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.
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.
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]
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
25 * Copyright (c) 2012, 2017 by Delphix. All rights reserved.
26 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
27 * Copyright 2016 Toomas Soome <tsoome@me.com>
33 #include <sys/zio_priority.h>
34 #include <sys/zfs_context.h>
38 #include <sys/kstat.h>
39 #include <sys/fs/zfs.h>
40 #include <sys/zio_impl.h>
49 #define ZEC_MAGIC 0x210da7ab10c7a11ULL
51 typedef struct zio_eck {
52 uint64_t zec_magic; /* for validation, endianness */
53 zio_cksum_t zec_cksum; /* 256-bit checksum */
57 * Gang block headers are self-checksumming and contain an array
60 #define SPA_GANGBLOCKSIZE SPA_MINBLOCKSIZE
61 #define SPA_GBH_NBLKPTRS ((SPA_GANGBLOCKSIZE - \
62 sizeof (zio_eck_t)) / sizeof (blkptr_t))
63 #define SPA_GBH_FILLER ((SPA_GANGBLOCKSIZE - \
64 sizeof (zio_eck_t) - \
65 (SPA_GBH_NBLKPTRS * sizeof (blkptr_t))) /\
68 typedef struct zio_gbh {
69 blkptr_t zg_blkptr[SPA_GBH_NBLKPTRS];
70 uint64_t zg_filler[SPA_GBH_FILLER];
75 ZIO_CHECKSUM_INHERIT = 0,
79 ZIO_CHECKSUM_GANG_HEADER,
81 ZIO_CHECKSUM_FLETCHER_2,
82 ZIO_CHECKSUM_FLETCHER_4,
85 ZIO_CHECKSUM_NOPARITY,
91 ZIO_CHECKSUM_FUNCTIONS
95 * The number of "legacy" compression functions which can be set on individual
98 #define ZIO_CHECKSUM_LEGACY_FUNCTIONS ZIO_CHECKSUM_ZILOG2
100 #define ZIO_CHECKSUM_ON_VALUE ZIO_CHECKSUM_FLETCHER_4
101 #define ZIO_CHECKSUM_DEFAULT ZIO_CHECKSUM_ON
103 #define ZIO_CHECKSUM_MASK 0xffULL
104 #define ZIO_CHECKSUM_VERIFY (1 << 8)
106 #define ZIO_DEDUPCHECKSUM ZIO_CHECKSUM_SHA256
107 #define ZIO_DEDUPDITTO_MIN 100
110 * The number of "legacy" compression functions which can be set on individual
113 #define ZIO_COMPRESS_LEGACY_FUNCTIONS ZIO_COMPRESS_LZ4
116 * The meaning of "compress = on" selected by the compression features enabled
119 #define ZIO_COMPRESS_LEGACY_ON_VALUE ZIO_COMPRESS_LZJB
120 #define ZIO_COMPRESS_LZ4_ON_VALUE ZIO_COMPRESS_LZ4
122 #define ZIO_COMPRESS_DEFAULT ZIO_COMPRESS_OFF
124 #define BOOTFS_COMPRESS_VALID(compress) \
125 ((compress) == ZIO_COMPRESS_LZJB || \
126 (compress) == ZIO_COMPRESS_LZ4 || \
127 (compress) == ZIO_COMPRESS_ON || \
128 (compress) == ZIO_COMPRESS_OFF)
130 #define ZIO_FAILURE_MODE_WAIT 0
131 #define ZIO_FAILURE_MODE_CONTINUE 1
132 #define ZIO_FAILURE_MODE_PANIC 2
136 * Flags inherited by gang, ddt, and vdev children,
137 * and that must be equal for two zios to aggregate
139 ZIO_FLAG_DONT_AGGREGATE = 1 << 0,
140 ZIO_FLAG_IO_REPAIR = 1 << 1,
141 ZIO_FLAG_SELF_HEAL = 1 << 2,
142 ZIO_FLAG_RESILVER = 1 << 3,
143 ZIO_FLAG_SCRUB = 1 << 4,
144 ZIO_FLAG_SCAN_THREAD = 1 << 5,
145 ZIO_FLAG_PHYSICAL = 1 << 6,
147 #define ZIO_FLAG_AGG_INHERIT (ZIO_FLAG_CANFAIL - 1)
150 * Flags inherited by ddt, gang, and vdev children.
