]>
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. |
6d974228 | 23 | * Copyright (c) 2011 by Delphix. All rights reserved. |
34dc7c2f BB |
24 | */ |
25 | ||
34dc7c2f BB |
26 | /* |
27 | * The objective of this program is to provide a DMU/ZAP/SPA stress test | |
28 | * that runs entirely in userland, is easy to use, and easy to extend. | |
29 | * | |
30 | * The overall design of the ztest program is as follows: | |
31 | * | |
32 | * (1) For each major functional area (e.g. adding vdevs to a pool, | |
33 | * creating and destroying datasets, reading and writing objects, etc) | |
34 | * we have a simple routine to test that functionality. These | |
35 | * individual routines do not have to do anything "stressful". | |
36 | * | |
37 | * (2) We turn these simple functionality tests into a stress test by | |
38 | * running them all in parallel, with as many threads as desired, | |
39 | * and spread across as many datasets, objects, and vdevs as desired. | |
40 | * | |
41 | * (3) While all this is happening, we inject faults into the pool to | |
42 | * verify that self-healing data really works. | |
43 | * | |
44 | * (4) Every time we open a dataset, we change its checksum and compression | |
45 | * functions. Thus even individual objects vary from block to block | |
46 | * in which checksum they use and whether they're compressed. | |
47 | * | |
48 | * (5) To verify that we never lose on-disk consistency after a crash, | |
49 | * we run the entire test in a child of the main process. | |
50 | * At random times, the child self-immolates with a SIGKILL. | |
51 | * This is the software equivalent of pulling the power cord. | |
52 | * The parent then runs the test again, using the existing | |
53 | * storage pool, as many times as desired. | |
54 | * | |
55 | * (6) To verify that we don't have future leaks or temporal incursions, | |
56 | * many of the functional tests record the transaction group number | |
57 | * as part of their data. When reading old data, they verify that | |
58 | * the transaction group number is less than the current, open txg. | |
59 | * If you add a new test, please do this if applicable. | |
60 | * | |
1e33ac1e BB |
61 | * (7) Threads are created with a reduced stack size, for sanity checking. |
62 | * Therefore, it's important not to allocate huge buffers on the stack. | |
63 | * | |
34dc7c2f BB |
64 | * When run with no arguments, ztest runs for about five minutes and |
65 | * produces no output if successful. To get a little bit of information, | |
66 | * specify -V. To get more information, specify -VV, and so on. | |
67 | * | |
68 | * To turn this into an overnight stress test, use -T to specify run time. | |
69 | * | |
70 | * You can ask more more vdevs [-v], datasets [-d], or threads [-t] | |
71 | * to increase the pool capacity, fanout, and overall stress level. | |
72 | * | |
73 | * The -N(okill) option will suppress kills, so each child runs to completion. | |
74 | * This can be useful when you're trying to distinguish temporal incursions | |
75 | * from plain old race conditions. | |
76 | */ | |
77 | ||
78 | #include <sys/zfs_context.h> | |
79 | #include <sys/spa.h> | |
80 | #include <sys/dmu.h> | |
81 | #include <sys/txg.h> | |
9babb374 | 82 | #include <sys/dbuf.h> |
34dc7c2f | 83 | #include <sys/zap.h> |
34dc7c2f BB |
84 | #include <sys/dmu_objset.h> |
85 | #include <sys/poll.h> | |
86 | #include <sys/stat.h> | |
87 | #include <sys/time.h> | |
88 | #include <sys/wait.h> | |
89 | #include <sys/mman.h> | |
90 | #include <sys/resource.h> | |
91 | #include <sys/zio.h> | |
34dc7c2f | 92 | #include <sys/zil.h> |
428870ff | 93 | #include <sys/zil_impl.h> |
34dc7c2f | 94 | #include <sys/vdev_impl.h> |
b128c09f | 95 | #include <sys/vdev_file.h> |
34dc7c2f | 96 | #include <sys/spa_impl.h> |
428870ff | 97 | #include <sys/metaslab_impl.h> |
34dc7c2f | 98 | #include <sys/dsl_prop.h> |
9babb374 | 99 | #include <sys/dsl_dataset.h> |
428870ff BB |
100 | #include <sys/dsl_scan.h> |
101 | #include <sys/zio_checksum.h> | |
34dc7c2f BB |
102 | #include <sys/refcount.h> |
103 | #include <stdio.h> | |
104 | #include <stdio_ext.h> | |
105 | #include <stdlib.h> | |
106 | #include <unistd.h> | |
107 | #include <signal.h> | |
108 | #include <umem.h> | |
109 | #include <dlfcn.h> | |
110 | #include <ctype.h> | |
111 | #include <math.h> | |
112 | #include <sys/fs/zfs.h> | |
428870ff | 113 | #include <libnvpair.h> |
34dc7c2f BB |
114 | |
115 | static char cmdname[] = "ztest"; | |
116 | static char *zopt_pool = cmdname; | |
117 | ||
118 | static uint64_t zopt_vdevs = 5; | |
119 | static uint64_t zopt_vdevtime; | |
120 | static int zopt_ashift = SPA_MINBLOCKSHIFT; | |
121 | static int zopt_mirrors = 2; | |
122 | static int zopt_raidz = 4; | |
123 | static int zopt_raidz_parity = 1; | |
124 | static size_t zopt_vdev_size = SPA_MINDEVSIZE; | |
125 | static int zopt_datasets = 7; | |
126 | static int zopt_threads = 23; | |
127 | static uint64_t zopt_passtime = 60; /* 60 seconds */ | |
128 | static uint64_t zopt_killrate = 70; /* 70% kill rate */ | |
129 | static int zopt_verbose = 0; | |
130 | static int zopt_init = 1; | |
131 | static char *zopt_dir = "/tmp"; | |
132 | static uint64_t zopt_time = 300; /* 5 minutes */ | |
428870ff BB |
133 | static uint64_t zopt_maxloops = 50; /* max loops during spa_freeze() */ |
134 | ||
135 | #define BT_MAGIC 0x123456789abcdefULL | |
136 | #define MAXFAULTS() (MAX(zs->zs_mirrors, 1) * (zopt_raidz_parity + 1) - 1) | |
137 | ||
138 | enum ztest_io_type { | |
139 | ZTEST_IO_WRITE_TAG, | |
140 | ZTEST_IO_WRITE_PATTERN, | |
141 | ZTEST_IO_WRITE_ZEROES, | |
142 | ZTEST_IO_TRUNCATE, | |
143 | ZTEST_IO_SETATTR, | |
144 | ZTEST_IO_TYPES | |
145 | }; | |
34dc7c2f BB |
146 | |
147 | typedef struct ztest_block_tag { | |
428870ff | 148 | uint64_t bt_magic; |
34dc7c2f BB |
149 | uint64_t bt_objset; |
150 | uint64_t bt_object; | |
151 | uint64_t bt_offset; | |
428870ff | 152 | uint64_t bt_gen; |
34dc7c2f | 153 | uint64_t bt_txg; |
428870ff | 154 | uint64_t bt_crtxg; |
34dc7c2f BB |
155 | } ztest_block_tag_t; |
156 | ||
428870ff BB |
157 | typedef struct bufwad { |
158 | uint64_t bw_index; | |
159 | uint64_t bw_txg; | |
160 | uint64_t bw_data; | |
161 | } bufwad_t; | |
162 | ||
163 | /* | |
164 | * XXX -- fix zfs range locks to be generic so we can use them here. | |
165 | */ | |
166 | typedef enum { | |
167 | RL_READER, | |
168 | RL_WRITER, | |
169 | RL_APPEND | |
170 | } rl_type_t; | |
171 | ||
172 | typedef struct rll { | |
173 | void *rll_writer; | |
174 | int rll_readers; | |
1e33ac1e BB |
175 | kmutex_t rll_lock; |
176 | kcondvar_t rll_cv; | |
428870ff BB |
177 | } rll_t; |
178 | ||
179 | typedef struct rl { | |
180 | uint64_t rl_object; | |
181 | uint64_t rl_offset; | |
182 | uint64_t rl_size; | |
183 | rll_t *rl_lock; | |
184 | } rl_t; | |
185 | ||
186 | #define ZTEST_RANGE_LOCKS 64 | |
187 | #define ZTEST_OBJECT_LOCKS 64 | |
188 | ||
189 | /* | |
190 | * Object descriptor. Used as a template for object lookup/create/remove. | |
191 | */ | |
192 | typedef struct ztest_od { | |
193 | uint64_t od_dir; | |
194 | uint64_t od_object; | |
195 | dmu_object_type_t od_type; | |
196 | dmu_object_type_t od_crtype; | |
197 | uint64_t od_blocksize; | |
198 | uint64_t od_crblocksize; | |
199 | uint64_t od_gen; | |
200 | uint64_t od_crgen; | |
201 | char od_name[MAXNAMELEN]; | |
202 | } ztest_od_t; | |
34dc7c2f | 203 | |
428870ff BB |
204 | /* |
205 | * Per-dataset state. | |
206 | */ | |
207 | typedef struct ztest_ds { | |
208 | objset_t *zd_os; | |
3e31d2b0 | 209 | krwlock_t zd_zilog_lock; |
428870ff BB |
210 | zilog_t *zd_zilog; |
211 | uint64_t zd_seq; | |
212 | ztest_od_t *zd_od; /* debugging aid */ | |
213 | char zd_name[MAXNAMELEN]; | |
1e33ac1e | 214 | kmutex_t zd_dirobj_lock; |
428870ff BB |
215 | rll_t zd_object_lock[ZTEST_OBJECT_LOCKS]; |
216 | rll_t zd_range_lock[ZTEST_RANGE_LOCKS]; | |
217 | } ztest_ds_t; | |
218 | ||
219 | /* | |
220 | * Per-iteration state. | |
221 | */ | |
222 | typedef void ztest_func_t(ztest_ds_t *zd, uint64_t id); | |
223 | ||
224 | typedef struct ztest_info { | |
225 | ztest_func_t *zi_func; /* test function */ | |
226 | uint64_t zi_iters; /* iterations per execution */ | |
227 | uint64_t *zi_interval; /* execute every <interval> seconds */ | |
228 | uint64_t zi_call_count; /* per-pass count */ | |
229 | uint64_t zi_call_time; /* per-pass time */ | |
230 | uint64_t zi_call_next; /* next time to call this function */ | |
231 | } ztest_info_t; | |
34dc7c2f BB |
232 | |
233 | /* | |
234 | * Note: these aren't static because we want dladdr() to work. | |
235 | */ | |
236 | ztest_func_t ztest_dmu_read_write; | |
237 | ztest_func_t ztest_dmu_write_parallel; | |
238 | ztest_func_t ztest_dmu_object_alloc_free; | |
428870ff | 239 | ztest_func_t ztest_dmu_commit_callbacks; |
34dc7c2f BB |
240 | ztest_func_t ztest_zap; |
241 | ztest_func_t ztest_zap_parallel; | |
428870ff | 242 | ztest_func_t ztest_zil_commit; |
3e31d2b0 | 243 | ztest_func_t ztest_zil_remount; |
428870ff | 244 | ztest_func_t ztest_dmu_read_write_zcopy; |
34dc7c2f | 245 | ztest_func_t ztest_dmu_objset_create_destroy; |
428870ff BB |
246 | ztest_func_t ztest_dmu_prealloc; |
247 | ztest_func_t ztest_fzap; | |
34dc7c2f | 248 | ztest_func_t ztest_dmu_snapshot_create_destroy; |
428870ff BB |
249 | ztest_func_t ztest_dsl_prop_get_set; |
250 | ztest_func_t ztest_spa_prop_get_set; | |
34dc7c2f BB |
251 | ztest_func_t ztest_spa_create_destroy; |
252 | ztest_func_t ztest_fault_inject; | |
428870ff BB |
253 | ztest_func_t ztest_ddt_repair; |
254 | ztest_func_t ztest_dmu_snapshot_hold; | |
b128c09f | 255 | ztest_func_t ztest_spa_rename; |
428870ff BB |
256 | ztest_func_t ztest_scrub; |
257 | ztest_func_t ztest_dsl_dataset_promote_busy; | |
34dc7c2f BB |
258 | ztest_func_t ztest_vdev_attach_detach; |
259 | ztest_func_t ztest_vdev_LUN_growth; | |
260 | ztest_func_t ztest_vdev_add_remove; | |
b128c09f | 261 | ztest_func_t ztest_vdev_aux_add_remove; |
428870ff | 262 | ztest_func_t ztest_split_pool; |
34dc7c2f | 263 | |
428870ff BB |
264 | uint64_t zopt_always = 0ULL * NANOSEC; /* all the time */ |
265 | uint64_t zopt_incessant = 1ULL * NANOSEC / 10; /* every 1/10 second */ | |
266 | uint64_t zopt_often = 1ULL * NANOSEC; /* every second */ | |
267 | uint64_t zopt_sometimes = 10ULL * NANOSEC; /* every 10 seconds */ | |
268 | uint64_t zopt_rarely = 60ULL * NANOSEC; /* every 60 seconds */ | |
34dc7c2f BB |
269 | |
270 | ztest_info_t ztest_info[] = { | |
271 | { ztest_dmu_read_write, 1, &zopt_always }, | |
428870ff | 272 | { ztest_dmu_write_parallel, 10, &zopt_always }, |
34dc7c2f | 273 | { ztest_dmu_object_alloc_free, 1, &zopt_always }, |
428870ff | 274 | { ztest_dmu_commit_callbacks, 1, &zopt_always }, |
34dc7c2f BB |
275 | { ztest_zap, 30, &zopt_always }, |
276 | { ztest_zap_parallel, 100, &zopt_always }, | |
428870ff BB |
277 | { ztest_split_pool, 1, &zopt_always }, |
278 | { ztest_zil_commit, 1, &zopt_incessant }, | |
3e31d2b0 | 279 | { ztest_zil_remount, 1, &zopt_sometimes }, |
428870ff BB |
280 | { ztest_dmu_read_write_zcopy, 1, &zopt_often }, |
281 | { ztest_dmu_objset_create_destroy, 1, &zopt_often }, | |
282 | { ztest_dsl_prop_get_set, 1, &zopt_often }, | |
283 | { ztest_spa_prop_get_set, 1, &zopt_sometimes }, | |
284 | #if 0 | |
285 | { ztest_dmu_prealloc, 1, &zopt_sometimes }, | |
286 | #endif | |
287 | { ztest_fzap, 1, &zopt_sometimes }, | |
288 | { ztest_dmu_snapshot_create_destroy, 1, &zopt_sometimes }, | |
289 | { ztest_spa_create_destroy, 1, &zopt_sometimes }, | |
34dc7c2f | 290 | { ztest_fault_inject, 1, &zopt_sometimes }, |
428870ff BB |
291 | { ztest_ddt_repair, 1, &zopt_sometimes }, |
292 | { ztest_dmu_snapshot_hold, 1, &zopt_sometimes }, | |
34dc7c2f | 293 | { ztest_spa_rename, 1, &zopt_rarely }, |
428870ff | 294 | { ztest_scrub, 1, &zopt_rarely }, |
9babb374 | 295 | { ztest_dsl_dataset_promote_busy, 1, &zopt_rarely }, |
428870ff BB |
296 | { ztest_vdev_attach_detach, 1, &zopt_rarely }, |
297 | { ztest_vdev_LUN_growth, 1, &zopt_rarely }, | |
298 | { ztest_vdev_add_remove, 1, &zopt_vdevtime }, | |
b128c09f | 299 | { ztest_vdev_aux_add_remove, 1, &zopt_vdevtime }, |
34dc7c2f BB |
300 | }; |
301 | ||
302 | #define ZTEST_FUNCS (sizeof (ztest_info) / sizeof (ztest_info_t)) | |
303 | ||
428870ff BB |
304 | /* |
305 | * The following struct is used to hold a list of uncalled commit callbacks. | |
306 | * The callbacks are ordered by txg number. | |
307 | */ | |
308 | typedef struct ztest_cb_list { | |
1e33ac1e BB |
309 | kmutex_t zcl_callbacks_lock; |
310 | list_t zcl_callbacks; | |
428870ff | 311 | } ztest_cb_list_t; |
34dc7c2f BB |
312 | |
313 | /* | |
314 | * Stuff we need to share writably between parent and child. | |
315 | */ | |
316 | typedef struct ztest_shared { | |
428870ff BB |
317 | char *zs_pool; |
318 | spa_t *zs_spa; | |
319 | hrtime_t zs_proc_start; | |
320 | hrtime_t zs_proc_stop; | |
321 | hrtime_t zs_thread_start; | |
322 | hrtime_t zs_thread_stop; | |
323 | hrtime_t zs_thread_kill; | |
34dc7c2f | 324 | uint64_t zs_enospc_count; |
428870ff BB |
325 | uint64_t zs_vdev_next_leaf; |
326 | uint64_t zs_vdev_aux; | |
34dc7c2f BB |
327 | uint64_t zs_alloc; |
328 | uint64_t zs_space; | |
1e33ac1e BB |
329 | kmutex_t zs_vdev_lock; |
330 | krwlock_t zs_name_lock; | |
34dc7c2f | 331 | ztest_info_t zs_info[ZTEST_FUNCS]; |
428870ff BB |
332 | uint64_t zs_splits; |
333 | uint64_t zs_mirrors; | |
334 | ztest_ds_t zs_zd[]; | |
34dc7c2f BB |
335 | } ztest_shared_t; |
336 | ||
428870ff BB |
337 | #define ID_PARALLEL -1ULL |
338 | ||
34dc7c2f | 339 | static char ztest_dev_template[] = "%s/%s.%llua"; |
b128c09f | 340 | static char ztest_aux_template[] = "%s/%s.%s.%llu"; |
428870ff BB |
341 | ztest_shared_t *ztest_shared; |
342 | uint64_t *ztest_seq; | |
34dc7c2f BB |
343 | |
344 | static int ztest_random_fd; | |
345 | static int ztest_dump_core = 1; | |
346 | ||
b128c09f | 347 | static boolean_t ztest_exiting; |
34dc7c2f | 348 | |
428870ff BB |
349 | /* Global commit callback list */ |
350 | static ztest_cb_list_t zcl; | |
090ff092 RC |
351 | /* Commit cb delay */ |
352 | static uint64_t zc_min_txg_delay = UINT64_MAX; | |
353 | static int zc_cb_counter = 0; | |
354 | ||
355 | /* | |
356 | * Minimum number of commit callbacks that need to be registered for us to check | |
357 | * whether the minimum txg delay is acceptable. | |
358 | */ | |
359 | #define ZTEST_COMMIT_CB_MIN_REG 100 | |
360 | ||
361 | /* | |
362 | * If a number of txgs equal to this threshold have been created after a commit | |
363 | * callback has been registered but not called, then we assume there is an | |
364 | * implementation bug. | |
365 | */ | |
366 | #define ZTEST_COMMIT_CB_THRESH (TXG_CONCURRENT_STATES + 1000) | |
428870ff | 367 | |
34dc7c2f | 368 | extern uint64_t metaslab_gang_bang; |
9babb374 | 369 | extern uint64_t metaslab_df_alloc_threshold; |
428870ff | 370 | static uint64_t metaslab_sz; |
34dc7c2f | 371 | |
428870ff BB |
372 | enum ztest_object { |
373 | ZTEST_META_DNODE = 0, | |
374 | ZTEST_DIROBJ, | |
375 | ZTEST_OBJECTS | |
376 | }; | |
34dc7c2f BB |
377 | |
378 | static void usage(boolean_t) __NORETURN; | |
379 | ||
380 | /* | |
381 | * These libumem hooks provide a reasonable set of defaults for the allocator's | |
382 | * debugging facilities. | |
383 | */ | |
384 | const char * | |
0bc8fd78 | 385 | _umem_debug_init(void) |
34dc7c2f BB |
386 | { |
387 | return ("default,verbose"); /* $UMEM_DEBUG setting */ | |
388 | } | |
389 | ||
390 | const char * | |
391 | _umem_logging_init(void) | |
392 | { | |
393 | return ("fail,contents"); /* $UMEM_LOGGING setting */ | |
394 | } | |
395 | ||
396 | #define FATAL_MSG_SZ 1024 | |
397 | ||
398 | char *fatal_msg; | |
399 | ||
400 | static void | |
401 | fatal(int do_perror, char *message, ...) | |
402 | { | |
403 | va_list args; | |
404 | int save_errno = errno; | |
40b84e7a | 405 | char *buf; |
34dc7c2f BB |
406 | |
407 | (void) fflush(stdout); | |
40b84e7a | 408 | buf = umem_alloc(FATAL_MSG_SZ, UMEM_NOFAIL); |
34dc7c2f BB |
409 | |
410 | va_start(args, message); | |
411 | (void) sprintf(buf, "ztest: "); | |
412 | /* LINTED */ | |
413 | (void) vsprintf(buf + strlen(buf), message, args); | |
414 | va_end(args); | |
415 | if (do_perror) { | |
416 | (void) snprintf(buf + strlen(buf), FATAL_MSG_SZ - strlen(buf), | |
417 | ": %s", strerror(save_errno)); | |
418 | } | |
419 | (void) fprintf(stderr, "%s\n", buf); | |
420 | fatal_msg = buf; /* to ease debugging */ | |
421 | if (ztest_dump_core) | |
422 | abort(); | |
423 | exit(3); | |
424 | } | |
425 | ||
426 | static int | |
427 | str2shift(const char *buf) | |
428 | { | |
429 | const char *ends = "BKMGTPEZ"; | |
430 | int i; | |
431 | ||
432 | if (buf[0] == '\0') | |
433 | return (0); | |
434 | for (i = 0; i < strlen(ends); i++) { | |
435 | if (toupper(buf[0]) == ends[i]) | |
436 | break; | |
437 | } | |
438 | if (i == strlen(ends)) { | |
439 | (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", | |
440 | buf); | |
441 | usage(B_FALSE); | |
442 | } | |
443 | if (buf[1] == '\0' || (toupper(buf[1]) == 'B' && buf[2] == '\0')) { | |
444 | return (10*i); | |
445 | } | |
446 | (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", buf); | |
447 | usage(B_FALSE); | |
448 | /* NOTREACHED */ | |
449 | } | |
450 | ||
451 | static uint64_t | |
452 | nicenumtoull(const char *buf) | |
453 | { | |
454 | char *end; | |
455 | uint64_t val; | |
456 | ||
457 | val = strtoull(buf, &end, 0); | |
458 | if (end == buf) { | |
459 | (void) fprintf(stderr, "ztest: bad numeric value: %s\n", buf); | |
460 | usage(B_FALSE); | |
461 | } else if (end[0] == '.') { | |
462 | double fval = strtod(buf, &end); | |
463 | fval *= pow(2, str2shift(end)); | |
464 | if (fval > UINT64_MAX) { | |
465 | (void) fprintf(stderr, "ztest: value too large: %s\n", | |
466 | buf); | |
467 | usage(B_FALSE); | |
468 | } | |
469 | val = (uint64_t)fval; | |
470 | } else { | |
471 | int shift = str2shift(end); | |
472 | if (shift >= 64 || (val << shift) >> shift != val) { | |
473 | (void) fprintf(stderr, "ztest: value too large: %s\n", | |
474 | buf); | |
475 | usage(B_FALSE); | |
476 | } | |
477 | val <<= shift; | |
478 | } | |
479 | return (val); | |
480 | } | |
481 | ||
482 | static void | |
483 | usage(boolean_t requested) | |
484 | { | |
485 | char nice_vdev_size[10]; | |
486 | char nice_gang_bang[10]; | |
487 | FILE *fp = requested ? stdout : stderr; | |
488 | ||
489 | nicenum(zopt_vdev_size, nice_vdev_size); | |
490 | nicenum(metaslab_gang_bang, nice_gang_bang); | |
491 | ||
492 | (void) fprintf(fp, "Usage: %s\n" | |
493 | "\t[-v vdevs (default: %llu)]\n" | |
494 | "\t[-s size_of_each_vdev (default: %s)]\n" | |
428870ff | 495 | "\t[-a alignment_shift (default: %d)] use 0 for random\n" |
34dc7c2f BB |
496 | "\t[-m mirror_copies (default: %d)]\n" |
497 | "\t[-r raidz_disks (default: %d)]\n" | |
498 | "\t[-R raidz_parity (default: %d)]\n" | |
499 | "\t[-d datasets (default: %d)]\n" | |
500 | "\t[-t threads (default: %d)]\n" | |
501 | "\t[-g gang_block_threshold (default: %s)]\n" | |
428870ff BB |
502 | "\t[-i init_count (default: %d)] initialize pool i times\n" |
503 | "\t[-k kill_percentage (default: %llu%%)]\n" | |
34dc7c2f | 504 | "\t[-p pool_name (default: %s)]\n" |
428870ff BB |
505 | "\t[-f dir (default: %s)] file directory for vdev files\n" |
506 | "\t[-V] verbose (use multiple times for ever more blather)\n" | |
507 | "\t[-E] use existing pool instead of creating new one\n" | |
508 | "\t[-T time (default: %llu sec)] total run time\n" | |
509 | "\t[-F freezeloops (default: %llu)] max loops in spa_freeze()\n" | |
510 | "\t[-P passtime (default: %llu sec)] time per pass\n" | |
34dc7c2f BB |
511 | "\t[-h] (print help)\n" |
512 | "", | |
513 | cmdname, | |
514 | (u_longlong_t)zopt_vdevs, /* -v */ | |
515 | nice_vdev_size, /* -s */ | |
516 | zopt_ashift, /* -a */ | |
517 | zopt_mirrors, /* -m */ | |
518 | zopt_raidz, /* -r */ | |
519 | zopt_raidz_parity, /* -R */ | |
520 | zopt_datasets, /* -d */ | |
521 | zopt_threads, /* -t */ | |
522 | nice_gang_bang, /* -g */ | |
523 | zopt_init, /* -i */ | |
524 | (u_longlong_t)zopt_killrate, /* -k */ | |
525 | zopt_pool, /* -p */ | |
526 | zopt_dir, /* -f */ | |
527 | (u_longlong_t)zopt_time, /* -T */ | |
428870ff | 528 | (u_longlong_t)zopt_maxloops, /* -F */ |
b128c09f | 529 | (u_longlong_t)zopt_passtime); /* -P */ |
34dc7c2f BB |
530 | exit(requested ? 0 : 1); |
531 | } | |
532 | ||
34dc7c2f BB |
533 | static void |
534 | process_options(int argc, char **argv) | |
535 | { | |
536 | int opt; | |
537 | uint64_t value; | |
538 | ||
539 | /* By default, test gang blocks for blocks 32K and greater */ | |
540 | metaslab_gang_bang = 32 << 10; | |
541 | ||
34dc7c2f | 542 | while ((opt = getopt(argc, argv, |
428870ff | 543 | "v:s:a:m:r:R:d:t:g:i:k:p:f:VET:P:hF:")) != EOF) { |
34dc7c2f BB |
544 | value = 0; |
545 | switch (opt) { | |
546 | case 'v': | |
547 | case 's': | |
548 | case 'a': | |
549 | case 'm': | |
550 | case 'r': | |
551 | case 'R': | |
552 | case 'd': | |
553 | case 't': | |
554 | case 'g': | |
555 | case 'i': | |
556 | case 'k': | |
557 | case 'T': | |
558 | case 'P': | |
428870ff | 559 | case 'F': |
34dc7c2f BB |
560 | value = nicenumtoull(optarg); |
561 | } | |
562 | switch (opt) { | |
563 | case 'v': | |
564 | zopt_vdevs = value; | |
565 | break; | |
566 | case 's': | |
567 | zopt_vdev_size = MAX(SPA_MINDEVSIZE, value); | |
568 | break; | |
569 | case 'a': | |
570 | zopt_ashift = value; | |
571 | break; | |
572 | case 'm': | |
573 | zopt_mirrors = value; | |
574 | break; | |
575 | case 'r': | |
576 | zopt_raidz = MAX(1, value); | |
577 | break; | |
578 | case 'R': | |
45d1cae3 | 579 | zopt_raidz_parity = MIN(MAX(value, 1), 3); |
34dc7c2f BB |
580 | break; |
581 | case 'd': | |
582 | zopt_datasets = MAX(1, value); | |
583 | break; | |
584 | case 't': | |
585 | zopt_threads = MAX(1, value); | |
586 | break; | |
587 | case 'g': | |
588 | metaslab_gang_bang = MAX(SPA_MINBLOCKSIZE << 1, value); | |
589 | break; | |
590 | case 'i': | |
591 | zopt_init = value; | |
592 | break; | |
593 | case 'k': | |
594 | zopt_killrate = value; | |
595 | break; | |
596 | case 'p': | |
597 | zopt_pool = strdup(optarg); | |
598 | break; | |
599 | case 'f': | |
600 | zopt_dir = strdup(optarg); | |
601 | break; | |
602 | case 'V': | |
603 | zopt_verbose++; | |
604 | break; | |
605 | case 'E': | |
606 | zopt_init = 0; | |
607 | break; | |
608 | case 'T': | |
609 | zopt_time = value; | |
610 | break; | |
611 | case 'P': | |
612 | zopt_passtime = MAX(1, value); | |
613 | break; | |
428870ff BB |
614 | case 'F': |
615 | zopt_maxloops = MAX(1, value); | |
616 | break; | |
34dc7c2f BB |
617 | case 'h': |
618 | usage(B_TRUE); | |
619 | break; | |
620 | case '?': | |
621 | default: | |
622 | usage(B_FALSE); | |
623 | break; | |
624 | } | |
625 | } | |
626 | ||
627 | zopt_raidz_parity = MIN(zopt_raidz_parity, zopt_raidz - 1); | |
628 | ||
428870ff BB |
629 | zopt_vdevtime = (zopt_vdevs > 0 ? zopt_time * NANOSEC / zopt_vdevs : |
630 | UINT64_MAX >> 2); | |
631 | } | |
632 | ||
633 | static void | |
634 | ztest_kill(ztest_shared_t *zs) | |
635 | { | |
636 | zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(zs->zs_spa)); | |
637 | zs->zs_space = metaslab_class_get_space(spa_normal_class(zs->zs_spa)); | |
638 | (void) kill(getpid(), SIGKILL); | |
639 | } | |
640 | ||
641 | static uint64_t | |
642 | ztest_random(uint64_t range) | |
643 | { | |
644 | uint64_t r; | |
645 | ||
646 | if (range == 0) | |
647 | return (0); | |
648 | ||
649 | if (read(ztest_random_fd, &r, sizeof (r)) != sizeof (r)) | |
650 | fatal(1, "short read from /dev/urandom"); | |
651 | ||
652 | return (r % range); | |
653 | } | |
654 | ||
655 | /* ARGSUSED */ | |
656 | static void | |
657 | ztest_record_enospc(const char *s) | |
658 | { | |
659 | ztest_shared->zs_enospc_count++; | |
34dc7c2f BB |
660 | } |
661 | ||
662 | static uint64_t | |
663 | ztest_get_ashift(void) | |
664 | { | |
665 | if (zopt_ashift == 0) | |
666 | return (SPA_MINBLOCKSHIFT + ztest_random(3)); | |
667 | return (zopt_ashift); | |
668 | } | |
669 | ||
670 | static nvlist_t * | |
b128c09f | 671 | make_vdev_file(char *path, char *aux, size_t size, uint64_t ashift) |
34dc7c2f | 672 | { |
40b84e7a | 673 | char *pathbuf; |
34dc7c2f | 674 | uint64_t vdev; |
34dc7c2f BB |
675 | nvlist_t *file; |
676 | ||
40b84e7a BB |
677 | pathbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
678 | ||
b128c09f BB |
679 | if (ashift == 0) |
680 | ashift = ztest_get_ashift(); | |
681 | ||
682 | if (path == NULL) { | |
683 | path = pathbuf; | |
684 | ||
685 | if (aux != NULL) { | |
686 | vdev = ztest_shared->zs_vdev_aux; | |
687 | (void) sprintf(path, ztest_aux_template, | |
688 | zopt_dir, zopt_pool, aux, vdev); | |
689 | } else { | |
428870ff | 690 | vdev = ztest_shared->zs_vdev_next_leaf++; |
b128c09f BB |
691 | (void) sprintf(path, ztest_dev_template, |
692 | zopt_dir, zopt_pool, vdev); | |
693 | } | |
694 | } | |
34dc7c2f | 695 | |
b128c09f BB |
696 | if (size != 0) { |
697 | int fd = open(path, O_RDWR | O_CREAT | O_TRUNC, 0666); | |
34dc7c2f | 698 | if (fd == -1) |
b128c09f | 699 | fatal(1, "can't open %s", path); |
34dc7c2f | 700 | if (ftruncate(fd, size) != 0) |
b128c09f | 701 | fatal(1, "can't ftruncate %s", path); |
34dc7c2f BB |
702 | (void) close(fd); |
703 | } | |
704 | ||
705 | VERIFY(nvlist_alloc(&file, NV_UNIQUE_NAME, 0) == 0); | |
706 | VERIFY(nvlist_add_string(file, ZPOOL_CONFIG_TYPE, VDEV_TYPE_FILE) == 0); | |
b128c09f | 707 | VERIFY(nvlist_add_string(file, ZPOOL_CONFIG_PATH, path) == 0); |
34dc7c2f | 708 | VERIFY(nvlist_add_uint64(file, ZPOOL_CONFIG_ASHIFT, ashift) == 0); |
40b84e7a | 709 | umem_free(pathbuf, MAXPATHLEN); |
34dc7c2f BB |
710 | |
711 | return (file); | |
712 | } | |
713 | ||
714 | static nvlist_t * | |
b128c09f | 715 | make_vdev_raidz(char *path, char *aux, size_t size, uint64_t ashift, int r) |
34dc7c2f BB |
716 | { |
717 | nvlist_t *raidz, **child; | |
718 | int c; | |
719 | ||
720 | if (r < 2) | |
b128c09f | 721 | return (make_vdev_file(path, aux, size, ashift)); |
34dc7c2f BB |
722 | child = umem_alloc(r * sizeof (nvlist_t *), UMEM_NOFAIL); |
723 | ||
724 | for (c = 0; c < r; c++) | |
b128c09f | 725 | child[c] = make_vdev_file(path, aux, size, ashift); |
34dc7c2f BB |
726 | |
727 | VERIFY(nvlist_alloc(&raidz, NV_UNIQUE_NAME, 0) == 0); | |
728 | VERIFY(nvlist_add_string(raidz, ZPOOL_CONFIG_TYPE, | |
729 | VDEV_TYPE_RAIDZ) == 0); | |
730 | VERIFY(nvlist_add_uint64(raidz, ZPOOL_CONFIG_NPARITY, | |
731 | zopt_raidz_parity) == 0); | |
732 | VERIFY(nvlist_add_nvlist_array(raidz, ZPOOL_CONFIG_CHILDREN, | |
733 | child, r) == 0); | |
734 | ||
735 | for (c = 0; c < r; c++) | |
736 | nvlist_free(child[c]); | |
737 | ||
738 | umem_free(child, r * sizeof (nvlist_t *)); | |
739 | ||
740 | return (raidz); | |
741 | } | |
742 | ||
743 | static nvlist_t * | |
b128c09f BB |
744 | make_vdev_mirror(char *path, char *aux, size_t size, uint64_t ashift, |
745 | int r, int m) | |
34dc7c2f BB |
746 | { |
747 | nvlist_t *mirror, **child; | |
748 | int c; | |
749 | ||
750 | if (m < 1) | |
b128c09f | 751 | return (make_vdev_raidz(path, aux, size, ashift, r)); |
34dc7c2f BB |
