]> git.proxmox.com Git - mirror_zfs.git/blob - include/sys/vdev_impl.h
Remove unused vdev_t fields
[mirror_zfs.git] / include / sys / vdev_impl.h
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) 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_removal.h>
42 #include <sys/zfs_ratelimit.h>
43
44 #ifdef __cplusplus
45 extern "C" {
46 #endif
47
48 /*
49 * Virtual device descriptors.
50 *
51 * All storage pool operations go through the virtual device framework,
52 * which provides data replication and I/O scheduling.
53 */
54
55 /*
56 * Forward declarations that lots of things need.
57 */
58 typedef struct vdev_queue vdev_queue_t;
59 typedef struct vdev_cache vdev_cache_t;
60 typedef struct vdev_cache_entry vdev_cache_entry_t;
61 struct abd;
62
63 extern int zfs_vdev_queue_depth_pct;
64 extern int zfs_vdev_def_queue_depth;
65 extern uint32_t zfs_vdev_async_write_max_active;
66
67 /*
68 * Virtual device operations
69 */
70 typedef int vdev_open_func_t(vdev_t *vd, uint64_t *size, uint64_t *max_size,
71 uint64_t *ashift);
72 typedef void vdev_close_func_t(vdev_t *vd);
73 typedef uint64_t vdev_asize_func_t(vdev_t *vd, uint64_t psize);
74 typedef void vdev_io_start_func_t(zio_t *zio);
75 typedef void vdev_io_done_func_t(zio_t *zio);
76 typedef void vdev_state_change_func_t(vdev_t *vd, int, int);
77 typedef boolean_t vdev_need_resilver_func_t(vdev_t *vd, uint64_t, size_t);
78 typedef void vdev_hold_func_t(vdev_t *vd);
79 typedef void vdev_rele_func_t(vdev_t *vd);
80
81 typedef void vdev_remap_cb_t(uint64_t inner_offset, vdev_t *vd,
82 uint64_t offset, uint64_t size, void *arg);
83 typedef void vdev_remap_func_t(vdev_t *vd, uint64_t offset, uint64_t size,
84 vdev_remap_cb_t callback, void *arg);
85 /*
86 * Given a target vdev, translates the logical range "in" to the physical
87 * range "res"
88 */
89 typedef void vdev_xlation_func_t(vdev_t *cvd, const range_seg_t *in,
90 range_seg_t *res);
91
92 typedef const struct vdev_ops {
93 vdev_open_func_t *vdev_op_open;
94 vdev_close_func_t *vdev_op_close;
95 vdev_asize_func_t *vdev_op_asize;
96 vdev_io_start_func_t *vdev_op_io_start;
97 vdev_io_done_func_t *vdev_op_io_done;
98 vdev_state_change_func_t *vdev_op_state_change;
99 vdev_need_resilver_func_t *vdev_op_need_resilver;
100 vdev_hold_func_t *vdev_op_hold;
101 vdev_rele_func_t *vdev_op_rele;
102 vdev_remap_func_t *vdev_op_remap;
103 /*
104 * For translating ranges from non-leaf vdevs (e.g. raidz) to leaves.
105 * Used when initializing vdevs. Isn't used by leaf ops.
106 */
107 vdev_xlation_func_t *vdev_op_xlate;
108 char vdev_op_type[16];
109 boolean_t vdev_op_leaf;
110 } vdev_ops_t;
111
112 /*
113 * Virtual device properties
114 */
115 struct vdev_cache_entry {
116 struct abd *ve_abd;
117 uint64_t ve_offset;
118 clock_t ve_lastused;
119 avl_node_t ve_offset_node;
120 avl_node_t ve_lastused_node;
121 uint32_t ve_hits;
122 uint16_t ve_missed_update;
123 zio_t *ve_fill_io;
124 };
125
126 struct vdev_cache {
127 avl_tree_t vc_offset_tree;
128 avl_tree_t vc_lastused_tree;
129 kmutex_t vc_lock;
130 };
131
132 typedef struct vdev_queue_class {
133 uint32_t vqc_active;
134
135 /*
136 * Sorted by offset or timestamp, depending on if the queue is
137 * LBA-ordered vs FIFO.
