]>
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 | |
1d3ba0bf | 9 | * or https://opensource.org/licenses/CDDL-1.0. |
6f1ffb06 MA |
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 | /* | |
196bee4c | 23 | * Copyright (c) 2012, 2020 by Delphix. All rights reserved. |
95fd54a1 | 24 | * Copyright (c) 2013 Steven Hartland. All rights reserved. |
d12f91fd | 25 | * Copyright 2017 RackTop Systems. |
d3f2cd7e | 26 | * Copyright (c) 2017 Open-E, Inc. All Rights Reserved. |
a73f361f | 27 | * Copyright (c) 2019, 2020 by Christian Schwarz. All rights reserved. |
e8cf3a4f | 28 | * Copyright (c) 2019 Datto Inc. |
6f1ffb06 MA |
29 | */ |
30 | ||
31 | /* | |
32 | * LibZFS_Core (lzc) is intended to replace most functionality in libzfs. | |
33 | * It has the following characteristics: | |
34 | * | |
35 | * - Thread Safe. libzfs_core is accessible concurrently from multiple | |
36 | * threads. This is accomplished primarily by avoiding global data | |
37 | * (e.g. caching). Since it's thread-safe, there is no reason for a | |
38 | * process to have multiple libzfs "instances". Therefore, we store | |
39 | * our few pieces of data (e.g. the file descriptor) in global | |
40 | * variables. The fd is reference-counted so that the libzfs_core | |
41 | * library can be "initialized" multiple times (e.g. by different | |
42 | * consumers within the same process). | |
43 | * | |
44 | * - Committed Interface. The libzfs_core interface will be committed, | |
45 | * therefore consumers can compile against it and be confident that | |
46 | * their code will continue to work on future releases of this code. | |
47 | * Currently, the interface is Evolving (not Committed), but we intend | |
48 | * to commit to it once it is more complete and we determine that it | |
49 | * meets the needs of all consumers. | |
50 | * | |
b8fce77b | 51 | * - Programmatic Error Handling. libzfs_core communicates errors with |
6f1ffb06 MA |
52 | * defined error numbers, and doesn't print anything to stdout/stderr. |
53 | * | |
54 | * - Thin Layer. libzfs_core is a thin layer, marshaling arguments | |
55 | * to/from the kernel ioctls. There is generally a 1:1 correspondence | |
fb0be12d | 56 | * between libzfs_core functions and ioctls to ZFS_DEV. |
6f1ffb06 MA |
57 | * |
58 | * - Clear Atomicity. Because libzfs_core functions are generally 1:1 | |
59 | * with kernel ioctls, and kernel ioctls are general atomic, each | |
60 | * libzfs_core function is atomic. For example, creating multiple | |
61 | * snapshots with a single call to lzc_snapshot() is atomic -- it | |
62 | * can't fail with only some of the requested snapshots created, even | |
63 | * in the event of power loss or system crash. | |
64 | * | |
65 | * - Continued libzfs Support. Some higher-level operations (e.g. | |
66 | * support for "zfs send -R") are too complicated to fit the scope of | |
67 | * libzfs_core. This functionality will continue to live in libzfs. | |
68 | * Where appropriate, libzfs will use the underlying atomic operations | |
69 | * of libzfs_core. For example, libzfs may implement "zfs send -R | | |
70 | * zfs receive" by using individual "send one snapshot", rename, | |
71 | * destroy, and "receive one snapshot" operations in libzfs_core. | |
f8b2ca6b | 72 | * /sbin/zfs and /sbin/zpool will link with both libzfs and |
6f1ffb06 MA |
73 | * libzfs_core. Other consumers should aim to use only libzfs_core, |
74 | * since that will be the supported, stable interface going forwards. | |
75 | */ | |
76 | ||
77 | #include <libzfs_core.h> | |
78 | #include <ctype.h> | |
79 | #include <unistd.h> | |
80 | #include <stdlib.h> | |
81 | #include <string.h> | |
b83a0e2d DB |
82 | #ifdef ZFS_DEBUG |
83 | #include <stdio.h> | |
84 | #endif | |
6f1ffb06 MA |
85 | #include <errno.h> |
86 | #include <fcntl.h> | |
87 | #include <pthread.h> | |
1f2f46be | 88 | #include <libzutil.h> |
6f1ffb06 MA |
89 | #include <sys/nvpair.h> |
90 | #include <sys/param.h> | |
91 | #include <sys/types.h> | |
92 | #include <sys/stat.h> | |
93 | #include <sys/zfs_ioctl.h> | |
fbbea09d AZ |
94 | #if __FreeBSD__ |
95 | #define BIG_PIPE_SIZE (64 * 1024) /* From sys/pipe.h */ | |
96 | #endif | |
6f1ffb06 | 97 | |
e2454897 | 98 | static int g_fd = -1; |
6f1ffb06 MA |
99 | static pthread_mutex_t g_lock = PTHREAD_MUTEX_INITIALIZER; |
100 | static int g_refcount; | |
101 | ||
b83a0e2d | 102 | #ifdef ZFS_DEBUG |
4f7fb135 | 103 | static zfs_ioc_t fail_ioc_cmd = ZFS_IOC_LAST; |
b83a0e2d DB |
104 | static zfs_errno_t fail_ioc_err; |
105 | ||
106 | static void | |
107 | libzfs_core_debug_ioc(void) | |
108 | { | |
109 | /* | |
110 | * To test running newer user space binaries with kernel's | |
111 | * that don't yet support an ioctl or a new ioctl arg we | |
112 | * provide an override to intentionally fail an ioctl. | |
113 | * | |
114 | * USAGE: | |
115 | * The override variable, ZFS_IOC_TEST, is of the form "cmd:err" | |
116 | * | |
117 | * For example, to fail a ZFS_IOC_POOL_CHECKPOINT with a | |
118 | * ZFS_ERR_IOC_CMD_UNAVAIL, the string would be "0x5a4d:1029" | |
119 | * | |
120 | * $ sudo sh -c "ZFS_IOC_TEST=0x5a4d:1029 zpool checkpoint tank" | |
121 | * cannot checkpoint 'tank': the loaded zfs module does not support | |
122 | * this operation. A reboot may be required to enable this operation. | |
123 | */ | |
4f7fb135 | 124 | if (fail_ioc_cmd == ZFS_IOC_LAST) { |
b83a0e2d DB |
125 | char *ioc_test = getenv("ZFS_IOC_TEST"); |
126 | unsigned int ioc_num = 0, ioc_err = 0; | |
127 | ||
128 | if (ioc_test != NULL && | |
129 | sscanf(ioc_test, "%i:%i", &ioc_num, &ioc_err) == 2 && | |
130 | ioc_num < ZFS_IOC_LAST) { | |
131 | fail_ioc_cmd = ioc_num; | |
132 | fail_ioc_err = ioc_err; | |
133 | } | |
134 | } | |
135 | } | |
136 | #endif | |
137 | ||
6f1ffb06 MA |
138 | int |
139 | libzfs_core_init(void) | |
140 | { | |
141 | (void) pthread_mutex_lock(&g_lock); | |
142 | if (g_refcount == 0) { | |
92ffd87a | 143 | g_fd = open(ZFS_DEV, O_RDWR|O_CLOEXEC); |
6f1ffb06 MA |
144 | if (g_fd < 0) { |
145 | (void) pthread_mutex_unlock(&g_lock); | |
146 | return (errno); | |
147 | } | |
148 | } | |
149 | g_refcount++; | |
b83a0e2d DB |
150 | |
151 | #ifdef ZFS_DEBUG | |
152 | libzfs_core_debug_ioc(); | |
153 | #endif | |
6f1ffb06 MA |
154 | (void) pthread_mutex_unlock(&g_lock); |
155 | return (0); | |
156 | } | |
157 | ||
158 | void | |
159 | libzfs_core_fini(void) | |
160 | { | |
161 | (void) pthread_mutex_lock(&g_lock); | |
162 | ASSERT3S(g_refcount, >, 0); | |
e2454897 | 163 | |
a0cb347c | 164 | g_refcount--; |
e2454897 GM |
165 | |
166 | if (g_refcount == 0 && g_fd != -1) { | |
6f1ffb06 | 167 | (void) close(g_fd); |
e2454897 GM |
168 | g_fd = -1; |
169 | } | |
6f1ffb06 MA |
170 | (void) pthread_mutex_unlock(&g_lock); |
171 | } | |
172 | ||
173 | static int | |
174 | lzc_ioctl(zfs_ioc_t ioc, const char *name, | |
175 | nvlist_t *source, nvlist_t **resultp) | |
176 | { | |
13fe0198 | 177 | zfs_cmd_t zc = {"\0"}; |
6f1ffb06 | 178 | int error = 0; |
bec1067d AP |
179 | char *packed = NULL; |
180 | size_t size = 0; | |
6f1ffb06 MA |
181 | |
182 | ASSERT3S(g_refcount, >, 0); | |
e2454897 | 183 | VERIFY3S(g_fd, !=, -1); |
6f1ffb06 | 184 | |
b83a0e2d DB |
185 | #ifdef ZFS_DEBUG |
186 | if (ioc == fail_ioc_cmd) | |
187 | return (fail_ioc_err); | |
188 | #endif | |
189 | ||
bec1067d AP |
190 | if (name != NULL) |
191 | (void) strlcpy(zc.zc_name, name, sizeof (zc.zc_name)); | |
6f1ffb06 | 192 | |
bec1067d AP |
193 | if (source != NULL) { |
194 | packed = fnvlist_pack(source, &size); | |
195 | zc.zc_nvlist_src = (uint64_t)(uintptr_t)packed; | |
196 | zc.zc_nvlist_src_size = size; | |
197 | } | |
6f1ffb06 MA |
198 | |
199 | if (resultp != NULL) { | |
13fe0198 | 200 | *resultp = NULL; |
234c91c5 CW |
201 | if (ioc == ZFS_IOC_CHANNEL_PROGRAM) { |
202 | zc.zc_nvlist_dst_size = fnvlist_lookup_uint64(source, | |
203 | ZCP_ARG_MEMLIMIT); | |
204 | } else { | |
205 | zc.zc_nvlist_dst_size = MAX(size * 2, 128 * 1024); | |
206 | } | |
6f1ffb06 MA |
207 | zc.zc_nvlist_dst = (uint64_t)(uintptr_t) |
208 | malloc(zc.zc_nvlist_dst_size); | |
209 | if (zc.zc_nvlist_dst == (uint64_t)0) { | |
210 | error = ENOMEM; | |
211 | goto out; | |
212 | } | |
213 | } | |
214 | ||
514498fe | 215 | while (lzc_ioctl_fd(g_fd, ioc, &zc) != 0) { |
d99a0153 CW |
216 | /* |
217 | * If ioctl exited with ENOMEM, we retry the ioctl after | |
218 | * increasing the size of the destination nvlist. | |
219 | * | |
234c91c5 | 220 | * Channel programs that exit with ENOMEM ran over the |
d99a0153 CW |
221 | * lua memory sandbox; they should not be retried. |
222 | */ | |
223 | if (errno == ENOMEM && resultp != NULL && | |
224 | ioc != ZFS_IOC_CHANNEL_PROGRAM) { | |
6f1ffb06 MA |
225 | free((void *)(uintptr_t)zc.zc_nvlist_dst); |
226 | zc.zc_nvlist_dst_size *= 2; | |
227 | zc.zc_nvlist_dst = (uint64_t)(uintptr_t) | |
228 | malloc(zc.zc_nvlist_dst_size); | |
229 | if (zc.zc_nvlist_dst == (uint64_t)0) { | |
230 | error = ENOMEM; | |
231 | goto out; | |
232 | } | |
233 | } else { | |
234 | error = errno; | |
235 | break; | |
236 | } | |
237 | } | |
238 | if (zc.zc_nvlist_dst_filled) { | |
239 | *resultp = fnvlist_unpack((void *)(uintptr_t)zc.zc_nvlist_dst, | |
240 | zc.zc_nvlist_dst_size); | |
6f1ffb06 MA |
241 | } |
242 | ||
243 | out: | |
b5256303 TC |
244 | if (packed != NULL) |
245 | fnvlist_pack_free(packed, size); | |
6f1ffb06 MA |
246 | free((void *)(uintptr_t)zc.zc_nvlist_dst); |
247 | return (error); | |
248 | } | |
249 | ||
250 | int | |
b5256303 TC |
251 | lzc_create(const char *fsname, enum lzc_dataset_type type, nvlist_t *props, |
