xref: /illumos-gate/usr/src/uts/common/fs/zfs/sys/zfs_ioctl.h (revision 12a8814c13fbb1d6d58616cf090ea5815dc107f9)
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) 2011-2012 Pawel Jakub Dawidek. All rights reserved.
24  * Copyright (c) 2012, 2017 by Delphix. All rights reserved.
25  * Copyright 2016 RackTop Systems.
26  * Copyright (c) 2014 Integros [integros.com]
27  */
28 
29 #ifndef	_SYS_ZFS_IOCTL_H
30 #define	_SYS_ZFS_IOCTL_H
31 
32 #include <sys/cred.h>
33 #include <sys/dmu.h>
34 #include <sys/zio.h>
35 #include <sys/dsl_deleg.h>
36 #include <sys/spa.h>
37 #include <sys/zfs_stat.h>
38 
39 #ifdef _KERNEL
40 #include <sys/nvpair.h>
41 #endif	/* _KERNEL */
42 
43 #ifdef	__cplusplus
44 extern "C" {
45 #endif
46 
47 /*
48  * The structures in this file are passed between userland and the
49  * kernel.  Userland may be running a 32-bit process, while the kernel
50  * is 64-bit.  Therefore, these structures need to compile the same in
51  * 32-bit and 64-bit.  This means not using type "long", and adding
52  * explicit padding so that the 32-bit structure will not be packed more
53  * tightly than the 64-bit structure (which requires 64-bit alignment).
54  */
55 
56 /*
57  * Property values for snapdir
58  */
59 #define	ZFS_SNAPDIR_HIDDEN		0
60 #define	ZFS_SNAPDIR_VISIBLE		1
61 
62 /*
63  * Field manipulation macros for the drr_versioninfo field of the
64  * send stream header.
65  */
66 
67 /*
68  * Header types for zfs send streams.
69  */
70 typedef enum drr_headertype {
71 	DMU_SUBSTREAM = 0x1,
72 	DMU_COMPOUNDSTREAM = 0x2
73 } drr_headertype_t;
74 
75 #define	DMU_GET_STREAM_HDRTYPE(vi)	BF64_GET((vi), 0, 2)
76 #define	DMU_SET_STREAM_HDRTYPE(vi, x)	BF64_SET((vi), 0, 2, x)
77 
78 #define	DMU_GET_FEATUREFLAGS(vi)	BF64_GET((vi), 2, 30)
79 #define	DMU_SET_FEATUREFLAGS(vi, x)	BF64_SET((vi), 2, 30, x)
80 
81 /*
82  * Feature flags for zfs send streams (flags in drr_versioninfo)
83  */
84 
85 #define	DMU_BACKUP_FEATURE_DEDUP		(1 << 0)
86 #define	DMU_BACKUP_FEATURE_DEDUPPROPS		(1 << 1)
87 #define	DMU_BACKUP_FEATURE_SA_SPILL		(1 << 2)
88 /* flags #3 - #15 are reserved for incompatible closed-source implementations */
89 #define	DMU_BACKUP_FEATURE_EMBED_DATA		(1 << 16)
90 #define	DMU_BACKUP_FEATURE_LZ4			(1 << 17)
91 /* flag #18 is reserved for a Delphix feature */
92 #define	DMU_BACKUP_FEATURE_LARGE_BLOCKS		(1 << 19)
93 #define	DMU_BACKUP_FEATURE_RESUMING		(1 << 20)
94 /* flag #21 is reserved for a Delphix feature */
95 #define	DMU_BACKUP_FEATURE_COMPRESSED		(1 << 22)
96 #define	DMU_BACKUP_FEATURE_LARGE_DNODE		(1 << 23)
97 /* flag #24 is reserved for the raw send feature */
98 /* flag #25 is reserved for the ZSTD compression feature */
99 
100 /*
101  * Mask of all supported backup features
102  */
103 #define	DMU_BACKUP_FEATURE_MASK	(DMU_BACKUP_FEATURE_DEDUP | \
104     DMU_BACKUP_FEATURE_DEDUPPROPS | DMU_BACKUP_FEATURE_SA_SPILL | \
105     DMU_BACKUP_FEATURE_EMBED_DATA | DMU_BACKUP_FEATURE_LZ4 | \
106     DMU_BACKUP_FEATURE_RESUMING | \