138 */
139 avl_tree_t vqc_queued_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_active_tree;
146 avl_tree_t vq_read_offset_tree;
147 avl_tree_t vq_write_offset_tree;
148 uint64_t vq_last_offset;
149 hrtime_t vq_io_complete_ts; /* time last i/o completed */
150 hrtime_t vq_io_delta_ts;
151 zio_t vq_io_search; /* used as local for stack reduction */
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 vdev_stat_ex_t vdev_stat_ex; /* extended statistics */
230 boolean_t vdev_expanding; /* expand the vdev? */
231 boolean_t vdev_reopening; /* reopen in progress? */
232 boolean_t vdev_nonrot; /* true if solid state */
233 int vdev_open_error; /* error on last open */
234 kthread_t *vdev_open_thread; /* thread opening children */
235 uint64_t vdev_crtxg; /* txg when top-level was added */
236
237 /*
238 * Top-level vdev state.
239 */
240 uint64_t vdev_ms_array; /* metaslab array object */
241 uint64_t vdev_ms_shift; /* metaslab size shift */
242 uint64_t vdev_ms_count; /* number of metaslabs */
243 metaslab_group_t *vdev_mg; /* metaslab group */
244 metaslab_t **vdev_ms; /* metaslab array */
245 uint64_t vdev_pending_fastwrite; /* allocated fastwrites */
246 txg_list_t vdev_ms_list; /* per-txg dirty metaslab lists */
247 txg_list_t vdev_dtl_list; /* per-txg dirty DTL lists */
248 txg_node_t vdev_txg_node; /* per-txg dirty vdev linkage */
249 boolean_t vdev_remove_wanted; /* async remove wanted? */
250 boolean_t vdev_probe_wanted; /* async probe wanted? */
251 list_node_t vdev_config_dirty_node; /* config dirty list */
252 list_node_t vdev_state_dirty_node; /* state dirty list */
253 uint64_t vdev_deflate_ratio; /* deflation ratio (x512) */
254 uint64_t vdev_islog; /* is an intent log device */
255 uint64_t vdev_removing; /* device is being removed? */
256 boolean_t vdev_ishole; /* is a hole in the namespace */
257 uint64_t vdev_top_zap;
258 vdev_alloc_bias_t vdev_alloc_bias; /* metaslab allocation bias */
259
260 /* pool checkpoint related */
261 space_map_t *vdev_checkpoint_sm; /* contains reserved blocks */
262
263 boolean_t vdev_initialize_exit_wanted;
264 vdev_initializing_state_t vdev_initialize_state;
265 list_node_t vdev_initialize_node;
266 kthread_t *vdev_initialize_thread;
267 /* Protects vdev_initialize_thread and vdev_initialize_state. */
268 kmutex_t vdev_initialize_lock;
269 kcondvar_t vdev_initialize_cv;
270 uint64_t vdev_initialize_offset[TXG_SIZE];
271 uint64_t vdev_initialize_last_offset;
272 range_tree_t *vdev_initialize_tree; /* valid while initializing */
273 uint64_t vdev_initialize_bytes_est;
274 uint64_t vdev_initialize_bytes_done;
275 time_t vdev_initialize_action_time; /* start and end time */
276
277 /* for limiting outstanding I/Os */
278 kmutex_t vdev_initialize_io_lock;
279 kcondvar_t vdev_initialize_io_cv;
280 uint64_t vdev_initialize_inflight;
281
282 /*
283 * Values stored in the config for an indirect or removing vdev.
284 */
285 vdev_indirect_config_t vdev_indirect_config;
286
287 /*
288 * The vdev_indirect_rwlock protects the vdev_indirect_mapping
289 * pointer from changing on indirect vdevs (when it is condensed).
