]>
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 | * | |
77 | * You can ask more more vdevs [-v], datasets [-d], or threads [-t] | |
78 | * to increase the pool capacity, fanout, and overall stress level. | |
79 | * | |
c242c188 CS |
80 | * Use the -k option to set the desired frequency of kills. |
81 | * | |
82 | * When ztest invokes itself it passes all relevant information through a | |
83 | * temporary file which is mmap-ed in the child process. This allows shared | |
84 | * memory to survive the exec syscall. The ztest_shared_hdr_t struct is always | |
85 | * stored at offset 0 of this file and contains information on the size and | |
86 | * number of shared structures in the file. The information stored in this file | |
87 | * must remain backwards compatible with older versions of ztest so that | |
88 | * ztest can invoke them during backwards compatibility testing (-B). | |
34dc7c2f BB |
89 | */ |
90 | ||
91 | #include <sys/zfs_context.h> | |
92 | #include <sys/spa.h> | |
93 | #include <sys/dmu.h> | |
94 | #include <sys/txg.h> | |
9babb374 | 95 | #include <sys/dbuf.h> |
34dc7c2f | 96 | #include <sys/zap.h> |
34dc7c2f BB |
97 | #include <sys/dmu_objset.h> |
98 | #include <sys/poll.h> | |
99 | #include <sys/stat.h> | |
100 | #include <sys/time.h> | |
101 | #include <sys/wait.h> | |
102 | #include <sys/mman.h> | |
103 | #include <sys/resource.h> | |
104 | #include <sys/zio.h> | |
34dc7c2f | 105 | #include <sys/zil.h> |
428870ff | 106 | #include <sys/zil_impl.h> |
34dc7c2f | 107 | #include <sys/vdev_impl.h> |
b128c09f | 108 | #include <sys/vdev_file.h> |
34dc7c2f | 109 | #include <sys/spa_impl.h> |
428870ff | 110 | #include <sys/metaslab_impl.h> |
34dc7c2f | 111 | #include <sys/dsl_prop.h> |
9babb374 | 112 | #include <sys/dsl_dataset.h> |
13fe0198 | 113 | #include <sys/dsl_destroy.h> |
428870ff BB |
114 | #include <sys/dsl_scan.h> |
115 | #include <sys/zio_checksum.h> | |
34dc7c2f | 116 | #include <sys/refcount.h> |
9ae529ec | 117 | #include <sys/zfeature.h> |
13fe0198 | 118 | #include <sys/dsl_userhold.h> |
a6255b7f | 119 | #include <sys/abd.h> |
34dc7c2f BB |
120 | #include <stdio.h> |
121 | #include <stdio_ext.h> | |
122 | #include <stdlib.h> | |
123 | #include <unistd.h> | |
124 | #include <signal.h> | |
125 | #include <umem.h> | |
34dc7c2f BB |
126 | #include <ctype.h> |
127 | #include <math.h> | |
128 | #include <sys/fs/zfs.h> | |
1eeb4562 | 129 | #include <zfs_fletcher.h> |
428870ff | 130 | #include <libnvpair.h> |
379ca9cf | 131 | #include <libzfs.h> |
010d1247 | 132 | #include <sys/crypto/icp.h> |
967798d0 | 133 | #ifdef __GLIBC__ |
e82cdc3a NB |
134 | #include <execinfo.h> /* for backtrace() */ |
135 | #endif | |
34dc7c2f | 136 | |
9d81146b ED |
137 | static int ztest_fd_data = -1; |
138 | static int ztest_fd_rand = -1; | |
c242c188 CS |
139 | |
140 | typedef struct ztest_shared_hdr { | |
141 | uint64_t zh_hdr_size; | |
142 | uint64_t zh_opts_size; | |
143 | uint64_t zh_size; | |
144 | uint64_t zh_stats_size; | |
145 | uint64_t zh_stats_count; | |
146 | uint64_t zh_ds_size; | |
147 | uint64_t zh_ds_count; | |
148 | } ztest_shared_hdr_t; | |
149 | ||
150 | static ztest_shared_hdr_t *ztest_shared_hdr; | |
151 | ||
cc99f275 DB |
152 | enum ztest_class_state { |
153 | ZTEST_VDEV_CLASS_OFF, | |
154 | ZTEST_VDEV_CLASS_ON, | |
155 | ZTEST_VDEV_CLASS_RND | |
156 | }; | |
157 | ||
c242c188 | 158 | typedef struct ztest_shared_opts { |
eca7b760 IK |
159 | char zo_pool[ZFS_MAX_DATASET_NAME_LEN]; |
160 | char zo_dir[ZFS_MAX_DATASET_NAME_LEN]; | |
c242c188 CS |
161 | char zo_alt_ztest[MAXNAMELEN]; |
162 | char zo_alt_libpath[MAXNAMELEN]; | |
163 | uint64_t zo_vdevs; | |
164 | uint64_t zo_vdevtime; | |
165 | size_t zo_vdev_size; | |
166 | int zo_ashift; | |
167 | int zo_mirrors; | |
168 | int zo_raidz; | |
169 | int zo_raidz_parity; | |
170 | int zo_datasets; | |
171 | int zo_threads; | |
172 | uint64_t zo_passtime; | |
173 | uint64_t zo_killrate; | |
174 | int zo_verbose; | |
175 | int zo_init; | |
176 | uint64_t zo_time; | |
177 | uint64_t zo_maxloops; | |
d830d479 | 178 | uint64_t zo_metaslab_force_ganging; |
379ca9cf | 179 | int zo_mmp_test; |
cc99f275 | 180 | int zo_special_vdevs; |
ab4c009e | 181 | int zo_dump_dbgmsg; |
c242c188 CS |
182 | } ztest_shared_opts_t; |
183 | ||
184 | static const ztest_shared_opts_t ztest_opts_defaults = { | |
185 | .zo_pool = { 'z', 't', 'e', 's', 't', '\0' }, | |
186 | .zo_dir = { '/', 't', 'm', 'p', '\0' }, | |
187 | .zo_alt_ztest = { '\0' }, | |
188 | .zo_alt_libpath = { '\0' }, | |
189 | .zo_vdevs = 5, | |
190 | .zo_ashift = SPA_MINBLOCKSHIFT, | |
191 | .zo_mirrors = 2, | |
192 | .zo_raidz = 4, | |
193 | .zo_raidz_parity = 1, | |
4e21fd06 | 194 | .zo_vdev_size = SPA_MINDEVSIZE * 4, /* 256m default size */ |
c242c188 CS |
195 | .zo_datasets = 7, |
196 | .zo_threads = 23, | |
197 | .zo_passtime = 60, /* 60 seconds */ | |
198 | .zo_killrate = 70, /* 70% kill rate */ | |
199 | .zo_verbose = 0, | |
379ca9cf | 200 | .zo_mmp_test = 0, |
c242c188 CS |
201 | .zo_init = 1, |
202 | .zo_time = 300, /* 5 minutes */ | |
203 | .zo_maxloops = 50, /* max loops during spa_freeze() */ | |
c3bd3fb4 | 204 | .zo_metaslab_force_ganging = 32 << 10, |
cc99f275 | 205 | .zo_special_vdevs = ZTEST_VDEV_CLASS_RND, |
c242c188 CS |
206 | }; |
207 | ||
d830d479 | 208 | extern uint64_t metaslab_force_ganging; |
c242c188 | 209 | extern uint64_t metaslab_df_alloc_threshold; |
8fb1ede1 | 210 | extern unsigned long zfs_deadman_synctime_ms; |
080b3100 | 211 | extern int metaslab_preload_limit; |
d3c2ae1c | 212 | extern boolean_t zfs_compressed_arc_enabled; |
047116ac TC |
213 | extern int zfs_abd_scatter_enabled; |
214 | extern int dmu_object_alloc_chunk_shift; | |
4d044c4c | 215 | extern boolean_t zfs_force_some_double_word_sm_entries; |
c3bd3fb4 | 216 | extern unsigned long zio_decompress_fail_fraction; |
1258bd77 | 217 | extern unsigned long zfs_reconstruct_indirect_damage_fraction; |
c242c188 CS |
218 | |
219 | static ztest_shared_opts_t *ztest_shared_opts; | |
220 | static ztest_shared_opts_t ztest_opts; | |
4807c0ba | 221 | static char *ztest_wkeydata = "abcdefghijklmnopqrstuvwxyz012345"; |
c242c188 CS |
222 | |
223 | typedef struct ztest_shared_ds { | |
224 | uint64_t zd_seq; | |
225 | } ztest_shared_ds_t; | |
226 | ||
227 | static ztest_shared_ds_t *ztest_shared_ds; | |
228 | #define ZTEST_GET_SHARED_DS(d) (&ztest_shared_ds[d]) | |
428870ff BB |
229 | |
230 | #define BT_MAGIC 0x123456789abcdefULL | |
60b82074 BB |
231 | #define MAXFAULTS(zs) \ |
232 | (MAX((zs)->zs_mirrors, 1) * (ztest_opts.zo_raidz_parity + 1) - 1) | |
428870ff BB |
233 | |
234 | enum ztest_io_type { | |
235 | ZTEST_IO_WRITE_TAG, | |
236 | ZTEST_IO_WRITE_PATTERN, | |
237 | ZTEST_IO_WRITE_ZEROES, | |
238 | ZTEST_IO_TRUNCATE, | |
239 | ZTEST_IO_SETATTR, | |
03c6040b | 240 | ZTEST_IO_REWRITE, |
428870ff BB |
241 | ZTEST_IO_TYPES |
242 | }; | |
34dc7c2f BB |
243 | |
244 | typedef struct ztest_block_tag { | |
428870ff | 245 | uint64_t bt_magic; |
34dc7c2f BB |
246 | uint64_t bt_objset; |
247 | uint64_t bt_object; | |
50c957f7 | 248 | uint64_t bt_dnodesize; |
34dc7c2f | 249 | uint64_t bt_offset; |
428870ff | 250 | uint64_t bt_gen; |
34dc7c2f | 251 | uint64_t bt_txg; |
428870ff | 252 | uint64_t bt_crtxg; |
34dc7c2f BB |
253 | } ztest_block_tag_t; |
254 | ||
428870ff BB |
255 | typedef struct bufwad { |
256 | uint64_t bw_index; | |
257 | uint64_t bw_txg; | |
258 | uint64_t bw_data; | |
259 | } bufwad_t; | |
260 | ||
5d43cc9a MA |
261 | /* |
262 | * It would be better to use a rangelock_t per object. Unfortunately | |
263 | * the rangelock_t is not a drop-in replacement for rl_t, because we | |
264 | * still need to map from object ID to rangelock_t. | |
265 | */ | |
266 | typedef enum { | |
267 | RL_READER, | |
268 | RL_WRITER, | |
269 | RL_APPEND | |
270 | } rl_type_t; | |
271 | ||
428870ff BB |
272 | typedef struct rll { |
273 | void *rll_writer; | |
274 | int rll_readers; | |
1e33ac1e BB |
275 | kmutex_t rll_lock; |
276 | kcondvar_t rll_cv; | |
428870ff BB |
277 | } rll_t; |
278 | ||
5d43cc9a MA |
279 | typedef struct rl { |
280 | uint64_t rl_object; | |
281 | uint64_t rl_offset; | |
282 | uint64_t rl_size; | |
283 | rll_t *rl_lock; | |
284 | } rl_t; | |
428870ff BB |
285 | |
286 | #define ZTEST_RANGE_LOCKS 64 | |
287 | #define ZTEST_OBJECT_LOCKS 64 | |
288 | ||
289 | /* | |
290 | * Object descriptor. Used as a template for object lookup/create/remove. | |
291 | */ | |
292 | typedef struct ztest_od { | |
293 | uint64_t od_dir; | |
294 | uint64_t od_object; | |
295 | dmu_object_type_t od_type; | |
296 | dmu_object_type_t od_crtype; | |
297 | uint64_t od_blocksize; | |
298 | uint64_t od_crblocksize; | |
50c957f7 | 299 | uint64_t od_crdnodesize; |
428870ff BB |
300 | uint64_t od_gen; |
301 | uint64_t od_crgen; | |
eca7b760 | 302 | char od_name[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff | 303 | } ztest_od_t; |
34dc7c2f | 304 | |
428870ff BB |
305 | /* |
306 | * Per-dataset state. | |
307 | */ | |
308 | typedef struct ztest_ds { | |
c242c188 | 309 | ztest_shared_ds_t *zd_shared; |
428870ff | 310 | objset_t *zd_os; |
2d9142c9 | 311 | pthread_rwlock_t zd_zilog_lock; |
428870ff | 312 | zilog_t *zd_zilog; |
428870ff | 313 | ztest_od_t *zd_od; /* debugging aid */ |
eca7b760 | 314 | char zd_name[ZFS_MAX_DATASET_NAME_LEN]; |
1e33ac1e | 315 | kmutex_t zd_dirobj_lock; |
428870ff | 316 | rll_t zd_object_lock[ZTEST_OBJECT_LOCKS]; |
5d43cc9a | 317 | rll_t zd_range_lock[ZTEST_RANGE_LOCKS]; |
428870ff BB |
318 | } ztest_ds_t; |
319 | ||
320 | /* | |
321 | * Per-iteration state. | |
322 | */ | |
323 | typedef void ztest_func_t(ztest_ds_t *zd, uint64_t id); | |
324 | ||
325 | typedef struct ztest_info { | |
326 | ztest_func_t *zi_func; /* test function */ | |
327 | uint64_t zi_iters; /* iterations per execution */ | |
328 | uint64_t *zi_interval; /* execute every <interval> seconds */ | |
fdc5d982 | 329 | const char *zi_funcname; /* name of test function */ |
428870ff | 330 | } ztest_info_t; |
34dc7c2f | 331 | |
c242c188 CS |
332 | typedef struct ztest_shared_callstate { |
333 | uint64_t zc_count; /* per-pass count */ | |
334 | uint64_t zc_time; /* per-pass time */ | |
335 | uint64_t zc_next; /* next time to call this function */ | |
336 | } ztest_shared_callstate_t; | |
337 | ||
338 | static ztest_shared_callstate_t *ztest_shared_callstate; | |
339 | #define ZTEST_GET_SHARED_CALLSTATE(c) (&ztest_shared_callstate[c]) | |
340 | ||
34dc7c2f BB |
341 | ztest_func_t ztest_dmu_read_write; |
342 | ztest_func_t ztest_dmu_write_parallel; | |
343 | ztest_func_t ztest_dmu_object_alloc_free; | |
047116ac | 344 | ztest_func_t ztest_dmu_object_next_chunk; |
428870ff | 345 | ztest_func_t ztest_dmu_commit_callbacks; |
34dc7c2f BB |
346 | ztest_func_t ztest_zap; |
347 | ztest_func_t ztest_zap_parallel; | |
428870ff | 348 | ztest_func_t ztest_zil_commit; |
3e31d2b0 | 349 | ztest_func_t ztest_zil_remount; |
428870ff | 350 | ztest_func_t ztest_dmu_read_write_zcopy; |
34dc7c2f | 351 | ztest_func_t ztest_dmu_objset_create_destroy; |
428870ff BB |
352 | ztest_func_t ztest_dmu_prealloc; |
353 | ztest_func_t ztest_fzap; | |
34dc7c2f | 354 | ztest_func_t ztest_dmu_snapshot_create_destroy; |
428870ff BB |
355 | ztest_func_t ztest_dsl_prop_get_set; |
356 | ztest_func_t ztest_spa_prop_get_set; | |
34dc7c2f BB |
357 | ztest_func_t ztest_spa_create_destroy; |
358 | ztest_func_t ztest_fault_inject; | |
428870ff BB |
359 | ztest_func_t ztest_ddt_repair; |
360 | ztest_func_t ztest_dmu_snapshot_hold; | |
0582e403 | 361 | ztest_func_t ztest_mmp_enable_disable; |
428870ff BB |
362 | ztest_func_t ztest_scrub; |
363 | ztest_func_t ztest_dsl_dataset_promote_busy; | |
34dc7c2f BB |
364 | ztest_func_t ztest_vdev_attach_detach; |
365 | ztest_func_t ztest_vdev_LUN_growth; | |
366 | ztest_func_t ztest_vdev_add_remove; | |
cc99f275 | 367 | ztest_func_t ztest_vdev_class_add; |
b128c09f | 368 | ztest_func_t ztest_vdev_aux_add_remove; |
428870ff | 369 | ztest_func_t ztest_split_pool; |
3541dc6d | 370 | ztest_func_t ztest_reguid; |
ea0b2538 | 371 | ztest_func_t ztest_spa_upgrade; |
a1d477c2 MA |
372 | ztest_func_t ztest_device_removal; |
373 | ztest_func_t ztest_remap_blocks; | |
d2734cce | 374 | ztest_func_t ztest_spa_checkpoint_create_discard; |
1eeb4562 | 375 | ztest_func_t ztest_fletcher; |
37f520db | 376 | ztest_func_t ztest_fletcher_incr; |
50c957f7 | 377 | ztest_func_t ztest_verify_dnode_bt; |
34dc7c2f | 378 | |
428870ff BB |
379 | uint64_t zopt_always = 0ULL * NANOSEC; /* all the time */ |
380 | uint64_t zopt_incessant = 1ULL * NANOSEC / 10; /* every 1/10 second */ | |
381 | uint64_t zopt_often = 1ULL * NANOSEC; /* every second */ | |
382 | uint64_t zopt_sometimes = 10ULL * NANOSEC; /* every 10 seconds */ | |
383 | uint64_t zopt_rarely = 60ULL * NANOSEC; /* every 60 seconds */ | |
34dc7c2f | 384 | |
fdc5d982 TC |
385 | #define ZTI_INIT(func, iters, interval) \ |
386 | { .zi_func = (func), \ | |
387 | .zi_iters = (iters), \ | |
388 | .zi_interval = (interval), \ | |
389 | .zi_funcname = # func } | |
390 | ||
34dc7c2f | 391 | ztest_info_t ztest_info[] = { |
fdc5d982 TC |
392 | ZTI_INIT(ztest_dmu_read_write, 1, &zopt_always), |
393 | ZTI_INIT(ztest_dmu_write_parallel, 10, &zopt_always), | |
394 | ZTI_INIT(ztest_dmu_object_alloc_free, 1, &zopt_always), | |
047116ac | 395 | ZTI_INIT(ztest_dmu_object_next_chunk, 1, &zopt_sometimes), |
fdc5d982 TC |
396 | ZTI_INIT(ztest_dmu_commit_callbacks, 1, &zopt_always), |
397 | ZTI_INIT(ztest_zap, 30, &zopt_always), | |
398 | ZTI_INIT(ztest_zap_parallel, 100, &zopt_always), | |
399 | ZTI_INIT(ztest_split_pool, 1, &zopt_always), | |
400 | ZTI_INIT(ztest_zil_commit, 1, &zopt_incessant), | |
401 | ZTI_INIT(ztest_zil_remount, 1, &zopt_sometimes), | |
402 | ZTI_INIT(ztest_dmu_read_write_zcopy, 1, &zopt_often), | |
403 | ZTI_INIT(ztest_dmu_objset_create_destroy, 1, &zopt_often), | |
404 | ZTI_INIT(ztest_dsl_prop_get_set, 1, &zopt_often), | |
405 | ZTI_INIT(ztest_spa_prop_get_set, 1, &zopt_sometimes), | |
428870ff | 406 | #if 0 |
fdc5d982 | 407 | ZTI_INIT(ztest_dmu_prealloc, 1, &zopt_sometimes), |
428870ff | 408 | #endif |
fdc5d982 TC |
409 | ZTI_INIT(ztest_fzap, 1, &zopt_sometimes), |
410 | ZTI_INIT(ztest_dmu_snapshot_create_destroy, 1, &zopt_sometimes), | |
411 | ZTI_INIT(ztest_spa_create_destroy, 1, &zopt_sometimes), | |
412 | ZTI_INIT(ztest_fault_inject, 1, &zopt_sometimes), | |
413 | ZTI_INIT(ztest_ddt_repair, 1, &zopt_sometimes), | |
414 | ZTI_INIT(ztest_dmu_snapshot_hold, 1, &zopt_sometimes), | |
0582e403 | 415 | ZTI_INIT(ztest_mmp_enable_disable, 1, &zopt_sometimes), |
fdc5d982 | 416 | ZTI_INIT(ztest_reguid, 1, &zopt_rarely), |
fdc5d982 TC |
417 | ZTI_INIT(ztest_scrub, 1, &zopt_rarely), |
418 | ZTI_INIT(ztest_spa_upgrade, 1, &zopt_rarely), | |
419 | ZTI_INIT(ztest_dsl_dataset_promote_busy, 1, &zopt_rarely), | |
420 | ZTI_INIT(ztest_vdev_attach_detach, 1, &zopt_sometimes), | |
421 | ZTI_INIT(ztest_vdev_LUN_growth, 1, &zopt_rarely), | |
422 | ZTI_INIT(ztest_vdev_add_remove, 1, &ztest_opts.zo_vdevtime), | |
cc99f275 | 423 | ZTI_INIT(ztest_vdev_class_add, 1, &ztest_opts.zo_vdevtime), |
fdc5d982 | 424 | ZTI_INIT(ztest_vdev_aux_add_remove, 1, &ztest_opts.zo_vdevtime), |
a1d477c2 MA |
425 | ZTI_INIT(ztest_device_removal, 1, &zopt_sometimes), |
426 | ZTI_INIT(ztest_remap_blocks, 1, &zopt_sometimes), | |
d2734cce | 427 | ZTI_INIT(ztest_spa_checkpoint_create_discard, 1, &zopt_rarely), |
1eeb4562 | 428 | ZTI_INIT(ztest_fletcher, 1, &zopt_rarely), |
37f520db | 429 | ZTI_INIT(ztest_fletcher_incr, 1, &zopt_rarely), |
50c957f7 | 430 | ZTI_INIT(ztest_verify_dnode_bt, 1, &zopt_sometimes), |
34dc7c2f BB |
431 | }; |
432 | ||
433 | #define ZTEST_FUNCS (sizeof (ztest_info) / sizeof (ztest_info_t)) | |
434 | ||
428870ff BB |
435 | /* |
436 | * The following struct is used to hold a list of uncalled commit callbacks. | |
437 | * The callbacks are ordered by txg number. | |
438 | */ | |
439 | typedef struct ztest_cb_list { | |
1e33ac1e BB |
440 | kmutex_t zcl_callbacks_lock; |
441 | list_t zcl_callbacks; | |
428870ff | 442 | } ztest_cb_list_t; |
34dc7c2f BB |
443 | |
444 | /* | |
445 | * Stuff we need to share writably between parent and child. | |
446 | */ | |
447 | typedef struct ztest_shared { | |
c242c188 | 448 | boolean_t zs_do_init; |
428870ff BB |
449 | hrtime_t zs_proc_start; |
450 | hrtime_t zs_proc_stop; | |
451 | hrtime_t zs_thread_start; | |
452 | hrtime_t zs_thread_stop; | |
453 | hrtime_t zs_thread_kill; | |
34dc7c2f | 454 | uint64_t zs_enospc_count; |
428870ff BB |
455 | uint64_t zs_vdev_next_leaf; |
456 | uint64_t zs_vdev_aux; | |
34dc7c2f BB |
457 | uint64_t zs_alloc; |
458 | uint64_t zs_space; | |
428870ff BB |
459 | uint64_t zs_splits; |
460 | uint64_t zs_mirrors; | |
c242c188 CS |
461 | uint64_t zs_metaslab_sz; |
462 | uint64_t zs_metaslab_df_alloc_threshold; | |
463 | uint64_t zs_guid; | |
34dc7c2f BB |
464 | } ztest_shared_t; |
465 | ||
428870ff BB |
466 | #define ID_PARALLEL -1ULL |
467 | ||
34dc7c2f | 468 | static char ztest_dev_template[] = "%s/%s.%llua"; |
b128c09f | 469 | static char ztest_aux_template[] = "%s/%s.%s.%llu"; |
428870ff | 470 | ztest_shared_t *ztest_shared; |
34dc7c2f | 471 | |
c242c188 CS |
472 | static spa_t *ztest_spa = NULL; |
473 | static ztest_ds_t *ztest_ds; | |
474 | ||
475 | static kmutex_t ztest_vdev_lock; | |
9e052db4 | 476 | static boolean_t ztest_device_removal_active = B_FALSE; |
d2734cce | 477 | static kmutex_t ztest_checkpoint_lock; |
3bc7e0fb GW |
478 | |
479 | /* | |
480 | * The ztest_name_lock protects the pool and dataset namespace used by | |
481 | * the individual tests. To modify the namespace, consumers must grab | |
482 | * this lock as writer. Grabbing the lock as reader will ensure that the | |
483 | * namespace does not change while the lock is held. | |
484 | */ | |
2d9142c9 | 485 | static pthread_rwlock_t ztest_name_lock; |
34dc7c2f | 486 | |
c242c188 | 487 | static boolean_t ztest_dump_core = B_TRUE; |
b128c09f | 488 | static boolean_t ztest_exiting; |
34dc7c2f | 489 | |
428870ff BB |
490 | /* Global commit callback list */ |
491 | static ztest_cb_list_t zcl; | |
090ff092 RC |
492 | /* Commit cb delay */ |
493 | static uint64_t zc_min_txg_delay = UINT64_MAX; | |
494 | static int zc_cb_counter = 0; | |
495 | ||
496 | /* | |
497 | * Minimum number of commit callbacks that need to be registered for us to check | |
498 | * whether the minimum txg delay is acceptable. | |
499 | */ | |
500 | #define ZTEST_COMMIT_CB_MIN_REG 100 | |
501 | ||
502 | /* | |
503 | * If a number of txgs equal to this threshold have been created after a commit | |
504 | * callback has been registered but not called, then we assume there is an | |
505 | * implementation bug. | |
506 | */ | |
507 | #define ZTEST_COMMIT_CB_THRESH (TXG_CONCURRENT_STATES + 1000) | |
428870ff | 508 | |
428870ff BB |
509 | enum ztest_object { |
510 | ZTEST_META_DNODE = 0, | |
511 | ZTEST_DIROBJ, | |
512 | ZTEST_OBJECTS | |
513 | }; | |
34dc7c2f BB |
514 | |
515 | static void usage(boolean_t) __NORETURN; | |
516 | ||
517 | /* | |
518 | * These libumem hooks provide a reasonable set of defaults for the allocator's | |
519 | * debugging facilities. | |
520 | */ | |
521 | const char * | |
0bc8fd78 | 522 | _umem_debug_init(void) |
34dc7c2f BB |
523 | { |
524 | return ("default,verbose"); /* $UMEM_DEBUG setting */ | |
525 | } | |
526 | ||
527 | const char * | |
528 | _umem_logging_init(void) | |
529 | { | |
530 | return ("fail,contents"); /* $UMEM_LOGGING setting */ | |
531 | } | |
532 | ||
8fb1ede1 BB |
533 | static void |
534 | dump_debug_buffer(void) | |
535 | { | |
ab4c009e TC |
536 | ssize_t ret __attribute__((unused)); |
537 | ||
538 | if (!ztest_opts.zo_dump_dbgmsg) | |
8fb1ede1 BB |
539 | return; |
540 | ||
ab4c009e TC |
541 | /* |
542 | * We use write() instead of printf() so that this function | |
543 | * is safe to call from a signal handler. | |
544 | */ | |
545 | ret = write(STDOUT_FILENO, "\n", 1); | |
8fb1ede1 BB |
546 | zfs_dbgmsg_print("ztest"); |
547 | } | |
548 | ||
e82cdc3a NB |
549 | #define BACKTRACE_SZ 100 |
550 | ||
551 | static void sig_handler(int signo) | |
552 | { | |
553 | struct sigaction action; | |
967798d0 | 554 | #ifdef __GLIBC__ /* backtrace() is a GNU extension */ |
e82cdc3a NB |
555 | int nptrs; |
556 | void *buffer[BACKTRACE_SZ]; | |
557 | ||
558 | nptrs = backtrace(buffer, BACKTRACE_SZ); | |
559 | backtrace_symbols_fd(buffer, nptrs, STDERR_FILENO); | |
560 | #endif | |
8fb1ede1 | 561 | dump_debug_buffer(); |
e82cdc3a NB |
562 | |
563 | /* | |
564 | * Restore default action and re-raise signal so SIGSEGV and | |
565 | * SIGABRT can trigger a core dump. | |
566 | */ | |
567 | action.sa_handler = SIG_DFL; | |
568 | sigemptyset(&action.sa_mask); | |
569 | action.sa_flags = 0; | |
570 | (void) sigaction(signo, &action, NULL); | |
571 | raise(signo); | |
572 | } | |
573 | ||
34dc7c2f BB |
574 | #define FATAL_MSG_SZ 1024 |
575 | ||
576 | char *fatal_msg; | |
577 | ||
578 | static void | |
579 | fatal(int do_perror, char *message, ...) | |
580 | { | |
581 | va_list args; | |
582 | int save_errno = errno; | |
40b84e7a | 583 | char *buf; |
34dc7c2f BB |
584 | |
585 | (void) fflush(stdout); | |
40b84e7a | 586 | buf = umem_alloc(FATAL_MSG_SZ, UMEM_NOFAIL); |
34dc7c2f BB |
587 | |
588 | va_start(args, message); | |
589 | (void) sprintf(buf, "ztest: "); | |
590 | /* LINTED */ | |
591 | (void) vsprintf(buf + strlen(buf), message, args); | |
592 | va_end(args); | |
593 | if (do_perror) { | |
594 | (void) snprintf(buf + strlen(buf), FATAL_MSG_SZ - strlen(buf), | |
595 | ": %s", strerror(save_errno)); | |
596 | } | |
597 | (void) fprintf(stderr, "%s\n", buf); | |
598 | fatal_msg = buf; /* to ease debugging */ | |
8fb1ede1 | 599 | |
34dc7c2f BB |
600 | if (ztest_dump_core) |
601 | abort(); | |
ab4c009e TC |
602 | else |
603 | dump_debug_buffer(); | |
604 | ||
34dc7c2f BB |
605 | exit(3); |
606 | } | |
607 | ||
608 | static int | |
609 | str2shift(const char *buf) | |
610 | { | |
611 | const char *ends = "BKMGTPEZ"; | |
612 | int i; | |
613 | ||
614 | if (buf[0] == '\0') | |
615 | return (0); | |
616 | for (i = 0; i < strlen(ends); i++) { | |
617 | if (toupper(buf[0]) == ends[i]) | |
618 | break; | |
619 | } | |
620 | if (i == strlen(ends)) { | |
621 | (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", | |
622 | buf); | |
623 | usage(B_FALSE); | |
624 | } | |
625 | if (buf[1] == '\0' || (toupper(buf[1]) == 'B' && buf[2] == '\0')) { | |
626 | return (10*i); | |
627 | } | |
628 | (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", buf); | |
629 | usage(B_FALSE); | |
630 | /* NOTREACHED */ | |
631 | } | |
632 | ||
633 | static uint64_t | |
634 | nicenumtoull(const char *buf) | |
635 | { | |
636 | char *end; | |
637 | uint64_t val; | |
638 | ||
639 | val = strtoull(buf, &end, 0); | |
640 | if (end == buf) { | |
641 | (void) fprintf(stderr, "ztest: bad numeric value: %s\n", buf); | |
642 | usage(B_FALSE); | |
643 | } else if (end[0] == '.') { | |
644 | double fval = strtod(buf, &end); | |
645 | fval *= pow(2, str2shift(end)); | |
646 | if (fval > UINT64_MAX) { | |
647 | (void) fprintf(stderr, "ztest: value too large: %s\n", | |
648 | buf); | |
649 | usage(B_FALSE); | |
650 | } | |
651 | val = (uint64_t)fval; | |
652 | } else { | |
653 | int shift = str2shift(end); | |
654 | if (shift >= 64 || (val << shift) >> shift != val) { | |
655 | (void) fprintf(stderr, "ztest: value too large: %s\n", | |
656 | buf); | |
657 | usage(B_FALSE); | |
658 | } | |
659 | val <<= shift; | |
660 | } | |
661 | return (val); | |
662 | } | |
663 | ||
664 | static void | |
665 | usage(boolean_t requested) | |
666 | { | |
c242c188 CS |
667 | const ztest_shared_opts_t *zo = &ztest_opts_defaults; |
668 | ||
f3c8c9e6 | 669 | char nice_vdev_size[NN_NUMBUF_SZ]; |
d830d479 | 670 | char nice_force_ganging[NN_NUMBUF_SZ]; |
34dc7c2f BB |
671 | FILE *fp = requested ? stdout : stderr; |
672 | ||
f3c8c9e6 | 673 | nicenum(zo->zo_vdev_size, nice_vdev_size, sizeof (nice_vdev_size)); |
d830d479 MA |
674 | nicenum(zo->zo_metaslab_force_ganging, nice_force_ganging, |
675 | sizeof (nice_force_ganging)); | |
34dc7c2f BB |
676 | |
677 | (void) fprintf(fp, "Usage: %s\n" | |
678 | "\t[-v vdevs (default: %llu)]\n" | |
679 | "\t[-s size_of_each_vdev (default: %s)]\n" | |
428870ff | 680 | "\t[-a alignment_shift (default: %d)] use 0 for random\n" |
34dc7c2f BB |
681 | "\t[-m mirror_copies (default: %d)]\n" |
682 | "\t[-r raidz_disks (default: %d)]\n" | |
683 | "\t[-R raidz_parity (default: %d)]\n" | |
684 | "\t[-d datasets (default: %d)]\n" | |
685 | "\t[-t threads (default: %d)]\n" | |
686 | "\t[-g gang_block_threshold (default: %s)]\n" | |
428870ff BB |
687 | "\t[-i init_count (default: %d)] initialize pool i times\n" |
688 | "\t[-k kill_percentage (default: %llu%%)]\n" | |
34dc7c2f | 689 | "\t[-p pool_name (default: %s)]\n" |
428870ff | 690 | "\t[-f dir (default: %s)] file directory for vdev files\n" |
379ca9cf | 691 | "\t[-M] Multi-host simulate pool imported on remote host\n" |
428870ff BB |
692 | "\t[-V] verbose (use multiple times for ever more blather)\n" |
693 | "\t[-E] use existing pool instead of creating new one\n" | |
694 | "\t[-T time (default: %llu sec)] total run time\n" | |
695 | "\t[-F freezeloops (default: %llu)] max loops in spa_freeze()\n" | |
696 | "\t[-P passtime (default: %llu sec)] time per pass\n" | |
c242c188 | 697 | "\t[-B alt_ztest (default: <none>)] alternate ztest path\n" |
cc99f275 | 698 | "\t[-C vdev class state (default: random)] special=on|off|random\n" |
ed828c0c GM |
699 | "\t[-o variable=value] ... set global variable to an unsigned\n" |
700 | "\t 32-bit integer value\n" | |
8fb1ede1 | 701 | "\t[-G dump zfs_dbgmsg buffer before exiting due to an error\n" |
34dc7c2f BB |
702 | "\t[-h] (print help)\n" |
703 | "", | |
c242c188 CS |
704 | zo->zo_pool, |
705 | (u_longlong_t)zo->zo_vdevs, /* -v */ | |
34dc7c2f | 706 | nice_vdev_size, /* -s */ |
c242c188 CS |
707 | zo->zo_ashift, /* -a */ |
708 | zo->zo_mirrors, /* -m */ | |
709 | zo->zo_raidz, /* -r */ | |
710 | zo->zo_raidz_parity, /* -R */ | |
711 | zo->zo_datasets, /* -d */ | |
712 | zo->zo_threads, /* -t */ | |
d830d479 | 713 | nice_force_ganging, /* -g */ |
c242c188 CS |
714 | zo->zo_init, /* -i */ |
715 | (u_longlong_t)zo->zo_killrate, /* -k */ | |
716 | zo->zo_pool, /* -p */ | |
717 | zo->zo_dir, /* -f */ | |
718 | (u_longlong_t)zo->zo_time, /* -T */ | |
719 | (u_longlong_t)zo->zo_maxloops, /* -F */ | |
720 | (u_longlong_t)zo->zo_passtime); | |
34dc7c2f BB |
721 | exit(requested ? 0 : 1); |
722 | } | |
723 | ||
cc99f275 DB |
724 | |
725 | static void | |
726 | ztest_parse_name_value(const char *input, ztest_shared_opts_t *zo) | |
727 | { | |
728 | char name[32]; | |
729 | char *value; | |
730 | int state = ZTEST_VDEV_CLASS_RND; | |
731 | ||
732 | (void) strlcpy(name, input, sizeof (name)); | |
733 | ||
734 | value = strchr(name, '='); | |
735 | if (value == NULL) { | |
736 | (void) fprintf(stderr, "missing value in property=value " | |
737 | "'-C' argument (%s)\n", input); | |
738 | usage(B_FALSE); | |
739 | } | |
740 | *(value) = '\0'; | |
741 | value++; | |
742 | ||
743 | if (strcmp(value, "on") == 0) { | |
744 | state = ZTEST_VDEV_CLASS_ON; | |
745 | } else if (strcmp(value, "off") == 0) { | |
746 | state = ZTEST_VDEV_CLASS_OFF; | |
747 | } else if (strcmp(value, "random") == 0) { | |
748 | state = ZTEST_VDEV_CLASS_RND; | |
749 | } else { | |
750 | (void) fprintf(stderr, "invalid property value '%s'\n", value); | |
751 | usage(B_FALSE); | |
752 | } | |
753 | ||
754 | if (strcmp(name, "special") == 0) { | |
755 | zo->zo_special_vdevs = state; | |
756 | } else { | |
757 | (void) fprintf(stderr, "invalid property name '%s'\n", name); | |
758 | usage(B_FALSE); | |
759 | } | |
760 | if (zo->zo_verbose >= 3) | |
761 | (void) printf("%s vdev state is '%s'\n", name, value); | |
762 | } | |
763 | ||
34dc7c2f BB |
764 | static void |
765 | process_options(int argc, char **argv) | |
766 | { | |
c242c188 CS |
767 | char *path; |
768 | ztest_shared_opts_t *zo = &ztest_opts; | |
769 | ||
34dc7c2f BB |
770 | int opt; |
771 | uint64_t value; | |
c242c188 | 772 | char altdir[MAXNAMELEN] = { 0 }; |
34dc7c2f | 773 | |
c242c188 | 774 | bcopy(&ztest_opts_defaults, zo, sizeof (*zo)); |
34dc7c2f | 775 | |
34dc7c2f | 776 | while ((opt = getopt(argc, argv, |
cc99f275 | 777 | "v:s:a:m:r:R:d:t:g:i:k:p:f:MVET:P:hF:B:C:o:G")) != EOF) { |
34dc7c2f BB |
778 | value = 0; |
779 | switch (opt) { | |
780 | case 'v': | |
781 | case 's': | |
782 | case 'a': | |
783 | case 'm': | |
784 | case 'r': | |
785 | case 'R': | |
786 | case 'd': | |
787 | case 't': | |
788 | case 'g': | |
789 | case 'i': | |
790 | case 'k': | |
791 | case 'T': | |
792 | case 'P': | |
428870ff | 793 | case 'F': |
34dc7c2f BB |
794 | value = nicenumtoull(optarg); |
795 | } | |
796 | switch (opt) { | |
797 | case 'v': | |
c242c188 | 798 | zo->zo_vdevs = value; |
34dc7c2f BB |
799 | break; |
800 | case 's': | |
c242c188 | 801 | zo->zo_vdev_size = MAX(SPA_MINDEVSIZE, value); |
34dc7c2f BB |
802 | break; |
803 | case 'a': | |
c242c188 | 804 | zo->zo_ashift = value; |
34dc7c2f BB |
805 | break; |
806 | case 'm': | |
c242c188 | 807 | zo->zo_mirrors = value; |
34dc7c2f BB |
808 | break; |
809 | case 'r': | |
c242c188 | 810 | zo->zo_raidz = MAX(1, value); |
34dc7c2f BB |
811 | break; |
812 | case 'R': | |
c242c188 | 813 | zo->zo_raidz_parity = MIN(MAX(value, 1), 3); |
34dc7c2f BB |
814 | break; |
815 | case 'd': | |
c242c188 | 816 | zo->zo_datasets = MAX(1, value); |
34dc7c2f BB |
817 | break; |
818 | case 't': | |
c242c188 | 819 | zo->zo_threads = MAX(1, value); |
34dc7c2f BB |
820 | break; |
821 | case 'g': | |
d830d479 MA |
822 | zo->zo_metaslab_force_ganging = |
823 | MAX(SPA_MINBLOCKSIZE << 1, value); | |
34dc7c2f BB |
824 | break; |
825 | case 'i': | |
c242c188 | 826 | zo->zo_init = value; |
34dc7c2f BB |
827 | break; |
828 | case 'k': | |
c242c188 | 829 | zo->zo_killrate = value; |
34dc7c2f BB |
830 | break; |
831 | case 'p': | |
c242c188 CS |
832 | (void) strlcpy(zo->zo_pool, optarg, |
833 | sizeof (zo->zo_pool)); | |
34dc7c2f BB |
834 | break; |
835 | case 'f': | |
c242c188 CS |
836 | path = realpath(optarg, NULL); |
837 | if (path == NULL) { | |
838 | (void) fprintf(stderr, "error: %s: %s\n", | |
839 | optarg, strerror(errno)); | |
840 | usage(B_FALSE); | |
841 | } else { | |
842 | (void) strlcpy(zo->zo_dir, path, | |
843 | sizeof (zo->zo_dir)); | |
df053d67 | 844 | free(path); |
c242c188 | 845 | } |
34dc7c2f | 846 | break; |
379ca9cf OF |
847 | case 'M': |
848 | zo->zo_mmp_test = 1; | |
849 | break; | |
34dc7c2f | 850 | case 'V': |
c242c188 | 851 | zo->zo_verbose++; |
34dc7c2f BB |
852 | break; |
853 | case 'E': | |
c242c188 | 854 | zo->zo_init = 0; |
34dc7c2f BB |
855 | break; |
856 | case 'T': | |
c242c188 | 857 | zo->zo_time = value; |
34dc7c2f BB |
858 | break; |
859 | case 'P': | |
c242c188 | 860 | zo->zo_passtime = MAX(1, value); |
34dc7c2f | 861 | break; |
428870ff | 862 | case 'F': |
c242c188 CS |
863 | zo->zo_maxloops = MAX(1, value); |
864 | break; | |
865 | case 'B': | |
866 | (void) strlcpy(altdir, optarg, sizeof (altdir)); | |
428870ff | 867 | break; |
cc99f275 DB |
868 | case 'C': |
869 | ztest_parse_name_value(optarg, zo); | |
870 | break; | |
ed828c0c GM |
871 | case 'o': |
872 | if (set_global_var(optarg) != 0) | |
873 | usage(B_FALSE); | |
874 | break; | |
8fb1ede1 | 875 | case 'G': |
ab4c009e | 876 | zo->zo_dump_dbgmsg = 1; |
8fb1ede1 | 877 | break; |
34dc7c2f BB |
878 | case 'h': |
879 | usage(B_TRUE); | |
880 | break; | |
881 | case '?': | |
882 | default: | |
883 | usage(B_FALSE); | |
884 | break; | |
885 | } | |
886 | } | |
887 | ||
c242c188 | 888 | zo->zo_raidz_parity = MIN(zo->zo_raidz_parity, zo->zo_raidz - 1); |
34dc7c2f | 889 | |
c242c188 CS |
890 | zo->zo_vdevtime = |
891 | (zo->zo_vdevs > 0 ? zo->zo_time * NANOSEC / zo->zo_vdevs : | |
428870ff | 892 | UINT64_MAX >> 2); |
c242c188 CS |
893 | |
894 | if (strlen(altdir) > 0) { | |
ae380cfa BB |
895 | char *cmd; |
896 | char *realaltdir; | |
c242c188 CS |
897 | char *bin; |
898 | char *ztest; | |
899 | char *isa; | |
900 | int isalen; | |
901 | ||
ae380cfa BB |
902 | cmd = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
903 | realaltdir = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
904 | ||
5be98cfe | 905 | VERIFY(NULL != realpath(getexecname(), cmd)); |
c242c188 CS |
906 | if (0 != access(altdir, F_OK)) { |
907 | ztest_dump_core = B_FALSE; | |
908 | fatal(B_TRUE, "invalid alternate ztest path: %s", | |
909 | altdir); | |
910 | } | |
911 | VERIFY(NULL != realpath(altdir, realaltdir)); | |
912 | ||
913 | /* | |
914 | * 'cmd' should be of the form "<anything>/usr/bin/<isa>/ztest". | |
915 | * We want to extract <isa> to determine if we should use | |
916 | * 32 or 64 bit binaries. | |
917 | */ | |
918 | bin = strstr(cmd, "/usr/bin/"); | |
919 | ztest = strstr(bin, "/ztest"); | |
920 | isa = bin + 9; | |
921 | isalen = ztest - isa; | |
922 | (void) snprintf(zo->zo_alt_ztest, sizeof (zo->zo_alt_ztest), | |
923 | "%s/usr/bin/%.*s/ztest", realaltdir, isalen, isa); | |
924 | (void) snprintf(zo->zo_alt_libpath, sizeof (zo->zo_alt_libpath), | |
925 | "%s/usr/lib/%.*s", realaltdir, isalen, isa); | |
926 | ||
927 | if (0 != access(zo->zo_alt_ztest, X_OK)) { | |
928 | ztest_dump_core = B_FALSE; | |
929 | fatal(B_TRUE, "invalid alternate ztest: %s", | |
930 | zo->zo_alt_ztest); | |
931 | } else if (0 != access(zo->zo_alt_libpath, X_OK)) { | |
932 | ztest_dump_core = B_FALSE; | |
933 | fatal(B_TRUE, "invalid alternate lib directory %s", | |
934 | zo->zo_alt_libpath); | |
935 | } | |
ae380cfa BB |
936 | |
937 | umem_free(cmd, MAXPATHLEN); | |
938 | umem_free(realaltdir, MAXPATHLEN); | |
c242c188 | 939 | } |
428870ff BB |
940 | } |
941 | ||
942 | static void | |
943 | ztest_kill(ztest_shared_t *zs) | |
944 | { | |
c242c188 CS |
945 | zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(ztest_spa)); |
946 | zs->zs_space = metaslab_class_get_space(spa_normal_class(ztest_spa)); | |
5d1f7fb6 GW |
947 | |
948 | /* | |
949 | * Before we kill off ztest, make sure that the config is updated. | |
a1d477c2 | 950 | * See comment above spa_write_cachefile(). |
5d1f7fb6 GW |
951 | */ |
952 | mutex_enter(&spa_namespace_lock); | |
a1d477c2 | 953 | spa_write_cachefile(ztest_spa, B_FALSE, B_FALSE); |
5d1f7fb6 GW |
954 | mutex_exit(&spa_namespace_lock); |
955 | ||
428870ff BB |
956 | (void) kill(getpid(), SIGKILL); |
957 | } | |
958 | ||
959 | static uint64_t | |
960 | ztest_random(uint64_t range) | |
961 | { | |
962 | uint64_t r; | |
963 | ||
9d81146b ED |
964 | ASSERT3S(ztest_fd_rand, >=, 0); |
965 | ||
428870ff BB |
966 | if (range == 0) |
967 | return (0); | |
968 | ||
9d81146b | 969 | if (read(ztest_fd_rand, &r, sizeof (r)) != sizeof (r)) |
428870ff BB |
970 | fatal(1, "short read from /dev/urandom"); |
971 | ||
972 | return (r % range); | |
973 | } | |
974 | ||
975 | /* ARGSUSED */ | |
976 | static void | |
977 | ztest_record_enospc(const char *s) | |
978 | { | |
979 | ztest_shared->zs_enospc_count++; | |
34dc7c2f BB |
980 | } |
981 | ||
982 | static uint64_t | |
983 | ztest_get_ashift(void) | |
984 | { | |
c242c188 | 985 | if (ztest_opts.zo_ashift == 0) |
b02fe35d | 986 | return (SPA_MINBLOCKSHIFT + ztest_random(5)); |
c242c188 | 987 | return (ztest_opts.zo_ashift); |
34dc7c2f BB |
988 | } |
989 | ||
990 | static nvlist_t * | |
ea0b2538 | 991 | make_vdev_file(char *path, char *aux, char *pool, size_t size, uint64_t ashift) |
34dc7c2f | 992 | { |
40b84e7a | 993 | char *pathbuf; |
34dc7c2f | 994 | uint64_t vdev; |
34dc7c2f BB |
995 | nvlist_t *file; |
996 | ||
40b84e7a BB |
997 | pathbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
998 | ||
b128c09f BB |
999 | if (ashift == 0) |
1000 | ashift = ztest_get_ashift(); | |
1001 | ||
1002 | if (path == NULL) { | |
1003 | path = pathbuf; | |
1004 | ||
1005 | if (aux != NULL) { | |
1006 | vdev = ztest_shared->zs_vdev_aux; | |
c242c188 CS |
1007 | (void) snprintf(path, MAXPATHLEN, |
1008 | ztest_aux_template, ztest_opts.zo_dir, | |
ea0b2538 GW |
1009 | pool == NULL ? ztest_opts.zo_pool : pool, |
1010 | aux, vdev); | |
b128c09f | 1011 | } else { |
