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.
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.
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]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2011, 2018 by Delphix. All rights reserved.
24 * Copyright (c) 2017, Intel Corporation.
27 #ifndef _SYS_VDEV_IMPL_H
28 #define _SYS_VDEV_IMPL_H
31 #include <sys/bpobj.h>
33 #include <sys/metaslab.h>
34 #include <sys/nvpair.h>
35 #include <sys/space_map.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>
48 * Virtual device descriptors.
50 * All storage pool operations go through the virtual device framework,
51 * which provides data replication and I/O scheduling.
55 * Forward declarations that lots of things need.
57 typedef struct vdev_queue vdev_queue_t;
58 typedef struct vdev_cache vdev_cache_t;
59 typedef struct vdev_cache_entry vdev_cache_entry_t;
62 extern int zfs_vdev_queue_depth_pct;
63 extern int zfs_vdev_def_queue_depth;
64 extern uint32_t zfs_vdev_async_write_max_active;
67 * Virtual device operations
69 typedef int vdev_open_func_t(vdev_t *vd, uint64_t *size, uint64_t *max_size,
70 uint64_t *logical_ashift, uint64_t *physical_ashift);
71 typedef void vdev_close_func_t(vdev_t *vd);
72 typedef uint64_t vdev_asize_func_t(vdev_t *vd, uint64_t psize);
73 typedef void vdev_io_start_func_t(zio_t *zio);
74 typedef void vdev_io_done_func_t(zio_t *zio);
75 typedef void vdev_state_change_func_t(vdev_t *vd, int, int);
76 typedef boolean_t vdev_need_resilver_func_t(vdev_t *vd, uint64_t, size_t);
77 typedef void vdev_hold_func_t(vdev_t *vd);
78 typedef void vdev_rele_func_t(vdev_t *vd);
80 typedef void vdev_remap_cb_t(uint64_t inner_offset, vdev_t *vd,
81 uint64_t offset, uint64_t size, void *arg);
82 typedef void vdev_remap_func_t(vdev_t *vd, uint64_t offset, uint64_t size,
83 vdev_remap_cb_t callback, void *arg);
85 * Given a target vdev, translates the logical range "in" to the physical
88 typedef void vdev_xlation_func_t(vdev_t *cvd, const range_seg_t *in,
91 typedef struct vdev_ops {
92 vdev_open_func_t *vdev_op_open;
93 vdev_close_func_t *vdev_op_close;
94 vdev_asize_func_t *vdev_op_asize;
95 vdev_io_start_func_t *vdev_op_io_start;
96 vdev_io_done_func_t *vdev_op_io_done;
97 vdev_state_change_func_t *vdev_op_state_change;
98 vdev_need_resilver_func_t *vdev_op_need_resilver;
99 vdev_hold_func_t *vdev_op_hold;
100 vdev_rele_func_t *vdev_op_rele;
101 vdev_remap_func_t *vdev_op_remap;
103 * For translating ranges from non-leaf vdevs (e.g. raidz) to leaves.
104 * Used when initializing vdevs. Isn't used by leaf ops.
106 vdev_xlation_func_t *vdev_op_xlate;
107 char vdev_op_type[16];
108 boolean_t vdev_op_leaf;
112 * Virtual device properties
114 struct vdev_cache_entry {
117 uint64_t ve_lastused;
118 avl_node_t ve_offset_node;
119 avl_node_t ve_lastused_node;
121 uint16_t ve_missed_update;
126 avl_tree_t vc_offset_tree;
127 avl_tree_t vc_lastused_tree;
131 typedef struct vdev_queue_class {
135 * Sorted by offset or timestamp, depending on if the queue is
136 * LBA-ordered vs FIFO.
138 avl_tree_t vqc_queued_tree;
139 } vdev_queue_class_t;
143 vdev_queue_class_t vq_class[ZIO_PRIORITY_NUM_QUEUEABLE];
144 avl_tree_t vq_active_tree;
145 avl_tree_t vq_read_offset_tree;
146 avl_tree_t vq_write_offset_tree;
147 uint64_t vq_last_offset;
148 hrtime_t vq_io_complete_ts; /* time last i/o completed */
150 uint64_t vq_lastoffset;
153 typedef enum vdev_alloc_bias {
155 VDEV_BIAS_LOG, /* dedicated to ZIL data (SLOG) */
156 VDEV_BIAS_SPECIAL, /* dedicated to ddt, metadata, and small blks */
157 VDEV_BIAS_DEDUP /* dedicated to dedup metadata */
162 * On-disk indirect vdev state.
