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