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 https://opensource.org/licenses/CDDL-1.0.
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, 2020 by Delphix. All rights reserved.
24  * Copyright (c) 2017, Intel Corporation.
25  * Copyright (c) 2023, Klara Inc.
26  */
27 
28 #ifndef _SYS_VDEV_IMPL_H
29 #define	_SYS_VDEV_IMPL_H
30 
31 #include <sys/avl.h>
32 #include <sys/bpobj.h>
33 #include <sys/dmu.h>
34 #include <sys/metaslab.h>
35 #include <sys/nvpair.h>
36 #include <sys/space_map.h>
37 #include <sys/vdev.h>
38 #include <sys/dkio.h>
39 #include <sys/uberblock_impl.h>
40 #include <sys/vdev_indirect_mapping.h>
41 #include <sys/vdev_indirect_births.h>
42 #include <sys/vdev_rebuild.h>
43 #include <sys/vdev_removal.h>
44 #include <sys/zfs_ratelimit.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 struct abd;
62 
63 extern uint_t zfs_vdev_queue_depth_pct;
64 extern uint_t zfs_vdev_def_queue_depth;
65 extern uint_t zfs_vdev_async_write_max_active;
66 
67 /*
68  * Virtual device operations
69  */
70 typedef int	vdev_init_func_t(spa_t *spa, nvlist_t *nv, void **tsd);
71 typedef void	vdev_kobj_post_evt_func_t(vdev_t *vd);
72 typedef void	vdev_fini_func_t(vdev_t *vd);
73 typedef int	vdev_open_func_t(vdev_t *vd, uint64_t *size, uint64_t *max_size,
74     uint64_t *ashift, uint64_t *pshift);
75 typedef void	vdev_close_func_t(vdev_t *vd);
76 typedef uint64_t vdev_asize_func_t(vdev_t *vd, uint64_t psize);
77 typedef uint64_t vdev_min_asize_func_t(vdev_t *vd);
78 typedef uint64_t vdev_min_alloc_func_t(vdev_t *vd);
79 typedef void	vdev_io_start_func_t(zio_t *zio);
80 typedef void	vdev_io_done_func_t(zio_t *zio);
81 typedef void	vdev_state_change_func_t(vdev_t *vd, int, int);
82 typedef boolean_t vdev_need_resilver_func_t(vdev_t *vd, const dva_t *dva,
83     size_t psize, uint64_t phys_birth);
84 typedef void	vdev_hold_func_t(vdev_t *vd);
85 typedef void	vdev_rele_func_t(vdev_t *vd);
86 
87 typedef void	vdev_remap_cb_t(uint64_t inner_offset, vdev_t *vd,
88     uint64_t offset, uint64_t size, void *arg);
89 typedef void	vdev_remap_func_t(vdev_t *vd, uint64_t offset, uint64_t size,
90     vdev_remap_cb_t callback, void *arg);
91 /*
92  * Given a target vdev, translates the logical range "in" to the physical
93  * range "res"
94  */
95 typedef void vdev_xlation_func_t(vdev_t *cvd, const range_seg64_t *logical,
96     range_seg64_t *physical, range_seg64_t *remain);
97 typedef uint64_t vdev_rebuild_asize_func_t(vdev_t *vd, uint64_t start,
98     uint64_t size, uint64_t max_segment);
99 typedef void vdev_metaslab_init_func_t(vdev_t *vd, uint64_t *startp,
100     uint64_t *sizep);
101 typedef void vdev_config_generate_func_t(vdev_t *vd, nvlist_t *nv);
102 typedef uint64_t vdev_nparity_func_t(vdev_t *vd);
103 typedef uint64_t vdev_ndisks_func_t(vdev_t *vd);
104 
105 typedef const struct vdev_ops {
106 	vdev_init_func_t		*vdev_op_init;
107 	vdev_fini_func_t		*vdev_op_fini;
108 	vdev_open_func_t		*vdev_op_open;
109 	vdev_close_func_t		*vdev_op_close;
110 	vdev_asize_func_t		*vdev_op_asize;
111 	vdev_min_asize_func_t		*vdev_op_min_asize;
112 	vdev_min_alloc_func_t		*vdev_op_min_alloc;
