]>
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. |
c242c188 | 23 | * Copyright (c) 2012 by Delphix. All rights reserved. |
3541dc6d | 24 | * Copyright 2011 Nexenta Systems, Inc. All rights reserved. |
34dc7c2f BB |
25 | */ |
26 | ||
34dc7c2f BB |
27 | /* |
28 | * The objective of this program is to provide a DMU/ZAP/SPA stress test | |
29 | * that runs entirely in userland, is easy to use, and easy to extend. | |
30 | * | |
31 | * The overall design of the ztest program is as follows: | |
32 | * | |
33 | * (1) For each major functional area (e.g. adding vdevs to a pool, | |
34 | * creating and destroying datasets, reading and writing objects, etc) | |
35 | * we have a simple routine to test that functionality. These | |
36 | * individual routines do not have to do anything "stressful". | |
37 | * | |
38 | * (2) We turn these simple functionality tests into a stress test by | |
39 | * running them all in parallel, with as many threads as desired, | |
40 | * and spread across as many datasets, objects, and vdevs as desired. | |
41 | * | |
42 | * (3) While all this is happening, we inject faults into the pool to | |
43 | * verify that self-healing data really works. | |
44 | * | |
45 | * (4) Every time we open a dataset, we change its checksum and compression | |
46 | * functions. Thus even individual objects vary from block to block | |
47 | * in which checksum they use and whether they're compressed. | |
48 | * | |
49 | * (5) To verify that we never lose on-disk consistency after a crash, | |
50 | * we run the entire test in a child of the main process. | |
51 | * At random times, the child self-immolates with a SIGKILL. | |
52 | * This is the software equivalent of pulling the power cord. | |
53 | * The parent then runs the test again, using the existing | |
c242c188 CS |
54 | * storage pool, as many times as desired. If backwards compatability |
55 | * testing is enabled ztest will sometimes run the "older" version | |
56 | * of ztest after a SIGKILL. | |
34dc7c2f BB |
57 | * |
58 | * (6) To verify that we don't have future leaks or temporal incursions, | |
59 | * many of the functional tests record the transaction group number | |
60 | * as part of their data. When reading old data, they verify that | |
61 | * the transaction group number is less than the current, open txg. | |
62 | * If you add a new test, please do this if applicable. | |
63 | * | |
1e33ac1e BB |
64 | * (7) Threads are created with a reduced stack size, for sanity checking. |
65 | * Therefore, it's important not to allocate huge buffers on the stack. | |
66 | * | |
34dc7c2f BB |
67 | * When run with no arguments, ztest runs for about five minutes and |
68 | * produces no output if successful. To get a little bit of information, | |
69 | * specify -V. To get more information, specify -VV, and so on. | |
70 | * | |
71 | * To turn this into an overnight stress test, use -T to specify run time. | |
72 | * | |
73 | * You can ask more more vdevs [-v], datasets [-d], or threads [-t] | |
74 | * to increase the pool capacity, fanout, and overall stress level. | |
75 | * | |
c242c188 CS |
76 | * Use the -k option to set the desired frequency of kills. |
77 | * | |
78 | * When ztest invokes itself it passes all relevant information through a | |
79 | * temporary file which is mmap-ed in the child process. This allows shared | |
80 | * memory to survive the exec syscall. The ztest_shared_hdr_t struct is always | |
81 | * stored at offset 0 of this file and contains information on the size and | |
82 | * number of shared structures in the file. The information stored in this file | |
83 | * must remain backwards compatible with older versions of ztest so that | |
84 | * ztest can invoke them during backwards compatibility testing (-B). | |
34dc7c2f BB |
85 | */ |
86 | ||
87 | #include <sys/zfs_context.h> | |
88 | #include <sys/spa.h> | |
89 | #include <sys/dmu.h> | |
90 | #include <sys/txg.h> | |
9babb374 | 91 | #include <sys/dbuf.h> |
34dc7c2f | 92 | #include <sys/zap.h> |
34dc7c2f BB |
93 | #include <sys/dmu_objset.h> |
94 | #include <sys/poll.h> | |
95 | #include <sys/stat.h> | |
96 | #include <sys/time.h> | |
97 | #include <sys/wait.h> | |
98 | #include <sys/mman.h> | |
99 | #include <sys/resource.h> | |
100 | #include <sys/zio.h> | |
34dc7c2f | 101 | #include <sys/zil.h> |
428870ff | 102 | #include <sys/zil_impl.h> |
34dc7c2f | 103 | #include <sys/vdev_impl.h> |
b128c09f | 104 | #include <sys/vdev_file.h> |
34dc7c2f | 105 | #include <sys/spa_impl.h> |
428870ff | 106 | #include <sys/metaslab_impl.h> |
34dc7c2f | 107 | #include <sys/dsl_prop.h> |
9babb374 | 108 | #include <sys/dsl_dataset.h> |
428870ff BB |
109 | #include <sys/dsl_scan.h> |
110 | #include <sys/zio_checksum.h> | |
34dc7c2f | 111 | #include <sys/refcount.h> |
9ae529ec | 112 | #include <sys/zfeature.h> |
34dc7c2f BB |
113 | #include <stdio.h> |
114 | #include <stdio_ext.h> | |
115 | #include <stdlib.h> | |
116 | #include <unistd.h> | |
117 | #include <signal.h> | |
118 | #include <umem.h> | |
119 | #include <dlfcn.h> | |
120 | #include <ctype.h> | |
121 | #include <math.h> | |
122 | #include <sys/fs/zfs.h> | |
428870ff | 123 | #include <libnvpair.h> |
34dc7c2f | 124 | |
9d81146b ED |
125 | static int ztest_fd_data = -1; |
126 | static int ztest_fd_rand = -1; | |
c242c188 CS |
127 | |
128 | typedef struct ztest_shared_hdr { | |
129 | uint64_t zh_hdr_size; | |
130 | uint64_t zh_opts_size; | |
131 | uint64_t zh_size; | |
132 | uint64_t zh_stats_size; | |
133 | uint64_t zh_stats_count; | |
134 | uint64_t zh_ds_size; | |
135 | uint64_t zh_ds_count; | |
136 | } ztest_shared_hdr_t; | |
137 | ||
138 | static ztest_shared_hdr_t *ztest_shared_hdr; | |
139 | ||
140 | typedef struct ztest_shared_opts { | |
141 | char zo_pool[MAXNAMELEN]; | |
142 | char zo_dir[MAXNAMELEN]; | |
143 | char zo_alt_ztest[MAXNAMELEN]; | |
144 | char zo_alt_libpath[MAXNAMELEN]; | |
145 | uint64_t zo_vdevs; | |
146 | uint64_t zo_vdevtime; | |
147 | size_t zo_vdev_size; | |
148 | int zo_ashift; | |
149 | int zo_mirrors; | |
150 | int zo_raidz; | |
151 | int zo_raidz_parity; | |
152 | int zo_datasets; | |
153 | int zo_threads; | |
154 | uint64_t zo_passtime; | |
155 | uint64_t zo_killrate; | |
156 | int zo_verbose; | |
157 | int zo_init; | |
158 | uint64_t zo_time; | |
159 | uint64_t zo_maxloops; | |
160 | uint64_t zo_metaslab_gang_bang; | |
161 | } ztest_shared_opts_t; | |
162 | ||
163 | static const ztest_shared_opts_t ztest_opts_defaults = { | |
164 | .zo_pool = { 'z', 't', 'e', 's', 't', '\0' }, | |
165 | .zo_dir = { '/', 't', 'm', 'p', '\0' }, | |
166 | .zo_alt_ztest = { '\0' }, | |
167 | .zo_alt_libpath = { '\0' }, | |
168 | .zo_vdevs = 5, | |
169 | .zo_ashift = SPA_MINBLOCKSHIFT, | |
170 | .zo_mirrors = 2, | |
171 | .zo_raidz = 4, | |
172 | .zo_raidz_parity = 1, | |
173 | .zo_vdev_size = SPA_MINDEVSIZE, | |
174 | .zo_datasets = 7, | |
175 | .zo_threads = 23, | |
176 | .zo_passtime = 60, /* 60 seconds */ | |
177 | .zo_killrate = 70, /* 70% kill rate */ | |
178 | .zo_verbose = 0, | |
179 | .zo_init = 1, | |
180 | .zo_time = 300, /* 5 minutes */ | |
181 | .zo_maxloops = 50, /* max loops during spa_freeze() */ | |
182 | .zo_metaslab_gang_bang = 32 << 10 | |
183 | }; | |
184 | ||
185 | extern uint64_t metaslab_gang_bang; | |
186 | extern uint64_t metaslab_df_alloc_threshold; | |
187 | ||
188 | static ztest_shared_opts_t *ztest_shared_opts; | |
189 | static ztest_shared_opts_t ztest_opts; | |
190 | ||
191 | typedef struct ztest_shared_ds { | |
192 | uint64_t zd_seq; | |
193 | } ztest_shared_ds_t; | |
194 | ||
195 | static ztest_shared_ds_t *ztest_shared_ds; | |
196 | #define ZTEST_GET_SHARED_DS(d) (&ztest_shared_ds[d]) | |
428870ff BB |
197 | |
198 | #define BT_MAGIC 0x123456789abcdefULL | |
c242c188 CS |
199 | #define MAXFAULTS() \ |
200 | (MAX(zs->zs_mirrors, 1) * (ztest_opts.zo_raidz_parity + 1) - 1) | |
428870ff BB |
201 | |
202 | enum ztest_io_type { | |
203 | ZTEST_IO_WRITE_TAG, | |
204 | ZTEST_IO_WRITE_PATTERN, | |
205 | ZTEST_IO_WRITE_ZEROES, | |
206 | ZTEST_IO_TRUNCATE, | |
207 | ZTEST_IO_SETATTR, | |
208 | ZTEST_IO_TYPES | |
209 | }; | |
34dc7c2f BB |
210 | |
211 | typedef struct ztest_block_tag { | |
428870ff | 212 | uint64_t bt_magic; |
34dc7c2f BB |
213 | uint64_t bt_objset; |
214 | uint64_t bt_object; | |
215 | uint64_t bt_offset; | |
428870ff | 216 | uint64_t bt_gen; |
34dc7c2f | 217 | uint64_t bt_txg; |
428870ff | 218 | uint64_t bt_crtxg; |
34dc7c2f BB |
219 | } ztest_block_tag_t; |
220 | ||
428870ff BB |
221 | typedef struct bufwad { |
222 | uint64_t bw_index; | |
223 | uint64_t bw_txg; | |
224 | uint64_t bw_data; | |
225 | } bufwad_t; | |
226 | ||
227 | /* | |
228 | * XXX -- fix zfs range locks to be generic so we can use them here. | |
229 | */ | |
230 | typedef enum { | |
231 | RL_READER, | |
232 | RL_WRITER, | |
233 | RL_APPEND | |
234 | } rl_type_t; | |
235 | ||
236 | typedef struct rll { | |
237 | void *rll_writer; | |
238 | int rll_readers; | |
1e33ac1e BB |
239 | kmutex_t rll_lock; |
240 | kcondvar_t rll_cv; | |
428870ff BB |
241 | } rll_t; |
242 | ||
243 | typedef struct rl { | |
244 | uint64_t rl_object; | |
245 | uint64_t rl_offset; | |
246 | uint64_t rl_size; | |
247 | rll_t *rl_lock; | |
248 | } rl_t; | |
249 | ||
250 | #define ZTEST_RANGE_LOCKS 64 | |
251 | #define ZTEST_OBJECT_LOCKS 64 | |
252 | ||
253 | /* | |
254 | * Object descriptor. Used as a template for object lookup/create/remove. | |
255 | */ | |
256 | typedef struct ztest_od { | |
257 | uint64_t od_dir; | |
258 | uint64_t od_object; | |
259 | dmu_object_type_t od_type; | |
260 | dmu_object_type_t od_crtype; | |
261 | uint64_t od_blocksize; | |
262 | uint64_t od_crblocksize; | |
263 | uint64_t od_gen; | |
264 | uint64_t od_crgen; | |
265 | char od_name[MAXNAMELEN]; | |
266 | } ztest_od_t; | |
34dc7c2f | 267 | |
428870ff BB |
268 | /* |
269 | * Per-dataset state. | |
270 | */ | |
271 | typedef struct ztest_ds { | |
c242c188 | 272 | ztest_shared_ds_t *zd_shared; |
428870ff | 273 | objset_t *zd_os; |
3e31d2b0 | 274 | krwlock_t zd_zilog_lock; |
428870ff | 275 | zilog_t *zd_zilog; |
428870ff BB |
276 | ztest_od_t *zd_od; /* debugging aid */ |
277 | char zd_name[MAXNAMELEN]; | |
1e33ac1e | 278 | kmutex_t zd_dirobj_lock; |
428870ff BB |
279 | rll_t zd_object_lock[ZTEST_OBJECT_LOCKS]; |
280 | rll_t zd_range_lock[ZTEST_RANGE_LOCKS]; | |
281 | } ztest_ds_t; | |
282 | ||
283 | /* | |
284 | * Per-iteration state. | |
285 | */ | |
286 | typedef void ztest_func_t(ztest_ds_t *zd, uint64_t id); | |
287 | ||
288 | typedef struct ztest_info { | |
289 | ztest_func_t *zi_func; /* test function */ | |
290 | uint64_t zi_iters; /* iterations per execution */ | |
291 | uint64_t *zi_interval; /* execute every <interval> seconds */ | |
428870ff | 292 | } ztest_info_t; |
34dc7c2f | 293 | |
c242c188 CS |
294 | typedef struct ztest_shared_callstate { |
295 | uint64_t zc_count; /* per-pass count */ | |
296 | uint64_t zc_time; /* per-pass time */ | |
297 | uint64_t zc_next; /* next time to call this function */ | |
298 | } ztest_shared_callstate_t; | |
299 | ||
300 | static ztest_shared_callstate_t *ztest_shared_callstate; | |
301 | #define ZTEST_GET_SHARED_CALLSTATE(c) (&ztest_shared_callstate[c]) | |
302 | ||
34dc7c2f BB |
303 | /* |
304 | * Note: these aren't static because we want dladdr() to work. | |
305 | */ | |
306 | ztest_func_t ztest_dmu_read_write; | |
307 | ztest_func_t ztest_dmu_write_parallel; | |
308 | ztest_func_t ztest_dmu_object_alloc_free; | |
428870ff | 309 | ztest_func_t ztest_dmu_commit_callbacks; |
34dc7c2f BB |
310 | ztest_func_t ztest_zap; |
311 | ztest_func_t ztest_zap_parallel; | |
428870ff | 312 | ztest_func_t ztest_zil_commit; |
3e31d2b0 | 313 | ztest_func_t ztest_zil_remount; |
428870ff | 314 | ztest_func_t ztest_dmu_read_write_zcopy; |
34dc7c2f | 315 | ztest_func_t ztest_dmu_objset_create_destroy; |
428870ff BB |
316 | ztest_func_t ztest_dmu_prealloc; |
317 | ztest_func_t ztest_fzap; | |
34dc7c2f | 318 | ztest_func_t ztest_dmu_snapshot_create_destroy; |
428870ff BB |
319 | ztest_func_t ztest_dsl_prop_get_set; |
320 | ztest_func_t ztest_spa_prop_get_set; | |
34dc7c2f BB |
321 | ztest_func_t ztest_spa_create_destroy; |
322 | ztest_func_t ztest_fault_inject; | |
428870ff BB |
323 | ztest_func_t ztest_ddt_repair; |
324 | ztest_func_t ztest_dmu_snapshot_hold; | |
b128c09f | 325 | ztest_func_t ztest_spa_rename; |
428870ff BB |
326 | ztest_func_t ztest_scrub; |
327 | ztest_func_t ztest_dsl_dataset_promote_busy; | |
34dc7c2f BB |
328 | ztest_func_t ztest_vdev_attach_detach; |
329 | ztest_func_t ztest_vdev_LUN_growth; | |
330 | ztest_func_t ztest_vdev_add_remove; | |
b128c09f | 331 | ztest_func_t ztest_vdev_aux_add_remove; |
428870ff | 332 | ztest_func_t ztest_split_pool; |
3541dc6d | 333 | ztest_func_t ztest_reguid; |
ea0b2538 | 334 | ztest_func_t ztest_spa_upgrade; |
34dc7c2f | 335 | |
428870ff BB |
336 | uint64_t zopt_always = 0ULL * NANOSEC; /* all the time */ |
337 | uint64_t zopt_incessant = 1ULL * NANOSEC / 10; /* every 1/10 second */ | |
338 | uint64_t zopt_often = 1ULL * NANOSEC; /* every second */ | |
339 | uint64_t zopt_sometimes = 10ULL * NANOSEC; /* every 10 seconds */ | |
340 | uint64_t zopt_rarely = 60ULL * NANOSEC; /* every 60 seconds */ | |
34dc7c2f BB |
341 | |
342 | ztest_info_t ztest_info[] = { | |
343 | { ztest_dmu_read_write, 1, &zopt_always }, | |
428870ff | 344 | { ztest_dmu_write_parallel, 10, &zopt_always }, |
34dc7c2f | 345 | { ztest_dmu_object_alloc_free, 1, &zopt_always }, |
428870ff | 346 | { ztest_dmu_commit_callbacks, 1, &zopt_always }, |
34dc7c2f BB |
347 | { ztest_zap, 30, &zopt_always }, |
348 | { ztest_zap_parallel, 100, &zopt_always }, | |
428870ff BB |
349 | { ztest_split_pool, 1, &zopt_always }, |
350 | { ztest_zil_commit, 1, &zopt_incessant }, | |
3e31d2b0 | 351 | { ztest_zil_remount, 1, &zopt_sometimes }, |
428870ff BB |
352 | { ztest_dmu_read_write_zcopy, 1, &zopt_often }, |
353 | { ztest_dmu_objset_create_destroy, 1, &zopt_often }, | |
354 | { ztest_dsl_prop_get_set, 1, &zopt_often }, | |
355 | { ztest_spa_prop_get_set, 1, &zopt_sometimes }, | |
356 | #if 0 | |
357 | { ztest_dmu_prealloc, 1, &zopt_sometimes }, | |
358 | #endif | |
359 | { ztest_fzap, 1, &zopt_sometimes }, | |
360 | { ztest_dmu_snapshot_create_destroy, 1, &zopt_sometimes }, | |
361 | { ztest_spa_create_destroy, 1, &zopt_sometimes }, | |
34dc7c2f | 362 | { ztest_fault_inject, 1, &zopt_sometimes }, |
428870ff BB |
363 | { ztest_ddt_repair, 1, &zopt_sometimes }, |
364 | { ztest_dmu_snapshot_hold, 1, &zopt_sometimes }, | |
3541dc6d | 365 | { ztest_reguid, 1, &zopt_sometimes }, |
34dc7c2f | 366 | { ztest_spa_rename, 1, &zopt_rarely }, |
428870ff | 367 | { ztest_scrub, 1, &zopt_rarely }, |
ea0b2538 | 368 | { ztest_spa_upgrade, 1, &zopt_rarely }, |
9babb374 | 369 | { ztest_dsl_dataset_promote_busy, 1, &zopt_rarely }, |
3bc7e0fb | 370 | { ztest_vdev_attach_detach, 1, &zopt_rarely }, |
428870ff | 371 | { ztest_vdev_LUN_growth, 1, &zopt_rarely }, |
c242c188 CS |
372 | { ztest_vdev_add_remove, 1, |
373 | &ztest_opts.zo_vdevtime }, | |
374 | { ztest_vdev_aux_add_remove, 1, | |
375 | &ztest_opts.zo_vdevtime }, | |
34dc7c2f BB |
376 | }; |
377 | ||
378 | #define ZTEST_FUNCS (sizeof (ztest_info) / sizeof (ztest_info_t)) | |
379 | ||
428870ff BB |
380 | /* |
381 | * The following struct is used to hold a list of uncalled commit callbacks. | |
382 | * The callbacks are ordered by txg number. | |
383 | */ | |
384 | typedef struct ztest_cb_list { | |
1e33ac1e BB |
385 | kmutex_t zcl_callbacks_lock; |
386 | list_t zcl_callbacks; | |
428870ff | 387 | } ztest_cb_list_t; |
34dc7c2f BB |
388 | |
389 | /* | |
390 | * Stuff we need to share writably between parent and child. | |
391 | */ | |
392 | typedef struct ztest_shared { | |
c242c188 | 393 | boolean_t zs_do_init; |
428870ff BB |
394 | hrtime_t zs_proc_start; |
395 | hrtime_t zs_proc_stop; | |
396 | hrtime_t zs_thread_start; | |
397 | hrtime_t zs_thread_stop; | |
398 | hrtime_t zs_thread_kill; | |
34dc7c2f | 399 | uint64_t zs_enospc_count; |
428870ff BB |
400 | uint64_t zs_vdev_next_leaf; |
401 | uint64_t zs_vdev_aux; | |
34dc7c2f BB |
402 | uint64_t zs_alloc; |
403 | uint64_t zs_space; | |
428870ff BB |
404 | uint64_t zs_splits; |
405 | uint64_t zs_mirrors; | |
c242c188 CS |
406 | uint64_t zs_metaslab_sz; |
407 | uint64_t zs_metaslab_df_alloc_threshold; | |
408 | uint64_t zs_guid; | |
34dc7c2f BB |
409 | } ztest_shared_t; |
410 | ||
428870ff BB |
411 | #define ID_PARALLEL -1ULL |
412 | ||
34dc7c2f | 413 | static char ztest_dev_template[] = "%s/%s.%llua"; |
b128c09f | 414 | static char ztest_aux_template[] = "%s/%s.%s.%llu"; |
428870ff | 415 | ztest_shared_t *ztest_shared; |
34dc7c2f | 416 | |
c242c188 CS |
417 | static spa_t *ztest_spa = NULL; |
418 | static ztest_ds_t *ztest_ds; | |
419 | ||
420 | static kmutex_t ztest_vdev_lock; | |
3bc7e0fb GW |
421 | |
422 | /* | |
423 | * The ztest_name_lock protects the pool and dataset namespace used by | |
424 | * the individual tests. To modify the namespace, consumers must grab | |
425 | * this lock as writer. Grabbing the lock as reader will ensure that the | |
426 | * namespace does not change while the lock is held. | |
427 | */ | |
c242c188 | 428 | static krwlock_t ztest_name_lock; |
34dc7c2f | 429 | |
c242c188 | 430 | static boolean_t ztest_dump_core = B_TRUE; |
b128c09f | 431 | static boolean_t ztest_exiting; |
34dc7c2f | 432 | |
428870ff BB |
433 | /* Global commit callback list */ |
434 | static ztest_cb_list_t zcl; | |
090ff092 RC |
435 | /* Commit cb delay */ |
436 | static uint64_t zc_min_txg_delay = UINT64_MAX; | |
437 | static int zc_cb_counter = 0; | |
438 | ||
439 | /* | |
440 | * Minimum number of commit callbacks that need to be registered for us to check | |
441 | * whether the minimum txg delay is acceptable. | |
442 | */ | |
443 | #define ZTEST_COMMIT_CB_MIN_REG 100 | |
444 | ||
445 | /* | |
446 | * If a number of txgs equal to this threshold have been created after a commit | |
447 | * callback has been registered but not called, then we assume there is an | |
448 | * implementation bug. | |
449 | */ | |
450 | #define ZTEST_COMMIT_CB_THRESH (TXG_CONCURRENT_STATES + 1000) | |
428870ff | 451 | |
34dc7c2f | 452 | extern uint64_t metaslab_gang_bang; |
9babb374 | 453 | extern uint64_t metaslab_df_alloc_threshold; |
34dc7c2f | 454 | |
428870ff BB |
455 | enum ztest_object { |
456 | ZTEST_META_DNODE = 0, | |
457 | ZTEST_DIROBJ, | |
458 | ZTEST_OBJECTS | |
459 | }; | |
34dc7c2f BB |
460 | |
461 | static void usage(boolean_t) __NORETURN; | |
462 | ||
463 | /* | |
464 | * These libumem hooks provide a reasonable set of defaults for the allocator's | |
465 | * debugging facilities. | |
466 | */ | |
467 | const char * | |
0bc8fd78 | 468 | _umem_debug_init(void) |
34dc7c2f BB |
469 | { |
470 | return ("default,verbose"); /* $UMEM_DEBUG setting */ | |
471 | } | |
472 | ||
473 | const char * | |
474 | _umem_logging_init(void) | |
475 | { | |
476 | return ("fail,contents"); /* $UMEM_LOGGING setting */ | |
477 | } | |
478 | ||
479 | #define FATAL_MSG_SZ 1024 | |
480 | ||
481 | char *fatal_msg; | |
482 | ||
483 | static void | |
484 | fatal(int do_perror, char *message, ...) | |
485 | { | |
486 | va_list args; | |
487 | int save_errno = errno; | |
40b84e7a | 488 | char *buf; |
34dc7c2f BB |
489 | |
490 | (void) fflush(stdout); | |
40b84e7a | 491 | buf = umem_alloc(FATAL_MSG_SZ, UMEM_NOFAIL); |
34dc7c2f BB |
492 | |
493 | va_start(args, message); | |
494 | (void) sprintf(buf, "ztest: "); | |
495 | /* LINTED */ | |
496 | (void) vsprintf(buf + strlen(buf), message, args); | |
497 | va_end(args); | |
498 | if (do_perror) { | |
499 | (void) snprintf(buf + strlen(buf), FATAL_MSG_SZ - strlen(buf), | |
500 | ": %s", strerror(save_errno)); | |
501 | } | |
502 | (void) fprintf(stderr, "%s\n", buf); | |
503 | fatal_msg = buf; /* to ease debugging */ | |
504 | if (ztest_dump_core) | |
505 | abort(); | |
506 | exit(3); | |
507 | } | |
508 | ||
509 | static int | |
510 | str2shift(const char *buf) | |
511 | { | |
512 | const char *ends = "BKMGTPEZ"; | |
513 | int i; | |
514 | ||
515 | if (buf[0] == '\0') | |
516 | return (0); | |
517 | for (i = 0; i < strlen(ends); i++) { | |
518 | if (toupper(buf[0]) == ends[i]) | |
519 | break; | |
520 | } | |
521 | if (i == strlen(ends)) { | |
522 | (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", | |
523 | buf); | |
524 | usage(B_FALSE); | |
525 | } | |
526 | if (buf[1] == '\0' || (toupper(buf[1]) == 'B' && buf[2] == '\0')) { | |
527 | return (10*i); | |
528 | } | |
529 | (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", buf); | |
530 | usage(B_FALSE); | |
531 | /* NOTREACHED */ | |
532 | } | |
533 | ||
534 | static uint64_t | |
535 | nicenumtoull(const char *buf) | |
536 | { | |
537 | char *end; | |
538 | uint64_t val; | |
539 | ||
540 | val = strtoull(buf, &end, 0); | |
541 | if (end == buf) { | |
542 | (void) fprintf(stderr, "ztest: bad numeric value: %s\n", buf); | |
543 | usage(B_FALSE); | |
544 | } else if (end[0] == '.') { | |
545 | double fval = strtod(buf, &end); | |
546 | fval *= pow(2, str2shift(end)); | |
547 | if (fval > UINT64_MAX) { | |
548 | (void) fprintf(stderr, "ztest: value too large: %s\n", | |
549 | buf); | |
550 | usage(B_FALSE); | |
551 | } | |
552 | val = (uint64_t)fval; | |
553 | } else { | |
554 | int shift = str2shift(end); | |
555 | if (shift >= 64 || (val << shift) >> shift != val) { | |
556 | (void) fprintf(stderr, "ztest: value too large: %s\n", | |
557 | buf); | |
558 | usage(B_FALSE); | |
559 | } | |
560 | val <<= shift; | |
561 | } | |
562 | return (val); | |
563 | } | |
564 | ||
565 | static void | |
566 | usage(boolean_t requested) | |
567 | { | |
c242c188 CS |
568 | const ztest_shared_opts_t *zo = &ztest_opts_defaults; |
569 | ||
34dc7c2f BB |
570 | char nice_vdev_size[10]; |
571 | char nice_gang_bang[10]; | |
572 | FILE *fp = requested ? stdout : stderr; | |
573 | ||
c242c188 CS |
574 | nicenum(zo->zo_vdev_size, nice_vdev_size); |
575 | nicenum(zo->zo_metaslab_gang_bang, nice_gang_bang); | |
34dc7c2f BB |
576 | |
577 | (void) fprintf(fp, "Usage: %s\n" | |
578 | "\t[-v vdevs (default: %llu)]\n" | |
579 | "\t[-s size_of_each_vdev (default: %s)]\n" | |
428870ff | 580 | "\t[-a alignment_shift (default: %d)] use 0 for random\n" |
34dc7c2f BB |
581 | "\t[-m mirror_copies (default: %d)]\n" |
582 | "\t[-r raidz_disks (default: %d)]\n" | |
583 | "\t[-R raidz_parity (default: %d)]\n" | |
584 | "\t[-d datasets (default: %d)]\n" | |
585 | "\t[-t threads (default: %d)]\n" | |
586 | "\t[-g gang_block_threshold (default: %s)]\n" | |
428870ff BB |
587 | "\t[-i init_count (default: %d)] initialize pool i times\n" |
588 | "\t[-k kill_percentage (default: %llu%%)]\n" | |
34dc7c2f | 589 | "\t[-p pool_name (default: %s)]\n" |
428870ff BB |
590 | "\t[-f dir (default: %s)] file directory for vdev files\n" |
591 | "\t[-V] verbose (use multiple times for ever more blather)\n" | |
592 | "\t[-E] use existing pool instead of creating new one\n" | |
593 | "\t[-T time (default: %llu sec)] total run time\n" | |
594 | "\t[-F freezeloops (default: %llu)] max loops in spa_freeze()\n" | |
595 | "\t[-P passtime (default: %llu sec)] time per pass\n" | |
c242c188 | 596 | "\t[-B alt_ztest (default: <none>)] alternate ztest path\n" |
34dc7c2f BB |
597 | "\t[-h] (print help)\n" |
598 | "", | |
c242c188 CS |
599 | zo->zo_pool, |
600 | (u_longlong_t)zo->zo_vdevs, /* -v */ | |
34dc7c2f | 601 | nice_vdev_size, /* -s */ |
c242c188 CS |
602 | zo->zo_ashift, /* -a */ |
603 | zo->zo_mirrors, /* -m */ | |
604 | zo->zo_raidz, /* -r */ | |
605 | zo->zo_raidz_parity, /* -R */ | |
606 | zo->zo_datasets, /* -d */ | |
607 | zo->zo_threads, /* -t */ | |
34dc7c2f | 608 | nice_gang_bang, /* -g */ |
c242c188 CS |
609 | zo->zo_init, /* -i */ |
610 | (u_longlong_t)zo->zo_killrate, /* -k */ | |
611 | zo->zo_pool, /* -p */ | |
612 | zo->zo_dir, /* -f */ | |
613 | (u_longlong_t)zo->zo_time, /* -T */ | |
614 | (u_longlong_t)zo->zo_maxloops, /* -F */ | |
615 | (u_longlong_t)zo->zo_passtime); | |
34dc7c2f BB |
616 | exit(requested ? 0 : 1); |
617 | } | |
618 | ||
34dc7c2f BB |
619 | static void |
620 | process_options(int argc, char **argv) | |
621 | { | |
c242c188 CS |
622 | char *path; |
623 | ztest_shared_opts_t *zo = &ztest_opts; | |
624 | ||
34dc7c2f BB |
625 | int opt; |
626 | uint64_t value; | |
c242c188 | 627 | char altdir[MAXNAMELEN] = { 0 }; |
34dc7c2f | 628 | |
c242c188 | 629 | bcopy(&ztest_opts_defaults, zo, sizeof (*zo)); |
34dc7c2f | 630 | |
34dc7c2f | 631 | while ((opt = getopt(argc, argv, |
c242c188 | 632 | "v:s:a:m:r:R:d:t:g:i:k:p:f:VET:P:hF:B:")) != EOF) { |
34dc7c2f BB |
633 | value = 0; |
634 | switch (opt) { | |
635 | case 'v': | |
636 | case 's': | |
637 | case 'a': | |
638 | case 'm': | |
639 | case 'r': | |
640 | case 'R': | |
641 | case 'd': | |
642 | case 't': | |
643 | case 'g': | |
644 | case 'i': | |
645 | case 'k': | |
646 | case 'T': | |
647 | case 'P': | |
428870ff | 648 | case 'F': |
34dc7c2f BB |
649 | value = nicenumtoull(optarg); |
650 | } | |
651 | switch (opt) { | |
652 | case 'v': | |
c242c188 | 653 | zo->zo_vdevs = value; |
34dc7c2f BB |
654 | break; |
655 | case 's': | |
c242c188 | 656 | zo->zo_vdev_size = MAX(SPA_MINDEVSIZE, value); |
34dc7c2f BB |
657 | break; |
658 | case 'a': | |
c242c188 | 659 | zo->zo_ashift = value; |
34dc7c2f BB |
660 | break; |
661 | case 'm': | |
c242c188 | 662 | zo->zo_mirrors = value; |
34dc7c2f BB |
663 | break; |
664 | case 'r': | |
c242c188 | 665 | zo->zo_raidz = MAX(1, value); |
34dc7c2f BB |
666 | break; |
667 | case 'R': | |
c242c188 | 668 | zo->zo_raidz_parity = MIN(MAX(value, 1), 3); |
34dc7c2f BB |
669 | break; |
670 | case 'd': | |
c242c188 | 671 | zo->zo_datasets = MAX(1, value); |
34dc7c2f BB |
672 | break; |
673 | case 't': | |
c242c188 | 674 | zo->zo_threads = MAX(1, value); |
34dc7c2f BB |
675 | break; |
676 | case 'g': | |
c242c188 CS |
677 | zo->zo_metaslab_gang_bang = MAX(SPA_MINBLOCKSIZE << 1, |
678 | value); | |
34dc7c2f BB |
679 | break; |
680 | case 'i': | |
c242c188 | 681 | zo->zo_init = value; |
34dc7c2f BB |
682 | break; |
683 | case 'k': | |
c242c188 | 684 | zo->zo_killrate = value; |
34dc7c2f BB |
685 | break; |
686 | case 'p': | |
c242c188 CS |
687 | (void) strlcpy(zo->zo_pool, optarg, |
688 | sizeof (zo->zo_pool)); | |
34dc7c2f BB |
689 | break; |
690 | case 'f': | |
c242c188 CS |
691 | path = realpath(optarg, NULL); |
692 | if (path == NULL) { | |
693 | (void) fprintf(stderr, "error: %s: %s\n", | |
694 | optarg, strerror(errno)); | |
695 | usage(B_FALSE); | |
696 | } else { | |
697 | (void) strlcpy(zo->zo_dir, path, | |
698 | sizeof (zo->zo_dir)); | |
699 | } | |
34dc7c2f BB |
700 | break; |
701 | case 'V': | |
c242c188 | 702 | zo->zo_verbose++; |
34dc7c2f BB |
703 | break; |
704 | case 'E': | |
c242c188 | 705 | zo->zo_init = 0; |
34dc7c2f BB |
706 | break; |
707 | case 'T': | |
c242c188 | 708 | zo->zo_time = value; |
34dc7c2f BB |
709 | break; |
710 | case 'P': | |
c242c188 | 711 | zo->zo_passtime = MAX(1, value); |
34dc7c2f | 712 | break; |
428870ff | 713 | case 'F': |
c242c188 CS |
714 | zo->zo_maxloops = MAX(1, value); |
715 | break; | |
716 | case 'B': | |
717 | (void) strlcpy(altdir, optarg, sizeof (altdir)); | |
428870ff | 718 | break; |
34dc7c2f BB |
719 | case 'h': |
720 | usage(B_TRUE); | |
721 | break; | |
722 | case '?': | |
723 | default: | |
724 | usage(B_FALSE); | |
725 | break; | |
726 | } | |
727 | } | |
728 | ||
c242c188 | 729 | zo->zo_raidz_parity = MIN(zo->zo_raidz_parity, zo->zo_raidz - 1); |
34dc7c2f | 730 | |
c242c188 CS |
731 | zo->zo_vdevtime = |
732 | (zo->zo_vdevs > 0 ? zo->zo_time * NANOSEC / zo->zo_vdevs : | |
428870ff | 733 | UINT64_MAX >> 2); |
c242c188 CS |
734 | |
735 | if (strlen(altdir) > 0) { | |
ae380cfa BB |
736 | char *cmd; |
737 | char *realaltdir; | |
c242c188 CS |
738 | char *bin; |
739 | char *ztest; | |
740 | char *isa; | |
741 | int isalen; | |
742 | ||
ae380cfa BB |
743 | cmd = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
744 | realaltdir = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
745 | ||
5be98cfe | 746 | VERIFY(NULL != realpath(getexecname(), cmd)); |
c242c188 CS |
747 | if (0 != access(altdir, F_OK)) { |
748 | ztest_dump_core = B_FALSE; | |
749 | fatal(B_TRUE, "invalid alternate ztest path: %s", | |
750 | altdir); | |
751 | } | |
752 | VERIFY(NULL != realpath(altdir, realaltdir)); | |
753 | ||
754 | /* | |
755 | * 'cmd' should be of the form "<anything>/usr/bin/<isa>/ztest". | |
756 | * We want to extract <isa> to determine if we should use | |
757 | * 32 or 64 bit binaries. | |
758 | */ | |
759 | bin = strstr(cmd, "/usr/bin/"); | |
760 | ztest = strstr(bin, "/ztest"); | |
761 | isa = bin + 9; | |
762 | isalen = ztest - isa; | |
763 | (void) snprintf(zo->zo_alt_ztest, sizeof (zo->zo_alt_ztest), | |
764 | "%s/usr/bin/%.*s/ztest", realaltdir, isalen, isa); | |
765 | (void) snprintf(zo->zo_alt_libpath, sizeof (zo->zo_alt_libpath), | |
766 | "%s/usr/lib/%.*s", realaltdir, isalen, isa); | |
767 | ||
768 | if (0 != access(zo->zo_alt_ztest, X_OK)) { | |
769 | ztest_dump_core = B_FALSE; | |
770 | fatal(B_TRUE, "invalid alternate ztest: %s", | |
771 | zo->zo_alt_ztest); | |
772 | } else if (0 != access(zo->zo_alt_libpath, X_OK)) { | |
773 | ztest_dump_core = B_FALSE; | |
774 | fatal(B_TRUE, "invalid alternate lib directory %s", | |
775 | zo->zo_alt_libpath); | |
776 | } | |
ae380cfa BB |
777 | |
778 | umem_free(cmd, MAXPATHLEN); | |
779 | umem_free(realaltdir, MAXPATHLEN); | |
c242c188 | 780 | } |
428870ff BB |
781 | } |
782 | ||
783 | static void | |
784 | ztest_kill(ztest_shared_t *zs) | |
785 | { | |
c242c188 CS |
786 | zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(ztest_spa)); |
787 | zs->zs_space = metaslab_class_get_space(spa_normal_class(ztest_spa)); | |
428870ff BB |
788 | (void) kill(getpid(), SIGKILL); |
789 | } | |
790 | ||
791 | static uint64_t | |
792 | ztest_random(uint64_t range) | |
793 | { | |
794 | uint64_t r; | |
795 | ||
9d81146b ED |
796 | ASSERT3S(ztest_fd_rand, >=, 0); |
797 | ||
428870ff BB |
798 | if (range == 0) |