152 ZIO_FLAG_CANFAIL = 1 << 7, /* must be first for INHERIT */
153 ZIO_FLAG_SPECULATIVE = 1 << 8,
154 ZIO_FLAG_CONFIG_WRITER = 1 << 9,
155 ZIO_FLAG_DONT_RETRY = 1 << 10,
156 ZIO_FLAG_DONT_CACHE = 1 << 11,
157 ZIO_FLAG_NODATA = 1 << 12,
158 ZIO_FLAG_INDUCE_DAMAGE = 1 << 13,
159 ZIO_FLAG_IO_ALLOCATING = 1 << 14,
161 #define ZIO_FLAG_DDT_INHERIT (ZIO_FLAG_IO_RETRY - 1)
162 #define ZIO_FLAG_GANG_INHERIT (ZIO_FLAG_IO_RETRY - 1)
165 * Flags inherited by vdev children.
167 ZIO_FLAG_IO_RETRY = 1 << 15, /* must be first for INHERIT */
168 ZIO_FLAG_PROBE = 1 << 16,
169 ZIO_FLAG_TRYHARD = 1 << 17,
170 ZIO_FLAG_OPTIONAL = 1 << 18,
172 #define ZIO_FLAG_VDEV_INHERIT (ZIO_FLAG_DONT_QUEUE - 1)
175 * Flags not inherited by any children.
177 ZIO_FLAG_DONT_QUEUE = 1 << 19, /* must be first for INHERIT */
178 ZIO_FLAG_DONT_PROPAGATE = 1 << 20,
179 ZIO_FLAG_IO_BYPASS = 1 << 21,
180 ZIO_FLAG_IO_REWRITE = 1 << 22,
181 ZIO_FLAG_RAW = 1 << 23,
182 ZIO_FLAG_GANG_CHILD = 1 << 24,
183 ZIO_FLAG_DDT_CHILD = 1 << 25,
184 ZIO_FLAG_GODFATHER = 1 << 26,
185 ZIO_FLAG_NOPWRITE = 1 << 27,
186 ZIO_FLAG_REEXECUTED = 1 << 28,
187 ZIO_FLAG_DELEGATED = 1 << 29,
190 #define ZIO_FLAG_MUSTSUCCEED 0
192 #define ZIO_DDT_CHILD_FLAGS(zio) \
193 (((zio)->io_flags & ZIO_FLAG_DDT_INHERIT) | \
194 ZIO_FLAG_DDT_CHILD | ZIO_FLAG_CANFAIL)
196 #define ZIO_GANG_CHILD_FLAGS(zio) \
197 (((zio)->io_flags & ZIO_FLAG_GANG_INHERIT) | \
198 ZIO_FLAG_GANG_CHILD | ZIO_FLAG_CANFAIL)
200 #define ZIO_VDEV_CHILD_FLAGS(zio) \
201 (((zio)->io_flags & ZIO_FLAG_VDEV_INHERIT) | \
219 * We'll take the number 122 and 123 to indicate checksum errors and
220 * fragmentation. Those doesn't collide with any errno values as they
221 * are greater than ELAST.
226 typedef void zio_done_func_t(zio_t *zio);
228 extern boolean_t zio_dva_throttle_enabled;
229 extern const char *zio_type_name[ZIO_TYPES];
232 * A bookmark is a four-tuple <objset, object, level, blkid> that uniquely
233 * identifies any block in the pool. By convention, the meta-objset (MOS)
234 * is objset 0, and the meta-dnode is object 0. This covers all blocks
235 * except root blocks and ZIL blocks, which are defined as follows:
237 * Root blocks (objset_phys_t) are object 0, level -1: <objset, 0, -1, 0>.
