xref: /freebsd/sys/contrib/openzfs/include/sys/zio.h (revision 2d76470b7011d0c2f50ae394972976fbca14bc95)
1 // SPDX-License-Identifier: CDDL-1.0
2 /*
3  * CDDL HEADER START
4  *
5  * The contents of this file are subject to the terms of the
6  * Common Development and Distribution License (the "License").
7  * You may not use this file except in compliance with the License.
8  *
9  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
10  * or https://opensource.org/licenses/CDDL-1.0.
11  * See the License for the specific language governing permissions
12  * and limitations under the License.
13  *
14  * When distributing Covered Code, include this CDDL HEADER in each
15  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
16  * If applicable, add the following below this CDDL HEADER, with the
17  * fields enclosed by brackets "[]" replaced with your own identifying
18  * information: Portions Copyright [yyyy] [name of copyright owner]
19  *
20  * CDDL HEADER END
21  */
22 
23 /*
24  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
25  * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
26  * Copyright (c) 2012, 2024 by Delphix. All rights reserved.
27  * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
28  * Copyright (c) 2013, Joyent, Inc. All rights reserved.
29  * Copyright 2016 Toomas Soome <tsoome@me.com>
30  * Copyright (c) 2019, Allan Jude
31  * Copyright (c) 2019, 2023, 2024, Klara Inc.
32  * Copyright (c) 2019-2020, Michael Niewöhner
33  * Copyright (c) 2024 by George Melikov. All rights reserved.
34  */
35 
36 #ifndef _ZIO_H
37 #define	_ZIO_H
38 
39 #include <sys/zfs_context.h>
40 #include <sys/spa.h>
41 #include <sys/txg.h>
42 #include <sys/avl.h>
43 #include <sys/fs/zfs.h>
44 #include <sys/zio_impl.h>
45 
46 #ifdef	__cplusplus
47 extern "C" {
48 #endif
49 
50 /*
51  * Embedded checksum
52  */
53 #define	ZEC_MAGIC	0x210da7ab10c7a11ULL
54 
55 typedef struct zio_eck {
56 	uint64_t	zec_magic;	/* for validation, endianness	*/
57 	zio_cksum_t	zec_cksum;	/* 256-bit checksum		*/
58 } zio_eck_t;
59 
60 /*
61  * Gang block headers are self-checksumming and contain an array
62  * of block pointers. The old gang block size has enough room for 3 blkptrs,
63  * while new gang blocks can store more.
64  *
65  * Layout:
66  * +--------+--------+--------+-----+---------+-----------+
67  * |        |        |        |     |         |           |
68  * | blkptr | blkptr | blkptr | ... | padding | zio_eck_t |
69  * |   1    |   2    |   3    |     |         |           |
70  * +--------+--------+--------+-----+---------+-----------+
71  *   128B     128B     128B             88B        40B
72  */
73 #define	SPA_OLD_GANGBLOCKSIZE	SPA_MINBLOCKSIZE
74 typedef void zio_gbh_phys_t;
75 
76 static inline uint64_t
gbh_nblkptrs(uint64_t size)77 gbh_nblkptrs(uint64_t size) {
78 	ASSERT(IS_P2ALIGNED(size, sizeof (blkptr_t)));
79 	return ((size - sizeof (zio_eck_t)) / sizeof (blkptr_t));
80 }
81 
82 static inline zio_eck_t *
gbh_eck(zio_gbh_phys_t * gbh,uint64_t size)83 gbh_eck(zio_gbh_phys_t *gbh, uint64_t size) {
84 	ASSERT(IS_P2ALIGNED(size, sizeof (blkptr_t)));
85 	return ((zio_eck_t *)((uintptr_t)gbh + (size_t)size - sizeof (zio_eck_t)));
86 }
87 
88 static inline blkptr_t *
gbh_bp(zio_gbh_phys_t * gbh,int bp)89 gbh_bp(zio_gbh_phys_t *gbh, int bp) {
90 	return (&((blkptr_t *)gbh)[bp]);
91 }
92 
93 enum zio_checksum {
94 	ZIO_CHECKSUM_INHERIT = 0,
95 	ZIO_CHECKSUM_ON,
96 	ZIO_CHECKSUM_OFF,
97 	ZIO_CHECKSUM_LABEL,
98 	ZIO_CHECKSUM_GANG_HEADER,
99 	ZIO_CHECKSUM_ZILOG,
100 	ZIO_CHECKSUM_FLETCHER_2,
101 	ZIO_CHECKSUM_FLETCHER_4,
102 	ZIO_CHECKSUM_SHA256,
103 	ZIO_CHECKSUM_ZILOG2,
104 	ZIO_CHECKSUM_NOPARITY,
105 	ZIO_CHECKSUM_SHA512,
106 	ZIO_CHECKSUM_SKEIN,
107 	ZIO_CHECKSUM_EDONR,
108 	ZIO_CHECKSUM_BLAKE3,
109 	ZIO_CHECKSUM_FUNCTIONS
110 };
111 
112 /*
113  * The number of "legacy" compression functions which can be set on individual
114  * objects.
