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, 2014 by Delphix. All rights reserved.
26 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
32 #include <sys/zfs_context.h>
36 #include <sys/kstat.h>
37 #include <sys/fs/zfs.h>
38 #include <sys/zio_impl.h>
47 #define ZEC_MAGIC 0x210da7ab10c7a11ULL
49 typedef struct zio_eck {
50 uint64_t zec_magic; /* for validation, endianness */
51 zio_cksum_t zec_cksum; /* 256-bit checksum */
55 * Gang block headers are self-checksumming and contain an array
58 #define SPA_GANGBLOCKSIZE SPA_MINBLOCKSIZE
59 #define SPA_GBH_NBLKPTRS ((SPA_GANGBLOCKSIZE - \
60 sizeof (zio_eck_t)) / sizeof (blkptr_t))
61 #define SPA_GBH_FILLER ((SPA_GANGBLOCKSIZE - \
62 sizeof (zio_eck_t) - \
63 (SPA_GBH_NBLKPTRS * sizeof (blkptr_t))) /\
66 typedef struct zio_gbh {
67 blkptr_t zg_blkptr[SPA_GBH_NBLKPTRS];
68 uint64_t zg_filler[SPA_GBH_FILLER];
73 ZIO_CHECKSUM_INHERIT = 0,
77 ZIO_CHECKSUM_GANG_HEADER,
79 ZIO_CHECKSUM_FLETCHER_2,
80 ZIO_CHECKSUM_FLETCHER_4,
83 ZIO_CHECKSUM_NOPARITY,
84 ZIO_CHECKSUM_FUNCTIONS
88 * The number of "legacy" compression functions which can be set on individual
91 #define ZIO_CHECKSUM_LEGACY_FUNCTIONS ZIO_CHECKSUM_ZILOG2
93 #define ZIO_CHECKSUM_ON_VALUE ZIO_CHECKSUM_FLETCHER_4
94 #define ZIO_CHECKSUM_DEFAULT ZIO_CHECKSUM_ON
96 #define ZIO_CHECKSUM_MASK 0xffULL
97 #define ZIO_CHECKSUM_VERIFY (1 << 8)
99 #define ZIO_DEDUPCHECKSUM ZIO_CHECKSUM_SHA256
100 #define ZIO_DEDUPDITTO_MIN 100
103 ZIO_COMPRESS_INHERIT = 0,
119 ZIO_COMPRESS_FUNCTIONS
123 * The number of "legacy" compression functions which can be set on individual
126 #define ZIO_COMPRESS_LEGACY_FUNCTIONS ZIO_COMPRESS_LZ4
129 * The meaning of "compress = on" selected by the compression features enabled
132 #define ZIO_COMPRESS_LEGACY_ON_VALUE ZIO_COMPRESS_LZJB
133 #define ZIO_COMPRESS_LZ4_ON_VALUE ZIO_COMPRESS_LZ4
135 #define ZIO_COMPRESS_DEFAULT ZIO_COMPRESS_OFF
137 #define BOOTFS_COMPRESS_VALID(compress) \
138 ((compress) == ZIO_COMPRESS_LZJB || \
139 (compress) == ZIO_COMPRESS_LZ4 || \
140 (compress) == ZIO_COMPRESS_ON || \
141 (compress) == ZIO_COMPRESS_OFF)
143 #define ZIO_FAILURE_MODE_WAIT 0
144 #define ZIO_FAILURE_MODE_CONTINUE 1
145 #define ZIO_FAILURE_MODE_PANIC 2
147 typedef enum zio_priority {
148 ZIO_PRIORITY_SYNC_READ,
149 ZIO_PRIORITY_SYNC_WRITE, /* ZIL */
150 ZIO_PRIORITY_ASYNC_READ, /* prefetch */
151 ZIO_PRIORITY_ASYNC_WRITE, /* spa_sync() */
152 ZIO_PRIORITY_SCRUB, /* asynchronous scrub/resilver reads */
153 ZIO_PRIORITY_TRIM, /* free requests used for TRIM */
154 ZIO_PRIORITY_NUM_QUEUEABLE,
156 ZIO_PRIORITY_NOW /* non-queued I/Os (e.g. ioctl) */
159 #define ZIO_PIPELINE_CONTINUE 0x100
160 #define ZIO_PIPELINE_STOP 0x101
164 * Flags inherited by gang, ddt, and vdev children,
165 * and that must be equal for two zios to aggregate
167 ZIO_FLAG_DONT_AGGREGATE = 1 << 0,
168 ZIO_FLAG_IO_REPAIR = 1 << 1,
169 ZIO_FLAG_SELF_HEAL = 1 << 2,
170 ZIO_FLAG_RESILVER = 1 << 3,
171 ZIO_FLAG_SCRUB = 1 << 4,
172 ZIO_FLAG_SCAN_THREAD = 1 << 5,
173 ZIO_FLAG_PHYSICAL = 1 << 6,
175 #define ZIO_FLAG_AGG_INHERIT (ZIO_FLAG_CANFAIL - 1)
178 * Flags inherited by ddt, gang, and vdev children.
