xref: /illumos-gate/usr/src/uts/common/fs/zfs/sys/zio.h (revision d48be21240dfd051b689384ce2b23479d757f2d8)
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 2019, Joyent, Inc.
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 #define	ZIO_CRYPT_ON_VALUE	ZIO_CRYPT_AES_256_CCM
108 #define	ZIO_CRYPT_DEFAULT	ZIO_CRYPT_OFF
109 
110 /* macros defining encryption lengths */
111 #define	ZIO_OBJSET_MAC_LEN		32
112 #define	ZIO_DATA_IV_LEN			12
113 #define	ZIO_DATA_SALT_LEN		8
114 #define	ZIO_DATA_MAC_LEN		16
115 
116 /*
117  * The number of "legacy" compression functions which can be set on individual
118  * objects.
119  */
120 #define	ZIO_COMPRESS_LEGACY_FUNCTIONS ZIO_COMPRESS_LZ4
121 
122 /*
123  * The meaning of "compress = on" selected by the compression features enabled
124  * on a given pool.
125  */
126 #define	ZIO_COMPRESS_LEGACY_ON_VALUE	ZIO_COMPRESS_LZJB
127 #define	ZIO_COMPRESS_LZ4_ON_VALUE	ZIO_COMPRESS_LZ4
128 
129 #define	ZIO_COMPRESS_DEFAULT		ZIO_COMPRESS_OFF
130 
131 #define	BOOTFS_COMPRESS_VALID(compress)			\
132 	((compress) == ZIO_COMPRESS_LZJB ||		\
133 	(compress) == ZIO_COMPRESS_LZ4 ||		\
134 	(compress) == ZIO_COMPRESS_GZIP_1 ||		\
135 	(compress) == ZIO_COMPRESS_GZIP_2 ||		\
136 	(compress) == ZIO_COMPRESS_GZIP_3 ||		\
137 	(compress) == ZIO_COMPRESS_GZIP_4 ||		\
138 	(compress) == ZIO_COMPRESS_GZIP_5 ||		\
139 	(compress) == ZIO_COMPRESS_GZIP_6 ||		\
140 	(compress) == ZIO_COMPRESS_GZIP_7 ||		\
141 	(compress) == ZIO_COMPRESS_GZIP_8 ||		\
142 	(compress) == ZIO_COMPRESS_GZIP_9 ||		\
143 	(compress) == ZIO_COMPRESS_ON ||		\
144 	(compress) == ZIO_COMPRESS_OFF)
145 
146 #define	ZIO_FAILURE_MODE_WAIT		0
147 #define	ZIO_FAILURE_MODE_CONTINUE	1
148 #define	ZIO_FAILURE_MODE_PANIC		2
149 
150 typedef enum zio_suspend_reason {
151 	ZIO_SUSPEND_NONE = 0,
152 	ZIO_SUSPEND_IOERR,
153 	ZIO_SUSPEND_MMP,
154 } zio_suspend_reason_t;
155 
156 enum zio_flag {
157 	/*
158 	 * Flags inherited by gang, ddt, and vdev children,
159 	 * and that must be equal for two zios to aggregate
160 	 */
161 	ZIO_FLAG_DONT_AGGREGATE	= 1 << 0,
162 	ZIO_FLAG_IO_REPAIR	= 1 << 1,
163 	ZIO_FLAG_SELF_HEAL	= 1 << 2,
164 	ZIO_FLAG_RESILVER	= 1 << 3,
165 	ZIO_FLAG_SCRUB		= 1 << 4,
166 	ZIO_FLAG_SCAN_THREAD	= 1 << 5,
167 	ZIO_FLAG_PHYSICAL	= 1 << 6,
168 
169 #define	ZIO_FLAG_AGG_INHERIT	(ZIO_FLAG_CANFAIL - 1)
170 
171 	/*
172 	 * Flags inherited by ddt, gang, and vdev children.
