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, 2017 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))
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 enum zio_flag {
142 /*
143 * Flags inherited by gang, ddt, and vdev children,
144 * and that must be equal for two zios to aggregate
145 */
146 ZIO_FLAG_DONT_AGGREGATE = 1 << 0,
147 ZIO_FLAG_IO_REPAIR = 1 << 1,
148 ZIO_FLAG_SELF_HEAL = 1 << 2,
149 ZIO_FLAG_RESILVER = 1 << 3,
150 ZIO_FLAG_SCRUB = 1 << 4,
151 ZIO_FLAG_SCAN_THREAD = 1 << 5,
152 ZIO_FLAG_PHYSICAL = 1 << 6,
153
154 #define ZIO_FLAG_AGG_INHERIT (ZIO_FLAG_CANFAIL - 1)
155
156 /*
157 * Flags inherited by ddt, gang, and vdev children.
158 */
159 ZIO_FLAG_CANFAIL = 1 << 7, /* must be first for INHERIT */
160 ZIO_FLAG_SPECULATIVE = 1 << 8,
189 ZIO_FLAG_GANG_CHILD = 1 << 24,
190 ZIO_FLAG_DDT_CHILD = 1 << 25,
191 ZIO_FLAG_GODFATHER = 1 << 26,
192 ZIO_FLAG_NOPWRITE = 1 << 27,
193 ZIO_FLAG_REEXECUTED = 1 << 28,
194 ZIO_FLAG_DELEGATED = 1 << 29,
195 };
196
197 #define ZIO_FLAG_MUSTSUCCEED 0
198
199 #define ZIO_DDT_CHILD_FLAGS(zio) \
200 (((zio)->io_flags & ZIO_FLAG_DDT_INHERIT) | \
201 ZIO_FLAG_DDT_CHILD | ZIO_FLAG_CANFAIL)
202
203 #define ZIO_GANG_CHILD_FLAGS(zio) \
204 (((zio)->io_flags & ZIO_FLAG_GANG_INHERIT) | \
205 ZIO_FLAG_GANG_CHILD | ZIO_FLAG_CANFAIL)
206
207 #define ZIO_VDEV_CHILD_FLAGS(zio) \
208 (((zio)->io_flags & ZIO_FLAG_VDEV_INHERIT) | \
209 ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_CANFAIL)
210
211 #define ZIO_CHILD_BIT(x) (1 << (x))
212 #define ZIO_CHILD_BIT_IS_SET(val, x) ((val) & (1 << (x)))
213
214 enum zio_child {
215 ZIO_CHILD_VDEV = 0,
216 ZIO_CHILD_GANG,
217 ZIO_CHILD_DDT,
218 ZIO_CHILD_LOGICAL,
219 ZIO_CHILD_TYPES
220 };
221
222 #define ZIO_CHILD_VDEV_BIT ZIO_CHILD_BIT(ZIO_CHILD_VDEV)
223 #define ZIO_CHILD_GANG_BIT ZIO_CHILD_BIT(ZIO_CHILD_GANG)
224 #define ZIO_CHILD_DDT_BIT ZIO_CHILD_BIT(ZIO_CHILD_DDT)
225 #define ZIO_CHILD_LOGICAL_BIT ZIO_CHILD_BIT(ZIO_CHILD_LOGICAL)
226 #define ZIO_CHILD_ALL_BITS \
227 (ZIO_CHILD_VDEV_BIT | ZIO_CHILD_GANG_BIT | \
228 ZIO_CHILD_DDT_BIT | ZIO_CHILD_LOGICAL_BIT)
229
230 enum zio_wait_type {
231 ZIO_WAIT_READY = 0,
232 ZIO_WAIT_DONE,
233 ZIO_WAIT_TYPES
234 };
235
236 /*
237 * We'll take the unused errnos, 'EBADE' and 'EBADR' (from the Convergent
238 * graveyard) to indicate checksum errors and fragmentation.
239 */
240 #define ECKSUM EBADE
241 #define EFRAGS EBADR
242
243 typedef void zio_done_func_t(zio_t *zio);
244
245 extern boolean_t zio_dva_throttle_enabled;
246 extern const char *zio_type_name[ZIO_TYPES];
247
248 /*
249 * A bookmark is a four-tuple <objset, object, level, blkid> that uniquely
250 * identifies any block in the pool. By convention, the meta-objset (MOS)
251 * is objset 0, and the meta-dnode is object 0. This covers all blocks
252 * except root blocks and ZIL blocks, which are defined as follows:
253 *
254 * Root blocks (objset_phys_t) are object 0, level -1: <objset, 0, -1, 0>.
