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  */
26 
27 #ifndef _SYS_VDEV_IMPL_H
28 #define	_SYS_VDEV_IMPL_H
29 
30 #include <sys/avl.h>
31 #include <sys/bpobj.h>
32 #include <sys/dmu.h>
33 #include <sys/metaslab.h>
34 #include <sys/nvpair.h>
35 #include <sys/space_map.h>
36 #include <sys/vdev.h>
37 #include <sys/dkio.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_rebuild.h>
42 #include <sys/vdev_removal.h>
43 #include <sys/zfs_ratelimit.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 struct abd;
61 
62 extern uint_t zfs_vdev_queue_depth_pct;
63 extern uint_t zfs_vdev_def_queue_depth;
64 extern uint_t zfs_vdev_async_write_max_active;
65 
66 /*
67  * Virtual device operations
68  */
69 typedef int	vdev_init_func_t(spa_t *spa, nvlist_t *nv, void **tsd);
70 typedef void	vdev_kobj_post_evt_func_t(vdev_t *vd);
71 typedef void	vdev_fini_func_t(vdev_t *vd);
72 typedef int	vdev_open_func_t(vdev_t *vd, uint64_t *size, uint64_t *max_size,
73     uint64_t *ashift, uint64_t *pshift);
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 uint64_t vdev_min_asize_func_t(vdev_t *vd);
77 typedef uint64_t vdev_min_alloc_func_t(vdev_t *vd);
78 typedef void	vdev_io_start_func_t(zio_t *zio);
79 typedef void	vdev_io_done_func_t(zio_t *zio);
80 typedef void	vdev_state_change_func_t(vdev_t *vd, int, int);
81 typedef boolean_t vdev_need_resilver_func_t(vdev_t *vd, const dva_t *dva,
82     size_t psize, uint64_t phys_birth);
83 typedef void	vdev_hold_func_t(vdev_t *vd);
84 typedef void	vdev_rele_func_t(vdev_t *vd);
85 
86 typedef void	vdev_remap_cb_t(uint64_t inner_offset, vdev_t *vd,
87     uint64_t offset, uint64_t size, void *arg);
88 typedef void	vdev_remap_func_t(vdev_t *vd, uint64_t offset, uint64_t size,
89     vdev_remap_cb_t callback, void *arg);
90 /*
91  * Given a target vdev, translates the logical range "in" to the physical
92  * range "res"
93  */
94 typedef void vdev_xlation_func_t(vdev_t *cvd, const range_seg64_t *logical,
95     range_seg64_t *physical, range_seg64_t *remain);
96 typedef uint64_t vdev_rebuild_asize_func_t(vdev_t *vd, uint64_t start,
97     uint64_t size, uint64_t max_segment);
98 typedef void vdev_metaslab_init_func_t(vdev_t *vd, uint64_t *startp,
99     uint64_t *sizep);
100 typedef void vdev_config_generate_func_t(vdev_t *vd, nvlist_t *nv);
101 typedef uint64_t vdev_nparity_func_t(vdev_t *vd);
102 typedef uint64_t vdev_ndisks_func_t(vdev_t *vd);
103 
104 typedef const struct vdev_ops {
105 	vdev_init_func_t		*vdev_op_init;
106 	vdev_fini_func_t		*vdev_op_fini;
107 	vdev_open_func_t		*vdev_op_open;
108 	vdev_close_func_t		*vdev_op_close;
109 	vdev_asize_func_t		*vdev_op_asize;
110 	vdev_min_asize_func_t		*vdev_op_min_asize;
111 	vdev_min_alloc_func_t		*vdev_op_min_alloc;
112 	vdev_io_start_func_t		*vdev_op_io_start;
113 	vdev_io_done_func_t		*vdev_op_io_done;
114 	vdev_state_change_func_t	*vdev_op_state_change;
115 	vdev_need_resilver_func_t	*vdev_op_need_resilver;
116 	vdev_hold_func_t		*vdev_op_hold;
117 	vdev_rele_func_t		*vdev_op_rele;
118 	vdev_remap_func_t		*vdev_op_remap;
