]>
Commit | Line | Data |
---|---|---|
34dc7c2f BB |
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 | /* | |
428870ff | 23 | * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. |
cae5b340 AX |
24 | * Copyright (c) 2013, 2015 by Delphix. All rights reserved. |
25 | * Copyright (c) 2016 Intel Corporation. | |
26 | * Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>. | |
34dc7c2f BB |
27 | */ |
28 | ||
34dc7c2f BB |
29 | /* |
30 | * Functions to convert between a list of vdevs and an nvlist representing the | |
31 | * configuration. Each entry in the list can be one of: | |
32 | * | |
33 | * Device vdevs | |
34 | * disk=(path=..., devid=...) | |
35 | * file=(path=...) | |
36 | * | |
37 | * Group vdevs | |
38 | * raidz[1|2]=(...) | |
39 | * mirror=(...) | |
40 | * | |
41 | * Hot spares | |
42 | * | |
43 | * While the underlying implementation supports it, group vdevs cannot contain | |
44 | * other group vdevs. All userland verification of devices is contained within | |
45 | * this file. If successful, the nvlist returned can be passed directly to the | |
46 | * kernel; we've done as much verification as possible in userland. | |
47 | * | |
48 | * Hot spares are a special case, and passed down as an array of disk vdevs, at | |
49 | * the same level as the root of the vdev tree. | |
50 | * | |
51 | * The only function exported by this file is 'make_root_vdev'. The | |
52 | * function performs several passes: | |
53 | * | |
54 | * 1. Construct the vdev specification. Performs syntax validation and | |
55 | * makes sure each device is valid. | |
d603ed6c | 56 | * 2. Check for devices in use. Using libblkid to make sure that no |
34dc7c2f BB |
57 | * devices are also in use. Some can be overridden using the 'force' |
58 | * flag, others cannot. | |
59 | * 3. Check for replication errors if the 'force' flag is not specified. | |
60 | * validates that the replication level is consistent across the | |
61 | * entire pool. | |
62 | * 4. Call libzfs to label any whole disks with an EFI label. | |
63 | */ | |
64 | ||
65 | #include <assert.h> | |
d603ed6c | 66 | #include <ctype.h> |
34dc7c2f BB |
67 | #include <devid.h> |
68 | #include <errno.h> | |
69 | #include <fcntl.h> | |
34dc7c2f BB |
70 | #include <libintl.h> |
71 | #include <libnvpair.h> | |
45d1cae3 | 72 | #include <limits.h> |
cae5b340 | 73 | #include <sys/spa.h> |
c06d4368 AX |
74 | #include <scsi/scsi.h> |
75 | #include <scsi/sg.h> | |
34dc7c2f BB |
76 | #include <stdio.h> |
77 | #include <string.h> | |
78 | #include <unistd.h> | |
79 | #include <sys/efi_partition.h> | |
80 | #include <sys/stat.h> | |
81 | #include <sys/vtoc.h> | |
82 | #include <sys/mntent.h> | |
d603ed6c | 83 | #include <uuid/uuid.h> |
d603ed6c | 84 | #include <blkid/blkid.h> |
34dc7c2f | 85 | #include "zpool_util.h" |
c06d4368 | 86 | #include <sys/zfs_context.h> |
34dc7c2f | 87 | |
34dc7c2f BB |
88 | /* |
89 | * For any given vdev specification, we can have multiple errors. The | |
90 | * vdev_error() function keeps track of whether we have seen an error yet, and | |
91 | * prints out a header if its the first error we've seen. | |
92 | */ | |
93 | boolean_t error_seen; | |
94 | boolean_t is_force; | |
95 | ||
c06d4368 AX |
96 | typedef struct vdev_disk_db_entry |
97 | { | |
98 | char id[24]; | |
99 | int sector_size; | |
100 | } vdev_disk_db_entry_t; | |
101 | ||
102 | /* | |
103 | * Database of block devices that lie about physical sector sizes. The | |
104 | * identification string must be precisely 24 characters to avoid false | |
105 | * negatives | |
106 | */ | |
107 | static vdev_disk_db_entry_t vdev_disk_database[] = { | |
a08ee875 LG |
108 | {"ATA ADATA SSD S396 3", 8192}, |
109 | {"ATA APPLE SSD SM128E", 8192}, | |
110 | {"ATA APPLE SSD SM256E", 8192}, | |
111 | {"ATA APPLE SSD SM512E", 8192}, | |
112 | {"ATA APPLE SSD SM768E", 8192}, | |
113 | {"ATA C400-MTFDDAC064M", 8192}, | |
114 | {"ATA C400-MTFDDAC128M", 8192}, | |
115 | {"ATA C400-MTFDDAC256M", 8192}, | |
116 | {"ATA C400-MTFDDAC512M", 8192}, | |
c06d4368 | 117 | {"ATA Corsair Force 3 ", 8192}, |
a08ee875 | 118 | {"ATA Corsair Force GS", 8192}, |
c06d4368 | 119 | {"ATA INTEL SSDSA2CT04", 8192}, |
a08ee875 LG |
120 | {"ATA INTEL SSDSA2BZ10", 8192}, |
121 | {"ATA INTEL SSDSA2BZ20", 8192}, | |
122 | {"ATA INTEL SSDSA2BZ30", 8192}, | |
123 | {"ATA INTEL SSDSA2CW04", 8192}, | |
124 | {"ATA INTEL SSDSA2CW08", 8192}, | |
125 | {"ATA INTEL SSDSA2CW12", 8192}, | |
c06d4368 | 126 | {"ATA INTEL SSDSA2CW16", 8192}, |
a08ee875 LG |
127 | {"ATA INTEL SSDSA2CW30", 8192}, |
128 | {"ATA INTEL SSDSA2CW60", 8192}, | |
a08ee875 LG |
129 | {"ATA INTEL SSDSC2CT06", 8192}, |
130 | {"ATA INTEL SSDSC2CT12", 8192}, | |
c06d4368 | 131 | {"ATA INTEL SSDSC2CT18", 8192}, |
a08ee875 LG |
132 | {"ATA INTEL SSDSC2CT24", 8192}, |
133 | {"ATA INTEL SSDSC2CW06", 8192}, | |
c06d4368 | 134 | {"ATA INTEL SSDSC2CW12", 8192}, |
a08ee875 LG |
135 | {"ATA INTEL SSDSC2CW18", 8192}, |
136 | {"ATA INTEL SSDSC2CW24", 8192}, | |
137 | {"ATA INTEL SSDSC2CW48", 8192}, | |
c06d4368 | 138 | {"ATA KINGSTON SH100S3", 8192}, |
a08ee875 | 139 | {"ATA KINGSTON SH103S3", 8192}, |
c06d4368 AX |
140 | {"ATA M4-CT064M4SSD2 ", 8192}, |
141 | {"ATA M4-CT128M4SSD2 ", 8192}, | |
142 | {"ATA M4-CT256M4SSD2 ", 8192}, | |
143 | {"ATA M4-CT512M4SSD2 ", 8192}, | |
144 | {"ATA OCZ-AGILITY2 ", 8192}, | |
a08ee875 | 145 | {"ATA OCZ-AGILITY3 ", 8192}, |
c06d4368 AX |
146 | {"ATA OCZ-VERTEX2 3.5 ", 8192}, |
147 | {"ATA OCZ-VERTEX3 ", 8192}, | |
148 | {"ATA OCZ-VERTEX3 LT ", 8192}, | |
149 | {"ATA OCZ-VERTEX3 MI ", 8192}, | |
a08ee875 LG |
150 | {"ATA OCZ-VERTEX4 ", 8192}, |
151 | {"ATA SAMSUNG MZ7WD120", 8192}, | |
152 | {"ATA SAMSUNG MZ7WD240", 8192}, | |
153 | {"ATA SAMSUNG MZ7WD480", 8192}, | |
154 | {"ATA SAMSUNG MZ7WD960", 8192}, | |
c06d4368 AX |
155 | {"ATA SAMSUNG SSD 830 ", 8192}, |
156 | {"ATA Samsung SSD 840 ", 8192}, | |
a08ee875 LG |
157 | {"ATA SanDisk SSD U100", 8192}, |
158 | {"ATA TOSHIBA THNSNH06", 8192}, | |
159 | {"ATA TOSHIBA THNSNH12", 8192}, | |
160 | {"ATA TOSHIBA THNSNH25", 8192}, | |
161 | {"ATA TOSHIBA THNSNH51", 8192}, | |
162 | {"ATA APPLE SSD TS064C", 4096}, | |
163 | {"ATA APPLE SSD TS128C", 4096}, | |
164 | {"ATA APPLE SSD TS256C", 4096}, | |
165 | {"ATA APPLE SSD TS512C", 4096}, | |
c06d4368 AX |
166 | {"ATA INTEL SSDSA2M040", 4096}, |
167 | {"ATA INTEL SSDSA2M080", 4096}, | |
168 | {"ATA INTEL SSDSA2M160", 4096}, | |
a08ee875 LG |
169 | {"ATA INTEL SSDSC2MH12", 4096}, |
170 | {"ATA INTEL SSDSC2MH25", 4096}, | |
171 | {"ATA OCZ CORE_SSD ", 4096}, | |
172 | {"ATA OCZ-VERTEX ", 4096}, | |
173 | {"ATA SAMSUNG MCCOE32G", 4096}, | |
174 | {"ATA SAMSUNG MCCOE64G", 4096}, | |
175 | {"ATA SAMSUNG SSD PM80", 4096}, | |
ea04106b AX |
176 | /* Flash drives optimized for 4KB IOs on larger pages */ |
177 | {"ATA INTEL SSDSC2BA10", 4096}, | |
178 | {"ATA INTEL SSDSC2BA20", 4096}, | |
179 | {"ATA INTEL SSDSC2BA40", 4096}, | |
180 | {"ATA INTEL SSDSC2BA80", 4096}, | |
181 | {"ATA INTEL SSDSC2BB08", 4096}, | |
182 | {"ATA INTEL SSDSC2BB12", 4096}, | |
183 | {"ATA INTEL SSDSC2BB16", 4096}, | |
184 | {"ATA INTEL SSDSC2BB24", 4096}, | |
185 | {"ATA INTEL SSDSC2BB30", 4096}, | |
186 | {"ATA INTEL SSDSC2BB40", 4096}, | |
187 | {"ATA INTEL SSDSC2BB48", 4096}, | |
188 | {"ATA INTEL SSDSC2BB60", 4096}, | |
189 | {"ATA INTEL SSDSC2BB80", 4096}, | |
190 | {"ATA INTEL SSDSC2BW24", 4096}, | |
191 | {"ATA INTEL SSDSC2BP24", 4096}, | |
192 | {"ATA INTEL SSDSC2BP48", 4096}, | |
193 | {"NA SmrtStorSDLKAE9W", 4096}, | |
a07c8b41 | 194 | {"NVMe Amazon EC2 NVMe ", 4096}, |
a08ee875 | 195 | /* Imported from Open Solaris */ |
c06d4368 AX |
196 | {"ATA MARVELL SD88SA02", 4096}, |
197 | /* Advanced format Hard drives */ | |
198 | {"ATA Hitachi HDS5C303", 4096}, | |
199 | {"ATA SAMSUNG HD204UI ", 4096}, | |
200 | {"ATA ST2000DL004 HD20", 4096}, | |
201 | {"ATA WDC WD10EARS-00M", 4096}, | |
202 | {"ATA WDC WD10EARS-00S", 4096}, | |
203 | {"ATA WDC WD10EARS-00Z", 4096}, | |
204 | {"ATA WDC WD15EARS-00M", 4096}, | |
205 | {"ATA WDC WD15EARS-00S", 4096}, | |
206 | {"ATA WDC WD15EARS-00Z", 4096}, | |
207 | {"ATA WDC WD20EARS-00M", 4096}, | |
208 | {"ATA WDC WD20EARS-00S", 4096}, | |