428870ff | 1012 | vdev = ztest_shared->zs_vdev_next_leaf++; |
c242c188 CS |
1013 | (void) snprintf(path, MAXPATHLEN, |
1014 | ztest_dev_template, ztest_opts.zo_dir, | |
ea0b2538 | 1015 | pool == NULL ? ztest_opts.zo_pool : pool, vdev); |
b128c09f BB |
1016 | } |
1017 | } | |
34dc7c2f | 1018 | |
b128c09f BB |
1019 | if (size != 0) { |
1020 | int fd = open(path, O_RDWR | O_CREAT | O_TRUNC, 0666); | |
34dc7c2f | 1021 | if (fd == -1) |
b128c09f | 1022 | fatal(1, "can't open %s", path); |
34dc7c2f | 1023 | if (ftruncate(fd, size) != 0) |
b128c09f | 1024 | fatal(1, "can't ftruncate %s", path); |
34dc7c2f BB |
1025 | (void) close(fd); |
1026 | } | |
1027 | ||
1028 | VERIFY(nvlist_alloc(&file, NV_UNIQUE_NAME, 0) == 0); | |
1029 | VERIFY(nvlist_add_string(file, ZPOOL_CONFIG_TYPE, VDEV_TYPE_FILE) == 0); | |
b128c09f | 1030 | VERIFY(nvlist_add_string(file, ZPOOL_CONFIG_PATH, path) == 0); |
34dc7c2f | 1031 | VERIFY(nvlist_add_uint64(file, ZPOOL_CONFIG_ASHIFT, ashift) == 0); |
40b84e7a | 1032 | umem_free(pathbuf, MAXPATHLEN); |
34dc7c2f BB |
1033 | |
1034 | return (file); | |
1035 | } | |
1036 | ||
1037 | static nvlist_t * | |
ea0b2538 GW |
1038 | make_vdev_raidz(char *path, char *aux, char *pool, size_t size, |
1039 | uint64_t ashift, int r) | |
34dc7c2f BB |
1040 | { |
1041 | nvlist_t *raidz, **child; | |
1042 | int c; | |
1043 | ||
1044 | if (r < 2) | |
ea0b2538 | 1045 | return (make_vdev_file(path, aux, pool, size, ashift)); |
34dc7c2f BB |
1046 | child = umem_alloc(r * sizeof (nvlist_t *), UMEM_NOFAIL); |
1047 | ||
1048 | for (c = 0; c < r; c++) | |
ea0b2538 | 1049 | child[c] = make_vdev_file(path, aux, pool, size, ashift); |
34dc7c2f BB |
1050 | |
1051 | VERIFY(nvlist_alloc(&raidz, NV_UNIQUE_NAME, 0) == 0); | |
1052 | VERIFY(nvlist_add_string(raidz, ZPOOL_CONFIG_TYPE, | |
1053 | VDEV_TYPE_RAIDZ) == 0); | |
1054 | VERIFY(nvlist_add_uint64(raidz, ZPOOL_CONFIG_NPARITY, | |
c242c188 | 1055 | ztest_opts.zo_raidz_parity) == 0); |
34dc7c2f BB |
1056 | VERIFY(nvlist_add_nvlist_array(raidz, ZPOOL_CONFIG_CHILDREN, |
1057 | child, r) == 0); | |
1058 | ||
1059 | for (c = 0; c < r; c++) | |
1060 | nvlist_free(child[c]); | |
1061 | ||
1062 | umem_free(child, r * sizeof (nvlist_t *)); | |
1063 | ||
1064 | return (raidz); | |
1065 | } | |
1066 | ||
1067 | static nvlist_t * | |
ea0b2538 GW |
1068 | make_vdev_mirror(char *path, char *aux, char *pool, size_t size, |
1069 | uint64_t ashift, int r, int m) | |
34dc7c2f BB |
1070 | { |
1071 | nvlist_t *mirror, **child; | |
1072 | int c; | |
1073 | ||
1074 | if (m < 1) | |
ea0b2538 | 1075 | return (make_vdev_raidz(path, aux, pool, size, ashift, r)); |
34dc7c2f BB |
1076 | |
1077 | child = umem_alloc(m * sizeof (nvlist_t *), UMEM_NOFAIL); | |
1078 | ||
1079 | for (c = 0; c < m; c++) | |
ea0b2538 | 1080 | child[c] = make_vdev_raidz(path, aux, pool, size, ashift, r); |
34dc7c2f BB |
1081 | |
1082 | VERIFY(nvlist_alloc(&mirror, NV_UNIQUE_NAME, 0) == 0); | |
1083 | VERIFY(nvlist_add_string(mirror, ZPOOL_CONFIG_TYPE, | |
1084 | VDEV_TYPE_MIRROR) == 0); | |
1085 | VERIFY(nvlist_add_nvlist_array(mirror, ZPOOL_CONFIG_CHILDREN, | |
1086 | child, m) == 0); | |
34dc7c2f BB |
1087 | |
1088 | for (c = 0; c < m; c++) | |
1089 | nvlist_free(child[c]); | |
1090 | ||
1091 | umem_free(child, m * sizeof (nvlist_t *)); | |
1092 | ||
1093 | return (mirror); | |
1094 | } | |
1095 | ||
1096 | static nvlist_t * | |
ea0b2538 | 1097 | make_vdev_root(char *path, char *aux, char *pool, size_t size, uint64_t ashift, |
cc99f275 | 1098 | const char *class, int r, int m, int t) |
34dc7c2f BB |
1099 | { |
1100 | nvlist_t *root, **child; | |
1101 | int c; | |
cc99f275 | 1102 | boolean_t log; |
34dc7c2f BB |
1103 | |
1104 | ASSERT(t > 0); | |
1105 | ||
cc99f275 DB |
1106 | log = (class != NULL && strcmp(class, "log") == 0); |
1107 | ||
34dc7c2f BB |
1108 | child = umem_alloc(t * sizeof (nvlist_t *), UMEM_NOFAIL); |
1109 | ||
b128c09f | 1110 | for (c = 0; c < t; c++) { |
ea0b2538 GW |
1111 | child[c] = make_vdev_mirror(path, aux, pool, size, ashift, |
1112 | r, m); | |
b128c09f BB |
1113 | VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_IS_LOG, |
1114 | log) == 0); | |
cc99f275 DB |
1115 | |
1116 | if (class != NULL && class[0] != '\0') { | |
1117 | ASSERT(m > 1 || log); /* expecting a mirror */ | |
1118 | VERIFY(nvlist_add_string(child[c], | |
1119 | ZPOOL_CONFIG_ALLOCATION_BIAS, class) == 0); | |
1120 | } | |
b128c09f | 1121 | } |
34dc7c2f BB |
1122 | |
1123 | VERIFY(nvlist_alloc(&root, NV_UNIQUE_NAME, 0) == 0); | |
1124 | VERIFY(nvlist_add_string(root, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT) == 0); | |
b128c09f | 1125 | VERIFY(nvlist_add_nvlist_array(root, aux ? aux : ZPOOL_CONFIG_CHILDREN, |
34dc7c2f BB |
1126 | child, t) == 0); |
1127 | ||
1128 | for (c = 0; c < t; c++) | |
1129 | nvlist_free(child[c]); | |
1130 | ||
1131 | umem_free(child, t * sizeof (nvlist_t *)); | |
1132 | ||
1133 | return (root); | |
1134 | } | |
1135 | ||
ea0b2538 GW |
1136 | /* |
1137 | * Find a random spa version. Returns back a random spa version in the | |
1138 | * range [initial_version, SPA_VERSION_FEATURES]. | |
1139 | */ | |
1140 | static uint64_t | |
1141 | ztest_random_spa_version(uint64_t initial_version) | |
1142 | { | |
1143 | uint64_t version = initial_version; | |
1144 | ||
1145 | if (version <= SPA_VERSION_BEFORE_FEATURES) { | |
1146 | version = version + | |
1147 | ztest_random(SPA_VERSION_BEFORE_FEATURES - version + 1); | |
1148 | } | |
1149 | ||
1150 | if (version > SPA_VERSION_BEFORE_FEATURES) | |
1151 | version = SPA_VERSION_FEATURES; | |
1152 | ||
1153 | ASSERT(SPA_VERSION_IS_SUPPORTED(version)); | |
1154 | return (version); | |
1155 | } | |
1156 | ||
428870ff BB |
1157 | static int |
1158 | ztest_random_blocksize(void) | |
34dc7c2f | 1159 | { |
cc99f275 DB |
1160 | ASSERT(ztest_spa->spa_max_ashift != 0); |
1161 | ||
f1512ee6 MA |
1162 | /* |
1163 | * Choose a block size >= the ashift. | |
1164 | * If the SPA supports new MAXBLOCKSIZE, test up to 1MB blocks. | |
1165 | */ | |
1166 | int maxbs = SPA_OLD_MAXBLOCKSHIFT; | |
1167 | if (spa_maxblocksize(ztest_spa) == SPA_MAXBLOCKSIZE) | |
1168 | maxbs = 20; | |
c3520e7f MA |
1169 | uint64_t block_shift = |
1170 | ztest_random(maxbs - ztest_spa->spa_max_ashift + 1); | |
b02fe35d | 1171 | return (1 << (SPA_MINBLOCKSHIFT + block_shift)); |
428870ff | 1172 | } |
34dc7c2f | 1173 | |
50c957f7 NB |
1174 | static int |
1175 | ztest_random_dnodesize(void) | |
1176 | { | |
1177 | int slots; | |
1178 | int max_slots = spa_maxdnodesize(ztest_spa) >> DNODE_SHIFT; | |
1179 | ||
1180 | if (max_slots == DNODE_MIN_SLOTS) | |
1181 | return (DNODE_MIN_SIZE); | |
1182 | ||
1183 | /* | |
1184 | * Weight the random distribution more heavily toward smaller | |
1185 | * dnode sizes since that is more likely to reflect real-world | |
1186 | * usage. | |
1187 | */ | |
1188 | ASSERT3U(max_slots, >, 4); | |
1189 | switch (ztest_random(10)) { | |
1190 | case 0: | |
1191 | slots = 5 + ztest_random(max_slots - 4); | |
1192 | break; | |
1193 | case 1 ... 4: | |
1194 | slots = 2 + ztest_random(3); | |
1195 | break; | |
1196 | default: | |
1197 | slots = 1; | |
1198 | break; | |
1199 | } | |
1200 | ||
1201 | return (slots << DNODE_SHIFT); | |
1202 | } | |
1203 | ||
428870ff BB |
1204 | static int |
1205 | ztest_random_ibshift(void) | |
1206 | { | |
1207 | return (DN_MIN_INDBLKSHIFT + | |
1208 | ztest_random(DN_MAX_INDBLKSHIFT - DN_MIN_INDBLKSHIFT + 1)); | |
34dc7c2f BB |
1209 | } |
1210 | ||
428870ff BB |
1211 | static uint64_t |
1212 | ztest_random_vdev_top(spa_t *spa, boolean_t log_ok) | |
34dc7c2f | 1213 | { |
428870ff BB |
1214 | uint64_t top; |
1215 | vdev_t *rvd = spa->spa_root_vdev; | |
1216 | vdev_t *tvd; | |
34dc7c2f | 1217 | |
428870ff | 1218 | ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); |
34dc7c2f | 1219 | |
428870ff BB |
1220 | do { |
1221 | top = ztest_random(rvd->vdev_children); | |
1222 | tvd = rvd->vdev_child[top]; | |
a1d477c2 | 1223 | } while (!vdev_is_concrete(tvd) || (tvd->vdev_islog && !log_ok) || |
428870ff | 1224 | tvd->vdev_mg == NULL || tvd->vdev_mg->mg_class == NULL); |
34dc7c2f | 1225 | |
428870ff | 1226 | return (top); |
34dc7c2f BB |
1227 | } |
1228 | ||
428870ff BB |
1229 | static uint64_t |
1230 | ztest_random_dsl_prop(zfs_prop_t prop) | |
34dc7c2f | 1231 | { |
428870ff BB |
1232 | uint64_t value; |
1233 | ||
1234 | do { | |
1235 | value = zfs_prop_random_value(prop, ztest_random(-1ULL)); | |
1236 | } while (prop == ZFS_PROP_CHECKSUM && value == ZIO_CHECKSUM_OFF); | |
1237 | ||
1238 | return (value); | |
34dc7c2f BB |
1239 | } |
1240 | ||
34dc7c2f | 1241 | static int |
428870ff BB |
1242 | ztest_dsl_prop_set_uint64(char *osname, zfs_prop_t prop, uint64_t value, |
1243 | boolean_t inherit) | |
34dc7c2f | 1244 | { |
428870ff BB |
1245 | const char *propname = zfs_prop_to_name(prop); |
1246 | const char *valname; | |
40b84e7a | 1247 | char *setpoint; |
428870ff | 1248 | uint64_t curval; |
34dc7c2f BB |
1249 | int error; |
1250 | ||
13fe0198 MA |
1251 | error = dsl_prop_set_int(osname, propname, |
1252 | (inherit ? ZPROP_SRC_NONE : ZPROP_SRC_LOCAL), value); | |
34dc7c2f | 1253 | |
428870ff BB |
1254 | if (error == ENOSPC) { |
1255 | ztest_record_enospc(FTAG); | |
34dc7c2f BB |
1256 | return (error); |
1257 | } | |
c99c9001 | 1258 | ASSERT0(error); |
34dc7c2f | 1259 | |
40b84e7a | 1260 | setpoint = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
13fe0198 | 1261 | VERIFY0(dsl_prop_get_integer(osname, propname, &curval, setpoint)); |
428870ff | 1262 | |
c242c188 | 1263 | if (ztest_opts.zo_verbose >= 6) { |
6bec4351 TC |
1264 | int err; |
1265 | ||
1266 | err = zfs_prop_index_to_string(prop, curval, &valname); | |
1267 | if (err) | |
02730c33 BB |
1268 | (void) printf("%s %s = %llu at '%s'\n", osname, |
1269 | propname, (unsigned long long)curval, setpoint); | |
6bec4351 TC |
1270 | else |
1271 | (void) printf("%s %s = %s at '%s'\n", | |
1272 | osname, propname, valname, setpoint); | |
34dc7c2f | 1273 | } |
40b84e7a | 1274 | umem_free(setpoint, MAXPATHLEN); |
34dc7c2f BB |
1275 | |
1276 | return (error); | |
1277 | } | |
1278 | ||
1279 | static int | |
c242c188 | 1280 | ztest_spa_prop_set_uint64(zpool_prop_t prop, uint64_t value) |
34dc7c2f | 1281 | { |
c242c188 | 1282 | spa_t *spa = ztest_spa; |
428870ff | 1283 | nvlist_t *props = NULL; |
34dc7c2f BB |
1284 | int error; |
1285 | ||
428870ff BB |
1286 | VERIFY(nvlist_alloc(&props, NV_UNIQUE_NAME, 0) == 0); |
1287 | VERIFY(nvlist_add_uint64(props, zpool_prop_to_name(prop), value) == 0); | |
34dc7c2f | 1288 | |
428870ff BB |
1289 | error = spa_prop_set(spa, props); |
1290 | ||
1291 | nvlist_free(props); | |
1292 | ||
1293 | if (error == ENOSPC) { | |
1294 | ztest_record_enospc(FTAG); | |
34dc7c2f BB |
1295 | return (error); |
1296 | } | |
c99c9001 | 1297 | ASSERT0(error); |
34dc7c2f BB |
1298 | |
1299 | return (error); | |
1300 | } | |
1301 | ||
4807c0ba TC |
1302 | static int |
1303 | ztest_dmu_objset_own(const char *name, dmu_objset_type_t type, | |
1304 | boolean_t readonly, boolean_t decrypt, void *tag, objset_t **osp) | |
1305 | { | |
1306 | int err; | |
1307 | ||
1308 | err = dmu_objset_own(name, type, readonly, decrypt, tag, osp); | |
1309 | if (decrypt && err == EACCES) { | |
1310 | char ddname[ZFS_MAX_DATASET_NAME_LEN]; | |
1311 | dsl_crypto_params_t *dcp; | |
1312 | nvlist_t *crypto_args = fnvlist_alloc(); | |
1313 | char *cp = NULL; | |
1314 | ||
1315 | /* spa_keystore_load_wkey() expects a dsl dir name */ | |
1316 | strcpy(ddname, name); | |
1317 | cp = strchr(ddname, '@'); | |
1318 | if (cp != NULL) | |
1319 | *cp = '\0'; | |
1320 | ||
1321 | fnvlist_add_uint8_array(crypto_args, "wkeydata", | |
1322 | (uint8_t *)ztest_wkeydata, WRAPPING_KEY_LEN); | |
1323 | VERIFY0(dsl_crypto_params_create_nvlist(DCP_CMD_NONE, NULL, | |
1324 | crypto_args, &dcp)); | |
1325 | err = spa_keystore_load_wkey(ddname, dcp, B_FALSE); | |
1326 | dsl_crypto_params_free(dcp, B_FALSE); | |
1327 | fnvlist_free(crypto_args); | |
1328 | ||
1329 | if (err != 0) | |
1330 | return (err); | |
1331 | ||
1332 | err = dmu_objset_own(name, type, readonly, decrypt, tag, osp); | |
1333 | } | |
1334 | ||
1335 | return (err); | |
1336 | } | |
1337 | ||
428870ff BB |
1338 | static void |
1339 | ztest_rll_init(rll_t *rll) | |
1340 | { | |
1341 | rll->rll_writer = NULL; | |
1342 | rll->rll_readers = 0; | |
1e33ac1e BB |
1343 | mutex_init(&rll->rll_lock, NULL, MUTEX_DEFAULT, NULL); |
1344 | cv_init(&rll->rll_cv, NULL, CV_DEFAULT, NULL); | |
428870ff | 1345 | } |
34dc7c2f | 1346 | |
428870ff BB |
1347 | static void |
1348 | ztest_rll_destroy(rll_t *rll) | |
34dc7c2f | 1349 | { |
428870ff BB |
1350 | ASSERT(rll->rll_writer == NULL); |
1351 | ASSERT(rll->rll_readers == 0); | |
1e33ac1e BB |
1352 | mutex_destroy(&rll->rll_lock); |
1353 | cv_destroy(&rll->rll_cv); | |
428870ff | 1354 | } |
34dc7c2f | 1355 | |
428870ff BB |
1356 | static void |
1357 | ztest_rll_lock(rll_t *rll, rl_type_t type) | |
1358 | { | |
1e33ac1e | 1359 | mutex_enter(&rll->rll_lock); |
34dc7c2f | 1360 | |
428870ff BB |
1361 | if (type == RL_READER) { |
1362 | while (rll->rll_writer != NULL) | |
1e33ac1e | 1363 | (void) cv_wait(&rll->rll_cv, &rll->rll_lock); |
428870ff BB |
1364 | rll->rll_readers++; |
1365 | } else { | |
1366 | while (rll->rll_writer != NULL || rll->rll_readers) | |
1e33ac1e | 1367 | (void) cv_wait(&rll->rll_cv, &rll->rll_lock); |
428870ff BB |
1368 | rll->rll_writer = curthread; |
1369 | } | |
34dc7c2f | 1370 | |
1e33ac1e | 1371 | mutex_exit(&rll->rll_lock); |
428870ff | 1372 | } |
34dc7c2f | 1373 | |
428870ff BB |
1374 | static void |
1375 | ztest_rll_unlock(rll_t *rll) | |
1376 | { | |
1e33ac1e | 1377 | mutex_enter(&rll->rll_lock); |
34dc7c2f | 1378 | |
428870ff BB |
1379 | if (rll->rll_writer) { |
1380 | ASSERT(rll->rll_readers == 0); | |
1381 | rll->rll_writer = NULL; | |
1382 | } else { | |
1383 | ASSERT(rll->rll_readers != 0); | |
1384 | ASSERT(rll->rll_writer == NULL); | |
1385 | rll->rll_readers--; | |
1386 | } | |
34dc7c2f | 1387 | |
428870ff | 1388 | if (rll->rll_writer == NULL && rll->rll_readers == 0) |
1e33ac1e | 1389 | cv_broadcast(&rll->rll_cv); |
428870ff | 1390 | |
1e33ac1e | 1391 | mutex_exit(&rll->rll_lock); |
34dc7c2f BB |
1392 | } |
1393 | ||
428870ff BB |
1394 | static void |
1395 | ztest_object_lock(ztest_ds_t *zd, uint64_t object, rl_type_t type) | |
b128c09f | 1396 | { |
428870ff | 1397 | rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; |
b128c09f | 1398 | |
428870ff BB |
1399 | ztest_rll_lock(rll, type); |
1400 | } | |
b128c09f | 1401 | |
428870ff BB |
1402 | static void |
1403 | ztest_object_unlock(ztest_ds_t *zd, uint64_t object) | |
1404 | { | |
1405 | rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; | |
b128c09f | 1406 | |
428870ff | 1407 | ztest_rll_unlock(rll); |
b128c09f BB |
1408 | } |
1409 | ||
5d43cc9a | 1410 | static rl_t * |
e3a07cd0 BP |
1411 | ztest_range_lock(ztest_ds_t *zd, uint64_t object, uint64_t offset, |
1412 | uint64_t size, rl_type_t type) | |
1413 | { | |
5d43cc9a MA |
1414 | uint64_t hash = object ^ (offset % (ZTEST_RANGE_LOCKS + 1)); |
1415 | rll_t *rll = &zd->zd_range_lock[hash & (ZTEST_RANGE_LOCKS - 1)]; | |
1416 | rl_t *rl; | |
1417 | ||
1418 | rl = umem_alloc(sizeof (*rl), UMEM_NOFAIL); | |
1419 | rl->rl_object = object; | |
1420 | rl->rl_offset = offset; | |
1421 | rl->rl_size = size; | |
1422 | rl->rl_lock = rll; | |
1423 | ||
1424 | ztest_rll_lock(rll, type); | |
1425 | ||
1426 | return (rl); | |
e3a07cd0 | 1427 | } |
34dc7c2f | 1428 | |
e3a07cd0 | 1429 | static void |
5d43cc9a | 1430 | ztest_range_unlock(rl_t *rl) |
e3a07cd0 | 1431 | { |
5d43cc9a MA |
1432 | rll_t *rll = rl->rl_lock; |
1433 | ||
1434 | ztest_rll_unlock(rll); | |
1435 | ||
1436 | umem_free(rl, sizeof (*rl)); | |
428870ff | 1437 | } |
34dc7c2f | 1438 | |
428870ff | 1439 | static void |
c242c188 | 1440 | ztest_zd_init(ztest_ds_t *zd, ztest_shared_ds_t *szd, objset_t *os) |
428870ff BB |
1441 | { |
1442 | zd->zd_os = os; | |
1443 | zd->zd_zilog = dmu_objset_zil(os); | |
c242c188 | 1444 | zd->zd_shared = szd; |
428870ff | 1445 | dmu_objset_name(os, zd->zd_name); |
d6320ddb | 1446 | int l; |
428870ff | 1447 | |
c242c188 CS |
1448 | if (zd->zd_shared != NULL) |
1449 | zd->zd_shared->zd_seq = 0; | |
1450 | ||
2d9142c9 | 1451 | VERIFY0(pthread_rwlock_init(&zd->zd_zilog_lock, NULL)); |
1e33ac1e | 1452 | mutex_init(&zd->zd_dirobj_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 1453 | |
d6320ddb | 1454 | for (l = 0; l < ZTEST_OBJECT_LOCKS; l++) |
428870ff | 1455 | ztest_rll_init(&zd->zd_object_lock[l]); |
34dc7c2f | 1456 | |
d6320ddb | 1457 | for (l = 0; l < ZTEST_RANGE_LOCKS; l++) |
5d43cc9a | 1458 | ztest_rll_init(&zd->zd_range_lock[l]); |
34dc7c2f BB |
1459 | } |
1460 | ||
428870ff BB |
1461 | static void |
1462 | ztest_zd_fini(ztest_ds_t *zd) | |
34dc7c2f | 1463 | { |
d6320ddb BB |
1464 | int l; |
1465 | ||
1e33ac1e | 1466 | mutex_destroy(&zd->zd_dirobj_lock); |
2d9142c9 | 1467 | (void) pthread_rwlock_destroy(&zd->zd_zilog_lock); |
34dc7c2f | 1468 | |
d6320ddb | 1469 | for (l = 0; l < ZTEST_OBJECT_LOCKS; l++) |
428870ff | 1470 | ztest_rll_destroy(&zd->zd_object_lock[l]); |
b128c09f | 1471 | |
d6320ddb | 1472 | for (l = 0; l < ZTEST_RANGE_LOCKS; l++) |
5d43cc9a | 1473 | ztest_rll_destroy(&zd->zd_range_lock[l]); |
428870ff | 1474 | } |
b128c09f | 1475 | |
428870ff | 1476 | #define TXG_MIGHTWAIT (ztest_random(10) == 0 ? TXG_NOWAIT : TXG_WAIT) |
b128c09f | 1477 | |
428870ff BB |
1478 | static uint64_t |
1479 | ztest_tx_assign(dmu_tx_t *tx, uint64_t txg_how, const char *tag) | |
1480 | { | |
1481 | uint64_t txg; | |
1482 | int error; | |
1483 | ||
1484 | /* | |
1485 | * Attempt to assign tx to some transaction group. | |
1486 | */ | |
1487 | error = dmu_tx_assign(tx, txg_how); | |
1488 | if (error) { | |
1489 | if (error == ERESTART) { | |
1490 | ASSERT(txg_how == TXG_NOWAIT); | |
1491 | dmu_tx_wait(tx); | |
1492 | } else { | |
1493 | ASSERT3U(error, ==, ENOSPC); | |
1494 | ztest_record_enospc(tag); | |
1495 | } | |
1496 | dmu_tx_abort(tx); | |
1497 | return (0); | |
1498 | } | |
1499 | txg = dmu_tx_get_txg(tx); | |
1500 | ASSERT(txg != 0); | |
1501 | return (txg); | |
1502 | } | |
1503 | ||
1504 | static void | |
1505 | ztest_pattern_set(void *buf, uint64_t size, uint64_t value) | |
1506 | { | |
1507 | uint64_t *ip = buf; | |
1508 | uint64_t *ip_end = (uint64_t *)((uintptr_t)buf + (uintptr_t)size); | |
1509 | ||
1510 | while (ip < ip_end) | |
1511 | *ip++ = value; | |
1512 | } | |
1513 | ||
1fde1e37 | 1514 | #ifndef NDEBUG |
428870ff BB |
1515 | static boolean_t |
1516 | ztest_pattern_match(void *buf, uint64_t size, uint64_t value) | |
1517 | { | |
1518 | uint64_t *ip = buf; | |
1519 | uint64_t *ip_end = (uint64_t *)((uintptr_t)buf + (uintptr_t)size); | |
1520 | uint64_t diff = 0; | |
1521 | ||
1522 | while (ip < ip_end) | |
1523 | diff |= (value - *ip++); | |
1524 | ||
1525 | return (diff == 0); | |
1526 | } | |
1fde1e37 | 1527 | #endif |
428870ff BB |
1528 | |
1529 | static void | |
1530 | ztest_bt_generate(ztest_block_tag_t *bt, objset_t *os, uint64_t object, | |
50c957f7 NB |
1531 | uint64_t dnodesize, uint64_t offset, uint64_t gen, uint64_t txg, |
1532 | uint64_t crtxg) | |
428870ff BB |
1533 | { |
1534 | bt->bt_magic = BT_MAGIC; | |
1535 | bt->bt_objset = dmu_objset_id(os); | |
1536 | bt->bt_object = object; | |
50c957f7 | 1537 | bt->bt_dnodesize = dnodesize; |
428870ff BB |
1538 | bt->bt_offset = offset; |
1539 | bt->bt_gen = gen; | |
1540 | bt->bt_txg = txg; | |
1541 | bt->bt_crtxg = crtxg; | |
1542 | } | |
1543 | ||
1544 | static void | |
1545 | ztest_bt_verify(ztest_block_tag_t *bt, objset_t *os, uint64_t object, | |
50c957f7 NB |
1546 | uint64_t dnodesize, uint64_t offset, uint64_t gen, uint64_t txg, |
1547 | uint64_t crtxg) | |
428870ff | 1548 | { |
9b67f605 MA |
1549 | ASSERT3U(bt->bt_magic, ==, BT_MAGIC); |
1550 | ASSERT3U(bt->bt_objset, ==, dmu_objset_id(os)); | |
1551 | ASSERT3U(bt->bt_object, ==, object); | |
50c957f7 | 1552 | ASSERT3U(bt->bt_dnodesize, ==, dnodesize); |
9b67f605 MA |
1553 | ASSERT3U(bt->bt_offset, ==, offset); |
1554 | ASSERT3U(bt->bt_gen, <=, gen); | |
1555 | ASSERT3U(bt->bt_txg, <=, txg); | |
1556 | ASSERT3U(bt->bt_crtxg, ==, crtxg); | |
428870ff BB |
1557 | } |
1558 | ||
1559 | static ztest_block_tag_t * | |
1560 | ztest_bt_bonus(dmu_buf_t *db) | |
1561 | { | |
1562 | dmu_object_info_t doi; | |
1563 | ztest_block_tag_t *bt; | |
1564 | ||
1565 | dmu_object_info_from_db(db, &doi); | |
1566 | ASSERT3U(doi.doi_bonus_size, <=, db->db_size); | |
1567 | ASSERT3U(doi.doi_bonus_size, >=, sizeof (*bt)); | |
1568 | bt = (void *)((char *)db->db_data + doi.doi_bonus_size - sizeof (*bt)); | |
1569 | ||
1570 | return (bt); | |
1571 | } | |
1572 | ||
50c957f7 NB |
1573 | /* |
1574 | * Generate a token to fill up unused bonus buffer space. Try to make | |
1575 | * it unique to the object, generation, and offset to verify that data | |
1576 | * is not getting overwritten by data from other dnodes. | |
1577 | */ | |
1578 | #define ZTEST_BONUS_FILL_TOKEN(obj, ds, gen, offset) \ | |
1579 | (((ds) << 48) | ((gen) << 32) | ((obj) << 8) | (offset)) | |
1580 | ||
1581 | /* | |
1582 | * Fill up the unused bonus buffer region before the block tag with a | |
1583 | * verifiable pattern. Filling the whole bonus area with non-zero data | |
1584 | * helps ensure that all dnode traversal code properly skips the | |
1585 | * interior regions of large dnodes. | |
1586 | */ | |
1587 | void | |
1588 | ztest_fill_unused_bonus(dmu_buf_t *db, void *end, uint64_t obj, | |
1589 | objset_t *os, uint64_t gen) | |
1590 | { | |
1591 | uint64_t *bonusp; | |
1592 | ||
1593 | ASSERT(IS_P2ALIGNED((char *)end - (char *)db->db_data, 8)); | |
1594 | ||
1595 | for (bonusp = db->db_data; bonusp < (uint64_t *)end; bonusp++) { | |
1596 | uint64_t token = ZTEST_BONUS_FILL_TOKEN(obj, dmu_objset_id(os), | |
1597 | gen, bonusp - (uint64_t *)db->db_data); | |
1598 | *bonusp = token; | |
1599 | } | |
1600 | } | |
1601 | ||
1602 | /* | |
1603 | * Verify that the unused area of a bonus buffer is filled with the | |
1604 | * expected tokens. | |
1605 | */ | |
1606 | void | |
1607 | ztest_verify_unused_bonus(dmu_buf_t *db, void *end, uint64_t obj, | |
1608 | objset_t *os, uint64_t gen) | |
1609 | { | |
1610 | uint64_t *bonusp; | |
1611 | ||
1612 | for (bonusp = db->db_data; bonusp < (uint64_t *)end; bonusp++) { | |
1613 | uint64_t token = ZTEST_BONUS_FILL_TOKEN(obj, dmu_objset_id(os), | |
1614 | gen, bonusp - (uint64_t *)db->db_data); | |
1615 | VERIFY3U(*bonusp, ==, token); | |
1616 | } | |
1617 | } | |
1618 | ||
428870ff BB |
1619 | /* |
1620 | * ZIL logging ops | |
1621 | */ | |
1622 | ||
1623 | #define lrz_type lr_mode | |
1624 | #define lrz_blocksize lr_uid | |
1625 | #define lrz_ibshift lr_gid | |
1626 | #define lrz_bonustype lr_rdev | |
50c957f7 | 1627 | #define lrz_dnodesize lr_crtime[1] |
428870ff | 1628 | |
572e2857 | 1629 | static void |
428870ff BB |
1630 | ztest_log_create(ztest_ds_t *zd, dmu_tx_t *tx, lr_create_t *lr) |
1631 | { | |
1632 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1633 | size_t namesize = strlen(name) + 1; | |
1634 | itx_t *itx; | |
1635 | ||
1636 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1637 | return; |
428870ff BB |
1638 | |
1639 | itx = zil_itx_create(TX_CREATE, sizeof (*lr) + namesize); | |
1640 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1641 | sizeof (*lr) + namesize - sizeof (lr_t)); | |
1642 | ||
572e2857 | 1643 | zil_itx_assign(zd->zd_zilog, itx, tx); |
428870ff BB |
1644 | } |
1645 | ||
572e2857 BB |
1646 | static void |
1647 | ztest_log_remove(ztest_ds_t *zd, dmu_tx_t *tx, lr_remove_t *lr, uint64_t object) | |
428870ff BB |
1648 | { |
1649 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1650 | size_t namesize = strlen(name) + 1; | |
1651 | itx_t *itx; | |
1652 | ||
1653 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1654 | return; |
428870ff BB |
1655 | |
1656 | itx = zil_itx_create(TX_REMOVE, sizeof (*lr) + namesize); | |
1657 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1658 | sizeof (*lr) + namesize - sizeof (lr_t)); | |
1659 | ||
572e2857 BB |
1660 | itx->itx_oid = object; |
1661 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1662 | } |
1663 | ||
572e2857 | 1664 | static void |
428870ff BB |
1665 | ztest_log_write(ztest_ds_t *zd, dmu_tx_t *tx, lr_write_t *lr) |
1666 | { | |
1667 | itx_t *itx; | |
1668 | itx_wr_state_t write_state = ztest_random(WR_NUM_STATES); | |
1669 | ||
1670 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1671 | return; |
428870ff BB |
1672 | |
1673 | if (lr->lr_length > ZIL_MAX_LOG_DATA) | |
1674 | write_state = WR_INDIRECT; | |
1675 | ||
1676 | itx = zil_itx_create(TX_WRITE, | |
1677 | sizeof (*lr) + (write_state == WR_COPIED ? lr->lr_length : 0)); | |
1678 | ||
1679 | if (write_state == WR_COPIED && | |
1680 | dmu_read(zd->zd_os, lr->lr_foid, lr->lr_offset, lr->lr_length, | |
1681 | ((lr_write_t *)&itx->itx_lr) + 1, DMU_READ_NO_PREFETCH) != 0) { | |
1682 | zil_itx_destroy(itx); | |
1683 | itx = zil_itx_create(TX_WRITE, sizeof (*lr)); | |
1684 | write_state = WR_NEED_COPY; | |
1685 | } | |
1686 | itx->itx_private = zd; | |
1687 | itx->itx_wr_state = write_state; | |
1688 | itx->itx_sync = (ztest_random(8) == 0); | |
428870ff BB |
1689 | |
1690 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1691 | sizeof (*lr) - sizeof (lr_t)); | |
1692 | ||
572e2857 | 1693 | zil_itx_assign(zd->zd_zilog, itx, tx); |
428870ff BB |
1694 | } |
1695 | ||
572e2857 | 1696 | static void |
428870ff BB |
1697 | ztest_log_truncate(ztest_ds_t *zd, dmu_tx_t *tx, lr_truncate_t *lr) |
1698 | { | |
1699 | itx_t *itx; | |
1700 | ||
1701 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1702 | return; |
428870ff BB |
1703 | |
1704 | itx = zil_itx_create(TX_TRUNCATE, sizeof (*lr)); | |
1705 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1706 | sizeof (*lr) - sizeof (lr_t)); | |
1707 | ||
572e2857 BB |
1708 | itx->itx_sync = B_FALSE; |
1709 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1710 | } |
1711 | ||
572e2857 | 1712 | static void |
428870ff BB |
1713 | ztest_log_setattr(ztest_ds_t *zd, dmu_tx_t *tx, lr_setattr_t *lr) |
1714 | { | |
1715 | itx_t *itx; | |
1716 | ||
1717 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1718 | return; |
428870ff BB |
1719 | |
1720 | itx = zil_itx_create(TX_SETATTR, sizeof (*lr)); | |
1721 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1722 | sizeof (*lr) - sizeof (lr_t)); | |
1723 | ||
572e2857 BB |
1724 | itx->itx_sync = B_FALSE; |
1725 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1726 | } |
1727 | ||
1728 | /* | |
1729 | * ZIL replay ops | |
1730 | */ | |
1731 | static int | |
867959b5 | 1732 | ztest_replay_create(void *arg1, void *arg2, boolean_t byteswap) |
428870ff | 1733 | { |
867959b5 BB |
1734 | ztest_ds_t *zd = arg1; |
1735 | lr_create_t *lr = arg2; | |
428870ff BB |
1736 | char *name = (void *)(lr + 1); /* name follows lr */ |
1737 | objset_t *os = zd->zd_os; | |
1738 | ztest_block_tag_t *bbt; | |
1739 | dmu_buf_t *db; | |
1740 | dmu_tx_t *tx; | |
1741 | uint64_t txg; | |
1742 | int error = 0; | |
50c957f7 | 1743 | int bonuslen; |
428870ff BB |
1744 | |
1745 | if (byteswap) | |
1746 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1747 | ||
1748 | ASSERT(lr->lr_doid == ZTEST_DIROBJ); | |
1749 | ASSERT(name[0] != '\0'); | |
1750 | ||
1751 | tx = dmu_tx_create(os); | |
1752 | ||
1753 | dmu_tx_hold_zap(tx, lr->lr_doid, B_TRUE, name); | |
1754 | ||
1755 | if (lr->lrz_type == DMU_OT_ZAP_OTHER) { | |
1756 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, B_TRUE, NULL); | |
1757 | } else { | |
1758 | dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT); | |
1759 | } | |
1760 | ||
1761 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1762 | if (txg == 0) | |
1763 | return (ENOSPC); | |
1764 | ||
1765 | ASSERT(dmu_objset_zil(os)->zl_replay == !!lr->lr_foid); | |
50c957f7 | 1766 | bonuslen = DN_BONUS_SIZE(lr->lrz_dnodesize); |
428870ff BB |
1767 | |
1768 | if (lr->lrz_type == DMU_OT_ZAP_OTHER) { | |
1769 | if (lr->lr_foid == 0) { | |
50c957f7 | 1770 | lr->lr_foid = zap_create_dnsize(os, |
428870ff | 1771 | lr->lrz_type, lr->lrz_bonustype, |
50c957f7 | 1772 | bonuslen, lr->lrz_dnodesize, tx); |
428870ff | 1773 | } else { |
50c957f7 | 1774 | error = zap_create_claim_dnsize(os, lr->lr_foid, |
428870ff | 1775 | lr->lrz_type, lr->lrz_bonustype, |
50c957f7 | 1776 | bonuslen, lr->lrz_dnodesize, tx); |
428870ff BB |
1777 | } |
1778 | } else { | |
1779 | if (lr->lr_foid == 0) { | |
50c957f7 | 1780 | lr->lr_foid = dmu_object_alloc_dnsize(os, |
428870ff | 1781 | lr->lrz_type, 0, lr->lrz_bonustype, |
50c957f7 | 1782 | bonuslen, lr->lrz_dnodesize, tx); |
428870ff | 1783 | } else { |
50c957f7 | 1784 | error = dmu_object_claim_dnsize(os, lr->lr_foid, |
428870ff | 1785 | lr->lrz_type, 0, lr->lrz_bonustype, |
50c957f7 | 1786 | bonuslen, lr->lrz_dnodesize, tx); |
428870ff BB |
1787 | } |
1788 | } | |
1789 | ||
1790 | if (error) { | |
1791 | ASSERT3U(error, ==, EEXIST); | |
1792 | ASSERT(zd->zd_zilog->zl_replay); | |
1793 | dmu_tx_commit(tx); | |
1794 | return (error); | |
1795 | } | |
1796 | ||
1797 | ASSERT(lr->lr_foid != 0); | |
1798 | ||
1799 | if (lr->lrz_type != DMU_OT_ZAP_OTHER) | |
1800 | VERIFY3U(0, ==, dmu_object_set_blocksize(os, lr->lr_foid, | |
1801 | lr->lrz_blocksize, lr->lrz_ibshift, tx)); | |
1802 | ||
1803 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
1804 | bbt = ztest_bt_bonus(db); | |
1805 | dmu_buf_will_dirty(db, tx); | |
50c957f7 NB |
1806 | ztest_bt_generate(bbt, os, lr->lr_foid, lr->lrz_dnodesize, -1ULL, |
1807 | lr->lr_gen, txg, txg); | |
1808 | ztest_fill_unused_bonus(db, bbt, lr->lr_foid, os, lr->lr_gen); | |
428870ff BB |
1809 | dmu_buf_rele(db, FTAG); |
1810 | ||
1811 | VERIFY3U(0, ==, zap_add(os, lr->lr_doid, name, sizeof (uint64_t), 1, | |
1812 | &lr->lr_foid, tx)); | |
1813 | ||
1814 | (void) ztest_log_create(zd, tx, lr); | |
1815 | ||
1816 | dmu_tx_commit(tx); | |
1817 | ||
1818 | return (0); | |
1819 | } | |
1820 | ||
1821 | static int | |
867959b5 | 1822 | ztest_replay_remove(void *arg1, void *arg2, boolean_t byteswap) |
428870ff | 1823 | { |
867959b5 BB |
1824 | ztest_ds_t *zd = arg1; |
1825 | lr_remove_t *lr = arg2; | |
428870ff BB |
1826 | char *name = (void *)(lr + 1); /* name follows lr */ |
1827 | objset_t *os = zd->zd_os; | |
1828 | dmu_object_info_t doi; | |
1829 | dmu_tx_t *tx; | |
1830 | uint64_t object, txg; | |
1831 | ||
1832 | if (byteswap) | |
1833 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1834 | ||
1835 | ASSERT(lr->lr_doid == ZTEST_DIROBJ); | |
1836 | ASSERT(name[0] != '\0'); | |
1837 | ||
1838 | VERIFY3U(0, ==, | |
1839 | zap_lookup(os, lr->lr_doid, name, sizeof (object), 1, &object)); | |
1840 | ASSERT(object != 0); | |
1841 | ||
1842 | ztest_object_lock(zd, object, RL_WRITER); | |
1843 | ||
1844 | VERIFY3U(0, ==, dmu_object_info(os, object, &doi)); | |
1845 | ||
1846 | tx = dmu_tx_create(os); | |
1847 | ||
1848 | dmu_tx_hold_zap(tx, lr->lr_doid, B_FALSE, name); | |
1849 | dmu_tx_hold_free(tx, object, 0, DMU_OBJECT_END); | |
1850 | ||
1851 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1852 | if (txg == 0) { | |
1853 | ztest_object_unlock(zd, object); | |
1854 | return (ENOSPC); | |
1855 | } | |
1856 | ||
1857 | if (doi.doi_type == DMU_OT_ZAP_OTHER) { | |
1858 | VERIFY3U(0, ==, zap_destroy(os, object, tx)); | |
1859 | } else { | |
1860 | VERIFY3U(0, ==, dmu_object_free(os, object, tx)); | |
1861 | } | |
1862 | ||
1863 | VERIFY3U(0, ==, zap_remove(os, lr->lr_doid, name, tx)); | |
1864 | ||
572e2857 | 1865 | (void) ztest_log_remove(zd, tx, lr, object); |
428870ff BB |
1866 | |
1867 | dmu_tx_commit(tx); | |
1868 | ||
1869 | ztest_object_unlock(zd, object); | |
1870 | ||
1871 | return (0); | |
1872 | } | |
1873 | ||
1874 | static int | |
867959b5 | 1875 | ztest_replay_write(void *arg1, void *arg2, boolean_t byteswap) |
428870ff | 1876 | { |
867959b5 BB |
1877 | ztest_ds_t *zd = arg1; |
1878 | lr_write_t *lr = arg2; | |
428870ff BB |
1879 | objset_t *os = zd->zd_os; |
1880 | void *data = lr + 1; /* data follows lr */ | |
1881 | uint64_t offset, length; | |
1882 | ztest_block_tag_t *bt = data; | |
1883 | ztest_block_tag_t *bbt; | |
1884 | uint64_t gen, txg, lrtxg, crtxg; | |
1885 | dmu_object_info_t doi; | |
1886 | dmu_tx_t *tx; | |
1887 | dmu_buf_t *db; | |
1888 | arc_buf_t *abuf = NULL; | |
5d43cc9a | 1889 | rl_t *rl; |
428870ff BB |
1890 | |
1891 | if (byteswap) | |
1892 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1893 | ||
1894 | offset = lr->lr_offset; | |
1895 | length = lr->lr_length; | |
1896 | ||
1897 | /* If it's a dmu_sync() block, write the whole block */ | |
1898 | if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) { | |
1899 | uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr); | |
1900 | if (length < blocksize) { | |
1901 | offset -= offset % blocksize; | |
1902 | length = blocksize; | |
1903 | } | |
1904 | } | |
1905 | ||
1906 | if (bt->bt_magic == BSWAP_64(BT_MAGIC)) | |
1907 | byteswap_uint64_array(bt, sizeof (*bt)); | |
1908 | ||
1909 | if (bt->bt_magic != BT_MAGIC) | |
1910 | bt = NULL; | |
1911 | ||
1912 | ztest_object_lock(zd, lr->lr_foid, RL_READER); | |
1913 | rl = ztest_range_lock(zd, lr->lr_foid, offset, length, RL_WRITER); | |
1914 | ||
1915 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
1916 | ||
1917 | dmu_object_info_from_db(db, &doi); | |
1918 | ||
1919 | bbt = ztest_bt_bonus(db); | |
1920 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
1921 | gen = bbt->bt_gen; | |
1922 | crtxg = bbt->bt_crtxg; | |
1923 | lrtxg = lr->lr_common.lrc_txg; | |
1924 | ||
1925 | tx = dmu_tx_create(os); | |
1926 | ||
1927 | dmu_tx_hold_write(tx, lr->lr_foid, offset, length); | |
1928 | ||
1929 | if (ztest_random(8) == 0 && length == doi.doi_data_block_size && | |
1930 | P2PHASE(offset, length) == 0) | |
1931 | abuf = dmu_request_arcbuf(db, length); | |
1932 | ||
1933 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1934 | if (txg == 0) { | |
1935 | if (abuf != NULL) | |
1936 | dmu_return_arcbuf(abuf); | |
1937 | dmu_buf_rele(db, FTAG); | |
5d43cc9a | 1938 | ztest_range_unlock(rl); |
428870ff BB |
1939 | ztest_object_unlock(zd, lr->lr_foid); |
1940 | return (ENOSPC); | |
1941 | } | |
1942 | ||
1943 | if (bt != NULL) { | |
1944 | /* | |
1945 | * Usually, verify the old data before writing new data -- | |
1946 | * but not always, because we also want to verify correct | |
1947 | * behavior when the data was not recently read into cache. | |
1948 | */ | |
1949 | ASSERT(offset % doi.doi_data_block_size == 0); | |
1950 | if (ztest_random(4) != 0) { | |
1951 | int prefetch = ztest_random(2) ? | |
1952 | DMU_READ_PREFETCH : DMU_READ_NO_PREFETCH; | |
1953 | ztest_block_tag_t rbt; | |
1954 | ||
1955 | VERIFY(dmu_read(os, lr->lr_foid, offset, | |
1956 | sizeof (rbt), &rbt, prefetch) == 0); | |
1957 | if (rbt.bt_magic == BT_MAGIC) { | |
50c957f7 | 1958 | ztest_bt_verify(&rbt, os, lr->lr_foid, 0, |
428870ff BB |
1959 | offset, gen, txg, crtxg); |
1960 | } | |
1961 | } | |
1962 | ||
1963 | /* | |
1964 | * Writes can appear to be newer than the bonus buffer because | |
1965 | * the ztest_get_data() callback does a dmu_read() of the | |
1966 | * open-context data, which may be different than the data | |
1967 | * as it was when the write was generated. | |
1968 | */ | |
1969 | if (zd->zd_zilog->zl_replay) { | |
50c957f7 | 1970 | ztest_bt_verify(bt, os, lr->lr_foid, 0, offset, |
428870ff BB |
1971 | MAX(gen, bt->bt_gen), MAX(txg, lrtxg), |
1972 | bt->bt_crtxg); | |
1973 | } | |
1974 | ||
1975 | /* | |
1976 | * Set the bt's gen/txg to the bonus buffer's gen/txg | |
1977 | * so that all of the usual ASSERTs will work. | |
1978 | */ | |
50c957f7 NB |
1979 | ztest_bt_generate(bt, os, lr->lr_foid, 0, offset, gen, txg, |
1980 | crtxg); | |
428870ff BB |
1981 | } |
1982 | ||
1983 | if (abuf == NULL) { | |
1984 | dmu_write(os, lr->lr_foid, offset, length, data, tx); | |
1985 | } else { | |
1986 | bcopy(data, abuf->b_data, length); | |
440a3eb9 | 1987 | dmu_assign_arcbuf_by_dbuf(db, offset, abuf, tx); |
428870ff BB |
