]>
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. |
cc99f275 | 23 | * Copyright (c) 2011, 2018 by Delphix. All rights reserved. |
3541dc6d | 24 | * Copyright 2011 Nexenta Systems, Inc. All rights reserved. |
95fd54a1 | 25 | * Copyright (c) 2013 Steven Hartland. All rights reserved. |
f3c8c9e6 JK |
26 | * Copyright (c) 2014 Integros [integros.com] |
27 | * Copyright 2017 Joyent, Inc. | |
cc99f275 | 28 | * Copyright (c) 2017, Intel Corporation. |
34dc7c2f BB |
29 | */ |
30 | ||
34dc7c2f BB |
31 | /* |
32 | * The objective of this program is to provide a DMU/ZAP/SPA stress test | |
33 | * that runs entirely in userland, is easy to use, and easy to extend. | |
34 | * | |
35 | * The overall design of the ztest program is as follows: | |
36 | * | |
37 | * (1) For each major functional area (e.g. adding vdevs to a pool, | |
38 | * creating and destroying datasets, reading and writing objects, etc) | |
39 | * we have a simple routine to test that functionality. These | |
40 | * individual routines do not have to do anything "stressful". | |
41 | * | |
42 | * (2) We turn these simple functionality tests into a stress test by | |
43 | * running them all in parallel, with as many threads as desired, | |
44 | * and spread across as many datasets, objects, and vdevs as desired. | |
45 | * | |
46 | * (3) While all this is happening, we inject faults into the pool to | |
47 | * verify that self-healing data really works. | |
48 | * | |
49 | * (4) Every time we open a dataset, we change its checksum and compression | |
50 | * functions. Thus even individual objects vary from block to block | |
51 | * in which checksum they use and whether they're compressed. | |
52 | * | |
53 | * (5) To verify that we never lose on-disk consistency after a crash, | |
54 | * we run the entire test in a child of the main process. | |
55 | * At random times, the child self-immolates with a SIGKILL. | |
56 | * This is the software equivalent of pulling the power cord. | |
57 | * The parent then runs the test again, using the existing | |
9b67f605 | 58 | * storage pool, as many times as desired. If backwards compatibility |
c242c188 CS |
59 | * testing is enabled ztest will sometimes run the "older" version |
60 | * of ztest after a SIGKILL. | |
34dc7c2f BB |
61 | * |
62 | * (6) To verify that we don't have future leaks or temporal incursions, | |
63 | * many of the functional tests record the transaction group number | |
64 | * as part of their data. When reading old data, they verify that | |
65 | * the transaction group number is less than the current, open txg. | |
66 | * If you add a new test, please do this if applicable. | |
67 | * | |
1e33ac1e BB |
68 | * (7) Threads are created with a reduced stack size, for sanity checking. |
69 | * Therefore, it's important not to allocate huge buffers on the stack. | |
70 | * | |
34dc7c2f BB |
71 | * When run with no arguments, ztest runs for about five minutes and |
72 | * produces no output if successful. To get a little bit of information, | |
73 | * specify -V. To get more information, specify -VV, and so on. | |
74 | * | |
75 | * To turn this into an overnight stress test, use -T to specify run time. | |
76 | * | |
ad0b23b1 | 77 | * You can ask more vdevs [-v], datasets [-d], or threads [-t] |
34dc7c2f BB |
78 | * to increase the pool capacity, fanout, and overall stress level. |
79 | * | |
c242c188 CS |
80 | * Use the -k option to set the desired frequency of kills. |
81 | * | |
82 | * When ztest invokes itself it passes all relevant information through a | |
83 | * temporary file which is mmap-ed in the child process. This allows shared | |
84 | * memory to survive the exec syscall. The ztest_shared_hdr_t struct is always | |
85 | * stored at offset 0 of this file and contains information on the size and | |
86 | * number of shared structures in the file. The information stored in this file | |
87 | * must remain backwards compatible with older versions of ztest so that | |
88 | * ztest can invoke them during backwards compatibility testing (-B). | |
34dc7c2f BB |
89 | */ |
90 | ||
91 | #include <sys/zfs_context.h> | |
92 | #include <sys/spa.h> | |
93 | #include <sys/dmu.h> | |
94 | #include <sys/txg.h> | |
9babb374 | 95 | #include <sys/dbuf.h> |
34dc7c2f | 96 | #include <sys/zap.h> |
34dc7c2f BB |
97 | #include <sys/dmu_objset.h> |
98 | #include <sys/poll.h> | |
99 | #include <sys/stat.h> | |
100 | #include <sys/time.h> | |
101 | #include <sys/wait.h> | |
102 | #include <sys/mman.h> | |
103 | #include <sys/resource.h> | |
104 | #include <sys/zio.h> | |
34dc7c2f | 105 | #include <sys/zil.h> |
428870ff | 106 | #include <sys/zil_impl.h> |
34dc7c2f | 107 | #include <sys/vdev_impl.h> |
b128c09f | 108 | #include <sys/vdev_file.h> |
619f0976 | 109 | #include <sys/vdev_initialize.h> |
e5db3134 | 110 | #include <sys/vdev_raidz.h> |
1b939560 | 111 | #include <sys/vdev_trim.h> |
34dc7c2f | 112 | #include <sys/spa_impl.h> |
428870ff | 113 | #include <sys/metaslab_impl.h> |
34dc7c2f | 114 | #include <sys/dsl_prop.h> |
9babb374 | 115 | #include <sys/dsl_dataset.h> |
13fe0198 | 116 | #include <sys/dsl_destroy.h> |
428870ff BB |
117 | #include <sys/dsl_scan.h> |
118 | #include <sys/zio_checksum.h> | |
27d96d22 | 119 | #include <sys/zfs_refcount.h> |
9ae529ec | 120 | #include <sys/zfeature.h> |
13fe0198 | 121 | #include <sys/dsl_userhold.h> |
a6255b7f | 122 | #include <sys/abd.h> |
34dc7c2f | 123 | #include <stdio.h> |
34dc7c2f BB |
124 | #include <stdlib.h> |
125 | #include <unistd.h> | |
126 | #include <signal.h> | |
127 | #include <umem.h> | |
34dc7c2f BB |
128 | #include <ctype.h> |
129 | #include <math.h> | |
130 | #include <sys/fs/zfs.h> | |
1eeb4562 | 131 | #include <zfs_fletcher.h> |
428870ff | 132 | #include <libnvpair.h> |
e89f1295 | 133 | #include <libzutil.h> |
010d1247 | 134 | #include <sys/crypto/icp.h> |
967798d0 | 135 | #ifdef __GLIBC__ |
e82cdc3a NB |
136 | #include <execinfo.h> /* for backtrace() */ |
137 | #endif | |
34dc7c2f | 138 | |
9d81146b ED |
139 | static int ztest_fd_data = -1; |
140 | static int ztest_fd_rand = -1; | |
c242c188 CS |
141 | |
142 | typedef struct ztest_shared_hdr { | |
143 | uint64_t zh_hdr_size; | |
144 | uint64_t zh_opts_size; | |
145 | uint64_t zh_size; | |
146 | uint64_t zh_stats_size; | |
147 | uint64_t zh_stats_count; | |
148 | uint64_t zh_ds_size; | |
149 | uint64_t zh_ds_count; | |
150 | } ztest_shared_hdr_t; | |
151 | ||
152 | static ztest_shared_hdr_t *ztest_shared_hdr; | |
153 | ||
cc99f275 DB |
154 | enum ztest_class_state { |
155 | ZTEST_VDEV_CLASS_OFF, | |
156 | ZTEST_VDEV_CLASS_ON, | |
157 | ZTEST_VDEV_CLASS_RND | |
158 | }; | |
159 | ||
c242c188 | 160 | typedef struct ztest_shared_opts { |
eca7b760 IK |
161 | char zo_pool[ZFS_MAX_DATASET_NAME_LEN]; |
162 | char zo_dir[ZFS_MAX_DATASET_NAME_LEN]; | |
c242c188 CS |
163 | char zo_alt_ztest[MAXNAMELEN]; |
164 | char zo_alt_libpath[MAXNAMELEN]; | |
165 | uint64_t zo_vdevs; | |
166 | uint64_t zo_vdevtime; | |
167 | size_t zo_vdev_size; | |
168 | int zo_ashift; | |
169 | int zo_mirrors; | |
170 | int zo_raidz; | |
171 | int zo_raidz_parity; | |
172 | int zo_datasets; | |
173 | int zo_threads; | |
174 | uint64_t zo_passtime; | |
175 | uint64_t zo_killrate; | |
176 | int zo_verbose; | |
177 | int zo_init; | |
178 | uint64_t zo_time; | |
179 | uint64_t zo_maxloops; | |
d830d479 | 180 | uint64_t zo_metaslab_force_ganging; |
379ca9cf | 181 | int zo_mmp_test; |
cc99f275 | 182 | int zo_special_vdevs; |
ab4c009e | 183 | int zo_dump_dbgmsg; |
c242c188 CS |
184 | } ztest_shared_opts_t; |
185 | ||
186 | static const ztest_shared_opts_t ztest_opts_defaults = { | |
caacc6e4 TH |
187 | .zo_pool = "ztest", |
188 | .zo_dir = "/tmp", | |
c242c188 CS |
189 | .zo_alt_ztest = { '\0' }, |
190 | .zo_alt_libpath = { '\0' }, | |
191 | .zo_vdevs = 5, | |
192 | .zo_ashift = SPA_MINBLOCKSHIFT, | |
193 | .zo_mirrors = 2, | |
194 | .zo_raidz = 4, | |
195 | .zo_raidz_parity = 1, | |
4e21fd06 | 196 | .zo_vdev_size = SPA_MINDEVSIZE * 4, /* 256m default size */ |
c242c188 CS |
197 | .zo_datasets = 7, |
198 | .zo_threads = 23, | |
199 | .zo_passtime = 60, /* 60 seconds */ | |
200 | .zo_killrate = 70, /* 70% kill rate */ | |
201 | .zo_verbose = 0, | |
379ca9cf | 202 | .zo_mmp_test = 0, |
c242c188 CS |
203 | .zo_init = 1, |
204 | .zo_time = 300, /* 5 minutes */ | |
205 | .zo_maxloops = 50, /* max loops during spa_freeze() */ | |
09b85f2d | 206 | .zo_metaslab_force_ganging = 64 << 10, |
cc99f275 | 207 | .zo_special_vdevs = ZTEST_VDEV_CLASS_RND, |
c242c188 CS |
208 | }; |
209 | ||
d830d479 | 210 | extern uint64_t metaslab_force_ganging; |
c242c188 | 211 | extern uint64_t metaslab_df_alloc_threshold; |
8fb1ede1 | 212 | extern unsigned long zfs_deadman_synctime_ms; |
080b3100 | 213 | extern int metaslab_preload_limit; |
d3c2ae1c | 214 | extern boolean_t zfs_compressed_arc_enabled; |
047116ac TC |
215 | extern int zfs_abd_scatter_enabled; |
216 | extern int dmu_object_alloc_chunk_shift; | |
4d044c4c | 217 | extern boolean_t zfs_force_some_double_word_sm_entries; |
c3bd3fb4 | 218 | extern unsigned long zio_decompress_fail_fraction; |
1258bd77 | 219 | extern unsigned long zfs_reconstruct_indirect_damage_fraction; |
65ca2c1e | 220 | |
c242c188 CS |
221 | |
222 | static ztest_shared_opts_t *ztest_shared_opts; | |
223 | static ztest_shared_opts_t ztest_opts; | |
4807c0ba | 224 | static char *ztest_wkeydata = "abcdefghijklmnopqrstuvwxyz012345"; |
c242c188 CS |
225 | |
226 | typedef struct ztest_shared_ds { | |
227 | uint64_t zd_seq; | |
228 | } ztest_shared_ds_t; | |
229 | ||
230 | static ztest_shared_ds_t *ztest_shared_ds; | |
231 | #define ZTEST_GET_SHARED_DS(d) (&ztest_shared_ds[d]) | |
428870ff BB |
232 | |
233 | #define BT_MAGIC 0x123456789abcdefULL | |
60b82074 BB |
234 | #define MAXFAULTS(zs) \ |
235 | (MAX((zs)->zs_mirrors, 1) * (ztest_opts.zo_raidz_parity + 1) - 1) | |
428870ff BB |
236 | |
237 | enum ztest_io_type { | |
238 | ZTEST_IO_WRITE_TAG, | |
239 | ZTEST_IO_WRITE_PATTERN, | |
240 | ZTEST_IO_WRITE_ZEROES, | |
241 | ZTEST_IO_TRUNCATE, | |
242 | ZTEST_IO_SETATTR, | |
03c6040b | 243 | ZTEST_IO_REWRITE, |
428870ff BB |
244 | ZTEST_IO_TYPES |
245 | }; | |
34dc7c2f BB |
246 | |
247 | typedef struct ztest_block_tag { | |
428870ff | 248 | uint64_t bt_magic; |
34dc7c2f BB |
249 | uint64_t bt_objset; |
250 | uint64_t bt_object; | |
50c957f7 | 251 | uint64_t bt_dnodesize; |
34dc7c2f | 252 | uint64_t bt_offset; |
428870ff | 253 | uint64_t bt_gen; |
34dc7c2f | 254 | uint64_t bt_txg; |
428870ff | 255 | uint64_t bt_crtxg; |
34dc7c2f BB |
256 | } ztest_block_tag_t; |
257 | ||
428870ff BB |
258 | typedef struct bufwad { |
259 | uint64_t bw_index; | |
260 | uint64_t bw_txg; | |
261 | uint64_t bw_data; | |
262 | } bufwad_t; | |
263 | ||
5d43cc9a MA |
264 | /* |
265 | * It would be better to use a rangelock_t per object. Unfortunately | |
266 | * the rangelock_t is not a drop-in replacement for rl_t, because we | |
267 | * still need to map from object ID to rangelock_t. | |
268 | */ | |
269 | typedef enum { | |
270 | RL_READER, | |
271 | RL_WRITER, | |
272 | RL_APPEND | |
273 | } rl_type_t; | |
274 | ||
428870ff BB |
275 | typedef struct rll { |
276 | void *rll_writer; | |
277 | int rll_readers; | |
1e33ac1e BB |
278 | kmutex_t rll_lock; |
279 | kcondvar_t rll_cv; | |
428870ff BB |
280 | } rll_t; |
281 | ||
5d43cc9a MA |
282 | typedef struct rl { |
283 | uint64_t rl_object; | |
284 | uint64_t rl_offset; | |
285 | uint64_t rl_size; | |
286 | rll_t *rl_lock; | |
287 | } rl_t; | |
428870ff BB |
288 | |
289 | #define ZTEST_RANGE_LOCKS 64 | |
290 | #define ZTEST_OBJECT_LOCKS 64 | |
291 | ||
292 | /* | |
293 | * Object descriptor. Used as a template for object lookup/create/remove. | |
294 | */ | |
295 | typedef struct ztest_od { | |
296 | uint64_t od_dir; | |
297 | uint64_t od_object; | |
298 | dmu_object_type_t od_type; | |
299 | dmu_object_type_t od_crtype; | |
300 | uint64_t od_blocksize; | |
301 | uint64_t od_crblocksize; | |
50c957f7 | 302 | uint64_t od_crdnodesize; |
428870ff BB |
303 | uint64_t od_gen; |
304 | uint64_t od_crgen; | |
eca7b760 | 305 | char od_name[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff | 306 | } ztest_od_t; |
34dc7c2f | 307 | |
428870ff BB |
308 | /* |
309 | * Per-dataset state. | |
310 | */ | |
311 | typedef struct ztest_ds { | |
c242c188 | 312 | ztest_shared_ds_t *zd_shared; |
428870ff | 313 | objset_t *zd_os; |
2d9142c9 | 314 | pthread_rwlock_t zd_zilog_lock; |
428870ff | 315 | zilog_t *zd_zilog; |
428870ff | 316 | ztest_od_t *zd_od; /* debugging aid */ |
eca7b760 | 317 | char zd_name[ZFS_MAX_DATASET_NAME_LEN]; |
1e33ac1e | 318 | kmutex_t zd_dirobj_lock; |
428870ff | 319 | rll_t zd_object_lock[ZTEST_OBJECT_LOCKS]; |
5d43cc9a | 320 | rll_t zd_range_lock[ZTEST_RANGE_LOCKS]; |
428870ff BB |
321 | } ztest_ds_t; |
322 | ||
323 | /* | |
324 | * Per-iteration state. | |
325 | */ | |
326 | typedef void ztest_func_t(ztest_ds_t *zd, uint64_t id); | |
327 | ||
328 | typedef struct ztest_info { | |
329 | ztest_func_t *zi_func; /* test function */ | |
330 | uint64_t zi_iters; /* iterations per execution */ | |
331 | uint64_t *zi_interval; /* execute every <interval> seconds */ | |
fdc5d982 | 332 | const char *zi_funcname; /* name of test function */ |
428870ff | 333 | } ztest_info_t; |
34dc7c2f | 334 | |
c242c188 CS |
335 | typedef struct ztest_shared_callstate { |
336 | uint64_t zc_count; /* per-pass count */ | |
337 | uint64_t zc_time; /* per-pass time */ | |
338 | uint64_t zc_next; /* next time to call this function */ | |
339 | } ztest_shared_callstate_t; | |
340 | ||
341 | static ztest_shared_callstate_t *ztest_shared_callstate; | |
342 | #define ZTEST_GET_SHARED_CALLSTATE(c) (&ztest_shared_callstate[c]) | |
343 | ||
34dc7c2f BB |
344 | ztest_func_t ztest_dmu_read_write; |
345 | ztest_func_t ztest_dmu_write_parallel; | |
346 | ztest_func_t ztest_dmu_object_alloc_free; | |
047116ac | 347 | ztest_func_t ztest_dmu_object_next_chunk; |
428870ff | 348 | ztest_func_t ztest_dmu_commit_callbacks; |
34dc7c2f BB |
349 | ztest_func_t ztest_zap; |
350 | ztest_func_t ztest_zap_parallel; | |
428870ff | 351 | ztest_func_t ztest_zil_commit; |
3e31d2b0 | 352 | ztest_func_t ztest_zil_remount; |
428870ff | 353 | ztest_func_t ztest_dmu_read_write_zcopy; |
34dc7c2f | 354 | ztest_func_t ztest_dmu_objset_create_destroy; |
428870ff BB |
355 | ztest_func_t ztest_dmu_prealloc; |
356 | ztest_func_t ztest_fzap; | |
34dc7c2f | 357 | ztest_func_t ztest_dmu_snapshot_create_destroy; |
428870ff BB |
358 | ztest_func_t ztest_dsl_prop_get_set; |
359 | ztest_func_t ztest_spa_prop_get_set; | |
34dc7c2f BB |
360 | ztest_func_t ztest_spa_create_destroy; |
361 | ztest_func_t ztest_fault_inject; | |
428870ff | 362 | ztest_func_t ztest_dmu_snapshot_hold; |
0582e403 | 363 | ztest_func_t ztest_mmp_enable_disable; |
428870ff BB |
364 | ztest_func_t ztest_scrub; |
365 | ztest_func_t ztest_dsl_dataset_promote_busy; | |
34dc7c2f BB |
366 | ztest_func_t ztest_vdev_attach_detach; |
367 | ztest_func_t ztest_vdev_LUN_growth; | |
368 | ztest_func_t ztest_vdev_add_remove; | |
cc99f275 | 369 | ztest_func_t ztest_vdev_class_add; |
b128c09f | 370 | ztest_func_t ztest_vdev_aux_add_remove; |
428870ff | 371 | ztest_func_t ztest_split_pool; |
3541dc6d | 372 | ztest_func_t ztest_reguid; |
ea0b2538 | 373 | ztest_func_t ztest_spa_upgrade; |
a1d477c2 | 374 | ztest_func_t ztest_device_removal; |
d2734cce | 375 | ztest_func_t ztest_spa_checkpoint_create_discard; |
619f0976 | 376 | ztest_func_t ztest_initialize; |
1b939560 | 377 | ztest_func_t ztest_trim; |
1eeb4562 | 378 | ztest_func_t ztest_fletcher; |
37f520db | 379 | ztest_func_t ztest_fletcher_incr; |
50c957f7 | 380 | ztest_func_t ztest_verify_dnode_bt; |
34dc7c2f | 381 | |
428870ff BB |
382 | uint64_t zopt_always = 0ULL * NANOSEC; /* all the time */ |
383 | uint64_t zopt_incessant = 1ULL * NANOSEC / 10; /* every 1/10 second */ | |
384 | uint64_t zopt_often = 1ULL * NANOSEC; /* every second */ | |
385 | uint64_t zopt_sometimes = 10ULL * NANOSEC; /* every 10 seconds */ | |
386 | uint64_t zopt_rarely = 60ULL * NANOSEC; /* every 60 seconds */ | |
34dc7c2f | 387 | |
fdc5d982 TC |
388 | #define ZTI_INIT(func, iters, interval) \ |
389 | { .zi_func = (func), \ | |
390 | .zi_iters = (iters), \ | |
391 | .zi_interval = (interval), \ | |
392 | .zi_funcname = # func } | |
393 | ||
34dc7c2f | 394 | ztest_info_t ztest_info[] = { |
fdc5d982 TC |
395 | ZTI_INIT(ztest_dmu_read_write, 1, &zopt_always), |
396 | ZTI_INIT(ztest_dmu_write_parallel, 10, &zopt_always), | |
397 | ZTI_INIT(ztest_dmu_object_alloc_free, 1, &zopt_always), | |
047116ac | 398 | ZTI_INIT(ztest_dmu_object_next_chunk, 1, &zopt_sometimes), |
fdc5d982 TC |
399 | ZTI_INIT(ztest_dmu_commit_callbacks, 1, &zopt_always), |
400 | ZTI_INIT(ztest_zap, 30, &zopt_always), | |
401 | ZTI_INIT(ztest_zap_parallel, 100, &zopt_always), | |
402 | ZTI_INIT(ztest_split_pool, 1, &zopt_always), | |
403 | ZTI_INIT(ztest_zil_commit, 1, &zopt_incessant), | |
404 | ZTI_INIT(ztest_zil_remount, 1, &zopt_sometimes), | |
405 | ZTI_INIT(ztest_dmu_read_write_zcopy, 1, &zopt_often), | |
406 | ZTI_INIT(ztest_dmu_objset_create_destroy, 1, &zopt_often), | |
407 | ZTI_INIT(ztest_dsl_prop_get_set, 1, &zopt_often), | |
408 | ZTI_INIT(ztest_spa_prop_get_set, 1, &zopt_sometimes), | |
428870ff | 409 | #if 0 |
fdc5d982 | 410 | ZTI_INIT(ztest_dmu_prealloc, 1, &zopt_sometimes), |
428870ff | 411 | #endif |
fdc5d982 TC |
412 | ZTI_INIT(ztest_fzap, 1, &zopt_sometimes), |
413 | ZTI_INIT(ztest_dmu_snapshot_create_destroy, 1, &zopt_sometimes), | |
414 | ZTI_INIT(ztest_spa_create_destroy, 1, &zopt_sometimes), | |
415 | ZTI_INIT(ztest_fault_inject, 1, &zopt_sometimes), | |
fdc5d982 | 416 | ZTI_INIT(ztest_dmu_snapshot_hold, 1, &zopt_sometimes), |
0582e403 | 417 | ZTI_INIT(ztest_mmp_enable_disable, 1, &zopt_sometimes), |
fdc5d982 | 418 | ZTI_INIT(ztest_reguid, 1, &zopt_rarely), |
fdc5d982 TC |
419 | ZTI_INIT(ztest_scrub, 1, &zopt_rarely), |
420 | ZTI_INIT(ztest_spa_upgrade, 1, &zopt_rarely), | |
421 | ZTI_INIT(ztest_dsl_dataset_promote_busy, 1, &zopt_rarely), | |
422 | ZTI_INIT(ztest_vdev_attach_detach, 1, &zopt_sometimes), | |
423 | ZTI_INIT(ztest_vdev_LUN_growth, 1, &zopt_rarely), | |
424 | ZTI_INIT(ztest_vdev_add_remove, 1, &ztest_opts.zo_vdevtime), | |
cc99f275 | 425 | ZTI_INIT(ztest_vdev_class_add, 1, &ztest_opts.zo_vdevtime), |
fdc5d982 | 426 | ZTI_INIT(ztest_vdev_aux_add_remove, 1, &ztest_opts.zo_vdevtime), |
a1d477c2 | 427 | ZTI_INIT(ztest_device_removal, 1, &zopt_sometimes), |
d2734cce | 428 | ZTI_INIT(ztest_spa_checkpoint_create_discard, 1, &zopt_rarely), |
619f0976 | 429 | ZTI_INIT(ztest_initialize, 1, &zopt_sometimes), |
1b939560 | 430 | ZTI_INIT(ztest_trim, 1, &zopt_sometimes), |
1eeb4562 | 431 | ZTI_INIT(ztest_fletcher, 1, &zopt_rarely), |
37f520db | 432 | ZTI_INIT(ztest_fletcher_incr, 1, &zopt_rarely), |
50c957f7 | 433 | ZTI_INIT(ztest_verify_dnode_bt, 1, &zopt_sometimes), |
34dc7c2f BB |
434 | }; |
435 | ||
436 | #define ZTEST_FUNCS (sizeof (ztest_info) / sizeof (ztest_info_t)) | |
437 | ||
428870ff BB |
438 | /* |
439 | * The following struct is used to hold a list of uncalled commit callbacks. | |
440 | * The callbacks are ordered by txg number. | |
441 | */ | |
442 | typedef struct ztest_cb_list { | |
1e33ac1e BB |
443 | kmutex_t zcl_callbacks_lock; |
444 | list_t zcl_callbacks; | |
428870ff | 445 | } ztest_cb_list_t; |
34dc7c2f BB |
446 | |
447 | /* | |
448 | * Stuff we need to share writably between parent and child. | |
449 | */ | |
450 | typedef struct ztest_shared { | |
c242c188 | 451 | boolean_t zs_do_init; |
428870ff BB |
452 | hrtime_t zs_proc_start; |
453 | hrtime_t zs_proc_stop; | |
454 | hrtime_t zs_thread_start; | |
455 | hrtime_t zs_thread_stop; | |
456 | hrtime_t zs_thread_kill; | |
34dc7c2f | 457 | uint64_t zs_enospc_count; |
428870ff BB |
458 | uint64_t zs_vdev_next_leaf; |
459 | uint64_t zs_vdev_aux; | |
34dc7c2f BB |
460 | uint64_t zs_alloc; |
461 | uint64_t zs_space; | |
428870ff BB |
462 | uint64_t zs_splits; |
463 | uint64_t zs_mirrors; | |
c242c188 CS |
464 | uint64_t zs_metaslab_sz; |
465 | uint64_t zs_metaslab_df_alloc_threshold; | |
466 | uint64_t zs_guid; | |
34dc7c2f BB |
467 | } ztest_shared_t; |
468 | ||
428870ff BB |
469 | #define ID_PARALLEL -1ULL |
470 | ||
34dc7c2f | 471 | static char ztest_dev_template[] = "%s/%s.%llua"; |
b128c09f | 472 | static char ztest_aux_template[] = "%s/%s.%s.%llu"; |
428870ff | 473 | ztest_shared_t *ztest_shared; |
34dc7c2f | 474 | |
c242c188 CS |
475 | static spa_t *ztest_spa = NULL; |
476 | static ztest_ds_t *ztest_ds; | |
477 | ||
478 | static kmutex_t ztest_vdev_lock; | |
9e052db4 | 479 | static boolean_t ztest_device_removal_active = B_FALSE; |
ce5fb2a7 | 480 | static boolean_t ztest_pool_scrubbed = B_FALSE; |
d2734cce | 481 | static kmutex_t ztest_checkpoint_lock; |
3bc7e0fb GW |
482 | |
483 | /* | |
484 | * The ztest_name_lock protects the pool and dataset namespace used by | |
485 | * the individual tests. To modify the namespace, consumers must grab | |
486 | * this lock as writer. Grabbing the lock as reader will ensure that the | |
487 | * namespace does not change while the lock is held. | |
488 | */ | |
2d9142c9 | 489 | static pthread_rwlock_t ztest_name_lock; |
34dc7c2f | 490 | |
c242c188 | 491 | static boolean_t ztest_dump_core = B_TRUE; |
b128c09f | 492 | static boolean_t ztest_exiting; |
34dc7c2f | 493 | |
428870ff BB |
494 | /* Global commit callback list */ |
495 | static ztest_cb_list_t zcl; | |
090ff092 RC |
496 | /* Commit cb delay */ |
497 | static uint64_t zc_min_txg_delay = UINT64_MAX; | |
498 | static int zc_cb_counter = 0; | |
499 | ||
500 | /* | |
501 | * Minimum number of commit callbacks that need to be registered for us to check | |
502 | * whether the minimum txg delay is acceptable. | |
503 | */ | |
504 | #define ZTEST_COMMIT_CB_MIN_REG 100 | |
505 | ||
506 | /* | |
507 | * If a number of txgs equal to this threshold have been created after a commit | |
508 | * callback has been registered but not called, then we assume there is an | |
509 | * implementation bug. | |
510 | */ | |
511 | #define ZTEST_COMMIT_CB_THRESH (TXG_CONCURRENT_STATES + 1000) | |
428870ff | 512 | |
428870ff BB |
513 | enum ztest_object { |
514 | ZTEST_META_DNODE = 0, | |
515 | ZTEST_DIROBJ, | |
516 | ZTEST_OBJECTS | |
517 | }; | |
34dc7c2f BB |
518 | |
519 | static void usage(boolean_t) __NORETURN; | |
ce5fb2a7 | 520 | static int ztest_scrub_impl(spa_t *spa); |
34dc7c2f BB |
521 | |
522 | /* | |
523 | * These libumem hooks provide a reasonable set of defaults for the allocator's | |
524 | * debugging facilities. | |
525 | */ | |
526 | const char * | |
0bc8fd78 | 527 | _umem_debug_init(void) |
34dc7c2f BB |
528 | { |
529 | return ("default,verbose"); /* $UMEM_DEBUG setting */ | |
530 | } | |
531 | ||
532 | const char * | |
533 | _umem_logging_init(void) | |
534 | { | |
535 | return ("fail,contents"); /* $UMEM_LOGGING setting */ | |
536 | } | |
537 | ||
8fb1ede1 BB |
538 | static void |
539 | dump_debug_buffer(void) | |
540 | { | |
ab4c009e TC |
541 | ssize_t ret __attribute__((unused)); |
542 | ||
543 | if (!ztest_opts.zo_dump_dbgmsg) | |
8fb1ede1 BB |
544 | return; |
545 | ||
ab4c009e TC |
546 | /* |
547 | * We use write() instead of printf() so that this function | |
548 | * is safe to call from a signal handler. | |
549 | */ | |
550 | ret = write(STDOUT_FILENO, "\n", 1); | |
8fb1ede1 BB |
551 | zfs_dbgmsg_print("ztest"); |
552 | } | |
553 | ||
e82cdc3a NB |
554 | #define BACKTRACE_SZ 100 |
555 | ||
556 | static void sig_handler(int signo) | |
557 | { | |
558 | struct sigaction action; | |
967798d0 | 559 | #ifdef __GLIBC__ /* backtrace() is a GNU extension */ |
e82cdc3a NB |
560 | int nptrs; |
561 | void *buffer[BACKTRACE_SZ]; | |
562 | ||
563 | nptrs = backtrace(buffer, BACKTRACE_SZ); | |
564 | backtrace_symbols_fd(buffer, nptrs, STDERR_FILENO); | |
565 | #endif | |
8fb1ede1 | 566 | dump_debug_buffer(); |
e82cdc3a NB |
567 | |
568 | /* | |
569 | * Restore default action and re-raise signal so SIGSEGV and | |
570 | * SIGABRT can trigger a core dump. | |
571 | */ | |
572 | action.sa_handler = SIG_DFL; | |
573 | sigemptyset(&action.sa_mask); | |
574 | action.sa_flags = 0; | |
575 | (void) sigaction(signo, &action, NULL); | |
576 | raise(signo); | |
577 | } | |
578 | ||
34dc7c2f BB |
579 | #define FATAL_MSG_SZ 1024 |
580 | ||
581 | char *fatal_msg; | |
582 | ||
583 | static void | |
584 | fatal(int do_perror, char *message, ...) | |
585 | { | |
586 | va_list args; | |
587 | int save_errno = errno; | |
40b84e7a | 588 | char *buf; |
34dc7c2f BB |
589 | |
590 | (void) fflush(stdout); | |
40b84e7a | 591 | buf = umem_alloc(FATAL_MSG_SZ, UMEM_NOFAIL); |
34dc7c2f BB |
592 | |
593 | va_start(args, message); | |
594 | (void) sprintf(buf, "ztest: "); | |
595 | /* LINTED */ | |
596 | (void) vsprintf(buf + strlen(buf), message, args); | |
597 | va_end(args); | |
598 | if (do_perror) { | |
599 | (void) snprintf(buf + strlen(buf), FATAL_MSG_SZ - strlen(buf), | |
600 | ": %s", strerror(save_errno)); | |
601 | } | |
602 | (void) fprintf(stderr, "%s\n", buf); | |
603 | fatal_msg = buf; /* to ease debugging */ | |
8fb1ede1 | 604 | |
34dc7c2f BB |
605 | if (ztest_dump_core) |
606 | abort(); | |
ab4c009e TC |
607 | else |
608 | dump_debug_buffer(); | |
609 | ||
34dc7c2f BB |
610 | exit(3); |
611 | } | |
612 | ||
613 | static int | |
614 | str2shift(const char *buf) | |
615 | { | |
616 | const char *ends = "BKMGTPEZ"; | |
617 | int i; | |
618 | ||
619 | if (buf[0] == '\0') | |
620 | return (0); | |
621 | for (i = 0; i < strlen(ends); i++) { | |
622 | if (toupper(buf[0]) == ends[i]) | |
623 | break; | |
624 | } | |
625 | if (i == strlen(ends)) { | |
626 | (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", | |
627 | buf); | |
628 | usage(B_FALSE); | |
629 | } | |
630 | if (buf[1] == '\0' || (toupper(buf[1]) == 'B' && buf[2] == '\0')) { | |
631 | return (10*i); | |
632 | } | |
633 | (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", buf); | |
634 | usage(B_FALSE); | |
635 | /* NOTREACHED */ | |
636 | } | |
637 | ||
638 | static uint64_t | |
639 | nicenumtoull(const char *buf) | |
640 | { | |
641 | char *end; | |
642 | uint64_t val; | |
643 | ||
644 | val = strtoull(buf, &end, 0); | |
645 | if (end == buf) { | |
646 | (void) fprintf(stderr, "ztest: bad numeric value: %s\n", buf); | |
647 | usage(B_FALSE); | |
648 | } else if (end[0] == '.') { | |
649 | double fval = strtod(buf, &end); | |
650 | fval *= pow(2, str2shift(end)); | |
4d32abaa RM |
651 | /* |
652 | * UINT64_MAX is not exactly representable as a double. | |
653 | * The closest representation is UINT64_MAX + 1, so we | |
654 | * use a >= comparison instead of > for the bounds check. | |
655 | */ | |
656 | if (fval >= (double)UINT64_MAX) { | |
34dc7c2f BB |
657 | (void) fprintf(stderr, "ztest: value too large: %s\n", |
658 | buf); | |
659 | usage(B_FALSE); | |
660 | } | |
661 | val = (uint64_t)fval; | |
662 | } else { | |
663 | int shift = str2shift(end); | |
664 | if (shift >= 64 || (val << shift) >> shift != val) { | |
665 | (void) fprintf(stderr, "ztest: value too large: %s\n", | |
666 | buf); | |
667 | usage(B_FALSE); | |
668 | } | |
669 | val <<= shift; | |
670 | } | |
671 | return (val); | |
672 | } | |
673 | ||
674 | static void | |
675 | usage(boolean_t requested) | |
676 | { | |
c242c188 CS |
677 | const ztest_shared_opts_t *zo = &ztest_opts_defaults; |
678 | ||
f3c8c9e6 | 679 | char nice_vdev_size[NN_NUMBUF_SZ]; |
d830d479 | 680 | char nice_force_ganging[NN_NUMBUF_SZ]; |
34dc7c2f BB |
681 | FILE *fp = requested ? stdout : stderr; |
682 | ||
f3c8c9e6 | 683 | nicenum(zo->zo_vdev_size, nice_vdev_size, sizeof (nice_vdev_size)); |
d830d479 MA |
684 | nicenum(zo->zo_metaslab_force_ganging, nice_force_ganging, |
685 | sizeof (nice_force_ganging)); | |
34dc7c2f BB |
686 | |
687 | (void) fprintf(fp, "Usage: %s\n" | |
688 | "\t[-v vdevs (default: %llu)]\n" | |
689 | "\t[-s size_of_each_vdev (default: %s)]\n" | |
428870ff | 690 | "\t[-a alignment_shift (default: %d)] use 0 for random\n" |
34dc7c2f BB |
691 | "\t[-m mirror_copies (default: %d)]\n" |
692 | "\t[-r raidz_disks (default: %d)]\n" | |
693 | "\t[-R raidz_parity (default: %d)]\n" | |
694 | "\t[-d datasets (default: %d)]\n" | |
695 | "\t[-t threads (default: %d)]\n" | |
696 | "\t[-g gang_block_threshold (default: %s)]\n" | |
428870ff BB |
697 | "\t[-i init_count (default: %d)] initialize pool i times\n" |
698 | "\t[-k kill_percentage (default: %llu%%)]\n" | |
34dc7c2f | 699 | "\t[-p pool_name (default: %s)]\n" |
428870ff | 700 | "\t[-f dir (default: %s)] file directory for vdev files\n" |
379ca9cf | 701 | "\t[-M] Multi-host simulate pool imported on remote host\n" |
428870ff BB |
702 | "\t[-V] verbose (use multiple times for ever more blather)\n" |
703 | "\t[-E] use existing pool instead of creating new one\n" | |
704 | "\t[-T time (default: %llu sec)] total run time\n" | |
705 | "\t[-F freezeloops (default: %llu)] max loops in spa_freeze()\n" | |
706 | "\t[-P passtime (default: %llu sec)] time per pass\n" | |
c242c188 | 707 | "\t[-B alt_ztest (default: <none>)] alternate ztest path\n" |
cc99f275 | 708 | "\t[-C vdev class state (default: random)] special=on|off|random\n" |
ed828c0c GM |
709 | "\t[-o variable=value] ... set global variable to an unsigned\n" |
710 | "\t 32-bit integer value\n" | |
8fb1ede1 | 711 | "\t[-G dump zfs_dbgmsg buffer before exiting due to an error\n" |
34dc7c2f BB |
712 | "\t[-h] (print help)\n" |
713 | "", | |
c242c188 CS |
714 | zo->zo_pool, |
715 | (u_longlong_t)zo->zo_vdevs, /* -v */ | |
34dc7c2f | 716 | nice_vdev_size, /* -s */ |
c242c188 CS |
717 | zo->zo_ashift, /* -a */ |
718 | zo->zo_mirrors, /* -m */ | |
719 | zo->zo_raidz, /* -r */ | |
720 | zo->zo_raidz_parity, /* -R */ | |
721 | zo->zo_datasets, /* -d */ | |
722 | zo->zo_threads, /* -t */ | |
d830d479 | 723 | nice_force_ganging, /* -g */ |
c242c188 CS |
724 | zo->zo_init, /* -i */ |
725 | (u_longlong_t)zo->zo_killrate, /* -k */ | |
726 | zo->zo_pool, /* -p */ | |
727 | zo->zo_dir, /* -f */ | |
728 | (u_longlong_t)zo->zo_time, /* -T */ | |
729 | (u_longlong_t)zo->zo_maxloops, /* -F */ | |
730 | (u_longlong_t)zo->zo_passtime); | |
34dc7c2f BB |
731 | exit(requested ? 0 : 1); |
732 | } | |
733 | ||
cc99f275 DB |
734 | |
735 | static void | |
736 | ztest_parse_name_value(const char *input, ztest_shared_opts_t *zo) | |
737 | { | |
738 | char name[32]; | |
739 | char *value; | |
740 | int state = ZTEST_VDEV_CLASS_RND; | |
741 | ||
742 | (void) strlcpy(name, input, sizeof (name)); | |
743 | ||
744 | value = strchr(name, '='); | |
745 | if (value == NULL) { | |
746 | (void) fprintf(stderr, "missing value in property=value " | |
747 | "'-C' argument (%s)\n", input); | |
748 | usage(B_FALSE); | |
749 | } | |
750 | *(value) = '\0'; | |
751 | value++; | |
752 | ||
753 | if (strcmp(value, "on") == 0) { | |
754 | state = ZTEST_VDEV_CLASS_ON; | |
755 | } else if (strcmp(value, "off") == 0) { | |
756 | state = ZTEST_VDEV_CLASS_OFF; | |
757 | } else if (strcmp(value, "random") == 0) { | |
758 | state = ZTEST_VDEV_CLASS_RND; | |
759 | } else { | |
760 | (void) fprintf(stderr, "invalid property value '%s'\n", value); | |
761 | usage(B_FALSE); | |
762 | } | |
763 | ||
764 | if (strcmp(name, "special") == 0) { | |
765 | zo->zo_special_vdevs = state; | |
766 | } else { | |
767 | (void) fprintf(stderr, "invalid property name '%s'\n", name); | |
768 | usage(B_FALSE); | |
769 | } | |
770 | if (zo->zo_verbose >= 3) | |
771 | (void) printf("%s vdev state is '%s'\n", name, value); | |
772 | } | |
773 | ||
34dc7c2f BB |
774 | static void |
775 | process_options(int argc, char **argv) | |
776 | { | |
c242c188 CS |
777 | char *path; |
778 | ztest_shared_opts_t *zo = &ztest_opts; | |
779 | ||
34dc7c2f BB |
780 | int opt; |
781 | uint64_t value; | |
c242c188 | 782 | char altdir[MAXNAMELEN] = { 0 }; |
34dc7c2f | 783 | |
c242c188 | 784 | bcopy(&ztest_opts_defaults, zo, sizeof (*zo)); |
34dc7c2f | 785 | |
34dc7c2f | 786 | while ((opt = getopt(argc, argv, |
cc99f275 | 787 | "v:s:a:m:r:R:d:t:g:i:k:p:f:MVET:P:hF:B:C:o:G")) != EOF) { |
34dc7c2f BB |
788 | value = 0; |
789 | switch (opt) { | |
790 | case 'v': | |
791 | case 's': | |
792 | case 'a': | |
793 | case 'm': | |
794 | case 'r': | |
795 | case 'R': | |
796 | case 'd': | |
797 | case 't': | |
798 | case 'g': | |
799 | case 'i': | |
800 | case 'k': | |
801 | case 'T': | |
802 | case 'P': | |
428870ff | 803 | case 'F': |
34dc7c2f BB |
804 | value = nicenumtoull(optarg); |
805 | } | |
806 | switch (opt) { | |
807 | case 'v': | |
c242c188 | 808 | zo->zo_vdevs = value; |
34dc7c2f BB |
809 | break; |
810 | case 's': | |
c242c188 | 811 | zo->zo_vdev_size = MAX(SPA_MINDEVSIZE, value); |
34dc7c2f BB |
812 | break; |
813 | case 'a': | |
c242c188 | 814 | zo->zo_ashift = value; |
34dc7c2f BB |
815 | break; |
816 | case 'm': | |
c242c188 | 817 | zo->zo_mirrors = value; |
34dc7c2f BB |
818 | break; |
819 | case 'r': | |
c242c188 | 820 | zo->zo_raidz = MAX(1, value); |
34dc7c2f BB |
821 | break; |
822 | case 'R': | |
c242c188 | 823 | zo->zo_raidz_parity = MIN(MAX(value, 1), 3); |
34dc7c2f BB |
824 | break; |
825 | case 'd': | |
c242c188 | 826 | zo->zo_datasets = MAX(1, value); |
34dc7c2f BB |
827 | break; |
828 | case 't': | |
c242c188 | 829 | zo->zo_threads = MAX(1, value); |
34dc7c2f BB |
830 | break; |
831 | case 'g': | |
d830d479 MA |
832 | zo->zo_metaslab_force_ganging = |
833 | MAX(SPA_MINBLOCKSIZE << 1, value); | |
34dc7c2f BB |
834 | break; |
835 | case 'i': | |
c242c188 | 836 | zo->zo_init = value; |
34dc7c2f BB |
837 | break; |
838 | case 'k': | |
c242c188 | 839 | zo->zo_killrate = value; |
34dc7c2f BB |
840 | break; |
841 | case 'p': | |
c242c188 CS |
842 | (void) strlcpy(zo->zo_pool, optarg, |
843 | sizeof (zo->zo_pool)); | |
34dc7c2f BB |
844 | break; |
845 | case 'f': | |
c242c188 CS |
846 | path = realpath(optarg, NULL); |
847 | if (path == NULL) { | |
848 | (void) fprintf(stderr, "error: %s: %s\n", | |
849 | optarg, strerror(errno)); | |
850 | usage(B_FALSE); | |
851 | } else { | |
852 | (void) strlcpy(zo->zo_dir, path, | |
853 | sizeof (zo->zo_dir)); | |
df053d67 | 854 | free(path); |
c242c188 | 855 | } |
34dc7c2f | 856 | break; |
379ca9cf OF |
857 | case 'M': |
858 | zo->zo_mmp_test = 1; | |
859 | break; | |
34dc7c2f | 860 | case 'V': |
c242c188 | 861 | zo->zo_verbose++; |
34dc7c2f BB |
862 | break; |
863 | case 'E': | |
c242c188 | 864 | zo->zo_init = 0; |
34dc7c2f BB |
865 | break; |
866 | case 'T': | |
c242c188 | 867 | zo->zo_time = value; |
34dc7c2f BB |
868 | break; |
869 | case 'P': | |
c242c188 | 870 | zo->zo_passtime = MAX(1, value); |
34dc7c2f | 871 | break; |
428870ff | 872 | case 'F': |
c242c188 CS |
873 | zo->zo_maxloops = MAX(1, value); |
874 | break; | |
875 | case 'B': | |
876 | (void) strlcpy(altdir, optarg, sizeof (altdir)); | |
428870ff | 877 | break; |
cc99f275 DB |
878 | case 'C': |
879 | ztest_parse_name_value(optarg, zo); | |
880 | break; | |
ed828c0c GM |
881 | case 'o': |
882 | if (set_global_var(optarg) != 0) | |
883 | usage(B_FALSE); | |
884 | break; | |
8fb1ede1 | 885 | case 'G': |
ab4c009e | 886 | zo->zo_dump_dbgmsg = 1; |
8fb1ede1 | 887 | break; |
34dc7c2f BB |
888 | case 'h': |
889 | usage(B_TRUE); | |
890 | break; | |
891 | case '?': | |
892 | default: | |
893 | usage(B_FALSE); | |
894 | break; | |
895 | } | |
896 | } | |
897 | ||
c242c188 | 898 | zo->zo_raidz_parity = MIN(zo->zo_raidz_parity, zo->zo_raidz - 1); |
34dc7c2f | 899 | |
c242c188 CS |
900 | zo->zo_vdevtime = |
901 | (zo->zo_vdevs > 0 ? zo->zo_time * NANOSEC / zo->zo_vdevs : | |
428870ff | 902 | UINT64_MAX >> 2); |
c242c188 CS |
903 | |
904 | if (strlen(altdir) > 0) { | |
ae380cfa BB |
905 | char *cmd; |
906 | char *realaltdir; | |
c242c188 CS |
907 | char *bin; |
908 | char *ztest; | |
909 | char *isa; | |
910 | int isalen; | |
911 | ||
ae380cfa BB |
912 | cmd = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
913 | realaltdir = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
914 | ||
5be98cfe | 915 | VERIFY(NULL != realpath(getexecname(), cmd)); |
c242c188 CS |
916 | if (0 != access(altdir, F_OK)) { |
917 | ztest_dump_core = B_FALSE; | |
918 | fatal(B_TRUE, "invalid alternate ztest path: %s", | |
919 | altdir); | |
920 | } | |
921 | VERIFY(NULL != realpath(altdir, realaltdir)); | |
922 | ||
923 | /* | |
924 | * 'cmd' should be of the form "<anything>/usr/bin/<isa>/ztest". | |
925 | * We want to extract <isa> to determine if we should use | |
926 | * 32 or 64 bit binaries. | |
927 | */ | |
928 | bin = strstr(cmd, "/usr/bin/"); | |
929 | ztest = strstr(bin, "/ztest"); | |
930 | isa = bin + 9; | |
931 | isalen = ztest - isa; | |
932 | (void) snprintf(zo->zo_alt_ztest, sizeof (zo->zo_alt_ztest), | |
933 | "%s/usr/bin/%.*s/ztest", realaltdir, isalen, isa); | |
934 | (void) snprintf(zo->zo_alt_libpath, sizeof (zo->zo_alt_libpath), | |
935 | "%s/usr/lib/%.*s", realaltdir, isalen, isa); | |
936 | ||
937 | if (0 != access(zo->zo_alt_ztest, X_OK)) { | |
938 | ztest_dump_core = B_FALSE; | |
939 | fatal(B_TRUE, "invalid alternate ztest: %s", | |
940 | zo->zo_alt_ztest); | |
941 | } else if (0 != access(zo->zo_alt_libpath, X_OK)) { | |
942 | ztest_dump_core = B_FALSE; | |
943 | fatal(B_TRUE, "invalid alternate lib directory %s", | |
944 | zo->zo_alt_libpath); | |
945 | } | |
ae380cfa BB |
946 | |
947 | umem_free(cmd, MAXPATHLEN); | |
948 | umem_free(realaltdir, MAXPATHLEN); | |
c242c188 | 949 | } |
428870ff BB |
950 | } |
951 | ||
952 | static void | |
953 | ztest_kill(ztest_shared_t *zs) | |
954 | { | |
c242c188 CS |
955 | zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(ztest_spa)); |
956 | zs->zs_space = metaslab_class_get_space(spa_normal_class(ztest_spa)); | |
5d1f7fb6 GW |