115  */
116 #define	ZIO_CHECKSUM_LEGACY_FUNCTIONS ZIO_CHECKSUM_ZILOG2
117 
118 #define	ZIO_CHECKSUM_ON_VALUE	ZIO_CHECKSUM_FLETCHER_4
119 #define	ZIO_CHECKSUM_DEFAULT	ZIO_CHECKSUM_ON
120 
121 #define	ZIO_CHECKSUM_MASK	0xffULL
122 #define	ZIO_CHECKSUM_VERIFY	(1U << 8)
123 
124 #define	ZIO_DEDUPCHECKSUM	ZIO_CHECKSUM_SHA256
125 
126 /* macros defining encryption lengths */
127 #define	ZIO_OBJSET_MAC_LEN		32
128 #define	ZIO_DATA_IV_LEN			12
129 #define	ZIO_DATA_SALT_LEN		8
130 #define	ZIO_DATA_MAC_LEN		16
131 
132 /*
133  * The number of "legacy" compression functions which can be set on individual
134  * objects.
135  */
136 #define	ZIO_COMPRESS_LEGACY_FUNCTIONS ZIO_COMPRESS_LZ4
137 
138 /*
139  * The meaning of "compress = on" selected by the compression features enabled
140  * on a given pool.
141  */
142 #define	ZIO_COMPRESS_LEGACY_ON_VALUE	ZIO_COMPRESS_LZJB
143 #define	ZIO_COMPRESS_LZ4_ON_VALUE	ZIO_COMPRESS_LZ4
144 
145 #define	ZIO_COMPRESS_DEFAULT		ZIO_COMPRESS_ON
146 
147 #define	BOOTFS_COMPRESS_VALID(compress)			\
148 	((compress) == ZIO_COMPRESS_LZJB ||		\
149 	(compress) == ZIO_COMPRESS_LZ4 ||		\
150 	(compress) == ZIO_COMPRESS_GZIP_1 ||		\
151 	(compress) == ZIO_COMPRESS_GZIP_2 ||		\
152 	(compress) == ZIO_COMPRESS_GZIP_3 ||		\
153 	(compress) == ZIO_COMPRESS_GZIP_4 ||		\
154 	(compress) == ZIO_COMPRESS_GZIP_5 ||		\
155 	(compress) == ZIO_COMPRESS_GZIP_6 ||		\
156 	(compress) == ZIO_COMPRESS_GZIP_7 ||		\
157 	(compress) == ZIO_COMPRESS_GZIP_8 ||		\
158 	(compress) == ZIO_COMPRESS_GZIP_9 ||		\
159 	(compress) == ZIO_COMPRESS_ZLE ||		\
160 	(compress) == ZIO_COMPRESS_ZSTD ||		\
161 	(compress) == ZIO_COMPRESS_ON ||		\
162 	(compress) == ZIO_COMPRESS_OFF)
163 
164 
165 #define	ZIO_COMPRESS_ALGO(x)	(x & SPA_COMPRESSMASK)
166 #define	ZIO_COMPRESS_LEVEL(x)	((x & ~SPA_COMPRESSMASK) >> SPA_COMPRESSBITS)
167 #define	ZIO_COMPRESS_RAW(type, level)	(type | ((level) << SPA_COMPRESSBITS))
168 
169 #define	ZIO_COMPLEVEL_ZSTD(level)	\
170 	ZIO_COMPRESS_RAW(ZIO_COMPRESS_ZSTD, level)
171 
172 #define	ZIO_FAILURE_MODE_WAIT		0
173 #define	ZIO_FAILURE_MODE_CONTINUE	1
174 #define	ZIO_FAILURE_MODE_PANIC		2
175 
176 typedef enum zio_suspend_reason {
177 	ZIO_SUSPEND_NONE = 0,
178 	ZIO_SUSPEND_IOERR,
179 	ZIO_SUSPEND_MMP,
180 } zio_suspend_reason_t;
181 
182 /*
183  * This was originally an enum type. However, those are 32-bit and there is no
184  * way to make a 64-bit enum type. Since we ran out of bits for flags, we were
185  * forced to upgrade it to a uint64_t.
186  *
187  * NOTE: PLEASE UPDATE THE BITFIELD STRINGS IN zfs_valstr.c IF YOU ADD ANOTHER
188  * FLAG.
