xref: /illumos-gate/usr/src/uts/common/fs/zfs/sys/zfs_ioctl.h (revision d1aea6f139360e9e7f1504facb24f8521047b15c)
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, 2015 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 /* flag #23 is reserved for the large dnode feature */
97 
98 /*
99  * Mask of all supported backup features
100  */
101 #define	DMU_BACKUP_FEATURE_MASK	(DMU_BACKUP_FEATURE_DEDUP | \
102     DMU_BACKUP_FEATURE_DEDUPPROPS | DMU_BACKUP_FEATURE_SA_SPILL | \
103     DMU_BACKUP_FEATURE_EMBED_DATA | DMU_BACKUP_FEATURE_LZ4 | \
104     DMU_BACKUP_FEATURE_RESUMING | \
105     DMU_BACKUP_FEATURE_LARGE_BLOCKS | \
106     DMU_BACKUP_FEATURE_COMPRESSED)
107 
108 /* Are all features in the given flag word currently supported? */
109 #define	DMU_STREAM_SUPPORTED(x)	(!((x) & ~DMU_BACKUP_FEATURE_MASK))
110 
111 typedef enum dmu_send_resume_token_version {
112 	ZFS_SEND_RESUME_TOKEN_VERSION = 1
113 } dmu_send_resume_token_version_t;
114 
115 /*
116  * The drr_versioninfo field of the dmu_replay_record has the
117  * following layout:
118  *
119  *	64	56	48	40	32	24	16	8	0
120  *	+-------+-------+-------+-------+-------+-------+-------+-------+
121  *  	|		reserved	|        feature-flags	    |C|S|
122  *	+-------+-------+-------+-------+-------+-------+-------+-------+
123  *
124  * The low order two bits indicate the header type: SUBSTREAM (0x1)
125  * or COMPOUNDSTREAM (0x2).  Using two bits for this is historical:
126  * this field used to be a version number, where the two version types
127  * were 1 and 2.  Using two bits for this allows earlier versions of
128  * the code to be able to recognize send streams that don't use any
129  * of the features indicated by feature flags.
130  */
131 
132 #define	DMU_BACKUP_MAGIC 0x2F5bacbacULL
133 
134 /*
135  * Send stream flags.  Bits 24-31 are reserved for vendor-specific
136  * implementations and should not be used.
137  */
138 #define	DRR_FLAG_CLONE		(1<<0)
139 #define	DRR_FLAG_CI_DATA	(1<<1)
140 /*
141  * This send stream, if it is a full send, includes the FREE and FREEOBJECT
142  * records that are created by the sending process.  This means that the send
143  * stream can be received as a clone, even though it is not an incremental.
144  * This is not implemented as a feature flag, because the receiving side does
145  * not need to have implemented it to receive this stream; it is fully backwards
146  * compatible.  We need a flag, though, because full send streams without it
147  * cannot necessarily be received as a clone correctly.