957 | |
958 | /* | |
959 | * Before we kill off ztest, make sure that the config is updated. | |
a1d477c2 | 960 | * See comment above spa_write_cachefile(). |
5d1f7fb6 GW |
961 | */ |
962 | mutex_enter(&spa_namespace_lock); | |
a1d477c2 | 963 | spa_write_cachefile(ztest_spa, B_FALSE, B_FALSE); |
5d1f7fb6 GW |
964 | mutex_exit(&spa_namespace_lock); |
965 | ||
428870ff BB |
966 | (void) kill(getpid(), SIGKILL); |
967 | } | |
968 | ||
969 | static uint64_t | |
970 | ztest_random(uint64_t range) | |
971 | { | |
972 | uint64_t r; | |
973 | ||
9d81146b ED |
974 | ASSERT3S(ztest_fd_rand, >=, 0); |
975 | ||
428870ff BB |
976 | if (range == 0) |
977 | return (0); | |
978 | ||
9d81146b | 979 | if (read(ztest_fd_rand, &r, sizeof (r)) != sizeof (r)) |
428870ff BB |
980 | fatal(1, "short read from /dev/urandom"); |
981 | ||
982 | return (r % range); | |
983 | } | |
984 | ||
985 | /* ARGSUSED */ | |
986 | static void | |
987 | ztest_record_enospc(const char *s) | |
988 | { | |
989 | ztest_shared->zs_enospc_count++; | |
34dc7c2f BB |
990 | } |
991 | ||
992 | static uint64_t | |
993 | ztest_get_ashift(void) | |
994 | { | |
c242c188 | 995 | if (ztest_opts.zo_ashift == 0) |
b02fe35d | 996 | return (SPA_MINBLOCKSHIFT + ztest_random(5)); |
c242c188 | 997 | return (ztest_opts.zo_ashift); |
34dc7c2f BB |
998 | } |
999 | ||
1000 | static nvlist_t * | |
ea0b2538 | 1001 | make_vdev_file(char *path, char *aux, char *pool, size_t size, uint64_t ashift) |
34dc7c2f | 1002 | { |
40b84e7a | 1003 | char *pathbuf; |
34dc7c2f | 1004 | uint64_t vdev; |
34dc7c2f BB |
1005 | nvlist_t *file; |
1006 | ||
40b84e7a BB |
1007 | pathbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
1008 | ||
b128c09f BB |
1009 | if (ashift == 0) |
1010 | ashift = ztest_get_ashift(); | |
1011 | ||
1012 | if (path == NULL) { | |
1013 | path = pathbuf; | |
1014 | ||
1015 | if (aux != NULL) { | |
1016 | vdev = ztest_shared->zs_vdev_aux; | |
c242c188 CS |
1017 | (void) snprintf(path, MAXPATHLEN, |
1018 | ztest_aux_template, ztest_opts.zo_dir, | |
ea0b2538 GW |
1019 | pool == NULL ? ztest_opts.zo_pool : pool, |
1020 | aux, vdev); | |
b128c09f | 1021 | } else { |
428870ff | 1022 | vdev = ztest_shared->zs_vdev_next_leaf++; |
c242c188 CS |
1023 | (void) snprintf(path, MAXPATHLEN, |
1024 | ztest_dev_template, ztest_opts.zo_dir, | |
ea0b2538 | 1025 | pool == NULL ? ztest_opts.zo_pool : pool, vdev); |
b128c09f BB |
1026 | } |
1027 | } | |
34dc7c2f | 1028 | |
b128c09f BB |
1029 | if (size != 0) { |
1030 | int fd = open(path, O_RDWR | O_CREAT | O_TRUNC, 0666); | |
34dc7c2f | 1031 | if (fd == -1) |
b128c09f | 1032 | fatal(1, "can't open %s", path); |
34dc7c2f | 1033 | if (ftruncate(fd, size) != 0) |
b128c09f | 1034 | fatal(1, "can't ftruncate %s", path); |
34dc7c2f BB |
1035 | (void) close(fd); |
1036 | } | |
1037 | ||
1038 | VERIFY(nvlist_alloc(&file, NV_UNIQUE_NAME, 0) == 0); | |
1039 | VERIFY(nvlist_add_string(file, ZPOOL_CONFIG_TYPE, VDEV_TYPE_FILE) == 0); | |
b128c09f | 1040 | VERIFY(nvlist_add_string(file, ZPOOL_CONFIG_PATH, path) == 0); |
34dc7c2f | 1041 | VERIFY(nvlist_add_uint64(file, ZPOOL_CONFIG_ASHIFT, ashift) == 0); |
40b84e7a | 1042 | umem_free(pathbuf, MAXPATHLEN); |
34dc7c2f BB |
1043 | |
1044 | return (file); | |
1045 | } | |
1046 | ||
1047 | static nvlist_t * | |
ea0b2538 GW |
1048 | make_vdev_raidz(char *path, char *aux, char *pool, size_t size, |
1049 | uint64_t ashift, int r) | |
34dc7c2f BB |
1050 | { |
1051 | nvlist_t *raidz, **child; | |
1052 | int c; | |
1053 | ||
1054 | if (r < 2) | |
ea0b2538 | 1055 | return (make_vdev_file(path, aux, pool, size, ashift)); |
34dc7c2f BB |
1056 | child = umem_alloc(r * sizeof (nvlist_t *), UMEM_NOFAIL); |
1057 | ||
1058 | for (c = 0; c < r; c++) | |
ea0b2538 | 1059 | child[c] = make_vdev_file(path, aux, pool, size, ashift); |
34dc7c2f BB |
1060 | |
1061 | VERIFY(nvlist_alloc(&raidz, NV_UNIQUE_NAME, 0) == 0); | |
1062 | VERIFY(nvlist_add_string(raidz, ZPOOL_CONFIG_TYPE, | |
1063 | VDEV_TYPE_RAIDZ) == 0); | |
1064 | VERIFY(nvlist_add_uint64(raidz, ZPOOL_CONFIG_NPARITY, | |
c242c188 | 1065 | ztest_opts.zo_raidz_parity) == 0); |
34dc7c2f BB |
1066 | VERIFY(nvlist_add_nvlist_array(raidz, ZPOOL_CONFIG_CHILDREN, |
1067 | child, r) == 0); | |
1068 | ||
1069 | for (c = 0; c < r; c++) | |
1070 | nvlist_free(child[c]); | |
1071 | ||
1072 | umem_free(child, r * sizeof (nvlist_t *)); | |
1073 | ||
1074 | return (raidz); | |
1075 | } | |
1076 | ||
1077 | static nvlist_t * | |
ea0b2538 GW |
1078 | make_vdev_mirror(char *path, char *aux, char *pool, size_t size, |
1079 | uint64_t ashift, int r, int m) | |
34dc7c2f BB |
1080 | { |
1081 | nvlist_t *mirror, **child; | |
1082 | int c; | |
1083 | ||
1084 | if (m < 1) | |
ea0b2538 | 1085 | return (make_vdev_raidz(path, aux, pool, size, ashift, r)); |
34dc7c2f BB |
1086 | |
1087 | child = umem_alloc(m * sizeof (nvlist_t *), UMEM_NOFAIL); | |
1088 | ||
1089 | for (c = 0; c < m; c++) | |
ea0b2538 | 1090 | child[c] = make_vdev_raidz(path, aux, pool, size, ashift, r); |
34dc7c2f BB |
1091 | |
1092 | VERIFY(nvlist_alloc(&mirror, NV_UNIQUE_NAME, 0) == 0); | |
1093 | VERIFY(nvlist_add_string(mirror, ZPOOL_CONFIG_TYPE, | |
1094 | VDEV_TYPE_MIRROR) == 0); | |
1095 | VERIFY(nvlist_add_nvlist_array(mirror, ZPOOL_CONFIG_CHILDREN, | |
1096 | child, m) == 0); | |
34dc7c2f BB |
1097 | |
1098 | for (c = 0; c < m; c++) | |
1099 | nvlist_free(child[c]); | |
1100 | ||
1101 | umem_free(child, m * sizeof (nvlist_t *)); | |
1102 | ||
1103 | return (mirror); | |
1104 | } | |
1105 | ||
1106 | static nvlist_t * | |
ea0b2538 | 1107 | make_vdev_root(char *path, char *aux, char *pool, size_t size, uint64_t ashift, |
cc99f275 | 1108 | const char *class, int r, int m, int t) |
34dc7c2f BB |
1109 | { |
1110 | nvlist_t *root, **child; | |
1111 | int c; | |
cc99f275 | 1112 | boolean_t log; |
34dc7c2f BB |
1113 | |
1114 | ASSERT(t > 0); | |
1115 | ||
cc99f275 DB |
1116 | log = (class != NULL && strcmp(class, "log") == 0); |
1117 | ||
34dc7c2f BB |
1118 | child = umem_alloc(t * sizeof (nvlist_t *), UMEM_NOFAIL); |
1119 | ||
b128c09f | 1120 | for (c = 0; c < t; c++) { |
ea0b2538 GW |
1121 | child[c] = make_vdev_mirror(path, aux, pool, size, ashift, |
1122 | r, m); | |
b128c09f BB |
1123 | VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_IS_LOG, |
1124 | log) == 0); | |
cc99f275 DB |
1125 | |
1126 | if (class != NULL && class[0] != '\0') { | |
1127 | ASSERT(m > 1 || log); /* expecting a mirror */ | |
1128 | VERIFY(nvlist_add_string(child[c], | |
1129 | ZPOOL_CONFIG_ALLOCATION_BIAS, class) == 0); | |
1130 | } | |
b128c09f | 1131 | } |
34dc7c2f BB |
1132 | |
1133 | VERIFY(nvlist_alloc(&root, NV_UNIQUE_NAME, 0) == 0); | |
1134 | VERIFY(nvlist_add_string(root, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT) == 0); | |
b128c09f | 1135 | VERIFY(nvlist_add_nvlist_array(root, aux ? aux : ZPOOL_CONFIG_CHILDREN, |
34dc7c2f BB |
1136 | child, t) == 0); |
1137 | ||
1138 | for (c = 0; c < t; c++) | |
1139 | nvlist_free(child[c]); | |
1140 | ||
1141 | umem_free(child, t * sizeof (nvlist_t *)); | |
1142 | ||
1143 | return (root); | |
1144 | } | |
1145 | ||
ea0b2538 GW |
1146 | /* |
1147 | * Find a random spa version. Returns back a random spa version in the | |
1148 | * range [initial_version, SPA_VERSION_FEATURES]. | |
1149 | */ | |
1150 | static uint64_t | |
1151 | ztest_random_spa_version(uint64_t initial_version) | |
1152 | { | |
1153 | uint64_t version = initial_version; | |
1154 | ||
1155 | if (version <= SPA_VERSION_BEFORE_FEATURES) { | |
1156 | version = version + | |
1157 | ztest_random(SPA_VERSION_BEFORE_FEATURES - version + 1); | |
1158 | } | |
1159 | ||
1160 | if (version > SPA_VERSION_BEFORE_FEATURES) | |
1161 | version = SPA_VERSION_FEATURES; | |
1162 | ||
1163 | ASSERT(SPA_VERSION_IS_SUPPORTED(version)); | |
1164 | return (version); | |
1165 | } | |
1166 | ||
428870ff BB |
1167 | static int |
1168 | ztest_random_blocksize(void) | |
34dc7c2f | 1169 | { |
cc99f275 DB |
1170 | ASSERT(ztest_spa->spa_max_ashift != 0); |
1171 | ||
f1512ee6 MA |
1172 | /* |
1173 | * Choose a block size >= the ashift. | |
1174 | * If the SPA supports new MAXBLOCKSIZE, test up to 1MB blocks. | |
1175 | */ | |
1176 | int maxbs = SPA_OLD_MAXBLOCKSHIFT; | |
1177 | if (spa_maxblocksize(ztest_spa) == SPA_MAXBLOCKSIZE) | |
1178 | maxbs = 20; | |
c3520e7f MA |
1179 | uint64_t block_shift = |
1180 | ztest_random(maxbs - ztest_spa->spa_max_ashift + 1); | |
b02fe35d | 1181 | return (1 << (SPA_MINBLOCKSHIFT + block_shift)); |
428870ff | 1182 | } |
34dc7c2f | 1183 | |
50c957f7 NB |
1184 | static int |
1185 | ztest_random_dnodesize(void) | |
1186 | { | |
1187 | int slots; | |
1188 | int max_slots = spa_maxdnodesize(ztest_spa) >> DNODE_SHIFT; | |
1189 | ||
1190 | if (max_slots == DNODE_MIN_SLOTS) | |
1191 | return (DNODE_MIN_SIZE); | |
1192 | ||
1193 | /* | |
1194 | * Weight the random distribution more heavily toward smaller | |
1195 | * dnode sizes since that is more likely to reflect real-world | |
1196 | * usage. | |
1197 | */ | |
1198 | ASSERT3U(max_slots, >, 4); | |
1199 | switch (ztest_random(10)) { | |
1200 | case 0: | |
1201 | slots = 5 + ztest_random(max_slots - 4); | |
1202 | break; | |
1203 | case 1 ... 4: | |
1204 | slots = 2 + ztest_random(3); | |
1205 | break; | |
1206 | default: | |
1207 | slots = 1; | |
1208 | break; | |
1209 | } | |
1210 | ||
1211 | return (slots << DNODE_SHIFT); | |
1212 | } | |
1213 | ||
428870ff BB |
1214 | static int |
1215 | ztest_random_ibshift(void) | |
1216 | { | |
1217 | return (DN_MIN_INDBLKSHIFT + | |
1218 | ztest_random(DN_MAX_INDBLKSHIFT - DN_MIN_INDBLKSHIFT + 1)); | |
34dc7c2f BB |
1219 | } |
1220 | ||
428870ff BB |
1221 | static uint64_t |
1222 | ztest_random_vdev_top(spa_t *spa, boolean_t log_ok) | |
34dc7c2f | 1223 | { |
428870ff BB |
1224 | uint64_t top; |
1225 | vdev_t *rvd = spa->spa_root_vdev; | |
1226 | vdev_t *tvd; | |
34dc7c2f | 1227 | |
428870ff | 1228 | ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); |
34dc7c2f | 1229 | |
428870ff BB |
1230 | do { |
1231 | top = ztest_random(rvd->vdev_children); | |
1232 | tvd = rvd->vdev_child[top]; | |
a1d477c2 | 1233 | } while (!vdev_is_concrete(tvd) || (tvd->vdev_islog && !log_ok) || |
428870ff | 1234 | tvd->vdev_mg == NULL || tvd->vdev_mg->mg_class == NULL); |
34dc7c2f | 1235 | |
428870ff | 1236 | return (top); |
34dc7c2f BB |
1237 | } |
1238 | ||
428870ff BB |
1239 | static uint64_t |
1240 | ztest_random_dsl_prop(zfs_prop_t prop) | |
34dc7c2f | 1241 | { |
428870ff BB |
1242 | uint64_t value; |
1243 | ||
1244 | do { | |
1245 | value = zfs_prop_random_value(prop, ztest_random(-1ULL)); | |
1246 | } while (prop == ZFS_PROP_CHECKSUM && value == ZIO_CHECKSUM_OFF); | |
1247 | ||
1248 | return (value); | |
34dc7c2f BB |
1249 | } |
1250 | ||
34dc7c2f | 1251 | static int |
428870ff BB |
1252 | ztest_dsl_prop_set_uint64(char *osname, zfs_prop_t prop, uint64_t value, |
1253 | boolean_t inherit) | |
34dc7c2f | 1254 | { |
428870ff BB |
1255 | const char *propname = zfs_prop_to_name(prop); |
1256 | const char *valname; | |
40b84e7a | 1257 | char *setpoint; |
428870ff | 1258 | uint64_t curval; |
34dc7c2f BB |
1259 | int error; |
1260 | ||
13fe0198 MA |
1261 | error = dsl_prop_set_int(osname, propname, |
1262 | (inherit ? ZPROP_SRC_NONE : ZPROP_SRC_LOCAL), value); | |
34dc7c2f | 1263 | |
428870ff BB |
1264 | if (error == ENOSPC) { |
1265 | ztest_record_enospc(FTAG); | |
34dc7c2f BB |
1266 | return (error); |
1267 | } | |
c99c9001 | 1268 | ASSERT0(error); |
34dc7c2f | 1269 | |
40b84e7a | 1270 | setpoint = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
13fe0198 | 1271 | VERIFY0(dsl_prop_get_integer(osname, propname, &curval, setpoint)); |
428870ff | 1272 | |
c242c188 | 1273 | if (ztest_opts.zo_verbose >= 6) { |
6bec4351 TC |
1274 | int err; |
1275 | ||
1276 | err = zfs_prop_index_to_string(prop, curval, &valname); | |
1277 | if (err) | |
02730c33 BB |
1278 | (void) printf("%s %s = %llu at '%s'\n", osname, |
1279 | propname, (unsigned long long)curval, setpoint); | |
6bec4351 TC |
1280 | else |
1281 | (void) printf("%s %s = %s at '%s'\n", | |
1282 | osname, propname, valname, setpoint); | |
34dc7c2f | 1283 | } |
40b84e7a | 1284 | umem_free(setpoint, MAXPATHLEN); |
34dc7c2f BB |
1285 | |
1286 | return (error); | |
1287 | } | |
1288 | ||
1289 | static int | |
c242c188 | 1290 | ztest_spa_prop_set_uint64(zpool_prop_t prop, uint64_t value) |
34dc7c2f | 1291 | { |
c242c188 | 1292 | spa_t *spa = ztest_spa; |
428870ff | 1293 | nvlist_t *props = NULL; |
34dc7c2f BB |
1294 | int error; |
1295 | ||
428870ff BB |
1296 | VERIFY(nvlist_alloc(&props, NV_UNIQUE_NAME, 0) == 0); |
1297 | VERIFY(nvlist_add_uint64(props, zpool_prop_to_name(prop), value) == 0); | |
34dc7c2f | 1298 | |
428870ff BB |
1299 | error = spa_prop_set(spa, props); |
1300 | ||
1301 | nvlist_free(props); | |
1302 | ||
1303 | if (error == ENOSPC) { | |
1304 | ztest_record_enospc(FTAG); | |
34dc7c2f BB |
1305 | return (error); |
1306 | } | |
c99c9001 | 1307 | ASSERT0(error); |
34dc7c2f BB |
1308 | |
1309 | return (error); | |
1310 | } | |
1311 | ||
4807c0ba TC |
1312 | static int |
1313 | ztest_dmu_objset_own(const char *name, dmu_objset_type_t type, | |
1314 | boolean_t readonly, boolean_t decrypt, void *tag, objset_t **osp) | |
1315 | { | |
1316 | int err; | |
f44ad929 TC |
1317 | char *cp = NULL; |
1318 | char ddname[ZFS_MAX_DATASET_NAME_LEN]; | |
1319 | ||
1320 | strcpy(ddname, name); | |
1321 | cp = strchr(ddname, '@'); | |
1322 | if (cp != NULL) | |
1323 | *cp = '\0'; | |
4807c0ba TC |
1324 | |
1325 | err = dmu_objset_own(name, type, readonly, decrypt, tag, osp); | |
f44ad929 | 1326 | while (decrypt && err == EACCES) { |
4807c0ba TC |
1327 | dsl_crypto_params_t *dcp; |
1328 | nvlist_t *crypto_args = fnvlist_alloc(); | |
4807c0ba TC |
1329 | |
1330 | fnvlist_add_uint8_array(crypto_args, "wkeydata", | |
1331 | (uint8_t *)ztest_wkeydata, WRAPPING_KEY_LEN); | |
1332 | VERIFY0(dsl_crypto_params_create_nvlist(DCP_CMD_NONE, NULL, | |
1333 | crypto_args, &dcp)); | |
1334 | err = spa_keystore_load_wkey(ddname, dcp, B_FALSE); | |
1335 | dsl_crypto_params_free(dcp, B_FALSE); | |
1336 | fnvlist_free(crypto_args); | |
1337 | ||
f44ad929 TC |
1338 | if (err == EINVAL) { |
1339 | /* | |
1340 | * We couldn't load a key for this dataset so try | |
1341 | * the parent. This loop will eventually hit the | |
1342 | * encryption root since ztest only makes clones | |
1343 | * as children of their origin datasets. | |
1344 | */ | |
1345 | cp = strrchr(ddname, '/'); | |
1346 | if (cp == NULL) | |
1347 | return (err); | |
1348 | ||
1349 | *cp = '\0'; | |
1350 | err = EACCES; | |
1351 | continue; | |
1352 | } else if (err != 0) { | |
1353 | break; | |
1354 | } | |
4807c0ba TC |
1355 | |
1356 | err = dmu_objset_own(name, type, readonly, decrypt, tag, osp); | |
f44ad929 | 1357 | break; |
4807c0ba TC |
1358 | } |
1359 | ||
1360 | return (err); | |
1361 | } | |
1362 | ||
428870ff BB |
1363 | static void |
1364 | ztest_rll_init(rll_t *rll) | |
1365 | { | |
1366 | rll->rll_writer = NULL; | |
1367 | rll->rll_readers = 0; | |
1e33ac1e BB |
1368 | mutex_init(&rll->rll_lock, NULL, MUTEX_DEFAULT, NULL); |
1369 | cv_init(&rll->rll_cv, NULL, CV_DEFAULT, NULL); | |
428870ff | 1370 | } |
34dc7c2f | 1371 | |
428870ff BB |
1372 | static void |
1373 | ztest_rll_destroy(rll_t *rll) | |
34dc7c2f | 1374 | { |
428870ff BB |
1375 | ASSERT(rll->rll_writer == NULL); |
1376 | ASSERT(rll->rll_readers == 0); | |
1e33ac1e BB |
1377 | mutex_destroy(&rll->rll_lock); |
1378 | cv_destroy(&rll->rll_cv); | |
428870ff | 1379 | } |
34dc7c2f | 1380 | |
428870ff BB |
1381 | static void |
1382 | ztest_rll_lock(rll_t *rll, rl_type_t type) | |
1383 | { | |
1e33ac1e | 1384 | mutex_enter(&rll->rll_lock); |
34dc7c2f | 1385 | |
428870ff BB |
1386 | if (type == RL_READER) { |
1387 | while (rll->rll_writer != NULL) | |
1e33ac1e | 1388 | (void) cv_wait(&rll->rll_cv, &rll->rll_lock); |
428870ff BB |
1389 | rll->rll_readers++; |
1390 | } else { | |
1391 | while (rll->rll_writer != NULL || rll->rll_readers) | |
1e33ac1e | 1392 | (void) cv_wait(&rll->rll_cv, &rll->rll_lock); |
428870ff BB |
1393 | rll->rll_writer = curthread; |
1394 | } | |
34dc7c2f | 1395 | |
1e33ac1e | 1396 | mutex_exit(&rll->rll_lock); |
428870ff | 1397 | } |
34dc7c2f | 1398 | |
428870ff BB |
1399 | static void |
1400 | ztest_rll_unlock(rll_t *rll) | |
1401 | { | |
1e33ac1e | 1402 | mutex_enter(&rll->rll_lock); |
34dc7c2f | 1403 | |
428870ff BB |
1404 | if (rll->rll_writer) { |
1405 | ASSERT(rll->rll_readers == 0); | |
1406 | rll->rll_writer = NULL; | |
1407 | } else { | |
1408 | ASSERT(rll->rll_readers != 0); | |
1409 | ASSERT(rll->rll_writer == NULL); | |
1410 | rll->rll_readers--; | |
1411 | } | |
34dc7c2f | 1412 | |
428870ff | 1413 | if (rll->rll_writer == NULL && rll->rll_readers == 0) |
1e33ac1e | 1414 | cv_broadcast(&rll->rll_cv); |
428870ff | 1415 | |
1e33ac1e | 1416 | mutex_exit(&rll->rll_lock); |
34dc7c2f BB |
1417 | } |
1418 | ||
428870ff BB |
1419 | static void |
1420 | ztest_object_lock(ztest_ds_t *zd, uint64_t object, rl_type_t type) | |
b128c09f | 1421 | { |
428870ff | 1422 | rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; |
b128c09f | 1423 | |
428870ff BB |
1424 | ztest_rll_lock(rll, type); |
1425 | } | |
b128c09f | 1426 | |
428870ff BB |
1427 | static void |
1428 | ztest_object_unlock(ztest_ds_t *zd, uint64_t object) | |
1429 | { | |
1430 | rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; | |
b128c09f | 1431 | |
428870ff | 1432 | ztest_rll_unlock(rll); |
b128c09f BB |
1433 | } |
1434 | ||
5d43cc9a | 1435 | static rl_t * |
e3a07cd0 BP |
1436 | ztest_range_lock(ztest_ds_t *zd, uint64_t object, uint64_t offset, |
1437 | uint64_t size, rl_type_t type) | |
1438 | { | |
5d43cc9a MA |
1439 | uint64_t hash = object ^ (offset % (ZTEST_RANGE_LOCKS + 1)); |
1440 | rll_t *rll = &zd->zd_range_lock[hash & (ZTEST_RANGE_LOCKS - 1)]; | |
1441 | rl_t *rl; | |
1442 | ||
1443 | rl = umem_alloc(sizeof (*rl), UMEM_NOFAIL); | |
1444 | rl->rl_object = object; | |
1445 | rl->rl_offset = offset; | |
1446 | rl->rl_size = size; | |
1447 | rl->rl_lock = rll; | |
1448 | ||
1449 | ztest_rll_lock(rll, type); | |
1450 | ||
1451 | return (rl); | |
e3a07cd0 | 1452 | } |
34dc7c2f | 1453 | |
e3a07cd0 | 1454 | static void |
5d43cc9a | 1455 | ztest_range_unlock(rl_t *rl) |
e3a07cd0 | 1456 | { |
5d43cc9a MA |
1457 | rll_t *rll = rl->rl_lock; |
1458 | ||
1459 | ztest_rll_unlock(rll); | |
1460 | ||
1461 | umem_free(rl, sizeof (*rl)); | |
428870ff | 1462 | } |
34dc7c2f | 1463 | |
428870ff | 1464 | static void |
c242c188 | 1465 | ztest_zd_init(ztest_ds_t *zd, ztest_shared_ds_t *szd, objset_t *os) |
428870ff BB |
1466 | { |
1467 | zd->zd_os = os; | |
1468 | zd->zd_zilog = dmu_objset_zil(os); | |
c242c188 | 1469 | zd->zd_shared = szd; |
428870ff | 1470 | dmu_objset_name(os, zd->zd_name); |
d6320ddb | 1471 | int l; |
428870ff | 1472 | |
c242c188 CS |
1473 | if (zd->zd_shared != NULL) |
1474 | zd->zd_shared->zd_seq = 0; | |
1475 | ||
2d9142c9 | 1476 | VERIFY0(pthread_rwlock_init(&zd->zd_zilog_lock, NULL)); |
1e33ac1e | 1477 | mutex_init(&zd->zd_dirobj_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 1478 | |
d6320ddb | 1479 | for (l = 0; l < ZTEST_OBJECT_LOCKS; l++) |
428870ff | 1480 | ztest_rll_init(&zd->zd_object_lock[l]); |
34dc7c2f | 1481 | |
d6320ddb | 1482 | for (l = 0; l < ZTEST_RANGE_LOCKS; l++) |
5d43cc9a | 1483 | ztest_rll_init(&zd->zd_range_lock[l]); |
34dc7c2f BB |
1484 | } |
1485 | ||
428870ff BB |
1486 | static void |
1487 | ztest_zd_fini(ztest_ds_t *zd) | |
34dc7c2f | 1488 | { |
d6320ddb BB |
1489 | int l; |
1490 | ||
1e33ac1e | 1491 | mutex_destroy(&zd->zd_dirobj_lock); |
2d9142c9 | 1492 | (void) pthread_rwlock_destroy(&zd->zd_zilog_lock); |
34dc7c2f | 1493 | |
d6320ddb | 1494 | for (l = 0; l < ZTEST_OBJECT_LOCKS; l++) |
428870ff | 1495 | ztest_rll_destroy(&zd->zd_object_lock[l]); |
b128c09f | 1496 | |
d6320ddb | 1497 | for (l = 0; l < ZTEST_RANGE_LOCKS; l++) |
5d43cc9a | 1498 | ztest_rll_destroy(&zd->zd_range_lock[l]); |
428870ff | 1499 | } |
b128c09f | 1500 | |
428870ff | 1501 | #define TXG_MIGHTWAIT (ztest_random(10) == 0 ? TXG_NOWAIT : TXG_WAIT) |
b128c09f | 1502 | |
428870ff BB |
1503 | static uint64_t |
1504 | ztest_tx_assign(dmu_tx_t *tx, uint64_t txg_how, const char *tag) | |
1505 | { | |
1506 | uint64_t txg; | |
1507 | int error; | |
1508 | ||
1509 | /* | |
1510 | * Attempt to assign tx to some transaction group. | |
1511 | */ | |
1512 | error = dmu_tx_assign(tx, txg_how); | |
1513 | if (error) { | |
1514 | if (error == ERESTART) { | |
1515 | ASSERT(txg_how == TXG_NOWAIT); | |
1516 | dmu_tx_wait(tx); | |
1517 | } else { | |
1518 | ASSERT3U(error, ==, ENOSPC); | |
1519 | ztest_record_enospc(tag); | |
1520 | } | |
1521 | dmu_tx_abort(tx); | |
1522 | return (0); | |
1523 | } | |
1524 | txg = dmu_tx_get_txg(tx); | |
1525 | ASSERT(txg != 0); | |
1526 | return (txg); | |
1527 | } | |
1528 | ||
428870ff BB |
1529 | static void |
1530 | ztest_bt_generate(ztest_block_tag_t *bt, objset_t *os, uint64_t object, | |
50c957f7 NB |
1531 | uint64_t dnodesize, uint64_t offset, uint64_t gen, uint64_t txg, |
1532 | uint64_t crtxg) | |
428870ff BB |
1533 | { |
1534 | bt->bt_magic = BT_MAGIC; | |
1535 | bt->bt_objset = dmu_objset_id(os); | |
1536 | bt->bt_object = object; | |
50c957f7 | 1537 | bt->bt_dnodesize = dnodesize; |
428870ff BB |
1538 | bt->bt_offset = offset; |
1539 | bt->bt_gen = gen; | |
1540 | bt->bt_txg = txg; | |
1541 | bt->bt_crtxg = crtxg; | |
1542 | } | |
1543 | ||
1544 | static void | |
1545 | ztest_bt_verify(ztest_block_tag_t *bt, objset_t *os, uint64_t object, | |
50c957f7 NB |
1546 | uint64_t dnodesize, uint64_t offset, uint64_t gen, uint64_t txg, |
1547 | uint64_t crtxg) | |
428870ff | 1548 | { |
9b67f605 MA |
1549 | ASSERT3U(bt->bt_magic, ==, BT_MAGIC); |
1550 | ASSERT3U(bt->bt_objset, ==, dmu_objset_id(os)); | |
1551 | ASSERT3U(bt->bt_object, ==, object); | |
50c957f7 | 1552 | ASSERT3U(bt->bt_dnodesize, ==, dnodesize); |
9b67f605 MA |
1553 | ASSERT3U(bt->bt_offset, ==, offset); |
1554 | ASSERT3U(bt->bt_gen, <=, gen); | |
1555 | ASSERT3U(bt->bt_txg, <=, txg); | |
1556 | ASSERT3U(bt->bt_crtxg, ==, crtxg); | |
428870ff BB |
1557 | } |
1558 | ||
1559 | static ztest_block_tag_t * | |
1560 | ztest_bt_bonus(dmu_buf_t *db) | |
1561 | { | |
1562 | dmu_object_info_t doi; | |
1563 | ztest_block_tag_t *bt; | |
1564 | ||
1565 | dmu_object_info_from_db(db, &doi); | |
1566 | ASSERT3U(doi.doi_bonus_size, <=, db->db_size); | |
1567 | ASSERT3U(doi.doi_bonus_size, >=, sizeof (*bt)); | |
1568 | bt = (void *)((char *)db->db_data + doi.doi_bonus_size - sizeof (*bt)); | |
1569 | ||
1570 | return (bt); | |
1571 | } | |
1572 | ||
50c957f7 NB |
1573 | /* |
1574 | * Generate a token to fill up unused bonus buffer space. Try to make | |
1575 | * it unique to the object, generation, and offset to verify that data | |
1576 | * is not getting overwritten by data from other dnodes. | |
1577 | */ | |
1578 | #define ZTEST_BONUS_FILL_TOKEN(obj, ds, gen, offset) \ | |
1579 | (((ds) << 48) | ((gen) << 32) | ((obj) << 8) | (offset)) | |
1580 | ||
1581 | /* | |
1582 | * Fill up the unused bonus buffer region before the block tag with a | |
1583 | * verifiable pattern. Filling the whole bonus area with non-zero data | |
1584 | * helps ensure that all dnode traversal code properly skips the | |
1585 | * interior regions of large dnodes. | |
1586 | */ | |
65c7cc49 | 1587 | static void |
50c957f7 NB |
1588 | ztest_fill_unused_bonus(dmu_buf_t *db, void *end, uint64_t obj, |
1589 | objset_t *os, uint64_t gen) | |
1590 | { | |
1591 | uint64_t *bonusp; | |
1592 | ||
1593 | ASSERT(IS_P2ALIGNED((char *)end - (char *)db->db_data, 8)); | |
1594 | ||
1595 | for (bonusp = db->db_data; bonusp < (uint64_t *)end; bonusp++) { | |
1596 | uint64_t token = ZTEST_BONUS_FILL_TOKEN(obj, dmu_objset_id(os), | |
1597 | gen, bonusp - (uint64_t *)db->db_data); | |
1598 | *bonusp = token; | |
1599 | } | |
1600 | } | |
1601 | ||
1602 | /* | |
1603 | * Verify that the unused area of a bonus buffer is filled with the | |
1604 | * expected tokens. | |
1605 | */ | |
65c7cc49 | 1606 | static void |
50c957f7 NB |
1607 | ztest_verify_unused_bonus(dmu_buf_t *db, void *end, uint64_t obj, |
1608 | objset_t *os, uint64_t gen) | |
1609 | { | |
1610 | uint64_t *bonusp; | |
1611 | ||
1612 | for (bonusp = db->db_data; bonusp < (uint64_t *)end; bonusp++) { | |
1613 | uint64_t token = ZTEST_BONUS_FILL_TOKEN(obj, dmu_objset_id(os), | |
1614 | gen, bonusp - (uint64_t *)db->db_data); | |
1615 | VERIFY3U(*bonusp, ==, token); | |
1616 | } | |
1617 | } | |
1618 | ||
428870ff BB |
1619 | /* |
1620 | * ZIL logging ops | |
1621 | */ | |
1622 | ||
1623 | #define lrz_type lr_mode | |
1624 | #define lrz_blocksize lr_uid | |
1625 | #define lrz_ibshift lr_gid | |
1626 | #define lrz_bonustype lr_rdev | |
50c957f7 | 1627 | #define lrz_dnodesize lr_crtime[1] |
428870ff | 1628 | |
572e2857 | 1629 | static void |
428870ff BB |
1630 | ztest_log_create(ztest_ds_t *zd, dmu_tx_t *tx, lr_create_t *lr) |
1631 | { | |
1632 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1633 | size_t namesize = strlen(name) + 1; | |
1634 | itx_t *itx; | |
1635 | ||
1636 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1637 | return; |
428870ff BB |
1638 | |
1639 | itx = zil_itx_create(TX_CREATE, sizeof (*lr) + namesize); | |
1640 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1641 | sizeof (*lr) + namesize - sizeof (lr_t)); | |
1642 | ||
572e2857 | 1643 | zil_itx_assign(zd->zd_zilog, itx, tx); |
428870ff BB |
1644 | } |
1645 | ||
572e2857 BB |
1646 | static void |
1647 | ztest_log_remove(ztest_ds_t *zd, dmu_tx_t *tx, lr_remove_t *lr, uint64_t object) | |
428870ff BB |
1648 | { |
1649 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1650 | size_t namesize = strlen(name) + 1; | |
1651 | itx_t *itx; | |
1652 | ||
1653 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1654 | return; |
428870ff BB |
1655 | |
1656 | itx = zil_itx_create(TX_REMOVE, sizeof (*lr) + namesize); | |
1657 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1658 | sizeof (*lr) + namesize - sizeof (lr_t)); | |
1659 | ||
572e2857 BB |
1660 | itx->itx_oid = object; |
1661 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1662 | } |
1663 | ||
572e2857 | 1664 | static void |
428870ff BB |
1665 | ztest_log_write(ztest_ds_t *zd, dmu_tx_t *tx, lr_write_t *lr) |
1666 | { | |
1667 | itx_t *itx; | |
1668 | itx_wr_state_t write_state = ztest_random(WR_NUM_STATES); | |
1669 | ||
1670 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1671 | return; |
428870ff | 1672 | |
b8738257 | 1673 | if (lr->lr_length > zil_max_log_data(zd->zd_zilog)) |
428870ff BB |
1674 | write_state = WR_INDIRECT; |
1675 | ||
1676 | itx = zil_itx_create(TX_WRITE, | |
1677 | sizeof (*lr) + (write_state == WR_COPIED ? lr->lr_length : 0)); | |
1678 | ||
1679 | if (write_state == WR_COPIED && | |
1680 | dmu_read(zd->zd_os, lr->lr_foid, lr->lr_offset, lr->lr_length, | |
1681 | ((lr_write_t *)&itx->itx_lr) + 1, DMU_READ_NO_PREFETCH) != 0) { | |
1682 | zil_itx_destroy(itx); | |
1683 | itx = zil_itx_create(TX_WRITE, sizeof (*lr)); | |
1684 | write_state = WR_NEED_COPY; | |
1685 | } | |
1686 | itx->itx_private = zd; | |
1687 | itx->itx_wr_state = write_state; | |
1688 | itx->itx_sync = (ztest_random(8) == 0); | |
428870ff BB |
1689 | |
1690 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1691 | sizeof (*lr) - sizeof (lr_t)); | |
1692 | ||
572e2857 | 1693 | zil_itx_assign(zd->zd_zilog, itx, tx); |
428870ff BB |
1694 | } |
1695 | ||
572e2857 | 1696 | static void |
428870ff BB |
1697 | ztest_log_truncate(ztest_ds_t *zd, dmu_tx_t *tx, lr_truncate_t *lr) |
1698 | { | |
1699 | itx_t *itx; | |
1700 | ||
1701 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1702 | return; |
428870ff BB |
1703 | |
1704 | itx = zil_itx_create(TX_TRUNCATE, sizeof (*lr)); | |
1705 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1706 | sizeof (*lr) - sizeof (lr_t)); | |
1707 | ||
572e2857 BB |
1708 | itx->itx_sync = B_FALSE; |
1709 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1710 | } |
1711 | ||
572e2857 | 1712 | static void |
428870ff BB |
1713 | ztest_log_setattr(ztest_ds_t *zd, dmu_tx_t *tx, lr_setattr_t *lr) |
1714 | { | |
1715 | itx_t *itx; | |
1716 | ||
1717 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1718 | return; |
428870ff BB |
1719 | |
1720 | itx = zil_itx_create(TX_SETATTR, sizeof (*lr)); | |
1721 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1722 | sizeof (*lr) - sizeof (lr_t)); | |
1723 | ||
572e2857 BB |
1724 | itx->itx_sync = B_FALSE; |
1725 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1726 | } |
1727 | ||
1728 | /* | |
1729 | * ZIL replay ops | |
1730 | */ | |
1731 | static int | |
867959b5 | 1732 | ztest_replay_create(void *arg1, void *arg2, boolean_t byteswap) |
428870ff | 1733 | { |
867959b5 BB |
1734 | ztest_ds_t *zd = arg1; |
1735 | lr_create_t *lr = arg2; | |
428870ff BB |
1736 | char *name = (void *)(lr + 1); /* name follows lr */ |
1737 | objset_t *os = zd->zd_os; | |
1738 | ztest_block_tag_t *bbt; | |
1739 | dmu_buf_t *db; | |
1740 | dmu_tx_t *tx; | |
1741 | uint64_t txg; | |
1742 | int error = 0; | |
50c957f7 | 1743 | int bonuslen; |
428870ff BB |
1744 | |
1745 | if (byteswap) | |
1746 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1747 | ||
1748 | ASSERT(lr->lr_doid == ZTEST_DIROBJ); | |
1749 | ASSERT(name[0] != '\0'); | |
1750 | ||
1751 | tx = dmu_tx_create(os); | |
1752 | ||
1753 | dmu_tx_hold_zap(tx, lr->lr_doid, B_TRUE, name); | |
1754 | ||
1755 | if (lr->lrz_type == DMU_OT_ZAP_OTHER) { | |
1756 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, B_TRUE, NULL); | |
1757 | } else { | |
1758 | dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT); | |
1759 | } | |
1760 | ||
1761 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1762 | if (txg == 0) | |
1763 | return (ENOSPC); | |
1764 | ||
1765 | ASSERT(dmu_objset_zil(os)->zl_replay == !!lr->lr_foid); | |
50c957f7 | 1766 | bonuslen = DN_BONUS_SIZE(lr->lrz_dnodesize); |
428870ff BB |
1767 | |
1768 | if (lr->lrz_type == DMU_OT_ZAP_OTHER) { | |
1769 | if (lr->lr_foid == 0) { | |
50c957f7 | 1770 | lr->lr_foid = zap_create_dnsize(os, |
428870ff | 1771 | lr->lrz_type, lr->lrz_bonustype, |
50c957f7 | 1772 | bonuslen, lr->lrz_dnodesize, tx); |
428870ff | 1773 | } else { |
50c957f7 | 1774 | error = zap_create_claim_dnsize(os, lr->lr_foid, |
428870ff | 1775 | lr->lrz_type, lr->lrz_bonustype, |
50c957f7 | 1776 | bonuslen, lr->lrz_dnodesize, tx); |
428870ff BB |
1777 | } |
1778 | } else { | |
1779 | if (lr->lr_foid == 0) { | |
50c957f7 | 1780 | lr->lr_foid = dmu_object_alloc_dnsize(os, |
428870ff | 1781 | lr->lrz_type, 0, lr->lrz_bonustype, |
50c957f7 | 1782 | bonuslen, lr->lrz_dnodesize, tx); |
428870ff | 1783 | } else { |
50c957f7 | 1784 | error = dmu_object_claim_dnsize(os, lr->lr_foid, |
428870ff | 1785 | lr->lrz_type, 0, lr->lrz_bonustype, |
50c957f7 | 1786 | bonuslen, lr->lrz_dnodesize, tx); |
428870ff BB |
1787 | } |
1788 | } | |
1789 | ||
1790 | if (error) { | |
1791 | ASSERT3U(error, ==, EEXIST); | |
1792 | ASSERT(zd->zd_zilog->zl_replay); | |
1793 | dmu_tx_commit(tx); | |
1794 | return (error); | |
1795 | } | |
1796 | ||
1797 | ASSERT(lr->lr_foid != 0); | |
1798 | ||
1799 | if (lr->lrz_type != DMU_OT_ZAP_OTHER) | |
1800 | VERIFY3U(0, ==, dmu_object_set_blocksize(os, lr->lr_foid, | |
1801 | lr->lrz_blocksize, lr->lrz_ibshift, tx)); | |
1802 | ||
1803 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
1804 | bbt = ztest_bt_bonus(db); | |
1805 | dmu_buf_will_dirty(db, tx); | |
50c957f7 NB |
1806 | ztest_bt_generate(bbt, os, lr->lr_foid, lr->lrz_dnodesize, -1ULL, |
1807 | lr->lr_gen, txg, txg); | |
1808 | ztest_fill_unused_bonus(db, bbt, lr->lr_foid, os, lr->lr_gen); | |
428870ff BB |
1809 | dmu_buf_rele(db, FTAG); |
1810 | ||
1811 | VERIFY3U(0, ==, zap_add(os, lr->lr_doid, name, sizeof (uint64_t), 1, | |
1812 | &lr->lr_foid, tx)); | |
1813 | ||
1814 | (void) ztest_log_create(zd, tx, lr); | |
1815 | ||
1816 | dmu_tx_commit(tx); | |
1817 | ||
1818 | return (0); | |
1819 | } | |
1820 | ||
1821 | static int | |
867959b5 | 1822 | ztest_replay_remove(void *arg1, void *arg2, boolean_t byteswap) |
428870ff | 1823 | { |
867959b5 BB |
1824 | ztest_ds_t *zd = arg1; |
1825 | lr_remove_t *lr = arg2; | |
428870ff BB |
1826 | char *name = (void *)(lr + 1); /* name follows lr */ |
1827 | objset_t *os = zd->zd_os; | |
1828 | dmu_object_info_t doi; | |
1829 | dmu_tx_t *tx; | |
1830 | uint64_t object, txg; | |
1831 | ||
1832 | if (byteswap) | |
1833 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1834 | ||
1835 | ASSERT(lr->lr_doid == ZTEST_DIROBJ); | |
1836 | ASSERT(name[0] != '\0'); | |
1837 | ||
1838 | VERIFY3U(0, ==, | |
1839 | zap_lookup(os, lr->lr_doid, name, sizeof (object), 1, &object)); | |
1840 | ASSERT(object != 0); | |
1841 | ||
1842 | ztest_object_lock(zd, object, RL_WRITER); | |
1843 | ||
1844 | VERIFY3U(0, ==, dmu_object_info(os, object, &doi)); | |
1845 | ||
1846 | tx = dmu_tx_create(os); | |
1847 | ||
1848 | dmu_tx_hold_zap(tx, lr->lr_doid, B_FALSE, name); | |
1849 | dmu_tx_hold_free(tx, object, 0, DMU_OBJECT_END); | |
1850 | ||
1851 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1852 | if (txg == 0) { | |
1853 | ztest_object_unlock(zd, object); | |
1854 | return (ENOSPC); | |
1855 | } | |
1856 | ||
1857 | if (doi.doi_type == DMU_OT_ZAP_OTHER) { | |
1858 | VERIFY3U(0, ==, zap_destroy(os, object, tx)); | |
1859 | } else { | |
1860 | VERIFY3U(0, ==, dmu_object_free(os, object, tx)); | |
1861 | } | |
1862 | ||
1863 | VERIFY3U(0, ==, zap_remove(os, lr->lr_doid, name, tx)); | |
1864 | ||
572e2857 | 1865 | (void) ztest_log_remove(zd, tx, lr, object); |
428870ff BB |
1866 | |
1867 | dmu_tx_commit(tx); | |
1868 | ||
1869 | ztest_object_unlock(zd, object); | |
1870 | ||
1871 | return (0); | |
1872 | } | |
1873 | ||
1874 | static int | |
867959b5 | 1875 | ztest_replay_write(void *arg1, void *arg2, boolean_t byteswap) |
428870ff | 1876 | { |
867959b5 BB |
1877 | ztest_ds_t *zd = arg1; |
1878 | lr_write_t *lr = arg2; | |
428870ff BB |
1879 | objset_t *os = zd->zd_os; |
1880 | void *data = lr + 1; /* data follows lr */ | |
1881 | uint64_t offset, length; | |
1882 | ztest_block_tag_t *bt = data; | |
1883 | ztest_block_tag_t *bbt; | |
1884 | uint64_t gen, txg, lrtxg, crtxg; | |
1885 | dmu_object_info_t doi; | |
1886 | dmu_tx_t *tx; | |
1887 | dmu_buf_t *db; | |
1888 | arc_buf_t *abuf = NULL; | |
5d43cc9a | 1889 | rl_t *rl; |
428870ff BB |
1890 | |
1891 | if (byteswap) | |
1892 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1893 | ||
1894 | offset = lr->lr_offset; | |
1895 | length = lr->lr_length; | |
1896 | ||
1897 | /* If it's a dmu_sync() block, write the whole block */ | |
1898 | if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) { | |
1899 | uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr); | |
1900 | if (length < blocksize) { | |
1901 | offset -= offset % blocksize; | |
1902 | length = blocksize; | |
1903 | } | |
1904 | } | |
1905 | ||
1906 | if (bt->bt_magic == BSWAP_64(BT_MAGIC)) | |
1907 | byteswap_uint64_array(bt, sizeof (*bt)); | |
1908 | ||
1909 | if (bt->bt_magic != BT_MAGIC) | |
1910 | bt = NULL; | |
1911 | ||
1912 | ztest_object_lock(zd, lr->lr_foid, RL_READER); | |
1913 | rl = ztest_range_lock(zd, lr->lr_foid, offset, length, RL_WRITER); | |
1914 | ||
1915 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
1916 | ||
1917 | dmu_object_info_from_db(db, &doi); | |
1918 | ||
1919 | bbt = ztest_bt_bonus(db); | |
1920 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
1921 | gen = bbt->bt_gen; | |
1922 | crtxg = bbt->bt_crtxg; | |
1923 | lrtxg = lr->lr_common.lrc_txg; | |
1924 | ||
1925 | tx = dmu_tx_create(os); | |
1926 | ||
1927 | dmu_tx_hold_write(tx, lr->lr_foid, offset, length); | |
1928 | ||
1929 | if (ztest_random(8) == 0 && length == doi.doi_data_block_size && | |
1930 | P2PHASE(offset, length) == 0) | |
1931 | abuf = dmu_request_arcbuf(db, length); | |
1932 | ||
1933 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1934 | if (txg == 0) { | |
1935 | if (abuf != NULL) | |
1936 | dmu_return_arcbuf(abuf); | |
1937 | dmu_buf_rele(db, FTAG); | |
5d43cc9a | 1938 | ztest_range_unlock(rl); |
428870ff BB |
1939 | ztest_object_unlock(zd, lr->lr_foid); |
1940 | return (ENOSPC); | |
1941 | } | |
1942 | ||
1943 | if (bt != NULL) { | |
1944 | /* | |
1945 | * Usually, verify the old data before writing new data -- | |
1946 | * but not always, because we also want to verify correct | |
1947 | * behavior when the data was not recently read into cache. | |
1948 | */ | |
1949 | ASSERT(offset % doi.doi_data_block_size == 0); | |
1950 | if (ztest_random(4) != 0) { | |
1951 | int prefetch = ztest_random(2) ? | |
1952 | DMU_READ_PREFETCH : DMU_READ_NO_PREFETCH; | |
1953 | ztest_block_tag_t rbt; | |
1954 | ||
1955 | VERIFY(dmu_read(os, lr->lr_foid, offset, | |
1956 | sizeof (rbt), &rbt, prefetch) == 0); | |
1957 | if (rbt.bt_magic == BT_MAGIC) { | |
50c957f7 | 1958 | ztest_bt_verify(&rbt, os, lr->lr_foid, 0, |
428870ff BB |
1959 | offset, gen, txg, crtxg); |
1960 | } | |
1961 | } | |
1962 | ||
1963 | /* | |
1964 | * Writes can appear to be newer than the bonus buffer because | |
1965 | * the ztest_get_data() callback does a dmu_read() of the | |
1966 | * open-context data, which may be different than the data | |
1967 | * as it was when the write was generated. | |
1968 | */ | |
1969 | if (zd->zd_zilog->zl_replay) { | |
50c957f7 | 1970 | ztest_bt_verify(bt, os, lr->lr_foid, 0, offset, |
428870ff BB |
1971 | MAX(gen, bt->bt_gen), MAX(txg, lrtxg), |
1972 | bt->bt_crtxg); | |
1973 | } | |
1974 | ||
1975 | /* | |
1976 | * Set the bt's gen/txg to the bonus buffer's gen/txg | |