209 | {"ATA WDC WD20EARS-00Z", 4096}, | |
ea04106b AX |
210 | {"ATA WDC WD1600BEVT-0", 4096}, |
211 | {"ATA WDC WD2500BEVT-0", 4096}, | |
212 | {"ATA WDC WD3200BEVT-0", 4096}, | |
213 | {"ATA WDC WD5000BEVT-0", 4096}, | |
c06d4368 AX |
214 | /* Virtual disks: Assume zvols with default volblocksize */ |
215 | #if 0 | |
216 | {"ATA QEMU HARDDISK ", 8192}, | |
217 | {"IET VIRTUAL-DISK ", 8192}, | |
218 | {"OI COMSTAR ", 8192}, | |
a08ee875 LG |
219 | {"SUN COMSTAR ", 8192}, |
220 | {"NETAPP LUN ", 8192}, | |
c06d4368 AX |
221 | #endif |
222 | }; | |
223 | ||
224 | static const int vdev_disk_database_size = | |
225 | sizeof (vdev_disk_database) / sizeof (vdev_disk_database[0]); | |
226 | ||
227 | #define INQ_REPLY_LEN 96 | |
228 | #define INQ_CMD_LEN 6 | |
229 | ||
230 | static boolean_t | |
231 | check_sector_size_database(char *path, int *sector_size) | |
232 | { | |
233 | unsigned char inq_buff[INQ_REPLY_LEN]; | |
234 | unsigned char sense_buffer[32]; | |
235 | unsigned char inq_cmd_blk[INQ_CMD_LEN] = | |
236 | {INQUIRY, 0, 0, 0, INQ_REPLY_LEN, 0}; | |
237 | sg_io_hdr_t io_hdr; | |
238 | int error; | |
239 | int fd; | |
240 | int i; | |
241 | ||
242 | /* Prepare INQUIRY command */ | |
a08ee875 | 243 | memset(&io_hdr, 0, sizeof (sg_io_hdr_t)); |
c06d4368 | 244 | io_hdr.interface_id = 'S'; |
a08ee875 LG |
245 | io_hdr.cmd_len = sizeof (inq_cmd_blk); |
246 | io_hdr.mx_sb_len = sizeof (sense_buffer); | |
c06d4368 AX |
247 | io_hdr.dxfer_direction = SG_DXFER_FROM_DEV; |
248 | io_hdr.dxfer_len = INQ_REPLY_LEN; | |
249 | io_hdr.dxferp = inq_buff; | |
250 | io_hdr.cmdp = inq_cmd_blk; | |
251 | io_hdr.sbp = sense_buffer; | |
a08ee875 | 252 | io_hdr.timeout = 10; /* 10 milliseconds is ample time */ |
c06d4368 AX |
253 | |
254 | if ((fd = open(path, O_RDONLY|O_DIRECT)) < 0) | |
255 | return (B_FALSE); | |
256 | ||
257 | error = ioctl(fd, SG_IO, (unsigned long) &io_hdr); | |
258 | ||
259 | (void) close(fd); | |
260 | ||
261 | if (error < 0) | |
262 | return (B_FALSE); | |
263 | ||
264 | if ((io_hdr.info & SG_INFO_OK_MASK) != SG_INFO_OK) | |
265 | return (B_FALSE); | |
266 | ||
267 | for (i = 0; i < vdev_disk_database_size; i++) { | |
268 | if (memcmp(inq_buff + 8, vdev_disk_database[i].id, 24)) | |
269 | continue; | |
270 | ||
271 | *sector_size = vdev_disk_database[i].sector_size; | |
272 | return (B_TRUE); | |
273 | } | |
274 | ||
275 | return (B_FALSE); | |
276 | } | |
277 | ||
34dc7c2f BB |
278 | /*PRINTFLIKE1*/ |
279 | static void | |
280 | vdev_error(const char *fmt, ...) | |
281 | { | |
282 | va_list ap; | |
283 | ||
284 | if (!error_seen) { | |
285 | (void) fprintf(stderr, gettext("invalid vdev specification\n")); | |
286 | if (!is_force) | |
287 | (void) fprintf(stderr, gettext("use '-f' to override " | |
288 | "the following errors:\n")); | |
289 | else | |
290 | (void) fprintf(stderr, gettext("the following errors " | |
291 | "must be manually repaired:\n")); | |
292 | error_seen = B_TRUE; | |
293 | } | |
294 | ||
295 | va_start(ap, fmt); | |
296 | (void) vfprintf(stderr, fmt, ap); | |
297 | va_end(ap); | |
298 | } | |
299 | ||
34dc7c2f BB |
300 | /* |
301 | * Check that a file is valid. All we can do in this case is check that it's | |
302 | * not in use by another pool, and not in use by swap. | |
303 | */ | |
304 | static int | |
305 | check_file(const char *file, boolean_t force, boolean_t isspare) | |
306 | { | |
307 | char *name; | |
308 | int fd; | |
309 | int ret = 0; | |
34dc7c2f BB |
310 | pool_state_t state; |
311 | boolean_t inuse; | |
312 | ||
34dc7c2f BB |
313 | if ((fd = open(file, O_RDONLY)) < 0) |
314 | return (0); | |
315 | ||
316 | if (zpool_in_use(g_zfs, fd, &state, &name, &inuse) == 0 && inuse) { | |
317 | const char *desc; | |
318 | ||
319 | switch (state) { | |
320 | case POOL_STATE_ACTIVE: | |
321 | desc = gettext("active"); | |
322 | break; | |
323 | ||
324 | case POOL_STATE_EXPORTED: | |
325 | desc = gettext("exported"); | |
326 | break; | |
327 | ||
328 | case POOL_STATE_POTENTIALLY_ACTIVE: | |
329 | desc = gettext("potentially active"); | |
330 | break; | |
331 | ||
332 | default: | |
333 | desc = gettext("unknown"); | |
334 | break; | |
335 | } | |
336 | ||
337 | /* | |
338 | * Allow hot spares to be shared between pools. | |
339 | */ | |
cae5b340 AX |
340 | if (state == POOL_STATE_SPARE && isspare) { |
341 | free(name); | |
342 | (void) close(fd); | |
34dc7c2f | 343 | return (0); |
cae5b340 | 344 | } |
34dc7c2f BB |
345 | |
346 | if (state == POOL_STATE_ACTIVE || | |
347 | state == POOL_STATE_SPARE || !force) { | |
348 | switch (state) { | |
349 | case POOL_STATE_SPARE: | |
350 | vdev_error(gettext("%s is reserved as a hot " | |
351 | "spare for pool %s\n"), file, name); | |
352 | break; | |
353 | default: | |
354 | vdev_error(gettext("%s is part of %s pool " | |
355 | "'%s'\n"), file, desc, name); | |
356 | break; | |
357 | } | |
358 | ret = -1; | |
359 | } | |
360 | ||
361 | free(name); | |
362 | } | |
363 | ||
364 | (void) close(fd); | |
365 | return (ret); | |
366 | } | |
367 | ||
d603ed6c BB |
368 | static int |
369 | check_slice(const char *path, blkid_cache cache, int force, boolean_t isspare) | |
370 | { | |
d603ed6c | 371 | int err; |
d603ed6c | 372 | char *value; |
d603ed6c | 373 | |
d603ed6c BB |
374 | /* No valid type detected device is safe to use */ |
375 | value = blkid_get_tag_value(cache, "TYPE", path); | |
376 | if (value == NULL) | |
377 | return (0); | |
378 | ||
379 | /* | |
380 | * If libblkid detects a ZFS device, we check the device | |
381 | * using check_file() to see if it's safe. The one safe | |
382 | * case is a spare device shared between multiple pools. | |
383 | */ | |
ea04106b | 384 | if (strcmp(value, "zfs_member") == 0) { |
d603ed6c BB |
385 | err = check_file(path, force, isspare); |
386 | } else { | |
387 | if (force) { | |
388 | err = 0; | |
389 | } else { | |
390 | err = -1; | |
391 | vdev_error(gettext("%s contains a filesystem of " | |
a08ee875 | 392 | "type '%s'\n"), path, value); |
d603ed6c BB |
393 | } |
394 | } | |
395 | ||
396 | free(value); | |
d603ed6c BB |
397 | |
398 | return (err); | |
399 | } | |
400 | ||
401 | /* | |
cae5b340 AX |
402 | * Validate that a disk including all partitions are safe to use. |
403 | * | |
404 | * For EFI labeled disks this can done relatively easily with the libefi | |
405 | * library. The partition numbers are extracted from the label and used | |
406 | * to generate the expected /dev/ paths. Each partition can then be | |
407 | * checked for conflicts. | |
408 | * | |
409 | * For non-EFI labeled disks (MBR/EBR/etc) the same process is possible | |
410 | * but due to the lack of a readily available libraries this scanning is | |
411 | * not implemented. Instead only the device path as given is checked. | |
d603ed6c BB |
412 | */ |
413 | static int | |
414 | check_disk(const char *path, blkid_cache cache, int force, | |
a08ee875 | 415 | boolean_t isspare, boolean_t iswholedisk) |
d603ed6c BB |
416 | { |
417 | struct dk_gpt *vtoc; | |
418 | char slice_path[MAXPATHLEN]; | |
419 | int err = 0; | |
420 | int fd, i; | |
421 | ||
d603ed6c | 422 | if (!iswholedisk) |
a08ee875 | 423 | return (check_slice(path, cache, force, isspare)); |
d603ed6c | 424 | |
cae5b340 AX |
425 | if ((fd = open(path, O_RDONLY|O_DIRECT|O_EXCL)) < 0) { |
426 | char *value = blkid_get_tag_value(cache, "TYPE", path); | |
427 | (void) fprintf(stderr, gettext("%s is in use and contains " | |
428 | "a %s filesystem.\n"), path, value ? value : "unknown"); | |
a08ee875 | 429 | return (-1); |
d603ed6c BB |
430 | } |
431 | ||
cae5b340 AX |
432 | /* |
433 | * Expected to fail for non-EFI labled disks. Just check the device | |
434 | * as given and do not attempt to detect and scan partitions. | |
435 | */ | |
436 | err = efi_alloc_and_read(fd, &vtoc); | |
437 | if (err) { | |
d603ed6c | 438 | (void) close(fd); |
cae5b340 | 439 | return (check_slice(path, cache, force, isspare)); |
d603ed6c BB |
440 | } |
441 | ||
442 | /* | |
443 | * The primary efi partition label is damaged however the secondary | |
444 | * label at the end of the device is intact. Rather than use this | |
