xref: /illumos-gate/usr/src/uts/common/fs/zfs/sys/zio.h (revision 344db6f401efe76f7e1d4f4c1a644ae593910219)
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 /*
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, 2018 by Delphix. All rights reserved.
26  * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
27  * Copyright (c) 2013, Joyent, Inc. All rights reserved.
28  * Copyright 2016 Toomas Soome <tsoome@me.com>
29  */
30 
31 #ifndef _ZIO_H
32 #define	_ZIO_H
33 
34 #include <sys/zio_priority.h>
35 #include <sys/zfs_context.h>
36 #include <sys/spa.h>
37 #include <sys/txg.h>
38 #include <sys/avl.h>
39 #include <sys/fs/zfs.h>
40 #include <sys/zio_impl.h>
41 
42 #ifdef	__cplusplus
43 extern "C" {
44 #endif
45 
46 /*
47  * Embedded checksum
48  */
49 #define	ZEC_MAGIC	0x210da7ab10c7a11ULL
50 
51 typedef struct zio_eck {
52 	uint64_t	zec_magic;	/* for validation, endianness	*/
53 	zio_cksum_t	zec_cksum;	/* 256-bit checksum		*/
54 } zio_eck_t;
55 
56 /*
57  * Gang block headers are self-checksumming and contain an array
58  * of block pointers.
59  */
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))) /\
66 	sizeof (uint64_t))
67 
68 typedef struct zio_gbh {
69 	blkptr_t		zg_blkptr[SPA_GBH_NBLKPTRS];
70 	uint64_t		zg_filler[SPA_GBH_FILLER];
71 	zio_eck_t		zg_tail;
72 } zio_gbh_phys_t;
73 
74 enum zio_checksum {
75 	ZIO_CHECKSUM_INHERIT = 0,
76 	ZIO_CHECKSUM_ON,
77 	ZIO_CHECKSUM_OFF,
78 	ZIO_CHECKSUM_LABEL,
79 	ZIO_CHECKSUM_GANG_HEADER,
80 	ZIO_CHECKSUM_ZILOG,
81 	ZIO_CHECKSUM_FLETCHER_2,
82 	ZIO_CHECKSUM_FLETCHER_4,
83 	ZIO_CHECKSUM_SHA256,
84 	ZIO_CHECKSUM_ZILOG2,
85 	ZIO_CHECKSUM_NOPARITY,
86 	ZIO_CHECKSUM_SHA512,
87 	ZIO_CHECKSUM_SKEIN,
88 	ZIO_CHECKSUM_EDONR,
89 	ZIO_CHECKSUM_FUNCTIONS
90 };
91 
92 /*
93  * The number of "legacy" compression functions which can be set on individual
94  * objects.
95  */
96 #define	ZIO_CHECKSUM_LEGACY_FUNCTIONS ZIO_CHECKSUM_ZILOG2
97 
98 #define	ZIO_CHECKSUM_ON_VALUE	ZIO_CHECKSUM_FLETCHER_4
99 #define	ZIO_CHECKSUM_DEFAULT	ZIO_CHECKSUM_ON
100 
101 #define	ZIO_CHECKSUM_MASK	0xffULL
102 #define	ZIO_CHECKSUM_VERIFY	(1 << 8)
103 
104 #define	ZIO_DEDUPCHECKSUM	ZIO_CHECKSUM_SHA256
105 #define	ZIO_DEDUPDITTO_MIN	100
106 
107 /*
108  * The number of "legacy" compression functions which can be set on individual
109  * objects.
110  */
111 #define	ZIO_COMPRESS_LEGACY_FUNCTIONS ZIO_COMPRESS_LZ4
112 
113 /*
114  * The meaning of "compress = on" selected by the compression features enabled
115  * on a given pool.
