]>
Commit | Line | Data |
---|---|---|
6f1ffb06 MA |
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 | /* | |
5dc8b736 | 23 | * Copyright (c) 2012, 2014 by Delphix. All rights reserved. |
95fd54a1 | 24 | * Copyright (c) 2013 Steven Hartland. All rights reserved. |
6f1ffb06 MA |
25 | */ |
26 | ||
27 | /* | |
28 | * LibZFS_Core (lzc) is intended to replace most functionality in libzfs. | |
29 | * It has the following characteristics: | |
30 | * | |
31 | * - Thread Safe. libzfs_core is accessible concurrently from multiple | |
32 | * threads. This is accomplished primarily by avoiding global data | |
33 | * (e.g. caching). Since it's thread-safe, there is no reason for a | |
34 | * process to have multiple libzfs "instances". Therefore, we store | |
35 | * our few pieces of data (e.g. the file descriptor) in global | |
36 | * variables. The fd is reference-counted so that the libzfs_core | |
37 | * library can be "initialized" multiple times (e.g. by different | |
38 | * consumers within the same process). | |
39 | * | |
40 | * - Committed Interface. The libzfs_core interface will be committed, | |
41 | * therefore consumers can compile against it and be confident that | |
42 | * their code will continue to work on future releases of this code. | |
43 | * Currently, the interface is Evolving (not Committed), but we intend | |
44 | * to commit to it once it is more complete and we determine that it | |
45 | * meets the needs of all consumers. | |
46 | * | |
b8fce77b | 47 | * - Programmatic Error Handling. libzfs_core communicates errors with |
6f1ffb06 MA |
48 | * defined error numbers, and doesn't print anything to stdout/stderr. |
49 | * | |
50 | * - Thin Layer. libzfs_core is a thin layer, marshaling arguments | |
51 | * to/from the kernel ioctls. There is generally a 1:1 correspondence | |
52 | * between libzfs_core functions and ioctls to /dev/zfs. | |
53 | * | |
54 | * - Clear Atomicity. Because libzfs_core functions are generally 1:1 | |
55 | * with kernel ioctls, and kernel ioctls are general atomic, each | |
56 | * libzfs_core function is atomic. For example, creating multiple | |
57 | * snapshots with a single call to lzc_snapshot() is atomic -- it | |
58 | * can't fail with only some of the requested snapshots created, even | |
59 | * in the event of power loss or system crash. | |
60 | * | |
61 | * - Continued libzfs Support. Some higher-level operations (e.g. | |
62 | * support for "zfs send -R") are too complicated to fit the scope of | |
63 | * libzfs_core. This functionality will continue to live in libzfs. | |
64 | * Where appropriate, libzfs will use the underlying atomic operations | |
65 | * of libzfs_core. For example, libzfs may implement "zfs send -R | | |
66 | * zfs receive" by using individual "send one snapshot", rename, | |
67 | * destroy, and "receive one snapshot" operations in libzfs_core. | |
68 | * /sbin/zfs and /zbin/zpool will link with both libzfs and | |
69 | * libzfs_core. Other consumers should aim to use only libzfs_core, | |
70 | * since that will be the supported, stable interface going forwards. | |
71 | */ | |
72 | ||
73 | #include <libzfs_core.h> | |
74 | #include <ctype.h> | |
75 | #include <unistd.h> | |
76 | #include <stdlib.h> | |
77 | #include <string.h> | |
78 | #include <errno.h> | |
79 | #include <fcntl.h> | |
80 | #include <pthread.h> | |
81 | #include <sys/nvpair.h> | |
82 | #include <sys/param.h> | |
83 | #include <sys/types.h> | |
84 | #include <sys/stat.h> | |
85 | #include <sys/zfs_ioctl.h> | |
86 | ||
87 | static int g_fd; | |
88 | static pthread_mutex_t g_lock = PTHREAD_MUTEX_INITIALIZER; | |
89 | static int g_refcount; | |
90 | ||
91 | int | |
92 | libzfs_core_init(void) | |
93 | { | |
94 | (void) pthread_mutex_lock(&g_lock); | |
95 | if (g_refcount == 0) { | |
96 | g_fd = open("/dev/zfs", O_RDWR); | |
97 | if (g_fd < 0) { | |
98 | (void) pthread_mutex_unlock(&g_lock); | |
99 | return (errno); | |
100 | } | |
101 | } | |
102 | g_refcount++; | |
103 | (void) pthread_mutex_unlock(&g_lock); | |
104 | return (0); | |
105 | } | |
106 | ||
107 | void | |
108 | libzfs_core_fini(void) | |
109 | { | |
110 | (void) pthread_mutex_lock(&g_lock); | |
111 | ASSERT3S(g_refcount, >, 0); | |
112 | g_refcount--; | |
113 | if (g_refcount == 0) | |
114 | (void) close(g_fd); | |
115 | (void) pthread_mutex_unlock(&g_lock); | |
116 | } | |
117 | ||
118 | static int | |
119 | lzc_ioctl(zfs_ioc_t ioc, const char *name, | |
120 | nvlist_t *source, nvlist_t **resultp) | |
121 | { | |
13fe0198 | 122 | zfs_cmd_t zc = {"\0"}; |
6f1ffb06 MA |
