]>
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> |
b2255edc | 107 | #include <sys/vdev_draid.h> |
34dc7c2f | 108 | #include <sys/vdev_impl.h> |
b128c09f | 109 | #include <sys/vdev_file.h> |
619f0976 | 110 | #include <sys/vdev_initialize.h> |
e5db3134 | 111 | #include <sys/vdev_raidz.h> |
1b939560 | 112 | #include <sys/vdev_trim.h> |
34dc7c2f | 113 | #include <sys/spa_impl.h> |
428870ff | 114 | #include <sys/metaslab_impl.h> |
34dc7c2f | 115 | #include <sys/dsl_prop.h> |
9babb374 | 116 | #include <sys/dsl_dataset.h> |
13fe0198 | 117 | #include <sys/dsl_destroy.h> |
428870ff BB |
118 | #include <sys/dsl_scan.h> |
119 | #include <sys/zio_checksum.h> | |
27d96d22 | 120 | #include <sys/zfs_refcount.h> |
9ae529ec | 121 | #include <sys/zfeature.h> |
13fe0198 | 122 | #include <sys/dsl_userhold.h> |
a6255b7f | 123 | #include <sys/abd.h> |
34dc7c2f | 124 | #include <stdio.h> |
34dc7c2f BB |
125 | #include <stdlib.h> |
126 | #include <unistd.h> | |
127 | #include <signal.h> | |
128 | #include <umem.h> | |
34dc7c2f BB |
129 | #include <ctype.h> |
130 | #include <math.h> | |
131 | #include <sys/fs/zfs.h> | |
1eeb4562 | 132 | #include <zfs_fletcher.h> |
428870ff | 133 | #include <libnvpair.h> |
e89f1295 | 134 | #include <libzutil.h> |
010d1247 | 135 | #include <sys/crypto/icp.h> |
967798d0 | 136 | #ifdef __GLIBC__ |
e82cdc3a NB |
137 | #include <execinfo.h> /* for backtrace() */ |
138 | #endif | |
34dc7c2f | 139 | |
9d81146b ED |
140 | static int ztest_fd_data = -1; |
141 | static int ztest_fd_rand = -1; | |
c242c188 CS |
142 | |
143 | typedef struct ztest_shared_hdr { | |
144 | uint64_t zh_hdr_size; | |
145 | uint64_t zh_opts_size; | |
146 | uint64_t zh_size; | |
147 | uint64_t zh_stats_size; | |
148 | uint64_t zh_stats_count; | |
149 | uint64_t zh_ds_size; | |
150 | uint64_t zh_ds_count; | |
151 | } ztest_shared_hdr_t; | |
152 | ||
153 | static ztest_shared_hdr_t *ztest_shared_hdr; | |
154 | ||
cc99f275 DB |
155 | enum ztest_class_state { |
156 | ZTEST_VDEV_CLASS_OFF, | |
157 | ZTEST_VDEV_CLASS_ON, | |
158 | ZTEST_VDEV_CLASS_RND | |
159 | }; | |
160 | ||
c242c188 | 161 | typedef struct ztest_shared_opts { |
eca7b760 IK |
162 | char zo_pool[ZFS_MAX_DATASET_NAME_LEN]; |
163 | char zo_dir[ZFS_MAX_DATASET_NAME_LEN]; | |
c242c188 CS |
164 | char zo_alt_ztest[MAXNAMELEN]; |
165 | char zo_alt_libpath[MAXNAMELEN]; | |
166 | uint64_t zo_vdevs; | |
167 | uint64_t zo_vdevtime; | |
168 | size_t zo_vdev_size; | |
169 | int zo_ashift; | |
170 | int zo_mirrors; | |
b2255edc BB |
171 | int zo_raid_children; |
172 | int zo_raid_parity; | |
173 | char zo_raid_type[8]; | |
174 | int zo_draid_data; | |
175 | int zo_draid_spares; | |
c242c188 CS |
176 | int zo_datasets; |
177 | int zo_threads; | |
178 | uint64_t zo_passtime; | |
179 | uint64_t zo_killrate; | |
180 | int zo_verbose; | |
181 | int zo_init; | |
182 | uint64_t zo_time; | |
183 | uint64_t zo_maxloops; | |
d830d479 | 184 | uint64_t zo_metaslab_force_ganging; |
379ca9cf | 185 | int zo_mmp_test; |
cc99f275 | 186 | int zo_special_vdevs; |
ab4c009e | 187 | int zo_dump_dbgmsg; |
c242c188 CS |
188 | } ztest_shared_opts_t; |
189 | ||
190 | static const ztest_shared_opts_t ztest_opts_defaults = { | |
caacc6e4 TH |
191 | .zo_pool = "ztest", |
192 | .zo_dir = "/tmp", | |
c242c188 CS |
193 | .zo_alt_ztest = { '\0' }, |
194 | .zo_alt_libpath = { '\0' }, | |
195 | .zo_vdevs = 5, | |
196 | .zo_ashift = SPA_MINBLOCKSHIFT, | |
197 | .zo_mirrors = 2, | |
b2255edc BB |
198 | .zo_raid_children = 4, |
199 | .zo_raid_parity = 1, | |
200 | .zo_raid_type = VDEV_TYPE_RAIDZ, | |
4e21fd06 | 201 | .zo_vdev_size = SPA_MINDEVSIZE * 4, /* 256m default size */ |
b2255edc BB |
202 | .zo_draid_data = 4, /* data drives */ |
203 | .zo_draid_spares = 1, /* distributed spares */ | |
c242c188 CS |
204 | .zo_datasets = 7, |
205 | .zo_threads = 23, | |
206 | .zo_passtime = 60, /* 60 seconds */ | |
207 | .zo_killrate = 70, /* 70% kill rate */ | |
208 | .zo_verbose = 0, | |
379ca9cf | 209 | .zo_mmp_test = 0, |
c242c188 CS |
210 | .zo_init = 1, |
211 | .zo_time = 300, /* 5 minutes */ | |
212 | .zo_maxloops = 50, /* max loops during spa_freeze() */ | |
09b85f2d | 213 | .zo_metaslab_force_ganging = 64 << 10, |
cc99f275 | 214 | .zo_special_vdevs = ZTEST_VDEV_CLASS_RND, |
c242c188 CS |
215 | }; |
216 | ||
d830d479 | 217 | extern uint64_t metaslab_force_ganging; |
c242c188 | 218 | extern uint64_t metaslab_df_alloc_threshold; |
8fb1ede1 | 219 | extern unsigned long zfs_deadman_synctime_ms; |
080b3100 | 220 | extern int metaslab_preload_limit; |
d3c2ae1c | 221 | extern boolean_t zfs_compressed_arc_enabled; |
047116ac TC |
222 | extern int zfs_abd_scatter_enabled; |
223 | extern int dmu_object_alloc_chunk_shift; | |
4d044c4c | 224 | extern boolean_t zfs_force_some_double_word_sm_entries; |
c3bd3fb4 | 225 | extern unsigned long zio_decompress_fail_fraction; |
1258bd77 | 226 | extern unsigned long zfs_reconstruct_indirect_damage_fraction; |
65ca2c1e | 227 | |
c242c188 CS |
228 | |
229 | static ztest_shared_opts_t *ztest_shared_opts; | |
230 | static ztest_shared_opts_t ztest_opts; | |
4807c0ba | 231 | static char *ztest_wkeydata = "abcdefghijklmnopqrstuvwxyz012345"; |
c242c188 CS |
232 | |
233 | typedef struct ztest_shared_ds { | |
234 | uint64_t zd_seq; | |
235 | } ztest_shared_ds_t; | |
236 | ||
237 | static ztest_shared_ds_t *ztest_shared_ds; | |
238 | #define ZTEST_GET_SHARED_DS(d) (&ztest_shared_ds[d]) | |
428870ff BB |
239 | |
240 | #define BT_MAGIC 0x123456789abcdefULL | |
60b82074 | 241 | #define MAXFAULTS(zs) \ |
b2255edc | 242 | (MAX((zs)->zs_mirrors, 1) * (ztest_opts.zo_raid_parity + 1) - 1) |
428870ff BB |
243 | |
244 | enum ztest_io_type { | |
245 | ZTEST_IO_WRITE_TAG, | |
246 | ZTEST_IO_WRITE_PATTERN, | |
247 | ZTEST_IO_WRITE_ZEROES, | |
248 | ZTEST_IO_TRUNCATE, | |
249 | ZTEST_IO_SETATTR, | |
03c6040b | 250 | ZTEST_IO_REWRITE, |
428870ff BB |
251 | ZTEST_IO_TYPES |
252 | }; | |
34dc7c2f BB |
253 | |
254 | typedef struct ztest_block_tag { | |
428870ff | 255 | uint64_t bt_magic; |
34dc7c2f BB |
256 | uint64_t bt_objset; |
257 | uint64_t bt_object; | |
50c957f7 | 258 | uint64_t bt_dnodesize; |
34dc7c2f | 259 | uint64_t bt_offset; |
428870ff | 260 | uint64_t bt_gen; |
34dc7c2f | 261 | uint64_t bt_txg; |
428870ff | 262 | uint64_t bt_crtxg; |
34dc7c2f BB |
263 | } ztest_block_tag_t; |
264 | ||
428870ff BB |
265 | typedef struct bufwad { |
266 | uint64_t bw_index; | |
267 | uint64_t bw_txg; | |
268 | uint64_t bw_data; | |
269 | } bufwad_t; | |
270 | ||
5d43cc9a MA |
271 | /* |
272 | * It would be better to use a rangelock_t per object. Unfortunately | |
273 | * the rangelock_t is not a drop-in replacement for rl_t, because we | |
274 | * still need to map from object ID to rangelock_t. | |
275 | */ | |
276 | typedef enum { | |
277 | RL_READER, | |
278 | RL_WRITER, | |
279 | RL_APPEND | |
280 | } rl_type_t; | |
281 | ||
428870ff BB |
282 | typedef struct rll { |
283 | void *rll_writer; | |
284 | int rll_readers; | |
1e33ac1e BB |
285 | kmutex_t rll_lock; |
286 | kcondvar_t rll_cv; | |
428870ff BB |
287 | } rll_t; |
288 | ||
5d43cc9a MA |
289 | typedef struct rl { |
290 | uint64_t rl_object; | |
291 | uint64_t rl_offset; | |
292 | uint64_t rl_size; | |
293 | rll_t *rl_lock; | |
294 | } rl_t; | |
428870ff BB |
295 | |
296 | #define ZTEST_RANGE_LOCKS 64 | |
297 | #define ZTEST_OBJECT_LOCKS 64 | |
298 | ||
299 | /* | |
300 | * Object descriptor. Used as a template for object lookup/create/remove. | |
301 | */ | |
302 | typedef struct ztest_od { | |
303 | uint64_t od_dir; | |
304 | uint64_t od_object; | |
305 | dmu_object_type_t od_type; | |
306 | dmu_object_type_t od_crtype; | |
307 | uint64_t od_blocksize; | |
308 | uint64_t od_crblocksize; | |
50c957f7 | 309 | uint64_t od_crdnodesize; |
428870ff BB |
310 | uint64_t od_gen; |
311 | uint64_t od_crgen; | |
eca7b760 | 312 | char od_name[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff | 313 | } ztest_od_t; |
34dc7c2f | 314 | |
428870ff BB |
315 | /* |
316 | * Per-dataset state. | |
317 | */ | |
318 | typedef struct ztest_ds { | |
c242c188 | 319 | ztest_shared_ds_t *zd_shared; |
428870ff | 320 | objset_t *zd_os; |
2d9142c9 | 321 | pthread_rwlock_t zd_zilog_lock; |
428870ff | 322 | zilog_t *zd_zilog; |
428870ff | 323 | ztest_od_t *zd_od; /* debugging aid */ |
eca7b760 | 324 | char zd_name[ZFS_MAX_DATASET_NAME_LEN]; |
1e33ac1e | 325 | kmutex_t zd_dirobj_lock; |
428870ff | 326 | rll_t zd_object_lock[ZTEST_OBJECT_LOCKS]; |
5d43cc9a | 327 | rll_t zd_range_lock[ZTEST_RANGE_LOCKS]; |
428870ff BB |
328 | } ztest_ds_t; |
329 | ||
330 | /* | |
331 | * Per-iteration state. | |
332 | */ | |
333 | typedef void ztest_func_t(ztest_ds_t *zd, uint64_t id); | |
334 | ||
335 | typedef struct ztest_info { | |
336 | ztest_func_t *zi_func; /* test function */ | |
337 | uint64_t zi_iters; /* iterations per execution */ | |
338 | uint64_t *zi_interval; /* execute every <interval> seconds */ | |
fdc5d982 | 339 | const char *zi_funcname; /* name of test function */ |
428870ff | 340 | } ztest_info_t; |
34dc7c2f | 341 | |
c242c188 CS |
342 | typedef struct ztest_shared_callstate { |
343 | uint64_t zc_count; /* per-pass count */ | |
344 | uint64_t zc_time; /* per-pass time */ | |
345 | uint64_t zc_next; /* next time to call this function */ | |
346 | } ztest_shared_callstate_t; | |
347 | ||
348 | static ztest_shared_callstate_t *ztest_shared_callstate; | |
349 | #define ZTEST_GET_SHARED_CALLSTATE(c) (&ztest_shared_callstate[c]) | |
350 | ||
34dc7c2f BB |
351 | ztest_func_t ztest_dmu_read_write; |
352 | ztest_func_t ztest_dmu_write_parallel; | |
353 | ztest_func_t ztest_dmu_object_alloc_free; | |
047116ac | 354 | ztest_func_t ztest_dmu_object_next_chunk; |
428870ff | 355 | ztest_func_t ztest_dmu_commit_callbacks; |
34dc7c2f BB |
356 | ztest_func_t ztest_zap; |
357 | ztest_func_t ztest_zap_parallel; | |
428870ff | 358 | ztest_func_t ztest_zil_commit; |
3e31d2b0 | 359 | ztest_func_t ztest_zil_remount; |
428870ff | 360 | ztest_func_t ztest_dmu_read_write_zcopy; |
34dc7c2f | 361 | ztest_func_t ztest_dmu_objset_create_destroy; |
428870ff BB |
362 | ztest_func_t ztest_dmu_prealloc; |
363 | ztest_func_t ztest_fzap; | |
34dc7c2f | 364 | ztest_func_t ztest_dmu_snapshot_create_destroy; |
428870ff BB |
365 | ztest_func_t ztest_dsl_prop_get_set; |
366 | ztest_func_t ztest_spa_prop_get_set; | |
34dc7c2f BB |
367 | ztest_func_t ztest_spa_create_destroy; |
368 | ztest_func_t ztest_fault_inject; | |
428870ff | 369 | ztest_func_t ztest_dmu_snapshot_hold; |
0582e403 | 370 | ztest_func_t ztest_mmp_enable_disable; |
428870ff BB |
371 | ztest_func_t ztest_scrub; |
372 | ztest_func_t ztest_dsl_dataset_promote_busy; | |
34dc7c2f BB |
373 | ztest_func_t ztest_vdev_attach_detach; |
374 | ztest_func_t ztest_vdev_LUN_growth; | |
375 | ztest_func_t ztest_vdev_add_remove; | |
cc99f275 | 376 | ztest_func_t ztest_vdev_class_add; |
b128c09f | 377 | ztest_func_t ztest_vdev_aux_add_remove; |
428870ff | 378 | ztest_func_t ztest_split_pool; |
3541dc6d | 379 | ztest_func_t ztest_reguid; |
ea0b2538 | 380 | ztest_func_t ztest_spa_upgrade; |
a1d477c2 | 381 | ztest_func_t ztest_device_removal; |
d2734cce | 382 | ztest_func_t ztest_spa_checkpoint_create_discard; |
619f0976 | 383 | ztest_func_t ztest_initialize; |
1b939560 | 384 | ztest_func_t ztest_trim; |
1eeb4562 | 385 | ztest_func_t ztest_fletcher; |
37f520db | 386 | ztest_func_t ztest_fletcher_incr; |
50c957f7 | 387 | ztest_func_t ztest_verify_dnode_bt; |
34dc7c2f | 388 | |
428870ff BB |
389 | uint64_t zopt_always = 0ULL * NANOSEC; /* all the time */ |
390 | uint64_t zopt_incessant = 1ULL * NANOSEC / 10; /* every 1/10 second */ | |
391 | uint64_t zopt_often = 1ULL * NANOSEC; /* every second */ | |
392 | uint64_t zopt_sometimes = 10ULL * NANOSEC; /* every 10 seconds */ | |
393 | uint64_t zopt_rarely = 60ULL * NANOSEC; /* every 60 seconds */ | |
34dc7c2f | 394 | |
fdc5d982 TC |
395 | #define ZTI_INIT(func, iters, interval) \ |
396 | { .zi_func = (func), \ | |
397 | .zi_iters = (iters), \ | |
398 | .zi_interval = (interval), \ | |
399 | .zi_funcname = # func } | |
400 | ||
34dc7c2f | 401 | ztest_info_t ztest_info[] = { |
fdc5d982 TC |
402 | ZTI_INIT(ztest_dmu_read_write, 1, &zopt_always), |
403 | ZTI_INIT(ztest_dmu_write_parallel, 10, &zopt_always), | |
404 | ZTI_INIT(ztest_dmu_object_alloc_free, 1, &zopt_always), | |
047116ac | 405 | ZTI_INIT(ztest_dmu_object_next_chunk, 1, &zopt_sometimes), |
fdc5d982 TC |
406 | ZTI_INIT(ztest_dmu_commit_callbacks, 1, &zopt_always), |
407 | ZTI_INIT(ztest_zap, 30, &zopt_always), | |
408 | ZTI_INIT(ztest_zap_parallel, 100, &zopt_always), | |
409 | ZTI_INIT(ztest_split_pool, 1, &zopt_always), | |
410 | ZTI_INIT(ztest_zil_commit, 1, &zopt_incessant), | |
411 | ZTI_INIT(ztest_zil_remount, 1, &zopt_sometimes), | |
412 | ZTI_INIT(ztest_dmu_read_write_zcopy, 1, &zopt_often), | |
413 | ZTI_INIT(ztest_dmu_objset_create_destroy, 1, &zopt_often), | |
414 | ZTI_INIT(ztest_dsl_prop_get_set, 1, &zopt_often), | |
415 | ZTI_INIT(ztest_spa_prop_get_set, 1, &zopt_sometimes), | |
428870ff | 416 | #if 0 |
fdc5d982 | 417 | ZTI_INIT(ztest_dmu_prealloc, 1, &zopt_sometimes), |
428870ff | 418 | #endif |
fdc5d982 TC |
419 | ZTI_INIT(ztest_fzap, 1, &zopt_sometimes), |
420 | ZTI_INIT(ztest_dmu_snapshot_create_destroy, 1, &zopt_sometimes), | |
421 | ZTI_INIT(ztest_spa_create_destroy, 1, &zopt_sometimes), | |
422 | ZTI_INIT(ztest_fault_inject, 1, &zopt_sometimes), | |
fdc5d982 | 423 | ZTI_INIT(ztest_dmu_snapshot_hold, 1, &zopt_sometimes), |
0582e403 | 424 | ZTI_INIT(ztest_mmp_enable_disable, 1, &zopt_sometimes), |
fdc5d982 | 425 | ZTI_INIT(ztest_reguid, 1, &zopt_rarely), |
fdc5d982 TC |
426 | ZTI_INIT(ztest_scrub, 1, &zopt_rarely), |
427 | ZTI_INIT(ztest_spa_upgrade, 1, &zopt_rarely), | |
428 | ZTI_INIT(ztest_dsl_dataset_promote_busy, 1, &zopt_rarely), | |
429 | ZTI_INIT(ztest_vdev_attach_detach, 1, &zopt_sometimes), | |
430 | ZTI_INIT(ztest_vdev_LUN_growth, 1, &zopt_rarely), | |
431 | ZTI_INIT(ztest_vdev_add_remove, 1, &ztest_opts.zo_vdevtime), | |
cc99f275 | 432 | ZTI_INIT(ztest_vdev_class_add, 1, &ztest_opts.zo_vdevtime), |
fdc5d982 | 433 | ZTI_INIT(ztest_vdev_aux_add_remove, 1, &ztest_opts.zo_vdevtime), |
a1d477c2 | 434 | ZTI_INIT(ztest_device_removal, 1, &zopt_sometimes), |
d2734cce | 435 | ZTI_INIT(ztest_spa_checkpoint_create_discard, 1, &zopt_rarely), |
619f0976 | 436 | ZTI_INIT(ztest_initialize, 1, &zopt_sometimes), |
1b939560 | 437 | ZTI_INIT(ztest_trim, 1, &zopt_sometimes), |
1eeb4562 | 438 | ZTI_INIT(ztest_fletcher, 1, &zopt_rarely), |
37f520db | 439 | ZTI_INIT(ztest_fletcher_incr, 1, &zopt_rarely), |
50c957f7 | 440 | ZTI_INIT(ztest_verify_dnode_bt, 1, &zopt_sometimes), |
34dc7c2f BB |
441 | }; |
442 | ||
443 | #define ZTEST_FUNCS (sizeof (ztest_info) / sizeof (ztest_info_t)) | |
444 | ||
428870ff BB |
445 | /* |
446 | * The following struct is used to hold a list of uncalled commit callbacks. | |
447 | * The callbacks are ordered by txg number. | |
448 | */ | |
449 | typedef struct ztest_cb_list { | |
1e33ac1e BB |
450 | kmutex_t zcl_callbacks_lock; |
451 | list_t zcl_callbacks; | |
428870ff | 452 | } ztest_cb_list_t; |
34dc7c2f BB |
453 | |
454 | /* | |
455 | * Stuff we need to share writably between parent and child. | |
456 | */ | |
457 | typedef struct ztest_shared { | |
c242c188 | 458 | boolean_t zs_do_init; |
428870ff BB |
459 | hrtime_t zs_proc_start; |
460 | hrtime_t zs_proc_stop; | |
461 | hrtime_t zs_thread_start; | |
462 | hrtime_t zs_thread_stop; | |
463 | hrtime_t zs_thread_kill; | |
34dc7c2f | 464 | uint64_t zs_enospc_count; |
428870ff BB |
465 | uint64_t zs_vdev_next_leaf; |
466 | uint64_t zs_vdev_aux; | |
34dc7c2f BB |
467 | uint64_t zs_alloc; |
468 | uint64_t zs_space; | |
428870ff BB |
469 | uint64_t zs_splits; |
470 | uint64_t zs_mirrors; | |
c242c188 CS |
471 | uint64_t zs_metaslab_sz; |
472 | uint64_t zs_metaslab_df_alloc_threshold; | |
473 | uint64_t zs_guid; | |
34dc7c2f BB |
474 | } ztest_shared_t; |
475 | ||
428870ff BB |
476 | #define ID_PARALLEL -1ULL |
477 | ||
34dc7c2f | 478 | static char ztest_dev_template[] = "%s/%s.%llua"; |
b128c09f | 479 | static char ztest_aux_template[] = "%s/%s.%s.%llu"; |
428870ff | 480 | ztest_shared_t *ztest_shared; |
34dc7c2f | 481 | |
c242c188 CS |
482 | static spa_t *ztest_spa = NULL; |
483 | static ztest_ds_t *ztest_ds; | |
484 | ||
485 | static kmutex_t ztest_vdev_lock; | |
9e052db4 | 486 | static boolean_t ztest_device_removal_active = B_FALSE; |
ce5fb2a7 | 487 | static boolean_t ztest_pool_scrubbed = B_FALSE; |
d2734cce | 488 | static kmutex_t ztest_checkpoint_lock; |
3bc7e0fb GW |
489 | |
490 | /* | |
491 | * The ztest_name_lock protects the pool and dataset namespace used by | |
492 | * the individual tests. To modify the namespace, consumers must grab | |
493 | * this lock as writer. Grabbing the lock as reader will ensure that the | |
494 | * namespace does not change while the lock is held. | |
495 | */ | |
2d9142c9 | 496 | static pthread_rwlock_t ztest_name_lock; |
34dc7c2f | 497 | |
c242c188 | 498 | static boolean_t ztest_dump_core = B_TRUE; |
b128c09f | 499 | static boolean_t ztest_exiting; |
34dc7c2f | 500 | |
428870ff BB |
501 | /* Global commit callback list */ |
502 | static ztest_cb_list_t zcl; | |
090ff092 RC |
503 | /* Commit cb delay */ |
504 | static uint64_t zc_min_txg_delay = UINT64_MAX; | |
505 | static int zc_cb_counter = 0; | |
506 | ||
507 | /* | |
508 | * Minimum number of commit callbacks that need to be registered for us to check | |
509 | * whether the minimum txg delay is acceptable. | |
510 | */ | |
511 | #define ZTEST_COMMIT_CB_MIN_REG 100 | |
512 | ||
513 | /* | |
514 | * If a number of txgs equal to this threshold have been created after a commit | |
515 | * callback has been registered but not called, then we assume there is an | |
516 | * implementation bug. | |
517 | */ | |
518 | #define ZTEST_COMMIT_CB_THRESH (TXG_CONCURRENT_STATES + 1000) | |
428870ff | 519 | |
428870ff BB |
520 | enum ztest_object { |
521 | ZTEST_META_DNODE = 0, | |
522 | ZTEST_DIROBJ, | |
523 | ZTEST_OBJECTS | |
524 | }; | |
34dc7c2f BB |
525 | |
526 | static void usage(boolean_t) __NORETURN; | |
ce5fb2a7 | 527 | static int ztest_scrub_impl(spa_t *spa); |
34dc7c2f BB |
528 | |
529 | /* | |
530 | * These libumem hooks provide a reasonable set of defaults for the allocator's | |
531 | * debugging facilities. | |
532 | */ | |
533 | const char * | |
0bc8fd78 | 534 | _umem_debug_init(void) |
34dc7c2f BB |
535 | { |
536 | return ("default,verbose"); /* $UMEM_DEBUG setting */ | |
537 | } | |
538 | ||
539 | const char * | |
540 | _umem_logging_init(void) | |
541 | { | |
542 | return ("fail,contents"); /* $UMEM_LOGGING setting */ | |
543 | } | |
544 | ||
8fb1ede1 BB |
545 | static void |
546 | dump_debug_buffer(void) | |
547 | { | |
ab4c009e TC |
548 | ssize_t ret __attribute__((unused)); |
549 | ||
550 | if (!ztest_opts.zo_dump_dbgmsg) | |
8fb1ede1 BB |
551 | return; |
552 | ||
ab4c009e TC |
553 | /* |
554 | * We use write() instead of printf() so that this function | |
555 | * is safe to call from a signal handler. | |
556 | */ | |
557 | ret = write(STDOUT_FILENO, "\n", 1); | |
8fb1ede1 BB |
558 | zfs_dbgmsg_print("ztest"); |
559 | } | |
560 | ||
e82cdc3a NB |
561 | #define BACKTRACE_SZ 100 |
562 | ||
563 | static void sig_handler(int signo) | |
564 | { | |
565 | struct sigaction action; | |
967798d0 | 566 | #ifdef __GLIBC__ /* backtrace() is a GNU extension */ |
e82cdc3a NB |
567 | int nptrs; |
568 | void *buffer[BACKTRACE_SZ]; | |
569 | ||
570 | nptrs = backtrace(buffer, BACKTRACE_SZ); | |
571 | backtrace_symbols_fd(buffer, nptrs, STDERR_FILENO); | |
572 | #endif | |
8fb1ede1 | 573 | dump_debug_buffer(); |
e82cdc3a NB |
574 | |
575 | /* | |
576 | * Restore default action and re-raise signal so SIGSEGV and | |
577 | * SIGABRT can trigger a core dump. | |
578 | */ | |
579 | action.sa_handler = SIG_DFL; | |
580 | sigemptyset(&action.sa_mask); | |
581 | action.sa_flags = 0; | |
582 | (void) sigaction(signo, &action, NULL); | |
583 | raise(signo); | |
584 | } | |
585 | ||
34dc7c2f BB |
586 | #define FATAL_MSG_SZ 1024 |
587 | ||
588 | char *fatal_msg; | |
589 | ||
590 | static void | |
591 | fatal(int do_perror, char *message, ...) | |
592 | { | |
593 | va_list args; | |
594 | int save_errno = errno; | |
40b84e7a | 595 | char *buf; |
34dc7c2f BB |
596 | |
597 | (void) fflush(stdout); | |
40b84e7a | 598 | buf = umem_alloc(FATAL_MSG_SZ, UMEM_NOFAIL); |
34dc7c2f BB |
599 | |
600 | va_start(args, message); | |
601 | (void) sprintf(buf, "ztest: "); | |
602 | /* LINTED */ | |
603 | (void) vsprintf(buf + strlen(buf), message, args); | |
604 | va_end(args); | |
605 | if (do_perror) { | |
606 | (void) snprintf(buf + strlen(buf), FATAL_MSG_SZ - strlen(buf), | |
607 | ": %s", strerror(save_errno)); | |
608 | } | |
609 | (void) fprintf(stderr, "%s\n", buf); | |
610 | fatal_msg = buf; /* to ease debugging */ | |
8fb1ede1 | 611 | |
34dc7c2f BB |
612 | if (ztest_dump_core) |
613 | abort(); | |
ab4c009e TC |
614 | else |
615 | dump_debug_buffer(); | |
616 | ||
34dc7c2f BB |
617 | exit(3); |
618 | } | |
619 | ||
620 | static int | |
621 | str2shift(const char *buf) | |
622 | { | |
623 | const char *ends = "BKMGTPEZ"; | |
624 | int i; | |
625 | ||
626 | if (buf[0] == '\0') | |
627 | return (0); | |
628 | for (i = 0; i < strlen(ends); i++) { | |
629 | if (toupper(buf[0]) == ends[i]) | |
630 | break; | |
631 | } | |
632 | if (i == strlen(ends)) { | |
633 | (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", | |
634 | buf); | |
635 | usage(B_FALSE); | |
636 | } | |
637 | if (buf[1] == '\0' || (toupper(buf[1]) == 'B' && buf[2] == '\0')) { | |
638 | return (10*i); | |
639 | } | |
640 | (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", buf); | |
641 | usage(B_FALSE); | |
642 | /* NOTREACHED */ | |
643 | } | |
644 | ||
645 | static uint64_t | |
646 | nicenumtoull(const char *buf) | |
647 | { | |
648 | char *end; | |
649 | uint64_t val; | |
650 | ||
651 | val = strtoull(buf, &end, 0); | |
652 | if (end == buf) { | |
653 | (void) fprintf(stderr, "ztest: bad numeric value: %s\n", buf); | |
654 | usage(B_FALSE); | |
655 | } else if (end[0] == '.') { | |
656 | double fval = strtod(buf, &end); | |
657 | fval *= pow(2, str2shift(end)); | |
4d32abaa RM |
658 | /* |
659 | * UINT64_MAX is not exactly representable as a double. | |
660 | * The closest representation is UINT64_MAX + 1, so we | |
661 | * use a >= comparison instead of > for the bounds check. | |
662 | */ | |
663 | if (fval >= (double)UINT64_MAX) { | |
34dc7c2f BB |
664 | (void) fprintf(stderr, "ztest: value too large: %s\n", |
665 | buf); | |
666 | usage(B_FALSE); | |
667 | } | |
668 | val = (uint64_t)fval; | |
669 | } else { | |
670 | int shift = str2shift(end); | |
671 | if (shift >= 64 || (val << shift) >> shift != val) { | |
672 | (void) fprintf(stderr, "ztest: value too large: %s\n", | |
673 | buf); | |
674 | usage(B_FALSE); | |
675 | } | |
676 | val <<= shift; | |
677 | } | |
678 | return (val); | |
679 | } | |
680 | ||
681 | static void | |
682 | usage(boolean_t requested) | |
683 | { | |
c242c188 CS |
684 | const ztest_shared_opts_t *zo = &ztest_opts_defaults; |
685 | ||
f3c8c9e6 | 686 | char nice_vdev_size[NN_NUMBUF_SZ]; |
d830d479 | 687 | char nice_force_ganging[NN_NUMBUF_SZ]; |
34dc7c2f BB |
688 | FILE *fp = requested ? stdout : stderr; |
689 | ||
f3c8c9e6 | 690 | nicenum(zo->zo_vdev_size, nice_vdev_size, sizeof (nice_vdev_size)); |
d830d479 MA |
691 | nicenum(zo->zo_metaslab_force_ganging, nice_force_ganging, |
692 | sizeof (nice_force_ganging)); | |
34dc7c2f BB |
693 | |
694 | (void) fprintf(fp, "Usage: %s\n" | |
695 | "\t[-v vdevs (default: %llu)]\n" | |
696 | "\t[-s size_of_each_vdev (default: %s)]\n" | |
428870ff | 697 | "\t[-a alignment_shift (default: %d)] use 0 for random\n" |
34dc7c2f | 698 | "\t[-m mirror_copies (default: %d)]\n" |
b2255edc BB |
699 | "\t[-r raidz_disks / draid_disks (default: %d)]\n" |
700 | "\t[-R raid_parity (default: %d)]\n" | |
701 | "\t[-K raid_kind (default: random)] raidz|draid|random\n" | |
702 | "\t[-D draid_data (default: %d)] in config\n" | |
703 | "\t[-S draid_spares (default: %d)]\n" | |
34dc7c2f BB |
704 | "\t[-d datasets (default: %d)]\n" |
705 | "\t[-t threads (default: %d)]\n" | |
706 | "\t[-g gang_block_threshold (default: %s)]\n" | |
428870ff BB |
707 | "\t[-i init_count (default: %d)] initialize pool i times\n" |
708 | "\t[-k kill_percentage (default: %llu%%)]\n" | |
34dc7c2f | 709 | "\t[-p pool_name (default: %s)]\n" |
428870ff | 710 | "\t[-f dir (default: %s)] file directory for vdev files\n" |
379ca9cf | 711 | "\t[-M] Multi-host simulate pool imported on remote host\n" |
428870ff BB |
712 | "\t[-V] verbose (use multiple times for ever more blather)\n" |
713 | "\t[-E] use existing pool instead of creating new one\n" | |
714 | "\t[-T time (default: %llu sec)] total run time\n" | |
715 | "\t[-F freezeloops (default: %llu)] max loops in spa_freeze()\n" | |
716 | "\t[-P passtime (default: %llu sec)] time per pass\n" | |
c242c188 | 717 | "\t[-B alt_ztest (default: <none>)] alternate ztest path\n" |
cc99f275 | 718 | "\t[-C vdev class state (default: random)] special=on|off|random\n" |
ed828c0c GM |
719 | "\t[-o variable=value] ... set global variable to an unsigned\n" |
720 | "\t 32-bit integer value\n" | |
8fb1ede1 | 721 | "\t[-G dump zfs_dbgmsg buffer before exiting due to an error\n" |
34dc7c2f BB |
722 | "\t[-h] (print help)\n" |
723 | "", | |
c242c188 CS |
724 | zo->zo_pool, |
725 | (u_longlong_t)zo->zo_vdevs, /* -v */ | |
34dc7c2f | 726 | nice_vdev_size, /* -s */ |
c242c188 CS |
727 | zo->zo_ashift, /* -a */ |
728 | zo->zo_mirrors, /* -m */ | |
b2255edc BB |
729 | zo->zo_raid_children, /* -r */ |
730 | zo->zo_raid_parity, /* -R */ | |
731 | zo->zo_draid_data, /* -D */ | |
732 | zo->zo_draid_spares, /* -S */ | |
c242c188 CS |
733 | zo->zo_datasets, /* -d */ |
734 | zo->zo_threads, /* -t */ | |
d830d479 | 735 | nice_force_ganging, /* -g */ |
c242c188 CS |
736 | zo->zo_init, /* -i */ |
737 | (u_longlong_t)zo->zo_killrate, /* -k */ | |
738 | zo->zo_pool, /* -p */ | |
739 | zo->zo_dir, /* -f */ | |
740 | (u_longlong_t)zo->zo_time, /* -T */ | |
741 | (u_longlong_t)zo->zo_maxloops, /* -F */ | |
742 | (u_longlong_t)zo->zo_passtime); | |
34dc7c2f BB |
743 | exit(requested ? 0 : 1); |
744 | } | |
745 | ||
b2255edc BB |
746 | static uint64_t |
747 | ztest_random(uint64_t range) | |
748 | { | |
749 | uint64_t r; | |
750 | ||
751 | ASSERT3S(ztest_fd_rand, >=, 0); | |
752 | ||
753 | if (range == 0) | |
754 | return (0); | |
755 | ||
756 | if (read(ztest_fd_rand, &r, sizeof (r)) != sizeof (r)) | |
757 | fatal(1, "short read from /dev/urandom"); | |
758 | ||
759 | return (r % range); | |
760 | } | |
cc99f275 DB |
761 | |
762 | static void | |
763 | ztest_parse_name_value(const char *input, ztest_shared_opts_t *zo) | |
764 | { | |
765 | char name[32]; | |
766 | char *value; | |
767 | int state = ZTEST_VDEV_CLASS_RND; | |
768 | ||
769 | (void) strlcpy(name, input, sizeof (name)); | |
770 | ||
771 | value = strchr(name, '='); | |
772 | if (value == NULL) { | |
773 | (void) fprintf(stderr, "missing value in property=value " | |
774 | "'-C' argument (%s)\n", input); | |
775 | usage(B_FALSE); | |
776 | } | |
777 | *(value) = '\0'; | |
778 | value++; | |
779 | ||
780 | if (strcmp(value, "on") == 0) { | |
781 | state = ZTEST_VDEV_CLASS_ON; | |
782 | } else if (strcmp(value, "off") == 0) { | |
783 | state = ZTEST_VDEV_CLASS_OFF; | |
784 | } else if (strcmp(value, "random") == 0) { | |
785 | state = ZTEST_VDEV_CLASS_RND; | |
786 | } else { | |
787 | (void) fprintf(stderr, "invalid property value '%s'\n", value); | |
788 | usage(B_FALSE); | |
789 | } | |
790 | ||
791 | if (strcmp(name, "special") == 0) { | |
792 | zo->zo_special_vdevs = state; | |
793 | } else { | |
794 | (void) fprintf(stderr, "invalid property name '%s'\n", name); | |
795 | usage(B_FALSE); | |
796 | } | |
797 | if (zo->zo_verbose >= 3) | |
798 | (void) printf("%s vdev state is '%s'\n", name, value); | |
799 | } | |
800 | ||
34dc7c2f BB |
801 | static void |
802 | process_options(int argc, char **argv) | |
803 | { | |
c242c188 CS |
804 | char *path; |
805 | ztest_shared_opts_t *zo = &ztest_opts; | |
806 | ||
34dc7c2f BB |
807 | int opt; |
808 | uint64_t value; | |
c242c188 | 809 | char altdir[MAXNAMELEN] = { 0 }; |
b2255edc | 810 | char raid_kind[8] = { "random" }; |
34dc7c2f | 811 | |
c242c188 | 812 | bcopy(&ztest_opts_defaults, zo, sizeof (*zo)); |
34dc7c2f | 813 | |
34dc7c2f | 814 | while ((opt = getopt(argc, argv, |
b2255edc | 815 | "v:s:a:m:r:R:K:D:S:d:t:g:i:k:p:f:MVET:P:hF:B:C:o:G")) != EOF) { |
34dc7c2f BB |
816 | value = 0; |
817 | switch (opt) { | |
818 | case 'v': | |
819 | case 's': | |
820 | case 'a': | |
821 | case 'm': | |
822 | case 'r': | |
823 | case 'R': | |
b2255edc BB |
824 | case 'D': |
825 | case 'S': | |
34dc7c2f BB |
826 | case 'd': |
827 | case 't': | |
828 | case 'g': | |
829 | case 'i': | |
830 | case 'k': | |
831 | case 'T': | |
832 | case 'P': | |
428870ff | 833 | case 'F': |
34dc7c2f BB |
834 | value = nicenumtoull(optarg); |
835 | } | |
836 | switch (opt) { | |
837 | case 'v': | |
c242c188 | 838 | zo->zo_vdevs = value; |
34dc7c2f BB |
839 | break; |
840 | case 's': | |
c242c188 | 841 | zo->zo_vdev_size = MAX(SPA_MINDEVSIZE, value); |
34dc7c2f BB |
842 | break; |
843 | case 'a': | |
c242c188 | 844 | zo->zo_ashift = value; |
34dc7c2f BB |
845 | break; |
846 | case 'm': | |
c242c188 | 847 | zo->zo_mirrors = value; |
34dc7c2f BB |
848 | break; |
849 | case 'r': | |
b2255edc | 850 | zo->zo_raid_children = MAX(1, value); |
34dc7c2f BB |
851 | break; |
852 | case 'R': | |
b2255edc BB |
853 | zo->zo_raid_parity = MIN(MAX(value, 1), 3); |
854 | break; | |
855 | case 'K': | |
856 | (void) strlcpy(raid_kind, optarg, sizeof (raid_kind)); | |
857 | break; | |
858 | case 'D': | |
859 | zo->zo_draid_data = MAX(1, value); | |
860 | break; | |
861 | case 'S': | |
862 | zo->zo_draid_spares = MAX(1, value); | |
34dc7c2f BB |
863 | break; |
864 | case 'd': | |
c242c188 | 865 | zo->zo_datasets = MAX(1, value); |
34dc7c2f BB |
866 | break; |
867 | case 't': | |
c242c188 | 868 | zo->zo_threads = MAX(1, value); |
34dc7c2f BB |
869 | break; |
870 | case 'g': | |
d830d479 MA |
871 | zo->zo_metaslab_force_ganging = |
872 | MAX(SPA_MINBLOCKSIZE << 1, value); | |
34dc7c2f BB |
873 | break; |
874 | case 'i': | |
c242c188 | 875 | zo->zo_init = value; |
34dc7c2f BB |
876 | break; |
877 | case 'k': | |
c242c188 | 878 | zo->zo_killrate = value; |
34dc7c2f BB |
879 | break; |
880 | case 'p': | |
c242c188 CS |
881 | (void) strlcpy(zo->zo_pool, optarg, |
882 | sizeof (zo->zo_pool)); | |
34dc7c2f BB |
883 | break; |
884 | case 'f': | |
c242c188 CS |
885 | path = realpath(optarg, NULL); |
886 | if (path == NULL) { | |
887 | (void) fprintf(stderr, "error: %s: %s\n", | |
888 | optarg, strerror(errno)); | |
889 | usage(B_FALSE); | |
890 | } else { | |
891 | (void) strlcpy(zo->zo_dir, path, | |
892 | sizeof (zo->zo_dir)); | |
df053d67 | 893 | free(path); |
c242c188 | 894 | } |
34dc7c2f | 895 | break; |
379ca9cf OF |
896 | case 'M': |
897 | zo->zo_mmp_test = 1; | |
898 | break; | |
34dc7c2f | 899 | case 'V': |
c242c188 | 900 | zo->zo_verbose++; |
34dc7c2f BB |
901 | break; |
902 | case 'E': | |
c242c188 | 903 | zo->zo_init = 0; |
34dc7c2f BB |
904 | break; |
905 | case 'T': | |
c242c188 | 906 | zo->zo_time = value; |
34dc7c2f BB |
907 | break; |
908 | case 'P': | |
c242c188 | 909 | zo->zo_passtime = MAX(1, value); |
34dc7c2f | 910 | break; |
428870ff | 911 | case 'F': |
c242c188 CS |
912 | zo->zo_maxloops = MAX(1, value); |
913 | break; | |
914 | case 'B': | |
915 | (void) strlcpy(altdir, optarg, sizeof (altdir)); | |
428870ff | 916 | break; |
cc99f275 DB |
917 | case 'C': |
918 | ztest_parse_name_value(optarg, zo); | |
919 | break; | |
ed828c0c GM |
920 | case 'o': |
921 | if (set_global_var(optarg) != 0) | |
922 | usage(B_FALSE); | |
923 | break; | |
8fb1ede1 | 924 | case 'G': |
ab4c009e | 925 | zo->zo_dump_dbgmsg = 1; |
8fb1ede1 | 926 | break; |
34dc7c2f BB |
927 | case 'h': |
928 | usage(B_TRUE); | |
929 | break; | |
930 | case '?': | |
931 | default: | |
932 | usage(B_FALSE); | |
933 | break; | |
934 | } | |
935 | } | |
936 | ||
b2255edc BB |
937 | /* When raid choice is 'random' add a draid pool 50% of the time */ |
938 | if (strcmp(raid_kind, "random") == 0) { | |
939 | (void) strlcpy(raid_kind, (ztest_random(2) == 0) ? | |
940 | "draid" : "raidz", sizeof (raid_kind)); | |
941 | ||
942 | if (ztest_opts.zo_verbose >= 3) | |
943 | (void) printf("choosing RAID type '%s'\n", raid_kind); | |
944 | } | |
945 | ||
946 | if (strcmp(raid_kind, "draid") == 0) { | |
947 | uint64_t min_devsize; | |
948 | ||
949 | /* With fewer disk use 256M, otherwise 128M is OK */ | |
950 | min_devsize = (ztest_opts.zo_raid_children < 16) ? | |
951 | (256ULL << 20) : (128ULL << 20); | |
952 | ||
953 | /* No top-level mirrors with dRAID for now */ | |
954 | zo->zo_mirrors = 0; | |
955 | ||
956 | /* Use more appropriate defaults for dRAID */ | |
957 | if (zo->zo_vdevs == ztest_opts_defaults.zo_vdevs) | |
958 | zo->zo_vdevs = 1; | |
959 | if (zo->zo_raid_children == | |
960 | ztest_opts_defaults.zo_raid_children) | |
961 | zo->zo_raid_children = 16; | |
962 | if (zo->zo_ashift < 12) | |
963 | zo->zo_ashift = 12; | |
964 | if (zo->zo_vdev_size < min_devsize) | |
965 | zo->zo_vdev_size = min_devsize; | |
966 | ||
967 | if (zo->zo_draid_data + zo->zo_raid_parity > | |
968 | zo->zo_raid_children - zo->zo_draid_spares) { | |
969 | (void) fprintf(stderr, "error: too few draid " | |
970 | "children (%d) for stripe width (%d)\n", | |
971 | zo->zo_raid_children, | |
972 | zo->zo_draid_data + zo->zo_raid_parity); | |
973 | usage(B_FALSE); | |
974 | } | |
975 | ||
976 | (void) strlcpy(zo->zo_raid_type, VDEV_TYPE_DRAID, | |
977 | sizeof (zo->zo_raid_type)); | |
978 | ||
979 | } else /* using raidz */ { | |
980 | ASSERT0(strcmp(raid_kind, "raidz")); | |
981 | ||
982 | zo->zo_raid_parity = MIN(zo->zo_raid_parity, | |
983 | zo->zo_raid_children - 1); | |
984 | } | |
34dc7c2f | 985 | |
c242c188 CS |
986 | zo->zo_vdevtime = |
987 | (zo->zo_vdevs > 0 ? zo->zo_time * NANOSEC / zo->zo_vdevs : | |
428870ff | 988 | UINT64_MAX >> 2); |
c242c188 CS |
989 | |
990 | if (strlen(altdir) > 0) { | |
ae380cfa BB |
991 | char *cmd; |
992 | char *realaltdir; | |
c242c188 CS |
993 | char *bin; |
994 | char *ztest; | |
995 | char *isa; | |
996 | int isalen; | |
997 | ||
ae380cfa BB |
998 | cmd = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
999 | realaltdir = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
1000 | ||
5be98cfe | 1001 | VERIFY(NULL != realpath(getexecname(), cmd)); |
c242c188 CS |
1002 | if (0 != access(altdir, F_OK)) { |
1003 | ztest_dump_core = B_FALSE; | |
1004 | fatal(B_TRUE, "invalid alternate ztest path: %s", | |
1005 | altdir); | |
1006 | } | |
1007 | VERIFY(NULL != realpath(altdir, realaltdir)); | |
1008 | ||
1009 | /* | |
1010 | * 'cmd' should be of the form "<anything>/usr/bin/<isa>/ztest". | |
1011 | * We want to extract <isa> to determine if we should use | |
1012 | * 32 or 64 bit binaries. | |
1013 | */ | |
1014 | bin = strstr(cmd, "/usr/bin/"); | |
1015 | ztest = strstr(bin, "/ztest"); | |
1016 | isa = bin + 9; | |
1017 | isalen = ztest - isa; | |
1018 | (void) snprintf(zo->zo_alt_ztest, sizeof (zo->zo_alt_ztest), | |
1019 | "%s/usr/bin/%.*s/ztest", realaltdir, isalen, isa); | |
1020 | (void) snprintf(zo->zo_alt_libpath, sizeof (zo->zo_alt_libpath), | |
1021 | "%s/usr/lib/%.*s", realaltdir, isalen, isa); | |
1022 | ||
1023 | if (0 != access(zo->zo_alt_ztest, X_OK)) { | |
1024 | ztest_dump_core = B_FALSE; | |
1025 | fatal(B_TRUE, "invalid alternate ztest: %s", | |
1026 | zo->zo_alt_ztest); | |
1027 | } else if (0 != access(zo->zo_alt_libpath, X_OK)) { | |
1028 | ztest_dump_core = B_FALSE; | |
1029 | fatal(B_TRUE, "invalid alternate lib directory %s", | |
1030 | zo->zo_alt_libpath); | |
1031 | } | |
ae380cfa BB |
1032 | |
1033 | umem_free(cmd, MAXPATHLEN); | |
1034 | umem_free(realaltdir, MAXPATHLEN); | |
c242c188 | 1035 | } |
428870ff BB |
1036 | } |
1037 | ||
1038 | static void | |
1039 | ztest_kill(ztest_shared_t *zs) | |
1040 | { | |
c242c188 CS |
1041 | zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(ztest_spa)); |
1042 | zs->zs_space = metaslab_class_get_space(spa_normal_class(ztest_spa)); | |
5d1f7fb6 GW |
1043 | |
1044 | /* | |
1045 | * Before we kill off ztest, make sure that the config is updated. | |
a1d477c2 | 1046 | * See comment above spa_write_cachefile(). |
5d1f7fb6 GW |
1047 | */ |
1048 | mutex_enter(&spa_namespace_lock); | |
a1d477c2 | 1049 | spa_write_cachefile(ztest_spa, B_FALSE, B_FALSE); |
5d1f7fb6 GW |
1050 | mutex_exit(&spa_namespace_lock); |
1051 | ||
428870ff BB |
1052 | (void) kill(getpid(), SIGKILL); |
1053 | } | |
1054 | ||
428870ff BB |
1055 | /* ARGSUSED */ |
1056 | static void | |
1057 | ztest_record_enospc(const char *s) | |
1058 | { | |
1059 | ztest_shared->zs_enospc_count++; | |
34dc7c2f BB |
1060 | } |
1061 | ||
1062 | static uint64_t | |
1063 | ztest_get_ashift(void) | |
1064 | { | |
c242c188 | 1065 | if (ztest_opts.zo_ashift == 0) |
b02fe35d | 1066 | return (SPA_MINBLOCKSHIFT + ztest_random(5)); |
c242c188 | 1067 | return (ztest_opts.zo_ashift); |
34dc7c2f BB |
1068 | } |
1069 | ||
b2255edc BB |
1070 | static boolean_t |
1071 | ztest_is_draid_spare(const char *name) | |
1072 | { | |
1073 | uint64_t spare_id = 0, parity = 0, vdev_id = 0; | |
1074 | ||
1075 | if (sscanf(name, VDEV_TYPE_DRAID "%llu-%llu-%llu", | |
1076 | (u_longlong_t *)&parity, (u_longlong_t *)&vdev_id, | |
1077 | (u_longlong_t *)&spare_id) == 3) { | |
1078 | return (B_TRUE); | |
1079 | } | |
1080 | ||
1081 | return (B_FALSE); | |
1082 | } | |
1083 | ||
34dc7c2f | 1084 | static nvlist_t * |
ea0b2538 | 1085 | make_vdev_file(char *path, char *aux, char *pool, size_t size, uint64_t ashift) |
