
 299
 300         /* for limiting outstanding I/Os (initialize and TRIM) */
 301         kmutex_t        vdev_initialize_io_lock;
 302         kcondvar_t      vdev_initialize_io_cv;
 303         uint64_t        vdev_initialize_inflight;
 304         kmutex_t        vdev_trim_io_lock;
 305         kcondvar_t      vdev_trim_io_cv;
 306         uint64_t        vdev_trim_inflight[2];
 307
 308         /*
 309          * Values stored in the config for an indirect or removing vdev.
 310          */
 311         vdev_indirect_config_t  vdev_indirect_config;
 312
 313         /*
 314          * The vdev_indirect_rwlock protects the vdev_indirect_mapping
 315          * pointer from changing on indirect vdevs (when it is condensed).
 316          * Note that removing (not yet indirect) vdevs have different
 317          * access patterns (the mapping is not accessed from open context,
 318          * e.g. from zio_read) and locking strategy (e.g. svr_lock).
 319          */
 320         krwlock_t vdev_indirect_rwlock;
 321         vdev_indirect_mapping_t *vdev_indirect_mapping;
 322         vdev_indirect_births_t *vdev_indirect_births;
 323
 324         /*
 325          * In memory data structures used to manage the obsolete sm, for
 326          * indirect or removing vdevs.
 327          *
 328          * The vdev_obsolete_segments is the in-core record of the segments
 329          * that are no longer referenced anywhere in the pool (due to
 330          * being freed or remapped and not referenced by any snapshots).
 331          * During a sync, segments are added to vdev_obsolete_segments
 332          * via vdev_indirect_mark_obsolete(); at the end of each sync
 333          * pass, this is appended to vdev_obsolete_sm via
 334          * vdev_indirect_sync_obsolete().  The vdev_obsolete_lock
 335          * protects against concurrent modifications of vdev_obsolete_segments
 336          * from multiple zio threads.
 337          */
 338         kmutex_t        vdev_obsolete_lock;
 339         range_tree_t    *vdev_obsolete_segments;
 340         space_map_t     *vdev_obsolete_sm;
 341
 342         /*
 343          * Protects the vdev_scan_io_queue field itself as well as the
 344          * structure's contents (when present).
 345          */
 346         kmutex_t                        vdev_scan_io_queue_lock;
 347         struct dsl_scan_io_queue        *vdev_scan_io_queue;
 348
 349         /*
 350          * Leaf vdev state.
 351          */
 352         range_tree_t    *vdev_dtl[DTL_TYPES]; /* dirty time logs        */
 353         space_map_t     *vdev_dtl_sm;   /* dirty time log space map     */
 354         txg_node_t      vdev_dtl_node;  /* per-txg dirty DTL linkage    */
 355         uint64_t        vdev_dtl_object; /* DTL object                  */
 356         uint64_t        vdev_psize;     /* physical device capacity     */
 357         uint64_t        vdev_wholedisk; /* true if this is a whole disk */
 358         uint64_t        vdev_offline;   /* persistent offline state     */
 359         uint64_t        vdev_faulted;   /* persistent faulted state     */
 360         uint64_t        vdev_degraded;  /* persistent degraded state    */
 361         uint64_t        vdev_removed;   /* persistent removed state     */
 362         uint64_t        vdev_resilver_txg; /* persistent resilvering state */
 363         uint64_t        vdev_nparity;   /* number of parity devices for raidz */
 364         char            *vdev_path;     /* vdev path (if any)           */
 365         char            *vdev_devid;    /* vdev devid (if any)          */
 366         char            *vdev_physpath; /* vdev device path (if any)    */
 367         char            *vdev_enc_sysfs_path;   /* enclosure sysfs path */
 368         char            *vdev_fru;      /* physical FRU location        */
 369         uint64_t        vdev_not_present; /* not present during import  */
 370         uint64_t        vdev_unspare;   /* unspare when resilvering done */
 371         boolean_t       vdev_nowritecache; /* true if flushwritecache failed */
 372         boolean_t       vdev_has_trim;  /* TRIM is supported            */
 373         boolean_t       vdev_has_securetrim; /* secure TRIM is supported */
 374         boolean_t       vdev_checkremove; /* temporary online test      */
 375         boolean_t       vdev_forcefault; /* force online fault          */
 376         boolean_t       vdev_splitting; /* split or repair in progress  */
