]>
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. |
34dc7c2f BB |
26 | */ |
27 | ||
34dc7c2f BB |
28 | /* |
29 | * The objective of this program is to provide a DMU/ZAP/SPA stress test | |
30 | * that runs entirely in userland, is easy to use, and easy to extend. | |
31 | * | |
32 | * The overall design of the ztest program is as follows: | |
33 | * | |
34 | * (1) For each major functional area (e.g. adding vdevs to a pool, | |
35 | * creating and destroying datasets, reading and writing objects, etc) | |
36 | * we have a simple routine to test that functionality. These | |
37 | * individual routines do not have to do anything "stressful". | |
38 | * | |
39 | * (2) We turn these simple functionality tests into a stress test by | |
40 | * running them all in parallel, with as many threads as desired, | |
41 | * and spread across as many datasets, objects, and vdevs as desired. | |
42 | * | |
43 | * (3) While all this is happening, we inject faults into the pool to | |
44 | * verify that self-healing data really works. | |
45 | * | |
46 | * (4) Every time we open a dataset, we change its checksum and compression | |
47 | * functions. Thus even individual objects vary from block to block | |
48 | * in which checksum they use and whether they're compressed. | |
49 | * | |
50 | * (5) To verify that we never lose on-disk consistency after a crash, | |
51 | * we run the entire test in a child of the main process. | |
52 | * At random times, the child self-immolates with a SIGKILL. | |
53 | * This is the software equivalent of pulling the power cord. | |
54 | * The parent then runs the test again, using the existing | |
9b67f605 | 55 | * storage pool, as many times as desired. If backwards compatibility |
c242c188 CS |
56 | * testing is enabled ztest will sometimes run the "older" version |
57 | * of ztest after a SIGKILL. | |
34dc7c2f BB |
58 | * |
59 | * (6) To verify that we don't have future leaks or temporal incursions, | |
60 | * many of the functional tests record the transaction group number | |
61 | * as part of their data. When reading old data, they verify that | |
62 | * the transaction group number is less than the current, open txg. | |
63 | * If you add a new test, please do this if applicable. | |
64 | * | |
1e33ac1e BB |
65 | * (7) Threads are created with a reduced stack size, for sanity checking. |
66 | * Therefore, it's important not to allocate huge buffers on the stack. | |
67 | * | |
34dc7c2f BB |
68 | * When run with no arguments, ztest runs for about five minutes and |
69 | * produces no output if successful. To get a little bit of information, | |
70 | * specify -V. To get more information, specify -VV, and so on. | |
71 | * | |
72 | * To turn this into an overnight stress test, use -T to specify run time. | |
73 | * | |
74 | * You can ask more more vdevs [-v], datasets [-d], or threads [-t] | |
75 | * to increase the pool capacity, fanout, and overall stress level. | |
76 | * | |
c242c188 CS |
77 | * Use the -k option to set the desired frequency of kills. |
78 | * | |
79 | * When ztest invokes itself it passes all relevant information through a | |
80 | * temporary file which is mmap-ed in the child process. This allows shared | |
81 | * memory to survive the exec syscall. The ztest_shared_hdr_t struct is always | |
82 | * stored at offset 0 of this file and contains information on the size and | |
83 | * number of shared structures in the file. The information stored in this file | |
84 | * must remain backwards compatible with older versions of ztest so that | |
85 | * ztest can invoke them during backwards compatibility testing (-B). | |
34dc7c2f BB |
86 | */ |
87 | ||
88 | #include <sys/zfs_context.h> | |
89 | #include <sys/spa.h> | |
90 | #include <sys/dmu.h> | |
91 | #include <sys/txg.h> | |
9babb374 | 92 | #include <sys/dbuf.h> |
34dc7c2f | 93 | #include <sys/zap.h> |
34dc7c2f BB |
94 | #include <sys/dmu_objset.h> |
95 | #include <sys/poll.h> | |
96 | #include <sys/stat.h> | |
97 | #include <sys/time.h> | |
98 | #include <sys/wait.h> | |
99 | #include <sys/mman.h> | |
100 | #include <sys/resource.h> | |
101 | #include <sys/zio.h> | |
34dc7c2f | 102 | #include <sys/zil.h> |
428870ff | 103 | #include <sys/zil_impl.h> |
e3a07cd0 | 104 | #include <sys/zfs_rlock.h> |
34dc7c2f | 105 | #include <sys/vdev_impl.h> |
b128c09f | 106 | #include <sys/vdev_file.h> |
34dc7c2f | 107 | #include <sys/spa_impl.h> |
428870ff | 108 | #include <sys/metaslab_impl.h> |
34dc7c2f | 109 | #include <sys/dsl_prop.h> |
9babb374 | 110 | #include <sys/dsl_dataset.h> |
13fe0198 | 111 | #include <sys/dsl_destroy.h> |
428870ff BB |
112 | #include <sys/dsl_scan.h> |
113 | #include <sys/zio_checksum.h> | |
34dc7c2f | 114 | #include <sys/refcount.h> |
9ae529ec | 115 | #include <sys/zfeature.h> |
13fe0198 | 116 | #include <sys/dsl_userhold.h> |
a6255b7f | 117 | #include <sys/abd.h> |
34dc7c2f BB |
118 | #include <stdio.h> |
119 | #include <stdio_ext.h> | |
120 | #include <stdlib.h> | |
121 | #include <unistd.h> | |
122 | #include <signal.h> | |
123 | #include <umem.h> | |
34dc7c2f BB |
124 | #include <ctype.h> |
125 | #include <math.h> | |
126 | #include <sys/fs/zfs.h> | |
1eeb4562 | 127 | #include <zfs_fletcher.h> |
428870ff | 128 | #include <libnvpair.h> |
379ca9cf | 129 | #include <libzfs.h> |
967798d0 | 130 | #ifdef __GLIBC__ |
e82cdc3a NB |
131 | #include <execinfo.h> /* for backtrace() */ |
132 | #endif | |
34dc7c2f | 133 | |
9d81146b ED |
134 | static int ztest_fd_data = -1; |
135 | static int ztest_fd_rand = -1; | |
c242c188 CS |
136 | |
137 | typedef struct ztest_shared_hdr { | |
138 | uint64_t zh_hdr_size; | |
139 | uint64_t zh_opts_size; | |
140 | uint64_t zh_size; | |
141 | uint64_t zh_stats_size; | |
142 | uint64_t zh_stats_count; | |
143 | uint64_t zh_ds_size; | |
144 | uint64_t zh_ds_count; | |
145 | } ztest_shared_hdr_t; | |
146 | ||
147 | static ztest_shared_hdr_t *ztest_shared_hdr; | |
148 | ||
149 | typedef struct ztest_shared_opts { | |
eca7b760 IK |
150 | char zo_pool[ZFS_MAX_DATASET_NAME_LEN]; |
151 | char zo_dir[ZFS_MAX_DATASET_NAME_LEN]; | |
c242c188 CS |
152 | char zo_alt_ztest[MAXNAMELEN]; |
153 | char zo_alt_libpath[MAXNAMELEN]; | |
154 | uint64_t zo_vdevs; | |
155 | uint64_t zo_vdevtime; | |
156 | size_t zo_vdev_size; | |
157 | int zo_ashift; | |
158 | int zo_mirrors; | |
159 | int zo_raidz; | |
160 | int zo_raidz_parity; | |
161 | int zo_datasets; | |
162 | int zo_threads; | |
163 | uint64_t zo_passtime; | |
164 | uint64_t zo_killrate; | |
165 | int zo_verbose; | |
166 | int zo_init; | |
167 | uint64_t zo_time; | |
168 | uint64_t zo_maxloops; | |
169 | uint64_t zo_metaslab_gang_bang; | |
379ca9cf | 170 | int zo_mmp_test; |
c242c188 CS |
171 | } ztest_shared_opts_t; |
172 | ||
173 | static const ztest_shared_opts_t ztest_opts_defaults = { | |
174 | .zo_pool = { 'z', 't', 'e', 's', 't', '\0' }, | |
175 | .zo_dir = { '/', 't', 'm', 'p', '\0' }, | |
176 | .zo_alt_ztest = { '\0' }, | |
177 | .zo_alt_libpath = { '\0' }, | |
178 | .zo_vdevs = 5, | |
179 | .zo_ashift = SPA_MINBLOCKSHIFT, | |
180 | .zo_mirrors = 2, | |
181 | .zo_raidz = 4, | |
182 | .zo_raidz_parity = 1, | |
4e21fd06 | 183 | .zo_vdev_size = SPA_MINDEVSIZE * 4, /* 256m default size */ |
c242c188 CS |
184 | .zo_datasets = 7, |
185 | .zo_threads = 23, | |
186 | .zo_passtime = 60, /* 60 seconds */ | |
187 | .zo_killrate = 70, /* 70% kill rate */ | |
188 | .zo_verbose = 0, | |
379ca9cf | 189 | .zo_mmp_test = 0, |
c242c188 CS |
190 | .zo_init = 1, |
191 | .zo_time = 300, /* 5 minutes */ | |
192 | .zo_maxloops = 50, /* max loops during spa_freeze() */ | |
193 | .zo_metaslab_gang_bang = 32 << 10 | |
194 | }; | |
195 | ||
196 | extern uint64_t metaslab_gang_bang; | |
197 | extern uint64_t metaslab_df_alloc_threshold; | |
080b3100 | 198 | extern int metaslab_preload_limit; |
d3c2ae1c | 199 | extern boolean_t zfs_compressed_arc_enabled; |
a6255b7f | 200 | extern int zfs_abd_scatter_enabled; |
c242c188 CS |
201 | |
202 | static ztest_shared_opts_t *ztest_shared_opts; | |
203 | static ztest_shared_opts_t ztest_opts; | |
204 | ||
205 | typedef struct ztest_shared_ds { | |
206 | uint64_t zd_seq; | |
207 | } ztest_shared_ds_t; | |
208 | ||
209 | static ztest_shared_ds_t *ztest_shared_ds; | |
210 | #define ZTEST_GET_SHARED_DS(d) (&ztest_shared_ds[d]) | |
428870ff BB |
211 | |
212 | #define BT_MAGIC 0x123456789abcdefULL | |
c242c188 CS |
213 | #define MAXFAULTS() \ |
214 | (MAX(zs->zs_mirrors, 1) * (ztest_opts.zo_raidz_parity + 1) - 1) | |
428870ff BB |
215 | |
216 | enum ztest_io_type { | |
217 | ZTEST_IO_WRITE_TAG, | |
218 | ZTEST_IO_WRITE_PATTERN, | |
219 | ZTEST_IO_WRITE_ZEROES, | |
220 | ZTEST_IO_TRUNCATE, | |
221 | ZTEST_IO_SETATTR, | |
03c6040b | 222 | ZTEST_IO_REWRITE, |
428870ff BB |
223 | ZTEST_IO_TYPES |
224 | }; | |
34dc7c2f BB |
225 | |
226 | typedef struct ztest_block_tag { | |
428870ff | 227 | uint64_t bt_magic; |
34dc7c2f BB |
228 | uint64_t bt_objset; |
229 | uint64_t bt_object; | |
50c957f7 | 230 | uint64_t bt_dnodesize; |
34dc7c2f | 231 | uint64_t bt_offset; |
428870ff | 232 | uint64_t bt_gen; |
34dc7c2f | 233 | uint64_t bt_txg; |
428870ff | 234 | uint64_t bt_crtxg; |
34dc7c2f BB |
235 | } ztest_block_tag_t; |
236 | ||
428870ff BB |
237 | typedef struct bufwad { |
238 | uint64_t bw_index; | |
239 | uint64_t bw_txg; | |
240 | uint64_t bw_data; | |
241 | } bufwad_t; | |
242 | ||
428870ff BB |
243 | typedef struct rll { |
244 | void *rll_writer; | |
245 | int rll_readers; | |
1e33ac1e BB |
246 | kmutex_t rll_lock; |
247 | kcondvar_t rll_cv; | |
428870ff BB |
248 | } rll_t; |
249 | ||
e3a07cd0 BP |
250 | typedef struct zll { |
251 | list_t z_list; | |
252 | kmutex_t z_lock; | |
253 | } zll_t; | |
428870ff BB |
254 | |
255 | #define ZTEST_RANGE_LOCKS 64 | |
256 | #define ZTEST_OBJECT_LOCKS 64 | |
257 | ||
258 | /* | |
259 | * Object descriptor. Used as a template for object lookup/create/remove. | |
260 | */ | |
261 | typedef struct ztest_od { | |
262 | uint64_t od_dir; | |
263 | uint64_t od_object; | |
264 | dmu_object_type_t od_type; | |
265 | dmu_object_type_t od_crtype; | |
266 | uint64_t od_blocksize; | |
267 | uint64_t od_crblocksize; | |
50c957f7 | 268 | uint64_t od_crdnodesize; |
428870ff BB |
269 | uint64_t od_gen; |
270 | uint64_t od_crgen; | |
eca7b760 | 271 | char od_name[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff | 272 | } ztest_od_t; |
34dc7c2f | 273 | |
428870ff BB |
274 | /* |
275 | * Per-dataset state. | |
276 | */ | |
277 | typedef struct ztest_ds { | |
c242c188 | 278 | ztest_shared_ds_t *zd_shared; |
428870ff | 279 | objset_t *zd_os; |
7809eb8b | 280 | rwlock_t zd_zilog_lock; |
428870ff | 281 | zilog_t *zd_zilog; |
428870ff | 282 | ztest_od_t *zd_od; /* debugging aid */ |
eca7b760 | 283 | char zd_name[ZFS_MAX_DATASET_NAME_LEN]; |
1e33ac1e | 284 | kmutex_t zd_dirobj_lock; |
428870ff | 285 | rll_t zd_object_lock[ZTEST_OBJECT_LOCKS]; |
e3a07cd0 | 286 | zll_t zd_range_lock[ZTEST_RANGE_LOCKS]; |
428870ff BB |
287 | } ztest_ds_t; |
288 | ||
289 | /* | |
290 | * Per-iteration state. | |
291 | */ | |
292 | typedef void ztest_func_t(ztest_ds_t *zd, uint64_t id); | |
293 | ||
294 | typedef struct ztest_info { | |
295 | ztest_func_t *zi_func; /* test function */ | |
296 | uint64_t zi_iters; /* iterations per execution */ | |
297 | uint64_t *zi_interval; /* execute every <interval> seconds */ | |
fdc5d982 | 298 | const char *zi_funcname; /* name of test function */ |
428870ff | 299 | } ztest_info_t; |
34dc7c2f | 300 | |
c242c188 CS |
301 | typedef struct ztest_shared_callstate { |
302 | uint64_t zc_count; /* per-pass count */ | |
303 | uint64_t zc_time; /* per-pass time */ | |
304 | uint64_t zc_next; /* next time to call this function */ | |
305 | } ztest_shared_callstate_t; | |
306 | ||
307 | static ztest_shared_callstate_t *ztest_shared_callstate; | |
308 | #define ZTEST_GET_SHARED_CALLSTATE(c) (&ztest_shared_callstate[c]) | |
309 | ||
34dc7c2f BB |
310 | ztest_func_t ztest_dmu_read_write; |
311 | ztest_func_t ztest_dmu_write_parallel; | |
312 | ztest_func_t ztest_dmu_object_alloc_free; | |
428870ff | 313 | ztest_func_t ztest_dmu_commit_callbacks; |
34dc7c2f BB |
314 | ztest_func_t ztest_zap; |
315 | ztest_func_t ztest_zap_parallel; | |
428870ff | 316 | ztest_func_t ztest_zil_commit; |
3e31d2b0 | 317 | ztest_func_t ztest_zil_remount; |
428870ff | 318 | ztest_func_t ztest_dmu_read_write_zcopy; |
34dc7c2f | 319 | ztest_func_t ztest_dmu_objset_create_destroy; |
428870ff BB |
320 | ztest_func_t ztest_dmu_prealloc; |
321 | ztest_func_t ztest_fzap; | |
34dc7c2f | 322 | ztest_func_t ztest_dmu_snapshot_create_destroy; |
428870ff BB |
323 | ztest_func_t ztest_dsl_prop_get_set; |
324 | ztest_func_t ztest_spa_prop_get_set; | |
34dc7c2f BB |
325 | ztest_func_t ztest_spa_create_destroy; |
326 | ztest_func_t ztest_fault_inject; | |
428870ff BB |
327 | ztest_func_t ztest_ddt_repair; |
328 | ztest_func_t ztest_dmu_snapshot_hold; | |
0582e403 | 329 | ztest_func_t ztest_mmp_enable_disable; |
b128c09f | 330 | ztest_func_t ztest_spa_rename; |
428870ff BB |
331 | ztest_func_t ztest_scrub; |
332 | ztest_func_t ztest_dsl_dataset_promote_busy; | |
34dc7c2f BB |
333 | ztest_func_t ztest_vdev_attach_detach; |
334 | ztest_func_t ztest_vdev_LUN_growth; | |
335 | ztest_func_t ztest_vdev_add_remove; | |
b128c09f | 336 | ztest_func_t ztest_vdev_aux_add_remove; |
428870ff | 337 | ztest_func_t ztest_split_pool; |
3541dc6d | 338 | ztest_func_t ztest_reguid; |
ea0b2538 | 339 | ztest_func_t ztest_spa_upgrade; |
1eeb4562 | 340 | ztest_func_t ztest_fletcher; |
37f520db | 341 | ztest_func_t ztest_fletcher_incr; |
50c957f7 | 342 | ztest_func_t ztest_verify_dnode_bt; |
34dc7c2f | 343 | |
428870ff BB |
344 | uint64_t zopt_always = 0ULL * NANOSEC; /* all the time */ |
345 | uint64_t zopt_incessant = 1ULL * NANOSEC / 10; /* every 1/10 second */ | |
346 | uint64_t zopt_often = 1ULL * NANOSEC; /* every second */ | |
347 | uint64_t zopt_sometimes = 10ULL * NANOSEC; /* every 10 seconds */ | |
348 | uint64_t zopt_rarely = 60ULL * NANOSEC; /* every 60 seconds */ | |
34dc7c2f | 349 | |
fdc5d982 TC |
350 | #define ZTI_INIT(func, iters, interval) \ |
351 | { .zi_func = (func), \ | |
352 | .zi_iters = (iters), \ | |
353 | .zi_interval = (interval), \ | |
354 | .zi_funcname = # func } | |
355 | ||
34dc7c2f | 356 | ztest_info_t ztest_info[] = { |
fdc5d982 TC |
357 | ZTI_INIT(ztest_dmu_read_write, 1, &zopt_always), |
358 | ZTI_INIT(ztest_dmu_write_parallel, 10, &zopt_always), | |
359 | ZTI_INIT(ztest_dmu_object_alloc_free, 1, &zopt_always), | |
360 | ZTI_INIT(ztest_dmu_commit_callbacks, 1, &zopt_always), | |
361 | ZTI_INIT(ztest_zap, 30, &zopt_always), | |
362 | ZTI_INIT(ztest_zap_parallel, 100, &zopt_always), | |
363 | ZTI_INIT(ztest_split_pool, 1, &zopt_always), | |
364 | ZTI_INIT(ztest_zil_commit, 1, &zopt_incessant), | |
365 | ZTI_INIT(ztest_zil_remount, 1, &zopt_sometimes), | |
366 | ZTI_INIT(ztest_dmu_read_write_zcopy, 1, &zopt_often), | |
367 | ZTI_INIT(ztest_dmu_objset_create_destroy, 1, &zopt_often), | |
368 | ZTI_INIT(ztest_dsl_prop_get_set, 1, &zopt_often), | |
369 | ZTI_INIT(ztest_spa_prop_get_set, 1, &zopt_sometimes), | |
428870ff | 370 | #if 0 |
fdc5d982 | 371 | ZTI_INIT(ztest_dmu_prealloc, 1, &zopt_sometimes), |
428870ff | 372 | #endif |
fdc5d982 TC |
373 | ZTI_INIT(ztest_fzap, 1, &zopt_sometimes), |
374 | ZTI_INIT(ztest_dmu_snapshot_create_destroy, 1, &zopt_sometimes), | |
375 | ZTI_INIT(ztest_spa_create_destroy, 1, &zopt_sometimes), | |
376 | ZTI_INIT(ztest_fault_inject, 1, &zopt_sometimes), | |
377 | ZTI_INIT(ztest_ddt_repair, 1, &zopt_sometimes), | |
378 | ZTI_INIT(ztest_dmu_snapshot_hold, 1, &zopt_sometimes), | |
0582e403 | 379 | ZTI_INIT(ztest_mmp_enable_disable, 1, &zopt_sometimes), |
fdc5d982 TC |
380 | ZTI_INIT(ztest_reguid, 1, &zopt_rarely), |
381 | ZTI_INIT(ztest_spa_rename, 1, &zopt_rarely), | |
382 | ZTI_INIT(ztest_scrub, 1, &zopt_rarely), | |
383 | ZTI_INIT(ztest_spa_upgrade, 1, &zopt_rarely), | |
384 | ZTI_INIT(ztest_dsl_dataset_promote_busy, 1, &zopt_rarely), | |
385 | ZTI_INIT(ztest_vdev_attach_detach, 1, &zopt_sometimes), | |
386 | ZTI_INIT(ztest_vdev_LUN_growth, 1, &zopt_rarely), | |
387 | ZTI_INIT(ztest_vdev_add_remove, 1, &ztest_opts.zo_vdevtime), | |
388 | ZTI_INIT(ztest_vdev_aux_add_remove, 1, &ztest_opts.zo_vdevtime), | |
1eeb4562 | 389 | ZTI_INIT(ztest_fletcher, 1, &zopt_rarely), |
37f520db | 390 | ZTI_INIT(ztest_fletcher_incr, 1, &zopt_rarely), |
50c957f7 | 391 | ZTI_INIT(ztest_verify_dnode_bt, 1, &zopt_sometimes), |
34dc7c2f BB |
392 | }; |
393 | ||
394 | #define ZTEST_FUNCS (sizeof (ztest_info) / sizeof (ztest_info_t)) | |
395 | ||
428870ff BB |
396 | /* |
397 | * The following struct is used to hold a list of uncalled commit callbacks. | |
398 | * The callbacks are ordered by txg number. | |
399 | */ | |
400 | typedef struct ztest_cb_list { | |
1e33ac1e BB |
401 | kmutex_t zcl_callbacks_lock; |
402 | list_t zcl_callbacks; | |
428870ff | 403 | } ztest_cb_list_t; |
34dc7c2f BB |
404 | |
405 | /* | |
406 | * Stuff we need to share writably between parent and child. | |
407 | */ | |
408 | typedef struct ztest_shared { | |
c242c188 | 409 | boolean_t zs_do_init; |
428870ff BB |
410 | hrtime_t zs_proc_start; |
411 | hrtime_t zs_proc_stop; | |
412 | hrtime_t zs_thread_start; | |
413 | hrtime_t zs_thread_stop; | |
414 | hrtime_t zs_thread_kill; | |
34dc7c2f | 415 | uint64_t zs_enospc_count; |
428870ff BB |
416 | uint64_t zs_vdev_next_leaf; |
417 | uint64_t zs_vdev_aux; | |
34dc7c2f BB |
418 | uint64_t zs_alloc; |
419 | uint64_t zs_space; | |
428870ff BB |
420 | uint64_t zs_splits; |
421 | uint64_t zs_mirrors; | |
c242c188 CS |
422 | uint64_t zs_metaslab_sz; |
423 | uint64_t zs_metaslab_df_alloc_threshold; | |
424 | uint64_t zs_guid; | |
34dc7c2f BB |
425 | } ztest_shared_t; |
426 | ||
428870ff BB |
427 | #define ID_PARALLEL -1ULL |
428 | ||
34dc7c2f | 429 | static char ztest_dev_template[] = "%s/%s.%llua"; |
b128c09f | 430 | static char ztest_aux_template[] = "%s/%s.%s.%llu"; |
428870ff | 431 | ztest_shared_t *ztest_shared; |
34dc7c2f | 432 | |
c242c188 CS |
433 | static spa_t *ztest_spa = NULL; |
434 | static ztest_ds_t *ztest_ds; | |
435 | ||
436 | static kmutex_t ztest_vdev_lock; | |
3bc7e0fb GW |
437 | |
438 | /* | |
439 | * The ztest_name_lock protects the pool and dataset namespace used by | |
440 | * the individual tests. To modify the namespace, consumers must grab | |
441 | * this lock as writer. Grabbing the lock as reader will ensure that the | |
442 | * namespace does not change while the lock is held. | |
443 | */ | |
7809eb8b | 444 | static rwlock_t ztest_name_lock; |
34dc7c2f | 445 | |
c242c188 | 446 | static boolean_t ztest_dump_core = B_TRUE; |
b128c09f | 447 | static boolean_t ztest_exiting; |
34dc7c2f | 448 | |
428870ff BB |
449 | /* Global commit callback list */ |
450 | static ztest_cb_list_t zcl; | |
090ff092 RC |
451 | /* Commit cb delay */ |
452 | static uint64_t zc_min_txg_delay = UINT64_MAX; | |
453 | static int zc_cb_counter = 0; | |
454 | ||
455 | /* | |
456 | * Minimum number of commit callbacks that need to be registered for us to check | |
457 | * whether the minimum txg delay is acceptable. | |
458 | */ | |
459 | #define ZTEST_COMMIT_CB_MIN_REG 100 | |
460 | ||
461 | /* | |
462 | * If a number of txgs equal to this threshold have been created after a commit | |
463 | * callback has been registered but not called, then we assume there is an | |
464 | * implementation bug. | |
465 | */ | |
466 | #define ZTEST_COMMIT_CB_THRESH (TXG_CONCURRENT_STATES + 1000) | |
428870ff | 467 | |
34dc7c2f | 468 | extern uint64_t metaslab_gang_bang; |
9babb374 | 469 | extern uint64_t metaslab_df_alloc_threshold; |
34dc7c2f | 470 | |
428870ff BB |
471 | enum ztest_object { |
472 | ZTEST_META_DNODE = 0, | |
473 | ZTEST_DIROBJ, | |
474 | ZTEST_OBJECTS | |
475 | }; | |
34dc7c2f BB |
476 | |
477 | static void usage(boolean_t) __NORETURN; | |
478 | ||
479 | /* | |
480 | * These libumem hooks provide a reasonable set of defaults for the allocator's | |
481 | * debugging facilities. | |
482 | */ | |
483 | const char * | |
0bc8fd78 | 484 | _umem_debug_init(void) |
34dc7c2f BB |
485 | { |
486 | return ("default,verbose"); /* $UMEM_DEBUG setting */ | |
487 | } | |
488 | ||
489 | const char * | |
490 | _umem_logging_init(void) | |
491 | { | |
492 | return ("fail,contents"); /* $UMEM_LOGGING setting */ | |
493 | } | |
494 | ||
e82cdc3a NB |
495 | #define BACKTRACE_SZ 100 |
496 | ||
497 | static void sig_handler(int signo) | |
498 | { | |
499 | struct sigaction action; | |
967798d0 | 500 | #ifdef __GLIBC__ /* backtrace() is a GNU extension */ |
e82cdc3a NB |
501 | int nptrs; |
502 | void *buffer[BACKTRACE_SZ]; | |
503 | ||
504 | nptrs = backtrace(buffer, BACKTRACE_SZ); | |
505 | backtrace_symbols_fd(buffer, nptrs, STDERR_FILENO); | |
506 | #endif | |
507 | ||
508 | /* | |
509 | * Restore default action and re-raise signal so SIGSEGV and | |
510 | * SIGABRT can trigger a core dump. | |
511 | */ | |
512 | action.sa_handler = SIG_DFL; | |
513 | sigemptyset(&action.sa_mask); | |
514 | action.sa_flags = 0; | |
515 | (void) sigaction(signo, &action, NULL); | |
516 | raise(signo); | |
517 | } | |
518 | ||
34dc7c2f BB |
519 | #define FATAL_MSG_SZ 1024 |
520 | ||
521 | char *fatal_msg; | |
522 | ||
523 | static void | |
524 | fatal(int do_perror, char *message, ...) | |
525 | { | |
526 | va_list args; | |
527 | int save_errno = errno; | |
40b84e7a | 528 | char *buf; |
34dc7c2f BB |
529 | |
530 | (void) fflush(stdout); | |
40b84e7a | 531 | buf = umem_alloc(FATAL_MSG_SZ, UMEM_NOFAIL); |
34dc7c2f BB |
532 | |
533 | va_start(args, message); | |
534 | (void) sprintf(buf, "ztest: "); | |
535 | /* LINTED */ | |
536 | (void) vsprintf(buf + strlen(buf), message, args); | |
537 | va_end(args); | |
538 | if (do_perror) { | |
539 | (void) snprintf(buf + strlen(buf), FATAL_MSG_SZ - strlen(buf), | |
540 | ": %s", strerror(save_errno)); | |
541 | } | |
542 | (void) fprintf(stderr, "%s\n", buf); | |
543 | fatal_msg = buf; /* to ease debugging */ | |
544 | if (ztest_dump_core) | |
545 | abort(); | |
546 | exit(3); | |
547 | } | |
548 | ||
549 | static int | |
550 | str2shift(const char *buf) | |
551 | { | |
552 | const char *ends = "BKMGTPEZ"; | |
553 | int i; | |
554 | ||
555 | if (buf[0] == '\0') | |
556 | return (0); | |
557 | for (i = 0; i < strlen(ends); i++) { | |
558 | if (toupper(buf[0]) == ends[i]) | |
559 | break; | |
560 | } | |
561 | if (i == strlen(ends)) { | |
562 | (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", | |
563 | buf); | |
564 | usage(B_FALSE); | |
565 | } | |
566 | if (buf[1] == '\0' || (toupper(buf[1]) == 'B' && buf[2] == '\0')) { | |
567 | return (10*i); | |
568 | } | |
569 | (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", buf); | |
570 | usage(B_FALSE); | |
571 | /* NOTREACHED */ | |
572 | } | |
573 | ||
574 | static uint64_t | |
575 | nicenumtoull(const char *buf) | |
576 | { | |
577 | char *end; | |
578 | uint64_t val; | |
579 | ||
580 | val = strtoull(buf, &end, 0); | |
581 | if (end == buf) { | |
582 | (void) fprintf(stderr, "ztest: bad numeric value: %s\n", buf); | |
583 | usage(B_FALSE); | |
584 | } else if (end[0] == '.') { | |
585 | double fval = strtod(buf, &end); | |
586 | fval *= pow(2, str2shift(end)); | |
587 | if (fval > UINT64_MAX) { | |
588 | (void) fprintf(stderr, "ztest: value too large: %s\n", | |
589 | buf); | |
590 | usage(B_FALSE); | |
591 | } | |
592 | val = (uint64_t)fval; | |
593 | } else { | |
594 | int shift = str2shift(end); | |
595 | if (shift >= 64 || (val << shift) >> shift != val) { | |
596 | (void) fprintf(stderr, "ztest: value too large: %s\n", | |
597 | buf); | |
598 | usage(B_FALSE); | |
599 | } | |
600 | val <<= shift; | |
601 | } | |
602 | return (val); | |
603 | } | |
604 | ||
605 | static void | |
606 | usage(boolean_t requested) | |
607 | { | |
c242c188 CS |
608 | const ztest_shared_opts_t *zo = &ztest_opts_defaults; |
609 | ||
34dc7c2f BB |
610 | char nice_vdev_size[10]; |
611 | char nice_gang_bang[10]; | |
612 | FILE *fp = requested ? stdout : stderr; | |
613 | ||
c242c188 CS |
614 | nicenum(zo->zo_vdev_size, nice_vdev_size); |
615 | nicenum(zo->zo_metaslab_gang_bang, nice_gang_bang); | |
34dc7c2f BB |
616 | |
617 | (void) fprintf(fp, "Usage: %s\n" | |
618 | "\t[-v vdevs (default: %llu)]\n" | |
619 | "\t[-s size_of_each_vdev (default: %s)]\n" | |
428870ff | 620 | "\t[-a alignment_shift (default: %d)] use 0 for random\n" |
34dc7c2f BB |
621 | "\t[-m mirror_copies (default: %d)]\n" |
622 | "\t[-r raidz_disks (default: %d)]\n" | |
623 | "\t[-R raidz_parity (default: %d)]\n" | |
624 | "\t[-d datasets (default: %d)]\n" | |
625 | "\t[-t threads (default: %d)]\n" | |
626 | "\t[-g gang_block_threshold (default: %s)]\n" | |
428870ff BB |
627 | "\t[-i init_count (default: %d)] initialize pool i times\n" |
628 | "\t[-k kill_percentage (default: %llu%%)]\n" | |
34dc7c2f | 629 | "\t[-p pool_name (default: %s)]\n" |
428870ff | 630 | "\t[-f dir (default: %s)] file directory for vdev files\n" |
379ca9cf | 631 | "\t[-M] Multi-host simulate pool imported on remote host\n" |
428870ff BB |
632 | "\t[-V] verbose (use multiple times for ever more blather)\n" |
633 | "\t[-E] use existing pool instead of creating new one\n" | |
634 | "\t[-T time (default: %llu sec)] total run time\n" | |
635 | "\t[-F freezeloops (default: %llu)] max loops in spa_freeze()\n" | |
636 | "\t[-P passtime (default: %llu sec)] time per pass\n" | |
c242c188 | 637 | "\t[-B alt_ztest (default: <none>)] alternate ztest path\n" |
ed828c0c GM |
638 | "\t[-o variable=value] ... set global variable to an unsigned\n" |
639 | "\t 32-bit integer value\n" | |
34dc7c2f BB |
640 | "\t[-h] (print help)\n" |
641 | "", | |
c242c188 CS |
642 | zo->zo_pool, |
643 | (u_longlong_t)zo->zo_vdevs, /* -v */ | |
34dc7c2f | 644 | nice_vdev_size, /* -s */ |
c242c188 CS |
645 | zo->zo_ashift, /* -a */ |
646 | zo->zo_mirrors, /* -m */ | |
647 | zo->zo_raidz, /* -r */ | |
648 | zo->zo_raidz_parity, /* -R */ | |
649 | zo->zo_datasets, /* -d */ | |
650 | zo->zo_threads, /* -t */ | |
34dc7c2f | 651 | nice_gang_bang, /* -g */ |
c242c188 CS |
652 | zo->zo_init, /* -i */ |
653 | (u_longlong_t)zo->zo_killrate, /* -k */ | |
654 | zo->zo_pool, /* -p */ | |
655 | zo->zo_dir, /* -f */ | |
656 | (u_longlong_t)zo->zo_time, /* -T */ | |
657 | (u_longlong_t)zo->zo_maxloops, /* -F */ | |
658 | (u_longlong_t)zo->zo_passtime); | |
34dc7c2f BB |
659 | exit(requested ? 0 : 1); |
660 | } | |
661 | ||
34dc7c2f BB |
662 | static void |
663 | process_options(int argc, char **argv) | |
664 | { | |
c242c188 CS |
665 | char *path; |
666 | ztest_shared_opts_t *zo = &ztest_opts; | |
667 | ||
34dc7c2f BB |
668 | int opt; |
669 | uint64_t value; | |
c242c188 | 670 | char altdir[MAXNAMELEN] = { 0 }; |
34dc7c2f | 671 | |
c242c188 | 672 | bcopy(&ztest_opts_defaults, zo, sizeof (*zo)); |
34dc7c2f | 673 | |
34dc7c2f | 674 | while ((opt = getopt(argc, argv, |
379ca9cf | 675 | "v:s:a:m:r:R:d:t:g:i:k:p:f:MVET:P:hF:B:o:")) != EOF) { |
34dc7c2f BB |
676 | value = 0; |
677 | switch (opt) { | |
678 | case 'v': | |
679 | case 's': | |
680 | case 'a': | |
681 | case 'm': | |
682 | case 'r': | |
683 | case 'R': | |
684 | case 'd': | |
685 | case 't': | |
686 | case 'g': | |
687 | case 'i': | |
688 | case 'k': | |
689 | case 'T': | |
690 | case 'P': | |
428870ff | 691 | case 'F': |
34dc7c2f BB |
692 | value = nicenumtoull(optarg); |
693 | } | |
694 | switch (opt) { | |
695 | case 'v': | |
c242c188 | 696 | zo->zo_vdevs = value; |
34dc7c2f BB |
697 | break; |
698 | case 's': | |
c242c188 | 699 | zo->zo_vdev_size = MAX(SPA_MINDEVSIZE, value); |
34dc7c2f BB |
700 | break; |
701 | case 'a': | |
c242c188 | 702 | zo->zo_ashift = value; |
34dc7c2f BB |
703 | break; |
704 | case 'm': | |
c242c188 | 705 | zo->zo_mirrors = value; |
34dc7c2f BB |
706 | break; |
707 | case 'r': | |
c242c188 | 708 | zo->zo_raidz = MAX(1, value); |
34dc7c2f BB |
709 | break; |
710 | case 'R': | |
c242c188 | 711 | zo->zo_raidz_parity = MIN(MAX(value, 1), 3); |
34dc7c2f BB |
712 | break; |
713 | case 'd': | |
c242c188 | 714 | zo->zo_datasets = MAX(1, value); |
34dc7c2f BB |
715 | break; |
716 | case 't': | |
c242c188 | 717 | zo->zo_threads = MAX(1, value); |
34dc7c2f BB |
718 | break; |
719 | case 'g': | |
c242c188 CS |
720 | zo->zo_metaslab_gang_bang = MAX(SPA_MINBLOCKSIZE << 1, |
721 | value); | |
34dc7c2f BB |
722 | break; |
723 | case 'i': | |
c242c188 | 724 | zo->zo_init = value; |
34dc7c2f BB |
725 | break; |
726 | case 'k': | |
c242c188 | 727 | zo->zo_killrate = value; |
34dc7c2f BB |
728 | break; |
729 | case 'p': | |
c242c188 CS |
730 | (void) strlcpy(zo->zo_pool, optarg, |
731 | sizeof (zo->zo_pool)); | |
34dc7c2f BB |
732 | break; |
733 | case 'f': | |
c242c188 CS |
734 | path = realpath(optarg, NULL); |
735 | if (path == NULL) { | |
736 | (void) fprintf(stderr, "error: %s: %s\n", | |
737 | optarg, strerror(errno)); | |
738 | usage(B_FALSE); | |
739 | } else { | |
740 | (void) strlcpy(zo->zo_dir, path, | |
741 | sizeof (zo->zo_dir)); | |
df053d67 | 742 | free(path); |
c242c188 | 743 | } |
34dc7c2f | 744 | break; |
379ca9cf OF |
745 | case 'M': |
746 | zo->zo_mmp_test = 1; | |
747 | break; | |
34dc7c2f | 748 | case 'V': |
c242c188 | 749 | zo->zo_verbose++; |
34dc7c2f BB |
750 | break; |
751 | case 'E': | |
c242c188 | 752 | zo->zo_init = 0; |
34dc7c2f BB |
753 | break; |
754 | case 'T': | |
c242c188 | 755 | zo->zo_time = value; |
34dc7c2f BB |
756 | break; |
757 | case 'P': | |
c242c188 | 758 | zo->zo_passtime = MAX(1, value); |
34dc7c2f | 759 | break; |
428870ff | 760 | case 'F': |
c242c188 CS |
761 | zo->zo_maxloops = MAX(1, value); |
762 | break; | |
763 | case 'B': | |
764 | (void) strlcpy(altdir, optarg, sizeof (altdir)); | |
428870ff | 765 | break; |
ed828c0c GM |
766 | case 'o': |
767 | if (set_global_var(optarg) != 0) | |
768 | usage(B_FALSE); | |
769 | break; | |
34dc7c2f BB |
770 | case 'h': |
771 | usage(B_TRUE); | |
772 | break; | |
773 | case '?': | |
774 | default: | |
775 | usage(B_FALSE); | |
776 | break; | |
777 | } | |
778 | } | |
779 | ||
c242c188 | 780 | zo->zo_raidz_parity = MIN(zo->zo_raidz_parity, zo->zo_raidz - 1); |
34dc7c2f | 781 | |
c242c188 CS |
782 | zo->zo_vdevtime = |
783 | (zo->zo_vdevs > 0 ? zo->zo_time * NANOSEC / zo->zo_vdevs : | |
428870ff | 784 | UINT64_MAX >> 2); |
c242c188 CS |
785 | |
786 | if (strlen(altdir) > 0) { | |
ae380cfa BB |
787 | char *cmd; |
788 | char *realaltdir; | |
c242c188 CS |
789 | char *bin; |
790 | char *ztest; | |
791 | char *isa; | |
792 | int isalen; | |
793 | ||
ae380cfa BB |
794 | cmd = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
795 | realaltdir = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
796 | ||
5be98cfe | 797 | VERIFY(NULL != realpath(getexecname(), cmd)); |
c242c188 CS |
798 | if (0 != access(altdir, F_OK)) { |
799 | ztest_dump_core = B_FALSE; | |
800 | fatal(B_TRUE, "invalid alternate ztest path: %s", | |
801 | altdir); | |
802 | } | |
803 | VERIFY(NULL != realpath(altdir, realaltdir)); | |
804 | ||
805 | /* | |
806 | * 'cmd' should be of the form "<anything>/usr/bin/<isa>/ztest". | |
807 | * We want to extract <isa> to determine if we should use | |
808 | * 32 or 64 bit binaries. | |
809 | */ | |
810 | bin = strstr(cmd, "/usr/bin/"); | |
811 | ztest = strstr(bin, "/ztest"); | |
812 | isa = bin + 9; | |
813 | isalen = ztest - isa; | |
814 | (void) snprintf(zo->zo_alt_ztest, sizeof (zo->zo_alt_ztest), | |
815 | "%s/usr/bin/%.*s/ztest", realaltdir, isalen, isa); | |
816 | (void) snprintf(zo->zo_alt_libpath, sizeof (zo->zo_alt_libpath), | |
817 | "%s/usr/lib/%.*s", realaltdir, isalen, isa); | |
818 | ||
819 | if (0 != access(zo->zo_alt_ztest, X_OK)) { | |
820 | ztest_dump_core = B_FALSE; | |
821 | fatal(B_TRUE, "invalid alternate ztest: %s", | |
822 | zo->zo_alt_ztest); | |
823 | } else if (0 != access(zo->zo_alt_libpath, X_OK)) { | |
824 | ztest_dump_core = B_FALSE; | |
825 | fatal(B_TRUE, "invalid alternate lib directory %s", | |
826 | zo->zo_alt_libpath); | |
827 | } | |
ae380cfa BB |
828 | |
829 | umem_free(cmd, MAXPATHLEN); | |
830 | umem_free(realaltdir, MAXPATHLEN); | |
c242c188 | 831 | } |
428870ff BB |
832 | } |
833 | ||
834 | static void | |
835 | ztest_kill(ztest_shared_t *zs) | |
836 | { | |
c242c188 CS |
837 | zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(ztest_spa)); |
838 | zs->zs_space = metaslab_class_get_space(spa_normal_class(ztest_spa)); | |
5d1f7fb6 GW |
839 | |
840 | /* | |
841 | * Before we kill off ztest, make sure that the config is updated. | |
842 | * See comment above spa_config_sync(). | |
843 | */ | |
844 | mutex_enter(&spa_namespace_lock); | |
845 | spa_config_sync(ztest_spa, B_FALSE, B_FALSE); | |
846 | mutex_exit(&spa_namespace_lock); | |
847 | ||
428870ff BB |
848 | (void) kill(getpid(), SIGKILL); |
849 | } | |
850 | ||
851 | static uint64_t | |
852 | ztest_random(uint64_t range) | |
853 | { | |
854 | uint64_t r; | |
855 | ||
9d81146b ED |
856 | ASSERT3S(ztest_fd_rand, >=, 0); |
857 | ||
428870ff BB |
858 | if (range == 0) |
859 | return (0); | |
860 | ||
9d81146b | 861 | if (read(ztest_fd_rand, &r, sizeof (r)) != sizeof (r)) |
428870ff BB |
862 | fatal(1, "short read from /dev/urandom"); |
863 | ||
864 | return (r % range); | |
865 | } | |
866 | ||
867 | /* ARGSUSED */ | |
868 | static void | |
869 | ztest_record_enospc(const char *s) | |
870 | { | |
871 | ztest_shared->zs_enospc_count++; | |
34dc7c2f BB |
872 | } |
873 | ||
874 | static uint64_t | |
875 | ztest_get_ashift(void) | |
876 | { | |
c242c188 | 877 | if (ztest_opts.zo_ashift == 0) |
b02fe35d | 878 | return (SPA_MINBLOCKSHIFT + ztest_random(5)); |
c242c188 | 879 | return (ztest_opts.zo_ashift); |
34dc7c2f BB |
880 | } |
881 | ||
882 | static nvlist_t * | |
ea0b2538 | 883 | make_vdev_file(char *path, char *aux, char *pool, size_t size, uint64_t ashift) |
34dc7c2f | 884 | { |
40b84e7a | 885 | char *pathbuf; |
34dc7c2f | 886 | uint64_t vdev; |
34dc7c2f BB |
887 | nvlist_t *file; |
888 | ||
40b84e7a BB |
889 | pathbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
890 | ||
b128c09f BB |
891 | if (ashift == 0) |
892 | ashift = ztest_get_ashift(); | |
893 | ||
894 | if (path == NULL) { | |
895 | path = pathbuf; | |