148  */
149 #define	DRR_FLAG_FREERECORDS	(1<<2)
150 
151 /*
152  * flags in the drr_checksumflags field in the DRR_WRITE and
153  * DRR_WRITE_BYREF blocks
154  */
155 #define	DRR_CHECKSUM_DEDUP	(1<<0)
156 
157 #define	DRR_IS_DEDUP_CAPABLE(flags)	((flags) & DRR_CHECKSUM_DEDUP)
158 
159 /* deal with compressed drr_write replay records */
160 #define	DRR_WRITE_COMPRESSED(drrw)	((drrw)->drr_compressiontype != 0)
161 #define	DRR_WRITE_PAYLOAD_SIZE(drrw) \
162 	(DRR_WRITE_COMPRESSED(drrw) ? (drrw)->drr_compressed_size : \
163 	(drrw)->drr_logical_size)
164 
165 /*
166  * zfs ioctl command structure
167  */
168 typedef struct dmu_replay_record {
169 	enum {
170 		DRR_BEGIN, DRR_OBJECT, DRR_FREEOBJECTS,
171 		DRR_WRITE, DRR_FREE, DRR_END, DRR_WRITE_BYREF,
172 		DRR_SPILL, DRR_WRITE_EMBEDDED, DRR_NUMTYPES
173 	} drr_type;
174 	uint32_t drr_payloadlen;
175 	union {
176 		struct drr_begin {
177 			uint64_t drr_magic;
178 			uint64_t drr_versioninfo; /* was drr_version */
179 			uint64_t drr_creation_time;
180 			dmu_objset_type_t drr_type;
181 			uint32_t drr_flags;
182 			uint64_t drr_toguid;
183 			uint64_t drr_fromguid;
184 			char drr_toname[MAXNAMELEN];
185 		} drr_begin;
186 		struct drr_end {
187 			zio_cksum_t drr_checksum;
188 			uint64_t drr_toguid;
189 		} drr_end;
190 		struct drr_object {
191 			uint64_t drr_object;
192 			dmu_object_type_t drr_type;
193 			dmu_object_type_t drr_bonustype;
194 			uint32_t drr_blksz;
195 			uint32_t drr_bonuslen;
196 			uint8_t drr_checksumtype;
197 			uint8_t drr_compress;
198 			uint8_t drr_pad[6];
199 			uint64_t drr_toguid;
200 			/* bonus content follows */
201 		} drr_object;
202 		struct drr_freeobjects {
203 			uint64_t drr_firstobj;
204 			uint64_t drr_numobjs;
205 			uint64_t drr_toguid;
206 		} drr_freeobjects;
207 		struct drr_write {
208 			uint64_t drr_object;
209 			dmu_object_type_t drr_type;
210 			uint32_t drr_pad;
211 			uint64_t drr_offset;
212 			uint64_t drr_logical_size;
213 			uint64_t drr_toguid;
214 			uint8_t drr_checksumtype;
215 			uint8_t drr_checksumflags;
216 			uint8_t drr_compressiontype;
217 			uint8_t drr_pad2[5];
218 			/* deduplication key */
219 			ddt_key_t drr_key;
220 			/* only nonzero if drr_compressiontype is not 0 */
221 			uint64_t drr_compressed_size;
222 			/* content follows */
223 		} drr_write;
224 		struct drr_free {
225 			uint64_t drr_object;
226 			uint64_t drr_offset;
227 			uint64_t drr_length;
228 			uint64_t drr_toguid;
229 		} drr_free;
230 		struct drr_write_byref {
231 			/* where to put the data */
232 			uint64_t drr_object;
233 			uint64_t drr_offset;
234 			uint64_t drr_length;
235 			uint64_t drr_toguid;
236 			/* where to find the prior copy of the data */
237 			uint64_t drr_refguid;
238 			uint64_t drr_refobject;
239 			uint64_t drr_refoffset;
240 			/* properties of the data */
241 			uint8_t drr_checksumtype;
242 			uint8_t drr_checksumflags;
243 			uint8_t drr_pad2[6];
244 			ddt_key_t drr_key; /* deduplication key */
245 		} drr_write_byref;
246 		struct drr_spill {
247 			uint64_t drr_object;
248 			uint64_t drr_length;
249 			uint64_t drr_toguid;
250 			uint64_t drr_pad[4]; /* needed for crypto */
251 			/* spill data follows */
252 		} drr_spill;
253 		struct drr_write_embedded {
254 			uint64_t drr_object;
255 			uint64_t drr_offset;
256 			/* logical length, should equal blocksize */
257 			uint64_t drr_length;
258 			uint64_t drr_toguid;
259 			uint8_t drr_compression;
260 			uint8_t drr_etype;
261 			uint8_t drr_pad[6];
262 			uint32_t drr_lsize; /* uncompressed size of payload */
263 			uint32_t drr_psize; /* compr. (real) size of payload */
264 			/* (possibly compressed) content follows */
265 		} drr_write_embedded;
266 
267 		/*
268 		 * Nore: drr_checksum is overlaid with all record types
269 		 * except DRR_BEGIN.  Therefore its (non-pad) members
270 		 * must not overlap with members from the other structs.