799 | return (0); | |
800 | ||
9d81146b | 801 | if (read(ztest_fd_rand, &r, sizeof (r)) != sizeof (r)) |
428870ff BB |
802 | fatal(1, "short read from /dev/urandom"); |
803 | ||
804 | return (r % range); | |
805 | } | |
806 | ||
807 | /* ARGSUSED */ | |
808 | static void | |
809 | ztest_record_enospc(const char *s) | |
810 | { | |
811 | ztest_shared->zs_enospc_count++; | |
34dc7c2f BB |
812 | } |
813 | ||
814 | static uint64_t | |
815 | ztest_get_ashift(void) | |
816 | { | |
c242c188 | 817 | if (ztest_opts.zo_ashift == 0) |
34dc7c2f | 818 | return (SPA_MINBLOCKSHIFT + ztest_random(3)); |
c242c188 | 819 | return (ztest_opts.zo_ashift); |
34dc7c2f BB |
820 | } |
821 | ||
822 | static nvlist_t * | |
ea0b2538 | 823 | make_vdev_file(char *path, char *aux, char *pool, size_t size, uint64_t ashift) |
34dc7c2f | 824 | { |
40b84e7a | 825 | char *pathbuf; |
34dc7c2f | 826 | uint64_t vdev; |
34dc7c2f BB |
827 | nvlist_t *file; |
828 | ||
40b84e7a BB |
829 | pathbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
830 | ||
b128c09f BB |
831 | if (ashift == 0) |
832 | ashift = ztest_get_ashift(); | |
833 | ||
834 | if (path == NULL) { | |
835 | path = pathbuf; | |
836 | ||
837 | if (aux != NULL) { | |
838 | vdev = ztest_shared->zs_vdev_aux; | |
c242c188 CS |
839 | (void) snprintf(path, MAXPATHLEN, |
840 | ztest_aux_template, ztest_opts.zo_dir, | |
ea0b2538 GW |
841 | pool == NULL ? ztest_opts.zo_pool : pool, |
842 | aux, vdev); | |
b128c09f | 843 | } else { |
428870ff | 844 | vdev = ztest_shared->zs_vdev_next_leaf++; |
c242c188 CS |
845 | (void) snprintf(path, MAXPATHLEN, |
846 | ztest_dev_template, ztest_opts.zo_dir, | |
ea0b2538 | 847 | pool == NULL ? ztest_opts.zo_pool : pool, vdev); |
b128c09f BB |
848 | } |
849 | } | |
34dc7c2f | 850 | |
b128c09f BB |
851 | if (size != 0) { |
852 | int fd = open(path, O_RDWR | O_CREAT | O_TRUNC, 0666); | |
34dc7c2f | 853 | if (fd == -1) |
b128c09f | 854 | fatal(1, "can't open %s", path); |
34dc7c2f | 855 | if (ftruncate(fd, size) != 0) |
b128c09f | 856 | fatal(1, "can't ftruncate %s", path); |
34dc7c2f BB |
857 | (void) close(fd); |
858 | } | |
859 | ||
860 | VERIFY(nvlist_alloc(&file, NV_UNIQUE_NAME, 0) == 0); | |
861 | VERIFY(nvlist_add_string(file, ZPOOL_CONFIG_TYPE, VDEV_TYPE_FILE) == 0); | |
b128c09f | 862 | VERIFY(nvlist_add_string(file, ZPOOL_CONFIG_PATH, path) == 0); |
34dc7c2f | 863 | VERIFY(nvlist_add_uint64(file, ZPOOL_CONFIG_ASHIFT, ashift) == 0); |
40b84e7a | 864 | umem_free(pathbuf, MAXPATHLEN); |
34dc7c2f BB |
865 | |
866 | return (file); | |
867 | } | |
868 | ||
869 | static nvlist_t * | |
ea0b2538 GW |
870 | make_vdev_raidz(char *path, char *aux, char *pool, size_t size, |
871 | uint64_t ashift, int r) | |
34dc7c2f BB |
872 | { |
873 | nvlist_t *raidz, **child; | |
874 | int c; | |
875 | ||
876 | if (r < 2) | |
ea0b2538 | 877 | return (make_vdev_file(path, aux, pool, size, ashift)); |
34dc7c2f BB |
878 | child = umem_alloc(r * sizeof (nvlist_t *), UMEM_NOFAIL); |
879 | ||
880 | for (c = 0; c < r; c++) | |
ea0b2538 | 881 | child[c] = make_vdev_file(path, aux, pool, size, ashift); |
34dc7c2f BB |
882 | |
883 | VERIFY(nvlist_alloc(&raidz, NV_UNIQUE_NAME, 0) == 0); | |
884 | VERIFY(nvlist_add_string(raidz, ZPOOL_CONFIG_TYPE, | |
885 | VDEV_TYPE_RAIDZ) == 0); | |
886 | VERIFY(nvlist_add_uint64(raidz, ZPOOL_CONFIG_NPARITY, | |
c242c188 | 887 | ztest_opts.zo_raidz_parity) == 0); |
34dc7c2f BB |
888 | VERIFY(nvlist_add_nvlist_array(raidz, ZPOOL_CONFIG_CHILDREN, |
889 | child, r) == 0); | |
890 | ||
891 | for (c = 0; c < r; c++) | |
892 | nvlist_free(child[c]); | |
893 | ||
894 | umem_free(child, r * sizeof (nvlist_t *)); | |
895 | ||
896 | return (raidz); | |
897 | } | |
898 | ||
899 | static nvlist_t * | |
ea0b2538 GW |
900 | make_vdev_mirror(char *path, char *aux, char *pool, size_t size, |
901 | uint64_t ashift, int r, int m) | |
34dc7c2f BB |
902 | { |
903 | nvlist_t *mirror, **child; | |
904 | int c; | |
905 | ||
906 | if (m < 1) | |
ea0b2538 | 907 | return (make_vdev_raidz(path, aux, pool, size, ashift, r)); |
34dc7c2f BB |
908 | |
909 | child = umem_alloc(m * sizeof (nvlist_t *), UMEM_NOFAIL); | |
910 | ||
911 | for (c = 0; c < m; c++) | |
ea0b2538 | 912 | child[c] = make_vdev_raidz(path, aux, pool, size, ashift, r); |
34dc7c2f BB |
913 | |
914 | VERIFY(nvlist_alloc(&mirror, NV_UNIQUE_NAME, 0) == 0); | |
915 | VERIFY(nvlist_add_string(mirror, ZPOOL_CONFIG_TYPE, | |
916 | VDEV_TYPE_MIRROR) == 0); | |
917 | VERIFY(nvlist_add_nvlist_array(mirror, ZPOOL_CONFIG_CHILDREN, | |
918 | child, m) == 0); | |
34dc7c2f BB |
919 | |
920 | for (c = 0; c < m; c++) | |
921 | nvlist_free(child[c]); | |
922 | ||
923 | umem_free(child, m * sizeof (nvlist_t *)); | |
924 | ||
925 | return (mirror); | |
926 | } | |
927 | ||
928 | static nvlist_t * | |
ea0b2538 GW |
929 | make_vdev_root(char *path, char *aux, char *pool, size_t size, uint64_t ashift, |
930 | int log, int r, int m, int t) | |
34dc7c2f BB |
931 | { |
932 | nvlist_t *root, **child; | |
933 | int c; | |
934 | ||
935 | ASSERT(t > 0); | |
936 | ||
937 | child = umem_alloc(t * sizeof (nvlist_t *), UMEM_NOFAIL); | |
938 | ||
b128c09f | 939 | for (c = 0; c < t; c++) { |
ea0b2538 GW |
940 | child[c] = make_vdev_mirror(path, aux, pool, size, ashift, |
941 | r, m); | |
b128c09f BB |
942 | VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_IS_LOG, |
943 | log) == 0); | |
944 | } | |
34dc7c2f BB |
945 | |
946 | VERIFY(nvlist_alloc(&root, NV_UNIQUE_NAME, 0) == 0); | |
947 | VERIFY(nvlist_add_string(root, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT) == 0); | |
b128c09f | 948 | VERIFY(nvlist_add_nvlist_array(root, aux ? aux : ZPOOL_CONFIG_CHILDREN, |
34dc7c2f BB |
949 | child, t) == 0); |
950 | ||
951 | for (c = 0; c < t; c++) | |
952 | nvlist_free(child[c]); | |
953 | ||
954 | umem_free(child, t * sizeof (nvlist_t *)); | |
955 | ||
956 | return (root); | |
957 | } | |
958 | ||
ea0b2538 GW |
959 | /* |
960 | * Find a random spa version. Returns back a random spa version in the | |
961 | * range [initial_version, SPA_VERSION_FEATURES]. | |
962 | */ | |
963 | static uint64_t | |
964 | ztest_random_spa_version(uint64_t initial_version) | |
965 | { | |
966 | uint64_t version = initial_version; | |
967 | ||
968 | if (version <= SPA_VERSION_BEFORE_FEATURES) { | |
969 | version = version + | |
970 | ztest_random(SPA_VERSION_BEFORE_FEATURES - version + 1); | |
971 | } | |
972 | ||
973 | if (version > SPA_VERSION_BEFORE_FEATURES) | |
974 | version = SPA_VERSION_FEATURES; | |
975 | ||
976 | ASSERT(SPA_VERSION_IS_SUPPORTED(version)); | |
977 | return (version); | |
978 | } | |
979 | ||
428870ff BB |
980 | static int |
981 | ztest_random_blocksize(void) | |
34dc7c2f | 982 | { |
428870ff BB |
983 | return (1 << (SPA_MINBLOCKSHIFT + |
984 | ztest_random(SPA_MAXBLOCKSHIFT - SPA_MINBLOCKSHIFT + 1))); | |
985 | } | |
34dc7c2f | 986 | |
428870ff BB |
987 | static int |
988 | ztest_random_ibshift(void) | |
989 | { | |
990 | return (DN_MIN_INDBLKSHIFT + | |
991 | ztest_random(DN_MAX_INDBLKSHIFT - DN_MIN_INDBLKSHIFT + 1)); | |
34dc7c2f BB |
992 | } |
993 | ||
428870ff BB |
994 | static uint64_t |
995 | ztest_random_vdev_top(spa_t *spa, boolean_t log_ok) | |
34dc7c2f | 996 | { |
428870ff BB |
997 | uint64_t top; |
998 | vdev_t *rvd = spa->spa_root_vdev; | |
999 | vdev_t *tvd; | |
34dc7c2f | 1000 | |
428870ff | 1001 | ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); |
34dc7c2f | 1002 | |
428870ff BB |
1003 | do { |
1004 | top = ztest_random(rvd->vdev_children); | |
1005 | tvd = rvd->vdev_child[top]; | |
1006 | } while (tvd->vdev_ishole || (tvd->vdev_islog && !log_ok) || | |
1007 | tvd->vdev_mg == NULL || tvd->vdev_mg->mg_class == NULL); | |
34dc7c2f | 1008 | |
428870ff | 1009 | return (top); |
34dc7c2f BB |
1010 | } |
1011 | ||
428870ff BB |
1012 | static uint64_t |
1013 | ztest_random_dsl_prop(zfs_prop_t prop) | |
34dc7c2f | 1014 | { |
428870ff BB |
1015 | uint64_t value; |
1016 | ||
1017 | do { | |
1018 | value = zfs_prop_random_value(prop, ztest_random(-1ULL)); | |
1019 | } while (prop == ZFS_PROP_CHECKSUM && value == ZIO_CHECKSUM_OFF); | |
1020 | ||
1021 | return (value); | |
34dc7c2f BB |
1022 | } |
1023 | ||
34dc7c2f | 1024 | static int |
428870ff BB |
1025 | ztest_dsl_prop_set_uint64(char *osname, zfs_prop_t prop, uint64_t value, |
1026 | boolean_t inherit) | |
34dc7c2f | 1027 | { |
428870ff BB |
1028 | const char *propname = zfs_prop_to_name(prop); |
1029 | const char *valname; | |
40b84e7a | 1030 | char *setpoint; |
428870ff | 1031 | uint64_t curval; |
34dc7c2f BB |
1032 | int error; |
1033 | ||
428870ff BB |
1034 | error = dsl_prop_set(osname, propname, |
1035 | (inherit ? ZPROP_SRC_NONE : ZPROP_SRC_LOCAL), | |
1036 | sizeof (value), 1, &value); | |
34dc7c2f | 1037 | |
428870ff BB |
1038 | if (error == ENOSPC) { |
1039 | ztest_record_enospc(FTAG); | |
34dc7c2f BB |
1040 | return (error); |
1041 | } | |
c99c9001 | 1042 | ASSERT0(error); |
34dc7c2f | 1043 | |
40b84e7a | 1044 | setpoint = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
428870ff BB |
1045 | VERIFY3U(dsl_prop_get(osname, propname, sizeof (curval), |
1046 | 1, &curval, setpoint), ==, 0); | |
1047 | ||
c242c188 | 1048 | if (ztest_opts.zo_verbose >= 6) { |
428870ff BB |
1049 | VERIFY(zfs_prop_index_to_string(prop, curval, &valname) == 0); |
1050 | (void) printf("%s %s = %s at '%s'\n", | |
1051 | osname, propname, valname, setpoint); | |
34dc7c2f | 1052 | } |
40b84e7a | 1053 | umem_free(setpoint, MAXPATHLEN); |
34dc7c2f BB |
1054 | |
1055 | return (error); | |
1056 | } | |
1057 | ||
1058 | static int | |
c242c188 | 1059 | ztest_spa_prop_set_uint64(zpool_prop_t prop, uint64_t value) |
34dc7c2f | 1060 | { |
c242c188 | 1061 | spa_t *spa = ztest_spa; |
428870ff | 1062 | nvlist_t *props = NULL; |
34dc7c2f BB |
1063 | int error; |
1064 | ||
428870ff BB |
1065 | VERIFY(nvlist_alloc(&props, NV_UNIQUE_NAME, 0) == 0); |
1066 | VERIFY(nvlist_add_uint64(props, zpool_prop_to_name(prop), value) == 0); | |
34dc7c2f | 1067 | |
428870ff BB |
1068 | error = spa_prop_set(spa, props); |
1069 | ||
1070 | nvlist_free(props); | |
1071 | ||
1072 | if (error == ENOSPC) { | |
1073 | ztest_record_enospc(FTAG); | |
34dc7c2f BB |
1074 | return (error); |
1075 | } | |
c99c9001 | 1076 | ASSERT0(error); |
34dc7c2f BB |
1077 | |
1078 | return (error); | |
1079 | } | |
1080 | ||
428870ff BB |
1081 | static void |
1082 | ztest_rll_init(rll_t *rll) | |
1083 | { | |
1084 | rll->rll_writer = NULL; | |
1085 | rll->rll_readers = 0; | |
1e33ac1e BB |
1086 | mutex_init(&rll->rll_lock, NULL, MUTEX_DEFAULT, NULL); |
1087 | cv_init(&rll->rll_cv, NULL, CV_DEFAULT, NULL); | |
428870ff | 1088 | } |
34dc7c2f | 1089 | |
428870ff BB |
1090 | static void |
1091 | ztest_rll_destroy(rll_t *rll) | |
34dc7c2f | 1092 | { |
428870ff BB |
1093 | ASSERT(rll->rll_writer == NULL); |
1094 | ASSERT(rll->rll_readers == 0); | |
1e33ac1e BB |
1095 | mutex_destroy(&rll->rll_lock); |
1096 | cv_destroy(&rll->rll_cv); | |
428870ff | 1097 | } |
34dc7c2f | 1098 | |
428870ff BB |
1099 | static void |
1100 | ztest_rll_lock(rll_t *rll, rl_type_t type) | |
1101 | { | |
1e33ac1e | 1102 | mutex_enter(&rll->rll_lock); |
34dc7c2f | 1103 | |
428870ff BB |
1104 | if (type == RL_READER) { |
1105 | while (rll->rll_writer != NULL) | |
1e33ac1e | 1106 | (void) cv_wait(&rll->rll_cv, &rll->rll_lock); |
428870ff BB |
1107 | rll->rll_readers++; |
1108 | } else { | |
1109 | while (rll->rll_writer != NULL || rll->rll_readers) | |
1e33ac1e | 1110 | (void) cv_wait(&rll->rll_cv, &rll->rll_lock); |
428870ff BB |
1111 | rll->rll_writer = curthread; |
1112 | } | |
34dc7c2f | 1113 | |
1e33ac1e | 1114 | mutex_exit(&rll->rll_lock); |
428870ff | 1115 | } |
34dc7c2f | 1116 | |
428870ff BB |
1117 | static void |
1118 | ztest_rll_unlock(rll_t *rll) | |
1119 | { | |
1e33ac1e | 1120 | mutex_enter(&rll->rll_lock); |
34dc7c2f | 1121 | |
428870ff BB |
1122 | if (rll->rll_writer) { |
1123 | ASSERT(rll->rll_readers == 0); | |
1124 | rll->rll_writer = NULL; | |
1125 | } else { | |
1126 | ASSERT(rll->rll_readers != 0); | |
1127 | ASSERT(rll->rll_writer == NULL); | |
1128 | rll->rll_readers--; | |
1129 | } | |
34dc7c2f | 1130 | |
428870ff | 1131 | if (rll->rll_writer == NULL && rll->rll_readers == 0) |
1e33ac1e | 1132 | cv_broadcast(&rll->rll_cv); |
428870ff | 1133 | |
1e33ac1e | 1134 | mutex_exit(&rll->rll_lock); |
34dc7c2f BB |
1135 | } |
1136 | ||
428870ff BB |
1137 | static void |
1138 | ztest_object_lock(ztest_ds_t *zd, uint64_t object, rl_type_t type) | |
b128c09f | 1139 | { |
428870ff | 1140 | rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; |
b128c09f | 1141 | |
428870ff BB |
1142 | ztest_rll_lock(rll, type); |
1143 | } | |
b128c09f | 1144 | |
428870ff BB |
1145 | static void |
1146 | ztest_object_unlock(ztest_ds_t *zd, uint64_t object) | |
1147 | { | |
1148 | rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; | |
b128c09f | 1149 | |
428870ff | 1150 | ztest_rll_unlock(rll); |
b128c09f BB |
1151 | } |
1152 | ||
428870ff BB |
1153 | static rl_t * |
1154 | ztest_range_lock(ztest_ds_t *zd, uint64_t object, uint64_t offset, | |
1155 | uint64_t size, rl_type_t type) | |
34dc7c2f | 1156 | { |
428870ff BB |
1157 | uint64_t hash = object ^ (offset % (ZTEST_RANGE_LOCKS + 1)); |
1158 | rll_t *rll = &zd->zd_range_lock[hash & (ZTEST_RANGE_LOCKS - 1)]; | |
1159 | rl_t *rl; | |
34dc7c2f | 1160 | |
428870ff BB |
1161 | rl = umem_alloc(sizeof (*rl), UMEM_NOFAIL); |
1162 | rl->rl_object = object; | |
1163 | rl->rl_offset = offset; | |
1164 | rl->rl_size = size; | |
1165 | rl->rl_lock = rll; | |
34dc7c2f | 1166 | |
428870ff BB |
1167 | ztest_rll_lock(rll, type); |
1168 | ||
1169 | return (rl); | |
1170 | } | |
34dc7c2f | 1171 | |
428870ff BB |
1172 | static void |
1173 | ztest_range_unlock(rl_t *rl) | |
1174 | { | |
1175 | rll_t *rll = rl->rl_lock; | |
34dc7c2f | 1176 | |
428870ff | 1177 | ztest_rll_unlock(rll); |
34dc7c2f | 1178 | |
428870ff BB |
1179 | umem_free(rl, sizeof (*rl)); |
1180 | } | |
34dc7c2f | 1181 | |
428870ff | 1182 | static void |
c242c188 | 1183 | ztest_zd_init(ztest_ds_t *zd, ztest_shared_ds_t *szd, objset_t *os) |
428870ff BB |
1184 | { |
1185 | zd->zd_os = os; | |
1186 | zd->zd_zilog = dmu_objset_zil(os); | |
c242c188 | 1187 | zd->zd_shared = szd; |
428870ff | 1188 | dmu_objset_name(os, zd->zd_name); |
d6320ddb | 1189 | int l; |
428870ff | 1190 | |
c242c188 CS |
1191 | if (zd->zd_shared != NULL) |
1192 | zd->zd_shared->zd_seq = 0; | |
1193 | ||
3e31d2b0 | 1194 | rw_init(&zd->zd_zilog_lock, NULL, RW_DEFAULT, NULL); |
1e33ac1e | 1195 | mutex_init(&zd->zd_dirobj_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 1196 | |
d6320ddb | 1197 | for (l = 0; l < ZTEST_OBJECT_LOCKS; l++) |
428870ff | 1198 | ztest_rll_init(&zd->zd_object_lock[l]); |
34dc7c2f | 1199 | |
d6320ddb | 1200 | for (l = 0; l < ZTEST_RANGE_LOCKS; l++) |
428870ff | 1201 | ztest_rll_init(&zd->zd_range_lock[l]); |
34dc7c2f BB |
1202 | } |
1203 | ||
428870ff BB |
1204 | static void |
1205 | ztest_zd_fini(ztest_ds_t *zd) | |
34dc7c2f | 1206 | { |
d6320ddb BB |
1207 | int l; |
1208 | ||
1e33ac1e | 1209 | mutex_destroy(&zd->zd_dirobj_lock); |
3e31d2b0 | 1210 | rw_destroy(&zd->zd_zilog_lock); |
34dc7c2f | 1211 | |
d6320ddb | 1212 | for (l = 0; l < ZTEST_OBJECT_LOCKS; l++) |
428870ff | 1213 | ztest_rll_destroy(&zd->zd_object_lock[l]); |
b128c09f | 1214 | |
d6320ddb | 1215 | for (l = 0; l < ZTEST_RANGE_LOCKS; l++) |
428870ff BB |
1216 | ztest_rll_destroy(&zd->zd_range_lock[l]); |
1217 | } | |
b128c09f | 1218 | |
428870ff | 1219 | #define TXG_MIGHTWAIT (ztest_random(10) == 0 ? TXG_NOWAIT : TXG_WAIT) |
b128c09f | 1220 | |
428870ff BB |
1221 | static uint64_t |
1222 | ztest_tx_assign(dmu_tx_t *tx, uint64_t txg_how, const char *tag) | |
1223 | { | |
1224 | uint64_t txg; | |
1225 | int error; | |
1226 | ||
1227 | /* | |
1228 | * Attempt to assign tx to some transaction group. | |
1229 | */ | |
1230 | error = dmu_tx_assign(tx, txg_how); | |
1231 | if (error) { | |
1232 | if (error == ERESTART) { | |
1233 | ASSERT(txg_how == TXG_NOWAIT); | |
1234 | dmu_tx_wait(tx); | |
1235 | } else { | |
1236 | ASSERT3U(error, ==, ENOSPC); | |
1237 | ztest_record_enospc(tag); | |
1238 | } | |
1239 | dmu_tx_abort(tx); | |
1240 | return (0); | |
1241 | } | |
1242 | txg = dmu_tx_get_txg(tx); | |
1243 | ASSERT(txg != 0); | |
1244 | return (txg); | |
1245 | } | |
1246 | ||
1247 | static void | |
1248 | ztest_pattern_set(void *buf, uint64_t size, uint64_t value) | |
1249 | { | |
1250 | uint64_t *ip = buf; | |
1251 | uint64_t *ip_end = (uint64_t *)((uintptr_t)buf + (uintptr_t)size); | |
1252 | ||
1253 | while (ip < ip_end) | |
1254 | *ip++ = value; | |
1255 | } | |
1256 | ||
1fde1e37 | 1257 | #ifndef NDEBUG |
428870ff BB |
1258 | static boolean_t |
1259 | ztest_pattern_match(void *buf, uint64_t size, uint64_t value) | |
1260 | { | |
1261 | uint64_t *ip = buf; | |
1262 | uint64_t *ip_end = (uint64_t *)((uintptr_t)buf + (uintptr_t)size); | |
1263 | uint64_t diff = 0; | |
1264 | ||
1265 | while (ip < ip_end) | |
1266 | diff |= (value - *ip++); | |
1267 | ||
1268 | return (diff == 0); | |
1269 | } | |
1fde1e37 | 1270 | #endif |
428870ff BB |
1271 | |
1272 | static void | |
1273 | ztest_bt_generate(ztest_block_tag_t *bt, objset_t *os, uint64_t object, | |
1274 | uint64_t offset, uint64_t gen, uint64_t txg, uint64_t crtxg) | |
1275 | { | |
1276 | bt->bt_magic = BT_MAGIC; | |
1277 | bt->bt_objset = dmu_objset_id(os); | |
1278 | bt->bt_object = object; | |
1279 | bt->bt_offset = offset; | |
1280 | bt->bt_gen = gen; | |
1281 | bt->bt_txg = txg; | |
1282 | bt->bt_crtxg = crtxg; | |
1283 | } | |
1284 | ||
1285 | static void | |
1286 | ztest_bt_verify(ztest_block_tag_t *bt, objset_t *os, uint64_t object, | |
1287 | uint64_t offset, uint64_t gen, uint64_t txg, uint64_t crtxg) | |
1288 | { | |
1289 | ASSERT(bt->bt_magic == BT_MAGIC); | |
1290 | ASSERT(bt->bt_objset == dmu_objset_id(os)); | |
1291 | ASSERT(bt->bt_object == object); | |
1292 | ASSERT(bt->bt_offset == offset); | |
1293 | ASSERT(bt->bt_gen <= gen); | |
1294 | ASSERT(bt->bt_txg <= txg); | |
1295 | ASSERT(bt->bt_crtxg == crtxg); | |
1296 | } | |
1297 | ||
1298 | static ztest_block_tag_t * | |
1299 | ztest_bt_bonus(dmu_buf_t *db) | |
1300 | { | |
1301 | dmu_object_info_t doi; | |
1302 | ztest_block_tag_t *bt; | |
1303 | ||
1304 | dmu_object_info_from_db(db, &doi); | |
1305 | ASSERT3U(doi.doi_bonus_size, <=, db->db_size); | |
1306 | ASSERT3U(doi.doi_bonus_size, >=, sizeof (*bt)); | |
1307 | bt = (void *)((char *)db->db_data + doi.doi_bonus_size - sizeof (*bt)); | |
1308 | ||
1309 | return (bt); | |
1310 | } | |
1311 | ||
1312 | /* | |
1313 | * ZIL logging ops | |
1314 | */ | |
1315 | ||
1316 | #define lrz_type lr_mode | |
1317 | #define lrz_blocksize lr_uid | |
1318 | #define lrz_ibshift lr_gid | |
1319 | #define lrz_bonustype lr_rdev | |
1320 | #define lrz_bonuslen lr_crtime[1] | |
1321 | ||
572e2857 | 1322 | static void |
428870ff BB |
1323 | ztest_log_create(ztest_ds_t *zd, dmu_tx_t *tx, lr_create_t *lr) |
1324 | { | |
1325 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1326 | size_t namesize = strlen(name) + 1; | |
1327 | itx_t *itx; | |
1328 | ||
1329 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1330 | return; |
428870ff BB |
1331 | |
1332 | itx = zil_itx_create(TX_CREATE, sizeof (*lr) + namesize); | |
1333 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1334 | sizeof (*lr) + namesize - sizeof (lr_t)); | |
1335 | ||
572e2857 | 1336 | zil_itx_assign(zd->zd_zilog, itx, tx); |
428870ff BB |
1337 | } |
1338 | ||
572e2857 BB |
1339 | static void |
1340 | ztest_log_remove(ztest_ds_t *zd, dmu_tx_t *tx, lr_remove_t *lr, uint64_t object) | |
428870ff BB |
1341 | { |
1342 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1343 | size_t namesize = strlen(name) + 1; | |
1344 | itx_t *itx; | |
1345 | ||
1346 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1347 | return; |
428870ff BB |
1348 | |
1349 | itx = zil_itx_create(TX_REMOVE, sizeof (*lr) + namesize); | |
1350 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1351 | sizeof (*lr) + namesize - sizeof (lr_t)); | |
1352 | ||
572e2857 BB |
1353 | itx->itx_oid = object; |
1354 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1355 | } |
1356 | ||
572e2857 | 1357 | static void |
428870ff BB |
1358 | ztest_log_write(ztest_ds_t *zd, dmu_tx_t *tx, lr_write_t *lr) |
1359 | { | |
1360 | itx_t *itx; | |
1361 | itx_wr_state_t write_state = ztest_random(WR_NUM_STATES); | |
1362 | ||
1363 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1364 | return; |
428870ff BB |
1365 | |
1366 | if (lr->lr_length > ZIL_MAX_LOG_DATA) | |
1367 | write_state = WR_INDIRECT; | |
1368 | ||
1369 | itx = zil_itx_create(TX_WRITE, | |
1370 | sizeof (*lr) + (write_state == WR_COPIED ? lr->lr_length : 0)); | |
1371 | ||
1372 | if (write_state == WR_COPIED && | |
1373 | dmu_read(zd->zd_os, lr->lr_foid, lr->lr_offset, lr->lr_length, | |
1374 | ((lr_write_t *)&itx->itx_lr) + 1, DMU_READ_NO_PREFETCH) != 0) { | |
1375 | zil_itx_destroy(itx); | |
1376 | itx = zil_itx_create(TX_WRITE, sizeof (*lr)); | |
1377 | write_state = WR_NEED_COPY; | |
1378 | } | |
1379 | itx->itx_private = zd; | |
1380 | itx->itx_wr_state = write_state; | |
1381 | itx->itx_sync = (ztest_random(8) == 0); | |
1382 | itx->itx_sod += (write_state == WR_NEED_COPY ? lr->lr_length : 0); | |
1383 | ||
1384 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1385 | sizeof (*lr) - sizeof (lr_t)); | |
1386 | ||
572e2857 | 1387 | zil_itx_assign(zd->zd_zilog, itx, tx); |
428870ff BB |
1388 | } |
1389 | ||
572e2857 | 1390 | static void |
428870ff BB |
1391 | ztest_log_truncate(ztest_ds_t *zd, dmu_tx_t *tx, lr_truncate_t *lr) |
1392 | { | |
1393 | itx_t *itx; | |
1394 | ||
1395 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1396 | return; |
428870ff BB |
1397 | |
1398 | itx = zil_itx_create(TX_TRUNCATE, sizeof (*lr)); | |
1399 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1400 | sizeof (*lr) - sizeof (lr_t)); | |
1401 | ||
572e2857 BB |
1402 | itx->itx_sync = B_FALSE; |
1403 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1404 | } |
1405 | ||
572e2857 | 1406 | static void |
428870ff BB |
1407 | ztest_log_setattr(ztest_ds_t *zd, dmu_tx_t *tx, lr_setattr_t *lr) |
1408 | { | |
1409 | itx_t *itx; | |
1410 | ||
1411 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1412 | return; |
428870ff BB |
1413 | |
1414 | itx = zil_itx_create(TX_SETATTR, sizeof (*lr)); | |
1415 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1416 | sizeof (*lr) - sizeof (lr_t)); | |
1417 | ||
572e2857 BB |
1418 | itx->itx_sync = B_FALSE; |
1419 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1420 | } |
1421 | ||
1422 | /* | |
1423 | * ZIL replay ops | |
1424 | */ | |
1425 | static int | |
1426 | ztest_replay_create(ztest_ds_t *zd, lr_create_t *lr, boolean_t byteswap) | |
1427 | { | |
1428 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1429 | objset_t *os = zd->zd_os; | |
1430 | ztest_block_tag_t *bbt; | |
1431 | dmu_buf_t *db; | |
1432 | dmu_tx_t *tx; | |
1433 | uint64_t txg; | |
1434 | int error = 0; | |
1435 | ||
1436 | if (byteswap) | |
1437 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1438 | ||
1439 | ASSERT(lr->lr_doid == ZTEST_DIROBJ); | |
1440 | ASSERT(name[0] != '\0'); | |
1441 | ||
1442 | tx = dmu_tx_create(os); | |
1443 | ||
1444 | dmu_tx_hold_zap(tx, lr->lr_doid, B_TRUE, name); | |
1445 | ||
1446 | if (lr->lrz_type == DMU_OT_ZAP_OTHER) { | |
1447 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, B_TRUE, NULL); | |
1448 | } else { | |
1449 | dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT); | |
1450 | } | |
1451 | ||
1452 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1453 | if (txg == 0) | |
1454 | return (ENOSPC); | |
1455 | ||
1456 | ASSERT(dmu_objset_zil(os)->zl_replay == !!lr->lr_foid); | |
1457 | ||
1458 | if (lr->lrz_type == DMU_OT_ZAP_OTHER) { | |
1459 | if (lr->lr_foid == 0) { | |
1460 | lr->lr_foid = zap_create(os, | |
1461 | lr->lrz_type, lr->lrz_bonustype, | |
1462 | lr->lrz_bonuslen, tx); | |
1463 | } else { | |
1464 | error = zap_create_claim(os, lr->lr_foid, | |
1465 | lr->lrz_type, lr->lrz_bonustype, | |
1466 | lr->lrz_bonuslen, tx); | |
1467 | } | |
1468 | } else { | |
1469 | if (lr->lr_foid == 0) { | |
1470 | lr->lr_foid = dmu_object_alloc(os, | |
1471 | lr->lrz_type, 0, lr->lrz_bonustype, | |
1472 | lr->lrz_bonuslen, tx); | |
1473 | } else { | |
1474 | error = dmu_object_claim(os, lr->lr_foid, | |
1475 | lr->lrz_type, 0, lr->lrz_bonustype, | |
1476 | lr->lrz_bonuslen, tx); | |
1477 | } | |
1478 | } | |
1479 | ||
1480 | if (error) { | |
1481 | ASSERT3U(error, ==, EEXIST); | |
1482 | ASSERT(zd->zd_zilog->zl_replay); | |
1483 | dmu_tx_commit(tx); | |
1484 | return (error); | |
1485 | } | |
1486 | ||
1487 | ASSERT(lr->lr_foid != 0); | |
1488 | ||
1489 | if (lr->lrz_type != DMU_OT_ZAP_OTHER) | |
1490 | VERIFY3U(0, ==, dmu_object_set_blocksize(os, lr->lr_foid, | |
1491 | lr->lrz_blocksize, lr->lrz_ibshift, tx)); | |
1492 | ||
1493 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
1494 | bbt = ztest_bt_bonus(db); | |
1495 | dmu_buf_will_dirty(db, tx); | |
1496 | ztest_bt_generate(bbt, os, lr->lr_foid, -1ULL, lr->lr_gen, txg, txg); | |
1497 | dmu_buf_rele(db, FTAG); | |
1498 | ||
1499 | VERIFY3U(0, ==, zap_add(os, lr->lr_doid, name, sizeof (uint64_t), 1, | |
1500 | &lr->lr_foid, tx)); | |
1501 | ||
1502 | (void) ztest_log_create(zd, tx, lr); | |
1503 | ||
1504 | dmu_tx_commit(tx); | |
1505 | ||
1506 | return (0); | |
1507 | } | |
1508 | ||
1509 | static int | |
1510 | ztest_replay_remove(ztest_ds_t *zd, lr_remove_t *lr, boolean_t byteswap) | |
1511 | { | |
1512 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1513 | objset_t *os = zd->zd_os; | |
1514 | dmu_object_info_t doi; | |
1515 | dmu_tx_t *tx; | |
1516 | uint64_t object, txg; | |
1517 | ||
1518 | if (byteswap) | |
1519 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1520 | ||
1521 | ASSERT(lr->lr_doid == ZTEST_DIROBJ); | |
1522 | ASSERT(name[0] != '\0'); | |
1523 | ||
1524 | VERIFY3U(0, ==, | |
1525 | zap_lookup(os, lr->lr_doid, name, sizeof (object), 1, &object)); | |
1526 | ASSERT(object != 0); | |
1527 | ||
1528 | ztest_object_lock(zd, object, RL_WRITER); | |
1529 | ||
1530 | VERIFY3U(0, ==, dmu_object_info(os, object, &doi)); | |
1531 | ||
1532 | tx = dmu_tx_create(os); | |
1533 | ||
1534 | dmu_tx_hold_zap(tx, lr->lr_doid, B_FALSE, name); | |
1535 | dmu_tx_hold_free(tx, object, 0, DMU_OBJECT_END); | |
1536 | ||
1537 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1538 | if (txg == 0) { | |
1539 | ztest_object_unlock(zd, object); | |
1540 | return (ENOSPC); | |
1541 | } | |
1542 | ||
1543 | if (doi.doi_type == DMU_OT_ZAP_OTHER) { | |
1544 | VERIFY3U(0, ==, zap_destroy(os, object, tx)); | |
1545 | } else { | |
1546 | VERIFY3U(0, ==, dmu_object_free(os, object, tx)); | |
1547 | } | |
1548 | ||
1549 | VERIFY3U(0, ==, zap_remove(os, lr->lr_doid, name, tx)); | |
1550 | ||
572e2857 | 1551 | (void) ztest_log_remove(zd, tx, lr, object); |
428870ff BB |
1552 | |
1553 | dmu_tx_commit(tx); | |
1554 | ||
1555 | ztest_object_unlock(zd, object); | |
1556 | ||
1557 | return (0); | |
1558 | } | |
1559 | ||
1560 | static int | |
1561 | ztest_replay_write(ztest_ds_t *zd, lr_write_t *lr, boolean_t byteswap) | |
1562 | { | |
1563 | objset_t *os = zd->zd_os; | |
1564 | void *data = lr + 1; /* data follows lr */ | |
1565 | uint64_t offset, length; | |
1566 | ztest_block_tag_t *bt = data; | |
1567 | ztest_block_tag_t *bbt; | |
1568 | uint64_t gen, txg, lrtxg, crtxg; | |
1569 | dmu_object_info_t doi; | |
1570 | dmu_tx_t *tx; | |
1571 | dmu_buf_t *db; | |
1572 | arc_buf_t *abuf = NULL; | |
1573 | rl_t *rl; | |
1574 | ||
1575 | if (byteswap) | |
1576 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1577 | ||
1578 | offset = lr->lr_offset; | |
1579 | length = lr->lr_length; | |
1580 | ||
1581 | /* If it's a dmu_sync() block, write the whole block */ | |
1582 | if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) { | |
1583 | uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr); | |
1584 | if (length < blocksize) { | |
1585 | offset -= offset % blocksize; | |
1586 | length = blocksize; | |
1587 | } | |
1588 | } | |
1589 | ||
1590 | if (bt->bt_magic == BSWAP_64(BT_MAGIC)) | |
1591 | byteswap_uint64_array(bt, sizeof (*bt)); | |
1592 | ||
1593 | if (bt->bt_magic != BT_MAGIC) | |
1594 | bt = NULL; | |
1595 | ||
1596 | ztest_object_lock(zd, lr->lr_foid, RL_READER); | |
1597 | rl = ztest_range_lock(zd, lr->lr_foid, offset, length, RL_WRITER); | |
1598 | ||
1599 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
1600 | ||
1601 | dmu_object_info_from_db(db, &doi); | |
1602 | ||
1603 | bbt = ztest_bt_bonus(db); | |
1604 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
1605 | gen = bbt->bt_gen; | |
1606 | crtxg = bbt->bt_crtxg; | |
1607 | lrtxg = lr->lr_common.lrc_txg; | |
1608 | ||
1609 | tx = dmu_tx_create(os); | |
1610 | ||
1611 | dmu_tx_hold_write(tx, lr->lr_foid, offset, length); | |
1612 | ||
1613 | if (ztest_random(8) == 0 && length == doi.doi_data_block_size && | |
1614 | P2PHASE(offset, length) == 0) | |
1615 | abuf = dmu_request_arcbuf(db, length); | |
1616 | ||
1617 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1618 | if (txg == 0) { | |
1619 | if (abuf != NULL) | |
1620 | dmu_return_arcbuf(abuf); | |
1621 | dmu_buf_rele(db, FTAG); | |
1622 | ztest_range_unlock(rl); | |
1623 | ztest_object_unlock(zd, lr->lr_foid); | |
1624 | return (ENOSPC); | |
1625 | } | |
1626 | ||
1627 | if (bt != NULL) { | |
1628 | /* | |
1629 | * Usually, verify the old data before writing new data -- | |
1630 | * but not always, because we also want to verify correct | |
1631 | * behavior when the data was not recently read into cache. | |
1632 | */ | |
1633 | ASSERT(offset % doi.doi_data_block_size == 0); | |
1634 | if (ztest_random(4) != 0) { | |
1635 | int prefetch = ztest_random(2) ? | |
1636 | DMU_READ_PREFETCH : DMU_READ_NO_PREFETCH; | |
1637 | ztest_block_tag_t rbt; | |
1638 | ||
1639 | VERIFY(dmu_read(os, lr->lr_foid, offset, | |
1640 | sizeof (rbt), &rbt, prefetch) == 0); | |