896 | ||
897 | if (aux != NULL) { | |
898 | vdev = ztest_shared->zs_vdev_aux; | |
c242c188 CS |
899 | (void) snprintf(path, MAXPATHLEN, |
900 | ztest_aux_template, ztest_opts.zo_dir, | |
ea0b2538 GW |
901 | pool == NULL ? ztest_opts.zo_pool : pool, |
902 | aux, vdev); | |
b128c09f | 903 | } else { |
428870ff | 904 | vdev = ztest_shared->zs_vdev_next_leaf++; |
c242c188 CS |
905 | (void) snprintf(path, MAXPATHLEN, |
906 | ztest_dev_template, ztest_opts.zo_dir, | |
ea0b2538 | 907 | pool == NULL ? ztest_opts.zo_pool : pool, vdev); |
b128c09f BB |
908 | } |
909 | } | |
34dc7c2f | 910 | |
b128c09f BB |
911 | if (size != 0) { |
912 | int fd = open(path, O_RDWR | O_CREAT | O_TRUNC, 0666); | |
34dc7c2f | 913 | if (fd == -1) |
b128c09f | 914 | fatal(1, "can't open %s", path); |
34dc7c2f | 915 | if (ftruncate(fd, size) != 0) |
b128c09f | 916 | fatal(1, "can't ftruncate %s", path); |
34dc7c2f BB |
917 | (void) close(fd); |
918 | } | |
919 | ||
920 | VERIFY(nvlist_alloc(&file, NV_UNIQUE_NAME, 0) == 0); | |
921 | VERIFY(nvlist_add_string(file, ZPOOL_CONFIG_TYPE, VDEV_TYPE_FILE) == 0); | |
b128c09f | 922 | VERIFY(nvlist_add_string(file, ZPOOL_CONFIG_PATH, path) == 0); |
34dc7c2f | 923 | VERIFY(nvlist_add_uint64(file, ZPOOL_CONFIG_ASHIFT, ashift) == 0); |
40b84e7a | 924 | umem_free(pathbuf, MAXPATHLEN); |
34dc7c2f BB |
925 | |
926 | return (file); | |
927 | } | |
928 | ||
929 | static nvlist_t * | |
ea0b2538 GW |
930 | make_vdev_raidz(char *path, char *aux, char *pool, size_t size, |
931 | uint64_t ashift, int r) | |
34dc7c2f BB |
932 | { |
933 | nvlist_t *raidz, **child; | |
934 | int c; | |
935 | ||
936 | if (r < 2) | |
ea0b2538 | 937 | return (make_vdev_file(path, aux, pool, size, ashift)); |
34dc7c2f BB |
938 | child = umem_alloc(r * sizeof (nvlist_t *), UMEM_NOFAIL); |
939 | ||
940 | for (c = 0; c < r; c++) | |
ea0b2538 | 941 | child[c] = make_vdev_file(path, aux, pool, size, ashift); |
34dc7c2f BB |
942 | |
943 | VERIFY(nvlist_alloc(&raidz, NV_UNIQUE_NAME, 0) == 0); | |
944 | VERIFY(nvlist_add_string(raidz, ZPOOL_CONFIG_TYPE, | |
945 | VDEV_TYPE_RAIDZ) == 0); | |
946 | VERIFY(nvlist_add_uint64(raidz, ZPOOL_CONFIG_NPARITY, | |
c242c188 | 947 | ztest_opts.zo_raidz_parity) == 0); |
34dc7c2f BB |
948 | VERIFY(nvlist_add_nvlist_array(raidz, ZPOOL_CONFIG_CHILDREN, |
949 | child, r) == 0); | |
950 | ||
951 | for (c = 0; c < r; c++) | |
952 | nvlist_free(child[c]); | |
953 | ||
954 | umem_free(child, r * sizeof (nvlist_t *)); | |
955 | ||
956 | return (raidz); | |
957 | } | |
958 | ||
959 | static nvlist_t * | |
ea0b2538 GW |
960 | make_vdev_mirror(char *path, char *aux, char *pool, size_t size, |
961 | uint64_t ashift, int r, int m) | |
34dc7c2f BB |
962 | { |
963 | nvlist_t *mirror, **child; | |
964 | int c; | |
965 | ||
966 | if (m < 1) | |
ea0b2538 | 967 | return (make_vdev_raidz(path, aux, pool, size, ashift, r)); |
34dc7c2f BB |
968 | |
969 | child = umem_alloc(m * sizeof (nvlist_t *), UMEM_NOFAIL); | |
970 | ||
971 | for (c = 0; c < m; c++) | |
ea0b2538 | 972 | child[c] = make_vdev_raidz(path, aux, pool, size, ashift, r); |
34dc7c2f BB |
973 | |
974 | VERIFY(nvlist_alloc(&mirror, NV_UNIQUE_NAME, 0) == 0); | |
975 | VERIFY(nvlist_add_string(mirror, ZPOOL_CONFIG_TYPE, | |
976 | VDEV_TYPE_MIRROR) == 0); | |
977 | VERIFY(nvlist_add_nvlist_array(mirror, ZPOOL_CONFIG_CHILDREN, | |
978 | child, m) == 0); | |
34dc7c2f BB |
979 | |
980 | for (c = 0; c < m; c++) | |
981 | nvlist_free(child[c]); | |
982 | ||
983 | umem_free(child, m * sizeof (nvlist_t *)); | |
984 | ||
985 | return (mirror); | |
986 | } | |
987 | ||
988 | static nvlist_t * | |
ea0b2538 GW |
989 | make_vdev_root(char *path, char *aux, char *pool, size_t size, uint64_t ashift, |
990 | int log, int r, int m, int t) | |
34dc7c2f BB |
991 | { |
992 | nvlist_t *root, **child; | |
993 | int c; | |
994 | ||
995 | ASSERT(t > 0); | |
996 | ||
997 | child = umem_alloc(t * sizeof (nvlist_t *), UMEM_NOFAIL); | |
998 | ||
b128c09f | 999 | for (c = 0; c < t; c++) { |
ea0b2538 GW |
1000 | child[c] = make_vdev_mirror(path, aux, pool, size, ashift, |
1001 | r, m); | |
b128c09f BB |
1002 | VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_IS_LOG, |
1003 | log) == 0); | |
1004 | } | |
34dc7c2f BB |
1005 | |
1006 | VERIFY(nvlist_alloc(&root, NV_UNIQUE_NAME, 0) == 0); | |
1007 | VERIFY(nvlist_add_string(root, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT) == 0); | |
b128c09f | 1008 | VERIFY(nvlist_add_nvlist_array(root, aux ? aux : ZPOOL_CONFIG_CHILDREN, |
34dc7c2f BB |
1009 | child, t) == 0); |
1010 | ||
1011 | for (c = 0; c < t; c++) | |
1012 | nvlist_free(child[c]); | |
1013 | ||
1014 | umem_free(child, t * sizeof (nvlist_t *)); | |
1015 | ||
1016 | return (root); | |
1017 | } | |
1018 | ||
ea0b2538 GW |
1019 | /* |
1020 | * Find a random spa version. Returns back a random spa version in the | |
1021 | * range [initial_version, SPA_VERSION_FEATURES]. | |
1022 | */ | |
1023 | static uint64_t | |
1024 | ztest_random_spa_version(uint64_t initial_version) | |
1025 | { | |
1026 | uint64_t version = initial_version; | |
1027 | ||
1028 | if (version <= SPA_VERSION_BEFORE_FEATURES) { | |
1029 | version = version + | |
1030 | ztest_random(SPA_VERSION_BEFORE_FEATURES - version + 1); | |
1031 | } | |
1032 | ||
1033 | if (version > SPA_VERSION_BEFORE_FEATURES) | |
1034 | version = SPA_VERSION_FEATURES; | |
1035 | ||
1036 | ASSERT(SPA_VERSION_IS_SUPPORTED(version)); | |
1037 | return (version); | |
1038 | } | |
1039 | ||
428870ff BB |
1040 | static int |
1041 | ztest_random_blocksize(void) | |
34dc7c2f | 1042 | { |
f1512ee6 MA |
1043 | /* |
1044 | * Choose a block size >= the ashift. | |
1045 | * If the SPA supports new MAXBLOCKSIZE, test up to 1MB blocks. | |
1046 | */ | |
1047 | int maxbs = SPA_OLD_MAXBLOCKSHIFT; | |
1048 | if (spa_maxblocksize(ztest_spa) == SPA_MAXBLOCKSIZE) | |
1049 | maxbs = 20; | |
c3520e7f MA |
1050 | uint64_t block_shift = |
1051 | ztest_random(maxbs - ztest_spa->spa_max_ashift + 1); | |
b02fe35d | 1052 | return (1 << (SPA_MINBLOCKSHIFT + block_shift)); |
428870ff | 1053 | } |
34dc7c2f | 1054 | |
50c957f7 NB |
1055 | static int |
1056 | ztest_random_dnodesize(void) | |
1057 | { | |
1058 | int slots; | |
1059 | int max_slots = spa_maxdnodesize(ztest_spa) >> DNODE_SHIFT; | |
1060 | ||
1061 | if (max_slots == DNODE_MIN_SLOTS) | |
1062 | return (DNODE_MIN_SIZE); | |
1063 | ||
1064 | /* | |
1065 | * Weight the random distribution more heavily toward smaller | |
1066 | * dnode sizes since that is more likely to reflect real-world | |
1067 | * usage. | |
1068 | */ | |
1069 | ASSERT3U(max_slots, >, 4); | |
1070 | switch (ztest_random(10)) { | |
1071 | case 0: | |
1072 | slots = 5 + ztest_random(max_slots - 4); | |
1073 | break; | |
1074 | case 1 ... 4: | |
1075 | slots = 2 + ztest_random(3); | |
1076 | break; | |
1077 | default: | |
1078 | slots = 1; | |
1079 | break; | |
1080 | } | |
1081 | ||
1082 | return (slots << DNODE_SHIFT); | |
1083 | } | |
1084 | ||
428870ff BB |
1085 | static int |
1086 | ztest_random_ibshift(void) | |
1087 | { | |
1088 | return (DN_MIN_INDBLKSHIFT + | |
1089 | ztest_random(DN_MAX_INDBLKSHIFT - DN_MIN_INDBLKSHIFT + 1)); | |
34dc7c2f BB |
1090 | } |
1091 | ||
428870ff BB |
1092 | static uint64_t |
1093 | ztest_random_vdev_top(spa_t *spa, boolean_t log_ok) | |
34dc7c2f | 1094 | { |
428870ff BB |
1095 | uint64_t top; |
1096 | vdev_t *rvd = spa->spa_root_vdev; | |
1097 | vdev_t *tvd; | |
34dc7c2f | 1098 | |
428870ff | 1099 | ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); |
34dc7c2f | 1100 | |
428870ff BB |
1101 | do { |
1102 | top = ztest_random(rvd->vdev_children); | |
1103 | tvd = rvd->vdev_child[top]; | |
1104 | } while (tvd->vdev_ishole || (tvd->vdev_islog && !log_ok) || | |
1105 | tvd->vdev_mg == NULL || tvd->vdev_mg->mg_class == NULL); | |
34dc7c2f | 1106 | |
428870ff | 1107 | return (top); |
34dc7c2f BB |
1108 | } |
1109 | ||
428870ff BB |
1110 | static uint64_t |
1111 | ztest_random_dsl_prop(zfs_prop_t prop) | |
34dc7c2f | 1112 | { |
428870ff BB |
1113 | uint64_t value; |
1114 | ||
1115 | do { | |
1116 | value = zfs_prop_random_value(prop, ztest_random(-1ULL)); | |
1117 | } while (prop == ZFS_PROP_CHECKSUM && value == ZIO_CHECKSUM_OFF); | |
1118 | ||
1119 | return (value); | |
34dc7c2f BB |
1120 | } |
1121 | ||
34dc7c2f | 1122 | static int |
428870ff BB |
1123 | ztest_dsl_prop_set_uint64(char *osname, zfs_prop_t prop, uint64_t value, |
1124 | boolean_t inherit) | |
34dc7c2f | 1125 | { |
428870ff BB |
1126 | const char *propname = zfs_prop_to_name(prop); |
1127 | const char *valname; | |
40b84e7a | 1128 | char *setpoint; |
428870ff | 1129 | uint64_t curval; |
34dc7c2f BB |
1130 | int error; |
1131 | ||
13fe0198 MA |
1132 | error = dsl_prop_set_int(osname, propname, |
1133 | (inherit ? ZPROP_SRC_NONE : ZPROP_SRC_LOCAL), value); | |
34dc7c2f | 1134 | |
428870ff BB |
1135 | if (error == ENOSPC) { |
1136 | ztest_record_enospc(FTAG); | |
34dc7c2f BB |
1137 | return (error); |
1138 | } | |
c99c9001 | 1139 | ASSERT0(error); |
34dc7c2f | 1140 | |
40b84e7a | 1141 | setpoint = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
13fe0198 | 1142 | VERIFY0(dsl_prop_get_integer(osname, propname, &curval, setpoint)); |
428870ff | 1143 | |
c242c188 | 1144 | if (ztest_opts.zo_verbose >= 6) { |
6bec4351 TC |
1145 | int err; |
1146 | ||
1147 | err = zfs_prop_index_to_string(prop, curval, &valname); | |
1148 | if (err) | |
02730c33 BB |
1149 | (void) printf("%s %s = %llu at '%s'\n", osname, |
1150 | propname, (unsigned long long)curval, setpoint); | |
6bec4351 TC |
1151 | else |
1152 | (void) printf("%s %s = %s at '%s'\n", | |
1153 | osname, propname, valname, setpoint); | |
34dc7c2f | 1154 | } |
40b84e7a | 1155 | umem_free(setpoint, MAXPATHLEN); |
34dc7c2f BB |
1156 | |
1157 | return (error); | |
1158 | } | |
1159 | ||
1160 | static int | |
c242c188 | 1161 | ztest_spa_prop_set_uint64(zpool_prop_t prop, uint64_t value) |
34dc7c2f | 1162 | { |
c242c188 | 1163 | spa_t *spa = ztest_spa; |
428870ff | 1164 | nvlist_t *props = NULL; |
34dc7c2f BB |
1165 | int error; |
1166 | ||
428870ff BB |
1167 | VERIFY(nvlist_alloc(&props, NV_UNIQUE_NAME, 0) == 0); |
1168 | VERIFY(nvlist_add_uint64(props, zpool_prop_to_name(prop), value) == 0); | |
34dc7c2f | 1169 | |
428870ff BB |
1170 | error = spa_prop_set(spa, props); |
1171 | ||
1172 | nvlist_free(props); | |
1173 | ||
1174 | if (error == ENOSPC) { | |
1175 | ztest_record_enospc(FTAG); | |
34dc7c2f BB |
1176 | return (error); |
1177 | } | |
c99c9001 | 1178 | ASSERT0(error); |
34dc7c2f BB |
1179 | |
1180 | return (error); | |
1181 | } | |
1182 | ||
e3a07cd0 BP |
1183 | |
1184 | /* | |
1185 | * Object and range lock mechanics | |
1186 | */ | |
1187 | typedef struct { | |
1188 | list_node_t z_lnode; | |
1189 | refcount_t z_refcnt; | |
1190 | uint64_t z_object; | |
1191 | zfs_rlock_t z_range_lock; | |
1192 | } ztest_znode_t; | |
1193 | ||
1194 | typedef struct { | |
1195 | rl_t *z_rl; | |
1196 | ztest_znode_t *z_ztznode; | |
1197 | } ztest_zrl_t; | |
1198 | ||
1199 | static ztest_znode_t * | |
1200 | ztest_znode_init(uint64_t object) | |
1201 | { | |
1202 | ztest_znode_t *zp = umem_alloc(sizeof (*zp), UMEM_NOFAIL); | |
1203 | ||
1204 | list_link_init(&zp->z_lnode); | |
1205 | refcount_create(&zp->z_refcnt); | |
1206 | zp->z_object = object; | |
1207 | zfs_rlock_init(&zp->z_range_lock); | |
1208 | ||
1209 | return (zp); | |
1210 | } | |
1211 | ||
1212 | static void | |
1213 | ztest_znode_fini(ztest_znode_t *zp) | |
1214 | { | |
1215 | ASSERT(refcount_is_zero(&zp->z_refcnt)); | |
1216 | zfs_rlock_destroy(&zp->z_range_lock); | |
1217 | zp->z_object = 0; | |
1218 | refcount_destroy(&zp->z_refcnt); | |
1219 | list_link_init(&zp->z_lnode); | |
1220 | umem_free(zp, sizeof (*zp)); | |
1221 | } | |
1222 | ||
1223 | static void | |
1224 | ztest_zll_init(zll_t *zll) | |
1225 | { | |
1226 | mutex_init(&zll->z_lock, NULL, MUTEX_DEFAULT, NULL); | |
1227 | list_create(&zll->z_list, sizeof (ztest_znode_t), | |
1228 | offsetof(ztest_znode_t, z_lnode)); | |
1229 | } | |
1230 | ||
1231 | static void | |
1232 | ztest_zll_destroy(zll_t *zll) | |
1233 | { | |
1234 | list_destroy(&zll->z_list); | |
1235 | mutex_destroy(&zll->z_lock); | |
1236 | } | |
1237 | ||
1238 | #define RL_TAG "range_lock" | |
1239 | static ztest_znode_t * | |
1240 | ztest_znode_get(ztest_ds_t *zd, uint64_t object) | |
1241 | { | |
1242 | zll_t *zll = &zd->zd_range_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; | |
1243 | ztest_znode_t *zp = NULL; | |
1244 | mutex_enter(&zll->z_lock); | |
1245 | for (zp = list_head(&zll->z_list); (zp); | |
1246 | zp = list_next(&zll->z_list, zp)) { | |
1247 | if (zp->z_object == object) { | |
1248 | refcount_add(&zp->z_refcnt, RL_TAG); | |
1249 | break; | |
1250 | } | |
1251 | } | |
1252 | if (zp == NULL) { | |
1253 | zp = ztest_znode_init(object); | |
1254 | refcount_add(&zp->z_refcnt, RL_TAG); | |
1255 | list_insert_head(&zll->z_list, zp); | |
1256 | } | |
1257 | mutex_exit(&zll->z_lock); | |
1258 | return (zp); | |
1259 | } | |
1260 | ||
1261 | static void | |
1262 | ztest_znode_put(ztest_ds_t *zd, ztest_znode_t *zp) | |
1263 | { | |
1264 | zll_t *zll = NULL; | |
1265 | ASSERT3U(zp->z_object, !=, 0); | |
1266 | zll = &zd->zd_range_lock[zp->z_object & (ZTEST_OBJECT_LOCKS - 1)]; | |
1267 | mutex_enter(&zll->z_lock); | |
1268 | refcount_remove(&zp->z_refcnt, RL_TAG); | |
1269 | if (refcount_is_zero(&zp->z_refcnt)) { | |
1270 | list_remove(&zll->z_list, zp); | |
1271 | ztest_znode_fini(zp); | |
1272 | } | |
1273 | mutex_exit(&zll->z_lock); | |
1274 | } | |
1275 | ||
1276 | ||
428870ff BB |
1277 | static void |
1278 | ztest_rll_init(rll_t *rll) | |
1279 | { | |
1280 | rll->rll_writer = NULL; | |
1281 | rll->rll_readers = 0; | |
1e33ac1e BB |
1282 | mutex_init(&rll->rll_lock, NULL, MUTEX_DEFAULT, NULL); |
1283 | cv_init(&rll->rll_cv, NULL, CV_DEFAULT, NULL); | |
428870ff | 1284 | } |
34dc7c2f | 1285 | |
428870ff BB |
1286 | static void |
1287 | ztest_rll_destroy(rll_t *rll) | |
34dc7c2f | 1288 | { |
428870ff BB |
1289 | ASSERT(rll->rll_writer == NULL); |
1290 | ASSERT(rll->rll_readers == 0); | |
1e33ac1e BB |
1291 | mutex_destroy(&rll->rll_lock); |
1292 | cv_destroy(&rll->rll_cv); | |
428870ff | 1293 | } |
34dc7c2f | 1294 | |
428870ff BB |
1295 | static void |
1296 | ztest_rll_lock(rll_t *rll, rl_type_t type) | |
1297 | { | |
1e33ac1e | 1298 | mutex_enter(&rll->rll_lock); |
34dc7c2f | 1299 | |
428870ff BB |
1300 | if (type == RL_READER) { |
1301 | while (rll->rll_writer != NULL) | |
1e33ac1e | 1302 | (void) cv_wait(&rll->rll_cv, &rll->rll_lock); |
428870ff BB |
1303 | rll->rll_readers++; |
1304 | } else { | |
1305 | while (rll->rll_writer != NULL || rll->rll_readers) | |
1e33ac1e | 1306 | (void) cv_wait(&rll->rll_cv, &rll->rll_lock); |
428870ff BB |
1307 | rll->rll_writer = curthread; |
1308 | } | |
34dc7c2f | 1309 | |
1e33ac1e | 1310 | mutex_exit(&rll->rll_lock); |
428870ff | 1311 | } |
34dc7c2f | 1312 | |
428870ff BB |
1313 | static void |
1314 | ztest_rll_unlock(rll_t *rll) | |
1315 | { | |
1e33ac1e | 1316 | mutex_enter(&rll->rll_lock); |
34dc7c2f | 1317 | |
428870ff BB |
1318 | if (rll->rll_writer) { |
1319 | ASSERT(rll->rll_readers == 0); | |
1320 | rll->rll_writer = NULL; | |
1321 | } else { | |
1322 | ASSERT(rll->rll_readers != 0); | |
1323 | ASSERT(rll->rll_writer == NULL); | |
1324 | rll->rll_readers--; | |
1325 | } | |
34dc7c2f | 1326 | |
428870ff | 1327 | if (rll->rll_writer == NULL && rll->rll_readers == 0) |
1e33ac1e | 1328 | cv_broadcast(&rll->rll_cv); |
428870ff | 1329 | |
1e33ac1e | 1330 | mutex_exit(&rll->rll_lock); |
34dc7c2f BB |
1331 | } |
1332 | ||
428870ff BB |
1333 | static void |
1334 | ztest_object_lock(ztest_ds_t *zd, uint64_t object, rl_type_t type) | |
b128c09f | 1335 | { |
428870ff | 1336 | rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; |
b128c09f | 1337 | |
428870ff BB |
1338 | ztest_rll_lock(rll, type); |
1339 | } | |
b128c09f | 1340 | |
428870ff BB |
1341 | static void |
1342 | ztest_object_unlock(ztest_ds_t *zd, uint64_t object) | |
1343 | { | |
1344 | rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; | |
b128c09f | 1345 | |
428870ff | 1346 | ztest_rll_unlock(rll); |
b128c09f BB |
1347 | } |
1348 | ||
e3a07cd0 BP |
1349 | static ztest_zrl_t * |
1350 | ztest_zrl_init(rl_t *rl, ztest_znode_t *zp) | |
34dc7c2f | 1351 | { |
e3a07cd0 BP |
1352 | ztest_zrl_t *zrl = umem_alloc(sizeof (*zrl), UMEM_NOFAIL); |
1353 | zrl->z_rl = rl; | |
1354 | zrl->z_ztznode = zp; | |
1355 | return (zrl); | |
428870ff | 1356 | } |
34dc7c2f | 1357 | |
428870ff | 1358 | static void |
e3a07cd0 | 1359 | ztest_zrl_fini(ztest_zrl_t *zrl) |
428870ff | 1360 | { |
e3a07cd0 BP |
1361 | umem_free(zrl, sizeof (*zrl)); |
1362 | } | |
34dc7c2f | 1363 | |
e3a07cd0 BP |
1364 | static ztest_zrl_t * |
1365 | ztest_range_lock(ztest_ds_t *zd, uint64_t object, uint64_t offset, | |
1366 | uint64_t size, rl_type_t type) | |
1367 | { | |
1368 | ztest_znode_t *zp = ztest_znode_get(zd, object); | |
1369 | rl_t *rl = zfs_range_lock(&zp->z_range_lock, offset, | |
1370 | size, type); | |
1371 | return (ztest_zrl_init(rl, zp)); | |
1372 | } | |
34dc7c2f | 1373 | |
e3a07cd0 BP |
1374 | static void |
1375 | ztest_range_unlock(ztest_ds_t *zd, ztest_zrl_t *zrl) | |
1376 | { | |
1377 | zfs_range_unlock(zrl->z_rl); | |
1378 | ztest_znode_put(zd, zrl->z_ztznode); | |
1379 | ztest_zrl_fini(zrl); | |
428870ff | 1380 | } |
34dc7c2f | 1381 | |
428870ff | 1382 | static void |
c242c188 | 1383 | ztest_zd_init(ztest_ds_t *zd, ztest_shared_ds_t *szd, objset_t *os) |
428870ff BB |
1384 | { |
1385 | zd->zd_os = os; | |
1386 | zd->zd_zilog = dmu_objset_zil(os); | |
c242c188 | 1387 | zd->zd_shared = szd; |
428870ff | 1388 | dmu_objset_name(os, zd->zd_name); |
d6320ddb | 1389 | int l; |
428870ff | 1390 | |
c242c188 CS |
1391 | if (zd->zd_shared != NULL) |
1392 | zd->zd_shared->zd_seq = 0; | |
1393 | ||
7809eb8b | 1394 | VERIFY(rwlock_init(&zd->zd_zilog_lock, USYNC_THREAD, NULL) == 0); |
1e33ac1e | 1395 | mutex_init(&zd->zd_dirobj_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 1396 | |
d6320ddb | 1397 | for (l = 0; l < ZTEST_OBJECT_LOCKS; l++) |
428870ff | 1398 | ztest_rll_init(&zd->zd_object_lock[l]); |
34dc7c2f | 1399 | |
d6320ddb | 1400 | for (l = 0; l < ZTEST_RANGE_LOCKS; l++) |
e3a07cd0 | 1401 | ztest_zll_init(&zd->zd_range_lock[l]); |
34dc7c2f BB |
1402 | } |
1403 | ||
428870ff BB |
1404 | static void |
1405 | ztest_zd_fini(ztest_ds_t *zd) | |
34dc7c2f | 1406 | { |
d6320ddb BB |
1407 | int l; |
1408 | ||
1e33ac1e | 1409 | mutex_destroy(&zd->zd_dirobj_lock); |
7809eb8b | 1410 | (void) rwlock_destroy(&zd->zd_zilog_lock); |
34dc7c2f | 1411 | |
d6320ddb | 1412 | for (l = 0; l < ZTEST_OBJECT_LOCKS; l++) |
428870ff | 1413 | ztest_rll_destroy(&zd->zd_object_lock[l]); |
b128c09f | 1414 | |
d6320ddb | 1415 | for (l = 0; l < ZTEST_RANGE_LOCKS; l++) |
e3a07cd0 | 1416 | ztest_zll_destroy(&zd->zd_range_lock[l]); |
428870ff | 1417 | } |
b128c09f | 1418 | |
428870ff | 1419 | #define TXG_MIGHTWAIT (ztest_random(10) == 0 ? TXG_NOWAIT : TXG_WAIT) |
b128c09f | 1420 | |
428870ff BB |
1421 | static uint64_t |
1422 | ztest_tx_assign(dmu_tx_t *tx, uint64_t txg_how, const char *tag) | |
1423 | { | |
1424 | uint64_t txg; | |
1425 | int error; | |
1426 | ||
1427 | /* | |
1428 | * Attempt to assign tx to some transaction group. | |
1429 | */ | |
1430 | error = dmu_tx_assign(tx, txg_how); | |
1431 | if (error) { | |
1432 | if (error == ERESTART) { | |
1433 | ASSERT(txg_how == TXG_NOWAIT); | |
1434 | dmu_tx_wait(tx); | |
1435 | } else { | |
1436 | ASSERT3U(error, ==, ENOSPC); | |
1437 | ztest_record_enospc(tag); | |
1438 | } | |
1439 | dmu_tx_abort(tx); | |
1440 | return (0); | |
1441 | } | |
1442 | txg = dmu_tx_get_txg(tx); | |
1443 | ASSERT(txg != 0); | |
1444 | return (txg); | |
1445 | } | |
1446 | ||
1447 | static void | |
1448 | ztest_pattern_set(void *buf, uint64_t size, uint64_t value) | |
1449 | { | |
1450 | uint64_t *ip = buf; | |
1451 | uint64_t *ip_end = (uint64_t *)((uintptr_t)buf + (uintptr_t)size); | |
1452 | ||
1453 | while (ip < ip_end) | |
1454 | *ip++ = value; | |
1455 | } | |
1456 | ||
1fde1e37 | 1457 | #ifndef NDEBUG |
428870ff BB |
1458 | static boolean_t |
1459 | ztest_pattern_match(void *buf, uint64_t size, uint64_t value) | |
1460 | { | |
1461 | uint64_t *ip = buf; | |
1462 | uint64_t *ip_end = (uint64_t *)((uintptr_t)buf + (uintptr_t)size); | |
1463 | uint64_t diff = 0; | |
1464 | ||
1465 | while (ip < ip_end) | |
1466 | diff |= (value - *ip++); | |
1467 | ||
1468 | return (diff == 0); | |
1469 | } | |
1fde1e37 | 1470 | #endif |
428870ff BB |
1471 | |
1472 | static void | |
1473 | ztest_bt_generate(ztest_block_tag_t *bt, objset_t *os, uint64_t object, | |
50c957f7 NB |
1474 | uint64_t dnodesize, uint64_t offset, uint64_t gen, uint64_t txg, |
1475 | uint64_t crtxg) | |
428870ff BB |
1476 | { |
1477 | bt->bt_magic = BT_MAGIC; | |
1478 | bt->bt_objset = dmu_objset_id(os); | |
1479 | bt->bt_object = object; | |
50c957f7 | 1480 | bt->bt_dnodesize = dnodesize; |
428870ff BB |
1481 | bt->bt_offset = offset; |
1482 | bt->bt_gen = gen; | |
1483 | bt->bt_txg = txg; | |
1484 | bt->bt_crtxg = crtxg; | |
1485 | } | |
1486 | ||
1487 | static void | |
1488 | ztest_bt_verify(ztest_block_tag_t *bt, objset_t *os, uint64_t object, | |
50c957f7 NB |
1489 | uint64_t dnodesize, uint64_t offset, uint64_t gen, uint64_t txg, |
1490 | uint64_t crtxg) | |
428870ff | 1491 | { |
9b67f605 MA |
1492 | ASSERT3U(bt->bt_magic, ==, BT_MAGIC); |
1493 | ASSERT3U(bt->bt_objset, ==, dmu_objset_id(os)); | |
1494 | ASSERT3U(bt->bt_object, ==, object); | |
50c957f7 | 1495 | ASSERT3U(bt->bt_dnodesize, ==, dnodesize); |
9b67f605 MA |
1496 | ASSERT3U(bt->bt_offset, ==, offset); |
1497 | ASSERT3U(bt->bt_gen, <=, gen); | |
1498 | ASSERT3U(bt->bt_txg, <=, txg); | |
1499 | ASSERT3U(bt->bt_crtxg, ==, crtxg); | |
428870ff BB |
1500 | } |
1501 | ||
1502 | static ztest_block_tag_t * | |
1503 | ztest_bt_bonus(dmu_buf_t *db) | |
1504 | { | |
1505 | dmu_object_info_t doi; | |
1506 | ztest_block_tag_t *bt; | |
1507 | ||
1508 | dmu_object_info_from_db(db, &doi); | |
1509 | ASSERT3U(doi.doi_bonus_size, <=, db->db_size); | |
1510 | ASSERT3U(doi.doi_bonus_size, >=, sizeof (*bt)); | |
1511 | bt = (void *)((char *)db->db_data + doi.doi_bonus_size - sizeof (*bt)); | |
1512 | ||
1513 | return (bt); | |
1514 | } | |
1515 | ||
50c957f7 NB |
1516 | /* |
1517 | * Generate a token to fill up unused bonus buffer space. Try to make | |
1518 | * it unique to the object, generation, and offset to verify that data | |
1519 | * is not getting overwritten by data from other dnodes. | |
1520 | */ | |
1521 | #define ZTEST_BONUS_FILL_TOKEN(obj, ds, gen, offset) \ | |
1522 | (((ds) << 48) | ((gen) << 32) | ((obj) << 8) | (offset)) | |
1523 | ||
1524 | /* | |
1525 | * Fill up the unused bonus buffer region before the block tag with a | |
1526 | * verifiable pattern. Filling the whole bonus area with non-zero data | |
1527 | * helps ensure that all dnode traversal code properly skips the | |
1528 | * interior regions of large dnodes. | |
1529 | */ | |
1530 | void | |
1531 | ztest_fill_unused_bonus(dmu_buf_t *db, void *end, uint64_t obj, | |
1532 | objset_t *os, uint64_t gen) | |
1533 | { | |
1534 | uint64_t *bonusp; | |
1535 | ||
1536 | ASSERT(IS_P2ALIGNED((char *)end - (char *)db->db_data, 8)); | |
1537 | ||
1538 | for (bonusp = db->db_data; bonusp < (uint64_t *)end; bonusp++) { | |
1539 | uint64_t token = ZTEST_BONUS_FILL_TOKEN(obj, dmu_objset_id(os), | |
1540 | gen, bonusp - (uint64_t *)db->db_data); | |
1541 | *bonusp = token; | |
1542 | } | |
1543 | } | |
1544 | ||
1545 | /* | |
1546 | * Verify that the unused area of a bonus buffer is filled with the | |
1547 | * expected tokens. | |
1548 | */ | |
1549 | void | |
1550 | ztest_verify_unused_bonus(dmu_buf_t *db, void *end, uint64_t obj, | |
1551 | objset_t *os, uint64_t gen) | |
1552 | { | |
1553 | uint64_t *bonusp; | |
1554 | ||
1555 | for (bonusp = db->db_data; bonusp < (uint64_t *)end; bonusp++) { | |
1556 | uint64_t token = ZTEST_BONUS_FILL_TOKEN(obj, dmu_objset_id(os), | |
1557 | gen, bonusp - (uint64_t *)db->db_data); | |
1558 | VERIFY3U(*bonusp, ==, token); | |
1559 | } | |
1560 | } | |
1561 | ||
428870ff BB |
1562 | /* |
1563 | * ZIL logging ops | |
1564 | */ | |
1565 | ||
1566 | #define lrz_type lr_mode | |
1567 | #define lrz_blocksize lr_uid | |
1568 | #define lrz_ibshift lr_gid | |
1569 | #define lrz_bonustype lr_rdev | |
50c957f7 | 1570 | #define lrz_dnodesize lr_crtime[1] |
428870ff | 1571 | |
572e2857 | 1572 | static void |
428870ff BB |
1573 | ztest_log_create(ztest_ds_t *zd, dmu_tx_t *tx, lr_create_t *lr) |
1574 | { | |
1575 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1576 | size_t namesize = strlen(name) + 1; | |
1577 | itx_t *itx; | |
1578 | ||
1579 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1580 | return; |
428870ff BB |
1581 | |
1582 | itx = zil_itx_create(TX_CREATE, sizeof (*lr) + namesize); | |
1583 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1584 | sizeof (*lr) + namesize - sizeof (lr_t)); | |
1585 | ||
572e2857 | 1586 | zil_itx_assign(zd->zd_zilog, itx, tx); |
428870ff BB |
1587 | } |
1588 | ||
572e2857 BB |
1589 | static void |
1590 | ztest_log_remove(ztest_ds_t *zd, dmu_tx_t *tx, lr_remove_t *lr, uint64_t object) | |
428870ff BB |
1591 | { |
1592 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1593 | size_t namesize = strlen(name) + 1; | |
1594 | itx_t *itx; | |
1595 | ||
1596 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1597 | return; |
428870ff BB |
1598 | |
1599 | itx = zil_itx_create(TX_REMOVE, sizeof (*lr) + namesize); | |
1600 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1601 | sizeof (*lr) + namesize - sizeof (lr_t)); | |
1602 | ||
572e2857 BB |
1603 | itx->itx_oid = object; |
1604 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1605 | } |
1606 | ||
572e2857 | 1607 | static void |
428870ff BB |
1608 | ztest_log_write(ztest_ds_t *zd, dmu_tx_t *tx, lr_write_t *lr) |
1609 | { | |
1610 | itx_t *itx; | |
1611 | itx_wr_state_t write_state = ztest_random(WR_NUM_STATES); | |
1612 | ||
1613 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1614 | return; |
428870ff BB |
1615 | |
1616 | if (lr->lr_length > ZIL_MAX_LOG_DATA) | |
1617 | write_state = WR_INDIRECT; | |
1618 | ||
1619 | itx = zil_itx_create(TX_WRITE, | |
1620 | sizeof (*lr) + (write_state == WR_COPIED ? lr->lr_length : 0)); | |
1621 | ||
1622 | if (write_state == WR_COPIED && | |
1623 | dmu_read(zd->zd_os, lr->lr_foid, lr->lr_offset, lr->lr_length, | |
1624 | ((lr_write_t *)&itx->itx_lr) + 1, DMU_READ_NO_PREFETCH) != 0) { | |
1625 | zil_itx_destroy(itx); | |
1626 | itx = zil_itx_create(TX_WRITE, sizeof (*lr)); | |
1627 | write_state = WR_NEED_COPY; | |
1628 | } | |
1629 | itx->itx_private = zd; | |
1630 | itx->itx_wr_state = write_state; | |
1631 | itx->itx_sync = (ztest_random(8) == 0); | |
428870ff BB |
1632 | |
1633 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1634 | sizeof (*lr) - sizeof (lr_t)); | |
1635 | ||
572e2857 | 1636 | zil_itx_assign(zd->zd_zilog, itx, tx); |
428870ff BB |
1637 | } |
1638 | ||
572e2857 | 1639 | static void |
428870ff BB |
1640 | ztest_log_truncate(ztest_ds_t *zd, dmu_tx_t *tx, lr_truncate_t *lr) |
1641 | { | |
1642 | itx_t *itx; | |
1643 | ||
1644 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1645 | return; |
428870ff BB |
1646 | |
1647 | itx = zil_itx_create(TX_TRUNCATE, sizeof (*lr)); | |
1648 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1649 | sizeof (*lr) - sizeof (lr_t)); | |
1650 | ||
572e2857 BB |
1651 | itx->itx_sync = B_FALSE; |
1652 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1653 | } |
1654 | ||
572e2857 | 1655 | static void |
428870ff BB |
1656 | ztest_log_setattr(ztest_ds_t *zd, dmu_tx_t *tx, lr_setattr_t *lr) |
1657 | { | |
1658 | itx_t *itx; | |
1659 | ||
1660 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1661 | return; |
428870ff BB |
1662 | |
1663 | itx = zil_itx_create(TX_SETATTR, sizeof (*lr)); | |
1664 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1665 | sizeof (*lr) - sizeof (lr_t)); | |
1666 | ||
572e2857 BB |
1667 | itx->itx_sync = B_FALSE; |
1668 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1669 | } |
1670 | ||
1671 | /* | |
1672 | * ZIL replay ops | |
1673 | */ | |
1674 | static int | |
1675 | ztest_replay_create(ztest_ds_t *zd, lr_create_t *lr, boolean_t byteswap) | |
1676 | { | |
1677 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1678 | objset_t *os = zd->zd_os; | |
1679 | ztest_block_tag_t *bbt; | |
1680 | dmu_buf_t *db; | |
1681 | dmu_tx_t *tx; | |
1682 | uint64_t txg; | |
1683 | int error = 0; | |
50c957f7 | 1684 | int bonuslen; |
428870ff BB |
1685 | |
1686 | if (byteswap) | |
1687 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1688 | ||
1689 | ASSERT(lr->lr_doid == ZTEST_DIROBJ); | |
1690 | ASSERT(name[0] != '\0'); | |
1691 | ||
1692 | tx = dmu_tx_create(os); | |
1693 | ||
1694 | dmu_tx_hold_zap(tx, lr->lr_doid, B_TRUE, name); | |
1695 | ||
1696 | if (lr->lrz_type == DMU_OT_ZAP_OTHER) { | |
1697 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, B_TRUE, NULL); | |
1698 | } else { | |
1699 | dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT); | |
1700 | } | |
1701 | ||
1702 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1703 | if (txg == 0) | |
1704 | return (ENOSPC); | |
1705 | ||
1706 | ASSERT(dmu_objset_zil(os)->zl_replay == !!lr->lr_foid); | |
50c957f7 | 1707 | bonuslen = DN_BONUS_SIZE(lr->lrz_dnodesize); |
428870ff BB |
1708 | |
1709 | if (lr->lrz_type == DMU_OT_ZAP_OTHER) { | |
1710 | if (lr->lr_foid == 0) { | |
50c957f7 | 1711 | lr->lr_foid = zap_create_dnsize(os, |
428870ff | 1712 | lr->lrz_type, lr->lrz_bonustype, |
50c957f7 | 1713 | bonuslen, lr->lrz_dnodesize, tx); |
428870ff | 1714 | } else { |
50c957f7 | 1715 | error = zap_create_claim_dnsize(os, lr->lr_foid, |
428870ff | 1716 | lr->lrz_type, lr->lrz_bonustype, |
50c957f7 | 1717 | bonuslen, lr->lrz_dnodesize, tx); |
428870ff BB |
1718 | } |
1719 | } else { | |
1720 | if (lr->lr_foid == 0) { | |
50c957f7 | 1721 | lr->lr_foid = dmu_object_alloc_dnsize(os, |
428870ff | 1722 | lr->lrz_type, 0, lr->lrz_bonustype, |
50c957f7 | 1723 | bonuslen, lr->lrz_dnodesize, tx); |
428870ff | 1724 | } else { |
50c957f7 | 1725 | error = dmu_object_claim_dnsize(os, lr->lr_foid, |
428870ff | 1726 | lr->lrz_type, 0, lr->lrz_bonustype, |
50c957f7 | 1727 | bonuslen, lr->lrz_dnodesize, tx); |
428870ff BB |
1728 | } |
1729 | } | |
1730 | ||
1731 | if (error) { | |
1732 | ASSERT3U(error, ==, EEXIST); | |
1733 | ASSERT(zd->zd_zilog->zl_replay); | |
1734 | dmu_tx_commit(tx); | |
1735 | return (error); | |
1736 | } | |
1737 | ||
1738 | ASSERT(lr->lr_foid != 0); | |
1739 | ||
1740 | if (lr->lrz_type != DMU_OT_ZAP_OTHER) | |
1741 | VERIFY3U(0, ==, dmu_object_set_blocksize(os, lr->lr_foid, | |
1742 | lr->lrz_blocksize, lr->lrz_ibshift, tx)); | |
1743 | ||
1744 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
1745 | bbt = ztest_bt_bonus(db); | |
1746 | dmu_buf_will_dirty(db, tx); | |
50c957f7 NB |
1747 | ztest_bt_generate(bbt, os, lr->lr_foid, lr->lrz_dnodesize, -1ULL, |
1748 | lr->lr_gen, txg, txg); | |
1749 | ztest_fill_unused_bonus(db, bbt, lr->lr_foid, os, lr->lr_gen); | |
428870ff BB |
1750 | dmu_buf_rele(db, FTAG); |
1751 | ||
1752 | VERIFY3U(0, ==, zap_add(os, lr->lr_doid, name, sizeof (uint64_t), 1, | |
1753 | &lr->lr_foid, tx)); | |
1754 | ||
1755 | (void) ztest_log_create(zd, tx, lr); | |
1756 | ||
1757 | dmu_tx_commit(tx); | |
1758 | ||
1759 | return (0); | |
1760 | } | |
1761 | ||
1762 | static int | |
1763 | ztest_replay_remove(ztest_ds_t *zd, lr_remove_t *lr, boolean_t byteswap) | |
1764 | { | |
1765 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1766 | objset_t *os = zd->zd_os; | |
1767 | dmu_object_info_t doi; | |
1768 | dmu_tx_t *tx; | |
1769 | uint64_t object, txg; | |
1770 | ||
1771 | if (byteswap) | |
1772 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1773 | ||
1774 | ASSERT(lr->lr_doid == ZTEST_DIROBJ); | |
1775 | ASSERT(name[0] != '\0'); | |
1776 | ||
1777 | VERIFY3U(0, ==, | |
1778 | zap_lookup(os, lr->lr_doid, name, sizeof (object), 1, &object)); | |
1779 | ASSERT(object != 0); | |
1780 | ||
1781 | ztest_object_lock(zd, object, RL_WRITER); | |
1782 | ||
1783 | VERIFY3U(0, ==, dmu_object_info(os, object, &doi)); | |
1784 | ||
1785 | tx = dmu_tx_create(os); | |
1786 | ||
1787 | dmu_tx_hold_zap(tx, lr->lr_doid, B_FALSE, name); | |
1788 | dmu_tx_hold_free(tx, object, 0, DMU_OBJECT_END); | |
1789 | ||
1790 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1791 | if (txg == 0) { | |
1792 | ztest_object_unlock(zd, object); | |
1793 | return (ENOSPC); | |
1794 | } | |
1795 | ||
1796 | if (doi.doi_type == DMU_OT_ZAP_OTHER) { | |
1797 | VERIFY3U(0, ==, zap_destroy(os, object, tx)); | |
1798 | } else { | |
1799 | VERIFY3U(0, ==, dmu_object_free(os, object, tx)); | |
1800 | } | |
1801 | ||
1802 | VERIFY3U(0, ==, zap_remove(os, lr->lr_doid, name, tx)); | |
1803 | ||
572e2857 | 1804 | (void) ztest_log_remove(zd, tx, lr, object); |
428870ff BB |
1805 | |
1806 | dmu_tx_commit(tx); | |
1807 | ||
1808 | ztest_object_unlock(zd, object); | |
1809 | ||
1810 | return (0); | |
1811 | } | |
1812 | ||
1813 | static int | |
1814 | ztest_replay_write(ztest_ds_t *zd, lr_write_t *lr, boolean_t byteswap) | |
1815 | { | |
1816 | objset_t *os = zd->zd_os; | |
1817 | void *data = lr + 1; /* data follows lr */ | |
1818 | uint64_t offset, length; | |
1819 | ztest_block_tag_t *bt = data; | |
1820 | ztest_block_tag_t *bbt; | |
1821 | uint64_t gen, txg, lrtxg, crtxg; | |
1822 | dmu_object_info_t doi; | |
1823 | dmu_tx_t *tx; | |
1824 | dmu_buf_t *db; | |
1825 | arc_buf_t *abuf = NULL; | |
e3a07cd0 | 1826 | ztest_zrl_t *rl; |
428870ff BB |
1827 | |
1828 | if (byteswap) | |
1829 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1830 | ||
1831 | offset = lr->lr_offset; | |
1832 | length = lr->lr_length; | |
1833 | ||
1834 | /* If it's a dmu_sync() block, write the whole block */ | |
1835 | if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) { | |
1836 | uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr); | |
1837 | if (length < blocksize) { | |
