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 (c) 2012 by Delphix. All rights reserved.
25 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
26 * Copyright (c) 2013, Joyent, Inc. 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
87 #define ZIO_CHECKSUM_ON_VALUE ZIO_CHECKSUM_FLETCHER_4
88 #define ZIO_CHECKSUM_DEFAULT ZIO_CHECKSUM_ON
90 #define ZIO_CHECKSUM_MASK 0xffULL
91 #define ZIO_CHECKSUM_VERIFY (1 << 8)
93 #define ZIO_DEDUPCHECKSUM ZIO_CHECKSUM_SHA256
94 #define ZIO_DEDUPDITTO_MIN 100
97 ZIO_COMPRESS_INHERIT = 0,
113 ZIO_COMPRESS_FUNCTIONS
116 /* N.B. when altering this value, also change BOOTFS_COMPRESS_VALID below */
117 #define ZIO_COMPRESS_ON_VALUE ZIO_COMPRESS_LZJB
118 #define ZIO_COMPRESS_DEFAULT ZIO_COMPRESS_OFF
120 #define BOOTFS_COMPRESS_VALID(compress) \
121 ((compress) == ZIO_COMPRESS_LZJB || \
122 (compress) == ZIO_COMPRESS_LZ4 || \
123 ((compress) == ZIO_COMPRESS_ON && \
124 ZIO_COMPRESS_ON_VALUE == ZIO_COMPRESS_LZJB) || \
125 (compress) == ZIO_COMPRESS_OFF)
127 #define ZIO_FAILURE_MODE_WAIT 0
128 #define ZIO_FAILURE_MODE_CONTINUE 1
129 #define ZIO_FAILURE_MODE_PANIC 2
131 #define ZIO_PRIORITY_NOW (zio_priority_table[0])
132 #define ZIO_PRIORITY_SYNC_READ (zio_priority_table[1])
133 #define ZIO_PRIORITY_SYNC_WRITE (zio_priority_table[2])
134 #define ZIO_PRIORITY_LOG_WRITE (zio_priority_table[3])
135 #define ZIO_PRIORITY_CACHE_FILL (zio_priority_table[4])
136 #define ZIO_PRIORITY_AGG (zio_priority_table[5])
137 #define ZIO_PRIORITY_FREE (zio_priority_table[6])
138 #define ZIO_PRIORITY_ASYNC_WRITE (zio_priority_table[7])
139 #define ZIO_PRIORITY_ASYNC_READ (zio_priority_table[8])
140 #define ZIO_PRIORITY_RESILVER (zio_priority_table[9])
141 #define ZIO_PRIORITY_SCRUB (zio_priority_table[10])
142 #define ZIO_PRIORITY_DDT_PREFETCH (zio_priority_table[11])
143 #define ZIO_PRIORITY_TRIM (zio_priority_table[12])
144 #define ZIO_PRIORITY_TABLE_SIZE 13
146 #define ZIO_PIPELINE_CONTINUE 0x100
147 #define ZIO_PIPELINE_STOP 0x101
151 * Flags inherited by gang, ddt, and vdev children,
152 * and that must be equal for two zios to aggregate
154 ZIO_FLAG_DONT_AGGREGATE = 1 << 0,
155 ZIO_FLAG_IO_REPAIR = 1 << 1,
156 ZIO_FLAG_SELF_HEAL = 1 << 2,
157 ZIO_FLAG_RESILVER = 1 << 3,
158 ZIO_FLAG_SCRUB = 1 << 4,
159 ZIO_FLAG_SCAN_THREAD = 1 << 5,
161 #define ZIO_FLAG_AGG_INHERIT (ZIO_FLAG_CANFAIL - 1)
164 * Flags inherited by ddt, gang, and vdev children.
166 ZIO_FLAG_CANFAIL = 1 << 6, /* must be first for INHERIT */
167 ZIO_FLAG_SPECULATIVE = 1 << 7,
168 ZIO_FLAG_CONFIG_WRITER = 1 << 8,
169 ZIO_FLAG_DONT_RETRY = 1 << 9,
170 ZIO_FLAG_DONT_CACHE = 1 << 10,
171 ZIO_FLAG_NODATA = 1 << 11,
172 ZIO_FLAG_INDUCE_DAMAGE = 1 << 12,
174 #define ZIO_FLAG_DDT_INHERIT (ZIO_FLAG_IO_RETRY - 1)
175 #define ZIO_FLAG_GANG_INHERIT (ZIO_FLAG_IO_RETRY - 1)
178 * Flags inherited by vdev children.
