]>
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. |
546d32ca | 23 | * Copyright (c) 2011, 2016 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. | |
34dc7c2f BB |
28 | */ |
29 | ||
34dc7c2f BB |
30 | /* |
31 | * The objective of this program is to provide a DMU/ZAP/SPA stress test | |
32 | * that runs entirely in userland, is easy to use, and easy to extend. | |
33 | * | |
34 | * The overall design of the ztest program is as follows: | |
35 | * | |
36 | * (1) For each major functional area (e.g. adding vdevs to a pool, | |
37 | * creating and destroying datasets, reading and writing objects, etc) | |
38 | * we have a simple routine to test that functionality. These | |
39 | * individual routines do not have to do anything "stressful". | |
40 | * | |
41 | * (2) We turn these simple functionality tests into a stress test by | |
42 | * running them all in parallel, with as many threads as desired, | |
43 | * and spread across as many datasets, objects, and vdevs as desired. | |
44 | * | |
45 | * (3) While all this is happening, we inject faults into the pool to | |
46 | * verify that self-healing data really works. | |
47 | * | |
48 | * (4) Every time we open a dataset, we change its checksum and compression | |
49 | * functions. Thus even individual objects vary from block to block | |
50 | * in which checksum they use and whether they're compressed. | |
51 | * | |
52 | * (5) To verify that we never lose on-disk consistency after a crash, | |
53 | * we run the entire test in a child of the main process. | |
54 | * At random times, the child self-immolates with a SIGKILL. | |
55 | * This is the software equivalent of pulling the power cord. | |
56 | * The parent then runs the test again, using the existing | |
9b67f605 | 57 | * storage pool, as many times as desired. If backwards compatibility |
c242c188 CS |
58 | * testing is enabled ztest will sometimes run the "older" version |
59 | * of ztest after a SIGKILL. | |
34dc7c2f BB |
60 | * |
61 | * (6) To verify that we don't have future leaks or temporal incursions, | |
62 | * many of the functional tests record the transaction group number | |
63 | * as part of their data. When reading old data, they verify that | |
64 | * the transaction group number is less than the current, open txg. | |
65 | * If you add a new test, please do this if applicable. | |
66 | * | |
1e33ac1e BB |
67 | * (7) Threads are created with a reduced stack size, for sanity checking. |
68 | * Therefore, it's important not to allocate huge buffers on the stack. | |
69 | * | |
34dc7c2f BB |
70 | * When run with no arguments, ztest runs for about five minutes and |
71 | * produces no output if successful. To get a little bit of information, | |
72 | * specify -V. To get more information, specify -VV, and so on. | |
73 | * | |
74 | * To turn this into an overnight stress test, use -T to specify run time. | |
75 | * | |
76 | * You can ask more more vdevs [-v], datasets [-d], or threads [-t] | |
77 | * to increase the pool capacity, fanout, and overall stress level. | |
78 | * | |
c242c188 CS |
79 | * Use the -k option to set the desired frequency of kills. |
80 | * | |
81 | * When ztest invokes itself it passes all relevant information through a | |
82 | * temporary file which is mmap-ed in the child process. This allows shared | |
83 | * memory to survive the exec syscall. The ztest_shared_hdr_t struct is always | |
84 | * stored at offset 0 of this file and contains information on the size and | |
85 | * number of shared structures in the file. The information stored in this file | |
86 | * must remain backwards compatible with older versions of ztest so that | |
87 | * ztest can invoke them during backwards compatibility testing (-B). | |
34dc7c2f BB |
88 | */ |
89 | ||
90 | #include <sys/zfs_context.h> | |
91 | #include <sys/spa.h> | |
92 | #include <sys/dmu.h> | |
93 | #include <sys/txg.h> | |
9babb374 | 94 | #include <sys/dbuf.h> |
34dc7c2f | 95 | #include <sys/zap.h> |
34dc7c2f BB |
96 | #include <sys/dmu_objset.h> |
97 | #include <sys/poll.h> | |
98 | #include <sys/stat.h> | |
99 | #include <sys/time.h> | |
100 | #include <sys/wait.h> | |
101 | #include <sys/mman.h> | |
102 | #include <sys/resource.h> | |
103 | #include <sys/zio.h> | |
34dc7c2f | 104 | #include <sys/zil.h> |
428870ff | 105 | #include <sys/zil_impl.h> |
e3a07cd0 | 106 | #include <sys/zfs_rlock.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> |
967798d0 | 132 | #ifdef __GLIBC__ |
e82cdc3a NB |
133 | #include <execinfo.h> /* for backtrace() */ |
134 | #endif | |
34dc7c2f | 135 | |
9d81146b ED |
136 | static int ztest_fd_data = -1; |
137 | static int ztest_fd_rand = -1; | |
c242c188 CS |
138 | |
139 | typedef struct ztest_shared_hdr { | |
140 | uint64_t zh_hdr_size; | |
141 | uint64_t zh_opts_size; | |
142 | uint64_t zh_size; | |
143 | uint64_t zh_stats_size; | |
144 | uint64_t zh_stats_count; | |
145 | uint64_t zh_ds_size; | |
146 | uint64_t zh_ds_count; | |
147 | } ztest_shared_hdr_t; | |
148 | ||
149 | static ztest_shared_hdr_t *ztest_shared_hdr; | |
150 | ||
151 | typedef struct ztest_shared_opts { | |
eca7b760 IK |
152 | char zo_pool[ZFS_MAX_DATASET_NAME_LEN]; |
153 | char zo_dir[ZFS_MAX_DATASET_NAME_LEN]; | |
c242c188 CS |
154 | char zo_alt_ztest[MAXNAMELEN]; |
155 | char zo_alt_libpath[MAXNAMELEN]; | |
156 | uint64_t zo_vdevs; | |
157 | uint64_t zo_vdevtime; | |
158 | size_t zo_vdev_size; | |
159 | int zo_ashift; | |
160 | int zo_mirrors; | |
161 | int zo_raidz; | |
162 | int zo_raidz_parity; | |
163 | int zo_datasets; | |
164 | int zo_threads; | |
165 | uint64_t zo_passtime; | |
166 | uint64_t zo_killrate; | |
167 | int zo_verbose; | |
168 | int zo_init; | |
169 | uint64_t zo_time; | |
170 | uint64_t zo_maxloops; | |
171 | uint64_t zo_metaslab_gang_bang; | |
379ca9cf | 172 | int zo_mmp_test; |
c242c188 CS |
173 | } ztest_shared_opts_t; |
174 | ||
175 | static const ztest_shared_opts_t ztest_opts_defaults = { | |
176 | .zo_pool = { 'z', 't', 'e', 's', 't', '\0' }, | |
177 | .zo_dir = { '/', 't', 'm', 'p', '\0' }, | |
178 | .zo_alt_ztest = { '\0' }, | |
179 | .zo_alt_libpath = { '\0' }, | |
180 | .zo_vdevs = 5, | |
181 | .zo_ashift = SPA_MINBLOCKSHIFT, | |
182 | .zo_mirrors = 2, | |
183 | .zo_raidz = 4, | |
184 | .zo_raidz_parity = 1, | |
4e21fd06 | 185 | .zo_vdev_size = SPA_MINDEVSIZE * 4, /* 256m default size */ |
c242c188 CS |
186 | .zo_datasets = 7, |
187 | .zo_threads = 23, | |
188 | .zo_passtime = 60, /* 60 seconds */ | |
189 | .zo_killrate = 70, /* 70% kill rate */ | |
190 | .zo_verbose = 0, | |
379ca9cf | 191 | .zo_mmp_test = 0, |
c242c188 CS |
192 | .zo_init = 1, |
193 | .zo_time = 300, /* 5 minutes */ | |
194 | .zo_maxloops = 50, /* max loops during spa_freeze() */ | |
195 | .zo_metaslab_gang_bang = 32 << 10 | |
196 | }; | |
197 | ||
198 | extern uint64_t metaslab_gang_bang; | |
199 | extern uint64_t metaslab_df_alloc_threshold; | |
8fb1ede1 | 200 | extern unsigned long zfs_deadman_synctime_ms; |
080b3100 | 201 | extern int metaslab_preload_limit; |
d3c2ae1c | 202 | extern boolean_t zfs_compressed_arc_enabled; |
047116ac TC |
203 | extern int zfs_abd_scatter_enabled; |
204 | extern int dmu_object_alloc_chunk_shift; | |
c242c188 CS |
205 | |
206 | static ztest_shared_opts_t *ztest_shared_opts; | |
207 | static ztest_shared_opts_t ztest_opts; | |
4807c0ba | 208 | static char *ztest_wkeydata = "abcdefghijklmnopqrstuvwxyz012345"; |
c242c188 CS |
209 | |
210 | typedef struct ztest_shared_ds { | |
211 | uint64_t zd_seq; | |
212 | } ztest_shared_ds_t; | |
213 | ||
214 | static ztest_shared_ds_t *ztest_shared_ds; | |
215 | #define ZTEST_GET_SHARED_DS(d) (&ztest_shared_ds[d]) | |
428870ff BB |
216 | |
217 | #define BT_MAGIC 0x123456789abcdefULL | |
60b82074 BB |
218 | #define MAXFAULTS(zs) \ |
219 | (MAX((zs)->zs_mirrors, 1) * (ztest_opts.zo_raidz_parity + 1) - 1) | |
428870ff BB |
220 | |
221 | enum ztest_io_type { | |
222 | ZTEST_IO_WRITE_TAG, | |
223 | ZTEST_IO_WRITE_PATTERN, | |
224 | ZTEST_IO_WRITE_ZEROES, | |
225 | ZTEST_IO_TRUNCATE, | |
226 | ZTEST_IO_SETATTR, | |
03c6040b | 227 | ZTEST_IO_REWRITE, |
428870ff BB |
228 | ZTEST_IO_TYPES |
229 | }; | |
34dc7c2f BB |
230 | |
231 | typedef struct ztest_block_tag { | |
428870ff | 232 | uint64_t bt_magic; |
34dc7c2f BB |
233 | uint64_t bt_objset; |
234 | uint64_t bt_object; | |
50c957f7 | 235 | uint64_t bt_dnodesize; |
34dc7c2f | 236 | uint64_t bt_offset; |
428870ff | 237 | uint64_t bt_gen; |
34dc7c2f | 238 | uint64_t bt_txg; |
428870ff | 239 | uint64_t bt_crtxg; |
34dc7c2f BB |
240 | } ztest_block_tag_t; |
241 | ||
428870ff BB |
242 | typedef struct bufwad { |
243 | uint64_t bw_index; | |
244 | uint64_t bw_txg; | |
245 | uint64_t bw_data; | |
246 | } bufwad_t; | |
247 | ||
428870ff BB |
248 | typedef struct rll { |
249 | void *rll_writer; | |
250 | int rll_readers; | |
1e33ac1e BB |
251 | kmutex_t rll_lock; |
252 | kcondvar_t rll_cv; | |
428870ff BB |
253 | } rll_t; |
254 | ||
e3a07cd0 BP |
255 | typedef struct zll { |
256 | list_t z_list; | |
257 | kmutex_t z_lock; | |
258 | } zll_t; | |
428870ff BB |
259 | |
260 | #define ZTEST_RANGE_LOCKS 64 | |
261 | #define ZTEST_OBJECT_LOCKS 64 | |
262 | ||
263 | /* | |
264 | * Object descriptor. Used as a template for object lookup/create/remove. | |
265 | */ | |
266 | typedef struct ztest_od { | |
267 | uint64_t od_dir; | |
268 | uint64_t od_object; | |
269 | dmu_object_type_t od_type; | |
270 | dmu_object_type_t od_crtype; | |
271 | uint64_t od_blocksize; | |
272 | uint64_t od_crblocksize; | |
50c957f7 | 273 | uint64_t od_crdnodesize; |
428870ff BB |
274 | uint64_t od_gen; |
275 | uint64_t od_crgen; | |
eca7b760 | 276 | char od_name[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff | 277 | } ztest_od_t; |
34dc7c2f | 278 | |
428870ff BB |
279 | /* |
280 | * Per-dataset state. | |
281 | */ | |
282 | typedef struct ztest_ds { | |
c242c188 | 283 | ztest_shared_ds_t *zd_shared; |
428870ff | 284 | objset_t *zd_os; |
7809eb8b | 285 | rwlock_t zd_zilog_lock; |
428870ff | 286 | zilog_t *zd_zilog; |
428870ff | 287 | ztest_od_t *zd_od; /* debugging aid */ |
eca7b760 | 288 | char zd_name[ZFS_MAX_DATASET_NAME_LEN]; |
1e33ac1e | 289 | kmutex_t zd_dirobj_lock; |
428870ff | 290 | rll_t zd_object_lock[ZTEST_OBJECT_LOCKS]; |
e3a07cd0 | 291 | zll_t zd_range_lock[ZTEST_RANGE_LOCKS]; |
428870ff BB |
292 | } ztest_ds_t; |
293 | ||
294 | /* | |
295 | * Per-iteration state. | |
296 | */ | |
297 | typedef void ztest_func_t(ztest_ds_t *zd, uint64_t id); | |
298 | ||
299 | typedef struct ztest_info { | |
300 | ztest_func_t *zi_func; /* test function */ | |
301 | uint64_t zi_iters; /* iterations per execution */ | |
302 | uint64_t *zi_interval; /* execute every <interval> seconds */ | |
fdc5d982 | 303 | const char *zi_funcname; /* name of test function */ |
428870ff | 304 | } ztest_info_t; |
34dc7c2f | 305 | |
c242c188 CS |
306 | typedef struct ztest_shared_callstate { |
307 | uint64_t zc_count; /* per-pass count */ | |
308 | uint64_t zc_time; /* per-pass time */ | |
309 | uint64_t zc_next; /* next time to call this function */ | |
310 | } ztest_shared_callstate_t; | |
311 | ||
312 | static ztest_shared_callstate_t *ztest_shared_callstate; | |
313 | #define ZTEST_GET_SHARED_CALLSTATE(c) (&ztest_shared_callstate[c]) | |
314 | ||
34dc7c2f BB |
315 | ztest_func_t ztest_dmu_read_write; |
316 | ztest_func_t ztest_dmu_write_parallel; | |
317 | ztest_func_t ztest_dmu_object_alloc_free; | |
047116ac | 318 | ztest_func_t ztest_dmu_object_next_chunk; |
428870ff | 319 | ztest_func_t ztest_dmu_commit_callbacks; |
34dc7c2f BB |
320 | ztest_func_t ztest_zap; |
321 | ztest_func_t ztest_zap_parallel; | |
428870ff | 322 | ztest_func_t ztest_zil_commit; |
3e31d2b0 | 323 | ztest_func_t ztest_zil_remount; |
428870ff | 324 | ztest_func_t ztest_dmu_read_write_zcopy; |
34dc7c2f | 325 | ztest_func_t ztest_dmu_objset_create_destroy; |
428870ff BB |
326 | ztest_func_t ztest_dmu_prealloc; |
327 | ztest_func_t ztest_fzap; | |
34dc7c2f | 328 | ztest_func_t ztest_dmu_snapshot_create_destroy; |
428870ff BB |
329 | ztest_func_t ztest_dsl_prop_get_set; |
330 | ztest_func_t ztest_spa_prop_get_set; | |
34dc7c2f BB |
331 | ztest_func_t ztest_spa_create_destroy; |
332 | ztest_func_t ztest_fault_inject; | |
428870ff BB |
333 | ztest_func_t ztest_ddt_repair; |
334 | ztest_func_t ztest_dmu_snapshot_hold; | |
0582e403 | 335 | ztest_func_t ztest_mmp_enable_disable; |
b128c09f | 336 | ztest_func_t ztest_spa_rename; |
428870ff BB |
337 | ztest_func_t ztest_scrub; |
338 | ztest_func_t ztest_dsl_dataset_promote_busy; | |
34dc7c2f BB |
339 | ztest_func_t ztest_vdev_attach_detach; |
340 | ztest_func_t ztest_vdev_LUN_growth; | |
341 | ztest_func_t ztest_vdev_add_remove; | |
b128c09f | 342 | ztest_func_t ztest_vdev_aux_add_remove; |
428870ff | 343 | ztest_func_t ztest_split_pool; |
3541dc6d | 344 | ztest_func_t ztest_reguid; |
ea0b2538 | 345 | ztest_func_t ztest_spa_upgrade; |
1eeb4562 | 346 | ztest_func_t ztest_fletcher; |
37f520db | 347 | ztest_func_t ztest_fletcher_incr; |
50c957f7 | 348 | ztest_func_t ztest_verify_dnode_bt; |
34dc7c2f | 349 | |
428870ff BB |
350 | uint64_t zopt_always = 0ULL * NANOSEC; /* all the time */ |
351 | uint64_t zopt_incessant = 1ULL * NANOSEC / 10; /* every 1/10 second */ | |
352 | uint64_t zopt_often = 1ULL * NANOSEC; /* every second */ | |
353 | uint64_t zopt_sometimes = 10ULL * NANOSEC; /* every 10 seconds */ | |
354 | uint64_t zopt_rarely = 60ULL * NANOSEC; /* every 60 seconds */ | |
34dc7c2f | 355 | |
fdc5d982 TC |
356 | #define ZTI_INIT(func, iters, interval) \ |
357 | { .zi_func = (func), \ | |
358 | .zi_iters = (iters), \ | |
359 | .zi_interval = (interval), \ | |
360 | .zi_funcname = # func } | |
361 | ||
34dc7c2f | 362 | ztest_info_t ztest_info[] = { |
fdc5d982 TC |
363 | ZTI_INIT(ztest_dmu_read_write, 1, &zopt_always), |
364 | ZTI_INIT(ztest_dmu_write_parallel, 10, &zopt_always), | |
365 | ZTI_INIT(ztest_dmu_object_alloc_free, 1, &zopt_always), | |
047116ac | 366 | ZTI_INIT(ztest_dmu_object_next_chunk, 1, &zopt_sometimes), |
fdc5d982 TC |
367 | ZTI_INIT(ztest_dmu_commit_callbacks, 1, &zopt_always), |
368 | ZTI_INIT(ztest_zap, 30, &zopt_always), | |
369 | ZTI_INIT(ztest_zap_parallel, 100, &zopt_always), | |
370 | ZTI_INIT(ztest_split_pool, 1, &zopt_always), | |
371 | ZTI_INIT(ztest_zil_commit, 1, &zopt_incessant), | |
372 | ZTI_INIT(ztest_zil_remount, 1, &zopt_sometimes), | |
373 | ZTI_INIT(ztest_dmu_read_write_zcopy, 1, &zopt_often), | |
374 | ZTI_INIT(ztest_dmu_objset_create_destroy, 1, &zopt_often), | |
375 | ZTI_INIT(ztest_dsl_prop_get_set, 1, &zopt_often), | |
376 | ZTI_INIT(ztest_spa_prop_get_set, 1, &zopt_sometimes), | |
428870ff | 377 | #if 0 |
fdc5d982 | 378 | ZTI_INIT(ztest_dmu_prealloc, 1, &zopt_sometimes), |
428870ff | 379 | #endif |
fdc5d982 TC |
380 | ZTI_INIT(ztest_fzap, 1, &zopt_sometimes), |
381 | ZTI_INIT(ztest_dmu_snapshot_create_destroy, 1, &zopt_sometimes), | |
382 | ZTI_INIT(ztest_spa_create_destroy, 1, &zopt_sometimes), | |
383 | ZTI_INIT(ztest_fault_inject, 1, &zopt_sometimes), | |
384 | ZTI_INIT(ztest_ddt_repair, 1, &zopt_sometimes), | |
385 | ZTI_INIT(ztest_dmu_snapshot_hold, 1, &zopt_sometimes), | |
0582e403 | 386 | ZTI_INIT(ztest_mmp_enable_disable, 1, &zopt_sometimes), |
fdc5d982 TC |
387 | ZTI_INIT(ztest_reguid, 1, &zopt_rarely), |
388 | ZTI_INIT(ztest_spa_rename, 1, &zopt_rarely), | |
389 | ZTI_INIT(ztest_scrub, 1, &zopt_rarely), | |
390 | ZTI_INIT(ztest_spa_upgrade, 1, &zopt_rarely), | |
391 | ZTI_INIT(ztest_dsl_dataset_promote_busy, 1, &zopt_rarely), | |
392 | ZTI_INIT(ztest_vdev_attach_detach, 1, &zopt_sometimes), | |
393 | ZTI_INIT(ztest_vdev_LUN_growth, 1, &zopt_rarely), | |
394 | ZTI_INIT(ztest_vdev_add_remove, 1, &ztest_opts.zo_vdevtime), | |
395 | ZTI_INIT(ztest_vdev_aux_add_remove, 1, &ztest_opts.zo_vdevtime), | |
1eeb4562 | 396 | ZTI_INIT(ztest_fletcher, 1, &zopt_rarely), |
37f520db | 397 | ZTI_INIT(ztest_fletcher_incr, 1, &zopt_rarely), |
50c957f7 | 398 | ZTI_INIT(ztest_verify_dnode_bt, 1, &zopt_sometimes), |
34dc7c2f BB |
399 | }; |
400 | ||
401 | #define ZTEST_FUNCS (sizeof (ztest_info) / sizeof (ztest_info_t)) | |
402 | ||
428870ff BB |
403 | /* |
404 | * The following struct is used to hold a list of uncalled commit callbacks. | |
405 | * The callbacks are ordered by txg number. | |
406 | */ | |
407 | typedef struct ztest_cb_list { | |
1e33ac1e BB |
408 | kmutex_t zcl_callbacks_lock; |
409 | list_t zcl_callbacks; | |
428870ff | 410 | } ztest_cb_list_t; |
34dc7c2f BB |
411 | |
412 | /* | |
413 | * Stuff we need to share writably between parent and child. | |
414 | */ | |
415 | typedef struct ztest_shared { | |
c242c188 | 416 | boolean_t zs_do_init; |
428870ff BB |
417 | hrtime_t zs_proc_start; |
418 | hrtime_t zs_proc_stop; | |
419 | hrtime_t zs_thread_start; | |
420 | hrtime_t zs_thread_stop; | |
421 | hrtime_t zs_thread_kill; | |
34dc7c2f | 422 | uint64_t zs_enospc_count; |
428870ff BB |
423 | uint64_t zs_vdev_next_leaf; |
424 | uint64_t zs_vdev_aux; | |
34dc7c2f BB |
425 | uint64_t zs_alloc; |
426 | uint64_t zs_space; | |
428870ff BB |
427 | uint64_t zs_splits; |
428 | uint64_t zs_mirrors; | |
c242c188 CS |
429 | uint64_t zs_metaslab_sz; |
430 | uint64_t zs_metaslab_df_alloc_threshold; | |
431 | uint64_t zs_guid; | |
34dc7c2f BB |
432 | } ztest_shared_t; |
433 | ||
428870ff BB |
434 | #define ID_PARALLEL -1ULL |
435 | ||
34dc7c2f | 436 | static char ztest_dev_template[] = "%s/%s.%llua"; |
b128c09f | 437 | static char ztest_aux_template[] = "%s/%s.%s.%llu"; |
428870ff | 438 | ztest_shared_t *ztest_shared; |
34dc7c2f | 439 | |
c242c188 CS |
440 | static spa_t *ztest_spa = NULL; |
441 | static ztest_ds_t *ztest_ds; | |
442 | ||
443 | static kmutex_t ztest_vdev_lock; | |
3bc7e0fb GW |
444 | |
445 | /* | |
446 | * The ztest_name_lock protects the pool and dataset namespace used by | |
447 | * the individual tests. To modify the namespace, consumers must grab | |
448 | * this lock as writer. Grabbing the lock as reader will ensure that the | |
449 | * namespace does not change while the lock is held. | |
450 | */ | |
7809eb8b | 451 | static rwlock_t ztest_name_lock; |
34dc7c2f | 452 | |
c242c188 | 453 | static boolean_t ztest_dump_core = B_TRUE; |
8fb1ede1 | 454 | static boolean_t ztest_dump_debug_buffer = B_FALSE; |
b128c09f | 455 | static boolean_t ztest_exiting; |
34dc7c2f | 456 | |
428870ff BB |
457 | /* Global commit callback list */ |
458 | static ztest_cb_list_t zcl; | |
090ff092 RC |
459 | /* Commit cb delay */ |
460 | static uint64_t zc_min_txg_delay = UINT64_MAX; | |
461 | static int zc_cb_counter = 0; | |
462 | ||
463 | /* | |
464 | * Minimum number of commit callbacks that need to be registered for us to check | |
465 | * whether the minimum txg delay is acceptable. | |
466 | */ | |
467 | #define ZTEST_COMMIT_CB_MIN_REG 100 | |
468 | ||
469 | /* | |
470 | * If a number of txgs equal to this threshold have been created after a commit | |
471 | * callback has been registered but not called, then we assume there is an | |
472 | * implementation bug. | |
473 | */ | |
474 | #define ZTEST_COMMIT_CB_THRESH (TXG_CONCURRENT_STATES + 1000) | |
428870ff | 475 | |
34dc7c2f | 476 | extern uint64_t metaslab_gang_bang; |
9babb374 | 477 | extern uint64_t metaslab_df_alloc_threshold; |
34dc7c2f | 478 | |
428870ff BB |
479 | enum ztest_object { |
480 | ZTEST_META_DNODE = 0, | |
481 | ZTEST_DIROBJ, | |
482 | ZTEST_OBJECTS | |
483 | }; | |
34dc7c2f BB |
484 | |
485 | static void usage(boolean_t) __NORETURN; | |
486 | ||
487 | /* | |
488 | * These libumem hooks provide a reasonable set of defaults for the allocator's | |
489 | * debugging facilities. | |
490 | */ | |
491 | const char * | |
0bc8fd78 | 492 | _umem_debug_init(void) |
34dc7c2f BB |
493 | { |
494 | return ("default,verbose"); /* $UMEM_DEBUG setting */ | |
495 | } | |
496 | ||
497 | const char * | |
498 | _umem_logging_init(void) | |
499 | { | |
500 | return ("fail,contents"); /* $UMEM_LOGGING setting */ | |
501 | } | |
502 | ||
8fb1ede1 BB |
503 | static void |
504 | dump_debug_buffer(void) | |
505 | { | |
506 | if (!ztest_dump_debug_buffer) | |
507 | return; | |
508 | ||
509 | (void) printf("\n"); | |
510 | zfs_dbgmsg_print("ztest"); | |
511 | } | |
512 | ||
e82cdc3a NB |
513 | #define BACKTRACE_SZ 100 |
514 | ||
515 | static void sig_handler(int signo) | |
516 | { | |
517 | struct sigaction action; | |
967798d0 | 518 | #ifdef __GLIBC__ /* backtrace() is a GNU extension */ |
e82cdc3a NB |
519 | int nptrs; |
520 | void *buffer[BACKTRACE_SZ]; | |
521 | ||
522 | nptrs = backtrace(buffer, BACKTRACE_SZ); | |
523 | backtrace_symbols_fd(buffer, nptrs, STDERR_FILENO); | |
524 | #endif | |
8fb1ede1 | 525 | dump_debug_buffer(); |
e82cdc3a NB |
526 | |
527 | /* | |
528 | * Restore default action and re-raise signal so SIGSEGV and | |
529 | * SIGABRT can trigger a core dump. | |
530 | */ | |
531 | action.sa_handler = SIG_DFL; | |
532 | sigemptyset(&action.sa_mask); | |
533 | action.sa_flags = 0; | |
534 | (void) sigaction(signo, &action, NULL); | |
535 | raise(signo); | |
536 | } | |
537 | ||
34dc7c2f BB |
538 | #define FATAL_MSG_SZ 1024 |
539 | ||
540 | char *fatal_msg; | |
541 | ||
542 | static void | |
543 | fatal(int do_perror, char *message, ...) | |
544 | { | |
545 | va_list args; | |
546 | int save_errno = errno; | |
40b84e7a | 547 | char *buf; |
34dc7c2f BB |
548 | |
549 | (void) fflush(stdout); | |
40b84e7a | 550 | buf = umem_alloc(FATAL_MSG_SZ, UMEM_NOFAIL); |
34dc7c2f BB |
551 | |
552 | va_start(args, message); | |
553 | (void) sprintf(buf, "ztest: "); | |
554 | /* LINTED */ | |
555 | (void) vsprintf(buf + strlen(buf), message, args); | |
556 | va_end(args); | |
557 | if (do_perror) { | |
558 | (void) snprintf(buf + strlen(buf), FATAL_MSG_SZ - strlen(buf), | |
559 | ": %s", strerror(save_errno)); | |
560 | } | |
561 | (void) fprintf(stderr, "%s\n", buf); | |
562 | fatal_msg = buf; /* to ease debugging */ | |
8fb1ede1 BB |
563 | |
564 | dump_debug_buffer(); | |
565 | ||
34dc7c2f BB |
566 | if (ztest_dump_core) |
567 | abort(); | |
568 | exit(3); | |
569 | } | |
570 | ||
571 | static int | |
572 | str2shift(const char *buf) | |
573 | { | |
574 | const char *ends = "BKMGTPEZ"; | |
575 | int i; | |
576 | ||
577 | if (buf[0] == '\0') | |
578 | return (0); | |
579 | for (i = 0; i < strlen(ends); i++) { | |
580 | if (toupper(buf[0]) == ends[i]) | |
581 | break; | |
582 | } | |
583 | if (i == strlen(ends)) { | |
584 | (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", | |
585 | buf); | |
586 | usage(B_FALSE); | |
587 | } | |
588 | if (buf[1] == '\0' || (toupper(buf[1]) == 'B' && buf[2] == '\0')) { | |
589 | return (10*i); | |
590 | } | |
591 | (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", buf); | |
592 | usage(B_FALSE); | |
593 | /* NOTREACHED */ | |
594 | } | |
595 | ||
596 | static uint64_t | |
597 | nicenumtoull(const char *buf) | |
598 | { | |
599 | char *end; | |
600 | uint64_t val; | |
601 | ||
602 | val = strtoull(buf, &end, 0); | |
603 | if (end == buf) { | |
604 | (void) fprintf(stderr, "ztest: bad numeric value: %s\n", buf); | |
605 | usage(B_FALSE); | |
606 | } else if (end[0] == '.') { | |
607 | double fval = strtod(buf, &end); | |
608 | fval *= pow(2, str2shift(end)); | |
609 | if (fval > UINT64_MAX) { | |
610 | (void) fprintf(stderr, "ztest: value too large: %s\n", | |
611 | buf); | |
612 | usage(B_FALSE); | |
613 | } | |
614 | val = (uint64_t)fval; | |
615 | } else { | |
616 | int shift = str2shift(end); | |
617 | if (shift >= 64 || (val << shift) >> shift != val) { | |
618 | (void) fprintf(stderr, "ztest: value too large: %s\n", | |
619 | buf); | |
620 | usage(B_FALSE); | |
621 | } | |
622 | val <<= shift; | |
623 | } | |
624 | return (val); | |
625 | } | |
626 | ||
627 | static void | |
628 | usage(boolean_t requested) | |
629 | { | |
c242c188 CS |
630 | const ztest_shared_opts_t *zo = &ztest_opts_defaults; |
631 | ||
f3c8c9e6 JK |
632 | char nice_vdev_size[NN_NUMBUF_SZ]; |
633 | char nice_gang_bang[NN_NUMBUF_SZ]; | |
34dc7c2f BB |
634 | FILE *fp = requested ? stdout : stderr; |
635 | ||
f3c8c9e6 JK |
636 | nicenum(zo->zo_vdev_size, nice_vdev_size, sizeof (nice_vdev_size)); |
637 | nicenum(zo->zo_metaslab_gang_bang, nice_gang_bang, | |
638 | sizeof (nice_gang_bang)); | |
34dc7c2f BB |
639 | |
640 | (void) fprintf(fp, "Usage: %s\n" | |
641 | "\t[-v vdevs (default: %llu)]\n" | |
642 | "\t[-s size_of_each_vdev (default: %s)]\n" | |
428870ff | 643 | "\t[-a alignment_shift (default: %d)] use 0 for random\n" |
34dc7c2f BB |
644 | "\t[-m mirror_copies (default: %d)]\n" |
645 | "\t[-r raidz_disks (default: %d)]\n" | |
646 | "\t[-R raidz_parity (default: %d)]\n" | |
647 | "\t[-d datasets (default: %d)]\n" | |
648 | "\t[-t threads (default: %d)]\n" | |
649 | "\t[-g gang_block_threshold (default: %s)]\n" | |
428870ff BB |
650 | "\t[-i init_count (default: %d)] initialize pool i times\n" |
651 | "\t[-k kill_percentage (default: %llu%%)]\n" | |
34dc7c2f | 652 | "\t[-p pool_name (default: %s)]\n" |
428870ff | 653 | "\t[-f dir (default: %s)] file directory for vdev files\n" |
379ca9cf | 654 | "\t[-M] Multi-host simulate pool imported on remote host\n" |
428870ff BB |
655 | "\t[-V] verbose (use multiple times for ever more blather)\n" |
656 | "\t[-E] use existing pool instead of creating new one\n" | |
657 | "\t[-T time (default: %llu sec)] total run time\n" | |
658 | "\t[-F freezeloops (default: %llu)] max loops in spa_freeze()\n" | |
659 | "\t[-P passtime (default: %llu sec)] time per pass\n" | |
c242c188 | 660 | "\t[-B alt_ztest (default: <none>)] alternate ztest path\n" |
ed828c0c GM |
661 | "\t[-o variable=value] ... set global variable to an unsigned\n" |
662 | "\t 32-bit integer value\n" | |
8fb1ede1 | 663 | "\t[-G dump zfs_dbgmsg buffer before exiting due to an error\n" |
34dc7c2f BB |
664 | "\t[-h] (print help)\n" |
665 | "", | |
c242c188 CS |
666 | zo->zo_pool, |
667 | (u_longlong_t)zo->zo_vdevs, /* -v */ | |
34dc7c2f | 668 | nice_vdev_size, /* -s */ |
c242c188 CS |
669 | zo->zo_ashift, /* -a */ |
670 | zo->zo_mirrors, /* -m */ | |
671 | zo->zo_raidz, /* -r */ | |
672 | zo->zo_raidz_parity, /* -R */ | |
673 | zo->zo_datasets, /* -d */ | |
674 | zo->zo_threads, /* -t */ | |
34dc7c2f | 675 | nice_gang_bang, /* -g */ |
c242c188 CS |
676 | zo->zo_init, /* -i */ |
677 | (u_longlong_t)zo->zo_killrate, /* -k */ | |
678 | zo->zo_pool, /* -p */ | |
679 | zo->zo_dir, /* -f */ | |
680 | (u_longlong_t)zo->zo_time, /* -T */ | |
681 | (u_longlong_t)zo->zo_maxloops, /* -F */ | |
682 | (u_longlong_t)zo->zo_passtime); | |
34dc7c2f BB |
683 | exit(requested ? 0 : 1); |
684 | } | |
685 | ||
34dc7c2f BB |
686 | static void |
687 | process_options(int argc, char **argv) | |
688 | { | |
c242c188 CS |
689 | char *path; |
690 | ztest_shared_opts_t *zo = &ztest_opts; | |
691 | ||
34dc7c2f BB |
692 | int opt; |
693 | uint64_t value; | |
c242c188 | 694 | char altdir[MAXNAMELEN] = { 0 }; |
34dc7c2f | 695 | |
c242c188 | 696 | bcopy(&ztest_opts_defaults, zo, sizeof (*zo)); |
34dc7c2f | 697 | |
34dc7c2f | 698 | while ((opt = getopt(argc, argv, |
8fb1ede1 | 699 | "v:s:a:m:r:R:d:t:g:i:k:p:f:MVET:P:hF:B:o:G")) != EOF) { |
34dc7c2f BB |
700 | value = 0; |
701 | switch (opt) { | |
702 | case 'v': | |
703 | case 's': | |
704 | case 'a': | |
705 | case 'm': | |
706 | case 'r': | |
707 | case 'R': | |
708 | case 'd': | |
709 | case 't': | |
710 | case 'g': | |
711 | case 'i': | |
712 | case 'k': | |
713 | case 'T': | |
714 | case 'P': | |
428870ff | 715 | case 'F': |
34dc7c2f BB |
716 | value = nicenumtoull(optarg); |
717 | } | |
718 | switch (opt) { | |
719 | case 'v': | |
c242c188 | 720 | zo->zo_vdevs = value; |
34dc7c2f BB |
721 | break; |
722 | case 's': | |
c242c188 | 723 | zo->zo_vdev_size = MAX(SPA_MINDEVSIZE, value); |
34dc7c2f BB |
724 | break; |
725 | case 'a': | |
c242c188 | 726 | zo->zo_ashift = value; |
34dc7c2f BB |
727 | break; |
728 | case 'm': | |
c242c188 | 729 | zo->zo_mirrors = value; |
34dc7c2f BB |
730 | break; |
731 | case 'r': | |
c242c188 | 732 | zo->zo_raidz = MAX(1, value); |
34dc7c2f BB |
733 | break; |
734 | case 'R': | |
c242c188 | 735 | zo->zo_raidz_parity = MIN(MAX(value, 1), 3); |
34dc7c2f BB |
736 | break; |
737 | case 'd': | |
c242c188 | 738 | zo->zo_datasets = MAX(1, value); |
34dc7c2f BB |
739 | break; |
740 | case 't': | |
c242c188 | 741 | zo->zo_threads = MAX(1, value); |
34dc7c2f BB |
742 | break; |
743 | case 'g': | |
c242c188 CS |
744 | zo->zo_metaslab_gang_bang = MAX(SPA_MINBLOCKSIZE << 1, |
745 | value); | |
34dc7c2f BB |
746 | break; |
747 | case 'i': | |
c242c188 | 748 | zo->zo_init = value; |
34dc7c2f BB |
749 | break; |
750 | case 'k': | |
c242c188 | 751 | zo->zo_killrate = value; |
34dc7c2f BB |
752 | break; |
753 | case 'p': | |
c242c188 CS |
754 | (void) strlcpy(zo->zo_pool, optarg, |
755 | sizeof (zo->zo_pool)); | |
34dc7c2f BB |
756 | break; |
757 | case 'f': | |
c242c188 CS |
758 | path = realpath(optarg, NULL); |
759 | if (path == NULL) { | |
760 | (void) fprintf(stderr, "error: %s: %s\n", | |
761 | optarg, strerror(errno)); | |
762 | usage(B_FALSE); | |
763 | } else { | |
764 | (void) strlcpy(zo->zo_dir, path, | |
765 | sizeof (zo->zo_dir)); | |
df053d67 | 766 | free(path); |
c242c188 | 767 | } |
34dc7c2f | 768 | break; |
379ca9cf OF |
769 | case 'M': |
770 | zo->zo_mmp_test = 1; | |
771 | break; | |
34dc7c2f | 772 | case 'V': |
c242c188 | 773 | zo->zo_verbose++; |
34dc7c2f BB |
774 | break; |
775 | case 'E': | |
c242c188 | 776 | zo->zo_init = 0; |
34dc7c2f BB |
777 | break; |
778 | case 'T': | |
c242c188 | 779 | zo->zo_time = value; |
34dc7c2f BB |
780 | break; |
781 | case 'P': | |
c242c188 | 782 | zo->zo_passtime = MAX(1, value); |
34dc7c2f | 783 | break; |
428870ff | 784 | case 'F': |
c242c188 CS |
785 | zo->zo_maxloops = MAX(1, value); |
786 | break; | |
787 | case 'B': | |
788 | (void) strlcpy(altdir, optarg, sizeof (altdir)); | |
428870ff | 789 | break; |
ed828c0c GM |
790 | case 'o': |
791 | if (set_global_var(optarg) != 0) | |
792 | usage(B_FALSE); | |
793 | break; | |
8fb1ede1 BB |
794 | case 'G': |
795 | ztest_dump_debug_buffer = B_TRUE; | |
796 | break; | |
34dc7c2f BB |
797 | case 'h': |
798 | usage(B_TRUE); | |
799 | break; | |
800 | case '?': | |
801 | default: | |
802 | usage(B_FALSE); | |
803 | break; | |
804 | } | |
805 | } | |
806 | ||
c242c188 | 807 | zo->zo_raidz_parity = MIN(zo->zo_raidz_parity, zo->zo_raidz - 1); |
34dc7c2f | 808 | |
c242c188 CS |
809 | zo->zo_vdevtime = |
810 | (zo->zo_vdevs > 0 ? zo->zo_time * NANOSEC / zo->zo_vdevs : | |
428870ff | 811 | UINT64_MAX >> 2); |
c242c188 CS |
812 | |
813 | if (strlen(altdir) > 0) { | |
ae380cfa BB |
814 | char *cmd; |
815 | char *realaltdir; | |
c242c188 CS |
816 | char *bin; |
817 | char *ztest; | |
818 | char *isa; | |
819 | int isalen; | |
820 | ||
ae380cfa BB |
821 | cmd = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
822 | realaltdir = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
823 | ||
5be98cfe | 824 | VERIFY(NULL != realpath(getexecname(), cmd)); |
c242c188 CS |
825 | if (0 != access(altdir, F_OK)) { |
826 | ztest_dump_core = B_FALSE; | |
827 | fatal(B_TRUE, "invalid alternate ztest path: %s", | |
828 | altdir); | |
829 | } | |
830 | VERIFY(NULL != realpath(altdir, realaltdir)); | |
831 | ||
832 | /* | |
833 | * 'cmd' should be of the form "<anything>/usr/bin/<isa>/ztest". | |
834 | * We want to extract <isa> to determine if we should use | |
835 | * 32 or 64 bit binaries. | |
836 | */ | |
837 | bin = strstr(cmd, "/usr/bin/"); | |
838 | ztest = strstr(bin, "/ztest"); | |
839 | isa = bin + 9; | |
840 | isalen = ztest - isa; | |
841 | (void) snprintf(zo->zo_alt_ztest, sizeof (zo->zo_alt_ztest), | |
842 | "%s/usr/bin/%.*s/ztest", realaltdir, isalen, isa); | |
843 | (void) snprintf(zo->zo_alt_libpath, sizeof (zo->zo_alt_libpath), | |
844 | "%s/usr/lib/%.*s", realaltdir, isalen, isa); | |
845 | ||
846 | if (0 != access(zo->zo_alt_ztest, X_OK)) { | |
847 | ztest_dump_core = B_FALSE; | |
848 | fatal(B_TRUE, "invalid alternate ztest: %s", | |
849 | zo->zo_alt_ztest); | |
850 | } else if (0 != access(zo->zo_alt_libpath, X_OK)) { | |
851 | ztest_dump_core = B_FALSE; | |
852 | fatal(B_TRUE, "invalid alternate lib directory %s", | |
853 | zo->zo_alt_libpath); | |
854 | } | |
ae380cfa BB |
855 | |
856 | umem_free(cmd, MAXPATHLEN); | |
857 | umem_free(realaltdir, MAXPATHLEN); | |
c242c188 | 858 | } |
428870ff BB |
859 | } |
860 | ||
861 | static void | |
862 | ztest_kill(ztest_shared_t *zs) | |
863 | { | |
c242c188 CS |
864 | zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(ztest_spa)); |
865 | zs->zs_space = metaslab_class_get_space(spa_normal_class(ztest_spa)); | |
5d1f7fb6 GW |
866 | |
867 | /* | |
868 | * Before we kill off ztest, make sure that the config is updated. | |
869 | * See comment above spa_config_sync(). | |
870 | */ | |
871 | mutex_enter(&spa_namespace_lock); | |
872 | spa_config_sync(ztest_spa, B_FALSE, B_FALSE); | |
873 | mutex_exit(&spa_namespace_lock); | |
874 | ||
428870ff BB |
875 | (void) kill(getpid(), SIGKILL); |
876 | } | |
877 | ||
878 | static uint64_t | |
879 | ztest_random(uint64_t range) | |
880 | { | |
881 | uint64_t r; | |
882 | ||
9d81146b ED |
883 | ASSERT3S(ztest_fd_rand, >=, 0); |
884 | ||
428870ff BB |
885 | if (range == 0) |
886 | return (0); | |
887 | ||
9d81146b | 888 | if (read(ztest_fd_rand, &r, sizeof (r)) != sizeof (r)) |
428870ff BB |
889 | fatal(1, "short read from /dev/urandom"); |
890 | ||
891 | return (r % range); | |
892 | } | |
893 | ||
894 | /* ARGSUSED */ | |
895 | static void | |
896 | ztest_record_enospc(const char *s) | |
897 | { | |
898 | ztest_shared->zs_enospc_count++; | |
34dc7c2f BB |
899 | } |
900 | ||
901 | static uint64_t | |
902 | ztest_get_ashift(void) | |
903 | { | |
c242c188 | 904 | if (ztest_opts.zo_ashift == 0) |
b02fe35d | 905 | return (SPA_MINBLOCKSHIFT + ztest_random(5)); |
c242c188 | 906 | return (ztest_opts.zo_ashift); |
34dc7c2f BB |
907 | } |
908 | ||
909 | static nvlist_t * | |
ea0b2538 | 910 | make_vdev_file(char *path, char *aux, char *pool, size_t size, uint64_t ashift) |
34dc7c2f | 911 | { |
40b84e7a | 912 | char *pathbuf; |
34dc7c2f | 913 | uint64_t vdev; |
34dc7c2f BB |
914 | nvlist_t *file; |
915 | ||
40b84e7a BB |
916 | pathbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
917 | ||
b128c09f BB |
918 | if (ashift == 0) |
919 | ashift = ztest_get_ashift(); | |
920 | ||
921 | if (path == NULL) { | |
922 | path = pathbuf; | |
923 | ||
924 | if (aux != NULL) { | |
925 | vdev = ztest_shared->zs_vdev_aux; | |
c242c188 CS |
926 | (void) snprintf(path, MAXPATHLEN, |
927 | ztest_aux_template, ztest_opts.zo_dir, | |
ea0b2538 GW |
928 | pool == NULL ? ztest_opts.zo_pool : pool, |
929 | aux, vdev); | |
b128c09f | 930 | } else { |
428870ff | 931 | vdev = ztest_shared->zs_vdev_next_leaf++; |
c242c188 CS |
932 | (void) snprintf(path, MAXPATHLEN, |
933 | ztest_dev_template, ztest_opts.zo_dir, | |
ea0b2538 | 934 | pool == NULL ? ztest_opts.zo_pool : pool, vdev); |
b128c09f BB |
935 | } |
936 | } | |
34dc7c2f | 937 | |
b128c09f BB |
938 | if (size != 0) { |
939 | int fd = open(path, O_RDWR | O_CREAT | O_TRUNC, 0666); | |
