Print this page
NEX-4934 Add capability to remove special vdev
Reviewed by: Alex Aizman <alex.aizman@nexenta.com>
Reviewed by: Alek Pinchuk <alek.pinchuk@nexenta.com>
| Split |
Close |
| Expand all |
| Collapse all |
--- old/usr/src/cmd/zinject/translate.c
+++ new/usr/src/cmd/zinject/translate.c
1 1 /*
2 2 * CDDL HEADER START
3 3 *
4 4 * The contents of this file are subject to the terms of the
5 5 * Common Development and Distribution License (the "License").
6 6 * You may not use this file except in compliance with the License.
7 7 *
8 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 9 * or http://www.opensolaris.org/os/licensing.
10 10 * See the License for the specific language governing permissions
11 11 * and limitations under the License.
12 12 *
13 13 * When distributing Covered Code, include this CDDL HEADER in each
|
↓ open down ↓ |
13 lines elided |
↑ open up ↑ |
14 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 15 * If applicable, add the following below this CDDL HEADER, with the
16 16 * fields enclosed by brackets "[]" replaced with your own identifying
17 17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 18 *
19 19 * CDDL HEADER END
20 20 */
21 21 /*
22 22 * Copyright (c) 2006, 2010, Oracle and/or its affiliates. All rights reserved.
23 23 * Copyright (c) 2012 by Delphix. All rights reserved.
24 + * Copyright 2015 Nexenta Systems, Inc. All rights reserved.
24 25 */
25 26
26 27 #include <libzfs.h>
27 28
28 29 #include <sys/zfs_context.h>
29 30
30 31 #include <errno.h>
31 32 #include <fcntl.h>
32 33 #include <stdarg.h>
33 34 #include <stddef.h>
34 35 #include <stdio.h>
35 36 #include <stdlib.h>
36 37 #include <strings.h>
37 38 #include <sys/file.h>
38 39 #include <sys/mntent.h>
39 40 #include <sys/mnttab.h>
40 41 #include <sys/param.h>
41 42 #include <sys/stat.h>
42 43
43 44 #include <sys/dmu.h>
44 45 #include <sys/dmu_objset.h>
45 46 #include <sys/dnode.h>
46 47 #include <sys/vdev_impl.h>
47 48
48 49 #include <sys/mkdev.h>
49 50
50 51 #include "zinject.h"
51 52
52 53 extern void kernel_init(int);
53 54 extern void kernel_fini(void);
54 55
55 56 static int debug;
56 57
57 58 static void
58 59 ziprintf(const char *fmt, ...)
59 60 {
60 61 va_list ap;
61 62
62 63 if (!debug)
63 64 return;
64 65
65 66 va_start(ap, fmt);
66 67 (void) vprintf(fmt, ap);
67 68 va_end(ap);
68 69 }
69 70
70 71 static void
71 72 compress_slashes(const char *src, char *dest)
72 73 {
73 74 while (*src != '\0') {
74 75 *dest = *src++;
75 76 while (*dest == '/' && *src == '/')
76 77 ++src;
77 78 ++dest;
78 79 }
79 80 *dest = '\0';
80 81 }
81 82
82 83 /*
83 84 * Given a full path to a file, translate into a dataset name and a relative
84 85 * path within the dataset. 'dataset' must be at least MAXNAMELEN characters,
85 86 * and 'relpath' must be at least MAXPATHLEN characters. We also pass a stat64
86 87 * buffer, which we need later to get the object ID.