1988 | } |
1989 | ||
1990 | (void) ztest_log_write(zd, tx, lr); | |
1991 | ||
1992 | dmu_buf_rele(db, FTAG); | |
1993 | ||
1994 | dmu_tx_commit(tx); | |
1995 | ||
5d43cc9a | 1996 | ztest_range_unlock(rl); |
428870ff BB |
1997 | ztest_object_unlock(zd, lr->lr_foid); |
1998 | ||
1999 | return (0); | |
2000 | } | |
2001 | ||
2002 | static int | |
867959b5 | 2003 | ztest_replay_truncate(void *arg1, void *arg2, boolean_t byteswap) |
428870ff | 2004 | { |
867959b5 BB |
2005 | ztest_ds_t *zd = arg1; |
2006 | lr_truncate_t *lr = arg2; | |
428870ff BB |
2007 | objset_t *os = zd->zd_os; |
2008 | dmu_tx_t *tx; | |
2009 | uint64_t txg; | |
5d43cc9a | 2010 | rl_t *rl; |
428870ff BB |
2011 | |
2012 | if (byteswap) | |
2013 | byteswap_uint64_array(lr, sizeof (*lr)); | |
2014 | ||
2015 | ztest_object_lock(zd, lr->lr_foid, RL_READER); | |
2016 | rl = ztest_range_lock(zd, lr->lr_foid, lr->lr_offset, lr->lr_length, | |
2017 | RL_WRITER); | |
2018 | ||
2019 | tx = dmu_tx_create(os); | |
2020 | ||
2021 | dmu_tx_hold_free(tx, lr->lr_foid, lr->lr_offset, lr->lr_length); | |
2022 | ||
2023 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
2024 | if (txg == 0) { | |
5d43cc9a | 2025 | ztest_range_unlock(rl); |
428870ff BB |
2026 | ztest_object_unlock(zd, lr->lr_foid); |
2027 | return (ENOSPC); | |
2028 | } | |
2029 | ||
2030 | VERIFY(dmu_free_range(os, lr->lr_foid, lr->lr_offset, | |
2031 | lr->lr_length, tx) == 0); | |
2032 | ||
2033 | (void) ztest_log_truncate(zd, tx, lr); | |
2034 | ||
2035 | dmu_tx_commit(tx); | |
2036 | ||
5d43cc9a | 2037 | ztest_range_unlock(rl); |
428870ff BB |
2038 | ztest_object_unlock(zd, lr->lr_foid); |
2039 | ||
2040 | return (0); | |
2041 | } | |
2042 | ||
2043 | static int | |
867959b5 | 2044 | ztest_replay_setattr(void *arg1, void *arg2, boolean_t byteswap) |
428870ff | 2045 | { |
867959b5 BB |
2046 | ztest_ds_t *zd = arg1; |
2047 | lr_setattr_t *lr = arg2; | |
428870ff BB |
2048 | objset_t *os = zd->zd_os; |
2049 | dmu_tx_t *tx; | |
2050 | dmu_buf_t *db; | |
2051 | ztest_block_tag_t *bbt; | |
50c957f7 | 2052 | uint64_t txg, lrtxg, crtxg, dnodesize; |
428870ff BB |
2053 | |
2054 | if (byteswap) | |
2055 | byteswap_uint64_array(lr, sizeof (*lr)); | |
2056 | ||
2057 | ztest_object_lock(zd, lr->lr_foid, RL_WRITER); | |
2058 | ||
2059 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
2060 | ||
2061 | tx = dmu_tx_create(os); | |
2062 | dmu_tx_hold_bonus(tx, lr->lr_foid); | |
2063 | ||
2064 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
2065 | if (txg == 0) { | |
2066 | dmu_buf_rele(db, FTAG); | |
2067 | ztest_object_unlock(zd, lr->lr_foid); | |
2068 | return (ENOSPC); | |
2069 | } | |
2070 | ||
2071 | bbt = ztest_bt_bonus(db); | |
2072 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
2073 | crtxg = bbt->bt_crtxg; | |
2074 | lrtxg = lr->lr_common.lrc_txg; | |
50c957f7 | 2075 | dnodesize = bbt->bt_dnodesize; |
428870ff BB |
2076 | |
2077 | if (zd->zd_zilog->zl_replay) { | |
2078 | ASSERT(lr->lr_size != 0); | |
2079 | ASSERT(lr->lr_mode != 0); | |
2080 | ASSERT(lrtxg != 0); | |
2081 | } else { | |
2082 | /* | |
2083 | * Randomly change the size and increment the generation. | |
2084 | */ | |
2085 | lr->lr_size = (ztest_random(db->db_size / sizeof (*bbt)) + 1) * | |
2086 | sizeof (*bbt); | |
2087 | lr->lr_mode = bbt->bt_gen + 1; | |
2088 | ASSERT(lrtxg == 0); | |
2089 | } | |
2090 | ||
2091 | /* | |
2092 | * Verify that the current bonus buffer is not newer than our txg. | |
2093 | */ | |
50c957f7 | 2094 | ztest_bt_verify(bbt, os, lr->lr_foid, dnodesize, -1ULL, lr->lr_mode, |
428870ff BB |
2095 | MAX(txg, lrtxg), crtxg); |
2096 | ||
2097 | dmu_buf_will_dirty(db, tx); | |
2098 | ||
2099 | ASSERT3U(lr->lr_size, >=, sizeof (*bbt)); | |
2100 | ASSERT3U(lr->lr_size, <=, db->db_size); | |
c99c9001 | 2101 | VERIFY0(dmu_set_bonus(db, lr->lr_size, tx)); |
428870ff BB |
2102 | bbt = ztest_bt_bonus(db); |
2103 | ||
50c957f7 NB |
2104 | ztest_bt_generate(bbt, os, lr->lr_foid, dnodesize, -1ULL, lr->lr_mode, |
2105 | txg, crtxg); | |
2106 | ztest_fill_unused_bonus(db, bbt, lr->lr_foid, os, bbt->bt_gen); | |
428870ff BB |
2107 | dmu_buf_rele(db, FTAG); |
2108 | ||
2109 | (void) ztest_log_setattr(zd, tx, lr); | |
2110 | ||
2111 | dmu_tx_commit(tx); | |
2112 | ||
2113 | ztest_object_unlock(zd, lr->lr_foid); | |
2114 | ||
2115 | return (0); | |
2116 | } | |
2117 | ||
867959b5 BB |
2118 | zil_replay_func_t *ztest_replay_vector[TX_MAX_TYPE] = { |
2119 | NULL, /* 0 no such transaction type */ | |
2120 | ztest_replay_create, /* TX_CREATE */ | |
2121 | NULL, /* TX_MKDIR */ | |
2122 | NULL, /* TX_MKXATTR */ | |
2123 | NULL, /* TX_SYMLINK */ | |
2124 | ztest_replay_remove, /* TX_REMOVE */ | |
2125 | NULL, /* TX_RMDIR */ | |
2126 | NULL, /* TX_LINK */ | |
2127 | NULL, /* TX_RENAME */ | |
2128 | ztest_replay_write, /* TX_WRITE */ | |
2129 | ztest_replay_truncate, /* TX_TRUNCATE */ | |
2130 | ztest_replay_setattr, /* TX_SETATTR */ | |
2131 | NULL, /* TX_ACL */ | |
2132 | NULL, /* TX_CREATE_ACL */ | |
2133 | NULL, /* TX_CREATE_ATTR */ | |
2134 | NULL, /* TX_CREATE_ACL_ATTR */ | |
2135 | NULL, /* TX_MKDIR_ACL */ | |
2136 | NULL, /* TX_MKDIR_ATTR */ | |
2137 | NULL, /* TX_MKDIR_ACL_ATTR */ | |
2138 | NULL, /* TX_WRITE2 */ | |
428870ff BB |
2139 | }; |
2140 | ||
2141 | /* | |
2142 | * ZIL get_data callbacks | |
2143 | */ | |
2144 | ||
2145 | static void | |
2146 | ztest_get_done(zgd_t *zgd, int error) | |
2147 | { | |
5d43cc9a MA |
2148 | ztest_ds_t *zd = zgd->zgd_private; |
2149 | uint64_t object = ((rl_t *)zgd->zgd_lr)->rl_object; | |
428870ff BB |
2150 | |
2151 | if (zgd->zgd_db) | |
2152 | dmu_buf_rele(zgd->zgd_db, zgd); | |
2153 | ||
5d43cc9a | 2154 | ztest_range_unlock((rl_t *)zgd->zgd_lr); |
428870ff BB |
2155 | ztest_object_unlock(zd, object); |
2156 | ||
2157 | if (error == 0 && zgd->zgd_bp) | |
1ce23dca | 2158 | zil_lwb_add_block(zgd->zgd_lwb, zgd->zgd_bp); |
428870ff BB |
2159 | |
2160 | umem_free(zgd, sizeof (*zgd)); | |
2161 | } | |
2162 | ||
2163 | static int | |
1ce23dca PS |
2164 | ztest_get_data(void *arg, lr_write_t *lr, char *buf, struct lwb *lwb, |
2165 | zio_t *zio) | |
428870ff BB |
2166 | { |
2167 | ztest_ds_t *zd = arg; | |
2168 | objset_t *os = zd->zd_os; | |
2169 | uint64_t object = lr->lr_foid; | |
2170 | uint64_t offset = lr->lr_offset; | |
2171 | uint64_t size = lr->lr_length; | |
428870ff BB |
2172 | uint64_t txg = lr->lr_common.lrc_txg; |
2173 | uint64_t crtxg; | |
2174 | dmu_object_info_t doi; | |
2175 | dmu_buf_t *db; | |
2176 | zgd_t *zgd; | |
2177 | int error; | |
2178 | ||
1ce23dca PS |
2179 | ASSERT3P(lwb, !=, NULL); |
2180 | ASSERT3P(zio, !=, NULL); | |
2181 | ASSERT3U(size, !=, 0); | |
2182 | ||
428870ff BB |
2183 | ztest_object_lock(zd, object, RL_READER); |
2184 | error = dmu_bonus_hold(os, object, FTAG, &db); | |
2185 | if (error) { | |
2186 | ztest_object_unlock(zd, object); | |
2187 | return (error); | |
2188 | } | |
2189 | ||
2190 | crtxg = ztest_bt_bonus(db)->bt_crtxg; | |
2191 | ||
2192 | if (crtxg == 0 || crtxg > txg) { | |
2193 | dmu_buf_rele(db, FTAG); | |
2194 | ztest_object_unlock(zd, object); | |
2195 | return (ENOENT); | |
2196 | } | |
2197 | ||
2198 | dmu_object_info_from_db(db, &doi); | |
2199 | dmu_buf_rele(db, FTAG); | |
2200 | db = NULL; | |
2201 | ||
2202 | zgd = umem_zalloc(sizeof (*zgd), UMEM_NOFAIL); | |
1ce23dca | 2203 | zgd->zgd_lwb = lwb; |
5d43cc9a | 2204 | zgd->zgd_private = zd; |
428870ff BB |
2205 | |
2206 | if (buf != NULL) { /* immediate write */ | |
5d43cc9a MA |
2207 | zgd->zgd_lr = (struct locked_range *)ztest_range_lock(zd, |
2208 | object, offset, size, RL_READER); | |
428870ff BB |
2209 | |
2210 | error = dmu_read(os, object, offset, size, buf, | |
2211 | DMU_READ_NO_PREFETCH); | |
2212 | ASSERT(error == 0); | |
2213 | } else { | |
2214 | size = doi.doi_data_block_size; | |
2215 | if (ISP2(size)) { | |
2216 | offset = P2ALIGN(offset, size); | |
2217 | } else { | |
2218 | ASSERT(offset < size); | |
2219 | offset = 0; | |
2220 | } | |
2221 | ||
5d43cc9a MA |
2222 | zgd->zgd_lr = (struct locked_range *)ztest_range_lock(zd, |
2223 | object, offset, size, RL_READER); | |
428870ff BB |
2224 | |
2225 | error = dmu_buf_hold(os, object, offset, zgd, &db, | |
2226 | DMU_READ_NO_PREFETCH); | |
2227 | ||
2228 | if (error == 0) { | |
02dc43bc | 2229 | blkptr_t *bp = &lr->lr_blkptr; |
03c6040b | 2230 | |
428870ff BB |
2231 | zgd->zgd_db = db; |
2232 | zgd->zgd_bp = bp; | |
2233 | ||
2234 | ASSERT(db->db_offset == offset); | |
2235 | ASSERT(db->db_size == size); | |
2236 | ||
2237 | error = dmu_sync(zio, lr->lr_common.lrc_txg, | |
2238 | ztest_get_done, zgd); | |
2239 | ||
2240 | if (error == 0) | |
2241 | return (0); | |
2242 | } | |
2243 | } | |
2244 | ||
2245 | ztest_get_done(zgd, error); | |
2246 | ||
2247 | return (error); | |
2248 | } | |
2249 | ||
2250 | static void * | |
2251 | ztest_lr_alloc(size_t lrsize, char *name) | |
2252 | { | |
2253 | char *lr; | |
2254 | size_t namesize = name ? strlen(name) + 1 : 0; | |
2255 | ||
2256 | lr = umem_zalloc(lrsize + namesize, UMEM_NOFAIL); | |
2257 | ||
2258 | if (name) | |
2259 | bcopy(name, lr + lrsize, namesize); | |
2260 | ||
2261 | return (lr); | |
2262 | } | |
2263 | ||
2264 | void | |
2265 | ztest_lr_free(void *lr, size_t lrsize, char *name) | |
2266 | { | |
2267 | size_t namesize = name ? strlen(name) + 1 : 0; | |
2268 | ||
2269 | umem_free(lr, lrsize + namesize); | |
2270 | } | |
2271 | ||
2272 | /* | |
2273 | * Lookup a bunch of objects. Returns the number of objects not found. | |
2274 | */ | |
2275 | static int | |
2276 | ztest_lookup(ztest_ds_t *zd, ztest_od_t *od, int count) | |
2277 | { | |
2278 | int missing = 0; | |
2279 | int error; | |
d6320ddb | 2280 | int i; |
428870ff | 2281 | |
c25b8f99 | 2282 | ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock)); |
428870ff | 2283 | |
d6320ddb | 2284 | for (i = 0; i < count; i++, od++) { |
428870ff BB |
2285 | od->od_object = 0; |
2286 | error = zap_lookup(zd->zd_os, od->od_dir, od->od_name, | |
2287 | sizeof (uint64_t), 1, &od->od_object); | |
2288 | if (error) { | |
2289 | ASSERT(error == ENOENT); | |
2290 | ASSERT(od->od_object == 0); | |
2291 | missing++; | |
2292 | } else { | |
2293 | dmu_buf_t *db; | |
2294 | ztest_block_tag_t *bbt; | |
2295 | dmu_object_info_t doi; | |
2296 | ||
2297 | ASSERT(od->od_object != 0); | |
2298 | ASSERT(missing == 0); /* there should be no gaps */ | |
2299 | ||
2300 | ztest_object_lock(zd, od->od_object, RL_READER); | |
2301 | VERIFY3U(0, ==, dmu_bonus_hold(zd->zd_os, | |
2302 | od->od_object, FTAG, &db)); | |
2303 | dmu_object_info_from_db(db, &doi); | |
2304 | bbt = ztest_bt_bonus(db); | |
2305 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
2306 | od->od_type = doi.doi_type; | |
2307 | od->od_blocksize = doi.doi_data_block_size; | |
2308 | od->od_gen = bbt->bt_gen; | |
2309 | dmu_buf_rele(db, FTAG); | |
2310 | ztest_object_unlock(zd, od->od_object); | |
2311 | } | |
2312 | } | |
2313 | ||
2314 | return (missing); | |
2315 | } | |
2316 | ||
2317 | static int | |
2318 | ztest_create(ztest_ds_t *zd, ztest_od_t *od, int count) | |
2319 | { | |
2320 | int missing = 0; | |
d6320ddb | 2321 | int i; |
428870ff | 2322 | |
c25b8f99 | 2323 | ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock)); |
428870ff | 2324 | |
d6320ddb | 2325 | for (i = 0; i < count; i++, od++) { |
428870ff BB |
2326 | if (missing) { |
2327 | od->od_object = 0; | |
2328 | missing++; | |
2329 | continue; | |
2330 | } | |
2331 | ||
2332 | lr_create_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name); | |
2333 | ||
2334 | lr->lr_doid = od->od_dir; | |
2335 | lr->lr_foid = 0; /* 0 to allocate, > 0 to claim */ | |
2336 | lr->lrz_type = od->od_crtype; | |
2337 | lr->lrz_blocksize = od->od_crblocksize; | |
2338 | lr->lrz_ibshift = ztest_random_ibshift(); | |
2339 | lr->lrz_bonustype = DMU_OT_UINT64_OTHER; | |
50c957f7 | 2340 | lr->lrz_dnodesize = od->od_crdnodesize; |
428870ff BB |
2341 | lr->lr_gen = od->od_crgen; |
2342 | lr->lr_crtime[0] = time(NULL); | |
2343 | ||
2344 | if (ztest_replay_create(zd, lr, B_FALSE) != 0) { | |
2345 | ASSERT(missing == 0); | |
2346 | od->od_object = 0; | |
2347 | missing++; | |
2348 | } else { | |
2349 | od->od_object = lr->lr_foid; | |
2350 | od->od_type = od->od_crtype; | |
2351 | od->od_blocksize = od->od_crblocksize; | |
2352 | od->od_gen = od->od_crgen; | |
2353 | ASSERT(od->od_object != 0); | |
2354 | } | |
2355 | ||
2356 | ztest_lr_free(lr, sizeof (*lr), od->od_name); | |
2357 | } | |
2358 | ||
2359 | return (missing); | |
2360 | } | |
2361 | ||
2362 | static int | |
2363 | ztest_remove(ztest_ds_t *zd, ztest_od_t *od, int count) | |
2364 | { | |
2365 | int missing = 0; | |
2366 | int error; | |
d6320ddb | 2367 | int i; |
428870ff | 2368 | |
c25b8f99 | 2369 | ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock)); |
428870ff BB |
2370 | |
2371 | od += count - 1; | |
2372 | ||
d6320ddb | 2373 | for (i = count - 1; i >= 0; i--, od--) { |
428870ff BB |
2374 | if (missing) { |
2375 | missing++; | |
2376 | continue; | |
2377 | } | |
2378 | ||
03c6040b GW |
2379 | /* |
2380 | * No object was found. | |
2381 | */ | |
428870ff BB |
2382 | if (od->od_object == 0) |
2383 | continue; | |
2384 | ||
2385 | lr_remove_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name); | |
2386 | ||
2387 | lr->lr_doid = od->od_dir; | |
2388 | ||
2389 | if ((error = ztest_replay_remove(zd, lr, B_FALSE)) != 0) { | |
2390 | ASSERT3U(error, ==, ENOSPC); | |
2391 | missing++; | |
2392 | } else { | |
2393 | od->od_object = 0; | |
2394 | } | |
2395 | ztest_lr_free(lr, sizeof (*lr), od->od_name); | |
2396 | } | |
2397 | ||
2398 | return (missing); | |
2399 | } | |
2400 | ||
2401 | static int | |
2402 | ztest_write(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size, | |
2403 | void *data) | |
2404 | { | |
2405 | lr_write_t *lr; | |
2406 | int error; | |
2407 | ||
2408 | lr = ztest_lr_alloc(sizeof (*lr) + size, NULL); | |
2409 | ||
2410 | lr->lr_foid = object; | |
2411 | lr->lr_offset = offset; | |
2412 | lr->lr_length = size; | |
2413 | lr->lr_blkoff = 0; | |
2414 | BP_ZERO(&lr->lr_blkptr); | |
2415 | ||
2416 | bcopy(data, lr + 1, size); | |
2417 | ||
2418 | error = ztest_replay_write(zd, lr, B_FALSE); | |
2419 | ||
2420 | ztest_lr_free(lr, sizeof (*lr) + size, NULL); | |
2421 | ||
2422 | return (error); | |
2423 | } | |
2424 | ||
2425 | static int | |
2426 | ztest_truncate(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size) | |
2427 | { | |
2428 | lr_truncate_t *lr; | |
2429 | int error; | |
2430 | ||
2431 | lr = ztest_lr_alloc(sizeof (*lr), NULL); | |
2432 | ||
2433 | lr->lr_foid = object; | |
2434 | lr->lr_offset = offset; | |
2435 | lr->lr_length = size; | |
2436 | ||
2437 | error = ztest_replay_truncate(zd, lr, B_FALSE); | |
2438 | ||
2439 | ztest_lr_free(lr, sizeof (*lr), NULL); | |
2440 | ||
2441 | return (error); | |
2442 | } | |
2443 | ||
2444 | static int | |
2445 | ztest_setattr(ztest_ds_t *zd, uint64_t object) | |
2446 | { | |
2447 | lr_setattr_t *lr; | |
2448 | int error; | |
2449 | ||
2450 | lr = ztest_lr_alloc(sizeof (*lr), NULL); | |
2451 | ||
2452 | lr->lr_foid = object; | |
2453 | lr->lr_size = 0; | |
2454 | lr->lr_mode = 0; | |
2455 | ||
2456 | error = ztest_replay_setattr(zd, lr, B_FALSE); | |
2457 | ||
2458 | ztest_lr_free(lr, sizeof (*lr), NULL); | |
2459 | ||
2460 | return (error); | |
2461 | } | |
2462 | ||
2463 | static void | |
2464 | ztest_prealloc(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size) | |
2465 | { | |
2466 | objset_t *os = zd->zd_os; | |
2467 | dmu_tx_t *tx; | |
2468 | uint64_t txg; | |
5d43cc9a | 2469 | rl_t *rl; |
428870ff BB |
2470 | |
2471 | txg_wait_synced(dmu_objset_pool(os), 0); | |
2472 | ||
2473 | ztest_object_lock(zd, object, RL_READER); | |
2474 | rl = ztest_range_lock(zd, object, offset, size, RL_WRITER); | |
2475 | ||
2476 | tx = dmu_tx_create(os); | |
2477 | ||
2478 | dmu_tx_hold_write(tx, object, offset, size); | |
2479 | ||
2480 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
2481 | ||
2482 | if (txg != 0) { | |
2483 | dmu_prealloc(os, object, offset, size, tx); | |
2484 | dmu_tx_commit(tx); | |
2485 | txg_wait_synced(dmu_objset_pool(os), txg); | |
2486 | } else { | |
2487 | (void) dmu_free_long_range(os, object, offset, size); | |
2488 | } | |
2489 | ||
5d43cc9a | 2490 | ztest_range_unlock(rl); |
428870ff BB |
2491 | ztest_object_unlock(zd, object); |
2492 | } | |
2493 | ||
2494 | static void | |
2495 | ztest_io(ztest_ds_t *zd, uint64_t object, uint64_t offset) | |
2496 | { | |
03c6040b | 2497 | int err; |
428870ff BB |
2498 | ztest_block_tag_t wbt; |
2499 | dmu_object_info_t doi; | |
2500 | enum ztest_io_type io_type; | |
2501 | uint64_t blocksize; | |
2502 | void *data; | |
2503 | ||
2504 | VERIFY(dmu_object_info(zd->zd_os, object, &doi) == 0); | |
2505 | blocksize = doi.doi_data_block_size; | |
2506 | data = umem_alloc(blocksize, UMEM_NOFAIL); | |
2507 | ||
2508 | /* | |
2509 | * Pick an i/o type at random, biased toward writing block tags. | |
2510 | */ | |
2511 | io_type = ztest_random(ZTEST_IO_TYPES); | |
2512 | if (ztest_random(2) == 0) | |
2513 | io_type = ZTEST_IO_WRITE_TAG; | |
2514 | ||
2d9142c9 | 2515 | (void) pthread_rwlock_rdlock(&zd->zd_zilog_lock); |
3e31d2b0 | 2516 | |
428870ff BB |
2517 | switch (io_type) { |
2518 | ||
2519 | case ZTEST_IO_WRITE_TAG: | |
50c957f7 NB |
2520 | ztest_bt_generate(&wbt, zd->zd_os, object, doi.doi_dnodesize, |
2521 | offset, 0, 0, 0); | |
428870ff BB |
2522 | (void) ztest_write(zd, object, offset, sizeof (wbt), &wbt); |
2523 | break; | |
2524 | ||
2525 | case ZTEST_IO_WRITE_PATTERN: | |
2526 | (void) memset(data, 'a' + (object + offset) % 5, blocksize); | |
2527 | if (ztest_random(2) == 0) { | |
2528 | /* | |
2529 | * Induce fletcher2 collisions to ensure that | |
2530 | * zio_ddt_collision() detects and resolves them | |
2531 | * when using fletcher2-verify for deduplication. | |
2532 | */ | |
2533 | ((uint64_t *)data)[0] ^= 1ULL << 63; | |
2534 | ((uint64_t *)data)[4] ^= 1ULL << 63; | |
2535 | } | |
2536 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2537 | break; | |
2538 | ||
2539 | case ZTEST_IO_WRITE_ZEROES: | |
2540 | bzero(data, blocksize); | |
2541 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2542 | break; | |
2543 | ||
2544 | case ZTEST_IO_TRUNCATE: | |
2545 | (void) ztest_truncate(zd, object, offset, blocksize); | |
2546 | break; | |
2547 | ||
2548 | case ZTEST_IO_SETATTR: | |
2549 | (void) ztest_setattr(zd, object); | |
2550 | break; | |
e75c13c3 BB |
2551 | default: |
2552 | break; | |
03c6040b GW |
2553 | |
2554 | case ZTEST_IO_REWRITE: | |
2d9142c9 | 2555 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
03c6040b GW |
2556 | err = ztest_dsl_prop_set_uint64(zd->zd_name, |
2557 | ZFS_PROP_CHECKSUM, spa_dedup_checksum(ztest_spa), | |
2558 | B_FALSE); | |
2559 | VERIFY(err == 0 || err == ENOSPC); | |
2560 | err = ztest_dsl_prop_set_uint64(zd->zd_name, | |
2561 | ZFS_PROP_COMPRESSION, | |
2562 | ztest_random_dsl_prop(ZFS_PROP_COMPRESSION), | |
2563 | B_FALSE); | |
2564 | VERIFY(err == 0 || err == ENOSPC); | |
2d9142c9 | 2565 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
03c6040b GW |
2566 | |
2567 | VERIFY0(dmu_read(zd->zd_os, object, offset, blocksize, data, | |
2568 | DMU_READ_NO_PREFETCH)); | |
2569 | ||
2570 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2571 | break; | |
428870ff BB |
2572 | } |
2573 | ||
2d9142c9 | 2574 | (void) pthread_rwlock_unlock(&zd->zd_zilog_lock); |
3e31d2b0 | 2575 | |
428870ff BB |
2576 | umem_free(data, blocksize); |
2577 | } | |
2578 | ||
2579 | /* | |
2580 | * Initialize an object description template. | |
2581 | */ | |
2582 | static void | |
2583 | ztest_od_init(ztest_od_t *od, uint64_t id, char *tag, uint64_t index, | |
50c957f7 NB |
2584 | dmu_object_type_t type, uint64_t blocksize, uint64_t dnodesize, |
2585 | uint64_t gen) | |
428870ff BB |
2586 | { |
2587 | od->od_dir = ZTEST_DIROBJ; | |
2588 | od->od_object = 0; | |
2589 | ||
2590 | od->od_crtype = type; | |
2591 | od->od_crblocksize = blocksize ? blocksize : ztest_random_blocksize(); | |
50c957f7 | 2592 | od->od_crdnodesize = dnodesize ? dnodesize : ztest_random_dnodesize(); |
428870ff BB |
2593 | od->od_crgen = gen; |
2594 | ||
2595 | od->od_type = DMU_OT_NONE; | |
2596 | od->od_blocksize = 0; | |
2597 | od->od_gen = 0; | |
2598 | ||
2599 | (void) snprintf(od->od_name, sizeof (od->od_name), "%s(%lld)[%llu]", | |
b8864a23 | 2600 | tag, (longlong_t)id, (u_longlong_t)index); |
428870ff BB |
2601 | } |
2602 | ||
2603 | /* | |
2604 | * Lookup or create the objects for a test using the od template. | |
2605 | * If the objects do not all exist, or if 'remove' is specified, | |
2606 | * remove any existing objects and create new ones. Otherwise, | |
2607 | * use the existing objects. | |
2608 | */ | |
2609 | static int | |
2610 | ztest_object_init(ztest_ds_t *zd, ztest_od_t *od, size_t size, boolean_t remove) | |
2611 | { | |
2612 | int count = size / sizeof (*od); | |
2613 | int rv = 0; | |
2614 | ||
1e33ac1e | 2615 | mutex_enter(&zd->zd_dirobj_lock); |
428870ff BB |
2616 | if ((ztest_lookup(zd, od, count) != 0 || remove) && |
2617 | (ztest_remove(zd, od, count) != 0 || | |
2618 | ztest_create(zd, od, count) != 0)) | |
2619 | rv = -1; | |
2620 | zd->zd_od = od; | |
1e33ac1e | 2621 | mutex_exit(&zd->zd_dirobj_lock); |
428870ff BB |
2622 | |
2623 | return (rv); | |
2624 | } | |
2625 | ||
2626 | /* ARGSUSED */ | |
2627 | void | |
2628 | ztest_zil_commit(ztest_ds_t *zd, uint64_t id) | |
2629 | { | |
2630 | zilog_t *zilog = zd->zd_zilog; | |
2631 | ||
2d9142c9 | 2632 | (void) pthread_rwlock_rdlock(&zd->zd_zilog_lock); |
3e31d2b0 | 2633 | |
572e2857 | 2634 | zil_commit(zilog, ztest_random(ZTEST_OBJECTS)); |
428870ff BB |
2635 | |
2636 | /* | |
2637 | * Remember the committed values in zd, which is in parent/child | |
2638 | * shared memory. If we die, the next iteration of ztest_run() | |
2639 | * will verify that the log really does contain this record. | |
2640 | */ | |
2641 | mutex_enter(&zilog->zl_lock); | |
c242c188 CS |
2642 | ASSERT(zd->zd_shared != NULL); |
2643 | ASSERT3U(zd->zd_shared->zd_seq, <=, zilog->zl_commit_lr_seq); | |
2644 | zd->zd_shared->zd_seq = zilog->zl_commit_lr_seq; | |
428870ff | 2645 | mutex_exit(&zilog->zl_lock); |
3e31d2b0 | 2646 | |
2d9142c9 | 2647 | (void) pthread_rwlock_unlock(&zd->zd_zilog_lock); |
3e31d2b0 ES |
2648 | } |
2649 | ||
2650 | /* | |
2651 | * This function is designed to simulate the operations that occur during a | |
2652 | * mount/unmount operation. We hold the dataset across these operations in an | |
2653 | * attempt to expose any implicit assumptions about ZIL management. | |
2654 | */ | |
2655 | /* ARGSUSED */ | |
2656 | void | |
2657 | ztest_zil_remount(ztest_ds_t *zd, uint64_t id) | |
2658 | { | |
2659 | objset_t *os = zd->zd_os; | |
2660 | ||
03c6040b GW |
2661 | /* |
2662 | * We grab the zd_dirobj_lock to ensure that no other thread is | |
2663 | * updating the zil (i.e. adding in-memory log records) and the | |
2664 | * zd_zilog_lock to block any I/O. | |
2665 | */ | |
29809a6c | 2666 | mutex_enter(&zd->zd_dirobj_lock); |
2d9142c9 | 2667 | (void) pthread_rwlock_wrlock(&zd->zd_zilog_lock); |
3e31d2b0 | 2668 | |
f298b24d | 2669 | /* zfsvfs_teardown() */ |
3e31d2b0 ES |
2670 | zil_close(zd->zd_zilog); |
2671 | ||
2672 | /* zfsvfs_setup() */ | |
2673 | VERIFY(zil_open(os, ztest_get_data) == zd->zd_zilog); | |
2674 | zil_replay(os, zd, ztest_replay_vector); | |
2675 | ||
2d9142c9 | 2676 | (void) pthread_rwlock_unlock(&zd->zd_zilog_lock); |
29809a6c | 2677 | mutex_exit(&zd->zd_dirobj_lock); |
428870ff BB |
2678 | } |
2679 | ||
2680 | /* | |
2681 | * Verify that we can't destroy an active pool, create an existing pool, | |
2682 | * or create a pool with a bad vdev spec. | |
2683 | */ | |
2684 | /* ARGSUSED */ | |
2685 | void | |
2686 | ztest_spa_create_destroy(ztest_ds_t *zd, uint64_t id) | |
2687 | { | |
c242c188 | 2688 | ztest_shared_opts_t *zo = &ztest_opts; |
428870ff BB |
2689 | spa_t *spa; |
2690 | nvlist_t *nvroot; | |
2691 | ||
379ca9cf OF |
2692 | if (zo->zo_mmp_test) |
2693 | return; | |
2694 | ||
428870ff BB |
2695 | /* |
2696 | * Attempt to create using a bad file. | |
2697 | */ | |
cc99f275 | 2698 | nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, NULL, 0, 0, 1); |
428870ff | 2699 | VERIFY3U(ENOENT, ==, |
b5256303 | 2700 | spa_create("ztest_bad_file", nvroot, NULL, NULL, NULL)); |
428870ff BB |
2701 | nvlist_free(nvroot); |
2702 | ||
2703 | /* | |
2704 | * Attempt to create using a bad mirror. | |
2705 | */ | |
cc99f275 | 2706 | nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, NULL, 0, 2, 1); |
428870ff | 2707 | VERIFY3U(ENOENT, ==, |
b5256303 | 2708 | spa_create("ztest_bad_mirror", nvroot, NULL, NULL, NULL)); |
428870ff BB |
2709 | nvlist_free(nvroot); |
2710 | ||
2711 | /* | |
2712 | * Attempt to create an existing pool. It shouldn't matter | |
2713 | * what's in the nvroot; we should fail with EEXIST. | |
2714 | */ | |
2d9142c9 | 2715 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
cc99f275 | 2716 | nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, NULL, 0, 0, 1); |
b5256303 TC |
2717 | VERIFY3U(EEXIST, ==, |
2718 | spa_create(zo->zo_pool, nvroot, NULL, NULL, NULL)); | |
428870ff | 2719 | nvlist_free(nvroot); |
c242c188 CS |
2720 | VERIFY3U(0, ==, spa_open(zo->zo_pool, &spa, FTAG)); |
2721 | VERIFY3U(EBUSY, ==, spa_destroy(zo->zo_pool)); | |
428870ff BB |
2722 | spa_close(spa, FTAG); |
2723 | ||
2d9142c9 | 2724 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
428870ff BB |
2725 | } |
2726 | ||
0582e403 OF |
2727 | /* |
2728 | * Start and then stop the MMP threads to ensure the startup and shutdown code | |
2729 | * works properly. Actual protection and property-related code tested via ZTS. | |
2730 | */ | |
2731 | /* ARGSUSED */ | |
2732 | void | |
2733 | ztest_mmp_enable_disable(ztest_ds_t *zd, uint64_t id) | |
2734 | { | |
2735 | ztest_shared_opts_t *zo = &ztest_opts; | |
2736 | spa_t *spa = ztest_spa; | |
2737 | ||
2738 | if (zo->zo_mmp_test) | |
2739 | return; | |
2740 | ||
2741 | spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); | |
2742 | mutex_enter(&spa->spa_props_lock); | |
2743 | ||
52002377 OF |
2744 | zfs_multihost_fail_intervals = 0; |
2745 | ||
0582e403 OF |
2746 | if (!spa_multihost(spa)) { |
2747 | spa->spa_multihost = B_TRUE; | |
2748 | mmp_thread_start(spa); | |
2749 | } | |
2750 | ||
2751 | mutex_exit(&spa->spa_props_lock); | |
2752 | spa_config_exit(spa, SCL_CONFIG, FTAG); | |
2753 | ||
2754 | txg_wait_synced(spa_get_dsl(spa), 0); | |
2755 | mmp_signal_all_threads(); | |
2756 | txg_wait_synced(spa_get_dsl(spa), 0); | |
2757 | ||
2758 | spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); | |
2759 | mutex_enter(&spa->spa_props_lock); | |
2760 | ||
2761 | if (spa_multihost(spa)) { | |
2762 | mmp_thread_stop(spa); | |
2763 | spa->spa_multihost = B_FALSE; | |
2764 | } | |
2765 | ||
2766 | mutex_exit(&spa->spa_props_lock); | |
2767 | spa_config_exit(spa, SCL_CONFIG, FTAG); | |
2768 | } | |
2769 | ||
ea0b2538 GW |
2770 | /* ARGSUSED */ |
2771 | void | |
2772 | ztest_spa_upgrade(ztest_ds_t *zd, uint64_t id) | |
2773 | { | |
2774 | spa_t *spa; | |
2775 | uint64_t initial_version = SPA_VERSION_INITIAL; | |
2776 | uint64_t version, newversion; | |
2777 | nvlist_t *nvroot, *props; | |
2778 | char *name; | |
2779 | ||
379ca9cf OF |
2780 | if (ztest_opts.zo_mmp_test) |
2781 | return; | |
2782 | ||
ea0b2538 GW |
2783 | mutex_enter(&ztest_vdev_lock); |
2784 | name = kmem_asprintf("%s_upgrade", ztest_opts.zo_pool); | |
2785 | ||
2786 | /* | |
2787 | * Clean up from previous runs. | |
2788 | */ | |
2789 | (void) spa_destroy(name); | |
2790 | ||
2791 | nvroot = make_vdev_root(NULL, NULL, name, ztest_opts.zo_vdev_size, 0, | |
cc99f275 | 2792 | NULL, ztest_opts.zo_raidz, ztest_opts.zo_mirrors, 1); |
ea0b2538 GW |
2793 | |
2794 | /* | |
2795 | * If we're configuring a RAIDZ device then make sure that the | |
2796 | * the initial version is capable of supporting that feature. | |
2797 | */ | |
2798 | switch (ztest_opts.zo_raidz_parity) { | |
2799 | case 0: | |
2800 | case 1: | |
2801 | initial_version = SPA_VERSION_INITIAL; | |
2802 | break; | |
2803 | case 2: | |
2804 | initial_version = SPA_VERSION_RAIDZ2; | |
2805 | break; | |
2806 | case 3: | |
2807 | initial_version = SPA_VERSION_RAIDZ3; | |
2808 | break; | |
2809 | } | |
2810 | ||
2811 | /* | |
2812 | * Create a pool with a spa version that can be upgraded. Pick | |
2813 | * a value between initial_version and SPA_VERSION_BEFORE_FEATURES. | |
2814 | */ | |
2815 | do { | |
2816 | version = ztest_random_spa_version(initial_version); | |
2817 | } while (version > SPA_VERSION_BEFORE_FEATURES); | |
2818 | ||
2819 | props = fnvlist_alloc(); | |
2820 | fnvlist_add_uint64(props, | |
2821 | zpool_prop_to_name(ZPOOL_PROP_VERSION), version); | |
b5256303 | 2822 | VERIFY3S(spa_create(name, nvroot, props, NULL, NULL), ==, 0); |
ea0b2538 GW |
2823 | fnvlist_free(nvroot); |
2824 | fnvlist_free(props); | |
2825 | ||
2826 | VERIFY3S(spa_open(name, &spa, FTAG), ==, 0); | |
2827 | VERIFY3U(spa_version(spa), ==, version); | |
2828 | newversion = ztest_random_spa_version(version + 1); | |
2829 | ||
2830 | if (ztest_opts.zo_verbose >= 4) { | |
2831 | (void) printf("upgrading spa version from %llu to %llu\n", | |
2832 | (u_longlong_t)version, (u_longlong_t)newversion); | |
2833 | } | |
2834 | ||
2835 | spa_upgrade(spa, newversion); | |
2836 | VERIFY3U(spa_version(spa), >, version); | |
2837 | VERIFY3U(spa_version(spa), ==, fnvlist_lookup_uint64(spa->spa_config, | |
2838 | zpool_prop_to_name(ZPOOL_PROP_VERSION))); | |
2839 | spa_close(spa, FTAG); | |
2840 | ||
2841 | strfree(name); | |
2842 | mutex_exit(&ztest_vdev_lock); | |
2843 | } | |
2844 | ||
d2734cce SD |
2845 | static void |
2846 | ztest_spa_checkpoint(spa_t *spa) | |
2847 | { | |
2848 | ASSERT(MUTEX_HELD(&ztest_checkpoint_lock)); | |
2849 | ||
2850 | int error = spa_checkpoint(spa->spa_name); | |
2851 | ||
2852 | switch (error) { | |
2853 | case 0: | |
2854 | case ZFS_ERR_DEVRM_IN_PROGRESS: | |
2855 | case ZFS_ERR_DISCARDING_CHECKPOINT: | |
2856 | case ZFS_ERR_CHECKPOINT_EXISTS: | |
2857 | break; | |
2858 | case ENOSPC: | |
2859 | ztest_record_enospc(FTAG); | |
2860 | break; | |
2861 | default: | |
2862 | fatal(0, "spa_checkpoint(%s) = %d", spa->spa_name, error); | |
2863 | } | |
2864 | } | |
2865 | ||
2866 | static void | |
2867 | ztest_spa_discard_checkpoint(spa_t *spa) | |
2868 | { | |
2869 | ASSERT(MUTEX_HELD(&ztest_checkpoint_lock)); | |
2870 | ||
2871 | int error = spa_checkpoint_discard(spa->spa_name); | |
2872 | ||
2873 | switch (error) { | |
2874 | case 0: | |
2875 | case ZFS_ERR_DISCARDING_CHECKPOINT: | |
2876 | case ZFS_ERR_NO_CHECKPOINT: | |
2877 | break; | |
2878 | default: | |
2879 | fatal(0, "spa_discard_checkpoint(%s) = %d", | |
2880 | spa->spa_name, error); | |
2881 | } | |
2882 | ||
2883 | } | |
2884 | ||
2885 | /* ARGSUSED */ | |
2886 | void | |
2887 | ztest_spa_checkpoint_create_discard(ztest_ds_t *zd, uint64_t id) | |
2888 | { | |
2889 | spa_t *spa = ztest_spa; | |
2890 | ||
2891 | mutex_enter(&ztest_checkpoint_lock); | |
2892 | if (ztest_random(2) == 0) { | |
2893 | ztest_spa_checkpoint(spa); | |
2894 | } else { | |
2895 | ztest_spa_discard_checkpoint(spa); | |
2896 | } | |
2897 | mutex_exit(&ztest_checkpoint_lock); | |
2898 | } | |
2899 | ||
2900 | ||
428870ff BB |
2901 | static vdev_t * |
2902 | vdev_lookup_by_path(vdev_t *vd, const char *path) | |
2903 | { | |
2904 | vdev_t *mvd; | |
d6320ddb | 2905 | int c; |
428870ff BB |
2906 | |
2907 | if (vd->vdev_path != NULL && strcmp(path, vd->vdev_path) == 0) | |
2908 | return (vd); | |
2909 | ||
d6320ddb | 2910 | for (c = 0; c < vd->vdev_children; c++) |
428870ff BB |
2911 | if ((mvd = vdev_lookup_by_path(vd->vdev_child[c], path)) != |
2912 | NULL) | |
2913 | return (mvd); | |
2914 | ||
2915 | return (NULL); | |
2916 | } | |
2917 | ||
2918 | /* | |
2919 | * Find the first available hole which can be used as a top-level. | |
2920 | */ | |
2921 | int | |
2922 | find_vdev_hole(spa_t *spa) | |
2923 | { | |
2924 | vdev_t *rvd = spa->spa_root_vdev; | |
2925 | int c; | |
2926 | ||
2927 | ASSERT(spa_config_held(spa, SCL_VDEV, RW_READER) == SCL_VDEV); | |
2928 | ||
2929 | for (c = 0; c < rvd->vdev_children; c++) { | |
2930 | vdev_t *cvd = rvd->vdev_child[c]; | |
2931 | ||
2932 | if (cvd->vdev_ishole) | |
2933 | break; | |
2934 | } | |
2935 | return (c); | |
2936 | } | |
2937 | ||
2938 | /* | |
2939 | * Verify that vdev_add() works as expected. | |
2940 | */ | |
2941 | /* ARGSUSED */ | |
2942 | void | |
2943 | ztest_vdev_add_remove(ztest_ds_t *zd, uint64_t id) | |
2944 | { | |
2945 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 2946 | spa_t *spa = ztest_spa; |
428870ff BB |
2947 | uint64_t leaves; |
2948 | uint64_t guid; | |
2949 | nvlist_t *nvroot; | |
2950 | int error; | |
2951 | ||
379ca9cf OF |
2952 | if (ztest_opts.zo_mmp_test) |
2953 | return; | |
2954 | ||
c242c188 | 2955 | mutex_enter(&ztest_vdev_lock); |
13fe0198 | 2956 | leaves = MAX(zs->zs_mirrors + zs->zs_splits, 1) * ztest_opts.zo_raidz; |
428870ff BB |
2957 | |
2958 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
2959 | ||
2960 | ztest_shared->zs_vdev_next_leaf = find_vdev_hole(spa) * leaves; | |
2961 | ||
2962 | /* | |
2963 | * If we have slogs then remove them 1/4 of the time. | |
2964 | */ | |
2965 | if (spa_has_slogs(spa) && ztest_random(4) == 0) { | |
cc99f275 DB |
2966 | metaslab_group_t *mg; |
2967 | ||
428870ff | 2968 | /* |
cc99f275 | 2969 | * find the first real slog in log allocation class |
428870ff | 2970 | */ |
cc99f275 DB |
2971 | mg = spa_log_class(spa)->mc_rotor; |
2972 | while (!mg->mg_vd->vdev_islog) | |
2973 | mg = mg->mg_next; | |
2974 | ||
2975 | guid = mg->mg_vd->vdev_guid; | |
428870ff BB |
2976 | |
2977 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
2978 | ||
2979 | /* | |
2980 | * We have to grab the zs_name_lock as writer to | |
2981 | * prevent a race between removing a slog (dmu_objset_find) | |
2982 | * and destroying a dataset. Removing the slog will | |
2983 | * grab a reference on the dataset which may cause | |
13fe0198 | 2984 | * dsl_destroy_head() to fail with EBUSY thus |
428870ff BB |
2985 | * leaving the dataset in an inconsistent state. |
2986 | */ | |
2d9142c9 | 2987 | pthread_rwlock_wrlock(&ztest_name_lock); |
428870ff | 2988 | error = spa_vdev_remove(spa, guid, B_FALSE); |
2d9142c9 | 2989 | pthread_rwlock_unlock(&ztest_name_lock); |
428870ff | 2990 | |
c91cf36f BB |
2991 | switch (error) { |
2992 | case 0: | |
2993 | case EEXIST: /* Generic zil_reset() error */ | |
2994 | case EBUSY: /* Replay required */ | |
2995 | case EACCES: /* Crypto key not loaded */ | |
d2734cce SD |
2996 | case ZFS_ERR_CHECKPOINT_EXISTS: |
2997 | case ZFS_ERR_DISCARDING_CHECKPOINT: | |
c91cf36f BB |
2998 | break; |
2999 | default: | |
428870ff | 3000 | fatal(0, "spa_vdev_remove() = %d", error); |
c91cf36f | 3001 | } |
428870ff BB |
3002 | } else { |
3003 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
3004 | ||
3005 | /* | |
cc99f275 | 3006 | * Make 1/4 of the devices be log devices |
428870ff | 3007 | */ |
ea0b2538 | 3008 | nvroot = make_vdev_root(NULL, NULL, NULL, |
cc99f275 DB |
3009 | ztest_opts.zo_vdev_size, 0, (ztest_random(4) == 0) ? |
3010 | "log" : NULL, ztest_opts.zo_raidz, zs->zs_mirrors, 1); | |
428870ff BB |
3011 | |
3012 | error = spa_vdev_add(spa, nvroot); | |
3013 | nvlist_free(nvroot); | |
3014 | ||
d2734cce SD |
3015 | switch (error) { |
3016 | case 0: | |
3017 | break; | |
3018 | case ENOSPC: | |
428870ff | 3019 | ztest_record_enospc("spa_vdev_add"); |
d2734cce SD |
3020 | break; |
3021 | default: | |
428870ff | 3022 | fatal(0, "spa_vdev_add() = %d", error); |
d2734cce | 3023 | } |
428870ff BB |
3024 | } |
3025 | ||
c242c188 | 3026 | mutex_exit(&ztest_vdev_lock); |
428870ff BB |
3027 | } |
3028 | ||
cc99f275 DB |
3029 | /* ARGSUSED */ |
3030 | void | |
3031 | ztest_vdev_class_add(ztest_ds_t *zd, uint64_t id) | |
3032 | { | |
3033 | ztest_shared_t *zs = ztest_shared; | |
3034 | spa_t *spa = ztest_spa; | |