113 	vdev_io_start_func_t		*vdev_op_io_start;
114 	vdev_io_done_func_t		*vdev_op_io_done;
115 	vdev_state_change_func_t	*vdev_op_state_change;
116 	vdev_need_resilver_func_t	*vdev_op_need_resilver;
117 	vdev_hold_func_t		*vdev_op_hold;
118 	vdev_rele_func_t		*vdev_op_rele;
119 	vdev_remap_func_t		*vdev_op_remap;
120 	vdev_xlation_func_t		*vdev_op_xlate;
121 	vdev_rebuild_asize_func_t	*vdev_op_rebuild_asize;
122 	vdev_metaslab_init_func_t	*vdev_op_metaslab_init;
123 	vdev_config_generate_func_t	*vdev_op_config_generate;
124 	vdev_nparity_func_t		*vdev_op_nparity;
125 	vdev_ndisks_func_t		*vdev_op_ndisks;
126 	vdev_kobj_post_evt_func_t	*vdev_op_kobj_evt_post;
127 	char				vdev_op_type[16];
128 	boolean_t			vdev_op_leaf;
129 } vdev_ops_t;
130 
131 /*
132  * Virtual device properties
133  */
134 typedef union vdev_queue_class {
135 	struct {
136 		ulong_t 	vqc_list_numnodes;
137 		list_t		vqc_list;
138 	};
139 	avl_tree_t	vqc_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_read_offset_tree;
146 	avl_tree_t	vq_write_offset_tree;
147 	uint64_t	vq_last_offset;
148 	zio_priority_t	vq_last_prio;	/* Last sent I/O priority. */
149 	uint32_t	vq_cqueued;	/* Classes with queued I/Os. */
150 	uint32_t	vq_cactive[ZIO_PRIORITY_NUM_QUEUEABLE];
151 	uint32_t	vq_active;	/* Number of active I/Os. */
152 	uint32_t	vq_ia_active;	/* Active interactive I/Os. */
153 	uint32_t	vq_nia_credit;	/* Non-interactive I/Os credit. */
154 	list_t		vq_active_list;	/* List of active I/Os. */
155 	hrtime_t	vq_io_complete_ts; /* time last i/o completed */
156 	hrtime_t	vq_io_delta_ts;
157 	zio_t		vq_io_search; /* used as local for stack reduction */
158 	kmutex_t	vq_lock;
159 };
160 
161 typedef enum vdev_alloc_bias {
162 	VDEV_BIAS_NONE,
163 	VDEV_BIAS_LOG,		/* dedicated to ZIL data (SLOG) */
164 	VDEV_BIAS_SPECIAL,	/* dedicated to ddt, metadata, and small blks */
165 	VDEV_BIAS_DEDUP		/* dedicated to dedup metadata */
166 } vdev_alloc_bias_t;
167 
168 
169 /*
170  * On-disk indirect vdev state.
171  *
172  * An indirect vdev is described exclusively in the MOS config of a pool.
173  * The config for an indirect vdev includes several fields, which are
174  * accessed in memory by a vdev_indirect_config_t.
175  */
176 typedef struct vdev_indirect_config {
177 	/*
178 	 * Object (in MOS) which contains the indirect mapping. This object
179 	 * contains an array of vdev_indirect_mapping_entry_phys_t ordered by
180 	 * vimep_src. The bonus buffer for this object is a
181 	 * vdev_indirect_mapping_phys_t. This object is allocated when a vdev
182 	 * removal is initiated.
183 	 *
184 	 * Note that this object can be empty if none of the data on the vdev
185 	 * has been copied yet.
186 	 */
187 	uint64_t	vic_mapping_object;
188 
189 	/*
190 	 * Object (in MOS) which contains the birth times for the mapping
191 	 * entries. This object contains an array of
192 	 * vdev_indirect_birth_entry_phys_t sorted by vibe_offset. The bonus
193 	 * buffer for this object is a vdev_indirect_birth_phys_t. This object
194 	 * is allocated when a vdev removal is initiated.
195 	 *
196 	 * Note that this object can be empty if none of the vdev has yet been
197 	 * copied.