87 88 */
88 89 static int
89 90 parse_pathname(const char *inpath, char *dataset, char *relpath,
90 91 struct stat64 *statbuf)
91 92 {
92 93 struct extmnttab mp;
93 94 FILE *fp;
94 95 int match;
95 96 const char *rel;
96 97 char fullpath[MAXPATHLEN];
97 98
98 99 compress_slashes(inpath, fullpath);
99 100
100 101 if (fullpath[0] != '/') {
101 102 (void) fprintf(stderr, "invalid object '%s': must be full "
102 103 "path\n", fullpath);
103 104 usage();
104 105 return (-1);
105 106 }
106 107
107 108 if (strlen(fullpath) >= MAXPATHLEN) {
108 109 (void) fprintf(stderr, "invalid object; pathname too long\n");
109 110 return (-1);
110 111 }
111 112
112 113 if (stat64(fullpath, statbuf) != 0) {
113 114 (void) fprintf(stderr, "cannot open '%s': %s\n",
114 115 fullpath, strerror(errno));
115 116 return (-1);
116 117 }
117 118
118 119 if ((fp = fopen(MNTTAB, "r")) == NULL) {
119 120 (void) fprintf(stderr, "cannot open /etc/mnttab\n");
120 121 return (-1);
121 122 }
122 123
123 124 match = 0;
124 125 while (getextmntent(fp, &mp, sizeof (mp)) == 0) {
125 126 if (makedev(mp.mnt_major, mp.mnt_minor) == statbuf->st_dev) {
126 127 match = 1;
127 128 break;
128 129 }
129 130 }
130 131
131 132 if (!match) {
132 133 (void) fprintf(stderr, "cannot find mountpoint for '%s'\n",
133 134 fullpath);
134 135 return (-1);
135 136 }
136 137
137 138 if (strcmp(mp.mnt_fstype, MNTTYPE_ZFS) != 0) {
138 139 (void) fprintf(stderr, "invalid path '%s': not a ZFS "
139 140 "filesystem\n", fullpath);
140 141 return (-1);
141 142 }
142 143
143 144 if (strncmp(fullpath, mp.mnt_mountp, strlen(mp.mnt_mountp)) != 0) {
144 145 (void) fprintf(stderr, "invalid path '%s': mountpoint "
145 146 "doesn't match path\n", fullpath);
146 147 return (-1);
147 148 }
148 149
149 150 (void) strcpy(dataset, mp.mnt_special);
150 151
151 152 rel = fullpath + strlen(mp.mnt_mountp);
152 153 if (rel[0] == '/')
153 154 rel++;
154 155 (void) strcpy(relpath, rel);
155 156
156 157 return (0);
157 158 }
158 159
159 160 /*
160 161 * Convert from a (dataset, path) pair into a (objset, object) pair. Note that
161 162 * we grab the object number from the inode number, since looking this up via
162 163 * libzpool is a real pain.
163 164 */
164 165 /* ARGSUSED */
165 166 static int
166 167 object_from_path(const char *dataset, const char *path, struct stat64 *statbuf,
167 168 zinject_record_t *record)
168 169 {
169 170 objset_t *os;
170 171 int err;
171 172
172 173 /*
173 174 * Before doing any libzpool operations, call sync() to ensure that the
174 175 * on-disk state is consistent with the in-core state.
175 176 */
176 177 sync();
177 178
178 179 err = dmu_objset_own(dataset, DMU_OST_ZFS, B_TRUE, FTAG, &os);
179 180 if (err != 0) {
180 181 (void) fprintf(stderr, "cannot open dataset '%s': %s\n",
181 182 dataset, strerror(err));
182 183 return (-1);
183 184 }
184 185
185 186 record->zi_objset = dmu_objset_id(os);
186 187 record->zi_object = statbuf->st_ino;
187 188
188 189 dmu_objset_disown(os, FTAG);
189 190
190 191 return (0);
191 192 }
192 193
193 194 /*
194 195 * Calculate the real range based on the type, level, and range given.
195 196 */
196 197 static int
197 198 calculate_range(const char *dataset, err_type_t type, int level, char *range,
198 199 zinject_record_t *record)
199 200 {
200 201 objset_t *os = NULL;
201 202 dnode_t *dn = NULL;
202 203 int err;
203 204 int ret = -1;
204 205
205 206 /*
206 207 * Determine the numeric range from the string.