180 ZIO_FLAG_IO_RETRY = 1 << 13, /* must be first for INHERIT */
181 ZIO_FLAG_PROBE = 1 << 14,
182 ZIO_FLAG_TRYHARD = 1 << 15,
183 ZIO_FLAG_OPTIONAL = 1 << 16,
185 #define ZIO_FLAG_VDEV_INHERIT (ZIO_FLAG_DONT_QUEUE - 1)
188 * Flags not inherited by any children.
190 ZIO_FLAG_DONT_QUEUE = 1 << 17, /* must be first for INHERIT */
191 ZIO_FLAG_DONT_PROPAGATE = 1 << 18,
192 ZIO_FLAG_IO_BYPASS = 1 << 19,
193 ZIO_FLAG_IO_REWRITE = 1 << 20,
194 ZIO_FLAG_RAW = 1 << 21,
195 ZIO_FLAG_GANG_CHILD = 1 << 22,
196 ZIO_FLAG_DDT_CHILD = 1 << 23,
197 ZIO_FLAG_GODFATHER = 1 << 24,
198 ZIO_FLAG_NOPWRITE = 1 << 25,
199 ZIO_FLAG_REEXECUTED = 1 << 26,
202 #define ZIO_FLAG_MUSTSUCCEED 0
204 #define ZIO_DDT_CHILD_FLAGS(zio) \
205 (((zio)->io_flags & ZIO_FLAG_DDT_INHERIT) | \
206 ZIO_FLAG_DDT_CHILD | ZIO_FLAG_CANFAIL)
208 #define ZIO_GANG_CHILD_FLAGS(zio) \
209 (((zio)->io_flags & ZIO_FLAG_GANG_INHERIT) | \
210 ZIO_FLAG_GANG_CHILD | ZIO_FLAG_CANFAIL)
212 #define ZIO_VDEV_CHILD_FLAGS(zio) \
213 (((zio)->io_flags & ZIO_FLAG_VDEV_INHERIT) | \
231 * We'll take the number 122 and 123 to indicate checksum errors and
232 * fragmentation. Those doesn't collide with any errno values as they
233 * are greater than ELAST.
238 typedef void zio_done_func_t(zio_t *zio);
240 extern uint8_t zio_priority_table[ZIO_PRIORITY_TABLE_SIZE];
241 extern char *zio_type_name[ZIO_TYPES];
244 * A bookmark is a four-tuple <objset, object, level, blkid> that uniquely
245 * identifies any block in the pool. By convention, the meta-objset (MOS)
246 * is objset 0, and the meta-dnode is object 0. This covers all blocks
247 * except root blocks and ZIL blocks, which are defined as follows:
249 * Root blocks (objset_phys_t) are object 0, level -1: <objset, 0, -1, 0>.
250 * ZIL blocks are bookmarked <objset, 0, -2, blkid == ZIL sequence number>.
251 * dmu_sync()ed ZIL data blocks are bookmarked <objset, object, -2, blkid>.
253 * Note: this structure is called a bookmark because its original purpose
254 * was to remember where to resume a pool-wide traverse.
256 * Note: this structure is passed between userland and the kernel.
257 * Therefore it must not change size or alignment between 32/64 bit
258 * compilation options.