180 ZIO_FLAG_CANFAIL = 1 << 7, /* must be first for INHERIT */
181 ZIO_FLAG_SPECULATIVE = 1 << 8,
182 ZIO_FLAG_CONFIG_WRITER = 1 << 9,
183 ZIO_FLAG_DONT_RETRY = 1 << 10,
184 ZIO_FLAG_DONT_CACHE = 1 << 11,
185 ZIO_FLAG_NODATA = 1 << 12,
186 ZIO_FLAG_INDUCE_DAMAGE = 1 << 13,
188 #define ZIO_FLAG_DDT_INHERIT (ZIO_FLAG_IO_RETRY - 1)
189 #define ZIO_FLAG_GANG_INHERIT (ZIO_FLAG_IO_RETRY - 1)
192 * Flags inherited by vdev children.
194 ZIO_FLAG_IO_RETRY = 1 << 14, /* must be first for INHERIT */
195 ZIO_FLAG_PROBE = 1 << 15,
196 ZIO_FLAG_TRYHARD = 1 << 16,
197 ZIO_FLAG_OPTIONAL = 1 << 17,
199 #define ZIO_FLAG_VDEV_INHERIT (ZIO_FLAG_DONT_QUEUE - 1)
202 * Flags not inherited by any children.
204 ZIO_FLAG_DONT_QUEUE = 1 << 18, /* must be first for INHERIT */
205 ZIO_FLAG_DONT_PROPAGATE = 1 << 19,
206 ZIO_FLAG_IO_BYPASS = 1 << 20,
207 ZIO_FLAG_IO_REWRITE = 1 << 21,
208 ZIO_FLAG_RAW = 1 << 22,
209 ZIO_FLAG_GANG_CHILD = 1 << 23,
210 ZIO_FLAG_DDT_CHILD = 1 << 24,
211 ZIO_FLAG_GODFATHER = 1 << 25,
212 ZIO_FLAG_NOPWRITE = 1 << 26,
213 ZIO_FLAG_REEXECUTED = 1 << 27,
214 ZIO_FLAG_DELEGATED = 1 << 28,
217 #define ZIO_FLAG_MUSTSUCCEED 0
219 #define ZIO_DDT_CHILD_FLAGS(zio) \
220 (((zio)->io_flags & ZIO_FLAG_DDT_INHERIT) | \
221 ZIO_FLAG_DDT_CHILD | ZIO_FLAG_CANFAIL)
223 #define ZIO_GANG_CHILD_FLAGS(zio) \
224 (((zio)->io_flags & ZIO_FLAG_GANG_INHERIT) | \
225 ZIO_FLAG_GANG_CHILD | ZIO_FLAG_CANFAIL)
227 #define ZIO_VDEV_CHILD_FLAGS(zio) \
228 (((zio)->io_flags & ZIO_FLAG_VDEV_INHERIT) | \
246 * We'll take the number 122 and 123 to indicate checksum errors and
247 * fragmentation. Those doesn't collide with any errno values as they
248 * are greater than ELAST.
253 typedef void zio_done_func_t(zio_t *zio);
255 extern const char *zio_type_name[ZIO_TYPES];
258 * A bookmark is a four-tuple <objset, object, level, blkid> that uniquely
259 * identifies any block in the pool. By convention, the meta-objset (MOS)
260 * is objset 0, and the meta-dnode is object 0. This covers all blocks
261 * except root blocks and ZIL blocks, which are defined as follows:
263 * Root blocks (objset_phys_t) are object 0, level -1: <objset, 0, -1, 0>.
264 * ZIL blocks are bookmarked <objset, 0, -2, blkid == ZIL sequence number>.
265 * dmu_sync()ed ZIL data blocks are bookmarked <objset, object, -2, blkid>.
267 * Note: this structure is called a bookmark because its original purpose
268 * was to remember where to resume a pool-wide traverse.
270 * Note: this structure is passed between userland and the kernel, and is
271 * stored on disk (by virtue of being incorporated into other on-disk
272 * structures, e.g. dsl_scan_phys_t).