34dc7c2f | 940 | if (fd == -1) |
b128c09f | 941 | fatal(1, "can't open %s", path); |
34dc7c2f | 942 | if (ftruncate(fd, size) != 0) |
b128c09f | 943 | fatal(1, "can't ftruncate %s", path); |
34dc7c2f BB |
944 | (void) close(fd); |
945 | } | |
946 | ||
947 | VERIFY(nvlist_alloc(&file, NV_UNIQUE_NAME, 0) == 0); | |
948 | VERIFY(nvlist_add_string(file, ZPOOL_CONFIG_TYPE, VDEV_TYPE_FILE) == 0); | |
b128c09f | 949 | VERIFY(nvlist_add_string(file, ZPOOL_CONFIG_PATH, path) == 0); |
34dc7c2f | 950 | VERIFY(nvlist_add_uint64(file, ZPOOL_CONFIG_ASHIFT, ashift) == 0); |
40b84e7a | 951 | umem_free(pathbuf, MAXPATHLEN); |
34dc7c2f BB |
952 | |
953 | return (file); | |
954 | } | |
955 | ||
956 | static nvlist_t * | |
ea0b2538 GW |
957 | make_vdev_raidz(char *path, char *aux, char *pool, size_t size, |
958 | uint64_t ashift, int r) | |
34dc7c2f BB |
959 | { |
960 | nvlist_t *raidz, **child; | |
961 | int c; | |
962 | ||
963 | if (r < 2) | |
ea0b2538 | 964 | return (make_vdev_file(path, aux, pool, size, ashift)); |
34dc7c2f BB |
965 | child = umem_alloc(r * sizeof (nvlist_t *), UMEM_NOFAIL); |
966 | ||
967 | for (c = 0; c < r; c++) | |
ea0b2538 | 968 | child[c] = make_vdev_file(path, aux, pool, size, ashift); |
34dc7c2f BB |
969 | |
970 | VERIFY(nvlist_alloc(&raidz, NV_UNIQUE_NAME, 0) == 0); | |
971 | VERIFY(nvlist_add_string(raidz, ZPOOL_CONFIG_TYPE, | |
972 | VDEV_TYPE_RAIDZ) == 0); | |
973 | VERIFY(nvlist_add_uint64(raidz, ZPOOL_CONFIG_NPARITY, | |
c242c188 | 974 | ztest_opts.zo_raidz_parity) == 0); |
34dc7c2f BB |
975 | VERIFY(nvlist_add_nvlist_array(raidz, ZPOOL_CONFIG_CHILDREN, |
976 | child, r) == 0); | |
977 | ||
978 | for (c = 0; c < r; c++) | |
979 | nvlist_free(child[c]); | |
980 | ||
981 | umem_free(child, r * sizeof (nvlist_t *)); | |
982 | ||
983 | return (raidz); | |
984 | } | |
985 | ||
986 | static nvlist_t * | |
ea0b2538 GW |
987 | make_vdev_mirror(char *path, char *aux, char *pool, size_t size, |
988 | uint64_t ashift, int r, int m) | |
34dc7c2f BB |
989 | { |
990 | nvlist_t *mirror, **child; | |
991 | int c; | |
992 | ||
993 | if (m < 1) | |
ea0b2538 | 994 | return (make_vdev_raidz(path, aux, pool, size, ashift, r)); |
34dc7c2f BB |
995 | |
996 | child = umem_alloc(m * sizeof (nvlist_t *), UMEM_NOFAIL); | |
997 | ||
998 | for (c = 0; c < m; c++) | |
ea0b2538 | 999 | child[c] = make_vdev_raidz(path, aux, pool, size, ashift, r); |
34dc7c2f BB |
1000 | |
1001 | VERIFY(nvlist_alloc(&mirror, NV_UNIQUE_NAME, 0) == 0); | |
1002 | VERIFY(nvlist_add_string(mirror, ZPOOL_CONFIG_TYPE, | |
1003 | VDEV_TYPE_MIRROR) == 0); | |
1004 | VERIFY(nvlist_add_nvlist_array(mirror, ZPOOL_CONFIG_CHILDREN, | |
1005 | child, m) == 0); | |
34dc7c2f BB |
1006 | |
1007 | for (c = 0; c < m; c++) | |
1008 | nvlist_free(child[c]); | |
1009 | ||
1010 | umem_free(child, m * sizeof (nvlist_t *)); | |
1011 | ||
1012 | return (mirror); | |
1013 | } | |
1014 | ||
1015 | static nvlist_t * | |
ea0b2538 GW |
1016 | make_vdev_root(char *path, char *aux, char *pool, size_t size, uint64_t ashift, |
1017 | int log, int r, int m, int t) | |
34dc7c2f BB |
1018 | { |
1019 | nvlist_t *root, **child; | |
1020 | int c; | |
1021 | ||
1022 | ASSERT(t > 0); | |
1023 | ||
1024 | child = umem_alloc(t * sizeof (nvlist_t *), UMEM_NOFAIL); | |
1025 | ||
b128c09f | 1026 | for (c = 0; c < t; c++) { |
ea0b2538 GW |
1027 | child[c] = make_vdev_mirror(path, aux, pool, size, ashift, |
1028 | r, m); | |
b128c09f BB |
1029 | VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_IS_LOG, |
1030 | log) == 0); | |
1031 | } | |
34dc7c2f BB |
1032 | |
1033 | VERIFY(nvlist_alloc(&root, NV_UNIQUE_NAME, 0) == 0); | |
1034 | VERIFY(nvlist_add_string(root, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT) == 0); | |
b128c09f | 1035 | VERIFY(nvlist_add_nvlist_array(root, aux ? aux : ZPOOL_CONFIG_CHILDREN, |
34dc7c2f BB |
1036 | child, t) == 0); |
1037 | ||
1038 | for (c = 0; c < t; c++) | |
1039 | nvlist_free(child[c]); | |
1040 | ||
1041 | umem_free(child, t * sizeof (nvlist_t *)); | |
1042 | ||
1043 | return (root); | |
1044 | } | |
1045 | ||
ea0b2538 GW |
1046 | /* |
1047 | * Find a random spa version. Returns back a random spa version in the | |
1048 | * range [initial_version, SPA_VERSION_FEATURES]. | |
1049 | */ | |
1050 | static uint64_t | |
1051 | ztest_random_spa_version(uint64_t initial_version) | |
1052 | { | |
1053 | uint64_t version = initial_version; | |
1054 | ||
1055 | if (version <= SPA_VERSION_BEFORE_FEATURES) { | |
1056 | version = version + | |
1057 | ztest_random(SPA_VERSION_BEFORE_FEATURES - version + 1); | |
1058 | } | |
1059 | ||
1060 | if (version > SPA_VERSION_BEFORE_FEATURES) | |
1061 | version = SPA_VERSION_FEATURES; | |
1062 | ||
1063 | ASSERT(SPA_VERSION_IS_SUPPORTED(version)); | |
1064 | return (version); | |
1065 | } | |
1066 | ||
428870ff BB |
1067 | static int |
1068 | ztest_random_blocksize(void) | |
34dc7c2f | 1069 | { |
f1512ee6 MA |
1070 | /* |
1071 | * Choose a block size >= the ashift. | |
1072 | * If the SPA supports new MAXBLOCKSIZE, test up to 1MB blocks. | |
1073 | */ | |
1074 | int maxbs = SPA_OLD_MAXBLOCKSHIFT; | |
1075 | if (spa_maxblocksize(ztest_spa) == SPA_MAXBLOCKSIZE) | |
1076 | maxbs = 20; | |
c3520e7f MA |
1077 | uint64_t block_shift = |
1078 | ztest_random(maxbs - ztest_spa->spa_max_ashift + 1); | |
b02fe35d | 1079 | return (1 << (SPA_MINBLOCKSHIFT + block_shift)); |
428870ff | 1080 | } |
34dc7c2f | 1081 | |
50c957f7 NB |
1082 | static int |
1083 | ztest_random_dnodesize(void) | |
1084 | { | |
1085 | int slots; | |
1086 | int max_slots = spa_maxdnodesize(ztest_spa) >> DNODE_SHIFT; | |
1087 | ||
1088 | if (max_slots == DNODE_MIN_SLOTS) | |
1089 | return (DNODE_MIN_SIZE); | |
1090 | ||
1091 | /* | |
1092 | * Weight the random distribution more heavily toward smaller | |
1093 | * dnode sizes since that is more likely to reflect real-world | |
1094 | * usage. | |
1095 | */ | |
1096 | ASSERT3U(max_slots, >, 4); | |
1097 | switch (ztest_random(10)) { | |
1098 | case 0: | |
1099 | slots = 5 + ztest_random(max_slots - 4); | |
1100 | break; | |
1101 | case 1 ... 4: | |
1102 | slots = 2 + ztest_random(3); | |
1103 | break; | |
1104 | default: | |
1105 | slots = 1; | |
1106 | break; | |
1107 | } | |
1108 | ||
1109 | return (slots << DNODE_SHIFT); | |
1110 | } | |
1111 | ||
428870ff BB |
1112 | static int |
1113 | ztest_random_ibshift(void) | |
1114 | { | |
1115 | return (DN_MIN_INDBLKSHIFT + | |
1116 | ztest_random(DN_MAX_INDBLKSHIFT - DN_MIN_INDBLKSHIFT + 1)); | |
34dc7c2f BB |
1117 | } |
1118 | ||
428870ff BB |
1119 | static uint64_t |
1120 | ztest_random_vdev_top(spa_t *spa, boolean_t log_ok) | |
34dc7c2f | 1121 | { |
428870ff BB |
1122 | uint64_t top; |
1123 | vdev_t *rvd = spa->spa_root_vdev; | |
1124 | vdev_t *tvd; | |
34dc7c2f | 1125 | |
428870ff | 1126 | ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); |
34dc7c2f | 1127 | |
428870ff BB |
1128 | do { |
1129 | top = ztest_random(rvd->vdev_children); | |
1130 | tvd = rvd->vdev_child[top]; | |
1131 | } while (tvd->vdev_ishole || (tvd->vdev_islog && !log_ok) || | |
1132 | tvd->vdev_mg == NULL || tvd->vdev_mg->mg_class == NULL); | |
34dc7c2f | 1133 | |
428870ff | 1134 | return (top); |
34dc7c2f BB |
1135 | } |
1136 | ||
428870ff BB |
1137 | static uint64_t |
1138 | ztest_random_dsl_prop(zfs_prop_t prop) | |
34dc7c2f | 1139 | { |
428870ff BB |
1140 | uint64_t value; |
1141 | ||
1142 | do { | |
1143 | value = zfs_prop_random_value(prop, ztest_random(-1ULL)); | |
1144 | } while (prop == ZFS_PROP_CHECKSUM && value == ZIO_CHECKSUM_OFF); | |
1145 | ||
1146 | return (value); | |
34dc7c2f BB |
1147 | } |
1148 | ||
34dc7c2f | 1149 | static int |
428870ff BB |
1150 | ztest_dsl_prop_set_uint64(char *osname, zfs_prop_t prop, uint64_t value, |
1151 | boolean_t inherit) | |
34dc7c2f | 1152 | { |
428870ff BB |
1153 | const char *propname = zfs_prop_to_name(prop); |
1154 | const char *valname; | |
40b84e7a | 1155 | char *setpoint; |
428870ff | 1156 | uint64_t curval; |
34dc7c2f BB |
1157 | int error; |
1158 | ||
13fe0198 MA |
1159 | error = dsl_prop_set_int(osname, propname, |
1160 | (inherit ? ZPROP_SRC_NONE : ZPROP_SRC_LOCAL), value); | |
34dc7c2f | 1161 | |
428870ff BB |
1162 | if (error == ENOSPC) { |
1163 | ztest_record_enospc(FTAG); | |
34dc7c2f BB |
1164 | return (error); |
1165 | } | |
c99c9001 | 1166 | ASSERT0(error); |
34dc7c2f | 1167 | |
40b84e7a | 1168 | setpoint = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
13fe0198 | 1169 | VERIFY0(dsl_prop_get_integer(osname, propname, &curval, setpoint)); |
428870ff | 1170 | |
c242c188 | 1171 | if (ztest_opts.zo_verbose >= 6) { |
6bec4351 TC |
1172 | int err; |
1173 | ||
1174 | err = zfs_prop_index_to_string(prop, curval, &valname); | |
1175 | if (err) | |
02730c33 BB |
1176 | (void) printf("%s %s = %llu at '%s'\n", osname, |
1177 | propname, (unsigned long long)curval, setpoint); | |
6bec4351 TC |
1178 | else |
1179 | (void) printf("%s %s = %s at '%s'\n", | |
1180 | osname, propname, valname, setpoint); | |
34dc7c2f | 1181 | } |
40b84e7a | 1182 | umem_free(setpoint, MAXPATHLEN); |
34dc7c2f BB |
1183 | |
1184 | return (error); | |
1185 | } | |
1186 | ||
1187 | static int | |
c242c188 | 1188 | ztest_spa_prop_set_uint64(zpool_prop_t prop, uint64_t value) |
34dc7c2f | 1189 | { |
c242c188 | 1190 | spa_t *spa = ztest_spa; |
428870ff | 1191 | nvlist_t *props = NULL; |
34dc7c2f BB |
1192 | int error; |
1193 | ||
428870ff BB |
1194 | VERIFY(nvlist_alloc(&props, NV_UNIQUE_NAME, 0) == 0); |
1195 | VERIFY(nvlist_add_uint64(props, zpool_prop_to_name(prop), value) == 0); | |
34dc7c2f | 1196 | |
428870ff BB |
1197 | error = spa_prop_set(spa, props); |
1198 | ||
1199 | nvlist_free(props); | |
1200 | ||
1201 | if (error == ENOSPC) { | |
1202 | ztest_record_enospc(FTAG); | |
34dc7c2f BB |
1203 | return (error); |
1204 | } | |
c99c9001 | 1205 | ASSERT0(error); |
34dc7c2f BB |
1206 | |
1207 | return (error); | |
1208 | } | |
1209 | ||
4807c0ba TC |
1210 | static int |
1211 | ztest_dmu_objset_own(const char *name, dmu_objset_type_t type, | |
1212 | boolean_t readonly, boolean_t decrypt, void *tag, objset_t **osp) | |
1213 | { | |
1214 | int err; | |
1215 | ||
1216 | err = dmu_objset_own(name, type, readonly, decrypt, tag, osp); | |
1217 | if (decrypt && err == EACCES) { | |
1218 | char ddname[ZFS_MAX_DATASET_NAME_LEN]; | |
1219 | dsl_crypto_params_t *dcp; | |
1220 | nvlist_t *crypto_args = fnvlist_alloc(); | |
1221 | char *cp = NULL; | |
1222 | ||
1223 | /* spa_keystore_load_wkey() expects a dsl dir name */ | |
1224 | strcpy(ddname, name); | |
1225 | cp = strchr(ddname, '@'); | |
1226 | if (cp != NULL) | |
1227 | *cp = '\0'; | |
1228 | ||
1229 | fnvlist_add_uint8_array(crypto_args, "wkeydata", | |
1230 | (uint8_t *)ztest_wkeydata, WRAPPING_KEY_LEN); | |
1231 | VERIFY0(dsl_crypto_params_create_nvlist(DCP_CMD_NONE, NULL, | |
1232 | crypto_args, &dcp)); | |
1233 | err = spa_keystore_load_wkey(ddname, dcp, B_FALSE); | |
1234 | dsl_crypto_params_free(dcp, B_FALSE); | |
1235 | fnvlist_free(crypto_args); | |
1236 | ||
1237 | if (err != 0) | |
1238 | return (err); | |
1239 | ||
1240 | err = dmu_objset_own(name, type, readonly, decrypt, tag, osp); | |
1241 | } | |
1242 | ||
1243 | return (err); | |
1244 | } | |
1245 | ||
e3a07cd0 BP |
1246 | |
1247 | /* | |
1248 | * Object and range lock mechanics | |
1249 | */ | |
1250 | typedef struct { | |
1251 | list_node_t z_lnode; | |
1252 | refcount_t z_refcnt; | |
1253 | uint64_t z_object; | |
1254 | zfs_rlock_t z_range_lock; | |
1255 | } ztest_znode_t; | |
1256 | ||
1257 | typedef struct { | |
1258 | rl_t *z_rl; | |
1259 | ztest_znode_t *z_ztznode; | |
1260 | } ztest_zrl_t; | |
1261 | ||
1262 | static ztest_znode_t * | |
1263 | ztest_znode_init(uint64_t object) | |
1264 | { | |
1265 | ztest_znode_t *zp = umem_alloc(sizeof (*zp), UMEM_NOFAIL); | |
1266 | ||
1267 | list_link_init(&zp->z_lnode); | |
1268 | refcount_create(&zp->z_refcnt); | |
1269 | zp->z_object = object; | |
1270 | zfs_rlock_init(&zp->z_range_lock); | |
1271 | ||
1272 | return (zp); | |
1273 | } | |
1274 | ||
1275 | static void | |
1276 | ztest_znode_fini(ztest_znode_t *zp) | |
1277 | { | |
1278 | ASSERT(refcount_is_zero(&zp->z_refcnt)); | |
1279 | zfs_rlock_destroy(&zp->z_range_lock); | |
1280 | zp->z_object = 0; | |
1281 | refcount_destroy(&zp->z_refcnt); | |
1282 | list_link_init(&zp->z_lnode); | |
1283 | umem_free(zp, sizeof (*zp)); | |
1284 | } | |
1285 | ||
1286 | static void | |
1287 | ztest_zll_init(zll_t *zll) | |
1288 | { | |
1289 | mutex_init(&zll->z_lock, NULL, MUTEX_DEFAULT, NULL); | |
1290 | list_create(&zll->z_list, sizeof (ztest_znode_t), | |
1291 | offsetof(ztest_znode_t, z_lnode)); | |
1292 | } | |
1293 | ||
1294 | static void | |
1295 | ztest_zll_destroy(zll_t *zll) | |
1296 | { | |
1297 | list_destroy(&zll->z_list); | |
1298 | mutex_destroy(&zll->z_lock); | |
1299 | } | |
1300 | ||
1301 | #define RL_TAG "range_lock" | |
1302 | static ztest_znode_t * | |
1303 | ztest_znode_get(ztest_ds_t *zd, uint64_t object) | |
1304 | { | |
1305 | zll_t *zll = &zd->zd_range_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; | |
1306 | ztest_znode_t *zp = NULL; | |
1307 | mutex_enter(&zll->z_lock); | |
1308 | for (zp = list_head(&zll->z_list); (zp); | |
1309 | zp = list_next(&zll->z_list, zp)) { | |
1310 | if (zp->z_object == object) { | |
1311 | refcount_add(&zp->z_refcnt, RL_TAG); | |
1312 | break; | |
1313 | } | |
1314 | } | |
1315 | if (zp == NULL) { | |
1316 | zp = ztest_znode_init(object); | |
1317 | refcount_add(&zp->z_refcnt, RL_TAG); | |
1318 | list_insert_head(&zll->z_list, zp); | |
1319 | } | |
1320 | mutex_exit(&zll->z_lock); | |
1321 | return (zp); | |
1322 | } | |
1323 | ||
1324 | static void | |
1325 | ztest_znode_put(ztest_ds_t *zd, ztest_znode_t *zp) | |
1326 | { | |
1327 | zll_t *zll = NULL; | |
1328 | ASSERT3U(zp->z_object, !=, 0); | |
1329 | zll = &zd->zd_range_lock[zp->z_object & (ZTEST_OBJECT_LOCKS - 1)]; | |
1330 | mutex_enter(&zll->z_lock); | |
1331 | refcount_remove(&zp->z_refcnt, RL_TAG); | |
1332 | if (refcount_is_zero(&zp->z_refcnt)) { | |
1333 | list_remove(&zll->z_list, zp); | |
1334 | ztest_znode_fini(zp); | |
1335 | } | |
1336 | mutex_exit(&zll->z_lock); | |
1337 | } | |
1338 | ||
1339 | ||
428870ff BB |
1340 | static void |
1341 | ztest_rll_init(rll_t *rll) | |
1342 | { | |
1343 | rll->rll_writer = NULL; | |
1344 | rll->rll_readers = 0; | |
1e33ac1e BB |
1345 | mutex_init(&rll->rll_lock, NULL, MUTEX_DEFAULT, NULL); |
1346 | cv_init(&rll->rll_cv, NULL, CV_DEFAULT, NULL); | |
428870ff | 1347 | } |
34dc7c2f | 1348 | |
428870ff BB |
1349 | static void |
1350 | ztest_rll_destroy(rll_t *rll) | |
34dc7c2f | 1351 | { |
428870ff BB |
1352 | ASSERT(rll->rll_writer == NULL); |
1353 | ASSERT(rll->rll_readers == 0); | |
1e33ac1e BB |
1354 | mutex_destroy(&rll->rll_lock); |
1355 | cv_destroy(&rll->rll_cv); | |
428870ff | 1356 | } |
34dc7c2f | 1357 | |
428870ff BB |
1358 | static void |
1359 | ztest_rll_lock(rll_t *rll, rl_type_t type) | |
1360 | { | |
1e33ac1e | 1361 | mutex_enter(&rll->rll_lock); |
34dc7c2f | 1362 | |
428870ff BB |
1363 | if (type == RL_READER) { |
1364 | while (rll->rll_writer != NULL) | |
1e33ac1e | 1365 | (void) cv_wait(&rll->rll_cv, &rll->rll_lock); |
428870ff BB |
1366 | rll->rll_readers++; |
1367 | } else { | |
1368 | while (rll->rll_writer != NULL || rll->rll_readers) | |
1e33ac1e | 1369 | (void) cv_wait(&rll->rll_cv, &rll->rll_lock); |
428870ff BB |
1370 | rll->rll_writer = curthread; |
1371 | } | |
34dc7c2f | 1372 | |
1e33ac1e | 1373 | mutex_exit(&rll->rll_lock); |
428870ff | 1374 | } |
34dc7c2f | 1375 | |
428870ff BB |
1376 | static void |
1377 | ztest_rll_unlock(rll_t *rll) | |
1378 | { | |
1e33ac1e | 1379 | mutex_enter(&rll->rll_lock); |
34dc7c2f | 1380 | |
428870ff BB |
1381 | if (rll->rll_writer) { |
1382 | ASSERT(rll->rll_readers == 0); | |
1383 | rll->rll_writer = NULL; | |
1384 | } else { | |
1385 | ASSERT(rll->rll_readers != 0); | |
1386 | ASSERT(rll->rll_writer == NULL); | |
1387 | rll->rll_readers--; | |
1388 | } | |
34dc7c2f | 1389 | |
428870ff | 1390 | if (rll->rll_writer == NULL && rll->rll_readers == 0) |
1e33ac1e | 1391 | cv_broadcast(&rll->rll_cv); |
428870ff | 1392 | |
1e33ac1e | 1393 | mutex_exit(&rll->rll_lock); |
34dc7c2f BB |
1394 | } |
1395 | ||
428870ff BB |
1396 | static void |
1397 | ztest_object_lock(ztest_ds_t *zd, uint64_t object, rl_type_t type) | |
b128c09f | 1398 | { |
428870ff | 1399 | rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; |
b128c09f | 1400 | |
428870ff BB |
1401 | ztest_rll_lock(rll, type); |
1402 | } | |
b128c09f | 1403 | |
428870ff BB |
1404 | static void |
1405 | ztest_object_unlock(ztest_ds_t *zd, uint64_t object) | |
1406 | { | |
1407 | rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; | |
b128c09f | 1408 | |
428870ff | 1409 | ztest_rll_unlock(rll); |
b128c09f BB |
1410 | } |
1411 | ||
e3a07cd0 BP |
1412 | static ztest_zrl_t * |
1413 | ztest_zrl_init(rl_t *rl, ztest_znode_t *zp) | |
34dc7c2f | 1414 | { |
e3a07cd0 BP |
1415 | ztest_zrl_t *zrl = umem_alloc(sizeof (*zrl), UMEM_NOFAIL); |
1416 | zrl->z_rl = rl; | |
1417 | zrl->z_ztznode = zp; | |
1418 | return (zrl); | |
428870ff | 1419 | } |
34dc7c2f | 1420 | |
428870ff | 1421 | static void |
e3a07cd0 | 1422 | ztest_zrl_fini(ztest_zrl_t *zrl) |
428870ff | 1423 | { |
e3a07cd0 BP |
1424 | umem_free(zrl, sizeof (*zrl)); |
1425 | } | |
34dc7c2f | 1426 | |
e3a07cd0 BP |
1427 | static ztest_zrl_t * |
1428 | ztest_range_lock(ztest_ds_t *zd, uint64_t object, uint64_t offset, | |
1429 | uint64_t size, rl_type_t type) | |
1430 | { | |
1431 | ztest_znode_t *zp = ztest_znode_get(zd, object); | |
1432 | rl_t *rl = zfs_range_lock(&zp->z_range_lock, offset, | |
1433 | size, type); | |
1434 | return (ztest_zrl_init(rl, zp)); | |
1435 | } | |
34dc7c2f | 1436 | |
e3a07cd0 BP |
1437 | static void |
1438 | ztest_range_unlock(ztest_ds_t *zd, ztest_zrl_t *zrl) | |
1439 | { | |
1440 | zfs_range_unlock(zrl->z_rl); | |
1441 | ztest_znode_put(zd, zrl->z_ztznode); | |
1442 | ztest_zrl_fini(zrl); | |
428870ff | 1443 | } |
34dc7c2f | 1444 | |
428870ff | 1445 | static void |
c242c188 | 1446 | ztest_zd_init(ztest_ds_t *zd, ztest_shared_ds_t *szd, objset_t *os) |
428870ff BB |
1447 | { |
1448 | zd->zd_os = os; | |
1449 | zd->zd_zilog = dmu_objset_zil(os); | |
c242c188 | 1450 | zd->zd_shared = szd; |
428870ff | 1451 | dmu_objset_name(os, zd->zd_name); |
d6320ddb | 1452 | int l; |
428870ff | 1453 | |
c242c188 CS |
1454 | if (zd->zd_shared != NULL) |
1455 | zd->zd_shared->zd_seq = 0; | |
1456 | ||
7809eb8b | 1457 | VERIFY(rwlock_init(&zd->zd_zilog_lock, USYNC_THREAD, NULL) == 0); |
1e33ac1e | 1458 | mutex_init(&zd->zd_dirobj_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 1459 | |
d6320ddb | 1460 | for (l = 0; l < ZTEST_OBJECT_LOCKS; l++) |
428870ff | 1461 | ztest_rll_init(&zd->zd_object_lock[l]); |
34dc7c2f | 1462 | |
d6320ddb | 1463 | for (l = 0; l < ZTEST_RANGE_LOCKS; l++) |
e3a07cd0 | 1464 | ztest_zll_init(&zd->zd_range_lock[l]); |
34dc7c2f BB |
1465 | } |
1466 | ||
428870ff BB |
1467 | static void |
1468 | ztest_zd_fini(ztest_ds_t *zd) | |
34dc7c2f | 1469 | { |
d6320ddb BB |
1470 | int l; |
1471 | ||
1e33ac1e | 1472 | mutex_destroy(&zd->zd_dirobj_lock); |
7809eb8b | 1473 | (void) rwlock_destroy(&zd->zd_zilog_lock); |
34dc7c2f | 1474 | |
d6320ddb | 1475 | for (l = 0; l < ZTEST_OBJECT_LOCKS; l++) |
428870ff | 1476 | ztest_rll_destroy(&zd->zd_object_lock[l]); |
b128c09f | 1477 | |
d6320ddb | 1478 | for (l = 0; l < ZTEST_RANGE_LOCKS; l++) |
e3a07cd0 | 1479 | ztest_zll_destroy(&zd->zd_range_lock[l]); |
428870ff | 1480 | } |
b128c09f | 1481 | |
428870ff | 1482 | #define TXG_MIGHTWAIT (ztest_random(10) == 0 ? TXG_NOWAIT : TXG_WAIT) |
b128c09f | 1483 | |
428870ff BB |
1484 | static uint64_t |
1485 | ztest_tx_assign(dmu_tx_t *tx, uint64_t txg_how, const char *tag) | |
1486 | { | |
1487 | uint64_t txg; | |
1488 | int error; | |
1489 | ||
1490 | /* | |
1491 | * Attempt to assign tx to some transaction group. | |
1492 | */ | |
1493 | error = dmu_tx_assign(tx, txg_how); | |
1494 | if (error) { | |
1495 | if (error == ERESTART) { | |
1496 | ASSERT(txg_how == TXG_NOWAIT); | |
1497 | dmu_tx_wait(tx); | |
1498 | } else { | |
1499 | ASSERT3U(error, ==, ENOSPC); | |
1500 | ztest_record_enospc(tag); | |
1501 | } | |
1502 | dmu_tx_abort(tx); | |
1503 | return (0); | |
1504 | } | |
1505 | txg = dmu_tx_get_txg(tx); | |
1506 | ASSERT(txg != 0); | |
1507 | return (txg); | |
1508 | } | |
1509 | ||
1510 | static void | |
1511 | ztest_pattern_set(void *buf, uint64_t size, uint64_t value) | |
1512 | { | |
1513 | uint64_t *ip = buf; | |
1514 | uint64_t *ip_end = (uint64_t *)((uintptr_t)buf + (uintptr_t)size); | |
1515 | ||
1516 | while (ip < ip_end) | |
1517 | *ip++ = value; | |
1518 | } | |
1519 | ||
1fde1e37 | 1520 | #ifndef NDEBUG |
428870ff BB |
1521 | static boolean_t |
1522 | ztest_pattern_match(void *buf, uint64_t size, uint64_t value) | |
1523 | { | |
1524 | uint64_t *ip = buf; | |
1525 | uint64_t *ip_end = (uint64_t *)((uintptr_t)buf + (uintptr_t)size); | |
1526 | uint64_t diff = 0; | |
1527 | ||
1528 | while (ip < ip_end) | |
1529 | diff |= (value - *ip++); | |
1530 | ||
1531 | return (diff == 0); | |
1532 | } | |
1fde1e37 | 1533 | #endif |
428870ff BB |
1534 | |
1535 | static void | |
1536 | ztest_bt_generate(ztest_block_tag_t *bt, objset_t *os, uint64_t object, | |
50c957f7 NB |
1537 | uint64_t dnodesize, uint64_t offset, uint64_t gen, uint64_t txg, |
1538 | uint64_t crtxg) | |
428870ff BB |
1539 | { |
1540 | bt->bt_magic = BT_MAGIC; | |
1541 | bt->bt_objset = dmu_objset_id(os); | |
1542 | bt->bt_object = object; | |
50c957f7 | 1543 | bt->bt_dnodesize = dnodesize; |
428870ff BB |
1544 | bt->bt_offset = offset; |
1545 | bt->bt_gen = gen; | |
1546 | bt->bt_txg = txg; | |
1547 | bt->bt_crtxg = crtxg; | |
1548 | } | |
1549 | ||
1550 | static void | |
1551 | ztest_bt_verify(ztest_block_tag_t *bt, objset_t *os, uint64_t object, | |
50c957f7 NB |
1552 | uint64_t dnodesize, uint64_t offset, uint64_t gen, uint64_t txg, |
1553 | uint64_t crtxg) | |
428870ff | 1554 | { |
9b67f605 MA |
1555 | ASSERT3U(bt->bt_magic, ==, BT_MAGIC); |
1556 | ASSERT3U(bt->bt_objset, ==, dmu_objset_id(os)); | |
1557 | ASSERT3U(bt->bt_object, ==, object); | |
50c957f7 | 1558 | ASSERT3U(bt->bt_dnodesize, ==, dnodesize); |
9b67f605 MA |
1559 | ASSERT3U(bt->bt_offset, ==, offset); |
1560 | ASSERT3U(bt->bt_gen, <=, gen); | |
1561 | ASSERT3U(bt->bt_txg, <=, txg); | |
1562 | ASSERT3U(bt->bt_crtxg, ==, crtxg); | |
428870ff BB |
1563 | } |
1564 | ||
1565 | static ztest_block_tag_t * | |
1566 | ztest_bt_bonus(dmu_buf_t *db) | |
1567 | { | |
1568 | dmu_object_info_t doi; | |
1569 | ztest_block_tag_t *bt; | |
1570 | ||
1571 | dmu_object_info_from_db(db, &doi); | |
1572 | ASSERT3U(doi.doi_bonus_size, <=, db->db_size); | |
1573 | ASSERT3U(doi.doi_bonus_size, >=, sizeof (*bt)); | |
1574 | bt = (void *)((char *)db->db_data + doi.doi_bonus_size - sizeof (*bt)); | |
1575 | ||
1576 | return (bt); | |
1577 | } | |
1578 | ||
50c957f7 NB |
1579 | /* |
1580 | * Generate a token to fill up unused bonus buffer space. Try to make | |
1581 | * it unique to the object, generation, and offset to verify that data | |
1582 | * is not getting overwritten by data from other dnodes. | |
1583 | */ | |
1584 | #define ZTEST_BONUS_FILL_TOKEN(obj, ds, gen, offset) \ | |
1585 | (((ds) << 48) | ((gen) << 32) | ((obj) << 8) | (offset)) | |
1586 | ||
1587 | /* | |
1588 | * Fill up the unused bonus buffer region before the block tag with a | |
1589 | * verifiable pattern. Filling the whole bonus area with non-zero data | |
1590 | * helps ensure that all dnode traversal code properly skips the | |
1591 | * interior regions of large dnodes. | |
1592 | */ | |
1593 | void | |
1594 | ztest_fill_unused_bonus(dmu_buf_t *db, void *end, uint64_t obj, | |
1595 | objset_t *os, uint64_t gen) | |
1596 | { | |
1597 | uint64_t *bonusp; | |
1598 | ||
1599 | ASSERT(IS_P2ALIGNED((char *)end - (char *)db->db_data, 8)); | |
1600 | ||
1601 | for (bonusp = db->db_data; bonusp < (uint64_t *)end; bonusp++) { | |
1602 | uint64_t token = ZTEST_BONUS_FILL_TOKEN(obj, dmu_objset_id(os), | |
1603 | gen, bonusp - (uint64_t *)db->db_data); | |
1604 | *bonusp = token; | |
1605 | } | |
1606 | } | |
1607 | ||
1608 | /* | |
1609 | * Verify that the unused area of a bonus buffer is filled with the | |
1610 | * expected tokens. | |
1611 | */ | |
1612 | void | |
1613 | ztest_verify_unused_bonus(dmu_buf_t *db, void *end, uint64_t obj, | |
1614 | objset_t *os, uint64_t gen) | |
1615 | { | |
1616 | uint64_t *bonusp; | |
1617 | ||
1618 | for (bonusp = db->db_data; bonusp < (uint64_t *)end; bonusp++) { | |
1619 | uint64_t token = ZTEST_BONUS_FILL_TOKEN(obj, dmu_objset_id(os), | |
1620 | gen, bonusp - (uint64_t *)db->db_data); | |
1621 | VERIFY3U(*bonusp, ==, token); | |
1622 | } | |
1623 | } | |
1624 | ||
428870ff BB |
1625 | /* |
1626 | * ZIL logging ops | |
1627 | */ | |
1628 | ||
1629 | #define lrz_type lr_mode | |
1630 | #define lrz_blocksize lr_uid | |
1631 | #define lrz_ibshift lr_gid | |
1632 | #define lrz_bonustype lr_rdev | |
50c957f7 | 1633 | #define lrz_dnodesize lr_crtime[1] |
428870ff | 1634 | |
572e2857 | 1635 | static void |
428870ff BB |
1636 | ztest_log_create(ztest_ds_t *zd, dmu_tx_t *tx, lr_create_t *lr) |
1637 | { | |
1638 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1639 | size_t namesize = strlen(name) + 1; | |
1640 | itx_t *itx; | |
1641 | ||
1642 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1643 | return; |
428870ff BB |
1644 | |
1645 | itx = zil_itx_create(TX_CREATE, sizeof (*lr) + namesize); | |
1646 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1647 | sizeof (*lr) + namesize - sizeof (lr_t)); | |
1648 | ||
572e2857 | 1649 | zil_itx_assign(zd->zd_zilog, itx, tx); |
428870ff BB |
1650 | } |
1651 | ||
572e2857 BB |
1652 | static void |
1653 | ztest_log_remove(ztest_ds_t *zd, dmu_tx_t *tx, lr_remove_t *lr, uint64_t object) | |
428870ff BB |
1654 | { |
1655 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1656 | size_t namesize = strlen(name) + 1; | |
1657 | itx_t *itx; | |
1658 | ||
1659 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1660 | return; |
428870ff BB |
1661 | |
1662 | itx = zil_itx_create(TX_REMOVE, sizeof (*lr) + namesize); | |
1663 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1664 | sizeof (*lr) + namesize - sizeof (lr_t)); | |
1665 | ||
572e2857 BB |
1666 | itx->itx_oid = object; |
1667 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1668 | } |
1669 | ||
572e2857 | 1670 | static void |
428870ff BB |
1671 | ztest_log_write(ztest_ds_t *zd, dmu_tx_t *tx, lr_write_t *lr) |
1672 | { | |
1673 | itx_t *itx; | |
1674 | itx_wr_state_t write_state = ztest_random(WR_NUM_STATES); | |
1675 | ||
1676 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1677 | return; |
428870ff BB |
1678 | |
1679 | if (lr->lr_length > ZIL_MAX_LOG_DATA) | |
1680 | write_state = WR_INDIRECT; | |
1681 | ||
1682 | itx = zil_itx_create(TX_WRITE, | |
1683 | sizeof (*lr) + (write_state == WR_COPIED ? lr->lr_length : 0)); | |
1684 | ||
1685 | if (write_state == WR_COPIED && | |
1686 | dmu_read(zd->zd_os, lr->lr_foid, lr->lr_offset, lr->lr_length, | |
1687 | ((lr_write_t *)&itx->itx_lr) + 1, DMU_READ_NO_PREFETCH) != 0) { | |
1688 | zil_itx_destroy(itx); | |
1689 | itx = zil_itx_create(TX_WRITE, sizeof (*lr)); | |
1690 | write_state = WR_NEED_COPY; | |
1691 | } | |
1692 | itx->itx_private = zd; | |
1693 | itx->itx_wr_state = write_state; | |
1694 | itx->itx_sync = (ztest_random(8) == 0); | |
428870ff BB |
1695 | |
1696 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1697 | sizeof (*lr) - sizeof (lr_t)); | |
1698 | ||
572e2857 | 1699 | zil_itx_assign(zd->zd_zilog, itx, tx); |
428870ff BB |
1700 | } |
1701 | ||
572e2857 | 1702 | static void |
428870ff BB |
1703 | ztest_log_truncate(ztest_ds_t *zd, dmu_tx_t *tx, lr_truncate_t *lr) |
1704 | { | |
1705 | itx_t *itx; | |
1706 | ||
1707 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1708 | return; |
428870ff BB |
1709 | |
1710 | itx = zil_itx_create(TX_TRUNCATE, sizeof (*lr)); | |
1711 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1712 | sizeof (*lr) - sizeof (lr_t)); | |
1713 | ||
572e2857 BB |
1714 | itx->itx_sync = B_FALSE; |
1715 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1716 | } |
1717 | ||
572e2857 | 1718 | static void |
428870ff BB |
1719 | ztest_log_setattr(ztest_ds_t *zd, dmu_tx_t *tx, lr_setattr_t *lr) |
1720 | { | |
1721 | itx_t *itx; | |
1722 | ||
1723 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1724 | return; |
428870ff BB |
1725 | |
1726 | itx = zil_itx_create(TX_SETATTR, sizeof (*lr)); | |
1727 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1728 | sizeof (*lr) - sizeof (lr_t)); | |
1729 | ||
572e2857 BB |
1730 | itx->itx_sync = B_FALSE; |
1731 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1732 | } |
1733 | ||
1734 | /* | |
1735 | * ZIL replay ops | |
1736 | */ | |
1737 | static int | |
867959b5 | 1738 | ztest_replay_create(void *arg1, void *arg2, boolean_t byteswap) |
428870ff | 1739 | { |
867959b5 BB |
1740 | ztest_ds_t *zd = arg1; |
1741 | lr_create_t *lr = arg2; | |
428870ff BB |
1742 | char *name = (void *)(lr + 1); /* name follows lr */ |
1743 | objset_t *os = zd->zd_os; | |
1744 | ztest_block_tag_t *bbt; | |
1745 | dmu_buf_t *db; | |
1746 | dmu_tx_t *tx; | |
1747 | uint64_t txg; | |
1748 | int error = 0; | |
50c957f7 | 1749 | int bonuslen; |
428870ff BB |
1750 | |
1751 | if (byteswap) | |
1752 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1753 | ||
1754 | ASSERT(lr->lr_doid == ZTEST_DIROBJ); | |
1755 | ASSERT(name[0] != '\0'); | |
1756 | ||
1757 | tx = dmu_tx_create(os); | |
1758 | ||
1759 | dmu_tx_hold_zap(tx, lr->lr_doid, B_TRUE, name); | |
1760 | ||
1761 | if (lr->lrz_type == DMU_OT_ZAP_OTHER) { | |
1762 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, B_TRUE, NULL); | |
1763 | } else { | |
1764 | dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT); | |
1765 | } | |
1766 | ||
1767 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1768 | if (txg == 0) | |
1769 | return (ENOSPC); | |
1770 | ||
1771 | ASSERT(dmu_objset_zil(os)->zl_replay == !!lr->lr_foid); | |
50c957f7 | 1772 | bonuslen = DN_BONUS_SIZE(lr->lrz_dnodesize); |
428870ff BB |
1773 | |
1774 | if (lr->lrz_type == DMU_OT_ZAP_OTHER) { | |
1775 | if (lr->lr_foid == 0) { | |
50c957f7 | 1776 | lr->lr_foid = zap_create_dnsize(os, |
428870ff | 1777 | lr->lrz_type, lr->lrz_bonustype, |
50c957f7 | 1778 | bonuslen, lr->lrz_dnodesize, tx); |
428870ff | 1779 | } else { |
50c957f7 | 1780 | error = zap_create_claim_dnsize(os, lr->lr_foid, |
428870ff | 1781 | lr->lrz_type, lr->lrz_bonustype, |
50c957f7 | 1782 | bonuslen, lr->lrz_dnodesize, tx); |
428870ff BB |
1783 | } |
1784 | } else { | |
1785 | if (lr->lr_foid == 0) { | |
50c957f7 | 1786 | lr->lr_foid = dmu_object_alloc_dnsize(os, |
428870ff | 1787 | lr->lrz_type, 0, lr->lrz_bonustype, |
50c957f7 | 1788 | bonuslen, lr->lrz_dnodesize, tx); |
428870ff | 1789 | } else { |
50c957f7 | 1790 | error = dmu_object_claim_dnsize(os, lr->lr_foid, |
428870ff | 1791 | lr->lrz_type, 0, lr->lrz_bonustype, |
50c957f7 | 1792 | bonuslen, lr->lrz_dnodesize, tx); |
428870ff BB |
1793 | } |
1794 | } | |
1795 | ||
1796 | if (error) { | |
1797 | ASSERT3U(error, ==, EEXIST); | |
1798 | ASSERT(zd->zd_zilog->zl_replay); | |
1799 | dmu_tx_commit(tx); | |
1800 | return (error); | |
1801 | } | |
1802 | ||
1803 | ASSERT(lr->lr_foid != 0); | |
1804 | ||
1805 | if (lr->lrz_type != DMU_OT_ZAP_OTHER) | |
1806 | VERIFY3U(0, ==, dmu_object_set_blocksize(os, lr->lr_foid, | |
1807 | lr->lrz_blocksize, lr->lrz_ibshift, tx)); | |
1808 | ||
1809 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
1810 | bbt = ztest_bt_bonus(db); | |
1811 | dmu_buf_will_dirty(db, tx); | |
50c957f7 NB |
1812 | ztest_bt_generate(bbt, os, lr->lr_foid, lr->lrz_dnodesize, -1ULL, |
1813 | lr->lr_gen, txg, txg); | |
1814 | ztest_fill_unused_bonus(db, bbt, lr->lr_foid, os, lr->lr_gen); | |
428870ff BB |
1815 | dmu_buf_rele(db, FTAG); |
1816 | ||
1817 | VERIFY3U(0, ==, zap_add(os, lr->lr_doid, name, sizeof (uint64_t), 1, | |
1818 | &lr->lr_foid, tx)); | |
1819 | ||
1820 | (void) ztest_log_create(zd, tx, lr); | |
1821 | ||
1822 | dmu_tx_commit(tx); | |
1823 | ||
1824 | return (0); | |
1825 | } | |
1826 | ||
1827 | static int | |
867959b5 | 1828 | ztest_replay_remove(void *arg1, void *arg2, boolean_t byteswap) |
428870ff | 1829 | { |
867959b5 BB |
1830 | ztest_ds_t *zd = arg1; |
1831 | lr_remove_t *lr = arg2; | |
428870ff BB |
1832 | char *name = (void *)(lr + 1); /* name follows lr */ |
1833 | objset_t *os = zd->zd_os; | |
1834 | dmu_object_info_t doi; | |
1835 | dmu_tx_t *tx; | |
1836 | uint64_t object, txg; | |
1837 | ||
1838 | if (byteswap) | |
1839 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1840 | ||
1841 | ASSERT(lr->lr_doid == ZTEST_DIROBJ); | |
1842 | ASSERT(name[0] != '\0'); | |
1843 | ||
1844 | VERIFY3U(0, ==, | |
1845 | zap_lookup(os, lr->lr_doid, name, sizeof (object), 1, &object)); | |
1846 | ASSERT(object != 0); | |
1847 | ||
1848 | ztest_object_lock(zd, object, RL_WRITER); | |
1849 | ||
1850 | VERIFY3U(0, ==, dmu_object_info(os, object, &doi)); | |
1851 | ||
1852 | tx = dmu_tx_create(os); | |
1853 | ||
1854 | dmu_tx_hold_zap(tx, lr->lr_doid, B_FALSE, name); | |
1855 | dmu_tx_hold_free(tx, object, 0, DMU_OBJECT_END); | |
1856 | ||
1857 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1858 | if (txg == 0) { | |
1859 | ztest_object_unlock(zd, object); | |
1860 | return (ENOSPC); | |
1861 | } | |
1862 | ||
1863 | if (doi.doi_type == DMU_OT_ZAP_OTHER) { | |
1864 | VERIFY3U(0, ==, zap_destroy(os, object, tx)); | |
1865 | } else { | |
1866 | VERIFY3U(0, ==, dmu_object_free(os, object, tx)); | |
1867 | } | |
1868 | ||
1869 | VERIFY3U(0, ==, zap_remove(os, lr->lr_doid, name, tx)); | |
1870 | ||
572e2857 | 1871 | (void) ztest_log_remove(zd, tx, lr, object); |
428870ff BB |
1872 | |
1873 | dmu_tx_commit(tx); | |
1874 | ||
1875 | ztest_object_unlock(zd, object); | |
1876 | ||
1877 | return (0); | |
1878 | } | |
1879 | ||
1880 | static int | |
867959b5 | 1881 | ztest_replay_write(void *arg1, void *arg2, boolean_t byteswap) |
428870ff | 1882 | { |
867959b5 BB |
1883 | ztest_ds_t *zd = arg1; |
1884 | lr_write_t *lr = arg2; | |
428870ff BB |
1885 | objset_t *os = zd->zd_os; |
1886 | void *data = lr + 1; /* data follows lr */ | |
1887 | uint64_t offset, length; | |
1888 | ztest_block_tag_t *bt = data; | |
1889 | ztest_block_tag_t *bbt; | |
1890 | uint64_t gen, txg, lrtxg, crtxg; | |
1891 | dmu_object_info_t doi; | |
1892 | dmu_tx_t *tx; | |
1893 | dmu_buf_t *db; | |
1894 | arc_buf_t *abuf = NULL; | |
e3a07cd0 | 1895 | ztest_zrl_t *rl; |
428870ff BB |
1896 | |
1897 | if (byteswap) | |
1898 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1899 | ||
1900 | offset = lr->lr_offset; | |
1901 | length = lr->lr_length; | |
1902 | ||
1903 | /* If it's a dmu_sync() block, write the whole block */ | |
1904 | if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) { | |
1905 | uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr); | |