252 | uint8_t *wkeydata, uint_t wkeylen) | |
6f1ffb06 MA |
253 | { |
254 | int error; | |
b5256303 | 255 | nvlist_t *hidden_args = NULL; |
6f1ffb06 | 256 | nvlist_t *args = fnvlist_alloc(); |
b5256303 | 257 | |
e67a7ffb | 258 | fnvlist_add_int32(args, "type", (dmu_objset_type_t)type); |
6f1ffb06 MA |
259 | if (props != NULL) |
260 | fnvlist_add_nvlist(args, "props", props); | |
b5256303 TC |
261 | |
262 | if (wkeydata != NULL) { | |
263 | hidden_args = fnvlist_alloc(); | |
264 | fnvlist_add_uint8_array(hidden_args, "wkeydata", wkeydata, | |
265 | wkeylen); | |
266 | fnvlist_add_nvlist(args, ZPOOL_HIDDEN_ARGS, hidden_args); | |
267 | } | |
268 | ||
6f1ffb06 | 269 | error = lzc_ioctl(ZFS_IOC_CREATE, fsname, args, NULL); |
b5256303 | 270 | nvlist_free(hidden_args); |
6f1ffb06 MA |
271 | nvlist_free(args); |
272 | return (error); | |
273 | } | |
274 | ||
275 | int | |
b5256303 | 276 | lzc_clone(const char *fsname, const char *origin, nvlist_t *props) |
6f1ffb06 MA |
277 | { |
278 | int error; | |
b5256303 | 279 | nvlist_t *hidden_args = NULL; |
6f1ffb06 | 280 | nvlist_t *args = fnvlist_alloc(); |
b5256303 | 281 | |
6f1ffb06 MA |
282 | fnvlist_add_string(args, "origin", origin); |
283 | if (props != NULL) | |
284 | fnvlist_add_nvlist(args, "props", props); | |
285 | error = lzc_ioctl(ZFS_IOC_CLONE, fsname, args, NULL); | |
b5256303 | 286 | nvlist_free(hidden_args); |
6f1ffb06 MA |
287 | nvlist_free(args); |
288 | return (error); | |
289 | } | |
290 | ||
d12f91fd GDN |
291 | int |
292 | lzc_promote(const char *fsname, char *snapnamebuf, int snapnamelen) | |
293 | { | |
294 | /* | |
295 | * The promote ioctl is still legacy, so we need to construct our | |
296 | * own zfs_cmd_t rather than using lzc_ioctl(). | |
297 | */ | |
659f4008 | 298 | zfs_cmd_t zc = {"\0"}; |
d12f91fd GDN |
299 | |
300 | ASSERT3S(g_refcount, >, 0); | |
301 | VERIFY3S(g_fd, !=, -1); | |
302 | ||
303 | (void) strlcpy(zc.zc_name, fsname, sizeof (zc.zc_name)); | |
514498fe | 304 | if (lzc_ioctl_fd(g_fd, ZFS_IOC_PROMOTE, &zc) != 0) { |
d12f91fd GDN |
305 | int error = errno; |
306 | if (error == EEXIST && snapnamebuf != NULL) | |
307 | (void) strlcpy(snapnamebuf, zc.zc_string, snapnamelen); | |
308 | return (error); | |
309 | } | |
310 | return (0); | |
311 | } | |
312 | ||
dc1c630b AG |
313 | int |
314 | lzc_rename(const char *source, const char *target) | |
315 | { | |
659f4008 | 316 | zfs_cmd_t zc = {"\0"}; |
dc1c630b | 317 | int error; |
659f4008 | 318 | |
dc1c630b AG |
319 | ASSERT3S(g_refcount, >, 0); |
320 | VERIFY3S(g_fd, !=, -1); | |
321 | (void) strlcpy(zc.zc_name, source, sizeof (zc.zc_name)); | |
322 | (void) strlcpy(zc.zc_value, target, sizeof (zc.zc_value)); | |
514498fe | 323 | error = lzc_ioctl_fd(g_fd, ZFS_IOC_RENAME, &zc); |
dc1c630b AG |
324 | if (error != 0) |
325 | error = errno; | |
326 | return (error); | |
327 | } | |
e1af0d0d | 328 | |
dc1c630b AG |
329 | int |
330 | lzc_destroy(const char *fsname) | |
331 | { | |
332 | int error; | |
333 | nvlist_t *args = fnvlist_alloc(); | |
334 | error = lzc_ioctl(ZFS_IOC_DESTROY, fsname, args, NULL); | |
335 | nvlist_free(args); | |
336 | return (error); | |
337 | } | |
338 | ||
6f1ffb06 MA |
339 | /* |
340 | * Creates snapshots. | |
341 | * | |
342 | * The keys in the snaps nvlist are the snapshots to be created. | |
343 | * They must all be in the same pool. | |
344 | * | |
345 | * The props nvlist is properties to set. Currently only user properties | |
346 | * are supported. { user:prop_name -> string value } | |
347 | * | |
348 | * The returned results nvlist will have an entry for each snapshot that failed. | |
349 | * The value will be the (int32) error code. | |
350 | * | |
351 | * The return value will be 0 if all snapshots were created, otherwise it will | |
13fe0198 | 352 | * be the errno of a (unspecified) snapshot that failed. |
6f1ffb06 MA |
353 | */ |
354 | int | |
355 | lzc_snapshot(nvlist_t *snaps, nvlist_t *props, nvlist_t **errlist) | |
356 | { | |
357 | nvpair_t *elem; | |
358 | nvlist_t *args; | |
359 | int error; | |
eca7b760 | 360 | char pool[ZFS_MAX_DATASET_NAME_LEN]; |
6f1ffb06 MA |
361 | |
362 | *errlist = NULL; | |
363 | ||
364 | /* determine the pool name */ | |
365 | elem = nvlist_next_nvpair(snaps, NULL); | |
366 | if (elem == NULL) | |
367 | return (0); | |
368 | (void) strlcpy(pool, nvpair_name(elem), sizeof (pool)); | |
369 | pool[strcspn(pool, "/@")] = '\0'; | |
370 | ||
371 | args = fnvlist_alloc(); | |
372 | fnvlist_add_nvlist(args, "snaps", snaps); | |
373 | if (props != NULL) | |
374 | fnvlist_add_nvlist(args, "props", props); | |
375 | ||
376 | error = lzc_ioctl(ZFS_IOC_SNAPSHOT, pool, args, errlist); | |
377 | nvlist_free(args); | |
378 | ||
379 | return (error); | |
380 | } | |
381 | ||
382 | /* | |
383 | * Destroys snapshots. | |
384 | * | |
385 | * The keys in the snaps nvlist are the snapshots to be destroyed. | |
386 | * They must all be in the same pool. | |
387 | * | |
388 | * Snapshots that do not exist will be silently ignored. | |
389 | * | |
390 | * If 'defer' is not set, and a snapshot has user holds or clones, the | |
391 | * destroy operation will fail and none of the snapshots will be | |
392 | * destroyed. | |
393 | * | |
394 | * If 'defer' is set, and a snapshot has user holds or clones, it will be | |
395 | * marked for deferred destruction, and will be destroyed when the last hold | |
396 | * or clone is removed/destroyed. | |
397 | * | |
398 | * The return value will be 0 if all snapshots were destroyed (or marked for | |
1a077756 | 399 | * later destruction if 'defer' is set) or didn't exist to begin with. |
6f1ffb06 | 400 | * |
13fe0198 | 401 | * Otherwise the return value will be the errno of a (unspecified) snapshot |
6f1ffb06 MA |
402 | * that failed, no snapshots will be destroyed, and the errlist will have an |
403 | * entry for each snapshot that failed. The value in the errlist will be | |
404 | * the (int32) error code. | |
405 | */ | |
406 | int | |
407 | lzc_destroy_snaps(nvlist_t *snaps, boolean_t defer, nvlist_t **errlist) | |
408 | { | |
409 | nvpair_t *elem; | |
410 | nvlist_t *args; | |
411 | int error; | |
eca7b760 | 412 | char pool[ZFS_MAX_DATASET_NAME_LEN]; |
6f1ffb06 MA |
413 | |
414 | /* determine the pool name */ | |
415 | elem = nvlist_next_nvpair(snaps, NULL); | |
416 | if (elem == NULL) | |
417 | return (0); | |
418 | (void) strlcpy(pool, nvpair_name(elem), sizeof (pool)); | |
419 | pool[strcspn(pool, "/@")] = '\0'; | |
420 | ||
421 | args = fnvlist_alloc(); | |
422 | fnvlist_add_nvlist(args, "snaps", snaps); | |
423 | if (defer) | |
424 | fnvlist_add_boolean(args, "defer"); | |
425 | ||
426 | error = lzc_ioctl(ZFS_IOC_DESTROY_SNAPS, pool, args, errlist); | |
427 | nvlist_free(args); | |
428 | ||
429 | return (error); | |
6f1ffb06 MA |
430 | } |
431 | ||
432 | int | |
433 | lzc_snaprange_space(const char *firstsnap, const char *lastsnap, | |
434 | uint64_t *usedp) | |
435 | { | |
436 | nvlist_t *args; | |
437 | nvlist_t *result; | |
438 | int err; | |
eca7b760 | 439 | char fs[ZFS_MAX_DATASET_NAME_LEN]; |
6f1ffb06 MA |
440 | char *atp; |
441 | ||
442 | /* determine the fs name */ | |
443 | (void) strlcpy(fs, firstsnap, sizeof (fs)); | |
444 | atp = strchr(fs, '@'); | |
445 | if (atp == NULL) | |
446 | return (EINVAL); | |
447 | *atp = '\0'; | |
448 | ||
449 | args = fnvlist_alloc(); | |
450 | fnvlist_add_string(args, "firstsnap", firstsnap); | |
451 | ||
452 | err = lzc_ioctl(ZFS_IOC_SPACE_SNAPS, lastsnap, args, &result); | |
453 | nvlist_free(args); | |
454 | if (err == 0) | |
455 | *usedp = fnvlist_lookup_uint64(result, "used"); | |
456 | fnvlist_free(result); | |
457 | ||
458 | return (err); | |
459 | } | |
460 | ||
461 | boolean_t | |
462 | lzc_exists(const char *dataset) | |
463 | { | |
464 | /* | |
465 | * The objset_stats ioctl is still legacy, so we need to construct our | |
d12f91fd | 466 | * own zfs_cmd_t rather than using lzc_ioctl(). |
6f1ffb06 | 467 | */ |
13fe0198 | 468 | zfs_cmd_t zc = {"\0"}; |
6f1ffb06 | 469 | |
e2454897 GM |
470 | ASSERT3S(g_refcount, >, 0); |
471 | VERIFY3S(g_fd, !=, -1); | |
472 | ||
6f1ffb06 | 473 | (void) strlcpy(zc.zc_name, dataset, sizeof (zc.zc_name)); |
514498fe | 474 | return (lzc_ioctl_fd(g_fd, ZFS_IOC_OBJSET_STATS, &zc) == 0); |
6f1ffb06 MA |
475 | } |
476 | ||
bec1067d AP |
477 | /* |
478 | * outnvl is unused. | |
479 | * It was added to preserve the function signature in case it is | |
480 | * needed in the future. | |
481 | */ | |
bec1067d AP |
482 | int |
483 | lzc_sync(const char *pool_name, nvlist_t *innvl, nvlist_t **outnvl) | |
484 | { | |
bfc17897 | 485 | (void) outnvl; |
bec1067d AP |
486 | return (lzc_ioctl(ZFS_IOC_POOL_SYNC, pool_name, innvl, NULL)); |
487 | } | |
488 | ||
13fe0198 MA |
489 | /* |
490 | * Create "user holds" on snapshots. If there is a hold on a snapshot, | |
491 | * the snapshot can not be destroyed. (However, it can be marked for deletion | |
492 | * by lzc_destroy_snaps(defer=B_TRUE).) | |
493 | * | |
494 | * The keys in the nvlist are snapshot names. | |
495 | * The snapshots must all be in the same pool. | |
496 | * The value is the name of the hold (string type). | |
497 | * | |
fb0be12d | 498 | * If cleanup_fd is not -1, it must be the result of open(ZFS_DEV, O_EXCL). |
13fe0198 MA |
499 | * In this case, when the cleanup_fd is closed (including on process |
500 | * termination), the holds will be released. If the system is shut down | |
501 | * uncleanly, the holds will be released when the pool is next opened | |
502 | * or imported. | |