445 | * label we should play it safe and treat this as a non efi device. | |
446 | */ | |
447 | if (vtoc->efi_flags & EFI_GPT_PRIMARY_CORRUPT) { | |
448 | efi_free(vtoc); | |
449 | (void) close(fd); | |
450 | ||
451 | if (force) { | |
cae5b340 | 452 | /* Partitions will now be created using the backup */ |
a08ee875 | 453 | return (0); |
d603ed6c BB |
454 | } else { |
455 | vdev_error(gettext("%s contains a corrupt primary " | |
456 | "EFI label.\n"), path); | |
a08ee875 | 457 | return (-1); |
d603ed6c BB |
458 | } |
459 | } | |
460 | ||
461 | for (i = 0; i < vtoc->efi_nparts; i++) { | |
462 | ||
463 | if (vtoc->efi_parts[i].p_tag == V_UNASSIGNED || | |
464 | uuid_is_null((uchar_t *)&vtoc->efi_parts[i].p_guid)) | |
465 | continue; | |
466 | ||
467 | if (strncmp(path, UDISK_ROOT, strlen(UDISK_ROOT)) == 0) | |
468 | (void) snprintf(slice_path, sizeof (slice_path), | |
469 | "%s%s%d", path, "-part", i+1); | |
470 | else | |
471 | (void) snprintf(slice_path, sizeof (slice_path), | |
472 | "%s%s%d", path, isdigit(path[strlen(path)-1]) ? | |
473 | "p" : "", i+1); | |
474 | ||
475 | err = check_slice(slice_path, cache, force, isspare); | |
476 | if (err) | |
477 | break; | |
478 | } | |
479 | ||
480 | efi_free(vtoc); | |
481 | (void) close(fd); | |
482 | ||
8128bd89 | 483 | return (err); |
d603ed6c BB |
484 | } |
485 | ||
486 | static int | |
487 | check_device(const char *path, boolean_t force, | |
a08ee875 | 488 | boolean_t isspare, boolean_t iswholedisk) |
d603ed6c | 489 | { |
cae5b340 AX |
490 | blkid_cache cache; |
491 | int error; | |
d603ed6c | 492 | |
cae5b340 AX |
493 | error = blkid_get_cache(&cache, NULL); |
494 | if (error != 0) { | |
495 | (void) fprintf(stderr, gettext("unable to access the blkid " | |
496 | "cache.\n")); | |
497 | return (-1); | |
d603ed6c | 498 | } |
34dc7c2f | 499 | |
cae5b340 AX |
500 | error = check_disk(path, cache, force, isspare, iswholedisk); |
501 | blkid_put_cache(cache); | |
34dc7c2f | 502 | |
cae5b340 | 503 | return (error); |
34dc7c2f BB |
504 | } |
505 | ||
d603ed6c BB |
506 | /* |
507 | * This may be a shorthand device path or it could be total gibberish. | |
eac47204 BB |
508 | * Check to see if it is a known device available in zfs_vdev_paths. |
509 | * As part of this check, see if we've been given an entire disk | |
510 | * (minus the slice number). | |
d603ed6c BB |
511 | */ |
512 | static int | |
cae5b340 | 513 | is_shorthand_path(const char *arg, char *path, size_t path_size, |
a08ee875 | 514 | struct stat64 *statbuf, boolean_t *wholedisk) |
d603ed6c | 515 | { |
eac47204 BB |
516 | int error; |
517 | ||
cae5b340 | 518 | error = zfs_resolve_shortname(arg, path, path_size); |
eac47204 | 519 | if (error == 0) { |
cae5b340 | 520 | *wholedisk = zfs_dev_is_whole_disk(path); |
79e7242a | 521 | if (*wholedisk || (stat64(path, statbuf) == 0)) |
d603ed6c BB |
522 | return (0); |
523 | } | |
524 | ||
cae5b340 | 525 | strlcpy(path, arg, path_size); |
a08ee875 | 526 | memset(statbuf, 0, sizeof (*statbuf)); |
d603ed6c BB |
527 | *wholedisk = B_FALSE; |
528 | ||
eac47204 | 529 | return (error); |
d603ed6c BB |
530 | } |
531 | ||
8128bd89 BB |
532 | /* |
533 | * Determine if the given path is a hot spare within the given configuration. | |
534 | * If no configuration is given we rely solely on the label. | |
535 | */ | |
536 | static boolean_t | |
537 | is_spare(nvlist_t *config, const char *path) | |
538 | { | |
539 | int fd; | |
540 | pool_state_t state; | |
541 | char *name = NULL; | |
542 | nvlist_t *label; | |
543 | uint64_t guid, spareguid; | |
544 | nvlist_t *nvroot; | |
545 | nvlist_t **spares; | |
546 | uint_t i, nspares; | |
547 | boolean_t inuse; | |
548 | ||
549 | if ((fd = open(path, O_RDONLY)) < 0) | |
550 | return (B_FALSE); | |
551 | ||
552 | if (zpool_in_use(g_zfs, fd, &state, &name, &inuse) != 0 || | |
553 | !inuse || | |
554 | state != POOL_STATE_SPARE || | |
ea04106b | 555 | zpool_read_label(fd, &label, NULL) != 0) { |
8128bd89 BB |
556 | free(name); |
557 | (void) close(fd); | |
558 | return (B_FALSE); | |
559 | } | |
560 | free(name); | |
561 | (void) close(fd); | |
562 | ||
cae5b340 AX |
563 | if (config == NULL) { |
564 | nvlist_free(label); | |
8128bd89 | 565 | return (B_TRUE); |
cae5b340 | 566 | } |
8128bd89 BB |
567 | |
568 | verify(nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID, &guid) == 0); | |
569 | nvlist_free(label); | |
570 | ||
571 | verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, | |
572 | &nvroot) == 0); | |
573 | if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, | |
574 | &spares, &nspares) == 0) { | |
575 | for (i = 0; i < nspares; i++) { | |
576 | verify(nvlist_lookup_uint64(spares[i], | |
577 | ZPOOL_CONFIG_GUID, &spareguid) == 0); | |
578 | if (spareguid == guid) | |
579 | return (B_TRUE); | |
580 | } | |
581 | } | |
582 | ||
583 | return (B_FALSE); | |
584 | } | |
585 | ||
34dc7c2f BB |
586 | /* |
587 | * Create a leaf vdev. Determine if this is a file or a device. If it's a | |
588 | * device, fill in the device id to make a complete nvlist. Valid forms for a | |
589 | * leaf vdev are: | |
590 | * | |
eac47204 BB |
591 | * /dev/xxx Complete disk path |
592 | * /xxx Full path to file | |
593 | * xxx Shorthand for <zfs_vdev_paths>/xxx | |
34dc7c2f BB |
594 | */ |
595 | static nvlist_t * | |
df30f566 | 596 | make_leaf_vdev(nvlist_t *props, const char *arg, uint64_t is_log) |
34dc7c2f BB |
597 | { |
598 | char path[MAXPATHLEN]; | |
599 | struct stat64 statbuf; | |
600 | nvlist_t *vdev = NULL; | |
601 | char *type = NULL; | |
602 | boolean_t wholedisk = B_FALSE; | |
c06d4368 | 603 | uint64_t ashift = 0; |
d603ed6c | 604 | int err; |
34dc7c2f BB |
605 | |
606 | /* | |
607 | * Determine what type of vdev this is, and put the full path into | |
608 | * 'path'. We detect whether this is a device of file afterwards by | |
609 | * checking the st_mode of the file. | |
610 | */ | |
611 | if (arg[0] == '/') { | |
612 | /* | |
613 | * Complete device or file path. Exact type is determined by | |
d603ed6c | 614 | * examining the file descriptor afterwards. Symbolic links |
cae5b340 | 615 | * are resolved to their real paths to determine whole disk |
d603ed6c BB |
616 | * and S_ISBLK/S_ISREG type checks. However, we are careful |
617 | * to store the given path as ZPOOL_CONFIG_PATH to ensure we | |
618 | * can leverage udev's persistent device labels. | |
34dc7c2f | 619 | */ |
d603ed6c | 620 | if (realpath(arg, path) == NULL) { |
34dc7c2f | 621 | (void) fprintf(stderr, |
d603ed6c | 622 | gettext("cannot resolve path '%s'\n"), arg); |
34dc7c2f BB |
623 | return (NULL); |
624 | } | |
625 | ||
cae5b340 | 626 | wholedisk = zfs_dev_is_whole_disk(path); |
34dc7c2f | 627 | if (!wholedisk && (stat64(path, &statbuf) != 0)) { |
d603ed6c BB |
628 | (void) fprintf(stderr, |
629 | gettext("cannot open '%s': %s\n"), | |
630 | path, strerror(errno)); | |
631 | return (NULL); | |
632 | } | |
633 | ||
cae5b340 AX |
634 | /* After whole disk check restore original passed path */ |
635 | strlcpy(path, arg, sizeof (path)); | |
d603ed6c | 636 | } else { |
cae5b340 AX |
637 | err = is_shorthand_path(arg, path, sizeof (path), |
638 | &statbuf, &wholedisk); | |
d603ed6c | 639 | if (err != 0) { |
34dc7c2f BB |
640 | /* |
641 | * If we got ENOENT, then the user gave us | |
642 | * gibberish, so try to direct them with a | |
643 | * reasonable error message. Otherwise, | |
644 | * regurgitate strerror() since it's the best we | |
645 | * can do. | |
646 | */ | |
d603ed6c | 647 | if (err == ENOENT) { |
34dc7c2f BB |
648 | (void) fprintf(stderr, |
649 | gettext("cannot open '%s': no such " | |
650 | "device in %s\n"), arg, DISK_ROOT); | |
651 | (void) fprintf(stderr, | |
652 | gettext("must be a full path or " | |
653 | "shorthand device name\n")); | |
654 | return (NULL); | |
655 | } else { | |
656 | (void) fprintf(stderr, | |
657 | gettext("cannot open '%s': %s\n"), | |
658 | path, strerror(errno)); | |
659 | return (NULL); | |
660 | } | |
661 | } | |
662 | } | |
663 | ||
664 | /* | |
665 | * Determine whether this is a device or a file. | |
666 | */ | |
667 | if (wholedisk || S_ISBLK(statbuf.st_mode)) { | |
668 | type = VDEV_TYPE_DISK; | |