116  */
117 #define	ZIO_COMPRESS_LEGACY_ON_VALUE	ZIO_COMPRESS_LZJB
118 #define	ZIO_COMPRESS_LZ4_ON_VALUE	ZIO_COMPRESS_LZ4
119 
120 #define	ZIO_COMPRESS_DEFAULT		ZIO_COMPRESS_OFF
121 
122 #define	BOOTFS_COMPRESS_VALID(compress)			\
123 	((compress) == ZIO_COMPRESS_LZJB ||		\
124 	(compress) == ZIO_COMPRESS_LZ4 ||		\
125 	(compress) == ZIO_COMPRESS_GZIP_1 ||		\
126 	(compress) == ZIO_COMPRESS_GZIP_2 ||		\
127 	(compress) == ZIO_COMPRESS_GZIP_3 ||		\
128 	(compress) == ZIO_COMPRESS_GZIP_4 ||		\
129 	(compress) == ZIO_COMPRESS_GZIP_5 ||		\
130 	(compress) == ZIO_COMPRESS_GZIP_6 ||		\
131 	(compress) == ZIO_COMPRESS_GZIP_7 ||		\
132 	(compress) == ZIO_COMPRESS_GZIP_8 ||		\
133 	(compress) == ZIO_COMPRESS_GZIP_9 ||		\
134 	(compress) == ZIO_COMPRESS_ON ||		\
135 	(compress) == ZIO_COMPRESS_OFF)
136 
137 #define	ZIO_FAILURE_MODE_WAIT		0
138 #define	ZIO_FAILURE_MODE_CONTINUE	1
139 #define	ZIO_FAILURE_MODE_PANIC		2
140 
141 typedef enum zio_suspend_reason {
142 	ZIO_SUSPEND_NONE = 0,
143 	ZIO_SUSPEND_IOERR,
144 	ZIO_SUSPEND_MMP,
145 } zio_suspend_reason_t;
146 
147 enum zio_flag {
148 	/*
149 	 * Flags inherited by gang, ddt, and vdev children,
150 	 * and that must be equal for two zios to aggregate
151 	 */
152 	ZIO_FLAG_DONT_AGGREGATE	= 1 << 0,
153 	ZIO_FLAG_IO_REPAIR	= 1 << 1,
154 	ZIO_FLAG_SELF_HEAL	= 1 << 2,
155 	ZIO_FLAG_RESILVER	= 1 << 3,
156 	ZIO_FLAG_SCRUB		= 1 << 4,
157 	ZIO_FLAG_SCAN_THREAD	= 1 << 5,
158 	ZIO_FLAG_PHYSICAL	= 1 << 6,
159 
160 #define	ZIO_FLAG_AGG_INHERIT	(ZIO_FLAG_CANFAIL - 1)
161 
162 	/*
163 	 * Flags inherited by ddt, gang, and vdev children.
164 	 */
165 	ZIO_FLAG_CANFAIL	= 1 << 7,	/* must be first for INHERIT */
166 	ZIO_FLAG_SPECULATIVE	= 1 << 8,
167 	ZIO_FLAG_CONFIG_WRITER	= 1 << 9,
168 	ZIO_FLAG_DONT_RETRY	= 1 << 10,
169 	ZIO_FLAG_DONT_CACHE	= 1 << 11,
170 	ZIO_FLAG_NODATA		= 1 << 12,
171 	ZIO_FLAG_INDUCE_DAMAGE	= 1 << 13,
172 	ZIO_FLAG_IO_ALLOCATING	= 1 << 14,
173 
174 #define	ZIO_FLAG_DDT_INHERIT	(ZIO_FLAG_IO_RETRY - 1)
175 #define	ZIO_FLAG_GANG_INHERIT	(ZIO_FLAG_IO_RETRY - 1)
176 
177 	/*
178 	 * Flags inherited by vdev children.
179 	 */
180 	ZIO_FLAG_IO_RETRY	= 1 << 15,	/* must be first for INHERIT */
181 	ZIO_FLAG_PROBE		= 1 << 16,
182 	ZIO_FLAG_TRYHARD	= 1 << 17,
183 	ZIO_FLAG_OPTIONAL	= 1 << 18,
184 
185 #define	ZIO_FLAG_VDEV_INHERIT	(ZIO_FLAG_DONT_QUEUE - 1)
186 
187 	/*
188 	 * Flags not inherited by any children.