503 | * | |
95fd54a1 | 504 | * Holds for snapshots which don't exist will be skipped and have an entry |
1a077756 | 505 | * added to errlist, but will not cause an overall failure. |
95fd54a1 | 506 | * |
1a077756 | 507 | * The return value will be 0 if all holds, for snapshots that existed, |
b8fce77b | 508 | * were successfully created. |
95fd54a1 SH |
509 | * |
510 | * Otherwise the return value will be the errno of a (unspecified) hold that | |
511 | * failed and no holds will be created. | |
512 | * | |
513 | * In all cases the errlist will have an entry for each hold that failed | |
514 | * (name = snapshot), with its value being the error code (int32). | |
13fe0198 MA |
515 | */ |
516 | int | |
517 | lzc_hold(nvlist_t *holds, int cleanup_fd, nvlist_t **errlist) | |
518 | { | |
eca7b760 | 519 | char pool[ZFS_MAX_DATASET_NAME_LEN]; |
13fe0198 MA |
520 | nvlist_t *args; |
521 | nvpair_t *elem; | |
522 | int error; | |
523 | ||
524 | /* determine the pool name */ | |
525 | elem = nvlist_next_nvpair(holds, NULL); | |
526 | if (elem == NULL) | |
527 | return (0); | |
528 | (void) strlcpy(pool, nvpair_name(elem), sizeof (pool)); | |
529 | pool[strcspn(pool, "/@")] = '\0'; | |
530 | ||
531 | args = fnvlist_alloc(); | |
532 | fnvlist_add_nvlist(args, "holds", holds); | |
533 | if (cleanup_fd != -1) | |
534 | fnvlist_add_int32(args, "cleanup_fd", cleanup_fd); | |
535 | ||
536 | error = lzc_ioctl(ZFS_IOC_HOLD, pool, args, errlist); | |
537 | nvlist_free(args); | |
538 | return (error); | |
539 | } | |
540 | ||
541 | /* | |
542 | * Release "user holds" on snapshots. If the snapshot has been marked for | |
543 | * deferred destroy (by lzc_destroy_snaps(defer=B_TRUE)), it does not have | |
544 | * any clones, and all the user holds are removed, then the snapshot will be | |
545 | * destroyed. | |
546 | * | |
547 | * The keys in the nvlist are snapshot names. | |
548 | * The snapshots must all be in the same pool. | |
d5884c34 | 549 | * The value is an nvlist whose keys are the holds to remove. |
13fe0198 | 550 | * |
95fd54a1 | 551 | * Holds which failed to release because they didn't exist will have an entry |
1a077756 | 552 | * added to errlist, but will not cause an overall failure. |
95fd54a1 SH |
553 | * |
554 | * The return value will be 0 if the nvl holds was empty or all holds that | |
1a077756 | 555 | * existed, were successfully removed. |
95fd54a1 SH |
556 | * |
557 | * Otherwise the return value will be the errno of a (unspecified) hold that | |
558 | * failed to release and no holds will be released. | |
559 | * | |
560 | * In all cases the errlist will have an entry for each hold that failed to | |
561 | * to release. | |
13fe0198 MA |
562 | */ |
563 | int | |
564 | lzc_release(nvlist_t *holds, nvlist_t **errlist) | |
565 | { | |
eca7b760 | 566 | char pool[ZFS_MAX_DATASET_NAME_LEN]; |
13fe0198 MA |
567 | nvpair_t *elem; |
568 | ||
569 | /* determine the pool name */ | |
570 | elem = nvlist_next_nvpair(holds, NULL); | |
571 | if (elem == NULL) | |
572 | return (0); | |
573 | (void) strlcpy(pool, nvpair_name(elem), sizeof (pool)); | |
574 | pool[strcspn(pool, "/@")] = '\0'; | |
575 | ||
576 | return (lzc_ioctl(ZFS_IOC_RELEASE, pool, holds, errlist)); | |
577 | } | |
578 | ||
579 | /* | |
580 | * Retrieve list of user holds on the specified snapshot. | |
581 | * | |
d5884c34 | 582 | * On success, *holdsp will be set to an nvlist which the caller must free. |
13fe0198 MA |
583 | * The keys are the names of the holds, and the value is the creation time |
584 | * of the hold (uint64) in seconds since the epoch. | |
585 | */ | |
586 | int | |
587 | lzc_get_holds(const char *snapname, nvlist_t **holdsp) | |
588 | { | |
bec1067d | 589 | return (lzc_ioctl(ZFS_IOC_GET_HOLDS, snapname, NULL, holdsp)); |
13fe0198 MA |
590 | } |
591 | ||
fbbea09d | 592 | static unsigned int |
8a3d7735 AZ |
593 | max_pipe_buffer(int infd) |
594 | { | |
595 | #if __linux__ | |
fbbea09d AZ |
596 | static unsigned int max; |
597 | if (max == 0) { | |
598 | max = 1048576; /* fs/pipe.c default */ | |
599 | ||
600 | FILE *procf = fopen("/proc/sys/fs/pipe-max-size", "re"); | |
601 | if (procf != NULL) { | |
602 | if (fscanf(procf, "%u", &max) <= 0) { | |
603 | /* ignore error: max untouched if parse fails */ | |
604 | } | |
605 | fclose(procf); | |
8a3d7735 | 606 | } |
8a3d7735 | 607 | } |
fbbea09d AZ |
608 | |
609 | unsigned int cur = fcntl(infd, F_GETPIPE_SZ); | |
a30927f7 RE |
610 | /* |
611 | * Sadly, Linux has an unfixed deadlock if you do SETPIPE_SZ on a pipe | |
612 | * with data in it. | |
613 | * cf. #13232, https://bugzilla.kernel.org/show_bug.cgi?id=212295 | |
614 | * | |
615 | * And since the problem is in waking up the writer, there's nothing | |
616 | * we can do about it from here. | |
617 | * | |
618 | * So if people want to, they can set this, but they | |
619 | * may regret it... | |
620 | */ | |
621 | if (getenv("ZFS_SET_PIPE_MAX") == NULL) | |
622 | return (cur); | |
fbbea09d AZ |
623 | if (cur < max && fcntl(infd, F_SETPIPE_SZ, max) != -1) |
624 | cur = max; | |
625 | return (cur); | |
8a3d7735 AZ |
626 | #else |
627 | /* FreeBSD automatically resizes */ | |
628 | (void) infd; | |
fbbea09d | 629 | return (BIG_PIPE_SIZE); |
8a3d7735 AZ |
630 | #endif |
631 | } | |
632 | ||
3a909fe3 AZ |
633 | #if __linux__ |
634 | struct send_worker_ctx { | |
635 | int from; /* read end of pipe, with send data; closed on exit */ | |
636 | int to; /* original arbitrary output fd; mustn't be a pipe */ | |
637 | }; | |
638 | ||
639 | static void * | |
640 | send_worker(void *arg) | |
641 | { | |
642 | struct send_worker_ctx *ctx = arg; | |
643 | unsigned int bufsiz = max_pipe_buffer(ctx->from); | |
644 | ssize_t rd; | |
645 | ||
646 | while ((rd = splice(ctx->from, NULL, ctx->to, NULL, bufsiz, | |
647 | SPLICE_F_MOVE | SPLICE_F_MORE)) > 0) | |
648 | ; | |
649 | ||
650 | int err = (rd == -1) ? errno : 0; | |
651 | close(ctx->from); | |
652 | return ((void *)(uintptr_t)err); | |
653 | } | |
654 | #endif | |
655 | ||
656 | /* | |
657 | * Since Linux 5.10, 4d03e3cc59828c82ee89ea6e27a2f3cdf95aaadf | |
658 | * ("fs: don't allow kernel reads and writes without iter ops"), | |
659 | * ZFS_IOC_SEND* will EINVAL when writing to /dev/null, /dev/zero, &c. | |
660 | * | |
661 | * This wrapper transparently executes func() with a pipe | |
662 | * by spawning a thread to copy from that pipe to the original output | |
663 | * in the background. | |
664 | * | |
665 | * Returns the error from func(), if nonzero, | |
666 | * otherwise the error from the thread. | |
667 | * | |
a86e0894 AZ |
668 | * No-op if orig_fd is -1, already a pipe (but the buffer size is bumped), |
669 | * and on not-Linux; as such, it is safe to wrap/call wrapped functions | |
670 | * in a wrapped context. | |
3a909fe3 AZ |
671 | */ |
672 | int | |
673 | lzc_send_wrapper(int (*func)(int, void *), int orig_fd, void *data) | |
674 | { | |
675 | #if __linux__ | |
676 | struct stat sb; | |
677 | if (orig_fd != -1 && fstat(orig_fd, &sb) == -1) | |
678 | return (errno); | |
a86e0894 AZ |
679 | if (orig_fd == -1 || S_ISFIFO(sb.st_mode)) { |
680 | if (orig_fd != -1) | |
681 | (void) max_pipe_buffer(orig_fd); | |
3a909fe3 | 682 | return (func(orig_fd, data)); |
a86e0894 | 683 | } |
3a909fe3 AZ |
684 | if ((fcntl(orig_fd, F_GETFL) & O_ACCMODE) == O_RDONLY) |
685 | return (errno = EBADF); | |
686 | ||
687 | int rw[2]; | |
688 | if (pipe2(rw, O_CLOEXEC) == -1) | |
689 | return (errno); | |
690 | ||
691 | int err; | |
692 | pthread_t send_thread; | |
693 | struct send_worker_ctx ctx = {.from = rw[0], .to = orig_fd}; | |
694 | if ((err = pthread_create(&send_thread, NULL, send_worker, &ctx)) | |
695 | != 0) { | |
696 | close(rw[0]); | |
697 | close(rw[1]); | |
698 | return (errno = err); | |
699 | } | |
700 | ||
701 | err = func(rw[1], data); | |
702 | ||
703 | void *send_err; | |
704 | close(rw[1]); | |
705 | pthread_join(send_thread, &send_err); | |
706 | if (err == 0 && send_err != 0) | |
707 | errno = err = (uintptr_t)send_err; | |
708 | ||
709 | return (err); | |
710 | #else | |
711 | return (func(orig_fd, data)); | |
712 | #endif | |
713 | } | |
714 | ||
6f1ffb06 | 715 | /* |
9b67f605 MA |
716 | * Generate a zfs send stream for the specified snapshot and write it to |
717 | * the specified file descriptor. | |
da536844 MA |
718 | * |
719 | * "snapname" is the full name of the snapshot to send (e.g. "pool/fs@snap") | |
720 | * | |
721 | * If "from" is NULL, a full (non-incremental) stream will be sent. | |
722 | * If "from" is non-NULL, it must be the full name of a snapshot or | |
723 | * bookmark to send an incremental from (e.g. "pool/fs@earlier_snap" or | |
724 | * "pool/fs#earlier_bmark"). If non-NULL, the specified snapshot or | |
725 | * bookmark must represent an earlier point in the history of "snapname"). | |
726 | * It can be an earlier snapshot in the same filesystem or zvol as "snapname", | |
727 | * or it can be the origin of "snapname"'s filesystem, or an earlier | |
728 | * snapshot in the origin, etc. | |
729 | * | |
730 | * "fd" is the file descriptor to write the send stream to. | |
9b67f605 | 731 | * |
f1512ee6 MA |
732 | * If "flags" contains LZC_SEND_FLAG_LARGE_BLOCK, the stream is permitted |
733 | * to contain DRR_WRITE records with drr_length > 128K, and DRR_OBJECT | |
734 | * records with drr_blksz > 128K. | |
735 | * | |
9b67f605 MA |
736 | * If "flags" contains LZC_SEND_FLAG_EMBED_DATA, the stream is permitted |
737 | * to contain DRR_WRITE_EMBEDDED records with drr_etype==BP_EMBEDDED_TYPE_DATA, | |