669 | } else if (S_ISREG(statbuf.st_mode)) { | |
670 | type = VDEV_TYPE_FILE; | |
671 | } else { | |
672 | (void) fprintf(stderr, gettext("cannot use '%s': must be a " | |
673 | "block device or regular file\n"), path); | |
674 | return (NULL); | |
675 | } | |
676 | ||
677 | /* | |
678 | * Finally, we have the complete device or file, and we know that it is | |
679 | * acceptable to use. Construct the nvlist to describe this vdev. All | |
680 | * vdevs have a 'path' element, and devices also have a 'devid' element. | |
681 | */ | |
682 | verify(nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) == 0); | |
683 | verify(nvlist_add_string(vdev, ZPOOL_CONFIG_PATH, path) == 0); | |
684 | verify(nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE, type) == 0); | |
685 | verify(nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_LOG, is_log) == 0); | |
686 | if (strcmp(type, VDEV_TYPE_DISK) == 0) | |
687 | verify(nvlist_add_uint64(vdev, ZPOOL_CONFIG_WHOLE_DISK, | |
688 | (uint64_t)wholedisk) == 0); | |
689 | ||
c06d4368 AX |
690 | /* |
691 | * Override defaults if custom properties are provided. | |
692 | */ | |
df30f566 | 693 | if (props != NULL) { |
df30f566 CK |
694 | char *value = NULL; |
695 | ||
696 | if (nvlist_lookup_string(props, | |
cae5b340 AX |
697 | zpool_prop_to_name(ZPOOL_PROP_ASHIFT), &value) == 0) { |
698 | if (zfs_nicestrtonum(NULL, value, &ashift) != 0) { | |
699 | (void) fprintf(stderr, | |
700 | gettext("ashift must be a number.\n")); | |
701 | return (NULL); | |
702 | } | |
703 | if (ashift != 0 && | |
704 | (ashift < ASHIFT_MIN || ashift > ASHIFT_MAX)) { | |
705 | (void) fprintf(stderr, | |
706 | gettext("invalid 'ashift=%" PRIu64 "' " | |
707 | "property: only values between %" PRId32 " " | |
708 | "and %" PRId32 " are allowed.\n"), | |
709 | ashift, ASHIFT_MIN, ASHIFT_MAX); | |
710 | return (NULL); | |
711 | } | |
712 | } | |
c06d4368 | 713 | } |
df30f566 | 714 | |
c06d4368 AX |
715 | /* |
716 | * If the device is known to incorrectly report its physical sector | |
717 | * size explicitly provide the known correct value. | |
718 | */ | |
719 | if (ashift == 0) { | |
720 | int sector_size; | |
721 | ||
722 | if (check_sector_size_database(path, §or_size) == B_TRUE) | |
ea04106b | 723 | ashift = highbit64(sector_size) - 1; |
df30f566 CK |
724 | } |
725 | ||
c06d4368 | 726 | if (ashift > 0) |
cae5b340 | 727 | (void) nvlist_add_uint64(vdev, ZPOOL_CONFIG_ASHIFT, ashift); |
c06d4368 | 728 | |
34dc7c2f BB |
729 | return (vdev); |
730 | } | |
731 | ||
732 | /* | |
733 | * Go through and verify the replication level of the pool is consistent. | |
734 | * Performs the following checks: | |
735 | * | |
736 | * For the new spec, verifies that devices in mirrors and raidz are the | |
737 | * same size. | |
738 | * | |
739 | * If the current configuration already has inconsistent replication | |
740 | * levels, ignore any other potential problems in the new spec. | |
741 | * | |
742 | * Otherwise, make sure that the current spec (if there is one) and the new | |
743 | * spec have consistent replication levels. | |
744 | */ | |
745 | typedef struct replication_level { | |
746 | char *zprl_type; | |
747 | uint64_t zprl_children; | |
748 | uint64_t zprl_parity; | |
749 | } replication_level_t; | |
750 | ||
751 | #define ZPOOL_FUZZ (16 * 1024 * 1024) | |
752 | ||
cae5b340 AX |
753 | static boolean_t |
754 | is_raidz_mirror(replication_level_t *a, replication_level_t *b, | |
755 | replication_level_t **raidz, replication_level_t **mirror) | |
756 | { | |
757 | if (strcmp(a->zprl_type, "raidz") == 0 && | |
758 | strcmp(b->zprl_type, "mirror") == 0) { | |
759 | *raidz = a; | |
760 | *mirror = b; | |
761 | return (B_TRUE); | |
762 | } | |
763 | return (B_FALSE); | |
764 | } | |
765 | ||
34dc7c2f BB |
766 | /* |
767 | * Given a list of toplevel vdevs, return the current replication level. If | |
768 | * the config is inconsistent, then NULL is returned. If 'fatal' is set, then | |
769 | * an error message will be displayed for each self-inconsistent vdev. | |
770 | */ | |
771 | static replication_level_t * | |
772 | get_replication(nvlist_t *nvroot, boolean_t fatal) | |
773 | { | |
774 | nvlist_t **top; | |
775 | uint_t t, toplevels; | |
776 | nvlist_t **child; | |
777 | uint_t c, children; | |
778 | nvlist_t *nv; | |
779 | char *type; | |
cae5b340 AX |
780 | replication_level_t lastrep = {0}; |
781 | replication_level_t rep; | |
782 | replication_level_t *ret; | |
783 | replication_level_t *raidz, *mirror; | |
34dc7c2f BB |
784 | boolean_t dontreport; |
785 | ||
786 | ret = safe_malloc(sizeof (replication_level_t)); | |
787 | ||
788 | verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, | |
789 | &top, &toplevels) == 0); | |
790 | ||
34dc7c2f BB |
791 | for (t = 0; t < toplevels; t++) { |
792 | uint64_t is_log = B_FALSE; | |
793 | ||
794 | nv = top[t]; | |
795 | ||
796 | /* | |
797 | * For separate logs we ignore the top level vdev replication | |
798 | * constraints. | |
799 | */ | |
800 | (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG, &is_log); | |
801 | if (is_log) | |
802 | continue; | |
803 | ||
41d74433 AX |
804 | /* Ignore holes introduced by removing aux devices */ |
805 | verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0); | |
806 | if (strcmp(type, VDEV_TYPE_HOLE) == 0) | |
807 | continue; | |
808 | ||
34dc7c2f BB |
809 | if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, |
810 | &child, &children) != 0) { | |
811 | /* | |
812 | * This is a 'file' or 'disk' vdev. | |
813 | */ | |
814 | rep.zprl_type = type; | |
815 | rep.zprl_children = 1; | |
816 | rep.zprl_parity = 0; | |
817 | } else { | |
818 | uint64_t vdev_size; | |
819 | ||
820 | /* | |
821 | * This is a mirror or RAID-Z vdev. Go through and make | |
822 | * sure the contents are all the same (files vs. disks), | |
823 | * keeping track of the number of elements in the | |
824 | * process. | |
825 | * | |
826 | * We also check that the size of each vdev (if it can | |
827 | * be determined) is the same. | |
828 | */ | |
829 | rep.zprl_type = type; | |
830 | rep.zprl_children = 0; | |
831 | ||
832 | if (strcmp(type, VDEV_TYPE_RAIDZ) == 0) { | |
833 | verify(nvlist_lookup_uint64(nv, | |
834 | ZPOOL_CONFIG_NPARITY, | |
835 | &rep.zprl_parity) == 0); | |
836 | assert(rep.zprl_parity != 0); | |
837 | } else { | |
838 | rep.zprl_parity = 0; | |
839 | } | |
840 | ||
841 | /* | |
842 | * The 'dontreport' variable indicates that we've | |
843 | * already reported an error for this spec, so don't | |
844 | * bother doing it again. | |
845 | */ | |
846 | type = NULL; | |
847 | dontreport = 0; | |
848 | vdev_size = -1ULL; | |
849 | for (c = 0; c < children; c++) { | |
850 | nvlist_t *cnv = child[c]; | |
851 | char *path; | |
852 | struct stat64 statbuf; | |
853 | uint64_t size = -1ULL; | |
854 | char *childtype; | |
855 | int fd, err; | |
856 | ||
857 | rep.zprl_children++; | |
858 | ||
859 | verify(nvlist_lookup_string(cnv, | |
860 | ZPOOL_CONFIG_TYPE, &childtype) == 0); | |
861 | ||
862 | /* | |
863 | * If this is a replacing or spare vdev, then | |
047218e2 AX |
864 | * get the real first child of the vdev: do this |
865 | * in a loop because replacing and spare vdevs | |
866 | * can be nested. | |
34dc7c2f | 867 | */ |
047218e2 | 868 | while (strcmp(childtype, |
34dc7c2f BB |
869 | VDEV_TYPE_REPLACING) == 0 || |
870 | strcmp(childtype, VDEV_TYPE_SPARE) == 0) { | |
871 | nvlist_t **rchild; | |
872 | uint_t rchildren; | |
873 | ||
874 | verify(nvlist_lookup_nvlist_array(cnv, | |
875 | ZPOOL_CONFIG_CHILDREN, &rchild, | |
876 | &rchildren) == 0); | |
877 | assert(rchildren == 2); | |
878 | cnv = rchild[0]; | |
879 | ||
880 | verify(nvlist_lookup_string(cnv, | |
881 | ZPOOL_CONFIG_TYPE, | |
882 | &childtype) == 0); | |
883 | } | |
884 | ||
885 | verify(nvlist_lookup_string(cnv, | |
886 | ZPOOL_CONFIG_PATH, &path) == 0); | |
887 | ||
888 | /* | |
889 | * If we have a raidz/mirror that combines disks | |
890 | * with files, report it as an error. | |
891 | */ | |
892 | if (!dontreport && type != NULL && | |
893 | strcmp(type, childtype) != 0) { | |
894 | if (ret != NULL) | |