 377         boolean_t       vdev_delayed_close; /* delayed device close?    */
 378         boolean_t       vdev_tmpoffline; /* device taken offline temporarily? */
 379         boolean_t       vdev_detached;  /* device detached?             */
 380         boolean_t       vdev_cant_read; /* vdev is failing all reads    */
 381         boolean_t       vdev_cant_write; /* vdev is failing all writes  */
 382         boolean_t       vdev_isspare;   /* was a hot spare              */
 383         boolean_t       vdev_isl2cache; /* was a l2cache device         */
 384         boolean_t       vdev_copy_uberblocks;  /* post expand copy uberblocks */
 385         boolean_t       vdev_resilver_deferred;  /* resilver deferred */
 386         vdev_queue_t    vdev_queue;     /* I/O deadline schedule queue  */
 387         vdev_cache_t    vdev_cache;     /* physical block cache         */
 388         spa_aux_vdev_t  *vdev_aux;      /* for l2cache and spares vdevs */
 389         zio_t           *vdev_probe_zio; /* root of current probe       */
 390         vdev_aux_t      vdev_label_aux; /* on-disk aux state            */
 391         uint64_t        vdev_leaf_zap;
 392         hrtime_t        vdev_mmp_pending; /* 0 if write finished        */
 393         uint64_t        vdev_mmp_kstat_id;      /* to find kstat entry */
 394         uint64_t        vdev_expansion_time;    /* vdev's last expansion time */
 395         list_node_t     vdev_leaf_node;         /* leaf vdev list */
 396
 397         /*
 398          * For DTrace to work in userland (libzpool) context, these fields must
 399          * remain at the end of the structure.  DTrace will use the kernel's
 400          * CTF definition for 'struct vdev', and since the size of a kmutex_t is
 401          * larger in userland, the offsets for the rest of the fields would be
 402          * incorrect.
 403          */
 404         kmutex_t        vdev_dtl_lock;  /* vdev_dtl_{map,resilver}      */
 405         kmutex_t        vdev_stat_lock; /* vdev_stat                    */
 406         kmutex_t        vdev_probe_lock; /* protects vdev_probe_zio     */
 407
 408         /*
 409          * We rate limit ZIO delay and ZIO checksum events, since they
 410          * can flood ZED with tons of events when a drive is acting up.
 411          */
 412         zfs_ratelimit_t vdev_delay_rl;
 413         zfs_ratelimit_t vdev_checksum_rl;
 414 };
 415
 416 #define VDEV_RAIDZ_MAXPARITY    3
 417
 418 #define VDEV_PAD_SIZE           (8 << 10)
 419 /* 2 padding areas (vl_pad1 and vl_pad2) to skip */
 420 #define VDEV_SKIP_SIZE          VDEV_PAD_SIZE * 2
 421 #define VDEV_PHYS_SIZE          (112 << 10)
 422 #define VDEV_UBERBLOCK_RING     (128 << 10)
 423
 424 /*
 425  * MMP blocks occupy the last MMP_BLOCKS_PER_LABEL slots in the uberblock
 426  * ring when MMP is enabled.
 427  */
 428 #define MMP_BLOCKS_PER_LABEL    1
 429
 430 /* The largest uberblock we support is 8k. */
 431 #define MAX_UBERBLOCK_SHIFT (13)
 432 #define VDEV_UBERBLOCK_SHIFT(vd)        \
 433         MIN(MAX((vd)->vdev_top->vdev_ashift, UBERBLOCK_SHIFT), \
 434             MAX_UBERBLOCK_SHIFT)
 435 #define VDEV_UBERBLOCK_COUNT(vd)        \
 436         (VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd))
 437 #define VDEV_UBERBLOCK_OFFSET(vd, n)    \
 438         offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)])
 439 #define VDEV_UBERBLOCK_SIZE(vd)         (1ULL << VDEV_UBERBLOCK_SHIFT(vd))
 440
 441 typedef struct vdev_phys {
 442         char            vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)];
 443         zio_eck_t       vp_zbt;
 444 } vdev_phys_t;
 445
 446 typedef struct vdev_label {
 447         char            vl_pad1[VDEV_PAD_SIZE];                 /*  8K */
 448         char            vl_pad2[VDEV_PAD_SIZE];                 /*  8K */
 449         vdev_phys_t     vl_vdev_phys;                           /* 112K */
 450         char            vl_uberblock[VDEV_UBERBLOCK_RING];      /* 128K */
 451 } vdev_label_t;                                                 /* 256K total */
 452
 453 /*
 454  * vdev_dirty() flags
 455  */
 456 #define VDD_METASLAB    0x01
 457 #define VDD_DTL         0x02
 458
 459 /* Offset of embedded boot loader region on each label */
 460 #define VDEV_BOOT_OFFSET        (2 * sizeof (vdev_label_t))
 461 /*
 462  * Size of embedded boot loader region on each label.
 463  * The total size of the first two labels plus the boot area is 4MB.
