]>
Commit | Line | Data |
---|---|---|
34dc7c2f BB |
1 | /* |
2 | * CDDL HEADER START | |
3 | * | |
4 | * The contents of this file are subject to the terms of the | |
5 | * Common Development and Distribution License (the "License"). | |
6 | * You may not use this file except in compliance with the License. | |
7 | * | |
8 | * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE | |
9 | * or http://www.opensolaris.org/os/licensing. | |
10 | * See the License for the specific language governing permissions | |
11 | * and limitations under the License. | |
12 | * | |
13 | * When distributing Covered Code, include this CDDL HEADER in each | |
14 | * file and include the License file at usr/src/OPENSOLARIS.LICENSE. | |
15 | * If applicable, add the following below this CDDL HEADER, with the | |
16 | * fields enclosed by brackets "[]" replaced with your own identifying | |
17 | * information: Portions Copyright [yyyy] [name of copyright owner] | |
18 | * | |
19 | * CDDL HEADER END | |
20 | */ | |
21 | /* | |
428870ff | 22 | * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. |
cc99f275 | 23 | * Copyright (c) 2011, 2018 by Delphix. All rights reserved. |
3541dc6d | 24 | * Copyright 2011 Nexenta Systems, Inc. All rights reserved. |
95fd54a1 | 25 | * Copyright (c) 2013 Steven Hartland. All rights reserved. |
f3c8c9e6 JK |
26 | * Copyright (c) 2014 Integros [integros.com] |
27 | * Copyright 2017 Joyent, Inc. | |
cc99f275 | 28 | * Copyright (c) 2017, Intel Corporation. |
34dc7c2f BB |
29 | */ |
30 | ||
34dc7c2f BB |
31 | /* |
32 | * The objective of this program is to provide a DMU/ZAP/SPA stress test | |
33 | * that runs entirely in userland, is easy to use, and easy to extend. | |
34 | * | |
35 | * The overall design of the ztest program is as follows: | |
36 | * | |
37 | * (1) For each major functional area (e.g. adding vdevs to a pool, | |
38 | * creating and destroying datasets, reading and writing objects, etc) | |
39 | * we have a simple routine to test that functionality. These | |
40 | * individual routines do not have to do anything "stressful". | |
41 | * | |
42 | * (2) We turn these simple functionality tests into a stress test by | |
43 | * running them all in parallel, with as many threads as desired, | |
44 | * and spread across as many datasets, objects, and vdevs as desired. | |
45 | * | |
46 | * (3) While all this is happening, we inject faults into the pool to | |
47 | * verify that self-healing data really works. | |
48 | * | |
49 | * (4) Every time we open a dataset, we change its checksum and compression | |
50 | * functions. Thus even individual objects vary from block to block | |
51 | * in which checksum they use and whether they're compressed. | |
52 | * | |
53 | * (5) To verify that we never lose on-disk consistency after a crash, | |
54 | * we run the entire test in a child of the main process. | |
55 | * At random times, the child self-immolates with a SIGKILL. | |
56 | * This is the software equivalent of pulling the power cord. | |
57 | * The parent then runs the test again, using the existing | |
9b67f605 | 58 | * storage pool, as many times as desired. If backwards compatibility |
c242c188 CS |
59 | * testing is enabled ztest will sometimes run the "older" version |
60 | * of ztest after a SIGKILL. | |
34dc7c2f BB |
61 | * |
62 | * (6) To verify that we don't have future leaks or temporal incursions, | |
63 | * many of the functional tests record the transaction group number | |
64 | * as part of their data. When reading old data, they verify that | |
65 | * the transaction group number is less than the current, open txg. | |
66 | * If you add a new test, please do this if applicable. | |
67 | * | |
1e33ac1e BB |
68 | * (7) Threads are created with a reduced stack size, for sanity checking. |
69 | * Therefore, it's important not to allocate huge buffers on the stack. | |
70 | * | |
34dc7c2f BB |
71 | * When run with no arguments, ztest runs for about five minutes and |
72 | * produces no output if successful. To get a little bit of information, | |
73 | * specify -V. To get more information, specify -VV, and so on. | |
74 | * | |
75 | * To turn this into an overnight stress test, use -T to specify run time. | |
76 | * | |
ad0b23b1 | 77 | * You can ask more vdevs [-v], datasets [-d], or threads [-t] |
34dc7c2f BB |
78 | * to increase the pool capacity, fanout, and overall stress level. |
79 | * | |
c242c188 CS |
80 | * Use the -k option to set the desired frequency of kills. |
81 | * | |
82 | * When ztest invokes itself it passes all relevant information through a | |
83 | * temporary file which is mmap-ed in the child process. This allows shared | |
84 | * memory to survive the exec syscall. The ztest_shared_hdr_t struct is always | |
85 | * stored at offset 0 of this file and contains information on the size and | |
86 | * number of shared structures in the file. The information stored in this file | |
87 | * must remain backwards compatible with older versions of ztest so that | |
88 | * ztest can invoke them during backwards compatibility testing (-B). | |