895 | free(ret); | |
896 | ret = NULL; | |
897 | if (fatal) | |
898 | vdev_error(gettext( | |
899 | "mismatched replication " | |
900 | "level: %s contains both " | |
901 | "files and devices\n"), | |
902 | rep.zprl_type); | |
903 | else | |
904 | return (NULL); | |
905 | dontreport = B_TRUE; | |
906 | } | |
907 | ||
908 | /* | |
909 | * According to stat(2), the value of 'st_size' | |
910 | * is undefined for block devices and character | |
911 | * devices. But there is no effective way to | |
912 | * determine the real size in userland. | |
913 | * | |
914 | * Instead, we'll take advantage of an | |
915 | * implementation detail of spec_size(). If the | |
916 | * device is currently open, then we (should) | |
917 | * return a valid size. | |
918 | * | |
919 | * If we still don't get a valid size (indicated | |
920 | * by a size of 0 or MAXOFFSET_T), then ignore | |
921 | * this device altogether. | |
922 | */ | |
923 | if ((fd = open(path, O_RDONLY)) >= 0) { | |
cae5b340 | 924 | err = fstat64_blk(fd, &statbuf); |
34dc7c2f BB |
925 | (void) close(fd); |
926 | } else { | |
927 | err = stat64(path, &statbuf); | |
928 | } | |
929 | ||
930 | if (err != 0 || | |
931 | statbuf.st_size == 0 || | |
932 | statbuf.st_size == MAXOFFSET_T) | |
933 | continue; | |
934 | ||
935 | size = statbuf.st_size; | |
936 | ||
937 | /* | |
938 | * Also make sure that devices and | |
939 | * slices have a consistent size. If | |
940 | * they differ by a significant amount | |
941 | * (~16MB) then report an error. | |
942 | */ | |
943 | if (!dontreport && | |
944 | (vdev_size != -1ULL && | |
945 | (labs(size - vdev_size) > | |
946 | ZPOOL_FUZZ))) { | |
947 | if (ret != NULL) | |
948 | free(ret); | |
949 | ret = NULL; | |
950 | if (fatal) | |
951 | vdev_error(gettext( | |
952 | "%s contains devices of " | |
953 | "different sizes\n"), | |
954 | rep.zprl_type); | |
955 | else | |
956 | return (NULL); | |
957 | dontreport = B_TRUE; | |
958 | } | |
959 | ||
960 | type = childtype; | |
961 | vdev_size = size; | |
962 | } | |
963 | } | |
964 | ||
965 | /* | |
966 | * At this point, we have the replication of the last toplevel | |
967 | * vdev in 'rep'. Compare it to 'lastrep' to see if its | |
968 | * different. | |
969 | */ | |
970 | if (lastrep.zprl_type != NULL) { | |
cae5b340 AX |
971 | if (is_raidz_mirror(&lastrep, &rep, &raidz, &mirror) || |
972 | is_raidz_mirror(&rep, &lastrep, &raidz, &mirror)) { | |
973 | /* | |
974 | * Accepted raidz and mirror when they can | |
975 | * handle the same number of disk failures. | |
976 | */ | |
977 | if (raidz->zprl_parity != | |
978 | mirror->zprl_children - 1) { | |
979 | if (ret != NULL) | |
980 | free(ret); | |
981 | ret = NULL; | |
982 | if (fatal) | |
983 | vdev_error(gettext( | |
984 | "mismatched replication " | |
985 | "level: " | |
986 | "%s and %s vdevs with " | |
987 | "different redundancy, " | |
988 | "%llu vs. %llu (%llu-way) " | |
989 | "are present\n"), | |
990 | raidz->zprl_type, | |
991 | mirror->zprl_type, | |
992 | raidz->zprl_parity, | |
993 | mirror->zprl_children - 1, | |
994 | mirror->zprl_children); | |
995 | else | |
996 | return (NULL); | |
997 | } | |
998 | } else if (strcmp(lastrep.zprl_type, rep.zprl_type) != | |
999 | 0) { | |
34dc7c2f BB |
1000 | if (ret != NULL) |
1001 | free(ret); | |
1002 | ret = NULL; | |
1003 | if (fatal) | |
1004 | vdev_error(gettext( | |
1005 | "mismatched replication level: " | |
1006 | "both %s and %s vdevs are " | |
1007 | "present\n"), | |
1008 | lastrep.zprl_type, rep.zprl_type); | |
1009 | else | |
1010 | return (NULL); | |
1011 | } else if (lastrep.zprl_parity != rep.zprl_parity) { | |
1012 | if (ret) | |
1013 | free(ret); | |
1014 | ret = NULL; | |
1015 | if (fatal) | |
1016 | vdev_error(gettext( | |
1017 | "mismatched replication level: " | |
1018 | "both %llu and %llu device parity " | |
1019 | "%s vdevs are present\n"), | |
1020 | lastrep.zprl_parity, | |
1021 | rep.zprl_parity, | |
1022 | rep.zprl_type); | |
1023 | else | |
1024 | return (NULL); | |
1025 | } else if (lastrep.zprl_children != rep.zprl_children) { | |
1026 | if (ret) | |
1027 | free(ret); | |
1028 | ret = NULL; | |
1029 | if (fatal) | |
1030 | vdev_error(gettext( | |
1031 | "mismatched replication level: " | |
1032 | "both %llu-way and %llu-way %s " | |
1033 | "vdevs are present\n"), | |
1034 | lastrep.zprl_children, | |
1035 | rep.zprl_children, | |
1036 | rep.zprl_type); | |
1037 | else | |
1038 | return (NULL); | |
1039 | } | |
1040 | } | |
1041 | lastrep = rep; | |
1042 | } | |
1043 | ||
1044 | if (ret != NULL) | |
1045 | *ret = rep; | |
1046 | ||
1047 | return (ret); | |
1048 | } | |
1049 | ||
1050 | /* | |
1051 | * Check the replication level of the vdev spec against the current pool. Calls | |
1052 | * get_replication() to make sure the new spec is self-consistent. If the pool | |
1053 | * has a consistent replication level, then we ignore any errors. Otherwise, | |
1054 | * report any difference between the two. | |
1055 | */ | |
1056 | static int | |
1057 | check_replication(nvlist_t *config, nvlist_t *newroot) | |
1058 | { | |
1059 | nvlist_t **child; | |
1060 | uint_t children; | |
1061 | replication_level_t *current = NULL, *new; | |
cae5b340 | 1062 | replication_level_t *raidz, *mirror; |
34dc7c2f BB |
1063 | int ret; |
1064 | ||
1065 | /* | |
1066 | * If we have a current pool configuration, check to see if it's | |
1067 | * self-consistent. If not, simply return success. | |
1068 | */ | |
1069 | if (config != NULL) { | |
1070 | nvlist_t *nvroot; | |
1071 | ||
1072 | verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, | |
1073 | &nvroot) == 0); | |
1074 | if ((current = get_replication(nvroot, B_FALSE)) == NULL) | |
1075 | return (0); | |
1076 | } | |
1077 | /* | |
1078 | * for spares there may be no children, and therefore no | |
1079 | * replication level to check | |
1080 | */ | |
1081 | if ((nvlist_lookup_nvlist_array(newroot, ZPOOL_CONFIG_CHILDREN, | |
1082 | &child, &children) != 0) || (children == 0)) { | |
1083 | free(current); | |
1084 | return (0); | |
1085 | } | |
1086 | ||
1087 | /* | |
1088 | * If all we have is logs then there's no replication level to check. | |
1089 | */ | |
1090 | if (num_logs(newroot) == children) { | |
1091 | free(current); | |
1092 | return (0); | |
1093 | } | |
1094 | ||
1095 | /* | |
1096 | * Get the replication level of the new vdev spec, reporting any | |
1097 | * inconsistencies found. | |
1098 | */ | |
1099 | if ((new = get_replication(newroot, B_TRUE)) == NULL) { | |
1100 | free(current); | |
1101 | return (-1); | |
1102 | } | |
1103 | ||
1104 | /* | |
1105 | * Check to see if the new vdev spec matches the replication level of | |
1106 | * the current pool. | |
1107 | */ | |
1108 | ret = 0; | |
1109 | if (current != NULL) { | |
cae5b340 AX |
1110 | if (is_raidz_mirror(current, new, &raidz, &mirror) || |
1111 | is_raidz_mirror(new, current, &raidz, &mirror)) { | |
1112 | if (raidz->zprl_parity != mirror->zprl_children - 1) { | |
1113 | vdev_error(gettext( | |
1114 | "mismatched replication level: pool and " | |
1115 | "new vdev with different redundancy, %s " | |
1116 | "and %s vdevs, %llu vs. %llu (%llu-way)\n"), | |
1117 | raidz->zprl_type, | |
1118 | mirror->zprl_type, | |
1119 | raidz->zprl_parity, | |
1120 | mirror->zprl_children - 1, | |
1121 | mirror->zprl_children); | |
1122 | ret = -1; | |
1123 | } | |
1124 | } else if (strcmp(current->zprl_type, new->zprl_type) != 0) { | |
34dc7c2f BB |
1125 | vdev_error(gettext( |
1126 | "mismatched replication level: pool uses %s " | |
1127 | "and new vdev is %s\n"), | |
1128 | current->zprl_type, new->zprl_type); | |
1129 | ret = -1; | |
1130 | } else if (current->zprl_parity != new->zprl_parity) { | |
1131 | vdev_error(gettext( | |
1132 | "mismatched replication level: pool uses %llu " | |
1133 | "device parity and new vdev uses %llu\n"), | |
1134 | current->zprl_parity, new->zprl_parity); | |
1135 | ret = -1; | |
1136 | } else if (current->zprl_children != new->zprl_children) { | |
1137 | vdev_error(gettext( | |
1138 | "mismatched replication level: pool uses %llu-way " | |
1139 | "%s and new vdev uses %llu-way %s\n"), | |