173 	 */
174 	ZIO_FLAG_CANFAIL	= 1 << 7,	/* must be first for INHERIT */
175 	ZIO_FLAG_SPECULATIVE	= 1 << 8,
176 	ZIO_FLAG_CONFIG_WRITER	= 1 << 9,
177 	ZIO_FLAG_DONT_RETRY	= 1 << 10,
178 	ZIO_FLAG_DONT_CACHE	= 1 << 11,
179 	ZIO_FLAG_NODATA		= 1 << 12,
180 	ZIO_FLAG_INDUCE_DAMAGE	= 1 << 13,
181 	ZIO_FLAG_IO_ALLOCATING	= 1 << 14,
182 
183 #define	ZIO_FLAG_DDT_INHERIT	(ZIO_FLAG_IO_RETRY - 1)
184 #define	ZIO_FLAG_GANG_INHERIT	(ZIO_FLAG_IO_RETRY - 1)
185 
186 	/*
187 	 * Flags inherited by vdev children.
188 	 */
189 	ZIO_FLAG_IO_RETRY	= 1 << 15,	/* must be first for INHERIT */
190 	ZIO_FLAG_PROBE		= 1 << 16,
191 	ZIO_FLAG_TRYHARD	= 1 << 17,
192 	ZIO_FLAG_OPTIONAL	= 1 << 18,
193 
194 #define	ZIO_FLAG_VDEV_INHERIT	(ZIO_FLAG_DONT_QUEUE - 1)
195 
196 	/*
197 	 * Flags not inherited by any children.
198 	 */
199 	ZIO_FLAG_DONT_QUEUE	= 1 << 19,	/* must be first for INHERIT */
200 	ZIO_FLAG_DONT_PROPAGATE	= 1 << 20,
201 	ZIO_FLAG_IO_BYPASS	= 1 << 21,
202 	ZIO_FLAG_IO_REWRITE	= 1 << 22,
203 	ZIO_FLAG_RAW_COMPRESS	= 1 << 23,
204 	ZIO_FLAG_RAW_ENCRYPT	= 1 << 24,
205 	ZIO_FLAG_GANG_CHILD	= 1 << 25,
206 	ZIO_FLAG_DDT_CHILD	= 1 << 26,
207 	ZIO_FLAG_GODFATHER	= 1 << 27,
208 	ZIO_FLAG_NOPWRITE	= 1 << 28,
209 	ZIO_FLAG_REEXECUTED	= 1 << 29,
210 	ZIO_FLAG_DELEGATED	= 1 << 30,
211 };
212 
213 #define	ZIO_FLAG_MUSTSUCCEED		0
214 #define	ZIO_FLAG_RAW	(ZIO_FLAG_RAW_COMPRESS | ZIO_FLAG_RAW_ENCRYPT)
215 
216 #define	ZIO_DDT_CHILD_FLAGS(zio)				\
217 	(((zio)->io_flags & ZIO_FLAG_DDT_INHERIT) |		\
218 	ZIO_FLAG_DDT_CHILD | ZIO_FLAG_CANFAIL)
219 
220 #define	ZIO_GANG_CHILD_FLAGS(zio)				\
221 	(((zio)->io_flags & ZIO_FLAG_GANG_INHERIT) |		\
222 	ZIO_FLAG_GANG_CHILD | ZIO_FLAG_CANFAIL)
223 
224 #define	ZIO_VDEV_CHILD_FLAGS(zio)				\
225 	(((zio)->io_flags & ZIO_FLAG_VDEV_INHERIT) |		\
226 	ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_CANFAIL)
227 
228 #define	ZIO_CHILD_BIT(x)		(1 << (x))
229 #define	ZIO_CHILD_BIT_IS_SET(val, x)	((val) & (1 << (x)))
230 
231 enum zio_child {
232 	ZIO_CHILD_VDEV = 0,
233 	ZIO_CHILD_GANG,
234 	ZIO_CHILD_DDT,
235 	ZIO_CHILD_LOGICAL,
236 	ZIO_CHILD_TYPES
237 };
238 
239 #define	ZIO_CHILD_VDEV_BIT		ZIO_CHILD_BIT(ZIO_CHILD_VDEV)
240 #define	ZIO_CHILD_GANG_BIT		ZIO_CHILD_BIT(ZIO_CHILD_GANG)
241 #define	ZIO_CHILD_DDT_BIT		ZIO_CHILD_BIT(ZIO_CHILD_DDT)
242 #define	ZIO_CHILD_LOGICAL_BIT		ZIO_CHILD_BIT(ZIO_CHILD_LOGICAL)
243 #define	ZIO_CHILD_ALL_BITS					\
244 	(ZIO_CHILD_VDEV_BIT | ZIO_CHILD_GANG_BIT |		\
245 	ZIO_CHILD_DDT_BIT | ZIO_CHILD_LOGICAL_BIT)
246 
247 enum zio_wait_type {
248 	ZIO_WAIT_READY = 0,
249 	ZIO_WAIT_DONE,
250 	ZIO_WAIT_TYPES
251 };
252 
253 /*
254  * We'll take the unused errnos, 'EBADE' and 'EBADR' (from the Convergent
255  * graveyard) to indicate checksum errors and fragmentation.