3035 | uint64_t leaves; | |
3036 | nvlist_t *nvroot; | |
3037 | const char *class = (ztest_random(2) == 0) ? | |
3038 | VDEV_ALLOC_BIAS_SPECIAL : VDEV_ALLOC_BIAS_DEDUP; | |
3039 | int error; | |
3040 | ||
3041 | /* | |
3042 | * By default add a special vdev 50% of the time | |
3043 | */ | |
3044 | if ((ztest_opts.zo_special_vdevs == ZTEST_VDEV_CLASS_OFF) || | |
3045 | (ztest_opts.zo_special_vdevs == ZTEST_VDEV_CLASS_RND && | |
3046 | ztest_random(2) == 0)) { | |
3047 | return; | |
3048 | } | |
3049 | ||
3050 | mutex_enter(&ztest_vdev_lock); | |
3051 | ||
3052 | /* Only test with mirrors */ | |
3053 | if (zs->zs_mirrors < 2) { | |
3054 | mutex_exit(&ztest_vdev_lock); | |
3055 | return; | |
3056 | } | |
3057 | ||
3058 | /* requires feature@allocation_classes */ | |
3059 | if (!spa_feature_is_enabled(spa, SPA_FEATURE_ALLOCATION_CLASSES)) { | |
3060 | mutex_exit(&ztest_vdev_lock); | |
3061 | return; | |
3062 | } | |
3063 | ||
3064 | leaves = MAX(zs->zs_mirrors + zs->zs_splits, 1) * ztest_opts.zo_raidz; | |
3065 | ||
3066 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
3067 | ztest_shared->zs_vdev_next_leaf = find_vdev_hole(spa) * leaves; | |
3068 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
3069 | ||
3070 | nvroot = make_vdev_root(NULL, NULL, NULL, ztest_opts.zo_vdev_size, 0, | |
3071 | class, ztest_opts.zo_raidz, zs->zs_mirrors, 1); | |
3072 | ||
3073 | error = spa_vdev_add(spa, nvroot); | |
3074 | nvlist_free(nvroot); | |
3075 | ||
3076 | if (error == ENOSPC) | |
3077 | ztest_record_enospc("spa_vdev_add"); | |
3078 | else if (error != 0) | |
3079 | fatal(0, "spa_vdev_add() = %d", error); | |
3080 | ||
3081 | /* | |
3082 | * 50% of the time allow small blocks in the special class | |
3083 | */ | |
3084 | if (error == 0 && | |
3085 | spa_special_class(spa)->mc_groups == 1 && ztest_random(2) == 0) { | |
3086 | if (ztest_opts.zo_verbose >= 3) | |
3087 | (void) printf("Enabling special VDEV small blocks\n"); | |
3088 | (void) ztest_dsl_prop_set_uint64(zd->zd_name, | |
3089 | ZFS_PROP_SPECIAL_SMALL_BLOCKS, 32768, B_FALSE); | |
3090 | } | |
3091 | ||
3092 | mutex_exit(&ztest_vdev_lock); | |
3093 | ||
3094 | if (ztest_opts.zo_verbose >= 3) { | |
3095 | metaslab_class_t *mc; | |
3096 | ||
3097 | if (strcmp(class, VDEV_ALLOC_BIAS_SPECIAL) == 0) | |
3098 | mc = spa_special_class(spa); | |
3099 | else | |
3100 | mc = spa_dedup_class(spa); | |
3101 | (void) printf("Added a %s mirrored vdev (of %d)\n", | |
3102 | class, (int)mc->mc_groups); | |
3103 | } | |
3104 | } | |
3105 | ||
428870ff BB |
3106 | /* |
3107 | * Verify that adding/removing aux devices (l2arc, hot spare) works as expected. | |
3108 | */ | |
3109 | /* ARGSUSED */ | |
3110 | void | |
3111 | ztest_vdev_aux_add_remove(ztest_ds_t *zd, uint64_t id) | |
3112 | { | |
3113 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 3114 | spa_t *spa = ztest_spa; |
428870ff BB |
3115 | vdev_t *rvd = spa->spa_root_vdev; |
3116 | spa_aux_vdev_t *sav; | |
3117 | char *aux; | |
40b84e7a | 3118 | char *path; |
428870ff BB |
3119 | uint64_t guid = 0; |
3120 | int error; | |
3121 | ||
379ca9cf OF |
3122 | if (ztest_opts.zo_mmp_test) |
3123 | return; | |
3124 | ||
40b84e7a BB |
3125 | path = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
3126 | ||
428870ff BB |
3127 | if (ztest_random(2) == 0) { |
3128 | sav = &spa->spa_spares; | |
3129 | aux = ZPOOL_CONFIG_SPARES; | |
3130 | } else { | |
3131 | sav = &spa->spa_l2cache; | |
3132 | aux = ZPOOL_CONFIG_L2CACHE; | |
3133 | } | |
3134 | ||
c242c188 | 3135 | mutex_enter(&ztest_vdev_lock); |
428870ff BB |
3136 | |
3137 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
3138 | ||
3139 | if (sav->sav_count != 0 && ztest_random(4) == 0) { | |
b128c09f BB |
3140 | /* |
3141 | * Pick a random device to remove. | |
3142 | */ | |
3143 | guid = sav->sav_vdevs[ztest_random(sav->sav_count)]->vdev_guid; | |
3144 | } else { | |
3145 | /* | |
3146 | * Find an unused device we can add. | |
3147 | */ | |
428870ff | 3148 | zs->zs_vdev_aux = 0; |
b128c09f | 3149 | for (;;) { |
b128c09f | 3150 | int c; |
3db3ff4a | 3151 | (void) snprintf(path, MAXPATHLEN, ztest_aux_template, |
c242c188 CS |
3152 | ztest_opts.zo_dir, ztest_opts.zo_pool, aux, |
3153 | zs->zs_vdev_aux); | |
b128c09f BB |
3154 | for (c = 0; c < sav->sav_count; c++) |
3155 | if (strcmp(sav->sav_vdevs[c]->vdev_path, | |
3156 | path) == 0) | |
3157 | break; | |
3158 | if (c == sav->sav_count && | |
3159 | vdev_lookup_by_path(rvd, path) == NULL) | |
3160 | break; | |
428870ff | 3161 | zs->zs_vdev_aux++; |
34dc7c2f BB |
3162 | } |
3163 | } | |
3164 | ||
b128c09f | 3165 | spa_config_exit(spa, SCL_VDEV, FTAG); |
34dc7c2f | 3166 | |
b128c09f BB |
3167 | if (guid == 0) { |
3168 | /* | |
3169 | * Add a new device. | |
3170 | */ | |
ea0b2538 | 3171 | nvlist_t *nvroot = make_vdev_root(NULL, aux, NULL, |
cc99f275 | 3172 | (ztest_opts.zo_vdev_size * 5) / 4, 0, NULL, 0, 0, 1); |
b128c09f | 3173 | error = spa_vdev_add(spa, nvroot); |
d2734cce SD |
3174 | |
3175 | switch (error) { | |
3176 | case 0: | |
3177 | break; | |
3178 | default: | |
b128c09f | 3179 | fatal(0, "spa_vdev_add(%p) = %d", nvroot, error); |
d2734cce | 3180 | } |
b128c09f BB |
3181 | nvlist_free(nvroot); |
3182 | } else { | |
3183 | /* | |
3184 | * Remove an existing device. Sometimes, dirty its | |
3185 | * vdev state first to make sure we handle removal | |
3186 | * of devices that have pending state changes. | |
3187 | */ | |
3188 | if (ztest_random(2) == 0) | |
9babb374 | 3189 | (void) vdev_online(spa, guid, 0, NULL); |
b128c09f BB |
3190 | |
3191 | error = spa_vdev_remove(spa, guid, B_FALSE); | |
d2734cce SD |
3192 | |
3193 | switch (error) { | |
3194 | case 0: | |
3195 | case EBUSY: | |
3196 | case ZFS_ERR_CHECKPOINT_EXISTS: | |
3197 | case ZFS_ERR_DISCARDING_CHECKPOINT: | |
3198 | break; | |
3199 | default: | |
b128c09f | 3200 | fatal(0, "spa_vdev_remove(%llu) = %d", guid, error); |
d2734cce | 3201 | } |
b128c09f BB |
3202 | } |
3203 | ||
c242c188 | 3204 | mutex_exit(&ztest_vdev_lock); |
40b84e7a BB |
3205 | |
3206 | umem_free(path, MAXPATHLEN); | |
428870ff BB |
3207 | } |
3208 | ||
3209 | /* | |
3210 | * split a pool if it has mirror tlvdevs | |
3211 | */ | |
3212 | /* ARGSUSED */ | |
3213 | void | |
3214 | ztest_split_pool(ztest_ds_t *zd, uint64_t id) | |
3215 | { | |
3216 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 3217 | spa_t *spa = ztest_spa; |
428870ff BB |
3218 | vdev_t *rvd = spa->spa_root_vdev; |
3219 | nvlist_t *tree, **child, *config, *split, **schild; | |
3220 | uint_t c, children, schildren = 0, lastlogid = 0; | |
3221 | int error = 0; | |
3222 | ||
379ca9cf OF |
3223 | if (ztest_opts.zo_mmp_test) |
3224 | return; | |
3225 | ||
c242c188 | 3226 | mutex_enter(&ztest_vdev_lock); |
428870ff BB |
3227 | |
3228 | /* ensure we have a useable config; mirrors of raidz aren't supported */ | |
c242c188 CS |
3229 | if (zs->zs_mirrors < 3 || ztest_opts.zo_raidz > 1) { |
3230 | mutex_exit(&ztest_vdev_lock); | |
428870ff BB |
3231 | return; |
3232 | } | |
3233 | ||
3234 | /* clean up the old pool, if any */ | |
3235 | (void) spa_destroy("splitp"); | |
3236 | ||
3237 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
3238 | ||
3239 | /* generate a config from the existing config */ | |
3240 | mutex_enter(&spa->spa_props_lock); | |
3241 | VERIFY(nvlist_lookup_nvlist(spa->spa_config, ZPOOL_CONFIG_VDEV_TREE, | |
3242 | &tree) == 0); | |
3243 | mutex_exit(&spa->spa_props_lock); | |
3244 | ||
3245 | VERIFY(nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child, | |
3246 | &children) == 0); | |
3247 | ||
3248 | schild = malloc(rvd->vdev_children * sizeof (nvlist_t *)); | |
3249 | for (c = 0; c < children; c++) { | |
3250 | vdev_t *tvd = rvd->vdev_child[c]; | |
3251 | nvlist_t **mchild; | |
3252 | uint_t mchildren; | |
3253 | ||
3254 | if (tvd->vdev_islog || tvd->vdev_ops == &vdev_hole_ops) { | |
3255 | VERIFY(nvlist_alloc(&schild[schildren], NV_UNIQUE_NAME, | |
3256 | 0) == 0); | |
3257 | VERIFY(nvlist_add_string(schild[schildren], | |
3258 | ZPOOL_CONFIG_TYPE, VDEV_TYPE_HOLE) == 0); | |
3259 | VERIFY(nvlist_add_uint64(schild[schildren], | |
3260 | ZPOOL_CONFIG_IS_HOLE, 1) == 0); | |
3261 | if (lastlogid == 0) | |
3262 | lastlogid = schildren; | |
3263 | ++schildren; | |
3264 | continue; | |
3265 | } | |
3266 | lastlogid = 0; | |
3267 | VERIFY(nvlist_lookup_nvlist_array(child[c], | |
3268 | ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0); | |
3269 | VERIFY(nvlist_dup(mchild[0], &schild[schildren++], 0) == 0); | |
3270 | } | |
3271 | ||
3272 | /* OK, create a config that can be used to split */ | |
3273 | VERIFY(nvlist_alloc(&split, NV_UNIQUE_NAME, 0) == 0); | |
3274 | VERIFY(nvlist_add_string(split, ZPOOL_CONFIG_TYPE, | |
3275 | VDEV_TYPE_ROOT) == 0); | |
3276 | VERIFY(nvlist_add_nvlist_array(split, ZPOOL_CONFIG_CHILDREN, schild, | |
3277 | lastlogid != 0 ? lastlogid : schildren) == 0); | |
3278 | ||
3279 | VERIFY(nvlist_alloc(&config, NV_UNIQUE_NAME, 0) == 0); | |
3280 | VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, split) == 0); | |
3281 | ||
3282 | for (c = 0; c < schildren; c++) | |
3283 | nvlist_free(schild[c]); | |
3284 | free(schild); | |
3285 | nvlist_free(split); | |
3286 | ||
3287 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
3288 | ||
2d9142c9 | 3289 | (void) pthread_rwlock_wrlock(&ztest_name_lock); |
428870ff | 3290 | error = spa_vdev_split_mirror(spa, "splitp", config, NULL, B_FALSE); |
2d9142c9 | 3291 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
428870ff BB |
3292 | |
3293 | nvlist_free(config); | |
3294 | ||
3295 | if (error == 0) { | |
3296 | (void) printf("successful split - results:\n"); | |
3297 | mutex_enter(&spa_namespace_lock); | |
3298 | show_pool_stats(spa); | |
3299 | show_pool_stats(spa_lookup("splitp")); | |
3300 | mutex_exit(&spa_namespace_lock); | |
3301 | ++zs->zs_splits; | |
3302 | --zs->zs_mirrors; | |
3303 | } | |
c242c188 | 3304 | mutex_exit(&ztest_vdev_lock); |
34dc7c2f BB |
3305 | } |
3306 | ||
3307 | /* | |
3308 | * Verify that we can attach and detach devices. | |
3309 | */ | |
428870ff | 3310 | /* ARGSUSED */ |
34dc7c2f | 3311 | void |
428870ff | 3312 | ztest_vdev_attach_detach(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3313 | { |
428870ff | 3314 | ztest_shared_t *zs = ztest_shared; |
c242c188 | 3315 | spa_t *spa = ztest_spa; |
b128c09f | 3316 | spa_aux_vdev_t *sav = &spa->spa_spares; |
34dc7c2f BB |
3317 | vdev_t *rvd = spa->spa_root_vdev; |
3318 | vdev_t *oldvd, *newvd, *pvd; | |
b128c09f | 3319 | nvlist_t *root; |
428870ff | 3320 | uint64_t leaves; |
34dc7c2f BB |
3321 | uint64_t leaf, top; |
3322 | uint64_t ashift = ztest_get_ashift(); | |
fb5f0bc8 | 3323 | uint64_t oldguid, pguid; |
5d1f7fb6 | 3324 | uint64_t oldsize, newsize; |
40b84e7a | 3325 | char *oldpath, *newpath; |
34dc7c2f | 3326 | int replacing; |
b128c09f BB |
3327 | int oldvd_has_siblings = B_FALSE; |
3328 | int newvd_is_spare = B_FALSE; | |
3329 | int oldvd_is_log; | |
34dc7c2f | 3330 | int error, expected_error; |
34dc7c2f | 3331 | |
379ca9cf OF |
3332 | if (ztest_opts.zo_mmp_test) |
3333 | return; | |
3334 | ||
40b84e7a BB |
3335 | oldpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
3336 | newpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
3337 | ||
c242c188 CS |
3338 | mutex_enter(&ztest_vdev_lock); |
3339 | leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raidz; | |
34dc7c2f | 3340 | |
a1d477c2 MA |
3341 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); |
3342 | ||
3343 | /* | |
3344 | * If a vdev is in the process of being removed, its removal may | |
3345 | * finish while we are in progress, leading to an unexpected error | |
3346 | * value. Don't bother trying to attach while we are in the middle | |
3347 | * of removal. | |
3348 | */ | |
9e052db4 | 3349 | if (ztest_device_removal_active) { |
a1d477c2 MA |
3350 | spa_config_exit(spa, SCL_ALL, FTAG); |
3351 | mutex_exit(&ztest_vdev_lock); | |
3352 | return; | |
3353 | } | |
34dc7c2f BB |
3354 | |
3355 | /* | |
3356 | * Decide whether to do an attach or a replace. | |
3357 | */ | |
3358 | replacing = ztest_random(2); | |
3359 | ||
3360 | /* | |
3361 | * Pick a random top-level vdev. | |
3362 | */ | |
428870ff | 3363 | top = ztest_random_vdev_top(spa, B_TRUE); |
34dc7c2f BB |
3364 | |
3365 | /* | |
3366 | * Pick a random leaf within it. | |
3367 | */ | |
3368 | leaf = ztest_random(leaves); | |
3369 | ||
3370 | /* | |
b128c09f | 3371 | * Locate this vdev. |
34dc7c2f | 3372 | */ |
b128c09f | 3373 | oldvd = rvd->vdev_child[top]; |
cc99f275 DB |
3374 | |
3375 | /* pick a child from the mirror */ | |
428870ff | 3376 | if (zs->zs_mirrors >= 1) { |
fb5f0bc8 | 3377 | ASSERT(oldvd->vdev_ops == &vdev_mirror_ops); |
428870ff | 3378 | ASSERT(oldvd->vdev_children >= zs->zs_mirrors); |
c242c188 | 3379 | oldvd = oldvd->vdev_child[leaf / ztest_opts.zo_raidz]; |
fb5f0bc8 | 3380 | } |
cc99f275 DB |
3381 | |
3382 | /* pick a child out of the raidz group */ | |
c242c188 | 3383 | if (ztest_opts.zo_raidz > 1) { |
fb5f0bc8 | 3384 | ASSERT(oldvd->vdev_ops == &vdev_raidz_ops); |
c242c188 CS |
3385 | ASSERT(oldvd->vdev_children == ztest_opts.zo_raidz); |
3386 | oldvd = oldvd->vdev_child[leaf % ztest_opts.zo_raidz]; | |
fb5f0bc8 | 3387 | } |
34dc7c2f BB |
3388 | |
3389 | /* | |
b128c09f BB |
3390 | * If we're already doing an attach or replace, oldvd may be a |
3391 | * mirror vdev -- in which case, pick a random child. | |
34dc7c2f | 3392 | */ |
b128c09f BB |
3393 | while (oldvd->vdev_children != 0) { |
3394 | oldvd_has_siblings = B_TRUE; | |
fb5f0bc8 BB |
3395 | ASSERT(oldvd->vdev_children >= 2); |
3396 | oldvd = oldvd->vdev_child[ztest_random(oldvd->vdev_children)]; | |
b128c09f BB |
3397 | } |
3398 | ||
3399 | oldguid = oldvd->vdev_guid; | |
9babb374 | 3400 | oldsize = vdev_get_min_asize(oldvd); |
b128c09f BB |
3401 | oldvd_is_log = oldvd->vdev_top->vdev_islog; |
3402 | (void) strcpy(oldpath, oldvd->vdev_path); | |
3403 | pvd = oldvd->vdev_parent; | |
fb5f0bc8 | 3404 | pguid = pvd->vdev_guid; |
34dc7c2f BB |
3405 | |
3406 | /* | |
b128c09f | 3407 | * If oldvd has siblings, then half of the time, detach it. |
34dc7c2f | 3408 | */ |
b128c09f | 3409 | if (oldvd_has_siblings && ztest_random(2) == 0) { |
a1d477c2 | 3410 | spa_config_exit(spa, SCL_ALL, FTAG); |
fb5f0bc8 BB |
3411 | error = spa_vdev_detach(spa, oldguid, pguid, B_FALSE); |
3412 | if (error != 0 && error != ENODEV && error != EBUSY && | |
d2734cce SD |
3413 | error != ENOTSUP && error != ZFS_ERR_CHECKPOINT_EXISTS && |
3414 | error != ZFS_ERR_DISCARDING_CHECKPOINT) | |
fb5f0bc8 | 3415 | fatal(0, "detach (%s) returned %d", oldpath, error); |
40b84e7a | 3416 | goto out; |
b128c09f | 3417 | } |
34dc7c2f BB |
3418 | |
3419 | /* | |
b128c09f BB |
3420 | * For the new vdev, choose with equal probability between the two |
3421 | * standard paths (ending in either 'a' or 'b') or a random hot spare. | |
34dc7c2f | 3422 | */ |
b128c09f BB |
3423 | if (sav->sav_count != 0 && ztest_random(3) == 0) { |
3424 | newvd = sav->sav_vdevs[ztest_random(sav->sav_count)]; | |
3425 | newvd_is_spare = B_TRUE; | |
3426 | (void) strcpy(newpath, newvd->vdev_path); | |
3427 | } else { | |
6aec1cd5 | 3428 | (void) snprintf(newpath, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
3429 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
3430 | top * leaves + leaf); | |
b128c09f BB |
3431 | if (ztest_random(2) == 0) |
3432 | newpath[strlen(newpath) - 1] = 'b'; | |
3433 | newvd = vdev_lookup_by_path(rvd, newpath); | |
3434 | } | |
3435 | ||
3436 | if (newvd) { | |
a1d477c2 MA |
3437 | /* |
3438 | * Reopen to ensure the vdev's asize field isn't stale. | |
3439 | */ | |
3440 | vdev_reopen(newvd); | |
9babb374 | 3441 | newsize = vdev_get_min_asize(newvd); |
b128c09f BB |
3442 | } else { |
3443 | /* | |
3444 | * Make newsize a little bigger or smaller than oldsize. | |
3445 | * If it's smaller, the attach should fail. | |
3446 | * If it's larger, and we're doing a replace, | |
3447 | * we should get dynamic LUN growth when we're done. | |
3448 | */ | |
3449 | newsize = 10 * oldsize / (9 + ztest_random(3)); | |
3450 | } | |
34dc7c2f BB |
3451 | |
3452 | /* | |
3453 | * If pvd is not a mirror or root, the attach should fail with ENOTSUP, | |
3454 | * unless it's a replace; in that case any non-replacing parent is OK. | |
3455 | * | |
3456 | * If newvd is already part of the pool, it should fail with EBUSY. | |
3457 | * | |
3458 | * If newvd is too small, it should fail with EOVERFLOW. | |
3459 | */ | |
b128c09f BB |
3460 | if (pvd->vdev_ops != &vdev_mirror_ops && |
3461 | pvd->vdev_ops != &vdev_root_ops && (!replacing || | |
3462 | pvd->vdev_ops == &vdev_replacing_ops || | |
3463 | pvd->vdev_ops == &vdev_spare_ops)) | |
34dc7c2f | 3464 | expected_error = ENOTSUP; |
b128c09f BB |
3465 | else if (newvd_is_spare && (!replacing || oldvd_is_log)) |
3466 | expected_error = ENOTSUP; | |
3467 | else if (newvd == oldvd) | |
3468 | expected_error = replacing ? 0 : EBUSY; | |
3469 | else if (vdev_lookup_by_path(rvd, newpath) != NULL) | |
3470 | expected_error = EBUSY; | |
34dc7c2f BB |
3471 | else if (newsize < oldsize) |
3472 | expected_error = EOVERFLOW; | |
3473 | else if (ashift > oldvd->vdev_top->vdev_ashift) | |
3474 | expected_error = EDOM; | |
3475 | else | |
3476 | expected_error = 0; | |
3477 | ||
a1d477c2 | 3478 | spa_config_exit(spa, SCL_ALL, FTAG); |
34dc7c2f BB |
3479 | |
3480 | /* | |
3481 | * Build the nvlist describing newpath. | |
3482 | */ | |
ea0b2538 | 3483 | root = make_vdev_root(newpath, NULL, NULL, newvd == NULL ? newsize : 0, |
cc99f275 | 3484 | ashift, NULL, 0, 0, 1); |
34dc7c2f | 3485 | |
b128c09f | 3486 | error = spa_vdev_attach(spa, oldguid, root, replacing); |
34dc7c2f | 3487 | |
34dc7c2f BB |
3488 | nvlist_free(root); |
3489 | ||
3490 | /* | |
3491 | * If our parent was the replacing vdev, but the replace completed, | |
3492 | * then instead of failing with ENOTSUP we may either succeed, | |
3493 | * fail with ENODEV, or fail with EOVERFLOW. | |
3494 | */ | |
3495 | if (expected_error == ENOTSUP && | |
3496 | (error == 0 || error == ENODEV || error == EOVERFLOW)) | |
3497 | expected_error = error; | |
3498 | ||
3499 | /* | |
3500 | * If someone grew the LUN, the replacement may be too small. | |
3501 | */ | |
b128c09f | 3502 | if (error == EOVERFLOW || error == EBUSY) |
34dc7c2f BB |
3503 | expected_error = error; |
3504 | ||
d2734cce SD |
3505 | if (error == ZFS_ERR_CHECKPOINT_EXISTS || |
3506 | error == ZFS_ERR_DISCARDING_CHECKPOINT) | |
3507 | expected_error = error; | |
3508 | ||
b128c09f BB |
3509 | /* XXX workaround 6690467 */ |
3510 | if (error != expected_error && expected_error != EBUSY) { | |
3511 | fatal(0, "attach (%s %llu, %s %llu, %d) " | |
3512 | "returned %d, expected %d", | |
5d1f7fb6 GW |
3513 | oldpath, oldsize, newpath, |
3514 | newsize, replacing, error, expected_error); | |
34dc7c2f | 3515 | } |
40b84e7a | 3516 | out: |
c242c188 | 3517 | mutex_exit(&ztest_vdev_lock); |
40b84e7a BB |
3518 | |
3519 | umem_free(oldpath, MAXPATHLEN); | |
3520 | umem_free(newpath, MAXPATHLEN); | |
34dc7c2f BB |
3521 | } |
3522 | ||
a1d477c2 MA |
3523 | /* ARGSUSED */ |
3524 | void | |
3525 | ztest_device_removal(ztest_ds_t *zd, uint64_t id) | |
3526 | { | |
3527 | spa_t *spa = ztest_spa; | |
3528 | vdev_t *vd; | |
3529 | uint64_t guid; | |
9e052db4 | 3530 | int error; |
a1d477c2 MA |
3531 | |
3532 | mutex_enter(&ztest_vdev_lock); | |
3533 | ||
9e052db4 MA |
3534 | if (ztest_device_removal_active) { |
3535 | mutex_exit(&ztest_vdev_lock); | |
3536 | return; | |
3537 | } | |
3538 | ||
3539 | /* | |
3540 | * Remove a random top-level vdev and wait for removal to finish. | |
3541 | */ | |
a1d477c2 MA |
3542 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); |
3543 | vd = vdev_lookup_top(spa, ztest_random_vdev_top(spa, B_FALSE)); | |
3544 | guid = vd->vdev_guid; | |
3545 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
3546 | ||
9e052db4 MA |
3547 | error = spa_vdev_remove(spa, guid, B_FALSE); |
3548 | if (error == 0) { | |
3549 | ztest_device_removal_active = B_TRUE; | |
3550 | mutex_exit(&ztest_vdev_lock); | |
3551 | ||
3552 | while (spa->spa_vdev_removal != NULL) | |
3553 | txg_wait_synced(spa_get_dsl(spa), 0); | |
3554 | } else { | |
3555 | mutex_exit(&ztest_vdev_lock); | |
3556 | return; | |
3557 | } | |
a1d477c2 | 3558 | |
9e052db4 MA |
3559 | /* |
3560 | * The pool needs to be scrubbed after completing device removal. | |
3561 | * Failure to do so may result in checksum errors due to the | |
3562 | * strategy employed by ztest_fault_inject() when selecting which | |
3563 | * offset are redundant and can be damaged. | |
3564 | */ | |
3565 | error = spa_scan(spa, POOL_SCAN_SCRUB); | |
3566 | if (error == 0) { | |
3567 | while (dsl_scan_scrubbing(spa_get_dsl(spa))) | |
3568 | txg_wait_synced(spa_get_dsl(spa), 0); | |
3569 | } | |
3570 | ||
3571 | mutex_enter(&ztest_vdev_lock); | |
3572 | ztest_device_removal_active = B_FALSE; | |
a1d477c2 MA |
3573 | mutex_exit(&ztest_vdev_lock); |
3574 | } | |
3575 | ||
9babb374 BB |
3576 | /* |
3577 | * Callback function which expands the physical size of the vdev. | |
3578 | */ | |
3579 | vdev_t * | |
3580 | grow_vdev(vdev_t *vd, void *arg) | |
3581 | { | |
1fde1e37 | 3582 | ASSERTV(spa_t *spa = vd->vdev_spa); |
9babb374 BB |
3583 | size_t *newsize = arg; |
3584 | size_t fsize; | |
3585 | int fd; | |
3586 | ||
3587 | ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE); | |
3588 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
3589 | ||
3590 | if ((fd = open(vd->vdev_path, O_RDWR)) == -1) | |
3591 | return (vd); | |
3592 | ||
3593 | fsize = lseek(fd, 0, SEEK_END); | |
0e5b68e0 | 3594 | VERIFY(ftruncate(fd, *newsize) == 0); |
9babb374 | 3595 | |
c242c188 | 3596 | if (ztest_opts.zo_verbose >= 6) { |
9babb374 BB |
3597 | (void) printf("%s grew from %lu to %lu bytes\n", |
3598 | vd->vdev_path, (ulong_t)fsize, (ulong_t)*newsize); | |
3599 | } | |
3600 | (void) close(fd); | |
3601 | return (NULL); | |
3602 | } | |
3603 | ||
3604 | /* | |
3605 | * Callback function which expands a given vdev by calling vdev_online(). | |
3606 | */ | |
3607 | /* ARGSUSED */ | |
3608 | vdev_t * | |
3609 | online_vdev(vdev_t *vd, void *arg) | |
3610 | { | |
3611 | spa_t *spa = vd->vdev_spa; | |
3612 | vdev_t *tvd = vd->vdev_top; | |
9babb374 | 3613 | uint64_t guid = vd->vdev_guid; |
428870ff BB |
3614 | uint64_t generation = spa->spa_config_generation + 1; |
3615 | vdev_state_t newstate = VDEV_STATE_UNKNOWN; | |
3616 | int error; | |
9babb374 BB |
3617 | |
3618 | ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE); | |
3619 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
3620 | ||
3621 | /* Calling vdev_online will initialize the new metaslabs */ | |
3622 | spa_config_exit(spa, SCL_STATE, spa); | |
428870ff | 3623 | error = vdev_online(spa, guid, ZFS_ONLINE_EXPAND, &newstate); |
9babb374 BB |
3624 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); |
3625 | ||
428870ff BB |
3626 | /* |
3627 | * If vdev_online returned an error or the underlying vdev_open | |
3628 | * failed then we abort the expand. The only way to know that | |
3629 | * vdev_open fails is by checking the returned newstate. | |
3630 | */ | |
3631 | if (error || newstate != VDEV_STATE_HEALTHY) { | |
c242c188 | 3632 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
3633 | (void) printf("Unable to expand vdev, state %llu, " |
3634 | "error %d\n", (u_longlong_t)newstate, error); | |
3635 | } | |
3636 | return (vd); | |
3637 | } | |
3638 | ASSERT3U(newstate, ==, VDEV_STATE_HEALTHY); | |
3639 | ||
9babb374 BB |
3640 | /* |
3641 | * Since we dropped the lock we need to ensure that we're | |
3642 | * still talking to the original vdev. It's possible this | |
3643 | * vdev may have been detached/replaced while we were | |
3644 | * trying to online it. | |
3645 | */ | |
428870ff | 3646 | if (generation != spa->spa_config_generation) { |
c242c188 | 3647 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
3648 | (void) printf("vdev configuration has changed, " |
3649 | "guid %llu, state %llu, expected gen %llu, " | |
3650 | "got gen %llu\n", | |
3651 | (u_longlong_t)guid, | |
3652 | (u_longlong_t)tvd->vdev_state, | |
3653 | (u_longlong_t)generation, | |
3654 | (u_longlong_t)spa->spa_config_generation); | |
9babb374 BB |
3655 | } |
3656 | return (vd); | |
3657 | } | |
3658 | return (NULL); | |
3659 | } | |
3660 | ||
3661 | /* | |
3662 | * Traverse the vdev tree calling the supplied function. | |
3663 | * We continue to walk the tree until we either have walked all | |
3664 | * children or we receive a non-NULL return from the callback. | |
3665 | * If a NULL callback is passed, then we just return back the first | |
3666 | * leaf vdev we encounter. | |
3667 | */ | |
3668 | vdev_t * | |
3669 | vdev_walk_tree(vdev_t *vd, vdev_t *(*func)(vdev_t *, void *), void *arg) | |
3670 | { | |
d6320ddb BB |
3671 | uint_t c; |
3672 | ||
9babb374 BB |
3673 | if (vd->vdev_ops->vdev_op_leaf) { |
3674 | if (func == NULL) | |
3675 | return (vd); | |
3676 | else | |
3677 | return (func(vd, arg)); | |
3678 | } | |
3679 | ||
d6320ddb | 3680 | for (c = 0; c < vd->vdev_children; c++) { |
9babb374 BB |
3681 | vdev_t *cvd = vd->vdev_child[c]; |
3682 | if ((cvd = vdev_walk_tree(cvd, func, arg)) != NULL) | |
3683 | return (cvd); | |
3684 | } | |
3685 | return (NULL); | |
3686 | } | |
3687 | ||
34dc7c2f BB |
3688 | /* |
3689 | * Verify that dynamic LUN growth works as expected. | |
3690 | */ | |
428870ff | 3691 | /* ARGSUSED */ |
34dc7c2f | 3692 | void |
428870ff | 3693 | ztest_vdev_LUN_growth(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3694 | { |
c242c188 | 3695 | spa_t *spa = ztest_spa; |
428870ff BB |
3696 | vdev_t *vd, *tvd; |
3697 | metaslab_class_t *mc; | |
3698 | metaslab_group_t *mg; | |
9babb374 | 3699 | size_t psize, newsize; |
428870ff BB |
3700 | uint64_t top; |
3701 | uint64_t old_class_space, new_class_space, old_ms_count, new_ms_count; | |
34dc7c2f | 3702 | |
d2734cce | 3703 | mutex_enter(&ztest_checkpoint_lock); |
c242c188 | 3704 | mutex_enter(&ztest_vdev_lock); |
9babb374 BB |
3705 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); |
3706 | ||
a1d477c2 MA |
3707 | /* |
3708 | * If there is a vdev removal in progress, it could complete while | |
3709 | * we are running, in which case we would not be able to verify | |
3710 | * that the metaslab_class space increased (because it decreases | |
3711 | * when the device removal completes). | |
3712 | */ | |
9e052db4 | 3713 | if (ztest_device_removal_active) { |
d2734cce | 3714 | spa_config_exit(spa, SCL_STATE, spa); |
a1d477c2 | 3715 | mutex_exit(&ztest_vdev_lock); |
d2734cce | 3716 | mutex_exit(&ztest_checkpoint_lock); |
a1d477c2 MA |
3717 | return; |
3718 | } | |
3719 | ||
428870ff | 3720 | top = ztest_random_vdev_top(spa, B_TRUE); |
9babb374 | 3721 | |
428870ff BB |
3722 | tvd = spa->spa_root_vdev->vdev_child[top]; |
3723 | mg = tvd->vdev_mg; | |
3724 | mc = mg->mg_class; | |
3725 | old_ms_count = tvd->vdev_ms_count; | |
3726 | old_class_space = metaslab_class_get_space(mc); | |
34dc7c2f BB |
3727 | |
3728 | /* | |
9babb374 BB |
3729 | * Determine the size of the first leaf vdev associated with |
3730 | * our top-level device. | |
34dc7c2f | 3731 | */ |
9babb374 BB |
3732 | vd = vdev_walk_tree(tvd, NULL, NULL); |
3733 | ASSERT3P(vd, !=, NULL); | |
3734 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
34dc7c2f | 3735 | |
9babb374 | 3736 | psize = vd->vdev_psize; |
34dc7c2f | 3737 | |
9babb374 | 3738 | /* |
428870ff BB |
3739 | * We only try to expand the vdev if it's healthy, less than 4x its |
3740 | * original size, and it has a valid psize. | |
9babb374 | 3741 | */ |
428870ff | 3742 | if (tvd->vdev_state != VDEV_STATE_HEALTHY || |
c242c188 | 3743 | psize == 0 || psize >= 4 * ztest_opts.zo_vdev_size) { |
9babb374 | 3744 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 3745 | mutex_exit(&ztest_vdev_lock); |
d2734cce | 3746 | mutex_exit(&ztest_checkpoint_lock); |
9babb374 BB |
3747 | return; |
3748 | } | |
3749 | ASSERT(psize > 0); | |
cc99f275 | 3750 | newsize = psize + MAX(psize / 8, SPA_MAXBLOCKSIZE); |
9babb374 | 3751 | ASSERT3U(newsize, >, psize); |
34dc7c2f | 3752 | |
c242c188 | 3753 | if (ztest_opts.zo_verbose >= 6) { |
428870ff | 3754 | (void) printf("Expanding LUN %s from %lu to %lu\n", |
9babb374 BB |
3755 | vd->vdev_path, (ulong_t)psize, (ulong_t)newsize); |
3756 | } | |
3757 | ||
9babb374 BB |
3758 | /* |
3759 | * Growing the vdev is a two step process: | |
3760 | * 1). expand the physical size (i.e. relabel) | |
3761 | * 2). online the vdev to create the new metaslabs | |
3762 | */ | |
3763 | if (vdev_walk_tree(tvd, grow_vdev, &newsize) != NULL || | |
3764 | vdev_walk_tree(tvd, online_vdev, NULL) != NULL || | |
3765 | tvd->vdev_state != VDEV_STATE_HEALTHY) { | |
c242c188 | 3766 | if (ztest_opts.zo_verbose >= 5) { |
9babb374 | 3767 | (void) printf("Could not expand LUN because " |
428870ff | 3768 | "the vdev configuration changed.\n"); |
34dc7c2f | 3769 | } |
428870ff | 3770 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 3771 | mutex_exit(&ztest_vdev_lock); |
d2734cce | 3772 | mutex_exit(&ztest_checkpoint_lock); |
9babb374 | 3773 | return; |
34dc7c2f BB |
3774 | } |
3775 | ||
428870ff | 3776 | spa_config_exit(spa, SCL_STATE, spa); |
9babb374 BB |
3777 | |
3778 | /* | |
3779 | * Expanding the LUN will update the config asynchronously, | |
3780 | * thus we must wait for the async thread to complete any | |
3781 | * pending tasks before proceeding. | |
3782 | */ | |
428870ff BB |
3783 | for (;;) { |
3784 | boolean_t done; | |
3785 | mutex_enter(&spa->spa_async_lock); | |
3786 | done = (spa->spa_async_thread == NULL && !spa->spa_async_tasks); | |
3787 | mutex_exit(&spa->spa_async_lock); | |
3788 | if (done) | |
3789 | break; | |
3790 | txg_wait_synced(spa_get_dsl(spa), 0); | |
3791 | (void) poll(NULL, 0, 100); | |
3792 | } | |
9babb374 BB |
3793 | |
3794 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); | |
428870ff BB |
3795 | |
3796 | tvd = spa->spa_root_vdev->vdev_child[top]; | |
3797 | new_ms_count = tvd->vdev_ms_count; | |
3798 | new_class_space = metaslab_class_get_space(mc); | |
3799 | ||
3800 | if (tvd->vdev_mg != mg || mg->mg_class != mc) { | |
c242c188 | 3801 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
3802 | (void) printf("Could not verify LUN expansion due to " |
3803 | "intervening vdev offline or remove.\n"); | |
3804 | } | |
3805 | spa_config_exit(spa, SCL_STATE, spa); | |
c242c188 | 3806 | mutex_exit(&ztest_vdev_lock); |
d2734cce | 3807 | mutex_exit(&ztest_checkpoint_lock); |
428870ff BB |
3808 | return; |
3809 | } | |
3810 | ||
3811 | /* | |
3812 | * Make sure we were able to grow the vdev. | |
3813 | */ | |
a1d477c2 MA |
3814 | if (new_ms_count <= old_ms_count) { |
3815 | fatal(0, "LUN expansion failed: ms_count %llu < %llu\n", | |
428870ff | 3816 | old_ms_count, new_ms_count); |
a1d477c2 | 3817 | } |
9babb374 BB |
3818 | |
3819 | /* | |
3820 | * Make sure we were able to grow the pool. | |
3821 | */ | |
a1d477c2 MA |
3822 | if (new_class_space <= old_class_space) { |
3823 | fatal(0, "LUN expansion failed: class_space %llu < %llu\n", | |
428870ff | 3824 | old_class_space, new_class_space); |
a1d477c2 | 3825 | } |
428870ff | 3826 | |
c242c188 | 3827 | if (ztest_opts.zo_verbose >= 5) { |
f3c8c9e6 | 3828 | char oldnumbuf[NN_NUMBUF_SZ], newnumbuf[NN_NUMBUF_SZ]; |
9babb374 | 3829 | |
f3c8c9e6 JK |
3830 | nicenum(old_class_space, oldnumbuf, sizeof (oldnumbuf)); |
3831 | nicenum(new_class_space, newnumbuf, sizeof (newnumbuf)); | |
9babb374 BB |
3832 | (void) printf("%s grew from %s to %s\n", |
3833 | spa->spa_name, oldnumbuf, newnumbuf); | |
3834 | } | |
428870ff | 3835 | |
9babb374 | 3836 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 3837 | mutex_exit(&ztest_vdev_lock); |
d2734cce | 3838 | mutex_exit(&ztest_checkpoint_lock); |
34dc7c2f BB |
3839 | } |
3840 | ||
428870ff BB |
3841 | /* |
3842 | * Verify that dmu_objset_{create,destroy,open,close} work as expected. | |
3843 | */ | |
34dc7c2f BB |
3844 | /* ARGSUSED */ |
3845 | static void | |
428870ff | 3846 | ztest_objset_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) |
34dc7c2f BB |
3847 | { |
3848 | /* | |
428870ff | 3849 | * Create the objects common to all ztest datasets. |
34dc7c2f | 3850 | */ |
428870ff | 3851 | VERIFY(zap_create_claim(os, ZTEST_DIROBJ, |
34dc7c2f | 3852 | DMU_OT_ZAP_OTHER, DMU_OT_NONE, 0, tx) == 0); |
428870ff | 3853 | } |
34dc7c2f | 3854 | |
428870ff BB |
3855 | static int |
3856 | ztest_dataset_create(char *dsname) | |
3857 | { | |
4807c0ba TC |
3858 | int err; |
3859 | uint64_t rand; | |
3860 | dsl_crypto_params_t *dcp = NULL; | |
3861 | ||
3862 | /* | |
3863 | * 50% of the time, we create encrypted datasets | |
3864 | * using a random cipher suite and a hard-coded | |
3865 | * wrapping key. | |
3866 | */ | |
3867 | rand = ztest_random(2); | |
3868 | if (rand != 0) { | |
3869 | nvlist_t *crypto_args = fnvlist_alloc(); | |
3870 | nvlist_t *props = fnvlist_alloc(); | |
3871 | ||
3872 | /* slight bias towards the default cipher suite */ | |
3873 | rand = ztest_random(ZIO_CRYPT_FUNCTIONS); | |
3874 | if (rand < ZIO_CRYPT_AES_128_CCM) | |
3875 | rand = ZIO_CRYPT_ON; | |
3876 | ||
3877 | fnvlist_add_uint64(props, | |
3878 | zfs_prop_to_name(ZFS_PROP_ENCRYPTION), rand); | |
3879 | fnvlist_add_uint8_array(crypto_args, "wkeydata", | |
3880 | (uint8_t *)ztest_wkeydata, WRAPPING_KEY_LEN); | |
3881 | ||
3882 | /* | |
3883 | * These parameters aren't really used by the kernel. They | |
3884 | * are simply stored so that userspace knows how to load | |
3885 | * the wrapping key. | |
3886 | */ | |
3887 | fnvlist_add_uint64(props, | |
3888 | zfs_prop_to_name(ZFS_PROP_KEYFORMAT), ZFS_KEYFORMAT_RAW); | |
3889 | fnvlist_add_string(props, | |
3890 | zfs_prop_to_name(ZFS_PROP_KEYLOCATION), "prompt"); | |
3891 | fnvlist_add_uint64(props, | |
3892 | zfs_prop_to_name(ZFS_PROP_PBKDF2_SALT), 0ULL); | |
3893 | fnvlist_add_uint64(props, | |
3894 | zfs_prop_to_name(ZFS_PROP_PBKDF2_ITERS), 0ULL); | |
3895 | ||
3896 | VERIFY0(dsl_crypto_params_create_nvlist(DCP_CMD_NONE, props, | |
3897 | crypto_args, &dcp)); | |
3898 | ||
010d1247 NL |
3899 | /* |
3900 | * Cycle through all available encryption implementations | |
3901 | * to verify interoperability. | |
3902 | */ | |
3903 | VERIFY0(gcm_impl_set("cycle")); | |
3904 | VERIFY0(aes_impl_set("cycle")); | |
3905 | ||
4807c0ba TC |
3906 | fnvlist_free(crypto_args); |
3907 | fnvlist_free(props); | |
3908 | } | |
3909 | ||
3910 | err = dmu_objset_create(dsname, DMU_OST_OTHER, 0, dcp, | |
428870ff | 3911 | ztest_objset_create_cb, NULL); |
4807c0ba | 3912 | dsl_crypto_params_free(dcp, !!err); |
428870ff | 3913 | |
4807c0ba TC |
3914 | rand = ztest_random(100); |
3915 | if (err || rand < 80) | |
428870ff BB |
3916 | return (err); |
3917 | ||
c242c188 | 3918 | if (ztest_opts.zo_verbose >= 5) |
b815ff9a | 3919 | (void) printf("Setting dataset %s to sync always\n", dsname); |
428870ff BB |
3920 | return (ztest_dsl_prop_set_uint64(dsname, ZFS_PROP_SYNC, |
3921 | ZFS_SYNC_ALWAYS, B_FALSE)); | |
34dc7c2f BB |
3922 | } |
3923 | ||
428870ff | 3924 | /* ARGSUSED */ |
34dc7c2f | 3925 | static int |
428870ff | 3926 | ztest_objset_destroy_cb(const char *name, void *arg) |
34dc7c2f | 3927 | { |
34dc7c2f | 3928 | objset_t *os; |
428870ff | 3929 | dmu_object_info_t doi; |
34dc7c2f BB |
3930 | int error; |
3931 | ||
3932 | /* | |
3933 | * Verify that the dataset contains a directory object. | |
3934 | */ | |
4807c0ba TC |
3935 | VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_TRUE, |
3936 | B_TRUE, FTAG, &os)); | |
428870ff | 3937 | error = dmu_object_info(os, ZTEST_DIROBJ, &doi); |
34dc7c2f BB |
3938 | if (error != ENOENT) { |
3939 | /* We could have crashed in the middle of destroying it */ | |
c99c9001 | 3940 | ASSERT0(error); |
428870ff BB |
3941 | ASSERT3U(doi.doi_type, ==, DMU_OT_ZAP_OTHER); |
3942 | ASSERT3S(doi.doi_physical_blocks_512, >=, 0); | |
34dc7c2f | 3943 | } |
b5256303 | 3944 | dmu_objset_disown(os, B_TRUE, FTAG); |
34dc7c2f BB |
3945 | |
3946 | /* | |