164 * An indirect vdev is described exclusively in the MOS config of a pool.
165 * The config for an indirect vdev includes several fields, which are
166 * accessed in memory by a vdev_indirect_config_t.
168 typedef struct vdev_indirect_config {
170 * Object (in MOS) which contains the indirect mapping. This object
171 * contains an array of vdev_indirect_mapping_entry_phys_t ordered by
172 * vimep_src. The bonus buffer for this object is a
173 * vdev_indirect_mapping_phys_t. This object is allocated when a vdev
174 * removal is initiated.
176 * Note that this object can be empty if none of the data on the vdev
177 * has been copied yet.
179 uint64_t vic_mapping_object;
182 * Object (in MOS) which contains the birth times for the mapping
183 * entries. This object contains an array of
184 * vdev_indirect_birth_entry_phys_t sorted by vibe_offset. The bonus
185 * buffer for this object is a vdev_indirect_birth_phys_t. This object
186 * is allocated when a vdev removal is initiated.
188 * Note that this object can be empty if none of the vdev has yet been
191 uint64_t vic_births_object;
194 * This is the vdev ID which was removed previous to this vdev, or
195 * UINT64_MAX if there are no previously removed vdevs.
197 uint64_t vic_prev_indirect_vdev;
198 } vdev_indirect_config_t;
201 * Virtual device descriptor
205 * Common to all vdev types.
207 uint64_t vdev_id; /* child number in vdev parent */
208 uint64_t vdev_guid; /* unique ID for this vdev */
209 uint64_t vdev_guid_sum; /* self guid + all child guids */
210 uint64_t vdev_orig_guid; /* orig. guid prior to remove */
211 uint64_t vdev_asize; /* allocatable device capacity */
212 uint64_t vdev_min_asize; /* min acceptable asize */
213 uint64_t vdev_max_asize; /* max acceptable asize */
214 uint64_t vdev_ashift; /* block alignment shift */
216 * Logical block alignment shift
218 * The smallest sized/aligned I/O supported by the device.
220 uint64_t vdev_logical_ashift;
222 * Physical block alignment shift
224 * The device supports logical I/Os with vdev_logical_ashift
225 * size/alignment, but optimum performance will be achieved by
226 * aligning/sizing requests to vdev_physical_ashift. Smaller
227 * requests may be inflated or incur device level read-modify-write
230 * May be 0 to indicate no preference (i.e. use vdev_logical_ashift).
232 uint64_t vdev_physical_ashift;
233 uint64_t vdev_state; /* see VDEV_STATE_* #defines */
234 uint64_t vdev_prevstate; /* used when reopening a vdev */
235 vdev_ops_t *vdev_ops; /* vdev operations */
236 spa_t *vdev_spa; /* spa for this vdev */
237 void *vdev_tsd; /* type-specific data */
238 vnode_t *vdev_name_vp; /* vnode for pathname */
239 vnode_t *vdev_devid_vp; /* vnode for devid */
240 vdev_t *vdev_top; /* top-level vdev */
241 vdev_t *vdev_parent; /* parent vdev */
242 vdev_t **vdev_child; /* array of children */
243 uint64_t vdev_children; /* number of children */
244 vdev_stat_t vdev_stat; /* virtual device statistics */
245 boolean_t vdev_expanding; /* expand the vdev? */
246 boolean_t vdev_reopening; /* reopen in progress? */
247 boolean_t vdev_nonrot; /* true if solid state */
248 int vdev_open_error; /* error on last open */
249 kthread_t *vdev_open_thread; /* thread opening children */
250 uint64_t vdev_crtxg; /* txg when top-level was added */
253 * Top-level vdev state.