1641 | if (rbt.bt_magic == BT_MAGIC) { | |
1642 | ztest_bt_verify(&rbt, os, lr->lr_foid, | |
1643 | offset, gen, txg, crtxg); | |
1644 | } | |
1645 | } | |
1646 | ||
1647 | /* | |
1648 | * Writes can appear to be newer than the bonus buffer because | |
1649 | * the ztest_get_data() callback does a dmu_read() of the | |
1650 | * open-context data, which may be different than the data | |
1651 | * as it was when the write was generated. | |
1652 | */ | |
1653 | if (zd->zd_zilog->zl_replay) { | |
1654 | ztest_bt_verify(bt, os, lr->lr_foid, offset, | |
1655 | MAX(gen, bt->bt_gen), MAX(txg, lrtxg), | |
1656 | bt->bt_crtxg); | |
1657 | } | |
1658 | ||
1659 | /* | |
1660 | * Set the bt's gen/txg to the bonus buffer's gen/txg | |
1661 | * so that all of the usual ASSERTs will work. | |
1662 | */ | |
1663 | ztest_bt_generate(bt, os, lr->lr_foid, offset, gen, txg, crtxg); | |
1664 | } | |
1665 | ||
1666 | if (abuf == NULL) { | |
1667 | dmu_write(os, lr->lr_foid, offset, length, data, tx); | |
1668 | } else { | |
1669 | bcopy(data, abuf->b_data, length); | |
1670 | dmu_assign_arcbuf(db, offset, abuf, tx); | |
1671 | } | |
1672 | ||
1673 | (void) ztest_log_write(zd, tx, lr); | |
1674 | ||
1675 | dmu_buf_rele(db, FTAG); | |
1676 | ||
1677 | dmu_tx_commit(tx); | |
1678 | ||
1679 | ztest_range_unlock(rl); | |
1680 | ztest_object_unlock(zd, lr->lr_foid); | |
1681 | ||
1682 | return (0); | |
1683 | } | |
1684 | ||
1685 | static int | |
1686 | ztest_replay_truncate(ztest_ds_t *zd, lr_truncate_t *lr, boolean_t byteswap) | |
1687 | { | |
1688 | objset_t *os = zd->zd_os; | |
1689 | dmu_tx_t *tx; | |
1690 | uint64_t txg; | |
1691 | rl_t *rl; | |
1692 | ||
1693 | if (byteswap) | |
1694 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1695 | ||
1696 | ztest_object_lock(zd, lr->lr_foid, RL_READER); | |
1697 | rl = ztest_range_lock(zd, lr->lr_foid, lr->lr_offset, lr->lr_length, | |
1698 | RL_WRITER); | |
1699 | ||
1700 | tx = dmu_tx_create(os); | |
1701 | ||
1702 | dmu_tx_hold_free(tx, lr->lr_foid, lr->lr_offset, lr->lr_length); | |
1703 | ||
1704 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1705 | if (txg == 0) { | |
1706 | ztest_range_unlock(rl); | |
1707 | ztest_object_unlock(zd, lr->lr_foid); | |
1708 | return (ENOSPC); | |
1709 | } | |
1710 | ||
1711 | VERIFY(dmu_free_range(os, lr->lr_foid, lr->lr_offset, | |
1712 | lr->lr_length, tx) == 0); | |
1713 | ||
1714 | (void) ztest_log_truncate(zd, tx, lr); | |
1715 | ||
1716 | dmu_tx_commit(tx); | |
1717 | ||
1718 | ztest_range_unlock(rl); | |
1719 | ztest_object_unlock(zd, lr->lr_foid); | |
1720 | ||
1721 | return (0); | |
1722 | } | |
1723 | ||
1724 | static int | |
1725 | ztest_replay_setattr(ztest_ds_t *zd, lr_setattr_t *lr, boolean_t byteswap) | |
1726 | { | |
1727 | objset_t *os = zd->zd_os; | |
1728 | dmu_tx_t *tx; | |
1729 | dmu_buf_t *db; | |
1730 | ztest_block_tag_t *bbt; | |
1731 | uint64_t txg, lrtxg, crtxg; | |
1732 | ||
1733 | if (byteswap) | |
1734 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1735 | ||
1736 | ztest_object_lock(zd, lr->lr_foid, RL_WRITER); | |
1737 | ||
1738 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
1739 | ||
1740 | tx = dmu_tx_create(os); | |
1741 | dmu_tx_hold_bonus(tx, lr->lr_foid); | |
1742 | ||
1743 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1744 | if (txg == 0) { | |
1745 | dmu_buf_rele(db, FTAG); | |
1746 | ztest_object_unlock(zd, lr->lr_foid); | |
1747 | return (ENOSPC); | |
1748 | } | |
1749 | ||
1750 | bbt = ztest_bt_bonus(db); | |
1751 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
1752 | crtxg = bbt->bt_crtxg; | |
1753 | lrtxg = lr->lr_common.lrc_txg; | |
1754 | ||
1755 | if (zd->zd_zilog->zl_replay) { | |
1756 | ASSERT(lr->lr_size != 0); | |
1757 | ASSERT(lr->lr_mode != 0); | |
1758 | ASSERT(lrtxg != 0); | |
1759 | } else { | |
1760 | /* | |
1761 | * Randomly change the size and increment the generation. | |
1762 | */ | |
1763 | lr->lr_size = (ztest_random(db->db_size / sizeof (*bbt)) + 1) * | |
1764 | sizeof (*bbt); | |
1765 | lr->lr_mode = bbt->bt_gen + 1; | |
1766 | ASSERT(lrtxg == 0); | |
1767 | } | |
1768 | ||
1769 | /* | |
1770 | * Verify that the current bonus buffer is not newer than our txg. | |
1771 | */ | |
1772 | ztest_bt_verify(bbt, os, lr->lr_foid, -1ULL, lr->lr_mode, | |
1773 | MAX(txg, lrtxg), crtxg); | |
1774 | ||
1775 | dmu_buf_will_dirty(db, tx); | |
1776 | ||
1777 | ASSERT3U(lr->lr_size, >=, sizeof (*bbt)); | |
1778 | ASSERT3U(lr->lr_size, <=, db->db_size); | |
c99c9001 | 1779 | VERIFY0(dmu_set_bonus(db, lr->lr_size, tx)); |
428870ff BB |
1780 | bbt = ztest_bt_bonus(db); |
1781 | ||
1782 | ztest_bt_generate(bbt, os, lr->lr_foid, -1ULL, lr->lr_mode, txg, crtxg); | |
1783 | ||
1784 | dmu_buf_rele(db, FTAG); | |
1785 | ||
1786 | (void) ztest_log_setattr(zd, tx, lr); | |
1787 | ||
1788 | dmu_tx_commit(tx); | |
1789 | ||
1790 | ztest_object_unlock(zd, lr->lr_foid); | |
1791 | ||
1792 | return (0); | |
1793 | } | |
1794 | ||
b01615d5 | 1795 | zil_replay_func_t ztest_replay_vector[TX_MAX_TYPE] = { |
0bc8fd78 | 1796 | NULL, /* 0 no such transaction type */ |
b01615d5 | 1797 | (zil_replay_func_t)ztest_replay_create, /* TX_CREATE */ |
0bc8fd78 BB |
1798 | NULL, /* TX_MKDIR */ |
1799 | NULL, /* TX_MKXATTR */ | |
1800 | NULL, /* TX_SYMLINK */ | |
b01615d5 | 1801 | (zil_replay_func_t)ztest_replay_remove, /* TX_REMOVE */ |
0bc8fd78 BB |
1802 | NULL, /* TX_RMDIR */ |
1803 | NULL, /* TX_LINK */ | |
1804 | NULL, /* TX_RENAME */ | |
b01615d5 RY |
1805 | (zil_replay_func_t)ztest_replay_write, /* TX_WRITE */ |
1806 | (zil_replay_func_t)ztest_replay_truncate, /* TX_TRUNCATE */ | |
1807 | (zil_replay_func_t)ztest_replay_setattr, /* TX_SETATTR */ | |
0bc8fd78 BB |
1808 | NULL, /* TX_ACL */ |
1809 | NULL, /* TX_CREATE_ACL */ | |
1810 | NULL, /* TX_CREATE_ATTR */ | |
1811 | NULL, /* TX_CREATE_ACL_ATTR */ | |
1812 | NULL, /* TX_MKDIR_ACL */ | |
1813 | NULL, /* TX_MKDIR_ATTR */ | |
1814 | NULL, /* TX_MKDIR_ACL_ATTR */ | |
1815 | NULL, /* TX_WRITE2 */ | |
428870ff BB |
1816 | }; |
1817 | ||
1818 | /* | |
1819 | * ZIL get_data callbacks | |
1820 | */ | |
1821 | ||
1822 | static void | |
1823 | ztest_get_done(zgd_t *zgd, int error) | |
1824 | { | |
1825 | ztest_ds_t *zd = zgd->zgd_private; | |
1826 | uint64_t object = zgd->zgd_rl->rl_object; | |
1827 | ||
1828 | if (zgd->zgd_db) | |
1829 | dmu_buf_rele(zgd->zgd_db, zgd); | |
1830 | ||
1831 | ztest_range_unlock(zgd->zgd_rl); | |
1832 | ztest_object_unlock(zd, object); | |
1833 | ||
1834 | if (error == 0 && zgd->zgd_bp) | |
1835 | zil_add_block(zgd->zgd_zilog, zgd->zgd_bp); | |
1836 | ||
1837 | umem_free(zgd, sizeof (*zgd)); | |
1838 | } | |
1839 | ||
1840 | static int | |
1841 | ztest_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio) | |
1842 | { | |
1843 | ztest_ds_t *zd = arg; | |
1844 | objset_t *os = zd->zd_os; | |
1845 | uint64_t object = lr->lr_foid; | |
1846 | uint64_t offset = lr->lr_offset; | |
1847 | uint64_t size = lr->lr_length; | |
1848 | blkptr_t *bp = &lr->lr_blkptr; | |
1849 | uint64_t txg = lr->lr_common.lrc_txg; | |
1850 | uint64_t crtxg; | |
1851 | dmu_object_info_t doi; | |
1852 | dmu_buf_t *db; | |
1853 | zgd_t *zgd; | |
1854 | int error; | |
1855 | ||
1856 | ztest_object_lock(zd, object, RL_READER); | |
1857 | error = dmu_bonus_hold(os, object, FTAG, &db); | |
1858 | if (error) { | |
1859 | ztest_object_unlock(zd, object); | |
1860 | return (error); | |
1861 | } | |
1862 | ||
1863 | crtxg = ztest_bt_bonus(db)->bt_crtxg; | |
1864 | ||
1865 | if (crtxg == 0 || crtxg > txg) { | |
1866 | dmu_buf_rele(db, FTAG); | |
1867 | ztest_object_unlock(zd, object); | |
1868 | return (ENOENT); | |
1869 | } | |
1870 | ||
1871 | dmu_object_info_from_db(db, &doi); | |
1872 | dmu_buf_rele(db, FTAG); | |
1873 | db = NULL; | |
1874 | ||
1875 | zgd = umem_zalloc(sizeof (*zgd), UMEM_NOFAIL); | |
1876 | zgd->zgd_zilog = zd->zd_zilog; | |
1877 | zgd->zgd_private = zd; | |
1878 | ||
1879 | if (buf != NULL) { /* immediate write */ | |
1880 | zgd->zgd_rl = ztest_range_lock(zd, object, offset, size, | |
1881 | RL_READER); | |
1882 | ||
1883 | error = dmu_read(os, object, offset, size, buf, | |
1884 | DMU_READ_NO_PREFETCH); | |
1885 | ASSERT(error == 0); | |
1886 | } else { | |
1887 | size = doi.doi_data_block_size; | |
1888 | if (ISP2(size)) { | |
1889 | offset = P2ALIGN(offset, size); | |
1890 | } else { | |
1891 | ASSERT(offset < size); | |
1892 | offset = 0; | |
1893 | } | |
1894 | ||
1895 | zgd->zgd_rl = ztest_range_lock(zd, object, offset, size, | |
1896 | RL_READER); | |
1897 | ||
1898 | error = dmu_buf_hold(os, object, offset, zgd, &db, | |
1899 | DMU_READ_NO_PREFETCH); | |
1900 | ||
1901 | if (error == 0) { | |
1902 | zgd->zgd_db = db; | |
1903 | zgd->zgd_bp = bp; | |
1904 | ||
1905 | ASSERT(db->db_offset == offset); | |
1906 | ASSERT(db->db_size == size); | |
1907 | ||
1908 | error = dmu_sync(zio, lr->lr_common.lrc_txg, | |
1909 | ztest_get_done, zgd); | |
1910 | ||
1911 | if (error == 0) | |
1912 | return (0); | |
1913 | } | |
1914 | } | |
1915 | ||
1916 | ztest_get_done(zgd, error); | |
1917 | ||
1918 | return (error); | |
1919 | } | |
1920 | ||
1921 | static void * | |
1922 | ztest_lr_alloc(size_t lrsize, char *name) | |
1923 | { | |
1924 | char *lr; | |
1925 | size_t namesize = name ? strlen(name) + 1 : 0; | |
1926 | ||
1927 | lr = umem_zalloc(lrsize + namesize, UMEM_NOFAIL); | |
1928 | ||
1929 | if (name) | |
1930 | bcopy(name, lr + lrsize, namesize); | |
1931 | ||
1932 | return (lr); | |
1933 | } | |
1934 | ||
1935 | void | |
1936 | ztest_lr_free(void *lr, size_t lrsize, char *name) | |
1937 | { | |
1938 | size_t namesize = name ? strlen(name) + 1 : 0; | |
1939 | ||
1940 | umem_free(lr, lrsize + namesize); | |
1941 | } | |
1942 | ||
1943 | /* | |
1944 | * Lookup a bunch of objects. Returns the number of objects not found. | |
1945 | */ | |
1946 | static int | |
1947 | ztest_lookup(ztest_ds_t *zd, ztest_od_t *od, int count) | |
1948 | { | |
1949 | int missing = 0; | |
1950 | int error; | |
d6320ddb | 1951 | int i; |
428870ff | 1952 | |
1e33ac1e | 1953 | ASSERT(mutex_held(&zd->zd_dirobj_lock)); |
428870ff | 1954 | |
d6320ddb | 1955 | for (i = 0; i < count; i++, od++) { |
428870ff BB |
1956 | od->od_object = 0; |
1957 | error = zap_lookup(zd->zd_os, od->od_dir, od->od_name, | |
1958 | sizeof (uint64_t), 1, &od->od_object); | |
1959 | if (error) { | |
1960 | ASSERT(error == ENOENT); | |
1961 | ASSERT(od->od_object == 0); | |
1962 | missing++; | |
1963 | } else { | |
1964 | dmu_buf_t *db; | |
1965 | ztest_block_tag_t *bbt; | |
1966 | dmu_object_info_t doi; | |
1967 | ||
1968 | ASSERT(od->od_object != 0); | |
1969 | ASSERT(missing == 0); /* there should be no gaps */ | |
1970 | ||
1971 | ztest_object_lock(zd, od->od_object, RL_READER); | |
1972 | VERIFY3U(0, ==, dmu_bonus_hold(zd->zd_os, | |
1973 | od->od_object, FTAG, &db)); | |
1974 | dmu_object_info_from_db(db, &doi); | |
1975 | bbt = ztest_bt_bonus(db); | |
1976 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
1977 | od->od_type = doi.doi_type; | |
1978 | od->od_blocksize = doi.doi_data_block_size; | |
1979 | od->od_gen = bbt->bt_gen; | |
1980 | dmu_buf_rele(db, FTAG); | |
1981 | ztest_object_unlock(zd, od->od_object); | |
1982 | } | |
1983 | } | |
1984 | ||
1985 | return (missing); | |
1986 | } | |
1987 | ||
1988 | static int | |
1989 | ztest_create(ztest_ds_t *zd, ztest_od_t *od, int count) | |
1990 | { | |
1991 | int missing = 0; | |
d6320ddb | 1992 | int i; |
428870ff | 1993 | |
1e33ac1e | 1994 | ASSERT(mutex_held(&zd->zd_dirobj_lock)); |
428870ff | 1995 | |
d6320ddb | 1996 | for (i = 0; i < count; i++, od++) { |
428870ff BB |
1997 | if (missing) { |
1998 | od->od_object = 0; | |
1999 | missing++; | |
2000 | continue; | |
2001 | } | |
2002 | ||
2003 | lr_create_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name); | |
2004 | ||
2005 | lr->lr_doid = od->od_dir; | |
2006 | lr->lr_foid = 0; /* 0 to allocate, > 0 to claim */ | |
2007 | lr->lrz_type = od->od_crtype; | |
2008 | lr->lrz_blocksize = od->od_crblocksize; | |
2009 | lr->lrz_ibshift = ztest_random_ibshift(); | |
2010 | lr->lrz_bonustype = DMU_OT_UINT64_OTHER; | |
2011 | lr->lrz_bonuslen = dmu_bonus_max(); | |
2012 | lr->lr_gen = od->od_crgen; | |
2013 | lr->lr_crtime[0] = time(NULL); | |
2014 | ||
2015 | if (ztest_replay_create(zd, lr, B_FALSE) != 0) { | |
2016 | ASSERT(missing == 0); | |
2017 | od->od_object = 0; | |
2018 | missing++; | |
2019 | } else { | |
2020 | od->od_object = lr->lr_foid; | |
2021 | od->od_type = od->od_crtype; | |
2022 | od->od_blocksize = od->od_crblocksize; | |
2023 | od->od_gen = od->od_crgen; | |
2024 | ASSERT(od->od_object != 0); | |
2025 | } | |
2026 | ||
2027 | ztest_lr_free(lr, sizeof (*lr), od->od_name); | |
2028 | } | |
2029 | ||
2030 | return (missing); | |
2031 | } | |
2032 | ||
2033 | static int | |
2034 | ztest_remove(ztest_ds_t *zd, ztest_od_t *od, int count) | |
2035 | { | |
2036 | int missing = 0; | |
2037 | int error; | |
d6320ddb | 2038 | int i; |
428870ff | 2039 | |
1e33ac1e | 2040 | ASSERT(mutex_held(&zd->zd_dirobj_lock)); |
428870ff BB |
2041 | |
2042 | od += count - 1; | |
2043 | ||
d6320ddb | 2044 | for (i = count - 1; i >= 0; i--, od--) { |
428870ff BB |
2045 | if (missing) { |
2046 | missing++; | |
2047 | continue; | |
2048 | } | |
2049 | ||
2050 | if (od->od_object == 0) | |
2051 | continue; | |
2052 | ||
2053 | lr_remove_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name); | |
2054 | ||
2055 | lr->lr_doid = od->od_dir; | |
2056 | ||
2057 | if ((error = ztest_replay_remove(zd, lr, B_FALSE)) != 0) { | |
2058 | ASSERT3U(error, ==, ENOSPC); | |
2059 | missing++; | |
2060 | } else { | |
2061 | od->od_object = 0; | |
2062 | } | |
2063 | ztest_lr_free(lr, sizeof (*lr), od->od_name); | |
2064 | } | |
2065 | ||
2066 | return (missing); | |
2067 | } | |
2068 | ||
2069 | static int | |
2070 | ztest_write(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size, | |
2071 | void *data) | |
2072 | { | |
2073 | lr_write_t *lr; | |
2074 | int error; | |
2075 | ||
2076 | lr = ztest_lr_alloc(sizeof (*lr) + size, NULL); | |
2077 | ||
2078 | lr->lr_foid = object; | |
2079 | lr->lr_offset = offset; | |
2080 | lr->lr_length = size; | |
2081 | lr->lr_blkoff = 0; | |
2082 | BP_ZERO(&lr->lr_blkptr); | |
2083 | ||
2084 | bcopy(data, lr + 1, size); | |
2085 | ||
2086 | error = ztest_replay_write(zd, lr, B_FALSE); | |
2087 | ||
2088 | ztest_lr_free(lr, sizeof (*lr) + size, NULL); | |
2089 | ||
2090 | return (error); | |
2091 | } | |
2092 | ||
2093 | static int | |
2094 | ztest_truncate(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size) | |
2095 | { | |
2096 | lr_truncate_t *lr; | |
2097 | int error; | |
2098 | ||
2099 | lr = ztest_lr_alloc(sizeof (*lr), NULL); | |
2100 | ||
2101 | lr->lr_foid = object; | |
2102 | lr->lr_offset = offset; | |
2103 | lr->lr_length = size; | |
2104 | ||
2105 | error = ztest_replay_truncate(zd, lr, B_FALSE); | |
2106 | ||
2107 | ztest_lr_free(lr, sizeof (*lr), NULL); | |
2108 | ||
2109 | return (error); | |
2110 | } | |
2111 | ||
2112 | static int | |
2113 | ztest_setattr(ztest_ds_t *zd, uint64_t object) | |
2114 | { | |
2115 | lr_setattr_t *lr; | |
2116 | int error; | |
2117 | ||
2118 | lr = ztest_lr_alloc(sizeof (*lr), NULL); | |
2119 | ||
2120 | lr->lr_foid = object; | |
2121 | lr->lr_size = 0; | |
2122 | lr->lr_mode = 0; | |
2123 | ||
2124 | error = ztest_replay_setattr(zd, lr, B_FALSE); | |
2125 | ||
2126 | ztest_lr_free(lr, sizeof (*lr), NULL); | |
2127 | ||
2128 | return (error); | |
2129 | } | |
2130 | ||
2131 | static void | |
2132 | ztest_prealloc(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size) | |
2133 | { | |
2134 | objset_t *os = zd->zd_os; | |
2135 | dmu_tx_t *tx; | |
2136 | uint64_t txg; | |
2137 | rl_t *rl; | |
2138 | ||
2139 | txg_wait_synced(dmu_objset_pool(os), 0); | |
2140 | ||
2141 | ztest_object_lock(zd, object, RL_READER); | |
2142 | rl = ztest_range_lock(zd, object, offset, size, RL_WRITER); | |
2143 | ||
2144 | tx = dmu_tx_create(os); | |
2145 | ||
2146 | dmu_tx_hold_write(tx, object, offset, size); | |
2147 | ||
2148 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
2149 | ||
2150 | if (txg != 0) { | |
2151 | dmu_prealloc(os, object, offset, size, tx); | |
2152 | dmu_tx_commit(tx); | |
2153 | txg_wait_synced(dmu_objset_pool(os), txg); | |
2154 | } else { | |
2155 | (void) dmu_free_long_range(os, object, offset, size); | |
2156 | } | |
2157 | ||
2158 | ztest_range_unlock(rl); | |
2159 | ztest_object_unlock(zd, object); | |
2160 | } | |
2161 | ||
2162 | static void | |
2163 | ztest_io(ztest_ds_t *zd, uint64_t object, uint64_t offset) | |
2164 | { | |
2165 | ztest_block_tag_t wbt; | |
2166 | dmu_object_info_t doi; | |
2167 | enum ztest_io_type io_type; | |
2168 | uint64_t blocksize; | |
2169 | void *data; | |
2170 | ||
2171 | VERIFY(dmu_object_info(zd->zd_os, object, &doi) == 0); | |
2172 | blocksize = doi.doi_data_block_size; | |
2173 | data = umem_alloc(blocksize, UMEM_NOFAIL); | |
2174 | ||
2175 | /* | |
2176 | * Pick an i/o type at random, biased toward writing block tags. | |
2177 | */ | |
2178 | io_type = ztest_random(ZTEST_IO_TYPES); | |
2179 | if (ztest_random(2) == 0) | |
2180 | io_type = ZTEST_IO_WRITE_TAG; | |
2181 | ||
3e31d2b0 ES |
2182 | (void) rw_enter(&zd->zd_zilog_lock, RW_READER); |
2183 | ||
428870ff BB |
2184 | switch (io_type) { |
2185 | ||
2186 | case ZTEST_IO_WRITE_TAG: | |
2187 | ztest_bt_generate(&wbt, zd->zd_os, object, offset, 0, 0, 0); | |
2188 | (void) ztest_write(zd, object, offset, sizeof (wbt), &wbt); | |
2189 | break; | |
2190 | ||
2191 | case ZTEST_IO_WRITE_PATTERN: | |
2192 | (void) memset(data, 'a' + (object + offset) % 5, blocksize); | |
2193 | if (ztest_random(2) == 0) { | |
2194 | /* | |
2195 | * Induce fletcher2 collisions to ensure that | |
2196 | * zio_ddt_collision() detects and resolves them | |
2197 | * when using fletcher2-verify for deduplication. | |
2198 | */ | |
2199 | ((uint64_t *)data)[0] ^= 1ULL << 63; | |
2200 | ((uint64_t *)data)[4] ^= 1ULL << 63; | |
2201 | } | |
2202 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2203 | break; | |
2204 | ||
2205 | case ZTEST_IO_WRITE_ZEROES: | |
2206 | bzero(data, blocksize); | |
2207 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2208 | break; | |
2209 | ||
2210 | case ZTEST_IO_TRUNCATE: | |
2211 | (void) ztest_truncate(zd, object, offset, blocksize); | |
2212 | break; | |
2213 | ||
2214 | case ZTEST_IO_SETATTR: | |
2215 | (void) ztest_setattr(zd, object); | |
2216 | break; | |
e75c13c3 BB |
2217 | default: |
2218 | break; | |
428870ff BB |
2219 | } |
2220 | ||
3e31d2b0 ES |
2221 | (void) rw_exit(&zd->zd_zilog_lock); |
2222 | ||
428870ff BB |
2223 | umem_free(data, blocksize); |
2224 | } | |
2225 | ||
2226 | /* | |
2227 | * Initialize an object description template. | |
2228 | */ | |
2229 | static void | |
2230 | ztest_od_init(ztest_od_t *od, uint64_t id, char *tag, uint64_t index, | |
2231 | dmu_object_type_t type, uint64_t blocksize, uint64_t gen) | |
2232 | { | |
2233 | od->od_dir = ZTEST_DIROBJ; | |
2234 | od->od_object = 0; | |
2235 | ||
2236 | od->od_crtype = type; | |
2237 | od->od_crblocksize = blocksize ? blocksize : ztest_random_blocksize(); | |
2238 | od->od_crgen = gen; | |
2239 | ||
2240 | od->od_type = DMU_OT_NONE; | |
2241 | od->od_blocksize = 0; | |
2242 | od->od_gen = 0; | |
2243 | ||
2244 | (void) snprintf(od->od_name, sizeof (od->od_name), "%s(%lld)[%llu]", | |
b8864a23 | 2245 | tag, (longlong_t)id, (u_longlong_t)index); |
428870ff BB |
2246 | } |
2247 | ||
2248 | /* | |
2249 | * Lookup or create the objects for a test using the od template. | |
2250 | * If the objects do not all exist, or if 'remove' is specified, | |
2251 | * remove any existing objects and create new ones. Otherwise, | |
2252 | * use the existing objects. | |
2253 | */ | |
2254 | static int | |
2255 | ztest_object_init(ztest_ds_t *zd, ztest_od_t *od, size_t size, boolean_t remove) | |
2256 | { | |
2257 | int count = size / sizeof (*od); | |
2258 | int rv = 0; | |
2259 | ||
1e33ac1e | 2260 | mutex_enter(&zd->zd_dirobj_lock); |
428870ff BB |
2261 | if ((ztest_lookup(zd, od, count) != 0 || remove) && |
2262 | (ztest_remove(zd, od, count) != 0 || | |
2263 | ztest_create(zd, od, count) != 0)) | |
2264 | rv = -1; | |
2265 | zd->zd_od = od; | |
1e33ac1e | 2266 | mutex_exit(&zd->zd_dirobj_lock); |
428870ff BB |
2267 | |
2268 | return (rv); | |
2269 | } | |
2270 | ||
2271 | /* ARGSUSED */ | |
2272 | void | |
2273 | ztest_zil_commit(ztest_ds_t *zd, uint64_t id) | |
2274 | { | |
2275 | zilog_t *zilog = zd->zd_zilog; | |
2276 | ||
3e31d2b0 ES |
2277 | (void) rw_enter(&zd->zd_zilog_lock, RW_READER); |
2278 | ||
572e2857 | 2279 | zil_commit(zilog, ztest_random(ZTEST_OBJECTS)); |
428870ff BB |
2280 | |
2281 | /* | |
2282 | * Remember the committed values in zd, which is in parent/child | |
2283 | * shared memory. If we die, the next iteration of ztest_run() | |
2284 | * will verify that the log really does contain this record. | |
2285 | */ | |
2286 | mutex_enter(&zilog->zl_lock); | |
c242c188 CS |
2287 | ASSERT(zd->zd_shared != NULL); |
2288 | ASSERT3U(zd->zd_shared->zd_seq, <=, zilog->zl_commit_lr_seq); | |
2289 | zd->zd_shared->zd_seq = zilog->zl_commit_lr_seq; | |
428870ff | 2290 | mutex_exit(&zilog->zl_lock); |
3e31d2b0 ES |
2291 | |
2292 | (void) rw_exit(&zd->zd_zilog_lock); | |
2293 | } | |
2294 | ||
2295 | /* | |
2296 | * This function is designed to simulate the operations that occur during a | |
2297 | * mount/unmount operation. We hold the dataset across these operations in an | |
2298 | * attempt to expose any implicit assumptions about ZIL management. | |
2299 | */ | |
2300 | /* ARGSUSED */ | |
2301 | void | |
2302 | ztest_zil_remount(ztest_ds_t *zd, uint64_t id) | |
2303 | { | |
2304 | objset_t *os = zd->zd_os; | |
2305 | ||
29809a6c | 2306 | mutex_enter(&zd->zd_dirobj_lock); |
3e31d2b0 ES |
2307 | (void) rw_enter(&zd->zd_zilog_lock, RW_WRITER); |
2308 | ||
86f35f34 | 2309 | /* zfs_sb_teardown() */ |
3e31d2b0 ES |
2310 | zil_close(zd->zd_zilog); |
2311 | ||
2312 | /* zfsvfs_setup() */ | |
2313 | VERIFY(zil_open(os, ztest_get_data) == zd->zd_zilog); | |
2314 | zil_replay(os, zd, ztest_replay_vector); | |
2315 | ||
2316 | (void) rw_exit(&zd->zd_zilog_lock); | |
29809a6c | 2317 | mutex_exit(&zd->zd_dirobj_lock); |
428870ff BB |
2318 | } |
2319 | ||
2320 | /* | |
2321 | * Verify that we can't destroy an active pool, create an existing pool, | |
2322 | * or create a pool with a bad vdev spec. | |
2323 | */ | |
2324 | /* ARGSUSED */ | |
2325 | void | |
2326 | ztest_spa_create_destroy(ztest_ds_t *zd, uint64_t id) | |
2327 | { | |
c242c188 | 2328 | ztest_shared_opts_t *zo = &ztest_opts; |
428870ff BB |
2329 | spa_t *spa; |
2330 | nvlist_t *nvroot; | |
2331 | ||
2332 | /* | |
2333 | * Attempt to create using a bad file. | |
2334 | */ | |
ea0b2538 | 2335 | nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, 0, 0, 0, 1); |
428870ff BB |
2336 | VERIFY3U(ENOENT, ==, |
2337 | spa_create("ztest_bad_file", nvroot, NULL, NULL, NULL)); | |
2338 | nvlist_free(nvroot); | |
2339 | ||
2340 | /* | |
2341 | * Attempt to create using a bad mirror. | |
2342 | */ | |
ea0b2538 | 2343 | nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, 0, 0, 2, 1); |
428870ff BB |
2344 | VERIFY3U(ENOENT, ==, |
2345 | spa_create("ztest_bad_mirror", nvroot, NULL, NULL, NULL)); | |
2346 | nvlist_free(nvroot); | |
2347 | ||
2348 | /* | |
2349 | * Attempt to create an existing pool. It shouldn't matter | |
2350 | * what's in the nvroot; we should fail with EEXIST. | |
2351 | */ | |
c242c188 | 2352 | (void) rw_enter(&ztest_name_lock, RW_READER); |
ea0b2538 | 2353 | nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, 0, 0, 0, 1); |
c242c188 | 2354 | VERIFY3U(EEXIST, ==, spa_create(zo->zo_pool, nvroot, NULL, NULL, NULL)); |
428870ff | 2355 | nvlist_free(nvroot); |
c242c188 CS |
2356 | VERIFY3U(0, ==, spa_open(zo->zo_pool, &spa, FTAG)); |
2357 | VERIFY3U(EBUSY, ==, spa_destroy(zo->zo_pool)); | |
428870ff BB |
2358 | spa_close(spa, FTAG); |
2359 | ||
c242c188 | 2360 | (void) rw_exit(&ztest_name_lock); |
428870ff BB |
2361 | } |
2362 | ||
ea0b2538 GW |
2363 | /* ARGSUSED */ |
2364 | void | |
2365 | ztest_spa_upgrade(ztest_ds_t *zd, uint64_t id) | |
2366 | { | |
2367 | spa_t *spa; | |
2368 | uint64_t initial_version = SPA_VERSION_INITIAL; | |
2369 | uint64_t version, newversion; | |
2370 | nvlist_t *nvroot, *props; | |
2371 | char *name; | |
2372 | ||
2373 | mutex_enter(&ztest_vdev_lock); | |
2374 | name = kmem_asprintf("%s_upgrade", ztest_opts.zo_pool); | |
2375 | ||
2376 | /* | |
2377 | * Clean up from previous runs. | |
2378 | */ | |
2379 | (void) spa_destroy(name); | |
2380 | ||
2381 | nvroot = make_vdev_root(NULL, NULL, name, ztest_opts.zo_vdev_size, 0, | |
2382 | 0, ztest_opts.zo_raidz, ztest_opts.zo_mirrors, 1); | |
2383 | ||
2384 | /* | |
2385 | * If we're configuring a RAIDZ device then make sure that the | |
2386 | * the initial version is capable of supporting that feature. | |
2387 | */ | |
2388 | switch (ztest_opts.zo_raidz_parity) { | |
2389 | case 0: | |
2390 | case 1: | |
2391 | initial_version = SPA_VERSION_INITIAL; | |
2392 | break; | |
2393 | case 2: | |
2394 | initial_version = SPA_VERSION_RAIDZ2; | |
2395 | break; | |
2396 | case 3: | |
2397 | initial_version = SPA_VERSION_RAIDZ3; | |
2398 | break; | |
2399 | } | |
2400 | ||
2401 | /* | |
2402 | * Create a pool with a spa version that can be upgraded. Pick | |
2403 | * a value between initial_version and SPA_VERSION_BEFORE_FEATURES. | |
2404 | */ | |
2405 | do { | |
2406 | version = ztest_random_spa_version(initial_version); | |
2407 | } while (version > SPA_VERSION_BEFORE_FEATURES); | |
2408 | ||
2409 | props = fnvlist_alloc(); | |
2410 | fnvlist_add_uint64(props, | |
2411 | zpool_prop_to_name(ZPOOL_PROP_VERSION), version); | |
2412 | VERIFY3S(spa_create(name, nvroot, props, NULL, NULL), ==, 0); | |
2413 | fnvlist_free(nvroot); | |
2414 | fnvlist_free(props); | |
2415 | ||
2416 | VERIFY3S(spa_open(name, &spa, FTAG), ==, 0); | |
2417 | VERIFY3U(spa_version(spa), ==, version); | |
2418 | newversion = ztest_random_spa_version(version + 1); | |
2419 | ||
2420 | if (ztest_opts.zo_verbose >= 4) { | |
2421 | (void) printf("upgrading spa version from %llu to %llu\n", | |
2422 | (u_longlong_t)version, (u_longlong_t)newversion); | |
2423 | } | |
2424 | ||
2425 | spa_upgrade(spa, newversion); | |
2426 | VERIFY3U(spa_version(spa), >, version); | |
2427 | VERIFY3U(spa_version(spa), ==, fnvlist_lookup_uint64(spa->spa_config, | |
2428 | zpool_prop_to_name(ZPOOL_PROP_VERSION))); | |
2429 | spa_close(spa, FTAG); | |
2430 | ||
2431 | strfree(name); | |
2432 | mutex_exit(&ztest_vdev_lock); | |
2433 | } | |
2434 | ||
428870ff BB |
2435 | static vdev_t * |
2436 | vdev_lookup_by_path(vdev_t *vd, const char *path) | |
2437 | { | |
2438 | vdev_t *mvd; | |
d6320ddb | 2439 | int c; |
428870ff BB |
2440 | |
2441 | if (vd->vdev_path != NULL && strcmp(path, vd->vdev_path) == 0) | |
2442 | return (vd); | |
2443 | ||
d6320ddb | 2444 | for (c = 0; c < vd->vdev_children; c++) |
428870ff BB |
2445 | if ((mvd = vdev_lookup_by_path(vd->vdev_child[c], path)) != |
2446 | NULL) | |
2447 | return (mvd); | |
2448 | ||
2449 | return (NULL); | |
2450 | } | |
2451 | ||
2452 | /* | |
2453 | * Find the first available hole which can be used as a top-level. | |
2454 | */ | |
2455 | int | |
2456 | find_vdev_hole(spa_t *spa) | |
2457 | { | |
2458 | vdev_t *rvd = spa->spa_root_vdev; | |
2459 | int c; | |
2460 | ||
2461 | ASSERT(spa_config_held(spa, SCL_VDEV, RW_READER) == SCL_VDEV); | |
2462 | ||
2463 | for (c = 0; c < rvd->vdev_children; c++) { | |
2464 | vdev_t *cvd = rvd->vdev_child[c]; | |
2465 | ||
2466 | if (cvd->vdev_ishole) | |
2467 | break; | |
2468 | } | |
2469 | return (c); | |
2470 | } | |
2471 | ||
2472 | /* | |
2473 | * Verify that vdev_add() works as expected. | |
2474 | */ | |
2475 | /* ARGSUSED */ | |
2476 | void | |
2477 | ztest_vdev_add_remove(ztest_ds_t *zd, uint64_t id) | |
2478 | { | |
2479 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 2480 | spa_t *spa = ztest_spa; |
428870ff BB |
2481 | uint64_t leaves; |
2482 | uint64_t guid; | |
2483 | nvlist_t *nvroot; | |
2484 | int error; | |
2485 | ||
c242c188 CS |
2486 | mutex_enter(&ztest_vdev_lock); |
2487 | leaves = | |
2488 | MAX(zs->zs_mirrors + zs->zs_splits, 1) * ztest_opts.zo_raidz; | |
428870ff BB |
2489 | |
2490 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
2491 | ||
2492 | ztest_shared->zs_vdev_next_leaf = find_vdev_hole(spa) * leaves; | |
2493 | ||
2494 | /* | |
2495 | * If we have slogs then remove them 1/4 of the time. | |
2496 | */ | |
2497 | if (spa_has_slogs(spa) && ztest_random(4) == 0) { | |
2498 | /* | |
2499 | * Grab the guid from the head of the log class rotor. | |
2500 | */ | |
2501 | guid = spa_log_class(spa)->mc_rotor->mg_vd->vdev_guid; | |
2502 | ||
2503 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
2504 | ||
2505 | /* | |
2506 | * We have to grab the zs_name_lock as writer to | |
2507 | * prevent a race between removing a slog (dmu_objset_find) | |
2508 | * and destroying a dataset. Removing the slog will | |
2509 | * grab a reference on the dataset which may cause | |
2510 | * dmu_objset_destroy() to fail with EBUSY thus | |
2511 | * leaving the dataset in an inconsistent state. | |
2512 | */ | |
c242c188 | 2513 | rw_enter(&ztest_name_lock, RW_WRITER); |
428870ff | 2514 | error = spa_vdev_remove(spa, guid, B_FALSE); |
c242c188 | 2515 | rw_exit(&ztest_name_lock); |
428870ff BB |
2516 | |
2517 | if (error && error != EEXIST) | |
2518 | fatal(0, "spa_vdev_remove() = %d", error); | |
2519 | } else { | |
2520 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
2521 | ||
2522 | /* | |
2523 | * Make 1/4 of the devices be log devices. | |
2524 | */ | |
ea0b2538 | 2525 | nvroot = make_vdev_root(NULL, NULL, NULL, |
c242c188 CS |
2526 | ztest_opts.zo_vdev_size, 0, |
2527 | ztest_random(4) == 0, ztest_opts.zo_raidz, | |
2528 | zs->zs_mirrors, 1); | |
428870ff BB |
2529 | |
2530 | error = spa_vdev_add(spa, nvroot); | |