256  */
257 #define	ECKSUM	EBADE
258 #define	EFRAGS	EBADR
259 
260 typedef void zio_done_func_t(zio_t *zio);
261 
262 extern boolean_t zio_dva_throttle_enabled;
263 extern const char *zio_type_name[ZIO_TYPES];
264 
265 /*
266  * A bookmark is a four-tuple <objset, object, level, blkid> that uniquely
267  * identifies any block in the pool.  By convention, the meta-objset (MOS)
268  * is objset 0, and the meta-dnode is object 0.  This covers all blocks
269  * except root blocks and ZIL blocks, which are defined as follows:
270  *
271  * Root blocks (objset_phys_t) are object 0, level -1:  <objset, 0, -1, 0>.
272  * ZIL blocks are bookmarked <objset, 0, -2, blkid == ZIL sequence number>.
273  * dmu_sync()ed ZIL data blocks are bookmarked <objset, object, -2, blkid>.
274  * dnode visit bookmarks are <objset, object id of dnode, -3, 0>.
275  *
276  * Note: this structure is called a bookmark because its original purpose
277  * was to remember where to resume a pool-wide traverse.
278  *
279  * Note: this structure is passed between userland and the kernel, and is
280  * stored on disk (by virtue of being incorporated into other on-disk
281  * structures, e.g. dsl_scan_phys_t).
282  */
283 typedef struct zbookmark_phys {
284 	uint64_t	zb_objset;
285 	uint64_t	zb_object;
286 	int64_t		zb_level;
287 	uint64_t	zb_blkid;
288 } zbookmark_phys_t;
289 
290 #define	SET_BOOKMARK(zb, objset, object, level, blkid)  \
291 {                                                       \
292 	(zb)->zb_objset = objset;                       \
293 	(zb)->zb_object = object;                       \
294 	(zb)->zb_level = level;                         \
295 	(zb)->zb_blkid = blkid;                         \
296 }
297 
298 #define	ZB_DESTROYED_OBJSET	(-1ULL)
299 
300 #define	ZB_ROOT_OBJECT		(0ULL)
301 #define	ZB_ROOT_LEVEL		(-1LL)
302 #define	ZB_ROOT_BLKID		(0ULL)
303 
304 #define	ZB_ZIL_OBJECT		(0ULL)
305 #define	ZB_ZIL_LEVEL		(-2LL)
306 
307 #define	ZB_DNODE_LEVEL		(-3LL)
308 #define	ZB_DNODE_BLKID		(0ULL)
309 
310 #define	ZB_IS_ZERO(zb)						\
311 	((zb)->zb_objset == 0 && (zb)->zb_object == 0 &&	\
312 	(zb)->zb_level == 0 && (zb)->zb_blkid == 0)
313 #define	ZB_IS_ROOT(zb)				\
314 	((zb)->zb_object == ZB_ROOT_OBJECT &&	\
315 	(zb)->zb_level == ZB_ROOT_LEVEL &&	\
316 	(zb)->zb_blkid == ZB_ROOT_BLKID)
317 
318 typedef struct zio_prop {
319 	enum zio_checksum	zp_checksum;
320 	enum zio_compress	zp_compress;
321 	dmu_object_type_t	zp_type;
322 	uint8_t			zp_level;
323 	uint8_t			zp_copies;
324 	boolean_t		zp_dedup;
325 	boolean_t		zp_dedup_verify;
326 	boolean_t		zp_nopwrite;