271 		 * We accomplish this by putting its members at the very
272 		 * end of the struct.
273 		 */
274 		struct drr_checksum {
275 			uint64_t drr_pad[34];
276 			/*
277 			 * fletcher-4 checksum of everything preceding the
278 			 * checksum.
279 			 */
280 			zio_cksum_t drr_checksum;
281 		} drr_checksum;
282 	} drr_u;
283 } dmu_replay_record_t;
284 
285 /* diff record range types */
286 typedef enum diff_type {
287 	DDR_NONE = 0x1,
288 	DDR_INUSE = 0x2,
289 	DDR_FREE = 0x4
290 } diff_type_t;
291 
292 /*
293  * The diff reports back ranges of free or in-use objects.
294  */
295 typedef struct dmu_diff_record {
296 	uint64_t ddr_type;
297 	uint64_t ddr_first;
298 	uint64_t ddr_last;
299 } dmu_diff_record_t;
300 
301 typedef struct zinject_record {
302 	uint64_t	zi_objset;
303 	uint64_t	zi_object;
304 	uint64_t	zi_start;
305 	uint64_t	zi_end;
306 	uint64_t	zi_guid;
307 	uint32_t	zi_level;
308 	uint32_t	zi_error;
309 	uint64_t	zi_type;
310 	uint32_t	zi_freq;
311 	uint32_t	zi_failfast;
312 	char		zi_func[MAXNAMELEN];
313 	uint32_t	zi_iotype;
314 	int32_t		zi_duration;
315 	uint64_t	zi_timer;
316 	uint64_t	zi_nlanes;
317 	uint32_t	zi_cmd;
318 	uint32_t	zi_pad;
319 } zinject_record_t;
320 
321 #define	ZINJECT_NULL		0x1
322 #define	ZINJECT_FLUSH_ARC	0x2
323 #define	ZINJECT_UNLOAD_SPA	0x4
324 
325 typedef enum zinject_type {
326 	ZINJECT_UNINITIALIZED,
327 	ZINJECT_DATA_FAULT,
328 	ZINJECT_DEVICE_FAULT,
329 	ZINJECT_LABEL_FAULT,
330 	ZINJECT_IGNORED_WRITES,
331 	ZINJECT_PANIC,
332 	ZINJECT_DELAY_IO,
333 } zinject_type_t;
334 
335 typedef struct zfs_share {
336 	uint64_t	z_exportdata;
337 	uint64_t	z_sharedata;
338 	uint64_t	z_sharetype;	/* 0 = share, 1 = unshare */
339 	uint64_t	z_sharemax;  /* max length of share string */
340 } zfs_share_t;
341 
342 /*
343  * ZFS file systems may behave the usual, POSIX-compliant way, where
344  * name lookups are case-sensitive.  They may also be set up so that
345  * all the name lookups are case-insensitive, or so that only some
346  * lookups, the ones that set an FIGNORECASE flag, are case-insensitive.
347  */
348 typedef enum zfs_case {
349 	ZFS_CASE_SENSITIVE,
350 	ZFS_CASE_INSENSITIVE,
351 	ZFS_CASE_MIXED
352 } zfs_case_t;
353 
354 /*
355  * Note: this struct must have the same layout in 32-bit and 64-bit, so
356  * that 32-bit processes (like /sbin/zfs) can pass it to the 64-bit
357  * kernel.  Therefore, we add padding to it so that no "hidden" padding
358  * is automatically added on 64-bit (but not on 32-bit).
359  */
360 typedef struct zfs_cmd {
361 	char		zc_name[MAXPATHLEN];	/* name of pool or dataset */
362 	uint64_t	zc_nvlist_src;		/* really (char *) */
363 	uint64_t	zc_nvlist_src_size;
364 	uint64_t	zc_nvlist_dst;		/* really (char *) */
365 	uint64_t	zc_nvlist_dst_size;
366 	boolean_t	zc_nvlist_dst_filled;	/* put an nvlist in dst? */
367 	int		zc_pad2;
368 
369 	/*
370 	 * The following members are for legacy ioctls which haven't been
371 	 * converted to the new method.