189  */
190 typedef uint64_t zio_flag_t;
191 	/*
192 	 * Flags inherited by gang, ddt, and vdev children,
193 	 * and that must be equal for two zios to aggregate
194 	 */
195 #define	ZIO_FLAG_DONT_AGGREGATE	(1ULL << 0)
196 #define	ZIO_FLAG_IO_REPAIR	(1ULL << 1)
197 #define	ZIO_FLAG_SELF_HEAL	(1ULL << 2)
198 #define	ZIO_FLAG_RESILVER	(1ULL << 3)
199 #define	ZIO_FLAG_SCRUB		(1ULL << 4)
200 #define	ZIO_FLAG_SCAN_THREAD	(1ULL << 5)
201 #define	ZIO_FLAG_PHYSICAL	(1ULL << 6)
202 
203 #define	ZIO_FLAG_AGG_INHERIT	(ZIO_FLAG_CANFAIL - 1)
204 
205 	/*
206 	 * Flags inherited by ddt, gang, and vdev children.
207 	 */
208 #define	ZIO_FLAG_CANFAIL	(1ULL << 7)	/* must be first for INHERIT */
209 #define	ZIO_FLAG_SPECULATIVE	(1ULL << 8)
210 #define	ZIO_FLAG_CONFIG_WRITER	(1ULL << 9)
211 #define	ZIO_FLAG_DONT_RETRY	(1ULL << 10)
212 #define	ZIO_FLAG_NODATA		(1ULL << 12)
213 #define	ZIO_FLAG_INDUCE_DAMAGE	(1ULL << 13)
214 #define	ZIO_FLAG_ALLOC_THROTTLED	(1ULL << 14)
215 
216 #define	ZIO_FLAG_DDT_INHERIT	(ZIO_FLAG_IO_RETRY - 1)
217 #define	ZIO_FLAG_GANG_INHERIT	(ZIO_FLAG_IO_RETRY - 1)
218 
219 	/*
220 	 * Flags inherited by vdev children.
221 	 */
222 #define	ZIO_FLAG_IO_RETRY	(1ULL << 15)	/* must be first for INHERIT */
223 #define	ZIO_FLAG_PROBE		(1ULL << 16)
224 #define	ZIO_FLAG_TRYHARD	(1ULL << 17)
225 #define	ZIO_FLAG_OPTIONAL	(1ULL << 18)
226 #define	ZIO_FLAG_DIO_READ	(1ULL << 19)
227 #define	ZIO_FLAG_VDEV_INHERIT	(ZIO_FLAG_DONT_QUEUE - 1)
228 
229 	/*
230 	 * Flags not inherited by any children.
231 	 */
232 #define	ZIO_FLAG_DONT_QUEUE	(1ULL << 20)	/* must be first for INHERIT */
233 #define	ZIO_FLAG_DONT_PROPAGATE	(1ULL << 21)
234 #define	ZIO_FLAG_IO_BYPASS	(1ULL << 22)
235 #define	ZIO_FLAG_IO_REWRITE	(1ULL << 23)
236 #define	ZIO_FLAG_RAW_COMPRESS	(1ULL << 24)
237 #define	ZIO_FLAG_RAW_ENCRYPT	(1ULL << 25)
238 #define	ZIO_FLAG_GANG_CHILD	(1ULL << 26)
239 #define	ZIO_FLAG_DDT_CHILD	(1ULL << 27)
240 #define	ZIO_FLAG_GODFATHER	(1ULL << 28)
241 #define	ZIO_FLAG_NOPWRITE	(1ULL << 29)
242 #define	ZIO_FLAG_REEXECUTED	(1ULL << 30)
243 #define	ZIO_FLAG_DELEGATED	(1ULL << 31)
244 #define	ZIO_FLAG_PREALLOCATED	(1ULL << 32)
245 
246 #define	ZIO_ALLOCATOR_NONE	(-1)
247 #define	ZIO_HAS_ALLOCATOR(zio)	((zio)->io_allocator != ZIO_ALLOCATOR_NONE)
248 
249 #define	ZIO_FLAG_MUSTSUCCEED		0
250 #define	ZIO_FLAG_RAW	(ZIO_FLAG_RAW_COMPRESS | ZIO_FLAG_RAW_ENCRYPT)
251 
252 #define	ZIO_DDT_CHILD_FLAGS(zio)				\
253 	(((zio)->io_flags & ZIO_FLAG_DDT_INHERIT) |		\
254 	ZIO_FLAG_DDT_CHILD | ZIO_FLAG_CANFAIL)
255 
256 #define	ZIO_GANG_CHILD_FLAGS(zio)				\
257 	(((zio)->io_flags & ZIO_FLAG_GANG_INHERIT) |		\
258 	ZIO_FLAG_GANG_CHILD | ZIO_FLAG_CANFAIL)
259 
260 #define	ZIO_VDEV_CHILD_FLAGS(zio)				\
261 	(((zio)->io_flags & ZIO_FLAG_VDEV_INHERIT) |		\
262 	ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_CANFAIL)
263 
264 #define	ZIO_CHILD_BIT(x)		(1U << (x))
265 #define	ZIO_CHILD_BIT_IS_SET(val, x)	((val) & (1U << (x)))
266 
267 enum zio_child {
268 	ZIO_CHILD_VDEV = 0,
269 	ZIO_CHILD_GANG,
270 	ZIO_CHILD_DDT,
271 	ZIO_CHILD_LOGICAL,
272 	ZIO_CHILD_TYPES
273 };
274 
275 #define	ZIO_CHILD_VDEV_BIT		ZIO_CHILD_BIT(ZIO_CHILD_VDEV)