738 | * which the receiving system must support (as indicated by support | |
739 | * for the "embedded_data" feature). | |
85ce3f4f | 740 | * |
741 | * If "flags" contains LZC_SEND_FLAG_COMPRESS, the stream is generated by using | |
742 | * compressed WRITE records for blocks which are compressed on disk and in | |
743 | * memory. If the lz4_compress feature is active on the sending system, then | |
744 | * the receiving system must have that feature enabled as well. | |
745 | * | |
746 | * If "flags" contains LZC_SEND_FLAG_RAW, the stream is generated, for encrypted | |
747 | * datasets, by sending data exactly as it exists on disk. This allows backups | |
748 | * to be taken even if encryption keys are not currently loaded. | |
6f1ffb06 MA |
749 | */ |
750 | int | |
9b67f605 MA |
751 | lzc_send(const char *snapname, const char *from, int fd, |
752 | enum lzc_send_flags flags) | |
47dfff3b | 753 | { |
30af21b0 PD |
754 | return (lzc_send_resume_redacted(snapname, from, fd, flags, 0, 0, |
755 | NULL)); | |
756 | } | |
757 | ||
758 | int | |
759 | lzc_send_redacted(const char *snapname, const char *from, int fd, | |
760 | enum lzc_send_flags flags, const char *redactbook) | |
761 | { | |
762 | return (lzc_send_resume_redacted(snapname, from, fd, flags, 0, 0, | |
763 | redactbook)); | |
47dfff3b MA |
764 | } |
765 | ||
766 | int | |
767 | lzc_send_resume(const char *snapname, const char *from, int fd, | |
768 | enum lzc_send_flags flags, uint64_t resumeobj, uint64_t resumeoff) | |
30af21b0 PD |
769 | { |
770 | return (lzc_send_resume_redacted(snapname, from, fd, flags, resumeobj, | |
771 | resumeoff, NULL)); | |
772 | } | |
773 | ||
774 | /* | |
775 | * snapname: The name of the "tosnap", or the snapshot whose contents we are | |
776 | * sending. | |
777 | * from: The name of the "fromsnap", or the incremental source. | |
778 | * fd: File descriptor to write the stream to. | |
779 | * flags: flags that determine features to be used by the stream. | |
780 | * resumeobj: Object to resume from, for resuming send | |
781 | * resumeoff: Offset to resume from, for resuming send. | |
782 | * redactnv: nvlist of string -> boolean(ignored) containing the names of all | |
783 | * the snapshots that we should redact with respect to. | |
784 | * redactbook: Name of the redaction bookmark to create. | |
3a909fe3 AZ |
785 | * |
786 | * Pre-wrapped. | |
30af21b0 | 787 | */ |
3a909fe3 AZ |
788 | static int |
789 | lzc_send_resume_redacted_cb_impl(const char *snapname, const char *from, int fd, | |
30af21b0 PD |
790 | enum lzc_send_flags flags, uint64_t resumeobj, uint64_t resumeoff, |
791 | const char *redactbook) | |
6f1ffb06 MA |
792 | { |
793 | nvlist_t *args; | |
794 | int err; | |
795 | ||
796 | args = fnvlist_alloc(); | |
797 | fnvlist_add_int32(args, "fd", fd); | |
da536844 MA |
798 | if (from != NULL) |
799 | fnvlist_add_string(args, "fromsnap", from); | |
f1512ee6 MA |
800 | if (flags & LZC_SEND_FLAG_LARGE_BLOCK) |
801 | fnvlist_add_boolean(args, "largeblockok"); | |
9b67f605 MA |
802 | if (flags & LZC_SEND_FLAG_EMBED_DATA) |
803 | fnvlist_add_boolean(args, "embedok"); | |
a7004725 DK |
804 | if (flags & LZC_SEND_FLAG_COMPRESS) |
805 | fnvlist_add_boolean(args, "compressok"); | |
b5256303 TC |
806 | if (flags & LZC_SEND_FLAG_RAW) |
807 | fnvlist_add_boolean(args, "rawok"); | |
ba0ba69e TC |
808 | if (flags & LZC_SEND_FLAG_SAVED) |
809 | fnvlist_add_boolean(args, "savedok"); | |
47dfff3b MA |
810 | if (resumeobj != 0 || resumeoff != 0) { |
811 | fnvlist_add_uint64(args, "resume_object", resumeobj); | |
812 | fnvlist_add_uint64(args, "resume_offset", resumeoff); | |
813 | } | |
30af21b0 PD |
814 | if (redactbook != NULL) |
815 | fnvlist_add_string(args, "redactbook", redactbook); | |
816 | ||
6f1ffb06 MA |
817 | err = lzc_ioctl(ZFS_IOC_SEND_NEW, snapname, args, NULL); |
818 | nvlist_free(args); | |
819 | return (err); | |
820 | } | |
821 | ||
3a909fe3 AZ |
822 | struct lzc_send_resume_redacted { |
823 | const char *snapname; | |
824 | const char *from; | |
825 | enum lzc_send_flags flags; | |
826 | uint64_t resumeobj; | |
827 | uint64_t resumeoff; | |
828 | const char *redactbook; | |
829 | }; | |
830 | ||
831 | static int | |
832 | lzc_send_resume_redacted_cb(int fd, void *arg) | |
833 | { | |
834 | struct lzc_send_resume_redacted *zsrr = arg; | |
835 | return (lzc_send_resume_redacted_cb_impl(zsrr->snapname, zsrr->from, | |
836 | fd, zsrr->flags, zsrr->resumeobj, zsrr->resumeoff, | |
837 | zsrr->redactbook)); | |
838 | } | |
839 | ||
840 | int | |
841 | lzc_send_resume_redacted(const char *snapname, const char *from, int fd, | |
842 | enum lzc_send_flags flags, uint64_t resumeobj, uint64_t resumeoff, | |
843 | const char *redactbook) | |
844 | { | |
845 | struct lzc_send_resume_redacted zsrr = { | |
846 | .snapname = snapname, | |
847 | .from = from, | |
848 | .flags = flags, | |
849 | .resumeobj = resumeobj, | |
850 | .resumeoff = resumeoff, | |
851 | .redactbook = redactbook, | |
852 | }; | |
853 | return (lzc_send_wrapper(lzc_send_resume_redacted_cb, fd, &zsrr)); | |
854 | } | |
855 | ||
6f1ffb06 | 856 | /* |
5dc8b736 MG |
857 | * "from" can be NULL, a snapshot, or a bookmark. |
858 | * | |
859 | * If from is NULL, a full (non-incremental) stream will be estimated. This | |
860 | * is calculated very efficiently. | |
861 | * | |
862 | * If from is a snapshot, lzc_send_space uses the deadlists attached to | |
863 | * each snapshot to efficiently estimate the stream size. | |
864 | * | |
865 | * If from is a bookmark, the indirect blocks in the destination snapshot | |
866 | * are traversed, looking for blocks with a birth time since the creation TXG of | |
867 | * the snapshot this bookmark was created from. This will result in | |
868 | * significantly more I/O and be less efficient than a send space estimation on | |
30af21b0 PD |
869 | * an equivalent snapshot. This process is also used if redact_snaps is |
870 | * non-null. | |
3a909fe3 AZ |
871 | * |
872 | * Pre-wrapped. | |
6f1ffb06 | 873 | */ |
3a909fe3 AZ |
874 | static int |
875 | lzc_send_space_resume_redacted_cb_impl(const char *snapname, const char *from, | |
30af21b0 PD |
876 | enum lzc_send_flags flags, uint64_t resumeobj, uint64_t resumeoff, |
877 | uint64_t resume_bytes, const char *redactbook, int fd, uint64_t *spacep) | |
6f1ffb06 MA |
878 | { |
879 | nvlist_t *args; | |
880 | nvlist_t *result; | |
881 | int err; | |
882 | ||
883 | args = fnvlist_alloc(); | |
5dc8b736 MG |
884 | if (from != NULL) |
885 | fnvlist_add_string(args, "from", from); | |
2aa34383 DK |
886 | if (flags & LZC_SEND_FLAG_LARGE_BLOCK) |
887 | fnvlist_add_boolean(args, "largeblockok"); | |
888 | if (flags & LZC_SEND_FLAG_EMBED_DATA) | |
889 | fnvlist_add_boolean(args, "embedok"); | |
890 | if (flags & LZC_SEND_FLAG_COMPRESS) | |
891 | fnvlist_add_boolean(args, "compressok"); | |
cf7684bc | 892 | if (flags & LZC_SEND_FLAG_RAW) |
893 | fnvlist_add_boolean(args, "rawok"); | |
30af21b0 PD |
894 | if (resumeobj != 0 || resumeoff != 0) { |
895 | fnvlist_add_uint64(args, "resume_object", resumeobj); | |
896 | fnvlist_add_uint64(args, "resume_offset", resumeoff); | |
897 | fnvlist_add_uint64(args, "bytes", resume_bytes); | |
898 | } | |
899 | if (redactbook != NULL) | |
900 | fnvlist_add_string(args, "redactbook", redactbook); | |
901 | if (fd != -1) | |
902 | fnvlist_add_int32(args, "fd", fd); | |
903 | ||
6f1ffb06 MA |
904 | err = lzc_ioctl(ZFS_IOC_SEND_SPACE, snapname, args, &result); |
905 | nvlist_free(args); | |
906 | if (err == 0) | |
907 | *spacep = fnvlist_lookup_uint64(result, "space"); | |
908 | nvlist_free(result); | |
909 | return (err); | |
910 | } | |
911 | ||
3a909fe3 AZ |
912 | struct lzc_send_space_resume_redacted { |
913 | const char *snapname; | |
914 | const char *from; | |
915 | enum lzc_send_flags flags; | |
916 | uint64_t resumeobj; | |
917 | uint64_t resumeoff; | |
918 | uint64_t resume_bytes; | |
919 | const char *redactbook; | |
920 | uint64_t *spacep; | |
921 | }; | |
922 | ||
923 | static int | |
924 | lzc_send_space_resume_redacted_cb(int fd, void *arg) | |
925 | { | |
926 | struct lzc_send_space_resume_redacted *zssrr = arg; | |
927 | return (lzc_send_space_resume_redacted_cb_impl(zssrr->snapname, | |
928 | zssrr->from, zssrr->flags, zssrr->resumeobj, zssrr->resumeoff, | |
929 | zssrr->resume_bytes, zssrr->redactbook, fd, zssrr->spacep)); | |
930 | } | |
931 | ||
932 | int | |
933 | lzc_send_space_resume_redacted(const char *snapname, const char *from, | |
934 | enum lzc_send_flags flags, uint64_t resumeobj, uint64_t resumeoff, | |
935 | uint64_t resume_bytes, const char *redactbook, int fd, uint64_t *spacep) | |
936 | { | |
937 | struct lzc_send_space_resume_redacted zssrr = { | |
938 | .snapname = snapname, | |
939 | .from = from, | |
940 | .flags = flags, | |
941 | .resumeobj = resumeobj, | |
942 | .resumeoff = resumeoff, | |
943 | .resume_bytes = resume_bytes, | |
944 | .redactbook = redactbook, | |
945 | .spacep = spacep, | |
946 | }; | |
947 | return (lzc_send_wrapper(lzc_send_space_resume_redacted_cb, | |
948 | fd, &zssrr)); | |
949 | } | |
950 | ||
30af21b0 PD |
951 | int |
952 | lzc_send_space(const char *snapname, const char *from, | |
953 | enum lzc_send_flags flags, uint64_t *spacep) | |
954 | { | |
955 | return (lzc_send_space_resume_redacted(snapname, from, flags, 0, 0, 0, | |
956 | NULL, -1, spacep)); | |
957 | } | |
958 | ||
6f1ffb06 MA |
959 | static int |
960 | recv_read(int fd, void *buf, int ilen) | |
961 | { | |
962 | char *cp = buf; | |
963 | int rv; | |
964 | int len = ilen; | |
