]>
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 | /* | |
428870ff | 22 | * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. |
34dc7c2f BB |
23 | */ |
24 | ||
34dc7c2f BB |
25 | /* |
26 | * The objective of this program is to provide a DMU/ZAP/SPA stress test | |
27 | * that runs entirely in userland, is easy to use, and easy to extend. | |
28 | * | |
29 | * The overall design of the ztest program is as follows: | |
30 | * | |
31 | * (1) For each major functional area (e.g. adding vdevs to a pool, | |
32 | * creating and destroying datasets, reading and writing objects, etc) | |
33 | * we have a simple routine to test that functionality. These | |
34 | * individual routines do not have to do anything "stressful". | |
35 | * | |
36 | * (2) We turn these simple functionality tests into a stress test by | |
37 | * running them all in parallel, with as many threads as desired, | |
38 | * and spread across as many datasets, objects, and vdevs as desired. | |
39 | * | |
40 | * (3) While all this is happening, we inject faults into the pool to | |
41 | * verify that self-healing data really works. | |
42 | * | |
43 | * (4) Every time we open a dataset, we change its checksum and compression | |
44 | * functions. Thus even individual objects vary from block to block | |
45 | * in which checksum they use and whether they're compressed. | |
46 | * | |
47 | * (5) To verify that we never lose on-disk consistency after a crash, | |
48 | * we run the entire test in a child of the main process. | |
49 | * At random times, the child self-immolates with a SIGKILL. | |
50 | * This is the software equivalent of pulling the power cord. | |
51 | * The parent then runs the test again, using the existing | |
52 | * storage pool, as many times as desired. | |
53 | * | |
54 | * (6) To verify that we don't have future leaks or temporal incursions, | |
55 | * many of the functional tests record the transaction group number | |
56 | * as part of their data. When reading old data, they verify that | |
57 | * the transaction group number is less than the current, open txg. | |
58 | * If you add a new test, please do this if applicable. | |
59 | * | |
1e33ac1e BB |
60 | * (7) Threads are created with a reduced stack size, for sanity checking. |
61 | * Therefore, it's important not to allocate huge buffers on the stack. | |
62 | * | |
34dc7c2f BB |
63 | * When run with no arguments, ztest runs for about five minutes and |
64 | * produces no output if successful. To get a little bit of information, | |
65 | * specify -V. To get more information, specify -VV, and so on. | |
66 | * | |
67 | * To turn this into an overnight stress test, use -T to specify run time. | |
68 | * | |
69 | * You can ask more more vdevs [-v], datasets [-d], or threads [-t] | |
70 | * to increase the pool capacity, fanout, and overall stress level. | |
71 | * | |
72 | * The -N(okill) option will suppress kills, so each child runs to completion. | |
73 | * This can be useful when you're trying to distinguish temporal incursions | |
74 | * from plain old race conditions. | |
75 | */ | |
76 | ||
77 | #include <sys/zfs_context.h> | |
78 | #include <sys/spa.h> | |
79 | #include <sys/dmu.h> | |
80 | #include <sys/txg.h> | |
9babb374 | 81 | #include <sys/dbuf.h> |
34dc7c2f | 82 | #include <sys/zap.h> |
34dc7c2f BB |
83 | #include <sys/dmu_objset.h> |
84 | #include <sys/poll.h> | |
85 | #include <sys/stat.h> | |
86 | #include <sys/time.h> | |
87 | #include <sys/wait.h> | |
88 | #include <sys/mman.h> | |
89 | #include <sys/resource.h> | |
90 | #include <sys/zio.h> | |
34dc7c2f | 91 | #include <sys/zil.h> |
428870ff | 92 | #include <sys/zil_impl.h> |
34dc7c2f | 93 | #include <sys/vdev_impl.h> |
b128c09f | 94 | #include <sys/vdev_file.h> |
34dc7c2f | 95 | #include <sys/spa_impl.h> |
428870ff | 96 | #include <sys/metaslab_impl.h> |
34dc7c2f | 97 | #include <sys/dsl_prop.h> |
9babb374 | 98 | #include <sys/dsl_dataset.h> |
428870ff BB |
99 | #include <sys/dsl_scan.h> |
100 | #include <sys/zio_checksum.h> | |
34dc7c2f BB |
101 | #include <sys/refcount.h> |
102 | #include <stdio.h> | |
103 | #include <stdio_ext.h> | |
104 | #include <stdlib.h> | |
105 | #include <unistd.h> | |
106 | #include <signal.h> | |
107 | #include <umem.h> | |
108 | #include <dlfcn.h> | |
109 | #include <ctype.h> | |
110 | #include <math.h> | |
111 | #include <sys/fs/zfs.h> | |
428870ff | 112 | #include <libnvpair.h> |
34dc7c2f BB |
113 | |
114 | static char cmdname[] = "ztest"; | |
115 | static char *zopt_pool = cmdname; | |
116 | ||
117 | static uint64_t zopt_vdevs = 5; | |
118 | static uint64_t zopt_vdevtime; | |
119 | static int zopt_ashift = SPA_MINBLOCKSHIFT; | |
120 | static int zopt_mirrors = 2; | |
121 | static int zopt_raidz = 4; | |
122 | static int zopt_raidz_parity = 1; | |
123 | static size_t zopt_vdev_size = SPA_MINDEVSIZE; | |
124 | static int zopt_datasets = 7; | |
125 | static int zopt_threads = 23; | |
126 | static uint64_t zopt_passtime = 60; /* 60 seconds */ | |
127 | static uint64_t zopt_killrate = 70; /* 70% kill rate */ | |
128 | static int zopt_verbose = 0; | |
129 | static int zopt_init = 1; | |
130 | static char *zopt_dir = "/tmp"; | |
131 | static uint64_t zopt_time = 300; /* 5 minutes */ | |
428870ff BB |
132 | static uint64_t zopt_maxloops = 50; /* max loops during spa_freeze() */ |
133 | ||
134 | #define BT_MAGIC 0x123456789abcdefULL | |
135 | #define MAXFAULTS() (MAX(zs->zs_mirrors, 1) * (zopt_raidz_parity + 1) - 1) | |
136 | ||
137 | enum ztest_io_type { | |
138 | ZTEST_IO_WRITE_TAG, | |
139 | ZTEST_IO_WRITE_PATTERN, | |
140 | ZTEST_IO_WRITE_ZEROES, | |
141 | ZTEST_IO_TRUNCATE, | |
142 | ZTEST_IO_SETATTR, | |
143 | ZTEST_IO_TYPES | |
144 | }; | |
34dc7c2f BB |
145 | |
146 | typedef struct ztest_block_tag { | |
428870ff | 147 | uint64_t bt_magic; |
34dc7c2f BB |
148 | uint64_t bt_objset; |
149 | uint64_t bt_object; | |
150 | uint64_t bt_offset; | |
428870ff | 151 | uint64_t bt_gen; |
34dc7c2f | 152 | uint64_t bt_txg; |
428870ff | 153 | uint64_t bt_crtxg; |
34dc7c2f BB |
154 | } ztest_block_tag_t; |
155 | ||
428870ff BB |
156 | typedef struct bufwad { |
157 | uint64_t bw_index; | |
158 | uint64_t bw_txg; | |
159 | uint64_t bw_data; | |
160 | } bufwad_t; | |
161 | ||
162 | /* | |
163 | * XXX -- fix zfs range locks to be generic so we can use them here. | |
164 | */ | |
165 | typedef enum { | |
166 | RL_READER, | |
167 | RL_WRITER, | |
168 | RL_APPEND | |
169 | } rl_type_t; | |
170 | ||
171 | typedef struct rll { | |
172 | void *rll_writer; | |
173 | int rll_readers; | |
1e33ac1e BB |
174 | kmutex_t rll_lock; |
175 | kcondvar_t rll_cv; | |
428870ff BB |
176 | } rll_t; |
177 | ||
178 | typedef struct rl { | |
179 | uint64_t rl_object; | |
180 | uint64_t rl_offset; | |
181 | uint64_t rl_size; | |
182 | rll_t *rl_lock; | |
183 | } rl_t; | |
184 | ||
185 | #define ZTEST_RANGE_LOCKS 64 | |
186 | #define ZTEST_OBJECT_LOCKS 64 | |
187 | ||
188 | /* | |
189 | * Object descriptor. Used as a template for object lookup/create/remove. | |
190 | */ | |
191 | typedef struct ztest_od { | |
192 | uint64_t od_dir; | |
193 | uint64_t od_object; | |
194 | dmu_object_type_t od_type; | |
195 | dmu_object_type_t od_crtype; | |
196 | uint64_t od_blocksize; | |
197 | uint64_t od_crblocksize; | |
198 | uint64_t od_gen; | |
199 | uint64_t od_crgen; | |
200 | char od_name[MAXNAMELEN]; | |
201 | } ztest_od_t; | |
34dc7c2f | 202 | |
428870ff BB |
203 | /* |
204 | * Per-dataset state. | |
205 | */ | |
206 | typedef struct ztest_ds { | |
207 | objset_t *zd_os; | |
208 | zilog_t *zd_zilog; | |
209 | uint64_t zd_seq; | |
210 | ztest_od_t *zd_od; /* debugging aid */ | |
211 | char zd_name[MAXNAMELEN]; | |
1e33ac1e | 212 | kmutex_t zd_dirobj_lock; |
428870ff BB |
213 | rll_t zd_object_lock[ZTEST_OBJECT_LOCKS]; |
214 | rll_t zd_range_lock[ZTEST_RANGE_LOCKS]; | |
215 | } ztest_ds_t; | |
216 | ||
217 | /* | |
218 | * Per-iteration state. | |
219 | */ | |
220 | typedef void ztest_func_t(ztest_ds_t *zd, uint64_t id); | |
221 | ||
222 | typedef struct ztest_info { | |
223 | ztest_func_t *zi_func; /* test function */ | |
224 | uint64_t zi_iters; /* iterations per execution */ | |
225 | uint64_t *zi_interval; /* execute every <interval> seconds */ | |
226 | uint64_t zi_call_count; /* per-pass count */ | |
227 | uint64_t zi_call_time; /* per-pass time */ | |
228 | uint64_t zi_call_next; /* next time to call this function */ | |
229 | } ztest_info_t; | |
34dc7c2f BB |
230 | |
231 | /* | |
232 | * Note: these aren't static because we want dladdr() to work. | |
233 | */ | |
234 | ztest_func_t ztest_dmu_read_write; | |
235 | ztest_func_t ztest_dmu_write_parallel; | |
236 | ztest_func_t ztest_dmu_object_alloc_free; | |
428870ff | 237 | ztest_func_t ztest_dmu_commit_callbacks; |
34dc7c2f BB |
238 | ztest_func_t ztest_zap; |
239 | ztest_func_t ztest_zap_parallel; | |
428870ff BB |
240 | ztest_func_t ztest_zil_commit; |
241 | ztest_func_t ztest_dmu_read_write_zcopy; | |
34dc7c2f | 242 | ztest_func_t ztest_dmu_objset_create_destroy; |
428870ff BB |
243 | ztest_func_t ztest_dmu_prealloc; |
244 | ztest_func_t ztest_fzap; | |
34dc7c2f | 245 | ztest_func_t ztest_dmu_snapshot_create_destroy; |
428870ff BB |
246 | ztest_func_t ztest_dsl_prop_get_set; |
247 | ztest_func_t ztest_spa_prop_get_set; | |
34dc7c2f BB |
248 | ztest_func_t ztest_spa_create_destroy; |
249 | ztest_func_t ztest_fault_inject; | |
428870ff BB |
250 | ztest_func_t ztest_ddt_repair; |
251 | ztest_func_t ztest_dmu_snapshot_hold; | |
b128c09f | 252 | ztest_func_t ztest_spa_rename; |
428870ff BB |
253 | ztest_func_t ztest_scrub; |
254 | ztest_func_t ztest_dsl_dataset_promote_busy; | |
34dc7c2f BB |
255 | ztest_func_t ztest_vdev_attach_detach; |
256 | ztest_func_t ztest_vdev_LUN_growth; | |
257 | ztest_func_t ztest_vdev_add_remove; | |
b128c09f | 258 | ztest_func_t ztest_vdev_aux_add_remove; |
428870ff | 259 | ztest_func_t ztest_split_pool; |
34dc7c2f | 260 | |
428870ff BB |
261 | uint64_t zopt_always = 0ULL * NANOSEC; /* all the time */ |
262 | uint64_t zopt_incessant = 1ULL * NANOSEC / 10; /* every 1/10 second */ | |
263 | uint64_t zopt_often = 1ULL * NANOSEC; /* every second */ | |
264 | uint64_t zopt_sometimes = 10ULL * NANOSEC; /* every 10 seconds */ | |
265 | uint64_t zopt_rarely = 60ULL * NANOSEC; /* every 60 seconds */ | |
34dc7c2f BB |
266 | |
267 | ztest_info_t ztest_info[] = { | |
268 | { ztest_dmu_read_write, 1, &zopt_always }, | |
428870ff | 269 | { ztest_dmu_write_parallel, 10, &zopt_always }, |
34dc7c2f | 270 | { ztest_dmu_object_alloc_free, 1, &zopt_always }, |
428870ff | 271 | { ztest_dmu_commit_callbacks, 1, &zopt_always }, |
34dc7c2f BB |
272 | { ztest_zap, 30, &zopt_always }, |
273 | { ztest_zap_parallel, 100, &zopt_always }, | |
428870ff BB |
274 | { ztest_split_pool, 1, &zopt_always }, |
275 | { ztest_zil_commit, 1, &zopt_incessant }, | |
276 | { ztest_dmu_read_write_zcopy, 1, &zopt_often }, | |
277 | { ztest_dmu_objset_create_destroy, 1, &zopt_often }, | |
278 | { ztest_dsl_prop_get_set, 1, &zopt_often }, | |
279 | { ztest_spa_prop_get_set, 1, &zopt_sometimes }, | |
280 | #if 0 | |
281 | { ztest_dmu_prealloc, 1, &zopt_sometimes }, | |
282 | #endif | |
283 | { ztest_fzap, 1, &zopt_sometimes }, | |
284 | { ztest_dmu_snapshot_create_destroy, 1, &zopt_sometimes }, | |
285 | { ztest_spa_create_destroy, 1, &zopt_sometimes }, | |
34dc7c2f | 286 | { ztest_fault_inject, 1, &zopt_sometimes }, |
428870ff BB |
287 | { ztest_ddt_repair, 1, &zopt_sometimes }, |
288 | { ztest_dmu_snapshot_hold, 1, &zopt_sometimes }, | |
34dc7c2f | 289 | { ztest_spa_rename, 1, &zopt_rarely }, |
428870ff | 290 | { ztest_scrub, 1, &zopt_rarely }, |
9babb374 | 291 | { ztest_dsl_dataset_promote_busy, 1, &zopt_rarely }, |
428870ff BB |
292 | { ztest_vdev_attach_detach, 1, &zopt_rarely }, |
293 | { ztest_vdev_LUN_growth, 1, &zopt_rarely }, | |
294 | { ztest_vdev_add_remove, 1, &zopt_vdevtime }, | |
b128c09f | 295 | { ztest_vdev_aux_add_remove, 1, &zopt_vdevtime }, |
34dc7c2f BB |
296 | }; |
297 | ||
298 | #define ZTEST_FUNCS (sizeof (ztest_info) / sizeof (ztest_info_t)) | |
299 | ||
428870ff BB |
300 | /* |
301 | * The following struct is used to hold a list of uncalled commit callbacks. | |
302 | * The callbacks are ordered by txg number. | |
303 | */ | |
304 | typedef struct ztest_cb_list { | |
1e33ac1e BB |
305 | kmutex_t zcl_callbacks_lock; |
306 | list_t zcl_callbacks; | |
428870ff | 307 | } ztest_cb_list_t; |
34dc7c2f BB |
308 | |
309 | /* | |
310 | * Stuff we need to share writably between parent and child. | |
311 | */ | |
312 | typedef struct ztest_shared { | |
428870ff BB |
313 | char *zs_pool; |
314 | spa_t *zs_spa; | |
315 | hrtime_t zs_proc_start; | |
316 | hrtime_t zs_proc_stop; | |
317 | hrtime_t zs_thread_start; | |
318 | hrtime_t zs_thread_stop; | |
319 | hrtime_t zs_thread_kill; | |
34dc7c2f | 320 | uint64_t zs_enospc_count; |
428870ff BB |
321 | uint64_t zs_vdev_next_leaf; |
322 | uint64_t zs_vdev_aux; | |
34dc7c2f BB |
323 | uint64_t zs_alloc; |
324 | uint64_t zs_space; | |
1e33ac1e BB |
325 | kmutex_t zs_vdev_lock; |
326 | krwlock_t zs_name_lock; | |
34dc7c2f | 327 | ztest_info_t zs_info[ZTEST_FUNCS]; |
428870ff BB |
328 | uint64_t zs_splits; |
329 | uint64_t zs_mirrors; | |
330 | ztest_ds_t zs_zd[]; | |
34dc7c2f BB |
331 | } ztest_shared_t; |
332 | ||
428870ff BB |
333 | #define ID_PARALLEL -1ULL |
334 | ||
34dc7c2f | 335 | static char ztest_dev_template[] = "%s/%s.%llua"; |
b128c09f | 336 | static char ztest_aux_template[] = "%s/%s.%s.%llu"; |
428870ff BB |
337 | ztest_shared_t *ztest_shared; |
338 | uint64_t *ztest_seq; | |
34dc7c2f BB |
339 | |
340 | static int ztest_random_fd; | |
341 | static int ztest_dump_core = 1; | |
342 | ||
b128c09f | 343 | static boolean_t ztest_exiting; |
34dc7c2f | 344 | |
428870ff BB |
345 | /* Global commit callback list */ |
346 | static ztest_cb_list_t zcl; | |
090ff092 RC |
347 | /* Commit cb delay */ |
348 | static uint64_t zc_min_txg_delay = UINT64_MAX; | |
349 | static int zc_cb_counter = 0; | |
350 | ||
351 | /* | |
352 | * Minimum number of commit callbacks that need to be registered for us to check | |
353 | * whether the minimum txg delay is acceptable. | |
354 | */ | |
355 | #define ZTEST_COMMIT_CB_MIN_REG 100 | |
356 | ||
357 | /* | |
358 | * If a number of txgs equal to this threshold have been created after a commit | |
359 | * callback has been registered but not called, then we assume there is an | |
360 | * implementation bug. | |
361 | */ | |
362 | #define ZTEST_COMMIT_CB_THRESH (TXG_CONCURRENT_STATES + 1000) | |
428870ff | 363 | |
34dc7c2f | 364 | extern uint64_t metaslab_gang_bang; |
9babb374 | 365 | extern uint64_t metaslab_df_alloc_threshold; |
428870ff | 366 | static uint64_t metaslab_sz; |
34dc7c2f | 367 | |
428870ff BB |
368 | enum ztest_object { |
369 | ZTEST_META_DNODE = 0, | |
370 | ZTEST_DIROBJ, | |
371 | ZTEST_OBJECTS | |
372 | }; | |
34dc7c2f BB |
373 | |
374 | static void usage(boolean_t) __NORETURN; | |
375 | ||
376 | /* | |
377 | * These libumem hooks provide a reasonable set of defaults for the allocator's | |
378 | * debugging facilities. | |
379 | */ | |
380 | const char * | |
0bc8fd78 | 381 | _umem_debug_init(void) |
34dc7c2f BB |
382 | { |
383 | return ("default,verbose"); /* $UMEM_DEBUG setting */ | |
384 | } | |
385 | ||
386 | const char * | |
387 | _umem_logging_init(void) | |
388 | { | |
389 | return ("fail,contents"); /* $UMEM_LOGGING setting */ | |
390 | } | |
391 | ||
392 | #define FATAL_MSG_SZ 1024 | |
393 | ||
394 | char *fatal_msg; | |
395 | ||
396 | static void | |
397 | fatal(int do_perror, char *message, ...) | |
398 | { | |
399 | va_list args; | |
400 | int save_errno = errno; | |
40b84e7a | 401 | char *buf; |
34dc7c2f BB |
402 | |
403 | (void) fflush(stdout); | |
40b84e7a | 404 | buf = umem_alloc(FATAL_MSG_SZ, UMEM_NOFAIL); |
34dc7c2f BB |
405 | |
406 | va_start(args, message); | |
407 | (void) sprintf(buf, "ztest: "); | |
408 | /* LINTED */ | |
409 | (void) vsprintf(buf + strlen(buf), message, args); | |
410 | va_end(args); | |
411 | if (do_perror) { | |
412 | (void) snprintf(buf + strlen(buf), FATAL_MSG_SZ - strlen(buf), | |
413 | ": %s", strerror(save_errno)); | |
414 | } | |
415 | (void) fprintf(stderr, "%s\n", buf); | |
416 | fatal_msg = buf; /* to ease debugging */ | |
417 | if (ztest_dump_core) | |
418 | abort(); | |
419 | exit(3); | |
420 | } | |
421 | ||
422 | static int | |
423 | str2shift(const char *buf) | |
424 | { | |
425 | const char *ends = "BKMGTPEZ"; | |
426 | int i; | |
427 | ||
428 | if (buf[0] == '\0') | |
429 | return (0); | |
430 | for (i = 0; i < strlen(ends); i++) { | |
431 | if (toupper(buf[0]) == ends[i]) | |
432 | break; | |
433 | } | |
434 | if (i == strlen(ends)) { | |
435 | (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", | |
436 | buf); | |
437 | usage(B_FALSE); | |
438 | } | |
439 | if (buf[1] == '\0' || (toupper(buf[1]) == 'B' && buf[2] == '\0')) { | |
440 | return (10*i); | |
441 | } | |
442 | (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", buf); | |
443 | usage(B_FALSE); | |
444 | /* NOTREACHED */ | |
445 | } | |
446 | ||
447 | static uint64_t | |
448 | nicenumtoull(const char *buf) | |
449 | { | |
450 | char *end; | |
451 | uint64_t val; | |
452 | ||
453 | val = strtoull(buf, &end, 0); | |
454 | if (end == buf) { | |
455 | (void) fprintf(stderr, "ztest: bad numeric value: %s\n", buf); | |
456 | usage(B_FALSE); | |
457 | } else if (end[0] == '.') { | |
458 | double fval = strtod(buf, &end); | |
459 | fval *= pow(2, str2shift(end)); | |
460 | if (fval > UINT64_MAX) { | |
461 | (void) fprintf(stderr, "ztest: value too large: %s\n", | |
462 | buf); | |
463 | usage(B_FALSE); | |
464 | } | |
465 | val = (uint64_t)fval; | |
466 | } else { | |
467 | int shift = str2shift(end); | |
468 | if (shift >= 64 || (val << shift) >> shift != val) { | |
469 | (void) fprintf(stderr, "ztest: value too large: %s\n", | |
470 | buf); | |
471 | usage(B_FALSE); | |
472 | } | |
473 | val <<= shift; | |
474 | } | |
475 | return (val); | |
476 | } | |
477 | ||
478 | static void | |
479 | usage(boolean_t requested) | |
480 | { | |
481 | char nice_vdev_size[10]; | |
482 | char nice_gang_bang[10]; | |
483 | FILE *fp = requested ? stdout : stderr; | |
484 | ||
485 | nicenum(zopt_vdev_size, nice_vdev_size); | |
486 | nicenum(metaslab_gang_bang, nice_gang_bang); | |
487 | ||
488 | (void) fprintf(fp, "Usage: %s\n" | |
489 | "\t[-v vdevs (default: %llu)]\n" | |
490 | "\t[-s size_of_each_vdev (default: %s)]\n" | |
428870ff | 491 | "\t[-a alignment_shift (default: %d)] use 0 for random\n" |
34dc7c2f BB |
492 | "\t[-m mirror_copies (default: %d)]\n" |
493 | "\t[-r raidz_disks (default: %d)]\n" | |
494 | "\t[-R raidz_parity (default: %d)]\n" | |
495 | "\t[-d datasets (default: %d)]\n" | |
496 | "\t[-t threads (default: %d)]\n" | |
497 | "\t[-g gang_block_threshold (default: %s)]\n" | |
428870ff BB |
498 | "\t[-i init_count (default: %d)] initialize pool i times\n" |
499 | "\t[-k kill_percentage (default: %llu%%)]\n" | |
34dc7c2f | 500 | "\t[-p pool_name (default: %s)]\n" |
428870ff BB |
501 | "\t[-f dir (default: %s)] file directory for vdev files\n" |
502 | "\t[-V] verbose (use multiple times for ever more blather)\n" | |
503 | "\t[-E] use existing pool instead of creating new one\n" | |
504 | "\t[-T time (default: %llu sec)] total run time\n" | |
505 | "\t[-F freezeloops (default: %llu)] max loops in spa_freeze()\n" | |
506 | "\t[-P passtime (default: %llu sec)] time per pass\n" | |
34dc7c2f BB |
507 | "\t[-h] (print help)\n" |
508 | "", | |
509 | cmdname, | |
510 | (u_longlong_t)zopt_vdevs, /* -v */ | |
511 | nice_vdev_size, /* -s */ | |
512 | zopt_ashift, /* -a */ | |
513 | zopt_mirrors, /* -m */ | |
514 | zopt_raidz, /* -r */ | |
515 | zopt_raidz_parity, /* -R */ | |
516 | zopt_datasets, /* -d */ | |
517 | zopt_threads, /* -t */ | |
518 | nice_gang_bang, /* -g */ | |
519 | zopt_init, /* -i */ | |
520 | (u_longlong_t)zopt_killrate, /* -k */ | |
521 | zopt_pool, /* -p */ | |
522 | zopt_dir, /* -f */ | |
523 | (u_longlong_t)zopt_time, /* -T */ | |
428870ff | 524 | (u_longlong_t)zopt_maxloops, /* -F */ |
b128c09f | 525 | (u_longlong_t)zopt_passtime); /* -P */ |
34dc7c2f BB |
526 | exit(requested ? 0 : 1); |
527 | } | |
528 | ||
34dc7c2f BB |
529 | static void |
530 | process_options(int argc, char **argv) | |
531 | { | |
532 | int opt; | |
533 | uint64_t value; | |
534 | ||
535 | /* By default, test gang blocks for blocks 32K and greater */ | |
536 | metaslab_gang_bang = 32 << 10; | |
537 | ||
34dc7c2f | 538 | while ((opt = getopt(argc, argv, |
428870ff | 539 | "v:s:a:m:r:R:d:t:g:i:k:p:f:VET:P:hF:")) != EOF) { |
34dc7c2f BB |
540 | value = 0; |
541 | switch (opt) { | |
542 | case 'v': | |
543 | case 's': | |
544 | case 'a': | |
545 | case 'm': | |
546 | case 'r': | |
547 | case 'R': | |
548 | case 'd': | |
549 | case 't': | |
550 | case 'g': | |
551 | case 'i': | |
552 | case 'k': | |
553 | case 'T': | |
554 | case 'P': | |
428870ff | 555 | case 'F': |
34dc7c2f BB |
556 | value = nicenumtoull(optarg); |
557 | } | |
558 | switch (opt) { | |
559 | case 'v': | |
560 | zopt_vdevs = value; | |
561 | break; | |
562 | case 's': | |
563 | zopt_vdev_size = MAX(SPA_MINDEVSIZE, value); | |
564 | break; | |
565 | case 'a': | |
566 | zopt_ashift = value; | |
567 | break; | |
568 | case 'm': | |
569 | zopt_mirrors = value; | |
570 | break; | |
571 | case 'r': | |
572 | zopt_raidz = MAX(1, value); | |
573 | break; | |
574 | case 'R': | |
45d1cae3 | 575 | zopt_raidz_parity = MIN(MAX(value, 1), 3); |
34dc7c2f BB |
576 | break; |
577 | case 'd': | |
578 | zopt_datasets = MAX(1, value); | |
579 | break; | |
580 | case 't': | |
581 | zopt_threads = MAX(1, value); | |
582 | break; | |
583 | case 'g': | |
584 | metaslab_gang_bang = MAX(SPA_MINBLOCKSIZE << 1, value); | |
585 | break; | |
586 | case 'i': | |
587 | zopt_init = value; | |
588 | break; | |
589 | case 'k': | |
590 | zopt_killrate = value; | |
591 | break; | |
592 | case 'p': | |
593 | zopt_pool = strdup(optarg); | |
594 | break; | |
595 | case 'f': | |
596 | zopt_dir = strdup(optarg); | |
597 | break; | |
598 | case 'V': | |
599 | zopt_verbose++; | |
600 | break; | |
601 | case 'E': | |
602 | zopt_init = 0; | |
603 | break; | |
604 | case 'T': | |
605 | zopt_time = value; | |
606 | break; | |
607 | case 'P': | |
608 | zopt_passtime = MAX(1, value); | |
609 | break; | |
428870ff BB |
610 | case 'F': |
611 | zopt_maxloops = MAX(1, value); | |
612 | break; | |
34dc7c2f BB |
613 | case 'h': |
614 | usage(B_TRUE); | |
615 | break; | |
616 | case '?': | |
617 | default: | |
618 | usage(B_FALSE); | |
619 | break; | |
620 | } | |
621 | } | |
622 | ||
623 | zopt_raidz_parity = MIN(zopt_raidz_parity, zopt_raidz - 1); | |
624 | ||
428870ff BB |
625 | zopt_vdevtime = (zopt_vdevs > 0 ? zopt_time * NANOSEC / zopt_vdevs : |
626 | UINT64_MAX >> 2); | |
627 | } | |
628 | ||
629 | static void | |
630 | ztest_kill(ztest_shared_t *zs) | |
631 | { | |
632 | zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(zs->zs_spa)); | |
633 | zs->zs_space = metaslab_class_get_space(spa_normal_class(zs->zs_spa)); | |
634 | (void) kill(getpid(), SIGKILL); | |
635 | } | |
636 | ||
637 | static uint64_t | |
638 | ztest_random(uint64_t range) | |
639 | { | |
640 | uint64_t r; | |
641 | ||
642 | if (range == 0) | |
643 | return (0); | |
644 | ||
645 | if (read(ztest_random_fd, &r, sizeof (r)) != sizeof (r)) | |
646 | fatal(1, "short read from /dev/urandom"); | |
647 | ||
648 | return (r % range); | |
649 | } | |
650 | ||
651 | /* ARGSUSED */ | |
652 | static void | |
653 | ztest_record_enospc(const char *s) | |
654 | { | |
655 | ztest_shared->zs_enospc_count++; | |
34dc7c2f BB |
656 | } |
657 | ||
658 | static uint64_t | |
659 | ztest_get_ashift(void) | |
660 | { | |
661 | if (zopt_ashift == 0) | |
662 | return (SPA_MINBLOCKSHIFT + ztest_random(3)); | |
663 | return (zopt_ashift); | |
664 | } | |
665 | ||
666 | static nvlist_t * | |
b128c09f | 667 | make_vdev_file(char *path, char *aux, size_t size, uint64_t ashift) |
34dc7c2f | 668 | { |
40b84e7a | 669 | char *pathbuf; |
34dc7c2f | 670 | uint64_t vdev; |
34dc7c2f BB |
671 | nvlist_t *file; |
672 | ||
40b84e7a BB |
673 | pathbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
674 | ||
b128c09f BB |
675 | if (ashift == 0) |
676 | ashift = ztest_get_ashift(); | |
677 | ||
678 | if (path == NULL) { | |
679 | path = pathbuf; | |
680 | ||
681 | if (aux != NULL) { | |
682 | vdev = ztest_shared->zs_vdev_aux; | |
683 | (void) sprintf(path, ztest_aux_template, | |
684 | zopt_dir, zopt_pool, aux, vdev); | |
685 | } else { | |
428870ff | 686 | vdev = ztest_shared->zs_vdev_next_leaf++; |
b128c09f BB |
687 | (void) sprintf(path, ztest_dev_template, |
688 | zopt_dir, zopt_pool, vdev); | |
689 | } | |
690 | } | |
34dc7c2f | 691 | |
b128c09f BB |
692 | if (size != 0) { |
693 | int fd = open(path, O_RDWR | O_CREAT | O_TRUNC, 0666); | |
34dc7c2f | 694 | if (fd == -1) |
b128c09f | 695 | fatal(1, "can't open %s", path); |
34dc7c2f | 696 | if (ftruncate(fd, size) != 0) |
b128c09f | 697 | fatal(1, "can't ftruncate %s", path); |
34dc7c2f BB |
698 | (void) close(fd); |
699 | } | |
700 | ||
701 | VERIFY(nvlist_alloc(&file, NV_UNIQUE_NAME, 0) == 0); | |
702 | VERIFY(nvlist_add_string(file, ZPOOL_CONFIG_TYPE, VDEV_TYPE_FILE) == 0); | |
b128c09f | 703 | VERIFY(nvlist_add_string(file, ZPOOL_CONFIG_PATH, path) == 0); |
34dc7c2f | 704 | VERIFY(nvlist_add_uint64(file, ZPOOL_CONFIG_ASHIFT, ashift) == 0); |
40b84e7a | 705 | umem_free(pathbuf, MAXPATHLEN); |
34dc7c2f BB |
706 | |
707 | return (file); | |
708 | } | |
709 | ||
710 | static nvlist_t * | |
b128c09f | 711 | make_vdev_raidz(char *path, char *aux, size_t size, uint64_t ashift, int r) |
34dc7c2f BB |
712 | { |
713 | nvlist_t *raidz, **child; | |
714 | int c; | |
715 | ||
716 | if (r < 2) | |
b128c09f | 717 | return (make_vdev_file(path, aux, size, ashift)); |
34dc7c2f BB |
718 | child = umem_alloc(r * sizeof (nvlist_t *), UMEM_NOFAIL); |
719 | ||
720 | for (c = 0; c < r; c++) | |
b128c09f | 721 | child[c] = make_vdev_file(path, aux, size, ashift); |
34dc7c2f BB |
722 | |
723 | VERIFY(nvlist_alloc(&raidz, NV_UNIQUE_NAME, 0) == 0); | |
724 | VERIFY(nvlist_add_string(raidz, ZPOOL_CONFIG_TYPE, | |
725 | VDEV_TYPE_RAIDZ) == 0); | |
726 | VERIFY(nvlist_add_uint64(raidz, ZPOOL_CONFIG_NPARITY, | |
727 | zopt_raidz_parity) == 0); | |
728 | VERIFY(nvlist_add_nvlist_array(raidz, ZPOOL_CONFIG_CHILDREN, | |
729 | child, r) == 0); | |
730 | ||
731 | for (c = 0; c < r; c++) | |
732 | nvlist_free(child[c]); | |
733 | ||
734 | umem_free(child, r * sizeof (nvlist_t *)); | |
735 | ||
736 | return (raidz); | |
737 | } | |
738 | ||
739 | static nvlist_t * | |
b128c09f BB |
740 | make_vdev_mirror(char *path, char *aux, size_t size, uint64_t ashift, |
741 | int r, int m) | |
34dc7c2f BB |
742 | { |
743 | nvlist_t *mirror, **child; | |
744 | int c; | |
745 | ||
746 | if (m < 1) | |