276 #define	ZIO_CHILD_GANG_BIT		ZIO_CHILD_BIT(ZIO_CHILD_GANG)
277 #define	ZIO_CHILD_DDT_BIT		ZIO_CHILD_BIT(ZIO_CHILD_DDT)
278 #define	ZIO_CHILD_LOGICAL_BIT		ZIO_CHILD_BIT(ZIO_CHILD_LOGICAL)
279 #define	ZIO_CHILD_ALL_BITS					\
280 	(ZIO_CHILD_VDEV_BIT | ZIO_CHILD_GANG_BIT |		\
281 	ZIO_CHILD_DDT_BIT | ZIO_CHILD_LOGICAL_BIT)
282 
283 enum zio_wait_type {
284 	ZIO_WAIT_READY = 0,
285 	ZIO_WAIT_DONE,
286 	ZIO_WAIT_TYPES
287 };
288 
289 typedef void zio_done_func_t(zio_t *zio);
290 
291 extern int zio_exclude_metadata;
292 extern int zio_dva_throttle_enabled;
293 extern const char *const zio_type_name[ZIO_TYPES];
294 
295 /*
296  * A bookmark is a four-tuple <objset, object, level, blkid> that uniquely
297  * identifies any block in the pool.  By convention, the meta-objset (MOS)
298  * is objset 0, and the meta-dnode is object 0.  This covers all blocks
299  * except root blocks and ZIL blocks, which are defined as follows:
300  *
301  * Root blocks (objset_phys_t) are object 0, level -1:  <objset, 0, -1, 0>.
302  * ZIL blocks are bookmarked <objset, 0, -2, blkid == ZIL sequence number>.
303  * dmu_sync()ed ZIL data blocks are bookmarked <objset, object, -2, blkid>.
304  * dnode visit bookmarks are <objset, object id of dnode, -3, 0>.
305  *
306  * Note: this structure is called a bookmark because its original purpose
307  * was to remember where to resume a pool-wide traverse.
308  *
309  * Note: this structure is passed between userland and the kernel, and is
310  * stored on disk (by virtue of being incorporated into other on-disk
311  * structures, e.g. dsl_scan_phys_t).
312  *
313  * If the head_errlog feature is enabled a different on-disk format for error
314  * logs is used. This introduces the use of an error bookmark, a four-tuple
315  * <object, level, blkid, birth> that uniquely identifies any error block
316  * in the pool. The birth transaction group is used to track whether the block
317  * has been overwritten by newer data or added to a snapshot since its marking
318  * as an error.
319  */
320 struct zbookmark_phys {
321 	uint64_t	zb_objset;
322 	uint64_t	zb_object;
323 	int64_t		zb_level;
324 	uint64_t	zb_blkid;
325 };
326 
327 struct zbookmark_err_phys {
328 	uint64_t	zb_object;
329 	int64_t		zb_level;
330 	uint64_t	zb_blkid;
331 	uint64_t	zb_birth;
332 };
333 
334 #define	SET_BOOKMARK(zb, objset, object, level, blkid)  \
335 {                                                       \
336 	(zb)->zb_objset = objset;                       \
337 	(zb)->zb_object = object;                       \
338 	(zb)->zb_level = level;                         \
339 	(zb)->zb_blkid = blkid;                         \
340 }
341 
342 #define	ZB_DESTROYED_OBJSET	(-1ULL)
343 
344 #define	ZB_ROOT_OBJECT		(0ULL)
345 #define	ZB_ROOT_LEVEL		(-1LL)
346 #define	ZB_ROOT_BLKID		(0ULL)
347 
348 #define	ZB_ZIL_OBJECT		(0ULL)
349 #define	ZB_ZIL_LEVEL		(-2LL)
350 
351 #define	ZB_DNODE_LEVEL		(-3LL)
352 #define	ZB_DNODE_BLKID		(0ULL)
353 
354 #define	ZB_IS_ZERO(zb)						\
355 	((zb)->zb_objset == 0 && (zb)->zb_object == 0 &&	\
356 	(zb)->zb_level == 0 && (zb)->zb_blkid == 0)
357 #define	ZB_IS_ROOT(zb)				\
358 	((zb)->zb_object == ZB_ROOT_OBJECT &&	\