123 | int error = 0; |
124 | char *packed; | |
125 | size_t size; | |
126 | ||
127 | ASSERT3S(g_refcount, >, 0); | |
128 | ||
129 | (void) strlcpy(zc.zc_name, name, sizeof (zc.zc_name)); | |
130 | ||
131 | packed = fnvlist_pack(source, &size); | |
132 | zc.zc_nvlist_src = (uint64_t)(uintptr_t)packed; | |
133 | zc.zc_nvlist_src_size = size; | |
134 | ||
135 | if (resultp != NULL) { | |
13fe0198 | 136 | *resultp = NULL; |
6f1ffb06 MA |
137 | zc.zc_nvlist_dst_size = MAX(size * 2, 128 * 1024); |
138 | zc.zc_nvlist_dst = (uint64_t)(uintptr_t) | |
139 | malloc(zc.zc_nvlist_dst_size); | |
140 | if (zc.zc_nvlist_dst == (uint64_t)0) { | |
141 | error = ENOMEM; | |
142 | goto out; | |
143 | } | |
144 | } | |
145 | ||
146 | while (ioctl(g_fd, ioc, &zc) != 0) { | |
147 | if (errno == ENOMEM && resultp != NULL) { | |
148 | free((void *)(uintptr_t)zc.zc_nvlist_dst); | |
149 | zc.zc_nvlist_dst_size *= 2; | |
150 | zc.zc_nvlist_dst = (uint64_t)(uintptr_t) | |
151 | malloc(zc.zc_nvlist_dst_size); | |
152 | if (zc.zc_nvlist_dst == (uint64_t)0) { | |
153 | error = ENOMEM; | |
154 | goto out; | |
155 | } | |
156 | } else { | |
157 | error = errno; | |
158 | break; | |
159 | } | |
160 | } | |
161 | if (zc.zc_nvlist_dst_filled) { | |
162 | *resultp = fnvlist_unpack((void *)(uintptr_t)zc.zc_nvlist_dst, | |
163 | zc.zc_nvlist_dst_size); | |
6f1ffb06 MA |
164 | } |
165 | ||
166 | out: | |
167 | fnvlist_pack_free(packed, size); | |
168 | free((void *)(uintptr_t)zc.zc_nvlist_dst); | |
169 | return (error); | |
170 | } | |
171 | ||
172 | int | |
173 | lzc_create(const char *fsname, dmu_objset_type_t type, nvlist_t *props) | |
174 | { | |
175 | int error; | |
176 | nvlist_t *args = fnvlist_alloc(); | |
177 | fnvlist_add_int32(args, "type", type); | |
178 | if (props != NULL) | |
179 | fnvlist_add_nvlist(args, "props", props); | |
180 | error = lzc_ioctl(ZFS_IOC_CREATE, fsname, args, NULL); | |
181 | nvlist_free(args); | |
182 | return (error); | |
183 | } | |
184 | ||
185 | int | |
186 | lzc_clone(const char *fsname, const char *origin, | |
187 | nvlist_t *props) | |
188 | { | |
189 | int error; | |
190 | nvlist_t *args = fnvlist_alloc(); | |
191 | fnvlist_add_string(args, "origin", origin); | |
192 | if (props != NULL) | |
193 | fnvlist_add_nvlist(args, "props", props); | |
194 | error = lzc_ioctl(ZFS_IOC_CLONE, fsname, args, NULL); | |
195 | nvlist_free(args); | |
196 | return (error); | |
197 | } | |
198 | ||
199 | /* | |
200 | * Creates snapshots. | |
201 | * | |
202 | * The keys in the snaps nvlist are the snapshots to be created. | |
203 | * They must all be in the same pool. | |
204 | * | |
205 | * The props nvlist is properties to set. Currently only user properties | |
206 | * are supported. { user:prop_name -> string value } | |
207 | * | |
208 | * The returned results nvlist will have an entry for each snapshot that failed. | |
209 | * The value will be the (int32) error code. | |
210 | * | |
211 | * The return value will be 0 if all snapshots were created, otherwise it will | |
13fe0198 | 212 | * be the errno of a (unspecified) snapshot that failed. |
6f1ffb06 MA |
213 | */ |
214 | int | |
215 | lzc_snapshot(nvlist_t *snaps, nvlist_t *props, nvlist_t **errlist) | |
216 | { | |
217 | nvpair_t *elem; | |
218 | nvlist_t *args; | |
219 | int error; | |
eca7b760 | 220 | char pool[ZFS_MAX_DATASET_NAME_LEN]; |
6f1ffb06 MA |
221 | |
222 | *errlist = NULL; | |
223 | ||
224 | /* determine the pool name */ | |
225 | elem = nvlist_next_nvpair(snaps, NULL); | |
226 | if (elem == NULL) | |
227 | return (0); | |
228 | (void) strlcpy(pool, nvpair_name(elem), sizeof (pool)); | |
229 | pool[strcspn(pool, "/@")] = '\0'; | |
230 | ||
231 | args = fnvlist_alloc(); | |
232 | fnvlist_add_nvlist(args, "snaps", snaps); | |
233 | if (props != NULL) | |
234 | fnvlist_add_nvlist(args, "props", props); | |
235 | ||
236 | error = lzc_ioctl(ZFS_IOC_SNAPSHOT, pool, args, errlist); | |
237 | nvlist_free(args); | |
238 | ||
239 | return (error); | |
240 | } | |
241 | ||
242 | /* | |
243 | * Destroys snapshots. | |
244 | * | |
245 | * The keys in the snaps nvlist are the snapshots to be destroyed. | |
246 | * They must all be in the same pool. | |
247 | * | |
248 | * Snapshots that do not exist will be silently ignored. | |
249 | * | |
250 | * If 'defer' is not set, and a snapshot has user holds or clones, the | |
251 | * destroy operation will fail and none of the snapshots will be | |
252 | * destroyed. | |
253 | * | |
254 | * If 'defer' is set, and a snapshot has user holds or clones, it will be | |