255 uint64_t vdev_ms_array; /* metaslab array object */
256 uint64_t vdev_ms_shift; /* metaslab size shift */
257 uint64_t vdev_ms_count; /* number of metaslabs */
258 metaslab_group_t *vdev_mg; /* metaslab group */
259 metaslab_t **vdev_ms; /* metaslab array */
260 txg_list_t vdev_ms_list; /* per-txg dirty metaslab lists */
261 txg_list_t vdev_dtl_list; /* per-txg dirty DTL lists */
262 txg_node_t vdev_txg_node; /* per-txg dirty vdev linkage */
263 boolean_t vdev_remove_wanted; /* async remove wanted? */
264 boolean_t vdev_probe_wanted; /* async probe wanted? */
265 list_node_t vdev_config_dirty_node; /* config dirty list */
266 list_node_t vdev_state_dirty_node; /* state dirty list */
267 uint64_t vdev_deflate_ratio; /* deflation ratio (x512) */
268 uint64_t vdev_islog; /* is an intent log device */
269 uint64_t vdev_removing; /* device is being removed? */
270 boolean_t vdev_ishole; /* is a hole in the namespace */
271 uint64_t vdev_top_zap;
272 vdev_alloc_bias_t vdev_alloc_bias; /* metaslab allocation bias */
274 /* pool checkpoint related */
275 space_map_t *vdev_checkpoint_sm; /* contains reserved blocks */
277 boolean_t vdev_initialize_exit_wanted;
278 vdev_initializing_state_t vdev_initialize_state;
279 kthread_t *vdev_initialize_thread;
280 /* Protects vdev_initialize_thread and vdev_initialize_state. */
281 kmutex_t vdev_initialize_lock;
282 kcondvar_t vdev_initialize_cv;
283 uint64_t vdev_initialize_offset[TXG_SIZE];
284 uint64_t vdev_initialize_last_offset;
285 range_tree_t *vdev_initialize_tree; /* valid while initializing */
286 uint64_t vdev_initialize_bytes_est;
287 uint64_t vdev_initialize_bytes_done;
288 time_t vdev_initialize_action_time; /* start and end time */
290 /* for limiting outstanding I/Os */
291 kmutex_t vdev_initialize_io_lock;
292 kcondvar_t vdev_initialize_io_cv;
293 uint64_t vdev_initialize_inflight;
296 * Values stored in the config for an indirect or removing vdev.
298 vdev_indirect_config_t vdev_indirect_config;
301 * The vdev_indirect_rwlock protects the vdev_indirect_mapping
302 * pointer from changing on indirect vdevs (when it is condensed).
303 * Note that removing (not yet indirect) vdevs have different
304 * access patterns (the mapping is not accessed from open context,
305 * e.g. from zio_read) and locking strategy (e.g. svr_lock).
307 krwlock_t vdev_indirect_rwlock;
308 vdev_indirect_mapping_t *vdev_indirect_mapping;
309 vdev_indirect_births_t *vdev_indirect_births;
312 * In memory data structures used to manage the obsolete sm, for
313 * indirect or removing vdevs.
315 * The vdev_obsolete_segments is the in-core record of the segments
316 * that are no longer referenced anywhere in the pool (due to
317 * being freed or remapped and not referenced by any snapshots).
318 * During a sync, segments are added to vdev_obsolete_segments
319 * via vdev_indirect_mark_obsolete(); at the end of each sync
320 * pass, this is appended to vdev_obsolete_sm via
321 * vdev_indirect_sync_obsolete(). The vdev_obsolete_lock
322 * protects against concurrent modifications of vdev_obsolete_segments
323 * from multiple zio threads.
325 kmutex_t vdev_obsolete_lock;
326 range_tree_t *vdev_obsolete_segments;
327 space_map_t *vdev_obsolete_sm;
330 * Protects the vdev_scan_io_queue field itself as well as the
331 * structure's contents (when present).