965 | ||
966 | do { | |
967 | rv = read(fd, cp, len); | |
968 | cp += rv; | |
969 | len -= rv; | |
970 | } while (rv > 0); | |
971 | ||
972 | if (rv < 0 || len != 0) | |
973 | return (EIO); | |
974 | ||
975 | return (0); | |
976 | } | |
977 | ||
43e52edd | 978 | /* |
b5256303 TC |
979 | * Linux adds ZFS_IOC_RECV_NEW for resumable and raw streams and preserves the |
980 | * legacy ZFS_IOC_RECV user/kernel interface. The new interface supports all | |
981 | * stream options but is currently only used for resumable streams. This way | |
982 | * updated user space utilities will interoperate with older kernel modules. | |
43e52edd BB |
983 | * |
984 | * Non-Linux OpenZFS platforms have opted to modify the legacy interface. | |
985 | */ | |
47dfff3b | 986 | static int |
a3eeab2d | 987 | recv_impl(const char *snapname, nvlist_t *recvdprops, nvlist_t *localprops, |
d9c460a0 | 988 | uint8_t *wkeydata, uint_t wkeylen, const char *origin, boolean_t force, |
e8cf3a4f | 989 | boolean_t heal, boolean_t resumable, boolean_t raw, int input_fd, |
196bee4c MA |
990 | const dmu_replay_record_t *begin_record, uint64_t *read_bytes, |
991 | uint64_t *errflags, nvlist_t **errors) | |
6f1ffb06 | 992 | { |
43e52edd BB |
993 | dmu_replay_record_t drr; |
994 | char fsname[MAXPATHLEN]; | |
6f1ffb06 | 995 | char *atp; |
6f1ffb06 | 996 | int error; |
30af21b0 | 997 | boolean_t payload = B_FALSE; |
6f1ffb06 | 998 | |
e2454897 GM |
999 | ASSERT3S(g_refcount, >, 0); |
1000 | VERIFY3S(g_fd, !=, -1); | |
1001 | ||
43e52edd BB |
1002 | /* Set 'fsname' to the name of containing filesystem */ |
1003 | (void) strlcpy(fsname, snapname, sizeof (fsname)); | |
1004 | atp = strchr(fsname, '@'); | |
6f1ffb06 MA |
1005 | if (atp == NULL) |
1006 | return (EINVAL); | |
1007 | *atp = '\0'; | |
1008 | ||
43e52edd BB |
1009 | /* If the fs does not exist, try its parent. */ |
1010 | if (!lzc_exists(fsname)) { | |
1011 | char *slashp = strrchr(fsname, '/'); | |
6f1ffb06 MA |
1012 | if (slashp == NULL) |
1013 | return (ENOENT); | |
1014 | *slashp = '\0'; | |
43e52edd | 1015 | } |
6f1ffb06 | 1016 | |
8a3d7735 AZ |
1017 | /* |
1018 | * It is not uncommon for gigabytes to be processed by zfs receive. | |
1019 | * Speculatively increase the buffer size if supported by the platform. | |
1020 | */ | |
1021 | struct stat sb; | |
1022 | if (fstat(input_fd, &sb) == -1) | |
1023 | return (errno); | |
1024 | if (S_ISFIFO(sb.st_mode)) | |
fbbea09d | 1025 | (void) max_pipe_buffer(input_fd); |
8a3d7735 | 1026 | |
43e52edd BB |
1027 | /* |
1028 | * The begin_record is normally a non-byteswapped BEGIN record. | |
1029 | * For resumable streams it may be set to any non-byteswapped | |
1030 | * dmu_replay_record_t. | |
1031 | */ | |
1032 | if (begin_record == NULL) { | |
1033 | error = recv_read(input_fd, &drr, sizeof (drr)); | |
1034 | if (error != 0) | |
1035 | return (error); | |
1036 | } else { | |
1037 | drr = *begin_record; | |
30af21b0 | 1038 | payload = (begin_record->drr_payloadlen != 0); |
6f1ffb06 MA |
1039 | } |
1040 | ||
d9c460a0 | 1041 | /* |
78595377 | 1042 | * All receives with a payload should use the new interface. |
d9c460a0 | 1043 | */ |
e8cf3a4f | 1044 | if (resumable || heal || raw || wkeydata != NULL || payload) { |
43e52edd BB |
1045 | nvlist_t *outnvl = NULL; |
1046 | nvlist_t *innvl = fnvlist_alloc(); | |
6f1ffb06 | 1047 | |
43e52edd | 1048 | fnvlist_add_string(innvl, "snapname", snapname); |
6f1ffb06 | 1049 | |
a3eeab2d | 1050 | if (recvdprops != NULL) |
1051 | fnvlist_add_nvlist(innvl, "props", recvdprops); | |
1052 | ||
1053 | if (localprops != NULL) | |
1054 | fnvlist_add_nvlist(innvl, "localprops", localprops); | |
6f1ffb06 | 1055 | |
d9c460a0 TC |
1056 | if (wkeydata != NULL) { |
1057 | /* | |
1058 | * wkeydata must be placed in the special | |
1059 | * ZPOOL_HIDDEN_ARGS nvlist so that it | |
1060 | * will not be printed to the zpool history. | |
1061 | */ | |
1062 | nvlist_t *hidden_args = fnvlist_alloc(); | |
1063 | fnvlist_add_uint8_array(hidden_args, "wkeydata", | |
1064 | wkeydata, wkeylen); | |
1065 | fnvlist_add_nvlist(innvl, ZPOOL_HIDDEN_ARGS, | |
1066 | hidden_args); | |
1067 | nvlist_free(hidden_args); | |
1068 | } | |
1069 | ||
43e52edd BB |
1070 | if (origin != NULL && strlen(origin)) |
1071 | fnvlist_add_string(innvl, "origin", origin); | |
1072 | ||
1073 | fnvlist_add_byte_array(innvl, "begin_record", | |
02730c33 | 1074 | (uchar_t *)&drr, sizeof (drr)); |
43e52edd BB |
1075 | |
1076 | fnvlist_add_int32(innvl, "input_fd", input_fd); | |
1077 | ||
1078 | if (force) | |
1079 | fnvlist_add_boolean(innvl, "force"); | |
1080 | ||
1081 | if (resumable) | |
1082 | fnvlist_add_boolean(innvl, "resumable"); | |
1083 | ||
e8cf3a4f AP |
1084 | if (heal) |
1085 | fnvlist_add_boolean(innvl, "heal"); | |
43e52edd BB |
1086 | |
1087 | error = lzc_ioctl(ZFS_IOC_RECV_NEW, fsname, innvl, &outnvl); | |
1088 | ||
1089 | if (error == 0 && read_bytes != NULL) | |
1090 | error = nvlist_lookup_uint64(outnvl, "read_bytes", | |
1091 | read_bytes); | |
1092 | ||
1093 | if (error == 0 && errflags != NULL) | |
1094 | error = nvlist_lookup_uint64(outnvl, "error_flags", | |
1095 | errflags); | |
1096 | ||
43e52edd BB |
1097 | if (error == 0 && errors != NULL) { |
1098 | nvlist_t *nvl; | |
1099 | error = nvlist_lookup_nvlist(outnvl, "errors", &nvl); | |
1100 | if (error == 0) | |
1101 | *errors = fnvlist_dup(nvl); | |
1102 | } | |
1103 | ||
1104 | fnvlist_free(innvl); | |
1105 | fnvlist_free(outnvl); | |
fd41e935 | 1106 | } else { |
43e52edd BB |
1107 | zfs_cmd_t zc = {"\0"}; |
1108 | char *packed = NULL; | |
1109 | size_t size; | |
6f1ffb06 | 1110 | |
43e52edd | 1111 | ASSERT3S(g_refcount, >, 0); |
6f1ffb06 | 1112 | |
ac4985e4 | 1113 | (void) strlcpy(zc.zc_name, fsname, sizeof (zc.zc_name)); |
43e52edd | 1114 | (void) strlcpy(zc.zc_value, snapname, sizeof (zc.zc_value)); |
6f1ffb06 | 1115 | |
a3eeab2d | 1116 | if (recvdprops != NULL) { |
1117 | packed = fnvlist_pack(recvdprops, &size); | |
43e52edd BB |
1118 | zc.zc_nvlist_src = (uint64_t)(uintptr_t)packed; |
1119 | zc.zc_nvlist_src_size = size; | |
1120 | } | |
47dfff3b | 1121 | |
a3eeab2d | 1122 | if (localprops != NULL) { |
1123 | packed = fnvlist_pack(localprops, &size); | |
1124 | zc.zc_nvlist_conf = (uint64_t)(uintptr_t)packed; | |
1125 | zc.zc_nvlist_conf_size = size; | |
1126 | } | |
1127 | ||
43e52edd BB |
1128 | if (origin != NULL) |
1129 | (void) strlcpy(zc.zc_string, origin, | |
1130 | sizeof (zc.zc_string)); | |
6f1ffb06 | 1131 | |
43e52edd BB |
1132 | ASSERT3S(drr.drr_type, ==, DRR_BEGIN); |
1133 | zc.zc_begin_record = drr.drr_u.drr_begin; | |
1134 | zc.zc_guid = force; | |
1135 | zc.zc_cookie = input_fd; | |
1136 | zc.zc_cleanup_fd = -1; | |
1137 | zc.zc_action_handle = 0; | |
1138 | ||
43e52edd BB |
1139 | zc.zc_nvlist_dst_size = 128 * 1024; |
1140 | zc.zc_nvlist_dst = (uint64_t)(uintptr_t) | |
1141 | malloc(zc.zc_nvlist_dst_size); | |
1142 | ||
514498fe | 1143 | error = lzc_ioctl_fd(g_fd, ZFS_IOC_RECV, &zc); |
43e52edd BB |
1144 | if (error != 0) { |
1145 | error = errno; | |
1146 | } else { | |
1147 | if (read_bytes != NULL) | |
1148 | *read_bytes = zc.zc_cookie; | |
1149 | ||
1150 | if (errflags != NULL) | |
1151 | *errflags = zc.zc_obj; | |
1152 | ||
43e52edd BB |
1153 | if (errors != NULL) |
1154 | VERIFY0(nvlist_unpack( | |
1155 | (void *)(uintptr_t)zc.zc_nvlist_dst, | |
1156 | zc.zc_nvlist_dst_size, errors, KM_SLEEP)); | |
1157 | } | |
1158 | ||
1159 | if (packed != NULL) | |
1160 | fnvlist_pack_free(packed, size); | |
1161 | free((void *)(uintptr_t)zc.zc_nvlist_dst); | |
1162 | } | |
6f1ffb06 | 1163 | |
6f1ffb06 MA |
1164 | return (error); |
1165 | } | |
46ba1e59 | 1166 | |
47dfff3b MA |
1167 | /* |
1168 | * The simplest receive case: receive from the specified fd, creating the | |
1169 | * specified snapshot. Apply the specified properties as "received" properties | |
1170 | * (which can be overridden by locally-set properties). If the stream is a | |
1171 | * clone, its origin snapshot must be specified by 'origin'. The 'force' | |
1172 | * flag will cause the target filesystem to be rolled back or destroyed if | |
1173 | * necessary to receive. | |
1174 | * | |
1175 | * Return 0 on success or an errno on failure. | |
1176 | * | |
1177 | * Note: this interface does not work on dedup'd streams | |
1178 | * (those with DMU_BACKUP_FEATURE_DEDUP). | |
1179 | */ | |
1180 | int | |
1181 | lzc_receive(const char *snapname, nvlist_t *props, const char *origin, | |
b5256303 | 1182 | boolean_t force, boolean_t raw, int fd) |
47dfff3b | 1183 | { |
d9c460a0 | 1184 | return (recv_impl(snapname, props, NULL, NULL, 0, origin, force, |
e8cf3a4f | 1185 | B_FALSE, B_FALSE, raw, fd, NULL, NULL, NULL, NULL)); |
47dfff3b MA |
1186 | } |
1187 | ||
1188 | /* | |
1189 | * Like lzc_receive, but if the receive fails due to premature stream | |
1190 | * termination, the intermediate state will be preserved on disk. In this | |
1191 | * case, ECKSUM will be returned. The receive may subsequently be resumed | |
1192 | * with a resuming send stream generated by lzc_send_resume(). | |
1193 | */ | |
1194 | int | |
1195 | lzc_receive_resumable(const char *snapname, nvlist_t *props, const char *origin, | |
b5256303 | 1196 | boolean_t force, boolean_t raw, int fd) |
47dfff3b | 1197 | { |
d9c460a0 | 1198 | return (recv_impl(snapname, props, NULL, NULL, 0, origin, force, |
e8cf3a4f | 1199 | B_FALSE, B_TRUE, raw, fd, NULL, NULL, NULL, NULL)); |
fd41e935 BB |
1200 | } |
1201 | ||
1202 | /* | |
1203 | * Like lzc_receive, but allows the caller to read the begin record and then to | |