255 | * marked for deferred destruction, and will be destroyed when the last hold | |
256 | * or clone is removed/destroyed. | |
257 | * | |
258 | * The return value will be 0 if all snapshots were destroyed (or marked for | |
1a077756 | 259 | * later destruction if 'defer' is set) or didn't exist to begin with. |
6f1ffb06 | 260 | * |
13fe0198 | 261 | * Otherwise the return value will be the errno of a (unspecified) snapshot |
6f1ffb06 MA |
262 | * that failed, no snapshots will be destroyed, and the errlist will have an |
263 | * entry for each snapshot that failed. The value in the errlist will be | |
264 | * the (int32) error code. | |
265 | */ | |
266 | int | |
267 | lzc_destroy_snaps(nvlist_t *snaps, boolean_t defer, nvlist_t **errlist) | |
268 | { | |
269 | nvpair_t *elem; | |
270 | nvlist_t *args; | |
271 | int error; | |
eca7b760 | 272 | char pool[ZFS_MAX_DATASET_NAME_LEN]; |
6f1ffb06 MA |
273 | |
274 | /* determine the pool name */ | |
275 | elem = nvlist_next_nvpair(snaps, NULL); | |
276 | if (elem == NULL) | |
277 | return (0); | |
278 | (void) strlcpy(pool, nvpair_name(elem), sizeof (pool)); | |
279 | pool[strcspn(pool, "/@")] = '\0'; | |
280 | ||
281 | args = fnvlist_alloc(); | |
282 | fnvlist_add_nvlist(args, "snaps", snaps); | |
283 | if (defer) | |
284 | fnvlist_add_boolean(args, "defer"); | |
285 | ||
286 | error = lzc_ioctl(ZFS_IOC_DESTROY_SNAPS, pool, args, errlist); | |
287 | nvlist_free(args); | |
288 | ||
289 | return (error); | |
6f1ffb06 MA |
290 | } |
291 | ||
292 | int | |
293 | lzc_snaprange_space(const char *firstsnap, const char *lastsnap, | |
294 | uint64_t *usedp) | |
295 | { | |
296 | nvlist_t *args; | |
297 | nvlist_t *result; | |
298 | int err; | |
eca7b760 | 299 | char fs[ZFS_MAX_DATASET_NAME_LEN]; |
6f1ffb06 MA |
300 | char *atp; |
301 | ||
302 | /* determine the fs name */ | |
303 | (void) strlcpy(fs, firstsnap, sizeof (fs)); | |
304 | atp = strchr(fs, '@'); | |
305 | if (atp == NULL) | |
306 | return (EINVAL); | |
307 | *atp = '\0'; | |
308 | ||
309 | args = fnvlist_alloc(); | |
310 | fnvlist_add_string(args, "firstsnap", firstsnap); | |
311 | ||
312 | err = lzc_ioctl(ZFS_IOC_SPACE_SNAPS, lastsnap, args, &result); | |
313 | nvlist_free(args); | |
314 | if (err == 0) | |
315 | *usedp = fnvlist_lookup_uint64(result, "used"); | |
316 | fnvlist_free(result); | |
317 | ||
318 | return (err); | |
319 | } | |
320 | ||
321 | boolean_t | |
322 | lzc_exists(const char *dataset) | |
323 | { | |
324 | /* | |
325 | * The objset_stats ioctl is still legacy, so we need to construct our | |
326 | * own zfs_cmd_t rather than using zfsc_ioctl(). | |
327 | */ | |
13fe0198 | 328 | zfs_cmd_t zc = {"\0"}; |
6f1ffb06 MA |
329 | |
330 | (void) strlcpy(zc.zc_name, dataset, sizeof (zc.zc_name)); | |
331 | return (ioctl(g_fd, ZFS_IOC_OBJSET_STATS, &zc) == 0); | |
332 | } | |
333 | ||
13fe0198 MA |
334 | /* |
335 | * Create "user holds" on snapshots. If there is a hold on a snapshot, | |
336 | * the snapshot can not be destroyed. (However, it can be marked for deletion | |
337 | * by lzc_destroy_snaps(defer=B_TRUE).) | |
338 | * | |
339 | * The keys in the nvlist are snapshot names. | |
340 | * The snapshots must all be in the same pool. | |
341 | * The value is the name of the hold (string type). | |
342 | * | |
343 | * If cleanup_fd is not -1, it must be the result of open("/dev/zfs", O_EXCL). | |
344 | * In this case, when the cleanup_fd is closed (including on process | |
345 | * termination), the holds will be released. If the system is shut down | |
346 | * uncleanly, the holds will be released when the pool is next opened | |
347 | * or imported. | |
348 | * | |
95fd54a1 | 349 | * Holds for snapshots which don't exist will be skipped and have an entry |
1a077756 | 350 | * added to errlist, but will not cause an overall failure. |
95fd54a1 | 351 | * |
1a077756 | 352 | * The return value will be 0 if all holds, for snapshots that existed, |
b8fce77b | 353 | * were successfully created. |
95fd54a1 SH |
354 | * |
355 | * Otherwise the return value will be the errno of a (unspecified) hold that | |
356 | * failed and no holds will be created. | |
357 | * | |
358 | * In all cases the errlist will have an entry for each hold that failed | |
359 | * (name = snapshot), with its value being the error code (int32). | |
13fe0198 MA |
360 | */ |
361 | int | |
362 | lzc_hold(nvlist_t *holds, int cleanup_fd, nvlist_t **errlist) | |
363 | { | |
eca7b760 | 364 | char pool[ZFS_MAX_DATASET_NAME_LEN]; |
13fe0198 MA |
365 | nvlist_t *args; |
366 | nvpair_t *elem; | |
367 | int error; | |
368 | ||
369 | /* determine the pool name */ | |
