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