34dc7c2f | 1086 | { |
40b84e7a | 1087 | char *pathbuf; |
34dc7c2f | 1088 | uint64_t vdev; |
34dc7c2f | 1089 | nvlist_t *file; |
b2255edc | 1090 | boolean_t draid_spare = B_FALSE; |
34dc7c2f | 1091 | |
40b84e7a BB |
1092 | pathbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
1093 | ||
b128c09f BB |
1094 | if (ashift == 0) |
1095 | ashift = ztest_get_ashift(); | |
1096 | ||
1097 | if (path == NULL) { | |
1098 | path = pathbuf; | |
1099 | ||
1100 | if (aux != NULL) { | |
1101 | vdev = ztest_shared->zs_vdev_aux; | |
c242c188 CS |
1102 | (void) snprintf(path, MAXPATHLEN, |
1103 | ztest_aux_template, ztest_opts.zo_dir, | |
ea0b2538 GW |
1104 | pool == NULL ? ztest_opts.zo_pool : pool, |
1105 | aux, vdev); | |
b128c09f | 1106 | } else { |
428870ff | 1107 | vdev = ztest_shared->zs_vdev_next_leaf++; |
c242c188 CS |
1108 | (void) snprintf(path, MAXPATHLEN, |
1109 | ztest_dev_template, ztest_opts.zo_dir, | |
ea0b2538 | 1110 | pool == NULL ? ztest_opts.zo_pool : pool, vdev); |
b128c09f | 1111 | } |
b2255edc BB |
1112 | } else { |
1113 | draid_spare = ztest_is_draid_spare(path); | |
b128c09f | 1114 | } |
34dc7c2f | 1115 | |
b2255edc | 1116 | if (size != 0 && !draid_spare) { |
b128c09f | 1117 | int fd = open(path, O_RDWR | O_CREAT | O_TRUNC, 0666); |
34dc7c2f | 1118 | if (fd == -1) |
b128c09f | 1119 | fatal(1, "can't open %s", path); |
34dc7c2f | 1120 | if (ftruncate(fd, size) != 0) |
b128c09f | 1121 | fatal(1, "can't ftruncate %s", path); |
34dc7c2f BB |
1122 | (void) close(fd); |
1123 | } | |
1124 | ||
b2255edc BB |
1125 | VERIFY0(nvlist_alloc(&file, NV_UNIQUE_NAME, 0)); |
1126 | VERIFY0(nvlist_add_string(file, ZPOOL_CONFIG_TYPE, | |
1127 | draid_spare ? VDEV_TYPE_DRAID_SPARE : VDEV_TYPE_FILE)); | |
1128 | VERIFY0(nvlist_add_string(file, ZPOOL_CONFIG_PATH, path)); | |
1129 | VERIFY0(nvlist_add_uint64(file, ZPOOL_CONFIG_ASHIFT, ashift)); | |
40b84e7a | 1130 | umem_free(pathbuf, MAXPATHLEN); |
34dc7c2f BB |
1131 | |
1132 | return (file); | |
1133 | } | |
1134 | ||
1135 | static nvlist_t * | |
b2255edc | 1136 | make_vdev_raid(char *path, char *aux, char *pool, size_t size, |
ea0b2538 | 1137 | uint64_t ashift, int r) |
34dc7c2f | 1138 | { |
b2255edc | 1139 | nvlist_t *raid, **child; |
34dc7c2f BB |
1140 | int c; |
1141 | ||
1142 | if (r < 2) | |
ea0b2538 | 1143 | return (make_vdev_file(path, aux, pool, size, ashift)); |
34dc7c2f BB |
1144 | child = umem_alloc(r * sizeof (nvlist_t *), UMEM_NOFAIL); |
1145 | ||
1146 | for (c = 0; c < r; c++) | |
ea0b2538 | 1147 | child[c] = make_vdev_file(path, aux, pool, size, ashift); |
34dc7c2f | 1148 | |
b2255edc BB |
1149 | VERIFY0(nvlist_alloc(&raid, NV_UNIQUE_NAME, 0)); |
1150 | VERIFY0(nvlist_add_string(raid, ZPOOL_CONFIG_TYPE, | |
1151 | ztest_opts.zo_raid_type)); | |
1152 | VERIFY0(nvlist_add_uint64(raid, ZPOOL_CONFIG_NPARITY, | |
1153 | ztest_opts.zo_raid_parity)); | |
1154 | VERIFY0(nvlist_add_nvlist_array(raid, ZPOOL_CONFIG_CHILDREN, | |
1155 | child, r)); | |
1156 | ||
1157 | if (strcmp(ztest_opts.zo_raid_type, VDEV_TYPE_DRAID) == 0) { | |
1158 | uint64_t ndata = ztest_opts.zo_draid_data; | |
1159 | uint64_t nparity = ztest_opts.zo_raid_parity; | |
1160 | uint64_t nspares = ztest_opts.zo_draid_spares; | |
1161 | uint64_t children = ztest_opts.zo_raid_children; | |
1162 | uint64_t ngroups = 1; | |
1163 | ||
1164 | /* | |
1165 | * Calculate the minimum number of groups required to fill a | |
1166 | * slice. This is the LCM of the stripe width (data + parity) | |
1167 | * and the number of data drives (children - spares). | |
1168 | */ | |
1169 | while (ngroups * (ndata + nparity) % (children - nspares) != 0) | |
1170 | ngroups++; | |
1171 | ||
1172 | /* Store the basic dRAID configuration. */ | |
1173 | fnvlist_add_uint64(raid, ZPOOL_CONFIG_DRAID_NDATA, ndata); | |
1174 | fnvlist_add_uint64(raid, ZPOOL_CONFIG_DRAID_NSPARES, nspares); | |
1175 | fnvlist_add_uint64(raid, ZPOOL_CONFIG_DRAID_NGROUPS, ngroups); | |
1176 | } | |
34dc7c2f BB |
1177 | |
1178 | for (c = 0; c < r; c++) | |
1179 | nvlist_free(child[c]); | |
1180 | ||
1181 | umem_free(child, r * sizeof (nvlist_t *)); | |
1182 | ||
b2255edc | 1183 | return (raid); |
34dc7c2f BB |
1184 | } |
1185 | ||
1186 | static nvlist_t * | |
ea0b2538 GW |
1187 | make_vdev_mirror(char *path, char *aux, char *pool, size_t size, |
1188 | uint64_t ashift, int r, int m) | |
34dc7c2f BB |
1189 | { |
1190 | nvlist_t *mirror, **child; | |
1191 | int c; | |
1192 | ||
1193 | if (m < 1) | |
b2255edc | 1194 | return (make_vdev_raid(path, aux, pool, size, ashift, r)); |
34dc7c2f BB |
1195 | |
1196 | child = umem_alloc(m * sizeof (nvlist_t *), UMEM_NOFAIL); | |
1197 | ||
1198 | for (c = 0; c < m; c++) | |
b2255edc | 1199 | child[c] = make_vdev_raid(path, aux, pool, size, ashift, r); |
34dc7c2f BB |
1200 | |
1201 | VERIFY(nvlist_alloc(&mirror, NV_UNIQUE_NAME, 0) == 0); | |
1202 | VERIFY(nvlist_add_string(mirror, ZPOOL_CONFIG_TYPE, | |
1203 | VDEV_TYPE_MIRROR) == 0); | |
1204 | VERIFY(nvlist_add_nvlist_array(mirror, ZPOOL_CONFIG_CHILDREN, | |
1205 | child, m) == 0); | |
34dc7c2f BB |
1206 | |
1207 | for (c = 0; c < m; c++) | |
1208 | nvlist_free(child[c]); | |
1209 | ||
1210 | umem_free(child, m * sizeof (nvlist_t *)); | |
1211 | ||
1212 | return (mirror); | |
1213 | } | |
1214 | ||
1215 | static nvlist_t * | |
ea0b2538 | 1216 | make_vdev_root(char *path, char *aux, char *pool, size_t size, uint64_t ashift, |
cc99f275 | 1217 | const char *class, int r, int m, int t) |
34dc7c2f BB |
1218 | { |
1219 | nvlist_t *root, **child; | |
1220 | int c; | |
cc99f275 | 1221 | boolean_t log; |
34dc7c2f BB |
1222 | |
1223 | ASSERT(t > 0); | |
1224 | ||
cc99f275 DB |
1225 | log = (class != NULL && strcmp(class, "log") == 0); |
1226 | ||
34dc7c2f BB |
1227 | child = umem_alloc(t * sizeof (nvlist_t *), UMEM_NOFAIL); |
1228 | ||
b128c09f | 1229 | for (c = 0; c < t; c++) { |
ea0b2538 GW |
1230 | child[c] = make_vdev_mirror(path, aux, pool, size, ashift, |
1231 | r, m); | |
b128c09f BB |
1232 | VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_IS_LOG, |
1233 | log) == 0); | |
cc99f275 DB |
1234 | |
1235 | if (class != NULL && class[0] != '\0') { | |
1236 | ASSERT(m > 1 || log); /* expecting a mirror */ | |
1237 | VERIFY(nvlist_add_string(child[c], | |
1238 | ZPOOL_CONFIG_ALLOCATION_BIAS, class) == 0); | |
1239 | } | |
b128c09f | 1240 | } |
34dc7c2f BB |
1241 | |
1242 | VERIFY(nvlist_alloc(&root, NV_UNIQUE_NAME, 0) == 0); | |
1243 | VERIFY(nvlist_add_string(root, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT) == 0); | |
b128c09f | 1244 | VERIFY(nvlist_add_nvlist_array(root, aux ? aux : ZPOOL_CONFIG_CHILDREN, |
34dc7c2f BB |
1245 | child, t) == 0); |
1246 | ||
1247 | for (c = 0; c < t; c++) | |
1248 | nvlist_free(child[c]); | |
1249 | ||
1250 | umem_free(child, t * sizeof (nvlist_t *)); | |
1251 | ||
1252 | return (root); | |
1253 | } | |
1254 | ||
ea0b2538 GW |
1255 | /* |
1256 | * Find a random spa version. Returns back a random spa version in the | |
1257 | * range [initial_version, SPA_VERSION_FEATURES]. | |
1258 | */ | |
1259 | static uint64_t | |
1260 | ztest_random_spa_version(uint64_t initial_version) | |
1261 | { | |
1262 | uint64_t version = initial_version; | |
1263 | ||
1264 | if (version <= SPA_VERSION_BEFORE_FEATURES) { | |
1265 | version = version + | |
1266 | ztest_random(SPA_VERSION_BEFORE_FEATURES - version + 1); | |
1267 | } | |
1268 | ||
1269 | if (version > SPA_VERSION_BEFORE_FEATURES) | |
1270 | version = SPA_VERSION_FEATURES; | |
1271 | ||
1272 | ASSERT(SPA_VERSION_IS_SUPPORTED(version)); | |
1273 | return (version); | |
1274 | } | |
1275 | ||
428870ff BB |
1276 | static int |
1277 | ztest_random_blocksize(void) | |
34dc7c2f | 1278 | { |
cc99f275 DB |
1279 | ASSERT(ztest_spa->spa_max_ashift != 0); |
1280 | ||
f1512ee6 MA |
1281 | /* |
1282 | * Choose a block size >= the ashift. | |
1283 | * If the SPA supports new MAXBLOCKSIZE, test up to 1MB blocks. | |
1284 | */ | |
1285 | int maxbs = SPA_OLD_MAXBLOCKSHIFT; | |
1286 | if (spa_maxblocksize(ztest_spa) == SPA_MAXBLOCKSIZE) | |
1287 | maxbs = 20; | |
c3520e7f MA |
1288 | uint64_t block_shift = |
1289 | ztest_random(maxbs - ztest_spa->spa_max_ashift + 1); | |
b02fe35d | 1290 | return (1 << (SPA_MINBLOCKSHIFT + block_shift)); |
428870ff | 1291 | } |
34dc7c2f | 1292 | |
50c957f7 NB |
1293 | static int |
1294 | ztest_random_dnodesize(void) | |
1295 | { | |
1296 | int slots; | |
1297 | int max_slots = spa_maxdnodesize(ztest_spa) >> DNODE_SHIFT; | |
1298 | ||
1299 | if (max_slots == DNODE_MIN_SLOTS) | |
1300 | return (DNODE_MIN_SIZE); | |
1301 | ||
1302 | /* | |
1303 | * Weight the random distribution more heavily toward smaller | |
1304 | * dnode sizes since that is more likely to reflect real-world | |
1305 | * usage. | |
1306 | */ | |
1307 | ASSERT3U(max_slots, >, 4); | |
1308 | switch (ztest_random(10)) { | |
1309 | case 0: | |
1310 | slots = 5 + ztest_random(max_slots - 4); | |
1311 | break; | |
1312 | case 1 ... 4: | |
1313 | slots = 2 + ztest_random(3); | |
1314 | break; | |
1315 | default: | |
1316 | slots = 1; | |
1317 | break; | |
1318 | } | |
1319 | ||
1320 | return (slots << DNODE_SHIFT); | |
1321 | } | |
1322 | ||
428870ff BB |
1323 | static int |
1324 | ztest_random_ibshift(void) | |
1325 | { | |
1326 | return (DN_MIN_INDBLKSHIFT + | |
1327 | ztest_random(DN_MAX_INDBLKSHIFT - DN_MIN_INDBLKSHIFT + 1)); | |
34dc7c2f BB |
1328 | } |
1329 | ||
428870ff BB |
1330 | static uint64_t |
1331 | ztest_random_vdev_top(spa_t *spa, boolean_t log_ok) | |
34dc7c2f | 1332 | { |
428870ff BB |
1333 | uint64_t top; |
1334 | vdev_t *rvd = spa->spa_root_vdev; | |
1335 | vdev_t *tvd; | |
34dc7c2f | 1336 | |
428870ff | 1337 | ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); |
34dc7c2f | 1338 | |
428870ff BB |
1339 | do { |
1340 | top = ztest_random(rvd->vdev_children); | |
1341 | tvd = rvd->vdev_child[top]; | |
a1d477c2 | 1342 | } while (!vdev_is_concrete(tvd) || (tvd->vdev_islog && !log_ok) || |
428870ff | 1343 | tvd->vdev_mg == NULL || tvd->vdev_mg->mg_class == NULL); |
34dc7c2f | 1344 | |
428870ff | 1345 | return (top); |
34dc7c2f BB |
1346 | } |
1347 | ||
428870ff BB |
1348 | static uint64_t |
1349 | ztest_random_dsl_prop(zfs_prop_t prop) | |
34dc7c2f | 1350 | { |
428870ff BB |
1351 | uint64_t value; |
1352 | ||
1353 | do { | |
1354 | value = zfs_prop_random_value(prop, ztest_random(-1ULL)); | |
1355 | } while (prop == ZFS_PROP_CHECKSUM && value == ZIO_CHECKSUM_OFF); | |
1356 | ||
1357 | return (value); | |
34dc7c2f BB |
1358 | } |
1359 | ||
34dc7c2f | 1360 | static int |
428870ff BB |
1361 | ztest_dsl_prop_set_uint64(char *osname, zfs_prop_t prop, uint64_t value, |
1362 | boolean_t inherit) | |
34dc7c2f | 1363 | { |
428870ff BB |
1364 | const char *propname = zfs_prop_to_name(prop); |
1365 | const char *valname; | |
40b84e7a | 1366 | char *setpoint; |
428870ff | 1367 | uint64_t curval; |
34dc7c2f BB |
1368 | int error; |
1369 | ||
13fe0198 MA |
1370 | error = dsl_prop_set_int(osname, propname, |
1371 | (inherit ? ZPROP_SRC_NONE : ZPROP_SRC_LOCAL), value); | |
34dc7c2f | 1372 | |
428870ff BB |
1373 | if (error == ENOSPC) { |
1374 | ztest_record_enospc(FTAG); | |
34dc7c2f BB |
1375 | return (error); |
1376 | } | |
c99c9001 | 1377 | ASSERT0(error); |
34dc7c2f | 1378 | |
40b84e7a | 1379 | setpoint = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
13fe0198 | 1380 | VERIFY0(dsl_prop_get_integer(osname, propname, &curval, setpoint)); |
428870ff | 1381 | |
c242c188 | 1382 | if (ztest_opts.zo_verbose >= 6) { |
6bec4351 TC |
1383 | int err; |
1384 | ||
1385 | err = zfs_prop_index_to_string(prop, curval, &valname); | |
1386 | if (err) | |
02730c33 BB |
1387 | (void) printf("%s %s = %llu at '%s'\n", osname, |
1388 | propname, (unsigned long long)curval, setpoint); | |
6bec4351 TC |
1389 | else |
1390 | (void) printf("%s %s = %s at '%s'\n", | |
1391 | osname, propname, valname, setpoint); | |
34dc7c2f | 1392 | } |
40b84e7a | 1393 | umem_free(setpoint, MAXPATHLEN); |
34dc7c2f BB |
1394 | |
1395 | return (error); | |
1396 | } | |
1397 | ||
1398 | static int | |
c242c188 | 1399 | ztest_spa_prop_set_uint64(zpool_prop_t prop, uint64_t value) |
34dc7c2f | 1400 | { |
c242c188 | 1401 | spa_t *spa = ztest_spa; |
428870ff | 1402 | nvlist_t *props = NULL; |
34dc7c2f BB |
1403 | int error; |
1404 | ||
428870ff BB |
1405 | VERIFY(nvlist_alloc(&props, NV_UNIQUE_NAME, 0) == 0); |
1406 | VERIFY(nvlist_add_uint64(props, zpool_prop_to_name(prop), value) == 0); | |
34dc7c2f | 1407 | |
428870ff BB |
1408 | error = spa_prop_set(spa, props); |
1409 | ||
1410 | nvlist_free(props); | |
1411 | ||
1412 | if (error == ENOSPC) { | |
1413 | ztest_record_enospc(FTAG); | |
34dc7c2f BB |
1414 | return (error); |
1415 | } | |
c99c9001 | 1416 | ASSERT0(error); |
34dc7c2f BB |
1417 | |
1418 | return (error); | |
1419 | } | |
1420 | ||
4807c0ba TC |
1421 | static int |
1422 | ztest_dmu_objset_own(const char *name, dmu_objset_type_t type, | |
1423 | boolean_t readonly, boolean_t decrypt, void *tag, objset_t **osp) | |
1424 | { | |
1425 | int err; | |
f44ad929 TC |
1426 | char *cp = NULL; |
1427 | char ddname[ZFS_MAX_DATASET_NAME_LEN]; | |
1428 | ||
1429 | strcpy(ddname, name); | |
1430 | cp = strchr(ddname, '@'); | |
1431 | if (cp != NULL) | |
1432 | *cp = '\0'; | |
4807c0ba TC |
1433 | |
1434 | err = dmu_objset_own(name, type, readonly, decrypt, tag, osp); | |
f44ad929 | 1435 | while (decrypt && err == EACCES) { |
4807c0ba TC |
1436 | dsl_crypto_params_t *dcp; |
1437 | nvlist_t *crypto_args = fnvlist_alloc(); | |
4807c0ba TC |
1438 | |
1439 | fnvlist_add_uint8_array(crypto_args, "wkeydata", | |
1440 | (uint8_t *)ztest_wkeydata, WRAPPING_KEY_LEN); | |
1441 | VERIFY0(dsl_crypto_params_create_nvlist(DCP_CMD_NONE, NULL, | |
1442 | crypto_args, &dcp)); | |
1443 | err = spa_keystore_load_wkey(ddname, dcp, B_FALSE); | |
1444 | dsl_crypto_params_free(dcp, B_FALSE); | |
1445 | fnvlist_free(crypto_args); | |
1446 | ||
f44ad929 TC |
1447 | if (err == EINVAL) { |
1448 | /* | |
1449 | * We couldn't load a key for this dataset so try | |
1450 | * the parent. This loop will eventually hit the | |
1451 | * encryption root since ztest only makes clones | |
1452 | * as children of their origin datasets. | |
1453 | */ | |
1454 | cp = strrchr(ddname, '/'); | |
1455 | if (cp == NULL) | |
1456 | return (err); | |
1457 | ||
1458 | *cp = '\0'; | |
1459 | err = EACCES; | |
1460 | continue; | |
1461 | } else if (err != 0) { | |
1462 | break; | |
1463 | } | |
4807c0ba TC |
1464 | |
1465 | err = dmu_objset_own(name, type, readonly, decrypt, tag, osp); | |
f44ad929 | 1466 | break; |
4807c0ba TC |
1467 | } |
1468 | ||
1469 | return (err); | |
1470 | } | |
1471 | ||
428870ff BB |
1472 | static void |
1473 | ztest_rll_init(rll_t *rll) | |
1474 | { | |
1475 | rll->rll_writer = NULL; | |
1476 | rll->rll_readers = 0; | |
1e33ac1e BB |
1477 | mutex_init(&rll->rll_lock, NULL, MUTEX_DEFAULT, NULL); |
1478 | cv_init(&rll->rll_cv, NULL, CV_DEFAULT, NULL); | |
428870ff | 1479 | } |
34dc7c2f | 1480 | |
428870ff BB |
1481 | static void |
1482 | ztest_rll_destroy(rll_t *rll) | |
34dc7c2f | 1483 | { |
428870ff BB |
1484 | ASSERT(rll->rll_writer == NULL); |
1485 | ASSERT(rll->rll_readers == 0); | |
1e33ac1e BB |
1486 | mutex_destroy(&rll->rll_lock); |
1487 | cv_destroy(&rll->rll_cv); | |
428870ff | 1488 | } |
34dc7c2f | 1489 | |
428870ff BB |
1490 | static void |
1491 | ztest_rll_lock(rll_t *rll, rl_type_t type) | |
1492 | { | |
1e33ac1e | 1493 | mutex_enter(&rll->rll_lock); |
34dc7c2f | 1494 | |
428870ff BB |
1495 | if (type == RL_READER) { |
1496 | while (rll->rll_writer != NULL) | |
1e33ac1e | 1497 | (void) cv_wait(&rll->rll_cv, &rll->rll_lock); |
428870ff BB |
1498 | rll->rll_readers++; |
1499 | } else { | |
1500 | while (rll->rll_writer != NULL || rll->rll_readers) | |
1e33ac1e | 1501 | (void) cv_wait(&rll->rll_cv, &rll->rll_lock); |
428870ff BB |
1502 | rll->rll_writer = curthread; |
1503 | } | |
34dc7c2f | 1504 | |
1e33ac1e | 1505 | mutex_exit(&rll->rll_lock); |
428870ff | 1506 | } |
34dc7c2f | 1507 | |
428870ff BB |
1508 | static void |
1509 | ztest_rll_unlock(rll_t *rll) | |
1510 | { | |
1e33ac1e | 1511 | mutex_enter(&rll->rll_lock); |
34dc7c2f | 1512 | |
428870ff BB |
1513 | if (rll->rll_writer) { |
1514 | ASSERT(rll->rll_readers == 0); | |
1515 | rll->rll_writer = NULL; | |
1516 | } else { | |
1517 | ASSERT(rll->rll_readers != 0); | |
1518 | ASSERT(rll->rll_writer == NULL); | |
1519 | rll->rll_readers--; | |
1520 | } | |
34dc7c2f | 1521 | |
428870ff | 1522 | if (rll->rll_writer == NULL && rll->rll_readers == 0) |
1e33ac1e | 1523 | cv_broadcast(&rll->rll_cv); |
428870ff | 1524 | |
1e33ac1e | 1525 | mutex_exit(&rll->rll_lock); |
34dc7c2f BB |
1526 | } |
1527 | ||
428870ff BB |
1528 | static void |
1529 | ztest_object_lock(ztest_ds_t *zd, uint64_t object, rl_type_t type) | |
b128c09f | 1530 | { |
428870ff | 1531 | rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; |
b128c09f | 1532 | |
428870ff BB |
1533 | ztest_rll_lock(rll, type); |
1534 | } | |
b128c09f | 1535 | |
428870ff BB |
1536 | static void |
1537 | ztest_object_unlock(ztest_ds_t *zd, uint64_t object) | |
1538 | { | |
1539 | rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; | |
b128c09f | 1540 | |
428870ff | 1541 | ztest_rll_unlock(rll); |
b128c09f BB |
1542 | } |
1543 | ||
5d43cc9a | 1544 | static rl_t * |
e3a07cd0 BP |
1545 | ztest_range_lock(ztest_ds_t *zd, uint64_t object, uint64_t offset, |
1546 | uint64_t size, rl_type_t type) | |
1547 | { | |
5d43cc9a MA |
1548 | uint64_t hash = object ^ (offset % (ZTEST_RANGE_LOCKS + 1)); |
1549 | rll_t *rll = &zd->zd_range_lock[hash & (ZTEST_RANGE_LOCKS - 1)]; | |
1550 | rl_t *rl; | |
1551 | ||
1552 | rl = umem_alloc(sizeof (*rl), UMEM_NOFAIL); | |
1553 | rl->rl_object = object; | |
1554 | rl->rl_offset = offset; | |
1555 | rl->rl_size = size; | |
1556 | rl->rl_lock = rll; | |
1557 | ||
1558 | ztest_rll_lock(rll, type); | |
1559 | ||
1560 | return (rl); | |
e3a07cd0 | 1561 | } |
34dc7c2f | 1562 | |
e3a07cd0 | 1563 | static void |
5d43cc9a | 1564 | ztest_range_unlock(rl_t *rl) |
e3a07cd0 | 1565 | { |
5d43cc9a MA |
1566 | rll_t *rll = rl->rl_lock; |
1567 | ||
1568 | ztest_rll_unlock(rll); | |
1569 | ||
1570 | umem_free(rl, sizeof (*rl)); | |
428870ff | 1571 | } |
34dc7c2f | 1572 | |
428870ff | 1573 | static void |
c242c188 | 1574 | ztest_zd_init(ztest_ds_t *zd, ztest_shared_ds_t *szd, objset_t *os) |
428870ff BB |
1575 | { |
1576 | zd->zd_os = os; | |
1577 | zd->zd_zilog = dmu_objset_zil(os); | |
c242c188 | 1578 | zd->zd_shared = szd; |
428870ff | 1579 | dmu_objset_name(os, zd->zd_name); |
d6320ddb | 1580 | int l; |
428870ff | 1581 | |
c242c188 CS |
1582 | if (zd->zd_shared != NULL) |
1583 | zd->zd_shared->zd_seq = 0; | |
1584 | ||
2d9142c9 | 1585 | VERIFY0(pthread_rwlock_init(&zd->zd_zilog_lock, NULL)); |
1e33ac1e | 1586 | mutex_init(&zd->zd_dirobj_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 1587 | |
d6320ddb | 1588 | for (l = 0; l < ZTEST_OBJECT_LOCKS; l++) |
428870ff | 1589 | ztest_rll_init(&zd->zd_object_lock[l]); |
34dc7c2f | 1590 | |
d6320ddb | 1591 | for (l = 0; l < ZTEST_RANGE_LOCKS; l++) |
5d43cc9a | 1592 | ztest_rll_init(&zd->zd_range_lock[l]); |
34dc7c2f BB |
1593 | } |
1594 | ||
428870ff BB |
1595 | static void |
1596 | ztest_zd_fini(ztest_ds_t *zd) | |
34dc7c2f | 1597 | { |
d6320ddb BB |
1598 | int l; |
1599 | ||
1e33ac1e | 1600 | mutex_destroy(&zd->zd_dirobj_lock); |
2d9142c9 | 1601 | (void) pthread_rwlock_destroy(&zd->zd_zilog_lock); |
34dc7c2f | 1602 | |
d6320ddb | 1603 | for (l = 0; l < ZTEST_OBJECT_LOCKS; l++) |
428870ff | 1604 | ztest_rll_destroy(&zd->zd_object_lock[l]); |
b128c09f | 1605 | |
d6320ddb | 1606 | for (l = 0; l < ZTEST_RANGE_LOCKS; l++) |
5d43cc9a | 1607 | ztest_rll_destroy(&zd->zd_range_lock[l]); |
428870ff | 1608 | } |
b128c09f | 1609 | |
428870ff | 1610 | #define TXG_MIGHTWAIT (ztest_random(10) == 0 ? TXG_NOWAIT : TXG_WAIT) |
b128c09f | 1611 | |
428870ff BB |
1612 | static uint64_t |
1613 | ztest_tx_assign(dmu_tx_t *tx, uint64_t txg_how, const char *tag) | |
1614 | { | |
1615 | uint64_t txg; | |
1616 | int error; | |
1617 | ||
1618 | /* | |
1619 | * Attempt to assign tx to some transaction group. | |
1620 | */ | |
1621 | error = dmu_tx_assign(tx, txg_how); | |
1622 | if (error) { | |
1623 | if (error == ERESTART) { | |
1624 | ASSERT(txg_how == TXG_NOWAIT); | |
1625 | dmu_tx_wait(tx); | |
1626 | } else { | |
1627 | ASSERT3U(error, ==, ENOSPC); | |
1628 | ztest_record_enospc(tag); | |
1629 | } | |
1630 | dmu_tx_abort(tx); | |
1631 | return (0); | |
1632 | } | |
1633 | txg = dmu_tx_get_txg(tx); | |
1634 | ASSERT(txg != 0); | |
1635 | return (txg); | |
1636 | } | |
1637 | ||
428870ff BB |
1638 | static void |
1639 | ztest_bt_generate(ztest_block_tag_t *bt, objset_t *os, uint64_t object, | |
50c957f7 NB |
1640 | uint64_t dnodesize, uint64_t offset, uint64_t gen, uint64_t txg, |
1641 | uint64_t crtxg) | |
428870ff BB |
1642 | { |
1643 | bt->bt_magic = BT_MAGIC; | |
1644 | bt->bt_objset = dmu_objset_id(os); | |
1645 | bt->bt_object = object; | |
50c957f7 | 1646 | bt->bt_dnodesize = dnodesize; |
428870ff BB |
1647 | bt->bt_offset = offset; |
1648 | bt->bt_gen = gen; | |
1649 | bt->bt_txg = txg; | |
1650 | bt->bt_crtxg = crtxg; | |
1651 | } | |
1652 | ||
1653 | static void | |
1654 | ztest_bt_verify(ztest_block_tag_t *bt, objset_t *os, uint64_t object, | |
50c957f7 NB |
1655 | uint64_t dnodesize, uint64_t offset, uint64_t gen, uint64_t txg, |
1656 | uint64_t crtxg) | |
428870ff | 1657 | { |
9b67f605 MA |
1658 | ASSERT3U(bt->bt_magic, ==, BT_MAGIC); |
1659 | ASSERT3U(bt->bt_objset, ==, dmu_objset_id(os)); | |
1660 | ASSERT3U(bt->bt_object, ==, object); | |
50c957f7 | 1661 | ASSERT3U(bt->bt_dnodesize, ==, dnodesize); |
9b67f605 MA |
1662 | ASSERT3U(bt->bt_offset, ==, offset); |
1663 | ASSERT3U(bt->bt_gen, <=, gen); | |
1664 | ASSERT3U(bt->bt_txg, <=, txg); | |
1665 | ASSERT3U(bt->bt_crtxg, ==, crtxg); | |
428870ff BB |
1666 | } |
1667 | ||
1668 | static ztest_block_tag_t * | |
1669 | ztest_bt_bonus(dmu_buf_t *db) | |
1670 | { | |
1671 | dmu_object_info_t doi; | |
1672 | ztest_block_tag_t *bt; | |
1673 | ||
1674 | dmu_object_info_from_db(db, &doi); | |
1675 | ASSERT3U(doi.doi_bonus_size, <=, db->db_size); | |
1676 | ASSERT3U(doi.doi_bonus_size, >=, sizeof (*bt)); | |
1677 | bt = (void *)((char *)db->db_data + doi.doi_bonus_size - sizeof (*bt)); | |
1678 | ||
1679 | return (bt); | |
1680 | } | |
1681 | ||
50c957f7 NB |
1682 | /* |
1683 | * Generate a token to fill up unused bonus buffer space. Try to make | |
1684 | * it unique to the object, generation, and offset to verify that data | |
1685 | * is not getting overwritten by data from other dnodes. | |
1686 | */ | |
1687 | #define ZTEST_BONUS_FILL_TOKEN(obj, ds, gen, offset) \ | |
1688 | (((ds) << 48) | ((gen) << 32) | ((obj) << 8) | (offset)) | |
1689 | ||
1690 | /* | |
1691 | * Fill up the unused bonus buffer region before the block tag with a | |
1692 | * verifiable pattern. Filling the whole bonus area with non-zero data | |
1693 | * helps ensure that all dnode traversal code properly skips the | |
1694 | * interior regions of large dnodes. | |
1695 | */ | |
65c7cc49 | 1696 | static void |
50c957f7 NB |
1697 | ztest_fill_unused_bonus(dmu_buf_t *db, void *end, uint64_t obj, |
1698 | objset_t *os, uint64_t gen) | |
1699 | { | |
1700 | uint64_t *bonusp; | |
1701 | ||
1702 | ASSERT(IS_P2ALIGNED((char *)end - (char *)db->db_data, 8)); | |
1703 | ||
1704 | for (bonusp = db->db_data; bonusp < (uint64_t *)end; bonusp++) { | |
1705 | uint64_t token = ZTEST_BONUS_FILL_TOKEN(obj, dmu_objset_id(os), | |
1706 | gen, bonusp - (uint64_t *)db->db_data); | |
1707 | *bonusp = token; | |
1708 | } | |
1709 | } | |
1710 | ||
1711 | /* | |
1712 | * Verify that the unused area of a bonus buffer is filled with the | |
1713 | * expected tokens. | |
1714 | */ | |
65c7cc49 | 1715 | static void |
50c957f7 NB |
1716 | ztest_verify_unused_bonus(dmu_buf_t *db, void *end, uint64_t obj, |
1717 | objset_t *os, uint64_t gen) | |
1718 | { | |
1719 | uint64_t *bonusp; | |
1720 | ||
1721 | for (bonusp = db->db_data; bonusp < (uint64_t *)end; bonusp++) { | |
1722 | uint64_t token = ZTEST_BONUS_FILL_TOKEN(obj, dmu_objset_id(os), | |
1723 | gen, bonusp - (uint64_t *)db->db_data); | |
1724 | VERIFY3U(*bonusp, ==, token); | |
1725 | } | |
1726 | } | |
1727 | ||
428870ff BB |
1728 | /* |
1729 | * ZIL logging ops | |
1730 | */ | |
1731 | ||
1732 | #define lrz_type lr_mode | |
1733 | #define lrz_blocksize lr_uid | |
1734 | #define lrz_ibshift lr_gid | |
1735 | #define lrz_bonustype lr_rdev | |
50c957f7 | 1736 | #define lrz_dnodesize lr_crtime[1] |
428870ff | 1737 | |
572e2857 | 1738 | static void |
428870ff BB |
1739 | ztest_log_create(ztest_ds_t *zd, dmu_tx_t *tx, lr_create_t *lr) |
1740 | { | |
1741 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1742 | size_t namesize = strlen(name) + 1; | |
1743 | itx_t *itx; | |
1744 | ||
1745 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1746 | return; |
428870ff BB |
1747 | |
1748 | itx = zil_itx_create(TX_CREATE, sizeof (*lr) + namesize); | |
1749 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1750 | sizeof (*lr) + namesize - sizeof (lr_t)); | |
1751 | ||
572e2857 | 1752 | zil_itx_assign(zd->zd_zilog, itx, tx); |
428870ff BB |
1753 | } |
1754 | ||
572e2857 BB |
1755 | static void |
1756 | ztest_log_remove(ztest_ds_t *zd, dmu_tx_t *tx, lr_remove_t *lr, uint64_t object) | |
428870ff BB |
1757 | { |
1758 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1759 | size_t namesize = strlen(name) + 1; | |
1760 | itx_t *itx; | |
1761 | ||
1762 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1763 | return; |
428870ff BB |
1764 | |
1765 | itx = zil_itx_create(TX_REMOVE, sizeof (*lr) + namesize); | |
1766 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1767 | sizeof (*lr) + namesize - sizeof (lr_t)); | |
1768 | ||
572e2857 BB |
1769 | itx->itx_oid = object; |
1770 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1771 | } |
1772 | ||
572e2857 | 1773 | static void |
428870ff BB |
1774 | ztest_log_write(ztest_ds_t *zd, dmu_tx_t *tx, lr_write_t *lr) |
1775 | { | |
1776 | itx_t *itx; | |
1777 | itx_wr_state_t write_state = ztest_random(WR_NUM_STATES); | |
1778 | ||
1779 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1780 | return; |
428870ff | 1781 | |
b8738257 | 1782 | if (lr->lr_length > zil_max_log_data(zd->zd_zilog)) |
428870ff BB |
1783 | write_state = WR_INDIRECT; |
1784 | ||
1785 | itx = zil_itx_create(TX_WRITE, | |
1786 | sizeof (*lr) + (write_state == WR_COPIED ? lr->lr_length : 0)); | |
1787 | ||
1788 | if (write_state == WR_COPIED && | |
1789 | dmu_read(zd->zd_os, lr->lr_foid, lr->lr_offset, lr->lr_length, | |
1790 | ((lr_write_t *)&itx->itx_lr) + 1, DMU_READ_NO_PREFETCH) != 0) { | |
1791 | zil_itx_destroy(itx); | |
1792 | itx = zil_itx_create(TX_WRITE, sizeof (*lr)); | |
1793 | write_state = WR_NEED_COPY; | |
1794 | } | |
1795 | itx->itx_private = zd; | |
1796 | itx->itx_wr_state = write_state; | |
1797 | itx->itx_sync = (ztest_random(8) == 0); | |
428870ff BB |
1798 | |
1799 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1800 | sizeof (*lr) - sizeof (lr_t)); | |
1801 | ||
572e2857 | 1802 | zil_itx_assign(zd->zd_zilog, itx, tx); |
428870ff BB |
1803 | } |
1804 | ||
572e2857 | 1805 | static void |
428870ff BB |
1806 | ztest_log_truncate(ztest_ds_t *zd, dmu_tx_t *tx, lr_truncate_t *lr) |
1807 | { | |
1808 | itx_t *itx; | |
1809 | ||
1810 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1811 | return; |
428870ff BB |
1812 | |
1813 | itx = zil_itx_create(TX_TRUNCATE, sizeof (*lr)); | |
1814 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1815 | sizeof (*lr) - sizeof (lr_t)); | |
1816 | ||
572e2857 BB |
1817 | itx->itx_sync = B_FALSE; |
1818 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1819 | } |
1820 | ||
572e2857 | 1821 | static void |
428870ff BB |
1822 | ztest_log_setattr(ztest_ds_t *zd, dmu_tx_t *tx, lr_setattr_t *lr) |
1823 | { | |
1824 | itx_t *itx; | |
1825 | ||
1826 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1827 | return; |
428870ff BB |
1828 | |
1829 | itx = zil_itx_create(TX_SETATTR, sizeof (*lr)); | |
1830 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1831 | sizeof (*lr) - sizeof (lr_t)); | |
1832 | ||
572e2857 BB |
1833 | itx->itx_sync = B_FALSE; |
1834 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1835 | } |
1836 | ||
1837 | /* | |
1838 | * ZIL replay ops | |
1839 | */ | |
1840 | static int | |
867959b5 | 1841 | ztest_replay_create(void *arg1, void *arg2, boolean_t byteswap) |
428870ff | 1842 | { |
867959b5 BB |
1843 | ztest_ds_t *zd = arg1; |
1844 | lr_create_t *lr = arg2; | |
428870ff BB |
1845 | char *name = (void *)(lr + 1); /* name follows lr */ |
1846 | objset_t *os = zd->zd_os; | |
1847 | ztest_block_tag_t *bbt; | |
1848 | dmu_buf_t *db; | |
1849 | dmu_tx_t *tx; | |
1850 | uint64_t txg; | |
1851 | int error = 0; | |
50c957f7 | 1852 | int bonuslen; |
428870ff BB |
1853 | |
1854 | if (byteswap) | |
1855 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1856 | ||
1857 | ASSERT(lr->lr_doid == ZTEST_DIROBJ); | |
1858 | ASSERT(name[0] != '\0'); | |
1859 | ||
1860 | tx = dmu_tx_create(os); | |
1861 | ||
1862 | dmu_tx_hold_zap(tx, lr->lr_doid, B_TRUE, name); | |
1863 | ||
1864 | if (lr->lrz_type == DMU_OT_ZAP_OTHER) { | |
1865 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, B_TRUE, NULL); | |
1866 | } else { | |
1867 | dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT); | |
1868 | } | |
1869 | ||
1870 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1871 | if (txg == 0) | |
1872 | return (ENOSPC); | |
1873 | ||
1874 | ASSERT(dmu_objset_zil(os)->zl_replay == !!lr->lr_foid); | |
50c957f7 | 1875 | bonuslen = DN_BONUS_SIZE(lr->lrz_dnodesize); |
428870ff BB |
1876 | |
1877 | if (lr->lrz_type == DMU_OT_ZAP_OTHER) { | |
1878 | if (lr->lr_foid == 0) { | |
50c957f7 | 1879 | lr->lr_foid = zap_create_dnsize(os, |
428870ff | 1880 | lr->lrz_type, lr->lrz_bonustype, |
50c957f7 | 1881 | bonuslen, lr->lrz_dnodesize, tx); |
428870ff | 1882 | } else { |
50c957f7 | 1883 | error = zap_create_claim_dnsize(os, lr->lr_foid, |
428870ff | 1884 | lr->lrz_type, lr->lrz_bonustype, |
50c957f7 | 1885 | bonuslen, lr->lrz_dnodesize, tx); |
428870ff BB |
1886 | } |
1887 | } else { | |
1888 | if (lr->lr_foid == 0) { | |
50c957f7 | 1889 | lr->lr_foid = dmu_object_alloc_dnsize(os, |
428870ff | 1890 | lr->lrz_type, 0, lr->lrz_bonustype, |
50c957f7 | 1891 | bonuslen, lr->lrz_dnodesize, tx); |
428870ff | 1892 | } else { |
50c957f7 | 1893 | error = dmu_object_claim_dnsize(os, lr->lr_foid, |
428870ff | 1894 | lr->lrz_type, 0, lr->lrz_bonustype, |
50c957f7 | 1895 | bonuslen, lr->lrz_dnodesize, tx); |
428870ff BB |
1896 | } |
1897 | } | |
1898 | ||
1899 | if (error) { | |
1900 | ASSERT3U(error, ==, EEXIST); | |
1901 | ASSERT(zd->zd_zilog->zl_replay); | |
1902 | dmu_tx_commit(tx); | |
1903 | return (error); | |
1904 | } | |
1905 | ||
1906 | ASSERT(lr->lr_foid != 0); | |
1907 | ||
1908 | if (lr->lrz_type != DMU_OT_ZAP_OTHER) | |
1909 | VERIFY3U(0, ==, dmu_object_set_blocksize(os, lr->lr_foid, | |
1910 | lr->lrz_blocksize, lr->lrz_ibshift, tx)); | |
1911 | ||
1912 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
1913 | bbt = ztest_bt_bonus(db); | |
1914 | dmu_buf_will_dirty(db, tx); | |
50c957f7 NB |
1915 | ztest_bt_generate(bbt, os, lr->lr_foid, lr->lrz_dnodesize, -1ULL, |
1916 | lr->lr_gen, txg, txg); | |
1917 | ztest_fill_unused_bonus(db, bbt, lr->lr_foid, os, lr->lr_gen); | |
428870ff BB |
1918 | dmu_buf_rele(db, FTAG); |
1919 | ||
1920 | VERIFY3U(0, ==, zap_add(os, lr->lr_doid, name, sizeof (uint64_t), 1, | |
1921 | &lr->lr_foid, tx)); | |
1922 | ||
1923 | (void) ztest_log_create(zd, tx, lr); | |
1924 | ||
1925 | dmu_tx_commit(tx); | |
1926 | ||
1927 | return (0); | |
1928 | } | |
1929 | ||
1930 | static int | |
867959b5 | 1931 | ztest_replay_remove(void *arg1, void *arg2, boolean_t byteswap) |
428870ff | 1932 | { |
867959b5 BB |
1933 | ztest_ds_t *zd = arg1; |
1934 | lr_remove_t *lr = arg2; | |
428870ff BB |
1935 | char *name = (void *)(lr + 1); /* name follows lr */ |
1936 | objset_t *os = zd->zd_os; | |
1937 | dmu_object_info_t doi; | |
1938 | dmu_tx_t *tx; | |
1939 | uint64_t object, txg; | |
1940 | ||
1941 | if (byteswap) | |
1942 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1943 | ||
1944 | ASSERT(lr->lr_doid == ZTEST_DIROBJ); | |
1945 | ASSERT(name[0] != '\0'); | |
1946 | ||
1947 | VERIFY3U(0, ==, | |
1948 | zap_lookup(os, lr->lr_doid, name, sizeof (object), 1, &object)); | |
1949 | ASSERT(object != 0); | |
1950 | ||
1951 | ztest_object_lock(zd, object, RL_WRITER); | |
1952 | ||
1953 | VERIFY3U(0, ==, dmu_object_info(os, object, &doi)); | |
1954 | ||
1955 | tx = dmu_tx_create(os); | |
1956 | ||
1957 | dmu_tx_hold_zap(tx, lr->lr_doid, B_FALSE, name); | |
1958 | dmu_tx_hold_free(tx, object, 0, DMU_OBJECT_END); | |
1959 | ||
1960 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1961 | if (txg == 0) { | |
1962 | ztest_object_unlock(zd, object); | |
1963 | return (ENOSPC); | |
1964 | } | |
1965 | ||
1966 | if (doi.doi_type == DMU_OT_ZAP_OTHER) { | |
1967 | VERIFY3U(0, ==, zap_destroy(os, object, tx)); | |
1968 | } else { | |
1969 | VERIFY3U(0, ==, dmu_object_free(os, object, tx)); | |
1970 | } | |
1971 | ||
1972 | VERIFY3U(0, ==, zap_remove(os, lr->lr_doid, name, tx)); | |
1973 | ||
572e2857 | 1974 | (void) ztest_log_remove(zd, tx, lr, object); |
428870ff BB |
1975 | |
1976 | dmu_tx_commit(tx); | |
1977 | ||
1978 | ztest_object_unlock(zd, object); | |
1979 | ||
1980 | return (0); | |
1981 | } | |
1982 | ||
1983 | static int | |
867959b5 | 1984 | ztest_replay_write(void *arg1, void *arg2, boolean_t byteswap) |
428870ff | 1985 | { |
867959b5 BB |
1986 | ztest_ds_t *zd = arg1; |
1987 | lr_write_t *lr = arg2; | |
428870ff BB |
1988 | objset_t *os = zd->zd_os; |
1989 | void *data = lr + 1; /* data follows lr */ | |
1990 | uint64_t offset, length; | |
1991 | ztest_block_tag_t *bt = data; | |
1992 | ztest_block_tag_t *bbt; | |
1993 | uint64_t gen, txg, lrtxg, crtxg; | |
1994 | dmu_object_info_t doi; | |
1995 | dmu_tx_t *tx; | |
1996 | dmu_buf_t *db; | |
1997 | arc_buf_t *abuf = NULL; | |
5d43cc9a | 1998 | rl_t *rl; |
428870ff BB |
1999 | |
2000 | if (byteswap) | |
2001 | byteswap_uint64_array(lr, sizeof (*lr)); | |
2002 | ||
2003 | offset = lr->lr_offset; | |
2004 | length = lr->lr_length; | |
2005 | ||
2006 | /* If it's a dmu_sync() block, write the whole block */ | |
2007 | if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) { | |
2008 | uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr); | |
2009 | if (length < blocksize) { | |
2010 | offset -= offset % blocksize; | |
2011 | length = blocksize; | |
2012 | } | |
2013 | } | |
2014 | ||
2015 | if (bt->bt_magic == BSWAP_64(BT_MAGIC)) | |
2016 | byteswap_uint64_array(bt, sizeof (*bt)); | |
2017 | ||
2018 | if (bt->bt_magic != BT_MAGIC) | |
2019 | bt = NULL; | |
2020 | ||
2021 | ztest_object_lock(zd, lr->lr_foid, RL_READER); | |
2022 | rl = ztest_range_lock(zd, lr->lr_foid, offset, length, RL_WRITER); | |
2023 | ||
2024 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
2025 | ||
2026 | dmu_object_info_from_db(db, &doi); | |
2027 | ||
2028 | bbt = ztest_bt_bonus(db); | |
2029 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
2030 | gen = bbt->bt_gen; | |
2031 | crtxg = bbt->bt_crtxg; | |
2032 | lrtxg = lr->lr_common.lrc_txg; | |
2033 | ||
2034 | tx = dmu_tx_create(os); | |
2035 | ||
2036 | dmu_tx_hold_write(tx, lr->lr_foid, offset, length); | |
2037 | ||
2038 | if (ztest_random(8) == 0 && length == doi.doi_data_block_size && | |
2039 | P2PHASE(offset, length) == 0) | |
2040 | abuf = dmu_request_arcbuf(db, length); | |
2041 | ||
2042 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
2043 | if (txg == 0) { | |
2044 | if (abuf != NULL) | |
2045 | dmu_return_arcbuf(abuf); | |
2046 | dmu_buf_rele(db, FTAG); | |
5d43cc9a | 2047 | ztest_range_unlock(rl); |
428870ff BB |
2048 | ztest_object_unlock(zd, lr->lr_foid); |
2049 | return (ENOSPC); | |
2050 | } | |
2051 | ||
2052 | if (bt != NULL) { | |
2053 | /* | |
2054 | * Usually, verify the old data before writing new data -- | |
2055 | * but not always, because we also want to verify correct | |
2056 | * behavior when the data was not recently read into cache. | |
2057 | */ | |
2058 | ASSERT(offset % doi.doi_data_block_size == 0); | |
2059 | if (ztest_random(4) != 0) { | |
2060 | int prefetch = ztest_random(2) ? | |
2061 | DMU_READ_PREFETCH : DMU_READ_NO_PREFETCH; | |
2062 | ztest_block_tag_t rbt; | |
2063 | ||
2064 | VERIFY(dmu_read(os, lr->lr_foid, offset, | |
2065 | sizeof (rbt), &rbt, prefetch) == 0); | |
2066 | if (rbt.bt_magic == BT_MAGIC) { | |
50c957f7 | 2067 | ztest_bt_verify(&rbt, os, lr->lr_foid, 0, |
428870ff BB |
2068 | offset, gen, txg, crtxg); |
2069 | } | |
2070 | } | |
2071 | ||
2072 | /* | |
2073 | * Writes can appear to be newer than the bonus buffer because | |
2074 | * the ztest_get_data() callback does a dmu_read() of the | |
2075 | * open-context data, which may be different than the data | |
2076 | * as it was when the write was generated. | |
2077 | */ | |
2078 | if (zd->zd_zilog->zl_replay) { | |
50c957f7 | 2079 | ztest_bt_verify(bt, os, lr->lr_foid, 0, offset, |
428870ff BB |
2080 | MAX(gen, bt->bt_gen), MAX(txg, lrtxg), |
2081 | bt->bt_crtxg); | |
2082 | } | |
2083 | ||
2084 | /* | |
2085 | * Set the bt's gen/txg to the bonus buffer's gen/txg | |
2086 | * so that all of the usual ASSERTs will work. | |
2087 | */ | |
50c957f7 NB |
2088 | ztest_bt_generate(bt, os, lr->lr_foid, 0, offset, gen, txg, |
2089 | crtxg); | |
428870ff BB |
2090 | } |
2091 | ||
2092 | if (abuf == NULL) { | |
2093 | dmu_write(os, lr->lr_foid, offset, length, data, tx); | |
2094 | } else { | |
2095 | bcopy(data, abuf->b_data, length); | |
440a3eb9 | 2096 | dmu_assign_arcbuf_by_dbuf(db, offset, abuf, tx); |