119 	vdev_xlation_func_t		*vdev_op_xlate;
120 	vdev_rebuild_asize_func_t	*vdev_op_rebuild_asize;
121 	vdev_metaslab_init_func_t	*vdev_op_metaslab_init;
122 	vdev_config_generate_func_t	*vdev_op_config_generate;
123 	vdev_nparity_func_t		*vdev_op_nparity;
124 	vdev_ndisks_func_t		*vdev_op_ndisks;
125 	vdev_kobj_post_evt_func_t	*vdev_op_kobj_evt_post;
126 	char				vdev_op_type[16];
127 	boolean_t			vdev_op_leaf;
128 } vdev_ops_t;
129 
130 /*
131  * Virtual device properties
132  */
133 typedef union vdev_queue_class {
134 	list_t		vqc_list;
135 	avl_tree_t	vqc_tree;
136 } vdev_queue_class_t;
137 
138 struct vdev_queue {
139 	vdev_t		*vq_vdev;
140 	vdev_queue_class_t vq_class[ZIO_PRIORITY_NUM_QUEUEABLE];
141 	avl_tree_t	vq_read_offset_tree;
142 	avl_tree_t	vq_write_offset_tree;
143 	uint64_t	vq_last_offset;
144 	zio_priority_t	vq_last_prio;	/* Last sent I/O priority. */
145 	uint32_t	vq_cqueued;	/* Classes with queued I/Os. */
146 	uint32_t	vq_cactive[ZIO_PRIORITY_NUM_QUEUEABLE];
147 	uint32_t	vq_active;	/* Number of active I/Os. */
148 	uint32_t	vq_ia_active;	/* Active interactive I/Os. */
149 	uint32_t	vq_nia_credit;	/* Non-interactive I/Os credit. */
150 	list_t		vq_active_list;	/* List of active I/Os. */
151 	hrtime_t	vq_io_complete_ts; /* time last i/o completed */
152 	hrtime_t	vq_io_delta_ts;
153 	zio_t		vq_io_search; /* used as local for stack reduction */
154 	kmutex_t	vq_lock;
155 };
156 
157 typedef enum vdev_alloc_bias {
158 	VDEV_BIAS_NONE,
159 	VDEV_BIAS_LOG,		/* dedicated to ZIL data (SLOG) */
160 	VDEV_BIAS_SPECIAL,	/* dedicated to ddt, metadata, and small blks */
161 	VDEV_BIAS_DEDUP		/* dedicated to dedup metadata */
162 } vdev_alloc_bias_t;
163 
164 
165 /*
166  * On-disk indirect vdev state.
167  *
168  * An indirect vdev is described exclusively in the MOS config of a pool.
169  * The config for an indirect vdev includes several fields, which are
170  * accessed in memory by a vdev_indirect_config_t.
171  */
172 typedef struct vdev_indirect_config {
173 	/*
174 	 * Object (in MOS) which contains the indirect mapping. This object
175 	 * contains an array of vdev_indirect_mapping_entry_phys_t ordered by
176 	 * vimep_src. The bonus buffer for this object is a
177 	 * vdev_indirect_mapping_phys_t. This object is allocated when a vdev
178 	 * removal is initiated.
179 	 *
180 	 * Note that this object can be empty if none of the data on the vdev
181 	 * has been copied yet.
182 	 */
183 	uint64_t	vic_mapping_object;
184 
185 	/*
186 	 * Object (in MOS) which contains the birth times for the mapping
187 	 * entries. This object contains an array of
188 	 * vdev_indirect_birth_entry_phys_t sorted by vibe_offset. The bonus
189 	 * buffer for this object is a vdev_indirect_birth_phys_t. This object
190 	 * is allocated when a vdev removal is initiated.
191 	 *
192 	 * Note that this object can be empty if none of the vdev has yet been
193 	 * copied.
194 	 */
195 	uint64_t	vic_births_object;
196 
197 	/*
198 	 * This is the vdev ID which was removed previous to this vdev, or
199 	 * UINT64_MAX if there are no previously removed vdevs.