107     DMU_BACKUP_FEATURE_LARGE_BLOCKS | DMU_BACKUP_FEATURE_LARGE_DNODE | \
108     DMU_BACKUP_FEATURE_COMPRESSED)
109 
110 /* Are all features in the given flag word currently supported? */
111 #define	DMU_STREAM_SUPPORTED(x)	(!((x) & ~DMU_BACKUP_FEATURE_MASK))
112 
113 typedef enum dmu_send_resume_token_version {
114 	ZFS_SEND_RESUME_TOKEN_VERSION = 1
115 } dmu_send_resume_token_version_t;
116 
117 /*
118  * The drr_versioninfo field of the dmu_replay_record has the
119  * following layout:
120  *
121  *	64	56	48	40	32	24	16	8	0
122  *	+-------+-------+-------+-------+-------+-------+-------+-------+
123  *	|		reserved	|        feature-flags	    |C|S|
124  *	+-------+-------+-------+-------+-------+-------+-------+-------+
125  *
126  * The low order two bits indicate the header type: SUBSTREAM (0x1)
127  * or COMPOUNDSTREAM (0x2).  Using two bits for this is historical:
128  * this field used to be a version number, where the two version types
129  * were 1 and 2.  Using two bits for this allows earlier versions of
130  * the code to be able to recognize send streams that don't use any
131  * of the features indicated by feature flags.
132  */
133 
134 #define	DMU_BACKUP_MAGIC 0x2F5bacbacULL
135 
136 /*
137  * Send stream flags.  Bits 24-31 are reserved for vendor-specific
138  * implementations and should not be used.
139  */
140 #define	DRR_FLAG_CLONE		(1<<0)
141 #define	DRR_FLAG_CI_DATA	(1<<1)
142 /*
143  * This send stream, if it is a full send, includes the FREE and FREEOBJECT
144  * records that are created by the sending process.  This means that the send
145  * stream can be received as a clone, even though it is not an incremental.
146  * This is not implemented as a feature flag, because the receiving side does
147  * not need to have implemented it to receive this stream; it is fully backwards
148  * compatible.  We need a flag, though, because full send streams without it
149  * cannot necessarily be received as a clone correctly.
150  */
151 #define	DRR_FLAG_FREERECORDS	(1<<2)
152 
153 /*
154  * flags in the drr_checksumflags field in the DRR_WRITE and
155  * DRR_WRITE_BYREF blocks
156  */
157 #define	DRR_CHECKSUM_DEDUP	(1<<0)
158 
159 #define	DRR_IS_DEDUP_CAPABLE(flags)	((flags) & DRR_CHECKSUM_DEDUP)
160 
161 /* deal with compressed drr_write replay records */
162 #define	DRR_WRITE_COMPRESSED(drrw)	((drrw)->drr_compressiontype != 0)
163 #define	DRR_WRITE_PAYLOAD_SIZE(drrw) \
164 	(DRR_WRITE_COMPRESSED(drrw) ? (drrw)->drr_compressed_size : \
165 	(drrw)->drr_logical_size)
166 
167 /*
168  * zfs ioctl command structure
169  */
170 typedef struct dmu_replay_record {
171 	enum {
172 		DRR_BEGIN, DRR_OBJECT, DRR_FREEOBJECTS,
173 		DRR_WRITE, DRR_FREE, DRR_END, DRR_WRITE_BYREF,
174 		DRR_SPILL, DRR_WRITE_EMBEDDED, DRR_NUMTYPES
175 	} drr_type;
176 	uint32_t drr_payloadlen;
177 	union {
178 		struct drr_begin {
179 			uint64_t drr_magic;
180 			uint64_t drr_versioninfo; /* was drr_version */
181 			uint64_t drr_creation_time;