238 * ZIL blocks are bookmarked <objset, 0, -2, blkid == ZIL sequence number>.
239 * dmu_sync()ed ZIL data blocks are bookmarked <objset, object, -2, blkid>.
240 * dnode visit bookmarks are <objset, object id of dnode, -3, 0>.
242 * Note: this structure is called a bookmark because its original purpose
243 * was to remember where to resume a pool-wide traverse.
245 * Note: this structure is passed between userland and the kernel, and is
246 * stored on disk (by virtue of being incorporated into other on-disk
247 * structures, e.g. dsl_scan_phys_t).
249 typedef struct zbookmark_phys {
256 #define SET_BOOKMARK(zb, objset, object, level, blkid) \
258 (zb)->zb_objset = objset; \
259 (zb)->zb_object = object; \
260 (zb)->zb_level = level; \
261 (zb)->zb_blkid = blkid; \
264 #define ZB_DESTROYED_OBJSET (-1ULL)
266 #define ZB_ROOT_OBJECT (0ULL)
267 #define ZB_ROOT_LEVEL (-1LL)
268 #define ZB_ROOT_BLKID (0ULL)
270 #define ZB_ZIL_OBJECT (0ULL)
271 #define ZB_ZIL_LEVEL (-2LL)
273 #define ZB_DNODE_LEVEL (-3LL)
274 #define ZB_DNODE_BLKID (0ULL)
276 #define ZB_IS_ZERO(zb) \
277 ((zb)->zb_objset == 0 && (zb)->zb_object == 0 && \
278 (zb)->zb_level == 0 && (zb)->zb_blkid == 0)
279 #define ZB_IS_ROOT(zb) \
280 ((zb)->zb_object == ZB_ROOT_OBJECT && \
281 (zb)->zb_level == ZB_ROOT_LEVEL && \
282 (zb)->zb_blkid == ZB_ROOT_BLKID)
284 typedef struct zio_prop {
285 enum zio_checksum zp_checksum;
286 enum zio_compress zp_compress;
287 dmu_object_type_t zp_type;
291 boolean_t zp_dedup_verify;
292 boolean_t zp_nopwrite;
295 typedef struct zio_cksum_report zio_cksum_report_t;
297 typedef void zio_cksum_finish_f(zio_cksum_report_t *rep,
298 const void *good_data);
299 typedef void zio_cksum_free_f(void *cbdata, size_t size);
301 struct zio_bad_cksum; /* defined in zio_checksum.h */
305 struct zio_cksum_report {
306 struct zio_cksum_report *zcr_next;
307 nvlist_t *zcr_ereport;
308 nvlist_t *zcr_detector;
310 size_t zcr_cbinfo; /* passed to zcr_free() */
313 zio_cksum_finish_f *zcr_finish;
314 zio_cksum_free_f *zcr_free;
316 /* internal use only */
317 struct zio_bad_cksum *zcr_ckinfo; /* information from failure */
320 typedef void zio_vsd_cksum_report_f(zio_t *zio, zio_cksum_report_t *zcr,
323 zio_vsd_cksum_report_f zio_vsd_default_cksum_report;
325 typedef struct zio_vsd_ops {
326 zio_done_func_t *vsd_free;
327 zio_vsd_cksum_report_f *vsd_cksum_report;
330 typedef struct zio_gang_node {
331 zio_gbh_phys_t *gn_gbh;
332 struct zio_gang_node *gn_child[SPA_GBH_NBLKPTRS];
335 typedef zio_t *zio_gang_issue_func_t(zio_t *zio, blkptr_t *bp,
336 zio_gang_node_t *gn, struct abd *data, uint64_t offset);
338 typedef void zio_transform_func_t(zio_t *zio, struct abd *data, uint64_t size);
340 typedef struct zio_transform {
341 struct abd *zt_orig_abd;
342 uint64_t zt_orig_size;
344 zio_transform_func_t *zt_transform;
345 struct zio_transform *zt_next;
348 typedef int zio_pipe_stage_t(zio_t *zio);
351 * The io_reexecute flags are distinct from io_flags because the child must
352 * be able to propagate them to the parent. The normal io_flags are local
353 * to the zio, not protected by any lock, and not modifiable by children;
354 * the reexecute flags are protected by io_lock, modifiable by children,
355 * and always propagated -- even when ZIO_FLAG_DONT_PROPAGATE is set.