200 	 */
201 	uint64_t	vic_prev_indirect_vdev;
202 } vdev_indirect_config_t;
203 
204 /*
205  * Virtual device descriptor
206  */
207 struct vdev {
208 	/*
209 	 * Common to all vdev types.
210 	 */
211 	uint64_t	vdev_id;	/* child number in vdev parent	*/
212 	uint64_t	vdev_guid;	/* unique ID for this vdev	*/
213 	uint64_t	vdev_guid_sum;	/* self guid + all child guids	*/
214 	uint64_t	vdev_orig_guid;	/* orig. guid prior to remove	*/
215 	uint64_t	vdev_asize;	/* allocatable device capacity	*/
216 	uint64_t	vdev_min_asize;	/* min acceptable asize		*/
217 	uint64_t	vdev_max_asize;	/* max acceptable asize		*/
218 	uint64_t	vdev_ashift;	/* block alignment shift	*/
219 
220 	/*
221 	 * Logical block alignment shift
222 	 *
223 	 * The smallest sized/aligned I/O supported by the device.
224 	 */
225 	uint64_t	vdev_logical_ashift;
226 	/*
227 	 * Physical block alignment shift
228 	 *
229 	 * The device supports logical I/Os with vdev_logical_ashift
230 	 * size/alignment, but optimum performance will be achieved by
231 	 * aligning/sizing requests to vdev_physical_ashift.  Smaller
232 	 * requests may be inflated or incur device level read-modify-write
233 	 * operations.
234 	 *
235 	 * May be 0 to indicate no preference (i.e. use vdev_logical_ashift).
236 	 */
237 	uint64_t	vdev_physical_ashift;
238 	uint64_t	vdev_state;	/* see VDEV_STATE_* #defines	*/
239 	uint64_t	vdev_prevstate;	/* used when reopening a vdev	*/
240 	vdev_ops_t	*vdev_ops;	/* vdev operations		*/
241 	spa_t		*vdev_spa;	/* spa for this vdev		*/
242 	void		*vdev_tsd;	/* type-specific data		*/
243 	vdev_t		*vdev_top;	/* top-level vdev		*/
244 	vdev_t		*vdev_parent;	/* parent vdev			*/
245 	vdev_t		**vdev_child;	/* array of children		*/
246 	uint64_t	vdev_children;	/* number of children		*/
247 	vdev_stat_t	vdev_stat;	/* virtual device statistics	*/
248 	vdev_stat_ex_t	vdev_stat_ex;	/* extended statistics		*/
249 	boolean_t	vdev_expanding;	/* expand the vdev?		*/
250 	boolean_t	vdev_reopening;	/* reopen in progress?		*/
251 	boolean_t	vdev_nonrot;	/* true if solid state		*/
252 	int		vdev_load_error; /* error on last load		*/
253 	int		vdev_open_error; /* error on last open		*/
254 	int		vdev_validate_error; /* error on last validate	*/
255 	kthread_t	*vdev_open_thread; /* thread opening children	*/
256 	kthread_t	*vdev_validate_thread; /* thread validating children */
257 	uint64_t	vdev_crtxg;	/* txg when top-level was added */
258 	uint64_t	vdev_root_zap;
259 
260 	/*
261 	 * Top-level vdev state.
