]> git.proxmox.com Git - mirror_zfs.git/blob - include/sys/zfs_ioctl.h
Fix send/recv lost spill block
[mirror_zfs.git] / include / sys / zfs_ioctl.h
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 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2012, 2017 by Delphix. All rights reserved.
24 * Copyright 2016 RackTop Systems.
25 * Copyright (c) 2017, Intel Corporation.
26 */
27
28 #ifndef _SYS_ZFS_IOCTL_H
29 #define _SYS_ZFS_IOCTL_H
30
31 #include <sys/cred.h>
32 #include <sys/dmu.h>
33 #include <sys/zio.h>
34 #include <sys/dsl_deleg.h>
35 #include <sys/spa.h>
36 #include <sys/zfs_stat.h>
37
38 #ifdef _KERNEL
39 #include <sys/nvpair.h>
40 #endif /* _KERNEL */
41
42 #ifdef __cplusplus
43 extern "C" {
44 #endif
45
46 /*
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).
53 */
54
55 /*
56 * Property values for snapdir
57 */
58 #define ZFS_SNAPDIR_HIDDEN 0
59 #define ZFS_SNAPDIR_VISIBLE 1
60
61 /*
62 * Property values for snapdev
63 */
64 #define ZFS_SNAPDEV_HIDDEN 0
65 #define ZFS_SNAPDEV_VISIBLE 1
66 /*
67 * Property values for acltype
68 */
69 #define ZFS_ACLTYPE_OFF 0
70 #define ZFS_ACLTYPE_POSIXACL 1
71
72 /*
73 * Field manipulation macros for the drr_versioninfo field of the
74 * send stream header.
75 */
76
77 /*
78 * Header types for zfs send streams.
79 */
80 typedef enum drr_headertype {
81 DMU_SUBSTREAM = 0x1,
82 DMU_COMPOUNDSTREAM = 0x2
83 } drr_headertype_t;
84
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)
87
88 #define DMU_GET_FEATUREFLAGS(vi) BF64_GET((vi), 2, 30)
89 #define DMU_SET_FEATUREFLAGS(vi, x) BF64_SET((vi), 2, 30, x)
90
91 /*
92 * Feature flags for zfs send streams (flags in drr_versioninfo)
93 */
94
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 /* flag #21 is reserved for the redacted send/receive feature */
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 /* flag #25 is reserved for the ZSTD compression feature */
109 #define DMU_BACKUP_FEATURE_HOLDS (1 << 26)
110
111 /*
112 * Mask of all supported backup features
113 */
114 #define DMU_BACKUP_FEATURE_MASK (DMU_BACKUP_FEATURE_DEDUP | \
115 DMU_BACKUP_FEATURE_DEDUPPROPS | DMU_BACKUP_FEATURE_SA_SPILL | \
116 DMU_BACKUP_FEATURE_EMBED_DATA | DMU_BACKUP_FEATURE_LZ4 | \
117 DMU_BACKUP_FEATURE_RESUMING | DMU_BACKUP_FEATURE_LARGE_BLOCKS | \
118 DMU_BACKUP_FEATURE_COMPRESSED | DMU_BACKUP_FEATURE_LARGE_DNODE | \
119 DMU_BACKUP_FEATURE_RAW | DMU_BACKUP_FEATURE_HOLDS)
120
121 /* Are all features in the given flag word currently supported? */
122 #define DMU_STREAM_SUPPORTED(x) (!((x) & ~DMU_BACKUP_FEATURE_MASK))
123
124 typedef enum dmu_send_resume_token_version {
125 ZFS_SEND_RESUME_TOKEN_VERSION = 1
126 } dmu_send_resume_token_version_t;
127
128 /*
129 * The drr_versioninfo field of the dmu_replay_record has the
130 * following layout:
131 *
132 * 64 56 48 40 32 24 16 8 0
133 * +-------+-------+-------+-------+-------+-------+-------+-------+
134 * | reserved | feature-flags |C|S|
135 * +-------+-------+-------+-------+-------+-------+-------+-------+
136 *
137 * The low order two bits indicate the header type: SUBSTREAM (0x1)
138 * or COMPOUNDSTREAM (0x2). Using two bits for this is historical:
139 * this field used to be a version number, where the two version types
140 * were 1 and 2. Using two bits for this allows earlier versions of
141 * the code to be able to recognize send streams that don't use any
142 * of the features indicated by feature flags.