1906 | if (length < blocksize) { | |
1907 | offset -= offset % blocksize; | |
1908 | length = blocksize; | |
1909 | } | |
1910 | } | |
1911 | ||
1912 | if (bt->bt_magic == BSWAP_64(BT_MAGIC)) | |
1913 | byteswap_uint64_array(bt, sizeof (*bt)); | |
1914 | ||
1915 | if (bt->bt_magic != BT_MAGIC) | |
1916 | bt = NULL; | |
1917 | ||
1918 | ztest_object_lock(zd, lr->lr_foid, RL_READER); | |
1919 | rl = ztest_range_lock(zd, lr->lr_foid, offset, length, RL_WRITER); | |
1920 | ||
1921 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
1922 | ||
1923 | dmu_object_info_from_db(db, &doi); | |
1924 | ||
1925 | bbt = ztest_bt_bonus(db); | |
1926 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
1927 | gen = bbt->bt_gen; | |
1928 | crtxg = bbt->bt_crtxg; | |
1929 | lrtxg = lr->lr_common.lrc_txg; | |
1930 | ||
1931 | tx = dmu_tx_create(os); | |
1932 | ||
1933 | dmu_tx_hold_write(tx, lr->lr_foid, offset, length); | |
1934 | ||
1935 | if (ztest_random(8) == 0 && length == doi.doi_data_block_size && | |
1936 | P2PHASE(offset, length) == 0) | |
1937 | abuf = dmu_request_arcbuf(db, length); | |
1938 | ||
1939 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1940 | if (txg == 0) { | |
1941 | if (abuf != NULL) | |
1942 | dmu_return_arcbuf(abuf); | |
1943 | dmu_buf_rele(db, FTAG); | |
e3a07cd0 | 1944 | ztest_range_unlock(zd, rl); |
428870ff BB |
1945 | ztest_object_unlock(zd, lr->lr_foid); |
1946 | return (ENOSPC); | |
1947 | } | |
1948 | ||
1949 | if (bt != NULL) { | |
1950 | /* | |
1951 | * Usually, verify the old data before writing new data -- | |
1952 | * but not always, because we also want to verify correct | |
1953 | * behavior when the data was not recently read into cache. | |
1954 | */ | |
1955 | ASSERT(offset % doi.doi_data_block_size == 0); | |
1956 | if (ztest_random(4) != 0) { | |
1957 | int prefetch = ztest_random(2) ? | |
1958 | DMU_READ_PREFETCH : DMU_READ_NO_PREFETCH; | |
1959 | ztest_block_tag_t rbt; | |
1960 | ||
1961 | VERIFY(dmu_read(os, lr->lr_foid, offset, | |
1962 | sizeof (rbt), &rbt, prefetch) == 0); | |
1963 | if (rbt.bt_magic == BT_MAGIC) { | |
50c957f7 | 1964 | ztest_bt_verify(&rbt, os, lr->lr_foid, 0, |
428870ff BB |
1965 | offset, gen, txg, crtxg); |
1966 | } | |
1967 | } | |
1968 | ||
1969 | /* | |
1970 | * Writes can appear to be newer than the bonus buffer because | |
1971 | * the ztest_get_data() callback does a dmu_read() of the | |
1972 | * open-context data, which may be different than the data | |
1973 | * as it was when the write was generated. | |
1974 | */ | |
1975 | if (zd->zd_zilog->zl_replay) { | |
50c957f7 | 1976 | ztest_bt_verify(bt, os, lr->lr_foid, 0, offset, |
428870ff BB |
1977 | MAX(gen, bt->bt_gen), MAX(txg, lrtxg), |
1978 | bt->bt_crtxg); | |
1979 | } | |
1980 | ||
1981 | /* | |
1982 | * Set the bt's gen/txg to the bonus buffer's gen/txg | |
1983 | * so that all of the usual ASSERTs will work. | |
1984 | */ | |
50c957f7 NB |
1985 | ztest_bt_generate(bt, os, lr->lr_foid, 0, offset, gen, txg, |
1986 | crtxg); | |
428870ff BB |
1987 | } |
1988 | ||
1989 | if (abuf == NULL) { | |
1990 | dmu_write(os, lr->lr_foid, offset, length, data, tx); | |
1991 | } else { | |
1992 | bcopy(data, abuf->b_data, length); | |
440a3eb9 | 1993 | dmu_assign_arcbuf_by_dbuf(db, offset, abuf, tx); |
428870ff BB |
1994 | } |
1995 | ||
1996 | (void) ztest_log_write(zd, tx, lr); | |
1997 | ||
1998 | dmu_buf_rele(db, FTAG); | |
1999 | ||
2000 | dmu_tx_commit(tx); | |
2001 | ||
e3a07cd0 | 2002 | ztest_range_unlock(zd, rl); |
428870ff BB |
2003 | ztest_object_unlock(zd, lr->lr_foid); |
2004 | ||
2005 | return (0); | |
2006 | } | |
2007 | ||
2008 | static int | |
867959b5 | 2009 | ztest_replay_truncate(void *arg1, void *arg2, boolean_t byteswap) |
428870ff | 2010 | { |
867959b5 BB |
2011 | ztest_ds_t *zd = arg1; |
2012 | lr_truncate_t *lr = arg2; | |
428870ff BB |
2013 | objset_t *os = zd->zd_os; |
2014 | dmu_tx_t *tx; | |
2015 | uint64_t txg; | |
e3a07cd0 | 2016 | ztest_zrl_t *rl; |
428870ff BB |
2017 | |
2018 | if (byteswap) | |
2019 | byteswap_uint64_array(lr, sizeof (*lr)); | |
2020 | ||
2021 | ztest_object_lock(zd, lr->lr_foid, RL_READER); | |
2022 | rl = ztest_range_lock(zd, lr->lr_foid, lr->lr_offset, lr->lr_length, | |
2023 | RL_WRITER); | |
2024 | ||
2025 | tx = dmu_tx_create(os); | |
2026 | ||
2027 | dmu_tx_hold_free(tx, lr->lr_foid, lr->lr_offset, lr->lr_length); | |
2028 | ||
2029 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
2030 | if (txg == 0) { | |
e3a07cd0 | 2031 | ztest_range_unlock(zd, rl); |
428870ff BB |
2032 | ztest_object_unlock(zd, lr->lr_foid); |
2033 | return (ENOSPC); | |
2034 | } | |
2035 | ||
2036 | VERIFY(dmu_free_range(os, lr->lr_foid, lr->lr_offset, | |
2037 | lr->lr_length, tx) == 0); | |
2038 | ||
2039 | (void) ztest_log_truncate(zd, tx, lr); | |
2040 | ||
2041 | dmu_tx_commit(tx); | |
2042 | ||
e3a07cd0 | 2043 | ztest_range_unlock(zd, rl); |
428870ff BB |
2044 | ztest_object_unlock(zd, lr->lr_foid); |
2045 | ||
2046 | return (0); | |
2047 | } | |
2048 | ||
2049 | static int | |
867959b5 | 2050 | ztest_replay_setattr(void *arg1, void *arg2, boolean_t byteswap) |
428870ff | 2051 | { |
867959b5 BB |
2052 | ztest_ds_t *zd = arg1; |
2053 | lr_setattr_t *lr = arg2; | |
428870ff BB |
2054 | objset_t *os = zd->zd_os; |
2055 | dmu_tx_t *tx; | |
2056 | dmu_buf_t *db; | |
2057 | ztest_block_tag_t *bbt; | |
50c957f7 | 2058 | uint64_t txg, lrtxg, crtxg, dnodesize; |
428870ff BB |
2059 | |
2060 | if (byteswap) | |
2061 | byteswap_uint64_array(lr, sizeof (*lr)); | |
2062 | ||
2063 | ztest_object_lock(zd, lr->lr_foid, RL_WRITER); | |
2064 | ||
2065 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
2066 | ||
2067 | tx = dmu_tx_create(os); | |
2068 | dmu_tx_hold_bonus(tx, lr->lr_foid); | |
2069 | ||
2070 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
2071 | if (txg == 0) { | |
2072 | dmu_buf_rele(db, FTAG); | |
2073 | ztest_object_unlock(zd, lr->lr_foid); | |
2074 | return (ENOSPC); | |
2075 | } | |
2076 | ||
2077 | bbt = ztest_bt_bonus(db); | |
2078 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
2079 | crtxg = bbt->bt_crtxg; | |
2080 | lrtxg = lr->lr_common.lrc_txg; | |
50c957f7 | 2081 | dnodesize = bbt->bt_dnodesize; |
428870ff BB |
2082 | |
2083 | if (zd->zd_zilog->zl_replay) { | |
2084 | ASSERT(lr->lr_size != 0); | |
2085 | ASSERT(lr->lr_mode != 0); | |
2086 | ASSERT(lrtxg != 0); | |
2087 | } else { | |
2088 | /* | |
2089 | * Randomly change the size and increment the generation. | |
2090 | */ | |
2091 | lr->lr_size = (ztest_random(db->db_size / sizeof (*bbt)) + 1) * | |
2092 | sizeof (*bbt); | |
2093 | lr->lr_mode = bbt->bt_gen + 1; | |
2094 | ASSERT(lrtxg == 0); | |
2095 | } | |
2096 | ||
2097 | /* | |
2098 | * Verify that the current bonus buffer is not newer than our txg. | |
2099 | */ | |
50c957f7 | 2100 | ztest_bt_verify(bbt, os, lr->lr_foid, dnodesize, -1ULL, lr->lr_mode, |
428870ff BB |
2101 | MAX(txg, lrtxg), crtxg); |
2102 | ||
2103 | dmu_buf_will_dirty(db, tx); | |
2104 | ||
2105 | ASSERT3U(lr->lr_size, >=, sizeof (*bbt)); | |
2106 | ASSERT3U(lr->lr_size, <=, db->db_size); | |
c99c9001 | 2107 | VERIFY0(dmu_set_bonus(db, lr->lr_size, tx)); |
428870ff BB |
2108 | bbt = ztest_bt_bonus(db); |
2109 | ||
50c957f7 NB |
2110 | ztest_bt_generate(bbt, os, lr->lr_foid, dnodesize, -1ULL, lr->lr_mode, |
2111 | txg, crtxg); | |
2112 | ztest_fill_unused_bonus(db, bbt, lr->lr_foid, os, bbt->bt_gen); | |
428870ff BB |
2113 | dmu_buf_rele(db, FTAG); |
2114 | ||
2115 | (void) ztest_log_setattr(zd, tx, lr); | |
2116 | ||
2117 | dmu_tx_commit(tx); | |
2118 | ||
2119 | ztest_object_unlock(zd, lr->lr_foid); | |
2120 | ||
2121 | return (0); | |
2122 | } | |
2123 | ||
867959b5 BB |
2124 | zil_replay_func_t *ztest_replay_vector[TX_MAX_TYPE] = { |
2125 | NULL, /* 0 no such transaction type */ | |
2126 | ztest_replay_create, /* TX_CREATE */ | |
2127 | NULL, /* TX_MKDIR */ | |
2128 | NULL, /* TX_MKXATTR */ | |
2129 | NULL, /* TX_SYMLINK */ | |
2130 | ztest_replay_remove, /* TX_REMOVE */ | |
2131 | NULL, /* TX_RMDIR */ | |
2132 | NULL, /* TX_LINK */ | |
2133 | NULL, /* TX_RENAME */ | |
2134 | ztest_replay_write, /* TX_WRITE */ | |
2135 | ztest_replay_truncate, /* TX_TRUNCATE */ | |
2136 | ztest_replay_setattr, /* TX_SETATTR */ | |
2137 | NULL, /* TX_ACL */ | |
2138 | NULL, /* TX_CREATE_ACL */ | |
2139 | NULL, /* TX_CREATE_ATTR */ | |
2140 | NULL, /* TX_CREATE_ACL_ATTR */ | |
2141 | NULL, /* TX_MKDIR_ACL */ | |
2142 | NULL, /* TX_MKDIR_ATTR */ | |
2143 | NULL, /* TX_MKDIR_ACL_ATTR */ | |
2144 | NULL, /* TX_WRITE2 */ | |
428870ff BB |
2145 | }; |
2146 | ||
2147 | /* | |
2148 | * ZIL get_data callbacks | |
2149 | */ | |
e3a07cd0 BP |
2150 | typedef struct ztest_zgd_private { |
2151 | ztest_ds_t *z_zd; | |
2152 | ztest_zrl_t *z_rl; | |
2153 | uint64_t z_object; | |
2154 | } ztest_zgd_private_t; | |
428870ff BB |
2155 | |
2156 | static void | |
2157 | ztest_get_done(zgd_t *zgd, int error) | |
2158 | { | |
e3a07cd0 BP |
2159 | ztest_zgd_private_t *zzp = zgd->zgd_private; |
2160 | ztest_ds_t *zd = zzp->z_zd; | |
2161 | uint64_t object = zzp->z_object; | |
428870ff BB |
2162 | |
2163 | if (zgd->zgd_db) | |
2164 | dmu_buf_rele(zgd->zgd_db, zgd); | |
2165 | ||
e3a07cd0 | 2166 | ztest_range_unlock(zd, zzp->z_rl); |
428870ff BB |
2167 | ztest_object_unlock(zd, object); |
2168 | ||
2169 | if (error == 0 && zgd->zgd_bp) | |
1ce23dca | 2170 | zil_lwb_add_block(zgd->zgd_lwb, zgd->zgd_bp); |
428870ff BB |
2171 | |
2172 | umem_free(zgd, sizeof (*zgd)); | |
e3a07cd0 | 2173 | umem_free(zzp, sizeof (*zzp)); |
428870ff BB |
2174 | } |
2175 | ||
2176 | static int | |
1ce23dca PS |
2177 | ztest_get_data(void *arg, lr_write_t *lr, char *buf, struct lwb *lwb, |
2178 | zio_t *zio) | |
428870ff BB |
2179 | { |
2180 | ztest_ds_t *zd = arg; | |
2181 | objset_t *os = zd->zd_os; | |
2182 | uint64_t object = lr->lr_foid; | |
2183 | uint64_t offset = lr->lr_offset; | |
2184 | uint64_t size = lr->lr_length; | |
428870ff BB |
2185 | uint64_t txg = lr->lr_common.lrc_txg; |
2186 | uint64_t crtxg; | |
2187 | dmu_object_info_t doi; | |
2188 | dmu_buf_t *db; | |
2189 | zgd_t *zgd; | |
2190 | int error; | |
e3a07cd0 | 2191 | ztest_zgd_private_t *zgd_private; |
428870ff | 2192 | |
1ce23dca PS |
2193 | ASSERT3P(lwb, !=, NULL); |
2194 | ASSERT3P(zio, !=, NULL); | |
2195 | ASSERT3U(size, !=, 0); | |
2196 | ||
428870ff BB |
2197 | ztest_object_lock(zd, object, RL_READER); |
2198 | error = dmu_bonus_hold(os, object, FTAG, &db); | |
2199 | if (error) { | |
2200 | ztest_object_unlock(zd, object); | |
2201 | return (error); | |
2202 | } | |
2203 | ||
2204 | crtxg = ztest_bt_bonus(db)->bt_crtxg; | |
2205 | ||
2206 | if (crtxg == 0 || crtxg > txg) { | |
2207 | dmu_buf_rele(db, FTAG); | |
2208 | ztest_object_unlock(zd, object); | |
2209 | return (ENOENT); | |
2210 | } | |
2211 | ||
2212 | dmu_object_info_from_db(db, &doi); | |
2213 | dmu_buf_rele(db, FTAG); | |
2214 | db = NULL; | |
2215 | ||
2216 | zgd = umem_zalloc(sizeof (*zgd), UMEM_NOFAIL); | |
1ce23dca | 2217 | zgd->zgd_lwb = lwb; |
e3a07cd0 BP |
2218 | zgd_private = umem_zalloc(sizeof (ztest_zgd_private_t), UMEM_NOFAIL); |
2219 | zgd_private->z_zd = zd; | |
2220 | zgd_private->z_object = object; | |
2221 | zgd->zgd_private = zgd_private; | |
428870ff BB |
2222 | |
2223 | if (buf != NULL) { /* immediate write */ | |
e3a07cd0 | 2224 | zgd_private->z_rl = ztest_range_lock(zd, object, offset, size, |
428870ff | 2225 | RL_READER); |
f763c3d1 | 2226 | zgd->zgd_rl = zgd_private->z_rl->z_rl; |
428870ff BB |
2227 | |
2228 | error = dmu_read(os, object, offset, size, buf, | |
2229 | DMU_READ_NO_PREFETCH); | |
2230 | ASSERT(error == 0); | |
2231 | } else { | |
2232 | size = doi.doi_data_block_size; | |
2233 | if (ISP2(size)) { | |
2234 | offset = P2ALIGN(offset, size); | |
2235 | } else { | |
2236 | ASSERT(offset < size); | |
2237 | offset = 0; | |
2238 | } | |
2239 | ||
e3a07cd0 | 2240 | zgd_private->z_rl = ztest_range_lock(zd, object, offset, size, |
428870ff | 2241 | RL_READER); |
f763c3d1 | 2242 | zgd->zgd_rl = zgd_private->z_rl->z_rl; |
428870ff BB |
2243 | |
2244 | error = dmu_buf_hold(os, object, offset, zgd, &db, | |
2245 | DMU_READ_NO_PREFETCH); | |
2246 | ||
2247 | if (error == 0) { | |
02dc43bc | 2248 | blkptr_t *bp = &lr->lr_blkptr; |
03c6040b | 2249 | |
428870ff BB |
2250 | zgd->zgd_db = db; |
2251 | zgd->zgd_bp = bp; | |
2252 | ||
2253 | ASSERT(db->db_offset == offset); | |
2254 | ASSERT(db->db_size == size); | |
2255 | ||
2256 | error = dmu_sync(zio, lr->lr_common.lrc_txg, | |
2257 | ztest_get_done, zgd); | |
2258 | ||
2259 | if (error == 0) | |
2260 | return (0); | |
2261 | } | |
2262 | } | |
2263 | ||
2264 | ztest_get_done(zgd, error); | |
2265 | ||
2266 | return (error); | |
2267 | } | |
2268 | ||
2269 | static void * | |
2270 | ztest_lr_alloc(size_t lrsize, char *name) | |
2271 | { | |
2272 | char *lr; | |
2273 | size_t namesize = name ? strlen(name) + 1 : 0; | |
2274 | ||
2275 | lr = umem_zalloc(lrsize + namesize, UMEM_NOFAIL); | |
2276 | ||
2277 | if (name) | |
2278 | bcopy(name, lr + lrsize, namesize); | |
2279 | ||
2280 | return (lr); | |
2281 | } | |
2282 | ||
2283 | void | |
2284 | ztest_lr_free(void *lr, size_t lrsize, char *name) | |
2285 | { | |
2286 | size_t namesize = name ? strlen(name) + 1 : 0; | |
2287 | ||
2288 | umem_free(lr, lrsize + namesize); | |
2289 | } | |
2290 | ||
2291 | /* | |
2292 | * Lookup a bunch of objects. Returns the number of objects not found. | |
2293 | */ | |
2294 | static int | |
2295 | ztest_lookup(ztest_ds_t *zd, ztest_od_t *od, int count) | |
2296 | { | |
2297 | int missing = 0; | |
2298 | int error; | |
d6320ddb | 2299 | int i; |
428870ff | 2300 | |
c25b8f99 | 2301 | ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock)); |
428870ff | 2302 | |
d6320ddb | 2303 | for (i = 0; i < count; i++, od++) { |
428870ff BB |
2304 | od->od_object = 0; |
2305 | error = zap_lookup(zd->zd_os, od->od_dir, od->od_name, | |
2306 | sizeof (uint64_t), 1, &od->od_object); | |
2307 | if (error) { | |
2308 | ASSERT(error == ENOENT); | |
2309 | ASSERT(od->od_object == 0); | |
2310 | missing++; | |
2311 | } else { | |
2312 | dmu_buf_t *db; | |
2313 | ztest_block_tag_t *bbt; | |
2314 | dmu_object_info_t doi; | |
2315 | ||
2316 | ASSERT(od->od_object != 0); | |
2317 | ASSERT(missing == 0); /* there should be no gaps */ | |
2318 | ||
2319 | ztest_object_lock(zd, od->od_object, RL_READER); | |
2320 | VERIFY3U(0, ==, dmu_bonus_hold(zd->zd_os, | |
2321 | od->od_object, FTAG, &db)); | |
2322 | dmu_object_info_from_db(db, &doi); | |
2323 | bbt = ztest_bt_bonus(db); | |
2324 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
2325 | od->od_type = doi.doi_type; | |
2326 | od->od_blocksize = doi.doi_data_block_size; | |
2327 | od->od_gen = bbt->bt_gen; | |
2328 | dmu_buf_rele(db, FTAG); | |
2329 | ztest_object_unlock(zd, od->od_object); | |
2330 | } | |
2331 | } | |
2332 | ||
2333 | return (missing); | |
2334 | } | |
2335 | ||
2336 | static int | |
2337 | ztest_create(ztest_ds_t *zd, ztest_od_t *od, int count) | |
2338 | { | |
2339 | int missing = 0; | |
d6320ddb | 2340 | int i; |
428870ff | 2341 | |
c25b8f99 | 2342 | ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock)); |
428870ff | 2343 | |
d6320ddb | 2344 | for (i = 0; i < count; i++, od++) { |
428870ff BB |
2345 | if (missing) { |
2346 | od->od_object = 0; | |
2347 | missing++; | |
2348 | continue; | |
2349 | } | |
2350 | ||
2351 | lr_create_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name); | |
2352 | ||
2353 | lr->lr_doid = od->od_dir; | |
2354 | lr->lr_foid = 0; /* 0 to allocate, > 0 to claim */ | |
2355 | lr->lrz_type = od->od_crtype; | |
2356 | lr->lrz_blocksize = od->od_crblocksize; | |
2357 | lr->lrz_ibshift = ztest_random_ibshift(); | |
2358 | lr->lrz_bonustype = DMU_OT_UINT64_OTHER; | |
50c957f7 | 2359 | lr->lrz_dnodesize = od->od_crdnodesize; |
428870ff BB |
2360 | lr->lr_gen = od->od_crgen; |
2361 | lr->lr_crtime[0] = time(NULL); | |
2362 | ||
2363 | if (ztest_replay_create(zd, lr, B_FALSE) != 0) { | |
2364 | ASSERT(missing == 0); | |
2365 | od->od_object = 0; | |
2366 | missing++; | |
2367 | } else { | |
2368 | od->od_object = lr->lr_foid; | |
2369 | od->od_type = od->od_crtype; | |
2370 | od->od_blocksize = od->od_crblocksize; | |
2371 | od->od_gen = od->od_crgen; | |
2372 | ASSERT(od->od_object != 0); | |
2373 | } | |
2374 | ||
2375 | ztest_lr_free(lr, sizeof (*lr), od->od_name); | |
2376 | } | |
2377 | ||
2378 | return (missing); | |
2379 | } | |
2380 | ||
2381 | static int | |
2382 | ztest_remove(ztest_ds_t *zd, ztest_od_t *od, int count) | |
2383 | { | |
2384 | int missing = 0; | |
2385 | int error; | |
d6320ddb | 2386 | int i; |
428870ff | 2387 | |
c25b8f99 | 2388 | ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock)); |
428870ff BB |
2389 | |
2390 | od += count - 1; | |
2391 | ||
d6320ddb | 2392 | for (i = count - 1; i >= 0; i--, od--) { |
428870ff BB |
2393 | if (missing) { |
2394 | missing++; | |
2395 | continue; | |
2396 | } | |
2397 | ||
03c6040b GW |
2398 | /* |
2399 | * No object was found. | |
2400 | */ | |
428870ff BB |
2401 | if (od->od_object == 0) |
2402 | continue; | |
2403 | ||
2404 | lr_remove_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name); | |
2405 | ||
2406 | lr->lr_doid = od->od_dir; | |
2407 | ||
2408 | if ((error = ztest_replay_remove(zd, lr, B_FALSE)) != 0) { | |
2409 | ASSERT3U(error, ==, ENOSPC); | |
2410 | missing++; | |
2411 | } else { | |
2412 | od->od_object = 0; | |
2413 | } | |
2414 | ztest_lr_free(lr, sizeof (*lr), od->od_name); | |
2415 | } | |
2416 | ||
2417 | return (missing); | |
2418 | } | |
2419 | ||
2420 | static int | |
2421 | ztest_write(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size, | |
2422 | void *data) | |
2423 | { | |
2424 | lr_write_t *lr; | |
2425 | int error; | |
2426 | ||
2427 | lr = ztest_lr_alloc(sizeof (*lr) + size, NULL); | |
2428 | ||
2429 | lr->lr_foid = object; | |
2430 | lr->lr_offset = offset; | |
2431 | lr->lr_length = size; | |
2432 | lr->lr_blkoff = 0; | |
2433 | BP_ZERO(&lr->lr_blkptr); | |
2434 | ||
2435 | bcopy(data, lr + 1, size); | |
2436 | ||
2437 | error = ztest_replay_write(zd, lr, B_FALSE); | |
2438 | ||
2439 | ztest_lr_free(lr, sizeof (*lr) + size, NULL); | |
2440 | ||
2441 | return (error); | |
2442 | } | |
2443 | ||
2444 | static int | |
2445 | ztest_truncate(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size) | |
2446 | { | |
2447 | lr_truncate_t *lr; | |
2448 | int error; | |
2449 | ||
2450 | lr = ztest_lr_alloc(sizeof (*lr), NULL); | |
2451 | ||
2452 | lr->lr_foid = object; | |
2453 | lr->lr_offset = offset; | |
2454 | lr->lr_length = size; | |
2455 | ||
2456 | error = ztest_replay_truncate(zd, lr, B_FALSE); | |
2457 | ||
2458 | ztest_lr_free(lr, sizeof (*lr), NULL); | |
2459 | ||
2460 | return (error); | |
2461 | } | |
2462 | ||
2463 | static int | |
2464 | ztest_setattr(ztest_ds_t *zd, uint64_t object) | |
2465 | { | |
2466 | lr_setattr_t *lr; | |
2467 | int error; | |
2468 | ||
2469 | lr = ztest_lr_alloc(sizeof (*lr), NULL); | |
2470 | ||
2471 | lr->lr_foid = object; | |
2472 | lr->lr_size = 0; | |
2473 | lr->lr_mode = 0; | |
2474 | ||
2475 | error = ztest_replay_setattr(zd, lr, B_FALSE); | |
2476 | ||
2477 | ztest_lr_free(lr, sizeof (*lr), NULL); | |
2478 | ||
2479 | return (error); | |
2480 | } | |
2481 | ||
2482 | static void | |
2483 | ztest_prealloc(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size) | |
2484 | { | |
2485 | objset_t *os = zd->zd_os; | |
2486 | dmu_tx_t *tx; | |
2487 | uint64_t txg; | |
e3a07cd0 | 2488 | ztest_zrl_t *rl; |
428870ff BB |
2489 | |
2490 | txg_wait_synced(dmu_objset_pool(os), 0); | |
2491 | ||
2492 | ztest_object_lock(zd, object, RL_READER); | |
2493 | rl = ztest_range_lock(zd, object, offset, size, RL_WRITER); | |
2494 | ||
2495 | tx = dmu_tx_create(os); | |
2496 | ||
2497 | dmu_tx_hold_write(tx, object, offset, size); | |
2498 | ||
2499 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
2500 | ||
2501 | if (txg != 0) { | |
2502 | dmu_prealloc(os, object, offset, size, tx); | |
2503 | dmu_tx_commit(tx); | |
2504 | txg_wait_synced(dmu_objset_pool(os), txg); | |
2505 | } else { | |
2506 | (void) dmu_free_long_range(os, object, offset, size); | |
2507 | } | |
2508 | ||
e3a07cd0 | 2509 | ztest_range_unlock(zd, rl); |
428870ff BB |
2510 | ztest_object_unlock(zd, object); |
2511 | } | |
2512 | ||
2513 | static void | |
2514 | ztest_io(ztest_ds_t *zd, uint64_t object, uint64_t offset) | |
2515 | { | |
03c6040b | 2516 | int err; |
428870ff BB |
2517 | ztest_block_tag_t wbt; |
2518 | dmu_object_info_t doi; | |
2519 | enum ztest_io_type io_type; | |
2520 | uint64_t blocksize; | |
2521 | void *data; | |
2522 | ||
2523 | VERIFY(dmu_object_info(zd->zd_os, object, &doi) == 0); | |
2524 | blocksize = doi.doi_data_block_size; | |
2525 | data = umem_alloc(blocksize, UMEM_NOFAIL); | |
2526 | ||
2527 | /* | |
2528 | * Pick an i/o type at random, biased toward writing block tags. | |
2529 | */ | |
2530 | io_type = ztest_random(ZTEST_IO_TYPES); | |
2531 | if (ztest_random(2) == 0) | |
2532 | io_type = ZTEST_IO_WRITE_TAG; | |
2533 | ||
7809eb8b | 2534 | (void) rw_rdlock(&zd->zd_zilog_lock); |
3e31d2b0 | 2535 | |
428870ff BB |
2536 | switch (io_type) { |
2537 | ||
2538 | case ZTEST_IO_WRITE_TAG: | |
50c957f7 NB |
2539 | ztest_bt_generate(&wbt, zd->zd_os, object, doi.doi_dnodesize, |
2540 | offset, 0, 0, 0); | |
428870ff BB |
2541 | (void) ztest_write(zd, object, offset, sizeof (wbt), &wbt); |
2542 | break; | |
2543 | ||
2544 | case ZTEST_IO_WRITE_PATTERN: | |
2545 | (void) memset(data, 'a' + (object + offset) % 5, blocksize); | |
2546 | if (ztest_random(2) == 0) { | |
2547 | /* | |
2548 | * Induce fletcher2 collisions to ensure that | |
2549 | * zio_ddt_collision() detects and resolves them | |
2550 | * when using fletcher2-verify for deduplication. | |
2551 | */ | |
2552 | ((uint64_t *)data)[0] ^= 1ULL << 63; | |
2553 | ((uint64_t *)data)[4] ^= 1ULL << 63; | |
2554 | } | |
2555 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2556 | break; | |
2557 | ||
2558 | case ZTEST_IO_WRITE_ZEROES: | |
2559 | bzero(data, blocksize); | |
2560 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2561 | break; | |
2562 | ||
2563 | case ZTEST_IO_TRUNCATE: | |
2564 | (void) ztest_truncate(zd, object, offset, blocksize); | |
2565 | break; | |
2566 | ||
2567 | case ZTEST_IO_SETATTR: | |
2568 | (void) ztest_setattr(zd, object); | |
2569 | break; | |
e75c13c3 BB |
2570 | default: |
2571 | break; | |
03c6040b GW |
2572 | |
2573 | case ZTEST_IO_REWRITE: | |
7809eb8b | 2574 | (void) rw_rdlock(&ztest_name_lock); |
03c6040b GW |
2575 | err = ztest_dsl_prop_set_uint64(zd->zd_name, |
2576 | ZFS_PROP_CHECKSUM, spa_dedup_checksum(ztest_spa), | |
2577 | B_FALSE); | |
2578 | VERIFY(err == 0 || err == ENOSPC); | |
2579 | err = ztest_dsl_prop_set_uint64(zd->zd_name, | |
2580 | ZFS_PROP_COMPRESSION, | |
2581 | ztest_random_dsl_prop(ZFS_PROP_COMPRESSION), | |
2582 | B_FALSE); | |
2583 | VERIFY(err == 0 || err == ENOSPC); | |
7809eb8b | 2584 | (void) rw_unlock(&ztest_name_lock); |
03c6040b GW |
2585 | |
2586 | VERIFY0(dmu_read(zd->zd_os, object, offset, blocksize, data, | |
2587 | DMU_READ_NO_PREFETCH)); | |
2588 | ||
2589 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2590 | break; | |
428870ff BB |
2591 | } |
2592 | ||
7809eb8b | 2593 | (void) rw_unlock(&zd->zd_zilog_lock); |
3e31d2b0 | 2594 | |
428870ff BB |
2595 | umem_free(data, blocksize); |
2596 | } | |
2597 | ||
2598 | /* | |
2599 | * Initialize an object description template. | |
2600 | */ | |
2601 | static void | |
2602 | ztest_od_init(ztest_od_t *od, uint64_t id, char *tag, uint64_t index, | |
50c957f7 NB |
2603 | dmu_object_type_t type, uint64_t blocksize, uint64_t dnodesize, |
2604 | uint64_t gen) | |
428870ff BB |
2605 | { |
2606 | od->od_dir = ZTEST_DIROBJ; | |
2607 | od->od_object = 0; | |
2608 | ||
2609 | od->od_crtype = type; | |
2610 | od->od_crblocksize = blocksize ? blocksize : ztest_random_blocksize(); | |
50c957f7 | 2611 | od->od_crdnodesize = dnodesize ? dnodesize : ztest_random_dnodesize(); |
428870ff BB |
2612 | od->od_crgen = gen; |
2613 | ||
2614 | od->od_type = DMU_OT_NONE; | |
2615 | od->od_blocksize = 0; | |
2616 | od->od_gen = 0; | |
2617 | ||
2618 | (void) snprintf(od->od_name, sizeof (od->od_name), "%s(%lld)[%llu]", | |
b8864a23 | 2619 | tag, (longlong_t)id, (u_longlong_t)index); |
428870ff BB |
2620 | } |
2621 | ||
2622 | /* | |
2623 | * Lookup or create the objects for a test using the od template. | |
2624 | * If the objects do not all exist, or if 'remove' is specified, | |
2625 | * remove any existing objects and create new ones. Otherwise, | |
2626 | * use the existing objects. | |
2627 | */ | |
2628 | static int | |
2629 | ztest_object_init(ztest_ds_t *zd, ztest_od_t *od, size_t size, boolean_t remove) | |
2630 | { | |
2631 | int count = size / sizeof (*od); | |
2632 | int rv = 0; | |
2633 | ||
1e33ac1e | 2634 | mutex_enter(&zd->zd_dirobj_lock); |
428870ff BB |
2635 | if ((ztest_lookup(zd, od, count) != 0 || remove) && |
2636 | (ztest_remove(zd, od, count) != 0 || | |
2637 | ztest_create(zd, od, count) != 0)) | |
2638 | rv = -1; | |
2639 | zd->zd_od = od; | |
1e33ac1e | 2640 | mutex_exit(&zd->zd_dirobj_lock); |
428870ff BB |
2641 | |
2642 | return (rv); | |
2643 | } | |
2644 | ||
2645 | /* ARGSUSED */ | |
2646 | void | |
2647 | ztest_zil_commit(ztest_ds_t *zd, uint64_t id) | |
2648 | { | |
2649 | zilog_t *zilog = zd->zd_zilog; | |
2650 | ||
7809eb8b | 2651 | (void) rw_rdlock(&zd->zd_zilog_lock); |
3e31d2b0 | 2652 | |
572e2857 | 2653 | zil_commit(zilog, ztest_random(ZTEST_OBJECTS)); |
428870ff BB |
2654 | |
2655 | /* | |
2656 | * Remember the committed values in zd, which is in parent/child | |
2657 | * shared memory. If we die, the next iteration of ztest_run() | |
2658 | * will verify that the log really does contain this record. | |
2659 | */ | |
2660 | mutex_enter(&zilog->zl_lock); | |
c242c188 CS |
2661 | ASSERT(zd->zd_shared != NULL); |
2662 | ASSERT3U(zd->zd_shared->zd_seq, <=, zilog->zl_commit_lr_seq); | |
2663 | zd->zd_shared->zd_seq = zilog->zl_commit_lr_seq; | |
428870ff | 2664 | mutex_exit(&zilog->zl_lock); |
3e31d2b0 | 2665 | |
7809eb8b | 2666 | (void) rw_unlock(&zd->zd_zilog_lock); |
3e31d2b0 ES |
2667 | } |
2668 | ||
2669 | /* | |
2670 | * This function is designed to simulate the operations that occur during a | |
2671 | * mount/unmount operation. We hold the dataset across these operations in an | |
2672 | * attempt to expose any implicit assumptions about ZIL management. | |
2673 | */ | |
2674 | /* ARGSUSED */ | |
2675 | void | |
2676 | ztest_zil_remount(ztest_ds_t *zd, uint64_t id) | |
2677 | { | |
2678 | objset_t *os = zd->zd_os; | |
2679 | ||
03c6040b GW |
2680 | /* |
2681 | * We grab the zd_dirobj_lock to ensure that no other thread is | |
2682 | * updating the zil (i.e. adding in-memory log records) and the | |
2683 | * zd_zilog_lock to block any I/O. | |
2684 | */ | |
29809a6c | 2685 | mutex_enter(&zd->zd_dirobj_lock); |
7809eb8b | 2686 | (void) rw_wrlock(&zd->zd_zilog_lock); |
3e31d2b0 | 2687 | |
f298b24d | 2688 | /* zfsvfs_teardown() */ |
3e31d2b0 ES |
2689 | zil_close(zd->zd_zilog); |
2690 | ||
2691 | /* zfsvfs_setup() */ | |
2692 | VERIFY(zil_open(os, ztest_get_data) == zd->zd_zilog); | |
2693 | zil_replay(os, zd, ztest_replay_vector); | |
2694 | ||
7809eb8b | 2695 | (void) rw_unlock(&zd->zd_zilog_lock); |
29809a6c | 2696 | mutex_exit(&zd->zd_dirobj_lock); |
428870ff BB |
2697 | } |
2698 | ||
2699 | /* | |
2700 | * Verify that we can't destroy an active pool, create an existing pool, | |
2701 | * or create a pool with a bad vdev spec. | |
2702 | */ | |
2703 | /* ARGSUSED */ | |
2704 | void | |
2705 | ztest_spa_create_destroy(ztest_ds_t *zd, uint64_t id) | |
2706 | { | |
c242c188 | 2707 | ztest_shared_opts_t *zo = &ztest_opts; |
428870ff BB |
2708 | spa_t *spa; |
2709 | nvlist_t *nvroot; | |
2710 | ||
379ca9cf OF |
2711 | if (zo->zo_mmp_test) |
2712 | return; | |
2713 | ||
428870ff BB |
2714 | /* |
2715 | * Attempt to create using a bad file. | |
2716 | */ | |
ea0b2538 | 2717 | nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, 0, 0, 0, 1); |
428870ff | 2718 | VERIFY3U(ENOENT, ==, |
b5256303 | 2719 | spa_create("ztest_bad_file", nvroot, NULL, NULL, NULL)); |
428870ff BB |
2720 | nvlist_free(nvroot); |
2721 | ||
2722 | /* | |
2723 | * Attempt to create using a bad mirror. | |
2724 | */ | |
ea0b2538 | 2725 | nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, 0, 0, 2, 1); |
428870ff | 2726 | VERIFY3U(ENOENT, ==, |
b5256303 | 2727 | spa_create("ztest_bad_mirror", nvroot, NULL, NULL, NULL)); |
428870ff BB |
2728 | nvlist_free(nvroot); |
2729 | ||
2730 | /* | |
2731 | * Attempt to create an existing pool. It shouldn't matter | |
2732 | * what's in the nvroot; we should fail with EEXIST. | |
2733 | */ | |
7809eb8b | 2734 | (void) rw_rdlock(&ztest_name_lock); |
ea0b2538 | 2735 | nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, 0, 0, 0, 1); |
b5256303 TC |
2736 | VERIFY3U(EEXIST, ==, |
2737 | spa_create(zo->zo_pool, nvroot, NULL, NULL, NULL)); | |
428870ff | 2738 | nvlist_free(nvroot); |
c242c188 CS |
2739 | VERIFY3U(0, ==, spa_open(zo->zo_pool, &spa, FTAG)); |
2740 | VERIFY3U(EBUSY, ==, spa_destroy(zo->zo_pool)); | |
428870ff BB |
2741 | spa_close(spa, FTAG); |
2742 | ||
7809eb8b | 2743 | (void) rw_unlock(&ztest_name_lock); |
428870ff BB |
2744 | } |
2745 | ||
0582e403 OF |
2746 | /* |
2747 | * Start and then stop the MMP threads to ensure the startup and shutdown code | |
2748 | * works properly. Actual protection and property-related code tested via ZTS. | |
2749 | */ | |
2750 | /* ARGSUSED */ | |
2751 | void | |
2752 | ztest_mmp_enable_disable(ztest_ds_t *zd, uint64_t id) | |
2753 | { | |
2754 | ztest_shared_opts_t *zo = &ztest_opts; | |
2755 | spa_t *spa = ztest_spa; | |
2756 | ||
2757 | if (zo->zo_mmp_test) | |
2758 | return; | |
2759 | ||
2760 | spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); | |
2761 | mutex_enter(&spa->spa_props_lock); | |
2762 | ||
2763 | if (!spa_multihost(spa)) { | |
2764 | spa->spa_multihost = B_TRUE; | |
2765 | mmp_thread_start(spa); | |
2766 | } | |
2767 | ||
2768 | mutex_exit(&spa->spa_props_lock); | |
2769 | spa_config_exit(spa, SCL_CONFIG, FTAG); | |
2770 | ||
2771 | txg_wait_synced(spa_get_dsl(spa), 0); | |
2772 | mmp_signal_all_threads(); | |
2773 | txg_wait_synced(spa_get_dsl(spa), 0); | |
2774 | ||
2775 | spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); | |
2776 | mutex_enter(&spa->spa_props_lock); | |
2777 | ||
2778 | if (spa_multihost(spa)) { | |
2779 | mmp_thread_stop(spa); | |
2780 | spa->spa_multihost = B_FALSE; | |
2781 | } | |
2782 | ||
2783 | mutex_exit(&spa->spa_props_lock); | |
2784 | spa_config_exit(spa, SCL_CONFIG, FTAG); | |
2785 | } | |
2786 | ||
ea0b2538 GW |
2787 | /* ARGSUSED */ |
2788 | void | |
2789 | ztest_spa_upgrade(ztest_ds_t *zd, uint64_t id) | |
2790 | { | |
2791 | spa_t *spa; | |
2792 | uint64_t initial_version = SPA_VERSION_INITIAL; | |
2793 | uint64_t version, newversion; | |
2794 | nvlist_t *nvroot, *props; | |
2795 | char *name; | |
2796 | ||
379ca9cf OF |
2797 | if (ztest_opts.zo_mmp_test) |
2798 | return; | |
2799 | ||
ea0b2538 GW |
2800 | mutex_enter(&ztest_vdev_lock); |
2801 | name = kmem_asprintf("%s_upgrade", ztest_opts.zo_pool); | |
2802 | ||
2803 | /* | |
2804 | * Clean up from previous runs. | |
2805 | */ | |
2806 | (void) spa_destroy(name); | |
2807 | ||
2808 | nvroot = make_vdev_root(NULL, NULL, name, ztest_opts.zo_vdev_size, 0, | |
2809 | 0, ztest_opts.zo_raidz, ztest_opts.zo_mirrors, 1); | |
2810 | ||
2811 | /* | |
2812 | * If we're configuring a RAIDZ device then make sure that the | |
2813 | * the initial version is capable of supporting that feature. | |
2814 | */ | |
2815 | switch (ztest_opts.zo_raidz_parity) { | |
2816 | case 0: | |
2817 | case 1: | |
2818 | initial_version = SPA_VERSION_INITIAL; | |
2819 | break; | |
2820 | case 2: | |
2821 | initial_version = SPA_VERSION_RAIDZ2; | |
2822 | break; | |
2823 | case 3: | |
2824 | initial_version = SPA_VERSION_RAIDZ3; | |
2825 | break; | |
2826 | } | |
2827 | ||
2828 | /* | |
2829 | * Create a pool with a spa version that can be upgraded. Pick | |
2830 | * a value between initial_version and SPA_VERSION_BEFORE_FEATURES. | |
2831 | */ | |
2832 | do { | |
2833 | version = ztest_random_spa_version(initial_version); | |
2834 | } while (version > SPA_VERSION_BEFORE_FEATURES); | |
2835 | ||
2836 | props = fnvlist_alloc(); | |
2837 | fnvlist_add_uint64(props, | |
2838 | zpool_prop_to_name(ZPOOL_PROP_VERSION), version); | |
b5256303 | 2839 | VERIFY3S(spa_create(name, nvroot, props, NULL, NULL), ==, 0); |
ea0b2538 GW |
2840 | fnvlist_free(nvroot); |
2841 | fnvlist_free(props); | |
2842 | ||
2843 | VERIFY3S(spa_open(name, &spa, FTAG), ==, 0); | |
2844 | VERIFY3U(spa_version(spa), ==, version); | |
2845 | newversion = ztest_random_spa_version(version + 1); | |
2846 | ||
2847 | if (ztest_opts.zo_verbose >= 4) { | |
2848 | (void) printf("upgrading spa version from %llu to %llu\n", | |
2849 | (u_longlong_t)version, (u_longlong_t)newversion); | |
2850 | } | |
2851 | ||
2852 | spa_upgrade(spa, newversion); | |
2853 | VERIFY3U(spa_version(spa), >, version); | |
2854 | VERIFY3U(spa_version(spa), ==, fnvlist_lookup_uint64(spa->spa_config, | |
2855 | zpool_prop_to_name(ZPOOL_PROP_VERSION))); | |
2856 | spa_close(spa, FTAG); | |
2857 | ||
2858 | strfree(name); | |
2859 | mutex_exit(&ztest_vdev_lock); | |
2860 | } | |
2861 | ||
428870ff BB |
2862 | static vdev_t * |
2863 | vdev_lookup_by_path(vdev_t *vd, const char *path) | |
2864 | { | |
2865 | vdev_t *mvd; | |
d6320ddb | 2866 | int c; |
428870ff BB |
2867 | |
2868 | if (vd->vdev_path != NULL && strcmp(path, vd->vdev_path) == 0) | |
2869 | return (vd); | |
2870 | ||
d6320ddb | 2871 | for (c = 0; c < vd->vdev_children; c++) |
428870ff BB |
2872 | if ((mvd = vdev_lookup_by_path(vd->vdev_child[c], path)) != |
2873 | NULL) | |
2874 | return (mvd); | |
2875 | ||
2876 | return (NULL); | |
2877 | } | |
2878 | ||
2879 | /* | |
2880 | * Find the first available hole which can be used as a top-level. | |
2881 | */ | |
2882 | int | |
2883 | find_vdev_hole(spa_t *spa) | |
2884 | { | |
2885 | vdev_t *rvd = spa->spa_root_vdev; | |
2886 | int c; | |
2887 | ||
2888 | ASSERT(spa_config_held(spa, SCL_VDEV, RW_READER) == SCL_VDEV); | |
2889 | ||
2890 | for (c = 0; c < rvd->vdev_children; c++) { | |
2891 | vdev_t *cvd = rvd->vdev_child[c]; | |
2892 | ||
2893 | if (cvd->vdev_ishole) | |
2894 | break; | |
2895 | } | |
2896 | return (c); | |
2897 | } | |
2898 | ||
2899 | /* | |
2900 | * Verify that vdev_add() works as expected. | |
2901 | */ | |
2902 | /* ARGSUSED */ | |
2903 | void | |
2904 | ztest_vdev_add_remove(ztest_ds_t *zd, uint64_t id) | |
2905 | { | |
2906 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 2907 | spa_t *spa = ztest_spa; |
428870ff BB |
2908 | uint64_t leaves; |
2909 | uint64_t guid; | |
2910 | nvlist_t *nvroot; | |
2911 | int error; | |
2912 | ||
379ca9cf OF |
2913 | if (ztest_opts.zo_mmp_test) |