b128c09f | 747 | return (make_vdev_raidz(path, aux, size, ashift, r)); |
34dc7c2f BB |
748 | |
749 | child = umem_alloc(m * sizeof (nvlist_t *), UMEM_NOFAIL); | |
750 | ||
751 | for (c = 0; c < m; c++) | |
b128c09f | 752 | child[c] = make_vdev_raidz(path, aux, size, ashift, r); |
34dc7c2f BB |
753 | |
754 | VERIFY(nvlist_alloc(&mirror, NV_UNIQUE_NAME, 0) == 0); | |
755 | VERIFY(nvlist_add_string(mirror, ZPOOL_CONFIG_TYPE, | |
756 | VDEV_TYPE_MIRROR) == 0); | |
757 | VERIFY(nvlist_add_nvlist_array(mirror, ZPOOL_CONFIG_CHILDREN, | |
758 | child, m) == 0); | |
34dc7c2f BB |
759 | |
760 | for (c = 0; c < m; c++) | |
761 | nvlist_free(child[c]); | |
762 | ||
763 | umem_free(child, m * sizeof (nvlist_t *)); | |
764 | ||
765 | return (mirror); | |
766 | } | |
767 | ||
768 | static nvlist_t * | |
b128c09f BB |
769 | make_vdev_root(char *path, char *aux, size_t size, uint64_t ashift, |
770 | int log, int r, int m, int t) | |
34dc7c2f BB |
771 | { |
772 | nvlist_t *root, **child; | |
773 | int c; | |
774 | ||
775 | ASSERT(t > 0); | |
776 | ||
777 | child = umem_alloc(t * sizeof (nvlist_t *), UMEM_NOFAIL); | |
778 | ||
b128c09f BB |
779 | for (c = 0; c < t; c++) { |
780 | child[c] = make_vdev_mirror(path, aux, size, ashift, r, m); | |
781 | VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_IS_LOG, | |
782 | log) == 0); | |
783 | } | |
34dc7c2f BB |
784 | |
785 | VERIFY(nvlist_alloc(&root, NV_UNIQUE_NAME, 0) == 0); | |
786 | VERIFY(nvlist_add_string(root, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT) == 0); | |
b128c09f | 787 | VERIFY(nvlist_add_nvlist_array(root, aux ? aux : ZPOOL_CONFIG_CHILDREN, |
34dc7c2f BB |
788 | child, t) == 0); |
789 | ||
790 | for (c = 0; c < t; c++) | |
791 | nvlist_free(child[c]); | |
792 | ||
793 | umem_free(child, t * sizeof (nvlist_t *)); | |
794 | ||
795 | return (root); | |
796 | } | |
797 | ||
428870ff BB |
798 | static int |
799 | ztest_random_blocksize(void) | |
34dc7c2f | 800 | { |
428870ff BB |
801 | return (1 << (SPA_MINBLOCKSHIFT + |
802 | ztest_random(SPA_MAXBLOCKSHIFT - SPA_MINBLOCKSHIFT + 1))); | |
803 | } | |
34dc7c2f | 804 | |
428870ff BB |
805 | static int |
806 | ztest_random_ibshift(void) | |
807 | { | |
808 | return (DN_MIN_INDBLKSHIFT + | |
809 | ztest_random(DN_MAX_INDBLKSHIFT - DN_MIN_INDBLKSHIFT + 1)); | |
34dc7c2f BB |
810 | } |
811 | ||
428870ff BB |
812 | static uint64_t |
813 | ztest_random_vdev_top(spa_t *spa, boolean_t log_ok) | |
34dc7c2f | 814 | { |
428870ff BB |
815 | uint64_t top; |
816 | vdev_t *rvd = spa->spa_root_vdev; | |
817 | vdev_t *tvd; | |
34dc7c2f | 818 | |
428870ff | 819 | ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); |
34dc7c2f | 820 | |
428870ff BB |
821 | do { |
822 | top = ztest_random(rvd->vdev_children); | |
823 | tvd = rvd->vdev_child[top]; | |
824 | } while (tvd->vdev_ishole || (tvd->vdev_islog && !log_ok) || | |
825 | tvd->vdev_mg == NULL || tvd->vdev_mg->mg_class == NULL); | |
34dc7c2f | 826 | |
428870ff | 827 | return (top); |
34dc7c2f BB |
828 | } |
829 | ||
428870ff BB |
830 | static uint64_t |
831 | ztest_random_dsl_prop(zfs_prop_t prop) | |
34dc7c2f | 832 | { |
428870ff BB |
833 | uint64_t value; |
834 | ||
835 | do { | |
836 | value = zfs_prop_random_value(prop, ztest_random(-1ULL)); | |
837 | } while (prop == ZFS_PROP_CHECKSUM && value == ZIO_CHECKSUM_OFF); | |
838 | ||
839 | return (value); | |
34dc7c2f BB |
840 | } |
841 | ||
34dc7c2f | 842 | static int |
428870ff BB |
843 | ztest_dsl_prop_set_uint64(char *osname, zfs_prop_t prop, uint64_t value, |
844 | boolean_t inherit) | |
34dc7c2f | 845 | { |
428870ff BB |
846 | const char *propname = zfs_prop_to_name(prop); |
847 | const char *valname; | |
40b84e7a | 848 | char *setpoint; |
428870ff | 849 | uint64_t curval; |
34dc7c2f BB |
850 | int error; |
851 | ||
428870ff BB |
852 | error = dsl_prop_set(osname, propname, |
853 | (inherit ? ZPROP_SRC_NONE : ZPROP_SRC_LOCAL), | |
854 | sizeof (value), 1, &value); | |
34dc7c2f | 855 | |
428870ff BB |
856 | if (error == ENOSPC) { |
857 | ztest_record_enospc(FTAG); | |
34dc7c2f BB |
858 | return (error); |
859 | } | |
34dc7c2f | 860 | ASSERT3U(error, ==, 0); |
34dc7c2f | 861 | |
40b84e7a | 862 | setpoint = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
428870ff BB |
863 | VERIFY3U(dsl_prop_get(osname, propname, sizeof (curval), |
864 | 1, &curval, setpoint), ==, 0); | |
865 | ||
866 | if (zopt_verbose >= 6) { | |
867 | VERIFY(zfs_prop_index_to_string(prop, curval, &valname) == 0); | |
868 | (void) printf("%s %s = %s at '%s'\n", | |
869 | osname, propname, valname, setpoint); | |
34dc7c2f | 870 | } |
40b84e7a | 871 | umem_free(setpoint, MAXPATHLEN); |
34dc7c2f BB |
872 | |
873 | return (error); | |
874 | } | |
875 | ||
876 | static int | |
428870ff | 877 | ztest_spa_prop_set_uint64(ztest_shared_t *zs, zpool_prop_t prop, uint64_t value) |
34dc7c2f | 878 | { |
428870ff BB |
879 | spa_t *spa = zs->zs_spa; |
880 | nvlist_t *props = NULL; | |
34dc7c2f BB |
881 | int error; |
882 | ||
428870ff BB |
883 | VERIFY(nvlist_alloc(&props, NV_UNIQUE_NAME, 0) == 0); |
884 | VERIFY(nvlist_add_uint64(props, zpool_prop_to_name(prop), value) == 0); | |
34dc7c2f | 885 | |
428870ff BB |
886 | error = spa_prop_set(spa, props); |
887 | ||
888 | nvlist_free(props); | |
889 | ||
890 | if (error == ENOSPC) { | |
891 | ztest_record_enospc(FTAG); | |
34dc7c2f BB |
892 | return (error); |
893 | } | |
428870ff | 894 | ASSERT3U(error, ==, 0); |
34dc7c2f BB |
895 | |
896 | return (error); | |
897 | } | |
898 | ||
428870ff BB |
899 | static void |
900 | ztest_rll_init(rll_t *rll) | |
901 | { | |
902 | rll->rll_writer = NULL; | |
903 | rll->rll_readers = 0; | |
1e33ac1e BB |
904 | mutex_init(&rll->rll_lock, NULL, MUTEX_DEFAULT, NULL); |
905 | cv_init(&rll->rll_cv, NULL, CV_DEFAULT, NULL); | |
428870ff | 906 | } |
34dc7c2f | 907 | |
428870ff BB |
908 | static void |
909 | ztest_rll_destroy(rll_t *rll) | |
34dc7c2f | 910 | { |
428870ff BB |
911 | ASSERT(rll->rll_writer == NULL); |
912 | ASSERT(rll->rll_readers == 0); | |
1e33ac1e BB |
913 | mutex_destroy(&rll->rll_lock); |
914 | cv_destroy(&rll->rll_cv); | |
428870ff | 915 | } |
34dc7c2f | 916 | |
428870ff BB |
917 | static void |
918 | ztest_rll_lock(rll_t *rll, rl_type_t type) | |
919 | { | |
1e33ac1e | 920 | mutex_enter(&rll->rll_lock); |
34dc7c2f | 921 | |
428870ff BB |
922 | if (type == RL_READER) { |
923 | while (rll->rll_writer != NULL) | |
1e33ac1e | 924 | (void) cv_wait(&rll->rll_cv, &rll->rll_lock); |
428870ff BB |
925 | rll->rll_readers++; |
926 | } else { | |
927 | while (rll->rll_writer != NULL || rll->rll_readers) | |
1e33ac1e | 928 | (void) cv_wait(&rll->rll_cv, &rll->rll_lock); |
428870ff BB |
929 | rll->rll_writer = curthread; |
930 | } | |
34dc7c2f | 931 | |
1e33ac1e | 932 | mutex_exit(&rll->rll_lock); |
428870ff | 933 | } |
34dc7c2f | 934 | |
428870ff BB |
935 | static void |
936 | ztest_rll_unlock(rll_t *rll) | |
937 | { | |
1e33ac1e | 938 | mutex_enter(&rll->rll_lock); |
34dc7c2f | 939 | |
428870ff BB |
940 | if (rll->rll_writer) { |
941 | ASSERT(rll->rll_readers == 0); | |
942 | rll->rll_writer = NULL; | |
943 | } else { | |
944 | ASSERT(rll->rll_readers != 0); | |
945 | ASSERT(rll->rll_writer == NULL); | |
946 | rll->rll_readers--; | |
947 | } | |
34dc7c2f | 948 | |
428870ff | 949 | if (rll->rll_writer == NULL && rll->rll_readers == 0) |
1e33ac1e | 950 | cv_broadcast(&rll->rll_cv); |
428870ff | 951 | |
1e33ac1e | 952 | mutex_exit(&rll->rll_lock); |
34dc7c2f BB |
953 | } |
954 | ||
428870ff BB |
955 | static void |
956 | ztest_object_lock(ztest_ds_t *zd, uint64_t object, rl_type_t type) | |
b128c09f | 957 | { |
428870ff | 958 | rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; |
b128c09f | 959 | |
428870ff BB |
960 | ztest_rll_lock(rll, type); |
961 | } | |
b128c09f | 962 | |
428870ff BB |
963 | static void |
964 | ztest_object_unlock(ztest_ds_t *zd, uint64_t object) | |
965 | { | |
966 | rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; | |
b128c09f | 967 | |
428870ff | 968 | ztest_rll_unlock(rll); |
b128c09f BB |
969 | } |
970 | ||
428870ff BB |
971 | static rl_t * |
972 | ztest_range_lock(ztest_ds_t *zd, uint64_t object, uint64_t offset, | |
973 | uint64_t size, rl_type_t type) | |
34dc7c2f | 974 | { |
428870ff BB |
975 | uint64_t hash = object ^ (offset % (ZTEST_RANGE_LOCKS + 1)); |
976 | rll_t *rll = &zd->zd_range_lock[hash & (ZTEST_RANGE_LOCKS - 1)]; | |
977 | rl_t *rl; | |
34dc7c2f | 978 | |
428870ff BB |
979 | rl = umem_alloc(sizeof (*rl), UMEM_NOFAIL); |
980 | rl->rl_object = object; | |
981 | rl->rl_offset = offset; | |
982 | rl->rl_size = size; | |
983 | rl->rl_lock = rll; | |
34dc7c2f | 984 | |
428870ff BB |
985 | ztest_rll_lock(rll, type); |
986 | ||
987 | return (rl); | |
988 | } | |
34dc7c2f | 989 | |
428870ff BB |
990 | static void |
991 | ztest_range_unlock(rl_t *rl) | |
992 | { | |
993 | rll_t *rll = rl->rl_lock; | |
34dc7c2f | 994 | |
428870ff | 995 | ztest_rll_unlock(rll); |
34dc7c2f | 996 | |
428870ff BB |
997 | umem_free(rl, sizeof (*rl)); |
998 | } | |
34dc7c2f | 999 | |
428870ff BB |
1000 | static void |
1001 | ztest_zd_init(ztest_ds_t *zd, objset_t *os) | |
1002 | { | |
1003 | zd->zd_os = os; | |
1004 | zd->zd_zilog = dmu_objset_zil(os); | |
1005 | zd->zd_seq = 0; | |
1006 | dmu_objset_name(os, zd->zd_name); | |
d6320ddb | 1007 | int l; |
428870ff | 1008 | |
1e33ac1e | 1009 | mutex_init(&zd->zd_dirobj_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 1010 | |
d6320ddb | 1011 | for (l = 0; l < ZTEST_OBJECT_LOCKS; l++) |
428870ff | 1012 | ztest_rll_init(&zd->zd_object_lock[l]); |
34dc7c2f | 1013 | |
d6320ddb | 1014 | for (l = 0; l < ZTEST_RANGE_LOCKS; l++) |
428870ff | 1015 | ztest_rll_init(&zd->zd_range_lock[l]); |
34dc7c2f BB |
1016 | } |
1017 | ||
428870ff BB |
1018 | static void |
1019 | ztest_zd_fini(ztest_ds_t *zd) | |
34dc7c2f | 1020 | { |
d6320ddb BB |
1021 | int l; |
1022 | ||
1e33ac1e | 1023 | mutex_destroy(&zd->zd_dirobj_lock); |
34dc7c2f | 1024 | |
d6320ddb | 1025 | for (l = 0; l < ZTEST_OBJECT_LOCKS; l++) |
428870ff | 1026 | ztest_rll_destroy(&zd->zd_object_lock[l]); |
b128c09f | 1027 | |
d6320ddb | 1028 | for (l = 0; l < ZTEST_RANGE_LOCKS; l++) |
428870ff BB |
1029 | ztest_rll_destroy(&zd->zd_range_lock[l]); |
1030 | } | |
b128c09f | 1031 | |
428870ff | 1032 | #define TXG_MIGHTWAIT (ztest_random(10) == 0 ? TXG_NOWAIT : TXG_WAIT) |
b128c09f | 1033 | |
428870ff BB |
1034 | static uint64_t |
1035 | ztest_tx_assign(dmu_tx_t *tx, uint64_t txg_how, const char *tag) | |
1036 | { | |
1037 | uint64_t txg; | |
1038 | int error; | |
1039 | ||
1040 | /* | |
1041 | * Attempt to assign tx to some transaction group. | |
1042 | */ | |
1043 | error = dmu_tx_assign(tx, txg_how); | |
1044 | if (error) { | |
1045 | if (error == ERESTART) { | |
1046 | ASSERT(txg_how == TXG_NOWAIT); | |
1047 | dmu_tx_wait(tx); | |
1048 | } else { | |
1049 | ASSERT3U(error, ==, ENOSPC); | |
1050 | ztest_record_enospc(tag); | |
1051 | } | |
1052 | dmu_tx_abort(tx); | |
1053 | return (0); | |
1054 | } | |
1055 | txg = dmu_tx_get_txg(tx); | |
1056 | ASSERT(txg != 0); | |
1057 | return (txg); | |
1058 | } | |
1059 | ||
1060 | static void | |
1061 | ztest_pattern_set(void *buf, uint64_t size, uint64_t value) | |
1062 | { | |
1063 | uint64_t *ip = buf; | |
1064 | uint64_t *ip_end = (uint64_t *)((uintptr_t)buf + (uintptr_t)size); | |
1065 | ||
1066 | while (ip < ip_end) | |
1067 | *ip++ = value; | |
1068 | } | |
1069 | ||
1fde1e37 | 1070 | #ifndef NDEBUG |
428870ff BB |
1071 | static boolean_t |
1072 | ztest_pattern_match(void *buf, uint64_t size, uint64_t value) | |
1073 | { | |
1074 | uint64_t *ip = buf; | |
1075 | uint64_t *ip_end = (uint64_t *)((uintptr_t)buf + (uintptr_t)size); | |
1076 | uint64_t diff = 0; | |
1077 | ||
1078 | while (ip < ip_end) | |
1079 | diff |= (value - *ip++); | |
1080 | ||
1081 | return (diff == 0); | |
1082 | } | |
1fde1e37 | 1083 | #endif |
428870ff BB |
1084 | |
1085 | static void | |
1086 | ztest_bt_generate(ztest_block_tag_t *bt, objset_t *os, uint64_t object, | |
1087 | uint64_t offset, uint64_t gen, uint64_t txg, uint64_t crtxg) | |
1088 | { | |
1089 | bt->bt_magic = BT_MAGIC; | |
1090 | bt->bt_objset = dmu_objset_id(os); | |
1091 | bt->bt_object = object; | |
1092 | bt->bt_offset = offset; | |
1093 | bt->bt_gen = gen; | |
1094 | bt->bt_txg = txg; | |
1095 | bt->bt_crtxg = crtxg; | |
1096 | } | |
1097 | ||
1098 | static void | |
1099 | ztest_bt_verify(ztest_block_tag_t *bt, objset_t *os, uint64_t object, | |
1100 | uint64_t offset, uint64_t gen, uint64_t txg, uint64_t crtxg) | |
1101 | { | |
1102 | ASSERT(bt->bt_magic == BT_MAGIC); | |
1103 | ASSERT(bt->bt_objset == dmu_objset_id(os)); | |
1104 | ASSERT(bt->bt_object == object); | |
1105 | ASSERT(bt->bt_offset == offset); | |
1106 | ASSERT(bt->bt_gen <= gen); | |
1107 | ASSERT(bt->bt_txg <= txg); | |
1108 | ASSERT(bt->bt_crtxg == crtxg); | |
1109 | } | |
1110 | ||
1111 | static ztest_block_tag_t * | |
1112 | ztest_bt_bonus(dmu_buf_t *db) | |
1113 | { | |
1114 | dmu_object_info_t doi; | |
1115 | ztest_block_tag_t *bt; | |
1116 | ||
1117 | dmu_object_info_from_db(db, &doi); | |
1118 | ASSERT3U(doi.doi_bonus_size, <=, db->db_size); | |
1119 | ASSERT3U(doi.doi_bonus_size, >=, sizeof (*bt)); | |
1120 | bt = (void *)((char *)db->db_data + doi.doi_bonus_size - sizeof (*bt)); | |
1121 | ||
1122 | return (bt); | |
1123 | } | |
1124 | ||
1125 | /* | |
1126 | * ZIL logging ops | |
1127 | */ | |
1128 | ||
1129 | #define lrz_type lr_mode | |
1130 | #define lrz_blocksize lr_uid | |
1131 | #define lrz_ibshift lr_gid | |
1132 | #define lrz_bonustype lr_rdev | |
1133 | #define lrz_bonuslen lr_crtime[1] | |
1134 | ||
572e2857 | 1135 | static void |
428870ff BB |
1136 | ztest_log_create(ztest_ds_t *zd, dmu_tx_t *tx, lr_create_t *lr) |
1137 | { | |
1138 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1139 | size_t namesize = strlen(name) + 1; | |
1140 | itx_t *itx; | |
1141 | ||
1142 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1143 | return; |
428870ff BB |
1144 | |
1145 | itx = zil_itx_create(TX_CREATE, sizeof (*lr) + namesize); | |
1146 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1147 | sizeof (*lr) + namesize - sizeof (lr_t)); | |
1148 | ||
572e2857 | 1149 | zil_itx_assign(zd->zd_zilog, itx, tx); |
428870ff BB |
1150 | } |
1151 | ||
572e2857 BB |
1152 | static void |
1153 | ztest_log_remove(ztest_ds_t *zd, dmu_tx_t *tx, lr_remove_t *lr, uint64_t object) | |
428870ff BB |
1154 | { |
1155 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1156 | size_t namesize = strlen(name) + 1; | |
1157 | itx_t *itx; | |
1158 | ||
1159 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1160 | return; |
428870ff BB |
1161 | |
1162 | itx = zil_itx_create(TX_REMOVE, sizeof (*lr) + namesize); | |
1163 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1164 | sizeof (*lr) + namesize - sizeof (lr_t)); | |
1165 | ||
572e2857 BB |
1166 | itx->itx_oid = object; |
1167 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1168 | } |
1169 | ||
572e2857 | 1170 | static void |
428870ff BB |
1171 | ztest_log_write(ztest_ds_t *zd, dmu_tx_t *tx, lr_write_t *lr) |
1172 | { | |
1173 | itx_t *itx; | |
1174 | itx_wr_state_t write_state = ztest_random(WR_NUM_STATES); | |
1175 | ||
1176 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1177 | return; |
428870ff BB |
1178 | |
1179 | if (lr->lr_length > ZIL_MAX_LOG_DATA) | |
1180 | write_state = WR_INDIRECT; | |
1181 | ||
1182 | itx = zil_itx_create(TX_WRITE, | |
1183 | sizeof (*lr) + (write_state == WR_COPIED ? lr->lr_length : 0)); | |
1184 | ||
1185 | if (write_state == WR_COPIED && | |
1186 | dmu_read(zd->zd_os, lr->lr_foid, lr->lr_offset, lr->lr_length, | |
1187 | ((lr_write_t *)&itx->itx_lr) + 1, DMU_READ_NO_PREFETCH) != 0) { | |
1188 | zil_itx_destroy(itx); | |
1189 | itx = zil_itx_create(TX_WRITE, sizeof (*lr)); | |
1190 | write_state = WR_NEED_COPY; | |
1191 | } | |
1192 | itx->itx_private = zd; | |
1193 | itx->itx_wr_state = write_state; | |
1194 | itx->itx_sync = (ztest_random(8) == 0); | |
1195 | itx->itx_sod += (write_state == WR_NEED_COPY ? lr->lr_length : 0); | |
1196 | ||
1197 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1198 | sizeof (*lr) - sizeof (lr_t)); | |
1199 | ||
572e2857 | 1200 | zil_itx_assign(zd->zd_zilog, itx, tx); |
428870ff BB |
1201 | } |
1202 | ||
572e2857 | 1203 | static void |
428870ff BB |
1204 | ztest_log_truncate(ztest_ds_t *zd, dmu_tx_t *tx, lr_truncate_t *lr) |
1205 | { | |
1206 | itx_t *itx; | |
1207 | ||
1208 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1209 | return; |
428870ff BB |
1210 | |
1211 | itx = zil_itx_create(TX_TRUNCATE, sizeof (*lr)); | |
1212 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1213 | sizeof (*lr) - sizeof (lr_t)); | |
1214 | ||
572e2857 BB |
1215 | itx->itx_sync = B_FALSE; |
1216 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1217 | } |
1218 | ||
572e2857 | 1219 | static void |
428870ff BB |
1220 | ztest_log_setattr(ztest_ds_t *zd, dmu_tx_t *tx, lr_setattr_t *lr) |
1221 | { | |
1222 | itx_t *itx; | |
1223 | ||
1224 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1225 | return; |
428870ff BB |
1226 | |
1227 | itx = zil_itx_create(TX_SETATTR, sizeof (*lr)); | |
1228 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1229 | sizeof (*lr) - sizeof (lr_t)); | |
1230 | ||
572e2857 BB |
1231 | itx->itx_sync = B_FALSE; |
1232 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1233 | } |
1234 | ||
1235 | /* | |
1236 | * ZIL replay ops | |
1237 | */ | |
1238 | static int | |
1239 | ztest_replay_create(ztest_ds_t *zd, lr_create_t *lr, boolean_t byteswap) | |
1240 | { | |
1241 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1242 | objset_t *os = zd->zd_os; | |
1243 | ztest_block_tag_t *bbt; | |
1244 | dmu_buf_t *db; | |
1245 | dmu_tx_t *tx; | |
1246 | uint64_t txg; | |
1247 | int error = 0; | |
1248 | ||
1249 | if (byteswap) | |
1250 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1251 | ||
1252 | ASSERT(lr->lr_doid == ZTEST_DIROBJ); | |
1253 | ASSERT(name[0] != '\0'); | |
1254 | ||
1255 | tx = dmu_tx_create(os); | |
1256 | ||
1257 | dmu_tx_hold_zap(tx, lr->lr_doid, B_TRUE, name); | |
1258 | ||
1259 | if (lr->lrz_type == DMU_OT_ZAP_OTHER) { | |
1260 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, B_TRUE, NULL); | |
1261 | } else { | |
1262 | dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT); | |
1263 | } | |
1264 | ||
1265 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1266 | if (txg == 0) | |
1267 | return (ENOSPC); | |
1268 | ||
1269 | ASSERT(dmu_objset_zil(os)->zl_replay == !!lr->lr_foid); | |
1270 | ||
1271 | if (lr->lrz_type == DMU_OT_ZAP_OTHER) { | |
1272 | if (lr->lr_foid == 0) { | |
1273 | lr->lr_foid = zap_create(os, | |
1274 | lr->lrz_type, lr->lrz_bonustype, | |
1275 | lr->lrz_bonuslen, tx); | |
1276 | } else { | |
1277 | error = zap_create_claim(os, lr->lr_foid, | |
1278 | lr->lrz_type, lr->lrz_bonustype, | |
1279 | lr->lrz_bonuslen, tx); | |
1280 | } | |
1281 | } else { | |
1282 | if (lr->lr_foid == 0) { | |
1283 | lr->lr_foid = dmu_object_alloc(os, | |
1284 | lr->lrz_type, 0, lr->lrz_bonustype, | |
1285 | lr->lrz_bonuslen, tx); | |
1286 | } else { | |
1287 | error = dmu_object_claim(os, lr->lr_foid, | |
1288 | lr->lrz_type, 0, lr->lrz_bonustype, | |
1289 | lr->lrz_bonuslen, tx); | |
1290 | } | |
1291 | } | |
1292 | ||
1293 | if (error) { | |
1294 | ASSERT3U(error, ==, EEXIST); | |
1295 | ASSERT(zd->zd_zilog->zl_replay); | |
1296 | dmu_tx_commit(tx); | |
1297 | return (error); | |
1298 | } | |
1299 | ||
1300 | ASSERT(lr->lr_foid != 0); | |
1301 | ||
1302 | if (lr->lrz_type != DMU_OT_ZAP_OTHER) | |
1303 | VERIFY3U(0, ==, dmu_object_set_blocksize(os, lr->lr_foid, | |
1304 | lr->lrz_blocksize, lr->lrz_ibshift, tx)); | |
1305 | ||
1306 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
1307 | bbt = ztest_bt_bonus(db); | |
1308 | dmu_buf_will_dirty(db, tx); | |
1309 | ztest_bt_generate(bbt, os, lr->lr_foid, -1ULL, lr->lr_gen, txg, txg); | |
1310 | dmu_buf_rele(db, FTAG); | |
1311 | ||
1312 | VERIFY3U(0, ==, zap_add(os, lr->lr_doid, name, sizeof (uint64_t), 1, | |
1313 | &lr->lr_foid, tx)); | |
1314 | ||
1315 | (void) ztest_log_create(zd, tx, lr); | |
1316 | ||
1317 | dmu_tx_commit(tx); | |
1318 | ||
1319 | return (0); | |
1320 | } | |
1321 | ||
1322 | static int | |
1323 | ztest_replay_remove(ztest_ds_t *zd, lr_remove_t *lr, boolean_t byteswap) | |
1324 | { | |
1325 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1326 | objset_t *os = zd->zd_os; | |
1327 | dmu_object_info_t doi; | |
1328 | dmu_tx_t *tx; | |
1329 | uint64_t object, txg; | |
1330 | ||
1331 | if (byteswap) | |
1332 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1333 | ||
1334 | ASSERT(lr->lr_doid == ZTEST_DIROBJ); | |
1335 | ASSERT(name[0] != '\0'); | |
1336 | ||
1337 | VERIFY3U(0, ==, | |
1338 | zap_lookup(os, lr->lr_doid, name, sizeof (object), 1, &object)); | |
1339 | ASSERT(object != 0); | |
1340 | ||
1341 | ztest_object_lock(zd, object, RL_WRITER); | |
1342 | ||
1343 | VERIFY3U(0, ==, dmu_object_info(os, object, &doi)); | |
1344 | ||
1345 | tx = dmu_tx_create(os); | |
1346 | ||
1347 | dmu_tx_hold_zap(tx, lr->lr_doid, B_FALSE, name); | |
1348 | dmu_tx_hold_free(tx, object, 0, DMU_OBJECT_END); | |
1349 | ||
1350 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1351 | if (txg == 0) { | |
1352 | ztest_object_unlock(zd, object); | |
1353 | return (ENOSPC); | |
1354 | } | |
1355 | ||
1356 | if (doi.doi_type == DMU_OT_ZAP_OTHER) { | |
1357 | VERIFY3U(0, ==, zap_destroy(os, object, tx)); | |
1358 | } else { | |
1359 | VERIFY3U(0, ==, dmu_object_free(os, object, tx)); | |
1360 | } | |
1361 | ||
1362 | VERIFY3U(0, ==, zap_remove(os, lr->lr_doid, name, tx)); | |
1363 | ||
572e2857 | 1364 | (void) ztest_log_remove(zd, tx, lr, object); |
428870ff BB |
1365 | |
1366 | dmu_tx_commit(tx); | |
1367 | ||
1368 | ztest_object_unlock(zd, object); | |
1369 | ||
1370 | return (0); | |
1371 | } | |
1372 | ||
1373 | static int | |
1374 | ztest_replay_write(ztest_ds_t *zd, lr_write_t *lr, boolean_t byteswap) | |
1375 | { | |
1376 | objset_t *os = zd->zd_os; | |
1377 | void *data = lr + 1; /* data follows lr */ | |
1378 | uint64_t offset, length; | |
1379 | ztest_block_tag_t *bt = data; | |
1380 | ztest_block_tag_t *bbt; | |
1381 | uint64_t gen, txg, lrtxg, crtxg; | |
1382 | dmu_object_info_t doi; | |
1383 | dmu_tx_t *tx; | |
1384 | dmu_buf_t *db; | |
1385 | arc_buf_t *abuf = NULL; | |
1386 | rl_t *rl; | |
1387 | ||
1388 | if (byteswap) | |
1389 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1390 | ||
1391 | offset = lr->lr_offset; | |
1392 | length = lr->lr_length; | |
1393 | ||
1394 | /* If it's a dmu_sync() block, write the whole block */ | |
1395 | if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) { | |
1396 | uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr); | |
1397 | if (length < blocksize) { | |
1398 | offset -= offset % blocksize; | |
1399 | length = blocksize; | |
1400 | } | |
1401 | } | |
1402 | ||
1403 | if (bt->bt_magic == BSWAP_64(BT_MAGIC)) | |
1404 | byteswap_uint64_array(bt, sizeof (*bt)); | |
1405 | ||
1406 | if (bt->bt_magic != BT_MAGIC) | |
1407 | bt = NULL; | |
1408 | ||
1409 | ztest_object_lock(zd, lr->lr_foid, RL_READER); | |
1410 | rl = ztest_range_lock(zd, lr->lr_foid, offset, length, RL_WRITER); | |
1411 | ||
1412 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
1413 | ||
1414 | dmu_object_info_from_db(db, &doi); | |
1415 | ||
1416 | bbt = ztest_bt_bonus(db); | |
1417 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
1418 | gen = bbt->bt_gen; | |
1419 | crtxg = bbt->bt_crtxg; | |
1420 | lrtxg = lr->lr_common.lrc_txg; | |
1421 | ||
1422 | tx = dmu_tx_create(os); | |
1423 | ||
1424 | dmu_tx_hold_write(tx, lr->lr_foid, offset, length); | |
1425 | ||
1426 | if (ztest_random(8) == 0 && length == doi.doi_data_block_size && | |
1427 | P2PHASE(offset, length) == 0) | |
1428 | abuf = dmu_request_arcbuf(db, length); | |
1429 | ||
1430 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1431 | if (txg == 0) { | |
1432 | if (abuf != NULL) | |
1433 | dmu_return_arcbuf(abuf); | |
1434 | dmu_buf_rele(db, FTAG); | |
1435 | ztest_range_unlock(rl); | |
1436 | ztest_object_unlock(zd, lr->lr_foid); | |
1437 | return (ENOSPC); | |
1438 | } | |
1439 | ||
1440 | if (bt != NULL) { | |
1441 | /* | |
1442 | * Usually, verify the old data before writing new data -- | |
1443 | * but not always, because we also want to verify correct | |
1444 | * behavior when the data was not recently read into cache. | |
1445 | */ | |
1446 | ASSERT(offset % doi.doi_data_block_size == 0); | |
1447 | if (ztest_random(4) != 0) { | |
1448 | int prefetch = ztest_random(2) ? | |
1449 | DMU_READ_PREFETCH : DMU_READ_NO_PREFETCH; | |
1450 | ztest_block_tag_t rbt; | |
1451 | ||
1452 | VERIFY(dmu_read(os, lr->lr_foid, offset, | |
1453 | sizeof (rbt), &rbt, prefetch) == 0); | |
1454 | if (rbt.bt_magic == BT_MAGIC) { | |
1455 | ztest_bt_verify(&rbt, os, lr->lr_foid, | |
1456 | offset, gen, txg, crtxg); | |
1457 | } | |
1458 | } | |
1459 | ||
1460 | /* | |
1461 | * Writes can appear to be newer than the bonus buffer because | |
1462 | * the ztest_get_data() callback does a dmu_read() of the | |
1463 | * open-context data, which may be different than the data | |
1464 | * as it was when the write was generated. | |
1465 | */ | |
1466 | if (zd->zd_zilog->zl_replay) { | |
1467 | ztest_bt_verify(bt, os, lr->lr_foid, offset, | |
1468 | MAX(gen, bt->bt_gen), MAX(txg, lrtxg), | |
1469 | bt->bt_crtxg); | |
1470 | } | |
1471 | ||
1472 | /* | |
1473 | * Set the bt's gen/txg to the bonus buffer's gen/txg | |
1474 | * so that all of the usual ASSERTs will work. | |
1475 | */ | |
1476 | ztest_bt_generate(bt, os, lr->lr_foid, offset, gen, txg, crtxg); | |
1477 | } | |
1478 | ||
1479 | if (abuf == NULL) { | |
1480 | dmu_write(os, lr->lr_foid, offset, length, data, tx); | |
1481 | } else { | |
1482 | bcopy(data, abuf->b_data, length); | |
1483 | dmu_assign_arcbuf(db, offset, abuf, tx); | |
1484 | } | |
1485 | ||
1486 | (void) ztest_log_write(zd, tx, lr); | |
1487 | ||
1488 | dmu_buf_rele(db, FTAG); | |
1489 | ||
1490 | dmu_tx_commit(tx); | |
1491 | ||
1492 | ztest_range_unlock(rl); | |
1493 | ztest_object_unlock(zd, lr->lr_foid); | |
1494 | ||
1495 | return (0); | |
1496 | } | |
1497 | ||
1498 | static int | |
1499 | ztest_replay_truncate(ztest_ds_t *zd, lr_truncate_t *lr, boolean_t byteswap) | |
1500 | { | |
1501 | objset_t *os = zd->zd_os; | |
1502 | dmu_tx_t *tx; | |
1503 | uint64_t txg; | |
1504 | rl_t *rl; | |
1505 | ||
1506 | if (byteswap) | |
1507 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1508 | ||
1509 | ztest_object_lock(zd, lr->lr_foid, RL_READER); | |
1510 | rl = ztest_range_lock(zd, lr->lr_foid, lr->lr_offset, lr->lr_length, | |
1511 | RL_WRITER); | |
1512 | ||
1513 | tx = dmu_tx_create(os); | |
1514 | ||
1515 | dmu_tx_hold_free(tx, lr->lr_foid, lr->lr_offset, lr->lr_length); | |