34dc7c2f BB |
89 | */ |
90 | ||
91 | #include <sys/zfs_context.h> | |
92 | #include <sys/spa.h> | |
93 | #include <sys/dmu.h> | |
94 | #include <sys/txg.h> | |
9babb374 | 95 | #include <sys/dbuf.h> |
34dc7c2f | 96 | #include <sys/zap.h> |
34dc7c2f BB |
97 | #include <sys/dmu_objset.h> |
98 | #include <sys/poll.h> | |
99 | #include <sys/stat.h> | |
100 | #include <sys/time.h> | |
101 | #include <sys/wait.h> | |
102 | #include <sys/mman.h> | |
103 | #include <sys/resource.h> | |
104 | #include <sys/zio.h> | |
34dc7c2f | 105 | #include <sys/zil.h> |
428870ff | 106 | #include <sys/zil_impl.h> |
b2255edc | 107 | #include <sys/vdev_draid.h> |
34dc7c2f | 108 | #include <sys/vdev_impl.h> |
b128c09f | 109 | #include <sys/vdev_file.h> |
619f0976 | 110 | #include <sys/vdev_initialize.h> |
e5db3134 | 111 | #include <sys/vdev_raidz.h> |
1b939560 | 112 | #include <sys/vdev_trim.h> |
34dc7c2f | 113 | #include <sys/spa_impl.h> |
428870ff | 114 | #include <sys/metaslab_impl.h> |
34dc7c2f | 115 | #include <sys/dsl_prop.h> |
9babb374 | 116 | #include <sys/dsl_dataset.h> |
13fe0198 | 117 | #include <sys/dsl_destroy.h> |
428870ff BB |
118 | #include <sys/dsl_scan.h> |
119 | #include <sys/zio_checksum.h> | |
27d96d22 | 120 | #include <sys/zfs_refcount.h> |
9ae529ec | 121 | #include <sys/zfeature.h> |
13fe0198 | 122 | #include <sys/dsl_userhold.h> |
a6255b7f | 123 | #include <sys/abd.h> |
34dc7c2f | 124 | #include <stdio.h> |
34dc7c2f BB |
125 | #include <stdlib.h> |
126 | #include <unistd.h> | |
127 | #include <signal.h> | |
128 | #include <umem.h> | |
34dc7c2f BB |
129 | #include <ctype.h> |
130 | #include <math.h> | |
131 | #include <sys/fs/zfs.h> | |
1eeb4562 | 132 | #include <zfs_fletcher.h> |
428870ff | 133 | #include <libnvpair.h> |
e89f1295 | 134 | #include <libzutil.h> |
010d1247 | 135 | #include <sys/crypto/icp.h> |
aef1830f | 136 | #if (__GLIBC__ && !__UCLIBC__) |
e82cdc3a NB |
137 | #include <execinfo.h> /* for backtrace() */ |
138 | #endif | |
34dc7c2f | 139 | |
9d81146b ED |
140 | static int ztest_fd_data = -1; |
141 | static int ztest_fd_rand = -1; | |
c242c188 CS |
142 | |
143 | typedef struct ztest_shared_hdr { | |
144 | uint64_t zh_hdr_size; | |
145 | uint64_t zh_opts_size; | |
146 | uint64_t zh_size; | |
147 | uint64_t zh_stats_size; | |
148 | uint64_t zh_stats_count; | |
149 | uint64_t zh_ds_size; | |
150 | uint64_t zh_ds_count; | |
151 | } ztest_shared_hdr_t; | |
152 | ||
153 | static ztest_shared_hdr_t *ztest_shared_hdr; | |
154 | ||
cc99f275 DB |
155 | enum ztest_class_state { |
156 | ZTEST_VDEV_CLASS_OFF, | |
157 | ZTEST_VDEV_CLASS_ON, | |
158 | ZTEST_VDEV_CLASS_RND | |
159 | }; | |
160 | ||
2801e68d CS |
161 | #define ZO_GVARS_MAX_ARGLEN ((size_t)64) |
162 | #define ZO_GVARS_MAX_COUNT ((size_t)10) | |
163 | ||
c242c188 | 164 | typedef struct ztest_shared_opts { |
eca7b760 IK |
165 | char zo_pool[ZFS_MAX_DATASET_NAME_LEN]; |
166 | char zo_dir[ZFS_MAX_DATASET_NAME_LEN]; | |
c242c188 CS |
167 | char zo_alt_ztest[MAXNAMELEN]; |
168 | char zo_alt_libpath[MAXNAMELEN]; | |
169 | uint64_t zo_vdevs; | |
170 | uint64_t zo_vdevtime; | |
171 | size_t zo_vdev_size; | |
172 | int zo_ashift; | |
173 | int zo_mirrors; | |
b2255edc BB |
174 | int zo_raid_children; |
175 | int zo_raid_parity; | |
176 | char zo_raid_type[8]; | |
177 | int zo_draid_data; | |
178 | int zo_draid_spares; | |
c242c188 CS |
179 | int zo_datasets; |
180 | int zo_threads; | |
181 | uint64_t zo_passtime; | |
182 | uint64_t zo_killrate; | |
183 | int zo_verbose; | |
184 | int zo_init; | |
185 | uint64_t zo_time; | |
186 | uint64_t zo_maxloops; | |
d830d479 | 187 | uint64_t zo_metaslab_force_ganging; |
379ca9cf | 188 | int zo_mmp_test; |
cc99f275 | 189 | int zo_special_vdevs; |
ab4c009e | 190 | int zo_dump_dbgmsg; |
2801e68d CS |
191 | int zo_gvars_count; |
192 | char zo_gvars[ZO_GVARS_MAX_COUNT][ZO_GVARS_MAX_ARGLEN]; | |
c242c188 CS |
193 | } ztest_shared_opts_t; |
194 | ||
195 | static const ztest_shared_opts_t ztest_opts_defaults = { | |
caacc6e4 TH |
196 | .zo_pool = "ztest", |
197 | .zo_dir = "/tmp", | |
c242c188 CS |
198 | .zo_alt_ztest = { '\0' }, |
199 | .zo_alt_libpath = { '\0' }, | |
200 | .zo_vdevs = 5, | |
201 | .zo_ashift = SPA_MINBLOCKSHIFT, | |
202 | .zo_mirrors = 2, | |
b2255edc BB |
203 | .zo_raid_children = 4, |
204 | .zo_raid_parity = 1, | |
205 | .zo_raid_type = VDEV_TYPE_RAIDZ, | |
4e21fd06 | 206 | .zo_vdev_size = SPA_MINDEVSIZE * 4, /* 256m default size */ |
b2255edc BB |
207 | .zo_draid_data = 4, /* data drives */ |
208 | .zo_draid_spares = 1, /* distributed spares */ | |
c242c188 CS |
209 | .zo_datasets = 7, |
210 | .zo_threads = 23, | |
211 | .zo_passtime = 60, /* 60 seconds */ | |
212 | .zo_killrate = 70, /* 70% kill rate */ | |
213 | .zo_verbose = 0, | |
379ca9cf | 214 | .zo_mmp_test = 0, |
c242c188 CS |
215 | .zo_init = 1, |
216 | .zo_time = 300, /* 5 minutes */ | |
217 | .zo_maxloops = 50, /* max loops during spa_freeze() */ | |
09b85f2d | 218 | .zo_metaslab_force_ganging = 64 << 10, |
cc99f275 | 219 | .zo_special_vdevs = ZTEST_VDEV_CLASS_RND, |