2914 | return; | |
2915 | ||
c242c188 | 2916 | mutex_enter(&ztest_vdev_lock); |
13fe0198 | 2917 | leaves = MAX(zs->zs_mirrors + zs->zs_splits, 1) * ztest_opts.zo_raidz; |
428870ff BB |
2918 | |
2919 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
2920 | ||
2921 | ztest_shared->zs_vdev_next_leaf = find_vdev_hole(spa) * leaves; | |
2922 | ||
2923 | /* | |
2924 | * If we have slogs then remove them 1/4 of the time. | |
2925 | */ | |
2926 | if (spa_has_slogs(spa) && ztest_random(4) == 0) { | |
2927 | /* | |
2928 | * Grab the guid from the head of the log class rotor. | |
2929 | */ | |
2930 | guid = spa_log_class(spa)->mc_rotor->mg_vd->vdev_guid; | |
2931 | ||
2932 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
2933 | ||
2934 | /* | |
2935 | * We have to grab the zs_name_lock as writer to | |
2936 | * prevent a race between removing a slog (dmu_objset_find) | |
2937 | * and destroying a dataset. Removing the slog will | |
2938 | * grab a reference on the dataset which may cause | |
13fe0198 | 2939 | * dsl_destroy_head() to fail with EBUSY thus |
428870ff BB |
2940 | * leaving the dataset in an inconsistent state. |
2941 | */ | |
7809eb8b | 2942 | rw_wrlock(&ztest_name_lock); |
428870ff | 2943 | error = spa_vdev_remove(spa, guid, B_FALSE); |
7809eb8b | 2944 | rw_unlock(&ztest_name_lock); |
428870ff BB |
2945 | |
2946 | if (error && error != EEXIST) | |
2947 | fatal(0, "spa_vdev_remove() = %d", error); | |
2948 | } else { | |
2949 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
2950 | ||
2951 | /* | |
2952 | * Make 1/4 of the devices be log devices. | |
2953 | */ | |
ea0b2538 | 2954 | nvroot = make_vdev_root(NULL, NULL, NULL, |
c242c188 CS |
2955 | ztest_opts.zo_vdev_size, 0, |
2956 | ztest_random(4) == 0, ztest_opts.zo_raidz, | |
2957 | zs->zs_mirrors, 1); | |
428870ff BB |
2958 | |
2959 | error = spa_vdev_add(spa, nvroot); | |
2960 | nvlist_free(nvroot); | |
2961 | ||
2962 | if (error == ENOSPC) | |
2963 | ztest_record_enospc("spa_vdev_add"); | |
2964 | else if (error != 0) | |
2965 | fatal(0, "spa_vdev_add() = %d", error); | |
2966 | } | |
2967 | ||
c242c188 | 2968 | mutex_exit(&ztest_vdev_lock); |
428870ff BB |
2969 | } |
2970 | ||
2971 | /* | |
2972 | * Verify that adding/removing aux devices (l2arc, hot spare) works as expected. | |
2973 | */ | |
2974 | /* ARGSUSED */ | |
2975 | void | |
2976 | ztest_vdev_aux_add_remove(ztest_ds_t *zd, uint64_t id) | |
2977 | { | |
2978 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 2979 | spa_t *spa = ztest_spa; |
428870ff BB |
2980 | vdev_t *rvd = spa->spa_root_vdev; |
2981 | spa_aux_vdev_t *sav; | |
2982 | char *aux; | |
40b84e7a | 2983 | char *path; |
428870ff BB |
2984 | uint64_t guid = 0; |
2985 | int error; | |
2986 | ||
379ca9cf OF |
2987 | if (ztest_opts.zo_mmp_test) |
2988 | return; | |
2989 | ||
40b84e7a BB |
2990 | path = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
2991 | ||
428870ff BB |
2992 | if (ztest_random(2) == 0) { |
2993 | sav = &spa->spa_spares; | |
2994 | aux = ZPOOL_CONFIG_SPARES; | |
2995 | } else { | |
2996 | sav = &spa->spa_l2cache; | |
2997 | aux = ZPOOL_CONFIG_L2CACHE; | |
2998 | } | |
2999 | ||
c242c188 | 3000 | mutex_enter(&ztest_vdev_lock); |
428870ff BB |
3001 | |
3002 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
3003 | ||
3004 | if (sav->sav_count != 0 && ztest_random(4) == 0) { | |
b128c09f BB |
3005 | /* |
3006 | * Pick a random device to remove. | |
3007 | */ | |
3008 | guid = sav->sav_vdevs[ztest_random(sav->sav_count)]->vdev_guid; | |
3009 | } else { | |
3010 | /* | |
3011 | * Find an unused device we can add. | |
3012 | */ | |
428870ff | 3013 | zs->zs_vdev_aux = 0; |
b128c09f | 3014 | for (;;) { |
b128c09f | 3015 | int c; |
3db3ff4a | 3016 | (void) snprintf(path, MAXPATHLEN, ztest_aux_template, |
c242c188 CS |
3017 | ztest_opts.zo_dir, ztest_opts.zo_pool, aux, |
3018 | zs->zs_vdev_aux); | |
b128c09f BB |
3019 | for (c = 0; c < sav->sav_count; c++) |
3020 | if (strcmp(sav->sav_vdevs[c]->vdev_path, | |
3021 | path) == 0) | |
3022 | break; | |
3023 | if (c == sav->sav_count && | |
3024 | vdev_lookup_by_path(rvd, path) == NULL) | |
3025 | break; | |
428870ff | 3026 | zs->zs_vdev_aux++; |
34dc7c2f BB |
3027 | } |
3028 | } | |
3029 | ||
b128c09f | 3030 | spa_config_exit(spa, SCL_VDEV, FTAG); |
34dc7c2f | 3031 | |
b128c09f BB |
3032 | if (guid == 0) { |
3033 | /* | |
3034 | * Add a new device. | |
3035 | */ | |
ea0b2538 | 3036 | nvlist_t *nvroot = make_vdev_root(NULL, aux, NULL, |
c242c188 | 3037 | (ztest_opts.zo_vdev_size * 5) / 4, 0, 0, 0, 0, 1); |
b128c09f BB |
3038 | error = spa_vdev_add(spa, nvroot); |
3039 | if (error != 0) | |
3040 | fatal(0, "spa_vdev_add(%p) = %d", nvroot, error); | |
3041 | nvlist_free(nvroot); | |
3042 | } else { | |
3043 | /* | |
3044 | * Remove an existing device. Sometimes, dirty its | |
3045 | * vdev state first to make sure we handle removal | |
3046 | * of devices that have pending state changes. | |
3047 | */ | |
3048 | if (ztest_random(2) == 0) | |
9babb374 | 3049 | (void) vdev_online(spa, guid, 0, NULL); |
b128c09f BB |
3050 | |
3051 | error = spa_vdev_remove(spa, guid, B_FALSE); | |
3052 | if (error != 0 && error != EBUSY) | |
3053 | fatal(0, "spa_vdev_remove(%llu) = %d", guid, error); | |
3054 | } | |
3055 | ||
c242c188 | 3056 | mutex_exit(&ztest_vdev_lock); |
40b84e7a BB |
3057 | |
3058 | umem_free(path, MAXPATHLEN); | |
428870ff BB |
3059 | } |
3060 | ||
3061 | /* | |
3062 | * split a pool if it has mirror tlvdevs | |
3063 | */ | |
3064 | /* ARGSUSED */ | |
3065 | void | |
3066 | ztest_split_pool(ztest_ds_t *zd, uint64_t id) | |
3067 | { | |
3068 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 3069 | spa_t *spa = ztest_spa; |
428870ff BB |
3070 | vdev_t *rvd = spa->spa_root_vdev; |
3071 | nvlist_t *tree, **child, *config, *split, **schild; | |
3072 | uint_t c, children, schildren = 0, lastlogid = 0; | |
3073 | int error = 0; | |
3074 | ||
379ca9cf OF |
3075 | if (ztest_opts.zo_mmp_test) |
3076 | return; | |
3077 | ||
c242c188 | 3078 | mutex_enter(&ztest_vdev_lock); |
428870ff BB |
3079 | |
3080 | /* ensure we have a useable config; mirrors of raidz aren't supported */ | |
c242c188 CS |
3081 | if (zs->zs_mirrors < 3 || ztest_opts.zo_raidz > 1) { |
3082 | mutex_exit(&ztest_vdev_lock); | |
428870ff BB |
3083 | return; |
3084 | } | |
3085 | ||
3086 | /* clean up the old pool, if any */ | |
3087 | (void) spa_destroy("splitp"); | |
3088 | ||
3089 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
3090 | ||
3091 | /* generate a config from the existing config */ | |
3092 | mutex_enter(&spa->spa_props_lock); | |
3093 | VERIFY(nvlist_lookup_nvlist(spa->spa_config, ZPOOL_CONFIG_VDEV_TREE, | |
3094 | &tree) == 0); | |
3095 | mutex_exit(&spa->spa_props_lock); | |
3096 | ||
3097 | VERIFY(nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child, | |
3098 | &children) == 0); | |
3099 | ||
3100 | schild = malloc(rvd->vdev_children * sizeof (nvlist_t *)); | |
3101 | for (c = 0; c < children; c++) { | |
3102 | vdev_t *tvd = rvd->vdev_child[c]; | |
3103 | nvlist_t **mchild; | |
3104 | uint_t mchildren; | |
3105 | ||
3106 | if (tvd->vdev_islog || tvd->vdev_ops == &vdev_hole_ops) { | |
3107 | VERIFY(nvlist_alloc(&schild[schildren], NV_UNIQUE_NAME, | |
3108 | 0) == 0); | |
3109 | VERIFY(nvlist_add_string(schild[schildren], | |
3110 | ZPOOL_CONFIG_TYPE, VDEV_TYPE_HOLE) == 0); | |
3111 | VERIFY(nvlist_add_uint64(schild[schildren], | |
3112 | ZPOOL_CONFIG_IS_HOLE, 1) == 0); | |
3113 | if (lastlogid == 0) | |
3114 | lastlogid = schildren; | |
3115 | ++schildren; | |
3116 | continue; | |
3117 | } | |
3118 | lastlogid = 0; | |
3119 | VERIFY(nvlist_lookup_nvlist_array(child[c], | |
3120 | ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0); | |
3121 | VERIFY(nvlist_dup(mchild[0], &schild[schildren++], 0) == 0); | |
3122 | } | |
3123 | ||
3124 | /* OK, create a config that can be used to split */ | |
3125 | VERIFY(nvlist_alloc(&split, NV_UNIQUE_NAME, 0) == 0); | |
3126 | VERIFY(nvlist_add_string(split, ZPOOL_CONFIG_TYPE, | |
3127 | VDEV_TYPE_ROOT) == 0); | |
3128 | VERIFY(nvlist_add_nvlist_array(split, ZPOOL_CONFIG_CHILDREN, schild, | |
3129 | lastlogid != 0 ? lastlogid : schildren) == 0); | |
3130 | ||
3131 | VERIFY(nvlist_alloc(&config, NV_UNIQUE_NAME, 0) == 0); | |
3132 | VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, split) == 0); | |
3133 | ||
3134 | for (c = 0; c < schildren; c++) | |
3135 | nvlist_free(schild[c]); | |
3136 | free(schild); | |
3137 | nvlist_free(split); | |
3138 | ||
3139 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
3140 | ||
7809eb8b | 3141 | (void) rw_wrlock(&ztest_name_lock); |
428870ff | 3142 | error = spa_vdev_split_mirror(spa, "splitp", config, NULL, B_FALSE); |
7809eb8b | 3143 | (void) rw_unlock(&ztest_name_lock); |
428870ff BB |
3144 | |
3145 | nvlist_free(config); | |
3146 | ||
3147 | if (error == 0) { | |
3148 | (void) printf("successful split - results:\n"); | |
3149 | mutex_enter(&spa_namespace_lock); | |
3150 | show_pool_stats(spa); | |
3151 | show_pool_stats(spa_lookup("splitp")); | |
3152 | mutex_exit(&spa_namespace_lock); | |
3153 | ++zs->zs_splits; | |
3154 | --zs->zs_mirrors; | |
3155 | } | |
c242c188 | 3156 | mutex_exit(&ztest_vdev_lock); |
428870ff | 3157 | |
34dc7c2f BB |
3158 | } |
3159 | ||
3160 | /* | |
3161 | * Verify that we can attach and detach devices. | |
3162 | */ | |
428870ff | 3163 | /* ARGSUSED */ |
34dc7c2f | 3164 | void |
428870ff | 3165 | ztest_vdev_attach_detach(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3166 | { |
428870ff | 3167 | ztest_shared_t *zs = ztest_shared; |
c242c188 | 3168 | spa_t *spa = ztest_spa; |
b128c09f | 3169 | spa_aux_vdev_t *sav = &spa->spa_spares; |
34dc7c2f BB |
3170 | vdev_t *rvd = spa->spa_root_vdev; |
3171 | vdev_t *oldvd, *newvd, *pvd; | |
b128c09f | 3172 | nvlist_t *root; |
428870ff | 3173 | uint64_t leaves; |
34dc7c2f BB |
3174 | uint64_t leaf, top; |
3175 | uint64_t ashift = ztest_get_ashift(); | |
fb5f0bc8 | 3176 | uint64_t oldguid, pguid; |
5d1f7fb6 | 3177 | uint64_t oldsize, newsize; |
40b84e7a | 3178 | char *oldpath, *newpath; |
34dc7c2f | 3179 | int replacing; |
b128c09f BB |
3180 | int oldvd_has_siblings = B_FALSE; |
3181 | int newvd_is_spare = B_FALSE; | |
3182 | int oldvd_is_log; | |
34dc7c2f | 3183 | int error, expected_error; |
34dc7c2f | 3184 | |
379ca9cf OF |
3185 | if (ztest_opts.zo_mmp_test) |
3186 | return; | |
3187 | ||
40b84e7a BB |
3188 | oldpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
3189 | newpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
3190 | ||
c242c188 CS |
3191 | mutex_enter(&ztest_vdev_lock); |
3192 | leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raidz; | |
34dc7c2f | 3193 | |
b128c09f | 3194 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); |
34dc7c2f BB |
3195 | |
3196 | /* | |
3197 | * Decide whether to do an attach or a replace. | |
3198 | */ | |
3199 | replacing = ztest_random(2); | |
3200 | ||
3201 | /* | |
3202 | * Pick a random top-level vdev. | |
3203 | */ | |
428870ff | 3204 | top = ztest_random_vdev_top(spa, B_TRUE); |
34dc7c2f BB |
3205 | |
3206 | /* | |
3207 | * Pick a random leaf within it. | |
3208 | */ | |
3209 | leaf = ztest_random(leaves); | |
3210 | ||
3211 | /* | |
b128c09f | 3212 | * Locate this vdev. |
34dc7c2f | 3213 | */ |
b128c09f | 3214 | oldvd = rvd->vdev_child[top]; |
428870ff | 3215 | if (zs->zs_mirrors >= 1) { |
fb5f0bc8 | 3216 | ASSERT(oldvd->vdev_ops == &vdev_mirror_ops); |
428870ff | 3217 | ASSERT(oldvd->vdev_children >= zs->zs_mirrors); |
c242c188 | 3218 | oldvd = oldvd->vdev_child[leaf / ztest_opts.zo_raidz]; |
fb5f0bc8 | 3219 | } |
c242c188 | 3220 | if (ztest_opts.zo_raidz > 1) { |
fb5f0bc8 | 3221 | ASSERT(oldvd->vdev_ops == &vdev_raidz_ops); |
c242c188 CS |
3222 | ASSERT(oldvd->vdev_children == ztest_opts.zo_raidz); |
3223 | oldvd = oldvd->vdev_child[leaf % ztest_opts.zo_raidz]; | |
fb5f0bc8 | 3224 | } |
34dc7c2f BB |
3225 | |
3226 | /* | |
b128c09f BB |
3227 | * If we're already doing an attach or replace, oldvd may be a |
3228 | * mirror vdev -- in which case, pick a random child. | |
34dc7c2f | 3229 | */ |
b128c09f BB |
3230 | while (oldvd->vdev_children != 0) { |
3231 | oldvd_has_siblings = B_TRUE; | |
fb5f0bc8 BB |
3232 | ASSERT(oldvd->vdev_children >= 2); |
3233 | oldvd = oldvd->vdev_child[ztest_random(oldvd->vdev_children)]; | |
b128c09f BB |
3234 | } |
3235 | ||
3236 | oldguid = oldvd->vdev_guid; | |
9babb374 | 3237 | oldsize = vdev_get_min_asize(oldvd); |
b128c09f BB |
3238 | oldvd_is_log = oldvd->vdev_top->vdev_islog; |
3239 | (void) strcpy(oldpath, oldvd->vdev_path); | |
3240 | pvd = oldvd->vdev_parent; | |
fb5f0bc8 | 3241 | pguid = pvd->vdev_guid; |
34dc7c2f BB |
3242 | |
3243 | /* | |
b128c09f | 3244 | * If oldvd has siblings, then half of the time, detach it. |
34dc7c2f | 3245 | */ |
b128c09f BB |
3246 | if (oldvd_has_siblings && ztest_random(2) == 0) { |
3247 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
fb5f0bc8 BB |
3248 | error = spa_vdev_detach(spa, oldguid, pguid, B_FALSE); |
3249 | if (error != 0 && error != ENODEV && error != EBUSY && | |
3250 | error != ENOTSUP) | |
3251 | fatal(0, "detach (%s) returned %d", oldpath, error); | |
40b84e7a | 3252 | goto out; |
b128c09f | 3253 | } |
34dc7c2f BB |
3254 | |
3255 | /* | |
b128c09f BB |
3256 | * For the new vdev, choose with equal probability between the two |
3257 | * standard paths (ending in either 'a' or 'b') or a random hot spare. | |
34dc7c2f | 3258 | */ |
b128c09f BB |
3259 | if (sav->sav_count != 0 && ztest_random(3) == 0) { |
3260 | newvd = sav->sav_vdevs[ztest_random(sav->sav_count)]; | |
3261 | newvd_is_spare = B_TRUE; | |
3262 | (void) strcpy(newpath, newvd->vdev_path); | |
3263 | } else { | |
6aec1cd5 | 3264 | (void) snprintf(newpath, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
3265 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
3266 | top * leaves + leaf); | |
b128c09f BB |
3267 | if (ztest_random(2) == 0) |
3268 | newpath[strlen(newpath) - 1] = 'b'; | |
3269 | newvd = vdev_lookup_by_path(rvd, newpath); | |
3270 | } | |
3271 | ||
3272 | if (newvd) { | |
9babb374 | 3273 | newsize = vdev_get_min_asize(newvd); |
b128c09f BB |
3274 | } else { |
3275 | /* | |
3276 | * Make newsize a little bigger or smaller than oldsize. | |
3277 | * If it's smaller, the attach should fail. | |
3278 | * If it's larger, and we're doing a replace, | |
3279 | * we should get dynamic LUN growth when we're done. | |
3280 | */ | |
3281 | newsize = 10 * oldsize / (9 + ztest_random(3)); | |
3282 | } | |
34dc7c2f BB |
3283 | |
3284 | /* | |
3285 | * If pvd is not a mirror or root, the attach should fail with ENOTSUP, | |
3286 | * unless it's a replace; in that case any non-replacing parent is OK. | |
3287 | * | |
3288 | * If newvd is already part of the pool, it should fail with EBUSY. | |
3289 | * | |
3290 | * If newvd is too small, it should fail with EOVERFLOW. | |
3291 | */ | |
b128c09f BB |
3292 | if (pvd->vdev_ops != &vdev_mirror_ops && |
3293 | pvd->vdev_ops != &vdev_root_ops && (!replacing || | |
3294 | pvd->vdev_ops == &vdev_replacing_ops || | |
3295 | pvd->vdev_ops == &vdev_spare_ops)) | |
34dc7c2f | 3296 | expected_error = ENOTSUP; |
b128c09f BB |
3297 | else if (newvd_is_spare && (!replacing || oldvd_is_log)) |
3298 | expected_error = ENOTSUP; | |
3299 | else if (newvd == oldvd) | |
3300 | expected_error = replacing ? 0 : EBUSY; | |
3301 | else if (vdev_lookup_by_path(rvd, newpath) != NULL) | |
3302 | expected_error = EBUSY; | |
34dc7c2f BB |
3303 | else if (newsize < oldsize) |
3304 | expected_error = EOVERFLOW; | |
3305 | else if (ashift > oldvd->vdev_top->vdev_ashift) | |
3306 | expected_error = EDOM; | |
3307 | else | |
3308 | expected_error = 0; | |
3309 | ||
b128c09f | 3310 | spa_config_exit(spa, SCL_VDEV, FTAG); |
34dc7c2f BB |
3311 | |
3312 | /* | |
3313 | * Build the nvlist describing newpath. | |
3314 | */ | |
ea0b2538 | 3315 | root = make_vdev_root(newpath, NULL, NULL, newvd == NULL ? newsize : 0, |
b128c09f | 3316 | ashift, 0, 0, 0, 1); |
34dc7c2f | 3317 | |
b128c09f | 3318 | error = spa_vdev_attach(spa, oldguid, root, replacing); |
34dc7c2f | 3319 | |
34dc7c2f BB |
3320 | nvlist_free(root); |
3321 | ||
3322 | /* | |
3323 | * If our parent was the replacing vdev, but the replace completed, | |
3324 | * then instead of failing with ENOTSUP we may either succeed, | |
3325 | * fail with ENODEV, or fail with EOVERFLOW. | |
3326 | */ | |
3327 | if (expected_error == ENOTSUP && | |
3328 | (error == 0 || error == ENODEV || error == EOVERFLOW)) | |
3329 | expected_error = error; | |
3330 | ||
3331 | /* | |
3332 | * If someone grew the LUN, the replacement may be too small. | |
3333 | */ | |
b128c09f | 3334 | if (error == EOVERFLOW || error == EBUSY) |
34dc7c2f BB |
3335 | expected_error = error; |
3336 | ||
b128c09f BB |
3337 | /* XXX workaround 6690467 */ |
3338 | if (error != expected_error && expected_error != EBUSY) { | |
3339 | fatal(0, "attach (%s %llu, %s %llu, %d) " | |
3340 | "returned %d, expected %d", | |
5d1f7fb6 GW |
3341 | oldpath, oldsize, newpath, |
3342 | newsize, replacing, error, expected_error); | |
34dc7c2f | 3343 | } |
40b84e7a | 3344 | out: |
c242c188 | 3345 | mutex_exit(&ztest_vdev_lock); |
40b84e7a BB |
3346 | |
3347 | umem_free(oldpath, MAXPATHLEN); | |
3348 | umem_free(newpath, MAXPATHLEN); | |
34dc7c2f BB |
3349 | } |
3350 | ||
9babb374 BB |
3351 | /* |
3352 | * Callback function which expands the physical size of the vdev. | |
3353 | */ | |
3354 | vdev_t * | |
3355 | grow_vdev(vdev_t *vd, void *arg) | |
3356 | { | |
1fde1e37 | 3357 | ASSERTV(spa_t *spa = vd->vdev_spa); |
9babb374 BB |
3358 | size_t *newsize = arg; |
3359 | size_t fsize; | |
3360 | int fd; | |
3361 | ||
3362 | ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE); | |
3363 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
3364 | ||
3365 | if ((fd = open(vd->vdev_path, O_RDWR)) == -1) | |
3366 | return (vd); | |
3367 | ||
3368 | fsize = lseek(fd, 0, SEEK_END); | |
0e5b68e0 | 3369 | VERIFY(ftruncate(fd, *newsize) == 0); |
9babb374 | 3370 | |
c242c188 | 3371 | if (ztest_opts.zo_verbose >= 6) { |
9babb374 BB |
3372 | (void) printf("%s grew from %lu to %lu bytes\n", |
3373 | vd->vdev_path, (ulong_t)fsize, (ulong_t)*newsize); | |
3374 | } | |
3375 | (void) close(fd); | |
3376 | return (NULL); | |
3377 | } | |
3378 | ||
3379 | /* | |
3380 | * Callback function which expands a given vdev by calling vdev_online(). | |
3381 | */ | |
3382 | /* ARGSUSED */ | |
3383 | vdev_t * | |
3384 | online_vdev(vdev_t *vd, void *arg) | |
3385 | { | |
3386 | spa_t *spa = vd->vdev_spa; | |
3387 | vdev_t *tvd = vd->vdev_top; | |
9babb374 | 3388 | uint64_t guid = vd->vdev_guid; |
428870ff BB |
3389 | uint64_t generation = spa->spa_config_generation + 1; |
3390 | vdev_state_t newstate = VDEV_STATE_UNKNOWN; | |
3391 | int error; | |
9babb374 BB |
3392 | |
3393 | ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE); | |
3394 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
3395 | ||
3396 | /* Calling vdev_online will initialize the new metaslabs */ | |
3397 | spa_config_exit(spa, SCL_STATE, spa); | |
428870ff | 3398 | error = vdev_online(spa, guid, ZFS_ONLINE_EXPAND, &newstate); |
9babb374 BB |
3399 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); |
3400 | ||
428870ff BB |
3401 | /* |
3402 | * If vdev_online returned an error or the underlying vdev_open | |
3403 | * failed then we abort the expand. The only way to know that | |
3404 | * vdev_open fails is by checking the returned newstate. | |
3405 | */ | |
3406 | if (error || newstate != VDEV_STATE_HEALTHY) { | |
c242c188 | 3407 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
3408 | (void) printf("Unable to expand vdev, state %llu, " |
3409 | "error %d\n", (u_longlong_t)newstate, error); | |
3410 | } | |
3411 | return (vd); | |
3412 | } | |
3413 | ASSERT3U(newstate, ==, VDEV_STATE_HEALTHY); | |
3414 | ||
9babb374 BB |
3415 | /* |
3416 | * Since we dropped the lock we need to ensure that we're | |
3417 | * still talking to the original vdev. It's possible this | |
3418 | * vdev may have been detached/replaced while we were | |
3419 | * trying to online it. | |
3420 | */ | |
428870ff | 3421 | if (generation != spa->spa_config_generation) { |
c242c188 | 3422 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
3423 | (void) printf("vdev configuration has changed, " |
3424 | "guid %llu, state %llu, expected gen %llu, " | |
3425 | "got gen %llu\n", | |
3426 | (u_longlong_t)guid, | |
3427 | (u_longlong_t)tvd->vdev_state, | |
3428 | (u_longlong_t)generation, | |
3429 | (u_longlong_t)spa->spa_config_generation); | |
9babb374 BB |
3430 | } |
3431 | return (vd); | |
3432 | } | |
3433 | return (NULL); | |
3434 | } | |
3435 | ||
3436 | /* | |
3437 | * Traverse the vdev tree calling the supplied function. | |
3438 | * We continue to walk the tree until we either have walked all | |
3439 | * children or we receive a non-NULL return from the callback. | |
3440 | * If a NULL callback is passed, then we just return back the first | |
3441 | * leaf vdev we encounter. | |
3442 | */ | |
3443 | vdev_t * | |
3444 | vdev_walk_tree(vdev_t *vd, vdev_t *(*func)(vdev_t *, void *), void *arg) | |
3445 | { | |
d6320ddb BB |
3446 | uint_t c; |
3447 | ||
9babb374 BB |
3448 | if (vd->vdev_ops->vdev_op_leaf) { |
3449 | if (func == NULL) | |
3450 | return (vd); | |
3451 | else | |
3452 | return (func(vd, arg)); | |
3453 | } | |
3454 | ||
d6320ddb | 3455 | for (c = 0; c < vd->vdev_children; c++) { |
9babb374 BB |
3456 | vdev_t *cvd = vd->vdev_child[c]; |
3457 | if ((cvd = vdev_walk_tree(cvd, func, arg)) != NULL) | |
3458 | return (cvd); | |
3459 | } | |
3460 | return (NULL); | |
3461 | } | |
3462 | ||
34dc7c2f BB |
3463 | /* |
3464 | * Verify that dynamic LUN growth works as expected. | |
3465 | */ | |
428870ff | 3466 | /* ARGSUSED */ |
34dc7c2f | 3467 | void |
428870ff | 3468 | ztest_vdev_LUN_growth(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3469 | { |
c242c188 | 3470 | spa_t *spa = ztest_spa; |
428870ff BB |
3471 | vdev_t *vd, *tvd; |
3472 | metaslab_class_t *mc; | |
3473 | metaslab_group_t *mg; | |
9babb374 | 3474 | size_t psize, newsize; |
428870ff BB |
3475 | uint64_t top; |
3476 | uint64_t old_class_space, new_class_space, old_ms_count, new_ms_count; | |
34dc7c2f | 3477 | |
c242c188 | 3478 | mutex_enter(&ztest_vdev_lock); |
9babb374 BB |
3479 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); |
3480 | ||
428870ff | 3481 | top = ztest_random_vdev_top(spa, B_TRUE); |
9babb374 | 3482 | |
428870ff BB |
3483 | tvd = spa->spa_root_vdev->vdev_child[top]; |
3484 | mg = tvd->vdev_mg; | |
3485 | mc = mg->mg_class; | |
3486 | old_ms_count = tvd->vdev_ms_count; | |
3487 | old_class_space = metaslab_class_get_space(mc); | |
34dc7c2f BB |
3488 | |
3489 | /* | |
9babb374 BB |
3490 | * Determine the size of the first leaf vdev associated with |
3491 | * our top-level device. | |
34dc7c2f | 3492 | */ |
9babb374 BB |
3493 | vd = vdev_walk_tree(tvd, NULL, NULL); |
3494 | ASSERT3P(vd, !=, NULL); | |
3495 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
34dc7c2f | 3496 | |
9babb374 | 3497 | psize = vd->vdev_psize; |
34dc7c2f | 3498 | |
9babb374 | 3499 | /* |
428870ff BB |
3500 | * We only try to expand the vdev if it's healthy, less than 4x its |
3501 | * original size, and it has a valid psize. | |
9babb374 | 3502 | */ |
428870ff | 3503 | if (tvd->vdev_state != VDEV_STATE_HEALTHY || |
c242c188 | 3504 | psize == 0 || psize >= 4 * ztest_opts.zo_vdev_size) { |
9babb374 | 3505 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 3506 | mutex_exit(&ztest_vdev_lock); |
9babb374 BB |
3507 | return; |
3508 | } | |
3509 | ASSERT(psize > 0); | |
3510 | newsize = psize + psize / 8; | |
3511 | ASSERT3U(newsize, >, psize); | |
34dc7c2f | 3512 | |
c242c188 | 3513 | if (ztest_opts.zo_verbose >= 6) { |
428870ff | 3514 | (void) printf("Expanding LUN %s from %lu to %lu\n", |
9babb374 BB |
3515 | vd->vdev_path, (ulong_t)psize, (ulong_t)newsize); |
3516 | } | |
3517 | ||
9babb374 BB |
3518 | /* |
3519 | * Growing the vdev is a two step process: | |
3520 | * 1). expand the physical size (i.e. relabel) | |
3521 | * 2). online the vdev to create the new metaslabs | |
3522 | */ | |
3523 | if (vdev_walk_tree(tvd, grow_vdev, &newsize) != NULL || | |
3524 | vdev_walk_tree(tvd, online_vdev, NULL) != NULL || | |
3525 | tvd->vdev_state != VDEV_STATE_HEALTHY) { | |
c242c188 | 3526 | if (ztest_opts.zo_verbose >= 5) { |
9babb374 | 3527 | (void) printf("Could not expand LUN because " |
428870ff | 3528 | "the vdev configuration changed.\n"); |
34dc7c2f | 3529 | } |
428870ff | 3530 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 3531 | mutex_exit(&ztest_vdev_lock); |
9babb374 | 3532 | return; |
34dc7c2f BB |
3533 | } |
3534 | ||
428870ff | 3535 | spa_config_exit(spa, SCL_STATE, spa); |
9babb374 BB |
3536 | |
3537 | /* | |
3538 | * Expanding the LUN will update the config asynchronously, | |
3539 | * thus we must wait for the async thread to complete any | |
3540 | * pending tasks before proceeding. | |
3541 | */ | |
428870ff BB |
3542 | for (;;) { |
3543 | boolean_t done; | |
3544 | mutex_enter(&spa->spa_async_lock); | |
3545 | done = (spa->spa_async_thread == NULL && !spa->spa_async_tasks); | |
3546 | mutex_exit(&spa->spa_async_lock); | |
3547 | if (done) | |
3548 | break; | |
3549 | txg_wait_synced(spa_get_dsl(spa), 0); | |
3550 | (void) poll(NULL, 0, 100); | |
3551 | } | |
9babb374 BB |
3552 | |
3553 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); | |
428870ff BB |
3554 | |
3555 | tvd = spa->spa_root_vdev->vdev_child[top]; | |
3556 | new_ms_count = tvd->vdev_ms_count; | |
3557 | new_class_space = metaslab_class_get_space(mc); | |
3558 | ||
3559 | if (tvd->vdev_mg != mg || mg->mg_class != mc) { | |
c242c188 | 3560 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
3561 | (void) printf("Could not verify LUN expansion due to " |
3562 | "intervening vdev offline or remove.\n"); | |
3563 | } | |
3564 | spa_config_exit(spa, SCL_STATE, spa); | |
c242c188 | 3565 | mutex_exit(&ztest_vdev_lock); |
428870ff BB |
3566 | return; |
3567 | } | |
3568 | ||
3569 | /* | |
3570 | * Make sure we were able to grow the vdev. | |
3571 | */ | |
3572 | if (new_ms_count <= old_ms_count) | |
3573 | fatal(0, "LUN expansion failed: ms_count %llu <= %llu\n", | |
3574 | old_ms_count, new_ms_count); | |
9babb374 BB |
3575 | |
3576 | /* | |
3577 | * Make sure we were able to grow the pool. | |
3578 | */ | |
428870ff BB |
3579 | if (new_class_space <= old_class_space) |
3580 | fatal(0, "LUN expansion failed: class_space %llu <= %llu\n", | |
3581 | old_class_space, new_class_space); | |
3582 | ||
c242c188 | 3583 | if (ztest_opts.zo_verbose >= 5) { |
f3c8c9e6 | 3584 | char oldnumbuf[NN_NUMBUF_SZ], newnumbuf[NN_NUMBUF_SZ]; |
9babb374 | 3585 | |
f3c8c9e6 JK |
3586 | nicenum(old_class_space, oldnumbuf, sizeof (oldnumbuf)); |
3587 | nicenum(new_class_space, newnumbuf, sizeof (newnumbuf)); | |
9babb374 BB |
3588 | (void) printf("%s grew from %s to %s\n", |
3589 | spa->spa_name, oldnumbuf, newnumbuf); | |
3590 | } | |
428870ff | 3591 | |
9babb374 | 3592 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 3593 | mutex_exit(&ztest_vdev_lock); |
34dc7c2f BB |
3594 | } |
3595 | ||
428870ff BB |
3596 | /* |
3597 | * Verify that dmu_objset_{create,destroy,open,close} work as expected. | |
3598 | */ | |
34dc7c2f BB |
3599 | /* ARGSUSED */ |
3600 | static void | |
428870ff | 3601 | ztest_objset_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) |
34dc7c2f BB |
3602 | { |
3603 | /* | |
428870ff | 3604 | * Create the objects common to all ztest datasets. |
34dc7c2f | 3605 | */ |
428870ff | 3606 | VERIFY(zap_create_claim(os, ZTEST_DIROBJ, |
34dc7c2f | 3607 | DMU_OT_ZAP_OTHER, DMU_OT_NONE, 0, tx) == 0); |
428870ff | 3608 | } |
34dc7c2f | 3609 | |
428870ff BB |
3610 | static int |
3611 | ztest_dataset_create(char *dsname) | |
3612 | { | |
4807c0ba TC |
3613 | int err; |
3614 | uint64_t rand; | |
3615 | dsl_crypto_params_t *dcp = NULL; | |
3616 | ||
3617 | /* | |
3618 | * 50% of the time, we create encrypted datasets | |
3619 | * using a random cipher suite and a hard-coded | |
3620 | * wrapping key. | |
3621 | */ | |
3622 | rand = ztest_random(2); | |
3623 | if (rand != 0) { | |
3624 | nvlist_t *crypto_args = fnvlist_alloc(); | |
3625 | nvlist_t *props = fnvlist_alloc(); | |
3626 | ||
3627 | /* slight bias towards the default cipher suite */ | |
3628 | rand = ztest_random(ZIO_CRYPT_FUNCTIONS); | |
3629 | if (rand < ZIO_CRYPT_AES_128_CCM) | |
3630 | rand = ZIO_CRYPT_ON; | |
3631 | ||
3632 | fnvlist_add_uint64(props, | |
3633 | zfs_prop_to_name(ZFS_PROP_ENCRYPTION), rand); | |
3634 | fnvlist_add_uint8_array(crypto_args, "wkeydata", | |
3635 | (uint8_t *)ztest_wkeydata, WRAPPING_KEY_LEN); | |
3636 | ||
3637 | /* | |
3638 | * These parameters aren't really used by the kernel. They | |
3639 | * are simply stored so that userspace knows how to load | |
3640 | * the wrapping key. | |
3641 | */ | |
3642 | fnvlist_add_uint64(props, | |
3643 | zfs_prop_to_name(ZFS_PROP_KEYFORMAT), ZFS_KEYFORMAT_RAW); | |
3644 | fnvlist_add_string(props, | |
3645 | zfs_prop_to_name(ZFS_PROP_KEYLOCATION), "prompt"); | |
3646 | fnvlist_add_uint64(props, | |
3647 | zfs_prop_to_name(ZFS_PROP_PBKDF2_SALT), 0ULL); | |
3648 | fnvlist_add_uint64(props, | |
3649 | zfs_prop_to_name(ZFS_PROP_PBKDF2_ITERS), 0ULL); | |
3650 | ||
3651 | VERIFY0(dsl_crypto_params_create_nvlist(DCP_CMD_NONE, props, | |
3652 | crypto_args, &dcp)); | |
3653 | ||
3654 | fnvlist_free(crypto_args); | |
3655 | fnvlist_free(props); | |
3656 | } | |
3657 | ||
3658 | err = dmu_objset_create(dsname, DMU_OST_OTHER, 0, dcp, | |
428870ff | 3659 | ztest_objset_create_cb, NULL); |
4807c0ba | 3660 | dsl_crypto_params_free(dcp, !!err); |
428870ff | 3661 | |
4807c0ba TC |
3662 | rand = ztest_random(100); |
3663 | if (err || rand < 80) | |
428870ff BB |
3664 | return (err); |
3665 | ||
c242c188 | 3666 | if (ztest_opts.zo_verbose >= 5) |
b815ff9a | 3667 | (void) printf("Setting dataset %s to sync always\n", dsname); |
428870ff BB |
3668 | return (ztest_dsl_prop_set_uint64(dsname, ZFS_PROP_SYNC, |
3669 | ZFS_SYNC_ALWAYS, B_FALSE)); | |
34dc7c2f BB |
3670 | } |
3671 | ||
428870ff | 3672 | /* ARGSUSED */ |
34dc7c2f | 3673 | static int |
428870ff | 3674 | ztest_objset_destroy_cb(const char *name, void *arg) |
34dc7c2f | 3675 | { |
34dc7c2f | 3676 | objset_t *os; |
428870ff | 3677 | dmu_object_info_t doi; |
34dc7c2f BB |
3678 | int error; |
3679 | ||
3680 | /* | |
3681 | * Verify that the dataset contains a directory object. | |
3682 | */ | |
4807c0ba TC |
3683 | VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_TRUE, |
3684 | B_TRUE, FTAG, &os)); | |
428870ff | 3685 | error = dmu_object_info(os, ZTEST_DIROBJ, &doi); |
34dc7c2f BB |
3686 | if (error != ENOENT) { |
3687 | /* We could have crashed in the middle of destroying it */ | |
c99c9001 | 3688 | ASSERT0(error); |
428870ff BB |
3689 | ASSERT3U(doi.doi_type, ==, DMU_OT_ZAP_OTHER); |
3690 | ASSERT3S(doi.doi_physical_blocks_512, >=, 0); | |
34dc7c2f | 3691 | } |
b5256303 | 3692 | dmu_objset_disown(os, B_TRUE, FTAG); |
34dc7c2f BB |
3693 | |
3694 | /* | |
3695 | * Destroy the dataset. | |
3696 | */ | |
13fe0198 | 3697 | if (strchr(name, '@') != NULL) { |
dc1fbc43 | 3698 | VERIFY0(dsl_destroy_snapshot(name, B_TRUE)); |
13fe0198 | 3699 | } else { |
dc1fbc43 GM |
3700 | error = dsl_destroy_head(name); |
3701 | /* There could be a hold on this dataset */ | |
3702 | if (error != EBUSY) | |
3703 | ASSERT0(error); | |
13fe0198 | 3704 | } |
34dc7c2f BB |
3705 | return (0); |
3706 | } | |
3707 | ||
428870ff BB |
3708 | static boolean_t |
3709 | ztest_snapshot_create(char *osname, uint64_t id) | |
34dc7c2f | 3710 | { |
eca7b760 | 3711 | char snapname[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
3712 | int error; |
3713 | ||
13fe0198 | 3714 | (void) snprintf(snapname, sizeof (snapname), "%llu", (u_longlong_t)id); |
428870ff | 3715 | |
13fe0198 | 3716 | error = dmu_objset_snapshot_one(osname, snapname); |
428870ff BB |
3717 | if (error == ENOSPC) { |
3718 | ztest_record_enospc(FTAG); | |
3719 | return (B_FALSE); | |
3720 | } | |
13fe0198 MA |
3721 | if (error != 0 && error != EEXIST) { |
3722 | fatal(0, "ztest_snapshot_create(%s@%s) = %d", osname, | |
3723 | snapname, error); | |
3724 | } | |
428870ff BB |
3725 | return (B_TRUE); |
3726 | } | |
3727 | ||
3728 | static boolean_t | |
3729 | ztest_snapshot_destroy(char *osname, uint64_t id) | |
3730 | { | |
eca7b760 | 3731 | char snapname[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
3732 | int error; |
3733 | ||
eca7b760 | 3734 | (void) snprintf(snapname, sizeof (snapname), "%s@%llu", osname, |
428870ff BB |
3735 | (u_longlong_t)id); |
3736 | ||
13fe0198 | 3737 | error = dsl_destroy_snapshot(snapname, B_FALSE); |
428870ff BB |
3738 | if (error != 0 && error != ENOENT) |
3739 | fatal(0, "ztest_snapshot_destroy(%s) = %d", snapname, error); | |
3740 | return (B_TRUE); | |
34dc7c2f BB |
3741 | } |
3742 | ||
428870ff | 3743 | /* ARGSUSED */ |
34dc7c2f | 3744 | void |
428870ff | 3745 | ztest_dmu_objset_create_destroy(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3746 | { |
40b84e7a | 3747 | ztest_ds_t *zdtmp; |
428870ff | 3748 | int iters; |
34dc7c2f | 3749 | int error; |
b128c09f | 3750 | objset_t *os, *os2; |
eca7b760 | 3751 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
34dc7c2f | 3752 | zilog_t *zilog; |
d6320ddb | 3753 | int i; |
34dc7c2f | 3754 | |
40b84e7a | 3755 | zdtmp = umem_alloc(sizeof (ztest_ds_t), UMEM_NOFAIL); |
40b84e7a | 3756 | |
7809eb8b | 3757 | (void) rw_rdlock(&ztest_name_lock); |
34dc7c2f | 3758 | |
eca7b760 | 3759 | (void) snprintf(name, sizeof (name), "%s/temp_%llu", |
c242c188 | 3760 | ztest_opts.zo_pool, (u_longlong_t)id); |
34dc7c2f BB |
3761 | |
3762 | /* | |
3763 | * If this dataset exists from a previous run, process its replay log | |
13fe0198 | 3764 | * half of the time. If we don't replay it, then dsl_destroy_head() |
428870ff | 3765 | * (invoked from ztest_objset_destroy_cb()) should just throw it away. |
34dc7c2f BB |
3766 | */ |
3767 | if (ztest_random(2) == 0 && | |
4807c0ba | 3768 | ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, |
b5256303 | 3769 | B_TRUE, FTAG, &os) == 0) { |
c242c188 | 3770 | ztest_zd_init(zdtmp, NULL, os); |
40b84e7a BB |
3771 | zil_replay(os, zdtmp, ztest_replay_vector); |
3772 | ztest_zd_fini(zdtmp); | |
4807c0ba | 3773 | txg_wait_synced(dmu_objset_pool(os), 0); |
b5256303 | 3774 | dmu_objset_disown(os, B_TRUE, FTAG); |
34dc7c2f BB |
3775 | } |
3776 | ||
3777 | /* | |
3778 | * There may be an old instance of the dataset we're about to | |
3779 | * create lying around from a previous run. If so, destroy it | |
3780 | * and all of its snapshots. | |
3781 | */ | |
428870ff | 3782 | (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL, |
34dc7c2f BB |
3783 | DS_FIND_CHILDREN | DS_FIND_SNAPSHOTS); |
3784 | ||
3785 | /* | |
3786 | * Verify that the destroyed dataset is no longer in the namespace. | |