1516 | ||
1517 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1518 | if (txg == 0) { | |
1519 | ztest_range_unlock(rl); | |
1520 | ztest_object_unlock(zd, lr->lr_foid); | |
1521 | return (ENOSPC); | |
1522 | } | |
1523 | ||
1524 | VERIFY(dmu_free_range(os, lr->lr_foid, lr->lr_offset, | |
1525 | lr->lr_length, tx) == 0); | |
1526 | ||
1527 | (void) ztest_log_truncate(zd, tx, lr); | |
1528 | ||
1529 | dmu_tx_commit(tx); | |
1530 | ||
1531 | ztest_range_unlock(rl); | |
1532 | ztest_object_unlock(zd, lr->lr_foid); | |
1533 | ||
1534 | return (0); | |
1535 | } | |
1536 | ||
1537 | static int | |
1538 | ztest_replay_setattr(ztest_ds_t *zd, lr_setattr_t *lr, boolean_t byteswap) | |
1539 | { | |
1540 | objset_t *os = zd->zd_os; | |
1541 | dmu_tx_t *tx; | |
1542 | dmu_buf_t *db; | |
1543 | ztest_block_tag_t *bbt; | |
1544 | uint64_t txg, lrtxg, crtxg; | |
1545 | ||
1546 | if (byteswap) | |
1547 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1548 | ||
1549 | ztest_object_lock(zd, lr->lr_foid, RL_WRITER); | |
1550 | ||
1551 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
1552 | ||
1553 | tx = dmu_tx_create(os); | |
1554 | dmu_tx_hold_bonus(tx, lr->lr_foid); | |
1555 | ||
1556 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1557 | if (txg == 0) { | |
1558 | dmu_buf_rele(db, FTAG); | |
1559 | ztest_object_unlock(zd, lr->lr_foid); | |
1560 | return (ENOSPC); | |
1561 | } | |
1562 | ||
1563 | bbt = ztest_bt_bonus(db); | |
1564 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
1565 | crtxg = bbt->bt_crtxg; | |
1566 | lrtxg = lr->lr_common.lrc_txg; | |
1567 | ||
1568 | if (zd->zd_zilog->zl_replay) { | |
1569 | ASSERT(lr->lr_size != 0); | |
1570 | ASSERT(lr->lr_mode != 0); | |
1571 | ASSERT(lrtxg != 0); | |
1572 | } else { | |
1573 | /* | |
1574 | * Randomly change the size and increment the generation. | |
1575 | */ | |
1576 | lr->lr_size = (ztest_random(db->db_size / sizeof (*bbt)) + 1) * | |
1577 | sizeof (*bbt); | |
1578 | lr->lr_mode = bbt->bt_gen + 1; | |
1579 | ASSERT(lrtxg == 0); | |
1580 | } | |
1581 | ||
1582 | /* | |
1583 | * Verify that the current bonus buffer is not newer than our txg. | |
1584 | */ | |
1585 | ztest_bt_verify(bbt, os, lr->lr_foid, -1ULL, lr->lr_mode, | |
1586 | MAX(txg, lrtxg), crtxg); | |
1587 | ||
1588 | dmu_buf_will_dirty(db, tx); | |
1589 | ||
1590 | ASSERT3U(lr->lr_size, >=, sizeof (*bbt)); | |
1591 | ASSERT3U(lr->lr_size, <=, db->db_size); | |
1592 | VERIFY3U(dmu_set_bonus(db, lr->lr_size, tx), ==, 0); | |
1593 | bbt = ztest_bt_bonus(db); | |
1594 | ||
1595 | ztest_bt_generate(bbt, os, lr->lr_foid, -1ULL, lr->lr_mode, txg, crtxg); | |
1596 | ||
1597 | dmu_buf_rele(db, FTAG); | |
1598 | ||
1599 | (void) ztest_log_setattr(zd, tx, lr); | |
1600 | ||
1601 | dmu_tx_commit(tx); | |
1602 | ||
1603 | ztest_object_unlock(zd, lr->lr_foid); | |
1604 | ||
1605 | return (0); | |
1606 | } | |
1607 | ||
1608 | zil_replay_func_t *ztest_replay_vector[TX_MAX_TYPE] = { | |
0bc8fd78 BB |
1609 | NULL, /* 0 no such transaction type */ |
1610 | (zil_replay_func_t *)ztest_replay_create, /* TX_CREATE */ | |
1611 | NULL, /* TX_MKDIR */ | |
1612 | NULL, /* TX_MKXATTR */ | |
1613 | NULL, /* TX_SYMLINK */ | |
1614 | (zil_replay_func_t *)ztest_replay_remove, /* TX_REMOVE */ | |
1615 | NULL, /* TX_RMDIR */ | |
1616 | NULL, /* TX_LINK */ | |
1617 | NULL, /* TX_RENAME */ | |
1618 | (zil_replay_func_t *)ztest_replay_write, /* TX_WRITE */ | |
1619 | (zil_replay_func_t *)ztest_replay_truncate, /* TX_TRUNCATE */ | |
1620 | (zil_replay_func_t *)ztest_replay_setattr, /* TX_SETATTR */ | |
1621 | NULL, /* TX_ACL */ | |
1622 | NULL, /* TX_CREATE_ACL */ | |
1623 | NULL, /* TX_CREATE_ATTR */ | |
1624 | NULL, /* TX_CREATE_ACL_ATTR */ | |
1625 | NULL, /* TX_MKDIR_ACL */ | |
1626 | NULL, /* TX_MKDIR_ATTR */ | |
1627 | NULL, /* TX_MKDIR_ACL_ATTR */ | |
1628 | NULL, /* TX_WRITE2 */ | |
428870ff BB |
1629 | }; |
1630 | ||
1631 | /* | |
1632 | * ZIL get_data callbacks | |
1633 | */ | |
1634 | ||
1635 | static void | |
1636 | ztest_get_done(zgd_t *zgd, int error) | |
1637 | { | |
1638 | ztest_ds_t *zd = zgd->zgd_private; | |
1639 | uint64_t object = zgd->zgd_rl->rl_object; | |
1640 | ||
1641 | if (zgd->zgd_db) | |
1642 | dmu_buf_rele(zgd->zgd_db, zgd); | |
1643 | ||
1644 | ztest_range_unlock(zgd->zgd_rl); | |
1645 | ztest_object_unlock(zd, object); | |
1646 | ||
1647 | if (error == 0 && zgd->zgd_bp) | |
1648 | zil_add_block(zgd->zgd_zilog, zgd->zgd_bp); | |
1649 | ||
1650 | umem_free(zgd, sizeof (*zgd)); | |
1651 | } | |
1652 | ||
1653 | static int | |
1654 | ztest_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio) | |
1655 | { | |
1656 | ztest_ds_t *zd = arg; | |
1657 | objset_t *os = zd->zd_os; | |
1658 | uint64_t object = lr->lr_foid; | |
1659 | uint64_t offset = lr->lr_offset; | |
1660 | uint64_t size = lr->lr_length; | |
1661 | blkptr_t *bp = &lr->lr_blkptr; | |
1662 | uint64_t txg = lr->lr_common.lrc_txg; | |
1663 | uint64_t crtxg; | |
1664 | dmu_object_info_t doi; | |
1665 | dmu_buf_t *db; | |
1666 | zgd_t *zgd; | |
1667 | int error; | |
1668 | ||
1669 | ztest_object_lock(zd, object, RL_READER); | |
1670 | error = dmu_bonus_hold(os, object, FTAG, &db); | |
1671 | if (error) { | |
1672 | ztest_object_unlock(zd, object); | |
1673 | return (error); | |
1674 | } | |
1675 | ||
1676 | crtxg = ztest_bt_bonus(db)->bt_crtxg; | |
1677 | ||
1678 | if (crtxg == 0 || crtxg > txg) { | |
1679 | dmu_buf_rele(db, FTAG); | |
1680 | ztest_object_unlock(zd, object); | |
1681 | return (ENOENT); | |
1682 | } | |
1683 | ||
1684 | dmu_object_info_from_db(db, &doi); | |
1685 | dmu_buf_rele(db, FTAG); | |
1686 | db = NULL; | |
1687 | ||
1688 | zgd = umem_zalloc(sizeof (*zgd), UMEM_NOFAIL); | |
1689 | zgd->zgd_zilog = zd->zd_zilog; | |
1690 | zgd->zgd_private = zd; | |
1691 | ||
1692 | if (buf != NULL) { /* immediate write */ | |
1693 | zgd->zgd_rl = ztest_range_lock(zd, object, offset, size, | |
1694 | RL_READER); | |
1695 | ||
1696 | error = dmu_read(os, object, offset, size, buf, | |
1697 | DMU_READ_NO_PREFETCH); | |
1698 | ASSERT(error == 0); | |
1699 | } else { | |
1700 | size = doi.doi_data_block_size; | |
1701 | if (ISP2(size)) { | |
1702 | offset = P2ALIGN(offset, size); | |
1703 | } else { | |
1704 | ASSERT(offset < size); | |
1705 | offset = 0; | |
1706 | } | |
1707 | ||
1708 | zgd->zgd_rl = ztest_range_lock(zd, object, offset, size, | |
1709 | RL_READER); | |
1710 | ||
1711 | error = dmu_buf_hold(os, object, offset, zgd, &db, | |
1712 | DMU_READ_NO_PREFETCH); | |
1713 | ||
1714 | if (error == 0) { | |
1715 | zgd->zgd_db = db; | |
1716 | zgd->zgd_bp = bp; | |
1717 | ||
1718 | ASSERT(db->db_offset == offset); | |
1719 | ASSERT(db->db_size == size); | |
1720 | ||
1721 | error = dmu_sync(zio, lr->lr_common.lrc_txg, | |
1722 | ztest_get_done, zgd); | |
1723 | ||
1724 | if (error == 0) | |
1725 | return (0); | |
1726 | } | |
1727 | } | |
1728 | ||
1729 | ztest_get_done(zgd, error); | |
1730 | ||
1731 | return (error); | |
1732 | } | |
1733 | ||
1734 | static void * | |
1735 | ztest_lr_alloc(size_t lrsize, char *name) | |
1736 | { | |
1737 | char *lr; | |
1738 | size_t namesize = name ? strlen(name) + 1 : 0; | |
1739 | ||
1740 | lr = umem_zalloc(lrsize + namesize, UMEM_NOFAIL); | |
1741 | ||
1742 | if (name) | |
1743 | bcopy(name, lr + lrsize, namesize); | |
1744 | ||
1745 | return (lr); | |
1746 | } | |
1747 | ||
1748 | void | |
1749 | ztest_lr_free(void *lr, size_t lrsize, char *name) | |
1750 | { | |
1751 | size_t namesize = name ? strlen(name) + 1 : 0; | |
1752 | ||
1753 | umem_free(lr, lrsize + namesize); | |
1754 | } | |
1755 | ||
1756 | /* | |
1757 | * Lookup a bunch of objects. Returns the number of objects not found. | |
1758 | */ | |
1759 | static int | |
1760 | ztest_lookup(ztest_ds_t *zd, ztest_od_t *od, int count) | |
1761 | { | |
1762 | int missing = 0; | |
1763 | int error; | |
d6320ddb | 1764 | int i; |
428870ff | 1765 | |
1e33ac1e | 1766 | ASSERT(mutex_held(&zd->zd_dirobj_lock)); |
428870ff | 1767 | |
d6320ddb | 1768 | for (i = 0; i < count; i++, od++) { |
428870ff BB |
1769 | od->od_object = 0; |
1770 | error = zap_lookup(zd->zd_os, od->od_dir, od->od_name, | |
1771 | sizeof (uint64_t), 1, &od->od_object); | |
1772 | if (error) { | |
1773 | ASSERT(error == ENOENT); | |
1774 | ASSERT(od->od_object == 0); | |
1775 | missing++; | |
1776 | } else { | |
1777 | dmu_buf_t *db; | |
1778 | ztest_block_tag_t *bbt; | |
1779 | dmu_object_info_t doi; | |
1780 | ||
1781 | ASSERT(od->od_object != 0); | |
1782 | ASSERT(missing == 0); /* there should be no gaps */ | |
1783 | ||
1784 | ztest_object_lock(zd, od->od_object, RL_READER); | |
1785 | VERIFY3U(0, ==, dmu_bonus_hold(zd->zd_os, | |
1786 | od->od_object, FTAG, &db)); | |
1787 | dmu_object_info_from_db(db, &doi); | |
1788 | bbt = ztest_bt_bonus(db); | |
1789 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
1790 | od->od_type = doi.doi_type; | |
1791 | od->od_blocksize = doi.doi_data_block_size; | |
1792 | od->od_gen = bbt->bt_gen; | |
1793 | dmu_buf_rele(db, FTAG); | |
1794 | ztest_object_unlock(zd, od->od_object); | |
1795 | } | |
1796 | } | |
1797 | ||
1798 | return (missing); | |
1799 | } | |
1800 | ||
1801 | static int | |
1802 | ztest_create(ztest_ds_t *zd, ztest_od_t *od, int count) | |
1803 | { | |
1804 | int missing = 0; | |
d6320ddb | 1805 | int i; |
428870ff | 1806 | |
1e33ac1e | 1807 | ASSERT(mutex_held(&zd->zd_dirobj_lock)); |
428870ff | 1808 | |
d6320ddb | 1809 | for (i = 0; i < count; i++, od++) { |
428870ff BB |
1810 | if (missing) { |
1811 | od->od_object = 0; | |
1812 | missing++; | |
1813 | continue; | |
1814 | } | |
1815 | ||
1816 | lr_create_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name); | |
1817 | ||
1818 | lr->lr_doid = od->od_dir; | |
1819 | lr->lr_foid = 0; /* 0 to allocate, > 0 to claim */ | |
1820 | lr->lrz_type = od->od_crtype; | |
1821 | lr->lrz_blocksize = od->od_crblocksize; | |
1822 | lr->lrz_ibshift = ztest_random_ibshift(); | |
1823 | lr->lrz_bonustype = DMU_OT_UINT64_OTHER; | |
1824 | lr->lrz_bonuslen = dmu_bonus_max(); | |
1825 | lr->lr_gen = od->od_crgen; | |
1826 | lr->lr_crtime[0] = time(NULL); | |
1827 | ||
1828 | if (ztest_replay_create(zd, lr, B_FALSE) != 0) { | |
1829 | ASSERT(missing == 0); | |
1830 | od->od_object = 0; | |
1831 | missing++; | |
1832 | } else { | |
1833 | od->od_object = lr->lr_foid; | |
1834 | od->od_type = od->od_crtype; | |
1835 | od->od_blocksize = od->od_crblocksize; | |
1836 | od->od_gen = od->od_crgen; | |
1837 | ASSERT(od->od_object != 0); | |
1838 | } | |
1839 | ||
1840 | ztest_lr_free(lr, sizeof (*lr), od->od_name); | |
1841 | } | |
1842 | ||
1843 | return (missing); | |
1844 | } | |
1845 | ||
1846 | static int | |
1847 | ztest_remove(ztest_ds_t *zd, ztest_od_t *od, int count) | |
1848 | { | |
1849 | int missing = 0; | |
1850 | int error; | |
d6320ddb | 1851 | int i; |
428870ff | 1852 | |
1e33ac1e | 1853 | ASSERT(mutex_held(&zd->zd_dirobj_lock)); |
428870ff BB |
1854 | |
1855 | od += count - 1; | |
1856 | ||
d6320ddb | 1857 | for (i = count - 1; i >= 0; i--, od--) { |
428870ff BB |
1858 | if (missing) { |
1859 | missing++; | |
1860 | continue; | |
1861 | } | |
1862 | ||
1863 | if (od->od_object == 0) | |
1864 | continue; | |
1865 | ||
1866 | lr_remove_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name); | |
1867 | ||
1868 | lr->lr_doid = od->od_dir; | |
1869 | ||
1870 | if ((error = ztest_replay_remove(zd, lr, B_FALSE)) != 0) { | |
1871 | ASSERT3U(error, ==, ENOSPC); | |
1872 | missing++; | |
1873 | } else { | |
1874 | od->od_object = 0; | |
1875 | } | |
1876 | ztest_lr_free(lr, sizeof (*lr), od->od_name); | |
1877 | } | |
1878 | ||
1879 | return (missing); | |
1880 | } | |
1881 | ||
1882 | static int | |
1883 | ztest_write(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size, | |
1884 | void *data) | |
1885 | { | |
1886 | lr_write_t *lr; | |
1887 | int error; | |
1888 | ||
1889 | lr = ztest_lr_alloc(sizeof (*lr) + size, NULL); | |
1890 | ||
1891 | lr->lr_foid = object; | |
1892 | lr->lr_offset = offset; | |
1893 | lr->lr_length = size; | |
1894 | lr->lr_blkoff = 0; | |
1895 | BP_ZERO(&lr->lr_blkptr); | |
1896 | ||
1897 | bcopy(data, lr + 1, size); | |
1898 | ||
1899 | error = ztest_replay_write(zd, lr, B_FALSE); | |
1900 | ||
1901 | ztest_lr_free(lr, sizeof (*lr) + size, NULL); | |
1902 | ||
1903 | return (error); | |
1904 | } | |
1905 | ||
1906 | static int | |
1907 | ztest_truncate(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size) | |
1908 | { | |
1909 | lr_truncate_t *lr; | |
1910 | int error; | |
1911 | ||
1912 | lr = ztest_lr_alloc(sizeof (*lr), NULL); | |
1913 | ||
1914 | lr->lr_foid = object; | |
1915 | lr->lr_offset = offset; | |
1916 | lr->lr_length = size; | |
1917 | ||
1918 | error = ztest_replay_truncate(zd, lr, B_FALSE); | |
1919 | ||
1920 | ztest_lr_free(lr, sizeof (*lr), NULL); | |
1921 | ||
1922 | return (error); | |
1923 | } | |
1924 | ||
1925 | static int | |
1926 | ztest_setattr(ztest_ds_t *zd, uint64_t object) | |
1927 | { | |
1928 | lr_setattr_t *lr; | |
1929 | int error; | |
1930 | ||
1931 | lr = ztest_lr_alloc(sizeof (*lr), NULL); | |
1932 | ||
1933 | lr->lr_foid = object; | |
1934 | lr->lr_size = 0; | |
1935 | lr->lr_mode = 0; | |
1936 | ||
1937 | error = ztest_replay_setattr(zd, lr, B_FALSE); | |
1938 | ||
1939 | ztest_lr_free(lr, sizeof (*lr), NULL); | |
1940 | ||
1941 | return (error); | |
1942 | } | |
1943 | ||
1944 | static void | |
1945 | ztest_prealloc(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size) | |
1946 | { | |
1947 | objset_t *os = zd->zd_os; | |
1948 | dmu_tx_t *tx; | |
1949 | uint64_t txg; | |
1950 | rl_t *rl; | |
1951 | ||
1952 | txg_wait_synced(dmu_objset_pool(os), 0); | |
1953 | ||
1954 | ztest_object_lock(zd, object, RL_READER); | |
1955 | rl = ztest_range_lock(zd, object, offset, size, RL_WRITER); | |
1956 | ||
1957 | tx = dmu_tx_create(os); | |
1958 | ||
1959 | dmu_tx_hold_write(tx, object, offset, size); | |
1960 | ||
1961 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1962 | ||
1963 | if (txg != 0) { | |
1964 | dmu_prealloc(os, object, offset, size, tx); | |
1965 | dmu_tx_commit(tx); | |
1966 | txg_wait_synced(dmu_objset_pool(os), txg); | |
1967 | } else { | |
1968 | (void) dmu_free_long_range(os, object, offset, size); | |
1969 | } | |
1970 | ||
1971 | ztest_range_unlock(rl); | |
1972 | ztest_object_unlock(zd, object); | |
1973 | } | |
1974 | ||
1975 | static void | |
1976 | ztest_io(ztest_ds_t *zd, uint64_t object, uint64_t offset) | |
1977 | { | |
1978 | ztest_block_tag_t wbt; | |
1979 | dmu_object_info_t doi; | |
1980 | enum ztest_io_type io_type; | |
1981 | uint64_t blocksize; | |
1982 | void *data; | |
1983 | ||
1984 | VERIFY(dmu_object_info(zd->zd_os, object, &doi) == 0); | |
1985 | blocksize = doi.doi_data_block_size; | |
1986 | data = umem_alloc(blocksize, UMEM_NOFAIL); | |
1987 | ||
1988 | /* | |
1989 | * Pick an i/o type at random, biased toward writing block tags. | |
1990 | */ | |
1991 | io_type = ztest_random(ZTEST_IO_TYPES); | |
1992 | if (ztest_random(2) == 0) | |
1993 | io_type = ZTEST_IO_WRITE_TAG; | |
1994 | ||
1995 | switch (io_type) { | |
1996 | ||
1997 | case ZTEST_IO_WRITE_TAG: | |
1998 | ztest_bt_generate(&wbt, zd->zd_os, object, offset, 0, 0, 0); | |
1999 | (void) ztest_write(zd, object, offset, sizeof (wbt), &wbt); | |
2000 | break; | |
2001 | ||
2002 | case ZTEST_IO_WRITE_PATTERN: | |
2003 | (void) memset(data, 'a' + (object + offset) % 5, blocksize); | |
2004 | if (ztest_random(2) == 0) { | |
2005 | /* | |
2006 | * Induce fletcher2 collisions to ensure that | |
2007 | * zio_ddt_collision() detects and resolves them | |
2008 | * when using fletcher2-verify for deduplication. | |
2009 | */ | |
2010 | ((uint64_t *)data)[0] ^= 1ULL << 63; | |
2011 | ((uint64_t *)data)[4] ^= 1ULL << 63; | |
2012 | } | |
2013 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2014 | break; | |
2015 | ||
2016 | case ZTEST_IO_WRITE_ZEROES: | |
2017 | bzero(data, blocksize); | |
2018 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2019 | break; | |
2020 | ||
2021 | case ZTEST_IO_TRUNCATE: | |
2022 | (void) ztest_truncate(zd, object, offset, blocksize); | |
2023 | break; | |
2024 | ||
2025 | case ZTEST_IO_SETATTR: | |
2026 | (void) ztest_setattr(zd, object); | |
2027 | break; | |
e75c13c3 BB |
2028 | default: |
2029 | break; | |
428870ff BB |
2030 | } |
2031 | ||
2032 | umem_free(data, blocksize); | |
2033 | } | |
2034 | ||
2035 | /* | |
2036 | * Initialize an object description template. | |
2037 | */ | |
2038 | static void | |
2039 | ztest_od_init(ztest_od_t *od, uint64_t id, char *tag, uint64_t index, | |
2040 | dmu_object_type_t type, uint64_t blocksize, uint64_t gen) | |
2041 | { | |
2042 | od->od_dir = ZTEST_DIROBJ; | |
2043 | od->od_object = 0; | |
2044 | ||
2045 | od->od_crtype = type; | |
2046 | od->od_crblocksize = blocksize ? blocksize : ztest_random_blocksize(); | |
2047 | od->od_crgen = gen; | |
2048 | ||
2049 | od->od_type = DMU_OT_NONE; | |
2050 | od->od_blocksize = 0; | |
2051 | od->od_gen = 0; | |
2052 | ||
2053 | (void) snprintf(od->od_name, sizeof (od->od_name), "%s(%lld)[%llu]", | |
b8864a23 | 2054 | tag, (longlong_t)id, (u_longlong_t)index); |
428870ff BB |
2055 | } |
2056 | ||
2057 | /* | |
2058 | * Lookup or create the objects for a test using the od template. | |
2059 | * If the objects do not all exist, or if 'remove' is specified, | |
2060 | * remove any existing objects and create new ones. Otherwise, | |
2061 | * use the existing objects. | |
2062 | */ | |
2063 | static int | |
2064 | ztest_object_init(ztest_ds_t *zd, ztest_od_t *od, size_t size, boolean_t remove) | |
2065 | { | |
2066 | int count = size / sizeof (*od); | |
2067 | int rv = 0; | |
2068 | ||
1e33ac1e | 2069 | mutex_enter(&zd->zd_dirobj_lock); |
428870ff BB |
2070 | if ((ztest_lookup(zd, od, count) != 0 || remove) && |
2071 | (ztest_remove(zd, od, count) != 0 || | |
2072 | ztest_create(zd, od, count) != 0)) | |
2073 | rv = -1; | |
2074 | zd->zd_od = od; | |
1e33ac1e | 2075 | mutex_exit(&zd->zd_dirobj_lock); |
428870ff BB |
2076 | |
2077 | return (rv); | |
2078 | } | |
2079 | ||
2080 | /* ARGSUSED */ | |
2081 | void | |
2082 | ztest_zil_commit(ztest_ds_t *zd, uint64_t id) | |
2083 | { | |
2084 | zilog_t *zilog = zd->zd_zilog; | |
2085 | ||
572e2857 | 2086 | zil_commit(zilog, ztest_random(ZTEST_OBJECTS)); |
428870ff BB |
2087 | |
2088 | /* | |
2089 | * Remember the committed values in zd, which is in parent/child | |
2090 | * shared memory. If we die, the next iteration of ztest_run() | |
2091 | * will verify that the log really does contain this record. | |
2092 | */ | |
2093 | mutex_enter(&zilog->zl_lock); | |
2094 | ASSERT(zd->zd_seq <= zilog->zl_commit_lr_seq); | |
2095 | zd->zd_seq = zilog->zl_commit_lr_seq; | |
2096 | mutex_exit(&zilog->zl_lock); | |
2097 | } | |
2098 | ||
2099 | /* | |
2100 | * Verify that we can't destroy an active pool, create an existing pool, | |
2101 | * or create a pool with a bad vdev spec. | |
2102 | */ | |
2103 | /* ARGSUSED */ | |
2104 | void | |
2105 | ztest_spa_create_destroy(ztest_ds_t *zd, uint64_t id) | |
2106 | { | |
2107 | ztest_shared_t *zs = ztest_shared; | |
2108 | spa_t *spa; | |
2109 | nvlist_t *nvroot; | |
2110 | ||
2111 | /* | |
2112 | * Attempt to create using a bad file. | |
2113 | */ | |
2114 | nvroot = make_vdev_root("/dev/bogus", NULL, 0, 0, 0, 0, 0, 1); | |
2115 | VERIFY3U(ENOENT, ==, | |
2116 | spa_create("ztest_bad_file", nvroot, NULL, NULL, NULL)); | |
2117 | nvlist_free(nvroot); | |
2118 | ||
2119 | /* | |
2120 | * Attempt to create using a bad mirror. | |
2121 | */ | |
2122 | nvroot = make_vdev_root("/dev/bogus", NULL, 0, 0, 0, 0, 2, 1); | |
2123 | VERIFY3U(ENOENT, ==, | |
2124 | spa_create("ztest_bad_mirror", nvroot, NULL, NULL, NULL)); | |
2125 | nvlist_free(nvroot); | |
2126 | ||
2127 | /* | |
2128 | * Attempt to create an existing pool. It shouldn't matter | |
2129 | * what's in the nvroot; we should fail with EEXIST. | |
2130 | */ | |
1e33ac1e | 2131 | (void) rw_enter(&zs->zs_name_lock, RW_READER); |
428870ff BB |
2132 | nvroot = make_vdev_root("/dev/bogus", NULL, 0, 0, 0, 0, 0, 1); |
2133 | VERIFY3U(EEXIST, ==, spa_create(zs->zs_pool, nvroot, NULL, NULL, NULL)); | |
2134 | nvlist_free(nvroot); | |
2135 | VERIFY3U(0, ==, spa_open(zs->zs_pool, &spa, FTAG)); | |
2136 | VERIFY3U(EBUSY, ==, spa_destroy(zs->zs_pool)); | |
2137 | spa_close(spa, FTAG); | |
2138 | ||
1e33ac1e | 2139 | (void) rw_exit(&zs->zs_name_lock); |
428870ff BB |
2140 | } |
2141 | ||
2142 | static vdev_t * | |
2143 | vdev_lookup_by_path(vdev_t *vd, const char *path) | |
2144 | { | |
2145 | vdev_t *mvd; | |
d6320ddb | 2146 | int c; |
428870ff BB |
2147 | |
2148 | if (vd->vdev_path != NULL && strcmp(path, vd->vdev_path) == 0) | |
2149 | return (vd); | |
2150 | ||
d6320ddb | 2151 | for (c = 0; c < vd->vdev_children; c++) |
428870ff BB |
2152 | if ((mvd = vdev_lookup_by_path(vd->vdev_child[c], path)) != |
2153 | NULL) | |
2154 | return (mvd); | |
2155 | ||
2156 | return (NULL); | |
2157 | } | |
2158 | ||
2159 | /* | |
2160 | * Find the first available hole which can be used as a top-level. | |
2161 | */ | |
2162 | int | |
2163 | find_vdev_hole(spa_t *spa) | |
2164 | { | |
2165 | vdev_t *rvd = spa->spa_root_vdev; | |
2166 | int c; | |
2167 | ||
2168 | ASSERT(spa_config_held(spa, SCL_VDEV, RW_READER) == SCL_VDEV); | |
2169 | ||
2170 | for (c = 0; c < rvd->vdev_children; c++) { | |
2171 | vdev_t *cvd = rvd->vdev_child[c]; | |
2172 | ||
2173 | if (cvd->vdev_ishole) | |
2174 | break; | |
2175 | } | |
2176 | return (c); | |
2177 | } | |
2178 | ||
2179 | /* | |
2180 | * Verify that vdev_add() works as expected. | |
2181 | */ | |
2182 | /* ARGSUSED */ | |
2183 | void | |
2184 | ztest_vdev_add_remove(ztest_ds_t *zd, uint64_t id) | |
2185 | { | |
2186 | ztest_shared_t *zs = ztest_shared; | |
2187 | spa_t *spa = zs->zs_spa; | |
2188 | uint64_t leaves; | |
2189 | uint64_t guid; | |
2190 | nvlist_t *nvroot; | |
2191 | int error; | |
2192 | ||
1e33ac1e | 2193 | mutex_enter(&zs->zs_vdev_lock); |
428870ff BB |
2194 | leaves = MAX(zs->zs_mirrors + zs->zs_splits, 1) * zopt_raidz; |
2195 | ||
2196 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
2197 | ||
2198 | ztest_shared->zs_vdev_next_leaf = find_vdev_hole(spa) * leaves; | |
2199 | ||
2200 | /* | |
2201 | * If we have slogs then remove them 1/4 of the time. | |
2202 | */ | |
2203 | if (spa_has_slogs(spa) && ztest_random(4) == 0) { | |
2204 | /* | |
2205 | * Grab the guid from the head of the log class rotor. | |
2206 | */ | |
2207 | guid = spa_log_class(spa)->mc_rotor->mg_vd->vdev_guid; | |
2208 | ||
2209 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
2210 | ||
2211 | /* | |
2212 | * We have to grab the zs_name_lock as writer to | |
2213 | * prevent a race between removing a slog (dmu_objset_find) | |
2214 | * and destroying a dataset. Removing the slog will | |
2215 | * grab a reference on the dataset which may cause | |
2216 | * dmu_objset_destroy() to fail with EBUSY thus | |
2217 | * leaving the dataset in an inconsistent state. | |
2218 | */ | |
1e33ac1e | 2219 | rw_enter(&ztest_shared->zs_name_lock, RW_WRITER); |
428870ff | 2220 | error = spa_vdev_remove(spa, guid, B_FALSE); |
1e33ac1e | 2221 | rw_exit(&ztest_shared->zs_name_lock); |
428870ff BB |
2222 | |
2223 | if (error && error != EEXIST) | |
2224 | fatal(0, "spa_vdev_remove() = %d", error); | |
2225 | } else { | |
2226 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
2227 | ||
2228 | /* | |
2229 | * Make 1/4 of the devices be log devices. | |
2230 | */ | |
2231 | nvroot = make_vdev_root(NULL, NULL, zopt_vdev_size, 0, | |
2232 | ztest_random(4) == 0, zopt_raidz, zs->zs_mirrors, 1); | |
2233 | ||
2234 | error = spa_vdev_add(spa, nvroot); | |
2235 | nvlist_free(nvroot); | |
2236 | ||
2237 | if (error == ENOSPC) | |
2238 | ztest_record_enospc("spa_vdev_add"); | |
2239 | else if (error != 0) | |
2240 | fatal(0, "spa_vdev_add() = %d", error); | |
2241 | } | |
2242 | ||
1e33ac1e | 2243 | mutex_exit(&ztest_shared->zs_vdev_lock); |
428870ff BB |
2244 | } |
2245 | ||
2246 | /* | |
2247 | * Verify that adding/removing aux devices (l2arc, hot spare) works as expected. | |
2248 | */ | |
2249 | /* ARGSUSED */ | |
2250 | void | |
2251 | ztest_vdev_aux_add_remove(ztest_ds_t *zd, uint64_t id) | |
2252 | { | |
2253 | ztest_shared_t *zs = ztest_shared; | |
2254 | spa_t *spa = zs->zs_spa; | |
2255 | vdev_t *rvd = spa->spa_root_vdev; | |
2256 | spa_aux_vdev_t *sav; | |
2257 | char *aux; | |
40b84e7a | 2258 | char *path; |
428870ff BB |
2259 | uint64_t guid = 0; |
2260 | int error; | |
2261 | ||
40b84e7a BB |
2262 | path = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
2263 | ||
428870ff BB |
2264 | if (ztest_random(2) == 0) { |
2265 | sav = &spa->spa_spares; | |
2266 | aux = ZPOOL_CONFIG_SPARES; | |
2267 | } else { | |
2268 | sav = &spa->spa_l2cache; | |
2269 | aux = ZPOOL_CONFIG_L2CACHE; | |
2270 | } | |
2271 | ||
1e33ac1e | 2272 | mutex_enter(&zs->zs_vdev_lock); |
428870ff BB |
2273 | |
2274 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
2275 | ||
2276 | if (sav->sav_count != 0 && ztest_random(4) == 0) { | |
b128c09f BB |
2277 | /* |
2278 | * Pick a random device to remove. | |
2279 | */ | |
2280 | guid = sav->sav_vdevs[ztest_random(sav->sav_count)]->vdev_guid; | |
2281 | } else { | |
2282 | /* | |
2283 | * Find an unused device we can add. | |
2284 | */ | |
428870ff | 2285 | zs->zs_vdev_aux = 0; |
b128c09f | 2286 | for (;;) { |
b128c09f BB |
2287 | int c; |
2288 | (void) sprintf(path, ztest_aux_template, zopt_dir, | |
428870ff | 2289 | zopt_pool, aux, zs->zs_vdev_aux); |
b128c09f BB |
2290 | for (c = 0; c < sav->sav_count; c++) |
2291 | if (strcmp(sav->sav_vdevs[c]->vdev_path, | |
2292 | path) == 0) | |
2293 | break; | |
2294 | if (c == sav->sav_count && | |
2295 | vdev_lookup_by_path(rvd, path) == NULL) | |
2296 | break; | |
428870ff | 2297 | zs->zs_vdev_aux++; |
34dc7c2f BB |
2298 | } |
2299 | } | |
2300 | ||
b128c09f | 2301 | spa_config_exit(spa, SCL_VDEV, FTAG); |
34dc7c2f | 2302 | |
b128c09f BB |
2303 | if (guid == 0) { |
2304 | /* | |
2305 | * Add a new device. | |
2306 | */ | |
2307 | nvlist_t *nvroot = make_vdev_root(NULL, aux, | |
2308 | (zopt_vdev_size * 5) / 4, 0, 0, 0, 0, 1); | |
2309 | error = spa_vdev_add(spa, nvroot); | |
2310 | if (error != 0) | |
2311 | fatal(0, "spa_vdev_add(%p) = %d", nvroot, error); | |
2312 | nvlist_free(nvroot); | |
2313 | } else { | |
2314 | /* | |
2315 | * Remove an existing device. Sometimes, dirty its | |
2316 | * vdev state first to make sure we handle removal | |
2317 | * of devices that have pending state changes. | |
2318 | */ | |
2319 | if (ztest_random(2) == 0) | |
9babb374 | 2320 | (void) vdev_online(spa, guid, 0, NULL); |
b128c09f BB |
2321 | |
2322 | error = spa_vdev_remove(spa, guid, B_FALSE); | |
2323 | if (error != 0 && error != EBUSY) | |
2324 | fatal(0, "spa_vdev_remove(%llu) = %d", guid, error); | |