207 208 */
208 209 if (range == NULL) {
209 210 /*
210 211 * If range is unspecified, set the range to [0,-1], which
211 212 * indicates that the whole object should be treated as an
212 213 * error.
213 214 */
214 215 record->zi_start = 0;
215 216 record->zi_end = -1ULL;
216 217 } else {
217 218 char *end;
218 219
219 220 /* XXX add support for suffixes */
220 221 record->zi_start = strtoull(range, &end, 10);
221 222
222 223
223 224 if (*end == '\0')
224 225 record->zi_end = record->zi_start + 1;
225 226 else if (*end == ',')
226 227 record->zi_end = strtoull(end + 1, &end, 10);
227 228
228 229 if (*end != '\0') {
229 230 (void) fprintf(stderr, "invalid range '%s': must be "
230 231 "a numeric range of the form 'start[,end]'\n",
231 232 range);
232 233 goto out;
233 234 }
234 235 }
235 236
236 237 switch (type) {
237 238 case TYPE_DATA:
238 239 break;
239 240
240 241 case TYPE_DNODE:
241 242 /*
242 243 * If this is a request to inject faults into the dnode, then we
243 244 * must translate the current (objset,object) pair into an
244 245 * offset within the metadnode for the objset. Specifying any
245 246 * kind of range with type 'dnode' is illegal.
246 247 */
247 248 if (range != NULL) {
248 249 (void) fprintf(stderr, "range cannot be specified when "
249 250 "type is 'dnode'\n");
250 251 goto out;
251 252 }
252 253
253 254 record->zi_start = record->zi_object * sizeof (dnode_phys_t);
254 255 record->zi_end = record->zi_start + sizeof (dnode_phys_t);
255 256 record->zi_object = 0;
256 257 break;
257 258 }
258 259
259 260 /*
260 261 * Get the dnode associated with object, so we can calculate the block
261 262 * size.
262 263 */
263 264 if ((err = dmu_objset_own(dataset, DMU_OST_ANY,
264 265 B_TRUE, FTAG, &os)) != 0) {
265 266 (void) fprintf(stderr, "cannot open dataset '%s': %s\n",
266 267 dataset, strerror(err));
267 268 goto out;
268 269 }
269 270
270 271 if (record->zi_object == 0) {
271 272 dn = DMU_META_DNODE(os);
272 273 } else {
273 274 err = dnode_hold(os, record->zi_object, FTAG, &dn);
274 275 if (err != 0) {
275 276 (void) fprintf(stderr, "failed to hold dnode "
276 277 "for object %llu\n",
277 278 (u_longlong_t)record->zi_object);
278 279 goto out;
279 280 }
280 281 }
281 282
282 283
283 284 ziprintf("data shift: %d\n", (int)dn->dn_datablkshift);
284 285 ziprintf(" ind shift: %d\n", (int)dn->dn_indblkshift);
285 286
286 287 /*
287 288 * Translate range into block IDs.
288 289 */
289 290 if (record->zi_start != 0 || record->zi_end != -1ULL) {
290 291 record->zi_start >>= dn->dn_datablkshift;
291 292 record->zi_end >>= dn->dn_datablkshift;
292 293 }
293 294
294 295 /*
295 296 * Check level, and then translate level 0 blkids into ranges
296 297 * appropriate for level of indirection.
