xref: /illumos-gate/usr/src/uts/common/fs/zfs/sys/vdev_impl.h (revision 4d9fdb46b215739778ebc12079842c9905586999)
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, 2018 by Delphix. All rights reserved.
24  * Copyright (c) 2011, 2019 by Delphix. All rights reserved.
25  * Copyright (c) 2017, Intel Corporation.
26  * Copyright 2019 Joyent, Inc.
27  * Copyright 2020 Joshua M. Clulow <josh@sysmgr.org>
28  */
29 
30 #ifndef _SYS_VDEV_IMPL_H
31 #define	_SYS_VDEV_IMPL_H
32 
33 #include <sys/avl.h>
34 #include <sys/bpobj.h>
35 #include <sys/dmu.h>
36 #include <sys/metaslab.h>
37 #include <sys/nvpair.h>
38 #include <sys/space_map.h>
39 #include <sys/vdev.h>
40 #include <sys/dkio.h>
41 #include <sys/uberblock_impl.h>
42 #include <sys/vdev_indirect_mapping.h>
43 #include <sys/vdev_indirect_births.h>
44 #include <sys/vdev_removal.h>
45 
46 #ifdef	__cplusplus
47 extern "C" {
48 #endif
49 
50 /*
51  * Virtual device descriptors.
52  *
53  * All storage pool operations go through the virtual device framework,
54  * which provides data replication and I/O scheduling.
55  */
56 
57 /*
58  * Forward declarations that lots of things need.
59  */
60 typedef struct vdev_queue vdev_queue_t;
61 typedef struct vdev_cache vdev_cache_t;
62 typedef struct vdev_cache_entry vdev_cache_entry_t;
63 struct abd;
64 
65 extern int zfs_vdev_queue_depth_pct;
66 extern int zfs_vdev_def_queue_depth;
67 extern uint32_t zfs_vdev_async_write_max_active;
68 
69 /*
70  * Virtual device operations
71  */
72 typedef int	vdev_open_func_t(vdev_t *vd, uint64_t *size, uint64_t *max_size,
73     uint64_t *ashift);
74 typedef void	vdev_close_func_t(vdev_t *vd);
75 typedef uint64_t vdev_asize_func_t(vdev_t *vd, uint64_t psize);
76 typedef void	vdev_io_start_func_t(zio_t *zio);
77 typedef void	vdev_io_done_func_t(zio_t *zio);
78 typedef void	vdev_state_change_func_t(vdev_t *vd, int, int);
79 typedef boolean_t vdev_need_resilver_func_t(vdev_t *vd, uint64_t, size_t);
80 typedef void	vdev_hold_func_t(vdev_t *vd);
81 typedef void	vdev_rele_func_t(vdev_t *vd);
82 
83 typedef void	vdev_remap_cb_t(uint64_t inner_offset, vdev_t *vd,
84     uint64_t offset, uint64_t size, void *arg);
85 typedef void	vdev_remap_func_t(vdev_t *vd, uint64_t offset, uint64_t size,
86     vdev_remap_cb_t callback, void *arg);
87 typedef int	vdev_dumpio_func_t(vdev_t *vd, caddr_t data, size_t size,
88     uint64_t offset, uint64_t origoffset, boolean_t doread, boolean_t isdump);
89 /*
90  * Given a target vdev, translates the logical range "in" to the physical
91  * range "res"
92  */
93 typedef void vdev_xlation_func_t(vdev_t *cvd, const range_seg64_t *in,
94     range_seg64_t *res);
95 
96 typedef struct vdev_ops {
97 	vdev_open_func_t		*vdev_op_open;
98 	vdev_close_func_t		*vdev_op_close;
99 	vdev_asize_func_t		*vdev_op_asize;
100 	vdev_io_start_func_t		*vdev_op_io_start;
101 	vdev_io_done_func_t		*vdev_op_io_done;
102 	vdev_state_change_func_t	*vdev_op_state_change;
103 	vdev_need_resilver_func_t	*vdev_op_need_resilver;
104 	vdev_hold_func_t		*vdev_op_hold;
105 	vdev_rele_func_t		*vdev_op_rele;
106 	vdev_remap_func_t		*vdev_op_remap;
107 	/*
108 	 * For translating ranges from non-leaf vdevs (e.g. raidz) to leaves.