198 	 */
199 	uint64_t	vic_births_object;
200 
201 	/*
202 	 * This is the vdev ID which was removed previous to this vdev, or
203 	 * UINT64_MAX if there are no previously removed vdevs.
204 	 */
205 	uint64_t	vic_prev_indirect_vdev;
206 } vdev_indirect_config_t;
207 
208 /*
209  * Virtual device descriptor
210  */
211 struct vdev {
212 	/*
213 	 * Common to all vdev types.
214 	 */
215 	uint64_t	vdev_id;	/* child number in vdev parent	*/
216 	uint64_t	vdev_guid;	/* unique ID for this vdev	*/
217 	uint64_t	vdev_guid_sum;	/* self guid + all child guids	*/
218 	uint64_t	vdev_orig_guid;	/* orig. guid prior to remove	*/
219 	uint64_t	vdev_asize;	/* allocatable device capacity	*/
220 	uint64_t	vdev_min_asize;	/* min acceptable asize		*/
221 	uint64_t	vdev_max_asize;	/* max acceptable asize		*/
222 	uint64_t	vdev_ashift;	/* block alignment shift	*/
223 
224 	/*
225 	 * Logical block alignment shift
226 	 *
227 	 * The smallest sized/aligned I/O supported by the device.
228 	 */
229 	uint64_t	vdev_logical_ashift;
230 	/*
231 	 * Physical block alignment shift
232 	 *
233 	 * The device supports logical I/Os with vdev_logical_ashift
234 	 * size/alignment, but optimum performance will be achieved by
235 	 * aligning/sizing requests to vdev_physical_ashift.  Smaller
236 	 * requests may be inflated or incur device level read-modify-write
237 	 * operations.
238 	 *
239 	 * May be 0 to indicate no preference (i.e. use vdev_logical_ashift).
240 	 */
241 	uint64_t	vdev_physical_ashift;
242 	uint64_t	vdev_state;	/* see VDEV_STATE_* #defines	*/
243 	uint64_t	vdev_prevstate;	/* used when reopening a vdev	*/
244 	vdev_ops_t	*vdev_ops;	/* vdev operations		*/
245 	spa_t		*vdev_spa;	/* spa for this vdev		*/
246 	void		*vdev_tsd;	/* type-specific data		*/
247 	vdev_t		*vdev_top;	/* top-level vdev		*/
248 	vdev_t		*vdev_parent;	/* parent vdev			*/
249 	vdev_t		**vdev_child;	/* array of children		*/
250 	uint64_t	vdev_children;	/* number of children		*/
251 	vdev_stat_t	vdev_stat;	/* virtual device statistics	*/
252 	vdev_stat_ex_t	vdev_stat_ex;	/* extended statistics		*/
253 	boolean_t	vdev_expanding;	/* expand the vdev?		*/
254 	boolean_t	vdev_reopening;	/* reopen in progress?		*/
255 	boolean_t	vdev_nonrot;	/* true if solid state		*/
256 	int		vdev_load_error; /* error on last load		*/
257 	int		vdev_open_error; /* error on last open		*/
258 	int		vdev_validate_error; /* error on last validate	*/
259 	kthread_t	*vdev_open_thread; /* thread opening children	*/
260 	kthread_t	*vdev_validate_thread; /* thread validating children */
261 	uint64_t	vdev_crtxg;	/* txg when top-level was added */
262 	uint64_t	vdev_root_zap;
263 
264 	/*
265 	 * Top-level vdev state.
