xref: /illumos-gate/usr/src/uts/common/fs/zfs/sys/spa_impl.h (revision eb633035)
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 2011 Nexenta Systems, Inc.  All rights reserved.
25  * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
26  * Copyright 2013 Saso Kiselkov. All rights reserved.
27  * Copyright (c) 2017 Datto Inc.
28  * Copyright (c) 2017, Intel Corporation.
29  */
30 
31 #ifndef _SYS_SPA_IMPL_H
32 #define	_SYS_SPA_IMPL_H
33 
34 #include <sys/spa.h>
35 #include <sys/spa_checkpoint.h>
36 #include <sys/vdev.h>
37 #include <sys/vdev_removal.h>
38 #include <sys/metaslab.h>
39 #include <sys/dmu.h>
40 #include <sys/dsl_pool.h>
41 #include <sys/uberblock_impl.h>
42 #include <sys/zfs_context.h>
43 #include <sys/avl.h>
44 #include <sys/refcount.h>
45 #include <sys/bplist.h>
46 #include <sys/bpobj.h>
47 #include <sys/dsl_crypt.h>
48 #include <sys/zfeature.h>
49 #include <sys/zthr.h>
50 #include <zfeature_common.h>
51 
52 #ifdef	__cplusplus
53 extern "C" {
54 #endif
55 
56 typedef struct spa_error_entry {
57 	zbookmark_phys_t	se_bookmark;
58 	char			*se_name;
59 	avl_node_t		se_avl;
60 } spa_error_entry_t;
61 
62 typedef struct spa_history_phys {
63 	uint64_t sh_pool_create_len;	/* ending offset of zpool create */
64 	uint64_t sh_phys_max_off;	/* physical EOF */
65 	uint64_t sh_bof;		/* logical BOF */
66 	uint64_t sh_eof;		/* logical EOF */
67 	uint64_t sh_records_lost;	/* num of records overwritten */
68 } spa_history_phys_t;
69 
70 /*
71  * All members must be uint64_t, for byteswap purposes.
72  */
73 typedef struct spa_removing_phys {
74 	uint64_t sr_state; /* dsl_scan_state_t */
75 
76 	/*
77 	 * The vdev ID that we most recently attempted to remove,
78 	 * or -1 if no removal has been attempted.
79 	 */
80 	uint64_t sr_removing_vdev;
81 
82 	/*
83 	 * The vdev ID that we most recently successfully removed,
84 	 * or -1 if no devices have been removed.
85 	 */
86 	uint64_t sr_prev_indirect_vdev;
87 
88 	uint64_t sr_start_time;
89 	uint64_t sr_end_time;
90 
91 	/*
92 	 * Note that we can not use the space map's or indirect mapping's
93 	 * accounting as a substitute for these values, because we need to
94 	 * count frees of not-yet-copied data as though it did the copy.
95 	 * Otherwise, we could get into a situation where copied > to_copy,
96 	 * or we complete before copied == to_copy.
97 	 */
98 	uint64_t sr_to_copy; /* bytes that need to be copied */
99 	uint64_t sr_copied; /* bytes that have been copied or freed */
100 } spa_removing_phys_t;
101 
102 /*
103  * This struct is stored as an entry in the DMU_POOL_DIRECTORY_OBJECT
104  * (with key DMU_POOL_CONDENSING_INDIRECT).  It is present if a condense
105  * of an indirect vdev's mapping object is in progress.
106  */
107 typedef struct spa_condensing_indirect_phys {
108 	/*
109 	 * The vdev ID of the indirect vdev whose indirect mapping is
110 	 * being condensed.
111 	 */
112 	uint64_t	scip_vdev;
113 
114 	/*
115 	 * The vdev's old obsolete spacemap.  This spacemap's contents are
116 	 * being integrated into the new mapping.
117 	 */
118 	uint64_t	scip_prev_obsolete_sm_object;
119 
120 	/*
121 	 * The new mapping object that is being created.