2325 | } | |
2326 | ||
1e33ac1e | 2327 | mutex_exit(&zs->zs_vdev_lock); |
40b84e7a BB |
2328 | |
2329 | umem_free(path, MAXPATHLEN); | |
428870ff BB |
2330 | } |
2331 | ||
2332 | /* | |
2333 | * split a pool if it has mirror tlvdevs | |
2334 | */ | |
2335 | /* ARGSUSED */ | |
2336 | void | |
2337 | ztest_split_pool(ztest_ds_t *zd, uint64_t id) | |
2338 | { | |
2339 | ztest_shared_t *zs = ztest_shared; | |
2340 | spa_t *spa = zs->zs_spa; | |
2341 | vdev_t *rvd = spa->spa_root_vdev; | |
2342 | nvlist_t *tree, **child, *config, *split, **schild; | |
2343 | uint_t c, children, schildren = 0, lastlogid = 0; | |
2344 | int error = 0; | |
2345 | ||
1e33ac1e | 2346 | mutex_enter(&zs->zs_vdev_lock); |
428870ff BB |
2347 | |
2348 | /* ensure we have a useable config; mirrors of raidz aren't supported */ | |
2349 | if (zs->zs_mirrors < 3 || zopt_raidz > 1) { | |
1e33ac1e | 2350 | mutex_exit(&zs->zs_vdev_lock); |
428870ff BB |
2351 | return; |
2352 | } | |
2353 | ||
2354 | /* clean up the old pool, if any */ | |
2355 | (void) spa_destroy("splitp"); | |
2356 | ||
2357 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
2358 | ||
2359 | /* generate a config from the existing config */ | |
2360 | mutex_enter(&spa->spa_props_lock); | |
2361 | VERIFY(nvlist_lookup_nvlist(spa->spa_config, ZPOOL_CONFIG_VDEV_TREE, | |
2362 | &tree) == 0); | |
2363 | mutex_exit(&spa->spa_props_lock); | |
2364 | ||
2365 | VERIFY(nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child, | |
2366 | &children) == 0); | |
2367 | ||
2368 | schild = malloc(rvd->vdev_children * sizeof (nvlist_t *)); | |
2369 | for (c = 0; c < children; c++) { | |
2370 | vdev_t *tvd = rvd->vdev_child[c]; | |
2371 | nvlist_t **mchild; | |
2372 | uint_t mchildren; | |
2373 | ||
2374 | if (tvd->vdev_islog || tvd->vdev_ops == &vdev_hole_ops) { | |
2375 | VERIFY(nvlist_alloc(&schild[schildren], NV_UNIQUE_NAME, | |
2376 | 0) == 0); | |
2377 | VERIFY(nvlist_add_string(schild[schildren], | |
2378 | ZPOOL_CONFIG_TYPE, VDEV_TYPE_HOLE) == 0); | |
2379 | VERIFY(nvlist_add_uint64(schild[schildren], | |
2380 | ZPOOL_CONFIG_IS_HOLE, 1) == 0); | |
2381 | if (lastlogid == 0) | |
2382 | lastlogid = schildren; | |
2383 | ++schildren; | |
2384 | continue; | |
2385 | } | |
2386 | lastlogid = 0; | |
2387 | VERIFY(nvlist_lookup_nvlist_array(child[c], | |
2388 | ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0); | |
2389 | VERIFY(nvlist_dup(mchild[0], &schild[schildren++], 0) == 0); | |
2390 | } | |
2391 | ||
2392 | /* OK, create a config that can be used to split */ | |
2393 | VERIFY(nvlist_alloc(&split, NV_UNIQUE_NAME, 0) == 0); | |
2394 | VERIFY(nvlist_add_string(split, ZPOOL_CONFIG_TYPE, | |
2395 | VDEV_TYPE_ROOT) == 0); | |
2396 | VERIFY(nvlist_add_nvlist_array(split, ZPOOL_CONFIG_CHILDREN, schild, | |
2397 | lastlogid != 0 ? lastlogid : schildren) == 0); | |
2398 | ||
2399 | VERIFY(nvlist_alloc(&config, NV_UNIQUE_NAME, 0) == 0); | |
2400 | VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, split) == 0); | |
2401 | ||
2402 | for (c = 0; c < schildren; c++) | |
2403 | nvlist_free(schild[c]); | |
2404 | free(schild); | |
2405 | nvlist_free(split); | |
2406 | ||
2407 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
2408 | ||
1e33ac1e | 2409 | (void) rw_enter(&zs->zs_name_lock, RW_WRITER); |
428870ff | 2410 | error = spa_vdev_split_mirror(spa, "splitp", config, NULL, B_FALSE); |
1e33ac1e | 2411 | (void) rw_exit(&zs->zs_name_lock); |
428870ff BB |
2412 | |
2413 | nvlist_free(config); | |
2414 | ||
2415 | if (error == 0) { | |
2416 | (void) printf("successful split - results:\n"); | |
2417 | mutex_enter(&spa_namespace_lock); | |
2418 | show_pool_stats(spa); | |
2419 | show_pool_stats(spa_lookup("splitp")); | |
2420 | mutex_exit(&spa_namespace_lock); | |
2421 | ++zs->zs_splits; | |
2422 | --zs->zs_mirrors; | |
2423 | } | |
1e33ac1e | 2424 | mutex_exit(&zs->zs_vdev_lock); |
428870ff | 2425 | |
34dc7c2f BB |
2426 | } |
2427 | ||
2428 | /* | |
2429 | * Verify that we can attach and detach devices. | |
2430 | */ | |
428870ff | 2431 | /* ARGSUSED */ |
34dc7c2f | 2432 | void |
428870ff | 2433 | ztest_vdev_attach_detach(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 2434 | { |
428870ff BB |
2435 | ztest_shared_t *zs = ztest_shared; |
2436 | spa_t *spa = zs->zs_spa; | |
b128c09f | 2437 | spa_aux_vdev_t *sav = &spa->spa_spares; |
34dc7c2f BB |
2438 | vdev_t *rvd = spa->spa_root_vdev; |
2439 | vdev_t *oldvd, *newvd, *pvd; | |
b128c09f | 2440 | nvlist_t *root; |
428870ff | 2441 | uint64_t leaves; |
34dc7c2f BB |
2442 | uint64_t leaf, top; |
2443 | uint64_t ashift = ztest_get_ashift(); | |
fb5f0bc8 | 2444 | uint64_t oldguid, pguid; |
34dc7c2f | 2445 | size_t oldsize, newsize; |
40b84e7a | 2446 | char *oldpath, *newpath; |
34dc7c2f | 2447 | int replacing; |
b128c09f BB |
2448 | int oldvd_has_siblings = B_FALSE; |
2449 | int newvd_is_spare = B_FALSE; | |
2450 | int oldvd_is_log; | |
34dc7c2f | 2451 | int error, expected_error; |
34dc7c2f | 2452 | |
40b84e7a BB |
2453 | oldpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
2454 | newpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
2455 | ||
1e33ac1e | 2456 | mutex_enter(&zs->zs_vdev_lock); |
428870ff | 2457 | leaves = MAX(zs->zs_mirrors, 1) * zopt_raidz; |
34dc7c2f | 2458 | |
b128c09f | 2459 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); |
34dc7c2f BB |
2460 | |
2461 | /* | |
2462 | * Decide whether to do an attach or a replace. | |
2463 | */ | |
2464 | replacing = ztest_random(2); | |
2465 | ||
2466 | /* | |
2467 | * Pick a random top-level vdev. | |
2468 | */ | |
428870ff | 2469 | top = ztest_random_vdev_top(spa, B_TRUE); |
34dc7c2f BB |
2470 | |
2471 | /* | |
2472 | * Pick a random leaf within it. | |
2473 | */ | |
2474 | leaf = ztest_random(leaves); | |
2475 | ||
2476 | /* | |
b128c09f | 2477 | * Locate this vdev. |
34dc7c2f | 2478 | */ |
b128c09f | 2479 | oldvd = rvd->vdev_child[top]; |
428870ff | 2480 | if (zs->zs_mirrors >= 1) { |
fb5f0bc8 | 2481 | ASSERT(oldvd->vdev_ops == &vdev_mirror_ops); |
428870ff | 2482 | ASSERT(oldvd->vdev_children >= zs->zs_mirrors); |
b128c09f | 2483 | oldvd = oldvd->vdev_child[leaf / zopt_raidz]; |
fb5f0bc8 BB |
2484 | } |
2485 | if (zopt_raidz > 1) { | |
2486 | ASSERT(oldvd->vdev_ops == &vdev_raidz_ops); | |
2487 | ASSERT(oldvd->vdev_children == zopt_raidz); | |
b128c09f | 2488 | oldvd = oldvd->vdev_child[leaf % zopt_raidz]; |
fb5f0bc8 | 2489 | } |
34dc7c2f BB |
2490 | |
2491 | /* | |
b128c09f BB |
2492 | * If we're already doing an attach or replace, oldvd may be a |
2493 | * mirror vdev -- in which case, pick a random child. | |
34dc7c2f | 2494 | */ |
b128c09f BB |
2495 | while (oldvd->vdev_children != 0) { |
2496 | oldvd_has_siblings = B_TRUE; | |
fb5f0bc8 BB |
2497 | ASSERT(oldvd->vdev_children >= 2); |
2498 | oldvd = oldvd->vdev_child[ztest_random(oldvd->vdev_children)]; | |
b128c09f BB |
2499 | } |
2500 | ||
2501 | oldguid = oldvd->vdev_guid; | |
9babb374 | 2502 | oldsize = vdev_get_min_asize(oldvd); |
b128c09f BB |
2503 | oldvd_is_log = oldvd->vdev_top->vdev_islog; |
2504 | (void) strcpy(oldpath, oldvd->vdev_path); | |
2505 | pvd = oldvd->vdev_parent; | |
fb5f0bc8 | 2506 | pguid = pvd->vdev_guid; |
34dc7c2f BB |
2507 | |
2508 | /* | |
b128c09f | 2509 | * If oldvd has siblings, then half of the time, detach it. |
34dc7c2f | 2510 | */ |
b128c09f BB |
2511 | if (oldvd_has_siblings && ztest_random(2) == 0) { |
2512 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
fb5f0bc8 BB |
2513 | error = spa_vdev_detach(spa, oldguid, pguid, B_FALSE); |
2514 | if (error != 0 && error != ENODEV && error != EBUSY && | |
2515 | error != ENOTSUP) | |
2516 | fatal(0, "detach (%s) returned %d", oldpath, error); | |
40b84e7a | 2517 | goto out; |
b128c09f | 2518 | } |
34dc7c2f BB |
2519 | |
2520 | /* | |
b128c09f BB |
2521 | * For the new vdev, choose with equal probability between the two |
2522 | * standard paths (ending in either 'a' or 'b') or a random hot spare. | |
34dc7c2f | 2523 | */ |
b128c09f BB |
2524 | if (sav->sav_count != 0 && ztest_random(3) == 0) { |
2525 | newvd = sav->sav_vdevs[ztest_random(sav->sav_count)]; | |
2526 | newvd_is_spare = B_TRUE; | |
2527 | (void) strcpy(newpath, newvd->vdev_path); | |
2528 | } else { | |
2529 | (void) snprintf(newpath, sizeof (newpath), ztest_dev_template, | |
2530 | zopt_dir, zopt_pool, top * leaves + leaf); | |
2531 | if (ztest_random(2) == 0) | |
2532 | newpath[strlen(newpath) - 1] = 'b'; | |
2533 | newvd = vdev_lookup_by_path(rvd, newpath); | |
2534 | } | |
2535 | ||
2536 | if (newvd) { | |
9babb374 | 2537 | newsize = vdev_get_min_asize(newvd); |
b128c09f BB |
2538 | } else { |
2539 | /* | |
2540 | * Make newsize a little bigger or smaller than oldsize. | |
2541 | * If it's smaller, the attach should fail. | |
2542 | * If it's larger, and we're doing a replace, | |
2543 | * we should get dynamic LUN growth when we're done. | |
2544 | */ | |
2545 | newsize = 10 * oldsize / (9 + ztest_random(3)); | |
2546 | } | |
34dc7c2f BB |
2547 | |
2548 | /* | |
2549 | * If pvd is not a mirror or root, the attach should fail with ENOTSUP, | |
2550 | * unless it's a replace; in that case any non-replacing parent is OK. | |
2551 | * | |
2552 | * If newvd is already part of the pool, it should fail with EBUSY. | |
2553 | * | |
2554 | * If newvd is too small, it should fail with EOVERFLOW. | |
2555 | */ | |
b128c09f BB |
2556 | if (pvd->vdev_ops != &vdev_mirror_ops && |
2557 | pvd->vdev_ops != &vdev_root_ops && (!replacing || | |
2558 | pvd->vdev_ops == &vdev_replacing_ops || | |
2559 | pvd->vdev_ops == &vdev_spare_ops)) | |
34dc7c2f | 2560 | expected_error = ENOTSUP; |
b128c09f BB |
2561 | else if (newvd_is_spare && (!replacing || oldvd_is_log)) |
2562 | expected_error = ENOTSUP; | |
2563 | else if (newvd == oldvd) | |
2564 | expected_error = replacing ? 0 : EBUSY; | |
2565 | else if (vdev_lookup_by_path(rvd, newpath) != NULL) | |
2566 | expected_error = EBUSY; | |
34dc7c2f BB |
2567 | else if (newsize < oldsize) |
2568 | expected_error = EOVERFLOW; | |
2569 | else if (ashift > oldvd->vdev_top->vdev_ashift) | |
2570 | expected_error = EDOM; | |
2571 | else | |
2572 | expected_error = 0; | |
2573 | ||
b128c09f | 2574 | spa_config_exit(spa, SCL_VDEV, FTAG); |
34dc7c2f BB |
2575 | |
2576 | /* | |
2577 | * Build the nvlist describing newpath. | |
2578 | */ | |
b128c09f BB |
2579 | root = make_vdev_root(newpath, NULL, newvd == NULL ? newsize : 0, |
2580 | ashift, 0, 0, 0, 1); | |
34dc7c2f | 2581 | |
b128c09f | 2582 | error = spa_vdev_attach(spa, oldguid, root, replacing); |
34dc7c2f | 2583 | |
34dc7c2f BB |
2584 | nvlist_free(root); |
2585 | ||
2586 | /* | |
2587 | * If our parent was the replacing vdev, but the replace completed, | |
2588 | * then instead of failing with ENOTSUP we may either succeed, | |
2589 | * fail with ENODEV, or fail with EOVERFLOW. | |
2590 | */ | |
2591 | if (expected_error == ENOTSUP && | |
2592 | (error == 0 || error == ENODEV || error == EOVERFLOW)) | |
2593 | expected_error = error; | |
2594 | ||
2595 | /* | |
2596 | * If someone grew the LUN, the replacement may be too small. | |
2597 | */ | |
b128c09f | 2598 | if (error == EOVERFLOW || error == EBUSY) |
34dc7c2f BB |
2599 | expected_error = error; |
2600 | ||
b128c09f BB |
2601 | /* XXX workaround 6690467 */ |
2602 | if (error != expected_error && expected_error != EBUSY) { | |
2603 | fatal(0, "attach (%s %llu, %s %llu, %d) " | |
2604 | "returned %d, expected %d", | |
2605 | oldpath, (longlong_t)oldsize, newpath, | |
2606 | (longlong_t)newsize, replacing, error, expected_error); | |
34dc7c2f | 2607 | } |
40b84e7a | 2608 | out: |
1e33ac1e | 2609 | mutex_exit(&zs->zs_vdev_lock); |
40b84e7a BB |
2610 | |
2611 | umem_free(oldpath, MAXPATHLEN); | |
2612 | umem_free(newpath, MAXPATHLEN); | |
34dc7c2f BB |
2613 | } |
2614 | ||
9babb374 BB |
2615 | /* |
2616 | * Callback function which expands the physical size of the vdev. | |
2617 | */ | |
2618 | vdev_t * | |
2619 | grow_vdev(vdev_t *vd, void *arg) | |
2620 | { | |
1fde1e37 | 2621 | ASSERTV(spa_t *spa = vd->vdev_spa); |
9babb374 BB |
2622 | size_t *newsize = arg; |
2623 | size_t fsize; | |
2624 | int fd; | |
2625 | ||
2626 | ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE); | |
2627 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
2628 | ||
2629 | if ((fd = open(vd->vdev_path, O_RDWR)) == -1) | |
2630 | return (vd); | |
2631 | ||
2632 | fsize = lseek(fd, 0, SEEK_END); | |
0e5b68e0 | 2633 | VERIFY(ftruncate(fd, *newsize) == 0); |
9babb374 BB |
2634 | |
2635 | if (zopt_verbose >= 6) { | |
2636 | (void) printf("%s grew from %lu to %lu bytes\n", | |
2637 | vd->vdev_path, (ulong_t)fsize, (ulong_t)*newsize); | |
2638 | } | |
2639 | (void) close(fd); | |
2640 | return (NULL); | |
2641 | } | |
2642 | ||
2643 | /* | |
2644 | * Callback function which expands a given vdev by calling vdev_online(). | |
2645 | */ | |
2646 | /* ARGSUSED */ | |
2647 | vdev_t * | |
2648 | online_vdev(vdev_t *vd, void *arg) | |
2649 | { | |
2650 | spa_t *spa = vd->vdev_spa; | |
2651 | vdev_t *tvd = vd->vdev_top; | |
9babb374 | 2652 | uint64_t guid = vd->vdev_guid; |
428870ff BB |
2653 | uint64_t generation = spa->spa_config_generation + 1; |
2654 | vdev_state_t newstate = VDEV_STATE_UNKNOWN; | |
2655 | int error; | |
9babb374 BB |
2656 | |
2657 | ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE); | |
2658 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
2659 | ||
2660 | /* Calling vdev_online will initialize the new metaslabs */ | |
2661 | spa_config_exit(spa, SCL_STATE, spa); | |
428870ff | 2662 | error = vdev_online(spa, guid, ZFS_ONLINE_EXPAND, &newstate); |
9babb374 BB |
2663 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); |
2664 | ||
428870ff BB |
2665 | /* |
2666 | * If vdev_online returned an error or the underlying vdev_open | |
2667 | * failed then we abort the expand. The only way to know that | |
2668 | * vdev_open fails is by checking the returned newstate. | |
2669 | */ | |
2670 | if (error || newstate != VDEV_STATE_HEALTHY) { | |
2671 | if (zopt_verbose >= 5) { | |
2672 | (void) printf("Unable to expand vdev, state %llu, " | |
2673 | "error %d\n", (u_longlong_t)newstate, error); | |
2674 | } | |
2675 | return (vd); | |
2676 | } | |
2677 | ASSERT3U(newstate, ==, VDEV_STATE_HEALTHY); | |
2678 | ||
9babb374 BB |
2679 | /* |
2680 | * Since we dropped the lock we need to ensure that we're | |
2681 | * still talking to the original vdev. It's possible this | |
2682 | * vdev may have been detached/replaced while we were | |
2683 | * trying to online it. | |
2684 | */ | |
428870ff BB |
2685 | if (generation != spa->spa_config_generation) { |
2686 | if (zopt_verbose >= 5) { | |
2687 | (void) printf("vdev configuration has changed, " | |
2688 | "guid %llu, state %llu, expected gen %llu, " | |
2689 | "got gen %llu\n", | |
2690 | (u_longlong_t)guid, | |
2691 | (u_longlong_t)tvd->vdev_state, | |
2692 | (u_longlong_t)generation, | |
2693 | (u_longlong_t)spa->spa_config_generation); | |
9babb374 BB |
2694 | } |
2695 | return (vd); | |
2696 | } | |
2697 | return (NULL); | |
2698 | } | |
2699 | ||
2700 | /* | |
2701 | * Traverse the vdev tree calling the supplied function. | |
2702 | * We continue to walk the tree until we either have walked all | |
2703 | * children or we receive a non-NULL return from the callback. | |
2704 | * If a NULL callback is passed, then we just return back the first | |
2705 | * leaf vdev we encounter. | |
2706 | */ | |
2707 | vdev_t * | |
2708 | vdev_walk_tree(vdev_t *vd, vdev_t *(*func)(vdev_t *, void *), void *arg) | |
2709 | { | |
d6320ddb BB |
2710 | uint_t c; |
2711 | ||
9babb374 BB |
2712 | if (vd->vdev_ops->vdev_op_leaf) { |
2713 | if (func == NULL) | |
2714 | return (vd); | |
2715 | else | |
2716 | return (func(vd, arg)); | |
2717 | } | |
2718 | ||
d6320ddb | 2719 | for (c = 0; c < vd->vdev_children; c++) { |
9babb374 BB |
2720 | vdev_t *cvd = vd->vdev_child[c]; |
2721 | if ((cvd = vdev_walk_tree(cvd, func, arg)) != NULL) | |
2722 | return (cvd); | |
2723 | } | |
2724 | return (NULL); | |
2725 | } | |
2726 | ||
34dc7c2f BB |
2727 | /* |
2728 | * Verify that dynamic LUN growth works as expected. | |
2729 | */ | |
428870ff | 2730 | /* ARGSUSED */ |
34dc7c2f | 2731 | void |
428870ff | 2732 | ztest_vdev_LUN_growth(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 2733 | { |
428870ff BB |
2734 | ztest_shared_t *zs = ztest_shared; |
2735 | spa_t *spa = zs->zs_spa; | |
2736 | vdev_t *vd, *tvd; | |
2737 | metaslab_class_t *mc; | |
2738 | metaslab_group_t *mg; | |
9babb374 | 2739 | size_t psize, newsize; |
428870ff BB |
2740 | uint64_t top; |
2741 | uint64_t old_class_space, new_class_space, old_ms_count, new_ms_count; | |
34dc7c2f | 2742 | |
1e33ac1e | 2743 | mutex_enter(&zs->zs_vdev_lock); |
9babb374 BB |
2744 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); |
2745 | ||
428870ff | 2746 | top = ztest_random_vdev_top(spa, B_TRUE); |
9babb374 | 2747 | |
428870ff BB |
2748 | tvd = spa->spa_root_vdev->vdev_child[top]; |
2749 | mg = tvd->vdev_mg; | |
2750 | mc = mg->mg_class; | |
2751 | old_ms_count = tvd->vdev_ms_count; | |
2752 | old_class_space = metaslab_class_get_space(mc); | |
34dc7c2f BB |
2753 | |
2754 | /* | |
9babb374 BB |
2755 | * Determine the size of the first leaf vdev associated with |
2756 | * our top-level device. | |
34dc7c2f | 2757 | */ |
9babb374 BB |
2758 | vd = vdev_walk_tree(tvd, NULL, NULL); |
2759 | ASSERT3P(vd, !=, NULL); | |
2760 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
34dc7c2f | 2761 | |
9babb374 | 2762 | psize = vd->vdev_psize; |
34dc7c2f | 2763 | |
9babb374 | 2764 | /* |
428870ff BB |
2765 | * We only try to expand the vdev if it's healthy, less than 4x its |
2766 | * original size, and it has a valid psize. | |
9babb374 | 2767 | */ |
428870ff BB |
2768 | if (tvd->vdev_state != VDEV_STATE_HEALTHY || |
2769 | psize == 0 || psize >= 4 * zopt_vdev_size) { | |
9babb374 | 2770 | spa_config_exit(spa, SCL_STATE, spa); |
1e33ac1e | 2771 | mutex_exit(&zs->zs_vdev_lock); |
9babb374 BB |
2772 | return; |
2773 | } | |
2774 | ASSERT(psize > 0); | |
2775 | newsize = psize + psize / 8; | |
2776 | ASSERT3U(newsize, >, psize); | |
34dc7c2f | 2777 | |
9babb374 | 2778 | if (zopt_verbose >= 6) { |
428870ff | 2779 | (void) printf("Expanding LUN %s from %lu to %lu\n", |
9babb374 BB |
2780 | vd->vdev_path, (ulong_t)psize, (ulong_t)newsize); |
2781 | } | |
2782 | ||
9babb374 BB |
2783 | /* |
2784 | * Growing the vdev is a two step process: | |
2785 | * 1). expand the physical size (i.e. relabel) | |
2786 | * 2). online the vdev to create the new metaslabs | |
2787 | */ | |
2788 | if (vdev_walk_tree(tvd, grow_vdev, &newsize) != NULL || | |
2789 | vdev_walk_tree(tvd, online_vdev, NULL) != NULL || | |
2790 | tvd->vdev_state != VDEV_STATE_HEALTHY) { | |
2791 | if (zopt_verbose >= 5) { | |
2792 | (void) printf("Could not expand LUN because " | |
428870ff | 2793 | "the vdev configuration changed.\n"); |
34dc7c2f | 2794 | } |
428870ff | 2795 | spa_config_exit(spa, SCL_STATE, spa); |
1e33ac1e | 2796 | mutex_exit(&zs->zs_vdev_lock); |
9babb374 | 2797 | return; |
34dc7c2f BB |
2798 | } |
2799 | ||
428870ff | 2800 | spa_config_exit(spa, SCL_STATE, spa); |
9babb374 BB |
2801 | |
2802 | /* | |
2803 | * Expanding the LUN will update the config asynchronously, | |
2804 | * thus we must wait for the async thread to complete any | |
2805 | * pending tasks before proceeding. | |
2806 | */ | |
428870ff BB |
2807 | for (;;) { |
2808 | boolean_t done; | |
2809 | mutex_enter(&spa->spa_async_lock); | |
2810 | done = (spa->spa_async_thread == NULL && !spa->spa_async_tasks); | |
2811 | mutex_exit(&spa->spa_async_lock); | |
2812 | if (done) | |
2813 | break; | |
2814 | txg_wait_synced(spa_get_dsl(spa), 0); | |
2815 | (void) poll(NULL, 0, 100); | |
2816 | } | |
9babb374 BB |
2817 | |
2818 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); | |
428870ff BB |
2819 | |
2820 | tvd = spa->spa_root_vdev->vdev_child[top]; | |
2821 | new_ms_count = tvd->vdev_ms_count; | |
2822 | new_class_space = metaslab_class_get_space(mc); | |
2823 | ||
2824 | if (tvd->vdev_mg != mg || mg->mg_class != mc) { | |
2825 | if (zopt_verbose >= 5) { | |
2826 | (void) printf("Could not verify LUN expansion due to " | |
2827 | "intervening vdev offline or remove.\n"); | |
2828 | } | |
2829 | spa_config_exit(spa, SCL_STATE, spa); | |
1e33ac1e | 2830 | mutex_exit(&zs->zs_vdev_lock); |
428870ff BB |
2831 | return; |
2832 | } | |
2833 | ||
2834 | /* | |
2835 | * Make sure we were able to grow the vdev. | |
2836 | */ | |
2837 | if (new_ms_count <= old_ms_count) | |
2838 | fatal(0, "LUN expansion failed: ms_count %llu <= %llu\n", | |
2839 | old_ms_count, new_ms_count); | |
9babb374 BB |
2840 | |
2841 | /* | |
2842 | * Make sure we were able to grow the pool. | |
2843 | */ | |
428870ff BB |
2844 | if (new_class_space <= old_class_space) |
2845 | fatal(0, "LUN expansion failed: class_space %llu <= %llu\n", | |
2846 | old_class_space, new_class_space); | |
2847 | ||
2848 | if (zopt_verbose >= 5) { | |
9babb374 BB |
2849 | char oldnumbuf[6], newnumbuf[6]; |
2850 | ||
428870ff BB |
2851 | nicenum(old_class_space, oldnumbuf); |
2852 | nicenum(new_class_space, newnumbuf); | |
9babb374 BB |
2853 | (void) printf("%s grew from %s to %s\n", |
2854 | spa->spa_name, oldnumbuf, newnumbuf); | |
2855 | } | |
428870ff | 2856 | |
9babb374 | 2857 | spa_config_exit(spa, SCL_STATE, spa); |
1e33ac1e | 2858 | mutex_exit(&zs->zs_vdev_lock); |
34dc7c2f BB |
2859 | } |
2860 | ||
428870ff BB |
2861 | /* |
2862 | * Verify that dmu_objset_{create,destroy,open,close} work as expected. | |
2863 | */ | |
34dc7c2f BB |
2864 | /* ARGSUSED */ |
2865 | static void | |
428870ff | 2866 | ztest_objset_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) |
34dc7c2f BB |
2867 | { |
2868 | /* | |
428870ff | 2869 | * Create the objects common to all ztest datasets. |
34dc7c2f | 2870 | */ |
428870ff | 2871 | VERIFY(zap_create_claim(os, ZTEST_DIROBJ, |
34dc7c2f | 2872 | DMU_OT_ZAP_OTHER, DMU_OT_NONE, 0, tx) == 0); |
428870ff | 2873 | } |
34dc7c2f | 2874 | |
428870ff BB |
2875 | static int |
2876 | ztest_dataset_create(char *dsname) | |
2877 | { | |
2878 | uint64_t zilset = ztest_random(100); | |
2879 | int err = dmu_objset_create(dsname, DMU_OST_OTHER, 0, | |
2880 | ztest_objset_create_cb, NULL); | |
2881 | ||
2882 | if (err || zilset < 80) | |
2883 | return (err); | |
2884 | ||
2885 | (void) printf("Setting dataset %s to sync always\n", dsname); | |
2886 | return (ztest_dsl_prop_set_uint64(dsname, ZFS_PROP_SYNC, | |
2887 | ZFS_SYNC_ALWAYS, B_FALSE)); | |
34dc7c2f BB |
2888 | } |
2889 | ||
428870ff | 2890 | /* ARGSUSED */ |
34dc7c2f | 2891 | static int |
428870ff | 2892 | ztest_objset_destroy_cb(const char *name, void *arg) |
34dc7c2f | 2893 | { |
34dc7c2f | 2894 | objset_t *os; |
428870ff | 2895 | dmu_object_info_t doi; |
34dc7c2f BB |
2896 | int error; |
2897 | ||
2898 | /* | |
2899 | * Verify that the dataset contains a directory object. | |
2900 | */ | |
428870ff BB |
2901 | VERIFY3U(0, ==, dmu_objset_hold(name, FTAG, &os)); |
2902 | error = dmu_object_info(os, ZTEST_DIROBJ, &doi); | |
34dc7c2f BB |
2903 | if (error != ENOENT) { |
2904 | /* We could have crashed in the middle of destroying it */ | |
2905 | ASSERT3U(error, ==, 0); | |
428870ff BB |
2906 | ASSERT3U(doi.doi_type, ==, DMU_OT_ZAP_OTHER); |
2907 | ASSERT3S(doi.doi_physical_blocks_512, >=, 0); | |
34dc7c2f | 2908 | } |
428870ff | 2909 | dmu_objset_rele(os, FTAG); |
34dc7c2f BB |
2910 | |
2911 | /* | |
2912 | * Destroy the dataset. | |
2913 | */ | |
428870ff | 2914 | VERIFY3U(0, ==, dmu_objset_destroy(name, B_FALSE)); |
34dc7c2f BB |
2915 | return (0); |
2916 | } | |
2917 | ||
428870ff BB |
2918 | static boolean_t |
2919 | ztest_snapshot_create(char *osname, uint64_t id) | |
34dc7c2f | 2920 | { |
428870ff BB |
2921 | char snapname[MAXNAMELEN]; |
2922 | int error; | |
2923 | ||
2924 | (void) snprintf(snapname, MAXNAMELEN, "%s@%llu", osname, | |
2925 | (u_longlong_t)id); | |
2926 | ||
2927 | error = dmu_objset_snapshot(osname, strchr(snapname, '@') + 1, | |
572e2857 | 2928 | NULL, NULL, B_FALSE, B_FALSE, -1); |
428870ff BB |
2929 | if (error == ENOSPC) { |
2930 | ztest_record_enospc(FTAG); | |
2931 | return (B_FALSE); | |
2932 | } | |
2933 | if (error != 0 && error != EEXIST) | |
2934 | fatal(0, "ztest_snapshot_create(%s) = %d", snapname, error); | |
2935 | return (B_TRUE); | |
2936 | } | |
2937 | ||
2938 | static boolean_t | |
2939 | ztest_snapshot_destroy(char *osname, uint64_t id) | |
2940 | { | |
2941 | char snapname[MAXNAMELEN]; | |
2942 | int error; | |
2943 | ||
2944 | (void) snprintf(snapname, MAXNAMELEN, "%s@%llu", osname, | |
2945 | (u_longlong_t)id); | |
2946 | ||
2947 | error = dmu_objset_destroy(snapname, B_FALSE); | |
2948 | if (error != 0 && error != ENOENT) | |
2949 | fatal(0, "ztest_snapshot_destroy(%s) = %d", snapname, error); | |
2950 | return (B_TRUE); | |
34dc7c2f BB |
2951 | } |
2952 | ||
428870ff | 2953 | /* ARGSUSED */ |
34dc7c2f | 2954 | void |
428870ff | 2955 | ztest_dmu_objset_create_destroy(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 2956 | { |
428870ff | 2957 | ztest_shared_t *zs = ztest_shared; |
40b84e7a | 2958 | ztest_ds_t *zdtmp; |
428870ff | 2959 | int iters; |
34dc7c2f | 2960 | int error; |
b128c09f | 2961 | objset_t *os, *os2; |
40b84e7a | 2962 | char *name; |
34dc7c2f | 2963 | zilog_t *zilog; |
d6320ddb | 2964 | int i; |
34dc7c2f | 2965 | |
40b84e7a BB |
2966 | zdtmp = umem_alloc(sizeof (ztest_ds_t), UMEM_NOFAIL); |
2967 | name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
2968 | ||
1e33ac1e | 2969 | (void) rw_enter(&zs->zs_name_lock, RW_READER); |
34dc7c2f | 2970 | |
428870ff BB |
2971 | (void) snprintf(name, MAXNAMELEN, "%s/temp_%llu", |
2972 | zs->zs_pool, (u_longlong_t)id); | |
34dc7c2f BB |
2973 | |
2974 | /* | |
2975 | * If this dataset exists from a previous run, process its replay log | |
2976 | * half of the time. If we don't replay it, then dmu_objset_destroy() | |
428870ff | 2977 | * (invoked from ztest_objset_destroy_cb()) should just throw it away. |
34dc7c2f BB |
2978 | */ |
2979 | if (ztest_random(2) == 0 && | |
428870ff | 2980 | dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, FTAG, &os) == 0) { |
40b84e7a BB |
2981 | ztest_zd_init(zdtmp, os); |
2982 | zil_replay(os, zdtmp, ztest_replay_vector); | |
2983 | ztest_zd_fini(zdtmp); | |
428870ff | 2984 | dmu_objset_disown(os, FTAG); |
34dc7c2f BB |
2985 | } |
2986 | ||
2987 | /* | |
2988 | * There may be an old instance of the dataset we're about to | |
2989 | * create lying around from a previous run. If so, destroy it | |
2990 | * and all of its snapshots. | |
2991 | */ | |
428870ff | 2992 | (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL, |
34dc7c2f BB |
2993 | DS_FIND_CHILDREN | DS_FIND_SNAPSHOTS); |
2994 | ||
2995 | /* | |
2996 | * Verify that the destroyed dataset is no longer in the namespace. | |