1838 | offset -= offset % blocksize; | |
1839 | length = blocksize; | |
1840 | } | |
1841 | } | |
1842 | ||
1843 | if (bt->bt_magic == BSWAP_64(BT_MAGIC)) | |
1844 | byteswap_uint64_array(bt, sizeof (*bt)); | |
1845 | ||
1846 | if (bt->bt_magic != BT_MAGIC) | |
1847 | bt = NULL; | |
1848 | ||
1849 | ztest_object_lock(zd, lr->lr_foid, RL_READER); | |
1850 | rl = ztest_range_lock(zd, lr->lr_foid, offset, length, RL_WRITER); | |
1851 | ||
1852 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
1853 | ||
1854 | dmu_object_info_from_db(db, &doi); | |
1855 | ||
1856 | bbt = ztest_bt_bonus(db); | |
1857 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
1858 | gen = bbt->bt_gen; | |
1859 | crtxg = bbt->bt_crtxg; | |
1860 | lrtxg = lr->lr_common.lrc_txg; | |
1861 | ||
1862 | tx = dmu_tx_create(os); | |
1863 | ||
1864 | dmu_tx_hold_write(tx, lr->lr_foid, offset, length); | |
1865 | ||
1866 | if (ztest_random(8) == 0 && length == doi.doi_data_block_size && | |
1867 | P2PHASE(offset, length) == 0) | |
1868 | abuf = dmu_request_arcbuf(db, length); | |
1869 | ||
1870 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1871 | if (txg == 0) { | |
1872 | if (abuf != NULL) | |
1873 | dmu_return_arcbuf(abuf); | |
1874 | dmu_buf_rele(db, FTAG); | |
e3a07cd0 | 1875 | ztest_range_unlock(zd, rl); |
428870ff BB |
1876 | ztest_object_unlock(zd, lr->lr_foid); |
1877 | return (ENOSPC); | |
1878 | } | |
1879 | ||
1880 | if (bt != NULL) { | |
1881 | /* | |
1882 | * Usually, verify the old data before writing new data -- | |
1883 | * but not always, because we also want to verify correct | |
1884 | * behavior when the data was not recently read into cache. | |
1885 | */ | |
1886 | ASSERT(offset % doi.doi_data_block_size == 0); | |
1887 | if (ztest_random(4) != 0) { | |
1888 | int prefetch = ztest_random(2) ? | |
1889 | DMU_READ_PREFETCH : DMU_READ_NO_PREFETCH; | |
1890 | ztest_block_tag_t rbt; | |
1891 | ||
1892 | VERIFY(dmu_read(os, lr->lr_foid, offset, | |
1893 | sizeof (rbt), &rbt, prefetch) == 0); | |
1894 | if (rbt.bt_magic == BT_MAGIC) { | |
50c957f7 | 1895 | ztest_bt_verify(&rbt, os, lr->lr_foid, 0, |
428870ff BB |
1896 | offset, gen, txg, crtxg); |
1897 | } | |
1898 | } | |
1899 | ||
1900 | /* | |
1901 | * Writes can appear to be newer than the bonus buffer because | |
1902 | * the ztest_get_data() callback does a dmu_read() of the | |
1903 | * open-context data, which may be different than the data | |
1904 | * as it was when the write was generated. | |
1905 | */ | |
1906 | if (zd->zd_zilog->zl_replay) { | |
50c957f7 | 1907 | ztest_bt_verify(bt, os, lr->lr_foid, 0, offset, |
428870ff BB |
1908 | MAX(gen, bt->bt_gen), MAX(txg, lrtxg), |
1909 | bt->bt_crtxg); | |
1910 | } | |
1911 | ||
1912 | /* | |
1913 | * Set the bt's gen/txg to the bonus buffer's gen/txg | |
1914 | * so that all of the usual ASSERTs will work. | |
1915 | */ | |
50c957f7 NB |
1916 | ztest_bt_generate(bt, os, lr->lr_foid, 0, offset, gen, txg, |
1917 | crtxg); | |
428870ff BB |
1918 | } |
1919 | ||
1920 | if (abuf == NULL) { | |
1921 | dmu_write(os, lr->lr_foid, offset, length, data, tx); | |
1922 | } else { | |
1923 | bcopy(data, abuf->b_data, length); | |
1924 | dmu_assign_arcbuf(db, offset, abuf, tx); | |
1925 | } | |
1926 | ||
1927 | (void) ztest_log_write(zd, tx, lr); | |
1928 | ||
1929 | dmu_buf_rele(db, FTAG); | |
1930 | ||
1931 | dmu_tx_commit(tx); | |
1932 | ||
e3a07cd0 | 1933 | ztest_range_unlock(zd, rl); |
428870ff BB |
1934 | ztest_object_unlock(zd, lr->lr_foid); |
1935 | ||
1936 | return (0); | |
1937 | } | |
1938 | ||
1939 | static int | |
1940 | ztest_replay_truncate(ztest_ds_t *zd, lr_truncate_t *lr, boolean_t byteswap) | |
1941 | { | |
1942 | objset_t *os = zd->zd_os; | |
1943 | dmu_tx_t *tx; | |
1944 | uint64_t txg; | |
e3a07cd0 | 1945 | ztest_zrl_t *rl; |
428870ff BB |
1946 | |
1947 | if (byteswap) | |
1948 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1949 | ||
1950 | ztest_object_lock(zd, lr->lr_foid, RL_READER); | |
1951 | rl = ztest_range_lock(zd, lr->lr_foid, lr->lr_offset, lr->lr_length, | |
1952 | RL_WRITER); | |
1953 | ||
1954 | tx = dmu_tx_create(os); | |
1955 | ||
1956 | dmu_tx_hold_free(tx, lr->lr_foid, lr->lr_offset, lr->lr_length); | |
1957 | ||
1958 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1959 | if (txg == 0) { | |
e3a07cd0 | 1960 | ztest_range_unlock(zd, rl); |
428870ff BB |
1961 | ztest_object_unlock(zd, lr->lr_foid); |
1962 | return (ENOSPC); | |
1963 | } | |
1964 | ||
1965 | VERIFY(dmu_free_range(os, lr->lr_foid, lr->lr_offset, | |
1966 | lr->lr_length, tx) == 0); | |
1967 | ||
1968 | (void) ztest_log_truncate(zd, tx, lr); | |
1969 | ||
1970 | dmu_tx_commit(tx); | |
1971 | ||
e3a07cd0 | 1972 | ztest_range_unlock(zd, rl); |
428870ff BB |
1973 | ztest_object_unlock(zd, lr->lr_foid); |
1974 | ||
1975 | return (0); | |
1976 | } | |
1977 | ||
1978 | static int | |
1979 | ztest_replay_setattr(ztest_ds_t *zd, lr_setattr_t *lr, boolean_t byteswap) | |
1980 | { | |
1981 | objset_t *os = zd->zd_os; | |
1982 | dmu_tx_t *tx; | |
1983 | dmu_buf_t *db; | |
1984 | ztest_block_tag_t *bbt; | |
50c957f7 | 1985 | uint64_t txg, lrtxg, crtxg, dnodesize; |
428870ff BB |
1986 | |
1987 | if (byteswap) | |
1988 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1989 | ||
1990 | ztest_object_lock(zd, lr->lr_foid, RL_WRITER); | |
1991 | ||
1992 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
1993 | ||
1994 | tx = dmu_tx_create(os); | |
1995 | dmu_tx_hold_bonus(tx, lr->lr_foid); | |
1996 | ||
1997 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1998 | if (txg == 0) { | |
1999 | dmu_buf_rele(db, FTAG); | |
2000 | ztest_object_unlock(zd, lr->lr_foid); | |
2001 | return (ENOSPC); | |
2002 | } | |
2003 | ||
2004 | bbt = ztest_bt_bonus(db); | |
2005 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
2006 | crtxg = bbt->bt_crtxg; | |
2007 | lrtxg = lr->lr_common.lrc_txg; | |
50c957f7 | 2008 | dnodesize = bbt->bt_dnodesize; |
428870ff BB |
2009 | |
2010 | if (zd->zd_zilog->zl_replay) { | |
2011 | ASSERT(lr->lr_size != 0); | |
2012 | ASSERT(lr->lr_mode != 0); | |
2013 | ASSERT(lrtxg != 0); | |
2014 | } else { | |
2015 | /* | |
2016 | * Randomly change the size and increment the generation. | |
2017 | */ | |
2018 | lr->lr_size = (ztest_random(db->db_size / sizeof (*bbt)) + 1) * | |
2019 | sizeof (*bbt); | |
2020 | lr->lr_mode = bbt->bt_gen + 1; | |
2021 | ASSERT(lrtxg == 0); | |
2022 | } | |
2023 | ||
2024 | /* | |
2025 | * Verify that the current bonus buffer is not newer than our txg. | |
2026 | */ | |
50c957f7 | 2027 | ztest_bt_verify(bbt, os, lr->lr_foid, dnodesize, -1ULL, lr->lr_mode, |
428870ff BB |
2028 | MAX(txg, lrtxg), crtxg); |
2029 | ||
2030 | dmu_buf_will_dirty(db, tx); | |
2031 | ||
2032 | ASSERT3U(lr->lr_size, >=, sizeof (*bbt)); | |
2033 | ASSERT3U(lr->lr_size, <=, db->db_size); | |
c99c9001 | 2034 | VERIFY0(dmu_set_bonus(db, lr->lr_size, tx)); |
428870ff BB |
2035 | bbt = ztest_bt_bonus(db); |
2036 | ||
50c957f7 NB |
2037 | ztest_bt_generate(bbt, os, lr->lr_foid, dnodesize, -1ULL, lr->lr_mode, |
2038 | txg, crtxg); | |
2039 | ztest_fill_unused_bonus(db, bbt, lr->lr_foid, os, bbt->bt_gen); | |
428870ff BB |
2040 | dmu_buf_rele(db, FTAG); |
2041 | ||
2042 | (void) ztest_log_setattr(zd, tx, lr); | |
2043 | ||
2044 | dmu_tx_commit(tx); | |
2045 | ||
2046 | ztest_object_unlock(zd, lr->lr_foid); | |
2047 | ||
2048 | return (0); | |
2049 | } | |
2050 | ||
b01615d5 | 2051 | zil_replay_func_t ztest_replay_vector[TX_MAX_TYPE] = { |
0bc8fd78 | 2052 | NULL, /* 0 no such transaction type */ |
b01615d5 | 2053 | (zil_replay_func_t)ztest_replay_create, /* TX_CREATE */ |
0bc8fd78 BB |
2054 | NULL, /* TX_MKDIR */ |
2055 | NULL, /* TX_MKXATTR */ | |
2056 | NULL, /* TX_SYMLINK */ | |
b01615d5 | 2057 | (zil_replay_func_t)ztest_replay_remove, /* TX_REMOVE */ |
0bc8fd78 BB |
2058 | NULL, /* TX_RMDIR */ |
2059 | NULL, /* TX_LINK */ | |
2060 | NULL, /* TX_RENAME */ | |
b01615d5 RY |
2061 | (zil_replay_func_t)ztest_replay_write, /* TX_WRITE */ |
2062 | (zil_replay_func_t)ztest_replay_truncate, /* TX_TRUNCATE */ | |
2063 | (zil_replay_func_t)ztest_replay_setattr, /* TX_SETATTR */ | |
0bc8fd78 BB |
2064 | NULL, /* TX_ACL */ |
2065 | NULL, /* TX_CREATE_ACL */ | |
2066 | NULL, /* TX_CREATE_ATTR */ | |
2067 | NULL, /* TX_CREATE_ACL_ATTR */ | |
2068 | NULL, /* TX_MKDIR_ACL */ | |
2069 | NULL, /* TX_MKDIR_ATTR */ | |
2070 | NULL, /* TX_MKDIR_ACL_ATTR */ | |
2071 | NULL, /* TX_WRITE2 */ | |
428870ff BB |
2072 | }; |
2073 | ||
2074 | /* | |
2075 | * ZIL get_data callbacks | |
2076 | */ | |
e3a07cd0 BP |
2077 | typedef struct ztest_zgd_private { |
2078 | ztest_ds_t *z_zd; | |
2079 | ztest_zrl_t *z_rl; | |
2080 | uint64_t z_object; | |
2081 | } ztest_zgd_private_t; | |
428870ff BB |
2082 | |
2083 | static void | |
2084 | ztest_get_done(zgd_t *zgd, int error) | |
2085 | { | |
e3a07cd0 BP |
2086 | ztest_zgd_private_t *zzp = zgd->zgd_private; |
2087 | ztest_ds_t *zd = zzp->z_zd; | |
2088 | uint64_t object = zzp->z_object; | |
428870ff BB |
2089 | |
2090 | if (zgd->zgd_db) | |
2091 | dmu_buf_rele(zgd->zgd_db, zgd); | |
2092 | ||
e3a07cd0 | 2093 | ztest_range_unlock(zd, zzp->z_rl); |
428870ff BB |
2094 | ztest_object_unlock(zd, object); |
2095 | ||
2096 | if (error == 0 && zgd->zgd_bp) | |
2097 | zil_add_block(zgd->zgd_zilog, zgd->zgd_bp); | |
2098 | ||
2099 | umem_free(zgd, sizeof (*zgd)); | |
e3a07cd0 | 2100 | umem_free(zzp, sizeof (*zzp)); |
428870ff BB |
2101 | } |
2102 | ||
2103 | static int | |
2104 | ztest_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio) | |
2105 | { | |
2106 | ztest_ds_t *zd = arg; | |
2107 | objset_t *os = zd->zd_os; | |
2108 | uint64_t object = lr->lr_foid; | |
2109 | uint64_t offset = lr->lr_offset; | |
2110 | uint64_t size = lr->lr_length; | |
428870ff BB |
2111 | uint64_t txg = lr->lr_common.lrc_txg; |
2112 | uint64_t crtxg; | |
2113 | dmu_object_info_t doi; | |
2114 | dmu_buf_t *db; | |
2115 | zgd_t *zgd; | |
2116 | int error; | |
e3a07cd0 | 2117 | ztest_zgd_private_t *zgd_private; |
428870ff BB |
2118 | |
2119 | ztest_object_lock(zd, object, RL_READER); | |
2120 | error = dmu_bonus_hold(os, object, FTAG, &db); | |
2121 | if (error) { | |
2122 | ztest_object_unlock(zd, object); | |
2123 | return (error); | |
2124 | } | |
2125 | ||
2126 | crtxg = ztest_bt_bonus(db)->bt_crtxg; | |
2127 | ||
2128 | if (crtxg == 0 || crtxg > txg) { | |
2129 | dmu_buf_rele(db, FTAG); | |
2130 | ztest_object_unlock(zd, object); | |
2131 | return (ENOENT); | |
2132 | } | |
2133 | ||
2134 | dmu_object_info_from_db(db, &doi); | |
2135 | dmu_buf_rele(db, FTAG); | |
2136 | db = NULL; | |
2137 | ||
2138 | zgd = umem_zalloc(sizeof (*zgd), UMEM_NOFAIL); | |
2139 | zgd->zgd_zilog = zd->zd_zilog; | |
e3a07cd0 BP |
2140 | zgd_private = umem_zalloc(sizeof (ztest_zgd_private_t), UMEM_NOFAIL); |
2141 | zgd_private->z_zd = zd; | |
2142 | zgd_private->z_object = object; | |
2143 | zgd->zgd_private = zgd_private; | |
428870ff BB |
2144 | |
2145 | if (buf != NULL) { /* immediate write */ | |
e3a07cd0 | 2146 | zgd_private->z_rl = ztest_range_lock(zd, object, offset, size, |
428870ff | 2147 | RL_READER); |
f763c3d1 | 2148 | zgd->zgd_rl = zgd_private->z_rl->z_rl; |
428870ff BB |
2149 | |
2150 | error = dmu_read(os, object, offset, size, buf, | |
2151 | DMU_READ_NO_PREFETCH); | |
2152 | ASSERT(error == 0); | |
2153 | } else { | |
2154 | size = doi.doi_data_block_size; | |
2155 | if (ISP2(size)) { | |
2156 | offset = P2ALIGN(offset, size); | |
2157 | } else { | |
2158 | ASSERT(offset < size); | |
2159 | offset = 0; | |
2160 | } | |
2161 | ||
e3a07cd0 | 2162 | zgd_private->z_rl = ztest_range_lock(zd, object, offset, size, |
428870ff | 2163 | RL_READER); |
f763c3d1 | 2164 | zgd->zgd_rl = zgd_private->z_rl->z_rl; |
428870ff BB |
2165 | |
2166 | error = dmu_buf_hold(os, object, offset, zgd, &db, | |
2167 | DMU_READ_NO_PREFETCH); | |
2168 | ||
2169 | if (error == 0) { | |
02dc43bc | 2170 | blkptr_t *bp = &lr->lr_blkptr; |
03c6040b | 2171 | |
428870ff BB |
2172 | zgd->zgd_db = db; |
2173 | zgd->zgd_bp = bp; | |
2174 | ||
2175 | ASSERT(db->db_offset == offset); | |
2176 | ASSERT(db->db_size == size); | |
2177 | ||
2178 | error = dmu_sync(zio, lr->lr_common.lrc_txg, | |
2179 | ztest_get_done, zgd); | |
2180 | ||
2181 | if (error == 0) | |
2182 | return (0); | |
2183 | } | |
2184 | } | |
2185 | ||
2186 | ztest_get_done(zgd, error); | |
2187 | ||
2188 | return (error); | |
2189 | } | |
2190 | ||
2191 | static void * | |
2192 | ztest_lr_alloc(size_t lrsize, char *name) | |
2193 | { | |
2194 | char *lr; | |
2195 | size_t namesize = name ? strlen(name) + 1 : 0; | |
2196 | ||
2197 | lr = umem_zalloc(lrsize + namesize, UMEM_NOFAIL); | |
2198 | ||
2199 | if (name) | |
2200 | bcopy(name, lr + lrsize, namesize); | |
2201 | ||
2202 | return (lr); | |
2203 | } | |
2204 | ||
2205 | void | |
2206 | ztest_lr_free(void *lr, size_t lrsize, char *name) | |
2207 | { | |
2208 | size_t namesize = name ? strlen(name) + 1 : 0; | |
2209 | ||
2210 | umem_free(lr, lrsize + namesize); | |
2211 | } | |
2212 | ||
2213 | /* | |
2214 | * Lookup a bunch of objects. Returns the number of objects not found. | |
2215 | */ | |
2216 | static int | |
2217 | ztest_lookup(ztest_ds_t *zd, ztest_od_t *od, int count) | |
2218 | { | |
2219 | int missing = 0; | |
2220 | int error; | |
d6320ddb | 2221 | int i; |
428870ff | 2222 | |
c25b8f99 | 2223 | ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock)); |
428870ff | 2224 | |
d6320ddb | 2225 | for (i = 0; i < count; i++, od++) { |
428870ff BB |
2226 | od->od_object = 0; |
2227 | error = zap_lookup(zd->zd_os, od->od_dir, od->od_name, | |
2228 | sizeof (uint64_t), 1, &od->od_object); | |
2229 | if (error) { | |
2230 | ASSERT(error == ENOENT); | |
2231 | ASSERT(od->od_object == 0); | |
2232 | missing++; | |
2233 | } else { | |
2234 | dmu_buf_t *db; | |
2235 | ztest_block_tag_t *bbt; | |
2236 | dmu_object_info_t doi; | |
2237 | ||
2238 | ASSERT(od->od_object != 0); | |
2239 | ASSERT(missing == 0); /* there should be no gaps */ | |
2240 | ||
2241 | ztest_object_lock(zd, od->od_object, RL_READER); | |
2242 | VERIFY3U(0, ==, dmu_bonus_hold(zd->zd_os, | |
2243 | od->od_object, FTAG, &db)); | |
2244 | dmu_object_info_from_db(db, &doi); | |
2245 | bbt = ztest_bt_bonus(db); | |
2246 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
2247 | od->od_type = doi.doi_type; | |
2248 | od->od_blocksize = doi.doi_data_block_size; | |
2249 | od->od_gen = bbt->bt_gen; | |
2250 | dmu_buf_rele(db, FTAG); | |
2251 | ztest_object_unlock(zd, od->od_object); | |
2252 | } | |
2253 | } | |
2254 | ||
2255 | return (missing); | |
2256 | } | |
2257 | ||
2258 | static int | |
2259 | ztest_create(ztest_ds_t *zd, ztest_od_t *od, int count) | |
2260 | { | |
2261 | int missing = 0; | |
d6320ddb | 2262 | int i; |
428870ff | 2263 | |
c25b8f99 | 2264 | ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock)); |
428870ff | 2265 | |
d6320ddb | 2266 | for (i = 0; i < count; i++, od++) { |
428870ff BB |
2267 | if (missing) { |
2268 | od->od_object = 0; | |
2269 | missing++; | |
2270 | continue; | |
2271 | } | |
2272 | ||
2273 | lr_create_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name); | |
2274 | ||
2275 | lr->lr_doid = od->od_dir; | |
2276 | lr->lr_foid = 0; /* 0 to allocate, > 0 to claim */ | |
2277 | lr->lrz_type = od->od_crtype; | |
2278 | lr->lrz_blocksize = od->od_crblocksize; | |
2279 | lr->lrz_ibshift = ztest_random_ibshift(); | |
2280 | lr->lrz_bonustype = DMU_OT_UINT64_OTHER; | |
50c957f7 | 2281 | lr->lrz_dnodesize = od->od_crdnodesize; |
428870ff BB |
2282 | lr->lr_gen = od->od_crgen; |
2283 | lr->lr_crtime[0] = time(NULL); | |
2284 | ||
2285 | if (ztest_replay_create(zd, lr, B_FALSE) != 0) { | |
2286 | ASSERT(missing == 0); | |
2287 | od->od_object = 0; | |
2288 | missing++; | |
2289 | } else { | |
2290 | od->od_object = lr->lr_foid; | |
2291 | od->od_type = od->od_crtype; | |
2292 | od->od_blocksize = od->od_crblocksize; | |
2293 | od->od_gen = od->od_crgen; | |
2294 | ASSERT(od->od_object != 0); | |
2295 | } | |
2296 | ||
2297 | ztest_lr_free(lr, sizeof (*lr), od->od_name); | |
2298 | } | |
2299 | ||
2300 | return (missing); | |
2301 | } | |
2302 | ||
2303 | static int | |
2304 | ztest_remove(ztest_ds_t *zd, ztest_od_t *od, int count) | |
2305 | { | |
2306 | int missing = 0; | |
2307 | int error; | |
d6320ddb | 2308 | int i; |
428870ff | 2309 | |
c25b8f99 | 2310 | ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock)); |
428870ff BB |
2311 | |
2312 | od += count - 1; | |
2313 | ||
d6320ddb | 2314 | for (i = count - 1; i >= 0; i--, od--) { |
428870ff BB |
2315 | if (missing) { |
2316 | missing++; | |
2317 | continue; | |
2318 | } | |
2319 | ||
03c6040b GW |
2320 | /* |
2321 | * No object was found. | |
2322 | */ | |
428870ff BB |
2323 | if (od->od_object == 0) |
2324 | continue; | |
2325 | ||
2326 | lr_remove_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name); | |
2327 | ||
2328 | lr->lr_doid = od->od_dir; | |
2329 | ||
2330 | if ((error = ztest_replay_remove(zd, lr, B_FALSE)) != 0) { | |
2331 | ASSERT3U(error, ==, ENOSPC); | |
2332 | missing++; | |
2333 | } else { | |
2334 | od->od_object = 0; | |
2335 | } | |
2336 | ztest_lr_free(lr, sizeof (*lr), od->od_name); | |
2337 | } | |
2338 | ||
2339 | return (missing); | |
2340 | } | |
2341 | ||
2342 | static int | |
2343 | ztest_write(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size, | |
2344 | void *data) | |
2345 | { | |
2346 | lr_write_t *lr; | |
2347 | int error; | |
2348 | ||
2349 | lr = ztest_lr_alloc(sizeof (*lr) + size, NULL); | |
2350 | ||
2351 | lr->lr_foid = object; | |
2352 | lr->lr_offset = offset; | |
2353 | lr->lr_length = size; | |
2354 | lr->lr_blkoff = 0; | |
2355 | BP_ZERO(&lr->lr_blkptr); | |
2356 | ||
2357 | bcopy(data, lr + 1, size); | |
2358 | ||
2359 | error = ztest_replay_write(zd, lr, B_FALSE); | |
2360 | ||
2361 | ztest_lr_free(lr, sizeof (*lr) + size, NULL); | |
2362 | ||
2363 | return (error); | |
2364 | } | |
2365 | ||
2366 | static int | |
2367 | ztest_truncate(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size) | |
2368 | { | |
2369 | lr_truncate_t *lr; | |
2370 | int error; | |
2371 | ||
2372 | lr = ztest_lr_alloc(sizeof (*lr), NULL); | |
2373 | ||
2374 | lr->lr_foid = object; | |
2375 | lr->lr_offset = offset; | |
2376 | lr->lr_length = size; | |
2377 | ||
2378 | error = ztest_replay_truncate(zd, lr, B_FALSE); | |
2379 | ||
2380 | ztest_lr_free(lr, sizeof (*lr), NULL); | |
2381 | ||
2382 | return (error); | |
2383 | } | |
2384 | ||
2385 | static int | |
2386 | ztest_setattr(ztest_ds_t *zd, uint64_t object) | |
2387 | { | |
2388 | lr_setattr_t *lr; | |
2389 | int error; | |
2390 | ||
2391 | lr = ztest_lr_alloc(sizeof (*lr), NULL); | |
2392 | ||
2393 | lr->lr_foid = object; | |
2394 | lr->lr_size = 0; | |
2395 | lr->lr_mode = 0; | |
2396 | ||
2397 | error = ztest_replay_setattr(zd, lr, B_FALSE); | |
2398 | ||
2399 | ztest_lr_free(lr, sizeof (*lr), NULL); | |
2400 | ||
2401 | return (error); | |
2402 | } | |
2403 | ||
2404 | static void | |
2405 | ztest_prealloc(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size) | |
2406 | { | |
2407 | objset_t *os = zd->zd_os; | |
2408 | dmu_tx_t *tx; | |
2409 | uint64_t txg; | |
e3a07cd0 | 2410 | ztest_zrl_t *rl; |
428870ff BB |
2411 | |
2412 | txg_wait_synced(dmu_objset_pool(os), 0); | |
2413 | ||
2414 | ztest_object_lock(zd, object, RL_READER); | |
2415 | rl = ztest_range_lock(zd, object, offset, size, RL_WRITER); | |
2416 | ||
2417 | tx = dmu_tx_create(os); | |
2418 | ||
2419 | dmu_tx_hold_write(tx, object, offset, size); | |
2420 | ||
2421 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
2422 | ||
2423 | if (txg != 0) { | |
2424 | dmu_prealloc(os, object, offset, size, tx); | |
2425 | dmu_tx_commit(tx); | |
2426 | txg_wait_synced(dmu_objset_pool(os), txg); | |
2427 | } else { | |
2428 | (void) dmu_free_long_range(os, object, offset, size); | |
2429 | } | |
2430 | ||
e3a07cd0 | 2431 | ztest_range_unlock(zd, rl); |
428870ff BB |
2432 | ztest_object_unlock(zd, object); |
2433 | } | |
2434 | ||
2435 | static void | |
2436 | ztest_io(ztest_ds_t *zd, uint64_t object, uint64_t offset) | |
2437 | { | |
03c6040b | 2438 | int err; |
428870ff BB |
2439 | ztest_block_tag_t wbt; |
2440 | dmu_object_info_t doi; | |
2441 | enum ztest_io_type io_type; | |
2442 | uint64_t blocksize; | |
2443 | void *data; | |
2444 | ||
2445 | VERIFY(dmu_object_info(zd->zd_os, object, &doi) == 0); | |
2446 | blocksize = doi.doi_data_block_size; | |
2447 | data = umem_alloc(blocksize, UMEM_NOFAIL); | |
2448 | ||
2449 | /* | |
2450 | * Pick an i/o type at random, biased toward writing block tags. | |
2451 | */ | |
2452 | io_type = ztest_random(ZTEST_IO_TYPES); | |
2453 | if (ztest_random(2) == 0) | |
2454 | io_type = ZTEST_IO_WRITE_TAG; | |
2455 | ||
7809eb8b | 2456 | (void) rw_rdlock(&zd->zd_zilog_lock); |
3e31d2b0 | 2457 | |
428870ff BB |
2458 | switch (io_type) { |
2459 | ||
2460 | case ZTEST_IO_WRITE_TAG: | |
50c957f7 NB |
2461 | ztest_bt_generate(&wbt, zd->zd_os, object, doi.doi_dnodesize, |
2462 | offset, 0, 0, 0); | |
428870ff BB |
2463 | (void) ztest_write(zd, object, offset, sizeof (wbt), &wbt); |
2464 | break; | |
2465 | ||
2466 | case ZTEST_IO_WRITE_PATTERN: | |
2467 | (void) memset(data, 'a' + (object + offset) % 5, blocksize); | |
2468 | if (ztest_random(2) == 0) { | |
2469 | /* | |
2470 | * Induce fletcher2 collisions to ensure that | |
2471 | * zio_ddt_collision() detects and resolves them | |
2472 | * when using fletcher2-verify for deduplication. | |
2473 | */ | |
2474 | ((uint64_t *)data)[0] ^= 1ULL << 63; | |
2475 | ((uint64_t *)data)[4] ^= 1ULL << 63; | |
2476 | } | |
2477 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2478 | break; | |
2479 | ||
2480 | case ZTEST_IO_WRITE_ZEROES: | |
2481 | bzero(data, blocksize); | |
2482 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2483 | break; | |
2484 | ||
2485 | case ZTEST_IO_TRUNCATE: | |
2486 | (void) ztest_truncate(zd, object, offset, blocksize); | |
2487 | break; | |
2488 | ||
2489 | case ZTEST_IO_SETATTR: | |
2490 | (void) ztest_setattr(zd, object); | |
2491 | break; | |
e75c13c3 BB |
2492 | default: |
2493 | break; | |
03c6040b GW |
2494 | |
2495 | case ZTEST_IO_REWRITE: | |
7809eb8b | 2496 | (void) rw_rdlock(&ztest_name_lock); |
03c6040b GW |
2497 | err = ztest_dsl_prop_set_uint64(zd->zd_name, |
2498 | ZFS_PROP_CHECKSUM, spa_dedup_checksum(ztest_spa), | |
2499 | B_FALSE); | |
2500 | VERIFY(err == 0 || err == ENOSPC); | |
2501 | err = ztest_dsl_prop_set_uint64(zd->zd_name, | |
2502 | ZFS_PROP_COMPRESSION, | |
2503 | ztest_random_dsl_prop(ZFS_PROP_COMPRESSION), | |
2504 | B_FALSE); | |
2505 | VERIFY(err == 0 || err == ENOSPC); | |
7809eb8b | 2506 | (void) rw_unlock(&ztest_name_lock); |
03c6040b GW |
2507 | |
2508 | VERIFY0(dmu_read(zd->zd_os, object, offset, blocksize, data, | |
2509 | DMU_READ_NO_PREFETCH)); | |
2510 | ||
2511 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2512 | break; | |
428870ff BB |
2513 | } |
2514 | ||
7809eb8b | 2515 | (void) rw_unlock(&zd->zd_zilog_lock); |
3e31d2b0 | 2516 | |
428870ff BB |
2517 | umem_free(data, blocksize); |
2518 | } | |
2519 | ||
2520 | /* | |
2521 | * Initialize an object description template. | |
2522 | */ | |
2523 | static void | |
2524 | ztest_od_init(ztest_od_t *od, uint64_t id, char *tag, uint64_t index, | |
50c957f7 NB |
2525 | dmu_object_type_t type, uint64_t blocksize, uint64_t dnodesize, |
2526 | uint64_t gen) | |
428870ff BB |
2527 | { |
2528 | od->od_dir = ZTEST_DIROBJ; | |
2529 | od->od_object = 0; | |
2530 | ||
2531 | od->od_crtype = type; | |
2532 | od->od_crblocksize = blocksize ? blocksize : ztest_random_blocksize(); | |
50c957f7 | 2533 | od->od_crdnodesize = dnodesize ? dnodesize : ztest_random_dnodesize(); |
428870ff BB |
2534 | od->od_crgen = gen; |
2535 | ||
2536 | od->od_type = DMU_OT_NONE; | |
2537 | od->od_blocksize = 0; | |
2538 | od->od_gen = 0; | |
2539 | ||
2540 | (void) snprintf(od->od_name, sizeof (od->od_name), "%s(%lld)[%llu]", | |
b8864a23 | 2541 | tag, (longlong_t)id, (u_longlong_t)index); |
428870ff BB |
2542 | } |
2543 | ||
2544 | /* | |
2545 | * Lookup or create the objects for a test using the od template. | |
2546 | * If the objects do not all exist, or if 'remove' is specified, | |
2547 | * remove any existing objects and create new ones. Otherwise, | |
2548 | * use the existing objects. | |
2549 | */ | |
2550 | static int | |
2551 | ztest_object_init(ztest_ds_t *zd, ztest_od_t *od, size_t size, boolean_t remove) | |
2552 | { | |
2553 | int count = size / sizeof (*od); | |
2554 | int rv = 0; | |
2555 | ||
1e33ac1e | 2556 | mutex_enter(&zd->zd_dirobj_lock); |
428870ff BB |
2557 | if ((ztest_lookup(zd, od, count) != 0 || remove) && |
2558 | (ztest_remove(zd, od, count) != 0 || | |
2559 | ztest_create(zd, od, count) != 0)) | |
2560 | rv = -1; | |
2561 | zd->zd_od = od; | |
1e33ac1e | 2562 | mutex_exit(&zd->zd_dirobj_lock); |
428870ff BB |
2563 | |
2564 | return (rv); | |
2565 | } | |
2566 | ||
2567 | /* ARGSUSED */ | |
2568 | void | |
2569 | ztest_zil_commit(ztest_ds_t *zd, uint64_t id) | |
2570 | { | |
2571 | zilog_t *zilog = zd->zd_zilog; | |
2572 | ||
7809eb8b | 2573 | (void) rw_rdlock(&zd->zd_zilog_lock); |
3e31d2b0 | 2574 | |
572e2857 | 2575 | zil_commit(zilog, ztest_random(ZTEST_OBJECTS)); |
428870ff BB |
2576 | |
2577 | /* | |
2578 | * Remember the committed values in zd, which is in parent/child | |
2579 | * shared memory. If we die, the next iteration of ztest_run() | |
2580 | * will verify that the log really does contain this record. | |
2581 | */ | |
2582 | mutex_enter(&zilog->zl_lock); | |
c242c188 CS |
2583 | ASSERT(zd->zd_shared != NULL); |
2584 | ASSERT3U(zd->zd_shared->zd_seq, <=, zilog->zl_commit_lr_seq); | |
2585 | zd->zd_shared->zd_seq = zilog->zl_commit_lr_seq; | |
428870ff | 2586 | mutex_exit(&zilog->zl_lock); |
3e31d2b0 | 2587 | |
7809eb8b | 2588 | (void) rw_unlock(&zd->zd_zilog_lock); |
3e31d2b0 ES |
2589 | } |
2590 | ||
2591 | /* | |
2592 | * This function is designed to simulate the operations that occur during a | |
2593 | * mount/unmount operation. We hold the dataset across these operations in an | |
2594 | * attempt to expose any implicit assumptions about ZIL management. | |
2595 | */ | |
2596 | /* ARGSUSED */ | |
2597 | void | |
2598 | ztest_zil_remount(ztest_ds_t *zd, uint64_t id) | |
2599 | { | |
2600 | objset_t *os = zd->zd_os; | |
2601 | ||
03c6040b GW |
2602 | /* |
2603 | * We grab the zd_dirobj_lock to ensure that no other thread is | |
2604 | * updating the zil (i.e. adding in-memory log records) and the | |
2605 | * zd_zilog_lock to block any I/O. | |
2606 | */ | |
29809a6c | 2607 | mutex_enter(&zd->zd_dirobj_lock); |
7809eb8b | 2608 | (void) rw_wrlock(&zd->zd_zilog_lock); |
3e31d2b0 | 2609 | |
f298b24d | 2610 | /* zfsvfs_teardown() */ |
3e31d2b0 ES |
2611 | zil_close(zd->zd_zilog); |
2612 | ||
2613 | /* zfsvfs_setup() */ | |
2614 | VERIFY(zil_open(os, ztest_get_data) == zd->zd_zilog); | |
2615 | zil_replay(os, zd, ztest_replay_vector); | |
2616 | ||
7809eb8b | 2617 | (void) rw_unlock(&zd->zd_zilog_lock); |
29809a6c | 2618 | mutex_exit(&zd->zd_dirobj_lock); |
428870ff BB |
2619 | } |
2620 | ||
2621 | /* | |
2622 | * Verify that we can't destroy an active pool, create an existing pool, | |
2623 | * or create a pool with a bad vdev spec. | |
2624 | */ | |
2625 | /* ARGSUSED */ | |
2626 | void | |
2627 | ztest_spa_create_destroy(ztest_ds_t *zd, uint64_t id) | |
2628 | { | |
c242c188 | 2629 | ztest_shared_opts_t *zo = &ztest_opts; |
428870ff BB |
2630 | spa_t *spa; |
2631 | nvlist_t *nvroot; | |
2632 | ||
379ca9cf OF |
2633 | if (zo->zo_mmp_test) |
2634 | return; | |
2635 | ||
428870ff BB |
2636 | /* |
2637 | * Attempt to create using a bad file. | |
2638 | */ | |
ea0b2538 | 2639 | nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, 0, 0, 0, 1); |
428870ff | 2640 | VERIFY3U(ENOENT, ==, |
b5256303 | 2641 | spa_create("ztest_bad_file", nvroot, NULL, NULL, NULL)); |
428870ff BB |
2642 | nvlist_free(nvroot); |
2643 | ||
2644 | /* | |
2645 | * Attempt to create using a bad mirror. | |
2646 | */ | |
ea0b2538 | 2647 | nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, 0, 0, 2, 1); |
428870ff | 2648 | VERIFY3U(ENOENT, ==, |
b5256303 | 2649 | spa_create("ztest_bad_mirror", nvroot, NULL, NULL, NULL)); |
428870ff BB |
2650 | nvlist_free(nvroot); |
2651 | ||
2652 | /* | |
2653 | * Attempt to create an existing pool. It shouldn't matter | |
2654 | * what's in the nvroot; we should fail with EEXIST. | |
2655 | */ | |
7809eb8b | 2656 | (void) rw_rdlock(&ztest_name_lock); |
ea0b2538 | 2657 | nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, 0, 0, 0, 1); |
b5256303 TC |
2658 | VERIFY3U(EEXIST, ==, |
2659 | spa_create(zo->zo_pool, nvroot, NULL, NULL, NULL)); | |
428870ff | 2660 | nvlist_free(nvroot); |
c242c188 CS |
2661 | VERIFY3U(0, ==, spa_open(zo->zo_pool, &spa, FTAG)); |
2662 | VERIFY3U(EBUSY, ==, spa_destroy(zo->zo_pool)); | |
428870ff BB |
2663 | spa_close(spa, FTAG); |
2664 | ||
7809eb8b | 2665 | (void) rw_unlock(&ztest_name_lock); |
428870ff BB |
2666 | } |
2667 | ||
0582e403 OF |
2668 | /* |
2669 | * Start and then stop the MMP threads to ensure the startup and shutdown code | |
2670 | * works properly. Actual protection and property-related code tested via ZTS. | |
2671 | */ | |
2672 | /* ARGSUSED */ | |
2673 | void | |
2674 | ztest_mmp_enable_disable(ztest_ds_t *zd, uint64_t id) | |
2675 | { | |
2676 | ztest_shared_opts_t *zo = &ztest_opts; | |
2677 | spa_t *spa = ztest_spa; | |
2678 | ||
2679 | if (zo->zo_mmp_test) | |
2680 | return; | |
2681 | ||
2682 | spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); | |
2683 | mutex_enter(&spa->spa_props_lock); | |
2684 | ||
2685 | if (!spa_multihost(spa)) { | |
2686 | spa->spa_multihost = B_TRUE; | |
2687 | mmp_thread_start(spa); | |
2688 | } | |
2689 | ||
2690 | mutex_exit(&spa->spa_props_lock); | |
2691 | spa_config_exit(spa, SCL_CONFIG, FTAG); | |
2692 | ||
2693 | txg_wait_synced(spa_get_dsl(spa), 0); | |
2694 | mmp_signal_all_threads(); | |
2695 | txg_wait_synced(spa_get_dsl(spa), 0); | |
2696 | ||
2697 | spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); | |
2698 | mutex_enter(&spa->spa_props_lock); | |
2699 | ||
2700 | if (spa_multihost(spa)) { | |
2701 | mmp_thread_stop(spa); | |
2702 | spa->spa_multihost = B_FALSE; | |
2703 | } | |
2704 | ||
2705 | mutex_exit(&spa->spa_props_lock); | |
2706 | spa_config_exit(spa, SCL_CONFIG, FTAG); | |
2707 | } | |
2708 | ||
ea0b2538 GW |
2709 | /* ARGSUSED */ |
2710 | void | |
2711 | ztest_spa_upgrade(ztest_ds_t *zd, uint64_t id) | |
2712 | { | |
2713 | spa_t *spa; | |
2714 | uint64_t initial_version = SPA_VERSION_INITIAL; | |
2715 | uint64_t version, newversion; | |
2716 | nvlist_t *nvroot, *props; | |
2717 | char *name; | |
2718 | ||
379ca9cf OF |
2719 | if (ztest_opts.zo_mmp_test) |
2720 | return; | |
2721 | ||
ea0b2538 GW |
2722 | mutex_enter(&ztest_vdev_lock); |
2723 | name = kmem_asprintf("%s_upgrade", ztest_opts.zo_pool); | |
2724 | ||
2725 | /* | |
2726 | * Clean up from previous runs. | |
2727 | */ | |
2728 | (void) spa_destroy(name); | |
2729 | ||
2730 | nvroot = make_vdev_root(NULL, NULL, name, ztest_opts.zo_vdev_size, 0, | |
2731 | 0, ztest_opts.zo_raidz, ztest_opts.zo_mirrors, 1); | |
2732 | ||
2733 | /* | |
2734 | * If we're configuring a RAIDZ device then make sure that the | |
2735 | * the initial version is capable of supporting that feature. | |
2736 | */ | |
2737 | switch (ztest_opts.zo_raidz_parity) { | |
2738 | case 0: | |
2739 | case 1: | |
2740 | initial_version = SPA_VERSION_INITIAL; | |
2741 | break; | |
2742 | case 2: | |
2743 | initial_version = SPA_VERSION_RAIDZ2; | |
2744 | break; | |
2745 | case 3: | |
2746 | initial_version = SPA_VERSION_RAIDZ3; | |
2747 | break; | |
2748 | } | |
2749 | ||
2750 | /* | |
2751 | * Create a pool with a spa version that can be upgraded. Pick | |
2752 | * a value between initial_version and SPA_VERSION_BEFORE_FEATURES. | |
2753 | */ | |
2754 | do { | |
2755 | version = ztest_random_spa_version(initial_version); | |
2756 | } while (version > SPA_VERSION_BEFORE_FEATURES); | |
2757 | ||
2758 | props = fnvlist_alloc(); | |
2759 | fnvlist_add_uint64(props, | |
2760 | zpool_prop_to_name(ZPOOL_PROP_VERSION), version); | |
b5256303 | 2761 | VERIFY3S(spa_create(name, nvroot, props, NULL, NULL), ==, 0); |
ea0b2538 GW |
2762 | fnvlist_free(nvroot); |
2763 | fnvlist_free(props); | |
2764 | ||
2765 | VERIFY3S(spa_open(name, &spa, FTAG), ==, 0); | |
2766 | VERIFY3U(spa_version(spa), ==, version); | |
2767 | newversion = ztest_random_spa_version(version + 1); | |
2768 | ||
2769 | if (ztest_opts.zo_verbose >= 4) { | |
2770 | (void) printf("upgrading spa version from %llu to %llu\n", | |
2771 | (u_longlong_t)version, (u_longlong_t)newversion); | |
2772 | } | |
2773 | ||
2774 | spa_upgrade(spa, newversion); | |
2775 | VERIFY3U(spa_version(spa), >, version); | |
2776 | VERIFY3U(spa_version(spa), ==, fnvlist_lookup_uint64(spa->spa_config, | |
2777 | zpool_prop_to_name(ZPOOL_PROP_VERSION))); | |
2778 | spa_close(spa, FTAG); | |
2779 | ||
2780 | strfree(name); | |
2781 | mutex_exit(&ztest_vdev_lock); | |
2782 | } | |
2783 | ||
428870ff BB |
2784 | static vdev_t * |
2785 | vdev_lookup_by_path(vdev_t *vd, const char *path) | |
2786 | { | |
2787 | vdev_t *mvd; | |
d6320ddb | 2788 | int c; |
428870ff BB |
2789 | |
2790 | if (vd->vdev_path != NULL && strcmp(path, vd->vdev_path) == 0) | |
2791 | return (vd); | |
2792 | ||
d6320ddb | 2793 | for (c = 0; c < vd->vdev_children; c++) |
428870ff BB |
2794 | if ((mvd = vdev_lookup_by_path(vd->vdev_child[c], path)) != |
2795 | NULL) | |
2796 | return (mvd); | |
2797 | ||
2798 | return (NULL); | |
2799 | } | |
2800 | ||
2801 | /* | |
2802 | * Find the first available hole which can be used as a top-level. | |
2803 | */ | |
2804 | int | |
2805 | find_vdev_hole(spa_t *spa) | |
2806 | { | |
2807 | vdev_t *rvd = spa->spa_root_vdev; | |
2808 | int c; | |
2809 | ||
2810 | ASSERT(spa_config_held(spa, SCL_VDEV, RW_READER) == SCL_VDEV); | |
2811 | ||
2812 | for (c = 0; c < rvd->vdev_children; c++) { | |
2813 | vdev_t *cvd = rvd->vdev_child[c]; | |
2814 | ||