2531 | nvlist_free(nvroot); | |
2532 | ||
2533 | if (error == ENOSPC) | |
2534 | ztest_record_enospc("spa_vdev_add"); | |
2535 | else if (error != 0) | |
2536 | fatal(0, "spa_vdev_add() = %d", error); | |
2537 | } | |
2538 | ||
c242c188 | 2539 | mutex_exit(&ztest_vdev_lock); |
428870ff BB |
2540 | } |
2541 | ||
2542 | /* | |
2543 | * Verify that adding/removing aux devices (l2arc, hot spare) works as expected. | |
2544 | */ | |
2545 | /* ARGSUSED */ | |
2546 | void | |
2547 | ztest_vdev_aux_add_remove(ztest_ds_t *zd, uint64_t id) | |
2548 | { | |
2549 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 2550 | spa_t *spa = ztest_spa; |
428870ff BB |
2551 | vdev_t *rvd = spa->spa_root_vdev; |
2552 | spa_aux_vdev_t *sav; | |
2553 | char *aux; | |
40b84e7a | 2554 | char *path; |
428870ff BB |
2555 | uint64_t guid = 0; |
2556 | int error; | |
2557 | ||
40b84e7a BB |
2558 | path = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
2559 | ||
428870ff BB |
2560 | if (ztest_random(2) == 0) { |
2561 | sav = &spa->spa_spares; | |
2562 | aux = ZPOOL_CONFIG_SPARES; | |
2563 | } else { | |
2564 | sav = &spa->spa_l2cache; | |
2565 | aux = ZPOOL_CONFIG_L2CACHE; | |
2566 | } | |
2567 | ||
c242c188 | 2568 | mutex_enter(&ztest_vdev_lock); |
428870ff BB |
2569 | |
2570 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
2571 | ||
2572 | if (sav->sav_count != 0 && ztest_random(4) == 0) { | |
b128c09f BB |
2573 | /* |
2574 | * Pick a random device to remove. | |
2575 | */ | |
2576 | guid = sav->sav_vdevs[ztest_random(sav->sav_count)]->vdev_guid; | |
2577 | } else { | |
2578 | /* | |
2579 | * Find an unused device we can add. | |
2580 | */ | |
428870ff | 2581 | zs->zs_vdev_aux = 0; |
b128c09f | 2582 | for (;;) { |
b128c09f | 2583 | int c; |
3db3ff4a | 2584 | (void) snprintf(path, MAXPATHLEN, ztest_aux_template, |
c242c188 CS |
2585 | ztest_opts.zo_dir, ztest_opts.zo_pool, aux, |
2586 | zs->zs_vdev_aux); | |
b128c09f BB |
2587 | for (c = 0; c < sav->sav_count; c++) |
2588 | if (strcmp(sav->sav_vdevs[c]->vdev_path, | |
2589 | path) == 0) | |
2590 | break; | |
2591 | if (c == sav->sav_count && | |
2592 | vdev_lookup_by_path(rvd, path) == NULL) | |
2593 | break; | |
428870ff | 2594 | zs->zs_vdev_aux++; |
34dc7c2f BB |
2595 | } |
2596 | } | |
2597 | ||
b128c09f | 2598 | spa_config_exit(spa, SCL_VDEV, FTAG); |
34dc7c2f | 2599 | |
b128c09f BB |
2600 | if (guid == 0) { |
2601 | /* | |
2602 | * Add a new device. | |
2603 | */ | |
ea0b2538 | 2604 | nvlist_t *nvroot = make_vdev_root(NULL, aux, NULL, |
c242c188 | 2605 | (ztest_opts.zo_vdev_size * 5) / 4, 0, 0, 0, 0, 1); |
b128c09f BB |
2606 | error = spa_vdev_add(spa, nvroot); |
2607 | if (error != 0) | |
2608 | fatal(0, "spa_vdev_add(%p) = %d", nvroot, error); | |
2609 | nvlist_free(nvroot); | |
2610 | } else { | |
2611 | /* | |
2612 | * Remove an existing device. Sometimes, dirty its | |
2613 | * vdev state first to make sure we handle removal | |
2614 | * of devices that have pending state changes. | |
2615 | */ | |
2616 | if (ztest_random(2) == 0) | |
9babb374 | 2617 | (void) vdev_online(spa, guid, 0, NULL); |
b128c09f BB |
2618 | |
2619 | error = spa_vdev_remove(spa, guid, B_FALSE); | |
2620 | if (error != 0 && error != EBUSY) | |
2621 | fatal(0, "spa_vdev_remove(%llu) = %d", guid, error); | |
2622 | } | |
2623 | ||
c242c188 | 2624 | mutex_exit(&ztest_vdev_lock); |
40b84e7a BB |
2625 | |
2626 | umem_free(path, MAXPATHLEN); | |
428870ff BB |
2627 | } |
2628 | ||
2629 | /* | |
2630 | * split a pool if it has mirror tlvdevs | |
2631 | */ | |
2632 | /* ARGSUSED */ | |
2633 | void | |
2634 | ztest_split_pool(ztest_ds_t *zd, uint64_t id) | |
2635 | { | |
2636 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 2637 | spa_t *spa = ztest_spa; |
428870ff BB |
2638 | vdev_t *rvd = spa->spa_root_vdev; |
2639 | nvlist_t *tree, **child, *config, *split, **schild; | |
2640 | uint_t c, children, schildren = 0, lastlogid = 0; | |
2641 | int error = 0; | |
2642 | ||
c242c188 | 2643 | mutex_enter(&ztest_vdev_lock); |
428870ff BB |
2644 | |
2645 | /* ensure we have a useable config; mirrors of raidz aren't supported */ | |
c242c188 CS |
2646 | if (zs->zs_mirrors < 3 || ztest_opts.zo_raidz > 1) { |
2647 | mutex_exit(&ztest_vdev_lock); | |
428870ff BB |
2648 | return; |
2649 | } | |
2650 | ||
2651 | /* clean up the old pool, if any */ | |
2652 | (void) spa_destroy("splitp"); | |
2653 | ||
2654 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
2655 | ||
2656 | /* generate a config from the existing config */ | |
2657 | mutex_enter(&spa->spa_props_lock); | |
2658 | VERIFY(nvlist_lookup_nvlist(spa->spa_config, ZPOOL_CONFIG_VDEV_TREE, | |
2659 | &tree) == 0); | |
2660 | mutex_exit(&spa->spa_props_lock); | |
2661 | ||
2662 | VERIFY(nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child, | |
2663 | &children) == 0); | |
2664 | ||
2665 | schild = malloc(rvd->vdev_children * sizeof (nvlist_t *)); | |
2666 | for (c = 0; c < children; c++) { | |
2667 | vdev_t *tvd = rvd->vdev_child[c]; | |
2668 | nvlist_t **mchild; | |
2669 | uint_t mchildren; | |
2670 | ||
2671 | if (tvd->vdev_islog || tvd->vdev_ops == &vdev_hole_ops) { | |
2672 | VERIFY(nvlist_alloc(&schild[schildren], NV_UNIQUE_NAME, | |
2673 | 0) == 0); | |
2674 | VERIFY(nvlist_add_string(schild[schildren], | |
2675 | ZPOOL_CONFIG_TYPE, VDEV_TYPE_HOLE) == 0); | |
2676 | VERIFY(nvlist_add_uint64(schild[schildren], | |
2677 | ZPOOL_CONFIG_IS_HOLE, 1) == 0); | |
2678 | if (lastlogid == 0) | |
2679 | lastlogid = schildren; | |
2680 | ++schildren; | |
2681 | continue; | |
2682 | } | |
2683 | lastlogid = 0; | |
2684 | VERIFY(nvlist_lookup_nvlist_array(child[c], | |
2685 | ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0); | |
2686 | VERIFY(nvlist_dup(mchild[0], &schild[schildren++], 0) == 0); | |
2687 | } | |
2688 | ||
2689 | /* OK, create a config that can be used to split */ | |
2690 | VERIFY(nvlist_alloc(&split, NV_UNIQUE_NAME, 0) == 0); | |
2691 | VERIFY(nvlist_add_string(split, ZPOOL_CONFIG_TYPE, | |
2692 | VDEV_TYPE_ROOT) == 0); | |
2693 | VERIFY(nvlist_add_nvlist_array(split, ZPOOL_CONFIG_CHILDREN, schild, | |
2694 | lastlogid != 0 ? lastlogid : schildren) == 0); | |
2695 | ||
2696 | VERIFY(nvlist_alloc(&config, NV_UNIQUE_NAME, 0) == 0); | |
2697 | VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, split) == 0); | |
2698 | ||
2699 | for (c = 0; c < schildren; c++) | |
2700 | nvlist_free(schild[c]); | |
2701 | free(schild); | |
2702 | nvlist_free(split); | |
2703 | ||
2704 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
2705 | ||
c242c188 | 2706 | (void) rw_enter(&ztest_name_lock, RW_WRITER); |
428870ff | 2707 | error = spa_vdev_split_mirror(spa, "splitp", config, NULL, B_FALSE); |
c242c188 | 2708 | (void) rw_exit(&ztest_name_lock); |
428870ff BB |
2709 | |
2710 | nvlist_free(config); | |
2711 | ||
2712 | if (error == 0) { | |
2713 | (void) printf("successful split - results:\n"); | |
2714 | mutex_enter(&spa_namespace_lock); | |
2715 | show_pool_stats(spa); | |
2716 | show_pool_stats(spa_lookup("splitp")); | |
2717 | mutex_exit(&spa_namespace_lock); | |
2718 | ++zs->zs_splits; | |
2719 | --zs->zs_mirrors; | |
2720 | } | |
c242c188 | 2721 | mutex_exit(&ztest_vdev_lock); |
428870ff | 2722 | |
34dc7c2f BB |
2723 | } |
2724 | ||
2725 | /* | |
2726 | * Verify that we can attach and detach devices. | |
2727 | */ | |
428870ff | 2728 | /* ARGSUSED */ |
34dc7c2f | 2729 | void |
428870ff | 2730 | ztest_vdev_attach_detach(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 2731 | { |
428870ff | 2732 | ztest_shared_t *zs = ztest_shared; |
c242c188 | 2733 | spa_t *spa = ztest_spa; |
b128c09f | 2734 | spa_aux_vdev_t *sav = &spa->spa_spares; |
34dc7c2f BB |
2735 | vdev_t *rvd = spa->spa_root_vdev; |
2736 | vdev_t *oldvd, *newvd, *pvd; | |
b128c09f | 2737 | nvlist_t *root; |
428870ff | 2738 | uint64_t leaves; |
34dc7c2f BB |
2739 | uint64_t leaf, top; |
2740 | uint64_t ashift = ztest_get_ashift(); | |
fb5f0bc8 | 2741 | uint64_t oldguid, pguid; |
34dc7c2f | 2742 | size_t oldsize, newsize; |
40b84e7a | 2743 | char *oldpath, *newpath; |
34dc7c2f | 2744 | int replacing; |
b128c09f BB |
2745 | int oldvd_has_siblings = B_FALSE; |
2746 | int newvd_is_spare = B_FALSE; | |
2747 | int oldvd_is_log; | |
34dc7c2f | 2748 | int error, expected_error; |
34dc7c2f | 2749 | |
40b84e7a BB |
2750 | oldpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
2751 | newpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
2752 | ||
c242c188 CS |
2753 | mutex_enter(&ztest_vdev_lock); |
2754 | leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raidz; | |
34dc7c2f | 2755 | |
b128c09f | 2756 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); |
34dc7c2f BB |
2757 | |
2758 | /* | |
2759 | * Decide whether to do an attach or a replace. | |
2760 | */ | |
2761 | replacing = ztest_random(2); | |
2762 | ||
2763 | /* | |
2764 | * Pick a random top-level vdev. | |
2765 | */ | |
428870ff | 2766 | top = ztest_random_vdev_top(spa, B_TRUE); |
34dc7c2f BB |
2767 | |
2768 | /* | |
2769 | * Pick a random leaf within it. | |
2770 | */ | |
2771 | leaf = ztest_random(leaves); | |
2772 | ||
2773 | /* | |
b128c09f | 2774 | * Locate this vdev. |
34dc7c2f | 2775 | */ |
b128c09f | 2776 | oldvd = rvd->vdev_child[top]; |
428870ff | 2777 | if (zs->zs_mirrors >= 1) { |
fb5f0bc8 | 2778 | ASSERT(oldvd->vdev_ops == &vdev_mirror_ops); |
428870ff | 2779 | ASSERT(oldvd->vdev_children >= zs->zs_mirrors); |
c242c188 | 2780 | oldvd = oldvd->vdev_child[leaf / ztest_opts.zo_raidz]; |
fb5f0bc8 | 2781 | } |
c242c188 | 2782 | if (ztest_opts.zo_raidz > 1) { |
fb5f0bc8 | 2783 | ASSERT(oldvd->vdev_ops == &vdev_raidz_ops); |
c242c188 CS |
2784 | ASSERT(oldvd->vdev_children == ztest_opts.zo_raidz); |
2785 | oldvd = oldvd->vdev_child[leaf % ztest_opts.zo_raidz]; | |
fb5f0bc8 | 2786 | } |
34dc7c2f BB |
2787 | |
2788 | /* | |
b128c09f BB |
2789 | * If we're already doing an attach or replace, oldvd may be a |
2790 | * mirror vdev -- in which case, pick a random child. | |
34dc7c2f | 2791 | */ |
b128c09f BB |
2792 | while (oldvd->vdev_children != 0) { |
2793 | oldvd_has_siblings = B_TRUE; | |
fb5f0bc8 BB |
2794 | ASSERT(oldvd->vdev_children >= 2); |
2795 | oldvd = oldvd->vdev_child[ztest_random(oldvd->vdev_children)]; | |
b128c09f BB |
2796 | } |
2797 | ||
2798 | oldguid = oldvd->vdev_guid; | |
9babb374 | 2799 | oldsize = vdev_get_min_asize(oldvd); |
b128c09f BB |
2800 | oldvd_is_log = oldvd->vdev_top->vdev_islog; |
2801 | (void) strcpy(oldpath, oldvd->vdev_path); | |
2802 | pvd = oldvd->vdev_parent; | |
fb5f0bc8 | 2803 | pguid = pvd->vdev_guid; |
34dc7c2f BB |
2804 | |
2805 | /* | |
b128c09f | 2806 | * If oldvd has siblings, then half of the time, detach it. |
34dc7c2f | 2807 | */ |
b128c09f BB |
2808 | if (oldvd_has_siblings && ztest_random(2) == 0) { |
2809 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
fb5f0bc8 BB |
2810 | error = spa_vdev_detach(spa, oldguid, pguid, B_FALSE); |
2811 | if (error != 0 && error != ENODEV && error != EBUSY && | |
2812 | error != ENOTSUP) | |
2813 | fatal(0, "detach (%s) returned %d", oldpath, error); | |
40b84e7a | 2814 | goto out; |
b128c09f | 2815 | } |
34dc7c2f BB |
2816 | |
2817 | /* | |
b128c09f BB |
2818 | * For the new vdev, choose with equal probability between the two |
2819 | * standard paths (ending in either 'a' or 'b') or a random hot spare. | |
34dc7c2f | 2820 | */ |
b128c09f BB |
2821 | if (sav->sav_count != 0 && ztest_random(3) == 0) { |
2822 | newvd = sav->sav_vdevs[ztest_random(sav->sav_count)]; | |
2823 | newvd_is_spare = B_TRUE; | |
2824 | (void) strcpy(newpath, newvd->vdev_path); | |
2825 | } else { | |
6aec1cd5 | 2826 | (void) snprintf(newpath, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
2827 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
2828 | top * leaves + leaf); | |
b128c09f BB |
2829 | if (ztest_random(2) == 0) |
2830 | newpath[strlen(newpath) - 1] = 'b'; | |
2831 | newvd = vdev_lookup_by_path(rvd, newpath); | |
2832 | } | |
2833 | ||
2834 | if (newvd) { | |
9babb374 | 2835 | newsize = vdev_get_min_asize(newvd); |
b128c09f BB |
2836 | } else { |
2837 | /* | |
2838 | * Make newsize a little bigger or smaller than oldsize. | |
2839 | * If it's smaller, the attach should fail. | |
2840 | * If it's larger, and we're doing a replace, | |
2841 | * we should get dynamic LUN growth when we're done. | |
2842 | */ | |
2843 | newsize = 10 * oldsize / (9 + ztest_random(3)); | |
2844 | } | |
34dc7c2f BB |
2845 | |
2846 | /* | |
2847 | * If pvd is not a mirror or root, the attach should fail with ENOTSUP, | |
2848 | * unless it's a replace; in that case any non-replacing parent is OK. | |
2849 | * | |
2850 | * If newvd is already part of the pool, it should fail with EBUSY. | |
2851 | * | |
2852 | * If newvd is too small, it should fail with EOVERFLOW. | |
2853 | */ | |
b128c09f BB |
2854 | if (pvd->vdev_ops != &vdev_mirror_ops && |
2855 | pvd->vdev_ops != &vdev_root_ops && (!replacing || | |
2856 | pvd->vdev_ops == &vdev_replacing_ops || | |
2857 | pvd->vdev_ops == &vdev_spare_ops)) | |
34dc7c2f | 2858 | expected_error = ENOTSUP; |
b128c09f BB |
2859 | else if (newvd_is_spare && (!replacing || oldvd_is_log)) |
2860 | expected_error = ENOTSUP; | |
2861 | else if (newvd == oldvd) | |
2862 | expected_error = replacing ? 0 : EBUSY; | |
2863 | else if (vdev_lookup_by_path(rvd, newpath) != NULL) | |
2864 | expected_error = EBUSY; | |
34dc7c2f BB |
2865 | else if (newsize < oldsize) |
2866 | expected_error = EOVERFLOW; | |
2867 | else if (ashift > oldvd->vdev_top->vdev_ashift) | |
2868 | expected_error = EDOM; | |
2869 | else | |
2870 | expected_error = 0; | |
2871 | ||
b128c09f | 2872 | spa_config_exit(spa, SCL_VDEV, FTAG); |
34dc7c2f BB |
2873 | |
2874 | /* | |
2875 | * Build the nvlist describing newpath. | |
2876 | */ | |
ea0b2538 | 2877 | root = make_vdev_root(newpath, NULL, NULL, newvd == NULL ? newsize : 0, |
b128c09f | 2878 | ashift, 0, 0, 0, 1); |
34dc7c2f | 2879 | |
b128c09f | 2880 | error = spa_vdev_attach(spa, oldguid, root, replacing); |
34dc7c2f | 2881 | |
34dc7c2f BB |
2882 | nvlist_free(root); |
2883 | ||
2884 | /* | |
2885 | * If our parent was the replacing vdev, but the replace completed, | |
2886 | * then instead of failing with ENOTSUP we may either succeed, | |
2887 | * fail with ENODEV, or fail with EOVERFLOW. | |
2888 | */ | |
2889 | if (expected_error == ENOTSUP && | |
2890 | (error == 0 || error == ENODEV || error == EOVERFLOW)) | |
2891 | expected_error = error; | |
2892 | ||
2893 | /* | |
2894 | * If someone grew the LUN, the replacement may be too small. | |
2895 | */ | |
b128c09f | 2896 | if (error == EOVERFLOW || error == EBUSY) |
34dc7c2f BB |
2897 | expected_error = error; |
2898 | ||
b128c09f BB |
2899 | /* XXX workaround 6690467 */ |
2900 | if (error != expected_error && expected_error != EBUSY) { | |
2901 | fatal(0, "attach (%s %llu, %s %llu, %d) " | |
2902 | "returned %d, expected %d", | |
2903 | oldpath, (longlong_t)oldsize, newpath, | |
2904 | (longlong_t)newsize, replacing, error, expected_error); | |
34dc7c2f | 2905 | } |
40b84e7a | 2906 | out: |
c242c188 | 2907 | mutex_exit(&ztest_vdev_lock); |
40b84e7a BB |
2908 | |
2909 | umem_free(oldpath, MAXPATHLEN); | |
2910 | umem_free(newpath, MAXPATHLEN); | |
34dc7c2f BB |
2911 | } |
2912 | ||
9babb374 BB |
2913 | /* |
2914 | * Callback function which expands the physical size of the vdev. | |
2915 | */ | |
2916 | vdev_t * | |
2917 | grow_vdev(vdev_t *vd, void *arg) | |
2918 | { | |
1fde1e37 | 2919 | ASSERTV(spa_t *spa = vd->vdev_spa); |
9babb374 BB |
2920 | size_t *newsize = arg; |
2921 | size_t fsize; | |
2922 | int fd; | |
2923 | ||
2924 | ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE); | |
2925 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
2926 | ||
2927 | if ((fd = open(vd->vdev_path, O_RDWR)) == -1) | |
2928 | return (vd); | |
2929 | ||
2930 | fsize = lseek(fd, 0, SEEK_END); | |
0e5b68e0 | 2931 | VERIFY(ftruncate(fd, *newsize) == 0); |
9babb374 | 2932 | |
c242c188 | 2933 | if (ztest_opts.zo_verbose >= 6) { |
9babb374 BB |
2934 | (void) printf("%s grew from %lu to %lu bytes\n", |
2935 | vd->vdev_path, (ulong_t)fsize, (ulong_t)*newsize); | |
2936 | } | |
2937 | (void) close(fd); | |
2938 | return (NULL); | |
2939 | } | |
2940 | ||
2941 | /* | |
2942 | * Callback function which expands a given vdev by calling vdev_online(). | |
2943 | */ | |
2944 | /* ARGSUSED */ | |
2945 | vdev_t * | |
2946 | online_vdev(vdev_t *vd, void *arg) | |
2947 | { | |
2948 | spa_t *spa = vd->vdev_spa; | |
2949 | vdev_t *tvd = vd->vdev_top; | |
9babb374 | 2950 | uint64_t guid = vd->vdev_guid; |
428870ff BB |
2951 | uint64_t generation = spa->spa_config_generation + 1; |
2952 | vdev_state_t newstate = VDEV_STATE_UNKNOWN; | |
2953 | int error; | |
9babb374 BB |
2954 | |
2955 | ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE); | |
2956 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
2957 | ||
2958 | /* Calling vdev_online will initialize the new metaslabs */ | |
2959 | spa_config_exit(spa, SCL_STATE, spa); | |
428870ff | 2960 | error = vdev_online(spa, guid, ZFS_ONLINE_EXPAND, &newstate); |
9babb374 BB |
2961 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); |
2962 | ||
428870ff BB |
2963 | /* |
2964 | * If vdev_online returned an error or the underlying vdev_open | |
2965 | * failed then we abort the expand. The only way to know that | |
2966 | * vdev_open fails is by checking the returned newstate. | |
2967 | */ | |
2968 | if (error || newstate != VDEV_STATE_HEALTHY) { | |
c242c188 | 2969 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
2970 | (void) printf("Unable to expand vdev, state %llu, " |
2971 | "error %d\n", (u_longlong_t)newstate, error); | |
2972 | } | |
2973 | return (vd); | |
2974 | } | |
2975 | ASSERT3U(newstate, ==, VDEV_STATE_HEALTHY); | |
2976 | ||
9babb374 BB |
2977 | /* |
2978 | * Since we dropped the lock we need to ensure that we're | |
2979 | * still talking to the original vdev. It's possible this | |
2980 | * vdev may have been detached/replaced while we were | |
2981 | * trying to online it. | |
2982 | */ | |
428870ff | 2983 | if (generation != spa->spa_config_generation) { |
c242c188 | 2984 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
2985 | (void) printf("vdev configuration has changed, " |
2986 | "guid %llu, state %llu, expected gen %llu, " | |
2987 | "got gen %llu\n", | |
2988 | (u_longlong_t)guid, | |
2989 | (u_longlong_t)tvd->vdev_state, | |
2990 | (u_longlong_t)generation, | |
2991 | (u_longlong_t)spa->spa_config_generation); | |
9babb374 BB |
2992 | } |
2993 | return (vd); | |
2994 | } | |
2995 | return (NULL); | |
2996 | } | |
2997 | ||
2998 | /* | |
2999 | * Traverse the vdev tree calling the supplied function. | |
3000 | * We continue to walk the tree until we either have walked all | |
3001 | * children or we receive a non-NULL return from the callback. | |
3002 | * If a NULL callback is passed, then we just return back the first | |
3003 | * leaf vdev we encounter. | |
3004 | */ | |
3005 | vdev_t * | |
3006 | vdev_walk_tree(vdev_t *vd, vdev_t *(*func)(vdev_t *, void *), void *arg) | |
3007 | { | |
d6320ddb BB |
3008 | uint_t c; |
3009 | ||
9babb374 BB |
3010 | if (vd->vdev_ops->vdev_op_leaf) { |
3011 | if (func == NULL) | |
3012 | return (vd); | |
3013 | else | |
3014 | return (func(vd, arg)); | |
3015 | } | |
3016 | ||
d6320ddb | 3017 | for (c = 0; c < vd->vdev_children; c++) { |
9babb374 BB |
3018 | vdev_t *cvd = vd->vdev_child[c]; |
3019 | if ((cvd = vdev_walk_tree(cvd, func, arg)) != NULL) | |
3020 | return (cvd); | |
3021 | } | |
3022 | return (NULL); | |
3023 | } | |
3024 | ||
34dc7c2f BB |
3025 | /* |
3026 | * Verify that dynamic LUN growth works as expected. | |
3027 | */ | |
428870ff | 3028 | /* ARGSUSED */ |
34dc7c2f | 3029 | void |
428870ff | 3030 | ztest_vdev_LUN_growth(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3031 | { |
c242c188 | 3032 | spa_t *spa = ztest_spa; |
428870ff BB |
3033 | vdev_t *vd, *tvd; |
3034 | metaslab_class_t *mc; | |
3035 | metaslab_group_t *mg; | |
9babb374 | 3036 | size_t psize, newsize; |
428870ff BB |
3037 | uint64_t top; |
3038 | uint64_t old_class_space, new_class_space, old_ms_count, new_ms_count; | |
34dc7c2f | 3039 | |
c242c188 | 3040 | mutex_enter(&ztest_vdev_lock); |
9babb374 BB |
3041 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); |
3042 | ||
428870ff | 3043 | top = ztest_random_vdev_top(spa, B_TRUE); |
9babb374 | 3044 | |
428870ff BB |
3045 | tvd = spa->spa_root_vdev->vdev_child[top]; |
3046 | mg = tvd->vdev_mg; | |
3047 | mc = mg->mg_class; | |
3048 | old_ms_count = tvd->vdev_ms_count; | |
3049 | old_class_space = metaslab_class_get_space(mc); | |
34dc7c2f BB |
3050 | |
3051 | /* | |
9babb374 BB |
3052 | * Determine the size of the first leaf vdev associated with |
3053 | * our top-level device. | |
34dc7c2f | 3054 | */ |
9babb374 BB |
3055 | vd = vdev_walk_tree(tvd, NULL, NULL); |
3056 | ASSERT3P(vd, !=, NULL); | |
3057 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
34dc7c2f | 3058 | |
9babb374 | 3059 | psize = vd->vdev_psize; |
34dc7c2f | 3060 | |
9babb374 | 3061 | /* |
428870ff BB |
3062 | * We only try to expand the vdev if it's healthy, less than 4x its |
3063 | * original size, and it has a valid psize. | |
9babb374 | 3064 | */ |
428870ff | 3065 | if (tvd->vdev_state != VDEV_STATE_HEALTHY || |
c242c188 | 3066 | psize == 0 || psize >= 4 * ztest_opts.zo_vdev_size) { |
9babb374 | 3067 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 3068 | mutex_exit(&ztest_vdev_lock); |
9babb374 BB |
3069 | return; |
3070 | } | |
3071 | ASSERT(psize > 0); | |
3072 | newsize = psize + psize / 8; | |
3073 | ASSERT3U(newsize, >, psize); | |
34dc7c2f | 3074 | |
c242c188 | 3075 | if (ztest_opts.zo_verbose >= 6) { |
428870ff | 3076 | (void) printf("Expanding LUN %s from %lu to %lu\n", |
9babb374 BB |
3077 | vd->vdev_path, (ulong_t)psize, (ulong_t)newsize); |
3078 | } | |
3079 | ||
9babb374 BB |
3080 | /* |
3081 | * Growing the vdev is a two step process: | |
3082 | * 1). expand the physical size (i.e. relabel) | |
3083 | * 2). online the vdev to create the new metaslabs | |
3084 | */ | |
3085 | if (vdev_walk_tree(tvd, grow_vdev, &newsize) != NULL || | |
3086 | vdev_walk_tree(tvd, online_vdev, NULL) != NULL || | |
3087 | tvd->vdev_state != VDEV_STATE_HEALTHY) { | |
c242c188 | 3088 | if (ztest_opts.zo_verbose >= 5) { |
9babb374 | 3089 | (void) printf("Could not expand LUN because " |
428870ff | 3090 | "the vdev configuration changed.\n"); |
34dc7c2f | 3091 | } |
428870ff | 3092 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 3093 | mutex_exit(&ztest_vdev_lock); |
9babb374 | 3094 | return; |
34dc7c2f BB |
3095 | } |
3096 | ||
428870ff | 3097 | spa_config_exit(spa, SCL_STATE, spa); |
9babb374 BB |
3098 | |
3099 | /* | |
3100 | * Expanding the LUN will update the config asynchronously, | |
3101 | * thus we must wait for the async thread to complete any | |
3102 | * pending tasks before proceeding. | |
3103 | */ | |
428870ff BB |
3104 | for (;;) { |
3105 | boolean_t done; | |
3106 | mutex_enter(&spa->spa_async_lock); | |
3107 | done = (spa->spa_async_thread == NULL && !spa->spa_async_tasks); | |
3108 | mutex_exit(&spa->spa_async_lock); | |
3109 | if (done) | |
3110 | break; | |
3111 | txg_wait_synced(spa_get_dsl(spa), 0); | |
3112 | (void) poll(NULL, 0, 100); | |
3113 | } | |
9babb374 BB |
3114 | |
3115 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); | |
428870ff BB |
3116 | |
3117 | tvd = spa->spa_root_vdev->vdev_child[top]; | |
3118 | new_ms_count = tvd->vdev_ms_count; | |
3119 | new_class_space = metaslab_class_get_space(mc); | |
3120 | ||
3121 | if (tvd->vdev_mg != mg || mg->mg_class != mc) { | |
c242c188 | 3122 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
3123 | (void) printf("Could not verify LUN expansion due to " |
3124 | "intervening vdev offline or remove.\n"); | |
3125 | } | |
3126 | spa_config_exit(spa, SCL_STATE, spa); | |
c242c188 | 3127 | mutex_exit(&ztest_vdev_lock); |
428870ff BB |
3128 | return; |
3129 | } | |
3130 | ||
3131 | /* | |
3132 | * Make sure we were able to grow the vdev. | |
3133 | */ | |
3134 | if (new_ms_count <= old_ms_count) | |
3135 | fatal(0, "LUN expansion failed: ms_count %llu <= %llu\n", | |
3136 | old_ms_count, new_ms_count); | |
9babb374 BB |
3137 | |
3138 | /* | |
3139 | * Make sure we were able to grow the pool. | |
3140 | */ | |
428870ff BB |
3141 | if (new_class_space <= old_class_space) |
3142 | fatal(0, "LUN expansion failed: class_space %llu <= %llu\n", | |
3143 | old_class_space, new_class_space); | |
3144 | ||
c242c188 | 3145 | if (ztest_opts.zo_verbose >= 5) { |
9babb374 BB |
3146 | char oldnumbuf[6], newnumbuf[6]; |
3147 | ||
428870ff BB |
3148 | nicenum(old_class_space, oldnumbuf); |
3149 | nicenum(new_class_space, newnumbuf); | |
9babb374 BB |
3150 | (void) printf("%s grew from %s to %s\n", |
3151 | spa->spa_name, oldnumbuf, newnumbuf); | |
3152 | } | |
428870ff | 3153 | |
9babb374 | 3154 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 3155 | mutex_exit(&ztest_vdev_lock); |
34dc7c2f BB |
3156 | } |
3157 | ||
428870ff BB |
3158 | /* |
3159 | * Verify that dmu_objset_{create,destroy,open,close} work as expected. | |
3160 | */ | |
34dc7c2f BB |
3161 | /* ARGSUSED */ |
3162 | static void | |
428870ff | 3163 | ztest_objset_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) |
34dc7c2f BB |
3164 | { |
3165 | /* | |
428870ff | 3166 | * Create the objects common to all ztest datasets. |
34dc7c2f | 3167 | */ |
428870ff | 3168 | VERIFY(zap_create_claim(os, ZTEST_DIROBJ, |
34dc7c2f | 3169 | DMU_OT_ZAP_OTHER, DMU_OT_NONE, 0, tx) == 0); |
428870ff | 3170 | } |
34dc7c2f | 3171 | |
428870ff BB |
3172 | static int |
3173 | ztest_dataset_create(char *dsname) | |
3174 | { | |
3175 | uint64_t zilset = ztest_random(100); | |
3176 | int err = dmu_objset_create(dsname, DMU_OST_OTHER, 0, | |
3177 | ztest_objset_create_cb, NULL); | |
3178 | ||
3179 | if (err || zilset < 80) | |
3180 | return (err); | |
3181 | ||
c242c188 | 3182 | if (ztest_opts.zo_verbose >= 5) |
b815ff9a | 3183 | (void) printf("Setting dataset %s to sync always\n", dsname); |
428870ff BB |
3184 | return (ztest_dsl_prop_set_uint64(dsname, ZFS_PROP_SYNC, |
3185 | ZFS_SYNC_ALWAYS, B_FALSE)); | |
34dc7c2f BB |
3186 | } |
3187 | ||
428870ff | 3188 | /* ARGSUSED */ |
34dc7c2f | 3189 | static int |
428870ff | 3190 | ztest_objset_destroy_cb(const char *name, void *arg) |
34dc7c2f | 3191 | { |
34dc7c2f | 3192 | objset_t *os; |
428870ff | 3193 | dmu_object_info_t doi; |
34dc7c2f BB |
3194 | int error; |
3195 | ||
3196 | /* | |
3197 | * Verify that the dataset contains a directory object. | |
3198 | */ | |
428870ff BB |
3199 | VERIFY3U(0, ==, dmu_objset_hold(name, FTAG, &os)); |
3200 | error = dmu_object_info(os, ZTEST_DIROBJ, &doi); | |
34dc7c2f BB |
3201 | if (error != ENOENT) { |
3202 | /* We could have crashed in the middle of destroying it */ | |
c99c9001 | 3203 | ASSERT0(error); |
428870ff BB |
3204 | ASSERT3U(doi.doi_type, ==, DMU_OT_ZAP_OTHER); |
3205 | ASSERT3S(doi.doi_physical_blocks_512, >=, 0); | |
34dc7c2f | 3206 | } |
428870ff | 3207 | dmu_objset_rele(os, FTAG); |
34dc7c2f BB |
3208 | |
3209 | /* | |
3210 | * Destroy the dataset. | |
3211 | */ | |
428870ff | 3212 | VERIFY3U(0, ==, dmu_objset_destroy(name, B_FALSE)); |
34dc7c2f BB |
3213 | return (0); |
3214 | } | |
3215 | ||
428870ff BB |
3216 | static boolean_t |
3217 | ztest_snapshot_create(char *osname, uint64_t id) | |
34dc7c2f | 3218 | { |
428870ff BB |
3219 | char snapname[MAXNAMELEN]; |
3220 | int error; | |
3221 | ||
3222 | (void) snprintf(snapname, MAXNAMELEN, "%s@%llu", osname, | |
3223 | (u_longlong_t)id); | |
3224 | ||
3225 | error = dmu_objset_snapshot(osname, strchr(snapname, '@') + 1, | |
572e2857 | 3226 | NULL, NULL, B_FALSE, B_FALSE, -1); |
428870ff BB |
3227 | if (error == ENOSPC) { |
3228 | ztest_record_enospc(FTAG); | |
3229 | return (B_FALSE); | |
3230 | } | |
3231 | if (error != 0 && error != EEXIST) | |
3232 | fatal(0, "ztest_snapshot_create(%s) = %d", snapname, error); | |
3233 | return (B_TRUE); | |
3234 | } | |
3235 | ||
3236 | static boolean_t | |
3237 | ztest_snapshot_destroy(char *osname, uint64_t id) | |
3238 | { | |
3239 | char snapname[MAXNAMELEN]; | |
3240 | int error; | |
3241 | ||
3242 | (void) snprintf(snapname, MAXNAMELEN, "%s@%llu", osname, | |
3243 | (u_longlong_t)id); | |
3244 | ||
3245 | error = dmu_objset_destroy(snapname, B_FALSE); | |
3246 | if (error != 0 && error != ENOENT) | |
3247 | fatal(0, "ztest_snapshot_destroy(%s) = %d", snapname, error); | |
3248 | return (B_TRUE); | |
34dc7c2f BB |
3249 | } |
3250 | ||
428870ff | 3251 | /* ARGSUSED */ |
34dc7c2f | 3252 | void |
428870ff | 3253 | ztest_dmu_objset_create_destroy(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3254 | { |
40b84e7a | 3255 | ztest_ds_t *zdtmp; |
428870ff | 3256 | int iters; |
34dc7c2f | 3257 | int error; |
b128c09f | 3258 | objset_t *os, *os2; |
40b84e7a | 3259 | char *name; |
34dc7c2f | 3260 | zilog_t *zilog; |
d6320ddb | 3261 | int i; |
34dc7c2f | 3262 | |
40b84e7a BB |
3263 | zdtmp = umem_alloc(sizeof (ztest_ds_t), UMEM_NOFAIL); |
3264 | name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3265 | ||
c242c188 | 3266 | (void) rw_enter(&ztest_name_lock, RW_READER); |
34dc7c2f | 3267 | |
428870ff | 3268 | (void) snprintf(name, MAXNAMELEN, "%s/temp_%llu", |
c242c188 | 3269 | ztest_opts.zo_pool, (u_longlong_t)id); |
34dc7c2f BB |
3270 | |
3271 | /* | |
3272 | * If this dataset exists from a previous run, process its replay log | |
3273 | * half of the time. If we don't replay it, then dmu_objset_destroy() | |
428870ff | 3274 | * (invoked from ztest_objset_destroy_cb()) should just throw it away. |
34dc7c2f BB |
3275 | */ |
3276 | if (ztest_random(2) == 0 && | |
428870ff | 3277 | dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, FTAG, &os) == 0) { |
c242c188 | 3278 | ztest_zd_init(zdtmp, NULL, os); |
40b84e7a BB |
3279 | zil_replay(os, zdtmp, ztest_replay_vector); |
3280 | ztest_zd_fini(zdtmp); | |