357 #define ZIO_REEXECUTE_NOW 0x01
358 #define ZIO_REEXECUTE_SUSPEND 0x02
360 typedef struct zio_alloc_list {
365 typedef struct zio_link {
368 list_node_t zl_parent_node;
369 list_node_t zl_child_node;
373 * Used for TRIM kstat.
375 typedef struct zio_trim_stats {
377 * Number of bytes successfully TRIMmed.
382 * Number of successful TRIM requests.
384 kstat_named_t success;
387 * Number of TRIM requests that failed because TRIM is not
390 kstat_named_t unsupported;
393 * Number of TRIM requests that failed for other reasons.
395 kstat_named_t failed;
398 extern zio_trim_stats_t zio_trim_stats;
400 #define ZIO_TRIM_STAT_INCR(stat, val) \
401 atomic_add_64(&zio_trim_stats.stat.value.ui64, (val));
402 #define ZIO_TRIM_STAT_BUMP(stat) \
403 ZIO_TRIM_STAT_INCR(stat, 1);
406 /* Core information about this I/O */
407 zbookmark_phys_t io_bookmark;
410 enum zio_child io_child_type;
412 zio_priority_t io_priority;
413 uint8_t io_reexecute;
414 uint8_t io_state[ZIO_WAIT_TYPES];
418 blkptr_t *io_bp_override;
420 list_t io_parent_list;
421 list_t io_child_list;
423 zio_transform_t *io_transform_stack;
426 zio_done_func_t *io_ready;
427 zio_done_func_t *io_children_ready;
428 zio_done_func_t *io_physdone;
429 zio_done_func_t *io_done;
431 int64_t io_prev_space_delta; /* DMU private */
434 /* Data represented by this I/O */
436 struct abd *io_orig_abd;
438 uint64_t io_orig_size;
439 /* io_lsize != io_orig_size iff this is a raw write */
442 /* Stuff for the vdev stack */
445 const zio_vsd_ops_t *io_vsd_ops;
448 hrtime_t io_timestamp;
449 hrtime_t io_queued_timestamp;
450 hrtime_t io_target_timestamp;
451 avl_node_t io_queue_node;
452 avl_node_t io_offset_node;
453 avl_node_t io_alloc_node;
454 zio_alloc_list_t io_alloc_list;
460 /* Internal pipeline state */
461 enum zio_flag io_flags;
462 enum zio_stage io_stage;
463 enum zio_stage io_pipeline;
464 enum zio_flag io_orig_flags;
465 enum zio_stage io_orig_stage;
466 enum zio_stage io_orig_pipeline;
467 enum zio_stage io_pipeline_trace;
469 int io_child_error[ZIO_CHILD_TYPES];
470 uint64_t io_children[ZIO_CHILD_TYPES][ZIO_WAIT_TYPES];
471 uint64_t io_child_count;
472 uint64_t io_phys_children;
473 uint64_t io_parent_count;
475 zio_t *io_gang_leader;
476 zio_gang_node_t *io_gang_tree;
483 zio_cksum_report_t *io_cksum_report;
486 /* Taskq dispatching state */
487 taskq_ent_t io_tqent;
489 avl_node_t io_trim_node;
490 list_node_t io_trim_link;
493 extern int zio_bookmark_compare(const void *, const void *);
495 extern zio_t *zio_null(zio_t *pio, spa_t *spa, vdev_t *vd,
496 zio_done_func_t *done, void *priv, enum zio_flag flags);
498 extern zio_t *zio_root(spa_t *spa,
499 zio_done_func_t *done, void *priv, enum zio_flag flags);
501 extern zio_t *zio_read(zio_t *pio, spa_t *spa, const blkptr_t *bp,
502 struct abd *data, uint64_t lsize, zio_done_func_t *done, void *priv,
503 zio_priority_t priority, enum zio_flag flags, const zbookmark_phys_t *zb);
505 extern zio_t *zio_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