1204 | * pass it in. That could be useful if the caller wants to derive, for example, | |
1205 | * the snapname or the origin parameters based on the information contained in | |
1206 | * the begin record. | |
1207 | * The begin record must be in its original form as read from the stream, | |
1208 | * in other words, it should not be byteswapped. | |
1209 | * | |
1210 | * The 'resumable' parameter allows to obtain the same behavior as with | |
1211 | * lzc_receive_resumable. | |
1212 | */ | |
1213 | int | |
1214 | lzc_receive_with_header(const char *snapname, nvlist_t *props, | |
b5256303 TC |
1215 | const char *origin, boolean_t force, boolean_t resumable, boolean_t raw, |
1216 | int fd, const dmu_replay_record_t *begin_record) | |
fd41e935 BB |
1217 | { |
1218 | if (begin_record == NULL) | |
1219 | return (EINVAL); | |
b5256303 | 1220 | |
d9c460a0 | 1221 | return (recv_impl(snapname, props, NULL, NULL, 0, origin, force, |
e8cf3a4f | 1222 | B_FALSE, resumable, raw, fd, begin_record, NULL, NULL, NULL)); |
43e52edd BB |
1223 | } |
1224 | ||
1225 | /* | |
1226 | * Like lzc_receive, but allows the caller to pass all supported arguments | |
1227 | * and retrieve all values returned. The only additional input parameter | |
1228 | * is 'cleanup_fd' which is used to set a cleanup-on-exit file descriptor. | |
1229 | * | |
1230 | * The following parameters all provide return values. Several may be set | |
1231 | * in the failure case and will contain additional information. | |
1232 | * | |
1233 | * The 'read_bytes' value will be set to the total number of bytes read. | |
1234 | * | |
1235 | * The 'errflags' value will contain zprop_errflags_t flags which are | |
1236 | * used to describe any failures. | |
1237 | * | |
196bee4c | 1238 | * The 'action_handle' and 'cleanup_fd' are no longer used, and are ignored. |
43e52edd BB |
1239 | * |
1240 | * The 'errors' nvlist contains an entry for each unapplied received | |
1241 | * property. Callers are responsible for freeing this nvlist. | |
1242 | */ | |
e1af0d0d RM |
1243 | int |
1244 | lzc_receive_one(const char *snapname, nvlist_t *props, | |
b5256303 TC |
1245 | const char *origin, boolean_t force, boolean_t resumable, boolean_t raw, |
1246 | int input_fd, const dmu_replay_record_t *begin_record, int cleanup_fd, | |
43e52edd BB |
1247 | uint64_t *read_bytes, uint64_t *errflags, uint64_t *action_handle, |
1248 | nvlist_t **errors) | |
1249 | { | |
bfc17897 | 1250 | (void) action_handle, (void) cleanup_fd; |
d9c460a0 | 1251 | return (recv_impl(snapname, props, NULL, NULL, 0, origin, force, |
e8cf3a4f | 1252 | B_FALSE, resumable, raw, input_fd, begin_record, |
196bee4c | 1253 | read_bytes, errflags, errors)); |
a3eeab2d | 1254 | } |
1255 | ||
1256 | /* | |
1257 | * Like lzc_receive_one, but allows the caller to pass an additional 'cmdprops' | |
1258 | * argument. | |
1259 | * | |
1260 | * The 'cmdprops' nvlist contains both override ('zfs receive -o') and | |
1261 | * exclude ('zfs receive -x') properties. Callers are responsible for freeing | |
1262 | * this nvlist | |
1263 | */ | |
e1af0d0d RM |
1264 | int |
1265 | lzc_receive_with_cmdprops(const char *snapname, nvlist_t *props, | |
d9c460a0 TC |
1266 | nvlist_t *cmdprops, uint8_t *wkeydata, uint_t wkeylen, const char *origin, |
1267 | boolean_t force, boolean_t resumable, boolean_t raw, int input_fd, | |
b5256303 TC |
1268 | const dmu_replay_record_t *begin_record, int cleanup_fd, |
1269 | uint64_t *read_bytes, uint64_t *errflags, uint64_t *action_handle, | |
1270 | nvlist_t **errors) | |
a3eeab2d | 1271 | { |
bfc17897 | 1272 | (void) action_handle, (void) cleanup_fd; |
d9c460a0 | 1273 | return (recv_impl(snapname, props, cmdprops, wkeydata, wkeylen, origin, |
e8cf3a4f AP |
1274 | force, B_FALSE, resumable, raw, input_fd, begin_record, |
1275 | read_bytes, errflags, errors)); | |
1276 | } | |
1277 | ||
1278 | /* | |
1279 | * Like lzc_receive_with_cmdprops, but allows the caller to pass an additional | |
1280 | * 'heal' argument. | |
1281 | * | |
1282 | * The heal arguments tells us to heal the provided snapshot using the provided | |
1283 | * send stream | |
1284 | */ | |
1285 | int lzc_receive_with_heal(const char *snapname, nvlist_t *props, | |
1286 | nvlist_t *cmdprops, uint8_t *wkeydata, uint_t wkeylen, const char *origin, | |
1287 | boolean_t force, boolean_t heal, boolean_t resumable, boolean_t raw, | |
1288 | int input_fd, const dmu_replay_record_t *begin_record, int cleanup_fd, | |
1289 | uint64_t *read_bytes, uint64_t *errflags, uint64_t *action_handle, | |
1290 | nvlist_t **errors) | |
1291 | { | |
1292 | (void) action_handle, (void) cleanup_fd; | |
1293 | return (recv_impl(snapname, props, cmdprops, wkeydata, wkeylen, origin, | |
1294 | force, heal, resumable, raw, input_fd, begin_record, | |
196bee4c | 1295 | read_bytes, errflags, errors)); |
47dfff3b MA |
1296 | } |
1297 | ||
46ba1e59 MA |
1298 | /* |
1299 | * Roll back this filesystem or volume to its most recent snapshot. | |
1300 | * If snapnamebuf is not NULL, it will be filled in with the name | |
1301 | * of the most recent snapshot. | |
8ca78ab0 AG |
1302 | * Note that the latest snapshot may change if a new one is concurrently |
1303 | * created or the current one is destroyed. lzc_rollback_to can be used | |
1304 | * to roll back to a specific latest snapshot. | |
46ba1e59 MA |
1305 | * |
1306 | * Return 0 on success or an errno on failure. | |
1307 | */ | |
1308 | int | |
1309 | lzc_rollback(const char *fsname, char *snapnamebuf, int snapnamelen) | |
1310 | { | |
1311 | nvlist_t *args; | |
1312 | nvlist_t *result; | |
1313 | int err; | |
1314 | ||
1315 | args = fnvlist_alloc(); | |
1316 | err = lzc_ioctl(ZFS_IOC_ROLLBACK, fsname, args, &result); | |
1317 | nvlist_free(args); | |
1318 | if (err == 0 && snapnamebuf != NULL) { | |
1319 | const char *snapname = fnvlist_lookup_string(result, "target"); | |
1320 | (void) strlcpy(snapnamebuf, snapname, snapnamelen); | |
1321 | } | |
bb7ffdaf GM |
1322 | nvlist_free(result); |
1323 | ||
46ba1e59 MA |
1324 | return (err); |
1325 | } | |
da536844 | 1326 | |
8ca78ab0 AG |
1327 | /* |
1328 | * Roll back this filesystem or volume to the specified snapshot, | |
1329 | * if possible. | |
1330 | * | |
1331 | * Return 0 on success or an errno on failure. | |
1332 | */ | |
1333 | int | |
1334 | lzc_rollback_to(const char *fsname, const char *snapname) | |
1335 | { | |
1336 | nvlist_t *args; | |
1337 | nvlist_t *result; | |
1338 | int err; | |
1339 | ||
1340 | args = fnvlist_alloc(); | |
1341 | fnvlist_add_string(args, "target", snapname); | |
1342 | err = lzc_ioctl(ZFS_IOC_ROLLBACK, fsname, args, &result); | |
1343 | nvlist_free(args); | |
1344 | nvlist_free(result); | |
1345 | return (err); | |
1346 | } | |
1347 | ||
da536844 | 1348 | /* |
a73f361f | 1349 | * Creates new bookmarks from existing snapshot or bookmark. |
da536844 | 1350 | * |
a73f361f CS |
1351 | * The bookmarks nvlist maps from the full name of the new bookmark to |
1352 | * the full name of the source snapshot or bookmark. | |
1353 | * All the bookmarks and snapshots must be in the same pool. | |
1354 | * The new bookmarks names must be unique. | |
1355 | * => see function dsl_bookmark_create_nvl_validate | |
da536844 MA |
1356 | * |
1357 | * The returned results nvlist will have an entry for each bookmark that failed. | |
1358 | * The value will be the (int32) error code. | |
1359 | * | |
1360 | * The return value will be 0 if all bookmarks were created, otherwise it will | |
1361 | * be the errno of a (undetermined) bookmarks that failed. | |
1362 | */ | |
1363 | int | |
1364 | lzc_bookmark(nvlist_t *bookmarks, nvlist_t **errlist) | |
1365 | { | |
1366 | nvpair_t *elem; | |
1367 | int error; | |
eca7b760 | 1368 | char pool[ZFS_MAX_DATASET_NAME_LEN]; |
da536844 | 1369 | |
a73f361f | 1370 | /* determine pool name from first bookmark */ |
da536844 MA |
1371 | elem = nvlist_next_nvpair(bookmarks, NULL); |
1372 | if (elem == NULL) | |
1373 | return (0); | |
1374 | (void) strlcpy(pool, nvpair_name(elem), sizeof (pool)); | |
1375 | pool[strcspn(pool, "/#")] = '\0'; | |
1376 | ||
1377 | error = lzc_ioctl(ZFS_IOC_BOOKMARK, pool, bookmarks, errlist); | |
1378 | ||
1379 | return (error); | |
1380 | } | |
1381 | ||
1382 | /* | |
1383 | * Retrieve bookmarks. | |
1384 | * | |
1385 | * Retrieve the list of bookmarks for the given file system. The props | |
1386 | * parameter is an nvlist of property names (with no values) that will be | |
1387 | * returned for each bookmark. | |
1388 | * | |
30af21b0 PD |
1389 | * The following are valid properties on bookmarks, most of which are numbers |
1390 | * (represented as uint64 in the nvlist), except redact_snaps, which is a | |
1391 | * uint64 array, and redact_complete, which is a boolean | |
da536844 MA |
1392 | * |
1393 | * "guid" - globally unique identifier of the snapshot it refers to | |
1394 | * "createtxg" - txg when the snapshot it refers to was created | |
1395 | * "creation" - timestamp when the snapshot it refers to was created | |
f00ab3f2 | 1396 | * "ivsetguid" - IVset guid for identifying encrypted snapshots |
30af21b0 PD |
1397 | * "redact_snaps" - list of guids of the redaction snapshots for the specified |
1398 | * bookmark. If the bookmark is not a redaction bookmark, the nvlist will | |
1399 | * not contain an entry for this value. If it is redacted with respect to | |
1400 | * no snapshots, it will contain value -> NULL uint64 array | |
1401 | * "redact_complete" - boolean value; true if the redaction bookmark is | |