1977 | * so that all of the usual ASSERTs will work. | |
1978 | */ | |
50c957f7 NB |
1979 | ztest_bt_generate(bt, os, lr->lr_foid, 0, offset, gen, txg, |
1980 | crtxg); | |
428870ff BB |
1981 | } |
1982 | ||
1983 | if (abuf == NULL) { | |
1984 | dmu_write(os, lr->lr_foid, offset, length, data, tx); | |
1985 | } else { | |
1986 | bcopy(data, abuf->b_data, length); | |
440a3eb9 | 1987 | dmu_assign_arcbuf_by_dbuf(db, offset, abuf, tx); |
428870ff BB |
1988 | } |
1989 | ||
1990 | (void) ztest_log_write(zd, tx, lr); | |
1991 | ||
1992 | dmu_buf_rele(db, FTAG); | |
1993 | ||
1994 | dmu_tx_commit(tx); | |
1995 | ||
5d43cc9a | 1996 | ztest_range_unlock(rl); |
428870ff BB |
1997 | ztest_object_unlock(zd, lr->lr_foid); |
1998 | ||
1999 | return (0); | |
2000 | } | |
2001 | ||
2002 | static int | |
867959b5 | 2003 | ztest_replay_truncate(void *arg1, void *arg2, boolean_t byteswap) |
428870ff | 2004 | { |
867959b5 BB |
2005 | ztest_ds_t *zd = arg1; |
2006 | lr_truncate_t *lr = arg2; | |
428870ff BB |
2007 | objset_t *os = zd->zd_os; |
2008 | dmu_tx_t *tx; | |
2009 | uint64_t txg; | |
5d43cc9a | 2010 | rl_t *rl; |
428870ff BB |
2011 | |
2012 | if (byteswap) | |
2013 | byteswap_uint64_array(lr, sizeof (*lr)); | |
2014 | ||
2015 | ztest_object_lock(zd, lr->lr_foid, RL_READER); | |
2016 | rl = ztest_range_lock(zd, lr->lr_foid, lr->lr_offset, lr->lr_length, | |
2017 | RL_WRITER); | |
2018 | ||
2019 | tx = dmu_tx_create(os); | |
2020 | ||
2021 | dmu_tx_hold_free(tx, lr->lr_foid, lr->lr_offset, lr->lr_length); | |
2022 | ||
2023 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
2024 | if (txg == 0) { | |
5d43cc9a | 2025 | ztest_range_unlock(rl); |
428870ff BB |
2026 | ztest_object_unlock(zd, lr->lr_foid); |
2027 | return (ENOSPC); | |
2028 | } | |
2029 | ||
2030 | VERIFY(dmu_free_range(os, lr->lr_foid, lr->lr_offset, | |
2031 | lr->lr_length, tx) == 0); | |
2032 | ||
2033 | (void) ztest_log_truncate(zd, tx, lr); | |
2034 | ||
2035 | dmu_tx_commit(tx); | |
2036 | ||
5d43cc9a | 2037 | ztest_range_unlock(rl); |
428870ff BB |
2038 | ztest_object_unlock(zd, lr->lr_foid); |
2039 | ||
2040 | return (0); | |
2041 | } | |
2042 | ||
2043 | static int | |
867959b5 | 2044 | ztest_replay_setattr(void *arg1, void *arg2, boolean_t byteswap) |
428870ff | 2045 | { |
867959b5 BB |
2046 | ztest_ds_t *zd = arg1; |
2047 | lr_setattr_t *lr = arg2; | |
428870ff BB |
2048 | objset_t *os = zd->zd_os; |
2049 | dmu_tx_t *tx; | |
2050 | dmu_buf_t *db; | |
2051 | ztest_block_tag_t *bbt; | |
50c957f7 | 2052 | uint64_t txg, lrtxg, crtxg, dnodesize; |
428870ff BB |
2053 | |
2054 | if (byteswap) | |
2055 | byteswap_uint64_array(lr, sizeof (*lr)); | |
2056 | ||
2057 | ztest_object_lock(zd, lr->lr_foid, RL_WRITER); | |
2058 | ||
2059 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
2060 | ||
2061 | tx = dmu_tx_create(os); | |
2062 | dmu_tx_hold_bonus(tx, lr->lr_foid); | |
2063 | ||
2064 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
2065 | if (txg == 0) { | |
2066 | dmu_buf_rele(db, FTAG); | |
2067 | ztest_object_unlock(zd, lr->lr_foid); | |
2068 | return (ENOSPC); | |
2069 | } | |
2070 | ||
2071 | bbt = ztest_bt_bonus(db); | |
2072 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
2073 | crtxg = bbt->bt_crtxg; | |
2074 | lrtxg = lr->lr_common.lrc_txg; | |
50c957f7 | 2075 | dnodesize = bbt->bt_dnodesize; |
428870ff BB |
2076 | |
2077 | if (zd->zd_zilog->zl_replay) { | |
2078 | ASSERT(lr->lr_size != 0); | |
2079 | ASSERT(lr->lr_mode != 0); | |
2080 | ASSERT(lrtxg != 0); | |
2081 | } else { | |
2082 | /* | |
2083 | * Randomly change the size and increment the generation. | |
2084 | */ | |
2085 | lr->lr_size = (ztest_random(db->db_size / sizeof (*bbt)) + 1) * | |
2086 | sizeof (*bbt); | |
2087 | lr->lr_mode = bbt->bt_gen + 1; | |
2088 | ASSERT(lrtxg == 0); | |
2089 | } | |
2090 | ||
2091 | /* | |
2092 | * Verify that the current bonus buffer is not newer than our txg. | |
2093 | */ | |
50c957f7 | 2094 | ztest_bt_verify(bbt, os, lr->lr_foid, dnodesize, -1ULL, lr->lr_mode, |
428870ff BB |
2095 | MAX(txg, lrtxg), crtxg); |
2096 | ||
2097 | dmu_buf_will_dirty(db, tx); | |
2098 | ||
2099 | ASSERT3U(lr->lr_size, >=, sizeof (*bbt)); | |
2100 | ASSERT3U(lr->lr_size, <=, db->db_size); | |
c99c9001 | 2101 | VERIFY0(dmu_set_bonus(db, lr->lr_size, tx)); |
428870ff BB |
2102 | bbt = ztest_bt_bonus(db); |
2103 | ||
50c957f7 NB |
2104 | ztest_bt_generate(bbt, os, lr->lr_foid, dnodesize, -1ULL, lr->lr_mode, |
2105 | txg, crtxg); | |
2106 | ztest_fill_unused_bonus(db, bbt, lr->lr_foid, os, bbt->bt_gen); | |
428870ff BB |
2107 | dmu_buf_rele(db, FTAG); |
2108 | ||
2109 | (void) ztest_log_setattr(zd, tx, lr); | |
2110 | ||
2111 | dmu_tx_commit(tx); | |
2112 | ||
2113 | ztest_object_unlock(zd, lr->lr_foid); | |
2114 | ||
2115 | return (0); | |
2116 | } | |
2117 | ||
867959b5 BB |
2118 | zil_replay_func_t *ztest_replay_vector[TX_MAX_TYPE] = { |
2119 | NULL, /* 0 no such transaction type */ | |
2120 | ztest_replay_create, /* TX_CREATE */ | |
2121 | NULL, /* TX_MKDIR */ | |
2122 | NULL, /* TX_MKXATTR */ | |
2123 | NULL, /* TX_SYMLINK */ | |
2124 | ztest_replay_remove, /* TX_REMOVE */ | |
2125 | NULL, /* TX_RMDIR */ | |
2126 | NULL, /* TX_LINK */ | |
2127 | NULL, /* TX_RENAME */ | |
2128 | ztest_replay_write, /* TX_WRITE */ | |
2129 | ztest_replay_truncate, /* TX_TRUNCATE */ | |
2130 | ztest_replay_setattr, /* TX_SETATTR */ | |
2131 | NULL, /* TX_ACL */ | |
2132 | NULL, /* TX_CREATE_ACL */ | |
2133 | NULL, /* TX_CREATE_ATTR */ | |
2134 | NULL, /* TX_CREATE_ACL_ATTR */ | |
2135 | NULL, /* TX_MKDIR_ACL */ | |
2136 | NULL, /* TX_MKDIR_ATTR */ | |
2137 | NULL, /* TX_MKDIR_ACL_ATTR */ | |
2138 | NULL, /* TX_WRITE2 */ | |
428870ff BB |
2139 | }; |
2140 | ||
2141 | /* | |
2142 | * ZIL get_data callbacks | |
2143 | */ | |
2144 | ||
900d09b2 | 2145 | /* ARGSUSED */ |
428870ff BB |
2146 | static void |
2147 | ztest_get_done(zgd_t *zgd, int error) | |
2148 | { | |
5d43cc9a MA |
2149 | ztest_ds_t *zd = zgd->zgd_private; |
2150 | uint64_t object = ((rl_t *)zgd->zgd_lr)->rl_object; | |
428870ff BB |
2151 | |
2152 | if (zgd->zgd_db) | |
2153 | dmu_buf_rele(zgd->zgd_db, zgd); | |
2154 | ||
5d43cc9a | 2155 | ztest_range_unlock((rl_t *)zgd->zgd_lr); |
428870ff BB |
2156 | ztest_object_unlock(zd, object); |
2157 | ||
428870ff BB |
2158 | umem_free(zgd, sizeof (*zgd)); |
2159 | } | |
2160 | ||
2161 | static int | |
1ce23dca PS |
2162 | ztest_get_data(void *arg, lr_write_t *lr, char *buf, struct lwb *lwb, |
2163 | zio_t *zio) | |
428870ff BB |
2164 | { |
2165 | ztest_ds_t *zd = arg; | |
2166 | objset_t *os = zd->zd_os; | |
2167 | uint64_t object = lr->lr_foid; | |
2168 | uint64_t offset = lr->lr_offset; | |
2169 | uint64_t size = lr->lr_length; | |
428870ff BB |
2170 | uint64_t txg = lr->lr_common.lrc_txg; |
2171 | uint64_t crtxg; | |
2172 | dmu_object_info_t doi; | |
2173 | dmu_buf_t *db; | |
2174 | zgd_t *zgd; | |
2175 | int error; | |
2176 | ||
1ce23dca PS |
2177 | ASSERT3P(lwb, !=, NULL); |
2178 | ASSERT3P(zio, !=, NULL); | |
2179 | ASSERT3U(size, !=, 0); | |
2180 | ||
428870ff BB |
2181 | ztest_object_lock(zd, object, RL_READER); |
2182 | error = dmu_bonus_hold(os, object, FTAG, &db); | |
2183 | if (error) { | |
2184 | ztest_object_unlock(zd, object); | |
2185 | return (error); | |
2186 | } | |
2187 | ||
2188 | crtxg = ztest_bt_bonus(db)->bt_crtxg; | |
2189 | ||
2190 | if (crtxg == 0 || crtxg > txg) { | |
2191 | dmu_buf_rele(db, FTAG); | |
2192 | ztest_object_unlock(zd, object); | |
2193 | return (ENOENT); | |
2194 | } | |
2195 | ||
2196 | dmu_object_info_from_db(db, &doi); | |
2197 | dmu_buf_rele(db, FTAG); | |
2198 | db = NULL; | |
2199 | ||
2200 | zgd = umem_zalloc(sizeof (*zgd), UMEM_NOFAIL); | |
1ce23dca | 2201 | zgd->zgd_lwb = lwb; |
5d43cc9a | 2202 | zgd->zgd_private = zd; |
428870ff BB |
2203 | |
2204 | if (buf != NULL) { /* immediate write */ | |
bd4dde8e | 2205 | zgd->zgd_lr = (struct zfs_locked_range *)ztest_range_lock(zd, |
5d43cc9a | 2206 | object, offset, size, RL_READER); |
428870ff BB |
2207 | |
2208 | error = dmu_read(os, object, offset, size, buf, | |
2209 | DMU_READ_NO_PREFETCH); | |
2210 | ASSERT(error == 0); | |
2211 | } else { | |
2212 | size = doi.doi_data_block_size; | |
2213 | if (ISP2(size)) { | |
2214 | offset = P2ALIGN(offset, size); | |
2215 | } else { | |
2216 | ASSERT(offset < size); | |
2217 | offset = 0; | |
2218 | } | |
2219 | ||
bd4dde8e | 2220 | zgd->zgd_lr = (struct zfs_locked_range *)ztest_range_lock(zd, |
5d43cc9a | 2221 | object, offset, size, RL_READER); |
428870ff BB |
2222 | |
2223 | error = dmu_buf_hold(os, object, offset, zgd, &db, | |
2224 | DMU_READ_NO_PREFETCH); | |
2225 | ||
2226 | if (error == 0) { | |
02dc43bc | 2227 | blkptr_t *bp = &lr->lr_blkptr; |
03c6040b | 2228 | |
428870ff BB |
2229 | zgd->zgd_db = db; |
2230 | zgd->zgd_bp = bp; | |
2231 | ||
2232 | ASSERT(db->db_offset == offset); | |
2233 | ASSERT(db->db_size == size); | |
2234 | ||
2235 | error = dmu_sync(zio, lr->lr_common.lrc_txg, | |
2236 | ztest_get_done, zgd); | |
2237 | ||
2238 | if (error == 0) | |
2239 | return (0); | |
2240 | } | |
2241 | } | |
2242 | ||
2243 | ztest_get_done(zgd, error); | |
2244 | ||
2245 | return (error); | |
2246 | } | |
2247 | ||
2248 | static void * | |
2249 | ztest_lr_alloc(size_t lrsize, char *name) | |
2250 | { | |
2251 | char *lr; | |
2252 | size_t namesize = name ? strlen(name) + 1 : 0; | |
2253 | ||
2254 | lr = umem_zalloc(lrsize + namesize, UMEM_NOFAIL); | |
2255 | ||
2256 | if (name) | |
2257 | bcopy(name, lr + lrsize, namesize); | |
2258 | ||
2259 | return (lr); | |
2260 | } | |
2261 | ||
65c7cc49 | 2262 | static void |
428870ff BB |
2263 | ztest_lr_free(void *lr, size_t lrsize, char *name) |
2264 | { | |
2265 | size_t namesize = name ? strlen(name) + 1 : 0; | |
2266 | ||
2267 | umem_free(lr, lrsize + namesize); | |
2268 | } | |
2269 | ||
2270 | /* | |
2271 | * Lookup a bunch of objects. Returns the number of objects not found. | |
2272 | */ | |
2273 | static int | |
2274 | ztest_lookup(ztest_ds_t *zd, ztest_od_t *od, int count) | |
2275 | { | |
2276 | int missing = 0; | |
2277 | int error; | |
d6320ddb | 2278 | int i; |
428870ff | 2279 | |
c25b8f99 | 2280 | ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock)); |
428870ff | 2281 | |
d6320ddb | 2282 | for (i = 0; i < count; i++, od++) { |
428870ff BB |
2283 | od->od_object = 0; |
2284 | error = zap_lookup(zd->zd_os, od->od_dir, od->od_name, | |
2285 | sizeof (uint64_t), 1, &od->od_object); | |
2286 | if (error) { | |
2287 | ASSERT(error == ENOENT); | |
2288 | ASSERT(od->od_object == 0); | |
2289 | missing++; | |
2290 | } else { | |
2291 | dmu_buf_t *db; | |
2292 | ztest_block_tag_t *bbt; | |
2293 | dmu_object_info_t doi; | |
2294 | ||
2295 | ASSERT(od->od_object != 0); | |
2296 | ASSERT(missing == 0); /* there should be no gaps */ | |
2297 | ||
2298 | ztest_object_lock(zd, od->od_object, RL_READER); | |
2299 | VERIFY3U(0, ==, dmu_bonus_hold(zd->zd_os, | |
2300 | od->od_object, FTAG, &db)); | |
2301 | dmu_object_info_from_db(db, &doi); | |
2302 | bbt = ztest_bt_bonus(db); | |
2303 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
2304 | od->od_type = doi.doi_type; | |
2305 | od->od_blocksize = doi.doi_data_block_size; | |
2306 | od->od_gen = bbt->bt_gen; | |
2307 | dmu_buf_rele(db, FTAG); | |
2308 | ztest_object_unlock(zd, od->od_object); | |
2309 | } | |
2310 | } | |
2311 | ||
2312 | return (missing); | |
2313 | } | |
2314 | ||
2315 | static int | |
2316 | ztest_create(ztest_ds_t *zd, ztest_od_t *od, int count) | |
2317 | { | |
2318 | int missing = 0; | |
d6320ddb | 2319 | int i; |
428870ff | 2320 | |
c25b8f99 | 2321 | ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock)); |
428870ff | 2322 | |
d6320ddb | 2323 | for (i = 0; i < count; i++, od++) { |
428870ff BB |
2324 | if (missing) { |
2325 | od->od_object = 0; | |
2326 | missing++; | |
2327 | continue; | |
2328 | } | |
2329 | ||
2330 | lr_create_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name); | |
2331 | ||
2332 | lr->lr_doid = od->od_dir; | |
2333 | lr->lr_foid = 0; /* 0 to allocate, > 0 to claim */ | |
2334 | lr->lrz_type = od->od_crtype; | |
2335 | lr->lrz_blocksize = od->od_crblocksize; | |
2336 | lr->lrz_ibshift = ztest_random_ibshift(); | |
2337 | lr->lrz_bonustype = DMU_OT_UINT64_OTHER; | |
50c957f7 | 2338 | lr->lrz_dnodesize = od->od_crdnodesize; |
428870ff BB |
2339 | lr->lr_gen = od->od_crgen; |
2340 | lr->lr_crtime[0] = time(NULL); | |
2341 | ||
2342 | if (ztest_replay_create(zd, lr, B_FALSE) != 0) { | |
2343 | ASSERT(missing == 0); | |
2344 | od->od_object = 0; | |
2345 | missing++; | |
2346 | } else { | |
2347 | od->od_object = lr->lr_foid; | |
2348 | od->od_type = od->od_crtype; | |
2349 | od->od_blocksize = od->od_crblocksize; | |
2350 | od->od_gen = od->od_crgen; | |
2351 | ASSERT(od->od_object != 0); | |
2352 | } | |
2353 | ||
2354 | ztest_lr_free(lr, sizeof (*lr), od->od_name); | |
2355 | } | |
2356 | ||
2357 | return (missing); | |
2358 | } | |
2359 | ||
2360 | static int | |
2361 | ztest_remove(ztest_ds_t *zd, ztest_od_t *od, int count) | |
2362 | { | |
2363 | int missing = 0; | |
2364 | int error; | |
d6320ddb | 2365 | int i; |
428870ff | 2366 | |
c25b8f99 | 2367 | ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock)); |
428870ff BB |
2368 | |
2369 | od += count - 1; | |
2370 | ||
d6320ddb | 2371 | for (i = count - 1; i >= 0; i--, od--) { |
428870ff BB |
2372 | if (missing) { |
2373 | missing++; | |
2374 | continue; | |
2375 | } | |
2376 | ||
03c6040b GW |
2377 | /* |
2378 | * No object was found. | |
2379 | */ | |
428870ff BB |
2380 | if (od->od_object == 0) |
2381 | continue; | |
2382 | ||
2383 | lr_remove_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name); | |
2384 | ||
2385 | lr->lr_doid = od->od_dir; | |
2386 | ||
2387 | if ((error = ztest_replay_remove(zd, lr, B_FALSE)) != 0) { | |
2388 | ASSERT3U(error, ==, ENOSPC); | |
2389 | missing++; | |
2390 | } else { | |
2391 | od->od_object = 0; | |
2392 | } | |
2393 | ztest_lr_free(lr, sizeof (*lr), od->od_name); | |
2394 | } | |
2395 | ||
2396 | return (missing); | |
2397 | } | |
2398 | ||
2399 | static int | |
2400 | ztest_write(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size, | |
2401 | void *data) | |
2402 | { | |
2403 | lr_write_t *lr; | |
2404 | int error; | |
2405 | ||
2406 | lr = ztest_lr_alloc(sizeof (*lr) + size, NULL); | |
2407 | ||
2408 | lr->lr_foid = object; | |
2409 | lr->lr_offset = offset; | |
2410 | lr->lr_length = size; | |
2411 | lr->lr_blkoff = 0; | |
2412 | BP_ZERO(&lr->lr_blkptr); | |
2413 | ||
2414 | bcopy(data, lr + 1, size); | |
2415 | ||
2416 | error = ztest_replay_write(zd, lr, B_FALSE); | |
2417 | ||
2418 | ztest_lr_free(lr, sizeof (*lr) + size, NULL); | |
2419 | ||
2420 | return (error); | |
2421 | } | |
2422 | ||
2423 | static int | |
2424 | ztest_truncate(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size) | |
2425 | { | |
2426 | lr_truncate_t *lr; | |
2427 | int error; | |
2428 | ||
2429 | lr = ztest_lr_alloc(sizeof (*lr), NULL); | |
2430 | ||
2431 | lr->lr_foid = object; | |
2432 | lr->lr_offset = offset; | |
2433 | lr->lr_length = size; | |
2434 | ||
2435 | error = ztest_replay_truncate(zd, lr, B_FALSE); | |
2436 | ||
2437 | ztest_lr_free(lr, sizeof (*lr), NULL); | |
2438 | ||
2439 | return (error); | |
2440 | } | |
2441 | ||
2442 | static int | |
2443 | ztest_setattr(ztest_ds_t *zd, uint64_t object) | |
2444 | { | |
2445 | lr_setattr_t *lr; | |
2446 | int error; | |
2447 | ||
2448 | lr = ztest_lr_alloc(sizeof (*lr), NULL); | |
2449 | ||
2450 | lr->lr_foid = object; | |
2451 | lr->lr_size = 0; | |
2452 | lr->lr_mode = 0; | |
2453 | ||
2454 | error = ztest_replay_setattr(zd, lr, B_FALSE); | |
2455 | ||
2456 | ztest_lr_free(lr, sizeof (*lr), NULL); | |
2457 | ||
2458 | return (error); | |
2459 | } | |
2460 | ||
2461 | static void | |
2462 | ztest_prealloc(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size) | |
2463 | { | |
2464 | objset_t *os = zd->zd_os; | |
2465 | dmu_tx_t *tx; | |
2466 | uint64_t txg; | |
5d43cc9a | 2467 | rl_t *rl; |
428870ff BB |
2468 | |
2469 | txg_wait_synced(dmu_objset_pool(os), 0); | |
2470 | ||
2471 | ztest_object_lock(zd, object, RL_READER); | |
2472 | rl = ztest_range_lock(zd, object, offset, size, RL_WRITER); | |
2473 | ||
2474 | tx = dmu_tx_create(os); | |
2475 | ||
2476 | dmu_tx_hold_write(tx, object, offset, size); | |
2477 | ||
2478 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
2479 | ||
2480 | if (txg != 0) { | |
2481 | dmu_prealloc(os, object, offset, size, tx); | |
2482 | dmu_tx_commit(tx); | |
2483 | txg_wait_synced(dmu_objset_pool(os), txg); | |
2484 | } else { | |
2485 | (void) dmu_free_long_range(os, object, offset, size); | |
2486 | } | |
2487 | ||
5d43cc9a | 2488 | ztest_range_unlock(rl); |
428870ff BB |
2489 | ztest_object_unlock(zd, object); |
2490 | } | |
2491 | ||
2492 | static void | |
2493 | ztest_io(ztest_ds_t *zd, uint64_t object, uint64_t offset) | |
2494 | { | |
03c6040b | 2495 | int err; |
428870ff BB |
2496 | ztest_block_tag_t wbt; |
2497 | dmu_object_info_t doi; | |
2498 | enum ztest_io_type io_type; | |
2499 | uint64_t blocksize; | |
2500 | void *data; | |
2501 | ||
2502 | VERIFY(dmu_object_info(zd->zd_os, object, &doi) == 0); | |
2503 | blocksize = doi.doi_data_block_size; | |
2504 | data = umem_alloc(blocksize, UMEM_NOFAIL); | |
2505 | ||
2506 | /* | |
2507 | * Pick an i/o type at random, biased toward writing block tags. | |
2508 | */ | |
2509 | io_type = ztest_random(ZTEST_IO_TYPES); | |
2510 | if (ztest_random(2) == 0) | |
2511 | io_type = ZTEST_IO_WRITE_TAG; | |
2512 | ||
2d9142c9 | 2513 | (void) pthread_rwlock_rdlock(&zd->zd_zilog_lock); |
3e31d2b0 | 2514 | |
428870ff BB |
2515 | switch (io_type) { |
2516 | ||
2517 | case ZTEST_IO_WRITE_TAG: | |
50c957f7 NB |
2518 | ztest_bt_generate(&wbt, zd->zd_os, object, doi.doi_dnodesize, |
2519 | offset, 0, 0, 0); | |
428870ff BB |
2520 | (void) ztest_write(zd, object, offset, sizeof (wbt), &wbt); |
2521 | break; | |
2522 | ||
2523 | case ZTEST_IO_WRITE_PATTERN: | |
2524 | (void) memset(data, 'a' + (object + offset) % 5, blocksize); | |
2525 | if (ztest_random(2) == 0) { | |
2526 | /* | |
2527 | * Induce fletcher2 collisions to ensure that | |
2528 | * zio_ddt_collision() detects and resolves them | |
2529 | * when using fletcher2-verify for deduplication. | |
2530 | */ | |
2531 | ((uint64_t *)data)[0] ^= 1ULL << 63; | |
2532 | ((uint64_t *)data)[4] ^= 1ULL << 63; | |
2533 | } | |
2534 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2535 | break; | |
2536 | ||
2537 | case ZTEST_IO_WRITE_ZEROES: | |
2538 | bzero(data, blocksize); | |
2539 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2540 | break; | |
2541 | ||
2542 | case ZTEST_IO_TRUNCATE: | |
2543 | (void) ztest_truncate(zd, object, offset, blocksize); | |
2544 | break; | |
2545 | ||
2546 | case ZTEST_IO_SETATTR: | |
2547 | (void) ztest_setattr(zd, object); | |
2548 | break; | |
e75c13c3 BB |
2549 | default: |
2550 | break; | |
03c6040b GW |
2551 | |
2552 | case ZTEST_IO_REWRITE: | |
2d9142c9 | 2553 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
03c6040b GW |
2554 | err = ztest_dsl_prop_set_uint64(zd->zd_name, |
2555 | ZFS_PROP_CHECKSUM, spa_dedup_checksum(ztest_spa), | |
2556 | B_FALSE); | |
2557 | VERIFY(err == 0 || err == ENOSPC); | |
2558 | err = ztest_dsl_prop_set_uint64(zd->zd_name, | |
2559 | ZFS_PROP_COMPRESSION, | |
2560 | ztest_random_dsl_prop(ZFS_PROP_COMPRESSION), | |
2561 | B_FALSE); | |
2562 | VERIFY(err == 0 || err == ENOSPC); | |
2d9142c9 | 2563 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
03c6040b GW |
2564 | |
2565 | VERIFY0(dmu_read(zd->zd_os, object, offset, blocksize, data, | |
2566 | DMU_READ_NO_PREFETCH)); | |
2567 | ||
2568 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2569 | break; | |
428870ff BB |
2570 | } |
2571 | ||
2d9142c9 | 2572 | (void) pthread_rwlock_unlock(&zd->zd_zilog_lock); |
3e31d2b0 | 2573 | |
428870ff BB |
2574 | umem_free(data, blocksize); |
2575 | } | |
2576 | ||
2577 | /* | |
2578 | * Initialize an object description template. | |
2579 | */ | |
2580 | static void | |
2581 | ztest_od_init(ztest_od_t *od, uint64_t id, char *tag, uint64_t index, | |
50c957f7 NB |
2582 | dmu_object_type_t type, uint64_t blocksize, uint64_t dnodesize, |
2583 | uint64_t gen) | |
428870ff BB |
2584 | { |
2585 | od->od_dir = ZTEST_DIROBJ; | |
2586 | od->od_object = 0; | |
2587 | ||
2588 | od->od_crtype = type; | |
2589 | od->od_crblocksize = blocksize ? blocksize : ztest_random_blocksize(); | |
50c957f7 | 2590 | od->od_crdnodesize = dnodesize ? dnodesize : ztest_random_dnodesize(); |
428870ff BB |
2591 | od->od_crgen = gen; |
2592 | ||
2593 | od->od_type = DMU_OT_NONE; | |
2594 | od->od_blocksize = 0; | |
2595 | od->od_gen = 0; | |
2596 | ||
2597 | (void) snprintf(od->od_name, sizeof (od->od_name), "%s(%lld)[%llu]", | |
b8864a23 | 2598 | tag, (longlong_t)id, (u_longlong_t)index); |
428870ff BB |
2599 | } |
2600 | ||
2601 | /* | |
2602 | * Lookup or create the objects for a test using the od template. | |
2603 | * If the objects do not all exist, or if 'remove' is specified, | |
2604 | * remove any existing objects and create new ones. Otherwise, | |
2605 | * use the existing objects. | |
2606 | */ | |
2607 | static int | |
2608 | ztest_object_init(ztest_ds_t *zd, ztest_od_t *od, size_t size, boolean_t remove) | |
2609 | { | |
2610 | int count = size / sizeof (*od); | |
2611 | int rv = 0; | |
2612 | ||
1e33ac1e | 2613 | mutex_enter(&zd->zd_dirobj_lock); |
428870ff BB |
2614 | if ((ztest_lookup(zd, od, count) != 0 || remove) && |
2615 | (ztest_remove(zd, od, count) != 0 || | |
2616 | ztest_create(zd, od, count) != 0)) | |
2617 | rv = -1; | |
2618 | zd->zd_od = od; | |
1e33ac1e | 2619 | mutex_exit(&zd->zd_dirobj_lock); |
428870ff BB |
2620 | |
2621 | return (rv); | |
2622 | } | |
2623 | ||
2624 | /* ARGSUSED */ | |
2625 | void | |
2626 | ztest_zil_commit(ztest_ds_t *zd, uint64_t id) | |
2627 | { | |
2628 | zilog_t *zilog = zd->zd_zilog; | |
2629 | ||
2d9142c9 | 2630 | (void) pthread_rwlock_rdlock(&zd->zd_zilog_lock); |
3e31d2b0 | 2631 | |
572e2857 | 2632 | zil_commit(zilog, ztest_random(ZTEST_OBJECTS)); |
428870ff BB |
2633 | |
2634 | /* | |
2635 | * Remember the committed values in zd, which is in parent/child | |
2636 | * shared memory. If we die, the next iteration of ztest_run() | |
2637 | * will verify that the log really does contain this record. | |
2638 | */ | |
2639 | mutex_enter(&zilog->zl_lock); | |
c242c188 CS |
2640 | ASSERT(zd->zd_shared != NULL); |
2641 | ASSERT3U(zd->zd_shared->zd_seq, <=, zilog->zl_commit_lr_seq); | |
2642 | zd->zd_shared->zd_seq = zilog->zl_commit_lr_seq; | |
428870ff | 2643 | mutex_exit(&zilog->zl_lock); |
3e31d2b0 | 2644 | |
2d9142c9 | 2645 | (void) pthread_rwlock_unlock(&zd->zd_zilog_lock); |
3e31d2b0 ES |
2646 | } |
2647 | ||
2648 | /* | |
2649 | * This function is designed to simulate the operations that occur during a | |
2650 | * mount/unmount operation. We hold the dataset across these operations in an | |
2651 | * attempt to expose any implicit assumptions about ZIL management. | |
2652 | */ | |
2653 | /* ARGSUSED */ | |
2654 | void | |
2655 | ztest_zil_remount(ztest_ds_t *zd, uint64_t id) | |
2656 | { | |
2657 | objset_t *os = zd->zd_os; | |
2658 | ||
0b606cb3 TC |
2659 | /* |
2660 | * We hold the ztest_vdev_lock so we don't cause problems with | |
2661 | * other threads that wish to remove a log device, such as | |
2662 | * ztest_device_removal(). | |
2663 | */ | |
2664 | mutex_enter(&ztest_vdev_lock); | |
2665 | ||
03c6040b GW |
2666 | /* |
2667 | * We grab the zd_dirobj_lock to ensure that no other thread is | |
2668 | * updating the zil (i.e. adding in-memory log records) and the | |
2669 | * zd_zilog_lock to block any I/O. | |
2670 | */ | |
29809a6c | 2671 | mutex_enter(&zd->zd_dirobj_lock); |
2d9142c9 | 2672 | (void) pthread_rwlock_wrlock(&zd->zd_zilog_lock); |
3e31d2b0 | 2673 | |
f298b24d | 2674 | /* zfsvfs_teardown() */ |
3e31d2b0 ES |
2675 | zil_close(zd->zd_zilog); |
2676 | ||
2677 | /* zfsvfs_setup() */ | |
2678 | VERIFY(zil_open(os, ztest_get_data) == zd->zd_zilog); | |
2679 | zil_replay(os, zd, ztest_replay_vector); | |
2680 | ||
2d9142c9 | 2681 | (void) pthread_rwlock_unlock(&zd->zd_zilog_lock); |
29809a6c | 2682 | mutex_exit(&zd->zd_dirobj_lock); |
0b606cb3 | 2683 | mutex_exit(&ztest_vdev_lock); |
428870ff BB |
2684 | } |
2685 | ||
2686 | /* | |
2687 | * Verify that we can't destroy an active pool, create an existing pool, | |
2688 | * or create a pool with a bad vdev spec. | |
2689 | */ | |
2690 | /* ARGSUSED */ | |
2691 | void | |
2692 | ztest_spa_create_destroy(ztest_ds_t *zd, uint64_t id) | |
2693 | { | |
c242c188 | 2694 | ztest_shared_opts_t *zo = &ztest_opts; |
428870ff BB |
2695 | spa_t *spa; |
2696 | nvlist_t *nvroot; | |
2697 | ||
379ca9cf OF |
2698 | if (zo->zo_mmp_test) |
2699 | return; | |
2700 | ||
428870ff BB |
2701 | /* |
2702 | * Attempt to create using a bad file. | |
2703 | */ | |
cc99f275 | 2704 | nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, NULL, 0, 0, 1); |
428870ff | 2705 | VERIFY3U(ENOENT, ==, |
b5256303 | 2706 | spa_create("ztest_bad_file", nvroot, NULL, NULL, NULL)); |
428870ff BB |
2707 | nvlist_free(nvroot); |
2708 | ||
2709 | /* | |
2710 | * Attempt to create using a bad mirror. | |
2711 | */ | |
cc99f275 | 2712 | nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, NULL, 0, 2, 1); |
428870ff | 2713 | VERIFY3U(ENOENT, ==, |
b5256303 | 2714 | spa_create("ztest_bad_mirror", nvroot, NULL, NULL, NULL)); |
428870ff BB |
2715 | nvlist_free(nvroot); |
2716 | ||
2717 | /* | |
2718 | * Attempt to create an existing pool. It shouldn't matter | |
2719 | * what's in the nvroot; we should fail with EEXIST. | |
2720 | */ | |
2d9142c9 | 2721 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
cc99f275 | 2722 | nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, NULL, 0, 0, 1); |
b5256303 TC |
2723 | VERIFY3U(EEXIST, ==, |
2724 | spa_create(zo->zo_pool, nvroot, NULL, NULL, NULL)); | |
428870ff | 2725 | nvlist_free(nvroot); |
43a85362 SD |
2726 | |
2727 | /* | |
2728 | * We open a reference to the spa and then we try to export it | |
2729 | * expecting one of the following errors: | |
2730 | * | |
2731 | * EBUSY | |
2732 | * Because of the reference we just opened. | |
2733 | * | |
2734 | * ZFS_ERR_EXPORT_IN_PROGRESS | |
2735 | * For the case that there is another ztest thread doing | |
2736 | * an export concurrently. | |
2737 | */ | |
c242c188 | 2738 | VERIFY3U(0, ==, spa_open(zo->zo_pool, &spa, FTAG)); |
43a85362 SD |
2739 | int error = spa_destroy(zo->zo_pool); |
2740 | if (error != EBUSY && error != ZFS_ERR_EXPORT_IN_PROGRESS) { | |
2741 | fatal(0, "spa_destroy(%s) returned unexpected value %d", | |
2742 | spa->spa_name, error); | |
2743 | } | |
428870ff BB |
2744 | spa_close(spa, FTAG); |
2745 | ||
2d9142c9 | 2746 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
428870ff BB |
2747 | } |
2748 | ||
0582e403 OF |
2749 | /* |
2750 | * Start and then stop the MMP threads to ensure the startup and shutdown code | |
2751 | * works properly. Actual protection and property-related code tested via ZTS. | |
2752 | */ | |
2753 | /* ARGSUSED */ | |
2754 | void | |
2755 | ztest_mmp_enable_disable(ztest_ds_t *zd, uint64_t id) | |
2756 | { | |
2757 | ztest_shared_opts_t *zo = &ztest_opts; | |
2758 | spa_t *spa = ztest_spa; | |
2759 | ||
2760 | if (zo->zo_mmp_test) | |
2761 | return; | |
2762 | ||
8133679f OF |
2763 | /* |
2764 | * Since enabling MMP involves setting a property, it could not be done | |
2765 | * while the pool is suspended. | |
2766 | */ | |
2767 | if (spa_suspended(spa)) | |
2768 | return; | |
2769 | ||
0582e403 OF |
2770 | spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); |
2771 | mutex_enter(&spa->spa_props_lock); | |
2772 | ||
52002377 OF |
2773 | zfs_multihost_fail_intervals = 0; |
2774 | ||
0582e403 OF |
2775 | if (!spa_multihost(spa)) { |
2776 | spa->spa_multihost = B_TRUE; | |
2777 | mmp_thread_start(spa); | |
2778 | } | |
2779 | ||
2780 | mutex_exit(&spa->spa_props_lock); | |
2781 | spa_config_exit(spa, SCL_CONFIG, FTAG); | |
2782 | ||
2783 | txg_wait_synced(spa_get_dsl(spa), 0); | |
2784 | mmp_signal_all_threads(); | |
2785 | txg_wait_synced(spa_get_dsl(spa), 0); | |
2786 | ||
2787 | spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); | |
2788 | mutex_enter(&spa->spa_props_lock); | |
2789 | ||
2790 | if (spa_multihost(spa)) { | |
2791 | mmp_thread_stop(spa); | |
2792 | spa->spa_multihost = B_FALSE; | |
2793 | } | |
2794 | ||
2795 | mutex_exit(&spa->spa_props_lock); | |
2796 | spa_config_exit(spa, SCL_CONFIG, FTAG); | |
2797 | } | |
2798 | ||
ea0b2538 GW |
2799 | /* ARGSUSED */ |
2800 | void | |
2801 | ztest_spa_upgrade(ztest_ds_t *zd, uint64_t id) | |
2802 | { | |
2803 | spa_t *spa; | |
2804 | uint64_t initial_version = SPA_VERSION_INITIAL; | |
2805 | uint64_t version, newversion; | |
2806 | nvlist_t *nvroot, *props; | |
2807 | char *name; | |
2808 | ||
379ca9cf OF |
2809 | if (ztest_opts.zo_mmp_test) |
2810 | return; | |
2811 | ||
ea0b2538 GW |
2812 | mutex_enter(&ztest_vdev_lock); |
2813 | name = kmem_asprintf("%s_upgrade", ztest_opts.zo_pool); | |
2814 | ||
2815 | /* | |
2816 | * Clean up from previous runs. | |
2817 | */ | |
2818 | (void) spa_destroy(name); | |
2819 | ||
2820 | nvroot = make_vdev_root(NULL, NULL, name, ztest_opts.zo_vdev_size, 0, | |
cc99f275 | 2821 | NULL, ztest_opts.zo_raidz, ztest_opts.zo_mirrors, 1); |
ea0b2538 GW |
2822 | |
2823 | /* | |
2824 | * If we're configuring a RAIDZ device then make sure that the | |
d611989f | 2825 | * initial version is capable of supporting that feature. |
ea0b2538 GW |
2826 | */ |
2827 | switch (ztest_opts.zo_raidz_parity) { | |
2828 | case 0: | |
2829 | case 1: | |
2830 | initial_version = SPA_VERSION_INITIAL; | |
2831 | break; | |
2832 | case 2: | |
2833 | initial_version = SPA_VERSION_RAIDZ2; | |
2834 | break; | |
2835 | case 3: | |
2836 | initial_version = SPA_VERSION_RAIDZ3; | |
2837 | break; | |
2838 | } | |
2839 | ||
2840 | /* | |
2841 | * Create a pool with a spa version that can be upgraded. Pick | |
2842 | * a value between initial_version and SPA_VERSION_BEFORE_FEATURES. | |
2843 | */ | |
2844 | do { | |
2845 | version = ztest_random_spa_version(initial_version); | |
2846 | } while (version > SPA_VERSION_BEFORE_FEATURES); | |
2847 | ||
2848 | props = fnvlist_alloc(); | |
2849 | fnvlist_add_uint64(props, | |
2850 | zpool_prop_to_name(ZPOOL_PROP_VERSION), version); | |
b5256303 | 2851 | VERIFY3S(spa_create(name, nvroot, props, NULL, NULL), ==, 0); |
ea0b2538 GW |
2852 | fnvlist_free(nvroot); |
2853 | fnvlist_free(props); | |
2854 | ||
2855 | VERIFY3S(spa_open(name, &spa, FTAG), ==, 0); | |
2856 | VERIFY3U(spa_version(spa), ==, version); | |
2857 | newversion = ztest_random_spa_version(version + 1); | |
2858 | ||
2859 | if (ztest_opts.zo_verbose >= 4) { | |
2860 | (void) printf("upgrading spa version from %llu to %llu\n", | |
2861 | (u_longlong_t)version, (u_longlong_t)newversion); | |
2862 | } | |
2863 | ||
2864 | spa_upgrade(spa, newversion); | |
2865 | VERIFY3U(spa_version(spa), >, version); | |
2866 | VERIFY3U(spa_version(spa), ==, fnvlist_lookup_uint64(spa->spa_config, | |
2867 | zpool_prop_to_name(ZPOOL_PROP_VERSION))); | |
2868 | spa_close(spa, FTAG); | |
2869 | ||
e4f5fa12 | 2870 | kmem_strfree(name); |
ea0b2538 GW |
2871 | mutex_exit(&ztest_vdev_lock); |
2872 | } | |
2873 | ||
d2734cce SD |
2874 | static void |
2875 | ztest_spa_checkpoint(spa_t *spa) | |
2876 | { | |
2877 | ASSERT(MUTEX_HELD(&ztest_checkpoint_lock)); | |
2878 | ||
2879 | int error = spa_checkpoint(spa->spa_name); | |
2880 | ||
2881 | switch (error) { | |
2882 | case 0: | |
2883 | case ZFS_ERR_DEVRM_IN_PROGRESS: | |
2884 | case ZFS_ERR_DISCARDING_CHECKPOINT: | |
2885 | case ZFS_ERR_CHECKPOINT_EXISTS: | |
2886 | break; | |
2887 | case ENOSPC: | |
2888 | ztest_record_enospc(FTAG); | |
2889 | break; | |
2890 | default: | |
2891 | fatal(0, "spa_checkpoint(%s) = %d", spa->spa_name, error); | |
2892 | } | |
2893 | } | |
2894 | ||
2895 | static void | |
2896 | ztest_spa_discard_checkpoint(spa_t *spa) | |
2897 | { | |
2898 | ASSERT(MUTEX_HELD(&ztest_checkpoint_lock)); | |
2899 | ||
2900 | int error = spa_checkpoint_discard(spa->spa_name); | |
2901 | ||
2902 | switch (error) { | |
2903 | case 0: | |
2904 | case ZFS_ERR_DISCARDING_CHECKPOINT: | |
2905 | case ZFS_ERR_NO_CHECKPOINT: | |
2906 | break; | |
2907 | default: | |
2908 | fatal(0, "spa_discard_checkpoint(%s) = %d", | |
2909 | spa->spa_name, error); | |
2910 | } | |
2911 | ||
2912 | } | |
2913 | ||
2914 | /* ARGSUSED */ | |
2915 | void | |
2916 | ztest_spa_checkpoint_create_discard(ztest_ds_t *zd, uint64_t id) | |
2917 | { | |
2918 | spa_t *spa = ztest_spa; | |
2919 | ||
2920 | mutex_enter(&ztest_checkpoint_lock); | |
2921 | if (ztest_random(2) == 0) { | |
2922 | ztest_spa_checkpoint(spa); | |
2923 | } else { | |
2924 | ztest_spa_discard_checkpoint(spa); | |
2925 | } | |
2926 | mutex_exit(&ztest_checkpoint_lock); | |
2927 | } | |
2928 | ||
2929 | ||
428870ff BB |
2930 | static vdev_t * |
2931 | vdev_lookup_by_path(vdev_t *vd, const char *path) | |
2932 | { | |
2933 | vdev_t *mvd; | |
d6320ddb | 2934 | int c; |
428870ff BB |
2935 | |
2936 | if (vd->vdev_path != NULL && strcmp(path, vd->vdev_path) == 0) | |
2937 | return (vd); | |
2938 | ||
d6320ddb | 2939 | for (c = 0; c < vd->vdev_children; c++) |
428870ff BB |
2940 | if ((mvd = vdev_lookup_by_path(vd->vdev_child[c], path)) != |
2941 | NULL) | |
2942 | return (mvd); | |
2943 | ||
2944 | return (NULL); | |
2945 | } | |
2946 | ||
93e28d66 SD |
2947 | static int |
2948 | spa_num_top_vdevs(spa_t *spa) | |
428870ff BB |
2949 | { |
2950 | vdev_t *rvd = spa->spa_root_vdev; | |
93e28d66 SD |
2951 | ASSERT3U(spa_config_held(spa, SCL_VDEV, RW_READER), ==, SCL_VDEV); |
2952 | return (rvd->vdev_children); | |
428870ff BB |
2953 | } |
2954 | ||
2955 | /* | |
2956 | * Verify that vdev_add() works as expected. | |
2957 | */ | |
2958 | /* ARGSUSED */ | |
2959 | void | |
2960 | ztest_vdev_add_remove(ztest_ds_t *zd, uint64_t id) | |
2961 | { | |
2962 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 2963 | spa_t *spa = ztest_spa; |
428870ff BB |
2964 | uint64_t leaves; |
2965 | uint64_t guid; | |
2966 | nvlist_t *nvroot; | |
2967 | int error; | |
2968 | ||
379ca9cf OF |
2969 | if (ztest_opts.zo_mmp_test) |
2970 | return; | |
2971 | ||
c242c188 | 2972 | mutex_enter(&ztest_vdev_lock); |
13fe0198 | 2973 | leaves = MAX(zs->zs_mirrors + zs->zs_splits, 1) * ztest_opts.zo_raidz; |
428870ff BB |
2974 | |
2975 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
2976 | ||
93e28d66 | 2977 | ztest_shared->zs_vdev_next_leaf = spa_num_top_vdevs(spa) * leaves; |
428870ff BB |
2978 | |
2979 | /* | |
2980 | * If we have slogs then remove them 1/4 of the time. | |
2981 | */ | |
2982 | if (spa_has_slogs(spa) && ztest_random(4) == 0) { | |
cc99f275 DB |
2983 | metaslab_group_t *mg; |
2984 | ||
428870ff | 2985 | /* |
cc99f275 | 2986 | * find the first real slog in log allocation class |
428870ff | 2987 | */ |
cc99f275 DB |
2988 | mg = spa_log_class(spa)->mc_rotor; |
2989 | while (!mg->mg_vd->vdev_islog) | |
2990 | mg = mg->mg_next; | |
2991 | ||
2992 | guid = mg->mg_vd->vdev_guid; | |
428870ff BB |
2993 | |
2994 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
2995 | ||
2996 | /* | |
2997 | * We have to grab the zs_name_lock as writer to | |
2998 | * prevent a race between removing a slog (dmu_objset_find) | |
2999 | * and destroying a dataset. Removing the slog will | |
3000 | * grab a reference on the dataset which may cause | |
13fe0198 | 3001 | * dsl_destroy_head() to fail with EBUSY thus |
428870ff BB |
3002 | * leaving the dataset in an inconsistent state. |
3003 | */ | |
2d9142c9 | 3004 | pthread_rwlock_wrlock(&ztest_name_lock); |
428870ff | 3005 | error = spa_vdev_remove(spa, guid, B_FALSE); |
2d9142c9 | 3006 | pthread_rwlock_unlock(&ztest_name_lock); |
428870ff | 3007 | |
c91cf36f BB |
3008 | switch (error) { |
3009 | case 0: | |
3010 | case EEXIST: /* Generic zil_reset() error */ | |
3011 | case EBUSY: /* Replay required */ | |
3012 | case EACCES: /* Crypto key not loaded */ | |
d2734cce SD |
3013 | case ZFS_ERR_CHECKPOINT_EXISTS: |
3014 | case ZFS_ERR_DISCARDING_CHECKPOINT: | |
c91cf36f BB |
3015 | break; |
3016 | default: | |
428870ff | 3017 | fatal(0, "spa_vdev_remove() = %d", error); |
c91cf36f | 3018 | } |
428870ff BB |
3019 | } else { |
3020 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
3021 | ||
3022 | /* | |
cc99f275 | 3023 | * Make 1/4 of the devices be log devices |
428870ff | 3024 | */ |
ea0b2538 | 3025 | nvroot = make_vdev_root(NULL, NULL, NULL, |
cc99f275 DB |
3026 | ztest_opts.zo_vdev_size, 0, (ztest_random(4) == 0) ? |
3027 | "log" : NULL, ztest_opts.zo_raidz, zs->zs_mirrors, 1); | |
428870ff BB |
3028 | |
3029 | error = spa_vdev_add(spa, nvroot); | |
3030 | nvlist_free(nvroot); | |
3031 | ||
d2734cce SD |
3032 | switch (error) { |
3033 | case 0: | |
3034 | break; | |
3035 | case ENOSPC: | |
428870ff | 3036 | ztest_record_enospc("spa_vdev_add"); |
d2734cce SD |
3037 | break; |
3038 | default: | |
428870ff | 3039 | fatal(0, "spa_vdev_add() = %d", error); |
d2734cce | 3040 | } |
428870ff BB |
3041 | } |
3042 | ||
c242c188 | 3043 | mutex_exit(&ztest_vdev_lock); |
428870ff BB |
3044 | } |
3045 | ||
cc99f275 DB |
3046 | /* ARGSUSED */ |
3047 | void | |
3048 | ztest_vdev_class_add(ztest_ds_t *zd, uint64_t id) | |
3049 | { | |
3050 | ztest_shared_t *zs = ztest_shared; | |
3051 | spa_t *spa = ztest_spa; | |
3052 | uint64_t leaves; | |
3053 | nvlist_t *nvroot; | |
3054 | const char *class = (ztest_random(2) == 0) ? | |