370 | elem = nvlist_next_nvpair(holds, NULL); | |
371 | if (elem == NULL) | |
372 | return (0); | |
373 | (void) strlcpy(pool, nvpair_name(elem), sizeof (pool)); | |
374 | pool[strcspn(pool, "/@")] = '\0'; | |
375 | ||
376 | args = fnvlist_alloc(); | |
377 | fnvlist_add_nvlist(args, "holds", holds); | |
378 | if (cleanup_fd != -1) | |
379 | fnvlist_add_int32(args, "cleanup_fd", cleanup_fd); | |
380 | ||
381 | error = lzc_ioctl(ZFS_IOC_HOLD, pool, args, errlist); | |
382 | nvlist_free(args); | |
383 | return (error); | |
384 | } | |
385 | ||
386 | /* | |
387 | * Release "user holds" on snapshots. If the snapshot has been marked for | |
388 | * deferred destroy (by lzc_destroy_snaps(defer=B_TRUE)), it does not have | |
389 | * any clones, and all the user holds are removed, then the snapshot will be | |
390 | * destroyed. | |
391 | * | |
392 | * The keys in the nvlist are snapshot names. | |
393 | * The snapshots must all be in the same pool. | |
d5884c34 | 394 | * The value is an nvlist whose keys are the holds to remove. |
13fe0198 | 395 | * |
95fd54a1 | 396 | * Holds which failed to release because they didn't exist will have an entry |
1a077756 | 397 | * added to errlist, but will not cause an overall failure. |
95fd54a1 SH |
398 | * |
399 | * The return value will be 0 if the nvl holds was empty or all holds that | |
1a077756 | 400 | * existed, were successfully removed. |
95fd54a1 SH |
401 | * |
402 | * Otherwise the return value will be the errno of a (unspecified) hold that | |
403 | * failed to release and no holds will be released. | |
404 | * | |
405 | * In all cases the errlist will have an entry for each hold that failed to | |
406 | * to release. | |
13fe0198 MA |
407 | */ |
408 | int | |
409 | lzc_release(nvlist_t *holds, nvlist_t **errlist) | |
410 | { | |
eca7b760 | 411 | char pool[ZFS_MAX_DATASET_NAME_LEN]; |
13fe0198 MA |
412 | nvpair_t *elem; |
413 | ||
414 | /* determine the pool name */ | |
415 | elem = nvlist_next_nvpair(holds, NULL); | |
416 | if (elem == NULL) | |
417 | return (0); | |
418 | (void) strlcpy(pool, nvpair_name(elem), sizeof (pool)); | |
419 | pool[strcspn(pool, "/@")] = '\0'; | |
420 | ||
421 | return (lzc_ioctl(ZFS_IOC_RELEASE, pool, holds, errlist)); | |
422 | } | |
423 | ||
424 | /* | |
425 | * Retrieve list of user holds on the specified snapshot. | |
426 | * | |
d5884c34 | 427 | * On success, *holdsp will be set to an nvlist which the caller must free. |
13fe0198 MA |
428 | * The keys are the names of the holds, and the value is the creation time |
429 | * of the hold (uint64) in seconds since the epoch. | |
430 | */ | |
431 | int | |
432 | lzc_get_holds(const char *snapname, nvlist_t **holdsp) | |
433 | { | |
434 | int error; | |
435 | nvlist_t *innvl = fnvlist_alloc(); | |
436 | error = lzc_ioctl(ZFS_IOC_GET_HOLDS, snapname, innvl, holdsp); | |
437 | fnvlist_free(innvl); | |
438 | return (error); | |
439 | } | |
440 | ||
6f1ffb06 | 441 | /* |
9b67f605 MA |
442 | * Generate a zfs send stream for the specified snapshot and write it to |
443 | * the specified file descriptor. | |
da536844 MA |
444 | * |
445 | * "snapname" is the full name of the snapshot to send (e.g. "pool/fs@snap") | |
446 | * | |
447 | * If "from" is NULL, a full (non-incremental) stream will be sent. | |
448 | * If "from" is non-NULL, it must be the full name of a snapshot or | |
449 | * bookmark to send an incremental from (e.g. "pool/fs@earlier_snap" or | |
450 | * "pool/fs#earlier_bmark"). If non-NULL, the specified snapshot or | |
451 | * bookmark must represent an earlier point in the history of "snapname"). | |
452 | * It can be an earlier snapshot in the same filesystem or zvol as "snapname", | |
453 | * or it can be the origin of "snapname"'s filesystem, or an earlier | |
454 | * snapshot in the origin, etc. | |
455 | * | |
456 | * "fd" is the file descriptor to write the send stream to. | |
9b67f605 | 457 | * |
f1512ee6 MA |
458 | * If "flags" contains LZC_SEND_FLAG_LARGE_BLOCK, the stream is permitted |
459 | * to contain DRR_WRITE records with drr_length > 128K, and DRR_OBJECT | |
460 | * records with drr_blksz > 128K. | |
461 | * | |
9b67f605 MA |
462 | * If "flags" contains LZC_SEND_FLAG_EMBED_DATA, the stream is permitted |
463 | * to contain DRR_WRITE_EMBEDDED records with drr_etype==BP_EMBEDDED_TYPE_DATA, | |
464 | * which the receiving system must support (as indicated by support | |
465 | * for the "embedded_data" feature). | |
6f1ffb06 MA |
466 | */ |
467 | int | |
9b67f605 MA |
468 | lzc_send(const char *snapname, const char *from, int fd, |