189 	 */
190 	ZIO_FLAG_DONT_QUEUE	= 1 << 19,	/* must be first for INHERIT */
191 	ZIO_FLAG_DONT_PROPAGATE	= 1 << 20,
192 	ZIO_FLAG_IO_BYPASS	= 1 << 21,
193 	ZIO_FLAG_IO_REWRITE	= 1 << 22,
194 	ZIO_FLAG_RAW		= 1 << 23,
195 	ZIO_FLAG_GANG_CHILD	= 1 << 24,
196 	ZIO_FLAG_DDT_CHILD	= 1 << 25,
197 	ZIO_FLAG_GODFATHER	= 1 << 26,
198 	ZIO_FLAG_NOPWRITE	= 1 << 27,
199 	ZIO_FLAG_REEXECUTED	= 1 << 28,
200 	ZIO_FLAG_DELEGATED	= 1 << 29,
201 };
202 
203 #define	ZIO_FLAG_MUSTSUCCEED		0
204 
205 #define	ZIO_DDT_CHILD_FLAGS(zio)				\
206 	(((zio)->io_flags & ZIO_FLAG_DDT_INHERIT) |		\
207 	ZIO_FLAG_DDT_CHILD | ZIO_FLAG_CANFAIL)
208 
209 #define	ZIO_GANG_CHILD_FLAGS(zio)				\
210 	(((zio)->io_flags & ZIO_FLAG_GANG_INHERIT) |		\
211 	ZIO_FLAG_GANG_CHILD | ZIO_FLAG_CANFAIL)
212 
213 #define	ZIO_VDEV_CHILD_FLAGS(zio)				\
214 	(((zio)->io_flags & ZIO_FLAG_VDEV_INHERIT) |		\
215 	ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_CANFAIL)
216 
217 #define	ZIO_CHILD_BIT(x)		(1 << (x))
218 #define	ZIO_CHILD_BIT_IS_SET(val, x)	((val) & (1 << (x)))
219 
220 enum zio_child {
221 	ZIO_CHILD_VDEV = 0,
222 	ZIO_CHILD_GANG,
223 	ZIO_CHILD_DDT,
224 	ZIO_CHILD_LOGICAL,
225 	ZIO_CHILD_TYPES
226 };
227 
228 #define	ZIO_CHILD_VDEV_BIT		ZIO_CHILD_BIT(ZIO_CHILD_VDEV)
229 #define	ZIO_CHILD_GANG_BIT		ZIO_CHILD_BIT(ZIO_CHILD_GANG)
230 #define	ZIO_CHILD_DDT_BIT		ZIO_CHILD_BIT(ZIO_CHILD_DDT)
231 #define	ZIO_CHILD_LOGICAL_BIT		ZIO_CHILD_BIT(ZIO_CHILD_LOGICAL)
232 #define	ZIO_CHILD_ALL_BITS					\
233 	(ZIO_CHILD_VDEV_BIT | ZIO_CHILD_GANG_BIT |		\
234 	ZIO_CHILD_DDT_BIT | ZIO_CHILD_LOGICAL_BIT)
235 
236 enum zio_wait_type {
237 	ZIO_WAIT_READY = 0,
238 	ZIO_WAIT_DONE,
239 	ZIO_WAIT_TYPES
240 };
241 
242 /*
243  * We'll take the unused errnos, 'EBADE' and 'EBADR' (from the Convergent
244  * graveyard) to indicate checksum errors and fragmentation.
245  */
246 #define	ECKSUM	EBADE
247 #define	EFRAGS	EBADR
248 
249 typedef void zio_done_func_t(zio_t *zio);
250 
251 extern boolean_t zio_dva_throttle_enabled;
252 extern const char *zio_type_name[ZIO_TYPES];
253 
254 /*
255  * A bookmark is a four-tuple <objset, object, level, blkid> that uniquely
256  * identifies any block in the pool.  By convention, the meta-objset (MOS)
257  * is objset 0, and the meta-dnode is object 0.  This covers all blocks
258  * except root blocks and ZIL blocks, which are defined as follows:
259  *
260  * Root blocks (objset_phys_t) are object 0, level -1:  <objset, 0, -1, 0>.
261  * ZIL blocks are bookmarked <objset, 0, -2, blkid == ZIL sequence number>.
262  * dmu_sync()ed ZIL data blocks are bookmarked <objset, object, -2, blkid>.
263  * dnode visit bookmarks are <objset, object id of dnode, -3, 0>.
264  *
265  * Note: this structure is called a bookmark because its original purpose
266  * was to remember where to resume a pool-wide traverse.
267  *
268  * Note: this structure is passed between userland and the kernel, and is
269  * stored on disk (by virtue of being incorporated into other on-disk
270  * structures, e.g. dsl_scan_phys_t).