2815 | if (cvd->vdev_ishole) | |
2816 | break; | |
2817 | } | |
2818 | return (c); | |
2819 | } | |
2820 | ||
2821 | /* | |
2822 | * Verify that vdev_add() works as expected. | |
2823 | */ | |
2824 | /* ARGSUSED */ | |
2825 | void | |
2826 | ztest_vdev_add_remove(ztest_ds_t *zd, uint64_t id) | |
2827 | { | |
2828 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 2829 | spa_t *spa = ztest_spa; |
428870ff BB |
2830 | uint64_t leaves; |
2831 | uint64_t guid; | |
2832 | nvlist_t *nvroot; | |
2833 | int error; | |
2834 | ||
379ca9cf OF |
2835 | if (ztest_opts.zo_mmp_test) |
2836 | return; | |
2837 | ||
c242c188 | 2838 | mutex_enter(&ztest_vdev_lock); |
13fe0198 | 2839 | leaves = MAX(zs->zs_mirrors + zs->zs_splits, 1) * ztest_opts.zo_raidz; |
428870ff BB |
2840 | |
2841 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
2842 | ||
2843 | ztest_shared->zs_vdev_next_leaf = find_vdev_hole(spa) * leaves; | |
2844 | ||
2845 | /* | |
2846 | * If we have slogs then remove them 1/4 of the time. | |
2847 | */ | |
2848 | if (spa_has_slogs(spa) && ztest_random(4) == 0) { | |
2849 | /* | |
2850 | * Grab the guid from the head of the log class rotor. | |
2851 | */ | |
2852 | guid = spa_log_class(spa)->mc_rotor->mg_vd->vdev_guid; | |
2853 | ||
2854 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
2855 | ||
2856 | /* | |
2857 | * We have to grab the zs_name_lock as writer to | |
2858 | * prevent a race between removing a slog (dmu_objset_find) | |
2859 | * and destroying a dataset. Removing the slog will | |
2860 | * grab a reference on the dataset which may cause | |
13fe0198 | 2861 | * dsl_destroy_head() to fail with EBUSY thus |
428870ff BB |
2862 | * leaving the dataset in an inconsistent state. |
2863 | */ | |
7809eb8b | 2864 | rw_wrlock(&ztest_name_lock); |
428870ff | 2865 | error = spa_vdev_remove(spa, guid, B_FALSE); |
7809eb8b | 2866 | rw_unlock(&ztest_name_lock); |
428870ff BB |
2867 | |
2868 | if (error && error != EEXIST) | |
2869 | fatal(0, "spa_vdev_remove() = %d", error); | |
2870 | } else { | |
2871 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
2872 | ||
2873 | /* | |
2874 | * Make 1/4 of the devices be log devices. | |
2875 | */ | |
ea0b2538 | 2876 | nvroot = make_vdev_root(NULL, NULL, NULL, |
c242c188 CS |
2877 | ztest_opts.zo_vdev_size, 0, |
2878 | ztest_random(4) == 0, ztest_opts.zo_raidz, | |
2879 | zs->zs_mirrors, 1); | |
428870ff BB |
2880 | |
2881 | error = spa_vdev_add(spa, nvroot); | |
2882 | nvlist_free(nvroot); | |
2883 | ||
2884 | if (error == ENOSPC) | |
2885 | ztest_record_enospc("spa_vdev_add"); | |
2886 | else if (error != 0) | |
2887 | fatal(0, "spa_vdev_add() = %d", error); | |
2888 | } | |
2889 | ||
c242c188 | 2890 | mutex_exit(&ztest_vdev_lock); |
428870ff BB |
2891 | } |
2892 | ||
2893 | /* | |
2894 | * Verify that adding/removing aux devices (l2arc, hot spare) works as expected. | |
2895 | */ | |
2896 | /* ARGSUSED */ | |
2897 | void | |
2898 | ztest_vdev_aux_add_remove(ztest_ds_t *zd, uint64_t id) | |
2899 | { | |
2900 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 2901 | spa_t *spa = ztest_spa; |
428870ff BB |
2902 | vdev_t *rvd = spa->spa_root_vdev; |
2903 | spa_aux_vdev_t *sav; | |
2904 | char *aux; | |
40b84e7a | 2905 | char *path; |
428870ff BB |
2906 | uint64_t guid = 0; |
2907 | int error; | |
2908 | ||
379ca9cf OF |
2909 | if (ztest_opts.zo_mmp_test) |
2910 | return; | |
2911 | ||
40b84e7a BB |
2912 | path = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
2913 | ||
428870ff BB |
2914 | if (ztest_random(2) == 0) { |
2915 | sav = &spa->spa_spares; | |
2916 | aux = ZPOOL_CONFIG_SPARES; | |
2917 | } else { | |
2918 | sav = &spa->spa_l2cache; | |
2919 | aux = ZPOOL_CONFIG_L2CACHE; | |
2920 | } | |
2921 | ||
c242c188 | 2922 | mutex_enter(&ztest_vdev_lock); |
428870ff BB |
2923 | |
2924 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
2925 | ||
2926 | if (sav->sav_count != 0 && ztest_random(4) == 0) { | |
b128c09f BB |
2927 | /* |
2928 | * Pick a random device to remove. | |
2929 | */ | |
2930 | guid = sav->sav_vdevs[ztest_random(sav->sav_count)]->vdev_guid; | |
2931 | } else { | |
2932 | /* | |
2933 | * Find an unused device we can add. | |
2934 | */ | |
428870ff | 2935 | zs->zs_vdev_aux = 0; |
b128c09f | 2936 | for (;;) { |
b128c09f | 2937 | int c; |
3db3ff4a | 2938 | (void) snprintf(path, MAXPATHLEN, ztest_aux_template, |
c242c188 CS |
2939 | ztest_opts.zo_dir, ztest_opts.zo_pool, aux, |
2940 | zs->zs_vdev_aux); | |
b128c09f BB |
2941 | for (c = 0; c < sav->sav_count; c++) |
2942 | if (strcmp(sav->sav_vdevs[c]->vdev_path, | |
2943 | path) == 0) | |
2944 | break; | |
2945 | if (c == sav->sav_count && | |
2946 | vdev_lookup_by_path(rvd, path) == NULL) | |
2947 | break; | |
428870ff | 2948 | zs->zs_vdev_aux++; |
34dc7c2f BB |
2949 | } |
2950 | } | |
2951 | ||
b128c09f | 2952 | spa_config_exit(spa, SCL_VDEV, FTAG); |
34dc7c2f | 2953 | |
b128c09f BB |
2954 | if (guid == 0) { |
2955 | /* | |
2956 | * Add a new device. | |
2957 | */ | |
ea0b2538 | 2958 | nvlist_t *nvroot = make_vdev_root(NULL, aux, NULL, |
c242c188 | 2959 | (ztest_opts.zo_vdev_size * 5) / 4, 0, 0, 0, 0, 1); |
b128c09f BB |
2960 | error = spa_vdev_add(spa, nvroot); |
2961 | if (error != 0) | |
2962 | fatal(0, "spa_vdev_add(%p) = %d", nvroot, error); | |
2963 | nvlist_free(nvroot); | |
2964 | } else { | |
2965 | /* | |
2966 | * Remove an existing device. Sometimes, dirty its | |
2967 | * vdev state first to make sure we handle removal | |
2968 | * of devices that have pending state changes. | |
2969 | */ | |
2970 | if (ztest_random(2) == 0) | |
9babb374 | 2971 | (void) vdev_online(spa, guid, 0, NULL); |
b128c09f BB |
2972 | |
2973 | error = spa_vdev_remove(spa, guid, B_FALSE); | |
2974 | if (error != 0 && error != EBUSY) | |
2975 | fatal(0, "spa_vdev_remove(%llu) = %d", guid, error); | |
2976 | } | |
2977 | ||
c242c188 | 2978 | mutex_exit(&ztest_vdev_lock); |
40b84e7a BB |
2979 | |
2980 | umem_free(path, MAXPATHLEN); | |
428870ff BB |
2981 | } |
2982 | ||
2983 | /* | |
2984 | * split a pool if it has mirror tlvdevs | |
2985 | */ | |
2986 | /* ARGSUSED */ | |
2987 | void | |
2988 | ztest_split_pool(ztest_ds_t *zd, uint64_t id) | |
2989 | { | |
2990 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 2991 | spa_t *spa = ztest_spa; |
428870ff BB |
2992 | vdev_t *rvd = spa->spa_root_vdev; |
2993 | nvlist_t *tree, **child, *config, *split, **schild; | |
2994 | uint_t c, children, schildren = 0, lastlogid = 0; | |
2995 | int error = 0; | |
2996 | ||
379ca9cf OF |
2997 | if (ztest_opts.zo_mmp_test) |
2998 | return; | |
2999 | ||
c242c188 | 3000 | mutex_enter(&ztest_vdev_lock); |
428870ff BB |
3001 | |
3002 | /* ensure we have a useable config; mirrors of raidz aren't supported */ | |
c242c188 CS |
3003 | if (zs->zs_mirrors < 3 || ztest_opts.zo_raidz > 1) { |
3004 | mutex_exit(&ztest_vdev_lock); | |
428870ff BB |
3005 | return; |
3006 | } | |
3007 | ||
3008 | /* clean up the old pool, if any */ | |
3009 | (void) spa_destroy("splitp"); | |
3010 | ||
3011 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
3012 | ||
3013 | /* generate a config from the existing config */ | |
3014 | mutex_enter(&spa->spa_props_lock); | |
3015 | VERIFY(nvlist_lookup_nvlist(spa->spa_config, ZPOOL_CONFIG_VDEV_TREE, | |
3016 | &tree) == 0); | |
3017 | mutex_exit(&spa->spa_props_lock); | |
3018 | ||
3019 | VERIFY(nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child, | |
3020 | &children) == 0); | |
3021 | ||
3022 | schild = malloc(rvd->vdev_children * sizeof (nvlist_t *)); | |
3023 | for (c = 0; c < children; c++) { | |
3024 | vdev_t *tvd = rvd->vdev_child[c]; | |
3025 | nvlist_t **mchild; | |
3026 | uint_t mchildren; | |
3027 | ||
3028 | if (tvd->vdev_islog || tvd->vdev_ops == &vdev_hole_ops) { | |
3029 | VERIFY(nvlist_alloc(&schild[schildren], NV_UNIQUE_NAME, | |
3030 | 0) == 0); | |
3031 | VERIFY(nvlist_add_string(schild[schildren], | |
3032 | ZPOOL_CONFIG_TYPE, VDEV_TYPE_HOLE) == 0); | |
3033 | VERIFY(nvlist_add_uint64(schild[schildren], | |
3034 | ZPOOL_CONFIG_IS_HOLE, 1) == 0); | |
3035 | if (lastlogid == 0) | |
3036 | lastlogid = schildren; | |
3037 | ++schildren; | |
3038 | continue; | |
3039 | } | |
3040 | lastlogid = 0; | |
3041 | VERIFY(nvlist_lookup_nvlist_array(child[c], | |
3042 | ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0); | |
3043 | VERIFY(nvlist_dup(mchild[0], &schild[schildren++], 0) == 0); | |
3044 | } | |
3045 | ||
3046 | /* OK, create a config that can be used to split */ | |
3047 | VERIFY(nvlist_alloc(&split, NV_UNIQUE_NAME, 0) == 0); | |
3048 | VERIFY(nvlist_add_string(split, ZPOOL_CONFIG_TYPE, | |
3049 | VDEV_TYPE_ROOT) == 0); | |
3050 | VERIFY(nvlist_add_nvlist_array(split, ZPOOL_CONFIG_CHILDREN, schild, | |
3051 | lastlogid != 0 ? lastlogid : schildren) == 0); | |
3052 | ||
3053 | VERIFY(nvlist_alloc(&config, NV_UNIQUE_NAME, 0) == 0); | |
3054 | VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, split) == 0); | |
3055 | ||
3056 | for (c = 0; c < schildren; c++) | |
3057 | nvlist_free(schild[c]); | |
3058 | free(schild); | |
3059 | nvlist_free(split); | |
3060 | ||
3061 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
3062 | ||
7809eb8b | 3063 | (void) rw_wrlock(&ztest_name_lock); |
428870ff | 3064 | error = spa_vdev_split_mirror(spa, "splitp", config, NULL, B_FALSE); |
7809eb8b | 3065 | (void) rw_unlock(&ztest_name_lock); |
428870ff BB |
3066 | |
3067 | nvlist_free(config); | |
3068 | ||
3069 | if (error == 0) { | |
3070 | (void) printf("successful split - results:\n"); | |
3071 | mutex_enter(&spa_namespace_lock); | |
3072 | show_pool_stats(spa); | |
3073 | show_pool_stats(spa_lookup("splitp")); | |
3074 | mutex_exit(&spa_namespace_lock); | |
3075 | ++zs->zs_splits; | |
3076 | --zs->zs_mirrors; | |
3077 | } | |
c242c188 | 3078 | mutex_exit(&ztest_vdev_lock); |
428870ff | 3079 | |
34dc7c2f BB |
3080 | } |
3081 | ||
3082 | /* | |
3083 | * Verify that we can attach and detach devices. | |
3084 | */ | |
428870ff | 3085 | /* ARGSUSED */ |
34dc7c2f | 3086 | void |
428870ff | 3087 | ztest_vdev_attach_detach(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3088 | { |
428870ff | 3089 | ztest_shared_t *zs = ztest_shared; |
c242c188 | 3090 | spa_t *spa = ztest_spa; |
b128c09f | 3091 | spa_aux_vdev_t *sav = &spa->spa_spares; |
34dc7c2f BB |
3092 | vdev_t *rvd = spa->spa_root_vdev; |
3093 | vdev_t *oldvd, *newvd, *pvd; | |
b128c09f | 3094 | nvlist_t *root; |
428870ff | 3095 | uint64_t leaves; |
34dc7c2f BB |
3096 | uint64_t leaf, top; |
3097 | uint64_t ashift = ztest_get_ashift(); | |
fb5f0bc8 | 3098 | uint64_t oldguid, pguid; |
5d1f7fb6 | 3099 | uint64_t oldsize, newsize; |
40b84e7a | 3100 | char *oldpath, *newpath; |
34dc7c2f | 3101 | int replacing; |
b128c09f BB |
3102 | int oldvd_has_siblings = B_FALSE; |
3103 | int newvd_is_spare = B_FALSE; | |
3104 | int oldvd_is_log; | |
34dc7c2f | 3105 | int error, expected_error; |
34dc7c2f | 3106 | |
379ca9cf OF |
3107 | if (ztest_opts.zo_mmp_test) |
3108 | return; | |
3109 | ||
40b84e7a BB |
3110 | oldpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
3111 | newpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
3112 | ||
c242c188 CS |
3113 | mutex_enter(&ztest_vdev_lock); |
3114 | leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raidz; | |
34dc7c2f | 3115 | |
b128c09f | 3116 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); |
34dc7c2f BB |
3117 | |
3118 | /* | |
3119 | * Decide whether to do an attach or a replace. | |
3120 | */ | |
3121 | replacing = ztest_random(2); | |
3122 | ||
3123 | /* | |
3124 | * Pick a random top-level vdev. | |
3125 | */ | |
428870ff | 3126 | top = ztest_random_vdev_top(spa, B_TRUE); |
34dc7c2f BB |
3127 | |
3128 | /* | |
3129 | * Pick a random leaf within it. | |
3130 | */ | |
3131 | leaf = ztest_random(leaves); | |
3132 | ||
3133 | /* | |
b128c09f | 3134 | * Locate this vdev. |
34dc7c2f | 3135 | */ |
b128c09f | 3136 | oldvd = rvd->vdev_child[top]; |
428870ff | 3137 | if (zs->zs_mirrors >= 1) { |
fb5f0bc8 | 3138 | ASSERT(oldvd->vdev_ops == &vdev_mirror_ops); |
428870ff | 3139 | ASSERT(oldvd->vdev_children >= zs->zs_mirrors); |
c242c188 | 3140 | oldvd = oldvd->vdev_child[leaf / ztest_opts.zo_raidz]; |
fb5f0bc8 | 3141 | } |
c242c188 | 3142 | if (ztest_opts.zo_raidz > 1) { |
fb5f0bc8 | 3143 | ASSERT(oldvd->vdev_ops == &vdev_raidz_ops); |
c242c188 CS |
3144 | ASSERT(oldvd->vdev_children == ztest_opts.zo_raidz); |
3145 | oldvd = oldvd->vdev_child[leaf % ztest_opts.zo_raidz]; | |
fb5f0bc8 | 3146 | } |
34dc7c2f BB |
3147 | |
3148 | /* | |
b128c09f BB |
3149 | * If we're already doing an attach or replace, oldvd may be a |
3150 | * mirror vdev -- in which case, pick a random child. | |
34dc7c2f | 3151 | */ |
b128c09f BB |
3152 | while (oldvd->vdev_children != 0) { |
3153 | oldvd_has_siblings = B_TRUE; | |
fb5f0bc8 BB |
3154 | ASSERT(oldvd->vdev_children >= 2); |
3155 | oldvd = oldvd->vdev_child[ztest_random(oldvd->vdev_children)]; | |
b128c09f BB |
3156 | } |
3157 | ||
3158 | oldguid = oldvd->vdev_guid; | |
9babb374 | 3159 | oldsize = vdev_get_min_asize(oldvd); |
b128c09f BB |
3160 | oldvd_is_log = oldvd->vdev_top->vdev_islog; |
3161 | (void) strcpy(oldpath, oldvd->vdev_path); | |
3162 | pvd = oldvd->vdev_parent; | |
fb5f0bc8 | 3163 | pguid = pvd->vdev_guid; |
34dc7c2f BB |
3164 | |
3165 | /* | |
b128c09f | 3166 | * If oldvd has siblings, then half of the time, detach it. |
34dc7c2f | 3167 | */ |
b128c09f BB |
3168 | if (oldvd_has_siblings && ztest_random(2) == 0) { |
3169 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
fb5f0bc8 BB |
3170 | error = spa_vdev_detach(spa, oldguid, pguid, B_FALSE); |
3171 | if (error != 0 && error != ENODEV && error != EBUSY && | |
3172 | error != ENOTSUP) | |
3173 | fatal(0, "detach (%s) returned %d", oldpath, error); | |
40b84e7a | 3174 | goto out; |
b128c09f | 3175 | } |
34dc7c2f BB |
3176 | |
3177 | /* | |
b128c09f BB |
3178 | * For the new vdev, choose with equal probability between the two |
3179 | * standard paths (ending in either 'a' or 'b') or a random hot spare. | |
34dc7c2f | 3180 | */ |
b128c09f BB |
3181 | if (sav->sav_count != 0 && ztest_random(3) == 0) { |
3182 | newvd = sav->sav_vdevs[ztest_random(sav->sav_count)]; | |
3183 | newvd_is_spare = B_TRUE; | |
3184 | (void) strcpy(newpath, newvd->vdev_path); | |
3185 | } else { | |
6aec1cd5 | 3186 | (void) snprintf(newpath, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
3187 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
3188 | top * leaves + leaf); | |
b128c09f BB |
3189 | if (ztest_random(2) == 0) |
3190 | newpath[strlen(newpath) - 1] = 'b'; | |
3191 | newvd = vdev_lookup_by_path(rvd, newpath); | |
3192 | } | |
3193 | ||
3194 | if (newvd) { | |
9babb374 | 3195 | newsize = vdev_get_min_asize(newvd); |
b128c09f BB |
3196 | } else { |
3197 | /* | |
3198 | * Make newsize a little bigger or smaller than oldsize. | |
3199 | * If it's smaller, the attach should fail. | |
3200 | * If it's larger, and we're doing a replace, | |
3201 | * we should get dynamic LUN growth when we're done. | |
3202 | */ | |
3203 | newsize = 10 * oldsize / (9 + ztest_random(3)); | |
3204 | } | |
34dc7c2f BB |
3205 | |
3206 | /* | |
3207 | * If pvd is not a mirror or root, the attach should fail with ENOTSUP, | |
3208 | * unless it's a replace; in that case any non-replacing parent is OK. | |
3209 | * | |
3210 | * If newvd is already part of the pool, it should fail with EBUSY. | |
3211 | * | |
3212 | * If newvd is too small, it should fail with EOVERFLOW. | |
3213 | */ | |
b128c09f BB |
3214 | if (pvd->vdev_ops != &vdev_mirror_ops && |
3215 | pvd->vdev_ops != &vdev_root_ops && (!replacing || | |
3216 | pvd->vdev_ops == &vdev_replacing_ops || | |
3217 | pvd->vdev_ops == &vdev_spare_ops)) | |
34dc7c2f | 3218 | expected_error = ENOTSUP; |
b128c09f BB |
3219 | else if (newvd_is_spare && (!replacing || oldvd_is_log)) |
3220 | expected_error = ENOTSUP; | |
3221 | else if (newvd == oldvd) | |
3222 | expected_error = replacing ? 0 : EBUSY; | |
3223 | else if (vdev_lookup_by_path(rvd, newpath) != NULL) | |
3224 | expected_error = EBUSY; | |
34dc7c2f BB |
3225 | else if (newsize < oldsize) |
3226 | expected_error = EOVERFLOW; | |
3227 | else if (ashift > oldvd->vdev_top->vdev_ashift) | |
3228 | expected_error = EDOM; | |
3229 | else | |
3230 | expected_error = 0; | |
3231 | ||
b128c09f | 3232 | spa_config_exit(spa, SCL_VDEV, FTAG); |
34dc7c2f BB |
3233 | |
3234 | /* | |
3235 | * Build the nvlist describing newpath. | |
3236 | */ | |
ea0b2538 | 3237 | root = make_vdev_root(newpath, NULL, NULL, newvd == NULL ? newsize : 0, |
b128c09f | 3238 | ashift, 0, 0, 0, 1); |
34dc7c2f | 3239 | |
b128c09f | 3240 | error = spa_vdev_attach(spa, oldguid, root, replacing); |
34dc7c2f | 3241 | |
34dc7c2f BB |
3242 | nvlist_free(root); |
3243 | ||
3244 | /* | |
3245 | * If our parent was the replacing vdev, but the replace completed, | |
3246 | * then instead of failing with ENOTSUP we may either succeed, | |
3247 | * fail with ENODEV, or fail with EOVERFLOW. | |
3248 | */ | |
3249 | if (expected_error == ENOTSUP && | |
3250 | (error == 0 || error == ENODEV || error == EOVERFLOW)) | |
3251 | expected_error = error; | |
3252 | ||
3253 | /* | |
3254 | * If someone grew the LUN, the replacement may be too small. | |
3255 | */ | |
b128c09f | 3256 | if (error == EOVERFLOW || error == EBUSY) |
34dc7c2f BB |
3257 | expected_error = error; |
3258 | ||
b128c09f BB |
3259 | /* XXX workaround 6690467 */ |
3260 | if (error != expected_error && expected_error != EBUSY) { | |
3261 | fatal(0, "attach (%s %llu, %s %llu, %d) " | |
3262 | "returned %d, expected %d", | |
5d1f7fb6 GW |
3263 | oldpath, oldsize, newpath, |
3264 | newsize, replacing, error, expected_error); | |
34dc7c2f | 3265 | } |
40b84e7a | 3266 | out: |
c242c188 | 3267 | mutex_exit(&ztest_vdev_lock); |
40b84e7a BB |
3268 | |
3269 | umem_free(oldpath, MAXPATHLEN); | |
3270 | umem_free(newpath, MAXPATHLEN); | |
34dc7c2f BB |
3271 | } |
3272 | ||
9babb374 BB |
3273 | /* |
3274 | * Callback function which expands the physical size of the vdev. | |
3275 | */ | |
3276 | vdev_t * | |
3277 | grow_vdev(vdev_t *vd, void *arg) | |
3278 | { | |
1fde1e37 | 3279 | ASSERTV(spa_t *spa = vd->vdev_spa); |
9babb374 BB |
3280 | size_t *newsize = arg; |
3281 | size_t fsize; | |
3282 | int fd; | |
3283 | ||
3284 | ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE); | |
3285 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
3286 | ||
3287 | if ((fd = open(vd->vdev_path, O_RDWR)) == -1) | |
3288 | return (vd); | |
3289 | ||
3290 | fsize = lseek(fd, 0, SEEK_END); | |
0e5b68e0 | 3291 | VERIFY(ftruncate(fd, *newsize) == 0); |
9babb374 | 3292 | |
c242c188 | 3293 | if (ztest_opts.zo_verbose >= 6) { |
9babb374 BB |
3294 | (void) printf("%s grew from %lu to %lu bytes\n", |
3295 | vd->vdev_path, (ulong_t)fsize, (ulong_t)*newsize); | |
3296 | } | |
3297 | (void) close(fd); | |
3298 | return (NULL); | |
3299 | } | |
3300 | ||
3301 | /* | |
3302 | * Callback function which expands a given vdev by calling vdev_online(). | |
3303 | */ | |
3304 | /* ARGSUSED */ | |
3305 | vdev_t * | |
3306 | online_vdev(vdev_t *vd, void *arg) | |
3307 | { | |
3308 | spa_t *spa = vd->vdev_spa; | |
3309 | vdev_t *tvd = vd->vdev_top; | |
9babb374 | 3310 | uint64_t guid = vd->vdev_guid; |
428870ff BB |
3311 | uint64_t generation = spa->spa_config_generation + 1; |
3312 | vdev_state_t newstate = VDEV_STATE_UNKNOWN; | |
3313 | int error; | |
9babb374 BB |
3314 | |
3315 | ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE); | |
3316 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
3317 | ||
3318 | /* Calling vdev_online will initialize the new metaslabs */ | |
3319 | spa_config_exit(spa, SCL_STATE, spa); | |
428870ff | 3320 | error = vdev_online(spa, guid, ZFS_ONLINE_EXPAND, &newstate); |
9babb374 BB |
3321 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); |
3322 | ||
428870ff BB |
3323 | /* |
3324 | * If vdev_online returned an error or the underlying vdev_open | |
3325 | * failed then we abort the expand. The only way to know that | |
3326 | * vdev_open fails is by checking the returned newstate. | |
3327 | */ | |
3328 | if (error || newstate != VDEV_STATE_HEALTHY) { | |
c242c188 | 3329 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
3330 | (void) printf("Unable to expand vdev, state %llu, " |
3331 | "error %d\n", (u_longlong_t)newstate, error); | |
3332 | } | |
3333 | return (vd); | |
3334 | } | |
3335 | ASSERT3U(newstate, ==, VDEV_STATE_HEALTHY); | |
3336 | ||
9babb374 BB |
3337 | /* |
3338 | * Since we dropped the lock we need to ensure that we're | |
3339 | * still talking to the original vdev. It's possible this | |
3340 | * vdev may have been detached/replaced while we were | |
3341 | * trying to online it. | |
3342 | */ | |
428870ff | 3343 | if (generation != spa->spa_config_generation) { |
c242c188 | 3344 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
3345 | (void) printf("vdev configuration has changed, " |
3346 | "guid %llu, state %llu, expected gen %llu, " | |
3347 | "got gen %llu\n", | |
3348 | (u_longlong_t)guid, | |
3349 | (u_longlong_t)tvd->vdev_state, | |
3350 | (u_longlong_t)generation, | |
3351 | (u_longlong_t)spa->spa_config_generation); | |
9babb374 BB |
3352 | } |
3353 | return (vd); | |
3354 | } | |
3355 | return (NULL); | |
3356 | } | |
3357 | ||
3358 | /* | |
3359 | * Traverse the vdev tree calling the supplied function. | |
3360 | * We continue to walk the tree until we either have walked all | |
3361 | * children or we receive a non-NULL return from the callback. | |
3362 | * If a NULL callback is passed, then we just return back the first | |
3363 | * leaf vdev we encounter. | |
3364 | */ | |
3365 | vdev_t * | |
3366 | vdev_walk_tree(vdev_t *vd, vdev_t *(*func)(vdev_t *, void *), void *arg) | |
3367 | { | |
d6320ddb BB |
3368 | uint_t c; |
3369 | ||
9babb374 BB |
3370 | if (vd->vdev_ops->vdev_op_leaf) { |
3371 | if (func == NULL) | |
3372 | return (vd); | |
3373 | else | |
3374 | return (func(vd, arg)); | |
3375 | } | |
3376 | ||
d6320ddb | 3377 | for (c = 0; c < vd->vdev_children; c++) { |
9babb374 BB |
3378 | vdev_t *cvd = vd->vdev_child[c]; |
3379 | if ((cvd = vdev_walk_tree(cvd, func, arg)) != NULL) | |
3380 | return (cvd); | |
3381 | } | |
3382 | return (NULL); | |
3383 | } | |
3384 | ||
34dc7c2f BB |
3385 | /* |
3386 | * Verify that dynamic LUN growth works as expected. | |
3387 | */ | |
428870ff | 3388 | /* ARGSUSED */ |
34dc7c2f | 3389 | void |
428870ff | 3390 | ztest_vdev_LUN_growth(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3391 | { |
c242c188 | 3392 | spa_t *spa = ztest_spa; |
428870ff BB |
3393 | vdev_t *vd, *tvd; |
3394 | metaslab_class_t *mc; | |
3395 | metaslab_group_t *mg; | |
9babb374 | 3396 | size_t psize, newsize; |
428870ff BB |
3397 | uint64_t top; |
3398 | uint64_t old_class_space, new_class_space, old_ms_count, new_ms_count; | |
34dc7c2f | 3399 | |
c242c188 | 3400 | mutex_enter(&ztest_vdev_lock); |
9babb374 BB |
3401 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); |
3402 | ||
428870ff | 3403 | top = ztest_random_vdev_top(spa, B_TRUE); |
9babb374 | 3404 | |
428870ff BB |
3405 | tvd = spa->spa_root_vdev->vdev_child[top]; |
3406 | mg = tvd->vdev_mg; | |
3407 | mc = mg->mg_class; | |
3408 | old_ms_count = tvd->vdev_ms_count; | |
3409 | old_class_space = metaslab_class_get_space(mc); | |
34dc7c2f BB |
3410 | |
3411 | /* | |
9babb374 BB |
3412 | * Determine the size of the first leaf vdev associated with |
3413 | * our top-level device. | |
34dc7c2f | 3414 | */ |
9babb374 BB |
3415 | vd = vdev_walk_tree(tvd, NULL, NULL); |
3416 | ASSERT3P(vd, !=, NULL); | |
3417 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
34dc7c2f | 3418 | |
9babb374 | 3419 | psize = vd->vdev_psize; |
34dc7c2f | 3420 | |
9babb374 | 3421 | /* |
428870ff BB |
3422 | * We only try to expand the vdev if it's healthy, less than 4x its |
3423 | * original size, and it has a valid psize. | |
9babb374 | 3424 | */ |
428870ff | 3425 | if (tvd->vdev_state != VDEV_STATE_HEALTHY || |
c242c188 | 3426 | psize == 0 || psize >= 4 * ztest_opts.zo_vdev_size) { |
9babb374 | 3427 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 3428 | mutex_exit(&ztest_vdev_lock); |
9babb374 BB |
3429 | return; |
3430 | } | |
3431 | ASSERT(psize > 0); | |
3432 | newsize = psize + psize / 8; | |
3433 | ASSERT3U(newsize, >, psize); | |
34dc7c2f | 3434 | |
c242c188 | 3435 | if (ztest_opts.zo_verbose >= 6) { |
428870ff | 3436 | (void) printf("Expanding LUN %s from %lu to %lu\n", |
9babb374 BB |
3437 | vd->vdev_path, (ulong_t)psize, (ulong_t)newsize); |
3438 | } | |
3439 | ||
9babb374 BB |
3440 | /* |
3441 | * Growing the vdev is a two step process: | |
3442 | * 1). expand the physical size (i.e. relabel) | |
3443 | * 2). online the vdev to create the new metaslabs | |
3444 | */ | |
3445 | if (vdev_walk_tree(tvd, grow_vdev, &newsize) != NULL || | |
3446 | vdev_walk_tree(tvd, online_vdev, NULL) != NULL || | |
3447 | tvd->vdev_state != VDEV_STATE_HEALTHY) { | |
c242c188 | 3448 | if (ztest_opts.zo_verbose >= 5) { |
9babb374 | 3449 | (void) printf("Could not expand LUN because " |
428870ff | 3450 | "the vdev configuration changed.\n"); |
34dc7c2f | 3451 | } |
428870ff | 3452 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 3453 | mutex_exit(&ztest_vdev_lock); |
9babb374 | 3454 | return; |
34dc7c2f BB |
3455 | } |
3456 | ||
428870ff | 3457 | spa_config_exit(spa, SCL_STATE, spa); |
9babb374 BB |
3458 | |
3459 | /* | |
3460 | * Expanding the LUN will update the config asynchronously, | |
3461 | * thus we must wait for the async thread to complete any | |
3462 | * pending tasks before proceeding. | |
3463 | */ | |
428870ff BB |
3464 | for (;;) { |
3465 | boolean_t done; | |
3466 | mutex_enter(&spa->spa_async_lock); | |
3467 | done = (spa->spa_async_thread == NULL && !spa->spa_async_tasks); | |
3468 | mutex_exit(&spa->spa_async_lock); | |
3469 | if (done) | |
3470 | break; | |
3471 | txg_wait_synced(spa_get_dsl(spa), 0); | |
3472 | (void) poll(NULL, 0, 100); | |
3473 | } | |
9babb374 BB |
3474 | |
3475 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); | |
428870ff BB |
3476 | |
3477 | tvd = spa->spa_root_vdev->vdev_child[top]; | |
3478 | new_ms_count = tvd->vdev_ms_count; | |
3479 | new_class_space = metaslab_class_get_space(mc); | |
3480 | ||
3481 | if (tvd->vdev_mg != mg || mg->mg_class != mc) { | |
c242c188 | 3482 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
3483 | (void) printf("Could not verify LUN expansion due to " |
3484 | "intervening vdev offline or remove.\n"); | |
3485 | } | |
3486 | spa_config_exit(spa, SCL_STATE, spa); | |
c242c188 | 3487 | mutex_exit(&ztest_vdev_lock); |
428870ff BB |
3488 | return; |
3489 | } | |
3490 | ||
3491 | /* | |
3492 | * Make sure we were able to grow the vdev. | |
3493 | */ | |
3494 | if (new_ms_count <= old_ms_count) | |
3495 | fatal(0, "LUN expansion failed: ms_count %llu <= %llu\n", | |
3496 | old_ms_count, new_ms_count); | |
9babb374 BB |
3497 | |
3498 | /* | |
3499 | * Make sure we were able to grow the pool. | |
3500 | */ | |
428870ff BB |
3501 | if (new_class_space <= old_class_space) |
3502 | fatal(0, "LUN expansion failed: class_space %llu <= %llu\n", | |
3503 | old_class_space, new_class_space); | |
3504 | ||
c242c188 | 3505 | if (ztest_opts.zo_verbose >= 5) { |
9babb374 BB |
3506 | char oldnumbuf[6], newnumbuf[6]; |
3507 | ||
428870ff BB |
3508 | nicenum(old_class_space, oldnumbuf); |
3509 | nicenum(new_class_space, newnumbuf); | |
9babb374 BB |
3510 | (void) printf("%s grew from %s to %s\n", |
3511 | spa->spa_name, oldnumbuf, newnumbuf); | |
3512 | } | |
428870ff | 3513 | |
9babb374 | 3514 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 3515 | mutex_exit(&ztest_vdev_lock); |
34dc7c2f BB |
3516 | } |
3517 | ||
428870ff BB |
3518 | /* |
3519 | * Verify that dmu_objset_{create,destroy,open,close} work as expected. | |
3520 | */ | |
34dc7c2f BB |
3521 | /* ARGSUSED */ |
3522 | static void | |
428870ff | 3523 | ztest_objset_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) |
34dc7c2f BB |
3524 | { |
3525 | /* | |
428870ff | 3526 | * Create the objects common to all ztest datasets. |
34dc7c2f | 3527 | */ |
428870ff | 3528 | VERIFY(zap_create_claim(os, ZTEST_DIROBJ, |
34dc7c2f | 3529 | DMU_OT_ZAP_OTHER, DMU_OT_NONE, 0, tx) == 0); |
428870ff | 3530 | } |
34dc7c2f | 3531 | |
428870ff BB |
3532 | static int |
3533 | ztest_dataset_create(char *dsname) | |
3534 | { | |
3535 | uint64_t zilset = ztest_random(100); | |
b5256303 | 3536 | int err = dmu_objset_create(dsname, DMU_OST_OTHER, 0, NULL, |
428870ff BB |
3537 | ztest_objset_create_cb, NULL); |
3538 | ||
3539 | if (err || zilset < 80) | |
3540 | return (err); | |
3541 | ||
c242c188 | 3542 | if (ztest_opts.zo_verbose >= 5) |
b815ff9a | 3543 | (void) printf("Setting dataset %s to sync always\n", dsname); |
428870ff BB |
3544 | return (ztest_dsl_prop_set_uint64(dsname, ZFS_PROP_SYNC, |
3545 | ZFS_SYNC_ALWAYS, B_FALSE)); | |
34dc7c2f BB |
3546 | } |
3547 | ||
428870ff | 3548 | /* ARGSUSED */ |
34dc7c2f | 3549 | static int |
428870ff | 3550 | ztest_objset_destroy_cb(const char *name, void *arg) |
34dc7c2f | 3551 | { |
34dc7c2f | 3552 | objset_t *os; |
428870ff | 3553 | dmu_object_info_t doi; |
34dc7c2f BB |
3554 | int error; |
3555 | ||
3556 | /* | |
3557 | * Verify that the dataset contains a directory object. | |
3558 | */ | |
b5256303 | 3559 | VERIFY0(dmu_objset_own(name, DMU_OST_OTHER, B_TRUE, B_TRUE, FTAG, &os)); |
428870ff | 3560 | error = dmu_object_info(os, ZTEST_DIROBJ, &doi); |
34dc7c2f BB |
3561 | if (error != ENOENT) { |
3562 | /* We could have crashed in the middle of destroying it */ | |
c99c9001 | 3563 | ASSERT0(error); |
428870ff BB |
3564 | ASSERT3U(doi.doi_type, ==, DMU_OT_ZAP_OTHER); |
3565 | ASSERT3S(doi.doi_physical_blocks_512, >=, 0); | |
34dc7c2f | 3566 | } |
b5256303 | 3567 | dmu_objset_disown(os, B_TRUE, FTAG); |
34dc7c2f BB |
3568 | |
3569 | /* | |
3570 | * Destroy the dataset. | |
3571 | */ | |
13fe0198 | 3572 | if (strchr(name, '@') != NULL) { |
dc1fbc43 | 3573 | VERIFY0(dsl_destroy_snapshot(name, B_TRUE)); |
13fe0198 | 3574 | } else { |
dc1fbc43 GM |
3575 | error = dsl_destroy_head(name); |
3576 | /* There could be a hold on this dataset */ | |
3577 | if (error != EBUSY) | |
3578 | ASSERT0(error); | |
13fe0198 | 3579 | } |
34dc7c2f BB |
3580 | return (0); |
3581 | } | |
3582 | ||
428870ff BB |
3583 | static boolean_t |
3584 | ztest_snapshot_create(char *osname, uint64_t id) | |
34dc7c2f | 3585 | { |
eca7b760 | 3586 | char snapname[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
3587 | int error; |
3588 | ||
13fe0198 | 3589 | (void) snprintf(snapname, sizeof (snapname), "%llu", (u_longlong_t)id); |
428870ff | 3590 | |
13fe0198 | 3591 | error = dmu_objset_snapshot_one(osname, snapname); |
428870ff BB |
3592 | if (error == ENOSPC) { |
3593 | ztest_record_enospc(FTAG); | |
3594 | return (B_FALSE); | |
3595 | } | |
13fe0198 MA |
3596 | if (error != 0 && error != EEXIST) { |
3597 | fatal(0, "ztest_snapshot_create(%s@%s) = %d", osname, | |
3598 | snapname, error); | |
3599 | } | |
428870ff BB |
3600 | return (B_TRUE); |
3601 | } | |
3602 | ||
3603 | static boolean_t | |
3604 | ztest_snapshot_destroy(char *osname, uint64_t id) | |
3605 | { | |
eca7b760 | 3606 | char snapname[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
3607 | int error; |
3608 | ||
eca7b760 | 3609 | (void) snprintf(snapname, sizeof (snapname), "%s@%llu", osname, |
428870ff BB |
3610 | (u_longlong_t)id); |
3611 | ||
13fe0198 | 3612 | error = dsl_destroy_snapshot(snapname, B_FALSE); |
428870ff BB |
3613 | if (error != 0 && error != ENOENT) |
3614 | fatal(0, "ztest_snapshot_destroy(%s) = %d", snapname, error); | |
3615 | return (B_TRUE); | |
34dc7c2f BB |
3616 | } |
3617 | ||
428870ff | 3618 | /* ARGSUSED */ |
34dc7c2f | 3619 | void |
428870ff | 3620 | ztest_dmu_objset_create_destroy(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3621 | { |
40b84e7a | 3622 | ztest_ds_t *zdtmp; |
428870ff | 3623 | int iters; |
34dc7c2f | 3624 | int error; |
b128c09f | 3625 | objset_t *os, *os2; |
eca7b760 | 3626 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
34dc7c2f | 3627 | zilog_t *zilog; |
d6320ddb | 3628 | int i; |
34dc7c2f | 3629 | |
40b84e7a | 3630 | zdtmp = umem_alloc(sizeof (ztest_ds_t), UMEM_NOFAIL); |
40b84e7a | 3631 | |
7809eb8b | 3632 | (void) rw_rdlock(&ztest_name_lock); |
34dc7c2f | 3633 | |
eca7b760 | 3634 | (void) snprintf(name, sizeof (name), "%s/temp_%llu", |
c242c188 | 3635 | ztest_opts.zo_pool, (u_longlong_t)id); |
34dc7c2f BB |
3636 | |
3637 | /* | |
3638 | * If this dataset exists from a previous run, process its replay log | |
13fe0198 | 3639 | * half of the time. If we don't replay it, then dsl_destroy_head() |
428870ff | 3640 | * (invoked from ztest_objset_destroy_cb()) should just throw it away. |
34dc7c2f BB |
3641 | */ |
3642 | if (ztest_random(2) == 0 && | |
b5256303 TC |
3643 | dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, |
3644 | B_TRUE, FTAG, &os) == 0) { | |
c242c188 | 3645 | ztest_zd_init(zdtmp, NULL, os); |
40b84e7a BB |
3646 | zil_replay(os, zdtmp, ztest_replay_vector); |
3647 | ztest_zd_fini(zdtmp); | |
b5256303 | 3648 | dmu_objset_disown(os, B_TRUE, FTAG); |
34dc7c2f BB |
3649 | } |
3650 | ||
3651 | /* | |
3652 | * There may be an old instance of the dataset we're about to | |
3653 | * create lying around from a previous run. If so, destroy it | |
3654 | * and all of its snapshots. | |
3655 | */ | |
428870ff | 3656 | (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL, |
34dc7c2f BB |
3657 | DS_FIND_CHILDREN | DS_FIND_SNAPSHOTS); |
3658 | ||
3659 | /* | |
3660 | * Verify that the destroyed dataset is no longer in the namespace. | |
3661 | */ | |
b5256303 | 3662 | VERIFY3U(ENOENT, ==, dmu_objset_own(name, DMU_OST_OTHER, B_TRUE, B_TRUE, |
13fe0198 | 3663 | FTAG, &os)); |
34dc7c2f BB |
3664 | |
3665 | /* | |
3666 | * Verify that we can create a new dataset. | |
3667 | */ | |
428870ff | 3668 | error = ztest_dataset_create(name); |
34dc7c2f BB |
3669 | if (error) { |
3670 | if (error == ENOSPC) { | |
428870ff | 3671 | ztest_record_enospc(FTAG); |