260 typedef struct zbookmark {
267 #define SET_BOOKMARK(zb, objset, object, level, blkid) \
269 (zb)->zb_objset = objset; \
270 (zb)->zb_object = object; \
271 (zb)->zb_level = level; \
272 (zb)->zb_blkid = blkid; \
275 #define ZB_DESTROYED_OBJSET (-1ULL)
277 #define ZB_ROOT_OBJECT (0ULL)
278 #define ZB_ROOT_LEVEL (-1LL)
279 #define ZB_ROOT_BLKID (0ULL)
281 #define ZB_ZIL_OBJECT (0ULL)
282 #define ZB_ZIL_LEVEL (-2LL)
284 #define ZB_IS_ZERO(zb) \
285 ((zb)->zb_objset == 0 && (zb)->zb_object == 0 && \
286 (zb)->zb_level == 0 && (zb)->zb_blkid == 0)
287 #define ZB_IS_ROOT(zb) \
288 ((zb)->zb_object == ZB_ROOT_OBJECT && \
289 (zb)->zb_level == ZB_ROOT_LEVEL && \
290 (zb)->zb_blkid == ZB_ROOT_BLKID)
292 typedef struct zio_prop {
293 enum zio_checksum zp_checksum;
294 enum zio_compress zp_compress;
295 dmu_object_type_t zp_type;
299 boolean_t zp_dedup_verify;
300 boolean_t zp_nopwrite;
303 typedef struct zio_cksum_report zio_cksum_report_t;
305 typedef void zio_cksum_finish_f(zio_cksum_report_t *rep,
306 const void *good_data);
307 typedef void zio_cksum_free_f(void *cbdata, size_t size);
309 struct zio_bad_cksum; /* defined in zio_checksum.h */
312 struct zio_cksum_report {
313 struct zio_cksum_report *zcr_next;
314 nvlist_t *zcr_ereport;
315 nvlist_t *zcr_detector;
317 size_t zcr_cbinfo; /* passed to zcr_free() */
320 zio_cksum_finish_f *zcr_finish;
321 zio_cksum_free_f *zcr_free;
323 /* internal use only */
324 struct zio_bad_cksum *zcr_ckinfo; /* information from failure */
327 typedef void zio_vsd_cksum_report_f(zio_t *zio, zio_cksum_report_t *zcr,
330 zio_vsd_cksum_report_f zio_vsd_default_cksum_report;
332 typedef struct zio_vsd_ops {
333 zio_done_func_t *vsd_free;
334 zio_vsd_cksum_report_f *vsd_cksum_report;
337 typedef struct zio_gang_node {
338 zio_gbh_phys_t *gn_gbh;
339 struct zio_gang_node *gn_child[SPA_GBH_NBLKPTRS];
342 typedef zio_t *zio_gang_issue_func_t(zio_t *zio, blkptr_t *bp,
343 zio_gang_node_t *gn, void *data);
345 typedef void zio_transform_func_t(zio_t *zio, void *data, uint64_t size);
347 typedef struct zio_transform {
349 uint64_t zt_orig_size;
351 zio_transform_func_t *zt_transform;
352 struct zio_transform *zt_next;
355 typedef int zio_pipe_stage_t(zio_t *zio);
358 * The io_reexecute flags are distinct from io_flags because the child must
359 * be able to propagate them to the parent. The normal io_flags are local
360 * to the zio, not protected by any lock, and not modifiable by children;
361 * the reexecute flags are protected by io_lock, modifiable by children,
362 * and always propagated -- even when ZIO_FLAG_DONT_PROPAGATE is set.
364 #define ZIO_REEXECUTE_NOW 0x01
365 #define ZIO_REEXECUTE_SUSPEND 0x02
367 typedef struct zio_link {
370 list_node_t zl_parent_node;
371 list_node_t zl_child_node;
375 * Used for TRIM kstat.
377 typedef struct zio_trim_stats {
379 * Number of bytes successfully TRIMmed.
384 * Number of successful TRIM requests.
386 kstat_named_t success;
389 * Number of TRIM requests that failed because TRIM is not
392 kstat_named_t unsupported;
395 * Number of TRIM requests that failed for other reasons.