359 	(zb)->zb_level == ZB_ROOT_LEVEL &&	\
360 	(zb)->zb_blkid == ZB_ROOT_BLKID)
361 
362 typedef struct zio_prop {
363 	enum zio_checksum	zp_checksum;
364 	enum zio_compress	zp_compress;
365 	uint8_t			zp_complevel;
366 	uint8_t			zp_level;
367 	uint8_t			zp_copies;
368 	uint8_t			zp_gang_copies;
369 	dmu_object_type_t	zp_type;
370 	boolean_t		zp_dedup;
371 	boolean_t		zp_dedup_verify;
372 	boolean_t		zp_nopwrite;
373 	boolean_t		zp_brtwrite;
374 	boolean_t		zp_encrypt;
375 	boolean_t		zp_byteorder;
376 	boolean_t		zp_direct_write;
377 	boolean_t		zp_rewrite;
378 	uint8_t			zp_salt[ZIO_DATA_SALT_LEN];
379 	uint8_t			zp_iv[ZIO_DATA_IV_LEN];
380 	uint8_t			zp_mac[ZIO_DATA_MAC_LEN];
381 	uint32_t		zp_zpl_smallblk;
382 	dmu_object_type_t	zp_storage_type;
383 } zio_prop_t;
384 
385 typedef struct zio_cksum_report zio_cksum_report_t;
386 
387 typedef void zio_cksum_finish_f(zio_cksum_report_t *rep,
388     const abd_t *good_data);
389 typedef void zio_cksum_free_f(void *cbdata, size_t size);
390 
391 struct zio_bad_cksum;				/* defined in zio_checksum.h */
392 struct dnode_phys;
393 struct abd;
394 
395 struct zio_cksum_report {
396 	struct zio_cksum_report *zcr_next;
397 	nvlist_t		*zcr_ereport;
398 	nvlist_t		*zcr_detector;
399 	void			*zcr_cbdata;
400 	size_t			zcr_cbinfo;	/* passed to zcr_free() */
401 	uint64_t		zcr_sector;
402 	uint64_t		zcr_align;
403 	uint64_t		zcr_length;
404 	zio_cksum_finish_f	*zcr_finish;
405 	zio_cksum_free_f	*zcr_free;
406 
407 	/* internal use only */
408 	struct zio_bad_cksum	*zcr_ckinfo;	/* information from failure */
409 };
410 
411 typedef struct zio_vsd_ops {
412 	zio_done_func_t		*vsd_free;
413 } zio_vsd_ops_t;
414 
415 typedef struct zio_gang_node {
416 	zio_gbh_phys_t		*gn_gbh;
417 	uint64_t		gn_gangblocksize;
418 	uint64_t		gn_allocsize;
419 	struct zio_gang_node	*gn_child[];
420 } zio_gang_node_t;
421 
422 typedef zio_t *zio_gang_issue_func_t(zio_t *zio, blkptr_t *bp,
423     zio_gang_node_t *gn, struct abd *data, uint64_t offset);
424 
425 typedef void zio_transform_func_t(zio_t *zio, struct abd *data, uint64_t size);
426 
427 typedef struct zio_transform {
428 	struct abd		*zt_orig_abd;
429 	uint64_t		zt_orig_size;
430 	uint64_t		zt_bufsize;
431 	zio_transform_func_t	*zt_transform;
432 	struct zio_transform	*zt_next;
433 } zio_transform_t;
434 
435 typedef zio_t *zio_pipe_stage_t(zio_t *zio);
436 
437 /*
438  * The io_post flags describe additional actions that a parent IO should
439  * consider or perform on behalf of a child. They are distinct from io_flags
440  * because the child must be able to propagate them to the parent. The normal
441  * io_flags are local to the zio, not protected by any lock, and not modifiable
442  * by children; the reexecute flags are protected by io_lock, modifiable by
443  * children, and always propagated -- even when ZIO_FLAG_DONT_PROPAGATE is set.
444  */
445 #define	ZIO_POST_REEXECUTE	(1 << 0)
446 #define	ZIO_POST_SUSPEND	(1 << 1)
447 #define	ZIO_POST_DIO_CHKSUM_ERR	(1 << 2)
448 
449 /*
450  * The io_trim flags are used to specify the type of TRIM to perform.  They
451  * only apply to ZIO_TYPE_TRIM zios are distinct from io_flags.