266 	 */
267 	uint64_t	vdev_ms_array;	/* metaslab array object	*/
268 	uint64_t	vdev_ms_shift;	/* metaslab size shift		*/
269 	uint64_t	vdev_ms_count;	/* number of metaslabs		*/
270 	metaslab_group_t *vdev_mg;	/* metaslab group		*/
271 	metaslab_group_t *vdev_log_mg;	/* embedded slog metaslab group	*/
272 	metaslab_t	**vdev_ms;	/* metaslab array		*/
273 	txg_list_t	vdev_ms_list;	/* per-txg dirty metaslab lists	*/
274 	txg_list_t	vdev_dtl_list;	/* per-txg dirty DTL lists	*/
275 	txg_node_t	vdev_txg_node;	/* per-txg dirty vdev linkage	*/
276 	boolean_t	vdev_remove_wanted; /* async remove wanted?	*/
277 	boolean_t	vdev_fault_wanted; /* async faulted wanted?	*/
278 	list_node_t	vdev_config_dirty_node; /* config dirty list	*/
279 	list_node_t	vdev_state_dirty_node; /* state dirty list	*/
280 	uint64_t	vdev_deflate_ratio; /* deflation ratio (x512)	*/
281 	uint64_t	vdev_islog;	/* is an intent log device	*/
282 	uint64_t	vdev_noalloc;	/* device is passivated?	*/
283 	uint64_t	vdev_removing;	/* device is being removed?	*/
284 	uint64_t	vdev_failfast;	/* device failfast setting	*/
285 	boolean_t	vdev_ishole;	/* is a hole in the namespace	*/
286 	uint64_t	vdev_top_zap;
287 	vdev_alloc_bias_t vdev_alloc_bias; /* metaslab allocation bias	*/
288 
289 	/* pool checkpoint related */
290 	space_map_t	*vdev_checkpoint_sm;	/* contains reserved blocks */
291 
292 	/* Initialize related */
293 	boolean_t	vdev_initialize_exit_wanted;
294 	vdev_initializing_state_t	vdev_initialize_state;
295 	list_node_t	vdev_initialize_node;
296 	kthread_t	*vdev_initialize_thread;
297 	/* Protects vdev_initialize_thread and vdev_initialize_state. */
298 	kmutex_t	vdev_initialize_lock;
299 	kcondvar_t	vdev_initialize_cv;
300 	uint64_t	vdev_initialize_offset[TXG_SIZE];
301 	uint64_t	vdev_initialize_last_offset;
302 	range_tree_t	*vdev_initialize_tree;	/* valid while initializing */
303 	uint64_t	vdev_initialize_bytes_est;
304 	uint64_t	vdev_initialize_bytes_done;
305 	uint64_t	vdev_initialize_action_time;	/* start and end time */
306 
307 	/* TRIM related */
308 	boolean_t	vdev_trim_exit_wanted;
309 	boolean_t	vdev_autotrim_exit_wanted;
310 	vdev_trim_state_t	vdev_trim_state;
311 	list_node_t	vdev_trim_node;
312 	kmutex_t	vdev_autotrim_lock;
313 	kcondvar_t	vdev_autotrim_cv;
314 	kcondvar_t	vdev_autotrim_kick_cv;
315 	kthread_t	*vdev_autotrim_thread;
316 	/* Protects vdev_trim_thread and vdev_trim_state. */
317 	kmutex_t	vdev_trim_lock;
318 	kcondvar_t	vdev_trim_cv;
319 	kthread_t	*vdev_trim_thread;
320 	uint64_t	vdev_trim_offset[TXG_SIZE];
321 	uint64_t	vdev_trim_last_offset;
322 	uint64_t	vdev_trim_bytes_est;
323 	uint64_t	vdev_trim_bytes_done;
324 	uint64_t	vdev_trim_rate;		/* requested rate (bytes/sec) */
325 	uint64_t	vdev_trim_partial;	/* requested partial TRIM */
326 	uint64_t	vdev_trim_secure;	/* requested secure TRIM */
327 	uint64_t	vdev_trim_action_time;	/* start and end time */
328 
329 	/* Rebuild related */
330 	boolean_t	vdev_rebuilding;
331 	boolean_t	vdev_rebuild_exit_wanted;
332 	boolean_t	vdev_rebuild_cancel_wanted;
333 	boolean_t	vdev_rebuild_reset_wanted;
334 	kmutex_t	vdev_rebuild_lock;
335 	kcondvar_t	vdev_rebuild_cv;
336 	kthread_t	*vdev_rebuild_thread;
337 	vdev_rebuild_t	vdev_rebuild_config;
338 
339 	/* For limiting outstanding I/Os (initialize, TRIM) */
340 	kmutex_t	vdev_initialize_io_lock;
341 	kcondvar_t	vdev_initialize_io_cv;
342 	uint64_t	vdev_initialize_inflight;
343 	kmutex_t	vdev_trim_io_lock;
344 	kcondvar_t	vdev_trim_io_cv;
345 	uint64_t	vdev_trim_inflight[3];
346 
347 	/*
348 	 * Values stored in the config for an indirect or removing vdev.