3055 | VDEV_ALLOC_BIAS_SPECIAL : VDEV_ALLOC_BIAS_DEDUP; | |
3056 | int error; | |
3057 | ||
3058 | /* | |
3059 | * By default add a special vdev 50% of the time | |
3060 | */ | |
3061 | if ((ztest_opts.zo_special_vdevs == ZTEST_VDEV_CLASS_OFF) || | |
3062 | (ztest_opts.zo_special_vdevs == ZTEST_VDEV_CLASS_RND && | |
3063 | ztest_random(2) == 0)) { | |
3064 | return; | |
3065 | } | |
3066 | ||
3067 | mutex_enter(&ztest_vdev_lock); | |
3068 | ||
3069 | /* Only test with mirrors */ | |
3070 | if (zs->zs_mirrors < 2) { | |
3071 | mutex_exit(&ztest_vdev_lock); | |
3072 | return; | |
3073 | } | |
3074 | ||
3075 | /* requires feature@allocation_classes */ | |
3076 | if (!spa_feature_is_enabled(spa, SPA_FEATURE_ALLOCATION_CLASSES)) { | |
3077 | mutex_exit(&ztest_vdev_lock); | |
3078 | return; | |
3079 | } | |
3080 | ||
3081 | leaves = MAX(zs->zs_mirrors + zs->zs_splits, 1) * ztest_opts.zo_raidz; | |
3082 | ||
3083 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
93e28d66 | 3084 | ztest_shared->zs_vdev_next_leaf = spa_num_top_vdevs(spa) * leaves; |
cc99f275 DB |
3085 | spa_config_exit(spa, SCL_VDEV, FTAG); |
3086 | ||
3087 | nvroot = make_vdev_root(NULL, NULL, NULL, ztest_opts.zo_vdev_size, 0, | |
3088 | class, ztest_opts.zo_raidz, zs->zs_mirrors, 1); | |
3089 | ||
3090 | error = spa_vdev_add(spa, nvroot); | |
3091 | nvlist_free(nvroot); | |
3092 | ||
3093 | if (error == ENOSPC) | |
3094 | ztest_record_enospc("spa_vdev_add"); | |
3095 | else if (error != 0) | |
3096 | fatal(0, "spa_vdev_add() = %d", error); | |
3097 | ||
3098 | /* | |
3099 | * 50% of the time allow small blocks in the special class | |
3100 | */ | |
3101 | if (error == 0 && | |
3102 | spa_special_class(spa)->mc_groups == 1 && ztest_random(2) == 0) { | |
3103 | if (ztest_opts.zo_verbose >= 3) | |
3104 | (void) printf("Enabling special VDEV small blocks\n"); | |
3105 | (void) ztest_dsl_prop_set_uint64(zd->zd_name, | |
3106 | ZFS_PROP_SPECIAL_SMALL_BLOCKS, 32768, B_FALSE); | |
3107 | } | |
3108 | ||
3109 | mutex_exit(&ztest_vdev_lock); | |
3110 | ||
3111 | if (ztest_opts.zo_verbose >= 3) { | |
3112 | metaslab_class_t *mc; | |
3113 | ||
3114 | if (strcmp(class, VDEV_ALLOC_BIAS_SPECIAL) == 0) | |
3115 | mc = spa_special_class(spa); | |
3116 | else | |
3117 | mc = spa_dedup_class(spa); | |
3118 | (void) printf("Added a %s mirrored vdev (of %d)\n", | |
3119 | class, (int)mc->mc_groups); | |
3120 | } | |
3121 | } | |
3122 | ||
428870ff BB |
3123 | /* |
3124 | * Verify that adding/removing aux devices (l2arc, hot spare) works as expected. | |
3125 | */ | |
3126 | /* ARGSUSED */ | |
3127 | void | |
3128 | ztest_vdev_aux_add_remove(ztest_ds_t *zd, uint64_t id) | |
3129 | { | |
3130 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 3131 | spa_t *spa = ztest_spa; |
428870ff BB |
3132 | vdev_t *rvd = spa->spa_root_vdev; |
3133 | spa_aux_vdev_t *sav; | |
3134 | char *aux; | |
40b84e7a | 3135 | char *path; |
428870ff BB |
3136 | uint64_t guid = 0; |
3137 | int error; | |
3138 | ||
379ca9cf OF |
3139 | if (ztest_opts.zo_mmp_test) |
3140 | return; | |
3141 | ||
40b84e7a BB |
3142 | path = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
3143 | ||
428870ff BB |
3144 | if (ztest_random(2) == 0) { |
3145 | sav = &spa->spa_spares; | |
3146 | aux = ZPOOL_CONFIG_SPARES; | |
3147 | } else { | |
3148 | sav = &spa->spa_l2cache; | |
3149 | aux = ZPOOL_CONFIG_L2CACHE; | |
3150 | } | |
3151 | ||
c242c188 | 3152 | mutex_enter(&ztest_vdev_lock); |
428870ff BB |
3153 | |
3154 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
3155 | ||
3156 | if (sav->sav_count != 0 && ztest_random(4) == 0) { | |
b128c09f BB |
3157 | /* |
3158 | * Pick a random device to remove. | |
3159 | */ | |
3160 | guid = sav->sav_vdevs[ztest_random(sav->sav_count)]->vdev_guid; | |
3161 | } else { | |
3162 | /* | |
3163 | * Find an unused device we can add. | |
3164 | */ | |
428870ff | 3165 | zs->zs_vdev_aux = 0; |
b128c09f | 3166 | for (;;) { |
b128c09f | 3167 | int c; |
3db3ff4a | 3168 | (void) snprintf(path, MAXPATHLEN, ztest_aux_template, |
c242c188 CS |
3169 | ztest_opts.zo_dir, ztest_opts.zo_pool, aux, |
3170 | zs->zs_vdev_aux); | |
b128c09f BB |
3171 | for (c = 0; c < sav->sav_count; c++) |
3172 | if (strcmp(sav->sav_vdevs[c]->vdev_path, | |
3173 | path) == 0) | |
3174 | break; | |
3175 | if (c == sav->sav_count && | |
3176 | vdev_lookup_by_path(rvd, path) == NULL) | |
3177 | break; | |
428870ff | 3178 | zs->zs_vdev_aux++; |
34dc7c2f BB |
3179 | } |
3180 | } | |
3181 | ||
b128c09f | 3182 | spa_config_exit(spa, SCL_VDEV, FTAG); |
34dc7c2f | 3183 | |
b128c09f BB |
3184 | if (guid == 0) { |
3185 | /* | |
3186 | * Add a new device. | |
3187 | */ | |
ea0b2538 | 3188 | nvlist_t *nvroot = make_vdev_root(NULL, aux, NULL, |
cc99f275 | 3189 | (ztest_opts.zo_vdev_size * 5) / 4, 0, NULL, 0, 0, 1); |
b128c09f | 3190 | error = spa_vdev_add(spa, nvroot); |
d2734cce SD |
3191 | |
3192 | switch (error) { | |
3193 | case 0: | |
3194 | break; | |
3195 | default: | |
b128c09f | 3196 | fatal(0, "spa_vdev_add(%p) = %d", nvroot, error); |
d2734cce | 3197 | } |
b128c09f BB |
3198 | nvlist_free(nvroot); |
3199 | } else { | |
3200 | /* | |
3201 | * Remove an existing device. Sometimes, dirty its | |
3202 | * vdev state first to make sure we handle removal | |
3203 | * of devices that have pending state changes. | |
3204 | */ | |
3205 | if (ztest_random(2) == 0) | |
9babb374 | 3206 | (void) vdev_online(spa, guid, 0, NULL); |
b128c09f BB |
3207 | |
3208 | error = spa_vdev_remove(spa, guid, B_FALSE); | |
d2734cce SD |
3209 | |
3210 | switch (error) { | |
3211 | case 0: | |
3212 | case EBUSY: | |
3213 | case ZFS_ERR_CHECKPOINT_EXISTS: | |
3214 | case ZFS_ERR_DISCARDING_CHECKPOINT: | |
3215 | break; | |
3216 | default: | |
b128c09f | 3217 | fatal(0, "spa_vdev_remove(%llu) = %d", guid, error); |
d2734cce | 3218 | } |
b128c09f BB |
3219 | } |
3220 | ||
c242c188 | 3221 | mutex_exit(&ztest_vdev_lock); |
40b84e7a BB |
3222 | |
3223 | umem_free(path, MAXPATHLEN); | |
428870ff BB |
3224 | } |
3225 | ||
3226 | /* | |
3227 | * split a pool if it has mirror tlvdevs | |
3228 | */ | |
3229 | /* ARGSUSED */ | |
3230 | void | |
3231 | ztest_split_pool(ztest_ds_t *zd, uint64_t id) | |
3232 | { | |
3233 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 3234 | spa_t *spa = ztest_spa; |
428870ff BB |
3235 | vdev_t *rvd = spa->spa_root_vdev; |
3236 | nvlist_t *tree, **child, *config, *split, **schild; | |
3237 | uint_t c, children, schildren = 0, lastlogid = 0; | |
3238 | int error = 0; | |
3239 | ||
379ca9cf OF |
3240 | if (ztest_opts.zo_mmp_test) |
3241 | return; | |
3242 | ||
c242c188 | 3243 | mutex_enter(&ztest_vdev_lock); |
428870ff | 3244 | |
d611989f | 3245 | /* ensure we have a usable config; mirrors of raidz aren't supported */ |
c242c188 CS |
3246 | if (zs->zs_mirrors < 3 || ztest_opts.zo_raidz > 1) { |
3247 | mutex_exit(&ztest_vdev_lock); | |
428870ff BB |
3248 | return; |
3249 | } | |
3250 | ||
3251 | /* clean up the old pool, if any */ | |
3252 | (void) spa_destroy("splitp"); | |
3253 | ||
3254 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
3255 | ||
3256 | /* generate a config from the existing config */ | |
3257 | mutex_enter(&spa->spa_props_lock); | |
3258 | VERIFY(nvlist_lookup_nvlist(spa->spa_config, ZPOOL_CONFIG_VDEV_TREE, | |
3259 | &tree) == 0); | |
3260 | mutex_exit(&spa->spa_props_lock); | |
3261 | ||
3262 | VERIFY(nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child, | |
3263 | &children) == 0); | |
3264 | ||
3265 | schild = malloc(rvd->vdev_children * sizeof (nvlist_t *)); | |
3266 | for (c = 0; c < children; c++) { | |
3267 | vdev_t *tvd = rvd->vdev_child[c]; | |
3268 | nvlist_t **mchild; | |
3269 | uint_t mchildren; | |
3270 | ||
3271 | if (tvd->vdev_islog || tvd->vdev_ops == &vdev_hole_ops) { | |
3272 | VERIFY(nvlist_alloc(&schild[schildren], NV_UNIQUE_NAME, | |
3273 | 0) == 0); | |
3274 | VERIFY(nvlist_add_string(schild[schildren], | |
3275 | ZPOOL_CONFIG_TYPE, VDEV_TYPE_HOLE) == 0); | |
3276 | VERIFY(nvlist_add_uint64(schild[schildren], | |
3277 | ZPOOL_CONFIG_IS_HOLE, 1) == 0); | |
3278 | if (lastlogid == 0) | |
3279 | lastlogid = schildren; | |
3280 | ++schildren; | |
3281 | continue; | |
3282 | } | |
3283 | lastlogid = 0; | |
3284 | VERIFY(nvlist_lookup_nvlist_array(child[c], | |
3285 | ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0); | |
3286 | VERIFY(nvlist_dup(mchild[0], &schild[schildren++], 0) == 0); | |
3287 | } | |
3288 | ||
3289 | /* OK, create a config that can be used to split */ | |
3290 | VERIFY(nvlist_alloc(&split, NV_UNIQUE_NAME, 0) == 0); | |
3291 | VERIFY(nvlist_add_string(split, ZPOOL_CONFIG_TYPE, | |
3292 | VDEV_TYPE_ROOT) == 0); | |
3293 | VERIFY(nvlist_add_nvlist_array(split, ZPOOL_CONFIG_CHILDREN, schild, | |
3294 | lastlogid != 0 ? lastlogid : schildren) == 0); | |
3295 | ||
3296 | VERIFY(nvlist_alloc(&config, NV_UNIQUE_NAME, 0) == 0); | |
3297 | VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, split) == 0); | |
3298 | ||
3299 | for (c = 0; c < schildren; c++) | |
3300 | nvlist_free(schild[c]); | |
3301 | free(schild); | |
3302 | nvlist_free(split); | |
3303 | ||
3304 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
3305 | ||
2d9142c9 | 3306 | (void) pthread_rwlock_wrlock(&ztest_name_lock); |
428870ff | 3307 | error = spa_vdev_split_mirror(spa, "splitp", config, NULL, B_FALSE); |
2d9142c9 | 3308 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
428870ff BB |
3309 | |
3310 | nvlist_free(config); | |
3311 | ||
3312 | if (error == 0) { | |
3313 | (void) printf("successful split - results:\n"); | |
3314 | mutex_enter(&spa_namespace_lock); | |
3315 | show_pool_stats(spa); | |
3316 | show_pool_stats(spa_lookup("splitp")); | |
3317 | mutex_exit(&spa_namespace_lock); | |
3318 | ++zs->zs_splits; | |
3319 | --zs->zs_mirrors; | |
3320 | } | |
c242c188 | 3321 | mutex_exit(&ztest_vdev_lock); |
34dc7c2f BB |
3322 | } |
3323 | ||
3324 | /* | |
3325 | * Verify that we can attach and detach devices. | |
3326 | */ | |
428870ff | 3327 | /* ARGSUSED */ |
34dc7c2f | 3328 | void |
428870ff | 3329 | ztest_vdev_attach_detach(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3330 | { |
428870ff | 3331 | ztest_shared_t *zs = ztest_shared; |
c242c188 | 3332 | spa_t *spa = ztest_spa; |
b128c09f | 3333 | spa_aux_vdev_t *sav = &spa->spa_spares; |
34dc7c2f BB |
3334 | vdev_t *rvd = spa->spa_root_vdev; |
3335 | vdev_t *oldvd, *newvd, *pvd; | |
b128c09f | 3336 | nvlist_t *root; |
428870ff | 3337 | uint64_t leaves; |
34dc7c2f BB |
3338 | uint64_t leaf, top; |
3339 | uint64_t ashift = ztest_get_ashift(); | |
fb5f0bc8 | 3340 | uint64_t oldguid, pguid; |
5d1f7fb6 | 3341 | uint64_t oldsize, newsize; |
40b84e7a | 3342 | char *oldpath, *newpath; |
34dc7c2f | 3343 | int replacing; |
b128c09f BB |
3344 | int oldvd_has_siblings = B_FALSE; |
3345 | int newvd_is_spare = B_FALSE; | |
3346 | int oldvd_is_log; | |
34dc7c2f | 3347 | int error, expected_error; |
34dc7c2f | 3348 | |
379ca9cf OF |
3349 | if (ztest_opts.zo_mmp_test) |
3350 | return; | |
3351 | ||
40b84e7a BB |
3352 | oldpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
3353 | newpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
3354 | ||
c242c188 CS |
3355 | mutex_enter(&ztest_vdev_lock); |
3356 | leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raidz; | |
34dc7c2f | 3357 | |
a1d477c2 MA |
3358 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); |
3359 | ||
3360 | /* | |
3361 | * If a vdev is in the process of being removed, its removal may | |
3362 | * finish while we are in progress, leading to an unexpected error | |
3363 | * value. Don't bother trying to attach while we are in the middle | |
3364 | * of removal. | |
3365 | */ | |
9e052db4 | 3366 | if (ztest_device_removal_active) { |
a1d477c2 MA |
3367 | spa_config_exit(spa, SCL_ALL, FTAG); |
3368 | mutex_exit(&ztest_vdev_lock); | |
3369 | return; | |
3370 | } | |
34dc7c2f BB |
3371 | |
3372 | /* | |
3373 | * Decide whether to do an attach or a replace. | |
3374 | */ | |
3375 | replacing = ztest_random(2); | |
3376 | ||
3377 | /* | |
3378 | * Pick a random top-level vdev. | |
3379 | */ | |
428870ff | 3380 | top = ztest_random_vdev_top(spa, B_TRUE); |
34dc7c2f BB |
3381 | |
3382 | /* | |
3383 | * Pick a random leaf within it. | |
3384 | */ | |
3385 | leaf = ztest_random(leaves); | |
3386 | ||
3387 | /* | |
b128c09f | 3388 | * Locate this vdev. |
34dc7c2f | 3389 | */ |
b128c09f | 3390 | oldvd = rvd->vdev_child[top]; |
cc99f275 DB |
3391 | |
3392 | /* pick a child from the mirror */ | |
428870ff | 3393 | if (zs->zs_mirrors >= 1) { |
fb5f0bc8 | 3394 | ASSERT(oldvd->vdev_ops == &vdev_mirror_ops); |
428870ff | 3395 | ASSERT(oldvd->vdev_children >= zs->zs_mirrors); |
c242c188 | 3396 | oldvd = oldvd->vdev_child[leaf / ztest_opts.zo_raidz]; |
fb5f0bc8 | 3397 | } |
cc99f275 DB |
3398 | |
3399 | /* pick a child out of the raidz group */ | |
c242c188 | 3400 | if (ztest_opts.zo_raidz > 1) { |
fb5f0bc8 | 3401 | ASSERT(oldvd->vdev_ops == &vdev_raidz_ops); |
c242c188 CS |
3402 | ASSERT(oldvd->vdev_children == ztest_opts.zo_raidz); |
3403 | oldvd = oldvd->vdev_child[leaf % ztest_opts.zo_raidz]; | |
fb5f0bc8 | 3404 | } |
34dc7c2f BB |
3405 | |
3406 | /* | |
b128c09f BB |
3407 | * If we're already doing an attach or replace, oldvd may be a |
3408 | * mirror vdev -- in which case, pick a random child. | |
34dc7c2f | 3409 | */ |
b128c09f BB |
3410 | while (oldvd->vdev_children != 0) { |
3411 | oldvd_has_siblings = B_TRUE; | |
fb5f0bc8 BB |
3412 | ASSERT(oldvd->vdev_children >= 2); |
3413 | oldvd = oldvd->vdev_child[ztest_random(oldvd->vdev_children)]; | |
b128c09f BB |
3414 | } |
3415 | ||
3416 | oldguid = oldvd->vdev_guid; | |
9babb374 | 3417 | oldsize = vdev_get_min_asize(oldvd); |
b128c09f BB |
3418 | oldvd_is_log = oldvd->vdev_top->vdev_islog; |
3419 | (void) strcpy(oldpath, oldvd->vdev_path); | |
3420 | pvd = oldvd->vdev_parent; | |
fb5f0bc8 | 3421 | pguid = pvd->vdev_guid; |
34dc7c2f BB |
3422 | |
3423 | /* | |
ce5fb2a7 BB |
3424 | * If oldvd has siblings, then half of the time, detach it. Prior |
3425 | * to the detach the pool is scrubbed in order to prevent creating | |
3426 | * unrepairable blocks as a result of the data corruption injection. | |
34dc7c2f | 3427 | */ |
b128c09f | 3428 | if (oldvd_has_siblings && ztest_random(2) == 0) { |
a1d477c2 | 3429 | spa_config_exit(spa, SCL_ALL, FTAG); |
ce5fb2a7 BB |
3430 | |
3431 | error = ztest_scrub_impl(spa); | |
3432 | if (error) | |
3433 | goto out; | |
3434 | ||
fb5f0bc8 BB |
3435 | error = spa_vdev_detach(spa, oldguid, pguid, B_FALSE); |
3436 | if (error != 0 && error != ENODEV && error != EBUSY && | |
d2734cce SD |
3437 | error != ENOTSUP && error != ZFS_ERR_CHECKPOINT_EXISTS && |
3438 | error != ZFS_ERR_DISCARDING_CHECKPOINT) | |
fb5f0bc8 | 3439 | fatal(0, "detach (%s) returned %d", oldpath, error); |
40b84e7a | 3440 | goto out; |
b128c09f | 3441 | } |
34dc7c2f BB |
3442 | |
3443 | /* | |
b128c09f BB |
3444 | * For the new vdev, choose with equal probability between the two |
3445 | * standard paths (ending in either 'a' or 'b') or a random hot spare. | |
34dc7c2f | 3446 | */ |
b128c09f BB |
3447 | if (sav->sav_count != 0 && ztest_random(3) == 0) { |
3448 | newvd = sav->sav_vdevs[ztest_random(sav->sav_count)]; | |
3449 | newvd_is_spare = B_TRUE; | |
3450 | (void) strcpy(newpath, newvd->vdev_path); | |
3451 | } else { | |
6aec1cd5 | 3452 | (void) snprintf(newpath, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
3453 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
3454 | top * leaves + leaf); | |
b128c09f BB |
3455 | if (ztest_random(2) == 0) |
3456 | newpath[strlen(newpath) - 1] = 'b'; | |
3457 | newvd = vdev_lookup_by_path(rvd, newpath); | |
3458 | } | |
3459 | ||
3460 | if (newvd) { | |
a1d477c2 MA |
3461 | /* |
3462 | * Reopen to ensure the vdev's asize field isn't stale. | |
3463 | */ | |
3464 | vdev_reopen(newvd); | |
9babb374 | 3465 | newsize = vdev_get_min_asize(newvd); |
b128c09f BB |
3466 | } else { |
3467 | /* | |
3468 | * Make newsize a little bigger or smaller than oldsize. | |
3469 | * If it's smaller, the attach should fail. | |
3470 | * If it's larger, and we're doing a replace, | |
3471 | * we should get dynamic LUN growth when we're done. | |
3472 | */ | |
3473 | newsize = 10 * oldsize / (9 + ztest_random(3)); | |
3474 | } | |
34dc7c2f BB |
3475 | |
3476 | /* | |
3477 | * If pvd is not a mirror or root, the attach should fail with ENOTSUP, | |
3478 | * unless it's a replace; in that case any non-replacing parent is OK. | |
3479 | * | |
3480 | * If newvd is already part of the pool, it should fail with EBUSY. | |
3481 | * | |
3482 | * If newvd is too small, it should fail with EOVERFLOW. | |
3483 | */ | |
b128c09f BB |
3484 | if (pvd->vdev_ops != &vdev_mirror_ops && |
3485 | pvd->vdev_ops != &vdev_root_ops && (!replacing || | |
3486 | pvd->vdev_ops == &vdev_replacing_ops || | |
3487 | pvd->vdev_ops == &vdev_spare_ops)) | |
34dc7c2f | 3488 | expected_error = ENOTSUP; |
b128c09f BB |
3489 | else if (newvd_is_spare && (!replacing || oldvd_is_log)) |
3490 | expected_error = ENOTSUP; | |
3491 | else if (newvd == oldvd) | |
3492 | expected_error = replacing ? 0 : EBUSY; | |
3493 | else if (vdev_lookup_by_path(rvd, newpath) != NULL) | |
3494 | expected_error = EBUSY; | |
34dc7c2f BB |
3495 | else if (newsize < oldsize) |
3496 | expected_error = EOVERFLOW; | |
3497 | else if (ashift > oldvd->vdev_top->vdev_ashift) | |
3498 | expected_error = EDOM; | |
3499 | else | |
3500 | expected_error = 0; | |
3501 | ||
a1d477c2 | 3502 | spa_config_exit(spa, SCL_ALL, FTAG); |
34dc7c2f BB |
3503 | |
3504 | /* | |
3505 | * Build the nvlist describing newpath. | |
3506 | */ | |
ea0b2538 | 3507 | root = make_vdev_root(newpath, NULL, NULL, newvd == NULL ? newsize : 0, |
cc99f275 | 3508 | ashift, NULL, 0, 0, 1); |
34dc7c2f | 3509 | |
9a49d3f3 BB |
3510 | /* |
3511 | * When supported select either a healing or sequential resilver. | |
3512 | */ | |
3513 | boolean_t rebuilding = B_FALSE; | |
3514 | if (pvd->vdev_ops == &vdev_mirror_ops || | |
3515 | pvd->vdev_ops == &vdev_root_ops) { | |
3516 | rebuilding = !!ztest_random(2); | |
3517 | } | |
3518 | ||
3519 | error = spa_vdev_attach(spa, oldguid, root, replacing, rebuilding); | |
34dc7c2f | 3520 | |
34dc7c2f BB |
3521 | nvlist_free(root); |
3522 | ||
3523 | /* | |
3524 | * If our parent was the replacing vdev, but the replace completed, | |
3525 | * then instead of failing with ENOTSUP we may either succeed, | |
3526 | * fail with ENODEV, or fail with EOVERFLOW. | |
3527 | */ | |
3528 | if (expected_error == ENOTSUP && | |
3529 | (error == 0 || error == ENODEV || error == EOVERFLOW)) | |
3530 | expected_error = error; | |
3531 | ||
3532 | /* | |
3533 | * If someone grew the LUN, the replacement may be too small. | |
3534 | */ | |
b128c09f | 3535 | if (error == EOVERFLOW || error == EBUSY) |
34dc7c2f BB |
3536 | expected_error = error; |
3537 | ||
d2734cce | 3538 | if (error == ZFS_ERR_CHECKPOINT_EXISTS || |
9a49d3f3 BB |
3539 | error == ZFS_ERR_DISCARDING_CHECKPOINT || |
3540 | error == ZFS_ERR_RESILVER_IN_PROGRESS || | |
3541 | error == ZFS_ERR_REBUILD_IN_PROGRESS) | |
d2734cce SD |
3542 | expected_error = error; |
3543 | ||
b128c09f BB |
3544 | if (error != expected_error && expected_error != EBUSY) { |
3545 | fatal(0, "attach (%s %llu, %s %llu, %d) " | |
3546 | "returned %d, expected %d", | |
5d1f7fb6 GW |
3547 | oldpath, oldsize, newpath, |
3548 | newsize, replacing, error, expected_error); | |
34dc7c2f | 3549 | } |
40b84e7a | 3550 | out: |
c242c188 | 3551 | mutex_exit(&ztest_vdev_lock); |
40b84e7a BB |
3552 | |
3553 | umem_free(oldpath, MAXPATHLEN); | |
3554 | umem_free(newpath, MAXPATHLEN); | |
34dc7c2f BB |
3555 | } |
3556 | ||
a1d477c2 MA |
3557 | /* ARGSUSED */ |
3558 | void | |
3559 | ztest_device_removal(ztest_ds_t *zd, uint64_t id) | |
3560 | { | |
3561 | spa_t *spa = ztest_spa; | |
3562 | vdev_t *vd; | |
3563 | uint64_t guid; | |
9e052db4 | 3564 | int error; |
a1d477c2 MA |
3565 | |
3566 | mutex_enter(&ztest_vdev_lock); | |
3567 | ||
9e052db4 MA |
3568 | if (ztest_device_removal_active) { |
3569 | mutex_exit(&ztest_vdev_lock); | |
3570 | return; | |
3571 | } | |
3572 | ||
3573 | /* | |
3574 | * Remove a random top-level vdev and wait for removal to finish. | |
3575 | */ | |
a1d477c2 MA |
3576 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); |
3577 | vd = vdev_lookup_top(spa, ztest_random_vdev_top(spa, B_FALSE)); | |
3578 | guid = vd->vdev_guid; | |
3579 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
3580 | ||
9e052db4 MA |
3581 | error = spa_vdev_remove(spa, guid, B_FALSE); |
3582 | if (error == 0) { | |
3583 | ztest_device_removal_active = B_TRUE; | |
3584 | mutex_exit(&ztest_vdev_lock); | |
3585 | ||
7d658d29 TC |
3586 | /* |
3587 | * spa->spa_vdev_removal is created in a sync task that | |
3588 | * is initiated via dsl_sync_task_nowait(). Since the | |
3589 | * task may not run before spa_vdev_remove() returns, we | |
3590 | * must wait at least 1 txg to ensure that the removal | |
3591 | * struct has been created. | |
3592 | */ | |
3593 | txg_wait_synced(spa_get_dsl(spa), 0); | |
3594 | ||
c40a1124 | 3595 | while (spa->spa_removing_phys.sr_state == DSS_SCANNING) |
9e052db4 MA |
3596 | txg_wait_synced(spa_get_dsl(spa), 0); |
3597 | } else { | |
3598 | mutex_exit(&ztest_vdev_lock); | |
3599 | return; | |
3600 | } | |
a1d477c2 | 3601 | |
9e052db4 MA |
3602 | /* |
3603 | * The pool needs to be scrubbed after completing device removal. | |
3604 | * Failure to do so may result in checksum errors due to the | |
3605 | * strategy employed by ztest_fault_inject() when selecting which | |
3606 | * offset are redundant and can be damaged. | |
3607 | */ | |
3608 | error = spa_scan(spa, POOL_SCAN_SCRUB); | |
3609 | if (error == 0) { | |
3610 | while (dsl_scan_scrubbing(spa_get_dsl(spa))) | |
3611 | txg_wait_synced(spa_get_dsl(spa), 0); | |
3612 | } | |
3613 | ||
3614 | mutex_enter(&ztest_vdev_lock); | |
3615 | ztest_device_removal_active = B_FALSE; | |
a1d477c2 MA |
3616 | mutex_exit(&ztest_vdev_lock); |
3617 | } | |
3618 | ||
9babb374 BB |
3619 | /* |
3620 | * Callback function which expands the physical size of the vdev. | |
3621 | */ | |
65c7cc49 | 3622 | static vdev_t * |
9babb374 BB |
3623 | grow_vdev(vdev_t *vd, void *arg) |
3624 | { | |
2a8ba608 | 3625 | spa_t *spa __maybe_unused = vd->vdev_spa; |
9babb374 BB |
3626 | size_t *newsize = arg; |
3627 | size_t fsize; | |
3628 | int fd; | |
3629 | ||
3630 | ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE); | |
3631 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
3632 | ||
3633 | if ((fd = open(vd->vdev_path, O_RDWR)) == -1) | |
3634 | return (vd); | |
3635 | ||
3636 | fsize = lseek(fd, 0, SEEK_END); | |
0e5b68e0 | 3637 | VERIFY(ftruncate(fd, *newsize) == 0); |
9babb374 | 3638 | |
c242c188 | 3639 | if (ztest_opts.zo_verbose >= 6) { |
9babb374 BB |
3640 | (void) printf("%s grew from %lu to %lu bytes\n", |
3641 | vd->vdev_path, (ulong_t)fsize, (ulong_t)*newsize); | |
3642 | } | |
3643 | (void) close(fd); | |
3644 | return (NULL); | |
3645 | } | |
3646 | ||
3647 | /* | |
3648 | * Callback function which expands a given vdev by calling vdev_online(). | |
3649 | */ | |
3650 | /* ARGSUSED */ | |
65c7cc49 | 3651 | static vdev_t * |
9babb374 BB |
3652 | online_vdev(vdev_t *vd, void *arg) |
3653 | { | |
3654 | spa_t *spa = vd->vdev_spa; | |
3655 | vdev_t *tvd = vd->vdev_top; | |
9babb374 | 3656 | uint64_t guid = vd->vdev_guid; |
428870ff BB |
3657 | uint64_t generation = spa->spa_config_generation + 1; |
3658 | vdev_state_t newstate = VDEV_STATE_UNKNOWN; | |
3659 | int error; | |
9babb374 BB |
3660 | |
3661 | ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE); | |
3662 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
3663 | ||
3664 | /* Calling vdev_online will initialize the new metaslabs */ | |
3665 | spa_config_exit(spa, SCL_STATE, spa); | |
428870ff | 3666 | error = vdev_online(spa, guid, ZFS_ONLINE_EXPAND, &newstate); |
9babb374 BB |
3667 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); |
3668 | ||
428870ff BB |
3669 | /* |
3670 | * If vdev_online returned an error or the underlying vdev_open | |
3671 | * failed then we abort the expand. The only way to know that | |
3672 | * vdev_open fails is by checking the returned newstate. | |
3673 | */ | |
3674 | if (error || newstate != VDEV_STATE_HEALTHY) { | |
c242c188 | 3675 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
3676 | (void) printf("Unable to expand vdev, state %llu, " |
3677 | "error %d\n", (u_longlong_t)newstate, error); | |
3678 | } | |
3679 | return (vd); | |
3680 | } | |
3681 | ASSERT3U(newstate, ==, VDEV_STATE_HEALTHY); | |
3682 | ||
9babb374 BB |
3683 | /* |
3684 | * Since we dropped the lock we need to ensure that we're | |
3685 | * still talking to the original vdev. It's possible this | |
3686 | * vdev may have been detached/replaced while we were | |
3687 | * trying to online it. | |
3688 | */ | |
428870ff | 3689 | if (generation != spa->spa_config_generation) { |
c242c188 | 3690 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
3691 | (void) printf("vdev configuration has changed, " |
3692 | "guid %llu, state %llu, expected gen %llu, " | |
3693 | "got gen %llu\n", | |
3694 | (u_longlong_t)guid, | |
3695 | (u_longlong_t)tvd->vdev_state, | |
3696 | (u_longlong_t)generation, | |
3697 | (u_longlong_t)spa->spa_config_generation); | |
9babb374 BB |
3698 | } |
3699 | return (vd); | |
3700 | } | |
3701 | return (NULL); | |
3702 | } | |
3703 | ||
3704 | /* | |
3705 | * Traverse the vdev tree calling the supplied function. | |
3706 | * We continue to walk the tree until we either have walked all | |
3707 | * children or we receive a non-NULL return from the callback. | |
3708 | * If a NULL callback is passed, then we just return back the first | |
3709 | * leaf vdev we encounter. | |
3710 | */ | |
65c7cc49 | 3711 | static vdev_t * |
9babb374 BB |
3712 | vdev_walk_tree(vdev_t *vd, vdev_t *(*func)(vdev_t *, void *), void *arg) |
3713 | { | |
d6320ddb BB |
3714 | uint_t c; |
3715 | ||
9babb374 BB |
3716 | if (vd->vdev_ops->vdev_op_leaf) { |
3717 | if (func == NULL) | |
3718 | return (vd); | |
3719 | else | |
3720 | return (func(vd, arg)); | |
3721 | } | |
3722 | ||
d6320ddb | 3723 | for (c = 0; c < vd->vdev_children; c++) { |
9babb374 BB |
3724 | vdev_t *cvd = vd->vdev_child[c]; |
3725 | if ((cvd = vdev_walk_tree(cvd, func, arg)) != NULL) | |
3726 | return (cvd); | |
3727 | } | |
3728 | return (NULL); | |
3729 | } | |
3730 | ||
34dc7c2f BB |
3731 | /* |
3732 | * Verify that dynamic LUN growth works as expected. | |
3733 | */ | |
428870ff | 3734 | /* ARGSUSED */ |
34dc7c2f | 3735 | void |
428870ff | 3736 | ztest_vdev_LUN_growth(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3737 | { |
c242c188 | 3738 | spa_t *spa = ztest_spa; |
428870ff BB |
3739 | vdev_t *vd, *tvd; |
3740 | metaslab_class_t *mc; | |
3741 | metaslab_group_t *mg; | |
9babb374 | 3742 | size_t psize, newsize; |
428870ff BB |
3743 | uint64_t top; |
3744 | uint64_t old_class_space, new_class_space, old_ms_count, new_ms_count; | |
34dc7c2f | 3745 | |
d2734cce | 3746 | mutex_enter(&ztest_checkpoint_lock); |
c242c188 | 3747 | mutex_enter(&ztest_vdev_lock); |
9babb374 BB |
3748 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); |
3749 | ||
a1d477c2 MA |
3750 | /* |
3751 | * If there is a vdev removal in progress, it could complete while | |
3752 | * we are running, in which case we would not be able to verify | |
3753 | * that the metaslab_class space increased (because it decreases | |
3754 | * when the device removal completes). | |
3755 | */ | |
9e052db4 | 3756 | if (ztest_device_removal_active) { |
d2734cce | 3757 | spa_config_exit(spa, SCL_STATE, spa); |
a1d477c2 | 3758 | mutex_exit(&ztest_vdev_lock); |
d2734cce | 3759 | mutex_exit(&ztest_checkpoint_lock); |
a1d477c2 MA |
3760 | return; |
3761 | } | |
3762 | ||
428870ff | 3763 | top = ztest_random_vdev_top(spa, B_TRUE); |
9babb374 | 3764 | |
428870ff BB |
3765 | tvd = spa->spa_root_vdev->vdev_child[top]; |
3766 | mg = tvd->vdev_mg; | |
3767 | mc = mg->mg_class; | |
3768 | old_ms_count = tvd->vdev_ms_count; | |
3769 | old_class_space = metaslab_class_get_space(mc); | |
34dc7c2f BB |
3770 | |
3771 | /* | |
9babb374 BB |
3772 | * Determine the size of the first leaf vdev associated with |
3773 | * our top-level device. | |
34dc7c2f | 3774 | */ |
9babb374 BB |
3775 | vd = vdev_walk_tree(tvd, NULL, NULL); |
3776 | ASSERT3P(vd, !=, NULL); | |
3777 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
34dc7c2f | 3778 | |
9babb374 | 3779 | psize = vd->vdev_psize; |
34dc7c2f | 3780 | |
9babb374 | 3781 | /* |
428870ff BB |
3782 | * We only try to expand the vdev if it's healthy, less than 4x its |
3783 | * original size, and it has a valid psize. | |
9babb374 | 3784 | */ |
428870ff | 3785 | if (tvd->vdev_state != VDEV_STATE_HEALTHY || |
c242c188 | 3786 | psize == 0 || psize >= 4 * ztest_opts.zo_vdev_size) { |
9babb374 | 3787 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 3788 | mutex_exit(&ztest_vdev_lock); |
d2734cce | 3789 | mutex_exit(&ztest_checkpoint_lock); |
9babb374 BB |
3790 | return; |
3791 | } | |
3792 | ASSERT(psize > 0); | |
cc99f275 | 3793 | newsize = psize + MAX(psize / 8, SPA_MAXBLOCKSIZE); |
9babb374 | 3794 | ASSERT3U(newsize, >, psize); |
34dc7c2f | 3795 | |
c242c188 | 3796 | if (ztest_opts.zo_verbose >= 6) { |
428870ff | 3797 | (void) printf("Expanding LUN %s from %lu to %lu\n", |
9babb374 BB |
3798 | vd->vdev_path, (ulong_t)psize, (ulong_t)newsize); |
3799 | } | |
3800 | ||
9babb374 BB |
3801 | /* |
3802 | * Growing the vdev is a two step process: | |
3803 | * 1). expand the physical size (i.e. relabel) | |
3804 | * 2). online the vdev to create the new metaslabs | |
3805 | */ | |
3806 | if (vdev_walk_tree(tvd, grow_vdev, &newsize) != NULL || | |
3807 | vdev_walk_tree(tvd, online_vdev, NULL) != NULL || | |
3808 | tvd->vdev_state != VDEV_STATE_HEALTHY) { | |
c242c188 | 3809 | if (ztest_opts.zo_verbose >= 5) { |
9babb374 | 3810 | (void) printf("Could not expand LUN because " |
428870ff | 3811 | "the vdev configuration changed.\n"); |
34dc7c2f | 3812 | } |
428870ff | 3813 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 3814 | mutex_exit(&ztest_vdev_lock); |
d2734cce | 3815 | mutex_exit(&ztest_checkpoint_lock); |
9babb374 | 3816 | return; |
34dc7c2f BB |
3817 | } |
3818 | ||
428870ff | 3819 | spa_config_exit(spa, SCL_STATE, spa); |
9babb374 BB |
3820 | |
3821 | /* | |
3822 | * Expanding the LUN will update the config asynchronously, | |
3823 | * thus we must wait for the async thread to complete any | |
3824 | * pending tasks before proceeding. | |
3825 | */ | |
428870ff BB |
3826 | for (;;) { |
3827 | boolean_t done; | |
3828 | mutex_enter(&spa->spa_async_lock); | |
3829 | done = (spa->spa_async_thread == NULL && !spa->spa_async_tasks); | |
3830 | mutex_exit(&spa->spa_async_lock); | |
3831 | if (done) | |
3832 | break; | |
3833 | txg_wait_synced(spa_get_dsl(spa), 0); | |
3834 | (void) poll(NULL, 0, 100); | |
3835 | } | |
9babb374 BB |
3836 | |
3837 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); | |
428870ff BB |
3838 | |
3839 | tvd = spa->spa_root_vdev->vdev_child[top]; | |
3840 | new_ms_count = tvd->vdev_ms_count; | |
3841 | new_class_space = metaslab_class_get_space(mc); | |
3842 | ||
3843 | if (tvd->vdev_mg != mg || mg->mg_class != mc) { | |
c242c188 | 3844 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
3845 | (void) printf("Could not verify LUN expansion due to " |
3846 | "intervening vdev offline or remove.\n"); | |
3847 | } | |
3848 | spa_config_exit(spa, SCL_STATE, spa); | |
c242c188 | 3849 | mutex_exit(&ztest_vdev_lock); |
d2734cce | 3850 | mutex_exit(&ztest_checkpoint_lock); |
428870ff BB |
3851 | return; |
3852 | } | |
3853 | ||
3854 | /* | |
3855 | * Make sure we were able to grow the vdev. | |
3856 | */ | |
a1d477c2 MA |
3857 | if (new_ms_count <= old_ms_count) { |
3858 | fatal(0, "LUN expansion failed: ms_count %llu < %llu\n", | |
428870ff | 3859 | old_ms_count, new_ms_count); |
a1d477c2 | 3860 | } |
9babb374 BB |
3861 | |
3862 | /* | |
3863 | * Make sure we were able to grow the pool. | |
3864 | */ | |
a1d477c2 MA |
3865 | if (new_class_space <= old_class_space) { |
3866 | fatal(0, "LUN expansion failed: class_space %llu < %llu\n", | |
428870ff | 3867 | old_class_space, new_class_space); |
a1d477c2 | 3868 | } |
428870ff | 3869 | |
c242c188 | 3870 | if (ztest_opts.zo_verbose >= 5) { |
f3c8c9e6 | 3871 | char oldnumbuf[NN_NUMBUF_SZ], newnumbuf[NN_NUMBUF_SZ]; |
9babb374 | 3872 | |
f3c8c9e6 JK |
3873 | nicenum(old_class_space, oldnumbuf, sizeof (oldnumbuf)); |
3874 | nicenum(new_class_space, newnumbuf, sizeof (newnumbuf)); | |
9babb374 BB |
3875 | (void) printf("%s grew from %s to %s\n", |
3876 | spa->spa_name, oldnumbuf, newnumbuf); | |
3877 | } | |
428870ff | 3878 | |
9babb374 | 3879 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 3880 | mutex_exit(&ztest_vdev_lock); |
d2734cce | 3881 | mutex_exit(&ztest_checkpoint_lock); |
34dc7c2f BB |
3882 | } |
3883 | ||
428870ff BB |
3884 | /* |
3885 | * Verify that dmu_objset_{create,destroy,open,close} work as expected. | |
3886 | */ | |
34dc7c2f BB |
3887 | /* ARGSUSED */ |
3888 | static void | |
428870ff | 3889 | ztest_objset_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) |
34dc7c2f BB |
3890 | { |
3891 | /* | |
428870ff | 3892 | * Create the objects common to all ztest datasets. |
34dc7c2f | 3893 | */ |
428870ff | 3894 | VERIFY(zap_create_claim(os, ZTEST_DIROBJ, |
34dc7c2f | 3895 | DMU_OT_ZAP_OTHER, DMU_OT_NONE, 0, tx) == 0); |
428870ff | 3896 | } |
34dc7c2f | 3897 | |
428870ff BB |
3898 | static int |
3899 | ztest_dataset_create(char *dsname) | |
3900 | { | |
4807c0ba TC |
3901 | int err; |
3902 | uint64_t rand; | |
3903 | dsl_crypto_params_t *dcp = NULL; | |
3904 | ||
3905 | /* | |
3906 | * 50% of the time, we create encrypted datasets | |
3907 | * using a random cipher suite and a hard-coded | |
3908 | * wrapping key. | |
3909 | */ | |
3910 | rand = ztest_random(2); | |
3911 | if (rand != 0) { | |
3912 | nvlist_t *crypto_args = fnvlist_alloc(); | |
3913 | nvlist_t *props = fnvlist_alloc(); | |
3914 | ||
3915 | /* slight bias towards the default cipher suite */ | |
3916 | rand = ztest_random(ZIO_CRYPT_FUNCTIONS); | |
3917 | if (rand < ZIO_CRYPT_AES_128_CCM) | |
3918 | rand = ZIO_CRYPT_ON; | |
3919 | ||
3920 | fnvlist_add_uint64(props, | |
3921 | zfs_prop_to_name(ZFS_PROP_ENCRYPTION), rand); | |
3922 | fnvlist_add_uint8_array(crypto_args, "wkeydata", | |
3923 | (uint8_t *)ztest_wkeydata, WRAPPING_KEY_LEN); | |
3924 | ||
3925 | /* | |
3926 | * These parameters aren't really used by the kernel. They | |
3927 | * are simply stored so that userspace knows how to load | |
3928 | * the wrapping key. | |
3929 | */ | |
3930 | fnvlist_add_uint64(props, | |
3931 | zfs_prop_to_name(ZFS_PROP_KEYFORMAT), ZFS_KEYFORMAT_RAW); | |
3932 | fnvlist_add_string(props, | |
3933 | zfs_prop_to_name(ZFS_PROP_KEYLOCATION), "prompt"); | |
3934 | fnvlist_add_uint64(props, | |
3935 | zfs_prop_to_name(ZFS_PROP_PBKDF2_SALT), 0ULL); | |
3936 | fnvlist_add_uint64(props, | |
3937 | zfs_prop_to_name(ZFS_PROP_PBKDF2_ITERS), 0ULL); | |
3938 | ||
3939 | VERIFY0(dsl_crypto_params_create_nvlist(DCP_CMD_NONE, props, | |
3940 | crypto_args, &dcp)); | |
3941 | ||
010d1247 NL |
3942 | /* |
3943 | * Cycle through all available encryption implementations | |
3944 | * to verify interoperability. | |
3945 | */ | |
3946 | VERIFY0(gcm_impl_set("cycle")); | |
3947 | VERIFY0(aes_impl_set("cycle")); | |
3948 | ||
4807c0ba TC |
3949 | fnvlist_free(crypto_args); |
3950 | fnvlist_free(props); | |
3951 | } | |
3952 | ||
3953 | err = dmu_objset_create(dsname, DMU_OST_OTHER, 0, dcp, | |
428870ff | 3954 | ztest_objset_create_cb, NULL); |
4807c0ba | 3955 | dsl_crypto_params_free(dcp, !!err); |
428870ff | 3956 | |
4807c0ba TC |
3957 | rand = ztest_random(100); |
3958 | if (err || rand < 80) | |
428870ff BB |
3959 | return (err); |
3960 | ||
c242c188 | 3961 | if (ztest_opts.zo_verbose >= 5) |
b815ff9a | 3962 | (void) printf("Setting dataset %s to sync always\n", dsname); |
428870ff BB |
3963 | return (ztest_dsl_prop_set_uint64(dsname, ZFS_PROP_SYNC, |
3964 | ZFS_SYNC_ALWAYS, B_FALSE)); | |
34dc7c2f BB |
3965 | } |
3966 | ||
428870ff | 3967 | /* ARGSUSED */ |
34dc7c2f | 3968 | static int |
428870ff | 3969 | ztest_objset_destroy_cb(const char *name, void *arg) |
34dc7c2f | 3970 | { |
34dc7c2f | 3971 | objset_t *os; |
428870ff | 3972 | dmu_object_info_t doi; |
34dc7c2f BB |
3973 | int error; |
3974 | ||
3975 | /* | |
3976 | * Verify that the dataset contains a directory object. | |
3977 | */ | |
4807c0ba TC |
3978 | VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_TRUE, |
3979 | B_TRUE, FTAG, &os)); | |
428870ff | 3980 | error = dmu_object_info(os, ZTEST_DIROBJ, &doi); |
34dc7c2f BB |
3981 | if (error != ENOENT) { |
3982 | /* We could have crashed in the middle of destroying it */ | |
c99c9001 | 3983 | ASSERT0(error); |
428870ff BB |
3984 | ASSERT3U(doi.doi_type, ==, DMU_OT_ZAP_OTHER); |
3985 | ASSERT3S(doi.doi_physical_blocks_512, >=, 0); | |
34dc7c2f | 3986 | } |
b5256303 | 3987 | dmu_objset_disown(os, B_TRUE, FTAG); |
34dc7c2f BB |
3988 | |
3989 | /* | |
3990 | * Destroy the dataset. | |
3991 | */ | |
13fe0198 | 3992 | if (strchr(name, '@') != NULL) { |
dc1fbc43 | 3993 | VERIFY0(dsl_destroy_snapshot(name, B_TRUE)); |
13fe0198 | 3994 | } else { |
dc1fbc43 | 3995 | error = dsl_destroy_head(name); |
0dd6b6bf BB |
3996 | if (error == ENOSPC) { |
3997 | /* There could be checkpoint or insufficient slop */ | |
3998 | ztest_record_enospc(FTAG); | |
3999 | } else if (error != EBUSY) { | |