428870ff BB |
2097 | } |
2098 | ||
2099 | (void) ztest_log_write(zd, tx, lr); | |
2100 | ||
2101 | dmu_buf_rele(db, FTAG); | |
2102 | ||
2103 | dmu_tx_commit(tx); | |
2104 | ||
5d43cc9a | 2105 | ztest_range_unlock(rl); |
428870ff BB |
2106 | ztest_object_unlock(zd, lr->lr_foid); |
2107 | ||
2108 | return (0); | |
2109 | } | |
2110 | ||
2111 | static int | |
867959b5 | 2112 | ztest_replay_truncate(void *arg1, void *arg2, boolean_t byteswap) |
428870ff | 2113 | { |
867959b5 BB |
2114 | ztest_ds_t *zd = arg1; |
2115 | lr_truncate_t *lr = arg2; | |
428870ff BB |
2116 | objset_t *os = zd->zd_os; |
2117 | dmu_tx_t *tx; | |
2118 | uint64_t txg; | |
5d43cc9a | 2119 | rl_t *rl; |
428870ff BB |
2120 | |
2121 | if (byteswap) | |
2122 | byteswap_uint64_array(lr, sizeof (*lr)); | |
2123 | ||
2124 | ztest_object_lock(zd, lr->lr_foid, RL_READER); | |
2125 | rl = ztest_range_lock(zd, lr->lr_foid, lr->lr_offset, lr->lr_length, | |
2126 | RL_WRITER); | |
2127 | ||
2128 | tx = dmu_tx_create(os); | |
2129 | ||
2130 | dmu_tx_hold_free(tx, lr->lr_foid, lr->lr_offset, lr->lr_length); | |
2131 | ||
2132 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
2133 | if (txg == 0) { | |
5d43cc9a | 2134 | ztest_range_unlock(rl); |
428870ff BB |
2135 | ztest_object_unlock(zd, lr->lr_foid); |
2136 | return (ENOSPC); | |
2137 | } | |
2138 | ||
2139 | VERIFY(dmu_free_range(os, lr->lr_foid, lr->lr_offset, | |
2140 | lr->lr_length, tx) == 0); | |
2141 | ||
2142 | (void) ztest_log_truncate(zd, tx, lr); | |
2143 | ||
2144 | dmu_tx_commit(tx); | |
2145 | ||
5d43cc9a | 2146 | ztest_range_unlock(rl); |
428870ff BB |
2147 | ztest_object_unlock(zd, lr->lr_foid); |
2148 | ||
2149 | return (0); | |
2150 | } | |
2151 | ||
2152 | static int | |
867959b5 | 2153 | ztest_replay_setattr(void *arg1, void *arg2, boolean_t byteswap) |
428870ff | 2154 | { |
867959b5 BB |
2155 | ztest_ds_t *zd = arg1; |
2156 | lr_setattr_t *lr = arg2; | |
428870ff BB |
2157 | objset_t *os = zd->zd_os; |
2158 | dmu_tx_t *tx; | |
2159 | dmu_buf_t *db; | |
2160 | ztest_block_tag_t *bbt; | |
50c957f7 | 2161 | uint64_t txg, lrtxg, crtxg, dnodesize; |
428870ff BB |
2162 | |
2163 | if (byteswap) | |
2164 | byteswap_uint64_array(lr, sizeof (*lr)); | |
2165 | ||
2166 | ztest_object_lock(zd, lr->lr_foid, RL_WRITER); | |
2167 | ||
2168 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
2169 | ||
2170 | tx = dmu_tx_create(os); | |
2171 | dmu_tx_hold_bonus(tx, lr->lr_foid); | |
2172 | ||
2173 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
2174 | if (txg == 0) { | |
2175 | dmu_buf_rele(db, FTAG); | |
2176 | ztest_object_unlock(zd, lr->lr_foid); | |
2177 | return (ENOSPC); | |
2178 | } | |
2179 | ||
2180 | bbt = ztest_bt_bonus(db); | |
2181 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
2182 | crtxg = bbt->bt_crtxg; | |
2183 | lrtxg = lr->lr_common.lrc_txg; | |
50c957f7 | 2184 | dnodesize = bbt->bt_dnodesize; |
428870ff BB |
2185 | |
2186 | if (zd->zd_zilog->zl_replay) { | |
2187 | ASSERT(lr->lr_size != 0); | |
2188 | ASSERT(lr->lr_mode != 0); | |
2189 | ASSERT(lrtxg != 0); | |
2190 | } else { | |
2191 | /* | |
2192 | * Randomly change the size and increment the generation. | |
2193 | */ | |
2194 | lr->lr_size = (ztest_random(db->db_size / sizeof (*bbt)) + 1) * | |
2195 | sizeof (*bbt); | |
2196 | lr->lr_mode = bbt->bt_gen + 1; | |
2197 | ASSERT(lrtxg == 0); | |
2198 | } | |
2199 | ||
2200 | /* | |
2201 | * Verify that the current bonus buffer is not newer than our txg. | |
2202 | */ | |
50c957f7 | 2203 | ztest_bt_verify(bbt, os, lr->lr_foid, dnodesize, -1ULL, lr->lr_mode, |
428870ff BB |
2204 | MAX(txg, lrtxg), crtxg); |
2205 | ||
2206 | dmu_buf_will_dirty(db, tx); | |
2207 | ||
2208 | ASSERT3U(lr->lr_size, >=, sizeof (*bbt)); | |
2209 | ASSERT3U(lr->lr_size, <=, db->db_size); | |
c99c9001 | 2210 | VERIFY0(dmu_set_bonus(db, lr->lr_size, tx)); |
428870ff BB |
2211 | bbt = ztest_bt_bonus(db); |
2212 | ||
50c957f7 NB |
2213 | ztest_bt_generate(bbt, os, lr->lr_foid, dnodesize, -1ULL, lr->lr_mode, |
2214 | txg, crtxg); | |
2215 | ztest_fill_unused_bonus(db, bbt, lr->lr_foid, os, bbt->bt_gen); | |
428870ff BB |
2216 | dmu_buf_rele(db, FTAG); |
2217 | ||
2218 | (void) ztest_log_setattr(zd, tx, lr); | |
2219 | ||
2220 | dmu_tx_commit(tx); | |
2221 | ||
2222 | ztest_object_unlock(zd, lr->lr_foid); | |
2223 | ||
2224 | return (0); | |
2225 | } | |
2226 | ||
867959b5 BB |
2227 | zil_replay_func_t *ztest_replay_vector[TX_MAX_TYPE] = { |
2228 | NULL, /* 0 no such transaction type */ | |
2229 | ztest_replay_create, /* TX_CREATE */ | |
2230 | NULL, /* TX_MKDIR */ | |
2231 | NULL, /* TX_MKXATTR */ | |
2232 | NULL, /* TX_SYMLINK */ | |
2233 | ztest_replay_remove, /* TX_REMOVE */ | |
2234 | NULL, /* TX_RMDIR */ | |
2235 | NULL, /* TX_LINK */ | |
2236 | NULL, /* TX_RENAME */ | |
2237 | ztest_replay_write, /* TX_WRITE */ | |
2238 | ztest_replay_truncate, /* TX_TRUNCATE */ | |
2239 | ztest_replay_setattr, /* TX_SETATTR */ | |
2240 | NULL, /* TX_ACL */ | |
2241 | NULL, /* TX_CREATE_ACL */ | |
2242 | NULL, /* TX_CREATE_ATTR */ | |
2243 | NULL, /* TX_CREATE_ACL_ATTR */ | |
2244 | NULL, /* TX_MKDIR_ACL */ | |
2245 | NULL, /* TX_MKDIR_ATTR */ | |
2246 | NULL, /* TX_MKDIR_ACL_ATTR */ | |
2247 | NULL, /* TX_WRITE2 */ | |
428870ff BB |
2248 | }; |
2249 | ||
2250 | /* | |
2251 | * ZIL get_data callbacks | |
2252 | */ | |
2253 | ||
900d09b2 | 2254 | /* ARGSUSED */ |
428870ff BB |
2255 | static void |
2256 | ztest_get_done(zgd_t *zgd, int error) | |
2257 | { | |
5d43cc9a MA |
2258 | ztest_ds_t *zd = zgd->zgd_private; |
2259 | uint64_t object = ((rl_t *)zgd->zgd_lr)->rl_object; | |
428870ff BB |
2260 | |
2261 | if (zgd->zgd_db) | |
2262 | dmu_buf_rele(zgd->zgd_db, zgd); | |
2263 | ||
5d43cc9a | 2264 | ztest_range_unlock((rl_t *)zgd->zgd_lr); |
428870ff BB |
2265 | ztest_object_unlock(zd, object); |
2266 | ||
428870ff BB |
2267 | umem_free(zgd, sizeof (*zgd)); |
2268 | } | |
2269 | ||
2270 | static int | |
1ce23dca PS |
2271 | ztest_get_data(void *arg, lr_write_t *lr, char *buf, struct lwb *lwb, |
2272 | zio_t *zio) | |
428870ff BB |
2273 | { |
2274 | ztest_ds_t *zd = arg; | |
2275 | objset_t *os = zd->zd_os; | |
2276 | uint64_t object = lr->lr_foid; | |
2277 | uint64_t offset = lr->lr_offset; | |
2278 | uint64_t size = lr->lr_length; | |
428870ff BB |
2279 | uint64_t txg = lr->lr_common.lrc_txg; |
2280 | uint64_t crtxg; | |
2281 | dmu_object_info_t doi; | |
2282 | dmu_buf_t *db; | |
2283 | zgd_t *zgd; | |
2284 | int error; | |
2285 | ||
1ce23dca PS |
2286 | ASSERT3P(lwb, !=, NULL); |
2287 | ASSERT3P(zio, !=, NULL); | |
2288 | ASSERT3U(size, !=, 0); | |
2289 | ||
428870ff BB |
2290 | ztest_object_lock(zd, object, RL_READER); |
2291 | error = dmu_bonus_hold(os, object, FTAG, &db); | |
2292 | if (error) { | |
2293 | ztest_object_unlock(zd, object); | |
2294 | return (error); | |
2295 | } | |
2296 | ||
2297 | crtxg = ztest_bt_bonus(db)->bt_crtxg; | |
2298 | ||
2299 | if (crtxg == 0 || crtxg > txg) { | |
2300 | dmu_buf_rele(db, FTAG); | |
2301 | ztest_object_unlock(zd, object); | |
2302 | return (ENOENT); | |
2303 | } | |
2304 | ||
2305 | dmu_object_info_from_db(db, &doi); | |
2306 | dmu_buf_rele(db, FTAG); | |
2307 | db = NULL; | |
2308 | ||
2309 | zgd = umem_zalloc(sizeof (*zgd), UMEM_NOFAIL); | |
1ce23dca | 2310 | zgd->zgd_lwb = lwb; |
5d43cc9a | 2311 | zgd->zgd_private = zd; |
428870ff BB |
2312 | |
2313 | if (buf != NULL) { /* immediate write */ | |
bd4dde8e | 2314 | zgd->zgd_lr = (struct zfs_locked_range *)ztest_range_lock(zd, |
5d43cc9a | 2315 | object, offset, size, RL_READER); |
428870ff BB |
2316 | |
2317 | error = dmu_read(os, object, offset, size, buf, | |
2318 | DMU_READ_NO_PREFETCH); | |
2319 | ASSERT(error == 0); | |
2320 | } else { | |
2321 | size = doi.doi_data_block_size; | |
2322 | if (ISP2(size)) { | |
2323 | offset = P2ALIGN(offset, size); | |
2324 | } else { | |
2325 | ASSERT(offset < size); | |
2326 | offset = 0; | |
2327 | } | |
2328 | ||
bd4dde8e | 2329 | zgd->zgd_lr = (struct zfs_locked_range *)ztest_range_lock(zd, |
5d43cc9a | 2330 | object, offset, size, RL_READER); |
428870ff BB |
2331 | |
2332 | error = dmu_buf_hold(os, object, offset, zgd, &db, | |
2333 | DMU_READ_NO_PREFETCH); | |
2334 | ||
2335 | if (error == 0) { | |
02dc43bc | 2336 | blkptr_t *bp = &lr->lr_blkptr; |
03c6040b | 2337 | |
428870ff BB |
2338 | zgd->zgd_db = db; |
2339 | zgd->zgd_bp = bp; | |
2340 | ||
2341 | ASSERT(db->db_offset == offset); | |
2342 | ASSERT(db->db_size == size); | |
2343 | ||
2344 | error = dmu_sync(zio, lr->lr_common.lrc_txg, | |
2345 | ztest_get_done, zgd); | |
2346 | ||
2347 | if (error == 0) | |
2348 | return (0); | |
2349 | } | |
2350 | } | |
2351 | ||
2352 | ztest_get_done(zgd, error); | |
2353 | ||
2354 | return (error); | |
2355 | } | |
2356 | ||
2357 | static void * | |
2358 | ztest_lr_alloc(size_t lrsize, char *name) | |
2359 | { | |
2360 | char *lr; | |
2361 | size_t namesize = name ? strlen(name) + 1 : 0; | |
2362 | ||
2363 | lr = umem_zalloc(lrsize + namesize, UMEM_NOFAIL); | |
2364 | ||
2365 | if (name) | |
2366 | bcopy(name, lr + lrsize, namesize); | |
2367 | ||
2368 | return (lr); | |
2369 | } | |
2370 | ||
65c7cc49 | 2371 | static void |
428870ff BB |
2372 | ztest_lr_free(void *lr, size_t lrsize, char *name) |
2373 | { | |
2374 | size_t namesize = name ? strlen(name) + 1 : 0; | |
2375 | ||
2376 | umem_free(lr, lrsize + namesize); | |
2377 | } | |
2378 | ||
2379 | /* | |
2380 | * Lookup a bunch of objects. Returns the number of objects not found. | |
2381 | */ | |
2382 | static int | |
2383 | ztest_lookup(ztest_ds_t *zd, ztest_od_t *od, int count) | |
2384 | { | |
2385 | int missing = 0; | |
2386 | int error; | |
d6320ddb | 2387 | int i; |
428870ff | 2388 | |
c25b8f99 | 2389 | ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock)); |
428870ff | 2390 | |
d6320ddb | 2391 | for (i = 0; i < count; i++, od++) { |
428870ff BB |
2392 | od->od_object = 0; |
2393 | error = zap_lookup(zd->zd_os, od->od_dir, od->od_name, | |
2394 | sizeof (uint64_t), 1, &od->od_object); | |
2395 | if (error) { | |
2396 | ASSERT(error == ENOENT); | |
2397 | ASSERT(od->od_object == 0); | |
2398 | missing++; | |
2399 | } else { | |
2400 | dmu_buf_t *db; | |
2401 | ztest_block_tag_t *bbt; | |
2402 | dmu_object_info_t doi; | |
2403 | ||
2404 | ASSERT(od->od_object != 0); | |
2405 | ASSERT(missing == 0); /* there should be no gaps */ | |
2406 | ||
2407 | ztest_object_lock(zd, od->od_object, RL_READER); | |
2408 | VERIFY3U(0, ==, dmu_bonus_hold(zd->zd_os, | |
2409 | od->od_object, FTAG, &db)); | |
2410 | dmu_object_info_from_db(db, &doi); | |
2411 | bbt = ztest_bt_bonus(db); | |
2412 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
2413 | od->od_type = doi.doi_type; | |
2414 | od->od_blocksize = doi.doi_data_block_size; | |
2415 | od->od_gen = bbt->bt_gen; | |
2416 | dmu_buf_rele(db, FTAG); | |
2417 | ztest_object_unlock(zd, od->od_object); | |
2418 | } | |
2419 | } | |
2420 | ||
2421 | return (missing); | |
2422 | } | |
2423 | ||
2424 | static int | |
2425 | ztest_create(ztest_ds_t *zd, ztest_od_t *od, int count) | |
2426 | { | |
2427 | int missing = 0; | |
d6320ddb | 2428 | int i; |
428870ff | 2429 | |
c25b8f99 | 2430 | ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock)); |
428870ff | 2431 | |
d6320ddb | 2432 | for (i = 0; i < count; i++, od++) { |
428870ff BB |
2433 | if (missing) { |
2434 | od->od_object = 0; | |
2435 | missing++; | |
2436 | continue; | |
2437 | } | |
2438 | ||
2439 | lr_create_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name); | |
2440 | ||
2441 | lr->lr_doid = od->od_dir; | |
2442 | lr->lr_foid = 0; /* 0 to allocate, > 0 to claim */ | |
2443 | lr->lrz_type = od->od_crtype; | |
2444 | lr->lrz_blocksize = od->od_crblocksize; | |
2445 | lr->lrz_ibshift = ztest_random_ibshift(); | |
2446 | lr->lrz_bonustype = DMU_OT_UINT64_OTHER; | |
50c957f7 | 2447 | lr->lrz_dnodesize = od->od_crdnodesize; |
428870ff BB |
2448 | lr->lr_gen = od->od_crgen; |
2449 | lr->lr_crtime[0] = time(NULL); | |
2450 | ||
2451 | if (ztest_replay_create(zd, lr, B_FALSE) != 0) { | |
2452 | ASSERT(missing == 0); | |
2453 | od->od_object = 0; | |
2454 | missing++; | |
2455 | } else { | |
2456 | od->od_object = lr->lr_foid; | |
2457 | od->od_type = od->od_crtype; | |
2458 | od->od_blocksize = od->od_crblocksize; | |
2459 | od->od_gen = od->od_crgen; | |
2460 | ASSERT(od->od_object != 0); | |
2461 | } | |
2462 | ||
2463 | ztest_lr_free(lr, sizeof (*lr), od->od_name); | |
2464 | } | |
2465 | ||
2466 | return (missing); | |
2467 | } | |
2468 | ||
2469 | static int | |
2470 | ztest_remove(ztest_ds_t *zd, ztest_od_t *od, int count) | |
2471 | { | |
2472 | int missing = 0; | |
2473 | int error; | |
d6320ddb | 2474 | int i; |
428870ff | 2475 | |
c25b8f99 | 2476 | ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock)); |
428870ff BB |
2477 | |
2478 | od += count - 1; | |
2479 | ||
d6320ddb | 2480 | for (i = count - 1; i >= 0; i--, od--) { |
428870ff BB |
2481 | if (missing) { |
2482 | missing++; | |
2483 | continue; | |
2484 | } | |
2485 | ||
03c6040b GW |
2486 | /* |
2487 | * No object was found. | |
2488 | */ | |
428870ff BB |
2489 | if (od->od_object == 0) |
2490 | continue; | |
2491 | ||
2492 | lr_remove_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name); | |
2493 | ||
2494 | lr->lr_doid = od->od_dir; | |
2495 | ||
2496 | if ((error = ztest_replay_remove(zd, lr, B_FALSE)) != 0) { | |
2497 | ASSERT3U(error, ==, ENOSPC); | |
2498 | missing++; | |
2499 | } else { | |
2500 | od->od_object = 0; | |
2501 | } | |
2502 | ztest_lr_free(lr, sizeof (*lr), od->od_name); | |
2503 | } | |
2504 | ||
2505 | return (missing); | |
2506 | } | |
2507 | ||
2508 | static int | |
2509 | ztest_write(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size, | |
2510 | void *data) | |
2511 | { | |
2512 | lr_write_t *lr; | |
2513 | int error; | |
2514 | ||
2515 | lr = ztest_lr_alloc(sizeof (*lr) + size, NULL); | |
2516 | ||
2517 | lr->lr_foid = object; | |
2518 | lr->lr_offset = offset; | |
2519 | lr->lr_length = size; | |
2520 | lr->lr_blkoff = 0; | |
2521 | BP_ZERO(&lr->lr_blkptr); | |
2522 | ||
2523 | bcopy(data, lr + 1, size); | |
2524 | ||
2525 | error = ztest_replay_write(zd, lr, B_FALSE); | |
2526 | ||
2527 | ztest_lr_free(lr, sizeof (*lr) + size, NULL); | |
2528 | ||
2529 | return (error); | |
2530 | } | |
2531 | ||
2532 | static int | |
2533 | ztest_truncate(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size) | |
2534 | { | |
2535 | lr_truncate_t *lr; | |
2536 | int error; | |
2537 | ||
2538 | lr = ztest_lr_alloc(sizeof (*lr), NULL); | |
2539 | ||
2540 | lr->lr_foid = object; | |
2541 | lr->lr_offset = offset; | |
2542 | lr->lr_length = size; | |
2543 | ||
2544 | error = ztest_replay_truncate(zd, lr, B_FALSE); | |
2545 | ||
2546 | ztest_lr_free(lr, sizeof (*lr), NULL); | |
2547 | ||
2548 | return (error); | |
2549 | } | |
2550 | ||
2551 | static int | |
2552 | ztest_setattr(ztest_ds_t *zd, uint64_t object) | |
2553 | { | |
2554 | lr_setattr_t *lr; | |
2555 | int error; | |
2556 | ||
2557 | lr = ztest_lr_alloc(sizeof (*lr), NULL); | |
2558 | ||
2559 | lr->lr_foid = object; | |
2560 | lr->lr_size = 0; | |
2561 | lr->lr_mode = 0; | |
2562 | ||
2563 | error = ztest_replay_setattr(zd, lr, B_FALSE); | |
2564 | ||
2565 | ztest_lr_free(lr, sizeof (*lr), NULL); | |
2566 | ||
2567 | return (error); | |
2568 | } | |
2569 | ||
2570 | static void | |
2571 | ztest_prealloc(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size) | |
2572 | { | |
2573 | objset_t *os = zd->zd_os; | |
2574 | dmu_tx_t *tx; | |
2575 | uint64_t txg; | |
5d43cc9a | 2576 | rl_t *rl; |
428870ff BB |
2577 | |
2578 | txg_wait_synced(dmu_objset_pool(os), 0); | |
2579 | ||
2580 | ztest_object_lock(zd, object, RL_READER); | |
2581 | rl = ztest_range_lock(zd, object, offset, size, RL_WRITER); | |
2582 | ||
2583 | tx = dmu_tx_create(os); | |
2584 | ||
2585 | dmu_tx_hold_write(tx, object, offset, size); | |
2586 | ||
2587 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
2588 | ||
2589 | if (txg != 0) { | |
2590 | dmu_prealloc(os, object, offset, size, tx); | |
2591 | dmu_tx_commit(tx); | |
2592 | txg_wait_synced(dmu_objset_pool(os), txg); | |
2593 | } else { | |
2594 | (void) dmu_free_long_range(os, object, offset, size); | |
2595 | } | |
2596 | ||
5d43cc9a | 2597 | ztest_range_unlock(rl); |
428870ff BB |
2598 | ztest_object_unlock(zd, object); |
2599 | } | |
2600 | ||
2601 | static void | |
2602 | ztest_io(ztest_ds_t *zd, uint64_t object, uint64_t offset) | |
2603 | { | |
03c6040b | 2604 | int err; |
428870ff BB |
2605 | ztest_block_tag_t wbt; |
2606 | dmu_object_info_t doi; | |
2607 | enum ztest_io_type io_type; | |
2608 | uint64_t blocksize; | |
2609 | void *data; | |
2610 | ||
2611 | VERIFY(dmu_object_info(zd->zd_os, object, &doi) == 0); | |
2612 | blocksize = doi.doi_data_block_size; | |
2613 | data = umem_alloc(blocksize, UMEM_NOFAIL); | |
2614 | ||
2615 | /* | |
2616 | * Pick an i/o type at random, biased toward writing block tags. | |
2617 | */ | |
2618 | io_type = ztest_random(ZTEST_IO_TYPES); | |
2619 | if (ztest_random(2) == 0) | |
2620 | io_type = ZTEST_IO_WRITE_TAG; | |
2621 | ||
2d9142c9 | 2622 | (void) pthread_rwlock_rdlock(&zd->zd_zilog_lock); |
3e31d2b0 | 2623 | |
428870ff BB |
2624 | switch (io_type) { |
2625 | ||
2626 | case ZTEST_IO_WRITE_TAG: | |
50c957f7 NB |
2627 | ztest_bt_generate(&wbt, zd->zd_os, object, doi.doi_dnodesize, |
2628 | offset, 0, 0, 0); | |
428870ff BB |
2629 | (void) ztest_write(zd, object, offset, sizeof (wbt), &wbt); |
2630 | break; | |
2631 | ||
2632 | case ZTEST_IO_WRITE_PATTERN: | |
2633 | (void) memset(data, 'a' + (object + offset) % 5, blocksize); | |
2634 | if (ztest_random(2) == 0) { | |
2635 | /* | |
2636 | * Induce fletcher2 collisions to ensure that | |
2637 | * zio_ddt_collision() detects and resolves them | |
2638 | * when using fletcher2-verify for deduplication. | |
2639 | */ | |
2640 | ((uint64_t *)data)[0] ^= 1ULL << 63; | |
2641 | ((uint64_t *)data)[4] ^= 1ULL << 63; | |
2642 | } | |
2643 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2644 | break; | |
2645 | ||
2646 | case ZTEST_IO_WRITE_ZEROES: | |
2647 | bzero(data, blocksize); | |
2648 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2649 | break; | |
2650 | ||
2651 | case ZTEST_IO_TRUNCATE: | |
2652 | (void) ztest_truncate(zd, object, offset, blocksize); | |
2653 | break; | |
2654 | ||
2655 | case ZTEST_IO_SETATTR: | |
2656 | (void) ztest_setattr(zd, object); | |
2657 | break; | |
e75c13c3 BB |
2658 | default: |
2659 | break; | |
03c6040b GW |
2660 | |
2661 | case ZTEST_IO_REWRITE: | |
2d9142c9 | 2662 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
03c6040b GW |
2663 | err = ztest_dsl_prop_set_uint64(zd->zd_name, |
2664 | ZFS_PROP_CHECKSUM, spa_dedup_checksum(ztest_spa), | |
2665 | B_FALSE); | |
2666 | VERIFY(err == 0 || err == ENOSPC); | |
2667 | err = ztest_dsl_prop_set_uint64(zd->zd_name, | |
2668 | ZFS_PROP_COMPRESSION, | |
2669 | ztest_random_dsl_prop(ZFS_PROP_COMPRESSION), | |
2670 | B_FALSE); | |
2671 | VERIFY(err == 0 || err == ENOSPC); | |
2d9142c9 | 2672 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
03c6040b GW |
2673 | |
2674 | VERIFY0(dmu_read(zd->zd_os, object, offset, blocksize, data, | |
2675 | DMU_READ_NO_PREFETCH)); | |
2676 | ||
2677 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2678 | break; | |
428870ff BB |
2679 | } |
2680 | ||
2d9142c9 | 2681 | (void) pthread_rwlock_unlock(&zd->zd_zilog_lock); |
3e31d2b0 | 2682 | |
428870ff BB |
2683 | umem_free(data, blocksize); |
2684 | } | |
2685 | ||
2686 | /* | |
2687 | * Initialize an object description template. | |
2688 | */ | |
2689 | static void | |
2690 | ztest_od_init(ztest_od_t *od, uint64_t id, char *tag, uint64_t index, | |
50c957f7 NB |
2691 | dmu_object_type_t type, uint64_t blocksize, uint64_t dnodesize, |
2692 | uint64_t gen) | |
428870ff BB |
2693 | { |
2694 | od->od_dir = ZTEST_DIROBJ; | |
2695 | od->od_object = 0; | |
2696 | ||
2697 | od->od_crtype = type; | |
2698 | od->od_crblocksize = blocksize ? blocksize : ztest_random_blocksize(); | |
50c957f7 | 2699 | od->od_crdnodesize = dnodesize ? dnodesize : ztest_random_dnodesize(); |
428870ff BB |
2700 | od->od_crgen = gen; |
2701 | ||
2702 | od->od_type = DMU_OT_NONE; | |
2703 | od->od_blocksize = 0; | |
2704 | od->od_gen = 0; | |
2705 | ||
2706 | (void) snprintf(od->od_name, sizeof (od->od_name), "%s(%lld)[%llu]", | |
b8864a23 | 2707 | tag, (longlong_t)id, (u_longlong_t)index); |
428870ff BB |
2708 | } |
2709 | ||
2710 | /* | |
2711 | * Lookup or create the objects for a test using the od template. | |
2712 | * If the objects do not all exist, or if 'remove' is specified, | |
2713 | * remove any existing objects and create new ones. Otherwise, | |
2714 | * use the existing objects. | |
2715 | */ | |
2716 | static int | |
2717 | ztest_object_init(ztest_ds_t *zd, ztest_od_t *od, size_t size, boolean_t remove) | |
2718 | { | |
2719 | int count = size / sizeof (*od); | |
2720 | int rv = 0; | |
2721 | ||
1e33ac1e | 2722 | mutex_enter(&zd->zd_dirobj_lock); |
428870ff BB |
2723 | if ((ztest_lookup(zd, od, count) != 0 || remove) && |
2724 | (ztest_remove(zd, od, count) != 0 || | |
2725 | ztest_create(zd, od, count) != 0)) | |
2726 | rv = -1; | |
2727 | zd->zd_od = od; | |
1e33ac1e | 2728 | mutex_exit(&zd->zd_dirobj_lock); |
428870ff BB |
2729 | |
2730 | return (rv); | |
2731 | } | |
2732 | ||
2733 | /* ARGSUSED */ | |
2734 | void | |
2735 | ztest_zil_commit(ztest_ds_t *zd, uint64_t id) | |
2736 | { | |
2737 | zilog_t *zilog = zd->zd_zilog; | |
2738 | ||
2d9142c9 | 2739 | (void) pthread_rwlock_rdlock(&zd->zd_zilog_lock); |
3e31d2b0 | 2740 | |
572e2857 | 2741 | zil_commit(zilog, ztest_random(ZTEST_OBJECTS)); |
428870ff BB |
2742 | |
2743 | /* | |
2744 | * Remember the committed values in zd, which is in parent/child | |
2745 | * shared memory. If we die, the next iteration of ztest_run() | |
2746 | * will verify that the log really does contain this record. | |
2747 | */ | |
2748 | mutex_enter(&zilog->zl_lock); | |
c242c188 CS |
2749 | ASSERT(zd->zd_shared != NULL); |
2750 | ASSERT3U(zd->zd_shared->zd_seq, <=, zilog->zl_commit_lr_seq); | |
2751 | zd->zd_shared->zd_seq = zilog->zl_commit_lr_seq; | |
428870ff | 2752 | mutex_exit(&zilog->zl_lock); |
3e31d2b0 | 2753 | |
2d9142c9 | 2754 | (void) pthread_rwlock_unlock(&zd->zd_zilog_lock); |
3e31d2b0 ES |
2755 | } |
2756 | ||
2757 | /* | |
2758 | * This function is designed to simulate the operations that occur during a | |
2759 | * mount/unmount operation. We hold the dataset across these operations in an | |
2760 | * attempt to expose any implicit assumptions about ZIL management. | |
2761 | */ | |
2762 | /* ARGSUSED */ | |
2763 | void | |
2764 | ztest_zil_remount(ztest_ds_t *zd, uint64_t id) | |
2765 | { | |
2766 | objset_t *os = zd->zd_os; | |
2767 | ||
0b606cb3 TC |
2768 | /* |
2769 | * We hold the ztest_vdev_lock so we don't cause problems with | |
2770 | * other threads that wish to remove a log device, such as | |
2771 | * ztest_device_removal(). | |
2772 | */ | |
2773 | mutex_enter(&ztest_vdev_lock); | |
2774 | ||
03c6040b GW |
2775 | /* |
2776 | * We grab the zd_dirobj_lock to ensure that no other thread is | |
2777 | * updating the zil (i.e. adding in-memory log records) and the | |
2778 | * zd_zilog_lock to block any I/O. | |
2779 | */ | |
29809a6c | 2780 | mutex_enter(&zd->zd_dirobj_lock); |
2d9142c9 | 2781 | (void) pthread_rwlock_wrlock(&zd->zd_zilog_lock); |
3e31d2b0 | 2782 | |
f298b24d | 2783 | /* zfsvfs_teardown() */ |
3e31d2b0 ES |
2784 | zil_close(zd->zd_zilog); |
2785 | ||
2786 | /* zfsvfs_setup() */ | |
2787 | VERIFY(zil_open(os, ztest_get_data) == zd->zd_zilog); | |
2788 | zil_replay(os, zd, ztest_replay_vector); | |
2789 | ||
2d9142c9 | 2790 | (void) pthread_rwlock_unlock(&zd->zd_zilog_lock); |
29809a6c | 2791 | mutex_exit(&zd->zd_dirobj_lock); |
0b606cb3 | 2792 | mutex_exit(&ztest_vdev_lock); |
428870ff BB |
2793 | } |
2794 | ||
2795 | /* | |
2796 | * Verify that we can't destroy an active pool, create an existing pool, | |
2797 | * or create a pool with a bad vdev spec. | |
2798 | */ | |
2799 | /* ARGSUSED */ | |
2800 | void | |
2801 | ztest_spa_create_destroy(ztest_ds_t *zd, uint64_t id) | |
2802 | { | |
c242c188 | 2803 | ztest_shared_opts_t *zo = &ztest_opts; |
428870ff BB |
2804 | spa_t *spa; |
2805 | nvlist_t *nvroot; | |
2806 | ||
379ca9cf OF |
2807 | if (zo->zo_mmp_test) |
2808 | return; | |
2809 | ||
428870ff BB |
2810 | /* |
2811 | * Attempt to create using a bad file. | |
2812 | */ | |
cc99f275 | 2813 | nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, NULL, 0, 0, 1); |
428870ff | 2814 | VERIFY3U(ENOENT, ==, |
b5256303 | 2815 | spa_create("ztest_bad_file", nvroot, NULL, NULL, NULL)); |
428870ff BB |
2816 | nvlist_free(nvroot); |
2817 | ||
2818 | /* | |
2819 | * Attempt to create using a bad mirror. | |
2820 | */ | |
cc99f275 | 2821 | nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, NULL, 0, 2, 1); |
428870ff | 2822 | VERIFY3U(ENOENT, ==, |
b5256303 | 2823 | spa_create("ztest_bad_mirror", nvroot, NULL, NULL, NULL)); |
428870ff BB |
2824 | nvlist_free(nvroot); |
2825 | ||
2826 | /* | |
2827 | * Attempt to create an existing pool. It shouldn't matter | |
2828 | * what's in the nvroot; we should fail with EEXIST. | |
2829 | */ | |
2d9142c9 | 2830 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
cc99f275 | 2831 | nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, NULL, 0, 0, 1); |
b5256303 TC |
2832 | VERIFY3U(EEXIST, ==, |
2833 | spa_create(zo->zo_pool, nvroot, NULL, NULL, NULL)); | |
428870ff | 2834 | nvlist_free(nvroot); |
43a85362 SD |
2835 | |
2836 | /* | |
2837 | * We open a reference to the spa and then we try to export it | |
2838 | * expecting one of the following errors: | |
2839 | * | |
2840 | * EBUSY | |
2841 | * Because of the reference we just opened. | |
2842 | * | |
2843 | * ZFS_ERR_EXPORT_IN_PROGRESS | |
2844 | * For the case that there is another ztest thread doing | |
2845 | * an export concurrently. | |
2846 | */ | |
c242c188 | 2847 | VERIFY3U(0, ==, spa_open(zo->zo_pool, &spa, FTAG)); |
43a85362 SD |
2848 | int error = spa_destroy(zo->zo_pool); |
2849 | if (error != EBUSY && error != ZFS_ERR_EXPORT_IN_PROGRESS) { | |
2850 | fatal(0, "spa_destroy(%s) returned unexpected value %d", | |
2851 | spa->spa_name, error); | |
2852 | } | |
428870ff BB |
2853 | spa_close(spa, FTAG); |
2854 | ||
2d9142c9 | 2855 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
428870ff BB |
2856 | } |
2857 | ||
0582e403 OF |
2858 | /* |
2859 | * Start and then stop the MMP threads to ensure the startup and shutdown code | |
2860 | * works properly. Actual protection and property-related code tested via ZTS. | |
2861 | */ | |
2862 | /* ARGSUSED */ | |
2863 | void | |
2864 | ztest_mmp_enable_disable(ztest_ds_t *zd, uint64_t id) | |
2865 | { | |
2866 | ztest_shared_opts_t *zo = &ztest_opts; | |
2867 | spa_t *spa = ztest_spa; | |
2868 | ||
2869 | if (zo->zo_mmp_test) | |
2870 | return; | |
2871 | ||
8133679f OF |
2872 | /* |
2873 | * Since enabling MMP involves setting a property, it could not be done | |
2874 | * while the pool is suspended. | |
2875 | */ | |
2876 | if (spa_suspended(spa)) | |
2877 | return; | |
2878 | ||
0582e403 OF |
2879 | spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); |
2880 | mutex_enter(&spa->spa_props_lock); | |
2881 | ||
52002377 OF |
2882 | zfs_multihost_fail_intervals = 0; |
2883 | ||
0582e403 OF |
2884 | if (!spa_multihost(spa)) { |
2885 | spa->spa_multihost = B_TRUE; | |
2886 | mmp_thread_start(spa); | |
2887 | } | |
2888 | ||
2889 | mutex_exit(&spa->spa_props_lock); | |
2890 | spa_config_exit(spa, SCL_CONFIG, FTAG); | |
2891 | ||
2892 | txg_wait_synced(spa_get_dsl(spa), 0); | |
2893 | mmp_signal_all_threads(); | |
2894 | txg_wait_synced(spa_get_dsl(spa), 0); | |
2895 | ||
2896 | spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); | |
2897 | mutex_enter(&spa->spa_props_lock); | |
2898 | ||
2899 | if (spa_multihost(spa)) { | |
2900 | mmp_thread_stop(spa); | |
2901 | spa->spa_multihost = B_FALSE; | |
2902 | } | |
2903 | ||
2904 | mutex_exit(&spa->spa_props_lock); | |
2905 | spa_config_exit(spa, SCL_CONFIG, FTAG); | |
2906 | } | |
2907 | ||
ea0b2538 GW |
2908 | /* ARGSUSED */ |
2909 | void | |
2910 | ztest_spa_upgrade(ztest_ds_t *zd, uint64_t id) | |
2911 | { | |
2912 | spa_t *spa; | |
2913 | uint64_t initial_version = SPA_VERSION_INITIAL; | |
2914 | uint64_t version, newversion; | |
2915 | nvlist_t *nvroot, *props; | |
2916 | char *name; | |
2917 | ||
379ca9cf OF |
2918 | if (ztest_opts.zo_mmp_test) |
2919 | return; | |
2920 | ||
b2255edc BB |
2921 | /* dRAID added after feature flags, skip upgrade test. */ |
2922 | if (strcmp(ztest_opts.zo_raid_type, VDEV_TYPE_DRAID) == 0) | |
2923 | return; | |
2924 | ||
ea0b2538 GW |
2925 | mutex_enter(&ztest_vdev_lock); |
2926 | name = kmem_asprintf("%s_upgrade", ztest_opts.zo_pool); | |
2927 | ||
2928 | /* | |
2929 | * Clean up from previous runs. | |
2930 | */ | |
2931 | (void) spa_destroy(name); | |
2932 | ||
2933 | nvroot = make_vdev_root(NULL, NULL, name, ztest_opts.zo_vdev_size, 0, | |
b2255edc | 2934 | NULL, ztest_opts.zo_raid_children, ztest_opts.zo_mirrors, 1); |
ea0b2538 GW |
2935 | |
2936 | /* | |
2937 | * If we're configuring a RAIDZ device then make sure that the | |
d611989f | 2938 | * initial version is capable of supporting that feature. |
ea0b2538 | 2939 | */ |
b2255edc | 2940 | switch (ztest_opts.zo_raid_parity) { |
ea0b2538 GW |
2941 | case 0: |
2942 | case 1: | |
2943 | initial_version = SPA_VERSION_INITIAL; | |
2944 | break; | |
2945 | case 2: | |
2946 | initial_version = SPA_VERSION_RAIDZ2; | |
2947 | break; | |
2948 | case 3: | |
2949 | initial_version = SPA_VERSION_RAIDZ3; | |
2950 | break; | |
2951 | } | |
2952 | ||
2953 | /* | |
2954 | * Create a pool with a spa version that can be upgraded. Pick | |
2955 | * a value between initial_version and SPA_VERSION_BEFORE_FEATURES. | |
2956 | */ | |
2957 | do { | |
2958 | version = ztest_random_spa_version(initial_version); | |
2959 | } while (version > SPA_VERSION_BEFORE_FEATURES); | |
2960 | ||
2961 | props = fnvlist_alloc(); | |
2962 | fnvlist_add_uint64(props, | |
2963 | zpool_prop_to_name(ZPOOL_PROP_VERSION), version); | |
b5256303 | 2964 | VERIFY3S(spa_create(name, nvroot, props, NULL, NULL), ==, 0); |
ea0b2538 GW |
2965 | fnvlist_free(nvroot); |
2966 | fnvlist_free(props); | |
2967 | ||
2968 | VERIFY3S(spa_open(name, &spa, FTAG), ==, 0); | |
2969 | VERIFY3U(spa_version(spa), ==, version); | |
2970 | newversion = ztest_random_spa_version(version + 1); | |
2971 | ||
2972 | if (ztest_opts.zo_verbose >= 4) { | |
2973 | (void) printf("upgrading spa version from %llu to %llu\n", | |
2974 | (u_longlong_t)version, (u_longlong_t)newversion); | |
2975 | } | |
2976 | ||
2977 | spa_upgrade(spa, newversion); | |
2978 | VERIFY3U(spa_version(spa), >, version); | |
2979 | VERIFY3U(spa_version(spa), ==, fnvlist_lookup_uint64(spa->spa_config, | |
2980 | zpool_prop_to_name(ZPOOL_PROP_VERSION))); | |
2981 | spa_close(spa, FTAG); | |
2982 | ||
e4f5fa12 | 2983 | kmem_strfree(name); |
ea0b2538 GW |
2984 | mutex_exit(&ztest_vdev_lock); |
2985 | } | |
2986 | ||
d2734cce SD |
2987 | static void |
2988 | ztest_spa_checkpoint(spa_t *spa) | |
2989 | { | |
2990 | ASSERT(MUTEX_HELD(&ztest_checkpoint_lock)); | |
2991 | ||
2992 | int error = spa_checkpoint(spa->spa_name); | |
2993 | ||
2994 | switch (error) { | |
2995 | case 0: | |
2996 | case ZFS_ERR_DEVRM_IN_PROGRESS: | |
2997 | case ZFS_ERR_DISCARDING_CHECKPOINT: | |
2998 | case ZFS_ERR_CHECKPOINT_EXISTS: | |
2999 | break; | |
3000 | case ENOSPC: | |
3001 | ztest_record_enospc(FTAG); | |
3002 | break; | |
3003 | default: | |
3004 | fatal(0, "spa_checkpoint(%s) = %d", spa->spa_name, error); | |
3005 | } | |
3006 | } | |
3007 | ||
3008 | static void | |
3009 | ztest_spa_discard_checkpoint(spa_t *spa) | |
3010 | { | |
3011 | ASSERT(MUTEX_HELD(&ztest_checkpoint_lock)); | |
3012 | ||
3013 | int error = spa_checkpoint_discard(spa->spa_name); | |
3014 | ||
3015 | switch (error) { | |
3016 | case 0: | |
3017 | case ZFS_ERR_DISCARDING_CHECKPOINT: | |
3018 | case ZFS_ERR_NO_CHECKPOINT: | |
3019 | break; | |
3020 | default: | |
3021 | fatal(0, "spa_discard_checkpoint(%s) = %d", | |
3022 | spa->spa_name, error); | |
3023 | } | |
3024 | ||
3025 | } | |
3026 | ||
3027 | /* ARGSUSED */ | |
3028 | void | |
3029 | ztest_spa_checkpoint_create_discard(ztest_ds_t *zd, uint64_t id) | |
3030 | { | |
3031 | spa_t *spa = ztest_spa; | |
3032 | ||
3033 | mutex_enter(&ztest_checkpoint_lock); | |
3034 | if (ztest_random(2) == 0) { | |
3035 | ztest_spa_checkpoint(spa); | |
3036 | } else { | |
3037 | ztest_spa_discard_checkpoint(spa); | |
3038 | } | |
3039 | mutex_exit(&ztest_checkpoint_lock); | |
3040 | } | |
3041 | ||
3042 | ||
428870ff BB |
3043 | static vdev_t * |
3044 | vdev_lookup_by_path(vdev_t *vd, const char *path) | |
3045 | { | |
3046 | vdev_t *mvd; | |
d6320ddb | 3047 | int c; |
428870ff BB |
3048 | |
3049 | if (vd->vdev_path != NULL && strcmp(path, vd->vdev_path) == 0) | |
3050 | return (vd); | |
3051 | ||
d6320ddb | 3052 | for (c = 0; c < vd->vdev_children; c++) |
428870ff BB |
3053 | if ((mvd = vdev_lookup_by_path(vd->vdev_child[c], path)) != |
3054 | NULL) | |
3055 | return (mvd); | |
3056 | ||
3057 | return (NULL); | |
3058 | } | |
3059 | ||
93e28d66 SD |
3060 | static int |
3061 | spa_num_top_vdevs(spa_t *spa) | |
428870ff BB |
3062 | { |
3063 | vdev_t *rvd = spa->spa_root_vdev; | |
93e28d66 SD |
3064 | ASSERT3U(spa_config_held(spa, SCL_VDEV, RW_READER), ==, SCL_VDEV); |
3065 | return (rvd->vdev_children); | |
428870ff BB |
3066 | } |
3067 | ||
3068 | /* | |
3069 | * Verify that vdev_add() works as expected. | |
3070 | */ | |
3071 | /* ARGSUSED */ | |
3072 | void | |
3073 | ztest_vdev_add_remove(ztest_ds_t *zd, uint64_t id) | |
3074 | { | |
3075 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 3076 | spa_t *spa = ztest_spa; |
428870ff BB |
3077 | uint64_t leaves; |
3078 | uint64_t guid; | |
3079 | nvlist_t *nvroot; | |
3080 | int error; | |
3081 | ||
379ca9cf OF |
3082 | if (ztest_opts.zo_mmp_test) |
3083 | return; | |
3084 | ||
c242c188 | 3085 | mutex_enter(&ztest_vdev_lock); |
b2255edc BB |
3086 | leaves = MAX(zs->zs_mirrors + zs->zs_splits, 1) * |
3087 | ztest_opts.zo_raid_children; | |
428870ff BB |
3088 | |
3089 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
3090 | ||
93e28d66 | 3091 | ztest_shared->zs_vdev_next_leaf = spa_num_top_vdevs(spa) * leaves; |
428870ff BB |
3092 | |
3093 | /* | |
3094 | * If we have slogs then remove them 1/4 of the time. | |
3095 | */ | |
3096 | if (spa_has_slogs(spa) && ztest_random(4) == 0) { | |
cc99f275 DB |
3097 | metaslab_group_t *mg; |
3098 | ||
428870ff | 3099 | /* |
cc99f275 | 3100 | * find the first real slog in log allocation class |
428870ff | 3101 | */ |
cc99f275 DB |
3102 | mg = spa_log_class(spa)->mc_rotor; |
3103 | while (!mg->mg_vd->vdev_islog) | |
3104 | mg = mg->mg_next; | |
3105 | ||
3106 | guid = mg->mg_vd->vdev_guid; | |
428870ff BB |
3107 | |
3108 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
3109 | ||
3110 | /* | |
3111 | * We have to grab the zs_name_lock as writer to | |
3112 | * prevent a race between removing a slog (dmu_objset_find) | |
3113 | * and destroying a dataset. Removing the slog will | |
3114 | * grab a reference on the dataset which may cause | |
13fe0198 | 3115 | * dsl_destroy_head() to fail with EBUSY thus |
428870ff BB |
3116 | * leaving the dataset in an inconsistent state. |
3117 | */ | |
2d9142c9 | 3118 | pthread_rwlock_wrlock(&ztest_name_lock); |
428870ff | 3119 | error = spa_vdev_remove(spa, guid, B_FALSE); |
2d9142c9 | 3120 | pthread_rwlock_unlock(&ztest_name_lock); |
428870ff | 3121 | |
c91cf36f BB |
3122 | switch (error) { |
3123 | case 0: | |
3124 | case EEXIST: /* Generic zil_reset() error */ | |
3125 | case EBUSY: /* Replay required */ | |
3126 | case EACCES: /* Crypto key not loaded */ | |
d2734cce SD |
3127 | case ZFS_ERR_CHECKPOINT_EXISTS: |
3128 | case ZFS_ERR_DISCARDING_CHECKPOINT: | |
c91cf36f BB |
3129 | break; |
3130 | default: | |
428870ff | 3131 | fatal(0, "spa_vdev_remove() = %d", error); |
c91cf36f | 3132 | } |
428870ff BB |
3133 | } else { |
3134 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
3135 | ||
3136 | /* | |
cc99f275 | 3137 | * Make 1/4 of the devices be log devices |
428870ff | 3138 | */ |
ea0b2538 | 3139 | nvroot = make_vdev_root(NULL, NULL, NULL, |
cc99f275 | 3140 | ztest_opts.zo_vdev_size, 0, (ztest_random(4) == 0) ? |
b2255edc BB |
3141 | "log" : NULL, ztest_opts.zo_raid_children, zs->zs_mirrors, |
3142 | 1); | |
428870ff BB |
3143 | |
3144 | error = spa_vdev_add(spa, nvroot); | |
3145 | nvlist_free(nvroot); | |
3146 | ||
d2734cce SD |
3147 | switch (error) { |
3148 | case 0: | |
3149 | break; | |
3150 | case ENOSPC: | |
428870ff | 3151 | ztest_record_enospc("spa_vdev_add"); |
d2734cce SD |
3152 | break; |
3153 | default: | |
428870ff | 3154 | fatal(0, "spa_vdev_add() = %d", error); |
d2734cce | 3155 | } |
428870ff BB |
3156 | } |
3157 | ||
c242c188 | 3158 | mutex_exit(&ztest_vdev_lock); |
428870ff BB |
3159 | } |
3160 | ||
cc99f275 DB |
3161 | /* ARGSUSED */ |
3162 | void | |
3163 | ztest_vdev_class_add(ztest_ds_t *zd, uint64_t id) | |
3164 | { | |
3165 | ztest_shared_t *zs = ztest_shared; | |
3166 | spa_t *spa = ztest_spa; | |
3167 | uint64_t leaves; | |
3168 | nvlist_t *nvroot; | |
3169 | const char *class = (ztest_random(2) == 0) ? | |
3170 | VDEV_ALLOC_BIAS_SPECIAL : VDEV_ALLOC_BIAS_DEDUP; | |
3171 | int error; | |
3172 | ||
3173 | /* | |
3174 | * By default add a special vdev 50% of the time | |
3175 | */ | |
3176 | if ((ztest_opts.zo_special_vdevs == ZTEST_VDEV_CLASS_OFF) || | |
3177 | (ztest_opts.zo_special_vdevs == ZTEST_VDEV_CLASS_RND && | |
3178 | ztest_random(2) == 0)) { | |