143 */
144
145 #define DMU_BACKUP_MAGIC 0x2F5bacbacULL
146
147 /*
148 * Send stream flags. Bits 24-31 are reserved for vendor-specific
149 * implementations and should not be used.
150 */
151 #define DRR_FLAG_CLONE (1<<0)
152 #define DRR_FLAG_CI_DATA (1<<1)
153 /*
154 * This send stream, if it is a full send, includes the FREE and FREEOBJECT
155 * records that are created by the sending process. This means that the send
156 * stream can be received as a clone, even though it is not an incremental.
157 * This is not implemented as a feature flag, because the receiving side does
158 * not need to have implemented it to receive this stream; it is fully backwards
159 * compatible. We need a flag, though, because full send streams without it
160 * cannot necessarily be received as a clone correctly.
161 */
162 #define DRR_FLAG_FREERECORDS (1<<2)
163 /*
164 * When DRR_FLAG_SPILL_BLOCK is set it indicates the DRR_OBJECT_SPILL
165 * and DRR_SPILL_UNMODIFIED flags are meaningful in the send stream.
166 *
167 * When DRR_FLAG_SPILL_BLOCK is set, DRR_OBJECT records will have
168 * DRR_OBJECT_SPILL set if and only if they should have a spill block
169 * (either an existing one, or a new one in the send stream). When clear
170 * the object does not have a spill block and any existing spill block
171 * should be freed.
172 *
173 * Similarly, when DRR_FLAG_SPILL_BLOCK is set, DRR_SPILL records will
174 * have DRR_SPILL_UNMODIFIED set if and only if they were included for
175 * backward compatibility purposes, and can be safely ignored by new versions
176 * of zfs receive. Previous versions of ZFS which do not understand the
177 * DRR_FLAG_SPILL_BLOCK will process this record and recreate any missing
178 * spill blocks.
179 */
180 #define DRR_FLAG_SPILL_BLOCK (1<<3)
181
182 /*
183 * flags in the drr_flags field in the DRR_WRITE, DRR_SPILL, DRR_OBJECT,
184 * DRR_WRITE_BYREF, and DRR_OBJECT_RANGE blocks
185 */
186 #define DRR_CHECKSUM_DEDUP (1<<0) /* not used for SPILL records */
187 #define DRR_RAW_BYTESWAP (1<<1)
188 #define DRR_OBJECT_SPILL (1<<2) /* OBJECT record has a spill block */
189 #define DRR_SPILL_UNMODIFIED (1<<2) /* SPILL record for unmodified block */
190
191 #define DRR_IS_DEDUP_CAPABLE(flags) ((flags) & DRR_CHECKSUM_DEDUP)
192 #define DRR_IS_RAW_BYTESWAPPED(flags) ((flags) & DRR_RAW_BYTESWAP)
193 #define DRR_OBJECT_HAS_SPILL(flags) ((flags) & DRR_OBJECT_SPILL)
194 #define DRR_SPILL_IS_UNMODIFIED(flags) ((flags) & DRR_SPILL_UNMODIFIED)
195
196 /* deal with compressed drr_write replay records */
197 #define DRR_WRITE_COMPRESSED(drrw) ((drrw)->drr_compressiontype != 0)
198 #define DRR_WRITE_PAYLOAD_SIZE(drrw) \
199 (DRR_WRITE_COMPRESSED(drrw) ? (drrw)->drr_compressed_size : \
200 (drrw)->drr_logical_size)
201 #define DRR_SPILL_PAYLOAD_SIZE(drrs) \
202 ((drrs)->drr_compressed_size ? \
203 (drrs)->drr_compressed_size : (drrs)->drr_length)
204 #define DRR_OBJECT_PAYLOAD_SIZE(drro) \
205 ((drro)->drr_raw_bonuslen != 0 ? \
206 (drro)->drr_raw_bonuslen : P2ROUNDUP((drro)->drr_bonuslen, 8))
207
208 /*
209 * zfs ioctl command structure
210 */
211 typedef struct dmu_replay_record {
212 enum {
213 DRR_BEGIN, DRR_OBJECT, DRR_FREEOBJECTS,
214 DRR_WRITE, DRR_FREE, DRR_END, DRR_WRITE_BYREF,
215 DRR_SPILL, DRR_WRITE_EMBEDDED, DRR_OBJECT_RANGE,
216 DRR_NUMTYPES
217 } drr_type;
218 uint32_t drr_payloadlen;
219 union {
220 struct drr_begin {
221 uint64_t drr_magic;
222 uint64_t drr_versioninfo; /* was drr_version */
223 uint64_t drr_creation_time;
224 dmu_objset_type_t drr_type;