271  */
272 typedef struct zbookmark_phys {
273 	uint64_t	zb_objset;
274 	uint64_t	zb_object;
275 	int64_t		zb_level;
276 	uint64_t	zb_blkid;
277 } zbookmark_phys_t;
278 
279 #define	SET_BOOKMARK(zb, objset, object, level, blkid)  \
280 {                                                       \
281 	(zb)->zb_objset = objset;                       \
282 	(zb)->zb_object = object;                       \
283 	(zb)->zb_level = level;                         \
284 	(zb)->zb_blkid = blkid;                         \
285 }
286 
287 #define	ZB_DESTROYED_OBJSET	(-1ULL)
288 
289 #define	ZB_ROOT_OBJECT		(0ULL)
290 #define	ZB_ROOT_LEVEL		(-1LL)
291 #define	ZB_ROOT_BLKID		(0ULL)
292 
293 #define	ZB_ZIL_OBJECT		(0ULL)
294 #define	ZB_ZIL_LEVEL		(-2LL)
295 
296 #define	ZB_DNODE_LEVEL		(-3LL)
297 #define	ZB_DNODE_BLKID		(0ULL)
298 
299 #define	ZB_IS_ZERO(zb)						\
300 	((zb)->zb_objset == 0 && (zb)->zb_object == 0 &&	\
301 	(zb)->zb_level == 0 && (zb)->zb_blkid == 0)
302 #define	ZB_IS_ROOT(zb)				\
303 	((zb)->zb_object == ZB_ROOT_OBJECT &&	\
304 	(zb)->zb_level == ZB_ROOT_LEVEL &&	\
305 	(zb)->zb_blkid == ZB_ROOT_BLKID)
306 
307 typedef struct zio_prop {
308 	enum zio_checksum	zp_checksum;
309 	enum zio_compress	zp_compress;
310 	dmu_object_type_t	zp_type;
311 	uint8_t			zp_level;
312 	uint8_t			zp_copies;
313 	boolean_t		zp_dedup;
314 	boolean_t		zp_dedup_verify;
315 	boolean_t		zp_nopwrite;
316 	uint32_t		zp_zpl_smallblk;
317 } zio_prop_t;
318 
319 typedef struct zio_cksum_report zio_cksum_report_t;
320 
321 typedef void zio_cksum_finish_f(zio_cksum_report_t *rep,
322     const void *good_data);
323 typedef void zio_cksum_free_f(void *cbdata, size_t size);
324 
325 struct zio_bad_cksum;				/* defined in zio_checksum.h */
326 struct dnode_phys;
327 struct abd;
328 
329 struct zio_cksum_report {
330 	struct zio_cksum_report *zcr_next;
331 	nvlist_t		*zcr_ereport;
332 	nvlist_t		*zcr_detector;
333 	void			*zcr_cbdata;
334 	size_t			zcr_cbinfo;	/* passed to zcr_free() */
335 	uint64_t		zcr_align;
336 	uint64_t		zcr_length;
337 	zio_cksum_finish_f	*zcr_finish;
338 	zio_cksum_free_f	*zcr_free;
339 
340 	/* internal use only */
341 	struct zio_bad_cksum	*zcr_ckinfo;	/* information from failure */
342 };
343 
344 typedef void zio_vsd_cksum_report_f(zio_t *zio, zio_cksum_report_t *zcr,
345     void *arg);
346 
347 zio_vsd_cksum_report_f	zio_vsd_default_cksum_report;
348 
349 typedef struct zio_vsd_ops {
350 	zio_done_func_t		*vsd_free;
351 	zio_vsd_cksum_report_f	*vsd_cksum_report;
352 } zio_vsd_ops_t;
353 
354 typedef struct zio_gang_node {
355 	zio_gbh_phys_t		*gn_gbh;
356 	struct zio_gang_node	*gn_child[SPA_GBH_NBLKPTRS];
357 } zio_gang_node_t;
358 
359 typedef zio_t *zio_gang_issue_func_t(zio_t *zio, blkptr_t *bp,
360     zio_gang_node_t *gn, struct abd *data, uint64_t offset);
361 
362 typedef void zio_transform_func_t(zio_t *zio, struct abd *data, uint64_t size);
363 
364 typedef struct zio_transform {
365 	struct abd		*zt_orig_abd;
366 	uint64_t		zt_orig_size;
367 	uint64_t		zt_bufsize;
368 	zio_transform_func_t	*zt_transform;
369 	struct zio_transform	*zt_next;
370 } zio_transform_t;
371 
372 typedef int zio_pipe_stage_t(zio_t *zio);
373 
374 /*
375  * The io_reexecute flags are distinct from io_flags because the child must
376  * be able to propagate them to the parent.  The normal io_flags are local
377  * to the zio, not protected by any lock, and not modifiable by children;
378  * the reexecute flags are protected by io_lock, modifiable by children,
379  * and always propagated -- even when ZIO_FLAG_DONT_PROPAGATE is set.