506 struct abd *data, uint64_t size, uint64_t psize, const zio_prop_t *zp,
507 zio_done_func_t *ready, zio_done_func_t *children_ready,
508 zio_done_func_t *physdone, zio_done_func_t *done,
509 void *priv, zio_priority_t priority, enum zio_flag flags,
510 const zbookmark_phys_t *zb);
512 extern zio_t *zio_rewrite(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
513 struct abd *data, uint64_t size, zio_done_func_t *done, void *priv,
514 zio_priority_t priority, enum zio_flag flags, zbookmark_phys_t *zb);
516 extern void zio_write_override(zio_t *zio, blkptr_t *bp, int copies,
519 extern void zio_free(spa_t *spa, uint64_t txg, const blkptr_t *bp);
521 extern zio_t *zio_claim(zio_t *pio, spa_t *spa, uint64_t txg,
523 zio_done_func_t *done, void *priv, enum zio_flag flags);
525 extern zio_t *zio_ioctl(zio_t *pio, spa_t *spa, vdev_t *vd, int cmd,
526 uint64_t offset, uint64_t size, zio_done_func_t *done, void *priv,
527 zio_priority_t priority, enum zio_flag flags);
529 extern zio_t *zio_read_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
530 uint64_t size, struct abd *data, int checksum,
531 zio_done_func_t *done, void *priv, zio_priority_t priority,
532 enum zio_flag flags, boolean_t labels);
534 extern zio_t *zio_write_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
535 uint64_t size, struct abd *data, int checksum,
536 zio_done_func_t *done, void *priv, zio_priority_t priority,
537 enum zio_flag flags, boolean_t labels);
539 extern zio_t *zio_free_sync(zio_t *pio, spa_t *spa, uint64_t txg,
540 const blkptr_t *bp, uint64_t size, enum zio_flag flags);
542 extern int zio_alloc_zil(spa_t *spa, uint64_t txg, blkptr_t *new_bp,
543 blkptr_t *old_bp, uint64_t size, boolean_t *slog);
544 extern void zio_free_zil(spa_t *spa, uint64_t txg, blkptr_t *bp);
545 extern void zio_flush(zio_t *zio, vdev_t *vd);
546 extern zio_t *zio_trim(zio_t *zio, spa_t *spa, vdev_t *vd, uint64_t offset,
548 extern void zio_shrink(zio_t *zio, uint64_t size);
550 extern int zio_wait(zio_t *zio);
551 extern void zio_nowait(zio_t *zio);
552 extern void zio_execute(zio_t *zio);
553 extern void zio_interrupt(zio_t *zio);
554 extern void zio_delay_init(zio_t *zio);
555 extern void zio_delay_interrupt(zio_t *zio);
557 extern zio_t *zio_walk_parents(zio_t *cio, zio_link_t **);
558 extern zio_t *zio_walk_children(zio_t *pio, zio_link_t **);
559 extern zio_t *zio_unique_parent(zio_t *cio);
560 extern void zio_add_child(zio_t *pio, zio_t *cio);
562 extern void *zio_buf_alloc(size_t size);
563 extern void zio_buf_free(void *buf, size_t size);
564 extern void *zio_data_buf_alloc(size_t size);
565 extern void zio_data_buf_free(void *buf, size_t size);
567 extern void zio_push_transform(zio_t *zio, struct abd *abd, uint64_t size,
568 uint64_t bufsize, zio_transform_func_t *transform);
569 extern void zio_pop_transforms(zio_t *zio);
571 extern void zio_resubmit_stage_async(void *);
573 extern zio_t *zio_vdev_child_io(zio_t *zio, blkptr_t *bp, vdev_t *vd,
574 uint64_t offset, struct abd *data, uint64_t size, int type,