225 uint32_t drr_flags;
226 uint64_t drr_toguid;
227 uint64_t drr_fromguid;
228 char drr_toname[MAXNAMELEN];
229 } drr_begin;
230 struct drr_end {
231 zio_cksum_t drr_checksum;
232 uint64_t drr_toguid;
233 } drr_end;
234 struct drr_object {
235 uint64_t drr_object;
236 dmu_object_type_t drr_type;
237 dmu_object_type_t drr_bonustype;
238 uint32_t drr_blksz;
239 uint32_t drr_bonuslen;
240 uint8_t drr_checksumtype;
241 uint8_t drr_compress;
242 uint8_t drr_dn_slots;
243 uint8_t drr_flags;
244 uint32_t drr_raw_bonuslen;
245 uint64_t drr_toguid;
246 /* only (possibly) nonzero for raw streams */
247 uint8_t drr_indblkshift;
248 uint8_t drr_nlevels;
249 uint8_t drr_nblkptr;
250 uint8_t drr_pad[5];
251 uint64_t drr_maxblkid;
252 /* bonus content follows */
253 } drr_object;
254 struct drr_freeobjects {
255 uint64_t drr_firstobj;
256 uint64_t drr_numobjs;
257 uint64_t drr_toguid;
258 } drr_freeobjects;
259 struct drr_write {
260 uint64_t drr_object;
261 dmu_object_type_t drr_type;
262 uint32_t drr_pad;
263 uint64_t drr_offset;
264 uint64_t drr_logical_size;
265 uint64_t drr_toguid;
266 uint8_t drr_checksumtype;
267 uint8_t drr_flags;
268 uint8_t drr_compressiontype;
269 uint8_t drr_pad2[5];
270 /* deduplication key */
271 ddt_key_t drr_key;
272 /* only nonzero if drr_compressiontype is not 0 */
273 uint64_t drr_compressed_size;
274 /* only nonzero for raw streams */
275 uint8_t drr_salt[ZIO_DATA_SALT_LEN];
276 uint8_t drr_iv[ZIO_DATA_IV_LEN];
277 uint8_t drr_mac[ZIO_DATA_MAC_LEN];
278 /* content follows */
279 } drr_write;
280 struct drr_free {
281 uint64_t drr_object;
282 uint64_t drr_offset;
283 uint64_t drr_length;
284 uint64_t drr_toguid;
285 } drr_free;
286 struct drr_write_byref {
287 /* where to put the data */
288 uint64_t drr_object;
289 uint64_t drr_offset;
290 uint64_t drr_length;
291 uint64_t drr_toguid;
292 /* where to find the prior copy of the data */
293 uint64_t drr_refguid;
294 uint64_t drr_refobject;
295 uint64_t drr_refoffset;
296 /* properties of the data */
297 uint8_t drr_checksumtype;
298 uint8_t drr_flags;
299 uint8_t drr_pad2[6];
300 ddt_key_t drr_key; /* deduplication key */
301 } drr_write_byref;
302 struct drr_spill {
303 uint64_t drr_object;
304 uint64_t drr_length;
305 uint64_t drr_toguid;
306 uint8_t drr_flags;
307 uint8_t drr_compressiontype;
308 uint8_t drr_pad[6];
309 /* only nonzero for raw streams */
310 uint64_t drr_compressed_size;
311 uint8_t drr_salt[ZIO_DATA_SALT_LEN];
312 uint8_t drr_iv[ZIO_DATA_IV_LEN];
313 uint8_t drr_mac[ZIO_DATA_MAC_LEN];
314 dmu_object_type_t drr_type;
315 /* spill data follows */
316 } drr_spill;
317 struct drr_write_embedded {
318 uint64_t drr_object;
319 uint64_t drr_offset;
320 /* logical length, should equal blocksize */
321 uint64_t drr_length;
322 uint64_t drr_toguid;
323 uint8_t drr_compression;
324 uint8_t drr_etype;
325 uint8_t drr_pad[6];
326 uint32_t drr_lsize; /* uncompressed size of payload */
327 uint32_t drr_psize; /* compr. (real) size of payload */
328 /* (possibly compressed) content follows */
329 } drr_write_embedded;
330 struct drr_object_range {
331 uint64_t drr_firstobj;
332 uint64_t drr_numslots;
333 uint64_t drr_toguid;
334 uint8_t drr_salt[ZIO_DATA_SALT_LEN];
335 uint8_t drr_iv[ZIO_DATA_IV_LEN];
336 uint8_t drr_mac[ZIO_DATA_MAC_LEN];
337 uint8_t drr_flags;
338 uint8_t drr_pad[3];
339 } drr_object_range;
340
341 /*
342 * Nore: drr_checksum is overlaid with all record types
343 * except DRR_BEGIN. Therefore its (non-pad) members
344 * must not overlap with members from the other structs.