255 * ZIL blocks are bookmarked <objset, 0, -2, blkid == ZIL sequence number>.
256 * dmu_sync()ed ZIL data blocks are bookmarked <objset, object, -2, blkid>.
257 * dnode visit bookmarks are <objset, object id of dnode, -3, 0>.
258 *
259 * Note: this structure is called a bookmark because its original purpose
260 * was to remember where to resume a pool-wide traverse.
261 *
262 * Note: this structure is passed between userland and the kernel, and is
263 * stored on disk (by virtue of being incorporated into other on-disk
264 * structures, e.g. dsl_scan_phys_t).
265 */
266 typedef struct zbookmark_phys {
267 uint64_t zb_objset;
287 #define ZB_ZIL_OBJECT (0ULL)
288 #define ZB_ZIL_LEVEL (-2LL)
289
290 #define ZB_DNODE_LEVEL (-3LL)
291 #define ZB_DNODE_BLKID (0ULL)
292
293 #define ZB_IS_ZERO(zb) \
294 ((zb)->zb_objset == 0 && (zb)->zb_object == 0 && \
295 (zb)->zb_level == 0 && (zb)->zb_blkid == 0)
296 #define ZB_IS_ROOT(zb) \
297 ((zb)->zb_object == ZB_ROOT_OBJECT && \
298 (zb)->zb_level == ZB_ROOT_LEVEL && \
299 (zb)->zb_blkid == ZB_ROOT_BLKID)
300
301 typedef struct zio_prop {
302 enum zio_checksum zp_checksum;
303 enum zio_compress zp_compress;
304 dmu_object_type_t zp_type;
305 uint8_t zp_level;
306 uint8_t zp_copies;
307 boolean_t zp_dedup;
308 boolean_t zp_dedup_verify;
309 boolean_t zp_nopwrite;
310 } zio_prop_t;
311
312 typedef struct zio_cksum_report zio_cksum_report_t;
313
314 typedef void zio_cksum_finish_f(zio_cksum_report_t *rep,
315 const void *good_data);
316 typedef void zio_cksum_free_f(void *cbdata, size_t size);
317
318 struct zio_bad_cksum; /* defined in zio_checksum.h */
319 struct dnode_phys;
320 struct abd;
321
322 struct zio_cksum_report {
323 struct zio_cksum_report *zcr_next;
324 nvlist_t *zcr_ereport;
325 nvlist_t *zcr_detector;
326 void *zcr_cbdata;
327 size_t zcr_cbinfo; /* passed to zcr_free() */
328 uint64_t zcr_align;
329 uint64_t zcr_length;
369 * be able to propagate them to the parent. The normal io_flags are local
370 * to the zio, not protected by any lock, and not modifiable by children;
371 * the reexecute flags are protected by io_lock, modifiable by children,
372 * and always propagated -- even when ZIO_FLAG_DONT_PROPAGATE is set.
373 */
374 #define ZIO_REEXECUTE_NOW 0x01
375 #define ZIO_REEXECUTE_SUSPEND 0x02
376
377 typedef struct zio_alloc_list {
378 list_t zal_list;
379 uint64_t zal_size;
380 } zio_alloc_list_t;
381
382 typedef struct zio_link {
383 zio_t *zl_parent;
384 zio_t *zl_child;
385 list_node_t zl_parent_node;
386 list_node_t zl_child_node;
387 } zio_link_t;
388
389 struct zio {
390 /* Core information about this I/O */
391 zbookmark_phys_t io_bookmark;
392 zio_prop_t io_prop;
393 zio_type_t io_type;
394 enum zio_child io_child_type;
395 int io_cmd;
396 zio_priority_t io_priority;
397 uint8_t io_reexecute;
398 uint8_t io_state[ZIO_WAIT_TYPES];
399 uint64_t io_txg;
400 spa_t *io_spa;
401 blkptr_t *io_bp;
402 blkptr_t *io_bp_override;
403 blkptr_t io_bp_copy;
404 list_t io_parent_list;
405 list_t io_child_list;
406 zio_t *io_logical;
407 zio_transform_t *io_transform_stack;
408
409 /* Callback info */
410 zio_done_func_t *io_ready;
411 zio_done_func_t *io_children_ready;
412 zio_done_func_t *io_physdone;
413 zio_done_func_t *io_done;
414 void *io_private;
415 int64_t io_prev_space_delta; /* DMU private */
416 blkptr_t io_bp_orig;
417
418 /* Data represented by this I/O */
419 struct abd *io_abd;