1140 | current->zprl_children, current->zprl_type, | |
1141 | new->zprl_children, new->zprl_type); | |
1142 | ret = -1; | |
1143 | } | |
1144 | } | |
1145 | ||
1146 | free(new); | |
1147 | if (current != NULL) | |
1148 | free(current); | |
1149 | ||
1150 | return (ret); | |
1151 | } | |
1152 | ||
d603ed6c BB |
1153 | static int |
1154 | zero_label(char *path) | |
1155 | { | |
1156 | const int size = 4096; | |
1157 | char buf[size]; | |
1158 | int err, fd; | |
1159 | ||
1160 | if ((fd = open(path, O_WRONLY|O_EXCL)) < 0) { | |
1161 | (void) fprintf(stderr, gettext("cannot open '%s': %s\n"), | |
1162 | path, strerror(errno)); | |
1163 | return (-1); | |
1164 | } | |
1165 | ||
1166 | memset(buf, 0, size); | |
1167 | err = write(fd, buf, size); | |
1168 | (void) fdatasync(fd); | |
1169 | (void) close(fd); | |
1170 | ||
1171 | if (err == -1) { | |
1172 | (void) fprintf(stderr, gettext("cannot zero first %d bytes " | |
1173 | "of '%s': %s\n"), size, path, strerror(errno)); | |
1174 | return (-1); | |
1175 | } | |
1176 | ||
1177 | if (err != size) { | |
1178 | (void) fprintf(stderr, gettext("could only zero %d/%d bytes " | |
1179 | "of '%s'\n"), err, size, path); | |
1180 | return (-1); | |
1181 | } | |
1182 | ||
a08ee875 | 1183 | return (0); |
d603ed6c BB |
1184 | } |
1185 | ||
34dc7c2f BB |
1186 | /* |
1187 | * Go through and find any whole disks in the vdev specification, labelling them | |
1188 | * as appropriate. When constructing the vdev spec, we were unable to open this | |
1189 | * device in order to provide a devid. Now that we have labelled the disk and | |
1190 | * know that slice 0 is valid, we can construct the devid now. | |
1191 | * | |
1192 | * If the disk was already labeled with an EFI label, we will have gotten the | |
1193 | * devid already (because we were able to open the whole disk). Otherwise, we | |
1194 | * need to get the devid after we label the disk. | |
1195 | */ | |
1196 | static int | |
1197 | make_disks(zpool_handle_t *zhp, nvlist_t *nv) | |
1198 | { | |
1199 | nvlist_t **child; | |
1200 | uint_t c, children; | |
8128bd89 | 1201 | char *type, *path; |
d877ac6b NB |
1202 | char devpath[MAXPATHLEN]; |
1203 | char udevpath[MAXPATHLEN]; | |
34dc7c2f | 1204 | uint64_t wholedisk; |
d877ac6b | 1205 | struct stat64 statbuf; |
8128bd89 BB |
1206 | int is_exclusive = 0; |
1207 | int fd; | |
34dc7c2f | 1208 | int ret; |
34dc7c2f BB |
1209 | |
1210 | verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0); | |
1211 | ||
1212 | if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, | |
1213 | &child, &children) != 0) { | |
1214 | ||
1215 | if (strcmp(type, VDEV_TYPE_DISK) != 0) | |
1216 | return (0); | |
1217 | ||
1218 | /* | |
d603ed6c BB |
1219 | * We have a disk device. If this is a whole disk write |
1220 | * out the efi partition table, otherwise write zero's to | |
1221 | * the first 4k of the partition. This is to ensure that | |
1222 | * libblkid will not misidentify the partition due to a | |
1223 | * magic value left by the previous filesystem. | |
34dc7c2f | 1224 | */ |
d603ed6c BB |
1225 | verify(!nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path)); |
1226 | verify(!nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, | |
1227 | &wholedisk)); | |
1228 | ||
1229 | if (!wholedisk) { | |
cae5b340 AX |
1230 | /* |
1231 | * Update device id string for mpath nodes (Linux only) | |
1232 | */ | |
1233 | if (is_mpath_whole_disk(path)) | |
1234 | update_vdev_config_dev_strs(nv); | |
1235 | ||
1236 | if (!is_spare(NULL, path)) | |
1237 | (void) zero_label(path); | |
8128bd89 | 1238 | return (0); |
d603ed6c BB |
1239 | } |
1240 | ||
d877ac6b | 1241 | if (realpath(path, devpath) == NULL) { |
d603ed6c BB |
1242 | ret = errno; |
1243 | (void) fprintf(stderr, | |
1244 | gettext("cannot resolve path '%s'\n"), path); | |
1245 | return (ret); | |
1246 | } | |
34dc7c2f | 1247 | |
d877ac6b NB |
1248 | /* |
1249 | * Remove any previously existing symlink from a udev path to | |
5eacc075 AX |
1250 | * the device before labeling the disk. This ensures that |
1251 | * only newly created links are used. Otherwise there is a | |
1252 | * window between when udev deletes and recreates the link | |
1253 | * during which access attempts will fail with ENOENT. | |
d877ac6b | 1254 | */ |
cae5b340 | 1255 | strlcpy(udevpath, path, MAXPATHLEN); |
eac47204 BB |
1256 | (void) zfs_append_partition(udevpath, MAXPATHLEN); |
1257 | ||
8128bd89 BB |
1258 | fd = open(devpath, O_RDWR|O_EXCL); |
1259 | if (fd == -1) { | |
1260 | if (errno == EBUSY) | |
1261 | is_exclusive = 1; | |
1262 | } else { | |
1263 | (void) close(fd); | |
1264 | } | |
34dc7c2f BB |
1265 | |
1266 | /* | |
8128bd89 BB |
1267 | * If the partition exists, contains a valid spare label, |
1268 | * and is opened exclusively there is no need to partition | |
1269 | * it. Hot spares have already been partitioned and are | |
1270 | * held open exclusively by the kernel as a safety measure. | |
1271 | * | |
1272 | * If the provided path is for a /dev/disk/ device its | |
1273 | * symbolic link will be removed, partition table created, | |
1274 | * and then block until udev creates the new link. | |
34dc7c2f | 1275 | */ |
8128bd89 | 1276 | if (!is_exclusive || !is_spare(NULL, udevpath)) { |
5eacc075 AX |
1277 | char *devnode = strrchr(devpath, '/') + 1; |
1278 | ||
a08ee875 | 1279 | ret = strncmp(udevpath, UDISK_ROOT, strlen(UDISK_ROOT)); |
8128bd89 BB |
1280 | if (ret == 0) { |
1281 | ret = lstat64(udevpath, &statbuf); | |
1282 | if (ret == 0 && S_ISLNK(statbuf.st_mode)) | |
1283 | (void) unlink(udevpath); | |
1284 | } | |
1285 | ||
5eacc075 AX |
1286 | /* |
1287 | * When labeling a pool the raw device node name | |
1288 | * is provided as it appears under /dev/. | |
1289 | */ | |
1290 | if (zpool_label_disk(g_zfs, zhp, devnode) == -1) | |
8128bd89 BB |
1291 | return (-1); |
1292 | ||
5eacc075 AX |
1293 | /* |
1294 | * Wait for udev to signal the device is available | |
1295 | * by the provided path. | |
1296 | */ | |
a08ee875 | 1297 | ret = zpool_label_disk_wait(udevpath, DISK_LABEL_WAIT); |
8128bd89 | 1298 | if (ret) { |
5eacc075 AX |
1299 | (void) fprintf(stderr, |
1300 | gettext("missing link: %s was " | |
1301 | "partitioned but %s is missing\n"), | |
1302 | devnode, udevpath); | |
1303 | return (ret); | |
8128bd89 BB |
1304 | } |
1305 | ||
5eacc075 AX |
1306 | ret = zero_label(udevpath); |
1307 | if (ret) | |
1308 | return (ret); | |
34dc7c2f BB |
1309 | } |
1310 | ||
34dc7c2f | 1311 | /* |
eac47204 | 1312 | * Update the path to refer to the partition. The presence of |
34dc7c2f | 1313 | * the 'whole_disk' field indicates to the CLI that we should |
eac47204 | 1314 | * chop off the partition number when displaying the device in |
34dc7c2f BB |
1315 | * future output. |
1316 | */ | |
d877ac6b | 1317 | verify(nvlist_add_string(nv, ZPOOL_CONFIG_PATH, udevpath) == 0); |
34dc7c2f | 1318 | |
cae5b340 AX |
1319 | /* |
1320 | * Update device id strings for whole disks (Linux only) | |
1321 | */ | |
1322 | update_vdev_config_dev_strs(nv); | |
1323 | ||
34dc7c2f BB |
1324 | return (0); |
1325 | } | |
1326 | ||
1327 | for (c = 0; c < children; c++) | |
1328 | if ((ret = make_disks(zhp, child[c])) != 0) | |
1329 | return (ret); | |
1330 | ||
1331 | if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES, | |
1332 | &child, &children) == 0) | |
1333 | for (c = 0; c < children; c++) | |
1334 | if ((ret = make_disks(zhp, child[c])) != 0) | |
1335 | return (ret); | |
1336 | ||
1337 | if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE, | |
1338 | &child, &children) == 0) | |
1339 | for (c = 0; c < children; c++) | |
1340 | if ((ret = make_disks(zhp, child[c])) != 0) | |
1341 | return (ret); | |
1342 | ||
1343 | return (0); | |
1344 | } | |
1345 | ||
34dc7c2f BB |
1346 | /* |
1347 | * Go through and find any devices that are in use. We rely on libdiskmgt for | |
1348 | * the majority of this task. | |
1349 | */ | |
e10b0808 AX |
1350 | static boolean_t |
1351 | is_device_in_use(nvlist_t *config, nvlist_t *nv, boolean_t force, | |
428870ff | 1352 | boolean_t replacing, boolean_t isspare) |
34dc7c2f BB |
1353 | { |