380  */
381 #define	ZIO_REEXECUTE_NOW	0x01
382 #define	ZIO_REEXECUTE_SUSPEND	0x02
383 
384 typedef struct zio_alloc_list {
385 	list_t  zal_list;
386 	uint64_t zal_size;
387 } zio_alloc_list_t;
388 
389 typedef struct zio_link {
390 	zio_t		*zl_parent;
391 	zio_t		*zl_child;
392 	list_node_t	zl_parent_node;
393 	list_node_t	zl_child_node;
394 } zio_link_t;
395 
396 struct zio {
397 	/* Core information about this I/O */
398 	zbookmark_phys_t	io_bookmark;
399 	zio_prop_t	io_prop;
400 	zio_type_t	io_type;
401 	enum zio_child	io_child_type;
402 	int		io_cmd;
403 	zio_priority_t	io_priority;
404 	uint8_t		io_reexecute;
405 	uint8_t		io_state[ZIO_WAIT_TYPES];
406 	uint64_t	io_txg;
407 	spa_t		*io_spa;
408 	blkptr_t	*io_bp;
409 	blkptr_t	*io_bp_override;
410 	blkptr_t	io_bp_copy;
411 	list_t		io_parent_list;
412 	list_t		io_child_list;
413 	zio_t		*io_logical;
414 	zio_transform_t *io_transform_stack;
415 
416 	/* Callback info */
417 	zio_done_func_t	*io_ready;
418 	zio_done_func_t	*io_children_ready;
419 	zio_done_func_t	*io_physdone;
420 	zio_done_func_t	*io_done;
421 	void		*io_private;
422 	int64_t		io_prev_space_delta;	/* DMU private */
423 	blkptr_t	io_bp_orig;
424 
425 	/* Data represented by this I/O */
426 	struct abd	*io_abd;
427 	struct abd	*io_orig_abd;
428 	uint64_t	io_size;
429 	uint64_t	io_orig_size;
430 	/* io_lsize != io_orig_size iff this is a raw write */
431 	uint64_t	io_lsize;
432 
433 	/* Stuff for the vdev stack */
434 	vdev_t		*io_vd;
435 	void		*io_vsd;
436 	const zio_vsd_ops_t *io_vsd_ops;
437 	metaslab_class_t *io_metaslab_class;	/* dva throttle class */
438 
439 	uint64_t	io_offset;
440 	hrtime_t	io_timestamp;
441 	hrtime_t	io_queued_timestamp;
442 	hrtime_t	io_target_timestamp;
443 	avl_node_t	io_queue_node;
444 	avl_node_t	io_offset_node;
445 	avl_node_t	io_alloc_node;
446 	zio_alloc_list_t	io_alloc_list;
447 
448 	/* Internal pipeline state */
449 	enum zio_flag	io_flags;
450 	enum zio_stage	io_stage;
451 	enum zio_stage	io_pipeline;
452 	enum zio_flag	io_orig_flags;
453 	enum zio_stage	io_orig_stage;
454 	enum zio_stage	io_orig_pipeline;
455 	enum zio_stage	io_pipeline_trace;
456 	int		io_error;
457 	int		io_child_error[ZIO_CHILD_TYPES];
458 	uint64_t	io_children[ZIO_CHILD_TYPES][ZIO_WAIT_TYPES];
459 	uint64_t	io_child_count;
460 	uint64_t	io_phys_children;
461 	uint64_t	io_parent_count;
462 	uint64_t	*io_stall;
463 	zio_t		*io_gang_leader;
464 	zio_gang_node_t	*io_gang_tree;
465 	void		*io_executor;
466 	void		*io_waiter;
467 	kmutex_t	io_lock;
468 	kcondvar_t	io_cv;
469 	int		io_allocator;
470 
471 	/* FMA state */
472 	zio_cksum_report_t *io_cksum_report;
473 	uint64_t	io_ena;
474 
475 	/* Taskq dispatching state */
476 	taskq_ent_t	io_tqent;
477 };
478 
479 extern int zio_bookmark_compare(const void *, const void *);
480 
481 extern zio_t *zio_null(zio_t *pio, spa_t *spa, vdev_t *vd,
482     zio_done_func_t *done, void *private, enum zio_flag flags);
483 
484 extern zio_t *zio_root(spa_t *spa,
485     zio_done_func_t *done, void *private, enum zio_flag flags);
486 
487 extern zio_t *zio_read(zio_t *pio, spa_t *spa, const blkptr_t *bp,
488     struct abd *data, uint64_t lsize, zio_done_func_t *done, void *private,
489     zio_priority_t priority, enum zio_flag flags, const zbookmark_phys_t *zb);
490 