2997 | */ | |
428870ff | 2998 | VERIFY3U(ENOENT, ==, dmu_objset_hold(name, FTAG, &os)); |
34dc7c2f BB |
2999 | |
3000 | /* | |
3001 | * Verify that we can create a new dataset. | |
3002 | */ | |
428870ff | 3003 | error = ztest_dataset_create(name); |
34dc7c2f BB |
3004 | if (error) { |
3005 | if (error == ENOSPC) { | |
428870ff | 3006 | ztest_record_enospc(FTAG); |
40b84e7a | 3007 | goto out; |
34dc7c2f BB |
3008 | } |
3009 | fatal(0, "dmu_objset_create(%s) = %d", name, error); | |
3010 | } | |
3011 | ||
428870ff BB |
3012 | VERIFY3U(0, ==, |
3013 | dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, FTAG, &os)); | |
3014 | ||
40b84e7a | 3015 | ztest_zd_init(zdtmp, os); |
34dc7c2f BB |
3016 | |
3017 | /* | |
3018 | * Open the intent log for it. | |
3019 | */ | |
428870ff | 3020 | zilog = zil_open(os, ztest_get_data); |
34dc7c2f BB |
3021 | |
3022 | /* | |
428870ff BB |
3023 | * Put some objects in there, do a little I/O to them, |
3024 | * and randomly take a couple of snapshots along the way. | |
34dc7c2f | 3025 | */ |
428870ff | 3026 | iters = ztest_random(5); |
d6320ddb | 3027 | for (i = 0; i < iters; i++) { |
40b84e7a | 3028 | ztest_dmu_object_alloc_free(zdtmp, id); |
428870ff BB |
3029 | if (ztest_random(iters) == 0) |
3030 | (void) ztest_snapshot_create(name, i); | |
34dc7c2f BB |
3031 | } |
3032 | ||
3033 | /* | |
3034 | * Verify that we cannot create an existing dataset. | |
3035 | */ | |
428870ff BB |
3036 | VERIFY3U(EEXIST, ==, |
3037 | dmu_objset_create(name, DMU_OST_OTHER, 0, NULL, NULL)); | |
34dc7c2f BB |
3038 | |
3039 | /* | |
428870ff | 3040 | * Verify that we can hold an objset that is also owned. |
b128c09f | 3041 | */ |
428870ff BB |
3042 | VERIFY3U(0, ==, dmu_objset_hold(name, FTAG, &os2)); |
3043 | dmu_objset_rele(os2, FTAG); | |
34dc7c2f | 3044 | |
428870ff BB |
3045 | /* |
3046 | * Verify that we cannot own an objset that is already owned. | |
3047 | */ | |
3048 | VERIFY3U(EBUSY, ==, | |
3049 | dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, FTAG, &os2)); | |
34dc7c2f | 3050 | |
428870ff BB |
3051 | zil_close(zilog); |
3052 | dmu_objset_disown(os, FTAG); | |
40b84e7a BB |
3053 | ztest_zd_fini(zdtmp); |
3054 | out: | |
1e33ac1e | 3055 | (void) rw_exit(&zs->zs_name_lock); |
40b84e7a BB |
3056 | |
3057 | umem_free(name, MAXNAMELEN); | |
3058 | umem_free(zdtmp, sizeof (ztest_ds_t)); | |
34dc7c2f BB |
3059 | } |
3060 | ||
3061 | /* | |
3062 | * Verify that dmu_snapshot_{create,destroy,open,close} work as expected. | |
3063 | */ | |
3064 | void | |
428870ff | 3065 | ztest_dmu_snapshot_create_destroy(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3066 | { |
428870ff | 3067 | ztest_shared_t *zs = ztest_shared; |
34dc7c2f | 3068 | |
1e33ac1e | 3069 | (void) rw_enter(&zs->zs_name_lock, RW_READER); |
428870ff BB |
3070 | (void) ztest_snapshot_destroy(zd->zd_name, id); |
3071 | (void) ztest_snapshot_create(zd->zd_name, id); | |
1e33ac1e | 3072 | (void) rw_exit(&zs->zs_name_lock); |
34dc7c2f BB |
3073 | } |
3074 | ||
9babb374 BB |
3075 | /* |
3076 | * Cleanup non-standard snapshots and clones. | |
3077 | */ | |
3078 | void | |
428870ff | 3079 | ztest_dsl_dataset_cleanup(char *osname, uint64_t id) |
9babb374 | 3080 | { |
40b84e7a BB |
3081 | char *snap1name; |
3082 | char *clone1name; | |
3083 | char *snap2name; | |
3084 | char *clone2name; | |
3085 | char *snap3name; | |
9babb374 BB |
3086 | int error; |
3087 | ||
40b84e7a BB |
3088 | snap1name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); |
3089 | clone1name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3090 | snap2name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3091 | clone2name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3092 | snap3name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3093 | ||
b8864a23 BB |
3094 | (void) snprintf(snap1name, MAXNAMELEN, "%s@s1_%llu", |
3095 | osname, (u_longlong_t)id); | |
3096 | (void) snprintf(clone1name, MAXNAMELEN, "%s/c1_%llu", | |
3097 | osname, (u_longlong_t)id); | |
3098 | (void) snprintf(snap2name, MAXNAMELEN, "%s@s2_%llu", | |
3099 | clone1name, (u_longlong_t)id); | |
3100 | (void) snprintf(clone2name, MAXNAMELEN, "%s/c2_%llu", | |
3101 | osname, (u_longlong_t)id); | |
3102 | (void) snprintf(snap3name, MAXNAMELEN, "%s@s3_%llu", | |
3103 | clone1name, (u_longlong_t)id); | |
9babb374 | 3104 | |
45d1cae3 | 3105 | error = dmu_objset_destroy(clone2name, B_FALSE); |
9babb374 BB |
3106 | if (error && error != ENOENT) |
3107 | fatal(0, "dmu_objset_destroy(%s) = %d", clone2name, error); | |
45d1cae3 | 3108 | error = dmu_objset_destroy(snap3name, B_FALSE); |
9babb374 BB |
3109 | if (error && error != ENOENT) |
3110 | fatal(0, "dmu_objset_destroy(%s) = %d", snap3name, error); | |
45d1cae3 | 3111 | error = dmu_objset_destroy(snap2name, B_FALSE); |
9babb374 BB |
3112 | if (error && error != ENOENT) |
3113 | fatal(0, "dmu_objset_destroy(%s) = %d", snap2name, error); | |
45d1cae3 | 3114 | error = dmu_objset_destroy(clone1name, B_FALSE); |
9babb374 BB |
3115 | if (error && error != ENOENT) |
3116 | fatal(0, "dmu_objset_destroy(%s) = %d", clone1name, error); | |
45d1cae3 | 3117 | error = dmu_objset_destroy(snap1name, B_FALSE); |
9babb374 BB |
3118 | if (error && error != ENOENT) |
3119 | fatal(0, "dmu_objset_destroy(%s) = %d", snap1name, error); | |
40b84e7a BB |
3120 | |
3121 | umem_free(snap1name, MAXNAMELEN); | |
3122 | umem_free(clone1name, MAXNAMELEN); | |
3123 | umem_free(snap2name, MAXNAMELEN); | |
3124 | umem_free(clone2name, MAXNAMELEN); | |
3125 | umem_free(snap3name, MAXNAMELEN); | |
9babb374 BB |
3126 | } |
3127 | ||
3128 | /* | |
3129 | * Verify dsl_dataset_promote handles EBUSY | |
3130 | */ | |
3131 | void | |
428870ff | 3132 | ztest_dsl_dataset_promote_busy(ztest_ds_t *zd, uint64_t id) |
9babb374 | 3133 | { |
428870ff | 3134 | ztest_shared_t *zs = ztest_shared; |
9babb374 BB |
3135 | objset_t *clone; |
3136 | dsl_dataset_t *ds; | |
40b84e7a BB |
3137 | char *snap1name; |
3138 | char *clone1name; | |
3139 | char *snap2name; | |
3140 | char *clone2name; | |
3141 | char *snap3name; | |
428870ff BB |
3142 | char *osname = zd->zd_name; |
3143 | int error; | |
9babb374 | 3144 | |
40b84e7a BB |
3145 | snap1name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); |
3146 | clone1name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3147 | snap2name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3148 | clone2name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3149 | snap3name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3150 | ||
1e33ac1e | 3151 | (void) rw_enter(&zs->zs_name_lock, RW_READER); |
9babb374 | 3152 | |
428870ff | 3153 | ztest_dsl_dataset_cleanup(osname, id); |
9babb374 | 3154 | |
b8864a23 BB |
3155 | (void) snprintf(snap1name, MAXNAMELEN, "%s@s1_%llu", |
3156 | osname, (u_longlong_t)id); | |
3157 | (void) snprintf(clone1name, MAXNAMELEN, "%s/c1_%llu", | |
3158 | osname, (u_longlong_t)id); | |
3159 | (void) snprintf(snap2name, MAXNAMELEN, "%s@s2_%llu", | |
3160 | clone1name, (u_longlong_t)id); | |
3161 | (void) snprintf(clone2name, MAXNAMELEN, "%s/c2_%llu", | |
3162 | osname, (u_longlong_t)id); | |
3163 | (void) snprintf(snap3name, MAXNAMELEN, "%s@s3_%llu", | |
3164 | clone1name, (u_longlong_t)id); | |
9babb374 BB |
3165 | |
3166 | error = dmu_objset_snapshot(osname, strchr(snap1name, '@')+1, | |
572e2857 | 3167 | NULL, NULL, B_FALSE, B_FALSE, -1); |
9babb374 BB |
3168 | if (error && error != EEXIST) { |
3169 | if (error == ENOSPC) { | |
428870ff | 3170 | ztest_record_enospc(FTAG); |
9babb374 BB |
3171 | goto out; |
3172 | } | |
3173 | fatal(0, "dmu_take_snapshot(%s) = %d", snap1name, error); | |
3174 | } | |
3175 | ||
428870ff | 3176 | error = dmu_objset_hold(snap1name, FTAG, &clone); |
9babb374 BB |
3177 | if (error) |
3178 | fatal(0, "dmu_open_snapshot(%s) = %d", snap1name, error); | |
3179 | ||
428870ff BB |
3180 | error = dmu_objset_clone(clone1name, dmu_objset_ds(clone), 0); |
3181 | dmu_objset_rele(clone, FTAG); | |
9babb374 BB |
3182 | if (error) { |
3183 | if (error == ENOSPC) { | |
428870ff | 3184 | ztest_record_enospc(FTAG); |
9babb374 BB |
3185 | goto out; |
3186 | } | |
3187 | fatal(0, "dmu_objset_create(%s) = %d", clone1name, error); | |
3188 | } | |
3189 | ||
428870ff | 3190 | error = dmu_objset_snapshot(clone1name, strchr(snap2name, '@')+1, |
572e2857 | 3191 | NULL, NULL, B_FALSE, B_FALSE, -1); |
428870ff BB |
3192 | if (error && error != EEXIST) { |
3193 | if (error == ENOSPC) { | |
3194 | ztest_record_enospc(FTAG); | |
3195 | goto out; | |
34dc7c2f | 3196 | } |
428870ff BB |
3197 | fatal(0, "dmu_open_snapshot(%s) = %d", snap2name, error); |
3198 | } | |
34dc7c2f | 3199 | |
428870ff | 3200 | error = dmu_objset_snapshot(clone1name, strchr(snap3name, '@')+1, |
572e2857 | 3201 | NULL, NULL, B_FALSE, B_FALSE, -1); |
428870ff BB |
3202 | if (error && error != EEXIST) { |
3203 | if (error == ENOSPC) { | |
3204 | ztest_record_enospc(FTAG); | |
3205 | goto out; | |
3206 | } | |
3207 | fatal(0, "dmu_open_snapshot(%s) = %d", snap3name, error); | |
3208 | } | |
34dc7c2f | 3209 | |
428870ff BB |
3210 | error = dmu_objset_hold(snap3name, FTAG, &clone); |
3211 | if (error) | |
3212 | fatal(0, "dmu_open_snapshot(%s) = %d", snap3name, error); | |
34dc7c2f | 3213 | |
428870ff BB |
3214 | error = dmu_objset_clone(clone2name, dmu_objset_ds(clone), 0); |
3215 | dmu_objset_rele(clone, FTAG); | |
3216 | if (error) { | |
3217 | if (error == ENOSPC) { | |
3218 | ztest_record_enospc(FTAG); | |
3219 | goto out; | |
3220 | } | |
3221 | fatal(0, "dmu_objset_create(%s) = %d", clone2name, error); | |
3222 | } | |
34dc7c2f | 3223 | |
428870ff BB |
3224 | error = dsl_dataset_own(snap2name, B_FALSE, FTAG, &ds); |
3225 | if (error) | |
3226 | fatal(0, "dsl_dataset_own(%s) = %d", snap2name, error); | |
3227 | error = dsl_dataset_promote(clone2name, NULL); | |
3228 | if (error != EBUSY) | |
3229 | fatal(0, "dsl_dataset_promote(%s), %d, not EBUSY", clone2name, | |
3230 | error); | |
3231 | dsl_dataset_disown(ds, FTAG); | |
34dc7c2f | 3232 | |
428870ff BB |
3233 | out: |
3234 | ztest_dsl_dataset_cleanup(osname, id); | |
34dc7c2f | 3235 | |
1e33ac1e | 3236 | (void) rw_exit(&zs->zs_name_lock); |
40b84e7a BB |
3237 | |
3238 | umem_free(snap1name, MAXNAMELEN); | |
3239 | umem_free(clone1name, MAXNAMELEN); | |
3240 | umem_free(snap2name, MAXNAMELEN); | |
3241 | umem_free(clone2name, MAXNAMELEN); | |
3242 | umem_free(snap3name, MAXNAMELEN); | |
428870ff | 3243 | } |
34dc7c2f | 3244 | |
40b84e7a BB |
3245 | #undef OD_ARRAY_SIZE |
3246 | #define OD_ARRAY_SIZE 4 | |
3247 | ||
428870ff BB |
3248 | /* |
3249 | * Verify that dmu_object_{alloc,free} work as expected. | |
3250 | */ | |
3251 | void | |
3252 | ztest_dmu_object_alloc_free(ztest_ds_t *zd, uint64_t id) | |
3253 | { | |
40b84e7a BB |
3254 | ztest_od_t *od; |
3255 | int batchsize; | |
3256 | int size; | |
d6320ddb | 3257 | int b; |
34dc7c2f | 3258 | |
40b84e7a BB |
3259 | size = sizeof(ztest_od_t) * OD_ARRAY_SIZE; |
3260 | od = umem_alloc(size, UMEM_NOFAIL); | |
3261 | batchsize = OD_ARRAY_SIZE; | |
3262 | ||
d6320ddb | 3263 | for (b = 0; b < batchsize; b++) |
40b84e7a | 3264 | ztest_od_init(od + b, id, FTAG, b, DMU_OT_UINT64_OTHER, 0, 0); |
34dc7c2f | 3265 | |
428870ff BB |
3266 | /* |
3267 | * Destroy the previous batch of objects, create a new batch, | |
3268 | * and do some I/O on the new objects. | |
3269 | */ | |
40b84e7a | 3270 | if (ztest_object_init(zd, od, size, B_TRUE) != 0) |
428870ff | 3271 | return; |
34dc7c2f | 3272 | |
428870ff BB |
3273 | while (ztest_random(4 * batchsize) != 0) |
3274 | ztest_io(zd, od[ztest_random(batchsize)].od_object, | |
3275 | ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
40b84e7a BB |
3276 | |
3277 | umem_free(od, size); | |
34dc7c2f BB |
3278 | } |
3279 | ||
40b84e7a BB |
3280 | #undef OD_ARRAY_SIZE |
3281 | #define OD_ARRAY_SIZE 2 | |
3282 | ||
34dc7c2f BB |
3283 | /* |
3284 | * Verify that dmu_{read,write} work as expected. | |
3285 | */ | |
34dc7c2f | 3286 | void |
428870ff | 3287 | ztest_dmu_read_write(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3288 | { |
40b84e7a BB |
3289 | int size; |
3290 | ztest_od_t *od; | |
3291 | ||
428870ff | 3292 | objset_t *os = zd->zd_os; |
40b84e7a BB |
3293 | size = sizeof(ztest_od_t) * OD_ARRAY_SIZE; |
3294 | od = umem_alloc(size, UMEM_NOFAIL); | |
34dc7c2f BB |
3295 | dmu_tx_t *tx; |
3296 | int i, freeit, error; | |
3297 | uint64_t n, s, txg; | |
3298 | bufwad_t *packbuf, *bigbuf, *pack, *bigH, *bigT; | |
428870ff BB |
3299 | uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize; |
3300 | uint64_t chunksize = (1000 + ztest_random(1000)) * sizeof (uint64_t); | |
34dc7c2f BB |
3301 | uint64_t regions = 997; |
3302 | uint64_t stride = 123456789ULL; | |
3303 | uint64_t width = 40; | |
3304 | int free_percent = 5; | |
3305 | ||
3306 | /* | |
3307 | * This test uses two objects, packobj and bigobj, that are always | |
3308 | * updated together (i.e. in the same tx) so that their contents are | |
3309 | * in sync and can be compared. Their contents relate to each other | |
3310 | * in a simple way: packobj is a dense array of 'bufwad' structures, | |
3311 | * while bigobj is a sparse array of the same bufwads. Specifically, | |
3312 | * for any index n, there are three bufwads that should be identical: | |
3313 | * | |
3314 | * packobj, at offset n * sizeof (bufwad_t) | |
3315 | * bigobj, at the head of the nth chunk | |
3316 | * bigobj, at the tail of the nth chunk | |
3317 | * | |
3318 | * The chunk size is arbitrary. It doesn't have to be a power of two, | |
3319 | * and it doesn't have any relation to the object blocksize. | |
3320 | * The only requirement is that it can hold at least two bufwads. | |
3321 | * | |
3322 | * Normally, we write the bufwad to each of these locations. | |
3323 | * However, free_percent of the time we instead write zeroes to | |
3324 | * packobj and perform a dmu_free_range() on bigobj. By comparing | |
3325 | * bigobj to packobj, we can verify that the DMU is correctly | |
3326 | * tracking which parts of an object are allocated and free, | |
3327 | * and that the contents of the allocated blocks are correct. | |
3328 | */ | |
3329 | ||
3330 | /* | |
3331 | * Read the directory info. If it's the first time, set things up. | |
3332 | */ | |
40b84e7a BB |
3333 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, chunksize); |
3334 | ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, chunksize); | |
34dc7c2f | 3335 | |
40b84e7a BB |
3336 | if (ztest_object_init(zd, od, size, B_FALSE) != 0) { |
3337 | umem_free(od, size); | |
428870ff | 3338 | return; |
40b84e7a | 3339 | } |
34dc7c2f | 3340 | |
428870ff BB |
3341 | bigobj = od[0].od_object; |
3342 | packobj = od[1].od_object; | |
3343 | chunksize = od[0].od_gen; | |
3344 | ASSERT(chunksize == od[1].od_gen); | |
34dc7c2f BB |
3345 | |
3346 | /* | |
3347 | * Prefetch a random chunk of the big object. | |
3348 | * Our aim here is to get some async reads in flight | |
3349 | * for blocks that we may free below; the DMU should | |
3350 | * handle this race correctly. | |
3351 | */ | |
3352 | n = ztest_random(regions) * stride + ztest_random(width); | |
3353 | s = 1 + ztest_random(2 * width - 1); | |
428870ff | 3354 | dmu_prefetch(os, bigobj, n * chunksize, s * chunksize); |
34dc7c2f BB |
3355 | |
3356 | /* | |
3357 | * Pick a random index and compute the offsets into packobj and bigobj. | |
3358 | */ | |
3359 | n = ztest_random(regions) * stride + ztest_random(width); | |
3360 | s = 1 + ztest_random(width - 1); | |
3361 | ||
3362 | packoff = n * sizeof (bufwad_t); | |
3363 | packsize = s * sizeof (bufwad_t); | |
3364 | ||
428870ff BB |
3365 | bigoff = n * chunksize; |
3366 | bigsize = s * chunksize; | |
34dc7c2f BB |
3367 | |
3368 | packbuf = umem_alloc(packsize, UMEM_NOFAIL); | |
3369 | bigbuf = umem_alloc(bigsize, UMEM_NOFAIL); | |
3370 | ||
3371 | /* | |
3372 | * free_percent of the time, free a range of bigobj rather than | |
3373 | * overwriting it. | |
3374 | */ | |
3375 | freeit = (ztest_random(100) < free_percent); | |
3376 | ||
3377 | /* | |
3378 | * Read the current contents of our objects. | |
3379 | */ | |
428870ff | 3380 | error = dmu_read(os, packobj, packoff, packsize, packbuf, |
9babb374 | 3381 | DMU_READ_PREFETCH); |
34dc7c2f | 3382 | ASSERT3U(error, ==, 0); |
428870ff | 3383 | error = dmu_read(os, bigobj, bigoff, bigsize, bigbuf, |
9babb374 | 3384 | DMU_READ_PREFETCH); |
34dc7c2f BB |
3385 | ASSERT3U(error, ==, 0); |
3386 | ||
3387 | /* | |
3388 | * Get a tx for the mods to both packobj and bigobj. | |
3389 | */ | |
3390 | tx = dmu_tx_create(os); | |
3391 | ||
428870ff | 3392 | dmu_tx_hold_write(tx, packobj, packoff, packsize); |
34dc7c2f BB |
3393 | |
3394 | if (freeit) | |
428870ff | 3395 | dmu_tx_hold_free(tx, bigobj, bigoff, bigsize); |
34dc7c2f | 3396 | else |
428870ff | 3397 | dmu_tx_hold_write(tx, bigobj, bigoff, bigsize); |
34dc7c2f | 3398 | |
428870ff BB |
3399 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); |
3400 | if (txg == 0) { | |
34dc7c2f BB |
3401 | umem_free(packbuf, packsize); |
3402 | umem_free(bigbuf, bigsize); | |
40b84e7a | 3403 | umem_free(od, size); |
34dc7c2f BB |
3404 | return; |
3405 | } | |
3406 | ||
428870ff BB |
3407 | dmu_object_set_checksum(os, bigobj, |
3408 | (enum zio_checksum)ztest_random_dsl_prop(ZFS_PROP_CHECKSUM), tx); | |
3409 | ||
3410 | dmu_object_set_compress(os, bigobj, | |
3411 | (enum zio_compress)ztest_random_dsl_prop(ZFS_PROP_COMPRESSION), tx); | |
34dc7c2f BB |
3412 | |
3413 | /* | |
3414 | * For each index from n to n + s, verify that the existing bufwad | |
3415 | * in packobj matches the bufwads at the head and tail of the | |
3416 | * corresponding chunk in bigobj. Then update all three bufwads | |
3417 | * with the new values we want to write out. | |
3418 | */ | |
3419 | for (i = 0; i < s; i++) { | |
3420 | /* LINTED */ | |
3421 | pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t)); | |
3422 | /* LINTED */ | |
428870ff | 3423 | bigH = (bufwad_t *)((char *)bigbuf + i * chunksize); |
34dc7c2f | 3424 | /* LINTED */ |
428870ff | 3425 | bigT = (bufwad_t *)((char *)bigH + chunksize) - 1; |
34dc7c2f BB |
3426 | |
3427 | ASSERT((uintptr_t)bigH - (uintptr_t)bigbuf < bigsize); | |
3428 | ASSERT((uintptr_t)bigT - (uintptr_t)bigbuf < bigsize); | |
3429 | ||
3430 | if (pack->bw_txg > txg) | |
3431 | fatal(0, "future leak: got %llx, open txg is %llx", | |
3432 | pack->bw_txg, txg); | |
3433 | ||
3434 | if (pack->bw_data != 0 && pack->bw_index != n + i) | |
3435 | fatal(0, "wrong index: got %llx, wanted %llx+%llx", | |
3436 | pack->bw_index, n, i); | |
3437 | ||
3438 | if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0) | |
3439 | fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH); | |
3440 | ||
3441 | if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0) | |
3442 | fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT); | |
3443 | ||
3444 | if (freeit) { | |
3445 | bzero(pack, sizeof (bufwad_t)); | |
3446 | } else { | |
3447 | pack->bw_index = n + i; | |
3448 | pack->bw_txg = txg; | |
3449 | pack->bw_data = 1 + ztest_random(-2ULL); | |
3450 | } | |
3451 | *bigH = *pack; | |
3452 | *bigT = *pack; | |
3453 | } | |
3454 | ||
3455 | /* | |
3456 | * We've verified all the old bufwads, and made new ones. | |
3457 | * Now write them out. | |
3458 | */ | |
428870ff | 3459 | dmu_write(os, packobj, packoff, packsize, packbuf, tx); |
34dc7c2f BB |
3460 | |
3461 | if (freeit) { | |
428870ff | 3462 | if (zopt_verbose >= 7) { |
34dc7c2f BB |
3463 | (void) printf("freeing offset %llx size %llx" |
3464 | " txg %llx\n", | |
3465 | (u_longlong_t)bigoff, | |
3466 | (u_longlong_t)bigsize, | |
3467 | (u_longlong_t)txg); | |
3468 | } | |
428870ff | 3469 | VERIFY(0 == dmu_free_range(os, bigobj, bigoff, bigsize, tx)); |
34dc7c2f | 3470 | } else { |
428870ff | 3471 | if (zopt_verbose >= 7) { |
34dc7c2f BB |
3472 | (void) printf("writing offset %llx size %llx" |
3473 | " txg %llx\n", | |
3474 | (u_longlong_t)bigoff, | |
3475 | (u_longlong_t)bigsize, | |
3476 | (u_longlong_t)txg); | |
3477 | } | |
428870ff | 3478 | dmu_write(os, bigobj, bigoff, bigsize, bigbuf, tx); |
34dc7c2f BB |
3479 | } |
3480 | ||
3481 | dmu_tx_commit(tx); | |
3482 | ||
3483 | /* | |
3484 | * Sanity check the stuff we just wrote. | |
3485 | */ | |
3486 | { | |
3487 | void *packcheck = umem_alloc(packsize, UMEM_NOFAIL); | |
3488 | void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL); | |
3489 | ||
428870ff | 3490 | VERIFY(0 == dmu_read(os, packobj, packoff, |
9babb374 | 3491 | packsize, packcheck, DMU_READ_PREFETCH)); |
428870ff | 3492 | VERIFY(0 == dmu_read(os, bigobj, bigoff, |
9babb374 | 3493 | bigsize, bigcheck, DMU_READ_PREFETCH)); |
34dc7c2f BB |
3494 | |
3495 | ASSERT(bcmp(packbuf, packcheck, packsize) == 0); | |
3496 | ASSERT(bcmp(bigbuf, bigcheck, bigsize) == 0); | |
3497 | ||
3498 | umem_free(packcheck, packsize); | |
3499 | umem_free(bigcheck, bigsize); | |
3500 | } | |
3501 | ||
3502 | umem_free(packbuf, packsize); | |
3503 | umem_free(bigbuf, bigsize); | |
40b84e7a | 3504 | umem_free(od, size); |
34dc7c2f BB |
3505 | } |
3506 | ||
9babb374 BB |
3507 | void |
3508 | compare_and_update_pbbufs(uint64_t s, bufwad_t *packbuf, bufwad_t *bigbuf, | |
428870ff | 3509 | uint64_t bigsize, uint64_t n, uint64_t chunksize, uint64_t txg) |
9babb374 BB |
3510 | { |
3511 | uint64_t i; | |
3512 | bufwad_t *pack; | |
3513 | bufwad_t *bigH; | |
3514 | bufwad_t *bigT; | |
3515 | ||
3516 | /* | |
3517 | * For each index from n to n + s, verify that the existing bufwad | |
3518 | * in packobj matches the bufwads at the head and tail of the | |
3519 | * corresponding chunk in bigobj. Then update all three bufwads | |
3520 | * with the new values we want to write out. | |
3521 | */ | |
3522 | for (i = 0; i < s; i++) { | |
3523 | /* LINTED */ | |
3524 | pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t)); | |
3525 | /* LINTED */ | |
428870ff | 3526 | bigH = (bufwad_t *)((char *)bigbuf + i * chunksize); |
9babb374 | 3527 | /* LINTED */ |
428870ff | 3528 | bigT = (bufwad_t *)((char *)bigH + chunksize) - 1; |
9babb374 BB |
3529 | |
3530 | ASSERT((uintptr_t)bigH - (uintptr_t)bigbuf < bigsize); | |
3531 | ASSERT((uintptr_t)bigT - (uintptr_t)bigbuf < bigsize); | |
3532 | ||
3533 | if (pack->bw_txg > txg) | |
3534 | fatal(0, "future leak: got %llx, open txg is %llx", | |
3535 | pack->bw_txg, txg); | |
3536 | ||
3537 | if (pack->bw_data != 0 && pack->bw_index != n + i) | |
3538 | fatal(0, "wrong index: got %llx, wanted %llx+%llx", | |
3539 | pack->bw_index, n, i); | |
3540 | ||
3541 | if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0) | |
3542 | fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH); | |
3543 | ||
3544 | if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0) | |
3545 | fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT); | |
3546 | ||
3547 | pack->bw_index = n + i; | |
3548 | pack->bw_txg = txg; | |
3549 | pack->bw_data = 1 + ztest_random(-2ULL); | |
3550 | ||
3551 | *bigH = *pack; | |
3552 | *bigT = *pack; | |
3553 | } | |
3554 | } | |
3555 | ||
40b84e7a BB |
3556 | #undef OD_ARRAY_SIZE |
3557 | #define OD_ARRAY_SIZE 2 | |
3558 | ||
9babb374 | 3559 | void |
428870ff | 3560 | ztest_dmu_read_write_zcopy(ztest_ds_t *zd, uint64_t id) |
9babb374 | 3561 | { |
428870ff | 3562 | objset_t *os = zd->zd_os; |
40b84e7a | 3563 | ztest_od_t *od; |
9babb374 BB |
3564 | dmu_tx_t *tx; |
3565 | uint64_t i; | |
3566 | int error; | |
40b84e7a | 3567 | int size; |
9babb374 BB |
3568 | uint64_t n, s, txg; |
3569 | bufwad_t *packbuf, *bigbuf; | |
428870ff BB |
3570 | uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize; |
3571 | uint64_t blocksize = ztest_random_blocksize(); | |
3572 | uint64_t chunksize = blocksize; | |
9babb374 BB |
3573 | uint64_t regions = 997; |
3574 | uint64_t stride = 123456789ULL; | |
3575 | uint64_t width = 9; | |
3576 | dmu_buf_t *bonus_db; | |
3577 | arc_buf_t **bigbuf_arcbufs; | |
428870ff | 3578 | dmu_object_info_t doi; |
9babb374 | 3579 | |
40b84e7a BB |
3580 | size = sizeof(ztest_od_t) * OD_ARRAY_SIZE; |
3581 | od = umem_alloc(size, UMEM_NOFAIL); | |
3582 | ||
9babb374 BB |
3583 | /* |
3584 | * This test uses two objects, packobj and bigobj, that are always | |
3585 | * updated together (i.e. in the same tx) so that their contents are | |
3586 | * in sync and can be compared. Their contents relate to each other | |
3587 | * in a simple way: packobj is a dense array of 'bufwad' structures, | |
3588 | * while bigobj is a sparse array of the same bufwads. Specifically, | |
3589 | * for any index n, there are three bufwads that should be identical: | |
3590 | * | |
3591 | * packobj, at offset n * sizeof (bufwad_t) | |
3592 | * bigobj, at the head of the nth chunk | |
3593 | * bigobj, at the tail of the nth chunk | |
3594 | * | |
3595 | * The chunk size is set equal to bigobj block size so that | |
3596 | * dmu_assign_arcbuf() can be tested for object updates. | |
3597 | */ | |
3598 | ||
3599 | /* | |
3600 | * Read the directory info. If it's the first time, set things up. | |
3601 | */ | |
40b84e7a BB |
3602 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0); |
3603 | ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, chunksize); | |
3604 | ||
9babb374 | 3605 | |
40b84e7a BB |
3606 | if (ztest_object_init(zd, od, size, B_FALSE) != 0) { |
3607 | umem_free(od, size); | |
428870ff | 3608 | return; |
40b84e7a | 3609 | } |
9babb374 | 3610 | |
428870ff BB |
3611 | bigobj = od[0].od_object; |
3612 | packobj = od[1].od_object; | |
3613 | blocksize = od[0].od_blocksize; | |
3614 | chunksize = blocksize; | |
3615 | ASSERT(chunksize == od[1].od_gen); | |
9babb374 | 3616 | |
428870ff BB |
3617 | VERIFY(dmu_object_info(os, bigobj, &doi) == 0); |
3618 | VERIFY(ISP2(doi.doi_data_block_size)); | |
3619 | VERIFY(chunksize == doi.doi_data_block_size); | |
3620 | VERIFY(chunksize >= 2 * sizeof (bufwad_t)); | |
9babb374 BB |
3621 | |
3622 | /* | |
3623 | * Pick a random index and compute the offsets into packobj and bigobj. | |
3624 | */ | |
3625 | n = ztest_random(regions) * stride + ztest_random(width); | |
3626 | s = 1 + ztest_random(width - 1); | |
3627 | ||
3628 | packoff = n * sizeof (bufwad_t); | |
3629 | packsize = s * sizeof (bufwad_t); | |
3630 | ||
428870ff BB |
3631 | bigoff = n * chunksize; |
3632 | bigsize = s * chunksize; | |
9babb374 BB |
3633 | |
3634 | packbuf = umem_zalloc(packsize, UMEM_NOFAIL); | |
3635 | bigbuf = umem_zalloc(bigsize, UMEM_NOFAIL); | |
3636 | ||
428870ff | 3637 | VERIFY3U(0, ==, dmu_bonus_hold(os, bigobj, FTAG, &bonus_db)); |
9babb374 BB |
3638 | |
3639 | bigbuf_arcbufs = umem_zalloc(2 * s * sizeof (arc_buf_t *), UMEM_NOFAIL); | |
3640 | ||
3641 | /* | |
3642 | * Iteration 0 test zcopy for DB_UNCACHED dbufs. | |
3643 | * Iteration 1 test zcopy to already referenced dbufs. | |
3644 | * Iteration 2 test zcopy to dirty dbuf in the same txg. | |