2801e68d | 220 | .zo_gvars_count = 0, |
c242c188 CS |
221 | }; |
222 | ||
d830d479 | 223 | extern uint64_t metaslab_force_ganging; |
c242c188 | 224 | extern uint64_t metaslab_df_alloc_threshold; |
8fb1ede1 | 225 | extern unsigned long zfs_deadman_synctime_ms; |
080b3100 | 226 | extern int metaslab_preload_limit; |
d3c2ae1c | 227 | extern boolean_t zfs_compressed_arc_enabled; |
047116ac TC |
228 | extern int zfs_abd_scatter_enabled; |
229 | extern int dmu_object_alloc_chunk_shift; | |
4d044c4c | 230 | extern boolean_t zfs_force_some_double_word_sm_entries; |
c3bd3fb4 | 231 | extern unsigned long zio_decompress_fail_fraction; |
1258bd77 | 232 | extern unsigned long zfs_reconstruct_indirect_damage_fraction; |
65ca2c1e | 233 | |
c242c188 CS |
234 | |
235 | static ztest_shared_opts_t *ztest_shared_opts; | |
236 | static ztest_shared_opts_t ztest_opts; | |
4807c0ba | 237 | static char *ztest_wkeydata = "abcdefghijklmnopqrstuvwxyz012345"; |
c242c188 CS |
238 | |
239 | typedef struct ztest_shared_ds { | |
240 | uint64_t zd_seq; | |
241 | } ztest_shared_ds_t; | |
242 | ||
243 | static ztest_shared_ds_t *ztest_shared_ds; | |
244 | #define ZTEST_GET_SHARED_DS(d) (&ztest_shared_ds[d]) | |
428870ff BB |
245 | |
246 | #define BT_MAGIC 0x123456789abcdefULL | |
60b82074 | 247 | #define MAXFAULTS(zs) \ |
b2255edc | 248 | (MAX((zs)->zs_mirrors, 1) * (ztest_opts.zo_raid_parity + 1) - 1) |
428870ff BB |
249 | |
250 | enum ztest_io_type { | |
251 | ZTEST_IO_WRITE_TAG, | |
252 | ZTEST_IO_WRITE_PATTERN, | |
253 | ZTEST_IO_WRITE_ZEROES, | |
254 | ZTEST_IO_TRUNCATE, | |
255 | ZTEST_IO_SETATTR, | |
03c6040b | 256 | ZTEST_IO_REWRITE, |
428870ff BB |
257 | ZTEST_IO_TYPES |
258 | }; | |
34dc7c2f BB |
259 | |
260 | typedef struct ztest_block_tag { | |
428870ff | 261 | uint64_t bt_magic; |
34dc7c2f BB |
262 | uint64_t bt_objset; |
263 | uint64_t bt_object; | |
50c957f7 | 264 | uint64_t bt_dnodesize; |
34dc7c2f | 265 | uint64_t bt_offset; |
428870ff | 266 | uint64_t bt_gen; |
34dc7c2f | 267 | uint64_t bt_txg; |
428870ff | 268 | uint64_t bt_crtxg; |
34dc7c2f BB |
269 | } ztest_block_tag_t; |
270 | ||
428870ff BB |
271 | typedef struct bufwad { |
272 | uint64_t bw_index; | |
273 | uint64_t bw_txg; | |
274 | uint64_t bw_data; | |
275 | } bufwad_t; | |
276 | ||
5d43cc9a MA |
277 | /* |
278 | * It would be better to use a rangelock_t per object. Unfortunately | |
279 | * the rangelock_t is not a drop-in replacement for rl_t, because we | |
280 | * still need to map from object ID to rangelock_t. | |
281 | */ | |
282 | typedef enum { | |
283 | RL_READER, | |
284 | RL_WRITER, | |
285 | RL_APPEND | |
286 | } rl_type_t; | |
287 | ||
428870ff BB |
288 | typedef struct rll { |
289 | void *rll_writer; | |
290 | int rll_readers; | |
1e33ac1e BB |
291 | kmutex_t rll_lock; |
292 | kcondvar_t rll_cv; | |
428870ff BB |
293 | } rll_t; |
294 | ||
5d43cc9a MA |
295 | typedef struct rl { |
296 | uint64_t rl_object; | |
297 | uint64_t rl_offset; | |
298 | uint64_t rl_size; | |
299 | rll_t *rl_lock; | |
300 | } rl_t; | |
428870ff BB |
301 | |
302 | #define ZTEST_RANGE_LOCKS 64 | |
303 | #define ZTEST_OBJECT_LOCKS 64 | |
304 | ||
305 | /* | |
306 | * Object descriptor. Used as a template for object lookup/create/remove. | |
307 | */ | |
308 | typedef struct ztest_od { | |
309 | uint64_t od_dir; | |
310 | uint64_t od_object; | |
311 | dmu_object_type_t od_type; | |
312 | dmu_object_type_t od_crtype; | |
313 | uint64_t od_blocksize; | |
314 | uint64_t od_crblocksize; | |
50c957f7 | 315 | uint64_t od_crdnodesize; |
428870ff BB |
316 | uint64_t od_gen; |
317 | uint64_t od_crgen; | |
eca7b760 | 318 | char od_name[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff | 319 | } ztest_od_t; |
34dc7c2f | 320 | |
428870ff BB |
321 | /* |
322 | * Per-dataset state. | |
323 | */ | |
324 | typedef struct ztest_ds { | |
c242c188 | 325 | ztest_shared_ds_t *zd_shared; |
428870ff | 326 | objset_t *zd_os; |
2d9142c9 | 327 | pthread_rwlock_t zd_zilog_lock; |
428870ff | 328 | zilog_t *zd_zilog; |
428870ff | 329 | ztest_od_t *zd_od; /* debugging aid */ |
eca7b760 | 330 | char zd_name[ZFS_MAX_DATASET_NAME_LEN]; |
1e33ac1e | 331 | kmutex_t zd_dirobj_lock; |
428870ff | 332 | rll_t zd_object_lock[ZTEST_OBJECT_LOCKS]; |
5d43cc9a | 333 | rll_t zd_range_lock[ZTEST_RANGE_LOCKS]; |
428870ff BB |
334 | } ztest_ds_t; |
335 | ||
336 | /* | |
337 | * Per-iteration state. | |
338 | */ | |
339 | typedef void ztest_func_t(ztest_ds_t *zd, uint64_t id); | |
340 | ||
341 | typedef struct ztest_info { | |
342 | ztest_func_t *zi_func; /* test function */ | |
343 | uint64_t zi_iters; /* iterations per execution */ | |
344 | uint64_t *zi_interval; /* execute every <interval> seconds */ | |
fdc5d982 | 345 | const char *zi_funcname; /* name of test function */ |
428870ff | 346 | } ztest_info_t; |
34dc7c2f | 347 | |
c242c188 CS |
348 | typedef struct ztest_shared_callstate { |
349 | uint64_t zc_count; /* per-pass count */ | |
350 | uint64_t zc_time; /* per-pass time */ | |
351 | uint64_t zc_next; /* next time to call this function */ | |
352 | } ztest_shared_callstate_t; | |
353 | ||