274 typedef struct zbookmark_phys {
281 #define SET_BOOKMARK(zb, objset, object, level, blkid) \
283 (zb)->zb_objset = objset; \
284 (zb)->zb_object = object; \
285 (zb)->zb_level = level; \
286 (zb)->zb_blkid = blkid; \
289 #define ZB_DESTROYED_OBJSET (-1ULL)
291 #define ZB_ROOT_OBJECT (0ULL)
292 #define ZB_ROOT_LEVEL (-1LL)
293 #define ZB_ROOT_BLKID (0ULL)
295 #define ZB_ZIL_OBJECT (0ULL)
296 #define ZB_ZIL_LEVEL (-2LL)
298 #define ZB_IS_ZERO(zb) \
299 ((zb)->zb_objset == 0 && (zb)->zb_object == 0 && \
300 (zb)->zb_level == 0 && (zb)->zb_blkid == 0)
301 #define ZB_IS_ROOT(zb) \
302 ((zb)->zb_object == ZB_ROOT_OBJECT && \
303 (zb)->zb_level == ZB_ROOT_LEVEL && \
304 (zb)->zb_blkid == ZB_ROOT_BLKID)
306 typedef struct zio_prop {
307 enum zio_checksum zp_checksum;
308 enum zio_compress zp_compress;
309 dmu_object_type_t zp_type;
313 boolean_t zp_dedup_verify;
314 boolean_t zp_nopwrite;
317 typedef struct zio_cksum_report zio_cksum_report_t;
319 typedef void zio_cksum_finish_f(zio_cksum_report_t *rep,
320 const void *good_data);
321 typedef void zio_cksum_free_f(void *cbdata, size_t size);
323 struct zio_bad_cksum; /* defined in zio_checksum.h */
326 struct zio_cksum_report {
327 struct zio_cksum_report *zcr_next;
328 nvlist_t *zcr_ereport;
329 nvlist_t *zcr_detector;
331 size_t zcr_cbinfo; /* passed to zcr_free() */
334 zio_cksum_finish_f *zcr_finish;
335 zio_cksum_free_f *zcr_free;
337 /* internal use only */
338 struct zio_bad_cksum *zcr_ckinfo; /* information from failure */
341 typedef void zio_vsd_cksum_report_f(zio_t *zio, zio_cksum_report_t *zcr,
344 zio_vsd_cksum_report_f zio_vsd_default_cksum_report;
346 typedef struct zio_vsd_ops {
347 zio_done_func_t *vsd_free;
348 zio_vsd_cksum_report_f *vsd_cksum_report;
351 typedef struct zio_gang_node {
352 zio_gbh_phys_t *gn_gbh;
353 struct zio_gang_node *gn_child[SPA_GBH_NBLKPTRS];
356 typedef zio_t *zio_gang_issue_func_t(zio_t *zio, blkptr_t *bp,
357 zio_gang_node_t *gn, void *data);
359 typedef void zio_transform_func_t(zio_t *zio, void *data, uint64_t size);
361 typedef struct zio_transform {
363 uint64_t zt_orig_size;
365 zio_transform_func_t *zt_transform;
366 struct zio_transform *zt_next;
369 typedef int zio_pipe_stage_t(zio_t *zio);
372 * The io_reexecute flags are distinct from io_flags because the child must
373 * be able to propagate them to the parent. The normal io_flags are local
374 * to the zio, not protected by any lock, and not modifiable by children;
375 * the reexecute flags are protected by io_lock, modifiable by children,
376 * and always propagated -- even when ZIO_FLAG_DONT_PROPAGATE is set.
378 #define ZIO_REEXECUTE_NOW 0x01
379 #define ZIO_REEXECUTE_SUSPEND 0x02
381 typedef struct zio_link {
384 list_node_t zl_parent_node;
385 list_node_t zl_child_node;
389 * Used for TRIM kstat.
391 typedef struct zio_trim_stats {
393 * Number of bytes successfully TRIMmed.
398 * Number of successful TRIM requests.
400 kstat_named_t success;
403 * Number of TRIM requests that failed because TRIM is not
406 kstat_named_t unsupported;
409 * Number of TRIM requests that failed for other reasons.