372 	 */
373 	uint64_t	zc_history;		/* really (char *) */
374 	char		zc_value[MAXPATHLEN * 2];
375 	char		zc_string[MAXNAMELEN];
376 	uint64_t	zc_guid;
377 	uint64_t	zc_nvlist_conf;		/* really (char *) */
378 	uint64_t	zc_nvlist_conf_size;
379 	uint64_t	zc_cookie;
380 	uint64_t	zc_objset_type;
381 	uint64_t	zc_perm_action;
382 	uint64_t	zc_history_len;
383 	uint64_t	zc_history_offset;
384 	uint64_t	zc_obj;
385 	uint64_t	zc_iflags;		/* internal to zfs(7fs) */
386 	zfs_share_t	zc_share;
387 	dmu_objset_stats_t zc_objset_stats;
388 	dmu_replay_record_t zc_begin_record;
389 	zinject_record_t zc_inject_record;
390 	uint32_t	zc_defer_destroy;
391 	uint32_t	zc_flags;
392 	uint64_t	zc_action_handle;
393 	int		zc_cleanup_fd;
394 	uint8_t		zc_simple;
395 	uint8_t		zc_pad3[3];
396 	boolean_t	zc_resumable;
397 	uint32_t	zc_pad4;
398 	uint64_t	zc_sendobj;
399 	uint64_t	zc_fromobj;
400 	uint64_t	zc_createtxg;
401 	zfs_stat_t	zc_stat;
402 } zfs_cmd_t;
403 
404 typedef struct zfs_useracct {
405 	char zu_domain[256];
406 	uid_t zu_rid;
407 	uint32_t zu_pad;
408 	uint64_t zu_space;
409 } zfs_useracct_t;
410 
411 #define	ZFSDEV_MAX_MINOR	(1 << 16)
412 #define	ZFS_MIN_MINOR	(ZFSDEV_MAX_MINOR + 1)
413 
414 #define	ZPOOL_EXPORT_AFTER_SPLIT 0x1
415 
416 #ifdef _KERNEL
417 
418 typedef struct zfs_creat {
419 	nvlist_t	*zct_zplprops;
420 	nvlist_t	*zct_props;
421 } zfs_creat_t;
422 
423 extern dev_info_t *zfs_dip;
424 
425 extern int zfs_secpolicy_snapshot_perms(const char *name, cred_t *cr);
426 extern int zfs_secpolicy_rename_perms(const char *from,
427     const char *to, cred_t *cr);
428 extern int zfs_secpolicy_destroy_perms(const char *name, cred_t *cr);
429 extern int zfs_busy(void);
430 extern int zfs_unmount_snap(const char *);
431 extern void zfs_destroy_unmount_origin(const char *);
432 
433 /*
434  * ZFS minor numbers can refer to either a control device instance or
435  * a zvol. Depending on the value of zss_type, zss_data points to either
436  * a zvol_state_t or a zfs_onexit_t.
437  */
438 enum zfs_soft_state_type {
439 	ZSST_ZVOL,
440 	ZSST_CTLDEV
441 };
442 
443 typedef struct zfs_soft_state {
444 	enum zfs_soft_state_type zss_type;
445 	void *zss_data;
446 } zfs_soft_state_t;
447 
448 extern void *zfsdev_get_soft_state(minor_t minor,
449     enum zfs_soft_state_type which);
450 extern minor_t zfsdev_minor_alloc(void);
451 
452 extern void *zfsdev_state;
453 extern kmutex_t zfsdev_state_lock;
454 
455 #endif	/* _KERNEL */
456 
457 #ifdef	__cplusplus
458 }
459 #endif
460 
461 #endif	/* _SYS_ZFS_IOCTL_H */
462