349 	 */
350 	vdev_indirect_config_t	vdev_indirect_config;
351 
352 	/*
353 	 * The vdev_indirect_rwlock protects the vdev_indirect_mapping
354 	 * pointer from changing on indirect vdevs (when it is condensed).
355 	 * Note that removing (not yet indirect) vdevs have different
356 	 * access patterns (the mapping is not accessed from open context,
357 	 * e.g. from zio_read) and locking strategy (e.g. svr_lock).
358 	 */
359 	krwlock_t vdev_indirect_rwlock;
360 	vdev_indirect_mapping_t *vdev_indirect_mapping;
361 	vdev_indirect_births_t *vdev_indirect_births;
362 
363 	/*
364 	 * In memory data structures used to manage the obsolete sm, for
365 	 * indirect or removing vdevs.
366 	 *
367 	 * The vdev_obsolete_segments is the in-core record of the segments
368 	 * that are no longer referenced anywhere in the pool (due to
369 	 * being freed or remapped and not referenced by any snapshots).
370 	 * During a sync, segments are added to vdev_obsolete_segments
371 	 * via vdev_indirect_mark_obsolete(); at the end of each sync
372 	 * pass, this is appended to vdev_obsolete_sm via
373 	 * vdev_indirect_sync_obsolete().  The vdev_obsolete_lock
374 	 * protects against concurrent modifications of vdev_obsolete_segments
375 	 * from multiple zio threads.
376 	 */
377 	kmutex_t	vdev_obsolete_lock;
378 	range_tree_t	*vdev_obsolete_segments;
379 	space_map_t	*vdev_obsolete_sm;
380 
381 	/*
382 	 * Protects the vdev_scan_io_queue field itself as well as the
383 	 * structure's contents (when present).
384 	 */
385 	kmutex_t			vdev_scan_io_queue_lock;
386 	struct dsl_scan_io_queue	*vdev_scan_io_queue;
387 
388 	/*
389 	 * Leaf vdev state.
390 	 */
391 	range_tree_t	*vdev_dtl[DTL_TYPES]; /* dirty time logs	*/
392 	space_map_t	*vdev_dtl_sm;	/* dirty time log space map	*/
393 	txg_node_t	vdev_dtl_node;	/* per-txg dirty DTL linkage	*/
394 	uint64_t	vdev_dtl_object; /* DTL object			*/
395 	uint64_t	vdev_psize;	/* physical device capacity	*/
396 	uint64_t	vdev_wholedisk;	/* true if this is a whole disk */
397 	uint64_t	vdev_offline;	/* persistent offline state	*/
398 	uint64_t	vdev_faulted;	/* persistent faulted state	*/
399 	uint64_t	vdev_degraded;	/* persistent degraded state	*/
400 	uint64_t	vdev_removed;	/* persistent removed state	*/
401 	uint64_t	vdev_resilver_txg; /* persistent resilvering state */
402 	uint64_t	vdev_rebuild_txg; /* persistent rebuilding state */
403 	char		*vdev_path;	/* vdev path (if any)		*/
404 	char		*vdev_devid;	/* vdev devid (if any)		*/
405 	char		*vdev_physpath;	/* vdev device path (if any)	*/
406 	char		*vdev_enc_sysfs_path;	/* enclosure sysfs path */
407 	char		*vdev_fru;	/* physical FRU location	*/
408 	uint64_t	vdev_not_present; /* not present during import	*/
409 	uint64_t	vdev_unspare;	/* unspare when resilvering done */
410 	boolean_t	vdev_nowritecache; /* true if flushwritecache failed */