345 * We accomplish this by putting its members at the very
346 * end of the struct.
347 */
348 struct drr_checksum {
349 uint64_t drr_pad[34];
350 /*
351 * fletcher-4 checksum of everything preceding the
352 * checksum.
353 */
354 zio_cksum_t drr_checksum;
355 } drr_checksum;
356 } drr_u;
357 } dmu_replay_record_t;
358
359 /* diff record range types */
360 typedef enum diff_type {
361 DDR_NONE = 0x1,
362 DDR_INUSE = 0x2,
363 DDR_FREE = 0x4
364 } diff_type_t;
365
366 /*
367 * The diff reports back ranges of free or in-use objects.
368 */
369 typedef struct dmu_diff_record {
370 uint64_t ddr_type;
371 uint64_t ddr_first;
372 uint64_t ddr_last;
373 } dmu_diff_record_t;
374
375 typedef struct zinject_record {
376 uint64_t zi_objset;
377 uint64_t zi_object;
378 uint64_t zi_start;
379 uint64_t zi_end;
380 uint64_t zi_guid;
381 uint32_t zi_level;
382 uint32_t zi_error;
383 uint64_t zi_type;
384 uint32_t zi_freq;
385 uint32_t zi_failfast;
386 char zi_func[MAXNAMELEN];
387 uint32_t zi_iotype;
388 int32_t zi_duration;
389 uint64_t zi_timer;
390 uint64_t zi_nlanes;
391 uint32_t zi_cmd;
392 uint32_t zi_dvas;
393 } zinject_record_t;
394
395 #define ZINJECT_NULL 0x1
396 #define ZINJECT_FLUSH_ARC 0x2
397 #define ZINJECT_UNLOAD_SPA 0x4
398 #define ZINJECT_CALC_RANGE 0x8
399
400 #define ZEVENT_NONE 0x0
401 #define ZEVENT_NONBLOCK 0x1
402 #define ZEVENT_SIZE 1024
403
404 #define ZEVENT_SEEK_START 0
405 #define ZEVENT_SEEK_END UINT64_MAX
406
407 /* scaled frequency ranges */
408 #define ZI_PERCENTAGE_MIN 4294UL
409 #define ZI_PERCENTAGE_MAX UINT32_MAX
410
411 #define ZI_NO_DVA (-1)
412
413 typedef enum zinject_type {
414 ZINJECT_UNINITIALIZED,
415 ZINJECT_DATA_FAULT,
416 ZINJECT_DEVICE_FAULT,
417 ZINJECT_LABEL_FAULT,
418 ZINJECT_IGNORED_WRITES,
419 ZINJECT_PANIC,
420 ZINJECT_DELAY_IO,
421 ZINJECT_DECRYPT_FAULT,
422 } zinject_type_t;
423
424 typedef struct zfs_share {
425 uint64_t z_exportdata;
426 uint64_t z_sharedata;
427 uint64_t z_sharetype; /* 0 = share, 1 = unshare */
428 uint64_t z_sharemax; /* max length of share string */
429 } zfs_share_t;
430
431 /*
432 * ZFS file systems may behave the usual, POSIX-compliant way, where
433 * name lookups are case-sensitive. They may also be set up so that
434 * all the name lookups are case-insensitive, or so that only some
435 * lookups, the ones that set an FIGNORECASE flag, are case-insensitive.
436 */
437 typedef enum zfs_case {
438 ZFS_CASE_SENSITIVE,
439 ZFS_CASE_INSENSITIVE,
440 ZFS_CASE_MIXED
441 } zfs_case_t;
442
443 /*
444 * Note: this struct must have the same layout in 32-bit and 64-bit, so
445 * that 32-bit processes (like /sbin/zfs) can pass it to the 64-bit
446 * kernel. Therefore, we add padding to it so that no "hidden" padding
447 * is automatically added on 64-bit (but not on 32-bit).