354 | static ztest_shared_callstate_t *ztest_shared_callstate; | |
355 | #define ZTEST_GET_SHARED_CALLSTATE(c) (&ztest_shared_callstate[c]) | |
356 | ||
34dc7c2f BB |
357 | ztest_func_t ztest_dmu_read_write; |
358 | ztest_func_t ztest_dmu_write_parallel; | |
359 | ztest_func_t ztest_dmu_object_alloc_free; | |
047116ac | 360 | ztest_func_t ztest_dmu_object_next_chunk; |
428870ff | 361 | ztest_func_t ztest_dmu_commit_callbacks; |
34dc7c2f BB |
362 | ztest_func_t ztest_zap; |
363 | ztest_func_t ztest_zap_parallel; | |
428870ff | 364 | ztest_func_t ztest_zil_commit; |
3e31d2b0 | 365 | ztest_func_t ztest_zil_remount; |
428870ff | 366 | ztest_func_t ztest_dmu_read_write_zcopy; |
34dc7c2f | 367 | ztest_func_t ztest_dmu_objset_create_destroy; |
428870ff BB |
368 | ztest_func_t ztest_dmu_prealloc; |
369 | ztest_func_t ztest_fzap; | |
34dc7c2f | 370 | ztest_func_t ztest_dmu_snapshot_create_destroy; |
428870ff BB |
371 | ztest_func_t ztest_dsl_prop_get_set; |
372 | ztest_func_t ztest_spa_prop_get_set; | |
34dc7c2f BB |
373 | ztest_func_t ztest_spa_create_destroy; |
374 | ztest_func_t ztest_fault_inject; | |
428870ff | 375 | ztest_func_t ztest_dmu_snapshot_hold; |
0582e403 | 376 | ztest_func_t ztest_mmp_enable_disable; |
428870ff BB |
377 | ztest_func_t ztest_scrub; |
378 | ztest_func_t ztest_dsl_dataset_promote_busy; | |
34dc7c2f BB |
379 | ztest_func_t ztest_vdev_attach_detach; |
380 | ztest_func_t ztest_vdev_LUN_growth; | |
381 | ztest_func_t ztest_vdev_add_remove; | |
cc99f275 | 382 | ztest_func_t ztest_vdev_class_add; |
b128c09f | 383 | ztest_func_t ztest_vdev_aux_add_remove; |
428870ff | 384 | ztest_func_t ztest_split_pool; |
3541dc6d | 385 | ztest_func_t ztest_reguid; |
ea0b2538 | 386 | ztest_func_t ztest_spa_upgrade; |
a1d477c2 | 387 | ztest_func_t ztest_device_removal; |
d2734cce | 388 | ztest_func_t ztest_spa_checkpoint_create_discard; |
619f0976 | 389 | ztest_func_t ztest_initialize; |
1b939560 | 390 | ztest_func_t ztest_trim; |
1eeb4562 | 391 | ztest_func_t ztest_fletcher; |
37f520db | 392 | ztest_func_t ztest_fletcher_incr; |
50c957f7 | 393 | ztest_func_t ztest_verify_dnode_bt; |
34dc7c2f | 394 | |
428870ff BB |
395 | uint64_t zopt_always = 0ULL * NANOSEC; /* all the time */ |
396 | uint64_t zopt_incessant = 1ULL * NANOSEC / 10; /* every 1/10 second */ | |
397 | uint64_t zopt_often = 1ULL * NANOSEC; /* every second */ | |
398 | uint64_t zopt_sometimes = 10ULL * NANOSEC; /* every 10 seconds */ | |
399 | uint64_t zopt_rarely = 60ULL * NANOSEC; /* every 60 seconds */ | |
34dc7c2f | 400 | |
fdc5d982 TC |
401 | #define ZTI_INIT(func, iters, interval) \ |
402 | { .zi_func = (func), \ | |
403 | .zi_iters = (iters), \ | |
404 | .zi_interval = (interval), \ | |
405 | .zi_funcname = # func } | |
406 | ||
34dc7c2f | 407 | ztest_info_t ztest_info[] = { |
fdc5d982 TC |
408 | ZTI_INIT(ztest_dmu_read_write, 1, &zopt_always), |
409 | ZTI_INIT(ztest_dmu_write_parallel, 10, &zopt_always), | |
410 | ZTI_INIT(ztest_dmu_object_alloc_free, 1, &zopt_always), | |
047116ac | 411 | ZTI_INIT(ztest_dmu_object_next_chunk, 1, &zopt_sometimes), |
fdc5d982 TC |
412 | ZTI_INIT(ztest_dmu_commit_callbacks, 1, &zopt_always), |
413 | ZTI_INIT(ztest_zap, 30, &zopt_always), | |
414 | ZTI_INIT(ztest_zap_parallel, 100, &zopt_always), | |
415 | ZTI_INIT(ztest_split_pool, 1, &zopt_always), | |
416 | ZTI_INIT(ztest_zil_commit, 1, &zopt_incessant), | |
417 | ZTI_INIT(ztest_zil_remount, 1, &zopt_sometimes), | |
418 | ZTI_INIT(ztest_dmu_read_write_zcopy, 1, &zopt_often), | |
419 | ZTI_INIT(ztest_dmu_objset_create_destroy, 1, &zopt_often), | |
420 | ZTI_INIT(ztest_dsl_prop_get_set, 1, &zopt_often), | |
421 | ZTI_INIT(ztest_spa_prop_get_set, 1, &zopt_sometimes), | |
428870ff | 422 | #if 0 |
fdc5d982 | 423 | ZTI_INIT(ztest_dmu_prealloc, 1, &zopt_sometimes), |
428870ff | 424 | #endif |
fdc5d982 TC |
425 | ZTI_INIT(ztest_fzap, 1, &zopt_sometimes), |
426 | ZTI_INIT(ztest_dmu_snapshot_create_destroy, 1, &zopt_sometimes), | |
427 | ZTI_INIT(ztest_spa_create_destroy, 1, &zopt_sometimes), | |
428 | ZTI_INIT(ztest_fault_inject, 1, &zopt_sometimes), | |
fdc5d982 | 429 | ZTI_INIT(ztest_dmu_snapshot_hold, 1, &zopt_sometimes), |
0582e403 | 430 | ZTI_INIT(ztest_mmp_enable_disable, 1, &zopt_sometimes), |
fdc5d982 | 431 | ZTI_INIT(ztest_reguid, 1, &zopt_rarely), |
fdc5d982 TC |
432 | ZTI_INIT(ztest_scrub, 1, &zopt_rarely), |
433 | ZTI_INIT(ztest_spa_upgrade, 1, &zopt_rarely), | |
434 | ZTI_INIT(ztest_dsl_dataset_promote_busy, 1, &zopt_rarely), | |
435 | ZTI_INIT(ztest_vdev_attach_detach, 1, &zopt_sometimes), | |
436 | ZTI_INIT(ztest_vdev_LUN_growth, 1, &zopt_rarely), | |
437 | ZTI_INIT(ztest_vdev_add_remove, 1, &ztest_opts.zo_vdevtime), | |
cc99f275 | 438 | ZTI_INIT(ztest_vdev_class_add, 1, &ztest_opts.zo_vdevtime), |
fdc5d982 | 439 | ZTI_INIT(ztest_vdev_aux_add_remove, 1, &ztest_opts.zo_vdevtime), |
a1d477c2 | 440 | ZTI_INIT(ztest_device_removal, 1, &zopt_sometimes), |
d2734cce | 441 | ZTI_INIT(ztest_spa_checkpoint_create_discard, 1, &zopt_rarely), |
619f0976 | 442 | ZTI_INIT(ztest_initialize, 1, &zopt_sometimes), |