290 * Note that removing (not yet indirect) vdevs have different
291 * access patterns (the mapping is not accessed from open context,
292 * e.g. from zio_read) and locking strategy (e.g. svr_lock).
293 */
294 krwlock_t vdev_indirect_rwlock;
295 vdev_indirect_mapping_t *vdev_indirect_mapping;
296 vdev_indirect_births_t *vdev_indirect_births;
297
298 /*
299 * In memory data structures used to manage the obsolete sm, for
300 * indirect or removing vdevs.
301 *
302 * The vdev_obsolete_segments is the in-core record of the segments
303 * that are no longer referenced anywhere in the pool (due to
304 * being freed or remapped and not referenced by any snapshots).
305 * During a sync, segments are added to vdev_obsolete_segments
306 * via vdev_indirect_mark_obsolete(); at the end of each sync
307 * pass, this is appended to vdev_obsolete_sm via
308 * vdev_indirect_sync_obsolete(). The vdev_obsolete_lock
309 * protects against concurrent modifications of vdev_obsolete_segments
310 * from multiple zio threads.
311 */
312 kmutex_t vdev_obsolete_lock;
313 range_tree_t *vdev_obsolete_segments;
314 space_map_t *vdev_obsolete_sm;
315
316 /*
317 * Protects the vdev_scan_io_queue field itself as well as the
318 * structure's contents (when present).
319 */
320 kmutex_t vdev_scan_io_queue_lock;
321 struct dsl_scan_io_queue *vdev_scan_io_queue;
322
323 /*
324 * Leaf vdev state.
325 */
326 range_tree_t *vdev_dtl[DTL_TYPES]; /* dirty time logs */
327 space_map_t *vdev_dtl_sm; /* dirty time log space map */
328 txg_node_t vdev_dtl_node; /* per-txg dirty DTL linkage */
329 uint64_t vdev_dtl_object; /* DTL object */
330 uint64_t vdev_psize; /* physical device capacity */
331 uint64_t vdev_wholedisk; /* true if this is a whole disk */
332 uint64_t vdev_offline; /* persistent offline state */
333 uint64_t vdev_faulted; /* persistent faulted state */
334 uint64_t vdev_degraded; /* persistent degraded state */
335 uint64_t vdev_removed; /* persistent removed state */
336 uint64_t vdev_resilver_txg; /* persistent resilvering state */
337 uint64_t vdev_nparity; /* number of parity devices for raidz */
338 char *vdev_path; /* vdev path (if any) */
339 char *vdev_devid; /* vdev devid (if any) */
340 char *vdev_physpath; /* vdev device path (if any) */
341 char *vdev_enc_sysfs_path; /* enclosure sysfs path */
342 char *vdev_fru; /* physical FRU location */
343 uint64_t vdev_not_present; /* not present during import */
344 uint64_t vdev_unspare; /* unspare when resilvering done */
345 boolean_t vdev_nowritecache; /* true if flushwritecache failed */
346 boolean_t vdev_checkremove; /* temporary online test */
347 boolean_t vdev_forcefault; /* force online fault */
348 boolean_t vdev_splitting; /* split or repair in progress */
349 boolean_t vdev_delayed_close; /* delayed device close? */
350 boolean_t vdev_tmpoffline; /* device taken offline temporarily? */
351 boolean_t vdev_detached; /* device detached? */
352 boolean_t vdev_cant_read; /* vdev is failing all reads */
353 boolean_t vdev_cant_write; /* vdev is failing all writes */
354 boolean_t vdev_isspare; /* was a hot spare */
355 boolean_t vdev_isl2cache; /* was a l2cache device */
356 boolean_t vdev_copy_uberblocks; /* post expand copy uberblocks */
357 boolean_t vdev_resilver_deferred; /* resilver deferred */
358 vdev_queue_t vdev_queue; /* I/O deadline schedule queue */
359 vdev_cache_t vdev_cache; /* physical block cache */
360 spa_aux_vdev_t *vdev_aux; /* for l2cache and spares vdevs */
361 zio_t *vdev_probe_zio; /* root of current probe */
362 vdev_aux_t vdev_label_aux; /* on-disk aux state */
363 uint64_t vdev_leaf_zap;
364 hrtime_t vdev_mmp_pending; /* 0 if write finished */
365 uint64_t vdev_mmp_kstat_id; /* to find kstat entry */
366 uint64_t vdev_expansion_time; /* vdev's last expansion time */
367
368 /*
369 * For DTrace to work in userland (libzpool) context, these fields must
370 * remain at the end of the structure. DTrace will use the kernel's
371 * CTF definition for 'struct vdev', and since the size of a kmutex_t is
372 * larger in userland, the offsets for the rest of the fields would be
373 * incorrect.
