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 https://opensource.org/licenses/CDDL-1.0.
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, 2024 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
71 #define ZFS_ACLTYPE_NFSV4 2
74 * Field manipulation macros for the drr_versioninfo field of the
79 * Header types for zfs send streams.
81 typedef enum drr_headertype {
83 DMU_COMPOUNDSTREAM = 0x2
86 #define DMU_GET_STREAM_HDRTYPE(vi) BF64_GET((vi), 0, 2)
87 #define DMU_SET_STREAM_HDRTYPE(vi, x) BF64_SET((vi), 0, 2, x)
89 #define DMU_GET_FEATUREFLAGS(vi) BF64_GET((vi), 2, 30)
90 #define DMU_SET_FEATUREFLAGS(vi, x) BF64_SET((vi), 2, 30, x)
93 * Feature flags for zfs send streams (flags in drr_versioninfo)
96 #define DMU_BACKUP_FEATURE_DEDUP (1 << 0)
97 #define DMU_BACKUP_FEATURE_DEDUPPROPS (1 << 1)
98 #define DMU_BACKUP_FEATURE_SA_SPILL (1 << 2)
99 /* flags #3 - #15 are reserved for incompatible closed-source implementations */
100 #define DMU_BACKUP_FEATURE_EMBED_DATA (1 << 16)
101 #define DMU_BACKUP_FEATURE_LZ4 (1 << 17)
102 /* flag #18 is reserved for a Delphix feature */
103 #define DMU_BACKUP_FEATURE_LARGE_BLOCKS (1 << 19)
104 #define DMU_BACKUP_FEATURE_RESUMING (1 << 20)
105 #define DMU_BACKUP_FEATURE_REDACTED (1 << 21)
106 #define DMU_BACKUP_FEATURE_COMPRESSED (1 << 22)
107 #define DMU_BACKUP_FEATURE_LARGE_DNODE (1 << 23)
108 #define DMU_BACKUP_FEATURE_RAW (1 << 24)
109 #define DMU_BACKUP_FEATURE_ZSTD (1 << 25)
110 #define DMU_BACKUP_FEATURE_HOLDS (1 << 26)
112 * The SWITCH_TO_LARGE_BLOCKS feature indicates that we can receive
113 * incremental LARGE_BLOCKS streams (those with WRITE records of >128KB) even
114 * if the previous send did not use LARGE_BLOCKS, and thus its large blocks
115 * were split into multiple 128KB WRITE records. (See
116 * flush_write_batch_impl() and receive_object()). Older software that does
117 * not support this flag may encounter a bug when switching to large blocks,
118 * which causes files to incorrectly be zeroed.
120 * This flag is currently not set on any send streams. In the future, we
121 * intend for incremental send streams of snapshots that have large blocks to
122 * use LARGE_BLOCKS by default, and these streams will also have the
123 * SWITCH_TO_LARGE_BLOCKS feature set. This ensures that streams from the
124 * default use of "zfs send" won't encounter the bug mentioned above.
126 #define DMU_BACKUP_FEATURE_SWITCH_TO_LARGE_BLOCKS (1 << 27)
127 /* flag #28 is reserved for a Nutanix feature */
129 * flag #29 is the last unused bit. It is reserved to indicate a to-be-designed
130 * extension to the stream format which will accomodate more feature flags.
131 * If you need to add another feature flag, please reach out to the OpenZFS
132 * community, e.g., on GitHub or Slack.
136 * Mask of all supported backup features
138 #define DMU_BACKUP_FEATURE_MASK (DMU_BACKUP_FEATURE_SA_SPILL | \
139 DMU_BACKUP_FEATURE_EMBED_DATA | DMU_BACKUP_FEATURE_LZ4 | \
140 DMU_BACKUP_FEATURE_RESUMING | DMU_BACKUP_FEATURE_LARGE_BLOCKS | \
141 DMU_BACKUP_FEATURE_COMPRESSED | DMU_BACKUP_FEATURE_LARGE_DNODE | \
142 DMU_BACKUP_FEATURE_RAW | DMU_BACKUP_FEATURE_HOLDS | \
143 DMU_BACKUP_FEATURE_REDACTED | DMU_BACKUP_FEATURE_SWITCH_TO_LARGE_BLOCKS | \
144 DMU_BACKUP_FEATURE_ZSTD)
146 /* Are all features in the given flag word currently supported? */
147 #define DMU_STREAM_SUPPORTED(x) (!((x) & ~DMU_BACKUP_FEATURE_MASK))
149 typedef enum dmu_send_resume_token_version {
150 ZFS_SEND_RESUME_TOKEN_VERSION = 1
151 } dmu_send_resume_token_version_t;
154 * The drr_versioninfo field of the dmu_replay_record has the
157 * 64 56 48 40 32 24 16 8 0
158 * +-------+-------+-------+-------+-------+-------+-------+-------+
159 * | reserved | feature-flags |C|S|
160 * +-------+-------+-------+-------+-------+-------+-------+-------+
162 * The low order two bits indicate the header type: SUBSTREAM (0x1)
163 * or COMPOUNDSTREAM (0x2). Using two bits for this is historical:
164 * this field used to be a version number, where the two version types
165 * were 1 and 2. Using two bits for this allows earlier versions of
166 * the code to be able to recognize send streams that don't use any
167 * of the features indicated by feature flags.