1354 | nvlist_t **child; | |
1355 | uint_t c, children; | |
1356 | char *type, *path; | |
d603ed6c | 1357 | int ret = 0; |
34dc7c2f | 1358 | char buf[MAXPATHLEN]; |
d603ed6c | 1359 | uint64_t wholedisk = B_FALSE; |
e10b0808 | 1360 | boolean_t anyinuse = B_FALSE; |
34dc7c2f BB |
1361 | |
1362 | verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0); | |
1363 | ||
1364 | if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, | |
1365 | &child, &children) != 0) { | |
1366 | ||
d603ed6c BB |
1367 | verify(!nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path)); |
1368 | if (strcmp(type, VDEV_TYPE_DISK) == 0) | |
1369 | verify(!nvlist_lookup_uint64(nv, | |
a08ee875 | 1370 | ZPOOL_CONFIG_WHOLE_DISK, &wholedisk)); |
34dc7c2f BB |
1371 | |
1372 | /* | |
1373 | * As a generic check, we look to see if this is a replace of a | |
1374 | * hot spare within the same pool. If so, we allow it | |
d603ed6c | 1375 | * regardless of what libblkid or zpool_in_use() says. |
34dc7c2f | 1376 | */ |
428870ff | 1377 | if (replacing) { |
8128bd89 BB |
1378 | (void) strlcpy(buf, path, sizeof (buf)); |
1379 | if (wholedisk) { | |
1380 | ret = zfs_append_partition(buf, sizeof (buf)); | |
1381 | if (ret == -1) | |
1382 | return (-1); | |
1383 | } | |
428870ff | 1384 | |
34dc7c2f | 1385 | if (is_spare(config, buf)) |
e10b0808 | 1386 | return (B_FALSE); |
34dc7c2f BB |
1387 | } |
1388 | ||
1389 | if (strcmp(type, VDEV_TYPE_DISK) == 0) | |
d603ed6c | 1390 | ret = check_device(path, force, isspare, wholedisk); |
34dc7c2f | 1391 | |
e10b0808 | 1392 | else if (strcmp(type, VDEV_TYPE_FILE) == 0) |
34dc7c2f BB |
1393 | ret = check_file(path, force, isspare); |
1394 | ||
e10b0808 | 1395 | return (ret != 0); |
34dc7c2f BB |
1396 | } |
1397 | ||
1398 | for (c = 0; c < children; c++) | |
e10b0808 AX |
1399 | if (is_device_in_use(config, child[c], force, replacing, |
1400 | B_FALSE)) | |
1401 | anyinuse = B_TRUE; | |
34dc7c2f BB |
1402 | |
1403 | if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES, | |
1404 | &child, &children) == 0) | |
1405 | for (c = 0; c < children; c++) | |
e10b0808 AX |
1406 | if (is_device_in_use(config, child[c], force, replacing, |
1407 | B_TRUE)) | |
1408 | anyinuse = B_TRUE; | |
34dc7c2f BB |
1409 | |
1410 | if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE, | |
1411 | &child, &children) == 0) | |
1412 | for (c = 0; c < children; c++) | |
e10b0808 AX |
1413 | if (is_device_in_use(config, child[c], force, replacing, |
1414 | B_FALSE)) | |
1415 | anyinuse = B_TRUE; | |
34dc7c2f | 1416 | |
e10b0808 | 1417 | return (anyinuse); |
34dc7c2f BB |
1418 | } |
1419 | ||
1420 | static const char * | |
45d1cae3 | 1421 | is_grouping(const char *type, int *mindev, int *maxdev) |
34dc7c2f | 1422 | { |
45d1cae3 BB |
1423 | if (strncmp(type, "raidz", 5) == 0) { |
1424 | const char *p = type + 5; | |
1425 | char *end; | |
1426 | long nparity; | |
1427 | ||
1428 | if (*p == '\0') { | |
1429 | nparity = 1; | |
1430 | } else if (*p == '0') { | |
1431 | return (NULL); /* no zero prefixes allowed */ | |
1432 | } else { | |
1433 | errno = 0; | |
1434 | nparity = strtol(p, &end, 10); | |
1435 | if (errno != 0 || nparity < 1 || nparity >= 255 || | |
1436 | *end != '\0') | |
1437 | return (NULL); | |
1438 | } | |
34dc7c2f | 1439 | |
34dc7c2f | 1440 | if (mindev != NULL) |
45d1cae3 BB |
1441 | *mindev = nparity + 1; |
1442 | if (maxdev != NULL) | |
1443 | *maxdev = 255; | |
34dc7c2f BB |
1444 | return (VDEV_TYPE_RAIDZ); |
1445 | } | |
1446 | ||
45d1cae3 BB |
1447 | if (maxdev != NULL) |
1448 | *maxdev = INT_MAX; | |
1449 | ||
34dc7c2f BB |
1450 | if (strcmp(type, "mirror") == 0) { |
1451 | if (mindev != NULL) | |
1452 | *mindev = 2; | |
1453 | return (VDEV_TYPE_MIRROR); | |
1454 | } | |
1455 | ||
1456 | if (strcmp(type, "spare") == 0) { | |
1457 | if (mindev != NULL) | |
1458 | *mindev = 1; | |
1459 | return (VDEV_TYPE_SPARE); | |
1460 | } | |
1461 | ||
1462 | if (strcmp(type, "log") == 0) { | |
1463 | if (mindev != NULL) | |
1464 | *mindev = 1; | |
1465 | return (VDEV_TYPE_LOG); | |
1466 | } | |
1467 | ||
1468 | if (strcmp(type, "cache") == 0) { | |
1469 | if (mindev != NULL) | |
1470 | *mindev = 1; | |
1471 | return (VDEV_TYPE_L2CACHE); | |
1472 | } | |
1473 | ||
1474 | return (NULL); | |
1475 | } | |
1476 | ||
1477 | /* | |
1478 | * Construct a syntactically valid vdev specification, | |
1479 | * and ensure that all devices and files exist and can be opened. | |
1480 | * Note: we don't bother freeing anything in the error paths | |
1481 | * because the program is just going to exit anyway. | |
1482 | */ | |
1483 | nvlist_t * | |
df30f566 | 1484 | construct_spec(nvlist_t *props, int argc, char **argv) |
34dc7c2f BB |
1485 | { |
1486 | nvlist_t *nvroot, *nv, **top, **spares, **l2cache; | |
45d1cae3 | 1487 | int t, toplevels, mindev, maxdev, nspares, nlogs, nl2cache; |
34dc7c2f BB |
1488 | const char *type; |
1489 | uint64_t is_log; | |
1490 | boolean_t seen_logs; | |
1491 | ||
1492 | top = NULL; | |
1493 | toplevels = 0; | |
1494 | spares = NULL; | |
1495 | l2cache = NULL; | |
1496 | nspares = 0; | |
1497 | nlogs = 0; | |
1498 | nl2cache = 0; | |
1499 | is_log = B_FALSE; | |
1500 | seen_logs = B_FALSE; | |
cae5b340 | 1501 | nvroot = NULL; |
34dc7c2f BB |
1502 | |
1503 | while (argc > 0) { | |
1504 | nv = NULL; | |
1505 | ||
1506 | /* | |
1507 | * If it's a mirror or raidz, the subsequent arguments are | |
1508 | * its leaves -- until we encounter the next mirror or raidz. | |
1509 | */ | |
45d1cae3 | 1510 | if ((type = is_grouping(argv[0], &mindev, &maxdev)) != NULL) { |
34dc7c2f BB |
1511 | nvlist_t **child = NULL; |
1512 | int c, children = 0; | |
1513 | ||
1514 | if (strcmp(type, VDEV_TYPE_SPARE) == 0) { | |
1515 | if (spares != NULL) { | |
1516 | (void) fprintf(stderr, | |
1517 | gettext("invalid vdev " | |
1518 | "specification: 'spare' can be " | |
1519 | "specified only once\n")); | |
cae5b340 | 1520 | goto spec_out; |
34dc7c2f BB |
1521 | } |
1522 | is_log = B_FALSE; | |
1523 | } | |
1524 | ||
1525 | if (strcmp(type, VDEV_TYPE_LOG) == 0) { | |
1526 | if (seen_logs) { | |
1527 | (void) fprintf(stderr, | |
1528 | gettext("invalid vdev " | |
1529 | "specification: 'log' can be " | |
1530 | "specified only once\n")); | |
cae5b340 | 1531 | goto spec_out; |
34dc7c2f BB |
1532 | } |
1533 | seen_logs = B_TRUE; | |
1534 | is_log = B_TRUE; | |
1535 | argc--; | |
1536 | argv++; | |
1537 | /* | |
1538 | * A log is not a real grouping device. | |
1539 | * We just set is_log and continue. | |
1540 | */ | |
1541 | continue; | |
1542 | } | |
1543 | ||
1544 | if (strcmp(type, VDEV_TYPE_L2CACHE) == 0) { | |
1545 | if (l2cache != NULL) { | |
1546 | (void) fprintf(stderr, | |
1547 | gettext("invalid vdev " | |
1548 | "specification: 'cache' can be " | |
1549 | "specified only once\n")); | |
cae5b340 | 1550 | goto spec_out; |
34dc7c2f BB |
1551 | } |
1552 | is_log = B_FALSE; | |
1553 | } | |
1554 | ||
1555 | if (is_log) { | |
1556 | if (strcmp(type, VDEV_TYPE_MIRROR) != 0) { | |
1557 | (void) fprintf(stderr, | |
1558 | gettext("invalid vdev " | |
1559 | "specification: unsupported 'log' " | |
1560 | "device: %s\n"), type); | |
cae5b340 | 1561 | goto spec_out; |
34dc7c2f BB |
1562 | } |
1563 | nlogs++; | |
1564 | } | |
1565 | ||
1566 | for (c = 1; c < argc; c++) { | |
45d1cae3 | 1567 | if (is_grouping(argv[c], NULL, NULL) != NULL) |
34dc7c2f BB |
1568 | break; |
1569 | children++; | |
1570 | child = realloc(child, | |
1571 | children * sizeof (nvlist_t *)); | |
1572 | if (child == NULL) | |
1573 | zpool_no_memory(); | |
a08ee875 | 1574 | if ((nv = make_leaf_vdev(props, argv[c], |
cae5b340 AX |
1575 | B_FALSE)) == NULL) { |
1576 | for (c = 0; c < children - 1; c++) | |
1577 | nvlist_free(child[c]); | |
1578 | free(child); | |
1579 | goto spec_out; | |
1580 | } | |
1581 | ||
34dc7c2f BB |
1582 | child[children - 1] = nv; |
1583 | } | |
1584 | ||
1585 | if (children < mindev) { | |
1586 | (void) fprintf(stderr, gettext("invalid vdev " | |
1587 | "specification: %s requires at least %d " | |
1588 | "devices\n"), argv[0], mindev); | |
cae5b340 AX |
1589 | for (c = 0; c < children; c++) |
1590 | nvlist_free(child[c]); | |