109 	 * Used when initializing vdevs. Isn't used by leaf ops.
110 	 */
111 	vdev_xlation_func_t		*vdev_op_xlate;
112 	vdev_dumpio_func_t		*vdev_op_dumpio;
113 	char				vdev_op_type[16];
114 	boolean_t			vdev_op_leaf;
115 } vdev_ops_t;
116 
117 /*
118  * Virtual device properties
119  */
120 struct vdev_cache_entry {
121 	struct abd	*ve_abd;
122 	uint64_t	ve_offset;
123 	uint64_t	ve_lastused;
124 	avl_node_t	ve_offset_node;
125 	avl_node_t	ve_lastused_node;
126 	uint32_t	ve_hits;
127 	uint16_t	ve_missed_update;
128 	zio_t		*ve_fill_io;
129 };
130 
131 struct vdev_cache {
132 	avl_tree_t	vc_offset_tree;
133 	avl_tree_t	vc_lastused_tree;
134 	kmutex_t	vc_lock;
135 };
136 
137 typedef struct vdev_queue_class {
138 	uint32_t	vqc_active;
139 
140 	/*
141 	 * Sorted by offset or timestamp, depending on if the queue is
142 	 * LBA-ordered vs FIFO.
143 	 */
144 	avl_tree_t	vqc_queued_tree;
145 } vdev_queue_class_t;
146 
147 struct vdev_queue {
148 	vdev_t		*vq_vdev;
149 	vdev_queue_class_t vq_class[ZIO_PRIORITY_NUM_QUEUEABLE];
150 	avl_tree_t	vq_active_tree;
151 	avl_tree_t	vq_read_offset_tree;
152 	avl_tree_t	vq_write_offset_tree;
153 	avl_tree_t	vq_trim_offset_tree;
154 	uint64_t	vq_last_offset;
155 	hrtime_t	vq_io_complete_ts; /* time last i/o completed */
156 	kmutex_t	vq_lock;
157 };
158 
159 typedef enum vdev_alloc_bias {
160 	VDEV_BIAS_NONE,
161 	VDEV_BIAS_LOG,		/* dedicated to ZIL data (SLOG) */
162 	VDEV_BIAS_SPECIAL,	/* dedicated to ddt, metadata, and small blks */
163 	VDEV_BIAS_DEDUP		/* dedicated to dedup metadata */
164 } vdev_alloc_bias_t;
165 
166 
167 /*
168  * On-disk indirect vdev state.
169  *
170  * An indirect vdev is described exclusively in the MOS config of a pool.
171  * The config for an indirect vdev includes several fields, which are
172  * accessed in memory by a vdev_indirect_config_t.
173  */
174 typedef struct vdev_indirect_config {
175 	/*
176 	 * Object (in MOS) which contains the indirect mapping. This object
177 	 * contains an array of vdev_indirect_mapping_entry_phys_t ordered by
178 	 * vimep_src. The bonus buffer for this object is a
179 	 * vdev_indirect_mapping_phys_t. This object is allocated when a vdev
180 	 * removal is initiated.
181 	 *
182 	 * Note that this object can be empty if none of the data on the vdev
183 	 * has been copied yet.
184 	 */
185 	uint64_t	vic_mapping_object;
186 
187 	/*
188 	 * Object (in MOS) which contains the birth times for the mapping
189 	 * entries. This object contains an array of
190 	 * vdev_indirect_birth_entry_phys_t sorted by vibe_offset. The bonus
191 	 * buffer for this object is a vdev_indirect_birth_phys_t. This object
192 	 * is allocated when a vdev removal is initiated.
193 	 *
194 	 * Note that this object can be empty if none of the vdev has yet been
195 	 * copied.
196 	 */
197 	uint64_t	vic_births_object;
198 
199 	/*
200 	 * This is the vdev ID which was removed previous to this vdev, or
201 	 * UINT64_MAX if there are no previously removed vdevs.