491 extern zio_t *zio_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
492     struct abd *data, uint64_t size, uint64_t psize, const zio_prop_t *zp,
493     zio_done_func_t *ready, zio_done_func_t *children_ready,
494     zio_done_func_t *physdone, zio_done_func_t *done,
495     void *private, zio_priority_t priority, enum zio_flag flags,
496     const zbookmark_phys_t *zb);
497 
498 extern zio_t *zio_rewrite(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
499     struct abd *data, uint64_t size, zio_done_func_t *done, void *private,
500     zio_priority_t priority, enum zio_flag flags, zbookmark_phys_t *zb);
501 
502 extern void zio_write_override(zio_t *zio, blkptr_t *bp, int copies,
503     boolean_t nopwrite);
504 
505 extern void zio_free(spa_t *spa, uint64_t txg, const blkptr_t *bp);
506 
507 extern zio_t *zio_claim(zio_t *pio, spa_t *spa, uint64_t txg,
508     const blkptr_t *bp,
509     zio_done_func_t *done, void *private, enum zio_flag flags);
510 
511 extern zio_t *zio_ioctl(zio_t *pio, spa_t *spa, vdev_t *vd, int cmd,
512     zio_done_func_t *done, void *private, enum zio_flag flags);
513 
514 extern zio_t *zio_read_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
515     uint64_t size, struct abd *data, int checksum,
516     zio_done_func_t *done, void *private, zio_priority_t priority,
517     enum zio_flag flags, boolean_t labels);
518 
519 extern zio_t *zio_write_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
520     uint64_t size, struct abd *data, int checksum,
521     zio_done_func_t *done, void *private, zio_priority_t priority,
522     enum zio_flag flags, boolean_t labels);
523 
524 extern zio_t *zio_free_sync(zio_t *pio, spa_t *spa, uint64_t txg,
525     const blkptr_t *bp, enum zio_flag flags);
526 
527 extern int zio_alloc_zil(spa_t *spa, uint64_t objset, uint64_t txg,
528     blkptr_t *new_bp, blkptr_t *old_bp, uint64_t size, boolean_t *slog);
529 extern void zio_flush(zio_t *zio, vdev_t *vd);
530 extern void zio_shrink(zio_t *zio, uint64_t size);
531 
532 extern int zio_wait(zio_t *zio);
533 extern void zio_nowait(zio_t *zio);
534 extern void zio_execute(zio_t *zio);
535 extern void zio_interrupt(zio_t *zio);
536 extern void zio_delay_init(zio_t *zio);
537 extern void zio_delay_interrupt(zio_t *zio);
538 
539 extern zio_t *zio_walk_parents(zio_t *cio, zio_link_t **);
540 extern zio_t *zio_walk_children(zio_t *pio, zio_link_t **);
541 extern zio_t *zio_unique_parent(zio_t *cio);
542 extern void zio_add_child(zio_t *pio, zio_t *cio);
543 
544 extern void *zio_buf_alloc(size_t size);
545 extern void zio_buf_free(void *buf, size_t size);
546 extern void *zio_data_buf_alloc(size_t size);
547 extern void zio_data_buf_free(void *buf, size_t size);
548 
549 extern void zio_push_transform(zio_t *zio, struct abd *abd, uint64_t size,
550     uint64_t bufsize, zio_transform_func_t *transform);
551 extern void zio_pop_transforms(zio_t *zio);
552 
553 extern void zio_resubmit_stage_async(void *);
554 
555 extern zio_t *zio_vdev_child_io(zio_t *zio, blkptr_t *bp, vdev_t *vd,
556     uint64_t offset, struct abd *data, uint64_t size, int type,
557     zio_priority_t priority, enum zio_flag flags,
558     zio_done_func_t *done, void *private);
559 
560 extern zio_t *zio_vdev_delegated_io(vdev_t *vd, uint64_t offset,