327 	uint32_t		zp_zpl_smallblk;
328 	boolean_t		zp_encrypt;
329 	boolean_t		zp_byteorder;
330 	uint8_t			zp_salt[ZIO_DATA_SALT_LEN];
331 	uint8_t			zp_iv[ZIO_DATA_IV_LEN];
332 	uint8_t			zp_mac[ZIO_DATA_MAC_LEN];
333 } zio_prop_t;
334 
335 typedef struct zio_cksum_report zio_cksum_report_t;
336 
337 typedef void zio_cksum_finish_f(zio_cksum_report_t *rep,
338     const abd_t *good_data);
339 typedef void zio_cksum_free_f(void *cbdata, size_t size);
340 
341 struct zio_bad_cksum;				/* defined in zio_checksum.h */
342 struct dnode_phys;
343 struct abd;
344 
345 struct zio_cksum_report {
346 	struct zio_cksum_report *zcr_next;
347 	nvlist_t		*zcr_ereport;
348 	nvlist_t		*zcr_detector;
349 	void			*zcr_cbdata;
350 	size_t			zcr_cbinfo;	/* passed to zcr_free() */
351 	uint64_t		zcr_align;
352 	uint64_t		zcr_length;
353 	zio_cksum_finish_f	*zcr_finish;
354 	zio_cksum_free_f	*zcr_free;
355 
356 	/* internal use only */
357 	struct zio_bad_cksum	*zcr_ckinfo;	/* information from failure */
358 };
359 
360 typedef void zio_vsd_cksum_report_f(zio_t *zio, zio_cksum_report_t *zcr,
361     void *arg);
362 
363 zio_vsd_cksum_report_f	zio_vsd_default_cksum_report;
364 
365 typedef struct zio_vsd_ops {
366 	zio_done_func_t		*vsd_free;
367 	zio_vsd_cksum_report_f	*vsd_cksum_report;
368 } zio_vsd_ops_t;
369 
370 typedef struct zio_gang_node {
371 	zio_gbh_phys_t		*gn_gbh;
372 	struct zio_gang_node	*gn_child[SPA_GBH_NBLKPTRS];
373 } zio_gang_node_t;
374 
375 typedef zio_t *zio_gang_issue_func_t(zio_t *zio, blkptr_t *bp,
376     zio_gang_node_t *gn, struct abd *data, uint64_t offset);
377 
378 typedef void zio_transform_func_t(zio_t *zio, struct abd *data, uint64_t size);
379 
380 typedef struct zio_transform {
381 	struct abd		*zt_orig_abd;
382 	uint64_t		zt_orig_size;
383 	uint64_t		zt_bufsize;
384 	zio_transform_func_t	*zt_transform;
385 	struct zio_transform	*zt_next;
386 } zio_transform_t;
387 
388 typedef int zio_pipe_stage_t(zio_t *zio);
389 
390 /*
391  * The io_reexecute flags are distinct from io_flags because the child must
392  * be able to propagate them to the parent.  The normal io_flags are local
393  * to the zio, not protected by any lock, and not modifiable by children;
394  * the reexecute flags are protected by io_lock, modifiable by children,
395  * and always propagated -- even when ZIO_FLAG_DONT_PROPAGATE is set.
396  */
397 #define	ZIO_REEXECUTE_NOW	0x01
398 #define	ZIO_REEXECUTE_SUSPEND	0x02
399 
400 /*
401  * The io_trim flags are used to specify the type of TRIM to perform.  They
402  * only apply to ZIO_TYPE_TRIM zios and are distinct from io_flags.