202 	 */
203 	uint64_t	vic_prev_indirect_vdev;
204 } vdev_indirect_config_t;
205 
206 /*
207  * Virtual device descriptor
208  */
209 struct vdev {
210 	/*
211 	 * Common to all vdev types.
212 	 */
213 	uint64_t	vdev_id;	/* child number in vdev parent	*/
214 	uint64_t	vdev_guid;	/* unique ID for this vdev	*/
215 	uint64_t	vdev_guid_sum;	/* self guid + all child guids	*/
216 	uint64_t	vdev_orig_guid;	/* orig. guid prior to remove	*/
217 	uint64_t	vdev_asize;	/* allocatable device capacity	*/
218 	uint64_t	vdev_min_asize;	/* min acceptable asize		*/
219 	uint64_t	vdev_max_asize;	/* max acceptable asize		*/
220 	uint64_t	vdev_ashift;	/* block alignment shift	*/
221 	uint64_t	vdev_state;	/* see VDEV_STATE_* #defines	*/
222 	uint64_t	vdev_prevstate;	/* used when reopening a vdev	*/
223 	vdev_ops_t	*vdev_ops;	/* vdev operations		*/
224 	spa_t		*vdev_spa;	/* spa for this vdev		*/
225 	void		*vdev_tsd;	/* type-specific data		*/
226 	vnode_t		*vdev_name_vp;	/* vnode for pathname		*/
227 	vnode_t		*vdev_devid_vp;	/* vnode for devid		*/
228 	vdev_t		*vdev_top;	/* top-level vdev		*/
229 	vdev_t		*vdev_parent;	/* parent vdev			*/
230 	vdev_t		**vdev_child;	/* array of children		*/
231 	uint64_t	vdev_children;	/* number of children		*/
232 	vdev_stat_t	vdev_stat;	/* virtual device statistics	*/
233 	vdev_stat_ex_t	vdev_stat_ex;	/* extended statistics		*/
234 	boolean_t	vdev_expanding;	/* expand the vdev?		*/
235 	boolean_t	vdev_reopening;	/* reopen in progress?		*/
236 	boolean_t	vdev_nonrot;	/* true if solid state		*/
237 	int		vdev_open_error; /* error on last open		*/
238 	kthread_t	*vdev_open_thread; /* thread opening children	*/
239 	uint64_t	vdev_crtxg;	/* txg when top-level was added */
240 
241 	/*
242 	 * Top-level vdev state.
243 	 */
244 	uint64_t	vdev_ms_array;	/* metaslab array object	*/
245 	uint64_t	vdev_ms_shift;	/* metaslab size shift		*/
246 	uint64_t	vdev_ms_count;	/* number of metaslabs		*/
247 	metaslab_group_t *vdev_mg;	/* metaslab group		*/
248 	metaslab_t	**vdev_ms;	/* metaslab array		*/
249 	txg_list_t	vdev_ms_list;	/* per-txg dirty metaslab lists	*/
250 	txg_list_t	vdev_dtl_list;	/* per-txg dirty DTL lists	*/
251 	txg_node_t	vdev_txg_node;	/* per-txg dirty vdev linkage	*/
252 	boolean_t	vdev_remove_wanted; /* async remove wanted?	*/
253 	boolean_t	vdev_probe_wanted; /* async probe wanted?	*/
254 	list_node_t	vdev_config_dirty_node; /* config dirty list	*/
255 	list_node_t	vdev_state_dirty_node; /* state dirty list	*/
256 	uint64_t	vdev_deflate_ratio; /* deflation ratio (x512)	*/
257 	uint64_t	vdev_islog;	/* is an intent log device	*/
258 	uint64_t	vdev_removing;	/* device is being removed?	*/
259 	boolean_t	vdev_ishole;	/* is a hole in the namespace	*/
260 	uint64_t	vdev_top_zap;
261 	vdev_alloc_bias_t vdev_alloc_bias; /* metaslab allocation bias	*/
262 
263 	/* pool checkpoint related */
264 	space_map_t	*vdev_checkpoint_sm;	/* contains reserved blocks */
265 
266 	/* Initialize related */
267 	boolean_t	vdev_initialize_exit_wanted;
268 	vdev_initializing_state_t	vdev_initialize_state;
269 	list_node_t	vdev_initialize_node;