3179 | return; | |
3180 | } | |
3181 | ||
3182 | mutex_enter(&ztest_vdev_lock); | |
3183 | ||
3184 | /* Only test with mirrors */ | |
3185 | if (zs->zs_mirrors < 2) { | |
3186 | mutex_exit(&ztest_vdev_lock); | |
3187 | return; | |
3188 | } | |
3189 | ||
3190 | /* requires feature@allocation_classes */ | |
3191 | if (!spa_feature_is_enabled(spa, SPA_FEATURE_ALLOCATION_CLASSES)) { | |
3192 | mutex_exit(&ztest_vdev_lock); | |
3193 | return; | |
3194 | } | |
3195 | ||
b2255edc BB |
3196 | leaves = MAX(zs->zs_mirrors + zs->zs_splits, 1) * |
3197 | ztest_opts.zo_raid_children; | |
cc99f275 DB |
3198 | |
3199 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
93e28d66 | 3200 | ztest_shared->zs_vdev_next_leaf = spa_num_top_vdevs(spa) * leaves; |
cc99f275 DB |
3201 | spa_config_exit(spa, SCL_VDEV, FTAG); |
3202 | ||
3203 | nvroot = make_vdev_root(NULL, NULL, NULL, ztest_opts.zo_vdev_size, 0, | |
b2255edc | 3204 | class, ztest_opts.zo_raid_children, zs->zs_mirrors, 1); |
cc99f275 DB |
3205 | |
3206 | error = spa_vdev_add(spa, nvroot); | |
3207 | nvlist_free(nvroot); | |
3208 | ||
3209 | if (error == ENOSPC) | |
3210 | ztest_record_enospc("spa_vdev_add"); | |
3211 | else if (error != 0) | |
3212 | fatal(0, "spa_vdev_add() = %d", error); | |
3213 | ||
3214 | /* | |
3215 | * 50% of the time allow small blocks in the special class | |
3216 | */ | |
3217 | if (error == 0 && | |
3218 | spa_special_class(spa)->mc_groups == 1 && ztest_random(2) == 0) { | |
3219 | if (ztest_opts.zo_verbose >= 3) | |
3220 | (void) printf("Enabling special VDEV small blocks\n"); | |
3221 | (void) ztest_dsl_prop_set_uint64(zd->zd_name, | |
3222 | ZFS_PROP_SPECIAL_SMALL_BLOCKS, 32768, B_FALSE); | |
3223 | } | |
3224 | ||
3225 | mutex_exit(&ztest_vdev_lock); | |
3226 | ||
3227 | if (ztest_opts.zo_verbose >= 3) { | |
3228 | metaslab_class_t *mc; | |
3229 | ||
3230 | if (strcmp(class, VDEV_ALLOC_BIAS_SPECIAL) == 0) | |
3231 | mc = spa_special_class(spa); | |
3232 | else | |
3233 | mc = spa_dedup_class(spa); | |
3234 | (void) printf("Added a %s mirrored vdev (of %d)\n", | |
3235 | class, (int)mc->mc_groups); | |
3236 | } | |
3237 | } | |
3238 | ||
428870ff BB |
3239 | /* |
3240 | * Verify that adding/removing aux devices (l2arc, hot spare) works as expected. | |
3241 | */ | |
3242 | /* ARGSUSED */ | |
3243 | void | |
3244 | ztest_vdev_aux_add_remove(ztest_ds_t *zd, uint64_t id) | |
3245 | { | |
3246 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 3247 | spa_t *spa = ztest_spa; |
428870ff BB |
3248 | vdev_t *rvd = spa->spa_root_vdev; |
3249 | spa_aux_vdev_t *sav; | |
3250 | char *aux; | |
40b84e7a | 3251 | char *path; |
428870ff | 3252 | uint64_t guid = 0; |
b2255edc | 3253 | int error, ignore_err = 0; |
428870ff | 3254 | |
379ca9cf OF |
3255 | if (ztest_opts.zo_mmp_test) |
3256 | return; | |
3257 | ||
40b84e7a BB |
3258 | path = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
3259 | ||
428870ff BB |
3260 | if (ztest_random(2) == 0) { |
3261 | sav = &spa->spa_spares; | |
3262 | aux = ZPOOL_CONFIG_SPARES; | |
3263 | } else { | |
3264 | sav = &spa->spa_l2cache; | |
3265 | aux = ZPOOL_CONFIG_L2CACHE; | |
3266 | } | |
3267 | ||
c242c188 | 3268 | mutex_enter(&ztest_vdev_lock); |
428870ff BB |
3269 | |
3270 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
3271 | ||
3272 | if (sav->sav_count != 0 && ztest_random(4) == 0) { | |
b128c09f BB |
3273 | /* |
3274 | * Pick a random device to remove. | |
3275 | */ | |
b2255edc BB |
3276 | vdev_t *svd = sav->sav_vdevs[ztest_random(sav->sav_count)]; |
3277 | ||
3278 | /* dRAID spares cannot be removed; try anyways to see ENOTSUP */ | |
3279 | if (strstr(svd->vdev_path, VDEV_TYPE_DRAID) != NULL) | |
3280 | ignore_err = ENOTSUP; | |
3281 | ||
3282 | guid = svd->vdev_guid; | |
b128c09f BB |
3283 | } else { |
3284 | /* | |
3285 | * Find an unused device we can add. | |
3286 | */ | |
428870ff | 3287 | zs->zs_vdev_aux = 0; |
b128c09f | 3288 | for (;;) { |
b128c09f | 3289 | int c; |
3db3ff4a | 3290 | (void) snprintf(path, MAXPATHLEN, ztest_aux_template, |
c242c188 CS |
3291 | ztest_opts.zo_dir, ztest_opts.zo_pool, aux, |
3292 | zs->zs_vdev_aux); | |
b128c09f BB |
3293 | for (c = 0; c < sav->sav_count; c++) |
3294 | if (strcmp(sav->sav_vdevs[c]->vdev_path, | |
3295 | path) == 0) | |
3296 | break; | |
3297 | if (c == sav->sav_count && | |
3298 | vdev_lookup_by_path(rvd, path) == NULL) | |
3299 | break; | |
428870ff | 3300 | zs->zs_vdev_aux++; |
34dc7c2f BB |
3301 | } |
3302 | } | |
3303 | ||
b128c09f | 3304 | spa_config_exit(spa, SCL_VDEV, FTAG); |
34dc7c2f | 3305 | |
b128c09f BB |
3306 | if (guid == 0) { |
3307 | /* | |
3308 | * Add a new device. | |
3309 | */ | |
ea0b2538 | 3310 | nvlist_t *nvroot = make_vdev_root(NULL, aux, NULL, |
cc99f275 | 3311 | (ztest_opts.zo_vdev_size * 5) / 4, 0, NULL, 0, 0, 1); |
b128c09f | 3312 | error = spa_vdev_add(spa, nvroot); |
d2734cce SD |
3313 | |
3314 | switch (error) { | |
3315 | case 0: | |
3316 | break; | |
3317 | default: | |
b128c09f | 3318 | fatal(0, "spa_vdev_add(%p) = %d", nvroot, error); |
d2734cce | 3319 | } |
b128c09f BB |
3320 | nvlist_free(nvroot); |
3321 | } else { | |
3322 | /* | |
3323 | * Remove an existing device. Sometimes, dirty its | |
3324 | * vdev state first to make sure we handle removal | |
3325 | * of devices that have pending state changes. | |
3326 | */ | |
3327 | if (ztest_random(2) == 0) | |
9babb374 | 3328 | (void) vdev_online(spa, guid, 0, NULL); |
b128c09f BB |
3329 | |
3330 | error = spa_vdev_remove(spa, guid, B_FALSE); | |
d2734cce SD |
3331 | |
3332 | switch (error) { | |
3333 | case 0: | |
3334 | case EBUSY: | |
3335 | case ZFS_ERR_CHECKPOINT_EXISTS: | |
3336 | case ZFS_ERR_DISCARDING_CHECKPOINT: | |
3337 | break; | |
3338 | default: | |
b2255edc BB |
3339 | if (error != ignore_err) |
3340 | fatal(0, "spa_vdev_remove(%llu) = %d", guid, | |
3341 | error); | |
d2734cce | 3342 | } |
b128c09f BB |
3343 | } |
3344 | ||
c242c188 | 3345 | mutex_exit(&ztest_vdev_lock); |
40b84e7a BB |
3346 | |
3347 | umem_free(path, MAXPATHLEN); | |
428870ff BB |
3348 | } |
3349 | ||
3350 | /* | |
3351 | * split a pool if it has mirror tlvdevs | |
3352 | */ | |
3353 | /* ARGSUSED */ | |
3354 | void | |
3355 | ztest_split_pool(ztest_ds_t *zd, uint64_t id) | |
3356 | { | |
3357 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 3358 | spa_t *spa = ztest_spa; |
428870ff BB |
3359 | vdev_t *rvd = spa->spa_root_vdev; |
3360 | nvlist_t *tree, **child, *config, *split, **schild; | |
3361 | uint_t c, children, schildren = 0, lastlogid = 0; | |
3362 | int error = 0; | |
3363 | ||
379ca9cf OF |
3364 | if (ztest_opts.zo_mmp_test) |
3365 | return; | |
3366 | ||
c242c188 | 3367 | mutex_enter(&ztest_vdev_lock); |
428870ff | 3368 | |
d611989f | 3369 | /* ensure we have a usable config; mirrors of raidz aren't supported */ |
b2255edc | 3370 | if (zs->zs_mirrors < 3 || ztest_opts.zo_raid_children > 1) { |
c242c188 | 3371 | mutex_exit(&ztest_vdev_lock); |
428870ff BB |
3372 | return; |
3373 | } | |
3374 | ||
3375 | /* clean up the old pool, if any */ | |
3376 | (void) spa_destroy("splitp"); | |
3377 | ||
3378 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
3379 | ||
3380 | /* generate a config from the existing config */ | |
3381 | mutex_enter(&spa->spa_props_lock); | |
3382 | VERIFY(nvlist_lookup_nvlist(spa->spa_config, ZPOOL_CONFIG_VDEV_TREE, | |
3383 | &tree) == 0); | |
3384 | mutex_exit(&spa->spa_props_lock); | |
3385 | ||
3386 | VERIFY(nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child, | |
3387 | &children) == 0); | |
3388 | ||
3389 | schild = malloc(rvd->vdev_children * sizeof (nvlist_t *)); | |
3390 | for (c = 0; c < children; c++) { | |
3391 | vdev_t *tvd = rvd->vdev_child[c]; | |
3392 | nvlist_t **mchild; | |
3393 | uint_t mchildren; | |
3394 | ||
3395 | if (tvd->vdev_islog || tvd->vdev_ops == &vdev_hole_ops) { | |
3396 | VERIFY(nvlist_alloc(&schild[schildren], NV_UNIQUE_NAME, | |
3397 | 0) == 0); | |
3398 | VERIFY(nvlist_add_string(schild[schildren], | |
3399 | ZPOOL_CONFIG_TYPE, VDEV_TYPE_HOLE) == 0); | |
3400 | VERIFY(nvlist_add_uint64(schild[schildren], | |
3401 | ZPOOL_CONFIG_IS_HOLE, 1) == 0); | |
3402 | if (lastlogid == 0) | |
3403 | lastlogid = schildren; | |
3404 | ++schildren; | |
3405 | continue; | |
3406 | } | |
3407 | lastlogid = 0; | |
3408 | VERIFY(nvlist_lookup_nvlist_array(child[c], | |
3409 | ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0); | |
3410 | VERIFY(nvlist_dup(mchild[0], &schild[schildren++], 0) == 0); | |
3411 | } | |
3412 | ||
3413 | /* OK, create a config that can be used to split */ | |
3414 | VERIFY(nvlist_alloc(&split, NV_UNIQUE_NAME, 0) == 0); | |
3415 | VERIFY(nvlist_add_string(split, ZPOOL_CONFIG_TYPE, | |
3416 | VDEV_TYPE_ROOT) == 0); | |
3417 | VERIFY(nvlist_add_nvlist_array(split, ZPOOL_CONFIG_CHILDREN, schild, | |
3418 | lastlogid != 0 ? lastlogid : schildren) == 0); | |
3419 | ||
3420 | VERIFY(nvlist_alloc(&config, NV_UNIQUE_NAME, 0) == 0); | |
3421 | VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, split) == 0); | |
3422 | ||
3423 | for (c = 0; c < schildren; c++) | |
3424 | nvlist_free(schild[c]); | |
3425 | free(schild); | |
3426 | nvlist_free(split); | |
3427 | ||
3428 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
3429 | ||
2d9142c9 | 3430 | (void) pthread_rwlock_wrlock(&ztest_name_lock); |
428870ff | 3431 | error = spa_vdev_split_mirror(spa, "splitp", config, NULL, B_FALSE); |
2d9142c9 | 3432 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
428870ff BB |
3433 | |
3434 | nvlist_free(config); | |
3435 | ||
3436 | if (error == 0) { | |
3437 | (void) printf("successful split - results:\n"); | |
3438 | mutex_enter(&spa_namespace_lock); | |
3439 | show_pool_stats(spa); | |
3440 | show_pool_stats(spa_lookup("splitp")); | |
3441 | mutex_exit(&spa_namespace_lock); | |
3442 | ++zs->zs_splits; | |
3443 | --zs->zs_mirrors; | |
3444 | } | |
c242c188 | 3445 | mutex_exit(&ztest_vdev_lock); |
34dc7c2f BB |
3446 | } |
3447 | ||
3448 | /* | |
3449 | * Verify that we can attach and detach devices. | |
3450 | */ | |
428870ff | 3451 | /* ARGSUSED */ |
34dc7c2f | 3452 | void |
428870ff | 3453 | ztest_vdev_attach_detach(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3454 | { |
428870ff | 3455 | ztest_shared_t *zs = ztest_shared; |
c242c188 | 3456 | spa_t *spa = ztest_spa; |
b128c09f | 3457 | spa_aux_vdev_t *sav = &spa->spa_spares; |
34dc7c2f BB |
3458 | vdev_t *rvd = spa->spa_root_vdev; |
3459 | vdev_t *oldvd, *newvd, *pvd; | |
b128c09f | 3460 | nvlist_t *root; |
428870ff | 3461 | uint64_t leaves; |
34dc7c2f BB |
3462 | uint64_t leaf, top; |
3463 | uint64_t ashift = ztest_get_ashift(); | |
fb5f0bc8 | 3464 | uint64_t oldguid, pguid; |
5d1f7fb6 | 3465 | uint64_t oldsize, newsize; |
40b84e7a | 3466 | char *oldpath, *newpath; |
34dc7c2f | 3467 | int replacing; |
b128c09f BB |
3468 | int oldvd_has_siblings = B_FALSE; |
3469 | int newvd_is_spare = B_FALSE; | |
b2255edc | 3470 | int newvd_is_dspare = B_FALSE; |
b128c09f | 3471 | int oldvd_is_log; |
34dc7c2f | 3472 | int error, expected_error; |
34dc7c2f | 3473 | |
379ca9cf OF |
3474 | if (ztest_opts.zo_mmp_test) |
3475 | return; | |
3476 | ||
40b84e7a BB |
3477 | oldpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
3478 | newpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
3479 | ||
c242c188 | 3480 | mutex_enter(&ztest_vdev_lock); |
b2255edc | 3481 | leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raid_children; |
34dc7c2f | 3482 | |
a1d477c2 MA |
3483 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); |
3484 | ||
3485 | /* | |
3486 | * If a vdev is in the process of being removed, its removal may | |
3487 | * finish while we are in progress, leading to an unexpected error | |
3488 | * value. Don't bother trying to attach while we are in the middle | |
3489 | * of removal. | |
3490 | */ | |
9e052db4 | 3491 | if (ztest_device_removal_active) { |
a1d477c2 MA |
3492 | spa_config_exit(spa, SCL_ALL, FTAG); |
3493 | mutex_exit(&ztest_vdev_lock); | |
3494 | return; | |
3495 | } | |
34dc7c2f BB |
3496 | |
3497 | /* | |
3498 | * Decide whether to do an attach or a replace. | |
3499 | */ | |
3500 | replacing = ztest_random(2); | |
3501 | ||
3502 | /* | |
3503 | * Pick a random top-level vdev. | |
3504 | */ | |
428870ff | 3505 | top = ztest_random_vdev_top(spa, B_TRUE); |
34dc7c2f BB |
3506 | |
3507 | /* | |
3508 | * Pick a random leaf within it. | |
3509 | */ | |
3510 | leaf = ztest_random(leaves); | |
3511 | ||
3512 | /* | |
b128c09f | 3513 | * Locate this vdev. |
34dc7c2f | 3514 | */ |
b128c09f | 3515 | oldvd = rvd->vdev_child[top]; |
cc99f275 DB |
3516 | |
3517 | /* pick a child from the mirror */ | |
428870ff | 3518 | if (zs->zs_mirrors >= 1) { |
fb5f0bc8 | 3519 | ASSERT(oldvd->vdev_ops == &vdev_mirror_ops); |
428870ff | 3520 | ASSERT(oldvd->vdev_children >= zs->zs_mirrors); |
b2255edc | 3521 | oldvd = oldvd->vdev_child[leaf / ztest_opts.zo_raid_children]; |
fb5f0bc8 | 3522 | } |
cc99f275 DB |
3523 | |
3524 | /* pick a child out of the raidz group */ | |
b2255edc BB |
3525 | if (ztest_opts.zo_raid_children > 1) { |
3526 | if (strcmp(oldvd->vdev_ops->vdev_op_type, "raidz") == 0) | |
3527 | ASSERT(oldvd->vdev_ops == &vdev_raidz_ops); | |
3528 | else | |
3529 | ASSERT(oldvd->vdev_ops == &vdev_draid_ops); | |
3530 | ASSERT(oldvd->vdev_children == ztest_opts.zo_raid_children); | |
3531 | oldvd = oldvd->vdev_child[leaf % ztest_opts.zo_raid_children]; | |
fb5f0bc8 | 3532 | } |
34dc7c2f BB |
3533 | |
3534 | /* | |
b128c09f BB |
3535 | * If we're already doing an attach or replace, oldvd may be a |
3536 | * mirror vdev -- in which case, pick a random child. | |
34dc7c2f | 3537 | */ |
b128c09f BB |
3538 | while (oldvd->vdev_children != 0) { |
3539 | oldvd_has_siblings = B_TRUE; | |
fb5f0bc8 BB |
3540 | ASSERT(oldvd->vdev_children >= 2); |
3541 | oldvd = oldvd->vdev_child[ztest_random(oldvd->vdev_children)]; | |
b128c09f BB |
3542 | } |
3543 | ||
3544 | oldguid = oldvd->vdev_guid; | |
9babb374 | 3545 | oldsize = vdev_get_min_asize(oldvd); |
b128c09f BB |
3546 | oldvd_is_log = oldvd->vdev_top->vdev_islog; |
3547 | (void) strcpy(oldpath, oldvd->vdev_path); | |
3548 | pvd = oldvd->vdev_parent; | |
fb5f0bc8 | 3549 | pguid = pvd->vdev_guid; |
34dc7c2f BB |
3550 | |
3551 | /* | |
ce5fb2a7 BB |
3552 | * If oldvd has siblings, then half of the time, detach it. Prior |
3553 | * to the detach the pool is scrubbed in order to prevent creating | |
3554 | * unrepairable blocks as a result of the data corruption injection. | |
34dc7c2f | 3555 | */ |
b128c09f | 3556 | if (oldvd_has_siblings && ztest_random(2) == 0) { |
a1d477c2 | 3557 | spa_config_exit(spa, SCL_ALL, FTAG); |
ce5fb2a7 BB |
3558 | |
3559 | error = ztest_scrub_impl(spa); | |
3560 | if (error) | |
3561 | goto out; | |
3562 | ||
fb5f0bc8 BB |
3563 | error = spa_vdev_detach(spa, oldguid, pguid, B_FALSE); |
3564 | if (error != 0 && error != ENODEV && error != EBUSY && | |
d2734cce SD |
3565 | error != ENOTSUP && error != ZFS_ERR_CHECKPOINT_EXISTS && |
3566 | error != ZFS_ERR_DISCARDING_CHECKPOINT) | |
fb5f0bc8 | 3567 | fatal(0, "detach (%s) returned %d", oldpath, error); |
40b84e7a | 3568 | goto out; |
b128c09f | 3569 | } |
34dc7c2f BB |
3570 | |
3571 | /* | |
b128c09f BB |
3572 | * For the new vdev, choose with equal probability between the two |
3573 | * standard paths (ending in either 'a' or 'b') or a random hot spare. | |
34dc7c2f | 3574 | */ |
b128c09f BB |
3575 | if (sav->sav_count != 0 && ztest_random(3) == 0) { |
3576 | newvd = sav->sav_vdevs[ztest_random(sav->sav_count)]; | |
3577 | newvd_is_spare = B_TRUE; | |
b2255edc BB |
3578 | |
3579 | if (newvd->vdev_ops == &vdev_draid_spare_ops) | |
3580 | newvd_is_dspare = B_TRUE; | |
3581 | ||
b128c09f BB |
3582 | (void) strcpy(newpath, newvd->vdev_path); |
3583 | } else { | |
6aec1cd5 | 3584 | (void) snprintf(newpath, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
3585 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
3586 | top * leaves + leaf); | |
b128c09f BB |
3587 | if (ztest_random(2) == 0) |
3588 | newpath[strlen(newpath) - 1] = 'b'; | |
3589 | newvd = vdev_lookup_by_path(rvd, newpath); | |
3590 | } | |
3591 | ||
3592 | if (newvd) { | |
a1d477c2 MA |
3593 | /* |
3594 | * Reopen to ensure the vdev's asize field isn't stale. | |
3595 | */ | |
3596 | vdev_reopen(newvd); | |
9babb374 | 3597 | newsize = vdev_get_min_asize(newvd); |
b128c09f BB |
3598 | } else { |
3599 | /* | |
3600 | * Make newsize a little bigger or smaller than oldsize. | |
3601 | * If it's smaller, the attach should fail. | |
3602 | * If it's larger, and we're doing a replace, | |
3603 | * we should get dynamic LUN growth when we're done. | |
3604 | */ | |
3605 | newsize = 10 * oldsize / (9 + ztest_random(3)); | |
3606 | } | |
34dc7c2f BB |
3607 | |
3608 | /* | |
3609 | * If pvd is not a mirror or root, the attach should fail with ENOTSUP, | |
3610 | * unless it's a replace; in that case any non-replacing parent is OK. | |
3611 | * | |
3612 | * If newvd is already part of the pool, it should fail with EBUSY. | |
3613 | * | |
3614 | * If newvd is too small, it should fail with EOVERFLOW. | |
b2255edc BB |
3615 | * |
3616 | * If newvd is a distributed spare and it's being attached to a | |
3617 | * dRAID which is not its parent it should fail with EINVAL. | |
34dc7c2f | 3618 | */ |
b128c09f BB |
3619 | if (pvd->vdev_ops != &vdev_mirror_ops && |
3620 | pvd->vdev_ops != &vdev_root_ops && (!replacing || | |
3621 | pvd->vdev_ops == &vdev_replacing_ops || | |
3622 | pvd->vdev_ops == &vdev_spare_ops)) | |
34dc7c2f | 3623 | expected_error = ENOTSUP; |
b128c09f BB |
3624 | else if (newvd_is_spare && (!replacing || oldvd_is_log)) |
3625 | expected_error = ENOTSUP; | |
3626 | else if (newvd == oldvd) | |
3627 | expected_error = replacing ? 0 : EBUSY; | |
3628 | else if (vdev_lookup_by_path(rvd, newpath) != NULL) | |
3629 | expected_error = EBUSY; | |
b2255edc | 3630 | else if (!newvd_is_dspare && newsize < oldsize) |
34dc7c2f BB |
3631 | expected_error = EOVERFLOW; |
3632 | else if (ashift > oldvd->vdev_top->vdev_ashift) | |
3633 | expected_error = EDOM; | |
b2255edc BB |
3634 | else if (newvd_is_dspare && pvd != vdev_draid_spare_get_parent(newvd)) |
3635 | expected_error = ENOTSUP; | |
34dc7c2f BB |
3636 | else |
3637 | expected_error = 0; | |
3638 | ||
a1d477c2 | 3639 | spa_config_exit(spa, SCL_ALL, FTAG); |
34dc7c2f BB |
3640 | |
3641 | /* | |
3642 | * Build the nvlist describing newpath. | |
3643 | */ | |
ea0b2538 | 3644 | root = make_vdev_root(newpath, NULL, NULL, newvd == NULL ? newsize : 0, |
cc99f275 | 3645 | ashift, NULL, 0, 0, 1); |
34dc7c2f | 3646 | |
9a49d3f3 BB |
3647 | /* |
3648 | * When supported select either a healing or sequential resilver. | |
3649 | */ | |
3650 | boolean_t rebuilding = B_FALSE; | |
3651 | if (pvd->vdev_ops == &vdev_mirror_ops || | |
3652 | pvd->vdev_ops == &vdev_root_ops) { | |
3653 | rebuilding = !!ztest_random(2); | |
3654 | } | |
3655 | ||
3656 | error = spa_vdev_attach(spa, oldguid, root, replacing, rebuilding); | |
34dc7c2f | 3657 | |
34dc7c2f BB |
3658 | nvlist_free(root); |
3659 | ||
3660 | /* | |
3661 | * If our parent was the replacing vdev, but the replace completed, | |
3662 | * then instead of failing with ENOTSUP we may either succeed, | |
3663 | * fail with ENODEV, or fail with EOVERFLOW. | |
3664 | */ | |
3665 | if (expected_error == ENOTSUP && | |
3666 | (error == 0 || error == ENODEV || error == EOVERFLOW)) | |
3667 | expected_error = error; | |
3668 | ||
3669 | /* | |
3670 | * If someone grew the LUN, the replacement may be too small. | |
3671 | */ | |
b128c09f | 3672 | if (error == EOVERFLOW || error == EBUSY) |
34dc7c2f BB |
3673 | expected_error = error; |
3674 | ||
d2734cce | 3675 | if (error == ZFS_ERR_CHECKPOINT_EXISTS || |
9a49d3f3 BB |
3676 | error == ZFS_ERR_DISCARDING_CHECKPOINT || |
3677 | error == ZFS_ERR_RESILVER_IN_PROGRESS || | |
3678 | error == ZFS_ERR_REBUILD_IN_PROGRESS) | |
d2734cce SD |
3679 | expected_error = error; |
3680 | ||
b128c09f BB |
3681 | if (error != expected_error && expected_error != EBUSY) { |
3682 | fatal(0, "attach (%s %llu, %s %llu, %d) " | |
3683 | "returned %d, expected %d", | |
5d1f7fb6 GW |
3684 | oldpath, oldsize, newpath, |
3685 | newsize, replacing, error, expected_error); | |
34dc7c2f | 3686 | } |
40b84e7a | 3687 | out: |
c242c188 | 3688 | mutex_exit(&ztest_vdev_lock); |
40b84e7a BB |
3689 | |
3690 | umem_free(oldpath, MAXPATHLEN); | |
3691 | umem_free(newpath, MAXPATHLEN); | |
34dc7c2f BB |
3692 | } |
3693 | ||
a1d477c2 MA |
3694 | /* ARGSUSED */ |
3695 | void | |
3696 | ztest_device_removal(ztest_ds_t *zd, uint64_t id) | |
3697 | { | |
3698 | spa_t *spa = ztest_spa; | |
3699 | vdev_t *vd; | |
3700 | uint64_t guid; | |
9e052db4 | 3701 | int error; |
a1d477c2 MA |
3702 | |
3703 | mutex_enter(&ztest_vdev_lock); | |
3704 | ||
9e052db4 MA |
3705 | if (ztest_device_removal_active) { |
3706 | mutex_exit(&ztest_vdev_lock); | |
3707 | return; | |
3708 | } | |
3709 | ||
3710 | /* | |
3711 | * Remove a random top-level vdev and wait for removal to finish. | |
3712 | */ | |
a1d477c2 MA |
3713 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); |
3714 | vd = vdev_lookup_top(spa, ztest_random_vdev_top(spa, B_FALSE)); | |
3715 | guid = vd->vdev_guid; | |
3716 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
3717 | ||
9e052db4 MA |
3718 | error = spa_vdev_remove(spa, guid, B_FALSE); |
3719 | if (error == 0) { | |
3720 | ztest_device_removal_active = B_TRUE; | |
3721 | mutex_exit(&ztest_vdev_lock); | |
3722 | ||
7d658d29 TC |
3723 | /* |
3724 | * spa->spa_vdev_removal is created in a sync task that | |
3725 | * is initiated via dsl_sync_task_nowait(). Since the | |
3726 | * task may not run before spa_vdev_remove() returns, we | |
3727 | * must wait at least 1 txg to ensure that the removal | |
3728 | * struct has been created. | |
3729 | */ | |
3730 | txg_wait_synced(spa_get_dsl(spa), 0); | |
3731 | ||
c40a1124 | 3732 | while (spa->spa_removing_phys.sr_state == DSS_SCANNING) |
9e052db4 MA |
3733 | txg_wait_synced(spa_get_dsl(spa), 0); |
3734 | } else { | |
3735 | mutex_exit(&ztest_vdev_lock); | |
3736 | return; | |
3737 | } | |
a1d477c2 | 3738 | |
9e052db4 MA |
3739 | /* |
3740 | * The pool needs to be scrubbed after completing device removal. | |
3741 | * Failure to do so may result in checksum errors due to the | |
3742 | * strategy employed by ztest_fault_inject() when selecting which | |
3743 | * offset are redundant and can be damaged. | |
3744 | */ | |
3745 | error = spa_scan(spa, POOL_SCAN_SCRUB); | |
3746 | if (error == 0) { | |
3747 | while (dsl_scan_scrubbing(spa_get_dsl(spa))) | |
3748 | txg_wait_synced(spa_get_dsl(spa), 0); | |
3749 | } | |
3750 | ||
3751 | mutex_enter(&ztest_vdev_lock); | |
3752 | ztest_device_removal_active = B_FALSE; | |
a1d477c2 MA |
3753 | mutex_exit(&ztest_vdev_lock); |
3754 | } | |
3755 | ||
9babb374 BB |
3756 | /* |
3757 | * Callback function which expands the physical size of the vdev. | |
3758 | */ | |
65c7cc49 | 3759 | static vdev_t * |
9babb374 BB |
3760 | grow_vdev(vdev_t *vd, void *arg) |
3761 | { | |
2a8ba608 | 3762 | spa_t *spa __maybe_unused = vd->vdev_spa; |
9babb374 BB |
3763 | size_t *newsize = arg; |
3764 | size_t fsize; | |
3765 | int fd; | |
3766 | ||
3767 | ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE); | |
3768 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
3769 | ||
3770 | if ((fd = open(vd->vdev_path, O_RDWR)) == -1) | |
3771 | return (vd); | |
3772 | ||
3773 | fsize = lseek(fd, 0, SEEK_END); | |
0e5b68e0 | 3774 | VERIFY(ftruncate(fd, *newsize) == 0); |
9babb374 | 3775 | |
c242c188 | 3776 | if (ztest_opts.zo_verbose >= 6) { |
9babb374 BB |
3777 | (void) printf("%s grew from %lu to %lu bytes\n", |
3778 | vd->vdev_path, (ulong_t)fsize, (ulong_t)*newsize); | |
3779 | } | |
3780 | (void) close(fd); | |
3781 | return (NULL); | |
3782 | } | |
3783 | ||
3784 | /* | |
3785 | * Callback function which expands a given vdev by calling vdev_online(). | |
3786 | */ | |
3787 | /* ARGSUSED */ | |
65c7cc49 | 3788 | static vdev_t * |
9babb374 BB |
3789 | online_vdev(vdev_t *vd, void *arg) |
3790 | { | |
3791 | spa_t *spa = vd->vdev_spa; | |
3792 | vdev_t *tvd = vd->vdev_top; | |
9babb374 | 3793 | uint64_t guid = vd->vdev_guid; |
428870ff BB |
3794 | uint64_t generation = spa->spa_config_generation + 1; |
3795 | vdev_state_t newstate = VDEV_STATE_UNKNOWN; | |
3796 | int error; | |
9babb374 BB |
3797 | |
3798 | ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE); | |
3799 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
3800 | ||
3801 | /* Calling vdev_online will initialize the new metaslabs */ | |
3802 | spa_config_exit(spa, SCL_STATE, spa); | |
428870ff | 3803 | error = vdev_online(spa, guid, ZFS_ONLINE_EXPAND, &newstate); |
9babb374 BB |
3804 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); |
3805 | ||
428870ff BB |
3806 | /* |
3807 | * If vdev_online returned an error or the underlying vdev_open | |
3808 | * failed then we abort the expand. The only way to know that | |
3809 | * vdev_open fails is by checking the returned newstate. | |
3810 | */ | |
3811 | if (error || newstate != VDEV_STATE_HEALTHY) { | |
c242c188 | 3812 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
3813 | (void) printf("Unable to expand vdev, state %llu, " |
3814 | "error %d\n", (u_longlong_t)newstate, error); | |
3815 | } | |
3816 | return (vd); | |
3817 | } | |
3818 | ASSERT3U(newstate, ==, VDEV_STATE_HEALTHY); | |
3819 | ||
9babb374 BB |
3820 | /* |
3821 | * Since we dropped the lock we need to ensure that we're | |
3822 | * still talking to the original vdev. It's possible this | |
3823 | * vdev may have been detached/replaced while we were | |
3824 | * trying to online it. | |
3825 | */ | |
428870ff | 3826 | if (generation != spa->spa_config_generation) { |
c242c188 | 3827 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
3828 | (void) printf("vdev configuration has changed, " |
3829 | "guid %llu, state %llu, expected gen %llu, " | |
3830 | "got gen %llu\n", | |
3831 | (u_longlong_t)guid, | |
3832 | (u_longlong_t)tvd->vdev_state, | |
3833 | (u_longlong_t)generation, | |
3834 | (u_longlong_t)spa->spa_config_generation); | |
9babb374 BB |
3835 | } |
3836 | return (vd); | |
3837 | } | |
3838 | return (NULL); | |
3839 | } | |
3840 | ||
3841 | /* | |
3842 | * Traverse the vdev tree calling the supplied function. | |
3843 | * We continue to walk the tree until we either have walked all | |
3844 | * children or we receive a non-NULL return from the callback. | |
3845 | * If a NULL callback is passed, then we just return back the first | |
3846 | * leaf vdev we encounter. | |
3847 | */ | |
65c7cc49 | 3848 | static vdev_t * |
9babb374 BB |
3849 | vdev_walk_tree(vdev_t *vd, vdev_t *(*func)(vdev_t *, void *), void *arg) |
3850 | { | |
d6320ddb BB |
3851 | uint_t c; |
3852 | ||
9babb374 BB |
3853 | if (vd->vdev_ops->vdev_op_leaf) { |
3854 | if (func == NULL) | |
3855 | return (vd); | |
3856 | else | |
3857 | return (func(vd, arg)); | |
3858 | } | |
3859 | ||
d6320ddb | 3860 | for (c = 0; c < vd->vdev_children; c++) { |
9babb374 BB |
3861 | vdev_t *cvd = vd->vdev_child[c]; |
3862 | if ((cvd = vdev_walk_tree(cvd, func, arg)) != NULL) | |
3863 | return (cvd); | |
3864 | } | |
3865 | return (NULL); | |
3866 | } | |
3867 | ||
34dc7c2f BB |
3868 | /* |
3869 | * Verify that dynamic LUN growth works as expected. | |
3870 | */ | |
428870ff | 3871 | /* ARGSUSED */ |
34dc7c2f | 3872 | void |
428870ff | 3873 | ztest_vdev_LUN_growth(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3874 | { |
c242c188 | 3875 | spa_t *spa = ztest_spa; |
428870ff BB |
3876 | vdev_t *vd, *tvd; |
3877 | metaslab_class_t *mc; | |
3878 | metaslab_group_t *mg; | |
9babb374 | 3879 | size_t psize, newsize; |
428870ff BB |
3880 | uint64_t top; |
3881 | uint64_t old_class_space, new_class_space, old_ms_count, new_ms_count; | |
34dc7c2f | 3882 | |
d2734cce | 3883 | mutex_enter(&ztest_checkpoint_lock); |
c242c188 | 3884 | mutex_enter(&ztest_vdev_lock); |
9babb374 BB |
3885 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); |
3886 | ||
a1d477c2 MA |
3887 | /* |
3888 | * If there is a vdev removal in progress, it could complete while | |
3889 | * we are running, in which case we would not be able to verify | |
3890 | * that the metaslab_class space increased (because it decreases | |
3891 | * when the device removal completes). | |
3892 | */ | |
9e052db4 | 3893 | if (ztest_device_removal_active) { |
d2734cce | 3894 | spa_config_exit(spa, SCL_STATE, spa); |
a1d477c2 | 3895 | mutex_exit(&ztest_vdev_lock); |
d2734cce | 3896 | mutex_exit(&ztest_checkpoint_lock); |
a1d477c2 MA |
3897 | return; |
3898 | } | |
3899 | ||
428870ff | 3900 | top = ztest_random_vdev_top(spa, B_TRUE); |
9babb374 | 3901 | |
428870ff BB |
3902 | tvd = spa->spa_root_vdev->vdev_child[top]; |
3903 | mg = tvd->vdev_mg; | |
3904 | mc = mg->mg_class; | |
3905 | old_ms_count = tvd->vdev_ms_count; | |
3906 | old_class_space = metaslab_class_get_space(mc); | |
34dc7c2f BB |
3907 | |
3908 | /* | |
9babb374 BB |
3909 | * Determine the size of the first leaf vdev associated with |
3910 | * our top-level device. | |
34dc7c2f | 3911 | */ |
9babb374 BB |
3912 | vd = vdev_walk_tree(tvd, NULL, NULL); |
3913 | ASSERT3P(vd, !=, NULL); | |
3914 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
34dc7c2f | 3915 | |
9babb374 | 3916 | psize = vd->vdev_psize; |
34dc7c2f | 3917 | |
9babb374 | 3918 | /* |
428870ff BB |
3919 | * We only try to expand the vdev if it's healthy, less than 4x its |
3920 | * original size, and it has a valid psize. | |
9babb374 | 3921 | */ |
428870ff | 3922 | if (tvd->vdev_state != VDEV_STATE_HEALTHY || |
c242c188 | 3923 | psize == 0 || psize >= 4 * ztest_opts.zo_vdev_size) { |
9babb374 | 3924 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 3925 | mutex_exit(&ztest_vdev_lock); |
d2734cce | 3926 | mutex_exit(&ztest_checkpoint_lock); |
9babb374 BB |
3927 | return; |
3928 | } | |
3929 | ASSERT(psize > 0); | |
cc99f275 | 3930 | newsize = psize + MAX(psize / 8, SPA_MAXBLOCKSIZE); |
9babb374 | 3931 | ASSERT3U(newsize, >, psize); |
34dc7c2f | 3932 | |
c242c188 | 3933 | if (ztest_opts.zo_verbose >= 6) { |
428870ff | 3934 | (void) printf("Expanding LUN %s from %lu to %lu\n", |
9babb374 BB |
3935 | vd->vdev_path, (ulong_t)psize, (ulong_t)newsize); |
3936 | } | |
3937 | ||
9babb374 BB |
3938 | /* |
3939 | * Growing the vdev is a two step process: | |
3940 | * 1). expand the physical size (i.e. relabel) | |
3941 | * 2). online the vdev to create the new metaslabs | |
3942 | */ | |
3943 | if (vdev_walk_tree(tvd, grow_vdev, &newsize) != NULL || | |
3944 | vdev_walk_tree(tvd, online_vdev, NULL) != NULL || | |
3945 | tvd->vdev_state != VDEV_STATE_HEALTHY) { | |
c242c188 | 3946 | if (ztest_opts.zo_verbose >= 5) { |
9babb374 | 3947 | (void) printf("Could not expand LUN because " |
428870ff | 3948 | "the vdev configuration changed.\n"); |
34dc7c2f | 3949 | } |
428870ff | 3950 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 3951 | mutex_exit(&ztest_vdev_lock); |
d2734cce | 3952 | mutex_exit(&ztest_checkpoint_lock); |
9babb374 | 3953 | return; |
34dc7c2f BB |
3954 | } |
3955 | ||
428870ff | 3956 | spa_config_exit(spa, SCL_STATE, spa); |
9babb374 BB |
3957 | |
3958 | /* | |
3959 | * Expanding the LUN will update the config asynchronously, | |
3960 | * thus we must wait for the async thread to complete any | |
3961 | * pending tasks before proceeding. | |
3962 | */ | |
428870ff BB |
3963 | for (;;) { |
3964 | boolean_t done; | |
3965 | mutex_enter(&spa->spa_async_lock); | |
3966 | done = (spa->spa_async_thread == NULL && !spa->spa_async_tasks); | |
3967 | mutex_exit(&spa->spa_async_lock); | |
3968 | if (done) | |
3969 | break; | |
3970 | txg_wait_synced(spa_get_dsl(spa), 0); | |
3971 | (void) poll(NULL, 0, 100); | |
3972 | } | |
9babb374 BB |
3973 | |
3974 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); | |
428870ff BB |
3975 | |
3976 | tvd = spa->spa_root_vdev->vdev_child[top]; | |
3977 | new_ms_count = tvd->vdev_ms_count; | |
3978 | new_class_space = metaslab_class_get_space(mc); | |
3979 | ||
3980 | if (tvd->vdev_mg != mg || mg->mg_class != mc) { | |
c242c188 | 3981 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
3982 | (void) printf("Could not verify LUN expansion due to " |
3983 | "intervening vdev offline or remove.\n"); | |
3984 | } | |
3985 | spa_config_exit(spa, SCL_STATE, spa); | |
c242c188 | 3986 | mutex_exit(&ztest_vdev_lock); |
d2734cce | 3987 | mutex_exit(&ztest_checkpoint_lock); |
428870ff BB |
3988 | return; |
3989 | } | |
3990 | ||
3991 | /* | |
3992 | * Make sure we were able to grow the vdev. | |
3993 | */ | |
a1d477c2 MA |
3994 | if (new_ms_count <= old_ms_count) { |
3995 | fatal(0, "LUN expansion failed: ms_count %llu < %llu\n", | |
428870ff | 3996 | old_ms_count, new_ms_count); |
a1d477c2 | 3997 | } |
9babb374 BB |
3998 | |
3999 | /* | |
4000 | * Make sure we were able to grow the pool. | |
4001 | */ | |
a1d477c2 MA |
4002 | if (new_class_space <= old_class_space) { |
4003 | fatal(0, "LUN expansion failed: class_space %llu < %llu\n", | |
428870ff | 4004 | old_class_space, new_class_space); |
a1d477c2 | 4005 | } |
428870ff | 4006 | |
c242c188 | 4007 | if (ztest_opts.zo_verbose >= 5) { |
f3c8c9e6 | 4008 | char oldnumbuf[NN_NUMBUF_SZ], newnumbuf[NN_NUMBUF_SZ]; |
9babb374 | 4009 | |
f3c8c9e6 JK |
4010 | nicenum(old_class_space, oldnumbuf, sizeof (oldnumbuf)); |
4011 | nicenum(new_class_space, newnumbuf, sizeof (newnumbuf)); | |
9babb374 BB |
4012 | (void) printf("%s grew from %s to %s\n", |
4013 | spa->spa_name, oldnumbuf, newnumbuf); | |
4014 | } | |
428870ff | 4015 | |
9babb374 | 4016 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 4017 | mutex_exit(&ztest_vdev_lock); |
d2734cce | 4018 | mutex_exit(&ztest_checkpoint_lock); |
34dc7c2f BB |
4019 | } |
4020 | ||
428870ff BB |
4021 | /* |
4022 | * Verify that dmu_objset_{create,destroy,open,close} work as expected. | |
4023 | */ | |
34dc7c2f BB |
4024 | /* ARGSUSED */ |
4025 | static void | |
428870ff | 4026 | ztest_objset_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) |
34dc7c2f BB |
4027 | { |
4028 | /* | |
428870ff | 4029 | * Create the objects common to all ztest datasets. |
34dc7c2f | 4030 | */ |
428870ff | 4031 | VERIFY(zap_create_claim(os, ZTEST_DIROBJ, |
34dc7c2f | 4032 | DMU_OT_ZAP_OTHER, DMU_OT_NONE, 0, tx) == 0); |
428870ff | 4033 | } |
34dc7c2f | 4034 | |
428870ff BB |
4035 | static int |
4036 | ztest_dataset_create(char *dsname) | |
4037 | { | |
4807c0ba TC |
4038 | int err; |
4039 | uint64_t rand; | |
4040 | dsl_crypto_params_t *dcp = NULL; | |
4041 | ||
4042 | /* | |
4043 | * 50% of the time, we create encrypted datasets | |
4044 | * using a random cipher suite and a hard-coded | |
4045 | * wrapping key. | |
4046 | */ | |
4047 | rand = ztest_random(2); | |
4048 | if (rand != 0) { | |
4049 | nvlist_t *crypto_args = fnvlist_alloc(); | |
4050 | nvlist_t *props = fnvlist_alloc(); | |
4051 | ||
4052 | /* slight bias towards the default cipher suite */ | |
4053 | rand = ztest_random(ZIO_CRYPT_FUNCTIONS); | |
4054 | if (rand < ZIO_CRYPT_AES_128_CCM) | |
4055 | rand = ZIO_CRYPT_ON; | |
4056 | ||
4057 | fnvlist_add_uint64(props, | |
4058 | zfs_prop_to_name(ZFS_PROP_ENCRYPTION), rand); | |
4059 | fnvlist_add_uint8_array(crypto_args, "wkeydata", | |
4060 | (uint8_t *)ztest_wkeydata, WRAPPING_KEY_LEN); | |
4061 | ||
4062 | /* | |
4063 | * These parameters aren't really used by the kernel. They | |
4064 | * are simply stored so that userspace knows how to load | |
4065 | * the wrapping key. | |
4066 | */ | |
4067 | fnvlist_add_uint64(props, | |
4068 | zfs_prop_to_name(ZFS_PROP_KEYFORMAT), ZFS_KEYFORMAT_RAW); | |
4069 | fnvlist_add_string(props, | |
4070 | zfs_prop_to_name(ZFS_PROP_KEYLOCATION), "prompt"); | |
4071 | fnvlist_add_uint64(props, | |
4072 | zfs_prop_to_name(ZFS_PROP_PBKDF2_SALT), 0ULL); | |
4073 | fnvlist_add_uint64(props, | |
4074 | zfs_prop_to_name(ZFS_PROP_PBKDF2_ITERS), 0ULL); | |
4075 | ||
4076 | VERIFY0(dsl_crypto_params_create_nvlist(DCP_CMD_NONE, props, | |
4077 | crypto_args, &dcp)); | |
4078 | ||
010d1247 NL |
4079 | /* |
4080 | * Cycle through all available encryption implementations | |
4081 | * to verify interoperability. | |
4082 | */ | |
4083 | VERIFY0(gcm_impl_set("cycle")); | |
4084 | VERIFY0(aes_impl_set("cycle")); | |
4085 | ||
4807c0ba TC |
4086 | fnvlist_free(crypto_args); |
4087 | fnvlist_free(props); | |
4088 | } | |
4089 | ||
4090 | err = dmu_objset_create(dsname, DMU_OST_OTHER, 0, dcp, | |
428870ff | 4091 | ztest_objset_create_cb, NULL); |
4807c0ba | 4092 | dsl_crypto_params_free(dcp, !!err); |
428870ff | 4093 | |
4807c0ba TC |
4094 | rand = ztest_random(100); |
4095 | if (err || rand < 80) | |
428870ff BB |
4096 | return (err); |
4097 | ||
c242c188 | 4098 | if (ztest_opts.zo_verbose >= 5) |
b815ff9a | 4099 | (void) printf("Setting dataset %s to sync always\n", dsname); |
428870ff BB |
4100 | return (ztest_dsl_prop_set_uint64(dsname, ZFS_PROP_SYNC, |
4101 | ZFS_SYNC_ALWAYS, B_FALSE)); | |
34dc7c2f BB |
4102 | } |
4103 | ||
428870ff | 4104 | /* ARGSUSED */ |
34dc7c2f | 4105 | static int |
428870ff | 4106 | ztest_objset_destroy_cb(const char *name, void *arg) |
34dc7c2f | 4107 | { |
34dc7c2f | 4108 | objset_t *os; |
428870ff | 4109 | dmu_object_info_t doi; |
34dc7c2f BB |
4110 | int error; |
4111 | ||
4112 | /* | |
4113 | * Verify that the dataset contains a directory object. | |
4114 | */ | |
4807c0ba TC |
4115 | VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_TRUE, |
4116 | B_TRUE, FTAG, &os)); | |
428870ff | 4117 | error = dmu_object_info(os, ZTEST_DIROBJ, &doi); |
34dc7c2f BB |
4118 | if (error != ENOENT) { |
4119 | /* We could have crashed in the middle of destroying it */ | |
c99c9001 | 4120 | ASSERT0(error); |
428870ff BB |