752 | |
753 | child = umem_alloc(m * sizeof (nvlist_t *), UMEM_NOFAIL); | |
754 | ||
755 | for (c = 0; c < m; c++) | |
b128c09f | 756 | child[c] = make_vdev_raidz(path, aux, size, ashift, r); |
34dc7c2f BB |
757 | |
758 | VERIFY(nvlist_alloc(&mirror, NV_UNIQUE_NAME, 0) == 0); | |
759 | VERIFY(nvlist_add_string(mirror, ZPOOL_CONFIG_TYPE, | |
760 | VDEV_TYPE_MIRROR) == 0); | |
761 | VERIFY(nvlist_add_nvlist_array(mirror, ZPOOL_CONFIG_CHILDREN, | |
762 | child, m) == 0); | |
34dc7c2f BB |
763 | |
764 | for (c = 0; c < m; c++) | |
765 | nvlist_free(child[c]); | |
766 | ||
767 | umem_free(child, m * sizeof (nvlist_t *)); | |
768 | ||
769 | return (mirror); | |
770 | } | |
771 | ||
772 | static nvlist_t * | |
b128c09f BB |
773 | make_vdev_root(char *path, char *aux, size_t size, uint64_t ashift, |
774 | int log, int r, int m, int t) | |
34dc7c2f BB |
775 | { |
776 | nvlist_t *root, **child; | |
777 | int c; | |
778 | ||
779 | ASSERT(t > 0); | |
780 | ||
781 | child = umem_alloc(t * sizeof (nvlist_t *), UMEM_NOFAIL); | |
782 | ||
b128c09f BB |
783 | for (c = 0; c < t; c++) { |
784 | child[c] = make_vdev_mirror(path, aux, size, ashift, r, m); | |
785 | VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_IS_LOG, | |
786 | log) == 0); | |
787 | } | |
34dc7c2f BB |
788 | |
789 | VERIFY(nvlist_alloc(&root, NV_UNIQUE_NAME, 0) == 0); | |
790 | VERIFY(nvlist_add_string(root, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT) == 0); | |
b128c09f | 791 | VERIFY(nvlist_add_nvlist_array(root, aux ? aux : ZPOOL_CONFIG_CHILDREN, |
34dc7c2f BB |
792 | child, t) == 0); |
793 | ||
794 | for (c = 0; c < t; c++) | |
795 | nvlist_free(child[c]); | |
796 | ||
797 | umem_free(child, t * sizeof (nvlist_t *)); | |
798 | ||
799 | return (root); | |
800 | } | |
801 | ||
428870ff BB |
802 | static int |
803 | ztest_random_blocksize(void) | |
34dc7c2f | 804 | { |
428870ff BB |
805 | return (1 << (SPA_MINBLOCKSHIFT + |
806 | ztest_random(SPA_MAXBLOCKSHIFT - SPA_MINBLOCKSHIFT + 1))); | |
807 | } | |
34dc7c2f | 808 | |
428870ff BB |
809 | static int |
810 | ztest_random_ibshift(void) | |
811 | { | |
812 | return (DN_MIN_INDBLKSHIFT + | |
813 | ztest_random(DN_MAX_INDBLKSHIFT - DN_MIN_INDBLKSHIFT + 1)); | |
34dc7c2f BB |
814 | } |
815 | ||
428870ff BB |
816 | static uint64_t |
817 | ztest_random_vdev_top(spa_t *spa, boolean_t log_ok) | |
34dc7c2f | 818 | { |
428870ff BB |
819 | uint64_t top; |
820 | vdev_t *rvd = spa->spa_root_vdev; | |
821 | vdev_t *tvd; | |
34dc7c2f | 822 | |
428870ff | 823 | ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); |
34dc7c2f | 824 | |
428870ff BB |
825 | do { |
826 | top = ztest_random(rvd->vdev_children); | |
827 | tvd = rvd->vdev_child[top]; | |
828 | } while (tvd->vdev_ishole || (tvd->vdev_islog && !log_ok) || | |
829 | tvd->vdev_mg == NULL || tvd->vdev_mg->mg_class == NULL); | |
34dc7c2f | 830 | |
428870ff | 831 | return (top); |
34dc7c2f BB |
832 | } |
833 | ||
428870ff BB |
834 | static uint64_t |
835 | ztest_random_dsl_prop(zfs_prop_t prop) | |
34dc7c2f | 836 | { |
428870ff BB |
837 | uint64_t value; |
838 | ||
839 | do { | |
840 | value = zfs_prop_random_value(prop, ztest_random(-1ULL)); | |
841 | } while (prop == ZFS_PROP_CHECKSUM && value == ZIO_CHECKSUM_OFF); | |
842 | ||
843 | return (value); | |
34dc7c2f BB |
844 | } |
845 | ||
34dc7c2f | 846 | static int |
428870ff BB |
847 | ztest_dsl_prop_set_uint64(char *osname, zfs_prop_t prop, uint64_t value, |
848 | boolean_t inherit) | |
34dc7c2f | 849 | { |
428870ff BB |
850 | const char *propname = zfs_prop_to_name(prop); |
851 | const char *valname; | |
40b84e7a | 852 | char *setpoint; |
428870ff | 853 | uint64_t curval; |
34dc7c2f BB |
854 | int error; |
855 | ||
428870ff BB |
856 | error = dsl_prop_set(osname, propname, |
857 | (inherit ? ZPROP_SRC_NONE : ZPROP_SRC_LOCAL), | |
858 | sizeof (value), 1, &value); | |
34dc7c2f | 859 | |
428870ff BB |
860 | if (error == ENOSPC) { |
861 | ztest_record_enospc(FTAG); | |
34dc7c2f BB |
862 | return (error); |
863 | } | |
34dc7c2f | 864 | ASSERT3U(error, ==, 0); |
34dc7c2f | 865 | |
40b84e7a | 866 | setpoint = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
428870ff BB |
867 | VERIFY3U(dsl_prop_get(osname, propname, sizeof (curval), |
868 | 1, &curval, setpoint), ==, 0); | |
869 | ||
870 | if (zopt_verbose >= 6) { | |
871 | VERIFY(zfs_prop_index_to_string(prop, curval, &valname) == 0); | |
872 | (void) printf("%s %s = %s at '%s'\n", | |
873 | osname, propname, valname, setpoint); | |
34dc7c2f | 874 | } |
40b84e7a | 875 | umem_free(setpoint, MAXPATHLEN); |
34dc7c2f BB |
876 | |
877 | return (error); | |
878 | } | |
879 | ||
880 | static int | |
428870ff | 881 | ztest_spa_prop_set_uint64(ztest_shared_t *zs, zpool_prop_t prop, uint64_t value) |
34dc7c2f | 882 | { |
428870ff BB |
883 | spa_t *spa = zs->zs_spa; |
884 | nvlist_t *props = NULL; | |
34dc7c2f BB |
885 | int error; |
886 | ||
428870ff BB |
887 | VERIFY(nvlist_alloc(&props, NV_UNIQUE_NAME, 0) == 0); |
888 | VERIFY(nvlist_add_uint64(props, zpool_prop_to_name(prop), value) == 0); | |
34dc7c2f | 889 | |
428870ff BB |
890 | error = spa_prop_set(spa, props); |
891 | ||
892 | nvlist_free(props); | |
893 | ||
894 | if (error == ENOSPC) { | |
895 | ztest_record_enospc(FTAG); | |
34dc7c2f BB |
896 | return (error); |
897 | } | |
428870ff | 898 | ASSERT3U(error, ==, 0); |
34dc7c2f BB |
899 | |
900 | return (error); | |
901 | } | |
902 | ||
428870ff BB |
903 | static void |
904 | ztest_rll_init(rll_t *rll) | |
905 | { | |
906 | rll->rll_writer = NULL; | |
907 | rll->rll_readers = 0; | |
1e33ac1e BB |
908 | mutex_init(&rll->rll_lock, NULL, MUTEX_DEFAULT, NULL); |
909 | cv_init(&rll->rll_cv, NULL, CV_DEFAULT, NULL); | |
428870ff | 910 | } |
34dc7c2f | 911 | |
428870ff BB |
912 | static void |
913 | ztest_rll_destroy(rll_t *rll) | |
34dc7c2f | 914 | { |
428870ff BB |
915 | ASSERT(rll->rll_writer == NULL); |
916 | ASSERT(rll->rll_readers == 0); | |
1e33ac1e BB |
917 | mutex_destroy(&rll->rll_lock); |
918 | cv_destroy(&rll->rll_cv); | |
428870ff | 919 | } |
34dc7c2f | 920 | |
428870ff BB |
921 | static void |
922 | ztest_rll_lock(rll_t *rll, rl_type_t type) | |
923 | { | |
1e33ac1e | 924 | mutex_enter(&rll->rll_lock); |
34dc7c2f | 925 | |
428870ff BB |
926 | if (type == RL_READER) { |
927 | while (rll->rll_writer != NULL) | |
1e33ac1e | 928 | (void) cv_wait(&rll->rll_cv, &rll->rll_lock); |
428870ff BB |
929 | rll->rll_readers++; |
930 | } else { | |
931 | while (rll->rll_writer != NULL || rll->rll_readers) | |
1e33ac1e | 932 | (void) cv_wait(&rll->rll_cv, &rll->rll_lock); |
428870ff BB |
933 | rll->rll_writer = curthread; |
934 | } | |
34dc7c2f | 935 | |
1e33ac1e | 936 | mutex_exit(&rll->rll_lock); |
428870ff | 937 | } |
34dc7c2f | 938 | |
428870ff BB |
939 | static void |
940 | ztest_rll_unlock(rll_t *rll) | |
941 | { | |
1e33ac1e | 942 | mutex_enter(&rll->rll_lock); |
34dc7c2f | 943 | |
428870ff BB |
944 | if (rll->rll_writer) { |
945 | ASSERT(rll->rll_readers == 0); | |
946 | rll->rll_writer = NULL; | |
947 | } else { | |
948 | ASSERT(rll->rll_readers != 0); | |
949 | ASSERT(rll->rll_writer == NULL); | |
950 | rll->rll_readers--; | |
951 | } | |
34dc7c2f | 952 | |
428870ff | 953 | if (rll->rll_writer == NULL && rll->rll_readers == 0) |
1e33ac1e | 954 | cv_broadcast(&rll->rll_cv); |
428870ff | 955 | |
1e33ac1e | 956 | mutex_exit(&rll->rll_lock); |
34dc7c2f BB |
957 | } |
958 | ||
428870ff BB |
959 | static void |
960 | ztest_object_lock(ztest_ds_t *zd, uint64_t object, rl_type_t type) | |
b128c09f | 961 | { |
428870ff | 962 | rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; |
b128c09f | 963 | |
428870ff BB |
964 | ztest_rll_lock(rll, type); |
965 | } | |
b128c09f | 966 | |
428870ff BB |
967 | static void |
968 | ztest_object_unlock(ztest_ds_t *zd, uint64_t object) | |
969 | { | |
970 | rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; | |
b128c09f | 971 | |
428870ff | 972 | ztest_rll_unlock(rll); |
b128c09f BB |
973 | } |
974 | ||
428870ff BB |
975 | static rl_t * |
976 | ztest_range_lock(ztest_ds_t *zd, uint64_t object, uint64_t offset, | |
977 | uint64_t size, rl_type_t type) | |
34dc7c2f | 978 | { |
428870ff BB |
979 | uint64_t hash = object ^ (offset % (ZTEST_RANGE_LOCKS + 1)); |
980 | rll_t *rll = &zd->zd_range_lock[hash & (ZTEST_RANGE_LOCKS - 1)]; | |
981 | rl_t *rl; | |
34dc7c2f | 982 | |
428870ff BB |
983 | rl = umem_alloc(sizeof (*rl), UMEM_NOFAIL); |
984 | rl->rl_object = object; | |
985 | rl->rl_offset = offset; | |
986 | rl->rl_size = size; | |
987 | rl->rl_lock = rll; | |
34dc7c2f | 988 | |
428870ff BB |
989 | ztest_rll_lock(rll, type); |
990 | ||
991 | return (rl); | |
992 | } | |
34dc7c2f | 993 | |
428870ff BB |
994 | static void |
995 | ztest_range_unlock(rl_t *rl) | |
996 | { | |
997 | rll_t *rll = rl->rl_lock; | |
34dc7c2f | 998 | |
428870ff | 999 | ztest_rll_unlock(rll); |
34dc7c2f | 1000 | |
428870ff BB |
1001 | umem_free(rl, sizeof (*rl)); |
1002 | } | |
34dc7c2f | 1003 | |
428870ff BB |
1004 | static void |
1005 | ztest_zd_init(ztest_ds_t *zd, objset_t *os) | |
1006 | { | |
1007 | zd->zd_os = os; | |
1008 | zd->zd_zilog = dmu_objset_zil(os); | |
1009 | zd->zd_seq = 0; | |
1010 | dmu_objset_name(os, zd->zd_name); | |
d6320ddb | 1011 | int l; |
428870ff | 1012 | |
3e31d2b0 | 1013 | rw_init(&zd->zd_zilog_lock, NULL, RW_DEFAULT, NULL); |
1e33ac1e | 1014 | mutex_init(&zd->zd_dirobj_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 1015 | |
d6320ddb | 1016 | for (l = 0; l < ZTEST_OBJECT_LOCKS; l++) |
428870ff | 1017 | ztest_rll_init(&zd->zd_object_lock[l]); |
34dc7c2f | 1018 | |
d6320ddb | 1019 | for (l = 0; l < ZTEST_RANGE_LOCKS; l++) |
428870ff | 1020 | ztest_rll_init(&zd->zd_range_lock[l]); |
34dc7c2f BB |
1021 | } |
1022 | ||
428870ff BB |
1023 | static void |
1024 | ztest_zd_fini(ztest_ds_t *zd) | |
34dc7c2f | 1025 | { |
d6320ddb BB |
1026 | int l; |
1027 | ||
1e33ac1e | 1028 | mutex_destroy(&zd->zd_dirobj_lock); |
3e31d2b0 | 1029 | rw_destroy(&zd->zd_zilog_lock); |
34dc7c2f | 1030 | |
d6320ddb | 1031 | for (l = 0; l < ZTEST_OBJECT_LOCKS; l++) |
428870ff | 1032 | ztest_rll_destroy(&zd->zd_object_lock[l]); |
b128c09f | 1033 | |
d6320ddb | 1034 | for (l = 0; l < ZTEST_RANGE_LOCKS; l++) |
428870ff BB |
1035 | ztest_rll_destroy(&zd->zd_range_lock[l]); |
1036 | } | |
b128c09f | 1037 | |
428870ff | 1038 | #define TXG_MIGHTWAIT (ztest_random(10) == 0 ? TXG_NOWAIT : TXG_WAIT) |
b128c09f | 1039 | |
428870ff BB |
1040 | static uint64_t |
1041 | ztest_tx_assign(dmu_tx_t *tx, uint64_t txg_how, const char *tag) | |
1042 | { | |
1043 | uint64_t txg; | |
1044 | int error; | |
1045 | ||
1046 | /* | |
1047 | * Attempt to assign tx to some transaction group. | |
1048 | */ | |
1049 | error = dmu_tx_assign(tx, txg_how); | |
1050 | if (error) { | |
1051 | if (error == ERESTART) { | |
1052 | ASSERT(txg_how == TXG_NOWAIT); | |
1053 | dmu_tx_wait(tx); | |
1054 | } else { | |
1055 | ASSERT3U(error, ==, ENOSPC); | |
1056 | ztest_record_enospc(tag); | |
1057 | } | |
1058 | dmu_tx_abort(tx); | |
1059 | return (0); | |
1060 | } | |
1061 | txg = dmu_tx_get_txg(tx); | |
1062 | ASSERT(txg != 0); | |
1063 | return (txg); | |
1064 | } | |
1065 | ||
1066 | static void | |
1067 | ztest_pattern_set(void *buf, uint64_t size, uint64_t value) | |
1068 | { | |
1069 | uint64_t *ip = buf; | |
1070 | uint64_t *ip_end = (uint64_t *)((uintptr_t)buf + (uintptr_t)size); | |
1071 | ||
1072 | while (ip < ip_end) | |
1073 | *ip++ = value; | |
1074 | } | |
1075 | ||
1fde1e37 | 1076 | #ifndef NDEBUG |
428870ff BB |
1077 | static boolean_t |
1078 | ztest_pattern_match(void *buf, uint64_t size, uint64_t value) | |
1079 | { | |
1080 | uint64_t *ip = buf; | |
1081 | uint64_t *ip_end = (uint64_t *)((uintptr_t)buf + (uintptr_t)size); | |
1082 | uint64_t diff = 0; | |
1083 | ||
1084 | while (ip < ip_end) | |
1085 | diff |= (value - *ip++); | |
1086 | ||
1087 | return (diff == 0); | |
1088 | } | |
1fde1e37 | 1089 | #endif |
428870ff BB |
1090 | |
1091 | static void | |
1092 | ztest_bt_generate(ztest_block_tag_t *bt, objset_t *os, uint64_t object, | |
1093 | uint64_t offset, uint64_t gen, uint64_t txg, uint64_t crtxg) | |
1094 | { | |
1095 | bt->bt_magic = BT_MAGIC; | |
1096 | bt->bt_objset = dmu_objset_id(os); | |
1097 | bt->bt_object = object; | |
1098 | bt->bt_offset = offset; | |
1099 | bt->bt_gen = gen; | |
1100 | bt->bt_txg = txg; | |
1101 | bt->bt_crtxg = crtxg; | |
1102 | } | |
1103 | ||
1104 | static void | |
1105 | ztest_bt_verify(ztest_block_tag_t *bt, objset_t *os, uint64_t object, | |
1106 | uint64_t offset, uint64_t gen, uint64_t txg, uint64_t crtxg) | |
1107 | { | |
1108 | ASSERT(bt->bt_magic == BT_MAGIC); | |
1109 | ASSERT(bt->bt_objset == dmu_objset_id(os)); | |
1110 | ASSERT(bt->bt_object == object); | |
1111 | ASSERT(bt->bt_offset == offset); | |
1112 | ASSERT(bt->bt_gen <= gen); | |
1113 | ASSERT(bt->bt_txg <= txg); | |
1114 | ASSERT(bt->bt_crtxg == crtxg); | |
1115 | } | |
1116 | ||
1117 | static ztest_block_tag_t * | |
1118 | ztest_bt_bonus(dmu_buf_t *db) | |
1119 | { | |
1120 | dmu_object_info_t doi; | |
1121 | ztest_block_tag_t *bt; | |
1122 | ||
1123 | dmu_object_info_from_db(db, &doi); | |
1124 | ASSERT3U(doi.doi_bonus_size, <=, db->db_size); | |
1125 | ASSERT3U(doi.doi_bonus_size, >=, sizeof (*bt)); | |
1126 | bt = (void *)((char *)db->db_data + doi.doi_bonus_size - sizeof (*bt)); | |
1127 | ||
1128 | return (bt); | |
1129 | } | |
1130 | ||
1131 | /* | |
1132 | * ZIL logging ops | |
1133 | */ | |
1134 | ||
1135 | #define lrz_type lr_mode | |
1136 | #define lrz_blocksize lr_uid | |
1137 | #define lrz_ibshift lr_gid | |
1138 | #define lrz_bonustype lr_rdev | |
1139 | #define lrz_bonuslen lr_crtime[1] | |
1140 | ||
572e2857 | 1141 | static void |
428870ff BB |
1142 | ztest_log_create(ztest_ds_t *zd, dmu_tx_t *tx, lr_create_t *lr) |
1143 | { | |
1144 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1145 | size_t namesize = strlen(name) + 1; | |
1146 | itx_t *itx; | |
1147 | ||
1148 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1149 | return; |
428870ff BB |
1150 | |
1151 | itx = zil_itx_create(TX_CREATE, sizeof (*lr) + namesize); | |
1152 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1153 | sizeof (*lr) + namesize - sizeof (lr_t)); | |
1154 | ||
572e2857 | 1155 | zil_itx_assign(zd->zd_zilog, itx, tx); |
428870ff BB |
1156 | } |
1157 | ||
572e2857 BB |
1158 | static void |
1159 | ztest_log_remove(ztest_ds_t *zd, dmu_tx_t *tx, lr_remove_t *lr, uint64_t object) | |
428870ff BB |
1160 | { |
1161 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1162 | size_t namesize = strlen(name) + 1; | |
1163 | itx_t *itx; | |
1164 | ||
1165 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1166 | return; |
428870ff BB |
1167 | |
1168 | itx = zil_itx_create(TX_REMOVE, sizeof (*lr) + namesize); | |
1169 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1170 | sizeof (*lr) + namesize - sizeof (lr_t)); | |
1171 | ||
572e2857 BB |
1172 | itx->itx_oid = object; |
1173 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1174 | } |
1175 | ||
572e2857 | 1176 | static void |
428870ff BB |
1177 | ztest_log_write(ztest_ds_t *zd, dmu_tx_t *tx, lr_write_t *lr) |
1178 | { | |
1179 | itx_t *itx; | |
1180 | itx_wr_state_t write_state = ztest_random(WR_NUM_STATES); | |
1181 | ||
1182 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1183 | return; |
428870ff BB |
1184 | |
1185 | if (lr->lr_length > ZIL_MAX_LOG_DATA) | |
1186 | write_state = WR_INDIRECT; | |
1187 | ||
1188 | itx = zil_itx_create(TX_WRITE, | |
1189 | sizeof (*lr) + (write_state == WR_COPIED ? lr->lr_length : 0)); | |
1190 | ||
1191 | if (write_state == WR_COPIED && | |
1192 | dmu_read(zd->zd_os, lr->lr_foid, lr->lr_offset, lr->lr_length, | |
1193 | ((lr_write_t *)&itx->itx_lr) + 1, DMU_READ_NO_PREFETCH) != 0) { | |
1194 | zil_itx_destroy(itx); | |
1195 | itx = zil_itx_create(TX_WRITE, sizeof (*lr)); | |
1196 | write_state = WR_NEED_COPY; | |
1197 | } | |
1198 | itx->itx_private = zd; | |
1199 | itx->itx_wr_state = write_state; | |
1200 | itx->itx_sync = (ztest_random(8) == 0); | |
1201 | itx->itx_sod += (write_state == WR_NEED_COPY ? lr->lr_length : 0); | |
1202 | ||
1203 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1204 | sizeof (*lr) - sizeof (lr_t)); | |
1205 | ||
572e2857 | 1206 | zil_itx_assign(zd->zd_zilog, itx, tx); |
428870ff BB |
1207 | } |
1208 | ||
572e2857 | 1209 | static void |
428870ff BB |
1210 | ztest_log_truncate(ztest_ds_t *zd, dmu_tx_t *tx, lr_truncate_t *lr) |
1211 | { | |
1212 | itx_t *itx; | |
1213 | ||
1214 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1215 | return; |
428870ff BB |
1216 | |
1217 | itx = zil_itx_create(TX_TRUNCATE, sizeof (*lr)); | |
1218 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1219 | sizeof (*lr) - sizeof (lr_t)); | |
1220 | ||
572e2857 BB |
1221 | itx->itx_sync = B_FALSE; |
1222 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1223 | } |
1224 | ||
572e2857 | 1225 | static void |
428870ff BB |
1226 | ztest_log_setattr(ztest_ds_t *zd, dmu_tx_t *tx, lr_setattr_t *lr) |
1227 | { | |
1228 | itx_t *itx; | |
1229 | ||
1230 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1231 | return; |
428870ff BB |
1232 | |
1233 | itx = zil_itx_create(TX_SETATTR, sizeof (*lr)); | |
1234 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1235 | sizeof (*lr) - sizeof (lr_t)); | |
1236 | ||
572e2857 BB |
1237 | itx->itx_sync = B_FALSE; |
1238 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1239 | } |
1240 | ||
1241 | /* | |
1242 | * ZIL replay ops | |
1243 | */ | |
1244 | static int | |
1245 | ztest_replay_create(ztest_ds_t *zd, lr_create_t *lr, boolean_t byteswap) | |
1246 | { | |
1247 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1248 | objset_t *os = zd->zd_os; | |
1249 | ztest_block_tag_t *bbt; | |
1250 | dmu_buf_t *db; | |
1251 | dmu_tx_t *tx; | |
1252 | uint64_t txg; | |
1253 | int error = 0; | |
1254 | ||
1255 | if (byteswap) | |
1256 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1257 | ||
1258 | ASSERT(lr->lr_doid == ZTEST_DIROBJ); | |
1259 | ASSERT(name[0] != '\0'); | |
1260 | ||
1261 | tx = dmu_tx_create(os); | |
1262 | ||
1263 | dmu_tx_hold_zap(tx, lr->lr_doid, B_TRUE, name); | |
1264 | ||
1265 | if (lr->lrz_type == DMU_OT_ZAP_OTHER) { | |
1266 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, B_TRUE, NULL); | |
1267 | } else { | |
1268 | dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT); | |
1269 | } | |
1270 | ||
1271 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1272 | if (txg == 0) | |
1273 | return (ENOSPC); | |
1274 | ||
1275 | ASSERT(dmu_objset_zil(os)->zl_replay == !!lr->lr_foid); | |
1276 | ||
1277 | if (lr->lrz_type == DMU_OT_ZAP_OTHER) { | |
1278 | if (lr->lr_foid == 0) { | |
1279 | lr->lr_foid = zap_create(os, | |
1280 | lr->lrz_type, lr->lrz_bonustype, | |
1281 | lr->lrz_bonuslen, tx); | |
1282 | } else { | |
1283 | error = zap_create_claim(os, lr->lr_foid, | |
1284 | lr->lrz_type, lr->lrz_bonustype, | |
1285 | lr->lrz_bonuslen, tx); | |
1286 | } | |
1287 | } else { | |
1288 | if (lr->lr_foid == 0) { | |
1289 | lr->lr_foid = dmu_object_alloc(os, | |
1290 | lr->lrz_type, 0, lr->lrz_bonustype, | |
1291 | lr->lrz_bonuslen, tx); | |
1292 | } else { | |
1293 | error = dmu_object_claim(os, lr->lr_foid, | |
1294 | lr->lrz_type, 0, lr->lrz_bonustype, | |
1295 | lr->lrz_bonuslen, tx); | |
1296 | } | |
1297 | } | |
1298 | ||
1299 | if (error) { | |
1300 | ASSERT3U(error, ==, EEXIST); | |
1301 | ASSERT(zd->zd_zilog->zl_replay); | |
1302 | dmu_tx_commit(tx); | |
1303 | return (error); | |
1304 | } | |
1305 | ||
1306 | ASSERT(lr->lr_foid != 0); | |
1307 | ||
1308 | if (lr->lrz_type != DMU_OT_ZAP_OTHER) | |
1309 | VERIFY3U(0, ==, dmu_object_set_blocksize(os, lr->lr_foid, | |
1310 | lr->lrz_blocksize, lr->lrz_ibshift, tx)); | |
1311 | ||
1312 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
1313 | bbt = ztest_bt_bonus(db); | |
1314 | dmu_buf_will_dirty(db, tx); | |
1315 | ztest_bt_generate(bbt, os, lr->lr_foid, -1ULL, lr->lr_gen, txg, txg); | |
1316 | dmu_buf_rele(db, FTAG); | |
1317 | ||
1318 | VERIFY3U(0, ==, zap_add(os, lr->lr_doid, name, sizeof (uint64_t), 1, | |
1319 | &lr->lr_foid, tx)); | |
1320 | ||
1321 | (void) ztest_log_create(zd, tx, lr); | |
1322 | ||
1323 | dmu_tx_commit(tx); | |
1324 | ||
1325 | return (0); | |
1326 | } | |
1327 | ||
1328 | static int | |
1329 | ztest_replay_remove(ztest_ds_t *zd, lr_remove_t *lr, boolean_t byteswap) | |
1330 | { | |
1331 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1332 | objset_t *os = zd->zd_os; | |
1333 | dmu_object_info_t doi; | |
1334 | dmu_tx_t *tx; | |
1335 | uint64_t object, txg; | |
1336 | ||
1337 | if (byteswap) | |
1338 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1339 | ||
1340 | ASSERT(lr->lr_doid == ZTEST_DIROBJ); | |
1341 | ASSERT(name[0] != '\0'); | |
1342 | ||
1343 | VERIFY3U(0, ==, | |
1344 | zap_lookup(os, lr->lr_doid, name, sizeof (object), 1, &object)); | |
1345 | ASSERT(object != 0); | |
1346 | ||
1347 | ztest_object_lock(zd, object, RL_WRITER); | |
1348 | ||
1349 | VERIFY3U(0, ==, dmu_object_info(os, object, &doi)); | |
1350 | ||
1351 | tx = dmu_tx_create(os); | |
1352 | ||
1353 | dmu_tx_hold_zap(tx, lr->lr_doid, B_FALSE, name); | |
1354 | dmu_tx_hold_free(tx, object, 0, DMU_OBJECT_END); | |
1355 | ||
1356 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1357 | if (txg == 0) { | |
1358 | ztest_object_unlock(zd, object); | |
1359 | return (ENOSPC); | |
1360 | } | |
1361 | ||
1362 | if (doi.doi_type == DMU_OT_ZAP_OTHER) { | |
1363 | VERIFY3U(0, ==, zap_destroy(os, object, tx)); | |
1364 | } else { | |
1365 | VERIFY3U(0, ==, dmu_object_free(os, object, tx)); | |
1366 | } | |
1367 | ||
1368 | VERIFY3U(0, ==, zap_remove(os, lr->lr_doid, name, tx)); | |
1369 | ||
572e2857 | 1370 | (void) ztest_log_remove(zd, tx, lr, object); |
428870ff BB |
1371 | |
1372 | dmu_tx_commit(tx); | |
1373 | ||
1374 | ztest_object_unlock(zd, object); | |
1375 | ||
1376 | return (0); | |
1377 | } | |
1378 | ||
1379 | static int | |
1380 | ztest_replay_write(ztest_ds_t *zd, lr_write_t *lr, boolean_t byteswap) | |
1381 | { | |
1382 | objset_t *os = zd->zd_os; | |
1383 | void *data = lr + 1; /* data follows lr */ | |
1384 | uint64_t offset, length; | |
1385 | ztest_block_tag_t *bt = data; | |
1386 | ztest_block_tag_t *bbt; | |
1387 | uint64_t gen, txg, lrtxg, crtxg; | |
1388 | dmu_object_info_t doi; | |
1389 | dmu_tx_t *tx; | |
1390 | dmu_buf_t *db; | |
1391 | arc_buf_t *abuf = NULL; | |
1392 | rl_t *rl; | |
1393 | ||
1394 | if (byteswap) | |
1395 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1396 | ||
1397 | offset = lr->lr_offset; | |
1398 | length = lr->lr_length; | |
1399 | ||
1400 | /* If it's a dmu_sync() block, write the whole block */ | |
1401 | if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) { | |
1402 | uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr); | |
1403 | if (length < blocksize) { | |
1404 | offset -= offset % blocksize; | |
1405 | length = blocksize; | |
1406 | } | |
1407 | } | |
1408 | ||
1409 | if (bt->bt_magic == BSWAP_64(BT_MAGIC)) | |
1410 | byteswap_uint64_array(bt, sizeof (*bt)); | |
1411 | ||
1412 | if (bt->bt_magic != BT_MAGIC) | |
1413 | bt = NULL; | |
1414 | ||
1415 | ztest_object_lock(zd, lr->lr_foid, RL_READER); | |
1416 | rl = ztest_range_lock(zd, lr->lr_foid, offset, length, RL_WRITER); | |
1417 | ||
1418 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
1419 | ||
1420 | dmu_object_info_from_db(db, &doi); | |
1421 | ||
1422 | bbt = ztest_bt_bonus(db); | |
1423 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
1424 | gen = bbt->bt_gen; | |
1425 | crtxg = bbt->bt_crtxg; | |
1426 | lrtxg = lr->lr_common.lrc_txg; | |
1427 | ||
1428 | tx = dmu_tx_create(os); | |
1429 | ||
1430 | dmu_tx_hold_write(tx, lr->lr_foid, offset, length); | |
1431 | ||
1432 | if (ztest_random(8) == 0 && length == doi.doi_data_block_size && | |
1433 | P2PHASE(offset, length) == 0) | |
1434 | abuf = dmu_request_arcbuf(db, length); | |
1435 | ||
1436 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1437 | if (txg == 0) { | |
1438 | if (abuf != NULL) | |
1439 | dmu_return_arcbuf(abuf); | |
1440 | dmu_buf_rele(db, FTAG); | |
1441 | ztest_range_unlock(rl); | |
1442 | ztest_object_unlock(zd, lr->lr_foid); | |
1443 | return (ENOSPC); | |
1444 | } | |
1445 | ||
1446 | if (bt != NULL) { | |
1447 | /* | |
1448 | * Usually, verify the old data before writing new data -- | |
1449 | * but not always, because we also want to verify correct | |
1450 | * behavior when the data was not recently read into cache. | |
1451 | */ | |
1452 | ASSERT(offset % doi.doi_data_block_size == 0); | |
1453 | if (ztest_random(4) != 0) { | |
1454 | int prefetch = ztest_random(2) ? | |
1455 | DMU_READ_PREFETCH : DMU_READ_NO_PREFETCH; | |
1456 | ztest_block_tag_t rbt; | |
1457 | ||
1458 | VERIFY(dmu_read(os, lr->lr_foid, offset, | |
1459 | sizeof (rbt), &rbt, prefetch) == 0); | |
1460 | if (rbt.bt_magic == BT_MAGIC) { | |
1461 | ztest_bt_verify(&rbt, os, lr->lr_foid, | |
1462 | offset, gen, txg, crtxg); | |
1463 | } | |
1464 | } | |
1465 | ||
1466 | /* | |
1467 | * Writes can appear to be newer than the bonus buffer because | |
1468 | * the ztest_get_data() callback does a dmu_read() of the | |
1469 | * open-context data, which may be different than the data | |
1470 | * as it was when the write was generated. | |
1471 | */ | |
1472 | if (zd->zd_zilog->zl_replay) { | |
1473 | ztest_bt_verify(bt, os, lr->lr_foid, offset, | |
1474 | MAX(gen, bt->bt_gen), MAX(txg, lrtxg), | |
1475 | bt->bt_crtxg); | |
1476 | } | |
1477 | ||
1478 | /* | |
1479 | * Set the bt's gen/txg to the bonus buffer's gen/txg | |
1480 | * so that all of the usual ASSERTs will work. | |
1481 | */ | |
1482 | ztest_bt_generate(bt, os, lr->lr_foid, offset, gen, txg, crtxg); | |
1483 | } | |
1484 | ||
1485 | if (abuf == NULL) { | |
1486 | dmu_write(os, lr->lr_foid, offset, length, data, tx); | |
1487 | } else { | |
1488 | bcopy(data, abuf->b_data, length); | |
1489 | dmu_assign_arcbuf(db, offset, abuf, tx); | |
1490 | } | |
1491 | ||
1492 | (void) ztest_log_write(zd, tx, lr); | |
1493 | ||
1494 | dmu_buf_rele(db, FTAG); | |
1495 | ||
1496 | dmu_tx_commit(tx); | |
1497 | ||
1498 | ztest_range_unlock(rl); | |
1499 | ztest_object_unlock(zd, lr->lr_foid); | |
1500 | ||
1501 | return (0); | |
1502 | } | |
1503 | ||
1504 | static int | |
1505 | ztest_replay_truncate(ztest_ds_t *zd, lr_truncate_t *lr, boolean_t byteswap) | |
1506 | { | |
1507 | objset_t *os = zd->zd_os; | |
1508 | dmu_tx_t *tx; | |
1509 | uint64_t txg; | |
1510 | rl_t *rl; | |
1511 | ||
1512 | if (byteswap) | |
1513 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1514 | ||
1515 | ztest_object_lock(zd, lr->lr_foid, RL_READER); | |
1516 | rl = ztest_range_lock(zd, lr->lr_foid, lr->lr_offset, lr->lr_length, | |
1517 | RL_WRITER); | |
1518 | ||
1519 | tx = dmu_tx_create(os); | |
1520 | ||
1521 | dmu_tx_hold_free(tx, lr->lr_foid, lr->lr_offset, lr->lr_length); | |
1522 | ||
1523 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1524 | if (txg == 0) { | |
1525 | ztest_range_unlock(rl); | |
1526 | ztest_object_unlock(zd, lr->lr_foid); | |