270 	kthread_t	*vdev_initialize_thread;
271 	/* Protects vdev_initialize_thread and vdev_initialize_state. */
272 	kmutex_t	vdev_initialize_lock;
273 	kcondvar_t	vdev_initialize_cv;
274 	uint64_t	vdev_initialize_offset[TXG_SIZE];
275 	uint64_t	vdev_initialize_last_offset;
276 	range_tree_t	*vdev_initialize_tree;	/* valid while initializing */
277 	uint64_t	vdev_initialize_bytes_est;
278 	uint64_t	vdev_initialize_bytes_done;
279 	time_t		vdev_initialize_action_time;	/* start and end time */
280 
281 	/* TRIM related */
282 	boolean_t	vdev_trim_exit_wanted;
283 	boolean_t	vdev_autotrim_exit_wanted;
284 	vdev_trim_state_t	vdev_trim_state;
285 	list_node_t	vdev_trim_node;
286 	kmutex_t	vdev_autotrim_lock;
287 	kcondvar_t	vdev_autotrim_cv;
288 	kthread_t	*vdev_autotrim_thread;
289 	/* Protects vdev_trim_thread and vdev_trim_state. */
290 	kmutex_t	vdev_trim_lock;
291 	kcondvar_t	vdev_trim_cv;
292 	kthread_t	*vdev_trim_thread;
293 	uint64_t	vdev_trim_offset[TXG_SIZE];
294 	uint64_t	vdev_trim_last_offset;
295 	uint64_t	vdev_trim_bytes_est;
296 	uint64_t	vdev_trim_bytes_done;
297 	uint64_t	vdev_trim_rate;		/* requested rate (bytes/sec) */
298 	uint64_t	vdev_trim_partial;	/* requested partial TRIM */
299 	uint64_t	vdev_trim_secure;	/* requested secure TRIM */
300 	time_t		vdev_trim_action_time;	/* start and end time */
301 
302 	/* The following is not in ZoL, but used for auto-trim test progress */
303 	uint64_t	vdev_autotrim_bytes_done;
304 
305 	/* for limiting outstanding I/Os (initialize and TRIM) */
306 	kmutex_t	vdev_initialize_io_lock;
307 	kcondvar_t	vdev_initialize_io_cv;
308 	uint64_t	vdev_initialize_inflight;
309 	kmutex_t	vdev_trim_io_lock;
310 	kcondvar_t	vdev_trim_io_cv;
311 	uint64_t	vdev_trim_inflight[2];
312 
313 	/*
314 	 * Values stored in the config for an indirect or removing vdev.
315 	 */
316 	vdev_indirect_config_t	vdev_indirect_config;
317 
318 	/*
319 	 * The vdev_indirect_rwlock protects the vdev_indirect_mapping
320 	 * pointer from changing on indirect vdevs (when it is condensed).
321 	 * Note that removing (not yet indirect) vdevs have different
322 	 * access patterns (the mapping is not accessed from open context,
323 	 * e.g. from zio_read) and locking strategy (e.g. svr_lock).
324 	 */
325 	krwlock_t vdev_indirect_rwlock;
326 	vdev_indirect_mapping_t *vdev_indirect_mapping;
327 	vdev_indirect_births_t *vdev_indirect_births;
328 
329 	/*
330 	 * In memory data structures used to manage the obsolete sm, for
331 	 * indirect or removing vdevs.
332 	 *
333 	 * The vdev_obsolete_segments is the in-core record of the segments
334 	 * that are no longer referenced anywhere in the pool (due to
335 	 * being freed or remapped and not referenced by any snapshots).
336 	 * During a sync, segments are added to vdev_obsolete_segments
337 	 * via vdev_indirect_mark_obsolete(); at the end of each sync
338 	 * pass, this is appended to vdev_obsolete_sm via
339 	 * vdev_indirect_sync_obsolete().  The vdev_obsolete_lock
340 	 * protects against concurrent modifications of vdev_obsolete_segments
341 	 * from multiple zio threads.