452  */
453 enum trim_flag {
454 	ZIO_TRIM_SECURE		= 1U << 0,
455 };
456 
457 typedef struct zio_alloc_list {
458 	list_t  zal_list;
459 	uint64_t zal_size;
460 } zio_alloc_list_t;
461 
462 typedef struct zio_link {
463 	zio_t		*zl_parent;
464 	zio_t		*zl_child;
465 	list_node_t	zl_parent_node;
466 	list_node_t	zl_child_node;
467 } zio_link_t;
468 
469 enum zio_qstate {
470 	ZIO_QS_NONE = 0,
471 	ZIO_QS_QUEUED,
472 	ZIO_QS_ACTIVE,
473 };
474 
475 struct zio {
476 	/* Core information about this I/O */
477 	zbookmark_phys_t	io_bookmark;
478 	zio_prop_t	io_prop;
479 	zio_type_t	io_type;
480 	enum zio_child	io_child_type;
481 	enum trim_flag	io_trim_flags;
482 	zio_priority_t	io_priority;
483 	uint8_t		io_post;
484 	uint8_t		io_state[ZIO_WAIT_TYPES];
485 	uint64_t	io_txg;
486 	spa_t		*io_spa;
487 	blkptr_t	*io_bp;
488 	blkptr_t	*io_bp_override;
489 	blkptr_t	io_bp_copy;
490 	list_t		io_parent_list;
491 	list_t		io_child_list;
492 	zio_t		*io_logical;
493 	zio_transform_t *io_transform_stack;
494 
495 	/* Callback info */
496 	zio_done_func_t	*io_ready;
497 	zio_done_func_t	*io_children_ready;
498 	zio_done_func_t	*io_done;
499 	void		*io_private;
500 	int64_t		io_prev_space_delta;	/* DMU private */
501 	blkptr_t	io_bp_orig;
502 	/* io_lsize != io_orig_size iff this is a raw write */
503 	uint64_t	io_lsize;
504 
505 	/* Data represented by this I/O */
506 	struct abd	*io_abd;
507 	struct abd	*io_orig_abd;
508 	uint64_t	io_size;
509 	uint64_t	io_orig_size;
510 
511 	/* Stuff for the vdev stack */
512 	vdev_t		*io_vd;
513 	void		*io_vsd;
514 	const zio_vsd_ops_t *io_vsd_ops;
515 	metaslab_class_t *io_metaslab_class;	/* dva throttle class */
516 
517 	enum zio_qstate	io_queue_state;	/* vdev queue state */
518 	union {
519 		list_node_t l;
520 		avl_node_t a;
521 	} io_queue_node ____cacheline_aligned;	/* allocator and vdev queues */
522 	avl_node_t	io_offset_node;	/* vdev offset queues */
523 	uint64_t	io_offset;
524 	hrtime_t	io_timestamp;	/* submitted at */
525 	hrtime_t	io_queued_timestamp;
526 	hrtime_t	io_target_timestamp;
527 	hrtime_t	io_delta;	/* vdev queue service delta */
528 	hrtime_t	io_delay;	/* Device access time (disk or */
529 					/* file). */
530 	zio_alloc_list_t 	io_alloc_list;
531 
532 	/* Internal pipeline state */
533 	zio_flag_t	io_flags;
534 	enum zio_stage	io_stage;
535 	enum zio_stage	io_pipeline;
536 	zio_flag_t	io_orig_flags;
537 	enum zio_stage	io_orig_stage;
538 	enum zio_stage	io_orig_pipeline;
539 	enum zio_stage	io_pipeline_trace;
540 	int		io_error;
541 	int		io_child_error[ZIO_CHILD_TYPES];
542 	uint64_t	io_children[ZIO_CHILD_TYPES][ZIO_WAIT_TYPES];
543 	uint64_t	*io_stall;
544 	zio_t		*io_gang_leader;
545 	zio_gang_node_t	*io_gang_tree;
546 	void		*io_executor;
547 	void		*io_waiter;
548 	void		*io_bio;
549 	kmutex_t	io_lock;
550 	kcondvar_t	io_cv;
551 	int		io_allocator;
552 
553 	/* FMA state */
554 	zio_cksum_report_t *io_cksum_report;
555 	uint64_t	io_ena;
556 
557 	/* Taskq dispatching state */
558 	taskq_ent_t	io_tqent;
559 };
560 
561 enum blk_verify_flag {
562 	BLK_VERIFY_ONLY,
563 	BLK_VERIFY_LOG,
564 	BLK_VERIFY_HALT
565 };
566 
567 enum blk_config_flag {
568 	BLK_CONFIG_HELD,   // SCL_VDEV held for writer
569 	BLK_CONFIG_NEEDED, // SCL_VDEV should be obtained for reader
570 	BLK_CONFIG_NEEDED_TRY, // Try with SCL_VDEV for reader
571 	BLK_CONFIG_SKIP,   // skip checks which require SCL_VDEV
572 };
573 
574 extern int zio_bookmark_compare(const void *, const void *);
575 
576 extern zio_t *zio_null(zio_t *pio, spa_t *spa, vdev_t *vd,
577     zio_done_func_t *done, void *priv, zio_flag_t flags);
578 
579 extern zio_t *zio_root(spa_t *spa,
580     zio_done_func_t *done, void *priv, zio_flag_t flags);
581 
582 extern void zio_destroy(zio_t *zio);
583 
584 extern zio_t *zio_read(zio_t *pio, spa_t *spa, const blkptr_t *bp,
585     struct abd *data, uint64_t lsize, zio_done_func_t *done, void *priv,
586     zio_priority_t priority, zio_flag_t flags, const zbookmark_phys_t *zb);
587 
588 extern zio_t *zio_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
589     struct abd *data, uint64_t size, uint64_t psize, const zio_prop_t *zp,
590     zio_done_func_t *ready, zio_done_func_t *children_ready,
591     zio_done_func_t *done, void *priv, zio_priority_t priority,
592     zio_flag_t flags, const zbookmark_phys_t *zb);
593 
594 extern zio_t *zio_rewrite(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