403  */
404 enum trim_flag {
405 	ZIO_TRIM_SECURE		= 1 << 0,
406 };
407 
408 typedef struct zio_alloc_list {
409 	list_t  zal_list;
410 	uint64_t zal_size;
411 } zio_alloc_list_t;
412 
413 typedef struct zio_link {
414 	zio_t		*zl_parent;
415 	zio_t		*zl_child;
416 	list_node_t	zl_parent_node;
417 	list_node_t	zl_child_node;
418 } zio_link_t;
419 
420 struct zio {
421 	/* Core information about this I/O */
422 	zbookmark_phys_t	io_bookmark;
423 	zio_prop_t	io_prop;
424 	zio_type_t	io_type;
425 	enum zio_child	io_child_type;
426 	enum trim_flag	io_trim_flags;
427 	int		io_cmd;
428 	zio_priority_t	io_priority;
429 	uint8_t		io_reexecute;
430 	uint8_t		io_state[ZIO_WAIT_TYPES];
431 	uint64_t	io_txg;
432 	spa_t		*io_spa;
433 	blkptr_t	*io_bp;
434 	blkptr_t	*io_bp_override;
435 	blkptr_t	io_bp_copy;
436 	list_t		io_parent_list;
437 	list_t		io_child_list;
438 	zio_t		*io_logical;
439 	zio_transform_t *io_transform_stack;
440 
441 	/* Callback info */
442 	zio_done_func_t	*io_ready;
443 	zio_done_func_t	*io_children_ready;
444 	zio_done_func_t	*io_physdone;
445 	zio_done_func_t	*io_done;
446 	void		*io_private;
447 	int64_t		io_prev_space_delta;	/* DMU private */
448 	blkptr_t	io_bp_orig;
449 
450 	/* Data represented by this I/O */
451 	struct abd	*io_abd;
452 	struct abd	*io_orig_abd;
453 	uint64_t	io_size;
454 	uint64_t	io_orig_size;
455 	/* io_lsize != io_orig_size iff this is a raw write */
456 	uint64_t	io_lsize;
457 
458 	/* Stuff for the vdev stack */
459 	vdev_t		*io_vd;
460 	void		*io_vsd;
461 	const zio_vsd_ops_t *io_vsd_ops;
462 	metaslab_class_t *io_metaslab_class;	/* dva throttle class */
463 
464 	uint64_t	io_offset;
465 	hrtime_t	io_timestamp;
466 	hrtime_t	io_queued_timestamp;
467 	hrtime_t	io_target_timestamp;
468 	hrtime_t	io_delta;	/* vdev queue service delta */
469 	hrtime_t	io_delay;	/* Device access time (disk or */
470 					/* file). */
471 	avl_node_t	io_queue_node;
472 	avl_node_t	io_offset_node;
473 	avl_node_t	io_alloc_node;
474 	zio_alloc_list_t	io_alloc_list;
475 
476 	/* Internal pipeline state */
477 	enum zio_flag	io_flags;
478 	enum zio_stage	io_stage;
479 	enum zio_stage	io_pipeline;
480 	enum zio_flag	io_orig_flags;
481 	enum zio_stage	io_orig_stage;
482 	enum zio_stage	io_orig_pipeline;
483 	enum zio_stage	io_pipeline_trace;
484 	int		io_error;
485 	int		io_child_error[ZIO_CHILD_TYPES];
486 	uint64_t	io_children[ZIO_CHILD_TYPES][ZIO_WAIT_TYPES];
487 	uint64_t	io_child_count;
488 	uint64_t	io_phys_children;
489 	uint64_t	io_parent_count;
490 	uint64_t	*io_stall;
491 	zio_t		*io_gang_leader;
492 	zio_gang_node_t	*io_gang_tree;
493 	void		*io_executor;
494 	void		*io_waiter;
495 	kmutex_t	io_lock;
496 	kcondvar_t	io_cv;
497 	int		io_allocator;
498 
499 	/* FMA state */
500 	zio_cksum_report_t *io_cksum_report;
501 	uint64_t	io_ena;
502 
503 	/* Taskq dispatching state */
504 	taskq_ent_t	io_tqent;
505 };
506 
507 extern int zio_bookmark_compare(const void *, const void *);
508 
509 extern zio_t *zio_null(zio_t *pio, spa_t *spa, vdev_t *vd,
510     zio_done_func_t *done, void *private, enum zio_flag flags);
511 
512 extern zio_t *zio_root(spa_t *spa,
513     zio_done_func_t *done, void *private, enum zio_flag flags);
514 
515 extern zio_t *zio_read(zio_t *pio, spa_t *spa, const blkptr_t *bp,
516     struct abd *data, uint64_t lsize, zio_done_func_t *done, void *private,
517     zio_priority_t priority, enum zio_flag flags, const zbookmark_phys_t *zb);
518 
519 extern zio_t *zio_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