420 struct abd *io_orig_abd;
421 uint64_t io_size;
422 uint64_t io_orig_size;
423 /* io_lsize != io_orig_size iff this is a raw write */
424 uint64_t io_lsize;
425
426 /* Stuff for the vdev stack */
427 vdev_t *io_vd;
448 int io_error;
449 int io_child_error[ZIO_CHILD_TYPES];
450 uint64_t io_children[ZIO_CHILD_TYPES][ZIO_WAIT_TYPES];
451 uint64_t io_child_count;
452 uint64_t io_phys_children;
453 uint64_t io_parent_count;
454 uint64_t *io_stall;
455 zio_t *io_gang_leader;
456 zio_gang_node_t *io_gang_tree;
457 void *io_executor;
458 void *io_waiter;
459 kmutex_t io_lock;
460 kcondvar_t io_cv;
461
462 /* FMA state */
463 zio_cksum_report_t *io_cksum_report;
464 uint64_t io_ena;
465
466 /* Taskq dispatching state */
467 taskq_ent_t io_tqent;
468 };
469
470 extern int zio_bookmark_compare(const void *, const void *);
471
472 extern zio_t *zio_null(zio_t *pio, spa_t *spa, vdev_t *vd,
473 zio_done_func_t *done, void *private, enum zio_flag flags);
474
475 extern zio_t *zio_root(spa_t *spa,
476 zio_done_func_t *done, void *private, enum zio_flag flags);
477
478 extern zio_t *zio_read(zio_t *pio, spa_t *spa, const blkptr_t *bp,
479 struct abd *data, uint64_t lsize, zio_done_func_t *done, void *private,
480 zio_priority_t priority, enum zio_flag flags, const zbookmark_phys_t *zb);
481
482 extern zio_t *zio_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
483 struct abd *data, uint64_t size, uint64_t psize, const zio_prop_t *zp,
484 zio_done_func_t *ready, zio_done_func_t *children_ready,
485 zio_done_func_t *physdone, zio_done_func_t *done,
486 void *private, zio_priority_t priority, enum zio_flag flags,
487 const zbookmark_phys_t *zb);
488
489 extern zio_t *zio_rewrite(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
490 struct abd *data, uint64_t size, zio_done_func_t *done, void *private,
491 zio_priority_t priority, enum zio_flag flags, zbookmark_phys_t *zb);
492
493 extern void zio_write_override(zio_t *zio, blkptr_t *bp, int copies,
494 boolean_t nopwrite);
495
496 extern void zio_free(spa_t *spa, uint64_t txg, const blkptr_t *bp);
497
498 extern zio_t *zio_claim(zio_t *pio, spa_t *spa, uint64_t txg,
499 const blkptr_t *bp,
500 zio_done_func_t *done, void *private, enum zio_flag flags);
501
502 extern zio_t *zio_ioctl(zio_t *pio, spa_t *spa, vdev_t *vd, int cmd,
503 zio_done_func_t *done, void *private, enum zio_flag flags);
504
505 extern zio_t *zio_read_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
506 uint64_t size, struct abd *data, int checksum,
507 zio_done_func_t *done, void *private, zio_priority_t priority,
508 enum zio_flag flags, boolean_t labels);
509
510 extern zio_t *zio_write_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
511 uint64_t size, struct abd *data, int checksum,
512 zio_done_func_t *done, void *private, zio_priority_t priority,
513 enum zio_flag flags, boolean_t labels);
514
515 extern zio_t *zio_free_sync(zio_t *pio, spa_t *spa, uint64_t txg,
516 const blkptr_t *bp, enum zio_flag flags);
517
518 extern int zio_alloc_zil(spa_t *spa, uint64_t txg, blkptr_t *new_bp,
519 blkptr_t *old_bp, uint64_t size, boolean_t *slog);
520 extern void zio_free_zil(spa_t *spa, uint64_t txg, blkptr_t *bp);
521 extern void zio_flush(zio_t *zio, vdev_t *vd);
522 extern void zio_shrink(zio_t *zio, uint64_t size);
523
524 extern int zio_wait(zio_t *zio);
525 extern void zio_nowait(zio_t *zio);