333 kmutex_t vdev_scan_io_queue_lock;
334 struct dsl_scan_io_queue *vdev_scan_io_queue;
339 range_tree_t *vdev_dtl[DTL_TYPES]; /* dirty time logs */
340 space_map_t *vdev_dtl_sm; /* dirty time log space map */
341 txg_node_t vdev_dtl_node; /* per-txg dirty DTL linkage */
342 uint64_t vdev_dtl_object; /* DTL object */
343 uint64_t vdev_psize; /* physical device capacity */
344 uint64_t vdev_wholedisk; /* true if this is a whole disk */
345 uint64_t vdev_offline; /* persistent offline state */
346 uint64_t vdev_faulted; /* persistent faulted state */
347 uint64_t vdev_degraded; /* persistent degraded state */
348 uint64_t vdev_removed; /* persistent removed state */
349 uint64_t vdev_resilver_txg; /* persistent resilvering state */
350 uint64_t vdev_nparity; /* number of parity devices for raidz */
351 char *vdev_path; /* vdev path (if any) */
352 char *vdev_devid; /* vdev devid (if any) */
353 char *vdev_physpath; /* vdev device path (if any) */
354 char *vdev_fru; /* physical FRU location */
355 uint64_t vdev_not_present; /* not present during import */
356 uint64_t vdev_unspare; /* unspare when resilvering done */
357 boolean_t vdev_nowritecache; /* true if flushwritecache failed */
358 boolean_t vdev_notrim; /* true if trim failed */
359 boolean_t vdev_checkremove; /* temporary online test */
360 boolean_t vdev_forcefault; /* force online fault */
361 boolean_t vdev_splitting; /* split or repair in progress */
362 boolean_t vdev_delayed_close; /* delayed device close? */
363 boolean_t vdev_tmpoffline; /* device taken offline temporarily? */
364 boolean_t vdev_detached; /* device detached? */
365 boolean_t vdev_cant_read; /* vdev is failing all reads */
366 boolean_t vdev_cant_write; /* vdev is failing all writes */
367 boolean_t vdev_isspare; /* was a hot spare */
368 boolean_t vdev_isl2cache; /* was a l2cache device */
369 vdev_queue_t vdev_queue; /* I/O deadline schedule queue */
370 vdev_cache_t vdev_cache; /* physical block cache */
371 spa_aux_vdev_t *vdev_aux; /* for l2cache and spares vdevs */
372 zio_t *vdev_probe_zio; /* root of current probe */
373 vdev_aux_t vdev_label_aux; /* on-disk aux state */
374 struct trim_map *vdev_trimmap; /* map on outstanding trims */
375 uint64_t vdev_leaf_zap;
376 hrtime_t vdev_mmp_pending; /* 0 if write finished */
377 uint64_t vdev_mmp_kstat_id; /* to find kstat entry */
378 list_node_t vdev_leaf_node; /* leaf vdev list */
381 * For DTrace to work in userland (libzpool) context, these fields must
382 * remain at the end of the structure. DTrace will use the kernel's
383 * CTF definition for 'struct vdev', and since the size of a kmutex_t is
384 * larger in userland, the offsets for the rest of the fields would be
387 kmutex_t vdev_dtl_lock; /* vdev_dtl_{map,resilver} */
388 kmutex_t vdev_stat_lock; /* vdev_stat */
389 kmutex_t vdev_probe_lock; /* protects vdev_probe_zio */
392 #define VDEV_RAIDZ_MAXPARITY 3
394 #define VDEV_PAD_SIZE (8 << 10)
395 /* 2 padding areas (vl_pad1 and vl_pad2) to skip */
396 #define VDEV_SKIP_SIZE VDEV_PAD_SIZE * 2
397 #define VDEV_PHYS_SIZE (112 << 10)
398 #define VDEV_UBERBLOCK_RING (128 << 10)
401 * MMP blocks occupy the last MMP_BLOCKS_PER_LABEL slots in the uberblock
402 * ring when MMP is enabled.
404 #define MMP_BLOCKS_PER_LABEL 1
406 /* The largest uberblock we support is 8k. */
407 #define MAX_UBERBLOCK_SHIFT (13)
408 #define VDEV_UBERBLOCK_SHIFT(vd) \
409 MIN(MAX((vd)->vdev_top->vdev_ashift, UBERBLOCK_SHIFT), \
411 #define VDEV_UBERBLOCK_COUNT(vd) \
412 (VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd))
413 #define VDEV_UBERBLOCK_OFFSET(vd, n) \
414 offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)])
415 #define VDEV_UBERBLOCK_SIZE(vd) (1ULL << VDEV_UBERBLOCK_SHIFT(vd))
417 typedef struct vdev_phys {
418 char vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)];
422 typedef struct vdev_label {
423 char vl_pad1[VDEV_PAD_SIZE]; /* 8K */
424 char vl_pad2[VDEV_PAD_SIZE]; /* 8K */
425 vdev_phys_t vl_vdev_phys; /* 112K */
426 char vl_uberblock[VDEV_UBERBLOCK_RING]; /* 128K */
427 } vdev_label_t; /* 256K total */
432 #define VDD_METASLAB 0x01
435 /* Offset of embedded boot loader region on each label */
436 #define VDEV_BOOT_OFFSET (2 * sizeof (vdev_label_t))
438 * Size of embedded boot loader region on each label.