1527 | return (ENOSPC); | |
1528 | } | |
1529 | ||
1530 | VERIFY(dmu_free_range(os, lr->lr_foid, lr->lr_offset, | |
1531 | lr->lr_length, tx) == 0); | |
1532 | ||
1533 | (void) ztest_log_truncate(zd, tx, lr); | |
1534 | ||
1535 | dmu_tx_commit(tx); | |
1536 | ||
1537 | ztest_range_unlock(rl); | |
1538 | ztest_object_unlock(zd, lr->lr_foid); | |
1539 | ||
1540 | return (0); | |
1541 | } | |
1542 | ||
1543 | static int | |
1544 | ztest_replay_setattr(ztest_ds_t *zd, lr_setattr_t *lr, boolean_t byteswap) | |
1545 | { | |
1546 | objset_t *os = zd->zd_os; | |
1547 | dmu_tx_t *tx; | |
1548 | dmu_buf_t *db; | |
1549 | ztest_block_tag_t *bbt; | |
1550 | uint64_t txg, lrtxg, crtxg; | |
1551 | ||
1552 | if (byteswap) | |
1553 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1554 | ||
1555 | ztest_object_lock(zd, lr->lr_foid, RL_WRITER); | |
1556 | ||
1557 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
1558 | ||
1559 | tx = dmu_tx_create(os); | |
1560 | dmu_tx_hold_bonus(tx, lr->lr_foid); | |
1561 | ||
1562 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1563 | if (txg == 0) { | |
1564 | dmu_buf_rele(db, FTAG); | |
1565 | ztest_object_unlock(zd, lr->lr_foid); | |
1566 | return (ENOSPC); | |
1567 | } | |
1568 | ||
1569 | bbt = ztest_bt_bonus(db); | |
1570 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
1571 | crtxg = bbt->bt_crtxg; | |
1572 | lrtxg = lr->lr_common.lrc_txg; | |
1573 | ||
1574 | if (zd->zd_zilog->zl_replay) { | |
1575 | ASSERT(lr->lr_size != 0); | |
1576 | ASSERT(lr->lr_mode != 0); | |
1577 | ASSERT(lrtxg != 0); | |
1578 | } else { | |
1579 | /* | |
1580 | * Randomly change the size and increment the generation. | |
1581 | */ | |
1582 | lr->lr_size = (ztest_random(db->db_size / sizeof (*bbt)) + 1) * | |
1583 | sizeof (*bbt); | |
1584 | lr->lr_mode = bbt->bt_gen + 1; | |
1585 | ASSERT(lrtxg == 0); | |
1586 | } | |
1587 | ||
1588 | /* | |
1589 | * Verify that the current bonus buffer is not newer than our txg. | |
1590 | */ | |
1591 | ztest_bt_verify(bbt, os, lr->lr_foid, -1ULL, lr->lr_mode, | |
1592 | MAX(txg, lrtxg), crtxg); | |
1593 | ||
1594 | dmu_buf_will_dirty(db, tx); | |
1595 | ||
1596 | ASSERT3U(lr->lr_size, >=, sizeof (*bbt)); | |
1597 | ASSERT3U(lr->lr_size, <=, db->db_size); | |
1598 | VERIFY3U(dmu_set_bonus(db, lr->lr_size, tx), ==, 0); | |
1599 | bbt = ztest_bt_bonus(db); | |
1600 | ||
1601 | ztest_bt_generate(bbt, os, lr->lr_foid, -1ULL, lr->lr_mode, txg, crtxg); | |
1602 | ||
1603 | dmu_buf_rele(db, FTAG); | |
1604 | ||
1605 | (void) ztest_log_setattr(zd, tx, lr); | |
1606 | ||
1607 | dmu_tx_commit(tx); | |
1608 | ||
1609 | ztest_object_unlock(zd, lr->lr_foid); | |
1610 | ||
1611 | return (0); | |
1612 | } | |
1613 | ||
1614 | zil_replay_func_t *ztest_replay_vector[TX_MAX_TYPE] = { | |
0bc8fd78 BB |
1615 | NULL, /* 0 no such transaction type */ |
1616 | (zil_replay_func_t *)ztest_replay_create, /* TX_CREATE */ | |
1617 | NULL, /* TX_MKDIR */ | |
1618 | NULL, /* TX_MKXATTR */ | |
1619 | NULL, /* TX_SYMLINK */ | |
1620 | (zil_replay_func_t *)ztest_replay_remove, /* TX_REMOVE */ | |
1621 | NULL, /* TX_RMDIR */ | |
1622 | NULL, /* TX_LINK */ | |
1623 | NULL, /* TX_RENAME */ | |
1624 | (zil_replay_func_t *)ztest_replay_write, /* TX_WRITE */ | |
1625 | (zil_replay_func_t *)ztest_replay_truncate, /* TX_TRUNCATE */ | |
1626 | (zil_replay_func_t *)ztest_replay_setattr, /* TX_SETATTR */ | |
1627 | NULL, /* TX_ACL */ | |
1628 | NULL, /* TX_CREATE_ACL */ | |
1629 | NULL, /* TX_CREATE_ATTR */ | |
1630 | NULL, /* TX_CREATE_ACL_ATTR */ | |
1631 | NULL, /* TX_MKDIR_ACL */ | |
1632 | NULL, /* TX_MKDIR_ATTR */ | |
1633 | NULL, /* TX_MKDIR_ACL_ATTR */ | |
1634 | NULL, /* TX_WRITE2 */ | |
428870ff BB |
1635 | }; |
1636 | ||
1637 | /* | |
1638 | * ZIL get_data callbacks | |
1639 | */ | |
1640 | ||
1641 | static void | |
1642 | ztest_get_done(zgd_t *zgd, int error) | |
1643 | { | |
1644 | ztest_ds_t *zd = zgd->zgd_private; | |
1645 | uint64_t object = zgd->zgd_rl->rl_object; | |
1646 | ||
1647 | if (zgd->zgd_db) | |
1648 | dmu_buf_rele(zgd->zgd_db, zgd); | |
1649 | ||
1650 | ztest_range_unlock(zgd->zgd_rl); | |
1651 | ztest_object_unlock(zd, object); | |
1652 | ||
1653 | if (error == 0 && zgd->zgd_bp) | |
1654 | zil_add_block(zgd->zgd_zilog, zgd->zgd_bp); | |
1655 | ||
1656 | umem_free(zgd, sizeof (*zgd)); | |
1657 | } | |
1658 | ||
1659 | static int | |
1660 | ztest_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio) | |
1661 | { | |
1662 | ztest_ds_t *zd = arg; | |
1663 | objset_t *os = zd->zd_os; | |
1664 | uint64_t object = lr->lr_foid; | |
1665 | uint64_t offset = lr->lr_offset; | |
1666 | uint64_t size = lr->lr_length; | |
1667 | blkptr_t *bp = &lr->lr_blkptr; | |
1668 | uint64_t txg = lr->lr_common.lrc_txg; | |
1669 | uint64_t crtxg; | |
1670 | dmu_object_info_t doi; | |
1671 | dmu_buf_t *db; | |
1672 | zgd_t *zgd; | |
1673 | int error; | |
1674 | ||
1675 | ztest_object_lock(zd, object, RL_READER); | |
1676 | error = dmu_bonus_hold(os, object, FTAG, &db); | |
1677 | if (error) { | |
1678 | ztest_object_unlock(zd, object); | |
1679 | return (error); | |
1680 | } | |
1681 | ||
1682 | crtxg = ztest_bt_bonus(db)->bt_crtxg; | |
1683 | ||
1684 | if (crtxg == 0 || crtxg > txg) { | |
1685 | dmu_buf_rele(db, FTAG); | |
1686 | ztest_object_unlock(zd, object); | |
1687 | return (ENOENT); | |
1688 | } | |
1689 | ||
1690 | dmu_object_info_from_db(db, &doi); | |
1691 | dmu_buf_rele(db, FTAG); | |
1692 | db = NULL; | |
1693 | ||
1694 | zgd = umem_zalloc(sizeof (*zgd), UMEM_NOFAIL); | |
1695 | zgd->zgd_zilog = zd->zd_zilog; | |
1696 | zgd->zgd_private = zd; | |
1697 | ||
1698 | if (buf != NULL) { /* immediate write */ | |
1699 | zgd->zgd_rl = ztest_range_lock(zd, object, offset, size, | |
1700 | RL_READER); | |
1701 | ||
1702 | error = dmu_read(os, object, offset, size, buf, | |
1703 | DMU_READ_NO_PREFETCH); | |
1704 | ASSERT(error == 0); | |
1705 | } else { | |
1706 | size = doi.doi_data_block_size; | |
1707 | if (ISP2(size)) { | |
1708 | offset = P2ALIGN(offset, size); | |
1709 | } else { | |
1710 | ASSERT(offset < size); | |
1711 | offset = 0; | |
1712 | } | |
1713 | ||
1714 | zgd->zgd_rl = ztest_range_lock(zd, object, offset, size, | |
1715 | RL_READER); | |
1716 | ||
1717 | error = dmu_buf_hold(os, object, offset, zgd, &db, | |
1718 | DMU_READ_NO_PREFETCH); | |
1719 | ||
1720 | if (error == 0) { | |
1721 | zgd->zgd_db = db; | |
1722 | zgd->zgd_bp = bp; | |
1723 | ||
1724 | ASSERT(db->db_offset == offset); | |
1725 | ASSERT(db->db_size == size); | |
1726 | ||
1727 | error = dmu_sync(zio, lr->lr_common.lrc_txg, | |
1728 | ztest_get_done, zgd); | |
1729 | ||
1730 | if (error == 0) | |
1731 | return (0); | |
1732 | } | |
1733 | } | |
1734 | ||
1735 | ztest_get_done(zgd, error); | |
1736 | ||
1737 | return (error); | |
1738 | } | |
1739 | ||
1740 | static void * | |
1741 | ztest_lr_alloc(size_t lrsize, char *name) | |
1742 | { | |
1743 | char *lr; | |
1744 | size_t namesize = name ? strlen(name) + 1 : 0; | |
1745 | ||
1746 | lr = umem_zalloc(lrsize + namesize, UMEM_NOFAIL); | |
1747 | ||
1748 | if (name) | |
1749 | bcopy(name, lr + lrsize, namesize); | |
1750 | ||
1751 | return (lr); | |
1752 | } | |
1753 | ||
1754 | void | |
1755 | ztest_lr_free(void *lr, size_t lrsize, char *name) | |
1756 | { | |
1757 | size_t namesize = name ? strlen(name) + 1 : 0; | |
1758 | ||
1759 | umem_free(lr, lrsize + namesize); | |
1760 | } | |
1761 | ||
1762 | /* | |
1763 | * Lookup a bunch of objects. Returns the number of objects not found. | |
1764 | */ | |
1765 | static int | |
1766 | ztest_lookup(ztest_ds_t *zd, ztest_od_t *od, int count) | |
1767 | { | |
1768 | int missing = 0; | |
1769 | int error; | |
d6320ddb | 1770 | int i; |
428870ff | 1771 | |
1e33ac1e | 1772 | ASSERT(mutex_held(&zd->zd_dirobj_lock)); |
428870ff | 1773 | |
d6320ddb | 1774 | for (i = 0; i < count; i++, od++) { |
428870ff BB |
1775 | od->od_object = 0; |
1776 | error = zap_lookup(zd->zd_os, od->od_dir, od->od_name, | |
1777 | sizeof (uint64_t), 1, &od->od_object); | |
1778 | if (error) { | |
1779 | ASSERT(error == ENOENT); | |
1780 | ASSERT(od->od_object == 0); | |
1781 | missing++; | |
1782 | } else { | |
1783 | dmu_buf_t *db; | |
1784 | ztest_block_tag_t *bbt; | |
1785 | dmu_object_info_t doi; | |
1786 | ||
1787 | ASSERT(od->od_object != 0); | |
1788 | ASSERT(missing == 0); /* there should be no gaps */ | |
1789 | ||
1790 | ztest_object_lock(zd, od->od_object, RL_READER); | |
1791 | VERIFY3U(0, ==, dmu_bonus_hold(zd->zd_os, | |
1792 | od->od_object, FTAG, &db)); | |
1793 | dmu_object_info_from_db(db, &doi); | |
1794 | bbt = ztest_bt_bonus(db); | |
1795 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
1796 | od->od_type = doi.doi_type; | |
1797 | od->od_blocksize = doi.doi_data_block_size; | |
1798 | od->od_gen = bbt->bt_gen; | |
1799 | dmu_buf_rele(db, FTAG); | |
1800 | ztest_object_unlock(zd, od->od_object); | |
1801 | } | |
1802 | } | |
1803 | ||
1804 | return (missing); | |
1805 | } | |
1806 | ||
1807 | static int | |
1808 | ztest_create(ztest_ds_t *zd, ztest_od_t *od, int count) | |
1809 | { | |
1810 | int missing = 0; | |
d6320ddb | 1811 | int i; |
428870ff | 1812 | |
1e33ac1e | 1813 | ASSERT(mutex_held(&zd->zd_dirobj_lock)); |
428870ff | 1814 | |
d6320ddb | 1815 | for (i = 0; i < count; i++, od++) { |
428870ff BB |
1816 | if (missing) { |
1817 | od->od_object = 0; | |
1818 | missing++; | |
1819 | continue; | |
1820 | } | |
1821 | ||
1822 | lr_create_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name); | |
1823 | ||
1824 | lr->lr_doid = od->od_dir; | |
1825 | lr->lr_foid = 0; /* 0 to allocate, > 0 to claim */ | |
1826 | lr->lrz_type = od->od_crtype; | |
1827 | lr->lrz_blocksize = od->od_crblocksize; | |
1828 | lr->lrz_ibshift = ztest_random_ibshift(); | |
1829 | lr->lrz_bonustype = DMU_OT_UINT64_OTHER; | |
1830 | lr->lrz_bonuslen = dmu_bonus_max(); | |
1831 | lr->lr_gen = od->od_crgen; | |
1832 | lr->lr_crtime[0] = time(NULL); | |
1833 | ||
1834 | if (ztest_replay_create(zd, lr, B_FALSE) != 0) { | |
1835 | ASSERT(missing == 0); | |
1836 | od->od_object = 0; | |
1837 | missing++; | |
1838 | } else { | |
1839 | od->od_object = lr->lr_foid; | |
1840 | od->od_type = od->od_crtype; | |
1841 | od->od_blocksize = od->od_crblocksize; | |
1842 | od->od_gen = od->od_crgen; | |
1843 | ASSERT(od->od_object != 0); | |
1844 | } | |
1845 | ||
1846 | ztest_lr_free(lr, sizeof (*lr), od->od_name); | |
1847 | } | |
1848 | ||
1849 | return (missing); | |
1850 | } | |
1851 | ||
1852 | static int | |
1853 | ztest_remove(ztest_ds_t *zd, ztest_od_t *od, int count) | |
1854 | { | |
1855 | int missing = 0; | |
1856 | int error; | |
d6320ddb | 1857 | int i; |
428870ff | 1858 | |
1e33ac1e | 1859 | ASSERT(mutex_held(&zd->zd_dirobj_lock)); |
428870ff BB |
1860 | |
1861 | od += count - 1; | |
1862 | ||
d6320ddb | 1863 | for (i = count - 1; i >= 0; i--, od--) { |
428870ff BB |
1864 | if (missing) { |
1865 | missing++; | |
1866 | continue; | |
1867 | } | |
1868 | ||
1869 | if (od->od_object == 0) | |
1870 | continue; | |
1871 | ||
1872 | lr_remove_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name); | |
1873 | ||
1874 | lr->lr_doid = od->od_dir; | |
1875 | ||
1876 | if ((error = ztest_replay_remove(zd, lr, B_FALSE)) != 0) { | |
1877 | ASSERT3U(error, ==, ENOSPC); | |
1878 | missing++; | |
1879 | } else { | |
1880 | od->od_object = 0; | |
1881 | } | |
1882 | ztest_lr_free(lr, sizeof (*lr), od->od_name); | |
1883 | } | |
1884 | ||
1885 | return (missing); | |
1886 | } | |
1887 | ||
1888 | static int | |
1889 | ztest_write(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size, | |
1890 | void *data) | |
1891 | { | |
1892 | lr_write_t *lr; | |
1893 | int error; | |
1894 | ||
1895 | lr = ztest_lr_alloc(sizeof (*lr) + size, NULL); | |
1896 | ||
1897 | lr->lr_foid = object; | |
1898 | lr->lr_offset = offset; | |
1899 | lr->lr_length = size; | |
1900 | lr->lr_blkoff = 0; | |
1901 | BP_ZERO(&lr->lr_blkptr); | |
1902 | ||
1903 | bcopy(data, lr + 1, size); | |
1904 | ||
1905 | error = ztest_replay_write(zd, lr, B_FALSE); | |
1906 | ||
1907 | ztest_lr_free(lr, sizeof (*lr) + size, NULL); | |
1908 | ||
1909 | return (error); | |
1910 | } | |
1911 | ||
1912 | static int | |
1913 | ztest_truncate(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size) | |
1914 | { | |
1915 | lr_truncate_t *lr; | |
1916 | int error; | |
1917 | ||
1918 | lr = ztest_lr_alloc(sizeof (*lr), NULL); | |
1919 | ||
1920 | lr->lr_foid = object; | |
1921 | lr->lr_offset = offset; | |
1922 | lr->lr_length = size; | |
1923 | ||
1924 | error = ztest_replay_truncate(zd, lr, B_FALSE); | |
1925 | ||
1926 | ztest_lr_free(lr, sizeof (*lr), NULL); | |
1927 | ||
1928 | return (error); | |
1929 | } | |
1930 | ||
1931 | static int | |
1932 | ztest_setattr(ztest_ds_t *zd, uint64_t object) | |
1933 | { | |
1934 | lr_setattr_t *lr; | |
1935 | int error; | |
1936 | ||
1937 | lr = ztest_lr_alloc(sizeof (*lr), NULL); | |
1938 | ||
1939 | lr->lr_foid = object; | |
1940 | lr->lr_size = 0; | |
1941 | lr->lr_mode = 0; | |
1942 | ||
1943 | error = ztest_replay_setattr(zd, lr, B_FALSE); | |
1944 | ||
1945 | ztest_lr_free(lr, sizeof (*lr), NULL); | |
1946 | ||
1947 | return (error); | |
1948 | } | |
1949 | ||
1950 | static void | |
1951 | ztest_prealloc(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size) | |
1952 | { | |
1953 | objset_t *os = zd->zd_os; | |
1954 | dmu_tx_t *tx; | |
1955 | uint64_t txg; | |
1956 | rl_t *rl; | |
1957 | ||
1958 | txg_wait_synced(dmu_objset_pool(os), 0); | |
1959 | ||
1960 | ztest_object_lock(zd, object, RL_READER); | |
1961 | rl = ztest_range_lock(zd, object, offset, size, RL_WRITER); | |
1962 | ||
1963 | tx = dmu_tx_create(os); | |
1964 | ||
1965 | dmu_tx_hold_write(tx, object, offset, size); | |
1966 | ||
1967 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1968 | ||
1969 | if (txg != 0) { | |
1970 | dmu_prealloc(os, object, offset, size, tx); | |
1971 | dmu_tx_commit(tx); | |
1972 | txg_wait_synced(dmu_objset_pool(os), txg); | |
1973 | } else { | |
1974 | (void) dmu_free_long_range(os, object, offset, size); | |
1975 | } | |
1976 | ||
1977 | ztest_range_unlock(rl); | |
1978 | ztest_object_unlock(zd, object); | |
1979 | } | |
1980 | ||
1981 | static void | |
1982 | ztest_io(ztest_ds_t *zd, uint64_t object, uint64_t offset) | |
1983 | { | |
1984 | ztest_block_tag_t wbt; | |
1985 | dmu_object_info_t doi; | |
1986 | enum ztest_io_type io_type; | |
1987 | uint64_t blocksize; | |
1988 | void *data; | |
1989 | ||
1990 | VERIFY(dmu_object_info(zd->zd_os, object, &doi) == 0); | |
1991 | blocksize = doi.doi_data_block_size; | |
1992 | data = umem_alloc(blocksize, UMEM_NOFAIL); | |
1993 | ||
1994 | /* | |
1995 | * Pick an i/o type at random, biased toward writing block tags. | |
1996 | */ | |
1997 | io_type = ztest_random(ZTEST_IO_TYPES); | |
1998 | if (ztest_random(2) == 0) | |
1999 | io_type = ZTEST_IO_WRITE_TAG; | |
2000 | ||
3e31d2b0 ES |
2001 | (void) rw_enter(&zd->zd_zilog_lock, RW_READER); |
2002 | ||
428870ff BB |
2003 | switch (io_type) { |
2004 | ||
2005 | case ZTEST_IO_WRITE_TAG: | |
2006 | ztest_bt_generate(&wbt, zd->zd_os, object, offset, 0, 0, 0); | |
2007 | (void) ztest_write(zd, object, offset, sizeof (wbt), &wbt); | |
2008 | break; | |
2009 | ||
2010 | case ZTEST_IO_WRITE_PATTERN: | |
2011 | (void) memset(data, 'a' + (object + offset) % 5, blocksize); | |
2012 | if (ztest_random(2) == 0) { | |
2013 | /* | |
2014 | * Induce fletcher2 collisions to ensure that | |
2015 | * zio_ddt_collision() detects and resolves them | |
2016 | * when using fletcher2-verify for deduplication. | |
2017 | */ | |
2018 | ((uint64_t *)data)[0] ^= 1ULL << 63; | |
2019 | ((uint64_t *)data)[4] ^= 1ULL << 63; | |
2020 | } | |
2021 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2022 | break; | |
2023 | ||
2024 | case ZTEST_IO_WRITE_ZEROES: | |
2025 | bzero(data, blocksize); | |
2026 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2027 | break; | |
2028 | ||
2029 | case ZTEST_IO_TRUNCATE: | |
2030 | (void) ztest_truncate(zd, object, offset, blocksize); | |
2031 | break; | |
2032 | ||
2033 | case ZTEST_IO_SETATTR: | |
2034 | (void) ztest_setattr(zd, object); | |
2035 | break; | |
e75c13c3 BB |
2036 | default: |
2037 | break; | |
428870ff BB |
2038 | } |
2039 | ||
3e31d2b0 ES |
2040 | (void) rw_exit(&zd->zd_zilog_lock); |
2041 | ||
428870ff BB |
2042 | umem_free(data, blocksize); |
2043 | } | |
2044 | ||
2045 | /* | |
2046 | * Initialize an object description template. | |
2047 | */ | |
2048 | static void | |
2049 | ztest_od_init(ztest_od_t *od, uint64_t id, char *tag, uint64_t index, | |
2050 | dmu_object_type_t type, uint64_t blocksize, uint64_t gen) | |
2051 | { | |
2052 | od->od_dir = ZTEST_DIROBJ; | |
2053 | od->od_object = 0; | |
2054 | ||
2055 | od->od_crtype = type; | |
2056 | od->od_crblocksize = blocksize ? blocksize : ztest_random_blocksize(); | |
2057 | od->od_crgen = gen; | |
2058 | ||
2059 | od->od_type = DMU_OT_NONE; | |
2060 | od->od_blocksize = 0; | |
2061 | od->od_gen = 0; | |
2062 | ||
2063 | (void) snprintf(od->od_name, sizeof (od->od_name), "%s(%lld)[%llu]", | |
b8864a23 | 2064 | tag, (longlong_t)id, (u_longlong_t)index); |
428870ff BB |
2065 | } |
2066 | ||
2067 | /* | |
2068 | * Lookup or create the objects for a test using the od template. | |
2069 | * If the objects do not all exist, or if 'remove' is specified, | |
2070 | * remove any existing objects and create new ones. Otherwise, | |
2071 | * use the existing objects. | |
2072 | */ | |
2073 | static int | |
2074 | ztest_object_init(ztest_ds_t *zd, ztest_od_t *od, size_t size, boolean_t remove) | |
2075 | { | |
2076 | int count = size / sizeof (*od); | |
2077 | int rv = 0; | |
2078 | ||
1e33ac1e | 2079 | mutex_enter(&zd->zd_dirobj_lock); |
428870ff BB |
2080 | if ((ztest_lookup(zd, od, count) != 0 || remove) && |
2081 | (ztest_remove(zd, od, count) != 0 || | |
2082 | ztest_create(zd, od, count) != 0)) | |
2083 | rv = -1; | |
2084 | zd->zd_od = od; | |
1e33ac1e | 2085 | mutex_exit(&zd->zd_dirobj_lock); |
428870ff BB |
2086 | |
2087 | return (rv); | |
2088 | } | |
2089 | ||
2090 | /* ARGSUSED */ | |
2091 | void | |
2092 | ztest_zil_commit(ztest_ds_t *zd, uint64_t id) | |
2093 | { | |
2094 | zilog_t *zilog = zd->zd_zilog; | |
2095 | ||
3e31d2b0 ES |
2096 | (void) rw_enter(&zd->zd_zilog_lock, RW_READER); |
2097 | ||
572e2857 | 2098 | zil_commit(zilog, ztest_random(ZTEST_OBJECTS)); |
428870ff BB |
2099 | |
2100 | /* | |
2101 | * Remember the committed values in zd, which is in parent/child | |
2102 | * shared memory. If we die, the next iteration of ztest_run() | |
2103 | * will verify that the log really does contain this record. | |
2104 | */ | |
2105 | mutex_enter(&zilog->zl_lock); | |
2106 | ASSERT(zd->zd_seq <= zilog->zl_commit_lr_seq); | |
2107 | zd->zd_seq = zilog->zl_commit_lr_seq; | |
2108 | mutex_exit(&zilog->zl_lock); | |
3e31d2b0 ES |
2109 | |
2110 | (void) rw_exit(&zd->zd_zilog_lock); | |
2111 | } | |
2112 | ||
2113 | /* | |
2114 | * This function is designed to simulate the operations that occur during a | |
2115 | * mount/unmount operation. We hold the dataset across these operations in an | |
2116 | * attempt to expose any implicit assumptions about ZIL management. | |
2117 | */ | |
2118 | /* ARGSUSED */ | |
2119 | void | |
2120 | ztest_zil_remount(ztest_ds_t *zd, uint64_t id) | |
2121 | { | |
2122 | objset_t *os = zd->zd_os; | |
2123 | ||
2124 | (void) rw_enter(&zd->zd_zilog_lock, RW_WRITER); | |
2125 | ||
2126 | /* zfsvfs_teardown() */ | |
2127 | zil_close(zd->zd_zilog); | |
2128 | ||
2129 | /* zfsvfs_setup() */ | |
2130 | VERIFY(zil_open(os, ztest_get_data) == zd->zd_zilog); | |
2131 | zil_replay(os, zd, ztest_replay_vector); | |
2132 | ||
2133 | (void) rw_exit(&zd->zd_zilog_lock); | |
428870ff BB |
2134 | } |
2135 | ||
2136 | /* | |
2137 | * Verify that we can't destroy an active pool, create an existing pool, | |
2138 | * or create a pool with a bad vdev spec. | |
2139 | */ | |
2140 | /* ARGSUSED */ | |
2141 | void | |
2142 | ztest_spa_create_destroy(ztest_ds_t *zd, uint64_t id) | |
2143 | { | |
2144 | ztest_shared_t *zs = ztest_shared; | |
2145 | spa_t *spa; | |
2146 | nvlist_t *nvroot; | |
2147 | ||
2148 | /* | |
2149 | * Attempt to create using a bad file. | |
2150 | */ | |
2151 | nvroot = make_vdev_root("/dev/bogus", NULL, 0, 0, 0, 0, 0, 1); | |
2152 | VERIFY3U(ENOENT, ==, | |
2153 | spa_create("ztest_bad_file", nvroot, NULL, NULL, NULL)); | |
2154 | nvlist_free(nvroot); | |
2155 | ||
2156 | /* | |
2157 | * Attempt to create using a bad mirror. | |
2158 | */ | |
2159 | nvroot = make_vdev_root("/dev/bogus", NULL, 0, 0, 0, 0, 2, 1); | |
2160 | VERIFY3U(ENOENT, ==, | |
2161 | spa_create("ztest_bad_mirror", nvroot, NULL, NULL, NULL)); | |
2162 | nvlist_free(nvroot); | |
2163 | ||
2164 | /* | |
2165 | * Attempt to create an existing pool. It shouldn't matter | |
2166 | * what's in the nvroot; we should fail with EEXIST. | |
2167 | */ | |
1e33ac1e | 2168 | (void) rw_enter(&zs->zs_name_lock, RW_READER); |
428870ff BB |
2169 | nvroot = make_vdev_root("/dev/bogus", NULL, 0, 0, 0, 0, 0, 1); |
2170 | VERIFY3U(EEXIST, ==, spa_create(zs->zs_pool, nvroot, NULL, NULL, NULL)); | |
2171 | nvlist_free(nvroot); | |
2172 | VERIFY3U(0, ==, spa_open(zs->zs_pool, &spa, FTAG)); | |
2173 | VERIFY3U(EBUSY, ==, spa_destroy(zs->zs_pool)); | |
2174 | spa_close(spa, FTAG); | |
2175 | ||
1e33ac1e | 2176 | (void) rw_exit(&zs->zs_name_lock); |
428870ff BB |
2177 | } |
2178 | ||
2179 | static vdev_t * | |
2180 | vdev_lookup_by_path(vdev_t *vd, const char *path) | |
2181 | { | |
2182 | vdev_t *mvd; | |
d6320ddb | 2183 | int c; |
428870ff BB |
2184 | |
2185 | if (vd->vdev_path != NULL && strcmp(path, vd->vdev_path) == 0) | |
2186 | return (vd); | |
2187 | ||
d6320ddb | 2188 | for (c = 0; c < vd->vdev_children; c++) |
428870ff BB |
2189 | if ((mvd = vdev_lookup_by_path(vd->vdev_child[c], path)) != |
2190 | NULL) | |
2191 | return (mvd); | |
2192 | ||
2193 | return (NULL); | |
2194 | } | |
2195 | ||
2196 | /* | |
2197 | * Find the first available hole which can be used as a top-level. | |
2198 | */ | |
2199 | int | |
2200 | find_vdev_hole(spa_t *spa) | |
2201 | { | |
2202 | vdev_t *rvd = spa->spa_root_vdev; | |
2203 | int c; | |
2204 | ||
2205 | ASSERT(spa_config_held(spa, SCL_VDEV, RW_READER) == SCL_VDEV); | |
2206 | ||
2207 | for (c = 0; c < rvd->vdev_children; c++) { | |
2208 | vdev_t *cvd = rvd->vdev_child[c]; | |
2209 | ||
2210 | if (cvd->vdev_ishole) | |
2211 | break; | |
2212 | } | |
2213 | return (c); | |
2214 | } | |
2215 | ||
2216 | /* | |
2217 | * Verify that vdev_add() works as expected. | |
2218 | */ | |
2219 | /* ARGSUSED */ | |
2220 | void | |
2221 | ztest_vdev_add_remove(ztest_ds_t *zd, uint64_t id) | |
2222 | { | |
2223 | ztest_shared_t *zs = ztest_shared; | |
2224 | spa_t *spa = zs->zs_spa; | |
2225 | uint64_t leaves; | |
2226 | uint64_t guid; | |
2227 | nvlist_t *nvroot; | |
2228 | int error; | |
2229 | ||
1e33ac1e | 2230 | mutex_enter(&zs->zs_vdev_lock); |
428870ff BB |
2231 | leaves = MAX(zs->zs_mirrors + zs->zs_splits, 1) * zopt_raidz; |
2232 | ||
2233 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
2234 | ||
2235 | ztest_shared->zs_vdev_next_leaf = find_vdev_hole(spa) * leaves; | |
2236 | ||
2237 | /* | |
2238 | * If we have slogs then remove them 1/4 of the time. | |
2239 | */ | |
2240 | if (spa_has_slogs(spa) && ztest_random(4) == 0) { | |
2241 | /* | |
2242 | * Grab the guid from the head of the log class rotor. | |
2243 | */ | |
2244 | guid = spa_log_class(spa)->mc_rotor->mg_vd->vdev_guid; | |
2245 | ||
2246 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
2247 | ||
2248 | /* | |
2249 | * We have to grab the zs_name_lock as writer to | |
2250 | * prevent a race between removing a slog (dmu_objset_find) | |
2251 | * and destroying a dataset. Removing the slog will | |
2252 | * grab a reference on the dataset which may cause | |
2253 | * dmu_objset_destroy() to fail with EBUSY thus | |
2254 | * leaving the dataset in an inconsistent state. | |
2255 | */ | |
1e33ac1e | 2256 | rw_enter(&ztest_shared->zs_name_lock, RW_WRITER); |
428870ff | 2257 | error = spa_vdev_remove(spa, guid, B_FALSE); |
1e33ac1e | 2258 | rw_exit(&ztest_shared->zs_name_lock); |
428870ff BB |
2259 | |
2260 | if (error && error != EEXIST) | |
2261 | fatal(0, "spa_vdev_remove() = %d", error); | |
2262 | } else { | |
2263 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
2264 | ||
2265 | /* | |
2266 | * Make 1/4 of the devices be log devices. | |
2267 | */ | |
2268 | nvroot = make_vdev_root(NULL, NULL, zopt_vdev_size, 0, | |
2269 | ztest_random(4) == 0, zopt_raidz, zs->zs_mirrors, 1); | |
2270 | ||
2271 | error = spa_vdev_add(spa, nvroot); | |
2272 | nvlist_free(nvroot); | |
2273 | ||
2274 | if (error == ENOSPC) | |
2275 | ztest_record_enospc("spa_vdev_add"); | |
2276 | else if (error != 0) | |
2277 | fatal(0, "spa_vdev_add() = %d", error); | |
2278 | } | |
2279 | ||
1e33ac1e | 2280 | mutex_exit(&ztest_shared->zs_vdev_lock); |
428870ff BB |
2281 | } |
2282 | ||
2283 | /* | |
2284 | * Verify that adding/removing aux devices (l2arc, hot spare) works as expected. | |
2285 | */ | |
2286 | /* ARGSUSED */ | |
2287 | void | |
2288 | ztest_vdev_aux_add_remove(ztest_ds_t *zd, uint64_t id) | |
2289 | { | |
2290 | ztest_shared_t *zs = ztest_shared; | |
2291 | spa_t *spa = zs->zs_spa; | |
2292 | vdev_t *rvd = spa->spa_root_vdev; | |
2293 | spa_aux_vdev_t *sav; | |
2294 | char *aux; | |
40b84e7a | 2295 | char *path; |
428870ff BB |
2296 | uint64_t guid = 0; |
2297 | int error; | |
2298 | ||
40b84e7a BB |
2299 | path = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
2300 | ||
428870ff BB |
2301 | if (ztest_random(2) == 0) { |
2302 | sav = &spa->spa_spares; | |
2303 | aux = ZPOOL_CONFIG_SPARES; | |
2304 | } else { | |
2305 | sav = &spa->spa_l2cache; | |
2306 | aux = ZPOOL_CONFIG_L2CACHE; | |
2307 | } | |
2308 | ||
1e33ac1e | 2309 | mutex_enter(&zs->zs_vdev_lock); |
428870ff BB |
2310 | |
2311 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
2312 | ||
2313 | if (sav->sav_count != 0 && ztest_random(4) == 0) { | |
b128c09f BB |
2314 | /* |
2315 | * Pick a random device to remove. | |
2316 | */ | |
2317 | guid = sav->sav_vdevs[ztest_random(sav->sav_count)]->vdev_guid; | |
2318 | } else { | |
2319 | /* | |
2320 | * Find an unused device we can add. | |
2321 | */ | |
428870ff | 2322 | zs->zs_vdev_aux = 0; |
b128c09f | 2323 | for (;;) { |
b128c09f BB |
2324 | int c; |
2325 | (void) sprintf(path, ztest_aux_template, zopt_dir, | |
428870ff | 2326 | zopt_pool, aux, zs->zs_vdev_aux); |
b128c09f BB |
2327 | for (c = 0; c < sav->sav_count; c++) |
2328 | if (strcmp(sav->sav_vdevs[c]->vdev_path, | |
2329 | path) == 0) | |
2330 | break; | |
2331 | if (c == sav->sav_count && | |
2332 | vdev_lookup_by_path(rvd, path) == NULL) | |
2333 | break; | |
428870ff | 2334 | zs->zs_vdev_aux++; |
34dc7c2f BB |
2335 | } |
2336 | } | |
2337 | ||
b128c09f | 2338 | spa_config_exit(spa, SCL_VDEV, FTAG); |
34dc7c2f | 2339 | |
b128c09f BB |
2340 | if (guid == 0) { |
2341 | /* | |
2342 | * Add a new device. | |
2343 | */ | |
2344 | nvlist_t *nvroot = make_vdev_root(NULL, aux, | |