575 zio_priority_t priority, enum zio_flag flags,
576 zio_done_func_t *done, void *priv);
578 extern zio_t *zio_vdev_delegated_io(vdev_t *vd, uint64_t offset,
579 struct abd *data, uint64_t size, int type, zio_priority_t priority,
580 enum zio_flag flags, zio_done_func_t *done, void *priv);
582 extern void zio_vdev_io_bypass(zio_t *zio);
583 extern void zio_vdev_io_reissue(zio_t *zio);
584 extern void zio_vdev_io_redone(zio_t *zio);
586 extern void zio_checksum_verified(zio_t *zio);
587 extern int zio_worst_error(int e1, int e2);
589 extern enum zio_checksum zio_checksum_select(enum zio_checksum child,
590 enum zio_checksum parent);
591 extern enum zio_checksum zio_checksum_dedup_select(spa_t *spa,
592 enum zio_checksum child, enum zio_checksum parent);
593 extern enum zio_compress zio_compress_select(spa_t *spa,
594 enum zio_compress child, enum zio_compress parent);
596 extern void zio_suspend(spa_t *spa, zio_t *zio);
597 extern int zio_resume(spa_t *spa);
598 extern void zio_resume_wait(spa_t *spa);
601 * Initial setup and teardown.
603 extern void zio_init(void);
604 extern void zio_fini(void);
609 struct zinject_record;
610 extern uint32_t zio_injection_enabled;
611 extern int zio_inject_fault(char *name, int flags, int *id,
612 struct zinject_record *record);
613 extern int zio_inject_list_next(int *id, char *name, size_t buflen,
614 struct zinject_record *record);
615 extern int zio_clear_fault(int id);
616 extern void zio_handle_panic_injection(spa_t *spa, char *tag, uint64_t type);
617 extern int zio_handle_fault_injection(zio_t *zio, int error);
618 extern int zio_handle_device_injection(vdev_t *vd, zio_t *zio, int error);
619 extern int zio_handle_label_injection(zio_t *zio, int error);
620 extern void zio_handle_ignored_writes(zio_t *zio);
621 extern hrtime_t zio_handle_io_delay(zio_t *zio);
624 * Checksum ereport functions
626 extern void zfs_ereport_start_checksum(spa_t *spa, vdev_t *vd, struct zio *zio,
627 uint64_t offset, uint64_t length, void *arg, struct zio_bad_cksum *info);
628 extern void zfs_ereport_finish_checksum(zio_cksum_report_t *report,
629 const void *good_data, const void *bad_data, boolean_t drop_if_identical);
631 extern void zfs_ereport_send_interim_checksum(zio_cksum_report_t *report);
632 extern void zfs_ereport_free_checksum(zio_cksum_report_t *report);
634 /* If we have the good data in hand, this function can be used */
635 extern void zfs_ereport_post_checksum(spa_t *spa, vdev_t *vd,
636 struct zio *zio, uint64_t offset, uint64_t length,
637 const void *good_data, const void *bad_data, struct zio_bad_cksum *info);
639 /* Called from spa_sync(), but primarily an injection handler */
640 extern void spa_handle_ignored_writes(spa_t *spa);
642 /* zbookmark_phys functions */
643 boolean_t zbookmark_subtree_completed(const struct dnode_phys *dnp,
644 const zbookmark_phys_t *subtree_root, const zbookmark_phys_t *last_block);
645 int zbookmark_compare(uint16_t dbss1, uint8_t ibs1, uint16_t dbss2,
646 uint8_t ibs2, const zbookmark_phys_t *zb1, const zbookmark_phys_t *zb2);