4121 | ASSERT3U(doi.doi_type, ==, DMU_OT_ZAP_OTHER); |
4122 | ASSERT3S(doi.doi_physical_blocks_512, >=, 0); | |
34dc7c2f | 4123 | } |
b5256303 | 4124 | dmu_objset_disown(os, B_TRUE, FTAG); |
34dc7c2f BB |
4125 | |
4126 | /* | |
4127 | * Destroy the dataset. | |
4128 | */ | |
13fe0198 | 4129 | if (strchr(name, '@') != NULL) { |
dc1fbc43 | 4130 | VERIFY0(dsl_destroy_snapshot(name, B_TRUE)); |
13fe0198 | 4131 | } else { |
dc1fbc43 | 4132 | error = dsl_destroy_head(name); |
0dd6b6bf BB |
4133 | if (error == ENOSPC) { |
4134 | /* There could be checkpoint or insufficient slop */ | |
4135 | ztest_record_enospc(FTAG); | |
4136 | } else if (error != EBUSY) { | |
4137 | /* There could be a hold on this dataset */ | |
dc1fbc43 | 4138 | ASSERT0(error); |
0dd6b6bf | 4139 | } |
13fe0198 | 4140 | } |
34dc7c2f BB |
4141 | return (0); |
4142 | } | |
4143 | ||
428870ff BB |
4144 | static boolean_t |
4145 | ztest_snapshot_create(char *osname, uint64_t id) | |
34dc7c2f | 4146 | { |
eca7b760 | 4147 | char snapname[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
4148 | int error; |
4149 | ||
13fe0198 | 4150 | (void) snprintf(snapname, sizeof (snapname), "%llu", (u_longlong_t)id); |
428870ff | 4151 | |
13fe0198 | 4152 | error = dmu_objset_snapshot_one(osname, snapname); |
428870ff BB |
4153 | if (error == ENOSPC) { |
4154 | ztest_record_enospc(FTAG); | |
4155 | return (B_FALSE); | |
4156 | } | |
13fe0198 MA |
4157 | if (error != 0 && error != EEXIST) { |
4158 | fatal(0, "ztest_snapshot_create(%s@%s) = %d", osname, | |
4159 | snapname, error); | |
4160 | } | |
428870ff BB |
4161 | return (B_TRUE); |
4162 | } | |
4163 | ||
4164 | static boolean_t | |
4165 | ztest_snapshot_destroy(char *osname, uint64_t id) | |
4166 | { | |
eca7b760 | 4167 | char snapname[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
4168 | int error; |
4169 | ||
eca7b760 | 4170 | (void) snprintf(snapname, sizeof (snapname), "%s@%llu", osname, |
428870ff BB |
4171 | (u_longlong_t)id); |
4172 | ||
13fe0198 | 4173 | error = dsl_destroy_snapshot(snapname, B_FALSE); |
428870ff BB |
4174 | if (error != 0 && error != ENOENT) |
4175 | fatal(0, "ztest_snapshot_destroy(%s) = %d", snapname, error); | |
4176 | return (B_TRUE); | |
34dc7c2f BB |
4177 | } |
4178 | ||
428870ff | 4179 | /* ARGSUSED */ |
34dc7c2f | 4180 | void |
428870ff | 4181 | ztest_dmu_objset_create_destroy(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4182 | { |
40b84e7a | 4183 | ztest_ds_t *zdtmp; |
428870ff | 4184 | int iters; |
34dc7c2f | 4185 | int error; |
b128c09f | 4186 | objset_t *os, *os2; |
eca7b760 | 4187 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
34dc7c2f | 4188 | zilog_t *zilog; |
d6320ddb | 4189 | int i; |
34dc7c2f | 4190 | |
40b84e7a | 4191 | zdtmp = umem_alloc(sizeof (ztest_ds_t), UMEM_NOFAIL); |
40b84e7a | 4192 | |
2d9142c9 | 4193 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
34dc7c2f | 4194 | |
eca7b760 | 4195 | (void) snprintf(name, sizeof (name), "%s/temp_%llu", |
c242c188 | 4196 | ztest_opts.zo_pool, (u_longlong_t)id); |
34dc7c2f BB |
4197 | |
4198 | /* | |
4199 | * If this dataset exists from a previous run, process its replay log | |
13fe0198 | 4200 | * half of the time. If we don't replay it, then dsl_destroy_head() |
428870ff | 4201 | * (invoked from ztest_objset_destroy_cb()) should just throw it away. |
34dc7c2f BB |
4202 | */ |
4203 | if (ztest_random(2) == 0 && | |
4807c0ba | 4204 | ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, |
b5256303 | 4205 | B_TRUE, FTAG, &os) == 0) { |
c242c188 | 4206 | ztest_zd_init(zdtmp, NULL, os); |
40b84e7a BB |
4207 | zil_replay(os, zdtmp, ztest_replay_vector); |
4208 | ztest_zd_fini(zdtmp); | |
b5256303 | 4209 | dmu_objset_disown(os, B_TRUE, FTAG); |
34dc7c2f BB |
4210 | } |
4211 | ||
4212 | /* | |
4213 | * There may be an old instance of the dataset we're about to | |
4214 | * create lying around from a previous run. If so, destroy it | |
4215 | * and all of its snapshots. | |
4216 | */ | |
428870ff | 4217 | (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL, |
34dc7c2f BB |
4218 | DS_FIND_CHILDREN | DS_FIND_SNAPSHOTS); |
4219 | ||
4220 | /* | |
4221 | * Verify that the destroyed dataset is no longer in the namespace. | |
4222 | */ | |
4807c0ba TC |
4223 | VERIFY3U(ENOENT, ==, ztest_dmu_objset_own(name, DMU_OST_OTHER, B_TRUE, |
4224 | B_TRUE, FTAG, &os)); | |
34dc7c2f BB |
4225 | |
4226 | /* | |
4227 | * Verify that we can create a new dataset. | |
4228 | */ | |
428870ff | 4229 | error = ztest_dataset_create(name); |
34dc7c2f BB |
4230 | if (error) { |
4231 | if (error == ENOSPC) { | |
428870ff | 4232 | ztest_record_enospc(FTAG); |
40b84e7a | 4233 | goto out; |
34dc7c2f BB |
4234 | } |
4235 | fatal(0, "dmu_objset_create(%s) = %d", name, error); | |
4236 | } | |
4237 | ||
4807c0ba | 4238 | VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, B_TRUE, |
b5256303 | 4239 | FTAG, &os)); |
428870ff | 4240 | |
c242c188 | 4241 | ztest_zd_init(zdtmp, NULL, os); |
34dc7c2f BB |
4242 | |
4243 | /* | |
4244 | * Open the intent log for it. | |
4245 | */ | |
428870ff | 4246 | zilog = zil_open(os, ztest_get_data); |
34dc7c2f BB |
4247 | |
4248 | /* | |
428870ff BB |
4249 | * Put some objects in there, do a little I/O to them, |
4250 | * and randomly take a couple of snapshots along the way. | |
34dc7c2f | 4251 | */ |
428870ff | 4252 | iters = ztest_random(5); |
d6320ddb | 4253 | for (i = 0; i < iters; i++) { |
40b84e7a | 4254 | ztest_dmu_object_alloc_free(zdtmp, id); |
428870ff BB |
4255 | if (ztest_random(iters) == 0) |
4256 | (void) ztest_snapshot_create(name, i); | |
34dc7c2f BB |
4257 | } |
4258 | ||
4259 | /* | |
4260 | * Verify that we cannot create an existing dataset. | |
4261 | */ | |
428870ff | 4262 | VERIFY3U(EEXIST, ==, |
b5256303 | 4263 | dmu_objset_create(name, DMU_OST_OTHER, 0, NULL, NULL, NULL)); |
34dc7c2f BB |
4264 | |
4265 | /* | |
428870ff | 4266 | * Verify that we can hold an objset that is also owned. |
b128c09f | 4267 | */ |
428870ff BB |
4268 | VERIFY3U(0, ==, dmu_objset_hold(name, FTAG, &os2)); |
4269 | dmu_objset_rele(os2, FTAG); | |
34dc7c2f | 4270 | |
428870ff BB |
4271 | /* |
4272 | * Verify that we cannot own an objset that is already owned. | |
4273 | */ | |
4807c0ba TC |
4274 | VERIFY3U(EBUSY, ==, ztest_dmu_objset_own(name, DMU_OST_OTHER, |
4275 | B_FALSE, B_TRUE, FTAG, &os2)); | |
34dc7c2f | 4276 | |
428870ff | 4277 | zil_close(zilog); |
b5256303 | 4278 | dmu_objset_disown(os, B_TRUE, FTAG); |
40b84e7a BB |
4279 | ztest_zd_fini(zdtmp); |
4280 | out: | |
2d9142c9 | 4281 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
40b84e7a | 4282 | |
40b84e7a | 4283 | umem_free(zdtmp, sizeof (ztest_ds_t)); |
34dc7c2f BB |
4284 | } |
4285 | ||
4286 | /* | |
4287 | * Verify that dmu_snapshot_{create,destroy,open,close} work as expected. | |
4288 | */ | |
4289 | void | |
428870ff | 4290 | ztest_dmu_snapshot_create_destroy(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4291 | { |
2d9142c9 | 4292 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
428870ff BB |
4293 | (void) ztest_snapshot_destroy(zd->zd_name, id); |
4294 | (void) ztest_snapshot_create(zd->zd_name, id); | |
2d9142c9 | 4295 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
34dc7c2f BB |
4296 | } |
4297 | ||
9babb374 BB |
4298 | /* |
4299 | * Cleanup non-standard snapshots and clones. | |
4300 | */ | |
65c7cc49 | 4301 | static void |
428870ff | 4302 | ztest_dsl_dataset_cleanup(char *osname, uint64_t id) |
9babb374 | 4303 | { |
40b84e7a BB |
4304 | char *snap1name; |
4305 | char *clone1name; | |
4306 | char *snap2name; | |
4307 | char *clone2name; | |
4308 | char *snap3name; | |
9babb374 BB |
4309 | int error; |
4310 | ||
eca7b760 IK |
4311 | snap1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); |
4312 | clone1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4313 | snap2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4314 | clone2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4315 | snap3name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4316 | ||
4317 | (void) snprintf(snap1name, ZFS_MAX_DATASET_NAME_LEN, | |
4318 | "%s@s1_%llu", osname, (u_longlong_t)id); | |
4319 | (void) snprintf(clone1name, ZFS_MAX_DATASET_NAME_LEN, | |
4320 | "%s/c1_%llu", osname, (u_longlong_t)id); | |
4321 | (void) snprintf(snap2name, ZFS_MAX_DATASET_NAME_LEN, | |
4322 | "%s@s2_%llu", clone1name, (u_longlong_t)id); | |
4323 | (void) snprintf(clone2name, ZFS_MAX_DATASET_NAME_LEN, | |
4324 | "%s/c2_%llu", osname, (u_longlong_t)id); | |
4325 | (void) snprintf(snap3name, ZFS_MAX_DATASET_NAME_LEN, | |
4326 | "%s@s3_%llu", clone1name, (u_longlong_t)id); | |
9babb374 | 4327 | |
13fe0198 | 4328 | error = dsl_destroy_head(clone2name); |
9babb374 | 4329 | if (error && error != ENOENT) |
13fe0198 MA |
4330 | fatal(0, "dsl_destroy_head(%s) = %d", clone2name, error); |
4331 | error = dsl_destroy_snapshot(snap3name, B_FALSE); | |
9babb374 | 4332 | if (error && error != ENOENT) |
13fe0198 MA |
4333 | fatal(0, "dsl_destroy_snapshot(%s) = %d", snap3name, error); |
4334 | error = dsl_destroy_snapshot(snap2name, B_FALSE); | |
9babb374 | 4335 | if (error && error != ENOENT) |
13fe0198 MA |
4336 | fatal(0, "dsl_destroy_snapshot(%s) = %d", snap2name, error); |
4337 | error = dsl_destroy_head(clone1name); | |
9babb374 | 4338 | if (error && error != ENOENT) |
13fe0198 MA |
4339 | fatal(0, "dsl_destroy_head(%s) = %d", clone1name, error); |
4340 | error = dsl_destroy_snapshot(snap1name, B_FALSE); | |
9babb374 | 4341 | if (error && error != ENOENT) |
13fe0198 | 4342 | fatal(0, "dsl_destroy_snapshot(%s) = %d", snap1name, error); |
40b84e7a | 4343 | |
eca7b760 IK |
4344 | umem_free(snap1name, ZFS_MAX_DATASET_NAME_LEN); |
4345 | umem_free(clone1name, ZFS_MAX_DATASET_NAME_LEN); | |
4346 | umem_free(snap2name, ZFS_MAX_DATASET_NAME_LEN); | |
4347 | umem_free(clone2name, ZFS_MAX_DATASET_NAME_LEN); | |
4348 | umem_free(snap3name, ZFS_MAX_DATASET_NAME_LEN); | |
9babb374 BB |
4349 | } |
4350 | ||
4351 | /* | |
4352 | * Verify dsl_dataset_promote handles EBUSY | |
4353 | */ | |
4354 | void | |
428870ff | 4355 | ztest_dsl_dataset_promote_busy(ztest_ds_t *zd, uint64_t id) |
9babb374 | 4356 | { |
13fe0198 | 4357 | objset_t *os; |
40b84e7a BB |
4358 | char *snap1name; |
4359 | char *clone1name; | |
4360 | char *snap2name; | |
4361 | char *clone2name; | |
4362 | char *snap3name; | |
428870ff BB |
4363 | char *osname = zd->zd_name; |
4364 | int error; | |
9babb374 | 4365 | |
eca7b760 IK |
4366 | snap1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); |
4367 | clone1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4368 | snap2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4369 | clone2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4370 | snap3name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
40b84e7a | 4371 | |
2d9142c9 | 4372 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
9babb374 | 4373 | |
428870ff | 4374 | ztest_dsl_dataset_cleanup(osname, id); |
9babb374 | 4375 | |
eca7b760 IK |
4376 | (void) snprintf(snap1name, ZFS_MAX_DATASET_NAME_LEN, |
4377 | "%s@s1_%llu", osname, (u_longlong_t)id); | |
4378 | (void) snprintf(clone1name, ZFS_MAX_DATASET_NAME_LEN, | |
4379 | "%s/c1_%llu", osname, (u_longlong_t)id); | |
4380 | (void) snprintf(snap2name, ZFS_MAX_DATASET_NAME_LEN, | |
4381 | "%s@s2_%llu", clone1name, (u_longlong_t)id); | |
4382 | (void) snprintf(clone2name, ZFS_MAX_DATASET_NAME_LEN, | |
4383 | "%s/c2_%llu", osname, (u_longlong_t)id); | |
4384 | (void) snprintf(snap3name, ZFS_MAX_DATASET_NAME_LEN, | |
4385 | "%s@s3_%llu", clone1name, (u_longlong_t)id); | |
9babb374 | 4386 | |
6f1ffb06 | 4387 | error = dmu_objset_snapshot_one(osname, strchr(snap1name, '@') + 1); |
9babb374 BB |
4388 | if (error && error != EEXIST) { |
4389 | if (error == ENOSPC) { | |
428870ff | 4390 | ztest_record_enospc(FTAG); |
9babb374 BB |
4391 | goto out; |
4392 | } | |
4393 | fatal(0, "dmu_take_snapshot(%s) = %d", snap1name, error); | |
4394 | } | |
4395 | ||
13fe0198 | 4396 | error = dmu_objset_clone(clone1name, snap1name); |
9babb374 BB |
4397 | if (error) { |
4398 | if (error == ENOSPC) { | |
428870ff | 4399 | ztest_record_enospc(FTAG); |
9babb374 BB |
4400 | goto out; |
4401 | } | |
4402 | fatal(0, "dmu_objset_create(%s) = %d", clone1name, error); | |
4403 | } | |
4404 | ||
6f1ffb06 | 4405 | error = dmu_objset_snapshot_one(clone1name, strchr(snap2name, '@') + 1); |
428870ff BB |
4406 | if (error && error != EEXIST) { |
4407 | if (error == ENOSPC) { | |
4408 | ztest_record_enospc(FTAG); | |
4409 | goto out; | |
34dc7c2f | 4410 | } |
428870ff BB |
4411 | fatal(0, "dmu_open_snapshot(%s) = %d", snap2name, error); |
4412 | } | |
34dc7c2f | 4413 | |
6f1ffb06 | 4414 | error = dmu_objset_snapshot_one(clone1name, strchr(snap3name, '@') + 1); |
428870ff BB |
4415 | if (error && error != EEXIST) { |
4416 | if (error == ENOSPC) { | |
4417 | ztest_record_enospc(FTAG); | |
4418 | goto out; | |
4419 | } | |
4420 | fatal(0, "dmu_open_snapshot(%s) = %d", snap3name, error); | |
4421 | } | |
34dc7c2f | 4422 | |
13fe0198 | 4423 | error = dmu_objset_clone(clone2name, snap3name); |
428870ff BB |
4424 | if (error) { |
4425 | if (error == ENOSPC) { | |
4426 | ztest_record_enospc(FTAG); | |
4427 | goto out; | |
4428 | } | |
4429 | fatal(0, "dmu_objset_create(%s) = %d", clone2name, error); | |
4430 | } | |
34dc7c2f | 4431 | |
4807c0ba | 4432 | error = ztest_dmu_objset_own(snap2name, DMU_OST_ANY, B_TRUE, B_TRUE, |
b5256303 | 4433 | FTAG, &os); |
428870ff | 4434 | if (error) |
13fe0198 | 4435 | fatal(0, "dmu_objset_own(%s) = %d", snap2name, error); |
428870ff | 4436 | error = dsl_dataset_promote(clone2name, NULL); |
9b67f605 | 4437 | if (error == ENOSPC) { |
b5256303 | 4438 | dmu_objset_disown(os, B_TRUE, FTAG); |
9b67f605 MA |
4439 | ztest_record_enospc(FTAG); |
4440 | goto out; | |
4441 | } | |
428870ff BB |
4442 | if (error != EBUSY) |
4443 | fatal(0, "dsl_dataset_promote(%s), %d, not EBUSY", clone2name, | |
4444 | error); | |
b5256303 | 4445 | dmu_objset_disown(os, B_TRUE, FTAG); |
34dc7c2f | 4446 | |
428870ff BB |
4447 | out: |
4448 | ztest_dsl_dataset_cleanup(osname, id); | |
34dc7c2f | 4449 | |
2d9142c9 | 4450 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
40b84e7a | 4451 | |
eca7b760 IK |
4452 | umem_free(snap1name, ZFS_MAX_DATASET_NAME_LEN); |
4453 | umem_free(clone1name, ZFS_MAX_DATASET_NAME_LEN); | |
4454 | umem_free(snap2name, ZFS_MAX_DATASET_NAME_LEN); | |
4455 | umem_free(clone2name, ZFS_MAX_DATASET_NAME_LEN); | |
4456 | umem_free(snap3name, ZFS_MAX_DATASET_NAME_LEN); | |
428870ff | 4457 | } |
34dc7c2f | 4458 | |
40b84e7a | 4459 | #undef OD_ARRAY_SIZE |
d1d7e268 | 4460 | #define OD_ARRAY_SIZE 4 |
40b84e7a | 4461 | |
428870ff BB |
4462 | /* |
4463 | * Verify that dmu_object_{alloc,free} work as expected. | |
4464 | */ | |
4465 | void | |
4466 | ztest_dmu_object_alloc_free(ztest_ds_t *zd, uint64_t id) | |
4467 | { | |
40b84e7a BB |
4468 | ztest_od_t *od; |
4469 | int batchsize; | |
4470 | int size; | |
d6320ddb | 4471 | int b; |
34dc7c2f | 4472 | |
d1d7e268 | 4473 | size = sizeof (ztest_od_t) * OD_ARRAY_SIZE; |
40b84e7a BB |
4474 | od = umem_alloc(size, UMEM_NOFAIL); |
4475 | batchsize = OD_ARRAY_SIZE; | |
4476 | ||
d6320ddb | 4477 | for (b = 0; b < batchsize; b++) |
50c957f7 NB |
4478 | ztest_od_init(od + b, id, FTAG, b, DMU_OT_UINT64_OTHER, |
4479 | 0, 0, 0); | |
34dc7c2f | 4480 | |
428870ff BB |
4481 | /* |
4482 | * Destroy the previous batch of objects, create a new batch, | |
4483 | * and do some I/O on the new objects. | |
4484 | */ | |
40b84e7a | 4485 | if (ztest_object_init(zd, od, size, B_TRUE) != 0) |
428870ff | 4486 | return; |
34dc7c2f | 4487 | |
428870ff BB |
4488 | while (ztest_random(4 * batchsize) != 0) |
4489 | ztest_io(zd, od[ztest_random(batchsize)].od_object, | |
4490 | ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
40b84e7a BB |
4491 | |
4492 | umem_free(od, size); | |
34dc7c2f BB |
4493 | } |
4494 | ||
047116ac TC |
4495 | /* |
4496 | * Rewind the global allocator to verify object allocation backfilling. | |
4497 | */ | |
4498 | void | |
4499 | ztest_dmu_object_next_chunk(ztest_ds_t *zd, uint64_t id) | |
4500 | { | |
4501 | objset_t *os = zd->zd_os; | |
4502 | int dnodes_per_chunk = 1 << dmu_object_alloc_chunk_shift; | |
4503 | uint64_t object; | |
4504 | ||
4505 | /* | |
4506 | * Rewind the global allocator randomly back to a lower object number | |
4507 | * to force backfilling and reclamation of recently freed dnodes. | |
4508 | */ | |
4509 | mutex_enter(&os->os_obj_lock); | |
4510 | object = ztest_random(os->os_obj_next_chunk); | |
4511 | os->os_obj_next_chunk = P2ALIGN(object, dnodes_per_chunk); | |
4512 | mutex_exit(&os->os_obj_lock); | |
4513 | } | |
4514 | ||
40b84e7a | 4515 | #undef OD_ARRAY_SIZE |
d1d7e268 | 4516 | #define OD_ARRAY_SIZE 2 |
40b84e7a | 4517 | |
34dc7c2f BB |
4518 | /* |
4519 | * Verify that dmu_{read,write} work as expected. | |
4520 | */ | |
34dc7c2f | 4521 | void |
428870ff | 4522 | ztest_dmu_read_write(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4523 | { |
40b84e7a BB |
4524 | int size; |
4525 | ztest_od_t *od; | |
4526 | ||
428870ff | 4527 | objset_t *os = zd->zd_os; |
d1d7e268 | 4528 | size = sizeof (ztest_od_t) * OD_ARRAY_SIZE; |
40b84e7a | 4529 | od = umem_alloc(size, UMEM_NOFAIL); |
34dc7c2f BB |
4530 | dmu_tx_t *tx; |
4531 | int i, freeit, error; | |
4532 | uint64_t n, s, txg; | |
4533 | bufwad_t *packbuf, *bigbuf, *pack, *bigH, *bigT; | |
428870ff BB |
4534 | uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize; |
4535 | uint64_t chunksize = (1000 + ztest_random(1000)) * sizeof (uint64_t); | |
34dc7c2f BB |
4536 | uint64_t regions = 997; |
4537 | uint64_t stride = 123456789ULL; | |
4538 | uint64_t width = 40; | |
4539 | int free_percent = 5; | |
4540 | ||
4541 | /* | |
4542 | * This test uses two objects, packobj and bigobj, that are always | |
4543 | * updated together (i.e. in the same tx) so that their contents are | |
4544 | * in sync and can be compared. Their contents relate to each other | |
4545 | * in a simple way: packobj is a dense array of 'bufwad' structures, | |
4546 | * while bigobj is a sparse array of the same bufwads. Specifically, | |
4547 | * for any index n, there are three bufwads that should be identical: | |
4548 | * | |
4549 | * packobj, at offset n * sizeof (bufwad_t) | |
4550 | * bigobj, at the head of the nth chunk | |
4551 | * bigobj, at the tail of the nth chunk | |
4552 | * | |
4553 | * The chunk size is arbitrary. It doesn't have to be a power of two, | |
4554 | * and it doesn't have any relation to the object blocksize. | |
4555 | * The only requirement is that it can hold at least two bufwads. | |
4556 | * | |
4557 | * Normally, we write the bufwad to each of these locations. | |
4558 | * However, free_percent of the time we instead write zeroes to | |
4559 | * packobj and perform a dmu_free_range() on bigobj. By comparing | |
4560 | * bigobj to packobj, we can verify that the DMU is correctly | |
4561 | * tracking which parts of an object are allocated and free, | |
4562 | * and that the contents of the allocated blocks are correct. | |
4563 | */ | |
4564 | ||
4565 | /* | |
4566 | * Read the directory info. If it's the first time, set things up. | |
4567 | */ | |
50c957f7 NB |
4568 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, chunksize); |
4569 | ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, 0, | |
4570 | chunksize); | |
34dc7c2f | 4571 | |
40b84e7a BB |
4572 | if (ztest_object_init(zd, od, size, B_FALSE) != 0) { |
4573 | umem_free(od, size); | |
428870ff | 4574 | return; |
40b84e7a | 4575 | } |
34dc7c2f | 4576 | |
428870ff BB |
4577 | bigobj = od[0].od_object; |
4578 | packobj = od[1].od_object; | |
4579 | chunksize = od[0].od_gen; | |
4580 | ASSERT(chunksize == od[1].od_gen); | |
34dc7c2f BB |
4581 | |
4582 | /* | |
4583 | * Prefetch a random chunk of the big object. | |
4584 | * Our aim here is to get some async reads in flight | |
4585 | * for blocks that we may free below; the DMU should | |
4586 | * handle this race correctly. | |
4587 | */ | |
4588 | n = ztest_random(regions) * stride + ztest_random(width); | |
4589 | s = 1 + ztest_random(2 * width - 1); | |
fcff0f35 PD |
4590 | dmu_prefetch(os, bigobj, 0, n * chunksize, s * chunksize, |
4591 | ZIO_PRIORITY_SYNC_READ); | |
34dc7c2f BB |
4592 | |
4593 | /* | |
4594 | * Pick a random index and compute the offsets into packobj and bigobj. | |
4595 | */ | |
4596 | n = ztest_random(regions) * stride + ztest_random(width); | |
4597 | s = 1 + ztest_random(width - 1); | |
4598 | ||
4599 | packoff = n * sizeof (bufwad_t); | |
4600 | packsize = s * sizeof (bufwad_t); | |
4601 | ||
428870ff BB |
4602 | bigoff = n * chunksize; |
4603 | bigsize = s * chunksize; | |
34dc7c2f BB |
4604 | |
4605 | packbuf = umem_alloc(packsize, UMEM_NOFAIL); | |
4606 | bigbuf = umem_alloc(bigsize, UMEM_NOFAIL); | |
4607 | ||
4608 | /* | |
4609 | * free_percent of the time, free a range of bigobj rather than | |
4610 | * overwriting it. | |
4611 | */ | |
4612 | freeit = (ztest_random(100) < free_percent); | |
4613 | ||
4614 | /* | |
4615 | * Read the current contents of our objects. | |
4616 | */ | |
428870ff | 4617 | error = dmu_read(os, packobj, packoff, packsize, packbuf, |
9babb374 | 4618 | DMU_READ_PREFETCH); |
c99c9001 | 4619 | ASSERT0(error); |
428870ff | 4620 | error = dmu_read(os, bigobj, bigoff, bigsize, bigbuf, |
9babb374 | 4621 | DMU_READ_PREFETCH); |
c99c9001 | 4622 | ASSERT0(error); |
34dc7c2f BB |
4623 | |
4624 | /* | |
4625 | * Get a tx for the mods to both packobj and bigobj. | |
4626 | */ | |
4627 | tx = dmu_tx_create(os); | |
4628 | ||
428870ff | 4629 | dmu_tx_hold_write(tx, packobj, packoff, packsize); |
34dc7c2f BB |
4630 | |
4631 | if (freeit) | |
428870ff | 4632 | dmu_tx_hold_free(tx, bigobj, bigoff, bigsize); |
34dc7c2f | 4633 | else |
428870ff | 4634 | dmu_tx_hold_write(tx, bigobj, bigoff, bigsize); |
34dc7c2f | 4635 | |
383fc4a9 MA |
4636 | /* This accounts for setting the checksum/compression. */ |
4637 | dmu_tx_hold_bonus(tx, bigobj); | |
4638 | ||
428870ff BB |
4639 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); |
4640 | if (txg == 0) { | |
34dc7c2f BB |
4641 | umem_free(packbuf, packsize); |
4642 | umem_free(bigbuf, bigsize); | |
40b84e7a | 4643 | umem_free(od, size); |
34dc7c2f BB |
4644 | return; |
4645 | } | |
4646 | ||
9b67f605 MA |
4647 | enum zio_checksum cksum; |
4648 | do { | |
4649 | cksum = (enum zio_checksum) | |
4650 | ztest_random_dsl_prop(ZFS_PROP_CHECKSUM); | |
4651 | } while (cksum >= ZIO_CHECKSUM_LEGACY_FUNCTIONS); | |
4652 | dmu_object_set_checksum(os, bigobj, cksum, tx); | |
428870ff | 4653 | |
9b67f605 MA |
4654 | enum zio_compress comp; |
4655 | do { | |
4656 | comp = (enum zio_compress) | |
4657 | ztest_random_dsl_prop(ZFS_PROP_COMPRESSION); | |
4658 | } while (comp >= ZIO_COMPRESS_LEGACY_FUNCTIONS); | |
4659 | dmu_object_set_compress(os, bigobj, comp, tx); | |
34dc7c2f BB |
4660 | |
4661 | /* | |
4662 | * For each index from n to n + s, verify that the existing bufwad | |
4663 | * in packobj matches the bufwads at the head and tail of the | |
4664 | * corresponding chunk in bigobj. Then update all three bufwads | |
4665 | * with the new values we want to write out. | |
4666 | */ | |
4667 | for (i = 0; i < s; i++) { | |
4668 | /* LINTED */ | |
4669 | pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t)); | |
4670 | /* LINTED */ | |
428870ff | 4671 | bigH = (bufwad_t *)((char *)bigbuf + i * chunksize); |
34dc7c2f | 4672 | /* LINTED */ |
428870ff | 4673 | bigT = (bufwad_t *)((char *)bigH + chunksize) - 1; |
34dc7c2f BB |
4674 | |
4675 | ASSERT((uintptr_t)bigH - (uintptr_t)bigbuf < bigsize); | |
4676 | ASSERT((uintptr_t)bigT - (uintptr_t)bigbuf < bigsize); | |
4677 | ||
4678 | if (pack->bw_txg > txg) | |
4679 | fatal(0, "future leak: got %llx, open txg is %llx", | |
4680 | pack->bw_txg, txg); | |
4681 | ||
4682 | if (pack->bw_data != 0 && pack->bw_index != n + i) | |
4683 | fatal(0, "wrong index: got %llx, wanted %llx+%llx", | |
4684 | pack->bw_index, n, i); | |
4685 | ||
4686 | if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0) | |
4687 | fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH); | |
4688 | ||
4689 | if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0) | |
4690 | fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT); | |
4691 | ||
4692 | if (freeit) { | |
4693 | bzero(pack, sizeof (bufwad_t)); | |
4694 | } else { | |
4695 | pack->bw_index = n + i; | |
4696 | pack->bw_txg = txg; | |
4697 | pack->bw_data = 1 + ztest_random(-2ULL); | |
4698 | } | |
4699 | *bigH = *pack; | |
4700 | *bigT = *pack; | |
4701 | } | |
4702 | ||
4703 | /* | |
4704 | * We've verified all the old bufwads, and made new ones. | |
4705 | * Now write them out. | |
4706 | */ | |
428870ff | 4707 | dmu_write(os, packobj, packoff, packsize, packbuf, tx); |
34dc7c2f BB |
4708 | |
4709 | if (freeit) { | |
c242c188 | 4710 | if (ztest_opts.zo_verbose >= 7) { |
34dc7c2f BB |
4711 | (void) printf("freeing offset %llx size %llx" |
4712 | " txg %llx\n", | |
4713 | (u_longlong_t)bigoff, | |
4714 | (u_longlong_t)bigsize, | |
4715 | (u_longlong_t)txg); | |
4716 | } | |
428870ff | 4717 | VERIFY(0 == dmu_free_range(os, bigobj, bigoff, bigsize, tx)); |
34dc7c2f | 4718 | } else { |
c242c188 | 4719 | if (ztest_opts.zo_verbose >= 7) { |
34dc7c2f BB |
4720 | (void) printf("writing offset %llx size %llx" |
4721 | " txg %llx\n", | |
4722 | (u_longlong_t)bigoff, | |
4723 | (u_longlong_t)bigsize, | |
4724 | (u_longlong_t)txg); | |
4725 | } | |
428870ff | 4726 | dmu_write(os, bigobj, bigoff, bigsize, bigbuf, tx); |
34dc7c2f BB |
4727 | } |
4728 | ||
4729 | dmu_tx_commit(tx); | |
4730 | ||
4731 | /* | |
4732 | * Sanity check the stuff we just wrote. | |
4733 | */ | |
4734 | { | |
4735 | void *packcheck = umem_alloc(packsize, UMEM_NOFAIL); | |
4736 | void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL); | |
4737 | ||
428870ff | 4738 | VERIFY(0 == dmu_read(os, packobj, packoff, |
9babb374 | 4739 | packsize, packcheck, DMU_READ_PREFETCH)); |
428870ff | 4740 | VERIFY(0 == dmu_read(os, bigobj, bigoff, |
9babb374 | 4741 | bigsize, bigcheck, DMU_READ_PREFETCH)); |
34dc7c2f BB |
4742 | |
4743 | ASSERT(bcmp(packbuf, packcheck, packsize) == 0); | |
4744 | ASSERT(bcmp(bigbuf, bigcheck, bigsize) == 0); | |
4745 | ||
4746 | umem_free(packcheck, packsize); | |
4747 | umem_free(bigcheck, bigsize); | |
4748 | } | |
4749 | ||
4750 | umem_free(packbuf, packsize); | |
4751 | umem_free(bigbuf, bigsize); | |
40b84e7a | 4752 | umem_free(od, size); |
34dc7c2f BB |
4753 | } |
4754 | ||
65c7cc49 | 4755 | static void |
9babb374 | 4756 | compare_and_update_pbbufs(uint64_t s, bufwad_t *packbuf, bufwad_t *bigbuf, |
428870ff | 4757 | uint64_t bigsize, uint64_t n, uint64_t chunksize, uint64_t txg) |
9babb374 BB |
4758 | { |
4759 | uint64_t i; | |
4760 | bufwad_t *pack; | |
4761 | bufwad_t *bigH; | |
4762 | bufwad_t *bigT; | |
4763 | ||
4764 | /* | |
4765 | * For each index from n to n + s, verify that the existing bufwad | |
4766 | * in packobj matches the bufwads at the head and tail of the | |
4767 | * corresponding chunk in bigobj. Then update all three bufwads | |
4768 | * with the new values we want to write out. | |
4769 | */ | |
4770 | for (i = 0; i < s; i++) { | |
4771 | /* LINTED */ | |
4772 | pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t)); | |
4773 | /* LINTED */ | |
428870ff | 4774 | bigH = (bufwad_t *)((char *)bigbuf + i * chunksize); |
9babb374 | 4775 | /* LINTED */ |
428870ff | 4776 | bigT = (bufwad_t *)((char *)bigH + chunksize) - 1; |
9babb374 BB |
4777 | |
4778 | ASSERT((uintptr_t)bigH - (uintptr_t)bigbuf < bigsize); | |
4779 | ASSERT((uintptr_t)bigT - (uintptr_t)bigbuf < bigsize); | |
4780 | ||
4781 | if (pack->bw_txg > txg) | |
4782 | fatal(0, "future leak: got %llx, open txg is %llx", | |
4783 | pack->bw_txg, txg); | |
4784 | ||
4785 | if (pack->bw_data != 0 && pack->bw_index != n + i) | |
4786 | fatal(0, "wrong index: got %llx, wanted %llx+%llx", | |
4787 | pack->bw_index, n, i); | |
4788 | ||
4789 | if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0) | |
4790 | fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH); | |
4791 | ||
4792 | if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0) | |
4793 | fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT); | |
4794 | ||
4795 | pack->bw_index = n + i; | |
4796 | pack->bw_txg = txg; | |
4797 | pack->bw_data = 1 + ztest_random(-2ULL); | |
4798 | ||
4799 | *bigH = *pack; | |
4800 | *bigT = *pack; | |
4801 | } | |
4802 | } | |
4803 | ||
40b84e7a | 4804 | #undef OD_ARRAY_SIZE |
d1d7e268 | 4805 | #define OD_ARRAY_SIZE 2 |
40b84e7a | 4806 | |
9babb374 | 4807 | void |
428870ff | 4808 | ztest_dmu_read_write_zcopy(ztest_ds_t *zd, uint64_t id) |
9babb374 | 4809 | { |
428870ff | 4810 | objset_t *os = zd->zd_os; |
40b84e7a | 4811 | ztest_od_t *od; |
9babb374 BB |
4812 | dmu_tx_t *tx; |
4813 | uint64_t i; | |
4814 | int error; | |
40b84e7a | 4815 | int size; |
9babb374 BB |
4816 | uint64_t n, s, txg; |
4817 | bufwad_t *packbuf, *bigbuf; | |
428870ff BB |
4818 | uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize; |
4819 | uint64_t blocksize = ztest_random_blocksize(); | |
4820 | uint64_t chunksize = blocksize; | |
9babb374 BB |
4821 | uint64_t regions = 997; |
4822 | uint64_t stride = 123456789ULL; | |
4823 | uint64_t width = 9; | |
4824 | dmu_buf_t *bonus_db; | |
4825 | arc_buf_t **bigbuf_arcbufs; | |
428870ff | 4826 | dmu_object_info_t doi; |
9babb374 | 4827 | |
d1d7e268 | 4828 | size = sizeof (ztest_od_t) * OD_ARRAY_SIZE; |
40b84e7a BB |
4829 | od = umem_alloc(size, UMEM_NOFAIL); |
4830 | ||
9babb374 BB |
4831 | /* |
4832 | * This test uses two objects, packobj and bigobj, that are always | |
4833 | * updated together (i.e. in the same tx) so that their contents are | |
4834 | * in sync and can be compared. Their contents relate to each other | |
4835 | * in a simple way: packobj is a dense array of 'bufwad' structures, | |
4836 | * while bigobj is a sparse array of the same bufwads. Specifically, | |
4837 | * for any index n, there are three bufwads that should be identical: | |
4838 | * | |
4839 | * packobj, at offset n * sizeof (bufwad_t) | |
4840 | * bigobj, at the head of the nth chunk | |
4841 | * bigobj, at the tail of the nth chunk | |
4842 | * | |
4843 | * The chunk size is set equal to bigobj block size so that | |
440a3eb9 | 4844 | * dmu_assign_arcbuf_by_dbuf() can be tested for object updates. |
9babb374 BB |
4845 | */ |
4846 | ||
4847 | /* | |
4848 | * Read the directory info. If it's the first time, set things up. | |
4849 | */ | |
50c957f7 NB |
4850 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0, 0); |
4851 | ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, 0, | |
4852 | chunksize); | |
40b84e7a | 4853 | |
9babb374 | 4854 | |
40b84e7a BB |
4855 | if (ztest_object_init(zd, od, size, B_FALSE) != 0) { |
4856 | umem_free(od, size); | |
428870ff | 4857 | return; |
40b84e7a | 4858 | } |
9babb374 | 4859 | |
428870ff BB |
4860 | bigobj = od[0].od_object; |
4861 | packobj = od[1].od_object; | |
4862 | blocksize = od[0].od_blocksize; | |
4863 | chunksize = blocksize; | |
4864 | ASSERT(chunksize == od[1].od_gen); | |
9babb374 | 4865 | |
428870ff BB |
4866 | VERIFY(dmu_object_info(os, bigobj, &doi) == 0); |
4867 | VERIFY(ISP2(doi.doi_data_block_size)); | |
4868 | VERIFY(chunksize == doi.doi_data_block_size); | |
4869 | VERIFY(chunksize >= 2 * sizeof (bufwad_t)); | |
9babb374 BB |
4870 | |
4871 | /* | |
4872 | * Pick a random index and compute the offsets into packobj and bigobj. | |
4873 | */ | |
4874 | n = ztest_random(regions) * stride + ztest_random(width); | |
4875 | s = 1 + ztest_random(width - 1); | |
4876 | ||
4877 | packoff = n * sizeof (bufwad_t); | |
4878 | packsize = s * sizeof (bufwad_t); | |
4879 | ||
428870ff BB |
4880 | bigoff = n * chunksize; |
4881 | bigsize = s * chunksize; | |
9babb374 BB |
4882 | |
4883 | packbuf = umem_zalloc(packsize, UMEM_NOFAIL); | |
4884 | bigbuf = umem_zalloc(bigsize, UMEM_NOFAIL); | |
4885 | ||
428870ff | 4886 | VERIFY3U(0, ==, dmu_bonus_hold(os, bigobj, FTAG, &bonus_db)); |
9babb374 BB |
4887 | |
4888 | bigbuf_arcbufs = umem_zalloc(2 * s * sizeof (arc_buf_t *), UMEM_NOFAIL); | |
4889 | ||
4890 | /* | |
4891 | * Iteration 0 test zcopy for DB_UNCACHED dbufs. | |
4892 | * Iteration 1 test zcopy to already referenced dbufs. | |
4893 | * Iteration 2 test zcopy to dirty dbuf in the same txg. | |
4894 | * Iteration 3 test zcopy to dbuf dirty in previous txg. | |
4895 | * Iteration 4 test zcopy when dbuf is no longer dirty. | |
4896 | * Iteration 5 test zcopy when it can't be done. | |
4897 | * Iteration 6 one more zcopy write. | |
4898 | */ | |
4899 | for (i = 0; i < 7; i++) { | |
4900 | uint64_t j; | |
4901 | uint64_t off; | |
4902 | ||
4903 | /* | |
4904 | * In iteration 5 (i == 5) use arcbufs | |
4905 | * that don't match bigobj blksz to test | |
440a3eb9 | 4906 | * dmu_assign_arcbuf_by_dbuf() when it can't directly |
9babb374 BB |
4907 | * assign an arcbuf to a dbuf. |
4908 | */ | |
4909 | for (j = 0; j < s; j++) { | |
b9541d6b | 4910 | if (i != 5 || chunksize < (SPA_MINBLOCKSIZE * 2)) { |
9babb374 | 4911 | bigbuf_arcbufs[j] = |
428870ff | 4912 | dmu_request_arcbuf(bonus_db, chunksize); |
9babb374 BB |
4913 | } else { |
4914 | bigbuf_arcbufs[2 * j] = | |
428870ff | 4915 | dmu_request_arcbuf(bonus_db, chunksize / 2); |
9babb374 | 4916 | bigbuf_arcbufs[2 * j + 1] = |
428870ff | 4917 | dmu_request_arcbuf(bonus_db, chunksize / 2); |
9babb374 BB |
4918 | } |
4919 | } | |
4920 | ||
4921 | /* | |
4922 | * Get a tx for the mods to both packobj and bigobj. | |
4923 | */ | |
4924 | tx = dmu_tx_create(os); | |
4925 | ||
428870ff BB |
4926 | dmu_tx_hold_write(tx, packobj, packoff, packsize); |
4927 | dmu_tx_hold_write(tx, bigobj, bigoff, bigsize); | |
9babb374 | 4928 | |
428870ff BB |
4929 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); |
4930 | if (txg == 0) { | |
9babb374 BB |
4931 | umem_free(packbuf, packsize); |
4932 | umem_free(bigbuf, bigsize); | |
4933 | for (j = 0; j < s; j++) { | |
b9541d6b CW |
4934 | if (i != 5 || |
4935 | chunksize < (SPA_MINBLOCKSIZE * 2)) { | |
9babb374 BB |
4936 | dmu_return_arcbuf(bigbuf_arcbufs[j]); |
4937 | } else { | |
4938 | dmu_return_arcbuf( | |
4939 | bigbuf_arcbufs[2 * j]); | |
4940 | dmu_return_arcbuf( | |
4941 | bigbuf_arcbufs[2 * j + 1]); | |
4942 | } | |
4943 | } | |
4944 | umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *)); | |
40b84e7a | 4945 | umem_free(od, size); |
9babb374 BB |
4946 | dmu_buf_rele(bonus_db, FTAG); |
4947 | return; | |
4948 | } | |
4949 | ||
9babb374 BB |
4950 | /* |
4951 | * 50% of the time don't read objects in the 1st iteration to | |
440a3eb9 TC |
4952 | * test dmu_assign_arcbuf_by_dbuf() for the case when there are |
4953 | * no existing dbufs for the specified offsets. | |
9babb374 BB |
4954 | */ |
4955 | if (i != 0 || ztest_random(2) != 0) { | |
428870ff | 4956 | error = dmu_read(os, packobj, packoff, |
9babb374 | 4957 | packsize, packbuf, DMU_READ_PREFETCH); |
c99c9001 | 4958 | ASSERT0(error); |
428870ff | 4959 | error = dmu_read(os, bigobj, bigoff, bigsize, |
9babb374 | 4960 | bigbuf, DMU_READ_PREFETCH); |
c99c9001 | 4961 | ASSERT0(error); |
9babb374 BB |
4962 | } |
4963 | compare_and_update_pbbufs(s, packbuf, bigbuf, bigsize, | |
428870ff | 4964 | n, chunksize, txg); |
9babb374 BB |
4965 | |
4966 | /* | |
4967 | * We've verified all the old bufwads, and made new ones. | |
4968 | * Now write them out. | |
4969 | */ | |
428870ff | 4970 | dmu_write(os, packobj, packoff, packsize, packbuf, tx); |
c242c188 | 4971 | if (ztest_opts.zo_verbose >= 7) { |
9babb374 BB |
4972 | (void) printf("writing offset %llx size %llx" |
4973 | " txg %llx\n", | |
4974 | (u_longlong_t)bigoff, | |
4975 | (u_longlong_t)bigsize, | |
4976 | (u_longlong_t)txg); | |
4977 | } | |
428870ff | 4978 | for (off = bigoff, j = 0; j < s; j++, off += chunksize) { |
9babb374 | 4979 | dmu_buf_t *dbt; |
b9541d6b | 4980 | if (i != 5 || chunksize < (SPA_MINBLOCKSIZE * 2)) { |
9babb374 | 4981 | bcopy((caddr_t)bigbuf + (off - bigoff), |
428870ff | 4982 | bigbuf_arcbufs[j]->b_data, chunksize); |
9babb374 BB |
4983 | } else { |
4984 | bcopy((caddr_t)bigbuf + (off - bigoff), | |
4985 | bigbuf_arcbufs[2 * j]->b_data, | |
428870ff | 4986 | chunksize / 2); |
9babb374 | 4987 | bcopy((caddr_t)bigbuf + (off - bigoff) + |
428870ff | 4988 | chunksize / 2, |
9babb374 | 4989 | bigbuf_arcbufs[2 * j + 1]->b_data, |
428870ff | 4990 | chunksize / 2); |
9babb374 BB |
4991 | } |
4992 | ||
4993 | if (i == 1) { | |
428870ff BB |
4994 | VERIFY(dmu_buf_hold(os, bigobj, off, |
4995 | FTAG, &dbt, DMU_READ_NO_PREFETCH) == 0); | |
9babb374 | 4996 | } |
b9541d6b | 4997 | if (i != 5 || chunksize < (SPA_MINBLOCKSIZE * 2)) { |
305781da TC |
4998 | VERIFY0(dmu_assign_arcbuf_by_dbuf(bonus_db, |
4999 | off, bigbuf_arcbufs[j], tx)); | |
9babb374 | 5000 | } else { |
305781da TC |
5001 | VERIFY0(dmu_assign_arcbuf_by_dbuf(bonus_db, |
5002 | off, bigbuf_arcbufs[2 * j], tx)); | |
5003 | VERIFY0(dmu_assign_arcbuf_by_dbuf(bonus_db, | |
428870ff | 5004 | off + chunksize / 2, |
305781da | 5005 | bigbuf_arcbufs[2 * j + 1], tx)); |
9babb374 BB |
5006 | } |
5007 | if (i == 1) { | |
5008 | dmu_buf_rele(dbt, FTAG); | |
5009 | } | |
5010 | } | |
5011 | dmu_tx_commit(tx); | |
5012 | ||
5013 | /* | |
5014 | * Sanity check the stuff we just wrote. | |
5015 | */ | |
5016 | { | |
5017 | void *packcheck = umem_alloc(packsize, UMEM_NOFAIL); | |
5018 | void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL); | |
5019 | ||
b2255edc | 5020 | VERIFY0(dmu_read(os, packobj, packoff, |
9babb374 | 5021 | packsize, packcheck, DMU_READ_PREFETCH)); |
b2255edc | 5022 | VERIFY0(dmu_read(os, bigobj, bigoff, |
9babb374 BB |
5023 | bigsize, bigcheck, DMU_READ_PREFETCH)); |
5024 | ||
b2255edc BB |