3947 | * Destroy the dataset. | |
3948 | */ | |
13fe0198 | 3949 | if (strchr(name, '@') != NULL) { |
dc1fbc43 | 3950 | VERIFY0(dsl_destroy_snapshot(name, B_TRUE)); |
13fe0198 | 3951 | } else { |
dc1fbc43 GM |
3952 | error = dsl_destroy_head(name); |
3953 | /* There could be a hold on this dataset */ | |
3954 | if (error != EBUSY) | |
3955 | ASSERT0(error); | |
13fe0198 | 3956 | } |
34dc7c2f BB |
3957 | return (0); |
3958 | } | |
3959 | ||
428870ff BB |
3960 | static boolean_t |
3961 | ztest_snapshot_create(char *osname, uint64_t id) | |
34dc7c2f | 3962 | { |
eca7b760 | 3963 | char snapname[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
3964 | int error; |
3965 | ||
13fe0198 | 3966 | (void) snprintf(snapname, sizeof (snapname), "%llu", (u_longlong_t)id); |
428870ff | 3967 | |
13fe0198 | 3968 | error = dmu_objset_snapshot_one(osname, snapname); |
428870ff BB |
3969 | if (error == ENOSPC) { |
3970 | ztest_record_enospc(FTAG); | |
3971 | return (B_FALSE); | |
3972 | } | |
13fe0198 MA |
3973 | if (error != 0 && error != EEXIST) { |
3974 | fatal(0, "ztest_snapshot_create(%s@%s) = %d", osname, | |
3975 | snapname, error); | |
3976 | } | |
428870ff BB |
3977 | return (B_TRUE); |
3978 | } | |
3979 | ||
3980 | static boolean_t | |
3981 | ztest_snapshot_destroy(char *osname, uint64_t id) | |
3982 | { | |
eca7b760 | 3983 | char snapname[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
3984 | int error; |
3985 | ||
eca7b760 | 3986 | (void) snprintf(snapname, sizeof (snapname), "%s@%llu", osname, |
428870ff BB |
3987 | (u_longlong_t)id); |
3988 | ||
13fe0198 | 3989 | error = dsl_destroy_snapshot(snapname, B_FALSE); |
428870ff BB |
3990 | if (error != 0 && error != ENOENT) |
3991 | fatal(0, "ztest_snapshot_destroy(%s) = %d", snapname, error); | |
3992 | return (B_TRUE); | |
34dc7c2f BB |
3993 | } |
3994 | ||
428870ff | 3995 | /* ARGSUSED */ |
34dc7c2f | 3996 | void |
428870ff | 3997 | ztest_dmu_objset_create_destroy(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3998 | { |
40b84e7a | 3999 | ztest_ds_t *zdtmp; |
428870ff | 4000 | int iters; |
34dc7c2f | 4001 | int error; |
b128c09f | 4002 | objset_t *os, *os2; |
eca7b760 | 4003 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
34dc7c2f | 4004 | zilog_t *zilog; |
d6320ddb | 4005 | int i; |
34dc7c2f | 4006 | |
40b84e7a | 4007 | zdtmp = umem_alloc(sizeof (ztest_ds_t), UMEM_NOFAIL); |
40b84e7a | 4008 | |
2d9142c9 | 4009 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
34dc7c2f | 4010 | |
eca7b760 | 4011 | (void) snprintf(name, sizeof (name), "%s/temp_%llu", |
c242c188 | 4012 | ztest_opts.zo_pool, (u_longlong_t)id); |
34dc7c2f BB |
4013 | |
4014 | /* | |
4015 | * If this dataset exists from a previous run, process its replay log | |
13fe0198 | 4016 | * half of the time. If we don't replay it, then dsl_destroy_head() |
428870ff | 4017 | * (invoked from ztest_objset_destroy_cb()) should just throw it away. |
34dc7c2f BB |
4018 | */ |
4019 | if (ztest_random(2) == 0 && | |
4807c0ba | 4020 | ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, |
b5256303 | 4021 | B_TRUE, FTAG, &os) == 0) { |
c242c188 | 4022 | ztest_zd_init(zdtmp, NULL, os); |
40b84e7a BB |
4023 | zil_replay(os, zdtmp, ztest_replay_vector); |
4024 | ztest_zd_fini(zdtmp); | |
b5256303 | 4025 | dmu_objset_disown(os, B_TRUE, FTAG); |
34dc7c2f BB |
4026 | } |
4027 | ||
4028 | /* | |
4029 | * There may be an old instance of the dataset we're about to | |
4030 | * create lying around from a previous run. If so, destroy it | |
4031 | * and all of its snapshots. | |
4032 | */ | |
428870ff | 4033 | (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL, |
34dc7c2f BB |
4034 | DS_FIND_CHILDREN | DS_FIND_SNAPSHOTS); |
4035 | ||
4036 | /* | |
4037 | * Verify that the destroyed dataset is no longer in the namespace. | |
4038 | */ | |
4807c0ba TC |
4039 | VERIFY3U(ENOENT, ==, ztest_dmu_objset_own(name, DMU_OST_OTHER, B_TRUE, |
4040 | B_TRUE, FTAG, &os)); | |
34dc7c2f BB |
4041 | |
4042 | /* | |
4043 | * Verify that we can create a new dataset. | |
4044 | */ | |
428870ff | 4045 | error = ztest_dataset_create(name); |
34dc7c2f BB |
4046 | if (error) { |
4047 | if (error == ENOSPC) { | |
428870ff | 4048 | ztest_record_enospc(FTAG); |
40b84e7a | 4049 | goto out; |
34dc7c2f BB |
4050 | } |
4051 | fatal(0, "dmu_objset_create(%s) = %d", name, error); | |
4052 | } | |
4053 | ||
4807c0ba | 4054 | VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, B_TRUE, |
b5256303 | 4055 | FTAG, &os)); |
428870ff | 4056 | |
c242c188 | 4057 | ztest_zd_init(zdtmp, NULL, os); |
34dc7c2f BB |
4058 | |
4059 | /* | |
4060 | * Open the intent log for it. | |
4061 | */ | |
428870ff | 4062 | zilog = zil_open(os, ztest_get_data); |
34dc7c2f BB |
4063 | |
4064 | /* | |
428870ff BB |
4065 | * Put some objects in there, do a little I/O to them, |
4066 | * and randomly take a couple of snapshots along the way. | |
34dc7c2f | 4067 | */ |
428870ff | 4068 | iters = ztest_random(5); |
d6320ddb | 4069 | for (i = 0; i < iters; i++) { |
40b84e7a | 4070 | ztest_dmu_object_alloc_free(zdtmp, id); |
428870ff BB |
4071 | if (ztest_random(iters) == 0) |
4072 | (void) ztest_snapshot_create(name, i); | |
34dc7c2f BB |
4073 | } |
4074 | ||
4075 | /* | |
4076 | * Verify that we cannot create an existing dataset. | |
4077 | */ | |
428870ff | 4078 | VERIFY3U(EEXIST, ==, |
b5256303 | 4079 | dmu_objset_create(name, DMU_OST_OTHER, 0, NULL, NULL, NULL)); |
34dc7c2f BB |
4080 | |
4081 | /* | |
428870ff | 4082 | * Verify that we can hold an objset that is also owned. |
b128c09f | 4083 | */ |
428870ff BB |
4084 | VERIFY3U(0, ==, dmu_objset_hold(name, FTAG, &os2)); |
4085 | dmu_objset_rele(os2, FTAG); | |
34dc7c2f | 4086 | |
428870ff BB |
4087 | /* |
4088 | * Verify that we cannot own an objset that is already owned. | |
4089 | */ | |
4807c0ba TC |
4090 | VERIFY3U(EBUSY, ==, ztest_dmu_objset_own(name, DMU_OST_OTHER, |
4091 | B_FALSE, B_TRUE, FTAG, &os2)); | |
34dc7c2f | 4092 | |
428870ff | 4093 | zil_close(zilog); |
b5256303 | 4094 | dmu_objset_disown(os, B_TRUE, FTAG); |
40b84e7a BB |
4095 | ztest_zd_fini(zdtmp); |
4096 | out: | |
2d9142c9 | 4097 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
40b84e7a | 4098 | |
40b84e7a | 4099 | umem_free(zdtmp, sizeof (ztest_ds_t)); |
34dc7c2f BB |
4100 | } |
4101 | ||
4102 | /* | |
4103 | * Verify that dmu_snapshot_{create,destroy,open,close} work as expected. | |
4104 | */ | |
4105 | void | |
428870ff | 4106 | ztest_dmu_snapshot_create_destroy(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4107 | { |
2d9142c9 | 4108 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
428870ff BB |
4109 | (void) ztest_snapshot_destroy(zd->zd_name, id); |
4110 | (void) ztest_snapshot_create(zd->zd_name, id); | |
2d9142c9 | 4111 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
34dc7c2f BB |
4112 | } |
4113 | ||
9babb374 BB |
4114 | /* |
4115 | * Cleanup non-standard snapshots and clones. | |
4116 | */ | |
4117 | void | |
428870ff | 4118 | ztest_dsl_dataset_cleanup(char *osname, uint64_t id) |
9babb374 | 4119 | { |
40b84e7a BB |
4120 | char *snap1name; |
4121 | char *clone1name; | |
4122 | char *snap2name; | |
4123 | char *clone2name; | |
4124 | char *snap3name; | |
9babb374 BB |
4125 | int error; |
4126 | ||
eca7b760 IK |
4127 | snap1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); |
4128 | clone1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4129 | snap2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4130 | clone2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4131 | snap3name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4132 | ||
4133 | (void) snprintf(snap1name, ZFS_MAX_DATASET_NAME_LEN, | |
4134 | "%s@s1_%llu", osname, (u_longlong_t)id); | |
4135 | (void) snprintf(clone1name, ZFS_MAX_DATASET_NAME_LEN, | |
4136 | "%s/c1_%llu", osname, (u_longlong_t)id); | |
4137 | (void) snprintf(snap2name, ZFS_MAX_DATASET_NAME_LEN, | |
4138 | "%s@s2_%llu", clone1name, (u_longlong_t)id); | |
4139 | (void) snprintf(clone2name, ZFS_MAX_DATASET_NAME_LEN, | |
4140 | "%s/c2_%llu", osname, (u_longlong_t)id); | |
4141 | (void) snprintf(snap3name, ZFS_MAX_DATASET_NAME_LEN, | |
4142 | "%s@s3_%llu", clone1name, (u_longlong_t)id); | |
9babb374 | 4143 | |
13fe0198 | 4144 | error = dsl_destroy_head(clone2name); |
9babb374 | 4145 | if (error && error != ENOENT) |
13fe0198 MA |
4146 | fatal(0, "dsl_destroy_head(%s) = %d", clone2name, error); |
4147 | error = dsl_destroy_snapshot(snap3name, B_FALSE); | |
9babb374 | 4148 | if (error && error != ENOENT) |
13fe0198 MA |
4149 | fatal(0, "dsl_destroy_snapshot(%s) = %d", snap3name, error); |
4150 | error = dsl_destroy_snapshot(snap2name, B_FALSE); | |
9babb374 | 4151 | if (error && error != ENOENT) |
13fe0198 MA |
4152 | fatal(0, "dsl_destroy_snapshot(%s) = %d", snap2name, error); |
4153 | error = dsl_destroy_head(clone1name); | |
9babb374 | 4154 | if (error && error != ENOENT) |
13fe0198 MA |
4155 | fatal(0, "dsl_destroy_head(%s) = %d", clone1name, error); |
4156 | error = dsl_destroy_snapshot(snap1name, B_FALSE); | |
9babb374 | 4157 | if (error && error != ENOENT) |
13fe0198 | 4158 | fatal(0, "dsl_destroy_snapshot(%s) = %d", snap1name, error); |
40b84e7a | 4159 | |
eca7b760 IK |
4160 | umem_free(snap1name, ZFS_MAX_DATASET_NAME_LEN); |
4161 | umem_free(clone1name, ZFS_MAX_DATASET_NAME_LEN); | |
4162 | umem_free(snap2name, ZFS_MAX_DATASET_NAME_LEN); | |
4163 | umem_free(clone2name, ZFS_MAX_DATASET_NAME_LEN); | |
4164 | umem_free(snap3name, ZFS_MAX_DATASET_NAME_LEN); | |
9babb374 BB |
4165 | } |
4166 | ||
4167 | /* | |
4168 | * Verify dsl_dataset_promote handles EBUSY | |
4169 | */ | |
4170 | void | |
428870ff | 4171 | ztest_dsl_dataset_promote_busy(ztest_ds_t *zd, uint64_t id) |
9babb374 | 4172 | { |
13fe0198 | 4173 | objset_t *os; |
40b84e7a BB |
4174 | char *snap1name; |
4175 | char *clone1name; | |
4176 | char *snap2name; | |
4177 | char *clone2name; | |
4178 | char *snap3name; | |
428870ff BB |
4179 | char *osname = zd->zd_name; |
4180 | int error; | |
9babb374 | 4181 | |
eca7b760 IK |
4182 | snap1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); |
4183 | clone1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4184 | snap2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4185 | clone2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4186 | snap3name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
40b84e7a | 4187 | |
2d9142c9 | 4188 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
9babb374 | 4189 | |
428870ff | 4190 | ztest_dsl_dataset_cleanup(osname, id); |
9babb374 | 4191 | |
eca7b760 IK |
4192 | (void) snprintf(snap1name, ZFS_MAX_DATASET_NAME_LEN, |
4193 | "%s@s1_%llu", osname, (u_longlong_t)id); | |
4194 | (void) snprintf(clone1name, ZFS_MAX_DATASET_NAME_LEN, | |
4195 | "%s/c1_%llu", osname, (u_longlong_t)id); | |
4196 | (void) snprintf(snap2name, ZFS_MAX_DATASET_NAME_LEN, | |
4197 | "%s@s2_%llu", clone1name, (u_longlong_t)id); | |
4198 | (void) snprintf(clone2name, ZFS_MAX_DATASET_NAME_LEN, | |
4199 | "%s/c2_%llu", osname, (u_longlong_t)id); | |
4200 | (void) snprintf(snap3name, ZFS_MAX_DATASET_NAME_LEN, | |
4201 | "%s@s3_%llu", clone1name, (u_longlong_t)id); | |
9babb374 | 4202 | |
6f1ffb06 | 4203 | error = dmu_objset_snapshot_one(osname, strchr(snap1name, '@') + 1); |
9babb374 BB |
4204 | if (error && error != EEXIST) { |
4205 | if (error == ENOSPC) { | |
428870ff | 4206 | ztest_record_enospc(FTAG); |
9babb374 BB |
4207 | goto out; |
4208 | } | |
4209 | fatal(0, "dmu_take_snapshot(%s) = %d", snap1name, error); | |
4210 | } | |
4211 | ||
13fe0198 | 4212 | error = dmu_objset_clone(clone1name, snap1name); |
9babb374 BB |
4213 | if (error) { |
4214 | if (error == ENOSPC) { | |
428870ff | 4215 | ztest_record_enospc(FTAG); |
9babb374 BB |
4216 | goto out; |
4217 | } | |
4218 | fatal(0, "dmu_objset_create(%s) = %d", clone1name, error); | |
4219 | } | |
4220 | ||
6f1ffb06 | 4221 | error = dmu_objset_snapshot_one(clone1name, strchr(snap2name, '@') + 1); |
428870ff BB |
4222 | if (error && error != EEXIST) { |
4223 | if (error == ENOSPC) { | |
4224 | ztest_record_enospc(FTAG); | |
4225 | goto out; | |
34dc7c2f | 4226 | } |
428870ff BB |
4227 | fatal(0, "dmu_open_snapshot(%s) = %d", snap2name, error); |
4228 | } | |
34dc7c2f | 4229 | |
6f1ffb06 | 4230 | error = dmu_objset_snapshot_one(clone1name, strchr(snap3name, '@') + 1); |
428870ff BB |
4231 | if (error && error != EEXIST) { |
4232 | if (error == ENOSPC) { | |
4233 | ztest_record_enospc(FTAG); | |
4234 | goto out; | |
4235 | } | |
4236 | fatal(0, "dmu_open_snapshot(%s) = %d", snap3name, error); | |
4237 | } | |
34dc7c2f | 4238 | |
13fe0198 | 4239 | error = dmu_objset_clone(clone2name, snap3name); |
428870ff BB |
4240 | if (error) { |
4241 | if (error == ENOSPC) { | |
4242 | ztest_record_enospc(FTAG); | |
4243 | goto out; | |
4244 | } | |
4245 | fatal(0, "dmu_objset_create(%s) = %d", clone2name, error); | |
4246 | } | |
34dc7c2f | 4247 | |
4807c0ba | 4248 | error = ztest_dmu_objset_own(snap2name, DMU_OST_ANY, B_TRUE, B_TRUE, |
b5256303 | 4249 | FTAG, &os); |
428870ff | 4250 | if (error) |
13fe0198 | 4251 | fatal(0, "dmu_objset_own(%s) = %d", snap2name, error); |
428870ff | 4252 | error = dsl_dataset_promote(clone2name, NULL); |
9b67f605 | 4253 | if (error == ENOSPC) { |
b5256303 | 4254 | dmu_objset_disown(os, B_TRUE, FTAG); |
9b67f605 MA |
4255 | ztest_record_enospc(FTAG); |
4256 | goto out; | |
4257 | } | |
428870ff BB |
4258 | if (error != EBUSY) |
4259 | fatal(0, "dsl_dataset_promote(%s), %d, not EBUSY", clone2name, | |
4260 | error); | |
b5256303 | 4261 | dmu_objset_disown(os, B_TRUE, FTAG); |
34dc7c2f | 4262 | |
428870ff BB |
4263 | out: |
4264 | ztest_dsl_dataset_cleanup(osname, id); | |
34dc7c2f | 4265 | |
2d9142c9 | 4266 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
40b84e7a | 4267 | |
eca7b760 IK |
4268 | umem_free(snap1name, ZFS_MAX_DATASET_NAME_LEN); |
4269 | umem_free(clone1name, ZFS_MAX_DATASET_NAME_LEN); | |
4270 | umem_free(snap2name, ZFS_MAX_DATASET_NAME_LEN); | |
4271 | umem_free(clone2name, ZFS_MAX_DATASET_NAME_LEN); | |
4272 | umem_free(snap3name, ZFS_MAX_DATASET_NAME_LEN); | |
428870ff | 4273 | } |
34dc7c2f | 4274 | |
40b84e7a | 4275 | #undef OD_ARRAY_SIZE |
d1d7e268 | 4276 | #define OD_ARRAY_SIZE 4 |
40b84e7a | 4277 | |
428870ff BB |
4278 | /* |
4279 | * Verify that dmu_object_{alloc,free} work as expected. | |
4280 | */ | |
4281 | void | |
4282 | ztest_dmu_object_alloc_free(ztest_ds_t *zd, uint64_t id) | |
4283 | { | |
40b84e7a BB |
4284 | ztest_od_t *od; |
4285 | int batchsize; | |
4286 | int size; | |
d6320ddb | 4287 | int b; |
34dc7c2f | 4288 | |
d1d7e268 | 4289 | size = sizeof (ztest_od_t) * OD_ARRAY_SIZE; |
40b84e7a BB |
4290 | od = umem_alloc(size, UMEM_NOFAIL); |
4291 | batchsize = OD_ARRAY_SIZE; | |
4292 | ||
d6320ddb | 4293 | for (b = 0; b < batchsize; b++) |
50c957f7 NB |
4294 | ztest_od_init(od + b, id, FTAG, b, DMU_OT_UINT64_OTHER, |
4295 | 0, 0, 0); | |
34dc7c2f | 4296 | |
428870ff BB |
4297 | /* |
4298 | * Destroy the previous batch of objects, create a new batch, | |
4299 | * and do some I/O on the new objects. | |
4300 | */ | |
40b84e7a | 4301 | if (ztest_object_init(zd, od, size, B_TRUE) != 0) |
428870ff | 4302 | return; |
34dc7c2f | 4303 | |
428870ff BB |
4304 | while (ztest_random(4 * batchsize) != 0) |
4305 | ztest_io(zd, od[ztest_random(batchsize)].od_object, | |
4306 | ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
40b84e7a BB |
4307 | |
4308 | umem_free(od, size); | |
34dc7c2f BB |
4309 | } |
4310 | ||
047116ac TC |
4311 | /* |
4312 | * Rewind the global allocator to verify object allocation backfilling. | |
4313 | */ | |
4314 | void | |
4315 | ztest_dmu_object_next_chunk(ztest_ds_t *zd, uint64_t id) | |
4316 | { | |
4317 | objset_t *os = zd->zd_os; | |
4318 | int dnodes_per_chunk = 1 << dmu_object_alloc_chunk_shift; | |
4319 | uint64_t object; | |
4320 | ||
4321 | /* | |
4322 | * Rewind the global allocator randomly back to a lower object number | |
4323 | * to force backfilling and reclamation of recently freed dnodes. | |
4324 | */ | |
4325 | mutex_enter(&os->os_obj_lock); | |
4326 | object = ztest_random(os->os_obj_next_chunk); | |
4327 | os->os_obj_next_chunk = P2ALIGN(object, dnodes_per_chunk); | |
4328 | mutex_exit(&os->os_obj_lock); | |
4329 | } | |
4330 | ||
40b84e7a | 4331 | #undef OD_ARRAY_SIZE |
d1d7e268 | 4332 | #define OD_ARRAY_SIZE 2 |
40b84e7a | 4333 | |
34dc7c2f BB |
4334 | /* |
4335 | * Verify that dmu_{read,write} work as expected. | |
4336 | */ | |
34dc7c2f | 4337 | void |
428870ff | 4338 | ztest_dmu_read_write(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4339 | { |
40b84e7a BB |
4340 | int size; |
4341 | ztest_od_t *od; | |
4342 | ||
428870ff | 4343 | objset_t *os = zd->zd_os; |
d1d7e268 | 4344 | size = sizeof (ztest_od_t) * OD_ARRAY_SIZE; |
40b84e7a | 4345 | od = umem_alloc(size, UMEM_NOFAIL); |
34dc7c2f BB |
4346 | dmu_tx_t *tx; |
4347 | int i, freeit, error; | |
4348 | uint64_t n, s, txg; | |
4349 | bufwad_t *packbuf, *bigbuf, *pack, *bigH, *bigT; | |
428870ff BB |
4350 | uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize; |
4351 | uint64_t chunksize = (1000 + ztest_random(1000)) * sizeof (uint64_t); | |
34dc7c2f BB |
4352 | uint64_t regions = 997; |
4353 | uint64_t stride = 123456789ULL; | |
4354 | uint64_t width = 40; | |
4355 | int free_percent = 5; | |
4356 | ||
4357 | /* | |
4358 | * This test uses two objects, packobj and bigobj, that are always | |
4359 | * updated together (i.e. in the same tx) so that their contents are | |
4360 | * in sync and can be compared. Their contents relate to each other | |
4361 | * in a simple way: packobj is a dense array of 'bufwad' structures, | |
4362 | * while bigobj is a sparse array of the same bufwads. Specifically, | |
4363 | * for any index n, there are three bufwads that should be identical: | |
4364 | * | |
4365 | * packobj, at offset n * sizeof (bufwad_t) | |
4366 | * bigobj, at the head of the nth chunk | |
4367 | * bigobj, at the tail of the nth chunk | |
4368 | * | |
4369 | * The chunk size is arbitrary. It doesn't have to be a power of two, | |
4370 | * and it doesn't have any relation to the object blocksize. | |
4371 | * The only requirement is that it can hold at least two bufwads. | |
4372 | * | |
4373 | * Normally, we write the bufwad to each of these locations. | |
4374 | * However, free_percent of the time we instead write zeroes to | |
4375 | * packobj and perform a dmu_free_range() on bigobj. By comparing | |
4376 | * bigobj to packobj, we can verify that the DMU is correctly | |
4377 | * tracking which parts of an object are allocated and free, | |
4378 | * and that the contents of the allocated blocks are correct. | |
4379 | */ | |
4380 | ||
4381 | /* | |
4382 | * Read the directory info. If it's the first time, set things up. | |
4383 | */ | |
50c957f7 NB |
4384 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, chunksize); |
4385 | ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, 0, | |
4386 | chunksize); | |
34dc7c2f | 4387 | |
40b84e7a BB |
4388 | if (ztest_object_init(zd, od, size, B_FALSE) != 0) { |
4389 | umem_free(od, size); | |
428870ff | 4390 | return; |
40b84e7a | 4391 | } |
34dc7c2f | 4392 | |
428870ff BB |
4393 | bigobj = od[0].od_object; |
4394 | packobj = od[1].od_object; | |
4395 | chunksize = od[0].od_gen; | |
4396 | ASSERT(chunksize == od[1].od_gen); | |
34dc7c2f BB |
4397 | |
4398 | /* | |
4399 | * Prefetch a random chunk of the big object. | |
4400 | * Our aim here is to get some async reads in flight | |
4401 | * for blocks that we may free below; the DMU should | |
4402 | * handle this race correctly. | |
4403 | */ | |
4404 | n = ztest_random(regions) * stride + ztest_random(width); | |
4405 | s = 1 + ztest_random(2 * width - 1); | |
fcff0f35 PD |
4406 | dmu_prefetch(os, bigobj, 0, n * chunksize, s * chunksize, |
4407 | ZIO_PRIORITY_SYNC_READ); | |
34dc7c2f BB |
4408 | |
4409 | /* | |
4410 | * Pick a random index and compute the offsets into packobj and bigobj. | |
4411 | */ | |
4412 | n = ztest_random(regions) * stride + ztest_random(width); | |
4413 | s = 1 + ztest_random(width - 1); | |
4414 | ||
4415 | packoff = n * sizeof (bufwad_t); | |
4416 | packsize = s * sizeof (bufwad_t); | |
4417 | ||
428870ff BB |
4418 | bigoff = n * chunksize; |
4419 | bigsize = s * chunksize; | |
34dc7c2f BB |
4420 | |
4421 | packbuf = umem_alloc(packsize, UMEM_NOFAIL); | |
4422 | bigbuf = umem_alloc(bigsize, UMEM_NOFAIL); | |
4423 | ||
4424 | /* | |
4425 | * free_percent of the time, free a range of bigobj rather than | |
4426 | * overwriting it. | |
4427 | */ | |
4428 | freeit = (ztest_random(100) < free_percent); | |
4429 | ||
4430 | /* | |
4431 | * Read the current contents of our objects. | |
4432 | */ | |
428870ff | 4433 | error = dmu_read(os, packobj, packoff, packsize, packbuf, |
9babb374 | 4434 | DMU_READ_PREFETCH); |
c99c9001 | 4435 | ASSERT0(error); |
428870ff | 4436 | error = dmu_read(os, bigobj, bigoff, bigsize, bigbuf, |
9babb374 | 4437 | DMU_READ_PREFETCH); |
c99c9001 | 4438 | ASSERT0(error); |
34dc7c2f BB |
4439 | |
4440 | /* | |
4441 | * Get a tx for the mods to both packobj and bigobj. | |
4442 | */ | |
4443 | tx = dmu_tx_create(os); | |
4444 | ||
428870ff | 4445 | dmu_tx_hold_write(tx, packobj, packoff, packsize); |
34dc7c2f BB |
4446 | |
4447 | if (freeit) | |
428870ff | 4448 | dmu_tx_hold_free(tx, bigobj, bigoff, bigsize); |
34dc7c2f | 4449 | else |
428870ff | 4450 | dmu_tx_hold_write(tx, bigobj, bigoff, bigsize); |
34dc7c2f | 4451 | |
383fc4a9 MA |
4452 | /* This accounts for setting the checksum/compression. */ |
4453 | dmu_tx_hold_bonus(tx, bigobj); | |
4454 | ||
428870ff BB |
4455 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); |
4456 | if (txg == 0) { | |
34dc7c2f BB |
4457 | umem_free(packbuf, packsize); |
4458 | umem_free(bigbuf, bigsize); | |
40b84e7a | 4459 | umem_free(od, size); |
34dc7c2f BB |
4460 | return; |
4461 | } | |
4462 | ||
9b67f605 MA |
4463 | enum zio_checksum cksum; |
4464 | do { | |
4465 | cksum = (enum zio_checksum) | |
4466 | ztest_random_dsl_prop(ZFS_PROP_CHECKSUM); | |
4467 | } while (cksum >= ZIO_CHECKSUM_LEGACY_FUNCTIONS); | |
4468 | dmu_object_set_checksum(os, bigobj, cksum, tx); | |
428870ff | 4469 | |
9b67f605 MA |
4470 | enum zio_compress comp; |
4471 | do { | |
4472 | comp = (enum zio_compress) | |
4473 | ztest_random_dsl_prop(ZFS_PROP_COMPRESSION); | |
4474 | } while (comp >= ZIO_COMPRESS_LEGACY_FUNCTIONS); | |
4475 | dmu_object_set_compress(os, bigobj, comp, tx); | |
34dc7c2f BB |
4476 | |
4477 | /* | |
4478 | * For each index from n to n + s, verify that the existing bufwad | |
4479 | * in packobj matches the bufwads at the head and tail of the | |
4480 | * corresponding chunk in bigobj. Then update all three bufwads | |
4481 | * with the new values we want to write out. | |
4482 | */ | |
4483 | for (i = 0; i < s; i++) { | |
4484 | /* LINTED */ | |
4485 | pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t)); | |
4486 | /* LINTED */ | |
428870ff | 4487 | bigH = (bufwad_t *)((char *)bigbuf + i * chunksize); |
34dc7c2f | 4488 | /* LINTED */ |
428870ff | 4489 | bigT = (bufwad_t *)((char *)bigH + chunksize) - 1; |
34dc7c2f BB |
4490 | |
4491 | ASSERT((uintptr_t)bigH - (uintptr_t)bigbuf < bigsize); | |
4492 | ASSERT((uintptr_t)bigT - (uintptr_t)bigbuf < bigsize); | |
4493 | ||
4494 | if (pack->bw_txg > txg) | |
4495 | fatal(0, "future leak: got %llx, open txg is %llx", | |
4496 | pack->bw_txg, txg); | |
4497 | ||
4498 | if (pack->bw_data != 0 && pack->bw_index != n + i) | |
4499 | fatal(0, "wrong index: got %llx, wanted %llx+%llx", | |
4500 | pack->bw_index, n, i); | |
4501 | ||
4502 | if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0) | |
4503 | fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH); | |
4504 | ||
4505 | if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0) | |
4506 | fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT); | |
4507 | ||
4508 | if (freeit) { | |
4509 | bzero(pack, sizeof (bufwad_t)); | |
4510 | } else { | |
4511 | pack->bw_index = n + i; | |
4512 | pack->bw_txg = txg; | |
4513 | pack->bw_data = 1 + ztest_random(-2ULL); | |
4514 | } | |
4515 | *bigH = *pack; | |
4516 | *bigT = *pack; | |
4517 | } | |
4518 | ||
4519 | /* | |
4520 | * We've verified all the old bufwads, and made new ones. | |
4521 | * Now write them out. | |
4522 | */ | |
428870ff | 4523 | dmu_write(os, packobj, packoff, packsize, packbuf, tx); |
34dc7c2f BB |
4524 | |
4525 | if (freeit) { | |
c242c188 | 4526 | if (ztest_opts.zo_verbose >= 7) { |
34dc7c2f BB |
4527 | (void) printf("freeing offset %llx size %llx" |
4528 | " txg %llx\n", | |
4529 | (u_longlong_t)bigoff, | |
4530 | (u_longlong_t)bigsize, | |
4531 | (u_longlong_t)txg); | |
4532 | } | |
428870ff | 4533 | VERIFY(0 == dmu_free_range(os, bigobj, bigoff, bigsize, tx)); |
34dc7c2f | 4534 | } else { |
c242c188 | 4535 | if (ztest_opts.zo_verbose >= 7) { |
34dc7c2f BB |
4536 | (void) printf("writing offset %llx size %llx" |
4537 | " txg %llx\n", | |
4538 | (u_longlong_t)bigoff, | |
4539 | (u_longlong_t)bigsize, | |
4540 | (u_longlong_t)txg); | |
4541 | } | |
428870ff | 4542 | dmu_write(os, bigobj, bigoff, bigsize, bigbuf, tx); |
34dc7c2f BB |
4543 | } |
4544 | ||
4545 | dmu_tx_commit(tx); | |
4546 | ||
4547 | /* | |
4548 | * Sanity check the stuff we just wrote. | |
4549 | */ | |
4550 | { | |
4551 | void *packcheck = umem_alloc(packsize, UMEM_NOFAIL); | |
4552 | void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL); | |
4553 | ||
428870ff | 4554 | VERIFY(0 == dmu_read(os, packobj, packoff, |
9babb374 | 4555 | packsize, packcheck, DMU_READ_PREFETCH)); |
428870ff | 4556 | VERIFY(0 == dmu_read(os, bigobj, bigoff, |
9babb374 | 4557 | bigsize, bigcheck, DMU_READ_PREFETCH)); |
34dc7c2f BB |
4558 | |
4559 | ASSERT(bcmp(packbuf, packcheck, packsize) == 0); | |
4560 | ASSERT(bcmp(bigbuf, bigcheck, bigsize) == 0); | |
4561 | ||
4562 | umem_free(packcheck, packsize); | |
4563 | umem_free(bigcheck, bigsize); | |
4564 | } | |
4565 | ||
4566 | umem_free(packbuf, packsize); | |
4567 | umem_free(bigbuf, bigsize); | |
40b84e7a | 4568 | umem_free(od, size); |
34dc7c2f BB |
4569 | } |
4570 | ||
9babb374 BB |
4571 | void |
4572 | compare_and_update_pbbufs(uint64_t s, bufwad_t *packbuf, bufwad_t *bigbuf, | |
428870ff | 4573 | uint64_t bigsize, uint64_t n, uint64_t chunksize, uint64_t txg) |
9babb374 BB |
4574 | { |
4575 | uint64_t i; | |
4576 | bufwad_t *pack; | |
4577 | bufwad_t *bigH; | |
4578 | bufwad_t *bigT; | |
4579 | ||
4580 | /* | |
4581 | * For each index from n to n + s, verify that the existing bufwad | |
4582 | * in packobj matches the bufwads at the head and tail of the | |
4583 | * corresponding chunk in bigobj. Then update all three bufwads | |
4584 | * with the new values we want to write out. | |
4585 | */ | |
4586 | for (i = 0; i < s; i++) { | |
4587 | /* LINTED */ | |
4588 | pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t)); | |
4589 | /* LINTED */ | |
428870ff | 4590 | bigH = (bufwad_t *)((char *)bigbuf + i * chunksize); |
9babb374 | 4591 | /* LINTED */ |
428870ff | 4592 | bigT = (bufwad_t *)((char *)bigH + chunksize) - 1; |
9babb374 BB |
4593 | |
4594 | ASSERT((uintptr_t)bigH - (uintptr_t)bigbuf < bigsize); | |
4595 | ASSERT((uintptr_t)bigT - (uintptr_t)bigbuf < bigsize); | |
4596 | ||
4597 | if (pack->bw_txg > txg) | |
4598 | fatal(0, "future leak: got %llx, open txg is %llx", | |
4599 | pack->bw_txg, txg); | |
4600 | ||
4601 | if (pack->bw_data != 0 && pack->bw_index != n + i) | |
4602 | fatal(0, "wrong index: got %llx, wanted %llx+%llx", | |
4603 | pack->bw_index, n, i); | |
4604 | ||
4605 | if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0) | |
4606 | fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH); | |
4607 | ||
4608 | if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0) | |
4609 | fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT); | |
4610 | ||
4611 | pack->bw_index = n + i; | |
4612 | pack->bw_txg = txg; | |
4613 | pack->bw_data = 1 + ztest_random(-2ULL); | |
4614 | ||
4615 | *bigH = *pack; | |
4616 | *bigT = *pack; | |
4617 | } | |
4618 | } | |
4619 | ||
40b84e7a | 4620 | #undef OD_ARRAY_SIZE |
d1d7e268 | 4621 | #define OD_ARRAY_SIZE 2 |
40b84e7a | 4622 | |
9babb374 | 4623 | void |
428870ff | 4624 | ztest_dmu_read_write_zcopy(ztest_ds_t *zd, uint64_t id) |
9babb374 | 4625 | { |
428870ff | 4626 | objset_t *os = zd->zd_os; |
40b84e7a | 4627 | ztest_od_t *od; |
9babb374 BB |
4628 | dmu_tx_t *tx; |
4629 | uint64_t i; | |
4630 | int error; | |
40b84e7a | 4631 | int size; |
9babb374 BB |
4632 | uint64_t n, s, txg; |
4633 | bufwad_t *packbuf, *bigbuf; | |
428870ff BB |
4634 | uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize; |
4635 | uint64_t blocksize = ztest_random_blocksize(); | |
4636 | uint64_t chunksize = blocksize; | |
9babb374 BB |
4637 | uint64_t regions = 997; |
4638 | uint64_t stride = 123456789ULL; | |
4639 | uint64_t width = 9; | |
4640 | dmu_buf_t *bonus_db; | |
4641 | arc_buf_t **bigbuf_arcbufs; | |
428870ff | 4642 | dmu_object_info_t doi; |
9babb374 | 4643 | |
d1d7e268 | 4644 | size = sizeof (ztest_od_t) * OD_ARRAY_SIZE; |
40b84e7a BB |
4645 | od = umem_alloc(size, UMEM_NOFAIL); |
4646 | ||
9babb374 BB |
4647 | /* |
4648 | * This test uses two objects, packobj and bigobj, that are always | |
4649 | * updated together (i.e. in the same tx) so that their contents are | |
4650 | * in sync and can be compared. Their contents relate to each other | |
4651 | * in a simple way: packobj is a dense array of 'bufwad' structures, | |
4652 | * while bigobj is a sparse array of the same bufwads. Specifically, | |
4653 | * for any index n, there are three bufwads that should be identical: | |
4654 | * | |
4655 | * packobj, at offset n * sizeof (bufwad_t) | |
4656 | * bigobj, at the head of the nth chunk | |
4657 | * bigobj, at the tail of the nth chunk | |
4658 | * | |
4659 | * The chunk size is set equal to bigobj block size so that | |
440a3eb9 | 4660 | * dmu_assign_arcbuf_by_dbuf() can be tested for object updates. |
9babb374 BB |
4661 | */ |
4662 | ||
4663 | /* | |
4664 | * Read the directory info. If it's the first time, set things up. | |
4665 | */ | |
50c957f7 NB |
4666 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0, 0); |
4667 | ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, 0, | |
4668 | chunksize); | |
40b84e7a | 4669 | |
9babb374 | 4670 | |
40b84e7a BB |
4671 | if (ztest_object_init(zd, od, size, B_FALSE) != 0) { |
4672 | umem_free(od, size); | |
428870ff | 4673 | return; |
40b84e7a | 4674 | } |
9babb374 | 4675 | |
428870ff BB |
4676 | bigobj = od[0].od_object; |
4677 | packobj = od[1].od_object; | |
4678 | blocksize = od[0].od_blocksize; | |
4679 | chunksize = blocksize; | |
4680 | ASSERT(chunksize == od[1].od_gen); | |
9babb374 | 4681 | |
428870ff BB |
4682 | VERIFY(dmu_object_info(os, bigobj, &doi) == 0); |
4683 | VERIFY(ISP2(doi.doi_data_block_size)); | |
4684 | VERIFY(chunksize == doi.doi_data_block_size); | |
4685 | VERIFY(chunksize >= 2 * sizeof (bufwad_t)); | |
9babb374 BB |
4686 | |
4687 | /* | |
4688 | * Pick a random index and compute the offsets into packobj and bigobj. | |
4689 | */ | |
4690 | n = ztest_random(regions) * stride + ztest_random(width); | |
4691 | s = 1 + ztest_random(width - 1); | |
4692 | ||
4693 | packoff = n * sizeof (bufwad_t); | |
4694 | packsize = s * sizeof (bufwad_t); | |
4695 | ||
428870ff BB |
4696 | bigoff = n * chunksize; |
4697 | bigsize = s * chunksize; | |
9babb374 BB |
4698 | |
4699 | packbuf = umem_zalloc(packsize, UMEM_NOFAIL); | |
4700 | bigbuf = umem_zalloc(bigsize, UMEM_NOFAIL); | |
4701 | ||
428870ff | 4702 | VERIFY3U(0, ==, dmu_bonus_hold(os, bigobj, FTAG, &bonus_db)); |
9babb374 BB |
4703 | |
4704 | bigbuf_arcbufs = umem_zalloc(2 * s * sizeof (arc_buf_t *), UMEM_NOFAIL); | |
4705 | ||
4706 | /* | |
4707 | * Iteration 0 test zcopy for DB_UNCACHED dbufs. | |
4708 | * Iteration 1 test zcopy to already referenced dbufs. | |
4709 | * Iteration 2 test zcopy to dirty dbuf in the same txg. | |
4710 | * Iteration 3 test zcopy to dbuf dirty in previous txg. | |
4711 | * Iteration 4 test zcopy when dbuf is no longer dirty. | |
4712 | * Iteration 5 test zcopy when it can't be done. | |
4713 | * Iteration 6 one more zcopy write. | |
4714 | */ | |
4715 | for (i = 0; i < 7; i++) { | |
4716 | uint64_t j; | |
4717 | uint64_t off; | |
4718 | ||
4719 | /* | |
4720 | * In iteration 5 (i == 5) use arcbufs | |
4721 | * that don't match bigobj blksz to test | |
440a3eb9 | 4722 | * dmu_assign_arcbuf_by_dbuf() when it can't directly |
9babb374 BB |
4723 | * assign an arcbuf to a dbuf. |
4724 | */ | |
4725 | for (j = 0; j < s; j++) { | |
b9541d6b | 4726 | if (i != 5 || chunksize < (SPA_MINBLOCKSIZE * 2)) { |
9babb374 | 4727 | bigbuf_arcbufs[j] = |
428870ff | 4728 | dmu_request_arcbuf(bonus_db, chunksize); |
9babb374 BB |
4729 | } else { |
4730 | bigbuf_arcbufs[2 * j] = | |
428870ff | 4731 | dmu_request_arcbuf(bonus_db, chunksize / 2); |
9babb374 | 4732 | bigbuf_arcbufs[2 * j + 1] = |
428870ff | 4733 | dmu_request_arcbuf(bonus_db, chunksize / 2); |
9babb374 BB |
4734 | } |
4735 | } | |
4736 | ||
4737 | /* | |
4738 | * Get a tx for the mods to both packobj and bigobj. | |
4739 | */ | |
4740 | tx = dmu_tx_create(os); | |
4741 | ||
428870ff BB |
4742 | dmu_tx_hold_write(tx, packobj, packoff, packsize); |
4743 | dmu_tx_hold_write(tx, bigobj, bigoff, bigsize); | |
9babb374 | 4744 | |
428870ff BB |
4745 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); |
4746 | if (txg == 0) { | |
9babb374 BB |
4747 | umem_free(packbuf, packsize); |
4748 | umem_free(bigbuf, bigsize); | |
4749 | for (j = 0; j < s; j++) { | |
b9541d6b CW |
4750 | if (i != 5 || |
4751 | chunksize < (SPA_MINBLOCKSIZE * 2)) { | |
9babb374 BB |
4752 | dmu_return_arcbuf(bigbuf_arcbufs[j]); |
4753 | } else { | |
4754 | dmu_return_arcbuf( | |
4755 | bigbuf_arcbufs[2 * j]); | |
4756 | dmu_return_arcbuf( | |
4757 | bigbuf_arcbufs[2 * j + 1]); | |
4758 | } | |
4759 | } | |
4760 | umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *)); | |
40b84e7a | 4761 | umem_free(od, size); |
9babb374 BB |
4762 | dmu_buf_rele(bonus_db, FTAG); |
4763 | return; | |
4764 | } | |
4765 | ||
9babb374 BB |
4766 | /* |
4767 | * 50% of the time don't read objects in the 1st iteration to | |
440a3eb9 TC |
4768 | * test dmu_assign_arcbuf_by_dbuf() for the case when there are |
4769 | * no existing dbufs for the specified offsets. | |
9babb374 BB |
4770 | */ |
4771 | if (i != 0 || ztest_random(2) != 0) { | |
428870ff | 4772 | error = dmu_read(os, packobj, packoff, |
9babb374 | 4773 | packsize, packbuf, DMU_READ_PREFETCH); |
c99c9001 | 4774 | ASSERT0(error); |
428870ff | 4775 | error = dmu_read(os, bigobj, bigoff, bigsize, |
9babb374 | 4776 | bigbuf, DMU_READ_PREFETCH); |
c99c9001 | 4777 | ASSERT0(error); |
9babb374 BB |
4778 | } |
4779 | compare_and_update_pbbufs(s, packbuf, bigbuf, bigsize, | |
428870ff | 4780 | n, chunksize, txg); |
9babb374 BB |
4781 | |
4782 | /* | |
4783 | * We've verified all the old bufwads, and made new ones. | |
4784 | * Now write them out. | |
4785 | */ | |
428870ff | 4786 | dmu_write(os, packobj, packoff, packsize, packbuf, tx); |
c242c188 | 4787 | if (ztest_opts.zo_verbose >= 7) { |
9babb374 BB |
4788 | (void) printf("writing offset %llx size %llx" |
4789 | " txg %llx\n", | |
4790 | (u_longlong_t)bigoff, | |
4791 | (u_longlong_t)bigsize, | |
4792 | (u_longlong_t)txg); | |
4793 | } | |
428870ff | 4794 | for (off = bigoff, j = 0; j < s; j++, off += chunksize) { |
9babb374 | 4795 | dmu_buf_t *dbt; |
b9541d6b | 4796 | if (i != 5 || chunksize < (SPA_MINBLOCKSIZE * 2)) { |
9babb374 | 4797 | bcopy((caddr_t)bigbuf + (off - bigoff), |
428870ff | 4798 | bigbuf_arcbufs[j]->b_data, chunksize); |
9babb374 BB |
4799 | } else { |
4800 | bcopy((caddr_t)bigbuf + (off - bigoff), | |
4801 | bigbuf_arcbufs[2 * j]->b_data, | |
428870ff | 4802 | chunksize / 2); |
9babb374 | 4803 | bcopy((caddr_t)bigbuf + (off - bigoff) + |
428870ff | 4804 | chunksize / 2, |
9babb374 | 4805 | bigbuf_arcbufs[2 * j + 1]->b_data, |
428870ff | 4806 | chunksize / 2); |
9babb374 BB |
4807 | } |
4808 | ||