411 	boolean_t	vdev_has_trim;	/* TRIM is supported		*/
412 	boolean_t	vdev_has_securetrim; /* secure TRIM is supported */
413 	boolean_t	vdev_checkremove; /* temporary online test	*/
414 	boolean_t	vdev_forcefault; /* force online fault		*/
415 	boolean_t	vdev_splitting;	/* split or repair in progress  */
416 	boolean_t	vdev_delayed_close; /* delayed device close?	*/
417 	boolean_t	vdev_tmpoffline; /* device taken offline temporarily? */
418 	boolean_t	vdev_detached;	/* device detached?		*/
419 	boolean_t	vdev_cant_read;	/* vdev is failing all reads	*/
420 	boolean_t	vdev_cant_write; /* vdev is failing all writes	*/
421 	boolean_t	vdev_isspare;	/* was a hot spare		*/
422 	boolean_t	vdev_isl2cache;	/* was a l2cache device		*/
423 	boolean_t	vdev_copy_uberblocks;  /* post expand copy uberblocks */
424 	boolean_t	vdev_resilver_deferred;  /* resilver deferred */
425 	boolean_t	vdev_kobj_flag; /* kobj event record */
426 	boolean_t	vdev_attaching; /* vdev attach ashift handling */
427 	vdev_queue_t	vdev_queue;	/* I/O deadline schedule queue	*/
428 	spa_aux_vdev_t	*vdev_aux;	/* for l2cache and spares vdevs	*/
429 	zio_t		*vdev_probe_zio; /* root of current probe	*/
430 	vdev_aux_t	vdev_label_aux;	/* on-disk aux state		*/
431 	uint64_t	vdev_leaf_zap;
432 	hrtime_t	vdev_mmp_pending; /* 0 if write finished	*/
433 	uint64_t	vdev_mmp_kstat_id;	/* to find kstat entry */
434 	uint64_t	vdev_expansion_time;	/* vdev's last expansion time */
435 	list_node_t	vdev_leaf_node;		/* leaf vdev list */
436 
437 	/*
438 	 * For DTrace to work in userland (libzpool) context, these fields must
439 	 * remain at the end of the structure.  DTrace will use the kernel's
440 	 * CTF definition for 'struct vdev', and since the size of a kmutex_t is
441 	 * larger in userland, the offsets for the rest of the fields would be
442 	 * incorrect.
443 	 */
444 	kmutex_t	vdev_dtl_lock;	/* vdev_dtl_{map,resilver}	*/
445 	kmutex_t	vdev_stat_lock;	/* vdev_stat			*/
446 	kmutex_t	vdev_probe_lock; /* protects vdev_probe_zio	*/
447 
448 	/*
449 	 * We rate limit ZIO delay, deadman, and checksum events, since they
450 	 * can flood ZED with tons of events when a drive is acting up.
451 	 */
452 	zfs_ratelimit_t vdev_delay_rl;
453 	zfs_ratelimit_t vdev_deadman_rl;
454 	zfs_ratelimit_t vdev_checksum_rl;
455 
456 	/*
457 	 * Vdev properties for tuning ZED or zfsd
458 	 */
459 	uint64_t	vdev_checksum_n;
460 	uint64_t	vdev_checksum_t;
461 	uint64_t	vdev_io_n;
462 	uint64_t	vdev_io_t;
463 	uint64_t	vdev_slow_io_n;
464 	uint64_t	vdev_slow_io_t;
465 };
466 
467 #define	VDEV_PAD_SIZE		(8 << 10)
468 /* 2 padding areas (vl_pad1 and vl_be) to skip */
469 #define	VDEV_SKIP_SIZE		VDEV_PAD_SIZE * 2
470 #define	VDEV_PHYS_SIZE		(112 << 10)
471 #define	VDEV_UBERBLOCK_RING	(128 << 10)
472 
473 /*
474  * MMP blocks occupy the last MMP_BLOCKS_PER_LABEL slots in the uberblock
475  * ring when MMP is enabled.