122 	 */
123 	uint64_t	scip_next_mapping_object;
124 } spa_condensing_indirect_phys_t;
125 
126 struct spa_aux_vdev {
127 	uint64_t	sav_object;		/* MOS object for device list */
128 	nvlist_t	*sav_config;		/* cached device config */
129 	vdev_t		**sav_vdevs;		/* devices */
130 	int		sav_count;		/* number devices */
131 	boolean_t	sav_sync;		/* sync the device list */
132 	nvlist_t	**sav_pending;		/* pending device additions */
133 	uint_t		sav_npending;		/* # pending devices */
134 };
135 
136 typedef struct spa_config_lock {
137 	kmutex_t	scl_lock;
138 	kthread_t	*scl_writer;
139 	int		scl_write_wanted;
140 	kcondvar_t	scl_cv;
141 	zfs_refcount_t	scl_count;
142 } spa_config_lock_t;
143 
144 typedef struct spa_config_dirent {
145 	list_node_t	scd_link;
146 	char		*scd_path;
147 } spa_config_dirent_t;
148 
149 typedef enum zio_taskq_type {
150 	ZIO_TASKQ_ISSUE = 0,
151 	ZIO_TASKQ_ISSUE_HIGH,
152 	ZIO_TASKQ_INTERRUPT,
153 	ZIO_TASKQ_INTERRUPT_HIGH,
154 	ZIO_TASKQ_TYPES
155 } zio_taskq_type_t;
156 
157 /*
158  * State machine for the zpool-poolname process.  The states transitions
159  * are done as follows:
160  *
161  *	From		   To			Routine
162  *	PROC_NONE	-> PROC_CREATED		spa_activate()
163  *	PROC_CREATED	-> PROC_ACTIVE		spa_thread()
164  *	PROC_ACTIVE	-> PROC_DEACTIVATE	spa_deactivate()
165  *	PROC_DEACTIVATE	-> PROC_GONE		spa_thread()
166  *	PROC_GONE	-> PROC_NONE		spa_deactivate()
167  */
168 typedef enum spa_proc_state {
169 	SPA_PROC_NONE,		/* spa_proc = &p0, no process created */
170 	SPA_PROC_CREATED,	/* spa_activate() has proc, is waiting */
171 	SPA_PROC_ACTIVE,	/* taskqs created, spa_proc set */
172 	SPA_PROC_DEACTIVATE,	/* spa_deactivate() requests process exit */
173 	SPA_PROC_GONE		/* spa_thread() is exiting, spa_proc = &p0 */
174 } spa_proc_state_t;
175 
176 typedef struct spa_taskqs {
177 	uint_t stqs_count;
178 	taskq_t **stqs_taskq;
179 } spa_taskqs_t;
180 
181 typedef enum spa_all_vdev_zap_action {
182 	AVZ_ACTION_NONE = 0,
183 	AVZ_ACTION_DESTROY,	/* Destroy all per-vdev ZAPs and the AVZ. */
184 	AVZ_ACTION_REBUILD,	/* Populate the new AVZ, see spa_avz_rebuild */
185 	AVZ_ACTION_INITIALIZE
186 } spa_avz_action_t;
187 
188 typedef enum spa_config_source {
189 	SPA_CONFIG_SRC_NONE = 0,
190 	SPA_CONFIG_SRC_SCAN,		/* scan of path (default: /dev/dsk) */
191 	SPA_CONFIG_SRC_CACHEFILE,	/* any cachefile */
192 	SPA_CONFIG_SRC_TRYIMPORT,	/* returned from call to tryimport */
193 	SPA_CONFIG_SRC_SPLIT,		/* new pool in a pool split */
194 	SPA_CONFIG_SRC_MOS		/* MOS, but not always from right txg */
195 } spa_config_source_t;
196 
197 struct spa {
198 	/*
199 	 * Fields protected by spa_namespace_lock.