595     struct abd *data, uint64_t size, zio_done_func_t *done, void *priv,
596     zio_priority_t priority, zio_flag_t flags, zbookmark_phys_t *zb);
597 
598 extern void zio_write_override(zio_t *zio, blkptr_t *bp, int copies,
599     int gang_copies, boolean_t nopwrite, boolean_t brtwrite);
600 
601 extern void zio_free(spa_t *spa, uint64_t txg, const blkptr_t *bp);
602 
603 extern zio_t *zio_claim(zio_t *pio, spa_t *spa, uint64_t txg,
604     const blkptr_t *bp,
605     zio_done_func_t *done, void *priv, zio_flag_t flags);
606 
607 extern zio_t *zio_trim(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size,
608     zio_done_func_t *done, void *priv, zio_priority_t priority,
609     zio_flag_t flags, enum trim_flag trim_flags);
610 
611 extern zio_t *zio_read_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
612     uint64_t size, struct abd *data, int checksum,
613     zio_done_func_t *done, void *priv, zio_priority_t priority,
614     zio_flag_t flags, boolean_t labels);
615 
616 extern zio_t *zio_write_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
617     uint64_t size, struct abd *data, int checksum,
618     zio_done_func_t *done, void *priv, zio_priority_t priority,
619     zio_flag_t flags, boolean_t labels);
620 
621 extern zio_t *zio_free_sync(zio_t *pio, spa_t *spa, uint64_t txg,
622     const blkptr_t *bp, zio_flag_t flags);
623 
624 extern int zio_alloc_zil(spa_t *spa, objset_t *os, uint64_t txg,
625     blkptr_t *new_bp, uint64_t size, boolean_t *slog);
626 extern void zio_flush(zio_t *zio, vdev_t *vd);
627 extern void zio_shrink(zio_t *zio, uint64_t size);
628 
629 extern size_t zio_get_compression_max_size(enum zio_compress compress,
630     uint64_t gcd_alloc, uint64_t min_alloc, size_t s_len);
631 extern int zio_wait(zio_t *zio);
632 extern void zio_nowait(zio_t *zio);
633 extern void zio_execute(void *zio);
634 extern void zio_interrupt(void *zio);
635 extern void zio_delay_init(zio_t *zio);
636 extern void zio_delay_interrupt(zio_t *zio);
637 extern void zio_deadman(zio_t *zio, const char *tag);
638 
639 extern zio_t *zio_walk_parents(zio_t *cio, zio_link_t **);
640 extern zio_t *zio_walk_children(zio_t *pio, zio_link_t **);
641 extern zio_t *zio_unique_parent(zio_t *cio);
642 extern void zio_add_child(zio_t *pio, zio_t *cio);
643 
644 extern void *zio_buf_alloc(size_t size);
645 extern void zio_buf_free(void *buf, size_t size);
646 extern void *zio_data_buf_alloc(size_t size);
647 extern void zio_data_buf_free(void *buf, size_t size);
648 
649 extern void zio_push_transform(zio_t *zio, struct abd *abd, uint64_t size,
650     uint64_t bufsize, zio_transform_func_t *transform);
651 extern void zio_pop_transforms(zio_t *zio);
652 
653 extern void zio_resubmit_stage_async(void *);
654 
655 extern zio_t *zio_vdev_child_io(zio_t *zio, blkptr_t *bp, vdev_t *vd,
656     uint64_t offset, struct abd *data, uint64_t size, int type,
657     zio_priority_t priority, zio_flag_t flags,
658     zio_done_func_t *done, void *priv);
659 
660 extern zio_t *zio_vdev_delegated_io(vdev_t *vd, uint64_t offset,
661     struct abd *data, uint64_t size, zio_type_t type, zio_priority_t priority,
662     zio_flag_t flags, zio_done_func_t *done, void *priv);
663 
664 extern void zio_vdev_io_bypass(zio_t *zio);
665 extern void zio_vdev_io_reissue(zio_t *zio);
666 extern void zio_vdev_io_redone(zio_t *zio);
667 
668 extern void zio_change_priority(zio_t *pio, zio_priority_t priority);
669 
670 extern void zio_checksum_verified(zio_t *zio);
671 extern void zio_dio_chksum_verify_error_report(zio_t *zio);
672 extern int zio_worst_error(int e1, int e2);
673 
674 extern enum zio_checksum zio_checksum_select(enum zio_checksum child,
675     enum zio_checksum parent);
676 extern enum zio_checksum zio_checksum_dedup_select(spa_t *spa,
677     enum zio_checksum child, enum zio_checksum parent);
678 extern enum zio_compress zio_compress_select(spa_t *spa,
679     enum zio_compress child, enum zio_compress parent);
680 extern uint8_t zio_complevel_select(spa_t *spa, enum zio_compress compress,
681     uint8_t child, uint8_t parent);
682 
683 extern void zio_suspend(spa_t *spa, zio_t *zio, zio_suspend_reason_t);
684 extern int zio_resume(spa_t *spa);
685 extern void zio_resume_wait(spa_t *spa);
686 
687 extern int zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp,
688     enum blk_config_flag blk_config, enum blk_verify_flag blk_verify);
689 
690 /*
691  * Initial setup and teardown.