520     struct abd *data, uint64_t size, uint64_t psize, const zio_prop_t *zp,
521     zio_done_func_t *ready, zio_done_func_t *children_ready,
522     zio_done_func_t *physdone, zio_done_func_t *done,
523     void *private, zio_priority_t priority, enum zio_flag flags,
524     const zbookmark_phys_t *zb);
525 
526 extern zio_t *zio_rewrite(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
527     struct abd *data, uint64_t size, zio_done_func_t *done, void *private,
528     zio_priority_t priority, enum zio_flag flags, zbookmark_phys_t *zb);
529 
530 extern void zio_write_override(zio_t *zio, blkptr_t *bp, int copies,
531     boolean_t nopwrite);
532 
533 extern void zio_free(spa_t *spa, uint64_t txg, const blkptr_t *bp);
534 
535 extern zio_t *zio_claim(zio_t *pio, spa_t *spa, uint64_t txg,
536     const blkptr_t *bp,
537     zio_done_func_t *done, void *private, enum zio_flag flags);
538 
539 extern zio_t *zio_ioctl(zio_t *pio, spa_t *spa, vdev_t *vd, int cmd,
540     zio_done_func_t *done, void *private, enum zio_flag flags);
541 
542 extern zio_t *zio_trim(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size,
543     zio_done_func_t *done, void *private, zio_priority_t priority,
544     enum zio_flag flags, enum trim_flag trim_flags);
545 
546 extern zio_t *zio_read_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
547     uint64_t size, struct abd *data, int checksum,
548     zio_done_func_t *done, void *private, zio_priority_t priority,
549     enum zio_flag flags, boolean_t labels);
550 
551 extern zio_t *zio_write_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
552     uint64_t size, struct abd *data, int checksum,
553     zio_done_func_t *done, void *private, zio_priority_t priority,
554     enum zio_flag flags, boolean_t labels);
555 
556 extern zio_t *zio_free_sync(zio_t *pio, spa_t *spa, uint64_t txg,
557     const blkptr_t *bp, enum zio_flag flags);
558 
559 extern int zio_alloc_zil(spa_t *spa, objset_t *os, uint64_t txg,
560     blkptr_t *new_bp, blkptr_t *old_bp, uint64_t size, boolean_t *slog);
561 extern void zio_free_zil(spa_t *spa, uint64_t txg, blkptr_t *bp);
562 extern void zio_flush(zio_t *zio, vdev_t *vd);
563 extern void zio_shrink(zio_t *zio, uint64_t size);
564 
565 extern int zio_wait(zio_t *zio);
566 extern void zio_nowait(zio_t *zio);
567 extern void zio_execute(zio_t *zio);
568 extern void zio_interrupt(zio_t *zio);
569 extern void zio_delay_init(zio_t *zio);
570 extern void zio_delay_interrupt(zio_t *zio);
571 
572 extern zio_t *zio_walk_parents(zio_t *cio, zio_link_t **);
573 extern zio_t *zio_walk_children(zio_t *pio, zio_link_t **);
574 extern zio_t *zio_unique_parent(zio_t *cio);
575 extern void zio_add_child(zio_t *pio, zio_t *cio);
576 
577 extern void *zio_buf_alloc(size_t size);
578 extern void zio_buf_free(void *buf, size_t size);
579 extern void *zio_data_buf_alloc(size_t size);
580 extern void zio_data_buf_free(void *buf, size_t size);
581 
582 extern void zio_push_transform(zio_t *zio, struct abd *abd, uint64_t size,
583     uint64_t bufsize, zio_transform_func_t *transform);
584 extern void zio_pop_transforms(zio_t *zio);
585 
586 extern void zio_resubmit_stage_async(void *);
587 
588 extern zio_t *zio_vdev_child_io(zio_t *zio, blkptr_t *bp, vdev_t *vd,
589     uint64_t offset, struct abd *data, uint64_t size, int type,
590     zio_priority_t priority, enum zio_flag flags,
591     zio_done_func_t *done, void *private);
592 