526 extern void zio_execute(zio_t *zio);
527 extern void zio_interrupt(zio_t *zio);
528 extern void zio_delay_init(zio_t *zio);
529 extern void zio_delay_interrupt(zio_t *zio);
530
531 extern zio_t *zio_walk_parents(zio_t *cio, zio_link_t **);
532 extern zio_t *zio_walk_children(zio_t *pio, zio_link_t **);
533 extern zio_t *zio_unique_parent(zio_t *cio);
534 extern void zio_add_child(zio_t *pio, zio_t *cio);
587 extern int zio_inject_list_next(int *id, char *name, size_t buflen,
588 struct zinject_record *record);
589 extern int zio_clear_fault(int id);
590 extern void zio_handle_panic_injection(spa_t *spa, char *tag, uint64_t type);
591 extern int zio_handle_fault_injection(zio_t *zio, int error);
592 extern int zio_handle_device_injection(vdev_t *vd, zio_t *zio, int error);
593 extern int zio_handle_label_injection(zio_t *zio, int error);
594 extern void zio_handle_ignored_writes(zio_t *zio);
595 extern hrtime_t zio_handle_io_delay(zio_t *zio);
596
597 /*
598 * Checksum ereport functions
599 */
600 extern void zfs_ereport_start_checksum(spa_t *spa, vdev_t *vd, struct zio *zio,
601 uint64_t offset, uint64_t length, void *arg, struct zio_bad_cksum *info);
602 extern void zfs_ereport_finish_checksum(zio_cksum_report_t *report,
603 const void *good_data, const void *bad_data, boolean_t drop_if_identical);
604
605 extern void zfs_ereport_send_interim_checksum(zio_cksum_report_t *report);
606 extern void zfs_ereport_free_checksum(zio_cksum_report_t *report);
607
608 /* If we have the good data in hand, this function can be used */
609 extern void zfs_ereport_post_checksum(spa_t *spa, vdev_t *vd,
610 struct zio *zio, uint64_t offset, uint64_t length,
611 const void *good_data, const void *bad_data, struct zio_bad_cksum *info);
612
613 /* Called from spa_sync(), but primarily an injection handler */
614 extern void spa_handle_ignored_writes(spa_t *spa);
615
616 /* zbookmark_phys functions */
617 boolean_t zbookmark_subtree_completed(const struct dnode_phys *dnp,
618 const zbookmark_phys_t *subtree_root, const zbookmark_phys_t *last_block);
619 int zbookmark_compare(uint16_t dbss1, uint8_t ibs1, uint16_t dbss2,
620 uint8_t ibs2, const zbookmark_phys_t *zb1, const zbookmark_phys_t *zb2);
621
622 #ifdef __cplusplus
623 }
624 #endif
625
626 #endif /* _ZIO_H */
|
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 2016 Nexenta Systems, Inc. All rights reserved.
25 * Copyright (c) 2012, 2017 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 * Checksum state w.r.t. SHA256 acceleration.
48 */
49 typedef enum {
50 CKSTATE_NONE = 0,
51 CKSTATE_WAITING,
52 CKSTATE_CHECKSUMMING,
53 CKSTATE_CHECKSUM_DONE
54 } zio_checksum_state_t;
55
56 /*
57 * Embedded checksum
58 */
59 #define ZEC_MAGIC 0x210da7ab10c7a11ULL
60
61 typedef struct zio_eck {
62 uint64_t zec_magic; /* for validation, endianness */
63 zio_cksum_t zec_cksum; /* 256-bit checksum */
64 } zio_eck_t;
65
66 /*
67 * Gang block headers are self-checksumming and contain an array
68 * of block pointers.
69 */
70 #define SPA_GANGBLOCKSIZE SPA_MINBLOCKSIZE
71 #define SPA_GBH_NBLKPTRS ((SPA_GANGBLOCKSIZE - \
72 sizeof (zio_eck_t)) / sizeof (blkptr_t))
73 #define SPA_GBH_FILLER ((SPA_GANGBLOCKSIZE - \
74 sizeof (zio_eck_t) - \
75 (SPA_GBH_NBLKPTRS * sizeof (blkptr_t))) /\
76 sizeof (uint64_t))
131
132 #define BOOTFS_COMPRESS_VALID(compress) \
133 ((compress) == ZIO_COMPRESS_LZJB || \
134 (compress) == ZIO_COMPRESS_LZ4 || \