692  */
693 extern void zio_init(void);
694 extern void zio_fini(void);
695 
696 /*
697  * Fault injection
698  */
699 struct zinject_record;
700 extern uint32_t zio_injection_enabled;
701 extern int zio_inject_fault(char *name, int flags, int *id,
702     struct zinject_record *record);
703 extern int zio_inject_list_next(int *id, char *name, size_t buflen,
704     struct zinject_record *record);
705 extern int zio_clear_fault(int id);
706 extern void zio_handle_panic_injection(spa_t *spa, const char *tag,
707     uint64_t type);
708 extern int zio_handle_decrypt_injection(spa_t *spa, const zbookmark_phys_t *zb,
709     uint64_t type, int error);
710 extern int zio_handle_fault_injection(zio_t *zio, int error);
711 extern int zio_handle_device_injection(vdev_t *vd, zio_t *zio, int error);
712 extern int zio_handle_device_injections(vdev_t *vd, zio_t *zio, int err1,
713     int err2);
714 extern int zio_handle_label_injection(zio_t *zio, int error);
715 extern void zio_handle_ignored_writes(zio_t *zio);
716 extern hrtime_t zio_handle_io_delay(zio_t *zio);
717 extern void zio_handle_import_delay(spa_t *spa, hrtime_t elapsed);
718 extern void zio_handle_export_delay(spa_t *spa, hrtime_t elapsed);
719 
720 /*
721  * Checksum ereport functions
722  */
723 extern int zfs_ereport_start_checksum(spa_t *spa, vdev_t *vd,
724     const zbookmark_phys_t *zb, struct zio *zio, uint64_t offset,
725     uint64_t length, struct zio_bad_cksum *info);
726 extern void zfs_ereport_finish_checksum(zio_cksum_report_t *report,
727     const abd_t *good_data, const abd_t *bad_data, boolean_t drop_if_identical);
728 
729 extern void zfs_ereport_free_checksum(zio_cksum_report_t *report);
730 
731 /* If we have the good data in hand, this function can be used */
732 extern int zfs_ereport_post_checksum(spa_t *spa, vdev_t *vd,
733     const zbookmark_phys_t *zb, struct zio *zio, uint64_t offset,
734     uint64_t length, const abd_t *good_data, const abd_t *bad_data,
735     struct zio_bad_cksum *info);
736 
737 void zio_vsd_default_cksum_report(zio_t *zio, zio_cksum_report_t *zcr);
738 extern void zfs_ereport_snapshot_post(const char *subclass, spa_t *spa,
739     const char *name);
740 
741 /* Called from spa_sync(), but primarily an injection handler */
742 extern void spa_handle_ignored_writes(spa_t *spa);
743 
744 /* zbookmark_phys functions */
745 boolean_t zbookmark_subtree_completed(const struct dnode_phys *dnp,
746     const zbookmark_phys_t *subtree_root, const zbookmark_phys_t *last_block);
747 boolean_t zbookmark_subtree_tbd(const struct dnode_phys *dnp,
748     const zbookmark_phys_t *subtree_root, const zbookmark_phys_t *last_block);
749 int zbookmark_compare(uint16_t dbss1, uint8_t ibs1, uint16_t dbss2,
750     uint8_t ibs2, const zbookmark_phys_t *zb1, const zbookmark_phys_t *zb2);
751 
752 #ifdef	__cplusplus
753 }
754 #endif
755 
756 #endif	/* _ZIO_H */
757