170 #define DMU_BACKUP_MAGIC 0x2F5bacbacULL
173 * Send stream flags. Bits 24-31 are reserved for vendor-specific
174 * implementations and should not be used.
176 #define DRR_FLAG_CLONE (1<<0)
177 #define DRR_FLAG_CI_DATA (1<<1)
179 * This send stream, if it is a full send, includes the FREE and FREEOBJECT
180 * records that are created by the sending process. This means that the send
181 * stream can be received as a clone, even though it is not an incremental.
182 * This is not implemented as a feature flag, because the receiving side does
183 * not need to have implemented it to receive this stream; it is fully backwards
184 * compatible. We need a flag, though, because full send streams without it
185 * cannot necessarily be received as a clone correctly.
187 #define DRR_FLAG_FREERECORDS (1<<2)
189 * When DRR_FLAG_SPILL_BLOCK is set it indicates the DRR_OBJECT_SPILL
190 * and DRR_SPILL_UNMODIFIED flags are meaningful in the send stream.
192 * When DRR_FLAG_SPILL_BLOCK is set, DRR_OBJECT records will have
193 * DRR_OBJECT_SPILL set if and only if they should have a spill block
194 * (either an existing one, or a new one in the send stream). When clear
195 * the object does not have a spill block and any existing spill block
198 * Similarly, when DRR_FLAG_SPILL_BLOCK is set, DRR_SPILL records will
199 * have DRR_SPILL_UNMODIFIED set if and only if they were included for
200 * backward compatibility purposes, and can be safely ignored by new versions
201 * of zfs receive. Previous versions of ZFS which do not understand the
202 * DRR_FLAG_SPILL_BLOCK will process this record and recreate any missing
205 #define DRR_FLAG_SPILL_BLOCK (1<<3)
208 * flags in the drr_flags field in the DRR_WRITE, DRR_SPILL, DRR_OBJECT,
209 * DRR_WRITE_BYREF, and DRR_OBJECT_RANGE blocks
211 #define DRR_CHECKSUM_DEDUP (1<<0) /* not used for SPILL records */
212 #define DRR_RAW_BYTESWAP (1<<1)
213 #define DRR_OBJECT_SPILL (1<<2) /* OBJECT record has a spill block */
214 #define DRR_SPILL_UNMODIFIED (1<<2) /* SPILL record for unmodified block */
216 #define DRR_IS_DEDUP_CAPABLE(flags) ((flags) & DRR_CHECKSUM_DEDUP)
217 #define DRR_IS_RAW_BYTESWAPPED(flags) ((flags) & DRR_RAW_BYTESWAP)
218 #define DRR_OBJECT_HAS_SPILL(flags) ((flags) & DRR_OBJECT_SPILL)
219 #define DRR_SPILL_IS_UNMODIFIED(flags) ((flags) & DRR_SPILL_UNMODIFIED)
221 /* deal with compressed drr_write replay records */
222 #define DRR_WRITE_COMPRESSED(drrw) ((drrw)->drr_compressiontype != 0)
223 #define DRR_WRITE_PAYLOAD_SIZE(drrw) \
224 (DRR_WRITE_COMPRESSED(drrw) ? (drrw)->drr_compressed_size : \
225 (drrw)->drr_logical_size)
226 #define DRR_SPILL_PAYLOAD_SIZE(drrs) \
227 ((drrs)->drr_compressed_size ? \
228 (drrs)->drr_compressed_size : (drrs)->drr_length)
229 #define DRR_OBJECT_PAYLOAD_SIZE(drro) \
230 ((drro)->drr_raw_bonuslen != 0 ? \
231 (drro)->drr_raw_bonuslen : P2ROUNDUP((drro)->drr_bonuslen, 8))
234 * zfs ioctl command structure
237 /* Header is used in C++ so can't forward declare untagged struct */