3787 | */ | |
4807c0ba TC |
3788 | VERIFY3U(ENOENT, ==, ztest_dmu_objset_own(name, DMU_OST_OTHER, B_TRUE, |
3789 | B_TRUE, FTAG, &os)); | |
34dc7c2f BB |
3790 | |
3791 | /* | |
3792 | * Verify that we can create a new dataset. | |
3793 | */ | |
428870ff | 3794 | error = ztest_dataset_create(name); |
34dc7c2f BB |
3795 | if (error) { |
3796 | if (error == ENOSPC) { | |
428870ff | 3797 | ztest_record_enospc(FTAG); |
40b84e7a | 3798 | goto out; |
34dc7c2f BB |
3799 | } |
3800 | fatal(0, "dmu_objset_create(%s) = %d", name, error); | |
3801 | } | |
3802 | ||
4807c0ba | 3803 | VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, B_TRUE, |
b5256303 | 3804 | FTAG, &os)); |
428870ff | 3805 | |
c242c188 | 3806 | ztest_zd_init(zdtmp, NULL, os); |
34dc7c2f BB |
3807 | |
3808 | /* | |
3809 | * Open the intent log for it. | |
3810 | */ | |
428870ff | 3811 | zilog = zil_open(os, ztest_get_data); |
34dc7c2f BB |
3812 | |
3813 | /* | |
428870ff BB |
3814 | * Put some objects in there, do a little I/O to them, |
3815 | * and randomly take a couple of snapshots along the way. | |
34dc7c2f | 3816 | */ |
428870ff | 3817 | iters = ztest_random(5); |
d6320ddb | 3818 | for (i = 0; i < iters; i++) { |
40b84e7a | 3819 | ztest_dmu_object_alloc_free(zdtmp, id); |
428870ff BB |
3820 | if (ztest_random(iters) == 0) |
3821 | (void) ztest_snapshot_create(name, i); | |
34dc7c2f BB |
3822 | } |
3823 | ||
3824 | /* | |
3825 | * Verify that we cannot create an existing dataset. | |
3826 | */ | |
428870ff | 3827 | VERIFY3U(EEXIST, ==, |
b5256303 | 3828 | dmu_objset_create(name, DMU_OST_OTHER, 0, NULL, NULL, NULL)); |
34dc7c2f BB |
3829 | |
3830 | /* | |
428870ff | 3831 | * Verify that we can hold an objset that is also owned. |
b128c09f | 3832 | */ |
428870ff BB |
3833 | VERIFY3U(0, ==, dmu_objset_hold(name, FTAG, &os2)); |
3834 | dmu_objset_rele(os2, FTAG); | |
34dc7c2f | 3835 | |
428870ff BB |
3836 | /* |
3837 | * Verify that we cannot own an objset that is already owned. | |
3838 | */ | |
4807c0ba TC |
3839 | VERIFY3U(EBUSY, ==, ztest_dmu_objset_own(name, DMU_OST_OTHER, |
3840 | B_FALSE, B_TRUE, FTAG, &os2)); | |
34dc7c2f | 3841 | |
428870ff | 3842 | zil_close(zilog); |
4807c0ba | 3843 | txg_wait_synced(spa_get_dsl(os->os_spa), 0); |
b5256303 | 3844 | dmu_objset_disown(os, B_TRUE, FTAG); |
40b84e7a BB |
3845 | ztest_zd_fini(zdtmp); |
3846 | out: | |
7809eb8b | 3847 | (void) rw_unlock(&ztest_name_lock); |
40b84e7a | 3848 | |
40b84e7a | 3849 | umem_free(zdtmp, sizeof (ztest_ds_t)); |
34dc7c2f BB |
3850 | } |
3851 | ||
3852 | /* | |
3853 | * Verify that dmu_snapshot_{create,destroy,open,close} work as expected. | |
3854 | */ | |
3855 | void | |
428870ff | 3856 | ztest_dmu_snapshot_create_destroy(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3857 | { |
7809eb8b | 3858 | (void) rw_rdlock(&ztest_name_lock); |
428870ff BB |
3859 | (void) ztest_snapshot_destroy(zd->zd_name, id); |
3860 | (void) ztest_snapshot_create(zd->zd_name, id); | |
7809eb8b | 3861 | (void) rw_unlock(&ztest_name_lock); |
34dc7c2f BB |
3862 | } |
3863 | ||
9babb374 BB |
3864 | /* |
3865 | * Cleanup non-standard snapshots and clones. | |
3866 | */ | |
3867 | void | |
428870ff | 3868 | ztest_dsl_dataset_cleanup(char *osname, uint64_t id) |
9babb374 | 3869 | { |
40b84e7a BB |
3870 | char *snap1name; |
3871 | char *clone1name; | |
3872 | char *snap2name; | |
3873 | char *clone2name; | |
3874 | char *snap3name; | |
9babb374 BB |
3875 | int error; |
3876 | ||
eca7b760 IK |
3877 | snap1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); |
3878 | clone1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
3879 | snap2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
3880 | clone2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
3881 | snap3name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
3882 | ||
3883 | (void) snprintf(snap1name, ZFS_MAX_DATASET_NAME_LEN, | |
3884 | "%s@s1_%llu", osname, (u_longlong_t)id); | |
3885 | (void) snprintf(clone1name, ZFS_MAX_DATASET_NAME_LEN, | |
3886 | "%s/c1_%llu", osname, (u_longlong_t)id); | |
3887 | (void) snprintf(snap2name, ZFS_MAX_DATASET_NAME_LEN, | |
3888 | "%s@s2_%llu", clone1name, (u_longlong_t)id); | |
3889 | (void) snprintf(clone2name, ZFS_MAX_DATASET_NAME_LEN, | |
3890 | "%s/c2_%llu", osname, (u_longlong_t)id); | |
3891 | (void) snprintf(snap3name, ZFS_MAX_DATASET_NAME_LEN, | |
3892 | "%s@s3_%llu", clone1name, (u_longlong_t)id); | |
9babb374 | 3893 | |
13fe0198 | 3894 | error = dsl_destroy_head(clone2name); |
9babb374 | 3895 | if (error && error != ENOENT) |
13fe0198 MA |
3896 | fatal(0, "dsl_destroy_head(%s) = %d", clone2name, error); |
3897 | error = dsl_destroy_snapshot(snap3name, B_FALSE); | |
9babb374 | 3898 | if (error && error != ENOENT) |
13fe0198 MA |
3899 | fatal(0, "dsl_destroy_snapshot(%s) = %d", snap3name, error); |
3900 | error = dsl_destroy_snapshot(snap2name, B_FALSE); | |
9babb374 | 3901 | if (error && error != ENOENT) |
13fe0198 MA |
3902 | fatal(0, "dsl_destroy_snapshot(%s) = %d", snap2name, error); |
3903 | error = dsl_destroy_head(clone1name); | |
9babb374 | 3904 | if (error && error != ENOENT) |
13fe0198 MA |
3905 | fatal(0, "dsl_destroy_head(%s) = %d", clone1name, error); |
3906 | error = dsl_destroy_snapshot(snap1name, B_FALSE); | |
9babb374 | 3907 | if (error && error != ENOENT) |
13fe0198 | 3908 | fatal(0, "dsl_destroy_snapshot(%s) = %d", snap1name, error); |
40b84e7a | 3909 | |
eca7b760 IK |
3910 | umem_free(snap1name, ZFS_MAX_DATASET_NAME_LEN); |
3911 | umem_free(clone1name, ZFS_MAX_DATASET_NAME_LEN); | |
3912 | umem_free(snap2name, ZFS_MAX_DATASET_NAME_LEN); | |
3913 | umem_free(clone2name, ZFS_MAX_DATASET_NAME_LEN); | |
3914 | umem_free(snap3name, ZFS_MAX_DATASET_NAME_LEN); | |
9babb374 BB |
3915 | } |
3916 | ||
3917 | /* | |
3918 | * Verify dsl_dataset_promote handles EBUSY | |
3919 | */ | |
3920 | void | |
428870ff | 3921 | ztest_dsl_dataset_promote_busy(ztest_ds_t *zd, uint64_t id) |
9babb374 | 3922 | { |
13fe0198 | 3923 | objset_t *os; |
40b84e7a BB |
3924 | char *snap1name; |
3925 | char *clone1name; | |
3926 | char *snap2name; | |
3927 | char *clone2name; | |
3928 | char *snap3name; | |
428870ff BB |
3929 | char *osname = zd->zd_name; |
3930 | int error; | |
9babb374 | 3931 | |
eca7b760 IK |
3932 | snap1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); |
3933 | clone1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
3934 | snap2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
3935 | clone2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
3936 | snap3name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
40b84e7a | 3937 | |
7809eb8b | 3938 | (void) rw_rdlock(&ztest_name_lock); |
9babb374 | 3939 | |
428870ff | 3940 | ztest_dsl_dataset_cleanup(osname, id); |
9babb374 | 3941 | |
eca7b760 IK |
3942 | (void) snprintf(snap1name, ZFS_MAX_DATASET_NAME_LEN, |
3943 | "%s@s1_%llu", osname, (u_longlong_t)id); | |
3944 | (void) snprintf(clone1name, ZFS_MAX_DATASET_NAME_LEN, | |
3945 | "%s/c1_%llu", osname, (u_longlong_t)id); | |
3946 | (void) snprintf(snap2name, ZFS_MAX_DATASET_NAME_LEN, | |
3947 | "%s@s2_%llu", clone1name, (u_longlong_t)id); | |
3948 | (void) snprintf(clone2name, ZFS_MAX_DATASET_NAME_LEN, | |
3949 | "%s/c2_%llu", osname, (u_longlong_t)id); | |
3950 | (void) snprintf(snap3name, ZFS_MAX_DATASET_NAME_LEN, | |
3951 | "%s@s3_%llu", clone1name, (u_longlong_t)id); | |
9babb374 | 3952 | |
6f1ffb06 | 3953 | error = dmu_objset_snapshot_one(osname, strchr(snap1name, '@') + 1); |
9babb374 BB |
3954 | if (error && error != EEXIST) { |
3955 | if (error == ENOSPC) { | |
428870ff | 3956 | ztest_record_enospc(FTAG); |
9babb374 BB |
3957 | goto out; |
3958 | } | |
3959 | fatal(0, "dmu_take_snapshot(%s) = %d", snap1name, error); | |
3960 | } | |
3961 | ||
13fe0198 | 3962 | error = dmu_objset_clone(clone1name, snap1name); |
9babb374 BB |
3963 | if (error) { |
3964 | if (error == ENOSPC) { | |
428870ff | 3965 | ztest_record_enospc(FTAG); |
9babb374 BB |
3966 | goto out; |
3967 | } | |
3968 | fatal(0, "dmu_objset_create(%s) = %d", clone1name, error); | |
3969 | } | |
3970 | ||
6f1ffb06 | 3971 | error = dmu_objset_snapshot_one(clone1name, strchr(snap2name, '@') + 1); |
428870ff BB |
3972 | if (error && error != EEXIST) { |
3973 | if (error == ENOSPC) { | |
3974 | ztest_record_enospc(FTAG); | |
3975 | goto out; | |
34dc7c2f | 3976 | } |
428870ff BB |
3977 | fatal(0, "dmu_open_snapshot(%s) = %d", snap2name, error); |
3978 | } | |
34dc7c2f | 3979 | |
6f1ffb06 | 3980 | error = dmu_objset_snapshot_one(clone1name, strchr(snap3name, '@') + 1); |
428870ff BB |
3981 | if (error && error != EEXIST) { |
3982 | if (error == ENOSPC) { | |
3983 | ztest_record_enospc(FTAG); | |
3984 | goto out; | |
3985 | } | |
3986 | fatal(0, "dmu_open_snapshot(%s) = %d", snap3name, error); | |
3987 | } | |
34dc7c2f | 3988 | |
13fe0198 | 3989 | error = dmu_objset_clone(clone2name, snap3name); |
428870ff BB |
3990 | if (error) { |
3991 | if (error == ENOSPC) { | |
3992 | ztest_record_enospc(FTAG); | |
3993 | goto out; | |
3994 | } | |
3995 | fatal(0, "dmu_objset_create(%s) = %d", clone2name, error); | |
3996 | } | |
34dc7c2f | 3997 | |
4807c0ba | 3998 | error = ztest_dmu_objset_own(snap2name, DMU_OST_ANY, B_TRUE, B_TRUE, |
b5256303 | 3999 | FTAG, &os); |
428870ff | 4000 | if (error) |
13fe0198 | 4001 | fatal(0, "dmu_objset_own(%s) = %d", snap2name, error); |
428870ff | 4002 | error = dsl_dataset_promote(clone2name, NULL); |
9b67f605 | 4003 | if (error == ENOSPC) { |
b5256303 | 4004 | dmu_objset_disown(os, B_TRUE, FTAG); |
9b67f605 MA |
4005 | ztest_record_enospc(FTAG); |
4006 | goto out; | |
4007 | } | |
428870ff BB |
4008 | if (error != EBUSY) |
4009 | fatal(0, "dsl_dataset_promote(%s), %d, not EBUSY", clone2name, | |
4010 | error); | |
b5256303 | 4011 | dmu_objset_disown(os, B_TRUE, FTAG); |
34dc7c2f | 4012 | |
428870ff BB |
4013 | out: |
4014 | ztest_dsl_dataset_cleanup(osname, id); | |
34dc7c2f | 4015 | |
7809eb8b | 4016 | (void) rw_unlock(&ztest_name_lock); |
40b84e7a | 4017 | |
eca7b760 IK |
4018 | umem_free(snap1name, ZFS_MAX_DATASET_NAME_LEN); |
4019 | umem_free(clone1name, ZFS_MAX_DATASET_NAME_LEN); | |
4020 | umem_free(snap2name, ZFS_MAX_DATASET_NAME_LEN); | |
4021 | umem_free(clone2name, ZFS_MAX_DATASET_NAME_LEN); | |
4022 | umem_free(snap3name, ZFS_MAX_DATASET_NAME_LEN); | |
428870ff | 4023 | } |
34dc7c2f | 4024 | |
40b84e7a | 4025 | #undef OD_ARRAY_SIZE |
d1d7e268 | 4026 | #define OD_ARRAY_SIZE 4 |
40b84e7a | 4027 | |
428870ff BB |
4028 | /* |
4029 | * Verify that dmu_object_{alloc,free} work as expected. | |
4030 | */ | |
4031 | void | |
4032 | ztest_dmu_object_alloc_free(ztest_ds_t *zd, uint64_t id) | |
4033 | { | |
40b84e7a BB |
4034 | ztest_od_t *od; |
4035 | int batchsize; | |
4036 | int size; | |
d6320ddb | 4037 | int b; |
34dc7c2f | 4038 | |
d1d7e268 | 4039 | size = sizeof (ztest_od_t) * OD_ARRAY_SIZE; |
40b84e7a BB |
4040 | od = umem_alloc(size, UMEM_NOFAIL); |
4041 | batchsize = OD_ARRAY_SIZE; | |
4042 | ||
d6320ddb | 4043 | for (b = 0; b < batchsize; b++) |
50c957f7 NB |
4044 | ztest_od_init(od + b, id, FTAG, b, DMU_OT_UINT64_OTHER, |
4045 | 0, 0, 0); | |
34dc7c2f | 4046 | |
428870ff BB |
4047 | /* |
4048 | * Destroy the previous batch of objects, create a new batch, | |
4049 | * and do some I/O on the new objects. | |
4050 | */ | |
40b84e7a | 4051 | if (ztest_object_init(zd, od, size, B_TRUE) != 0) |
428870ff | 4052 | return; |
34dc7c2f | 4053 | |
428870ff BB |
4054 | while (ztest_random(4 * batchsize) != 0) |
4055 | ztest_io(zd, od[ztest_random(batchsize)].od_object, | |
4056 | ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
40b84e7a BB |
4057 | |
4058 | umem_free(od, size); | |
34dc7c2f BB |
4059 | } |
4060 | ||
047116ac TC |
4061 | /* |
4062 | * Rewind the global allocator to verify object allocation backfilling. | |
4063 | */ | |
4064 | void | |
4065 | ztest_dmu_object_next_chunk(ztest_ds_t *zd, uint64_t id) | |
4066 | { | |
4067 | objset_t *os = zd->zd_os; | |
4068 | int dnodes_per_chunk = 1 << dmu_object_alloc_chunk_shift; | |
4069 | uint64_t object; | |
4070 | ||
4071 | /* | |
4072 | * Rewind the global allocator randomly back to a lower object number | |
4073 | * to force backfilling and reclamation of recently freed dnodes. | |
4074 | */ | |
4075 | mutex_enter(&os->os_obj_lock); | |
4076 | object = ztest_random(os->os_obj_next_chunk); | |
4077 | os->os_obj_next_chunk = P2ALIGN(object, dnodes_per_chunk); | |
4078 | mutex_exit(&os->os_obj_lock); | |
4079 | } | |
4080 | ||
40b84e7a | 4081 | #undef OD_ARRAY_SIZE |
d1d7e268 | 4082 | #define OD_ARRAY_SIZE 2 |
40b84e7a | 4083 | |
34dc7c2f BB |
4084 | /* |
4085 | * Verify that dmu_{read,write} work as expected. | |
4086 | */ | |
34dc7c2f | 4087 | void |
428870ff | 4088 | ztest_dmu_read_write(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4089 | { |
40b84e7a BB |
4090 | int size; |
4091 | ztest_od_t *od; | |
4092 | ||
428870ff | 4093 | objset_t *os = zd->zd_os; |
d1d7e268 | 4094 | size = sizeof (ztest_od_t) * OD_ARRAY_SIZE; |
40b84e7a | 4095 | od = umem_alloc(size, UMEM_NOFAIL); |
34dc7c2f BB |
4096 | dmu_tx_t *tx; |
4097 | int i, freeit, error; | |
4098 | uint64_t n, s, txg; | |
4099 | bufwad_t *packbuf, *bigbuf, *pack, *bigH, *bigT; | |
428870ff BB |
4100 | uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize; |
4101 | uint64_t chunksize = (1000 + ztest_random(1000)) * sizeof (uint64_t); | |
34dc7c2f BB |
4102 | uint64_t regions = 997; |
4103 | uint64_t stride = 123456789ULL; | |
4104 | uint64_t width = 40; | |
4105 | int free_percent = 5; | |
4106 | ||
4107 | /* | |
4108 | * This test uses two objects, packobj and bigobj, that are always | |
4109 | * updated together (i.e. in the same tx) so that their contents are | |
4110 | * in sync and can be compared. Their contents relate to each other | |
4111 | * in a simple way: packobj is a dense array of 'bufwad' structures, | |
4112 | * while bigobj is a sparse array of the same bufwads. Specifically, | |
4113 | * for any index n, there are three bufwads that should be identical: | |
4114 | * | |
4115 | * packobj, at offset n * sizeof (bufwad_t) | |
4116 | * bigobj, at the head of the nth chunk | |
4117 | * bigobj, at the tail of the nth chunk | |
4118 | * | |
4119 | * The chunk size is arbitrary. It doesn't have to be a power of two, | |
4120 | * and it doesn't have any relation to the object blocksize. | |
4121 | * The only requirement is that it can hold at least two bufwads. | |
4122 | * | |
4123 | * Normally, we write the bufwad to each of these locations. | |
4124 | * However, free_percent of the time we instead write zeroes to | |
4125 | * packobj and perform a dmu_free_range() on bigobj. By comparing | |
4126 | * bigobj to packobj, we can verify that the DMU is correctly | |
4127 | * tracking which parts of an object are allocated and free, | |
4128 | * and that the contents of the allocated blocks are correct. | |
4129 | */ | |
4130 | ||
4131 | /* | |
4132 | * Read the directory info. If it's the first time, set things up. | |
4133 | */ | |
50c957f7 NB |
4134 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, chunksize); |
4135 | ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, 0, | |
4136 | chunksize); | |
34dc7c2f | 4137 | |
40b84e7a BB |
4138 | if (ztest_object_init(zd, od, size, B_FALSE) != 0) { |
4139 | umem_free(od, size); | |
428870ff | 4140 | return; |
40b84e7a | 4141 | } |
34dc7c2f | 4142 | |
428870ff BB |
4143 | bigobj = od[0].od_object; |
4144 | packobj = od[1].od_object; | |
4145 | chunksize = od[0].od_gen; | |
4146 | ASSERT(chunksize == od[1].od_gen); | |
34dc7c2f BB |
4147 | |
4148 | /* | |
4149 | * Prefetch a random chunk of the big object. | |
4150 | * Our aim here is to get some async reads in flight | |
4151 | * for blocks that we may free below; the DMU should | |
4152 | * handle this race correctly. | |
4153 | */ | |
4154 | n = ztest_random(regions) * stride + ztest_random(width); | |
4155 | s = 1 + ztest_random(2 * width - 1); | |
fcff0f35 PD |
4156 | dmu_prefetch(os, bigobj, 0, n * chunksize, s * chunksize, |
4157 | ZIO_PRIORITY_SYNC_READ); | |
34dc7c2f BB |
4158 | |
4159 | /* | |
4160 | * Pick a random index and compute the offsets into packobj and bigobj. | |
4161 | */ | |
4162 | n = ztest_random(regions) * stride + ztest_random(width); | |
4163 | s = 1 + ztest_random(width - 1); | |
4164 | ||
4165 | packoff = n * sizeof (bufwad_t); | |
4166 | packsize = s * sizeof (bufwad_t); | |
4167 | ||
428870ff BB |
4168 | bigoff = n * chunksize; |
4169 | bigsize = s * chunksize; | |
34dc7c2f BB |
4170 | |
4171 | packbuf = umem_alloc(packsize, UMEM_NOFAIL); | |
4172 | bigbuf = umem_alloc(bigsize, UMEM_NOFAIL); | |
4173 | ||
4174 | /* | |
4175 | * free_percent of the time, free a range of bigobj rather than | |
4176 | * overwriting it. | |
4177 | */ | |
4178 | freeit = (ztest_random(100) < free_percent); | |
4179 | ||
4180 | /* | |
4181 | * Read the current contents of our objects. | |
4182 | */ | |
428870ff | 4183 | error = dmu_read(os, packobj, packoff, packsize, packbuf, |
9babb374 | 4184 | DMU_READ_PREFETCH); |
c99c9001 | 4185 | ASSERT0(error); |
428870ff | 4186 | error = dmu_read(os, bigobj, bigoff, bigsize, bigbuf, |
9babb374 | 4187 | DMU_READ_PREFETCH); |
c99c9001 | 4188 | ASSERT0(error); |
34dc7c2f BB |
4189 | |
4190 | /* | |
4191 | * Get a tx for the mods to both packobj and bigobj. | |
4192 | */ | |
4193 | tx = dmu_tx_create(os); | |
4194 | ||
428870ff | 4195 | dmu_tx_hold_write(tx, packobj, packoff, packsize); |
34dc7c2f BB |
4196 | |
4197 | if (freeit) | |
428870ff | 4198 | dmu_tx_hold_free(tx, bigobj, bigoff, bigsize); |
34dc7c2f | 4199 | else |
428870ff | 4200 | dmu_tx_hold_write(tx, bigobj, bigoff, bigsize); |
34dc7c2f | 4201 | |
383fc4a9 MA |
4202 | /* This accounts for setting the checksum/compression. */ |
4203 | dmu_tx_hold_bonus(tx, bigobj); | |
4204 | ||
428870ff BB |
4205 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); |
4206 | if (txg == 0) { | |
34dc7c2f BB |
4207 | umem_free(packbuf, packsize); |
4208 | umem_free(bigbuf, bigsize); | |
40b84e7a | 4209 | umem_free(od, size); |
34dc7c2f BB |
4210 | return; |
4211 | } | |
4212 | ||
9b67f605 MA |
4213 | enum zio_checksum cksum; |
4214 | do { | |
4215 | cksum = (enum zio_checksum) | |
4216 | ztest_random_dsl_prop(ZFS_PROP_CHECKSUM); | |
4217 | } while (cksum >= ZIO_CHECKSUM_LEGACY_FUNCTIONS); | |
4218 | dmu_object_set_checksum(os, bigobj, cksum, tx); | |
428870ff | 4219 | |
9b67f605 MA |
4220 | enum zio_compress comp; |
4221 | do { | |
4222 | comp = (enum zio_compress) | |
4223 | ztest_random_dsl_prop(ZFS_PROP_COMPRESSION); | |
4224 | } while (comp >= ZIO_COMPRESS_LEGACY_FUNCTIONS); | |
4225 | dmu_object_set_compress(os, bigobj, comp, tx); | |
34dc7c2f BB |
4226 | |
4227 | /* | |
4228 | * For each index from n to n + s, verify that the existing bufwad | |
4229 | * in packobj matches the bufwads at the head and tail of the | |
4230 | * corresponding chunk in bigobj. Then update all three bufwads | |
4231 | * with the new values we want to write out. | |
4232 | */ | |
4233 | for (i = 0; i < s; i++) { | |
4234 | /* LINTED */ | |
4235 | pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t)); | |
4236 | /* LINTED */ | |
428870ff | 4237 | bigH = (bufwad_t *)((char *)bigbuf + i * chunksize); |
34dc7c2f | 4238 | /* LINTED */ |
428870ff | 4239 | bigT = (bufwad_t *)((char *)bigH + chunksize) - 1; |
34dc7c2f BB |
4240 | |
4241 | ASSERT((uintptr_t)bigH - (uintptr_t)bigbuf < bigsize); | |
4242 | ASSERT((uintptr_t)bigT - (uintptr_t)bigbuf < bigsize); | |
4243 | ||
4244 | if (pack->bw_txg > txg) | |
4245 | fatal(0, "future leak: got %llx, open txg is %llx", | |
4246 | pack->bw_txg, txg); | |
4247 | ||
4248 | if (pack->bw_data != 0 && pack->bw_index != n + i) | |
4249 | fatal(0, "wrong index: got %llx, wanted %llx+%llx", | |
4250 | pack->bw_index, n, i); | |
4251 | ||
4252 | if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0) | |
4253 | fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH); | |
4254 | ||
4255 | if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0) | |
4256 | fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT); | |
4257 | ||
4258 | if (freeit) { | |
4259 | bzero(pack, sizeof (bufwad_t)); | |
4260 | } else { | |
4261 | pack->bw_index = n + i; | |
4262 | pack->bw_txg = txg; | |
4263 | pack->bw_data = 1 + ztest_random(-2ULL); | |
4264 | } | |
4265 | *bigH = *pack; | |
4266 | *bigT = *pack; | |
4267 | } | |
4268 | ||
4269 | /* | |
4270 | * We've verified all the old bufwads, and made new ones. | |
4271 | * Now write them out. | |
4272 | */ | |
428870ff | 4273 | dmu_write(os, packobj, packoff, packsize, packbuf, tx); |
34dc7c2f BB |
4274 | |
4275 | if (freeit) { | |
c242c188 | 4276 | if (ztest_opts.zo_verbose >= 7) { |
34dc7c2f BB |
4277 | (void) printf("freeing offset %llx size %llx" |
4278 | " txg %llx\n", | |
4279 | (u_longlong_t)bigoff, | |
4280 | (u_longlong_t)bigsize, | |
4281 | (u_longlong_t)txg); | |
4282 | } | |
428870ff | 4283 | VERIFY(0 == dmu_free_range(os, bigobj, bigoff, bigsize, tx)); |
34dc7c2f | 4284 | } else { |
c242c188 | 4285 | if (ztest_opts.zo_verbose >= 7) { |
34dc7c2f BB |
4286 | (void) printf("writing offset %llx size %llx" |
4287 | " txg %llx\n", | |
4288 | (u_longlong_t)bigoff, | |
4289 | (u_longlong_t)bigsize, | |
4290 | (u_longlong_t)txg); | |
4291 | } | |
428870ff | 4292 | dmu_write(os, bigobj, bigoff, bigsize, bigbuf, tx); |
34dc7c2f BB |
4293 | } |
4294 | ||
4295 | dmu_tx_commit(tx); | |
4296 | ||
4297 | /* | |
4298 | * Sanity check the stuff we just wrote. | |
4299 | */ | |
4300 | { | |
4301 | void *packcheck = umem_alloc(packsize, UMEM_NOFAIL); | |
4302 | void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL); | |
4303 | ||
428870ff | 4304 | VERIFY(0 == dmu_read(os, packobj, packoff, |
9babb374 | 4305 | packsize, packcheck, DMU_READ_PREFETCH)); |
428870ff | 4306 | VERIFY(0 == dmu_read(os, bigobj, bigoff, |
9babb374 | 4307 | bigsize, bigcheck, DMU_READ_PREFETCH)); |
34dc7c2f BB |
4308 | |
4309 | ASSERT(bcmp(packbuf, packcheck, packsize) == 0); | |
4310 | ASSERT(bcmp(bigbuf, bigcheck, bigsize) == 0); | |
4311 | ||
4312 | umem_free(packcheck, packsize); | |
4313 | umem_free(bigcheck, bigsize); | |
4314 | } | |
4315 | ||
4316 | umem_free(packbuf, packsize); | |
4317 | umem_free(bigbuf, bigsize); | |
40b84e7a | 4318 | umem_free(od, size); |
34dc7c2f BB |
4319 | } |
4320 | ||
9babb374 BB |
4321 | void |
4322 | compare_and_update_pbbufs(uint64_t s, bufwad_t *packbuf, bufwad_t *bigbuf, | |
428870ff | 4323 | uint64_t bigsize, uint64_t n, uint64_t chunksize, uint64_t txg) |
9babb374 BB |
4324 | { |
4325 | uint64_t i; | |
4326 | bufwad_t *pack; | |
4327 | bufwad_t *bigH; | |
4328 | bufwad_t *bigT; | |
4329 | ||
4330 | /* | |
4331 | * For each index from n to n + s, verify that the existing bufwad | |
4332 | * in packobj matches the bufwads at the head and tail of the | |
4333 | * corresponding chunk in bigobj. Then update all three bufwads | |
4334 | * with the new values we want to write out. | |
4335 | */ | |
4336 | for (i = 0; i < s; i++) { | |
4337 | /* LINTED */ | |
4338 | pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t)); | |
4339 | /* LINTED */ | |
428870ff | 4340 | bigH = (bufwad_t *)((char *)bigbuf + i * chunksize); |
9babb374 | 4341 | /* LINTED */ |
428870ff | 4342 | bigT = (bufwad_t *)((char *)bigH + chunksize) - 1; |
9babb374 BB |
4343 | |
4344 | ASSERT((uintptr_t)bigH - (uintptr_t)bigbuf < bigsize); | |
4345 | ASSERT((uintptr_t)bigT - (uintptr_t)bigbuf < bigsize); | |
4346 | ||
4347 | if (pack->bw_txg > txg) | |
4348 | fatal(0, "future leak: got %llx, open txg is %llx", | |
4349 | pack->bw_txg, txg); | |
4350 | ||
4351 | if (pack->bw_data != 0 && pack->bw_index != n + i) | |
4352 | fatal(0, "wrong index: got %llx, wanted %llx+%llx", | |
4353 | pack->bw_index, n, i); | |
4354 | ||
4355 | if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0) | |
4356 | fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH); | |
4357 | ||
4358 | if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0) | |
4359 | fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT); | |
4360 | ||
4361 | pack->bw_index = n + i; | |
4362 | pack->bw_txg = txg; | |
4363 | pack->bw_data = 1 + ztest_random(-2ULL); | |
4364 | ||
4365 | *bigH = *pack; | |
4366 | *bigT = *pack; | |
4367 | } | |
4368 | } | |
4369 | ||
40b84e7a | 4370 | #undef OD_ARRAY_SIZE |
d1d7e268 | 4371 | #define OD_ARRAY_SIZE 2 |
40b84e7a | 4372 | |
9babb374 | 4373 | void |
428870ff | 4374 | ztest_dmu_read_write_zcopy(ztest_ds_t *zd, uint64_t id) |
9babb374 | 4375 | { |
428870ff | 4376 | objset_t *os = zd->zd_os; |
40b84e7a | 4377 | ztest_od_t *od; |
9babb374 BB |
4378 | dmu_tx_t *tx; |
4379 | uint64_t i; | |
4380 | int error; | |
40b84e7a | 4381 | int size; |
9babb374 BB |
4382 | uint64_t n, s, txg; |
4383 | bufwad_t *packbuf, *bigbuf; | |
428870ff BB |
4384 | uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize; |
4385 | uint64_t blocksize = ztest_random_blocksize(); | |
4386 | uint64_t chunksize = blocksize; | |
9babb374 BB |
4387 | uint64_t regions = 997; |
4388 | uint64_t stride = 123456789ULL; | |
4389 | uint64_t width = 9; | |
4390 | dmu_buf_t *bonus_db; | |
4391 | arc_buf_t **bigbuf_arcbufs; | |
428870ff | 4392 | dmu_object_info_t doi; |
9babb374 | 4393 | |
d1d7e268 | 4394 | size = sizeof (ztest_od_t) * OD_ARRAY_SIZE; |
40b84e7a BB |
4395 | od = umem_alloc(size, UMEM_NOFAIL); |
4396 | ||
9babb374 BB |
4397 | /* |
4398 | * This test uses two objects, packobj and bigobj, that are always | |
4399 | * updated together (i.e. in the same tx) so that their contents are | |
4400 | * in sync and can be compared. Their contents relate to each other | |
4401 | * in a simple way: packobj is a dense array of 'bufwad' structures, | |
4402 | * while bigobj is a sparse array of the same bufwads. Specifically, | |
4403 | * for any index n, there are three bufwads that should be identical: | |
4404 | * | |
4405 | * packobj, at offset n * sizeof (bufwad_t) | |
4406 | * bigobj, at the head of the nth chunk | |
4407 | * bigobj, at the tail of the nth chunk | |
4408 | * | |
4409 | * The chunk size is set equal to bigobj block size so that | |
440a3eb9 | 4410 | * dmu_assign_arcbuf_by_dbuf() can be tested for object updates. |
9babb374 BB |
4411 | */ |
4412 | ||
4413 | /* | |
4414 | * Read the directory info. If it's the first time, set things up. | |
4415 | */ | |
50c957f7 NB |
4416 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0, 0); |
4417 | ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, 0, | |
4418 | chunksize); | |
40b84e7a | 4419 | |
9babb374 | 4420 | |
40b84e7a BB |
4421 | if (ztest_object_init(zd, od, size, B_FALSE) != 0) { |
4422 | umem_free(od, size); | |
428870ff | 4423 | return; |
40b84e7a | 4424 | } |
9babb374 | 4425 | |
428870ff BB |
4426 | bigobj = od[0].od_object; |
4427 | packobj = od[1].od_object; | |
4428 | blocksize = od[0].od_blocksize; | |
4429 | chunksize = blocksize; | |
4430 | ASSERT(chunksize == od[1].od_gen); | |
9babb374 | 4431 | |
428870ff BB |
4432 | VERIFY(dmu_object_info(os, bigobj, &doi) == 0); |
4433 | VERIFY(ISP2(doi.doi_data_block_size)); | |
4434 | VERIFY(chunksize == doi.doi_data_block_size); | |
4435 | VERIFY(chunksize >= 2 * sizeof (bufwad_t)); | |
9babb374 BB |
4436 | |
4437 | /* | |
4438 | * Pick a random index and compute the offsets into packobj and bigobj. | |
4439 | */ | |
4440 | n = ztest_random(regions) * stride + ztest_random(width); | |
4441 | s = 1 + ztest_random(width - 1); | |
4442 | ||
4443 | packoff = n * sizeof (bufwad_t); | |
4444 | packsize = s * sizeof (bufwad_t); | |
4445 | ||
428870ff BB |
4446 | bigoff = n * chunksize; |
4447 | bigsize = s * chunksize; | |
9babb374 BB |
4448 | |
4449 | packbuf = umem_zalloc(packsize, UMEM_NOFAIL); | |
4450 | bigbuf = umem_zalloc(bigsize, UMEM_NOFAIL); | |
4451 | ||
428870ff | 4452 | VERIFY3U(0, ==, dmu_bonus_hold(os, bigobj, FTAG, &bonus_db)); |
9babb374 BB |
4453 | |
4454 | bigbuf_arcbufs = umem_zalloc(2 * s * sizeof (arc_buf_t *), UMEM_NOFAIL); | |
4455 | ||
4456 | /* | |
4457 | * Iteration 0 test zcopy for DB_UNCACHED dbufs. | |
4458 | * Iteration 1 test zcopy to already referenced dbufs. | |
4459 | * Iteration 2 test zcopy to dirty dbuf in the same txg. | |
4460 | * Iteration 3 test zcopy to dbuf dirty in previous txg. | |
4461 | * Iteration 4 test zcopy when dbuf is no longer dirty. | |
4462 | * Iteration 5 test zcopy when it can't be done. | |
4463 | * Iteration 6 one more zcopy write. | |
4464 | */ | |
4465 | for (i = 0; i < 7; i++) { | |
4466 | uint64_t j; | |
4467 | uint64_t off; | |
4468 | ||
4469 | /* | |
4470 | * In iteration 5 (i == 5) use arcbufs | |
4471 | * that don't match bigobj blksz to test | |
440a3eb9 | 4472 | * dmu_assign_arcbuf_by_dbuf() when it can't directly |
9babb374 BB |
4473 | * assign an arcbuf to a dbuf. |
4474 | */ | |
4475 | for (j = 0; j < s; j++) { | |
b9541d6b | 4476 | if (i != 5 || chunksize < (SPA_MINBLOCKSIZE * 2)) { |
9babb374 | 4477 | bigbuf_arcbufs[j] = |
428870ff | 4478 | dmu_request_arcbuf(bonus_db, chunksize); |
9babb374 BB |
4479 | } else { |
4480 | bigbuf_arcbufs[2 * j] = | |
428870ff | 4481 | dmu_request_arcbuf(bonus_db, chunksize / 2); |
9babb374 | 4482 | bigbuf_arcbufs[2 * j + 1] = |
428870ff | 4483 | dmu_request_arcbuf(bonus_db, chunksize / 2); |
9babb374 BB |
4484 | } |
4485 | } | |
4486 | ||
4487 | /* | |
4488 | * Get a tx for the mods to both packobj and bigobj. | |
4489 | */ | |
4490 | tx = dmu_tx_create(os); | |
4491 | ||
428870ff BB |
4492 | dmu_tx_hold_write(tx, packobj, packoff, packsize); |
4493 | dmu_tx_hold_write(tx, bigobj, bigoff, bigsize); | |
9babb374 | 4494 | |
428870ff BB |
4495 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); |
4496 | if (txg == 0) { | |
9babb374 BB |
4497 | umem_free(packbuf, packsize); |
4498 | umem_free(bigbuf, bigsize); | |
4499 | for (j = 0; j < s; j++) { | |
b9541d6b CW |
4500 | if (i != 5 || |
4501 | chunksize < (SPA_MINBLOCKSIZE * 2)) { | |
9babb374 BB |
4502 | dmu_return_arcbuf(bigbuf_arcbufs[j]); |
4503 | } else { | |
4504 | dmu_return_arcbuf( | |
4505 | bigbuf_arcbufs[2 * j]); | |
4506 | dmu_return_arcbuf( | |
4507 | bigbuf_arcbufs[2 * j + 1]); | |
4508 | } | |
4509 | } | |
4510 | umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *)); | |
40b84e7a | 4511 | umem_free(od, size); |
9babb374 BB |
4512 | dmu_buf_rele(bonus_db, FTAG); |
4513 | return; | |
4514 | } | |
4515 | ||
9babb374 BB |
4516 | /* |
4517 | * 50% of the time don't read objects in the 1st iteration to | |
440a3eb9 TC |
4518 | * test dmu_assign_arcbuf_by_dbuf() for the case when there are |
4519 | * no existing dbufs for the specified offsets. | |
9babb374 BB |
4520 | */ |
4521 | if (i != 0 || ztest_random(2) != 0) { | |
428870ff | 4522 | error = dmu_read(os, packobj, packoff, |
9babb374 | 4523 | packsize, packbuf, DMU_READ_PREFETCH); |
c99c9001 | 4524 | ASSERT0(error); |
428870ff | 4525 | error = dmu_read(os, bigobj, bigoff, bigsize, |
9babb374 | 4526 | bigbuf, DMU_READ_PREFETCH); |
c99c9001 | 4527 | ASSERT0(error); |
9babb374 BB |
4528 | } |
4529 | compare_and_update_pbbufs(s, packbuf, bigbuf, bigsize, | |
428870ff | 4530 | n, chunksize, txg); |
9babb374 BB |
4531 | |
4532 | /* | |
4533 | * We've verified all the old bufwads, and made new ones. | |
4534 | * Now write them out. | |
4535 | */ | |
428870ff | 4536 | dmu_write(os, packobj, packoff, packsize, packbuf, tx); |
c242c188 | 4537 | if (ztest_opts.zo_verbose >= 7) { |
9babb374 BB |
4538 | (void) printf("writing offset %llx size %llx" |
4539 | " txg %llx\n", | |
4540 | (u_longlong_t)bigoff, | |
4541 | (u_longlong_t)bigsize, | |
4542 | (u_longlong_t)txg); | |
4543 | } | |
428870ff | 4544 | for (off = bigoff, j = 0; j < s; j++, off += chunksize) { |
9babb374 | 4545 | dmu_buf_t *dbt; |
b9541d6b | 4546 | if (i != 5 || chunksize < (SPA_MINBLOCKSIZE * 2)) { |
9babb374 | 4547 | bcopy((caddr_t)bigbuf + (off - bigoff), |
428870ff | 4548 | bigbuf_arcbufs[j]->b_data, chunksize); |
9babb374 BB |
4549 | } else { |
4550 | bcopy((caddr_t)bigbuf + (off - bigoff), | |
4551 | bigbuf_arcbufs[2 * j]->b_data, | |
428870ff | 4552 | chunksize / 2); |
9babb374 | 4553 | bcopy((caddr_t)bigbuf + (off - bigoff) + |
428870ff | 4554 | chunksize / 2, |
9babb374 | 4555 | bigbuf_arcbufs[2 * j + 1]->b_data, |
428870ff | 4556 | chunksize / 2); |
9babb374 BB |
4557 | } |
4558 | ||
4559 | if (i == 1) { | |
428870ff BB |
4560 | VERIFY(dmu_buf_hold(os, bigobj, off, |
4561 | FTAG, &dbt, DMU_READ_NO_PREFETCH) == 0); | |
9babb374 | 4562 | } |
b9541d6b | 4563 | if (i != 5 || chunksize < (SPA_MINBLOCKSIZE * 2)) { |
440a3eb9 | 4564 | dmu_assign_arcbuf_by_dbuf(bonus_db, off, |
9babb374 BB |
4565 | bigbuf_arcbufs[j], tx); |
4566 | } else { | |
440a3eb9 | 4567 | dmu_assign_arcbuf_by_dbuf(bonus_db, off, |
9babb374 | 4568 | bigbuf_arcbufs[2 * j], tx); |
440a3eb9 | 4569 | dmu_assign_arcbuf_by_dbuf(bonus_db, |
428870ff | 4570 | off + chunksize / 2, |
9babb374 BB |
4571 | bigbuf_arcbufs[2 * j + 1], tx); |
4572 | } | |
4573 | if (i == 1) { | |
4574 | dmu_buf_rele(dbt, FTAG); | |
4575 | } | |
4576 | } | |
4577 | dmu_tx_commit(tx); | |
4578 | ||
4579 | /* | |
4580 | * Sanity check the stuff we just wrote. | |
4581 | */ | |
4582 | { | |
4583 | void *packcheck = umem_alloc(packsize, UMEM_NOFAIL); | |
4584 | void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL); | |
4585 | ||
428870ff | 4586 | VERIFY(0 == dmu_read(os, packobj, packoff, |
9babb374 | 4587 | packsize, packcheck, DMU_READ_PREFETCH)); |
428870ff | 4588 | VERIFY(0 == dmu_read(os, bigobj, bigoff, |
9babb374 BB |
4589 | bigsize, bigcheck, DMU_READ_PREFETCH)); |
4590 | ||
4591 | ASSERT(bcmp(packbuf, packcheck, packsize) == 0); | |
4592 | ASSERT(bcmp(bigbuf, bigcheck, bigsize) == 0); | |
4593 | ||
4594 | umem_free(packcheck, packsize); | |
4595 | umem_free(bigcheck, bigsize); | |
4596 | } | |
4597 | if (i == 2) { | |
4598 | txg_wait_open(dmu_objset_pool(os), 0); | |
4599 | } else if (i == 3) { | |
4600 | txg_wait_synced(dmu_objset_pool(os), 0); | |
4601 | } | |
4602 | } | |
4603 | ||
4604 | dmu_buf_rele(bonus_db, FTAG); | |
4605 | umem_free(packbuf, packsize); | |
4606 | umem_free(bigbuf, bigsize); | |
4607 | umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *)); | |
40b84e7a | 4608 | umem_free(od, size); |
9babb374 BB |
4609 | } |
4610 | ||
428870ff | 4611 | /* ARGSUSED */ |
34dc7c2f | 4612 | void |
428870ff | 4613 | ztest_dmu_write_parallel(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4614 | { |
40b84e7a BB |
4615 | ztest_od_t *od; |
4616 | ||
d1d7e268 | 4617 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
428870ff BB |
4618 | uint64_t offset = (1ULL << (ztest_random(20) + 43)) + |
4619 | (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
34dc7c2f BB |
4620 | |
4621 | /* | |
428870ff BB |
4622 | * Have multiple threads write to large offsets in an object |
4623 | * to verify that parallel writes to an object -- even to the | |
4624 | * same blocks within the object -- doesn't cause any trouble. | |
34dc7c2f | 4625 | */ |
50c957f7 | 4626 | ztest_od_init(od, ID_PARALLEL, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0); |