182 			dmu_objset_type_t drr_type;
183 			uint32_t drr_flags;
184 			uint64_t drr_toguid;
185 			uint64_t drr_fromguid;
186 			char drr_toname[MAXNAMELEN];
187 		} drr_begin;
188 		struct drr_end {
189 			zio_cksum_t drr_checksum;
190 			uint64_t drr_toguid;
191 		} drr_end;
192 		struct drr_object {
193 			uint64_t drr_object;
194 			dmu_object_type_t drr_type;
195 			dmu_object_type_t drr_bonustype;
196 			uint32_t drr_blksz;
197 			uint32_t drr_bonuslen;
198 			uint8_t drr_checksumtype;
199 			uint8_t drr_compress;
200 			uint8_t drr_dn_slots;
201 			uint8_t drr_pad[5];
202 			uint64_t drr_toguid;
203 			/* bonus content follows */
204 		} drr_object;
205 		struct drr_freeobjects {
206 			uint64_t drr_firstobj;
207 			uint64_t drr_numobjs;
208 			uint64_t drr_toguid;
209 		} drr_freeobjects;
210 		struct drr_write {
211 			uint64_t drr_object;
212 			dmu_object_type_t drr_type;
213 			uint32_t drr_pad;
214 			uint64_t drr_offset;
215 			uint64_t drr_logical_size;
216 			uint64_t drr_toguid;
217 			uint8_t drr_checksumtype;
218 			uint8_t drr_checksumflags;
219 			uint8_t drr_compressiontype;
220 			uint8_t drr_pad2[5];
221 			/* deduplication key */
222 			ddt_key_t drr_key;
223 			/* only nonzero if drr_compressiontype is not 0 */
224 			uint64_t drr_compressed_size;
225 			/* content follows */
226 		} drr_write;
227 		struct drr_free {
228 			uint64_t drr_object;
229 			uint64_t drr_offset;
230 			uint64_t drr_length;
231 			uint64_t drr_toguid;
232 		} drr_free;
233 		struct drr_write_byref {
234 			/* where to put the data */
235 			uint64_t drr_object;
236 			uint64_t drr_offset;
237 			uint64_t drr_length;
238 			uint64_t drr_toguid;
239 			/* where to find the prior copy of the data */
240 			uint64_t drr_refguid;
241 			uint64_t drr_refobject;
242 			uint64_t drr_refoffset;
243 			/* properties of the data */
244 			uint8_t drr_checksumtype;
245 			uint8_t drr_checksumflags;
246 			uint8_t drr_pad2[6];
247 			ddt_key_t drr_key; /* deduplication key */
248 		} drr_write_byref;
249 		struct drr_spill {
250 			uint64_t drr_object;
251 			uint64_t drr_length;
252 			uint64_t drr_toguid;
253 			uint64_t drr_pad[4]; /* needed for crypto */
254 			/* spill data follows */
255 		} drr_spill;
256 		struct drr_write_embedded {
257 			uint64_t drr_object;
258 			uint64_t drr_offset;
259 			/* logical length, should equal blocksize */
260 			uint64_t drr_length;
261 			uint64_t drr_toguid;
262 			uint8_t drr_compression;
263 			uint8_t drr_etype;
264 			uint8_t drr_pad[6];
265 			uint32_t drr_lsize; /* uncompressed size of payload */
266 			uint32_t drr_psize; /* compr. (real) size of payload */
267 			/* (possibly compressed) content follows */
268 		} drr_write_embedded;
269 
270 		/*
271 		 * Nore: drr_checksum is overlaid with all record types
272 		 * except DRR_BEGIN.  Therefore its (non-pad) members
273 		 * must not overlap with members from the other structs.
274 		 * We accomplish this by putting its members at the very
275 		 * end of the struct.