40b84e7a | 3672 | goto out; |
34dc7c2f BB |
3673 | } |
3674 | fatal(0, "dmu_objset_create(%s) = %d", name, error); | |
3675 | } | |
3676 | ||
b5256303 TC |
3677 | VERIFY0(dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, B_TRUE, |
3678 | FTAG, &os)); | |
428870ff | 3679 | |
c242c188 | 3680 | ztest_zd_init(zdtmp, NULL, os); |
34dc7c2f BB |
3681 | |
3682 | /* | |
3683 | * Open the intent log for it. | |
3684 | */ | |
428870ff | 3685 | zilog = zil_open(os, ztest_get_data); |
34dc7c2f BB |
3686 | |
3687 | /* | |
428870ff BB |
3688 | * Put some objects in there, do a little I/O to them, |
3689 | * and randomly take a couple of snapshots along the way. | |
34dc7c2f | 3690 | */ |
428870ff | 3691 | iters = ztest_random(5); |
d6320ddb | 3692 | for (i = 0; i < iters; i++) { |
40b84e7a | 3693 | ztest_dmu_object_alloc_free(zdtmp, id); |
428870ff BB |
3694 | if (ztest_random(iters) == 0) |
3695 | (void) ztest_snapshot_create(name, i); | |
34dc7c2f BB |
3696 | } |
3697 | ||
3698 | /* | |
3699 | * Verify that we cannot create an existing dataset. | |
3700 | */ | |
428870ff | 3701 | VERIFY3U(EEXIST, ==, |
b5256303 | 3702 | dmu_objset_create(name, DMU_OST_OTHER, 0, NULL, NULL, NULL)); |
34dc7c2f BB |
3703 | |
3704 | /* | |
428870ff | 3705 | * Verify that we can hold an objset that is also owned. |
b128c09f | 3706 | */ |
428870ff BB |
3707 | VERIFY3U(0, ==, dmu_objset_hold(name, FTAG, &os2)); |
3708 | dmu_objset_rele(os2, FTAG); | |
34dc7c2f | 3709 | |
428870ff BB |
3710 | /* |
3711 | * Verify that we cannot own an objset that is already owned. | |
3712 | */ | |
3713 | VERIFY3U(EBUSY, ==, | |
b5256303 | 3714 | dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, B_TRUE, FTAG, &os2)); |
34dc7c2f | 3715 | |
428870ff | 3716 | zil_close(zilog); |
b5256303 | 3717 | dmu_objset_disown(os, B_TRUE, FTAG); |
40b84e7a BB |
3718 | ztest_zd_fini(zdtmp); |
3719 | out: | |
7809eb8b | 3720 | (void) rw_unlock(&ztest_name_lock); |
40b84e7a | 3721 | |
40b84e7a | 3722 | umem_free(zdtmp, sizeof (ztest_ds_t)); |
34dc7c2f BB |
3723 | } |
3724 | ||
3725 | /* | |
3726 | * Verify that dmu_snapshot_{create,destroy,open,close} work as expected. | |
3727 | */ | |
3728 | void | |
428870ff | 3729 | ztest_dmu_snapshot_create_destroy(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3730 | { |
7809eb8b | 3731 | (void) rw_rdlock(&ztest_name_lock); |
428870ff BB |
3732 | (void) ztest_snapshot_destroy(zd->zd_name, id); |
3733 | (void) ztest_snapshot_create(zd->zd_name, id); | |
7809eb8b | 3734 | (void) rw_unlock(&ztest_name_lock); |
34dc7c2f BB |
3735 | } |
3736 | ||
9babb374 BB |
3737 | /* |
3738 | * Cleanup non-standard snapshots and clones. | |
3739 | */ | |
3740 | void | |
428870ff | 3741 | ztest_dsl_dataset_cleanup(char *osname, uint64_t id) |
9babb374 | 3742 | { |
40b84e7a BB |
3743 | char *snap1name; |
3744 | char *clone1name; | |
3745 | char *snap2name; | |
3746 | char *clone2name; | |
3747 | char *snap3name; | |
9babb374 BB |
3748 | int error; |
3749 | ||
eca7b760 IK |
3750 | snap1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); |
3751 | clone1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
3752 | snap2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
3753 | clone2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
3754 | snap3name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
3755 | ||
3756 | (void) snprintf(snap1name, ZFS_MAX_DATASET_NAME_LEN, | |
3757 | "%s@s1_%llu", osname, (u_longlong_t)id); | |
3758 | (void) snprintf(clone1name, ZFS_MAX_DATASET_NAME_LEN, | |
3759 | "%s/c1_%llu", osname, (u_longlong_t)id); | |
3760 | (void) snprintf(snap2name, ZFS_MAX_DATASET_NAME_LEN, | |
3761 | "%s@s2_%llu", clone1name, (u_longlong_t)id); | |
3762 | (void) snprintf(clone2name, ZFS_MAX_DATASET_NAME_LEN, | |
3763 | "%s/c2_%llu", osname, (u_longlong_t)id); | |
3764 | (void) snprintf(snap3name, ZFS_MAX_DATASET_NAME_LEN, | |
3765 | "%s@s3_%llu", clone1name, (u_longlong_t)id); | |
9babb374 | 3766 | |
13fe0198 | 3767 | error = dsl_destroy_head(clone2name); |
9babb374 | 3768 | if (error && error != ENOENT) |
13fe0198 MA |
3769 | fatal(0, "dsl_destroy_head(%s) = %d", clone2name, error); |
3770 | error = dsl_destroy_snapshot(snap3name, B_FALSE); | |
9babb374 | 3771 | if (error && error != ENOENT) |
13fe0198 MA |
3772 | fatal(0, "dsl_destroy_snapshot(%s) = %d", snap3name, error); |
3773 | error = dsl_destroy_snapshot(snap2name, B_FALSE); | |
9babb374 | 3774 | if (error && error != ENOENT) |
13fe0198 MA |
3775 | fatal(0, "dsl_destroy_snapshot(%s) = %d", snap2name, error); |
3776 | error = dsl_destroy_head(clone1name); | |
9babb374 | 3777 | if (error && error != ENOENT) |
13fe0198 MA |
3778 | fatal(0, "dsl_destroy_head(%s) = %d", clone1name, error); |
3779 | error = dsl_destroy_snapshot(snap1name, B_FALSE); | |
9babb374 | 3780 | if (error && error != ENOENT) |
13fe0198 | 3781 | fatal(0, "dsl_destroy_snapshot(%s) = %d", snap1name, error); |
40b84e7a | 3782 | |
eca7b760 IK |
3783 | umem_free(snap1name, ZFS_MAX_DATASET_NAME_LEN); |
3784 | umem_free(clone1name, ZFS_MAX_DATASET_NAME_LEN); | |
3785 | umem_free(snap2name, ZFS_MAX_DATASET_NAME_LEN); | |
3786 | umem_free(clone2name, ZFS_MAX_DATASET_NAME_LEN); | |
3787 | umem_free(snap3name, ZFS_MAX_DATASET_NAME_LEN); | |
9babb374 BB |
3788 | } |
3789 | ||
3790 | /* | |
3791 | * Verify dsl_dataset_promote handles EBUSY | |
3792 | */ | |
3793 | void | |
428870ff | 3794 | ztest_dsl_dataset_promote_busy(ztest_ds_t *zd, uint64_t id) |
9babb374 | 3795 | { |
13fe0198 | 3796 | objset_t *os; |
40b84e7a BB |
3797 | char *snap1name; |
3798 | char *clone1name; | |
3799 | char *snap2name; | |
3800 | char *clone2name; | |
3801 | char *snap3name; | |
428870ff BB |
3802 | char *osname = zd->zd_name; |
3803 | int error; | |
9babb374 | 3804 | |
eca7b760 IK |
3805 | snap1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); |
3806 | clone1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
3807 | snap2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
3808 | clone2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
3809 | snap3name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
40b84e7a | 3810 | |
7809eb8b | 3811 | (void) rw_rdlock(&ztest_name_lock); |
9babb374 | 3812 | |
428870ff | 3813 | ztest_dsl_dataset_cleanup(osname, id); |
9babb374 | 3814 | |
eca7b760 IK |
3815 | (void) snprintf(snap1name, ZFS_MAX_DATASET_NAME_LEN, |
3816 | "%s@s1_%llu", osname, (u_longlong_t)id); | |
3817 | (void) snprintf(clone1name, ZFS_MAX_DATASET_NAME_LEN, | |
3818 | "%s/c1_%llu", osname, (u_longlong_t)id); | |
3819 | (void) snprintf(snap2name, ZFS_MAX_DATASET_NAME_LEN, | |
3820 | "%s@s2_%llu", clone1name, (u_longlong_t)id); | |
3821 | (void) snprintf(clone2name, ZFS_MAX_DATASET_NAME_LEN, | |
3822 | "%s/c2_%llu", osname, (u_longlong_t)id); | |
3823 | (void) snprintf(snap3name, ZFS_MAX_DATASET_NAME_LEN, | |
3824 | "%s@s3_%llu", clone1name, (u_longlong_t)id); | |
9babb374 | 3825 | |
6f1ffb06 | 3826 | error = dmu_objset_snapshot_one(osname, strchr(snap1name, '@') + 1); |
9babb374 BB |
3827 | if (error && error != EEXIST) { |
3828 | if (error == ENOSPC) { | |
428870ff | 3829 | ztest_record_enospc(FTAG); |
9babb374 BB |
3830 | goto out; |
3831 | } | |
3832 | fatal(0, "dmu_take_snapshot(%s) = %d", snap1name, error); | |
3833 | } | |
3834 | ||
13fe0198 | 3835 | error = dmu_objset_clone(clone1name, snap1name); |
9babb374 BB |
3836 | if (error) { |
3837 | if (error == ENOSPC) { | |
428870ff | 3838 | ztest_record_enospc(FTAG); |
9babb374 BB |
3839 | goto out; |
3840 | } | |
3841 | fatal(0, "dmu_objset_create(%s) = %d", clone1name, error); | |
3842 | } | |
3843 | ||
6f1ffb06 | 3844 | error = dmu_objset_snapshot_one(clone1name, strchr(snap2name, '@') + 1); |
428870ff BB |
3845 | if (error && error != EEXIST) { |
3846 | if (error == ENOSPC) { | |
3847 | ztest_record_enospc(FTAG); | |
3848 | goto out; | |
34dc7c2f | 3849 | } |
428870ff BB |
3850 | fatal(0, "dmu_open_snapshot(%s) = %d", snap2name, error); |
3851 | } | |
34dc7c2f | 3852 | |
6f1ffb06 | 3853 | error = dmu_objset_snapshot_one(clone1name, strchr(snap3name, '@') + 1); |
428870ff BB |
3854 | if (error && error != EEXIST) { |
3855 | if (error == ENOSPC) { | |
3856 | ztest_record_enospc(FTAG); | |
3857 | goto out; | |
3858 | } | |
3859 | fatal(0, "dmu_open_snapshot(%s) = %d", snap3name, error); | |
3860 | } | |
34dc7c2f | 3861 | |
13fe0198 | 3862 | error = dmu_objset_clone(clone2name, snap3name); |
428870ff BB |
3863 | if (error) { |
3864 | if (error == ENOSPC) { | |
3865 | ztest_record_enospc(FTAG); | |
3866 | goto out; | |
3867 | } | |
3868 | fatal(0, "dmu_objset_create(%s) = %d", clone2name, error); | |
3869 | } | |
34dc7c2f | 3870 | |
b5256303 TC |
3871 | error = dmu_objset_own(snap2name, DMU_OST_ANY, B_TRUE, B_TRUE, |
3872 | FTAG, &os); | |
428870ff | 3873 | if (error) |
13fe0198 | 3874 | fatal(0, "dmu_objset_own(%s) = %d", snap2name, error); |
428870ff | 3875 | error = dsl_dataset_promote(clone2name, NULL); |
9b67f605 | 3876 | if (error == ENOSPC) { |
b5256303 | 3877 | dmu_objset_disown(os, B_TRUE, FTAG); |
9b67f605 MA |
3878 | ztest_record_enospc(FTAG); |
3879 | goto out; | |
3880 | } | |
428870ff BB |
3881 | if (error != EBUSY) |
3882 | fatal(0, "dsl_dataset_promote(%s), %d, not EBUSY", clone2name, | |
3883 | error); | |
b5256303 | 3884 | dmu_objset_disown(os, B_TRUE, FTAG); |
34dc7c2f | 3885 | |
428870ff BB |
3886 | out: |
3887 | ztest_dsl_dataset_cleanup(osname, id); | |
34dc7c2f | 3888 | |
7809eb8b | 3889 | (void) rw_unlock(&ztest_name_lock); |
40b84e7a | 3890 | |
eca7b760 IK |
3891 | umem_free(snap1name, ZFS_MAX_DATASET_NAME_LEN); |
3892 | umem_free(clone1name, ZFS_MAX_DATASET_NAME_LEN); | |
3893 | umem_free(snap2name, ZFS_MAX_DATASET_NAME_LEN); | |
3894 | umem_free(clone2name, ZFS_MAX_DATASET_NAME_LEN); | |
3895 | umem_free(snap3name, ZFS_MAX_DATASET_NAME_LEN); | |
428870ff | 3896 | } |
34dc7c2f | 3897 | |
40b84e7a | 3898 | #undef OD_ARRAY_SIZE |
d1d7e268 | 3899 | #define OD_ARRAY_SIZE 4 |
40b84e7a | 3900 | |
428870ff BB |
3901 | /* |
3902 | * Verify that dmu_object_{alloc,free} work as expected. | |
3903 | */ | |
3904 | void | |
3905 | ztest_dmu_object_alloc_free(ztest_ds_t *zd, uint64_t id) | |
3906 | { | |
40b84e7a BB |
3907 | ztest_od_t *od; |
3908 | int batchsize; | |
3909 | int size; | |
d6320ddb | 3910 | int b; |
34dc7c2f | 3911 | |
d1d7e268 | 3912 | size = sizeof (ztest_od_t) * OD_ARRAY_SIZE; |
40b84e7a BB |
3913 | od = umem_alloc(size, UMEM_NOFAIL); |
3914 | batchsize = OD_ARRAY_SIZE; | |
3915 | ||
d6320ddb | 3916 | for (b = 0; b < batchsize; b++) |
50c957f7 NB |
3917 | ztest_od_init(od + b, id, FTAG, b, DMU_OT_UINT64_OTHER, |
3918 | 0, 0, 0); | |
34dc7c2f | 3919 | |
428870ff BB |
3920 | /* |
3921 | * Destroy the previous batch of objects, create a new batch, | |
3922 | * and do some I/O on the new objects. | |
3923 | */ | |
40b84e7a | 3924 | if (ztest_object_init(zd, od, size, B_TRUE) != 0) |
428870ff | 3925 | return; |
34dc7c2f | 3926 | |
428870ff BB |
3927 | while (ztest_random(4 * batchsize) != 0) |
3928 | ztest_io(zd, od[ztest_random(batchsize)].od_object, | |
3929 | ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
40b84e7a BB |
3930 | |
3931 | umem_free(od, size); | |
34dc7c2f BB |
3932 | } |
3933 | ||
40b84e7a | 3934 | #undef OD_ARRAY_SIZE |
d1d7e268 | 3935 | #define OD_ARRAY_SIZE 2 |
40b84e7a | 3936 | |
34dc7c2f BB |
3937 | /* |
3938 | * Verify that dmu_{read,write} work as expected. | |
3939 | */ | |
34dc7c2f | 3940 | void |
428870ff | 3941 | ztest_dmu_read_write(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3942 | { |
40b84e7a BB |
3943 | int size; |
3944 | ztest_od_t *od; | |
3945 | ||
428870ff | 3946 | objset_t *os = zd->zd_os; |
d1d7e268 | 3947 | size = sizeof (ztest_od_t) * OD_ARRAY_SIZE; |
40b84e7a | 3948 | od = umem_alloc(size, UMEM_NOFAIL); |
34dc7c2f BB |
3949 | dmu_tx_t *tx; |
3950 | int i, freeit, error; | |
3951 | uint64_t n, s, txg; | |
3952 | bufwad_t *packbuf, *bigbuf, *pack, *bigH, *bigT; | |
428870ff BB |
3953 | uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize; |
3954 | uint64_t chunksize = (1000 + ztest_random(1000)) * sizeof (uint64_t); | |
34dc7c2f BB |
3955 | uint64_t regions = 997; |
3956 | uint64_t stride = 123456789ULL; | |
3957 | uint64_t width = 40; | |
3958 | int free_percent = 5; | |
3959 | ||
3960 | /* | |
3961 | * This test uses two objects, packobj and bigobj, that are always | |
3962 | * updated together (i.e. in the same tx) so that their contents are | |
3963 | * in sync and can be compared. Their contents relate to each other | |
3964 | * in a simple way: packobj is a dense array of 'bufwad' structures, | |
3965 | * while bigobj is a sparse array of the same bufwads. Specifically, | |
3966 | * for any index n, there are three bufwads that should be identical: | |
3967 | * | |
3968 | * packobj, at offset n * sizeof (bufwad_t) | |
3969 | * bigobj, at the head of the nth chunk | |
3970 | * bigobj, at the tail of the nth chunk | |
3971 | * | |
3972 | * The chunk size is arbitrary. It doesn't have to be a power of two, | |
3973 | * and it doesn't have any relation to the object blocksize. | |
3974 | * The only requirement is that it can hold at least two bufwads. | |
3975 | * | |
3976 | * Normally, we write the bufwad to each of these locations. | |
3977 | * However, free_percent of the time we instead write zeroes to | |
3978 | * packobj and perform a dmu_free_range() on bigobj. By comparing | |
3979 | * bigobj to packobj, we can verify that the DMU is correctly | |
3980 | * tracking which parts of an object are allocated and free, | |
3981 | * and that the contents of the allocated blocks are correct. | |
3982 | */ | |
3983 | ||
3984 | /* | |
3985 | * Read the directory info. If it's the first time, set things up. | |
3986 | */ | |
50c957f7 NB |
3987 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, chunksize); |
3988 | ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, 0, | |
3989 | chunksize); | |
34dc7c2f | 3990 | |
40b84e7a BB |
3991 | if (ztest_object_init(zd, od, size, B_FALSE) != 0) { |
3992 | umem_free(od, size); | |
428870ff | 3993 | return; |
40b84e7a | 3994 | } |
34dc7c2f | 3995 | |
428870ff BB |
3996 | bigobj = od[0].od_object; |
3997 | packobj = od[1].od_object; | |
3998 | chunksize = od[0].od_gen; | |
3999 | ASSERT(chunksize == od[1].od_gen); | |
34dc7c2f BB |
4000 | |
4001 | /* | |
4002 | * Prefetch a random chunk of the big object. | |
4003 | * Our aim here is to get some async reads in flight | |
4004 | * for blocks that we may free below; the DMU should | |
4005 | * handle this race correctly. | |
4006 | */ | |
4007 | n = ztest_random(regions) * stride + ztest_random(width); | |
4008 | s = 1 + ztest_random(2 * width - 1); | |
fcff0f35 PD |
4009 | dmu_prefetch(os, bigobj, 0, n * chunksize, s * chunksize, |
4010 | ZIO_PRIORITY_SYNC_READ); | |
34dc7c2f BB |
4011 | |
4012 | /* | |
4013 | * Pick a random index and compute the offsets into packobj and bigobj. | |
4014 | */ | |
4015 | n = ztest_random(regions) * stride + ztest_random(width); | |
4016 | s = 1 + ztest_random(width - 1); | |
4017 | ||
4018 | packoff = n * sizeof (bufwad_t); | |
4019 | packsize = s * sizeof (bufwad_t); | |
4020 | ||
428870ff BB |
4021 | bigoff = n * chunksize; |
4022 | bigsize = s * chunksize; | |
34dc7c2f BB |
4023 | |
4024 | packbuf = umem_alloc(packsize, UMEM_NOFAIL); | |
4025 | bigbuf = umem_alloc(bigsize, UMEM_NOFAIL); | |
4026 | ||
4027 | /* | |
4028 | * free_percent of the time, free a range of bigobj rather than | |
4029 | * overwriting it. | |
4030 | */ | |
4031 | freeit = (ztest_random(100) < free_percent); | |
4032 | ||
4033 | /* | |
4034 | * Read the current contents of our objects. | |
4035 | */ | |
428870ff | 4036 | error = dmu_read(os, packobj, packoff, packsize, packbuf, |
9babb374 | 4037 | DMU_READ_PREFETCH); |
c99c9001 | 4038 | ASSERT0(error); |
428870ff | 4039 | error = dmu_read(os, bigobj, bigoff, bigsize, bigbuf, |
9babb374 | 4040 | DMU_READ_PREFETCH); |
c99c9001 | 4041 | ASSERT0(error); |
34dc7c2f BB |
4042 | |
4043 | /* | |
4044 | * Get a tx for the mods to both packobj and bigobj. | |
4045 | */ | |
4046 | tx = dmu_tx_create(os); | |
4047 | ||
428870ff | 4048 | dmu_tx_hold_write(tx, packobj, packoff, packsize); |
34dc7c2f BB |
4049 | |
4050 | if (freeit) | |
428870ff | 4051 | dmu_tx_hold_free(tx, bigobj, bigoff, bigsize); |
34dc7c2f | 4052 | else |
428870ff | 4053 | dmu_tx_hold_write(tx, bigobj, bigoff, bigsize); |
34dc7c2f | 4054 | |
383fc4a9 MA |
4055 | /* This accounts for setting the checksum/compression. */ |
4056 | dmu_tx_hold_bonus(tx, bigobj); | |
4057 | ||
428870ff BB |
4058 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); |
4059 | if (txg == 0) { | |
34dc7c2f BB |
4060 | umem_free(packbuf, packsize); |
4061 | umem_free(bigbuf, bigsize); | |
40b84e7a | 4062 | umem_free(od, size); |
34dc7c2f BB |
4063 | return; |
4064 | } | |
4065 | ||
9b67f605 MA |
4066 | enum zio_checksum cksum; |
4067 | do { | |
4068 | cksum = (enum zio_checksum) | |
4069 | ztest_random_dsl_prop(ZFS_PROP_CHECKSUM); | |
4070 | } while (cksum >= ZIO_CHECKSUM_LEGACY_FUNCTIONS); | |
4071 | dmu_object_set_checksum(os, bigobj, cksum, tx); | |
428870ff | 4072 | |
9b67f605 MA |
4073 | enum zio_compress comp; |
4074 | do { | |
4075 | comp = (enum zio_compress) | |
4076 | ztest_random_dsl_prop(ZFS_PROP_COMPRESSION); | |
4077 | } while (comp >= ZIO_COMPRESS_LEGACY_FUNCTIONS); | |
4078 | dmu_object_set_compress(os, bigobj, comp, tx); | |
34dc7c2f BB |
4079 | |
4080 | /* | |
4081 | * For each index from n to n + s, verify that the existing bufwad | |
4082 | * in packobj matches the bufwads at the head and tail of the | |
4083 | * corresponding chunk in bigobj. Then update all three bufwads | |
4084 | * with the new values we want to write out. | |
4085 | */ | |
4086 | for (i = 0; i < s; i++) { | |
4087 | /* LINTED */ | |
4088 | pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t)); | |
4089 | /* LINTED */ | |
428870ff | 4090 | bigH = (bufwad_t *)((char *)bigbuf + i * chunksize); |
34dc7c2f | 4091 | /* LINTED */ |
428870ff | 4092 | bigT = (bufwad_t *)((char *)bigH + chunksize) - 1; |
34dc7c2f BB |
4093 | |
4094 | ASSERT((uintptr_t)bigH - (uintptr_t)bigbuf < bigsize); | |
4095 | ASSERT((uintptr_t)bigT - (uintptr_t)bigbuf < bigsize); | |
4096 | ||
4097 | if (pack->bw_txg > txg) | |
4098 | fatal(0, "future leak: got %llx, open txg is %llx", | |
4099 | pack->bw_txg, txg); | |
4100 | ||
4101 | if (pack->bw_data != 0 && pack->bw_index != n + i) | |
4102 | fatal(0, "wrong index: got %llx, wanted %llx+%llx", | |
4103 | pack->bw_index, n, i); | |
4104 | ||
4105 | if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0) | |
4106 | fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH); | |
4107 | ||
4108 | if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0) | |
4109 | fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT); | |
4110 | ||
4111 | if (freeit) { | |
4112 | bzero(pack, sizeof (bufwad_t)); | |
4113 | } else { | |
4114 | pack->bw_index = n + i; | |
4115 | pack->bw_txg = txg; | |
4116 | pack->bw_data = 1 + ztest_random(-2ULL); | |
4117 | } | |
4118 | *bigH = *pack; | |
4119 | *bigT = *pack; | |
4120 | } | |
4121 | ||
4122 | /* | |
4123 | * We've verified all the old bufwads, and made new ones. | |
4124 | * Now write them out. | |
4125 | */ | |
428870ff | 4126 | dmu_write(os, packobj, packoff, packsize, packbuf, tx); |
34dc7c2f BB |
4127 | |
4128 | if (freeit) { | |
c242c188 | 4129 | if (ztest_opts.zo_verbose >= 7) { |
34dc7c2f BB |
4130 | (void) printf("freeing offset %llx size %llx" |
4131 | " txg %llx\n", | |
4132 | (u_longlong_t)bigoff, | |
4133 | (u_longlong_t)bigsize, | |
4134 | (u_longlong_t)txg); | |
4135 | } | |
428870ff | 4136 | VERIFY(0 == dmu_free_range(os, bigobj, bigoff, bigsize, tx)); |
34dc7c2f | 4137 | } else { |
c242c188 | 4138 | if (ztest_opts.zo_verbose >= 7) { |
34dc7c2f BB |
4139 | (void) printf("writing offset %llx size %llx" |
4140 | " txg %llx\n", | |
4141 | (u_longlong_t)bigoff, | |
4142 | (u_longlong_t)bigsize, | |
4143 | (u_longlong_t)txg); | |
4144 | } | |
428870ff | 4145 | dmu_write(os, bigobj, bigoff, bigsize, bigbuf, tx); |
34dc7c2f BB |
4146 | } |
4147 | ||
4148 | dmu_tx_commit(tx); | |
4149 | ||
4150 | /* | |
4151 | * Sanity check the stuff we just wrote. | |
4152 | */ | |
4153 | { | |
4154 | void *packcheck = umem_alloc(packsize, UMEM_NOFAIL); | |
4155 | void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL); | |
4156 | ||
428870ff | 4157 | VERIFY(0 == dmu_read(os, packobj, packoff, |
9babb374 | 4158 | packsize, packcheck, DMU_READ_PREFETCH)); |
428870ff | 4159 | VERIFY(0 == dmu_read(os, bigobj, bigoff, |
9babb374 | 4160 | bigsize, bigcheck, DMU_READ_PREFETCH)); |
34dc7c2f BB |
4161 | |
4162 | ASSERT(bcmp(packbuf, packcheck, packsize) == 0); | |
4163 | ASSERT(bcmp(bigbuf, bigcheck, bigsize) == 0); | |
4164 | ||
4165 | umem_free(packcheck, packsize); | |
4166 | umem_free(bigcheck, bigsize); | |
4167 | } | |
4168 | ||
4169 | umem_free(packbuf, packsize); | |
4170 | umem_free(bigbuf, bigsize); | |
40b84e7a | 4171 | umem_free(od, size); |
34dc7c2f BB |
4172 | } |
4173 | ||
9babb374 BB |
4174 | void |
4175 | compare_and_update_pbbufs(uint64_t s, bufwad_t *packbuf, bufwad_t *bigbuf, | |
428870ff | 4176 | uint64_t bigsize, uint64_t n, uint64_t chunksize, uint64_t txg) |
9babb374 BB |
4177 | { |
4178 | uint64_t i; | |
4179 | bufwad_t *pack; | |
4180 | bufwad_t *bigH; | |
4181 | bufwad_t *bigT; | |
4182 | ||
4183 | /* | |
4184 | * For each index from n to n + s, verify that the existing bufwad | |
4185 | * in packobj matches the bufwads at the head and tail of the | |
4186 | * corresponding chunk in bigobj. Then update all three bufwads | |
4187 | * with the new values we want to write out. | |
4188 | */ | |
4189 | for (i = 0; i < s; i++) { | |
4190 | /* LINTED */ | |
4191 | pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t)); | |
4192 | /* LINTED */ | |
428870ff | 4193 | bigH = (bufwad_t *)((char *)bigbuf + i * chunksize); |
9babb374 | 4194 | /* LINTED */ |
428870ff | 4195 | bigT = (bufwad_t *)((char *)bigH + chunksize) - 1; |
9babb374 BB |
4196 | |
4197 | ASSERT((uintptr_t)bigH - (uintptr_t)bigbuf < bigsize); | |
4198 | ASSERT((uintptr_t)bigT - (uintptr_t)bigbuf < bigsize); | |
4199 | ||
4200 | if (pack->bw_txg > txg) | |
4201 | fatal(0, "future leak: got %llx, open txg is %llx", | |
4202 | pack->bw_txg, txg); | |
4203 | ||
4204 | if (pack->bw_data != 0 && pack->bw_index != n + i) | |
4205 | fatal(0, "wrong index: got %llx, wanted %llx+%llx", | |
4206 | pack->bw_index, n, i); | |
4207 | ||
4208 | if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0) | |
4209 | fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH); | |
4210 | ||
4211 | if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0) | |
4212 | fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT); | |
4213 | ||
4214 | pack->bw_index = n + i; | |
4215 | pack->bw_txg = txg; | |
4216 | pack->bw_data = 1 + ztest_random(-2ULL); | |
4217 | ||
4218 | *bigH = *pack; | |
4219 | *bigT = *pack; | |
4220 | } | |
4221 | } | |
4222 | ||
40b84e7a | 4223 | #undef OD_ARRAY_SIZE |
d1d7e268 | 4224 | #define OD_ARRAY_SIZE 2 |
40b84e7a | 4225 | |
9babb374 | 4226 | void |
428870ff | 4227 | ztest_dmu_read_write_zcopy(ztest_ds_t *zd, uint64_t id) |
9babb374 | 4228 | { |
428870ff | 4229 | objset_t *os = zd->zd_os; |
40b84e7a | 4230 | ztest_od_t *od; |
9babb374 BB |
4231 | dmu_tx_t *tx; |
4232 | uint64_t i; | |
4233 | int error; | |
40b84e7a | 4234 | int size; |
9babb374 BB |
4235 | uint64_t n, s, txg; |
4236 | bufwad_t *packbuf, *bigbuf; | |
428870ff BB |
4237 | uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize; |
4238 | uint64_t blocksize = ztest_random_blocksize(); | |
4239 | uint64_t chunksize = blocksize; | |
9babb374 BB |
4240 | uint64_t regions = 997; |
4241 | uint64_t stride = 123456789ULL; | |
4242 | uint64_t width = 9; | |
4243 | dmu_buf_t *bonus_db; | |
4244 | arc_buf_t **bigbuf_arcbufs; | |
428870ff | 4245 | dmu_object_info_t doi; |
9babb374 | 4246 | |
d1d7e268 | 4247 | size = sizeof (ztest_od_t) * OD_ARRAY_SIZE; |
40b84e7a BB |
4248 | od = umem_alloc(size, UMEM_NOFAIL); |
4249 | ||
9babb374 BB |
4250 | /* |
4251 | * This test uses two objects, packobj and bigobj, that are always | |
4252 | * updated together (i.e. in the same tx) so that their contents are | |
4253 | * in sync and can be compared. Their contents relate to each other | |
4254 | * in a simple way: packobj is a dense array of 'bufwad' structures, | |
4255 | * while bigobj is a sparse array of the same bufwads. Specifically, | |
4256 | * for any index n, there are three bufwads that should be identical: | |
4257 | * | |
4258 | * packobj, at offset n * sizeof (bufwad_t) | |
4259 | * bigobj, at the head of the nth chunk | |
4260 | * bigobj, at the tail of the nth chunk | |
4261 | * | |
4262 | * The chunk size is set equal to bigobj block size so that | |
4263 | * dmu_assign_arcbuf() can be tested for object updates. | |
4264 | */ | |
4265 | ||
4266 | /* | |
4267 | * Read the directory info. If it's the first time, set things up. | |
4268 | */ | |
50c957f7 NB |
4269 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0, 0); |
4270 | ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, 0, | |
4271 | chunksize); | |
40b84e7a | 4272 | |
9babb374 | 4273 | |
40b84e7a BB |
4274 | if (ztest_object_init(zd, od, size, B_FALSE) != 0) { |
4275 | umem_free(od, size); | |
428870ff | 4276 | return; |
40b84e7a | 4277 | } |
9babb374 | 4278 | |
428870ff BB |
4279 | bigobj = od[0].od_object; |
4280 | packobj = od[1].od_object; | |
4281 | blocksize = od[0].od_blocksize; | |
4282 | chunksize = blocksize; | |
4283 | ASSERT(chunksize == od[1].od_gen); | |
9babb374 | 4284 | |
428870ff BB |
4285 | VERIFY(dmu_object_info(os, bigobj, &doi) == 0); |
4286 | VERIFY(ISP2(doi.doi_data_block_size)); | |
4287 | VERIFY(chunksize == doi.doi_data_block_size); | |
4288 | VERIFY(chunksize >= 2 * sizeof (bufwad_t)); | |
9babb374 BB |
4289 | |
4290 | /* | |
4291 | * Pick a random index and compute the offsets into packobj and bigobj. | |
4292 | */ | |
4293 | n = ztest_random(regions) * stride + ztest_random(width); | |
4294 | s = 1 + ztest_random(width - 1); | |
4295 | ||
4296 | packoff = n * sizeof (bufwad_t); | |
4297 | packsize = s * sizeof (bufwad_t); | |
4298 | ||
428870ff BB |
4299 | bigoff = n * chunksize; |
4300 | bigsize = s * chunksize; | |
9babb374 BB |
4301 | |
4302 | packbuf = umem_zalloc(packsize, UMEM_NOFAIL); | |
4303 | bigbuf = umem_zalloc(bigsize, UMEM_NOFAIL); | |
4304 | ||
428870ff | 4305 | VERIFY3U(0, ==, dmu_bonus_hold(os, bigobj, FTAG, &bonus_db)); |
9babb374 BB |
4306 | |
4307 | bigbuf_arcbufs = umem_zalloc(2 * s * sizeof (arc_buf_t *), UMEM_NOFAIL); | |
4308 | ||
4309 | /* | |
4310 | * Iteration 0 test zcopy for DB_UNCACHED dbufs. | |
4311 | * Iteration 1 test zcopy to already referenced dbufs. | |
4312 | * Iteration 2 test zcopy to dirty dbuf in the same txg. | |
4313 | * Iteration 3 test zcopy to dbuf dirty in previous txg. | |
4314 | * Iteration 4 test zcopy when dbuf is no longer dirty. | |
4315 | * Iteration 5 test zcopy when it can't be done. | |
4316 | * Iteration 6 one more zcopy write. | |
4317 | */ | |
4318 | for (i = 0; i < 7; i++) { | |
4319 | uint64_t j; | |
4320 | uint64_t off; | |
4321 | ||
4322 | /* | |
4323 | * In iteration 5 (i == 5) use arcbufs | |
4324 | * that don't match bigobj blksz to test | |
4325 | * dmu_assign_arcbuf() when it can't directly | |
4326 | * assign an arcbuf to a dbuf. | |
4327 | */ | |
4328 | for (j = 0; j < s; j++) { | |
b9541d6b | 4329 | if (i != 5 || chunksize < (SPA_MINBLOCKSIZE * 2)) { |
9babb374 | 4330 | bigbuf_arcbufs[j] = |
428870ff | 4331 | dmu_request_arcbuf(bonus_db, chunksize); |
9babb374 BB |
4332 | } else { |
4333 | bigbuf_arcbufs[2 * j] = | |
428870ff | 4334 | dmu_request_arcbuf(bonus_db, chunksize / 2); |
9babb374 | 4335 | bigbuf_arcbufs[2 * j + 1] = |
428870ff | 4336 | dmu_request_arcbuf(bonus_db, chunksize / 2); |
9babb374 BB |
4337 | } |
4338 | } | |
4339 | ||
4340 | /* | |
4341 | * Get a tx for the mods to both packobj and bigobj. | |
4342 | */ | |
4343 | tx = dmu_tx_create(os); | |
4344 | ||
428870ff BB |
4345 | dmu_tx_hold_write(tx, packobj, packoff, packsize); |
4346 | dmu_tx_hold_write(tx, bigobj, bigoff, bigsize); | |
9babb374 | 4347 | |
428870ff BB |
4348 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); |
4349 | if (txg == 0) { | |
9babb374 BB |
4350 | umem_free(packbuf, packsize); |
4351 | umem_free(bigbuf, bigsize); | |
4352 | for (j = 0; j < s; j++) { | |
b9541d6b CW |
4353 | if (i != 5 || |
4354 | chunksize < (SPA_MINBLOCKSIZE * 2)) { | |
9babb374 BB |
4355 | dmu_return_arcbuf(bigbuf_arcbufs[j]); |
4356 | } else { | |
4357 | dmu_return_arcbuf( | |
4358 | bigbuf_arcbufs[2 * j]); | |
4359 | dmu_return_arcbuf( | |
4360 | bigbuf_arcbufs[2 * j + 1]); | |
4361 | } | |
4362 | } | |
4363 | umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *)); | |
40b84e7a | 4364 | umem_free(od, size); |
9babb374 BB |
4365 | dmu_buf_rele(bonus_db, FTAG); |
4366 | return; | |
4367 | } | |
4368 | ||
9babb374 BB |
4369 | /* |
4370 | * 50% of the time don't read objects in the 1st iteration to | |
4371 | * test dmu_assign_arcbuf() for the case when there're no | |
4372 | * existing dbufs for the specified offsets. | |
4373 | */ | |
4374 | if (i != 0 || ztest_random(2) != 0) { | |
428870ff | 4375 | error = dmu_read(os, packobj, packoff, |
9babb374 | 4376 | packsize, packbuf, DMU_READ_PREFETCH); |
c99c9001 | 4377 | ASSERT0(error); |
428870ff | 4378 | error = dmu_read(os, bigobj, bigoff, bigsize, |
9babb374 | 4379 | bigbuf, DMU_READ_PREFETCH); |
c99c9001 | 4380 | ASSERT0(error); |
9babb374 BB |
4381 | } |
4382 | compare_and_update_pbbufs(s, packbuf, bigbuf, bigsize, | |
428870ff | 4383 | n, chunksize, txg); |
9babb374 BB |
4384 | |
4385 | /* | |
4386 | * We've verified all the old bufwads, and made new ones. | |
4387 | * Now write them out. | |
4388 | */ | |
428870ff | 4389 | dmu_write(os, packobj, packoff, packsize, packbuf, tx); |
c242c188 | 4390 | if (ztest_opts.zo_verbose >= 7) { |
9babb374 BB |
4391 | (void) printf("writing offset %llx size %llx" |
4392 | " txg %llx\n", | |
4393 | (u_longlong_t)bigoff, | |
4394 | (u_longlong_t)bigsize, | |
4395 | (u_longlong_t)txg); | |
4396 | } | |
428870ff | 4397 | for (off = bigoff, j = 0; j < s; j++, off += chunksize) { |
9babb374 | 4398 | dmu_buf_t *dbt; |
b9541d6b | 4399 | if (i != 5 || chunksize < (SPA_MINBLOCKSIZE * 2)) { |
9babb374 | 4400 | bcopy((caddr_t)bigbuf + (off - bigoff), |
428870ff | 4401 | bigbuf_arcbufs[j]->b_data, chunksize); |
9babb374 BB |
4402 | } else { |
4403 | bcopy((caddr_t)bigbuf + (off - bigoff), | |
4404 | bigbuf_arcbufs[2 * j]->b_data, | |
428870ff | 4405 | chunksize / 2); |
9babb374 | 4406 | bcopy((caddr_t)bigbuf + (off - bigoff) + |
428870ff | 4407 | chunksize / 2, |
9babb374 | 4408 | bigbuf_arcbufs[2 * j + 1]->b_data, |
428870ff | 4409 | chunksize / 2); |
9babb374 BB |
4410 | } |
4411 | ||
4412 | if (i == 1) { | |
428870ff BB |
4413 | VERIFY(dmu_buf_hold(os, bigobj, off, |
4414 | FTAG, &dbt, DMU_READ_NO_PREFETCH) == 0); | |
9babb374 | 4415 | } |
b9541d6b | 4416 | if (i != 5 || chunksize < (SPA_MINBLOCKSIZE * 2)) { |
9babb374 BB |
4417 | dmu_assign_arcbuf(bonus_db, off, |
4418 | bigbuf_arcbufs[j], tx); | |
4419 | } else { | |
4420 | dmu_assign_arcbuf(bonus_db, off, | |
4421 | bigbuf_arcbufs[2 * j], tx); | |
4422 | dmu_assign_arcbuf(bonus_db, | |
428870ff | 4423 | off + chunksize / 2, |
9babb374 BB |
4424 | bigbuf_arcbufs[2 * j + 1], tx); |
4425 | } | |
4426 | if (i == 1) { | |
4427 | dmu_buf_rele(dbt, FTAG); | |
4428 | } | |
4429 | } | |
4430 | dmu_tx_commit(tx); | |
4431 | ||
4432 | /* | |
4433 | * Sanity check the stuff we just wrote. | |
4434 | */ | |
4435 | { | |
4436 | void *packcheck = umem_alloc(packsize, UMEM_NOFAIL); | |
4437 | void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL); | |
4438 | ||
428870ff | 4439 | VERIFY(0 == dmu_read(os, packobj, packoff, |
9babb374 | 4440 | packsize, packcheck, DMU_READ_PREFETCH)); |
428870ff | 4441 | VERIFY(0 == dmu_read(os, bigobj, bigoff, |
9babb374 BB |
4442 | bigsize, bigcheck, DMU_READ_PREFETCH)); |
4443 | ||
4444 | ASSERT(bcmp(packbuf, packcheck, packsize) == 0); | |
4445 | ASSERT(bcmp(bigbuf, bigcheck, bigsize) == 0); | |
4446 | ||
4447 | umem_free(packcheck, packsize); | |
4448 | umem_free(bigcheck, bigsize); | |
4449 | } | |
4450 | if (i == 2) { | |
4451 | txg_wait_open(dmu_objset_pool(os), 0); | |
4452 | } else if (i == 3) { | |
4453 | txg_wait_synced(dmu_objset_pool(os), 0); | |
4454 | } | |
4455 | } | |
4456 | ||
4457 | dmu_buf_rele(bonus_db, FTAG); | |
4458 | umem_free(packbuf, packsize); | |
4459 | umem_free(bigbuf, bigsize); | |
4460 | umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *)); | |
40b84e7a | 4461 | umem_free(od, size); |
9babb374 BB |
4462 | } |
4463 | ||
428870ff | 4464 | /* ARGSUSED */ |
34dc7c2f | 4465 | void |
428870ff | 4466 | ztest_dmu_write_parallel(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4467 | { |
40b84e7a BB |
4468 | ztest_od_t *od; |
4469 | ||
d1d7e268 | 4470 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
428870ff BB |
4471 | uint64_t offset = (1ULL << (ztest_random(20) + 43)) + |
4472 | (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
34dc7c2f BB |
4473 | |
4474 | /* | |
428870ff BB |
4475 | * Have multiple threads write to large offsets in an object |
4476 | * to verify that parallel writes to an object -- even to the | |
4477 | * same blocks within the object -- doesn't cause any trouble. | |
34dc7c2f | 4478 | */ |
50c957f7 | 4479 | ztest_od_init(od, ID_PARALLEL, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0); |
9babb374 | 4480 | |
40b84e7a | 4481 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) |
34dc7c2f | 4482 | return; |
34dc7c2f | 4483 | |
428870ff | 4484 | while (ztest_random(10) != 0) |
40b84e7a BB |
4485 | ztest_io(zd, od->od_object, offset); |
4486 | ||
d1d7e268 | 4487 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 4488 | } |
34dc7c2f | 4489 | |
428870ff BB |
4490 | void |
4491 | ztest_dmu_prealloc(ztest_ds_t *zd, uint64_t id) | |
4492 | { | |