200 	 */
201 	char		spa_name[ZFS_MAX_DATASET_NAME_LEN];	/* pool name */
202 	char		*spa_comment;		/* comment */
203 	avl_node_t	spa_avl;		/* node in spa_namespace_avl */
204 	nvlist_t	*spa_config;		/* last synced config */
205 	nvlist_t	*spa_config_syncing;	/* currently syncing config */
206 	nvlist_t	*spa_config_splitting;	/* config for splitting */
207 	nvlist_t	*spa_load_info;		/* info and errors from load */
208 	uint64_t	spa_config_txg;		/* txg of last config change */
209 	int		spa_sync_pass;		/* iterate-to-convergence */
210 	pool_state_t	spa_state;		/* pool state */
211 	int		spa_inject_ref;		/* injection references */
212 	uint8_t		spa_sync_on;		/* sync threads are running */
213 	spa_load_state_t spa_load_state;	/* current load operation */
214 	boolean_t	spa_indirect_vdevs_loaded; /* mappings loaded? */
215 	boolean_t	spa_trust_config;	/* do we trust vdev tree? */
216 	spa_config_source_t spa_config_source;	/* where config comes from? */
217 	uint64_t	spa_import_flags;	/* import specific flags */
218 	spa_taskqs_t	spa_zio_taskq[ZIO_TYPES][ZIO_TASKQ_TYPES];
219 	dsl_pool_t	*spa_dsl_pool;
220 	boolean_t	spa_is_initializing;	/* true while opening pool */
221 	metaslab_class_t *spa_normal_class;	/* normal data class */
222 	metaslab_class_t *spa_log_class;	/* intent log data class */
223 	metaslab_class_t *spa_special_class;	/* special allocation class */
224 	metaslab_class_t *spa_dedup_class;	/* dedup allocation class */
225 	uint64_t	spa_first_txg;		/* first txg after spa_open() */
226 	uint64_t	spa_final_txg;		/* txg of export/destroy */
227 	uint64_t	spa_freeze_txg;		/* freeze pool at this txg */
228 	uint64_t	spa_load_max_txg;	/* best initial ub_txg */
229 	uint64_t	spa_claim_max_txg;	/* highest claimed birth txg */
230 	timespec_t	spa_loaded_ts;		/* 1st successful open time */
231 	objset_t	*spa_meta_objset;	/* copy of dp->dp_meta_objset */
232 	kmutex_t	spa_evicting_os_lock;	/* Evicting objset list lock */
233 	list_t		spa_evicting_os_list;	/* Objsets being evicted. */
234 	kcondvar_t	spa_evicting_os_cv;	/* Objset Eviction Completion */
235 	txg_list_t	spa_vdev_txg_list;	/* per-txg dirty vdev list */
236 	vdev_t		*spa_root_vdev;		/* top-level vdev container */
237 	int		spa_min_ashift;		/* of vdevs in normal class */
238 	int		spa_max_ashift;		/* of vdevs in normal class */
239 	uint64_t	spa_config_guid;	/* config pool guid */
240 	uint64_t	spa_load_guid;		/* spa_load initialized guid */
241 	uint64_t	spa_last_synced_guid;	/* last synced guid */
242 	list_t		spa_config_dirty_list;	/* vdevs with dirty config */
243 	list_t		spa_state_dirty_list;	/* vdevs with dirty state */
244 	/*
245 	 * spa_alloc_locks and spa_alloc_trees are arrays, whose lengths are
246 	 * stored in spa_alloc_count. There is one tree and one lock for each
247 	 * allocator, to help improve allocation performance in write-heavy
248 	 * workloads.
249 	 */
250 	kmutex_t	*spa_alloc_locks;
251 	avl_tree_t	*spa_alloc_trees;
252 	int		spa_alloc_count;
253 
254 	spa_aux_vdev_t	spa_spares;		/* hot spares */
255 	spa_aux_vdev_t	spa_l2cache;		/* L2ARC cache devices */
256 	nvlist_t	*spa_label_features;	/* Features for reading MOS */
257 	uint64_t	spa_config_object;	/* MOS object for pool config */
258 	uint64_t	spa_config_generation;	/* config generation number */
259 	uint64_t	spa_syncing_txg;	/* txg currently syncing */
260 	bpobj_t		spa_deferred_bpobj;	/* deferred-free bplist */
261 	bplist_t	spa_free_bplist[TXG_SIZE]; /* bplist of stuff to free */
262 	zio_cksum_salt_t spa_cksum_salt;	/* secret salt for cksum */
263 	/* checksum context templates */
264 	kmutex_t	spa_cksum_tmpls_lock;
265 	void		*spa_cksum_tmpls[ZIO_CHECKSUM_FUNCTIONS];
266 	uberblock_t	spa_ubsync;		/* last synced uberblock */
267 	uberblock_t	spa_uberblock;		/* current uberblock */
268 	boolean_t	spa_extreme_rewind;	/* rewind past deferred frees */
269 	kmutex_t	spa_scrub_lock;		/* resilver/scrub lock */
270 	uint64_t	spa_scrub_inflight;	/* in-flight scrub bytes */
271 	uint64_t	spa_load_verify_ios;	/* in-flight verification IOs */
272 	kcondvar_t	spa_scrub_io_cv;	/* scrub I/O completion */
273 	uint8_t		spa_scrub_active;	/* active or suspended? */
274 	uint8_t		spa_scrub_type;		/* type of scrub we're doing */
275 	uint8_t		spa_scrub_finished;	/* indicator to rotate logs */
276 	uint8_t		spa_scrub_started;	/* started since last boot */
277 	uint8_t		spa_scrub_reopen;	/* scrub doing vdev_reopen */
278 	uint64_t	spa_scan_pass_start;	/* start time per pass/reboot */
279 	uint64_t	spa_scan_pass_scrub_pause; /* scrub pause time */
280 	uint64_t	spa_scan_pass_scrub_spent_paused; /* total paused */
281 	uint64_t	spa_scan_pass_exam;	/* examined bytes per pass */
282 	uint64_t	spa_scan_pass_issued;	/* issued bytes per pass */
283 
284 	/*
285 	 * We are in the middle of a resilver, and another resilver
286 	 * is needed once this one completes. This is set iff any
287 	 * vdev_resilver_deferred is set.