4809 | if (i == 1) { | |
428870ff BB |
4810 | VERIFY(dmu_buf_hold(os, bigobj, off, |
4811 | FTAG, &dbt, DMU_READ_NO_PREFETCH) == 0); | |
9babb374 | 4812 | } |
b9541d6b | 4813 | if (i != 5 || chunksize < (SPA_MINBLOCKSIZE * 2)) { |
440a3eb9 | 4814 | dmu_assign_arcbuf_by_dbuf(bonus_db, off, |
9babb374 BB |
4815 | bigbuf_arcbufs[j], tx); |
4816 | } else { | |
440a3eb9 | 4817 | dmu_assign_arcbuf_by_dbuf(bonus_db, off, |
9babb374 | 4818 | bigbuf_arcbufs[2 * j], tx); |
440a3eb9 | 4819 | dmu_assign_arcbuf_by_dbuf(bonus_db, |
428870ff | 4820 | off + chunksize / 2, |
9babb374 BB |
4821 | bigbuf_arcbufs[2 * j + 1], tx); |
4822 | } | |
4823 | if (i == 1) { | |
4824 | dmu_buf_rele(dbt, FTAG); | |
4825 | } | |
4826 | } | |
4827 | dmu_tx_commit(tx); | |
4828 | ||
4829 | /* | |
4830 | * Sanity check the stuff we just wrote. | |
4831 | */ | |
4832 | { | |
4833 | void *packcheck = umem_alloc(packsize, UMEM_NOFAIL); | |
4834 | void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL); | |
4835 | ||
428870ff | 4836 | VERIFY(0 == dmu_read(os, packobj, packoff, |
9babb374 | 4837 | packsize, packcheck, DMU_READ_PREFETCH)); |
428870ff | 4838 | VERIFY(0 == dmu_read(os, bigobj, bigoff, |
9babb374 BB |
4839 | bigsize, bigcheck, DMU_READ_PREFETCH)); |
4840 | ||
4841 | ASSERT(bcmp(packbuf, packcheck, packsize) == 0); | |
4842 | ASSERT(bcmp(bigbuf, bigcheck, bigsize) == 0); | |
4843 | ||
4844 | umem_free(packcheck, packsize); | |
4845 | umem_free(bigcheck, bigsize); | |
4846 | } | |
4847 | if (i == 2) { | |
4848 | txg_wait_open(dmu_objset_pool(os), 0); | |
4849 | } else if (i == 3) { | |
4850 | txg_wait_synced(dmu_objset_pool(os), 0); | |
4851 | } | |
4852 | } | |
4853 | ||
4854 | dmu_buf_rele(bonus_db, FTAG); | |
4855 | umem_free(packbuf, packsize); | |
4856 | umem_free(bigbuf, bigsize); | |
4857 | umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *)); | |
40b84e7a | 4858 | umem_free(od, size); |
9babb374 BB |
4859 | } |
4860 | ||
428870ff | 4861 | /* ARGSUSED */ |
34dc7c2f | 4862 | void |
428870ff | 4863 | ztest_dmu_write_parallel(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4864 | { |
40b84e7a BB |
4865 | ztest_od_t *od; |
4866 | ||
d1d7e268 | 4867 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
428870ff BB |
4868 | uint64_t offset = (1ULL << (ztest_random(20) + 43)) + |
4869 | (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
34dc7c2f BB |
4870 | |
4871 | /* | |
428870ff BB |
4872 | * Have multiple threads write to large offsets in an object |
4873 | * to verify that parallel writes to an object -- even to the | |
4874 | * same blocks within the object -- doesn't cause any trouble. | |
34dc7c2f | 4875 | */ |
50c957f7 | 4876 | ztest_od_init(od, ID_PARALLEL, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0); |
9babb374 | 4877 | |
40b84e7a | 4878 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) |
34dc7c2f | 4879 | return; |
34dc7c2f | 4880 | |
428870ff | 4881 | while (ztest_random(10) != 0) |
40b84e7a BB |
4882 | ztest_io(zd, od->od_object, offset); |
4883 | ||
d1d7e268 | 4884 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 4885 | } |
34dc7c2f | 4886 | |
428870ff BB |
4887 | void |
4888 | ztest_dmu_prealloc(ztest_ds_t *zd, uint64_t id) | |
4889 | { | |
40b84e7a | 4890 | ztest_od_t *od; |
428870ff BB |
4891 | uint64_t offset = (1ULL << (ztest_random(4) + SPA_MAXBLOCKSHIFT)) + |
4892 | (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
4893 | uint64_t count = ztest_random(20) + 1; | |
4894 | uint64_t blocksize = ztest_random_blocksize(); | |
4895 | void *data; | |
34dc7c2f | 4896 | |
d1d7e268 | 4897 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
34dc7c2f | 4898 | |
50c957f7 | 4899 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0, 0); |
40b84e7a | 4900 | |
d1d7e268 MK |
4901 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
4902 | !ztest_random(2)) != 0) { | |
4903 | umem_free(od, sizeof (ztest_od_t)); | |
34dc7c2f | 4904 | return; |
40b84e7a | 4905 | } |
34dc7c2f | 4906 | |
40b84e7a | 4907 | if (ztest_truncate(zd, od->od_object, offset, count * blocksize) != 0) { |
d1d7e268 | 4908 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f | 4909 | return; |
40b84e7a | 4910 | } |
34dc7c2f | 4911 | |
40b84e7a | 4912 | ztest_prealloc(zd, od->od_object, offset, count * blocksize); |
34dc7c2f | 4913 | |
428870ff | 4914 | data = umem_zalloc(blocksize, UMEM_NOFAIL); |
34dc7c2f | 4915 | |
428870ff BB |
4916 | while (ztest_random(count) != 0) { |
4917 | uint64_t randoff = offset + (ztest_random(count) * blocksize); | |
40b84e7a | 4918 | if (ztest_write(zd, od->od_object, randoff, blocksize, |
428870ff BB |
4919 | data) != 0) |
4920 | break; | |
4921 | while (ztest_random(4) != 0) | |
40b84e7a | 4922 | ztest_io(zd, od->od_object, randoff); |
9babb374 | 4923 | } |
34dc7c2f | 4924 | |
428870ff | 4925 | umem_free(data, blocksize); |
d1d7e268 | 4926 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f BB |
4927 | } |
4928 | ||
4929 | /* | |
4930 | * Verify that zap_{create,destroy,add,remove,update} work as expected. | |
4931 | */ | |
4932 | #define ZTEST_ZAP_MIN_INTS 1 | |
4933 | #define ZTEST_ZAP_MAX_INTS 4 | |
4934 | #define ZTEST_ZAP_MAX_PROPS 1000 | |
4935 | ||
4936 | void | |
428870ff | 4937 | ztest_zap(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4938 | { |
428870ff | 4939 | objset_t *os = zd->zd_os; |
40b84e7a | 4940 | ztest_od_t *od; |
34dc7c2f BB |
4941 | uint64_t object; |
4942 | uint64_t txg, last_txg; | |
4943 | uint64_t value[ZTEST_ZAP_MAX_INTS]; | |
4944 | uint64_t zl_ints, zl_intsize, prop; | |
4945 | int i, ints; | |
4946 | dmu_tx_t *tx; | |
4947 | char propname[100], txgname[100]; | |
4948 | int error; | |
34dc7c2f BB |
4949 | char *hc[2] = { "s.acl.h", ".s.open.h.hyLZlg" }; |
4950 | ||
d1d7e268 | 4951 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 4952 | ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0, 0); |
34dc7c2f | 4953 | |
40b84e7a | 4954 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
02730c33 | 4955 | !ztest_random(2)) != 0) |
40b84e7a | 4956 | goto out; |
34dc7c2f | 4957 | |
40b84e7a | 4958 | object = od->od_object; |
34dc7c2f | 4959 | |
428870ff BB |
4960 | /* |
4961 | * Generate a known hash collision, and verify that | |
4962 | * we can lookup and remove both entries. | |
4963 | */ | |
4964 | tx = dmu_tx_create(os); | |
4965 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); | |
4966 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4967 | if (txg == 0) | |
40b84e7a | 4968 | goto out; |
428870ff BB |
4969 | for (i = 0; i < 2; i++) { |
4970 | value[i] = i; | |
4971 | VERIFY3U(0, ==, zap_add(os, object, hc[i], sizeof (uint64_t), | |
4972 | 1, &value[i], tx)); | |
4973 | } | |
4974 | for (i = 0; i < 2; i++) { | |
4975 | VERIFY3U(EEXIST, ==, zap_add(os, object, hc[i], | |
4976 | sizeof (uint64_t), 1, &value[i], tx)); | |
4977 | VERIFY3U(0, ==, | |
4978 | zap_length(os, object, hc[i], &zl_intsize, &zl_ints)); | |
4979 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
4980 | ASSERT3U(zl_ints, ==, 1); | |
4981 | } | |
4982 | for (i = 0; i < 2; i++) { | |
4983 | VERIFY3U(0, ==, zap_remove(os, object, hc[i], tx)); | |
34dc7c2f | 4984 | } |
428870ff | 4985 | dmu_tx_commit(tx); |
34dc7c2f | 4986 | |
428870ff BB |
4987 | /* |
4988 | * Generate a buch of random entries. | |
4989 | */ | |
34dc7c2f BB |
4990 | ints = MAX(ZTEST_ZAP_MIN_INTS, object % ZTEST_ZAP_MAX_INTS); |
4991 | ||
4992 | prop = ztest_random(ZTEST_ZAP_MAX_PROPS); | |
4993 | (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop); | |
4994 | (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop); | |
4995 | bzero(value, sizeof (value)); | |
4996 | last_txg = 0; | |
4997 | ||
4998 | /* | |
4999 | * If these zap entries already exist, validate their contents. | |
5000 | */ | |
5001 | error = zap_length(os, object, txgname, &zl_intsize, &zl_ints); | |
5002 | if (error == 0) { | |
5003 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
5004 | ASSERT3U(zl_ints, ==, 1); | |
5005 | ||
5006 | VERIFY(zap_lookup(os, object, txgname, zl_intsize, | |
5007 | zl_ints, &last_txg) == 0); | |
5008 | ||
5009 | VERIFY(zap_length(os, object, propname, &zl_intsize, | |
5010 | &zl_ints) == 0); | |
5011 | ||
5012 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
5013 | ASSERT3U(zl_ints, ==, ints); | |
5014 | ||
5015 | VERIFY(zap_lookup(os, object, propname, zl_intsize, | |
5016 | zl_ints, value) == 0); | |
5017 | ||
5018 | for (i = 0; i < ints; i++) { | |
5019 | ASSERT3U(value[i], ==, last_txg + object + i); | |
5020 | } | |
5021 | } else { | |
5022 | ASSERT3U(error, ==, ENOENT); | |
5023 | } | |
5024 | ||
5025 | /* | |
5026 | * Atomically update two entries in our zap object. | |
5027 | * The first is named txg_%llu, and contains the txg | |
5028 | * in which the property was last updated. The second | |
5029 | * is named prop_%llu, and the nth element of its value | |
5030 | * should be txg + object + n. | |
5031 | */ | |
5032 | tx = dmu_tx_create(os); | |
428870ff BB |
5033 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
5034 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
5035 | if (txg == 0) | |
40b84e7a | 5036 | goto out; |
34dc7c2f BB |
5037 | |
5038 | if (last_txg > txg) | |
5039 | fatal(0, "zap future leak: old %llu new %llu", last_txg, txg); | |
5040 | ||
5041 | for (i = 0; i < ints; i++) | |
5042 | value[i] = txg + object + i; | |
5043 | ||
428870ff BB |
5044 | VERIFY3U(0, ==, zap_update(os, object, txgname, sizeof (uint64_t), |
5045 | 1, &txg, tx)); | |
5046 | VERIFY3U(0, ==, zap_update(os, object, propname, sizeof (uint64_t), | |
5047 | ints, value, tx)); | |
34dc7c2f BB |
5048 | |
5049 | dmu_tx_commit(tx); | |
5050 | ||
5051 | /* | |
5052 | * Remove a random pair of entries. | |
5053 | */ | |
5054 | prop = ztest_random(ZTEST_ZAP_MAX_PROPS); | |
5055 | (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop); | |
5056 | (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop); | |
5057 | ||
5058 | error = zap_length(os, object, txgname, &zl_intsize, &zl_ints); | |
5059 | ||
5060 | if (error == ENOENT) | |
40b84e7a | 5061 | goto out; |
34dc7c2f | 5062 | |
c99c9001 | 5063 | ASSERT0(error); |
34dc7c2f BB |
5064 | |
5065 | tx = dmu_tx_create(os); | |
428870ff BB |
5066 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
5067 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
5068 | if (txg == 0) | |
40b84e7a | 5069 | goto out; |
428870ff BB |
5070 | VERIFY3U(0, ==, zap_remove(os, object, txgname, tx)); |
5071 | VERIFY3U(0, ==, zap_remove(os, object, propname, tx)); | |
5072 | dmu_tx_commit(tx); | |
40b84e7a | 5073 | out: |
d1d7e268 | 5074 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 5075 | } |
34dc7c2f | 5076 | |
428870ff BB |
5077 | /* |
5078 | * Testcase to test the upgrading of a microzap to fatzap. | |
5079 | */ | |
5080 | void | |
5081 | ztest_fzap(ztest_ds_t *zd, uint64_t id) | |
5082 | { | |
5083 | objset_t *os = zd->zd_os; | |
40b84e7a | 5084 | ztest_od_t *od; |
428870ff | 5085 | uint64_t object, txg; |
d6320ddb | 5086 | int i; |
34dc7c2f | 5087 | |
d1d7e268 | 5088 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 5089 | ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0, 0); |
428870ff | 5090 | |
40b84e7a | 5091 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
02730c33 | 5092 | !ztest_random(2)) != 0) |
40b84e7a BB |
5093 | goto out; |
5094 | object = od->od_object; | |
34dc7c2f BB |
5095 | |
5096 | /* | |
428870ff BB |
5097 | * Add entries to this ZAP and make sure it spills over |
5098 | * and gets upgraded to a fatzap. Also, since we are adding | |
5099 | * 2050 entries we should see ptrtbl growth and leaf-block split. | |
34dc7c2f | 5100 | */ |
d6320ddb | 5101 | for (i = 0; i < 2050; i++) { |
eca7b760 | 5102 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
5103 | uint64_t value = i; |
5104 | dmu_tx_t *tx; | |
5105 | int error; | |
34dc7c2f | 5106 | |
428870ff | 5107 | (void) snprintf(name, sizeof (name), "fzap-%llu-%llu", |
b8864a23 | 5108 | (u_longlong_t)id, (u_longlong_t)value); |
428870ff BB |
5109 | |
5110 | tx = dmu_tx_create(os); | |
5111 | dmu_tx_hold_zap(tx, object, B_TRUE, name); | |
5112 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
5113 | if (txg == 0) | |
40b84e7a | 5114 | goto out; |
428870ff BB |
5115 | error = zap_add(os, object, name, sizeof (uint64_t), 1, |
5116 | &value, tx); | |
5117 | ASSERT(error == 0 || error == EEXIST); | |
5118 | dmu_tx_commit(tx); | |
34dc7c2f | 5119 | } |
40b84e7a | 5120 | out: |
d1d7e268 | 5121 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f BB |
5122 | } |
5123 | ||
428870ff | 5124 | /* ARGSUSED */ |
34dc7c2f | 5125 | void |
428870ff | 5126 | ztest_zap_parallel(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 5127 | { |
428870ff | 5128 | objset_t *os = zd->zd_os; |
40b84e7a | 5129 | ztest_od_t *od; |
34dc7c2f BB |
5130 | uint64_t txg, object, count, wsize, wc, zl_wsize, zl_wc; |
5131 | dmu_tx_t *tx; | |
5132 | int i, namelen, error; | |
428870ff | 5133 | int micro = ztest_random(2); |
34dc7c2f BB |
5134 | char name[20], string_value[20]; |
5135 | void *data; | |
5136 | ||
d1d7e268 | 5137 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 5138 | ztest_od_init(od, ID_PARALLEL, FTAG, micro, DMU_OT_ZAP_OTHER, 0, 0, 0); |
428870ff | 5139 | |
40b84e7a | 5140 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
d1d7e268 | 5141 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 5142 | return; |
40b84e7a | 5143 | } |
428870ff | 5144 | |
40b84e7a | 5145 | object = od->od_object; |
428870ff | 5146 | |
34dc7c2f BB |
5147 | /* |
5148 | * Generate a random name of the form 'xxx.....' where each | |
5149 | * x is a random printable character and the dots are dots. | |
5150 | * There are 94 such characters, and the name length goes from | |
5151 | * 6 to 20, so there are 94^3 * 15 = 12,458,760 possible names. | |
5152 | */ | |
5153 | namelen = ztest_random(sizeof (name) - 5) + 5 + 1; | |
5154 | ||
5155 | for (i = 0; i < 3; i++) | |
5156 | name[i] = '!' + ztest_random('~' - '!' + 1); | |
5157 | for (; i < namelen - 1; i++) | |
5158 | name[i] = '.'; | |
5159 | name[i] = '\0'; | |
5160 | ||
428870ff | 5161 | if ((namelen & 1) || micro) { |
34dc7c2f BB |
5162 | wsize = sizeof (txg); |
5163 | wc = 1; | |
5164 | data = &txg; | |
5165 | } else { | |
5166 | wsize = 1; | |
5167 | wc = namelen; | |
5168 | data = string_value; | |
5169 | } | |
5170 | ||
5171 | count = -1ULL; | |
13fe0198 | 5172 | VERIFY0(zap_count(os, object, &count)); |
34dc7c2f BB |
5173 | ASSERT(count != -1ULL); |
5174 | ||
5175 | /* | |
5176 | * Select an operation: length, lookup, add, update, remove. | |
5177 | */ | |
5178 | i = ztest_random(5); | |
5179 | ||
5180 | if (i >= 2) { | |
5181 | tx = dmu_tx_create(os); | |
428870ff BB |
5182 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
5183 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
df053d67 GN |
5184 | if (txg == 0) { |
5185 | umem_free(od, sizeof (ztest_od_t)); | |
34dc7c2f | 5186 | return; |
df053d67 | 5187 | } |
34dc7c2f BB |
5188 | bcopy(name, string_value, namelen); |
5189 | } else { | |
5190 | tx = NULL; | |
5191 | txg = 0; | |
5192 | bzero(string_value, namelen); | |
5193 | } | |
5194 | ||
5195 | switch (i) { | |
5196 | ||
5197 | case 0: | |
5198 | error = zap_length(os, object, name, &zl_wsize, &zl_wc); | |
5199 | if (error == 0) { | |
5200 | ASSERT3U(wsize, ==, zl_wsize); | |
5201 | ASSERT3U(wc, ==, zl_wc); | |
5202 | } else { | |
5203 | ASSERT3U(error, ==, ENOENT); | |
5204 | } | |
5205 | break; | |
5206 | ||
5207 | case 1: | |
5208 | error = zap_lookup(os, object, name, wsize, wc, data); | |
5209 | if (error == 0) { | |
5210 | if (data == string_value && | |
5211 | bcmp(name, data, namelen) != 0) | |
5212 | fatal(0, "name '%s' != val '%s' len %d", | |
5213 | name, data, namelen); | |
5214 | } else { | |
5215 | ASSERT3U(error, ==, ENOENT); | |
5216 | } | |
5217 | break; | |
5218 | ||
5219 | case 2: | |
5220 | error = zap_add(os, object, name, wsize, wc, data, tx); | |
5221 | ASSERT(error == 0 || error == EEXIST); | |
5222 | break; | |
5223 | ||
5224 | case 3: | |
5225 | VERIFY(zap_update(os, object, name, wsize, wc, data, tx) == 0); | |
5226 | break; | |
5227 | ||
5228 | case 4: | |
5229 | error = zap_remove(os, object, name, tx); | |
5230 | ASSERT(error == 0 || error == ENOENT); | |
5231 | break; | |
5232 | } | |
5233 | ||
5234 | if (tx != NULL) | |
5235 | dmu_tx_commit(tx); | |
40b84e7a | 5236 | |
d1d7e268 | 5237 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f BB |
5238 | } |
5239 | ||
428870ff BB |
5240 | /* |
5241 | * Commit callback data. | |
5242 | */ | |
5243 | typedef struct ztest_cb_data { | |
5244 | list_node_t zcd_node; | |
5245 | uint64_t zcd_txg; | |
5246 | int zcd_expected_err; | |
5247 | boolean_t zcd_added; | |
5248 | boolean_t zcd_called; | |
5249 | spa_t *zcd_spa; | |
5250 | } ztest_cb_data_t; | |
5251 | ||
5252 | /* This is the actual commit callback function */ | |
5253 | static void | |
5254 | ztest_commit_callback(void *arg, int error) | |
5255 | { | |
5256 | ztest_cb_data_t *data = arg; | |
5257 | uint64_t synced_txg; | |
5258 | ||
5259 | VERIFY(data != NULL); | |
5260 | VERIFY3S(data->zcd_expected_err, ==, error); | |
5261 | VERIFY(!data->zcd_called); | |
5262 | ||
5263 | synced_txg = spa_last_synced_txg(data->zcd_spa); | |
5264 | if (data->zcd_txg > synced_txg) | |
5265 | fatal(0, "commit callback of txg %" PRIu64 " called prematurely" | |
5266 | ", last synced txg = %" PRIu64 "\n", data->zcd_txg, | |
5267 | synced_txg); | |
5268 | ||
5269 | data->zcd_called = B_TRUE; | |
5270 | ||
5271 | if (error == ECANCELED) { | |
c99c9001 | 5272 | ASSERT0(data->zcd_txg); |
428870ff BB |
5273 | ASSERT(!data->zcd_added); |
5274 | ||
5275 | /* | |
5276 | * The private callback data should be destroyed here, but | |
5277 | * since we are going to check the zcd_called field after | |
5278 | * dmu_tx_abort(), we will destroy it there. | |
5279 | */ | |
5280 | return; | |
5281 | } | |
5282 | ||
090ff092 | 5283 | ASSERT(data->zcd_added); |
428870ff BB |
5284 | ASSERT3U(data->zcd_txg, !=, 0); |
5285 | ||
1e33ac1e | 5286 | (void) mutex_enter(&zcl.zcl_callbacks_lock); |
090ff092 RC |
5287 | |
5288 | /* See if this cb was called more quickly */ | |
5289 | if ((synced_txg - data->zcd_txg) < zc_min_txg_delay) | |
5290 | zc_min_txg_delay = synced_txg - data->zcd_txg; | |
5291 | ||
5292 | /* Remove our callback from the list */ | |
428870ff | 5293 | list_remove(&zcl.zcl_callbacks, data); |
090ff092 | 5294 | |
1e33ac1e | 5295 | (void) mutex_exit(&zcl.zcl_callbacks_lock); |
428870ff | 5296 | |
428870ff BB |
5297 | umem_free(data, sizeof (ztest_cb_data_t)); |
5298 | } | |
5299 | ||
5300 | /* Allocate and initialize callback data structure */ | |
5301 | static ztest_cb_data_t * | |
5302 | ztest_create_cb_data(objset_t *os, uint64_t txg) | |
5303 | { | |
5304 | ztest_cb_data_t *cb_data; | |
5305 | ||
5306 | cb_data = umem_zalloc(sizeof (ztest_cb_data_t), UMEM_NOFAIL); | |
5307 | ||
5308 | cb_data->zcd_txg = txg; | |
5309 | cb_data->zcd_spa = dmu_objset_spa(os); | |
1e33ac1e | 5310 | list_link_init(&cb_data->zcd_node); |
428870ff BB |
5311 | |
5312 | return (cb_data); | |
5313 | } | |
5314 | ||
428870ff BB |
5315 | /* |
5316 | * Commit callback test. | |
5317 | */ | |
34dc7c2f | 5318 | void |
428870ff BB |
5319 | ztest_dmu_commit_callbacks(ztest_ds_t *zd, uint64_t id) |
5320 | { | |
5321 | objset_t *os = zd->zd_os; | |
40b84e7a | 5322 | ztest_od_t *od; |
428870ff BB |
5323 | dmu_tx_t *tx; |
5324 | ztest_cb_data_t *cb_data[3], *tmp_cb; | |
5325 | uint64_t old_txg, txg; | |
090ff092 | 5326 | int i, error = 0; |
428870ff | 5327 | |
d1d7e268 | 5328 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 5329 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0); |
428870ff | 5330 | |
40b84e7a | 5331 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
d1d7e268 | 5332 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 5333 | return; |
40b84e7a | 5334 | } |
428870ff BB |
5335 | |
5336 | tx = dmu_tx_create(os); | |
5337 | ||
5338 | cb_data[0] = ztest_create_cb_data(os, 0); | |
5339 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[0]); | |
5340 | ||
40b84e7a | 5341 | dmu_tx_hold_write(tx, od->od_object, 0, sizeof (uint64_t)); |
428870ff BB |
5342 | |
5343 | /* Every once in a while, abort the transaction on purpose */ | |
5344 | if (ztest_random(100) == 0) | |
5345 | error = -1; | |
5346 | ||
5347 | if (!error) | |
5348 | error = dmu_tx_assign(tx, TXG_NOWAIT); | |
5349 | ||
5350 | txg = error ? 0 : dmu_tx_get_txg(tx); | |
5351 | ||
5352 | cb_data[0]->zcd_txg = txg; | |
5353 | cb_data[1] = ztest_create_cb_data(os, txg); | |
5354 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[1]); | |
5355 | ||
5356 | if (error) { | |
5357 | /* | |
5358 | * It's not a strict requirement to call the registered | |
5359 | * callbacks from inside dmu_tx_abort(), but that's what | |
5360 | * it's supposed to happen in the current implementation | |
5361 | * so we will check for that. | |
5362 | */ | |
5363 | for (i = 0; i < 2; i++) { | |
5364 | cb_data[i]->zcd_expected_err = ECANCELED; | |
5365 | VERIFY(!cb_data[i]->zcd_called); | |
5366 | } | |
5367 | ||
5368 | dmu_tx_abort(tx); | |
5369 | ||
5370 | for (i = 0; i < 2; i++) { | |
5371 | VERIFY(cb_data[i]->zcd_called); | |
5372 | umem_free(cb_data[i], sizeof (ztest_cb_data_t)); | |
5373 | } | |
5374 | ||
d1d7e268 | 5375 | umem_free(od, sizeof (ztest_od_t)); |
428870ff BB |
5376 | return; |
5377 | } | |
5378 | ||
5379 | cb_data[2] = ztest_create_cb_data(os, txg); | |
5380 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[2]); | |
5381 | ||
5382 | /* | |
5383 | * Read existing data to make sure there isn't a future leak. | |
5384 | */ | |
40b84e7a | 5385 | VERIFY(0 == dmu_read(os, od->od_object, 0, sizeof (uint64_t), |
428870ff BB |
5386 | &old_txg, DMU_READ_PREFETCH)); |
5387 | ||
5388 | if (old_txg > txg) | |
5389 | fatal(0, "future leak: got %" PRIu64 ", open txg is %" PRIu64, | |
5390 | old_txg, txg); | |
5391 | ||
40b84e7a | 5392 | dmu_write(os, od->od_object, 0, sizeof (uint64_t), &txg, tx); |
428870ff | 5393 | |
1e33ac1e | 5394 | (void) mutex_enter(&zcl.zcl_callbacks_lock); |
428870ff BB |
5395 | |
5396 | /* | |
5397 | * Since commit callbacks don't have any ordering requirement and since | |
5398 | * it is theoretically possible for a commit callback to be called | |
5399 | * after an arbitrary amount of time has elapsed since its txg has been | |
5400 | * synced, it is difficult to reliably determine whether a commit | |
5401 | * callback hasn't been called due to high load or due to a flawed | |
5402 | * implementation. | |
5403 | * | |
5404 | * In practice, we will assume that if after a certain number of txgs a | |
5405 | * commit callback hasn't been called, then most likely there's an | |
5406 | * implementation bug.. | |
5407 | */ | |
5408 | tmp_cb = list_head(&zcl.zcl_callbacks); | |
5409 | if (tmp_cb != NULL && | |
090ff092 | 5410 | tmp_cb->zcd_txg + ZTEST_COMMIT_CB_THRESH < txg) { |
428870ff BB |
5411 | fatal(0, "Commit callback threshold exceeded, oldest txg: %" |
5412 | PRIu64 ", open txg: %" PRIu64 "\n", tmp_cb->zcd_txg, txg); | |
5413 | } | |
5414 | ||
5415 | /* | |
5416 | * Let's find the place to insert our callbacks. | |
5417 | * | |
5418 | * Even though the list is ordered by txg, it is possible for the | |
5419 | * insertion point to not be the end because our txg may already be | |
5420 | * quiescing at this point and other callbacks in the open txg | |
5421 | * (from other objsets) may have sneaked in. | |
5422 | */ | |
5423 | tmp_cb = list_tail(&zcl.zcl_callbacks); | |
5424 | while (tmp_cb != NULL && tmp_cb->zcd_txg > txg) | |
5425 | tmp_cb = list_prev(&zcl.zcl_callbacks, tmp_cb); | |
5426 | ||
5427 | /* Add the 3 callbacks to the list */ | |
5428 | for (i = 0; i < 3; i++) { | |
5429 | if (tmp_cb == NULL) | |
5430 | list_insert_head(&zcl.zcl_callbacks, cb_data[i]); | |
5431 | else | |
5432 | list_insert_after(&zcl.zcl_callbacks, tmp_cb, | |
5433 | cb_data[i]); | |
5434 | ||
5435 | cb_data[i]->zcd_added = B_TRUE; | |
5436 | VERIFY(!cb_data[i]->zcd_called); | |
5437 | ||
5438 | tmp_cb = cb_data[i]; | |
5439 | } | |
5440 | ||
090ff092 RC |
5441 | zc_cb_counter += 3; |
5442 | ||
1e33ac1e | 5443 | (void) mutex_exit(&zcl.zcl_callbacks_lock); |
428870ff BB |
5444 | |
5445 | dmu_tx_commit(tx); | |
40b84e7a | 5446 | |
d1d7e268 | 5447 | umem_free(od, sizeof (ztest_od_t)); |
428870ff BB |
5448 | } |
5449 | ||
50c957f7 NB |
5450 | /* |
5451 | * Visit each object in the dataset. Verify that its properties | |
5452 | * are consistent what was stored in the block tag when it was created, | |
5453 | * and that its unused bonus buffer space has not been overwritten. | |
5454 | */ | |
817b1b6e | 5455 | /* ARGSUSED */ |
50c957f7 NB |
5456 | void |
5457 | ztest_verify_dnode_bt(ztest_ds_t *zd, uint64_t id) | |
5458 | { | |
5459 | objset_t *os = zd->zd_os; | |
5460 | uint64_t obj; | |
5461 | int err = 0; | |
5462 | ||
5463 | for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) { | |
5464 | ztest_block_tag_t *bt = NULL; | |
5465 | dmu_object_info_t doi; | |
5466 | dmu_buf_t *db; | |
5467 | ||
06401e42 BB |
5468 | ztest_object_lock(zd, obj, RL_READER); |
5469 | if (dmu_bonus_hold(os, obj, FTAG, &db) != 0) { | |
5470 | ztest_object_unlock(zd, obj); | |
50c957f7 | 5471 | continue; |
06401e42 | 5472 | } |
50c957f7 NB |
5473 | |
5474 | dmu_object_info_from_db(db, &doi); | |
5475 | if (doi.doi_bonus_size >= sizeof (*bt)) | |
5476 | bt = ztest_bt_bonus(db); | |
5477 | ||
5478 | if (bt && bt->bt_magic == BT_MAGIC) { | |
5479 | ztest_bt_verify(bt, os, obj, doi.doi_dnodesize, | |
5480 | bt->bt_offset, bt->bt_gen, bt->bt_txg, | |
5481 | bt->bt_crtxg); | |
5482 | ztest_verify_unused_bonus(db, bt, obj, os, bt->bt_gen); | |
5483 | } | |
5484 | ||
5485 | dmu_buf_rele(db, FTAG); | |
06401e42 | 5486 | ztest_object_unlock(zd, obj); |
50c957f7 NB |
5487 | } |
5488 | } | |
5489 | ||
428870ff BB |
5490 | /* ARGSUSED */ |
5491 | void | |
5492 | ztest_dsl_prop_get_set(ztest_ds_t *zd, uint64_t id) | |
5493 | { | |
5494 | zfs_prop_t proplist[] = { | |
5495 | ZFS_PROP_CHECKSUM, | |
5496 | ZFS_PROP_COMPRESSION, | |
5497 | ZFS_PROP_COPIES, | |
5498 | ZFS_PROP_DEDUP | |
5499 | }; | |
d6320ddb | 5500 | int p; |
428870ff | 5501 | |
2d9142c9 | 5502 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
428870ff | 5503 | |
d6320ddb | 5504 | for (p = 0; p < sizeof (proplist) / sizeof (proplist[0]); p++) |
428870ff BB |
5505 | (void) ztest_dsl_prop_set_uint64(zd->zd_name, proplist[p], |
5506 | ztest_random_dsl_prop(proplist[p]), (int)ztest_random(2)); | |
5507 | ||
76d52067 BB |
5508 | VERIFY0(ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_RECORDSIZE, |
5509 | ztest_random_blocksize(), (int)ztest_random(2))); | |
5510 | ||
2d9142c9 | 5511 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
428870ff BB |
5512 | } |
5513 | ||
a1d477c2 MA |
5514 | /* ARGSUSED */ |
5515 | void | |
5516 | ztest_remap_blocks(ztest_ds_t *zd, uint64_t id) | |
5517 | { | |
2d9142c9 | 5518 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
a1d477c2 MA |
5519 | |
5520 | int error = dmu_objset_remap_indirects(zd->zd_name); | |
5521 | if (error == ENOSPC) | |
5522 | error = 0; | |
5523 | ASSERT0(error); | |
5524 | ||
2d9142c9 | 5525 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
a1d477c2 MA |
5526 | } |
5527 | ||
428870ff BB |
5528 | /* ARGSUSED */ |
5529 | void | |
5530 | ztest_spa_prop_get_set(ztest_ds_t *zd, uint64_t id) | |
5531 | { | |
428870ff BB |
5532 | nvlist_t *props = NULL; |
5533 | ||
2d9142c9 | 5534 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
428870ff | 5535 | |
c242c188 | 5536 | (void) ztest_spa_prop_set_uint64(ZPOOL_PROP_DEDUPDITTO, |
428870ff BB |
5537 | ZIO_DEDUPDITTO_MIN + ztest_random(ZIO_DEDUPDITTO_MIN)); |
5538 | ||
c99c9001 | 5539 | VERIFY0(spa_prop_get(ztest_spa, &props)); |
428870ff | 5540 | |
c242c188 | 5541 | if (ztest_opts.zo_verbose >= 6) |
428870ff BB |
5542 | dump_nvlist(props, 4); |
5543 | ||
5544 | nvlist_free(props); | |
5545 | ||
2d9142c9 | 5546 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
428870ff BB |
5547 | } |
5548 | ||
13fe0198 MA |
5549 | static int |
5550 | user_release_one(const char *snapname, const char *holdname) | |
5551 | { | |
5552 | nvlist_t *snaps, *holds; | |
5553 | int error; | |
5554 | ||
5555 | snaps = fnvlist_alloc(); | |
5556 | holds = fnvlist_alloc(); | |
5557 | fnvlist_add_boolean(holds, holdname); | |
5558 | fnvlist_add_nvlist(snaps, snapname, holds); | |
5559 | fnvlist_free(holds); | |
5560 | error = dsl_dataset_user_release(snaps, NULL); | |
5561 | fnvlist_free(snaps); | |
5562 | return (error); | |
5563 | } | |
5564 | ||
428870ff BB |
5565 | /* |
5566 | * Test snapshot hold/release and deferred destroy. | |
5567 | */ | |
5568 | void | |
5569 | ztest_dmu_snapshot_hold(ztest_ds_t *zd, uint64_t id) | |
34dc7c2f | 5570 | { |
34dc7c2f | 5571 | int error; |
428870ff BB |
5572 | objset_t *os = zd->zd_os; |
5573 | objset_t *origin; | |
5574 | char snapname[100]; | |
5575 | char fullname[100]; | |
5576 | char clonename[100]; | |
5577 | char tag[100]; | |
eca7b760 | 5578 | char osname[ZFS_MAX_DATASET_NAME_LEN]; |
13fe0198 | 5579 | nvlist_t *holds; |
34dc7c2f | 5580 | |
2d9142c9 | 5581 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
34dc7c2f BB |
5582 | |
5583 | dmu_objset_name(os, osname); | |
5584 | ||
d1d7e268 MK |
5585 | (void) snprintf(snapname, sizeof (snapname), "sh1_%llu", |
5586 | (u_longlong_t)id); | |
13fe0198 MA |
5587 | (void) snprintf(fullname, sizeof (fullname), "%s@%s", osname, snapname); |
5588 | (void) snprintf(clonename, sizeof (clonename), | |
d1d7e268 MK |
5589 | "%s/ch1_%llu", osname, (u_longlong_t)id); |
5590 | (void) snprintf(tag, sizeof (tag), "tag_%llu", (u_longlong_t)id); | |
428870ff BB |
5591 | |
5592 | /* | |
5593 | * Clean up from any previous run. | |
5594 | */ | |
13fe0198 MA |
5595 | error = dsl_destroy_head(clonename); |
5596 | if (error != ENOENT) | |
5597 | ASSERT0(error); | |
5598 | error = user_release_one(fullname, tag); | |
5599 | if (error != ESRCH && error != ENOENT) | |
5600 | ASSERT0(error); | |
5601 | error = dsl_destroy_snapshot(fullname, B_FALSE); | |
5602 | if (error != ENOENT) | |
5603 | ASSERT0(error); | |
428870ff BB |
5604 | |
5605 | /* | |
5606 | * Create snapshot, clone it, mark snap for deferred destroy, | |
5607 | * destroy clone, verify snap was also destroyed. | |
5608 | */ | |
6f1ffb06 | 5609 | error = dmu_objset_snapshot_one(osname, snapname); |
428870ff BB |
5610 | if (error) { |
5611 | if (error == ENOSPC) { | |
5612 | ztest_record_enospc("dmu_objset_snapshot"); | |
5613 | goto out; | |
34dc7c2f | 5614 | } |
428870ff BB |
5615 | fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error); |
5616 | } | |
34dc7c2f | 5617 | |
13fe0198 | 5618 | error = dmu_objset_clone(clonename, fullname); |
428870ff | 5619 | if (error) { |
34dc7c2f | 5620 | if (error == ENOSPC) { |
428870ff BB |
5621 | ztest_record_enospc("dmu_objset_clone"); |
5622 | goto out; | |
34dc7c2f | 5623 | } |
428870ff BB |
5624 | fatal(0, "dmu_objset_clone(%s) = %d", clonename, error); |
5625 | } | |
34dc7c2f | 5626 | |
13fe0198 | 5627 | error = dsl_destroy_snapshot(fullname, B_TRUE); |
428870ff | 5628 | if (error) { |
13fe0198 | 5629 | fatal(0, "dsl_destroy_snapshot(%s, B_TRUE) = %d", |
428870ff BB |
5630 | fullname, error); |
5631 | } | |
34dc7c2f | 5632 | |
13fe0198 | 5633 | error = dsl_destroy_head(clonename); |
428870ff | 5634 | if (error) |
13fe0198 | 5635 | fatal(0, "dsl_destroy_head(%s) = %d", clonename, error); |
34dc7c2f | 5636 | |
428870ff BB |
5637 | error = dmu_objset_hold(fullname, FTAG, &origin); |
5638 | if (error != ENOENT) | |
5639 | fatal(0, "dmu_objset_hold(%s) = %d", fullname, error); | |
34dc7c2f | 5640 | |
428870ff BB |
5641 | /* |
5642 | * Create snapshot, add temporary hold, verify that we can't | |
5643 | * destroy a held snapshot, mark for deferred destroy, | |
5644 | * release hold, verify snapshot was destroyed. | |
5645 | */ | |
6f1ffb06 | 5646 | error = dmu_objset_snapshot_one(osname, snapname); |
428870ff BB |
5647 | if (error) { |
5648 | if (error == ENOSPC) { | |
5649 | ztest_record_enospc("dmu_objset_snapshot"); | |
5650 | goto out; | |
34dc7c2f | 5651 | } |
428870ff BB |
5652 | fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error); |
5653 | } | |
5654 | ||
13fe0198 MA |
5655 | holds = fnvlist_alloc(); |
5656 | fnvlist_add_string(holds, fullname, tag); | |
5657 | error = dsl_dataset_user_hold(holds, 0, NULL); | |
5658 | fnvlist_free(holds); | |
5659 | ||
9b67f605 MA |
5660 | if (error == ENOSPC) { |
5661 | ztest_record_enospc("dsl_dataset_user_hold"); | |
5662 | goto out; | |
5663 | } else if (error) { | |
5664 | fatal(0, "dsl_dataset_user_hold(%s, %s) = %u", | |
5665 | fullname, tag, error); | |
5666 | } | |
428870ff | 5667 | |
13fe0198 | 5668 | error = dsl_destroy_snapshot(fullname, B_FALSE); |
428870ff | 5669 | if (error != EBUSY) { |
13fe0198 | 5670 | fatal(0, "dsl_destroy_snapshot(%s, B_FALSE) = %d", |
428870ff BB |
5671 | fullname, error); |
5672 | } | |
5673 | ||
13fe0198 | 5674 | error = dsl_destroy_snapshot(fullname, B_TRUE); |
428870ff | 5675 | if (error) { |
13fe0198 | 5676 | fatal(0, "dsl_destroy_snapshot(%s, B_TRUE) = %d", |
428870ff | 5677 | fullname, error); |
34dc7c2f BB |
5678 | } |
5679 | ||
13fe0198 | 5680 | error = user_release_one(fullname, tag); |
428870ff | 5681 | if (error) |
95fd54a1 | 5682 | fatal(0, "user_release_one(%s, %s) = %d", fullname, tag, error); |
428870ff | 5683 | |
13fe0198 | 5684 | VERIFY3U(dmu_objset_hold(fullname, FTAG, &origin), ==, ENOENT); |
428870ff BB |
5685 | |
5686 | out: | |
2d9142c9 | 5687 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
34dc7c2f BB |
5688 | } |
5689 | ||
34dc7c2f BB |
5690 | /* |
5691 | * Inject random faults into the on-disk data. | |
5692 | */ | |
428870ff | 5693 | /* ARGSUSED */ |
34dc7c2f | 5694 | void |
428870ff | 5695 | ztest_fault_inject(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 5696 | { |
428870ff | 5697 | ztest_shared_t *zs = ztest_shared; |
c242c188 | 5698 | spa_t *spa = ztest_spa; |
34dc7c2f BB |
5699 | int fd; |
5700 | uint64_t offset; | |
428870ff | 5701 | uint64_t leaves; |
c5b3a7bb | 5702 | uint64_t bad = 0x1990c0ffeedecadeull; |
34dc7c2f | 5703 | uint64_t top, leaf; |
40b84e7a BB |
5704 | char *path0; |
5705 | char *pathrand; | |
34dc7c2f | 5706 | size_t fsize; |
2014c09f | 5707 | int bshift = SPA_MAXBLOCKSHIFT + 2; |
34dc7c2f | 5708 | int iters = 1000; |
428870ff BB |
5709 | int maxfaults; |
5710 | int mirror_save; | |
b128c09f | 5711 | vdev_t *vd0 = NULL; |
34dc7c2f | 5712 | uint64_t guid0 = 0; |
428870ff BB |
5713 | boolean_t islog = B_FALSE; |
5714 | ||
40b84e7a BB |
5715 | path0 = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
5716 | pathrand = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
5717 | ||
c242c188 | 5718 | mutex_enter(&ztest_vdev_lock); |
9e052db4 MA |
5719 | |
5720 | /* | |
5721 | * Device removal is in progress, fault injection must be disabled | |
5722 | * until it completes and the pool is scrubbed. The fault injection | |
5723 | * strategy for damaging blocks does not take in to account evacuated | |
5724 | * blocks which may have already been damaged. | |
5725 | */ | |
5726 | if (ztest_device_removal_active) { | |
5727 | mutex_exit(&ztest_vdev_lock); | |
5728 | goto out; | |
5729 | } | |
5730 | ||
60b82074 | 5731 | maxfaults = MAXFAULTS(zs); |
c242c188 | 5732 | leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raidz; |
428870ff | 5733 | mirror_save = zs->zs_mirrors; |
c242c188 | 5734 | mutex_exit(&ztest_vdev_lock); |
34dc7c2f | 5735 | |
b128c09f | 5736 | ASSERT(leaves >= 1); |
34dc7c2f | 5737 | |
621dd7bb GW |
5738 | /* |
5739 | * Grab the name lock as reader. There are some operations | |
5740 | * which don't like to have their vdevs changed while | |
5741 | * they are in progress (i.e. spa_change_guid). Those | |
5742 | * operations will have grabbed the name lock as writer. | |
5743 | */ | |
2d9142c9 | 5744 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