240 uint64_t drr_versioninfo; /* was drr_version */
241 uint64_t drr_creation_time;
242 dmu_objset_type_t drr_type;
245 uint64_t drr_fromguid;
246 char drr_toname[MAXNAMELEN];
249 typedef struct dmu_replay_record {
251 DRR_BEGIN, DRR_OBJECT, DRR_FREEOBJECTS,
252 DRR_WRITE, DRR_FREE, DRR_END, DRR_WRITE_BYREF,
253 DRR_SPILL, DRR_WRITE_EMBEDDED, DRR_OBJECT_RANGE, DRR_REDACT,
256 uint32_t drr_payloadlen;
258 struct drr_begin drr_begin;
260 zio_cksum_t drr_checksum;
265 dmu_object_type_t drr_type;
266 dmu_object_type_t drr_bonustype;
268 uint32_t drr_bonuslen;
269 uint8_t drr_checksumtype;
270 uint8_t drr_compress;
271 uint8_t drr_dn_slots;
273 uint32_t drr_raw_bonuslen;
275 /* only (possibly) nonzero for raw streams */
276 uint8_t drr_indblkshift;
280 uint64_t drr_maxblkid;
281 /* bonus content follows */
283 struct drr_freeobjects {
284 uint64_t drr_firstobj;
285 uint64_t drr_numobjs;
290 dmu_object_type_t drr_type;
293 uint64_t drr_logical_size;
295 uint8_t drr_checksumtype;
297 uint8_t drr_compressiontype;
299 /* deduplication key */
301 /* only nonzero if drr_compressiontype is not 0 */
302 uint64_t drr_compressed_size;
303 /* only nonzero for raw streams */
304 uint8_t drr_salt[ZIO_DATA_SALT_LEN];
305 uint8_t drr_iv[ZIO_DATA_IV_LEN];
306 uint8_t drr_mac[ZIO_DATA_MAC_LEN];
307 /* content follows */
315 struct drr_write_byref {
316 /* where to put the data */
321 /* where to find the prior copy of the data */
322 uint64_t drr_refguid;
323 uint64_t drr_refobject;
324 uint64_t drr_refoffset;
325 /* properties of the data */
326 uint8_t drr_checksumtype;
329 ddt_key_t drr_key; /* deduplication key */
336 uint8_t drr_compressiontype;
338 /* only nonzero for raw streams */
339 uint64_t drr_compressed_size;
340 uint8_t drr_salt[ZIO_DATA_SALT_LEN];
341 uint8_t drr_iv[ZIO_DATA_IV_LEN];
342 uint8_t drr_mac[ZIO_DATA_MAC_LEN];
343 dmu_object_type_t drr_type;
344 /* spill data follows */
346 struct drr_write_embedded {
349 /* logical length, should equal blocksize */
352 uint8_t drr_compression;
355 uint32_t drr_lsize; /* uncompressed size of payload */
356 uint32_t drr_psize; /* compr. (real) size of payload */
357 /* (possibly compressed) content follows */
358 } drr_write_embedded;
359 struct drr_object_range {
360 uint64_t drr_firstobj;
361 uint64_t drr_numslots;
363 uint8_t drr_salt[ZIO_DATA_SALT_LEN];
364 uint8_t drr_iv[ZIO_DATA_IV_LEN];
365 uint8_t drr_mac[ZIO_DATA_MAC_LEN];
377 * Note: drr_checksum is overlaid with all record types
378 * except DRR_BEGIN. Therefore its (non-pad) members
379 * must not overlap with members from the other structs.
380 * We accomplish this by putting its members at the very
383 struct drr_checksum {
384 uint64_t drr_pad[34];
386 * fletcher-4 checksum of everything preceding the
389 zio_cksum_t drr_checksum;
392 } dmu_replay_record_t;
394 /* diff record range types */
395 typedef enum diff_type {
402 * The diff reports back ranges of free or in-use objects.