2345 | (zopt_vdev_size * 5) / 4, 0, 0, 0, 0, 1); | |
2346 | error = spa_vdev_add(spa, nvroot); | |
2347 | if (error != 0) | |
2348 | fatal(0, "spa_vdev_add(%p) = %d", nvroot, error); | |
2349 | nvlist_free(nvroot); | |
2350 | } else { | |
2351 | /* | |
2352 | * Remove an existing device. Sometimes, dirty its | |
2353 | * vdev state first to make sure we handle removal | |
2354 | * of devices that have pending state changes. | |
2355 | */ | |
2356 | if (ztest_random(2) == 0) | |
9babb374 | 2357 | (void) vdev_online(spa, guid, 0, NULL); |
b128c09f BB |
2358 | |
2359 | error = spa_vdev_remove(spa, guid, B_FALSE); | |
2360 | if (error != 0 && error != EBUSY) | |
2361 | fatal(0, "spa_vdev_remove(%llu) = %d", guid, error); | |
2362 | } | |
2363 | ||
1e33ac1e | 2364 | mutex_exit(&zs->zs_vdev_lock); |
40b84e7a BB |
2365 | |
2366 | umem_free(path, MAXPATHLEN); | |
428870ff BB |
2367 | } |
2368 | ||
2369 | /* | |
2370 | * split a pool if it has mirror tlvdevs | |
2371 | */ | |
2372 | /* ARGSUSED */ | |
2373 | void | |
2374 | ztest_split_pool(ztest_ds_t *zd, uint64_t id) | |
2375 | { | |
2376 | ztest_shared_t *zs = ztest_shared; | |
2377 | spa_t *spa = zs->zs_spa; | |
2378 | vdev_t *rvd = spa->spa_root_vdev; | |
2379 | nvlist_t *tree, **child, *config, *split, **schild; | |
2380 | uint_t c, children, schildren = 0, lastlogid = 0; | |
2381 | int error = 0; | |
2382 | ||
1e33ac1e | 2383 | mutex_enter(&zs->zs_vdev_lock); |
428870ff BB |
2384 | |
2385 | /* ensure we have a useable config; mirrors of raidz aren't supported */ | |
2386 | if (zs->zs_mirrors < 3 || zopt_raidz > 1) { | |
1e33ac1e | 2387 | mutex_exit(&zs->zs_vdev_lock); |
428870ff BB |
2388 | return; |
2389 | } | |
2390 | ||
2391 | /* clean up the old pool, if any */ | |
2392 | (void) spa_destroy("splitp"); | |
2393 | ||
2394 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
2395 | ||
2396 | /* generate a config from the existing config */ | |
2397 | mutex_enter(&spa->spa_props_lock); | |
2398 | VERIFY(nvlist_lookup_nvlist(spa->spa_config, ZPOOL_CONFIG_VDEV_TREE, | |
2399 | &tree) == 0); | |
2400 | mutex_exit(&spa->spa_props_lock); | |
2401 | ||
2402 | VERIFY(nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child, | |
2403 | &children) == 0); | |
2404 | ||
2405 | schild = malloc(rvd->vdev_children * sizeof (nvlist_t *)); | |
2406 | for (c = 0; c < children; c++) { | |
2407 | vdev_t *tvd = rvd->vdev_child[c]; | |
2408 | nvlist_t **mchild; | |
2409 | uint_t mchildren; | |
2410 | ||
2411 | if (tvd->vdev_islog || tvd->vdev_ops == &vdev_hole_ops) { | |
2412 | VERIFY(nvlist_alloc(&schild[schildren], NV_UNIQUE_NAME, | |
2413 | 0) == 0); | |
2414 | VERIFY(nvlist_add_string(schild[schildren], | |
2415 | ZPOOL_CONFIG_TYPE, VDEV_TYPE_HOLE) == 0); | |
2416 | VERIFY(nvlist_add_uint64(schild[schildren], | |
2417 | ZPOOL_CONFIG_IS_HOLE, 1) == 0); | |
2418 | if (lastlogid == 0) | |
2419 | lastlogid = schildren; | |
2420 | ++schildren; | |
2421 | continue; | |
2422 | } | |
2423 | lastlogid = 0; | |
2424 | VERIFY(nvlist_lookup_nvlist_array(child[c], | |
2425 | ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0); | |
2426 | VERIFY(nvlist_dup(mchild[0], &schild[schildren++], 0) == 0); | |
2427 | } | |
2428 | ||
2429 | /* OK, create a config that can be used to split */ | |
2430 | VERIFY(nvlist_alloc(&split, NV_UNIQUE_NAME, 0) == 0); | |
2431 | VERIFY(nvlist_add_string(split, ZPOOL_CONFIG_TYPE, | |
2432 | VDEV_TYPE_ROOT) == 0); | |
2433 | VERIFY(nvlist_add_nvlist_array(split, ZPOOL_CONFIG_CHILDREN, schild, | |
2434 | lastlogid != 0 ? lastlogid : schildren) == 0); | |
2435 | ||
2436 | VERIFY(nvlist_alloc(&config, NV_UNIQUE_NAME, 0) == 0); | |
2437 | VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, split) == 0); | |
2438 | ||
2439 | for (c = 0; c < schildren; c++) | |
2440 | nvlist_free(schild[c]); | |
2441 | free(schild); | |
2442 | nvlist_free(split); | |
2443 | ||
2444 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
2445 | ||
1e33ac1e | 2446 | (void) rw_enter(&zs->zs_name_lock, RW_WRITER); |
428870ff | 2447 | error = spa_vdev_split_mirror(spa, "splitp", config, NULL, B_FALSE); |
1e33ac1e | 2448 | (void) rw_exit(&zs->zs_name_lock); |
428870ff BB |
2449 | |
2450 | nvlist_free(config); | |
2451 | ||
2452 | if (error == 0) { | |
2453 | (void) printf("successful split - results:\n"); | |
2454 | mutex_enter(&spa_namespace_lock); | |
2455 | show_pool_stats(spa); | |
2456 | show_pool_stats(spa_lookup("splitp")); | |
2457 | mutex_exit(&spa_namespace_lock); | |
2458 | ++zs->zs_splits; | |
2459 | --zs->zs_mirrors; | |
2460 | } | |
1e33ac1e | 2461 | mutex_exit(&zs->zs_vdev_lock); |
428870ff | 2462 | |
34dc7c2f BB |
2463 | } |
2464 | ||
2465 | /* | |
2466 | * Verify that we can attach and detach devices. | |
2467 | */ | |
428870ff | 2468 | /* ARGSUSED */ |
34dc7c2f | 2469 | void |
428870ff | 2470 | ztest_vdev_attach_detach(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 2471 | { |
428870ff BB |
2472 | ztest_shared_t *zs = ztest_shared; |
2473 | spa_t *spa = zs->zs_spa; | |
b128c09f | 2474 | spa_aux_vdev_t *sav = &spa->spa_spares; |
34dc7c2f BB |
2475 | vdev_t *rvd = spa->spa_root_vdev; |
2476 | vdev_t *oldvd, *newvd, *pvd; | |
b128c09f | 2477 | nvlist_t *root; |
428870ff | 2478 | uint64_t leaves; |
34dc7c2f BB |
2479 | uint64_t leaf, top; |
2480 | uint64_t ashift = ztest_get_ashift(); | |
fb5f0bc8 | 2481 | uint64_t oldguid, pguid; |
34dc7c2f | 2482 | size_t oldsize, newsize; |
40b84e7a | 2483 | char *oldpath, *newpath; |
34dc7c2f | 2484 | int replacing; |
b128c09f BB |
2485 | int oldvd_has_siblings = B_FALSE; |
2486 | int newvd_is_spare = B_FALSE; | |
2487 | int oldvd_is_log; | |
34dc7c2f | 2488 | int error, expected_error; |
34dc7c2f | 2489 | |
40b84e7a BB |
2490 | oldpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
2491 | newpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
2492 | ||
1e33ac1e | 2493 | mutex_enter(&zs->zs_vdev_lock); |
428870ff | 2494 | leaves = MAX(zs->zs_mirrors, 1) * zopt_raidz; |
34dc7c2f | 2495 | |
b128c09f | 2496 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); |
34dc7c2f BB |
2497 | |
2498 | /* | |
2499 | * Decide whether to do an attach or a replace. | |
2500 | */ | |
2501 | replacing = ztest_random(2); | |
2502 | ||
2503 | /* | |
2504 | * Pick a random top-level vdev. | |
2505 | */ | |
428870ff | 2506 | top = ztest_random_vdev_top(spa, B_TRUE); |
34dc7c2f BB |
2507 | |
2508 | /* | |
2509 | * Pick a random leaf within it. | |
2510 | */ | |
2511 | leaf = ztest_random(leaves); | |
2512 | ||
2513 | /* | |
b128c09f | 2514 | * Locate this vdev. |
34dc7c2f | 2515 | */ |
b128c09f | 2516 | oldvd = rvd->vdev_child[top]; |
428870ff | 2517 | if (zs->zs_mirrors >= 1) { |
fb5f0bc8 | 2518 | ASSERT(oldvd->vdev_ops == &vdev_mirror_ops); |
428870ff | 2519 | ASSERT(oldvd->vdev_children >= zs->zs_mirrors); |
b128c09f | 2520 | oldvd = oldvd->vdev_child[leaf / zopt_raidz]; |
fb5f0bc8 BB |
2521 | } |
2522 | if (zopt_raidz > 1) { | |
2523 | ASSERT(oldvd->vdev_ops == &vdev_raidz_ops); | |
2524 | ASSERT(oldvd->vdev_children == zopt_raidz); | |
b128c09f | 2525 | oldvd = oldvd->vdev_child[leaf % zopt_raidz]; |
fb5f0bc8 | 2526 | } |
34dc7c2f BB |
2527 | |
2528 | /* | |
b128c09f BB |
2529 | * If we're already doing an attach or replace, oldvd may be a |
2530 | * mirror vdev -- in which case, pick a random child. | |
34dc7c2f | 2531 | */ |
b128c09f BB |
2532 | while (oldvd->vdev_children != 0) { |
2533 | oldvd_has_siblings = B_TRUE; | |
fb5f0bc8 BB |
2534 | ASSERT(oldvd->vdev_children >= 2); |
2535 | oldvd = oldvd->vdev_child[ztest_random(oldvd->vdev_children)]; | |
b128c09f BB |
2536 | } |
2537 | ||
2538 | oldguid = oldvd->vdev_guid; | |
9babb374 | 2539 | oldsize = vdev_get_min_asize(oldvd); |
b128c09f BB |
2540 | oldvd_is_log = oldvd->vdev_top->vdev_islog; |
2541 | (void) strcpy(oldpath, oldvd->vdev_path); | |
2542 | pvd = oldvd->vdev_parent; | |
fb5f0bc8 | 2543 | pguid = pvd->vdev_guid; |
34dc7c2f BB |
2544 | |
2545 | /* | |
b128c09f | 2546 | * If oldvd has siblings, then half of the time, detach it. |
34dc7c2f | 2547 | */ |
b128c09f BB |
2548 | if (oldvd_has_siblings && ztest_random(2) == 0) { |
2549 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
fb5f0bc8 BB |
2550 | error = spa_vdev_detach(spa, oldguid, pguid, B_FALSE); |
2551 | if (error != 0 && error != ENODEV && error != EBUSY && | |
2552 | error != ENOTSUP) | |
2553 | fatal(0, "detach (%s) returned %d", oldpath, error); | |
40b84e7a | 2554 | goto out; |
b128c09f | 2555 | } |
34dc7c2f BB |
2556 | |
2557 | /* | |
b128c09f BB |
2558 | * For the new vdev, choose with equal probability between the two |
2559 | * standard paths (ending in either 'a' or 'b') or a random hot spare. | |
34dc7c2f | 2560 | */ |
b128c09f BB |
2561 | if (sav->sav_count != 0 && ztest_random(3) == 0) { |
2562 | newvd = sav->sav_vdevs[ztest_random(sav->sav_count)]; | |
2563 | newvd_is_spare = B_TRUE; | |
2564 | (void) strcpy(newpath, newvd->vdev_path); | |
2565 | } else { | |
2566 | (void) snprintf(newpath, sizeof (newpath), ztest_dev_template, | |
2567 | zopt_dir, zopt_pool, top * leaves + leaf); | |
2568 | if (ztest_random(2) == 0) | |
2569 | newpath[strlen(newpath) - 1] = 'b'; | |
2570 | newvd = vdev_lookup_by_path(rvd, newpath); | |
2571 | } | |
2572 | ||
2573 | if (newvd) { | |
9babb374 | 2574 | newsize = vdev_get_min_asize(newvd); |
b128c09f BB |
2575 | } else { |
2576 | /* | |
2577 | * Make newsize a little bigger or smaller than oldsize. | |
2578 | * If it's smaller, the attach should fail. | |
2579 | * If it's larger, and we're doing a replace, | |
2580 | * we should get dynamic LUN growth when we're done. | |
2581 | */ | |
2582 | newsize = 10 * oldsize / (9 + ztest_random(3)); | |
2583 | } | |
34dc7c2f BB |
2584 | |
2585 | /* | |
2586 | * If pvd is not a mirror or root, the attach should fail with ENOTSUP, | |
2587 | * unless it's a replace; in that case any non-replacing parent is OK. | |
2588 | * | |
2589 | * If newvd is already part of the pool, it should fail with EBUSY. | |
2590 | * | |
2591 | * If newvd is too small, it should fail with EOVERFLOW. | |
2592 | */ | |
b128c09f BB |
2593 | if (pvd->vdev_ops != &vdev_mirror_ops && |
2594 | pvd->vdev_ops != &vdev_root_ops && (!replacing || | |
2595 | pvd->vdev_ops == &vdev_replacing_ops || | |
2596 | pvd->vdev_ops == &vdev_spare_ops)) | |
34dc7c2f | 2597 | expected_error = ENOTSUP; |
b128c09f BB |
2598 | else if (newvd_is_spare && (!replacing || oldvd_is_log)) |
2599 | expected_error = ENOTSUP; | |
2600 | else if (newvd == oldvd) | |
2601 | expected_error = replacing ? 0 : EBUSY; | |
2602 | else if (vdev_lookup_by_path(rvd, newpath) != NULL) | |
2603 | expected_error = EBUSY; | |
34dc7c2f BB |
2604 | else if (newsize < oldsize) |
2605 | expected_error = EOVERFLOW; | |
2606 | else if (ashift > oldvd->vdev_top->vdev_ashift) | |
2607 | expected_error = EDOM; | |
2608 | else | |
2609 | expected_error = 0; | |
2610 | ||
b128c09f | 2611 | spa_config_exit(spa, SCL_VDEV, FTAG); |
34dc7c2f BB |
2612 | |
2613 | /* | |
2614 | * Build the nvlist describing newpath. | |
2615 | */ | |
b128c09f BB |
2616 | root = make_vdev_root(newpath, NULL, newvd == NULL ? newsize : 0, |
2617 | ashift, 0, 0, 0, 1); | |
34dc7c2f | 2618 | |
b128c09f | 2619 | error = spa_vdev_attach(spa, oldguid, root, replacing); |
34dc7c2f | 2620 | |
34dc7c2f BB |
2621 | nvlist_free(root); |
2622 | ||
2623 | /* | |
2624 | * If our parent was the replacing vdev, but the replace completed, | |
2625 | * then instead of failing with ENOTSUP we may either succeed, | |
2626 | * fail with ENODEV, or fail with EOVERFLOW. | |
2627 | */ | |
2628 | if (expected_error == ENOTSUP && | |
2629 | (error == 0 || error == ENODEV || error == EOVERFLOW)) | |
2630 | expected_error = error; | |
2631 | ||
2632 | /* | |
2633 | * If someone grew the LUN, the replacement may be too small. | |
2634 | */ | |
b128c09f | 2635 | if (error == EOVERFLOW || error == EBUSY) |
34dc7c2f BB |
2636 | expected_error = error; |
2637 | ||
b128c09f BB |
2638 | /* XXX workaround 6690467 */ |
2639 | if (error != expected_error && expected_error != EBUSY) { | |
2640 | fatal(0, "attach (%s %llu, %s %llu, %d) " | |
2641 | "returned %d, expected %d", | |
2642 | oldpath, (longlong_t)oldsize, newpath, | |
2643 | (longlong_t)newsize, replacing, error, expected_error); | |
34dc7c2f | 2644 | } |
40b84e7a | 2645 | out: |
1e33ac1e | 2646 | mutex_exit(&zs->zs_vdev_lock); |
40b84e7a BB |
2647 | |
2648 | umem_free(oldpath, MAXPATHLEN); | |
2649 | umem_free(newpath, MAXPATHLEN); | |
34dc7c2f BB |
2650 | } |
2651 | ||
9babb374 BB |
2652 | /* |
2653 | * Callback function which expands the physical size of the vdev. | |
2654 | */ | |
2655 | vdev_t * | |
2656 | grow_vdev(vdev_t *vd, void *arg) | |
2657 | { | |
1fde1e37 | 2658 | ASSERTV(spa_t *spa = vd->vdev_spa); |
9babb374 BB |
2659 | size_t *newsize = arg; |
2660 | size_t fsize; | |
2661 | int fd; | |
2662 | ||
2663 | ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE); | |
2664 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
2665 | ||
2666 | if ((fd = open(vd->vdev_path, O_RDWR)) == -1) | |
2667 | return (vd); | |
2668 | ||
2669 | fsize = lseek(fd, 0, SEEK_END); | |
0e5b68e0 | 2670 | VERIFY(ftruncate(fd, *newsize) == 0); |
9babb374 BB |
2671 | |
2672 | if (zopt_verbose >= 6) { | |
2673 | (void) printf("%s grew from %lu to %lu bytes\n", | |
2674 | vd->vdev_path, (ulong_t)fsize, (ulong_t)*newsize); | |
2675 | } | |
2676 | (void) close(fd); | |
2677 | return (NULL); | |
2678 | } | |
2679 | ||
2680 | /* | |
2681 | * Callback function which expands a given vdev by calling vdev_online(). | |
2682 | */ | |
2683 | /* ARGSUSED */ | |
2684 | vdev_t * | |
2685 | online_vdev(vdev_t *vd, void *arg) | |
2686 | { | |
2687 | spa_t *spa = vd->vdev_spa; | |
2688 | vdev_t *tvd = vd->vdev_top; | |
9babb374 | 2689 | uint64_t guid = vd->vdev_guid; |
428870ff BB |
2690 | uint64_t generation = spa->spa_config_generation + 1; |
2691 | vdev_state_t newstate = VDEV_STATE_UNKNOWN; | |
2692 | int error; | |
9babb374 BB |
2693 | |
2694 | ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE); | |
2695 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
2696 | ||
2697 | /* Calling vdev_online will initialize the new metaslabs */ | |
2698 | spa_config_exit(spa, SCL_STATE, spa); | |
428870ff | 2699 | error = vdev_online(spa, guid, ZFS_ONLINE_EXPAND, &newstate); |
9babb374 BB |
2700 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); |
2701 | ||
428870ff BB |
2702 | /* |
2703 | * If vdev_online returned an error or the underlying vdev_open | |
2704 | * failed then we abort the expand. The only way to know that | |
2705 | * vdev_open fails is by checking the returned newstate. | |
2706 | */ | |
2707 | if (error || newstate != VDEV_STATE_HEALTHY) { | |
2708 | if (zopt_verbose >= 5) { | |
2709 | (void) printf("Unable to expand vdev, state %llu, " | |
2710 | "error %d\n", (u_longlong_t)newstate, error); | |
2711 | } | |
2712 | return (vd); | |
2713 | } | |
2714 | ASSERT3U(newstate, ==, VDEV_STATE_HEALTHY); | |
2715 | ||
9babb374 BB |
2716 | /* |
2717 | * Since we dropped the lock we need to ensure that we're | |
2718 | * still talking to the original vdev. It's possible this | |
2719 | * vdev may have been detached/replaced while we were | |
2720 | * trying to online it. | |
2721 | */ | |
428870ff BB |
2722 | if (generation != spa->spa_config_generation) { |
2723 | if (zopt_verbose >= 5) { | |
2724 | (void) printf("vdev configuration has changed, " | |
2725 | "guid %llu, state %llu, expected gen %llu, " | |
2726 | "got gen %llu\n", | |
2727 | (u_longlong_t)guid, | |
2728 | (u_longlong_t)tvd->vdev_state, | |
2729 | (u_longlong_t)generation, | |
2730 | (u_longlong_t)spa->spa_config_generation); | |
9babb374 BB |
2731 | } |
2732 | return (vd); | |
2733 | } | |
2734 | return (NULL); | |
2735 | } | |
2736 | ||
2737 | /* | |
2738 | * Traverse the vdev tree calling the supplied function. | |
2739 | * We continue to walk the tree until we either have walked all | |
2740 | * children or we receive a non-NULL return from the callback. | |
2741 | * If a NULL callback is passed, then we just return back the first | |
2742 | * leaf vdev we encounter. | |
2743 | */ | |
2744 | vdev_t * | |
2745 | vdev_walk_tree(vdev_t *vd, vdev_t *(*func)(vdev_t *, void *), void *arg) | |
2746 | { | |
d6320ddb BB |
2747 | uint_t c; |
2748 | ||
9babb374 BB |
2749 | if (vd->vdev_ops->vdev_op_leaf) { |
2750 | if (func == NULL) | |
2751 | return (vd); | |
2752 | else | |
2753 | return (func(vd, arg)); | |
2754 | } | |
2755 | ||
d6320ddb | 2756 | for (c = 0; c < vd->vdev_children; c++) { |
9babb374 BB |
2757 | vdev_t *cvd = vd->vdev_child[c]; |
2758 | if ((cvd = vdev_walk_tree(cvd, func, arg)) != NULL) | |
2759 | return (cvd); | |
2760 | } | |
2761 | return (NULL); | |
2762 | } | |
2763 | ||
34dc7c2f BB |
2764 | /* |
2765 | * Verify that dynamic LUN growth works as expected. | |
2766 | */ | |
428870ff | 2767 | /* ARGSUSED */ |
34dc7c2f | 2768 | void |
428870ff | 2769 | ztest_vdev_LUN_growth(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 2770 | { |
428870ff BB |
2771 | ztest_shared_t *zs = ztest_shared; |
2772 | spa_t *spa = zs->zs_spa; | |
2773 | vdev_t *vd, *tvd; | |
2774 | metaslab_class_t *mc; | |
2775 | metaslab_group_t *mg; | |
9babb374 | 2776 | size_t psize, newsize; |
428870ff BB |
2777 | uint64_t top; |
2778 | uint64_t old_class_space, new_class_space, old_ms_count, new_ms_count; | |
34dc7c2f | 2779 | |
1e33ac1e | 2780 | mutex_enter(&zs->zs_vdev_lock); |
9babb374 BB |
2781 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); |
2782 | ||
428870ff | 2783 | top = ztest_random_vdev_top(spa, B_TRUE); |
9babb374 | 2784 | |
428870ff BB |
2785 | tvd = spa->spa_root_vdev->vdev_child[top]; |
2786 | mg = tvd->vdev_mg; | |
2787 | mc = mg->mg_class; | |
2788 | old_ms_count = tvd->vdev_ms_count; | |
2789 | old_class_space = metaslab_class_get_space(mc); | |
34dc7c2f BB |
2790 | |
2791 | /* | |
9babb374 BB |
2792 | * Determine the size of the first leaf vdev associated with |
2793 | * our top-level device. | |
34dc7c2f | 2794 | */ |
9babb374 BB |
2795 | vd = vdev_walk_tree(tvd, NULL, NULL); |
2796 | ASSERT3P(vd, !=, NULL); | |
2797 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
34dc7c2f | 2798 | |
9babb374 | 2799 | psize = vd->vdev_psize; |
34dc7c2f | 2800 | |
9babb374 | 2801 | /* |
428870ff BB |
2802 | * We only try to expand the vdev if it's healthy, less than 4x its |
2803 | * original size, and it has a valid psize. | |
9babb374 | 2804 | */ |
428870ff BB |
2805 | if (tvd->vdev_state != VDEV_STATE_HEALTHY || |
2806 | psize == 0 || psize >= 4 * zopt_vdev_size) { | |
9babb374 | 2807 | spa_config_exit(spa, SCL_STATE, spa); |
1e33ac1e | 2808 | mutex_exit(&zs->zs_vdev_lock); |
9babb374 BB |
2809 | return; |
2810 | } | |
2811 | ASSERT(psize > 0); | |
2812 | newsize = psize + psize / 8; | |
2813 | ASSERT3U(newsize, >, psize); | |
34dc7c2f | 2814 | |
9babb374 | 2815 | if (zopt_verbose >= 6) { |
428870ff | 2816 | (void) printf("Expanding LUN %s from %lu to %lu\n", |
9babb374 BB |
2817 | vd->vdev_path, (ulong_t)psize, (ulong_t)newsize); |
2818 | } | |
2819 | ||
9babb374 BB |
2820 | /* |
2821 | * Growing the vdev is a two step process: | |
2822 | * 1). expand the physical size (i.e. relabel) | |
2823 | * 2). online the vdev to create the new metaslabs | |
2824 | */ | |
2825 | if (vdev_walk_tree(tvd, grow_vdev, &newsize) != NULL || | |
2826 | vdev_walk_tree(tvd, online_vdev, NULL) != NULL || | |
2827 | tvd->vdev_state != VDEV_STATE_HEALTHY) { | |
2828 | if (zopt_verbose >= 5) { | |
2829 | (void) printf("Could not expand LUN because " | |
428870ff | 2830 | "the vdev configuration changed.\n"); |
34dc7c2f | 2831 | } |
428870ff | 2832 | spa_config_exit(spa, SCL_STATE, spa); |
1e33ac1e | 2833 | mutex_exit(&zs->zs_vdev_lock); |
9babb374 | 2834 | return; |
34dc7c2f BB |
2835 | } |
2836 | ||
428870ff | 2837 | spa_config_exit(spa, SCL_STATE, spa); |
9babb374 BB |
2838 | |
2839 | /* | |
2840 | * Expanding the LUN will update the config asynchronously, | |
2841 | * thus we must wait for the async thread to complete any | |
2842 | * pending tasks before proceeding. | |
2843 | */ | |
428870ff BB |
2844 | for (;;) { |
2845 | boolean_t done; | |
2846 | mutex_enter(&spa->spa_async_lock); | |
2847 | done = (spa->spa_async_thread == NULL && !spa->spa_async_tasks); | |
2848 | mutex_exit(&spa->spa_async_lock); | |
2849 | if (done) | |
2850 | break; | |
2851 | txg_wait_synced(spa_get_dsl(spa), 0); | |
2852 | (void) poll(NULL, 0, 100); | |
2853 | } | |
9babb374 BB |
2854 | |
2855 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); | |
428870ff BB |
2856 | |
2857 | tvd = spa->spa_root_vdev->vdev_child[top]; | |
2858 | new_ms_count = tvd->vdev_ms_count; | |
2859 | new_class_space = metaslab_class_get_space(mc); | |
2860 | ||
2861 | if (tvd->vdev_mg != mg || mg->mg_class != mc) { | |
2862 | if (zopt_verbose >= 5) { | |
2863 | (void) printf("Could not verify LUN expansion due to " | |
2864 | "intervening vdev offline or remove.\n"); | |
2865 | } | |
2866 | spa_config_exit(spa, SCL_STATE, spa); | |
1e33ac1e | 2867 | mutex_exit(&zs->zs_vdev_lock); |
428870ff BB |
2868 | return; |
2869 | } | |
2870 | ||
2871 | /* | |
2872 | * Make sure we were able to grow the vdev. | |
2873 | */ | |
2874 | if (new_ms_count <= old_ms_count) | |
2875 | fatal(0, "LUN expansion failed: ms_count %llu <= %llu\n", | |
2876 | old_ms_count, new_ms_count); | |
9babb374 BB |
2877 | |
2878 | /* | |
2879 | * Make sure we were able to grow the pool. | |
2880 | */ | |
428870ff BB |
2881 | if (new_class_space <= old_class_space) |
2882 | fatal(0, "LUN expansion failed: class_space %llu <= %llu\n", | |
2883 | old_class_space, new_class_space); | |
2884 | ||
2885 | if (zopt_verbose >= 5) { | |
9babb374 BB |
2886 | char oldnumbuf[6], newnumbuf[6]; |
2887 | ||
428870ff BB |
2888 | nicenum(old_class_space, oldnumbuf); |
2889 | nicenum(new_class_space, newnumbuf); | |
9babb374 BB |
2890 | (void) printf("%s grew from %s to %s\n", |
2891 | spa->spa_name, oldnumbuf, newnumbuf); | |
2892 | } | |
428870ff | 2893 | |
9babb374 | 2894 | spa_config_exit(spa, SCL_STATE, spa); |
1e33ac1e | 2895 | mutex_exit(&zs->zs_vdev_lock); |
34dc7c2f BB |
2896 | } |
2897 | ||
428870ff BB |
2898 | /* |
2899 | * Verify that dmu_objset_{create,destroy,open,close} work as expected. | |
2900 | */ | |
34dc7c2f BB |
2901 | /* ARGSUSED */ |
2902 | static void | |
428870ff | 2903 | ztest_objset_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) |
34dc7c2f BB |
2904 | { |
2905 | /* | |
428870ff | 2906 | * Create the objects common to all ztest datasets. |
34dc7c2f | 2907 | */ |
428870ff | 2908 | VERIFY(zap_create_claim(os, ZTEST_DIROBJ, |
34dc7c2f | 2909 | DMU_OT_ZAP_OTHER, DMU_OT_NONE, 0, tx) == 0); |
428870ff | 2910 | } |
34dc7c2f | 2911 | |
428870ff BB |
2912 | static int |
2913 | ztest_dataset_create(char *dsname) | |
2914 | { | |
2915 | uint64_t zilset = ztest_random(100); | |
2916 | int err = dmu_objset_create(dsname, DMU_OST_OTHER, 0, | |
2917 | ztest_objset_create_cb, NULL); | |
2918 | ||
2919 | if (err || zilset < 80) | |
2920 | return (err); | |
2921 | ||
2922 | (void) printf("Setting dataset %s to sync always\n", dsname); | |
2923 | return (ztest_dsl_prop_set_uint64(dsname, ZFS_PROP_SYNC, | |
2924 | ZFS_SYNC_ALWAYS, B_FALSE)); | |
34dc7c2f BB |
2925 | } |
2926 | ||
428870ff | 2927 | /* ARGSUSED */ |
34dc7c2f | 2928 | static int |
428870ff | 2929 | ztest_objset_destroy_cb(const char *name, void *arg) |
34dc7c2f | 2930 | { |
34dc7c2f | 2931 | objset_t *os; |
428870ff | 2932 | dmu_object_info_t doi; |
34dc7c2f BB |
2933 | int error; |
2934 | ||
2935 | /* | |
2936 | * Verify that the dataset contains a directory object. | |
2937 | */ | |
428870ff BB |
2938 | VERIFY3U(0, ==, dmu_objset_hold(name, FTAG, &os)); |
2939 | error = dmu_object_info(os, ZTEST_DIROBJ, &doi); | |
34dc7c2f BB |
2940 | if (error != ENOENT) { |
2941 | /* We could have crashed in the middle of destroying it */ | |
2942 | ASSERT3U(error, ==, 0); | |
428870ff BB |
2943 | ASSERT3U(doi.doi_type, ==, DMU_OT_ZAP_OTHER); |
2944 | ASSERT3S(doi.doi_physical_blocks_512, >=, 0); | |
34dc7c2f | 2945 | } |
428870ff | 2946 | dmu_objset_rele(os, FTAG); |
34dc7c2f BB |
2947 | |
2948 | /* | |
2949 | * Destroy the dataset. | |
2950 | */ | |
428870ff | 2951 | VERIFY3U(0, ==, dmu_objset_destroy(name, B_FALSE)); |
34dc7c2f BB |
2952 | return (0); |
2953 | } | |
2954 | ||
428870ff BB |
2955 | static boolean_t |
2956 | ztest_snapshot_create(char *osname, uint64_t id) | |
34dc7c2f | 2957 | { |
428870ff BB |
2958 | char snapname[MAXNAMELEN]; |
2959 | int error; | |
2960 | ||
2961 | (void) snprintf(snapname, MAXNAMELEN, "%s@%llu", osname, | |
2962 | (u_longlong_t)id); | |
2963 | ||
2964 | error = dmu_objset_snapshot(osname, strchr(snapname, '@') + 1, | |
572e2857 | 2965 | NULL, NULL, B_FALSE, B_FALSE, -1); |
428870ff BB |
2966 | if (error == ENOSPC) { |
2967 | ztest_record_enospc(FTAG); | |
2968 | return (B_FALSE); | |
2969 | } | |
2970 | if (error != 0 && error != EEXIST) | |
2971 | fatal(0, "ztest_snapshot_create(%s) = %d", snapname, error); | |
2972 | return (B_TRUE); | |
2973 | } | |
2974 | ||
2975 | static boolean_t | |
2976 | ztest_snapshot_destroy(char *osname, uint64_t id) | |
2977 | { | |
2978 | char snapname[MAXNAMELEN]; | |
2979 | int error; | |
2980 | ||
2981 | (void) snprintf(snapname, MAXNAMELEN, "%s@%llu", osname, | |
2982 | (u_longlong_t)id); | |
2983 | ||
2984 | error = dmu_objset_destroy(snapname, B_FALSE); | |
2985 | if (error != 0 && error != ENOENT) | |
2986 | fatal(0, "ztest_snapshot_destroy(%s) = %d", snapname, error); | |
2987 | return (B_TRUE); | |
34dc7c2f BB |
2988 | } |
2989 | ||
428870ff | 2990 | /* ARGSUSED */ |
34dc7c2f | 2991 | void |
428870ff | 2992 | ztest_dmu_objset_create_destroy(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 2993 | { |
428870ff | 2994 | ztest_shared_t *zs = ztest_shared; |
40b84e7a | 2995 | ztest_ds_t *zdtmp; |
428870ff | 2996 | int iters; |
34dc7c2f | 2997 | int error; |
b128c09f | 2998 | objset_t *os, *os2; |
40b84e7a | 2999 | char *name; |
34dc7c2f | 3000 | zilog_t *zilog; |
d6320ddb | 3001 | int i; |
34dc7c2f | 3002 | |
40b84e7a BB |
3003 | zdtmp = umem_alloc(sizeof (ztest_ds_t), UMEM_NOFAIL); |
3004 | name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3005 | ||
1e33ac1e | 3006 | (void) rw_enter(&zs->zs_name_lock, RW_READER); |
34dc7c2f | 3007 | |
428870ff BB |
3008 | (void) snprintf(name, MAXNAMELEN, "%s/temp_%llu", |
3009 | zs->zs_pool, (u_longlong_t)id); | |
34dc7c2f BB |
3010 | |
3011 | /* | |
3012 | * If this dataset exists from a previous run, process its replay log | |
3013 | * half of the time. If we don't replay it, then dmu_objset_destroy() | |
428870ff | 3014 | * (invoked from ztest_objset_destroy_cb()) should just throw it away. |
34dc7c2f BB |
3015 | */ |
3016 | if (ztest_random(2) == 0 && | |
428870ff | 3017 | dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, FTAG, &os) == 0) { |
40b84e7a BB |
3018 | ztest_zd_init(zdtmp, os); |
3019 | zil_replay(os, zdtmp, ztest_replay_vector); | |
3020 | ztest_zd_fini(zdtmp); | |
428870ff | 3021 | dmu_objset_disown(os, FTAG); |
34dc7c2f BB |