1591 | free(child); | |
1592 | goto spec_out; | |
34dc7c2f BB |
1593 | } |
1594 | ||
45d1cae3 BB |
1595 | if (children > maxdev) { |
1596 | (void) fprintf(stderr, gettext("invalid vdev " | |
1597 | "specification: %s supports no more than " | |
1598 | "%d devices\n"), argv[0], maxdev); | |
cae5b340 AX |
1599 | for (c = 0; c < children; c++) |
1600 | nvlist_free(child[c]); | |
1601 | free(child); | |
1602 | goto spec_out; | |
45d1cae3 BB |
1603 | } |
1604 | ||
34dc7c2f BB |
1605 | argc -= c; |
1606 | argv += c; | |
1607 | ||
1608 | if (strcmp(type, VDEV_TYPE_SPARE) == 0) { | |
1609 | spares = child; | |
1610 | nspares = children; | |
1611 | continue; | |
1612 | } else if (strcmp(type, VDEV_TYPE_L2CACHE) == 0) { | |
1613 | l2cache = child; | |
1614 | nl2cache = children; | |
1615 | continue; | |
1616 | } else { | |
1617 | verify(nvlist_alloc(&nv, NV_UNIQUE_NAME, | |
1618 | 0) == 0); | |
1619 | verify(nvlist_add_string(nv, ZPOOL_CONFIG_TYPE, | |
1620 | type) == 0); | |
1621 | verify(nvlist_add_uint64(nv, | |
1622 | ZPOOL_CONFIG_IS_LOG, is_log) == 0); | |
1623 | if (strcmp(type, VDEV_TYPE_RAIDZ) == 0) { | |
1624 | verify(nvlist_add_uint64(nv, | |
1625 | ZPOOL_CONFIG_NPARITY, | |
1626 | mindev - 1) == 0); | |
1627 | } | |
1628 | verify(nvlist_add_nvlist_array(nv, | |
1629 | ZPOOL_CONFIG_CHILDREN, child, | |
1630 | children) == 0); | |
1631 | ||
1632 | for (c = 0; c < children; c++) | |
1633 | nvlist_free(child[c]); | |
1634 | free(child); | |
1635 | } | |
1636 | } else { | |
1637 | /* | |
1638 | * We have a device. Pass off to make_leaf_vdev() to | |
1639 | * construct the appropriate nvlist describing the vdev. | |
1640 | */ | |
a08ee875 LG |
1641 | if ((nv = make_leaf_vdev(props, argv[0], |
1642 | is_log)) == NULL) | |
cae5b340 AX |
1643 | goto spec_out; |
1644 | ||
34dc7c2f BB |
1645 | if (is_log) |
1646 | nlogs++; | |
1647 | argc--; | |
1648 | argv++; | |
1649 | } | |
1650 | ||
1651 | toplevels++; | |
1652 | top = realloc(top, toplevels * sizeof (nvlist_t *)); | |
1653 | if (top == NULL) | |
1654 | zpool_no_memory(); | |
1655 | top[toplevels - 1] = nv; | |
1656 | } | |
1657 | ||
1658 | if (toplevels == 0 && nspares == 0 && nl2cache == 0) { | |
1659 | (void) fprintf(stderr, gettext("invalid vdev " | |
1660 | "specification: at least one toplevel vdev must be " | |
1661 | "specified\n")); | |
cae5b340 | 1662 | goto spec_out; |
34dc7c2f BB |
1663 | } |
1664 | ||
1665 | if (seen_logs && nlogs == 0) { | |
1666 | (void) fprintf(stderr, gettext("invalid vdev specification: " | |
1667 | "log requires at least 1 device\n")); | |
cae5b340 | 1668 | goto spec_out; |
34dc7c2f BB |
1669 | } |
1670 | ||
1671 | /* | |
1672 | * Finally, create nvroot and add all top-level vdevs to it. | |
1673 | */ | |
1674 | verify(nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) == 0); | |
1675 | verify(nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE, | |
1676 | VDEV_TYPE_ROOT) == 0); | |
1677 | verify(nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, | |
1678 | top, toplevels) == 0); | |
1679 | if (nspares != 0) | |
1680 | verify(nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, | |
1681 | spares, nspares) == 0); | |
1682 | if (nl2cache != 0) | |
1683 | verify(nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, | |
1684 | l2cache, nl2cache) == 0); | |
1685 | ||
cae5b340 | 1686 | spec_out: |
34dc7c2f BB |
1687 | for (t = 0; t < toplevels; t++) |
1688 | nvlist_free(top[t]); | |
1689 | for (t = 0; t < nspares; t++) | |
1690 | nvlist_free(spares[t]); | |
1691 | for (t = 0; t < nl2cache; t++) | |
1692 | nvlist_free(l2cache[t]); | |
cae5b340 AX |
1693 | |
1694 | free(spares); | |
1695 | free(l2cache); | |
34dc7c2f BB |
1696 | free(top); |
1697 | ||
1698 | return (nvroot); | |
1699 | } | |
1700 | ||
428870ff BB |
1701 | nvlist_t * |
1702 | split_mirror_vdev(zpool_handle_t *zhp, char *newname, nvlist_t *props, | |
1703 | splitflags_t flags, int argc, char **argv) | |
1704 | { | |
1705 | nvlist_t *newroot = NULL, **child; | |
1706 | uint_t c, children; | |
1707 | ||
1708 | if (argc > 0) { | |
df30f566 | 1709 | if ((newroot = construct_spec(props, argc, argv)) == NULL) { |
428870ff BB |
1710 | (void) fprintf(stderr, gettext("Unable to build a " |
1711 | "pool from the specified devices\n")); | |
1712 | return (NULL); | |
1713 | } | |
1714 | ||
1715 | if (!flags.dryrun && make_disks(zhp, newroot) != 0) { | |
1716 | nvlist_free(newroot); | |
1717 | return (NULL); | |
1718 | } | |
1719 | ||
1720 | /* avoid any tricks in the spec */ | |
1721 | verify(nvlist_lookup_nvlist_array(newroot, | |
1722 | ZPOOL_CONFIG_CHILDREN, &child, &children) == 0); | |
1723 | for (c = 0; c < children; c++) { | |
1724 | char *path; | |
1725 | const char *type; | |
1726 | int min, max; | |
1727 | ||
1728 | verify(nvlist_lookup_string(child[c], | |
1729 | ZPOOL_CONFIG_PATH, &path) == 0); | |
1730 | if ((type = is_grouping(path, &min, &max)) != NULL) { | |
1731 | (void) fprintf(stderr, gettext("Cannot use " | |
1732 | "'%s' as a device for splitting\n"), type); | |
1733 | nvlist_free(newroot); | |
1734 | return (NULL); | |
1735 | } | |
1736 | } | |
1737 | } | |
1738 | ||
1739 | if (zpool_vdev_split(zhp, newname, &newroot, props, flags) != 0) { | |
cae5b340 | 1740 | nvlist_free(newroot); |
428870ff BB |
1741 | return (NULL); |
1742 | } | |
1743 | ||
1744 | return (newroot); | |
1745 | } | |
34dc7c2f BB |
1746 | |
1747 | /* | |
1748 | * Get and validate the contents of the given vdev specification. This ensures | |
1749 | * that the nvlist returned is well-formed, that all the devices exist, and that | |
1750 | * they are not currently in use by any other known consumer. The 'poolconfig' | |
1751 | * parameter is the current configuration of the pool when adding devices | |
1752 | * existing pool, and is used to perform additional checks, such as changing the | |
1753 | * replication level of the pool. It can be 'NULL' to indicate that this is a | |
1754 | * new pool. The 'force' flag controls whether devices should be forcefully | |
1755 | * added, even if they appear in use. | |
1756 | */ | |
1757 | nvlist_t * | |
df30f566 | 1758 | make_root_vdev(zpool_handle_t *zhp, nvlist_t *props, int force, int check_rep, |
428870ff | 1759 | boolean_t replacing, boolean_t dryrun, int argc, char **argv) |
34dc7c2f BB |
1760 | { |
1761 | nvlist_t *newroot; | |
1762 | nvlist_t *poolconfig = NULL; | |
1763 | is_force = force; | |
1764 | ||
1765 | /* | |
1766 | * Construct the vdev specification. If this is successful, we know | |
1767 | * that we have a valid specification, and that all devices can be | |
1768 | * opened. | |
1769 | */ | |
df30f566 | 1770 | if ((newroot = construct_spec(props, argc, argv)) == NULL) |
34dc7c2f BB |
1771 | return (NULL); |
1772 | ||
e10b0808 AX |
1773 | if (zhp && ((poolconfig = zpool_get_config(zhp, NULL)) == NULL)) { |
1774 | nvlist_free(newroot); | |
34dc7c2f | 1775 | return (NULL); |
e10b0808 | 1776 | } |
34dc7c2f BB |
1777 | |
1778 | /* | |
1779 | * Validate each device to make sure that its not shared with another | |
1780 | * subsystem. We do this even if 'force' is set, because there are some | |
1781 | * uses (such as a dedicated dump device) that even '-f' cannot | |
1782 | * override. | |
1783 | */ | |
e10b0808 | 1784 | if (is_device_in_use(poolconfig, newroot, force, replacing, B_FALSE)) { |
34dc7c2f BB |
1785 | nvlist_free(newroot); |
1786 | return (NULL); | |
1787 | } | |
1788 | ||
1789 | /* | |
1790 | * Check the replication level of the given vdevs and report any errors | |
1791 | * found. We include the existing pool spec, if any, as we need to | |
1792 | * catch changes against the existing replication level. | |
1793 | */ | |
1794 | if (check_rep && check_replication(poolconfig, newroot) != 0) { | |
1795 | nvlist_free(newroot); | |
1796 | return (NULL); | |
1797 | } | |
1798 | ||
1799 | /* | |
1800 | * Run through the vdev specification and label any whole disks found. | |
1801 | */ | |
b128c09f | 1802 | if (!dryrun && make_disks(zhp, newroot) != 0) { |
34dc7c2f BB |
1803 | nvlist_free(newroot); |
1804 | return (NULL); | |
1805 | } | |
1806 | ||
1807 | return (newroot); | |
1808 | } |