404 typedef struct dmu_diff_record {
410 typedef struct zinject_record {
420 uint32_t zi_failfast;
421 char zi_func[MAXNAMELEN];
430 #define ZINJECT_NULL 0x1
431 #define ZINJECT_FLUSH_ARC 0x2
432 #define ZINJECT_UNLOAD_SPA 0x4
433 #define ZINJECT_CALC_RANGE 0x8
435 #define ZEVENT_NONE 0x0
436 #define ZEVENT_NONBLOCK 0x1
437 #define ZEVENT_SIZE 1024
439 #define ZEVENT_SEEK_START 0
440 #define ZEVENT_SEEK_END UINT64_MAX
442 /* scaled frequency ranges */
443 #define ZI_PERCENTAGE_MIN 4294UL
444 #define ZI_PERCENTAGE_MAX UINT32_MAX
446 #define ZI_NO_DVA (-1)
448 typedef enum zinject_type {
449 ZINJECT_UNINITIALIZED,
451 ZINJECT_DEVICE_FAULT,
453 ZINJECT_IGNORED_WRITES,
456 ZINJECT_DECRYPT_FAULT,
457 ZINJECT_DELAY_IMPORT,
458 ZINJECT_DELAY_EXPORT,
461 typedef struct zfs_share {
462 uint64_t z_exportdata;
463 uint64_t z_sharedata;
464 uint64_t z_sharetype; /* 0 = share, 1 = unshare */
465 uint64_t z_sharemax; /* max length of share string */
469 * ZFS file systems may behave the usual, POSIX-compliant way, where
470 * name lookups are case-sensitive. They may also be set up so that
471 * all the name lookups are case-insensitive, or so that only some
472 * lookups, the ones that set an FIGNORECASE flag, are case-insensitive.
474 typedef enum zfs_case {
476 ZFS_CASE_INSENSITIVE,
481 * Note: this struct must have the same layout in 32-bit and 64-bit, so
482 * that 32-bit processes (like /sbin/zfs) can pass it to the 64-bit
483 * kernel. Therefore, we add padding to it so that no "hidden" padding
484 * is automatically added on 64-bit (but not on 32-bit).
486 typedef struct zfs_cmd {
487 char zc_name[MAXPATHLEN]; /* name of pool or dataset */
488 uint64_t zc_nvlist_src; /* really (char *) */
489 uint64_t zc_nvlist_src_size;
490 uint64_t zc_nvlist_dst; /* really (char *) */
491 uint64_t zc_nvlist_dst_size;
492 boolean_t zc_nvlist_dst_filled; /* put an nvlist in dst? */
496 * The following members are for legacy ioctls which haven't been
497 * converted to the new method.
499 uint64_t zc_history; /* really (char *) */
500 char zc_value[MAXPATHLEN * 2];
501 char zc_string[MAXNAMELEN];
503 uint64_t zc_nvlist_conf; /* really (char *) */
504 uint64_t zc_nvlist_conf_size;
506 uint64_t zc_objset_type;
507 uint64_t zc_perm_action;
508 uint64_t zc_history_len;
509 uint64_t zc_history_offset;
511 uint64_t zc_iflags; /* internal to zfs(7fs) */
512 zfs_share_t zc_share;
513 dmu_objset_stats_t zc_objset_stats;
514 struct drr_begin zc_begin_record;
515 zinject_record_t zc_inject_record;
516 uint32_t zc_defer_destroy;
518 uint64_t zc_action_handle;
521 uint8_t zc_pad[3]; /* alignment */
524 uint64_t zc_createtxg;
529 typedef struct zfs_useracct {
536 #define ZFSDEV_MAX_MINOR (1 << 16)
538 #define ZPOOL_EXPORT_AFTER_SPLIT 0x1
544 typedef struct zfs_creat {
545 nvlist_t *zct_zplprops;
549 extern int zfs_secpolicy_snapshot_perms(const char *, cred_t *);
550 extern int zfs_secpolicy_rename_perms(const char *, const char *, cred_t *);
551 extern int zfs_secpolicy_destroy_perms(const char *, cred_t *);
552 extern void zfs_unmount_snap(const char *);
553 extern void zfs_destroy_unmount_origin(const char *);
554 extern int getzfsvfs_impl(struct objset *, struct zfsvfs **);
555 extern int getzfsvfs(const char *, struct zfsvfs **);
557 enum zfsdev_state_type {
564 * The zfsdev_state_t structure is managed as a singly-linked list
565 * from which items are never deleted. This allows for lock-free
566 * reading of the list so long as assignments to the zs_next and
567 * reads from zs_minor are performed atomically. Empty items are
568 * indicated by storing -1 into zs_minor.
570 typedef struct zfsdev_state {
571 struct zfsdev_state *zs_next; /* next zfsdev_state_t link */
572 minor_t zs_minor; /* made up minor number */
573 void *zs_onexit; /* onexit data */
574 void *zs_zevent; /* zevent data */
577 extern void *zfsdev_get_state(minor_t minor, enum zfsdev_state_type which);
578 extern int zfsdev_getminor(zfs_file_t *fp, minor_t *minorp);
580 extern uint_t zfs_allow_log_key;
588 #endif /* _SYS_ZFS_IOCTL_H */