374 */
375 kmutex_t vdev_dtl_lock; /* vdev_dtl_{map,resilver} */
376 kmutex_t vdev_stat_lock; /* vdev_stat */
377 kmutex_t vdev_probe_lock; /* protects vdev_probe_zio */
378
379 /*
380 * We rate limit ZIO delay and ZIO checksum events, since they
381 * can flood ZED with tons of events when a drive is acting up.
382 */
383 zfs_ratelimit_t vdev_delay_rl;
384 zfs_ratelimit_t vdev_checksum_rl;
385 };
386
387 #define VDEV_RAIDZ_MAXPARITY 3
388
389 #define VDEV_PAD_SIZE (8 << 10)
390 /* 2 padding areas (vl_pad1 and vl_pad2) to skip */
391 #define VDEV_SKIP_SIZE VDEV_PAD_SIZE * 2
392 #define VDEV_PHYS_SIZE (112 << 10)
393 #define VDEV_UBERBLOCK_RING (128 << 10)
394
395 /*
396 * MMP blocks occupy the last MMP_BLOCKS_PER_LABEL slots in the uberblock
397 * ring when MMP is enabled.
398 */
399 #define MMP_BLOCKS_PER_LABEL 1
400
401 /* The largest uberblock we support is 8k. */
402 #define MAX_UBERBLOCK_SHIFT (13)
403 #define VDEV_UBERBLOCK_SHIFT(vd) \
404 MIN(MAX((vd)->vdev_top->vdev_ashift, UBERBLOCK_SHIFT), \
405 MAX_UBERBLOCK_SHIFT)
406 #define VDEV_UBERBLOCK_COUNT(vd) \
407 (VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd))
408 #define VDEV_UBERBLOCK_OFFSET(vd, n) \
409 offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)])
410 #define VDEV_UBERBLOCK_SIZE(vd) (1ULL << VDEV_UBERBLOCK_SHIFT(vd))
411
412 typedef struct vdev_phys {
413 char vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)];
414 zio_eck_t vp_zbt;
415 } vdev_phys_t;
416
417 typedef struct vdev_label {
418 char vl_pad1[VDEV_PAD_SIZE]; /* 8K */
419 char vl_pad2[VDEV_PAD_SIZE]; /* 8K */
420 vdev_phys_t vl_vdev_phys; /* 112K */
421 char vl_uberblock[VDEV_UBERBLOCK_RING]; /* 128K */
422 } vdev_label_t; /* 256K total */
423
424 /*
425 * vdev_dirty() flags
426 */
427 #define VDD_METASLAB 0x01
428 #define VDD_DTL 0x02
429
430 /* Offset of embedded boot loader region on each label */
431 #define VDEV_BOOT_OFFSET (2 * sizeof (vdev_label_t))
432 /*
433 * Size of embedded boot loader region on each label.
434 * The total size of the first two labels plus the boot area is 4MB.
435 */
436 #define VDEV_BOOT_SIZE (7ULL << 19) /* 3.5M */
437
438 /*
439 * Size of label regions at the start and end of each leaf device.