469 | enum lzc_send_flags flags) | |
47dfff3b MA |
470 | { |
471 | return (lzc_send_resume(snapname, from, fd, flags, 0, 0)); | |
472 | } | |
473 | ||
474 | int | |
475 | lzc_send_resume(const char *snapname, const char *from, int fd, | |
476 | enum lzc_send_flags flags, uint64_t resumeobj, uint64_t resumeoff) | |
6f1ffb06 MA |
477 | { |
478 | nvlist_t *args; | |
479 | int err; | |
480 | ||
481 | args = fnvlist_alloc(); | |
482 | fnvlist_add_int32(args, "fd", fd); | |
da536844 MA |
483 | if (from != NULL) |
484 | fnvlist_add_string(args, "fromsnap", from); | |
f1512ee6 MA |
485 | if (flags & LZC_SEND_FLAG_LARGE_BLOCK) |
486 | fnvlist_add_boolean(args, "largeblockok"); | |
2aa34383 DK |
487 | if (flags & LZC_SEND_FLAG_COMPRESS) |
488 | fnvlist_add_boolean(args, "compressok"); | |
9b67f605 MA |
489 | if (flags & LZC_SEND_FLAG_EMBED_DATA) |
490 | fnvlist_add_boolean(args, "embedok"); | |
47dfff3b MA |
491 | if (resumeobj != 0 || resumeoff != 0) { |
492 | fnvlist_add_uint64(args, "resume_object", resumeobj); | |
493 | fnvlist_add_uint64(args, "resume_offset", resumeoff); | |
494 | } | |
6f1ffb06 MA |
495 | err = lzc_ioctl(ZFS_IOC_SEND_NEW, snapname, args, NULL); |
496 | nvlist_free(args); | |
497 | return (err); | |
498 | } | |
499 | ||
500 | /* | |
5dc8b736 MG |
501 | * "from" can be NULL, a snapshot, or a bookmark. |
502 | * | |
503 | * If from is NULL, a full (non-incremental) stream will be estimated. This | |
504 | * is calculated very efficiently. | |
505 | * | |
506 | * If from is a snapshot, lzc_send_space uses the deadlists attached to | |
507 | * each snapshot to efficiently estimate the stream size. | |
508 | * | |
509 | * If from is a bookmark, the indirect blocks in the destination snapshot | |
510 | * are traversed, looking for blocks with a birth time since the creation TXG of | |
511 | * the snapshot this bookmark was created from. This will result in | |
512 | * significantly more I/O and be less efficient than a send space estimation on | |
513 | * an equivalent snapshot. | |
6f1ffb06 MA |
514 | */ |
515 | int | |
2aa34383 DK |
516 | lzc_send_space(const char *snapname, const char *from, |
517 | enum lzc_send_flags flags, uint64_t *spacep) | |
6f1ffb06 MA |
518 | { |
519 | nvlist_t *args; | |
520 | nvlist_t *result; | |
521 | int err; | |
522 | ||
523 | args = fnvlist_alloc(); | |
5dc8b736 MG |
524 | if (from != NULL) |
525 | fnvlist_add_string(args, "from", from); | |
2aa34383 DK |
526 | if (flags & LZC_SEND_FLAG_LARGE_BLOCK) |
527 | fnvlist_add_boolean(args, "largeblockok"); | |
528 | if (flags & LZC_SEND_FLAG_EMBED_DATA) | |
529 | fnvlist_add_boolean(args, "embedok"); | |
530 | if (flags & LZC_SEND_FLAG_COMPRESS) | |
531 | fnvlist_add_boolean(args, "compressok"); | |
6f1ffb06 MA |
532 | err = lzc_ioctl(ZFS_IOC_SEND_SPACE, snapname, args, &result); |
533 | nvlist_free(args); | |
534 | if (err == 0) | |
535 | *spacep = fnvlist_lookup_uint64(result, "space"); | |
536 | nvlist_free(result); | |
537 | return (err); | |
538 | } | |
539 | ||
540 | static int | |
541 | recv_read(int fd, void *buf, int ilen) | |
542 | { | |
543 | char *cp = buf; | |
544 | int rv; | |
545 | int len = ilen; | |
546 | ||
547 | do { | |
548 | rv = read(fd, cp, len); | |
549 | cp += rv; | |
550 | len -= rv; | |
551 | } while (rv > 0); | |
552 | ||
553 | if (rv < 0 || len != 0) | |
554 | return (EIO); | |
555 | ||
556 | return (0); | |
557 | } | |
558 | ||
43e52edd BB |
559 | /* |
560 | * Linux adds ZFS_IOC_RECV_NEW for resumable streams and preserves the legacy | |
561 | * ZFS_IOC_RECV user/kernel interface. The new interface supports all stream | |
562 | * options but is currently only used for resumable streams. This way updated | |
563 | * user space utilities will interoperate with older kernel modules. | |
564 | * | |
565 | * Non-Linux OpenZFS platforms have opted to modify the legacy interface. | |
566 | */ | |
47dfff3b | 567 | static int |
fd41e935 | 568 | recv_impl(const char *snapname, nvlist_t *props, const char *origin, |
43e52edd BB |
569 | boolean_t force, boolean_t resumable, int input_fd, |
570 | const dmu_replay_record_t *begin_record, int cleanup_fd, | |
571 | uint64_t *read_bytes, uint64_t *errflags, uint64_t *action_handle, | |
572 | nvlist_t **errors) | |
6f1ffb06 | 573 | { |
43e52edd BB |
574 | dmu_replay_record_t drr; |
575 | char fsname[MAXPATHLEN]; | |
6f1ffb06 | 576 | char *atp; |
6f1ffb06 MA |
577 | int error; |
578 | ||
43e52edd BB |
579 | /* Set 'fsname' to the name of containing filesystem */ |
580 | (void) strlcpy(fsname, snapname, sizeof (fsname)); | |