262 	 */
263 	uint64_t	vdev_ms_array;	/* metaslab array object	*/
264 	uint64_t	vdev_ms_shift;	/* metaslab size shift		*/
265 	uint64_t	vdev_ms_count;	/* number of metaslabs		*/
266 	metaslab_group_t *vdev_mg;	/* metaslab group		*/
267 	metaslab_group_t *vdev_log_mg;	/* embedded slog metaslab group	*/
268 	metaslab_t	**vdev_ms;	/* metaslab array		*/
269 	txg_list_t	vdev_ms_list;	/* per-txg dirty metaslab lists	*/
270 	txg_list_t	vdev_dtl_list;	/* per-txg dirty DTL lists	*/
271 	txg_node_t	vdev_txg_node;	/* per-txg dirty vdev linkage	*/
272 	boolean_t	vdev_remove_wanted; /* async remove wanted?	*/
273 	boolean_t	vdev_probe_wanted; /* async probe wanted?	*/
274 	list_node_t	vdev_config_dirty_node; /* config dirty list	*/
275 	list_node_t	vdev_state_dirty_node; /* state dirty list	*/
276 	uint64_t	vdev_deflate_ratio; /* deflation ratio (x512)	*/
277 	uint64_t	vdev_islog;	/* is an intent log device	*/
278 	uint64_t	vdev_noalloc;	/* device is passivated?	*/
279 	uint64_t	vdev_removing;	/* device is being removed?	*/
280 	uint64_t	vdev_failfast;	/* device failfast setting	*/
281 	boolean_t	vdev_ishole;	/* is a hole in the namespace	*/
282 	uint64_t	vdev_top_zap;
283 	vdev_alloc_bias_t vdev_alloc_bias; /* metaslab allocation bias	*/
284 
285 	/* pool checkpoint related */
286 	space_map_t	*vdev_checkpoint_sm;	/* contains reserved blocks */
287 
288 	/* Initialize related */
289 	boolean_t	vdev_initialize_exit_wanted;
290 	vdev_initializing_state_t	vdev_initialize_state;
291 	list_node_t	vdev_initialize_node;
292 	kthread_t	*vdev_initialize_thread;
293 	/* Protects vdev_initialize_thread and vdev_initialize_state. */
294 	kmutex_t	vdev_initialize_lock;
295 	kcondvar_t	vdev_initialize_cv;
296 	uint64_t	vdev_initialize_offset[TXG_SIZE];
297 	uint64_t	vdev_initialize_last_offset;
298 	range_tree_t	*vdev_initialize_tree;	/* valid while initializing */
299 	uint64_t	vdev_initialize_bytes_est;
300 	uint64_t	vdev_initialize_bytes_done;
301 	uint64_t	vdev_initialize_action_time;	/* start and end time */
302 
303 	/* TRIM related */
304 	boolean_t	vdev_trim_exit_wanted;
305 	boolean_t	vdev_autotrim_exit_wanted;
306 	vdev_trim_state_t	vdev_trim_state;
307 	list_node_t	vdev_trim_node;
308 	kmutex_t	vdev_autotrim_lock;
309 	kcondvar_t	vdev_autotrim_cv;
310 	kcondvar_t	vdev_autotrim_kick_cv;
311 	kthread_t	*vdev_autotrim_thread;
312 	/* Protects vdev_trim_thread and vdev_trim_state. */
313 	kmutex_t	vdev_trim_lock;
314 	kcondvar_t	vdev_trim_cv;
315 	kthread_t	*vdev_trim_thread;
316 	uint64_t	vdev_trim_offset[TXG_SIZE];
317 	uint64_t	vdev_trim_last_offset;
318 	uint64_t	vdev_trim_bytes_est;
319 	uint64_t	vdev_trim_bytes_done;
320 	uint64_t	vdev_trim_rate;		/* requested rate (bytes/sec) */
321 	uint64_t	vdev_trim_partial;	/* requested partial TRIM */
322 	uint64_t	vdev_trim_secure;	/* requested secure TRIM */
323 	uint64_t	vdev_trim_action_time;	/* start and end time */
324 
325 	/* Rebuild related */
326 	boolean_t	vdev_rebuilding;
327 	boolean_t	vdev_rebuild_exit_wanted;
328 	boolean_t	vdev_rebuild_cancel_wanted;
329 	boolean_t	vdev_rebuild_reset_wanted;
330 	kmutex_t	vdev_rebuild_lock;
331 	kcondvar_t	vdev_rebuild_cv;
332 	kthread_t	*vdev_rebuild_thread;
333 	vdev_rebuild_t	vdev_rebuild_config;
334 
335 	/* For limiting outstanding I/Os (initialize, TRIM) */
336 	kmutex_t	vdev_initialize_io_lock;
337 	kcondvar_t	vdev_initialize_io_cv;
338 	uint64_t	vdev_initialize_inflight;
339 	kmutex_t	vdev_trim_io_lock;
340 	kcondvar_t	vdev_trim_io_cv;
341 	uint64_t	vdev_trim_inflight[3];
342 
343 	/*
344 	 * Values stored in the config for an indirect or removing vdev.