4000 | /* There could be a hold on this dataset */ | |
dc1fbc43 | 4001 | ASSERT0(error); |
0dd6b6bf | 4002 | } |
13fe0198 | 4003 | } |
34dc7c2f BB |
4004 | return (0); |
4005 | } | |
4006 | ||
428870ff BB |
4007 | static boolean_t |
4008 | ztest_snapshot_create(char *osname, uint64_t id) | |
34dc7c2f | 4009 | { |
eca7b760 | 4010 | char snapname[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
4011 | int error; |
4012 | ||
13fe0198 | 4013 | (void) snprintf(snapname, sizeof (snapname), "%llu", (u_longlong_t)id); |
428870ff | 4014 | |
13fe0198 | 4015 | error = dmu_objset_snapshot_one(osname, snapname); |
428870ff BB |
4016 | if (error == ENOSPC) { |
4017 | ztest_record_enospc(FTAG); | |
4018 | return (B_FALSE); | |
4019 | } | |
13fe0198 MA |
4020 | if (error != 0 && error != EEXIST) { |
4021 | fatal(0, "ztest_snapshot_create(%s@%s) = %d", osname, | |
4022 | snapname, error); | |
4023 | } | |
428870ff BB |
4024 | return (B_TRUE); |
4025 | } | |
4026 | ||
4027 | static boolean_t | |
4028 | ztest_snapshot_destroy(char *osname, uint64_t id) | |
4029 | { | |
eca7b760 | 4030 | char snapname[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
4031 | int error; |
4032 | ||
eca7b760 | 4033 | (void) snprintf(snapname, sizeof (snapname), "%s@%llu", osname, |
428870ff BB |
4034 | (u_longlong_t)id); |
4035 | ||
13fe0198 | 4036 | error = dsl_destroy_snapshot(snapname, B_FALSE); |
428870ff BB |
4037 | if (error != 0 && error != ENOENT) |
4038 | fatal(0, "ztest_snapshot_destroy(%s) = %d", snapname, error); | |
4039 | return (B_TRUE); | |
34dc7c2f BB |
4040 | } |
4041 | ||
428870ff | 4042 | /* ARGSUSED */ |
34dc7c2f | 4043 | void |
428870ff | 4044 | ztest_dmu_objset_create_destroy(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4045 | { |
40b84e7a | 4046 | ztest_ds_t *zdtmp; |
428870ff | 4047 | int iters; |
34dc7c2f | 4048 | int error; |
b128c09f | 4049 | objset_t *os, *os2; |
eca7b760 | 4050 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
34dc7c2f | 4051 | zilog_t *zilog; |
d6320ddb | 4052 | int i; |
34dc7c2f | 4053 | |
40b84e7a | 4054 | zdtmp = umem_alloc(sizeof (ztest_ds_t), UMEM_NOFAIL); |
40b84e7a | 4055 | |
2d9142c9 | 4056 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
34dc7c2f | 4057 | |
eca7b760 | 4058 | (void) snprintf(name, sizeof (name), "%s/temp_%llu", |
c242c188 | 4059 | ztest_opts.zo_pool, (u_longlong_t)id); |
34dc7c2f BB |
4060 | |
4061 | /* | |
4062 | * If this dataset exists from a previous run, process its replay log | |
13fe0198 | 4063 | * half of the time. If we don't replay it, then dsl_destroy_head() |
428870ff | 4064 | * (invoked from ztest_objset_destroy_cb()) should just throw it away. |
34dc7c2f BB |
4065 | */ |
4066 | if (ztest_random(2) == 0 && | |
4807c0ba | 4067 | ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, |
b5256303 | 4068 | B_TRUE, FTAG, &os) == 0) { |
c242c188 | 4069 | ztest_zd_init(zdtmp, NULL, os); |
40b84e7a BB |
4070 | zil_replay(os, zdtmp, ztest_replay_vector); |
4071 | ztest_zd_fini(zdtmp); | |
b5256303 | 4072 | dmu_objset_disown(os, B_TRUE, FTAG); |
34dc7c2f BB |
4073 | } |
4074 | ||
4075 | /* | |
4076 | * There may be an old instance of the dataset we're about to | |
4077 | * create lying around from a previous run. If so, destroy it | |
4078 | * and all of its snapshots. | |
4079 | */ | |
428870ff | 4080 | (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL, |
34dc7c2f BB |
4081 | DS_FIND_CHILDREN | DS_FIND_SNAPSHOTS); |
4082 | ||
4083 | /* | |
4084 | * Verify that the destroyed dataset is no longer in the namespace. | |
4085 | */ | |
4807c0ba TC |
4086 | VERIFY3U(ENOENT, ==, ztest_dmu_objset_own(name, DMU_OST_OTHER, B_TRUE, |
4087 | B_TRUE, FTAG, &os)); | |
34dc7c2f BB |
4088 | |
4089 | /* | |
4090 | * Verify that we can create a new dataset. | |
4091 | */ | |
428870ff | 4092 | error = ztest_dataset_create(name); |
34dc7c2f BB |
4093 | if (error) { |
4094 | if (error == ENOSPC) { | |
428870ff | 4095 | ztest_record_enospc(FTAG); |
40b84e7a | 4096 | goto out; |
34dc7c2f BB |
4097 | } |
4098 | fatal(0, "dmu_objset_create(%s) = %d", name, error); | |
4099 | } | |
4100 | ||
4807c0ba | 4101 | VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, B_TRUE, |
b5256303 | 4102 | FTAG, &os)); |
428870ff | 4103 | |
c242c188 | 4104 | ztest_zd_init(zdtmp, NULL, os); |
34dc7c2f BB |
4105 | |
4106 | /* | |
4107 | * Open the intent log for it. | |
4108 | */ | |
428870ff | 4109 | zilog = zil_open(os, ztest_get_data); |
34dc7c2f BB |
4110 | |
4111 | /* | |
428870ff BB |
4112 | * Put some objects in there, do a little I/O to them, |
4113 | * and randomly take a couple of snapshots along the way. | |
34dc7c2f | 4114 | */ |
428870ff | 4115 | iters = ztest_random(5); |
d6320ddb | 4116 | for (i = 0; i < iters; i++) { |
40b84e7a | 4117 | ztest_dmu_object_alloc_free(zdtmp, id); |
428870ff BB |
4118 | if (ztest_random(iters) == 0) |
4119 | (void) ztest_snapshot_create(name, i); | |
34dc7c2f BB |
4120 | } |
4121 | ||
4122 | /* | |
4123 | * Verify that we cannot create an existing dataset. | |
4124 | */ | |
428870ff | 4125 | VERIFY3U(EEXIST, ==, |
b5256303 | 4126 | dmu_objset_create(name, DMU_OST_OTHER, 0, NULL, NULL, NULL)); |
34dc7c2f BB |
4127 | |
4128 | /* | |
428870ff | 4129 | * Verify that we can hold an objset that is also owned. |
b128c09f | 4130 | */ |
428870ff BB |
4131 | VERIFY3U(0, ==, dmu_objset_hold(name, FTAG, &os2)); |
4132 | dmu_objset_rele(os2, FTAG); | |
34dc7c2f | 4133 | |
428870ff BB |
4134 | /* |
4135 | * Verify that we cannot own an objset that is already owned. | |
4136 | */ | |
4807c0ba TC |
4137 | VERIFY3U(EBUSY, ==, ztest_dmu_objset_own(name, DMU_OST_OTHER, |
4138 | B_FALSE, B_TRUE, FTAG, &os2)); | |
34dc7c2f | 4139 | |
428870ff | 4140 | zil_close(zilog); |
b5256303 | 4141 | dmu_objset_disown(os, B_TRUE, FTAG); |
40b84e7a BB |
4142 | ztest_zd_fini(zdtmp); |
4143 | out: | |
2d9142c9 | 4144 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
40b84e7a | 4145 | |
40b84e7a | 4146 | umem_free(zdtmp, sizeof (ztest_ds_t)); |
34dc7c2f BB |
4147 | } |
4148 | ||
4149 | /* | |
4150 | * Verify that dmu_snapshot_{create,destroy,open,close} work as expected. | |
4151 | */ | |
4152 | void | |
428870ff | 4153 | ztest_dmu_snapshot_create_destroy(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4154 | { |
2d9142c9 | 4155 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
428870ff BB |
4156 | (void) ztest_snapshot_destroy(zd->zd_name, id); |
4157 | (void) ztest_snapshot_create(zd->zd_name, id); | |
2d9142c9 | 4158 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
34dc7c2f BB |
4159 | } |
4160 | ||
9babb374 BB |
4161 | /* |
4162 | * Cleanup non-standard snapshots and clones. | |
4163 | */ | |
65c7cc49 | 4164 | static void |
428870ff | 4165 | ztest_dsl_dataset_cleanup(char *osname, uint64_t id) |
9babb374 | 4166 | { |
40b84e7a BB |
4167 | char *snap1name; |
4168 | char *clone1name; | |
4169 | char *snap2name; | |
4170 | char *clone2name; | |
4171 | char *snap3name; | |
9babb374 BB |
4172 | int error; |
4173 | ||
eca7b760 IK |
4174 | snap1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); |
4175 | clone1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4176 | snap2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4177 | clone2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4178 | snap3name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4179 | ||
4180 | (void) snprintf(snap1name, ZFS_MAX_DATASET_NAME_LEN, | |
4181 | "%s@s1_%llu", osname, (u_longlong_t)id); | |
4182 | (void) snprintf(clone1name, ZFS_MAX_DATASET_NAME_LEN, | |
4183 | "%s/c1_%llu", osname, (u_longlong_t)id); | |
4184 | (void) snprintf(snap2name, ZFS_MAX_DATASET_NAME_LEN, | |
4185 | "%s@s2_%llu", clone1name, (u_longlong_t)id); | |
4186 | (void) snprintf(clone2name, ZFS_MAX_DATASET_NAME_LEN, | |
4187 | "%s/c2_%llu", osname, (u_longlong_t)id); | |
4188 | (void) snprintf(snap3name, ZFS_MAX_DATASET_NAME_LEN, | |
4189 | "%s@s3_%llu", clone1name, (u_longlong_t)id); | |
9babb374 | 4190 | |
13fe0198 | 4191 | error = dsl_destroy_head(clone2name); |
9babb374 | 4192 | if (error && error != ENOENT) |
13fe0198 MA |
4193 | fatal(0, "dsl_destroy_head(%s) = %d", clone2name, error); |
4194 | error = dsl_destroy_snapshot(snap3name, B_FALSE); | |
9babb374 | 4195 | if (error && error != ENOENT) |
13fe0198 MA |
4196 | fatal(0, "dsl_destroy_snapshot(%s) = %d", snap3name, error); |
4197 | error = dsl_destroy_snapshot(snap2name, B_FALSE); | |
9babb374 | 4198 | if (error && error != ENOENT) |
13fe0198 MA |
4199 | fatal(0, "dsl_destroy_snapshot(%s) = %d", snap2name, error); |
4200 | error = dsl_destroy_head(clone1name); | |
9babb374 | 4201 | if (error && error != ENOENT) |
13fe0198 MA |
4202 | fatal(0, "dsl_destroy_head(%s) = %d", clone1name, error); |
4203 | error = dsl_destroy_snapshot(snap1name, B_FALSE); | |
9babb374 | 4204 | if (error && error != ENOENT) |
13fe0198 | 4205 | fatal(0, "dsl_destroy_snapshot(%s) = %d", snap1name, error); |
40b84e7a | 4206 | |
eca7b760 IK |
4207 | umem_free(snap1name, ZFS_MAX_DATASET_NAME_LEN); |
4208 | umem_free(clone1name, ZFS_MAX_DATASET_NAME_LEN); | |
4209 | umem_free(snap2name, ZFS_MAX_DATASET_NAME_LEN); | |
4210 | umem_free(clone2name, ZFS_MAX_DATASET_NAME_LEN); | |
4211 | umem_free(snap3name, ZFS_MAX_DATASET_NAME_LEN); | |
9babb374 BB |
4212 | } |
4213 | ||
4214 | /* | |
4215 | * Verify dsl_dataset_promote handles EBUSY | |
4216 | */ | |
4217 | void | |
428870ff | 4218 | ztest_dsl_dataset_promote_busy(ztest_ds_t *zd, uint64_t id) |
9babb374 | 4219 | { |
13fe0198 | 4220 | objset_t *os; |
40b84e7a BB |
4221 | char *snap1name; |
4222 | char *clone1name; | |
4223 | char *snap2name; | |
4224 | char *clone2name; | |
4225 | char *snap3name; | |
428870ff BB |
4226 | char *osname = zd->zd_name; |
4227 | int error; | |
9babb374 | 4228 | |
eca7b760 IK |
4229 | snap1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); |
4230 | clone1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4231 | snap2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4232 | clone2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4233 | snap3name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
40b84e7a | 4234 | |
2d9142c9 | 4235 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
9babb374 | 4236 | |
428870ff | 4237 | ztest_dsl_dataset_cleanup(osname, id); |
9babb374 | 4238 | |
eca7b760 IK |
4239 | (void) snprintf(snap1name, ZFS_MAX_DATASET_NAME_LEN, |
4240 | "%s@s1_%llu", osname, (u_longlong_t)id); | |
4241 | (void) snprintf(clone1name, ZFS_MAX_DATASET_NAME_LEN, | |
4242 | "%s/c1_%llu", osname, (u_longlong_t)id); | |
4243 | (void) snprintf(snap2name, ZFS_MAX_DATASET_NAME_LEN, | |
4244 | "%s@s2_%llu", clone1name, (u_longlong_t)id); | |
4245 | (void) snprintf(clone2name, ZFS_MAX_DATASET_NAME_LEN, | |
4246 | "%s/c2_%llu", osname, (u_longlong_t)id); | |
4247 | (void) snprintf(snap3name, ZFS_MAX_DATASET_NAME_LEN, | |
4248 | "%s@s3_%llu", clone1name, (u_longlong_t)id); | |
9babb374 | 4249 | |
6f1ffb06 | 4250 | error = dmu_objset_snapshot_one(osname, strchr(snap1name, '@') + 1); |
9babb374 BB |
4251 | if (error && error != EEXIST) { |
4252 | if (error == ENOSPC) { | |
428870ff | 4253 | ztest_record_enospc(FTAG); |
9babb374 BB |
4254 | goto out; |
4255 | } | |
4256 | fatal(0, "dmu_take_snapshot(%s) = %d", snap1name, error); | |
4257 | } | |
4258 | ||
13fe0198 | 4259 | error = dmu_objset_clone(clone1name, snap1name); |
9babb374 BB |
4260 | if (error) { |
4261 | if (error == ENOSPC) { | |
428870ff | 4262 | ztest_record_enospc(FTAG); |
9babb374 BB |
4263 | goto out; |
4264 | } | |
4265 | fatal(0, "dmu_objset_create(%s) = %d", clone1name, error); | |
4266 | } | |
4267 | ||
6f1ffb06 | 4268 | error = dmu_objset_snapshot_one(clone1name, strchr(snap2name, '@') + 1); |
428870ff BB |
4269 | if (error && error != EEXIST) { |
4270 | if (error == ENOSPC) { | |
4271 | ztest_record_enospc(FTAG); | |
4272 | goto out; | |
34dc7c2f | 4273 | } |
428870ff BB |
4274 | fatal(0, "dmu_open_snapshot(%s) = %d", snap2name, error); |
4275 | } | |
34dc7c2f | 4276 | |
6f1ffb06 | 4277 | error = dmu_objset_snapshot_one(clone1name, strchr(snap3name, '@') + 1); |
428870ff BB |
4278 | if (error && error != EEXIST) { |
4279 | if (error == ENOSPC) { | |
4280 | ztest_record_enospc(FTAG); | |
4281 | goto out; | |
4282 | } | |
4283 | fatal(0, "dmu_open_snapshot(%s) = %d", snap3name, error); | |
4284 | } | |
34dc7c2f | 4285 | |
13fe0198 | 4286 | error = dmu_objset_clone(clone2name, snap3name); |
428870ff BB |
4287 | if (error) { |
4288 | if (error == ENOSPC) { | |
4289 | ztest_record_enospc(FTAG); | |
4290 | goto out; | |
4291 | } | |
4292 | fatal(0, "dmu_objset_create(%s) = %d", clone2name, error); | |
4293 | } | |
34dc7c2f | 4294 | |
4807c0ba | 4295 | error = ztest_dmu_objset_own(snap2name, DMU_OST_ANY, B_TRUE, B_TRUE, |
b5256303 | 4296 | FTAG, &os); |
428870ff | 4297 | if (error) |
13fe0198 | 4298 | fatal(0, "dmu_objset_own(%s) = %d", snap2name, error); |
428870ff | 4299 | error = dsl_dataset_promote(clone2name, NULL); |
9b67f605 | 4300 | if (error == ENOSPC) { |
b5256303 | 4301 | dmu_objset_disown(os, B_TRUE, FTAG); |
9b67f605 MA |
4302 | ztest_record_enospc(FTAG); |
4303 | goto out; | |
4304 | } | |
428870ff BB |
4305 | if (error != EBUSY) |
4306 | fatal(0, "dsl_dataset_promote(%s), %d, not EBUSY", clone2name, | |
4307 | error); | |
b5256303 | 4308 | dmu_objset_disown(os, B_TRUE, FTAG); |
34dc7c2f | 4309 | |
428870ff BB |
4310 | out: |
4311 | ztest_dsl_dataset_cleanup(osname, id); | |
34dc7c2f | 4312 | |
2d9142c9 | 4313 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
40b84e7a | 4314 | |
eca7b760 IK |
4315 | umem_free(snap1name, ZFS_MAX_DATASET_NAME_LEN); |
4316 | umem_free(clone1name, ZFS_MAX_DATASET_NAME_LEN); | |
4317 | umem_free(snap2name, ZFS_MAX_DATASET_NAME_LEN); | |
4318 | umem_free(clone2name, ZFS_MAX_DATASET_NAME_LEN); | |
4319 | umem_free(snap3name, ZFS_MAX_DATASET_NAME_LEN); | |
428870ff | 4320 | } |
34dc7c2f | 4321 | |
40b84e7a | 4322 | #undef OD_ARRAY_SIZE |
d1d7e268 | 4323 | #define OD_ARRAY_SIZE 4 |
40b84e7a | 4324 | |
428870ff BB |
4325 | /* |
4326 | * Verify that dmu_object_{alloc,free} work as expected. | |
4327 | */ | |
4328 | void | |
4329 | ztest_dmu_object_alloc_free(ztest_ds_t *zd, uint64_t id) | |
4330 | { | |
40b84e7a BB |
4331 | ztest_od_t *od; |
4332 | int batchsize; | |
4333 | int size; | |
d6320ddb | 4334 | int b; |
34dc7c2f | 4335 | |
d1d7e268 | 4336 | size = sizeof (ztest_od_t) * OD_ARRAY_SIZE; |
40b84e7a BB |
4337 | od = umem_alloc(size, UMEM_NOFAIL); |
4338 | batchsize = OD_ARRAY_SIZE; | |
4339 | ||
d6320ddb | 4340 | for (b = 0; b < batchsize; b++) |
50c957f7 NB |
4341 | ztest_od_init(od + b, id, FTAG, b, DMU_OT_UINT64_OTHER, |
4342 | 0, 0, 0); | |
34dc7c2f | 4343 | |
428870ff BB |
4344 | /* |
4345 | * Destroy the previous batch of objects, create a new batch, | |
4346 | * and do some I/O on the new objects. | |
4347 | */ | |
40b84e7a | 4348 | if (ztest_object_init(zd, od, size, B_TRUE) != 0) |
428870ff | 4349 | return; |
34dc7c2f | 4350 | |
428870ff BB |
4351 | while (ztest_random(4 * batchsize) != 0) |
4352 | ztest_io(zd, od[ztest_random(batchsize)].od_object, | |
4353 | ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
40b84e7a BB |
4354 | |
4355 | umem_free(od, size); | |
34dc7c2f BB |
4356 | } |
4357 | ||
047116ac TC |
4358 | /* |
4359 | * Rewind the global allocator to verify object allocation backfilling. | |
4360 | */ | |
4361 | void | |
4362 | ztest_dmu_object_next_chunk(ztest_ds_t *zd, uint64_t id) | |
4363 | { | |
4364 | objset_t *os = zd->zd_os; | |
4365 | int dnodes_per_chunk = 1 << dmu_object_alloc_chunk_shift; | |
4366 | uint64_t object; | |
4367 | ||
4368 | /* | |
4369 | * Rewind the global allocator randomly back to a lower object number | |
4370 | * to force backfilling and reclamation of recently freed dnodes. | |
4371 | */ | |
4372 | mutex_enter(&os->os_obj_lock); | |
4373 | object = ztest_random(os->os_obj_next_chunk); | |
4374 | os->os_obj_next_chunk = P2ALIGN(object, dnodes_per_chunk); | |
4375 | mutex_exit(&os->os_obj_lock); | |
4376 | } | |
4377 | ||
40b84e7a | 4378 | #undef OD_ARRAY_SIZE |
d1d7e268 | 4379 | #define OD_ARRAY_SIZE 2 |
40b84e7a | 4380 | |
34dc7c2f BB |
4381 | /* |
4382 | * Verify that dmu_{read,write} work as expected. | |
4383 | */ | |
34dc7c2f | 4384 | void |
428870ff | 4385 | ztest_dmu_read_write(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4386 | { |
40b84e7a BB |
4387 | int size; |
4388 | ztest_od_t *od; | |
4389 | ||
428870ff | 4390 | objset_t *os = zd->zd_os; |
d1d7e268 | 4391 | size = sizeof (ztest_od_t) * OD_ARRAY_SIZE; |
40b84e7a | 4392 | od = umem_alloc(size, UMEM_NOFAIL); |
34dc7c2f BB |
4393 | dmu_tx_t *tx; |
4394 | int i, freeit, error; | |
4395 | uint64_t n, s, txg; | |
4396 | bufwad_t *packbuf, *bigbuf, *pack, *bigH, *bigT; | |
428870ff BB |
4397 | uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize; |
4398 | uint64_t chunksize = (1000 + ztest_random(1000)) * sizeof (uint64_t); | |
34dc7c2f BB |
4399 | uint64_t regions = 997; |
4400 | uint64_t stride = 123456789ULL; | |
4401 | uint64_t width = 40; | |
4402 | int free_percent = 5; | |
4403 | ||
4404 | /* | |
4405 | * This test uses two objects, packobj and bigobj, that are always | |
4406 | * updated together (i.e. in the same tx) so that their contents are | |
4407 | * in sync and can be compared. Their contents relate to each other | |
4408 | * in a simple way: packobj is a dense array of 'bufwad' structures, | |
4409 | * while bigobj is a sparse array of the same bufwads. Specifically, | |
4410 | * for any index n, there are three bufwads that should be identical: | |
4411 | * | |
4412 | * packobj, at offset n * sizeof (bufwad_t) | |
4413 | * bigobj, at the head of the nth chunk | |
4414 | * bigobj, at the tail of the nth chunk | |
4415 | * | |
4416 | * The chunk size is arbitrary. It doesn't have to be a power of two, | |
4417 | * and it doesn't have any relation to the object blocksize. | |
4418 | * The only requirement is that it can hold at least two bufwads. | |
4419 | * | |
4420 | * Normally, we write the bufwad to each of these locations. | |
4421 | * However, free_percent of the time we instead write zeroes to | |
4422 | * packobj and perform a dmu_free_range() on bigobj. By comparing | |
4423 | * bigobj to packobj, we can verify that the DMU is correctly | |
4424 | * tracking which parts of an object are allocated and free, | |
4425 | * and that the contents of the allocated blocks are correct. | |
4426 | */ | |
4427 | ||
4428 | /* | |
4429 | * Read the directory info. If it's the first time, set things up. | |
4430 | */ | |
50c957f7 NB |
4431 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, chunksize); |
4432 | ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, 0, | |
4433 | chunksize); | |
34dc7c2f | 4434 | |
40b84e7a BB |
4435 | if (ztest_object_init(zd, od, size, B_FALSE) != 0) { |
4436 | umem_free(od, size); | |
428870ff | 4437 | return; |
40b84e7a | 4438 | } |
34dc7c2f | 4439 | |
428870ff BB |
4440 | bigobj = od[0].od_object; |
4441 | packobj = od[1].od_object; | |
4442 | chunksize = od[0].od_gen; | |
4443 | ASSERT(chunksize == od[1].od_gen); | |
34dc7c2f BB |
4444 | |
4445 | /* | |
4446 | * Prefetch a random chunk of the big object. | |
4447 | * Our aim here is to get some async reads in flight | |
4448 | * for blocks that we may free below; the DMU should | |
4449 | * handle this race correctly. | |
4450 | */ | |
4451 | n = ztest_random(regions) * stride + ztest_random(width); | |
4452 | s = 1 + ztest_random(2 * width - 1); | |
fcff0f35 PD |
4453 | dmu_prefetch(os, bigobj, 0, n * chunksize, s * chunksize, |
4454 | ZIO_PRIORITY_SYNC_READ); | |
34dc7c2f BB |
4455 | |
4456 | /* | |
4457 | * Pick a random index and compute the offsets into packobj and bigobj. | |
4458 | */ | |
4459 | n = ztest_random(regions) * stride + ztest_random(width); | |
4460 | s = 1 + ztest_random(width - 1); | |
4461 | ||
4462 | packoff = n * sizeof (bufwad_t); | |
4463 | packsize = s * sizeof (bufwad_t); | |
4464 | ||
428870ff BB |
4465 | bigoff = n * chunksize; |
4466 | bigsize = s * chunksize; | |
34dc7c2f BB |
4467 | |
4468 | packbuf = umem_alloc(packsize, UMEM_NOFAIL); | |
4469 | bigbuf = umem_alloc(bigsize, UMEM_NOFAIL); | |
4470 | ||
4471 | /* | |
4472 | * free_percent of the time, free a range of bigobj rather than | |
4473 | * overwriting it. | |
4474 | */ | |
4475 | freeit = (ztest_random(100) < free_percent); | |
4476 | ||
4477 | /* | |
4478 | * Read the current contents of our objects. | |
4479 | */ | |
428870ff | 4480 | error = dmu_read(os, packobj, packoff, packsize, packbuf, |
9babb374 | 4481 | DMU_READ_PREFETCH); |
c99c9001 | 4482 | ASSERT0(error); |
428870ff | 4483 | error = dmu_read(os, bigobj, bigoff, bigsize, bigbuf, |
9babb374 | 4484 | DMU_READ_PREFETCH); |
c99c9001 | 4485 | ASSERT0(error); |
34dc7c2f BB |
4486 | |
4487 | /* | |
4488 | * Get a tx for the mods to both packobj and bigobj. | |
4489 | */ | |
4490 | tx = dmu_tx_create(os); | |
4491 | ||
428870ff | 4492 | dmu_tx_hold_write(tx, packobj, packoff, packsize); |
34dc7c2f BB |
4493 | |
4494 | if (freeit) | |
428870ff | 4495 | dmu_tx_hold_free(tx, bigobj, bigoff, bigsize); |
34dc7c2f | 4496 | else |
428870ff | 4497 | dmu_tx_hold_write(tx, bigobj, bigoff, bigsize); |
34dc7c2f | 4498 | |
383fc4a9 MA |
4499 | /* This accounts for setting the checksum/compression. */ |
4500 | dmu_tx_hold_bonus(tx, bigobj); | |
4501 | ||
428870ff BB |
4502 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); |
4503 | if (txg == 0) { | |
34dc7c2f BB |
4504 | umem_free(packbuf, packsize); |
4505 | umem_free(bigbuf, bigsize); | |
40b84e7a | 4506 | umem_free(od, size); |
34dc7c2f BB |
4507 | return; |
4508 | } | |
4509 | ||
9b67f605 MA |
4510 | enum zio_checksum cksum; |
4511 | do { | |
4512 | cksum = (enum zio_checksum) | |
4513 | ztest_random_dsl_prop(ZFS_PROP_CHECKSUM); | |
4514 | } while (cksum >= ZIO_CHECKSUM_LEGACY_FUNCTIONS); | |
4515 | dmu_object_set_checksum(os, bigobj, cksum, tx); | |
428870ff | 4516 | |
9b67f605 MA |
4517 | enum zio_compress comp; |
4518 | do { | |
4519 | comp = (enum zio_compress) | |
4520 | ztest_random_dsl_prop(ZFS_PROP_COMPRESSION); | |
4521 | } while (comp >= ZIO_COMPRESS_LEGACY_FUNCTIONS); | |
4522 | dmu_object_set_compress(os, bigobj, comp, tx); | |
34dc7c2f BB |
4523 | |
4524 | /* | |
4525 | * For each index from n to n + s, verify that the existing bufwad | |
4526 | * in packobj matches the bufwads at the head and tail of the | |
4527 | * corresponding chunk in bigobj. Then update all three bufwads | |
4528 | * with the new values we want to write out. | |
4529 | */ | |
4530 | for (i = 0; i < s; i++) { | |
4531 | /* LINTED */ | |
4532 | pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t)); | |
4533 | /* LINTED */ | |
428870ff | 4534 | bigH = (bufwad_t *)((char *)bigbuf + i * chunksize); |
34dc7c2f | 4535 | /* LINTED */ |
428870ff | 4536 | bigT = (bufwad_t *)((char *)bigH + chunksize) - 1; |
34dc7c2f BB |
4537 | |
4538 | ASSERT((uintptr_t)bigH - (uintptr_t)bigbuf < bigsize); | |
4539 | ASSERT((uintptr_t)bigT - (uintptr_t)bigbuf < bigsize); | |
4540 | ||
4541 | if (pack->bw_txg > txg) | |
4542 | fatal(0, "future leak: got %llx, open txg is %llx", | |
4543 | pack->bw_txg, txg); | |
4544 | ||
4545 | if (pack->bw_data != 0 && pack->bw_index != n + i) | |
4546 | fatal(0, "wrong index: got %llx, wanted %llx+%llx", | |
4547 | pack->bw_index, n, i); | |
4548 | ||
4549 | if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0) | |
4550 | fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH); | |
4551 | ||
4552 | if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0) | |
4553 | fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT); | |
4554 | ||
4555 | if (freeit) { | |
4556 | bzero(pack, sizeof (bufwad_t)); | |
4557 | } else { | |
4558 | pack->bw_index = n + i; | |
4559 | pack->bw_txg = txg; | |
4560 | pack->bw_data = 1 + ztest_random(-2ULL); | |
4561 | } | |
4562 | *bigH = *pack; | |
4563 | *bigT = *pack; | |
4564 | } | |
4565 | ||
4566 | /* | |
4567 | * We've verified all the old bufwads, and made new ones. | |
4568 | * Now write them out. | |
4569 | */ | |
428870ff | 4570 | dmu_write(os, packobj, packoff, packsize, packbuf, tx); |
34dc7c2f BB |
4571 | |
4572 | if (freeit) { | |
c242c188 | 4573 | if (ztest_opts.zo_verbose >= 7) { |
34dc7c2f BB |
4574 | (void) printf("freeing offset %llx size %llx" |
4575 | " txg %llx\n", | |
4576 | (u_longlong_t)bigoff, | |
4577 | (u_longlong_t)bigsize, | |
4578 | (u_longlong_t)txg); | |
4579 | } | |
428870ff | 4580 | VERIFY(0 == dmu_free_range(os, bigobj, bigoff, bigsize, tx)); |
34dc7c2f | 4581 | } else { |
c242c188 | 4582 | if (ztest_opts.zo_verbose >= 7) { |
34dc7c2f BB |
4583 | (void) printf("writing offset %llx size %llx" |
4584 | " txg %llx\n", | |
4585 | (u_longlong_t)bigoff, | |
4586 | (u_longlong_t)bigsize, | |
4587 | (u_longlong_t)txg); | |
4588 | } | |
428870ff | 4589 | dmu_write(os, bigobj, bigoff, bigsize, bigbuf, tx); |
34dc7c2f BB |
4590 | } |
4591 | ||
4592 | dmu_tx_commit(tx); | |
4593 | ||
4594 | /* | |
4595 | * Sanity check the stuff we just wrote. | |
4596 | */ | |
4597 | { | |
4598 | void *packcheck = umem_alloc(packsize, UMEM_NOFAIL); | |
4599 | void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL); | |
4600 | ||
428870ff | 4601 | VERIFY(0 == dmu_read(os, packobj, packoff, |
9babb374 | 4602 | packsize, packcheck, DMU_READ_PREFETCH)); |
428870ff | 4603 | VERIFY(0 == dmu_read(os, bigobj, bigoff, |
9babb374 | 4604 | bigsize, bigcheck, DMU_READ_PREFETCH)); |
34dc7c2f BB |
4605 | |
4606 | ASSERT(bcmp(packbuf, packcheck, packsize) == 0); | |
4607 | ASSERT(bcmp(bigbuf, bigcheck, bigsize) == 0); | |
4608 | ||
4609 | umem_free(packcheck, packsize); | |
4610 | umem_free(bigcheck, bigsize); | |
4611 | } | |
4612 | ||
4613 | umem_free(packbuf, packsize); | |
4614 | umem_free(bigbuf, bigsize); | |
40b84e7a | 4615 | umem_free(od, size); |
34dc7c2f BB |
4616 | } |
4617 | ||
65c7cc49 | 4618 | static void |
9babb374 | 4619 | compare_and_update_pbbufs(uint64_t s, bufwad_t *packbuf, bufwad_t *bigbuf, |
428870ff | 4620 | uint64_t bigsize, uint64_t n, uint64_t chunksize, uint64_t txg) |
9babb374 BB |
4621 | { |
4622 | uint64_t i; | |
4623 | bufwad_t *pack; | |
4624 | bufwad_t *bigH; | |
4625 | bufwad_t *bigT; | |
4626 | ||
4627 | /* | |
4628 | * For each index from n to n + s, verify that the existing bufwad | |
4629 | * in packobj matches the bufwads at the head and tail of the | |
4630 | * corresponding chunk in bigobj. Then update all three bufwads | |
4631 | * with the new values we want to write out. | |
4632 | */ | |
4633 | for (i = 0; i < s; i++) { | |
4634 | /* LINTED */ | |
4635 | pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t)); | |
4636 | /* LINTED */ | |
428870ff | 4637 | bigH = (bufwad_t *)((char *)bigbuf + i * chunksize); |
9babb374 | 4638 | /* LINTED */ |
428870ff | 4639 | bigT = (bufwad_t *)((char *)bigH + chunksize) - 1; |
9babb374 BB |
4640 | |
4641 | ASSERT((uintptr_t)bigH - (uintptr_t)bigbuf < bigsize); | |
4642 | ASSERT((uintptr_t)bigT - (uintptr_t)bigbuf < bigsize); | |
4643 | ||
4644 | if (pack->bw_txg > txg) | |
4645 | fatal(0, "future leak: got %llx, open txg is %llx", | |
4646 | pack->bw_txg, txg); | |
4647 | ||
4648 | if (pack->bw_data != 0 && pack->bw_index != n + i) | |
4649 | fatal(0, "wrong index: got %llx, wanted %llx+%llx", | |
4650 | pack->bw_index, n, i); | |
4651 | ||
4652 | if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0) | |
4653 | fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH); | |
4654 | ||
4655 | if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0) | |
4656 | fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT); | |
4657 | ||
4658 | pack->bw_index = n + i; | |
4659 | pack->bw_txg = txg; | |
4660 | pack->bw_data = 1 + ztest_random(-2ULL); | |
4661 | ||
4662 | *bigH = *pack; | |
4663 | *bigT = *pack; | |
4664 | } | |
4665 | } | |
4666 | ||
40b84e7a | 4667 | #undef OD_ARRAY_SIZE |
d1d7e268 | 4668 | #define OD_ARRAY_SIZE 2 |
40b84e7a | 4669 | |
9babb374 | 4670 | void |
428870ff | 4671 | ztest_dmu_read_write_zcopy(ztest_ds_t *zd, uint64_t id) |
9babb374 | 4672 | { |
428870ff | 4673 | objset_t *os = zd->zd_os; |
40b84e7a | 4674 | ztest_od_t *od; |
9babb374 BB |
4675 | dmu_tx_t *tx; |
4676 | uint64_t i; | |
4677 | int error; | |
40b84e7a | 4678 | int size; |
9babb374 BB |
4679 | uint64_t n, s, txg; |
4680 | bufwad_t *packbuf, *bigbuf; | |
428870ff BB |
4681 | uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize; |
4682 | uint64_t blocksize = ztest_random_blocksize(); | |
4683 | uint64_t chunksize = blocksize; | |
9babb374 BB |
4684 | uint64_t regions = 997; |
4685 | uint64_t stride = 123456789ULL; | |
4686 | uint64_t width = 9; | |
4687 | dmu_buf_t *bonus_db; | |
4688 | arc_buf_t **bigbuf_arcbufs; | |
428870ff | 4689 | dmu_object_info_t doi; |
9babb374 | 4690 | |
d1d7e268 | 4691 | size = sizeof (ztest_od_t) * OD_ARRAY_SIZE; |
40b84e7a BB |
4692 | od = umem_alloc(size, UMEM_NOFAIL); |
4693 | ||
9babb374 BB |
4694 | /* |
4695 | * This test uses two objects, packobj and bigobj, that are always | |
4696 | * updated together (i.e. in the same tx) so that their contents are | |
4697 | * in sync and can be compared. Their contents relate to each other | |
4698 | * in a simple way: packobj is a dense array of 'bufwad' structures, | |
4699 | * while bigobj is a sparse array of the same bufwads. Specifically, | |
4700 | * for any index n, there are three bufwads that should be identical: | |
4701 | * | |
4702 | * packobj, at offset n * sizeof (bufwad_t) | |
4703 | * bigobj, at the head of the nth chunk | |
4704 | * bigobj, at the tail of the nth chunk | |
4705 | * | |
4706 | * The chunk size is set equal to bigobj block size so that | |
440a3eb9 | 4707 | * dmu_assign_arcbuf_by_dbuf() can be tested for object updates. |
9babb374 BB |
4708 | */ |
4709 | ||
4710 | /* | |
4711 | * Read the directory info. If it's the first time, set things up. | |
4712 | */ | |
50c957f7 NB |
4713 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0, 0); |
4714 | ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, 0, | |
4715 | chunksize); | |
40b84e7a | 4716 | |
9babb374 | 4717 | |
40b84e7a BB |
4718 | if (ztest_object_init(zd, od, size, B_FALSE) != 0) { |
4719 | umem_free(od, size); | |
428870ff | 4720 | return; |
40b84e7a | 4721 | } |
9babb374 | 4722 | |
428870ff BB |
4723 | bigobj = od[0].od_object; |
4724 | packobj = od[1].od_object; | |
4725 | blocksize = od[0].od_blocksize; | |
4726 | chunksize = blocksize; | |
4727 | ASSERT(chunksize == od[1].od_gen); | |
9babb374 | 4728 | |
428870ff BB |
4729 | VERIFY(dmu_object_info(os, bigobj, &doi) == 0); |
4730 | VERIFY(ISP2(doi.doi_data_block_size)); | |
4731 | VERIFY(chunksize == doi.doi_data_block_size); | |
4732 | VERIFY(chunksize >= 2 * sizeof (bufwad_t)); | |
9babb374 BB |
4733 | |
4734 | /* | |
4735 | * Pick a random index and compute the offsets into packobj and bigobj. | |
4736 | */ | |
4737 | n = ztest_random(regions) * stride + ztest_random(width); | |
4738 | s = 1 + ztest_random(width - 1); | |
4739 | ||
4740 | packoff = n * sizeof (bufwad_t); | |
4741 | packsize = s * sizeof (bufwad_t); | |
4742 | ||
428870ff BB |
4743 | bigoff = n * chunksize; |
4744 | bigsize = s * chunksize; | |
9babb374 BB |
4745 | |
4746 | packbuf = umem_zalloc(packsize, UMEM_NOFAIL); | |
4747 | bigbuf = umem_zalloc(bigsize, UMEM_NOFAIL); | |
4748 | ||
428870ff | 4749 | VERIFY3U(0, ==, dmu_bonus_hold(os, bigobj, FTAG, &bonus_db)); |
9babb374 BB |
4750 | |
4751 | bigbuf_arcbufs = umem_zalloc(2 * s * sizeof (arc_buf_t *), UMEM_NOFAIL); | |
4752 | ||
4753 | /* | |
4754 | * Iteration 0 test zcopy for DB_UNCACHED dbufs. | |
4755 | * Iteration 1 test zcopy to already referenced dbufs. | |
4756 | * Iteration 2 test zcopy to dirty dbuf in the same txg. | |
4757 | * Iteration 3 test zcopy to dbuf dirty in previous txg. | |
4758 | * Iteration 4 test zcopy when dbuf is no longer dirty. | |
4759 | * Iteration 5 test zcopy when it can't be done. | |
4760 | * Iteration 6 one more zcopy write. | |
4761 | */ | |
4762 | for (i = 0; i < 7; i++) { | |
4763 | uint64_t j; | |
4764 | uint64_t off; | |
4765 | ||
4766 | /* | |
4767 | * In iteration 5 (i == 5) use arcbufs | |
4768 | * that don't match bigobj blksz to test | |
440a3eb9 | 4769 | * dmu_assign_arcbuf_by_dbuf() when it can't directly |
9babb374 BB |
4770 | * assign an arcbuf to a dbuf. |
4771 | */ | |
4772 | for (j = 0; j < s; j++) { | |
b9541d6b | 4773 | if (i != 5 || chunksize < (SPA_MINBLOCKSIZE * 2)) { |
9babb374 | 4774 | bigbuf_arcbufs[j] = |
428870ff | 4775 | dmu_request_arcbuf(bonus_db, chunksize); |
9babb374 BB |
4776 | } else { |
4777 | bigbuf_arcbufs[2 * j] = | |
428870ff | 4778 | dmu_request_arcbuf(bonus_db, chunksize / 2); |
9babb374 | 4779 | bigbuf_arcbufs[2 * j + 1] = |
428870ff | 4780 | dmu_request_arcbuf(bonus_db, chunksize / 2); |
9babb374 BB |
4781 | } |
4782 | } | |
4783 | ||
4784 | /* | |
4785 | * Get a tx for the mods to both packobj and bigobj. | |
4786 | */ | |
4787 | tx = dmu_tx_create(os); | |
4788 | ||
428870ff BB |
4789 | dmu_tx_hold_write(tx, packobj, packoff, packsize); |
4790 | dmu_tx_hold_write(tx, bigobj, bigoff, bigsize); | |
9babb374 | 4791 | |
428870ff BB |
4792 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); |
4793 | if (txg == 0) { | |
9babb374 BB |
4794 | umem_free(packbuf, packsize); |
4795 | umem_free(bigbuf, bigsize); | |
4796 | for (j = 0; j < s; j++) { | |
b9541d6b CW |
4797 | if (i != 5 || |
4798 | chunksize < (SPA_MINBLOCKSIZE * 2)) { | |
9babb374 BB |
4799 | dmu_return_arcbuf(bigbuf_arcbufs[j]); |
4800 | } else { | |
4801 | dmu_return_arcbuf( | |
4802 | bigbuf_arcbufs[2 * j]); | |
4803 | dmu_return_arcbuf( | |
4804 | bigbuf_arcbufs[2 * j + 1]); | |
4805 | } | |
4806 | } | |
4807 | umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *)); | |
40b84e7a | 4808 | umem_free(od, size); |
9babb374 BB |
4809 | dmu_buf_rele(bonus_db, FTAG); |
4810 | return; | |
4811 | } | |
4812 | ||
9babb374 BB |
4813 | /* |
4814 | * 50% of the time don't read objects in the 1st iteration to | |
440a3eb9 TC |
4815 | * test dmu_assign_arcbuf_by_dbuf() for the case when there are |
4816 | * no existing dbufs for the specified offsets. | |
9babb374 BB |
4817 | */ |
4818 | if (i != 0 || ztest_random(2) != 0) { | |
428870ff | 4819 | error = dmu_read(os, packobj, packoff, |
9babb374 | 4820 | packsize, packbuf, DMU_READ_PREFETCH); |
c99c9001 | 4821 | ASSERT0(error); |
428870ff | 4822 | error = dmu_read(os, bigobj, bigoff, bigsize, |
9babb374 | 4823 | bigbuf, DMU_READ_PREFETCH); |
c99c9001 | 4824 | ASSERT0(error); |
9babb374 BB |
4825 | } |
4826 | compare_and_update_pbbufs(s, packbuf, bigbuf, bigsize, | |
428870ff | 4827 | n, chunksize, txg); |
9babb374 BB |
4828 | |
4829 | /* | |
4830 | * We've verified all the old bufwads, and made new ones. | |
4831 | * Now write them out. | |
4832 | */ | |
428870ff | 4833 | dmu_write(os, packobj, packoff, packsize, packbuf, tx); |
c242c188 | 4834 | if (ztest_opts.zo_verbose >= 7) { |
9babb374 BB |
4835 | (void) printf("writing offset %llx size %llx" |
4836 | " txg %llx\n", | |
4837 | (u_longlong_t)bigoff, | |
4838 | (u_longlong_t)bigsize, | |
4839 | (u_longlong_t)txg); | |
4840 | } | |
428870ff | 4841 | for (off = bigoff, j = 0; j < s; j++, off += chunksize) { |
9babb374 | 4842 | dmu_buf_t *dbt; |
b9541d6b | 4843 | if (i != 5 || chunksize < (SPA_MINBLOCKSIZE * 2)) { |
9babb374 | 4844 | bcopy((caddr_t)bigbuf + (off - bigoff), |
428870ff | 4845 | bigbuf_arcbufs[j]->b_data, chunksize); |
9babb374 BB |
4846 | } else { |
4847 | bcopy((caddr_t)bigbuf + (off - bigoff), | |
4848 | bigbuf_arcbufs[2 * j]->b_data, | |
428870ff | 4849 | chunksize / 2); |
9babb374 | 4850 | bcopy((caddr_t)bigbuf + (off - bigoff) + |
428870ff | 4851 | chunksize / 2, |
9babb374 | 4852 | bigbuf_arcbufs[2 * j + 1]->b_data, |
428870ff | 4853 | chunksize / 2); |
9babb374 BB |
4854 | } |
4855 | ||
4856 | if (i == 1) { | |
428870ff BB |
4857 | VERIFY(dmu_buf_hold(os, bigobj, off, |
4858 | FTAG, &dbt, DMU_READ_NO_PREFETCH) == 0); | |
9babb374 | 4859 | } |
b9541d6b | 4860 | if (i != 5 || chunksize < (SPA_MINBLOCKSIZE * 2)) { |
305781da TC |
4861 | VERIFY0(dmu_assign_arcbuf_by_dbuf(bonus_db, |
4862 | off, bigbuf_arcbufs[j], tx)); | |
9babb374 | 4863 | } else { |
305781da TC |
4864 | VERIFY0(dmu_assign_arcbuf_by_dbuf(bonus_db, |
4865 | off, bigbuf_arcbufs[2 * j], tx)); | |
4866 | VERIFY0(dmu_assign_arcbuf_by_dbuf(bonus_db, | |
428870ff | 4867 | off + chunksize / 2, |
305781da | 4868 | bigbuf_arcbufs[2 * j + 1], tx)); |
9babb374 BB |
4869 | } |
4870 | if (i == 1) { | |
4871 | dmu_buf_rele(dbt, FTAG); | |
4872 | } | |
4873 | } | |
4874 | dmu_tx_commit(tx); | |
4875 | ||
4876 | /* | |
4877 | * Sanity check the stuff we just wrote. | |
4878 | */ | |
4879 | { | |
4880 | void *packcheck = umem_alloc(packsize, UMEM_NOFAIL); | |
4881 | void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL); | |
4882 | ||
428870ff | 4883 | VERIFY(0 == dmu_read(os, packobj, packoff, |
9babb374 | 4884 | packsize, packcheck, DMU_READ_PREFETCH)); |
428870ff | 4885 | VERIFY(0 == dmu_read(os, bigobj, bigoff, |
9babb374 BB |
4886 | bigsize, bigcheck, DMU_READ_PREFETCH)); |
4887 | ||
4888 | ASSERT(bcmp(packbuf, packcheck, packsize) == 0); | |
4889 | ASSERT(bcmp(bigbuf, bigcheck, bigsize) == 0); | |
4890 | ||
4891 | umem_free(packcheck, packsize); | |
4892 | umem_free(bigcheck, bigsize); | |
4893 | } | |
4894 | if (i == 2) { | |
1b939560 | 4895 | txg_wait_open(dmu_objset_pool(os), 0, B_TRUE); |
9babb374 BB |
4896 | } else if (i == 3) { |
4897 | txg_wait_synced(dmu_objset_pool(os), 0); | |
4898 | } | |
4899 | } | |
4900 | ||