428870ff | 3281 | dmu_objset_disown(os, FTAG); |
34dc7c2f BB |
3282 | } |
3283 | ||
3284 | /* | |
3285 | * There may be an old instance of the dataset we're about to | |
3286 | * create lying around from a previous run. If so, destroy it | |
3287 | * and all of its snapshots. | |
3288 | */ | |
428870ff | 3289 | (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL, |
34dc7c2f BB |
3290 | DS_FIND_CHILDREN | DS_FIND_SNAPSHOTS); |
3291 | ||
3292 | /* | |
3293 | * Verify that the destroyed dataset is no longer in the namespace. | |
3294 | */ | |
428870ff | 3295 | VERIFY3U(ENOENT, ==, dmu_objset_hold(name, FTAG, &os)); |
34dc7c2f BB |
3296 | |
3297 | /* | |
3298 | * Verify that we can create a new dataset. | |
3299 | */ | |
428870ff | 3300 | error = ztest_dataset_create(name); |
34dc7c2f BB |
3301 | if (error) { |
3302 | if (error == ENOSPC) { | |
428870ff | 3303 | ztest_record_enospc(FTAG); |
40b84e7a | 3304 | goto out; |
34dc7c2f BB |
3305 | } |
3306 | fatal(0, "dmu_objset_create(%s) = %d", name, error); | |
3307 | } | |
3308 | ||
428870ff BB |
3309 | VERIFY3U(0, ==, |
3310 | dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, FTAG, &os)); | |
3311 | ||
c242c188 | 3312 | ztest_zd_init(zdtmp, NULL, os); |
34dc7c2f BB |
3313 | |
3314 | /* | |
3315 | * Open the intent log for it. | |
3316 | */ | |
428870ff | 3317 | zilog = zil_open(os, ztest_get_data); |
34dc7c2f BB |
3318 | |
3319 | /* | |
428870ff BB |
3320 | * Put some objects in there, do a little I/O to them, |
3321 | * and randomly take a couple of snapshots along the way. | |
34dc7c2f | 3322 | */ |
428870ff | 3323 | iters = ztest_random(5); |
d6320ddb | 3324 | for (i = 0; i < iters; i++) { |
40b84e7a | 3325 | ztest_dmu_object_alloc_free(zdtmp, id); |
428870ff BB |
3326 | if (ztest_random(iters) == 0) |
3327 | (void) ztest_snapshot_create(name, i); | |
34dc7c2f BB |
3328 | } |
3329 | ||
3330 | /* | |
3331 | * Verify that we cannot create an existing dataset. | |
3332 | */ | |
428870ff BB |
3333 | VERIFY3U(EEXIST, ==, |
3334 | dmu_objset_create(name, DMU_OST_OTHER, 0, NULL, NULL)); | |
34dc7c2f BB |
3335 | |
3336 | /* | |
428870ff | 3337 | * Verify that we can hold an objset that is also owned. |
b128c09f | 3338 | */ |
428870ff BB |
3339 | VERIFY3U(0, ==, dmu_objset_hold(name, FTAG, &os2)); |
3340 | dmu_objset_rele(os2, FTAG); | |
34dc7c2f | 3341 | |
428870ff BB |
3342 | /* |
3343 | * Verify that we cannot own an objset that is already owned. | |
3344 | */ | |
3345 | VERIFY3U(EBUSY, ==, | |
3346 | dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, FTAG, &os2)); | |
34dc7c2f | 3347 | |
428870ff BB |
3348 | zil_close(zilog); |
3349 | dmu_objset_disown(os, FTAG); | |
40b84e7a BB |
3350 | ztest_zd_fini(zdtmp); |
3351 | out: | |
c242c188 | 3352 | (void) rw_exit(&ztest_name_lock); |
40b84e7a BB |
3353 | |
3354 | umem_free(name, MAXNAMELEN); | |
3355 | umem_free(zdtmp, sizeof (ztest_ds_t)); | |
34dc7c2f BB |
3356 | } |
3357 | ||
3358 | /* | |
3359 | * Verify that dmu_snapshot_{create,destroy,open,close} work as expected. | |
3360 | */ | |
3361 | void | |
428870ff | 3362 | ztest_dmu_snapshot_create_destroy(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3363 | { |
c242c188 | 3364 | (void) rw_enter(&ztest_name_lock, RW_READER); |
428870ff BB |
3365 | (void) ztest_snapshot_destroy(zd->zd_name, id); |
3366 | (void) ztest_snapshot_create(zd->zd_name, id); | |
c242c188 | 3367 | (void) rw_exit(&ztest_name_lock); |
34dc7c2f BB |
3368 | } |
3369 | ||
9babb374 BB |
3370 | /* |
3371 | * Cleanup non-standard snapshots and clones. | |
3372 | */ | |
3373 | void | |
428870ff | 3374 | ztest_dsl_dataset_cleanup(char *osname, uint64_t id) |
9babb374 | 3375 | { |
40b84e7a BB |
3376 | char *snap1name; |
3377 | char *clone1name; | |
3378 | char *snap2name; | |
3379 | char *clone2name; | |
3380 | char *snap3name; | |
9babb374 BB |
3381 | int error; |
3382 | ||
40b84e7a BB |
3383 | snap1name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); |
3384 | clone1name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3385 | snap2name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3386 | clone2name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3387 | snap3name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3388 | ||
b8864a23 BB |
3389 | (void) snprintf(snap1name, MAXNAMELEN, "%s@s1_%llu", |
3390 | osname, (u_longlong_t)id); | |
3391 | (void) snprintf(clone1name, MAXNAMELEN, "%s/c1_%llu", | |
3392 | osname, (u_longlong_t)id); | |
3393 | (void) snprintf(snap2name, MAXNAMELEN, "%s@s2_%llu", | |
3394 | clone1name, (u_longlong_t)id); | |
3395 | (void) snprintf(clone2name, MAXNAMELEN, "%s/c2_%llu", | |
3396 | osname, (u_longlong_t)id); | |
3397 | (void) snprintf(snap3name, MAXNAMELEN, "%s@s3_%llu", | |
3398 | clone1name, (u_longlong_t)id); | |
9babb374 | 3399 | |
45d1cae3 | 3400 | error = dmu_objset_destroy(clone2name, B_FALSE); |
9babb374 BB |
3401 | if (error && error != ENOENT) |
3402 | fatal(0, "dmu_objset_destroy(%s) = %d", clone2name, error); | |
45d1cae3 | 3403 | error = dmu_objset_destroy(snap3name, B_FALSE); |
9babb374 BB |
3404 | if (error && error != ENOENT) |
3405 | fatal(0, "dmu_objset_destroy(%s) = %d", snap3name, error); | |
45d1cae3 | 3406 | error = dmu_objset_destroy(snap2name, B_FALSE); |
9babb374 BB |
3407 | if (error && error != ENOENT) |
3408 | fatal(0, "dmu_objset_destroy(%s) = %d", snap2name, error); | |
45d1cae3 | 3409 | error = dmu_objset_destroy(clone1name, B_FALSE); |
9babb374 BB |
3410 | if (error && error != ENOENT) |
3411 | fatal(0, "dmu_objset_destroy(%s) = %d", clone1name, error); | |
45d1cae3 | 3412 | error = dmu_objset_destroy(snap1name, B_FALSE); |
9babb374 BB |
3413 | if (error && error != ENOENT) |
3414 | fatal(0, "dmu_objset_destroy(%s) = %d", snap1name, error); | |
40b84e7a BB |
3415 | |
3416 | umem_free(snap1name, MAXNAMELEN); | |
3417 | umem_free(clone1name, MAXNAMELEN); | |
3418 | umem_free(snap2name, MAXNAMELEN); | |
3419 | umem_free(clone2name, MAXNAMELEN); | |
3420 | umem_free(snap3name, MAXNAMELEN); | |
9babb374 BB |
3421 | } |
3422 | ||
3423 | /* | |
3424 | * Verify dsl_dataset_promote handles EBUSY | |
3425 | */ | |
3426 | void | |
428870ff | 3427 | ztest_dsl_dataset_promote_busy(ztest_ds_t *zd, uint64_t id) |
9babb374 | 3428 | { |
9babb374 BB |
3429 | objset_t *clone; |
3430 | dsl_dataset_t *ds; | |
40b84e7a BB |
3431 | char *snap1name; |
3432 | char *clone1name; | |
3433 | char *snap2name; | |
3434 | char *clone2name; | |
3435 | char *snap3name; | |
428870ff BB |
3436 | char *osname = zd->zd_name; |
3437 | int error; | |
9babb374 | 3438 | |
40b84e7a BB |
3439 | snap1name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); |
3440 | clone1name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3441 | snap2name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3442 | clone2name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3443 | snap3name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3444 | ||
c242c188 | 3445 | (void) rw_enter(&ztest_name_lock, RW_READER); |
9babb374 | 3446 | |
428870ff | 3447 | ztest_dsl_dataset_cleanup(osname, id); |
9babb374 | 3448 | |
b8864a23 BB |
3449 | (void) snprintf(snap1name, MAXNAMELEN, "%s@s1_%llu", |
3450 | osname, (u_longlong_t)id); | |
3451 | (void) snprintf(clone1name, MAXNAMELEN, "%s/c1_%llu", | |
3452 | osname, (u_longlong_t)id); | |
3453 | (void) snprintf(snap2name, MAXNAMELEN, "%s@s2_%llu", | |
3454 | clone1name, (u_longlong_t)id); | |
3455 | (void) snprintf(clone2name, MAXNAMELEN, "%s/c2_%llu", | |
3456 | osname, (u_longlong_t)id); | |
3457 | (void) snprintf(snap3name, MAXNAMELEN, "%s@s3_%llu", | |
3458 | clone1name, (u_longlong_t)id); | |
9babb374 BB |
3459 | |
3460 | error = dmu_objset_snapshot(osname, strchr(snap1name, '@')+1, | |
572e2857 | 3461 | NULL, NULL, B_FALSE, B_FALSE, -1); |
9babb374 BB |
3462 | if (error && error != EEXIST) { |
3463 | if (error == ENOSPC) { | |
428870ff | 3464 | ztest_record_enospc(FTAG); |
9babb374 BB |
3465 | goto out; |
3466 | } | |
3467 | fatal(0, "dmu_take_snapshot(%s) = %d", snap1name, error); | |
3468 | } | |
3469 | ||
428870ff | 3470 | error = dmu_objset_hold(snap1name, FTAG, &clone); |
9babb374 BB |
3471 | if (error) |
3472 | fatal(0, "dmu_open_snapshot(%s) = %d", snap1name, error); | |
3473 | ||
428870ff BB |
3474 | error = dmu_objset_clone(clone1name, dmu_objset_ds(clone), 0); |
3475 | dmu_objset_rele(clone, FTAG); | |
9babb374 BB |
3476 | if (error) { |
3477 | if (error == ENOSPC) { | |
428870ff | 3478 | ztest_record_enospc(FTAG); |
9babb374 BB |
3479 | goto out; |
3480 | } | |
3481 | fatal(0, "dmu_objset_create(%s) = %d", clone1name, error); | |
3482 | } | |
3483 | ||
428870ff | 3484 | error = dmu_objset_snapshot(clone1name, strchr(snap2name, '@')+1, |
572e2857 | 3485 | NULL, NULL, B_FALSE, B_FALSE, -1); |
428870ff BB |
3486 | if (error && error != EEXIST) { |
3487 | if (error == ENOSPC) { | |
3488 | ztest_record_enospc(FTAG); | |
3489 | goto out; | |
34dc7c2f | 3490 | } |
428870ff BB |
3491 | fatal(0, "dmu_open_snapshot(%s) = %d", snap2name, error); |
3492 | } | |
34dc7c2f | 3493 | |
428870ff | 3494 | error = dmu_objset_snapshot(clone1name, strchr(snap3name, '@')+1, |
572e2857 | 3495 | NULL, NULL, B_FALSE, B_FALSE, -1); |
428870ff BB |
3496 | if (error && error != EEXIST) { |
3497 | if (error == ENOSPC) { | |
3498 | ztest_record_enospc(FTAG); | |
3499 | goto out; | |
3500 | } | |
3501 | fatal(0, "dmu_open_snapshot(%s) = %d", snap3name, error); | |
3502 | } | |
34dc7c2f | 3503 | |
428870ff BB |
3504 | error = dmu_objset_hold(snap3name, FTAG, &clone); |
3505 | if (error) | |
3506 | fatal(0, "dmu_open_snapshot(%s) = %d", snap3name, error); | |
34dc7c2f | 3507 | |
428870ff BB |
3508 | error = dmu_objset_clone(clone2name, dmu_objset_ds(clone), 0); |
3509 | dmu_objset_rele(clone, FTAG); | |
3510 | if (error) { | |
3511 | if (error == ENOSPC) { | |
3512 | ztest_record_enospc(FTAG); | |
3513 | goto out; | |
3514 | } | |
3515 | fatal(0, "dmu_objset_create(%s) = %d", clone2name, error); | |
3516 | } | |
34dc7c2f | 3517 | |
428870ff BB |
3518 | error = dsl_dataset_own(snap2name, B_FALSE, FTAG, &ds); |
3519 | if (error) | |
3520 | fatal(0, "dsl_dataset_own(%s) = %d", snap2name, error); | |
3521 | error = dsl_dataset_promote(clone2name, NULL); | |
3522 | if (error != EBUSY) | |
3523 | fatal(0, "dsl_dataset_promote(%s), %d, not EBUSY", clone2name, | |
3524 | error); | |
3525 | dsl_dataset_disown(ds, FTAG); | |
34dc7c2f | 3526 | |
428870ff BB |
3527 | out: |
3528 | ztest_dsl_dataset_cleanup(osname, id); | |
34dc7c2f | 3529 | |
c242c188 | 3530 | (void) rw_exit(&ztest_name_lock); |
40b84e7a BB |
3531 | |
3532 | umem_free(snap1name, MAXNAMELEN); | |
3533 | umem_free(clone1name, MAXNAMELEN); | |
3534 | umem_free(snap2name, MAXNAMELEN); | |
3535 | umem_free(clone2name, MAXNAMELEN); | |
3536 | umem_free(snap3name, MAXNAMELEN); | |
428870ff | 3537 | } |
34dc7c2f | 3538 | |
40b84e7a BB |
3539 | #undef OD_ARRAY_SIZE |
3540 | #define OD_ARRAY_SIZE 4 | |
3541 | ||
428870ff BB |
3542 | /* |
3543 | * Verify that dmu_object_{alloc,free} work as expected. | |
3544 | */ | |
3545 | void | |
3546 | ztest_dmu_object_alloc_free(ztest_ds_t *zd, uint64_t id) | |
3547 | { | |
40b84e7a BB |
3548 | ztest_od_t *od; |
3549 | int batchsize; | |
3550 | int size; | |
d6320ddb | 3551 | int b; |
34dc7c2f | 3552 | |
40b84e7a BB |
3553 | size = sizeof(ztest_od_t) * OD_ARRAY_SIZE; |
3554 | od = umem_alloc(size, UMEM_NOFAIL); | |
3555 | batchsize = OD_ARRAY_SIZE; | |
3556 | ||
d6320ddb | 3557 | for (b = 0; b < batchsize; b++) |
40b84e7a | 3558 | ztest_od_init(od + b, id, FTAG, b, DMU_OT_UINT64_OTHER, 0, 0); |
34dc7c2f | 3559 | |
428870ff BB |
3560 | /* |
3561 | * Destroy the previous batch of objects, create a new batch, | |
3562 | * and do some I/O on the new objects. | |
3563 | */ | |
40b84e7a | 3564 | if (ztest_object_init(zd, od, size, B_TRUE) != 0) |
428870ff | 3565 | return; |
34dc7c2f | 3566 | |
428870ff BB |
3567 | while (ztest_random(4 * batchsize) != 0) |
3568 | ztest_io(zd, od[ztest_random(batchsize)].od_object, | |
3569 | ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
40b84e7a BB |
3570 | |
3571 | umem_free(od, size); | |
34dc7c2f BB |
3572 | } |
3573 | ||
40b84e7a BB |
3574 | #undef OD_ARRAY_SIZE |
3575 | #define OD_ARRAY_SIZE 2 | |
3576 | ||
34dc7c2f BB |
3577 | /* |
3578 | * Verify that dmu_{read,write} work as expected. | |
3579 | */ | |
34dc7c2f | 3580 | void |
428870ff | 3581 | ztest_dmu_read_write(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3582 | { |
40b84e7a BB |
3583 | int size; |
3584 | ztest_od_t *od; | |
3585 | ||
428870ff | 3586 | objset_t *os = zd->zd_os; |
40b84e7a BB |
3587 | size = sizeof(ztest_od_t) * OD_ARRAY_SIZE; |
3588 | od = umem_alloc(size, UMEM_NOFAIL); | |
34dc7c2f BB |
3589 | dmu_tx_t *tx; |
3590 | int i, freeit, error; | |
3591 | uint64_t n, s, txg; | |
3592 | bufwad_t *packbuf, *bigbuf, *pack, *bigH, *bigT; | |
428870ff BB |
3593 | uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize; |
3594 | uint64_t chunksize = (1000 + ztest_random(1000)) * sizeof (uint64_t); | |
34dc7c2f BB |
3595 | uint64_t regions = 997; |
3596 | uint64_t stride = 123456789ULL; | |
3597 | uint64_t width = 40; | |
3598 | int free_percent = 5; | |
3599 | ||
3600 | /* | |
3601 | * This test uses two objects, packobj and bigobj, that are always | |
3602 | * updated together (i.e. in the same tx) so that their contents are | |
3603 | * in sync and can be compared. Their contents relate to each other | |
3604 | * in a simple way: packobj is a dense array of 'bufwad' structures, | |
3605 | * while bigobj is a sparse array of the same bufwads. Specifically, | |
3606 | * for any index n, there are three bufwads that should be identical: | |
3607 | * | |
3608 | * packobj, at offset n * sizeof (bufwad_t) | |
3609 | * bigobj, at the head of the nth chunk | |
3610 | * bigobj, at the tail of the nth chunk | |
3611 | * | |
3612 | * The chunk size is arbitrary. It doesn't have to be a power of two, | |
3613 | * and it doesn't have any relation to the object blocksize. | |
3614 | * The only requirement is that it can hold at least two bufwads. | |
3615 | * | |
3616 | * Normally, we write the bufwad to each of these locations. | |
3617 | * However, free_percent of the time we instead write zeroes to | |
3618 | * packobj and perform a dmu_free_range() on bigobj. By comparing | |
3619 | * bigobj to packobj, we can verify that the DMU is correctly | |
3620 | * tracking which parts of an object are allocated and free, | |
3621 | * and that the contents of the allocated blocks are correct. | |
3622 | */ | |
3623 | ||
3624 | /* | |
3625 | * Read the directory info. If it's the first time, set things up. | |
3626 | */ | |
40b84e7a BB |
3627 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, chunksize); |
3628 | ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, chunksize); | |
34dc7c2f | 3629 | |
40b84e7a BB |
3630 | if (ztest_object_init(zd, od, size, B_FALSE) != 0) { |
3631 | umem_free(od, size); | |
428870ff | 3632 | return; |
40b84e7a | 3633 | } |
34dc7c2f | 3634 | |
428870ff BB |
3635 | bigobj = od[0].od_object; |
3636 | packobj = od[1].od_object; | |
3637 | chunksize = od[0].od_gen; | |
3638 | ASSERT(chunksize == od[1].od_gen); | |
34dc7c2f BB |
3639 | |
3640 | /* | |
3641 | * Prefetch a random chunk of the big object. | |
3642 | * Our aim here is to get some async reads in flight | |
3643 | * for blocks that we may free below; the DMU should | |
3644 | * handle this race correctly. | |
3645 | */ | |
3646 | n = ztest_random(regions) * stride + ztest_random(width); | |
3647 | s = 1 + ztest_random(2 * width - 1); | |
428870ff | 3648 | dmu_prefetch(os, bigobj, n * chunksize, s * chunksize); |
34dc7c2f BB |
3649 | |
3650 | /* | |
3651 | * Pick a random index and compute the offsets into packobj and bigobj. | |
3652 | */ | |
3653 | n = ztest_random(regions) * stride + ztest_random(width); | |
3654 | s = 1 + ztest_random(width - 1); | |
3655 | ||
3656 | packoff = n * sizeof (bufwad_t); | |
3657 | packsize = s * sizeof (bufwad_t); | |
3658 | ||
428870ff BB |
3659 | bigoff = n * chunksize; |
3660 | bigsize = s * chunksize; | |
34dc7c2f BB |
3661 | |
3662 | packbuf = umem_alloc(packsize, UMEM_NOFAIL); | |
3663 | bigbuf = umem_alloc(bigsize, UMEM_NOFAIL); | |
3664 | ||
3665 | /* | |
3666 | * free_percent of the time, free a range of bigobj rather than | |
3667 | * overwriting it. | |
3668 | */ | |
3669 | freeit = (ztest_random(100) < free_percent); | |
3670 | ||
3671 | /* | |
3672 | * Read the current contents of our objects. | |
3673 | */ | |
428870ff | 3674 | error = dmu_read(os, packobj, packoff, packsize, packbuf, |
9babb374 | 3675 | DMU_READ_PREFETCH); |
c99c9001 | 3676 | ASSERT0(error); |
428870ff | 3677 | error = dmu_read(os, bigobj, bigoff, bigsize, bigbuf, |
9babb374 | 3678 | DMU_READ_PREFETCH); |
c99c9001 | 3679 | ASSERT0(error); |
34dc7c2f BB |
3680 | |
3681 | /* | |
3682 | * Get a tx for the mods to both packobj and bigobj. | |
3683 | */ | |
3684 | tx = dmu_tx_create(os); | |
3685 | ||
428870ff | 3686 | dmu_tx_hold_write(tx, packobj, packoff, packsize); |
34dc7c2f BB |
3687 | |
3688 | if (freeit) | |
428870ff | 3689 | dmu_tx_hold_free(tx, bigobj, bigoff, bigsize); |
34dc7c2f | 3690 | else |
428870ff | 3691 | dmu_tx_hold_write(tx, bigobj, bigoff, bigsize); |
34dc7c2f | 3692 | |
428870ff BB |
3693 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); |
3694 | if (txg == 0) { | |
34dc7c2f BB |
3695 | umem_free(packbuf, packsize); |
3696 | umem_free(bigbuf, bigsize); | |
40b84e7a | 3697 | umem_free(od, size); |
34dc7c2f BB |
3698 | return; |
3699 | } | |
3700 | ||
428870ff BB |
3701 | dmu_object_set_checksum(os, bigobj, |
3702 | (enum zio_checksum)ztest_random_dsl_prop(ZFS_PROP_CHECKSUM), tx); | |
3703 | ||
3704 | dmu_object_set_compress(os, bigobj, | |
3705 | (enum zio_compress)ztest_random_dsl_prop(ZFS_PROP_COMPRESSION), tx); | |
34dc7c2f BB |
3706 | |
3707 | /* | |
3708 | * For each index from n to n + s, verify that the existing bufwad | |
3709 | * in packobj matches the bufwads at the head and tail of the | |
3710 | * corresponding chunk in bigobj. Then update all three bufwads | |
3711 | * with the new values we want to write out. | |
3712 | */ | |
3713 | for (i = 0; i < s; i++) { | |
3714 | /* LINTED */ | |
3715 | pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t)); | |
3716 | /* LINTED */ | |
428870ff | 3717 | bigH = (bufwad_t *)((char *)bigbuf + i * chunksize); |
34dc7c2f | 3718 | /* LINTED */ |
428870ff | 3719 | bigT = (bufwad_t *)((char *)bigH + chunksize) - 1; |
34dc7c2f BB |
3720 | |
3721 | ASSERT((uintptr_t)bigH - (uintptr_t)bigbuf < bigsize); | |
3722 | ASSERT((uintptr_t)bigT - (uintptr_t)bigbuf < bigsize); | |
3723 | ||
3724 | if (pack->bw_txg > txg) | |
3725 | fatal(0, "future leak: got %llx, open txg is %llx", | |
3726 | pack->bw_txg, txg); | |
3727 | ||
3728 | if (pack->bw_data != 0 && pack->bw_index != n + i) | |
3729 | fatal(0, "wrong index: got %llx, wanted %llx+%llx", | |
3730 | pack->bw_index, n, i); | |
3731 | ||
3732 | if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0) | |
3733 | fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH); | |
3734 | ||
3735 | if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0) | |
3736 | fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT); | |
3737 | ||
3738 | if (freeit) { | |
3739 | bzero(pack, sizeof (bufwad_t)); | |
3740 | } else { | |
3741 | pack->bw_index = n + i; | |
3742 | pack->bw_txg = txg; | |
3743 | pack->bw_data = 1 + ztest_random(-2ULL); | |
3744 | } | |
3745 | *bigH = *pack; | |
3746 | *bigT = *pack; | |
3747 | } | |
3748 | ||
3749 | /* | |
3750 | * We've verified all the old bufwads, and made new ones. | |
3751 | * Now write them out. | |
3752 | */ | |
428870ff | 3753 | dmu_write(os, packobj, packoff, packsize, packbuf, tx); |
34dc7c2f BB |
3754 | |
3755 | if (freeit) { | |
c242c188 | 3756 | if (ztest_opts.zo_verbose >= 7) { |
34dc7c2f BB |
3757 | (void) printf("freeing offset %llx size %llx" |
3758 | " txg %llx\n", | |
3759 | (u_longlong_t)bigoff, | |
3760 | (u_longlong_t)bigsize, | |
3761 | (u_longlong_t)txg); | |
3762 | } | |
428870ff | 3763 | VERIFY(0 == dmu_free_range(os, bigobj, bigoff, bigsize, tx)); |
34dc7c2f | 3764 | } else { |
c242c188 | 3765 | if (ztest_opts.zo_verbose >= 7) { |
34dc7c2f BB |
3766 | (void) printf("writing offset %llx size %llx" |
3767 | " txg %llx\n", | |
3768 | (u_longlong_t)bigoff, | |
3769 | (u_longlong_t)bigsize, | |
3770 | (u_longlong_t)txg); | |
3771 | } | |
428870ff | 3772 | dmu_write(os, bigobj, bigoff, bigsize, bigbuf, tx); |
34dc7c2f BB |
3773 | } |
3774 | ||
3775 | dmu_tx_commit(tx); | |
3776 | ||
3777 | /* | |
3778 | * Sanity check the stuff we just wrote. | |
3779 | */ | |
3780 | { | |
3781 | void *packcheck = umem_alloc(packsize, UMEM_NOFAIL); | |
3782 | void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL); | |
3783 | ||
428870ff | 3784 | VERIFY(0 == dmu_read(os, packobj, packoff, |
9babb374 | 3785 | packsize, packcheck, DMU_READ_PREFETCH)); |
428870ff | 3786 | VERIFY(0 == dmu_read(os, bigobj, bigoff, |
9babb374 | 3787 | bigsize, bigcheck, DMU_READ_PREFETCH)); |
34dc7c2f BB |
3788 | |
3789 | ASSERT(bcmp(packbuf, packcheck, packsize) == 0); | |
3790 | ASSERT(bcmp(bigbuf, bigcheck, bigsize) == 0); | |
3791 | ||
3792 | umem_free(packcheck, packsize); | |
3793 | umem_free(bigcheck, bigsize); | |
3794 | } | |
3795 | ||
3796 | umem_free(packbuf, packsize); | |
3797 | umem_free(bigbuf, bigsize); | |
40b84e7a | 3798 | umem_free(od, size); |
34dc7c2f BB |
3799 | } |
3800 | ||
9babb374 BB |
3801 | void |
3802 | compare_and_update_pbbufs(uint64_t s, bufwad_t *packbuf, bufwad_t *bigbuf, | |
428870ff | 3803 | uint64_t bigsize, uint64_t n, uint64_t chunksize, uint64_t txg) |
9babb374 BB |
3804 | { |
3805 | uint64_t i; | |
3806 | bufwad_t *pack; | |
3807 | bufwad_t *bigH; | |
3808 | bufwad_t *bigT; | |
3809 | ||
3810 | /* | |
3811 | * For each index from n to n + s, verify that the existing bufwad | |
3812 | * in packobj matches the bufwads at the head and tail of the | |
3813 | * corresponding chunk in bigobj. Then update all three bufwads | |
3814 | * with the new values we want to write out. | |
3815 | */ | |
3816 | for (i = 0; i < s; i++) { | |
3817 | /* LINTED */ | |
3818 | pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t)); | |
3819 | /* LINTED */ | |
428870ff | 3820 | bigH = (bufwad_t *)((char *)bigbuf + i * chunksize); |
9babb374 | 3821 | /* LINTED */ |
428870ff | 3822 | bigT = (bufwad_t *)((char *)bigH + chunksize) - 1; |
9babb374 BB |
3823 | |
3824 | ASSERT((uintptr_t)bigH - (uintptr_t)bigbuf < bigsize); | |
3825 | ASSERT((uintptr_t)bigT - (uintptr_t)bigbuf < bigsize); | |
3826 | ||
3827 | if (pack->bw_txg > txg) | |
3828 | fatal(0, "future leak: got %llx, open txg is %llx", | |
3829 | pack->bw_txg, txg); | |
3830 | ||
3831 | if (pack->bw_data != 0 && pack->bw_index != n + i) | |
3832 | fatal(0, "wrong index: got %llx, wanted %llx+%llx", | |
3833 | pack->bw_index, n, i); | |
3834 | ||
3835 | if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0) | |
3836 | fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH); | |
3837 | ||
3838 | if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0) | |
3839 | fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT); | |
3840 | ||
3841 | pack->bw_index = n + i; | |
3842 | pack->bw_txg = txg; | |
3843 | pack->bw_data = 1 + ztest_random(-2ULL); | |
3844 | ||
3845 | *bigH = *pack; | |
3846 | *bigT = *pack; | |
3847 | } | |
3848 | } | |
3849 | ||
40b84e7a BB |
3850 | #undef OD_ARRAY_SIZE |
3851 | #define OD_ARRAY_SIZE 2 | |
3852 | ||
9babb374 | 3853 | void |
428870ff | 3854 | ztest_dmu_read_write_zcopy(ztest_ds_t *zd, uint64_t id) |
9babb374 | 3855 | { |
428870ff | 3856 | objset_t *os = zd->zd_os; |
40b84e7a | 3857 | ztest_od_t *od; |
9babb374 BB |
3858 | dmu_tx_t *tx; |
3859 | uint64_t i; | |
3860 | int error; | |
40b84e7a | 3861 | int size; |
9babb374 BB |
3862 | uint64_t n, s, txg; |
3863 | bufwad_t *packbuf, *bigbuf; | |
428870ff BB |
3864 | uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize; |
3865 | uint64_t blocksize = ztest_random_blocksize(); | |
3866 | uint64_t chunksize = blocksize; | |
9babb374 BB |
3867 | uint64_t regions = 997; |
3868 | uint64_t stride = 123456789ULL; | |
3869 | uint64_t width = 9; | |
3870 | dmu_buf_t *bonus_db; | |
3871 | arc_buf_t **bigbuf_arcbufs; | |
428870ff | 3872 | dmu_object_info_t doi; |
9babb374 | 3873 | |
40b84e7a BB |
3874 | size = sizeof(ztest_od_t) * OD_ARRAY_SIZE; |
3875 | od = umem_alloc(size, UMEM_NOFAIL); | |
3876 | ||
9babb374 BB |
3877 | /* |
3878 | * This test uses two objects, packobj and bigobj, that are always | |
3879 | * updated together (i.e. in the same tx) so that their contents are | |
3880 | * in sync and can be compared. Their contents relate to each other | |
3881 | * in a simple way: packobj is a dense array of 'bufwad' structures, | |
3882 | * while bigobj is a sparse array of the same bufwads. Specifically, | |
3883 | * for any index n, there are three bufwads that should be identical: | |
3884 | * | |
3885 | * packobj, at offset n * sizeof (bufwad_t) | |
3886 | * bigobj, at the head of the nth chunk | |
3887 | * bigobj, at the tail of the nth chunk | |
3888 | * | |
3889 | * The chunk size is set equal to bigobj block size so that | |
3890 | * dmu_assign_arcbuf() can be tested for object updates. | |
3891 | */ | |
3892 | ||
3893 | /* | |
3894 | * Read the directory info. If it's the first time, set things up. | |
3895 | */ | |
40b84e7a BB |
3896 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0); |
3897 | ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, chunksize); | |
3898 | ||
9babb374 | 3899 | |
40b84e7a BB |
3900 | if (ztest_object_init(zd, od, size, B_FALSE) != 0) { |
3901 | umem_free(od, size); | |
428870ff | 3902 | return; |
40b84e7a | 3903 | } |
9babb374 | 3904 | |
428870ff BB |
3905 | bigobj = od[0].od_object; |
3906 | packobj = od[1].od_object; | |
3907 | blocksize = od[0].od_blocksize; | |
3908 | chunksize = blocksize; | |
3909 | ASSERT(chunksize == od[1].od_gen); | |
9babb374 | 3910 | |
428870ff BB |
3911 | VERIFY(dmu_object_info(os, bigobj, &doi) == 0); |
3912 | VERIFY(ISP2(doi.doi_data_block_size)); | |
3913 | VERIFY(chunksize == doi.doi_data_block_size); | |
3914 | VERIFY(chunksize >= 2 * sizeof (bufwad_t)); | |
9babb374 BB |
3915 | |
3916 | /* | |
3917 | * Pick a random index and compute the offsets into packobj and bigobj. | |
3918 | */ | |
3919 | n = ztest_random(regions) * stride + ztest_random(width); | |
3920 | s = 1 + ztest_random(width - 1); | |
3921 | ||
3922 | packoff = n * sizeof (bufwad_t); | |
3923 | packsize = s * sizeof (bufwad_t); | |
3924 | ||
428870ff BB |
3925 | bigoff = n * chunksize; |
3926 | bigsize = s * chunksize; | |
9babb374 BB |
3927 | |
3928 | packbuf = umem_zalloc(packsize, UMEM_NOFAIL); | |
3929 | bigbuf = umem_zalloc(bigsize, UMEM_NOFAIL); | |
3930 | ||
428870ff | 3931 | VERIFY3U(0, ==, dmu_bonus_hold(os, bigobj, FTAG, &bonus_db)); |
9babb374 BB |
3932 | |
3933 | bigbuf_arcbufs = umem_zalloc(2 * s * sizeof (arc_buf_t *), UMEM_NOFAIL); | |
3934 | ||
3935 | /* | |
3936 | * Iteration 0 test zcopy for DB_UNCACHED dbufs. | |
3937 | * Iteration 1 test zcopy to already referenced dbufs. | |
3938 | * Iteration 2 test zcopy to dirty dbuf in the same txg. | |
3939 | * Iteration 3 test zcopy to dbuf dirty in previous txg. | |
3940 | * Iteration 4 test zcopy when dbuf is no longer dirty. | |
3941 | * Iteration 5 test zcopy when it can't be done. | |
3942 | * Iteration 6 one more zcopy write. | |
3943 | */ | |
3944 | for (i = 0; i < 7; i++) { | |
3945 | uint64_t j; | |
3946 | uint64_t off; | |
3947 | ||
3948 | /* | |
3949 | * In iteration 5 (i == 5) use arcbufs | |
3950 | * that don't match bigobj blksz to test | |
3951 | * dmu_assign_arcbuf() when it can't directly | |
3952 | * assign an arcbuf to a dbuf. | |
3953 | */ | |
3954 | for (j = 0; j < s; j++) { | |
3955 | if (i != 5) { | |
3956 | bigbuf_arcbufs[j] = | |
428870ff | 3957 | dmu_request_arcbuf(bonus_db, chunksize); |
9babb374 BB |
3958 | } else { |
3959 | bigbuf_arcbufs[2 * j] = | |
428870ff | 3960 | dmu_request_arcbuf(bonus_db, chunksize / 2); |
9babb374 | 3961 | bigbuf_arcbufs[2 * j + 1] = |
428870ff | 3962 | dmu_request_arcbuf(bonus_db, chunksize / 2); |
9babb374 BB |
3963 | } |
3964 | } | |
3965 | ||
3966 | /* | |
3967 | * Get a tx for the mods to both packobj and bigobj. | |
3968 | */ | |
3969 | tx = dmu_tx_create(os); | |
3970 | ||
428870ff BB |
3971 | dmu_tx_hold_write(tx, packobj, packoff, packsize); |
3972 | dmu_tx_hold_write(tx, bigobj, bigoff, bigsize); | |
9babb374 | 3973 | |
428870ff BB |
3974 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); |
3975 | if (txg == 0) { | |
9babb374 BB |
3976 | umem_free(packbuf, packsize); |
3977 | umem_free(bigbuf, bigsize); | |
3978 | for (j = 0; j < s; j++) { | |
3979 | if (i != 5) { | |
3980 | dmu_return_arcbuf(bigbuf_arcbufs[j]); | |
3981 | } else { | |
3982 | dmu_return_arcbuf( | |
3983 | bigbuf_arcbufs[2 * j]); | |
3984 | dmu_return_arcbuf( | |
3985 | bigbuf_arcbufs[2 * j + 1]); | |
3986 | } | |
3987 | } | |
3988 | umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *)); | |
40b84e7a | 3989 | umem_free(od, size); |
9babb374 BB |
3990 | dmu_buf_rele(bonus_db, FTAG); |
3991 | return; | |
3992 | } | |
3993 | ||
9babb374 BB |
3994 | /* |
3995 | * 50% of the time don't read objects in the 1st iteration to | |
3996 | * test dmu_assign_arcbuf() for the case when there're no | |
3997 | * existing dbufs for the specified offsets. | |
3998 | */ | |
3999 | if (i != 0 || ztest_random(2) != 0) { | |
428870ff | 4000 | error = dmu_read(os, packobj, packoff, |
9babb374 | 4001 | packsize, packbuf, DMU_READ_PREFETCH); |
c99c9001 | 4002 | ASSERT0(error); |
428870ff | 4003 | error = dmu_read(os, bigobj, bigoff, bigsize, |
9babb374 | 4004 | bigbuf, DMU_READ_PREFETCH); |
c99c9001 | 4005 | ASSERT0(error); |
9babb374 BB |
4006 | } |
4007 | compare_and_update_pbbufs(s, packbuf, bigbuf, bigsize, | |