3645 | * Iteration 3 test zcopy to dbuf dirty in previous txg. | |
3646 | * Iteration 4 test zcopy when dbuf is no longer dirty. | |
3647 | * Iteration 5 test zcopy when it can't be done. | |
3648 | * Iteration 6 one more zcopy write. | |
3649 | */ | |
3650 | for (i = 0; i < 7; i++) { | |
3651 | uint64_t j; | |
3652 | uint64_t off; | |
3653 | ||
3654 | /* | |
3655 | * In iteration 5 (i == 5) use arcbufs | |
3656 | * that don't match bigobj blksz to test | |
3657 | * dmu_assign_arcbuf() when it can't directly | |
3658 | * assign an arcbuf to a dbuf. | |
3659 | */ | |
3660 | for (j = 0; j < s; j++) { | |
3661 | if (i != 5) { | |
3662 | bigbuf_arcbufs[j] = | |
428870ff | 3663 | dmu_request_arcbuf(bonus_db, chunksize); |
9babb374 BB |
3664 | } else { |
3665 | bigbuf_arcbufs[2 * j] = | |
428870ff | 3666 | dmu_request_arcbuf(bonus_db, chunksize / 2); |
9babb374 | 3667 | bigbuf_arcbufs[2 * j + 1] = |
428870ff | 3668 | dmu_request_arcbuf(bonus_db, chunksize / 2); |
9babb374 BB |
3669 | } |
3670 | } | |
3671 | ||
3672 | /* | |
3673 | * Get a tx for the mods to both packobj and bigobj. | |
3674 | */ | |
3675 | tx = dmu_tx_create(os); | |
3676 | ||
428870ff BB |
3677 | dmu_tx_hold_write(tx, packobj, packoff, packsize); |
3678 | dmu_tx_hold_write(tx, bigobj, bigoff, bigsize); | |
9babb374 | 3679 | |
428870ff BB |
3680 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); |
3681 | if (txg == 0) { | |
9babb374 BB |
3682 | umem_free(packbuf, packsize); |
3683 | umem_free(bigbuf, bigsize); | |
3684 | for (j = 0; j < s; j++) { | |
3685 | if (i != 5) { | |
3686 | dmu_return_arcbuf(bigbuf_arcbufs[j]); | |
3687 | } else { | |
3688 | dmu_return_arcbuf( | |
3689 | bigbuf_arcbufs[2 * j]); | |
3690 | dmu_return_arcbuf( | |
3691 | bigbuf_arcbufs[2 * j + 1]); | |
3692 | } | |
3693 | } | |
3694 | umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *)); | |
40b84e7a | 3695 | umem_free(od, size); |
9babb374 BB |
3696 | dmu_buf_rele(bonus_db, FTAG); |
3697 | return; | |
3698 | } | |
3699 | ||
9babb374 BB |
3700 | /* |
3701 | * 50% of the time don't read objects in the 1st iteration to | |
3702 | * test dmu_assign_arcbuf() for the case when there're no | |
3703 | * existing dbufs for the specified offsets. | |
3704 | */ | |
3705 | if (i != 0 || ztest_random(2) != 0) { | |
428870ff | 3706 | error = dmu_read(os, packobj, packoff, |
9babb374 BB |
3707 | packsize, packbuf, DMU_READ_PREFETCH); |
3708 | ASSERT3U(error, ==, 0); | |
428870ff | 3709 | error = dmu_read(os, bigobj, bigoff, bigsize, |
9babb374 BB |
3710 | bigbuf, DMU_READ_PREFETCH); |
3711 | ASSERT3U(error, ==, 0); | |
3712 | } | |
3713 | compare_and_update_pbbufs(s, packbuf, bigbuf, bigsize, | |
428870ff | 3714 | n, chunksize, txg); |
9babb374 BB |
3715 | |
3716 | /* | |
3717 | * We've verified all the old bufwads, and made new ones. | |
3718 | * Now write them out. | |
3719 | */ | |
428870ff BB |
3720 | dmu_write(os, packobj, packoff, packsize, packbuf, tx); |
3721 | if (zopt_verbose >= 7) { | |
9babb374 BB |
3722 | (void) printf("writing offset %llx size %llx" |
3723 | " txg %llx\n", | |
3724 | (u_longlong_t)bigoff, | |
3725 | (u_longlong_t)bigsize, | |
3726 | (u_longlong_t)txg); | |
3727 | } | |
428870ff | 3728 | for (off = bigoff, j = 0; j < s; j++, off += chunksize) { |
9babb374 BB |
3729 | dmu_buf_t *dbt; |
3730 | if (i != 5) { | |
3731 | bcopy((caddr_t)bigbuf + (off - bigoff), | |
428870ff | 3732 | bigbuf_arcbufs[j]->b_data, chunksize); |
9babb374 BB |
3733 | } else { |
3734 | bcopy((caddr_t)bigbuf + (off - bigoff), | |
3735 | bigbuf_arcbufs[2 * j]->b_data, | |
428870ff | 3736 | chunksize / 2); |
9babb374 | 3737 | bcopy((caddr_t)bigbuf + (off - bigoff) + |
428870ff | 3738 | chunksize / 2, |
9babb374 | 3739 | bigbuf_arcbufs[2 * j + 1]->b_data, |
428870ff | 3740 | chunksize / 2); |
9babb374 BB |
3741 | } |
3742 | ||
3743 | if (i == 1) { | |
428870ff BB |
3744 | VERIFY(dmu_buf_hold(os, bigobj, off, |
3745 | FTAG, &dbt, DMU_READ_NO_PREFETCH) == 0); | |
9babb374 BB |
3746 | } |
3747 | if (i != 5) { | |
3748 | dmu_assign_arcbuf(bonus_db, off, | |
3749 | bigbuf_arcbufs[j], tx); | |
3750 | } else { | |
3751 | dmu_assign_arcbuf(bonus_db, off, | |
3752 | bigbuf_arcbufs[2 * j], tx); | |
3753 | dmu_assign_arcbuf(bonus_db, | |
428870ff | 3754 | off + chunksize / 2, |
9babb374 BB |
3755 | bigbuf_arcbufs[2 * j + 1], tx); |
3756 | } | |
3757 | if (i == 1) { | |
3758 | dmu_buf_rele(dbt, FTAG); | |
3759 | } | |
3760 | } | |
3761 | dmu_tx_commit(tx); | |
3762 | ||
3763 | /* | |
3764 | * Sanity check the stuff we just wrote. | |
3765 | */ | |
3766 | { | |
3767 | void *packcheck = umem_alloc(packsize, UMEM_NOFAIL); | |
3768 | void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL); | |
3769 | ||
428870ff | 3770 | VERIFY(0 == dmu_read(os, packobj, packoff, |
9babb374 | 3771 | packsize, packcheck, DMU_READ_PREFETCH)); |
428870ff | 3772 | VERIFY(0 == dmu_read(os, bigobj, bigoff, |
9babb374 BB |
3773 | bigsize, bigcheck, DMU_READ_PREFETCH)); |
3774 | ||
3775 | ASSERT(bcmp(packbuf, packcheck, packsize) == 0); | |
3776 | ASSERT(bcmp(bigbuf, bigcheck, bigsize) == 0); | |
3777 | ||
3778 | umem_free(packcheck, packsize); | |
3779 | umem_free(bigcheck, bigsize); | |
3780 | } | |
3781 | if (i == 2) { | |
3782 | txg_wait_open(dmu_objset_pool(os), 0); | |
3783 | } else if (i == 3) { | |
3784 | txg_wait_synced(dmu_objset_pool(os), 0); | |
3785 | } | |
3786 | } | |
3787 | ||
3788 | dmu_buf_rele(bonus_db, FTAG); | |
3789 | umem_free(packbuf, packsize); | |
3790 | umem_free(bigbuf, bigsize); | |
3791 | umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *)); | |
40b84e7a | 3792 | umem_free(od, size); |
9babb374 BB |
3793 | } |
3794 | ||
428870ff | 3795 | /* ARGSUSED */ |
34dc7c2f | 3796 | void |
428870ff | 3797 | ztest_dmu_write_parallel(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3798 | { |
40b84e7a BB |
3799 | ztest_od_t *od; |
3800 | ||
3801 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); | |
428870ff BB |
3802 | uint64_t offset = (1ULL << (ztest_random(20) + 43)) + |
3803 | (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
34dc7c2f BB |
3804 | |
3805 | /* | |
428870ff BB |
3806 | * Have multiple threads write to large offsets in an object |
3807 | * to verify that parallel writes to an object -- even to the | |
3808 | * same blocks within the object -- doesn't cause any trouble. | |
34dc7c2f | 3809 | */ |
40b84e7a | 3810 | ztest_od_init(od, ID_PARALLEL, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0); |
9babb374 | 3811 | |
40b84e7a | 3812 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) |
34dc7c2f | 3813 | return; |
34dc7c2f | 3814 | |
428870ff | 3815 | while (ztest_random(10) != 0) |
40b84e7a BB |
3816 | ztest_io(zd, od->od_object, offset); |
3817 | ||
3818 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff | 3819 | } |
34dc7c2f | 3820 | |
428870ff BB |
3821 | void |
3822 | ztest_dmu_prealloc(ztest_ds_t *zd, uint64_t id) | |
3823 | { | |
40b84e7a | 3824 | ztest_od_t *od; |
428870ff BB |
3825 | uint64_t offset = (1ULL << (ztest_random(4) + SPA_MAXBLOCKSHIFT)) + |
3826 | (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
3827 | uint64_t count = ztest_random(20) + 1; | |
3828 | uint64_t blocksize = ztest_random_blocksize(); | |
3829 | void *data; | |
34dc7c2f | 3830 | |
40b84e7a | 3831 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
34dc7c2f | 3832 | |
40b84e7a BB |
3833 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0); |
3834 | ||
3835 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), !ztest_random(2)) != 0) { | |
3836 | umem_free(od, sizeof(ztest_od_t)); | |
34dc7c2f | 3837 | return; |
40b84e7a | 3838 | } |
34dc7c2f | 3839 | |
40b84e7a BB |
3840 | if (ztest_truncate(zd, od->od_object, offset, count * blocksize) != 0) { |
3841 | umem_free(od, sizeof(ztest_od_t)); | |
34dc7c2f | 3842 | return; |
40b84e7a | 3843 | } |
34dc7c2f | 3844 | |
40b84e7a | 3845 | ztest_prealloc(zd, od->od_object, offset, count * blocksize); |
34dc7c2f | 3846 | |
428870ff | 3847 | data = umem_zalloc(blocksize, UMEM_NOFAIL); |
34dc7c2f | 3848 | |
428870ff BB |
3849 | while (ztest_random(count) != 0) { |
3850 | uint64_t randoff = offset + (ztest_random(count) * blocksize); | |
40b84e7a | 3851 | if (ztest_write(zd, od->od_object, randoff, blocksize, |
428870ff BB |
3852 | data) != 0) |
3853 | break; | |
3854 | while (ztest_random(4) != 0) | |
40b84e7a | 3855 | ztest_io(zd, od->od_object, randoff); |
9babb374 | 3856 | } |
34dc7c2f | 3857 | |
428870ff | 3858 | umem_free(data, blocksize); |
40b84e7a | 3859 | umem_free(od, sizeof(ztest_od_t)); |
34dc7c2f BB |
3860 | } |
3861 | ||
3862 | /* | |
3863 | * Verify that zap_{create,destroy,add,remove,update} work as expected. | |
3864 | */ | |
3865 | #define ZTEST_ZAP_MIN_INTS 1 | |
3866 | #define ZTEST_ZAP_MAX_INTS 4 | |
3867 | #define ZTEST_ZAP_MAX_PROPS 1000 | |
3868 | ||
3869 | void | |
428870ff | 3870 | ztest_zap(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3871 | { |
428870ff | 3872 | objset_t *os = zd->zd_os; |
40b84e7a | 3873 | ztest_od_t *od; |
34dc7c2f BB |
3874 | uint64_t object; |
3875 | uint64_t txg, last_txg; | |
3876 | uint64_t value[ZTEST_ZAP_MAX_INTS]; | |
3877 | uint64_t zl_ints, zl_intsize, prop; | |
3878 | int i, ints; | |
3879 | dmu_tx_t *tx; | |
3880 | char propname[100], txgname[100]; | |
3881 | int error; | |
34dc7c2f BB |
3882 | char *hc[2] = { "s.acl.h", ".s.open.h.hyLZlg" }; |
3883 | ||
40b84e7a BB |
3884 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
3885 | ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0); | |
34dc7c2f | 3886 | |
40b84e7a BB |
3887 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
3888 | !ztest_random(2)) != 0) | |
3889 | goto out; | |
34dc7c2f | 3890 | |
40b84e7a | 3891 | object = od->od_object; |
34dc7c2f | 3892 | |
428870ff BB |
3893 | /* |
3894 | * Generate a known hash collision, and verify that | |
3895 | * we can lookup and remove both entries. | |
3896 | */ | |
3897 | tx = dmu_tx_create(os); | |
3898 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); | |
3899 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
3900 | if (txg == 0) | |
40b84e7a | 3901 | goto out; |
428870ff BB |
3902 | for (i = 0; i < 2; i++) { |
3903 | value[i] = i; | |
3904 | VERIFY3U(0, ==, zap_add(os, object, hc[i], sizeof (uint64_t), | |
3905 | 1, &value[i], tx)); | |
3906 | } | |
3907 | for (i = 0; i < 2; i++) { | |
3908 | VERIFY3U(EEXIST, ==, zap_add(os, object, hc[i], | |
3909 | sizeof (uint64_t), 1, &value[i], tx)); | |
3910 | VERIFY3U(0, ==, | |
3911 | zap_length(os, object, hc[i], &zl_intsize, &zl_ints)); | |
3912 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
3913 | ASSERT3U(zl_ints, ==, 1); | |
3914 | } | |
3915 | for (i = 0; i < 2; i++) { | |
3916 | VERIFY3U(0, ==, zap_remove(os, object, hc[i], tx)); | |
34dc7c2f | 3917 | } |
428870ff | 3918 | dmu_tx_commit(tx); |
34dc7c2f | 3919 | |
428870ff BB |
3920 | /* |
3921 | * Generate a buch of random entries. | |
3922 | */ | |
34dc7c2f BB |
3923 | ints = MAX(ZTEST_ZAP_MIN_INTS, object % ZTEST_ZAP_MAX_INTS); |
3924 | ||
3925 | prop = ztest_random(ZTEST_ZAP_MAX_PROPS); | |
3926 | (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop); | |
3927 | (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop); | |
3928 | bzero(value, sizeof (value)); | |
3929 | last_txg = 0; | |
3930 | ||
3931 | /* | |
3932 | * If these zap entries already exist, validate their contents. | |
3933 | */ | |
3934 | error = zap_length(os, object, txgname, &zl_intsize, &zl_ints); | |
3935 | if (error == 0) { | |
3936 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
3937 | ASSERT3U(zl_ints, ==, 1); | |
3938 | ||
3939 | VERIFY(zap_lookup(os, object, txgname, zl_intsize, | |
3940 | zl_ints, &last_txg) == 0); | |
3941 | ||
3942 | VERIFY(zap_length(os, object, propname, &zl_intsize, | |
3943 | &zl_ints) == 0); | |
3944 | ||
3945 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
3946 | ASSERT3U(zl_ints, ==, ints); | |
3947 | ||
3948 | VERIFY(zap_lookup(os, object, propname, zl_intsize, | |
3949 | zl_ints, value) == 0); | |
3950 | ||
3951 | for (i = 0; i < ints; i++) { | |
3952 | ASSERT3U(value[i], ==, last_txg + object + i); | |
3953 | } | |
3954 | } else { | |
3955 | ASSERT3U(error, ==, ENOENT); | |
3956 | } | |
3957 | ||
3958 | /* | |
3959 | * Atomically update two entries in our zap object. | |
3960 | * The first is named txg_%llu, and contains the txg | |
3961 | * in which the property was last updated. The second | |
3962 | * is named prop_%llu, and the nth element of its value | |
3963 | * should be txg + object + n. | |
3964 | */ | |
3965 | tx = dmu_tx_create(os); | |
428870ff BB |
3966 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
3967 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
3968 | if (txg == 0) | |
40b84e7a | 3969 | goto out; |
34dc7c2f BB |
3970 | |
3971 | if (last_txg > txg) | |
3972 | fatal(0, "zap future leak: old %llu new %llu", last_txg, txg); | |
3973 | ||
3974 | for (i = 0; i < ints; i++) | |
3975 | value[i] = txg + object + i; | |
3976 | ||
428870ff BB |
3977 | VERIFY3U(0, ==, zap_update(os, object, txgname, sizeof (uint64_t), |
3978 | 1, &txg, tx)); | |
3979 | VERIFY3U(0, ==, zap_update(os, object, propname, sizeof (uint64_t), | |
3980 | ints, value, tx)); | |
34dc7c2f BB |
3981 | |
3982 | dmu_tx_commit(tx); | |
3983 | ||
3984 | /* | |
3985 | * Remove a random pair of entries. | |
3986 | */ | |
3987 | prop = ztest_random(ZTEST_ZAP_MAX_PROPS); | |
3988 | (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop); | |
3989 | (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop); | |
3990 | ||
3991 | error = zap_length(os, object, txgname, &zl_intsize, &zl_ints); | |
3992 | ||
3993 | if (error == ENOENT) | |
40b84e7a | 3994 | goto out; |
34dc7c2f BB |
3995 | |
3996 | ASSERT3U(error, ==, 0); | |
3997 | ||
3998 | tx = dmu_tx_create(os); | |
428870ff BB |
3999 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
4000 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4001 | if (txg == 0) | |
40b84e7a | 4002 | goto out; |
428870ff BB |
4003 | VERIFY3U(0, ==, zap_remove(os, object, txgname, tx)); |
4004 | VERIFY3U(0, ==, zap_remove(os, object, propname, tx)); | |
4005 | dmu_tx_commit(tx); | |
40b84e7a BB |
4006 | out: |
4007 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff | 4008 | } |
34dc7c2f | 4009 | |
428870ff BB |
4010 | /* |
4011 | * Testcase to test the upgrading of a microzap to fatzap. | |
4012 | */ | |
4013 | void | |
4014 | ztest_fzap(ztest_ds_t *zd, uint64_t id) | |
4015 | { | |
4016 | objset_t *os = zd->zd_os; | |
40b84e7a | 4017 | ztest_od_t *od; |
428870ff | 4018 | uint64_t object, txg; |
d6320ddb | 4019 | int i; |
34dc7c2f | 4020 | |
40b84e7a BB |
4021 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
4022 | ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0); | |
428870ff | 4023 | |
40b84e7a BB |
4024 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
4025 | !ztest_random(2)) != 0) | |
4026 | goto out; | |
4027 | object = od->od_object; | |
34dc7c2f BB |
4028 | |
4029 | /* | |
428870ff BB |
4030 | * Add entries to this ZAP and make sure it spills over |
4031 | * and gets upgraded to a fatzap. Also, since we are adding | |
4032 | * 2050 entries we should see ptrtbl growth and leaf-block split. | |
34dc7c2f | 4033 | */ |
d6320ddb | 4034 | for (i = 0; i < 2050; i++) { |
428870ff BB |
4035 | char name[MAXNAMELEN]; |
4036 | uint64_t value = i; | |
4037 | dmu_tx_t *tx; | |
4038 | int error; | |
34dc7c2f | 4039 | |
428870ff | 4040 | (void) snprintf(name, sizeof (name), "fzap-%llu-%llu", |
b8864a23 | 4041 | (u_longlong_t)id, (u_longlong_t)value); |
428870ff BB |
4042 | |
4043 | tx = dmu_tx_create(os); | |
4044 | dmu_tx_hold_zap(tx, object, B_TRUE, name); | |
4045 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4046 | if (txg == 0) | |
40b84e7a | 4047 | goto out; |
428870ff BB |
4048 | error = zap_add(os, object, name, sizeof (uint64_t), 1, |
4049 | &value, tx); | |
4050 | ASSERT(error == 0 || error == EEXIST); | |
4051 | dmu_tx_commit(tx); | |
34dc7c2f | 4052 | } |
40b84e7a BB |
4053 | out: |
4054 | umem_free(od, sizeof(ztest_od_t)); | |
34dc7c2f BB |
4055 | } |
4056 | ||
428870ff | 4057 | /* ARGSUSED */ |
34dc7c2f | 4058 | void |
428870ff | 4059 | ztest_zap_parallel(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4060 | { |
428870ff | 4061 | objset_t *os = zd->zd_os; |
40b84e7a | 4062 | ztest_od_t *od; |
34dc7c2f BB |
4063 | uint64_t txg, object, count, wsize, wc, zl_wsize, zl_wc; |
4064 | dmu_tx_t *tx; | |
4065 | int i, namelen, error; | |
428870ff | 4066 | int micro = ztest_random(2); |
34dc7c2f BB |
4067 | char name[20], string_value[20]; |
4068 | void *data; | |
4069 | ||
40b84e7a BB |
4070 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
4071 | ztest_od_init(od, ID_PARALLEL, FTAG, micro, DMU_OT_ZAP_OTHER, 0, 0); | |
428870ff | 4072 | |
40b84e7a BB |
4073 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
4074 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff | 4075 | return; |
40b84e7a | 4076 | } |
428870ff | 4077 | |
40b84e7a | 4078 | object = od->od_object; |
428870ff | 4079 | |
34dc7c2f BB |
4080 | /* |
4081 | * Generate a random name of the form 'xxx.....' where each | |
4082 | * x is a random printable character and the dots are dots. | |
4083 | * There are 94 such characters, and the name length goes from | |
4084 | * 6 to 20, so there are 94^3 * 15 = 12,458,760 possible names. | |
4085 | */ | |
4086 | namelen = ztest_random(sizeof (name) - 5) + 5 + 1; | |
4087 | ||
4088 | for (i = 0; i < 3; i++) | |
4089 | name[i] = '!' + ztest_random('~' - '!' + 1); | |
4090 | for (; i < namelen - 1; i++) | |
4091 | name[i] = '.'; | |
4092 | name[i] = '\0'; | |
4093 | ||
428870ff | 4094 | if ((namelen & 1) || micro) { |
34dc7c2f BB |
4095 | wsize = sizeof (txg); |
4096 | wc = 1; | |
4097 | data = &txg; | |
4098 | } else { | |
4099 | wsize = 1; | |
4100 | wc = namelen; | |
4101 | data = string_value; | |
4102 | } | |
4103 | ||
4104 | count = -1ULL; | |
4105 | VERIFY(zap_count(os, object, &count) == 0); | |
4106 | ASSERT(count != -1ULL); | |
4107 | ||
4108 | /* | |
4109 | * Select an operation: length, lookup, add, update, remove. | |
4110 | */ | |
4111 | i = ztest_random(5); | |
4112 | ||
4113 | if (i >= 2) { | |
4114 | tx = dmu_tx_create(os); | |
428870ff BB |
4115 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
4116 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4117 | if (txg == 0) | |
34dc7c2f | 4118 | return; |
34dc7c2f BB |
4119 | bcopy(name, string_value, namelen); |
4120 | } else { | |
4121 | tx = NULL; | |
4122 | txg = 0; | |
4123 | bzero(string_value, namelen); | |
4124 | } | |
4125 | ||
4126 | switch (i) { | |
4127 | ||
4128 | case 0: | |
4129 | error = zap_length(os, object, name, &zl_wsize, &zl_wc); | |
4130 | if (error == 0) { | |
4131 | ASSERT3U(wsize, ==, zl_wsize); | |
4132 | ASSERT3U(wc, ==, zl_wc); | |
4133 | } else { | |
4134 | ASSERT3U(error, ==, ENOENT); | |
4135 | } | |
4136 | break; | |
4137 | ||
4138 | case 1: | |
4139 | error = zap_lookup(os, object, name, wsize, wc, data); | |
4140 | if (error == 0) { | |
4141 | if (data == string_value && | |
4142 | bcmp(name, data, namelen) != 0) | |
4143 | fatal(0, "name '%s' != val '%s' len %d", | |
4144 | name, data, namelen); | |
4145 | } else { | |
4146 | ASSERT3U(error, ==, ENOENT); | |
4147 | } | |
4148 | break; | |
4149 | ||
4150 | case 2: | |
4151 | error = zap_add(os, object, name, wsize, wc, data, tx); | |
4152 | ASSERT(error == 0 || error == EEXIST); | |
4153 | break; | |
4154 | ||
4155 | case 3: | |
4156 | VERIFY(zap_update(os, object, name, wsize, wc, data, tx) == 0); | |
4157 | break; | |
4158 | ||
4159 | case 4: | |
4160 | error = zap_remove(os, object, name, tx); | |
4161 | ASSERT(error == 0 || error == ENOENT); | |
4162 | break; | |
4163 | } | |
4164 | ||
4165 | if (tx != NULL) | |
4166 | dmu_tx_commit(tx); | |
40b84e7a BB |
4167 | |
4168 | umem_free(od, sizeof(ztest_od_t)); | |
34dc7c2f BB |
4169 | } |
4170 | ||
428870ff BB |
4171 | /* |
4172 | * Commit callback data. | |
4173 | */ | |
4174 | typedef struct ztest_cb_data { | |
4175 | list_node_t zcd_node; | |
4176 | uint64_t zcd_txg; | |
4177 | int zcd_expected_err; | |
4178 | boolean_t zcd_added; | |
4179 | boolean_t zcd_called; | |
4180 | spa_t *zcd_spa; | |
4181 | } ztest_cb_data_t; | |
4182 | ||
4183 | /* This is the actual commit callback function */ | |
4184 | static void | |
4185 | ztest_commit_callback(void *arg, int error) | |
4186 | { | |
4187 | ztest_cb_data_t *data = arg; | |
4188 | uint64_t synced_txg; | |
4189 | ||
4190 | VERIFY(data != NULL); | |
4191 | VERIFY3S(data->zcd_expected_err, ==, error); | |
4192 | VERIFY(!data->zcd_called); | |
4193 | ||
4194 | synced_txg = spa_last_synced_txg(data->zcd_spa); | |
4195 | if (data->zcd_txg > synced_txg) | |
4196 | fatal(0, "commit callback of txg %" PRIu64 " called prematurely" | |
4197 | ", last synced txg = %" PRIu64 "\n", data->zcd_txg, | |
4198 | synced_txg); | |
4199 | ||
4200 | data->zcd_called = B_TRUE; | |
4201 | ||
4202 | if (error == ECANCELED) { | |
4203 | ASSERT3U(data->zcd_txg, ==, 0); | |
4204 | ASSERT(!data->zcd_added); | |
4205 | ||
4206 | /* | |
4207 | * The private callback data should be destroyed here, but | |
4208 | * since we are going to check the zcd_called field after | |
4209 | * dmu_tx_abort(), we will destroy it there. | |
4210 | */ | |
4211 | return; | |
4212 | } | |
4213 | ||
090ff092 | 4214 | ASSERT(data->zcd_added); |
428870ff BB |
4215 | ASSERT3U(data->zcd_txg, !=, 0); |
4216 | ||
1e33ac1e | 4217 | (void) mutex_enter(&zcl.zcl_callbacks_lock); |
090ff092 RC |
4218 | |
4219 | /* See if this cb was called more quickly */ | |
4220 | if ((synced_txg - data->zcd_txg) < zc_min_txg_delay) | |
4221 | zc_min_txg_delay = synced_txg - data->zcd_txg; | |
4222 | ||
4223 | /* Remove our callback from the list */ | |
428870ff | 4224 | list_remove(&zcl.zcl_callbacks, data); |
090ff092 | 4225 | |
1e33ac1e | 4226 | (void) mutex_exit(&zcl.zcl_callbacks_lock); |
428870ff | 4227 | |
428870ff BB |
4228 | umem_free(data, sizeof (ztest_cb_data_t)); |
4229 | } | |
4230 | ||
4231 | /* Allocate and initialize callback data structure */ | |
4232 | static ztest_cb_data_t * | |
4233 | ztest_create_cb_data(objset_t *os, uint64_t txg) | |
4234 | { | |
4235 | ztest_cb_data_t *cb_data; | |
4236 | ||
4237 | cb_data = umem_zalloc(sizeof (ztest_cb_data_t), UMEM_NOFAIL); | |
4238 | ||
4239 | cb_data->zcd_txg = txg; | |
4240 | cb_data->zcd_spa = dmu_objset_spa(os); | |
1e33ac1e | 4241 | list_link_init(&cb_data->zcd_node); |
428870ff BB |
4242 | |
4243 | return (cb_data); | |
4244 | } | |
4245 | ||
428870ff BB |
4246 | /* |
4247 | * Commit callback test. | |
4248 | */ | |
34dc7c2f | 4249 | void |
428870ff BB |
4250 | ztest_dmu_commit_callbacks(ztest_ds_t *zd, uint64_t id) |
4251 | { | |
4252 | objset_t *os = zd->zd_os; | |
40b84e7a | 4253 | ztest_od_t *od; |
428870ff BB |
4254 | dmu_tx_t *tx; |
4255 | ztest_cb_data_t *cb_data[3], *tmp_cb; | |
4256 | uint64_t old_txg, txg; | |
090ff092 | 4257 | int i, error = 0; |
428870ff | 4258 | |
40b84e7a BB |
4259 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
4260 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0); | |
428870ff | 4261 | |
40b84e7a BB |
4262 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
4263 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff | 4264 | return; |
40b84e7a | 4265 | } |
428870ff BB |
4266 | |
4267 | tx = dmu_tx_create(os); | |
4268 | ||
4269 | cb_data[0] = ztest_create_cb_data(os, 0); | |
4270 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[0]); | |
4271 | ||
40b84e7a | 4272 | dmu_tx_hold_write(tx, od->od_object, 0, sizeof (uint64_t)); |
428870ff BB |
4273 | |
4274 | /* Every once in a while, abort the transaction on purpose */ | |
4275 | if (ztest_random(100) == 0) | |
4276 | error = -1; | |
4277 | ||
4278 | if (!error) | |
4279 | error = dmu_tx_assign(tx, TXG_NOWAIT); | |
4280 | ||
4281 | txg = error ? 0 : dmu_tx_get_txg(tx); | |
4282 | ||
4283 | cb_data[0]->zcd_txg = txg; | |
4284 | cb_data[1] = ztest_create_cb_data(os, txg); | |
4285 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[1]); | |
4286 | ||
4287 | if (error) { | |
4288 | /* | |
4289 | * It's not a strict requirement to call the registered | |
4290 | * callbacks from inside dmu_tx_abort(), but that's what | |
4291 | * it's supposed to happen in the current implementation | |
4292 | * so we will check for that. | |
4293 | */ | |
4294 | for (i = 0; i < 2; i++) { | |
4295 | cb_data[i]->zcd_expected_err = ECANCELED; | |
4296 | VERIFY(!cb_data[i]->zcd_called); | |
4297 | } | |
4298 | ||
4299 | dmu_tx_abort(tx); | |
4300 | ||
4301 | for (i = 0; i < 2; i++) { | |
4302 | VERIFY(cb_data[i]->zcd_called); | |
4303 | umem_free(cb_data[i], sizeof (ztest_cb_data_t)); | |
4304 | } | |
4305 | ||
40b84e7a | 4306 | umem_free(od, sizeof(ztest_od_t)); |
428870ff BB |
4307 | return; |
4308 | } | |
4309 | ||
4310 | cb_data[2] = ztest_create_cb_data(os, txg); | |
4311 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[2]); | |
4312 | ||
4313 | /* | |
4314 | * Read existing data to make sure there isn't a future leak. | |
4315 | */ | |
40b84e7a | 4316 | VERIFY(0 == dmu_read(os, od->od_object, 0, sizeof (uint64_t), |
428870ff BB |
4317 | &old_txg, DMU_READ_PREFETCH)); |
4318 | ||
4319 | if (old_txg > txg) | |
4320 | fatal(0, "future leak: got %" PRIu64 ", open txg is %" PRIu64, | |
4321 | old_txg, txg); | |
4322 | ||
40b84e7a | 4323 | dmu_write(os, od->od_object, 0, sizeof (uint64_t), &txg, tx); |
428870ff | 4324 | |
1e33ac1e | 4325 | (void) mutex_enter(&zcl.zcl_callbacks_lock); |
428870ff BB |
4326 | |
4327 | /* | |
4328 | * Since commit callbacks don't have any ordering requirement and since | |
4329 | * it is theoretically possible for a commit callback to be called | |
4330 | * after an arbitrary amount of time has elapsed since its txg has been | |
4331 | * synced, it is difficult to reliably determine whether a commit | |
4332 | * callback hasn't been called due to high load or due to a flawed | |
4333 | * implementation. | |
4334 | * | |
4335 | * In practice, we will assume that if after a certain number of txgs a | |
4336 | * commit callback hasn't been called, then most likely there's an | |
4337 | * implementation bug.. | |
4338 | */ | |
4339 | tmp_cb = list_head(&zcl.zcl_callbacks); | |
4340 | if (tmp_cb != NULL && | |
090ff092 | 4341 | tmp_cb->zcd_txg + ZTEST_COMMIT_CB_THRESH < txg) { |
428870ff BB |
4342 | fatal(0, "Commit callback threshold exceeded, oldest txg: %" |
4343 | PRIu64 ", open txg: %" PRIu64 "\n", tmp_cb->zcd_txg, txg); | |
4344 | } | |
4345 | ||
4346 | /* | |
4347 | * Let's find the place to insert our callbacks. | |
4348 | * | |
4349 | * Even though the list is ordered by txg, it is possible for the | |
4350 | * insertion point to not be the end because our txg may already be | |