476  */
477 #define	MMP_BLOCKS_PER_LABEL	1
478 
479 /* The largest uberblock we support is 8k. */
480 #define	MAX_UBERBLOCK_SHIFT (13)
481 #define	VDEV_UBERBLOCK_SHIFT(vd)	\
482 	MIN(MAX((vd)->vdev_top->vdev_ashift, UBERBLOCK_SHIFT), \
483 	    MAX_UBERBLOCK_SHIFT)
484 #define	VDEV_UBERBLOCK_COUNT(vd)	\
485 	(VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd))
486 #define	VDEV_UBERBLOCK_OFFSET(vd, n)	\
487 	offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)])
488 #define	VDEV_UBERBLOCK_SIZE(vd)		(1ULL << VDEV_UBERBLOCK_SHIFT(vd))
489 
490 typedef struct vdev_phys {
491 	char		vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)];
492 	zio_eck_t	vp_zbt;
493 } vdev_phys_t;
494 
495 typedef enum vbe_vers {
496 	/*
497 	 * The bootenv file is stored as ascii text in the envblock.
498 	 * It is used by the GRUB bootloader used on Linux to store the
499 	 * contents of the grubenv file. The file is stored as raw ASCII,
500 	 * and is protected by an embedded checksum. By default, GRUB will
501 	 * check if the boot filesystem supports storing the environment data
502 	 * in a special location, and if so, will invoke filesystem specific
503 	 * logic to retrieve it. This can be overridden by a variable, should
504 	 * the user so desire.
505 	 */
506 	VB_RAW = 0,
507 
508 	/*
509 	 * The bootenv file is converted to an nvlist and then packed into the
510 	 * envblock.
511 	 */
512 	VB_NVLIST = 1
513 } vbe_vers_t;
514 
515 typedef struct vdev_boot_envblock {
516 	uint64_t	vbe_version;
517 	char		vbe_bootenv[VDEV_PAD_SIZE - sizeof (uint64_t) -
518 			sizeof (zio_eck_t)];
519 	zio_eck_t	vbe_zbt;
520 } vdev_boot_envblock_t;
521 _Static_assert(sizeof (vdev_boot_envblock_t) == VDEV_PAD_SIZE,
522 	"vdev_boot_envblock_t wrong size");
523 
524 typedef struct vdev_label {
525 	char		vl_pad1[VDEV_PAD_SIZE];			/*  8K */
526 	vdev_boot_envblock_t	vl_be;				/*  8K */
527 	vdev_phys_t	vl_vdev_phys;				/* 112K	*/
528 	char		vl_uberblock[VDEV_UBERBLOCK_RING];	/* 128K	*/
529 } vdev_label_t;						/* 256K total */
530 
531 /*
532  * vdev_dirty() flags
533  */
534 #define	VDD_METASLAB	0x01
535 #define	VDD_DTL		0x02
536 
537 /* Offset of embedded boot loader region on each label */
538 #define	VDEV_BOOT_OFFSET	(2 * sizeof (vdev_label_t))
539 /*
540  * Size of embedded boot loader region on each label.
541  * The total size of the first two labels plus the boot area is 4MB.
542  */
543 #define	VDEV_BOOT_SIZE		(7ULL << 19)			/* 3.5M */
544 
545 /*
546  * Size of label regions at the start and end of each leaf device.
547  */
548 #define	VDEV_LABEL_START_SIZE	(2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE)
549 #define	VDEV_LABEL_END_SIZE	(2 * sizeof (vdev_label_t))
550 #define	VDEV_LABELS		4
551 #define	VDEV_BEST_LABEL		VDEV_LABELS
552 #define	VDEV_OFFSET_IS_LABEL(vd, off)                           \
553 	(((off) < VDEV_LABEL_START_SIZE) ||                     \
554 	((off) >= ((vd)->vdev_psize - VDEV_LABEL_END_SIZE)))
555 
556 #define	VDEV_ALLOC_LOAD		0
557 #define	VDEV_ALLOC_ADD		1
558 #define	VDEV_ALLOC_SPARE	2
559 #define	VDEV_ALLOC_L2CACHE	3
560 #define	VDEV_ALLOC_ROOTPOOL	4
561 #define	VDEV_ALLOC_SPLIT	5
562 #define	VDEV_ALLOC_ATTACH	6
563 
564 /*
565  * Allocate or free a vdev
566  */
567 extern vdev_t *vdev_alloc_common(spa_t *spa, uint_t id, uint64_t guid,
568     vdev_ops_t *ops);
569 extern int vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *config,
570     vdev_t *parent, uint_t id, int alloctype);
571 extern void vdev_free(vdev_t *vd);
572 
573 /*
574  * Add or remove children and parents
575  */
576 extern void vdev_add_child(vdev_t *pvd, vdev_t *cvd);
577 extern void vdev_remove_child(vdev_t *pvd, vdev_t *cvd);
578 extern void vdev_compact_children(vdev_t *pvd);
579 extern vdev_t *vdev_add_parent(vdev_t *cvd, vdev_ops_t *ops);
580 extern void vdev_remove_parent(vdev_t *cvd);
581 
582 /*
583  * vdev sync load and sync
584  */
585 extern boolean_t vdev_log_state_valid(vdev_t *vd);
586 extern int vdev_load(vdev_t *vd);
587 extern int vdev_dtl_load(vdev_t *vd);
588 extern void vdev_sync(vdev_t *vd, uint64_t txg);
589 extern void vdev_sync_done(vdev_t *vd, uint64_t txg);
590 extern void vdev_dirty(vdev_t *vd, int flags, void *arg, uint64_t txg);
591 extern void vdev_dirty_leaves(vdev_t *vd, int flags, uint64_t txg);
592 
593 /*
594  * Available vdev types.