135 (compress) == ZIO_COMPRESS_GZIP_1 || \
136 (compress) == ZIO_COMPRESS_GZIP_2 || \
137 (compress) == ZIO_COMPRESS_GZIP_3 || \
138 (compress) == ZIO_COMPRESS_GZIP_4 || \
139 (compress) == ZIO_COMPRESS_GZIP_5 || \
140 (compress) == ZIO_COMPRESS_GZIP_6 || \
141 (compress) == ZIO_COMPRESS_GZIP_7 || \
142 (compress) == ZIO_COMPRESS_GZIP_8 || \
143 (compress) == ZIO_COMPRESS_GZIP_9 || \
144 (compress) == ZIO_COMPRESS_ON || \
145 (compress) == ZIO_COMPRESS_OFF)
146
147 #define ZIO_FAILURE_MODE_WAIT 0
148 #define ZIO_FAILURE_MODE_CONTINUE 1
149 #define ZIO_FAILURE_MODE_PANIC 2
150
151 /*
152 * Macro for asserting validity of the priorities obtained by conversion
153 * from CoS/vdev properties
154 */
155 #define ZIO_PRIORITY_QUEUEABLE_VALID(prio) \
156 (((prio) >= ZIO_PRIORITY_SYNC_READ) && \
157 ((prio) < ZIO_PRIORITY_NUM_QUEUEABLE))
158
159 #define ZIO_PIPELINE_CONTINUE 0x100
160 #define ZIO_PIPELINE_STOP 0x101
161 #define ZIO_PIPELINE_RESTART_STAGE 0x102
162
163 enum zio_flag {
164 /*
165 * Flags inherited by gang, ddt, and vdev children,
166 * and that must be equal for two zios to aggregate
167 */
168 ZIO_FLAG_DONT_AGGREGATE = 1 << 0,
169 ZIO_FLAG_IO_REPAIR = 1 << 1,
170 ZIO_FLAG_SELF_HEAL = 1 << 2,
171 ZIO_FLAG_RESILVER = 1 << 3,
172 ZIO_FLAG_SCRUB = 1 << 4,
173 ZIO_FLAG_SCAN_THREAD = 1 << 5,
174 ZIO_FLAG_PHYSICAL = 1 << 6,
175
176 #define ZIO_FLAG_AGG_INHERIT (ZIO_FLAG_CANFAIL - 1)
177
178 /*
179 * Flags inherited by ddt, gang, and vdev children.
180 */
181 ZIO_FLAG_CANFAIL = 1 << 7, /* must be first for INHERIT */
182 ZIO_FLAG_SPECULATIVE = 1 << 8,
211 ZIO_FLAG_GANG_CHILD = 1 << 24,
212 ZIO_FLAG_DDT_CHILD = 1 << 25,
213 ZIO_FLAG_GODFATHER = 1 << 26,
214 ZIO_FLAG_NOPWRITE = 1 << 27,
215 ZIO_FLAG_REEXECUTED = 1 << 28,
216 ZIO_FLAG_DELEGATED = 1 << 29,
217 };
218
219 #define ZIO_FLAG_MUSTSUCCEED 0
220
221 #define ZIO_DDT_CHILD_FLAGS(zio) \
222 (((zio)->io_flags & ZIO_FLAG_DDT_INHERIT) | \
223 ZIO_FLAG_DDT_CHILD | ZIO_FLAG_CANFAIL)
224
225 #define ZIO_GANG_CHILD_FLAGS(zio) \
226 (((zio)->io_flags & ZIO_FLAG_GANG_INHERIT) | \
227 ZIO_FLAG_GANG_CHILD | ZIO_FLAG_CANFAIL)
228
229 #define ZIO_VDEV_CHILD_FLAGS(zio) \
230 (((zio)->io_flags & ZIO_FLAG_VDEV_INHERIT) | \
231 ZIO_FLAG_CANFAIL)
232
233 enum zio_child {
234 ZIO_CHILD_VDEV = 0,
235 ZIO_CHILD_GANG,
236 ZIO_CHILD_DDT,
237 ZIO_CHILD_LOGICAL,
238 ZIO_CHILD_TYPES
239 };
240
241 enum zio_wait_type {
242 ZIO_WAIT_READY = 0,
243 ZIO_WAIT_DONE,
244 ZIO_WAIT_TYPES
245 };
246
247 /*
248 * We'll take the unused errnos, 'EBADE' and 'EBADR' (from the Convergent
249 * graveyard) to indicate checksum errors and fragmentation.
250 */
251 #define ECKSUM EBADE
252 #define EFRAGS EBADR
253
254 typedef void zio_done_func_t(zio_t *zio);
255
256 extern boolean_t zio_dva_throttle_enabled;
257 extern const char *zio_type_name[ZIO_TYPES];
258
259 struct range_tree;
260
261 /*
262 * A bookmark is a four-tuple <objset, object, level, blkid> that uniquely
263 * identifies any block in the pool. By convention, the meta-objset (MOS)
264 * is objset 0, and the meta-dnode is object 0. This covers all blocks
265 * except root blocks and ZIL blocks, which are defined as follows:
266 *
267 * Root blocks (objset_phys_t) are object 0, level -1: <objset, 0, -1, 0>.
268 * ZIL blocks are bookmarked <objset, 0, -2, blkid == ZIL sequence number>.
269 * dmu_sync()ed ZIL data blocks are bookmarked <objset, object, -2, blkid>.