621dd7bb | 5745 | |
34dc7c2f | 5746 | /* |
b128c09f | 5747 | * We need SCL_STATE here because we're going to look at vd0->vdev_tsd. |
34dc7c2f | 5748 | */ |
b128c09f | 5749 | spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); |
34dc7c2f | 5750 | |
b128c09f BB |
5751 | if (ztest_random(2) == 0) { |
5752 | /* | |
428870ff | 5753 | * Inject errors on a normal data device or slog device. |
b128c09f | 5754 | */ |
428870ff BB |
5755 | top = ztest_random_vdev_top(spa, B_TRUE); |
5756 | leaf = ztest_random(leaves) + zs->zs_splits; | |
34dc7c2f | 5757 | |
b128c09f BB |
5758 | /* |
5759 | * Generate paths to the first leaf in this top-level vdev, | |
5760 | * and to the random leaf we selected. We'll induce transient | |
5761 | * write failures and random online/offline activity on leaf 0, | |
5762 | * and we'll write random garbage to the randomly chosen leaf. | |
5763 | */ | |
6aec1cd5 | 5764 | (void) snprintf(path0, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
5765 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
5766 | top * leaves + zs->zs_splits); | |
6aec1cd5 | 5767 | (void) snprintf(pathrand, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
5768 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
5769 | top * leaves + leaf); | |
34dc7c2f | 5770 | |
b128c09f | 5771 | vd0 = vdev_lookup_by_path(spa->spa_root_vdev, path0); |
428870ff BB |
5772 | if (vd0 != NULL && vd0->vdev_top->vdev_islog) |
5773 | islog = B_TRUE; | |
5774 | ||
621dd7bb GW |
5775 | /* |
5776 | * If the top-level vdev needs to be resilvered | |
5777 | * then we only allow faults on the device that is | |
5778 | * resilvering. | |
5779 | */ | |
5780 | if (vd0 != NULL && maxfaults != 1 && | |
5781 | (!vdev_resilver_needed(vd0->vdev_top, NULL, NULL) || | |
5d1f7fb6 | 5782 | vd0->vdev_resilver_txg != 0)) { |
b128c09f BB |
5783 | /* |
5784 | * Make vd0 explicitly claim to be unreadable, | |
5785 | * or unwriteable, or reach behind its back | |
5786 | * and close the underlying fd. We can do this if | |
5787 | * maxfaults == 0 because we'll fail and reexecute, | |
5788 | * and we can do it if maxfaults >= 2 because we'll | |
5789 | * have enough redundancy. If maxfaults == 1, the | |
5790 | * combination of this with injection of random data | |
5791 | * corruption below exceeds the pool's fault tolerance. | |
5792 | */ | |
5793 | vdev_file_t *vf = vd0->vdev_tsd; | |
5794 | ||
a1d477c2 MA |
5795 | zfs_dbgmsg("injecting fault to vdev %llu; maxfaults=%d", |
5796 | (long long)vd0->vdev_id, (int)maxfaults); | |
5797 | ||
b128c09f BB |
5798 | if (vf != NULL && ztest_random(3) == 0) { |
5799 | (void) close(vf->vf_vnode->v_fd); | |
5800 | vf->vf_vnode->v_fd = -1; | |
5801 | } else if (ztest_random(2) == 0) { | |
5802 | vd0->vdev_cant_read = B_TRUE; | |
5803 | } else { | |
5804 | vd0->vdev_cant_write = B_TRUE; | |
5805 | } | |
5806 | guid0 = vd0->vdev_guid; | |
5807 | } | |
5808 | } else { | |
5809 | /* | |
5810 | * Inject errors on an l2cache device. | |
5811 | */ | |
5812 | spa_aux_vdev_t *sav = &spa->spa_l2cache; | |
34dc7c2f | 5813 | |
b128c09f BB |
5814 | if (sav->sav_count == 0) { |
5815 | spa_config_exit(spa, SCL_STATE, FTAG); | |
2d9142c9 | 5816 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
40b84e7a | 5817 | goto out; |
b128c09f BB |
5818 | } |
5819 | vd0 = sav->sav_vdevs[ztest_random(sav->sav_count)]; | |
34dc7c2f | 5820 | guid0 = vd0->vdev_guid; |
b128c09f BB |
5821 | (void) strcpy(path0, vd0->vdev_path); |
5822 | (void) strcpy(pathrand, vd0->vdev_path); | |
5823 | ||
5824 | leaf = 0; | |
5825 | leaves = 1; | |
5826 | maxfaults = INT_MAX; /* no limit on cache devices */ | |
34dc7c2f BB |
5827 | } |
5828 | ||
b128c09f | 5829 | spa_config_exit(spa, SCL_STATE, FTAG); |
2d9142c9 | 5830 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
b128c09f | 5831 | |
34dc7c2f | 5832 | /* |
428870ff BB |
5833 | * If we can tolerate two or more faults, or we're dealing |
5834 | * with a slog, randomly online/offline vd0. | |
34dc7c2f | 5835 | */ |
428870ff | 5836 | if ((maxfaults >= 2 || islog) && guid0 != 0) { |
fb5f0bc8 BB |
5837 | if (ztest_random(10) < 6) { |
5838 | int flags = (ztest_random(2) == 0 ? | |
5839 | ZFS_OFFLINE_TEMPORARY : 0); | |
428870ff BB |
5840 | |
5841 | /* | |
5842 | * We have to grab the zs_name_lock as writer to | |
5843 | * prevent a race between offlining a slog and | |
5844 | * destroying a dataset. Offlining the slog will | |
5845 | * grab a reference on the dataset which may cause | |
13fe0198 | 5846 | * dsl_destroy_head() to fail with EBUSY thus |
428870ff BB |
5847 | * leaving the dataset in an inconsistent state. |
5848 | */ | |
5849 | if (islog) | |
2d9142c9 | 5850 | (void) pthread_rwlock_wrlock(&ztest_name_lock); |
428870ff | 5851 | |
fb5f0bc8 | 5852 | VERIFY(vdev_offline(spa, guid0, flags) != EBUSY); |
428870ff BB |
5853 | |
5854 | if (islog) | |
2d9142c9 | 5855 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
fb5f0bc8 | 5856 | } else { |
1eb5bfa3 GW |
5857 | /* |
5858 | * Ideally we would like to be able to randomly | |
5859 | * call vdev_[on|off]line without holding locks | |
5860 | * to force unpredictable failures but the side | |
5861 | * effects of vdev_[on|off]line prevent us from | |
5862 | * doing so. We grab the ztest_vdev_lock here to | |
5863 | * prevent a race between injection testing and | |
5864 | * aux_vdev removal. | |
5865 | */ | |
5866 | mutex_enter(&ztest_vdev_lock); | |
fb5f0bc8 | 5867 | (void) vdev_online(spa, guid0, 0, NULL); |
1eb5bfa3 | 5868 | mutex_exit(&ztest_vdev_lock); |
fb5f0bc8 | 5869 | } |
34dc7c2f BB |
5870 | } |
5871 | ||
428870ff | 5872 | if (maxfaults == 0) |
40b84e7a | 5873 | goto out; |
428870ff | 5874 | |
34dc7c2f BB |
5875 | /* |
5876 | * We have at least single-fault tolerance, so inject data corruption. | |
5877 | */ | |
5878 | fd = open(pathrand, O_RDWR); | |
5879 | ||
d2734cce | 5880 | if (fd == -1) /* we hit a gap in the device namespace */ |
40b84e7a | 5881 | goto out; |
34dc7c2f BB |
5882 | |
5883 | fsize = lseek(fd, 0, SEEK_END); | |
5884 | ||
5885 | while (--iters != 0) { | |
91d88843 MA |
5886 | /* |
5887 | * The offset must be chosen carefully to ensure that | |
5888 | * we do not inject a given logical block with errors | |
5889 | * on two different leaf devices, because ZFS can not | |
5890 | * tolerate that (if maxfaults==1). | |
5891 | * | |
5892 | * We divide each leaf into chunks of size | |
5893 | * (# leaves * SPA_MAXBLOCKSIZE * 4). Within each chunk | |
5894 | * there is a series of ranges to which we can inject errors. | |
5895 | * Each range can accept errors on only a single leaf vdev. | |
5896 | * The error injection ranges are separated by ranges | |
5897 | * which we will not inject errors on any device (DMZs). | |
5898 | * Each DMZ must be large enough such that a single block | |
5899 | * can not straddle it, so that a single block can not be | |
5900 | * a target in two different injection ranges (on different | |
5901 | * leaf vdevs). | |
5902 | * | |
5903 | * For example, with 3 leaves, each chunk looks like: | |
5904 | * 0 to 32M: injection range for leaf 0 | |
5905 | * 32M to 64M: DMZ - no injection allowed | |
5906 | * 64M to 96M: injection range for leaf 1 | |
5907 | * 96M to 128M: DMZ - no injection allowed | |
5908 | * 128M to 160M: injection range for leaf 2 | |
5909 | * 160M to 192M: DMZ - no injection allowed | |
5910 | */ | |
34dc7c2f BB |
5911 | offset = ztest_random(fsize / (leaves << bshift)) * |
5912 | (leaves << bshift) + (leaf << bshift) + | |
5913 | (ztest_random(1ULL << (bshift - 1)) & -8ULL); | |
5914 | ||
2014c09f GM |
5915 | /* |
5916 | * Only allow damage to the labels at one end of the vdev. | |
5917 | * | |
5918 | * If all labels are damaged, the device will be totally | |
5919 | * inaccessible, which will result in loss of data, | |
5920 | * because we also damage (parts of) the other side of | |
5921 | * the mirror/raidz. | |
5922 | * | |
5923 | * Additionally, we will always have both an even and an | |
5924 | * odd label, so that we can handle crashes in the | |
5925 | * middle of vdev_config_sync(). | |
5926 | */ | |
5927 | if ((leaf & 1) == 0 && offset < VDEV_LABEL_START_SIZE) | |
5928 | continue; | |
5929 | ||
5930 | /* | |
5931 | * The two end labels are stored at the "end" of the disk, but | |
5932 | * the end of the disk (vdev_psize) is aligned to | |
5933 | * sizeof (vdev_label_t). | |
5934 | */ | |
5935 | uint64_t psize = P2ALIGN(fsize, sizeof (vdev_label_t)); | |
5936 | if ((leaf & 1) == 1 && | |
5937 | offset + sizeof (bad) > psize - VDEV_LABEL_END_SIZE) | |
34dc7c2f BB |
5938 | continue; |
5939 | ||
c242c188 | 5940 | mutex_enter(&ztest_vdev_lock); |
428870ff | 5941 | if (mirror_save != zs->zs_mirrors) { |
c242c188 | 5942 | mutex_exit(&ztest_vdev_lock); |
428870ff | 5943 | (void) close(fd); |
40b84e7a | 5944 | goto out; |
428870ff | 5945 | } |
34dc7c2f BB |
5946 | |
5947 | if (pwrite(fd, &bad, sizeof (bad), offset) != sizeof (bad)) | |
5948 | fatal(1, "can't inject bad word at 0x%llx in %s", | |
5949 | offset, pathrand); | |
428870ff | 5950 | |
c242c188 | 5951 | mutex_exit(&ztest_vdev_lock); |
428870ff | 5952 | |
c242c188 | 5953 | if (ztest_opts.zo_verbose >= 7) |
428870ff BB |
5954 | (void) printf("injected bad word into %s," |
5955 | " offset 0x%llx\n", pathrand, (u_longlong_t)offset); | |
34dc7c2f BB |
5956 | } |
5957 | ||
5958 | (void) close(fd); | |
40b84e7a BB |
5959 | out: |
5960 | umem_free(path0, MAXPATHLEN); | |
5961 | umem_free(pathrand, MAXPATHLEN); | |
34dc7c2f BB |
5962 | } |
5963 | ||
5964 | /* | |
428870ff | 5965 | * Verify that DDT repair works as expected. |
34dc7c2f BB |
5966 | */ |
5967 | void | |
428870ff | 5968 | ztest_ddt_repair(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 5969 | { |
428870ff | 5970 | ztest_shared_t *zs = ztest_shared; |
c242c188 | 5971 | spa_t *spa = ztest_spa; |
428870ff | 5972 | objset_t *os = zd->zd_os; |
40b84e7a | 5973 | ztest_od_t *od; |
428870ff BB |
5974 | uint64_t object, blocksize, txg, pattern, psize; |
5975 | enum zio_checksum checksum = spa_dedup_checksum(spa); | |
5976 | dmu_buf_t *db; | |
5977 | dmu_tx_t *tx; | |
a6255b7f | 5978 | abd_t *abd; |
428870ff BB |
5979 | blkptr_t blk; |
5980 | int copies = 2 * ZIO_DEDUPDITTO_MIN; | |
d6320ddb | 5981 | int i; |
34dc7c2f | 5982 | |
428870ff BB |
5983 | blocksize = ztest_random_blocksize(); |
5984 | blocksize = MIN(blocksize, 2048); /* because we write so many */ | |
34dc7c2f | 5985 | |
d1d7e268 | 5986 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 5987 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0, 0); |
34dc7c2f | 5988 | |
40b84e7a | 5989 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
d1d7e268 | 5990 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 5991 | return; |
40b84e7a | 5992 | } |
34dc7c2f BB |
5993 | |
5994 | /* | |
428870ff BB |
5995 | * Take the name lock as writer to prevent anyone else from changing |
5996 | * the pool and dataset properies we need to maintain during this test. | |
34dc7c2f | 5997 | */ |
2d9142c9 | 5998 | (void) pthread_rwlock_wrlock(&ztest_name_lock); |
34dc7c2f | 5999 | |
428870ff BB |
6000 | if (ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_DEDUP, checksum, |
6001 | B_FALSE) != 0 || | |
6002 | ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_COPIES, 1, | |
6003 | B_FALSE) != 0) { | |
2d9142c9 | 6004 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
d1d7e268 | 6005 | umem_free(od, sizeof (ztest_od_t)); |
428870ff BB |
6006 | return; |
6007 | } | |
6008 | ||
546d32ca BB |
6009 | dmu_objset_stats_t dds; |
6010 | dsl_pool_config_enter(dmu_objset_pool(os), FTAG); | |
6011 | dmu_objset_fast_stat(os, &dds); | |
6012 | dsl_pool_config_exit(dmu_objset_pool(os), FTAG); | |
6013 | ||
428870ff BB |
6014 | object = od[0].od_object; |
6015 | blocksize = od[0].od_blocksize; | |
546d32ca | 6016 | pattern = zs->zs_guid ^ dds.dds_guid; |
428870ff BB |
6017 | |
6018 | ASSERT(object != 0); | |
6019 | ||
6020 | tx = dmu_tx_create(os); | |
6021 | dmu_tx_hold_write(tx, object, 0, copies * blocksize); | |
6022 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
6023 | if (txg == 0) { | |
2d9142c9 | 6024 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
d1d7e268 | 6025 | umem_free(od, sizeof (ztest_od_t)); |
428870ff BB |
6026 | return; |
6027 | } | |
34dc7c2f BB |
6028 | |
6029 | /* | |
428870ff | 6030 | * Write all the copies of our block. |
34dc7c2f | 6031 | */ |
d6320ddb | 6032 | for (i = 0; i < copies; i++) { |
428870ff | 6033 | uint64_t offset = i * blocksize; |
13fe0198 MA |
6034 | int error = dmu_buf_hold(os, object, offset, FTAG, &db, |
6035 | DMU_READ_NO_PREFETCH); | |
6036 | if (error != 0) { | |
6037 | fatal(B_FALSE, "dmu_buf_hold(%p, %llu, %llu) = %u", | |
6038 | os, (long long)object, (long long) offset, error); | |
6039 | } | |
428870ff BB |
6040 | ASSERT(db->db_offset == offset); |
6041 | ASSERT(db->db_size == blocksize); | |
6042 | ASSERT(ztest_pattern_match(db->db_data, db->db_size, pattern) || | |
6043 | ztest_pattern_match(db->db_data, db->db_size, 0ULL)); | |
6044 | dmu_buf_will_fill(db, tx); | |
6045 | ztest_pattern_set(db->db_data, db->db_size, pattern); | |
6046 | dmu_buf_rele(db, FTAG); | |
6047 | } | |
34dc7c2f | 6048 | |
428870ff BB |
6049 | dmu_tx_commit(tx); |
6050 | txg_wait_synced(spa_get_dsl(spa), txg); | |
34dc7c2f BB |
6051 | |
6052 | /* | |
428870ff | 6053 | * Find out what block we got. |
34dc7c2f | 6054 | */ |
03c6040b GW |
6055 | VERIFY0(dmu_buf_hold(os, object, 0, FTAG, &db, |
6056 | DMU_READ_NO_PREFETCH)); | |
428870ff BB |
6057 | blk = *((dmu_buf_impl_t *)db)->db_blkptr; |
6058 | dmu_buf_rele(db, FTAG); | |
34dc7c2f BB |
6059 | |
6060 | /* | |
428870ff | 6061 | * Damage the block. Dedup-ditto will save us when we read it later. |
34dc7c2f | 6062 | */ |
428870ff | 6063 | psize = BP_GET_PSIZE(&blk); |
a6255b7f DQ |
6064 | abd = abd_alloc_linear(psize, B_TRUE); |
6065 | ztest_pattern_set(abd_to_buf(abd), psize, ~pattern); | |
34dc7c2f | 6066 | |
428870ff | 6067 | (void) zio_wait(zio_rewrite(NULL, spa, 0, &blk, |
a6255b7f | 6068 | abd, psize, NULL, NULL, ZIO_PRIORITY_SYNC_WRITE, |
428870ff | 6069 | ZIO_FLAG_CANFAIL | ZIO_FLAG_INDUCE_DAMAGE, NULL)); |
34dc7c2f | 6070 | |
a6255b7f | 6071 | abd_free(abd); |
34dc7c2f | 6072 | |
2d9142c9 | 6073 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
d1d7e268 | 6074 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f BB |
6075 | } |
6076 | ||
34dc7c2f | 6077 | /* |
428870ff | 6078 | * Scrub the pool. |
34dc7c2f | 6079 | */ |
428870ff BB |
6080 | /* ARGSUSED */ |
6081 | void | |
6082 | ztest_scrub(ztest_ds_t *zd, uint64_t id) | |
34dc7c2f | 6083 | { |
c242c188 | 6084 | spa_t *spa = ztest_spa; |
34dc7c2f | 6085 | |
9e052db4 MA |
6086 | /* |
6087 | * Scrub in progress by device removal. | |
6088 | */ | |
6089 | if (ztest_device_removal_active) | |
6090 | return; | |
6091 | ||
428870ff BB |
6092 | (void) spa_scan(spa, POOL_SCAN_SCRUB); |
6093 | (void) poll(NULL, 0, 100); /* wait a moment, then force a restart */ | |
6094 | (void) spa_scan(spa, POOL_SCAN_SCRUB); | |
6095 | } | |
34dc7c2f | 6096 | |
3541dc6d GA |
6097 | /* |
6098 | * Change the guid for the pool. | |
6099 | */ | |
6100 | /* ARGSUSED */ | |
6101 | void | |
6102 | ztest_reguid(ztest_ds_t *zd, uint64_t id) | |
6103 | { | |
c242c188 | 6104 | spa_t *spa = ztest_spa; |
3541dc6d | 6105 | uint64_t orig, load; |
3bc7e0fb | 6106 | int error; |
3541dc6d | 6107 | |
b33d668d OF |
6108 | if (ztest_opts.zo_mmp_test) |
6109 | return; | |
6110 | ||
3541dc6d GA |
6111 | orig = spa_guid(spa); |
6112 | load = spa_load_guid(spa); | |
3bc7e0fb | 6113 | |
2d9142c9 | 6114 | (void) pthread_rwlock_wrlock(&ztest_name_lock); |
3bc7e0fb | 6115 | error = spa_change_guid(spa); |
2d9142c9 | 6116 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
3bc7e0fb GW |
6117 | |
6118 | if (error != 0) | |
3541dc6d GA |
6119 | return; |
6120 | ||
ea0b2538 | 6121 | if (ztest_opts.zo_verbose >= 4) { |
3541dc6d GA |
6122 | (void) printf("Changed guid old %llu -> %llu\n", |
6123 | (u_longlong_t)orig, (u_longlong_t)spa_guid(spa)); | |
6124 | } | |
6125 | ||
6126 | VERIFY3U(orig, !=, spa_guid(spa)); | |
6127 | VERIFY3U(load, ==, spa_load_guid(spa)); | |
6128 | } | |
6129 | ||
1eeb4562 JX |
6130 | void |
6131 | ztest_fletcher(ztest_ds_t *zd, uint64_t id) | |
6132 | { | |
6133 | hrtime_t end = gethrtime() + NANOSEC; | |
6134 | ||
6135 | while (gethrtime() <= end) { | |
6136 | int run_count = 100; | |
6137 | void *buf; | |
2fe36b0b | 6138 | struct abd *abd_data, *abd_meta; |
1eeb4562 JX |
6139 | uint32_t size; |
6140 | int *ptr; | |
6141 | int i; | |
6142 | zio_cksum_t zc_ref; | |
6143 | zio_cksum_t zc_ref_byteswap; | |
6144 | ||
6145 | size = ztest_random_blocksize(); | |
2fe36b0b | 6146 | |
1eeb4562 | 6147 | buf = umem_alloc(size, UMEM_NOFAIL); |
2fe36b0b DQ |
6148 | abd_data = abd_alloc(size, B_FALSE); |
6149 | abd_meta = abd_alloc(size, B_TRUE); | |
1eeb4562 JX |
6150 | |
6151 | for (i = 0, ptr = buf; i < size / sizeof (*ptr); i++, ptr++) | |
6152 | *ptr = ztest_random(UINT_MAX); | |
6153 | ||
2fe36b0b DQ |
6154 | abd_copy_from_buf_off(abd_data, buf, 0, size); |
6155 | abd_copy_from_buf_off(abd_meta, buf, 0, size); | |
6156 | ||
1eeb4562 | 6157 | VERIFY0(fletcher_4_impl_set("scalar")); |
3c67d83a TH |
6158 | fletcher_4_native(buf, size, NULL, &zc_ref); |
6159 | fletcher_4_byteswap(buf, size, NULL, &zc_ref_byteswap); | |
1eeb4562 JX |
6160 | |
6161 | VERIFY0(fletcher_4_impl_set("cycle")); | |
6162 | while (run_count-- > 0) { | |
6163 | zio_cksum_t zc; | |
6164 | zio_cksum_t zc_byteswap; | |
6165 | ||
3c67d83a TH |
6166 | fletcher_4_byteswap(buf, size, NULL, &zc_byteswap); |
6167 | fletcher_4_native(buf, size, NULL, &zc); | |
1eeb4562 JX |
6168 | |
6169 | VERIFY0(bcmp(&zc, &zc_ref, sizeof (zc))); | |
6170 | VERIFY0(bcmp(&zc_byteswap, &zc_ref_byteswap, | |
6171 | sizeof (zc_byteswap))); | |
2fe36b0b DQ |
6172 | |
6173 | /* Test ABD - data */ | |
6174 | abd_fletcher_4_byteswap(abd_data, size, NULL, | |
6175 | &zc_byteswap); | |
6176 | abd_fletcher_4_native(abd_data, size, NULL, &zc); | |
6177 | ||
6178 | VERIFY0(bcmp(&zc, &zc_ref, sizeof (zc))); | |
6179 | VERIFY0(bcmp(&zc_byteswap, &zc_ref_byteswap, | |
6180 | sizeof (zc_byteswap))); | |
6181 | ||
6182 | /* Test ABD - metadata */ | |
6183 | abd_fletcher_4_byteswap(abd_meta, size, NULL, | |
6184 | &zc_byteswap); | |
6185 | abd_fletcher_4_native(abd_meta, size, NULL, &zc); | |
6186 | ||
6187 | VERIFY0(bcmp(&zc, &zc_ref, sizeof (zc))); | |
6188 | VERIFY0(bcmp(&zc_byteswap, &zc_ref_byteswap, | |
6189 | sizeof (zc_byteswap))); | |
6190 | ||
1eeb4562 JX |
6191 | } |
6192 | ||
6193 | umem_free(buf, size); | |
2fe36b0b DQ |
6194 | abd_free(abd_data); |
6195 | abd_free(abd_meta); | |
1eeb4562 JX |
6196 | } |
6197 | } | |
6198 | ||
37f520db GN |
6199 | void |
6200 | ztest_fletcher_incr(ztest_ds_t *zd, uint64_t id) | |
6201 | { | |
6202 | void *buf; | |
6203 | size_t size; | |
6204 | int *ptr; | |
6205 | int i; | |
6206 | zio_cksum_t zc_ref; | |
6207 | zio_cksum_t zc_ref_bswap; | |
6208 | ||
6209 | hrtime_t end = gethrtime() + NANOSEC; | |
6210 | ||
6211 | while (gethrtime() <= end) { | |
6212 | int run_count = 100; | |
6213 | ||
6214 | size = ztest_random_blocksize(); | |
6215 | buf = umem_alloc(size, UMEM_NOFAIL); | |
6216 | ||
6217 | for (i = 0, ptr = buf; i < size / sizeof (*ptr); i++, ptr++) | |
6218 | *ptr = ztest_random(UINT_MAX); | |
6219 | ||
6220 | VERIFY0(fletcher_4_impl_set("scalar")); | |
6221 | fletcher_4_native(buf, size, NULL, &zc_ref); | |
6222 | fletcher_4_byteswap(buf, size, NULL, &zc_ref_bswap); | |
6223 | ||
6224 | VERIFY0(fletcher_4_impl_set("cycle")); | |
6225 | ||
6226 | while (run_count-- > 0) { | |
6227 | zio_cksum_t zc; | |
6228 | zio_cksum_t zc_bswap; | |
6229 | size_t pos = 0; | |
6230 | ||
6231 | ZIO_SET_CHECKSUM(&zc, 0, 0, 0, 0); | |
6232 | ZIO_SET_CHECKSUM(&zc_bswap, 0, 0, 0, 0); | |
6233 | ||
6234 | while (pos < size) { | |
6235 | size_t inc = 64 * ztest_random(size / 67); | |
6236 | /* sometimes add few bytes to test non-simd */ | |
6237 | if (ztest_random(100) < 10) | |
6238 | inc += P2ALIGN(ztest_random(64), | |
6239 | sizeof (uint32_t)); | |
6240 | ||
6241 | if (inc > (size - pos)) | |
6242 | inc = size - pos; | |
6243 | ||
6244 | fletcher_4_incremental_native(buf + pos, inc, | |
6245 | &zc); | |
6246 | fletcher_4_incremental_byteswap(buf + pos, inc, | |
6247 | &zc_bswap); | |
6248 | ||
6249 | pos += inc; | |
6250 | } | |
6251 | ||
6252 | VERIFY3U(pos, ==, size); | |
6253 | ||
6254 | VERIFY(ZIO_CHECKSUM_EQUAL(zc, zc_ref)); | |
6255 | VERIFY(ZIO_CHECKSUM_EQUAL(zc_bswap, zc_ref_bswap)); | |
6256 | ||
6257 | /* | |
6258 | * verify if incremental on the whole buffer is | |
6259 | * equivalent to non-incremental version | |
6260 | */ | |
6261 | ZIO_SET_CHECKSUM(&zc, 0, 0, 0, 0); | |
6262 | ZIO_SET_CHECKSUM(&zc_bswap, 0, 0, 0, 0); | |
6263 | ||
6264 | fletcher_4_incremental_native(buf, size, &zc); | |
6265 | fletcher_4_incremental_byteswap(buf, size, &zc_bswap); | |
6266 | ||
6267 | VERIFY(ZIO_CHECKSUM_EQUAL(zc, zc_ref)); | |
6268 | VERIFY(ZIO_CHECKSUM_EQUAL(zc_bswap, zc_ref_bswap)); | |
6269 | } | |
6270 | ||
6271 | umem_free(buf, size); | |
6272 | } | |
6273 | } | |
6274 | ||
90aa094d CC |
6275 | static int |
6276 | ztest_check_path(char *path) | |
6277 | { | |
6278 | struct stat s; | |
6279 | /* return true on success */ | |
6280 | return (!stat(path, &s)); | |
6281 | } | |
6282 | ||
6283 | static void | |
6284 | ztest_get_zdb_bin(char *bin, int len) | |
6285 | { | |
6286 | char *zdb_path; | |
6287 | /* | |
6288 | * Try to use ZDB_PATH and in-tree zdb path. If not successful, just | |
6289 | * let popen to search through PATH. | |
6290 | */ | |
6291 | if ((zdb_path = getenv("ZDB_PATH"))) { | |
6292 | strlcpy(bin, zdb_path, len); /* In env */ | |
6293 | if (!ztest_check_path(bin)) { | |
6294 | ztest_dump_core = 0; | |
6295 | fatal(1, "invalid ZDB_PATH '%s'", bin); | |
6296 | } | |
6297 | return; | |
6298 | } | |
6299 | ||
6300 | VERIFY(realpath(getexecname(), bin) != NULL); | |
6301 | if (strstr(bin, "/ztest/")) { | |
6302 | strstr(bin, "/ztest/")[0] = '\0'; /* In-tree */ | |
6303 | strcat(bin, "/zdb/zdb"); | |
6304 | if (ztest_check_path(bin)) | |
6305 | return; | |
6306 | } | |
6307 | strcpy(bin, "zdb"); | |
6308 | } | |
6309 | ||
428870ff BB |
6310 | /* |
6311 | * Verify pool integrity by running zdb. | |
6312 | */ | |
34dc7c2f | 6313 | static void |
428870ff | 6314 | ztest_run_zdb(char *pool) |
34dc7c2f BB |
6315 | { |
6316 | int status; | |
34dc7c2f | 6317 | char *bin; |
0e8d1b2d BB |
6318 | char *zdb; |
6319 | char *zbuf; | |
90aa094d | 6320 | const int len = MAXPATHLEN + MAXNAMELEN + 20; |
34dc7c2f BB |
6321 | FILE *fp; |
6322 | ||
90aa094d CC |
6323 | bin = umem_alloc(len, UMEM_NOFAIL); |
6324 | zdb = umem_alloc(len, UMEM_NOFAIL); | |
0e8d1b2d | 6325 | zbuf = umem_alloc(1024, UMEM_NOFAIL); |
34dc7c2f | 6326 | |
90aa094d | 6327 | ztest_get_zdb_bin(bin, len); |
0e8d1b2d BB |
6328 | |
6329 | (void) sprintf(zdb, | |
4589f3ae | 6330 | "%s -bcc%s%s -G -d -U %s " |
1258bd77 | 6331 | "-o zfs_reconstruct_indirect_combinations_max=65536 %s", |
0e8d1b2d | 6332 | bin, |
c242c188 CS |
6333 | ztest_opts.zo_verbose >= 3 ? "s" : "", |
6334 | ztest_opts.zo_verbose >= 4 ? "v" : "", | |
428870ff | 6335 | spa_config_path, |
b128c09f | 6336 | pool); |
34dc7c2f | 6337 | |
c242c188 | 6338 | if (ztest_opts.zo_verbose >= 5) |
34dc7c2f BB |
6339 | (void) printf("Executing %s\n", strstr(zdb, "zdb ")); |
6340 | ||
6341 | fp = popen(zdb, "r"); | |
6342 | ||
6aec1cd5 | 6343 | while (fgets(zbuf, 1024, fp) != NULL) |
c242c188 | 6344 | if (ztest_opts.zo_verbose >= 3) |
34dc7c2f BB |
6345 | (void) printf("%s", zbuf); |
6346 | ||
6347 | status = pclose(fp); | |
6348 | ||
6349 | if (status == 0) | |
0e8d1b2d | 6350 | goto out; |
34dc7c2f BB |
6351 | |
6352 | ztest_dump_core = 0; | |
6353 | if (WIFEXITED(status)) | |
6354 | fatal(0, "'%s' exit code %d", zdb, WEXITSTATUS(status)); | |
6355 | else | |
6356 | fatal(0, "'%s' died with signal %d", zdb, WTERMSIG(status)); | |
0e8d1b2d | 6357 | out: |
90aa094d CC |
6358 | umem_free(bin, len); |
6359 | umem_free(zdb, len); | |
0e8d1b2d | 6360 | umem_free(zbuf, 1024); |
34dc7c2f BB |
6361 | } |
6362 | ||
6363 | static void | |
6364 | ztest_walk_pool_directory(char *header) | |
6365 | { | |
6366 | spa_t *spa = NULL; | |
6367 | ||
c242c188 | 6368 | if (ztest_opts.zo_verbose >= 6) |
34dc7c2f BB |
6369 | (void) printf("%s\n", header); |
6370 | ||
6371 | mutex_enter(&spa_namespace_lock); | |
6372 | while ((spa = spa_next(spa)) != NULL) | |
c242c188 | 6373 | if (ztest_opts.zo_verbose >= 6) |
34dc7c2f BB |
6374 | (void) printf("\t%s\n", spa_name(spa)); |
6375 | mutex_exit(&spa_namespace_lock); | |
6376 | } | |
6377 | ||
6378 | static void | |
6379 | ztest_spa_import_export(char *oldname, char *newname) | |
6380 | { | |
fb5f0bc8 | 6381 | nvlist_t *config, *newconfig; |
34dc7c2f BB |
6382 | uint64_t pool_guid; |
6383 | spa_t *spa; | |
13fe0198 | 6384 | int error; |
34dc7c2f | 6385 | |
c242c188 | 6386 | if (ztest_opts.zo_verbose >= 4) { |
34dc7c2f BB |
6387 | (void) printf("import/export: old = %s, new = %s\n", |
6388 | oldname, newname); | |
6389 | } | |
6390 | ||
6391 | /* | |
6392 | * Clean up from previous runs. | |
6393 | */ | |
6394 | (void) spa_destroy(newname); | |
6395 | ||
6396 | /* | |
6397 | * Get the pool's configuration and guid. | |
6398 | */ | |
428870ff | 6399 | VERIFY3U(0, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f | 6400 | |
fb5f0bc8 BB |
6401 | /* |
6402 | * Kick off a scrub to tickle scrub/export races. | |
6403 | */ | |
6404 | if (ztest_random(2) == 0) | |
428870ff | 6405 | (void) spa_scan(spa, POOL_SCAN_SCRUB); |
fb5f0bc8 | 6406 | |
34dc7c2f BB |
6407 | pool_guid = spa_guid(spa); |
6408 | spa_close(spa, FTAG); | |
6409 | ||
6410 | ztest_walk_pool_directory("pools before export"); | |
6411 | ||
6412 | /* | |
6413 | * Export it. | |
6414 | */ | |
428870ff | 6415 | VERIFY3U(0, ==, spa_export(oldname, &config, B_FALSE, B_FALSE)); |
34dc7c2f BB |
6416 | |
6417 | ztest_walk_pool_directory("pools after export"); | |
6418 | ||
fb5f0bc8 BB |
6419 | /* |
6420 | * Try to import it. | |
6421 | */ | |
6422 | newconfig = spa_tryimport(config); | |
6423 | ASSERT(newconfig != NULL); | |
6424 | nvlist_free(newconfig); | |
6425 | ||
34dc7c2f BB |
6426 | /* |
6427 | * Import it under the new name. | |
6428 | */ | |
13fe0198 MA |
6429 | error = spa_import(newname, config, NULL, 0); |
6430 | if (error != 0) { | |
6431 | dump_nvlist(config, 0); | |
6432 | fatal(B_FALSE, "couldn't import pool %s as %s: error %u", | |
6433 | oldname, newname, error); | |
6434 | } | |
34dc7c2f BB |
6435 | |
6436 | ztest_walk_pool_directory("pools after import"); | |
6437 | ||
6438 | /* | |
6439 | * Try to import it again -- should fail with EEXIST. | |
6440 | */ | |
572e2857 | 6441 | VERIFY3U(EEXIST, ==, spa_import(newname, config, NULL, 0)); |
34dc7c2f BB |
6442 | |
6443 | /* | |
6444 | * Try to import it under a different name -- should fail with EEXIST. | |
6445 | */ | |
572e2857 | 6446 | VERIFY3U(EEXIST, ==, spa_import(oldname, config, NULL, 0)); |
34dc7c2f BB |
6447 | |
6448 | /* | |
6449 | * Verify that the pool is no longer visible under the old name. | |
6450 | */ | |
428870ff | 6451 | VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f BB |
6452 | |
6453 | /* | |
6454 | * Verify that we can open and close the pool using the new name. | |
6455 | */ | |
428870ff | 6456 | VERIFY3U(0, ==, spa_open(newname, &spa, FTAG)); |
34dc7c2f BB |
6457 | ASSERT(pool_guid == spa_guid(spa)); |
6458 | spa_close(spa, FTAG); | |
6459 | ||
6460 | nvlist_free(config); | |
6461 | } | |
6462 | ||
fb5f0bc8 BB |
6463 | static void |
6464 | ztest_resume(spa_t *spa) | |
6465 | { | |
c242c188 | 6466 | if (spa_suspended(spa) && ztest_opts.zo_verbose >= 6) |
428870ff BB |
6467 | (void) printf("resuming from suspended state\n"); |
6468 | spa_vdev_state_enter(spa, SCL_NONE); | |
6469 | vdev_clear(spa, NULL); | |
6470 | (void) spa_vdev_state_exit(spa, NULL, 0); | |
6471 | (void) zio_resume(spa); | |
fb5f0bc8 BB |
6472 | } |
6473 | ||
c25b8f99 | 6474 | static void |
fb5f0bc8 | 6475 | ztest_resume_thread(void *arg) |
34dc7c2f | 6476 | { |
b128c09f | 6477 | spa_t *spa = arg; |
34dc7c2f BB |
6478 | |
6479 | while (!ztest_exiting) { | |
428870ff BB |
6480 | if (spa_suspended(spa)) |
6481 | ztest_resume(spa); | |
6482 | (void) poll(NULL, 0, 100); | |
d3c2ae1c GW |
6483 | |
6484 | /* | |
6485 | * Periodically change the zfs_compressed_arc_enabled setting. | |
6486 | */ | |
6487 | if (ztest_random(10) == 0) | |
6488 | zfs_compressed_arc_enabled = ztest_random(2); | |
a6255b7f DQ |
6489 | |
6490 | /* | |
6491 | * Periodically change the zfs_abd_scatter_enabled setting. | |
6492 | */ | |
6493 | if (ztest_random(10) == 0) | |
6494 | zfs_abd_scatter_enabled = ztest_random(2); | |
34dc7c2f | 6495 | } |
34dc7c2f | 6496 | |
1e33ac1e | 6497 | thread_exit(); |
1e33ac1e | 6498 | } |
428870ff | 6499 | |
1e33ac1e | 6500 | static void |
8fb1ede1 | 6501 | ztest_deadman_thread(void *arg) |
1e33ac1e | 6502 | { |
8fb1ede1 BB |
6503 | ztest_shared_t *zs = arg; |
6504 | spa_t *spa = ztest_spa; | |
94102578 | 6505 | hrtime_t delay, overdue, last_run = gethrtime(); |
8fb1ede1 | 6506 | |
94102578 TC |
6507 | delay = (zs->zs_thread_stop - zs->zs_thread_start) + |
6508 | MSEC2NSEC(zfs_deadman_synctime_ms); | |
8fb1ede1 | 6509 | |
94102578 TC |
6510 | while (!ztest_exiting) { |
6511 | /* | |
6512 | * Wait for the delay timer while checking occasionally | |
6513 | * if we should stop. | |
6514 | */ | |
6515 | if (gethrtime() < last_run + delay) { | |
6516 | (void) poll(NULL, 0, 1000); | |
6517 | continue; | |
6518 | } | |
8fb1ede1 BB |
6519 | |
6520 | /* | |
6521 | * If the pool is suspended then fail immediately. Otherwise, | |
6522 | * check to see if the pool is making any progress. If | |
6523 | * vdev_deadman() discovers that there hasn't been any recent | |
6524 | * I/Os then it will end up aborting the tests. | |
6525 | */ | |
6526 | if (spa_suspended(spa) || spa->spa_root_vdev == NULL) { | |
6527 | fatal(0, "aborting test after %llu seconds because " | |
6528 | "pool has transitioned to a suspended state.", | |
6529 | zfs_deadman_synctime_ms / 1000); | |
6530 | } | |
6531 | vdev_deadman(spa->spa_root_vdev, FTAG); | |
6532 | ||
6533 | /* | |
6534 | * If the process doesn't complete within a grace period of | |
6535 | * zfs_deadman_synctime_ms over the expected finish time, | |
6536 | * then it may be hung and is terminated. | |
6537 | */ | |
6538 | overdue = zs->zs_proc_stop + MSEC2NSEC(zfs_deadman_synctime_ms); | |
8fb1ede1 BB |
6539 | if (gethrtime() > overdue) { |
6540 | fatal(0, "aborting test after %llu seconds because " | |
94102578 TC |
6541 | "the process is overdue for termination.", |
6542 | (gethrtime() - zs->zs_proc_start) / NANOSEC); | |
8fb1ede1 BB |
6543 | } |
6544 | ||
6545 | (void) printf("ztest has been running for %lld seconds\n", | |
94102578 TC |
6546 | (gethrtime() - zs->zs_proc_start) / NANOSEC); |
6547 | ||
6548 | last_run = gethrtime(); | |
6549 | delay = MSEC2NSEC(zfs_deadman_checktime_ms); | |
8fb1ede1 | 6550 | } |
94102578 TC |
6551 | |
6552 | thread_exit(); | |
428870ff BB |
6553 | } |
6554 | ||
6555 | static void | |
c242c188 | 6556 | ztest_execute(int test, ztest_info_t *zi, uint64_t id) |
428870ff | 6557 | { |
c242c188 CS |
6558 | ztest_ds_t *zd = &ztest_ds[id % ztest_opts.zo_datasets]; |
6559 | ztest_shared_callstate_t *zc = ZTEST_GET_SHARED_CALLSTATE(test); | |
428870ff | 6560 | hrtime_t functime = gethrtime(); |
d6320ddb | 6561 | int i; |
428870ff | 6562 | |
d6320ddb | 6563 | for (i = 0; i < zi->zi_iters; i++) |
428870ff BB |
6564 | zi->zi_func(zd, id); |
6565 | ||
6566 | functime = gethrtime() - functime; | |
6567 | ||
c242c188 CS |
6568 | atomic_add_64(&zc->zc_count, 1); |
6569 | atomic_add_64(&zc->zc_time, functime); | |
428870ff | 6570 | |
fdc5d982 | 6571 | if (ztest_opts.zo_verbose >= 4) |
428870ff | 6572 | (void) printf("%6.2f sec in %s\n", |
fdc5d982 | 6573 | (double)functime / NANOSEC, zi->zi_funcname); |
428870ff BB |
6574 | } |
6575 | ||
c25b8f99 | 6576 | static void |
34dc7c2f BB |
6577 | ztest_thread(void *arg) |
6578 | { | |
c242c188 | 6579 | int rand; |
428870ff | 6580 | uint64_t id = (uintptr_t)arg; |
34dc7c2f | 6581 | ztest_shared_t *zs = ztest_shared; |
428870ff BB |
6582 | uint64_t call_next; |
6583 | hrtime_t now; | |
34dc7c2f | 6584 | ztest_info_t *zi; |
c242c188 | 6585 | ztest_shared_callstate_t *zc; |
34dc7c2f | 6586 | |
428870ff | 6587 | while ((now = gethrtime()) < zs->zs_thread_stop) { |
34dc7c2f BB |
6588 | /* |
6589 | * See if it's time to force a crash. | |
6590 | */ | |
428870ff BB |
6591 | if (now > zs->zs_thread_kill) |
6592 | ztest_kill(zs); | |
34dc7c2f BB |
6593 | |
6594 | /* | |
428870ff | 6595 | * If we're getting ENOSPC with some regularity, stop. |
34dc7c2f | 6596 | */ |
428870ff BB |
6597 | if (zs->zs_enospc_count > 10) |
6598 | break; | |
34dc7c2f BB |
6599 | |
6600 | /* | |
428870ff | 6601 | * Pick a random function to execute. |
34dc7c2f | 6602 | */ |
c242c188 CS |
6603 | rand = ztest_random(ZTEST_FUNCS); |
6604 | zi = &ztest_info[rand]; | |
6605 | zc = ZTEST_GET_SHARED_CALLSTATE(rand); | |
6606 | call_next = zc->zc_next; | |
34dc7c2f | 6607 | |
428870ff | 6608 | if (now >= call_next && |
c242c188 CS |
6609 | atomic_cas_64(&zc->zc_next, call_next, call_next + |
6610 | ztest_random(2 * zi->zi_interval[0] + 1)) == call_next) { | |
6611 | ztest_execute(rand, zi, id); | |
6612 | } | |
428870ff | 6613 | } |
34dc7c2f | 6614 | |
1e33ac1e | 6615 | thread_exit(); |
428870ff | 6616 | } |
34dc7c2f | 6617 | |
428870ff BB |
6618 | static void |
6619 | ztest_dataset_name(char *dsname, char *pool, int d) | |
6620 | { | |
eca7b760 | 6621 | (void) snprintf(dsname, ZFS_MAX_DATASET_NAME_LEN, "%s/ds_%d", pool, d); |
428870ff | 6622 | } |
34dc7c2f | 6623 | |
428870ff | 6624 | static void |
c242c188 | 6625 | ztest_dataset_destroy(int d) |
428870ff | 6626 | { |
eca7b760 | 6627 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
d6320ddb | 6628 | int t; |
34dc7c2f | 6629 | |
c242c188 | 6630 | ztest_dataset_name(name, ztest_opts.zo_pool, d); |
34dc7c2f | 6631 | |
c242c188 | 6632 | if (ztest_opts.zo_verbose >= 3) |
428870ff | 6633 | (void) printf("Destroying %s to free up space\n", name); |
34dc7c2f | 6634 | |
428870ff BB |
6635 | /* |
6636 | * Cleanup any non-standard clones and snapshots. In general, | |
6637 | * ztest thread t operates on dataset (t % zopt_datasets), | |
6638 | * so there may be more than one thing to clean up. | |
6639 | */ | |
c242c188 CS |
6640 | for (t = d; t < ztest_opts.zo_threads; |
6641 | t += ztest_opts.zo_datasets) | |
428870ff BB |
6642 | ztest_dsl_dataset_cleanup(name, t); |
6643 | ||
6644 | (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL, | |
6645 | DS_FIND_SNAPSHOTS | DS_FIND_CHILDREN); | |
6646 | } | |
6647 | ||
6648 | static void | |
6649 | ztest_dataset_dirobj_verify(ztest_ds_t *zd) | |
6650 | { | |
6651 | uint64_t usedobjs, dirobjs, scratch; | |
6652 | ||
6653 | /* | |
6654 | * ZTEST_DIROBJ is the object directory for the entire dataset. | |
6655 | * Therefore, the number of objects in use should equal the | |
6656 | * number of ZTEST_DIROBJ entries, +1 for ZTEST_DIROBJ itself. | |
6657 | * If not, we have an object leak. | |
6658 | * | |
6659 | * Note that we can only check this in ztest_dataset_open(), | |
6660 | * when the open-context and syncing-context values agree. | |
6661 | * That's because zap_count() returns the open-context value, | |
6662 | * while dmu_objset_space() returns the rootbp fill count. | |
6663 | */ | |
6664 | VERIFY3U(0, ==, zap_count(zd->zd_os, ZTEST_DIROBJ, &dirobjs)); | |
6665 | dmu_objset_space(zd->zd_os, &scratch, &scratch, &usedobjs, &scratch); | |
6666 | ASSERT3U(dirobjs + 1, ==, usedobjs); | |
6667 | } | |
6668 | ||
6669 | static int | |
c242c188 | 6670 | ztest_dataset_open(int d) |
428870ff | 6671 | { |
c242c188 CS |
6672 | ztest_ds_t *zd = &ztest_ds[d]; |
6673 | uint64_t committed_seq = ZTEST_GET_SHARED_DS(d)->zd_seq; | |
428870ff BB |
6674 | objset_t *os; |
6675 | zilog_t *zilog; | |