428870ff | 4008 | n, chunksize, txg); |
9babb374 BB |
4009 | |
4010 | /* | |
4011 | * We've verified all the old bufwads, and made new ones. | |
4012 | * Now write them out. | |
4013 | */ | |
428870ff | 4014 | dmu_write(os, packobj, packoff, packsize, packbuf, tx); |
c242c188 | 4015 | if (ztest_opts.zo_verbose >= 7) { |
9babb374 BB |
4016 | (void) printf("writing offset %llx size %llx" |
4017 | " txg %llx\n", | |
4018 | (u_longlong_t)bigoff, | |
4019 | (u_longlong_t)bigsize, | |
4020 | (u_longlong_t)txg); | |
4021 | } | |
428870ff | 4022 | for (off = bigoff, j = 0; j < s; j++, off += chunksize) { |
9babb374 BB |
4023 | dmu_buf_t *dbt; |
4024 | if (i != 5) { | |
4025 | bcopy((caddr_t)bigbuf + (off - bigoff), | |
428870ff | 4026 | bigbuf_arcbufs[j]->b_data, chunksize); |
9babb374 BB |
4027 | } else { |
4028 | bcopy((caddr_t)bigbuf + (off - bigoff), | |
4029 | bigbuf_arcbufs[2 * j]->b_data, | |
428870ff | 4030 | chunksize / 2); |
9babb374 | 4031 | bcopy((caddr_t)bigbuf + (off - bigoff) + |
428870ff | 4032 | chunksize / 2, |
9babb374 | 4033 | bigbuf_arcbufs[2 * j + 1]->b_data, |
428870ff | 4034 | chunksize / 2); |
9babb374 BB |
4035 | } |
4036 | ||
4037 | if (i == 1) { | |
428870ff BB |
4038 | VERIFY(dmu_buf_hold(os, bigobj, off, |
4039 | FTAG, &dbt, DMU_READ_NO_PREFETCH) == 0); | |
9babb374 BB |
4040 | } |
4041 | if (i != 5) { | |
4042 | dmu_assign_arcbuf(bonus_db, off, | |
4043 | bigbuf_arcbufs[j], tx); | |
4044 | } else { | |
4045 | dmu_assign_arcbuf(bonus_db, off, | |
4046 | bigbuf_arcbufs[2 * j], tx); | |
4047 | dmu_assign_arcbuf(bonus_db, | |
428870ff | 4048 | off + chunksize / 2, |
9babb374 BB |
4049 | bigbuf_arcbufs[2 * j + 1], tx); |
4050 | } | |
4051 | if (i == 1) { | |
4052 | dmu_buf_rele(dbt, FTAG); | |
4053 | } | |
4054 | } | |
4055 | dmu_tx_commit(tx); | |
4056 | ||
4057 | /* | |
4058 | * Sanity check the stuff we just wrote. | |
4059 | */ | |
4060 | { | |
4061 | void *packcheck = umem_alloc(packsize, UMEM_NOFAIL); | |
4062 | void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL); | |
4063 | ||
428870ff | 4064 | VERIFY(0 == dmu_read(os, packobj, packoff, |
9babb374 | 4065 | packsize, packcheck, DMU_READ_PREFETCH)); |
428870ff | 4066 | VERIFY(0 == dmu_read(os, bigobj, bigoff, |
9babb374 BB |
4067 | bigsize, bigcheck, DMU_READ_PREFETCH)); |
4068 | ||
4069 | ASSERT(bcmp(packbuf, packcheck, packsize) == 0); | |
4070 | ASSERT(bcmp(bigbuf, bigcheck, bigsize) == 0); | |
4071 | ||
4072 | umem_free(packcheck, packsize); | |
4073 | umem_free(bigcheck, bigsize); | |
4074 | } | |
4075 | if (i == 2) { | |
4076 | txg_wait_open(dmu_objset_pool(os), 0); | |
4077 | } else if (i == 3) { | |
4078 | txg_wait_synced(dmu_objset_pool(os), 0); | |
4079 | } | |
4080 | } | |
4081 | ||
4082 | dmu_buf_rele(bonus_db, FTAG); | |
4083 | umem_free(packbuf, packsize); | |
4084 | umem_free(bigbuf, bigsize); | |
4085 | umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *)); | |
40b84e7a | 4086 | umem_free(od, size); |
9babb374 BB |
4087 | } |
4088 | ||
428870ff | 4089 | /* ARGSUSED */ |
34dc7c2f | 4090 | void |
428870ff | 4091 | ztest_dmu_write_parallel(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4092 | { |
40b84e7a BB |
4093 | ztest_od_t *od; |
4094 | ||
4095 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); | |
428870ff BB |
4096 | uint64_t offset = (1ULL << (ztest_random(20) + 43)) + |
4097 | (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
34dc7c2f BB |
4098 | |
4099 | /* | |
428870ff BB |
4100 | * Have multiple threads write to large offsets in an object |
4101 | * to verify that parallel writes to an object -- even to the | |
4102 | * same blocks within the object -- doesn't cause any trouble. | |
34dc7c2f | 4103 | */ |
40b84e7a | 4104 | ztest_od_init(od, ID_PARALLEL, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0); |
9babb374 | 4105 | |
40b84e7a | 4106 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) |
34dc7c2f | 4107 | return; |
34dc7c2f | 4108 | |
428870ff | 4109 | while (ztest_random(10) != 0) |
40b84e7a BB |
4110 | ztest_io(zd, od->od_object, offset); |
4111 | ||
4112 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff | 4113 | } |
34dc7c2f | 4114 | |
428870ff BB |
4115 | void |
4116 | ztest_dmu_prealloc(ztest_ds_t *zd, uint64_t id) | |
4117 | { | |
40b84e7a | 4118 | ztest_od_t *od; |
428870ff BB |
4119 | uint64_t offset = (1ULL << (ztest_random(4) + SPA_MAXBLOCKSHIFT)) + |
4120 | (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
4121 | uint64_t count = ztest_random(20) + 1; | |
4122 | uint64_t blocksize = ztest_random_blocksize(); | |
4123 | void *data; | |
34dc7c2f | 4124 | |
40b84e7a | 4125 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
34dc7c2f | 4126 | |
40b84e7a BB |
4127 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0); |
4128 | ||
4129 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), !ztest_random(2)) != 0) { | |
4130 | umem_free(od, sizeof(ztest_od_t)); | |
34dc7c2f | 4131 | return; |
40b84e7a | 4132 | } |
34dc7c2f | 4133 | |
40b84e7a BB |
4134 | if (ztest_truncate(zd, od->od_object, offset, count * blocksize) != 0) { |
4135 | umem_free(od, sizeof(ztest_od_t)); | |
34dc7c2f | 4136 | return; |
40b84e7a | 4137 | } |
34dc7c2f | 4138 | |
40b84e7a | 4139 | ztest_prealloc(zd, od->od_object, offset, count * blocksize); |
34dc7c2f | 4140 | |
428870ff | 4141 | data = umem_zalloc(blocksize, UMEM_NOFAIL); |
34dc7c2f | 4142 | |
428870ff BB |
4143 | while (ztest_random(count) != 0) { |
4144 | uint64_t randoff = offset + (ztest_random(count) * blocksize); | |
40b84e7a | 4145 | if (ztest_write(zd, od->od_object, randoff, blocksize, |
428870ff BB |
4146 | data) != 0) |
4147 | break; | |
4148 | while (ztest_random(4) != 0) | |
40b84e7a | 4149 | ztest_io(zd, od->od_object, randoff); |
9babb374 | 4150 | } |
34dc7c2f | 4151 | |
428870ff | 4152 | umem_free(data, blocksize); |
40b84e7a | 4153 | umem_free(od, sizeof(ztest_od_t)); |
34dc7c2f BB |
4154 | } |
4155 | ||
4156 | /* | |
4157 | * Verify that zap_{create,destroy,add,remove,update} work as expected. | |
4158 | */ | |
4159 | #define ZTEST_ZAP_MIN_INTS 1 | |
4160 | #define ZTEST_ZAP_MAX_INTS 4 | |
4161 | #define ZTEST_ZAP_MAX_PROPS 1000 | |
4162 | ||
4163 | void | |
428870ff | 4164 | ztest_zap(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4165 | { |
428870ff | 4166 | objset_t *os = zd->zd_os; |
40b84e7a | 4167 | ztest_od_t *od; |
34dc7c2f BB |
4168 | uint64_t object; |
4169 | uint64_t txg, last_txg; | |
4170 | uint64_t value[ZTEST_ZAP_MAX_INTS]; | |
4171 | uint64_t zl_ints, zl_intsize, prop; | |
4172 | int i, ints; | |
4173 | dmu_tx_t *tx; | |
4174 | char propname[100], txgname[100]; | |
4175 | int error; | |
34dc7c2f BB |
4176 | char *hc[2] = { "s.acl.h", ".s.open.h.hyLZlg" }; |
4177 | ||
40b84e7a BB |
4178 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
4179 | ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0); | |
34dc7c2f | 4180 | |
40b84e7a BB |
4181 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
4182 | !ztest_random(2)) != 0) | |
4183 | goto out; | |
34dc7c2f | 4184 | |
40b84e7a | 4185 | object = od->od_object; |
34dc7c2f | 4186 | |
428870ff BB |
4187 | /* |
4188 | * Generate a known hash collision, and verify that | |
4189 | * we can lookup and remove both entries. | |
4190 | */ | |
4191 | tx = dmu_tx_create(os); | |
4192 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); | |
4193 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4194 | if (txg == 0) | |
40b84e7a | 4195 | goto out; |
428870ff BB |
4196 | for (i = 0; i < 2; i++) { |
4197 | value[i] = i; | |
4198 | VERIFY3U(0, ==, zap_add(os, object, hc[i], sizeof (uint64_t), | |
4199 | 1, &value[i], tx)); | |
4200 | } | |
4201 | for (i = 0; i < 2; i++) { | |
4202 | VERIFY3U(EEXIST, ==, zap_add(os, object, hc[i], | |
4203 | sizeof (uint64_t), 1, &value[i], tx)); | |
4204 | VERIFY3U(0, ==, | |
4205 | zap_length(os, object, hc[i], &zl_intsize, &zl_ints)); | |
4206 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
4207 | ASSERT3U(zl_ints, ==, 1); | |
4208 | } | |
4209 | for (i = 0; i < 2; i++) { | |
4210 | VERIFY3U(0, ==, zap_remove(os, object, hc[i], tx)); | |
34dc7c2f | 4211 | } |
428870ff | 4212 | dmu_tx_commit(tx); |
34dc7c2f | 4213 | |
428870ff BB |
4214 | /* |
4215 | * Generate a buch of random entries. | |
4216 | */ | |
34dc7c2f BB |
4217 | ints = MAX(ZTEST_ZAP_MIN_INTS, object % ZTEST_ZAP_MAX_INTS); |
4218 | ||
4219 | prop = ztest_random(ZTEST_ZAP_MAX_PROPS); | |
4220 | (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop); | |
4221 | (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop); | |
4222 | bzero(value, sizeof (value)); | |
4223 | last_txg = 0; | |
4224 | ||
4225 | /* | |
4226 | * If these zap entries already exist, validate their contents. | |
4227 | */ | |
4228 | error = zap_length(os, object, txgname, &zl_intsize, &zl_ints); | |
4229 | if (error == 0) { | |
4230 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
4231 | ASSERT3U(zl_ints, ==, 1); | |
4232 | ||
4233 | VERIFY(zap_lookup(os, object, txgname, zl_intsize, | |
4234 | zl_ints, &last_txg) == 0); | |
4235 | ||
4236 | VERIFY(zap_length(os, object, propname, &zl_intsize, | |
4237 | &zl_ints) == 0); | |
4238 | ||
4239 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
4240 | ASSERT3U(zl_ints, ==, ints); | |
4241 | ||
4242 | VERIFY(zap_lookup(os, object, propname, zl_intsize, | |
4243 | zl_ints, value) == 0); | |
4244 | ||
4245 | for (i = 0; i < ints; i++) { | |
4246 | ASSERT3U(value[i], ==, last_txg + object + i); | |
4247 | } | |
4248 | } else { | |
4249 | ASSERT3U(error, ==, ENOENT); | |
4250 | } | |
4251 | ||
4252 | /* | |
4253 | * Atomically update two entries in our zap object. | |
4254 | * The first is named txg_%llu, and contains the txg | |
4255 | * in which the property was last updated. The second | |
4256 | * is named prop_%llu, and the nth element of its value | |
4257 | * should be txg + object + n. | |
4258 | */ | |
4259 | tx = dmu_tx_create(os); | |
428870ff BB |
4260 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
4261 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4262 | if (txg == 0) | |
40b84e7a | 4263 | goto out; |
34dc7c2f BB |
4264 | |
4265 | if (last_txg > txg) | |
4266 | fatal(0, "zap future leak: old %llu new %llu", last_txg, txg); | |
4267 | ||
4268 | for (i = 0; i < ints; i++) | |
4269 | value[i] = txg + object + i; | |
4270 | ||
428870ff BB |
4271 | VERIFY3U(0, ==, zap_update(os, object, txgname, sizeof (uint64_t), |
4272 | 1, &txg, tx)); | |
4273 | VERIFY3U(0, ==, zap_update(os, object, propname, sizeof (uint64_t), | |
4274 | ints, value, tx)); | |
34dc7c2f BB |
4275 | |
4276 | dmu_tx_commit(tx); | |
4277 | ||
4278 | /* | |
4279 | * Remove a random pair of entries. | |
4280 | */ | |
4281 | prop = ztest_random(ZTEST_ZAP_MAX_PROPS); | |
4282 | (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop); | |
4283 | (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop); | |
4284 | ||
4285 | error = zap_length(os, object, txgname, &zl_intsize, &zl_ints); | |
4286 | ||
4287 | if (error == ENOENT) | |
40b84e7a | 4288 | goto out; |
34dc7c2f | 4289 | |
c99c9001 | 4290 | ASSERT0(error); |
34dc7c2f BB |
4291 | |
4292 | tx = dmu_tx_create(os); | |
428870ff BB |
4293 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
4294 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4295 | if (txg == 0) | |
40b84e7a | 4296 | goto out; |
428870ff BB |
4297 | VERIFY3U(0, ==, zap_remove(os, object, txgname, tx)); |
4298 | VERIFY3U(0, ==, zap_remove(os, object, propname, tx)); | |
4299 | dmu_tx_commit(tx); | |
40b84e7a BB |
4300 | out: |
4301 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff | 4302 | } |
34dc7c2f | 4303 | |
428870ff BB |
4304 | /* |
4305 | * Testcase to test the upgrading of a microzap to fatzap. | |
4306 | */ | |
4307 | void | |
4308 | ztest_fzap(ztest_ds_t *zd, uint64_t id) | |
4309 | { | |
4310 | objset_t *os = zd->zd_os; | |
40b84e7a | 4311 | ztest_od_t *od; |
428870ff | 4312 | uint64_t object, txg; |
d6320ddb | 4313 | int i; |
34dc7c2f | 4314 | |
40b84e7a BB |
4315 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
4316 | ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0); | |
428870ff | 4317 | |
40b84e7a BB |
4318 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
4319 | !ztest_random(2)) != 0) | |
4320 | goto out; | |
4321 | object = od->od_object; | |
34dc7c2f BB |
4322 | |
4323 | /* | |
428870ff BB |
4324 | * Add entries to this ZAP and make sure it spills over |
4325 | * and gets upgraded to a fatzap. Also, since we are adding | |
4326 | * 2050 entries we should see ptrtbl growth and leaf-block split. | |
34dc7c2f | 4327 | */ |
d6320ddb | 4328 | for (i = 0; i < 2050; i++) { |
428870ff BB |
4329 | char name[MAXNAMELEN]; |
4330 | uint64_t value = i; | |
4331 | dmu_tx_t *tx; | |
4332 | int error; | |
34dc7c2f | 4333 | |
428870ff | 4334 | (void) snprintf(name, sizeof (name), "fzap-%llu-%llu", |
b8864a23 | 4335 | (u_longlong_t)id, (u_longlong_t)value); |
428870ff BB |
4336 | |
4337 | tx = dmu_tx_create(os); | |
4338 | dmu_tx_hold_zap(tx, object, B_TRUE, name); | |
4339 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4340 | if (txg == 0) | |
40b84e7a | 4341 | goto out; |
428870ff BB |
4342 | error = zap_add(os, object, name, sizeof (uint64_t), 1, |
4343 | &value, tx); | |
4344 | ASSERT(error == 0 || error == EEXIST); | |
4345 | dmu_tx_commit(tx); | |
34dc7c2f | 4346 | } |
40b84e7a BB |
4347 | out: |
4348 | umem_free(od, sizeof(ztest_od_t)); | |
34dc7c2f BB |
4349 | } |
4350 | ||
428870ff | 4351 | /* ARGSUSED */ |
34dc7c2f | 4352 | void |
428870ff | 4353 | ztest_zap_parallel(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4354 | { |
428870ff | 4355 | objset_t *os = zd->zd_os; |
40b84e7a | 4356 | ztest_od_t *od; |
34dc7c2f BB |
4357 | uint64_t txg, object, count, wsize, wc, zl_wsize, zl_wc; |
4358 | dmu_tx_t *tx; | |
4359 | int i, namelen, error; | |
428870ff | 4360 | int micro = ztest_random(2); |
34dc7c2f BB |
4361 | char name[20], string_value[20]; |
4362 | void *data; | |
4363 | ||
40b84e7a BB |
4364 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
4365 | ztest_od_init(od, ID_PARALLEL, FTAG, micro, DMU_OT_ZAP_OTHER, 0, 0); | |
428870ff | 4366 | |
40b84e7a BB |
4367 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
4368 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff | 4369 | return; |
40b84e7a | 4370 | } |
428870ff | 4371 | |
40b84e7a | 4372 | object = od->od_object; |
428870ff | 4373 | |
34dc7c2f BB |
4374 | /* |
4375 | * Generate a random name of the form 'xxx.....' where each | |
4376 | * x is a random printable character and the dots are dots. | |
4377 | * There are 94 such characters, and the name length goes from | |
4378 | * 6 to 20, so there are 94^3 * 15 = 12,458,760 possible names. | |
4379 | */ | |
4380 | namelen = ztest_random(sizeof (name) - 5) + 5 + 1; | |
4381 | ||
4382 | for (i = 0; i < 3; i++) | |
4383 | name[i] = '!' + ztest_random('~' - '!' + 1); | |
4384 | for (; i < namelen - 1; i++) | |
4385 | name[i] = '.'; | |
4386 | name[i] = '\0'; | |
4387 | ||
428870ff | 4388 | if ((namelen & 1) || micro) { |
34dc7c2f BB |
4389 | wsize = sizeof (txg); |
4390 | wc = 1; | |
4391 | data = &txg; | |
4392 | } else { | |
4393 | wsize = 1; | |
4394 | wc = namelen; | |
4395 | data = string_value; | |
4396 | } | |
4397 | ||
4398 | count = -1ULL; | |
4399 | VERIFY(zap_count(os, object, &count) == 0); | |
4400 | ASSERT(count != -1ULL); | |
4401 | ||
4402 | /* | |
4403 | * Select an operation: length, lookup, add, update, remove. | |
4404 | */ | |
4405 | i = ztest_random(5); | |
4406 | ||
4407 | if (i >= 2) { | |
4408 | tx = dmu_tx_create(os); | |
428870ff BB |
4409 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
4410 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4411 | if (txg == 0) | |
34dc7c2f | 4412 | return; |
34dc7c2f BB |
4413 | bcopy(name, string_value, namelen); |
4414 | } else { | |
4415 | tx = NULL; | |
4416 | txg = 0; | |
4417 | bzero(string_value, namelen); | |
4418 | } | |
4419 | ||
4420 | switch (i) { | |
4421 | ||
4422 | case 0: | |
4423 | error = zap_length(os, object, name, &zl_wsize, &zl_wc); | |
4424 | if (error == 0) { | |
4425 | ASSERT3U(wsize, ==, zl_wsize); | |
4426 | ASSERT3U(wc, ==, zl_wc); | |
4427 | } else { | |
4428 | ASSERT3U(error, ==, ENOENT); | |
4429 | } | |
4430 | break; | |
4431 | ||
4432 | case 1: | |
4433 | error = zap_lookup(os, object, name, wsize, wc, data); | |
4434 | if (error == 0) { | |
4435 | if (data == string_value && | |
4436 | bcmp(name, data, namelen) != 0) | |
4437 | fatal(0, "name '%s' != val '%s' len %d", | |
4438 | name, data, namelen); | |
4439 | } else { | |
4440 | ASSERT3U(error, ==, ENOENT); | |
4441 | } | |
4442 | break; | |
4443 | ||
4444 | case 2: | |
4445 | error = zap_add(os, object, name, wsize, wc, data, tx); | |
4446 | ASSERT(error == 0 || error == EEXIST); | |
4447 | break; | |
4448 | ||
4449 | case 3: | |
4450 | VERIFY(zap_update(os, object, name, wsize, wc, data, tx) == 0); | |
4451 | break; | |
4452 | ||
4453 | case 4: | |
4454 | error = zap_remove(os, object, name, tx); | |
4455 | ASSERT(error == 0 || error == ENOENT); | |
4456 | break; | |
4457 | } | |
4458 | ||
4459 | if (tx != NULL) | |
4460 | dmu_tx_commit(tx); | |
40b84e7a BB |
4461 | |
4462 | umem_free(od, sizeof(ztest_od_t)); | |
34dc7c2f BB |
4463 | } |
4464 | ||
428870ff BB |
4465 | /* |
4466 | * Commit callback data. | |
4467 | */ | |
4468 | typedef struct ztest_cb_data { | |
4469 | list_node_t zcd_node; | |
4470 | uint64_t zcd_txg; | |
4471 | int zcd_expected_err; | |
4472 | boolean_t zcd_added; | |
4473 | boolean_t zcd_called; | |
4474 | spa_t *zcd_spa; | |
4475 | } ztest_cb_data_t; | |
4476 | ||
4477 | /* This is the actual commit callback function */ | |
4478 | static void | |
4479 | ztest_commit_callback(void *arg, int error) | |
4480 | { | |
4481 | ztest_cb_data_t *data = arg; | |
4482 | uint64_t synced_txg; | |
4483 | ||
4484 | VERIFY(data != NULL); | |
4485 | VERIFY3S(data->zcd_expected_err, ==, error); | |
4486 | VERIFY(!data->zcd_called); | |
4487 | ||
4488 | synced_txg = spa_last_synced_txg(data->zcd_spa); | |
4489 | if (data->zcd_txg > synced_txg) | |
4490 | fatal(0, "commit callback of txg %" PRIu64 " called prematurely" | |
4491 | ", last synced txg = %" PRIu64 "\n", data->zcd_txg, | |
4492 | synced_txg); | |
4493 | ||
4494 | data->zcd_called = B_TRUE; | |
4495 | ||
4496 | if (error == ECANCELED) { | |
c99c9001 | 4497 | ASSERT0(data->zcd_txg); |
428870ff BB |
4498 | ASSERT(!data->zcd_added); |
4499 | ||
4500 | /* | |
4501 | * The private callback data should be destroyed here, but | |
4502 | * since we are going to check the zcd_called field after | |
4503 | * dmu_tx_abort(), we will destroy it there. | |
4504 | */ | |
4505 | return; | |
4506 | } | |
4507 | ||
090ff092 | 4508 | ASSERT(data->zcd_added); |
428870ff BB |
4509 | ASSERT3U(data->zcd_txg, !=, 0); |
4510 | ||
1e33ac1e | 4511 | (void) mutex_enter(&zcl.zcl_callbacks_lock); |
090ff092 RC |
4512 | |
4513 | /* See if this cb was called more quickly */ | |
4514 | if ((synced_txg - data->zcd_txg) < zc_min_txg_delay) | |
4515 | zc_min_txg_delay = synced_txg - data->zcd_txg; | |
4516 | ||
4517 | /* Remove our callback from the list */ | |
428870ff | 4518 | list_remove(&zcl.zcl_callbacks, data); |
090ff092 | 4519 | |
1e33ac1e | 4520 | (void) mutex_exit(&zcl.zcl_callbacks_lock); |
428870ff | 4521 | |
428870ff BB |
4522 | umem_free(data, sizeof (ztest_cb_data_t)); |
4523 | } | |
4524 | ||
4525 | /* Allocate and initialize callback data structure */ | |
4526 | static ztest_cb_data_t * | |
4527 | ztest_create_cb_data(objset_t *os, uint64_t txg) | |
4528 | { | |
4529 | ztest_cb_data_t *cb_data; | |
4530 | ||
4531 | cb_data = umem_zalloc(sizeof (ztest_cb_data_t), UMEM_NOFAIL); | |
4532 | ||
4533 | cb_data->zcd_txg = txg; | |
4534 | cb_data->zcd_spa = dmu_objset_spa(os); | |
1e33ac1e | 4535 | list_link_init(&cb_data->zcd_node); |
428870ff BB |
4536 | |
4537 | return (cb_data); | |
4538 | } | |
4539 | ||
428870ff BB |
4540 | /* |
4541 | * Commit callback test. | |
4542 | */ | |
34dc7c2f | 4543 | void |
428870ff BB |
4544 | ztest_dmu_commit_callbacks(ztest_ds_t *zd, uint64_t id) |
4545 | { | |
4546 | objset_t *os = zd->zd_os; | |
40b84e7a | 4547 | ztest_od_t *od; |
428870ff BB |
4548 | dmu_tx_t *tx; |
4549 | ztest_cb_data_t *cb_data[3], *tmp_cb; | |
4550 | uint64_t old_txg, txg; | |
090ff092 | 4551 | int i, error = 0; |
428870ff | 4552 | |
40b84e7a BB |
4553 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
4554 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0); | |
428870ff | 4555 | |
40b84e7a BB |
4556 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
4557 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff | 4558 | return; |
40b84e7a | 4559 | } |
428870ff BB |
4560 | |
4561 | tx = dmu_tx_create(os); | |
4562 | ||
4563 | cb_data[0] = ztest_create_cb_data(os, 0); | |
4564 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[0]); | |
4565 | ||
40b84e7a | 4566 | dmu_tx_hold_write(tx, od->od_object, 0, sizeof (uint64_t)); |
428870ff BB |
4567 | |
4568 | /* Every once in a while, abort the transaction on purpose */ | |
4569 | if (ztest_random(100) == 0) | |
4570 | error = -1; | |
4571 | ||
4572 | if (!error) | |
4573 | error = dmu_tx_assign(tx, TXG_NOWAIT); | |
4574 | ||
4575 | txg = error ? 0 : dmu_tx_get_txg(tx); | |
4576 | ||
4577 | cb_data[0]->zcd_txg = txg; | |
4578 | cb_data[1] = ztest_create_cb_data(os, txg); | |
4579 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[1]); | |
4580 | ||
4581 | if (error) { | |
4582 | /* | |
4583 | * It's not a strict requirement to call the registered | |
4584 | * callbacks from inside dmu_tx_abort(), but that's what | |
4585 | * it's supposed to happen in the current implementation | |
4586 | * so we will check for that. | |
4587 | */ | |
4588 | for (i = 0; i < 2; i++) { | |
4589 | cb_data[i]->zcd_expected_err = ECANCELED; | |
4590 | VERIFY(!cb_data[i]->zcd_called); | |
4591 | } | |
4592 | ||
4593 | dmu_tx_abort(tx); | |
4594 | ||
4595 | for (i = 0; i < 2; i++) { | |
4596 | VERIFY(cb_data[i]->zcd_called); | |
4597 | umem_free(cb_data[i], sizeof (ztest_cb_data_t)); | |
4598 | } | |
4599 | ||
40b84e7a | 4600 | umem_free(od, sizeof(ztest_od_t)); |
428870ff BB |
4601 | return; |
4602 | } | |
4603 | ||
4604 | cb_data[2] = ztest_create_cb_data(os, txg); | |
4605 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[2]); | |
4606 | ||
4607 | /* | |
4608 | * Read existing data to make sure there isn't a future leak. | |
4609 | */ | |
40b84e7a | 4610 | VERIFY(0 == dmu_read(os, od->od_object, 0, sizeof (uint64_t), |
428870ff BB |
4611 | &old_txg, DMU_READ_PREFETCH)); |
4612 | ||
4613 | if (old_txg > txg) | |
4614 | fatal(0, "future leak: got %" PRIu64 ", open txg is %" PRIu64, | |
4615 | old_txg, txg); | |
4616 | ||
40b84e7a | 4617 | dmu_write(os, od->od_object, 0, sizeof (uint64_t), &txg, tx); |
428870ff | 4618 | |
1e33ac1e | 4619 | (void) mutex_enter(&zcl.zcl_callbacks_lock); |
428870ff BB |
4620 | |
4621 | /* | |
4622 | * Since commit callbacks don't have any ordering requirement and since | |
4623 | * it is theoretically possible for a commit callback to be called | |
4624 | * after an arbitrary amount of time has elapsed since its txg has been | |
4625 | * synced, it is difficult to reliably determine whether a commit | |
4626 | * callback hasn't been called due to high load or due to a flawed | |
4627 | * implementation. | |
4628 | * | |
4629 | * In practice, we will assume that if after a certain number of txgs a | |
4630 | * commit callback hasn't been called, then most likely there's an | |
4631 | * implementation bug.. | |
4632 | */ | |
4633 | tmp_cb = list_head(&zcl.zcl_callbacks); | |
4634 | if (tmp_cb != NULL && | |
090ff092 | 4635 | tmp_cb->zcd_txg + ZTEST_COMMIT_CB_THRESH < txg) { |
428870ff BB |
4636 | fatal(0, "Commit callback threshold exceeded, oldest txg: %" |
4637 | PRIu64 ", open txg: %" PRIu64 "\n", tmp_cb->zcd_txg, txg); | |
4638 | } | |
4639 | ||
4640 | /* | |
4641 | * Let's find the place to insert our callbacks. | |
4642 | * | |
4643 | * Even though the list is ordered by txg, it is possible for the | |
4644 | * insertion point to not be the end because our txg may already be | |
4645 | * quiescing at this point and other callbacks in the open txg | |
4646 | * (from other objsets) may have sneaked in. | |
4647 | */ | |
4648 | tmp_cb = list_tail(&zcl.zcl_callbacks); | |
4649 | while (tmp_cb != NULL && tmp_cb->zcd_txg > txg) | |
4650 | tmp_cb = list_prev(&zcl.zcl_callbacks, tmp_cb); | |
4651 | ||
4652 | /* Add the 3 callbacks to the list */ | |
4653 | for (i = 0; i < 3; i++) { | |
4654 | if (tmp_cb == NULL) | |
4655 | list_insert_head(&zcl.zcl_callbacks, cb_data[i]); | |
4656 | else | |
4657 | list_insert_after(&zcl.zcl_callbacks, tmp_cb, | |
4658 | cb_data[i]); | |
4659 | ||
4660 | cb_data[i]->zcd_added = B_TRUE; | |
4661 | VERIFY(!cb_data[i]->zcd_called); | |
4662 | ||
4663 | tmp_cb = cb_data[i]; | |
4664 | } | |
4665 | ||
090ff092 RC |
4666 | zc_cb_counter += 3; |
4667 | ||
1e33ac1e | 4668 | (void) mutex_exit(&zcl.zcl_callbacks_lock); |
428870ff BB |
4669 | |
4670 | dmu_tx_commit(tx); | |
40b84e7a BB |
4671 | |
4672 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff BB |
4673 | } |
4674 | ||
4675 | /* ARGSUSED */ | |
4676 | void | |
4677 | ztest_dsl_prop_get_set(ztest_ds_t *zd, uint64_t id) | |
4678 | { | |
4679 | zfs_prop_t proplist[] = { | |
4680 | ZFS_PROP_CHECKSUM, | |
4681 | ZFS_PROP_COMPRESSION, | |
4682 | ZFS_PROP_COPIES, | |
4683 | ZFS_PROP_DEDUP | |
4684 | }; | |
d6320ddb | 4685 | int p; |
428870ff | 4686 | |
c242c188 | 4687 | (void) rw_enter(&ztest_name_lock, RW_READER); |
428870ff | 4688 | |
d6320ddb | 4689 | for (p = 0; p < sizeof (proplist) / sizeof (proplist[0]); p++) |
428870ff BB |
4690 | (void) ztest_dsl_prop_set_uint64(zd->zd_name, proplist[p], |
4691 | ztest_random_dsl_prop(proplist[p]), (int)ztest_random(2)); | |
4692 | ||
c242c188 | 4693 | (void) rw_exit(&ztest_name_lock); |
428870ff BB |
4694 | } |
4695 | ||
4696 | /* ARGSUSED */ | |
4697 | void | |
4698 | ztest_spa_prop_get_set(ztest_ds_t *zd, uint64_t id) | |
4699 | { | |
428870ff BB |
4700 | nvlist_t *props = NULL; |
4701 | ||
c242c188 | 4702 | (void) rw_enter(&ztest_name_lock, RW_READER); |
428870ff | 4703 | |
c242c188 | 4704 | (void) ztest_spa_prop_set_uint64(ZPOOL_PROP_DEDUPDITTO, |
428870ff BB |
4705 | ZIO_DEDUPDITTO_MIN + ztest_random(ZIO_DEDUPDITTO_MIN)); |
4706 | ||
c99c9001 | 4707 | VERIFY0(spa_prop_get(ztest_spa, &props)); |
428870ff | 4708 | |
c242c188 | 4709 | if (ztest_opts.zo_verbose >= 6) |
428870ff BB |
4710 | dump_nvlist(props, 4); |
4711 | ||
4712 | nvlist_free(props); | |
4713 | ||
c242c188 | 4714 | (void) rw_exit(&ztest_name_lock); |
428870ff BB |
4715 | } |
4716 | ||
4717 | /* | |
4718 | * Test snapshot hold/release and deferred destroy. | |
4719 | */ | |
4720 | void | |
4721 | ztest_dmu_snapshot_hold(ztest_ds_t *zd, uint64_t id) | |
34dc7c2f | 4722 | { |
34dc7c2f | 4723 | int error; |
428870ff BB |
4724 | objset_t *os = zd->zd_os; |
4725 | objset_t *origin; | |
4726 | char snapname[100]; | |
4727 | char fullname[100]; | |
4728 | char clonename[100]; | |
4729 | char tag[100]; | |
4730 | char osname[MAXNAMELEN]; | |
34dc7c2f | 4731 | |
c242c188 | 4732 | (void) rw_enter(&ztest_name_lock, RW_READER); |
34dc7c2f BB |
4733 | |
4734 | dmu_objset_name(os, osname); | |
4735 | ||
1e33ac1e | 4736 | (void) snprintf(snapname, 100, "sh1_%llu", (u_longlong_t)id); |
428870ff | 4737 | (void) snprintf(fullname, 100, "%s@%s", osname, snapname); |
1e33ac1e BB |
4738 | (void) snprintf(clonename, 100, "%s/ch1_%llu",osname,(u_longlong_t)id); |
4739 | (void) snprintf(tag, 100, "tag_%llu", (u_longlong_t)id); | |
428870ff BB |
4740 | |
4741 | /* | |
4742 | * Clean up from any previous run. | |
4743 | */ | |
4744 | (void) dmu_objset_destroy(clonename, B_FALSE); | |
4745 | (void) dsl_dataset_user_release(osname, snapname, tag, B_FALSE); | |
4746 | (void) dmu_objset_destroy(fullname, B_FALSE); | |
4747 | ||
4748 | /* | |
4749 | * Create snapshot, clone it, mark snap for deferred destroy, | |
4750 | * destroy clone, verify snap was also destroyed. | |
4751 | */ | |
572e2857 BB |
4752 | error = dmu_objset_snapshot(osname, snapname, NULL, NULL, FALSE, |
4753 | FALSE, -1); | |
428870ff BB |
4754 | if (error) { |
4755 | if (error == ENOSPC) { | |
4756 | ztest_record_enospc("dmu_objset_snapshot"); | |
4757 | goto out; | |
34dc7c2f | 4758 | } |
428870ff BB |
4759 | fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error); |
4760 | } | |
34dc7c2f | 4761 | |
428870ff BB |
4762 | error = dmu_objset_hold(fullname, FTAG, &origin); |
4763 | if (error) | |
4764 | fatal(0, "dmu_objset_hold(%s) = %d", fullname, error); | |
34dc7c2f | 4765 | |
428870ff BB |
4766 | error = dmu_objset_clone(clonename, dmu_objset_ds(origin), 0); |
4767 | dmu_objset_rele(origin, FTAG); | |
4768 | if (error) { | |
34dc7c2f | 4769 | if (error == ENOSPC) { |
428870ff BB |
4770 | ztest_record_enospc("dmu_objset_clone"); |
4771 | goto out; | |
34dc7c2f | 4772 | } |
428870ff BB |
4773 | fatal(0, "dmu_objset_clone(%s) = %d", clonename, error); |
4774 | } | |
34dc7c2f | 4775 | |
428870ff BB |
4776 | error = dmu_objset_destroy(fullname, B_TRUE); |
4777 | if (error) { | |
4778 | fatal(0, "dmu_objset_destroy(%s, B_TRUE) = %d", | |
4779 | fullname, error); | |
4780 | } | |
34dc7c2f | 4781 | |
428870ff BB |
4782 | error = dmu_objset_destroy(clonename, B_FALSE); |
4783 | if (error) | |
4784 | fatal(0, "dmu_objset_destroy(%s) = %d", clonename, error); | |
34dc7c2f | 4785 | |
428870ff BB |
4786 | error = dmu_objset_hold(fullname, FTAG, &origin); |
4787 | if (error != ENOENT) | |
4788 | fatal(0, "dmu_objset_hold(%s) = %d", fullname, error); | |
34dc7c2f | 4789 | |
428870ff BB |
4790 | /* |
4791 | * Create snapshot, add temporary hold, verify that we can't | |
4792 | * destroy a held snapshot, mark for deferred destroy, | |
4793 | * release hold, verify snapshot was destroyed. | |
4794 | */ | |
572e2857 BB |
4795 | error = dmu_objset_snapshot(osname, snapname, NULL, NULL, FALSE, |