270 * dnode visit bookmarks are <objset, object id of dnode, -3, 0>.
271 *
272 * Note: this structure is called a bookmark because its original purpose
273 * was to remember where to resume a pool-wide traverse.
274 *
275 * Note: this structure is passed between userland and the kernel, and is
276 * stored on disk (by virtue of being incorporated into other on-disk
277 * structures, e.g. dsl_scan_phys_t).
278 */
279 typedef struct zbookmark_phys {
280 uint64_t zb_objset;
300 #define ZB_ZIL_OBJECT (0ULL)
301 #define ZB_ZIL_LEVEL (-2LL)
302
303 #define ZB_DNODE_LEVEL (-3LL)
304 #define ZB_DNODE_BLKID (0ULL)
305
306 #define ZB_IS_ZERO(zb) \
307 ((zb)->zb_objset == 0 && (zb)->zb_object == 0 && \
308 (zb)->zb_level == 0 && (zb)->zb_blkid == 0)
309 #define ZB_IS_ROOT(zb) \
310 ((zb)->zb_object == ZB_ROOT_OBJECT && \
311 (zb)->zb_level == ZB_ROOT_LEVEL && \
312 (zb)->zb_blkid == ZB_ROOT_BLKID)
313
314 typedef struct zio_prop {
315 enum zio_checksum zp_checksum;
316 enum zio_compress zp_compress;
317 dmu_object_type_t zp_type;
318 uint8_t zp_level;
319 uint8_t zp_copies;
320 uint8_t zp_dedup;
321 uint8_t zp_dedup_verify;
322 boolean_t zp_nopwrite;
323 boolean_t zp_metadata;
324 boolean_t zp_usesc;
325 boolean_t zp_usewbc;
326 uint64_t zp_zpl_meta_to_special;
327 } zio_prop_t;
328
329 typedef struct zio_cksum_report zio_cksum_report_t;
330
331 typedef void zio_cksum_finish_f(zio_cksum_report_t *rep,
332 const void *good_data);
333 typedef void zio_cksum_free_f(void *cbdata, size_t size);
334
335 struct zio_bad_cksum; /* defined in zio_checksum.h */
336 struct dnode_phys;
337 struct abd;
338
339 struct zio_cksum_report {
340 struct zio_cksum_report *zcr_next;
341 nvlist_t *zcr_ereport;
342 nvlist_t *zcr_detector;
343 void *zcr_cbdata;
344 size_t zcr_cbinfo; /* passed to zcr_free() */
345 uint64_t zcr_align;
346 uint64_t zcr_length;
386 * be able to propagate them to the parent. The normal io_flags are local
387 * to the zio, not protected by any lock, and not modifiable by children;
388 * the reexecute flags are protected by io_lock, modifiable by children,
389 * and always propagated -- even when ZIO_FLAG_DONT_PROPAGATE is set.
390 */
391 #define ZIO_REEXECUTE_NOW 0x01
392 #define ZIO_REEXECUTE_SUSPEND 0x02
393
394 typedef struct zio_alloc_list {
395 list_t zal_list;
396 uint64_t zal_size;
397 } zio_alloc_list_t;
398
399 typedef struct zio_link {
400 zio_t *zl_parent;
401 zio_t *zl_child;
402 list_node_t zl_parent_node;
403 list_node_t zl_child_node;
404 } zio_link_t;
405
406 /*
407 * When smart compression is enabled, this callback info structure is
408 * passed to write zio's to monitor per-object compression performance.
409 *
410 * When zio_write determines that the `compression' setting for the dataset
411 * is not `off', if `sc_ask' is not NULL, it will call the `sc_ask' callback
412 * function, asking the upper layers whether it should really try to compress
413 * the object in question. If the function returns B_TRUE, compression is
414 * attempted. Once compression is done, sc_result is called to inform the
415 * upper layers of the compression result. By comparing the zio's io_size to
416 * io_orig_size it can monitor compression performance on the particular
417 * object in question (if io_size == io_orig_size, then compression failed).
418 * It is not legal to pass a NULL sc_ask but non-NULL sc_result to zio_write.