397 kstat_named_t failed;
400 extern zio_trim_stats_t zio_trim_stats;
402 #define ZIO_TRIM_STAT_INCR(stat, val) \
403 atomic_add_64(&zio_trim_stats.stat.value.ui64, (val));
404 #define ZIO_TRIM_STAT_BUMP(stat) \
405 ZIO_TRIM_STAT_INCR(stat, 1);
408 /* Core information about this I/O */
409 zbookmark_t io_bookmark;
412 enum zio_child io_child_type;
415 uint8_t io_reexecute;
416 uint8_t io_state[ZIO_WAIT_TYPES];
420 blkptr_t *io_bp_override;
422 list_t io_parent_list;
423 list_t io_child_list;
424 zio_link_t *io_walk_link;
426 zio_transform_t *io_transform_stack;
429 zio_done_func_t *io_ready;
430 zio_done_func_t *io_done;
432 int64_t io_prev_space_delta; /* DMU private */
435 /* Data represented by this I/O */
439 uint64_t io_orig_size;
441 /* Stuff for the vdev stack */
444 const zio_vsd_ops_t *io_vsd_ops;
447 uint64_t io_deadline;
448 hrtime_t io_timestamp;
449 avl_node_t io_offset_node;
450 avl_node_t io_deadline_node;
451 avl_tree_t *io_vdev_tree;
453 /* Internal pipeline state */
454 enum zio_flag io_flags;
455 enum zio_stage io_stage;
456 enum zio_stage io_pipeline;
457 enum zio_flag io_orig_flags;
458 enum zio_stage io_orig_stage;
459 enum zio_stage io_orig_pipeline;
461 int io_child_error[ZIO_CHILD_TYPES];
462 uint64_t io_children[ZIO_CHILD_TYPES][ZIO_WAIT_TYPES];
463 uint64_t io_child_count;
464 uint64_t io_parent_count;
466 zio_t *io_gang_leader;
467 zio_gang_node_t *io_gang_tree;
474 zio_cksum_report_t *io_cksum_report;
479 struct ostask io_task;
481 avl_node_t io_trim_node;
482 list_node_t io_trim_link;
485 extern zio_t *zio_null(zio_t *pio, spa_t *spa, vdev_t *vd,
486 zio_done_func_t *done, void *priv, enum zio_flag flags);
488 extern zio_t *zio_root(spa_t *spa,
489 zio_done_func_t *done, void *priv, enum zio_flag flags);
491 extern zio_t *zio_read(zio_t *pio, spa_t *spa, const blkptr_t *bp, void *data,
492 uint64_t size, zio_done_func_t *done, void *priv,
493 int priority, enum zio_flag flags, const zbookmark_t *zb);
495 extern zio_t *zio_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
496 void *data, uint64_t size, const zio_prop_t *zp,
497 zio_done_func_t *ready, zio_done_func_t *done, void *priv,
498 int priority, enum zio_flag flags, const zbookmark_t *zb);
500 extern zio_t *zio_rewrite(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
501 void *data, uint64_t size, zio_done_func_t *done, void *priv,
502 int priority, enum zio_flag flags, zbookmark_t *zb);
504 extern void zio_write_override(zio_t *zio, blkptr_t *bp, int copies,
507 extern void zio_free(spa_t *spa, uint64_t txg, const blkptr_t *bp);
509 extern zio_t *zio_claim(zio_t *pio, spa_t *spa, uint64_t txg,
511 zio_done_func_t *done, void *priv, enum zio_flag flags);
513 extern zio_t *zio_ioctl(zio_t *pio, spa_t *spa, vdev_t *vd, int cmd,
514 uint64_t offset, uint64_t size, zio_done_func_t *done, void *priv,
515 int priority, enum zio_flag flags);
517 extern zio_t *zio_read_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
518 uint64_t size, void *data, int checksum,
519 zio_done_func_t *done, void *priv, int priority, enum zio_flag flags,
522 extern zio_t *zio_write_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
523 uint64_t size, void *data, int checksum,
524 zio_done_func_t *done, void *priv, int priority, enum zio_flag flags,
527 extern zio_t *zio_free_sync(zio_t *pio, spa_t *spa, uint64_t txg,
528 const blkptr_t *bp, uint64_t size, enum zio_flag flags);
530 extern int zio_alloc_zil(spa_t *spa, uint64_t txg, blkptr_t *new_bp,
531 blkptr_t *old_bp, uint64_t size, boolean_t use_slog);
532 extern void zio_free_zil(spa_t *spa, uint64_t txg, blkptr_t *bp);
533 extern void zio_flush(zio_t *zio, vdev_t *vd);