345 	 */
346 	vdev_indirect_config_t	vdev_indirect_config;
347 
348 	/*
349 	 * The vdev_indirect_rwlock protects the vdev_indirect_mapping
350 	 * pointer from changing on indirect vdevs (when it is condensed).
351 	 * Note that removing (not yet indirect) vdevs have different
352 	 * access patterns (the mapping is not accessed from open context,
353 	 * e.g. from zio_read) and locking strategy (e.g. svr_lock).
354 	 */
355 	krwlock_t vdev_indirect_rwlock;
356 	vdev_indirect_mapping_t *vdev_indirect_mapping;
357 	vdev_indirect_births_t *vdev_indirect_births;
358 
359 	/*
360 	 * In memory data structures used to manage the obsolete sm, for
361 	 * indirect or removing vdevs.
362 	 *
363 	 * The vdev_obsolete_segments is the in-core record of the segments
364 	 * that are no longer referenced anywhere in the pool (due to
365 	 * being freed or remapped and not referenced by any snapshots).
366 	 * During a sync, segments are added to vdev_obsolete_segments
367 	 * via vdev_indirect_mark_obsolete(); at the end of each sync
368 	 * pass, this is appended to vdev_obsolete_sm via
369 	 * vdev_indirect_sync_obsolete().  The vdev_obsolete_lock
370 	 * protects against concurrent modifications of vdev_obsolete_segments
371 	 * from multiple zio threads.
372 	 */
373 	kmutex_t	vdev_obsolete_lock;
374 	range_tree_t	*vdev_obsolete_segments;
375 	space_map_t	*vdev_obsolete_sm;
376 
377 	/*
378 	 * Protects the vdev_scan_io_queue field itself as well as the
379 	 * structure's contents (when present).
380 	 */
381 	kmutex_t			vdev_scan_io_queue_lock;
382 	struct dsl_scan_io_queue	*vdev_scan_io_queue;
383 
384 	/*
385 	 * Leaf vdev state.
386 	 */
387 	range_tree_t	*vdev_dtl[DTL_TYPES]; /* dirty time logs	*/
388 	space_map_t	*vdev_dtl_sm;	/* dirty time log space map	*/
389 	txg_node_t	vdev_dtl_node;	/* per-txg dirty DTL linkage	*/
390 	uint64_t	vdev_dtl_object; /* DTL object			*/
391 	uint64_t	vdev_psize;	/* physical device capacity	*/
392 	uint64_t	vdev_wholedisk;	/* true if this is a whole disk */
393 	uint64_t	vdev_offline;	/* persistent offline state	*/
394 	uint64_t	vdev_faulted;	/* persistent faulted state	*/
395 	uint64_t	vdev_degraded;	/* persistent degraded state	*/
396 	uint64_t	vdev_removed;	/* persistent removed state	*/
397 	uint64_t	vdev_resilver_txg; /* persistent resilvering state */
398 	uint64_t	vdev_rebuild_txg; /* persistent rebuilding state */
399 	char		*vdev_path;	/* vdev path (if any)		*/
400 	char		*vdev_devid;	/* vdev devid (if any)		*/
401 	char		*vdev_physpath;	/* vdev device path (if any)	*/
402 	char		*vdev_enc_sysfs_path;	/* enclosure sysfs path */
403 	char		*vdev_fru;	/* physical FRU location	*/
404 	uint64_t	vdev_not_present; /* not present during import	*/
405 	uint64_t	vdev_unspare;	/* unspare when resilvering done */
406 	boolean_t	vdev_nowritecache; /* true if flushwritecache failed */
407 	boolean_t	vdev_has_trim;	/* TRIM is supported		*/
408 	boolean_t	vdev_has_securetrim; /* secure TRIM is supported */
409 	boolean_t	vdev_checkremove; /* temporary online test	*/
410 	boolean_t	vdev_forcefault; /* force online fault		*/
411 	boolean_t	vdev_splitting;	/* split or repair in progress  */
412 	boolean_t	vdev_delayed_close; /* delayed device close?	*/