419 */
420 typedef struct zio_smartcomp_info {
421 boolean_t (*sc_ask)(void *userinfo, const zio_t *);
422 void (*sc_result)(void *userinfo, const zio_t *);
423 void *sc_userinfo;
424 } zio_smartcomp_info_t;
425
426 #define ZIO_SHOULD_COMPRESS(zio) \
427 ((zio)->io_smartcomp.sc_ask == NULL || \
428 (zio)->io_smartcomp.sc_ask((zio)->io_smartcomp.sc_userinfo, (zio)))
429
430 struct zio {
431 /* Core information about this I/O */
432 zbookmark_phys_t io_bookmark;
433 zio_prop_t io_prop;
434 zio_type_t io_type;
435 enum zio_child io_child_type;
436 int io_cmd;
437 zio_priority_t io_priority;
438 uint8_t io_reexecute;
439 uint8_t io_state[ZIO_WAIT_TYPES];
440 uint64_t io_txg;
441 spa_t *io_spa;
442 blkptr_t *io_bp;
443 blkptr_t *io_bp_override;
444 blkptr_t io_bp_copy;
445 list_t io_parent_list;
446 list_t io_child_list;
447 zio_t *io_logical;
448 zio_transform_t *io_transform_stack;
449 zio_smartcomp_info_t io_smartcomp;
450
451 /* Callback info */
452 zio_done_func_t *io_ready;
453 zio_done_func_t *io_children_ready;
454 zio_done_func_t *io_physdone;
455 zio_done_func_t *io_done;
456 void *io_private;
457 int64_t io_prev_space_delta; /* DMU private */
458 blkptr_t io_bp_orig;
459
460 /* Data represented by this I/O */
461 struct abd *io_abd;
462 struct abd *io_orig_abd;
463 uint64_t io_size;
464 uint64_t io_orig_size;
465 /* io_lsize != io_orig_size iff this is a raw write */
466 uint64_t io_lsize;
467
468 /* Stuff for the vdev stack */
469 vdev_t *io_vd;
490 int io_error;
491 int io_child_error[ZIO_CHILD_TYPES];
492 uint64_t io_children[ZIO_CHILD_TYPES][ZIO_WAIT_TYPES];
493 uint64_t io_child_count;
494 uint64_t io_phys_children;
495 uint64_t io_parent_count;
496 uint64_t *io_stall;
497 zio_t *io_gang_leader;
498 zio_gang_node_t *io_gang_tree;
499 void *io_executor;
500 void *io_waiter;
501 kmutex_t io_lock;
502 kcondvar_t io_cv;
503
504 /* FMA state */
505 zio_cksum_report_t *io_cksum_report;
506 uint64_t io_ena;
507
508 /* Taskq dispatching state */
509 taskq_ent_t io_tqent;
510
511 /* Timestamp for tracking vdev I/O latency */
512 hrtime_t io_vd_timestamp;
513
514 /* Checksum acceleration */
515 zio_checksum_state_t zio_checksum_state;
516 zio_cksum_t *zio_checksump;
517 void *zio_checksum_datap;
518 uint64_t zio_checksum_data_size;
519 struct zio *zio_checksum_next;
520 zio_cksum_t actual_cksum;
521
522 /* Metaslab class that will be used */
523 metaslab_class_t *io_mc;
524 };
525
526 extern int zio_bookmark_compare(const void *, const void *);
527
528 extern zio_t *zio_null(zio_t *pio, spa_t *spa, vdev_t *vd,
529 zio_done_func_t *done, void *private, enum zio_flag flags);
530
531 extern zio_t *zio_root(spa_t *spa,
532 zio_done_func_t *done, void *private, enum zio_flag flags);
533
534 extern zio_t *zio_read(zio_t *pio, spa_t *spa, const blkptr_t *bp,
535 struct abd *data, uint64_t lsize, zio_done_func_t *done, void *private,
536 zio_priority_t priority, enum zio_flag flags, const zbookmark_phys_t *zb);
537
538 extern zio_t *zio_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
539 struct abd *data, uint64_t size, uint64_t psize, const zio_prop_t *zp,
540 zio_done_func_t *ready, zio_done_func_t *children_ready,
541 zio_done_func_t *physdone, zio_done_func_t *done,
542 void *private, zio_priority_t priority, enum zio_flag flags,
543 const zbookmark_phys_t *zb,
544 const zio_smartcomp_info_t *smartcomp);
545
546 extern zio_t *zio_rewrite(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
547 struct abd *data, uint64_t size, zio_done_func_t *done, void *private,
548 zio_priority_t priority, enum zio_flag flags, zbookmark_phys_t *zb);
549
550 extern void zio_write_override(zio_t *zio, blkptr_t *bp, int copies,
551 boolean_t nopwrite);
552
553 extern void zio_free(spa_t *spa, uint64_t txg, const blkptr_t *bp);
554
555 extern zio_t *zio_claim(zio_t *pio, spa_t *spa, uint64_t txg,
556 const blkptr_t *bp,