40b84e7a | 4493 | ztest_od_t *od; |
428870ff BB |
4494 | uint64_t offset = (1ULL << (ztest_random(4) + SPA_MAXBLOCKSHIFT)) + |
4495 | (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
4496 | uint64_t count = ztest_random(20) + 1; | |
4497 | uint64_t blocksize = ztest_random_blocksize(); | |
4498 | void *data; | |
34dc7c2f | 4499 | |
d1d7e268 | 4500 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
34dc7c2f | 4501 | |
50c957f7 | 4502 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0, 0); |
40b84e7a | 4503 | |
d1d7e268 MK |
4504 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
4505 | !ztest_random(2)) != 0) { | |
4506 | umem_free(od, sizeof (ztest_od_t)); | |
34dc7c2f | 4507 | return; |
40b84e7a | 4508 | } |
34dc7c2f | 4509 | |
40b84e7a | 4510 | if (ztest_truncate(zd, od->od_object, offset, count * blocksize) != 0) { |
d1d7e268 | 4511 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f | 4512 | return; |
40b84e7a | 4513 | } |
34dc7c2f | 4514 | |
40b84e7a | 4515 | ztest_prealloc(zd, od->od_object, offset, count * blocksize); |
34dc7c2f | 4516 | |
428870ff | 4517 | data = umem_zalloc(blocksize, UMEM_NOFAIL); |
34dc7c2f | 4518 | |
428870ff BB |
4519 | while (ztest_random(count) != 0) { |
4520 | uint64_t randoff = offset + (ztest_random(count) * blocksize); | |
40b84e7a | 4521 | if (ztest_write(zd, od->od_object, randoff, blocksize, |
428870ff BB |
4522 | data) != 0) |
4523 | break; | |
4524 | while (ztest_random(4) != 0) | |
40b84e7a | 4525 | ztest_io(zd, od->od_object, randoff); |
9babb374 | 4526 | } |
34dc7c2f | 4527 | |
428870ff | 4528 | umem_free(data, blocksize); |
d1d7e268 | 4529 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f BB |
4530 | } |
4531 | ||
4532 | /* | |
4533 | * Verify that zap_{create,destroy,add,remove,update} work as expected. | |
4534 | */ | |
4535 | #define ZTEST_ZAP_MIN_INTS 1 | |
4536 | #define ZTEST_ZAP_MAX_INTS 4 | |
4537 | #define ZTEST_ZAP_MAX_PROPS 1000 | |
4538 | ||
4539 | void | |
428870ff | 4540 | ztest_zap(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4541 | { |
428870ff | 4542 | objset_t *os = zd->zd_os; |
40b84e7a | 4543 | ztest_od_t *od; |
34dc7c2f BB |
4544 | uint64_t object; |
4545 | uint64_t txg, last_txg; | |
4546 | uint64_t value[ZTEST_ZAP_MAX_INTS]; | |
4547 | uint64_t zl_ints, zl_intsize, prop; | |
4548 | int i, ints; | |
4549 | dmu_tx_t *tx; | |
4550 | char propname[100], txgname[100]; | |
4551 | int error; | |
34dc7c2f BB |
4552 | char *hc[2] = { "s.acl.h", ".s.open.h.hyLZlg" }; |
4553 | ||
d1d7e268 | 4554 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 4555 | ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0, 0); |
34dc7c2f | 4556 | |
40b84e7a | 4557 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
02730c33 | 4558 | !ztest_random(2)) != 0) |
40b84e7a | 4559 | goto out; |
34dc7c2f | 4560 | |
40b84e7a | 4561 | object = od->od_object; |
34dc7c2f | 4562 | |
428870ff BB |
4563 | /* |
4564 | * Generate a known hash collision, and verify that | |
4565 | * we can lookup and remove both entries. | |
4566 | */ | |
4567 | tx = dmu_tx_create(os); | |
4568 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); | |
4569 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4570 | if (txg == 0) | |
40b84e7a | 4571 | goto out; |
428870ff BB |
4572 | for (i = 0; i < 2; i++) { |
4573 | value[i] = i; | |
4574 | VERIFY3U(0, ==, zap_add(os, object, hc[i], sizeof (uint64_t), | |
4575 | 1, &value[i], tx)); | |
4576 | } | |
4577 | for (i = 0; i < 2; i++) { | |
4578 | VERIFY3U(EEXIST, ==, zap_add(os, object, hc[i], | |
4579 | sizeof (uint64_t), 1, &value[i], tx)); | |
4580 | VERIFY3U(0, ==, | |
4581 | zap_length(os, object, hc[i], &zl_intsize, &zl_ints)); | |
4582 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
4583 | ASSERT3U(zl_ints, ==, 1); | |
4584 | } | |
4585 | for (i = 0; i < 2; i++) { | |
4586 | VERIFY3U(0, ==, zap_remove(os, object, hc[i], tx)); | |
34dc7c2f | 4587 | } |
428870ff | 4588 | dmu_tx_commit(tx); |
34dc7c2f | 4589 | |
428870ff BB |
4590 | /* |
4591 | * Generate a buch of random entries. | |
4592 | */ | |
34dc7c2f BB |
4593 | ints = MAX(ZTEST_ZAP_MIN_INTS, object % ZTEST_ZAP_MAX_INTS); |
4594 | ||
4595 | prop = ztest_random(ZTEST_ZAP_MAX_PROPS); | |
4596 | (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop); | |
4597 | (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop); | |
4598 | bzero(value, sizeof (value)); | |
4599 | last_txg = 0; | |
4600 | ||
4601 | /* | |
4602 | * If these zap entries already exist, validate their contents. | |
4603 | */ | |
4604 | error = zap_length(os, object, txgname, &zl_intsize, &zl_ints); | |
4605 | if (error == 0) { | |
4606 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
4607 | ASSERT3U(zl_ints, ==, 1); | |
4608 | ||
4609 | VERIFY(zap_lookup(os, object, txgname, zl_intsize, | |
4610 | zl_ints, &last_txg) == 0); | |
4611 | ||
4612 | VERIFY(zap_length(os, object, propname, &zl_intsize, | |
4613 | &zl_ints) == 0); | |
4614 | ||
4615 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
4616 | ASSERT3U(zl_ints, ==, ints); | |
4617 | ||
4618 | VERIFY(zap_lookup(os, object, propname, zl_intsize, | |
4619 | zl_ints, value) == 0); | |
4620 | ||
4621 | for (i = 0; i < ints; i++) { | |
4622 | ASSERT3U(value[i], ==, last_txg + object + i); | |
4623 | } | |
4624 | } else { | |
4625 | ASSERT3U(error, ==, ENOENT); | |
4626 | } | |
4627 | ||
4628 | /* | |
4629 | * Atomically update two entries in our zap object. | |
4630 | * The first is named txg_%llu, and contains the txg | |
4631 | * in which the property was last updated. The second | |
4632 | * is named prop_%llu, and the nth element of its value | |
4633 | * should be txg + object + n. | |
4634 | */ | |
4635 | tx = dmu_tx_create(os); | |
428870ff BB |
4636 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
4637 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4638 | if (txg == 0) | |
40b84e7a | 4639 | goto out; |
34dc7c2f BB |
4640 | |
4641 | if (last_txg > txg) | |
4642 | fatal(0, "zap future leak: old %llu new %llu", last_txg, txg); | |
4643 | ||
4644 | for (i = 0; i < ints; i++) | |
4645 | value[i] = txg + object + i; | |
4646 | ||
428870ff BB |
4647 | VERIFY3U(0, ==, zap_update(os, object, txgname, sizeof (uint64_t), |
4648 | 1, &txg, tx)); | |
4649 | VERIFY3U(0, ==, zap_update(os, object, propname, sizeof (uint64_t), | |
4650 | ints, value, tx)); | |
34dc7c2f BB |
4651 | |
4652 | dmu_tx_commit(tx); | |
4653 | ||
4654 | /* | |
4655 | * Remove a random pair of entries. | |
4656 | */ | |
4657 | prop = ztest_random(ZTEST_ZAP_MAX_PROPS); | |
4658 | (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop); | |
4659 | (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop); | |
4660 | ||
4661 | error = zap_length(os, object, txgname, &zl_intsize, &zl_ints); | |
4662 | ||
4663 | if (error == ENOENT) | |
40b84e7a | 4664 | goto out; |
34dc7c2f | 4665 | |
c99c9001 | 4666 | ASSERT0(error); |
34dc7c2f BB |
4667 | |
4668 | tx = dmu_tx_create(os); | |
428870ff BB |
4669 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
4670 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4671 | if (txg == 0) | |
40b84e7a | 4672 | goto out; |
428870ff BB |
4673 | VERIFY3U(0, ==, zap_remove(os, object, txgname, tx)); |
4674 | VERIFY3U(0, ==, zap_remove(os, object, propname, tx)); | |
4675 | dmu_tx_commit(tx); | |
40b84e7a | 4676 | out: |
d1d7e268 | 4677 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 4678 | } |
34dc7c2f | 4679 | |
428870ff BB |
4680 | /* |
4681 | * Testcase to test the upgrading of a microzap to fatzap. | |
4682 | */ | |
4683 | void | |
4684 | ztest_fzap(ztest_ds_t *zd, uint64_t id) | |
4685 | { | |
4686 | objset_t *os = zd->zd_os; | |
40b84e7a | 4687 | ztest_od_t *od; |
428870ff | 4688 | uint64_t object, txg; |
d6320ddb | 4689 | int i; |
34dc7c2f | 4690 | |
d1d7e268 | 4691 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 4692 | ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0, 0); |
428870ff | 4693 | |
40b84e7a | 4694 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
02730c33 | 4695 | !ztest_random(2)) != 0) |
40b84e7a BB |
4696 | goto out; |
4697 | object = od->od_object; | |
34dc7c2f BB |
4698 | |
4699 | /* | |
428870ff BB |
4700 | * Add entries to this ZAP and make sure it spills over |
4701 | * and gets upgraded to a fatzap. Also, since we are adding | |
4702 | * 2050 entries we should see ptrtbl growth and leaf-block split. | |
34dc7c2f | 4703 | */ |
d6320ddb | 4704 | for (i = 0; i < 2050; i++) { |
eca7b760 | 4705 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
4706 | uint64_t value = i; |
4707 | dmu_tx_t *tx; | |
4708 | int error; | |
34dc7c2f | 4709 | |
428870ff | 4710 | (void) snprintf(name, sizeof (name), "fzap-%llu-%llu", |
b8864a23 | 4711 | (u_longlong_t)id, (u_longlong_t)value); |
428870ff BB |
4712 | |
4713 | tx = dmu_tx_create(os); | |
4714 | dmu_tx_hold_zap(tx, object, B_TRUE, name); | |
4715 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4716 | if (txg == 0) | |
40b84e7a | 4717 | goto out; |
428870ff BB |
4718 | error = zap_add(os, object, name, sizeof (uint64_t), 1, |
4719 | &value, tx); | |
4720 | ASSERT(error == 0 || error == EEXIST); | |
4721 | dmu_tx_commit(tx); | |
34dc7c2f | 4722 | } |
40b84e7a | 4723 | out: |
d1d7e268 | 4724 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f BB |
4725 | } |
4726 | ||
428870ff | 4727 | /* ARGSUSED */ |
34dc7c2f | 4728 | void |
428870ff | 4729 | ztest_zap_parallel(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4730 | { |
428870ff | 4731 | objset_t *os = zd->zd_os; |
40b84e7a | 4732 | ztest_od_t *od; |
34dc7c2f BB |
4733 | uint64_t txg, object, count, wsize, wc, zl_wsize, zl_wc; |
4734 | dmu_tx_t *tx; | |
4735 | int i, namelen, error; | |
428870ff | 4736 | int micro = ztest_random(2); |
34dc7c2f BB |
4737 | char name[20], string_value[20]; |
4738 | void *data; | |
4739 | ||
d1d7e268 | 4740 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 4741 | ztest_od_init(od, ID_PARALLEL, FTAG, micro, DMU_OT_ZAP_OTHER, 0, 0, 0); |
428870ff | 4742 | |
40b84e7a | 4743 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
d1d7e268 | 4744 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 4745 | return; |
40b84e7a | 4746 | } |
428870ff | 4747 | |
40b84e7a | 4748 | object = od->od_object; |
428870ff | 4749 | |
34dc7c2f BB |
4750 | /* |
4751 | * Generate a random name of the form 'xxx.....' where each | |
4752 | * x is a random printable character and the dots are dots. | |
4753 | * There are 94 such characters, and the name length goes from | |
4754 | * 6 to 20, so there are 94^3 * 15 = 12,458,760 possible names. | |
4755 | */ | |
4756 | namelen = ztest_random(sizeof (name) - 5) + 5 + 1; | |
4757 | ||
4758 | for (i = 0; i < 3; i++) | |
4759 | name[i] = '!' + ztest_random('~' - '!' + 1); | |
4760 | for (; i < namelen - 1; i++) | |
4761 | name[i] = '.'; | |
4762 | name[i] = '\0'; | |
4763 | ||
428870ff | 4764 | if ((namelen & 1) || micro) { |
34dc7c2f BB |
4765 | wsize = sizeof (txg); |
4766 | wc = 1; | |
4767 | data = &txg; | |
4768 | } else { | |
4769 | wsize = 1; | |
4770 | wc = namelen; | |
4771 | data = string_value; | |
4772 | } | |
4773 | ||
4774 | count = -1ULL; | |
13fe0198 | 4775 | VERIFY0(zap_count(os, object, &count)); |
34dc7c2f BB |
4776 | ASSERT(count != -1ULL); |
4777 | ||
4778 | /* | |
4779 | * Select an operation: length, lookup, add, update, remove. | |
4780 | */ | |
4781 | i = ztest_random(5); | |
4782 | ||
4783 | if (i >= 2) { | |
4784 | tx = dmu_tx_create(os); | |
428870ff BB |
4785 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
4786 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
df053d67 GN |
4787 | if (txg == 0) { |
4788 | umem_free(od, sizeof (ztest_od_t)); | |
34dc7c2f | 4789 | return; |
df053d67 | 4790 | } |
34dc7c2f BB |
4791 | bcopy(name, string_value, namelen); |
4792 | } else { | |
4793 | tx = NULL; | |
4794 | txg = 0; | |
4795 | bzero(string_value, namelen); | |
4796 | } | |
4797 | ||
4798 | switch (i) { | |
4799 | ||
4800 | case 0: | |
4801 | error = zap_length(os, object, name, &zl_wsize, &zl_wc); | |
4802 | if (error == 0) { | |
4803 | ASSERT3U(wsize, ==, zl_wsize); | |
4804 | ASSERT3U(wc, ==, zl_wc); | |
4805 | } else { | |
4806 | ASSERT3U(error, ==, ENOENT); | |
4807 | } | |
4808 | break; | |
4809 | ||
4810 | case 1: | |
4811 | error = zap_lookup(os, object, name, wsize, wc, data); | |
4812 | if (error == 0) { | |
4813 | if (data == string_value && | |
4814 | bcmp(name, data, namelen) != 0) | |
4815 | fatal(0, "name '%s' != val '%s' len %d", | |
4816 | name, data, namelen); | |
4817 | } else { | |
4818 | ASSERT3U(error, ==, ENOENT); | |
4819 | } | |
4820 | break; | |
4821 | ||
4822 | case 2: | |
4823 | error = zap_add(os, object, name, wsize, wc, data, tx); | |
4824 | ASSERT(error == 0 || error == EEXIST); | |
4825 | break; | |
4826 | ||
4827 | case 3: | |
4828 | VERIFY(zap_update(os, object, name, wsize, wc, data, tx) == 0); | |
4829 | break; | |
4830 | ||
4831 | case 4: | |
4832 | error = zap_remove(os, object, name, tx); | |
4833 | ASSERT(error == 0 || error == ENOENT); | |
4834 | break; | |
4835 | } | |
4836 | ||
4837 | if (tx != NULL) | |
4838 | dmu_tx_commit(tx); | |
40b84e7a | 4839 | |
d1d7e268 | 4840 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f BB |
4841 | } |
4842 | ||
428870ff BB |
4843 | /* |
4844 | * Commit callback data. | |
4845 | */ | |
4846 | typedef struct ztest_cb_data { | |
4847 | list_node_t zcd_node; | |
4848 | uint64_t zcd_txg; | |
4849 | int zcd_expected_err; | |
4850 | boolean_t zcd_added; | |
4851 | boolean_t zcd_called; | |
4852 | spa_t *zcd_spa; | |
4853 | } ztest_cb_data_t; | |
4854 | ||
4855 | /* This is the actual commit callback function */ | |
4856 | static void | |
4857 | ztest_commit_callback(void *arg, int error) | |
4858 | { | |
4859 | ztest_cb_data_t *data = arg; | |
4860 | uint64_t synced_txg; | |
4861 | ||
4862 | VERIFY(data != NULL); | |
4863 | VERIFY3S(data->zcd_expected_err, ==, error); | |
4864 | VERIFY(!data->zcd_called); | |
4865 | ||
4866 | synced_txg = spa_last_synced_txg(data->zcd_spa); | |
4867 | if (data->zcd_txg > synced_txg) | |
4868 | fatal(0, "commit callback of txg %" PRIu64 " called prematurely" | |
4869 | ", last synced txg = %" PRIu64 "\n", data->zcd_txg, | |
4870 | synced_txg); | |
4871 | ||
4872 | data->zcd_called = B_TRUE; | |
4873 | ||
4874 | if (error == ECANCELED) { | |
c99c9001 | 4875 | ASSERT0(data->zcd_txg); |
428870ff BB |
4876 | ASSERT(!data->zcd_added); |
4877 | ||
4878 | /* | |
4879 | * The private callback data should be destroyed here, but | |
4880 | * since we are going to check the zcd_called field after | |
4881 | * dmu_tx_abort(), we will destroy it there. | |
4882 | */ | |
4883 | return; | |
4884 | } | |
4885 | ||
090ff092 | 4886 | ASSERT(data->zcd_added); |
428870ff BB |
4887 | ASSERT3U(data->zcd_txg, !=, 0); |
4888 | ||
1e33ac1e | 4889 | (void) mutex_enter(&zcl.zcl_callbacks_lock); |
090ff092 RC |
4890 | |
4891 | /* See if this cb was called more quickly */ | |
4892 | if ((synced_txg - data->zcd_txg) < zc_min_txg_delay) | |
4893 | zc_min_txg_delay = synced_txg - data->zcd_txg; | |
4894 | ||
4895 | /* Remove our callback from the list */ | |
428870ff | 4896 | list_remove(&zcl.zcl_callbacks, data); |
090ff092 | 4897 | |
1e33ac1e | 4898 | (void) mutex_exit(&zcl.zcl_callbacks_lock); |
428870ff | 4899 | |
428870ff BB |
4900 | umem_free(data, sizeof (ztest_cb_data_t)); |
4901 | } | |
4902 | ||
4903 | /* Allocate and initialize callback data structure */ | |
4904 | static ztest_cb_data_t * | |
4905 | ztest_create_cb_data(objset_t *os, uint64_t txg) | |
4906 | { | |
4907 | ztest_cb_data_t *cb_data; | |
4908 | ||
4909 | cb_data = umem_zalloc(sizeof (ztest_cb_data_t), UMEM_NOFAIL); | |
4910 | ||
4911 | cb_data->zcd_txg = txg; | |
4912 | cb_data->zcd_spa = dmu_objset_spa(os); | |
1e33ac1e | 4913 | list_link_init(&cb_data->zcd_node); |
428870ff BB |
4914 | |
4915 | return (cb_data); | |
4916 | } | |
4917 | ||
428870ff BB |
4918 | /* |
4919 | * Commit callback test. | |
4920 | */ | |
34dc7c2f | 4921 | void |
428870ff BB |
4922 | ztest_dmu_commit_callbacks(ztest_ds_t *zd, uint64_t id) |
4923 | { | |
4924 | objset_t *os = zd->zd_os; | |
40b84e7a | 4925 | ztest_od_t *od; |
428870ff BB |
4926 | dmu_tx_t *tx; |
4927 | ztest_cb_data_t *cb_data[3], *tmp_cb; | |
4928 | uint64_t old_txg, txg; | |
090ff092 | 4929 | int i, error = 0; |
428870ff | 4930 | |
d1d7e268 | 4931 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 4932 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0); |
428870ff | 4933 | |
40b84e7a | 4934 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
d1d7e268 | 4935 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 4936 | return; |
40b84e7a | 4937 | } |
428870ff BB |
4938 | |
4939 | tx = dmu_tx_create(os); | |
4940 | ||
4941 | cb_data[0] = ztest_create_cb_data(os, 0); | |
4942 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[0]); | |
4943 | ||
40b84e7a | 4944 | dmu_tx_hold_write(tx, od->od_object, 0, sizeof (uint64_t)); |
428870ff BB |
4945 | |
4946 | /* Every once in a while, abort the transaction on purpose */ | |
4947 | if (ztest_random(100) == 0) | |
4948 | error = -1; | |
4949 | ||
4950 | if (!error) | |
4951 | error = dmu_tx_assign(tx, TXG_NOWAIT); | |
4952 | ||
4953 | txg = error ? 0 : dmu_tx_get_txg(tx); | |
4954 | ||
4955 | cb_data[0]->zcd_txg = txg; | |
4956 | cb_data[1] = ztest_create_cb_data(os, txg); | |
4957 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[1]); | |
4958 | ||
4959 | if (error) { | |
4960 | /* | |
4961 | * It's not a strict requirement to call the registered | |
4962 | * callbacks from inside dmu_tx_abort(), but that's what | |
4963 | * it's supposed to happen in the current implementation | |
4964 | * so we will check for that. | |
4965 | */ | |
4966 | for (i = 0; i < 2; i++) { | |
4967 | cb_data[i]->zcd_expected_err = ECANCELED; | |
4968 | VERIFY(!cb_data[i]->zcd_called); | |
4969 | } | |
4970 | ||
4971 | dmu_tx_abort(tx); | |
4972 | ||
4973 | for (i = 0; i < 2; i++) { | |
4974 | VERIFY(cb_data[i]->zcd_called); | |
4975 | umem_free(cb_data[i], sizeof (ztest_cb_data_t)); | |
4976 | } | |
4977 | ||
d1d7e268 | 4978 | umem_free(od, sizeof (ztest_od_t)); |
428870ff BB |
4979 | return; |
4980 | } | |
4981 | ||
4982 | cb_data[2] = ztest_create_cb_data(os, txg); | |
4983 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[2]); | |
4984 | ||
4985 | /* | |
4986 | * Read existing data to make sure there isn't a future leak. | |
4987 | */ | |
40b84e7a | 4988 | VERIFY(0 == dmu_read(os, od->od_object, 0, sizeof (uint64_t), |
428870ff BB |
4989 | &old_txg, DMU_READ_PREFETCH)); |
4990 | ||
4991 | if (old_txg > txg) | |
4992 | fatal(0, "future leak: got %" PRIu64 ", open txg is %" PRIu64, | |
4993 | old_txg, txg); | |
4994 | ||
40b84e7a | 4995 | dmu_write(os, od->od_object, 0, sizeof (uint64_t), &txg, tx); |
428870ff | 4996 | |
1e33ac1e | 4997 | (void) mutex_enter(&zcl.zcl_callbacks_lock); |
428870ff BB |
4998 | |
4999 | /* | |
5000 | * Since commit callbacks don't have any ordering requirement and since | |
5001 | * it is theoretically possible for a commit callback to be called | |
5002 | * after an arbitrary amount of time has elapsed since its txg has been | |
5003 | * synced, it is difficult to reliably determine whether a commit | |
5004 | * callback hasn't been called due to high load or due to a flawed | |
5005 | * implementation. | |
5006 | * | |
5007 | * In practice, we will assume that if after a certain number of txgs a | |
5008 | * commit callback hasn't been called, then most likely there's an | |
5009 | * implementation bug.. | |
5010 | */ | |
5011 | tmp_cb = list_head(&zcl.zcl_callbacks); | |
5012 | if (tmp_cb != NULL && | |
090ff092 | 5013 | tmp_cb->zcd_txg + ZTEST_COMMIT_CB_THRESH < txg) { |
428870ff BB |
5014 | fatal(0, "Commit callback threshold exceeded, oldest txg: %" |
5015 | PRIu64 ", open txg: %" PRIu64 "\n", tmp_cb->zcd_txg, txg); | |
5016 | } | |
5017 | ||
5018 | /* | |
5019 | * Let's find the place to insert our callbacks. | |
5020 | * | |
5021 | * Even though the list is ordered by txg, it is possible for the | |
5022 | * insertion point to not be the end because our txg may already be | |
5023 | * quiescing at this point and other callbacks in the open txg | |
5024 | * (from other objsets) may have sneaked in. | |
5025 | */ | |
5026 | tmp_cb = list_tail(&zcl.zcl_callbacks); | |
5027 | while (tmp_cb != NULL && tmp_cb->zcd_txg > txg) | |
5028 | tmp_cb = list_prev(&zcl.zcl_callbacks, tmp_cb); | |
5029 | ||
5030 | /* Add the 3 callbacks to the list */ | |
5031 | for (i = 0; i < 3; i++) { | |
5032 | if (tmp_cb == NULL) | |
5033 | list_insert_head(&zcl.zcl_callbacks, cb_data[i]); | |
5034 | else | |
5035 | list_insert_after(&zcl.zcl_callbacks, tmp_cb, | |
5036 | cb_data[i]); | |
5037 | ||
5038 | cb_data[i]->zcd_added = B_TRUE; | |
5039 | VERIFY(!cb_data[i]->zcd_called); | |
5040 | ||
5041 | tmp_cb = cb_data[i]; | |
5042 | } | |
5043 | ||
090ff092 RC |
5044 | zc_cb_counter += 3; |
5045 | ||
1e33ac1e | 5046 | (void) mutex_exit(&zcl.zcl_callbacks_lock); |
428870ff BB |
5047 | |
5048 | dmu_tx_commit(tx); | |
40b84e7a | 5049 | |
d1d7e268 | 5050 | umem_free(od, sizeof (ztest_od_t)); |
428870ff BB |
5051 | } |
5052 | ||
50c957f7 NB |
5053 | /* |
5054 | * Visit each object in the dataset. Verify that its properties | |
5055 | * are consistent what was stored in the block tag when it was created, | |
5056 | * and that its unused bonus buffer space has not been overwritten. | |
5057 | */ | |
817b1b6e | 5058 | /* ARGSUSED */ |
50c957f7 NB |
5059 | void |
5060 | ztest_verify_dnode_bt(ztest_ds_t *zd, uint64_t id) | |
5061 | { | |
5062 | objset_t *os = zd->zd_os; | |
5063 | uint64_t obj; | |
5064 | int err = 0; | |
5065 | ||
5066 | for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) { | |
5067 | ztest_block_tag_t *bt = NULL; | |
5068 | dmu_object_info_t doi; | |
5069 | dmu_buf_t *db; | |
5070 | ||
5071 | if (dmu_bonus_hold(os, obj, FTAG, &db) != 0) | |
5072 | continue; | |
5073 | ||
5074 | dmu_object_info_from_db(db, &doi); | |
5075 | if (doi.doi_bonus_size >= sizeof (*bt)) | |
5076 | bt = ztest_bt_bonus(db); | |
5077 | ||
5078 | if (bt && bt->bt_magic == BT_MAGIC) { | |
5079 | ztest_bt_verify(bt, os, obj, doi.doi_dnodesize, | |
5080 | bt->bt_offset, bt->bt_gen, bt->bt_txg, | |
5081 | bt->bt_crtxg); | |
5082 | ztest_verify_unused_bonus(db, bt, obj, os, bt->bt_gen); | |
5083 | } | |
5084 | ||
5085 | dmu_buf_rele(db, FTAG); | |
5086 | } | |
5087 | } | |
5088 | ||
428870ff BB |
5089 | /* ARGSUSED */ |
5090 | void | |
5091 | ztest_dsl_prop_get_set(ztest_ds_t *zd, uint64_t id) | |
5092 | { | |
5093 | zfs_prop_t proplist[] = { | |
5094 | ZFS_PROP_CHECKSUM, | |
5095 | ZFS_PROP_COMPRESSION, | |
5096 | ZFS_PROP_COPIES, | |
5097 | ZFS_PROP_DEDUP | |
5098 | }; | |
d6320ddb | 5099 | int p; |
428870ff | 5100 | |
7809eb8b | 5101 | (void) rw_rdlock(&ztest_name_lock); |
428870ff | 5102 | |
d6320ddb | 5103 | for (p = 0; p < sizeof (proplist) / sizeof (proplist[0]); p++) |
428870ff BB |
5104 | (void) ztest_dsl_prop_set_uint64(zd->zd_name, proplist[p], |
5105 | ztest_random_dsl_prop(proplist[p]), (int)ztest_random(2)); | |
5106 | ||
76d52067 BB |
5107 | VERIFY0(ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_RECORDSIZE, |
5108 | ztest_random_blocksize(), (int)ztest_random(2))); | |
5109 | ||
7809eb8b | 5110 | (void) rw_unlock(&ztest_name_lock); |
428870ff BB |
5111 | } |
5112 | ||
5113 | /* ARGSUSED */ | |
5114 | void | |
5115 | ztest_spa_prop_get_set(ztest_ds_t *zd, uint64_t id) | |
5116 | { | |
428870ff BB |
5117 | nvlist_t *props = NULL; |
5118 | ||
7809eb8b | 5119 | (void) rw_rdlock(&ztest_name_lock); |
428870ff | 5120 | |
c242c188 | 5121 | (void) ztest_spa_prop_set_uint64(ZPOOL_PROP_DEDUPDITTO, |
428870ff BB |
5122 | ZIO_DEDUPDITTO_MIN + ztest_random(ZIO_DEDUPDITTO_MIN)); |
5123 | ||
c99c9001 | 5124 | VERIFY0(spa_prop_get(ztest_spa, &props)); |
428870ff | 5125 | |
c242c188 | 5126 | if (ztest_opts.zo_verbose >= 6) |
428870ff BB |
5127 | dump_nvlist(props, 4); |
5128 | ||
5129 | nvlist_free(props); | |
5130 | ||
7809eb8b | 5131 | (void) rw_unlock(&ztest_name_lock); |
428870ff BB |
5132 | } |
5133 | ||
13fe0198 MA |
5134 | static int |
5135 | user_release_one(const char *snapname, const char *holdname) | |
5136 | { | |
5137 | nvlist_t *snaps, *holds; | |
5138 | int error; | |
5139 | ||
5140 | snaps = fnvlist_alloc(); | |
5141 | holds = fnvlist_alloc(); | |
5142 | fnvlist_add_boolean(holds, holdname); | |
5143 | fnvlist_add_nvlist(snaps, snapname, holds); | |
5144 | fnvlist_free(holds); | |
5145 | error = dsl_dataset_user_release(snaps, NULL); | |
5146 | fnvlist_free(snaps); | |
5147 | return (error); | |
5148 | } | |
5149 | ||
428870ff BB |
5150 | /* |
5151 | * Test snapshot hold/release and deferred destroy. | |
5152 | */ | |
5153 | void | |
5154 | ztest_dmu_snapshot_hold(ztest_ds_t *zd, uint64_t id) | |
34dc7c2f | 5155 | { |
34dc7c2f | 5156 | int error; |
428870ff BB |
5157 | objset_t *os = zd->zd_os; |
5158 | objset_t *origin; | |
5159 | char snapname[100]; | |
5160 | char fullname[100]; | |
5161 | char clonename[100]; | |
5162 | char tag[100]; | |
eca7b760 | 5163 | char osname[ZFS_MAX_DATASET_NAME_LEN]; |
13fe0198 | 5164 | nvlist_t *holds; |
34dc7c2f | 5165 | |
7809eb8b | 5166 | (void) rw_rdlock(&ztest_name_lock); |
34dc7c2f BB |
5167 | |
5168 | dmu_objset_name(os, osname); | |
5169 | ||
d1d7e268 MK |
5170 | (void) snprintf(snapname, sizeof (snapname), "sh1_%llu", |
5171 | (u_longlong_t)id); | |
13fe0198 MA |
5172 | (void) snprintf(fullname, sizeof (fullname), "%s@%s", osname, snapname); |
5173 | (void) snprintf(clonename, sizeof (clonename), | |
d1d7e268 MK |
5174 | "%s/ch1_%llu", osname, (u_longlong_t)id); |
5175 | (void) snprintf(tag, sizeof (tag), "tag_%llu", (u_longlong_t)id); | |
428870ff BB |
5176 | |
5177 | /* | |
5178 | * Clean up from any previous run. | |
5179 | */ | |
13fe0198 MA |
5180 | error = dsl_destroy_head(clonename); |
5181 | if (error != ENOENT) | |
5182 | ASSERT0(error); | |
5183 | error = user_release_one(fullname, tag); | |
5184 | if (error != ESRCH && error != ENOENT) | |
5185 | ASSERT0(error); | |
5186 | error = dsl_destroy_snapshot(fullname, B_FALSE); | |
5187 | if (error != ENOENT) | |
5188 | ASSERT0(error); | |
428870ff BB |
5189 | |
5190 | /* | |
5191 | * Create snapshot, clone it, mark snap for deferred destroy, | |
5192 | * destroy clone, verify snap was also destroyed. | |
5193 | */ | |
6f1ffb06 | 5194 | error = dmu_objset_snapshot_one(osname, snapname); |
428870ff BB |
5195 | if (error) { |
5196 | if (error == ENOSPC) { | |
5197 | ztest_record_enospc("dmu_objset_snapshot"); | |
5198 | goto out; | |
34dc7c2f | 5199 | } |
428870ff BB |
5200 | fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error); |
5201 | } | |
34dc7c2f | 5202 | |
13fe0198 | 5203 | error = dmu_objset_clone(clonename, fullname); |
428870ff | 5204 | if (error) { |
34dc7c2f | 5205 | if (error == ENOSPC) { |
428870ff BB |
5206 | ztest_record_enospc("dmu_objset_clone"); |
5207 | goto out; | |
34dc7c2f | 5208 | } |
428870ff BB |
5209 | fatal(0, "dmu_objset_clone(%s) = %d", clonename, error); |
5210 | } | |
34dc7c2f | 5211 | |
13fe0198 | 5212 | error = dsl_destroy_snapshot(fullname, B_TRUE); |
428870ff | 5213 | if (error) { |
13fe0198 | 5214 | fatal(0, "dsl_destroy_snapshot(%s, B_TRUE) = %d", |
428870ff BB |
5215 | fullname, error); |
5216 | } | |
34dc7c2f | 5217 | |
13fe0198 | 5218 | error = dsl_destroy_head(clonename); |
428870ff | 5219 | if (error) |
13fe0198 | 5220 | fatal(0, "dsl_destroy_head(%s) = %d", clonename, error); |
34dc7c2f | 5221 | |
428870ff BB |
5222 | error = dmu_objset_hold(fullname, FTAG, &origin); |
5223 | if (error != ENOENT) | |
5224 | fatal(0, "dmu_objset_hold(%s) = %d", fullname, error); | |
34dc7c2f | 5225 | |
428870ff BB |
5226 | /* |
5227 | * Create snapshot, add temporary hold, verify that we can't | |
5228 | * destroy a held snapshot, mark for deferred destroy, | |
5229 | * release hold, verify snapshot was destroyed. | |
5230 | */ | |
6f1ffb06 | 5231 | error = dmu_objset_snapshot_one(osname, snapname); |
428870ff BB |
5232 | if (error) { |
5233 | if (error == ENOSPC) { | |
5234 | ztest_record_enospc("dmu_objset_snapshot"); | |
5235 | goto out; | |
34dc7c2f | 5236 | } |
428870ff BB |
5237 | fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error); |
5238 | } | |
5239 | ||
13fe0198 MA |
5240 | holds = fnvlist_alloc(); |
5241 | fnvlist_add_string(holds, fullname, tag); | |
5242 | error = dsl_dataset_user_hold(holds, 0, NULL); | |
5243 | fnvlist_free(holds); | |
5244 | ||
9b67f605 MA |
5245 | if (error == ENOSPC) { |
5246 | ztest_record_enospc("dsl_dataset_user_hold"); | |
5247 | goto out; | |
5248 | } else if (error) { | |
5249 | fatal(0, "dsl_dataset_user_hold(%s, %s) = %u", | |
5250 | fullname, tag, error); | |
5251 | } | |
428870ff | 5252 | |
13fe0198 | 5253 | error = dsl_destroy_snapshot(fullname, B_FALSE); |
428870ff | 5254 | if (error != EBUSY) { |
13fe0198 | 5255 | fatal(0, "dsl_destroy_snapshot(%s, B_FALSE) = %d", |
428870ff BB |
5256 | fullname, error); |
5257 | } | |
5258 | ||
13fe0198 | 5259 | error = dsl_destroy_snapshot(fullname, B_TRUE); |
428870ff | 5260 | if (error) { |
13fe0198 | 5261 | fatal(0, "dsl_destroy_snapshot(%s, B_TRUE) = %d", |
428870ff | 5262 | fullname, error); |
34dc7c2f BB |
5263 | } |
5264 | ||
13fe0198 | 5265 | error = user_release_one(fullname, tag); |
428870ff | 5266 | if (error) |
95fd54a1 | 5267 | fatal(0, "user_release_one(%s, %s) = %d", fullname, tag, error); |
428870ff | 5268 | |
13fe0198 | 5269 | VERIFY3U(dmu_objset_hold(fullname, FTAG, &origin), ==, ENOENT); |
428870ff BB |
5270 | |
5271 | out: | |
7809eb8b | 5272 | (void) rw_unlock(&ztest_name_lock); |
34dc7c2f BB |
5273 | } |
5274 | ||
34dc7c2f BB |
5275 | /* |
5276 | * Inject random faults into the on-disk data. | |
5277 | */ | |
428870ff | 5278 | /* ARGSUSED */ |
34dc7c2f | 5279 | void |
428870ff | 5280 | ztest_fault_inject(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 5281 | { |
428870ff | 5282 | ztest_shared_t *zs = ztest_shared; |
c242c188 | 5283 | spa_t *spa = ztest_spa; |
34dc7c2f BB |
5284 | int fd; |
5285 | uint64_t offset; | |
428870ff | 5286 | uint64_t leaves; |
c5b3a7bb | 5287 | uint64_t bad = 0x1990c0ffeedecadeull; |
34dc7c2f | 5288 | uint64_t top, leaf; |
40b84e7a BB |
5289 | char *path0; |
5290 | char *pathrand; | |
34dc7c2f | 5291 | size_t fsize; |
2014c09f | 5292 | int bshift = SPA_MAXBLOCKSHIFT + 2; |
34dc7c2f | 5293 | int iters = 1000; |
428870ff BB |
5294 | int maxfaults; |
5295 | int mirror_save; | |
b128c09f | 5296 | vdev_t *vd0 = NULL; |
34dc7c2f | 5297 | uint64_t guid0 = 0; |
428870ff BB |
5298 | boolean_t islog = B_FALSE; |
5299 | ||
40b84e7a BB |
5300 | path0 = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
5301 | pathrand = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
5302 | ||
c242c188 | 5303 | mutex_enter(&ztest_vdev_lock); |
428870ff | 5304 | maxfaults = MAXFAULTS(); |
c242c188 | 5305 | leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raidz; |
428870ff | 5306 | mirror_save = zs->zs_mirrors; |
c242c188 | 5307 | mutex_exit(&ztest_vdev_lock); |
34dc7c2f | 5308 | |
b128c09f | 5309 | ASSERT(leaves >= 1); |
34dc7c2f | 5310 | |
621dd7bb GW |
5311 | /* |
5312 | * Grab the name lock as reader. There are some operations | |
5313 | * which don't like to have their vdevs changed while | |
5314 | * they are in progress (i.e. spa_change_guid). Those | |
5315 | * operations will have grabbed the name lock as writer. | |
5316 | */ | |
7809eb8b | 5317 | (void) rw_rdlock(&ztest_name_lock); |
621dd7bb | 5318 | |
34dc7c2f | 5319 | /* |
b128c09f | 5320 | * We need SCL_STATE here because we're going to look at vd0->vdev_tsd. |
34dc7c2f | 5321 | */ |
b128c09f | 5322 | spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); |
34dc7c2f | 5323 | |
b128c09f BB |
5324 | if (ztest_random(2) == 0) { |
5325 | /* | |
428870ff | 5326 | * Inject errors on a normal data device or slog device. |
b128c09f | 5327 | */ |
428870ff BB |
5328 | top = ztest_random_vdev_top(spa, B_TRUE); |
5329 | leaf = ztest_random(leaves) + zs->zs_splits; | |
34dc7c2f | 5330 | |
b128c09f BB |
5331 | /* |
5332 | * Generate paths to the first leaf in this top-level vdev, | |
5333 | * and to the random leaf we selected. We'll induce transient | |
5334 | * write failures and random online/offline activity on leaf 0, | |
5335 | * and we'll write random garbage to the randomly chosen leaf. | |
5336 | */ | |
6aec1cd5 | 5337 | (void) snprintf(path0, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
5338 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
5339 | top * leaves + zs->zs_splits); | |
6aec1cd5 | 5340 | (void) snprintf(pathrand, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
5341 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
5342 | top * leaves + leaf); | |
34dc7c2f | 5343 | |
b128c09f | 5344 | vd0 = vdev_lookup_by_path(spa->spa_root_vdev, path0); |
428870ff BB |
5345 | if (vd0 != NULL && vd0->vdev_top->vdev_islog) |
5346 | islog = B_TRUE; | |
5347 | ||
621dd7bb GW |
5348 | /* |
5349 | * If the top-level vdev needs to be resilvered | |
5350 | * then we only allow faults on the device that is | |
5351 | * resilvering. | |
5352 | */ | |
5353 | if (vd0 != NULL && maxfaults != 1 && | |
5354 | (!vdev_resilver_needed(vd0->vdev_top, NULL, NULL) || | |
5d1f7fb6 | 5355 | vd0->vdev_resilver_txg != 0)) { |
b128c09f BB |
5356 | /* |
5357 | * Make vd0 explicitly claim to be unreadable, | |
5358 | * or unwriteable, or reach behind its back | |
5359 | * and close the underlying fd. We can do this if | |
5360 | * maxfaults == 0 because we'll fail and reexecute, | |
5361 | * and we can do it if maxfaults >= 2 because we'll | |
5362 | * have enough redundancy. If maxfaults == 1, the | |
5363 | * combination of this with injection of random data | |
5364 | * corruption below exceeds the pool's fault tolerance. | |
5365 | */ | |
5366 | vdev_file_t *vf = vd0->vdev_tsd; | |
5367 | ||
5368 | if (vf != NULL && ztest_random(3) == 0) { | |
5369 | (void) close(vf->vf_vnode->v_fd); | |
5370 | vf->vf_vnode->v_fd = -1; | |