eca7b760 | 6676 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
6677 | int error; |
6678 | ||
c242c188 | 6679 | ztest_dataset_name(name, ztest_opts.zo_pool, d); |
428870ff | 6680 | |
2d9142c9 | 6681 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
428870ff BB |
6682 | |
6683 | error = ztest_dataset_create(name); | |
6684 | if (error == ENOSPC) { | |
2d9142c9 | 6685 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
428870ff BB |
6686 | ztest_record_enospc(FTAG); |
6687 | return (error); | |
34dc7c2f | 6688 | } |
428870ff | 6689 | ASSERT(error == 0 || error == EEXIST); |
34dc7c2f | 6690 | |
4807c0ba TC |
6691 | VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, |
6692 | B_TRUE, zd, &os)); | |
2d9142c9 | 6693 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
428870ff | 6694 | |
c242c188 | 6695 | ztest_zd_init(zd, ZTEST_GET_SHARED_DS(d), os); |
428870ff BB |
6696 | |
6697 | zilog = zd->zd_zilog; | |
6698 | ||
6699 | if (zilog->zl_header->zh_claim_lr_seq != 0 && | |
6700 | zilog->zl_header->zh_claim_lr_seq < committed_seq) | |
6701 | fatal(0, "missing log records: claimed %llu < committed %llu", | |
6702 | zilog->zl_header->zh_claim_lr_seq, committed_seq); | |
6703 | ||
6704 | ztest_dataset_dirobj_verify(zd); | |
6705 | ||
6706 | zil_replay(os, zd, ztest_replay_vector); | |
6707 | ||
6708 | ztest_dataset_dirobj_verify(zd); | |
6709 | ||
c242c188 | 6710 | if (ztest_opts.zo_verbose >= 6) |
428870ff BB |
6711 | (void) printf("%s replay %llu blocks, %llu records, seq %llu\n", |
6712 | zd->zd_name, | |
6713 | (u_longlong_t)zilog->zl_parse_blk_count, | |
6714 | (u_longlong_t)zilog->zl_parse_lr_count, | |
6715 | (u_longlong_t)zilog->zl_replaying_seq); | |
6716 | ||
6717 | zilog = zil_open(os, ztest_get_data); | |
6718 | ||
6719 | if (zilog->zl_replaying_seq != 0 && | |
6720 | zilog->zl_replaying_seq < committed_seq) | |
6721 | fatal(0, "missing log records: replayed %llu < committed %llu", | |
6722 | zilog->zl_replaying_seq, committed_seq); | |
6723 | ||
6724 | return (0); | |
6725 | } | |
6726 | ||
6727 | static void | |
c242c188 | 6728 | ztest_dataset_close(int d) |
428870ff | 6729 | { |
c242c188 | 6730 | ztest_ds_t *zd = &ztest_ds[d]; |
428870ff BB |
6731 | |
6732 | zil_close(zd->zd_zilog); | |
b5256303 | 6733 | dmu_objset_disown(zd->zd_os, B_TRUE, zd); |
428870ff BB |
6734 | |
6735 | ztest_zd_fini(zd); | |
34dc7c2f BB |
6736 | } |
6737 | ||
6738 | /* | |
6739 | * Kick off threads to run tests on all datasets in parallel. | |
6740 | */ | |
6741 | static void | |
428870ff | 6742 | ztest_run(ztest_shared_t *zs) |
34dc7c2f | 6743 | { |
34dc7c2f | 6744 | spa_t *spa; |
3541dc6d | 6745 | objset_t *os; |
94102578 | 6746 | kthread_t *resume_thread, *deadman_thread; |
c25b8f99 | 6747 | kthread_t **run_threads; |
1e33ac1e | 6748 | uint64_t object; |
428870ff | 6749 | int error; |
d6320ddb | 6750 | int t, d; |
b128c09f BB |
6751 | |
6752 | ztest_exiting = B_FALSE; | |
34dc7c2f | 6753 | |
34dc7c2f | 6754 | /* |
428870ff | 6755 | * Initialize parent/child shared state. |
34dc7c2f | 6756 | */ |
c242c188 | 6757 | mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); |
d2734cce | 6758 | mutex_init(&ztest_checkpoint_lock, NULL, MUTEX_DEFAULT, NULL); |
2d9142c9 | 6759 | VERIFY0(pthread_rwlock_init(&ztest_name_lock, NULL)); |
34dc7c2f | 6760 | |
428870ff | 6761 | zs->zs_thread_start = gethrtime(); |
c242c188 CS |
6762 | zs->zs_thread_stop = |
6763 | zs->zs_thread_start + ztest_opts.zo_passtime * NANOSEC; | |
428870ff BB |
6764 | zs->zs_thread_stop = MIN(zs->zs_thread_stop, zs->zs_proc_stop); |
6765 | zs->zs_thread_kill = zs->zs_thread_stop; | |
c242c188 CS |
6766 | if (ztest_random(100) < ztest_opts.zo_killrate) { |
6767 | zs->zs_thread_kill -= | |
6768 | ztest_random(ztest_opts.zo_passtime * NANOSEC); | |
6769 | } | |
34dc7c2f | 6770 | |
1e33ac1e | 6771 | mutex_init(&zcl.zcl_callbacks_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 6772 | |
428870ff BB |
6773 | list_create(&zcl.zcl_callbacks, sizeof (ztest_cb_data_t), |
6774 | offsetof(ztest_cb_data_t, zcd_node)); | |
34dc7c2f BB |
6775 | |
6776 | /* | |
b128c09f | 6777 | * Open our pool. |
34dc7c2f | 6778 | */ |
428870ff | 6779 | kernel_init(FREAD | FWRITE); |
13fe0198 | 6780 | VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
080b3100 | 6781 | metaslab_preload_limit = ztest_random(20) + 1; |
c242c188 | 6782 | ztest_spa = spa; |
428870ff | 6783 | |
546d32ca | 6784 | dmu_objset_stats_t dds; |
4807c0ba | 6785 | VERIFY0(ztest_dmu_objset_own(ztest_opts.zo_pool, |
b5256303 | 6786 | DMU_OST_ANY, B_TRUE, B_TRUE, FTAG, &os)); |
546d32ca BB |
6787 | dsl_pool_config_enter(dmu_objset_pool(os), FTAG); |
6788 | dmu_objset_fast_stat(os, &dds); | |
6789 | dsl_pool_config_exit(dmu_objset_pool(os), FTAG); | |
6790 | zs->zs_guid = dds.dds_guid; | |
b5256303 | 6791 | dmu_objset_disown(os, B_TRUE, FTAG); |
3541dc6d | 6792 | |
428870ff | 6793 | spa->spa_dedup_ditto = 2 * ZIO_DEDUPDITTO_MIN; |
34dc7c2f BB |
6794 | |
6795 | /* | |
b128c09f | 6796 | * Create a thread to periodically resume suspended I/O. |
34dc7c2f | 6797 | */ |
c25b8f99 BB |
6798 | resume_thread = thread_create(NULL, 0, ztest_resume_thread, |
6799 | spa, 0, NULL, TS_RUN | TS_JOINABLE, defclsyspri); | |
34dc7c2f | 6800 | |
428870ff | 6801 | /* |
8fb1ede1 | 6802 | * Create a deadman thread and set to panic if we hang. |
428870ff | 6803 | */ |
94102578 | 6804 | deadman_thread = thread_create(NULL, 0, ztest_deadman_thread, |
8fb1ede1 BB |
6805 | zs, 0, NULL, TS_RUN | TS_JOINABLE, defclsyspri); |
6806 | ||
6807 | spa->spa_deadman_failmode = ZIO_FAILURE_MODE_PANIC; | |
428870ff | 6808 | |
34dc7c2f | 6809 | /* |
d2734cce | 6810 | * Verify that we can safely inquire about any object, |
34dc7c2f BB |
6811 | * whether it's allocated or not. To make it interesting, |
6812 | * we probe a 5-wide window around each power of two. | |
6813 | * This hits all edge cases, including zero and the max. | |
6814 | */ | |
d6320ddb BB |
6815 | for (t = 0; t < 64; t++) { |
6816 | for (d = -5; d <= 5; d++) { | |
34dc7c2f BB |
6817 | error = dmu_object_info(spa->spa_meta_objset, |
6818 | (1ULL << t) + d, NULL); | |
6819 | ASSERT(error == 0 || error == ENOENT || | |
6820 | error == EINVAL); | |
6821 | } | |
6822 | } | |
6823 | ||
6824 | /* | |
428870ff | 6825 | * If we got any ENOSPC errors on the previous run, destroy something. |
34dc7c2f | 6826 | */ |
428870ff | 6827 | if (zs->zs_enospc_count != 0) { |
c242c188 CS |
6828 | int d = ztest_random(ztest_opts.zo_datasets); |
6829 | ztest_dataset_destroy(d); | |
428870ff | 6830 | } |
34dc7c2f BB |
6831 | zs->zs_enospc_count = 0; |
6832 | ||
c25b8f99 | 6833 | run_threads = umem_zalloc(ztest_opts.zo_threads * sizeof (kthread_t *), |
c242c188 | 6834 | UMEM_NOFAIL); |
34dc7c2f | 6835 | |
c242c188 | 6836 | if (ztest_opts.zo_verbose >= 4) |
34dc7c2f BB |
6837 | (void) printf("starting main threads...\n"); |
6838 | ||
428870ff BB |
6839 | /* |
6840 | * Kick off all the tests that run in parallel. | |
6841 | */ | |
c242c188 | 6842 | for (t = 0; t < ztest_opts.zo_threads; t++) { |
c25b8f99 BB |
6843 | if (t < ztest_opts.zo_datasets && ztest_dataset_open(t) != 0) { |
6844 | umem_free(run_threads, ztest_opts.zo_threads * | |
6845 | sizeof (kthread_t *)); | |
428870ff | 6846 | return; |
06cf4d98 | 6847 | } |
1e33ac1e | 6848 | |
c25b8f99 BB |
6849 | run_threads[t] = thread_create(NULL, 0, ztest_thread, |
6850 | (void *)(uintptr_t)t, 0, NULL, TS_RUN | TS_JOINABLE, | |
6851 | defclsyspri); | |
34dc7c2f BB |
6852 | } |
6853 | ||
428870ff BB |
6854 | /* |
6855 | * Wait for all of the tests to complete. We go in reverse order | |
6856 | * so we don't close datasets while threads are still using them. | |
6857 | */ | |
c242c188 | 6858 | for (t = ztest_opts.zo_threads - 1; t >= 0; t--) { |
c25b8f99 | 6859 | VERIFY0(thread_join(run_threads[t])); |
c242c188 CS |
6860 | if (t < ztest_opts.zo_datasets) |
6861 | ztest_dataset_close(t); | |
34dc7c2f BB |
6862 | } |
6863 | ||
34dc7c2f BB |
6864 | txg_wait_synced(spa_get_dsl(spa), 0); |
6865 | ||
428870ff BB |
6866 | zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(spa)); |
6867 | zs->zs_space = metaslab_class_get_space(spa_normal_class(spa)); | |
6868 | ||
c25b8f99 | 6869 | umem_free(run_threads, ztest_opts.zo_threads * sizeof (kthread_t *)); |
428870ff | 6870 | |
94102578 | 6871 | /* Kill the resume and deadman threads */ |
428870ff | 6872 | ztest_exiting = B_TRUE; |
c25b8f99 | 6873 | VERIFY0(thread_join(resume_thread)); |
94102578 | 6874 | VERIFY0(thread_join(deadman_thread)); |
428870ff | 6875 | ztest_resume(spa); |
34dc7c2f BB |
6876 | |
6877 | /* | |
428870ff BB |
6878 | * Right before closing the pool, kick off a bunch of async I/O; |
6879 | * spa_close() should wait for it to complete. | |
34dc7c2f | 6880 | */ |
fcff0f35 PD |
6881 | for (object = 1; object < 50; object++) { |
6882 | dmu_prefetch(spa->spa_meta_objset, object, 0, 0, 1ULL << 20, | |
6883 | ZIO_PRIORITY_SYNC_READ); | |
6884 | } | |
428870ff | 6885 | |
090ff092 RC |
6886 | /* Verify that at least one commit cb was called in a timely fashion */ |
6887 | if (zc_cb_counter >= ZTEST_COMMIT_CB_MIN_REG) | |
c99c9001 | 6888 | VERIFY0(zc_min_txg_delay); |
090ff092 | 6889 | |
428870ff BB |
6890 | spa_close(spa, FTAG); |
6891 | ||
6892 | /* | |
6893 | * Verify that we can loop over all pools. | |
6894 | */ | |
6895 | mutex_enter(&spa_namespace_lock); | |
6896 | for (spa = spa_next(NULL); spa != NULL; spa = spa_next(spa)) | |
c242c188 | 6897 | if (ztest_opts.zo_verbose > 3) |
428870ff BB |
6898 | (void) printf("spa_next: found %s\n", spa_name(spa)); |
6899 | mutex_exit(&spa_namespace_lock); | |
6900 | ||
6901 | /* | |
6902 | * Verify that we can export the pool and reimport it under a | |
6903 | * different name. | |
6904 | */ | |
379ca9cf | 6905 | if ((ztest_random(2) == 0) && !ztest_opts.zo_mmp_test) { |
eca7b760 IK |
6906 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
6907 | (void) snprintf(name, sizeof (name), "%s_import", | |
c242c188 CS |
6908 | ztest_opts.zo_pool); |
6909 | ztest_spa_import_export(ztest_opts.zo_pool, name); | |
6910 | ztest_spa_import_export(name, ztest_opts.zo_pool); | |
428870ff BB |
6911 | } |
6912 | ||
6913 | kernel_fini(); | |
572e2857 BB |
6914 | |
6915 | list_destroy(&zcl.zcl_callbacks); | |
0e8d1b2d | 6916 | mutex_destroy(&zcl.zcl_callbacks_lock); |
2d9142c9 | 6917 | (void) pthread_rwlock_destroy(&ztest_name_lock); |
c242c188 | 6918 | mutex_destroy(&ztest_vdev_lock); |
d2734cce | 6919 | mutex_destroy(&ztest_checkpoint_lock); |
428870ff BB |
6920 | } |
6921 | ||
6922 | static void | |
c242c188 | 6923 | ztest_freeze(void) |
428870ff | 6924 | { |
c242c188 | 6925 | ztest_ds_t *zd = &ztest_ds[0]; |
428870ff BB |
6926 | spa_t *spa; |
6927 | int numloops = 0; | |
6928 | ||
c242c188 | 6929 | if (ztest_opts.zo_verbose >= 3) |
428870ff | 6930 | (void) printf("testing spa_freeze()...\n"); |
9babb374 | 6931 | |
428870ff | 6932 | kernel_init(FREAD | FWRITE); |
c242c188 CS |
6933 | VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
6934 | VERIFY3U(0, ==, ztest_dataset_open(0)); | |
03c6040b | 6935 | ztest_spa = spa; |
9babb374 | 6936 | |
428870ff BB |
6937 | /* |
6938 | * Force the first log block to be transactionally allocated. | |
6939 | * We have to do this before we freeze the pool -- otherwise | |
6940 | * the log chain won't be anchored. | |
6941 | */ | |
6942 | while (BP_IS_HOLE(&zd->zd_zilog->zl_header->zh_log)) { | |
6943 | ztest_dmu_object_alloc_free(zd, 0); | |
572e2857 | 6944 | zil_commit(zd->zd_zilog, 0); |
34dc7c2f BB |
6945 | } |
6946 | ||
6947 | txg_wait_synced(spa_get_dsl(spa), 0); | |
6948 | ||
428870ff BB |
6949 | /* |
6950 | * Freeze the pool. This stops spa_sync() from doing anything, | |
6951 | * so that the only way to record changes from now on is the ZIL. | |
6952 | */ | |
6953 | spa_freeze(spa); | |
b128c09f | 6954 | |
b02fe35d AR |
6955 | /* |
6956 | * Because it is hard to predict how much space a write will actually | |
6957 | * require beforehand, we leave ourselves some fudge space to write over | |
6958 | * capacity. | |
6959 | */ | |
6960 | uint64_t capacity = metaslab_class_get_space(spa_normal_class(spa)) / 2; | |
6961 | ||
428870ff BB |
6962 | /* |
6963 | * Run tests that generate log records but don't alter the pool config | |
6964 | * or depend on DSL sync tasks (snapshots, objset create/destroy, etc). | |
6965 | * We do a txg_wait_synced() after each iteration to force the txg | |
6966 | * to increase well beyond the last synced value in the uberblock. | |
6967 | * The ZIL should be OK with that. | |
b02fe35d AR |
6968 | * |
6969 | * Run a random number of times less than zo_maxloops and ensure we do | |
6970 | * not run out of space on the pool. | |
428870ff | 6971 | */ |
c242c188 | 6972 | while (ztest_random(10) != 0 && |
b02fe35d AR |
6973 | numloops++ < ztest_opts.zo_maxloops && |
6974 | metaslab_class_get_alloc(spa_normal_class(spa)) < capacity) { | |
6975 | ztest_od_t od; | |
50c957f7 | 6976 | ztest_od_init(&od, 0, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0); |
b02fe35d AR |
6977 | VERIFY0(ztest_object_init(zd, &od, sizeof (od), B_FALSE)); |
6978 | ztest_io(zd, od.od_object, | |
6979 | ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
428870ff BB |
6980 | txg_wait_synced(spa_get_dsl(spa), 0); |
6981 | } | |
b128c09f | 6982 | |
34dc7c2f | 6983 | /* |
428870ff | 6984 | * Commit all of the changes we just generated. |
34dc7c2f | 6985 | */ |
572e2857 | 6986 | zil_commit(zd->zd_zilog, 0); |
428870ff | 6987 | txg_wait_synced(spa_get_dsl(spa), 0); |
34dc7c2f | 6988 | |
428870ff BB |
6989 | /* |
6990 | * Close our dataset and close the pool. | |
6991 | */ | |
c242c188 | 6992 | ztest_dataset_close(0); |
34dc7c2f | 6993 | spa_close(spa, FTAG); |
428870ff | 6994 | kernel_fini(); |
34dc7c2f | 6995 | |
428870ff BB |
6996 | /* |
6997 | * Open and close the pool and dataset to induce log replay. | |
6998 | */ | |
6999 | kernel_init(FREAD | FWRITE); | |
c242c188 | 7000 | VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
29809a6c | 7001 | ASSERT(spa_freeze_txg(spa) == UINT64_MAX); |
c242c188 | 7002 | VERIFY3U(0, ==, ztest_dataset_open(0)); |
3bc7e0fb GW |
7003 | ztest_spa = spa; |
7004 | txg_wait_synced(spa_get_dsl(spa), 0); | |
4807c0ba | 7005 | ztest_dataset_close(0); |
3bc7e0fb GW |
7006 | ztest_reguid(NULL, 0); |
7007 | ||
428870ff | 7008 | spa_close(spa, FTAG); |
34dc7c2f BB |
7009 | kernel_fini(); |
7010 | } | |
7011 | ||
7012 | void | |
7013 | print_time(hrtime_t t, char *timebuf) | |
7014 | { | |
7015 | hrtime_t s = t / NANOSEC; | |
7016 | hrtime_t m = s / 60; | |
7017 | hrtime_t h = m / 60; | |
7018 | hrtime_t d = h / 24; | |
7019 | ||
7020 | s -= m * 60; | |
7021 | m -= h * 60; | |
7022 | h -= d * 24; | |
7023 | ||
7024 | timebuf[0] = '\0'; | |
7025 | ||
7026 | if (d) | |
7027 | (void) sprintf(timebuf, | |
7028 | "%llud%02lluh%02llum%02llus", d, h, m, s); | |
7029 | else if (h) | |
7030 | (void) sprintf(timebuf, "%lluh%02llum%02llus", h, m, s); | |
7031 | else if (m) | |
7032 | (void) sprintf(timebuf, "%llum%02llus", m, s); | |
7033 | else | |
7034 | (void) sprintf(timebuf, "%llus", s); | |
7035 | } | |
7036 | ||
428870ff | 7037 | static nvlist_t * |
0bc8fd78 | 7038 | make_random_props(void) |
428870ff BB |
7039 | { |
7040 | nvlist_t *props; | |
7041 | ||
60b82074 | 7042 | VERIFY0(nvlist_alloc(&props, NV_UNIQUE_NAME, 0)); |
cc99f275 | 7043 | |
9ae529ec CS |
7044 | if (ztest_random(2) == 0) |
7045 | return (props); | |
60b82074 BB |
7046 | |
7047 | VERIFY0(nvlist_add_uint64(props, | |
7048 | zpool_prop_to_name(ZPOOL_PROP_AUTOREPLACE), 1)); | |
428870ff | 7049 | |
428870ff BB |
7050 | return (props); |
7051 | } | |
7052 | ||
379ca9cf OF |
7053 | /* |
7054 | * Import a storage pool with the given name. | |
7055 | */ | |
7056 | static void | |
7057 | ztest_import(ztest_shared_t *zs) | |
7058 | { | |
7059 | libzfs_handle_t *hdl; | |
7060 | importargs_t args = { 0 }; | |
7061 | spa_t *spa; | |
7062 | nvlist_t *cfg = NULL; | |
7063 | int nsearch = 1; | |
7064 | char *searchdirs[nsearch]; | |
7065 | char *name = ztest_opts.zo_pool; | |
7066 | int flags = ZFS_IMPORT_MISSING_LOG; | |
7067 | int error; | |
7068 | ||
7069 | mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); | |
d2734cce | 7070 | mutex_init(&ztest_checkpoint_lock, NULL, MUTEX_DEFAULT, NULL); |
2d9142c9 | 7071 | VERIFY0(pthread_rwlock_init(&ztest_name_lock, NULL)); |
379ca9cf OF |
7072 | |
7073 | kernel_init(FREAD | FWRITE); | |
7074 | hdl = libzfs_init(); | |
7075 | ||
7076 | searchdirs[0] = ztest_opts.zo_dir; | |
7077 | args.paths = nsearch; | |
7078 | args.path = searchdirs; | |
7079 | args.can_be_active = B_FALSE; | |
7080 | ||
7081 | error = zpool_tryimport(hdl, name, &cfg, &args); | |
7082 | if (error) | |
7083 | (void) fatal(0, "No pools found\n"); | |
7084 | ||
7085 | VERIFY0(spa_import(name, cfg, NULL, flags)); | |
7086 | VERIFY0(spa_open(name, &spa, FTAG)); | |
7087 | zs->zs_metaslab_sz = | |
7088 | 1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift; | |
7089 | spa_close(spa, FTAG); | |
7090 | ||
7091 | libzfs_fini(hdl); | |
7092 | kernel_fini(); | |
7093 | ||
7094 | if (!ztest_opts.zo_mmp_test) { | |
7095 | ztest_run_zdb(ztest_opts.zo_pool); | |
7096 | ztest_freeze(); | |
7097 | ztest_run_zdb(ztest_opts.zo_pool); | |
7098 | } | |
7099 | ||
2d9142c9 | 7100 | (void) pthread_rwlock_destroy(&ztest_name_lock); |
379ca9cf | 7101 | mutex_destroy(&ztest_vdev_lock); |
d2734cce | 7102 | mutex_destroy(&ztest_checkpoint_lock); |
379ca9cf OF |
7103 | } |
7104 | ||
34dc7c2f BB |
7105 | /* |
7106 | * Create a storage pool with the given name and initial vdev size. | |
428870ff | 7107 | * Then test spa_freeze() functionality. |
34dc7c2f BB |
7108 | */ |
7109 | static void | |
428870ff | 7110 | ztest_init(ztest_shared_t *zs) |
34dc7c2f BB |
7111 | { |
7112 | spa_t *spa; | |
428870ff | 7113 | nvlist_t *nvroot, *props; |
9ae529ec | 7114 | int i; |
428870ff | 7115 | |
c242c188 | 7116 | mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); |
d2734cce | 7117 | mutex_init(&ztest_checkpoint_lock, NULL, MUTEX_DEFAULT, NULL); |
2d9142c9 | 7118 | VERIFY0(pthread_rwlock_init(&ztest_name_lock, NULL)); |
34dc7c2f BB |
7119 | |
7120 | kernel_init(FREAD | FWRITE); | |
7121 | ||
7122 | /* | |
7123 | * Create the storage pool. | |
7124 | */ | |
c242c188 | 7125 | (void) spa_destroy(ztest_opts.zo_pool); |
428870ff BB |
7126 | ztest_shared->zs_vdev_next_leaf = 0; |
7127 | zs->zs_splits = 0; | |
c242c188 | 7128 | zs->zs_mirrors = ztest_opts.zo_mirrors; |
ea0b2538 | 7129 | nvroot = make_vdev_root(NULL, NULL, NULL, ztest_opts.zo_vdev_size, 0, |
cc99f275 | 7130 | NULL, ztest_opts.zo_raidz, zs->zs_mirrors, 1); |
428870ff | 7131 | props = make_random_props(); |
60b82074 BB |
7132 | |
7133 | /* | |
7134 | * We don't expect the pool to suspend unless maxfaults == 0, | |
7135 | * in which case ztest_fault_inject() temporarily takes away | |
7136 | * the only valid replica. | |
7137 | */ | |
7138 | VERIFY0(nvlist_add_uint64(props, | |
7139 | zpool_prop_to_name(ZPOOL_PROP_FAILUREMODE), | |
7140 | MAXFAULTS(zs) ? ZIO_FAILURE_MODE_PANIC : ZIO_FAILURE_MODE_WAIT)); | |
7141 | ||
9ae529ec CS |
7142 | for (i = 0; i < SPA_FEATURES; i++) { |
7143 | char *buf; | |
7144 | VERIFY3S(-1, !=, asprintf(&buf, "feature@%s", | |
7145 | spa_feature_table[i].fi_uname)); | |
7146 | VERIFY3U(0, ==, nvlist_add_uint64(props, buf, 0)); | |
7147 | free(buf); | |
7148 | } | |
60b82074 BB |
7149 | |
7150 | VERIFY0(spa_create(ztest_opts.zo_pool, nvroot, props, NULL, NULL)); | |
34dc7c2f | 7151 | nvlist_free(nvroot); |
85802aa4 | 7152 | nvlist_free(props); |
34dc7c2f | 7153 | |
c242c188 CS |
7154 | VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
7155 | zs->zs_metaslab_sz = | |
7156 | 1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift; | |
34dc7c2f BB |
7157 | spa_close(spa, FTAG); |
7158 | ||
7159 | kernel_fini(); | |
428870ff | 7160 | |
379ca9cf OF |
7161 | if (!ztest_opts.zo_mmp_test) { |
7162 | ztest_run_zdb(ztest_opts.zo_pool); | |
7163 | ztest_freeze(); | |
7164 | ztest_run_zdb(ztest_opts.zo_pool); | |
7165 | } | |
c242c188 | 7166 | |
2d9142c9 | 7167 | (void) pthread_rwlock_destroy(&ztest_name_lock); |
c242c188 | 7168 | mutex_destroy(&ztest_vdev_lock); |
d2734cce | 7169 | mutex_destroy(&ztest_checkpoint_lock); |
c242c188 CS |
7170 | } |
7171 | ||
7172 | static void | |
9d81146b | 7173 | setup_data_fd(void) |
c242c188 | 7174 | { |
facbbe43 BB |
7175 | static char ztest_name_data[] = "/tmp/ztest.data.XXXXXX"; |
7176 | ||
7177 | ztest_fd_data = mkstemp(ztest_name_data); | |
9d81146b | 7178 | ASSERT3S(ztest_fd_data, >=, 0); |
facbbe43 | 7179 | (void) unlink(ztest_name_data); |
c242c188 CS |
7180 | } |
7181 | ||
22257dc0 CS |
7182 | static int |
7183 | shared_data_size(ztest_shared_hdr_t *hdr) | |
7184 | { | |
7185 | int size; | |
7186 | ||
7187 | size = hdr->zh_hdr_size; | |
7188 | size += hdr->zh_opts_size; | |
7189 | size += hdr->zh_size; | |
7190 | size += hdr->zh_stats_size * hdr->zh_stats_count; | |
7191 | size += hdr->zh_ds_size * hdr->zh_ds_count; | |
7192 | ||
7193 | return (size); | |
7194 | } | |
7195 | ||
c242c188 CS |
7196 | static void |
7197 | setup_hdr(void) | |
7198 | { | |
22257dc0 | 7199 | int size; |
c242c188 CS |
7200 | ztest_shared_hdr_t *hdr; |
7201 | ||
7202 | hdr = (void *)mmap(0, P2ROUNDUP(sizeof (*hdr), getpagesize()), | |
9d81146b | 7203 | PROT_READ | PROT_WRITE, MAP_SHARED, ztest_fd_data, 0); |
799402d8 | 7204 | ASSERT(hdr != MAP_FAILED); |
c242c188 | 7205 | |
9d81146b | 7206 | VERIFY3U(0, ==, ftruncate(ztest_fd_data, sizeof (ztest_shared_hdr_t))); |
22257dc0 | 7207 | |
c242c188 CS |
7208 | hdr->zh_hdr_size = sizeof (ztest_shared_hdr_t); |
7209 | hdr->zh_opts_size = sizeof (ztest_shared_opts_t); | |
7210 | hdr->zh_size = sizeof (ztest_shared_t); | |
7211 | hdr->zh_stats_size = sizeof (ztest_shared_callstate_t); | |
7212 | hdr->zh_stats_count = ZTEST_FUNCS; | |
7213 | hdr->zh_ds_size = sizeof (ztest_shared_ds_t); | |
7214 | hdr->zh_ds_count = ztest_opts.zo_datasets; | |
7215 | ||
22257dc0 | 7216 | size = shared_data_size(hdr); |
9d81146b | 7217 | VERIFY3U(0, ==, ftruncate(ztest_fd_data, size)); |
22257dc0 | 7218 | |
c242c188 CS |
7219 | (void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize())); |
7220 | } | |
7221 | ||
7222 | static void | |
7223 | setup_data(void) | |
7224 | { | |
7225 | int size, offset; | |
7226 | ztest_shared_hdr_t *hdr; | |
7227 | uint8_t *buf; | |
7228 | ||
7229 | hdr = (void *)mmap(0, P2ROUNDUP(sizeof (*hdr), getpagesize()), | |
9d81146b | 7230 | PROT_READ, MAP_SHARED, ztest_fd_data, 0); |
799402d8 | 7231 | ASSERT(hdr != MAP_FAILED); |
c242c188 | 7232 | |
22257dc0 | 7233 | size = shared_data_size(hdr); |
c242c188 CS |
7234 | |
7235 | (void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize())); | |
7236 | hdr = ztest_shared_hdr = (void *)mmap(0, P2ROUNDUP(size, getpagesize()), | |
9d81146b | 7237 | PROT_READ | PROT_WRITE, MAP_SHARED, ztest_fd_data, 0); |
799402d8 | 7238 | ASSERT(hdr != MAP_FAILED); |
c242c188 CS |
7239 | buf = (uint8_t *)hdr; |
7240 | ||
7241 | offset = hdr->zh_hdr_size; | |
7242 | ztest_shared_opts = (void *)&buf[offset]; | |
7243 | offset += hdr->zh_opts_size; | |
7244 | ztest_shared = (void *)&buf[offset]; | |
7245 | offset += hdr->zh_size; | |
7246 | ztest_shared_callstate = (void *)&buf[offset]; | |
7247 | offset += hdr->zh_stats_size * hdr->zh_stats_count; | |
7248 | ztest_shared_ds = (void *)&buf[offset]; | |
7249 | } | |
7250 | ||
7251 | static boolean_t | |
7252 | exec_child(char *cmd, char *libpath, boolean_t ignorekill, int *statusp) | |
7253 | { | |
7254 | pid_t pid; | |
7255 | int status; | |
483106eb | 7256 | char *cmdbuf = NULL; |
c242c188 CS |
7257 | |
7258 | pid = fork(); | |
7259 | ||
7260 | if (cmd == NULL) { | |
483106eb BB |
7261 | cmdbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
7262 | (void) strlcpy(cmdbuf, getexecname(), MAXPATHLEN); | |
c242c188 CS |
7263 | cmd = cmdbuf; |
7264 | } | |
7265 | ||
7266 | if (pid == -1) | |
7267 | fatal(1, "fork failed"); | |
7268 | ||
7269 | if (pid == 0) { /* child */ | |
7270 | char *emptyargv[2] = { cmd, NULL }; | |
9d81146b | 7271 | char fd_data_str[12]; |
c242c188 CS |
7272 | |
7273 | struct rlimit rl = { 1024, 1024 }; | |
7274 | (void) setrlimit(RLIMIT_NOFILE, &rl); | |
9d81146b ED |
7275 | |
7276 | (void) close(ztest_fd_rand); | |
7277 | VERIFY(11 >= snprintf(fd_data_str, 12, "%d", ztest_fd_data)); | |
7278 | VERIFY(0 == setenv("ZTEST_FD_DATA", fd_data_str, 1)); | |
7279 | ||
c242c188 CS |
7280 | (void) enable_extended_FILE_stdio(-1, -1); |
7281 | if (libpath != NULL) | |
7282 | VERIFY(0 == setenv("LD_LIBRARY_PATH", libpath, 1)); | |
7283 | (void) execv(cmd, emptyargv); | |
7284 | ztest_dump_core = B_FALSE; | |
7285 | fatal(B_TRUE, "exec failed: %s", cmd); | |
7286 | } | |
7287 | ||
483106eb BB |
7288 | if (cmdbuf != NULL) { |
7289 | umem_free(cmdbuf, MAXPATHLEN); | |
7290 | cmd = NULL; | |
7291 | } | |
7292 | ||
c242c188 CS |
7293 | while (waitpid(pid, &status, 0) != pid) |
7294 | continue; | |
7295 | if (statusp != NULL) | |
7296 | *statusp = status; | |
7297 | ||
7298 | if (WIFEXITED(status)) { | |
7299 | if (WEXITSTATUS(status) != 0) { | |
7300 | (void) fprintf(stderr, "child exited with code %d\n", | |
7301 | WEXITSTATUS(status)); | |
7302 | exit(2); | |
7303 | } | |
7304 | return (B_FALSE); | |
7305 | } else if (WIFSIGNALED(status)) { | |
7306 | if (!ignorekill || WTERMSIG(status) != SIGKILL) { | |
7307 | (void) fprintf(stderr, "child died with signal %d\n", | |
7308 | WTERMSIG(status)); | |
7309 | exit(3); | |
7310 | } | |
7311 | return (B_TRUE); | |
7312 | } else { | |
7313 | (void) fprintf(stderr, "something strange happened to child\n"); | |
7314 | exit(4); | |
7315 | /* NOTREACHED */ | |
7316 | } | |
7317 | } | |
7318 | ||
7319 | static void | |
7320 | ztest_run_init(void) | |
7321 | { | |
7322 | int i; | |
428870ff | 7323 | |
c242c188 | 7324 | ztest_shared_t *zs = ztest_shared; |
428870ff | 7325 | |
c242c188 CS |
7326 | /* |
7327 | * Blow away any existing copy of zpool.cache | |
7328 | */ | |
7329 | (void) remove(spa_config_path); | |
7330 | ||
379ca9cf OF |
7331 | if (ztest_opts.zo_init == 0) { |
7332 | if (ztest_opts.zo_verbose >= 1) | |
7333 | (void) printf("Importing pool %s\n", | |
7334 | ztest_opts.zo_pool); | |
7335 | ztest_import(zs); | |
7336 | return; | |
7337 | } | |
7338 | ||
c242c188 CS |
7339 | /* |
7340 | * Create and initialize our storage pool. | |
7341 | */ | |
7342 | for (i = 1; i <= ztest_opts.zo_init; i++) { | |
7343 | bzero(zs, sizeof (ztest_shared_t)); | |
7344 | if (ztest_opts.zo_verbose >= 3 && | |
7345 | ztest_opts.zo_init != 1) { | |
7346 | (void) printf("ztest_init(), pass %d\n", i); | |
7347 | } | |
7348 | ztest_init(zs); | |
7349 | } | |
34dc7c2f BB |
7350 | } |
7351 | ||
7352 | int | |
7353 | main(int argc, char **argv) | |
7354 | { | |
7355 | int kills = 0; | |
7356 | int iters = 0; | |
c242c188 CS |
7357 | int older = 0; |
7358 | int newer = 0; | |
34dc7c2f BB |
7359 | ztest_shared_t *zs; |
7360 | ztest_info_t *zi; | |
c242c188 | 7361 | ztest_shared_callstate_t *zc; |
34dc7c2f | 7362 | char timebuf[100]; |
f3c8c9e6 | 7363 | char numbuf[NN_NUMBUF_SZ]; |
483106eb | 7364 | char *cmd; |
c242c188 CS |
7365 | boolean_t hasalt; |
7366 | int f; | |
9d81146b | 7367 | char *fd_data_str = getenv("ZTEST_FD_DATA"); |
e82cdc3a | 7368 | struct sigaction action; |
34dc7c2f BB |
7369 | |
7370 | (void) setvbuf(stdout, NULL, _IOLBF, 0); | |
7371 | ||
498877ba | 7372 | dprintf_setup(&argc, argv); |
8fb1ede1 | 7373 | zfs_deadman_synctime_ms = 300000; |
94102578 | 7374 | zfs_deadman_checktime_ms = 30000; |
4d044c4c SD |
7375 | /* |
7376 | * As two-word space map entries may not come up often (especially | |
7377 | * if pool and vdev sizes are small) we want to force at least some | |
7378 | * of them so the feature get tested. | |
7379 | */ | |
7380 | zfs_force_some_double_word_sm_entries = B_TRUE; | |
498877ba | 7381 | |
1258bd77 BB |
7382 | /* |
7383 | * Verify that even extensively damaged split blocks with many | |
7384 | * segments can be reconstructed in a reasonable amount of time | |
7385 | * when reconstruction is known to be possible. | |
7386 | */ | |
7387 | zfs_reconstruct_indirect_damage_fraction = 4; | |
7388 | ||
e82cdc3a NB |
7389 | action.sa_handler = sig_handler; |
7390 | sigemptyset(&action.sa_mask); | |
7391 | action.sa_flags = 0; | |
7392 | ||
7393 | if (sigaction(SIGSEGV, &action, NULL) < 0) { | |
7394 | (void) fprintf(stderr, "ztest: cannot catch SIGSEGV: %s.\n", | |
7395 | strerror(errno)); | |
7396 | exit(EXIT_FAILURE); | |
7397 | } | |
7398 | ||
7399 | if (sigaction(SIGABRT, &action, NULL) < 0) { | |
7400 | (void) fprintf(stderr, "ztest: cannot catch SIGABRT: %s.\n", | |
7401 | strerror(errno)); | |
7402 | exit(EXIT_FAILURE); | |
7403 | } | |
7404 | ||
e1a0850c BB |
7405 | /* |
7406 | * Force random_get_bytes() to use /dev/urandom in order to prevent | |
7407 | * ztest from needlessly depleting the system entropy pool. | |
7408 | */ | |
7409 | random_path = "/dev/urandom"; | |
7410 | ztest_fd_rand = open(random_path, O_RDONLY); | |
9d81146b ED |
7411 | ASSERT3S(ztest_fd_rand, >=, 0); |
7412 | ||
7413 | if (!fd_data_str) { | |
c242c188 | 7414 | process_options(argc, argv); |
34dc7c2f | 7415 | |
9d81146b | 7416 | setup_data_fd(); |
c242c188 CS |
7417 | setup_hdr(); |
7418 | setup_data(); | |
7419 | bcopy(&ztest_opts, ztest_shared_opts, | |
7420 | sizeof (*ztest_shared_opts)); | |
7421 | } else { | |
9d81146b | 7422 | ztest_fd_data = atoi(fd_data_str); |
c242c188 CS |
7423 | setup_data(); |
7424 | bcopy(ztest_shared_opts, &ztest_opts, sizeof (ztest_opts)); | |
7425 | } | |
7426 | ASSERT3U(ztest_opts.zo_datasets, ==, ztest_shared_hdr->zh_ds_count); | |
34dc7c2f | 7427 | |
428870ff | 7428 | /* Override location of zpool.cache */ |
5be98cfe BB |
7429 | VERIFY(asprintf((char **)&spa_config_path, "%s/zpool.cache", |
7430 | ztest_opts.zo_dir) != -1); | |
9d81146b | 7431 | |
c242c188 CS |
7432 | ztest_ds = umem_alloc(ztest_opts.zo_datasets * sizeof (ztest_ds_t), |
7433 | UMEM_NOFAIL); | |
7434 | zs = ztest_shared; | |
7435 | ||
9d81146b | 7436 | if (fd_data_str) { |
d830d479 | 7437 | metaslab_force_ganging = ztest_opts.zo_metaslab_force_ganging; |
c242c188 CS |
7438 | metaslab_df_alloc_threshold = |
7439 | zs->zs_metaslab_df_alloc_threshold; | |
7440 | ||
7441 | if (zs->zs_do_init) | |
7442 | ztest_run_init(); | |
7443 | else | |
7444 | ztest_run(zs); | |
7445 | exit(0); | |
7446 | } | |
34dc7c2f | 7447 | |
c242c188 | 7448 | hasalt = (strlen(ztest_opts.zo_alt_ztest) != 0); |
34dc7c2f | 7449 | |
c242c188 | 7450 | if (ztest_opts.zo_verbose >= 1) { |
34dc7c2f BB |
7451 | (void) printf("%llu vdevs, %d datasets, %d threads," |
7452 | " %llu seconds...\n", | |
c242c188 CS |
7453 | (u_longlong_t)ztest_opts.zo_vdevs, |
7454 | ztest_opts.zo_datasets, | |
7455 | ztest_opts.zo_threads, | |
7456 | (u_longlong_t)ztest_opts.zo_time); | |
34dc7c2f BB |
7457 | } |
7458 | ||
483106eb BB |
7459 | cmd = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); |
7460 | (void) strlcpy(cmd, getexecname(), MAXNAMELEN); | |
c242c188 CS |
7461 | |
7462 | zs->zs_do_init = B_TRUE; | |
7463 | if (strlen(ztest_opts.zo_alt_ztest) != 0) { | |
7464 | if (ztest_opts.zo_verbose >= 1) { | |
7465 | (void) printf("Executing older ztest for " | |
7466 | "initialization: %s\n", ztest_opts.zo_alt_ztest); | |
7467 | } | |
7468 | VERIFY(!exec_child(ztest_opts.zo_alt_ztest, | |
7469 | ztest_opts.zo_alt_libpath, B_FALSE, NULL)); | |
7470 | } else { | |
7471 | VERIFY(!exec_child(NULL, NULL, B_FALSE, NULL)); | |
34dc7c2f | 7472 | } |
c242c188 | 7473 | zs->zs_do_init = B_FALSE; |
34dc7c2f | 7474 | |
428870ff | 7475 | zs->zs_proc_start = gethrtime(); |
c242c188 | 7476 | zs->zs_proc_stop = zs->zs_proc_start + ztest_opts.zo_time * NANOSEC; |
34dc7c2f | 7477 | |
d6320ddb | 7478 | for (f = 0; f < ZTEST_FUNCS; f++) { |
c242c188 CS |
7479 | zi = &ztest_info[f]; |
7480 | zc = ZTEST_GET_SHARED_CALLSTATE(f); | |
428870ff | 7481 | if (zs->zs_proc_start + zi->zi_interval[0] > zs->zs_proc_stop) |
c242c188 | 7482 | zc->zc_next = UINT64_MAX; |
34dc7c2f | 7483 | else |
c242c188 | 7484 | zc->zc_next = zs->zs_proc_start + |
428870ff | 7485 | ztest_random(2 * zi->zi_interval[0] + 1); |
34dc7c2f BB |
7486 | } |
7487 | ||
34dc7c2f BB |
7488 | /* |
7489 | * Run the tests in a loop. These tests include fault injection | |
7490 | * to verify that self-healing data works, and forced crashes | |
7491 | * to verify that we never lose on-disk consistency. | |
7492 | */ | |
428870ff | 7493 | while (gethrtime() < zs->zs_proc_stop) { |
34dc7c2f | 7494 | int status; |
c242c188 | 7495 | boolean_t killed; |
34dc7c2f BB |
7496 | |
7497 | /* | |
7498 | * Initialize the workload counters for each function. | |
7499 | */ | |
d6320ddb | 7500 | for (f = 0; f < ZTEST_FUNCS; f++) { |
c242c188 CS |
7501 | zc = ZTEST_GET_SHARED_CALLSTATE(f); |
7502 | zc->zc_count = 0; | |
7503 | zc->zc_time = 0; | |
34dc7c2f BB |
7504 | } |
7505 | ||
9babb374 | 7506 | /* Set the allocation switch size */ |
c242c188 CS |
7507 | zs->zs_metaslab_df_alloc_threshold = |
7508 | ztest_random(zs->zs_metaslab_sz / 4) + 1; | |
9babb374 | 7509 | |
c242c188 CS |
7510 | if (!hasalt || ztest_random(2) == 0) { |
7511 | if (hasalt && ztest_opts.zo_verbose >= 1) { | |
7512 | (void) printf("Executing newer ztest: %s\n", | |
7513 | cmd); | |
34dc7c2f | 7514 | } |
c242c188 CS |
7515 | newer++; |
7516 | killed = exec_child(cmd, NULL, B_TRUE, &status); | |
34dc7c2f | 7517 | } else { |
c242c188 CS |
7518 | if (hasalt && ztest_opts.zo_verbose >= 1) { |
7519 | (void) printf("Executing older ztest: %s\n", | |
7520 | ztest_opts.zo_alt_ztest); | |
7521 | } | |
7522 | older++; | |
7523 | killed = exec_child(ztest_opts.zo_alt_ztest, | |
7524 | ztest_opts.zo_alt_libpath, B_TRUE, &status); | |
34dc7c2f BB |
7525 | } |
7526 | ||
c242c188 CS |
7527 | if (killed) |
7528 | kills++; | |
34dc7c2f BB |
7529 | iters++; |
7530 | ||
c242c188 | 7531 | if (ztest_opts.zo_verbose >= 1) { |
34dc7c2f BB |
7532 | hrtime_t now = gethrtime(); |
7533 | ||
428870ff BB |
7534 | now = MIN(now, zs->zs_proc_stop); |
7535 | print_time(zs->zs_proc_stop - now, timebuf); | |
f3c8c9e6 | 7536 | nicenum(zs->zs_space, numbuf, sizeof (numbuf)); |
34dc7c2f BB |
7537 | |
7538 | (void) printf("Pass %3d, %8s, %3llu ENOSPC, " | |
7539 | "%4.1f%% of %5s used, %3.0f%% done, %8s to go\n", | |
7540 | iters, | |
7541 | WIFEXITED(status) ? "Complete" : "SIGKILL", | |
7542 | (u_longlong_t)zs->zs_enospc_count, | |
7543 | 100.0 * zs->zs_alloc / zs->zs_space, | |
7544 | numbuf, | |
428870ff | 7545 | 100.0 * (now - zs->zs_proc_start) / |
c242c188 | 7546 | (ztest_opts.zo_time * NANOSEC), timebuf); |
34dc7c2f BB |
7547 | } |
7548 | ||
c242c188 | 7549 | if (ztest_opts.zo_verbose >= 2) { |
34dc7c2f BB |
7550 | (void) printf("\nWorkload summary:\n\n"); |
7551 | (void) printf("%7s %9s %s\n", | |
7552 | "Calls", "Time", "Function"); | |
7553 | (void) printf("%7s %9s %s\n", | |
7554 | "-----", "----", "--------"); | |
d6320ddb | 7555 | for (f = 0; f < ZTEST_FUNCS; f++) { |
c242c188 CS |
7556 | zi = &ztest_info[f]; |
7557 | zc = ZTEST_GET_SHARED_CALLSTATE(f); | |
7558 | print_time(zc->zc_time, timebuf); | |
34dc7c2f | 7559 | (void) printf("%7llu %9s %s\n", |
c242c188 | 7560 | (u_longlong_t)zc->zc_count, timebuf, |
fdc5d982 | 7561 | zi->zi_funcname); |
34dc7c2f BB |
7562 | } |
7563 | (void) printf("\n"); | |
7564 | } | |
7565 | ||
379ca9cf OF |
7566 | if (!ztest_opts.zo_mmp_test) |
7567 | ztest_run_zdb(ztest_opts.zo_pool); | |
428870ff | 7568 | } |
34dc7c2f | 7569 | |
c242c188 CS |
7570 | if (ztest_opts.zo_verbose >= 1) { |
7571 | if (hasalt) { | |
7572 | (void) printf("%d runs of older ztest: %s\n", older, | |
7573 | ztest_opts.zo_alt_ztest); | |
7574 | (void) printf("%d runs of newer ztest: %s\n", newer, | |
7575 | cmd); | |
7576 | } | |
34dc7c2f BB |
7577 | (void) printf("%d killed, %d completed, %.0f%% kill rate\n", |
7578 | kills, iters - kills, (100.0 * kills) / MAX(1, iters)); | |
7579 | } | |
7580 | ||
483106eb BB |
7581 | umem_free(cmd, MAXNAMELEN); |
7582 | ||
34dc7c2f BB |
7583 | return (0); |
7584 | } |