413 	boolean_t	vdev_tmpoffline; /* device taken offline temporarily? */
414 	boolean_t	vdev_detached;	/* device detached?		*/
415 	boolean_t	vdev_cant_read;	/* vdev is failing all reads	*/
416 	boolean_t	vdev_cant_write; /* vdev is failing all writes	*/
417 	boolean_t	vdev_isspare;	/* was a hot spare		*/
418 	boolean_t	vdev_isl2cache;	/* was a l2cache device		*/
419 	boolean_t	vdev_copy_uberblocks;  /* post expand copy uberblocks */
420 	boolean_t	vdev_resilver_deferred;  /* resilver deferred */
421 	boolean_t	vdev_kobj_flag; /* kobj event record */
422 	boolean_t	vdev_attaching; /* vdev attach ashift handling */
423 	vdev_queue_t	vdev_queue;	/* I/O deadline schedule queue	*/
424 	spa_aux_vdev_t	*vdev_aux;	/* for l2cache and spares vdevs	*/
425 	zio_t		*vdev_probe_zio; /* root of current probe	*/
426 	vdev_aux_t	vdev_label_aux;	/* on-disk aux state		*/
427 	uint64_t	vdev_leaf_zap;
428 	hrtime_t	vdev_mmp_pending; /* 0 if write finished	*/
429 	uint64_t	vdev_mmp_kstat_id;	/* to find kstat entry */
430 	uint64_t	vdev_expansion_time;	/* vdev's last expansion time */
431 	list_node_t	vdev_leaf_node;		/* leaf vdev list */
432 
433 	/*
434 	 * For DTrace to work in userland (libzpool) context, these fields must
435 	 * remain at the end of the structure.  DTrace will use the kernel's
436 	 * CTF definition for 'struct vdev', and since the size of a kmutex_t is
437 	 * larger in userland, the offsets for the rest of the fields would be
438 	 * incorrect.
439 	 */
440 	kmutex_t	vdev_dtl_lock;	/* vdev_dtl_{map,resilver}	*/
441 	kmutex_t	vdev_stat_lock;	/* vdev_stat			*/
442 	kmutex_t	vdev_probe_lock; /* protects vdev_probe_zio	*/
443 
444 	/*
445 	 * We rate limit ZIO delay, deadman, and checksum events, since they
446 	 * can flood ZED with tons of events when a drive is acting up.
447 	 */
448 	zfs_ratelimit_t vdev_delay_rl;
449 	zfs_ratelimit_t vdev_deadman_rl;
450 	zfs_ratelimit_t vdev_checksum_rl;
451 
452 	/*
453 	 * Checksum and IO thresholds for tuning ZED
454 	 */
455 	uint64_t	vdev_checksum_n;
456 	uint64_t	vdev_checksum_t;
457 	uint64_t	vdev_io_n;
458 	uint64_t	vdev_io_t;
459 };
460 
461 #define	VDEV_PAD_SIZE		(8 << 10)
462 /* 2 padding areas (vl_pad1 and vl_be) to skip */
463 #define	VDEV_SKIP_SIZE		VDEV_PAD_SIZE * 2
464 #define	VDEV_PHYS_SIZE		(112 << 10)
465 #define	VDEV_UBERBLOCK_RING	(128 << 10)
466 
467 /*
468  * MMP blocks occupy the last MMP_BLOCKS_PER_LABEL slots in the uberblock
469  * ring when MMP is enabled.
470  */
471 #define	MMP_BLOCKS_PER_LABEL	1
472 
473 /* The largest uberblock we support is 8k. */
474 #define	MAX_UBERBLOCK_SHIFT (13)
475 #define	VDEV_UBERBLOCK_SHIFT(vd)	\
476 	MIN(MAX((vd)->vdev_top->vdev_ashift, UBERBLOCK_SHIFT), \
477 	    MAX_UBERBLOCK_SHIFT)
478 #define	VDEV_UBERBLOCK_COUNT(vd)	\
479 	(VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd))
480 #define	VDEV_UBERBLOCK_OFFSET(vd, n)	\
481 	offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)])
482 #define	VDEV_UBERBLOCK_SIZE(vd)		(1ULL << VDEV_UBERBLOCK_SHIFT(vd))
483 
484 typedef struct vdev_phys {
485 	char		vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)];
486 	zio_eck_t	vp_zbt;
487 } vdev_phys_t;
488 
489 typedef enum vbe_vers {
490 	/*
491 	 * The bootenv file is stored as ascii text in the envblock.