3022 | } |
3023 | ||
3024 | /* | |
3025 | * There may be an old instance of the dataset we're about to | |
3026 | * create lying around from a previous run. If so, destroy it | |
3027 | * and all of its snapshots. | |
3028 | */ | |
428870ff | 3029 | (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL, |
34dc7c2f BB |
3030 | DS_FIND_CHILDREN | DS_FIND_SNAPSHOTS); |
3031 | ||
3032 | /* | |
3033 | * Verify that the destroyed dataset is no longer in the namespace. | |
3034 | */ | |
428870ff | 3035 | VERIFY3U(ENOENT, ==, dmu_objset_hold(name, FTAG, &os)); |
34dc7c2f BB |
3036 | |
3037 | /* | |
3038 | * Verify that we can create a new dataset. | |
3039 | */ | |
428870ff | 3040 | error = ztest_dataset_create(name); |
34dc7c2f BB |
3041 | if (error) { |
3042 | if (error == ENOSPC) { | |
428870ff | 3043 | ztest_record_enospc(FTAG); |
40b84e7a | 3044 | goto out; |
34dc7c2f BB |
3045 | } |
3046 | fatal(0, "dmu_objset_create(%s) = %d", name, error); | |
3047 | } | |
3048 | ||
428870ff BB |
3049 | VERIFY3U(0, ==, |
3050 | dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, FTAG, &os)); | |
3051 | ||
40b84e7a | 3052 | ztest_zd_init(zdtmp, os); |
34dc7c2f BB |
3053 | |
3054 | /* | |
3055 | * Open the intent log for it. | |
3056 | */ | |
428870ff | 3057 | zilog = zil_open(os, ztest_get_data); |
34dc7c2f BB |
3058 | |
3059 | /* | |
428870ff BB |
3060 | * Put some objects in there, do a little I/O to them, |
3061 | * and randomly take a couple of snapshots along the way. | |
34dc7c2f | 3062 | */ |
428870ff | 3063 | iters = ztest_random(5); |
d6320ddb | 3064 | for (i = 0; i < iters; i++) { |
40b84e7a | 3065 | ztest_dmu_object_alloc_free(zdtmp, id); |
428870ff BB |
3066 | if (ztest_random(iters) == 0) |
3067 | (void) ztest_snapshot_create(name, i); | |
34dc7c2f BB |
3068 | } |
3069 | ||
3070 | /* | |
3071 | * Verify that we cannot create an existing dataset. | |
3072 | */ | |
428870ff BB |
3073 | VERIFY3U(EEXIST, ==, |
3074 | dmu_objset_create(name, DMU_OST_OTHER, 0, NULL, NULL)); | |
34dc7c2f BB |
3075 | |
3076 | /* | |
428870ff | 3077 | * Verify that we can hold an objset that is also owned. |
b128c09f | 3078 | */ |
428870ff BB |
3079 | VERIFY3U(0, ==, dmu_objset_hold(name, FTAG, &os2)); |
3080 | dmu_objset_rele(os2, FTAG); | |
34dc7c2f | 3081 | |
428870ff BB |
3082 | /* |
3083 | * Verify that we cannot own an objset that is already owned. | |
3084 | */ | |
3085 | VERIFY3U(EBUSY, ==, | |
3086 | dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, FTAG, &os2)); | |
34dc7c2f | 3087 | |
428870ff BB |
3088 | zil_close(zilog); |
3089 | dmu_objset_disown(os, FTAG); | |
40b84e7a BB |
3090 | ztest_zd_fini(zdtmp); |
3091 | out: | |
1e33ac1e | 3092 | (void) rw_exit(&zs->zs_name_lock); |
40b84e7a BB |
3093 | |
3094 | umem_free(name, MAXNAMELEN); | |
3095 | umem_free(zdtmp, sizeof (ztest_ds_t)); | |
34dc7c2f BB |
3096 | } |
3097 | ||
3098 | /* | |
3099 | * Verify that dmu_snapshot_{create,destroy,open,close} work as expected. | |
3100 | */ | |
3101 | void | |
428870ff | 3102 | ztest_dmu_snapshot_create_destroy(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3103 | { |
428870ff | 3104 | ztest_shared_t *zs = ztest_shared; |
34dc7c2f | 3105 | |
1e33ac1e | 3106 | (void) rw_enter(&zs->zs_name_lock, RW_READER); |
428870ff BB |
3107 | (void) ztest_snapshot_destroy(zd->zd_name, id); |
3108 | (void) ztest_snapshot_create(zd->zd_name, id); | |
1e33ac1e | 3109 | (void) rw_exit(&zs->zs_name_lock); |
34dc7c2f BB |
3110 | } |
3111 | ||
9babb374 BB |
3112 | /* |
3113 | * Cleanup non-standard snapshots and clones. | |
3114 | */ | |
3115 | void | |
428870ff | 3116 | ztest_dsl_dataset_cleanup(char *osname, uint64_t id) |
9babb374 | 3117 | { |
40b84e7a BB |
3118 | char *snap1name; |
3119 | char *clone1name; | |
3120 | char *snap2name; | |
3121 | char *clone2name; | |
3122 | char *snap3name; | |
9babb374 BB |
3123 | int error; |
3124 | ||
40b84e7a BB |
3125 | snap1name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); |
3126 | clone1name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3127 | snap2name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3128 | clone2name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3129 | snap3name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3130 | ||
b8864a23 BB |
3131 | (void) snprintf(snap1name, MAXNAMELEN, "%s@s1_%llu", |
3132 | osname, (u_longlong_t)id); | |
3133 | (void) snprintf(clone1name, MAXNAMELEN, "%s/c1_%llu", | |
3134 | osname, (u_longlong_t)id); | |
3135 | (void) snprintf(snap2name, MAXNAMELEN, "%s@s2_%llu", | |
3136 | clone1name, (u_longlong_t)id); | |
3137 | (void) snprintf(clone2name, MAXNAMELEN, "%s/c2_%llu", | |
3138 | osname, (u_longlong_t)id); | |
3139 | (void) snprintf(snap3name, MAXNAMELEN, "%s@s3_%llu", | |
3140 | clone1name, (u_longlong_t)id); | |
9babb374 | 3141 | |
45d1cae3 | 3142 | error = dmu_objset_destroy(clone2name, B_FALSE); |
9babb374 BB |
3143 | if (error && error != ENOENT) |
3144 | fatal(0, "dmu_objset_destroy(%s) = %d", clone2name, error); | |
45d1cae3 | 3145 | error = dmu_objset_destroy(snap3name, B_FALSE); |
9babb374 BB |
3146 | if (error && error != ENOENT) |
3147 | fatal(0, "dmu_objset_destroy(%s) = %d", snap3name, error); | |
45d1cae3 | 3148 | error = dmu_objset_destroy(snap2name, B_FALSE); |
9babb374 BB |
3149 | if (error && error != ENOENT) |
3150 | fatal(0, "dmu_objset_destroy(%s) = %d", snap2name, error); | |
45d1cae3 | 3151 | error = dmu_objset_destroy(clone1name, B_FALSE); |
9babb374 BB |
3152 | if (error && error != ENOENT) |
3153 | fatal(0, "dmu_objset_destroy(%s) = %d", clone1name, error); | |
45d1cae3 | 3154 | error = dmu_objset_destroy(snap1name, B_FALSE); |
9babb374 BB |
3155 | if (error && error != ENOENT) |
3156 | fatal(0, "dmu_objset_destroy(%s) = %d", snap1name, error); | |
40b84e7a BB |
3157 | |
3158 | umem_free(snap1name, MAXNAMELEN); | |
3159 | umem_free(clone1name, MAXNAMELEN); | |
3160 | umem_free(snap2name, MAXNAMELEN); | |
3161 | umem_free(clone2name, MAXNAMELEN); | |
3162 | umem_free(snap3name, MAXNAMELEN); | |
9babb374 BB |
3163 | } |
3164 | ||
3165 | /* | |
3166 | * Verify dsl_dataset_promote handles EBUSY | |
3167 | */ | |
3168 | void | |
428870ff | 3169 | ztest_dsl_dataset_promote_busy(ztest_ds_t *zd, uint64_t id) |
9babb374 | 3170 | { |
428870ff | 3171 | ztest_shared_t *zs = ztest_shared; |
9babb374 BB |
3172 | objset_t *clone; |
3173 | dsl_dataset_t *ds; | |
40b84e7a BB |
3174 | char *snap1name; |
3175 | char *clone1name; | |
3176 | char *snap2name; | |
3177 | char *clone2name; | |
3178 | char *snap3name; | |
428870ff BB |
3179 | char *osname = zd->zd_name; |
3180 | int error; | |
9babb374 | 3181 | |
40b84e7a BB |
3182 | snap1name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); |
3183 | clone1name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3184 | snap2name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3185 | clone2name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3186 | snap3name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3187 | ||
1e33ac1e | 3188 | (void) rw_enter(&zs->zs_name_lock, RW_READER); |
9babb374 | 3189 | |
428870ff | 3190 | ztest_dsl_dataset_cleanup(osname, id); |
9babb374 | 3191 | |
b8864a23 BB |
3192 | (void) snprintf(snap1name, MAXNAMELEN, "%s@s1_%llu", |
3193 | osname, (u_longlong_t)id); | |
3194 | (void) snprintf(clone1name, MAXNAMELEN, "%s/c1_%llu", | |
3195 | osname, (u_longlong_t)id); | |
3196 | (void) snprintf(snap2name, MAXNAMELEN, "%s@s2_%llu", | |
3197 | clone1name, (u_longlong_t)id); | |
3198 | (void) snprintf(clone2name, MAXNAMELEN, "%s/c2_%llu", | |
3199 | osname, (u_longlong_t)id); | |
3200 | (void) snprintf(snap3name, MAXNAMELEN, "%s@s3_%llu", | |
3201 | clone1name, (u_longlong_t)id); | |
9babb374 BB |
3202 | |
3203 | error = dmu_objset_snapshot(osname, strchr(snap1name, '@')+1, | |
572e2857 | 3204 | NULL, NULL, B_FALSE, B_FALSE, -1); |
9babb374 BB |
3205 | if (error && error != EEXIST) { |
3206 | if (error == ENOSPC) { | |
428870ff | 3207 | ztest_record_enospc(FTAG); |
9babb374 BB |
3208 | goto out; |
3209 | } | |
3210 | fatal(0, "dmu_take_snapshot(%s) = %d", snap1name, error); | |
3211 | } | |
3212 | ||
428870ff | 3213 | error = dmu_objset_hold(snap1name, FTAG, &clone); |
9babb374 BB |
3214 | if (error) |
3215 | fatal(0, "dmu_open_snapshot(%s) = %d", snap1name, error); | |
3216 | ||
428870ff BB |
3217 | error = dmu_objset_clone(clone1name, dmu_objset_ds(clone), 0); |
3218 | dmu_objset_rele(clone, FTAG); | |
9babb374 BB |
3219 | if (error) { |
3220 | if (error == ENOSPC) { | |
428870ff | 3221 | ztest_record_enospc(FTAG); |
9babb374 BB |
3222 | goto out; |
3223 | } | |
3224 | fatal(0, "dmu_objset_create(%s) = %d", clone1name, error); | |
3225 | } | |
3226 | ||
428870ff | 3227 | error = dmu_objset_snapshot(clone1name, strchr(snap2name, '@')+1, |
572e2857 | 3228 | NULL, NULL, B_FALSE, B_FALSE, -1); |
428870ff BB |
3229 | if (error && error != EEXIST) { |
3230 | if (error == ENOSPC) { | |
3231 | ztest_record_enospc(FTAG); | |
3232 | goto out; | |
34dc7c2f | 3233 | } |
428870ff BB |
3234 | fatal(0, "dmu_open_snapshot(%s) = %d", snap2name, error); |
3235 | } | |
34dc7c2f | 3236 | |
428870ff | 3237 | error = dmu_objset_snapshot(clone1name, strchr(snap3name, '@')+1, |
572e2857 | 3238 | NULL, NULL, B_FALSE, B_FALSE, -1); |
428870ff BB |
3239 | if (error && error != EEXIST) { |
3240 | if (error == ENOSPC) { | |
3241 | ztest_record_enospc(FTAG); | |
3242 | goto out; | |
3243 | } | |
3244 | fatal(0, "dmu_open_snapshot(%s) = %d", snap3name, error); | |
3245 | } | |
34dc7c2f | 3246 | |
428870ff BB |
3247 | error = dmu_objset_hold(snap3name, FTAG, &clone); |
3248 | if (error) | |
3249 | fatal(0, "dmu_open_snapshot(%s) = %d", snap3name, error); | |
34dc7c2f | 3250 | |
428870ff BB |
3251 | error = dmu_objset_clone(clone2name, dmu_objset_ds(clone), 0); |
3252 | dmu_objset_rele(clone, FTAG); | |
3253 | if (error) { | |
3254 | if (error == ENOSPC) { | |
3255 | ztest_record_enospc(FTAG); | |
3256 | goto out; | |
3257 | } | |
3258 | fatal(0, "dmu_objset_create(%s) = %d", clone2name, error); | |
3259 | } | |
34dc7c2f | 3260 | |
428870ff BB |
3261 | error = dsl_dataset_own(snap2name, B_FALSE, FTAG, &ds); |
3262 | if (error) | |
3263 | fatal(0, "dsl_dataset_own(%s) = %d", snap2name, error); | |
3264 | error = dsl_dataset_promote(clone2name, NULL); | |
3265 | if (error != EBUSY) | |
3266 | fatal(0, "dsl_dataset_promote(%s), %d, not EBUSY", clone2name, | |
3267 | error); | |
3268 | dsl_dataset_disown(ds, FTAG); | |
34dc7c2f | 3269 | |
428870ff BB |
3270 | out: |
3271 | ztest_dsl_dataset_cleanup(osname, id); | |
34dc7c2f | 3272 | |
1e33ac1e | 3273 | (void) rw_exit(&zs->zs_name_lock); |
40b84e7a BB |
3274 | |
3275 | umem_free(snap1name, MAXNAMELEN); | |
3276 | umem_free(clone1name, MAXNAMELEN); | |
3277 | umem_free(snap2name, MAXNAMELEN); | |
3278 | umem_free(clone2name, MAXNAMELEN); | |
3279 | umem_free(snap3name, MAXNAMELEN); | |
428870ff | 3280 | } |
34dc7c2f | 3281 | |
40b84e7a BB |
3282 | #undef OD_ARRAY_SIZE |
3283 | #define OD_ARRAY_SIZE 4 | |
3284 | ||
428870ff BB |
3285 | /* |
3286 | * Verify that dmu_object_{alloc,free} work as expected. | |
3287 | */ | |
3288 | void | |
3289 | ztest_dmu_object_alloc_free(ztest_ds_t *zd, uint64_t id) | |
3290 | { | |
40b84e7a BB |
3291 | ztest_od_t *od; |
3292 | int batchsize; | |
3293 | int size; | |
d6320ddb | 3294 | int b; |
34dc7c2f | 3295 | |
40b84e7a BB |
3296 | size = sizeof(ztest_od_t) * OD_ARRAY_SIZE; |
3297 | od = umem_alloc(size, UMEM_NOFAIL); | |
3298 | batchsize = OD_ARRAY_SIZE; | |
3299 | ||
d6320ddb | 3300 | for (b = 0; b < batchsize; b++) |
40b84e7a | 3301 | ztest_od_init(od + b, id, FTAG, b, DMU_OT_UINT64_OTHER, 0, 0); |
34dc7c2f | 3302 | |
428870ff BB |
3303 | /* |
3304 | * Destroy the previous batch of objects, create a new batch, | |
3305 | * and do some I/O on the new objects. | |
3306 | */ | |
40b84e7a | 3307 | if (ztest_object_init(zd, od, size, B_TRUE) != 0) |
428870ff | 3308 | return; |
34dc7c2f | 3309 | |
428870ff BB |
3310 | while (ztest_random(4 * batchsize) != 0) |
3311 | ztest_io(zd, od[ztest_random(batchsize)].od_object, | |
3312 | ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
40b84e7a BB |
3313 | |
3314 | umem_free(od, size); | |
34dc7c2f BB |
3315 | } |
3316 | ||
40b84e7a BB |
3317 | #undef OD_ARRAY_SIZE |
3318 | #define OD_ARRAY_SIZE 2 | |
3319 | ||
34dc7c2f BB |
3320 | /* |
3321 | * Verify that dmu_{read,write} work as expected. | |
3322 | */ | |
34dc7c2f | 3323 | void |
428870ff | 3324 | ztest_dmu_read_write(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3325 | { |
40b84e7a BB |
3326 | int size; |
3327 | ztest_od_t *od; | |
3328 | ||
428870ff | 3329 | objset_t *os = zd->zd_os; |
40b84e7a BB |
3330 | size = sizeof(ztest_od_t) * OD_ARRAY_SIZE; |
3331 | od = umem_alloc(size, UMEM_NOFAIL); | |
34dc7c2f BB |
3332 | dmu_tx_t *tx; |
3333 | int i, freeit, error; | |
3334 | uint64_t n, s, txg; | |
3335 | bufwad_t *packbuf, *bigbuf, *pack, *bigH, *bigT; | |
428870ff BB |
3336 | uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize; |
3337 | uint64_t chunksize = (1000 + ztest_random(1000)) * sizeof (uint64_t); | |
34dc7c2f BB |
3338 | uint64_t regions = 997; |
3339 | uint64_t stride = 123456789ULL; | |
3340 | uint64_t width = 40; | |
3341 | int free_percent = 5; | |
3342 | ||
3343 | /* | |
3344 | * This test uses two objects, packobj and bigobj, that are always | |
3345 | * updated together (i.e. in the same tx) so that their contents are | |
3346 | * in sync and can be compared. Their contents relate to each other | |
3347 | * in a simple way: packobj is a dense array of 'bufwad' structures, | |
3348 | * while bigobj is a sparse array of the same bufwads. Specifically, | |
3349 | * for any index n, there are three bufwads that should be identical: | |
3350 | * | |
3351 | * packobj, at offset n * sizeof (bufwad_t) | |
3352 | * bigobj, at the head of the nth chunk | |
3353 | * bigobj, at the tail of the nth chunk | |
3354 | * | |
3355 | * The chunk size is arbitrary. It doesn't have to be a power of two, | |
3356 | * and it doesn't have any relation to the object blocksize. | |
3357 | * The only requirement is that it can hold at least two bufwads. | |
3358 | * | |
3359 | * Normally, we write the bufwad to each of these locations. | |
3360 | * However, free_percent of the time we instead write zeroes to | |
3361 | * packobj and perform a dmu_free_range() on bigobj. By comparing | |
3362 | * bigobj to packobj, we can verify that the DMU is correctly | |
3363 | * tracking which parts of an object are allocated and free, | |
3364 | * and that the contents of the allocated blocks are correct. | |
3365 | */ | |
3366 | ||
3367 | /* | |
3368 | * Read the directory info. If it's the first time, set things up. | |
3369 | */ | |
40b84e7a BB |
3370 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, chunksize); |
3371 | ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, chunksize); | |
34dc7c2f | 3372 | |
40b84e7a BB |
3373 | if (ztest_object_init(zd, od, size, B_FALSE) != 0) { |
3374 | umem_free(od, size); | |
428870ff | 3375 | return; |
40b84e7a | 3376 | } |
34dc7c2f | 3377 | |
428870ff BB |
3378 | bigobj = od[0].od_object; |
3379 | packobj = od[1].od_object; | |
3380 | chunksize = od[0].od_gen; | |
3381 | ASSERT(chunksize == od[1].od_gen); | |
34dc7c2f BB |
3382 | |
3383 | /* | |
3384 | * Prefetch a random chunk of the big object. | |
3385 | * Our aim here is to get some async reads in flight | |
3386 | * for blocks that we may free below; the DMU should | |
3387 | * handle this race correctly. | |
3388 | */ | |
3389 | n = ztest_random(regions) * stride + ztest_random(width); | |
3390 | s = 1 + ztest_random(2 * width - 1); | |
428870ff | 3391 | dmu_prefetch(os, bigobj, n * chunksize, s * chunksize); |
34dc7c2f BB |
3392 | |
3393 | /* | |
3394 | * Pick a random index and compute the offsets into packobj and bigobj. | |
3395 | */ | |
3396 | n = ztest_random(regions) * stride + ztest_random(width); | |
3397 | s = 1 + ztest_random(width - 1); | |
3398 | ||
3399 | packoff = n * sizeof (bufwad_t); | |
3400 | packsize = s * sizeof (bufwad_t); | |
3401 | ||
428870ff BB |
3402 | bigoff = n * chunksize; |
3403 | bigsize = s * chunksize; | |
34dc7c2f BB |
3404 | |
3405 | packbuf = umem_alloc(packsize, UMEM_NOFAIL); | |
3406 | bigbuf = umem_alloc(bigsize, UMEM_NOFAIL); | |
3407 | ||
3408 | /* | |
3409 | * free_percent of the time, free a range of bigobj rather than | |
3410 | * overwriting it. | |
3411 | */ | |
3412 | freeit = (ztest_random(100) < free_percent); | |
3413 | ||
3414 | /* | |
3415 | * Read the current contents of our objects. | |
3416 | */ | |
428870ff | 3417 | error = dmu_read(os, packobj, packoff, packsize, packbuf, |
9babb374 | 3418 | DMU_READ_PREFETCH); |
34dc7c2f | 3419 | ASSERT3U(error, ==, 0); |
428870ff | 3420 | error = dmu_read(os, bigobj, bigoff, bigsize, bigbuf, |
9babb374 | 3421 | DMU_READ_PREFETCH); |
34dc7c2f BB |
3422 | ASSERT3U(error, ==, 0); |
3423 | ||
3424 | /* | |
3425 | * Get a tx for the mods to both packobj and bigobj. | |
3426 | */ | |
3427 | tx = dmu_tx_create(os); | |
3428 | ||
428870ff | 3429 | dmu_tx_hold_write(tx, packobj, packoff, packsize); |
34dc7c2f BB |
3430 | |
3431 | if (freeit) | |
428870ff | 3432 | dmu_tx_hold_free(tx, bigobj, bigoff, bigsize); |
34dc7c2f | 3433 | else |
428870ff | 3434 | dmu_tx_hold_write(tx, bigobj, bigoff, bigsize); |
34dc7c2f | 3435 | |
428870ff BB |
3436 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); |
3437 | if (txg == 0) { | |
34dc7c2f BB |
3438 | umem_free(packbuf, packsize); |
3439 | umem_free(bigbuf, bigsize); | |
40b84e7a | 3440 | umem_free(od, size); |
34dc7c2f BB |
3441 | return; |
3442 | } | |
3443 | ||
428870ff BB |
3444 | dmu_object_set_checksum(os, bigobj, |
3445 | (enum zio_checksum)ztest_random_dsl_prop(ZFS_PROP_CHECKSUM), tx); | |
3446 | ||
3447 | dmu_object_set_compress(os, bigobj, | |
3448 | (enum zio_compress)ztest_random_dsl_prop(ZFS_PROP_COMPRESSION), tx); | |
34dc7c2f BB |
3449 | |
3450 | /* | |
3451 | * For each index from n to n + s, verify that the existing bufwad | |
3452 | * in packobj matches the bufwads at the head and tail of the | |
3453 | * corresponding chunk in bigobj. Then update all three bufwads | |
3454 | * with the new values we want to write out. | |
3455 | */ | |
3456 | for (i = 0; i < s; i++) { | |
3457 | /* LINTED */ | |
3458 | pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t)); | |
3459 | /* LINTED */ | |
428870ff | 3460 | bigH = (bufwad_t *)((char *)bigbuf + i * chunksize); |
34dc7c2f | 3461 | /* LINTED */ |
428870ff | 3462 | bigT = (bufwad_t *)((char *)bigH + chunksize) - 1; |
34dc7c2f BB |
3463 | |
3464 | ASSERT((uintptr_t)bigH - (uintptr_t)bigbuf < bigsize); | |
3465 | ASSERT((uintptr_t)bigT - (uintptr_t)bigbuf < bigsize); | |
3466 | ||
3467 | if (pack->bw_txg > txg) | |
3468 | fatal(0, "future leak: got %llx, open txg is %llx", | |
3469 | pack->bw_txg, txg); | |
3470 | ||
3471 | if (pack->bw_data != 0 && pack->bw_index != n + i) | |
3472 | fatal(0, "wrong index: got %llx, wanted %llx+%llx", | |
3473 | pack->bw_index, n, i); | |
3474 | ||
3475 | if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0) | |
3476 | fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH); | |
3477 | ||
3478 | if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0) | |
3479 | fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT); | |
3480 | ||
3481 | if (freeit) { | |
3482 | bzero(pack, sizeof (bufwad_t)); | |
3483 | } else { | |
3484 | pack->bw_index = n + i; | |
3485 | pack->bw_txg = txg; | |
3486 | pack->bw_data = 1 + ztest_random(-2ULL); | |
3487 | } | |
3488 | *bigH = *pack; | |
3489 | *bigT = *pack; | |
3490 | } | |
3491 | ||
3492 | /* | |
3493 | * We've verified all the old bufwads, and made new ones. | |
3494 | * Now write them out. | |
3495 | */ | |
428870ff | 3496 | dmu_write(os, packobj, packoff, packsize, packbuf, tx); |
34dc7c2f BB |
3497 | |
3498 | if (freeit) { | |
428870ff | 3499 | if (zopt_verbose >= 7) { |
34dc7c2f BB |
3500 | (void) printf("freeing offset %llx size %llx" |
3501 | " txg %llx\n", | |
3502 | (u_longlong_t)bigoff, | |
3503 | (u_longlong_t)bigsize, | |
3504 | (u_longlong_t)txg); | |
3505 | } | |
428870ff | 3506 | VERIFY(0 == dmu_free_range(os, bigobj, bigoff, bigsize, tx)); |
34dc7c2f | 3507 | } else { |
428870ff | 3508 | if (zopt_verbose >= 7) { |
34dc7c2f BB |
3509 | (void) printf("writing offset %llx size %llx" |
3510 | " txg %llx\n", | |
3511 | (u_longlong_t)bigoff, | |
3512 | (u_longlong_t)bigsize, | |
3513 | (u_longlong_t)txg); | |
3514 | } | |
428870ff | 3515 | dmu_write(os, bigobj, bigoff, bigsize, bigbuf, tx); |
34dc7c2f BB |
3516 | } |
3517 | ||
3518 | dmu_tx_commit(tx); | |
3519 | ||
3520 | /* | |
3521 | * Sanity check the stuff we just wrote. | |
3522 | */ | |
3523 | { | |
3524 | void *packcheck = umem_alloc(packsize, UMEM_NOFAIL); | |
3525 | void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL); | |
3526 | ||
428870ff | 3527 | VERIFY(0 == dmu_read(os, packobj, packoff, |
9babb374 | 3528 | packsize, packcheck, DMU_READ_PREFETCH)); |
428870ff | 3529 | VERIFY(0 == dmu_read(os, bigobj, bigoff, |
9babb374 | 3530 | bigsize, bigcheck, DMU_READ_PREFETCH)); |
34dc7c2f BB |
3531 | |
3532 | ASSERT(bcmp(packbuf, packcheck, packsize) == 0); | |
3533 | ASSERT(bcmp(bigbuf, bigcheck, bigsize) == 0); | |
3534 | ||
3535 | umem_free(packcheck, packsize); | |
3536 | umem_free(bigcheck, bigsize); | |
3537 | } | |
3538 | ||
3539 | umem_free(packbuf, packsize); | |
3540 | umem_free(bigbuf, bigsize); | |
40b84e7a | 3541 | umem_free(od, size); |
34dc7c2f BB |
3542 | } |
3543 | ||
9babb374 BB |
3544 | void |
3545 | compare_and_update_pbbufs(uint64_t s, bufwad_t *packbuf, bufwad_t *bigbuf, | |
428870ff | 3546 | uint64_t bigsize, uint64_t n, uint64_t chunksize, uint64_t txg) |
9babb374 BB |
3547 | { |
3548 | uint64_t i; | |
3549 | bufwad_t *pack; | |
3550 | bufwad_t *bigH; | |
3551 | bufwad_t *bigT; | |
3552 | ||
3553 | /* | |
3554 | * For each index from n to n + s, verify that the existing bufwad | |
3555 | * in packobj matches the bufwads at the head and tail of the | |
3556 | * corresponding chunk in bigobj. Then update all three bufwads | |
3557 | * with the new values we want to write out. | |
3558 | */ | |
3559 | for (i = 0; i < s; i++) { | |
3560 | /* LINTED */ | |
3561 | pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t)); | |
3562 | /* LINTED */ | |
428870ff | 3563 | bigH = (bufwad_t *)((char *)bigbuf + i * chunksize); |
9babb374 | 3564 | /* LINTED */ |
428870ff | 3565 | bigT = (bufwad_t *)((char *)bigH + chunksize) - 1; |
9babb374 BB |
3566 | |
3567 | ASSERT((uintptr_t)bigH - (uintptr_t)bigbuf < bigsize); | |
3568 | ASSERT((uintptr_t)bigT - (uintptr_t)bigbuf < bigsize); | |
3569 | ||
3570 | if (pack->bw_txg > txg) | |
3571 | fatal(0, "future leak: got %llx, open txg is %llx", | |
3572 | pack->bw_txg, txg); | |
3573 | ||
3574 | if (pack->bw_data != 0 && pack->bw_index != n + i) | |
3575 | fatal(0, "wrong index: got %llx, wanted %llx+%llx", | |
3576 | pack->bw_index, n, i); | |
3577 | ||
3578 | if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0) | |
3579 | fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH); | |
3580 | ||
3581 | if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0) | |
3582 | fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT); | |
3583 | ||
3584 | pack->bw_index = n + i; | |
3585 | pack->bw_txg = txg; | |
3586 | pack->bw_data = 1 + ztest_random(-2ULL); | |
3587 | ||
3588 | *bigH = *pack; | |
3589 | *bigT = *pack; | |
3590 | } | |
3591 | } | |
3592 | ||
40b84e7a BB |
3593 | #undef OD_ARRAY_SIZE |
3594 | #define OD_ARRAY_SIZE 2 | |
3595 | ||
9babb374 | 3596 | void |
428870ff | 3597 | ztest_dmu_read_write_zcopy(ztest_ds_t *zd, uint64_t id) |
9babb374 | 3598 | { |
428870ff | 3599 | objset_t *os = zd->zd_os; |
40b84e7a | 3600 | ztest_od_t *od; |
9babb374 BB |
3601 | dmu_tx_t *tx; |
3602 | uint64_t i; | |
3603 | int error; | |
40b84e7a | 3604 | int size; |
9babb374 BB |
3605 | uint64_t n, s, txg; |
3606 | bufwad_t *packbuf, *bigbuf; | |
428870ff BB |
3607 | uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize; |
3608 | uint64_t blocksize = ztest_random_blocksize(); | |
3609 | uint64_t chunksize = blocksize; | |
9babb374 BB |
3610 | uint64_t regions = 997; |
3611 | uint64_t stride = 123456789ULL; | |
3612 | uint64_t width = 9; | |
3613 | dmu_buf_t *bonus_db; | |
3614 | arc_buf_t **bigbuf_arcbufs; | |
428870ff | 3615 | dmu_object_info_t doi; |
9babb374 | 3616 | |
40b84e7a BB |
3617 | size = sizeof(ztest_od_t) * OD_ARRAY_SIZE; |
3618 | od = umem_alloc(size, UMEM_NOFAIL); | |
3619 | ||
9babb374 BB |
3620 | /* |
3621 | * This test uses two objects, packobj and bigobj, that are always | |
3622 | * updated together (i.e. in the same tx) so that their contents are | |
3623 | * in sync and can be compared. Their contents relate to each other | |
3624 | * in a simple way: packobj is a dense array of 'bufwad' structures, | |
3625 | * while bigobj is a sparse array of the same bufwads. Specifically, | |
3626 | * for any index n, there are three bufwads that should be identical: | |
3627 | * | |
3628 | * packobj, at offset n * sizeof (bufwad_t) | |
3629 | * bigobj, at the head of the nth chunk | |
3630 | * bigobj, at the tail of the nth chunk | |
3631 | * | |
3632 | * The chunk size is set equal to bigobj block size so that | |
3633 | * dmu_assign_arcbuf() can be tested for object updates. | |
3634 | */ | |
3635 | ||
3636 | /* | |
3637 | * Read the directory info. If it's the first time, set things up. | |
3638 | */ | |
40b84e7a BB |
3639 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0); |
3640 | ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, chunksize); | |
3641 | ||
9babb374 | 3642 | |
40b84e7a BB |
3643 | if (ztest_object_init(zd, od, size, B_FALSE) != 0) { |
3644 | umem_free(od, size); | |
428870ff | 3645 | return; |
40b84e7a | 3646 | } |
9babb374 | 3647 | |
428870ff BB |
3648 | bigobj = od[0].od_object; |
3649 | packobj = od[1].od_object; | |
3650 | blocksize = od[0].od_blocksize; | |
3651 | chunksize = blocksize; | |
3652 | ASSERT(chunksize == od[1].od_gen); | |
9babb374 | 3653 | |
428870ff BB |
3654 | VERIFY(dmu_object_info(os, bigobj, &doi) == 0); |
3655 | VERIFY(ISP2(doi.doi_data_block_size)); | |
3656 | VERIFY(chunksize == doi.doi_data_block_size); | |
3657 | VERIFY(chunksize >= 2 * sizeof (bufwad_t)); | |
9babb374 BB |
3658 | |
3659 | /* | |
3660 | * Pick a random index and compute the offsets into packobj and bigobj. | |
3661 | */ | |
3662 | n = ztest_random(regions) * stride + ztest_random(width); | |
3663 | s = 1 + ztest_random(width - 1); | |
3664 | ||
3665 | packoff = n * sizeof (bufwad_t); | |
3666 | packsize = s * sizeof (bufwad_t); | |
3667 | ||
428870ff BB |
3668 | bigoff = n * chunksize; |
3669 | bigsize = s * chunksize; | |
9babb374 BB |
3670 | |
3671 | packbuf = umem_zalloc(packsize, UMEM_NOFAIL); | |
3672 | bigbuf = umem_zalloc(bigsize, UMEM_NOFAIL); | |
3673 | ||
428870ff | 3674 | VERIFY3U(0, ==, dmu_bonus_hold(os, bigobj, FTAG, &bonus_db)); |
9babb374 BB |
3675 | |
3676 | bigbuf_arcbufs = umem_zalloc(2 * s * sizeof (arc_buf_t *), UMEM_NOFAIL); | |