4796 | FALSE, -1); | |
428870ff BB |
4797 | if (error) { |
4798 | if (error == ENOSPC) { | |
4799 | ztest_record_enospc("dmu_objset_snapshot"); | |
4800 | goto out; | |
34dc7c2f | 4801 | } |
428870ff BB |
4802 | fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error); |
4803 | } | |
4804 | ||
572e2857 BB |
4805 | error = dsl_dataset_user_hold(osname, snapname, tag, B_FALSE, |
4806 | B_TRUE, -1); | |
428870ff BB |
4807 | if (error) |
4808 | fatal(0, "dsl_dataset_user_hold(%s)", fullname, tag); | |
4809 | ||
4810 | error = dmu_objset_destroy(fullname, B_FALSE); | |
4811 | if (error != EBUSY) { | |
4812 | fatal(0, "dmu_objset_destroy(%s, B_FALSE) = %d", | |
4813 | fullname, error); | |
4814 | } | |
4815 | ||
4816 | error = dmu_objset_destroy(fullname, B_TRUE); | |
4817 | if (error) { | |
4818 | fatal(0, "dmu_objset_destroy(%s, B_TRUE) = %d", | |
4819 | fullname, error); | |
34dc7c2f BB |
4820 | } |
4821 | ||
428870ff BB |
4822 | error = dsl_dataset_user_release(osname, snapname, tag, B_FALSE); |
4823 | if (error) | |
4824 | fatal(0, "dsl_dataset_user_release(%s)", fullname, tag); | |
4825 | ||
4826 | VERIFY(dmu_objset_hold(fullname, FTAG, &origin) == ENOENT); | |
4827 | ||
4828 | out: | |
c242c188 | 4829 | (void) rw_exit(&ztest_name_lock); |
34dc7c2f BB |
4830 | } |
4831 | ||
34dc7c2f BB |
4832 | /* |
4833 | * Inject random faults into the on-disk data. | |
4834 | */ | |
428870ff | 4835 | /* ARGSUSED */ |
34dc7c2f | 4836 | void |
428870ff | 4837 | ztest_fault_inject(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4838 | { |
428870ff | 4839 | ztest_shared_t *zs = ztest_shared; |
c242c188 | 4840 | spa_t *spa = ztest_spa; |
34dc7c2f BB |
4841 | int fd; |
4842 | uint64_t offset; | |
428870ff | 4843 | uint64_t leaves; |
c5b3a7bb | 4844 | uint64_t bad = 0x1990c0ffeedecadeull; |
34dc7c2f | 4845 | uint64_t top, leaf; |
40b84e7a BB |
4846 | char *path0; |
4847 | char *pathrand; | |
34dc7c2f | 4848 | size_t fsize; |
34dc7c2f BB |
4849 | int bshift = SPA_MAXBLOCKSHIFT + 2; /* don't scrog all labels */ |
4850 | int iters = 1000; | |
428870ff BB |
4851 | int maxfaults; |
4852 | int mirror_save; | |
b128c09f | 4853 | vdev_t *vd0 = NULL; |
34dc7c2f | 4854 | uint64_t guid0 = 0; |
428870ff BB |
4855 | boolean_t islog = B_FALSE; |
4856 | ||
40b84e7a BB |
4857 | path0 = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
4858 | pathrand = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
4859 | ||
c242c188 | 4860 | mutex_enter(&ztest_vdev_lock); |
428870ff | 4861 | maxfaults = MAXFAULTS(); |
c242c188 | 4862 | leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raidz; |
428870ff | 4863 | mirror_save = zs->zs_mirrors; |
c242c188 | 4864 | mutex_exit(&ztest_vdev_lock); |
34dc7c2f | 4865 | |
b128c09f | 4866 | ASSERT(leaves >= 1); |
34dc7c2f BB |
4867 | |
4868 | /* | |
b128c09f | 4869 | * We need SCL_STATE here because we're going to look at vd0->vdev_tsd. |
34dc7c2f | 4870 | */ |
b128c09f | 4871 | spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); |
34dc7c2f | 4872 | |
b128c09f BB |
4873 | if (ztest_random(2) == 0) { |
4874 | /* | |
428870ff | 4875 | * Inject errors on a normal data device or slog device. |
b128c09f | 4876 | */ |
428870ff BB |
4877 | top = ztest_random_vdev_top(spa, B_TRUE); |
4878 | leaf = ztest_random(leaves) + zs->zs_splits; | |
34dc7c2f | 4879 | |
b128c09f BB |
4880 | /* |
4881 | * Generate paths to the first leaf in this top-level vdev, | |
4882 | * and to the random leaf we selected. We'll induce transient | |
4883 | * write failures and random online/offline activity on leaf 0, | |
4884 | * and we'll write random garbage to the randomly chosen leaf. | |
4885 | */ | |
6aec1cd5 | 4886 | (void) snprintf(path0, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
4887 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
4888 | top * leaves + zs->zs_splits); | |
6aec1cd5 | 4889 | (void) snprintf(pathrand, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
4890 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
4891 | top * leaves + leaf); | |
34dc7c2f | 4892 | |
b128c09f | 4893 | vd0 = vdev_lookup_by_path(spa->spa_root_vdev, path0); |
428870ff BB |
4894 | if (vd0 != NULL && vd0->vdev_top->vdev_islog) |
4895 | islog = B_TRUE; | |
4896 | ||
b128c09f BB |
4897 | if (vd0 != NULL && maxfaults != 1) { |
4898 | /* | |
4899 | * Make vd0 explicitly claim to be unreadable, | |
4900 | * or unwriteable, or reach behind its back | |
4901 | * and close the underlying fd. We can do this if | |
4902 | * maxfaults == 0 because we'll fail and reexecute, | |
4903 | * and we can do it if maxfaults >= 2 because we'll | |
4904 | * have enough redundancy. If maxfaults == 1, the | |
4905 | * combination of this with injection of random data | |
4906 | * corruption below exceeds the pool's fault tolerance. | |
4907 | */ | |
4908 | vdev_file_t *vf = vd0->vdev_tsd; | |
4909 | ||
4910 | if (vf != NULL && ztest_random(3) == 0) { | |
4911 | (void) close(vf->vf_vnode->v_fd); | |
4912 | vf->vf_vnode->v_fd = -1; | |
4913 | } else if (ztest_random(2) == 0) { | |
4914 | vd0->vdev_cant_read = B_TRUE; | |
4915 | } else { | |
4916 | vd0->vdev_cant_write = B_TRUE; | |
4917 | } | |
4918 | guid0 = vd0->vdev_guid; | |
4919 | } | |
4920 | } else { | |
4921 | /* | |
4922 | * Inject errors on an l2cache device. | |
4923 | */ | |
4924 | spa_aux_vdev_t *sav = &spa->spa_l2cache; | |
34dc7c2f | 4925 | |
b128c09f BB |
4926 | if (sav->sav_count == 0) { |
4927 | spa_config_exit(spa, SCL_STATE, FTAG); | |
40b84e7a | 4928 | goto out; |
b128c09f BB |
4929 | } |
4930 | vd0 = sav->sav_vdevs[ztest_random(sav->sav_count)]; | |
34dc7c2f | 4931 | guid0 = vd0->vdev_guid; |
b128c09f BB |
4932 | (void) strcpy(path0, vd0->vdev_path); |
4933 | (void) strcpy(pathrand, vd0->vdev_path); | |
4934 | ||
4935 | leaf = 0; | |
4936 | leaves = 1; | |
4937 | maxfaults = INT_MAX; /* no limit on cache devices */ | |
34dc7c2f BB |
4938 | } |
4939 | ||
b128c09f BB |
4940 | spa_config_exit(spa, SCL_STATE, FTAG); |
4941 | ||
34dc7c2f | 4942 | /* |
428870ff BB |
4943 | * If we can tolerate two or more faults, or we're dealing |
4944 | * with a slog, randomly online/offline vd0. | |
34dc7c2f | 4945 | */ |
428870ff | 4946 | if ((maxfaults >= 2 || islog) && guid0 != 0) { |
fb5f0bc8 BB |
4947 | if (ztest_random(10) < 6) { |
4948 | int flags = (ztest_random(2) == 0 ? | |
4949 | ZFS_OFFLINE_TEMPORARY : 0); | |
428870ff BB |
4950 | |
4951 | /* | |
4952 | * We have to grab the zs_name_lock as writer to | |
4953 | * prevent a race between offlining a slog and | |
4954 | * destroying a dataset. Offlining the slog will | |
4955 | * grab a reference on the dataset which may cause | |
4956 | * dmu_objset_destroy() to fail with EBUSY thus | |
4957 | * leaving the dataset in an inconsistent state. | |
4958 | */ | |
4959 | if (islog) | |
c242c188 | 4960 | (void) rw_enter(&ztest_name_lock, |
1e33ac1e | 4961 | RW_WRITER); |
428870ff | 4962 | |
fb5f0bc8 | 4963 | VERIFY(vdev_offline(spa, guid0, flags) != EBUSY); |
428870ff BB |
4964 | |
4965 | if (islog) | |
c242c188 | 4966 | (void) rw_exit(&ztest_name_lock); |
fb5f0bc8 | 4967 | } else { |
1eb5bfa3 GW |
4968 | /* |
4969 | * Ideally we would like to be able to randomly | |
4970 | * call vdev_[on|off]line without holding locks | |
4971 | * to force unpredictable failures but the side | |
4972 | * effects of vdev_[on|off]line prevent us from | |
4973 | * doing so. We grab the ztest_vdev_lock here to | |
4974 | * prevent a race between injection testing and | |
4975 | * aux_vdev removal. | |
4976 | */ | |
4977 | mutex_enter(&ztest_vdev_lock); | |
fb5f0bc8 | 4978 | (void) vdev_online(spa, guid0, 0, NULL); |
1eb5bfa3 | 4979 | mutex_exit(&ztest_vdev_lock); |
fb5f0bc8 | 4980 | } |
34dc7c2f BB |
4981 | } |
4982 | ||
428870ff | 4983 | if (maxfaults == 0) |
40b84e7a | 4984 | goto out; |
428870ff | 4985 | |
34dc7c2f BB |
4986 | /* |
4987 | * We have at least single-fault tolerance, so inject data corruption. | |
4988 | */ | |
4989 | fd = open(pathrand, O_RDWR); | |
4990 | ||
4991 | if (fd == -1) /* we hit a gap in the device namespace */ | |
40b84e7a | 4992 | goto out; |
34dc7c2f BB |
4993 | |
4994 | fsize = lseek(fd, 0, SEEK_END); | |
4995 | ||
4996 | while (--iters != 0) { | |
4997 | offset = ztest_random(fsize / (leaves << bshift)) * | |
4998 | (leaves << bshift) + (leaf << bshift) + | |
4999 | (ztest_random(1ULL << (bshift - 1)) & -8ULL); | |
5000 | ||
5001 | if (offset >= fsize) | |
5002 | continue; | |
5003 | ||
c242c188 | 5004 | mutex_enter(&ztest_vdev_lock); |
428870ff | 5005 | if (mirror_save != zs->zs_mirrors) { |
c242c188 | 5006 | mutex_exit(&ztest_vdev_lock); |
428870ff | 5007 | (void) close(fd); |
40b84e7a | 5008 | goto out; |
428870ff | 5009 | } |
34dc7c2f BB |
5010 | |
5011 | if (pwrite(fd, &bad, sizeof (bad), offset) != sizeof (bad)) | |
5012 | fatal(1, "can't inject bad word at 0x%llx in %s", | |
5013 | offset, pathrand); | |
428870ff | 5014 | |
c242c188 | 5015 | mutex_exit(&ztest_vdev_lock); |
428870ff | 5016 | |
c242c188 | 5017 | if (ztest_opts.zo_verbose >= 7) |
428870ff BB |
5018 | (void) printf("injected bad word into %s," |
5019 | " offset 0x%llx\n", pathrand, (u_longlong_t)offset); | |
34dc7c2f BB |
5020 | } |
5021 | ||
5022 | (void) close(fd); | |
40b84e7a BB |
5023 | out: |
5024 | umem_free(path0, MAXPATHLEN); | |
5025 | umem_free(pathrand, MAXPATHLEN); | |
34dc7c2f BB |
5026 | } |
5027 | ||
5028 | /* | |
428870ff | 5029 | * Verify that DDT repair works as expected. |
34dc7c2f BB |
5030 | */ |
5031 | void | |
428870ff | 5032 | ztest_ddt_repair(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 5033 | { |
428870ff | 5034 | ztest_shared_t *zs = ztest_shared; |
c242c188 | 5035 | spa_t *spa = ztest_spa; |
428870ff | 5036 | objset_t *os = zd->zd_os; |
40b84e7a | 5037 | ztest_od_t *od; |
428870ff BB |
5038 | uint64_t object, blocksize, txg, pattern, psize; |
5039 | enum zio_checksum checksum = spa_dedup_checksum(spa); | |
5040 | dmu_buf_t *db; | |
5041 | dmu_tx_t *tx; | |
5042 | void *buf; | |
5043 | blkptr_t blk; | |
5044 | int copies = 2 * ZIO_DEDUPDITTO_MIN; | |
d6320ddb | 5045 | int i; |
34dc7c2f | 5046 | |
428870ff BB |
5047 | blocksize = ztest_random_blocksize(); |
5048 | blocksize = MIN(blocksize, 2048); /* because we write so many */ | |
34dc7c2f | 5049 | |
40b84e7a BB |
5050 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
5051 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0); | |
34dc7c2f | 5052 | |
40b84e7a BB |
5053 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
5054 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff | 5055 | return; |
40b84e7a | 5056 | } |
34dc7c2f BB |
5057 | |
5058 | /* | |
428870ff BB |
5059 | * Take the name lock as writer to prevent anyone else from changing |
5060 | * the pool and dataset properies we need to maintain during this test. | |
34dc7c2f | 5061 | */ |
c242c188 | 5062 | (void) rw_enter(&ztest_name_lock, RW_WRITER); |
34dc7c2f | 5063 | |
428870ff BB |
5064 | if (ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_DEDUP, checksum, |
5065 | B_FALSE) != 0 || | |
5066 | ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_COPIES, 1, | |
5067 | B_FALSE) != 0) { | |
c242c188 | 5068 | (void) rw_exit(&ztest_name_lock); |
40b84e7a | 5069 | umem_free(od, sizeof(ztest_od_t)); |
428870ff BB |
5070 | return; |
5071 | } | |
5072 | ||
5073 | object = od[0].od_object; | |
5074 | blocksize = od[0].od_blocksize; | |
3541dc6d | 5075 | pattern = zs->zs_guid ^ dmu_objset_fsid_guid(os); |
428870ff BB |
5076 | |
5077 | ASSERT(object != 0); | |
5078 | ||
5079 | tx = dmu_tx_create(os); | |
5080 | dmu_tx_hold_write(tx, object, 0, copies * blocksize); | |
5081 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
5082 | if (txg == 0) { | |
c242c188 | 5083 | (void) rw_exit(&ztest_name_lock); |
40b84e7a | 5084 | umem_free(od, sizeof(ztest_od_t)); |
428870ff BB |
5085 | return; |
5086 | } | |
34dc7c2f BB |
5087 | |
5088 | /* | |
428870ff | 5089 | * Write all the copies of our block. |
34dc7c2f | 5090 | */ |
d6320ddb | 5091 | for (i = 0; i < copies; i++) { |
428870ff BB |
5092 | uint64_t offset = i * blocksize; |
5093 | VERIFY(dmu_buf_hold(os, object, offset, FTAG, &db, | |
5094 | DMU_READ_NO_PREFETCH) == 0); | |
5095 | ASSERT(db->db_offset == offset); | |
5096 | ASSERT(db->db_size == blocksize); | |
5097 | ASSERT(ztest_pattern_match(db->db_data, db->db_size, pattern) || | |
5098 | ztest_pattern_match(db->db_data, db->db_size, 0ULL)); | |
5099 | dmu_buf_will_fill(db, tx); | |
5100 | ztest_pattern_set(db->db_data, db->db_size, pattern); | |
5101 | dmu_buf_rele(db, FTAG); | |
5102 | } | |
34dc7c2f | 5103 | |
428870ff BB |
5104 | dmu_tx_commit(tx); |
5105 | txg_wait_synced(spa_get_dsl(spa), txg); | |
34dc7c2f BB |
5106 | |
5107 | /* | |
428870ff | 5108 | * Find out what block we got. |
34dc7c2f | 5109 | */ |
428870ff BB |
5110 | VERIFY(dmu_buf_hold(os, object, 0, FTAG, &db, |
5111 | DMU_READ_NO_PREFETCH) == 0); | |
5112 | blk = *((dmu_buf_impl_t *)db)->db_blkptr; | |
5113 | dmu_buf_rele(db, FTAG); | |
34dc7c2f BB |
5114 | |
5115 | /* | |
428870ff | 5116 | * Damage the block. Dedup-ditto will save us when we read it later. |
34dc7c2f | 5117 | */ |
428870ff BB |
5118 | psize = BP_GET_PSIZE(&blk); |
5119 | buf = zio_buf_alloc(psize); | |
5120 | ztest_pattern_set(buf, psize, ~pattern); | |
34dc7c2f | 5121 | |
428870ff BB |
5122 | (void) zio_wait(zio_rewrite(NULL, spa, 0, &blk, |
5123 | buf, psize, NULL, NULL, ZIO_PRIORITY_SYNC_WRITE, | |
5124 | ZIO_FLAG_CANFAIL | ZIO_FLAG_INDUCE_DAMAGE, NULL)); | |
34dc7c2f | 5125 | |
428870ff | 5126 | zio_buf_free(buf, psize); |
34dc7c2f | 5127 | |
c242c188 | 5128 | (void) rw_exit(&ztest_name_lock); |
40b84e7a | 5129 | umem_free(od, sizeof(ztest_od_t)); |
34dc7c2f BB |
5130 | } |
5131 | ||
34dc7c2f | 5132 | /* |
428870ff | 5133 | * Scrub the pool. |
34dc7c2f | 5134 | */ |
428870ff BB |
5135 | /* ARGSUSED */ |
5136 | void | |
5137 | ztest_scrub(ztest_ds_t *zd, uint64_t id) | |
34dc7c2f | 5138 | { |
c242c188 | 5139 | spa_t *spa = ztest_spa; |
34dc7c2f | 5140 | |
428870ff BB |
5141 | (void) spa_scan(spa, POOL_SCAN_SCRUB); |
5142 | (void) poll(NULL, 0, 100); /* wait a moment, then force a restart */ | |
5143 | (void) spa_scan(spa, POOL_SCAN_SCRUB); | |
5144 | } | |
34dc7c2f | 5145 | |
3541dc6d GA |
5146 | /* |
5147 | * Change the guid for the pool. | |
5148 | */ | |
5149 | /* ARGSUSED */ | |
5150 | void | |
5151 | ztest_reguid(ztest_ds_t *zd, uint64_t id) | |
5152 | { | |
c242c188 | 5153 | spa_t *spa = ztest_spa; |
3541dc6d | 5154 | uint64_t orig, load; |
3bc7e0fb | 5155 | int error; |
3541dc6d GA |
5156 | |
5157 | orig = spa_guid(spa); | |
5158 | load = spa_load_guid(spa); | |
3bc7e0fb GW |
5159 | |
5160 | (void) rw_enter(&ztest_name_lock, RW_WRITER); | |
5161 | error = spa_change_guid(spa); | |
5162 | (void) rw_exit(&ztest_name_lock); | |
5163 | ||
5164 | if (error != 0) | |
3541dc6d GA |
5165 | return; |
5166 | ||
ea0b2538 | 5167 | if (ztest_opts.zo_verbose >= 4) { |
3541dc6d GA |
5168 | (void) printf("Changed guid old %llu -> %llu\n", |
5169 | (u_longlong_t)orig, (u_longlong_t)spa_guid(spa)); | |
5170 | } | |
5171 | ||
5172 | VERIFY3U(orig, !=, spa_guid(spa)); | |
5173 | VERIFY3U(load, ==, spa_load_guid(spa)); | |
5174 | } | |
5175 | ||
428870ff BB |
5176 | /* |
5177 | * Rename the pool to a different name and then rename it back. | |
5178 | */ | |
5179 | /* ARGSUSED */ | |
5180 | void | |
5181 | ztest_spa_rename(ztest_ds_t *zd, uint64_t id) | |
5182 | { | |
428870ff BB |
5183 | char *oldname, *newname; |
5184 | spa_t *spa; | |
34dc7c2f | 5185 | |
c242c188 | 5186 | (void) rw_enter(&ztest_name_lock, RW_WRITER); |
34dc7c2f | 5187 | |
c242c188 | 5188 | oldname = ztest_opts.zo_pool; |
428870ff BB |
5189 | newname = umem_alloc(strlen(oldname) + 5, UMEM_NOFAIL); |
5190 | (void) strcpy(newname, oldname); | |
5191 | (void) strcat(newname, "_tmp"); | |
34dc7c2f BB |
5192 | |
5193 | /* | |
428870ff | 5194 | * Do the rename |
34dc7c2f | 5195 | */ |
428870ff | 5196 | VERIFY3U(0, ==, spa_rename(oldname, newname)); |
34dc7c2f BB |
5197 | |
5198 | /* | |
428870ff | 5199 | * Try to open it under the old name, which shouldn't exist |
34dc7c2f | 5200 | */ |
428870ff | 5201 | VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f BB |
5202 | |
5203 | /* | |
428870ff BB |
5204 | * Open it under the new name and make sure it's still the same spa_t. |
5205 | */ | |
5206 | VERIFY3U(0, ==, spa_open(newname, &spa, FTAG)); | |
34dc7c2f | 5207 | |
c242c188 | 5208 | ASSERT(spa == ztest_spa); |
428870ff | 5209 | spa_close(spa, FTAG); |
34dc7c2f BB |
5210 | |
5211 | /* | |
428870ff | 5212 | * Rename it back to the original |
34dc7c2f | 5213 | */ |
428870ff | 5214 | VERIFY3U(0, ==, spa_rename(newname, oldname)); |
34dc7c2f | 5215 | |
428870ff BB |
5216 | /* |
5217 | * Make sure it can still be opened | |
5218 | */ | |
5219 | VERIFY3U(0, ==, spa_open(oldname, &spa, FTAG)); | |
34dc7c2f | 5220 | |
c242c188 | 5221 | ASSERT(spa == ztest_spa); |
428870ff BB |
5222 | spa_close(spa, FTAG); |
5223 | ||
5224 | umem_free(newname, strlen(newname) + 1); | |
5225 | ||
c242c188 | 5226 | (void) rw_exit(&ztest_name_lock); |
34dc7c2f BB |
5227 | } |
5228 | ||
428870ff BB |
5229 | /* |
5230 | * Verify pool integrity by running zdb. | |
5231 | */ | |
34dc7c2f | 5232 | static void |
428870ff | 5233 | ztest_run_zdb(char *pool) |
34dc7c2f BB |
5234 | { |
5235 | int status; | |
34dc7c2f | 5236 | char *bin; |
0e8d1b2d BB |
5237 | char *zdb; |
5238 | char *zbuf; | |
34dc7c2f BB |
5239 | FILE *fp; |
5240 | ||
0e8d1b2d BB |
5241 | bin = umem_alloc(MAXPATHLEN + MAXNAMELEN + 20, UMEM_NOFAIL); |
5242 | zdb = umem_alloc(MAXPATHLEN + MAXNAMELEN + 20, UMEM_NOFAIL); | |
5243 | zbuf = umem_alloc(1024, UMEM_NOFAIL); | |
34dc7c2f | 5244 | |
0e8d1b2d | 5245 | VERIFY(realpath(getexecname(), bin) != NULL); |
341b5f1d | 5246 | if (strncmp(bin, "/usr/sbin/ztest", 15) == 0) { |
0e8d1b2d | 5247 | strcpy(bin, "/usr/sbin/zdb"); /* Installed */ |
341b5f1d BB |
5248 | } else if (strncmp(bin, "/sbin/ztest", 11) == 0) { |
5249 | strcpy(bin, "/sbin/zdb"); /* Installed */ | |
0e8d1b2d BB |
5250 | } else { |
5251 | strstr(bin, "/ztest/")[0] = '\0'; /* In-tree */ | |
5252 | strcat(bin, "/zdb/zdb"); | |
5253 | } | |
5254 | ||
5255 | (void) sprintf(zdb, | |
5256 | "%s -bcc%s%s -U %s %s", | |
5257 | bin, | |
c242c188 CS |
5258 | ztest_opts.zo_verbose >= 3 ? "s" : "", |
5259 | ztest_opts.zo_verbose >= 4 ? "v" : "", | |
428870ff | 5260 | spa_config_path, |
b128c09f | 5261 | pool); |
34dc7c2f | 5262 | |
c242c188 | 5263 | if (ztest_opts.zo_verbose >= 5) |
34dc7c2f BB |
5264 | (void) printf("Executing %s\n", strstr(zdb, "zdb ")); |
5265 | ||
5266 | fp = popen(zdb, "r"); | |
5267 | ||
6aec1cd5 | 5268 | while (fgets(zbuf, 1024, fp) != NULL) |
c242c188 | 5269 | if (ztest_opts.zo_verbose >= 3) |
34dc7c2f BB |
5270 | (void) printf("%s", zbuf); |
5271 | ||
5272 | status = pclose(fp); | |
5273 | ||
5274 | if (status == 0) | |
0e8d1b2d | 5275 | goto out; |
34dc7c2f BB |
5276 | |
5277 | ztest_dump_core = 0; | |
5278 | if (WIFEXITED(status)) | |
5279 | fatal(0, "'%s' exit code %d", zdb, WEXITSTATUS(status)); | |
5280 | else | |
5281 | fatal(0, "'%s' died with signal %d", zdb, WTERMSIG(status)); | |
0e8d1b2d BB |
5282 | out: |
5283 | umem_free(bin, MAXPATHLEN + MAXNAMELEN + 20); | |
5284 | umem_free(zdb, MAXPATHLEN + MAXNAMELEN + 20); | |
5285 | umem_free(zbuf, 1024); | |
34dc7c2f BB |
5286 | } |
5287 | ||
5288 | static void | |
5289 | ztest_walk_pool_directory(char *header) | |
5290 | { | |
5291 | spa_t *spa = NULL; | |
5292 | ||
c242c188 | 5293 | if (ztest_opts.zo_verbose >= 6) |
34dc7c2f BB |
5294 | (void) printf("%s\n", header); |
5295 | ||
5296 | mutex_enter(&spa_namespace_lock); | |
5297 | while ((spa = spa_next(spa)) != NULL) | |
c242c188 | 5298 | if (ztest_opts.zo_verbose >= 6) |
34dc7c2f BB |
5299 | (void) printf("\t%s\n", spa_name(spa)); |
5300 | mutex_exit(&spa_namespace_lock); | |
5301 | } | |
5302 | ||
5303 | static void | |
5304 | ztest_spa_import_export(char *oldname, char *newname) | |
5305 | { | |
fb5f0bc8 | 5306 | nvlist_t *config, *newconfig; |
34dc7c2f BB |
5307 | uint64_t pool_guid; |
5308 | spa_t *spa; | |
34dc7c2f | 5309 | |
c242c188 | 5310 | if (ztest_opts.zo_verbose >= 4) { |
34dc7c2f BB |
5311 | (void) printf("import/export: old = %s, new = %s\n", |
5312 | oldname, newname); | |
5313 | } | |
5314 | ||
5315 | /* | |
5316 | * Clean up from previous runs. | |
5317 | */ | |
5318 | (void) spa_destroy(newname); | |
5319 | ||
5320 | /* | |
5321 | * Get the pool's configuration and guid. | |
5322 | */ | |
428870ff | 5323 | VERIFY3U(0, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f | 5324 | |
fb5f0bc8 BB |
5325 | /* |
5326 | * Kick off a scrub to tickle scrub/export races. | |
5327 | */ | |
5328 | if (ztest_random(2) == 0) | |
428870ff | 5329 | (void) spa_scan(spa, POOL_SCAN_SCRUB); |
fb5f0bc8 | 5330 | |
34dc7c2f BB |
5331 | pool_guid = spa_guid(spa); |
5332 | spa_close(spa, FTAG); | |
5333 | ||
5334 | ztest_walk_pool_directory("pools before export"); | |
5335 | ||
5336 | /* | |
5337 | * Export it. | |
5338 | */ | |
428870ff | 5339 | VERIFY3U(0, ==, spa_export(oldname, &config, B_FALSE, B_FALSE)); |
34dc7c2f BB |
5340 | |
5341 | ztest_walk_pool_directory("pools after export"); | |
5342 | ||
fb5f0bc8 BB |
5343 | /* |
5344 | * Try to import it. | |
5345 | */ | |
5346 | newconfig = spa_tryimport(config); | |
5347 | ASSERT(newconfig != NULL); | |
5348 | nvlist_free(newconfig); | |
5349 | ||
34dc7c2f BB |
5350 | /* |
5351 | * Import it under the new name. | |
5352 | */ | |
572e2857 | 5353 | VERIFY3U(0, ==, spa_import(newname, config, NULL, 0)); |
34dc7c2f BB |
5354 | |
5355 | ztest_walk_pool_directory("pools after import"); | |
5356 | ||
5357 | /* | |
5358 | * Try to import it again -- should fail with EEXIST. | |
5359 | */ | |
572e2857 | 5360 | VERIFY3U(EEXIST, ==, spa_import(newname, config, NULL, 0)); |
34dc7c2f BB |
5361 | |
5362 | /* | |
5363 | * Try to import it under a different name -- should fail with EEXIST. | |
5364 | */ | |
572e2857 | 5365 | VERIFY3U(EEXIST, ==, spa_import(oldname, config, NULL, 0)); |
34dc7c2f BB |
5366 | |
5367 | /* | |
5368 | * Verify that the pool is no longer visible under the old name. | |
5369 | */ | |
428870ff | 5370 | VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f BB |
5371 | |
5372 | /* | |
5373 | * Verify that we can open and close the pool using the new name. | |
5374 | */ | |
428870ff | 5375 | VERIFY3U(0, ==, spa_open(newname, &spa, FTAG)); |
34dc7c2f BB |
5376 | ASSERT(pool_guid == spa_guid(spa)); |
5377 | spa_close(spa, FTAG); | |
5378 | ||
5379 | nvlist_free(config); | |
5380 | } | |
5381 | ||
fb5f0bc8 BB |
5382 | static void |
5383 | ztest_resume(spa_t *spa) | |
5384 | { | |
c242c188 | 5385 | if (spa_suspended(spa) && ztest_opts.zo_verbose >= 6) |
428870ff BB |
5386 | (void) printf("resuming from suspended state\n"); |
5387 | spa_vdev_state_enter(spa, SCL_NONE); | |
5388 | vdev_clear(spa, NULL); | |
5389 | (void) spa_vdev_state_exit(spa, NULL, 0); | |
5390 | (void) zio_resume(spa); | |
fb5f0bc8 BB |
5391 | } |
5392 | ||
34dc7c2f | 5393 | static void * |
fb5f0bc8 | 5394 | ztest_resume_thread(void *arg) |
34dc7c2f | 5395 | { |
b128c09f | 5396 | spa_t *spa = arg; |
34dc7c2f BB |
5397 | |
5398 | while (!ztest_exiting) { | |
428870ff BB |
5399 | if (spa_suspended(spa)) |
5400 | ztest_resume(spa); | |
5401 | (void) poll(NULL, 0, 100); | |
34dc7c2f | 5402 | } |
34dc7c2f | 5403 | |
1e33ac1e | 5404 | thread_exit(); |
428870ff | 5405 | |
1e33ac1e BB |
5406 | return (NULL); |
5407 | } | |
428870ff | 5408 | |
1e33ac1e | 5409 | #define GRACE 300 |
428870ff | 5410 | |
26099167 | 5411 | #if 0 |
1e33ac1e BB |
5412 | static void |
5413 | ztest_deadman_alarm(int sig) | |
5414 | { | |
5415 | fatal(0, "failed to complete within %d seconds of deadline", GRACE); | |
428870ff | 5416 | } |
26099167 | 5417 | #endif |
428870ff BB |
5418 | |
5419 | static void | |
c242c188 | 5420 | ztest_execute(int test, ztest_info_t *zi, uint64_t id) |
428870ff | 5421 | { |
c242c188 CS |
5422 | ztest_ds_t *zd = &ztest_ds[id % ztest_opts.zo_datasets]; |
5423 | ztest_shared_callstate_t *zc = ZTEST_GET_SHARED_CALLSTATE(test); | |
428870ff | 5424 | hrtime_t functime = gethrtime(); |
d6320ddb | 5425 | int i; |
428870ff | 5426 | |
d6320ddb | 5427 | for (i = 0; i < zi->zi_iters; i++) |
428870ff BB |
5428 | zi->zi_func(zd, id); |
5429 | ||
5430 | functime = gethrtime() - functime; | |
5431 | ||
c242c188 CS |
5432 | atomic_add_64(&zc->zc_count, 1); |
5433 | atomic_add_64(&zc->zc_time, functime); | |
428870ff | 5434 | |
c242c188 | 5435 | if (ztest_opts.zo_verbose >= 4) { |
428870ff BB |
5436 | Dl_info dli; |
5437 | (void) dladdr((void *)zi->zi_func, &dli); | |
5438 | (void) printf("%6.2f sec in %s\n", | |
5439 | (double)functime / NANOSEC, dli.dli_sname); | |
5440 | } | |
5441 | } | |
5442 | ||
34dc7c2f BB |
5443 | static void * |
5444 | ztest_thread(void *arg) | |
5445 | { | |
c242c188 | 5446 | int rand; |
428870ff | 5447 | uint64_t id = (uintptr_t)arg; |
34dc7c2f | 5448 | ztest_shared_t *zs = ztest_shared; |
428870ff BB |
5449 | uint64_t call_next; |
5450 | hrtime_t now; | |
34dc7c2f | 5451 | ztest_info_t *zi; |
c242c188 | 5452 | ztest_shared_callstate_t *zc; |
34dc7c2f | 5453 | |
428870ff | 5454 | while ((now = gethrtime()) < zs->zs_thread_stop) { |
34dc7c2f BB |
5455 | /* |
5456 | * See if it's time to force a crash. | |
5457 | */ | |
428870ff BB |
5458 | if (now > zs->zs_thread_kill) |
5459 | ztest_kill(zs); | |
34dc7c2f BB |
5460 | |
5461 | /* | |
428870ff | 5462 | * If we're getting ENOSPC with some regularity, stop. |
34dc7c2f | 5463 | */ |
428870ff BB |
5464 | if (zs->zs_enospc_count > 10) |
5465 | break; | |
34dc7c2f BB |
5466 | |
5467 | /* | |
428870ff | 5468 | * Pick a random function to execute. |
34dc7c2f | 5469 | */ |
c242c188 CS |
5470 | rand = ztest_random(ZTEST_FUNCS); |
5471 | zi = &ztest_info[rand]; | |
5472 | zc = ZTEST_GET_SHARED_CALLSTATE(rand); | |
5473 | call_next = zc->zc_next; | |
34dc7c2f | 5474 | |
428870ff | 5475 | if (now >= call_next && |
c242c188 CS |
5476 | atomic_cas_64(&zc->zc_next, call_next, call_next + |
5477 | ztest_random(2 * zi->zi_interval[0] + 1)) == call_next) { | |
5478 | ztest_execute(rand, zi, id); | |
5479 | } | |
428870ff | 5480 | } |
34dc7c2f | 5481 | |
1e33ac1e BB |
5482 | thread_exit(); |
5483 | ||
428870ff BB |
5484 | return (NULL); |
5485 | } | |
34dc7c2f | 5486 | |
428870ff BB |
5487 | static void |
5488 | ztest_dataset_name(char *dsname, char *pool, int d) | |
5489 | { | |
5490 | (void) snprintf(dsname, MAXNAMELEN, "%s/ds_%d", pool, d); | |
5491 | } | |
34dc7c2f | 5492 | |
428870ff | 5493 | static void |
c242c188 | 5494 | ztest_dataset_destroy(int d) |
428870ff BB |
5495 | { |
5496 | char name[MAXNAMELEN]; | |
d6320ddb | 5497 | int t; |
34dc7c2f | 5498 | |
c242c188 | 5499 | ztest_dataset_name(name, ztest_opts.zo_pool, d); |
34dc7c2f | 5500 | |
c242c188 | 5501 | if (ztest_opts.zo_verbose >= 3) |
428870ff | 5502 | (void) printf("Destroying %s to free up space\n", name); |
34dc7c2f | 5503 | |
428870ff BB |
5504 | /* |
5505 | * Cleanup any non-standard clones and snapshots. In general, | |
5506 | * ztest thread t operates on dataset (t % zopt_datasets), | |
5507 | * so there may be more than one thing to clean up. | |
5508 | */ | |
c242c188 CS |
5509 | for (t = d; t < ztest_opts.zo_threads; |
5510 | t += ztest_opts.zo_datasets) | |
428870ff BB |
5511 | ztest_dsl_dataset_cleanup(name, t); |
5512 | ||
5513 | (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL, | |
5514 | DS_FIND_SNAPSHOTS | DS_FIND_CHILDREN); | |
5515 | } | |
5516 | ||
5517 | static void | |
5518 | ztest_dataset_dirobj_verify(ztest_ds_t *zd) | |
5519 | { | |
5520 | uint64_t usedobjs, dirobjs, scratch; | |
5521 | ||
5522 | /* | |
5523 | * ZTEST_DIROBJ is the object directory for the entire dataset. | |
5524 | * Therefore, the number of objects in use should equal the | |
5525 | * number of ZTEST_DIROBJ entries, +1 for ZTEST_DIROBJ itself. | |
5526 | * If not, we have an object leak. | |
5527 | * | |
5528 | * Note that we can only check this in ztest_dataset_open(), | |
5529 | * when the open-context and syncing-context values agree. | |
5530 | * That's because zap_count() returns the open-context value, | |
5531 | * while dmu_objset_space() returns the rootbp fill count. | |
5532 | */ | |
5533 | VERIFY3U(0, ==, zap_count(zd->zd_os, ZTEST_DIROBJ, &dirobjs)); | |
5534 | dmu_objset_space(zd->zd_os, &scratch, &scratch, &usedobjs, &scratch); | |
5535 | ASSERT3U(dirobjs + 1, ==, usedobjs); | |
5536 | } | |
5537 | ||
5538 | static int | |
c242c188 | 5539 | ztest_dataset_open(int d) |
428870ff | 5540 | { |
c242c188 CS |
5541 | ztest_ds_t *zd = &ztest_ds[d]; |
5542 | uint64_t committed_seq = ZTEST_GET_SHARED_DS(d)->zd_seq; | |
428870ff BB |
5543 | objset_t *os; |
5544 | zilog_t *zilog; | |
5545 | char name[MAXNAMELEN]; | |
5546 | int error; | |
5547 | ||
c242c188 | 5548 | ztest_dataset_name(name, ztest_opts.zo_pool, d); |
428870ff | 5549 | |
c242c188 | 5550 | (void) rw_enter(&ztest_name_lock, RW_READER); |
428870ff BB |
5551 | |
5552 | error = ztest_dataset_create(name); | |
5553 | if (error == ENOSPC) { | |
c242c188 | 5554 | (void) rw_exit(&ztest_name_lock); |
428870ff BB |
5555 | ztest_record_enospc(FTAG); |
5556 | return (error); | |
34dc7c2f | 5557 | } |
428870ff | 5558 | ASSERT(error == 0 || error == EEXIST); |
34dc7c2f | 5559 | |
c99c9001 | 5560 | VERIFY0(dmu_objset_hold(name, zd, &os)); |
c242c188 | 5561 | (void) rw_exit(&ztest_name_lock); |
428870ff | 5562 | |
c242c188 | 5563 | ztest_zd_init(zd, ZTEST_GET_SHARED_DS(d), os); |
428870ff BB |
5564 | |
5565 | zilog = zd->zd_zilog; | |
5566 | ||
5567 | if (zilog->zl_header->zh_claim_lr_seq != 0 && | |
5568 | zilog->zl_header->zh_claim_lr_seq < committed_seq) | |
5569 | fatal(0, "missing log records: claimed %llu < committed %llu", | |
5570 | zilog->zl_header->zh_claim_lr_seq, committed_seq); | |
5571 | ||
5572 | ztest_dataset_dirobj_verify(zd); | |
5573 | ||
5574 | zil_replay(os, zd, ztest_replay_vector); | |
5575 | ||