593 extern zio_t *zio_vdev_delegated_io(vdev_t *vd, uint64_t offset,
594     struct abd *data, uint64_t size, zio_type_t type, zio_priority_t priority,
595     enum zio_flag flags, zio_done_func_t *done, void *private);
596 
597 extern void zio_vdev_io_bypass(zio_t *zio);
598 extern void zio_vdev_io_reissue(zio_t *zio);
599 extern void zio_vdev_io_redone(zio_t *zio);
600 
601 extern void zio_change_priority(zio_t *pio, zio_priority_t priority);
602 
603 extern void zio_checksum_verified(zio_t *zio);
604 extern int zio_worst_error(int e1, int e2);
605 
606 extern enum zio_checksum zio_checksum_select(enum zio_checksum child,
607     enum zio_checksum parent);
608 extern enum zio_checksum zio_checksum_dedup_select(spa_t *spa,
609     enum zio_checksum child, enum zio_checksum parent);
610 extern enum zio_compress zio_compress_select(spa_t *spa,
611     enum zio_compress child, enum zio_compress parent);
612 
613 extern void zio_suspend(spa_t *spa, zio_t *zio, zio_suspend_reason_t);
614 extern int zio_resume(spa_t *spa);
615 extern void zio_resume_wait(spa_t *spa);
616 
617 /*
618  * Initial setup and teardown.
619  */
620 extern void zio_init(void);
621 extern void zio_fini(void);
622 
623 /*
624  * Fault injection
625  */
626 struct zinject_record;
627 extern uint32_t zio_injection_enabled;
628 extern int zio_inject_fault(char *name, int flags, int *id,
629     struct zinject_record *record);
630 extern int zio_inject_list_next(int *id, char *name, size_t buflen,
631     struct zinject_record *record);
632 extern int zio_clear_fault(int id);
633 extern void zio_handle_panic_injection(spa_t *spa, char *tag, uint64_t type);
634 extern int zio_handle_decrypt_injection(spa_t *spa, const zbookmark_phys_t *zb,
635     uint64_t type, int error);
636 extern int zio_handle_fault_injection(zio_t *zio, int error);
637 extern int zio_handle_device_injection(vdev_t *vd, zio_t *zio, int error);
638 extern int zio_handle_label_injection(zio_t *zio, int error);
639 extern void zio_handle_ignored_writes(zio_t *zio);
640 extern hrtime_t zio_handle_io_delay(zio_t *zio);
641 
642 /*
643  * Checksum ereport functions
644  */
645 extern void zfs_ereport_start_checksum(spa_t *spa, vdev_t *vd,
646     const zbookmark_phys_t *zb, struct zio *zio, uint64_t offset,
647     uint64_t length, void *arg, struct zio_bad_cksum *info);
648 extern void zfs_ereport_finish_checksum(zio_cksum_report_t *report,
649     const abd_t *good_data, const abd_t *bad_data, boolean_t drop_if_identical);
650 
651 extern void zfs_ereport_send_interim_checksum(zio_cksum_report_t *report);
652 extern void zfs_ereport_free_checksum(zio_cksum_report_t *report);
653 
654 /* If we have the good data in hand, this function can be used */
655 extern int zfs_ereport_post_checksum(spa_t *spa, vdev_t *vd,
656     const zbookmark_phys_t *zb, struct zio *zio, uint64_t offset,
657     uint64_t length, const abd_t *good_data, const abd_t *bad_data,
658     struct zio_bad_cksum *info);
659 
660 /* Called from spa_sync(), but primarily an injection handler */
661 extern void spa_handle_ignored_writes(spa_t *spa);
662 
663 /* zbookmark_phys functions */
664 boolean_t zbookmark_subtree_completed(const struct dnode_phys *dnp,
665     const zbookmark_phys_t *subtree_root, const zbookmark_phys_t *last_block);
666 int zbookmark_compare(uint16_t dbss1, uint8_t ibs1, uint16_t dbss2,
667     uint8_t ibs2, const zbookmark_phys_t *zb1, const zbookmark_phys_t *zb2);
668 
669 #ifdef	__cplusplus
670 }
671 #endif
672 
673 #endif	/* _ZIO_H */
674