5371 | } else if (ztest_random(2) == 0) { | |
5372 | vd0->vdev_cant_read = B_TRUE; | |
5373 | } else { | |
5374 | vd0->vdev_cant_write = B_TRUE; | |
5375 | } | |
5376 | guid0 = vd0->vdev_guid; | |
5377 | } | |
5378 | } else { | |
5379 | /* | |
5380 | * Inject errors on an l2cache device. | |
5381 | */ | |
5382 | spa_aux_vdev_t *sav = &spa->spa_l2cache; | |
34dc7c2f | 5383 | |
b128c09f BB |
5384 | if (sav->sav_count == 0) { |
5385 | spa_config_exit(spa, SCL_STATE, FTAG); | |
7809eb8b | 5386 | (void) rw_unlock(&ztest_name_lock); |
40b84e7a | 5387 | goto out; |
b128c09f BB |
5388 | } |
5389 | vd0 = sav->sav_vdevs[ztest_random(sav->sav_count)]; | |
34dc7c2f | 5390 | guid0 = vd0->vdev_guid; |
b128c09f BB |
5391 | (void) strcpy(path0, vd0->vdev_path); |
5392 | (void) strcpy(pathrand, vd0->vdev_path); | |
5393 | ||
5394 | leaf = 0; | |
5395 | leaves = 1; | |
5396 | maxfaults = INT_MAX; /* no limit on cache devices */ | |
34dc7c2f BB |
5397 | } |
5398 | ||
b128c09f | 5399 | spa_config_exit(spa, SCL_STATE, FTAG); |
7809eb8b | 5400 | (void) rw_unlock(&ztest_name_lock); |
b128c09f | 5401 | |
34dc7c2f | 5402 | /* |
428870ff BB |
5403 | * If we can tolerate two or more faults, or we're dealing |
5404 | * with a slog, randomly online/offline vd0. | |
34dc7c2f | 5405 | */ |
428870ff | 5406 | if ((maxfaults >= 2 || islog) && guid0 != 0) { |
fb5f0bc8 BB |
5407 | if (ztest_random(10) < 6) { |
5408 | int flags = (ztest_random(2) == 0 ? | |
5409 | ZFS_OFFLINE_TEMPORARY : 0); | |
428870ff BB |
5410 | |
5411 | /* | |
5412 | * We have to grab the zs_name_lock as writer to | |
5413 | * prevent a race between offlining a slog and | |
5414 | * destroying a dataset. Offlining the slog will | |
5415 | * grab a reference on the dataset which may cause | |
13fe0198 | 5416 | * dsl_destroy_head() to fail with EBUSY thus |
428870ff BB |
5417 | * leaving the dataset in an inconsistent state. |
5418 | */ | |
5419 | if (islog) | |
7809eb8b | 5420 | (void) rw_wrlock(&ztest_name_lock); |
428870ff | 5421 | |
fb5f0bc8 | 5422 | VERIFY(vdev_offline(spa, guid0, flags) != EBUSY); |
428870ff BB |
5423 | |
5424 | if (islog) | |
7809eb8b | 5425 | (void) rw_unlock(&ztest_name_lock); |
fb5f0bc8 | 5426 | } else { |
1eb5bfa3 GW |
5427 | /* |
5428 | * Ideally we would like to be able to randomly | |
5429 | * call vdev_[on|off]line without holding locks | |
5430 | * to force unpredictable failures but the side | |
5431 | * effects of vdev_[on|off]line prevent us from | |
5432 | * doing so. We grab the ztest_vdev_lock here to | |
5433 | * prevent a race between injection testing and | |
5434 | * aux_vdev removal. | |
5435 | */ | |
5436 | mutex_enter(&ztest_vdev_lock); | |
fb5f0bc8 | 5437 | (void) vdev_online(spa, guid0, 0, NULL); |
1eb5bfa3 | 5438 | mutex_exit(&ztest_vdev_lock); |
fb5f0bc8 | 5439 | } |
34dc7c2f BB |
5440 | } |
5441 | ||
428870ff | 5442 | if (maxfaults == 0) |
40b84e7a | 5443 | goto out; |
428870ff | 5444 | |
34dc7c2f BB |
5445 | /* |
5446 | * We have at least single-fault tolerance, so inject data corruption. | |
5447 | */ | |
5448 | fd = open(pathrand, O_RDWR); | |
5449 | ||
5450 | if (fd == -1) /* we hit a gap in the device namespace */ | |
40b84e7a | 5451 | goto out; |
34dc7c2f BB |
5452 | |
5453 | fsize = lseek(fd, 0, SEEK_END); | |
5454 | ||
5455 | while (--iters != 0) { | |
91d88843 MA |
5456 | /* |
5457 | * The offset must be chosen carefully to ensure that | |
5458 | * we do not inject a given logical block with errors | |
5459 | * on two different leaf devices, because ZFS can not | |
5460 | * tolerate that (if maxfaults==1). | |
5461 | * | |
5462 | * We divide each leaf into chunks of size | |
5463 | * (# leaves * SPA_MAXBLOCKSIZE * 4). Within each chunk | |
5464 | * there is a series of ranges to which we can inject errors. | |
5465 | * Each range can accept errors on only a single leaf vdev. | |
5466 | * The error injection ranges are separated by ranges | |
5467 | * which we will not inject errors on any device (DMZs). | |
5468 | * Each DMZ must be large enough such that a single block | |
5469 | * can not straddle it, so that a single block can not be | |
5470 | * a target in two different injection ranges (on different | |
5471 | * leaf vdevs). | |
5472 | * | |
5473 | * For example, with 3 leaves, each chunk looks like: | |
5474 | * 0 to 32M: injection range for leaf 0 | |
5475 | * 32M to 64M: DMZ - no injection allowed | |
5476 | * 64M to 96M: injection range for leaf 1 | |
5477 | * 96M to 128M: DMZ - no injection allowed | |
5478 | * 128M to 160M: injection range for leaf 2 | |
5479 | * 160M to 192M: DMZ - no injection allowed | |
5480 | */ | |
34dc7c2f BB |
5481 | offset = ztest_random(fsize / (leaves << bshift)) * |
5482 | (leaves << bshift) + (leaf << bshift) + | |
5483 | (ztest_random(1ULL << (bshift - 1)) & -8ULL); | |
5484 | ||
2014c09f GM |
5485 | /* |
5486 | * Only allow damage to the labels at one end of the vdev. | |
5487 | * | |
5488 | * If all labels are damaged, the device will be totally | |
5489 | * inaccessible, which will result in loss of data, | |
5490 | * because we also damage (parts of) the other side of | |
5491 | * the mirror/raidz. | |
5492 | * | |
5493 | * Additionally, we will always have both an even and an | |
5494 | * odd label, so that we can handle crashes in the | |
5495 | * middle of vdev_config_sync(). | |
5496 | */ | |
5497 | if ((leaf & 1) == 0 && offset < VDEV_LABEL_START_SIZE) | |
5498 | continue; | |
5499 | ||
5500 | /* | |
5501 | * The two end labels are stored at the "end" of the disk, but | |
5502 | * the end of the disk (vdev_psize) is aligned to | |
5503 | * sizeof (vdev_label_t). | |
5504 | */ | |
5505 | uint64_t psize = P2ALIGN(fsize, sizeof (vdev_label_t)); | |
5506 | if ((leaf & 1) == 1 && | |
5507 | offset + sizeof (bad) > psize - VDEV_LABEL_END_SIZE) | |
34dc7c2f BB |
5508 | continue; |
5509 | ||
c242c188 | 5510 | mutex_enter(&ztest_vdev_lock); |
428870ff | 5511 | if (mirror_save != zs->zs_mirrors) { |
c242c188 | 5512 | mutex_exit(&ztest_vdev_lock); |
428870ff | 5513 | (void) close(fd); |
40b84e7a | 5514 | goto out; |
428870ff | 5515 | } |
34dc7c2f BB |
5516 | |
5517 | if (pwrite(fd, &bad, sizeof (bad), offset) != sizeof (bad)) | |
5518 | fatal(1, "can't inject bad word at 0x%llx in %s", | |
5519 | offset, pathrand); | |
428870ff | 5520 | |
c242c188 | 5521 | mutex_exit(&ztest_vdev_lock); |
428870ff | 5522 | |
c242c188 | 5523 | if (ztest_opts.zo_verbose >= 7) |
428870ff BB |
5524 | (void) printf("injected bad word into %s," |
5525 | " offset 0x%llx\n", pathrand, (u_longlong_t)offset); | |
34dc7c2f BB |
5526 | } |
5527 | ||
5528 | (void) close(fd); | |
40b84e7a BB |
5529 | out: |
5530 | umem_free(path0, MAXPATHLEN); | |
5531 | umem_free(pathrand, MAXPATHLEN); | |
34dc7c2f BB |
5532 | } |
5533 | ||
5534 | /* | |
428870ff | 5535 | * Verify that DDT repair works as expected. |
34dc7c2f BB |
5536 | */ |
5537 | void | |
428870ff | 5538 | ztest_ddt_repair(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 5539 | { |
428870ff | 5540 | ztest_shared_t *zs = ztest_shared; |
c242c188 | 5541 | spa_t *spa = ztest_spa; |
428870ff | 5542 | objset_t *os = zd->zd_os; |
40b84e7a | 5543 | ztest_od_t *od; |
428870ff BB |
5544 | uint64_t object, blocksize, txg, pattern, psize; |
5545 | enum zio_checksum checksum = spa_dedup_checksum(spa); | |
5546 | dmu_buf_t *db; | |
5547 | dmu_tx_t *tx; | |
a6255b7f | 5548 | abd_t *abd; |
428870ff BB |
5549 | blkptr_t blk; |
5550 | int copies = 2 * ZIO_DEDUPDITTO_MIN; | |
d6320ddb | 5551 | int i; |
34dc7c2f | 5552 | |
428870ff BB |
5553 | blocksize = ztest_random_blocksize(); |
5554 | blocksize = MIN(blocksize, 2048); /* because we write so many */ | |
34dc7c2f | 5555 | |
d1d7e268 | 5556 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 5557 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0, 0); |
34dc7c2f | 5558 | |
40b84e7a | 5559 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
d1d7e268 | 5560 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 5561 | return; |
40b84e7a | 5562 | } |
34dc7c2f BB |
5563 | |
5564 | /* | |
428870ff BB |
5565 | * Take the name lock as writer to prevent anyone else from changing |
5566 | * the pool and dataset properies we need to maintain during this test. | |
34dc7c2f | 5567 | */ |
7809eb8b | 5568 | (void) rw_wrlock(&ztest_name_lock); |
34dc7c2f | 5569 | |
428870ff BB |
5570 | if (ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_DEDUP, checksum, |
5571 | B_FALSE) != 0 || | |
5572 | ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_COPIES, 1, | |
5573 | B_FALSE) != 0) { | |
7809eb8b | 5574 | (void) rw_unlock(&ztest_name_lock); |
d1d7e268 | 5575 | umem_free(od, sizeof (ztest_od_t)); |
428870ff BB |
5576 | return; |
5577 | } | |
5578 | ||
546d32ca BB |
5579 | dmu_objset_stats_t dds; |
5580 | dsl_pool_config_enter(dmu_objset_pool(os), FTAG); | |
5581 | dmu_objset_fast_stat(os, &dds); | |
5582 | dsl_pool_config_exit(dmu_objset_pool(os), FTAG); | |
5583 | ||
428870ff BB |
5584 | object = od[0].od_object; |
5585 | blocksize = od[0].od_blocksize; | |
546d32ca | 5586 | pattern = zs->zs_guid ^ dds.dds_guid; |
428870ff BB |
5587 | |
5588 | ASSERT(object != 0); | |
5589 | ||
5590 | tx = dmu_tx_create(os); | |
5591 | dmu_tx_hold_write(tx, object, 0, copies * blocksize); | |
5592 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
5593 | if (txg == 0) { | |
7809eb8b | 5594 | (void) rw_unlock(&ztest_name_lock); |
d1d7e268 | 5595 | umem_free(od, sizeof (ztest_od_t)); |
428870ff BB |
5596 | return; |
5597 | } | |
34dc7c2f BB |
5598 | |
5599 | /* | |
428870ff | 5600 | * Write all the copies of our block. |
34dc7c2f | 5601 | */ |
d6320ddb | 5602 | for (i = 0; i < copies; i++) { |
428870ff | 5603 | uint64_t offset = i * blocksize; |
13fe0198 MA |
5604 | int error = dmu_buf_hold(os, object, offset, FTAG, &db, |
5605 | DMU_READ_NO_PREFETCH); | |
5606 | if (error != 0) { | |
5607 | fatal(B_FALSE, "dmu_buf_hold(%p, %llu, %llu) = %u", | |
5608 | os, (long long)object, (long long) offset, error); | |
5609 | } | |
428870ff BB |
5610 | ASSERT(db->db_offset == offset); |
5611 | ASSERT(db->db_size == blocksize); | |
5612 | ASSERT(ztest_pattern_match(db->db_data, db->db_size, pattern) || | |
5613 | ztest_pattern_match(db->db_data, db->db_size, 0ULL)); | |
5614 | dmu_buf_will_fill(db, tx); | |
5615 | ztest_pattern_set(db->db_data, db->db_size, pattern); | |
5616 | dmu_buf_rele(db, FTAG); | |
5617 | } | |
34dc7c2f | 5618 | |
428870ff BB |
5619 | dmu_tx_commit(tx); |
5620 | txg_wait_synced(spa_get_dsl(spa), txg); | |
34dc7c2f BB |
5621 | |
5622 | /* | |
428870ff | 5623 | * Find out what block we got. |
34dc7c2f | 5624 | */ |
03c6040b GW |
5625 | VERIFY0(dmu_buf_hold(os, object, 0, FTAG, &db, |
5626 | DMU_READ_NO_PREFETCH)); | |
428870ff BB |
5627 | blk = *((dmu_buf_impl_t *)db)->db_blkptr; |
5628 | dmu_buf_rele(db, FTAG); | |
34dc7c2f BB |
5629 | |
5630 | /* | |
428870ff | 5631 | * Damage the block. Dedup-ditto will save us when we read it later. |
34dc7c2f | 5632 | */ |
428870ff | 5633 | psize = BP_GET_PSIZE(&blk); |
a6255b7f DQ |
5634 | abd = abd_alloc_linear(psize, B_TRUE); |
5635 | ztest_pattern_set(abd_to_buf(abd), psize, ~pattern); | |
34dc7c2f | 5636 | |
428870ff | 5637 | (void) zio_wait(zio_rewrite(NULL, spa, 0, &blk, |
a6255b7f | 5638 | abd, psize, NULL, NULL, ZIO_PRIORITY_SYNC_WRITE, |
428870ff | 5639 | ZIO_FLAG_CANFAIL | ZIO_FLAG_INDUCE_DAMAGE, NULL)); |
34dc7c2f | 5640 | |
a6255b7f | 5641 | abd_free(abd); |
34dc7c2f | 5642 | |
7809eb8b | 5643 | (void) rw_unlock(&ztest_name_lock); |
d1d7e268 | 5644 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f BB |
5645 | } |
5646 | ||
34dc7c2f | 5647 | /* |
428870ff | 5648 | * Scrub the pool. |
34dc7c2f | 5649 | */ |
428870ff BB |
5650 | /* ARGSUSED */ |
5651 | void | |
5652 | ztest_scrub(ztest_ds_t *zd, uint64_t id) | |
34dc7c2f | 5653 | { |
c242c188 | 5654 | spa_t *spa = ztest_spa; |
34dc7c2f | 5655 | |
428870ff BB |
5656 | (void) spa_scan(spa, POOL_SCAN_SCRUB); |
5657 | (void) poll(NULL, 0, 100); /* wait a moment, then force a restart */ | |
5658 | (void) spa_scan(spa, POOL_SCAN_SCRUB); | |
5659 | } | |
34dc7c2f | 5660 | |
3541dc6d GA |
5661 | /* |
5662 | * Change the guid for the pool. | |
5663 | */ | |
5664 | /* ARGSUSED */ | |
5665 | void | |
5666 | ztest_reguid(ztest_ds_t *zd, uint64_t id) | |
5667 | { | |
c242c188 | 5668 | spa_t *spa = ztest_spa; |
3541dc6d | 5669 | uint64_t orig, load; |
3bc7e0fb | 5670 | int error; |
3541dc6d GA |
5671 | |
5672 | orig = spa_guid(spa); | |
5673 | load = spa_load_guid(spa); | |
3bc7e0fb | 5674 | |
7809eb8b | 5675 | (void) rw_wrlock(&ztest_name_lock); |
3bc7e0fb | 5676 | error = spa_change_guid(spa); |
7809eb8b | 5677 | (void) rw_unlock(&ztest_name_lock); |
3bc7e0fb GW |
5678 | |
5679 | if (error != 0) | |
3541dc6d GA |
5680 | return; |
5681 | ||
ea0b2538 | 5682 | if (ztest_opts.zo_verbose >= 4) { |
3541dc6d GA |
5683 | (void) printf("Changed guid old %llu -> %llu\n", |
5684 | (u_longlong_t)orig, (u_longlong_t)spa_guid(spa)); | |
5685 | } | |
5686 | ||
5687 | VERIFY3U(orig, !=, spa_guid(spa)); | |
5688 | VERIFY3U(load, ==, spa_load_guid(spa)); | |
5689 | } | |
5690 | ||
428870ff BB |
5691 | /* |
5692 | * Rename the pool to a different name and then rename it back. | |
5693 | */ | |
5694 | /* ARGSUSED */ | |
5695 | void | |
5696 | ztest_spa_rename(ztest_ds_t *zd, uint64_t id) | |
5697 | { | |
428870ff BB |
5698 | char *oldname, *newname; |
5699 | spa_t *spa; | |
34dc7c2f | 5700 | |
379ca9cf OF |
5701 | if (ztest_opts.zo_mmp_test) |
5702 | return; | |
5703 | ||
7809eb8b | 5704 | (void) rw_wrlock(&ztest_name_lock); |
34dc7c2f | 5705 | |
c242c188 | 5706 | oldname = ztest_opts.zo_pool; |
428870ff BB |
5707 | newname = umem_alloc(strlen(oldname) + 5, UMEM_NOFAIL); |
5708 | (void) strcpy(newname, oldname); | |
5709 | (void) strcat(newname, "_tmp"); | |
34dc7c2f BB |
5710 | |
5711 | /* | |
428870ff | 5712 | * Do the rename |
34dc7c2f | 5713 | */ |
428870ff | 5714 | VERIFY3U(0, ==, spa_rename(oldname, newname)); |
34dc7c2f BB |
5715 | |
5716 | /* | |
428870ff | 5717 | * Try to open it under the old name, which shouldn't exist |
34dc7c2f | 5718 | */ |
428870ff | 5719 | VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f BB |
5720 | |
5721 | /* | |
428870ff BB |
5722 | * Open it under the new name and make sure it's still the same spa_t. |
5723 | */ | |
5724 | VERIFY3U(0, ==, spa_open(newname, &spa, FTAG)); | |
34dc7c2f | 5725 | |
c242c188 | 5726 | ASSERT(spa == ztest_spa); |
428870ff | 5727 | spa_close(spa, FTAG); |
34dc7c2f BB |
5728 | |
5729 | /* | |
428870ff | 5730 | * Rename it back to the original |
34dc7c2f | 5731 | */ |
428870ff | 5732 | VERIFY3U(0, ==, spa_rename(newname, oldname)); |
34dc7c2f | 5733 | |
428870ff BB |
5734 | /* |
5735 | * Make sure it can still be opened | |
5736 | */ | |
5737 | VERIFY3U(0, ==, spa_open(oldname, &spa, FTAG)); | |
34dc7c2f | 5738 | |
c242c188 | 5739 | ASSERT(spa == ztest_spa); |
428870ff BB |
5740 | spa_close(spa, FTAG); |
5741 | ||
5742 | umem_free(newname, strlen(newname) + 1); | |
5743 | ||
7809eb8b | 5744 | (void) rw_unlock(&ztest_name_lock); |
34dc7c2f BB |
5745 | } |
5746 | ||
1eeb4562 JX |
5747 | void |
5748 | ztest_fletcher(ztest_ds_t *zd, uint64_t id) | |
5749 | { | |
5750 | hrtime_t end = gethrtime() + NANOSEC; | |
5751 | ||
5752 | while (gethrtime() <= end) { | |
5753 | int run_count = 100; | |
5754 | void *buf; | |
2fe36b0b | 5755 | struct abd *abd_data, *abd_meta; |
1eeb4562 JX |
5756 | uint32_t size; |
5757 | int *ptr; | |
5758 | int i; | |
5759 | zio_cksum_t zc_ref; | |
5760 | zio_cksum_t zc_ref_byteswap; | |
5761 | ||
5762 | size = ztest_random_blocksize(); | |
2fe36b0b | 5763 | |
1eeb4562 | 5764 | buf = umem_alloc(size, UMEM_NOFAIL); |
2fe36b0b DQ |
5765 | abd_data = abd_alloc(size, B_FALSE); |
5766 | abd_meta = abd_alloc(size, B_TRUE); | |
1eeb4562 JX |
5767 | |
5768 | for (i = 0, ptr = buf; i < size / sizeof (*ptr); i++, ptr++) | |
5769 | *ptr = ztest_random(UINT_MAX); | |
5770 | ||
2fe36b0b DQ |
5771 | abd_copy_from_buf_off(abd_data, buf, 0, size); |
5772 | abd_copy_from_buf_off(abd_meta, buf, 0, size); | |
5773 | ||
1eeb4562 | 5774 | VERIFY0(fletcher_4_impl_set("scalar")); |
3c67d83a TH |
5775 | fletcher_4_native(buf, size, NULL, &zc_ref); |
5776 | fletcher_4_byteswap(buf, size, NULL, &zc_ref_byteswap); | |
1eeb4562 JX |
5777 | |
5778 | VERIFY0(fletcher_4_impl_set("cycle")); | |
5779 | while (run_count-- > 0) { | |
5780 | zio_cksum_t zc; | |
5781 | zio_cksum_t zc_byteswap; | |
5782 | ||
3c67d83a TH |
5783 | fletcher_4_byteswap(buf, size, NULL, &zc_byteswap); |
5784 | fletcher_4_native(buf, size, NULL, &zc); | |
1eeb4562 JX |
5785 | |
5786 | VERIFY0(bcmp(&zc, &zc_ref, sizeof (zc))); | |
5787 | VERIFY0(bcmp(&zc_byteswap, &zc_ref_byteswap, | |
5788 | sizeof (zc_byteswap))); | |
2fe36b0b DQ |
5789 | |
5790 | /* Test ABD - data */ | |
5791 | abd_fletcher_4_byteswap(abd_data, size, NULL, | |
5792 | &zc_byteswap); | |
5793 | abd_fletcher_4_native(abd_data, size, NULL, &zc); | |
5794 | ||
5795 | VERIFY0(bcmp(&zc, &zc_ref, sizeof (zc))); | |
5796 | VERIFY0(bcmp(&zc_byteswap, &zc_ref_byteswap, | |
5797 | sizeof (zc_byteswap))); | |
5798 | ||
5799 | /* Test ABD - metadata */ | |
5800 | abd_fletcher_4_byteswap(abd_meta, size, NULL, | |
5801 | &zc_byteswap); | |
5802 | abd_fletcher_4_native(abd_meta, size, NULL, &zc); | |
5803 | ||
5804 | VERIFY0(bcmp(&zc, &zc_ref, sizeof (zc))); | |
5805 | VERIFY0(bcmp(&zc_byteswap, &zc_ref_byteswap, | |
5806 | sizeof (zc_byteswap))); | |
5807 | ||
1eeb4562 JX |
5808 | } |
5809 | ||
5810 | umem_free(buf, size); | |
2fe36b0b DQ |
5811 | abd_free(abd_data); |
5812 | abd_free(abd_meta); | |
1eeb4562 JX |
5813 | } |
5814 | } | |
5815 | ||
37f520db GN |
5816 | void |
5817 | ztest_fletcher_incr(ztest_ds_t *zd, uint64_t id) | |
5818 | { | |
5819 | void *buf; | |
5820 | size_t size; | |
5821 | int *ptr; | |
5822 | int i; | |
5823 | zio_cksum_t zc_ref; | |
5824 | zio_cksum_t zc_ref_bswap; | |
5825 | ||
5826 | hrtime_t end = gethrtime() + NANOSEC; | |
5827 | ||
5828 | while (gethrtime() <= end) { | |
5829 | int run_count = 100; | |
5830 | ||
5831 | size = ztest_random_blocksize(); | |
5832 | buf = umem_alloc(size, UMEM_NOFAIL); | |
5833 | ||
5834 | for (i = 0, ptr = buf; i < size / sizeof (*ptr); i++, ptr++) | |
5835 | *ptr = ztest_random(UINT_MAX); | |
5836 | ||
5837 | VERIFY0(fletcher_4_impl_set("scalar")); | |
5838 | fletcher_4_native(buf, size, NULL, &zc_ref); | |
5839 | fletcher_4_byteswap(buf, size, NULL, &zc_ref_bswap); | |
5840 | ||
5841 | VERIFY0(fletcher_4_impl_set("cycle")); | |
5842 | ||
5843 | while (run_count-- > 0) { | |
5844 | zio_cksum_t zc; | |
5845 | zio_cksum_t zc_bswap; | |
5846 | size_t pos = 0; | |
5847 | ||
5848 | ZIO_SET_CHECKSUM(&zc, 0, 0, 0, 0); | |
5849 | ZIO_SET_CHECKSUM(&zc_bswap, 0, 0, 0, 0); | |
5850 | ||
5851 | while (pos < size) { | |
5852 | size_t inc = 64 * ztest_random(size / 67); | |
5853 | /* sometimes add few bytes to test non-simd */ | |
5854 | if (ztest_random(100) < 10) | |
5855 | inc += P2ALIGN(ztest_random(64), | |
5856 | sizeof (uint32_t)); | |
5857 | ||
5858 | if (inc > (size - pos)) | |
5859 | inc = size - pos; | |
5860 | ||
5861 | fletcher_4_incremental_native(buf + pos, inc, | |
5862 | &zc); | |
5863 | fletcher_4_incremental_byteswap(buf + pos, inc, | |
5864 | &zc_bswap); | |
5865 | ||
5866 | pos += inc; | |
5867 | } | |
5868 | ||
5869 | VERIFY3U(pos, ==, size); | |
5870 | ||
5871 | VERIFY(ZIO_CHECKSUM_EQUAL(zc, zc_ref)); | |
5872 | VERIFY(ZIO_CHECKSUM_EQUAL(zc_bswap, zc_ref_bswap)); | |
5873 | ||
5874 | /* | |
5875 | * verify if incremental on the whole buffer is | |
5876 | * equivalent to non-incremental version | |
5877 | */ | |
5878 | ZIO_SET_CHECKSUM(&zc, 0, 0, 0, 0); | |
5879 | ZIO_SET_CHECKSUM(&zc_bswap, 0, 0, 0, 0); | |
5880 | ||
5881 | fletcher_4_incremental_native(buf, size, &zc); | |
5882 | fletcher_4_incremental_byteswap(buf, size, &zc_bswap); | |
5883 | ||
5884 | VERIFY(ZIO_CHECKSUM_EQUAL(zc, zc_ref)); | |
5885 | VERIFY(ZIO_CHECKSUM_EQUAL(zc_bswap, zc_ref_bswap)); | |
5886 | } | |
5887 | ||
5888 | umem_free(buf, size); | |
5889 | } | |
5890 | } | |
5891 | ||
90aa094d CC |
5892 | static int |
5893 | ztest_check_path(char *path) | |
5894 | { | |
5895 | struct stat s; | |
5896 | /* return true on success */ | |
5897 | return (!stat(path, &s)); | |
5898 | } | |
5899 | ||
5900 | static void | |
5901 | ztest_get_zdb_bin(char *bin, int len) | |
5902 | { | |
5903 | char *zdb_path; | |
5904 | /* | |
5905 | * Try to use ZDB_PATH and in-tree zdb path. If not successful, just | |
5906 | * let popen to search through PATH. | |
5907 | */ | |
5908 | if ((zdb_path = getenv("ZDB_PATH"))) { | |
5909 | strlcpy(bin, zdb_path, len); /* In env */ | |
5910 | if (!ztest_check_path(bin)) { | |
5911 | ztest_dump_core = 0; | |
5912 | fatal(1, "invalid ZDB_PATH '%s'", bin); | |
5913 | } | |
5914 | return; | |
5915 | } | |
5916 | ||
5917 | VERIFY(realpath(getexecname(), bin) != NULL); | |
5918 | if (strstr(bin, "/ztest/")) { | |
5919 | strstr(bin, "/ztest/")[0] = '\0'; /* In-tree */ | |
5920 | strcat(bin, "/zdb/zdb"); | |
5921 | if (ztest_check_path(bin)) | |
5922 | return; | |
5923 | } | |
5924 | strcpy(bin, "zdb"); | |
5925 | } | |
5926 | ||
428870ff BB |
5927 | /* |
5928 | * Verify pool integrity by running zdb. | |
5929 | */ | |
34dc7c2f | 5930 | static void |
428870ff | 5931 | ztest_run_zdb(char *pool) |
34dc7c2f BB |
5932 | { |
5933 | int status; | |
34dc7c2f | 5934 | char *bin; |
0e8d1b2d BB |
5935 | char *zdb; |
5936 | char *zbuf; | |
90aa094d | 5937 | const int len = MAXPATHLEN + MAXNAMELEN + 20; |
34dc7c2f BB |
5938 | FILE *fp; |
5939 | ||
90aa094d CC |
5940 | bin = umem_alloc(len, UMEM_NOFAIL); |
5941 | zdb = umem_alloc(len, UMEM_NOFAIL); | |
0e8d1b2d | 5942 | zbuf = umem_alloc(1024, UMEM_NOFAIL); |
34dc7c2f | 5943 | |
90aa094d | 5944 | ztest_get_zdb_bin(bin, len); |
0e8d1b2d BB |
5945 | |
5946 | (void) sprintf(zdb, | |
456079d4 | 5947 | "%s -bcc%s%s -G -d -U %s %s", |
0e8d1b2d | 5948 | bin, |
c242c188 CS |
5949 | ztest_opts.zo_verbose >= 3 ? "s" : "", |
5950 | ztest_opts.zo_verbose >= 4 ? "v" : "", | |
428870ff | 5951 | spa_config_path, |
b128c09f | 5952 | pool); |
34dc7c2f | 5953 | |
c242c188 | 5954 | if (ztest_opts.zo_verbose >= 5) |
34dc7c2f BB |
5955 | (void) printf("Executing %s\n", strstr(zdb, "zdb ")); |
5956 | ||
5957 | fp = popen(zdb, "r"); | |
5958 | ||
6aec1cd5 | 5959 | while (fgets(zbuf, 1024, fp) != NULL) |
c242c188 | 5960 | if (ztest_opts.zo_verbose >= 3) |
34dc7c2f BB |
5961 | (void) printf("%s", zbuf); |
5962 | ||
5963 | status = pclose(fp); | |
5964 | ||
5965 | if (status == 0) | |
0e8d1b2d | 5966 | goto out; |
34dc7c2f BB |
5967 | |
5968 | ztest_dump_core = 0; | |
5969 | if (WIFEXITED(status)) | |
5970 | fatal(0, "'%s' exit code %d", zdb, WEXITSTATUS(status)); | |
5971 | else | |
5972 | fatal(0, "'%s' died with signal %d", zdb, WTERMSIG(status)); | |
0e8d1b2d | 5973 | out: |
90aa094d CC |
5974 | umem_free(bin, len); |
5975 | umem_free(zdb, len); | |
0e8d1b2d | 5976 | umem_free(zbuf, 1024); |
34dc7c2f BB |
5977 | } |
5978 | ||
5979 | static void | |
5980 | ztest_walk_pool_directory(char *header) | |
5981 | { | |
5982 | spa_t *spa = NULL; | |
5983 | ||
c242c188 | 5984 | if (ztest_opts.zo_verbose >= 6) |
34dc7c2f BB |
5985 | (void) printf("%s\n", header); |
5986 | ||
5987 | mutex_enter(&spa_namespace_lock); | |
5988 | while ((spa = spa_next(spa)) != NULL) | |
c242c188 | 5989 | if (ztest_opts.zo_verbose >= 6) |
34dc7c2f BB |
5990 | (void) printf("\t%s\n", spa_name(spa)); |
5991 | mutex_exit(&spa_namespace_lock); | |
5992 | } | |
5993 | ||
5994 | static void | |
5995 | ztest_spa_import_export(char *oldname, char *newname) | |
5996 | { | |
fb5f0bc8 | 5997 | nvlist_t *config, *newconfig; |
34dc7c2f BB |
5998 | uint64_t pool_guid; |
5999 | spa_t *spa; | |
13fe0198 | 6000 | int error; |
34dc7c2f | 6001 | |
c242c188 | 6002 | if (ztest_opts.zo_verbose >= 4) { |
34dc7c2f BB |
6003 | (void) printf("import/export: old = %s, new = %s\n", |
6004 | oldname, newname); | |
6005 | } | |
6006 | ||
6007 | /* | |
6008 | * Clean up from previous runs. | |
6009 | */ | |
6010 | (void) spa_destroy(newname); | |
6011 | ||
6012 | /* | |
6013 | * Get the pool's configuration and guid. | |
6014 | */ | |
428870ff | 6015 | VERIFY3U(0, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f | 6016 | |
fb5f0bc8 BB |
6017 | /* |
6018 | * Kick off a scrub to tickle scrub/export races. | |
6019 | */ | |
6020 | if (ztest_random(2) == 0) | |
428870ff | 6021 | (void) spa_scan(spa, POOL_SCAN_SCRUB); |
fb5f0bc8 | 6022 | |
34dc7c2f BB |
6023 | pool_guid = spa_guid(spa); |
6024 | spa_close(spa, FTAG); | |
6025 | ||
6026 | ztest_walk_pool_directory("pools before export"); | |
6027 | ||
6028 | /* | |
6029 | * Export it. | |
6030 | */ | |
428870ff | 6031 | VERIFY3U(0, ==, spa_export(oldname, &config, B_FALSE, B_FALSE)); |
34dc7c2f BB |
6032 | |
6033 | ztest_walk_pool_directory("pools after export"); | |
6034 | ||
fb5f0bc8 BB |
6035 | /* |
6036 | * Try to import it. | |
6037 | */ | |
6038 | newconfig = spa_tryimport(config); | |
6039 | ASSERT(newconfig != NULL); | |
6040 | nvlist_free(newconfig); | |
6041 | ||
34dc7c2f BB |
6042 | /* |
6043 | * Import it under the new name. | |
6044 | */ | |
13fe0198 MA |
6045 | error = spa_import(newname, config, NULL, 0); |
6046 | if (error != 0) { | |
6047 | dump_nvlist(config, 0); | |
6048 | fatal(B_FALSE, "couldn't import pool %s as %s: error %u", | |
6049 | oldname, newname, error); | |
6050 | } | |
34dc7c2f BB |
6051 | |
6052 | ztest_walk_pool_directory("pools after import"); | |
6053 | ||
6054 | /* | |
6055 | * Try to import it again -- should fail with EEXIST. | |
6056 | */ | |
572e2857 | 6057 | VERIFY3U(EEXIST, ==, spa_import(newname, config, NULL, 0)); |
34dc7c2f BB |
6058 | |
6059 | /* | |
6060 | * Try to import it under a different name -- should fail with EEXIST. | |
6061 | */ | |
572e2857 | 6062 | VERIFY3U(EEXIST, ==, spa_import(oldname, config, NULL, 0)); |
34dc7c2f BB |
6063 | |
6064 | /* | |
6065 | * Verify that the pool is no longer visible under the old name. | |
6066 | */ | |
428870ff | 6067 | VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f BB |
6068 | |
6069 | /* | |
6070 | * Verify that we can open and close the pool using the new name. | |
6071 | */ | |
428870ff | 6072 | VERIFY3U(0, ==, spa_open(newname, &spa, FTAG)); |
34dc7c2f BB |
6073 | ASSERT(pool_guid == spa_guid(spa)); |
6074 | spa_close(spa, FTAG); | |
6075 | ||
6076 | nvlist_free(config); | |
6077 | } | |
6078 | ||
fb5f0bc8 BB |
6079 | static void |
6080 | ztest_resume(spa_t *spa) | |
6081 | { | |
c242c188 | 6082 | if (spa_suspended(spa) && ztest_opts.zo_verbose >= 6) |
428870ff BB |
6083 | (void) printf("resuming from suspended state\n"); |
6084 | spa_vdev_state_enter(spa, SCL_NONE); | |
6085 | vdev_clear(spa, NULL); | |
6086 | (void) spa_vdev_state_exit(spa, NULL, 0); | |
6087 | (void) zio_resume(spa); | |
fb5f0bc8 BB |
6088 | } |
6089 | ||
c25b8f99 | 6090 | static void |
fb5f0bc8 | 6091 | ztest_resume_thread(void *arg) |
34dc7c2f | 6092 | { |
b128c09f | 6093 | spa_t *spa = arg; |
34dc7c2f BB |
6094 | |
6095 | while (!ztest_exiting) { | |
428870ff BB |
6096 | if (spa_suspended(spa)) |
6097 | ztest_resume(spa); | |
6098 | (void) poll(NULL, 0, 100); | |
d3c2ae1c GW |
6099 | |
6100 | /* | |
6101 | * Periodically change the zfs_compressed_arc_enabled setting. | |
6102 | */ | |
6103 | if (ztest_random(10) == 0) | |
6104 | zfs_compressed_arc_enabled = ztest_random(2); | |
a6255b7f DQ |
6105 | |
6106 | /* | |
6107 | * Periodically change the zfs_abd_scatter_enabled setting. | |
6108 | */ | |
6109 | if (ztest_random(10) == 0) | |
6110 | zfs_abd_scatter_enabled = ztest_random(2); | |
34dc7c2f | 6111 | } |
34dc7c2f | 6112 | |
1e33ac1e | 6113 | thread_exit(); |
1e33ac1e | 6114 | } |
428870ff | 6115 | |
d1d7e268 | 6116 | #define GRACE 300 |
428870ff | 6117 | |
26099167 | 6118 | #if 0 |
1e33ac1e BB |
6119 | static void |
6120 | ztest_deadman_alarm(int sig) | |
6121 | { | |
6122 | fatal(0, "failed to complete within %d seconds of deadline", GRACE); | |
428870ff | 6123 | } |
26099167 | 6124 | #endif |
428870ff BB |
6125 | |
6126 | static void | |
c242c188 | 6127 | ztest_execute(int test, ztest_info_t *zi, uint64_t id) |
428870ff | 6128 | { |
c242c188 CS |
6129 | ztest_ds_t *zd = &ztest_ds[id % ztest_opts.zo_datasets]; |
6130 | ztest_shared_callstate_t *zc = ZTEST_GET_SHARED_CALLSTATE(test); | |
428870ff | 6131 | hrtime_t functime = gethrtime(); |
d6320ddb | 6132 | int i; |
428870ff | 6133 | |
d6320ddb | 6134 | for (i = 0; i < zi->zi_iters; i++) |
428870ff BB |
6135 | zi->zi_func(zd, id); |
6136 | ||
6137 | functime = gethrtime() - functime; | |
6138 | ||
c242c188 CS |
6139 | atomic_add_64(&zc->zc_count, 1); |
6140 | atomic_add_64(&zc->zc_time, functime); | |
428870ff | 6141 | |
fdc5d982 | 6142 | if (ztest_opts.zo_verbose >= 4) |
428870ff | 6143 | (void) printf("%6.2f sec in %s\n", |
fdc5d982 | 6144 | (double)functime / NANOSEC, zi->zi_funcname); |
428870ff BB |
6145 | } |
6146 | ||
c25b8f99 | 6147 | static void |
34dc7c2f BB |
6148 | ztest_thread(void *arg) |
6149 | { | |
c242c188 | 6150 | int rand; |
428870ff | 6151 | uint64_t id = (uintptr_t)arg; |
34dc7c2f | 6152 | ztest_shared_t *zs = ztest_shared; |
428870ff BB |
6153 | uint64_t call_next; |
6154 | hrtime_t now; | |
34dc7c2f | 6155 | ztest_info_t *zi; |
c242c188 | 6156 | ztest_shared_callstate_t *zc; |
34dc7c2f | 6157 | |
428870ff | 6158 | while ((now = gethrtime()) < zs->zs_thread_stop) { |
34dc7c2f BB |
6159 | /* |
6160 | * See if it's time to force a crash. | |
6161 | */ | |
428870ff BB |
6162 | if (now > zs->zs_thread_kill) |
6163 | ztest_kill(zs); | |
34dc7c2f BB |
6164 | |
6165 | /* | |
428870ff | 6166 | * If we're getting ENOSPC with some regularity, stop. |
34dc7c2f | 6167 | */ |
428870ff BB |
6168 | if (zs->zs_enospc_count > 10) |
6169 | break; | |
34dc7c2f BB |
6170 | |
6171 | /* | |
428870ff | 6172 | * Pick a random function to execute. |
34dc7c2f | 6173 | */ |
c242c188 CS |
6174 | rand = ztest_random(ZTEST_FUNCS); |
6175 | zi = &ztest_info[rand]; | |
6176 | zc = ZTEST_GET_SHARED_CALLSTATE(rand); | |
6177 | call_next = zc->zc_next; | |
34dc7c2f | 6178 | |
428870ff | 6179 | if (now >= call_next && |
c242c188 CS |
6180 | atomic_cas_64(&zc->zc_next, call_next, call_next + |
6181 | ztest_random(2 * zi->zi_interval[0] + 1)) == call_next) { | |
6182 | ztest_execute(rand, zi, id); | |
6183 | } | |
428870ff | 6184 | } |
34dc7c2f | 6185 | |
1e33ac1e | 6186 | thread_exit(); |
428870ff | 6187 | } |
34dc7c2f | 6188 | |
428870ff BB |
6189 | static void |
6190 | ztest_dataset_name(char *dsname, char *pool, int d) | |
6191 | { | |
eca7b760 | 6192 | (void) snprintf(dsname, ZFS_MAX_DATASET_NAME_LEN, "%s/ds_%d", pool, d); |
428870ff | 6193 | } |
34dc7c2f | 6194 | |
428870ff | 6195 | static void |
c242c188 | 6196 | ztest_dataset_destroy(int d) |
428870ff | 6197 | { |
eca7b760 | 6198 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
d6320ddb | 6199 | int t; |
34dc7c2f | 6200 | |
c242c188 | 6201 | ztest_dataset_name(name, ztest_opts.zo_pool, d); |
34dc7c2f | 6202 | |
c242c188 | 6203 | if (ztest_opts.zo_verbose >= 3) |
428870ff | 6204 | (void) printf("Destroying %s to free up space\n", name); |
34dc7c2f | 6205 | |
428870ff BB |
6206 | /* |
6207 | * Cleanup any non-standard clones and snapshots. In general, | |
6208 | * ztest thread t operates on dataset (t % zopt_datasets), | |
6209 | * so there may be more than one thing to clean up. | |
6210 | */ | |
c242c188 CS |
6211 | for (t = d; t < ztest_opts.zo_threads; |
6212 | t += ztest_opts.zo_datasets) | |
428870ff BB |
6213 | ztest_dsl_dataset_cleanup(name, t); |
6214 | ||
6215 | (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL, | |
6216 | DS_FIND_SNAPSHOTS | DS_FIND_CHILDREN); | |
6217 | } | |
6218 | ||
6219 | static void | |
6220 | ztest_dataset_dirobj_verify(ztest_ds_t *zd) | |
6221 | { | |
6222 | uint64_t usedobjs, dirobjs, scratch; | |
6223 | ||
6224 | /* | |
6225 | * ZTEST_DIROBJ is the object directory for the entire dataset. | |
6226 | * Therefore, the number of objects in use should equal the | |
6227 | * number of ZTEST_DIROBJ entries, +1 for ZTEST_DIROBJ itself. | |
6228 | * If not, we have an object leak. | |
6229 | * | |
6230 | * Note that we can only check this in ztest_dataset_open(), | |
6231 | * when the open-context and syncing-context values agree. | |
6232 | * That's because zap_count() returns the open-context value, | |
6233 | * while dmu_objset_space() returns the rootbp fill count. | |
6234 | */ | |
6235 | VERIFY3U(0, ==, zap_count(zd->zd_os, ZTEST_DIROBJ, &dirobjs)); | |
6236 | dmu_objset_space(zd->zd_os, &scratch, &scratch, &usedobjs, &scratch); | |
6237 | ASSERT3U(dirobjs + 1, ==, usedobjs); | |
6238 | } | |
6239 | ||
6240 | static int | |
c242c188 | 6241 | ztest_dataset_open(int d) |
428870ff | 6242 | { |
c242c188 CS |
6243 | ztest_ds_t *zd = &ztest_ds[d]; |
6244 | uint64_t committed_seq = ZTEST_GET_SHARED_DS(d)->zd_seq; | |
428870ff BB |
6245 | objset_t *os; |
6246 | zilog_t *zilog; | |
eca7b760 | 6247 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
6248 | int error; |
6249 | ||
c242c188 | 6250 | ztest_dataset_name(name, ztest_opts.zo_pool, d); |
428870ff | 6251 | |
7809eb8b | 6252 | (void) rw_rdlock(&ztest_name_lock); |
428870ff BB |
6253 | |
6254 | error = ztest_dataset_create(name); | |
6255 | if (error == ENOSPC) { | |
7809eb8b | 6256 | (void) rw_unlock(&ztest_name_lock); |
428870ff BB |
6257 | ztest_record_enospc(FTAG); |
6258 | return (error); | |
34dc7c2f | 6259 | } |
428870ff | 6260 | ASSERT(error == 0 || error == EEXIST); |
34dc7c2f | 6261 | |
b5256303 | 6262 | VERIFY0(dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, B_TRUE, zd, &os)); |
7809eb8b | 6263 | (void) rw_unlock(&ztest_name_lock); |
428870ff | 6264 | |
c242c188 | 6265 | ztest_zd_init(zd, ZTEST_GET_SHARED_DS(d), os); |
428870ff BB |
6266 | |
6267 | zilog = zd->zd_zilog; | |
6268 | ||
6269 | if (zilog->zl_header->zh_claim_lr_seq != 0 && | |
6270 | zilog->zl_header->zh_claim_lr_seq < committed_seq) | |
6271 | fatal(0, "missing log records: claimed %llu < committed %llu", | |
6272 | zilog->zl_header->zh_claim_lr_seq, committed_seq); | |
6273 | ||