492 	 * It is used by the GRUB bootloader used on Linux to store the
493 	 * contents of the grubenv file. The file is stored as raw ASCII,
494 	 * and is protected by an embedded checksum. By default, GRUB will
495 	 * check if the boot filesystem supports storing the environment data
496 	 * in a special location, and if so, will invoke filesystem specific
497 	 * logic to retrieve it. This can be overridden by a variable, should
498 	 * the user so desire.
499 	 */
500 	VB_RAW = 0,
501 
502 	/*
503 	 * The bootenv file is converted to an nvlist and then packed into the
504 	 * envblock.
505 	 */
506 	VB_NVLIST = 1
507 } vbe_vers_t;
508 
509 typedef struct vdev_boot_envblock {
510 	uint64_t	vbe_version;
511 	char		vbe_bootenv[VDEV_PAD_SIZE - sizeof (uint64_t) -
512 			sizeof (zio_eck_t)];
513 	zio_eck_t	vbe_zbt;
514 } vdev_boot_envblock_t;
515 _Static_assert(sizeof (vdev_boot_envblock_t) == VDEV_PAD_SIZE,
516 	"vdev_boot_envblock_t wrong size");
517 
518 typedef struct vdev_label {
519 	char		vl_pad1[VDEV_PAD_SIZE];			/*  8K */
520 	vdev_boot_envblock_t	vl_be;				/*  8K */
521 	vdev_phys_t	vl_vdev_phys;				/* 112K	*/
522 	char		vl_uberblock[VDEV_UBERBLOCK_RING];	/* 128K	*/
523 } vdev_label_t;						/* 256K total */
524 
525 /*
526  * vdev_dirty() flags
527  */
528 #define	VDD_METASLAB	0x01
529 #define	VDD_DTL		0x02
530 
531 /* Offset of embedded boot loader region on each label */
532 #define	VDEV_BOOT_OFFSET	(2 * sizeof (vdev_label_t))
533 /*
534  * Size of embedded boot loader region on each label.
535  * The total size of the first two labels plus the boot area is 4MB.
536  */
537 #define	VDEV_BOOT_SIZE		(7ULL << 19)			/* 3.5M */
538 
539 /*
540  * Size of label regions at the start and end of each leaf device.
541  */
542 #define	VDEV_LABEL_START_SIZE	(2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE)
543 #define	VDEV_LABEL_END_SIZE	(2 * sizeof (vdev_label_t))
544 #define	VDEV_LABELS		4
545 #define	VDEV_BEST_LABEL		VDEV_LABELS
546 #define	VDEV_OFFSET_IS_LABEL(vd, off)                           \
547 	(((off) < VDEV_LABEL_START_SIZE) ||                     \
548 	((off) >= ((vd)->vdev_psize - VDEV_LABEL_END_SIZE)))
549 
550 #define	VDEV_ALLOC_LOAD		0
551 #define	VDEV_ALLOC_ADD		1
552 #define	VDEV_ALLOC_SPARE	2
553 #define	VDEV_ALLOC_L2CACHE	3
554 #define	VDEV_ALLOC_ROOTPOOL	4
555 #define	VDEV_ALLOC_SPLIT	5
556 #define	VDEV_ALLOC_ATTACH	6
557 
558 /*
559  * Allocate or free a vdev
560  */
561 extern vdev_t *vdev_alloc_common(spa_t *spa, uint_t id, uint64_t guid,
562     vdev_ops_t *ops);
563 extern int vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *config,
564     vdev_t *parent, uint_t id, int alloctype);
565 extern void vdev_free(vdev_t *vd);
566 
567 /*
568  * Add or remove children and parents
569  */
570 extern void vdev_add_child(vdev_t *pvd, vdev_t *cvd);
571 extern void vdev_remove_child(vdev_t *pvd, vdev_t *cvd);
572 extern void vdev_compact_children(vdev_t *pvd);
573 extern vdev_t *vdev_add_parent(vdev_t *cvd, vdev_ops_t *ops);
574 extern void vdev_remove_parent(vdev_t *cvd);
575 
576 /*
577  * vdev sync load and sync
578  */
579 extern boolean_t vdev_log_state_valid(vdev_t *vd);
580 extern int vdev_load(vdev_t *vd);
581 extern int vdev_dtl_load(vdev_t *vd);
582 extern void vdev_sync(vdev_t *vd, uint64_t txg);
583 extern void vdev_sync_done(vdev_t *vd, uint64_t txg);
584 extern void vdev_dirty(vdev_t *vd, int flags, void *arg, uint64_t txg);
585 extern void vdev_dirty_leaves(vdev_t *vd, int flags, uint64_t txg);
586 
587 /*
588  * Available vdev types.