5025 | ASSERT0(bcmp(packbuf, packcheck, packsize)); |
5026 | ASSERT0(bcmp(bigbuf, bigcheck, bigsize)); | |
9babb374 BB |
5027 | |
5028 | umem_free(packcheck, packsize); | |
5029 | umem_free(bigcheck, bigsize); | |
5030 | } | |
5031 | if (i == 2) { | |
1b939560 | 5032 | txg_wait_open(dmu_objset_pool(os), 0, B_TRUE); |
9babb374 BB |
5033 | } else if (i == 3) { |
5034 | txg_wait_synced(dmu_objset_pool(os), 0); | |
5035 | } | |
5036 | } | |
5037 | ||
5038 | dmu_buf_rele(bonus_db, FTAG); | |
5039 | umem_free(packbuf, packsize); | |
5040 | umem_free(bigbuf, bigsize); | |
5041 | umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *)); | |
40b84e7a | 5042 | umem_free(od, size); |
9babb374 BB |
5043 | } |
5044 | ||
428870ff | 5045 | /* ARGSUSED */ |
34dc7c2f | 5046 | void |
428870ff | 5047 | ztest_dmu_write_parallel(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 5048 | { |
40b84e7a BB |
5049 | ztest_od_t *od; |
5050 | ||
d1d7e268 | 5051 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
428870ff BB |
5052 | uint64_t offset = (1ULL << (ztest_random(20) + 43)) + |
5053 | (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
34dc7c2f BB |
5054 | |
5055 | /* | |
428870ff BB |
5056 | * Have multiple threads write to large offsets in an object |
5057 | * to verify that parallel writes to an object -- even to the | |
5058 | * same blocks within the object -- doesn't cause any trouble. | |
34dc7c2f | 5059 | */ |
50c957f7 | 5060 | ztest_od_init(od, ID_PARALLEL, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0); |
9babb374 | 5061 | |
40b84e7a | 5062 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) |
34dc7c2f | 5063 | return; |
34dc7c2f | 5064 | |
428870ff | 5065 | while (ztest_random(10) != 0) |
40b84e7a BB |
5066 | ztest_io(zd, od->od_object, offset); |
5067 | ||
d1d7e268 | 5068 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 5069 | } |
34dc7c2f | 5070 | |
428870ff BB |
5071 | void |
5072 | ztest_dmu_prealloc(ztest_ds_t *zd, uint64_t id) | |
5073 | { | |
40b84e7a | 5074 | ztest_od_t *od; |
428870ff BB |
5075 | uint64_t offset = (1ULL << (ztest_random(4) + SPA_MAXBLOCKSHIFT)) + |
5076 | (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
5077 | uint64_t count = ztest_random(20) + 1; | |
5078 | uint64_t blocksize = ztest_random_blocksize(); | |
5079 | void *data; | |
34dc7c2f | 5080 | |
d1d7e268 | 5081 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
34dc7c2f | 5082 | |
50c957f7 | 5083 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0, 0); |
40b84e7a | 5084 | |
d1d7e268 MK |
5085 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
5086 | !ztest_random(2)) != 0) { | |
5087 | umem_free(od, sizeof (ztest_od_t)); | |
34dc7c2f | 5088 | return; |
40b84e7a | 5089 | } |
34dc7c2f | 5090 | |
40b84e7a | 5091 | if (ztest_truncate(zd, od->od_object, offset, count * blocksize) != 0) { |
d1d7e268 | 5092 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f | 5093 | return; |
40b84e7a | 5094 | } |
34dc7c2f | 5095 | |
40b84e7a | 5096 | ztest_prealloc(zd, od->od_object, offset, count * blocksize); |
34dc7c2f | 5097 | |
428870ff | 5098 | data = umem_zalloc(blocksize, UMEM_NOFAIL); |
34dc7c2f | 5099 | |
428870ff BB |
5100 | while (ztest_random(count) != 0) { |
5101 | uint64_t randoff = offset + (ztest_random(count) * blocksize); | |
40b84e7a | 5102 | if (ztest_write(zd, od->od_object, randoff, blocksize, |
428870ff BB |
5103 | data) != 0) |
5104 | break; | |
5105 | while (ztest_random(4) != 0) | |
40b84e7a | 5106 | ztest_io(zd, od->od_object, randoff); |
9babb374 | 5107 | } |
34dc7c2f | 5108 | |
428870ff | 5109 | umem_free(data, blocksize); |
d1d7e268 | 5110 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f BB |
5111 | } |
5112 | ||
5113 | /* | |
5114 | * Verify that zap_{create,destroy,add,remove,update} work as expected. | |
5115 | */ | |
5116 | #define ZTEST_ZAP_MIN_INTS 1 | |
5117 | #define ZTEST_ZAP_MAX_INTS 4 | |
5118 | #define ZTEST_ZAP_MAX_PROPS 1000 | |
5119 | ||
5120 | void | |
428870ff | 5121 | ztest_zap(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 5122 | { |
428870ff | 5123 | objset_t *os = zd->zd_os; |
40b84e7a | 5124 | ztest_od_t *od; |
34dc7c2f BB |
5125 | uint64_t object; |
5126 | uint64_t txg, last_txg; | |
5127 | uint64_t value[ZTEST_ZAP_MAX_INTS]; | |
5128 | uint64_t zl_ints, zl_intsize, prop; | |
5129 | int i, ints; | |
5130 | dmu_tx_t *tx; | |
5131 | char propname[100], txgname[100]; | |
5132 | int error; | |
34dc7c2f BB |
5133 | char *hc[2] = { "s.acl.h", ".s.open.h.hyLZlg" }; |
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); |
34dc7c2f | 5137 | |
40b84e7a | 5138 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
02730c33 | 5139 | !ztest_random(2)) != 0) |
40b84e7a | 5140 | goto out; |
34dc7c2f | 5141 | |
40b84e7a | 5142 | object = od->od_object; |
34dc7c2f | 5143 | |
428870ff BB |
5144 | /* |
5145 | * Generate a known hash collision, and verify that | |
5146 | * we can lookup and remove both entries. | |
5147 | */ | |
5148 | tx = dmu_tx_create(os); | |
5149 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); | |
5150 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
5151 | if (txg == 0) | |
40b84e7a | 5152 | goto out; |
428870ff BB |
5153 | for (i = 0; i < 2; i++) { |
5154 | value[i] = i; | |
5155 | VERIFY3U(0, ==, zap_add(os, object, hc[i], sizeof (uint64_t), | |
5156 | 1, &value[i], tx)); | |
5157 | } | |
5158 | for (i = 0; i < 2; i++) { | |
5159 | VERIFY3U(EEXIST, ==, zap_add(os, object, hc[i], | |
5160 | sizeof (uint64_t), 1, &value[i], tx)); | |
5161 | VERIFY3U(0, ==, | |
5162 | zap_length(os, object, hc[i], &zl_intsize, &zl_ints)); | |
5163 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
5164 | ASSERT3U(zl_ints, ==, 1); | |
5165 | } | |
5166 | for (i = 0; i < 2; i++) { | |
5167 | VERIFY3U(0, ==, zap_remove(os, object, hc[i], tx)); | |
34dc7c2f | 5168 | } |
428870ff | 5169 | dmu_tx_commit(tx); |
34dc7c2f | 5170 | |
428870ff | 5171 | /* |
d611989f | 5172 | * Generate a bunch of random entries. |
428870ff | 5173 | */ |
34dc7c2f BB |
5174 | ints = MAX(ZTEST_ZAP_MIN_INTS, object % ZTEST_ZAP_MAX_INTS); |
5175 | ||
5176 | prop = ztest_random(ZTEST_ZAP_MAX_PROPS); | |
5177 | (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop); | |
5178 | (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop); | |
5179 | bzero(value, sizeof (value)); | |
5180 | last_txg = 0; | |
5181 | ||
5182 | /* | |
5183 | * If these zap entries already exist, validate their contents. | |
5184 | */ | |
5185 | error = zap_length(os, object, txgname, &zl_intsize, &zl_ints); | |
5186 | if (error == 0) { | |
5187 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
5188 | ASSERT3U(zl_ints, ==, 1); | |
5189 | ||
5190 | VERIFY(zap_lookup(os, object, txgname, zl_intsize, | |
5191 | zl_ints, &last_txg) == 0); | |
5192 | ||
5193 | VERIFY(zap_length(os, object, propname, &zl_intsize, | |
5194 | &zl_ints) == 0); | |
5195 | ||
5196 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
5197 | ASSERT3U(zl_ints, ==, ints); | |
5198 | ||
5199 | VERIFY(zap_lookup(os, object, propname, zl_intsize, | |
5200 | zl_ints, value) == 0); | |
5201 | ||
5202 | for (i = 0; i < ints; i++) { | |
5203 | ASSERT3U(value[i], ==, last_txg + object + i); | |
5204 | } | |
5205 | } else { | |
5206 | ASSERT3U(error, ==, ENOENT); | |
5207 | } | |
5208 | ||
5209 | /* | |
5210 | * Atomically update two entries in our zap object. | |
5211 | * The first is named txg_%llu, and contains the txg | |
5212 | * in which the property was last updated. The second | |
5213 | * is named prop_%llu, and the nth element of its value | |
5214 | * should be txg + object + n. | |
5215 | */ | |
5216 | tx = dmu_tx_create(os); | |
428870ff BB |
5217 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
5218 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
5219 | if (txg == 0) | |
40b84e7a | 5220 | goto out; |
34dc7c2f BB |
5221 | |
5222 | if (last_txg > txg) | |
5223 | fatal(0, "zap future leak: old %llu new %llu", last_txg, txg); | |
5224 | ||
5225 | for (i = 0; i < ints; i++) | |
5226 | value[i] = txg + object + i; | |
5227 | ||
428870ff BB |
5228 | VERIFY3U(0, ==, zap_update(os, object, txgname, sizeof (uint64_t), |
5229 | 1, &txg, tx)); | |
5230 | VERIFY3U(0, ==, zap_update(os, object, propname, sizeof (uint64_t), | |
5231 | ints, value, tx)); | |
34dc7c2f BB |
5232 | |
5233 | dmu_tx_commit(tx); | |
5234 | ||
5235 | /* | |
5236 | * Remove a random pair of entries. | |
5237 | */ | |
5238 | prop = ztest_random(ZTEST_ZAP_MAX_PROPS); | |
5239 | (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop); | |
5240 | (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop); | |
5241 | ||
5242 | error = zap_length(os, object, txgname, &zl_intsize, &zl_ints); | |
5243 | ||
5244 | if (error == ENOENT) | |
40b84e7a | 5245 | goto out; |
34dc7c2f | 5246 | |
c99c9001 | 5247 | ASSERT0(error); |
34dc7c2f BB |
5248 | |
5249 | tx = dmu_tx_create(os); | |
428870ff BB |
5250 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
5251 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
5252 | if (txg == 0) | |
40b84e7a | 5253 | goto out; |
428870ff BB |
5254 | VERIFY3U(0, ==, zap_remove(os, object, txgname, tx)); |
5255 | VERIFY3U(0, ==, zap_remove(os, object, propname, tx)); | |
5256 | dmu_tx_commit(tx); | |
40b84e7a | 5257 | out: |
d1d7e268 | 5258 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 5259 | } |
34dc7c2f | 5260 | |
428870ff | 5261 | /* |
d611989f | 5262 | * Test case to test the upgrading of a microzap to fatzap. |
428870ff BB |
5263 | */ |
5264 | void | |
5265 | ztest_fzap(ztest_ds_t *zd, uint64_t id) | |
5266 | { | |
5267 | objset_t *os = zd->zd_os; | |
40b84e7a | 5268 | ztest_od_t *od; |
428870ff | 5269 | uint64_t object, txg; |
d6320ddb | 5270 | int i; |
34dc7c2f | 5271 | |
d1d7e268 | 5272 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 5273 | ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0, 0); |
428870ff | 5274 | |
40b84e7a | 5275 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
02730c33 | 5276 | !ztest_random(2)) != 0) |
40b84e7a BB |
5277 | goto out; |
5278 | object = od->od_object; | |
34dc7c2f BB |
5279 | |
5280 | /* | |
428870ff BB |
5281 | * Add entries to this ZAP and make sure it spills over |
5282 | * and gets upgraded to a fatzap. Also, since we are adding | |
5283 | * 2050 entries we should see ptrtbl growth and leaf-block split. | |
34dc7c2f | 5284 | */ |
d6320ddb | 5285 | for (i = 0; i < 2050; i++) { |
eca7b760 | 5286 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
5287 | uint64_t value = i; |
5288 | dmu_tx_t *tx; | |
5289 | int error; | |
34dc7c2f | 5290 | |
428870ff | 5291 | (void) snprintf(name, sizeof (name), "fzap-%llu-%llu", |
b8864a23 | 5292 | (u_longlong_t)id, (u_longlong_t)value); |
428870ff BB |
5293 | |
5294 | tx = dmu_tx_create(os); | |
5295 | dmu_tx_hold_zap(tx, object, B_TRUE, name); | |
5296 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
5297 | if (txg == 0) | |
40b84e7a | 5298 | goto out; |
428870ff BB |
5299 | error = zap_add(os, object, name, sizeof (uint64_t), 1, |
5300 | &value, tx); | |
5301 | ASSERT(error == 0 || error == EEXIST); | |
5302 | dmu_tx_commit(tx); | |
34dc7c2f | 5303 | } |
40b84e7a | 5304 | out: |
d1d7e268 | 5305 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f BB |
5306 | } |
5307 | ||
428870ff | 5308 | /* ARGSUSED */ |
34dc7c2f | 5309 | void |
428870ff | 5310 | ztest_zap_parallel(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 5311 | { |
428870ff | 5312 | objset_t *os = zd->zd_os; |
40b84e7a | 5313 | ztest_od_t *od; |
34dc7c2f BB |
5314 | uint64_t txg, object, count, wsize, wc, zl_wsize, zl_wc; |
5315 | dmu_tx_t *tx; | |
5316 | int i, namelen, error; | |
428870ff | 5317 | int micro = ztest_random(2); |
34dc7c2f BB |
5318 | char name[20], string_value[20]; |
5319 | void *data; | |
5320 | ||
d1d7e268 | 5321 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 5322 | ztest_od_init(od, ID_PARALLEL, FTAG, micro, DMU_OT_ZAP_OTHER, 0, 0, 0); |
428870ff | 5323 | |
40b84e7a | 5324 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
d1d7e268 | 5325 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 5326 | return; |
40b84e7a | 5327 | } |
428870ff | 5328 | |
40b84e7a | 5329 | object = od->od_object; |
428870ff | 5330 | |
34dc7c2f BB |
5331 | /* |
5332 | * Generate a random name of the form 'xxx.....' where each | |
5333 | * x is a random printable character and the dots are dots. | |
5334 | * There are 94 such characters, and the name length goes from | |
5335 | * 6 to 20, so there are 94^3 * 15 = 12,458,760 possible names. | |
5336 | */ | |
5337 | namelen = ztest_random(sizeof (name) - 5) + 5 + 1; | |
5338 | ||
5339 | for (i = 0; i < 3; i++) | |
5340 | name[i] = '!' + ztest_random('~' - '!' + 1); | |
5341 | for (; i < namelen - 1; i++) | |
5342 | name[i] = '.'; | |
5343 | name[i] = '\0'; | |
5344 | ||
428870ff | 5345 | if ((namelen & 1) || micro) { |
34dc7c2f BB |
5346 | wsize = sizeof (txg); |
5347 | wc = 1; | |
5348 | data = &txg; | |
5349 | } else { | |
5350 | wsize = 1; | |
5351 | wc = namelen; | |
5352 | data = string_value; | |
5353 | } | |
5354 | ||
5355 | count = -1ULL; | |
13fe0198 | 5356 | VERIFY0(zap_count(os, object, &count)); |
34dc7c2f BB |
5357 | ASSERT(count != -1ULL); |
5358 | ||
5359 | /* | |
5360 | * Select an operation: length, lookup, add, update, remove. | |
5361 | */ | |
5362 | i = ztest_random(5); | |
5363 | ||
5364 | if (i >= 2) { | |
5365 | tx = dmu_tx_create(os); | |
428870ff BB |
5366 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
5367 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
df053d67 GN |
5368 | if (txg == 0) { |
5369 | umem_free(od, sizeof (ztest_od_t)); | |
34dc7c2f | 5370 | return; |
df053d67 | 5371 | } |
34dc7c2f BB |
5372 | bcopy(name, string_value, namelen); |
5373 | } else { | |
5374 | tx = NULL; | |
5375 | txg = 0; | |
5376 | bzero(string_value, namelen); | |
5377 | } | |
5378 | ||
5379 | switch (i) { | |
5380 | ||
5381 | case 0: | |
5382 | error = zap_length(os, object, name, &zl_wsize, &zl_wc); | |
5383 | if (error == 0) { | |
5384 | ASSERT3U(wsize, ==, zl_wsize); | |
5385 | ASSERT3U(wc, ==, zl_wc); | |
5386 | } else { | |
5387 | ASSERT3U(error, ==, ENOENT); | |
5388 | } | |
5389 | break; | |
5390 | ||
5391 | case 1: | |
5392 | error = zap_lookup(os, object, name, wsize, wc, data); | |
5393 | if (error == 0) { | |
5394 | if (data == string_value && | |
5395 | bcmp(name, data, namelen) != 0) | |
5396 | fatal(0, "name '%s' != val '%s' len %d", | |
5397 | name, data, namelen); | |
5398 | } else { | |
5399 | ASSERT3U(error, ==, ENOENT); | |
5400 | } | |
5401 | break; | |
5402 | ||
5403 | case 2: | |
5404 | error = zap_add(os, object, name, wsize, wc, data, tx); | |
5405 | ASSERT(error == 0 || error == EEXIST); | |
5406 | break; | |
5407 | ||
5408 | case 3: | |
5409 | VERIFY(zap_update(os, object, name, wsize, wc, data, tx) == 0); | |
5410 | break; | |
5411 | ||
5412 | case 4: | |
5413 | error = zap_remove(os, object, name, tx); | |
5414 | ASSERT(error == 0 || error == ENOENT); | |
5415 | break; | |
5416 | } | |
5417 | ||
5418 | if (tx != NULL) | |
5419 | dmu_tx_commit(tx); | |
40b84e7a | 5420 | |
d1d7e268 | 5421 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f BB |
5422 | } |
5423 | ||
428870ff BB |
5424 | /* |
5425 | * Commit callback data. | |
5426 | */ | |
5427 | typedef struct ztest_cb_data { | |
5428 | list_node_t zcd_node; | |
5429 | uint64_t zcd_txg; | |
5430 | int zcd_expected_err; | |
5431 | boolean_t zcd_added; | |
5432 | boolean_t zcd_called; | |
5433 | spa_t *zcd_spa; | |
5434 | } ztest_cb_data_t; | |
5435 | ||
5436 | /* This is the actual commit callback function */ | |
5437 | static void | |
5438 | ztest_commit_callback(void *arg, int error) | |
5439 | { | |
5440 | ztest_cb_data_t *data = arg; | |
5441 | uint64_t synced_txg; | |
5442 | ||
5443 | VERIFY(data != NULL); | |
5444 | VERIFY3S(data->zcd_expected_err, ==, error); | |
5445 | VERIFY(!data->zcd_called); | |
5446 | ||
5447 | synced_txg = spa_last_synced_txg(data->zcd_spa); | |
5448 | if (data->zcd_txg > synced_txg) | |
5449 | fatal(0, "commit callback of txg %" PRIu64 " called prematurely" | |
5450 | ", last synced txg = %" PRIu64 "\n", data->zcd_txg, | |
5451 | synced_txg); | |
5452 | ||
5453 | data->zcd_called = B_TRUE; | |
5454 | ||
5455 | if (error == ECANCELED) { | |
c99c9001 | 5456 | ASSERT0(data->zcd_txg); |
428870ff BB |
5457 | ASSERT(!data->zcd_added); |
5458 | ||
5459 | /* | |
5460 | * The private callback data should be destroyed here, but | |
5461 | * since we are going to check the zcd_called field after | |
5462 | * dmu_tx_abort(), we will destroy it there. | |
5463 | */ | |
5464 | return; | |
5465 | } | |
5466 | ||
090ff092 | 5467 | ASSERT(data->zcd_added); |
428870ff BB |
5468 | ASSERT3U(data->zcd_txg, !=, 0); |
5469 | ||
1e33ac1e | 5470 | (void) mutex_enter(&zcl.zcl_callbacks_lock); |
090ff092 RC |
5471 | |
5472 | /* See if this cb was called more quickly */ | |
5473 | if ((synced_txg - data->zcd_txg) < zc_min_txg_delay) | |
5474 | zc_min_txg_delay = synced_txg - data->zcd_txg; | |
5475 | ||
5476 | /* Remove our callback from the list */ | |
428870ff | 5477 | list_remove(&zcl.zcl_callbacks, data); |
090ff092 | 5478 | |
1e33ac1e | 5479 | (void) mutex_exit(&zcl.zcl_callbacks_lock); |
428870ff | 5480 | |
428870ff BB |
5481 | umem_free(data, sizeof (ztest_cb_data_t)); |
5482 | } | |
5483 | ||
5484 | /* Allocate and initialize callback data structure */ | |
5485 | static ztest_cb_data_t * | |
5486 | ztest_create_cb_data(objset_t *os, uint64_t txg) | |
5487 | { | |
5488 | ztest_cb_data_t *cb_data; | |
5489 | ||
5490 | cb_data = umem_zalloc(sizeof (ztest_cb_data_t), UMEM_NOFAIL); | |
5491 | ||
5492 | cb_data->zcd_txg = txg; | |
5493 | cb_data->zcd_spa = dmu_objset_spa(os); | |
1e33ac1e | 5494 | list_link_init(&cb_data->zcd_node); |
428870ff BB |
5495 | |
5496 | return (cb_data); | |
5497 | } | |
5498 | ||
428870ff BB |
5499 | /* |
5500 | * Commit callback test. | |
5501 | */ | |
34dc7c2f | 5502 | void |
428870ff BB |
5503 | ztest_dmu_commit_callbacks(ztest_ds_t *zd, uint64_t id) |
5504 | { | |
5505 | objset_t *os = zd->zd_os; | |
40b84e7a | 5506 | ztest_od_t *od; |
428870ff BB |
5507 | dmu_tx_t *tx; |
5508 | ztest_cb_data_t *cb_data[3], *tmp_cb; | |
5509 | uint64_t old_txg, txg; | |
090ff092 | 5510 | int i, error = 0; |
428870ff | 5511 | |
d1d7e268 | 5512 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 5513 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0); |
428870ff | 5514 | |
40b84e7a | 5515 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
d1d7e268 | 5516 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 5517 | return; |
40b84e7a | 5518 | } |
428870ff BB |
5519 | |
5520 | tx = dmu_tx_create(os); | |
5521 | ||
5522 | cb_data[0] = ztest_create_cb_data(os, 0); | |
5523 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[0]); | |
5524 | ||
40b84e7a | 5525 | dmu_tx_hold_write(tx, od->od_object, 0, sizeof (uint64_t)); |
428870ff BB |
5526 | |
5527 | /* Every once in a while, abort the transaction on purpose */ | |
5528 | if (ztest_random(100) == 0) | |
5529 | error = -1; | |
5530 | ||
5531 | if (!error) | |
5532 | error = dmu_tx_assign(tx, TXG_NOWAIT); | |
5533 | ||
5534 | txg = error ? 0 : dmu_tx_get_txg(tx); | |
5535 | ||
5536 | cb_data[0]->zcd_txg = txg; | |
5537 | cb_data[1] = ztest_create_cb_data(os, txg); | |
5538 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[1]); | |
5539 | ||
5540 | if (error) { | |
5541 | /* | |
5542 | * It's not a strict requirement to call the registered | |
5543 | * callbacks from inside dmu_tx_abort(), but that's what | |
5544 | * it's supposed to happen in the current implementation | |
5545 | * so we will check for that. | |
5546 | */ | |
5547 | for (i = 0; i < 2; i++) { | |
5548 | cb_data[i]->zcd_expected_err = ECANCELED; | |
5549 | VERIFY(!cb_data[i]->zcd_called); | |
5550 | } | |
5551 | ||
5552 | dmu_tx_abort(tx); | |
5553 | ||
5554 | for (i = 0; i < 2; i++) { | |
5555 | VERIFY(cb_data[i]->zcd_called); | |
5556 | umem_free(cb_data[i], sizeof (ztest_cb_data_t)); | |
5557 | } | |
5558 | ||
d1d7e268 | 5559 | umem_free(od, sizeof (ztest_od_t)); |
428870ff BB |
5560 | return; |
5561 | } | |
5562 | ||
5563 | cb_data[2] = ztest_create_cb_data(os, txg); | |
5564 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[2]); | |
5565 | ||
5566 | /* | |
5567 | * Read existing data to make sure there isn't a future leak. | |
5568 | */ | |
40b84e7a | 5569 | VERIFY(0 == dmu_read(os, od->od_object, 0, sizeof (uint64_t), |
428870ff BB |
5570 | &old_txg, DMU_READ_PREFETCH)); |
5571 | ||
5572 | if (old_txg > txg) | |
5573 | fatal(0, "future leak: got %" PRIu64 ", open txg is %" PRIu64, | |
5574 | old_txg, txg); | |
5575 | ||
40b84e7a | 5576 | dmu_write(os, od->od_object, 0, sizeof (uint64_t), &txg, tx); |
428870ff | 5577 | |
1e33ac1e | 5578 | (void) mutex_enter(&zcl.zcl_callbacks_lock); |
428870ff BB |
5579 | |
5580 | /* | |
5581 | * Since commit callbacks don't have any ordering requirement and since | |
5582 | * it is theoretically possible for a commit callback to be called | |
5583 | * after an arbitrary amount of time has elapsed since its txg has been | |
5584 | * synced, it is difficult to reliably determine whether a commit | |
5585 | * callback hasn't been called due to high load or due to a flawed | |
5586 | * implementation. | |
5587 | * | |
5588 | * In practice, we will assume that if after a certain number of txgs a | |
5589 | * commit callback hasn't been called, then most likely there's an | |
5590 | * implementation bug.. | |
5591 | */ | |
5592 | tmp_cb = list_head(&zcl.zcl_callbacks); | |
5593 | if (tmp_cb != NULL && | |
090ff092 | 5594 | tmp_cb->zcd_txg + ZTEST_COMMIT_CB_THRESH < txg) { |
428870ff BB |
5595 | fatal(0, "Commit callback threshold exceeded, oldest txg: %" |
5596 | PRIu64 ", open txg: %" PRIu64 "\n", tmp_cb->zcd_txg, txg); | |
5597 | } | |
5598 | ||
5599 | /* | |
5600 | * Let's find the place to insert our callbacks. | |
5601 | * | |
5602 | * Even though the list is ordered by txg, it is possible for the | |
5603 | * insertion point to not be the end because our txg may already be | |
5604 | * quiescing at this point and other callbacks in the open txg | |
5605 | * (from other objsets) may have sneaked in. | |
5606 | */ | |
5607 | tmp_cb = list_tail(&zcl.zcl_callbacks); | |
5608 | while (tmp_cb != NULL && tmp_cb->zcd_txg > txg) | |
5609 | tmp_cb = list_prev(&zcl.zcl_callbacks, tmp_cb); | |
5610 | ||
5611 | /* Add the 3 callbacks to the list */ | |
5612 | for (i = 0; i < 3; i++) { | |
5613 | if (tmp_cb == NULL) | |
5614 | list_insert_head(&zcl.zcl_callbacks, cb_data[i]); | |
5615 | else | |
5616 | list_insert_after(&zcl.zcl_callbacks, tmp_cb, | |
5617 | cb_data[i]); | |
5618 | ||
5619 | cb_data[i]->zcd_added = B_TRUE; | |
5620 | VERIFY(!cb_data[i]->zcd_called); | |
5621 | ||
5622 | tmp_cb = cb_data[i]; | |
5623 | } | |
5624 | ||
090ff092 RC |
5625 | zc_cb_counter += 3; |
5626 | ||
1e33ac1e | 5627 | (void) mutex_exit(&zcl.zcl_callbacks_lock); |
428870ff BB |
5628 | |
5629 | dmu_tx_commit(tx); | |
40b84e7a | 5630 | |
d1d7e268 | 5631 | umem_free(od, sizeof (ztest_od_t)); |
428870ff BB |
5632 | } |
5633 | ||
50c957f7 NB |
5634 | /* |
5635 | * Visit each object in the dataset. Verify that its properties | |
5636 | * are consistent what was stored in the block tag when it was created, | |
5637 | * and that its unused bonus buffer space has not been overwritten. | |
5638 | */ | |
817b1b6e | 5639 | /* ARGSUSED */ |
50c957f7 NB |
5640 | void |
5641 | ztest_verify_dnode_bt(ztest_ds_t *zd, uint64_t id) | |
5642 | { | |
5643 | objset_t *os = zd->zd_os; | |
5644 | uint64_t obj; | |
5645 | int err = 0; | |
5646 | ||
5647 | for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) { | |
5648 | ztest_block_tag_t *bt = NULL; | |
5649 | dmu_object_info_t doi; | |
5650 | dmu_buf_t *db; | |
5651 | ||
06401e42 BB |
5652 | ztest_object_lock(zd, obj, RL_READER); |
5653 | if (dmu_bonus_hold(os, obj, FTAG, &db) != 0) { | |
5654 | ztest_object_unlock(zd, obj); | |
50c957f7 | 5655 | continue; |
06401e42 | 5656 | } |
50c957f7 NB |
5657 | |
5658 | dmu_object_info_from_db(db, &doi); | |
5659 | if (doi.doi_bonus_size >= sizeof (*bt)) | |
5660 | bt = ztest_bt_bonus(db); | |
5661 | ||
5662 | if (bt && bt->bt_magic == BT_MAGIC) { | |
5663 | ztest_bt_verify(bt, os, obj, doi.doi_dnodesize, | |
5664 | bt->bt_offset, bt->bt_gen, bt->bt_txg, | |
5665 | bt->bt_crtxg); | |
5666 | ztest_verify_unused_bonus(db, bt, obj, os, bt->bt_gen); | |
5667 | } | |
5668 | ||
5669 | dmu_buf_rele(db, FTAG); | |
06401e42 | 5670 | ztest_object_unlock(zd, obj); |
50c957f7 NB |
5671 | } |
5672 | } | |
5673 | ||
428870ff BB |
5674 | /* ARGSUSED */ |
5675 | void | |
5676 | ztest_dsl_prop_get_set(ztest_ds_t *zd, uint64_t id) | |
5677 | { | |
5678 | zfs_prop_t proplist[] = { | |
5679 | ZFS_PROP_CHECKSUM, | |
5680 | ZFS_PROP_COMPRESSION, | |
5681 | ZFS_PROP_COPIES, | |
5682 | ZFS_PROP_DEDUP | |
5683 | }; | |
d6320ddb | 5684 | int p; |
428870ff | 5685 | |
2d9142c9 | 5686 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
428870ff | 5687 | |
d6320ddb | 5688 | for (p = 0; p < sizeof (proplist) / sizeof (proplist[0]); p++) |
428870ff BB |
5689 | (void) ztest_dsl_prop_set_uint64(zd->zd_name, proplist[p], |
5690 | ztest_random_dsl_prop(proplist[p]), (int)ztest_random(2)); | |
5691 | ||
76d52067 BB |
5692 | VERIFY0(ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_RECORDSIZE, |
5693 | ztest_random_blocksize(), (int)ztest_random(2))); | |
5694 | ||
2d9142c9 | 5695 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
428870ff BB |
5696 | } |
5697 | ||
5698 | /* ARGSUSED */ | |
5699 | void | |
5700 | ztest_spa_prop_get_set(ztest_ds_t *zd, uint64_t id) | |
5701 | { | |
428870ff BB |
5702 | nvlist_t *props = NULL; |
5703 | ||
2d9142c9 | 5704 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
428870ff | 5705 | |
1b939560 BB |
5706 | (void) ztest_spa_prop_set_uint64(ZPOOL_PROP_AUTOTRIM, ztest_random(2)); |
5707 | ||
c99c9001 | 5708 | VERIFY0(spa_prop_get(ztest_spa, &props)); |
428870ff | 5709 | |
c242c188 | 5710 | if (ztest_opts.zo_verbose >= 6) |
428870ff BB |
5711 | dump_nvlist(props, 4); |
5712 | ||
5713 | nvlist_free(props); | |
5714 | ||
2d9142c9 | 5715 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
428870ff BB |
5716 | } |
5717 | ||
13fe0198 MA |
5718 | static int |
5719 | user_release_one(const char *snapname, const char *holdname) | |
5720 | { | |
5721 | nvlist_t *snaps, *holds; | |
5722 | int error; | |
5723 | ||
5724 | snaps = fnvlist_alloc(); | |
5725 | holds = fnvlist_alloc(); | |
5726 | fnvlist_add_boolean(holds, holdname); | |
5727 | fnvlist_add_nvlist(snaps, snapname, holds); | |
5728 | fnvlist_free(holds); | |
5729 | error = dsl_dataset_user_release(snaps, NULL); | |
5730 | fnvlist_free(snaps); | |
5731 | return (error); | |
5732 | } | |
5733 | ||
428870ff BB |
5734 | /* |
5735 | * Test snapshot hold/release and deferred destroy. | |
5736 | */ | |
5737 | void | |
5738 | ztest_dmu_snapshot_hold(ztest_ds_t *zd, uint64_t id) | |
34dc7c2f | 5739 | { |
34dc7c2f | 5740 | int error; |
428870ff BB |
5741 | objset_t *os = zd->zd_os; |
5742 | objset_t *origin; | |
5743 | char snapname[100]; | |
5744 | char fullname[100]; | |
5745 | char clonename[100]; | |
5746 | char tag[100]; | |
eca7b760 | 5747 | char osname[ZFS_MAX_DATASET_NAME_LEN]; |
13fe0198 | 5748 | nvlist_t *holds; |
34dc7c2f | 5749 | |
2d9142c9 | 5750 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
34dc7c2f BB |
5751 | |
5752 | dmu_objset_name(os, osname); | |
5753 | ||
d1d7e268 MK |
5754 | (void) snprintf(snapname, sizeof (snapname), "sh1_%llu", |
5755 | (u_longlong_t)id); | |
13fe0198 MA |
5756 | (void) snprintf(fullname, sizeof (fullname), "%s@%s", osname, snapname); |
5757 | (void) snprintf(clonename, sizeof (clonename), | |
d1d7e268 MK |
5758 | "%s/ch1_%llu", osname, (u_longlong_t)id); |
5759 | (void) snprintf(tag, sizeof (tag), "tag_%llu", (u_longlong_t)id); | |
428870ff BB |
5760 | |
5761 | /* | |
5762 | * Clean up from any previous run. | |
5763 | */ | |
13fe0198 MA |
5764 | error = dsl_destroy_head(clonename); |
5765 | if (error != ENOENT) | |
5766 | ASSERT0(error); | |
5767 | error = user_release_one(fullname, tag); | |
5768 | if (error != ESRCH && error != ENOENT) | |
5769 | ASSERT0(error); | |
5770 | error = dsl_destroy_snapshot(fullname, B_FALSE); | |
5771 | if (error != ENOENT) | |
5772 | ASSERT0(error); | |
428870ff BB |
5773 | |
5774 | /* | |
5775 | * Create snapshot, clone it, mark snap for deferred destroy, | |
5776 | * destroy clone, verify snap was also destroyed. | |
5777 | */ | |
6f1ffb06 | 5778 | error = dmu_objset_snapshot_one(osname, snapname); |
428870ff BB |
5779 | if (error) { |
5780 | if (error == ENOSPC) { | |
5781 | ztest_record_enospc("dmu_objset_snapshot"); | |
5782 | goto out; | |
34dc7c2f | 5783 | } |
428870ff BB |
5784 | fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error); |
5785 | } | |
34dc7c2f | 5786 | |
13fe0198 | 5787 | error = dmu_objset_clone(clonename, fullname); |
428870ff | 5788 | if (error) { |
34dc7c2f | 5789 | if (error == ENOSPC) { |
428870ff BB |
5790 | ztest_record_enospc("dmu_objset_clone"); |
5791 | goto out; | |
34dc7c2f | 5792 | } |
428870ff BB |
5793 | fatal(0, "dmu_objset_clone(%s) = %d", clonename, error); |
5794 | } | |
34dc7c2f | 5795 | |
13fe0198 | 5796 | error = dsl_destroy_snapshot(fullname, B_TRUE); |
428870ff | 5797 | if (error) { |
13fe0198 | 5798 | fatal(0, "dsl_destroy_snapshot(%s, B_TRUE) = %d", |
428870ff BB |
5799 | fullname, error); |
5800 | } | |
34dc7c2f | 5801 | |
13fe0198 | 5802 | error = dsl_destroy_head(clonename); |
428870ff | 5803 | if (error) |
13fe0198 | 5804 | fatal(0, "dsl_destroy_head(%s) = %d", clonename, error); |
34dc7c2f | 5805 | |
428870ff BB |
5806 | error = dmu_objset_hold(fullname, FTAG, &origin); |
5807 | if (error != ENOENT) | |
5808 | fatal(0, "dmu_objset_hold(%s) = %d", fullname, error); | |
34dc7c2f | 5809 | |
428870ff BB |
5810 | /* |
5811 | * Create snapshot, add temporary hold, verify that we can't | |
5812 | * destroy a held snapshot, mark for deferred destroy, | |
5813 | * release hold, verify snapshot was destroyed. | |
5814 | */ | |
6f1ffb06 | 5815 | error = dmu_objset_snapshot_one(osname, snapname); |
428870ff BB |
5816 | if (error) { |
5817 | if (error == ENOSPC) { | |
5818 | ztest_record_enospc("dmu_objset_snapshot"); | |
5819 | goto out; | |
34dc7c2f | 5820 | } |
428870ff BB |
5821 | fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error); |
5822 | } | |
5823 | ||
13fe0198 MA |
5824 | holds = fnvlist_alloc(); |
5825 | fnvlist_add_string(holds, fullname, tag); | |
5826 | error = dsl_dataset_user_hold(holds, 0, NULL); | |
5827 | fnvlist_free(holds); | |
5828 | ||
9b67f605 MA |
5829 | if (error == ENOSPC) { |
5830 | ztest_record_enospc("dsl_dataset_user_hold"); | |
5831 | goto out; | |
5832 | } else if (error) { | |
5833 | fatal(0, "dsl_dataset_user_hold(%s, %s) = %u", | |
5834 | fullname, tag, error); | |
5835 | } | |
428870ff | 5836 | |
13fe0198 | 5837 | error = dsl_destroy_snapshot(fullname, B_FALSE); |
428870ff | 5838 | if (error != EBUSY) { |
13fe0198 | 5839 | fatal(0, "dsl_destroy_snapshot(%s, B_FALSE) = %d", |
428870ff BB |
5840 | fullname, error); |
5841 | } | |
5842 | ||
13fe0198 | 5843 | error = dsl_destroy_snapshot(fullname, B_TRUE); |
428870ff | 5844 | if (error) { |
13fe0198 | 5845 | fatal(0, "dsl_destroy_snapshot(%s, B_TRUE) = %d", |
428870ff | 5846 | fullname, error); |
34dc7c2f BB |
5847 | } |
5848 | ||
13fe0198 | 5849 | error = user_release_one(fullname, tag); |
428870ff | 5850 | if (error) |
95fd54a1 | 5851 | fatal(0, "user_release_one(%s, %s) = %d", fullname, tag, error); |
428870ff | 5852 | |
13fe0198 | 5853 | VERIFY3U(dmu_objset_hold(fullname, FTAG, &origin), ==, ENOENT); |
428870ff BB |
5854 | |
5855 | out: | |
2d9142c9 | 5856 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
34dc7c2f BB |
5857 | } |
5858 | ||
34dc7c2f BB |
5859 | /* |
5860 | * Inject random faults into the on-disk data. | |
5861 | */ | |
428870ff | 5862 | /* ARGSUSED */ |
34dc7c2f | 5863 | void |
428870ff | 5864 | ztest_fault_inject(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 5865 | { |
428870ff | 5866 | ztest_shared_t *zs = ztest_shared; |
c242c188 | 5867 | spa_t *spa = ztest_spa; |
34dc7c2f BB |
5868 | int fd; |
5869 | uint64_t offset; | |
428870ff | 5870 | uint64_t leaves; |
c5b3a7bb | 5871 | uint64_t bad = 0x1990c0ffeedecadeull; |
34dc7c2f | 5872 | uint64_t top, leaf; |
40b84e7a BB |
5873 | char *path0; |
5874 | char *pathrand; | |
34dc7c2f | 5875 | size_t fsize; |
2014c09f | 5876 | int bshift = SPA_MAXBLOCKSHIFT + 2; |
34dc7c2f | 5877 | int iters = 1000; |
428870ff BB |
5878 | int maxfaults; |
5879 | int mirror_save; | |
b128c09f | 5880 | vdev_t *vd0 = NULL; |
34dc7c2f | 5881 | uint64_t guid0 = 0; |
428870ff BB |
5882 | boolean_t islog = B_FALSE; |
5883 | ||
40b84e7a BB |
5884 | path0 = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
5885 | pathrand = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
5886 | ||
c242c188 | 5887 | mutex_enter(&ztest_vdev_lock); |
9e052db4 MA |
5888 | |
5889 | /* | |
5890 | * Device removal is in progress, fault injection must be disabled | |
5891 | * until it completes and the pool is scrubbed. The fault injection | |
5892 | * strategy for damaging blocks does not take in to account evacuated | |
5893 | * blocks which may have already been damaged. | |
5894 | */ | |
5895 | if (ztest_device_removal_active) { | |
5896 | mutex_exit(&ztest_vdev_lock); | |
5897 | goto out; | |
5898 | } | |
5899 | ||
60b82074 | 5900 | maxfaults = MAXFAULTS(zs); |
b2255edc | 5901 | leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raid_children; |
428870ff | 5902 | mirror_save = zs->zs_mirrors; |
c242c188 | 5903 | mutex_exit(&ztest_vdev_lock); |
34dc7c2f | 5904 | |
b128c09f | 5905 | ASSERT(leaves >= 1); |
34dc7c2f | 5906 | |
ce5fb2a7 BB |
5907 | /* |
5908 | * While ztest is running the number of leaves will not change. This | |
5909 | * is critical for the fault injection logic as it determines where | |
5910 | * errors can be safely injected such that they are always repairable. | |
5911 | * | |
5912 | * When restarting ztest a different number of leaves may be requested | |
5913 | * which will shift the regions to be damaged. This is fine as long | |
5914 | * as the pool has been scrubbed prior to using the new mapping. | |
5915 | * Failure to do can result in non-repairable damage being injected. | |
5916 | */ | |
5917 | if (ztest_pool_scrubbed == B_FALSE) | |
5918 | goto out; | |
5919 | ||
621dd7bb GW |
5920 | /* |
5921 | * Grab the name lock as reader. There are some operations | |
5922 | * which don't like to have their vdevs changed while | |
5923 | * they are in progress (i.e. spa_change_guid). Those | |
5924 | * operations will have grabbed the name lock as writer. | |
5925 | */ | |
2d9142c9 | 5926 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
621dd7bb | 5927 | |
34dc7c2f | 5928 | /* |
b128c09f | 5929 | * We need SCL_STATE here because we're going to look at vd0->vdev_tsd. |
34dc7c2f | 5930 | */ |
b128c09f | 5931 | spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); |
34dc7c2f | 5932 | |
b128c09f BB |
5933 | if (ztest_random(2) == 0) { |
5934 | /* | |
428870ff | 5935 | * Inject errors on a normal data device or slog device. |
b128c09f | 5936 | */ |
428870ff BB |
5937 | top = ztest_random_vdev_top(spa, B_TRUE); |
5938 | leaf = ztest_random(leaves) + zs->zs_splits; | |
34dc7c2f | 5939 | |
b128c09f BB |
5940 | /* |
5941 | * Generate paths to the first leaf in this top-level vdev, | |
5942 | * and to the random leaf we selected. We'll induce transient | |
5943 | * write failures and random online/offline activity on leaf 0, | |
5944 | * and we'll write random garbage to the randomly chosen leaf. | |
5945 | */ | |
6aec1cd5 | 5946 | (void) snprintf(path0, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
5947 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
5948 | top * leaves + zs->zs_splits); | |
6aec1cd5 | 5949 | (void) snprintf(pathrand, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
5950 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
5951 | top * leaves + leaf); | |
34dc7c2f | 5952 | |
b128c09f | 5953 | vd0 = vdev_lookup_by_path(spa->spa_root_vdev, path0); |
428870ff BB |
5954 | if (vd0 != NULL && vd0->vdev_top->vdev_islog) |
5955 | islog = B_TRUE; | |
5956 | ||
621dd7bb GW |
5957 | /* |
5958 | * If the top-level vdev needs to be resilvered | |
5959 | * then we only allow faults on the device that is | |
5960 | * resilvering. | |
5961 | */ | |
5962 | if (vd0 != NULL && maxfaults != 1 && | |
5963 | (!vdev_resilver_needed(vd0->vdev_top, NULL, NULL) || | |
5d1f7fb6 | 5964 | vd0->vdev_resilver_txg != 0)) { |
b128c09f BB |
5965 | /* |
5966 | * Make vd0 explicitly claim to be unreadable, | |
5967 | * or unwriteable, or reach behind its back | |
5968 | * and close the underlying fd. We can do this if | |
5969 | * maxfaults == 0 because we'll fail and reexecute, | |
5970 | * and we can do it if maxfaults >= 2 because we'll | |
5971 | * have enough redundancy. If maxfaults == 1, the | |
5972 | * combination of this with injection of random data | |
5973 | * corruption below exceeds the pool's fault tolerance. | |
5974 | */ | |
5975 | vdev_file_t *vf = vd0->vdev_tsd; | |
5976 | ||
a1d477c2 MA |
5977 | zfs_dbgmsg("injecting fault to vdev %llu; maxfaults=%d", |
5978 | (long long)vd0->vdev_id, (int)maxfaults); | |
5979 | ||
b128c09f | 5980 | if (vf != NULL && ztest_random(3) == 0) { |
da92d5cb MM |
5981 | (void) close(vf->vf_file->f_fd); |
5982 | vf->vf_file->f_fd = -1; | |
b128c09f BB |
5983 | } else if (ztest_random(2) == 0) { |
5984 | vd0->vdev_cant_read = B_TRUE; | |
5985 | } else { | |
5986 | vd0->vdev_cant_write = B_TRUE; | |