9babb374 | 4627 | |
40b84e7a | 4628 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) |
34dc7c2f | 4629 | return; |
34dc7c2f | 4630 | |
428870ff | 4631 | while (ztest_random(10) != 0) |
40b84e7a BB |
4632 | ztest_io(zd, od->od_object, offset); |
4633 | ||
d1d7e268 | 4634 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 4635 | } |
34dc7c2f | 4636 | |
428870ff BB |
4637 | void |
4638 | ztest_dmu_prealloc(ztest_ds_t *zd, uint64_t id) | |
4639 | { | |
40b84e7a | 4640 | ztest_od_t *od; |
428870ff BB |
4641 | uint64_t offset = (1ULL << (ztest_random(4) + SPA_MAXBLOCKSHIFT)) + |
4642 | (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
4643 | uint64_t count = ztest_random(20) + 1; | |
4644 | uint64_t blocksize = ztest_random_blocksize(); | |
4645 | void *data; | |
34dc7c2f | 4646 | |
d1d7e268 | 4647 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
34dc7c2f | 4648 | |
50c957f7 | 4649 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0, 0); |
40b84e7a | 4650 | |
d1d7e268 MK |
4651 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
4652 | !ztest_random(2)) != 0) { | |
4653 | umem_free(od, sizeof (ztest_od_t)); | |
34dc7c2f | 4654 | return; |
40b84e7a | 4655 | } |
34dc7c2f | 4656 | |
40b84e7a | 4657 | if (ztest_truncate(zd, od->od_object, offset, count * blocksize) != 0) { |
d1d7e268 | 4658 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f | 4659 | return; |
40b84e7a | 4660 | } |
34dc7c2f | 4661 | |
40b84e7a | 4662 | ztest_prealloc(zd, od->od_object, offset, count * blocksize); |
34dc7c2f | 4663 | |
428870ff | 4664 | data = umem_zalloc(blocksize, UMEM_NOFAIL); |
34dc7c2f | 4665 | |
428870ff BB |
4666 | while (ztest_random(count) != 0) { |
4667 | uint64_t randoff = offset + (ztest_random(count) * blocksize); | |
40b84e7a | 4668 | if (ztest_write(zd, od->od_object, randoff, blocksize, |
428870ff BB |
4669 | data) != 0) |
4670 | break; | |
4671 | while (ztest_random(4) != 0) | |
40b84e7a | 4672 | ztest_io(zd, od->od_object, randoff); |
9babb374 | 4673 | } |
34dc7c2f | 4674 | |
428870ff | 4675 | umem_free(data, blocksize); |
d1d7e268 | 4676 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f BB |
4677 | } |
4678 | ||
4679 | /* | |
4680 | * Verify that zap_{create,destroy,add,remove,update} work as expected. | |
4681 | */ | |
4682 | #define ZTEST_ZAP_MIN_INTS 1 | |
4683 | #define ZTEST_ZAP_MAX_INTS 4 | |
4684 | #define ZTEST_ZAP_MAX_PROPS 1000 | |
4685 | ||
4686 | void | |
428870ff | 4687 | ztest_zap(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4688 | { |
428870ff | 4689 | objset_t *os = zd->zd_os; |
40b84e7a | 4690 | ztest_od_t *od; |
34dc7c2f BB |
4691 | uint64_t object; |
4692 | uint64_t txg, last_txg; | |
4693 | uint64_t value[ZTEST_ZAP_MAX_INTS]; | |
4694 | uint64_t zl_ints, zl_intsize, prop; | |
4695 | int i, ints; | |
4696 | dmu_tx_t *tx; | |
4697 | char propname[100], txgname[100]; | |
4698 | int error; | |
34dc7c2f BB |
4699 | char *hc[2] = { "s.acl.h", ".s.open.h.hyLZlg" }; |
4700 | ||
d1d7e268 | 4701 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 4702 | ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0, 0); |
34dc7c2f | 4703 | |
40b84e7a | 4704 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
02730c33 | 4705 | !ztest_random(2)) != 0) |
40b84e7a | 4706 | goto out; |
34dc7c2f | 4707 | |
40b84e7a | 4708 | object = od->od_object; |
34dc7c2f | 4709 | |
428870ff BB |
4710 | /* |
4711 | * Generate a known hash collision, and verify that | |
4712 | * we can lookup and remove both entries. | |
4713 | */ | |
4714 | tx = dmu_tx_create(os); | |
4715 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); | |
4716 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4717 | if (txg == 0) | |
40b84e7a | 4718 | goto out; |
428870ff BB |
4719 | for (i = 0; i < 2; i++) { |
4720 | value[i] = i; | |
4721 | VERIFY3U(0, ==, zap_add(os, object, hc[i], sizeof (uint64_t), | |
4722 | 1, &value[i], tx)); | |
4723 | } | |
4724 | for (i = 0; i < 2; i++) { | |
4725 | VERIFY3U(EEXIST, ==, zap_add(os, object, hc[i], | |
4726 | sizeof (uint64_t), 1, &value[i], tx)); | |
4727 | VERIFY3U(0, ==, | |
4728 | zap_length(os, object, hc[i], &zl_intsize, &zl_ints)); | |
4729 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
4730 | ASSERT3U(zl_ints, ==, 1); | |
4731 | } | |
4732 | for (i = 0; i < 2; i++) { | |
4733 | VERIFY3U(0, ==, zap_remove(os, object, hc[i], tx)); | |
34dc7c2f | 4734 | } |
428870ff | 4735 | dmu_tx_commit(tx); |
34dc7c2f | 4736 | |
428870ff BB |
4737 | /* |
4738 | * Generate a buch of random entries. | |
4739 | */ | |
34dc7c2f BB |
4740 | ints = MAX(ZTEST_ZAP_MIN_INTS, object % ZTEST_ZAP_MAX_INTS); |
4741 | ||
4742 | prop = ztest_random(ZTEST_ZAP_MAX_PROPS); | |
4743 | (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop); | |
4744 | (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop); | |
4745 | bzero(value, sizeof (value)); | |
4746 | last_txg = 0; | |
4747 | ||
4748 | /* | |
4749 | * If these zap entries already exist, validate their contents. | |
4750 | */ | |
4751 | error = zap_length(os, object, txgname, &zl_intsize, &zl_ints); | |
4752 | if (error == 0) { | |
4753 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
4754 | ASSERT3U(zl_ints, ==, 1); | |
4755 | ||
4756 | VERIFY(zap_lookup(os, object, txgname, zl_intsize, | |
4757 | zl_ints, &last_txg) == 0); | |
4758 | ||
4759 | VERIFY(zap_length(os, object, propname, &zl_intsize, | |
4760 | &zl_ints) == 0); | |
4761 | ||
4762 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
4763 | ASSERT3U(zl_ints, ==, ints); | |
4764 | ||
4765 | VERIFY(zap_lookup(os, object, propname, zl_intsize, | |
4766 | zl_ints, value) == 0); | |
4767 | ||
4768 | for (i = 0; i < ints; i++) { | |
4769 | ASSERT3U(value[i], ==, last_txg + object + i); | |
4770 | } | |
4771 | } else { | |
4772 | ASSERT3U(error, ==, ENOENT); | |
4773 | } | |
4774 | ||
4775 | /* | |
4776 | * Atomically update two entries in our zap object. | |
4777 | * The first is named txg_%llu, and contains the txg | |
4778 | * in which the property was last updated. The second | |
4779 | * is named prop_%llu, and the nth element of its value | |
4780 | * should be txg + object + n. | |
4781 | */ | |
4782 | tx = dmu_tx_create(os); | |
428870ff BB |
4783 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
4784 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4785 | if (txg == 0) | |
40b84e7a | 4786 | goto out; |
34dc7c2f BB |
4787 | |
4788 | if (last_txg > txg) | |
4789 | fatal(0, "zap future leak: old %llu new %llu", last_txg, txg); | |
4790 | ||
4791 | for (i = 0; i < ints; i++) | |
4792 | value[i] = txg + object + i; | |
4793 | ||
428870ff BB |
4794 | VERIFY3U(0, ==, zap_update(os, object, txgname, sizeof (uint64_t), |
4795 | 1, &txg, tx)); | |
4796 | VERIFY3U(0, ==, zap_update(os, object, propname, sizeof (uint64_t), | |
4797 | ints, value, tx)); | |
34dc7c2f BB |
4798 | |
4799 | dmu_tx_commit(tx); | |
4800 | ||
4801 | /* | |
4802 | * Remove a random pair of entries. | |
4803 | */ | |
4804 | prop = ztest_random(ZTEST_ZAP_MAX_PROPS); | |
4805 | (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop); | |
4806 | (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop); | |
4807 | ||
4808 | error = zap_length(os, object, txgname, &zl_intsize, &zl_ints); | |
4809 | ||
4810 | if (error == ENOENT) | |
40b84e7a | 4811 | goto out; |
34dc7c2f | 4812 | |
c99c9001 | 4813 | ASSERT0(error); |
34dc7c2f BB |
4814 | |
4815 | tx = dmu_tx_create(os); | |
428870ff BB |
4816 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
4817 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4818 | if (txg == 0) | |
40b84e7a | 4819 | goto out; |
428870ff BB |
4820 | VERIFY3U(0, ==, zap_remove(os, object, txgname, tx)); |
4821 | VERIFY3U(0, ==, zap_remove(os, object, propname, tx)); | |
4822 | dmu_tx_commit(tx); | |
40b84e7a | 4823 | out: |
d1d7e268 | 4824 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 4825 | } |
34dc7c2f | 4826 | |
428870ff BB |
4827 | /* |
4828 | * Testcase to test the upgrading of a microzap to fatzap. | |
4829 | */ | |
4830 | void | |
4831 | ztest_fzap(ztest_ds_t *zd, uint64_t id) | |
4832 | { | |
4833 | objset_t *os = zd->zd_os; | |
40b84e7a | 4834 | ztest_od_t *od; |
428870ff | 4835 | uint64_t object, txg; |
d6320ddb | 4836 | int i; |
34dc7c2f | 4837 | |
d1d7e268 | 4838 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 4839 | ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0, 0); |
428870ff | 4840 | |
40b84e7a | 4841 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
02730c33 | 4842 | !ztest_random(2)) != 0) |
40b84e7a BB |
4843 | goto out; |
4844 | object = od->od_object; | |
34dc7c2f BB |
4845 | |
4846 | /* | |
428870ff BB |
4847 | * Add entries to this ZAP and make sure it spills over |
4848 | * and gets upgraded to a fatzap. Also, since we are adding | |
4849 | * 2050 entries we should see ptrtbl growth and leaf-block split. | |
34dc7c2f | 4850 | */ |
d6320ddb | 4851 | for (i = 0; i < 2050; i++) { |
eca7b760 | 4852 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
4853 | uint64_t value = i; |
4854 | dmu_tx_t *tx; | |
4855 | int error; | |
34dc7c2f | 4856 | |
428870ff | 4857 | (void) snprintf(name, sizeof (name), "fzap-%llu-%llu", |
b8864a23 | 4858 | (u_longlong_t)id, (u_longlong_t)value); |
428870ff BB |
4859 | |
4860 | tx = dmu_tx_create(os); | |
4861 | dmu_tx_hold_zap(tx, object, B_TRUE, name); | |
4862 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4863 | if (txg == 0) | |
40b84e7a | 4864 | goto out; |
428870ff BB |
4865 | error = zap_add(os, object, name, sizeof (uint64_t), 1, |
4866 | &value, tx); | |
4867 | ASSERT(error == 0 || error == EEXIST); | |
4868 | dmu_tx_commit(tx); | |
34dc7c2f | 4869 | } |
40b84e7a | 4870 | out: |
d1d7e268 | 4871 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f BB |
4872 | } |
4873 | ||
428870ff | 4874 | /* ARGSUSED */ |
34dc7c2f | 4875 | void |
428870ff | 4876 | ztest_zap_parallel(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4877 | { |
428870ff | 4878 | objset_t *os = zd->zd_os; |
40b84e7a | 4879 | ztest_od_t *od; |
34dc7c2f BB |
4880 | uint64_t txg, object, count, wsize, wc, zl_wsize, zl_wc; |
4881 | dmu_tx_t *tx; | |
4882 | int i, namelen, error; | |
428870ff | 4883 | int micro = ztest_random(2); |
34dc7c2f BB |
4884 | char name[20], string_value[20]; |
4885 | void *data; | |
4886 | ||
d1d7e268 | 4887 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 4888 | ztest_od_init(od, ID_PARALLEL, FTAG, micro, DMU_OT_ZAP_OTHER, 0, 0, 0); |
428870ff | 4889 | |
40b84e7a | 4890 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
d1d7e268 | 4891 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 4892 | return; |
40b84e7a | 4893 | } |
428870ff | 4894 | |
40b84e7a | 4895 | object = od->od_object; |
428870ff | 4896 | |
34dc7c2f BB |
4897 | /* |
4898 | * Generate a random name of the form 'xxx.....' where each | |
4899 | * x is a random printable character and the dots are dots. | |
4900 | * There are 94 such characters, and the name length goes from | |
4901 | * 6 to 20, so there are 94^3 * 15 = 12,458,760 possible names. | |
4902 | */ | |
4903 | namelen = ztest_random(sizeof (name) - 5) + 5 + 1; | |
4904 | ||
4905 | for (i = 0; i < 3; i++) | |
4906 | name[i] = '!' + ztest_random('~' - '!' + 1); | |
4907 | for (; i < namelen - 1; i++) | |
4908 | name[i] = '.'; | |
4909 | name[i] = '\0'; | |
4910 | ||
428870ff | 4911 | if ((namelen & 1) || micro) { |
34dc7c2f BB |
4912 | wsize = sizeof (txg); |
4913 | wc = 1; | |
4914 | data = &txg; | |
4915 | } else { | |
4916 | wsize = 1; | |
4917 | wc = namelen; | |
4918 | data = string_value; | |
4919 | } | |
4920 | ||
4921 | count = -1ULL; | |
13fe0198 | 4922 | VERIFY0(zap_count(os, object, &count)); |
34dc7c2f BB |
4923 | ASSERT(count != -1ULL); |
4924 | ||
4925 | /* | |
4926 | * Select an operation: length, lookup, add, update, remove. | |
4927 | */ | |
4928 | i = ztest_random(5); | |
4929 | ||
4930 | if (i >= 2) { | |
4931 | tx = dmu_tx_create(os); | |
428870ff BB |
4932 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
4933 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
df053d67 GN |
4934 | if (txg == 0) { |
4935 | umem_free(od, sizeof (ztest_od_t)); | |
34dc7c2f | 4936 | return; |
df053d67 | 4937 | } |
34dc7c2f BB |
4938 | bcopy(name, string_value, namelen); |
4939 | } else { | |
4940 | tx = NULL; | |
4941 | txg = 0; | |
4942 | bzero(string_value, namelen); | |
4943 | } | |
4944 | ||
4945 | switch (i) { | |
4946 | ||
4947 | case 0: | |
4948 | error = zap_length(os, object, name, &zl_wsize, &zl_wc); | |
4949 | if (error == 0) { | |
4950 | ASSERT3U(wsize, ==, zl_wsize); | |
4951 | ASSERT3U(wc, ==, zl_wc); | |
4952 | } else { | |
4953 | ASSERT3U(error, ==, ENOENT); | |
4954 | } | |
4955 | break; | |
4956 | ||
4957 | case 1: | |
4958 | error = zap_lookup(os, object, name, wsize, wc, data); | |
4959 | if (error == 0) { | |
4960 | if (data == string_value && | |
4961 | bcmp(name, data, namelen) != 0) | |
4962 | fatal(0, "name '%s' != val '%s' len %d", | |
4963 | name, data, namelen); | |
4964 | } else { | |
4965 | ASSERT3U(error, ==, ENOENT); | |
4966 | } | |
4967 | break; | |
4968 | ||
4969 | case 2: | |
4970 | error = zap_add(os, object, name, wsize, wc, data, tx); | |
4971 | ASSERT(error == 0 || error == EEXIST); | |
4972 | break; | |
4973 | ||
4974 | case 3: | |
4975 | VERIFY(zap_update(os, object, name, wsize, wc, data, tx) == 0); | |
4976 | break; | |
4977 | ||
4978 | case 4: | |
4979 | error = zap_remove(os, object, name, tx); | |
4980 | ASSERT(error == 0 || error == ENOENT); | |
4981 | break; | |
4982 | } | |
4983 | ||
4984 | if (tx != NULL) | |
4985 | dmu_tx_commit(tx); | |
40b84e7a | 4986 | |
d1d7e268 | 4987 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f BB |
4988 | } |
4989 | ||
428870ff BB |
4990 | /* |
4991 | * Commit callback data. | |
4992 | */ | |
4993 | typedef struct ztest_cb_data { | |
4994 | list_node_t zcd_node; | |
4995 | uint64_t zcd_txg; | |
4996 | int zcd_expected_err; | |
4997 | boolean_t zcd_added; | |
4998 | boolean_t zcd_called; | |
4999 | spa_t *zcd_spa; | |
5000 | } ztest_cb_data_t; | |
5001 | ||
5002 | /* This is the actual commit callback function */ | |
5003 | static void | |
5004 | ztest_commit_callback(void *arg, int error) | |
5005 | { | |
5006 | ztest_cb_data_t *data = arg; | |
5007 | uint64_t synced_txg; | |
5008 | ||
5009 | VERIFY(data != NULL); | |
5010 | VERIFY3S(data->zcd_expected_err, ==, error); | |
5011 | VERIFY(!data->zcd_called); | |
5012 | ||
5013 | synced_txg = spa_last_synced_txg(data->zcd_spa); | |
5014 | if (data->zcd_txg > synced_txg) | |
5015 | fatal(0, "commit callback of txg %" PRIu64 " called prematurely" | |
5016 | ", last synced txg = %" PRIu64 "\n", data->zcd_txg, | |
5017 | synced_txg); | |
5018 | ||
5019 | data->zcd_called = B_TRUE; | |
5020 | ||
5021 | if (error == ECANCELED) { | |
c99c9001 | 5022 | ASSERT0(data->zcd_txg); |
428870ff BB |
5023 | ASSERT(!data->zcd_added); |
5024 | ||
5025 | /* | |
5026 | * The private callback data should be destroyed here, but | |
5027 | * since we are going to check the zcd_called field after | |
5028 | * dmu_tx_abort(), we will destroy it there. | |
5029 | */ | |
5030 | return; | |
5031 | } | |
5032 | ||
090ff092 | 5033 | ASSERT(data->zcd_added); |
428870ff BB |
5034 | ASSERT3U(data->zcd_txg, !=, 0); |
5035 | ||
1e33ac1e | 5036 | (void) mutex_enter(&zcl.zcl_callbacks_lock); |
090ff092 RC |
5037 | |
5038 | /* See if this cb was called more quickly */ | |
5039 | if ((synced_txg - data->zcd_txg) < zc_min_txg_delay) | |
5040 | zc_min_txg_delay = synced_txg - data->zcd_txg; | |
5041 | ||
5042 | /* Remove our callback from the list */ | |
428870ff | 5043 | list_remove(&zcl.zcl_callbacks, data); |
090ff092 | 5044 | |
1e33ac1e | 5045 | (void) mutex_exit(&zcl.zcl_callbacks_lock); |
428870ff | 5046 | |
428870ff BB |
5047 | umem_free(data, sizeof (ztest_cb_data_t)); |
5048 | } | |
5049 | ||
5050 | /* Allocate and initialize callback data structure */ | |
5051 | static ztest_cb_data_t * | |
5052 | ztest_create_cb_data(objset_t *os, uint64_t txg) | |
5053 | { | |
5054 | ztest_cb_data_t *cb_data; | |
5055 | ||
5056 | cb_data = umem_zalloc(sizeof (ztest_cb_data_t), UMEM_NOFAIL); | |
5057 | ||
5058 | cb_data->zcd_txg = txg; | |
5059 | cb_data->zcd_spa = dmu_objset_spa(os); | |
1e33ac1e | 5060 | list_link_init(&cb_data->zcd_node); |
428870ff BB |
5061 | |
5062 | return (cb_data); | |
5063 | } | |
5064 | ||
428870ff BB |
5065 | /* |
5066 | * Commit callback test. | |
5067 | */ | |
34dc7c2f | 5068 | void |
428870ff BB |
5069 | ztest_dmu_commit_callbacks(ztest_ds_t *zd, uint64_t id) |
5070 | { | |
5071 | objset_t *os = zd->zd_os; | |
40b84e7a | 5072 | ztest_od_t *od; |
428870ff BB |
5073 | dmu_tx_t *tx; |
5074 | ztest_cb_data_t *cb_data[3], *tmp_cb; | |
5075 | uint64_t old_txg, txg; | |
090ff092 | 5076 | int i, error = 0; |
428870ff | 5077 | |
d1d7e268 | 5078 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 5079 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0); |
428870ff | 5080 | |
40b84e7a | 5081 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
d1d7e268 | 5082 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 5083 | return; |
40b84e7a | 5084 | } |
428870ff BB |
5085 | |
5086 | tx = dmu_tx_create(os); | |
5087 | ||
5088 | cb_data[0] = ztest_create_cb_data(os, 0); | |
5089 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[0]); | |
5090 | ||
40b84e7a | 5091 | dmu_tx_hold_write(tx, od->od_object, 0, sizeof (uint64_t)); |
428870ff BB |
5092 | |
5093 | /* Every once in a while, abort the transaction on purpose */ | |
5094 | if (ztest_random(100) == 0) | |
5095 | error = -1; | |
5096 | ||
5097 | if (!error) | |
5098 | error = dmu_tx_assign(tx, TXG_NOWAIT); | |
5099 | ||
5100 | txg = error ? 0 : dmu_tx_get_txg(tx); | |
5101 | ||
5102 | cb_data[0]->zcd_txg = txg; | |
5103 | cb_data[1] = ztest_create_cb_data(os, txg); | |
5104 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[1]); | |
5105 | ||
5106 | if (error) { | |
5107 | /* | |
5108 | * It's not a strict requirement to call the registered | |
5109 | * callbacks from inside dmu_tx_abort(), but that's what | |
5110 | * it's supposed to happen in the current implementation | |
5111 | * so we will check for that. | |
5112 | */ | |
5113 | for (i = 0; i < 2; i++) { | |
5114 | cb_data[i]->zcd_expected_err = ECANCELED; | |
5115 | VERIFY(!cb_data[i]->zcd_called); | |
5116 | } | |
5117 | ||
5118 | dmu_tx_abort(tx); | |
5119 | ||
5120 | for (i = 0; i < 2; i++) { | |
5121 | VERIFY(cb_data[i]->zcd_called); | |
5122 | umem_free(cb_data[i], sizeof (ztest_cb_data_t)); | |
5123 | } | |
5124 | ||
d1d7e268 | 5125 | umem_free(od, sizeof (ztest_od_t)); |
428870ff BB |
5126 | return; |
5127 | } | |
5128 | ||
5129 | cb_data[2] = ztest_create_cb_data(os, txg); | |
5130 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[2]); | |
5131 | ||
5132 | /* | |
5133 | * Read existing data to make sure there isn't a future leak. | |
5134 | */ | |
40b84e7a | 5135 | VERIFY(0 == dmu_read(os, od->od_object, 0, sizeof (uint64_t), |
428870ff BB |
5136 | &old_txg, DMU_READ_PREFETCH)); |
5137 | ||
5138 | if (old_txg > txg) | |
5139 | fatal(0, "future leak: got %" PRIu64 ", open txg is %" PRIu64, | |
5140 | old_txg, txg); | |
5141 | ||
40b84e7a | 5142 | dmu_write(os, od->od_object, 0, sizeof (uint64_t), &txg, tx); |
428870ff | 5143 | |
1e33ac1e | 5144 | (void) mutex_enter(&zcl.zcl_callbacks_lock); |
428870ff BB |
5145 | |
5146 | /* | |
5147 | * Since commit callbacks don't have any ordering requirement and since | |
5148 | * it is theoretically possible for a commit callback to be called | |
5149 | * after an arbitrary amount of time has elapsed since its txg has been | |
5150 | * synced, it is difficult to reliably determine whether a commit | |
5151 | * callback hasn't been called due to high load or due to a flawed | |
5152 | * implementation. | |
5153 | * | |
5154 | * In practice, we will assume that if after a certain number of txgs a | |
5155 | * commit callback hasn't been called, then most likely there's an | |
5156 | * implementation bug.. | |
5157 | */ | |
5158 | tmp_cb = list_head(&zcl.zcl_callbacks); | |
5159 | if (tmp_cb != NULL && | |
090ff092 | 5160 | tmp_cb->zcd_txg + ZTEST_COMMIT_CB_THRESH < txg) { |
428870ff BB |
5161 | fatal(0, "Commit callback threshold exceeded, oldest txg: %" |
5162 | PRIu64 ", open txg: %" PRIu64 "\n", tmp_cb->zcd_txg, txg); | |
5163 | } | |
5164 | ||
5165 | /* | |
5166 | * Let's find the place to insert our callbacks. | |
5167 | * | |
5168 | * Even though the list is ordered by txg, it is possible for the | |
5169 | * insertion point to not be the end because our txg may already be | |
5170 | * quiescing at this point and other callbacks in the open txg | |
5171 | * (from other objsets) may have sneaked in. | |
5172 | */ | |
5173 | tmp_cb = list_tail(&zcl.zcl_callbacks); | |
5174 | while (tmp_cb != NULL && tmp_cb->zcd_txg > txg) | |
5175 | tmp_cb = list_prev(&zcl.zcl_callbacks, tmp_cb); | |
5176 | ||
5177 | /* Add the 3 callbacks to the list */ | |
5178 | for (i = 0; i < 3; i++) { | |
5179 | if (tmp_cb == NULL) | |
5180 | list_insert_head(&zcl.zcl_callbacks, cb_data[i]); | |
5181 | else | |
5182 | list_insert_after(&zcl.zcl_callbacks, tmp_cb, | |
5183 | cb_data[i]); | |
5184 | ||
5185 | cb_data[i]->zcd_added = B_TRUE; | |
5186 | VERIFY(!cb_data[i]->zcd_called); | |
5187 | ||
5188 | tmp_cb = cb_data[i]; | |
5189 | } | |
5190 | ||
090ff092 RC |
5191 | zc_cb_counter += 3; |
5192 | ||
1e33ac1e | 5193 | (void) mutex_exit(&zcl.zcl_callbacks_lock); |
428870ff BB |
5194 | |
5195 | dmu_tx_commit(tx); | |
40b84e7a | 5196 | |
d1d7e268 | 5197 | umem_free(od, sizeof (ztest_od_t)); |
428870ff BB |
5198 | } |
5199 | ||
50c957f7 NB |
5200 | /* |
5201 | * Visit each object in the dataset. Verify that its properties | |
5202 | * are consistent what was stored in the block tag when it was created, | |
5203 | * and that its unused bonus buffer space has not been overwritten. | |
5204 | */ | |
817b1b6e | 5205 | /* ARGSUSED */ |
50c957f7 NB |
5206 | void |
5207 | ztest_verify_dnode_bt(ztest_ds_t *zd, uint64_t id) | |
5208 | { | |
5209 | objset_t *os = zd->zd_os; | |
5210 | uint64_t obj; | |
5211 | int err = 0; | |
5212 | ||
5213 | for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) { | |
5214 | ztest_block_tag_t *bt = NULL; | |
5215 | dmu_object_info_t doi; | |
5216 | dmu_buf_t *db; | |
5217 | ||
06401e42 BB |
5218 | ztest_object_lock(zd, obj, RL_READER); |
5219 | if (dmu_bonus_hold(os, obj, FTAG, &db) != 0) { | |
5220 | ztest_object_unlock(zd, obj); | |
50c957f7 | 5221 | continue; |
06401e42 | 5222 | } |
50c957f7 NB |
5223 | |
5224 | dmu_object_info_from_db(db, &doi); | |
5225 | if (doi.doi_bonus_size >= sizeof (*bt)) | |
5226 | bt = ztest_bt_bonus(db); | |
5227 | ||
5228 | if (bt && bt->bt_magic == BT_MAGIC) { | |
5229 | ztest_bt_verify(bt, os, obj, doi.doi_dnodesize, | |
5230 | bt->bt_offset, bt->bt_gen, bt->bt_txg, | |
5231 | bt->bt_crtxg); | |
5232 | ztest_verify_unused_bonus(db, bt, obj, os, bt->bt_gen); | |
5233 | } | |
5234 | ||
5235 | dmu_buf_rele(db, FTAG); | |
06401e42 | 5236 | ztest_object_unlock(zd, obj); |
50c957f7 NB |
5237 | } |
5238 | } | |
5239 | ||
428870ff BB |
5240 | /* ARGSUSED */ |
5241 | void | |
5242 | ztest_dsl_prop_get_set(ztest_ds_t *zd, uint64_t id) | |
5243 | { | |
5244 | zfs_prop_t proplist[] = { | |
5245 | ZFS_PROP_CHECKSUM, | |
5246 | ZFS_PROP_COMPRESSION, | |
5247 | ZFS_PROP_COPIES, | |
5248 | ZFS_PROP_DEDUP | |
5249 | }; | |
d6320ddb | 5250 | int p; |
428870ff | 5251 | |
7809eb8b | 5252 | (void) rw_rdlock(&ztest_name_lock); |
428870ff | 5253 | |
d6320ddb | 5254 | for (p = 0; p < sizeof (proplist) / sizeof (proplist[0]); p++) |
428870ff BB |
5255 | (void) ztest_dsl_prop_set_uint64(zd->zd_name, proplist[p], |
5256 | ztest_random_dsl_prop(proplist[p]), (int)ztest_random(2)); | |
5257 | ||
76d52067 BB |
5258 | VERIFY0(ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_RECORDSIZE, |
5259 | ztest_random_blocksize(), (int)ztest_random(2))); | |
5260 | ||
7809eb8b | 5261 | (void) rw_unlock(&ztest_name_lock); |
428870ff BB |
5262 | } |
5263 | ||
5264 | /* ARGSUSED */ | |
5265 | void | |
5266 | ztest_spa_prop_get_set(ztest_ds_t *zd, uint64_t id) | |
5267 | { | |
428870ff BB |
5268 | nvlist_t *props = NULL; |
5269 | ||
7809eb8b | 5270 | (void) rw_rdlock(&ztest_name_lock); |
428870ff | 5271 | |
c242c188 | 5272 | (void) ztest_spa_prop_set_uint64(ZPOOL_PROP_DEDUPDITTO, |
428870ff BB |
5273 | ZIO_DEDUPDITTO_MIN + ztest_random(ZIO_DEDUPDITTO_MIN)); |
5274 | ||
c99c9001 | 5275 | VERIFY0(spa_prop_get(ztest_spa, &props)); |
428870ff | 5276 | |
c242c188 | 5277 | if (ztest_opts.zo_verbose >= 6) |
428870ff BB |
5278 | dump_nvlist(props, 4); |
5279 | ||
5280 | nvlist_free(props); | |
5281 | ||
7809eb8b | 5282 | (void) rw_unlock(&ztest_name_lock); |
428870ff BB |
5283 | } |
5284 | ||
13fe0198 MA |
5285 | static int |
5286 | user_release_one(const char *snapname, const char *holdname) | |
5287 | { | |
5288 | nvlist_t *snaps, *holds; | |
5289 | int error; | |
5290 | ||
5291 | snaps = fnvlist_alloc(); | |
5292 | holds = fnvlist_alloc(); | |
5293 | fnvlist_add_boolean(holds, holdname); | |
5294 | fnvlist_add_nvlist(snaps, snapname, holds); | |
5295 | fnvlist_free(holds); | |
5296 | error = dsl_dataset_user_release(snaps, NULL); | |
5297 | fnvlist_free(snaps); | |
5298 | return (error); | |
5299 | } | |
5300 | ||
428870ff BB |
5301 | /* |
5302 | * Test snapshot hold/release and deferred destroy. | |
5303 | */ | |
5304 | void | |
5305 | ztest_dmu_snapshot_hold(ztest_ds_t *zd, uint64_t id) | |
34dc7c2f | 5306 | { |
34dc7c2f | 5307 | int error; |
428870ff BB |
5308 | objset_t *os = zd->zd_os; |
5309 | objset_t *origin; | |
5310 | char snapname[100]; | |
5311 | char fullname[100]; | |
5312 | char clonename[100]; | |
5313 | char tag[100]; | |
eca7b760 | 5314 | char osname[ZFS_MAX_DATASET_NAME_LEN]; |
13fe0198 | 5315 | nvlist_t *holds; |
34dc7c2f | 5316 | |
7809eb8b | 5317 | (void) rw_rdlock(&ztest_name_lock); |
34dc7c2f BB |
5318 | |
5319 | dmu_objset_name(os, osname); | |
5320 | ||
d1d7e268 MK |
5321 | (void) snprintf(snapname, sizeof (snapname), "sh1_%llu", |
5322 | (u_longlong_t)id); | |
13fe0198 MA |
5323 | (void) snprintf(fullname, sizeof (fullname), "%s@%s", osname, snapname); |
5324 | (void) snprintf(clonename, sizeof (clonename), | |
d1d7e268 MK |
5325 | "%s/ch1_%llu", osname, (u_longlong_t)id); |
5326 | (void) snprintf(tag, sizeof (tag), "tag_%llu", (u_longlong_t)id); | |
428870ff BB |
5327 | |
5328 | /* | |
5329 | * Clean up from any previous run. | |
5330 | */ | |
13fe0198 MA |
5331 | error = dsl_destroy_head(clonename); |
5332 | if (error != ENOENT) | |
5333 | ASSERT0(error); | |
5334 | error = user_release_one(fullname, tag); | |
5335 | if (error != ESRCH && error != ENOENT) | |
5336 | ASSERT0(error); | |
5337 | error = dsl_destroy_snapshot(fullname, B_FALSE); | |
5338 | if (error != ENOENT) | |
5339 | ASSERT0(error); | |
428870ff BB |
5340 | |
5341 | /* | |
5342 | * Create snapshot, clone it, mark snap for deferred destroy, | |
5343 | * destroy clone, verify snap was also destroyed. | |
5344 | */ | |
6f1ffb06 | 5345 | error = dmu_objset_snapshot_one(osname, snapname); |
428870ff BB |
5346 | if (error) { |
5347 | if (error == ENOSPC) { | |
5348 | ztest_record_enospc("dmu_objset_snapshot"); | |
5349 | goto out; | |
34dc7c2f | 5350 | } |
428870ff BB |
5351 | fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error); |
5352 | } | |
34dc7c2f | 5353 | |
13fe0198 | 5354 | error = dmu_objset_clone(clonename, fullname); |
428870ff | 5355 | if (error) { |
34dc7c2f | 5356 | if (error == ENOSPC) { |
428870ff BB |
5357 | ztest_record_enospc("dmu_objset_clone"); |
5358 | goto out; | |
34dc7c2f | 5359 | } |
428870ff BB |
5360 | fatal(0, "dmu_objset_clone(%s) = %d", clonename, error); |
5361 | } | |
34dc7c2f | 5362 | |
13fe0198 | 5363 | error = dsl_destroy_snapshot(fullname, B_TRUE); |
428870ff | 5364 | if (error) { |
13fe0198 | 5365 | fatal(0, "dsl_destroy_snapshot(%s, B_TRUE) = %d", |
428870ff BB |
5366 | fullname, error); |
5367 | } | |
34dc7c2f | 5368 | |
13fe0198 | 5369 | error = dsl_destroy_head(clonename); |
428870ff | 5370 | if (error) |
13fe0198 | 5371 | fatal(0, "dsl_destroy_head(%s) = %d", clonename, error); |
34dc7c2f | 5372 | |
428870ff BB |
5373 | error = dmu_objset_hold(fullname, FTAG, &origin); |
5374 | if (error != ENOENT) | |
5375 | fatal(0, "dmu_objset_hold(%s) = %d", fullname, error); | |
34dc7c2f | 5376 | |
428870ff BB |
5377 | /* |
5378 | * Create snapshot, add temporary hold, verify that we can't | |
5379 | * destroy a held snapshot, mark for deferred destroy, | |
5380 | * release hold, verify snapshot was destroyed. | |
5381 | */ | |
6f1ffb06 | 5382 | error = dmu_objset_snapshot_one(osname, snapname); |
428870ff BB |
5383 | if (error) { |
5384 | if (error == ENOSPC) { | |
5385 | ztest_record_enospc("dmu_objset_snapshot"); | |
5386 | goto out; | |
34dc7c2f | 5387 | } |
428870ff BB |
5388 | fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error); |
5389 | } | |
5390 | ||
13fe0198 MA |
5391 | holds = fnvlist_alloc(); |
5392 | fnvlist_add_string(holds, fullname, tag); | |
5393 | error = dsl_dataset_user_hold(holds, 0, NULL); | |
5394 | fnvlist_free(holds); | |
5395 | ||
9b67f605 MA |
5396 | if (error == ENOSPC) { |
5397 | ztest_record_enospc("dsl_dataset_user_hold"); | |
5398 | goto out; | |
5399 | } else if (error) { | |
5400 | fatal(0, "dsl_dataset_user_hold(%s, %s) = %u", | |
5401 | fullname, tag, error); | |
5402 | } | |
428870ff | 5403 | |
13fe0198 | 5404 | error = dsl_destroy_snapshot(fullname, B_FALSE); |
428870ff | 5405 | if (error != EBUSY) { |
13fe0198 | 5406 | fatal(0, "dsl_destroy_snapshot(%s, B_FALSE) = %d", |
428870ff BB |
5407 | fullname, error); |
5408 | } | |
5409 | ||
13fe0198 | 5410 | error = dsl_destroy_snapshot(fullname, B_TRUE); |
428870ff | 5411 | if (error) { |
13fe0198 | 5412 | fatal(0, "dsl_destroy_snapshot(%s, B_TRUE) = %d", |
428870ff | 5413 | fullname, error); |
34dc7c2f BB |
5414 | } |
5415 | ||
13fe0198 | 5416 | error = user_release_one(fullname, tag); |
428870ff | 5417 | if (error) |
95fd54a1 | 5418 | fatal(0, "user_release_one(%s, %s) = %d", fullname, tag, error); |
428870ff | 5419 | |
13fe0198 | 5420 | VERIFY3U(dmu_objset_hold(fullname, FTAG, &origin), ==, ENOENT); |
428870ff BB |
5421 | |
5422 | out: | |
7809eb8b | 5423 | (void) rw_unlock(&ztest_name_lock); |
34dc7c2f BB |
5424 | } |
5425 | ||
34dc7c2f BB |
5426 | /* |
5427 | * Inject random faults into the on-disk data. | |
5428 | */ | |
428870ff | 5429 | /* ARGSUSED */ |
34dc7c2f | 5430 | void |
428870ff | 5431 | ztest_fault_inject(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 5432 | { |
428870ff | 5433 | ztest_shared_t *zs = ztest_shared; |
c242c188 | 5434 | spa_t *spa = ztest_spa; |
34dc7c2f BB |
5435 | int fd; |
5436 | uint64_t offset; | |
428870ff | 5437 | uint64_t leaves; |
c5b3a7bb | 5438 | uint64_t bad = 0x1990c0ffeedecadeull; |
34dc7c2f | 5439 | uint64_t top, leaf; |
40b84e7a BB |
5440 | char *path0; |
5441 | char *pathrand; | |
34dc7c2f | 5442 | size_t fsize; |
2014c09f | 5443 | int bshift = SPA_MAXBLOCKSHIFT + 2; |
34dc7c2f | 5444 | int iters = 1000; |
428870ff BB |
5445 | int maxfaults; |
5446 | int mirror_save; | |
b128c09f | 5447 | vdev_t *vd0 = NULL; |
34dc7c2f | 5448 | uint64_t guid0 = 0; |
428870ff BB |
5449 | boolean_t islog = B_FALSE; |
5450 | ||
40b84e7a BB |
5451 | path0 = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
5452 | pathrand = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
5453 | ||
c242c188 | 5454 | mutex_enter(&ztest_vdev_lock); |
60b82074 | 5455 | maxfaults = MAXFAULTS(zs); |
c242c188 | 5456 | leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raidz; |
428870ff | 5457 | mirror_save = zs->zs_mirrors; |
c242c188 | 5458 | mutex_exit(&ztest_vdev_lock); |
34dc7c2f | 5459 | |
b128c09f | 5460 | ASSERT(leaves >= 1); |
34dc7c2f | 5461 | |
621dd7bb GW |
5462 | /* |
5463 | * Grab the name lock as reader. There are some operations | |
5464 | * which don't like to have their vdevs changed while | |
5465 | * they are in progress (i.e. spa_change_guid). Those | |
5466 | * operations will have grabbed the name lock as writer. | |
5467 | */ | |
7809eb8b | 5468 | (void) rw_rdlock(&ztest_name_lock); |
621dd7bb | 5469 | |
34dc7c2f | 5470 | /* |
b128c09f | 5471 | * We need SCL_STATE here because we're going to look at vd0->vdev_tsd. |
34dc7c2f | 5472 | */ |
b128c09f | 5473 | spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); |
34dc7c2f | 5474 | |
b128c09f BB |
5475 | if (ztest_random(2) == 0) { |
5476 | /* | |
428870ff | 5477 | * Inject errors on a normal data device or slog device. |
b128c09f | 5478 | */ |
428870ff BB |
5479 | top = ztest_random_vdev_top(spa, B_TRUE); |
5480 | leaf = ztest_random(leaves) + zs->zs_splits; | |
34dc7c2f | 5481 | |
b128c09f BB |
5482 | /* |
5483 | * Generate paths to the first leaf in this top-level vdev, | |
5484 | * and to the random leaf we selected. We'll induce transient | |
5485 | * write failures and random online/offline activity on leaf 0, | |
5486 | * and we'll write random garbage to the randomly chosen leaf. | |
5487 | */ | |
6aec1cd5 | 5488 | (void) snprintf(path0, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
5489 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
5490 | top * leaves + zs->zs_splits); | |
6aec1cd5 | 5491 | (void) snprintf(pathrand, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
5492 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
5493 | top * leaves + leaf); | |
34dc7c2f | 5494 | |
b128c09f | 5495 | vd0 = vdev_lookup_by_path(spa->spa_root_vdev, path0); |
428870ff BB |
5496 | if (vd0 != NULL && vd0->vdev_top->vdev_islog) |
5497 | islog = B_TRUE; | |
5498 | ||
621dd7bb GW |
5499 | /* |
5500 | * If the top-level vdev needs to be resilvered | |
5501 | * then we only allow faults on the device that is | |
5502 | * resilvering. | |
5503 | */ | |
5504 | if (vd0 != NULL && maxfaults != 1 && | |
5505 | (!vdev_resilver_needed(vd0->vdev_top, NULL, NULL) || | |
5d1f7fb6 | 5506 | vd0->vdev_resilver_txg != 0)) { |
b128c09f BB |
5507 | /* |
5508 | * Make vd0 explicitly claim to be unreadable, | |
5509 | * or unwriteable, or reach behind its back | |
5510 | * and close the underlying fd. We can do this if | |
5511 | * maxfaults == 0 because we'll fail and reexecute, | |
5512 | * and we can do it if maxfaults >= 2 because we'll | |
5513 | * have enough redundancy. If maxfaults == 1, the | |
5514 | * combination of this with injection of random data | |
5515 | * corruption below exceeds the pool's fault tolerance. | |
5516 | */ | |
5517 | vdev_file_t *vf = vd0->vdev_tsd; | |
5518 | ||
5519 | if (vf != NULL && ztest_random(3) == 0) { | |
5520 | (void) close(vf->vf_vnode->v_fd); | |
5521 | vf->vf_vnode->v_fd = -1; | |
5522 | } else if (ztest_random(2) == 0) { | |
5523 | vd0->vdev_cant_read = B_TRUE; | |
5524 | } else { | |
5525 | vd0->vdev_cant_write = B_TRUE; | |
5526 | } | |
5527 | guid0 = vd0->vdev_guid; | |