1b939560 | 443 | ZTI_INIT(ztest_trim, 1, &zopt_sometimes), |
1eeb4562 | 444 | ZTI_INIT(ztest_fletcher, 1, &zopt_rarely), |
37f520db | 445 | ZTI_INIT(ztest_fletcher_incr, 1, &zopt_rarely), |
50c957f7 | 446 | ZTI_INIT(ztest_verify_dnode_bt, 1, &zopt_sometimes), |
34dc7c2f BB |
447 | }; |
448 | ||
449 | #define ZTEST_FUNCS (sizeof (ztest_info) / sizeof (ztest_info_t)) | |
450 | ||
428870ff BB |
451 | /* |
452 | * The following struct is used to hold a list of uncalled commit callbacks. | |
453 | * The callbacks are ordered by txg number. | |
454 | */ | |
455 | typedef struct ztest_cb_list { | |
1e33ac1e BB |
456 | kmutex_t zcl_callbacks_lock; |
457 | list_t zcl_callbacks; | |
428870ff | 458 | } ztest_cb_list_t; |
34dc7c2f BB |
459 | |
460 | /* | |
461 | * Stuff we need to share writably between parent and child. | |
462 | */ | |
463 | typedef struct ztest_shared { | |
c242c188 | 464 | boolean_t zs_do_init; |
428870ff BB |
465 | hrtime_t zs_proc_start; |
466 | hrtime_t zs_proc_stop; | |
467 | hrtime_t zs_thread_start; | |
468 | hrtime_t zs_thread_stop; | |
469 | hrtime_t zs_thread_kill; | |
34dc7c2f | 470 | uint64_t zs_enospc_count; |
428870ff BB |
471 | uint64_t zs_vdev_next_leaf; |
472 | uint64_t zs_vdev_aux; | |
34dc7c2f BB |
473 | uint64_t zs_alloc; |
474 | uint64_t zs_space; | |
428870ff BB |
475 | uint64_t zs_splits; |
476 | uint64_t zs_mirrors; | |
c242c188 CS |
477 | uint64_t zs_metaslab_sz; |
478 | uint64_t zs_metaslab_df_alloc_threshold; | |
479 | uint64_t zs_guid; | |
34dc7c2f BB |
480 | } ztest_shared_t; |
481 | ||
428870ff BB |
482 | #define ID_PARALLEL -1ULL |
483 | ||
34dc7c2f | 484 | static char ztest_dev_template[] = "%s/%s.%llua"; |
b128c09f | 485 | static char ztest_aux_template[] = "%s/%s.%s.%llu"; |
428870ff | 486 | ztest_shared_t *ztest_shared; |
34dc7c2f | 487 | |
c242c188 CS |
488 | static spa_t *ztest_spa = NULL; |
489 | static ztest_ds_t *ztest_ds; | |
490 | ||
491 | static kmutex_t ztest_vdev_lock; | |
9e052db4 | 492 | static boolean_t ztest_device_removal_active = B_FALSE; |
ce5fb2a7 | 493 | static boolean_t ztest_pool_scrubbed = B_FALSE; |
d2734cce | 494 | static kmutex_t ztest_checkpoint_lock; |
3bc7e0fb GW |
495 | |
496 | /* | |
497 | * The ztest_name_lock protects the pool and dataset namespace used by | |
498 | * the individual tests. To modify the namespace, consumers must grab | |
499 | * this lock as writer. Grabbing the lock as reader will ensure that the | |
500 | * namespace does not change while the lock is held. | |
501 | */ | |
2d9142c9 | 502 | static pthread_rwlock_t ztest_name_lock; |
34dc7c2f | 503 | |
c242c188 | 504 | static boolean_t ztest_dump_core = B_TRUE; |
b128c09f | 505 | static boolean_t ztest_exiting; |
34dc7c2f | 506 | |
428870ff BB |
507 | /* Global commit callback list */ |
508 | static ztest_cb_list_t zcl; | |
090ff092 RC |
509 | /* Commit cb delay */ |
510 | static uint64_t zc_min_txg_delay = UINT64_MAX; | |
511 | static int zc_cb_counter = 0; | |
512 | ||
513 | /* | |
514 | * Minimum number of commit callbacks that need to be registered for us to check | |
515 | * whether the minimum txg delay is acceptable. | |
516 | */ | |
517 | #define ZTEST_COMMIT_CB_MIN_REG 100 | |
518 | ||
519 | /* | |
520 | * If a number of txgs equal to this threshold have been created after a commit | |
521 | * callback has been registered but not called, then we assume there is an | |
522 | * implementation bug. | |
523 | */ | |
524 | #define ZTEST_COMMIT_CB_THRESH (TXG_CONCURRENT_STATES + 1000) | |
428870ff | 525 | |
428870ff BB |
526 | enum ztest_object { |
527 | ZTEST_META_DNODE = 0, | |
528 | ZTEST_DIROBJ, | |
529 | ZTEST_OBJECTS | |
530 | }; | |
34dc7c2f BB |
531 | |
532 | static void usage(boolean_t) __NORETURN; | |
ce5fb2a7 | 533 | static int ztest_scrub_impl(spa_t *spa); |
34dc7c2f BB |
534 | |
535 | /* | |
536 | * These libumem hooks provide a reasonable set of defaults for the allocator's | |
537 | * debugging facilities. | |
538 | */ | |
539 | const char * | |
0bc8fd78 | 540 | _umem_debug_init(void) |
34dc7c2f BB |
541 | { |
542 | return ("default,verbose"); /* $UMEM_DEBUG setting */ | |
543 | } | |
544 | ||
545 | const char * | |
546 | _umem_logging_init(void) | |
547 | { | |
548 | return ("fail,contents"); /* $UMEM_LOGGING setting */ | |
549 | } | |
550 | ||
8fb1ede1 BB |
551 | static void |
552 | dump_debug_buffer(void) | |
553 | { | |
ab4c009e TC |
554 | ssize_t ret __attribute__((unused)); |
555 | ||
556 | if (!ztest_opts.zo_dump_dbgmsg) | |
8fb1ede1 BB |
557 | return; |
558 | ||
ab4c009e TC |
559 | /* |
560 | * We use write() instead of printf() so that this function | |
561 | * is safe to call from a signal handler. | |
562 | */ | |
563 | ret = write(STDOUT_FILENO, "\n", 1); | |
8fb1ede1 BB |
564 | zfs_dbgmsg_print("ztest"); |
565 | } | |
566 | ||
e82cdc3a NB |
567 | #define BACKTRACE_SZ 100 |
568 | ||
569 | static void sig_handler(int signo) | |
570 | { | |
571 | struct sigaction action; | |
aef1830f | 572 | #if (__GLIBC__ && !__UCLIBC__) /* backtrace() is a GNU extension */ |
e82cdc3a NB |
573 | int nptrs; |