5987 | } | |
5988 | guid0 = vd0->vdev_guid; | |
5989 | } | |
5990 | } else { | |
5991 | /* | |
5992 | * Inject errors on an l2cache device. | |
5993 | */ | |
5994 | spa_aux_vdev_t *sav = &spa->spa_l2cache; | |
34dc7c2f | 5995 | |
b128c09f BB |
5996 | if (sav->sav_count == 0) { |
5997 | spa_config_exit(spa, SCL_STATE, FTAG); | |
2d9142c9 | 5998 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
40b84e7a | 5999 | goto out; |
b128c09f BB |
6000 | } |
6001 | vd0 = sav->sav_vdevs[ztest_random(sav->sav_count)]; | |
34dc7c2f | 6002 | guid0 = vd0->vdev_guid; |
b128c09f BB |
6003 | (void) strcpy(path0, vd0->vdev_path); |
6004 | (void) strcpy(pathrand, vd0->vdev_path); | |
6005 | ||
6006 | leaf = 0; | |
6007 | leaves = 1; | |
6008 | maxfaults = INT_MAX; /* no limit on cache devices */ | |
34dc7c2f BB |
6009 | } |
6010 | ||
b128c09f | 6011 | spa_config_exit(spa, SCL_STATE, FTAG); |
2d9142c9 | 6012 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
b128c09f | 6013 | |
34dc7c2f | 6014 | /* |
428870ff BB |
6015 | * If we can tolerate two or more faults, or we're dealing |
6016 | * with a slog, randomly online/offline vd0. | |
34dc7c2f | 6017 | */ |
428870ff | 6018 | if ((maxfaults >= 2 || islog) && guid0 != 0) { |
fb5f0bc8 BB |
6019 | if (ztest_random(10) < 6) { |
6020 | int flags = (ztest_random(2) == 0 ? | |
6021 | ZFS_OFFLINE_TEMPORARY : 0); | |
428870ff BB |
6022 | |
6023 | /* | |
6024 | * We have to grab the zs_name_lock as writer to | |
6025 | * prevent a race between offlining a slog and | |
6026 | * destroying a dataset. Offlining the slog will | |
6027 | * grab a reference on the dataset which may cause | |
13fe0198 | 6028 | * dsl_destroy_head() to fail with EBUSY thus |
428870ff BB |
6029 | * leaving the dataset in an inconsistent state. |
6030 | */ | |
6031 | if (islog) | |
2d9142c9 | 6032 | (void) pthread_rwlock_wrlock(&ztest_name_lock); |
428870ff | 6033 | |
fb5f0bc8 | 6034 | VERIFY(vdev_offline(spa, guid0, flags) != EBUSY); |
428870ff BB |
6035 | |
6036 | if (islog) | |
2d9142c9 | 6037 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
fb5f0bc8 | 6038 | } else { |
1eb5bfa3 GW |
6039 | /* |
6040 | * Ideally we would like to be able to randomly | |
6041 | * call vdev_[on|off]line without holding locks | |
6042 | * to force unpredictable failures but the side | |
6043 | * effects of vdev_[on|off]line prevent us from | |
6044 | * doing so. We grab the ztest_vdev_lock here to | |
6045 | * prevent a race between injection testing and | |
6046 | * aux_vdev removal. | |
6047 | */ | |
6048 | mutex_enter(&ztest_vdev_lock); | |
fb5f0bc8 | 6049 | (void) vdev_online(spa, guid0, 0, NULL); |
1eb5bfa3 | 6050 | mutex_exit(&ztest_vdev_lock); |
fb5f0bc8 | 6051 | } |
34dc7c2f BB |
6052 | } |
6053 | ||
428870ff | 6054 | if (maxfaults == 0) |
40b84e7a | 6055 | goto out; |
428870ff | 6056 | |
34dc7c2f BB |
6057 | /* |
6058 | * We have at least single-fault tolerance, so inject data corruption. | |
6059 | */ | |
6060 | fd = open(pathrand, O_RDWR); | |
6061 | ||
d2734cce | 6062 | if (fd == -1) /* we hit a gap in the device namespace */ |
40b84e7a | 6063 | goto out; |
34dc7c2f BB |
6064 | |
6065 | fsize = lseek(fd, 0, SEEK_END); | |
6066 | ||
6067 | while (--iters != 0) { | |
91d88843 MA |
6068 | /* |
6069 | * The offset must be chosen carefully to ensure that | |
6070 | * we do not inject a given logical block with errors | |
6071 | * on two different leaf devices, because ZFS can not | |
6072 | * tolerate that (if maxfaults==1). | |
6073 | * | |
42d8d1d6 SD |
6074 | * To achieve this we divide each leaf device into |
6075 | * chunks of size (# leaves * SPA_MAXBLOCKSIZE * 4). | |
6076 | * Each chunk is further divided into error-injection | |
6077 | * ranges (can accept errors) and clear ranges (we do | |
6078 | * not inject errors in those). Each error-injection | |
6079 | * range can accept errors only for a single leaf vdev. | |
6080 | * Error-injection ranges are separated by clear ranges. | |
91d88843 MA |
6081 | * |
6082 | * For example, with 3 leaves, each chunk looks like: | |
6083 | * 0 to 32M: injection range for leaf 0 | |
42d8d1d6 | 6084 | * 32M to 64M: clear range - no injection allowed |
91d88843 | 6085 | * 64M to 96M: injection range for leaf 1 |
42d8d1d6 | 6086 | * 96M to 128M: clear range - no injection allowed |
91d88843 | 6087 | * 128M to 160M: injection range for leaf 2 |
42d8d1d6 SD |
6088 | * 160M to 192M: clear range - no injection allowed |
6089 | * | |
6090 | * Each clear range must be large enough such that a | |
6091 | * single block cannot straddle it. This way a block | |
6092 | * can't be a target in two different injection ranges | |
6093 | * (on different leaf vdevs). | |
91d88843 | 6094 | */ |
34dc7c2f BB |
6095 | offset = ztest_random(fsize / (leaves << bshift)) * |
6096 | (leaves << bshift) + (leaf << bshift) + | |
6097 | (ztest_random(1ULL << (bshift - 1)) & -8ULL); | |
6098 | ||
2014c09f GM |
6099 | /* |
6100 | * Only allow damage to the labels at one end of the vdev. | |
6101 | * | |
6102 | * If all labels are damaged, the device will be totally | |
6103 | * inaccessible, which will result in loss of data, | |
6104 | * because we also damage (parts of) the other side of | |
6105 | * the mirror/raidz. | |
6106 | * | |
6107 | * Additionally, we will always have both an even and an | |
6108 | * odd label, so that we can handle crashes in the | |
6109 | * middle of vdev_config_sync(). | |
6110 | */ | |
6111 | if ((leaf & 1) == 0 && offset < VDEV_LABEL_START_SIZE) | |
6112 | continue; | |
6113 | ||
6114 | /* | |
6115 | * The two end labels are stored at the "end" of the disk, but | |
6116 | * the end of the disk (vdev_psize) is aligned to | |
6117 | * sizeof (vdev_label_t). | |
6118 | */ | |
6119 | uint64_t psize = P2ALIGN(fsize, sizeof (vdev_label_t)); | |
6120 | if ((leaf & 1) == 1 && | |
6121 | offset + sizeof (bad) > psize - VDEV_LABEL_END_SIZE) | |
34dc7c2f BB |
6122 | continue; |
6123 | ||
c242c188 | 6124 | mutex_enter(&ztest_vdev_lock); |
428870ff | 6125 | if (mirror_save != zs->zs_mirrors) { |
c242c188 | 6126 | mutex_exit(&ztest_vdev_lock); |
428870ff | 6127 | (void) close(fd); |
40b84e7a | 6128 | goto out; |
428870ff | 6129 | } |
34dc7c2f BB |
6130 | |
6131 | if (pwrite(fd, &bad, sizeof (bad), offset) != sizeof (bad)) | |
6132 | fatal(1, "can't inject bad word at 0x%llx in %s", | |
6133 | offset, pathrand); | |
428870ff | 6134 | |
c242c188 | 6135 | mutex_exit(&ztest_vdev_lock); |
428870ff | 6136 | |
c242c188 | 6137 | if (ztest_opts.zo_verbose >= 7) |
428870ff BB |
6138 | (void) printf("injected bad word into %s," |
6139 | " offset 0x%llx\n", pathrand, (u_longlong_t)offset); | |
34dc7c2f BB |
6140 | } |
6141 | ||
6142 | (void) close(fd); | |
40b84e7a BB |
6143 | out: |
6144 | umem_free(path0, MAXPATHLEN); | |
6145 | umem_free(pathrand, MAXPATHLEN); | |
34dc7c2f BB |
6146 | } |
6147 | ||
ce5fb2a7 BB |
6148 | /* |
6149 | * By design ztest will never inject uncorrectable damage in to the pool. | |
6150 | * Issue a scrub, wait for it to complete, and verify there is never any | |
b2255edc | 6151 | * persistent damage. |
ce5fb2a7 BB |
6152 | * |
6153 | * Only after a full scrub has been completed is it safe to start injecting | |
6154 | * data corruption. See the comment in zfs_fault_inject(). | |
6155 | */ | |
6156 | static int | |
6157 | ztest_scrub_impl(spa_t *spa) | |
6158 | { | |
6159 | int error = spa_scan(spa, POOL_SCAN_SCRUB); | |
6160 | if (error) | |
6161 | return (error); | |
6162 | ||
6163 | while (dsl_scan_scrubbing(spa_get_dsl(spa))) | |
6164 | txg_wait_synced(spa_get_dsl(spa), 0); | |
6165 | ||
6166 | if (spa_get_errlog_size(spa) > 0) | |
6167 | return (ECKSUM); | |
6168 | ||
6169 | ztest_pool_scrubbed = B_TRUE; | |
6170 | ||
6171 | return (0); | |
6172 | } | |
6173 | ||
34dc7c2f | 6174 | /* |
428870ff | 6175 | * Scrub the pool. |
34dc7c2f | 6176 | */ |
428870ff BB |
6177 | /* ARGSUSED */ |
6178 | void | |
6179 | ztest_scrub(ztest_ds_t *zd, uint64_t id) | |
34dc7c2f | 6180 | { |
c242c188 | 6181 | spa_t *spa = ztest_spa; |
ce5fb2a7 | 6182 | int error; |
34dc7c2f | 6183 | |
9e052db4 MA |
6184 | /* |
6185 | * Scrub in progress by device removal. | |
6186 | */ | |
6187 | if (ztest_device_removal_active) | |
6188 | return; | |
6189 | ||
ce5fb2a7 BB |
6190 | /* |
6191 | * Start a scrub, wait a moment, then force a restart. | |
6192 | */ | |
428870ff | 6193 | (void) spa_scan(spa, POOL_SCAN_SCRUB); |
ce5fb2a7 BB |
6194 | (void) poll(NULL, 0, 100); |
6195 | ||
6196 | error = ztest_scrub_impl(spa); | |
6197 | if (error == EBUSY) | |
6198 | error = 0; | |
6199 | ASSERT0(error); | |
428870ff | 6200 | } |
34dc7c2f | 6201 | |
3541dc6d GA |
6202 | /* |
6203 | * Change the guid for the pool. | |
6204 | */ | |
6205 | /* ARGSUSED */ | |
6206 | void | |
6207 | ztest_reguid(ztest_ds_t *zd, uint64_t id) | |
6208 | { | |
c242c188 | 6209 | spa_t *spa = ztest_spa; |
3541dc6d | 6210 | uint64_t orig, load; |
3bc7e0fb | 6211 | int error; |
3541dc6d | 6212 | |
b33d668d OF |
6213 | if (ztest_opts.zo_mmp_test) |
6214 | return; | |
6215 | ||
3541dc6d GA |
6216 | orig = spa_guid(spa); |
6217 | load = spa_load_guid(spa); | |
3bc7e0fb | 6218 | |
2d9142c9 | 6219 | (void) pthread_rwlock_wrlock(&ztest_name_lock); |
3bc7e0fb | 6220 | error = spa_change_guid(spa); |
2d9142c9 | 6221 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
3bc7e0fb GW |
6222 | |
6223 | if (error != 0) | |
3541dc6d GA |
6224 | return; |
6225 | ||
ea0b2538 | 6226 | if (ztest_opts.zo_verbose >= 4) { |
3541dc6d GA |
6227 | (void) printf("Changed guid old %llu -> %llu\n", |
6228 | (u_longlong_t)orig, (u_longlong_t)spa_guid(spa)); | |
6229 | } | |
6230 | ||
6231 | VERIFY3U(orig, !=, spa_guid(spa)); | |
6232 | VERIFY3U(load, ==, spa_load_guid(spa)); | |
6233 | } | |
6234 | ||
1eeb4562 JX |
6235 | void |
6236 | ztest_fletcher(ztest_ds_t *zd, uint64_t id) | |
6237 | { | |
6238 | hrtime_t end = gethrtime() + NANOSEC; | |
6239 | ||
6240 | while (gethrtime() <= end) { | |
6241 | int run_count = 100; | |
6242 | void *buf; | |
2fe36b0b | 6243 | struct abd *abd_data, *abd_meta; |
1eeb4562 JX |
6244 | uint32_t size; |
6245 | int *ptr; | |
6246 | int i; | |
6247 | zio_cksum_t zc_ref; | |
6248 | zio_cksum_t zc_ref_byteswap; | |
6249 | ||
6250 | size = ztest_random_blocksize(); | |
2fe36b0b | 6251 | |
1eeb4562 | 6252 | buf = umem_alloc(size, UMEM_NOFAIL); |
2fe36b0b DQ |
6253 | abd_data = abd_alloc(size, B_FALSE); |
6254 | abd_meta = abd_alloc(size, B_TRUE); | |
1eeb4562 JX |
6255 | |
6256 | for (i = 0, ptr = buf; i < size / sizeof (*ptr); i++, ptr++) | |
6257 | *ptr = ztest_random(UINT_MAX); | |
6258 | ||
2fe36b0b DQ |
6259 | abd_copy_from_buf_off(abd_data, buf, 0, size); |
6260 | abd_copy_from_buf_off(abd_meta, buf, 0, size); | |
6261 | ||
1eeb4562 | 6262 | VERIFY0(fletcher_4_impl_set("scalar")); |
3c67d83a TH |
6263 | fletcher_4_native(buf, size, NULL, &zc_ref); |
6264 | fletcher_4_byteswap(buf, size, NULL, &zc_ref_byteswap); | |
1eeb4562 JX |
6265 | |
6266 | VERIFY0(fletcher_4_impl_set("cycle")); | |
6267 | while (run_count-- > 0) { | |
6268 | zio_cksum_t zc; | |
6269 | zio_cksum_t zc_byteswap; | |
6270 | ||
3c67d83a TH |
6271 | fletcher_4_byteswap(buf, size, NULL, &zc_byteswap); |
6272 | fletcher_4_native(buf, size, NULL, &zc); | |
1eeb4562 JX |
6273 | |
6274 | VERIFY0(bcmp(&zc, &zc_ref, sizeof (zc))); | |
6275 | VERIFY0(bcmp(&zc_byteswap, &zc_ref_byteswap, | |
6276 | sizeof (zc_byteswap))); | |
2fe36b0b DQ |
6277 | |
6278 | /* Test ABD - data */ | |
6279 | abd_fletcher_4_byteswap(abd_data, size, NULL, | |
6280 | &zc_byteswap); | |
6281 | abd_fletcher_4_native(abd_data, size, NULL, &zc); | |
6282 | ||
6283 | VERIFY0(bcmp(&zc, &zc_ref, sizeof (zc))); | |
6284 | VERIFY0(bcmp(&zc_byteswap, &zc_ref_byteswap, | |
6285 | sizeof (zc_byteswap))); | |
6286 | ||
6287 | /* Test ABD - metadata */ | |
6288 | abd_fletcher_4_byteswap(abd_meta, size, NULL, | |
6289 | &zc_byteswap); | |
6290 | abd_fletcher_4_native(abd_meta, size, NULL, &zc); | |
6291 | ||
6292 | VERIFY0(bcmp(&zc, &zc_ref, sizeof (zc))); | |
6293 | VERIFY0(bcmp(&zc_byteswap, &zc_ref_byteswap, | |
6294 | sizeof (zc_byteswap))); | |
6295 | ||
1eeb4562 JX |
6296 | } |
6297 | ||
6298 | umem_free(buf, size); | |
2fe36b0b DQ |
6299 | abd_free(abd_data); |
6300 | abd_free(abd_meta); | |
1eeb4562 JX |
6301 | } |
6302 | } | |
6303 | ||
37f520db GN |
6304 | void |
6305 | ztest_fletcher_incr(ztest_ds_t *zd, uint64_t id) | |
6306 | { | |
6307 | void *buf; | |
6308 | size_t size; | |
6309 | int *ptr; | |
6310 | int i; | |
6311 | zio_cksum_t zc_ref; | |
6312 | zio_cksum_t zc_ref_bswap; | |
6313 | ||
6314 | hrtime_t end = gethrtime() + NANOSEC; | |
6315 | ||
6316 | while (gethrtime() <= end) { | |
6317 | int run_count = 100; | |
6318 | ||
6319 | size = ztest_random_blocksize(); | |
6320 | buf = umem_alloc(size, UMEM_NOFAIL); | |
6321 | ||
6322 | for (i = 0, ptr = buf; i < size / sizeof (*ptr); i++, ptr++) | |
6323 | *ptr = ztest_random(UINT_MAX); | |
6324 | ||
6325 | VERIFY0(fletcher_4_impl_set("scalar")); | |
6326 | fletcher_4_native(buf, size, NULL, &zc_ref); | |
6327 | fletcher_4_byteswap(buf, size, NULL, &zc_ref_bswap); | |
6328 | ||
6329 | VERIFY0(fletcher_4_impl_set("cycle")); | |
6330 | ||
6331 | while (run_count-- > 0) { | |
6332 | zio_cksum_t zc; | |
6333 | zio_cksum_t zc_bswap; | |
6334 | size_t pos = 0; | |
6335 | ||
6336 | ZIO_SET_CHECKSUM(&zc, 0, 0, 0, 0); | |
6337 | ZIO_SET_CHECKSUM(&zc_bswap, 0, 0, 0, 0); | |
6338 | ||
6339 | while (pos < size) { | |
6340 | size_t inc = 64 * ztest_random(size / 67); | |
6341 | /* sometimes add few bytes to test non-simd */ | |
6342 | if (ztest_random(100) < 10) | |
6343 | inc += P2ALIGN(ztest_random(64), | |
6344 | sizeof (uint32_t)); | |
6345 | ||
6346 | if (inc > (size - pos)) | |
6347 | inc = size - pos; | |
6348 | ||
6349 | fletcher_4_incremental_native(buf + pos, inc, | |
6350 | &zc); | |
6351 | fletcher_4_incremental_byteswap(buf + pos, inc, | |
6352 | &zc_bswap); | |
6353 | ||
6354 | pos += inc; | |
6355 | } | |
6356 | ||
6357 | VERIFY3U(pos, ==, size); | |
6358 | ||
6359 | VERIFY(ZIO_CHECKSUM_EQUAL(zc, zc_ref)); | |
6360 | VERIFY(ZIO_CHECKSUM_EQUAL(zc_bswap, zc_ref_bswap)); | |
6361 | ||
6362 | /* | |
6363 | * verify if incremental on the whole buffer is | |
6364 | * equivalent to non-incremental version | |
6365 | */ | |
6366 | ZIO_SET_CHECKSUM(&zc, 0, 0, 0, 0); | |
6367 | ZIO_SET_CHECKSUM(&zc_bswap, 0, 0, 0, 0); | |
6368 | ||
6369 | fletcher_4_incremental_native(buf, size, &zc); | |
6370 | fletcher_4_incremental_byteswap(buf, size, &zc_bswap); | |
6371 | ||
6372 | VERIFY(ZIO_CHECKSUM_EQUAL(zc, zc_ref)); | |
6373 | VERIFY(ZIO_CHECKSUM_EQUAL(zc_bswap, zc_ref_bswap)); | |
6374 | } | |
6375 | ||
6376 | umem_free(buf, size); | |
6377 | } | |
6378 | } | |
6379 | ||
90aa094d CC |
6380 | static int |
6381 | ztest_check_path(char *path) | |
6382 | { | |
6383 | struct stat s; | |
6384 | /* return true on success */ | |
6385 | return (!stat(path, &s)); | |
6386 | } | |
6387 | ||
6388 | static void | |
6389 | ztest_get_zdb_bin(char *bin, int len) | |
6390 | { | |
6391 | char *zdb_path; | |
6392 | /* | |
6393 | * Try to use ZDB_PATH and in-tree zdb path. If not successful, just | |
6394 | * let popen to search through PATH. | |
6395 | */ | |
6396 | if ((zdb_path = getenv("ZDB_PATH"))) { | |
6397 | strlcpy(bin, zdb_path, len); /* In env */ | |
6398 | if (!ztest_check_path(bin)) { | |
6399 | ztest_dump_core = 0; | |
6400 | fatal(1, "invalid ZDB_PATH '%s'", bin); | |
6401 | } | |
6402 | return; | |
6403 | } | |
6404 | ||
6405 | VERIFY(realpath(getexecname(), bin) != NULL); | |
6406 | if (strstr(bin, "/ztest/")) { | |
6407 | strstr(bin, "/ztest/")[0] = '\0'; /* In-tree */ | |
6408 | strcat(bin, "/zdb/zdb"); | |
6409 | if (ztest_check_path(bin)) | |
6410 | return; | |
6411 | } | |
6412 | strcpy(bin, "zdb"); | |
6413 | } | |
6414 | ||
619f0976 GW |
6415 | static vdev_t * |
6416 | ztest_random_concrete_vdev_leaf(vdev_t *vd) | |
6417 | { | |
6418 | if (vd == NULL) | |
6419 | return (NULL); | |
6420 | ||
6421 | if (vd->vdev_children == 0) | |
6422 | return (vd); | |
6423 | ||
6424 | vdev_t *eligible[vd->vdev_children]; | |
6425 | int eligible_idx = 0, i; | |
6426 | for (i = 0; i < vd->vdev_children; i++) { | |
6427 | vdev_t *cvd = vd->vdev_child[i]; | |
6428 | if (cvd->vdev_top->vdev_removing) | |
6429 | continue; | |
6430 | if (cvd->vdev_children > 0 || | |
6431 | (vdev_is_concrete(cvd) && !cvd->vdev_detached)) { | |
6432 | eligible[eligible_idx++] = cvd; | |
6433 | } | |
6434 | } | |
6435 | VERIFY(eligible_idx > 0); | |
6436 | ||
6437 | uint64_t child_no = ztest_random(eligible_idx); | |
6438 | return (ztest_random_concrete_vdev_leaf(eligible[child_no])); | |
6439 | } | |
6440 | ||
6441 | /* ARGSUSED */ | |
6442 | void | |
6443 | ztest_initialize(ztest_ds_t *zd, uint64_t id) | |
6444 | { | |
6445 | spa_t *spa = ztest_spa; | |
6446 | int error = 0; | |
6447 | ||
6448 | mutex_enter(&ztest_vdev_lock); | |
6449 | ||
6450 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
6451 | ||
6452 | /* Random leaf vdev */ | |
6453 | vdev_t *rand_vd = ztest_random_concrete_vdev_leaf(spa->spa_root_vdev); | |
6454 | if (rand_vd == NULL) { | |
6455 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
6456 | mutex_exit(&ztest_vdev_lock); | |
6457 | return; | |
6458 | } | |
6459 | ||
6460 | /* | |
6461 | * The random vdev we've selected may change as soon as we | |
6462 | * drop the spa_config_lock. We create local copies of things | |
6463 | * we're interested in. | |
6464 | */ | |
6465 | uint64_t guid = rand_vd->vdev_guid; | |
6466 | char *path = strdup(rand_vd->vdev_path); | |
6467 | boolean_t active = rand_vd->vdev_initialize_thread != NULL; | |
6468 | ||
a887d653 | 6469 | zfs_dbgmsg("vd %px, guid %llu", rand_vd, guid); |
619f0976 GW |
6470 | spa_config_exit(spa, SCL_VDEV, FTAG); |
6471 | ||
6472 | uint64_t cmd = ztest_random(POOL_INITIALIZE_FUNCS); | |
c10d37dd GW |
6473 | |
6474 | nvlist_t *vdev_guids = fnvlist_alloc(); | |
6475 | nvlist_t *vdev_errlist = fnvlist_alloc(); | |
6476 | fnvlist_add_uint64(vdev_guids, path, guid); | |
6477 | error = spa_vdev_initialize(spa, vdev_guids, cmd, vdev_errlist); | |
6478 | fnvlist_free(vdev_guids); | |
6479 | fnvlist_free(vdev_errlist); | |
6480 | ||
619f0976 GW |
6481 | switch (cmd) { |
6482 | case POOL_INITIALIZE_CANCEL: | |
6483 | if (ztest_opts.zo_verbose >= 4) { | |
6484 | (void) printf("Cancel initialize %s", path); | |
6485 | if (!active) | |
6486 | (void) printf(" failed (no initialize active)"); | |
6487 | (void) printf("\n"); | |
6488 | } | |
6489 | break; | |
1b939560 | 6490 | case POOL_INITIALIZE_START: |
619f0976 GW |
6491 | if (ztest_opts.zo_verbose >= 4) { |
6492 | (void) printf("Start initialize %s", path); | |
6493 | if (active && error == 0) | |
6494 | (void) printf(" failed (already active)"); | |
6495 | else if (error != 0) | |
6496 | (void) printf(" failed (error %d)", error); | |
6497 | (void) printf("\n"); | |
6498 | } | |
6499 | break; | |
6500 | case POOL_INITIALIZE_SUSPEND: | |
6501 | if (ztest_opts.zo_verbose >= 4) { | |
6502 | (void) printf("Suspend initialize %s", path); | |
6503 | if (!active) | |
6504 | (void) printf(" failed (no initialize active)"); | |
6505 | (void) printf("\n"); | |
6506 | } | |
6507 | break; | |
6508 | } | |
6509 | free(path); | |
6510 | mutex_exit(&ztest_vdev_lock); | |
6511 | } | |
6512 | ||
1b939560 BB |
6513 | /* ARGSUSED */ |
6514 | void | |
6515 | ztest_trim(ztest_ds_t *zd, uint64_t id) | |
6516 | { | |
6517 | spa_t *spa = ztest_spa; | |
6518 | int error = 0; | |
6519 | ||
6520 | mutex_enter(&ztest_vdev_lock); | |
6521 | ||
6522 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
6523 | ||
6524 | /* Random leaf vdev */ | |
6525 | vdev_t *rand_vd = ztest_random_concrete_vdev_leaf(spa->spa_root_vdev); | |
6526 | if (rand_vd == NULL) { | |
6527 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
6528 | mutex_exit(&ztest_vdev_lock); | |
6529 | return; | |
6530 | } | |
6531 | ||
6532 | /* | |
6533 | * The random vdev we've selected may change as soon as we | |
6534 | * drop the spa_config_lock. We create local copies of things | |
6535 | * we're interested in. | |
6536 | */ | |
6537 | uint64_t guid = rand_vd->vdev_guid; | |
6538 | char *path = strdup(rand_vd->vdev_path); | |
6539 | boolean_t active = rand_vd->vdev_trim_thread != NULL; | |
6540 | ||
6541 | zfs_dbgmsg("vd %p, guid %llu", rand_vd, guid); | |
6542 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
6543 | ||
6544 | uint64_t cmd = ztest_random(POOL_TRIM_FUNCS); | |
6545 | uint64_t rate = 1 << ztest_random(30); | |
6546 | boolean_t partial = (ztest_random(5) > 0); | |
6547 | boolean_t secure = (ztest_random(5) > 0); | |
6548 | ||
6549 | nvlist_t *vdev_guids = fnvlist_alloc(); | |
6550 | nvlist_t *vdev_errlist = fnvlist_alloc(); | |
6551 | fnvlist_add_uint64(vdev_guids, path, guid); | |
6552 | error = spa_vdev_trim(spa, vdev_guids, cmd, rate, partial, | |
6553 | secure, vdev_errlist); | |
6554 | fnvlist_free(vdev_guids); | |
6555 | fnvlist_free(vdev_errlist); | |
6556 | ||
6557 | switch (cmd) { | |
6558 | case POOL_TRIM_CANCEL: | |
6559 | if (ztest_opts.zo_verbose >= 4) { | |
6560 | (void) printf("Cancel TRIM %s", path); | |
6561 | if (!active) | |
6562 | (void) printf(" failed (no TRIM active)"); | |
6563 | (void) printf("\n"); | |
6564 | } | |
6565 | break; | |
6566 | case POOL_TRIM_START: | |
6567 | if (ztest_opts.zo_verbose >= 4) { | |
6568 | (void) printf("Start TRIM %s", path); | |
6569 | if (active && error == 0) | |
6570 | (void) printf(" failed (already active)"); | |
6571 | else if (error != 0) | |
6572 | (void) printf(" failed (error %d)", error); | |
6573 | (void) printf("\n"); | |
6574 | } | |
6575 | break; | |
6576 | case POOL_TRIM_SUSPEND: | |
6577 | if (ztest_opts.zo_verbose >= 4) { | |
6578 | (void) printf("Suspend TRIM %s", path); | |
6579 | if (!active) | |
6580 | (void) printf(" failed (no TRIM active)"); | |
6581 | (void) printf("\n"); | |
6582 | } | |
6583 | break; | |
6584 | } | |
6585 | free(path); | |
6586 | mutex_exit(&ztest_vdev_lock); | |
6587 | } | |
6588 | ||
428870ff BB |
6589 | /* |
6590 | * Verify pool integrity by running zdb. | |
6591 | */ | |
34dc7c2f | 6592 | static void |
428870ff | 6593 | ztest_run_zdb(char *pool) |
34dc7c2f BB |
6594 | { |
6595 | int status; | |
34dc7c2f | 6596 | char *bin; |
0e8d1b2d BB |
6597 | char *zdb; |
6598 | char *zbuf; | |
90aa094d | 6599 | const int len = MAXPATHLEN + MAXNAMELEN + 20; |
34dc7c2f BB |
6600 | FILE *fp; |
6601 | ||
90aa094d CC |
6602 | bin = umem_alloc(len, UMEM_NOFAIL); |
6603 | zdb = umem_alloc(len, UMEM_NOFAIL); | |
0e8d1b2d | 6604 | zbuf = umem_alloc(1024, UMEM_NOFAIL); |
34dc7c2f | 6605 | |
90aa094d | 6606 | ztest_get_zdb_bin(bin, len); |
0e8d1b2d BB |
6607 | |
6608 | (void) sprintf(zdb, | |
6774931d | 6609 | "%s -bcc%s%s -G -d -Y -e -y -p %s %s", |
0e8d1b2d | 6610 | bin, |
c242c188 CS |
6611 | ztest_opts.zo_verbose >= 3 ? "s" : "", |
6612 | ztest_opts.zo_verbose >= 4 ? "v" : "", | |
3d93161b | 6613 | ztest_opts.zo_dir, |
b128c09f | 6614 | pool); |
34dc7c2f | 6615 | |
c242c188 | 6616 | if (ztest_opts.zo_verbose >= 5) |
34dc7c2f BB |
6617 | (void) printf("Executing %s\n", strstr(zdb, "zdb ")); |
6618 | ||
6619 | fp = popen(zdb, "r"); | |
6620 | ||
6aec1cd5 | 6621 | while (fgets(zbuf, 1024, fp) != NULL) |
c242c188 | 6622 | if (ztest_opts.zo_verbose >= 3) |
34dc7c2f BB |
6623 | (void) printf("%s", zbuf); |
6624 | ||
6625 | status = pclose(fp); | |
6626 | ||
6627 | if (status == 0) | |
0e8d1b2d | 6628 | goto out; |
34dc7c2f BB |
6629 | |
6630 | ztest_dump_core = 0; | |
6631 | if (WIFEXITED(status)) | |
6632 | fatal(0, "'%s' exit code %d", zdb, WEXITSTATUS(status)); | |
6633 | else | |
6634 | fatal(0, "'%s' died with signal %d", zdb, WTERMSIG(status)); | |
0e8d1b2d | 6635 | out: |
90aa094d CC |
6636 | umem_free(bin, len); |
6637 | umem_free(zdb, len); | |
0e8d1b2d | 6638 | umem_free(zbuf, 1024); |
34dc7c2f BB |
6639 | } |
6640 | ||
6641 | static void | |
6642 | ztest_walk_pool_directory(char *header) | |
6643 | { | |
6644 | spa_t *spa = NULL; | |
6645 | ||
c242c188 | 6646 | if (ztest_opts.zo_verbose >= 6) |
34dc7c2f BB |
6647 | (void) printf("%s\n", header); |
6648 | ||
6649 | mutex_enter(&spa_namespace_lock); | |
6650 | while ((spa = spa_next(spa)) != NULL) | |
c242c188 | 6651 | if (ztest_opts.zo_verbose >= 6) |
34dc7c2f BB |
6652 | (void) printf("\t%s\n", spa_name(spa)); |
6653 | mutex_exit(&spa_namespace_lock); | |
6654 | } | |
6655 | ||
6656 | static void | |
6657 | ztest_spa_import_export(char *oldname, char *newname) | |
6658 | { | |
fb5f0bc8 | 6659 | nvlist_t *config, *newconfig; |
34dc7c2f BB |
6660 | uint64_t pool_guid; |
6661 | spa_t *spa; | |
13fe0198 | 6662 | int error; |
34dc7c2f | 6663 | |
c242c188 | 6664 | if (ztest_opts.zo_verbose >= 4) { |
34dc7c2f BB |
6665 | (void) printf("import/export: old = %s, new = %s\n", |
6666 | oldname, newname); | |
6667 | } | |
6668 | ||
6669 | /* | |
6670 | * Clean up from previous runs. | |
6671 | */ | |
6672 | (void) spa_destroy(newname); | |
6673 | ||
6674 | /* | |
6675 | * Get the pool's configuration and guid. | |
6676 | */ | |
428870ff | 6677 | VERIFY3U(0, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f | 6678 | |
fb5f0bc8 BB |
6679 | /* |
6680 | * Kick off a scrub to tickle scrub/export races. | |
6681 | */ | |
6682 | if (ztest_random(2) == 0) | |
428870ff | 6683 | (void) spa_scan(spa, POOL_SCAN_SCRUB); |
fb5f0bc8 | 6684 | |
34dc7c2f BB |
6685 | pool_guid = spa_guid(spa); |
6686 | spa_close(spa, FTAG); | |
6687 | ||
6688 | ztest_walk_pool_directory("pools before export"); | |
6689 | ||
6690 | /* | |
6691 | * Export it. | |
6692 | */ | |
428870ff | 6693 | VERIFY3U(0, ==, spa_export(oldname, &config, B_FALSE, B_FALSE)); |
34dc7c2f BB |
6694 | |
6695 | ztest_walk_pool_directory("pools after export"); | |
6696 | ||
fb5f0bc8 BB |
6697 | /* |
6698 | * Try to import it. | |
6699 | */ | |
6700 | newconfig = spa_tryimport(config); | |
6701 | ASSERT(newconfig != NULL); | |
6702 | nvlist_free(newconfig); | |
6703 | ||
34dc7c2f BB |
6704 | /* |
6705 | * Import it under the new name. | |
6706 | */ | |
13fe0198 MA |
6707 | error = spa_import(newname, config, NULL, 0); |
6708 | if (error != 0) { | |
6709 | dump_nvlist(config, 0); | |
6710 | fatal(B_FALSE, "couldn't import pool %s as %s: error %u", | |
6711 | oldname, newname, error); | |
6712 | } | |
34dc7c2f BB |
6713 | |
6714 | ztest_walk_pool_directory("pools after import"); | |
6715 | ||
6716 | /* | |
6717 | * Try to import it again -- should fail with EEXIST. | |
6718 | */ | |
572e2857 | 6719 | VERIFY3U(EEXIST, ==, spa_import(newname, config, NULL, 0)); |
34dc7c2f BB |
6720 | |
6721 | /* | |
6722 | * Try to import it under a different name -- should fail with EEXIST. | |
6723 | */ | |
572e2857 | 6724 | VERIFY3U(EEXIST, ==, spa_import(oldname, config, NULL, 0)); |
34dc7c2f BB |
6725 | |
6726 | /* | |
6727 | * Verify that the pool is no longer visible under the old name. | |
6728 | */ | |
428870ff | 6729 | VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f BB |
6730 | |
6731 | /* | |
6732 | * Verify that we can open and close the pool using the new name. | |
6733 | */ | |
428870ff | 6734 | VERIFY3U(0, ==, spa_open(newname, &spa, FTAG)); |
34dc7c2f BB |
6735 | ASSERT(pool_guid == spa_guid(spa)); |
6736 | spa_close(spa, FTAG); | |
6737 | ||
6738 | nvlist_free(config); | |
6739 | } | |
6740 | ||
fb5f0bc8 BB |
6741 | static void |
6742 | ztest_resume(spa_t *spa) | |
6743 | { | |
c242c188 | 6744 | if (spa_suspended(spa) && ztest_opts.zo_verbose >= 6) |
428870ff BB |
6745 | (void) printf("resuming from suspended state\n"); |
6746 | spa_vdev_state_enter(spa, SCL_NONE); | |
6747 | vdev_clear(spa, NULL); | |
6748 | (void) spa_vdev_state_exit(spa, NULL, 0); | |
6749 | (void) zio_resume(spa); | |
fb5f0bc8 BB |
6750 | } |
6751 | ||
c25b8f99 | 6752 | static void |
fb5f0bc8 | 6753 | ztest_resume_thread(void *arg) |
34dc7c2f | 6754 | { |
b128c09f | 6755 | spa_t *spa = arg; |
34dc7c2f BB |
6756 | |
6757 | while (!ztest_exiting) { | |
428870ff BB |
6758 | if (spa_suspended(spa)) |
6759 | ztest_resume(spa); | |
6760 | (void) poll(NULL, 0, 100); | |
d3c2ae1c GW |
6761 | |
6762 | /* | |
6763 | * Periodically change the zfs_compressed_arc_enabled setting. | |
6764 | */ | |
6765 | if (ztest_random(10) == 0) | |
6766 | zfs_compressed_arc_enabled = ztest_random(2); | |
a6255b7f DQ |
6767 | |
6768 | /* | |
6769 | * Periodically change the zfs_abd_scatter_enabled setting. | |
6770 | */ | |
6771 | if (ztest_random(10) == 0) | |
6772 | zfs_abd_scatter_enabled = ztest_random(2); | |
34dc7c2f | 6773 | } |
34dc7c2f | 6774 | |
1e33ac1e | 6775 | thread_exit(); |
1e33ac1e | 6776 | } |
428870ff | 6777 | |
1e33ac1e | 6778 | static void |
8fb1ede1 | 6779 | ztest_deadman_thread(void *arg) |
1e33ac1e | 6780 | { |
8fb1ede1 BB |
6781 | ztest_shared_t *zs = arg; |
6782 | spa_t *spa = ztest_spa; | |
94102578 | 6783 | hrtime_t delay, overdue, last_run = gethrtime(); |
8fb1ede1 | 6784 | |
94102578 TC |
6785 | delay = (zs->zs_thread_stop - zs->zs_thread_start) + |
6786 | MSEC2NSEC(zfs_deadman_synctime_ms); | |
8fb1ede1 | 6787 | |
94102578 TC |
6788 | while (!ztest_exiting) { |
6789 | /* | |
6790 | * Wait for the delay timer while checking occasionally | |
6791 | * if we should stop. | |
6792 | */ | |
6793 | if (gethrtime() < last_run + delay) { | |
6794 | (void) poll(NULL, 0, 1000); | |
6795 | continue; | |
6796 | } | |
8fb1ede1 BB |
6797 | |
6798 | /* | |
6799 | * If the pool is suspended then fail immediately. Otherwise, | |
6800 | * check to see if the pool is making any progress. If | |
6801 | * vdev_deadman() discovers that there hasn't been any recent | |
6802 | * I/Os then it will end up aborting the tests. | |
6803 | */ | |
6804 | if (spa_suspended(spa) || spa->spa_root_vdev == NULL) { | |
6805 | fatal(0, "aborting test after %llu seconds because " | |
6806 | "pool has transitioned to a suspended state.", | |
6807 | zfs_deadman_synctime_ms / 1000); | |
6808 | } | |
6809 | vdev_deadman(spa->spa_root_vdev, FTAG); | |
6810 | ||
6811 | /* | |
6812 | * If the process doesn't complete within a grace period of | |
6813 | * zfs_deadman_synctime_ms over the expected finish time, | |
6814 | * then it may be hung and is terminated. | |
6815 | */ | |
6816 | overdue = zs->zs_proc_stop + MSEC2NSEC(zfs_deadman_synctime_ms); | |
8fb1ede1 BB |
6817 | if (gethrtime() > overdue) { |
6818 | fatal(0, "aborting test after %llu seconds because " | |
94102578 TC |
6819 | "the process is overdue for termination.", |
6820 | (gethrtime() - zs->zs_proc_start) / NANOSEC); | |
8fb1ede1 BB |
6821 | } |
6822 | ||
6823 | (void) printf("ztest has been running for %lld seconds\n", | |
94102578 TC |
6824 | (gethrtime() - zs->zs_proc_start) / NANOSEC); |
6825 | ||
6826 | last_run = gethrtime(); | |
6827 | delay = MSEC2NSEC(zfs_deadman_checktime_ms); | |
8fb1ede1 | 6828 | } |
94102578 TC |
6829 | |
6830 | thread_exit(); | |
428870ff BB |
6831 | } |
6832 | ||
6833 | static void | |
c242c188 | 6834 | ztest_execute(int test, ztest_info_t *zi, uint64_t id) |
428870ff | 6835 | { |
c242c188 CS |
6836 | ztest_ds_t *zd = &ztest_ds[id % ztest_opts.zo_datasets]; |
6837 | ztest_shared_callstate_t *zc = ZTEST_GET_SHARED_CALLSTATE(test); | |
428870ff | 6838 | hrtime_t functime = gethrtime(); |
d6320ddb | 6839 | int i; |
428870ff | 6840 | |
d6320ddb | 6841 | for (i = 0; i < zi->zi_iters; i++) |
428870ff BB |
6842 | zi->zi_func(zd, id); |
6843 | ||
6844 | functime = gethrtime() - functime; | |
6845 | ||
c242c188 CS |
6846 | atomic_add_64(&zc->zc_count, 1); |
6847 | atomic_add_64(&zc->zc_time, functime); | |
428870ff | 6848 | |
fdc5d982 | 6849 | if (ztest_opts.zo_verbose >= 4) |
428870ff | 6850 | (void) printf("%6.2f sec in %s\n", |
fdc5d982 | 6851 | (double)functime / NANOSEC, zi->zi_funcname); |
428870ff BB |
6852 | } |
6853 | ||
c25b8f99 | 6854 | static void |
34dc7c2f BB |
6855 | ztest_thread(void *arg) |
6856 | { | |
c242c188 | 6857 | int rand; |
428870ff | 6858 | uint64_t id = (uintptr_t)arg; |
34dc7c2f | 6859 | ztest_shared_t *zs = ztest_shared; |
428870ff BB |
6860 | uint64_t call_next; |
6861 | hrtime_t now; | |
34dc7c2f | 6862 | ztest_info_t *zi; |
c242c188 | 6863 | ztest_shared_callstate_t *zc; |
34dc7c2f | 6864 | |
428870ff | 6865 | while ((now = gethrtime()) < zs->zs_thread_stop) { |
34dc7c2f BB |
6866 | /* |
6867 | * See if it's time to force a crash. | |
6868 | */ | |
428870ff BB |
6869 | if (now > zs->zs_thread_kill) |
6870 | ztest_kill(zs); | |
34dc7c2f BB |
6871 | |
6872 | /* | |
428870ff | 6873 | * If we're getting ENOSPC with some regularity, stop. |
34dc7c2f | 6874 | */ |
428870ff BB |
6875 | if (zs->zs_enospc_count > 10) |
6876 | break; | |
34dc7c2f BB |
6877 | |
6878 | /* | |
428870ff | 6879 | * Pick a random function to execute. |
34dc7c2f | 6880 | */ |
c242c188 CS |
6881 | rand = ztest_random(ZTEST_FUNCS); |
6882 | zi = &ztest_info[rand]; | |
6883 | zc = ZTEST_GET_SHARED_CALLSTATE(rand); | |
6884 | call_next = zc->zc_next; | |
34dc7c2f | 6885 | |
428870ff | 6886 | if (now >= call_next && |
c242c188 CS |
6887 | atomic_cas_64(&zc->zc_next, call_next, call_next + |
6888 | ztest_random(2 * zi->zi_interval[0] + 1)) == call_next) { | |
6889 | ztest_execute(rand, zi, id); | |
6890 | } | |
428870ff | 6891 | } |
34dc7c2f | 6892 | |
1e33ac1e | 6893 | thread_exit(); |
428870ff | 6894 | } |
34dc7c2f | 6895 | |
428870ff BB |
6896 | static void |
6897 | ztest_dataset_name(char *dsname, char *pool, int d) | |
6898 | { | |
eca7b760 | 6899 | (void) snprintf(dsname, ZFS_MAX_DATASET_NAME_LEN, "%s/ds_%d", pool, d); |
428870ff | 6900 | } |
34dc7c2f | 6901 | |
428870ff | 6902 | static void |
c242c188 | 6903 | ztest_dataset_destroy(int d) |
428870ff | 6904 | { |
eca7b760 | 6905 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
d6320ddb | 6906 | int t; |
34dc7c2f | 6907 | |
c242c188 | 6908 | ztest_dataset_name(name, ztest_opts.zo_pool, d); |
34dc7c2f | 6909 | |
c242c188 | 6910 | if (ztest_opts.zo_verbose >= 3) |
428870ff | 6911 | (void) printf("Destroying %s to free up space\n", name); |
34dc7c2f | 6912 | |
428870ff BB |
6913 | /* |
6914 | * Cleanup any non-standard clones and snapshots. In general, | |
6915 | * ztest thread t operates on dataset (t % zopt_datasets), | |
6916 | * so there may be more than one thing to clean up. | |
6917 | */ | |
c242c188 CS |
6918 | for (t = d; t < ztest_opts.zo_threads; |
6919 | t += ztest_opts.zo_datasets) | |
428870ff BB |
6920 | ztest_dsl_dataset_cleanup(name, t); |
6921 | ||
6922 | (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL, | |
6923 | DS_FIND_SNAPSHOTS | DS_FIND_CHILDREN); | |
6924 | } | |
6925 | ||
6926 | static void | |
6927 | ztest_dataset_dirobj_verify(ztest_ds_t *zd) | |
6928 | { | |
6929 | uint64_t usedobjs, dirobjs, scratch; | |
6930 | ||
6931 | /* | |
6932 | * ZTEST_DIROBJ is the object directory for the entire dataset. | |
6933 | * Therefore, the number of objects in use should equal the | |
6934 | * number of ZTEST_DIROBJ entries, +1 for ZTEST_DIROBJ itself. | |
6935 | * If not, we have an object leak. | |
6936 | * | |
6937 | * Note that we can only check this in ztest_dataset_open(), | |
6938 | * when the open-context and syncing-context values agree. | |
6939 | * That's because zap_count() returns the open-context value, | |
6940 | * while dmu_objset_space() returns the rootbp fill count. | |
6941 | */ | |
6942 | VERIFY3U(0, ==, zap_count(zd->zd_os, ZTEST_DIROBJ, &dirobjs)); | |
6943 | dmu_objset_space(zd->zd_os, &scratch, &scratch, &usedobjs, &scratch); | |
6944 | ASSERT3U(dirobjs + 1, ==, usedobjs); | |
6945 | } | |
6946 | ||
6947 | static int | |
c242c188 | 6948 | ztest_dataset_open(int d) |
428870ff | 6949 | { |
c242c188 CS |
6950 | ztest_ds_t *zd = &ztest_ds[d]; |
6951 | uint64_t committed_seq = ZTEST_GET_SHARED_DS(d)->zd_seq; | |
428870ff BB |
6952 | objset_t *os; |
6953 | zilog_t *zilog; | |
eca7b760 | 6954 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
6955 | int error; |
6956 | ||
c242c188 | 6957 | ztest_dataset_name(name, ztest_opts.zo_pool, d); |
428870ff | 6958 | |
2d9142c9 | 6959 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
428870ff BB |
6960 | |
6961 | error = ztest_dataset_create(name); | |
6962 | if (error == ENOSPC) { | |
2d9142c9 | 6963 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
428870ff BB |
6964 | ztest_record_enospc(FTAG); |
6965 | return (error); | |
34dc7c2f | 6966 | } |
428870ff | 6967 | ASSERT(error == 0 || error == EEXIST); |
34dc7c2f | 6968 | |
4807c0ba TC |
6969 | VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, |
6970 | B_TRUE, zd, &os)); | |
2d9142c9 | 6971 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
428870ff | 6972 | |
c242c188 | 6973 | ztest_zd_init(zd, ZTEST_GET_SHARED_DS(d), os); |
428870ff BB |
6974 | |
6975 | zilog = zd->zd_zilog; | |
6976 | ||
6977 | if (zilog->zl_header->zh_claim_lr_seq != 0 && | |
6978 | zilog->zl_header->zh_claim_lr_seq < committed_seq) | |
6979 | fatal(0, "missing log records: claimed %llu < committed %llu", | |
6980 | zilog->zl_header->zh_claim_lr_seq, committed_seq); | |
6981 | ||
6982 | ztest_dataset_dirobj_verify(zd); | |
6983 | ||
6984 | zil_replay(os, zd, ztest_replay_vector); | |
6985 | ||
6986 | ztest_dataset_dirobj_verify(zd); | |
6987 | ||
c242c188 | 6988 | if (ztest_opts.zo_verbose >= 6) |
428870ff BB |
6989 | (void) printf("%s replay %llu blocks, %llu records, seq %llu\n", |
6990 | zd->zd_name, | |
6991 | (u_longlong_t)zilog->zl_parse_blk_count, | |
6992 | (u_longlong_t)zilog->zl_parse_lr_count, | |
6993 | (u_longlong_t)zilog->zl_replaying_seq); | |
6994 | ||
6995 | zilog = zil_open(os, ztest_get_data); | |
6996 | ||
6997 | if (zilog->zl_replaying_seq != 0 && | |
6998 | zilog->zl_replaying_seq < committed_seq) | |
6999 | fatal(0, "missing log records: replayed %llu < committed %llu", | |
7000 | zilog->zl_replaying_seq, committed_seq); | |
7001 | ||