342 	 */
343 	kmutex_t	vdev_obsolete_lock;
344 	range_tree_t	*vdev_obsolete_segments;
345 	space_map_t	*vdev_obsolete_sm;
346 
347 	/*
348 	 * Protects the vdev_scan_io_queue field itself as well as the
349 	 * structure's contents (when present).
350 	 */
351 	kmutex_t	vdev_scan_io_queue_lock;
352 	struct dsl_scan_io_queue	*vdev_scan_io_queue;
353 
354 	/*
355 	 * Leaf vdev state.
356 	 */
357 	range_tree_t	*vdev_dtl[DTL_TYPES]; /* dirty time logs	*/
358 	space_map_t	*vdev_dtl_sm;	/* dirty time log space map	*/
359 	txg_node_t	vdev_dtl_node;	/* per-txg dirty DTL linkage	*/
360 	uint64_t	vdev_dtl_object; /* DTL object			*/
361 	uint64_t	vdev_psize;	/* physical device capacity	*/
362 	uint64_t	vdev_wholedisk;	/* true if this is a whole disk */
363 	uint64_t	vdev_offline;	/* persistent offline state	*/
364 	uint64_t	vdev_faulted;	/* persistent faulted state	*/
365 	uint64_t	vdev_degraded;	/* persistent degraded state	*/
366 	uint64_t	vdev_removed;	/* persistent removed state	*/
367 	uint64_t	vdev_resilver_txg; /* persistent resilvering state */
368 	uint64_t	vdev_nparity;	/* number of parity devices for raidz */
369 	char		*vdev_path;	/* vdev path (if any)		*/
370 	char		*vdev_devid;	/* vdev devid (if any)		*/
371 	char		*vdev_physpath;	/* vdev device path (if any)	*/
372 	char		*vdev_fru;	/* physical FRU location	*/
373 	uint64_t	vdev_not_present; /* not present during import	*/
374 	uint64_t	vdev_unspare;	/* unspare when resilvering done */
375 	boolean_t	vdev_nowritecache; /* true if flushwritecache failed */
376 	boolean_t	vdev_has_trim;	/* TRIM is supported		*/
377 	boolean_t	vdev_has_securetrim; /* secure TRIM is supported */
378 	boolean_t	vdev_checkremove; /* temporary online test	*/
379 	boolean_t	vdev_forcefault; /* force online fault		*/
380 	boolean_t	vdev_splitting;	/* split or repair in progress  */
381 	boolean_t	vdev_delayed_close; /* delayed device close?	*/
382 	boolean_t	vdev_tmpoffline; /* device taken offline temporarily? */
383 	boolean_t	vdev_detached;	/* device detached?		*/
384 	boolean_t	vdev_cant_read;	/* vdev is failing all reads	*/
385 	boolean_t	vdev_cant_write; /* vdev is failing all writes	*/
386 	boolean_t	vdev_isspare;	/* was a hot spare		*/
387 	boolean_t	vdev_isl2cache;	/* was a l2cache device		*/
388 	boolean_t	vdev_resilver_deferred;  /* resilver deferred */
389 	vdev_queue_t	vdev_queue;	/* I/O deadline schedule queue	*/
390 	vdev_cache_t	vdev_cache;	/* physical block cache		*/
391 	spa_aux_vdev_t	*vdev_aux;	/* for l2cache and spares vdevs	*/
392 	zio_t		*vdev_probe_zio; /* root of current probe	*/
393 	vdev_aux_t	vdev_label_aux;	/* on-disk aux state		*/
394 	uint64_t	vdev_leaf_zap;
395 	hrtime_t	vdev_mmp_pending; /* 0 if write finished	*/
396 	uint64_t	vdev_mmp_kstat_id;	/* to find kstat entry */
397 	list_node_t	vdev_leaf_node;		/* leaf vdev list */
398 
399 	/*
400 	 * For DTrace to work in userland (libzpool) context, these fields must
401 	 * remain at the end of the structure.  DTrace will use the kernel's
402 	 * CTF definition for 'struct vdev', and since the size of a kmutex_t is
403 	 * larger in userland, the offsets for the rest of the fields would be
404 	 * incorrect.