6274 | ztest_dataset_dirobj_verify(zd); | |
6275 | ||
6276 | zil_replay(os, zd, ztest_replay_vector); | |
6277 | ||
6278 | ztest_dataset_dirobj_verify(zd); | |
6279 | ||
c242c188 | 6280 | if (ztest_opts.zo_verbose >= 6) |
428870ff BB |
6281 | (void) printf("%s replay %llu blocks, %llu records, seq %llu\n", |
6282 | zd->zd_name, | |
6283 | (u_longlong_t)zilog->zl_parse_blk_count, | |
6284 | (u_longlong_t)zilog->zl_parse_lr_count, | |
6285 | (u_longlong_t)zilog->zl_replaying_seq); | |
6286 | ||
6287 | zilog = zil_open(os, ztest_get_data); | |
6288 | ||
6289 | if (zilog->zl_replaying_seq != 0 && | |
6290 | zilog->zl_replaying_seq < committed_seq) | |
6291 | fatal(0, "missing log records: replayed %llu < committed %llu", | |
6292 | zilog->zl_replaying_seq, committed_seq); | |
6293 | ||
6294 | return (0); | |
6295 | } | |
6296 | ||
6297 | static void | |
c242c188 | 6298 | ztest_dataset_close(int d) |
428870ff | 6299 | { |
c242c188 | 6300 | ztest_ds_t *zd = &ztest_ds[d]; |
428870ff BB |
6301 | |
6302 | zil_close(zd->zd_zilog); | |
b5256303 | 6303 | dmu_objset_disown(zd->zd_os, B_TRUE, zd); |
428870ff BB |
6304 | |
6305 | ztest_zd_fini(zd); | |
34dc7c2f BB |
6306 | } |
6307 | ||
6308 | /* | |
6309 | * Kick off threads to run tests on all datasets in parallel. | |
6310 | */ | |
6311 | static void | |
428870ff | 6312 | ztest_run(ztest_shared_t *zs) |
34dc7c2f | 6313 | { |
34dc7c2f | 6314 | spa_t *spa; |
3541dc6d | 6315 | objset_t *os; |
1e33ac1e | 6316 | kthread_t *resume_thread; |
c25b8f99 | 6317 | kthread_t **run_threads; |
1e33ac1e | 6318 | uint64_t object; |
428870ff | 6319 | int error; |
d6320ddb | 6320 | int t, d; |
b128c09f BB |
6321 | |
6322 | ztest_exiting = B_FALSE; | |
34dc7c2f | 6323 | |
34dc7c2f | 6324 | /* |
428870ff | 6325 | * Initialize parent/child shared state. |
34dc7c2f | 6326 | */ |
c242c188 | 6327 | mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); |
7809eb8b | 6328 | VERIFY(rwlock_init(&ztest_name_lock, USYNC_THREAD, NULL) == 0); |
34dc7c2f | 6329 | |
428870ff | 6330 | zs->zs_thread_start = gethrtime(); |
c242c188 CS |
6331 | zs->zs_thread_stop = |
6332 | zs->zs_thread_start + ztest_opts.zo_passtime * NANOSEC; | |
428870ff BB |
6333 | zs->zs_thread_stop = MIN(zs->zs_thread_stop, zs->zs_proc_stop); |
6334 | zs->zs_thread_kill = zs->zs_thread_stop; | |
c242c188 CS |
6335 | if (ztest_random(100) < ztest_opts.zo_killrate) { |
6336 | zs->zs_thread_kill -= | |
6337 | ztest_random(ztest_opts.zo_passtime * NANOSEC); | |
6338 | } | |
34dc7c2f | 6339 | |
1e33ac1e | 6340 | mutex_init(&zcl.zcl_callbacks_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 6341 | |
428870ff BB |
6342 | list_create(&zcl.zcl_callbacks, sizeof (ztest_cb_data_t), |
6343 | offsetof(ztest_cb_data_t, zcd_node)); | |
34dc7c2f BB |
6344 | |
6345 | /* | |
b128c09f | 6346 | * Open our pool. |
34dc7c2f | 6347 | */ |
428870ff | 6348 | kernel_init(FREAD | FWRITE); |
13fe0198 | 6349 | VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
6d974228 | 6350 | spa->spa_debug = B_TRUE; |
080b3100 | 6351 | metaslab_preload_limit = ztest_random(20) + 1; |
c242c188 | 6352 | ztest_spa = spa; |
428870ff | 6353 | |
546d32ca | 6354 | dmu_objset_stats_t dds; |
13fe0198 | 6355 | VERIFY0(dmu_objset_own(ztest_opts.zo_pool, |
b5256303 | 6356 | DMU_OST_ANY, B_TRUE, B_TRUE, FTAG, &os)); |
546d32ca BB |
6357 | dsl_pool_config_enter(dmu_objset_pool(os), FTAG); |
6358 | dmu_objset_fast_stat(os, &dds); | |
6359 | dsl_pool_config_exit(dmu_objset_pool(os), FTAG); | |
6360 | zs->zs_guid = dds.dds_guid; | |
b5256303 | 6361 | dmu_objset_disown(os, B_TRUE, FTAG); |
3541dc6d | 6362 | |
428870ff | 6363 | spa->spa_dedup_ditto = 2 * ZIO_DEDUPDITTO_MIN; |
34dc7c2f | 6364 | |
fb5f0bc8 BB |
6365 | /* |
6366 | * We don't expect the pool to suspend unless maxfaults == 0, | |
6367 | * in which case ztest_fault_inject() temporarily takes away | |
6368 | * the only valid replica. | |
6369 | */ | |
428870ff | 6370 | if (MAXFAULTS() == 0) |
fb5f0bc8 BB |
6371 | spa->spa_failmode = ZIO_FAILURE_MODE_WAIT; |
6372 | else | |
6373 | spa->spa_failmode = ZIO_FAILURE_MODE_PANIC; | |
6374 | ||
34dc7c2f | 6375 | /* |
b128c09f | 6376 | * Create a thread to periodically resume suspended I/O. |
34dc7c2f | 6377 | */ |
c25b8f99 BB |
6378 | resume_thread = thread_create(NULL, 0, ztest_resume_thread, |
6379 | spa, 0, NULL, TS_RUN | TS_JOINABLE, defclsyspri); | |
34dc7c2f | 6380 | |
26099167 | 6381 | #if 0 |
428870ff | 6382 | /* |
1e33ac1e | 6383 | * Set a deadman alarm to abort() if we hang. |
428870ff | 6384 | */ |
1e33ac1e BB |
6385 | signal(SIGALRM, ztest_deadman_alarm); |
6386 | alarm((zs->zs_thread_stop - zs->zs_thread_start) / NANOSEC + GRACE); | |
26099167 | 6387 | #endif |
428870ff | 6388 | |
34dc7c2f BB |
6389 | /* |
6390 | * Verify that we can safely inquire about about any object, | |
6391 | * whether it's allocated or not. To make it interesting, | |
6392 | * we probe a 5-wide window around each power of two. | |
6393 | * This hits all edge cases, including zero and the max. | |
6394 | */ | |
d6320ddb BB |
6395 | for (t = 0; t < 64; t++) { |
6396 | for (d = -5; d <= 5; d++) { | |
34dc7c2f BB |
6397 | error = dmu_object_info(spa->spa_meta_objset, |
6398 | (1ULL << t) + d, NULL); | |
6399 | ASSERT(error == 0 || error == ENOENT || | |
6400 | error == EINVAL); | |
6401 | } | |
6402 | } | |
6403 | ||
6404 | /* | |
428870ff | 6405 | * If we got any ENOSPC errors on the previous run, destroy something. |
34dc7c2f | 6406 | */ |
428870ff | 6407 | if (zs->zs_enospc_count != 0) { |
c242c188 CS |
6408 | int d = ztest_random(ztest_opts.zo_datasets); |
6409 | ztest_dataset_destroy(d); | |
428870ff | 6410 | } |
34dc7c2f BB |
6411 | zs->zs_enospc_count = 0; |
6412 | ||
c25b8f99 | 6413 | run_threads = umem_zalloc(ztest_opts.zo_threads * sizeof (kthread_t *), |
c242c188 | 6414 | UMEM_NOFAIL); |
34dc7c2f | 6415 | |
c242c188 | 6416 | if (ztest_opts.zo_verbose >= 4) |
34dc7c2f BB |
6417 | (void) printf("starting main threads...\n"); |
6418 | ||
428870ff BB |
6419 | /* |
6420 | * Kick off all the tests that run in parallel. | |
6421 | */ | |
c242c188 | 6422 | for (t = 0; t < ztest_opts.zo_threads; t++) { |
c25b8f99 BB |
6423 | if (t < ztest_opts.zo_datasets && ztest_dataset_open(t) != 0) { |
6424 | umem_free(run_threads, ztest_opts.zo_threads * | |
6425 | sizeof (kthread_t *)); | |
428870ff | 6426 | return; |
06cf4d98 | 6427 | } |
1e33ac1e | 6428 | |
c25b8f99 BB |
6429 | run_threads[t] = thread_create(NULL, 0, ztest_thread, |
6430 | (void *)(uintptr_t)t, 0, NULL, TS_RUN | TS_JOINABLE, | |
6431 | defclsyspri); | |
34dc7c2f BB |
6432 | } |
6433 | ||
428870ff BB |
6434 | /* |
6435 | * Wait for all of the tests to complete. We go in reverse order | |
6436 | * so we don't close datasets while threads are still using them. | |
6437 | */ | |
c242c188 | 6438 | for (t = ztest_opts.zo_threads - 1; t >= 0; t--) { |
c25b8f99 | 6439 | VERIFY0(thread_join(run_threads[t])); |
c242c188 CS |
6440 | if (t < ztest_opts.zo_datasets) |
6441 | ztest_dataset_close(t); | |
34dc7c2f BB |
6442 | } |
6443 | ||
34dc7c2f BB |
6444 | txg_wait_synced(spa_get_dsl(spa), 0); |
6445 | ||
428870ff BB |
6446 | zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(spa)); |
6447 | zs->zs_space = metaslab_class_get_space(spa_normal_class(spa)); | |
6448 | ||
c25b8f99 | 6449 | umem_free(run_threads, ztest_opts.zo_threads * sizeof (kthread_t *)); |
428870ff BB |
6450 | |
6451 | /* Kill the resume thread */ | |
6452 | ztest_exiting = B_TRUE; | |
c25b8f99 | 6453 | VERIFY0(thread_join(resume_thread)); |
428870ff | 6454 | ztest_resume(spa); |
34dc7c2f BB |
6455 | |
6456 | /* | |
428870ff BB |
6457 | * Right before closing the pool, kick off a bunch of async I/O; |
6458 | * spa_close() should wait for it to complete. | |
34dc7c2f | 6459 | */ |
fcff0f35 PD |
6460 | for (object = 1; object < 50; object++) { |
6461 | dmu_prefetch(spa->spa_meta_objset, object, 0, 0, 1ULL << 20, | |
6462 | ZIO_PRIORITY_SYNC_READ); | |
6463 | } | |
428870ff | 6464 | |
090ff092 RC |
6465 | /* Verify that at least one commit cb was called in a timely fashion */ |
6466 | if (zc_cb_counter >= ZTEST_COMMIT_CB_MIN_REG) | |
c99c9001 | 6467 | VERIFY0(zc_min_txg_delay); |
090ff092 | 6468 | |
428870ff BB |
6469 | spa_close(spa, FTAG); |
6470 | ||
6471 | /* | |
6472 | * Verify that we can loop over all pools. | |
6473 | */ | |
6474 | mutex_enter(&spa_namespace_lock); | |
6475 | for (spa = spa_next(NULL); spa != NULL; spa = spa_next(spa)) | |
c242c188 | 6476 | if (ztest_opts.zo_verbose > 3) |
428870ff BB |
6477 | (void) printf("spa_next: found %s\n", spa_name(spa)); |
6478 | mutex_exit(&spa_namespace_lock); | |
6479 | ||
6480 | /* | |
6481 | * Verify that we can export the pool and reimport it under a | |
6482 | * different name. | |
6483 | */ | |
379ca9cf | 6484 | if ((ztest_random(2) == 0) && !ztest_opts.zo_mmp_test) { |
eca7b760 IK |
6485 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
6486 | (void) snprintf(name, sizeof (name), "%s_import", | |
c242c188 CS |
6487 | ztest_opts.zo_pool); |
6488 | ztest_spa_import_export(ztest_opts.zo_pool, name); | |
6489 | ztest_spa_import_export(name, ztest_opts.zo_pool); | |
428870ff BB |
6490 | } |
6491 | ||
6492 | kernel_fini(); | |
572e2857 BB |
6493 | |
6494 | list_destroy(&zcl.zcl_callbacks); | |
0e8d1b2d | 6495 | mutex_destroy(&zcl.zcl_callbacks_lock); |
7809eb8b | 6496 | (void) rwlock_destroy(&ztest_name_lock); |
c242c188 | 6497 | mutex_destroy(&ztest_vdev_lock); |
428870ff BB |
6498 | } |
6499 | ||
6500 | static void | |
c242c188 | 6501 | ztest_freeze(void) |
428870ff | 6502 | { |
c242c188 | 6503 | ztest_ds_t *zd = &ztest_ds[0]; |
428870ff BB |
6504 | spa_t *spa; |
6505 | int numloops = 0; | |
6506 | ||
c242c188 | 6507 | if (ztest_opts.zo_verbose >= 3) |
428870ff | 6508 | (void) printf("testing spa_freeze()...\n"); |
9babb374 | 6509 | |
428870ff | 6510 | kernel_init(FREAD | FWRITE); |
c242c188 CS |
6511 | VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
6512 | VERIFY3U(0, ==, ztest_dataset_open(0)); | |
03c6040b GW |
6513 | spa->spa_debug = B_TRUE; |
6514 | ztest_spa = spa; | |
9babb374 | 6515 | |
428870ff BB |
6516 | /* |
6517 | * Force the first log block to be transactionally allocated. | |
6518 | * We have to do this before we freeze the pool -- otherwise | |
6519 | * the log chain won't be anchored. | |
6520 | */ | |
6521 | while (BP_IS_HOLE(&zd->zd_zilog->zl_header->zh_log)) { | |
6522 | ztest_dmu_object_alloc_free(zd, 0); | |
572e2857 | 6523 | zil_commit(zd->zd_zilog, 0); |
34dc7c2f BB |
6524 | } |
6525 | ||
6526 | txg_wait_synced(spa_get_dsl(spa), 0); | |
6527 | ||
428870ff BB |
6528 | /* |
6529 | * Freeze the pool. This stops spa_sync() from doing anything, | |
6530 | * so that the only way to record changes from now on is the ZIL. | |
6531 | */ | |
6532 | spa_freeze(spa); | |
b128c09f | 6533 | |
b02fe35d AR |
6534 | /* |
6535 | * Because it is hard to predict how much space a write will actually | |
6536 | * require beforehand, we leave ourselves some fudge space to write over | |
6537 | * capacity. | |
6538 | */ | |
6539 | uint64_t capacity = metaslab_class_get_space(spa_normal_class(spa)) / 2; | |
6540 | ||
428870ff BB |
6541 | /* |
6542 | * Run tests that generate log records but don't alter the pool config | |
6543 | * or depend on DSL sync tasks (snapshots, objset create/destroy, etc). | |
6544 | * We do a txg_wait_synced() after each iteration to force the txg | |
6545 | * to increase well beyond the last synced value in the uberblock. | |
6546 | * The ZIL should be OK with that. | |
b02fe35d AR |
6547 | * |
6548 | * Run a random number of times less than zo_maxloops and ensure we do | |
6549 | * not run out of space on the pool. | |
428870ff | 6550 | */ |
c242c188 | 6551 | while (ztest_random(10) != 0 && |
b02fe35d AR |
6552 | numloops++ < ztest_opts.zo_maxloops && |
6553 | metaslab_class_get_alloc(spa_normal_class(spa)) < capacity) { | |
6554 | ztest_od_t od; | |
50c957f7 | 6555 | ztest_od_init(&od, 0, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0); |
b02fe35d AR |
6556 | VERIFY0(ztest_object_init(zd, &od, sizeof (od), B_FALSE)); |
6557 | ztest_io(zd, od.od_object, | |
6558 | ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
428870ff BB |
6559 | txg_wait_synced(spa_get_dsl(spa), 0); |
6560 | } | |
b128c09f | 6561 | |
34dc7c2f | 6562 | /* |
428870ff | 6563 | * Commit all of the changes we just generated. |
34dc7c2f | 6564 | */ |
572e2857 | 6565 | zil_commit(zd->zd_zilog, 0); |
428870ff | 6566 | txg_wait_synced(spa_get_dsl(spa), 0); |
34dc7c2f | 6567 | |
428870ff BB |
6568 | /* |
6569 | * Close our dataset and close the pool. | |
6570 | */ | |
c242c188 | 6571 | ztest_dataset_close(0); |
34dc7c2f | 6572 | spa_close(spa, FTAG); |
428870ff | 6573 | kernel_fini(); |
34dc7c2f | 6574 | |
428870ff BB |
6575 | /* |
6576 | * Open and close the pool and dataset to induce log replay. | |
6577 | */ | |
6578 | kernel_init(FREAD | FWRITE); | |
c242c188 | 6579 | VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
29809a6c | 6580 | ASSERT(spa_freeze_txg(spa) == UINT64_MAX); |
c242c188 CS |
6581 | VERIFY3U(0, ==, ztest_dataset_open(0)); |
6582 | ztest_dataset_close(0); | |
3bc7e0fb GW |
6583 | |
6584 | spa->spa_debug = B_TRUE; | |
6585 | ztest_spa = spa; | |
6586 | txg_wait_synced(spa_get_dsl(spa), 0); | |
6587 | ztest_reguid(NULL, 0); | |
6588 | ||
428870ff | 6589 | spa_close(spa, FTAG); |
34dc7c2f BB |
6590 | kernel_fini(); |
6591 | } | |
6592 | ||
6593 | void | |
6594 | print_time(hrtime_t t, char *timebuf) | |
6595 | { | |
6596 | hrtime_t s = t / NANOSEC; | |
6597 | hrtime_t m = s / 60; | |
6598 | hrtime_t h = m / 60; | |
6599 | hrtime_t d = h / 24; | |
6600 | ||
6601 | s -= m * 60; | |
6602 | m -= h * 60; | |
6603 | h -= d * 24; | |
6604 | ||
6605 | timebuf[0] = '\0'; | |
6606 | ||
6607 | if (d) | |
6608 | (void) sprintf(timebuf, | |
6609 | "%llud%02lluh%02llum%02llus", d, h, m, s); | |
6610 | else if (h) | |
6611 | (void) sprintf(timebuf, "%lluh%02llum%02llus", h, m, s); | |
6612 | else if (m) | |
6613 | (void) sprintf(timebuf, "%llum%02llus", m, s); | |
6614 | else | |
6615 | (void) sprintf(timebuf, "%llus", s); | |
6616 | } | |
6617 | ||
428870ff | 6618 | static nvlist_t * |
0bc8fd78 | 6619 | make_random_props(void) |
428870ff BB |
6620 | { |
6621 | nvlist_t *props; | |
6622 | ||
428870ff | 6623 | VERIFY(nvlist_alloc(&props, NV_UNIQUE_NAME, 0) == 0); |
9ae529ec CS |
6624 | if (ztest_random(2) == 0) |
6625 | return (props); | |
428870ff BB |
6626 | VERIFY(nvlist_add_uint64(props, "autoreplace", 1) == 0); |
6627 | ||
428870ff BB |
6628 | return (props); |
6629 | } | |
6630 | ||
379ca9cf OF |
6631 | /* |
6632 | * Import a storage pool with the given name. | |
6633 | */ | |
6634 | static void | |
6635 | ztest_import(ztest_shared_t *zs) | |
6636 | { | |
6637 | libzfs_handle_t *hdl; | |
6638 | importargs_t args = { 0 }; | |
6639 | spa_t *spa; | |
6640 | nvlist_t *cfg = NULL; | |
6641 | int nsearch = 1; | |
6642 | char *searchdirs[nsearch]; | |
6643 | char *name = ztest_opts.zo_pool; | |
6644 | int flags = ZFS_IMPORT_MISSING_LOG; | |
6645 | int error; | |
6646 | ||
6647 | mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); | |
6648 | VERIFY(rwlock_init(&ztest_name_lock, USYNC_THREAD, NULL) == 0); | |
6649 | ||
6650 | kernel_init(FREAD | FWRITE); | |
6651 | hdl = libzfs_init(); | |
6652 | ||
6653 | searchdirs[0] = ztest_opts.zo_dir; | |
6654 | args.paths = nsearch; | |
6655 | args.path = searchdirs; | |
6656 | args.can_be_active = B_FALSE; | |
6657 | ||
6658 | error = zpool_tryimport(hdl, name, &cfg, &args); | |
6659 | if (error) | |
6660 | (void) fatal(0, "No pools found\n"); | |
6661 | ||
6662 | VERIFY0(spa_import(name, cfg, NULL, flags)); | |
6663 | VERIFY0(spa_open(name, &spa, FTAG)); | |
6664 | zs->zs_metaslab_sz = | |
6665 | 1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift; | |
6666 | spa_close(spa, FTAG); | |
6667 | ||
6668 | libzfs_fini(hdl); | |
6669 | kernel_fini(); | |
6670 | ||
6671 | if (!ztest_opts.zo_mmp_test) { | |
6672 | ztest_run_zdb(ztest_opts.zo_pool); | |
6673 | ztest_freeze(); | |
6674 | ztest_run_zdb(ztest_opts.zo_pool); | |
6675 | } | |
6676 | ||
6677 | (void) rwlock_destroy(&ztest_name_lock); | |
6678 | mutex_destroy(&ztest_vdev_lock); | |
6679 | } | |
6680 | ||
34dc7c2f BB |
6681 | /* |
6682 | * Create a storage pool with the given name and initial vdev size. | |
428870ff | 6683 | * Then test spa_freeze() functionality. |
34dc7c2f BB |
6684 | */ |
6685 | static void | |
428870ff | 6686 | ztest_init(ztest_shared_t *zs) |
34dc7c2f BB |
6687 | { |
6688 | spa_t *spa; | |
428870ff | 6689 | nvlist_t *nvroot, *props; |
9ae529ec | 6690 | int i; |
428870ff | 6691 | |
c242c188 | 6692 | mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); |
7809eb8b | 6693 | VERIFY(rwlock_init(&ztest_name_lock, USYNC_THREAD, NULL) == 0); |
34dc7c2f BB |
6694 | |
6695 | kernel_init(FREAD | FWRITE); | |
6696 | ||
6697 | /* | |
6698 | * Create the storage pool. | |
6699 | */ | |
c242c188 | 6700 | (void) spa_destroy(ztest_opts.zo_pool); |
428870ff BB |
6701 | ztest_shared->zs_vdev_next_leaf = 0; |
6702 | zs->zs_splits = 0; | |
c242c188 | 6703 | zs->zs_mirrors = ztest_opts.zo_mirrors; |
ea0b2538 | 6704 | nvroot = make_vdev_root(NULL, NULL, NULL, ztest_opts.zo_vdev_size, 0, |
c242c188 | 6705 | 0, ztest_opts.zo_raidz, zs->zs_mirrors, 1); |
428870ff | 6706 | props = make_random_props(); |
9ae529ec CS |
6707 | for (i = 0; i < SPA_FEATURES; i++) { |
6708 | char *buf; | |
6709 | VERIFY3S(-1, !=, asprintf(&buf, "feature@%s", | |
6710 | spa_feature_table[i].fi_uname)); | |
6711 | VERIFY3U(0, ==, nvlist_add_uint64(props, buf, 0)); | |
6712 | free(buf); | |
6713 | } | |
b5256303 TC |
6714 | VERIFY3U(0, ==, |
6715 | spa_create(ztest_opts.zo_pool, nvroot, props, NULL, NULL)); | |
34dc7c2f | 6716 | nvlist_free(nvroot); |
85802aa4 | 6717 | nvlist_free(props); |
34dc7c2f | 6718 | |
c242c188 CS |
6719 | VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
6720 | zs->zs_metaslab_sz = | |
6721 | 1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift; | |
34dc7c2f BB |
6722 | spa_close(spa, FTAG); |
6723 | ||
6724 | kernel_fini(); | |
428870ff | 6725 | |
379ca9cf OF |
6726 | if (!ztest_opts.zo_mmp_test) { |
6727 | ztest_run_zdb(ztest_opts.zo_pool); | |
6728 | ztest_freeze(); | |
6729 | ztest_run_zdb(ztest_opts.zo_pool); | |
6730 | } | |
c242c188 | 6731 | |
7809eb8b | 6732 | (void) rwlock_destroy(&ztest_name_lock); |
c242c188 CS |
6733 | mutex_destroy(&ztest_vdev_lock); |
6734 | } | |
6735 | ||
6736 | static void | |
9d81146b | 6737 | setup_data_fd(void) |
c242c188 | 6738 | { |
facbbe43 BB |
6739 | static char ztest_name_data[] = "/tmp/ztest.data.XXXXXX"; |
6740 | ||
6741 | ztest_fd_data = mkstemp(ztest_name_data); | |
9d81146b | 6742 | ASSERT3S(ztest_fd_data, >=, 0); |
facbbe43 | 6743 | (void) unlink(ztest_name_data); |
c242c188 CS |
6744 | } |
6745 | ||
22257dc0 CS |
6746 | static int |
6747 | shared_data_size(ztest_shared_hdr_t *hdr) | |
6748 | { | |
6749 | int size; | |
6750 | ||
6751 | size = hdr->zh_hdr_size; | |
6752 | size += hdr->zh_opts_size; | |
6753 | size += hdr->zh_size; | |
6754 | size += hdr->zh_stats_size * hdr->zh_stats_count; | |
6755 | size += hdr->zh_ds_size * hdr->zh_ds_count; | |
6756 | ||
6757 | return (size); | |
6758 | } | |
6759 | ||
c242c188 CS |
6760 | static void |
6761 | setup_hdr(void) | |
6762 | { | |
22257dc0 | 6763 | int size; |
c242c188 CS |
6764 | ztest_shared_hdr_t *hdr; |
6765 | ||
6766 | hdr = (void *)mmap(0, P2ROUNDUP(sizeof (*hdr), getpagesize()), | |
9d81146b | 6767 | PROT_READ | PROT_WRITE, MAP_SHARED, ztest_fd_data, 0); |
799402d8 | 6768 | ASSERT(hdr != MAP_FAILED); |
c242c188 | 6769 | |
9d81146b | 6770 | VERIFY3U(0, ==, ftruncate(ztest_fd_data, sizeof (ztest_shared_hdr_t))); |
22257dc0 | 6771 | |
c242c188 CS |
6772 | hdr->zh_hdr_size = sizeof (ztest_shared_hdr_t); |
6773 | hdr->zh_opts_size = sizeof (ztest_shared_opts_t); | |
6774 | hdr->zh_size = sizeof (ztest_shared_t); | |
6775 | hdr->zh_stats_size = sizeof (ztest_shared_callstate_t); | |
6776 | hdr->zh_stats_count = ZTEST_FUNCS; | |
6777 | hdr->zh_ds_size = sizeof (ztest_shared_ds_t); | |
6778 | hdr->zh_ds_count = ztest_opts.zo_datasets; | |
6779 | ||
22257dc0 | 6780 | size = shared_data_size(hdr); |
9d81146b | 6781 | VERIFY3U(0, ==, ftruncate(ztest_fd_data, size)); |
22257dc0 | 6782 | |
c242c188 CS |
6783 | (void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize())); |
6784 | } | |
6785 | ||
6786 | static void | |
6787 | setup_data(void) | |
6788 | { | |
6789 | int size, offset; | |
6790 | ztest_shared_hdr_t *hdr; | |
6791 | uint8_t *buf; | |
6792 | ||
6793 | hdr = (void *)mmap(0, P2ROUNDUP(sizeof (*hdr), getpagesize()), | |
9d81146b | 6794 | PROT_READ, MAP_SHARED, ztest_fd_data, 0); |
799402d8 | 6795 | ASSERT(hdr != MAP_FAILED); |
c242c188 | 6796 | |
22257dc0 | 6797 | size = shared_data_size(hdr); |
c242c188 CS |
6798 | |
6799 | (void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize())); | |
6800 | hdr = ztest_shared_hdr = (void *)mmap(0, P2ROUNDUP(size, getpagesize()), | |
9d81146b | 6801 | PROT_READ | PROT_WRITE, MAP_SHARED, ztest_fd_data, 0); |
799402d8 | 6802 | ASSERT(hdr != MAP_FAILED); |
c242c188 CS |
6803 | buf = (uint8_t *)hdr; |
6804 | ||
6805 | offset = hdr->zh_hdr_size; | |
6806 | ztest_shared_opts = (void *)&buf[offset]; | |
6807 | offset += hdr->zh_opts_size; | |
6808 | ztest_shared = (void *)&buf[offset]; | |
6809 | offset += hdr->zh_size; | |
6810 | ztest_shared_callstate = (void *)&buf[offset]; | |
6811 | offset += hdr->zh_stats_size * hdr->zh_stats_count; | |
6812 | ztest_shared_ds = (void *)&buf[offset]; | |
6813 | } | |
6814 | ||
6815 | static boolean_t | |
6816 | exec_child(char *cmd, char *libpath, boolean_t ignorekill, int *statusp) | |
6817 | { | |
6818 | pid_t pid; | |
6819 | int status; | |
483106eb | 6820 | char *cmdbuf = NULL; |
c242c188 CS |
6821 | |
6822 | pid = fork(); | |
6823 | ||
6824 | if (cmd == NULL) { | |
483106eb BB |
6825 | cmdbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
6826 | (void) strlcpy(cmdbuf, getexecname(), MAXPATHLEN); | |
c242c188 CS |
6827 | cmd = cmdbuf; |
6828 | } | |
6829 | ||
6830 | if (pid == -1) | |
6831 | fatal(1, "fork failed"); | |
6832 | ||
6833 | if (pid == 0) { /* child */ | |
6834 | char *emptyargv[2] = { cmd, NULL }; | |
9d81146b | 6835 | char fd_data_str[12]; |
c242c188 CS |
6836 | |
6837 | struct rlimit rl = { 1024, 1024 }; | |
6838 | (void) setrlimit(RLIMIT_NOFILE, &rl); | |
9d81146b ED |
6839 | |
6840 | (void) close(ztest_fd_rand); | |
6841 | VERIFY(11 >= snprintf(fd_data_str, 12, "%d", ztest_fd_data)); | |
6842 | VERIFY(0 == setenv("ZTEST_FD_DATA", fd_data_str, 1)); | |
6843 | ||
c242c188 CS |
6844 | (void) enable_extended_FILE_stdio(-1, -1); |
6845 | if (libpath != NULL) | |
6846 | VERIFY(0 == setenv("LD_LIBRARY_PATH", libpath, 1)); | |
6847 | (void) execv(cmd, emptyargv); | |
6848 | ztest_dump_core = B_FALSE; | |
6849 | fatal(B_TRUE, "exec failed: %s", cmd); | |
6850 | } | |
6851 | ||
483106eb BB |
6852 | if (cmdbuf != NULL) { |
6853 | umem_free(cmdbuf, MAXPATHLEN); | |
6854 | cmd = NULL; | |
6855 | } | |
6856 | ||
c242c188 CS |
6857 | while (waitpid(pid, &status, 0) != pid) |
6858 | continue; | |
6859 | if (statusp != NULL) | |
6860 | *statusp = status; | |
6861 | ||
6862 | if (WIFEXITED(status)) { | |
6863 | if (WEXITSTATUS(status) != 0) { | |
6864 | (void) fprintf(stderr, "child exited with code %d\n", | |
6865 | WEXITSTATUS(status)); | |
6866 | exit(2); | |
6867 | } | |
6868 | return (B_FALSE); | |
6869 | } else if (WIFSIGNALED(status)) { | |
6870 | if (!ignorekill || WTERMSIG(status) != SIGKILL) { | |
6871 | (void) fprintf(stderr, "child died with signal %d\n", | |
6872 | WTERMSIG(status)); | |
6873 | exit(3); | |
6874 | } | |
6875 | return (B_TRUE); | |
6876 | } else { | |
6877 | (void) fprintf(stderr, "something strange happened to child\n"); | |
6878 | exit(4); | |
6879 | /* NOTREACHED */ | |
6880 | } | |
6881 | } | |
6882 | ||
6883 | static void | |
6884 | ztest_run_init(void) | |
6885 | { | |
6886 | int i; | |
428870ff | 6887 | |
c242c188 | 6888 | ztest_shared_t *zs = ztest_shared; |
428870ff | 6889 | |
c242c188 CS |
6890 | /* |
6891 | * Blow away any existing copy of zpool.cache | |
6892 | */ | |
6893 | (void) remove(spa_config_path); | |
6894 | ||
379ca9cf OF |
6895 | if (ztest_opts.zo_init == 0) { |
6896 | if (ztest_opts.zo_verbose >= 1) | |
6897 | (void) printf("Importing pool %s\n", | |
6898 | ztest_opts.zo_pool); | |
6899 | ztest_import(zs); | |
6900 | return; | |
6901 | } | |
6902 | ||
c242c188 CS |
6903 | /* |
6904 | * Create and initialize our storage pool. | |
6905 | */ | |
6906 | for (i = 1; i <= ztest_opts.zo_init; i++) { | |
6907 | bzero(zs, sizeof (ztest_shared_t)); | |
6908 | if (ztest_opts.zo_verbose >= 3 && | |
6909 | ztest_opts.zo_init != 1) { | |
6910 | (void) printf("ztest_init(), pass %d\n", i); | |
6911 | } | |
6912 | ztest_init(zs); | |
6913 | } | |
34dc7c2f BB |
6914 | } |
6915 | ||
6916 | int | |
6917 | main(int argc, char **argv) | |
6918 | { | |
6919 | int kills = 0; | |
6920 | int iters = 0; | |
c242c188 CS |
6921 | int older = 0; |
6922 | int newer = 0; | |
34dc7c2f BB |
6923 | ztest_shared_t *zs; |
6924 | ztest_info_t *zi; | |
c242c188 | 6925 | ztest_shared_callstate_t *zc; |
34dc7c2f BB |
6926 | char timebuf[100]; |
6927 | char numbuf[6]; | |
428870ff | 6928 | spa_t *spa; |
483106eb | 6929 | char *cmd; |
c242c188 CS |
6930 | boolean_t hasalt; |
6931 | int f; | |
9d81146b | 6932 | char *fd_data_str = getenv("ZTEST_FD_DATA"); |
e82cdc3a | 6933 | struct sigaction action; |
34dc7c2f BB |
6934 | |
6935 | (void) setvbuf(stdout, NULL, _IOLBF, 0); | |
6936 | ||
498877ba MA |
6937 | dprintf_setup(&argc, argv); |
6938 | ||
e82cdc3a NB |
6939 | action.sa_handler = sig_handler; |
6940 | sigemptyset(&action.sa_mask); | |
6941 | action.sa_flags = 0; | |
6942 | ||
6943 | if (sigaction(SIGSEGV, &action, NULL) < 0) { | |
6944 | (void) fprintf(stderr, "ztest: cannot catch SIGSEGV: %s.\n", | |
6945 | strerror(errno)); | |
6946 | exit(EXIT_FAILURE); | |
6947 | } | |
6948 | ||
6949 | if (sigaction(SIGABRT, &action, NULL) < 0) { | |
6950 | (void) fprintf(stderr, "ztest: cannot catch SIGABRT: %s.\n", | |
6951 | strerror(errno)); | |
6952 | exit(EXIT_FAILURE); | |
6953 | } | |
6954 | ||
9d81146b ED |
6955 | ztest_fd_rand = open("/dev/urandom", O_RDONLY); |
6956 | ASSERT3S(ztest_fd_rand, >=, 0); | |
6957 | ||
6958 | if (!fd_data_str) { | |
c242c188 | 6959 | process_options(argc, argv); |
34dc7c2f | 6960 | |
9d81146b | 6961 | setup_data_fd(); |
c242c188 CS |
6962 | setup_hdr(); |
6963 | setup_data(); | |
6964 | bcopy(&ztest_opts, ztest_shared_opts, | |
6965 | sizeof (*ztest_shared_opts)); | |
6966 | } else { | |
9d81146b | 6967 | ztest_fd_data = atoi(fd_data_str); |
c242c188 CS |
6968 | setup_data(); |
6969 | bcopy(ztest_shared_opts, &ztest_opts, sizeof (ztest_opts)); | |
6970 | } | |
6971 | ASSERT3U(ztest_opts.zo_datasets, ==, ztest_shared_hdr->zh_ds_count); | |
34dc7c2f | 6972 | |
428870ff | 6973 | /* Override location of zpool.cache */ |
5be98cfe BB |
6974 | VERIFY(asprintf((char **)&spa_config_path, "%s/zpool.cache", |
6975 | ztest_opts.zo_dir) != -1); | |
9d81146b | 6976 | |
c242c188 CS |
6977 | ztest_ds = umem_alloc(ztest_opts.zo_datasets * sizeof (ztest_ds_t), |
6978 | UMEM_NOFAIL); | |
6979 | zs = ztest_shared; | |
6980 | ||
9d81146b | 6981 | if (fd_data_str) { |
c242c188 CS |
6982 | metaslab_gang_bang = ztest_opts.zo_metaslab_gang_bang; |
6983 | metaslab_df_alloc_threshold = | |
6984 | zs->zs_metaslab_df_alloc_threshold; | |
6985 | ||
6986 | if (zs->zs_do_init) | |
6987 | ztest_run_init(); | |
6988 | else | |
6989 | ztest_run(zs); | |
6990 | exit(0); | |
6991 | } | |
34dc7c2f | 6992 | |
c242c188 | 6993 | hasalt = (strlen(ztest_opts.zo_alt_ztest) != 0); |
34dc7c2f | 6994 | |
c242c188 | 6995 | if (ztest_opts.zo_verbose >= 1) { |
34dc7c2f BB |
6996 | (void) printf("%llu vdevs, %d datasets, %d threads," |
6997 | " %llu seconds...\n", | |
c242c188 CS |
6998 | (u_longlong_t)ztest_opts.zo_vdevs, |
6999 | ztest_opts.zo_datasets, | |
7000 | ztest_opts.zo_threads, | |
7001 | (u_longlong_t)ztest_opts.zo_time); | |
34dc7c2f BB |
7002 | } |
7003 | ||
483106eb BB |
7004 | cmd = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); |
7005 | (void) strlcpy(cmd, getexecname(), MAXNAMELEN); | |
c242c188 CS |
7006 | |
7007 | zs->zs_do_init = B_TRUE; | |
7008 | if (strlen(ztest_opts.zo_alt_ztest) != 0) { | |
7009 | if (ztest_opts.zo_verbose >= 1) { | |
7010 | (void) printf("Executing older ztest for " | |
7011 | "initialization: %s\n", ztest_opts.zo_alt_ztest); | |
7012 | } | |
7013 | VERIFY(!exec_child(ztest_opts.zo_alt_ztest, | |
7014 | ztest_opts.zo_alt_libpath, B_FALSE, NULL)); | |
7015 | } else { | |
7016 | VERIFY(!exec_child(NULL, NULL, B_FALSE, NULL)); | |
34dc7c2f | 7017 | } |
c242c188 | 7018 | zs->zs_do_init = B_FALSE; |
34dc7c2f | 7019 | |
428870ff | 7020 | zs->zs_proc_start = gethrtime(); |
c242c188 | 7021 | zs->zs_proc_stop = zs->zs_proc_start + ztest_opts.zo_time * NANOSEC; |
34dc7c2f | 7022 | |
d6320ddb | 7023 | for (f = 0; f < ZTEST_FUNCS; f++) { |
c242c188 CS |
7024 | zi = &ztest_info[f]; |
7025 | zc = ZTEST_GET_SHARED_CALLSTATE(f); | |
428870ff | 7026 | if (zs->zs_proc_start + zi->zi_interval[0] > zs->zs_proc_stop) |
c242c188 | 7027 | zc->zc_next = UINT64_MAX; |
34dc7c2f | 7028 | else |
c242c188 | 7029 | zc->zc_next = zs->zs_proc_start + |
428870ff | 7030 | ztest_random(2 * zi->zi_interval[0] + 1); |
34dc7c2f BB |
7031 | } |
7032 | ||
34dc7c2f BB |
7033 | /* |
7034 | * Run the tests in a loop. These tests include fault injection | |
7035 | * to verify that self-healing data works, and forced crashes | |
7036 | * to verify that we never lose on-disk consistency. | |
7037 | */ | |
428870ff | 7038 | while (gethrtime() < zs->zs_proc_stop) { |
34dc7c2f | 7039 | int status; |
c242c188 | 7040 | boolean_t killed; |
34dc7c2f BB |
7041 | |
7042 | /* | |
7043 | * Initialize the workload counters for each function. | |
7044 | */ | |
d6320ddb | 7045 | for (f = 0; f < ZTEST_FUNCS; f++) { |
c242c188 CS |
7046 | zc = ZTEST_GET_SHARED_CALLSTATE(f); |
7047 | zc->zc_count = 0; | |
7048 | zc->zc_time = 0; | |
34dc7c2f BB |
7049 | } |
7050 | ||
9babb374 | 7051 | /* Set the allocation switch size */ |
c242c188 CS |
7052 | zs->zs_metaslab_df_alloc_threshold = |
7053 | ztest_random(zs->zs_metaslab_sz / 4) + 1; | |
9babb374 | 7054 | |
c242c188 CS |
7055 | if (!hasalt || ztest_random(2) == 0) { |
7056 | if (hasalt && ztest_opts.zo_verbose >= 1) { | |
7057 | (void) printf("Executing newer ztest: %s\n", | |
7058 | cmd); | |
34dc7c2f | 7059 | } |
c242c188 CS |
7060 | newer++; |
7061 | killed = exec_child(cmd, NULL, B_TRUE, &status); | |
34dc7c2f | 7062 | } else { |
c242c188 CS |
7063 | if (hasalt && ztest_opts.zo_verbose >= 1) { |
7064 | (void) printf("Executing older ztest: %s\n", | |
7065 | ztest_opts.zo_alt_ztest); | |
7066 | } | |
7067 | older++; | |
7068 | killed = exec_child(ztest_opts.zo_alt_ztest, | |
7069 | ztest_opts.zo_alt_libpath, B_TRUE, &status); | |
34dc7c2f BB |
7070 | } |
7071 | ||
c242c188 CS |
7072 | if (killed) |
7073 | kills++; | |
34dc7c2f BB |
7074 | iters++; |
7075 | ||
c242c188 | 7076 | if (ztest_opts.zo_verbose >= 1) { |
34dc7c2f BB |
7077 | hrtime_t now = gethrtime(); |
7078 | ||
428870ff BB |
7079 | now = MIN(now, zs->zs_proc_stop); |
7080 | print_time(zs->zs_proc_stop - now, timebuf); | |
34dc7c2f BB |
7081 | nicenum(zs->zs_space, numbuf); |
7082 | ||
7083 | (void) printf("Pass %3d, %8s, %3llu ENOSPC, " | |
7084 | "%4.1f%% of %5s used, %3.0f%% done, %8s to go\n", | |
7085 | iters, | |
7086 | WIFEXITED(status) ? "Complete" : "SIGKILL", | |
7087 | (u_longlong_t)zs->zs_enospc_count, | |
7088 | 100.0 * zs->zs_alloc / zs->zs_space, | |
7089 | numbuf, | |
428870ff | 7090 | 100.0 * (now - zs->zs_proc_start) / |
c242c188 | 7091 | (ztest_opts.zo_time * NANOSEC), timebuf); |
34dc7c2f BB |
7092 | } |
7093 | ||
c242c188 | 7094 | if (ztest_opts.zo_verbose >= 2) { |
34dc7c2f BB |
7095 | (void) printf("\nWorkload summary:\n\n"); |
7096 | (void) printf("%7s %9s %s\n", | |
7097 | "Calls", "Time", "Function"); | |
7098 | (void) printf("%7s %9s %s\n", | |
7099 | "-----", "----", "--------"); | |
d6320ddb | 7100 | for (f = 0; f < ZTEST_FUNCS; f++) { |
c242c188 CS |
7101 | zi = &ztest_info[f]; |
7102 | zc = ZTEST_GET_SHARED_CALLSTATE(f); | |
7103 | print_time(zc->zc_time, timebuf); | |
34dc7c2f | 7104 | (void) printf("%7llu %9s %s\n", |
c242c188 | 7105 | (u_longlong_t)zc->zc_count, timebuf, |
fdc5d982 | 7106 | zi->zi_funcname); |
34dc7c2f BB |
7107 | } |
7108 | (void) printf("\n"); | |
7109 | } | |
7110 | ||
7111 | /* | |
428870ff BB |
7112 | * It's possible that we killed a child during a rename test, |
7113 | * in which case we'll have a 'ztest_tmp' pool lying around | |
7114 | * instead of 'ztest'. Do a blind rename in case this happened. | |
34dc7c2f | 7115 | */ |
428870ff | 7116 | kernel_init(FREAD); |
c242c188 | 7117 | if (spa_open(ztest_opts.zo_pool, &spa, FTAG) == 0) { |
428870ff BB |
7118 | spa_close(spa, FTAG); |
7119 | } else { | |
eca7b760 | 7120 | char tmpname[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
7121 | kernel_fini(); |
7122 | kernel_init(FREAD | FWRITE); | |
7123 | (void) snprintf(tmpname, sizeof (tmpname), "%s_tmp", | |
c242c188 CS |
7124 | ztest_opts.zo_pool); |
7125 | (void) spa_rename(tmpname, ztest_opts.zo_pool); | |
428870ff | 7126 | } |
34dc7c2f | 7127 | kernel_fini(); |
34dc7c2f | 7128 | |
379ca9cf OF |
7129 | if (!ztest_opts.zo_mmp_test) |
7130 | ztest_run_zdb(ztest_opts.zo_pool); | |
428870ff | 7131 | } |
34dc7c2f | 7132 | |
c242c188 CS |
7133 | if (ztest_opts.zo_verbose >= 1) { |
7134 | if (hasalt) { | |
7135 | (void) printf("%d runs of older ztest: %s\n", older, | |
7136 | ztest_opts.zo_alt_ztest); | |
7137 | (void) printf("%d runs of newer ztest: %s\n", newer, | |
7138 | cmd); | |
7139 | } | |
34dc7c2f BB |
7140 | (void) printf("%d killed, %d completed, %.0f%% kill rate\n", |
7141 | kills, iters - kills, (100.0 * kills) / MAX(1, iters)); | |
7142 | } | |
7143 | ||
483106eb BB |
7144 | umem_free(cmd, MAXNAMELEN); |
7145 | ||
34dc7c2f BB |
7146 | return (0); |
7147 | } |