276 		 */
277 		struct drr_checksum {
278 			uint64_t drr_pad[34];
279 			/*
280 			 * fletcher-4 checksum of everything preceding the
281 			 * checksum.
282 			 */
283 			zio_cksum_t drr_checksum;
284 		} drr_checksum;
285 	} drr_u;
286 } dmu_replay_record_t;
287 
288 /* diff record range types */
289 typedef enum diff_type {
290 	DDR_NONE = 0x1,
291 	DDR_INUSE = 0x2,
292 	DDR_FREE = 0x4
293 } diff_type_t;
294 
295 /*
296  * The diff reports back ranges of free or in-use objects.
297  */
298 typedef struct dmu_diff_record {
299 	uint64_t ddr_type;
300 	uint64_t ddr_first;
301 	uint64_t ddr_last;
302 } dmu_diff_record_t;
303 
304 typedef struct zinject_record {
305 	uint64_t	zi_objset;
306 	uint64_t	zi_object;
307 	uint64_t	zi_start;
308 	uint64_t	zi_end;
309 	uint64_t	zi_guid;
310 	uint32_t	zi_level;
311 	uint32_t	zi_error;
312 	uint64_t	zi_type;
313 	uint32_t	zi_freq;
314 	uint32_t	zi_failfast;
315 	char		zi_func[MAXNAMELEN];
316 	uint32_t	zi_iotype;
317 	int32_t		zi_duration;
318 	uint64_t	zi_timer;
319 	uint64_t	zi_nlanes;
320 	uint32_t	zi_cmd;
321 	uint32_t	zi_dvas;
322 } zinject_record_t;
323 
324 #define	ZINJECT_NULL		0x1
325 #define	ZINJECT_FLUSH_ARC	0x2
326 #define	ZINJECT_UNLOAD_SPA	0x4
327 
328 #define	ZI_NO_DVA		(-1)
329 
330 typedef enum zinject_type {
331 	ZINJECT_UNINITIALIZED,
332 	ZINJECT_DATA_FAULT,
333 	ZINJECT_DEVICE_FAULT,
334 	ZINJECT_LABEL_FAULT,
335 	ZINJECT_IGNORED_WRITES,
336 	ZINJECT_PANIC,
337 	ZINJECT_DELAY_IO,
338 } zinject_type_t;
339 
340 typedef struct zfs_share {
341 	uint64_t	z_exportdata;
342 	uint64_t	z_sharedata;
343 	uint64_t	z_sharetype;	/* 0 = share, 1 = unshare */
344 	uint64_t	z_sharemax;  /* max length of share string */
345 } zfs_share_t;
346 
347 /*
348  * ZFS file systems may behave the usual, POSIX-compliant way, where
349  * name lookups are case-sensitive.  They may also be set up so that
350  * all the name lookups are case-insensitive, or so that only some
351  * lookups, the ones that set an FIGNORECASE flag, are case-insensitive.
352  */
353 typedef enum zfs_case {
354 	ZFS_CASE_SENSITIVE,
355 	ZFS_CASE_INSENSITIVE,
356 	ZFS_CASE_MIXED
357 } zfs_case_t;
358 
359 /*
360  * Note: this struct must have the same layout in 32-bit and 64-bit, so
361  * that 32-bit processes (like /sbin/zfs) can pass it to the 64-bit
362  * kernel.  Therefore, we add padding to it so that no "hidden" padding
363  * is automatically added on 64-bit (but not on 32-bit).
364  */
365 typedef struct zfs_cmd {
366 	char		zc_name[MAXPATHLEN];	/* name of pool or dataset */
367 	uint64_t	zc_nvlist_src;		/* really (char *) */
368 	uint64_t	zc_nvlist_src_size;
369 	uint64_t	zc_nvlist_dst;		/* really (char *) */
370 	uint64_t	zc_nvlist_dst_size;
371 	boolean_t	zc_nvlist_dst_filled;	/* put an nvlist in dst? */
372 	int		zc_pad2;
373 
374 	/*
375 	 * The following members are for legacy ioctls which haven't been
376 	 * converted to the new method.