3677 | ||
3678 | /* | |
3679 | * Iteration 0 test zcopy for DB_UNCACHED dbufs. | |
3680 | * Iteration 1 test zcopy to already referenced dbufs. | |
3681 | * Iteration 2 test zcopy to dirty dbuf in the same txg. | |
3682 | * Iteration 3 test zcopy to dbuf dirty in previous txg. | |
3683 | * Iteration 4 test zcopy when dbuf is no longer dirty. | |
3684 | * Iteration 5 test zcopy when it can't be done. | |
3685 | * Iteration 6 one more zcopy write. | |
3686 | */ | |
3687 | for (i = 0; i < 7; i++) { | |
3688 | uint64_t j; | |
3689 | uint64_t off; | |
3690 | ||
3691 | /* | |
3692 | * In iteration 5 (i == 5) use arcbufs | |
3693 | * that don't match bigobj blksz to test | |
3694 | * dmu_assign_arcbuf() when it can't directly | |
3695 | * assign an arcbuf to a dbuf. | |
3696 | */ | |
3697 | for (j = 0; j < s; j++) { | |
3698 | if (i != 5) { | |
3699 | bigbuf_arcbufs[j] = | |
428870ff | 3700 | dmu_request_arcbuf(bonus_db, chunksize); |
9babb374 BB |
3701 | } else { |
3702 | bigbuf_arcbufs[2 * j] = | |
428870ff | 3703 | dmu_request_arcbuf(bonus_db, chunksize / 2); |
9babb374 | 3704 | bigbuf_arcbufs[2 * j + 1] = |
428870ff | 3705 | dmu_request_arcbuf(bonus_db, chunksize / 2); |
9babb374 BB |
3706 | } |
3707 | } | |
3708 | ||
3709 | /* | |
3710 | * Get a tx for the mods to both packobj and bigobj. | |
3711 | */ | |
3712 | tx = dmu_tx_create(os); | |
3713 | ||
428870ff BB |
3714 | dmu_tx_hold_write(tx, packobj, packoff, packsize); |
3715 | dmu_tx_hold_write(tx, bigobj, bigoff, bigsize); | |
9babb374 | 3716 | |
428870ff BB |
3717 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); |
3718 | if (txg == 0) { | |
9babb374 BB |
3719 | umem_free(packbuf, packsize); |
3720 | umem_free(bigbuf, bigsize); | |
3721 | for (j = 0; j < s; j++) { | |
3722 | if (i != 5) { | |
3723 | dmu_return_arcbuf(bigbuf_arcbufs[j]); | |
3724 | } else { | |
3725 | dmu_return_arcbuf( | |
3726 | bigbuf_arcbufs[2 * j]); | |
3727 | dmu_return_arcbuf( | |
3728 | bigbuf_arcbufs[2 * j + 1]); | |
3729 | } | |
3730 | } | |
3731 | umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *)); | |
40b84e7a | 3732 | umem_free(od, size); |
9babb374 BB |
3733 | dmu_buf_rele(bonus_db, FTAG); |
3734 | return; | |
3735 | } | |
3736 | ||
9babb374 BB |
3737 | /* |
3738 | * 50% of the time don't read objects in the 1st iteration to | |
3739 | * test dmu_assign_arcbuf() for the case when there're no | |
3740 | * existing dbufs for the specified offsets. | |
3741 | */ | |
3742 | if (i != 0 || ztest_random(2) != 0) { | |
428870ff | 3743 | error = dmu_read(os, packobj, packoff, |
9babb374 BB |
3744 | packsize, packbuf, DMU_READ_PREFETCH); |
3745 | ASSERT3U(error, ==, 0); | |
428870ff | 3746 | error = dmu_read(os, bigobj, bigoff, bigsize, |
9babb374 BB |
3747 | bigbuf, DMU_READ_PREFETCH); |
3748 | ASSERT3U(error, ==, 0); | |
3749 | } | |
3750 | compare_and_update_pbbufs(s, packbuf, bigbuf, bigsize, | |
428870ff | 3751 | n, chunksize, txg); |
9babb374 BB |
3752 | |
3753 | /* | |
3754 | * We've verified all the old bufwads, and made new ones. | |
3755 | * Now write them out. | |
3756 | */ | |
428870ff BB |
3757 | dmu_write(os, packobj, packoff, packsize, packbuf, tx); |
3758 | if (zopt_verbose >= 7) { | |
9babb374 BB |
3759 | (void) printf("writing offset %llx size %llx" |
3760 | " txg %llx\n", | |
3761 | (u_longlong_t)bigoff, | |
3762 | (u_longlong_t)bigsize, | |
3763 | (u_longlong_t)txg); | |
3764 | } | |
428870ff | 3765 | for (off = bigoff, j = 0; j < s; j++, off += chunksize) { |
9babb374 BB |
3766 | dmu_buf_t *dbt; |
3767 | if (i != 5) { | |
3768 | bcopy((caddr_t)bigbuf + (off - bigoff), | |
428870ff | 3769 | bigbuf_arcbufs[j]->b_data, chunksize); |
9babb374 BB |
3770 | } else { |
3771 | bcopy((caddr_t)bigbuf + (off - bigoff), | |
3772 | bigbuf_arcbufs[2 * j]->b_data, | |
428870ff | 3773 | chunksize / 2); |
9babb374 | 3774 | bcopy((caddr_t)bigbuf + (off - bigoff) + |
428870ff | 3775 | chunksize / 2, |
9babb374 | 3776 | bigbuf_arcbufs[2 * j + 1]->b_data, |
428870ff | 3777 | chunksize / 2); |
9babb374 BB |
3778 | } |
3779 | ||
3780 | if (i == 1) { | |
428870ff BB |
3781 | VERIFY(dmu_buf_hold(os, bigobj, off, |
3782 | FTAG, &dbt, DMU_READ_NO_PREFETCH) == 0); | |
9babb374 BB |
3783 | } |
3784 | if (i != 5) { | |
3785 | dmu_assign_arcbuf(bonus_db, off, | |
3786 | bigbuf_arcbufs[j], tx); | |
3787 | } else { | |
3788 | dmu_assign_arcbuf(bonus_db, off, | |
3789 | bigbuf_arcbufs[2 * j], tx); | |
3790 | dmu_assign_arcbuf(bonus_db, | |
428870ff | 3791 | off + chunksize / 2, |
9babb374 BB |
3792 | bigbuf_arcbufs[2 * j + 1], tx); |
3793 | } | |
3794 | if (i == 1) { | |
3795 | dmu_buf_rele(dbt, FTAG); | |
3796 | } | |
3797 | } | |
3798 | dmu_tx_commit(tx); | |
3799 | ||
3800 | /* | |
3801 | * Sanity check the stuff we just wrote. | |
3802 | */ | |
3803 | { | |
3804 | void *packcheck = umem_alloc(packsize, UMEM_NOFAIL); | |
3805 | void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL); | |
3806 | ||
428870ff | 3807 | VERIFY(0 == dmu_read(os, packobj, packoff, |
9babb374 | 3808 | packsize, packcheck, DMU_READ_PREFETCH)); |
428870ff | 3809 | VERIFY(0 == dmu_read(os, bigobj, bigoff, |
9babb374 BB |
3810 | bigsize, bigcheck, DMU_READ_PREFETCH)); |
3811 | ||
3812 | ASSERT(bcmp(packbuf, packcheck, packsize) == 0); | |
3813 | ASSERT(bcmp(bigbuf, bigcheck, bigsize) == 0); | |
3814 | ||
3815 | umem_free(packcheck, packsize); | |
3816 | umem_free(bigcheck, bigsize); | |
3817 | } | |
3818 | if (i == 2) { | |
3819 | txg_wait_open(dmu_objset_pool(os), 0); | |
3820 | } else if (i == 3) { | |
3821 | txg_wait_synced(dmu_objset_pool(os), 0); | |
3822 | } | |
3823 | } | |
3824 | ||
3825 | dmu_buf_rele(bonus_db, FTAG); | |
3826 | umem_free(packbuf, packsize); | |
3827 | umem_free(bigbuf, bigsize); | |
3828 | umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *)); | |
40b84e7a | 3829 | umem_free(od, size); |
9babb374 BB |
3830 | } |
3831 | ||
428870ff | 3832 | /* ARGSUSED */ |
34dc7c2f | 3833 | void |
428870ff | 3834 | ztest_dmu_write_parallel(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3835 | { |
40b84e7a BB |
3836 | ztest_od_t *od; |
3837 | ||
3838 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); | |
428870ff BB |
3839 | uint64_t offset = (1ULL << (ztest_random(20) + 43)) + |
3840 | (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
34dc7c2f BB |
3841 | |
3842 | /* | |
428870ff BB |
3843 | * Have multiple threads write to large offsets in an object |
3844 | * to verify that parallel writes to an object -- even to the | |
3845 | * same blocks within the object -- doesn't cause any trouble. | |
34dc7c2f | 3846 | */ |
40b84e7a | 3847 | ztest_od_init(od, ID_PARALLEL, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0); |
9babb374 | 3848 | |
40b84e7a | 3849 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) |
34dc7c2f | 3850 | return; |
34dc7c2f | 3851 | |
428870ff | 3852 | while (ztest_random(10) != 0) |
40b84e7a BB |
3853 | ztest_io(zd, od->od_object, offset); |
3854 | ||
3855 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff | 3856 | } |
34dc7c2f | 3857 | |
428870ff BB |
3858 | void |
3859 | ztest_dmu_prealloc(ztest_ds_t *zd, uint64_t id) | |
3860 | { | |
40b84e7a | 3861 | ztest_od_t *od; |
428870ff BB |
3862 | uint64_t offset = (1ULL << (ztest_random(4) + SPA_MAXBLOCKSHIFT)) + |
3863 | (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
3864 | uint64_t count = ztest_random(20) + 1; | |
3865 | uint64_t blocksize = ztest_random_blocksize(); | |
3866 | void *data; | |
34dc7c2f | 3867 | |
40b84e7a | 3868 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
34dc7c2f | 3869 | |
40b84e7a BB |
3870 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0); |
3871 | ||
3872 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), !ztest_random(2)) != 0) { | |
3873 | umem_free(od, sizeof(ztest_od_t)); | |
34dc7c2f | 3874 | return; |
40b84e7a | 3875 | } |
34dc7c2f | 3876 | |
40b84e7a BB |
3877 | if (ztest_truncate(zd, od->od_object, offset, count * blocksize) != 0) { |
3878 | umem_free(od, sizeof(ztest_od_t)); | |
34dc7c2f | 3879 | return; |
40b84e7a | 3880 | } |
34dc7c2f | 3881 | |
40b84e7a | 3882 | ztest_prealloc(zd, od->od_object, offset, count * blocksize); |
34dc7c2f | 3883 | |
428870ff | 3884 | data = umem_zalloc(blocksize, UMEM_NOFAIL); |
34dc7c2f | 3885 | |
428870ff BB |
3886 | while (ztest_random(count) != 0) { |
3887 | uint64_t randoff = offset + (ztest_random(count) * blocksize); | |
40b84e7a | 3888 | if (ztest_write(zd, od->od_object, randoff, blocksize, |
428870ff BB |
3889 | data) != 0) |
3890 | break; | |
3891 | while (ztest_random(4) != 0) | |
40b84e7a | 3892 | ztest_io(zd, od->od_object, randoff); |
9babb374 | 3893 | } |
34dc7c2f | 3894 | |
428870ff | 3895 | umem_free(data, blocksize); |
40b84e7a | 3896 | umem_free(od, sizeof(ztest_od_t)); |
34dc7c2f BB |
3897 | } |
3898 | ||
3899 | /* | |
3900 | * Verify that zap_{create,destroy,add,remove,update} work as expected. | |
3901 | */ | |
3902 | #define ZTEST_ZAP_MIN_INTS 1 | |
3903 | #define ZTEST_ZAP_MAX_INTS 4 | |
3904 | #define ZTEST_ZAP_MAX_PROPS 1000 | |
3905 | ||
3906 | void | |
428870ff | 3907 | ztest_zap(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3908 | { |
428870ff | 3909 | objset_t *os = zd->zd_os; |
40b84e7a | 3910 | ztest_od_t *od; |
34dc7c2f BB |
3911 | uint64_t object; |
3912 | uint64_t txg, last_txg; | |
3913 | uint64_t value[ZTEST_ZAP_MAX_INTS]; | |
3914 | uint64_t zl_ints, zl_intsize, prop; | |
3915 | int i, ints; | |
3916 | dmu_tx_t *tx; | |
3917 | char propname[100], txgname[100]; | |
3918 | int error; | |
34dc7c2f BB |
3919 | char *hc[2] = { "s.acl.h", ".s.open.h.hyLZlg" }; |
3920 | ||
40b84e7a BB |
3921 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
3922 | ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0); | |
34dc7c2f | 3923 | |
40b84e7a BB |
3924 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
3925 | !ztest_random(2)) != 0) | |
3926 | goto out; | |
34dc7c2f | 3927 | |
40b84e7a | 3928 | object = od->od_object; |
34dc7c2f | 3929 | |
428870ff BB |
3930 | /* |
3931 | * Generate a known hash collision, and verify that | |
3932 | * we can lookup and remove both entries. | |
3933 | */ | |
3934 | tx = dmu_tx_create(os); | |
3935 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); | |
3936 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
3937 | if (txg == 0) | |
40b84e7a | 3938 | goto out; |
428870ff BB |
3939 | for (i = 0; i < 2; i++) { |
3940 | value[i] = i; | |
3941 | VERIFY3U(0, ==, zap_add(os, object, hc[i], sizeof (uint64_t), | |
3942 | 1, &value[i], tx)); | |
3943 | } | |
3944 | for (i = 0; i < 2; i++) { | |
3945 | VERIFY3U(EEXIST, ==, zap_add(os, object, hc[i], | |
3946 | sizeof (uint64_t), 1, &value[i], tx)); | |
3947 | VERIFY3U(0, ==, | |
3948 | zap_length(os, object, hc[i], &zl_intsize, &zl_ints)); | |
3949 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
3950 | ASSERT3U(zl_ints, ==, 1); | |
3951 | } | |
3952 | for (i = 0; i < 2; i++) { | |
3953 | VERIFY3U(0, ==, zap_remove(os, object, hc[i], tx)); | |
34dc7c2f | 3954 | } |
428870ff | 3955 | dmu_tx_commit(tx); |
34dc7c2f | 3956 | |
428870ff BB |
3957 | /* |
3958 | * Generate a buch of random entries. | |
3959 | */ | |
34dc7c2f BB |
3960 | ints = MAX(ZTEST_ZAP_MIN_INTS, object % ZTEST_ZAP_MAX_INTS); |
3961 | ||
3962 | prop = ztest_random(ZTEST_ZAP_MAX_PROPS); | |
3963 | (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop); | |
3964 | (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop); | |
3965 | bzero(value, sizeof (value)); | |
3966 | last_txg = 0; | |
3967 | ||
3968 | /* | |
3969 | * If these zap entries already exist, validate their contents. | |
3970 | */ | |
3971 | error = zap_length(os, object, txgname, &zl_intsize, &zl_ints); | |
3972 | if (error == 0) { | |
3973 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
3974 | ASSERT3U(zl_ints, ==, 1); | |
3975 | ||
3976 | VERIFY(zap_lookup(os, object, txgname, zl_intsize, | |
3977 | zl_ints, &last_txg) == 0); | |
3978 | ||
3979 | VERIFY(zap_length(os, object, propname, &zl_intsize, | |
3980 | &zl_ints) == 0); | |
3981 | ||
3982 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
3983 | ASSERT3U(zl_ints, ==, ints); | |
3984 | ||
3985 | VERIFY(zap_lookup(os, object, propname, zl_intsize, | |
3986 | zl_ints, value) == 0); | |
3987 | ||
3988 | for (i = 0; i < ints; i++) { | |
3989 | ASSERT3U(value[i], ==, last_txg + object + i); | |
3990 | } | |
3991 | } else { | |
3992 | ASSERT3U(error, ==, ENOENT); | |
3993 | } | |
3994 | ||
3995 | /* | |
3996 | * Atomically update two entries in our zap object. | |
3997 | * The first is named txg_%llu, and contains the txg | |
3998 | * in which the property was last updated. The second | |
3999 | * is named prop_%llu, and the nth element of its value | |
4000 | * should be txg + object + n. | |
4001 | */ | |
4002 | tx = dmu_tx_create(os); | |
428870ff BB |
4003 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
4004 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4005 | if (txg == 0) | |
40b84e7a | 4006 | goto out; |
34dc7c2f BB |
4007 | |
4008 | if (last_txg > txg) | |
4009 | fatal(0, "zap future leak: old %llu new %llu", last_txg, txg); | |
4010 | ||
4011 | for (i = 0; i < ints; i++) | |
4012 | value[i] = txg + object + i; | |
4013 | ||
428870ff BB |
4014 | VERIFY3U(0, ==, zap_update(os, object, txgname, sizeof (uint64_t), |
4015 | 1, &txg, tx)); | |
4016 | VERIFY3U(0, ==, zap_update(os, object, propname, sizeof (uint64_t), | |
4017 | ints, value, tx)); | |
34dc7c2f BB |
4018 | |
4019 | dmu_tx_commit(tx); | |
4020 | ||
4021 | /* | |
4022 | * Remove a random pair of entries. | |
4023 | */ | |
4024 | prop = ztest_random(ZTEST_ZAP_MAX_PROPS); | |
4025 | (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop); | |
4026 | (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop); | |
4027 | ||
4028 | error = zap_length(os, object, txgname, &zl_intsize, &zl_ints); | |
4029 | ||
4030 | if (error == ENOENT) | |
40b84e7a | 4031 | goto out; |
34dc7c2f BB |
4032 | |
4033 | ASSERT3U(error, ==, 0); | |
4034 | ||
4035 | tx = dmu_tx_create(os); | |
428870ff BB |
4036 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
4037 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4038 | if (txg == 0) | |
40b84e7a | 4039 | goto out; |
428870ff BB |
4040 | VERIFY3U(0, ==, zap_remove(os, object, txgname, tx)); |
4041 | VERIFY3U(0, ==, zap_remove(os, object, propname, tx)); | |
4042 | dmu_tx_commit(tx); | |
40b84e7a BB |
4043 | out: |
4044 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff | 4045 | } |
34dc7c2f | 4046 | |
428870ff BB |
4047 | /* |
4048 | * Testcase to test the upgrading of a microzap to fatzap. | |
4049 | */ | |
4050 | void | |
4051 | ztest_fzap(ztest_ds_t *zd, uint64_t id) | |
4052 | { | |
4053 | objset_t *os = zd->zd_os; | |
40b84e7a | 4054 | ztest_od_t *od; |
428870ff | 4055 | uint64_t object, txg; |
d6320ddb | 4056 | int i; |
34dc7c2f | 4057 | |
40b84e7a BB |
4058 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
4059 | ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0); | |
428870ff | 4060 | |
40b84e7a BB |
4061 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
4062 | !ztest_random(2)) != 0) | |
4063 | goto out; | |
4064 | object = od->od_object; | |
34dc7c2f BB |
4065 | |
4066 | /* | |
428870ff BB |
4067 | * Add entries to this ZAP and make sure it spills over |
4068 | * and gets upgraded to a fatzap. Also, since we are adding | |
4069 | * 2050 entries we should see ptrtbl growth and leaf-block split. | |
34dc7c2f | 4070 | */ |
d6320ddb | 4071 | for (i = 0; i < 2050; i++) { |
428870ff BB |
4072 | char name[MAXNAMELEN]; |
4073 | uint64_t value = i; | |
4074 | dmu_tx_t *tx; | |
4075 | int error; | |
34dc7c2f | 4076 | |
428870ff | 4077 | (void) snprintf(name, sizeof (name), "fzap-%llu-%llu", |
b8864a23 | 4078 | (u_longlong_t)id, (u_longlong_t)value); |
428870ff BB |
4079 | |
4080 | tx = dmu_tx_create(os); | |
4081 | dmu_tx_hold_zap(tx, object, B_TRUE, name); | |
4082 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4083 | if (txg == 0) | |
40b84e7a | 4084 | goto out; |
428870ff BB |
4085 | error = zap_add(os, object, name, sizeof (uint64_t), 1, |
4086 | &value, tx); | |
4087 | ASSERT(error == 0 || error == EEXIST); | |
4088 | dmu_tx_commit(tx); | |
34dc7c2f | 4089 | } |
40b84e7a BB |
4090 | out: |
4091 | umem_free(od, sizeof(ztest_od_t)); | |
34dc7c2f BB |
4092 | } |
4093 | ||
428870ff | 4094 | /* ARGSUSED */ |
34dc7c2f | 4095 | void |
428870ff | 4096 | ztest_zap_parallel(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4097 | { |
428870ff | 4098 | objset_t *os = zd->zd_os; |
40b84e7a | 4099 | ztest_od_t *od; |
34dc7c2f BB |
4100 | uint64_t txg, object, count, wsize, wc, zl_wsize, zl_wc; |
4101 | dmu_tx_t *tx; | |
4102 | int i, namelen, error; | |
428870ff | 4103 | int micro = ztest_random(2); |
34dc7c2f BB |
4104 | char name[20], string_value[20]; |
4105 | void *data; | |
4106 | ||
40b84e7a BB |
4107 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
4108 | ztest_od_init(od, ID_PARALLEL, FTAG, micro, DMU_OT_ZAP_OTHER, 0, 0); | |
428870ff | 4109 | |
40b84e7a BB |
4110 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
4111 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff | 4112 | return; |
40b84e7a | 4113 | } |
428870ff | 4114 | |
40b84e7a | 4115 | object = od->od_object; |
428870ff | 4116 | |
34dc7c2f BB |
4117 | /* |
4118 | * Generate a random name of the form 'xxx.....' where each | |
4119 | * x is a random printable character and the dots are dots. | |
4120 | * There are 94 such characters, and the name length goes from | |
4121 | * 6 to 20, so there are 94^3 * 15 = 12,458,760 possible names. | |
4122 | */ | |
4123 | namelen = ztest_random(sizeof (name) - 5) + 5 + 1; | |
4124 | ||
4125 | for (i = 0; i < 3; i++) | |
4126 | name[i] = '!' + ztest_random('~' - '!' + 1); | |
4127 | for (; i < namelen - 1; i++) | |
4128 | name[i] = '.'; | |
4129 | name[i] = '\0'; | |
4130 | ||
428870ff | 4131 | if ((namelen & 1) || micro) { |
34dc7c2f BB |
4132 | wsize = sizeof (txg); |
4133 | wc = 1; | |
4134 | data = &txg; | |
4135 | } else { | |
4136 | wsize = 1; | |
4137 | wc = namelen; | |
4138 | data = string_value; | |
4139 | } | |
4140 | ||
4141 | count = -1ULL; | |
4142 | VERIFY(zap_count(os, object, &count) == 0); | |
4143 | ASSERT(count != -1ULL); | |
4144 | ||
4145 | /* | |
4146 | * Select an operation: length, lookup, add, update, remove. | |
4147 | */ | |
4148 | i = ztest_random(5); | |
4149 | ||
4150 | if (i >= 2) { | |
4151 | tx = dmu_tx_create(os); | |
428870ff BB |
4152 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
4153 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4154 | if (txg == 0) | |
34dc7c2f | 4155 | return; |
34dc7c2f BB |
4156 | bcopy(name, string_value, namelen); |
4157 | } else { | |
4158 | tx = NULL; | |
4159 | txg = 0; | |
4160 | bzero(string_value, namelen); | |
4161 | } | |
4162 | ||
4163 | switch (i) { | |
4164 | ||
4165 | case 0: | |
4166 | error = zap_length(os, object, name, &zl_wsize, &zl_wc); | |
4167 | if (error == 0) { | |
4168 | ASSERT3U(wsize, ==, zl_wsize); | |
4169 | ASSERT3U(wc, ==, zl_wc); | |
4170 | } else { | |
4171 | ASSERT3U(error, ==, ENOENT); | |
4172 | } | |
4173 | break; | |
4174 | ||
4175 | case 1: | |
4176 | error = zap_lookup(os, object, name, wsize, wc, data); | |
4177 | if (error == 0) { | |
4178 | if (data == string_value && | |
4179 | bcmp(name, data, namelen) != 0) | |
4180 | fatal(0, "name '%s' != val '%s' len %d", | |
4181 | name, data, namelen); | |
4182 | } else { | |
4183 | ASSERT3U(error, ==, ENOENT); | |
4184 | } | |
4185 | break; | |
4186 | ||
4187 | case 2: | |
4188 | error = zap_add(os, object, name, wsize, wc, data, tx); | |
4189 | ASSERT(error == 0 || error == EEXIST); | |
4190 | break; | |
4191 | ||
4192 | case 3: | |
4193 | VERIFY(zap_update(os, object, name, wsize, wc, data, tx) == 0); | |
4194 | break; | |
4195 | ||
4196 | case 4: | |
4197 | error = zap_remove(os, object, name, tx); | |
4198 | ASSERT(error == 0 || error == ENOENT); | |
4199 | break; | |
4200 | } | |
4201 | ||
4202 | if (tx != NULL) | |
4203 | dmu_tx_commit(tx); | |
40b84e7a BB |
4204 | |
4205 | umem_free(od, sizeof(ztest_od_t)); | |
34dc7c2f BB |
4206 | } |
4207 | ||
428870ff BB |
4208 | /* |
4209 | * Commit callback data. | |
4210 | */ | |
4211 | typedef struct ztest_cb_data { | |
4212 | list_node_t zcd_node; | |
4213 | uint64_t zcd_txg; | |
4214 | int zcd_expected_err; | |
4215 | boolean_t zcd_added; | |
4216 | boolean_t zcd_called; | |
4217 | spa_t *zcd_spa; | |
4218 | } ztest_cb_data_t; | |
4219 | ||
4220 | /* This is the actual commit callback function */ | |
4221 | static void | |
4222 | ztest_commit_callback(void *arg, int error) | |
4223 | { | |
4224 | ztest_cb_data_t *data = arg; | |
4225 | uint64_t synced_txg; | |
4226 | ||
4227 | VERIFY(data != NULL); | |
4228 | VERIFY3S(data->zcd_expected_err, ==, error); | |
4229 | VERIFY(!data->zcd_called); | |
4230 | ||
4231 | synced_txg = spa_last_synced_txg(data->zcd_spa); | |
4232 | if (data->zcd_txg > synced_txg) | |
4233 | fatal(0, "commit callback of txg %" PRIu64 " called prematurely" | |
4234 | ", last synced txg = %" PRIu64 "\n", data->zcd_txg, | |
4235 | synced_txg); | |
4236 | ||
4237 | data->zcd_called = B_TRUE; | |
4238 | ||
4239 | if (error == ECANCELED) { | |
4240 | ASSERT3U(data->zcd_txg, ==, 0); | |
4241 | ASSERT(!data->zcd_added); | |
4242 | ||
4243 | /* | |
4244 | * The private callback data should be destroyed here, but | |
4245 | * since we are going to check the zcd_called field after | |
4246 | * dmu_tx_abort(), we will destroy it there. | |
4247 | */ | |
4248 | return; | |
4249 | } | |
4250 | ||
090ff092 | 4251 | ASSERT(data->zcd_added); |
428870ff BB |
4252 | ASSERT3U(data->zcd_txg, !=, 0); |
4253 | ||
1e33ac1e | 4254 | (void) mutex_enter(&zcl.zcl_callbacks_lock); |
090ff092 RC |
4255 | |
4256 | /* See if this cb was called more quickly */ | |
4257 | if ((synced_txg - data->zcd_txg) < zc_min_txg_delay) | |
4258 | zc_min_txg_delay = synced_txg - data->zcd_txg; | |
4259 | ||
4260 | /* Remove our callback from the list */ | |
428870ff | 4261 | list_remove(&zcl.zcl_callbacks, data); |
090ff092 | 4262 | |
1e33ac1e | 4263 | (void) mutex_exit(&zcl.zcl_callbacks_lock); |
428870ff | 4264 | |
428870ff BB |
4265 | umem_free(data, sizeof (ztest_cb_data_t)); |
4266 | } | |
4267 | ||
4268 | /* Allocate and initialize callback data structure */ | |
4269 | static ztest_cb_data_t * | |
4270 | ztest_create_cb_data(objset_t *os, uint64_t txg) | |
4271 | { | |
4272 | ztest_cb_data_t *cb_data; | |
4273 | ||
4274 | cb_data = umem_zalloc(sizeof (ztest_cb_data_t), UMEM_NOFAIL); | |
4275 | ||
4276 | cb_data->zcd_txg = txg; | |
4277 | cb_data->zcd_spa = dmu_objset_spa(os); | |
1e33ac1e | 4278 | list_link_init(&cb_data->zcd_node); |
428870ff BB |
4279 | |
4280 | return (cb_data); | |
4281 | } | |
4282 | ||
428870ff BB |
4283 | /* |
4284 | * Commit callback test. | |
4285 | */ | |
34dc7c2f | 4286 | void |
428870ff BB |
4287 | ztest_dmu_commit_callbacks(ztest_ds_t *zd, uint64_t id) |
4288 | { | |
4289 | objset_t *os = zd->zd_os; | |
40b84e7a | 4290 | ztest_od_t *od; |
428870ff BB |
4291 | dmu_tx_t *tx; |
4292 | ztest_cb_data_t *cb_data[3], *tmp_cb; | |
4293 | uint64_t old_txg, txg; | |
090ff092 | 4294 | int i, error = 0; |
428870ff | 4295 | |
40b84e7a BB |
4296 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
4297 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0); | |
428870ff | 4298 | |
40b84e7a BB |
4299 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
4300 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff | 4301 | return; |
40b84e7a | 4302 | } |
428870ff BB |
4303 | |
4304 | tx = dmu_tx_create(os); | |
4305 | ||
4306 | cb_data[0] = ztest_create_cb_data(os, 0); | |
4307 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[0]); | |
4308 | ||
40b84e7a | 4309 | dmu_tx_hold_write(tx, od->od_object, 0, sizeof (uint64_t)); |
428870ff BB |
4310 | |
4311 | /* Every once in a while, abort the transaction on purpose */ | |
4312 | if (ztest_random(100) == 0) | |
4313 | error = -1; | |
4314 | ||
4315 | if (!error) | |
4316 | error = dmu_tx_assign(tx, TXG_NOWAIT); | |
4317 | ||
4318 | txg = error ? 0 : dmu_tx_get_txg(tx); | |
4319 | ||
4320 | cb_data[0]->zcd_txg = txg; | |
4321 | cb_data[1] = ztest_create_cb_data(os, txg); | |
4322 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[1]); | |
4323 | ||
4324 | if (error) { | |
4325 | /* | |
4326 | * It's not a strict requirement to call the registered | |
4327 | * callbacks from inside dmu_tx_abort(), but that's what | |
4328 | * it's supposed to happen in the current implementation | |
4329 | * so we will check for that. | |
4330 | */ | |
4331 | for (i = 0; i < 2; i++) { | |
4332 | cb_data[i]->zcd_expected_err = ECANCELED; | |
4333 | VERIFY(!cb_data[i]->zcd_called); | |
4334 | } | |
4335 | ||
4336 | dmu_tx_abort(tx); | |
4337 | ||
4338 | for (i = 0; i < 2; i++) { | |
4339 | VERIFY(cb_data[i]->zcd_called); | |
4340 | umem_free(cb_data[i], sizeof (ztest_cb_data_t)); | |
4341 | } | |
4342 | ||
40b84e7a | 4343 | umem_free(od, sizeof(ztest_od_t)); |
428870ff BB |
4344 | return; |
4345 | } | |
4346 | ||
4347 | cb_data[2] = ztest_create_cb_data(os, txg); | |
4348 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[2]); | |
4349 | ||
4350 | /* | |
4351 | * Read existing data to make sure there isn't a future leak. | |
4352 | */ | |
40b84e7a | 4353 | VERIFY(0 == dmu_read(os, od->od_object, 0, sizeof (uint64_t), |
428870ff BB |
4354 | &old_txg, DMU_READ_PREFETCH)); |
4355 | ||
4356 | if (old_txg > txg) | |
4357 | fatal(0, "future leak: got %" PRIu64 ", open txg is %" PRIu64, | |
4358 | old_txg, txg); | |
4359 | ||
40b84e7a | 4360 | dmu_write(os, od->od_object, 0, sizeof (uint64_t), &txg, tx); |
428870ff | 4361 | |
1e33ac1e | 4362 | (void) mutex_enter(&zcl.zcl_callbacks_lock); |
428870ff BB |
4363 | |
4364 | /* | |
4365 | * Since commit callbacks don't have any ordering requirement and since | |
4366 | * it is theoretically possible for a commit callback to be called | |
4367 | * after an arbitrary amount of time has elapsed since its txg has been | |
4368 | * synced, it is difficult to reliably determine whether a commit | |
4369 | * callback hasn't been called due to high load or due to a flawed | |
4370 | * implementation. | |
4371 | * | |
4372 | * In practice, we will assume that if after a certain number of txgs a | |
4373 | * commit callback hasn't been called, then most likely there's an | |
4374 | * implementation bug.. | |
4375 | */ | |
4376 | tmp_cb = list_head(&zcl.zcl_callbacks); | |
4377 | if (tmp_cb != NULL && | |
090ff092 | 4378 | tmp_cb->zcd_txg + ZTEST_COMMIT_CB_THRESH < txg) { |
428870ff BB |
4379 | fatal(0, "Commit callback threshold exceeded, oldest txg: %" |
4380 | PRIu64 ", open txg: %" PRIu64 "\n", tmp_cb->zcd_txg, txg); | |
4381 | } | |
4382 | ||
4383 | /* | |
4384 | * Let's find the place to insert our callbacks. | |
4385 | * | |
4386 | * Even though the list is ordered by txg, it is possible for the | |
4387 | * insertion point to not be the end because our txg may already be | |