574 | void *buffer[BACKTRACE_SZ]; | |
575 | ||
576 | nptrs = backtrace(buffer, BACKTRACE_SZ); | |
577 | backtrace_symbols_fd(buffer, nptrs, STDERR_FILENO); | |
578 | #endif | |
8fb1ede1 | 579 | dump_debug_buffer(); |
e82cdc3a NB |
580 | |
581 | /* | |
582 | * Restore default action and re-raise signal so SIGSEGV and | |
583 | * SIGABRT can trigger a core dump. | |
584 | */ | |
585 | action.sa_handler = SIG_DFL; | |
586 | sigemptyset(&action.sa_mask); | |
587 | action.sa_flags = 0; | |
588 | (void) sigaction(signo, &action, NULL); | |
589 | raise(signo); | |
590 | } | |
591 | ||
34dc7c2f BB |
592 | #define FATAL_MSG_SZ 1024 |
593 | ||
594 | char *fatal_msg; | |
595 | ||
596 | static void | |
597 | fatal(int do_perror, char *message, ...) | |
598 | { | |
599 | va_list args; | |
600 | int save_errno = errno; | |
40b84e7a | 601 | char *buf; |
34dc7c2f BB |
602 | |
603 | (void) fflush(stdout); | |
40b84e7a | 604 | buf = umem_alloc(FATAL_MSG_SZ, UMEM_NOFAIL); |
34dc7c2f BB |
605 | |
606 | va_start(args, message); | |
607 | (void) sprintf(buf, "ztest: "); | |
608 | /* LINTED */ | |
609 | (void) vsprintf(buf + strlen(buf), message, args); | |
610 | va_end(args); | |
611 | if (do_perror) { | |
612 | (void) snprintf(buf + strlen(buf), FATAL_MSG_SZ - strlen(buf), | |
613 | ": %s", strerror(save_errno)); | |
614 | } | |
615 | (void) fprintf(stderr, "%s\n", buf); | |
616 | fatal_msg = buf; /* to ease debugging */ | |
8fb1ede1 | 617 | |
34dc7c2f BB |
618 | if (ztest_dump_core) |
619 | abort(); | |
ab4c009e TC |
620 | else |
621 | dump_debug_buffer(); | |
622 | ||
34dc7c2f BB |
623 | exit(3); |
624 | } | |
625 | ||
626 | static int | |
627 | str2shift(const char *buf) | |
628 | { | |
629 | const char *ends = "BKMGTPEZ"; | |
630 | int i; | |
631 | ||
632 | if (buf[0] == '\0') | |
633 | return (0); | |
634 | for (i = 0; i < strlen(ends); i++) { | |
635 | if (toupper(buf[0]) == ends[i]) | |
636 | break; | |
637 | } | |
638 | if (i == strlen(ends)) { | |
639 | (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", | |
640 | buf); | |
641 | usage(B_FALSE); | |
642 | } | |
643 | if (buf[1] == '\0' || (toupper(buf[1]) == 'B' && buf[2] == '\0')) { | |
644 | return (10*i); | |
645 | } | |
646 | (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", buf); | |
647 | usage(B_FALSE); | |
648 | /* NOTREACHED */ | |
649 | } | |
650 | ||
651 | static uint64_t | |
652 | nicenumtoull(const char *buf) | |
653 | { | |
654 | char *end; | |
655 | uint64_t val; | |
656 | ||
657 | val = strtoull(buf, &end, 0); | |
658 | if (end == buf) { | |
659 | (void) fprintf(stderr, "ztest: bad numeric value: %s\n", buf); | |
660 | usage(B_FALSE); | |
661 | } else if (end[0] == '.') { | |
662 | double fval = strtod(buf, &end); | |
663 | fval *= pow(2, str2shift(end)); | |
4d32abaa RM |
664 | /* |
665 | * UINT64_MAX is not exactly representable as a double. | |
666 | * The closest representation is UINT64_MAX + 1, so we | |
667 | * use a >= comparison instead of > for the bounds check. | |
668 | */ | |
669 | if (fval >= (double)UINT64_MAX) { | |
34dc7c2f BB |
670 | (void) fprintf(stderr, "ztest: value too large: %s\n", |
671 | buf); | |
672 | usage(B_FALSE); | |
673 | } | |
674 | val = (uint64_t)fval; | |
675 | } else { | |
676 | int shift = str2shift(end); | |
677 | if (shift >= 64 || (val << shift) >> shift != val) { | |
678 | (void) fprintf(stderr, "ztest: value too large: %s\n", | |
679 | buf); | |
680 | usage(B_FALSE); | |
681 | } | |
682 | val <<= shift; | |
683 | } | |
684 | return (val); | |
685 | } | |
686 | ||
687 | static void | |
688 | usage(boolean_t requested) | |
689 | { | |
c242c188 CS |
690 | const ztest_shared_opts_t *zo = &ztest_opts_defaults; |
691 | ||
f3c8c9e6 | 692 | char nice_vdev_size[NN_NUMBUF_SZ]; |
d830d479 | 693 | char nice_force_ganging[NN_NUMBUF_SZ]; |
34dc7c2f BB |
694 | FILE *fp = requested ? stdout : stderr; |
695 | ||
f3c8c9e6 | 696 | nicenum(zo->zo_vdev_size, nice_vdev_size, sizeof (nice_vdev_size)); |
d830d479 MA |
697 | nicenum(zo->zo_metaslab_force_ganging, nice_force_ganging, |
698 | sizeof (nice_force_ganging)); | |
34dc7c2f BB |
699 | |
700 | (void) fprintf(fp, "Usage: %s\n" | |
701 | "\t[-v vdevs (default: %llu)]\n" | |
702 | "\t[-s size_of_each_vdev (default: %s)]\n" | |
428870ff | 703 | "\t[-a alignment_shift (default: %d)] use 0 for random\n" |
34dc7c2f | 704 | "\t[-m mirror_copies (default: %d)]\n" |
b2255edc BB |
705 | "\t[-r raidz_disks / draid_disks (default: %d)]\n" |
706 | "\t[-R raid_parity (default: %d)]\n" | |
707 | "\t[-K raid_kind (default: random)] raidz|draid|random\n" | |
708 | "\t[-D draid_data (default: %d)] in config\n" | |
709 | "\t[-S draid_spares (default: %d)]\n" | |
34dc7c2f BB |
710 | "\t[-d datasets (default: %d)]\n" |
711 | "\t[-t threads (default: %d)]\n" | |
712 | "\t[-g gang_block_threshold (default: %s)]\n" | |
428870ff BB |
713 | "\t[-i init_count (default: %d)] initialize pool i times\n" |
714 | "\t[-k kill_percentage (default: %llu%%)]\n" | |
34dc7c2f | 715 | "\t[-p pool_name (default: %s)]\n" |
428870ff | 716 | "\t[-f dir (default: %s)] file directory for vdev files\n" |