439 * The total size of the first two labels plus the boot area is 4MB.
441 #define VDEV_BOOT_SIZE (7ULL << 19) /* 3.5M */
444 * Size of label regions at the start and end of each leaf device.
446 #define VDEV_LABEL_START_SIZE (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE)
447 #define VDEV_LABEL_END_SIZE (2 * sizeof (vdev_label_t))
448 #define VDEV_LABELS 4
449 #define VDEV_BEST_LABEL VDEV_LABELS
451 #define VDEV_ALLOC_LOAD 0
452 #define VDEV_ALLOC_ADD 1
453 #define VDEV_ALLOC_SPARE 2
454 #define VDEV_ALLOC_L2CACHE 3
455 #define VDEV_ALLOC_ROOTPOOL 4
456 #define VDEV_ALLOC_SPLIT 5
457 #define VDEV_ALLOC_ATTACH 6
460 * Allocate or free a vdev
462 extern vdev_t *vdev_alloc_common(spa_t *spa, uint_t id, uint64_t guid,
464 extern int vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *config,
465 vdev_t *parent, uint_t id, int alloctype);
466 extern void vdev_free(vdev_t *vd);
469 * Add or remove children and parents
471 extern void vdev_add_child(vdev_t *pvd, vdev_t *cvd);
472 extern void vdev_remove_child(vdev_t *pvd, vdev_t *cvd);
473 extern void vdev_compact_children(vdev_t *pvd);
474 extern vdev_t *vdev_add_parent(vdev_t *cvd, vdev_ops_t *ops);
475 extern void vdev_remove_parent(vdev_t *cvd);
478 * vdev sync load and sync
480 extern boolean_t vdev_log_state_valid(vdev_t *vd);
481 extern int vdev_load(vdev_t *vd);
482 extern int vdev_dtl_load(vdev_t *vd);
483 extern void vdev_sync(vdev_t *vd, uint64_t txg);
484 extern void vdev_sync_done(vdev_t *vd, uint64_t txg);
485 extern void vdev_dirty(vdev_t *vd, int flags, void *arg, uint64_t txg);
486 extern void vdev_dirty_leaves(vdev_t *vd, int flags, uint64_t txg);
489 * Available vdev types.
491 extern vdev_ops_t vdev_root_ops;
492 extern vdev_ops_t vdev_mirror_ops;
493 extern vdev_ops_t vdev_replacing_ops;
494 extern vdev_ops_t vdev_raidz_ops;
496 extern vdev_ops_t vdev_geom_ops;
498 extern vdev_ops_t vdev_disk_ops;
500 extern vdev_ops_t vdev_file_ops;
501 extern vdev_ops_t vdev_missing_ops;
502 extern vdev_ops_t vdev_hole_ops;
503 extern vdev_ops_t vdev_spare_ops;
504 extern vdev_ops_t vdev_indirect_ops;
507 * Common size functions
509 extern void vdev_default_xlate(vdev_t *vd, const range_seg_t *in,
511 extern uint64_t vdev_default_asize(vdev_t *vd, uint64_t psize);
512 extern uint64_t vdev_get_min_asize(vdev_t *vd);
513 extern void vdev_set_min_asize(vdev_t *vd);
518 extern int vdev_standard_sm_blksz;
519 /* zdb uses this tunable, so it must be declared here to make lint happy. */
520 extern int zfs_vdev_cache_size;
521 extern uint_t zfs_geom_probe_vdev_key;
524 * Functions from vdev_indirect.c
526 extern void vdev_indirect_sync_obsolete(vdev_t *vd, dmu_tx_t *tx);
527 extern boolean_t vdev_indirect_should_condense(vdev_t *vd);
528 extern void spa_condense_indirect_start_sync(vdev_t *vd, dmu_tx_t *tx);
529 extern int vdev_obsolete_sm_object(vdev_t *vd);
530 extern boolean_t vdev_obsolete_counts_are_precise(vdev_t *vd);
534 * Other miscellaneous functions
536 int vdev_checkpoint_sm_object(vdev_t *vd);
539 * The vdev_buf_t is used to translate between zio_t and buf_t, and back again.
541 typedef struct vdev_buf {
542 buf_t vb_buf; /* buffer that describes the io */
543 zio_t *vb_io; /* pointer back to the original zio_t */
551 #endif /* _SYS_VDEV_IMPL_H */