4901 | dmu_buf_rele(bonus_db, FTAG); | |
4902 | umem_free(packbuf, packsize); | |
4903 | umem_free(bigbuf, bigsize); | |
4904 | umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *)); | |
40b84e7a | 4905 | umem_free(od, size); |
9babb374 BB |
4906 | } |
4907 | ||
428870ff | 4908 | /* ARGSUSED */ |
34dc7c2f | 4909 | void |
428870ff | 4910 | ztest_dmu_write_parallel(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4911 | { |
40b84e7a BB |
4912 | ztest_od_t *od; |
4913 | ||
d1d7e268 | 4914 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
428870ff BB |
4915 | uint64_t offset = (1ULL << (ztest_random(20) + 43)) + |
4916 | (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
34dc7c2f BB |
4917 | |
4918 | /* | |
428870ff BB |
4919 | * Have multiple threads write to large offsets in an object |
4920 | * to verify that parallel writes to an object -- even to the | |
4921 | * same blocks within the object -- doesn't cause any trouble. | |
34dc7c2f | 4922 | */ |
50c957f7 | 4923 | ztest_od_init(od, ID_PARALLEL, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0); |
9babb374 | 4924 | |
40b84e7a | 4925 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) |
34dc7c2f | 4926 | return; |
34dc7c2f | 4927 | |
428870ff | 4928 | while (ztest_random(10) != 0) |
40b84e7a BB |
4929 | ztest_io(zd, od->od_object, offset); |
4930 | ||
d1d7e268 | 4931 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 4932 | } |
34dc7c2f | 4933 | |
428870ff BB |
4934 | void |
4935 | ztest_dmu_prealloc(ztest_ds_t *zd, uint64_t id) | |
4936 | { | |
40b84e7a | 4937 | ztest_od_t *od; |
428870ff BB |
4938 | uint64_t offset = (1ULL << (ztest_random(4) + SPA_MAXBLOCKSHIFT)) + |
4939 | (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
4940 | uint64_t count = ztest_random(20) + 1; | |
4941 | uint64_t blocksize = ztest_random_blocksize(); | |
4942 | void *data; | |
34dc7c2f | 4943 | |
d1d7e268 | 4944 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
34dc7c2f | 4945 | |
50c957f7 | 4946 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0, 0); |
40b84e7a | 4947 | |
d1d7e268 MK |
4948 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
4949 | !ztest_random(2)) != 0) { | |
4950 | umem_free(od, sizeof (ztest_od_t)); | |
34dc7c2f | 4951 | return; |
40b84e7a | 4952 | } |
34dc7c2f | 4953 | |
40b84e7a | 4954 | if (ztest_truncate(zd, od->od_object, offset, count * blocksize) != 0) { |
d1d7e268 | 4955 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f | 4956 | return; |
40b84e7a | 4957 | } |
34dc7c2f | 4958 | |
40b84e7a | 4959 | ztest_prealloc(zd, od->od_object, offset, count * blocksize); |
34dc7c2f | 4960 | |
428870ff | 4961 | data = umem_zalloc(blocksize, UMEM_NOFAIL); |
34dc7c2f | 4962 | |
428870ff BB |
4963 | while (ztest_random(count) != 0) { |
4964 | uint64_t randoff = offset + (ztest_random(count) * blocksize); | |
40b84e7a | 4965 | if (ztest_write(zd, od->od_object, randoff, blocksize, |
428870ff BB |
4966 | data) != 0) |
4967 | break; | |
4968 | while (ztest_random(4) != 0) | |
40b84e7a | 4969 | ztest_io(zd, od->od_object, randoff); |
9babb374 | 4970 | } |
34dc7c2f | 4971 | |
428870ff | 4972 | umem_free(data, blocksize); |
d1d7e268 | 4973 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f BB |
4974 | } |
4975 | ||
4976 | /* | |
4977 | * Verify that zap_{create,destroy,add,remove,update} work as expected. | |
4978 | */ | |
4979 | #define ZTEST_ZAP_MIN_INTS 1 | |
4980 | #define ZTEST_ZAP_MAX_INTS 4 | |
4981 | #define ZTEST_ZAP_MAX_PROPS 1000 | |
4982 | ||
4983 | void | |
428870ff | 4984 | ztest_zap(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4985 | { |
428870ff | 4986 | objset_t *os = zd->zd_os; |
40b84e7a | 4987 | ztest_od_t *od; |
34dc7c2f BB |
4988 | uint64_t object; |
4989 | uint64_t txg, last_txg; | |
4990 | uint64_t value[ZTEST_ZAP_MAX_INTS]; | |
4991 | uint64_t zl_ints, zl_intsize, prop; | |
4992 | int i, ints; | |
4993 | dmu_tx_t *tx; | |
4994 | char propname[100], txgname[100]; | |
4995 | int error; | |
34dc7c2f BB |
4996 | char *hc[2] = { "s.acl.h", ".s.open.h.hyLZlg" }; |
4997 | ||
d1d7e268 | 4998 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 4999 | ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0, 0); |
34dc7c2f | 5000 | |
40b84e7a | 5001 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
02730c33 | 5002 | !ztest_random(2)) != 0) |
40b84e7a | 5003 | goto out; |
34dc7c2f | 5004 | |
40b84e7a | 5005 | object = od->od_object; |
34dc7c2f | 5006 | |
428870ff BB |
5007 | /* |
5008 | * Generate a known hash collision, and verify that | |
5009 | * we can lookup and remove both entries. | |
5010 | */ | |
5011 | tx = dmu_tx_create(os); | |
5012 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); | |
5013 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
5014 | if (txg == 0) | |
40b84e7a | 5015 | goto out; |
428870ff BB |
5016 | for (i = 0; i < 2; i++) { |
5017 | value[i] = i; | |
5018 | VERIFY3U(0, ==, zap_add(os, object, hc[i], sizeof (uint64_t), | |
5019 | 1, &value[i], tx)); | |
5020 | } | |
5021 | for (i = 0; i < 2; i++) { | |
5022 | VERIFY3U(EEXIST, ==, zap_add(os, object, hc[i], | |
5023 | sizeof (uint64_t), 1, &value[i], tx)); | |
5024 | VERIFY3U(0, ==, | |
5025 | zap_length(os, object, hc[i], &zl_intsize, &zl_ints)); | |
5026 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
5027 | ASSERT3U(zl_ints, ==, 1); | |
5028 | } | |
5029 | for (i = 0; i < 2; i++) { | |
5030 | VERIFY3U(0, ==, zap_remove(os, object, hc[i], tx)); | |
34dc7c2f | 5031 | } |
428870ff | 5032 | dmu_tx_commit(tx); |
34dc7c2f | 5033 | |
428870ff | 5034 | /* |
d611989f | 5035 | * Generate a bunch of random entries. |
428870ff | 5036 | */ |
34dc7c2f BB |
5037 | ints = MAX(ZTEST_ZAP_MIN_INTS, object % ZTEST_ZAP_MAX_INTS); |
5038 | ||
5039 | prop = ztest_random(ZTEST_ZAP_MAX_PROPS); | |
5040 | (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop); | |
5041 | (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop); | |
5042 | bzero(value, sizeof (value)); | |
5043 | last_txg = 0; | |
5044 | ||
5045 | /* | |
5046 | * If these zap entries already exist, validate their contents. | |
5047 | */ | |
5048 | error = zap_length(os, object, txgname, &zl_intsize, &zl_ints); | |
5049 | if (error == 0) { | |
5050 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
5051 | ASSERT3U(zl_ints, ==, 1); | |
5052 | ||
5053 | VERIFY(zap_lookup(os, object, txgname, zl_intsize, | |
5054 | zl_ints, &last_txg) == 0); | |
5055 | ||
5056 | VERIFY(zap_length(os, object, propname, &zl_intsize, | |
5057 | &zl_ints) == 0); | |
5058 | ||
5059 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
5060 | ASSERT3U(zl_ints, ==, ints); | |
5061 | ||
5062 | VERIFY(zap_lookup(os, object, propname, zl_intsize, | |
5063 | zl_ints, value) == 0); | |
5064 | ||
5065 | for (i = 0; i < ints; i++) { | |
5066 | ASSERT3U(value[i], ==, last_txg + object + i); | |
5067 | } | |
5068 | } else { | |
5069 | ASSERT3U(error, ==, ENOENT); | |
5070 | } | |
5071 | ||
5072 | /* | |
5073 | * Atomically update two entries in our zap object. | |
5074 | * The first is named txg_%llu, and contains the txg | |
5075 | * in which the property was last updated. The second | |
5076 | * is named prop_%llu, and the nth element of its value | |
5077 | * should be txg + object + n. | |
5078 | */ | |
5079 | tx = dmu_tx_create(os); | |
428870ff BB |
5080 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
5081 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
5082 | if (txg == 0) | |
40b84e7a | 5083 | goto out; |
34dc7c2f BB |
5084 | |
5085 | if (last_txg > txg) | |
5086 | fatal(0, "zap future leak: old %llu new %llu", last_txg, txg); | |
5087 | ||
5088 | for (i = 0; i < ints; i++) | |
5089 | value[i] = txg + object + i; | |
5090 | ||
428870ff BB |
5091 | VERIFY3U(0, ==, zap_update(os, object, txgname, sizeof (uint64_t), |
5092 | 1, &txg, tx)); | |
5093 | VERIFY3U(0, ==, zap_update(os, object, propname, sizeof (uint64_t), | |
5094 | ints, value, tx)); | |
34dc7c2f BB |
5095 | |
5096 | dmu_tx_commit(tx); | |
5097 | ||
5098 | /* | |
5099 | * Remove a random pair of entries. | |
5100 | */ | |
5101 | prop = ztest_random(ZTEST_ZAP_MAX_PROPS); | |
5102 | (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop); | |
5103 | (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop); | |
5104 | ||
5105 | error = zap_length(os, object, txgname, &zl_intsize, &zl_ints); | |
5106 | ||
5107 | if (error == ENOENT) | |
40b84e7a | 5108 | goto out; |
34dc7c2f | 5109 | |
c99c9001 | 5110 | ASSERT0(error); |
34dc7c2f BB |
5111 | |
5112 | tx = dmu_tx_create(os); | |
428870ff BB |
5113 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
5114 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
5115 | if (txg == 0) | |
40b84e7a | 5116 | goto out; |
428870ff BB |
5117 | VERIFY3U(0, ==, zap_remove(os, object, txgname, tx)); |
5118 | VERIFY3U(0, ==, zap_remove(os, object, propname, tx)); | |
5119 | dmu_tx_commit(tx); | |
40b84e7a | 5120 | out: |
d1d7e268 | 5121 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 5122 | } |
34dc7c2f | 5123 | |
428870ff | 5124 | /* |
d611989f | 5125 | * Test case to test the upgrading of a microzap to fatzap. |
428870ff BB |
5126 | */ |
5127 | void | |
5128 | ztest_fzap(ztest_ds_t *zd, uint64_t id) | |
5129 | { | |
5130 | objset_t *os = zd->zd_os; | |
40b84e7a | 5131 | ztest_od_t *od; |
428870ff | 5132 | uint64_t object, txg; |
d6320ddb | 5133 | int i; |
34dc7c2f | 5134 | |
d1d7e268 | 5135 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 5136 | ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0, 0); |
428870ff | 5137 | |
40b84e7a | 5138 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
02730c33 | 5139 | !ztest_random(2)) != 0) |
40b84e7a BB |
5140 | goto out; |
5141 | object = od->od_object; | |
34dc7c2f BB |
5142 | |
5143 | /* | |
428870ff BB |
5144 | * Add entries to this ZAP and make sure it spills over |
5145 | * and gets upgraded to a fatzap. Also, since we are adding | |
5146 | * 2050 entries we should see ptrtbl growth and leaf-block split. | |
34dc7c2f | 5147 | */ |
d6320ddb | 5148 | for (i = 0; i < 2050; i++) { |
eca7b760 | 5149 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
5150 | uint64_t value = i; |
5151 | dmu_tx_t *tx; | |
5152 | int error; | |
34dc7c2f | 5153 | |
428870ff | 5154 | (void) snprintf(name, sizeof (name), "fzap-%llu-%llu", |
b8864a23 | 5155 | (u_longlong_t)id, (u_longlong_t)value); |
428870ff BB |
5156 | |
5157 | tx = dmu_tx_create(os); | |
5158 | dmu_tx_hold_zap(tx, object, B_TRUE, name); | |
5159 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
5160 | if (txg == 0) | |
40b84e7a | 5161 | goto out; |
428870ff BB |
5162 | error = zap_add(os, object, name, sizeof (uint64_t), 1, |
5163 | &value, tx); | |
5164 | ASSERT(error == 0 || error == EEXIST); | |
5165 | dmu_tx_commit(tx); | |
34dc7c2f | 5166 | } |
40b84e7a | 5167 | out: |
d1d7e268 | 5168 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f BB |
5169 | } |
5170 | ||
428870ff | 5171 | /* ARGSUSED */ |
34dc7c2f | 5172 | void |
428870ff | 5173 | ztest_zap_parallel(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 5174 | { |
428870ff | 5175 | objset_t *os = zd->zd_os; |
40b84e7a | 5176 | ztest_od_t *od; |
34dc7c2f BB |
5177 | uint64_t txg, object, count, wsize, wc, zl_wsize, zl_wc; |
5178 | dmu_tx_t *tx; | |
5179 | int i, namelen, error; | |
428870ff | 5180 | int micro = ztest_random(2); |
34dc7c2f BB |
5181 | char name[20], string_value[20]; |
5182 | void *data; | |
5183 | ||
d1d7e268 | 5184 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 5185 | ztest_od_init(od, ID_PARALLEL, FTAG, micro, DMU_OT_ZAP_OTHER, 0, 0, 0); |
428870ff | 5186 | |
40b84e7a | 5187 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
d1d7e268 | 5188 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 5189 | return; |
40b84e7a | 5190 | } |
428870ff | 5191 | |
40b84e7a | 5192 | object = od->od_object; |
428870ff | 5193 | |
34dc7c2f BB |
5194 | /* |
5195 | * Generate a random name of the form 'xxx.....' where each | |
5196 | * x is a random printable character and the dots are dots. | |
5197 | * There are 94 such characters, and the name length goes from | |
5198 | * 6 to 20, so there are 94^3 * 15 = 12,458,760 possible names. | |
5199 | */ | |
5200 | namelen = ztest_random(sizeof (name) - 5) + 5 + 1; | |
5201 | ||
5202 | for (i = 0; i < 3; i++) | |
5203 | name[i] = '!' + ztest_random('~' - '!' + 1); | |
5204 | for (; i < namelen - 1; i++) | |
5205 | name[i] = '.'; | |
5206 | name[i] = '\0'; | |
5207 | ||
428870ff | 5208 | if ((namelen & 1) || micro) { |
34dc7c2f BB |
5209 | wsize = sizeof (txg); |
5210 | wc = 1; | |
5211 | data = &txg; | |
5212 | } else { | |
5213 | wsize = 1; | |
5214 | wc = namelen; | |
5215 | data = string_value; | |
5216 | } | |
5217 | ||
5218 | count = -1ULL; | |
13fe0198 | 5219 | VERIFY0(zap_count(os, object, &count)); |
34dc7c2f BB |
5220 | ASSERT(count != -1ULL); |
5221 | ||
5222 | /* | |
5223 | * Select an operation: length, lookup, add, update, remove. | |
5224 | */ | |
5225 | i = ztest_random(5); | |
5226 | ||
5227 | if (i >= 2) { | |
5228 | tx = dmu_tx_create(os); | |
428870ff BB |
5229 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
5230 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
df053d67 GN |
5231 | if (txg == 0) { |
5232 | umem_free(od, sizeof (ztest_od_t)); | |
34dc7c2f | 5233 | return; |
df053d67 | 5234 | } |
34dc7c2f BB |
5235 | bcopy(name, string_value, namelen); |
5236 | } else { | |
5237 | tx = NULL; | |
5238 | txg = 0; | |
5239 | bzero(string_value, namelen); | |
5240 | } | |
5241 | ||
5242 | switch (i) { | |
5243 | ||
5244 | case 0: | |
5245 | error = zap_length(os, object, name, &zl_wsize, &zl_wc); | |
5246 | if (error == 0) { | |
5247 | ASSERT3U(wsize, ==, zl_wsize); | |
5248 | ASSERT3U(wc, ==, zl_wc); | |
5249 | } else { | |
5250 | ASSERT3U(error, ==, ENOENT); | |
5251 | } | |
5252 | break; | |
5253 | ||
5254 | case 1: | |
5255 | error = zap_lookup(os, object, name, wsize, wc, data); | |
5256 | if (error == 0) { | |
5257 | if (data == string_value && | |
5258 | bcmp(name, data, namelen) != 0) | |
5259 | fatal(0, "name '%s' != val '%s' len %d", | |
5260 | name, data, namelen); | |
5261 | } else { | |
5262 | ASSERT3U(error, ==, ENOENT); | |
5263 | } | |
5264 | break; | |
5265 | ||
5266 | case 2: | |
5267 | error = zap_add(os, object, name, wsize, wc, data, tx); | |
5268 | ASSERT(error == 0 || error == EEXIST); | |
5269 | break; | |
5270 | ||
5271 | case 3: | |
5272 | VERIFY(zap_update(os, object, name, wsize, wc, data, tx) == 0); | |
5273 | break; | |
5274 | ||
5275 | case 4: | |
5276 | error = zap_remove(os, object, name, tx); | |
5277 | ASSERT(error == 0 || error == ENOENT); | |
5278 | break; | |
5279 | } | |
5280 | ||
5281 | if (tx != NULL) | |
5282 | dmu_tx_commit(tx); | |
40b84e7a | 5283 | |
d1d7e268 | 5284 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f BB |
5285 | } |
5286 | ||
428870ff BB |
5287 | /* |
5288 | * Commit callback data. | |
5289 | */ | |
5290 | typedef struct ztest_cb_data { | |
5291 | list_node_t zcd_node; | |
5292 | uint64_t zcd_txg; | |
5293 | int zcd_expected_err; | |
5294 | boolean_t zcd_added; | |
5295 | boolean_t zcd_called; | |
5296 | spa_t *zcd_spa; | |
5297 | } ztest_cb_data_t; | |
5298 | ||
5299 | /* This is the actual commit callback function */ | |
5300 | static void | |
5301 | ztest_commit_callback(void *arg, int error) | |
5302 | { | |
5303 | ztest_cb_data_t *data = arg; | |
5304 | uint64_t synced_txg; | |
5305 | ||
5306 | VERIFY(data != NULL); | |
5307 | VERIFY3S(data->zcd_expected_err, ==, error); | |
5308 | VERIFY(!data->zcd_called); | |
5309 | ||
5310 | synced_txg = spa_last_synced_txg(data->zcd_spa); | |
5311 | if (data->zcd_txg > synced_txg) | |
5312 | fatal(0, "commit callback of txg %" PRIu64 " called prematurely" | |
5313 | ", last synced txg = %" PRIu64 "\n", data->zcd_txg, | |
5314 | synced_txg); | |
5315 | ||
5316 | data->zcd_called = B_TRUE; | |
5317 | ||
5318 | if (error == ECANCELED) { | |
c99c9001 | 5319 | ASSERT0(data->zcd_txg); |
428870ff BB |
5320 | ASSERT(!data->zcd_added); |
5321 | ||
5322 | /* | |
5323 | * The private callback data should be destroyed here, but | |
5324 | * since we are going to check the zcd_called field after | |
5325 | * dmu_tx_abort(), we will destroy it there. | |
5326 | */ | |
5327 | return; | |
5328 | } | |
5329 | ||
090ff092 | 5330 | ASSERT(data->zcd_added); |
428870ff BB |
5331 | ASSERT3U(data->zcd_txg, !=, 0); |
5332 | ||
1e33ac1e | 5333 | (void) mutex_enter(&zcl.zcl_callbacks_lock); |
090ff092 RC |
5334 | |
5335 | /* See if this cb was called more quickly */ | |
5336 | if ((synced_txg - data->zcd_txg) < zc_min_txg_delay) | |
5337 | zc_min_txg_delay = synced_txg - data->zcd_txg; | |
5338 | ||
5339 | /* Remove our callback from the list */ | |
428870ff | 5340 | list_remove(&zcl.zcl_callbacks, data); |
090ff092 | 5341 | |
1e33ac1e | 5342 | (void) mutex_exit(&zcl.zcl_callbacks_lock); |
428870ff | 5343 | |
428870ff BB |
5344 | umem_free(data, sizeof (ztest_cb_data_t)); |
5345 | } | |
5346 | ||
5347 | /* Allocate and initialize callback data structure */ | |
5348 | static ztest_cb_data_t * | |
5349 | ztest_create_cb_data(objset_t *os, uint64_t txg) | |
5350 | { | |
5351 | ztest_cb_data_t *cb_data; | |
5352 | ||
5353 | cb_data = umem_zalloc(sizeof (ztest_cb_data_t), UMEM_NOFAIL); | |
5354 | ||
5355 | cb_data->zcd_txg = txg; | |
5356 | cb_data->zcd_spa = dmu_objset_spa(os); | |
1e33ac1e | 5357 | list_link_init(&cb_data->zcd_node); |
428870ff BB |
5358 | |
5359 | return (cb_data); | |
5360 | } | |
5361 | ||
428870ff BB |
5362 | /* |
5363 | * Commit callback test. | |
5364 | */ | |
34dc7c2f | 5365 | void |
428870ff BB |
5366 | ztest_dmu_commit_callbacks(ztest_ds_t *zd, uint64_t id) |
5367 | { | |
5368 | objset_t *os = zd->zd_os; | |
40b84e7a | 5369 | ztest_od_t *od; |
428870ff BB |
5370 | dmu_tx_t *tx; |
5371 | ztest_cb_data_t *cb_data[3], *tmp_cb; | |
5372 | uint64_t old_txg, txg; | |
090ff092 | 5373 | int i, error = 0; |
428870ff | 5374 | |
d1d7e268 | 5375 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 5376 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0); |
428870ff | 5377 | |
40b84e7a | 5378 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
d1d7e268 | 5379 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 5380 | return; |
40b84e7a | 5381 | } |
428870ff BB |
5382 | |
5383 | tx = dmu_tx_create(os); | |
5384 | ||
5385 | cb_data[0] = ztest_create_cb_data(os, 0); | |
5386 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[0]); | |
5387 | ||
40b84e7a | 5388 | dmu_tx_hold_write(tx, od->od_object, 0, sizeof (uint64_t)); |
428870ff BB |
5389 | |
5390 | /* Every once in a while, abort the transaction on purpose */ | |
5391 | if (ztest_random(100) == 0) | |
5392 | error = -1; | |
5393 | ||
5394 | if (!error) | |
5395 | error = dmu_tx_assign(tx, TXG_NOWAIT); | |
5396 | ||
5397 | txg = error ? 0 : dmu_tx_get_txg(tx); | |
5398 | ||
5399 | cb_data[0]->zcd_txg = txg; | |
5400 | cb_data[1] = ztest_create_cb_data(os, txg); | |
5401 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[1]); | |
5402 | ||
5403 | if (error) { | |
5404 | /* | |
5405 | * It's not a strict requirement to call the registered | |
5406 | * callbacks from inside dmu_tx_abort(), but that's what | |
5407 | * it's supposed to happen in the current implementation | |
5408 | * so we will check for that. | |
5409 | */ | |
5410 | for (i = 0; i < 2; i++) { | |
5411 | cb_data[i]->zcd_expected_err = ECANCELED; | |
5412 | VERIFY(!cb_data[i]->zcd_called); | |
5413 | } | |
5414 | ||
5415 | dmu_tx_abort(tx); | |
5416 | ||
5417 | for (i = 0; i < 2; i++) { | |
5418 | VERIFY(cb_data[i]->zcd_called); | |
5419 | umem_free(cb_data[i], sizeof (ztest_cb_data_t)); | |
5420 | } | |
5421 | ||
d1d7e268 | 5422 | umem_free(od, sizeof (ztest_od_t)); |
428870ff BB |
5423 | return; |
5424 | } | |
5425 | ||
5426 | cb_data[2] = ztest_create_cb_data(os, txg); | |
5427 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[2]); | |
5428 | ||
5429 | /* | |
5430 | * Read existing data to make sure there isn't a future leak. | |
5431 | */ | |
40b84e7a | 5432 | VERIFY(0 == dmu_read(os, od->od_object, 0, sizeof (uint64_t), |
428870ff BB |
5433 | &old_txg, DMU_READ_PREFETCH)); |
5434 | ||
5435 | if (old_txg > txg) | |
5436 | fatal(0, "future leak: got %" PRIu64 ", open txg is %" PRIu64, | |
5437 | old_txg, txg); | |
5438 | ||
40b84e7a | 5439 | dmu_write(os, od->od_object, 0, sizeof (uint64_t), &txg, tx); |
428870ff | 5440 | |
1e33ac1e | 5441 | (void) mutex_enter(&zcl.zcl_callbacks_lock); |
428870ff BB |
5442 | |
5443 | /* | |
5444 | * Since commit callbacks don't have any ordering requirement and since | |
5445 | * it is theoretically possible for a commit callback to be called | |
5446 | * after an arbitrary amount of time has elapsed since its txg has been | |
5447 | * synced, it is difficult to reliably determine whether a commit | |
5448 | * callback hasn't been called due to high load or due to a flawed | |
5449 | * implementation. | |
5450 | * | |
5451 | * In practice, we will assume that if after a certain number of txgs a | |
5452 | * commit callback hasn't been called, then most likely there's an | |
5453 | * implementation bug.. | |
5454 | */ | |
5455 | tmp_cb = list_head(&zcl.zcl_callbacks); | |
5456 | if (tmp_cb != NULL && | |
090ff092 | 5457 | tmp_cb->zcd_txg + ZTEST_COMMIT_CB_THRESH < txg) { |
428870ff BB |
5458 | fatal(0, "Commit callback threshold exceeded, oldest txg: %" |
5459 | PRIu64 ", open txg: %" PRIu64 "\n", tmp_cb->zcd_txg, txg); | |
5460 | } | |
5461 | ||
5462 | /* | |
5463 | * Let's find the place to insert our callbacks. | |
5464 | * | |
5465 | * Even though the list is ordered by txg, it is possible for the | |
5466 | * insertion point to not be the end because our txg may already be | |
5467 | * quiescing at this point and other callbacks in the open txg | |
5468 | * (from other objsets) may have sneaked in. | |
5469 | */ | |
5470 | tmp_cb = list_tail(&zcl.zcl_callbacks); | |
5471 | while (tmp_cb != NULL && tmp_cb->zcd_txg > txg) | |
5472 | tmp_cb = list_prev(&zcl.zcl_callbacks, tmp_cb); | |
5473 | ||
5474 | /* Add the 3 callbacks to the list */ | |
5475 | for (i = 0; i < 3; i++) { | |
5476 | if (tmp_cb == NULL) | |
5477 | list_insert_head(&zcl.zcl_callbacks, cb_data[i]); | |
5478 | else | |
5479 | list_insert_after(&zcl.zcl_callbacks, tmp_cb, | |
5480 | cb_data[i]); | |
5481 | ||
5482 | cb_data[i]->zcd_added = B_TRUE; | |
5483 | VERIFY(!cb_data[i]->zcd_called); | |
5484 | ||
5485 | tmp_cb = cb_data[i]; | |
5486 | } | |
5487 | ||
090ff092 RC |
5488 | zc_cb_counter += 3; |
5489 | ||
1e33ac1e | 5490 | (void) mutex_exit(&zcl.zcl_callbacks_lock); |
428870ff BB |
5491 | |
5492 | dmu_tx_commit(tx); | |
40b84e7a | 5493 | |
d1d7e268 | 5494 | umem_free(od, sizeof (ztest_od_t)); |
428870ff BB |
5495 | } |
5496 | ||
50c957f7 NB |
5497 | /* |
5498 | * Visit each object in the dataset. Verify that its properties | |
5499 | * are consistent what was stored in the block tag when it was created, | |
5500 | * and that its unused bonus buffer space has not been overwritten. | |
5501 | */ | |
817b1b6e | 5502 | /* ARGSUSED */ |
50c957f7 NB |
5503 | void |
5504 | ztest_verify_dnode_bt(ztest_ds_t *zd, uint64_t id) | |
5505 | { | |
5506 | objset_t *os = zd->zd_os; | |
5507 | uint64_t obj; | |
5508 | int err = 0; | |
5509 | ||
5510 | for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) { | |
5511 | ztest_block_tag_t *bt = NULL; | |
5512 | dmu_object_info_t doi; | |
5513 | dmu_buf_t *db; | |
5514 | ||
06401e42 BB |
5515 | ztest_object_lock(zd, obj, RL_READER); |
5516 | if (dmu_bonus_hold(os, obj, FTAG, &db) != 0) { | |
5517 | ztest_object_unlock(zd, obj); | |
50c957f7 | 5518 | continue; |
06401e42 | 5519 | } |
50c957f7 NB |
5520 | |
5521 | dmu_object_info_from_db(db, &doi); | |
5522 | if (doi.doi_bonus_size >= sizeof (*bt)) | |
5523 | bt = ztest_bt_bonus(db); | |
5524 | ||
5525 | if (bt && bt->bt_magic == BT_MAGIC) { | |
5526 | ztest_bt_verify(bt, os, obj, doi.doi_dnodesize, | |
5527 | bt->bt_offset, bt->bt_gen, bt->bt_txg, | |
5528 | bt->bt_crtxg); | |
5529 | ztest_verify_unused_bonus(db, bt, obj, os, bt->bt_gen); | |
5530 | } | |
5531 | ||
5532 | dmu_buf_rele(db, FTAG); | |
06401e42 | 5533 | ztest_object_unlock(zd, obj); |
50c957f7 NB |
5534 | } |
5535 | } | |
5536 | ||
428870ff BB |
5537 | /* ARGSUSED */ |
5538 | void | |
5539 | ztest_dsl_prop_get_set(ztest_ds_t *zd, uint64_t id) | |
5540 | { | |
5541 | zfs_prop_t proplist[] = { | |
5542 | ZFS_PROP_CHECKSUM, | |
5543 | ZFS_PROP_COMPRESSION, | |
5544 | ZFS_PROP_COPIES, | |
5545 | ZFS_PROP_DEDUP | |
5546 | }; | |
d6320ddb | 5547 | int p; |
428870ff | 5548 | |
2d9142c9 | 5549 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
428870ff | 5550 | |
d6320ddb | 5551 | for (p = 0; p < sizeof (proplist) / sizeof (proplist[0]); p++) |
428870ff BB |
5552 | (void) ztest_dsl_prop_set_uint64(zd->zd_name, proplist[p], |
5553 | ztest_random_dsl_prop(proplist[p]), (int)ztest_random(2)); | |
5554 | ||
76d52067 BB |
5555 | VERIFY0(ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_RECORDSIZE, |
5556 | ztest_random_blocksize(), (int)ztest_random(2))); | |
5557 | ||
2d9142c9 | 5558 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
428870ff BB |
5559 | } |
5560 | ||
5561 | /* ARGSUSED */ | |
5562 | void | |
5563 | ztest_spa_prop_get_set(ztest_ds_t *zd, uint64_t id) | |
5564 | { | |
428870ff BB |
5565 | nvlist_t *props = NULL; |
5566 | ||
2d9142c9 | 5567 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
428870ff | 5568 | |
1b939560 BB |
5569 | (void) ztest_spa_prop_set_uint64(ZPOOL_PROP_AUTOTRIM, ztest_random(2)); |
5570 | ||
c99c9001 | 5571 | VERIFY0(spa_prop_get(ztest_spa, &props)); |
428870ff | 5572 | |
c242c188 | 5573 | if (ztest_opts.zo_verbose >= 6) |
428870ff BB |
5574 | dump_nvlist(props, 4); |
5575 | ||
5576 | nvlist_free(props); | |
5577 | ||
2d9142c9 | 5578 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
428870ff BB |
5579 | } |
5580 | ||
13fe0198 MA |
5581 | static int |
5582 | user_release_one(const char *snapname, const char *holdname) | |
5583 | { | |
5584 | nvlist_t *snaps, *holds; | |
5585 | int error; | |
5586 | ||
5587 | snaps = fnvlist_alloc(); | |
5588 | holds = fnvlist_alloc(); | |
5589 | fnvlist_add_boolean(holds, holdname); | |
5590 | fnvlist_add_nvlist(snaps, snapname, holds); | |
5591 | fnvlist_free(holds); | |
5592 | error = dsl_dataset_user_release(snaps, NULL); | |
5593 | fnvlist_free(snaps); | |
5594 | return (error); | |
5595 | } | |
5596 | ||
428870ff BB |
5597 | /* |
5598 | * Test snapshot hold/release and deferred destroy. | |
5599 | */ | |
5600 | void | |
5601 | ztest_dmu_snapshot_hold(ztest_ds_t *zd, uint64_t id) | |
34dc7c2f | 5602 | { |
34dc7c2f | 5603 | int error; |
428870ff BB |
5604 | objset_t *os = zd->zd_os; |
5605 | objset_t *origin; | |
5606 | char snapname[100]; | |
5607 | char fullname[100]; | |
5608 | char clonename[100]; | |
5609 | char tag[100]; | |
eca7b760 | 5610 | char osname[ZFS_MAX_DATASET_NAME_LEN]; |
13fe0198 | 5611 | nvlist_t *holds; |
34dc7c2f | 5612 | |
2d9142c9 | 5613 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
34dc7c2f BB |
5614 | |
5615 | dmu_objset_name(os, osname); | |
5616 | ||
d1d7e268 MK |
5617 | (void) snprintf(snapname, sizeof (snapname), "sh1_%llu", |
5618 | (u_longlong_t)id); | |
13fe0198 MA |
5619 | (void) snprintf(fullname, sizeof (fullname), "%s@%s", osname, snapname); |
5620 | (void) snprintf(clonename, sizeof (clonename), | |
d1d7e268 MK |
5621 | "%s/ch1_%llu", osname, (u_longlong_t)id); |
5622 | (void) snprintf(tag, sizeof (tag), "tag_%llu", (u_longlong_t)id); | |
428870ff BB |
5623 | |
5624 | /* | |
5625 | * Clean up from any previous run. | |
5626 | */ | |
13fe0198 MA |
5627 | error = dsl_destroy_head(clonename); |
5628 | if (error != ENOENT) | |
5629 | ASSERT0(error); | |
5630 | error = user_release_one(fullname, tag); | |
5631 | if (error != ESRCH && error != ENOENT) | |
5632 | ASSERT0(error); | |
5633 | error = dsl_destroy_snapshot(fullname, B_FALSE); | |
5634 | if (error != ENOENT) | |
5635 | ASSERT0(error); | |
428870ff BB |
5636 | |
5637 | /* | |
5638 | * Create snapshot, clone it, mark snap for deferred destroy, | |
5639 | * destroy clone, verify snap was also destroyed. | |
5640 | */ | |
6f1ffb06 | 5641 | error = dmu_objset_snapshot_one(osname, snapname); |
428870ff BB |
5642 | if (error) { |
5643 | if (error == ENOSPC) { | |
5644 | ztest_record_enospc("dmu_objset_snapshot"); | |
5645 | goto out; | |
34dc7c2f | 5646 | } |
428870ff BB |
5647 | fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error); |
5648 | } | |
34dc7c2f | 5649 | |
13fe0198 | 5650 | error = dmu_objset_clone(clonename, fullname); |
428870ff | 5651 | if (error) { |
34dc7c2f | 5652 | if (error == ENOSPC) { |
428870ff BB |
5653 | ztest_record_enospc("dmu_objset_clone"); |
5654 | goto out; | |
34dc7c2f | 5655 | } |
428870ff BB |
5656 | fatal(0, "dmu_objset_clone(%s) = %d", clonename, error); |
5657 | } | |
34dc7c2f | 5658 | |
13fe0198 | 5659 | error = dsl_destroy_snapshot(fullname, B_TRUE); |
428870ff | 5660 | if (error) { |
13fe0198 | 5661 | fatal(0, "dsl_destroy_snapshot(%s, B_TRUE) = %d", |
428870ff BB |
5662 | fullname, error); |
5663 | } | |
34dc7c2f | 5664 | |
13fe0198 | 5665 | error = dsl_destroy_head(clonename); |
428870ff | 5666 | if (error) |
13fe0198 | 5667 | fatal(0, "dsl_destroy_head(%s) = %d", clonename, error); |
34dc7c2f | 5668 | |
428870ff BB |
5669 | error = dmu_objset_hold(fullname, FTAG, &origin); |
5670 | if (error != ENOENT) | |
5671 | fatal(0, "dmu_objset_hold(%s) = %d", fullname, error); | |
34dc7c2f | 5672 | |
428870ff BB |
5673 | /* |
5674 | * Create snapshot, add temporary hold, verify that we can't | |
5675 | * destroy a held snapshot, mark for deferred destroy, | |
5676 | * release hold, verify snapshot was destroyed. | |
5677 | */ | |
6f1ffb06 | 5678 | error = dmu_objset_snapshot_one(osname, snapname); |
428870ff BB |
5679 | if (error) { |
5680 | if (error == ENOSPC) { | |
5681 | ztest_record_enospc("dmu_objset_snapshot"); | |
5682 | goto out; | |
34dc7c2f | 5683 | } |
428870ff BB |
5684 | fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error); |
5685 | } | |
5686 | ||
13fe0198 MA |
5687 | holds = fnvlist_alloc(); |
5688 | fnvlist_add_string(holds, fullname, tag); | |
5689 | error = dsl_dataset_user_hold(holds, 0, NULL); | |
5690 | fnvlist_free(holds); | |
5691 | ||
9b67f605 MA |
5692 | if (error == ENOSPC) { |
5693 | ztest_record_enospc("dsl_dataset_user_hold"); | |
5694 | goto out; | |
5695 | } else if (error) { | |
5696 | fatal(0, "dsl_dataset_user_hold(%s, %s) = %u", | |
5697 | fullname, tag, error); | |
5698 | } | |
428870ff | 5699 | |
13fe0198 | 5700 | error = dsl_destroy_snapshot(fullname, B_FALSE); |
428870ff | 5701 | if (error != EBUSY) { |
13fe0198 | 5702 | fatal(0, "dsl_destroy_snapshot(%s, B_FALSE) = %d", |
428870ff BB |
5703 | fullname, error); |
5704 | } | |
5705 | ||
13fe0198 | 5706 | error = dsl_destroy_snapshot(fullname, B_TRUE); |
428870ff | 5707 | if (error) { |
13fe0198 | 5708 | fatal(0, "dsl_destroy_snapshot(%s, B_TRUE) = %d", |
428870ff | 5709 | fullname, error); |
34dc7c2f BB |
5710 | } |
5711 | ||
13fe0198 | 5712 | error = user_release_one(fullname, tag); |
428870ff | 5713 | if (error) |
95fd54a1 | 5714 | fatal(0, "user_release_one(%s, %s) = %d", fullname, tag, error); |
428870ff | 5715 | |
13fe0198 | 5716 | VERIFY3U(dmu_objset_hold(fullname, FTAG, &origin), ==, ENOENT); |
428870ff BB |
5717 | |
5718 | out: | |
2d9142c9 | 5719 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
34dc7c2f BB |
5720 | } |
5721 | ||
34dc7c2f BB |
5722 | /* |
5723 | * Inject random faults into the on-disk data. | |
5724 | */ | |
428870ff | 5725 | /* ARGSUSED */ |
34dc7c2f | 5726 | void |
428870ff | 5727 | ztest_fault_inject(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 5728 | { |
428870ff | 5729 | ztest_shared_t *zs = ztest_shared; |
c242c188 | 5730 | spa_t *spa = ztest_spa; |
34dc7c2f BB |
5731 | int fd; |
5732 | uint64_t offset; | |
428870ff | 5733 | uint64_t leaves; |
c5b3a7bb | 5734 | uint64_t bad = 0x1990c0ffeedecadeull; |
34dc7c2f | 5735 | uint64_t top, leaf; |
40b84e7a BB |
5736 | char *path0; |
5737 | char *pathrand; | |
34dc7c2f | 5738 | size_t fsize; |
2014c09f | 5739 | int bshift = SPA_MAXBLOCKSHIFT + 2; |
34dc7c2f | 5740 | int iters = 1000; |
428870ff BB |
5741 | int maxfaults; |
5742 | int mirror_save; | |
b128c09f | 5743 | vdev_t *vd0 = NULL; |
34dc7c2f | 5744 | uint64_t guid0 = 0; |
428870ff BB |
5745 | boolean_t islog = B_FALSE; |
5746 | ||
40b84e7a BB |
5747 | path0 = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
5748 | pathrand = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
5749 | ||
c242c188 | 5750 | mutex_enter(&ztest_vdev_lock); |
9e052db4 MA |
5751 | |
5752 | /* | |
5753 | * Device removal is in progress, fault injection must be disabled | |
5754 | * until it completes and the pool is scrubbed. The fault injection | |
5755 | * strategy for damaging blocks does not take in to account evacuated | |
5756 | * blocks which may have already been damaged. | |
5757 | */ | |
5758 | if (ztest_device_removal_active) { | |
5759 | mutex_exit(&ztest_vdev_lock); | |
5760 | goto out; | |
5761 | } | |
5762 | ||
60b82074 | 5763 | maxfaults = MAXFAULTS(zs); |
c242c188 | 5764 | leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raidz; |
428870ff | 5765 | mirror_save = zs->zs_mirrors; |
c242c188 | 5766 | mutex_exit(&ztest_vdev_lock); |
34dc7c2f | 5767 | |
b128c09f | 5768 | ASSERT(leaves >= 1); |
34dc7c2f | 5769 | |
ce5fb2a7 BB |
5770 | /* |
5771 | * While ztest is running the number of leaves will not change. This | |
5772 | * is critical for the fault injection logic as it determines where | |
5773 | * errors can be safely injected such that they are always repairable. | |
5774 | * | |
5775 | * When restarting ztest a different number of leaves may be requested | |
5776 | * which will shift the regions to be damaged. This is fine as long | |
5777 | * as the pool has been scrubbed prior to using the new mapping. | |
5778 | * Failure to do can result in non-repairable damage being injected. | |
5779 | */ | |
5780 | if (ztest_pool_scrubbed == B_FALSE) | |
5781 | goto out; | |
5782 | ||
621dd7bb GW |
5783 | /* |
5784 | * Grab the name lock as reader. There are some operations | |
5785 | * which don't like to have their vdevs changed while | |
5786 | * they are in progress (i.e. spa_change_guid). Those | |
5787 | * operations will have grabbed the name lock as writer. | |
5788 | */ | |
2d9142c9 | 5789 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
621dd7bb | 5790 | |
34dc7c2f | 5791 | /* |
b128c09f | 5792 | * We need SCL_STATE here because we're going to look at vd0->vdev_tsd. |
34dc7c2f | 5793 | */ |
b128c09f | 5794 | spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); |
34dc7c2f | 5795 | |
b128c09f BB |
5796 | if (ztest_random(2) == 0) { |
5797 | /* | |
428870ff | 5798 | * Inject errors on a normal data device or slog device. |
b128c09f | 5799 | */ |
428870ff BB |
5800 | top = ztest_random_vdev_top(spa, B_TRUE); |
5801 | leaf = ztest_random(leaves) + zs->zs_splits; | |
34dc7c2f | 5802 | |
b128c09f BB |
5803 | /* |
5804 | * Generate paths to the first leaf in this top-level vdev, | |
5805 | * and to the random leaf we selected. We'll induce transient | |
5806 | * write failures and random online/offline activity on leaf 0, | |
5807 | * and we'll write random garbage to the randomly chosen leaf. | |
5808 | */ | |
6aec1cd5 | 5809 | (void) snprintf(path0, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
5810 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
5811 | top * leaves + zs->zs_splits); | |
6aec1cd5 | 5812 | (void) snprintf(pathrand, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
5813 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
5814 | top * leaves + leaf); | |
34dc7c2f | 5815 | |
b128c09f | 5816 | vd0 = vdev_lookup_by_path(spa->spa_root_vdev, path0); |
428870ff BB |
5817 | if (vd0 != NULL && vd0->vdev_top->vdev_islog) |
5818 | islog = B_TRUE; | |
5819 | ||
621dd7bb GW |
5820 | /* |
5821 | * If the top-level vdev needs to be resilvered | |
5822 | * then we only allow faults on the device that is | |
5823 | * resilvering. | |
5824 | */ | |
5825 | if (vd0 != NULL && maxfaults != 1 && | |
5826 | (!vdev_resilver_needed(vd0->vdev_top, NULL, NULL) || | |
5d1f7fb6 | 5827 | vd0->vdev_resilver_txg != 0)) { |
b128c09f BB |
5828 | /* |
5829 | * Make vd0 explicitly claim to be unreadable, | |
5830 | * or unwriteable, or reach behind its back | |
5831 | * and close the underlying fd. We can do this if | |
5832 | * maxfaults == 0 because we'll fail and reexecute, | |
5833 | * and we can do it if maxfaults >= 2 because we'll | |
5834 | * have enough redundancy. If maxfaults == 1, the | |
5835 | * combination of this with injection of random data | |
5836 | * corruption below exceeds the pool's fault tolerance. | |
5837 | */ | |
5838 | vdev_file_t *vf = vd0->vdev_tsd; | |
5839 | ||
a1d477c2 MA |
5840 | zfs_dbgmsg("injecting fault to vdev %llu; maxfaults=%d", |
5841 | (long long)vd0->vdev_id, (int)maxfaults); | |
5842 | ||
b128c09f | 5843 | if (vf != NULL && ztest_random(3) == 0) { |
da92d5cb MM |
5844 | (void) close(vf->vf_file->f_fd); |
5845 | vf->vf_file->f_fd = -1; | |
b128c09f BB |
5846 | } else if (ztest_random(2) == 0) { |
5847 | vd0->vdev_cant_read = B_TRUE; | |
5848 | } else { | |
5849 | vd0->vdev_cant_write = B_TRUE; | |
5850 | } | |
5851 | guid0 = vd0->vdev_guid; | |
5852 | } | |
5853 | } else { | |
5854 | /* | |
5855 | * Inject errors on an l2cache device. | |
5856 | */ | |
5857 | spa_aux_vdev_t *sav = &spa->spa_l2cache; | |