4388 | * quiescing at this point and other callbacks in the open txg | |
4389 | * (from other objsets) may have sneaked in. | |
4390 | */ | |
4391 | tmp_cb = list_tail(&zcl.zcl_callbacks); | |
4392 | while (tmp_cb != NULL && tmp_cb->zcd_txg > txg) | |
4393 | tmp_cb = list_prev(&zcl.zcl_callbacks, tmp_cb); | |
4394 | ||
4395 | /* Add the 3 callbacks to the list */ | |
4396 | for (i = 0; i < 3; i++) { | |
4397 | if (tmp_cb == NULL) | |
4398 | list_insert_head(&zcl.zcl_callbacks, cb_data[i]); | |
4399 | else | |
4400 | list_insert_after(&zcl.zcl_callbacks, tmp_cb, | |
4401 | cb_data[i]); | |
4402 | ||
4403 | cb_data[i]->zcd_added = B_TRUE; | |
4404 | VERIFY(!cb_data[i]->zcd_called); | |
4405 | ||
4406 | tmp_cb = cb_data[i]; | |
4407 | } | |
4408 | ||
090ff092 RC |
4409 | zc_cb_counter += 3; |
4410 | ||
1e33ac1e | 4411 | (void) mutex_exit(&zcl.zcl_callbacks_lock); |
428870ff BB |
4412 | |
4413 | dmu_tx_commit(tx); | |
40b84e7a BB |
4414 | |
4415 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff BB |
4416 | } |
4417 | ||
4418 | /* ARGSUSED */ | |
4419 | void | |
4420 | ztest_dsl_prop_get_set(ztest_ds_t *zd, uint64_t id) | |
4421 | { | |
4422 | zfs_prop_t proplist[] = { | |
4423 | ZFS_PROP_CHECKSUM, | |
4424 | ZFS_PROP_COMPRESSION, | |
4425 | ZFS_PROP_COPIES, | |
4426 | ZFS_PROP_DEDUP | |
4427 | }; | |
4428 | ztest_shared_t *zs = ztest_shared; | |
d6320ddb | 4429 | int p; |
428870ff | 4430 | |
1e33ac1e | 4431 | (void) rw_enter(&zs->zs_name_lock, RW_READER); |
428870ff | 4432 | |
d6320ddb | 4433 | for (p = 0; p < sizeof (proplist) / sizeof (proplist[0]); p++) |
428870ff BB |
4434 | (void) ztest_dsl_prop_set_uint64(zd->zd_name, proplist[p], |
4435 | ztest_random_dsl_prop(proplist[p]), (int)ztest_random(2)); | |
4436 | ||
1e33ac1e | 4437 | (void) rw_exit(&zs->zs_name_lock); |
428870ff BB |
4438 | } |
4439 | ||
4440 | /* ARGSUSED */ | |
4441 | void | |
4442 | ztest_spa_prop_get_set(ztest_ds_t *zd, uint64_t id) | |
4443 | { | |
4444 | ztest_shared_t *zs = ztest_shared; | |
4445 | nvlist_t *props = NULL; | |
4446 | ||
1e33ac1e | 4447 | (void) rw_enter(&zs->zs_name_lock, RW_READER); |
428870ff BB |
4448 | |
4449 | (void) ztest_spa_prop_set_uint64(zs, ZPOOL_PROP_DEDUPDITTO, | |
4450 | ZIO_DEDUPDITTO_MIN + ztest_random(ZIO_DEDUPDITTO_MIN)); | |
4451 | ||
4452 | VERIFY3U(spa_prop_get(zs->zs_spa, &props), ==, 0); | |
4453 | ||
4454 | if (zopt_verbose >= 6) | |
4455 | dump_nvlist(props, 4); | |
4456 | ||
4457 | nvlist_free(props); | |
4458 | ||
1e33ac1e | 4459 | (void) rw_exit(&zs->zs_name_lock); |
428870ff BB |
4460 | } |
4461 | ||
4462 | /* | |
4463 | * Test snapshot hold/release and deferred destroy. | |
4464 | */ | |
4465 | void | |
4466 | ztest_dmu_snapshot_hold(ztest_ds_t *zd, uint64_t id) | |
34dc7c2f | 4467 | { |
34dc7c2f | 4468 | int error; |
428870ff BB |
4469 | objset_t *os = zd->zd_os; |
4470 | objset_t *origin; | |
4471 | char snapname[100]; | |
4472 | char fullname[100]; | |
4473 | char clonename[100]; | |
4474 | char tag[100]; | |
4475 | char osname[MAXNAMELEN]; | |
34dc7c2f | 4476 | |
1e33ac1e | 4477 | (void) rw_enter(&ztest_shared->zs_name_lock, RW_READER); |
34dc7c2f BB |
4478 | |
4479 | dmu_objset_name(os, osname); | |
4480 | ||
1e33ac1e | 4481 | (void) snprintf(snapname, 100, "sh1_%llu", (u_longlong_t)id); |
428870ff | 4482 | (void) snprintf(fullname, 100, "%s@%s", osname, snapname); |
1e33ac1e BB |
4483 | (void) snprintf(clonename, 100, "%s/ch1_%llu",osname,(u_longlong_t)id); |
4484 | (void) snprintf(tag, 100, "tag_%llu", (u_longlong_t)id); | |
428870ff BB |
4485 | |
4486 | /* | |
4487 | * Clean up from any previous run. | |
4488 | */ | |
4489 | (void) dmu_objset_destroy(clonename, B_FALSE); | |
4490 | (void) dsl_dataset_user_release(osname, snapname, tag, B_FALSE); | |
4491 | (void) dmu_objset_destroy(fullname, B_FALSE); | |
4492 | ||
4493 | /* | |
4494 | * Create snapshot, clone it, mark snap for deferred destroy, | |
4495 | * destroy clone, verify snap was also destroyed. | |
4496 | */ | |
572e2857 BB |
4497 | error = dmu_objset_snapshot(osname, snapname, NULL, NULL, FALSE, |
4498 | FALSE, -1); | |
428870ff BB |
4499 | if (error) { |
4500 | if (error == ENOSPC) { | |
4501 | ztest_record_enospc("dmu_objset_snapshot"); | |
4502 | goto out; | |
34dc7c2f | 4503 | } |
428870ff BB |
4504 | fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error); |
4505 | } | |
34dc7c2f | 4506 | |
428870ff BB |
4507 | error = dmu_objset_hold(fullname, FTAG, &origin); |
4508 | if (error) | |
4509 | fatal(0, "dmu_objset_hold(%s) = %d", fullname, error); | |
34dc7c2f | 4510 | |
428870ff BB |
4511 | error = dmu_objset_clone(clonename, dmu_objset_ds(origin), 0); |
4512 | dmu_objset_rele(origin, FTAG); | |
4513 | if (error) { | |
34dc7c2f | 4514 | if (error == ENOSPC) { |
428870ff BB |
4515 | ztest_record_enospc("dmu_objset_clone"); |
4516 | goto out; | |
34dc7c2f | 4517 | } |
428870ff BB |
4518 | fatal(0, "dmu_objset_clone(%s) = %d", clonename, error); |
4519 | } | |
34dc7c2f | 4520 | |
428870ff BB |
4521 | error = dmu_objset_destroy(fullname, B_TRUE); |
4522 | if (error) { | |
4523 | fatal(0, "dmu_objset_destroy(%s, B_TRUE) = %d", | |
4524 | fullname, error); | |
4525 | } | |
34dc7c2f | 4526 | |
428870ff BB |
4527 | error = dmu_objset_destroy(clonename, B_FALSE); |
4528 | if (error) | |
4529 | fatal(0, "dmu_objset_destroy(%s) = %d", clonename, error); | |
34dc7c2f | 4530 | |
428870ff BB |
4531 | error = dmu_objset_hold(fullname, FTAG, &origin); |
4532 | if (error != ENOENT) | |
4533 | fatal(0, "dmu_objset_hold(%s) = %d", fullname, error); | |
34dc7c2f | 4534 | |
428870ff BB |
4535 | /* |
4536 | * Create snapshot, add temporary hold, verify that we can't | |
4537 | * destroy a held snapshot, mark for deferred destroy, | |
4538 | * release hold, verify snapshot was destroyed. | |
4539 | */ | |
572e2857 BB |
4540 | error = dmu_objset_snapshot(osname, snapname, NULL, NULL, FALSE, |
4541 | FALSE, -1); | |
428870ff BB |
4542 | if (error) { |
4543 | if (error == ENOSPC) { | |
4544 | ztest_record_enospc("dmu_objset_snapshot"); | |
4545 | goto out; | |
34dc7c2f | 4546 | } |
428870ff BB |
4547 | fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error); |
4548 | } | |
4549 | ||
572e2857 BB |
4550 | error = dsl_dataset_user_hold(osname, snapname, tag, B_FALSE, |
4551 | B_TRUE, -1); | |
428870ff BB |
4552 | if (error) |
4553 | fatal(0, "dsl_dataset_user_hold(%s)", fullname, tag); | |
4554 | ||
4555 | error = dmu_objset_destroy(fullname, B_FALSE); | |
4556 | if (error != EBUSY) { | |
4557 | fatal(0, "dmu_objset_destroy(%s, B_FALSE) = %d", | |
4558 | fullname, error); | |
4559 | } | |
4560 | ||
4561 | error = dmu_objset_destroy(fullname, B_TRUE); | |
4562 | if (error) { | |
4563 | fatal(0, "dmu_objset_destroy(%s, B_TRUE) = %d", | |
4564 | fullname, error); | |
34dc7c2f BB |
4565 | } |
4566 | ||
428870ff BB |
4567 | error = dsl_dataset_user_release(osname, snapname, tag, B_FALSE); |
4568 | if (error) | |
4569 | fatal(0, "dsl_dataset_user_release(%s)", fullname, tag); | |
4570 | ||
4571 | VERIFY(dmu_objset_hold(fullname, FTAG, &origin) == ENOENT); | |
4572 | ||
4573 | out: | |
1e33ac1e | 4574 | (void) rw_exit(&ztest_shared->zs_name_lock); |
34dc7c2f BB |
4575 | } |
4576 | ||
34dc7c2f BB |
4577 | /* |
4578 | * Inject random faults into the on-disk data. | |
4579 | */ | |
428870ff | 4580 | /* ARGSUSED */ |
34dc7c2f | 4581 | void |
428870ff | 4582 | ztest_fault_inject(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4583 | { |
428870ff BB |
4584 | ztest_shared_t *zs = ztest_shared; |
4585 | spa_t *spa = zs->zs_spa; | |
34dc7c2f BB |
4586 | int fd; |
4587 | uint64_t offset; | |
428870ff | 4588 | uint64_t leaves; |
c5b3a7bb | 4589 | uint64_t bad = 0x1990c0ffeedecadeull; |
34dc7c2f | 4590 | uint64_t top, leaf; |
40b84e7a BB |
4591 | char *path0; |
4592 | char *pathrand; | |
34dc7c2f | 4593 | size_t fsize; |
34dc7c2f BB |
4594 | int bshift = SPA_MAXBLOCKSHIFT + 2; /* don't scrog all labels */ |
4595 | int iters = 1000; | |
428870ff BB |
4596 | int maxfaults; |
4597 | int mirror_save; | |
b128c09f | 4598 | vdev_t *vd0 = NULL; |
34dc7c2f | 4599 | uint64_t guid0 = 0; |
428870ff BB |
4600 | boolean_t islog = B_FALSE; |
4601 | ||
40b84e7a BB |
4602 | path0 = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
4603 | pathrand = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
4604 | ||
1e33ac1e | 4605 | mutex_enter(&zs->zs_vdev_lock); |
428870ff BB |
4606 | maxfaults = MAXFAULTS(); |
4607 | leaves = MAX(zs->zs_mirrors, 1) * zopt_raidz; | |
4608 | mirror_save = zs->zs_mirrors; | |
1e33ac1e | 4609 | mutex_exit(&zs->zs_vdev_lock); |
34dc7c2f | 4610 | |
b128c09f | 4611 | ASSERT(leaves >= 1); |
34dc7c2f BB |
4612 | |
4613 | /* | |
b128c09f | 4614 | * We need SCL_STATE here because we're going to look at vd0->vdev_tsd. |
34dc7c2f | 4615 | */ |
b128c09f | 4616 | spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); |
34dc7c2f | 4617 | |
b128c09f BB |
4618 | if (ztest_random(2) == 0) { |
4619 | /* | |
428870ff | 4620 | * Inject errors on a normal data device or slog device. |
b128c09f | 4621 | */ |
428870ff BB |
4622 | top = ztest_random_vdev_top(spa, B_TRUE); |
4623 | leaf = ztest_random(leaves) + zs->zs_splits; | |
34dc7c2f | 4624 | |
b128c09f BB |
4625 | /* |
4626 | * Generate paths to the first leaf in this top-level vdev, | |
4627 | * and to the random leaf we selected. We'll induce transient | |
4628 | * write failures and random online/offline activity on leaf 0, | |
4629 | * and we'll write random garbage to the randomly chosen leaf. | |
4630 | */ | |
4631 | (void) snprintf(path0, sizeof (path0), ztest_dev_template, | |
428870ff | 4632 | zopt_dir, zopt_pool, top * leaves + zs->zs_splits); |
b128c09f BB |
4633 | (void) snprintf(pathrand, sizeof (pathrand), ztest_dev_template, |
4634 | zopt_dir, zopt_pool, top * leaves + leaf); | |
34dc7c2f | 4635 | |
b128c09f | 4636 | vd0 = vdev_lookup_by_path(spa->spa_root_vdev, path0); |
428870ff BB |
4637 | if (vd0 != NULL && vd0->vdev_top->vdev_islog) |
4638 | islog = B_TRUE; | |
4639 | ||
b128c09f BB |
4640 | if (vd0 != NULL && maxfaults != 1) { |
4641 | /* | |
4642 | * Make vd0 explicitly claim to be unreadable, | |
4643 | * or unwriteable, or reach behind its back | |
4644 | * and close the underlying fd. We can do this if | |
4645 | * maxfaults == 0 because we'll fail and reexecute, | |
4646 | * and we can do it if maxfaults >= 2 because we'll | |
4647 | * have enough redundancy. If maxfaults == 1, the | |
4648 | * combination of this with injection of random data | |
4649 | * corruption below exceeds the pool's fault tolerance. | |
4650 | */ | |
4651 | vdev_file_t *vf = vd0->vdev_tsd; | |
4652 | ||
4653 | if (vf != NULL && ztest_random(3) == 0) { | |
4654 | (void) close(vf->vf_vnode->v_fd); | |
4655 | vf->vf_vnode->v_fd = -1; | |
4656 | } else if (ztest_random(2) == 0) { | |
4657 | vd0->vdev_cant_read = B_TRUE; | |
4658 | } else { | |
4659 | vd0->vdev_cant_write = B_TRUE; | |
4660 | } | |
4661 | guid0 = vd0->vdev_guid; | |
4662 | } | |
4663 | } else { | |
4664 | /* | |
4665 | * Inject errors on an l2cache device. | |
4666 | */ | |
4667 | spa_aux_vdev_t *sav = &spa->spa_l2cache; | |
34dc7c2f | 4668 | |
b128c09f BB |
4669 | if (sav->sav_count == 0) { |
4670 | spa_config_exit(spa, SCL_STATE, FTAG); | |
40b84e7a | 4671 | goto out; |
b128c09f BB |
4672 | } |
4673 | vd0 = sav->sav_vdevs[ztest_random(sav->sav_count)]; | |
34dc7c2f | 4674 | guid0 = vd0->vdev_guid; |
b128c09f BB |
4675 | (void) strcpy(path0, vd0->vdev_path); |
4676 | (void) strcpy(pathrand, vd0->vdev_path); | |
4677 | ||
4678 | leaf = 0; | |
4679 | leaves = 1; | |
4680 | maxfaults = INT_MAX; /* no limit on cache devices */ | |
34dc7c2f BB |
4681 | } |
4682 | ||
b128c09f BB |
4683 | spa_config_exit(spa, SCL_STATE, FTAG); |
4684 | ||
34dc7c2f | 4685 | /* |
428870ff BB |
4686 | * If we can tolerate two or more faults, or we're dealing |
4687 | * with a slog, randomly online/offline vd0. | |
34dc7c2f | 4688 | */ |
428870ff | 4689 | if ((maxfaults >= 2 || islog) && guid0 != 0) { |
fb5f0bc8 BB |
4690 | if (ztest_random(10) < 6) { |
4691 | int flags = (ztest_random(2) == 0 ? | |
4692 | ZFS_OFFLINE_TEMPORARY : 0); | |
428870ff BB |
4693 | |
4694 | /* | |
4695 | * We have to grab the zs_name_lock as writer to | |
4696 | * prevent a race between offlining a slog and | |
4697 | * destroying a dataset. Offlining the slog will | |
4698 | * grab a reference on the dataset which may cause | |
4699 | * dmu_objset_destroy() to fail with EBUSY thus | |
4700 | * leaving the dataset in an inconsistent state. | |
4701 | */ | |
4702 | if (islog) | |
1e33ac1e BB |
4703 | (void) rw_enter(&ztest_shared->zs_name_lock, |
4704 | RW_WRITER); | |
428870ff | 4705 | |
fb5f0bc8 | 4706 | VERIFY(vdev_offline(spa, guid0, flags) != EBUSY); |
428870ff BB |
4707 | |
4708 | if (islog) | |
1e33ac1e | 4709 | (void) rw_exit(&ztest_shared->zs_name_lock); |
fb5f0bc8 BB |
4710 | } else { |
4711 | (void) vdev_online(spa, guid0, 0, NULL); | |
4712 | } | |
34dc7c2f BB |
4713 | } |
4714 | ||
428870ff | 4715 | if (maxfaults == 0) |
40b84e7a | 4716 | goto out; |
428870ff | 4717 | |
34dc7c2f BB |
4718 | /* |
4719 | * We have at least single-fault tolerance, so inject data corruption. | |
4720 | */ | |
4721 | fd = open(pathrand, O_RDWR); | |
4722 | ||
4723 | if (fd == -1) /* we hit a gap in the device namespace */ | |
40b84e7a | 4724 | goto out; |
34dc7c2f BB |
4725 | |
4726 | fsize = lseek(fd, 0, SEEK_END); | |
4727 | ||
4728 | while (--iters != 0) { | |
4729 | offset = ztest_random(fsize / (leaves << bshift)) * | |
4730 | (leaves << bshift) + (leaf << bshift) + | |
4731 | (ztest_random(1ULL << (bshift - 1)) & -8ULL); | |
4732 | ||
4733 | if (offset >= fsize) | |
4734 | continue; | |
4735 | ||
1e33ac1e | 4736 | mutex_enter(&zs->zs_vdev_lock); |
428870ff | 4737 | if (mirror_save != zs->zs_mirrors) { |
1e33ac1e | 4738 | mutex_exit(&zs->zs_vdev_lock); |
428870ff | 4739 | (void) close(fd); |
40b84e7a | 4740 | goto out; |
428870ff | 4741 | } |
34dc7c2f BB |
4742 | |
4743 | if (pwrite(fd, &bad, sizeof (bad), offset) != sizeof (bad)) | |
4744 | fatal(1, "can't inject bad word at 0x%llx in %s", | |
4745 | offset, pathrand); | |
428870ff | 4746 | |
1e33ac1e | 4747 | mutex_exit(&zs->zs_vdev_lock); |
428870ff BB |
4748 | |
4749 | if (zopt_verbose >= 7) | |
4750 | (void) printf("injected bad word into %s," | |
4751 | " offset 0x%llx\n", pathrand, (u_longlong_t)offset); | |
34dc7c2f BB |
4752 | } |
4753 | ||
4754 | (void) close(fd); | |
40b84e7a BB |
4755 | out: |
4756 | umem_free(path0, MAXPATHLEN); | |
4757 | umem_free(pathrand, MAXPATHLEN); | |
34dc7c2f BB |
4758 | } |
4759 | ||
4760 | /* | |
428870ff | 4761 | * Verify that DDT repair works as expected. |
34dc7c2f BB |
4762 | */ |
4763 | void | |
428870ff | 4764 | ztest_ddt_repair(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4765 | { |
428870ff BB |
4766 | ztest_shared_t *zs = ztest_shared; |
4767 | spa_t *spa = zs->zs_spa; | |
4768 | objset_t *os = zd->zd_os; | |
40b84e7a | 4769 | ztest_od_t *od; |
428870ff BB |
4770 | uint64_t object, blocksize, txg, pattern, psize; |
4771 | enum zio_checksum checksum = spa_dedup_checksum(spa); | |
4772 | dmu_buf_t *db; | |
4773 | dmu_tx_t *tx; | |
4774 | void *buf; | |
4775 | blkptr_t blk; | |
4776 | int copies = 2 * ZIO_DEDUPDITTO_MIN; | |
d6320ddb | 4777 | int i; |
34dc7c2f | 4778 | |
428870ff BB |
4779 | blocksize = ztest_random_blocksize(); |
4780 | blocksize = MIN(blocksize, 2048); /* because we write so many */ | |
34dc7c2f | 4781 | |
40b84e7a BB |
4782 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
4783 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0); | |
34dc7c2f | 4784 | |
40b84e7a BB |
4785 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
4786 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff | 4787 | return; |
40b84e7a | 4788 | } |
34dc7c2f BB |
4789 | |
4790 | /* | |
428870ff BB |
4791 | * Take the name lock as writer to prevent anyone else from changing |
4792 | * the pool and dataset properies we need to maintain during this test. | |
34dc7c2f | 4793 | */ |
1e33ac1e | 4794 | (void) rw_enter(&zs->zs_name_lock, RW_WRITER); |
34dc7c2f | 4795 | |
428870ff BB |
4796 | if (ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_DEDUP, checksum, |
4797 | B_FALSE) != 0 || | |
4798 | ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_COPIES, 1, | |
4799 | B_FALSE) != 0) { | |
1e33ac1e | 4800 | (void) rw_exit(&zs->zs_name_lock); |
40b84e7a | 4801 | umem_free(od, sizeof(ztest_od_t)); |
428870ff BB |
4802 | return; |
4803 | } | |
4804 | ||
4805 | object = od[0].od_object; | |
4806 | blocksize = od[0].od_blocksize; | |
4807 | pattern = spa_guid(spa) ^ dmu_objset_fsid_guid(os); | |
4808 | ||
4809 | ASSERT(object != 0); | |
4810 | ||
4811 | tx = dmu_tx_create(os); | |
4812 | dmu_tx_hold_write(tx, object, 0, copies * blocksize); | |
4813 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
4814 | if (txg == 0) { | |
1e33ac1e | 4815 | (void) rw_exit(&zs->zs_name_lock); |
40b84e7a | 4816 | umem_free(od, sizeof(ztest_od_t)); |
428870ff BB |
4817 | return; |
4818 | } | |
34dc7c2f BB |
4819 | |
4820 | /* | |
428870ff | 4821 | * Write all the copies of our block. |
34dc7c2f | 4822 | */ |
d6320ddb | 4823 | for (i = 0; i < copies; i++) { |
428870ff BB |
4824 | uint64_t offset = i * blocksize; |
4825 | VERIFY(dmu_buf_hold(os, object, offset, FTAG, &db, | |
4826 | DMU_READ_NO_PREFETCH) == 0); | |
4827 | ASSERT(db->db_offset == offset); | |
4828 | ASSERT(db->db_size == blocksize); | |
4829 | ASSERT(ztest_pattern_match(db->db_data, db->db_size, pattern) || | |
4830 | ztest_pattern_match(db->db_data, db->db_size, 0ULL)); | |
4831 | dmu_buf_will_fill(db, tx); | |
4832 | ztest_pattern_set(db->db_data, db->db_size, pattern); | |
4833 | dmu_buf_rele(db, FTAG); | |
4834 | } | |
34dc7c2f | 4835 | |
428870ff BB |
4836 | dmu_tx_commit(tx); |
4837 | txg_wait_synced(spa_get_dsl(spa), txg); | |
34dc7c2f BB |
4838 | |
4839 | /* | |
428870ff | 4840 | * Find out what block we got. |
34dc7c2f | 4841 | */ |
428870ff BB |
4842 | VERIFY(dmu_buf_hold(os, object, 0, FTAG, &db, |
4843 | DMU_READ_NO_PREFETCH) == 0); | |
4844 | blk = *((dmu_buf_impl_t *)db)->db_blkptr; | |
4845 | dmu_buf_rele(db, FTAG); | |
34dc7c2f BB |
4846 | |
4847 | /* | |
428870ff | 4848 | * Damage the block. Dedup-ditto will save us when we read it later. |
34dc7c2f | 4849 | */ |
428870ff BB |
4850 | psize = BP_GET_PSIZE(&blk); |
4851 | buf = zio_buf_alloc(psize); | |
4852 | ztest_pattern_set(buf, psize, ~pattern); | |
34dc7c2f | 4853 | |
428870ff BB |
4854 | (void) zio_wait(zio_rewrite(NULL, spa, 0, &blk, |
4855 | buf, psize, NULL, NULL, ZIO_PRIORITY_SYNC_WRITE, | |
4856 | ZIO_FLAG_CANFAIL | ZIO_FLAG_INDUCE_DAMAGE, NULL)); | |
34dc7c2f | 4857 | |
428870ff | 4858 | zio_buf_free(buf, psize); |
34dc7c2f | 4859 | |
1e33ac1e | 4860 | (void) rw_exit(&zs->zs_name_lock); |
40b84e7a | 4861 | umem_free(od, sizeof(ztest_od_t)); |
34dc7c2f BB |
4862 | } |
4863 | ||
34dc7c2f | 4864 | /* |
428870ff | 4865 | * Scrub the pool. |
34dc7c2f | 4866 | */ |
428870ff BB |
4867 | /* ARGSUSED */ |
4868 | void | |
4869 | ztest_scrub(ztest_ds_t *zd, uint64_t id) | |
34dc7c2f | 4870 | { |
428870ff BB |
4871 | ztest_shared_t *zs = ztest_shared; |
4872 | spa_t *spa = zs->zs_spa; | |
34dc7c2f | 4873 | |
428870ff BB |
4874 | (void) spa_scan(spa, POOL_SCAN_SCRUB); |
4875 | (void) poll(NULL, 0, 100); /* wait a moment, then force a restart */ | |
4876 | (void) spa_scan(spa, POOL_SCAN_SCRUB); | |
4877 | } | |
34dc7c2f | 4878 | |
428870ff BB |
4879 | /* |
4880 | * Rename the pool to a different name and then rename it back. | |
4881 | */ | |
4882 | /* ARGSUSED */ | |
4883 | void | |
4884 | ztest_spa_rename(ztest_ds_t *zd, uint64_t id) | |
4885 | { | |
4886 | ztest_shared_t *zs = ztest_shared; | |
4887 | char *oldname, *newname; | |
4888 | spa_t *spa; | |
34dc7c2f | 4889 | |
1e33ac1e | 4890 | (void) rw_enter(&zs->zs_name_lock, RW_WRITER); |
34dc7c2f | 4891 | |
428870ff BB |
4892 | oldname = zs->zs_pool; |
4893 | newname = umem_alloc(strlen(oldname) + 5, UMEM_NOFAIL); | |
4894 | (void) strcpy(newname, oldname); | |
4895 | (void) strcat(newname, "_tmp"); | |
34dc7c2f BB |
4896 | |
4897 | /* | |
428870ff | 4898 | * Do the rename |
34dc7c2f | 4899 | */ |
428870ff | 4900 | VERIFY3U(0, ==, spa_rename(oldname, newname)); |
34dc7c2f BB |
4901 | |
4902 | /* | |
428870ff | 4903 | * Try to open it under the old name, which shouldn't exist |
34dc7c2f | 4904 | */ |
428870ff | 4905 | VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f BB |
4906 | |
4907 | /* | |
428870ff BB |
4908 | * Open it under the new name and make sure it's still the same spa_t. |
4909 | */ | |
4910 | VERIFY3U(0, ==, spa_open(newname, &spa, FTAG)); | |
34dc7c2f | 4911 | |
428870ff BB |
4912 | ASSERT(spa == zs->zs_spa); |
4913 | spa_close(spa, FTAG); | |
34dc7c2f BB |
4914 | |
4915 | /* | |
428870ff | 4916 | * Rename it back to the original |
34dc7c2f | 4917 | */ |
428870ff | 4918 | VERIFY3U(0, ==, spa_rename(newname, oldname)); |
34dc7c2f | 4919 | |
428870ff BB |
4920 | /* |
4921 | * Make sure it can still be opened | |
4922 | */ | |
4923 | VERIFY3U(0, ==, spa_open(oldname, &spa, FTAG)); | |
34dc7c2f | 4924 | |
428870ff BB |
4925 | ASSERT(spa == zs->zs_spa); |
4926 | spa_close(spa, FTAG); | |
4927 | ||
4928 | umem_free(newname, strlen(newname) + 1); | |
4929 | ||
1e33ac1e | 4930 | (void) rw_exit(&zs->zs_name_lock); |
34dc7c2f BB |
4931 | } |
4932 | ||
428870ff BB |
4933 | /* |
4934 | * Verify pool integrity by running zdb. | |
4935 | */ | |
34dc7c2f | 4936 | static void |
428870ff | 4937 | ztest_run_zdb(char *pool) |
34dc7c2f BB |
4938 | { |
4939 | int status; | |
34dc7c2f | 4940 | char *bin; |
0e8d1b2d BB |
4941 | char *zdb; |
4942 | char *zbuf; | |
34dc7c2f BB |
4943 | FILE *fp; |
4944 | ||
0e8d1b2d BB |
4945 | bin = umem_alloc(MAXPATHLEN + MAXNAMELEN + 20, UMEM_NOFAIL); |
4946 | zdb = umem_alloc(MAXPATHLEN + MAXNAMELEN + 20, UMEM_NOFAIL); | |
4947 | zbuf = umem_alloc(1024, UMEM_NOFAIL); | |
34dc7c2f | 4948 | |
0e8d1b2d | 4949 | VERIFY(realpath(getexecname(), bin) != NULL); |
341b5f1d | 4950 | if (strncmp(bin, "/usr/sbin/ztest", 15) == 0) { |
0e8d1b2d | 4951 | strcpy(bin, "/usr/sbin/zdb"); /* Installed */ |
341b5f1d BB |
4952 | } else if (strncmp(bin, "/sbin/ztest", 11) == 0) { |
4953 | strcpy(bin, "/sbin/zdb"); /* Installed */ | |
0e8d1b2d BB |
4954 | } else { |
4955 | strstr(bin, "/ztest/")[0] = '\0'; /* In-tree */ | |
4956 | strcat(bin, "/zdb/zdb"); | |
4957 | } | |
4958 | ||
4959 | (void) sprintf(zdb, | |
4960 | "%s -bcc%s%s -U %s %s", | |
4961 | bin, | |
34dc7c2f BB |
4962 | zopt_verbose >= 3 ? "s" : "", |
4963 | zopt_verbose >= 4 ? "v" : "", | |
428870ff | 4964 | spa_config_path, |
b128c09f | 4965 | pool); |
34dc7c2f BB |
4966 | |
4967 | if (zopt_verbose >= 5) | |
4968 | (void) printf("Executing %s\n", strstr(zdb, "zdb ")); | |
4969 | ||
4970 | fp = popen(zdb, "r"); | |
4971 | ||
4972 | while (fgets(zbuf, sizeof (zbuf), fp) != NULL) | |
4973 | if (zopt_verbose >= 3) | |
4974 | (void) printf("%s", zbuf); | |
4975 | ||
4976 | status = pclose(fp); | |
4977 | ||
4978 | if (status == 0) | |
0e8d1b2d | 4979 | goto out; |
34dc7c2f BB |
4980 | |
4981 | ztest_dump_core = 0; | |
4982 | if (WIFEXITED(status)) | |
4983 | fatal(0, "'%s' exit code %d", zdb, WEXITSTATUS(status)); | |
4984 | else | |
4985 | fatal(0, "'%s' died with signal %d", zdb, WTERMSIG(status)); | |
0e8d1b2d BB |
4986 | out: |
4987 | umem_free(bin, MAXPATHLEN + MAXNAMELEN + 20); | |
4988 | umem_free(zdb, MAXPATHLEN + MAXNAMELEN + 20); | |
4989 | umem_free(zbuf, 1024); | |
34dc7c2f BB |
4990 | } |
4991 | ||
4992 | static void | |
4993 | ztest_walk_pool_directory(char *header) | |
4994 | { | |
4995 | spa_t *spa = NULL; | |
4996 | ||
4997 | if (zopt_verbose >= 6) | |
4998 | (void) printf("%s\n", header); | |
4999 | ||
5000 | mutex_enter(&spa_namespace_lock); | |
5001 | while ((spa = spa_next(spa)) != NULL) | |
5002 | if (zopt_verbose >= 6) | |
5003 | (void) printf("\t%s\n", spa_name(spa)); | |
5004 | mutex_exit(&spa_namespace_lock); | |
5005 | } | |
5006 | ||
5007 | static void | |
5008 | ztest_spa_import_export(char *oldname, char *newname) | |
5009 | { | |
fb5f0bc8 | 5010 | nvlist_t *config, *newconfig; |
34dc7c2f BB |
5011 | uint64_t pool_guid; |
5012 | spa_t *spa; | |
34dc7c2f BB |
5013 | |
5014 | if (zopt_verbose >= 4) { | |
5015 | (void) printf("import/export: old = %s, new = %s\n", | |
5016 | oldname, newname); | |
5017 | } | |
5018 | ||
5019 | /* | |
5020 | * Clean up from previous runs. | |
5021 | */ | |
5022 | (void) spa_destroy(newname); | |
5023 | ||
5024 | /* | |
5025 | * Get the pool's configuration and guid. | |
5026 | */ | |
428870ff | 5027 | VERIFY3U(0, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f | 5028 | |
fb5f0bc8 BB |
5029 | /* |
5030 | * Kick off a scrub to tickle scrub/export races. | |
5031 | */ | |
5032 | if (ztest_random(2) == 0) | |
428870ff | 5033 | (void) spa_scan(spa, POOL_SCAN_SCRUB); |
fb5f0bc8 | 5034 | |
34dc7c2f BB |
5035 | pool_guid = spa_guid(spa); |
5036 | spa_close(spa, FTAG); | |
5037 | ||
5038 | ztest_walk_pool_directory("pools before export"); | |
5039 | ||
5040 | /* | |
5041 | * Export it. | |
5042 | */ | |
428870ff | 5043 | VERIFY3U(0, ==, spa_export(oldname, &config, B_FALSE, B_FALSE)); |
34dc7c2f BB |
5044 | |
5045 | ztest_walk_pool_directory("pools after export"); | |
5046 | ||
fb5f0bc8 BB |
5047 | /* |
5048 | * Try to import it. | |
5049 | */ | |
5050 | newconfig = spa_tryimport(config); | |
5051 | ASSERT(newconfig != NULL); | |
5052 | nvlist_free(newconfig); | |
5053 | ||
34dc7c2f BB |
5054 | /* |
5055 | * Import it under the new name. | |
5056 | */ | |
572e2857 | 5057 | VERIFY3U(0, ==, spa_import(newname, config, NULL, 0)); |
34dc7c2f BB |
5058 | |
5059 | ztest_walk_pool_directory("pools after import"); | |
5060 | ||
5061 | /* | |
5062 | * Try to import it again -- should fail with EEXIST. | |
5063 | */ | |
572e2857 | 5064 | VERIFY3U(EEXIST, ==, spa_import(newname, config, NULL, 0)); |
34dc7c2f BB |
5065 | |
5066 | /* | |
5067 | * Try to import it under a different name -- should fail with EEXIST. | |
5068 | */ | |
572e2857 | 5069 | VERIFY3U(EEXIST, ==, spa_import(oldname, config, NULL, 0)); |
34dc7c2f BB |
5070 | |
5071 | /* | |
5072 | * Verify that the pool is no longer visible under the old name. | |
5073 | */ | |
428870ff | 5074 | VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f BB |
5075 | |
5076 | /* | |
5077 | * Verify that we can open and close the pool using the new name. | |
5078 | */ | |
428870ff | 5079 | VERIFY3U(0, ==, spa_open(newname, &spa, FTAG)); |
34dc7c2f BB |
5080 | ASSERT(pool_guid == spa_guid(spa)); |
5081 | spa_close(spa, FTAG); | |
5082 | ||
5083 | nvlist_free(config); | |
5084 | } | |
5085 | ||
fb5f0bc8 BB |
5086 | static void |
5087 | ztest_resume(spa_t *spa) | |
5088 | { | |
428870ff BB |
5089 | if (spa_suspended(spa) && zopt_verbose >= 6) |
5090 | (void) printf("resuming from suspended state\n"); | |
5091 | spa_vdev_state_enter(spa, SCL_NONE); | |
5092 | vdev_clear(spa, NULL); | |
5093 | (void) spa_vdev_state_exit(spa, NULL, 0); | |
5094 | (void) zio_resume(spa); | |
fb5f0bc8 BB |
5095 | } |
5096 | ||