581 | atp = strchr(fsname, '@'); | |
6f1ffb06 MA |
582 | if (atp == NULL) |
583 | return (EINVAL); | |
584 | *atp = '\0'; | |
585 | ||
43e52edd BB |
586 | /* If the fs does not exist, try its parent. */ |
587 | if (!lzc_exists(fsname)) { | |
588 | char *slashp = strrchr(fsname, '/'); | |
6f1ffb06 MA |
589 | if (slashp == NULL) |
590 | return (ENOENT); | |
591 | *slashp = '\0'; | |
43e52edd | 592 | } |
6f1ffb06 | 593 | |
43e52edd BB |
594 | /* |
595 | * The begin_record is normally a non-byteswapped BEGIN record. | |
596 | * For resumable streams it may be set to any non-byteswapped | |
597 | * dmu_replay_record_t. | |
598 | */ | |
599 | if (begin_record == NULL) { | |
600 | error = recv_read(input_fd, &drr, sizeof (drr)); | |
601 | if (error != 0) | |
602 | return (error); | |
603 | } else { | |
604 | drr = *begin_record; | |
6f1ffb06 MA |
605 | } |
606 | ||
43e52edd BB |
607 | if (resumable) { |
608 | nvlist_t *outnvl = NULL; | |
609 | nvlist_t *innvl = fnvlist_alloc(); | |
6f1ffb06 | 610 | |
43e52edd | 611 | fnvlist_add_string(innvl, "snapname", snapname); |
6f1ffb06 | 612 | |
43e52edd BB |
613 | if (props != NULL) |
614 | fnvlist_add_nvlist(innvl, "props", props); | |
6f1ffb06 | 615 | |
43e52edd BB |
616 | if (origin != NULL && strlen(origin)) |
617 | fnvlist_add_string(innvl, "origin", origin); | |
618 | ||
619 | fnvlist_add_byte_array(innvl, "begin_record", | |
02730c33 | 620 | (uchar_t *)&drr, sizeof (drr)); |
43e52edd BB |
621 | |
622 | fnvlist_add_int32(innvl, "input_fd", input_fd); | |
623 | ||
624 | if (force) | |
625 | fnvlist_add_boolean(innvl, "force"); | |
626 | ||
627 | if (resumable) | |
628 | fnvlist_add_boolean(innvl, "resumable"); | |
629 | ||
630 | if (cleanup_fd >= 0) | |
631 | fnvlist_add_int32(innvl, "cleanup_fd", cleanup_fd); | |
632 | ||
633 | if (action_handle != NULL) | |
634 | fnvlist_add_uint64(innvl, "action_handle", | |
635 | *action_handle); | |
636 | ||
637 | error = lzc_ioctl(ZFS_IOC_RECV_NEW, fsname, innvl, &outnvl); | |
638 | ||
639 | if (error == 0 && read_bytes != NULL) | |
640 | error = nvlist_lookup_uint64(outnvl, "read_bytes", | |
641 | read_bytes); | |
642 | ||
643 | if (error == 0 && errflags != NULL) | |
644 | error = nvlist_lookup_uint64(outnvl, "error_flags", | |
645 | errflags); | |
646 | ||
647 | if (error == 0 && action_handle != NULL) | |
648 | error = nvlist_lookup_uint64(outnvl, "action_handle", | |
649 | action_handle); | |
650 | ||
651 | if (error == 0 && errors != NULL) { | |
652 | nvlist_t *nvl; | |
653 | error = nvlist_lookup_nvlist(outnvl, "errors", &nvl); | |
654 | if (error == 0) | |
655 | *errors = fnvlist_dup(nvl); | |
656 | } | |
657 | ||
658 | fnvlist_free(innvl); | |
659 | fnvlist_free(outnvl); | |
fd41e935 | 660 | } else { |
43e52edd BB |
661 | zfs_cmd_t zc = {"\0"}; |
662 | char *packed = NULL; | |
663 | size_t size; | |
6f1ffb06 | 664 | |
43e52edd | 665 | ASSERT3S(g_refcount, >, 0); |
6f1ffb06 | 666 | |
43e52edd BB |
667 | (void) strlcpy(zc.zc_name, fsname, sizeof (zc.zc_value)); |
668 | (void) strlcpy(zc.zc_value, snapname, sizeof (zc.zc_value)); | |
6f1ffb06 | 669 | |
43e52edd BB |
670 | if (props != NULL) { |
671 | packed = fnvlist_pack(props, &size); | |
672 | zc.zc_nvlist_src = (uint64_t)(uintptr_t)packed; | |
673 | zc.zc_nvlist_src_size = size; | |
674 | } | |
47dfff3b | 675 | |
43e52edd BB |
676 | if (origin != NULL) |
677 | (void) strlcpy(zc.zc_string, origin, | |
678 | sizeof (zc.zc_string)); | |
6f1ffb06 | 679 | |
43e52edd BB |
680 | ASSERT3S(drr.drr_type, ==, DRR_BEGIN); |
681 | zc.zc_begin_record = drr.drr_u.drr_begin; | |
682 | zc.zc_guid = force; | |
683 | zc.zc_cookie = input_fd; | |
684 | zc.zc_cleanup_fd = -1; | |
685 | zc.zc_action_handle = 0; | |
686 | ||
687 | if (cleanup_fd >= 0) | |
688 | zc.zc_cleanup_fd = cleanup_fd; | |
689 | ||
690 | if (action_handle != NULL) | |
691 | zc.zc_action_handle = *action_handle; | |
692 | ||
693 | zc.zc_nvlist_dst_size = 128 * 1024; | |
694 | zc.zc_nvlist_dst = (uint64_t)(uintptr_t) | |
695 | malloc(zc.zc_nvlist_dst_size); | |
696 | ||
697 | error = ioctl(g_fd, ZFS_IOC_RECV, &zc); | |
698 | if (error != 0) { | |
699 | error = errno; | |
700 | } else { | |
701 | if (read_bytes != NULL) | |
702 | *read_bytes = zc.zc_cookie; | |
703 | ||
704 | if (errflags != NULL) | |
705 | *errflags = zc.zc_obj; | |
706 | ||
707 | if (action_handle != NULL) | |
708 | *action_handle = zc.zc_action_handle; | |
709 | ||
710 | if (errors != NULL) | |
711 | VERIFY0(nvlist_unpack( | |
712 | (void *)(uintptr_t)zc.zc_nvlist_dst, | |
713 | zc.zc_nvlist_dst_size, errors, KM_SLEEP)); | |