440 */
441 #define VDEV_LABEL_START_SIZE (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE)
442 #define VDEV_LABEL_END_SIZE (2 * sizeof (vdev_label_t))
443 #define VDEV_LABELS 4
444 #define VDEV_BEST_LABEL VDEV_LABELS
445
446 #define VDEV_ALLOC_LOAD 0
447 #define VDEV_ALLOC_ADD 1
448 #define VDEV_ALLOC_SPARE 2
449 #define VDEV_ALLOC_L2CACHE 3
450 #define VDEV_ALLOC_ROOTPOOL 4
451 #define VDEV_ALLOC_SPLIT 5
452 #define VDEV_ALLOC_ATTACH 6
453
454 /*
455 * Allocate or free a vdev
456 */
457 extern vdev_t *vdev_alloc_common(spa_t *spa, uint_t id, uint64_t guid,
458 vdev_ops_t *ops);
459 extern int vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *config,
460 vdev_t *parent, uint_t id, int alloctype);
461 extern void vdev_free(vdev_t *vd);
462
463 /*
464 * Add or remove children and parents
465 */
466 extern void vdev_add_child(vdev_t *pvd, vdev_t *cvd);
467 extern void vdev_remove_child(vdev_t *pvd, vdev_t *cvd);
468 extern void vdev_compact_children(vdev_t *pvd);
469 extern vdev_t *vdev_add_parent(vdev_t *cvd, vdev_ops_t *ops);
470 extern void vdev_remove_parent(vdev_t *cvd);
471
472 /*
473 * vdev sync load and sync
474 */
475 extern boolean_t vdev_log_state_valid(vdev_t *vd);
476 extern int vdev_load(vdev_t *vd);
477 extern int vdev_dtl_load(vdev_t *vd);
478 extern void vdev_sync(vdev_t *vd, uint64_t txg);
479 extern void vdev_sync_done(vdev_t *vd, uint64_t txg);
480 extern void vdev_dirty(vdev_t *vd, int flags, void *arg, uint64_t txg);
481 extern void vdev_dirty_leaves(vdev_t *vd, int flags, uint64_t txg);
482
483 /*
484 * Available vdev types.
485 */
486 extern vdev_ops_t vdev_root_ops;
487 extern vdev_ops_t vdev_mirror_ops;
488 extern vdev_ops_t vdev_replacing_ops;
489 extern vdev_ops_t vdev_raidz_ops;
490 extern vdev_ops_t vdev_disk_ops;
491 extern vdev_ops_t vdev_file_ops;
492 extern vdev_ops_t vdev_missing_ops;
493 extern vdev_ops_t vdev_hole_ops;
494 extern vdev_ops_t vdev_spare_ops;
495 extern vdev_ops_t vdev_indirect_ops;
496
497 /*
498 * Common size functions
499 */
500 extern void vdev_default_xlate(vdev_t *vd, const range_seg_t *in,
501 range_seg_t *out);
502 extern uint64_t vdev_default_asize(vdev_t *vd, uint64_t psize);
503 extern uint64_t vdev_get_min_asize(vdev_t *vd);
504 extern void vdev_set_min_asize(vdev_t *vd);
505
506 /*
507 * Global variables
508 */
509 extern int vdev_standard_sm_blksz;
510 /* zdb uses this tunable, so it must be declared here to make lint happy. */
511 extern int zfs_vdev_cache_size;
512
513 /*
514 * Functions from vdev_indirect.c
515 */
516 extern void vdev_indirect_sync_obsolete(vdev_t *vd, dmu_tx_t *tx);
517 extern boolean_t vdev_indirect_should_condense(vdev_t *vd);
518 extern void spa_condense_indirect_start_sync(vdev_t *vd, dmu_tx_t *tx);
519 extern int vdev_obsolete_sm_object(vdev_t *vd, uint64_t *sm_obj);
520 extern int vdev_obsolete_counts_are_precise(vdev_t *vd, boolean_t *are_precise);
521
522 /*
523 * Other miscellaneous functions
524 */
525 int vdev_checkpoint_sm_object(vdev_t *vd, uint64_t *sm_obj);
526
527 #ifdef __cplusplus
528 }
529 #endif
530
531 #endif /* _SYS_VDEV_IMPL_H */