288 	 */
289 	boolean_t	spa_resilver_deferred;
290 	kmutex_t	spa_async_lock;		/* protect async state */
291 	kthread_t	*spa_async_thread;	/* thread doing async task */
292 	int		spa_async_suspended;	/* async tasks suspended */
293 	kcondvar_t	spa_async_cv;		/* wait for thread_exit() */
294 	uint16_t	spa_async_tasks;	/* async task mask */
295 	uint64_t	spa_missing_tvds;	/* unopenable tvds on load */
296 	uint64_t	spa_missing_tvds_allowed; /* allow loading spa? */
297 
298 	spa_removing_phys_t spa_removing_phys;
299 	spa_vdev_removal_t *spa_vdev_removal;
300 
301 	spa_condensing_indirect_phys_t	spa_condensing_indirect_phys;
302 	spa_condensing_indirect_t	*spa_condensing_indirect;
303 	zthr_t		*spa_condense_zthr;	/* zthr doing condense. */
304 
305 	uint64_t	spa_checkpoint_txg;	/* the txg of the checkpoint */
306 	spa_checkpoint_info_t spa_checkpoint_info; /* checkpoint accounting */
307 	zthr_t		*spa_checkpoint_discard_zthr;
308 
309 	char		*spa_root;		/* alternate root directory */
310 	uint64_t	spa_ena;		/* spa-wide ereport ENA */
311 	int		spa_last_open_failed;	/* error if last open failed */
312 	uint64_t	spa_last_ubsync_txg;	/* "best" uberblock txg */
313 	uint64_t	spa_last_ubsync_txg_ts;	/* timestamp from that ub */
314 	uint64_t	spa_load_txg;		/* ub txg that loaded */
315 	uint64_t	spa_load_txg_ts;	/* timestamp from that ub */
316 	uint64_t	spa_load_meta_errors;	/* verify metadata err count */
317 	uint64_t	spa_load_data_errors;	/* verify data err count */
318 	uint64_t	spa_verify_min_txg;	/* start txg of verify scrub */
319 	kmutex_t	spa_errlog_lock;	/* error log lock */
320 	uint64_t	spa_errlog_last;	/* last error log object */
321 	uint64_t	spa_errlog_scrub;	/* scrub error log object */
322 	kmutex_t	spa_errlist_lock;	/* error list/ereport lock */
323 	avl_tree_t	spa_errlist_last;	/* last error list */
324 	avl_tree_t	spa_errlist_scrub;	/* scrub error list */
325 	uint64_t	spa_deflate;		/* should we deflate? */
326 	uint64_t	spa_history;		/* history object */
327 	kmutex_t	spa_history_lock;	/* history lock */
328 	vdev_t		*spa_pending_vdev;	/* pending vdev additions */
329 	kmutex_t	spa_props_lock;		/* property lock */
330 	uint64_t	spa_pool_props_object;	/* object for properties */
331 	uint64_t	spa_bootfs;		/* default boot filesystem */
332 	uint64_t	spa_failmode;		/* failure mode for the pool */
333 	uint64_t	spa_delegation;		/* delegation on/off */
334 	list_t		spa_config_list;	/* previous cache file(s) */
335 	/* per-CPU array of root of async I/O: */
336 	zio_t		**spa_async_zio_root;
337 	zio_t		*spa_suspend_zio_root;	/* root of all suspended I/O */
338 	zio_t		*spa_txg_zio[TXG_SIZE]; /* spa_sync() waits for this */
339 	kmutex_t	spa_suspend_lock;	/* protects suspend_zio_root */
340 	kcondvar_t	spa_suspend_cv;		/* notification of resume */
341 	zio_suspend_reason_t	spa_suspended;	/* pool is suspended */
342 	uint8_t		spa_claiming;		/* pool is doing zil_claim() */
343 	boolean_t	spa_is_root;		/* pool is root */
344 	int		spa_minref;		/* num refs when first opened */
345 	int		spa_mode;		/* FREAD | FWRITE */
346 	spa_log_state_t spa_log_state;		/* log state */
347 	uint64_t	spa_autoexpand;		/* lun expansion on/off */
348 	uint64_t	spa_bootsize;		/* efi system partition size */
349 	ddt_t		*spa_ddt[ZIO_CHECKSUM_FUNCTIONS]; /* in-core DDTs */