714 | } | |
715 | ||
716 | if (packed != NULL) | |
717 | fnvlist_pack_free(packed, size); | |
718 | free((void *)(uintptr_t)zc.zc_nvlist_dst); | |
719 | } | |
6f1ffb06 | 720 | |
6f1ffb06 MA |
721 | return (error); |
722 | } | |
46ba1e59 | 723 | |
47dfff3b MA |
724 | /* |
725 | * The simplest receive case: receive from the specified fd, creating the | |
726 | * specified snapshot. Apply the specified properties as "received" properties | |
727 | * (which can be overridden by locally-set properties). If the stream is a | |
728 | * clone, its origin snapshot must be specified by 'origin'. The 'force' | |
729 | * flag will cause the target filesystem to be rolled back or destroyed if | |
730 | * necessary to receive. | |
731 | * | |
732 | * Return 0 on success or an errno on failure. | |
733 | * | |
734 | * Note: this interface does not work on dedup'd streams | |
735 | * (those with DMU_BACKUP_FEATURE_DEDUP). | |
736 | */ | |
737 | int | |
738 | lzc_receive(const char *snapname, nvlist_t *props, const char *origin, | |
739 | boolean_t force, int fd) | |
740 | { | |
43e52edd BB |
741 | return (recv_impl(snapname, props, origin, force, B_FALSE, fd, |
742 | NULL, -1, NULL, NULL, NULL, NULL)); | |
47dfff3b MA |
743 | } |
744 | ||
745 | /* | |
746 | * Like lzc_receive, but if the receive fails due to premature stream | |
747 | * termination, the intermediate state will be preserved on disk. In this | |
748 | * case, ECKSUM will be returned. The receive may subsequently be resumed | |
749 | * with a resuming send stream generated by lzc_send_resume(). | |
750 | */ | |
751 | int | |
752 | lzc_receive_resumable(const char *snapname, nvlist_t *props, const char *origin, | |
753 | boolean_t force, int fd) | |
754 | { | |
43e52edd BB |
755 | return (recv_impl(snapname, props, origin, force, B_TRUE, fd, |
756 | NULL, -1, NULL, NULL, NULL, NULL)); | |
fd41e935 BB |
757 | } |
758 | ||
759 | /* | |
760 | * Like lzc_receive, but allows the caller to read the begin record and then to | |
761 | * pass it in. That could be useful if the caller wants to derive, for example, | |
762 | * the snapname or the origin parameters based on the information contained in | |
763 | * the begin record. | |
764 | * The begin record must be in its original form as read from the stream, | |
765 | * in other words, it should not be byteswapped. | |
766 | * | |
767 | * The 'resumable' parameter allows to obtain the same behavior as with | |
768 | * lzc_receive_resumable. | |
769 | */ | |
770 | int | |
771 | lzc_receive_with_header(const char *snapname, nvlist_t *props, | |
772 | const char *origin, boolean_t force, boolean_t resumable, int fd, | |
773 | const dmu_replay_record_t *begin_record) | |
774 | { | |
775 | if (begin_record == NULL) | |
776 | return (EINVAL); | |
777 | return (recv_impl(snapname, props, origin, force, resumable, fd, | |
43e52edd BB |
778 | begin_record, -1, NULL, NULL, NULL, NULL)); |
779 | } | |
780 | ||
781 | /* | |
782 | * Like lzc_receive, but allows the caller to pass all supported arguments | |
783 | * and retrieve all values returned. The only additional input parameter | |
784 | * is 'cleanup_fd' which is used to set a cleanup-on-exit file descriptor. | |
785 | * | |
786 | * The following parameters all provide return values. Several may be set | |
787 | * in the failure case and will contain additional information. | |
788 | * | |
789 | * The 'read_bytes' value will be set to the total number of bytes read. | |
790 | * | |
791 | * The 'errflags' value will contain zprop_errflags_t flags which are | |
792 | * used to describe any failures. | |
793 | * | |
794 | * The 'action_handle' is used to pass the handle for this guid/ds mapping. | |
795 | * It should be set to zero on first call and will contain an updated handle | |
796 | * on success, it should be passed in subsequent calls. | |
797 | * | |
798 | * The 'errors' nvlist contains an entry for each unapplied received | |
799 | * property. Callers are responsible for freeing this nvlist. | |
800 | */ | |
801 | int lzc_receive_one(const char *snapname, nvlist_t *props, | |
802 | const char *origin, boolean_t force, boolean_t resumable, int input_fd, | |
803 | const dmu_replay_record_t *begin_record, int cleanup_fd, | |
804 | uint64_t *read_bytes, uint64_t *errflags, uint64_t *action_handle, | |
805 | nvlist_t **errors) | |
806 | { | |
807 | return (recv_impl(snapname, props, origin, force, resumable, | |
808 | input_fd, begin_record, cleanup_fd, read_bytes, errflags, | |
809 | action_handle, errors)); | |
47dfff3b MA |
810 | } |