5528 | } | |
5529 | } else { | |
5530 | /* | |
5531 | * Inject errors on an l2cache device. | |
5532 | */ | |
5533 | spa_aux_vdev_t *sav = &spa->spa_l2cache; | |
34dc7c2f | 5534 | |
b128c09f BB |
5535 | if (sav->sav_count == 0) { |
5536 | spa_config_exit(spa, SCL_STATE, FTAG); | |
7809eb8b | 5537 | (void) rw_unlock(&ztest_name_lock); |
40b84e7a | 5538 | goto out; |
b128c09f BB |
5539 | } |
5540 | vd0 = sav->sav_vdevs[ztest_random(sav->sav_count)]; | |
34dc7c2f | 5541 | guid0 = vd0->vdev_guid; |
b128c09f BB |
5542 | (void) strcpy(path0, vd0->vdev_path); |
5543 | (void) strcpy(pathrand, vd0->vdev_path); | |
5544 | ||
5545 | leaf = 0; | |
5546 | leaves = 1; | |
5547 | maxfaults = INT_MAX; /* no limit on cache devices */ | |
34dc7c2f BB |
5548 | } |
5549 | ||
b128c09f | 5550 | spa_config_exit(spa, SCL_STATE, FTAG); |
7809eb8b | 5551 | (void) rw_unlock(&ztest_name_lock); |
b128c09f | 5552 | |
34dc7c2f | 5553 | /* |
428870ff BB |
5554 | * If we can tolerate two or more faults, or we're dealing |
5555 | * with a slog, randomly online/offline vd0. | |
34dc7c2f | 5556 | */ |
428870ff | 5557 | if ((maxfaults >= 2 || islog) && guid0 != 0) { |
fb5f0bc8 BB |
5558 | if (ztest_random(10) < 6) { |
5559 | int flags = (ztest_random(2) == 0 ? | |
5560 | ZFS_OFFLINE_TEMPORARY : 0); | |
428870ff BB |
5561 | |
5562 | /* | |
5563 | * We have to grab the zs_name_lock as writer to | |
5564 | * prevent a race between offlining a slog and | |
5565 | * destroying a dataset. Offlining the slog will | |
5566 | * grab a reference on the dataset which may cause | |
13fe0198 | 5567 | * dsl_destroy_head() to fail with EBUSY thus |
428870ff BB |
5568 | * leaving the dataset in an inconsistent state. |
5569 | */ | |
5570 | if (islog) | |
7809eb8b | 5571 | (void) rw_wrlock(&ztest_name_lock); |
428870ff | 5572 | |
fb5f0bc8 | 5573 | VERIFY(vdev_offline(spa, guid0, flags) != EBUSY); |
428870ff BB |
5574 | |
5575 | if (islog) | |
7809eb8b | 5576 | (void) rw_unlock(&ztest_name_lock); |
fb5f0bc8 | 5577 | } else { |
1eb5bfa3 GW |
5578 | /* |
5579 | * Ideally we would like to be able to randomly | |
5580 | * call vdev_[on|off]line without holding locks | |
5581 | * to force unpredictable failures but the side | |
5582 | * effects of vdev_[on|off]line prevent us from | |
5583 | * doing so. We grab the ztest_vdev_lock here to | |
5584 | * prevent a race between injection testing and | |
5585 | * aux_vdev removal. | |
5586 | */ | |
5587 | mutex_enter(&ztest_vdev_lock); | |
fb5f0bc8 | 5588 | (void) vdev_online(spa, guid0, 0, NULL); |
1eb5bfa3 | 5589 | mutex_exit(&ztest_vdev_lock); |
fb5f0bc8 | 5590 | } |
34dc7c2f BB |
5591 | } |
5592 | ||
428870ff | 5593 | if (maxfaults == 0) |
40b84e7a | 5594 | goto out; |
428870ff | 5595 | |
34dc7c2f BB |
5596 | /* |
5597 | * We have at least single-fault tolerance, so inject data corruption. | |
5598 | */ | |
5599 | fd = open(pathrand, O_RDWR); | |
5600 | ||
5601 | if (fd == -1) /* we hit a gap in the device namespace */ | |
40b84e7a | 5602 | goto out; |
34dc7c2f BB |
5603 | |
5604 | fsize = lseek(fd, 0, SEEK_END); | |
5605 | ||
5606 | while (--iters != 0) { | |
91d88843 MA |
5607 | /* |
5608 | * The offset must be chosen carefully to ensure that | |
5609 | * we do not inject a given logical block with errors | |
5610 | * on two different leaf devices, because ZFS can not | |
5611 | * tolerate that (if maxfaults==1). | |
5612 | * | |
5613 | * We divide each leaf into chunks of size | |
5614 | * (# leaves * SPA_MAXBLOCKSIZE * 4). Within each chunk | |
5615 | * there is a series of ranges to which we can inject errors. | |
5616 | * Each range can accept errors on only a single leaf vdev. | |
5617 | * The error injection ranges are separated by ranges | |
5618 | * which we will not inject errors on any device (DMZs). | |
5619 | * Each DMZ must be large enough such that a single block | |
5620 | * can not straddle it, so that a single block can not be | |
5621 | * a target in two different injection ranges (on different | |
5622 | * leaf vdevs). | |
5623 | * | |
5624 | * For example, with 3 leaves, each chunk looks like: | |
5625 | * 0 to 32M: injection range for leaf 0 | |
5626 | * 32M to 64M: DMZ - no injection allowed | |
5627 | * 64M to 96M: injection range for leaf 1 | |
5628 | * 96M to 128M: DMZ - no injection allowed | |
5629 | * 128M to 160M: injection range for leaf 2 | |
5630 | * 160M to 192M: DMZ - no injection allowed | |
5631 | */ | |
34dc7c2f BB |
5632 | offset = ztest_random(fsize / (leaves << bshift)) * |
5633 | (leaves << bshift) + (leaf << bshift) + | |
5634 | (ztest_random(1ULL << (bshift - 1)) & -8ULL); | |
5635 | ||
2014c09f GM |
5636 | /* |
5637 | * Only allow damage to the labels at one end of the vdev. | |
5638 | * | |
5639 | * If all labels are damaged, the device will be totally | |
5640 | * inaccessible, which will result in loss of data, | |
5641 | * because we also damage (parts of) the other side of | |
5642 | * the mirror/raidz. | |
5643 | * | |
5644 | * Additionally, we will always have both an even and an | |
5645 | * odd label, so that we can handle crashes in the | |
5646 | * middle of vdev_config_sync(). | |
5647 | */ | |
5648 | if ((leaf & 1) == 0 && offset < VDEV_LABEL_START_SIZE) | |
5649 | continue; | |
5650 | ||
5651 | /* | |
5652 | * The two end labels are stored at the "end" of the disk, but | |
5653 | * the end of the disk (vdev_psize) is aligned to | |
5654 | * sizeof (vdev_label_t). | |
5655 | */ | |
5656 | uint64_t psize = P2ALIGN(fsize, sizeof (vdev_label_t)); | |
5657 | if ((leaf & 1) == 1 && | |
5658 | offset + sizeof (bad) > psize - VDEV_LABEL_END_SIZE) | |
34dc7c2f BB |
5659 | continue; |
5660 | ||
c242c188 | 5661 | mutex_enter(&ztest_vdev_lock); |
428870ff | 5662 | if (mirror_save != zs->zs_mirrors) { |
c242c188 | 5663 | mutex_exit(&ztest_vdev_lock); |
428870ff | 5664 | (void) close(fd); |
40b84e7a | 5665 | goto out; |
428870ff | 5666 | } |
34dc7c2f BB |
5667 | |
5668 | if (pwrite(fd, &bad, sizeof (bad), offset) != sizeof (bad)) | |
5669 | fatal(1, "can't inject bad word at 0x%llx in %s", | |
5670 | offset, pathrand); | |
428870ff | 5671 | |
c242c188 | 5672 | mutex_exit(&ztest_vdev_lock); |
428870ff | 5673 | |
c242c188 | 5674 | if (ztest_opts.zo_verbose >= 7) |
428870ff BB |
5675 | (void) printf("injected bad word into %s," |
5676 | " offset 0x%llx\n", pathrand, (u_longlong_t)offset); | |
34dc7c2f BB |
5677 | } |
5678 | ||
5679 | (void) close(fd); | |
40b84e7a BB |
5680 | out: |
5681 | umem_free(path0, MAXPATHLEN); | |
5682 | umem_free(pathrand, MAXPATHLEN); | |
34dc7c2f BB |
5683 | } |
5684 | ||
5685 | /* | |
428870ff | 5686 | * Verify that DDT repair works as expected. |
34dc7c2f BB |
5687 | */ |
5688 | void | |
428870ff | 5689 | ztest_ddt_repair(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 5690 | { |
428870ff | 5691 | ztest_shared_t *zs = ztest_shared; |
c242c188 | 5692 | spa_t *spa = ztest_spa; |
428870ff | 5693 | objset_t *os = zd->zd_os; |
40b84e7a | 5694 | ztest_od_t *od; |
428870ff BB |
5695 | uint64_t object, blocksize, txg, pattern, psize; |
5696 | enum zio_checksum checksum = spa_dedup_checksum(spa); | |
5697 | dmu_buf_t *db; | |
5698 | dmu_tx_t *tx; | |
a6255b7f | 5699 | abd_t *abd; |
428870ff BB |
5700 | blkptr_t blk; |
5701 | int copies = 2 * ZIO_DEDUPDITTO_MIN; | |
d6320ddb | 5702 | int i; |
34dc7c2f | 5703 | |
428870ff BB |
5704 | blocksize = ztest_random_blocksize(); |
5705 | blocksize = MIN(blocksize, 2048); /* because we write so many */ | |
34dc7c2f | 5706 | |
d1d7e268 | 5707 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 5708 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0, 0); |
34dc7c2f | 5709 | |
40b84e7a | 5710 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
d1d7e268 | 5711 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 5712 | return; |
40b84e7a | 5713 | } |
34dc7c2f BB |
5714 | |
5715 | /* | |
428870ff BB |
5716 | * Take the name lock as writer to prevent anyone else from changing |
5717 | * the pool and dataset properies we need to maintain during this test. | |
34dc7c2f | 5718 | */ |
7809eb8b | 5719 | (void) rw_wrlock(&ztest_name_lock); |
34dc7c2f | 5720 | |
428870ff BB |
5721 | if (ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_DEDUP, checksum, |
5722 | B_FALSE) != 0 || | |
5723 | ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_COPIES, 1, | |
5724 | B_FALSE) != 0) { | |
7809eb8b | 5725 | (void) rw_unlock(&ztest_name_lock); |
d1d7e268 | 5726 | umem_free(od, sizeof (ztest_od_t)); |
428870ff BB |
5727 | return; |
5728 | } | |
5729 | ||
546d32ca BB |
5730 | dmu_objset_stats_t dds; |
5731 | dsl_pool_config_enter(dmu_objset_pool(os), FTAG); | |
5732 | dmu_objset_fast_stat(os, &dds); | |
5733 | dsl_pool_config_exit(dmu_objset_pool(os), FTAG); | |
5734 | ||
428870ff BB |
5735 | object = od[0].od_object; |
5736 | blocksize = od[0].od_blocksize; | |
546d32ca | 5737 | pattern = zs->zs_guid ^ dds.dds_guid; |
428870ff BB |
5738 | |
5739 | ASSERT(object != 0); | |
5740 | ||
5741 | tx = dmu_tx_create(os); | |
5742 | dmu_tx_hold_write(tx, object, 0, copies * blocksize); | |
5743 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
5744 | if (txg == 0) { | |
7809eb8b | 5745 | (void) rw_unlock(&ztest_name_lock); |
d1d7e268 | 5746 | umem_free(od, sizeof (ztest_od_t)); |
428870ff BB |
5747 | return; |
5748 | } | |
34dc7c2f BB |
5749 | |
5750 | /* | |
428870ff | 5751 | * Write all the copies of our block. |
34dc7c2f | 5752 | */ |
d6320ddb | 5753 | for (i = 0; i < copies; i++) { |
428870ff | 5754 | uint64_t offset = i * blocksize; |
13fe0198 MA |
5755 | int error = dmu_buf_hold(os, object, offset, FTAG, &db, |
5756 | DMU_READ_NO_PREFETCH); | |
5757 | if (error != 0) { | |
5758 | fatal(B_FALSE, "dmu_buf_hold(%p, %llu, %llu) = %u", | |
5759 | os, (long long)object, (long long) offset, error); | |
5760 | } | |
428870ff BB |
5761 | ASSERT(db->db_offset == offset); |
5762 | ASSERT(db->db_size == blocksize); | |
5763 | ASSERT(ztest_pattern_match(db->db_data, db->db_size, pattern) || | |
5764 | ztest_pattern_match(db->db_data, db->db_size, 0ULL)); | |
5765 | dmu_buf_will_fill(db, tx); | |
5766 | ztest_pattern_set(db->db_data, db->db_size, pattern); | |
5767 | dmu_buf_rele(db, FTAG); | |
5768 | } | |
34dc7c2f | 5769 | |
428870ff BB |
5770 | dmu_tx_commit(tx); |
5771 | txg_wait_synced(spa_get_dsl(spa), txg); | |
34dc7c2f BB |
5772 | |
5773 | /* | |
428870ff | 5774 | * Find out what block we got. |
34dc7c2f | 5775 | */ |
03c6040b GW |
5776 | VERIFY0(dmu_buf_hold(os, object, 0, FTAG, &db, |
5777 | DMU_READ_NO_PREFETCH)); | |
428870ff BB |
5778 | blk = *((dmu_buf_impl_t *)db)->db_blkptr; |
5779 | dmu_buf_rele(db, FTAG); | |
34dc7c2f BB |
5780 | |
5781 | /* | |
428870ff | 5782 | * Damage the block. Dedup-ditto will save us when we read it later. |
34dc7c2f | 5783 | */ |
428870ff | 5784 | psize = BP_GET_PSIZE(&blk); |
a6255b7f DQ |
5785 | abd = abd_alloc_linear(psize, B_TRUE); |
5786 | ztest_pattern_set(abd_to_buf(abd), psize, ~pattern); | |
34dc7c2f | 5787 | |
428870ff | 5788 | (void) zio_wait(zio_rewrite(NULL, spa, 0, &blk, |
a6255b7f | 5789 | abd, psize, NULL, NULL, ZIO_PRIORITY_SYNC_WRITE, |
428870ff | 5790 | ZIO_FLAG_CANFAIL | ZIO_FLAG_INDUCE_DAMAGE, NULL)); |
34dc7c2f | 5791 | |
a6255b7f | 5792 | abd_free(abd); |
34dc7c2f | 5793 | |
7809eb8b | 5794 | (void) rw_unlock(&ztest_name_lock); |
d1d7e268 | 5795 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f BB |
5796 | } |
5797 | ||
34dc7c2f | 5798 | /* |
428870ff | 5799 | * Scrub the pool. |
34dc7c2f | 5800 | */ |
428870ff BB |
5801 | /* ARGSUSED */ |
5802 | void | |
5803 | ztest_scrub(ztest_ds_t *zd, uint64_t id) | |
34dc7c2f | 5804 | { |
c242c188 | 5805 | spa_t *spa = ztest_spa; |
34dc7c2f | 5806 | |
428870ff BB |
5807 | (void) spa_scan(spa, POOL_SCAN_SCRUB); |
5808 | (void) poll(NULL, 0, 100); /* wait a moment, then force a restart */ | |
5809 | (void) spa_scan(spa, POOL_SCAN_SCRUB); | |
5810 | } | |
34dc7c2f | 5811 | |
3541dc6d GA |
5812 | /* |
5813 | * Change the guid for the pool. | |
5814 | */ | |
5815 | /* ARGSUSED */ | |
5816 | void | |
5817 | ztest_reguid(ztest_ds_t *zd, uint64_t id) | |
5818 | { | |
c242c188 | 5819 | spa_t *spa = ztest_spa; |
3541dc6d | 5820 | uint64_t orig, load; |
3bc7e0fb | 5821 | int error; |
3541dc6d | 5822 | |
b33d668d OF |
5823 | if (ztest_opts.zo_mmp_test) |
5824 | return; | |
5825 | ||
3541dc6d GA |
5826 | orig = spa_guid(spa); |
5827 | load = spa_load_guid(spa); | |
3bc7e0fb | 5828 | |
7809eb8b | 5829 | (void) rw_wrlock(&ztest_name_lock); |
3bc7e0fb | 5830 | error = spa_change_guid(spa); |
7809eb8b | 5831 | (void) rw_unlock(&ztest_name_lock); |
3bc7e0fb GW |
5832 | |
5833 | if (error != 0) | |
3541dc6d GA |
5834 | return; |
5835 | ||
ea0b2538 | 5836 | if (ztest_opts.zo_verbose >= 4) { |
3541dc6d GA |
5837 | (void) printf("Changed guid old %llu -> %llu\n", |
5838 | (u_longlong_t)orig, (u_longlong_t)spa_guid(spa)); | |
5839 | } | |
5840 | ||
5841 | VERIFY3U(orig, !=, spa_guid(spa)); | |
5842 | VERIFY3U(load, ==, spa_load_guid(spa)); | |
5843 | } | |
5844 | ||
428870ff BB |
5845 | /* |
5846 | * Rename the pool to a different name and then rename it back. | |
5847 | */ | |
5848 | /* ARGSUSED */ | |
5849 | void | |
5850 | ztest_spa_rename(ztest_ds_t *zd, uint64_t id) | |
5851 | { | |
428870ff BB |
5852 | char *oldname, *newname; |
5853 | spa_t *spa; | |
34dc7c2f | 5854 | |
379ca9cf OF |
5855 | if (ztest_opts.zo_mmp_test) |
5856 | return; | |
5857 | ||
7809eb8b | 5858 | (void) rw_wrlock(&ztest_name_lock); |
34dc7c2f | 5859 | |
c242c188 | 5860 | oldname = ztest_opts.zo_pool; |
428870ff BB |
5861 | newname = umem_alloc(strlen(oldname) + 5, UMEM_NOFAIL); |
5862 | (void) strcpy(newname, oldname); | |
5863 | (void) strcat(newname, "_tmp"); | |
34dc7c2f BB |
5864 | |
5865 | /* | |
428870ff | 5866 | * Do the rename |
34dc7c2f | 5867 | */ |
428870ff | 5868 | VERIFY3U(0, ==, spa_rename(oldname, newname)); |
34dc7c2f BB |
5869 | |
5870 | /* | |
428870ff | 5871 | * Try to open it under the old name, which shouldn't exist |
34dc7c2f | 5872 | */ |
428870ff | 5873 | VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f BB |
5874 | |
5875 | /* | |
428870ff BB |
5876 | * Open it under the new name and make sure it's still the same spa_t. |
5877 | */ | |
5878 | VERIFY3U(0, ==, spa_open(newname, &spa, FTAG)); | |
34dc7c2f | 5879 | |
c242c188 | 5880 | ASSERT(spa == ztest_spa); |
428870ff | 5881 | spa_close(spa, FTAG); |
34dc7c2f BB |
5882 | |
5883 | /* | |
428870ff | 5884 | * Rename it back to the original |
34dc7c2f | 5885 | */ |
428870ff | 5886 | VERIFY3U(0, ==, spa_rename(newname, oldname)); |
34dc7c2f | 5887 | |
428870ff BB |
5888 | /* |
5889 | * Make sure it can still be opened | |
5890 | */ | |
5891 | VERIFY3U(0, ==, spa_open(oldname, &spa, FTAG)); | |
34dc7c2f | 5892 | |
c242c188 | 5893 | ASSERT(spa == ztest_spa); |
428870ff BB |
5894 | spa_close(spa, FTAG); |
5895 | ||
5896 | umem_free(newname, strlen(newname) + 1); | |
5897 | ||
7809eb8b | 5898 | (void) rw_unlock(&ztest_name_lock); |
34dc7c2f BB |
5899 | } |
5900 | ||
1eeb4562 JX |
5901 | void |
5902 | ztest_fletcher(ztest_ds_t *zd, uint64_t id) | |
5903 | { | |
5904 | hrtime_t end = gethrtime() + NANOSEC; | |
5905 | ||
5906 | while (gethrtime() <= end) { | |
5907 | int run_count = 100; | |
5908 | void *buf; | |
2fe36b0b | 5909 | struct abd *abd_data, *abd_meta; |
1eeb4562 JX |
5910 | uint32_t size; |
5911 | int *ptr; | |
5912 | int i; | |
5913 | zio_cksum_t zc_ref; | |
5914 | zio_cksum_t zc_ref_byteswap; | |
5915 | ||
5916 | size = ztest_random_blocksize(); | |
2fe36b0b | 5917 | |
1eeb4562 | 5918 | buf = umem_alloc(size, UMEM_NOFAIL); |
2fe36b0b DQ |
5919 | abd_data = abd_alloc(size, B_FALSE); |
5920 | abd_meta = abd_alloc(size, B_TRUE); | |
1eeb4562 JX |
5921 | |
5922 | for (i = 0, ptr = buf; i < size / sizeof (*ptr); i++, ptr++) | |
5923 | *ptr = ztest_random(UINT_MAX); | |
5924 | ||
2fe36b0b DQ |
5925 | abd_copy_from_buf_off(abd_data, buf, 0, size); |
5926 | abd_copy_from_buf_off(abd_meta, buf, 0, size); | |
5927 | ||
1eeb4562 | 5928 | VERIFY0(fletcher_4_impl_set("scalar")); |
3c67d83a TH |
5929 | fletcher_4_native(buf, size, NULL, &zc_ref); |
5930 | fletcher_4_byteswap(buf, size, NULL, &zc_ref_byteswap); | |
1eeb4562 JX |
5931 | |
5932 | VERIFY0(fletcher_4_impl_set("cycle")); | |
5933 | while (run_count-- > 0) { | |
5934 | zio_cksum_t zc; | |
5935 | zio_cksum_t zc_byteswap; | |
5936 | ||
3c67d83a TH |
5937 | fletcher_4_byteswap(buf, size, NULL, &zc_byteswap); |
5938 | fletcher_4_native(buf, size, NULL, &zc); | |
1eeb4562 JX |
5939 | |
5940 | VERIFY0(bcmp(&zc, &zc_ref, sizeof (zc))); | |
5941 | VERIFY0(bcmp(&zc_byteswap, &zc_ref_byteswap, | |
5942 | sizeof (zc_byteswap))); | |
2fe36b0b DQ |
5943 | |
5944 | /* Test ABD - data */ | |
5945 | abd_fletcher_4_byteswap(abd_data, size, NULL, | |
5946 | &zc_byteswap); | |
5947 | abd_fletcher_4_native(abd_data, size, NULL, &zc); | |
5948 | ||
5949 | VERIFY0(bcmp(&zc, &zc_ref, sizeof (zc))); | |
5950 | VERIFY0(bcmp(&zc_byteswap, &zc_ref_byteswap, | |
5951 | sizeof (zc_byteswap))); | |
5952 | ||
5953 | /* Test ABD - metadata */ | |
5954 | abd_fletcher_4_byteswap(abd_meta, size, NULL, | |
5955 | &zc_byteswap); | |
5956 | abd_fletcher_4_native(abd_meta, size, NULL, &zc); | |
5957 | ||
5958 | VERIFY0(bcmp(&zc, &zc_ref, sizeof (zc))); | |
5959 | VERIFY0(bcmp(&zc_byteswap, &zc_ref_byteswap, | |
5960 | sizeof (zc_byteswap))); | |
5961 | ||
1eeb4562 JX |
5962 | } |
5963 | ||
5964 | umem_free(buf, size); | |
2fe36b0b DQ |
5965 | abd_free(abd_data); |
5966 | abd_free(abd_meta); | |
1eeb4562 JX |
5967 | } |
5968 | } | |
5969 | ||
37f520db GN |
5970 | void |
5971 | ztest_fletcher_incr(ztest_ds_t *zd, uint64_t id) | |
5972 | { | |
5973 | void *buf; | |
5974 | size_t size; | |
5975 | int *ptr; | |
5976 | int i; | |
5977 | zio_cksum_t zc_ref; | |
5978 | zio_cksum_t zc_ref_bswap; | |
5979 | ||
5980 | hrtime_t end = gethrtime() + NANOSEC; | |
5981 | ||
5982 | while (gethrtime() <= end) { | |
5983 | int run_count = 100; | |
5984 | ||
5985 | size = ztest_random_blocksize(); | |
5986 | buf = umem_alloc(size, UMEM_NOFAIL); | |
5987 | ||
5988 | for (i = 0, ptr = buf; i < size / sizeof (*ptr); i++, ptr++) | |
5989 | *ptr = ztest_random(UINT_MAX); | |
5990 | ||
5991 | VERIFY0(fletcher_4_impl_set("scalar")); | |
5992 | fletcher_4_native(buf, size, NULL, &zc_ref); | |
5993 | fletcher_4_byteswap(buf, size, NULL, &zc_ref_bswap); | |
5994 | ||
5995 | VERIFY0(fletcher_4_impl_set("cycle")); | |
5996 | ||
5997 | while (run_count-- > 0) { | |
5998 | zio_cksum_t zc; | |
5999 | zio_cksum_t zc_bswap; | |
6000 | size_t pos = 0; | |
6001 | ||
6002 | ZIO_SET_CHECKSUM(&zc, 0, 0, 0, 0); | |
6003 | ZIO_SET_CHECKSUM(&zc_bswap, 0, 0, 0, 0); | |
6004 | ||
6005 | while (pos < size) { | |
6006 | size_t inc = 64 * ztest_random(size / 67); | |
6007 | /* sometimes add few bytes to test non-simd */ | |
6008 | if (ztest_random(100) < 10) | |
6009 | inc += P2ALIGN(ztest_random(64), | |
6010 | sizeof (uint32_t)); | |
6011 | ||
6012 | if (inc > (size - pos)) | |
6013 | inc = size - pos; | |
6014 | ||
6015 | fletcher_4_incremental_native(buf + pos, inc, | |
6016 | &zc); | |
6017 | fletcher_4_incremental_byteswap(buf + pos, inc, | |
6018 | &zc_bswap); | |
6019 | ||
6020 | pos += inc; | |
6021 | } | |
6022 | ||
6023 | VERIFY3U(pos, ==, size); | |
6024 | ||
6025 | VERIFY(ZIO_CHECKSUM_EQUAL(zc, zc_ref)); | |
6026 | VERIFY(ZIO_CHECKSUM_EQUAL(zc_bswap, zc_ref_bswap)); | |
6027 | ||
6028 | /* | |
6029 | * verify if incremental on the whole buffer is | |
6030 | * equivalent to non-incremental version | |
6031 | */ | |
6032 | ZIO_SET_CHECKSUM(&zc, 0, 0, 0, 0); | |
6033 | ZIO_SET_CHECKSUM(&zc_bswap, 0, 0, 0, 0); | |
6034 | ||
6035 | fletcher_4_incremental_native(buf, size, &zc); | |
6036 | fletcher_4_incremental_byteswap(buf, size, &zc_bswap); | |
6037 | ||
6038 | VERIFY(ZIO_CHECKSUM_EQUAL(zc, zc_ref)); | |
6039 | VERIFY(ZIO_CHECKSUM_EQUAL(zc_bswap, zc_ref_bswap)); | |
6040 | } | |
6041 | ||
6042 | umem_free(buf, size); | |
6043 | } | |
6044 | } | |
6045 | ||
90aa094d CC |
6046 | static int |
6047 | ztest_check_path(char *path) | |
6048 | { | |
6049 | struct stat s; | |
6050 | /* return true on success */ | |
6051 | return (!stat(path, &s)); | |
6052 | } | |
6053 | ||
6054 | static void | |
6055 | ztest_get_zdb_bin(char *bin, int len) | |
6056 | { | |
6057 | char *zdb_path; | |
6058 | /* | |
6059 | * Try to use ZDB_PATH and in-tree zdb path. If not successful, just | |
6060 | * let popen to search through PATH. | |
6061 | */ | |
6062 | if ((zdb_path = getenv("ZDB_PATH"))) { | |
6063 | strlcpy(bin, zdb_path, len); /* In env */ | |
6064 | if (!ztest_check_path(bin)) { | |
6065 | ztest_dump_core = 0; | |
6066 | fatal(1, "invalid ZDB_PATH '%s'", bin); | |
6067 | } | |
6068 | return; | |
6069 | } | |
6070 | ||
6071 | VERIFY(realpath(getexecname(), bin) != NULL); | |
6072 | if (strstr(bin, "/ztest/")) { | |
6073 | strstr(bin, "/ztest/")[0] = '\0'; /* In-tree */ | |
6074 | strcat(bin, "/zdb/zdb"); | |
6075 | if (ztest_check_path(bin)) | |
6076 | return; | |
6077 | } | |
6078 | strcpy(bin, "zdb"); | |
6079 | } | |
6080 | ||
428870ff BB |
6081 | /* |
6082 | * Verify pool integrity by running zdb. | |
6083 | */ | |
34dc7c2f | 6084 | static void |
428870ff | 6085 | ztest_run_zdb(char *pool) |
34dc7c2f BB |
6086 | { |
6087 | int status; | |
34dc7c2f | 6088 | char *bin; |
0e8d1b2d BB |
6089 | char *zdb; |
6090 | char *zbuf; | |
90aa094d | 6091 | const int len = MAXPATHLEN + MAXNAMELEN + 20; |
34dc7c2f BB |
6092 | FILE *fp; |
6093 | ||
90aa094d CC |
6094 | bin = umem_alloc(len, UMEM_NOFAIL); |
6095 | zdb = umem_alloc(len, UMEM_NOFAIL); | |
0e8d1b2d | 6096 | zbuf = umem_alloc(1024, UMEM_NOFAIL); |
34dc7c2f | 6097 | |
90aa094d | 6098 | ztest_get_zdb_bin(bin, len); |
0e8d1b2d BB |
6099 | |
6100 | (void) sprintf(zdb, | |
456079d4 | 6101 | "%s -bcc%s%s -G -d -U %s %s", |
0e8d1b2d | 6102 | bin, |
c242c188 CS |
6103 | ztest_opts.zo_verbose >= 3 ? "s" : "", |
6104 | ztest_opts.zo_verbose >= 4 ? "v" : "", | |
428870ff | 6105 | spa_config_path, |
b128c09f | 6106 | pool); |
34dc7c2f | 6107 | |
c242c188 | 6108 | if (ztest_opts.zo_verbose >= 5) |
34dc7c2f BB |
6109 | (void) printf("Executing %s\n", strstr(zdb, "zdb ")); |
6110 | ||
6111 | fp = popen(zdb, "r"); | |
6112 | ||
6aec1cd5 | 6113 | while (fgets(zbuf, 1024, fp) != NULL) |
c242c188 | 6114 | if (ztest_opts.zo_verbose >= 3) |
34dc7c2f BB |
6115 | (void) printf("%s", zbuf); |
6116 | ||
6117 | status = pclose(fp); | |
6118 | ||
6119 | if (status == 0) | |
0e8d1b2d | 6120 | goto out; |
34dc7c2f BB |
6121 | |
6122 | ztest_dump_core = 0; | |
6123 | if (WIFEXITED(status)) | |
6124 | fatal(0, "'%s' exit code %d", zdb, WEXITSTATUS(status)); | |
6125 | else | |
6126 | fatal(0, "'%s' died with signal %d", zdb, WTERMSIG(status)); | |
0e8d1b2d | 6127 | out: |
90aa094d CC |
6128 | umem_free(bin, len); |
6129 | umem_free(zdb, len); | |
0e8d1b2d | 6130 | umem_free(zbuf, 1024); |
34dc7c2f BB |
6131 | } |
6132 | ||
6133 | static void | |
6134 | ztest_walk_pool_directory(char *header) | |
6135 | { | |
6136 | spa_t *spa = NULL; | |
6137 | ||
c242c188 | 6138 | if (ztest_opts.zo_verbose >= 6) |
34dc7c2f BB |
6139 | (void) printf("%s\n", header); |
6140 | ||
6141 | mutex_enter(&spa_namespace_lock); | |
6142 | while ((spa = spa_next(spa)) != NULL) | |
c242c188 | 6143 | if (ztest_opts.zo_verbose >= 6) |
34dc7c2f BB |
6144 | (void) printf("\t%s\n", spa_name(spa)); |
6145 | mutex_exit(&spa_namespace_lock); | |
6146 | } | |
6147 | ||
6148 | static void | |
6149 | ztest_spa_import_export(char *oldname, char *newname) | |
6150 | { | |
fb5f0bc8 | 6151 | nvlist_t *config, *newconfig; |
34dc7c2f BB |
6152 | uint64_t pool_guid; |
6153 | spa_t *spa; | |
13fe0198 | 6154 | int error; |
34dc7c2f | 6155 | |
c242c188 | 6156 | if (ztest_opts.zo_verbose >= 4) { |
34dc7c2f BB |
6157 | (void) printf("import/export: old = %s, new = %s\n", |
6158 | oldname, newname); | |
6159 | } | |
6160 | ||
6161 | /* | |
6162 | * Clean up from previous runs. | |
6163 | */ | |
6164 | (void) spa_destroy(newname); | |
6165 | ||
6166 | /* | |
6167 | * Get the pool's configuration and guid. | |
6168 | */ | |
428870ff | 6169 | VERIFY3U(0, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f | 6170 | |
fb5f0bc8 BB |
6171 | /* |
6172 | * Kick off a scrub to tickle scrub/export races. | |
6173 | */ | |
6174 | if (ztest_random(2) == 0) | |
428870ff | 6175 | (void) spa_scan(spa, POOL_SCAN_SCRUB); |
fb5f0bc8 | 6176 | |
34dc7c2f BB |
6177 | pool_guid = spa_guid(spa); |
6178 | spa_close(spa, FTAG); | |
6179 | ||
6180 | ztest_walk_pool_directory("pools before export"); | |
6181 | ||
6182 | /* | |
6183 | * Export it. | |
6184 | */ | |
428870ff | 6185 | VERIFY3U(0, ==, spa_export(oldname, &config, B_FALSE, B_FALSE)); |
34dc7c2f BB |
6186 | |
6187 | ztest_walk_pool_directory("pools after export"); | |
6188 | ||
fb5f0bc8 BB |
6189 | /* |
6190 | * Try to import it. | |
6191 | */ | |
6192 | newconfig = spa_tryimport(config); | |
6193 | ASSERT(newconfig != NULL); | |
6194 | nvlist_free(newconfig); | |
6195 | ||
34dc7c2f BB |
6196 | /* |
6197 | * Import it under the new name. | |
6198 | */ | |
13fe0198 MA |
6199 | error = spa_import(newname, config, NULL, 0); |
6200 | if (error != 0) { | |
6201 | dump_nvlist(config, 0); | |
6202 | fatal(B_FALSE, "couldn't import pool %s as %s: error %u", | |
6203 | oldname, newname, error); | |
6204 | } | |
34dc7c2f BB |
6205 | |
6206 | ztest_walk_pool_directory("pools after import"); | |
6207 | ||
6208 | /* | |
6209 | * Try to import it again -- should fail with EEXIST. | |
6210 | */ | |
572e2857 | 6211 | VERIFY3U(EEXIST, ==, spa_import(newname, config, NULL, 0)); |
34dc7c2f BB |
6212 | |
6213 | /* | |
6214 | * Try to import it under a different name -- should fail with EEXIST. | |
6215 | */ | |
572e2857 | 6216 | VERIFY3U(EEXIST, ==, spa_import(oldname, config, NULL, 0)); |
34dc7c2f BB |
6217 | |
6218 | /* | |
6219 | * Verify that the pool is no longer visible under the old name. | |
6220 | */ | |
428870ff | 6221 | VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f BB |
6222 | |
6223 | /* | |
6224 | * Verify that we can open and close the pool using the new name. | |
6225 | */ | |
428870ff | 6226 | VERIFY3U(0, ==, spa_open(newname, &spa, FTAG)); |
34dc7c2f BB |
6227 | ASSERT(pool_guid == spa_guid(spa)); |
6228 | spa_close(spa, FTAG); | |
6229 | ||
6230 | nvlist_free(config); | |
6231 | } | |
6232 | ||
fb5f0bc8 BB |
6233 | static void |
6234 | ztest_resume(spa_t *spa) | |
6235 | { | |
c242c188 | 6236 | if (spa_suspended(spa) && ztest_opts.zo_verbose >= 6) |
428870ff BB |
6237 | (void) printf("resuming from suspended state\n"); |
6238 | spa_vdev_state_enter(spa, SCL_NONE); | |
6239 | vdev_clear(spa, NULL); | |
6240 | (void) spa_vdev_state_exit(spa, NULL, 0); | |
6241 | (void) zio_resume(spa); | |
fb5f0bc8 BB |
6242 | } |
6243 | ||
c25b8f99 | 6244 | static void |
fb5f0bc8 | 6245 | ztest_resume_thread(void *arg) |
34dc7c2f | 6246 | { |
b128c09f | 6247 | spa_t *spa = arg; |
34dc7c2f BB |
6248 | |
6249 | while (!ztest_exiting) { | |
428870ff BB |
6250 | if (spa_suspended(spa)) |
6251 | ztest_resume(spa); | |
6252 | (void) poll(NULL, 0, 100); | |
d3c2ae1c GW |
6253 | |
6254 | /* | |
6255 | * Periodically change the zfs_compressed_arc_enabled setting. | |
6256 | */ | |
6257 | if (ztest_random(10) == 0) | |
6258 | zfs_compressed_arc_enabled = ztest_random(2); | |
a6255b7f DQ |
6259 | |
6260 | /* | |
6261 | * Periodically change the zfs_abd_scatter_enabled setting. | |
6262 | */ | |
6263 | if (ztest_random(10) == 0) | |
6264 | zfs_abd_scatter_enabled = ztest_random(2); | |
34dc7c2f | 6265 | } |
34dc7c2f | 6266 | |
1e33ac1e | 6267 | thread_exit(); |
1e33ac1e | 6268 | } |
428870ff | 6269 | |
1e33ac1e | 6270 | static void |
8fb1ede1 | 6271 | ztest_deadman_thread(void *arg) |
1e33ac1e | 6272 | { |
8fb1ede1 BB |
6273 | ztest_shared_t *zs = arg; |
6274 | spa_t *spa = ztest_spa; | |
6275 | hrtime_t delta, overdue, total = 0; | |
6276 | ||
6277 | for (;;) { | |
6278 | delta = zs->zs_thread_stop - zs->zs_thread_start + | |
6279 | MSEC2NSEC(zfs_deadman_synctime_ms); | |
6280 | ||
6281 | (void) poll(NULL, 0, (int)NSEC2MSEC(delta)); | |
6282 | ||
6283 | /* | |
6284 | * If the pool is suspended then fail immediately. Otherwise, | |
6285 | * check to see if the pool is making any progress. If | |
6286 | * vdev_deadman() discovers that there hasn't been any recent | |
6287 | * I/Os then it will end up aborting the tests. | |
6288 | */ | |
6289 | if (spa_suspended(spa) || spa->spa_root_vdev == NULL) { | |
6290 | fatal(0, "aborting test after %llu seconds because " | |
6291 | "pool has transitioned to a suspended state.", | |
6292 | zfs_deadman_synctime_ms / 1000); | |
6293 | } | |
6294 | vdev_deadman(spa->spa_root_vdev, FTAG); | |
6295 | ||
6296 | /* | |
6297 | * If the process doesn't complete within a grace period of | |
6298 | * zfs_deadman_synctime_ms over the expected finish time, | |
6299 | * then it may be hung and is terminated. | |
6300 | */ | |
6301 | overdue = zs->zs_proc_stop + MSEC2NSEC(zfs_deadman_synctime_ms); | |
6302 | total += zfs_deadman_synctime_ms / 1000; | |
6303 | if (gethrtime() > overdue) { | |
6304 | fatal(0, "aborting test after %llu seconds because " | |
6305 | "the process is overdue for termination.", total); | |
6306 | } | |
6307 | ||
6308 | (void) printf("ztest has been running for %lld seconds\n", | |
6309 | total); | |
6310 | } | |
428870ff BB |
6311 | } |
6312 | ||
6313 | static void | |
c242c188 | 6314 | ztest_execute(int test, ztest_info_t *zi, uint64_t id) |
428870ff | 6315 | { |
c242c188 CS |
6316 | ztest_ds_t *zd = &ztest_ds[id % ztest_opts.zo_datasets]; |
6317 | ztest_shared_callstate_t *zc = ZTEST_GET_SHARED_CALLSTATE(test); | |
428870ff | 6318 | hrtime_t functime = gethrtime(); |
d6320ddb | 6319 | int i; |
428870ff | 6320 | |
d6320ddb | 6321 | for (i = 0; i < zi->zi_iters; i++) |
428870ff BB |
6322 | zi->zi_func(zd, id); |
6323 | ||
6324 | functime = gethrtime() - functime; | |
6325 | ||
c242c188 CS |
6326 | atomic_add_64(&zc->zc_count, 1); |
6327 | atomic_add_64(&zc->zc_time, functime); | |
428870ff | 6328 | |
fdc5d982 | 6329 | if (ztest_opts.zo_verbose >= 4) |
428870ff | 6330 | (void) printf("%6.2f sec in %s\n", |
fdc5d982 | 6331 | (double)functime / NANOSEC, zi->zi_funcname); |
428870ff BB |
6332 | } |
6333 | ||
c25b8f99 | 6334 | static void |
34dc7c2f BB |
6335 | ztest_thread(void *arg) |
6336 | { | |
c242c188 | 6337 | int rand; |
428870ff | 6338 | uint64_t id = (uintptr_t)arg; |
34dc7c2f | 6339 | ztest_shared_t *zs = ztest_shared; |
428870ff BB |
6340 | uint64_t call_next; |
6341 | hrtime_t now; | |
34dc7c2f | 6342 | ztest_info_t *zi; |
c242c188 | 6343 | ztest_shared_callstate_t *zc; |
34dc7c2f | 6344 | |
428870ff | 6345 | while ((now = gethrtime()) < zs->zs_thread_stop) { |
34dc7c2f BB |
6346 | /* |
6347 | * See if it's time to force a crash. | |
6348 | */ | |
428870ff BB |
6349 | if (now > zs->zs_thread_kill) |
6350 | ztest_kill(zs); | |
34dc7c2f BB |
6351 | |
6352 | /* | |
428870ff | 6353 | * If we're getting ENOSPC with some regularity, stop. |
34dc7c2f | 6354 | */ |
428870ff BB |
6355 | if (zs->zs_enospc_count > 10) |
6356 | break; | |
34dc7c2f BB |
6357 | |
6358 | /* | |
428870ff | 6359 | * Pick a random function to execute. |
34dc7c2f | 6360 | */ |
c242c188 CS |
6361 | rand = ztest_random(ZTEST_FUNCS); |
6362 | zi = &ztest_info[rand]; | |
6363 | zc = ZTEST_GET_SHARED_CALLSTATE(rand); | |
6364 | call_next = zc->zc_next; | |
34dc7c2f | 6365 | |
428870ff | 6366 | if (now >= call_next && |
c242c188 CS |
6367 | atomic_cas_64(&zc->zc_next, call_next, call_next + |
6368 | ztest_random(2 * zi->zi_interval[0] + 1)) == call_next) { | |
6369 | ztest_execute(rand, zi, id); | |
6370 | } | |
428870ff | 6371 | } |
34dc7c2f | 6372 | |
1e33ac1e | 6373 | thread_exit(); |
428870ff | 6374 | } |
34dc7c2f | 6375 | |
428870ff BB |
6376 | static void |
6377 | ztest_dataset_name(char *dsname, char *pool, int d) | |
6378 | { | |
eca7b760 | 6379 | (void) snprintf(dsname, ZFS_MAX_DATASET_NAME_LEN, "%s/ds_%d", pool, d); |
428870ff | 6380 | } |
34dc7c2f | 6381 | |
428870ff | 6382 | static void |
c242c188 | 6383 | ztest_dataset_destroy(int d) |
428870ff | 6384 | { |
eca7b760 | 6385 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
d6320ddb | 6386 | int t; |
34dc7c2f | 6387 | |
c242c188 | 6388 | ztest_dataset_name(name, ztest_opts.zo_pool, d); |
34dc7c2f | 6389 | |
c242c188 | 6390 | if (ztest_opts.zo_verbose >= 3) |
428870ff | 6391 | (void) printf("Destroying %s to free up space\n", name); |
34dc7c2f | 6392 | |
428870ff BB |
6393 | /* |
6394 | * Cleanup any non-standard clones and snapshots. In general, | |
6395 | * ztest thread t operates on dataset (t % zopt_datasets), | |
6396 | * so there may be more than one thing to clean up. | |
6397 | */ | |
c242c188 CS |
6398 | for (t = d; t < ztest_opts.zo_threads; |
6399 | t += ztest_opts.zo_datasets) | |
428870ff BB |
6400 | ztest_dsl_dataset_cleanup(name, t); |
6401 | ||
6402 | (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL, | |
6403 | DS_FIND_SNAPSHOTS | DS_FIND_CHILDREN); | |
6404 | } | |
6405 | ||
6406 | static void | |
6407 | ztest_dataset_dirobj_verify(ztest_ds_t *zd) | |
6408 | { | |
6409 | uint64_t usedobjs, dirobjs, scratch; | |
6410 | ||
6411 | /* | |
6412 | * ZTEST_DIROBJ is the object directory for the entire dataset. | |
6413 | * Therefore, the number of objects in use should equal the | |
6414 | * number of ZTEST_DIROBJ entries, +1 for ZTEST_DIROBJ itself. | |
6415 | * If not, we have an object leak. | |
6416 | * | |
6417 | * Note that we can only check this in ztest_dataset_open(), | |
6418 | * when the open-context and syncing-context values agree. | |
6419 | * That's because zap_count() returns the open-context value, | |
6420 | * while dmu_objset_space() returns the rootbp fill count. | |
6421 | */ | |
6422 | VERIFY3U(0, ==, zap_count(zd->zd_os, ZTEST_DIROBJ, &dirobjs)); | |
6423 | dmu_objset_space(zd->zd_os, &scratch, &scratch, &usedobjs, &scratch); | |
6424 | ASSERT3U(dirobjs + 1, ==, usedobjs); | |
6425 | } | |
6426 | ||
6427 | static int | |
c242c188 | 6428 | ztest_dataset_open(int d) |
428870ff | 6429 | { |
c242c188 CS |
6430 | ztest_ds_t *zd = &ztest_ds[d]; |
6431 | uint64_t committed_seq = ZTEST_GET_SHARED_DS(d)->zd_seq; | |
428870ff BB |
6432 | objset_t *os; |
6433 | zilog_t *zilog; | |
eca7b760 | 6434 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
6435 | int error; |
6436 | ||
c242c188 | 6437 | ztest_dataset_name(name, ztest_opts.zo_pool, d); |
428870ff | 6438 | |
7809eb8b | 6439 | (void) rw_rdlock(&ztest_name_lock); |
428870ff BB |
6440 | |
6441 | error = ztest_dataset_create(name); | |
6442 | if (error == ENOSPC) { | |
7809eb8b | 6443 | (void) rw_unlock(&ztest_name_lock); |
428870ff BB |
6444 | ztest_record_enospc(FTAG); |
6445 | return (error); | |
34dc7c2f | 6446 | } |
428870ff | 6447 | ASSERT(error == 0 || error == EEXIST); |
34dc7c2f | 6448 | |
4807c0ba TC |
6449 | VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, |
6450 | B_TRUE, zd, &os)); | |
7809eb8b | 6451 | (void) rw_unlock(&ztest_name_lock); |