377 	 */
378 	uint64_t	zc_history;		/* really (char *) */
379 	char		zc_value[MAXPATHLEN * 2];
380 	char		zc_string[MAXNAMELEN];
381 	uint64_t	zc_guid;
382 	uint64_t	zc_nvlist_conf;		/* really (char *) */
383 	uint64_t	zc_nvlist_conf_size;
384 	uint64_t	zc_cookie;
385 	uint64_t	zc_objset_type;
386 	uint64_t	zc_perm_action;
387 	uint64_t	zc_history_len;
388 	uint64_t	zc_history_offset;
389 	uint64_t	zc_obj;
390 	uint64_t	zc_iflags;		/* internal to zfs(7fs) */
391 	zfs_share_t	zc_share;
392 	dmu_objset_stats_t zc_objset_stats;
393 	dmu_replay_record_t zc_begin_record;
394 	zinject_record_t zc_inject_record;
395 	uint32_t	zc_defer_destroy;
396 	uint32_t	zc_flags;
397 	uint64_t	zc_action_handle;
398 	int		zc_cleanup_fd;
399 	uint8_t		zc_simple;
400 	uint8_t		zc_pad3[3];
401 	boolean_t	zc_resumable;
402 	uint32_t	zc_pad4;
403 	uint64_t	zc_sendobj;
404 	uint64_t	zc_fromobj;
405 	uint64_t	zc_createtxg;
406 	zfs_stat_t	zc_stat;
407 } zfs_cmd_t;
408 
409 typedef struct zfs_useracct {
410 	char zu_domain[256];
411 	uid_t zu_rid;
412 	uint32_t zu_pad;
413 	uint64_t zu_space;
414 } zfs_useracct_t;
415 
416 #define	ZFSDEV_MAX_MINOR	(1 << 16)
417 #define	ZFS_MIN_MINOR	(ZFSDEV_MAX_MINOR + 1)
418 
419 #define	ZPOOL_EXPORT_AFTER_SPLIT 0x1
420 
421 #ifdef _KERNEL
422 struct objset;
423 struct zfsvfs;
424 
425 typedef struct zfs_creat {
426 	nvlist_t	*zct_zplprops;
427 	nvlist_t	*zct_props;
428 } zfs_creat_t;
429 
430 extern dev_info_t *zfs_dip;
431 
432 extern int zfs_secpolicy_snapshot_perms(const char *, cred_t *);
433 extern int zfs_secpolicy_rename_perms(const char *, const char *, cred_t *);
434 extern int zfs_secpolicy_destroy_perms(const char *, cred_t *);
435 extern int zfs_busy(void);
436 extern void zfs_unmount_snap(const char *);
437 extern void zfs_destroy_unmount_origin(const char *);
438 extern int getzfsvfs_impl(struct objset *, struct zfsvfs **);
439 extern int getzfsvfs(const char *, struct zfsvfs **);
440 
441 /*
442  * ZFS minor numbers can refer to either a control device instance or
443  * a zvol. Depending on the value of zss_type, zss_data points to either
444  * a zvol_state_t or a zfs_onexit_t.
445  */
446 enum zfs_soft_state_type {
447 	ZSST_ZVOL,
448 	ZSST_CTLDEV
449 };
450 
451 typedef struct zfs_soft_state {
452 	enum zfs_soft_state_type zss_type;
453 	void *zss_data;
454 } zfs_soft_state_t;
455 
456 extern void *zfsdev_get_soft_state(minor_t minor,
457     enum zfs_soft_state_type which);
458 extern minor_t zfsdev_minor_alloc(void);
459 
460 extern void *zfsdev_state;
461 extern kmutex_t zfsdev_state_lock;
462 
463 #endif	/* _KERNEL */
464 
465 #ifdef	__cplusplus
466 }
467 #endif
468 
469 #endif	/* _SYS_ZFS_IOCTL_H */
470