405 	 */
406 	kmutex_t	vdev_dtl_lock;	/* vdev_dtl_{map,resilver}	*/
407 	kmutex_t	vdev_stat_lock;	/* vdev_stat			*/
408 	kmutex_t	vdev_probe_lock; /* protects vdev_probe_zio	*/
409 };
410 
411 #define	VDEV_RAIDZ_MAXPARITY	3
412 
413 #define	VDEV_PAD_SIZE		(8 << 10)
414 /* 2 padding areas (vl_pad1 and vl_pad2) to skip */
415 #define	VDEV_SKIP_SIZE		VDEV_PAD_SIZE * 2
416 #define	VDEV_PHYS_SIZE		(112 << 10)
417 #define	VDEV_UBERBLOCK_RING	(128 << 10)
418 
419 /*
420  * MMP blocks occupy the last MMP_BLOCKS_PER_LABEL slots in the uberblock
421  * ring when MMP is enabled.
422  */
423 #define	MMP_BLOCKS_PER_LABEL	1
424 
425 /* The largest uberblock we support is 8k. */
426 #define	MAX_UBERBLOCK_SHIFT (13)
427 #define	VDEV_UBERBLOCK_SHIFT(vd)	\
428 	MIN(MAX((vd)->vdev_top->vdev_ashift, UBERBLOCK_SHIFT), \
429 	    MAX_UBERBLOCK_SHIFT)
430 #define	VDEV_UBERBLOCK_COUNT(vd)	\
431 	(VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd))
432 #define	VDEV_UBERBLOCK_OFFSET(vd, n)	\
433 	offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)])
434 #define	VDEV_UBERBLOCK_SIZE(vd)		(1ULL << VDEV_UBERBLOCK_SHIFT(vd))
435 
436 typedef struct vdev_phys {
437 	char		vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)];
438 	zio_eck_t	vp_zbt;
439 } vdev_phys_t;
440 
441 typedef struct vdev_label {
442 	char		vl_pad1[VDEV_PAD_SIZE];			/*  8K */
443 	char		vl_pad2[VDEV_PAD_SIZE];			/*  8K */
444 	vdev_phys_t	vl_vdev_phys;				/* 112K	*/
445 	char		vl_uberblock[VDEV_UBERBLOCK_RING];	/* 128K	*/
446 } vdev_label_t;							/* 256K total */
447 
448 /*
449  * vdev_dirty() flags
450  */
451 #define	VDD_METASLAB	0x01
452 #define	VDD_DTL		0x02
453 
454 /* Offset of embedded boot loader region on each label */
455 #define	VDEV_BOOT_OFFSET	(2 * sizeof (vdev_label_t))
456 /*
457  * Size of embedded boot loader region on each label.
458  * The total size of the first two labels plus the boot area is 4MB.
459  */
460 #define	VDEV_BOOT_SIZE		(7ULL << 19)			/* 3.5M */
461 
462 /*
463  * Size of label regions at the start and end of each leaf device.