4351 | * quiescing at this point and other callbacks in the open txg | |
4352 | * (from other objsets) may have sneaked in. | |
4353 | */ | |
4354 | tmp_cb = list_tail(&zcl.zcl_callbacks); | |
4355 | while (tmp_cb != NULL && tmp_cb->zcd_txg > txg) | |
4356 | tmp_cb = list_prev(&zcl.zcl_callbacks, tmp_cb); | |
4357 | ||
4358 | /* Add the 3 callbacks to the list */ | |
4359 | for (i = 0; i < 3; i++) { | |
4360 | if (tmp_cb == NULL) | |
4361 | list_insert_head(&zcl.zcl_callbacks, cb_data[i]); | |
4362 | else | |
4363 | list_insert_after(&zcl.zcl_callbacks, tmp_cb, | |
4364 | cb_data[i]); | |
4365 | ||
4366 | cb_data[i]->zcd_added = B_TRUE; | |
4367 | VERIFY(!cb_data[i]->zcd_called); | |
4368 | ||
4369 | tmp_cb = cb_data[i]; | |
4370 | } | |
4371 | ||
090ff092 RC |
4372 | zc_cb_counter += 3; |
4373 | ||
1e33ac1e | 4374 | (void) mutex_exit(&zcl.zcl_callbacks_lock); |
428870ff BB |
4375 | |
4376 | dmu_tx_commit(tx); | |
40b84e7a BB |
4377 | |
4378 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff BB |
4379 | } |
4380 | ||
4381 | /* ARGSUSED */ | |
4382 | void | |
4383 | ztest_dsl_prop_get_set(ztest_ds_t *zd, uint64_t id) | |
4384 | { | |
4385 | zfs_prop_t proplist[] = { | |
4386 | ZFS_PROP_CHECKSUM, | |
4387 | ZFS_PROP_COMPRESSION, | |
4388 | ZFS_PROP_COPIES, | |
4389 | ZFS_PROP_DEDUP | |
4390 | }; | |
4391 | ztest_shared_t *zs = ztest_shared; | |
d6320ddb | 4392 | int p; |
428870ff | 4393 | |
1e33ac1e | 4394 | (void) rw_enter(&zs->zs_name_lock, RW_READER); |
428870ff | 4395 | |
d6320ddb | 4396 | for (p = 0; p < sizeof (proplist) / sizeof (proplist[0]); p++) |
428870ff BB |
4397 | (void) ztest_dsl_prop_set_uint64(zd->zd_name, proplist[p], |
4398 | ztest_random_dsl_prop(proplist[p]), (int)ztest_random(2)); | |
4399 | ||
1e33ac1e | 4400 | (void) rw_exit(&zs->zs_name_lock); |
428870ff BB |
4401 | } |
4402 | ||
4403 | /* ARGSUSED */ | |
4404 | void | |
4405 | ztest_spa_prop_get_set(ztest_ds_t *zd, uint64_t id) | |
4406 | { | |
4407 | ztest_shared_t *zs = ztest_shared; | |
4408 | nvlist_t *props = NULL; | |
4409 | ||
1e33ac1e | 4410 | (void) rw_enter(&zs->zs_name_lock, RW_READER); |
428870ff BB |
4411 | |
4412 | (void) ztest_spa_prop_set_uint64(zs, ZPOOL_PROP_DEDUPDITTO, | |
4413 | ZIO_DEDUPDITTO_MIN + ztest_random(ZIO_DEDUPDITTO_MIN)); | |
4414 | ||
4415 | VERIFY3U(spa_prop_get(zs->zs_spa, &props), ==, 0); | |
4416 | ||
4417 | if (zopt_verbose >= 6) | |
4418 | dump_nvlist(props, 4); | |
4419 | ||
4420 | nvlist_free(props); | |
4421 | ||
1e33ac1e | 4422 | (void) rw_exit(&zs->zs_name_lock); |
428870ff BB |
4423 | } |
4424 | ||
4425 | /* | |
4426 | * Test snapshot hold/release and deferred destroy. | |
4427 | */ | |
4428 | void | |
4429 | ztest_dmu_snapshot_hold(ztest_ds_t *zd, uint64_t id) | |
34dc7c2f | 4430 | { |
34dc7c2f | 4431 | int error; |
428870ff BB |
4432 | objset_t *os = zd->zd_os; |
4433 | objset_t *origin; | |
4434 | char snapname[100]; | |
4435 | char fullname[100]; | |
4436 | char clonename[100]; | |
4437 | char tag[100]; | |
4438 | char osname[MAXNAMELEN]; | |
34dc7c2f | 4439 | |
1e33ac1e | 4440 | (void) rw_enter(&ztest_shared->zs_name_lock, RW_READER); |
34dc7c2f BB |
4441 | |
4442 | dmu_objset_name(os, osname); | |
4443 | ||
1e33ac1e | 4444 | (void) snprintf(snapname, 100, "sh1_%llu", (u_longlong_t)id); |
428870ff | 4445 | (void) snprintf(fullname, 100, "%s@%s", osname, snapname); |
1e33ac1e BB |
4446 | (void) snprintf(clonename, 100, "%s/ch1_%llu",osname,(u_longlong_t)id); |
4447 | (void) snprintf(tag, 100, "tag_%llu", (u_longlong_t)id); | |
428870ff BB |
4448 | |
4449 | /* | |
4450 | * Clean up from any previous run. | |
4451 | */ | |
4452 | (void) dmu_objset_destroy(clonename, B_FALSE); | |
4453 | (void) dsl_dataset_user_release(osname, snapname, tag, B_FALSE); | |
4454 | (void) dmu_objset_destroy(fullname, B_FALSE); | |
4455 | ||
4456 | /* | |
4457 | * Create snapshot, clone it, mark snap for deferred destroy, | |
4458 | * destroy clone, verify snap was also destroyed. | |
4459 | */ | |
572e2857 BB |
4460 | error = dmu_objset_snapshot(osname, snapname, NULL, NULL, FALSE, |
4461 | FALSE, -1); | |
428870ff BB |
4462 | if (error) { |
4463 | if (error == ENOSPC) { | |
4464 | ztest_record_enospc("dmu_objset_snapshot"); | |
4465 | goto out; | |
34dc7c2f | 4466 | } |
428870ff BB |
4467 | fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error); |
4468 | } | |
34dc7c2f | 4469 | |
428870ff BB |
4470 | error = dmu_objset_hold(fullname, FTAG, &origin); |
4471 | if (error) | |
4472 | fatal(0, "dmu_objset_hold(%s) = %d", fullname, error); | |
34dc7c2f | 4473 | |
428870ff BB |
4474 | error = dmu_objset_clone(clonename, dmu_objset_ds(origin), 0); |
4475 | dmu_objset_rele(origin, FTAG); | |
4476 | if (error) { | |
34dc7c2f | 4477 | if (error == ENOSPC) { |
428870ff BB |
4478 | ztest_record_enospc("dmu_objset_clone"); |
4479 | goto out; | |
34dc7c2f | 4480 | } |
428870ff BB |
4481 | fatal(0, "dmu_objset_clone(%s) = %d", clonename, error); |
4482 | } | |
34dc7c2f | 4483 | |
428870ff BB |
4484 | error = dmu_objset_destroy(fullname, B_TRUE); |
4485 | if (error) { | |
4486 | fatal(0, "dmu_objset_destroy(%s, B_TRUE) = %d", | |
4487 | fullname, error); | |
4488 | } | |
34dc7c2f | 4489 | |
428870ff BB |
4490 | error = dmu_objset_destroy(clonename, B_FALSE); |
4491 | if (error) | |
4492 | fatal(0, "dmu_objset_destroy(%s) = %d", clonename, error); | |
34dc7c2f | 4493 | |
428870ff BB |
4494 | error = dmu_objset_hold(fullname, FTAG, &origin); |
4495 | if (error != ENOENT) | |
4496 | fatal(0, "dmu_objset_hold(%s) = %d", fullname, error); | |
34dc7c2f | 4497 | |
428870ff BB |
4498 | /* |
4499 | * Create snapshot, add temporary hold, verify that we can't | |
4500 | * destroy a held snapshot, mark for deferred destroy, | |
4501 | * release hold, verify snapshot was destroyed. | |
4502 | */ | |
572e2857 BB |
4503 | error = dmu_objset_snapshot(osname, snapname, NULL, NULL, FALSE, |
4504 | FALSE, -1); | |
428870ff BB |
4505 | if (error) { |
4506 | if (error == ENOSPC) { | |
4507 | ztest_record_enospc("dmu_objset_snapshot"); | |
4508 | goto out; | |
34dc7c2f | 4509 | } |
428870ff BB |
4510 | fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error); |
4511 | } | |
4512 | ||
572e2857 BB |
4513 | error = dsl_dataset_user_hold(osname, snapname, tag, B_FALSE, |
4514 | B_TRUE, -1); | |
428870ff BB |
4515 | if (error) |
4516 | fatal(0, "dsl_dataset_user_hold(%s)", fullname, tag); | |
4517 | ||
4518 | error = dmu_objset_destroy(fullname, B_FALSE); | |
4519 | if (error != EBUSY) { | |
4520 | fatal(0, "dmu_objset_destroy(%s, B_FALSE) = %d", | |
4521 | fullname, error); | |
4522 | } | |
4523 | ||
4524 | error = dmu_objset_destroy(fullname, B_TRUE); | |
4525 | if (error) { | |
4526 | fatal(0, "dmu_objset_destroy(%s, B_TRUE) = %d", | |
4527 | fullname, error); | |
34dc7c2f BB |
4528 | } |
4529 | ||
428870ff BB |
4530 | error = dsl_dataset_user_release(osname, snapname, tag, B_FALSE); |
4531 | if (error) | |
4532 | fatal(0, "dsl_dataset_user_release(%s)", fullname, tag); | |
4533 | ||
4534 | VERIFY(dmu_objset_hold(fullname, FTAG, &origin) == ENOENT); | |
4535 | ||
4536 | out: | |
1e33ac1e | 4537 | (void) rw_exit(&ztest_shared->zs_name_lock); |
34dc7c2f BB |
4538 | } |
4539 | ||
34dc7c2f BB |
4540 | /* |
4541 | * Inject random faults into the on-disk data. | |
4542 | */ | |
428870ff | 4543 | /* ARGSUSED */ |
34dc7c2f | 4544 | void |
428870ff | 4545 | ztest_fault_inject(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4546 | { |
428870ff BB |
4547 | ztest_shared_t *zs = ztest_shared; |
4548 | spa_t *spa = zs->zs_spa; | |
34dc7c2f BB |
4549 | int fd; |
4550 | uint64_t offset; | |
428870ff | 4551 | uint64_t leaves; |
c5b3a7bb | 4552 | uint64_t bad = 0x1990c0ffeedecadeull; |
34dc7c2f | 4553 | uint64_t top, leaf; |
40b84e7a BB |
4554 | char *path0; |
4555 | char *pathrand; | |
34dc7c2f | 4556 | size_t fsize; |
34dc7c2f BB |
4557 | int bshift = SPA_MAXBLOCKSHIFT + 2; /* don't scrog all labels */ |
4558 | int iters = 1000; | |
428870ff BB |
4559 | int maxfaults; |
4560 | int mirror_save; | |
b128c09f | 4561 | vdev_t *vd0 = NULL; |
34dc7c2f | 4562 | uint64_t guid0 = 0; |
428870ff BB |
4563 | boolean_t islog = B_FALSE; |
4564 | ||
40b84e7a BB |
4565 | path0 = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
4566 | pathrand = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
4567 | ||
1e33ac1e | 4568 | mutex_enter(&zs->zs_vdev_lock); |
428870ff BB |
4569 | maxfaults = MAXFAULTS(); |
4570 | leaves = MAX(zs->zs_mirrors, 1) * zopt_raidz; | |
4571 | mirror_save = zs->zs_mirrors; | |
1e33ac1e | 4572 | mutex_exit(&zs->zs_vdev_lock); |
34dc7c2f | 4573 | |
b128c09f | 4574 | ASSERT(leaves >= 1); |
34dc7c2f BB |
4575 | |
4576 | /* | |
b128c09f | 4577 | * We need SCL_STATE here because we're going to look at vd0->vdev_tsd. |
34dc7c2f | 4578 | */ |
b128c09f | 4579 | spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); |
34dc7c2f | 4580 | |
b128c09f BB |
4581 | if (ztest_random(2) == 0) { |
4582 | /* | |
428870ff | 4583 | * Inject errors on a normal data device or slog device. |
b128c09f | 4584 | */ |
428870ff BB |
4585 | top = ztest_random_vdev_top(spa, B_TRUE); |
4586 | leaf = ztest_random(leaves) + zs->zs_splits; | |
34dc7c2f | 4587 | |
b128c09f BB |
4588 | /* |
4589 | * Generate paths to the first leaf in this top-level vdev, | |
4590 | * and to the random leaf we selected. We'll induce transient | |
4591 | * write failures and random online/offline activity on leaf 0, | |
4592 | * and we'll write random garbage to the randomly chosen leaf. | |
4593 | */ | |
4594 | (void) snprintf(path0, sizeof (path0), ztest_dev_template, | |
428870ff | 4595 | zopt_dir, zopt_pool, top * leaves + zs->zs_splits); |
b128c09f BB |
4596 | (void) snprintf(pathrand, sizeof (pathrand), ztest_dev_template, |
4597 | zopt_dir, zopt_pool, top * leaves + leaf); | |
34dc7c2f | 4598 | |
b128c09f | 4599 | vd0 = vdev_lookup_by_path(spa->spa_root_vdev, path0); |
428870ff BB |
4600 | if (vd0 != NULL && vd0->vdev_top->vdev_islog) |
4601 | islog = B_TRUE; | |
4602 | ||
b128c09f BB |
4603 | if (vd0 != NULL && maxfaults != 1) { |
4604 | /* | |
4605 | * Make vd0 explicitly claim to be unreadable, | |
4606 | * or unwriteable, or reach behind its back | |
4607 | * and close the underlying fd. We can do this if | |
4608 | * maxfaults == 0 because we'll fail and reexecute, | |
4609 | * and we can do it if maxfaults >= 2 because we'll | |
4610 | * have enough redundancy. If maxfaults == 1, the | |
4611 | * combination of this with injection of random data | |
4612 | * corruption below exceeds the pool's fault tolerance. | |
4613 | */ | |
4614 | vdev_file_t *vf = vd0->vdev_tsd; | |
4615 | ||
4616 | if (vf != NULL && ztest_random(3) == 0) { | |
4617 | (void) close(vf->vf_vnode->v_fd); | |
4618 | vf->vf_vnode->v_fd = -1; | |
4619 | } else if (ztest_random(2) == 0) { | |
4620 | vd0->vdev_cant_read = B_TRUE; | |
4621 | } else { | |
4622 | vd0->vdev_cant_write = B_TRUE; | |
4623 | } | |
4624 | guid0 = vd0->vdev_guid; | |
4625 | } | |
4626 | } else { | |
4627 | /* | |
4628 | * Inject errors on an l2cache device. | |
4629 | */ | |
4630 | spa_aux_vdev_t *sav = &spa->spa_l2cache; | |
34dc7c2f | 4631 | |
b128c09f BB |
4632 | if (sav->sav_count == 0) { |
4633 | spa_config_exit(spa, SCL_STATE, FTAG); | |
40b84e7a | 4634 | goto out; |
b128c09f BB |
4635 | } |
4636 | vd0 = sav->sav_vdevs[ztest_random(sav->sav_count)]; | |
34dc7c2f | 4637 | guid0 = vd0->vdev_guid; |
b128c09f BB |
4638 | (void) strcpy(path0, vd0->vdev_path); |
4639 | (void) strcpy(pathrand, vd0->vdev_path); | |
4640 | ||
4641 | leaf = 0; | |
4642 | leaves = 1; | |
4643 | maxfaults = INT_MAX; /* no limit on cache devices */ | |
34dc7c2f BB |
4644 | } |
4645 | ||
b128c09f BB |
4646 | spa_config_exit(spa, SCL_STATE, FTAG); |
4647 | ||
34dc7c2f | 4648 | /* |
428870ff BB |
4649 | * If we can tolerate two or more faults, or we're dealing |
4650 | * with a slog, randomly online/offline vd0. | |
34dc7c2f | 4651 | */ |
428870ff | 4652 | if ((maxfaults >= 2 || islog) && guid0 != 0) { |
fb5f0bc8 BB |
4653 | if (ztest_random(10) < 6) { |
4654 | int flags = (ztest_random(2) == 0 ? | |
4655 | ZFS_OFFLINE_TEMPORARY : 0); | |
428870ff BB |
4656 | |
4657 | /* | |
4658 | * We have to grab the zs_name_lock as writer to | |
4659 | * prevent a race between offlining a slog and | |
4660 | * destroying a dataset. Offlining the slog will | |
4661 | * grab a reference on the dataset which may cause | |
4662 | * dmu_objset_destroy() to fail with EBUSY thus | |
4663 | * leaving the dataset in an inconsistent state. | |
4664 | */ | |
4665 | if (islog) | |
1e33ac1e BB |
4666 | (void) rw_enter(&ztest_shared->zs_name_lock, |
4667 | RW_WRITER); | |
428870ff | 4668 | |
fb5f0bc8 | 4669 | VERIFY(vdev_offline(spa, guid0, flags) != EBUSY); |
428870ff BB |
4670 | |
4671 | if (islog) | |
1e33ac1e | 4672 | (void) rw_exit(&ztest_shared->zs_name_lock); |
fb5f0bc8 BB |
4673 | } else { |
4674 | (void) vdev_online(spa, guid0, 0, NULL); | |
4675 | } | |
34dc7c2f BB |
4676 | } |
4677 | ||
428870ff | 4678 | if (maxfaults == 0) |
40b84e7a | 4679 | goto out; |
428870ff | 4680 | |
34dc7c2f BB |
4681 | /* |
4682 | * We have at least single-fault tolerance, so inject data corruption. | |
4683 | */ | |
4684 | fd = open(pathrand, O_RDWR); | |
4685 | ||
4686 | if (fd == -1) /* we hit a gap in the device namespace */ | |
40b84e7a | 4687 | goto out; |
34dc7c2f BB |
4688 | |
4689 | fsize = lseek(fd, 0, SEEK_END); | |
4690 | ||
4691 | while (--iters != 0) { | |
4692 | offset = ztest_random(fsize / (leaves << bshift)) * | |
4693 | (leaves << bshift) + (leaf << bshift) + | |
4694 | (ztest_random(1ULL << (bshift - 1)) & -8ULL); | |
4695 | ||
4696 | if (offset >= fsize) | |
4697 | continue; | |
4698 | ||
1e33ac1e | 4699 | mutex_enter(&zs->zs_vdev_lock); |
428870ff | 4700 | if (mirror_save != zs->zs_mirrors) { |
1e33ac1e | 4701 | mutex_exit(&zs->zs_vdev_lock); |
428870ff | 4702 | (void) close(fd); |
40b84e7a | 4703 | goto out; |
428870ff | 4704 | } |
34dc7c2f BB |
4705 | |
4706 | if (pwrite(fd, &bad, sizeof (bad), offset) != sizeof (bad)) | |
4707 | fatal(1, "can't inject bad word at 0x%llx in %s", | |
4708 | offset, pathrand); | |
428870ff | 4709 | |
1e33ac1e | 4710 | mutex_exit(&zs->zs_vdev_lock); |
428870ff BB |
4711 | |
4712 | if (zopt_verbose >= 7) | |
4713 | (void) printf("injected bad word into %s," | |
4714 | " offset 0x%llx\n", pathrand, (u_longlong_t)offset); | |
34dc7c2f BB |
4715 | } |
4716 | ||
4717 | (void) close(fd); | |
40b84e7a BB |
4718 | out: |
4719 | umem_free(path0, MAXPATHLEN); | |
4720 | umem_free(pathrand, MAXPATHLEN); | |
34dc7c2f BB |
4721 | } |
4722 | ||
4723 | /* | |
428870ff | 4724 | * Verify that DDT repair works as expected. |
34dc7c2f BB |
4725 | */ |
4726 | void | |
428870ff | 4727 | ztest_ddt_repair(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4728 | { |
428870ff BB |
4729 | ztest_shared_t *zs = ztest_shared; |
4730 | spa_t *spa = zs->zs_spa; | |
4731 | objset_t *os = zd->zd_os; | |
40b84e7a | 4732 | ztest_od_t *od; |
428870ff BB |
4733 | uint64_t object, blocksize, txg, pattern, psize; |
4734 | enum zio_checksum checksum = spa_dedup_checksum(spa); | |
4735 | dmu_buf_t *db; | |
4736 | dmu_tx_t *tx; | |
4737 | void *buf; | |
4738 | blkptr_t blk; | |
4739 | int copies = 2 * ZIO_DEDUPDITTO_MIN; | |
d6320ddb | 4740 | int i; |
34dc7c2f | 4741 | |
428870ff BB |
4742 | blocksize = ztest_random_blocksize(); |
4743 | blocksize = MIN(blocksize, 2048); /* because we write so many */ | |
34dc7c2f | 4744 | |
40b84e7a BB |
4745 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
4746 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0); | |
34dc7c2f | 4747 | |
40b84e7a BB |
4748 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
4749 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff | 4750 | return; |
40b84e7a | 4751 | } |
34dc7c2f BB |
4752 | |
4753 | /* | |
428870ff BB |
4754 | * Take the name lock as writer to prevent anyone else from changing |
4755 | * the pool and dataset properies we need to maintain during this test. | |
34dc7c2f | 4756 | */ |
1e33ac1e | 4757 | (void) rw_enter(&zs->zs_name_lock, RW_WRITER); |
34dc7c2f | 4758 | |
428870ff BB |
4759 | if (ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_DEDUP, checksum, |
4760 | B_FALSE) != 0 || | |
4761 | ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_COPIES, 1, | |
4762 | B_FALSE) != 0) { | |
1e33ac1e | 4763 | (void) rw_exit(&zs->zs_name_lock); |
40b84e7a | 4764 | umem_free(od, sizeof(ztest_od_t)); |
428870ff BB |
4765 | return; |
4766 | } | |
4767 | ||
4768 | object = od[0].od_object; | |
4769 | blocksize = od[0].od_blocksize; | |
4770 | pattern = spa_guid(spa) ^ dmu_objset_fsid_guid(os); | |
4771 | ||
4772 | ASSERT(object != 0); | |
4773 | ||
4774 | tx = dmu_tx_create(os); | |
4775 | dmu_tx_hold_write(tx, object, 0, copies * blocksize); | |
4776 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
4777 | if (txg == 0) { | |
1e33ac1e | 4778 | (void) rw_exit(&zs->zs_name_lock); |
40b84e7a | 4779 | umem_free(od, sizeof(ztest_od_t)); |
428870ff BB |
4780 | return; |
4781 | } | |
34dc7c2f BB |
4782 | |
4783 | /* | |
428870ff | 4784 | * Write all the copies of our block. |
34dc7c2f | 4785 | */ |
d6320ddb | 4786 | for (i = 0; i < copies; i++) { |
428870ff BB |
4787 | uint64_t offset = i * blocksize; |
4788 | VERIFY(dmu_buf_hold(os, object, offset, FTAG, &db, | |
4789 | DMU_READ_NO_PREFETCH) == 0); | |
4790 | ASSERT(db->db_offset == offset); | |
4791 | ASSERT(db->db_size == blocksize); | |
4792 | ASSERT(ztest_pattern_match(db->db_data, db->db_size, pattern) || | |
4793 | ztest_pattern_match(db->db_data, db->db_size, 0ULL)); | |
4794 | dmu_buf_will_fill(db, tx); | |
4795 | ztest_pattern_set(db->db_data, db->db_size, pattern); | |
4796 | dmu_buf_rele(db, FTAG); | |
4797 | } | |
34dc7c2f | 4798 | |
428870ff BB |
4799 | dmu_tx_commit(tx); |
4800 | txg_wait_synced(spa_get_dsl(spa), txg); | |
34dc7c2f BB |
4801 | |
4802 | /* | |
428870ff | 4803 | * Find out what block we got. |
34dc7c2f | 4804 | */ |
428870ff BB |
4805 | VERIFY(dmu_buf_hold(os, object, 0, FTAG, &db, |
4806 | DMU_READ_NO_PREFETCH) == 0); | |
4807 | blk = *((dmu_buf_impl_t *)db)->db_blkptr; | |
4808 | dmu_buf_rele(db, FTAG); | |
34dc7c2f BB |
4809 | |
4810 | /* | |
428870ff | 4811 | * Damage the block. Dedup-ditto will save us when we read it later. |
34dc7c2f | 4812 | */ |
428870ff BB |
4813 | psize = BP_GET_PSIZE(&blk); |
4814 | buf = zio_buf_alloc(psize); | |
4815 | ztest_pattern_set(buf, psize, ~pattern); | |
34dc7c2f | 4816 | |
428870ff BB |
4817 | (void) zio_wait(zio_rewrite(NULL, spa, 0, &blk, |
4818 | buf, psize, NULL, NULL, ZIO_PRIORITY_SYNC_WRITE, | |
4819 | ZIO_FLAG_CANFAIL | ZIO_FLAG_INDUCE_DAMAGE, NULL)); | |
34dc7c2f | 4820 | |
428870ff | 4821 | zio_buf_free(buf, psize); |
34dc7c2f | 4822 | |
1e33ac1e | 4823 | (void) rw_exit(&zs->zs_name_lock); |
40b84e7a | 4824 | umem_free(od, sizeof(ztest_od_t)); |
34dc7c2f BB |
4825 | } |
4826 | ||
34dc7c2f | 4827 | /* |
428870ff | 4828 | * Scrub the pool. |
34dc7c2f | 4829 | */ |
428870ff BB |
4830 | /* ARGSUSED */ |
4831 | void | |
4832 | ztest_scrub(ztest_ds_t *zd, uint64_t id) | |
34dc7c2f | 4833 | { |
428870ff BB |
4834 | ztest_shared_t *zs = ztest_shared; |
4835 | spa_t *spa = zs->zs_spa; | |
34dc7c2f | 4836 | |
428870ff BB |
4837 | (void) spa_scan(spa, POOL_SCAN_SCRUB); |
4838 | (void) poll(NULL, 0, 100); /* wait a moment, then force a restart */ | |
4839 | (void) spa_scan(spa, POOL_SCAN_SCRUB); | |
4840 | } | |
34dc7c2f | 4841 | |
428870ff BB |
4842 | /* |
4843 | * Rename the pool to a different name and then rename it back. | |
4844 | */ | |
4845 | /* ARGSUSED */ | |
4846 | void | |
4847 | ztest_spa_rename(ztest_ds_t *zd, uint64_t id) | |
4848 | { | |
4849 | ztest_shared_t *zs = ztest_shared; | |
4850 | char *oldname, *newname; | |
4851 | spa_t *spa; | |
34dc7c2f | 4852 | |
1e33ac1e | 4853 | (void) rw_enter(&zs->zs_name_lock, RW_WRITER); |
34dc7c2f | 4854 | |
428870ff BB |
4855 | oldname = zs->zs_pool; |
4856 | newname = umem_alloc(strlen(oldname) + 5, UMEM_NOFAIL); | |
4857 | (void) strcpy(newname, oldname); | |
4858 | (void) strcat(newname, "_tmp"); | |
34dc7c2f BB |
4859 | |
4860 | /* | |
428870ff | 4861 | * Do the rename |
34dc7c2f | 4862 | */ |
428870ff | 4863 | VERIFY3U(0, ==, spa_rename(oldname, newname)); |
34dc7c2f BB |
4864 | |
4865 | /* | |
428870ff | 4866 | * Try to open it under the old name, which shouldn't exist |
34dc7c2f | 4867 | */ |
428870ff | 4868 | VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f BB |
4869 | |
4870 | /* | |
428870ff BB |
4871 | * Open it under the new name and make sure it's still the same spa_t. |
4872 | */ | |
4873 | VERIFY3U(0, ==, spa_open(newname, &spa, FTAG)); | |
34dc7c2f | 4874 | |
428870ff BB |
4875 | ASSERT(spa == zs->zs_spa); |
4876 | spa_close(spa, FTAG); | |
34dc7c2f BB |
4877 | |
4878 | /* | |
428870ff | 4879 | * Rename it back to the original |
34dc7c2f | 4880 | */ |
428870ff | 4881 | VERIFY3U(0, ==, spa_rename(newname, oldname)); |
34dc7c2f | 4882 | |
428870ff BB |
4883 | /* |
4884 | * Make sure it can still be opened | |
4885 | */ | |
4886 | VERIFY3U(0, ==, spa_open(oldname, &spa, FTAG)); | |
34dc7c2f | 4887 | |
428870ff BB |
4888 | ASSERT(spa == zs->zs_spa); |
4889 | spa_close(spa, FTAG); | |
4890 | ||
4891 | umem_free(newname, strlen(newname) + 1); | |
4892 | ||
1e33ac1e | 4893 | (void) rw_exit(&zs->zs_name_lock); |
34dc7c2f BB |
4894 | } |
4895 | ||
428870ff BB |
4896 | /* |
4897 | * Verify pool integrity by running zdb. | |
4898 | */ | |
34dc7c2f | 4899 | static void |
428870ff | 4900 | ztest_run_zdb(char *pool) |
34dc7c2f BB |
4901 | { |
4902 | int status; | |
4903 | char zdb[MAXPATHLEN + MAXNAMELEN + 20]; | |
4904 | char zbuf[1024]; | |
4905 | char *bin; | |
4906 | char *ztest; | |
4907 | char *isa; | |
4908 | int isalen; | |
4909 | FILE *fp; | |
4910 | ||
4911 | (void) realpath(getexecname(), zdb); | |
4912 | ||
4913 | /* zdb lives in /usr/sbin, while ztest lives in /usr/bin */ | |
4914 | bin = strstr(zdb, "/usr/bin/"); | |
4915 | ztest = strstr(bin, "/ztest"); | |
4916 | isa = bin + 8; | |
4917 | isalen = ztest - isa; | |
4918 | isa = strdup(isa); | |
4919 | /* LINTED */ | |
4920 | (void) sprintf(bin, | |
428870ff | 4921 | "/usr/sbin%.*s/zdb -bcc%s%s -U %s %s", |
34dc7c2f BB |
4922 | isalen, |
4923 | isa, | |
4924 | zopt_verbose >= 3 ? "s" : "", | |
4925 | zopt_verbose >= 4 ? "v" : "", | |
428870ff | 4926 | spa_config_path, |
b128c09f | 4927 | pool); |
34dc7c2f BB |
4928 | free(isa); |
4929 | ||
4930 | if (zopt_verbose >= 5) | |
4931 | (void) printf("Executing %s\n", strstr(zdb, "zdb ")); | |
4932 | ||
4933 | fp = popen(zdb, "r"); | |
4934 | ||
4935 | while (fgets(zbuf, sizeof (zbuf), fp) != NULL) | |
4936 | if (zopt_verbose >= 3) | |
4937 | (void) printf("%s", zbuf); | |
4938 | ||
4939 | status = pclose(fp); | |
4940 | ||
4941 | if (status == 0) | |
4942 | return; | |
4943 | ||
4944 | ztest_dump_core = 0; | |
4945 | if (WIFEXITED(status)) | |
4946 | fatal(0, "'%s' exit code %d", zdb, WEXITSTATUS(status)); | |
4947 | else | |
4948 | fatal(0, "'%s' died with signal %d", zdb, WTERMSIG(status)); | |
4949 | } | |
4950 | ||
4951 | static void | |
4952 | ztest_walk_pool_directory(char *header) | |
4953 | { | |
4954 | spa_t *spa = NULL; | |
4955 | ||
4956 | if (zopt_verbose >= 6) | |
4957 | (void) printf("%s\n", header); | |
4958 | ||
4959 | mutex_enter(&spa_namespace_lock); | |
4960 | while ((spa = spa_next(spa)) != NULL) | |
4961 | if (zopt_verbose >= 6) | |
4962 | (void) printf("\t%s\n", spa_name(spa)); | |
4963 | mutex_exit(&spa_namespace_lock); | |
4964 | } | |
4965 | ||
4966 | static void | |
4967 | ztest_spa_import_export(char *oldname, char *newname) | |
4968 | { | |
fb5f0bc8 | 4969 | nvlist_t *config, *newconfig; |
34dc7c2f BB |
4970 | uint64_t pool_guid; |
4971 | spa_t *spa; | |
34dc7c2f BB |
4972 | |
4973 | if (zopt_verbose >= 4) { | |
4974 | (void) printf("import/export: old = %s, new = %s\n", | |
4975 | oldname, newname); | |
4976 | } | |
4977 | ||
4978 | /* | |
4979 | * Clean up from previous runs. | |
4980 | */ | |
4981 | (void) spa_destroy(newname); | |
4982 | ||
4983 | /* | |
4984 | * Get the pool's configuration and guid. | |
4985 | */ | |
428870ff | 4986 | VERIFY3U(0, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f | 4987 | |
fb5f0bc8 BB |
4988 | /* |
4989 | * Kick off a scrub to tickle scrub/export races. | |
4990 | */ | |
4991 | if (ztest_random(2) == 0) | |
428870ff | 4992 | (void) spa_scan(spa, POOL_SCAN_SCRUB); |
fb5f0bc8 | 4993 | |
34dc7c2f BB |
4994 | pool_guid = spa_guid(spa); |
4995 | spa_close(spa, FTAG); | |
4996 | ||
4997 | ztest_walk_pool_directory("pools before export"); | |
4998 | ||
4999 | /* | |
5000 | * Export it. | |
5001 | */ | |
428870ff | 5002 | VERIFY3U(0, ==, spa_export(oldname, &config, B_FALSE, B_FALSE)); |
34dc7c2f BB |
5003 | |
5004 | ztest_walk_pool_directory("pools after export"); | |
5005 | ||
fb5f0bc8 BB |
5006 | /* |
5007 | * Try to import it. | |
5008 | */ | |
5009 | newconfig = spa_tryimport(config); | |
5010 | ASSERT(newconfig != NULL); | |
5011 | nvlist_free(newconfig); | |
5012 | ||
34dc7c2f BB |
5013 | /* |
5014 | * Import it under the new name. | |
5015 | */ | |
572e2857 | 5016 | VERIFY3U(0, ==, spa_import(newname, config, NULL, 0)); |
34dc7c2f BB |
5017 | |
5018 | ztest_walk_pool_directory("pools after import"); | |
5019 | ||
5020 | /* | |
5021 | * Try to import it again -- should fail with EEXIST. | |
5022 | */ | |
572e2857 | 5023 | VERIFY3U(EEXIST, ==, spa_import(newname, config, NULL, 0)); |
34dc7c2f BB |
5024 | |
5025 | /* | |
5026 | * Try to import it under a different name -- should fail with EEXIST. | |
5027 | */ | |
572e2857 | 5028 | VERIFY3U(EEXIST, ==, spa_import(oldname, config, NULL, 0)); |
34dc7c2f BB |
5029 | |
5030 | /* | |
5031 | * Verify that the pool is no longer visible under the old name. | |
5032 | */ | |
428870ff | 5033 | VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f BB |
5034 | |
5035 | /* | |
5036 | * Verify that we can open and close the pool using the new name. | |
5037 | */ | |
428870ff | 5038 | VERIFY3U(0, ==, spa_open(newname, &spa, FTAG)); |
34dc7c2f BB |
5039 | ASSERT(pool_guid == spa_guid(spa)); |
5040 | spa_close(spa, FTAG); | |
5041 | ||
5042 | nvlist_free(config); | |
5043 | } | |
5044 | ||
fb5f0bc8 BB |
5045 | static void |
5046 | ztest_resume(spa_t *spa) | |
5047 | { | |
428870ff BB |
5048 | if (spa_suspended(spa) && zopt_verbose >= 6) |
5049 | (void) printf("resuming from suspended state\n"); | |
5050 | spa_vdev_state_enter(spa, SCL_NONE); | |
5051 | vdev_clear(spa, NULL); | |
5052 | (void) spa_vdev_state_exit(spa, NULL, 0); | |
5053 | (void) zio_resume(spa); | |
fb5f0bc8 BB |
5054 | } |
5055 | ||
34dc7c2f | 5056 | static void * |
fb5f0bc8 | 5057 | ztest_resume_thread(void *arg) |
34dc7c2f | 5058 | { |
b128c09f | 5059 | spa_t *spa = arg; |
34dc7c2f BB |
5060 | |