350 	uint64_t	spa_ddt_stat_object;	/* DDT statistics */
351 	uint64_t	spa_dedup_ditto;	/* dedup ditto threshold */
352 	uint64_t	spa_dedup_checksum;	/* default dedup checksum */
353 	uint64_t	spa_dspace;		/* dspace in normal class */
354 	kmutex_t	spa_vdev_top_lock;	/* dueling offline/remove */
355 	kmutex_t	spa_proc_lock;		/* protects spa_proc* */
356 	kcondvar_t	spa_proc_cv;		/* spa_proc_state transitions */
357 	spa_proc_state_t spa_proc_state;	/* see definition */
358 	struct proc	*spa_proc;		/* "zpool-poolname" process */
359 	uint64_t	spa_did;		/* if procp != p0, did of t1 */
360 	boolean_t	spa_autoreplace;	/* autoreplace set in open */
361 	int		spa_vdev_locks;		/* locks grabbed */
362 	uint64_t	spa_creation_version;	/* version at pool creation */
363 	uint64_t	spa_prev_software_version; /* See ub_software_version */
364 	uint64_t	spa_feat_for_write_obj;	/* required to write to pool */
365 	uint64_t	spa_feat_for_read_obj;	/* required to read from pool */
366 	uint64_t	spa_feat_desc_obj;	/* Feature descriptions */
367 	uint64_t	spa_feat_enabled_txg_obj; /* Feature enabled txg */
368 	/* cache feature refcounts */
369 	uint64_t	spa_feat_refcount_cache[SPA_FEATURES];
370 	cyclic_id_t	spa_deadman_cycid;	/* cyclic id */
371 	uint64_t	spa_deadman_calls;	/* number of deadman calls */
372 	hrtime_t	spa_sync_starttime;	/* starting time fo spa_sync */
373 	uint64_t	spa_deadman_synctime;	/* deadman expiration timer */
374 	uint64_t	spa_all_vdev_zaps;	/* ZAP of per-vd ZAP obj #s */
375 	spa_avz_action_t	spa_avz_action;	/* destroy/rebuild AVZ? */
376 	spa_keystore_t	spa_keystore;	/* loaded crypto keys */
377 	uint64_t	spa_errata;	/* errata issues detected */
378 
379 	/*
380 	 * spa_iokstat_lock protects spa_iokstat and
381 	 * spa_queue_stats[].
382 	 */
383 	kmutex_t	spa_iokstat_lock;
384 	struct kstat	*spa_iokstat;		/* kstat of io to this pool */
385 	struct {
386 		int spa_active;
387 		int spa_queued;
388 	} spa_queue_stats[ZIO_PRIORITY_NUM_QUEUEABLE];
389 
390 	/* arc_memory_throttle() parameters during low memory condition */
391 	uint64_t	spa_lowmem_page_load;	/* memory load during txg */
392 	uint64_t	spa_lowmem_last_txg;	/* txg window start */
393 
394 	hrtime_t	spa_ccw_fail_time;	/* Conf cache write fail time */
395 
396 	uint64_t	spa_multihost;		/* multihost aware (mmp) */
397 	mmp_thread_t	spa_mmp;		/* multihost mmp thread */
398 	list_t		spa_leaf_list;		/* list of leaf vdevs */
399 	uint64_t	spa_leaf_list_gen;	/* track leaf_list changes */
400 
401 	/*
402 	 * spa_refcount & spa_config_lock must be the last elements
403 	 * because refcount_t changes size based on compilation options.
404 	 * because zfs_refcount_t changes size based on compilation options.
405 	 * In order for the MDB module to function correctly, the other
406 	 * fields must remain in the same location.
407 	 */
408 	spa_config_lock_t spa_config_lock[SCL_LOCKS]; /* config changes */
409 	zfs_refcount_t	spa_refcount;		/* number of opens */
410 };
411 
412 extern const char *spa_config_path;
413 
414 extern void spa_taskq_dispatch_ent(spa_t *spa, zio_type_t t, zio_taskq_type_t q,
415     task_func_t *func, void *arg, uint_t flags, taskq_ent_t *ent);
416 extern void spa_load_spares(spa_t *spa);
417 extern void spa_load_l2cache(spa_t *spa);
418 
419 #ifdef	__cplusplus
420 }
421 #endif
422 
423 #endif	/* _SYS_SPA_IMPL_H */
424