7002 | return (0); | |
7003 | } | |
7004 | ||
7005 | static void | |
c242c188 | 7006 | ztest_dataset_close(int d) |
428870ff | 7007 | { |
c242c188 | 7008 | ztest_ds_t *zd = &ztest_ds[d]; |
428870ff BB |
7009 | |
7010 | zil_close(zd->zd_zilog); | |
b5256303 | 7011 | dmu_objset_disown(zd->zd_os, B_TRUE, zd); |
428870ff BB |
7012 | |
7013 | ztest_zd_fini(zd); | |
34dc7c2f BB |
7014 | } |
7015 | ||
f44ad929 TC |
7016 | /* ARGSUSED */ |
7017 | static int | |
7018 | ztest_replay_zil_cb(const char *name, void *arg) | |
7019 | { | |
7020 | objset_t *os; | |
7021 | ztest_ds_t *zdtmp; | |
7022 | ||
7023 | VERIFY0(ztest_dmu_objset_own(name, DMU_OST_ANY, B_TRUE, | |
7024 | B_TRUE, FTAG, &os)); | |
7025 | ||
7026 | zdtmp = umem_alloc(sizeof (ztest_ds_t), UMEM_NOFAIL); | |
7027 | ||
7028 | ztest_zd_init(zdtmp, NULL, os); | |
7029 | zil_replay(os, zdtmp, ztest_replay_vector); | |
7030 | ztest_zd_fini(zdtmp); | |
7031 | ||
7032 | if (dmu_objset_zil(os)->zl_parse_lr_count != 0 && | |
7033 | ztest_opts.zo_verbose >= 6) { | |
7034 | zilog_t *zilog = dmu_objset_zil(os); | |
7035 | ||
7036 | (void) printf("%s replay %llu blocks, %llu records, seq %llu\n", | |
7037 | name, | |
7038 | (u_longlong_t)zilog->zl_parse_blk_count, | |
7039 | (u_longlong_t)zilog->zl_parse_lr_count, | |
7040 | (u_longlong_t)zilog->zl_replaying_seq); | |
7041 | } | |
7042 | ||
7043 | umem_free(zdtmp, sizeof (ztest_ds_t)); | |
7044 | ||
7045 | dmu_objset_disown(os, B_TRUE, FTAG); | |
7046 | return (0); | |
7047 | } | |
7048 | ||
c1f3de18 BB |
7049 | static void |
7050 | ztest_freeze(void) | |
7051 | { | |
7052 | ztest_ds_t *zd = &ztest_ds[0]; | |
7053 | spa_t *spa; | |
7054 | int numloops = 0; | |
7055 | ||
7056 | if (ztest_opts.zo_verbose >= 3) | |
7057 | (void) printf("testing spa_freeze()...\n"); | |
7058 | ||
7059 | kernel_init(SPA_MODE_READ | SPA_MODE_WRITE); | |
7060 | VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); | |
7061 | VERIFY3U(0, ==, ztest_dataset_open(0)); | |
7062 | ztest_spa = spa; | |
7063 | ||
7064 | /* | |
7065 | * Force the first log block to be transactionally allocated. | |
7066 | * We have to do this before we freeze the pool -- otherwise | |
7067 | * the log chain won't be anchored. | |
7068 | */ | |
7069 | while (BP_IS_HOLE(&zd->zd_zilog->zl_header->zh_log)) { | |
7070 | ztest_dmu_object_alloc_free(zd, 0); | |
7071 | zil_commit(zd->zd_zilog, 0); | |
7072 | } | |
7073 | ||
7074 | txg_wait_synced(spa_get_dsl(spa), 0); | |
7075 | ||
7076 | /* | |
7077 | * Freeze the pool. This stops spa_sync() from doing anything, | |
7078 | * so that the only way to record changes from now on is the ZIL. | |
7079 | */ | |
7080 | spa_freeze(spa); | |
7081 | ||
7082 | /* | |
7083 | * Because it is hard to predict how much space a write will actually | |
7084 | * require beforehand, we leave ourselves some fudge space to write over | |
7085 | * capacity. | |
7086 | */ | |
7087 | uint64_t capacity = metaslab_class_get_space(spa_normal_class(spa)) / 2; | |
7088 | ||
7089 | /* | |
7090 | * Run tests that generate log records but don't alter the pool config | |
7091 | * or depend on DSL sync tasks (snapshots, objset create/destroy, etc). | |
7092 | * We do a txg_wait_synced() after each iteration to force the txg | |
7093 | * to increase well beyond the last synced value in the uberblock. | |
7094 | * The ZIL should be OK with that. | |
7095 | * | |
7096 | * Run a random number of times less than zo_maxloops and ensure we do | |
7097 | * not run out of space on the pool. | |
7098 | */ | |
7099 | while (ztest_random(10) != 0 && | |
7100 | numloops++ < ztest_opts.zo_maxloops && | |
7101 | metaslab_class_get_alloc(spa_normal_class(spa)) < capacity) { | |
7102 | ztest_od_t od; | |
7103 | ztest_od_init(&od, 0, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0); | |
7104 | VERIFY0(ztest_object_init(zd, &od, sizeof (od), B_FALSE)); | |
7105 | ztest_io(zd, od.od_object, | |
7106 | ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
7107 | txg_wait_synced(spa_get_dsl(spa), 0); | |
7108 | } | |
7109 | ||
7110 | /* | |
7111 | * Commit all of the changes we just generated. | |
7112 | */ | |
7113 | zil_commit(zd->zd_zilog, 0); | |
7114 | txg_wait_synced(spa_get_dsl(spa), 0); | |
7115 | ||
7116 | /* | |
7117 | * Close our dataset and close the pool. | |
7118 | */ | |
7119 | ztest_dataset_close(0); | |
7120 | spa_close(spa, FTAG); | |
7121 | kernel_fini(); | |
7122 | ||
7123 | /* | |
7124 | * Open and close the pool and dataset to induce log replay. | |
7125 | */ | |
7126 | kernel_init(SPA_MODE_READ | SPA_MODE_WRITE); | |
7127 | VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); | |
7128 | ASSERT(spa_freeze_txg(spa) == UINT64_MAX); | |
7129 | VERIFY3U(0, ==, ztest_dataset_open(0)); | |
7130 | ztest_spa = spa; | |
7131 | txg_wait_synced(spa_get_dsl(spa), 0); | |
7132 | ztest_dataset_close(0); | |
7133 | ztest_reguid(NULL, 0); | |
7134 | ||
7135 | spa_close(spa, FTAG); | |
7136 | kernel_fini(); | |
7137 | } | |
7138 | ||
7139 | static void | |
7140 | ztest_import_impl(ztest_shared_t *zs) | |
7141 | { | |
7142 | importargs_t args = { 0 }; | |
7143 | nvlist_t *cfg = NULL; | |
7144 | int nsearch = 1; | |
7145 | char *searchdirs[nsearch]; | |
7146 | int flags = ZFS_IMPORT_MISSING_LOG; | |
7147 | ||
7148 | searchdirs[0] = ztest_opts.zo_dir; | |
7149 | args.paths = nsearch; | |
7150 | args.path = searchdirs; | |
7151 | args.can_be_active = B_FALSE; | |
7152 | ||
7153 | VERIFY0(zpool_find_config(NULL, ztest_opts.zo_pool, &cfg, &args, | |
7154 | &libzpool_config_ops)); | |
7155 | VERIFY0(spa_import(ztest_opts.zo_pool, cfg, NULL, flags)); | |
7156 | } | |
7157 | ||
7158 | /* | |
7159 | * Import a storage pool with the given name. | |
7160 | */ | |
7161 | static void | |
7162 | ztest_import(ztest_shared_t *zs) | |
7163 | { | |
7164 | spa_t *spa; | |
7165 | ||
7166 | mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); | |
7167 | mutex_init(&ztest_checkpoint_lock, NULL, MUTEX_DEFAULT, NULL); | |
7168 | VERIFY0(pthread_rwlock_init(&ztest_name_lock, NULL)); | |
7169 | ||
7170 | kernel_init(SPA_MODE_READ | SPA_MODE_WRITE); | |
7171 | ||
7172 | ztest_import_impl(zs); | |
7173 | ||
7174 | VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG)); | |
7175 | zs->zs_metaslab_sz = | |
7176 | 1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift; | |
7177 | spa_close(spa, FTAG); | |
7178 | ||
7179 | kernel_fini(); | |
7180 | ||
7181 | if (!ztest_opts.zo_mmp_test) { | |
7182 | ztest_run_zdb(ztest_opts.zo_pool); | |
7183 | ztest_freeze(); | |
7184 | ztest_run_zdb(ztest_opts.zo_pool); | |
7185 | } | |
7186 | ||
7187 | (void) pthread_rwlock_destroy(&ztest_name_lock); | |
7188 | mutex_destroy(&ztest_vdev_lock); | |
7189 | mutex_destroy(&ztest_checkpoint_lock); | |
7190 | } | |
7191 | ||
34dc7c2f BB |
7192 | /* |
7193 | * Kick off threads to run tests on all datasets in parallel. | |
7194 | */ | |
7195 | static void | |
428870ff | 7196 | ztest_run(ztest_shared_t *zs) |
34dc7c2f | 7197 | { |
34dc7c2f | 7198 | spa_t *spa; |
3541dc6d | 7199 | objset_t *os; |
94102578 | 7200 | kthread_t *resume_thread, *deadman_thread; |
c25b8f99 | 7201 | kthread_t **run_threads; |
1e33ac1e | 7202 | uint64_t object; |
428870ff | 7203 | int error; |
d6320ddb | 7204 | int t, d; |
b128c09f BB |
7205 | |
7206 | ztest_exiting = B_FALSE; | |
34dc7c2f | 7207 | |
34dc7c2f | 7208 | /* |
428870ff | 7209 | * Initialize parent/child shared state. |
34dc7c2f | 7210 | */ |
c242c188 | 7211 | mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); |
d2734cce | 7212 | mutex_init(&ztest_checkpoint_lock, NULL, MUTEX_DEFAULT, NULL); |
2d9142c9 | 7213 | VERIFY0(pthread_rwlock_init(&ztest_name_lock, NULL)); |
34dc7c2f | 7214 | |
428870ff | 7215 | zs->zs_thread_start = gethrtime(); |
c242c188 CS |
7216 | zs->zs_thread_stop = |
7217 | zs->zs_thread_start + ztest_opts.zo_passtime * NANOSEC; | |
428870ff BB |
7218 | zs->zs_thread_stop = MIN(zs->zs_thread_stop, zs->zs_proc_stop); |
7219 | zs->zs_thread_kill = zs->zs_thread_stop; | |
c242c188 CS |
7220 | if (ztest_random(100) < ztest_opts.zo_killrate) { |
7221 | zs->zs_thread_kill -= | |
7222 | ztest_random(ztest_opts.zo_passtime * NANOSEC); | |
7223 | } | |
34dc7c2f | 7224 | |
1e33ac1e | 7225 | mutex_init(&zcl.zcl_callbacks_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 7226 | |
428870ff BB |
7227 | list_create(&zcl.zcl_callbacks, sizeof (ztest_cb_data_t), |
7228 | offsetof(ztest_cb_data_t, zcd_node)); | |
34dc7c2f BB |
7229 | |
7230 | /* | |
c1f3de18 BB |
7231 | * Open our pool. It may need to be imported first depending on |
7232 | * what tests were running when the previous pass was terminated. | |
34dc7c2f | 7233 | */ |
da92d5cb | 7234 | kernel_init(SPA_MODE_READ | SPA_MODE_WRITE); |
c1f3de18 BB |
7235 | error = spa_open(ztest_opts.zo_pool, &spa, FTAG); |
7236 | if (error) { | |
7237 | VERIFY3S(error, ==, ENOENT); | |
7238 | ztest_import_impl(zs); | |
7239 | VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG)); | |
7240 | zs->zs_metaslab_sz = | |
7241 | 1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift; | |
7242 | } | |
7243 | ||
080b3100 | 7244 | metaslab_preload_limit = ztest_random(20) + 1; |
c242c188 | 7245 | ztest_spa = spa; |
428870ff | 7246 | |
e5db3134 BB |
7247 | VERIFY0(vdev_raidz_impl_set("cycle")); |
7248 | ||
546d32ca | 7249 | dmu_objset_stats_t dds; |
4807c0ba | 7250 | VERIFY0(ztest_dmu_objset_own(ztest_opts.zo_pool, |
b5256303 | 7251 | DMU_OST_ANY, B_TRUE, B_TRUE, FTAG, &os)); |
546d32ca BB |
7252 | dsl_pool_config_enter(dmu_objset_pool(os), FTAG); |
7253 | dmu_objset_fast_stat(os, &dds); | |
7254 | dsl_pool_config_exit(dmu_objset_pool(os), FTAG); | |
7255 | zs->zs_guid = dds.dds_guid; | |
b5256303 | 7256 | dmu_objset_disown(os, B_TRUE, FTAG); |
3541dc6d | 7257 | |
34dc7c2f | 7258 | /* |
b128c09f | 7259 | * Create a thread to periodically resume suspended I/O. |
34dc7c2f | 7260 | */ |
c25b8f99 BB |
7261 | resume_thread = thread_create(NULL, 0, ztest_resume_thread, |
7262 | spa, 0, NULL, TS_RUN | TS_JOINABLE, defclsyspri); | |
34dc7c2f | 7263 | |
428870ff | 7264 | /* |
8fb1ede1 | 7265 | * Create a deadman thread and set to panic if we hang. |
428870ff | 7266 | */ |
94102578 | 7267 | deadman_thread = thread_create(NULL, 0, ztest_deadman_thread, |
8fb1ede1 BB |
7268 | zs, 0, NULL, TS_RUN | TS_JOINABLE, defclsyspri); |
7269 | ||
7270 | spa->spa_deadman_failmode = ZIO_FAILURE_MODE_PANIC; | |
428870ff | 7271 | |
34dc7c2f | 7272 | /* |
d2734cce | 7273 | * Verify that we can safely inquire about any object, |
34dc7c2f BB |
7274 | * whether it's allocated or not. To make it interesting, |
7275 | * we probe a 5-wide window around each power of two. | |
7276 | * This hits all edge cases, including zero and the max. | |
7277 | */ | |
d6320ddb BB |
7278 | for (t = 0; t < 64; t++) { |
7279 | for (d = -5; d <= 5; d++) { | |
34dc7c2f BB |
7280 | error = dmu_object_info(spa->spa_meta_objset, |
7281 | (1ULL << t) + d, NULL); | |
7282 | ASSERT(error == 0 || error == ENOENT || | |
7283 | error == EINVAL); | |
7284 | } | |
7285 | } | |
7286 | ||
7287 | /* | |
428870ff | 7288 | * If we got any ENOSPC errors on the previous run, destroy something. |
34dc7c2f | 7289 | */ |
428870ff | 7290 | if (zs->zs_enospc_count != 0) { |
c242c188 CS |
7291 | int d = ztest_random(ztest_opts.zo_datasets); |
7292 | ztest_dataset_destroy(d); | |
428870ff | 7293 | } |
34dc7c2f BB |
7294 | zs->zs_enospc_count = 0; |
7295 | ||
c40a1124 TC |
7296 | /* |
7297 | * If we were in the middle of ztest_device_removal() and were killed | |
7298 | * we need to ensure the removal and scrub complete before running | |
7299 | * any tests that check ztest_device_removal_active. The removal will | |
7300 | * be restarted automatically when the spa is opened, but we need to | |
d611989f | 7301 | * initiate the scrub manually if it is not already in progress. Note |
c40a1124 TC |
7302 | * that we always run the scrub whenever an indirect vdev exists |
7303 | * because we have no way of knowing for sure if ztest_device_removal() | |
7304 | * fully completed its scrub before the pool was reimported. | |
7305 | */ | |
7306 | if (spa->spa_removing_phys.sr_state == DSS_SCANNING || | |
7307 | spa->spa_removing_phys.sr_prev_indirect_vdev != -1) { | |
7308 | while (spa->spa_removing_phys.sr_state == DSS_SCANNING) | |
7309 | txg_wait_synced(spa_get_dsl(spa), 0); | |
7310 | ||
ce5fb2a7 BB |
7311 | error = ztest_scrub_impl(spa); |
7312 | if (error == EBUSY) | |
7313 | error = 0; | |
7314 | ASSERT0(error); | |
c40a1124 TC |
7315 | } |
7316 | ||
c25b8f99 | 7317 | run_threads = umem_zalloc(ztest_opts.zo_threads * sizeof (kthread_t *), |
c242c188 | 7318 | UMEM_NOFAIL); |
34dc7c2f | 7319 | |
c242c188 | 7320 | if (ztest_opts.zo_verbose >= 4) |
34dc7c2f BB |
7321 | (void) printf("starting main threads...\n"); |
7322 | ||
f44ad929 TC |
7323 | /* |
7324 | * Replay all logs of all datasets in the pool. This is primarily for | |
7325 | * temporary datasets which wouldn't otherwise get replayed, which | |
7326 | * can trigger failures when attempting to offline a SLOG in | |
7327 | * ztest_fault_inject(). | |
7328 | */ | |
7329 | (void) dmu_objset_find(ztest_opts.zo_pool, ztest_replay_zil_cb, | |
7330 | NULL, DS_FIND_CHILDREN); | |
7331 | ||
428870ff BB |
7332 | /* |
7333 | * Kick off all the tests that run in parallel. | |
7334 | */ | |
c242c188 | 7335 | for (t = 0; t < ztest_opts.zo_threads; t++) { |
c25b8f99 BB |
7336 | if (t < ztest_opts.zo_datasets && ztest_dataset_open(t) != 0) { |
7337 | umem_free(run_threads, ztest_opts.zo_threads * | |
7338 | sizeof (kthread_t *)); | |
428870ff | 7339 | return; |
06cf4d98 | 7340 | } |
1e33ac1e | 7341 | |
c25b8f99 BB |
7342 | run_threads[t] = thread_create(NULL, 0, ztest_thread, |
7343 | (void *)(uintptr_t)t, 0, NULL, TS_RUN | TS_JOINABLE, | |
7344 | defclsyspri); | |
34dc7c2f BB |
7345 | } |
7346 | ||
428870ff | 7347 | /* |
a2d88f77 | 7348 | * Wait for all of the tests to complete. |
428870ff | 7349 | */ |
a2d88f77 | 7350 | for (t = 0; t < ztest_opts.zo_threads; t++) |
c25b8f99 | 7351 | VERIFY0(thread_join(run_threads[t])); |
a2d88f77 TC |
7352 | |
7353 | /* | |
7354 | * Close all datasets. This must be done after all the threads | |
7355 | * are joined so we can be sure none of the datasets are in-use | |
7356 | * by any of the threads. | |
7357 | */ | |
7358 | for (t = 0; t < ztest_opts.zo_threads; t++) { | |
c242c188 CS |
7359 | if (t < ztest_opts.zo_datasets) |
7360 | ztest_dataset_close(t); | |
34dc7c2f BB |
7361 | } |
7362 | ||
34dc7c2f BB |
7363 | txg_wait_synced(spa_get_dsl(spa), 0); |
7364 | ||
428870ff BB |
7365 | zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(spa)); |
7366 | zs->zs_space = metaslab_class_get_space(spa_normal_class(spa)); | |
7367 | ||
c25b8f99 | 7368 | umem_free(run_threads, ztest_opts.zo_threads * sizeof (kthread_t *)); |
428870ff | 7369 | |
94102578 | 7370 | /* Kill the resume and deadman threads */ |
428870ff | 7371 | ztest_exiting = B_TRUE; |
c25b8f99 | 7372 | VERIFY0(thread_join(resume_thread)); |
94102578 | 7373 | VERIFY0(thread_join(deadman_thread)); |
428870ff | 7374 | ztest_resume(spa); |
34dc7c2f BB |
7375 | |
7376 | /* | |
428870ff BB |
7377 | * Right before closing the pool, kick off a bunch of async I/O; |
7378 | * spa_close() should wait for it to complete. | |
34dc7c2f | 7379 | */ |
fcff0f35 PD |
7380 | for (object = 1; object < 50; object++) { |
7381 | dmu_prefetch(spa->spa_meta_objset, object, 0, 0, 1ULL << 20, | |
7382 | ZIO_PRIORITY_SYNC_READ); | |
7383 | } | |
428870ff | 7384 | |
090ff092 RC |
7385 | /* Verify that at least one commit cb was called in a timely fashion */ |
7386 | if (zc_cb_counter >= ZTEST_COMMIT_CB_MIN_REG) | |
c99c9001 | 7387 | VERIFY0(zc_min_txg_delay); |
090ff092 | 7388 | |
428870ff BB |
7389 | spa_close(spa, FTAG); |
7390 | ||
7391 | /* | |
7392 | * Verify that we can loop over all pools. | |
7393 | */ | |
7394 | mutex_enter(&spa_namespace_lock); | |
7395 | for (spa = spa_next(NULL); spa != NULL; spa = spa_next(spa)) | |
c242c188 | 7396 | if (ztest_opts.zo_verbose > 3) |
428870ff BB |
7397 | (void) printf("spa_next: found %s\n", spa_name(spa)); |
7398 | mutex_exit(&spa_namespace_lock); | |
7399 | ||
7400 | /* | |
7401 | * Verify that we can export the pool and reimport it under a | |
7402 | * different name. | |
7403 | */ | |
379ca9cf | 7404 | if ((ztest_random(2) == 0) && !ztest_opts.zo_mmp_test) { |
eca7b760 IK |
7405 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
7406 | (void) snprintf(name, sizeof (name), "%s_import", | |
c242c188 CS |
7407 | ztest_opts.zo_pool); |
7408 | ztest_spa_import_export(ztest_opts.zo_pool, name); | |
7409 | ztest_spa_import_export(name, ztest_opts.zo_pool); | |
428870ff BB |
7410 | } |
7411 | ||
7412 | kernel_fini(); | |
572e2857 BB |
7413 | |
7414 | list_destroy(&zcl.zcl_callbacks); | |
0e8d1b2d | 7415 | mutex_destroy(&zcl.zcl_callbacks_lock); |
2d9142c9 | 7416 | (void) pthread_rwlock_destroy(&ztest_name_lock); |
c242c188 | 7417 | mutex_destroy(&ztest_vdev_lock); |
d2734cce | 7418 | mutex_destroy(&ztest_checkpoint_lock); |
428870ff BB |
7419 | } |
7420 | ||
65c7cc49 | 7421 | static void |
34dc7c2f BB |
7422 | print_time(hrtime_t t, char *timebuf) |
7423 | { | |
7424 | hrtime_t s = t / NANOSEC; | |
7425 | hrtime_t m = s / 60; | |
7426 | hrtime_t h = m / 60; | |
7427 | hrtime_t d = h / 24; | |
7428 | ||
7429 | s -= m * 60; | |
7430 | m -= h * 60; | |
7431 | h -= d * 24; | |
7432 | ||
7433 | timebuf[0] = '\0'; | |
7434 | ||
7435 | if (d) | |
7436 | (void) sprintf(timebuf, | |
7437 | "%llud%02lluh%02llum%02llus", d, h, m, s); | |
7438 | else if (h) | |
7439 | (void) sprintf(timebuf, "%lluh%02llum%02llus", h, m, s); | |
7440 | else if (m) | |
7441 | (void) sprintf(timebuf, "%llum%02llus", m, s); | |
7442 | else | |
7443 | (void) sprintf(timebuf, "%llus", s); | |
7444 | } | |
7445 | ||
428870ff | 7446 | static nvlist_t * |
0bc8fd78 | 7447 | make_random_props(void) |
428870ff BB |
7448 | { |
7449 | nvlist_t *props; | |
7450 | ||
60b82074 | 7451 | VERIFY0(nvlist_alloc(&props, NV_UNIQUE_NAME, 0)); |
cc99f275 | 7452 | |
9ae529ec CS |
7453 | if (ztest_random(2) == 0) |
7454 | return (props); | |
60b82074 BB |
7455 | |
7456 | VERIFY0(nvlist_add_uint64(props, | |
7457 | zpool_prop_to_name(ZPOOL_PROP_AUTOREPLACE), 1)); | |
428870ff | 7458 | |
428870ff BB |
7459 | return (props); |
7460 | } | |
7461 | ||
34dc7c2f BB |
7462 | /* |
7463 | * Create a storage pool with the given name and initial vdev size. | |
428870ff | 7464 | * Then test spa_freeze() functionality. |
34dc7c2f BB |
7465 | */ |
7466 | static void | |
428870ff | 7467 | ztest_init(ztest_shared_t *zs) |
34dc7c2f BB |
7468 | { |
7469 | spa_t *spa; | |
428870ff | 7470 | nvlist_t *nvroot, *props; |
9ae529ec | 7471 | int i; |
428870ff | 7472 | |
c242c188 | 7473 | mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); |
d2734cce | 7474 | mutex_init(&ztest_checkpoint_lock, NULL, MUTEX_DEFAULT, NULL); |
2d9142c9 | 7475 | VERIFY0(pthread_rwlock_init(&ztest_name_lock, NULL)); |
34dc7c2f | 7476 | |
da92d5cb | 7477 | kernel_init(SPA_MODE_READ | SPA_MODE_WRITE); |
34dc7c2f BB |
7478 | |
7479 | /* | |
7480 | * Create the storage pool. | |
7481 | */ | |
c242c188 | 7482 | (void) spa_destroy(ztest_opts.zo_pool); |
428870ff BB |
7483 | ztest_shared->zs_vdev_next_leaf = 0; |
7484 | zs->zs_splits = 0; | |
c242c188 | 7485 | zs->zs_mirrors = ztest_opts.zo_mirrors; |
ea0b2538 | 7486 | nvroot = make_vdev_root(NULL, NULL, NULL, ztest_opts.zo_vdev_size, 0, |
b2255edc | 7487 | NULL, ztest_opts.zo_raid_children, zs->zs_mirrors, 1); |
428870ff | 7488 | props = make_random_props(); |
60b82074 BB |
7489 | |
7490 | /* | |
7491 | * We don't expect the pool to suspend unless maxfaults == 0, | |
7492 | * in which case ztest_fault_inject() temporarily takes away | |
7493 | * the only valid replica. | |
7494 | */ | |
7495 | VERIFY0(nvlist_add_uint64(props, | |
7496 | zpool_prop_to_name(ZPOOL_PROP_FAILUREMODE), | |
7497 | MAXFAULTS(zs) ? ZIO_FAILURE_MODE_PANIC : ZIO_FAILURE_MODE_WAIT)); | |
7498 | ||
9ae529ec CS |
7499 | for (i = 0; i < SPA_FEATURES; i++) { |
7500 | char *buf; | |
93e28d66 SD |
7501 | |
7502 | /* | |
7503 | * 75% chance of using the log space map feature. We want ztest | |
7504 | * to exercise both the code paths that use the log space map | |
7505 | * feature and the ones that don't. | |
7506 | */ | |
7507 | if (i == SPA_FEATURE_LOG_SPACEMAP && ztest_random(4) == 0) | |
7508 | continue; | |
7509 | ||
9ae529ec CS |
7510 | VERIFY3S(-1, !=, asprintf(&buf, "feature@%s", |
7511 | spa_feature_table[i].fi_uname)); | |
7512 | VERIFY3U(0, ==, nvlist_add_uint64(props, buf, 0)); | |
7513 | free(buf); | |
7514 | } | |
60b82074 BB |
7515 | |
7516 | VERIFY0(spa_create(ztest_opts.zo_pool, nvroot, props, NULL, NULL)); | |
34dc7c2f | 7517 | nvlist_free(nvroot); |
85802aa4 | 7518 | nvlist_free(props); |
34dc7c2f | 7519 | |
c242c188 CS |
7520 | VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
7521 | zs->zs_metaslab_sz = | |
7522 | 1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift; | |
34dc7c2f BB |
7523 | spa_close(spa, FTAG); |
7524 | ||
7525 | kernel_fini(); | |
428870ff | 7526 | |
379ca9cf OF |
7527 | if (!ztest_opts.zo_mmp_test) { |
7528 | ztest_run_zdb(ztest_opts.zo_pool); | |
7529 | ztest_freeze(); | |
7530 | ztest_run_zdb(ztest_opts.zo_pool); | |
7531 | } | |
c242c188 | 7532 | |
2d9142c9 | 7533 | (void) pthread_rwlock_destroy(&ztest_name_lock); |
c242c188 | 7534 | mutex_destroy(&ztest_vdev_lock); |
d2734cce | 7535 | mutex_destroy(&ztest_checkpoint_lock); |
c242c188 CS |
7536 | } |
7537 | ||
7538 | static void | |
9d81146b | 7539 | setup_data_fd(void) |
c242c188 | 7540 | { |
facbbe43 BB |
7541 | static char ztest_name_data[] = "/tmp/ztest.data.XXXXXX"; |
7542 | ||
7543 | ztest_fd_data = mkstemp(ztest_name_data); | |
9d81146b | 7544 | ASSERT3S(ztest_fd_data, >=, 0); |
facbbe43 | 7545 | (void) unlink(ztest_name_data); |
c242c188 CS |
7546 | } |
7547 | ||
22257dc0 CS |
7548 | static int |
7549 | shared_data_size(ztest_shared_hdr_t *hdr) | |
7550 | { | |
7551 | int size; | |
7552 | ||
7553 | size = hdr->zh_hdr_size; | |
7554 | size += hdr->zh_opts_size; | |
7555 | size += hdr->zh_size; | |
7556 | size += hdr->zh_stats_size * hdr->zh_stats_count; | |
7557 | size += hdr->zh_ds_size * hdr->zh_ds_count; | |
7558 | ||
7559 | return (size); | |
7560 | } | |
7561 | ||
c242c188 CS |
7562 | static void |
7563 | setup_hdr(void) | |
7564 | { | |
22257dc0 | 7565 | int size; |
c242c188 CS |
7566 | ztest_shared_hdr_t *hdr; |
7567 | ||
7568 | hdr = (void *)mmap(0, P2ROUNDUP(sizeof (*hdr), getpagesize()), | |
9d81146b | 7569 | PROT_READ | PROT_WRITE, MAP_SHARED, ztest_fd_data, 0); |
799402d8 | 7570 | ASSERT(hdr != MAP_FAILED); |
c242c188 | 7571 | |
9d81146b | 7572 | VERIFY3U(0, ==, ftruncate(ztest_fd_data, sizeof (ztest_shared_hdr_t))); |
22257dc0 | 7573 | |
c242c188 CS |
7574 | hdr->zh_hdr_size = sizeof (ztest_shared_hdr_t); |
7575 | hdr->zh_opts_size = sizeof (ztest_shared_opts_t); | |
7576 | hdr->zh_size = sizeof (ztest_shared_t); | |
7577 | hdr->zh_stats_size = sizeof (ztest_shared_callstate_t); | |
7578 | hdr->zh_stats_count = ZTEST_FUNCS; | |
7579 | hdr->zh_ds_size = sizeof (ztest_shared_ds_t); | |
7580 | hdr->zh_ds_count = ztest_opts.zo_datasets; | |
7581 | ||
22257dc0 | 7582 | size = shared_data_size(hdr); |
9d81146b | 7583 | VERIFY3U(0, ==, ftruncate(ztest_fd_data, size)); |
22257dc0 | 7584 | |
c242c188 CS |
7585 | (void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize())); |
7586 | } | |
7587 | ||
7588 | static void | |
7589 | setup_data(void) | |
7590 | { | |
7591 | int size, offset; | |
7592 | ztest_shared_hdr_t *hdr; | |
7593 | uint8_t *buf; | |
7594 | ||
7595 | hdr = (void *)mmap(0, P2ROUNDUP(sizeof (*hdr), getpagesize()), | |
9d81146b | 7596 | PROT_READ, MAP_SHARED, ztest_fd_data, 0); |
799402d8 | 7597 | ASSERT(hdr != MAP_FAILED); |
c242c188 | 7598 | |
22257dc0 | 7599 | size = shared_data_size(hdr); |
c242c188 CS |
7600 | |
7601 | (void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize())); | |
7602 | hdr = ztest_shared_hdr = (void *)mmap(0, P2ROUNDUP(size, getpagesize()), | |
9d81146b | 7603 | PROT_READ | PROT_WRITE, MAP_SHARED, ztest_fd_data, 0); |
799402d8 | 7604 | ASSERT(hdr != MAP_FAILED); |
c242c188 CS |
7605 | buf = (uint8_t *)hdr; |
7606 | ||
7607 | offset = hdr->zh_hdr_size; | |
7608 | ztest_shared_opts = (void *)&buf[offset]; | |
7609 | offset += hdr->zh_opts_size; | |
7610 | ztest_shared = (void *)&buf[offset]; | |
7611 | offset += hdr->zh_size; | |
7612 | ztest_shared_callstate = (void *)&buf[offset]; | |
7613 | offset += hdr->zh_stats_size * hdr->zh_stats_count; | |
7614 | ztest_shared_ds = (void *)&buf[offset]; | |
7615 | } | |
7616 | ||
7617 | static boolean_t | |
7618 | exec_child(char *cmd, char *libpath, boolean_t ignorekill, int *statusp) | |
7619 | { | |
7620 | pid_t pid; | |
7621 | int status; | |
483106eb | 7622 | char *cmdbuf = NULL; |
c242c188 CS |
7623 | |
7624 | pid = fork(); | |
7625 | ||
7626 | if (cmd == NULL) { | |
483106eb BB |
7627 | cmdbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
7628 | (void) strlcpy(cmdbuf, getexecname(), MAXPATHLEN); | |
c242c188 CS |
7629 | cmd = cmdbuf; |
7630 | } | |
7631 | ||
7632 | if (pid == -1) | |
7633 | fatal(1, "fork failed"); | |
7634 | ||
7635 | if (pid == 0) { /* child */ | |
7636 | char *emptyargv[2] = { cmd, NULL }; | |
9d81146b | 7637 | char fd_data_str[12]; |
c242c188 CS |
7638 | |
7639 | struct rlimit rl = { 1024, 1024 }; | |
7640 | (void) setrlimit(RLIMIT_NOFILE, &rl); | |
9d81146b ED |
7641 | |
7642 | (void) close(ztest_fd_rand); | |
7643 | VERIFY(11 >= snprintf(fd_data_str, 12, "%d", ztest_fd_data)); | |
7644 | VERIFY(0 == setenv("ZTEST_FD_DATA", fd_data_str, 1)); | |
7645 | ||
c242c188 CS |
7646 | (void) enable_extended_FILE_stdio(-1, -1); |
7647 | if (libpath != NULL) | |
7648 | VERIFY(0 == setenv("LD_LIBRARY_PATH", libpath, 1)); | |
7649 | (void) execv(cmd, emptyargv); | |
7650 | ztest_dump_core = B_FALSE; | |
7651 | fatal(B_TRUE, "exec failed: %s", cmd); | |
7652 | } | |
7653 | ||
483106eb BB |
7654 | if (cmdbuf != NULL) { |
7655 | umem_free(cmdbuf, MAXPATHLEN); | |
7656 | cmd = NULL; | |
7657 | } | |
7658 | ||
c242c188 CS |
7659 | while (waitpid(pid, &status, 0) != pid) |
7660 | continue; | |
7661 | if (statusp != NULL) | |
7662 | *statusp = status; | |
7663 | ||
7664 | if (WIFEXITED(status)) { | |
7665 | if (WEXITSTATUS(status) != 0) { | |
7666 | (void) fprintf(stderr, "child exited with code %d\n", | |
7667 | WEXITSTATUS(status)); | |
7668 | exit(2); | |
7669 | } | |
7670 | return (B_FALSE); | |
7671 | } else if (WIFSIGNALED(status)) { | |
7672 | if (!ignorekill || WTERMSIG(status) != SIGKILL) { | |
7673 | (void) fprintf(stderr, "child died with signal %d\n", | |
7674 | WTERMSIG(status)); | |
7675 | exit(3); | |
7676 | } | |
7677 | return (B_TRUE); | |
7678 | } else { | |
7679 | (void) fprintf(stderr, "something strange happened to child\n"); | |
7680 | exit(4); | |
7681 | /* NOTREACHED */ | |
7682 | } | |
7683 | } | |
7684 | ||
7685 | static void | |
7686 | ztest_run_init(void) | |
7687 | { | |
7688 | int i; | |
428870ff | 7689 | |
c242c188 | 7690 | ztest_shared_t *zs = ztest_shared; |
428870ff | 7691 | |
c242c188 CS |
7692 | /* |
7693 | * Blow away any existing copy of zpool.cache | |
7694 | */ | |
7695 | (void) remove(spa_config_path); | |
7696 | ||
379ca9cf OF |
7697 | if (ztest_opts.zo_init == 0) { |
7698 | if (ztest_opts.zo_verbose >= 1) | |
7699 | (void) printf("Importing pool %s\n", | |
7700 | ztest_opts.zo_pool); | |
7701 | ztest_import(zs); | |
7702 | return; | |
7703 | } | |
7704 | ||
c242c188 CS |
7705 | /* |
7706 | * Create and initialize our storage pool. | |
7707 | */ | |
7708 | for (i = 1; i <= ztest_opts.zo_init; i++) { | |
7709 | bzero(zs, sizeof (ztest_shared_t)); | |
7710 | if (ztest_opts.zo_verbose >= 3 && | |
7711 | ztest_opts.zo_init != 1) { | |
7712 | (void) printf("ztest_init(), pass %d\n", i); | |
7713 | } | |
7714 | ztest_init(zs); | |
7715 | } | |
34dc7c2f BB |
7716 | } |
7717 | ||
7718 | int | |
7719 | main(int argc, char **argv) | |
7720 | { | |
7721 | int kills = 0; | |
7722 | int iters = 0; | |
c242c188 CS |
7723 | int older = 0; |
7724 | int newer = 0; | |
34dc7c2f BB |
7725 | ztest_shared_t *zs; |
7726 | ztest_info_t *zi; | |
c242c188 | 7727 | ztest_shared_callstate_t *zc; |
34dc7c2f | 7728 | char timebuf[100]; |
f3c8c9e6 | 7729 | char numbuf[NN_NUMBUF_SZ]; |
483106eb | 7730 | char *cmd; |
c242c188 CS |
7731 | boolean_t hasalt; |
7732 | int f; | |
9d81146b | 7733 | char *fd_data_str = getenv("ZTEST_FD_DATA"); |
e82cdc3a | 7734 | struct sigaction action; |
34dc7c2f BB |
7735 | |
7736 | (void) setvbuf(stdout, NULL, _IOLBF, 0); | |
7737 | ||
498877ba | 7738 | dprintf_setup(&argc, argv); |
8fb1ede1 | 7739 | zfs_deadman_synctime_ms = 300000; |
94102578 | 7740 | zfs_deadman_checktime_ms = 30000; |
4d044c4c SD |
7741 | /* |
7742 | * As two-word space map entries may not come up often (especially | |
7743 | * if pool and vdev sizes are small) we want to force at least some | |
7744 | * of them so the feature get tested. | |
7745 | */ | |
7746 | zfs_force_some_double_word_sm_entries = B_TRUE; | |
498877ba | 7747 | |
1258bd77 BB |
7748 | /* |
7749 | * Verify that even extensively damaged split blocks with many | |
7750 | * segments can be reconstructed in a reasonable amount of time | |
7751 | * when reconstruction is known to be possible. | |
4a7eb69a TC |
7752 | * |
7753 | * Note: the lower this value is, the more damage we inflict, and | |
7754 | * the more time ztest spends in recovering that damage. We chose | |
7755 | * to induce damage 1/100th of the time so recovery is tested but | |
7756 | * not so frequently that ztest doesn't get to test other code paths. | |
1258bd77 | 7757 | */ |
4a7eb69a | 7758 | zfs_reconstruct_indirect_damage_fraction = 100; |
1258bd77 | 7759 | |
e82cdc3a NB |
7760 | action.sa_handler = sig_handler; |
7761 | sigemptyset(&action.sa_mask); | |
7762 | action.sa_flags = 0; | |
7763 | ||
7764 | if (sigaction(SIGSEGV, &action, NULL) < 0) { | |
7765 | (void) fprintf(stderr, "ztest: cannot catch SIGSEGV: %s.\n", | |
7766 | strerror(errno)); | |
7767 | exit(EXIT_FAILURE); | |
7768 | } | |
7769 | ||
7770 | if (sigaction(SIGABRT, &action, NULL) < 0) { | |
7771 | (void) fprintf(stderr, "ztest: cannot catch SIGABRT: %s.\n", | |
7772 | strerror(errno)); | |
7773 | exit(EXIT_FAILURE); | |
7774 | } | |
7775 | ||
e1a0850c BB |
7776 | /* |
7777 | * Force random_get_bytes() to use /dev/urandom in order to prevent | |
7778 | * ztest from needlessly depleting the system entropy pool. | |
7779 | */ | |
7780 | random_path = "/dev/urandom"; | |
7781 | ztest_fd_rand = open(random_path, O_RDONLY); | |
9d81146b ED |
7782 | ASSERT3S(ztest_fd_rand, >=, 0); |
7783 | ||
7784 | if (!fd_data_str) { | |
c242c188 | 7785 | process_options(argc, argv); |
34dc7c2f | 7786 | |
9d81146b | 7787 | setup_data_fd(); |
c242c188 CS |
7788 | setup_hdr(); |
7789 | setup_data(); | |
7790 | bcopy(&ztest_opts, ztest_shared_opts, | |
7791 | sizeof (*ztest_shared_opts)); | |
7792 | } else { | |
9d81146b | 7793 | ztest_fd_data = atoi(fd_data_str); |
c242c188 CS |
7794 | setup_data(); |
7795 | bcopy(ztest_shared_opts, &ztest_opts, sizeof (ztest_opts)); | |
7796 | } | |
7797 | ASSERT3U(ztest_opts.zo_datasets, ==, ztest_shared_hdr->zh_ds_count); | |
34dc7c2f | 7798 | |
428870ff | 7799 | /* Override location of zpool.cache */ |
5be98cfe BB |
7800 | VERIFY(asprintf((char **)&spa_config_path, "%s/zpool.cache", |
7801 | ztest_opts.zo_dir) != -1); | |
9d81146b | 7802 | |
c242c188 CS |
7803 | ztest_ds = umem_alloc(ztest_opts.zo_datasets * sizeof (ztest_ds_t), |
7804 | UMEM_NOFAIL); | |
7805 | zs = ztest_shared; | |
7806 | ||
9d81146b | 7807 | if (fd_data_str) { |
d830d479 | 7808 | metaslab_force_ganging = ztest_opts.zo_metaslab_force_ganging; |
c242c188 CS |
7809 | metaslab_df_alloc_threshold = |
7810 | zs->zs_metaslab_df_alloc_threshold; | |
7811 | ||
7812 | if (zs->zs_do_init) | |
7813 | ztest_run_init(); | |
7814 | else | |
7815 | ztest_run(zs); | |
7816 | exit(0); | |
7817 | } | |
34dc7c2f | 7818 | |
c242c188 | 7819 | hasalt = (strlen(ztest_opts.zo_alt_ztest) != 0); |
34dc7c2f | 7820 | |
c242c188 | 7821 | if (ztest_opts.zo_verbose >= 1) { |
34dc7c2f | 7822 | (void) printf("%llu vdevs, %d datasets, %d threads," |
b2255edc | 7823 | "%d %s disks, %llu seconds...\n\n", |
c242c188 CS |
7824 | (u_longlong_t)ztest_opts.zo_vdevs, |
7825 | ztest_opts.zo_datasets, | |
7826 | ztest_opts.zo_threads, | |
b2255edc BB |
7827 | ztest_opts.zo_raid_children, |
7828 | ztest_opts.zo_raid_type, | |
c242c188 | 7829 | (u_longlong_t)ztest_opts.zo_time); |
34dc7c2f BB |
7830 | } |
7831 | ||
483106eb BB |
7832 | cmd = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); |
7833 | (void) strlcpy(cmd, getexecname(), MAXNAMELEN); | |
c242c188 CS |
7834 | |
7835 | zs->zs_do_init = B_TRUE; | |
7836 | if (strlen(ztest_opts.zo_alt_ztest) != 0) { | |
7837 | if (ztest_opts.zo_verbose >= 1) { | |
7838 | (void) printf("Executing older ztest for " | |
7839 | "initialization: %s\n", ztest_opts.zo_alt_ztest); | |
7840 | } | |
7841 | VERIFY(!exec_child(ztest_opts.zo_alt_ztest, | |
7842 | ztest_opts.zo_alt_libpath, B_FALSE, NULL)); | |
7843 | } else { | |
7844 | VERIFY(!exec_child(NULL, NULL, B_FALSE, NULL)); | |
34dc7c2f | 7845 | } |
c242c188 | 7846 | zs->zs_do_init = B_FALSE; |
34dc7c2f | 7847 | |
428870ff | 7848 | zs->zs_proc_start = gethrtime(); |
c242c188 | 7849 | zs->zs_proc_stop = zs->zs_proc_start + ztest_opts.zo_time * NANOSEC; |
34dc7c2f | 7850 | |
d6320ddb | 7851 | for (f = 0; f < ZTEST_FUNCS; f++) { |
c242c188 CS |
7852 | zi = &ztest_info[f]; |
7853 | zc = ZTEST_GET_SHARED_CALLSTATE(f); | |
428870ff | 7854 | if (zs->zs_proc_start + zi->zi_interval[0] > zs->zs_proc_stop) |
c242c188 | 7855 | zc->zc_next = UINT64_MAX; |
34dc7c2f | 7856 | else |
c242c188 | 7857 | zc->zc_next = zs->zs_proc_start + |
428870ff | 7858 | ztest_random(2 * zi->zi_interval[0] + 1); |
34dc7c2f BB |
7859 | } |
7860 | ||
34dc7c2f BB |
7861 | /* |
7862 | * Run the tests in a loop. These tests include fault injection | |
7863 | * to verify that self-healing data works, and forced crashes | |
7864 | * to verify that we never lose on-disk consistency. | |
7865 | */ | |
428870ff | 7866 | while (gethrtime() < zs->zs_proc_stop) { |
34dc7c2f | 7867 | int status; |
c242c188 | 7868 | boolean_t killed; |
34dc7c2f BB |
7869 | |
7870 | /* | |
7871 | * Initialize the workload counters for each function. | |
7872 | */ | |
d6320ddb | 7873 | for (f = 0; f < ZTEST_FUNCS; f++) { |
c242c188 CS |
7874 | zc = ZTEST_GET_SHARED_CALLSTATE(f); |
7875 | zc->zc_count = 0; | |
7876 | zc->zc_time = 0; | |
34dc7c2f BB |
7877 | } |
7878 | ||
9babb374 | 7879 | /* Set the allocation switch size */ |
c242c188 CS |
7880 | zs->zs_metaslab_df_alloc_threshold = |
7881 | ztest_random(zs->zs_metaslab_sz / 4) + 1; | |
9babb374 | 7882 | |
c242c188 CS |
7883 | if (!hasalt || ztest_random(2) == 0) { |
7884 | if (hasalt && ztest_opts.zo_verbose >= 1) { | |
7885 | (void) printf("Executing newer ztest: %s\n", | |
7886 | cmd); | |
34dc7c2f | 7887 | } |
c242c188 CS |
7888 | newer++; |
7889 | killed = exec_child(cmd, NULL, B_TRUE, &status); | |
34dc7c2f | 7890 | } else { |
c242c188 CS |
7891 | if (hasalt && ztest_opts.zo_verbose >= 1) { |
7892 | (void) printf("Executing older ztest: %s\n", | |
7893 | ztest_opts.zo_alt_ztest); | |
7894 | } | |
7895 | older++; | |
7896 | killed = exec_child(ztest_opts.zo_alt_ztest, | |
7897 | ztest_opts.zo_alt_libpath, B_TRUE, &status); | |
34dc7c2f BB |
7898 | } |
7899 | ||
c242c188 CS |
7900 | if (killed) |
7901 | kills++; | |
34dc7c2f BB |
7902 | iters++; |
7903 | ||
c242c188 | 7904 | if (ztest_opts.zo_verbose >= 1) { |
34dc7c2f BB |
7905 | hrtime_t now = gethrtime(); |
7906 | ||
428870ff BB |
7907 | now = MIN(now, zs->zs_proc_stop); |
7908 | print_time(zs->zs_proc_stop - now, timebuf); | |
f3c8c9e6 | 7909 | nicenum(zs->zs_space, numbuf, sizeof (numbuf)); |
34dc7c2f BB |
7910 | |
7911 | (void) printf("Pass %3d, %8s, %3llu ENOSPC, " | |
7912 | "%4.1f%% of %5s used, %3.0f%% done, %8s to go\n", | |
7913 | iters, | |
7914 | WIFEXITED(status) ? "Complete" : "SIGKILL", | |
7915 | (u_longlong_t)zs->zs_enospc_count, | |
7916 | 100.0 * zs->zs_alloc / zs->zs_space, | |
7917 | numbuf, | |
428870ff | 7918 | 100.0 * (now - zs->zs_proc_start) / |
c242c188 | 7919 | (ztest_opts.zo_time * NANOSEC), timebuf); |
34dc7c2f BB |
7920 | } |
7921 | ||
c242c188 | 7922 | if (ztest_opts.zo_verbose >= 2) { |
34dc7c2f BB |
7923 | (void) printf("\nWorkload summary:\n\n"); |
7924 | (void) printf("%7s %9s %s\n", | |
7925 | "Calls", "Time", "Function"); | |
7926 | (void) printf("%7s %9s %s\n", | |
7927 | "-----", "----", "--------"); | |
d6320ddb | 7928 | for (f = 0; f < ZTEST_FUNCS; f++) { |
c242c188 CS |
7929 | zi = &ztest_info[f]; |
7930 | zc = ZTEST_GET_SHARED_CALLSTATE(f); | |
7931 | print_time(zc->zc_time, timebuf); | |
34dc7c2f | 7932 | (void) printf("%7llu %9s %s\n", |
c242c188 | 7933 | (u_longlong_t)zc->zc_count, timebuf, |
fdc5d982 | 7934 | zi->zi_funcname); |
34dc7c2f BB |
7935 | } |
7936 | (void) printf("\n"); | |
7937 | } | |
7938 | ||
379ca9cf OF |
7939 | if (!ztest_opts.zo_mmp_test) |
7940 | ztest_run_zdb(ztest_opts.zo_pool); | |
428870ff | 7941 | } |
34dc7c2f | 7942 | |
c242c188 CS |
7943 | if (ztest_opts.zo_verbose >= 1) { |
7944 | if (hasalt) { | |
7945 | (void) printf("%d runs of older ztest: %s\n", older, | |
7946 | ztest_opts.zo_alt_ztest); | |
7947 | (void) printf("%d runs of newer ztest: %s\n", newer, | |
7948 | cmd); | |
7949 | } | |
34dc7c2f BB |
7950 | (void) printf("%d killed, %d completed, %.0f%% kill rate\n", |
7951 | kills, iters - kills, (100.0 * kills) / MAX(1, iters)); | |
7952 | } | |
7953 | ||
483106eb BB |
7954 | umem_free(cmd, MAXNAMELEN); |
7955 | ||
34dc7c2f BB |
7956 | return (0); |
7957 | } |