1402 | * complete, false otherwise. | |
da536844 MA |
1403 | * |
1404 | * The format of the returned nvlist as follows: | |
1405 | * <short name of bookmark> -> { | |
1406 | * <name of property> -> { | |
1407 | * "value" -> uint64 | |
1408 | * } | |
30af21b0 PD |
1409 | * ... |
1410 | * "redact_snaps" -> { | |
1411 | * "value" -> uint64 array | |
1412 | * } | |
1413 | * "redact_complete" -> { | |
1414 | * "value" -> boolean value | |
1415 | * } | |
da536844 MA |
1416 | * } |
1417 | */ | |
1418 | int | |
1419 | lzc_get_bookmarks(const char *fsname, nvlist_t *props, nvlist_t **bmarks) | |
1420 | { | |
1421 | return (lzc_ioctl(ZFS_IOC_GET_BOOKMARKS, fsname, props, bmarks)); | |
1422 | } | |
1423 | ||
30af21b0 PD |
1424 | /* |
1425 | * Get bookmark properties. | |
1426 | * | |
1427 | * Given a bookmark's full name, retrieve all properties for the bookmark. | |
1428 | * | |
1429 | * The format of the returned property list is as follows: | |
1430 | * { | |
1431 | * <name of property> -> { | |
1432 | * "value" -> uint64 | |
1433 | * } | |
1434 | * ... | |
1435 | * "redact_snaps" -> { | |
1436 | * "value" -> uint64 array | |
1437 | * } | |
1438 | */ | |
1439 | int | |
1440 | lzc_get_bookmark_props(const char *bookmark, nvlist_t **props) | |
1441 | { | |
1442 | int error; | |
1443 | ||
1444 | nvlist_t *innvl = fnvlist_alloc(); | |
1445 | error = lzc_ioctl(ZFS_IOC_GET_BOOKMARK_PROPS, bookmark, innvl, props); | |
1446 | fnvlist_free(innvl); | |
1447 | ||
1448 | return (error); | |
1449 | } | |
1450 | ||
da536844 MA |
1451 | /* |
1452 | * Destroys bookmarks. | |
1453 | * | |
1454 | * The keys in the bmarks nvlist are the bookmarks to be destroyed. | |
1455 | * They must all be in the same pool. Bookmarks are specified as | |
1456 | * <fs>#<bmark>. | |
1457 | * | |
1458 | * Bookmarks that do not exist will be silently ignored. | |
1459 | * | |
1460 | * The return value will be 0 if all bookmarks that existed were destroyed. | |
1461 | * | |
1462 | * Otherwise the return value will be the errno of a (undetermined) bookmark | |
1463 | * that failed, no bookmarks will be destroyed, and the errlist will have an | |
1464 | * entry for each bookmarks that failed. The value in the errlist will be | |
1465 | * the (int32) error code. | |
1466 | */ | |
1467 | int | |
1468 | lzc_destroy_bookmarks(nvlist_t *bmarks, nvlist_t **errlist) | |
1469 | { | |
1470 | nvpair_t *elem; | |
1471 | int error; | |
eca7b760 | 1472 | char pool[ZFS_MAX_DATASET_NAME_LEN]; |
da536844 MA |
1473 | |
1474 | /* determine the pool name */ | |
1475 | elem = nvlist_next_nvpair(bmarks, NULL); | |
1476 | if (elem == NULL) | |
1477 | return (0); | |
1478 | (void) strlcpy(pool, nvpair_name(elem), sizeof (pool)); | |
1479 | pool[strcspn(pool, "/#")] = '\0'; | |
1480 | ||
1481 | error = lzc_ioctl(ZFS_IOC_DESTROY_BOOKMARKS, pool, bmarks, errlist); | |
1482 | ||
1483 | return (error); | |
1484 | } | |
b5256303 | 1485 | |
5b72a38d SD |
1486 | static int |
1487 | lzc_channel_program_impl(const char *pool, const char *program, boolean_t sync, | |
1488 | uint64_t instrlimit, uint64_t memlimit, nvlist_t *argnvl, nvlist_t **outnvl) | |
1489 | { | |
1490 | int error; | |
1491 | nvlist_t *args; | |
1492 | ||
1493 | args = fnvlist_alloc(); | |
1494 | fnvlist_add_string(args, ZCP_ARG_PROGRAM, program); | |
1495 | fnvlist_add_nvlist(args, ZCP_ARG_ARGLIST, argnvl); | |
1496 | fnvlist_add_boolean_value(args, ZCP_ARG_SYNC, sync); | |
1497 | fnvlist_add_uint64(args, ZCP_ARG_INSTRLIMIT, instrlimit); | |
1498 | fnvlist_add_uint64(args, ZCP_ARG_MEMLIMIT, memlimit); | |
1499 | error = lzc_ioctl(ZFS_IOC_CHANNEL_PROGRAM, pool, args, outnvl); | |
1500 | fnvlist_free(args); | |
1501 | ||
1502 | return (error); | |
1503 | } | |
1504 | ||
d99a0153 CW |
1505 | /* |
1506 | * Executes a channel program. | |
1507 | * | |
1508 | * If this function returns 0 the channel program was successfully loaded and | |
1509 | * ran without failing. Note that individual commands the channel program ran | |
1510 | * may have failed and the channel program is responsible for reporting such | |
1511 | * errors through outnvl if they are important. | |
1512 | * | |
1513 | * This method may also return: | |
1514 | * | |
1515 | * EINVAL The program contains syntax errors, or an invalid memory or time | |
1516 | * limit was given. No part of the channel program was executed. | |
1517 | * If caused by syntax errors, 'outnvl' contains information about the | |
1518 | * errors. | |
1519 | * | |
1520 | * ECHRNG The program was executed, but encountered a runtime error, such as | |
1521 | * calling a function with incorrect arguments, invoking the error() | |
1522 | * function directly, failing an assert() command, etc. Some portion | |
1523 | * of the channel program may have executed and committed changes. | |
1524 | * Information about the failure can be found in 'outnvl'. | |
1525 | * | |
1526 | * ENOMEM The program fully executed, but the output buffer was not large | |
1527 | * enough to store the returned value. No output is returned through | |
1528 | * 'outnvl'. | |
1529 | * | |
1530 | * ENOSPC The program was terminated because it exceeded its memory usage | |
1531 | * limit. Some portion of the channel program may have executed and | |
1532 | * committed changes to disk. No output is returned through 'outnvl'. | |
1533 | * | |
1534 | * ETIME The program was terminated because it exceeded its Lua instruction | |
1535 | * limit. Some portion of the channel program may have executed and | |
1536 | * committed changes to disk. No output is returned through 'outnvl'. | |
1537 | */ | |
1538 | int | |
1539 | lzc_channel_program(const char *pool, const char *program, uint64_t instrlimit, | |
1540 | uint64_t memlimit, nvlist_t *argnvl, nvlist_t **outnvl) | |
1541 | { | |
5b72a38d SD |
1542 | return (lzc_channel_program_impl(pool, program, B_TRUE, instrlimit, |
1543 | memlimit, argnvl, outnvl)); | |
1544 | } | |
d99a0153 | 1545 | |
d2734cce SD |
1546 | /* |
1547 | * Creates a checkpoint for the specified pool. | |
1548 | * | |
1549 | * If this function returns 0 the pool was successfully checkpointed. | |
1550 | * | |
1551 | * This method may also return: | |
1552 | * | |
1553 | * ZFS_ERR_CHECKPOINT_EXISTS | |
1554 | * The pool already has a checkpoint. A pools can only have one | |
1555 | * checkpoint at most, at any given time. | |
1556 | * | |
1557 | * ZFS_ERR_DISCARDING_CHECKPOINT | |
1558 | * ZFS is in the middle of discarding a checkpoint for this pool. | |
1559 | * The pool can be checkpointed again once the discard is done. | |
1560 | * | |
1561 | * ZFS_DEVRM_IN_PROGRESS | |
1562 | * A vdev is currently being removed. The pool cannot be | |
1563 | * checkpointed until the device removal is done. | |
1564 | * | |
1565 | * ZFS_VDEV_TOO_BIG | |
1566 | * One or more top-level vdevs exceed the maximum vdev size | |
1567 | * supported for this feature. | |
1568 | */ | |
1569 | int | |
1570 | lzc_pool_checkpoint(const char *pool) | |
1571 | { | |
1572 | int error; | |
1573 | ||
1574 | nvlist_t *result = NULL; | |
1575 | nvlist_t *args = fnvlist_alloc(); | |
1576 | ||
1577 | error = lzc_ioctl(ZFS_IOC_POOL_CHECKPOINT, pool, args, &result); | |
1578 | ||
1579 | fnvlist_free(args); | |
1580 | fnvlist_free(result); | |
1581 | ||
1582 | return (error); | |
1583 | } | |
1584 | ||
1585 | /* | |
1586 | * Discard the checkpoint from the specified pool. | |
1587 | * | |
1588 | * If this function returns 0 the checkpoint was successfully discarded. | |
1589 | * | |
1590 | * This method may also return: | |
1591 | * | |
1592 | * ZFS_ERR_NO_CHECKPOINT | |
1593 | * The pool does not have a checkpoint. | |
1594 | * | |
1595 | * ZFS_ERR_DISCARDING_CHECKPOINT | |
1596 | * ZFS is already in the middle of discarding the checkpoint. | |
1597 | */ | |
1598 | int | |
1599 | lzc_pool_checkpoint_discard(const char *pool) | |
1600 | { | |
1601 | int error; | |
1602 | ||
1603 | nvlist_t *result = NULL; | |
1604 | nvlist_t *args = fnvlist_alloc(); | |
1605 | ||
1606 | error = lzc_ioctl(ZFS_IOC_POOL_DISCARD_CHECKPOINT, pool, args, &result); | |
1607 | ||
1608 | fnvlist_free(args); | |
1609 | fnvlist_free(result); | |
1610 | ||
1611 | return (error); | |
1612 | } | |
1613 | ||
5b72a38d SD |
1614 | /* |
1615 | * Executes a read-only channel program. | |
1616 | * | |
1617 | * A read-only channel program works programmatically the same way as a | |
1618 | * normal channel program executed with lzc_channel_program(). The only | |
1619 | * difference is it runs exclusively in open-context and therefore can | |
1620 | * return faster. The downside to that, is that the program cannot change | |
1621 | * on-disk state by calling functions from the zfs.sync submodule. | |
1622 | * | |
1623 | * The return values of this function (and their meaning) are exactly the | |
1624 | * same as the ones described in lzc_channel_program(). | |
1625 | */ | |
1626 | int | |
1627 | lzc_channel_program_nosync(const char *pool, const char *program, | |
1628 | uint64_t timeout, uint64_t memlimit, nvlist_t *argnvl, nvlist_t **outnvl) | |
1629 | { | |
1630 | return (lzc_channel_program_impl(pool, program, B_FALSE, timeout, | |
1631 | memlimit, argnvl, outnvl)); | |
d99a0153 CW |
1632 | } |
1633 | ||
2a673e76 AJ |
1634 | int |
1635 | lzc_get_vdev_prop(const char *poolname, nvlist_t *innvl, nvlist_t **outnvl) | |
1636 | { | |
1637 | return (lzc_ioctl(ZFS_IOC_VDEV_GET_PROPS, poolname, innvl, outnvl)); | |
1638 | } | |
1639 | ||
1640 | int | |
1641 | lzc_set_vdev_prop(const char *poolname, nvlist_t *innvl, nvlist_t **outnvl) | |