595  */
596 extern vdev_ops_t vdev_root_ops;
597 extern vdev_ops_t vdev_mirror_ops;
598 extern vdev_ops_t vdev_replacing_ops;
599 extern vdev_ops_t vdev_raidz_ops;
600 extern vdev_ops_t vdev_draid_ops;
601 extern vdev_ops_t vdev_draid_spare_ops;
602 extern vdev_ops_t vdev_disk_ops;
603 extern vdev_ops_t vdev_file_ops;
604 extern vdev_ops_t vdev_missing_ops;
605 extern vdev_ops_t vdev_hole_ops;
606 extern vdev_ops_t vdev_spare_ops;
607 extern vdev_ops_t vdev_indirect_ops;
608 
609 /*
610  * Common size functions
611  */
612 extern void vdev_default_xlate(vdev_t *vd, const range_seg64_t *logical_rs,
613     range_seg64_t *physical_rs, range_seg64_t *remain_rs);
614 extern uint64_t vdev_default_asize(vdev_t *vd, uint64_t psize);
615 extern uint64_t vdev_default_min_asize(vdev_t *vd);
616 extern uint64_t vdev_get_min_asize(vdev_t *vd);
617 extern void vdev_set_min_asize(vdev_t *vd);
618 extern uint64_t vdev_get_min_alloc(vdev_t *vd);
619 extern uint64_t vdev_get_nparity(vdev_t *vd);
620 extern uint64_t vdev_get_ndisks(vdev_t *vd);
621 
622 /*
623  * Global variables
624  */
625 extern int zfs_vdev_standard_sm_blksz;
626 
627 /*
628  * Functions from vdev_indirect.c
629  */
630 extern void vdev_indirect_sync_obsolete(vdev_t *vd, dmu_tx_t *tx);
631 extern boolean_t vdev_indirect_should_condense(vdev_t *vd);
632 extern void spa_condense_indirect_start_sync(vdev_t *vd, dmu_tx_t *tx);
633 extern int vdev_obsolete_sm_object(vdev_t *vd, uint64_t *sm_obj);
634 extern int vdev_obsolete_counts_are_precise(vdev_t *vd, boolean_t *are_precise);
635 
636 /*
637  * Other miscellaneous functions
638  */
639 int vdev_checkpoint_sm_object(vdev_t *vd, uint64_t *sm_obj);
640 void vdev_metaslab_group_create(vdev_t *vd);
641 uint64_t vdev_best_ashift(uint64_t logical, uint64_t a, uint64_t b);
642 
643 /*
644  * Vdev ashift optimization tunables
645  */
646 extern uint_t zfs_vdev_min_auto_ashift;
647 extern uint_t zfs_vdev_max_auto_ashift;
648 int param_set_min_auto_ashift(ZFS_MODULE_PARAM_ARGS);
649 int param_set_max_auto_ashift(ZFS_MODULE_PARAM_ARGS);
650 
651 #ifdef	__cplusplus
652 }
653 #endif
654 
655 #endif	/* _SYS_VDEV_IMPL_H */
656