297 298 */
298 299 record->zi_level = level;
299 300 if (level > 0) {
300 301 ziprintf("level 0 blkid range: [%llu, %llu]\n",
301 302 record->zi_start, record->zi_end);
302 303
303 304 if (level >= dn->dn_nlevels) {
304 305 (void) fprintf(stderr, "level %d exceeds max level "
305 306 "of object (%d)\n", level, dn->dn_nlevels - 1);
306 307 goto out;
307 308 }
308 309
309 310 if (record->zi_start != 0 || record->zi_end != 0) {
310 311 int shift = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
311 312
312 313 for (; level > 0; level--) {
313 314 record->zi_start >>= shift;
314 315 record->zi_end >>= shift;
315 316 }
316 317 }
317 318 }
318 319
319 320 ret = 0;
320 321 out:
321 322 if (dn) {
322 323 if (dn != DMU_META_DNODE(os))
323 324 dnode_rele(dn, FTAG);
324 325 }
325 326 if (os)
326 327 dmu_objset_disown(os, FTAG);
327 328
328 329 return (ret);
329 330 }
330 331
331 332 int
332 333 translate_record(err_type_t type, const char *object, const char *range,
333 334 int level, zinject_record_t *record, char *poolname, char *dataset)
334 335 {
335 336 char path[MAXPATHLEN];
336 337 char *slash;
337 338 struct stat64 statbuf;
338 339 int ret = -1;
339 340
340 341 kernel_init(FREAD);
341 342
342 343 debug = (getenv("ZINJECT_DEBUG") != NULL);
343 344
344 345 ziprintf("translating: %s\n", object);
345 346
346 347 if (MOS_TYPE(type)) {
347 348 /*
348 349 * MOS objects are treated specially.
349 350 */
350 351 switch (type) {
351 352 case TYPE_MOS:
352 353 record->zi_type = 0;
353 354 break;
354 355 case TYPE_MOSDIR:
355 356 record->zi_type = DMU_OT_OBJECT_DIRECTORY;
356 357 break;
357 358 case TYPE_METASLAB:
358 359 record->zi_type = DMU_OT_OBJECT_ARRAY;
359 360 break;
360 361 case TYPE_CONFIG:
361 362 record->zi_type = DMU_OT_PACKED_NVLIST;
362 363 break;
363 364 case TYPE_BPOBJ:
364 365 record->zi_type = DMU_OT_BPOBJ;
365 366 break;
366 367 case TYPE_SPACEMAP:
367 368 record->zi_type = DMU_OT_SPACE_MAP;
368 369 break;
369 370 case TYPE_ERRLOG:
370 371 record->zi_type = DMU_OT_ERROR_LOG;
371 372 break;
372 373 }
373 374
374 375 dataset[0] = '\0';
375 376 (void) strcpy(poolname, object);
376 377 return (0);
377 378 }
378 379
379 380 /*
380 381 * Convert a full path into a (dataset, file) pair.
381 382 */
382 383 if (parse_pathname(object, dataset, path, &statbuf) != 0)
383 384 goto err;
384 385
385 386 ziprintf(" dataset: %s\n", dataset);
386 387 ziprintf(" path: %s\n", path);
387 388
388 389 /*
389 390 * Convert (dataset, file) into (objset, object)
390 391 */
391 392 if (object_from_path(dataset, path, &statbuf, record) != 0)
392 393 goto err;
393 394
394 395 ziprintf("raw objset: %llu\n", record->zi_objset);
395 396 ziprintf("raw object: %llu\n", record->zi_object);
396 397
397 398 /*
398 399 * For the given object, calculate the real (type, level, range)
399 400 */
400 401 if (calculate_range(dataset, type, level, (char *)range, record) != 0)
401 402 goto err;
402 403
403 404 ziprintf(" objset: %llu\n", record->zi_objset);
404 405 ziprintf(" object: %llu\n", record->zi_object);
405 406 if (record->zi_start == 0 &&
406 407 record->zi_end == -1ULL)
407 408 ziprintf(" range: all\n");
408 409 else
409 410 ziprintf(" range: [%llu, %llu]\n", record->zi_start,
410 411 record->zi_end);
411 412
412 413 /*
413 414 * Copy the pool name
414 415 */
415 416 (void) strcpy(poolname, dataset);
416 417 if ((slash = strchr(poolname, '/')) != NULL)
417 418 *slash = '\0';
418 419
419 420 ret = 0;
420 421
421 422 err:
422 423 kernel_fini();
423 424 return (ret);
424 425 }
425 426
426 427 int
427 428 translate_raw(const char *str, zinject_record_t *record)
428 429 {
429 430 /*
430 431 * A raw bookmark of the form objset:object:level:blkid, where each
431 432 * number is a hexidecimal value.