411 kstat_named_t failed;
414 extern zio_trim_stats_t zio_trim_stats;
416 #define ZIO_TRIM_STAT_INCR(stat, val) \
417 atomic_add_64(&zio_trim_stats.stat.value.ui64, (val));
418 #define ZIO_TRIM_STAT_BUMP(stat) \
419 ZIO_TRIM_STAT_INCR(stat, 1);
422 /* Core information about this I/O */
423 zbookmark_phys_t io_bookmark;
426 enum zio_child io_child_type;
428 zio_priority_t io_priority;
429 uint8_t io_reexecute;
430 uint8_t io_state[ZIO_WAIT_TYPES];
434 blkptr_t *io_bp_override;
436 list_t io_parent_list;
437 list_t io_child_list;
438 zio_link_t *io_walk_link;
440 zio_transform_t *io_transform_stack;
443 zio_done_func_t *io_ready;
444 zio_done_func_t *io_physdone;
445 zio_done_func_t *io_done;
447 int64_t io_prev_space_delta; /* DMU private */
450 /* Data represented by this I/O */
454 uint64_t io_orig_size;
456 /* Stuff for the vdev stack */
459 const zio_vsd_ops_t *io_vsd_ops;
462 hrtime_t io_timestamp;
463 avl_node_t io_queue_node;
465 /* Internal pipeline state */
466 enum zio_flag io_flags;
467 enum zio_stage io_stage;
468 enum zio_stage io_pipeline;
469 enum zio_flag io_orig_flags;
470 enum zio_stage io_orig_stage;
471 enum zio_stage io_orig_pipeline;
473 int io_child_error[ZIO_CHILD_TYPES];
474 uint64_t io_children[ZIO_CHILD_TYPES][ZIO_WAIT_TYPES];
475 uint64_t io_child_count;
476 uint64_t io_phys_children;
477 uint64_t io_parent_count;
479 zio_t *io_gang_leader;
480 zio_gang_node_t *io_gang_tree;
487 zio_cksum_report_t *io_cksum_report;
490 /* Taskq dispatching state */
491 taskq_ent_t io_tqent;
493 avl_node_t io_trim_node;
494 list_node_t io_trim_link;
497 extern zio_t *zio_null(zio_t *pio, spa_t *spa, vdev_t *vd,
498 zio_done_func_t *done, void *priv, enum zio_flag flags);
500 extern zio_t *zio_root(spa_t *spa,
501 zio_done_func_t *done, void *priv, enum zio_flag flags);
503 extern zio_t *zio_read(zio_t *pio, spa_t *spa, const blkptr_t *bp, void *data,
504 uint64_t size, zio_done_func_t *done, void *priv,
505 zio_priority_t priority, enum zio_flag flags, const zbookmark_phys_t *zb);
507 extern zio_t *zio_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
508 void *data, uint64_t size, const zio_prop_t *zp,
509 zio_done_func_t *ready, zio_done_func_t *physdone, zio_done_func_t *done,
511 zio_priority_t priority, enum zio_flag flags, const zbookmark_phys_t *zb);
513 extern zio_t *zio_rewrite(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
514 void *data, uint64_t size, zio_done_func_t *done, void *priv,
515 zio_priority_t priority, enum zio_flag flags, zbookmark_phys_t *zb);
517 extern void zio_write_override(zio_t *zio, blkptr_t *bp, int copies,
520 extern void zio_free(spa_t *spa, uint64_t txg, const blkptr_t *bp);
522 extern zio_t *zio_claim(zio_t *pio, spa_t *spa, uint64_t txg,
524 zio_done_func_t *done, void *priv, enum zio_flag flags);
526 extern zio_t *zio_ioctl(zio_t *pio, spa_t *spa, vdev_t *vd, int cmd,
527 uint64_t offset, uint64_t size, zio_done_func_t *done, void *priv,
528 zio_priority_t priority, enum zio_flag flags);
530 extern zio_t *zio_read_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
531 uint64_t size, void *data, int checksum,
532 zio_done_func_t *done, void *priv, zio_priority_t priority,
533 enum zio_flag flags, boolean_t labels);
535 extern zio_t *zio_write_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
536 uint64_t size, void *data, int checksum,
537 zio_done_func_t *done, void *priv, zio_priority_t priority,
538 enum zio_flag flags, boolean_t labels);
540 extern zio_t *zio_free_sync(zio_t *pio, spa_t *spa, uint64_t txg,
541 const blkptr_t *bp, uint64_t size, enum zio_flag flags);
543 extern int zio_alloc_zil(spa_t *spa, uint64_t txg, blkptr_t *new_bp,
544 blkptr_t *old_bp, uint64_t size, boolean_t use_slog);
545 extern void zio_free_zil(spa_t *spa, uint64_t txg, blkptr_t *bp);