428870ff | 6452 | |
c242c188 | 6453 | ztest_zd_init(zd, ZTEST_GET_SHARED_DS(d), os); |
428870ff BB |
6454 | |
6455 | zilog = zd->zd_zilog; | |
6456 | ||
6457 | if (zilog->zl_header->zh_claim_lr_seq != 0 && | |
6458 | zilog->zl_header->zh_claim_lr_seq < committed_seq) | |
6459 | fatal(0, "missing log records: claimed %llu < committed %llu", | |
6460 | zilog->zl_header->zh_claim_lr_seq, committed_seq); | |
6461 | ||
6462 | ztest_dataset_dirobj_verify(zd); | |
6463 | ||
6464 | zil_replay(os, zd, ztest_replay_vector); | |
6465 | ||
6466 | ztest_dataset_dirobj_verify(zd); | |
6467 | ||
c242c188 | 6468 | if (ztest_opts.zo_verbose >= 6) |
428870ff BB |
6469 | (void) printf("%s replay %llu blocks, %llu records, seq %llu\n", |
6470 | zd->zd_name, | |
6471 | (u_longlong_t)zilog->zl_parse_blk_count, | |
6472 | (u_longlong_t)zilog->zl_parse_lr_count, | |
6473 | (u_longlong_t)zilog->zl_replaying_seq); | |
6474 | ||
6475 | zilog = zil_open(os, ztest_get_data); | |
6476 | ||
6477 | if (zilog->zl_replaying_seq != 0 && | |
6478 | zilog->zl_replaying_seq < committed_seq) | |
6479 | fatal(0, "missing log records: replayed %llu < committed %llu", | |
6480 | zilog->zl_replaying_seq, committed_seq); | |
6481 | ||
6482 | return (0); | |
6483 | } | |
6484 | ||
6485 | static void | |
c242c188 | 6486 | ztest_dataset_close(int d) |
428870ff | 6487 | { |
c242c188 | 6488 | ztest_ds_t *zd = &ztest_ds[d]; |
428870ff BB |
6489 | |
6490 | zil_close(zd->zd_zilog); | |
4807c0ba | 6491 | txg_wait_synced(spa_get_dsl(zd->zd_os->os_spa), 0); |
b5256303 | 6492 | dmu_objset_disown(zd->zd_os, B_TRUE, zd); |
428870ff BB |
6493 | |
6494 | ztest_zd_fini(zd); | |
34dc7c2f BB |
6495 | } |
6496 | ||
6497 | /* | |
6498 | * Kick off threads to run tests on all datasets in parallel. | |
6499 | */ | |
6500 | static void | |
428870ff | 6501 | ztest_run(ztest_shared_t *zs) |
34dc7c2f | 6502 | { |
34dc7c2f | 6503 | spa_t *spa; |
3541dc6d | 6504 | objset_t *os; |
1e33ac1e | 6505 | kthread_t *resume_thread; |
c25b8f99 | 6506 | kthread_t **run_threads; |
1e33ac1e | 6507 | uint64_t object; |
428870ff | 6508 | int error; |
d6320ddb | 6509 | int t, d; |
b128c09f BB |
6510 | |
6511 | ztest_exiting = B_FALSE; | |
34dc7c2f | 6512 | |
34dc7c2f | 6513 | /* |
428870ff | 6514 | * Initialize parent/child shared state. |
34dc7c2f | 6515 | */ |
c242c188 | 6516 | mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); |
7809eb8b | 6517 | VERIFY(rwlock_init(&ztest_name_lock, USYNC_THREAD, NULL) == 0); |
34dc7c2f | 6518 | |
428870ff | 6519 | zs->zs_thread_start = gethrtime(); |
c242c188 CS |
6520 | zs->zs_thread_stop = |
6521 | zs->zs_thread_start + ztest_opts.zo_passtime * NANOSEC; | |
428870ff BB |
6522 | zs->zs_thread_stop = MIN(zs->zs_thread_stop, zs->zs_proc_stop); |
6523 | zs->zs_thread_kill = zs->zs_thread_stop; | |
c242c188 CS |
6524 | if (ztest_random(100) < ztest_opts.zo_killrate) { |
6525 | zs->zs_thread_kill -= | |
6526 | ztest_random(ztest_opts.zo_passtime * NANOSEC); | |
6527 | } | |
34dc7c2f | 6528 | |
1e33ac1e | 6529 | mutex_init(&zcl.zcl_callbacks_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 6530 | |
428870ff BB |
6531 | list_create(&zcl.zcl_callbacks, sizeof (ztest_cb_data_t), |
6532 | offsetof(ztest_cb_data_t, zcd_node)); | |
34dc7c2f BB |
6533 | |
6534 | /* | |
b128c09f | 6535 | * Open our pool. |
34dc7c2f | 6536 | */ |
428870ff | 6537 | kernel_init(FREAD | FWRITE); |
13fe0198 | 6538 | VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
6d974228 | 6539 | spa->spa_debug = B_TRUE; |
080b3100 | 6540 | metaslab_preload_limit = ztest_random(20) + 1; |
c242c188 | 6541 | ztest_spa = spa; |
428870ff | 6542 | |
546d32ca | 6543 | dmu_objset_stats_t dds; |
4807c0ba | 6544 | VERIFY0(ztest_dmu_objset_own(ztest_opts.zo_pool, |
b5256303 | 6545 | DMU_OST_ANY, B_TRUE, B_TRUE, FTAG, &os)); |
546d32ca BB |
6546 | dsl_pool_config_enter(dmu_objset_pool(os), FTAG); |
6547 | dmu_objset_fast_stat(os, &dds); | |
6548 | dsl_pool_config_exit(dmu_objset_pool(os), FTAG); | |
6549 | zs->zs_guid = dds.dds_guid; | |
b5256303 | 6550 | dmu_objset_disown(os, B_TRUE, FTAG); |
3541dc6d | 6551 | |
428870ff | 6552 | spa->spa_dedup_ditto = 2 * ZIO_DEDUPDITTO_MIN; |
34dc7c2f BB |
6553 | |
6554 | /* | |
b128c09f | 6555 | * Create a thread to periodically resume suspended I/O. |
34dc7c2f | 6556 | */ |
c25b8f99 BB |
6557 | resume_thread = thread_create(NULL, 0, ztest_resume_thread, |
6558 | spa, 0, NULL, TS_RUN | TS_JOINABLE, defclsyspri); | |
34dc7c2f | 6559 | |
428870ff | 6560 | /* |
8fb1ede1 | 6561 | * Create a deadman thread and set to panic if we hang. |
428870ff | 6562 | */ |
8fb1ede1 BB |
6563 | (void) thread_create(NULL, 0, ztest_deadman_thread, |
6564 | zs, 0, NULL, TS_RUN | TS_JOINABLE, defclsyspri); | |
6565 | ||
6566 | spa->spa_deadman_failmode = ZIO_FAILURE_MODE_PANIC; | |
428870ff | 6567 | |
34dc7c2f BB |
6568 | /* |
6569 | * Verify that we can safely inquire about about any object, | |
6570 | * whether it's allocated or not. To make it interesting, | |
6571 | * we probe a 5-wide window around each power of two. | |
6572 | * This hits all edge cases, including zero and the max. | |
6573 | */ | |
d6320ddb BB |
6574 | for (t = 0; t < 64; t++) { |
6575 | for (d = -5; d <= 5; d++) { | |
34dc7c2f BB |
6576 | error = dmu_object_info(spa->spa_meta_objset, |
6577 | (1ULL << t) + d, NULL); | |
6578 | ASSERT(error == 0 || error == ENOENT || | |
6579 | error == EINVAL); | |
6580 | } | |
6581 | } | |
6582 | ||
6583 | /* | |
428870ff | 6584 | * If we got any ENOSPC errors on the previous run, destroy something. |
34dc7c2f | 6585 | */ |
428870ff | 6586 | if (zs->zs_enospc_count != 0) { |
c242c188 CS |
6587 | int d = ztest_random(ztest_opts.zo_datasets); |
6588 | ztest_dataset_destroy(d); | |
428870ff | 6589 | } |
34dc7c2f BB |
6590 | zs->zs_enospc_count = 0; |
6591 | ||
c25b8f99 | 6592 | run_threads = umem_zalloc(ztest_opts.zo_threads * sizeof (kthread_t *), |
c242c188 | 6593 | UMEM_NOFAIL); |
34dc7c2f | 6594 | |
c242c188 | 6595 | if (ztest_opts.zo_verbose >= 4) |
34dc7c2f BB |
6596 | (void) printf("starting main threads...\n"); |
6597 | ||
428870ff BB |
6598 | /* |
6599 | * Kick off all the tests that run in parallel. | |
6600 | */ | |
c242c188 | 6601 | for (t = 0; t < ztest_opts.zo_threads; t++) { |
c25b8f99 BB |
6602 | if (t < ztest_opts.zo_datasets && ztest_dataset_open(t) != 0) { |
6603 | umem_free(run_threads, ztest_opts.zo_threads * | |
6604 | sizeof (kthread_t *)); | |
428870ff | 6605 | return; |
06cf4d98 | 6606 | } |
1e33ac1e | 6607 | |
c25b8f99 BB |
6608 | run_threads[t] = thread_create(NULL, 0, ztest_thread, |
6609 | (void *)(uintptr_t)t, 0, NULL, TS_RUN | TS_JOINABLE, | |
6610 | defclsyspri); | |
34dc7c2f BB |
6611 | } |
6612 | ||
428870ff BB |
6613 | /* |
6614 | * Wait for all of the tests to complete. We go in reverse order | |
6615 | * so we don't close datasets while threads are still using them. | |
6616 | */ | |
c242c188 | 6617 | for (t = ztest_opts.zo_threads - 1; t >= 0; t--) { |
c25b8f99 | 6618 | VERIFY0(thread_join(run_threads[t])); |
c242c188 CS |
6619 | if (t < ztest_opts.zo_datasets) |
6620 | ztest_dataset_close(t); | |
34dc7c2f BB |
6621 | } |
6622 | ||
34dc7c2f BB |
6623 | txg_wait_synced(spa_get_dsl(spa), 0); |
6624 | ||
428870ff BB |
6625 | zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(spa)); |
6626 | zs->zs_space = metaslab_class_get_space(spa_normal_class(spa)); | |
6627 | ||
c25b8f99 | 6628 | umem_free(run_threads, ztest_opts.zo_threads * sizeof (kthread_t *)); |
428870ff BB |
6629 | |
6630 | /* Kill the resume thread */ | |
6631 | ztest_exiting = B_TRUE; | |
c25b8f99 | 6632 | VERIFY0(thread_join(resume_thread)); |
428870ff | 6633 | ztest_resume(spa); |
34dc7c2f BB |
6634 | |
6635 | /* | |
428870ff BB |
6636 | * Right before closing the pool, kick off a bunch of async I/O; |
6637 | * spa_close() should wait for it to complete. | |
34dc7c2f | 6638 | */ |
fcff0f35 PD |
6639 | for (object = 1; object < 50; object++) { |
6640 | dmu_prefetch(spa->spa_meta_objset, object, 0, 0, 1ULL << 20, | |
6641 | ZIO_PRIORITY_SYNC_READ); | |
6642 | } | |
428870ff | 6643 | |
090ff092 RC |
6644 | /* Verify that at least one commit cb was called in a timely fashion */ |
6645 | if (zc_cb_counter >= ZTEST_COMMIT_CB_MIN_REG) | |
c99c9001 | 6646 | VERIFY0(zc_min_txg_delay); |
090ff092 | 6647 | |
428870ff BB |
6648 | spa_close(spa, FTAG); |
6649 | ||
6650 | /* | |
6651 | * Verify that we can loop over all pools. | |
6652 | */ | |
6653 | mutex_enter(&spa_namespace_lock); | |
6654 | for (spa = spa_next(NULL); spa != NULL; spa = spa_next(spa)) | |
c242c188 | 6655 | if (ztest_opts.zo_verbose > 3) |
428870ff BB |
6656 | (void) printf("spa_next: found %s\n", spa_name(spa)); |
6657 | mutex_exit(&spa_namespace_lock); | |
6658 | ||
6659 | /* | |
6660 | * Verify that we can export the pool and reimport it under a | |
6661 | * different name. | |
6662 | */ | |
379ca9cf | 6663 | if ((ztest_random(2) == 0) && !ztest_opts.zo_mmp_test) { |
eca7b760 IK |
6664 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
6665 | (void) snprintf(name, sizeof (name), "%s_import", | |
c242c188 CS |
6666 | ztest_opts.zo_pool); |
6667 | ztest_spa_import_export(ztest_opts.zo_pool, name); | |
6668 | ztest_spa_import_export(name, ztest_opts.zo_pool); | |
428870ff BB |
6669 | } |
6670 | ||
6671 | kernel_fini(); | |
572e2857 BB |
6672 | |
6673 | list_destroy(&zcl.zcl_callbacks); | |
0e8d1b2d | 6674 | mutex_destroy(&zcl.zcl_callbacks_lock); |
7809eb8b | 6675 | (void) rwlock_destroy(&ztest_name_lock); |
c242c188 | 6676 | mutex_destroy(&ztest_vdev_lock); |
428870ff BB |
6677 | } |
6678 | ||
6679 | static void | |
c242c188 | 6680 | ztest_freeze(void) |
428870ff | 6681 | { |
c242c188 | 6682 | ztest_ds_t *zd = &ztest_ds[0]; |
428870ff BB |
6683 | spa_t *spa; |
6684 | int numloops = 0; | |
6685 | ||
c242c188 | 6686 | if (ztest_opts.zo_verbose >= 3) |
428870ff | 6687 | (void) printf("testing spa_freeze()...\n"); |
9babb374 | 6688 | |
428870ff | 6689 | kernel_init(FREAD | FWRITE); |
c242c188 CS |
6690 | VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
6691 | VERIFY3U(0, ==, ztest_dataset_open(0)); | |
03c6040b GW |
6692 | spa->spa_debug = B_TRUE; |
6693 | ztest_spa = spa; | |
9babb374 | 6694 | |
428870ff BB |
6695 | /* |
6696 | * Force the first log block to be transactionally allocated. | |
6697 | * We have to do this before we freeze the pool -- otherwise | |
6698 | * the log chain won't be anchored. | |
6699 | */ | |
6700 | while (BP_IS_HOLE(&zd->zd_zilog->zl_header->zh_log)) { | |
6701 | ztest_dmu_object_alloc_free(zd, 0); | |
572e2857 | 6702 | zil_commit(zd->zd_zilog, 0); |
34dc7c2f BB |
6703 | } |
6704 | ||
6705 | txg_wait_synced(spa_get_dsl(spa), 0); | |
6706 | ||
428870ff BB |
6707 | /* |
6708 | * Freeze the pool. This stops spa_sync() from doing anything, | |
6709 | * so that the only way to record changes from now on is the ZIL. | |
6710 | */ | |
6711 | spa_freeze(spa); | |
b128c09f | 6712 | |
b02fe35d AR |
6713 | /* |
6714 | * Because it is hard to predict how much space a write will actually | |
6715 | * require beforehand, we leave ourselves some fudge space to write over | |
6716 | * capacity. | |
6717 | */ | |
6718 | uint64_t capacity = metaslab_class_get_space(spa_normal_class(spa)) / 2; | |
6719 | ||
428870ff BB |
6720 | /* |
6721 | * Run tests that generate log records but don't alter the pool config | |
6722 | * or depend on DSL sync tasks (snapshots, objset create/destroy, etc). | |
6723 | * We do a txg_wait_synced() after each iteration to force the txg | |
6724 | * to increase well beyond the last synced value in the uberblock. | |
6725 | * The ZIL should be OK with that. | |
b02fe35d AR |
6726 | * |
6727 | * Run a random number of times less than zo_maxloops and ensure we do | |
6728 | * not run out of space on the pool. | |
428870ff | 6729 | */ |
c242c188 | 6730 | while (ztest_random(10) != 0 && |
b02fe35d AR |
6731 | numloops++ < ztest_opts.zo_maxloops && |
6732 | metaslab_class_get_alloc(spa_normal_class(spa)) < capacity) { | |
6733 | ztest_od_t od; | |
50c957f7 | 6734 | ztest_od_init(&od, 0, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0); |
b02fe35d AR |
6735 | VERIFY0(ztest_object_init(zd, &od, sizeof (od), B_FALSE)); |
6736 | ztest_io(zd, od.od_object, | |
6737 | ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
428870ff BB |
6738 | txg_wait_synced(spa_get_dsl(spa), 0); |
6739 | } | |
b128c09f | 6740 | |
34dc7c2f | 6741 | /* |
428870ff | 6742 | * Commit all of the changes we just generated. |
34dc7c2f | 6743 | */ |
572e2857 | 6744 | zil_commit(zd->zd_zilog, 0); |
428870ff | 6745 | txg_wait_synced(spa_get_dsl(spa), 0); |
34dc7c2f | 6746 | |
428870ff BB |
6747 | /* |
6748 | * Close our dataset and close the pool. | |
6749 | */ | |
c242c188 | 6750 | ztest_dataset_close(0); |
34dc7c2f | 6751 | spa_close(spa, FTAG); |
428870ff | 6752 | kernel_fini(); |
34dc7c2f | 6753 | |
428870ff BB |
6754 | /* |
6755 | * Open and close the pool and dataset to induce log replay. | |
6756 | */ | |
6757 | kernel_init(FREAD | FWRITE); | |
c242c188 | 6758 | VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
29809a6c | 6759 | ASSERT(spa_freeze_txg(spa) == UINT64_MAX); |
c242c188 | 6760 | VERIFY3U(0, ==, ztest_dataset_open(0)); |
3bc7e0fb GW |
6761 | spa->spa_debug = B_TRUE; |
6762 | ztest_spa = spa; | |
6763 | txg_wait_synced(spa_get_dsl(spa), 0); | |
4807c0ba | 6764 | ztest_dataset_close(0); |
3bc7e0fb GW |
6765 | ztest_reguid(NULL, 0); |
6766 | ||
428870ff | 6767 | spa_close(spa, FTAG); |
34dc7c2f BB |
6768 | kernel_fini(); |
6769 | } | |
6770 | ||
6771 | void | |
6772 | print_time(hrtime_t t, char *timebuf) | |
6773 | { | |
6774 | hrtime_t s = t / NANOSEC; | |
6775 | hrtime_t m = s / 60; | |
6776 | hrtime_t h = m / 60; | |
6777 | hrtime_t d = h / 24; | |
6778 | ||
6779 | s -= m * 60; | |
6780 | m -= h * 60; | |
6781 | h -= d * 24; | |
6782 | ||
6783 | timebuf[0] = '\0'; | |
6784 | ||
6785 | if (d) | |
6786 | (void) sprintf(timebuf, | |
6787 | "%llud%02lluh%02llum%02llus", d, h, m, s); | |
6788 | else if (h) | |
6789 | (void) sprintf(timebuf, "%lluh%02llum%02llus", h, m, s); | |
6790 | else if (m) | |
6791 | (void) sprintf(timebuf, "%llum%02llus", m, s); | |
6792 | else | |
6793 | (void) sprintf(timebuf, "%llus", s); | |
6794 | } | |
6795 | ||
428870ff | 6796 | static nvlist_t * |
0bc8fd78 | 6797 | make_random_props(void) |
428870ff BB |
6798 | { |
6799 | nvlist_t *props; | |
6800 | ||
60b82074 | 6801 | VERIFY0(nvlist_alloc(&props, NV_UNIQUE_NAME, 0)); |
9ae529ec CS |
6802 | if (ztest_random(2) == 0) |
6803 | return (props); | |
60b82074 BB |
6804 | |
6805 | VERIFY0(nvlist_add_uint64(props, | |
6806 | zpool_prop_to_name(ZPOOL_PROP_AUTOREPLACE), 1)); | |
428870ff | 6807 | |
428870ff BB |
6808 | return (props); |
6809 | } | |
6810 | ||
379ca9cf OF |
6811 | /* |
6812 | * Import a storage pool with the given name. | |
6813 | */ | |
6814 | static void | |
6815 | ztest_import(ztest_shared_t *zs) | |
6816 | { | |
6817 | libzfs_handle_t *hdl; | |
6818 | importargs_t args = { 0 }; | |
6819 | spa_t *spa; | |
6820 | nvlist_t *cfg = NULL; | |
6821 | int nsearch = 1; | |
6822 | char *searchdirs[nsearch]; | |
6823 | char *name = ztest_opts.zo_pool; | |
6824 | int flags = ZFS_IMPORT_MISSING_LOG; | |
6825 | int error; | |
6826 | ||
6827 | mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); | |
6828 | VERIFY(rwlock_init(&ztest_name_lock, USYNC_THREAD, NULL) == 0); | |
6829 | ||
6830 | kernel_init(FREAD | FWRITE); | |
6831 | hdl = libzfs_init(); | |
6832 | ||
6833 | searchdirs[0] = ztest_opts.zo_dir; | |
6834 | args.paths = nsearch; | |
6835 | args.path = searchdirs; | |
6836 | args.can_be_active = B_FALSE; | |
6837 | ||
6838 | error = zpool_tryimport(hdl, name, &cfg, &args); | |
6839 | if (error) | |
6840 | (void) fatal(0, "No pools found\n"); | |
6841 | ||
6842 | VERIFY0(spa_import(name, cfg, NULL, flags)); | |
6843 | VERIFY0(spa_open(name, &spa, FTAG)); | |
6844 | zs->zs_metaslab_sz = | |
6845 | 1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift; | |
6846 | spa_close(spa, FTAG); | |
6847 | ||
6848 | libzfs_fini(hdl); | |
6849 | kernel_fini(); | |
6850 | ||
6851 | if (!ztest_opts.zo_mmp_test) { | |
6852 | ztest_run_zdb(ztest_opts.zo_pool); | |
6853 | ztest_freeze(); | |
6854 | ztest_run_zdb(ztest_opts.zo_pool); | |
6855 | } | |
6856 | ||
6857 | (void) rwlock_destroy(&ztest_name_lock); | |
6858 | mutex_destroy(&ztest_vdev_lock); | |
6859 | } | |
6860 | ||
34dc7c2f BB |
6861 | /* |
6862 | * Create a storage pool with the given name and initial vdev size. | |
428870ff | 6863 | * Then test spa_freeze() functionality. |
34dc7c2f BB |
6864 | */ |
6865 | static void | |
428870ff | 6866 | ztest_init(ztest_shared_t *zs) |
34dc7c2f BB |
6867 | { |
6868 | spa_t *spa; | |
428870ff | 6869 | nvlist_t *nvroot, *props; |
9ae529ec | 6870 | int i; |
428870ff | 6871 | |
c242c188 | 6872 | mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); |
7809eb8b | 6873 | VERIFY(rwlock_init(&ztest_name_lock, USYNC_THREAD, NULL) == 0); |
34dc7c2f BB |
6874 | |
6875 | kernel_init(FREAD | FWRITE); | |
6876 | ||
6877 | /* | |
6878 | * Create the storage pool. | |
6879 | */ | |
c242c188 | 6880 | (void) spa_destroy(ztest_opts.zo_pool); |
428870ff BB |
6881 | ztest_shared->zs_vdev_next_leaf = 0; |
6882 | zs->zs_splits = 0; | |
c242c188 | 6883 | zs->zs_mirrors = ztest_opts.zo_mirrors; |
ea0b2538 | 6884 | nvroot = make_vdev_root(NULL, NULL, NULL, ztest_opts.zo_vdev_size, 0, |
c242c188 | 6885 | 0, ztest_opts.zo_raidz, zs->zs_mirrors, 1); |
428870ff | 6886 | props = make_random_props(); |
60b82074 BB |
6887 | |
6888 | /* | |
6889 | * We don't expect the pool to suspend unless maxfaults == 0, | |
6890 | * in which case ztest_fault_inject() temporarily takes away | |
6891 | * the only valid replica. | |
6892 | */ | |
6893 | VERIFY0(nvlist_add_uint64(props, | |
6894 | zpool_prop_to_name(ZPOOL_PROP_FAILUREMODE), | |
6895 | MAXFAULTS(zs) ? ZIO_FAILURE_MODE_PANIC : ZIO_FAILURE_MODE_WAIT)); | |
6896 | ||
9ae529ec CS |
6897 | for (i = 0; i < SPA_FEATURES; i++) { |
6898 | char *buf; | |
6899 | VERIFY3S(-1, !=, asprintf(&buf, "feature@%s", | |
6900 | spa_feature_table[i].fi_uname)); | |
6901 | VERIFY3U(0, ==, nvlist_add_uint64(props, buf, 0)); | |
6902 | free(buf); | |
6903 | } | |
60b82074 BB |
6904 | |
6905 | VERIFY0(spa_create(ztest_opts.zo_pool, nvroot, props, NULL, NULL)); | |
34dc7c2f | 6906 | nvlist_free(nvroot); |
85802aa4 | 6907 | nvlist_free(props); |
34dc7c2f | 6908 | |
c242c188 CS |
6909 | VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
6910 | zs->zs_metaslab_sz = | |
6911 | 1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift; | |
34dc7c2f BB |
6912 | spa_close(spa, FTAG); |
6913 | ||
6914 | kernel_fini(); | |
428870ff | 6915 | |
379ca9cf OF |
6916 | if (!ztest_opts.zo_mmp_test) { |
6917 | ztest_run_zdb(ztest_opts.zo_pool); | |
6918 | ztest_freeze(); | |
6919 | ztest_run_zdb(ztest_opts.zo_pool); | |
6920 | } | |
c242c188 | 6921 | |
7809eb8b | 6922 | (void) rwlock_destroy(&ztest_name_lock); |
c242c188 CS |
6923 | mutex_destroy(&ztest_vdev_lock); |
6924 | } | |
6925 | ||
6926 | static void | |
9d81146b | 6927 | setup_data_fd(void) |
c242c188 | 6928 | { |
facbbe43 BB |
6929 | static char ztest_name_data[] = "/tmp/ztest.data.XXXXXX"; |
6930 | ||
6931 | ztest_fd_data = mkstemp(ztest_name_data); | |
9d81146b | 6932 | ASSERT3S(ztest_fd_data, >=, 0); |
facbbe43 | 6933 | (void) unlink(ztest_name_data); |
c242c188 CS |
6934 | } |
6935 | ||
22257dc0 CS |
6936 | static int |
6937 | shared_data_size(ztest_shared_hdr_t *hdr) | |
6938 | { | |
6939 | int size; | |
6940 | ||
6941 | size = hdr->zh_hdr_size; | |
6942 | size += hdr->zh_opts_size; | |
6943 | size += hdr->zh_size; | |
6944 | size += hdr->zh_stats_size * hdr->zh_stats_count; | |
6945 | size += hdr->zh_ds_size * hdr->zh_ds_count; | |
6946 | ||
6947 | return (size); | |
6948 | } | |
6949 | ||
c242c188 CS |
6950 | static void |
6951 | setup_hdr(void) | |
6952 | { | |
22257dc0 | 6953 | int size; |
c242c188 CS |
6954 | ztest_shared_hdr_t *hdr; |
6955 | ||
6956 | hdr = (void *)mmap(0, P2ROUNDUP(sizeof (*hdr), getpagesize()), | |
9d81146b | 6957 | PROT_READ | PROT_WRITE, MAP_SHARED, ztest_fd_data, 0); |
799402d8 | 6958 | ASSERT(hdr != MAP_FAILED); |
c242c188 | 6959 | |
9d81146b | 6960 | VERIFY3U(0, ==, ftruncate(ztest_fd_data, sizeof (ztest_shared_hdr_t))); |
22257dc0 | 6961 | |
c242c188 CS |
6962 | hdr->zh_hdr_size = sizeof (ztest_shared_hdr_t); |
6963 | hdr->zh_opts_size = sizeof (ztest_shared_opts_t); | |
6964 | hdr->zh_size = sizeof (ztest_shared_t); | |
6965 | hdr->zh_stats_size = sizeof (ztest_shared_callstate_t); | |
6966 | hdr->zh_stats_count = ZTEST_FUNCS; | |
6967 | hdr->zh_ds_size = sizeof (ztest_shared_ds_t); | |
6968 | hdr->zh_ds_count = ztest_opts.zo_datasets; | |
6969 | ||
22257dc0 | 6970 | size = shared_data_size(hdr); |
9d81146b | 6971 | VERIFY3U(0, ==, ftruncate(ztest_fd_data, size)); |
22257dc0 | 6972 | |
c242c188 CS |
6973 | (void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize())); |
6974 | } | |
6975 | ||
6976 | static void | |
6977 | setup_data(void) | |
6978 | { | |
6979 | int size, offset; | |
6980 | ztest_shared_hdr_t *hdr; | |
6981 | uint8_t *buf; | |
6982 | ||
6983 | hdr = (void *)mmap(0, P2ROUNDUP(sizeof (*hdr), getpagesize()), | |
9d81146b | 6984 | PROT_READ, MAP_SHARED, ztest_fd_data, 0); |
799402d8 | 6985 | ASSERT(hdr != MAP_FAILED); |
c242c188 | 6986 | |
22257dc0 | 6987 | size = shared_data_size(hdr); |
c242c188 CS |
6988 | |
6989 | (void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize())); | |
6990 | hdr = ztest_shared_hdr = (void *)mmap(0, P2ROUNDUP(size, getpagesize()), | |
9d81146b | 6991 | PROT_READ | PROT_WRITE, MAP_SHARED, ztest_fd_data, 0); |
799402d8 | 6992 | ASSERT(hdr != MAP_FAILED); |
c242c188 CS |
6993 | buf = (uint8_t *)hdr; |
6994 | ||
6995 | offset = hdr->zh_hdr_size; | |
6996 | ztest_shared_opts = (void *)&buf[offset]; | |
6997 | offset += hdr->zh_opts_size; | |
6998 | ztest_shared = (void *)&buf[offset]; | |
6999 | offset += hdr->zh_size; | |
7000 | ztest_shared_callstate = (void *)&buf[offset]; | |
7001 | offset += hdr->zh_stats_size * hdr->zh_stats_count; | |
7002 | ztest_shared_ds = (void *)&buf[offset]; | |
7003 | } | |
7004 | ||
7005 | static boolean_t | |
7006 | exec_child(char *cmd, char *libpath, boolean_t ignorekill, int *statusp) | |
7007 | { | |
7008 | pid_t pid; | |
7009 | int status; | |
483106eb | 7010 | char *cmdbuf = NULL; |
c242c188 CS |
7011 | |
7012 | pid = fork(); | |
7013 | ||
7014 | if (cmd == NULL) { | |
483106eb BB |
7015 | cmdbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
7016 | (void) strlcpy(cmdbuf, getexecname(), MAXPATHLEN); | |
c242c188 CS |
7017 | cmd = cmdbuf; |
7018 | } | |
7019 | ||
7020 | if (pid == -1) | |
7021 | fatal(1, "fork failed"); | |
7022 | ||
7023 | if (pid == 0) { /* child */ | |
7024 | char *emptyargv[2] = { cmd, NULL }; | |
9d81146b | 7025 | char fd_data_str[12]; |
c242c188 CS |
7026 | |
7027 | struct rlimit rl = { 1024, 1024 }; | |
7028 | (void) setrlimit(RLIMIT_NOFILE, &rl); | |
9d81146b ED |
7029 | |
7030 | (void) close(ztest_fd_rand); | |
7031 | VERIFY(11 >= snprintf(fd_data_str, 12, "%d", ztest_fd_data)); | |
7032 | VERIFY(0 == setenv("ZTEST_FD_DATA", fd_data_str, 1)); | |
7033 | ||
c242c188 CS |
7034 | (void) enable_extended_FILE_stdio(-1, -1); |
7035 | if (libpath != NULL) | |
7036 | VERIFY(0 == setenv("LD_LIBRARY_PATH", libpath, 1)); | |
7037 | (void) execv(cmd, emptyargv); | |
7038 | ztest_dump_core = B_FALSE; | |
7039 | fatal(B_TRUE, "exec failed: %s", cmd); | |
7040 | } | |
7041 | ||
483106eb BB |
7042 | if (cmdbuf != NULL) { |
7043 | umem_free(cmdbuf, MAXPATHLEN); | |
7044 | cmd = NULL; | |
7045 | } | |
7046 | ||
c242c188 CS |
7047 | while (waitpid(pid, &status, 0) != pid) |
7048 | continue; | |
7049 | if (statusp != NULL) | |
7050 | *statusp = status; | |
7051 | ||
7052 | if (WIFEXITED(status)) { | |
7053 | if (WEXITSTATUS(status) != 0) { | |
7054 | (void) fprintf(stderr, "child exited with code %d\n", | |
7055 | WEXITSTATUS(status)); | |
7056 | exit(2); | |
7057 | } | |
7058 | return (B_FALSE); | |
7059 | } else if (WIFSIGNALED(status)) { | |
7060 | if (!ignorekill || WTERMSIG(status) != SIGKILL) { | |
7061 | (void) fprintf(stderr, "child died with signal %d\n", | |
7062 | WTERMSIG(status)); | |
7063 | exit(3); | |
7064 | } | |
7065 | return (B_TRUE); | |
7066 | } else { | |
7067 | (void) fprintf(stderr, "something strange happened to child\n"); | |
7068 | exit(4); | |
7069 | /* NOTREACHED */ | |
7070 | } | |
7071 | } | |
7072 | ||
7073 | static void | |
7074 | ztest_run_init(void) | |
7075 | { | |
7076 | int i; | |
428870ff | 7077 | |
c242c188 | 7078 | ztest_shared_t *zs = ztest_shared; |
428870ff | 7079 | |
c242c188 CS |
7080 | /* |
7081 | * Blow away any existing copy of zpool.cache | |
7082 | */ | |
7083 | (void) remove(spa_config_path); | |
7084 | ||
379ca9cf OF |
7085 | if (ztest_opts.zo_init == 0) { |
7086 | if (ztest_opts.zo_verbose >= 1) | |
7087 | (void) printf("Importing pool %s\n", | |
7088 | ztest_opts.zo_pool); | |
7089 | ztest_import(zs); | |
7090 | return; | |
7091 | } | |
7092 | ||
c242c188 CS |
7093 | /* |
7094 | * Create and initialize our storage pool. | |
7095 | */ | |
7096 | for (i = 1; i <= ztest_opts.zo_init; i++) { | |
7097 | bzero(zs, sizeof (ztest_shared_t)); | |
7098 | if (ztest_opts.zo_verbose >= 3 && | |
7099 | ztest_opts.zo_init != 1) { | |
7100 | (void) printf("ztest_init(), pass %d\n", i); | |
7101 | } | |
7102 | ztest_init(zs); | |
7103 | } | |
34dc7c2f BB |
7104 | } |
7105 | ||
7106 | int | |
7107 | main(int argc, char **argv) | |
7108 | { | |
7109 | int kills = 0; | |
7110 | int iters = 0; | |
c242c188 CS |
7111 | int older = 0; |
7112 | int newer = 0; | |
34dc7c2f BB |
7113 | ztest_shared_t *zs; |
7114 | ztest_info_t *zi; | |
c242c188 | 7115 | ztest_shared_callstate_t *zc; |
34dc7c2f | 7116 | char timebuf[100]; |
f3c8c9e6 | 7117 | char numbuf[NN_NUMBUF_SZ]; |
428870ff | 7118 | spa_t *spa; |
483106eb | 7119 | char *cmd; |
c242c188 CS |
7120 | boolean_t hasalt; |
7121 | int f; | |
9d81146b | 7122 | char *fd_data_str = getenv("ZTEST_FD_DATA"); |
e82cdc3a | 7123 | struct sigaction action; |
34dc7c2f BB |
7124 | |
7125 | (void) setvbuf(stdout, NULL, _IOLBF, 0); | |
7126 | ||
498877ba | 7127 | dprintf_setup(&argc, argv); |
8fb1ede1 | 7128 | zfs_deadman_synctime_ms = 300000; |
498877ba | 7129 | |
e82cdc3a NB |
7130 | action.sa_handler = sig_handler; |
7131 | sigemptyset(&action.sa_mask); | |
7132 | action.sa_flags = 0; | |
7133 | ||
7134 | if (sigaction(SIGSEGV, &action, NULL) < 0) { | |
7135 | (void) fprintf(stderr, "ztest: cannot catch SIGSEGV: %s.\n", | |
7136 | strerror(errno)); | |
7137 | exit(EXIT_FAILURE); | |
7138 | } | |
7139 | ||
7140 | if (sigaction(SIGABRT, &action, NULL) < 0) { | |
7141 | (void) fprintf(stderr, "ztest: cannot catch SIGABRT: %s.\n", | |
7142 | strerror(errno)); | |
7143 | exit(EXIT_FAILURE); | |
7144 | } | |
7145 | ||
e1a0850c BB |
7146 | /* |
7147 | * Force random_get_bytes() to use /dev/urandom in order to prevent | |
7148 | * ztest from needlessly depleting the system entropy pool. | |
7149 | */ | |
7150 | random_path = "/dev/urandom"; | |
7151 | ztest_fd_rand = open(random_path, O_RDONLY); | |
9d81146b ED |
7152 | ASSERT3S(ztest_fd_rand, >=, 0); |
7153 | ||
7154 | if (!fd_data_str) { | |
c242c188 | 7155 | process_options(argc, argv); |
34dc7c2f | 7156 | |
9d81146b | 7157 | setup_data_fd(); |
c242c188 CS |
7158 | setup_hdr(); |
7159 | setup_data(); | |
7160 | bcopy(&ztest_opts, ztest_shared_opts, | |
7161 | sizeof (*ztest_shared_opts)); | |
7162 | } else { | |
9d81146b | 7163 | ztest_fd_data = atoi(fd_data_str); |
c242c188 CS |
7164 | setup_data(); |
7165 | bcopy(ztest_shared_opts, &ztest_opts, sizeof (ztest_opts)); | |
7166 | } | |
7167 | ASSERT3U(ztest_opts.zo_datasets, ==, ztest_shared_hdr->zh_ds_count); | |
34dc7c2f | 7168 | |
428870ff | 7169 | /* Override location of zpool.cache */ |
5be98cfe BB |
7170 | VERIFY(asprintf((char **)&spa_config_path, "%s/zpool.cache", |
7171 | ztest_opts.zo_dir) != -1); | |
9d81146b | 7172 | |
c242c188 CS |
7173 | ztest_ds = umem_alloc(ztest_opts.zo_datasets * sizeof (ztest_ds_t), |
7174 | UMEM_NOFAIL); | |
7175 | zs = ztest_shared; | |
7176 | ||
9d81146b | 7177 | if (fd_data_str) { |
c242c188 CS |
7178 | metaslab_gang_bang = ztest_opts.zo_metaslab_gang_bang; |
7179 | metaslab_df_alloc_threshold = | |
7180 | zs->zs_metaslab_df_alloc_threshold; | |
7181 | ||
7182 | if (zs->zs_do_init) | |
7183 | ztest_run_init(); | |
7184 | else | |
7185 | ztest_run(zs); | |
7186 | exit(0); | |
7187 | } | |
34dc7c2f | 7188 | |
c242c188 | 7189 | hasalt = (strlen(ztest_opts.zo_alt_ztest) != 0); |
34dc7c2f | 7190 | |
c242c188 | 7191 | if (ztest_opts.zo_verbose >= 1) { |
34dc7c2f BB |
7192 | (void) printf("%llu vdevs, %d datasets, %d threads," |
7193 | " %llu seconds...\n", | |
c242c188 CS |
7194 | (u_longlong_t)ztest_opts.zo_vdevs, |
7195 | ztest_opts.zo_datasets, | |
7196 | ztest_opts.zo_threads, | |
7197 | (u_longlong_t)ztest_opts.zo_time); | |
34dc7c2f BB |
7198 | } |
7199 | ||
483106eb BB |
7200 | cmd = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); |
7201 | (void) strlcpy(cmd, getexecname(), MAXNAMELEN); | |
c242c188 CS |
7202 | |
7203 | zs->zs_do_init = B_TRUE; | |
7204 | if (strlen(ztest_opts.zo_alt_ztest) != 0) { | |
7205 | if (ztest_opts.zo_verbose >= 1) { | |
7206 | (void) printf("Executing older ztest for " | |
7207 | "initialization: %s\n", ztest_opts.zo_alt_ztest); | |
7208 | } | |
7209 | VERIFY(!exec_child(ztest_opts.zo_alt_ztest, | |
7210 | ztest_opts.zo_alt_libpath, B_FALSE, NULL)); | |
7211 | } else { | |
7212 | VERIFY(!exec_child(NULL, NULL, B_FALSE, NULL)); | |
34dc7c2f | 7213 | } |
c242c188 | 7214 | zs->zs_do_init = B_FALSE; |
34dc7c2f | 7215 | |
428870ff | 7216 | zs->zs_proc_start = gethrtime(); |
c242c188 | 7217 | zs->zs_proc_stop = zs->zs_proc_start + ztest_opts.zo_time * NANOSEC; |
34dc7c2f | 7218 | |
d6320ddb | 7219 | for (f = 0; f < ZTEST_FUNCS; f++) { |
c242c188 CS |
7220 | zi = &ztest_info[f]; |
7221 | zc = ZTEST_GET_SHARED_CALLSTATE(f); | |
428870ff | 7222 | if (zs->zs_proc_start + zi->zi_interval[0] > zs->zs_proc_stop) |
c242c188 | 7223 | zc->zc_next = UINT64_MAX; |
34dc7c2f | 7224 | else |
c242c188 | 7225 | zc->zc_next = zs->zs_proc_start + |
428870ff | 7226 | ztest_random(2 * zi->zi_interval[0] + 1); |
34dc7c2f BB |
7227 | } |
7228 | ||
34dc7c2f BB |
7229 | /* |
7230 | * Run the tests in a loop. These tests include fault injection | |
7231 | * to verify that self-healing data works, and forced crashes | |
7232 | * to verify that we never lose on-disk consistency. | |
7233 | */ | |
428870ff | 7234 | while (gethrtime() < zs->zs_proc_stop) { |
34dc7c2f | 7235 | int status; |
c242c188 | 7236 | boolean_t killed; |
34dc7c2f BB |
7237 | |
7238 | /* | |
7239 | * Initialize the workload counters for each function. | |
7240 | */ | |
d6320ddb | 7241 | for (f = 0; f < ZTEST_FUNCS; f++) { |
c242c188 CS |
7242 | zc = ZTEST_GET_SHARED_CALLSTATE(f); |
7243 | zc->zc_count = 0; | |
7244 | zc->zc_time = 0; | |
34dc7c2f BB |
7245 | } |
7246 | ||
9babb374 | 7247 | /* Set the allocation switch size */ |
c242c188 CS |
7248 | zs->zs_metaslab_df_alloc_threshold = |
7249 | ztest_random(zs->zs_metaslab_sz / 4) + 1; | |
9babb374 | 7250 | |
c242c188 CS |
7251 | if (!hasalt || ztest_random(2) == 0) { |
7252 | if (hasalt && ztest_opts.zo_verbose >= 1) { | |
7253 | (void) printf("Executing newer ztest: %s\n", | |
7254 | cmd); | |
34dc7c2f | 7255 | } |
c242c188 CS |
7256 | newer++; |
7257 | killed = exec_child(cmd, NULL, B_TRUE, &status); | |
34dc7c2f | 7258 | } else { |
c242c188 CS |
7259 | if (hasalt && ztest_opts.zo_verbose >= 1) { |
7260 | (void) printf("Executing older ztest: %s\n", | |
7261 | ztest_opts.zo_alt_ztest); | |
7262 | } | |
7263 | older++; | |
7264 | killed = exec_child(ztest_opts.zo_alt_ztest, | |
7265 | ztest_opts.zo_alt_libpath, B_TRUE, &status); | |
34dc7c2f BB |
7266 | } |
7267 | ||
c242c188 CS |
7268 | if (killed) |
7269 | kills++; | |
34dc7c2f BB |
7270 | iters++; |
7271 | ||
c242c188 | 7272 | if (ztest_opts.zo_verbose >= 1) { |
34dc7c2f BB |
7273 | hrtime_t now = gethrtime(); |
7274 | ||
428870ff BB |
7275 | now = MIN(now, zs->zs_proc_stop); |
7276 | print_time(zs->zs_proc_stop - now, timebuf); | |
f3c8c9e6 | 7277 | nicenum(zs->zs_space, numbuf, sizeof (numbuf)); |
34dc7c2f BB |
7278 | |
7279 | (void) printf("Pass %3d, %8s, %3llu ENOSPC, " | |
7280 | "%4.1f%% of %5s used, %3.0f%% done, %8s to go\n", | |
7281 | iters, | |
7282 | WIFEXITED(status) ? "Complete" : "SIGKILL", | |
7283 | (u_longlong_t)zs->zs_enospc_count, | |
7284 | 100.0 * zs->zs_alloc / zs->zs_space, | |
7285 | numbuf, | |
428870ff | 7286 | 100.0 * (now - zs->zs_proc_start) / |
c242c188 | 7287 | (ztest_opts.zo_time * NANOSEC), timebuf); |
34dc7c2f BB |
7288 | } |
7289 | ||
c242c188 | 7290 | if (ztest_opts.zo_verbose >= 2) { |
34dc7c2f BB |
7291 | (void) printf("\nWorkload summary:\n\n"); |
7292 | (void) printf("%7s %9s %s\n", | |
7293 | "Calls", "Time", "Function"); | |
7294 | (void) printf("%7s %9s %s\n", | |
7295 | "-----", "----", "--------"); | |
d6320ddb | 7296 | for (f = 0; f < ZTEST_FUNCS; f++) { |
c242c188 CS |
7297 | zi = &ztest_info[f]; |
7298 | zc = ZTEST_GET_SHARED_CALLSTATE(f); | |
7299 | print_time(zc->zc_time, timebuf); | |
34dc7c2f | 7300 | (void) printf("%7llu %9s %s\n", |
c242c188 | 7301 | (u_longlong_t)zc->zc_count, timebuf, |
fdc5d982 | 7302 | zi->zi_funcname); |
34dc7c2f BB |
7303 | } |
7304 | (void) printf("\n"); | |
7305 | } | |
7306 | ||
7307 | /* | |
428870ff BB |
7308 | * It's possible that we killed a child during a rename test, |
7309 | * in which case we'll have a 'ztest_tmp' pool lying around | |
7310 | * instead of 'ztest'. Do a blind rename in case this happened. | |
34dc7c2f | 7311 | */ |
428870ff | 7312 | kernel_init(FREAD); |
c242c188 | 7313 | if (spa_open(ztest_opts.zo_pool, &spa, FTAG) == 0) { |
428870ff BB |
7314 | spa_close(spa, FTAG); |
7315 | } else { | |
eca7b760 | 7316 | char tmpname[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
7317 | kernel_fini(); |
7318 | kernel_init(FREAD | FWRITE); | |
7319 | (void) snprintf(tmpname, sizeof (tmpname), "%s_tmp", | |
c242c188 CS |
7320 | ztest_opts.zo_pool); |
7321 | (void) spa_rename(tmpname, ztest_opts.zo_pool); | |
428870ff | 7322 | } |
34dc7c2f | 7323 | kernel_fini(); |
34dc7c2f | 7324 | |
379ca9cf OF |
7325 | if (!ztest_opts.zo_mmp_test) |
7326 | ztest_run_zdb(ztest_opts.zo_pool); | |
428870ff | 7327 | } |
34dc7c2f | 7328 | |
c242c188 CS |
7329 | if (ztest_opts.zo_verbose >= 1) { |
7330 | if (hasalt) { | |
7331 | (void) printf("%d runs of older ztest: %s\n", older, | |
7332 | ztest_opts.zo_alt_ztest); | |
7333 | (void) printf("%d runs of newer ztest: %s\n", newer, | |
7334 | cmd); | |
7335 | } | |
34dc7c2f BB |
7336 | (void) printf("%d killed, %d completed, %.0f%% kill rate\n", |
7337 | kills, iters - kills, (100.0 * kills) / MAX(1, iters)); | |
7338 | } | |
7339 | ||
483106eb BB |
7340 | umem_free(cmd, MAXNAMELEN); |
7341 | ||
34dc7c2f BB |
7342 | return (0); |
7343 | } |