561     struct abd *data, uint64_t size, zio_type_t type, zio_priority_t priority,
562     enum zio_flag flags, zio_done_func_t *done, void *private);
563 
564 extern void zio_vdev_io_bypass(zio_t *zio);
565 extern void zio_vdev_io_reissue(zio_t *zio);
566 extern void zio_vdev_io_redone(zio_t *zio);
567 
568 extern void zio_change_priority(zio_t *pio, zio_priority_t priority);
569 
570 extern void zio_checksum_verified(zio_t *zio);
571 extern int zio_worst_error(int e1, int e2);
572 
573 extern enum zio_checksum zio_checksum_select(enum zio_checksum child,
574     enum zio_checksum parent);
575 extern enum zio_checksum zio_checksum_dedup_select(spa_t *spa,
576     enum zio_checksum child, enum zio_checksum parent);
577 extern enum zio_compress zio_compress_select(spa_t *spa,
578     enum zio_compress child, enum zio_compress parent);
579 
580 extern void zio_suspend(spa_t *spa, zio_t *zio, zio_suspend_reason_t);
581 extern int zio_resume(spa_t *spa);
582 extern void zio_resume_wait(spa_t *spa);
583 
584 /*
585  * Initial setup and teardown.
586  */
587 extern void zio_init(void);
588 extern void zio_fini(void);
589 
590 /*
591  * Fault injection
592  */
593 struct zinject_record;
594 extern uint32_t zio_injection_enabled;
595 extern int zio_inject_fault(char *name, int flags, int *id,
596     struct zinject_record *record);
597 extern int zio_inject_list_next(int *id, char *name, size_t buflen,
598     struct zinject_record *record);
599 extern int zio_clear_fault(int id);
600 extern void zio_handle_panic_injection(spa_t *spa, char *tag, uint64_t type);
601 extern int zio_handle_fault_injection(zio_t *zio, int error);
602 extern int zio_handle_device_injection(vdev_t *vd, zio_t *zio, int error);
603 extern int zio_handle_label_injection(zio_t *zio, int error);
604 extern void zio_handle_ignored_writes(zio_t *zio);
605 extern hrtime_t zio_handle_io_delay(zio_t *zio);
606 
607 /*
608  * Checksum ereport functions
609  */
610 extern void zfs_ereport_start_checksum(spa_t *spa, vdev_t *vd, struct zio *zio,
611     uint64_t offset, uint64_t length, void *arg, struct zio_bad_cksum *info);
612 extern void zfs_ereport_finish_checksum(zio_cksum_report_t *report,
613     const void *good_data, const void *bad_data, boolean_t drop_if_identical);
614 
615 extern void zfs_ereport_send_interim_checksum(zio_cksum_report_t *report);
616 extern void zfs_ereport_free_checksum(zio_cksum_report_t *report);
617 
618 /* If we have the good data in hand, this function can be used */
619 extern void zfs_ereport_post_checksum(spa_t *spa, vdev_t *vd,
620     struct zio *zio, uint64_t offset, uint64_t length,
621     const void *good_data, const void *bad_data, struct zio_bad_cksum *info);
622 
623 /* Called from spa_sync(), but primarily an injection handler */
624 extern void spa_handle_ignored_writes(spa_t *spa);
625 
626 /* zbookmark_phys functions */
627 boolean_t zbookmark_subtree_completed(const struct dnode_phys *dnp,
628     const zbookmark_phys_t *subtree_root, const zbookmark_phys_t *last_block);
629 int zbookmark_compare(uint16_t dbss1, uint8_t ibs1, uint16_t dbss2,
630     uint8_t ibs2, const zbookmark_phys_t *zb1, const zbookmark_phys_t *zb2);
631 
632 #ifdef	__cplusplus
633 }
634 #endif
635 
636 #endif	/* _ZIO_H */
637