534 extern zio_t *zio_trim(zio_t *zio, spa_t *spa, vdev_t *vd, uint64_t offset,
536 extern void zio_shrink(zio_t *zio, uint64_t size);
538 extern int zio_wait(zio_t *zio);
539 extern void zio_nowait(zio_t *zio);
540 extern void zio_execute(zio_t *zio);
541 extern void zio_interrupt(zio_t *zio);
543 extern zio_t *zio_walk_parents(zio_t *cio);
544 extern zio_t *zio_walk_children(zio_t *pio);
545 extern zio_t *zio_unique_parent(zio_t *cio);
546 extern void zio_add_child(zio_t *pio, zio_t *cio);
548 extern void *zio_buf_alloc(size_t size);
549 extern void zio_buf_free(void *buf, size_t size);
550 extern void *zio_data_buf_alloc(size_t size);
551 extern void zio_data_buf_free(void *buf, size_t size);
553 extern void zio_resubmit_stage_async(void *);
555 extern zio_t *zio_vdev_child_io(zio_t *zio, blkptr_t *bp, vdev_t *vd,
556 uint64_t offset, void *data, uint64_t size, int type, int priority,
557 enum zio_flag flags, zio_done_func_t *done, void *priv);
559 extern zio_t *zio_vdev_delegated_io(vdev_t *vd, uint64_t offset,
560 void *data, uint64_t size, int type, int priority,
561 enum zio_flag flags, zio_done_func_t *done, void *priv);
563 extern void zio_vdev_io_bypass(zio_t *zio);
564 extern void zio_vdev_io_reissue(zio_t *zio);
565 extern void zio_vdev_io_redone(zio_t *zio);
567 extern void zio_checksum_verified(zio_t *zio);
568 extern int zio_worst_error(int e1, int e2);
570 extern enum zio_checksum zio_checksum_select(enum zio_checksum child,
571 enum zio_checksum parent);
572 extern enum zio_checksum zio_checksum_dedup_select(spa_t *spa,
573 enum zio_checksum child, enum zio_checksum parent);
574 extern enum zio_compress zio_compress_select(enum zio_compress child,
575 enum zio_compress parent);
577 extern void zio_suspend(spa_t *spa, zio_t *zio);
578 extern int zio_resume(spa_t *spa);
579 extern void zio_resume_wait(spa_t *spa);
582 * Initial setup and teardown.
584 extern void zio_init(void);
585 extern void zio_fini(void);
590 struct zinject_record;
591 extern uint32_t zio_injection_enabled;
592 extern int zio_inject_fault(char *name, int flags, int *id,
593 struct zinject_record *record);
594 extern int zio_inject_list_next(int *id, char *name, size_t buflen,
595 struct zinject_record *record);
596 extern int zio_clear_fault(int id);
597 extern void zio_handle_panic_injection(spa_t *spa, char *tag, uint64_t type);
598 extern int zio_handle_fault_injection(zio_t *zio, int error);
599 extern int zio_handle_device_injection(vdev_t *vd, zio_t *zio, int error);
600 extern int zio_handle_label_injection(zio_t *zio, int error);
601 extern void zio_handle_ignored_writes(zio_t *zio);
602 extern uint64_t zio_handle_io_delay(zio_t *zio);
605 * Checksum ereport functions
607 extern void zfs_ereport_start_checksum(spa_t *spa, vdev_t *vd, struct zio *zio,
608 uint64_t offset, uint64_t length, void *arg, struct zio_bad_cksum *info);
609 extern void zfs_ereport_finish_checksum(zio_cksum_report_t *report,
610 const void *good_data, const void *bad_data, boolean_t drop_if_identical);
612 extern void zfs_ereport_send_interim_checksum(zio_cksum_report_t *report);
613 extern void zfs_ereport_free_checksum(zio_cksum_report_t *report);
615 /* If we have the good data in hand, this function can be used */
616 extern void zfs_ereport_post_checksum(spa_t *spa, vdev_t *vd,
617 struct zio *zio, uint64_t offset, uint64_t length,
618 const void *good_data, const void *bad_data, struct zio_bad_cksum *info);
620 /* Called from spa_sync(), but primarily an injection handler */
621 extern void spa_handle_ignored_writes(spa_t *spa);
623 /* zbookmark functions */
624 boolean_t zbookmark_is_before(const struct dnode_phys *dnp,
625 const zbookmark_t *zb1, const zbookmark_t *zb2);