5061 | while (!ztest_exiting) { | |
428870ff BB |
5062 | if (spa_suspended(spa)) |
5063 | ztest_resume(spa); | |
5064 | (void) poll(NULL, 0, 100); | |
34dc7c2f | 5065 | } |
34dc7c2f | 5066 | |
1e33ac1e | 5067 | thread_exit(); |
428870ff | 5068 | |
1e33ac1e BB |
5069 | return (NULL); |
5070 | } | |
428870ff | 5071 | |
1e33ac1e | 5072 | #define GRACE 300 |
428870ff | 5073 | |
1e33ac1e BB |
5074 | static void |
5075 | ztest_deadman_alarm(int sig) | |
5076 | { | |
5077 | fatal(0, "failed to complete within %d seconds of deadline", GRACE); | |
428870ff BB |
5078 | } |
5079 | ||
5080 | static void | |
5081 | ztest_execute(ztest_info_t *zi, uint64_t id) | |
5082 | { | |
5083 | ztest_shared_t *zs = ztest_shared; | |
5084 | ztest_ds_t *zd = &zs->zs_zd[id % zopt_datasets]; | |
5085 | hrtime_t functime = gethrtime(); | |
d6320ddb | 5086 | int i; |
428870ff | 5087 | |
d6320ddb | 5088 | for (i = 0; i < zi->zi_iters; i++) |
428870ff BB |
5089 | zi->zi_func(zd, id); |
5090 | ||
5091 | functime = gethrtime() - functime; | |
5092 | ||
5093 | atomic_add_64(&zi->zi_call_count, 1); | |
5094 | atomic_add_64(&zi->zi_call_time, functime); | |
5095 | ||
5096 | if (zopt_verbose >= 4) { | |
5097 | Dl_info dli; | |
5098 | (void) dladdr((void *)zi->zi_func, &dli); | |
5099 | (void) printf("%6.2f sec in %s\n", | |
5100 | (double)functime / NANOSEC, dli.dli_sname); | |
5101 | } | |
5102 | } | |
5103 | ||
34dc7c2f BB |
5104 | static void * |
5105 | ztest_thread(void *arg) | |
5106 | { | |
428870ff | 5107 | uint64_t id = (uintptr_t)arg; |
34dc7c2f | 5108 | ztest_shared_t *zs = ztest_shared; |
428870ff BB |
5109 | uint64_t call_next; |
5110 | hrtime_t now; | |
34dc7c2f | 5111 | ztest_info_t *zi; |
34dc7c2f | 5112 | |
428870ff | 5113 | while ((now = gethrtime()) < zs->zs_thread_stop) { |
34dc7c2f BB |
5114 | /* |
5115 | * See if it's time to force a crash. | |
5116 | */ | |
428870ff BB |
5117 | if (now > zs->zs_thread_kill) |
5118 | ztest_kill(zs); | |
34dc7c2f BB |
5119 | |
5120 | /* | |
428870ff | 5121 | * If we're getting ENOSPC with some regularity, stop. |
34dc7c2f | 5122 | */ |
428870ff BB |
5123 | if (zs->zs_enospc_count > 10) |
5124 | break; | |
34dc7c2f BB |
5125 | |
5126 | /* | |
428870ff | 5127 | * Pick a random function to execute. |
34dc7c2f | 5128 | */ |
428870ff BB |
5129 | zi = &zs->zs_info[ztest_random(ZTEST_FUNCS)]; |
5130 | call_next = zi->zi_call_next; | |
34dc7c2f | 5131 | |
428870ff BB |
5132 | if (now >= call_next && |
5133 | atomic_cas_64(&zi->zi_call_next, call_next, call_next + | |
5134 | ztest_random(2 * zi->zi_interval[0] + 1)) == call_next) | |
5135 | ztest_execute(zi, id); | |
5136 | } | |
34dc7c2f | 5137 | |
1e33ac1e BB |
5138 | thread_exit(); |
5139 | ||
428870ff BB |
5140 | return (NULL); |
5141 | } | |
34dc7c2f | 5142 | |
428870ff BB |
5143 | static void |
5144 | ztest_dataset_name(char *dsname, char *pool, int d) | |
5145 | { | |
5146 | (void) snprintf(dsname, MAXNAMELEN, "%s/ds_%d", pool, d); | |
5147 | } | |
34dc7c2f | 5148 | |
428870ff BB |
5149 | static void |
5150 | ztest_dataset_destroy(ztest_shared_t *zs, int d) | |
5151 | { | |
5152 | char name[MAXNAMELEN]; | |
d6320ddb | 5153 | int t; |
34dc7c2f | 5154 | |
428870ff | 5155 | ztest_dataset_name(name, zs->zs_pool, d); |
34dc7c2f | 5156 | |
428870ff BB |
5157 | if (zopt_verbose >= 3) |
5158 | (void) printf("Destroying %s to free up space\n", name); | |
34dc7c2f | 5159 | |
428870ff BB |
5160 | /* |
5161 | * Cleanup any non-standard clones and snapshots. In general, | |
5162 | * ztest thread t operates on dataset (t % zopt_datasets), | |
5163 | * so there may be more than one thing to clean up. | |
5164 | */ | |
d6320ddb | 5165 | for (t = d; t < zopt_threads; t += zopt_datasets) |
428870ff BB |
5166 | ztest_dsl_dataset_cleanup(name, t); |
5167 | ||
5168 | (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL, | |
5169 | DS_FIND_SNAPSHOTS | DS_FIND_CHILDREN); | |
5170 | } | |
5171 | ||
5172 | static void | |
5173 | ztest_dataset_dirobj_verify(ztest_ds_t *zd) | |
5174 | { | |
5175 | uint64_t usedobjs, dirobjs, scratch; | |
5176 | ||
5177 | /* | |
5178 | * ZTEST_DIROBJ is the object directory for the entire dataset. | |
5179 | * Therefore, the number of objects in use should equal the | |
5180 | * number of ZTEST_DIROBJ entries, +1 for ZTEST_DIROBJ itself. | |
5181 | * If not, we have an object leak. | |
5182 | * | |
5183 | * Note that we can only check this in ztest_dataset_open(), | |
5184 | * when the open-context and syncing-context values agree. | |
5185 | * That's because zap_count() returns the open-context value, | |
5186 | * while dmu_objset_space() returns the rootbp fill count. | |
5187 | */ | |
5188 | VERIFY3U(0, ==, zap_count(zd->zd_os, ZTEST_DIROBJ, &dirobjs)); | |
5189 | dmu_objset_space(zd->zd_os, &scratch, &scratch, &usedobjs, &scratch); | |
5190 | ASSERT3U(dirobjs + 1, ==, usedobjs); | |
5191 | } | |
5192 | ||
5193 | static int | |
5194 | ztest_dataset_open(ztest_shared_t *zs, int d) | |
5195 | { | |
5196 | ztest_ds_t *zd = &zs->zs_zd[d]; | |
5197 | uint64_t committed_seq = zd->zd_seq; | |
5198 | objset_t *os; | |
5199 | zilog_t *zilog; | |
5200 | char name[MAXNAMELEN]; | |
5201 | int error; | |
5202 | ||
5203 | ztest_dataset_name(name, zs->zs_pool, d); | |
5204 | ||
1e33ac1e | 5205 | (void) rw_enter(&zs->zs_name_lock, RW_READER); |
428870ff BB |
5206 | |
5207 | error = ztest_dataset_create(name); | |
5208 | if (error == ENOSPC) { | |
1e33ac1e | 5209 | (void) rw_exit(&zs->zs_name_lock); |
428870ff BB |
5210 | ztest_record_enospc(FTAG); |
5211 | return (error); | |
34dc7c2f | 5212 | } |
428870ff | 5213 | ASSERT(error == 0 || error == EEXIST); |
34dc7c2f | 5214 | |
428870ff | 5215 | VERIFY3U(dmu_objset_hold(name, zd, &os), ==, 0); |
1e33ac1e | 5216 | (void) rw_exit(&zs->zs_name_lock); |
428870ff BB |
5217 | |
5218 | ztest_zd_init(zd, os); | |
5219 | ||
5220 | zilog = zd->zd_zilog; | |
5221 | ||
5222 | if (zilog->zl_header->zh_claim_lr_seq != 0 && | |
5223 | zilog->zl_header->zh_claim_lr_seq < committed_seq) | |
5224 | fatal(0, "missing log records: claimed %llu < committed %llu", | |
5225 | zilog->zl_header->zh_claim_lr_seq, committed_seq); | |
5226 | ||
5227 | ztest_dataset_dirobj_verify(zd); | |
5228 | ||
5229 | zil_replay(os, zd, ztest_replay_vector); | |
5230 | ||
5231 | ztest_dataset_dirobj_verify(zd); | |
5232 | ||
5233 | if (zopt_verbose >= 6) | |
5234 | (void) printf("%s replay %llu blocks, %llu records, seq %llu\n", | |
5235 | zd->zd_name, | |
5236 | (u_longlong_t)zilog->zl_parse_blk_count, | |
5237 | (u_longlong_t)zilog->zl_parse_lr_count, | |
5238 | (u_longlong_t)zilog->zl_replaying_seq); | |
5239 | ||
5240 | zilog = zil_open(os, ztest_get_data); | |
5241 | ||
5242 | if (zilog->zl_replaying_seq != 0 && | |
5243 | zilog->zl_replaying_seq < committed_seq) | |
5244 | fatal(0, "missing log records: replayed %llu < committed %llu", | |
5245 | zilog->zl_replaying_seq, committed_seq); | |
5246 | ||
5247 | return (0); | |
5248 | } | |
5249 | ||
5250 | static void | |
5251 | ztest_dataset_close(ztest_shared_t *zs, int d) | |
5252 | { | |
5253 | ztest_ds_t *zd = &zs->zs_zd[d]; | |
5254 | ||
5255 | zil_close(zd->zd_zilog); | |
5256 | dmu_objset_rele(zd->zd_os, zd); | |
5257 | ||
5258 | ztest_zd_fini(zd); | |
34dc7c2f BB |
5259 | } |
5260 | ||
5261 | /* | |
5262 | * Kick off threads to run tests on all datasets in parallel. | |
5263 | */ | |
5264 | static void | |
428870ff | 5265 | ztest_run(ztest_shared_t *zs) |
34dc7c2f | 5266 | { |
1e33ac1e | 5267 | kt_did_t *tid; |
34dc7c2f | 5268 | spa_t *spa; |
1e33ac1e BB |
5269 | kthread_t *resume_thread; |
5270 | uint64_t object; | |
428870ff | 5271 | int error; |
d6320ddb | 5272 | int t, d; |
b128c09f BB |
5273 | |
5274 | ztest_exiting = B_FALSE; | |
34dc7c2f | 5275 | |
34dc7c2f | 5276 | /* |
428870ff | 5277 | * Initialize parent/child shared state. |
34dc7c2f | 5278 | */ |
1e33ac1e BB |
5279 | mutex_init(&zs->zs_vdev_lock, NULL, MUTEX_DEFAULT, NULL); |
5280 | rw_init(&zs->zs_name_lock, NULL, RW_DEFAULT, NULL); | |
34dc7c2f | 5281 | |
428870ff BB |
5282 | zs->zs_thread_start = gethrtime(); |
5283 | zs->zs_thread_stop = zs->zs_thread_start + zopt_passtime * NANOSEC; | |
5284 | zs->zs_thread_stop = MIN(zs->zs_thread_stop, zs->zs_proc_stop); | |
5285 | zs->zs_thread_kill = zs->zs_thread_stop; | |
5286 | if (ztest_random(100) < zopt_killrate) | |
5287 | zs->zs_thread_kill -= ztest_random(zopt_passtime * NANOSEC); | |
34dc7c2f | 5288 | |
1e33ac1e | 5289 | mutex_init(&zcl.zcl_callbacks_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 5290 | |
428870ff BB |
5291 | list_create(&zcl.zcl_callbacks, sizeof (ztest_cb_data_t), |
5292 | offsetof(ztest_cb_data_t, zcd_node)); | |
34dc7c2f BB |
5293 | |
5294 | /* | |
b128c09f | 5295 | * Open our pool. |
34dc7c2f | 5296 | */ |
428870ff BB |
5297 | kernel_init(FREAD | FWRITE); |
5298 | VERIFY(spa_open(zs->zs_pool, &spa, FTAG) == 0); | |
5299 | zs->zs_spa = spa; | |
5300 | ||
5301 | spa->spa_dedup_ditto = 2 * ZIO_DEDUPDITTO_MIN; | |
34dc7c2f | 5302 | |
fb5f0bc8 BB |
5303 | /* |
5304 | * We don't expect the pool to suspend unless maxfaults == 0, | |
5305 | * in which case ztest_fault_inject() temporarily takes away | |
5306 | * the only valid replica. | |
5307 | */ | |
428870ff | 5308 | if (MAXFAULTS() == 0) |
fb5f0bc8 BB |
5309 | spa->spa_failmode = ZIO_FAILURE_MODE_WAIT; |
5310 | else | |
5311 | spa->spa_failmode = ZIO_FAILURE_MODE_PANIC; | |
5312 | ||
34dc7c2f | 5313 | /* |
b128c09f | 5314 | * Create a thread to periodically resume suspended I/O. |
34dc7c2f | 5315 | */ |
1e33ac1e BB |
5316 | VERIFY3P((resume_thread = thread_create(NULL, 0, ztest_resume_thread, |
5317 | spa, TS_RUN, NULL, 0, 0)), !=, NULL); | |
34dc7c2f | 5318 | |
428870ff | 5319 | /* |
1e33ac1e | 5320 | * Set a deadman alarm to abort() if we hang. |
428870ff | 5321 | */ |
1e33ac1e BB |
5322 | signal(SIGALRM, ztest_deadman_alarm); |
5323 | alarm((zs->zs_thread_stop - zs->zs_thread_start) / NANOSEC + GRACE); | |
428870ff | 5324 | |
34dc7c2f BB |
5325 | /* |
5326 | * Verify that we can safely inquire about about any object, | |
5327 | * whether it's allocated or not. To make it interesting, | |
5328 | * we probe a 5-wide window around each power of two. | |
5329 | * This hits all edge cases, including zero and the max. | |
5330 | */ | |
d6320ddb BB |
5331 | for (t = 0; t < 64; t++) { |
5332 | for (d = -5; d <= 5; d++) { | |
34dc7c2f BB |
5333 | error = dmu_object_info(spa->spa_meta_objset, |
5334 | (1ULL << t) + d, NULL); | |
5335 | ASSERT(error == 0 || error == ENOENT || | |
5336 | error == EINVAL); | |
5337 | } | |
5338 | } | |
5339 | ||
5340 | /* | |
428870ff | 5341 | * If we got any ENOSPC errors on the previous run, destroy something. |
34dc7c2f | 5342 | */ |
428870ff BB |
5343 | if (zs->zs_enospc_count != 0) { |
5344 | int d = ztest_random(zopt_datasets); | |
5345 | ztest_dataset_destroy(zs, d); | |
5346 | } | |
34dc7c2f BB |
5347 | zs->zs_enospc_count = 0; |
5348 | ||
1e33ac1e | 5349 | tid = umem_zalloc(zopt_threads * sizeof (kt_did_t), UMEM_NOFAIL); |
34dc7c2f BB |
5350 | |
5351 | if (zopt_verbose >= 4) | |
5352 | (void) printf("starting main threads...\n"); | |
5353 | ||
428870ff BB |
5354 | /* |
5355 | * Kick off all the tests that run in parallel. | |
5356 | */ | |
d6320ddb | 5357 | for (t = 0; t < zopt_threads; t++) { |
1e33ac1e BB |
5358 | kthread_t *thread; |
5359 | ||
428870ff BB |
5360 | if (t < zopt_datasets && ztest_dataset_open(zs, t) != 0) |
5361 | return; | |
1e33ac1e BB |
5362 | |
5363 | VERIFY3P(thread = thread_create(NULL, 0, ztest_thread, | |
5364 | (void *)(uintptr_t)t, TS_RUN, NULL, 0, 0), !=, NULL); | |
5365 | tid[t] = thread->t_tid; | |
34dc7c2f BB |
5366 | } |
5367 | ||
428870ff BB |
5368 | /* |
5369 | * Wait for all of the tests to complete. We go in reverse order | |
5370 | * so we don't close datasets while threads are still using them. | |
5371 | */ | |
d6320ddb | 5372 | for (t = zopt_threads - 1; t >= 0; t--) { |
1e33ac1e | 5373 | thread_join(tid[t]); |
428870ff BB |
5374 | if (t < zopt_datasets) |
5375 | ztest_dataset_close(zs, t); | |
34dc7c2f BB |
5376 | } |
5377 | ||
34dc7c2f BB |
5378 | txg_wait_synced(spa_get_dsl(spa), 0); |
5379 | ||
428870ff BB |
5380 | zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(spa)); |
5381 | zs->zs_space = metaslab_class_get_space(spa_normal_class(spa)); | |
5382 | ||
1e33ac1e | 5383 | umem_free(tid, zopt_threads * sizeof (kt_did_t)); |
428870ff BB |
5384 | |
5385 | /* Kill the resume thread */ | |
5386 | ztest_exiting = B_TRUE; | |
1e33ac1e | 5387 | thread_join(resume_thread->t_tid); |
428870ff | 5388 | ztest_resume(spa); |
34dc7c2f BB |
5389 | |
5390 | /* | |
428870ff BB |
5391 | * Right before closing the pool, kick off a bunch of async I/O; |
5392 | * spa_close() should wait for it to complete. | |
34dc7c2f | 5393 | */ |
1e33ac1e | 5394 | for (object = 1; object < 50; object++) |
428870ff BB |
5395 | dmu_prefetch(spa->spa_meta_objset, object, 0, 1ULL << 20); |
5396 | ||
090ff092 RC |
5397 | /* Verify that at least one commit cb was called in a timely fashion */ |
5398 | if (zc_cb_counter >= ZTEST_COMMIT_CB_MIN_REG) | |
5399 | VERIFY3U(zc_min_txg_delay, ==, 0); | |
5400 | ||
428870ff BB |
5401 | spa_close(spa, FTAG); |
5402 | ||
5403 | /* | |
5404 | * Verify that we can loop over all pools. | |
5405 | */ | |
5406 | mutex_enter(&spa_namespace_lock); | |
5407 | for (spa = spa_next(NULL); spa != NULL; spa = spa_next(spa)) | |
5408 | if (zopt_verbose > 3) | |
5409 | (void) printf("spa_next: found %s\n", spa_name(spa)); | |
5410 | mutex_exit(&spa_namespace_lock); | |
5411 | ||
5412 | /* | |
5413 | * Verify that we can export the pool and reimport it under a | |
5414 | * different name. | |
5415 | */ | |
5416 | if (ztest_random(2) == 0) { | |
5417 | char name[MAXNAMELEN]; | |
5418 | (void) snprintf(name, MAXNAMELEN, "%s_import", zs->zs_pool); | |
5419 | ztest_spa_import_export(zs->zs_pool, name); | |
5420 | ztest_spa_import_export(name, zs->zs_pool); | |
5421 | } | |
5422 | ||
5423 | kernel_fini(); | |
572e2857 BB |
5424 | |
5425 | list_destroy(&zcl.zcl_callbacks); | |
5426 | ||
5427 | (void) _mutex_destroy(&zcl.zcl_callbacks_lock); | |
5428 | ||
5429 | (void) rwlock_destroy(&zs->zs_name_lock); | |
5430 | (void) _mutex_destroy(&zs->zs_vdev_lock); | |
428870ff BB |
5431 | } |
5432 | ||
5433 | static void | |
5434 | ztest_freeze(ztest_shared_t *zs) | |
5435 | { | |
5436 | ztest_ds_t *zd = &zs->zs_zd[0]; | |
5437 | spa_t *spa; | |
5438 | int numloops = 0; | |
5439 | ||
5440 | if (zopt_verbose >= 3) | |
5441 | (void) printf("testing spa_freeze()...\n"); | |
9babb374 | 5442 | |
428870ff BB |
5443 | kernel_init(FREAD | FWRITE); |
5444 | VERIFY3U(0, ==, spa_open(zs->zs_pool, &spa, FTAG)); | |
5445 | VERIFY3U(0, ==, ztest_dataset_open(zs, 0)); | |
9babb374 | 5446 | |
428870ff BB |
5447 | /* |
5448 | * Force the first log block to be transactionally allocated. | |
5449 | * We have to do this before we freeze the pool -- otherwise | |
5450 | * the log chain won't be anchored. | |
5451 | */ | |
5452 | while (BP_IS_HOLE(&zd->zd_zilog->zl_header->zh_log)) { | |
5453 | ztest_dmu_object_alloc_free(zd, 0); | |
572e2857 | 5454 | zil_commit(zd->zd_zilog, 0); |
34dc7c2f BB |
5455 | } |
5456 | ||
5457 | txg_wait_synced(spa_get_dsl(spa), 0); | |
5458 | ||
428870ff BB |
5459 | /* |
5460 | * Freeze the pool. This stops spa_sync() from doing anything, | |
5461 | * so that the only way to record changes from now on is the ZIL. | |
5462 | */ | |
5463 | spa_freeze(spa); | |
b128c09f | 5464 | |
428870ff BB |
5465 | /* |
5466 | * Run tests that generate log records but don't alter the pool config | |
5467 | * or depend on DSL sync tasks (snapshots, objset create/destroy, etc). | |
5468 | * We do a txg_wait_synced() after each iteration to force the txg | |
5469 | * to increase well beyond the last synced value in the uberblock. | |
5470 | * The ZIL should be OK with that. | |
5471 | */ | |
5472 | while (ztest_random(10) != 0 && numloops++ < zopt_maxloops) { | |
5473 | ztest_dmu_write_parallel(zd, 0); | |
5474 | ztest_dmu_object_alloc_free(zd, 0); | |
5475 | txg_wait_synced(spa_get_dsl(spa), 0); | |
5476 | } | |
b128c09f | 5477 | |
34dc7c2f | 5478 | /* |
428870ff | 5479 | * Commit all of the changes we just generated. |
34dc7c2f | 5480 | */ |
572e2857 | 5481 | zil_commit(zd->zd_zilog, 0); |
428870ff | 5482 | txg_wait_synced(spa_get_dsl(spa), 0); |
34dc7c2f | 5483 | |
428870ff BB |
5484 | /* |
5485 | * Close our dataset and close the pool. | |
5486 | */ | |
5487 | ztest_dataset_close(zs, 0); | |
34dc7c2f | 5488 | spa_close(spa, FTAG); |
428870ff | 5489 | kernel_fini(); |
34dc7c2f | 5490 | |
428870ff BB |
5491 | /* |
5492 | * Open and close the pool and dataset to induce log replay. | |
5493 | */ | |
5494 | kernel_init(FREAD | FWRITE); | |
5495 | VERIFY3U(0, ==, spa_open(zs->zs_pool, &spa, FTAG)); | |
5496 | VERIFY3U(0, ==, ztest_dataset_open(zs, 0)); | |
5497 | ztest_dataset_close(zs, 0); | |
5498 | spa_close(spa, FTAG); | |
34dc7c2f BB |
5499 | kernel_fini(); |
5500 | } | |
5501 | ||
5502 | void | |
5503 | print_time(hrtime_t t, char *timebuf) | |
5504 | { | |
5505 | hrtime_t s = t / NANOSEC; | |
5506 | hrtime_t m = s / 60; | |
5507 | hrtime_t h = m / 60; | |
5508 | hrtime_t d = h / 24; | |
5509 | ||
5510 | s -= m * 60; | |
5511 | m -= h * 60; | |
5512 | h -= d * 24; | |
5513 | ||
5514 | timebuf[0] = '\0'; | |
5515 | ||
5516 | if (d) | |
5517 | (void) sprintf(timebuf, | |
5518 | "%llud%02lluh%02llum%02llus", d, h, m, s); | |
5519 | else if (h) | |
5520 | (void) sprintf(timebuf, "%lluh%02llum%02llus", h, m, s); | |
5521 | else if (m) | |
5522 | (void) sprintf(timebuf, "%llum%02llus", m, s); | |
5523 | else | |
5524 | (void) sprintf(timebuf, "%llus", s); | |
5525 | } | |
5526 | ||
428870ff | 5527 | static nvlist_t * |
0bc8fd78 | 5528 | make_random_props(void) |
428870ff BB |
5529 | { |
5530 | nvlist_t *props; | |
5531 | ||
5532 | if (ztest_random(2) == 0) | |
5533 | return (NULL); | |
5534 | ||
5535 | VERIFY(nvlist_alloc(&props, NV_UNIQUE_NAME, 0) == 0); | |
5536 | VERIFY(nvlist_add_uint64(props, "autoreplace", 1) == 0); | |
5537 | ||
5538 | (void) printf("props:\n"); | |
5539 | dump_nvlist(props, 4); | |
5540 | ||
5541 | return (props); | |
5542 | } | |
5543 | ||
34dc7c2f BB |
5544 | /* |
5545 | * Create a storage pool with the given name and initial vdev size. | |
428870ff | 5546 | * Then test spa_freeze() functionality. |
34dc7c2f BB |
5547 | */ |
5548 | static void | |
428870ff | 5549 | ztest_init(ztest_shared_t *zs) |
34dc7c2f BB |
5550 | { |
5551 | spa_t *spa; | |
428870ff BB |
5552 | nvlist_t *nvroot, *props; |
5553 | ||
1e33ac1e BB |
5554 | mutex_init(&zs->zs_vdev_lock, NULL, MUTEX_DEFAULT, NULL); |
5555 | rw_init(&zs->zs_name_lock, NULL, RW_DEFAULT, NULL); | |
34dc7c2f BB |
5556 | |
5557 | kernel_init(FREAD | FWRITE); | |
5558 | ||
5559 | /* | |
5560 | * Create the storage pool. | |
5561 | */ | |
428870ff BB |
5562 | (void) spa_destroy(zs->zs_pool); |
5563 | ztest_shared->zs_vdev_next_leaf = 0; | |
5564 | zs->zs_splits = 0; | |
5565 | zs->zs_mirrors = zopt_mirrors; | |
b128c09f | 5566 | nvroot = make_vdev_root(NULL, NULL, zopt_vdev_size, 0, |
428870ff BB |
5567 | 0, zopt_raidz, zs->zs_mirrors, 1); |
5568 | props = make_random_props(); | |
5569 | VERIFY3U(0, ==, spa_create(zs->zs_pool, nvroot, props, NULL, NULL)); | |
34dc7c2f BB |
5570 | nvlist_free(nvroot); |
5571 | ||
428870ff | 5572 | VERIFY3U(0, ==, spa_open(zs->zs_pool, &spa, FTAG)); |
9babb374 | 5573 | metaslab_sz = 1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift; |
34dc7c2f BB |
5574 | spa_close(spa, FTAG); |
5575 | ||
5576 | kernel_fini(); | |
428870ff BB |
5577 | |
5578 | ztest_run_zdb(zs->zs_pool); | |
5579 | ||
5580 | ztest_freeze(zs); | |
5581 | ||
5582 | ztest_run_zdb(zs->zs_pool); | |
572e2857 | 5583 | |
1e33ac1e BB |
5584 | (void) rw_destroy(&zs->zs_name_lock); |
5585 | (void) mutex_destroy(&zs->zs_vdev_lock); | |
34dc7c2f BB |
5586 | } |
5587 | ||
5588 | int | |
5589 | main(int argc, char **argv) | |
5590 | { | |
5591 | int kills = 0; | |
5592 | int iters = 0; | |
34dc7c2f | 5593 | ztest_shared_t *zs; |
428870ff | 5594 | size_t shared_size; |
34dc7c2f BB |
5595 | ztest_info_t *zi; |
5596 | char timebuf[100]; | |
5597 | char numbuf[6]; | |
428870ff | 5598 | spa_t *spa; |
d6320ddb | 5599 | int i, f; |
34dc7c2f BB |
5600 | |
5601 | (void) setvbuf(stdout, NULL, _IOLBF, 0); | |
5602 | ||
34dc7c2f BB |
5603 | ztest_random_fd = open("/dev/urandom", O_RDONLY); |
5604 | ||
5605 | process_options(argc, argv); | |
5606 | ||
428870ff | 5607 | /* Override location of zpool.cache */ |
0e5b68e0 BB |
5608 | VERIFY(asprintf((char **)&spa_config_path, "%s/zpool.cache", |
5609 | zopt_dir) != -1); | |
428870ff | 5610 | |
34dc7c2f BB |
5611 | /* |
5612 | * Blow away any existing copy of zpool.cache | |
5613 | */ | |
5614 | if (zopt_init != 0) | |
428870ff BB |
5615 | (void) remove(spa_config_path); |
5616 | ||
5617 | shared_size = sizeof (*zs) + zopt_datasets * sizeof (ztest_ds_t); | |
34dc7c2f BB |
5618 | |
5619 | zs = ztest_shared = (void *)mmap(0, | |
428870ff | 5620 | P2ROUNDUP(shared_size, getpagesize()), |
34dc7c2f BB |
5621 | PROT_READ | PROT_WRITE, MAP_SHARED | MAP_ANON, -1, 0); |
5622 | ||
5623 | if (zopt_verbose >= 1) { | |
5624 | (void) printf("%llu vdevs, %d datasets, %d threads," | |
5625 | " %llu seconds...\n", | |
5626 | (u_longlong_t)zopt_vdevs, zopt_datasets, zopt_threads, | |
5627 | (u_longlong_t)zopt_time); | |
5628 | } | |
5629 | ||
5630 | /* | |
5631 | * Create and initialize our storage pool. | |
5632 | */ | |
d6320ddb | 5633 | for (i = 1; i <= zopt_init; i++) { |
34dc7c2f BB |
5634 | bzero(zs, sizeof (ztest_shared_t)); |
5635 | if (zopt_verbose >= 3 && zopt_init != 1) | |
5636 | (void) printf("ztest_init(), pass %d\n", i); | |
428870ff BB |
5637 | zs->zs_pool = zopt_pool; |
5638 | ztest_init(zs); | |
34dc7c2f BB |
5639 | } |
5640 | ||
428870ff BB |
5641 | zs->zs_pool = zopt_pool; |
5642 | zs->zs_proc_start = gethrtime(); | |
5643 | zs->zs_proc_stop = zs->zs_proc_start + zopt_time * NANOSEC; | |
34dc7c2f | 5644 | |
d6320ddb | 5645 | for (f = 0; f < ZTEST_FUNCS; f++) { |
428870ff | 5646 | zi = &zs->zs_info[f]; |
34dc7c2f | 5647 | *zi = ztest_info[f]; |
428870ff BB |
5648 | if (zs->zs_proc_start + zi->zi_interval[0] > zs->zs_proc_stop) |
5649 | zi->zi_call_next = UINT64_MAX; | |
34dc7c2f | 5650 | else |
428870ff BB |
5651 | zi->zi_call_next = zs->zs_proc_start + |
5652 | ztest_random(2 * zi->zi_interval[0] + 1); | |
34dc7c2f BB |
5653 | } |
5654 | ||
34dc7c2f BB |
5655 | /* |
5656 | * Run the tests in a loop. These tests include fault injection | |
5657 | * to verify that self-healing data works, and forced crashes | |
5658 | * to verify that we never lose on-disk consistency. | |
5659 | */ | |
428870ff | 5660 | while (gethrtime() < zs->zs_proc_stop) { |
34dc7c2f BB |
5661 | int status; |
5662 | pid_t pid; | |
34dc7c2f BB |
5663 | |
5664 | /* | |
5665 | * Initialize the workload counters for each function. | |
5666 | */ | |
d6320ddb | 5667 | for (f = 0; f < ZTEST_FUNCS; f++) { |
34dc7c2f | 5668 | zi = &zs->zs_info[f]; |
428870ff | 5669 | zi->zi_call_count = 0; |
34dc7c2f BB |
5670 | zi->zi_call_time = 0; |
5671 | } | |
5672 | ||
9babb374 BB |
5673 | /* Set the allocation switch size */ |
5674 | metaslab_df_alloc_threshold = ztest_random(metaslab_sz / 4) + 1; | |
5675 | ||
34dc7c2f BB |
5676 | pid = fork(); |
5677 | ||
5678 | if (pid == -1) | |
5679 | fatal(1, "fork failed"); | |
5680 | ||
5681 | if (pid == 0) { /* child */ | |
5682 | struct rlimit rl = { 1024, 1024 }; | |
5683 | (void) setrlimit(RLIMIT_NOFILE, &rl); | |
5684 | (void) enable_extended_FILE_stdio(-1, -1); | |
428870ff | 5685 | ztest_run(zs); |
34dc7c2f BB |
5686 | exit(0); |
5687 | } | |
5688 | ||
5689 | while (waitpid(pid, &status, 0) != pid) | |
5690 | continue; | |
5691 | ||
5692 | if (WIFEXITED(status)) { | |
5693 | if (WEXITSTATUS(status) != 0) { | |
5694 | (void) fprintf(stderr, | |
5695 | "child exited with code %d\n", | |
5696 | WEXITSTATUS(status)); | |
5697 | exit(2); | |
5698 | } | |
5699 | } else if (WIFSIGNALED(status)) { | |
5700 | if (WTERMSIG(status) != SIGKILL) { | |
5701 | (void) fprintf(stderr, | |
5702 | "child died with signal %d\n", | |
5703 | WTERMSIG(status)); | |
5704 | exit(3); | |
5705 | } | |
5706 | kills++; | |
5707 | } else { | |
5708 | (void) fprintf(stderr, "something strange happened " | |
5709 | "to child\n"); | |
5710 | exit(4); | |
5711 | } | |
5712 | ||
5713 | iters++; | |
5714 | ||
5715 | if (zopt_verbose >= 1) { | |
5716 | hrtime_t now = gethrtime(); | |
5717 | ||
428870ff BB |
5718 | now = MIN(now, zs->zs_proc_stop); |
5719 | print_time(zs->zs_proc_stop - now, timebuf); | |
34dc7c2f BB |
5720 | nicenum(zs->zs_space, numbuf); |
5721 | ||
5722 | (void) printf("Pass %3d, %8s, %3llu ENOSPC, " | |
5723 | "%4.1f%% of %5s used, %3.0f%% done, %8s to go\n", | |
5724 | iters, | |
5725 | WIFEXITED(status) ? "Complete" : "SIGKILL", | |
5726 | (u_longlong_t)zs->zs_enospc_count, | |
5727 | 100.0 * zs->zs_alloc / zs->zs_space, | |
5728 | numbuf, | |
428870ff | 5729 | 100.0 * (now - zs->zs_proc_start) / |
34dc7c2f BB |
5730 | (zopt_time * NANOSEC), timebuf); |
5731 | } | |
5732 | ||
5733 | if (zopt_verbose >= 2) { | |
5734 | (void) printf("\nWorkload summary:\n\n"); | |
5735 | (void) printf("%7s %9s %s\n", | |
5736 | "Calls", "Time", "Function"); | |
5737 | (void) printf("%7s %9s %s\n", | |
5738 | "-----", "----", "--------"); | |
d6320ddb | 5739 | for (f = 0; f < ZTEST_FUNCS; f++) { |
34dc7c2f BB |
5740 | Dl_info dli; |
5741 | ||
5742 | zi = &zs->zs_info[f]; | |
5743 | print_time(zi->zi_call_time, timebuf); | |
5744 | (void) dladdr((void *)zi->zi_func, &dli); | |
5745 | (void) printf("%7llu %9s %s\n", | |
428870ff | 5746 | (u_longlong_t)zi->zi_call_count, timebuf, |
34dc7c2f BB |
5747 | dli.dli_sname); |
5748 | } | |
5749 | (void) printf("\n"); | |
5750 | } | |
5751 | ||
5752 | /* | |
428870ff BB |
5753 | * It's possible that we killed a child during a rename test, |
5754 | * in which case we'll have a 'ztest_tmp' pool lying around | |
5755 | * instead of 'ztest'. Do a blind rename in case this happened. | |
34dc7c2f | 5756 | */ |
428870ff BB |
5757 | kernel_init(FREAD); |
5758 | if (spa_open(zopt_pool, &spa, FTAG) == 0) { | |
5759 | spa_close(spa, FTAG); | |
5760 | } else { | |
5761 | char tmpname[MAXNAMELEN]; | |
5762 | kernel_fini(); | |
5763 | kernel_init(FREAD | FWRITE); | |
5764 | (void) snprintf(tmpname, sizeof (tmpname), "%s_tmp", | |
5765 | zopt_pool); | |
5766 | (void) spa_rename(tmpname, zopt_pool); | |
5767 | } | |
34dc7c2f | 5768 | kernel_fini(); |
34dc7c2f | 5769 | |
428870ff BB |
5770 | ztest_run_zdb(zopt_pool); |
5771 | } | |
34dc7c2f BB |
5772 | |
5773 | if (zopt_verbose >= 1) { | |
5774 | (void) printf("%d killed, %d completed, %.0f%% kill rate\n", | |
5775 | kills, iters - kills, (100.0 * kills) / MAX(1, iters)); | |
5776 | } | |
5777 | ||
5778 | return (0); | |
5779 | } |