379ca9cf | 717 | "\t[-M] Multi-host simulate pool imported on remote host\n" |
428870ff BB |
718 | "\t[-V] verbose (use multiple times for ever more blather)\n" |
719 | "\t[-E] use existing pool instead of creating new one\n" | |
720 | "\t[-T time (default: %llu sec)] total run time\n" | |
721 | "\t[-F freezeloops (default: %llu)] max loops in spa_freeze()\n" | |
722 | "\t[-P passtime (default: %llu sec)] time per pass\n" | |
c242c188 | 723 | "\t[-B alt_ztest (default: <none>)] alternate ztest path\n" |
cc99f275 | 724 | "\t[-C vdev class state (default: random)] special=on|off|random\n" |
ed828c0c GM |
725 | "\t[-o variable=value] ... set global variable to an unsigned\n" |
726 | "\t 32-bit integer value\n" | |
8fb1ede1 | 727 | "\t[-G dump zfs_dbgmsg buffer before exiting due to an error\n" |
34dc7c2f BB |
728 | "\t[-h] (print help)\n" |
729 | "", | |
c242c188 CS |
730 | zo->zo_pool, |
731 | (u_longlong_t)zo->zo_vdevs, /* -v */ | |
34dc7c2f | 732 | nice_vdev_size, /* -s */ |
c242c188 CS |
733 | zo->zo_ashift, /* -a */ |
734 | zo->zo_mirrors, /* -m */ | |
b2255edc BB |
735 | zo->zo_raid_children, /* -r */ |
736 | zo->zo_raid_parity, /* -R */ | |
737 | zo->zo_draid_data, /* -D */ | |
738 | zo->zo_draid_spares, /* -S */ | |
c242c188 CS |
739 | zo->zo_datasets, /* -d */ |
740 | zo->zo_threads, /* -t */ | |
d830d479 | 741 | nice_force_ganging, /* -g */ |
c242c188 CS |
742 | zo->zo_init, /* -i */ |
743 | (u_longlong_t)zo->zo_killrate, /* -k */ | |
744 | zo->zo_pool, /* -p */ | |
745 | zo->zo_dir, /* -f */ | |
746 | (u_longlong_t)zo->zo_time, /* -T */ | |
747 | (u_longlong_t)zo->zo_maxloops, /* -F */ | |
748 | (u_longlong_t)zo->zo_passtime); | |
34dc7c2f BB |
749 | exit(requested ? 0 : 1); |
750 | } | |
751 | ||
b2255edc BB |
752 | static uint64_t |
753 | ztest_random(uint64_t range) | |
754 | { | |
755 | uint64_t r; | |
756 | ||
757 | ASSERT3S(ztest_fd_rand, >=, 0); | |
758 | ||
759 | if (range == 0) | |
760 | return (0); | |
761 | ||
762 | if (read(ztest_fd_rand, &r, sizeof (r)) != sizeof (r)) | |
763 | fatal(1, "short read from /dev/urandom"); | |
764 | ||
765 | return (r % range); | |
766 | } | |
cc99f275 DB |
767 | |
768 | static void | |
769 | ztest_parse_name_value(const char *input, ztest_shared_opts_t *zo) | |
770 | { | |
771 | char name[32]; | |
772 | char *value; | |
773 | int state = ZTEST_VDEV_CLASS_RND; | |
774 | ||
775 | (void) strlcpy(name, input, sizeof (name)); | |
776 | ||
777 | value = strchr(name, '='); | |
778 | if (value == NULL) { | |
779 | (void) fprintf(stderr, "missing value in property=value " | |
780 | "'-C' argument (%s)\n", input); | |
781 | usage(B_FALSE); | |
782 | } | |
783 | *(value) = '\0'; | |
784 | value++; | |
785 | ||
786 | if (strcmp(value, "on") == 0) { | |
787 | state = ZTEST_VDEV_CLASS_ON; | |
788 | } else if (strcmp(value, "off") == 0) { | |
789 | state = ZTEST_VDEV_CLASS_OFF; | |
790 | } else if (strcmp(value, "random") == 0) { | |
791 | state = ZTEST_VDEV_CLASS_RND; | |
792 | } else { | |
793 | (void) fprintf(stderr, "invalid property value '%s'\n", value); | |
794 | usage(B_FALSE); | |
795 | } | |
796 | ||
797 | if (strcmp(name, "special") == 0) { | |
798 | zo->zo_special_vdevs = state; | |
799 | } else { | |
800 | (void) fprintf(stderr, "invalid property name '%s'\n", name); | |
801 | usage(B_FALSE); | |
802 | } | |
803 | if (zo->zo_verbose >= 3) | |
804 | (void) printf("%s vdev state is '%s'\n", name, value); | |
805 | } | |
806 | ||
34dc7c2f BB |
807 | static void |
808 | process_options(int argc, char **argv) | |
809 | { | |
c242c188 CS |
810 | char *path; |
811 | ztest_shared_opts_t *zo = &ztest_opts; | |
812 | ||
34dc7c2f BB |
813 | int opt; |
814 | uint64_t value; | |
c242c188 | 815 | char altdir[MAXNAMELEN] = { 0 }; |
b2255edc | 816 | char raid_kind[8] = { "random" }; |
34dc7c2f | 817 | |
c242c188 | 818 | bcopy(&ztest_opts_defaults, zo, sizeof (*zo)); |
34dc7c2f | 819 | |
34dc7c2f | 820 | while ((opt = getopt(argc, argv, |
b2255edc | 821 | "v:s:a:m:r:R:K:D:S:d:t:g:i:k:p:f:MVET:P:hF:B:C:o:G")) != EOF) { |
34dc7c2f BB |
822 | value = 0; |
823 | switch (opt) { | |
824 | case 'v': | |
825 | case 's': | |
826 | case 'a': | |
827 | case 'm': | |
828 | case 'r': | |
829 | case 'R': | |
b2255edc BB |
830 | case 'D': |
831 | case 'S': | |
34dc7c2f BB |
832 | case 'd': |
833 | case 't': | |
834 | case 'g': | |
835 | case 'i': | |
836 | case 'k': | |
837 | case 'T': | |
838 | case 'P': | |
428870ff | 839 | case 'F': |
34dc7c2f BB |
840 | value = nicenumtoull(optarg); |
841 | } | |
842 | switch (opt) { | |
843 | case 'v': | |
c242c188 | 844 | zo->zo_vdevs = value; |
34dc7c2f BB |
845 | break; |
846 | case 's': | |
c242c188 | 847 | zo->zo_vdev_size = MAX(SPA_MINDEVSIZE, value); |
34dc7c2f BB |
848 | break; |
849 | case 'a': | |
c242c188 | 850 | zo->zo_ashift = value; |
34dc7c2f BB |
851 | break; |
852 | case 'm': | |
c242c188 | 853 | zo->zo_mirrors = value; |
34dc7c2f BB |
854 | break; |
855 | case 'r': | |
b2255edc | 856 | zo->zo_raid_children = MAX(1, value); |
34dc7c2f BB |
857 | break; |
858 | case 'R': | |