34dc7c2f | 5097 | static void * |
fb5f0bc8 | 5098 | ztest_resume_thread(void *arg) |
34dc7c2f | 5099 | { |
b128c09f | 5100 | spa_t *spa = arg; |
34dc7c2f BB |
5101 | |
5102 | while (!ztest_exiting) { | |
428870ff BB |
5103 | if (spa_suspended(spa)) |
5104 | ztest_resume(spa); | |
5105 | (void) poll(NULL, 0, 100); | |
34dc7c2f | 5106 | } |
34dc7c2f | 5107 | |
1e33ac1e | 5108 | thread_exit(); |
428870ff | 5109 | |
1e33ac1e BB |
5110 | return (NULL); |
5111 | } | |
428870ff | 5112 | |
1e33ac1e | 5113 | #define GRACE 300 |
428870ff | 5114 | |
1e33ac1e BB |
5115 | static void |
5116 | ztest_deadman_alarm(int sig) | |
5117 | { | |
5118 | fatal(0, "failed to complete within %d seconds of deadline", GRACE); | |
428870ff BB |
5119 | } |
5120 | ||
5121 | static void | |
5122 | ztest_execute(ztest_info_t *zi, uint64_t id) | |
5123 | { | |
5124 | ztest_shared_t *zs = ztest_shared; | |
5125 | ztest_ds_t *zd = &zs->zs_zd[id % zopt_datasets]; | |
5126 | hrtime_t functime = gethrtime(); | |
d6320ddb | 5127 | int i; |
428870ff | 5128 | |
d6320ddb | 5129 | for (i = 0; i < zi->zi_iters; i++) |
428870ff BB |
5130 | zi->zi_func(zd, id); |
5131 | ||
5132 | functime = gethrtime() - functime; | |
5133 | ||
5134 | atomic_add_64(&zi->zi_call_count, 1); | |
5135 | atomic_add_64(&zi->zi_call_time, functime); | |
5136 | ||
5137 | if (zopt_verbose >= 4) { | |
5138 | Dl_info dli; | |
5139 | (void) dladdr((void *)zi->zi_func, &dli); | |
5140 | (void) printf("%6.2f sec in %s\n", | |
5141 | (double)functime / NANOSEC, dli.dli_sname); | |
5142 | } | |
5143 | } | |
5144 | ||
34dc7c2f BB |
5145 | static void * |
5146 | ztest_thread(void *arg) | |
5147 | { | |
428870ff | 5148 | uint64_t id = (uintptr_t)arg; |
34dc7c2f | 5149 | ztest_shared_t *zs = ztest_shared; |
428870ff BB |
5150 | uint64_t call_next; |
5151 | hrtime_t now; | |
34dc7c2f | 5152 | ztest_info_t *zi; |
34dc7c2f | 5153 | |
428870ff | 5154 | while ((now = gethrtime()) < zs->zs_thread_stop) { |
34dc7c2f BB |
5155 | /* |
5156 | * See if it's time to force a crash. | |
5157 | */ | |
428870ff BB |
5158 | if (now > zs->zs_thread_kill) |
5159 | ztest_kill(zs); | |
34dc7c2f BB |
5160 | |
5161 | /* | |
428870ff | 5162 | * If we're getting ENOSPC with some regularity, stop. |
34dc7c2f | 5163 | */ |
428870ff BB |
5164 | if (zs->zs_enospc_count > 10) |
5165 | break; | |
34dc7c2f BB |
5166 | |
5167 | /* | |
428870ff | 5168 | * Pick a random function to execute. |
34dc7c2f | 5169 | */ |
428870ff BB |
5170 | zi = &zs->zs_info[ztest_random(ZTEST_FUNCS)]; |
5171 | call_next = zi->zi_call_next; | |
34dc7c2f | 5172 | |
428870ff BB |
5173 | if (now >= call_next && |
5174 | atomic_cas_64(&zi->zi_call_next, call_next, call_next + | |
5175 | ztest_random(2 * zi->zi_interval[0] + 1)) == call_next) | |
5176 | ztest_execute(zi, id); | |
5177 | } | |
34dc7c2f | 5178 | |
1e33ac1e BB |
5179 | thread_exit(); |
5180 | ||
428870ff BB |
5181 | return (NULL); |
5182 | } | |
34dc7c2f | 5183 | |
428870ff BB |
5184 | static void |
5185 | ztest_dataset_name(char *dsname, char *pool, int d) | |
5186 | { | |
5187 | (void) snprintf(dsname, MAXNAMELEN, "%s/ds_%d", pool, d); | |
5188 | } | |
34dc7c2f | 5189 | |
428870ff BB |
5190 | static void |
5191 | ztest_dataset_destroy(ztest_shared_t *zs, int d) | |
5192 | { | |
5193 | char name[MAXNAMELEN]; | |
d6320ddb | 5194 | int t; |
34dc7c2f | 5195 | |
428870ff | 5196 | ztest_dataset_name(name, zs->zs_pool, d); |
34dc7c2f | 5197 | |
428870ff BB |
5198 | if (zopt_verbose >= 3) |
5199 | (void) printf("Destroying %s to free up space\n", name); | |
34dc7c2f | 5200 | |
428870ff BB |
5201 | /* |
5202 | * Cleanup any non-standard clones and snapshots. In general, | |
5203 | * ztest thread t operates on dataset (t % zopt_datasets), | |
5204 | * so there may be more than one thing to clean up. | |
5205 | */ | |
d6320ddb | 5206 | for (t = d; t < zopt_threads; t += zopt_datasets) |
428870ff BB |
5207 | ztest_dsl_dataset_cleanup(name, t); |
5208 | ||
5209 | (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL, | |
5210 | DS_FIND_SNAPSHOTS | DS_FIND_CHILDREN); | |
5211 | } | |
5212 | ||
5213 | static void | |
5214 | ztest_dataset_dirobj_verify(ztest_ds_t *zd) | |
5215 | { | |
5216 | uint64_t usedobjs, dirobjs, scratch; | |
5217 | ||
5218 | /* | |
5219 | * ZTEST_DIROBJ is the object directory for the entire dataset. | |
5220 | * Therefore, the number of objects in use should equal the | |
5221 | * number of ZTEST_DIROBJ entries, +1 for ZTEST_DIROBJ itself. | |
5222 | * If not, we have an object leak. | |
5223 | * | |
5224 | * Note that we can only check this in ztest_dataset_open(), | |
5225 | * when the open-context and syncing-context values agree. | |
5226 | * That's because zap_count() returns the open-context value, | |
5227 | * while dmu_objset_space() returns the rootbp fill count. | |
5228 | */ | |
5229 | VERIFY3U(0, ==, zap_count(zd->zd_os, ZTEST_DIROBJ, &dirobjs)); | |
5230 | dmu_objset_space(zd->zd_os, &scratch, &scratch, &usedobjs, &scratch); | |
5231 | ASSERT3U(dirobjs + 1, ==, usedobjs); | |
5232 | } | |
5233 | ||
5234 | static int | |
5235 | ztest_dataset_open(ztest_shared_t *zs, int d) | |
5236 | { | |
5237 | ztest_ds_t *zd = &zs->zs_zd[d]; | |
5238 | uint64_t committed_seq = zd->zd_seq; | |
5239 | objset_t *os; | |
5240 | zilog_t *zilog; | |
5241 | char name[MAXNAMELEN]; | |
5242 | int error; | |
5243 | ||
5244 | ztest_dataset_name(name, zs->zs_pool, d); | |
5245 | ||
1e33ac1e | 5246 | (void) rw_enter(&zs->zs_name_lock, RW_READER); |
428870ff BB |
5247 | |
5248 | error = ztest_dataset_create(name); | |
5249 | if (error == ENOSPC) { | |
1e33ac1e | 5250 | (void) rw_exit(&zs->zs_name_lock); |
428870ff BB |
5251 | ztest_record_enospc(FTAG); |
5252 | return (error); | |
34dc7c2f | 5253 | } |
428870ff | 5254 | ASSERT(error == 0 || error == EEXIST); |
34dc7c2f | 5255 | |
428870ff | 5256 | VERIFY3U(dmu_objset_hold(name, zd, &os), ==, 0); |
1e33ac1e | 5257 | (void) rw_exit(&zs->zs_name_lock); |
428870ff BB |
5258 | |
5259 | ztest_zd_init(zd, os); | |
5260 | ||
5261 | zilog = zd->zd_zilog; | |
5262 | ||
5263 | if (zilog->zl_header->zh_claim_lr_seq != 0 && | |
5264 | zilog->zl_header->zh_claim_lr_seq < committed_seq) | |
5265 | fatal(0, "missing log records: claimed %llu < committed %llu", | |
5266 | zilog->zl_header->zh_claim_lr_seq, committed_seq); | |
5267 | ||
5268 | ztest_dataset_dirobj_verify(zd); | |
5269 | ||
5270 | zil_replay(os, zd, ztest_replay_vector); | |
5271 | ||
5272 | ztest_dataset_dirobj_verify(zd); | |
5273 | ||
5274 | if (zopt_verbose >= 6) | |
5275 | (void) printf("%s replay %llu blocks, %llu records, seq %llu\n", | |
5276 | zd->zd_name, | |
5277 | (u_longlong_t)zilog->zl_parse_blk_count, | |
5278 | (u_longlong_t)zilog->zl_parse_lr_count, | |
5279 | (u_longlong_t)zilog->zl_replaying_seq); | |
5280 | ||
5281 | zilog = zil_open(os, ztest_get_data); | |
5282 | ||
5283 | if (zilog->zl_replaying_seq != 0 && | |
5284 | zilog->zl_replaying_seq < committed_seq) | |
5285 | fatal(0, "missing log records: replayed %llu < committed %llu", | |
5286 | zilog->zl_replaying_seq, committed_seq); | |
5287 | ||
5288 | return (0); | |
5289 | } | |
5290 | ||
5291 | static void | |
5292 | ztest_dataset_close(ztest_shared_t *zs, int d) | |
5293 | { | |
5294 | ztest_ds_t *zd = &zs->zs_zd[d]; | |
5295 | ||
5296 | zil_close(zd->zd_zilog); | |
5297 | dmu_objset_rele(zd->zd_os, zd); | |
5298 | ||
5299 | ztest_zd_fini(zd); | |
34dc7c2f BB |
5300 | } |
5301 | ||
5302 | /* | |
5303 | * Kick off threads to run tests on all datasets in parallel. | |
5304 | */ | |
5305 | static void | |
428870ff | 5306 | ztest_run(ztest_shared_t *zs) |
34dc7c2f | 5307 | { |
1e33ac1e | 5308 | kt_did_t *tid; |
34dc7c2f | 5309 | spa_t *spa; |
1e33ac1e BB |
5310 | kthread_t *resume_thread; |
5311 | uint64_t object; | |
428870ff | 5312 | int error; |
d6320ddb | 5313 | int t, d; |
b128c09f BB |
5314 | |
5315 | ztest_exiting = B_FALSE; | |
34dc7c2f | 5316 | |
34dc7c2f | 5317 | /* |
428870ff | 5318 | * Initialize parent/child shared state. |
34dc7c2f | 5319 | */ |
1e33ac1e BB |
5320 | mutex_init(&zs->zs_vdev_lock, NULL, MUTEX_DEFAULT, NULL); |
5321 | rw_init(&zs->zs_name_lock, NULL, RW_DEFAULT, NULL); | |
34dc7c2f | 5322 | |
428870ff BB |
5323 | zs->zs_thread_start = gethrtime(); |
5324 | zs->zs_thread_stop = zs->zs_thread_start + zopt_passtime * NANOSEC; | |
5325 | zs->zs_thread_stop = MIN(zs->zs_thread_stop, zs->zs_proc_stop); | |
5326 | zs->zs_thread_kill = zs->zs_thread_stop; | |
5327 | if (ztest_random(100) < zopt_killrate) | |
5328 | zs->zs_thread_kill -= ztest_random(zopt_passtime * NANOSEC); | |
34dc7c2f | 5329 | |
1e33ac1e | 5330 | mutex_init(&zcl.zcl_callbacks_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 5331 | |
428870ff BB |
5332 | list_create(&zcl.zcl_callbacks, sizeof (ztest_cb_data_t), |
5333 | offsetof(ztest_cb_data_t, zcd_node)); | |
34dc7c2f BB |
5334 | |
5335 | /* | |
b128c09f | 5336 | * Open our pool. |
34dc7c2f | 5337 | */ |
428870ff BB |
5338 | kernel_init(FREAD | FWRITE); |
5339 | VERIFY(spa_open(zs->zs_pool, &spa, FTAG) == 0); | |
6d974228 | 5340 | spa->spa_debug = B_TRUE; |
428870ff BB |
5341 | zs->zs_spa = spa; |
5342 | ||
5343 | spa->spa_dedup_ditto = 2 * ZIO_DEDUPDITTO_MIN; | |
34dc7c2f | 5344 | |
fb5f0bc8 BB |
5345 | /* |
5346 | * We don't expect the pool to suspend unless maxfaults == 0, | |
5347 | * in which case ztest_fault_inject() temporarily takes away | |
5348 | * the only valid replica. | |
5349 | */ | |
428870ff | 5350 | if (MAXFAULTS() == 0) |
fb5f0bc8 BB |
5351 | spa->spa_failmode = ZIO_FAILURE_MODE_WAIT; |
5352 | else | |
5353 | spa->spa_failmode = ZIO_FAILURE_MODE_PANIC; | |
5354 | ||
34dc7c2f | 5355 | /* |
b128c09f | 5356 | * Create a thread to periodically resume suspended I/O. |
34dc7c2f | 5357 | */ |
1e33ac1e BB |
5358 | VERIFY3P((resume_thread = thread_create(NULL, 0, ztest_resume_thread, |
5359 | spa, TS_RUN, NULL, 0, 0)), !=, NULL); | |
34dc7c2f | 5360 | |
428870ff | 5361 | /* |
1e33ac1e | 5362 | * Set a deadman alarm to abort() if we hang. |
428870ff | 5363 | */ |
1e33ac1e BB |
5364 | signal(SIGALRM, ztest_deadman_alarm); |
5365 | alarm((zs->zs_thread_stop - zs->zs_thread_start) / NANOSEC + GRACE); | |
428870ff | 5366 | |
34dc7c2f BB |
5367 | /* |
5368 | * Verify that we can safely inquire about about any object, | |
5369 | * whether it's allocated or not. To make it interesting, | |
5370 | * we probe a 5-wide window around each power of two. | |
5371 | * This hits all edge cases, including zero and the max. | |
5372 | */ | |
d6320ddb BB |
5373 | for (t = 0; t < 64; t++) { |
5374 | for (d = -5; d <= 5; d++) { | |
34dc7c2f BB |
5375 | error = dmu_object_info(spa->spa_meta_objset, |
5376 | (1ULL << t) + d, NULL); | |
5377 | ASSERT(error == 0 || error == ENOENT || | |
5378 | error == EINVAL); | |
5379 | } | |
5380 | } | |
5381 | ||
5382 | /* | |
428870ff | 5383 | * If we got any ENOSPC errors on the previous run, destroy something. |
34dc7c2f | 5384 | */ |
428870ff BB |
5385 | if (zs->zs_enospc_count != 0) { |
5386 | int d = ztest_random(zopt_datasets); | |
5387 | ztest_dataset_destroy(zs, d); | |
5388 | } | |
34dc7c2f BB |
5389 | zs->zs_enospc_count = 0; |
5390 | ||
1e33ac1e | 5391 | tid = umem_zalloc(zopt_threads * sizeof (kt_did_t), UMEM_NOFAIL); |
34dc7c2f BB |
5392 | |
5393 | if (zopt_verbose >= 4) | |
5394 | (void) printf("starting main threads...\n"); | |
5395 | ||
428870ff BB |
5396 | /* |
5397 | * Kick off all the tests that run in parallel. | |
5398 | */ | |
d6320ddb | 5399 | for (t = 0; t < zopt_threads; t++) { |
1e33ac1e BB |
5400 | kthread_t *thread; |
5401 | ||
428870ff BB |
5402 | if (t < zopt_datasets && ztest_dataset_open(zs, t) != 0) |
5403 | return; | |
1e33ac1e BB |
5404 | |
5405 | VERIFY3P(thread = thread_create(NULL, 0, ztest_thread, | |
5406 | (void *)(uintptr_t)t, TS_RUN, NULL, 0, 0), !=, NULL); | |
5407 | tid[t] = thread->t_tid; | |
34dc7c2f BB |
5408 | } |
5409 | ||
428870ff BB |
5410 | /* |
5411 | * Wait for all of the tests to complete. We go in reverse order | |
5412 | * so we don't close datasets while threads are still using them. | |
5413 | */ | |
d6320ddb | 5414 | for (t = zopt_threads - 1; t >= 0; t--) { |
1e33ac1e | 5415 | thread_join(tid[t]); |
428870ff BB |
5416 | if (t < zopt_datasets) |
5417 | ztest_dataset_close(zs, t); | |
34dc7c2f BB |
5418 | } |
5419 | ||
34dc7c2f BB |
5420 | txg_wait_synced(spa_get_dsl(spa), 0); |
5421 | ||
428870ff BB |
5422 | zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(spa)); |
5423 | zs->zs_space = metaslab_class_get_space(spa_normal_class(spa)); | |
5424 | ||
1e33ac1e | 5425 | umem_free(tid, zopt_threads * sizeof (kt_did_t)); |
428870ff BB |
5426 | |
5427 | /* Kill the resume thread */ | |
5428 | ztest_exiting = B_TRUE; | |
1e33ac1e | 5429 | thread_join(resume_thread->t_tid); |
428870ff | 5430 | ztest_resume(spa); |
34dc7c2f BB |
5431 | |
5432 | /* | |
428870ff BB |
5433 | * Right before closing the pool, kick off a bunch of async I/O; |
5434 | * spa_close() should wait for it to complete. | |
34dc7c2f | 5435 | */ |
1e33ac1e | 5436 | for (object = 1; object < 50; object++) |
428870ff BB |
5437 | dmu_prefetch(spa->spa_meta_objset, object, 0, 1ULL << 20); |
5438 | ||
090ff092 RC |
5439 | /* Verify that at least one commit cb was called in a timely fashion */ |
5440 | if (zc_cb_counter >= ZTEST_COMMIT_CB_MIN_REG) | |
5441 | VERIFY3U(zc_min_txg_delay, ==, 0); | |
5442 | ||
428870ff BB |
5443 | spa_close(spa, FTAG); |
5444 | ||
5445 | /* | |
5446 | * Verify that we can loop over all pools. | |
5447 | */ | |
5448 | mutex_enter(&spa_namespace_lock); | |
5449 | for (spa = spa_next(NULL); spa != NULL; spa = spa_next(spa)) | |
5450 | if (zopt_verbose > 3) | |
5451 | (void) printf("spa_next: found %s\n", spa_name(spa)); | |
5452 | mutex_exit(&spa_namespace_lock); | |
5453 | ||
5454 | /* | |
5455 | * Verify that we can export the pool and reimport it under a | |
5456 | * different name. | |
5457 | */ | |
5458 | if (ztest_random(2) == 0) { | |
5459 | char name[MAXNAMELEN]; | |
5460 | (void) snprintf(name, MAXNAMELEN, "%s_import", zs->zs_pool); | |
5461 | ztest_spa_import_export(zs->zs_pool, name); | |
5462 | ztest_spa_import_export(name, zs->zs_pool); | |
5463 | } | |
5464 | ||
5465 | kernel_fini(); | |
572e2857 BB |
5466 | |
5467 | list_destroy(&zcl.zcl_callbacks); | |
0e8d1b2d BB |
5468 | mutex_destroy(&zcl.zcl_callbacks_lock); |
5469 | rw_destroy(&zs->zs_name_lock); | |
5470 | mutex_destroy(&zs->zs_vdev_lock); | |
428870ff BB |
5471 | } |
5472 | ||
5473 | static void | |
5474 | ztest_freeze(ztest_shared_t *zs) | |
5475 | { | |
5476 | ztest_ds_t *zd = &zs->zs_zd[0]; | |
5477 | spa_t *spa; | |
5478 | int numloops = 0; | |
5479 | ||
5480 | if (zopt_verbose >= 3) | |
5481 | (void) printf("testing spa_freeze()...\n"); | |
9babb374 | 5482 | |
428870ff BB |
5483 | kernel_init(FREAD | FWRITE); |
5484 | VERIFY3U(0, ==, spa_open(zs->zs_pool, &spa, FTAG)); | |
5485 | VERIFY3U(0, ==, ztest_dataset_open(zs, 0)); | |
9babb374 | 5486 | |
428870ff BB |
5487 | /* |
5488 | * Force the first log block to be transactionally allocated. | |
5489 | * We have to do this before we freeze the pool -- otherwise | |
5490 | * the log chain won't be anchored. | |
5491 | */ | |
5492 | while (BP_IS_HOLE(&zd->zd_zilog->zl_header->zh_log)) { | |
5493 | ztest_dmu_object_alloc_free(zd, 0); | |
572e2857 | 5494 | zil_commit(zd->zd_zilog, 0); |
34dc7c2f BB |
5495 | } |
5496 | ||
5497 | txg_wait_synced(spa_get_dsl(spa), 0); | |
5498 | ||
428870ff BB |
5499 | /* |
5500 | * Freeze the pool. This stops spa_sync() from doing anything, | |
5501 | * so that the only way to record changes from now on is the ZIL. | |
5502 | */ | |
5503 | spa_freeze(spa); | |
b128c09f | 5504 | |
428870ff BB |
5505 | /* |
5506 | * Run tests that generate log records but don't alter the pool config | |
5507 | * or depend on DSL sync tasks (snapshots, objset create/destroy, etc). | |
5508 | * We do a txg_wait_synced() after each iteration to force the txg | |
5509 | * to increase well beyond the last synced value in the uberblock. | |
5510 | * The ZIL should be OK with that. | |
5511 | */ | |
5512 | while (ztest_random(10) != 0 && numloops++ < zopt_maxloops) { | |
5513 | ztest_dmu_write_parallel(zd, 0); | |
5514 | ztest_dmu_object_alloc_free(zd, 0); | |
5515 | txg_wait_synced(spa_get_dsl(spa), 0); | |
5516 | } | |
b128c09f | 5517 | |
34dc7c2f | 5518 | /* |
428870ff | 5519 | * Commit all of the changes we just generated. |
34dc7c2f | 5520 | */ |
572e2857 | 5521 | zil_commit(zd->zd_zilog, 0); |
428870ff | 5522 | txg_wait_synced(spa_get_dsl(spa), 0); |
34dc7c2f | 5523 | |
428870ff BB |
5524 | /* |
5525 | * Close our dataset and close the pool. | |
5526 | */ | |
5527 | ztest_dataset_close(zs, 0); | |
34dc7c2f | 5528 | spa_close(spa, FTAG); |
428870ff | 5529 | kernel_fini(); |
34dc7c2f | 5530 | |
428870ff BB |
5531 | /* |
5532 | * Open and close the pool and dataset to induce log replay. | |
5533 | */ | |
5534 | kernel_init(FREAD | FWRITE); | |
5535 | VERIFY3U(0, ==, spa_open(zs->zs_pool, &spa, FTAG)); | |
5536 | VERIFY3U(0, ==, ztest_dataset_open(zs, 0)); | |
5537 | ztest_dataset_close(zs, 0); | |
5538 | spa_close(spa, FTAG); | |
34dc7c2f BB |
5539 | kernel_fini(); |
5540 | } | |
5541 | ||
5542 | void | |
5543 | print_time(hrtime_t t, char *timebuf) | |
5544 | { | |
5545 | hrtime_t s = t / NANOSEC; | |
5546 | hrtime_t m = s / 60; | |
5547 | hrtime_t h = m / 60; | |
5548 | hrtime_t d = h / 24; | |
5549 | ||
5550 | s -= m * 60; | |
5551 | m -= h * 60; | |
5552 | h -= d * 24; | |
5553 | ||
5554 | timebuf[0] = '\0'; | |
5555 | ||
5556 | if (d) | |
5557 | (void) sprintf(timebuf, | |
5558 | "%llud%02lluh%02llum%02llus", d, h, m, s); | |
5559 | else if (h) | |
5560 | (void) sprintf(timebuf, "%lluh%02llum%02llus", h, m, s); | |
5561 | else if (m) | |
5562 | (void) sprintf(timebuf, "%llum%02llus", m, s); | |
5563 | else | |
5564 | (void) sprintf(timebuf, "%llus", s); | |
5565 | } | |
5566 | ||
428870ff | 5567 | static nvlist_t * |
0bc8fd78 | 5568 | make_random_props(void) |
428870ff BB |
5569 | { |
5570 | nvlist_t *props; | |
5571 | ||
5572 | if (ztest_random(2) == 0) | |
5573 | return (NULL); | |
5574 | ||
5575 | VERIFY(nvlist_alloc(&props, NV_UNIQUE_NAME, 0) == 0); | |
5576 | VERIFY(nvlist_add_uint64(props, "autoreplace", 1) == 0); | |
5577 | ||
5578 | (void) printf("props:\n"); | |
5579 | dump_nvlist(props, 4); | |
5580 | ||
5581 | return (props); | |
5582 | } | |
5583 | ||
34dc7c2f BB |
5584 | /* |
5585 | * Create a storage pool with the given name and initial vdev size. | |
428870ff | 5586 | * Then test spa_freeze() functionality. |
34dc7c2f BB |
5587 | */ |
5588 | static void | |
428870ff | 5589 | ztest_init(ztest_shared_t *zs) |
34dc7c2f BB |
5590 | { |
5591 | spa_t *spa; | |
428870ff BB |
5592 | nvlist_t *nvroot, *props; |
5593 | ||
1e33ac1e BB |
5594 | mutex_init(&zs->zs_vdev_lock, NULL, MUTEX_DEFAULT, NULL); |
5595 | rw_init(&zs->zs_name_lock, NULL, RW_DEFAULT, NULL); | |
34dc7c2f BB |
5596 | |
5597 | kernel_init(FREAD | FWRITE); | |
5598 | ||
5599 | /* | |
5600 | * Create the storage pool. | |
5601 | */ | |
428870ff BB |
5602 | (void) spa_destroy(zs->zs_pool); |
5603 | ztest_shared->zs_vdev_next_leaf = 0; | |
5604 | zs->zs_splits = 0; | |
5605 | zs->zs_mirrors = zopt_mirrors; | |
b128c09f | 5606 | nvroot = make_vdev_root(NULL, NULL, zopt_vdev_size, 0, |
428870ff BB |
5607 | 0, zopt_raidz, zs->zs_mirrors, 1); |
5608 | props = make_random_props(); | |
5609 | VERIFY3U(0, ==, spa_create(zs->zs_pool, nvroot, props, NULL, NULL)); | |
34dc7c2f BB |
5610 | nvlist_free(nvroot); |
5611 | ||
428870ff | 5612 | VERIFY3U(0, ==, spa_open(zs->zs_pool, &spa, FTAG)); |
9babb374 | 5613 | metaslab_sz = 1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift; |
34dc7c2f BB |
5614 | spa_close(spa, FTAG); |
5615 | ||
5616 | kernel_fini(); | |
428870ff BB |
5617 | |
5618 | ztest_run_zdb(zs->zs_pool); | |
5619 | ||
5620 | ztest_freeze(zs); | |
5621 | ||
5622 | ztest_run_zdb(zs->zs_pool); | |
572e2857 | 5623 | |
1e33ac1e BB |
5624 | (void) rw_destroy(&zs->zs_name_lock); |
5625 | (void) mutex_destroy(&zs->zs_vdev_lock); | |
34dc7c2f BB |
5626 | } |
5627 | ||
5628 | int | |
5629 | main(int argc, char **argv) | |
5630 | { | |
5631 | int kills = 0; | |
5632 | int iters = 0; | |
34dc7c2f | 5633 | ztest_shared_t *zs; |
428870ff | 5634 | size_t shared_size; |
34dc7c2f BB |
5635 | ztest_info_t *zi; |
5636 | char timebuf[100]; | |
5637 | char numbuf[6]; | |
428870ff | 5638 | spa_t *spa; |
d6320ddb | 5639 | int i, f; |
34dc7c2f BB |
5640 | |
5641 | (void) setvbuf(stdout, NULL, _IOLBF, 0); | |
5642 | ||
34dc7c2f BB |
5643 | ztest_random_fd = open("/dev/urandom", O_RDONLY); |
5644 | ||
0e8d1b2d | 5645 | dprintf_setup(&argc, argv); |
34dc7c2f BB |
5646 | process_options(argc, argv); |
5647 | ||
428870ff | 5648 | /* Override location of zpool.cache */ |
0e5b68e0 BB |
5649 | VERIFY(asprintf((char **)&spa_config_path, "%s/zpool.cache", |
5650 | zopt_dir) != -1); | |
428870ff | 5651 | |
34dc7c2f BB |
5652 | /* |
5653 | * Blow away any existing copy of zpool.cache | |
5654 | */ | |
5655 | if (zopt_init != 0) | |
428870ff BB |
5656 | (void) remove(spa_config_path); |
5657 | ||
5658 | shared_size = sizeof (*zs) + zopt_datasets * sizeof (ztest_ds_t); | |
34dc7c2f BB |
5659 | |
5660 | zs = ztest_shared = (void *)mmap(0, | |
428870ff | 5661 | P2ROUNDUP(shared_size, getpagesize()), |
34dc7c2f BB |
5662 | PROT_READ | PROT_WRITE, MAP_SHARED | MAP_ANON, -1, 0); |
5663 | ||
5664 | if (zopt_verbose >= 1) { | |
5665 | (void) printf("%llu vdevs, %d datasets, %d threads," | |
5666 | " %llu seconds...\n", | |
5667 | (u_longlong_t)zopt_vdevs, zopt_datasets, zopt_threads, | |
5668 | (u_longlong_t)zopt_time); | |
5669 | } | |
5670 | ||
5671 | /* | |
5672 | * Create and initialize our storage pool. | |
5673 | */ | |
d6320ddb | 5674 | for (i = 1; i <= zopt_init; i++) { |
34dc7c2f BB |
5675 | bzero(zs, sizeof (ztest_shared_t)); |
5676 | if (zopt_verbose >= 3 && zopt_init != 1) | |
5677 | (void) printf("ztest_init(), pass %d\n", i); | |
428870ff BB |
5678 | zs->zs_pool = zopt_pool; |
5679 | ztest_init(zs); | |
34dc7c2f BB |
5680 | } |
5681 | ||
428870ff BB |
5682 | zs->zs_pool = zopt_pool; |
5683 | zs->zs_proc_start = gethrtime(); | |
5684 | zs->zs_proc_stop = zs->zs_proc_start + zopt_time * NANOSEC; | |
34dc7c2f | 5685 | |
d6320ddb | 5686 | for (f = 0; f < ZTEST_FUNCS; f++) { |
428870ff | 5687 | zi = &zs->zs_info[f]; |
34dc7c2f | 5688 | *zi = ztest_info[f]; |
428870ff BB |
5689 | if (zs->zs_proc_start + zi->zi_interval[0] > zs->zs_proc_stop) |
5690 | zi->zi_call_next = UINT64_MAX; | |
34dc7c2f | 5691 | else |
428870ff BB |
5692 | zi->zi_call_next = zs->zs_proc_start + |
5693 | ztest_random(2 * zi->zi_interval[0] + 1); | |
34dc7c2f BB |
5694 | } |
5695 | ||
34dc7c2f BB |
5696 | /* |
5697 | * Run the tests in a loop. These tests include fault injection | |
5698 | * to verify that self-healing data works, and forced crashes | |
5699 | * to verify that we never lose on-disk consistency. | |
5700 | */ | |
428870ff | 5701 | while (gethrtime() < zs->zs_proc_stop) { |
34dc7c2f BB |
5702 | int status; |
5703 | pid_t pid; | |
34dc7c2f BB |
5704 | |
5705 | /* | |
5706 | * Initialize the workload counters for each function. | |
5707 | */ | |
d6320ddb | 5708 | for (f = 0; f < ZTEST_FUNCS; f++) { |
34dc7c2f | 5709 | zi = &zs->zs_info[f]; |
428870ff | 5710 | zi->zi_call_count = 0; |
34dc7c2f BB |
5711 | zi->zi_call_time = 0; |
5712 | } | |
5713 | ||
9babb374 BB |
5714 | /* Set the allocation switch size */ |
5715 | metaslab_df_alloc_threshold = ztest_random(metaslab_sz / 4) + 1; | |
5716 | ||
34dc7c2f BB |
5717 | pid = fork(); |
5718 | ||
5719 | if (pid == -1) | |
5720 | fatal(1, "fork failed"); | |
5721 | ||
5722 | if (pid == 0) { /* child */ | |
5723 | struct rlimit rl = { 1024, 1024 }; | |
5724 | (void) setrlimit(RLIMIT_NOFILE, &rl); | |
5725 | (void) enable_extended_FILE_stdio(-1, -1); | |
428870ff | 5726 | ztest_run(zs); |
34dc7c2f BB |
5727 | exit(0); |
5728 | } | |
5729 | ||
5730 | while (waitpid(pid, &status, 0) != pid) | |
5731 | continue; | |
5732 | ||
5733 | if (WIFEXITED(status)) { | |
5734 | if (WEXITSTATUS(status) != 0) { | |
5735 | (void) fprintf(stderr, | |
5736 | "child exited with code %d\n", | |
5737 | WEXITSTATUS(status)); | |
5738 | exit(2); | |
5739 | } | |
5740 | } else if (WIFSIGNALED(status)) { | |
5741 | if (WTERMSIG(status) != SIGKILL) { | |
5742 | (void) fprintf(stderr, | |
5743 | "child died with signal %d\n", | |
5744 | WTERMSIG(status)); | |
5745 | exit(3); | |
5746 | } | |
5747 | kills++; | |
5748 | } else { | |
5749 | (void) fprintf(stderr, "something strange happened " | |
5750 | "to child\n"); | |
5751 | exit(4); | |
5752 | } | |
5753 | ||
5754 | iters++; | |
5755 | ||
5756 | if (zopt_verbose >= 1) { | |
5757 | hrtime_t now = gethrtime(); | |
5758 | ||
428870ff BB |
5759 | now = MIN(now, zs->zs_proc_stop); |
5760 | print_time(zs->zs_proc_stop - now, timebuf); | |
34dc7c2f BB |
5761 | nicenum(zs->zs_space, numbuf); |
5762 | ||
5763 | (void) printf("Pass %3d, %8s, %3llu ENOSPC, " | |
5764 | "%4.1f%% of %5s used, %3.0f%% done, %8s to go\n", | |
5765 | iters, | |
5766 | WIFEXITED(status) ? "Complete" : "SIGKILL", | |
5767 | (u_longlong_t)zs->zs_enospc_count, | |
5768 | 100.0 * zs->zs_alloc / zs->zs_space, | |
5769 | numbuf, | |
428870ff | 5770 | 100.0 * (now - zs->zs_proc_start) / |
34dc7c2f BB |
5771 | (zopt_time * NANOSEC), timebuf); |
5772 | } | |
5773 | ||
5774 | if (zopt_verbose >= 2) { | |
5775 | (void) printf("\nWorkload summary:\n\n"); | |
5776 | (void) printf("%7s %9s %s\n", | |
5777 | "Calls", "Time", "Function"); | |
5778 | (void) printf("%7s %9s %s\n", | |
5779 | "-----", "----", "--------"); | |
d6320ddb | 5780 | for (f = 0; f < ZTEST_FUNCS; f++) { |
34dc7c2f BB |
5781 | Dl_info dli; |
5782 | ||
5783 | zi = &zs->zs_info[f]; | |
5784 | print_time(zi->zi_call_time, timebuf); | |
5785 | (void) dladdr((void *)zi->zi_func, &dli); | |
5786 | (void) printf("%7llu %9s %s\n", | |
428870ff | 5787 | (u_longlong_t)zi->zi_call_count, timebuf, |
34dc7c2f BB |
5788 | dli.dli_sname); |
5789 | } | |
5790 | (void) printf("\n"); | |
5791 | } | |
5792 | ||
5793 | /* | |
428870ff BB |
5794 | * It's possible that we killed a child during a rename test, |
5795 | * in which case we'll have a 'ztest_tmp' pool lying around | |
5796 | * instead of 'ztest'. Do a blind rename in case this happened. | |
34dc7c2f | 5797 | */ |
428870ff BB |
5798 | kernel_init(FREAD); |
5799 | if (spa_open(zopt_pool, &spa, FTAG) == 0) { | |
5800 | spa_close(spa, FTAG); | |
5801 | } else { | |
5802 | char tmpname[MAXNAMELEN]; | |
5803 | kernel_fini(); | |
5804 | kernel_init(FREAD | FWRITE); | |
5805 | (void) snprintf(tmpname, sizeof (tmpname), "%s_tmp", | |
5806 | zopt_pool); | |
5807 | (void) spa_rename(tmpname, zopt_pool); | |
5808 | } | |
34dc7c2f | 5809 | kernel_fini(); |
34dc7c2f | 5810 | |
428870ff BB |
5811 | ztest_run_zdb(zopt_pool); |
5812 | } | |
34dc7c2f BB |
5813 | |
5814 | if (zopt_verbose >= 1) { | |
5815 | (void) printf("%d killed, %d completed, %.0f%% kill rate\n", | |
5816 | kills, iters - kills, (100.0 * kills) / MAX(1, iters)); | |
5817 | } | |
5818 | ||
5819 | return (0); | |
5820 | } |