464  */
465 #define	VDEV_LABEL_START_SIZE	(2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE)
466 #define	VDEV_LABEL_END_SIZE	(2 * sizeof (vdev_label_t))
467 #define	VDEV_LABELS		4
468 #define	VDEV_BEST_LABEL		VDEV_LABELS
469 
470 #define	VDEV_ALLOC_LOAD		0
471 #define	VDEV_ALLOC_ADD		1
472 #define	VDEV_ALLOC_SPARE	2
473 #define	VDEV_ALLOC_L2CACHE	3
474 #define	VDEV_ALLOC_ROOTPOOL	4
475 #define	VDEV_ALLOC_SPLIT	5
476 #define	VDEV_ALLOC_ATTACH	6
477 
478 /*
479  * Allocate or free a vdev
480  */
481 extern vdev_t *vdev_alloc_common(spa_t *spa, uint_t id, uint64_t guid,
482     vdev_ops_t *ops);
483 extern int vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *config,
484     vdev_t *parent, uint_t id, int alloctype);
485 extern void vdev_free(vdev_t *vd);
486 
487 /*
488  * Add or remove children and parents
489  */
490 extern void vdev_add_child(vdev_t *pvd, vdev_t *cvd);
491 extern void vdev_remove_child(vdev_t *pvd, vdev_t *cvd);
492 extern void vdev_compact_children(vdev_t *pvd);
493 extern vdev_t *vdev_add_parent(vdev_t *cvd, vdev_ops_t *ops);
494 extern void vdev_remove_parent(vdev_t *cvd);
495 
496 /*
497  * vdev sync load and sync
498  */
499 extern boolean_t vdev_log_state_valid(vdev_t *vd);
500 extern int vdev_load(vdev_t *vd);
501 extern int vdev_dtl_load(vdev_t *vd);
502 extern void vdev_sync(vdev_t *vd, uint64_t txg);
503 extern void vdev_sync_done(vdev_t *vd, uint64_t txg);
504 extern void vdev_dirty(vdev_t *vd, int flags, void *arg, uint64_t txg);
505 extern void vdev_dirty_leaves(vdev_t *vd, int flags, uint64_t txg);
506 
507 /*
508  * Available vdev types.
509  */
510 extern vdev_ops_t vdev_root_ops;
511 extern vdev_ops_t vdev_mirror_ops;
512 extern vdev_ops_t vdev_replacing_ops;
513 extern vdev_ops_t vdev_raidz_ops;
514 extern vdev_ops_t vdev_disk_ops;
515 extern vdev_ops_t vdev_file_ops;
516 extern vdev_ops_t vdev_missing_ops;
517 extern vdev_ops_t vdev_hole_ops;
518 extern vdev_ops_t vdev_spare_ops;
519 extern vdev_ops_t vdev_indirect_ops;
520 
521 /*
522  * Common size functions
523  */
524 extern void vdev_default_xlate(vdev_t *vd, const range_seg64_t *in,
525     range_seg64_t *out);
526 extern uint64_t vdev_default_asize(vdev_t *vd, uint64_t psize);
527 extern uint64_t vdev_get_min_asize(vdev_t *vd);
528 extern void vdev_set_min_asize(vdev_t *vd);
529 
530 /*
531  * Global variables
532  */
533 extern int zfs_vdev_standard_sm_blksz;
534 /* zdb uses this tunable, so it must be declared here to make lint happy. */
535 extern int zfs_vdev_cache_size;
536 
537 /*
538  * Functions from vdev_indirect.c
539  */
540 extern void vdev_indirect_sync_obsolete(vdev_t *vd, dmu_tx_t *tx);
541 extern boolean_t vdev_indirect_should_condense(vdev_t *vd);
542 extern void spa_condense_indirect_start_sync(vdev_t *vd, dmu_tx_t *tx);
543 extern int vdev_obsolete_sm_object(vdev_t *vd);
544 extern boolean_t vdev_obsolete_counts_are_precise(vdev_t *vd);
545 
546 /*
547  * Other miscellaneous functions
548  */
549 int vdev_checkpoint_sm_object(vdev_t *vd);
550 
551 /*
552  * The vdev_buf_t is used to translate between zio_t and buf_t, and back again.
553  */
554 typedef struct vdev_buf {
555 	buf_t	vb_buf;		/* buffer that describes the io */
556 	zio_t	*vb_io;		/* pointer back to the original zio_t */
557 } vdev_buf_t;
558 
559 /*
560  * Support routines used during boot from a ZFS pool
561  */
562 extern int vdev_disk_read_rootlabel(const char *, const char *, nvlist_t **);
563 extern void vdev_disk_preroot_init(void);
564 extern void vdev_disk_preroot_fini(void);
565 extern const char *vdev_disk_preroot_lookup(uint64_t, uint64_t);
566 
567 #ifdef	__cplusplus
568 }
569 #endif
570 
571 #endif	/* _SYS_VDEV_IMPL_H */
572