448 */
449 typedef struct zfs_cmd {
450 char zc_name[MAXPATHLEN]; /* name of pool or dataset */
451 uint64_t zc_nvlist_src; /* really (char *) */
452 uint64_t zc_nvlist_src_size;
453 uint64_t zc_nvlist_dst; /* really (char *) */
454 uint64_t zc_nvlist_dst_size;
455 boolean_t zc_nvlist_dst_filled; /* put an nvlist in dst? */
456 int zc_pad2;
457
458 /*
459 * The following members are for legacy ioctls which haven't been
460 * converted to the new method.
461 */
462 uint64_t zc_history; /* really (char *) */
463 char zc_value[MAXPATHLEN * 2];
464 char zc_string[MAXNAMELEN];
465 uint64_t zc_guid;
466 uint64_t zc_nvlist_conf; /* really (char *) */
467 uint64_t zc_nvlist_conf_size;
468 uint64_t zc_cookie;
469 uint64_t zc_objset_type;
470 uint64_t zc_perm_action;
471 uint64_t zc_history_len;
472 uint64_t zc_history_offset;
473 uint64_t zc_obj;
474 uint64_t zc_iflags; /* internal to zfs(7fs) */
475 zfs_share_t zc_share;
476 dmu_objset_stats_t zc_objset_stats;
477 struct drr_begin zc_begin_record;
478 zinject_record_t zc_inject_record;
479 uint32_t zc_defer_destroy;
480 uint32_t zc_flags;
481 uint64_t zc_action_handle;
482 int zc_cleanup_fd;
483 uint8_t zc_simple;
484 uint8_t zc_pad[3]; /* alignment */
485 uint64_t zc_sendobj;
486 uint64_t zc_fromobj;
487 uint64_t zc_createtxg;
488 zfs_stat_t zc_stat;
489 } zfs_cmd_t;
490
491 typedef struct zfs_useracct {
492 char zu_domain[256];
493 uid_t zu_rid;
494 uint32_t zu_pad;
495 uint64_t zu_space;
496 } zfs_useracct_t;
497
498 #define ZFSDEV_MAX_MINOR (1 << 16)
499 #define ZFS_MIN_MINOR (ZFSDEV_MAX_MINOR + 1)
500
501 #define ZPOOL_EXPORT_AFTER_SPLIT 0x1
502
503 #ifdef _KERNEL
504 struct objset;
505 struct zfsvfs;
506
507 typedef struct zfs_creat {
508 nvlist_t *zct_zplprops;
509 nvlist_t *zct_props;
510 } zfs_creat_t;
511
512 extern int zfs_secpolicy_snapshot_perms(const char *, cred_t *);
513 extern int zfs_secpolicy_rename_perms(const char *, const char *, cred_t *);
514 extern int zfs_secpolicy_destroy_perms(const char *, cred_t *);
515 extern void zfs_unmount_snap(const char *);
516 extern void zfs_destroy_unmount_origin(const char *);
517 extern int getzfsvfs_impl(struct objset *, struct zfsvfs **);
518 extern int getzfsvfs(const char *, struct zfsvfs **);
519
520 enum zfsdev_state_type {
521 ZST_ONEXIT,
522 ZST_ZEVENT,
523 ZST_ALL,
524 };
525
526 /*
527 * The zfsdev_state_t structure is managed as a singly-linked list
528 * from which items are never deleted. This allows for lock-free
529 * reading of the list so long as assignments to the zs_next and
530 * reads from zs_minor are performed atomically. Empty items are
531 * indicated by storing -1 into zs_minor.
532 */
533 typedef struct zfsdev_state {
534 struct zfsdev_state *zs_next; /* next zfsdev_state_t link */
535 struct file *zs_file; /* associated file struct */
536 minor_t zs_minor; /* made up minor number */
537 void *zs_onexit; /* onexit data */
538 void *zs_zevent; /* zevent data */
539 } zfsdev_state_t;
540
541 extern void *zfsdev_get_state(minor_t minor, enum zfsdev_state_type which);
542 extern int zfsdev_getminor(struct file *filp, minor_t *minorp);
543 extern minor_t zfsdev_minor_alloc(void);
544
545 #endif /* _KERNEL */
546
547 #ifdef __cplusplus
548 }
549 #endif
550
551 #endif /* _SYS_ZFS_IOCTL_H */