811 | ||
46ba1e59 MA |
812 | /* |
813 | * Roll back this filesystem or volume to its most recent snapshot. | |
814 | * If snapnamebuf is not NULL, it will be filled in with the name | |
815 | * of the most recent snapshot. | |
816 | * | |
817 | * Return 0 on success or an errno on failure. | |
818 | */ | |
819 | int | |
820 | lzc_rollback(const char *fsname, char *snapnamebuf, int snapnamelen) | |
821 | { | |
822 | nvlist_t *args; | |
823 | nvlist_t *result; | |
824 | int err; | |
825 | ||
826 | args = fnvlist_alloc(); | |
827 | err = lzc_ioctl(ZFS_IOC_ROLLBACK, fsname, args, &result); | |
828 | nvlist_free(args); | |
829 | if (err == 0 && snapnamebuf != NULL) { | |
830 | const char *snapname = fnvlist_lookup_string(result, "target"); | |
831 | (void) strlcpy(snapnamebuf, snapname, snapnamelen); | |
832 | } | |
bb7ffdaf GM |
833 | nvlist_free(result); |
834 | ||
46ba1e59 MA |
835 | return (err); |
836 | } | |
da536844 MA |
837 | |
838 | /* | |
839 | * Creates bookmarks. | |
840 | * | |
841 | * The bookmarks nvlist maps from name of the bookmark (e.g. "pool/fs#bmark") to | |
842 | * the name of the snapshot (e.g. "pool/fs@snap"). All the bookmarks and | |
843 | * snapshots must be in the same pool. | |
844 | * | |
845 | * The returned results nvlist will have an entry for each bookmark that failed. | |
846 | * The value will be the (int32) error code. | |
847 | * | |
848 | * The return value will be 0 if all bookmarks were created, otherwise it will | |
849 | * be the errno of a (undetermined) bookmarks that failed. | |
850 | */ | |
851 | int | |
852 | lzc_bookmark(nvlist_t *bookmarks, nvlist_t **errlist) | |
853 | { | |
854 | nvpair_t *elem; | |
855 | int error; | |
eca7b760 | 856 | char pool[ZFS_MAX_DATASET_NAME_LEN]; |
da536844 MA |
857 | |
858 | /* determine the pool name */ | |
859 | elem = nvlist_next_nvpair(bookmarks, NULL); | |
860 | if (elem == NULL) | |
861 | return (0); | |
862 | (void) strlcpy(pool, nvpair_name(elem), sizeof (pool)); | |
863 | pool[strcspn(pool, "/#")] = '\0'; | |
864 | ||
865 | error = lzc_ioctl(ZFS_IOC_BOOKMARK, pool, bookmarks, errlist); | |
866 | ||
867 | return (error); | |
868 | } | |
869 | ||
870 | /* | |
871 | * Retrieve bookmarks. | |
872 | * | |
873 | * Retrieve the list of bookmarks for the given file system. The props | |
874 | * parameter is an nvlist of property names (with no values) that will be | |
875 | * returned for each bookmark. | |
876 | * | |
877 | * The following are valid properties on bookmarks, all of which are numbers | |
878 | * (represented as uint64 in the nvlist) | |
879 | * | |
880 | * "guid" - globally unique identifier of the snapshot it refers to | |
881 | * "createtxg" - txg when the snapshot it refers to was created | |
882 | * "creation" - timestamp when the snapshot it refers to was created | |
883 | * | |
884 | * The format of the returned nvlist as follows: | |
885 | * <short name of bookmark> -> { | |
886 | * <name of property> -> { | |
887 | * "value" -> uint64 | |
888 | * } | |
889 | * } | |
890 | */ | |
891 | int | |
892 | lzc_get_bookmarks(const char *fsname, nvlist_t *props, nvlist_t **bmarks) | |
893 | { | |
894 | return (lzc_ioctl(ZFS_IOC_GET_BOOKMARKS, fsname, props, bmarks)); | |
895 | } | |
896 | ||
897 | /* | |
898 | * Destroys bookmarks. | |
899 | * | |
900 | * The keys in the bmarks nvlist are the bookmarks to be destroyed. | |
901 | * They must all be in the same pool. Bookmarks are specified as | |
902 | * <fs>#<bmark>. | |
903 | * | |
904 | * Bookmarks that do not exist will be silently ignored. | |
905 | * | |
906 | * The return value will be 0 if all bookmarks that existed were destroyed. | |
907 | * | |
908 | * Otherwise the return value will be the errno of a (undetermined) bookmark | |
909 | * that failed, no bookmarks will be destroyed, and the errlist will have an | |
910 | * entry for each bookmarks that failed. The value in the errlist will be | |
911 | * the (int32) error code. | |
912 | */ | |
913 | int | |
914 | lzc_destroy_bookmarks(nvlist_t *bmarks, nvlist_t **errlist) | |
915 | { | |
916 | nvpair_t *elem; | |
917 | int error; | |
eca7b760 | 918 | char pool[ZFS_MAX_DATASET_NAME_LEN]; |
da536844 MA |
919 | |
920 | /* determine the pool name */ | |
921 | elem = nvlist_next_nvpair(bmarks, NULL); | |
922 | if (elem == NULL) | |
923 | return (0); | |
924 | (void) strlcpy(pool, nvpair_name(elem), sizeof (pool)); | |
925 | pool[strcspn(pool, "/#")] = '\0'; | |
926 | ||
927 | error = lzc_ioctl(ZFS_IOC_DESTROY_BOOKMARKS, pool, bmarks, errlist); | |
928 | ||
929 | return (error); | |
930 | } |