557 zio_done_func_t *done, void *private, enum zio_flag flags);
558
559 extern zio_t *zio_ioctl(zio_t *pio, spa_t *spa, vdev_t *vd, int cmd,
560 zio_done_func_t *done, void *private, enum zio_flag flags);
561
562 extern zio_t *zio_trim(spa_t *spa, vdev_t *vd, struct range_tree *tree,
563 zio_done_func_t *done, void *private, enum zio_flag flags,
564 int dkiocfree_flags, metaslab_t *msp);
565
566 extern zio_t *zio_read_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
567 uint64_t size, struct abd *data, int checksum,
568 zio_done_func_t *done, void *private, zio_priority_t priority,
569 enum zio_flag flags, boolean_t labels);
570
571 extern zio_t *zio_write_phys(zio_t *pio, vdev_t *vd, uint64_t offset,
572 uint64_t size, struct abd *data, int checksum,
573 zio_done_func_t *done, void *private, zio_priority_t priority,
574 enum zio_flag flags, boolean_t labels);
575
576 extern zio_t *zio_wbc(zio_type_t type, vdev_t *vd, abd_t *data,
577 uint64_t size, uint64_t offset);
578
579 extern zio_t *zio_free_sync(zio_t *pio, spa_t *spa, uint64_t txg,
580 const blkptr_t *bp, enum zio_flag flags);
581
582 extern int zio_alloc_zil(spa_t *spa, uint64_t txg, blkptr_t *new_bp,
583 blkptr_t *old_bp, uint64_t size, boolean_t *slog);
584 extern void zio_free_zil(spa_t *spa, uint64_t txg, blkptr_t *bp);
585 extern void zio_flush(zio_t *zio, vdev_t *vd);
586 extern void zio_shrink(zio_t *zio, uint64_t size);
587
588 extern int zio_wait(zio_t *zio);
589 extern void zio_nowait(zio_t *zio);
590 extern void zio_execute(zio_t *zio);
591 extern void zio_interrupt(zio_t *zio);
592 extern void zio_delay_init(zio_t *zio);
593 extern void zio_delay_interrupt(zio_t *zio);
594
595 extern zio_t *zio_walk_parents(zio_t *cio, zio_link_t **);
596 extern zio_t *zio_walk_children(zio_t *pio, zio_link_t **);
597 extern zio_t *zio_unique_parent(zio_t *cio);
598 extern void zio_add_child(zio_t *pio, zio_t *cio);
651 extern int zio_inject_list_next(int *id, char *name, size_t buflen,
652 struct zinject_record *record);
653 extern int zio_clear_fault(int id);
654 extern void zio_handle_panic_injection(spa_t *spa, char *tag, uint64_t type);
655 extern int zio_handle_fault_injection(zio_t *zio, int error);
656 extern int zio_handle_device_injection(vdev_t *vd, zio_t *zio, int error);
657 extern int zio_handle_label_injection(zio_t *zio, int error);
658 extern void zio_handle_ignored_writes(zio_t *zio);
659 extern hrtime_t zio_handle_io_delay(zio_t *zio);
660
661 /*
662 * Checksum ereport functions
663 */
664 extern void zfs_ereport_start_checksum(spa_t *spa, vdev_t *vd, struct zio *zio,
665 uint64_t offset, uint64_t length, void *arg, struct zio_bad_cksum *info);
666 extern void zfs_ereport_finish_checksum(zio_cksum_report_t *report,
667 const void *good_data, const void *bad_data, boolean_t drop_if_identical);
668
669 extern void zfs_ereport_send_interim_checksum(zio_cksum_report_t *report);
670 extern void zfs_ereport_free_checksum(zio_cksum_report_t *report);
671 /* If we have the good data in hand, this function can be used */
672 extern void zfs_ereport_post_checksum(spa_t *spa, vdev_t *vd,
673 struct zio *zio, uint64_t offset, uint64_t length,
674 const void *good_data, const void *bad_data, struct zio_bad_cksum *info);
675
676 /* Called from spa_sync(), but primarily an injection handler */
677 extern void spa_handle_ignored_writes(spa_t *spa);
678
679 /* zbookmark_phys functions */
680 boolean_t zbookmark_subtree_completed(const struct dnode_phys *dnp,
681 const zbookmark_phys_t *subtree_root, const zbookmark_phys_t *last_block);
682 int zbookmark_compare(uint16_t dbss1, uint8_t ibs1, uint16_t dbss2,
683 uint8_t ibs2, const zbookmark_phys_t *zb1, const zbookmark_phys_t *zb2);
684
685 /* best effort dedup */
686 void zio_best_effort_dedup(zio_t *zio);
687
688 #ifdef __cplusplus
689 }
690 #endif
691
692 #endif /* _ZIO_H */
|