1642 | { | |
1643 | return (lzc_ioctl(ZFS_IOC_VDEV_SET_PROPS, poolname, innvl, outnvl)); | |
1644 | } | |
1645 | ||
b5256303 TC |
1646 | /* |
1647 | * Performs key management functions | |
1648 | * | |
85ce3f4f | 1649 | * crypto_cmd should be a value from dcp_cmd_t. If the command specifies to |
1650 | * load or change a wrapping key, the key should be specified in the | |
1651 | * hidden_args nvlist so that it is not logged. | |
b5256303 TC |
1652 | */ |
1653 | int | |
1654 | lzc_load_key(const char *fsname, boolean_t noop, uint8_t *wkeydata, | |
1655 | uint_t wkeylen) | |
1656 | { | |
1657 | int error; | |
1658 | nvlist_t *ioc_args; | |
1659 | nvlist_t *hidden_args; | |
1660 | ||
1661 | if (wkeydata == NULL) | |
1662 | return (EINVAL); | |
1663 | ||
1664 | ioc_args = fnvlist_alloc(); | |
1665 | hidden_args = fnvlist_alloc(); | |
1666 | fnvlist_add_uint8_array(hidden_args, "wkeydata", wkeydata, wkeylen); | |
1667 | fnvlist_add_nvlist(ioc_args, ZPOOL_HIDDEN_ARGS, hidden_args); | |
1668 | if (noop) | |
1669 | fnvlist_add_boolean(ioc_args, "noop"); | |
1670 | error = lzc_ioctl(ZFS_IOC_LOAD_KEY, fsname, ioc_args, NULL); | |
1671 | nvlist_free(hidden_args); | |
1672 | nvlist_free(ioc_args); | |
1673 | ||
1674 | return (error); | |
1675 | } | |
1676 | ||
1677 | int | |
1678 | lzc_unload_key(const char *fsname) | |
1679 | { | |
1680 | return (lzc_ioctl(ZFS_IOC_UNLOAD_KEY, fsname, NULL, NULL)); | |
1681 | } | |
1682 | ||
1683 | int | |
1684 | lzc_change_key(const char *fsname, uint64_t crypt_cmd, nvlist_t *props, | |
1685 | uint8_t *wkeydata, uint_t wkeylen) | |
1686 | { | |
1687 | int error; | |
1688 | nvlist_t *ioc_args = fnvlist_alloc(); | |
1689 | nvlist_t *hidden_args = NULL; | |
1690 | ||
1691 | fnvlist_add_uint64(ioc_args, "crypt_cmd", crypt_cmd); | |
1692 | ||
1693 | if (wkeydata != NULL) { | |
1694 | hidden_args = fnvlist_alloc(); | |
1695 | fnvlist_add_uint8_array(hidden_args, "wkeydata", wkeydata, | |
1696 | wkeylen); | |
1697 | fnvlist_add_nvlist(ioc_args, ZPOOL_HIDDEN_ARGS, hidden_args); | |
1698 | } | |
1699 | ||
1700 | if (props != NULL) | |
1701 | fnvlist_add_nvlist(ioc_args, "props", props); | |
1702 | ||
1703 | error = lzc_ioctl(ZFS_IOC_CHANGE_KEY, fsname, ioc_args, NULL); | |
1704 | nvlist_free(hidden_args); | |
1705 | nvlist_free(ioc_args); | |
d99a0153 | 1706 | |
b5256303 TC |
1707 | return (error); |
1708 | } | |
d3f2cd7e AB |
1709 | |
1710 | int | |
1711 | lzc_reopen(const char *pool_name, boolean_t scrub_restart) | |
1712 | { | |
1713 | nvlist_t *args = fnvlist_alloc(); | |
1714 | int error; | |
1715 | ||
1716 | fnvlist_add_boolean_value(args, "scrub_restart", scrub_restart); | |
1717 | ||
1718 | error = lzc_ioctl(ZFS_IOC_POOL_REOPEN, pool_name, args, NULL); | |
1719 | nvlist_free(args); | |
1720 | return (error); | |
1721 | } | |
619f0976 GW |
1722 | |
1723 | /* | |
1724 | * Changes initializing state. | |
1725 | * | |
1726 | * vdevs should be a list of (<key>, guid) where guid is a uint64 vdev GUID. | |
1727 | * The key is ignored. | |
1728 | * | |
1729 | * If there are errors related to vdev arguments, per-vdev errors are returned | |
1730 | * in an nvlist with the key "vdevs". Each error is a (guid, errno) pair where | |
1731 | * guid is stringified with PRIu64, and errno is one of the following as | |
1732 | * an int64_t: | |
1733 | * - ENODEV if the device was not found | |
1734 | * - EINVAL if the devices is not a leaf or is not concrete (e.g. missing) | |
1735 | * - EROFS if the device is not writeable | |
1b939560 BB |
1736 | * - EBUSY start requested but the device is already being either |
1737 | * initialized or trimmed | |
619f0976 GW |
1738 | * - ESRCH cancel/suspend requested but device is not being initialized |
1739 | * | |
1740 | * If the errlist is empty, then return value will be: | |
1741 | * - EINVAL if one or more arguments was invalid | |
1742 | * - Other spa_open failures | |
1743 | * - 0 if the operation succeeded | |
1744 | */ | |
1745 | int | |
1746 | lzc_initialize(const char *poolname, pool_initialize_func_t cmd_type, | |
1747 | nvlist_t *vdevs, nvlist_t **errlist) | |
1748 | { | |
1749 | int error; | |
1b939560 | 1750 | |
619f0976 GW |
1751 | nvlist_t *args = fnvlist_alloc(); |
1752 | fnvlist_add_uint64(args, ZPOOL_INITIALIZE_COMMAND, (uint64_t)cmd_type); | |
1753 | fnvlist_add_nvlist(args, ZPOOL_INITIALIZE_VDEVS, vdevs); | |
1754 | ||
1755 | error = lzc_ioctl(ZFS_IOC_POOL_INITIALIZE, poolname, args, errlist); | |
1756 | ||
1757 | fnvlist_free(args); | |
1758 | ||
1759 | return (error); | |
1760 | } | |
1b939560 BB |
1761 | |
1762 | /* | |
1763 | * Changes TRIM state. | |
1764 | * | |
1765 | * vdevs should be a list of (<key>, guid) where guid is a uint64 vdev GUID. | |
1766 | * The key is ignored. | |
1767 | * | |
1768 | * If there are errors related to vdev arguments, per-vdev errors are returned | |
1769 | * in an nvlist with the key "vdevs". Each error is a (guid, errno) pair where | |
1770 | * guid is stringified with PRIu64, and errno is one of the following as | |
1771 | * an int64_t: | |
1772 | * - ENODEV if the device was not found | |
1773 | * - EINVAL if the devices is not a leaf or is not concrete (e.g. missing) | |
1774 | * - EROFS if the device is not writeable | |
1775 | * - EBUSY start requested but the device is already being either trimmed | |
1776 | * or initialized | |
1777 | * - ESRCH cancel/suspend requested but device is not being initialized | |
1778 | * - EOPNOTSUPP if the device does not support TRIM (or secure TRIM) | |
1779 | * | |
1780 | * If the errlist is empty, then return value will be: | |
1781 | * - EINVAL if one or more arguments was invalid | |
1782 | * - Other spa_open failures | |
1783 | * - 0 if the operation succeeded | |
1784 | */ | |
1785 | int | |
1786 | lzc_trim(const char *poolname, pool_trim_func_t cmd_type, uint64_t rate, | |
1787 | boolean_t secure, nvlist_t *vdevs, nvlist_t **errlist) | |
1788 | { | |
1789 | int error; | |
1790 | ||
1791 | nvlist_t *args = fnvlist_alloc(); | |
1792 | fnvlist_add_uint64(args, ZPOOL_TRIM_COMMAND, (uint64_t)cmd_type); | |
1793 | fnvlist_add_nvlist(args, ZPOOL_TRIM_VDEVS, vdevs); | |
1794 | fnvlist_add_uint64(args, ZPOOL_TRIM_RATE, rate); | |
1795 | fnvlist_add_boolean_value(args, ZPOOL_TRIM_SECURE, secure); | |
1796 | ||
1797 | error = lzc_ioctl(ZFS_IOC_POOL_TRIM, poolname, args, errlist); | |
1798 | ||
1799 | fnvlist_free(args); | |
1800 | ||
1801 | return (error); | |
1802 | } | |
30af21b0 PD |
1803 | |
1804 | /* | |
1805 | * Create a redaction bookmark named bookname by redacting snapshot with respect | |
1806 | * to all the snapshots in snapnv. | |
1807 | */ | |
1808 | int | |
1809 | lzc_redact(const char *snapshot, const char *bookname, nvlist_t *snapnv) | |
1810 | { | |
1811 | nvlist_t *args = fnvlist_alloc(); | |
1812 | fnvlist_add_string(args, "bookname", bookname); | |
1813 | fnvlist_add_nvlist(args, "snapnv", snapnv); | |
1814 | int error = lzc_ioctl(ZFS_IOC_REDACT, snapshot, args, NULL); | |
1815 | fnvlist_free(args); | |
1816 | return (error); | |
1817 | } | |
e60e158e JG |
1818 | |
1819 | static int | |
1820 | wait_common(const char *pool, zpool_wait_activity_t activity, boolean_t use_tag, | |
1821 | uint64_t tag, boolean_t *waited) | |
1822 | { | |
1823 | nvlist_t *args = fnvlist_alloc(); | |
1824 | nvlist_t *result = NULL; | |
1825 | ||
1826 | fnvlist_add_int32(args, ZPOOL_WAIT_ACTIVITY, activity); | |
1827 | if (use_tag) | |
1828 | fnvlist_add_uint64(args, ZPOOL_WAIT_TAG, tag); | |
1829 | ||
1830 | int error = lzc_ioctl(ZFS_IOC_WAIT, pool, args, &result); | |
1831 | ||
1832 | if (error == 0 && waited != NULL) | |
1833 | *waited = fnvlist_lookup_boolean_value(result, | |
1834 | ZPOOL_WAIT_WAITED); | |
1835 | ||
1836 | fnvlist_free(args); | |
1837 | fnvlist_free(result); | |
1838 | ||
1839 | return (error); | |
1840 | } | |
1841 | ||
1842 | int | |
1843 | lzc_wait(const char *pool, zpool_wait_activity_t activity, boolean_t *waited) | |
1844 | { | |
1845 | return (wait_common(pool, activity, B_FALSE, 0, waited)); | |
1846 | } | |
1847 | ||
1848 | int | |
1849 | lzc_wait_tag(const char *pool, zpool_wait_activity_t activity, uint64_t tag, | |
1850 | boolean_t *waited) | |
1851 | { | |
1852 | return (wait_common(pool, activity, B_TRUE, tag, waited)); | |
1853 | } | |
5a42ef04 PD |
1854 | |
1855 | int | |
1856 | lzc_wait_fs(const char *fs, zfs_wait_activity_t activity, boolean_t *waited) | |
1857 | { | |
1858 | nvlist_t *args = fnvlist_alloc(); | |
1859 | nvlist_t *result = NULL; | |
1860 | ||
1861 | fnvlist_add_int32(args, ZFS_WAIT_ACTIVITY, activity); | |
1862 | ||
1863 | int error = lzc_ioctl(ZFS_IOC_WAIT_FS, fs, args, &result); | |
1864 | ||
1865 | if (error == 0 && waited != NULL) | |
1866 | *waited = fnvlist_lookup_boolean_value(result, | |
1867 | ZFS_WAIT_WAITED); | |
1868 | ||
1869 | fnvlist_free(args); | |
1870 | fnvlist_free(result); | |
1871 | ||
1872 | return (error); | |
1873 | } | |
108a454a PD |
1874 | |
1875 | /* | |
1876 | * Set the bootenv contents for the given pool. | |
1877 | */ | |
1878 | int | |
1db9e6e4 | 1879 | lzc_set_bootenv(const char *pool, const nvlist_t *env) |
108a454a | 1880 | { |
1db9e6e4 | 1881 | return (lzc_ioctl(ZFS_IOC_SET_BOOTENV, pool, (nvlist_t *)env, NULL)); |
108a454a PD |
1882 | } |
1883 | ||
1884 | /* | |
1885 | * Get the contents of the bootenv of the given pool. | |
1886 | */ | |
1887 | int | |
1888 | lzc_get_bootenv(const char *pool, nvlist_t **outnvl) | |
1889 | { | |
1890 | return (lzc_ioctl(ZFS_IOC_GET_BOOTENV, pool, NULL, outnvl)); | |
1891 | } |