 464  */
 465 #define VDEV_BOOT_SIZE          (7ULL << 19)                    /* 3.5M */
 466
 467 /*
 468  * Size of label regions at the start and end of each leaf device.
 469  */
 470 #define VDEV_LABEL_START_SIZE   (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE)
 471 #define VDEV_LABEL_END_SIZE     (2 * sizeof (vdev_label_t))
 472 #define VDEV_LABELS             4
 473 #define VDEV_BEST_LABEL         VDEV_LABELS
 474
 475 #define VDEV_ALLOC_LOAD         0
 476 #define VDEV_ALLOC_ADD          1
 477 #define VDEV_ALLOC_SPARE        2
 478 #define VDEV_ALLOC_L2CACHE      3
 479 #define VDEV_ALLOC_ROOTPOOL     4
 480 #define VDEV_ALLOC_SPLIT        5
 481 #define VDEV_ALLOC_ATTACH       6
 482
 483 /*
 484  * Allocate or free a vdev
 485  */
 486 extern vdev_t *vdev_alloc_common(spa_t *spa, uint_t id, uint64_t guid,
 487     vdev_ops_t *ops);
 488 extern int vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *config,
 489     vdev_t *parent, uint_t id, int alloctype);
 490 extern void vdev_free(vdev_t *vd);
 491
 492 /*
 493  * Add or remove children and parents
 494  */
 495 extern void vdev_add_child(vdev_t *pvd, vdev_t *cvd);
 496 extern void vdev_remove_child(vdev_t *pvd, vdev_t *cvd);
 497 extern void vdev_compact_children(vdev_t *pvd);
 498 extern vdev_t *vdev_add_parent(vdev_t *cvd, vdev_ops_t *ops);
 499 extern void vdev_remove_parent(vdev_t *cvd);
 500
 501 /*
 502  * vdev sync load and sync
 503  */
 504 extern boolean_t vdev_log_state_valid(vdev_t *vd);
 505 extern int vdev_load(vdev_t *vd);
 506 extern int vdev_dtl_load(vdev_t *vd);
 507 extern void vdev_sync(vdev_t *vd, uint64_t txg);
 508 extern void vdev_sync_done(vdev_t *vd, uint64_t txg);
 509 extern void vdev_dirty(vdev_t *vd, int flags, void *arg, uint64_t txg);
 510 extern void vdev_dirty_leaves(vdev_t *vd, int flags, uint64_t txg);
 511
 512 /*
 513  * Available vdev types.
 514  */
 515 extern vdev_ops_t vdev_root_ops;
 516 extern vdev_ops_t vdev_mirror_ops;
 517 extern vdev_ops_t vdev_replacing_ops;
 518 extern vdev_ops_t vdev_raidz_ops;
 519 extern vdev_ops_t vdev_disk_ops;
 520 extern vdev_ops_t vdev_file_ops;
 521 extern vdev_ops_t vdev_missing_ops;
 522 extern vdev_ops_t vdev_hole_ops;
 523 extern vdev_ops_t vdev_spare_ops;
 524 extern vdev_ops_t vdev_indirect_ops;
 525
 526 /*
 527  * Common size functions
 528  */
 529 extern void vdev_default_xlate(vdev_t *vd, const range_seg_t *in,
 530     range_seg_t *out);
 531 extern uint64_t vdev_default_asize(vdev_t *vd, uint64_t psize);
 532 extern uint64_t vdev_get_min_asize(vdev_t *vd);
 533 extern void vdev_set_min_asize(vdev_t *vd);
 534
 535 /*
 536  * Global variables
 537  */
 538 extern int zfs_vdev_standard_sm_blksz;
 539 /* zdb uses this tunable, so it must be declared here to make lint happy. */
 540 extern int zfs_vdev_cache_size;
 541
 542 /*
 543  * Functions from vdev_indirect.c
 544  */
 545 extern void vdev_indirect_sync_obsolete(vdev_t *vd, dmu_tx_t *tx);
 546 extern boolean_t vdev_indirect_should_condense(vdev_t *vd);
 547 extern void spa_condense_indirect_start_sync(vdev_t *vd, dmu_tx_t *tx);
 548 extern int vdev_obsolete_sm_object(vdev_t *vd, uint64_t *sm_obj);
 549 extern int vdev_obsolete_counts_are_precise(vdev_t *vd, boolean_t *are_precise);
 550
 551 /*
 552  * Other miscellaneous functions
 553  */
 554 int vdev_checkpoint_sm_object(vdev_t *vd, uint64_t *sm_obj);
 555
 556 #ifdef  __cplusplus
 557 }
 558 #endif
 559
 560 #endif  /* _SYS_VDEV_IMPL_H */