432 433 */
433 434 if (sscanf(str, "%llx:%llx:%x:%llx", (u_longlong_t *)&record->zi_objset,
434 435 (u_longlong_t *)&record->zi_object, &record->zi_level,
435 436 (u_longlong_t *)&record->zi_start) != 4) {
436 437 (void) fprintf(stderr, "bad raw spec '%s': must be of the form "
437 438 "'objset:object:level:blkid'\n", str);
438 439 return (-1);
439 440 }
440 441
441 442 record->zi_end = record->zi_start;
442 443
443 444 return (0);
444 445 }
445 446
446 447 int
447 448 translate_device(const char *pool, const char *device, err_type_t label_type,
448 449 zinject_record_t *record)
449 450 {
450 451 char *end;
451 452 zpool_handle_t *zhp;
452 453 nvlist_t *tgt;
453 454 boolean_t isspare, iscache;
|
↓ open down ↓ |
420 lines elided |
↑ open up ↑ |
454 455
455 456 /*
456 457 * Given a device name or GUID, create an appropriate injection record
457 458 * with zi_guid set.
458 459 */
459 460 if ((zhp = zpool_open(g_zfs, pool)) == NULL)
460 461 return (-1);
461 462
462 463 record->zi_guid = strtoull(device, &end, 16);
463 464 if (record->zi_guid == 0 || *end != '\0') {
464 - tgt = zpool_find_vdev(zhp, device, &isspare, &iscache, NULL);
465 + tgt = zpool_find_vdev(zhp, device, &isspare, &iscache, NULL,
466 + NULL);
465 467
466 468 if (tgt == NULL) {
467 469 (void) fprintf(stderr, "cannot find device '%s' in "
468 470 "pool '%s'\n", device, pool);
469 471 return (-1);
470 472 }
471 473
472 474 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID,
473 475 &record->zi_guid) == 0);
474 476 }
475 477
476 478 /*
477 479 * Device faults can take on three different forms:
478 480 * 1). delayed or hanging I/O
479 481 * 2). zfs label faults
480 482 * 3). generic disk faults
481 483 */
482 484 if (record->zi_timer != 0) {
483 485 record->zi_cmd = ZINJECT_DELAY_IO;
484 486 } else if (label_type != TYPE_INVAL) {
485 487 record->zi_cmd = ZINJECT_LABEL_FAULT;
486 488 } else {
487 489 record->zi_cmd = ZINJECT_DEVICE_FAULT;
488 490 }
489 491
490 492 switch (label_type) {
491 493 case TYPE_LABEL_UBERBLOCK:
492 494 record->zi_start = offsetof(vdev_label_t, vl_uberblock[0]);
493 495 record->zi_end = record->zi_start + VDEV_UBERBLOCK_RING - 1;
494 496 break;
495 497 case TYPE_LABEL_NVLIST:
496 498 record->zi_start = offsetof(vdev_label_t, vl_vdev_phys);
497 499 record->zi_end = record->zi_start + VDEV_PHYS_SIZE - 1;
498 500 break;
499 501 case TYPE_LABEL_PAD1:
500 502 record->zi_start = offsetof(vdev_label_t, vl_pad1);
501 503 record->zi_end = record->zi_start + VDEV_PAD_SIZE - 1;
502 504 break;
503 505 case TYPE_LABEL_PAD2:
504 506 record->zi_start = offsetof(vdev_label_t, vl_pad2);
505 507 record->zi_end = record->zi_start + VDEV_PAD_SIZE - 1;
506 508 break;
507 509 }
508 510 return (0);
509 511 }
|
↓ open down ↓ |
35 lines elided |
↑ open up ↑ |
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX