4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2012, 2020 by Delphix. All rights reserved.
24 * Copyright 2016 RackTop Systems.
25 * Copyright (c) 2017, Intel Corporation.
28 #ifndef _SYS_ZFS_IOCTL_H
29 #define _SYS_ZFS_IOCTL_H
34 #include <sys/dsl_deleg.h>
36 #include <sys/zfs_stat.h>
39 #include <sys/nvpair.h>
47 * The structures in this file are passed between userland and the
48 * kernel. Userland may be running a 32-bit process, while the kernel
49 * is 64-bit. Therefore, these structures need to compile the same in
50 * 32-bit and 64-bit. This means not using type "long", and adding
51 * explicit padding so that the 32-bit structure will not be packed more
52 * tightly than the 64-bit structure (which requires 64-bit alignment).
56 * Property values for snapdir
58 #define ZFS_SNAPDIR_HIDDEN 0
59 #define ZFS_SNAPDIR_VISIBLE 1
62 * Property values for snapdev
64 #define ZFS_SNAPDEV_HIDDEN 0
65 #define ZFS_SNAPDEV_VISIBLE 1
67 * Property values for acltype
69 #define ZFS_ACLTYPE_OFF 0
70 #define ZFS_ACLTYPE_POSIX 1
73 * Field manipulation macros for the drr_versioninfo field of the
78 * Header types for zfs send streams.
80 typedef enum drr_headertype {
82 DMU_COMPOUNDSTREAM = 0x2
85 #define DMU_GET_STREAM_HDRTYPE(vi) BF64_GET((vi), 0, 2)
86 #define DMU_SET_STREAM_HDRTYPE(vi, x) BF64_SET((vi), 0, 2, x)
88 #define DMU_GET_FEATUREFLAGS(vi) BF64_GET((vi), 2, 30)
89 #define DMU_SET_FEATUREFLAGS(vi, x) BF64_SET((vi), 2, 30, x)
92 * Feature flags for zfs send streams (flags in drr_versioninfo)
95 #define DMU_BACKUP_FEATURE_DEDUP (1 << 0)
96 #define DMU_BACKUP_FEATURE_DEDUPPROPS (1 << 1)
97 #define DMU_BACKUP_FEATURE_SA_SPILL (1 << 2)
98 /* flags #3 - #15 are reserved for incompatible closed-source implementations */
99 #define DMU_BACKUP_FEATURE_EMBED_DATA (1 << 16)
100 #define DMU_BACKUP_FEATURE_LZ4 (1 << 17)
101 /* flag #18 is reserved for a Delphix feature */
102 #define DMU_BACKUP_FEATURE_LARGE_BLOCKS (1 << 19)
103 #define DMU_BACKUP_FEATURE_RESUMING (1 << 20)
104 #define DMU_BACKUP_FEATURE_REDACTED (1 << 21)
105 #define DMU_BACKUP_FEATURE_COMPRESSED (1 << 22)
106 #define DMU_BACKUP_FEATURE_LARGE_DNODE (1 << 23)
107 #define DMU_BACKUP_FEATURE_RAW (1 << 24)
108 #define DMU_BACKUP_FEATURE_ZSTD (1 << 25)
109 #define DMU_BACKUP_FEATURE_HOLDS (1 << 26)
111 * The SWITCH_TO_LARGE_BLOCKS feature indicates that we can receive
112 * incremental LARGE_BLOCKS streams (those with WRITE records of >128KB) even
113 * if the previous send did not use LARGE_BLOCKS, and thus its large blocks
114 * were split into multiple 128KB WRITE records. (See
115 * flush_write_batch_impl() and receive_object()). Older software that does
116 * not support this flag may encounter a bug when switching to large blocks,
117 * which causes files to incorrectly be zeroed.
119 * This flag is currently not set on any send streams. In the future, we
120 * intend for incremental send streams of snapshots that have large blocks to
121 * use LARGE_BLOCKS by default, and these streams will also have the
122 * SWITCH_TO_LARGE_BLOCKS feature set. This ensures that streams from the
123 * default use of "zfs send" won't encounter the bug mentioned above.
125 #define DMU_BACKUP_FEATURE_SWITCH_TO_LARGE_BLOCKS (1 << 27)
128 * Mask of all supported backup features
130 #define DMU_BACKUP_FEATURE_MASK (DMU_BACKUP_FEATURE_SA_SPILL | \
131 DMU_BACKUP_FEATURE_EMBED_DATA | DMU_BACKUP_FEATURE_LZ4 | \
132 DMU_BACKUP_FEATURE_RESUMING | DMU_BACKUP_FEATURE_LARGE_BLOCKS | \
133 DMU_BACKUP_FEATURE_COMPRESSED | DMU_BACKUP_FEATURE_LARGE_DNODE | \
134 DMU_BACKUP_FEATURE_RAW | DMU_BACKUP_FEATURE_HOLDS | \
135 DMU_BACKUP_FEATURE_REDACTED | DMU_BACKUP_FEATURE_SWITCH_TO_LARGE_BLOCKS | \
136 DMU_BACKUP_FEATURE_ZSTD)
138 /* Are all features in the given flag word currently supported? */
139 #define DMU_STREAM_SUPPORTED(x) (!((x) & ~DMU_BACKUP_FEATURE_MASK))
141 typedef enum dmu_send_resume_token_version {
142 ZFS_SEND_RESUME_TOKEN_VERSION = 1
143 } dmu_send_resume_token_version_t;
146 * The drr_versioninfo field of the dmu_replay_record has the
149 * 64 56 48 40 32 24 16 8 0
150 * +-------+-------+-------+-------+-------+-------+-------+-------+
151 * | reserved | feature-flags |C|S|
152 * +-------+-------+-------+-------+-------+-------+-------+-------+
154 * The low order two bits indicate the header type: SUBSTREAM (0x1)
155 * or COMPOUNDSTREAM (0x2). Using two bits for this is historical:
156 * this field used to be a version number, where the two version types
157 * were 1 and 2. Using two bits for this allows earlier versions of
158 * the code to be able to recognize send streams that don't use any
159 * of the features indicated by feature flags.
162 #define DMU_BACKUP_MAGIC 0x2F5bacbacULL
165 * Send stream flags. Bits 24-31 are reserved for vendor-specific
166 * implementations and should not be used.
168 #define DRR_FLAG_CLONE (1<<0)
169 #define DRR_FLAG_CI_DATA (1<<1)
171 * This send stream, if it is a full send, includes the FREE and FREEOBJECT
172 * records that are created by the sending process. This means that the send
173 * stream can be received as a clone, even though it is not an incremental.
174 * This is not implemented as a feature flag, because the receiving side does
175 * not need to have implemented it to receive this stream; it is fully backwards
176 * compatible. We need a flag, though, because full send streams without it
177 * cannot necessarily be received as a clone correctly.
179 #define DRR_FLAG_FREERECORDS (1<<2)
181 * When DRR_FLAG_SPILL_BLOCK is set it indicates the DRR_OBJECT_SPILL
182 * and DRR_SPILL_UNMODIFIED flags are meaningful in the send stream.
184 * When DRR_FLAG_SPILL_BLOCK is set, DRR_OBJECT records will have
185 * DRR_OBJECT_SPILL set if and only if they should have a spill block
186 * (either an existing one, or a new one in the send stream). When clear
187 * the object does not have a spill block and any existing spill block
190 * Similarly, when DRR_FLAG_SPILL_BLOCK is set, DRR_SPILL records will
191 * have DRR_SPILL_UNMODIFIED set if and only if they were included for
192 * backward compatibility purposes, and can be safely ignored by new versions
193 * of zfs receive. Previous versions of ZFS which do not understand the
194 * DRR_FLAG_SPILL_BLOCK will process this record and recreate any missing
197 #define DRR_FLAG_SPILL_BLOCK (1<<3)
200 * flags in the drr_flags field in the DRR_WRITE, DRR_SPILL, DRR_OBJECT,
201 * DRR_WRITE_BYREF, and DRR_OBJECT_RANGE blocks
203 #define DRR_CHECKSUM_DEDUP (1<<0) /* not used for SPILL records */
204 #define DRR_RAW_BYTESWAP (1<<1)
205 #define DRR_OBJECT_SPILL (1<<2) /* OBJECT record has a spill block */
206 #define DRR_SPILL_UNMODIFIED (1<<2) /* SPILL record for unmodified block */
208 #define DRR_IS_DEDUP_CAPABLE(flags) ((flags) & DRR_CHECKSUM_DEDUP)
209 #define DRR_IS_RAW_BYTESWAPPED(flags) ((flags) & DRR_RAW_BYTESWAP)
210 #define DRR_OBJECT_HAS_SPILL(flags) ((flags) & DRR_OBJECT_SPILL)
211 #define DRR_SPILL_IS_UNMODIFIED(flags) ((flags) & DRR_SPILL_UNMODIFIED)
213 /* deal with compressed drr_write replay records */
214 #define DRR_WRITE_COMPRESSED(drrw) ((drrw)->drr_compressiontype != 0)
215 #define DRR_WRITE_PAYLOAD_SIZE(drrw) \
216 (DRR_WRITE_COMPRESSED(drrw) ? (drrw)->drr_compressed_size : \
217 (drrw)->drr_logical_size)
218 #define DRR_SPILL_PAYLOAD_SIZE(drrs) \
219 ((drrs)->drr_compressed_size ? \
220 (drrs)->drr_compressed_size : (drrs)->drr_length)
221 #define DRR_OBJECT_PAYLOAD_SIZE(drro) \
222 ((drro)->drr_raw_bonuslen != 0 ? \
223 (drro)->drr_raw_bonuslen : P2ROUNDUP((drro)->drr_bonuslen, 8))
226 * zfs ioctl command structure
229 /* Header is used in C++ so can't forward declare untagged struct */
232 uint64_t drr_versioninfo; /* was drr_version */
233 uint64_t drr_creation_time;
234 dmu_objset_type_t drr_type;
237 uint64_t drr_fromguid;
238 char drr_toname[MAXNAMELEN];
241 typedef struct dmu_replay_record {
243 DRR_BEGIN, DRR_OBJECT, DRR_FREEOBJECTS,
244 DRR_WRITE, DRR_FREE, DRR_END, DRR_WRITE_BYREF,
245 DRR_SPILL, DRR_WRITE_EMBEDDED, DRR_OBJECT_RANGE, DRR_REDACT,
248 uint32_t drr_payloadlen;
250 struct drr_begin drr_begin;
252 zio_cksum_t drr_checksum;
257 dmu_object_type_t drr_type;
258 dmu_object_type_t drr_bonustype;
260 uint32_t drr_bonuslen;
261 uint8_t drr_checksumtype;
262 uint8_t drr_compress;
263 uint8_t drr_dn_slots;
265 uint32_t drr_raw_bonuslen;
267 /* only (possibly) nonzero for raw streams */
268 uint8_t drr_indblkshift;
272 uint64_t drr_maxblkid;
273 /* bonus content follows */
275 struct drr_freeobjects {
276 uint64_t drr_firstobj;
277 uint64_t drr_numobjs;
282 dmu_object_type_t drr_type;
285 uint64_t drr_logical_size;
287 uint8_t drr_checksumtype;
289 uint8_t drr_compressiontype;
291 /* deduplication key */
293 /* only nonzero if drr_compressiontype is not 0 */
294 uint64_t drr_compressed_size;
295 /* only nonzero for raw streams */
296 uint8_t drr_salt[ZIO_DATA_SALT_LEN];
297 uint8_t drr_iv[ZIO_DATA_IV_LEN];
298 uint8_t drr_mac[ZIO_DATA_MAC_LEN];
299 /* content follows */
307 struct drr_write_byref {
308 /* where to put the data */
313 /* where to find the prior copy of the data */
314 uint64_t drr_refguid;
315 uint64_t drr_refobject;
316 uint64_t drr_refoffset;
317 /* properties of the data */
318 uint8_t drr_checksumtype;
321 ddt_key_t drr_key; /* deduplication key */
328 uint8_t drr_compressiontype;
330 /* only nonzero for raw streams */
331 uint64_t drr_compressed_size;
332 uint8_t drr_salt[ZIO_DATA_SALT_LEN];
333 uint8_t drr_iv[ZIO_DATA_IV_LEN];
334 uint8_t drr_mac[ZIO_DATA_MAC_LEN];
335 dmu_object_type_t drr_type;
336 /* spill data follows */
338 struct drr_write_embedded {
341 /* logical length, should equal blocksize */
344 uint8_t drr_compression;
347 uint32_t drr_lsize; /* uncompressed size of payload */
348 uint32_t drr_psize; /* compr. (real) size of payload */
349 /* (possibly compressed) content follows */
350 } drr_write_embedded;
351 struct drr_object_range {
352 uint64_t drr_firstobj;
353 uint64_t drr_numslots;
355 uint8_t drr_salt[ZIO_DATA_SALT_LEN];
356 uint8_t drr_iv[ZIO_DATA_IV_LEN];
357 uint8_t drr_mac[ZIO_DATA_MAC_LEN];
369 * Note: drr_checksum is overlaid with all record types
370 * except DRR_BEGIN. Therefore its (non-pad) members
371 * must not overlap with members from the other structs.
372 * We accomplish this by putting its members at the very
375 struct drr_checksum {
376 uint64_t drr_pad[34];
378 * fletcher-4 checksum of everything preceding the
381 zio_cksum_t drr_checksum;
384 } dmu_replay_record_t;
386 /* diff record range types */
387 typedef enum diff_type {
394 * The diff reports back ranges of free or in-use objects.
396 typedef struct dmu_diff_record {
402 typedef struct zinject_record {
412 uint32_t zi_failfast;
413 char zi_func[MAXNAMELEN];
422 #define ZINJECT_NULL 0x1
423 #define ZINJECT_FLUSH_ARC 0x2
424 #define ZINJECT_UNLOAD_SPA 0x4
425 #define ZINJECT_CALC_RANGE 0x8
427 #define ZEVENT_NONE 0x0
428 #define ZEVENT_NONBLOCK 0x1
429 #define ZEVENT_SIZE 1024
431 #define ZEVENT_SEEK_START 0
432 #define ZEVENT_SEEK_END UINT64_MAX
434 /* scaled frequency ranges */
435 #define ZI_PERCENTAGE_MIN 4294UL
436 #define ZI_PERCENTAGE_MAX UINT32_MAX
438 #define ZI_NO_DVA (-1)
440 typedef enum zinject_type {
441 ZINJECT_UNINITIALIZED,
443 ZINJECT_DEVICE_FAULT,
445 ZINJECT_IGNORED_WRITES,
448 ZINJECT_DECRYPT_FAULT,
451 typedef struct zfs_share {
452 uint64_t z_exportdata;
453 uint64_t z_sharedata;
454 uint64_t z_sharetype; /* 0 = share, 1 = unshare */
455 uint64_t z_sharemax; /* max length of share string */
459 * ZFS file systems may behave the usual, POSIX-compliant way, where
460 * name lookups are case-sensitive. They may also be set up so that
461 * all the name lookups are case-insensitive, or so that only some
462 * lookups, the ones that set an FIGNORECASE flag, are case-insensitive.
464 typedef enum zfs_case {
466 ZFS_CASE_INSENSITIVE,
471 * Note: this struct must have the same layout in 32-bit and 64-bit, so
472 * that 32-bit processes (like /sbin/zfs) can pass it to the 64-bit
473 * kernel. Therefore, we add padding to it so that no "hidden" padding
474 * is automatically added on 64-bit (but not on 32-bit).
476 typedef struct zfs_cmd {
477 char zc_name[MAXPATHLEN]; /* name of pool or dataset */
478 uint64_t zc_nvlist_src; /* really (char *) */
479 uint64_t zc_nvlist_src_size;
480 uint64_t zc_nvlist_dst; /* really (char *) */
481 uint64_t zc_nvlist_dst_size;
482 boolean_t zc_nvlist_dst_filled; /* put an nvlist in dst? */
486 * The following members are for legacy ioctls which haven't been
487 * converted to the new method.
489 uint64_t zc_history; /* really (char *) */
490 char zc_value[MAXPATHLEN * 2];
491 char zc_string[MAXNAMELEN];
493 uint64_t zc_nvlist_conf; /* really (char *) */
494 uint64_t zc_nvlist_conf_size;
496 uint64_t zc_objset_type;
497 uint64_t zc_perm_action;
498 uint64_t zc_history_len;
499 uint64_t zc_history_offset;
501 uint64_t zc_iflags; /* internal to zfs(7fs) */
502 zfs_share_t zc_share;
503 dmu_objset_stats_t zc_objset_stats;
504 struct drr_begin zc_begin_record;
505 zinject_record_t zc_inject_record;
506 uint32_t zc_defer_destroy;
508 uint64_t zc_action_handle;
511 uint8_t zc_pad[3]; /* alignment */
514 uint64_t zc_createtxg;
519 typedef struct zfs_useracct {
526 #define ZFSDEV_MAX_MINOR (1 << 16)
527 #define ZFS_MIN_MINOR (ZFSDEV_MAX_MINOR + 1)
529 #define ZPOOL_EXPORT_AFTER_SPLIT 0x1
535 typedef struct zfs_creat {
536 nvlist_t *zct_zplprops;
540 extern int zfs_secpolicy_snapshot_perms(const char *, cred_t *);
541 extern int zfs_secpolicy_rename_perms(const char *, const char *, cred_t *);
542 extern int zfs_secpolicy_destroy_perms(const char *, cred_t *);
543 extern void zfs_unmount_snap(const char *);
544 extern void zfs_destroy_unmount_origin(const char *);
545 extern int getzfsvfs_impl(struct objset *, struct zfsvfs **);
546 extern int getzfsvfs(const char *, struct zfsvfs **);
548 enum zfsdev_state_type {
555 * The zfsdev_state_t structure is managed as a singly-linked list
556 * from which items are never deleted. This allows for lock-free
557 * reading of the list so long as assignments to the zs_next and
558 * reads from zs_minor are performed atomically. Empty items are
559 * indicated by storing -1 into zs_minor.
561 typedef struct zfsdev_state {
562 struct zfsdev_state *zs_next; /* next zfsdev_state_t link */
563 minor_t zs_minor; /* made up minor number */
564 void *zs_onexit; /* onexit data */
565 void *zs_zevent; /* zevent data */
568 extern void *zfsdev_get_state(minor_t minor, enum zfsdev_state_type which);
569 extern int zfsdev_getminor(int fd, minor_t *minorp);
570 extern minor_t zfsdev_minor_alloc(void);
572 extern uint_t zfs_fsyncer_key;
573 extern uint_t zfs_allow_log_key;
581 #endif /* _SYS_ZFS_IOCTL_H */