5576 | ztest_dataset_dirobj_verify(zd); | |
5577 | ||
c242c188 | 5578 | if (ztest_opts.zo_verbose >= 6) |
428870ff BB |
5579 | (void) printf("%s replay %llu blocks, %llu records, seq %llu\n", |
5580 | zd->zd_name, | |
5581 | (u_longlong_t)zilog->zl_parse_blk_count, | |
5582 | (u_longlong_t)zilog->zl_parse_lr_count, | |
5583 | (u_longlong_t)zilog->zl_replaying_seq); | |
5584 | ||
5585 | zilog = zil_open(os, ztest_get_data); | |
5586 | ||
5587 | if (zilog->zl_replaying_seq != 0 && | |
5588 | zilog->zl_replaying_seq < committed_seq) | |
5589 | fatal(0, "missing log records: replayed %llu < committed %llu", | |
5590 | zilog->zl_replaying_seq, committed_seq); | |
5591 | ||
5592 | return (0); | |
5593 | } | |
5594 | ||
5595 | static void | |
c242c188 | 5596 | ztest_dataset_close(int d) |
428870ff | 5597 | { |
c242c188 | 5598 | ztest_ds_t *zd = &ztest_ds[d]; |
428870ff BB |
5599 | |
5600 | zil_close(zd->zd_zilog); | |
5601 | dmu_objset_rele(zd->zd_os, zd); | |
5602 | ||
5603 | ztest_zd_fini(zd); | |
34dc7c2f BB |
5604 | } |
5605 | ||
5606 | /* | |
5607 | * Kick off threads to run tests on all datasets in parallel. | |
5608 | */ | |
5609 | static void | |
428870ff | 5610 | ztest_run(ztest_shared_t *zs) |
34dc7c2f | 5611 | { |
1e33ac1e | 5612 | kt_did_t *tid; |
34dc7c2f | 5613 | spa_t *spa; |
3541dc6d | 5614 | objset_t *os; |
1e33ac1e BB |
5615 | kthread_t *resume_thread; |
5616 | uint64_t object; | |
428870ff | 5617 | int error; |
d6320ddb | 5618 | int t, d; |
b128c09f BB |
5619 | |
5620 | ztest_exiting = B_FALSE; | |
34dc7c2f | 5621 | |
34dc7c2f | 5622 | /* |
428870ff | 5623 | * Initialize parent/child shared state. |
34dc7c2f | 5624 | */ |
c242c188 CS |
5625 | mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); |
5626 | rw_init(&ztest_name_lock, NULL, RW_DEFAULT, NULL); | |
34dc7c2f | 5627 | |
428870ff | 5628 | zs->zs_thread_start = gethrtime(); |
c242c188 CS |
5629 | zs->zs_thread_stop = |
5630 | zs->zs_thread_start + ztest_opts.zo_passtime * NANOSEC; | |
428870ff BB |
5631 | zs->zs_thread_stop = MIN(zs->zs_thread_stop, zs->zs_proc_stop); |
5632 | zs->zs_thread_kill = zs->zs_thread_stop; | |
c242c188 CS |
5633 | if (ztest_random(100) < ztest_opts.zo_killrate) { |
5634 | zs->zs_thread_kill -= | |
5635 | ztest_random(ztest_opts.zo_passtime * NANOSEC); | |
5636 | } | |
34dc7c2f | 5637 | |
1e33ac1e | 5638 | mutex_init(&zcl.zcl_callbacks_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 5639 | |
428870ff BB |
5640 | list_create(&zcl.zcl_callbacks, sizeof (ztest_cb_data_t), |
5641 | offsetof(ztest_cb_data_t, zcd_node)); | |
34dc7c2f BB |
5642 | |
5643 | /* | |
b128c09f | 5644 | * Open our pool. |
34dc7c2f | 5645 | */ |
428870ff | 5646 | kernel_init(FREAD | FWRITE); |
c242c188 | 5647 | VERIFY(spa_open(ztest_opts.zo_pool, &spa, FTAG) == 0); |
6d974228 | 5648 | spa->spa_debug = B_TRUE; |
c242c188 | 5649 | ztest_spa = spa; |
428870ff | 5650 | |
c242c188 | 5651 | VERIFY3U(0, ==, dmu_objset_hold(ztest_opts.zo_pool, FTAG, &os)); |
3541dc6d GA |
5652 | zs->zs_guid = dmu_objset_fsid_guid(os); |
5653 | dmu_objset_rele(os, FTAG); | |
5654 | ||
428870ff | 5655 | spa->spa_dedup_ditto = 2 * ZIO_DEDUPDITTO_MIN; |
34dc7c2f | 5656 | |
fb5f0bc8 BB |
5657 | /* |
5658 | * We don't expect the pool to suspend unless maxfaults == 0, | |
5659 | * in which case ztest_fault_inject() temporarily takes away | |
5660 | * the only valid replica. | |
5661 | */ | |
428870ff | 5662 | if (MAXFAULTS() == 0) |
fb5f0bc8 BB |
5663 | spa->spa_failmode = ZIO_FAILURE_MODE_WAIT; |
5664 | else | |
5665 | spa->spa_failmode = ZIO_FAILURE_MODE_PANIC; | |
5666 | ||
34dc7c2f | 5667 | /* |
b128c09f | 5668 | * Create a thread to periodically resume suspended I/O. |
34dc7c2f | 5669 | */ |
0aebd4f9 ED |
5670 | VERIFY3P((resume_thread = zk_thread_create(NULL, 0, |
5671 | (thread_func_t)ztest_resume_thread, spa, TS_RUN, NULL, 0, 0, | |
5672 | PTHREAD_CREATE_JOINABLE)), !=, NULL); | |
34dc7c2f | 5673 | |
26099167 | 5674 | #if 0 |
428870ff | 5675 | /* |
1e33ac1e | 5676 | * Set a deadman alarm to abort() if we hang. |
428870ff | 5677 | */ |
1e33ac1e BB |
5678 | signal(SIGALRM, ztest_deadman_alarm); |
5679 | alarm((zs->zs_thread_stop - zs->zs_thread_start) / NANOSEC + GRACE); | |
26099167 | 5680 | #endif |
428870ff | 5681 | |
34dc7c2f BB |
5682 | /* |
5683 | * Verify that we can safely inquire about about any object, | |
5684 | * whether it's allocated or not. To make it interesting, | |
5685 | * we probe a 5-wide window around each power of two. | |
5686 | * This hits all edge cases, including zero and the max. | |
5687 | */ | |
d6320ddb BB |
5688 | for (t = 0; t < 64; t++) { |
5689 | for (d = -5; d <= 5; d++) { | |
34dc7c2f BB |
5690 | error = dmu_object_info(spa->spa_meta_objset, |
5691 | (1ULL << t) + d, NULL); | |
5692 | ASSERT(error == 0 || error == ENOENT || | |
5693 | error == EINVAL); | |
5694 | } | |
5695 | } | |
5696 | ||
5697 | /* | |
428870ff | 5698 | * If we got any ENOSPC errors on the previous run, destroy something. |
34dc7c2f | 5699 | */ |
428870ff | 5700 | if (zs->zs_enospc_count != 0) { |
c242c188 CS |
5701 | int d = ztest_random(ztest_opts.zo_datasets); |
5702 | ztest_dataset_destroy(d); | |
428870ff | 5703 | } |
34dc7c2f BB |
5704 | zs->zs_enospc_count = 0; |
5705 | ||
c242c188 CS |
5706 | tid = umem_zalloc(ztest_opts.zo_threads * sizeof (kt_did_t), |
5707 | UMEM_NOFAIL); | |
34dc7c2f | 5708 | |
c242c188 | 5709 | if (ztest_opts.zo_verbose >= 4) |
34dc7c2f BB |
5710 | (void) printf("starting main threads...\n"); |
5711 | ||
428870ff BB |
5712 | /* |
5713 | * Kick off all the tests that run in parallel. | |
5714 | */ | |
c242c188 | 5715 | for (t = 0; t < ztest_opts.zo_threads; t++) { |
1e33ac1e BB |
5716 | kthread_t *thread; |
5717 | ||
c242c188 CS |
5718 | if (t < ztest_opts.zo_datasets && |
5719 | ztest_dataset_open(t) != 0) | |
428870ff | 5720 | return; |
1e33ac1e | 5721 | |
0aebd4f9 ED |
5722 | VERIFY3P(thread = zk_thread_create(NULL, 0, |
5723 | (thread_func_t)ztest_thread, | |
5724 | (void *)(uintptr_t)t, TS_RUN, NULL, 0, 0, | |
5725 | PTHREAD_CREATE_JOINABLE), !=, NULL); | |
1e33ac1e | 5726 | tid[t] = thread->t_tid; |
34dc7c2f BB |
5727 | } |
5728 | ||
428870ff BB |
5729 | /* |
5730 | * Wait for all of the tests to complete. We go in reverse order | |
5731 | * so we don't close datasets while threads are still using them. | |
5732 | */ | |
c242c188 | 5733 | for (t = ztest_opts.zo_threads - 1; t >= 0; t--) { |
1e33ac1e | 5734 | thread_join(tid[t]); |
c242c188 CS |
5735 | if (t < ztest_opts.zo_datasets) |
5736 | ztest_dataset_close(t); | |
34dc7c2f BB |
5737 | } |
5738 | ||
34dc7c2f BB |
5739 | txg_wait_synced(spa_get_dsl(spa), 0); |
5740 | ||
428870ff BB |
5741 | zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(spa)); |
5742 | zs->zs_space = metaslab_class_get_space(spa_normal_class(spa)); | |
5743 | ||
c242c188 | 5744 | umem_free(tid, ztest_opts.zo_threads * sizeof (kt_did_t)); |
428870ff BB |
5745 | |
5746 | /* Kill the resume thread */ | |
5747 | ztest_exiting = B_TRUE; | |
1e33ac1e | 5748 | thread_join(resume_thread->t_tid); |
428870ff | 5749 | ztest_resume(spa); |
34dc7c2f BB |
5750 | |
5751 | /* | |
428870ff BB |
5752 | * Right before closing the pool, kick off a bunch of async I/O; |
5753 | * spa_close() should wait for it to complete. | |
34dc7c2f | 5754 | */ |
1e33ac1e | 5755 | for (object = 1; object < 50; object++) |
428870ff BB |
5756 | dmu_prefetch(spa->spa_meta_objset, object, 0, 1ULL << 20); |
5757 | ||
090ff092 RC |
5758 | /* Verify that at least one commit cb was called in a timely fashion */ |
5759 | if (zc_cb_counter >= ZTEST_COMMIT_CB_MIN_REG) | |
c99c9001 | 5760 | VERIFY0(zc_min_txg_delay); |
090ff092 | 5761 | |
428870ff BB |
5762 | spa_close(spa, FTAG); |
5763 | ||
5764 | /* | |
5765 | * Verify that we can loop over all pools. | |
5766 | */ | |
5767 | mutex_enter(&spa_namespace_lock); | |
5768 | for (spa = spa_next(NULL); spa != NULL; spa = spa_next(spa)) | |
c242c188 | 5769 | if (ztest_opts.zo_verbose > 3) |
428870ff BB |
5770 | (void) printf("spa_next: found %s\n", spa_name(spa)); |
5771 | mutex_exit(&spa_namespace_lock); | |
5772 | ||
5773 | /* | |
5774 | * Verify that we can export the pool and reimport it under a | |
5775 | * different name. | |
5776 | */ | |
5777 | if (ztest_random(2) == 0) { | |
5778 | char name[MAXNAMELEN]; | |
c242c188 CS |
5779 | (void) snprintf(name, MAXNAMELEN, "%s_import", |
5780 | ztest_opts.zo_pool); | |
5781 | ztest_spa_import_export(ztest_opts.zo_pool, name); | |
5782 | ztest_spa_import_export(name, ztest_opts.zo_pool); | |
428870ff BB |
5783 | } |
5784 | ||
5785 | kernel_fini(); | |
572e2857 BB |
5786 | |
5787 | list_destroy(&zcl.zcl_callbacks); | |
0e8d1b2d | 5788 | mutex_destroy(&zcl.zcl_callbacks_lock); |
c242c188 CS |
5789 | rw_destroy(&ztest_name_lock); |
5790 | mutex_destroy(&ztest_vdev_lock); | |
428870ff BB |
5791 | } |
5792 | ||
5793 | static void | |
c242c188 | 5794 | ztest_freeze(void) |
428870ff | 5795 | { |
c242c188 | 5796 | ztest_ds_t *zd = &ztest_ds[0]; |
428870ff BB |
5797 | spa_t *spa; |
5798 | int numloops = 0; | |
5799 | ||
c242c188 | 5800 | if (ztest_opts.zo_verbose >= 3) |
428870ff | 5801 | (void) printf("testing spa_freeze()...\n"); |
9babb374 | 5802 | |
428870ff | 5803 | kernel_init(FREAD | FWRITE); |
c242c188 CS |
5804 | VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
5805 | VERIFY3U(0, ==, ztest_dataset_open(0)); | |
9babb374 | 5806 | |
428870ff BB |
5807 | /* |
5808 | * Force the first log block to be transactionally allocated. | |
5809 | * We have to do this before we freeze the pool -- otherwise | |
5810 | * the log chain won't be anchored. | |
5811 | */ | |
5812 | while (BP_IS_HOLE(&zd->zd_zilog->zl_header->zh_log)) { | |
5813 | ztest_dmu_object_alloc_free(zd, 0); | |
572e2857 | 5814 | zil_commit(zd->zd_zilog, 0); |
34dc7c2f BB |
5815 | } |
5816 | ||
5817 | txg_wait_synced(spa_get_dsl(spa), 0); | |
5818 | ||
428870ff BB |
5819 | /* |
5820 | * Freeze the pool. This stops spa_sync() from doing anything, | |
5821 | * so that the only way to record changes from now on is the ZIL. | |
5822 | */ | |
5823 | spa_freeze(spa); | |
b128c09f | 5824 | |
428870ff BB |
5825 | /* |
5826 | * Run tests that generate log records but don't alter the pool config | |
5827 | * or depend on DSL sync tasks (snapshots, objset create/destroy, etc). | |
5828 | * We do a txg_wait_synced() after each iteration to force the txg | |
5829 | * to increase well beyond the last synced value in the uberblock. | |
5830 | * The ZIL should be OK with that. | |
5831 | */ | |
c242c188 CS |
5832 | while (ztest_random(10) != 0 && |
5833 | numloops++ < ztest_opts.zo_maxloops) { | |
428870ff BB |
5834 | ztest_dmu_write_parallel(zd, 0); |
5835 | ztest_dmu_object_alloc_free(zd, 0); | |
5836 | txg_wait_synced(spa_get_dsl(spa), 0); | |
5837 | } | |
b128c09f | 5838 | |
34dc7c2f | 5839 | /* |
428870ff | 5840 | * Commit all of the changes we just generated. |
34dc7c2f | 5841 | */ |
572e2857 | 5842 | zil_commit(zd->zd_zilog, 0); |
428870ff | 5843 | txg_wait_synced(spa_get_dsl(spa), 0); |
34dc7c2f | 5844 | |
428870ff BB |
5845 | /* |
5846 | * Close our dataset and close the pool. | |
5847 | */ | |
c242c188 | 5848 | ztest_dataset_close(0); |
34dc7c2f | 5849 | spa_close(spa, FTAG); |
428870ff | 5850 | kernel_fini(); |
34dc7c2f | 5851 | |
428870ff BB |
5852 | /* |
5853 | * Open and close the pool and dataset to induce log replay. | |
5854 | */ | |
5855 | kernel_init(FREAD | FWRITE); | |
c242c188 | 5856 | VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
29809a6c | 5857 | ASSERT(spa_freeze_txg(spa) == UINT64_MAX); |
c242c188 CS |
5858 | VERIFY3U(0, ==, ztest_dataset_open(0)); |
5859 | ztest_dataset_close(0); | |
3bc7e0fb GW |
5860 | |
5861 | spa->spa_debug = B_TRUE; | |
5862 | ztest_spa = spa; | |
5863 | txg_wait_synced(spa_get_dsl(spa), 0); | |
5864 | ztest_reguid(NULL, 0); | |
5865 | ||
428870ff | 5866 | spa_close(spa, FTAG); |
34dc7c2f BB |
5867 | kernel_fini(); |
5868 | } | |
5869 | ||
5870 | void | |
5871 | print_time(hrtime_t t, char *timebuf) | |
5872 | { | |
5873 | hrtime_t s = t / NANOSEC; | |
5874 | hrtime_t m = s / 60; | |
5875 | hrtime_t h = m / 60; | |
5876 | hrtime_t d = h / 24; | |
5877 | ||
5878 | s -= m * 60; | |
5879 | m -= h * 60; | |
5880 | h -= d * 24; | |
5881 | ||
5882 | timebuf[0] = '\0'; | |
5883 | ||
5884 | if (d) | |
5885 | (void) sprintf(timebuf, | |
5886 | "%llud%02lluh%02llum%02llus", d, h, m, s); | |
5887 | else if (h) | |
5888 | (void) sprintf(timebuf, "%lluh%02llum%02llus", h, m, s); | |
5889 | else if (m) | |
5890 | (void) sprintf(timebuf, "%llum%02llus", m, s); | |
5891 | else | |
5892 | (void) sprintf(timebuf, "%llus", s); | |
5893 | } | |
5894 | ||
428870ff | 5895 | static nvlist_t * |
0bc8fd78 | 5896 | make_random_props(void) |
428870ff BB |
5897 | { |
5898 | nvlist_t *props; | |
5899 | ||
428870ff | 5900 | VERIFY(nvlist_alloc(&props, NV_UNIQUE_NAME, 0) == 0); |
9ae529ec CS |
5901 | if (ztest_random(2) == 0) |
5902 | return (props); | |
428870ff BB |
5903 | VERIFY(nvlist_add_uint64(props, "autoreplace", 1) == 0); |
5904 | ||
428870ff BB |
5905 | return (props); |
5906 | } | |
5907 | ||
34dc7c2f BB |
5908 | /* |
5909 | * Create a storage pool with the given name and initial vdev size. | |
428870ff | 5910 | * Then test spa_freeze() functionality. |
34dc7c2f BB |
5911 | */ |
5912 | static void | |
428870ff | 5913 | ztest_init(ztest_shared_t *zs) |
34dc7c2f BB |
5914 | { |
5915 | spa_t *spa; | |
428870ff | 5916 | nvlist_t *nvroot, *props; |
9ae529ec | 5917 | int i; |
428870ff | 5918 | |
c242c188 CS |
5919 | mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); |
5920 | rw_init(&ztest_name_lock, NULL, RW_DEFAULT, NULL); | |
34dc7c2f BB |
5921 | |
5922 | kernel_init(FREAD | FWRITE); | |
5923 | ||
5924 | /* | |
5925 | * Create the storage pool. | |
5926 | */ | |
c242c188 | 5927 | (void) spa_destroy(ztest_opts.zo_pool); |
428870ff BB |
5928 | ztest_shared->zs_vdev_next_leaf = 0; |
5929 | zs->zs_splits = 0; | |
c242c188 | 5930 | zs->zs_mirrors = ztest_opts.zo_mirrors; |
ea0b2538 | 5931 | nvroot = make_vdev_root(NULL, NULL, NULL, ztest_opts.zo_vdev_size, 0, |
c242c188 | 5932 | 0, ztest_opts.zo_raidz, zs->zs_mirrors, 1); |
428870ff | 5933 | props = make_random_props(); |
9ae529ec CS |
5934 | for (i = 0; i < SPA_FEATURES; i++) { |
5935 | char *buf; | |
5936 | VERIFY3S(-1, !=, asprintf(&buf, "feature@%s", | |
5937 | spa_feature_table[i].fi_uname)); | |
5938 | VERIFY3U(0, ==, nvlist_add_uint64(props, buf, 0)); | |
5939 | free(buf); | |
5940 | } | |
c242c188 CS |
5941 | VERIFY3U(0, ==, spa_create(ztest_opts.zo_pool, nvroot, props, |
5942 | NULL, NULL)); | |
34dc7c2f BB |
5943 | nvlist_free(nvroot); |
5944 | ||
c242c188 CS |
5945 | VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
5946 | zs->zs_metaslab_sz = | |
5947 | 1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift; | |
34dc7c2f BB |
5948 | spa_close(spa, FTAG); |
5949 | ||
5950 | kernel_fini(); | |
428870ff | 5951 | |
c242c188 CS |
5952 | ztest_run_zdb(ztest_opts.zo_pool); |
5953 | ||
5954 | ztest_freeze(); | |
5955 | ||
5956 | ztest_run_zdb(ztest_opts.zo_pool); | |
5957 | ||
5958 | rw_destroy(&ztest_name_lock); | |
5959 | mutex_destroy(&ztest_vdev_lock); | |
5960 | } | |
5961 | ||
5962 | static void | |
9d81146b | 5963 | setup_data_fd(void) |
c242c188 | 5964 | { |
facbbe43 BB |
5965 | static char ztest_name_data[] = "/tmp/ztest.data.XXXXXX"; |
5966 | ||
5967 | ztest_fd_data = mkstemp(ztest_name_data); | |
9d81146b | 5968 | ASSERT3S(ztest_fd_data, >=, 0); |
facbbe43 | 5969 | (void) unlink(ztest_name_data); |
c242c188 CS |
5970 | } |
5971 | ||
22257dc0 CS |
5972 | static int |
5973 | shared_data_size(ztest_shared_hdr_t *hdr) | |
5974 | { | |
5975 | int size; | |
5976 | ||
5977 | size = hdr->zh_hdr_size; | |
5978 | size += hdr->zh_opts_size; | |
5979 | size += hdr->zh_size; | |
5980 | size += hdr->zh_stats_size * hdr->zh_stats_count; | |
5981 | size += hdr->zh_ds_size * hdr->zh_ds_count; | |
5982 | ||
5983 | return (size); | |
5984 | } | |
5985 | ||
c242c188 CS |
5986 | static void |
5987 | setup_hdr(void) | |
5988 | { | |
22257dc0 | 5989 | int size; |
c242c188 CS |
5990 | ztest_shared_hdr_t *hdr; |
5991 | ||
5992 | hdr = (void *)mmap(0, P2ROUNDUP(sizeof (*hdr), getpagesize()), | |
9d81146b | 5993 | PROT_READ | PROT_WRITE, MAP_SHARED, ztest_fd_data, 0); |
c242c188 CS |
5994 | ASSERT(hdr != MAP_FAILED); |
5995 | ||
9d81146b | 5996 | VERIFY3U(0, ==, ftruncate(ztest_fd_data, sizeof (ztest_shared_hdr_t))); |
22257dc0 | 5997 | |
c242c188 CS |
5998 | hdr->zh_hdr_size = sizeof (ztest_shared_hdr_t); |
5999 | hdr->zh_opts_size = sizeof (ztest_shared_opts_t); | |
6000 | hdr->zh_size = sizeof (ztest_shared_t); | |
6001 | hdr->zh_stats_size = sizeof (ztest_shared_callstate_t); | |
6002 | hdr->zh_stats_count = ZTEST_FUNCS; | |
6003 | hdr->zh_ds_size = sizeof (ztest_shared_ds_t); | |
6004 | hdr->zh_ds_count = ztest_opts.zo_datasets; | |
6005 | ||
22257dc0 | 6006 | size = shared_data_size(hdr); |
9d81146b | 6007 | VERIFY3U(0, ==, ftruncate(ztest_fd_data, size)); |
22257dc0 | 6008 | |
c242c188 CS |
6009 | (void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize())); |
6010 | } | |
6011 | ||
6012 | static void | |
6013 | setup_data(void) | |
6014 | { | |
6015 | int size, offset; | |
6016 | ztest_shared_hdr_t *hdr; | |
6017 | uint8_t *buf; | |
6018 | ||
6019 | hdr = (void *)mmap(0, P2ROUNDUP(sizeof (*hdr), getpagesize()), | |
9d81146b | 6020 | PROT_READ, MAP_SHARED, ztest_fd_data, 0); |
c242c188 CS |
6021 | ASSERT(hdr != MAP_FAILED); |
6022 | ||
22257dc0 | 6023 | size = shared_data_size(hdr); |
c242c188 CS |
6024 | |
6025 | (void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize())); | |
6026 | hdr = ztest_shared_hdr = (void *)mmap(0, P2ROUNDUP(size, getpagesize()), | |
9d81146b | 6027 | PROT_READ | PROT_WRITE, MAP_SHARED, ztest_fd_data, 0); |
c242c188 CS |
6028 | ASSERT(hdr != MAP_FAILED); |
6029 | buf = (uint8_t *)hdr; | |
6030 | ||
6031 | offset = hdr->zh_hdr_size; | |
6032 | ztest_shared_opts = (void *)&buf[offset]; | |
6033 | offset += hdr->zh_opts_size; | |
6034 | ztest_shared = (void *)&buf[offset]; | |
6035 | offset += hdr->zh_size; | |
6036 | ztest_shared_callstate = (void *)&buf[offset]; | |
6037 | offset += hdr->zh_stats_size * hdr->zh_stats_count; | |
6038 | ztest_shared_ds = (void *)&buf[offset]; | |
6039 | } | |
6040 | ||
6041 | static boolean_t | |
6042 | exec_child(char *cmd, char *libpath, boolean_t ignorekill, int *statusp) | |
6043 | { | |
6044 | pid_t pid; | |
6045 | int status; | |
483106eb | 6046 | char *cmdbuf = NULL; |
c242c188 CS |
6047 | |
6048 | pid = fork(); | |
6049 | ||
6050 | if (cmd == NULL) { | |
483106eb BB |
6051 | cmdbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
6052 | (void) strlcpy(cmdbuf, getexecname(), MAXPATHLEN); | |
c242c188 CS |
6053 | cmd = cmdbuf; |
6054 | } | |
6055 | ||
6056 | if (pid == -1) | |
6057 | fatal(1, "fork failed"); | |
6058 | ||
6059 | if (pid == 0) { /* child */ | |
6060 | char *emptyargv[2] = { cmd, NULL }; | |
9d81146b | 6061 | char fd_data_str[12]; |
c242c188 CS |
6062 | |
6063 | struct rlimit rl = { 1024, 1024 }; | |
6064 | (void) setrlimit(RLIMIT_NOFILE, &rl); | |
9d81146b ED |
6065 | |
6066 | (void) close(ztest_fd_rand); | |
6067 | VERIFY(11 >= snprintf(fd_data_str, 12, "%d", ztest_fd_data)); | |
6068 | VERIFY(0 == setenv("ZTEST_FD_DATA", fd_data_str, 1)); | |
6069 | ||
c242c188 CS |
6070 | (void) enable_extended_FILE_stdio(-1, -1); |
6071 | if (libpath != NULL) | |
6072 | VERIFY(0 == setenv("LD_LIBRARY_PATH", libpath, 1)); | |
6073 | (void) execv(cmd, emptyargv); | |
6074 | ztest_dump_core = B_FALSE; | |
6075 | fatal(B_TRUE, "exec failed: %s", cmd); | |
6076 | } | |
6077 | ||
483106eb BB |
6078 | if (cmdbuf != NULL) { |
6079 | umem_free(cmdbuf, MAXPATHLEN); | |
6080 | cmd = NULL; | |
6081 | } | |
6082 | ||
c242c188 CS |
6083 | while (waitpid(pid, &status, 0) != pid) |
6084 | continue; | |
6085 | if (statusp != NULL) | |
6086 | *statusp = status; | |
6087 | ||
6088 | if (WIFEXITED(status)) { | |
6089 | if (WEXITSTATUS(status) != 0) { | |
6090 | (void) fprintf(stderr, "child exited with code %d\n", | |
6091 | WEXITSTATUS(status)); | |
6092 | exit(2); | |
6093 | } | |
6094 | return (B_FALSE); | |
6095 | } else if (WIFSIGNALED(status)) { | |
6096 | if (!ignorekill || WTERMSIG(status) != SIGKILL) { | |
6097 | (void) fprintf(stderr, "child died with signal %d\n", | |
6098 | WTERMSIG(status)); | |
6099 | exit(3); | |
6100 | } | |
6101 | return (B_TRUE); | |
6102 | } else { | |
6103 | (void) fprintf(stderr, "something strange happened to child\n"); | |
6104 | exit(4); | |
6105 | /* NOTREACHED */ | |
6106 | } | |
6107 | } | |
6108 | ||
6109 | static void | |
6110 | ztest_run_init(void) | |
6111 | { | |
6112 | int i; | |
428870ff | 6113 | |
c242c188 | 6114 | ztest_shared_t *zs = ztest_shared; |
428870ff | 6115 | |
c242c188 | 6116 | ASSERT(ztest_opts.zo_init != 0); |
572e2857 | 6117 | |
c242c188 CS |
6118 | /* |
6119 | * Blow away any existing copy of zpool.cache | |
6120 | */ | |
6121 | (void) remove(spa_config_path); | |
6122 | ||
6123 | /* | |
6124 | * Create and initialize our storage pool. | |
6125 | */ | |
6126 | for (i = 1; i <= ztest_opts.zo_init; i++) { | |
6127 | bzero(zs, sizeof (ztest_shared_t)); | |
6128 | if (ztest_opts.zo_verbose >= 3 && | |
6129 | ztest_opts.zo_init != 1) { | |
6130 | (void) printf("ztest_init(), pass %d\n", i); | |
6131 | } | |
6132 | ztest_init(zs); | |
6133 | } | |
34dc7c2f BB |
6134 | } |
6135 | ||
6136 | int | |
6137 | main(int argc, char **argv) | |
6138 | { | |
6139 | int kills = 0; | |
6140 | int iters = 0; | |
c242c188 CS |
6141 | int older = 0; |
6142 | int newer = 0; | |
34dc7c2f BB |
6143 | ztest_shared_t *zs; |
6144 | ztest_info_t *zi; | |
c242c188 | 6145 | ztest_shared_callstate_t *zc; |
34dc7c2f BB |
6146 | char timebuf[100]; |
6147 | char numbuf[6]; | |
428870ff | 6148 | spa_t *spa; |
483106eb | 6149 | char *cmd; |
c242c188 CS |
6150 | boolean_t hasalt; |
6151 | int f; | |
9d81146b | 6152 | char *fd_data_str = getenv("ZTEST_FD_DATA"); |
34dc7c2f BB |
6153 | |
6154 | (void) setvbuf(stdout, NULL, _IOLBF, 0); | |
6155 | ||
9d81146b ED |
6156 | ztest_fd_rand = open("/dev/urandom", O_RDONLY); |
6157 | ASSERT3S(ztest_fd_rand, >=, 0); | |
6158 | ||
6159 | if (!fd_data_str) { | |
c242c188 CS |
6160 | dprintf_setup(&argc, argv); |
6161 | process_options(argc, argv); | |
34dc7c2f | 6162 | |
9d81146b | 6163 | setup_data_fd(); |
c242c188 CS |
6164 | setup_hdr(); |
6165 | setup_data(); | |
6166 | bcopy(&ztest_opts, ztest_shared_opts, | |
6167 | sizeof (*ztest_shared_opts)); | |
6168 | } else { | |
9d81146b | 6169 | ztest_fd_data = atoi(fd_data_str); |
c242c188 CS |
6170 | setup_data(); |
6171 | bcopy(ztest_shared_opts, &ztest_opts, sizeof (ztest_opts)); | |
6172 | } | |
6173 | ASSERT3U(ztest_opts.zo_datasets, ==, ztest_shared_hdr->zh_ds_count); | |
34dc7c2f | 6174 | |
428870ff | 6175 | /* Override location of zpool.cache */ |
5be98cfe BB |
6176 | VERIFY(asprintf((char **)&spa_config_path, "%s/zpool.cache", |
6177 | ztest_opts.zo_dir) != -1); | |
9d81146b | 6178 | |
c242c188 CS |
6179 | ztest_ds = umem_alloc(ztest_opts.zo_datasets * sizeof (ztest_ds_t), |
6180 | UMEM_NOFAIL); | |
6181 | zs = ztest_shared; | |
6182 | ||
9d81146b | 6183 | if (fd_data_str) { |
c242c188 CS |
6184 | metaslab_gang_bang = ztest_opts.zo_metaslab_gang_bang; |
6185 | metaslab_df_alloc_threshold = | |
6186 | zs->zs_metaslab_df_alloc_threshold; | |
6187 | ||
6188 | if (zs->zs_do_init) | |
6189 | ztest_run_init(); | |
6190 | else | |
6191 | ztest_run(zs); | |
6192 | exit(0); | |
6193 | } | |
34dc7c2f | 6194 | |
c242c188 | 6195 | hasalt = (strlen(ztest_opts.zo_alt_ztest) != 0); |
34dc7c2f | 6196 | |
c242c188 | 6197 | if (ztest_opts.zo_verbose >= 1) { |
34dc7c2f BB |
6198 | (void) printf("%llu vdevs, %d datasets, %d threads," |
6199 | " %llu seconds...\n", | |
c242c188 CS |
6200 | (u_longlong_t)ztest_opts.zo_vdevs, |
6201 | ztest_opts.zo_datasets, | |
6202 | ztest_opts.zo_threads, | |
6203 | (u_longlong_t)ztest_opts.zo_time); | |
34dc7c2f BB |
6204 | } |
6205 | ||
483106eb BB |
6206 | cmd = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); |
6207 | (void) strlcpy(cmd, getexecname(), MAXNAMELEN); | |
c242c188 CS |
6208 | |
6209 | zs->zs_do_init = B_TRUE; | |
6210 | if (strlen(ztest_opts.zo_alt_ztest) != 0) { | |
6211 | if (ztest_opts.zo_verbose >= 1) { | |
6212 | (void) printf("Executing older ztest for " | |
6213 | "initialization: %s\n", ztest_opts.zo_alt_ztest); | |
6214 | } | |
6215 | VERIFY(!exec_child(ztest_opts.zo_alt_ztest, | |
6216 | ztest_opts.zo_alt_libpath, B_FALSE, NULL)); | |
6217 | } else { | |
6218 | VERIFY(!exec_child(NULL, NULL, B_FALSE, NULL)); | |
34dc7c2f | 6219 | } |
c242c188 | 6220 | zs->zs_do_init = B_FALSE; |
34dc7c2f | 6221 | |
428870ff | 6222 | zs->zs_proc_start = gethrtime(); |
c242c188 | 6223 | zs->zs_proc_stop = zs->zs_proc_start + ztest_opts.zo_time * NANOSEC; |
34dc7c2f | 6224 | |
d6320ddb | 6225 | for (f = 0; f < ZTEST_FUNCS; f++) { |
c242c188 CS |
6226 | zi = &ztest_info[f]; |
6227 | zc = ZTEST_GET_SHARED_CALLSTATE(f); | |
428870ff | 6228 | if (zs->zs_proc_start + zi->zi_interval[0] > zs->zs_proc_stop) |
c242c188 | 6229 | zc->zc_next = UINT64_MAX; |
34dc7c2f | 6230 | else |
c242c188 | 6231 | zc->zc_next = zs->zs_proc_start + |
428870ff | 6232 | ztest_random(2 * zi->zi_interval[0] + 1); |
34dc7c2f BB |
6233 | } |
6234 | ||
34dc7c2f BB |
6235 | /* |
6236 | * Run the tests in a loop. These tests include fault injection | |
6237 | * to verify that self-healing data works, and forced crashes | |
6238 | * to verify that we never lose on-disk consistency. | |
6239 | */ | |
428870ff | 6240 | while (gethrtime() < zs->zs_proc_stop) { |
34dc7c2f | 6241 | int status; |
c242c188 | 6242 | boolean_t killed; |
34dc7c2f BB |
6243 | |
6244 | /* | |
6245 | * Initialize the workload counters for each function. | |
6246 | */ | |
d6320ddb | 6247 | for (f = 0; f < ZTEST_FUNCS; f++) { |
c242c188 CS |
6248 | zc = ZTEST_GET_SHARED_CALLSTATE(f); |
6249 | zc->zc_count = 0; | |
6250 | zc->zc_time = 0; | |
34dc7c2f BB |
6251 | } |
6252 | ||
9babb374 | 6253 | /* Set the allocation switch size */ |
c242c188 CS |
6254 | zs->zs_metaslab_df_alloc_threshold = |
6255 | ztest_random(zs->zs_metaslab_sz / 4) + 1; | |
9babb374 | 6256 | |
c242c188 CS |
6257 | if (!hasalt || ztest_random(2) == 0) { |
6258 | if (hasalt && ztest_opts.zo_verbose >= 1) { | |
6259 | (void) printf("Executing newer ztest: %s\n", | |
6260 | cmd); | |
34dc7c2f | 6261 | } |
c242c188 CS |
6262 | newer++; |
6263 | killed = exec_child(cmd, NULL, B_TRUE, &status); | |
34dc7c2f | 6264 | } else { |
c242c188 CS |
6265 | if (hasalt && ztest_opts.zo_verbose >= 1) { |
6266 | (void) printf("Executing older ztest: %s\n", | |
6267 | ztest_opts.zo_alt_ztest); | |
6268 | } | |
6269 | older++; | |
6270 | killed = exec_child(ztest_opts.zo_alt_ztest, | |
6271 | ztest_opts.zo_alt_libpath, B_TRUE, &status); | |
34dc7c2f BB |
6272 | } |
6273 | ||
c242c188 CS |
6274 | if (killed) |
6275 | kills++; | |
34dc7c2f BB |
6276 | iters++; |
6277 | ||
c242c188 | 6278 | if (ztest_opts.zo_verbose >= 1) { |
34dc7c2f BB |
6279 | hrtime_t now = gethrtime(); |
6280 | ||
428870ff BB |
6281 | now = MIN(now, zs->zs_proc_stop); |
6282 | print_time(zs->zs_proc_stop - now, timebuf); | |
34dc7c2f BB |
6283 | nicenum(zs->zs_space, numbuf); |
6284 | ||
6285 | (void) printf("Pass %3d, %8s, %3llu ENOSPC, " | |
6286 | "%4.1f%% of %5s used, %3.0f%% done, %8s to go\n", | |
6287 | iters, | |
6288 | WIFEXITED(status) ? "Complete" : "SIGKILL", | |
6289 | (u_longlong_t)zs->zs_enospc_count, | |
6290 | 100.0 * zs->zs_alloc / zs->zs_space, | |
6291 | numbuf, | |
428870ff | 6292 | 100.0 * (now - zs->zs_proc_start) / |
c242c188 | 6293 | (ztest_opts.zo_time * NANOSEC), timebuf); |
34dc7c2f BB |
6294 | } |
6295 | ||
c242c188 | 6296 | if (ztest_opts.zo_verbose >= 2) { |
34dc7c2f BB |
6297 | (void) printf("\nWorkload summary:\n\n"); |
6298 | (void) printf("%7s %9s %s\n", | |
6299 | "Calls", "Time", "Function"); | |
6300 | (void) printf("%7s %9s %s\n", | |
6301 | "-----", "----", "--------"); | |
d6320ddb | 6302 | for (f = 0; f < ZTEST_FUNCS; f++) { |
34dc7c2f BB |
6303 | Dl_info dli; |
6304 | ||
c242c188 CS |
6305 | zi = &ztest_info[f]; |
6306 | zc = ZTEST_GET_SHARED_CALLSTATE(f); | |
6307 | print_time(zc->zc_time, timebuf); | |
34dc7c2f BB |
6308 | (void) dladdr((void *)zi->zi_func, &dli); |
6309 | (void) printf("%7llu %9s %s\n", | |
c242c188 | 6310 | (u_longlong_t)zc->zc_count, timebuf, |
34dc7c2f BB |
6311 | dli.dli_sname); |
6312 | } | |
6313 | (void) printf("\n"); | |
6314 | } | |
6315 | ||
6316 | /* | |
428870ff BB |
6317 | * It's possible that we killed a child during a rename test, |
6318 | * in which case we'll have a 'ztest_tmp' pool lying around | |
6319 | * instead of 'ztest'. Do a blind rename in case this happened. | |
34dc7c2f | 6320 | */ |
428870ff | 6321 | kernel_init(FREAD); |
c242c188 | 6322 | if (spa_open(ztest_opts.zo_pool, &spa, FTAG) == 0) { |
428870ff BB |
6323 | spa_close(spa, FTAG); |
6324 | } else { | |
6325 | char tmpname[MAXNAMELEN]; | |
6326 | kernel_fini(); | |
6327 | kernel_init(FREAD | FWRITE); | |
6328 | (void) snprintf(tmpname, sizeof (tmpname), "%s_tmp", | |
c242c188 CS |
6329 | ztest_opts.zo_pool); |
6330 | (void) spa_rename(tmpname, ztest_opts.zo_pool); | |
428870ff | 6331 | } |
34dc7c2f | 6332 | kernel_fini(); |
34dc7c2f | 6333 | |
c242c188 | 6334 | ztest_run_zdb(ztest_opts.zo_pool); |
428870ff | 6335 | } |
34dc7c2f | 6336 | |
c242c188 CS |
6337 | if (ztest_opts.zo_verbose >= 1) { |
6338 | if (hasalt) { | |
6339 | (void) printf("%d runs of older ztest: %s\n", older, | |
6340 | ztest_opts.zo_alt_ztest); | |
6341 | (void) printf("%d runs of newer ztest: %s\n", newer, | |
6342 | cmd); | |
6343 | } | |
34dc7c2f BB |
6344 | (void) printf("%d killed, %d completed, %.0f%% kill rate\n", |
6345 | kills, iters - kills, (100.0 * kills) / MAX(1, iters)); | |
6346 | } | |
6347 | ||
483106eb BB |
6348 | umem_free(cmd, MAXNAMELEN); |
6349 | ||
34dc7c2f BB |
6350 | return (0); |
6351 | } |