546 extern void zio_flush(zio_t *zio, vdev_t *vd);
547 extern zio_t *zio_trim(zio_t *zio, spa_t *spa, vdev_t *vd, uint64_t offset,
549 extern void zio_shrink(zio_t *zio, uint64_t size);
551 extern int zio_wait(zio_t *zio);
552 extern void zio_nowait(zio_t *zio);
553 extern void zio_execute(zio_t *zio);
554 extern void zio_interrupt(zio_t *zio);
556 extern zio_t *zio_walk_parents(zio_t *cio);
557 extern zio_t *zio_walk_children(zio_t *pio);
558 extern zio_t *zio_unique_parent(zio_t *cio);
559 extern void zio_add_child(zio_t *pio, zio_t *cio);
561 extern void *zio_buf_alloc(size_t size);
562 extern void zio_buf_free(void *buf, size_t size);
563 extern void *zio_data_buf_alloc(size_t size);
564 extern void zio_data_buf_free(void *buf, size_t size);
566 extern void zio_resubmit_stage_async(void *);
568 extern zio_t *zio_vdev_child_io(zio_t *zio, blkptr_t *bp, vdev_t *vd,
569 uint64_t offset, void *data, uint64_t size, int type,
570 zio_priority_t priority, enum zio_flag flags,
571 zio_done_func_t *done, void *priv);
573 extern zio_t *zio_vdev_delegated_io(vdev_t *vd, uint64_t offset,
574 void *data, uint64_t size, int type, zio_priority_t priority,
575 enum zio_flag flags, zio_done_func_t *done, void *priv);
577 extern void zio_vdev_io_bypass(zio_t *zio);
578 extern void zio_vdev_io_reissue(zio_t *zio);
579 extern void zio_vdev_io_redone(zio_t *zio);
581 extern void zio_checksum_verified(zio_t *zio);
582 extern int zio_worst_error(int e1, int e2);
584 extern enum zio_checksum zio_checksum_select(enum zio_checksum child,
585 enum zio_checksum parent);
586 extern enum zio_checksum zio_checksum_dedup_select(spa_t *spa,
587 enum zio_checksum child, enum zio_checksum parent);
588 extern enum zio_compress zio_compress_select(spa_t *spa,
589 enum zio_compress child, enum zio_compress parent);
591 extern void zio_suspend(spa_t *spa, zio_t *zio);
592 extern int zio_resume(spa_t *spa);
593 extern void zio_resume_wait(spa_t *spa);
596 * Initial setup and teardown.
598 extern void zio_init(void);
599 extern void zio_fini(void);
604 struct zinject_record;
605 extern uint32_t zio_injection_enabled;
606 extern int zio_inject_fault(char *name, int flags, int *id,
607 struct zinject_record *record);
608 extern int zio_inject_list_next(int *id, char *name, size_t buflen,
609 struct zinject_record *record);
610 extern int zio_clear_fault(int id);
611 extern void zio_handle_panic_injection(spa_t *spa, char *tag, uint64_t type);
612 extern int zio_handle_fault_injection(zio_t *zio, int error);
613 extern int zio_handle_device_injection(vdev_t *vd, zio_t *zio, int error);
614 extern int zio_handle_label_injection(zio_t *zio, int error);
615 extern void zio_handle_ignored_writes(zio_t *zio);
616 extern uint64_t zio_handle_io_delay(zio_t *zio);
619 * Checksum ereport functions
621 extern void zfs_ereport_start_checksum(spa_t *spa, vdev_t *vd, struct zio *zio,
622 uint64_t offset, uint64_t length, void *arg, struct zio_bad_cksum *info);
623 extern void zfs_ereport_finish_checksum(zio_cksum_report_t *report,
624 const void *good_data, const void *bad_data, boolean_t drop_if_identical);
626 extern void zfs_ereport_send_interim_checksum(zio_cksum_report_t *report);
627 extern void zfs_ereport_free_checksum(zio_cksum_report_t *report);
629 /* If we have the good data in hand, this function can be used */
630 extern void zfs_ereport_post_checksum(spa_t *spa, vdev_t *vd,
631 struct zio *zio, uint64_t offset, uint64_t length,
632 const void *good_data, const void *bad_data, struct zio_bad_cksum *info);
634 /* Called from spa_sync(), but primarily an injection handler */
635 extern void spa_handle_ignored_writes(spa_t *spa);
637 /* zbookmark_phys functions */
638 boolean_t zbookmark_is_before(const struct dnode_phys *dnp,
639 const zbookmark_phys_t *zb1, const zbookmark_phys_t *zb2);