589  */
590 extern vdev_ops_t vdev_root_ops;
591 extern vdev_ops_t vdev_mirror_ops;
592 extern vdev_ops_t vdev_replacing_ops;
593 extern vdev_ops_t vdev_raidz_ops;
594 extern vdev_ops_t vdev_draid_ops;
595 extern vdev_ops_t vdev_draid_spare_ops;
596 extern vdev_ops_t vdev_disk_ops;
597 extern vdev_ops_t vdev_file_ops;
598 extern vdev_ops_t vdev_missing_ops;
599 extern vdev_ops_t vdev_hole_ops;
600 extern vdev_ops_t vdev_spare_ops;
601 extern vdev_ops_t vdev_indirect_ops;
602 
603 /*
604  * Common size functions
605  */
606 extern void vdev_default_xlate(vdev_t *vd, const range_seg64_t *logical_rs,
607     range_seg64_t *physical_rs, range_seg64_t *remain_rs);
608 extern uint64_t vdev_default_asize(vdev_t *vd, uint64_t psize);
609 extern uint64_t vdev_default_min_asize(vdev_t *vd);
610 extern uint64_t vdev_get_min_asize(vdev_t *vd);
611 extern void vdev_set_min_asize(vdev_t *vd);
612 extern uint64_t vdev_get_min_alloc(vdev_t *vd);
613 extern uint64_t vdev_get_nparity(vdev_t *vd);
614 extern uint64_t vdev_get_ndisks(vdev_t *vd);
615 
616 /*
617  * Global variables
618  */
619 extern int zfs_vdev_standard_sm_blksz;
620 
621 /*
622  * Functions from vdev_indirect.c
623  */
624 extern void vdev_indirect_sync_obsolete(vdev_t *vd, dmu_tx_t *tx);
625 extern boolean_t vdev_indirect_should_condense(vdev_t *vd);
626 extern void spa_condense_indirect_start_sync(vdev_t *vd, dmu_tx_t *tx);
627 extern int vdev_obsolete_sm_object(vdev_t *vd, uint64_t *sm_obj);
628 extern int vdev_obsolete_counts_are_precise(vdev_t *vd, boolean_t *are_precise);
629 
630 /*
631  * Other miscellaneous functions
632  */
633 int vdev_checkpoint_sm_object(vdev_t *vd, uint64_t *sm_obj);
634 void vdev_metaslab_group_create(vdev_t *vd);
635 uint64_t vdev_best_ashift(uint64_t logical, uint64_t a, uint64_t b);
636 
637 /*
638  * Vdev ashift optimization tunables
639  */
640 extern uint_t zfs_vdev_min_auto_ashift;
641 extern uint_t zfs_vdev_max_auto_ashift;
642 int param_set_min_auto_ashift(ZFS_MODULE_PARAM_ARGS);
643 int param_set_max_auto_ashift(ZFS_MODULE_PARAM_ARGS);
644 
645 #ifdef	__cplusplus
646 }
647 #endif
648 
649 #endif	/* _SYS_VDEV_IMPL_H */
650