34dc7c2f | 5858 | |
b128c09f BB |
5859 | if (sav->sav_count == 0) { |
5860 | spa_config_exit(spa, SCL_STATE, FTAG); | |
2d9142c9 | 5861 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
40b84e7a | 5862 | goto out; |
b128c09f BB |
5863 | } |
5864 | vd0 = sav->sav_vdevs[ztest_random(sav->sav_count)]; | |
34dc7c2f | 5865 | guid0 = vd0->vdev_guid; |
b128c09f BB |
5866 | (void) strcpy(path0, vd0->vdev_path); |
5867 | (void) strcpy(pathrand, vd0->vdev_path); | |
5868 | ||
5869 | leaf = 0; | |
5870 | leaves = 1; | |
5871 | maxfaults = INT_MAX; /* no limit on cache devices */ | |
34dc7c2f BB |
5872 | } |
5873 | ||
b128c09f | 5874 | spa_config_exit(spa, SCL_STATE, FTAG); |
2d9142c9 | 5875 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
b128c09f | 5876 | |
34dc7c2f | 5877 | /* |
428870ff BB |
5878 | * If we can tolerate two or more faults, or we're dealing |
5879 | * with a slog, randomly online/offline vd0. | |
34dc7c2f | 5880 | */ |
428870ff | 5881 | if ((maxfaults >= 2 || islog) && guid0 != 0) { |
fb5f0bc8 BB |
5882 | if (ztest_random(10) < 6) { |
5883 | int flags = (ztest_random(2) == 0 ? | |
5884 | ZFS_OFFLINE_TEMPORARY : 0); | |
428870ff BB |
5885 | |
5886 | /* | |
5887 | * We have to grab the zs_name_lock as writer to | |
5888 | * prevent a race between offlining a slog and | |
5889 | * destroying a dataset. Offlining the slog will | |
5890 | * grab a reference on the dataset which may cause | |
13fe0198 | 5891 | * dsl_destroy_head() to fail with EBUSY thus |
428870ff BB |
5892 | * leaving the dataset in an inconsistent state. |
5893 | */ | |
5894 | if (islog) | |
2d9142c9 | 5895 | (void) pthread_rwlock_wrlock(&ztest_name_lock); |
428870ff | 5896 | |
fb5f0bc8 | 5897 | VERIFY(vdev_offline(spa, guid0, flags) != EBUSY); |
428870ff BB |
5898 | |
5899 | if (islog) | |
2d9142c9 | 5900 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
fb5f0bc8 | 5901 | } else { |
1eb5bfa3 GW |
5902 | /* |
5903 | * Ideally we would like to be able to randomly | |
5904 | * call vdev_[on|off]line without holding locks | |
5905 | * to force unpredictable failures but the side | |
5906 | * effects of vdev_[on|off]line prevent us from | |
5907 | * doing so. We grab the ztest_vdev_lock here to | |
5908 | * prevent a race between injection testing and | |
5909 | * aux_vdev removal. | |
5910 | */ | |
5911 | mutex_enter(&ztest_vdev_lock); | |
fb5f0bc8 | 5912 | (void) vdev_online(spa, guid0, 0, NULL); |
1eb5bfa3 | 5913 | mutex_exit(&ztest_vdev_lock); |
fb5f0bc8 | 5914 | } |
34dc7c2f BB |
5915 | } |
5916 | ||
428870ff | 5917 | if (maxfaults == 0) |
40b84e7a | 5918 | goto out; |
428870ff | 5919 | |
34dc7c2f BB |
5920 | /* |
5921 | * We have at least single-fault tolerance, so inject data corruption. | |
5922 | */ | |
5923 | fd = open(pathrand, O_RDWR); | |
5924 | ||
d2734cce | 5925 | if (fd == -1) /* we hit a gap in the device namespace */ |
40b84e7a | 5926 | goto out; |
34dc7c2f BB |
5927 | |
5928 | fsize = lseek(fd, 0, SEEK_END); | |
5929 | ||
5930 | while (--iters != 0) { | |
91d88843 MA |
5931 | /* |
5932 | * The offset must be chosen carefully to ensure that | |
5933 | * we do not inject a given logical block with errors | |
5934 | * on two different leaf devices, because ZFS can not | |
5935 | * tolerate that (if maxfaults==1). | |
5936 | * | |
42d8d1d6 SD |
5937 | * To achieve this we divide each leaf device into |
5938 | * chunks of size (# leaves * SPA_MAXBLOCKSIZE * 4). | |
5939 | * Each chunk is further divided into error-injection | |
5940 | * ranges (can accept errors) and clear ranges (we do | |
5941 | * not inject errors in those). Each error-injection | |
5942 | * range can accept errors only for a single leaf vdev. | |
5943 | * Error-injection ranges are separated by clear ranges. | |
91d88843 MA |
5944 | * |
5945 | * For example, with 3 leaves, each chunk looks like: | |
5946 | * 0 to 32M: injection range for leaf 0 | |
42d8d1d6 | 5947 | * 32M to 64M: clear range - no injection allowed |
91d88843 | 5948 | * 64M to 96M: injection range for leaf 1 |
42d8d1d6 | 5949 | * 96M to 128M: clear range - no injection allowed |
91d88843 | 5950 | * 128M to 160M: injection range for leaf 2 |
42d8d1d6 SD |
5951 | * 160M to 192M: clear range - no injection allowed |
5952 | * | |
5953 | * Each clear range must be large enough such that a | |
5954 | * single block cannot straddle it. This way a block | |
5955 | * can't be a target in two different injection ranges | |
5956 | * (on different leaf vdevs). | |
91d88843 | 5957 | */ |
34dc7c2f BB |
5958 | offset = ztest_random(fsize / (leaves << bshift)) * |
5959 | (leaves << bshift) + (leaf << bshift) + | |
5960 | (ztest_random(1ULL << (bshift - 1)) & -8ULL); | |
5961 | ||
2014c09f GM |
5962 | /* |
5963 | * Only allow damage to the labels at one end of the vdev. | |
5964 | * | |
5965 | * If all labels are damaged, the device will be totally | |
5966 | * inaccessible, which will result in loss of data, | |
5967 | * because we also damage (parts of) the other side of | |
5968 | * the mirror/raidz. | |
5969 | * | |
5970 | * Additionally, we will always have both an even and an | |
5971 | * odd label, so that we can handle crashes in the | |
5972 | * middle of vdev_config_sync(). | |
5973 | */ | |
5974 | if ((leaf & 1) == 0 && offset < VDEV_LABEL_START_SIZE) | |
5975 | continue; | |
5976 | ||
5977 | /* | |
5978 | * The two end labels are stored at the "end" of the disk, but | |
5979 | * the end of the disk (vdev_psize) is aligned to | |
5980 | * sizeof (vdev_label_t). | |
5981 | */ | |
5982 | uint64_t psize = P2ALIGN(fsize, sizeof (vdev_label_t)); | |
5983 | if ((leaf & 1) == 1 && | |
5984 | offset + sizeof (bad) > psize - VDEV_LABEL_END_SIZE) | |
34dc7c2f BB |
5985 | continue; |
5986 | ||
c242c188 | 5987 | mutex_enter(&ztest_vdev_lock); |
428870ff | 5988 | if (mirror_save != zs->zs_mirrors) { |
c242c188 | 5989 | mutex_exit(&ztest_vdev_lock); |
428870ff | 5990 | (void) close(fd); |
40b84e7a | 5991 | goto out; |
428870ff | 5992 | } |
34dc7c2f BB |
5993 | |
5994 | if (pwrite(fd, &bad, sizeof (bad), offset) != sizeof (bad)) | |
5995 | fatal(1, "can't inject bad word at 0x%llx in %s", | |
5996 | offset, pathrand); | |
428870ff | 5997 | |
c242c188 | 5998 | mutex_exit(&ztest_vdev_lock); |
428870ff | 5999 | |
c242c188 | 6000 | if (ztest_opts.zo_verbose >= 7) |
428870ff BB |
6001 | (void) printf("injected bad word into %s," |
6002 | " offset 0x%llx\n", pathrand, (u_longlong_t)offset); | |
34dc7c2f BB |
6003 | } |
6004 | ||
6005 | (void) close(fd); | |
40b84e7a BB |
6006 | out: |
6007 | umem_free(path0, MAXPATHLEN); | |
6008 | umem_free(pathrand, MAXPATHLEN); | |
34dc7c2f BB |
6009 | } |
6010 | ||
ce5fb2a7 BB |
6011 | /* |
6012 | * By design ztest will never inject uncorrectable damage in to the pool. | |
6013 | * Issue a scrub, wait for it to complete, and verify there is never any | |
6014 | * any persistent damage. | |
6015 | * | |
6016 | * Only after a full scrub has been completed is it safe to start injecting | |
6017 | * data corruption. See the comment in zfs_fault_inject(). | |
6018 | */ | |
6019 | static int | |
6020 | ztest_scrub_impl(spa_t *spa) | |
6021 | { | |
6022 | int error = spa_scan(spa, POOL_SCAN_SCRUB); | |
6023 | if (error) | |
6024 | return (error); | |
6025 | ||
6026 | while (dsl_scan_scrubbing(spa_get_dsl(spa))) | |
6027 | txg_wait_synced(spa_get_dsl(spa), 0); | |
6028 | ||
6029 | if (spa_get_errlog_size(spa) > 0) | |
6030 | return (ECKSUM); | |
6031 | ||
6032 | ztest_pool_scrubbed = B_TRUE; | |
6033 | ||
6034 | return (0); | |
6035 | } | |
6036 | ||
34dc7c2f | 6037 | /* |
428870ff | 6038 | * Scrub the pool. |
34dc7c2f | 6039 | */ |
428870ff BB |
6040 | /* ARGSUSED */ |
6041 | void | |
6042 | ztest_scrub(ztest_ds_t *zd, uint64_t id) | |
34dc7c2f | 6043 | { |
c242c188 | 6044 | spa_t *spa = ztest_spa; |
ce5fb2a7 | 6045 | int error; |
34dc7c2f | 6046 | |
9e052db4 MA |
6047 | /* |
6048 | * Scrub in progress by device removal. | |
6049 | */ | |
6050 | if (ztest_device_removal_active) | |
6051 | return; | |
6052 | ||
ce5fb2a7 BB |
6053 | /* |
6054 | * Start a scrub, wait a moment, then force a restart. | |
6055 | */ | |
428870ff | 6056 | (void) spa_scan(spa, POOL_SCAN_SCRUB); |
ce5fb2a7 BB |
6057 | (void) poll(NULL, 0, 100); |
6058 | ||
6059 | error = ztest_scrub_impl(spa); | |
6060 | if (error == EBUSY) | |
6061 | error = 0; | |
6062 | ASSERT0(error); | |
428870ff | 6063 | } |
34dc7c2f | 6064 | |
3541dc6d GA |
6065 | /* |
6066 | * Change the guid for the pool. | |
6067 | */ | |
6068 | /* ARGSUSED */ | |
6069 | void | |
6070 | ztest_reguid(ztest_ds_t *zd, uint64_t id) | |
6071 | { | |
c242c188 | 6072 | spa_t *spa = ztest_spa; |
3541dc6d | 6073 | uint64_t orig, load; |
3bc7e0fb | 6074 | int error; |
3541dc6d | 6075 | |
b33d668d OF |
6076 | if (ztest_opts.zo_mmp_test) |
6077 | return; | |
6078 | ||
3541dc6d GA |
6079 | orig = spa_guid(spa); |
6080 | load = spa_load_guid(spa); | |
3bc7e0fb | 6081 | |
2d9142c9 | 6082 | (void) pthread_rwlock_wrlock(&ztest_name_lock); |
3bc7e0fb | 6083 | error = spa_change_guid(spa); |
2d9142c9 | 6084 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
3bc7e0fb GW |
6085 | |
6086 | if (error != 0) | |
3541dc6d GA |
6087 | return; |
6088 | ||
ea0b2538 | 6089 | if (ztest_opts.zo_verbose >= 4) { |
3541dc6d GA |
6090 | (void) printf("Changed guid old %llu -> %llu\n", |
6091 | (u_longlong_t)orig, (u_longlong_t)spa_guid(spa)); | |
6092 | } | |
6093 | ||
6094 | VERIFY3U(orig, !=, spa_guid(spa)); | |
6095 | VERIFY3U(load, ==, spa_load_guid(spa)); | |
6096 | } | |
6097 | ||
1eeb4562 JX |
6098 | void |
6099 | ztest_fletcher(ztest_ds_t *zd, uint64_t id) | |
6100 | { | |
6101 | hrtime_t end = gethrtime() + NANOSEC; | |
6102 | ||
6103 | while (gethrtime() <= end) { | |
6104 | int run_count = 100; | |
6105 | void *buf; | |
2fe36b0b | 6106 | struct abd *abd_data, *abd_meta; |
1eeb4562 JX |
6107 | uint32_t size; |
6108 | int *ptr; | |
6109 | int i; | |
6110 | zio_cksum_t zc_ref; | |
6111 | zio_cksum_t zc_ref_byteswap; | |
6112 | ||
6113 | size = ztest_random_blocksize(); | |
2fe36b0b | 6114 | |
1eeb4562 | 6115 | buf = umem_alloc(size, UMEM_NOFAIL); |
2fe36b0b DQ |
6116 | abd_data = abd_alloc(size, B_FALSE); |
6117 | abd_meta = abd_alloc(size, B_TRUE); | |
1eeb4562 JX |
6118 | |
6119 | for (i = 0, ptr = buf; i < size / sizeof (*ptr); i++, ptr++) | |
6120 | *ptr = ztest_random(UINT_MAX); | |
6121 | ||
2fe36b0b DQ |
6122 | abd_copy_from_buf_off(abd_data, buf, 0, size); |
6123 | abd_copy_from_buf_off(abd_meta, buf, 0, size); | |
6124 | ||
1eeb4562 | 6125 | VERIFY0(fletcher_4_impl_set("scalar")); |
3c67d83a TH |
6126 | fletcher_4_native(buf, size, NULL, &zc_ref); |
6127 | fletcher_4_byteswap(buf, size, NULL, &zc_ref_byteswap); | |
1eeb4562 JX |
6128 | |
6129 | VERIFY0(fletcher_4_impl_set("cycle")); | |
6130 | while (run_count-- > 0) { | |
6131 | zio_cksum_t zc; | |
6132 | zio_cksum_t zc_byteswap; | |
6133 | ||
3c67d83a TH |
6134 | fletcher_4_byteswap(buf, size, NULL, &zc_byteswap); |
6135 | fletcher_4_native(buf, size, NULL, &zc); | |
1eeb4562 JX |
6136 | |
6137 | VERIFY0(bcmp(&zc, &zc_ref, sizeof (zc))); | |
6138 | VERIFY0(bcmp(&zc_byteswap, &zc_ref_byteswap, | |
6139 | sizeof (zc_byteswap))); | |
2fe36b0b DQ |
6140 | |
6141 | /* Test ABD - data */ | |
6142 | abd_fletcher_4_byteswap(abd_data, size, NULL, | |
6143 | &zc_byteswap); | |
6144 | abd_fletcher_4_native(abd_data, size, NULL, &zc); | |
6145 | ||
6146 | VERIFY0(bcmp(&zc, &zc_ref, sizeof (zc))); | |
6147 | VERIFY0(bcmp(&zc_byteswap, &zc_ref_byteswap, | |
6148 | sizeof (zc_byteswap))); | |
6149 | ||
6150 | /* Test ABD - metadata */ | |
6151 | abd_fletcher_4_byteswap(abd_meta, size, NULL, | |
6152 | &zc_byteswap); | |
6153 | abd_fletcher_4_native(abd_meta, size, NULL, &zc); | |
6154 | ||
6155 | VERIFY0(bcmp(&zc, &zc_ref, sizeof (zc))); | |
6156 | VERIFY0(bcmp(&zc_byteswap, &zc_ref_byteswap, | |
6157 | sizeof (zc_byteswap))); | |
6158 | ||
1eeb4562 JX |
6159 | } |
6160 | ||
6161 | umem_free(buf, size); | |
2fe36b0b DQ |
6162 | abd_free(abd_data); |
6163 | abd_free(abd_meta); | |
1eeb4562 JX |
6164 | } |
6165 | } | |
6166 | ||
37f520db GN |
6167 | void |
6168 | ztest_fletcher_incr(ztest_ds_t *zd, uint64_t id) | |
6169 | { | |
6170 | void *buf; | |
6171 | size_t size; | |
6172 | int *ptr; | |
6173 | int i; | |
6174 | zio_cksum_t zc_ref; | |
6175 | zio_cksum_t zc_ref_bswap; | |
6176 | ||
6177 | hrtime_t end = gethrtime() + NANOSEC; | |
6178 | ||
6179 | while (gethrtime() <= end) { | |
6180 | int run_count = 100; | |
6181 | ||
6182 | size = ztest_random_blocksize(); | |
6183 | buf = umem_alloc(size, UMEM_NOFAIL); | |
6184 | ||
6185 | for (i = 0, ptr = buf; i < size / sizeof (*ptr); i++, ptr++) | |
6186 | *ptr = ztest_random(UINT_MAX); | |
6187 | ||
6188 | VERIFY0(fletcher_4_impl_set("scalar")); | |
6189 | fletcher_4_native(buf, size, NULL, &zc_ref); | |
6190 | fletcher_4_byteswap(buf, size, NULL, &zc_ref_bswap); | |
6191 | ||
6192 | VERIFY0(fletcher_4_impl_set("cycle")); | |
6193 | ||
6194 | while (run_count-- > 0) { | |
6195 | zio_cksum_t zc; | |
6196 | zio_cksum_t zc_bswap; | |
6197 | size_t pos = 0; | |
6198 | ||
6199 | ZIO_SET_CHECKSUM(&zc, 0, 0, 0, 0); | |
6200 | ZIO_SET_CHECKSUM(&zc_bswap, 0, 0, 0, 0); | |
6201 | ||
6202 | while (pos < size) { | |
6203 | size_t inc = 64 * ztest_random(size / 67); | |
6204 | /* sometimes add few bytes to test non-simd */ | |
6205 | if (ztest_random(100) < 10) | |
6206 | inc += P2ALIGN(ztest_random(64), | |
6207 | sizeof (uint32_t)); | |
6208 | ||
6209 | if (inc > (size - pos)) | |
6210 | inc = size - pos; | |
6211 | ||
6212 | fletcher_4_incremental_native(buf + pos, inc, | |
6213 | &zc); | |
6214 | fletcher_4_incremental_byteswap(buf + pos, inc, | |
6215 | &zc_bswap); | |
6216 | ||
6217 | pos += inc; | |
6218 | } | |
6219 | ||
6220 | VERIFY3U(pos, ==, size); | |
6221 | ||
6222 | VERIFY(ZIO_CHECKSUM_EQUAL(zc, zc_ref)); | |
6223 | VERIFY(ZIO_CHECKSUM_EQUAL(zc_bswap, zc_ref_bswap)); | |
6224 | ||
6225 | /* | |
6226 | * verify if incremental on the whole buffer is | |
6227 | * equivalent to non-incremental version | |
6228 | */ | |
6229 | ZIO_SET_CHECKSUM(&zc, 0, 0, 0, 0); | |
6230 | ZIO_SET_CHECKSUM(&zc_bswap, 0, 0, 0, 0); | |
6231 | ||
6232 | fletcher_4_incremental_native(buf, size, &zc); | |
6233 | fletcher_4_incremental_byteswap(buf, size, &zc_bswap); | |
6234 | ||
6235 | VERIFY(ZIO_CHECKSUM_EQUAL(zc, zc_ref)); | |
6236 | VERIFY(ZIO_CHECKSUM_EQUAL(zc_bswap, zc_ref_bswap)); | |
6237 | } | |
6238 | ||
6239 | umem_free(buf, size); | |
6240 | } | |
6241 | } | |
6242 | ||
90aa094d CC |
6243 | static int |
6244 | ztest_check_path(char *path) | |
6245 | { | |
6246 | struct stat s; | |
6247 | /* return true on success */ | |
6248 | return (!stat(path, &s)); | |
6249 | } | |
6250 | ||
6251 | static void | |
6252 | ztest_get_zdb_bin(char *bin, int len) | |
6253 | { | |
6254 | char *zdb_path; | |
6255 | /* | |
6256 | * Try to use ZDB_PATH and in-tree zdb path. If not successful, just | |
6257 | * let popen to search through PATH. | |
6258 | */ | |
6259 | if ((zdb_path = getenv("ZDB_PATH"))) { | |
6260 | strlcpy(bin, zdb_path, len); /* In env */ | |
6261 | if (!ztest_check_path(bin)) { | |
6262 | ztest_dump_core = 0; | |
6263 | fatal(1, "invalid ZDB_PATH '%s'", bin); | |
6264 | } | |
6265 | return; | |
6266 | } | |
6267 | ||
6268 | VERIFY(realpath(getexecname(), bin) != NULL); | |
6269 | if (strstr(bin, "/ztest/")) { | |
6270 | strstr(bin, "/ztest/")[0] = '\0'; /* In-tree */ | |
6271 | strcat(bin, "/zdb/zdb"); | |
6272 | if (ztest_check_path(bin)) | |
6273 | return; | |
6274 | } | |
6275 | strcpy(bin, "zdb"); | |
6276 | } | |
6277 | ||
619f0976 GW |
6278 | static vdev_t * |
6279 | ztest_random_concrete_vdev_leaf(vdev_t *vd) | |
6280 | { | |
6281 | if (vd == NULL) | |
6282 | return (NULL); | |
6283 | ||
6284 | if (vd->vdev_children == 0) | |
6285 | return (vd); | |
6286 | ||
6287 | vdev_t *eligible[vd->vdev_children]; | |
6288 | int eligible_idx = 0, i; | |
6289 | for (i = 0; i < vd->vdev_children; i++) { | |
6290 | vdev_t *cvd = vd->vdev_child[i]; | |
6291 | if (cvd->vdev_top->vdev_removing) | |
6292 | continue; | |
6293 | if (cvd->vdev_children > 0 || | |
6294 | (vdev_is_concrete(cvd) && !cvd->vdev_detached)) { | |
6295 | eligible[eligible_idx++] = cvd; | |
6296 | } | |
6297 | } | |
6298 | VERIFY(eligible_idx > 0); | |
6299 | ||
6300 | uint64_t child_no = ztest_random(eligible_idx); | |
6301 | return (ztest_random_concrete_vdev_leaf(eligible[child_no])); | |
6302 | } | |
6303 | ||
6304 | /* ARGSUSED */ | |
6305 | void | |
6306 | ztest_initialize(ztest_ds_t *zd, uint64_t id) | |
6307 | { | |
6308 | spa_t *spa = ztest_spa; | |
6309 | int error = 0; | |
6310 | ||
6311 | mutex_enter(&ztest_vdev_lock); | |
6312 | ||
6313 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
6314 | ||
6315 | /* Random leaf vdev */ | |
6316 | vdev_t *rand_vd = ztest_random_concrete_vdev_leaf(spa->spa_root_vdev); | |
6317 | if (rand_vd == NULL) { | |
6318 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
6319 | mutex_exit(&ztest_vdev_lock); | |
6320 | return; | |
6321 | } | |
6322 | ||
6323 | /* | |
6324 | * The random vdev we've selected may change as soon as we | |
6325 | * drop the spa_config_lock. We create local copies of things | |
6326 | * we're interested in. | |
6327 | */ | |
6328 | uint64_t guid = rand_vd->vdev_guid; | |
6329 | char *path = strdup(rand_vd->vdev_path); | |
6330 | boolean_t active = rand_vd->vdev_initialize_thread != NULL; | |
6331 | ||
a887d653 | 6332 | zfs_dbgmsg("vd %px, guid %llu", rand_vd, guid); |
619f0976 GW |
6333 | spa_config_exit(spa, SCL_VDEV, FTAG); |
6334 | ||
6335 | uint64_t cmd = ztest_random(POOL_INITIALIZE_FUNCS); | |
c10d37dd GW |
6336 | |
6337 | nvlist_t *vdev_guids = fnvlist_alloc(); | |
6338 | nvlist_t *vdev_errlist = fnvlist_alloc(); | |
6339 | fnvlist_add_uint64(vdev_guids, path, guid); | |
6340 | error = spa_vdev_initialize(spa, vdev_guids, cmd, vdev_errlist); | |
6341 | fnvlist_free(vdev_guids); | |
6342 | fnvlist_free(vdev_errlist); | |
6343 | ||
619f0976 GW |
6344 | switch (cmd) { |
6345 | case POOL_INITIALIZE_CANCEL: | |
6346 | if (ztest_opts.zo_verbose >= 4) { | |
6347 | (void) printf("Cancel initialize %s", path); | |
6348 | if (!active) | |
6349 | (void) printf(" failed (no initialize active)"); | |
6350 | (void) printf("\n"); | |
6351 | } | |
6352 | break; | |
1b939560 | 6353 | case POOL_INITIALIZE_START: |
619f0976 GW |
6354 | if (ztest_opts.zo_verbose >= 4) { |
6355 | (void) printf("Start initialize %s", path); | |
6356 | if (active && error == 0) | |
6357 | (void) printf(" failed (already active)"); | |
6358 | else if (error != 0) | |
6359 | (void) printf(" failed (error %d)", error); | |
6360 | (void) printf("\n"); | |
6361 | } | |
6362 | break; | |
6363 | case POOL_INITIALIZE_SUSPEND: | |
6364 | if (ztest_opts.zo_verbose >= 4) { | |
6365 | (void) printf("Suspend initialize %s", path); | |
6366 | if (!active) | |
6367 | (void) printf(" failed (no initialize active)"); | |
6368 | (void) printf("\n"); | |
6369 | } | |
6370 | break; | |
6371 | } | |
6372 | free(path); | |
6373 | mutex_exit(&ztest_vdev_lock); | |
6374 | } | |
6375 | ||
1b939560 BB |
6376 | /* ARGSUSED */ |
6377 | void | |
6378 | ztest_trim(ztest_ds_t *zd, uint64_t id) | |
6379 | { | |
6380 | spa_t *spa = ztest_spa; | |
6381 | int error = 0; | |
6382 | ||
6383 | mutex_enter(&ztest_vdev_lock); | |
6384 | ||
6385 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
6386 | ||
6387 | /* Random leaf vdev */ | |
6388 | vdev_t *rand_vd = ztest_random_concrete_vdev_leaf(spa->spa_root_vdev); | |
6389 | if (rand_vd == NULL) { | |
6390 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
6391 | mutex_exit(&ztest_vdev_lock); | |
6392 | return; | |
6393 | } | |
6394 | ||
6395 | /* | |
6396 | * The random vdev we've selected may change as soon as we | |
6397 | * drop the spa_config_lock. We create local copies of things | |
6398 | * we're interested in. | |
6399 | */ | |
6400 | uint64_t guid = rand_vd->vdev_guid; | |
6401 | char *path = strdup(rand_vd->vdev_path); | |
6402 | boolean_t active = rand_vd->vdev_trim_thread != NULL; | |
6403 | ||
6404 | zfs_dbgmsg("vd %p, guid %llu", rand_vd, guid); | |
6405 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
6406 | ||
6407 | uint64_t cmd = ztest_random(POOL_TRIM_FUNCS); | |
6408 | uint64_t rate = 1 << ztest_random(30); | |
6409 | boolean_t partial = (ztest_random(5) > 0); | |
6410 | boolean_t secure = (ztest_random(5) > 0); | |
6411 | ||
6412 | nvlist_t *vdev_guids = fnvlist_alloc(); | |
6413 | nvlist_t *vdev_errlist = fnvlist_alloc(); | |
6414 | fnvlist_add_uint64(vdev_guids, path, guid); | |
6415 | error = spa_vdev_trim(spa, vdev_guids, cmd, rate, partial, | |
6416 | secure, vdev_errlist); | |
6417 | fnvlist_free(vdev_guids); | |
6418 | fnvlist_free(vdev_errlist); | |
6419 | ||
6420 | switch (cmd) { | |
6421 | case POOL_TRIM_CANCEL: | |
6422 | if (ztest_opts.zo_verbose >= 4) { | |
6423 | (void) printf("Cancel TRIM %s", path); | |
6424 | if (!active) | |
6425 | (void) printf(" failed (no TRIM active)"); | |
6426 | (void) printf("\n"); | |
6427 | } | |
6428 | break; | |
6429 | case POOL_TRIM_START: | |
6430 | if (ztest_opts.zo_verbose >= 4) { | |
6431 | (void) printf("Start TRIM %s", path); | |
6432 | if (active && error == 0) | |
6433 | (void) printf(" failed (already active)"); | |
6434 | else if (error != 0) | |
6435 | (void) printf(" failed (error %d)", error); | |
6436 | (void) printf("\n"); | |
6437 | } | |
6438 | break; | |
6439 | case POOL_TRIM_SUSPEND: | |
6440 | if (ztest_opts.zo_verbose >= 4) { | |
6441 | (void) printf("Suspend TRIM %s", path); | |
6442 | if (!active) | |
6443 | (void) printf(" failed (no TRIM active)"); | |
6444 | (void) printf("\n"); | |
6445 | } | |
6446 | break; | |
6447 | } | |
6448 | free(path); | |
6449 | mutex_exit(&ztest_vdev_lock); | |
6450 | } | |
6451 | ||
428870ff BB |
6452 | /* |
6453 | * Verify pool integrity by running zdb. | |
6454 | */ | |
34dc7c2f | 6455 | static void |
428870ff | 6456 | ztest_run_zdb(char *pool) |
34dc7c2f BB |
6457 | { |
6458 | int status; | |
34dc7c2f | 6459 | char *bin; |
0e8d1b2d BB |
6460 | char *zdb; |
6461 | char *zbuf; | |
90aa094d | 6462 | const int len = MAXPATHLEN + MAXNAMELEN + 20; |
34dc7c2f BB |
6463 | FILE *fp; |
6464 | ||
90aa094d CC |
6465 | bin = umem_alloc(len, UMEM_NOFAIL); |
6466 | zdb = umem_alloc(len, UMEM_NOFAIL); | |
0e8d1b2d | 6467 | zbuf = umem_alloc(1024, UMEM_NOFAIL); |
34dc7c2f | 6468 | |
90aa094d | 6469 | ztest_get_zdb_bin(bin, len); |
0e8d1b2d BB |
6470 | |
6471 | (void) sprintf(zdb, | |
6774931d | 6472 | "%s -bcc%s%s -G -d -Y -e -y -p %s %s", |
0e8d1b2d | 6473 | bin, |
c242c188 CS |
6474 | ztest_opts.zo_verbose >= 3 ? "s" : "", |
6475 | ztest_opts.zo_verbose >= 4 ? "v" : "", | |
3d93161b | 6476 | ztest_opts.zo_dir, |
b128c09f | 6477 | pool); |
34dc7c2f | 6478 | |
c242c188 | 6479 | if (ztest_opts.zo_verbose >= 5) |
34dc7c2f BB |
6480 | (void) printf("Executing %s\n", strstr(zdb, "zdb ")); |
6481 | ||
6482 | fp = popen(zdb, "r"); | |
6483 | ||
6aec1cd5 | 6484 | while (fgets(zbuf, 1024, fp) != NULL) |
c242c188 | 6485 | if (ztest_opts.zo_verbose >= 3) |
34dc7c2f BB |
6486 | (void) printf("%s", zbuf); |
6487 | ||
6488 | status = pclose(fp); | |
6489 | ||
6490 | if (status == 0) | |
0e8d1b2d | 6491 | goto out; |
34dc7c2f BB |
6492 | |
6493 | ztest_dump_core = 0; | |
6494 | if (WIFEXITED(status)) | |
6495 | fatal(0, "'%s' exit code %d", zdb, WEXITSTATUS(status)); | |
6496 | else | |
6497 | fatal(0, "'%s' died with signal %d", zdb, WTERMSIG(status)); | |
0e8d1b2d | 6498 | out: |
90aa094d CC |
6499 | umem_free(bin, len); |
6500 | umem_free(zdb, len); | |
0e8d1b2d | 6501 | umem_free(zbuf, 1024); |
34dc7c2f BB |
6502 | } |
6503 | ||
6504 | static void | |
6505 | ztest_walk_pool_directory(char *header) | |
6506 | { | |
6507 | spa_t *spa = NULL; | |
6508 | ||
c242c188 | 6509 | if (ztest_opts.zo_verbose >= 6) |
34dc7c2f BB |
6510 | (void) printf("%s\n", header); |
6511 | ||
6512 | mutex_enter(&spa_namespace_lock); | |
6513 | while ((spa = spa_next(spa)) != NULL) | |
c242c188 | 6514 | if (ztest_opts.zo_verbose >= 6) |
34dc7c2f BB |
6515 | (void) printf("\t%s\n", spa_name(spa)); |
6516 | mutex_exit(&spa_namespace_lock); | |
6517 | } | |
6518 | ||
6519 | static void | |
6520 | ztest_spa_import_export(char *oldname, char *newname) | |
6521 | { | |
fb5f0bc8 | 6522 | nvlist_t *config, *newconfig; |
34dc7c2f BB |
6523 | uint64_t pool_guid; |
6524 | spa_t *spa; | |
13fe0198 | 6525 | int error; |
34dc7c2f | 6526 | |
c242c188 | 6527 | if (ztest_opts.zo_verbose >= 4) { |
34dc7c2f BB |
6528 | (void) printf("import/export: old = %s, new = %s\n", |
6529 | oldname, newname); | |
6530 | } | |
6531 | ||
6532 | /* | |
6533 | * Clean up from previous runs. | |
6534 | */ | |
6535 | (void) spa_destroy(newname); | |
6536 | ||
6537 | /* | |
6538 | * Get the pool's configuration and guid. | |
6539 | */ | |
428870ff | 6540 | VERIFY3U(0, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f | 6541 | |
fb5f0bc8 BB |
6542 | /* |
6543 | * Kick off a scrub to tickle scrub/export races. | |
6544 | */ | |
6545 | if (ztest_random(2) == 0) | |
428870ff | 6546 | (void) spa_scan(spa, POOL_SCAN_SCRUB); |
fb5f0bc8 | 6547 | |
34dc7c2f BB |
6548 | pool_guid = spa_guid(spa); |
6549 | spa_close(spa, FTAG); | |
6550 | ||
6551 | ztest_walk_pool_directory("pools before export"); | |
6552 | ||
6553 | /* | |
6554 | * Export it. | |
6555 | */ | |
428870ff | 6556 | VERIFY3U(0, ==, spa_export(oldname, &config, B_FALSE, B_FALSE)); |
34dc7c2f BB |
6557 | |
6558 | ztest_walk_pool_directory("pools after export"); | |
6559 | ||
fb5f0bc8 BB |
6560 | /* |
6561 | * Try to import it. | |
6562 | */ | |
6563 | newconfig = spa_tryimport(config); | |
6564 | ASSERT(newconfig != NULL); | |
6565 | nvlist_free(newconfig); | |
6566 | ||
34dc7c2f BB |
6567 | /* |
6568 | * Import it under the new name. | |
6569 | */ | |
13fe0198 MA |
6570 | error = spa_import(newname, config, NULL, 0); |
6571 | if (error != 0) { | |
6572 | dump_nvlist(config, 0); | |
6573 | fatal(B_FALSE, "couldn't import pool %s as %s: error %u", | |
6574 | oldname, newname, error); | |
6575 | } | |
34dc7c2f BB |
6576 | |
6577 | ztest_walk_pool_directory("pools after import"); | |
6578 | ||
6579 | /* | |
6580 | * Try to import it again -- should fail with EEXIST. | |
6581 | */ | |
572e2857 | 6582 | VERIFY3U(EEXIST, ==, spa_import(newname, config, NULL, 0)); |
34dc7c2f BB |
6583 | |
6584 | /* | |
6585 | * Try to import it under a different name -- should fail with EEXIST. | |
6586 | */ | |
572e2857 | 6587 | VERIFY3U(EEXIST, ==, spa_import(oldname, config, NULL, 0)); |
34dc7c2f BB |
6588 | |
6589 | /* | |
6590 | * Verify that the pool is no longer visible under the old name. | |
6591 | */ | |
428870ff | 6592 | VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f BB |
6593 | |
6594 | /* | |
6595 | * Verify that we can open and close the pool using the new name. | |
6596 | */ | |
428870ff | 6597 | VERIFY3U(0, ==, spa_open(newname, &spa, FTAG)); |
34dc7c2f BB |
6598 | ASSERT(pool_guid == spa_guid(spa)); |
6599 | spa_close(spa, FTAG); | |
6600 | ||
6601 | nvlist_free(config); | |
6602 | } | |
6603 | ||
fb5f0bc8 BB |
6604 | static void |
6605 | ztest_resume(spa_t *spa) | |
6606 | { | |
c242c188 | 6607 | if (spa_suspended(spa) && ztest_opts.zo_verbose >= 6) |
428870ff BB |
6608 | (void) printf("resuming from suspended state\n"); |
6609 | spa_vdev_state_enter(spa, SCL_NONE); | |
6610 | vdev_clear(spa, NULL); | |
6611 | (void) spa_vdev_state_exit(spa, NULL, 0); | |
6612 | (void) zio_resume(spa); | |
fb5f0bc8 BB |
6613 | } |
6614 | ||
c25b8f99 | 6615 | static void |
fb5f0bc8 | 6616 | ztest_resume_thread(void *arg) |
34dc7c2f | 6617 | { |
b128c09f | 6618 | spa_t *spa = arg; |
34dc7c2f BB |
6619 | |
6620 | while (!ztest_exiting) { | |
428870ff BB |
6621 | if (spa_suspended(spa)) |
6622 | ztest_resume(spa); | |
6623 | (void) poll(NULL, 0, 100); | |
d3c2ae1c GW |
6624 | |
6625 | /* | |
6626 | * Periodically change the zfs_compressed_arc_enabled setting. | |
6627 | */ | |
6628 | if (ztest_random(10) == 0) | |
6629 | zfs_compressed_arc_enabled = ztest_random(2); | |
a6255b7f DQ |
6630 | |
6631 | /* | |
6632 | * Periodically change the zfs_abd_scatter_enabled setting. | |
6633 | */ | |
6634 | if (ztest_random(10) == 0) | |
6635 | zfs_abd_scatter_enabled = ztest_random(2); | |
34dc7c2f | 6636 | } |
34dc7c2f | 6637 | |
1e33ac1e | 6638 | thread_exit(); |
1e33ac1e | 6639 | } |
428870ff | 6640 | |
1e33ac1e | 6641 | static void |
8fb1ede1 | 6642 | ztest_deadman_thread(void *arg) |
1e33ac1e | 6643 | { |
8fb1ede1 BB |
6644 | ztest_shared_t *zs = arg; |
6645 | spa_t *spa = ztest_spa; | |
94102578 | 6646 | hrtime_t delay, overdue, last_run = gethrtime(); |
8fb1ede1 | 6647 | |
94102578 TC |
6648 | delay = (zs->zs_thread_stop - zs->zs_thread_start) + |
6649 | MSEC2NSEC(zfs_deadman_synctime_ms); | |
8fb1ede1 | 6650 | |
94102578 TC |
6651 | while (!ztest_exiting) { |
6652 | /* | |
6653 | * Wait for the delay timer while checking occasionally | |
6654 | * if we should stop. | |
6655 | */ | |
6656 | if (gethrtime() < last_run + delay) { | |
6657 | (void) poll(NULL, 0, 1000); | |
6658 | continue; | |
6659 | } | |
8fb1ede1 BB |
6660 | |
6661 | /* | |
6662 | * If the pool is suspended then fail immediately. Otherwise, | |
6663 | * check to see if the pool is making any progress. If | |
6664 | * vdev_deadman() discovers that there hasn't been any recent | |
6665 | * I/Os then it will end up aborting the tests. | |
6666 | */ | |
6667 | if (spa_suspended(spa) || spa->spa_root_vdev == NULL) { | |
6668 | fatal(0, "aborting test after %llu seconds because " | |
6669 | "pool has transitioned to a suspended state.", | |
6670 | zfs_deadman_synctime_ms / 1000); | |
6671 | } | |
6672 | vdev_deadman(spa->spa_root_vdev, FTAG); | |
6673 | ||
6674 | /* | |
6675 | * If the process doesn't complete within a grace period of | |
6676 | * zfs_deadman_synctime_ms over the expected finish time, | |
6677 | * then it may be hung and is terminated. | |
6678 | */ | |
6679 | overdue = zs->zs_proc_stop + MSEC2NSEC(zfs_deadman_synctime_ms); | |
8fb1ede1 BB |
6680 | if (gethrtime() > overdue) { |
6681 | fatal(0, "aborting test after %llu seconds because " | |
94102578 TC |
6682 | "the process is overdue for termination.", |
6683 | (gethrtime() - zs->zs_proc_start) / NANOSEC); | |
8fb1ede1 BB |
6684 | } |
6685 | ||
6686 | (void) printf("ztest has been running for %lld seconds\n", | |
94102578 TC |
6687 | (gethrtime() - zs->zs_proc_start) / NANOSEC); |
6688 | ||
6689 | last_run = gethrtime(); | |
6690 | delay = MSEC2NSEC(zfs_deadman_checktime_ms); | |
8fb1ede1 | 6691 | } |
94102578 TC |
6692 | |
6693 | thread_exit(); | |
428870ff BB |
6694 | } |
6695 | ||
6696 | static void | |
c242c188 | 6697 | ztest_execute(int test, ztest_info_t *zi, uint64_t id) |
428870ff | 6698 | { |
c242c188 CS |
6699 | ztest_ds_t *zd = &ztest_ds[id % ztest_opts.zo_datasets]; |
6700 | ztest_shared_callstate_t *zc = ZTEST_GET_SHARED_CALLSTATE(test); | |
428870ff | 6701 | hrtime_t functime = gethrtime(); |
d6320ddb | 6702 | int i; |
428870ff | 6703 | |
d6320ddb | 6704 | for (i = 0; i < zi->zi_iters; i++) |
428870ff BB |
6705 | zi->zi_func(zd, id); |
6706 | ||
6707 | functime = gethrtime() - functime; | |
6708 | ||
c242c188 CS |
6709 | atomic_add_64(&zc->zc_count, 1); |
6710 | atomic_add_64(&zc->zc_time, functime); | |
428870ff | 6711 | |
fdc5d982 | 6712 | if (ztest_opts.zo_verbose >= 4) |
428870ff | 6713 | (void) printf("%6.2f sec in %s\n", |
fdc5d982 | 6714 | (double)functime / NANOSEC, zi->zi_funcname); |
428870ff BB |
6715 | } |
6716 | ||
c25b8f99 | 6717 | static void |
34dc7c2f BB |
6718 | ztest_thread(void *arg) |
6719 | { | |
c242c188 | 6720 | int rand; |
428870ff | 6721 | uint64_t id = (uintptr_t)arg; |
34dc7c2f | 6722 | ztest_shared_t *zs = ztest_shared; |
428870ff BB |
6723 | uint64_t call_next; |
6724 | hrtime_t now; | |
34dc7c2f | 6725 | ztest_info_t *zi; |
c242c188 | 6726 | ztest_shared_callstate_t *zc; |
34dc7c2f | 6727 | |
428870ff | 6728 | while ((now = gethrtime()) < zs->zs_thread_stop) { |
34dc7c2f BB |
6729 | /* |
6730 | * See if it's time to force a crash. | |
6731 | */ | |
428870ff BB |
6732 | if (now > zs->zs_thread_kill) |
6733 | ztest_kill(zs); | |
34dc7c2f BB |
6734 | |
6735 | /* | |
428870ff | 6736 | * If we're getting ENOSPC with some regularity, stop. |
34dc7c2f | 6737 | */ |
428870ff BB |
6738 | if (zs->zs_enospc_count > 10) |
6739 | break; | |
34dc7c2f BB |
6740 | |
6741 | /* | |
428870ff | 6742 | * Pick a random function to execute. |
34dc7c2f | 6743 | */ |
c242c188 CS |
6744 | rand = ztest_random(ZTEST_FUNCS); |
6745 | zi = &ztest_info[rand]; | |
6746 | zc = ZTEST_GET_SHARED_CALLSTATE(rand); | |
6747 | call_next = zc->zc_next; | |
34dc7c2f | 6748 | |
428870ff | 6749 | if (now >= call_next && |
c242c188 CS |
6750 | atomic_cas_64(&zc->zc_next, call_next, call_next + |
6751 | ztest_random(2 * zi->zi_interval[0] + 1)) == call_next) { | |
6752 | ztest_execute(rand, zi, id); | |
6753 | } | |
428870ff | 6754 | } |
34dc7c2f | 6755 | |
1e33ac1e | 6756 | thread_exit(); |
428870ff | 6757 | } |
34dc7c2f | 6758 | |
428870ff BB |
6759 | static void |
6760 | ztest_dataset_name(char *dsname, char *pool, int d) | |
6761 | { | |
eca7b760 | 6762 | (void) snprintf(dsname, ZFS_MAX_DATASET_NAME_LEN, "%s/ds_%d", pool, d); |
428870ff | 6763 | } |
34dc7c2f | 6764 | |
428870ff | 6765 | static void |
c242c188 | 6766 | ztest_dataset_destroy(int d) |
428870ff | 6767 | { |
eca7b760 | 6768 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
d6320ddb | 6769 | int t; |
34dc7c2f | 6770 | |
c242c188 | 6771 | ztest_dataset_name(name, ztest_opts.zo_pool, d); |
34dc7c2f | 6772 | |
c242c188 | 6773 | if (ztest_opts.zo_verbose >= 3) |
428870ff | 6774 | (void) printf("Destroying %s to free up space\n", name); |
34dc7c2f | 6775 | |
428870ff BB |
6776 | /* |
6777 | * Cleanup any non-standard clones and snapshots. In general, | |
6778 | * ztest thread t operates on dataset (t % zopt_datasets), | |
6779 | * so there may be more than one thing to clean up. | |
6780 | */ | |
c242c188 CS |
6781 | for (t = d; t < ztest_opts.zo_threads; |
6782 | t += ztest_opts.zo_datasets) | |
428870ff BB |
6783 | ztest_dsl_dataset_cleanup(name, t); |
6784 | ||
6785 | (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL, | |
6786 | DS_FIND_SNAPSHOTS | DS_FIND_CHILDREN); | |
6787 | } | |
6788 | ||
6789 | static void | |
6790 | ztest_dataset_dirobj_verify(ztest_ds_t *zd) | |
6791 | { | |
6792 | uint64_t usedobjs, dirobjs, scratch; | |
6793 | ||
6794 | /* | |
6795 | * ZTEST_DIROBJ is the object directory for the entire dataset. | |
6796 | * Therefore, the number of objects in use should equal the | |
6797 | * number of ZTEST_DIROBJ entries, +1 for ZTEST_DIROBJ itself. | |
6798 | * If not, we have an object leak. | |
6799 | * | |
6800 | * Note that we can only check this in ztest_dataset_open(), | |
6801 | * when the open-context and syncing-context values agree. | |
6802 | * That's because zap_count() returns the open-context value, | |
6803 | * while dmu_objset_space() returns the rootbp fill count. | |
6804 | */ | |
6805 | VERIFY3U(0, ==, zap_count(zd->zd_os, ZTEST_DIROBJ, &dirobjs)); | |
6806 | dmu_objset_space(zd->zd_os, &scratch, &scratch, &usedobjs, &scratch); | |
6807 | ASSERT3U(dirobjs + 1, ==, usedobjs); | |
6808 | } | |
6809 | ||
6810 | static int | |
c242c188 | 6811 | ztest_dataset_open(int d) |
428870ff | 6812 | { |
c242c188 CS |
6813 | ztest_ds_t *zd = &ztest_ds[d]; |
6814 | uint64_t committed_seq = ZTEST_GET_SHARED_DS(d)->zd_seq; | |
428870ff BB |
6815 | objset_t *os; |
6816 | zilog_t *zilog; | |
eca7b760 | 6817 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
6818 | int error; |
6819 | ||
c242c188 | 6820 | ztest_dataset_name(name, ztest_opts.zo_pool, d); |
428870ff | 6821 | |
2d9142c9 | 6822 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
428870ff BB |
6823 | |
6824 | error = ztest_dataset_create(name); | |
6825 | if (error == ENOSPC) { | |
2d9142c9 | 6826 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
428870ff BB |
6827 | ztest_record_enospc(FTAG); |
6828 | return (error); | |
34dc7c2f | 6829 | } |
428870ff | 6830 | ASSERT(error == 0 || error == EEXIST); |
34dc7c2f | 6831 | |
4807c0ba TC |
6832 | VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, |
6833 | B_TRUE, zd, &os)); | |
2d9142c9 | 6834 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
428870ff | 6835 | |
c242c188 | 6836 | ztest_zd_init(zd, ZTEST_GET_SHARED_DS(d), os); |
428870ff BB |
6837 | |
6838 | zilog = zd->zd_zilog; | |
6839 | ||
6840 | if (zilog->zl_header->zh_claim_lr_seq != 0 && | |
6841 | zilog->zl_header->zh_claim_lr_seq < committed_seq) | |
6842 | fatal(0, "missing log records: claimed %llu < committed %llu", | |
6843 | zilog->zl_header->zh_claim_lr_seq, committed_seq); | |
6844 | ||
6845 | ztest_dataset_dirobj_verify(zd); | |
6846 | ||
6847 | zil_replay(os, zd, ztest_replay_vector); | |
6848 | ||
6849 | ztest_dataset_dirobj_verify(zd); | |
6850 | ||
c242c188 | 6851 | if (ztest_opts.zo_verbose >= 6) |
428870ff BB |
6852 | (void) printf("%s replay %llu blocks, %llu records, seq %llu\n", |
6853 | zd->zd_name, | |
6854 | (u_longlong_t)zilog->zl_parse_blk_count, | |
6855 | (u_longlong_t)zilog->zl_parse_lr_count, | |
6856 | (u_longlong_t)zilog->zl_replaying_seq); | |
6857 | ||
6858 | zilog = zil_open(os, ztest_get_data); | |
6859 | ||
6860 | if (zilog->zl_replaying_seq != 0 && | |
6861 | zilog->zl_replaying_seq < committed_seq) | |
6862 | fatal(0, "missing log records: replayed %llu < committed %llu", | |
6863 | zilog->zl_replaying_seq, committed_seq); | |
6864 | ||
6865 | return (0); | |
6866 | } | |
6867 | ||
6868 | static void | |
c242c188 | 6869 | ztest_dataset_close(int d) |
428870ff | 6870 | { |
c242c188 | 6871 | ztest_ds_t *zd = &ztest_ds[d]; |
428870ff BB |
6872 | |
6873 | zil_close(zd->zd_zilog); | |
b5256303 | 6874 | dmu_objset_disown(zd->zd_os, B_TRUE, zd); |
428870ff BB |
6875 | |
6876 | ztest_zd_fini(zd); | |
34dc7c2f BB |
6877 | } |
6878 | ||
f44ad929 TC |
6879 | /* ARGSUSED */ |
6880 | static int | |
6881 | ztest_replay_zil_cb(const char *name, void *arg) | |
6882 | { | |
6883 | objset_t *os; | |
6884 | ztest_ds_t *zdtmp; | |
6885 | ||
6886 | VERIFY0(ztest_dmu_objset_own(name, DMU_OST_ANY, B_TRUE, | |
6887 | B_TRUE, FTAG, &os)); | |
6888 | ||
6889 | zdtmp = umem_alloc(sizeof (ztest_ds_t), UMEM_NOFAIL); | |
6890 | ||
6891 | ztest_zd_init(zdtmp, NULL, os); | |
6892 | zil_replay(os, zdtmp, ztest_replay_vector); | |
6893 | ztest_zd_fini(zdtmp); | |
6894 | ||
6895 | if (dmu_objset_zil(os)->zl_parse_lr_count != 0 && | |
6896 | ztest_opts.zo_verbose >= 6) { | |
6897 | zilog_t *zilog = dmu_objset_zil(os); | |
6898 | ||
6899 | (void) printf("%s replay %llu blocks, %llu records, seq %llu\n", | |
6900 | name, | |
6901 | (u_longlong_t)zilog->zl_parse_blk_count, | |
6902 | (u_longlong_t)zilog->zl_parse_lr_count, | |
6903 | (u_longlong_t)zilog->zl_replaying_seq); | |
6904 | } | |
6905 | ||
6906 | umem_free(zdtmp, sizeof (ztest_ds_t)); | |
6907 | ||
6908 | dmu_objset_disown(os, B_TRUE, FTAG); | |
6909 | return (0); | |
6910 | } | |
6911 | ||
c1f3de18 BB |
6912 | static void |
6913 | ztest_freeze(void) | |
6914 | { | |
6915 | ztest_ds_t *zd = &ztest_ds[0]; | |
6916 | spa_t *spa; | |
6917 | int numloops = 0; | |
6918 | ||
6919 | if (ztest_opts.zo_verbose >= 3) | |
6920 | (void) printf("testing spa_freeze()...\n"); | |
6921 | ||
6922 | kernel_init(SPA_MODE_READ | SPA_MODE_WRITE); | |
6923 | VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); | |
6924 | VERIFY3U(0, ==, ztest_dataset_open(0)); | |
6925 | ztest_spa = spa; | |
6926 | ||
6927 | /* | |
6928 | * Force the first log block to be transactionally allocated. | |
6929 | * We have to do this before we freeze the pool -- otherwise | |
6930 | * the log chain won't be anchored. | |
6931 | */ | |
6932 | while (BP_IS_HOLE(&zd->zd_zilog->zl_header->zh_log)) { | |
6933 | ztest_dmu_object_alloc_free(zd, 0); | |
6934 | zil_commit(zd->zd_zilog, 0); | |
6935 | } | |
6936 | ||
6937 | txg_wait_synced(spa_get_dsl(spa), 0); | |
6938 | ||
6939 | /* | |
6940 | * Freeze the pool. This stops spa_sync() from doing anything, | |
6941 | * so that the only way to record changes from now on is the ZIL. | |
6942 | */ | |
6943 | spa_freeze(spa); | |
6944 | ||
6945 | /* | |
6946 | * Because it is hard to predict how much space a write will actually | |
6947 | * require beforehand, we leave ourselves some fudge space to write over | |
6948 | * capacity. | |
6949 | */ | |
6950 | uint64_t capacity = metaslab_class_get_space(spa_normal_class(spa)) / 2; | |
6951 | ||
6952 | /* | |
6953 | * Run tests that generate log records but don't alter the pool config | |
6954 | * or depend on DSL sync tasks (snapshots, objset create/destroy, etc). | |
6955 | * We do a txg_wait_synced() after each iteration to force the txg | |
6956 | * to increase well beyond the last synced value in the uberblock. | |
6957 | * The ZIL should be OK with that. | |
6958 | * | |
6959 | * Run a random number of times less than zo_maxloops and ensure we do | |
6960 | * not run out of space on the pool. | |
6961 | */ | |
6962 | while (ztest_random(10) != 0 && | |
6963 | numloops++ < ztest_opts.zo_maxloops && | |
6964 | metaslab_class_get_alloc(spa_normal_class(spa)) < capacity) { | |
6965 | ztest_od_t od; | |
6966 | ztest_od_init(&od, 0, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0); | |
6967 | VERIFY0(ztest_object_init(zd, &od, sizeof (od), B_FALSE)); | |
6968 | ztest_io(zd, od.od_object, | |
6969 | ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
6970 | txg_wait_synced(spa_get_dsl(spa), 0); | |
6971 | } | |
6972 | ||
6973 | /* | |
6974 | * Commit all of the changes we just generated. | |
6975 | */ | |
6976 | zil_commit(zd->zd_zilog, 0); | |
6977 | txg_wait_synced(spa_get_dsl(spa), 0); | |
6978 | ||
6979 | /* | |
6980 | * Close our dataset and close the pool. | |
6981 | */ | |
6982 | ztest_dataset_close(0); | |
6983 | spa_close(spa, FTAG); | |
6984 | kernel_fini(); | |
6985 | ||
6986 | /* | |
6987 | * Open and close the pool and dataset to induce log replay. | |
6988 | */ | |
6989 | kernel_init(SPA_MODE_READ | SPA_MODE_WRITE); | |
6990 | VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); | |
6991 | ASSERT(spa_freeze_txg(spa) == UINT64_MAX); | |
6992 | VERIFY3U(0, ==, ztest_dataset_open(0)); | |
6993 | ztest_spa = spa; | |
6994 | txg_wait_synced(spa_get_dsl(spa), 0); | |
6995 | ztest_dataset_close(0); | |
6996 | ztest_reguid(NULL, 0); | |
6997 | ||
6998 | spa_close(spa, FTAG); | |
6999 | kernel_fini(); | |
7000 | } | |
7001 | ||
7002 | static void | |
7003 | ztest_import_impl(ztest_shared_t *zs) | |
7004 | { | |
7005 | importargs_t args = { 0 }; | |
7006 | nvlist_t *cfg = NULL; | |
7007 | int nsearch = 1; | |
7008 | char *searchdirs[nsearch]; | |
7009 | int flags = ZFS_IMPORT_MISSING_LOG; | |
7010 | ||
7011 | searchdirs[0] = ztest_opts.zo_dir; | |
7012 | args.paths = nsearch; | |
7013 | args.path = searchdirs; | |
7014 | args.can_be_active = B_FALSE; | |
7015 | ||
7016 | VERIFY0(zpool_find_config(NULL, ztest_opts.zo_pool, &cfg, &args, | |
7017 | &libzpool_config_ops)); | |
7018 | VERIFY0(spa_import(ztest_opts.zo_pool, cfg, NULL, flags)); | |
7019 | } | |
7020 | ||
7021 | /* | |
7022 | * Import a storage pool with the given name. | |
7023 | */ | |
7024 | static void | |
7025 | ztest_import(ztest_shared_t *zs) | |
7026 | { | |
7027 | spa_t *spa; | |
7028 | ||
7029 | mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); | |
7030 | mutex_init(&ztest_checkpoint_lock, NULL, MUTEX_DEFAULT, NULL); | |
7031 | VERIFY0(pthread_rwlock_init(&ztest_name_lock, NULL)); | |
7032 | ||
7033 | kernel_init(SPA_MODE_READ | SPA_MODE_WRITE); | |
7034 | ||
7035 | ztest_import_impl(zs); | |
7036 | ||
7037 | VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG)); | |
7038 | zs->zs_metaslab_sz = | |
7039 | 1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift; | |
7040 | spa_close(spa, FTAG); | |
7041 | ||
7042 | kernel_fini(); | |
7043 | ||
7044 | if (!ztest_opts.zo_mmp_test) { | |
7045 | ztest_run_zdb(ztest_opts.zo_pool); | |
7046 | ztest_freeze(); | |
7047 | ztest_run_zdb(ztest_opts.zo_pool); | |
7048 | } | |
7049 | ||
7050 | (void) pthread_rwlock_destroy(&ztest_name_lock); | |
7051 | mutex_destroy(&ztest_vdev_lock); | |
7052 | mutex_destroy(&ztest_checkpoint_lock); | |
7053 | } | |
7054 | ||
34dc7c2f BB |
7055 | /* |
7056 | * Kick off threads to run tests on all datasets in parallel. | |
7057 | */ | |
7058 | static void | |
428870ff | 7059 | ztest_run(ztest_shared_t *zs) |
34dc7c2f | 7060 | { |
34dc7c2f | 7061 | spa_t *spa; |
3541dc6d | 7062 | objset_t *os; |
94102578 | 7063 | kthread_t *resume_thread, *deadman_thread; |
c25b8f99 | 7064 | kthread_t **run_threads; |
1e33ac1e | 7065 | uint64_t object; |
428870ff | 7066 | int error; |
d6320ddb | 7067 | int t, d; |
b128c09f BB |
7068 | |
7069 | ztest_exiting = B_FALSE; | |
34dc7c2f | 7070 | |
34dc7c2f | 7071 | /* |
428870ff | 7072 | * Initialize parent/child shared state. |
34dc7c2f | 7073 | */ |
c242c188 | 7074 | mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); |
d2734cce | 7075 | mutex_init(&ztest_checkpoint_lock, NULL, MUTEX_DEFAULT, NULL); |
2d9142c9 | 7076 | VERIFY0(pthread_rwlock_init(&ztest_name_lock, NULL)); |
34dc7c2f | 7077 | |
428870ff | 7078 | zs->zs_thread_start = gethrtime(); |
c242c188 CS |
7079 | zs->zs_thread_stop = |
7080 | zs->zs_thread_start + ztest_opts.zo_passtime * NANOSEC; | |
428870ff BB |
7081 | zs->zs_thread_stop = MIN(zs->zs_thread_stop, zs->zs_proc_stop); |
7082 | zs->zs_thread_kill = zs->zs_thread_stop; | |
c242c188 CS |
7083 | if (ztest_random(100) < ztest_opts.zo_killrate) { |
7084 | zs->zs_thread_kill -= | |
7085 | ztest_random(ztest_opts.zo_passtime * NANOSEC); | |
7086 | } | |
34dc7c2f | 7087 | |
1e33ac1e | 7088 | mutex_init(&zcl.zcl_callbacks_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 7089 | |
428870ff BB |
7090 | list_create(&zcl.zcl_callbacks, sizeof (ztest_cb_data_t), |
7091 | offsetof(ztest_cb_data_t, zcd_node)); | |
34dc7c2f BB |
7092 | |
7093 | /* | |
c1f3de18 BB |
7094 | * Open our pool. It may need to be imported first depending on |
7095 | * what tests were running when the previous pass was terminated. | |
34dc7c2f | 7096 | */ |
da92d5cb | 7097 | kernel_init(SPA_MODE_READ | SPA_MODE_WRITE); |
c1f3de18 BB |
7098 | error = spa_open(ztest_opts.zo_pool, &spa, FTAG); |
7099 | if (error) { | |
7100 | VERIFY3S(error, ==, ENOENT); | |
7101 | ztest_import_impl(zs); | |
7102 | VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG)); | |
7103 | zs->zs_metaslab_sz = | |
7104 | 1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift; | |
7105 | } | |
7106 | ||
080b3100 | 7107 | metaslab_preload_limit = ztest_random(20) + 1; |
c242c188 | 7108 | ztest_spa = spa; |
428870ff | 7109 | |
e5db3134 BB |
7110 | VERIFY0(vdev_raidz_impl_set("cycle")); |
7111 | ||
546d32ca | 7112 | dmu_objset_stats_t dds; |
4807c0ba | 7113 | VERIFY0(ztest_dmu_objset_own(ztest_opts.zo_pool, |
b5256303 | 7114 | DMU_OST_ANY, B_TRUE, B_TRUE, FTAG, &os)); |
546d32ca BB |
7115 | dsl_pool_config_enter(dmu_objset_pool(os), FTAG); |
7116 | dmu_objset_fast_stat(os, &dds); | |
7117 | dsl_pool_config_exit(dmu_objset_pool(os), FTAG); | |
7118 | zs->zs_guid = dds.dds_guid; | |
b5256303 | 7119 | dmu_objset_disown(os, B_TRUE, FTAG); |
3541dc6d | 7120 | |
34dc7c2f | 7121 | /* |
b128c09f | 7122 | * Create a thread to periodically resume suspended I/O. |
34dc7c2f | 7123 | */ |
c25b8f99 BB |
7124 | resume_thread = thread_create(NULL, 0, ztest_resume_thread, |
7125 | spa, 0, NULL, TS_RUN | TS_JOINABLE, defclsyspri); | |
34dc7c2f | 7126 | |
428870ff | 7127 | /* |
8fb1ede1 | 7128 | * Create a deadman thread and set to panic if we hang. |
428870ff | 7129 | */ |
94102578 | 7130 | deadman_thread = thread_create(NULL, 0, ztest_deadman_thread, |
8fb1ede1 BB |
7131 | zs, 0, NULL, TS_RUN | TS_JOINABLE, defclsyspri); |
7132 | ||
7133 | spa->spa_deadman_failmode = ZIO_FAILURE_MODE_PANIC; | |
428870ff | 7134 | |
34dc7c2f | 7135 | /* |
d2734cce | 7136 | * Verify that we can safely inquire about any object, |
34dc7c2f BB |
7137 | * whether it's allocated or not. To make it interesting, |
7138 | * we probe a 5-wide window around each power of two. | |
7139 | * This hits all edge cases, including zero and the max. | |
7140 | */ | |
d6320ddb BB |
7141 | for (t = 0; t < 64; t++) { |
7142 | for (d = -5; d <= 5; d++) { | |
34dc7c2f BB |
7143 | error = dmu_object_info(spa->spa_meta_objset, |
7144 | (1ULL << t) + d, NULL); | |
7145 | ASSERT(error == 0 || error == ENOENT || | |
7146 | error == EINVAL); | |
7147 | } | |
7148 | } | |
7149 | ||
7150 | /* | |
428870ff | 7151 | * If we got any ENOSPC errors on the previous run, destroy something. |
34dc7c2f | 7152 | */ |
428870ff | 7153 | if (zs->zs_enospc_count != 0) { |
c242c188 CS |
7154 | int d = ztest_random(ztest_opts.zo_datasets); |
7155 | ztest_dataset_destroy(d); | |
428870ff | 7156 | } |
34dc7c2f BB |
7157 | zs->zs_enospc_count = 0; |
7158 | ||
c40a1124 TC |
7159 | /* |
7160 | * If we were in the middle of ztest_device_removal() and were killed | |
7161 | * we need to ensure the removal and scrub complete before running | |
7162 | * any tests that check ztest_device_removal_active. The removal will | |
7163 | * be restarted automatically when the spa is opened, but we need to | |
d611989f | 7164 | * initiate the scrub manually if it is not already in progress. Note |
c40a1124 TC |
7165 | * that we always run the scrub whenever an indirect vdev exists |
7166 | * because we have no way of knowing for sure if ztest_device_removal() | |
7167 | * fully completed its scrub before the pool was reimported. | |
7168 | */ | |
7169 | if (spa->spa_removing_phys.sr_state == DSS_SCANNING || | |
7170 | spa->spa_removing_phys.sr_prev_indirect_vdev != -1) { | |
7171 | while (spa->spa_removing_phys.sr_state == DSS_SCANNING) | |
7172 | txg_wait_synced(spa_get_dsl(spa), 0); | |
7173 | ||
ce5fb2a7 BB |
7174 | error = ztest_scrub_impl(spa); |
7175 | if (error == EBUSY) | |
7176 | error = 0; | |
7177 | ASSERT0(error); | |
c40a1124 TC |
7178 | } |
7179 | ||
c25b8f99 | 7180 | run_threads = umem_zalloc(ztest_opts.zo_threads * sizeof (kthread_t *), |
c242c188 | 7181 | UMEM_NOFAIL); |
34dc7c2f | 7182 | |
c242c188 | 7183 | if (ztest_opts.zo_verbose >= 4) |
34dc7c2f BB |
7184 | (void) printf("starting main threads...\n"); |
7185 | ||
f44ad929 TC |
7186 | /* |
7187 | * Replay all logs of all datasets in the pool. This is primarily for | |
7188 | * temporary datasets which wouldn't otherwise get replayed, which | |
7189 | * can trigger failures when attempting to offline a SLOG in | |
7190 | * ztest_fault_inject(). | |
7191 | */ | |
7192 | (void) dmu_objset_find(ztest_opts.zo_pool, ztest_replay_zil_cb, | |
7193 | NULL, DS_FIND_CHILDREN); | |
7194 | ||
428870ff BB |
7195 | /* |
7196 | * Kick off all the tests that run in parallel. | |
7197 | */ | |
c242c188 | 7198 | for (t = 0; t < ztest_opts.zo_threads; t++) { |
c25b8f99 BB |
7199 | if (t < ztest_opts.zo_datasets && ztest_dataset_open(t) != 0) { |
7200 | umem_free(run_threads, ztest_opts.zo_threads * | |
7201 | sizeof (kthread_t *)); | |
428870ff | 7202 | return; |
06cf4d98 | 7203 | } |
1e33ac1e | 7204 | |
c25b8f99 BB |
7205 | run_threads[t] = thread_create(NULL, 0, ztest_thread, |
7206 | (void *)(uintptr_t)t, 0, NULL, TS_RUN | TS_JOINABLE, | |
7207 | defclsyspri); | |
34dc7c2f BB |
7208 | } |
7209 | ||
428870ff | 7210 | /* |
a2d88f77 | 7211 | * Wait for all of the tests to complete. |
428870ff | 7212 | */ |
a2d88f77 | 7213 | for (t = 0; t < ztest_opts.zo_threads; t++) |
c25b8f99 | 7214 | VERIFY0(thread_join(run_threads[t])); |
a2d88f77 TC |
7215 | |
7216 | /* | |
7217 | * Close all datasets. This must be done after all the threads | |
7218 | * are joined so we can be sure none of the datasets are in-use | |
7219 | * by any of the threads. | |
7220 | */ | |
7221 | for (t = 0; t < ztest_opts.zo_threads; t++) { | |
c242c188 CS |
7222 | if (t < ztest_opts.zo_datasets) |
7223 | ztest_dataset_close(t); | |
34dc7c2f BB |
7224 | } |
7225 | ||
34dc7c2f BB |
7226 | txg_wait_synced(spa_get_dsl(spa), 0); |
7227 | ||
428870ff BB |
7228 | zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(spa)); |
7229 | zs->zs_space = metaslab_class_get_space(spa_normal_class(spa)); | |
7230 | ||
c25b8f99 | 7231 | umem_free(run_threads, ztest_opts.zo_threads * sizeof (kthread_t *)); |
428870ff | 7232 | |
94102578 | 7233 | /* Kill the resume and deadman threads */ |
428870ff | 7234 | ztest_exiting = B_TRUE; |
c25b8f99 | 7235 | VERIFY0(thread_join(resume_thread)); |
94102578 | 7236 | VERIFY0(thread_join(deadman_thread)); |
428870ff | 7237 | ztest_resume(spa); |
34dc7c2f BB |
7238 | |
7239 | /* | |
428870ff BB |
7240 | * Right before closing the pool, kick off a bunch of async I/O; |
7241 | * spa_close() should wait for it to complete. | |
34dc7c2f | 7242 | */ |
fcff0f35 PD |
7243 | for (object = 1; object < 50; object++) { |
7244 | dmu_prefetch(spa->spa_meta_objset, object, 0, 0, 1ULL << 20, | |
7245 | ZIO_PRIORITY_SYNC_READ); | |
7246 | } | |
428870ff | 7247 | |
090ff092 RC |
7248 | /* Verify that at least one commit cb was called in a timely fashion */ |
7249 | if (zc_cb_counter >= ZTEST_COMMIT_CB_MIN_REG) | |
c99c9001 | 7250 | VERIFY0(zc_min_txg_delay); |
090ff092 | 7251 | |
428870ff BB |
7252 | spa_close(spa, FTAG); |
7253 | ||
7254 | /* | |
7255 | * Verify that we can loop over all pools. | |
7256 | */ | |
7257 | mutex_enter(&spa_namespace_lock); | |
7258 | for (spa = spa_next(NULL); spa != NULL; spa = spa_next(spa)) | |
c242c188 | 7259 | if (ztest_opts.zo_verbose > 3) |
428870ff BB |
7260 | (void) printf("spa_next: found %s\n", spa_name(spa)); |
7261 | mutex_exit(&spa_namespace_lock); | |
7262 | ||
7263 | /* | |
7264 | * Verify that we can export the pool and reimport it under a | |
7265 | * different name. | |
7266 | */ | |
379ca9cf | 7267 | if ((ztest_random(2) == 0) && !ztest_opts.zo_mmp_test) { |
eca7b760 IK |
7268 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
7269 | (void) snprintf(name, sizeof (name), "%s_import", | |
c242c188 CS |
7270 | ztest_opts.zo_pool); |
7271 | ztest_spa_import_export(ztest_opts.zo_pool, name); | |
7272 | ztest_spa_import_export(name, ztest_opts.zo_pool); | |
428870ff BB |
7273 | } |
7274 | ||
7275 | kernel_fini(); | |
572e2857 BB |
7276 | |
7277 | list_destroy(&zcl.zcl_callbacks); | |
0e8d1b2d | 7278 | mutex_destroy(&zcl.zcl_callbacks_lock); |
2d9142c9 | 7279 | (void) pthread_rwlock_destroy(&ztest_name_lock); |
c242c188 | 7280 | mutex_destroy(&ztest_vdev_lock); |
d2734cce | 7281 | mutex_destroy(&ztest_checkpoint_lock); |
428870ff BB |
7282 | } |
7283 | ||
65c7cc49 | 7284 | static void |
34dc7c2f BB |
7285 | print_time(hrtime_t t, char *timebuf) |
7286 | { | |
7287 | hrtime_t s = t / NANOSEC; | |
7288 | hrtime_t m = s / 60; | |
7289 | hrtime_t h = m / 60; | |
7290 | hrtime_t d = h / 24; | |
7291 | ||
7292 | s -= m * 60; | |
7293 | m -= h * 60; | |
7294 | h -= d * 24; | |
7295 | ||
7296 | timebuf[0] = '\0'; | |
7297 | ||
7298 | if (d) | |
7299 | (void) sprintf(timebuf, | |
7300 | "%llud%02lluh%02llum%02llus", d, h, m, s); | |
7301 | else if (h) | |
7302 | (void) sprintf(timebuf, "%lluh%02llum%02llus", h, m, s); | |
7303 | else if (m) | |
7304 | (void) sprintf(timebuf, "%llum%02llus", m, s); | |
7305 | else | |
7306 | (void) sprintf(timebuf, "%llus", s); | |
7307 | } | |
7308 | ||
428870ff | 7309 | static nvlist_t * |
0bc8fd78 | 7310 | make_random_props(void) |
428870ff BB |
7311 | { |
7312 | nvlist_t *props; | |
7313 | ||
60b82074 | 7314 | VERIFY0(nvlist_alloc(&props, NV_UNIQUE_NAME, 0)); |
cc99f275 | 7315 | |
9ae529ec CS |
7316 | if (ztest_random(2) == 0) |
7317 | return (props); | |
60b82074 BB |
7318 | |
7319 | VERIFY0(nvlist_add_uint64(props, | |
7320 | zpool_prop_to_name(ZPOOL_PROP_AUTOREPLACE), 1)); | |
428870ff | 7321 | |
428870ff BB |
7322 | return (props); |
7323 | } | |
7324 | ||
34dc7c2f BB |
7325 | /* |
7326 | * Create a storage pool with the given name and initial vdev size. | |
428870ff | 7327 | * Then test spa_freeze() functionality. |
34dc7c2f BB |
7328 | */ |
7329 | static void | |
428870ff | 7330 | ztest_init(ztest_shared_t *zs) |
34dc7c2f BB |
7331 | { |
7332 | spa_t *spa; | |
428870ff | 7333 | nvlist_t *nvroot, *props; |
9ae529ec | 7334 | int i; |
428870ff | 7335 | |
c242c188 | 7336 | mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); |
d2734cce | 7337 | mutex_init(&ztest_checkpoint_lock, NULL, MUTEX_DEFAULT, NULL); |
2d9142c9 | 7338 | VERIFY0(pthread_rwlock_init(&ztest_name_lock, NULL)); |
34dc7c2f | 7339 | |
da92d5cb | 7340 | kernel_init(SPA_MODE_READ | SPA_MODE_WRITE); |
34dc7c2f BB |
7341 | |
7342 | /* | |
7343 | * Create the storage pool. | |
7344 | */ | |
c242c188 | 7345 | (void) spa_destroy(ztest_opts.zo_pool); |
428870ff BB |
7346 | ztest_shared->zs_vdev_next_leaf = 0; |
7347 | zs->zs_splits = 0; | |
c242c188 | 7348 | zs->zs_mirrors = ztest_opts.zo_mirrors; |
ea0b2538 | 7349 | nvroot = make_vdev_root(NULL, NULL, NULL, ztest_opts.zo_vdev_size, 0, |
cc99f275 | 7350 | NULL, ztest_opts.zo_raidz, zs->zs_mirrors, 1); |
428870ff | 7351 | props = make_random_props(); |
60b82074 BB |
7352 | |
7353 | /* | |
7354 | * We don't expect the pool to suspend unless maxfaults == 0, | |
7355 | * in which case ztest_fault_inject() temporarily takes away | |
7356 | * the only valid replica. | |
7357 | */ | |
7358 | VERIFY0(nvlist_add_uint64(props, | |
7359 | zpool_prop_to_name(ZPOOL_PROP_FAILUREMODE), | |
7360 | MAXFAULTS(zs) ? ZIO_FAILURE_MODE_PANIC : ZIO_FAILURE_MODE_WAIT)); | |
7361 | ||
9ae529ec CS |
7362 | for (i = 0; i < SPA_FEATURES; i++) { |
7363 | char *buf; | |
93e28d66 SD |
7364 | |
7365 | /* | |
7366 | * 75% chance of using the log space map feature. We want ztest | |
7367 | * to exercise both the code paths that use the log space map | |
7368 | * feature and the ones that don't. | |
7369 | */ | |
7370 | if (i == SPA_FEATURE_LOG_SPACEMAP && ztest_random(4) == 0) | |
7371 | continue; | |
7372 | ||
9ae529ec CS |
7373 | VERIFY3S(-1, !=, asprintf(&buf, "feature@%s", |
7374 | spa_feature_table[i].fi_uname)); | |
7375 | VERIFY3U(0, ==, nvlist_add_uint64(props, buf, 0)); | |
7376 | free(buf); | |
7377 | } | |
60b82074 BB |
7378 | |
7379 | VERIFY0(spa_create(ztest_opts.zo_pool, nvroot, props, NULL, NULL)); | |
34dc7c2f | 7380 | nvlist_free(nvroot); |
85802aa4 | 7381 | nvlist_free(props); |
34dc7c2f | 7382 | |
c242c188 CS |
7383 | VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
7384 | zs->zs_metaslab_sz = | |
7385 | 1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift; | |
34dc7c2f BB |
7386 | spa_close(spa, FTAG); |
7387 | ||
7388 | kernel_fini(); | |
428870ff | 7389 | |
379ca9cf OF |
7390 | if (!ztest_opts.zo_mmp_test) { |
7391 | ztest_run_zdb(ztest_opts.zo_pool); | |
7392 | ztest_freeze(); | |
7393 | ztest_run_zdb(ztest_opts.zo_pool); | |
7394 | } | |
c242c188 | 7395 | |
2d9142c9 | 7396 | (void) pthread_rwlock_destroy(&ztest_name_lock); |
c242c188 | 7397 | mutex_destroy(&ztest_vdev_lock); |
d2734cce | 7398 | mutex_destroy(&ztest_checkpoint_lock); |
c242c188 CS |
7399 | } |
7400 | ||
7401 | static void | |
9d81146b | 7402 | setup_data_fd(void) |
c242c188 | 7403 | { |
facbbe43 BB |
7404 | static char ztest_name_data[] = "/tmp/ztest.data.XXXXXX"; |
7405 | ||
7406 | ztest_fd_data = mkstemp(ztest_name_data); | |
9d81146b | 7407 | ASSERT3S(ztest_fd_data, >=, 0); |
facbbe43 | 7408 | (void) unlink(ztest_name_data); |
c242c188 CS |
7409 | } |
7410 | ||
22257dc0 CS |
7411 | static int |
7412 | shared_data_size(ztest_shared_hdr_t *hdr) | |
7413 | { | |
7414 | int size; | |
7415 | ||
7416 | size = hdr->zh_hdr_size; | |
7417 | size += hdr->zh_opts_size; | |
7418 | size += hdr->zh_size; | |
7419 | size += hdr->zh_stats_size * hdr->zh_stats_count; | |
7420 | size += hdr->zh_ds_size * hdr->zh_ds_count; | |
7421 | ||
7422 | return (size); | |
7423 | } | |
7424 | ||
c242c188 CS |
7425 | static void |
7426 | setup_hdr(void) | |
7427 | { | |
22257dc0 | 7428 | int size; |
c242c188 CS |
7429 | ztest_shared_hdr_t *hdr; |
7430 | ||
7431 | hdr = (void *)mmap(0, P2ROUNDUP(sizeof (*hdr), getpagesize()), | |
9d81146b | 7432 | PROT_READ | PROT_WRITE, MAP_SHARED, ztest_fd_data, 0); |
799402d8 | 7433 | ASSERT(hdr != MAP_FAILED); |
c242c188 | 7434 | |
9d81146b | 7435 | VERIFY3U(0, ==, ftruncate(ztest_fd_data, sizeof (ztest_shared_hdr_t))); |
22257dc0 | 7436 | |
c242c188 CS |
7437 | hdr->zh_hdr_size = sizeof (ztest_shared_hdr_t); |
7438 | hdr->zh_opts_size = sizeof (ztest_shared_opts_t); | |
7439 | hdr->zh_size = sizeof (ztest_shared_t); | |
7440 | hdr->zh_stats_size = sizeof (ztest_shared_callstate_t); | |
7441 | hdr->zh_stats_count = ZTEST_FUNCS; | |
7442 | hdr->zh_ds_size = sizeof (ztest_shared_ds_t); | |
7443 | hdr->zh_ds_count = ztest_opts.zo_datasets; | |
7444 | ||
22257dc0 | 7445 | size = shared_data_size(hdr); |
9d81146b | 7446 | VERIFY3U(0, ==, ftruncate(ztest_fd_data, size)); |
22257dc0 | 7447 | |
c242c188 CS |
7448 | (void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize())); |
7449 | } | |
7450 | ||
7451 | static void | |
7452 | setup_data(void) | |
7453 | { | |
7454 | int size, offset; | |
7455 | ztest_shared_hdr_t *hdr; | |
7456 | uint8_t *buf; | |
7457 | ||
7458 | hdr = (void *)mmap(0, P2ROUNDUP(sizeof (*hdr), getpagesize()), | |
9d81146b | 7459 | PROT_READ, MAP_SHARED, ztest_fd_data, 0); |
799402d8 | 7460 | ASSERT(hdr != MAP_FAILED); |
c242c188 | 7461 | |
22257dc0 | 7462 | size = shared_data_size(hdr); |
c242c188 CS |
7463 | |
7464 | (void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize())); | |
7465 | hdr = ztest_shared_hdr = (void *)mmap(0, P2ROUNDUP(size, getpagesize()), | |
9d81146b | 7466 | PROT_READ | PROT_WRITE, MAP_SHARED, ztest_fd_data, 0); |
799402d8 | 7467 | ASSERT(hdr != MAP_FAILED); |
c242c188 CS |
7468 | buf = (uint8_t *)hdr; |
7469 | ||
7470 | offset = hdr->zh_hdr_size; | |
7471 | ztest_shared_opts = (void *)&buf[offset]; | |
7472 | offset += hdr->zh_opts_size; | |
7473 | ztest_shared = (void *)&buf[offset]; | |
7474 | offset += hdr->zh_size; | |
7475 | ztest_shared_callstate = (void *)&buf[offset]; | |
7476 | offset += hdr->zh_stats_size * hdr->zh_stats_count; | |
7477 | ztest_shared_ds = (void *)&buf[offset]; | |
7478 | } | |
7479 | ||
7480 | static boolean_t | |
7481 | exec_child(char *cmd, char *libpath, boolean_t ignorekill, int *statusp) | |
7482 | { | |
7483 | pid_t pid; | |
7484 | int status; | |
483106eb | 7485 | char *cmdbuf = NULL; |
c242c188 CS |
7486 | |
7487 | pid = fork(); | |
7488 | ||
7489 | if (cmd == NULL) { | |
483106eb BB |
7490 | cmdbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
7491 | (void) strlcpy(cmdbuf, getexecname(), MAXPATHLEN); | |
c242c188 CS |
7492 | cmd = cmdbuf; |
7493 | } | |
7494 | ||
7495 | if (pid == -1) | |
7496 | fatal(1, "fork failed"); | |
7497 | ||
7498 | if (pid == 0) { /* child */ | |
7499 | char *emptyargv[2] = { cmd, NULL }; | |
9d81146b | 7500 | char fd_data_str[12]; |
c242c188 CS |
7501 | |
7502 | struct rlimit rl = { 1024, 1024 }; | |
7503 | (void) setrlimit(RLIMIT_NOFILE, &rl); | |
9d81146b ED |
7504 | |
7505 | (void) close(ztest_fd_rand); | |
7506 | VERIFY(11 >= snprintf(fd_data_str, 12, "%d", ztest_fd_data)); | |
7507 | VERIFY(0 == setenv("ZTEST_FD_DATA", fd_data_str, 1)); | |
7508 | ||
c242c188 CS |
7509 | (void) enable_extended_FILE_stdio(-1, -1); |
7510 | if (libpath != NULL) | |
7511 | VERIFY(0 == setenv("LD_LIBRARY_PATH", libpath, 1)); | |
7512 | (void) execv(cmd, emptyargv); | |
7513 | ztest_dump_core = B_FALSE; | |
7514 | fatal(B_TRUE, "exec failed: %s", cmd); | |
7515 | } | |
7516 | ||
483106eb BB |
7517 | if (cmdbuf != NULL) { |
7518 | umem_free(cmdbuf, MAXPATHLEN); | |
7519 | cmd = NULL; | |
7520 | } | |
7521 | ||
c242c188 CS |
7522 | while (waitpid(pid, &status, 0) != pid) |
7523 | continue; | |
7524 | if (statusp != NULL) | |
7525 | *statusp = status; | |
7526 | ||
7527 | if (WIFEXITED(status)) { | |
7528 | if (WEXITSTATUS(status) != 0) { | |
7529 | (void) fprintf(stderr, "child exited with code %d\n", | |
7530 | WEXITSTATUS(status)); | |
7531 | exit(2); | |
7532 | } | |
7533 | return (B_FALSE); | |
7534 | } else if (WIFSIGNALED(status)) { | |
7535 | if (!ignorekill || WTERMSIG(status) != SIGKILL) { | |
7536 | (void) fprintf(stderr, "child died with signal %d\n", | |
7537 | WTERMSIG(status)); | |
7538 | exit(3); | |
7539 | } | |
7540 | return (B_TRUE); | |
7541 | } else { | |
7542 | (void) fprintf(stderr, "something strange happened to child\n"); | |
7543 | exit(4); | |
7544 | /* NOTREACHED */ | |
7545 | } | |
7546 | } | |
7547 | ||
7548 | static void | |
7549 | ztest_run_init(void) | |
7550 | { | |
7551 | int i; | |
428870ff | 7552 | |
c242c188 | 7553 | ztest_shared_t *zs = ztest_shared; |
428870ff | 7554 | |
c242c188 CS |
7555 | /* |
7556 | * Blow away any existing copy of zpool.cache | |
7557 | */ | |
7558 | (void) remove(spa_config_path); | |
7559 | ||
379ca9cf OF |
7560 | if (ztest_opts.zo_init == 0) { |
7561 | if (ztest_opts.zo_verbose >= 1) | |
7562 | (void) printf("Importing pool %s\n", | |
7563 | ztest_opts.zo_pool); | |
7564 | ztest_import(zs); | |
7565 | return; | |
7566 | } | |
7567 | ||
c242c188 CS |
7568 | /* |
7569 | * Create and initialize our storage pool. | |
7570 | */ | |
7571 | for (i = 1; i <= ztest_opts.zo_init; i++) { | |
7572 | bzero(zs, sizeof (ztest_shared_t)); | |
7573 | if (ztest_opts.zo_verbose >= 3 && | |
7574 | ztest_opts.zo_init != 1) { | |
7575 | (void) printf("ztest_init(), pass %d\n", i); | |
7576 | } | |
7577 | ztest_init(zs); | |
7578 | } | |
34dc7c2f BB |
7579 | } |
7580 | ||
7581 | int | |
7582 | main(int argc, char **argv) | |
7583 | { | |
7584 | int kills = 0; | |
7585 | int iters = 0; | |
c242c188 CS |
7586 | int older = 0; |
7587 | int newer = 0; | |
34dc7c2f BB |
7588 | ztest_shared_t *zs; |
7589 | ztest_info_t *zi; | |
c242c188 | 7590 | ztest_shared_callstate_t *zc; |
34dc7c2f | 7591 | char timebuf[100]; |
f3c8c9e6 | 7592 | char numbuf[NN_NUMBUF_SZ]; |
483106eb | 7593 | char *cmd; |
c242c188 CS |
7594 | boolean_t hasalt; |
7595 | int f; | |
9d81146b | 7596 | char *fd_data_str = getenv("ZTEST_FD_DATA"); |
e82cdc3a | 7597 | struct sigaction action; |
34dc7c2f BB |
7598 | |
7599 | (void) setvbuf(stdout, NULL, _IOLBF, 0); | |
7600 | ||
498877ba | 7601 | dprintf_setup(&argc, argv); |
8fb1ede1 | 7602 | zfs_deadman_synctime_ms = 300000; |
94102578 | 7603 | zfs_deadman_checktime_ms = 30000; |
4d044c4c SD |
7604 | /* |
7605 | * As two-word space map entries may not come up often (especially | |
7606 | * if pool and vdev sizes are small) we want to force at least some | |
7607 | * of them so the feature get tested. | |
7608 | */ | |
7609 | zfs_force_some_double_word_sm_entries = B_TRUE; | |
498877ba | 7610 | |
1258bd77 BB |
7611 | /* |
7612 | * Verify that even extensively damaged split blocks with many | |
7613 | * segments can be reconstructed in a reasonable amount of time | |
7614 | * when reconstruction is known to be possible. | |
4a7eb69a TC |
7615 | * |
7616 | * Note: the lower this value is, the more damage we inflict, and | |
7617 | * the more time ztest spends in recovering that damage. We chose | |
7618 | * to induce damage 1/100th of the time so recovery is tested but | |
7619 | * not so frequently that ztest doesn't get to test other code paths. | |
1258bd77 | 7620 | */ |
4a7eb69a | 7621 | zfs_reconstruct_indirect_damage_fraction = 100; |
1258bd77 | 7622 | |
e82cdc3a NB |
7623 | action.sa_handler = sig_handler; |
7624 | sigemptyset(&action.sa_mask); | |
7625 | action.sa_flags = 0; | |
7626 | ||
7627 | if (sigaction(SIGSEGV, &action, NULL) < 0) { | |
7628 | (void) fprintf(stderr, "ztest: cannot catch SIGSEGV: %s.\n", | |
7629 | strerror(errno)); | |
7630 | exit(EXIT_FAILURE); | |
7631 | } | |
7632 | ||
7633 | if (sigaction(SIGABRT, &action, NULL) < 0) { | |
7634 | (void) fprintf(stderr, "ztest: cannot catch SIGABRT: %s.\n", | |
7635 | strerror(errno)); | |
7636 | exit(EXIT_FAILURE); | |
7637 | } | |
7638 | ||
e1a0850c BB |
7639 | /* |
7640 | * Force random_get_bytes() to use /dev/urandom in order to prevent | |
7641 | * ztest from needlessly depleting the system entropy pool. | |
7642 | */ | |
7643 | random_path = "/dev/urandom"; | |
7644 | ztest_fd_rand = open(random_path, O_RDONLY); | |
9d81146b ED |
7645 | ASSERT3S(ztest_fd_rand, >=, 0); |
7646 | ||
7647 | if (!fd_data_str) { | |
c242c188 | 7648 | process_options(argc, argv); |
34dc7c2f | 7649 | |
9d81146b | 7650 | setup_data_fd(); |
c242c188 CS |
7651 | setup_hdr(); |
7652 | setup_data(); | |
7653 | bcopy(&ztest_opts, ztest_shared_opts, | |
7654 | sizeof (*ztest_shared_opts)); | |
7655 | } else { | |
9d81146b | 7656 | ztest_fd_data = atoi(fd_data_str); |
c242c188 CS |
7657 | setup_data(); |
7658 | bcopy(ztest_shared_opts, &ztest_opts, sizeof (ztest_opts)); | |
7659 | } | |
7660 | ASSERT3U(ztest_opts.zo_datasets, ==, ztest_shared_hdr->zh_ds_count); | |
34dc7c2f | 7661 | |
428870ff | 7662 | /* Override location of zpool.cache */ |
5be98cfe BB |
7663 | VERIFY(asprintf((char **)&spa_config_path, "%s/zpool.cache", |
7664 | ztest_opts.zo_dir) != -1); | |
9d81146b | 7665 | |
c242c188 CS |
7666 | ztest_ds = umem_alloc(ztest_opts.zo_datasets * sizeof (ztest_ds_t), |
7667 | UMEM_NOFAIL); | |
7668 | zs = ztest_shared; | |
7669 | ||
9d81146b | 7670 | if (fd_data_str) { |
d830d479 | 7671 | metaslab_force_ganging = ztest_opts.zo_metaslab_force_ganging; |
c242c188 CS |
7672 | metaslab_df_alloc_threshold = |
7673 | zs->zs_metaslab_df_alloc_threshold; | |
7674 | ||
7675 | if (zs->zs_do_init) | |
7676 | ztest_run_init(); | |
7677 | else | |
7678 | ztest_run(zs); | |
7679 | exit(0); | |
7680 | } | |
34dc7c2f | 7681 | |
c242c188 | 7682 | hasalt = (strlen(ztest_opts.zo_alt_ztest) != 0); |
34dc7c2f | 7683 | |
c242c188 | 7684 | if (ztest_opts.zo_verbose >= 1) { |
34dc7c2f BB |
7685 | (void) printf("%llu vdevs, %d datasets, %d threads," |
7686 | " %llu seconds...\n", | |
c242c188 CS |
7687 | (u_longlong_t)ztest_opts.zo_vdevs, |
7688 | ztest_opts.zo_datasets, | |
7689 | ztest_opts.zo_threads, | |
7690 | (u_longlong_t)ztest_opts.zo_time); | |
34dc7c2f BB |
7691 | } |
7692 | ||
483106eb BB |
7693 | cmd = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); |
7694 | (void) strlcpy(cmd, getexecname(), MAXNAMELEN); | |
c242c188 CS |
7695 | |
7696 | zs->zs_do_init = B_TRUE; | |
7697 | if (strlen(ztest_opts.zo_alt_ztest) != 0) { | |
7698 | if (ztest_opts.zo_verbose >= 1) { | |
7699 | (void) printf("Executing older ztest for " | |
7700 | "initialization: %s\n", ztest_opts.zo_alt_ztest); | |
7701 | } | |
7702 | VERIFY(!exec_child(ztest_opts.zo_alt_ztest, | |
7703 | ztest_opts.zo_alt_libpath, B_FALSE, NULL)); | |
7704 | } else { | |
7705 | VERIFY(!exec_child(NULL, NULL, B_FALSE, NULL)); | |
34dc7c2f | 7706 | } |
c242c188 | 7707 | zs->zs_do_init = B_FALSE; |
34dc7c2f | 7708 | |
428870ff | 7709 | zs->zs_proc_start = gethrtime(); |
c242c188 | 7710 | zs->zs_proc_stop = zs->zs_proc_start + ztest_opts.zo_time * NANOSEC; |
34dc7c2f | 7711 | |
d6320ddb | 7712 | for (f = 0; f < ZTEST_FUNCS; f++) { |
c242c188 CS |
7713 | zi = &ztest_info[f]; |
7714 | zc = ZTEST_GET_SHARED_CALLSTATE(f); | |
428870ff | 7715 | if (zs->zs_proc_start + zi->zi_interval[0] > zs->zs_proc_stop) |
c242c188 | 7716 | zc->zc_next = UINT64_MAX; |
34dc7c2f | 7717 | else |
c242c188 | 7718 | zc->zc_next = zs->zs_proc_start + |
428870ff | 7719 | ztest_random(2 * zi->zi_interval[0] + 1); |
34dc7c2f BB |
7720 | } |
7721 | ||
34dc7c2f BB |
7722 | /* |
7723 | * Run the tests in a loop. These tests include fault injection | |
7724 | * to verify that self-healing data works, and forced crashes | |
7725 | * to verify that we never lose on-disk consistency. | |
7726 | */ | |
428870ff | 7727 | while (gethrtime() < zs->zs_proc_stop) { |
34dc7c2f | 7728 | int status; |
c242c188 | 7729 | boolean_t killed; |
34dc7c2f BB |
7730 | |
7731 | /* | |
7732 | * Initialize the workload counters for each function. | |
7733 | */ | |
d6320ddb | 7734 | for (f = 0; f < ZTEST_FUNCS; f++) { |
c242c188 CS |
7735 | zc = ZTEST_GET_SHARED_CALLSTATE(f); |
7736 | zc->zc_count = 0; | |
7737 | zc->zc_time = 0; | |
34dc7c2f BB |
7738 | } |
7739 | ||
9babb374 | 7740 | /* Set the allocation switch size */ |
c242c188 CS |
7741 | zs->zs_metaslab_df_alloc_threshold = |
7742 | ztest_random(zs->zs_metaslab_sz / 4) + 1; | |
9babb374 | 7743 | |
c242c188 CS |
7744 | if (!hasalt || ztest_random(2) == 0) { |
7745 | if (hasalt && ztest_opts.zo_verbose >= 1) { | |
7746 | (void) printf("Executing newer ztest: %s\n", | |
7747 | cmd); | |
34dc7c2f | 7748 | } |
c242c188 CS |
7749 | newer++; |
7750 | killed = exec_child(cmd, NULL, B_TRUE, &status); | |
34dc7c2f | 7751 | } else { |
c242c188 CS |
7752 | if (hasalt && ztest_opts.zo_verbose >= 1) { |
7753 | (void) printf("Executing older ztest: %s\n", | |
7754 | ztest_opts.zo_alt_ztest); | |
7755 | } | |
7756 | older++; | |
7757 | killed = exec_child(ztest_opts.zo_alt_ztest, | |
7758 | ztest_opts.zo_alt_libpath, B_TRUE, &status); | |
34dc7c2f BB |
7759 | } |
7760 | ||
c242c188 CS |
7761 | if (killed) |
7762 | kills++; | |
34dc7c2f BB |
7763 | iters++; |
7764 | ||
c242c188 | 7765 | if (ztest_opts.zo_verbose >= 1) { |
34dc7c2f BB |
7766 | hrtime_t now = gethrtime(); |
7767 | ||
428870ff BB |
7768 | now = MIN(now, zs->zs_proc_stop); |
7769 | print_time(zs->zs_proc_stop - now, timebuf); | |
f3c8c9e6 | 7770 | nicenum(zs->zs_space, numbuf, sizeof (numbuf)); |
34dc7c2f BB |
7771 | |
7772 | (void) printf("Pass %3d, %8s, %3llu ENOSPC, " | |
7773 | "%4.1f%% of %5s used, %3.0f%% done, %8s to go\n", | |
7774 | iters, | |
7775 | WIFEXITED(status) ? "Complete" : "SIGKILL", | |
7776 | (u_longlong_t)zs->zs_enospc_count, | |
7777 | 100.0 * zs->zs_alloc / zs->zs_space, | |
7778 | numbuf, | |
428870ff | 7779 | 100.0 * (now - zs->zs_proc_start) / |
c242c188 | 7780 | (ztest_opts.zo_time * NANOSEC), timebuf); |
34dc7c2f BB |
7781 | } |
7782 | ||
c242c188 | 7783 | if (ztest_opts.zo_verbose >= 2) { |
34dc7c2f BB |
7784 | (void) printf("\nWorkload summary:\n\n"); |
7785 | (void) printf("%7s %9s %s\n", | |
7786 | "Calls", "Time", "Function"); | |
7787 | (void) printf("%7s %9s %s\n", | |
7788 | "-----", "----", "--------"); | |
d6320ddb | 7789 | for (f = 0; f < ZTEST_FUNCS; f++) { |
c242c188 CS |
7790 | zi = &ztest_info[f]; |
7791 | zc = ZTEST_GET_SHARED_CALLSTATE(f); | |
7792 | print_time(zc->zc_time, timebuf); | |
34dc7c2f | 7793 | (void) printf("%7llu %9s %s\n", |
c242c188 | 7794 | (u_longlong_t)zc->zc_count, timebuf, |
fdc5d982 | 7795 | zi->zi_funcname); |
34dc7c2f BB |
7796 | } |
7797 | (void) printf("\n"); | |
7798 | } | |
7799 | ||
379ca9cf OF |
7800 | if (!ztest_opts.zo_mmp_test) |
7801 | ztest_run_zdb(ztest_opts.zo_pool); | |
428870ff | 7802 | } |
34dc7c2f | 7803 | |
c242c188 CS |
7804 | if (ztest_opts.zo_verbose >= 1) { |
7805 | if (hasalt) { | |
7806 | (void) printf("%d runs of older ztest: %s\n", older, | |
7807 | ztest_opts.zo_alt_ztest); | |
7808 | (void) printf("%d runs of newer ztest: %s\n", newer, | |
7809 | cmd); | |
7810 | } | |
34dc7c2f BB |
7811 | (void) printf("%d killed, %d completed, %.0f%% kill rate\n", |
7812 | kills, iters - kills, (100.0 * kills) / MAX(1, iters)); | |
7813 | } | |
7814 | ||
483106eb BB |
7815 | umem_free(cmd, MAXNAMELEN); |
7816 | ||
34dc7c2f BB |
7817 | return (0); |
7818 | } |