b2255edc BB |
859 | zo->zo_raid_parity = MIN(MAX(value, 1), 3); |
860 | break; | |
861 | case 'K': | |
862 | (void) strlcpy(raid_kind, optarg, sizeof (raid_kind)); | |
863 | break; | |
864 | case 'D': | |
865 | zo->zo_draid_data = MAX(1, value); | |
866 | break; | |
867 | case 'S': | |
868 | zo->zo_draid_spares = MAX(1, value); | |
34dc7c2f BB |
869 | break; |
870 | case 'd': | |
c242c188 | 871 | zo->zo_datasets = MAX(1, value); |
34dc7c2f BB |
872 | break; |
873 | case 't': | |
c242c188 | 874 | zo->zo_threads = MAX(1, value); |
34dc7c2f BB |
875 | break; |
876 | case 'g': | |
d830d479 MA |
877 | zo->zo_metaslab_force_ganging = |
878 | MAX(SPA_MINBLOCKSIZE << 1, value); | |
34dc7c2f BB |
879 | break; |
880 | case 'i': | |
c242c188 | 881 | zo->zo_init = value; |
34dc7c2f BB |
882 | break; |
883 | case 'k': | |
c242c188 | 884 | zo->zo_killrate = value; |
34dc7c2f BB |
885 | break; |
886 | case 'p': | |
c242c188 CS |
887 | (void) strlcpy(zo->zo_pool, optarg, |
888 | sizeof (zo->zo_pool)); | |
34dc7c2f BB |
889 | break; |
890 | case 'f': | |
c242c188 CS |
891 | path = realpath(optarg, NULL); |
892 | if (path == NULL) { | |
893 | (void) fprintf(stderr, "error: %s: %s\n", | |
894 | optarg, strerror(errno)); | |
895 | usage(B_FALSE); | |
896 | } else { | |
897 | (void) strlcpy(zo->zo_dir, path, | |
898 | sizeof (zo->zo_dir)); | |
df053d67 | 899 | free(path); |
c242c188 | 900 | } |
34dc7c2f | 901 | break; |
379ca9cf OF |
902 | case 'M': |
903 | zo->zo_mmp_test = 1; | |
904 | break; | |
34dc7c2f | 905 | case 'V': |
c242c188 | 906 | zo->zo_verbose++; |
34dc7c2f BB |
907 | break; |
908 | case 'E': | |
c242c188 | 909 | zo->zo_init = 0; |
34dc7c2f BB |
910 | break; |
911 | case 'T': | |
c242c188 | 912 | zo->zo_time = value; |
34dc7c2f BB |
913 | break; |
914 | case 'P': | |
c242c188 | 915 | zo->zo_passtime = MAX(1, value); |
34dc7c2f | 916 | break; |
428870ff | 917 | case 'F': |
c242c188 CS |
918 | zo->zo_maxloops = MAX(1, value); |
919 | break; | |
920 | case 'B': | |
921 | (void) strlcpy(altdir, optarg, sizeof (altdir)); | |
428870ff | 922 | break; |
cc99f275 DB |
923 | case 'C': |
924 | ztest_parse_name_value(optarg, zo); | |
925 | break; | |
ed828c0c | 926 | case 'o': |
2801e68d CS |
927 | if (zo->zo_gvars_count >= ZO_GVARS_MAX_COUNT) { |
928 | (void) fprintf(stderr, | |
929 | "max global var count (%zu) exceeded\n", | |
930 | ZO_GVARS_MAX_COUNT); | |
931 | usage(B_FALSE); | |
932 | } | |
933 | char *v = zo->zo_gvars[zo->zo_gvars_count]; | |
934 | if (strlcpy(v, optarg, ZO_GVARS_MAX_ARGLEN) >= | |
935 | ZO_GVARS_MAX_ARGLEN) { | |
936 | (void) fprintf(stderr, | |
937 | "global var option '%s' is too long\n", | |
938 | optarg); | |
ed828c0c | 939 | usage(B_FALSE); |
2801e68d CS |
940 | } |
941 | zo->zo_gvars_count++; | |
ed828c0c | 942 | break; |
8fb1ede1 | 943 | case 'G': |
ab4c009e | 944 | zo->zo_dump_dbgmsg = 1; |
8fb1ede1 | 945 | break; |
34dc7c2f BB |
946 | case 'h': |
947 | usage(B_TRUE); | |
948 | break; | |
949 | case '?': | |
950 | default: | |
951 | usage(B_FALSE); | |
952 | break; | |
953 | } | |
954 | } | |
955 | ||
b2255edc BB |
956 | /* When raid choice is 'random' add a draid pool 50% of the time */ |
957 | if (strcmp(raid_kind, "random") == 0) { | |
958 | (void) strlcpy(raid_kind, (ztest_random(2) == 0) ? | |
959 | "draid" : "raidz", sizeof (raid_kind)); | |
960 | ||
961 | if (ztest_opts.zo_verbose >= 3) | |
962 | (void) printf("choosing RAID type '%s'\n", raid_kind); | |
963 | } | |
964 | ||
965 | if (strcmp(raid_kind, "draid") == 0) { | |
966 | uint64_t min_devsize; | |
967 | ||
968 | /* With fewer disk use 256M, otherwise 128M is OK */ | |
969 | min_devsize = (ztest_opts.zo_raid_children < 16) ? | |
970 | (256ULL << 20) : (128ULL << 20); | |
971 | ||
972 | /* No top-level mirrors with dRAID for now */ | |
973 | zo->zo_mirrors = 0; | |
974 | ||
975 | /* Use more appropriate defaults for dRAID */ | |
976 | if (zo->zo_vdevs == ztest_opts_defaults.zo_vdevs) | |
977 | zo->zo_vdevs = 1; | |
978 | if (zo->zo_raid_children == | |
979 | ztest_opts_defaults.zo_raid_children) | |
980 | zo->zo_raid_children = 16; | |
981 | if (zo->zo_ashift < 12) | |
982 | zo->zo_ashift = 12; | |
983 | if (zo->zo_vdev_size < min_devsize) | |
984 | zo->zo_vdev_size = min_devsize; | |
985 | ||
986 | if (zo->zo_draid_data + zo->zo_raid_parity > | |
987 | zo->zo_raid_children - zo->zo_draid_spares) { | |
988 | (void) fprintf(stderr, "error: too few draid " | |
989 | "children (%d) for stripe width (%d)\n", | |
990 | zo->zo_raid_children, | |
991 | zo->zo_draid_data + zo->zo_raid_parity); | |
992 | usage(B_FALSE); | |
993 | } | |
994 | ||
995 | (void) strlcpy(zo->zo_raid_type, VDEV_TYPE_DRAID, | |
996 | sizeof (zo->zo_raid_type)); | |
997 | ||
998 | } else /* using raidz */ { | |
999 | ASSERT0(strcmp(raid_kind, "raidz")); | |
1000 | ||
1001 | zo->zo_raid_parity = MIN(zo->zo_raid_parity, | |
1002 | zo->zo_raid_children - 1); | |
1003 | } | |
34dc7c2f | 1004 | |
c242c188 CS |
1005 | zo->zo_vdevtime = |
1006 | (zo->zo_vdevs > 0 ? zo->zo_time * NANOSEC / zo->zo_vdevs : | |
428870ff | 1007 | UINT64_MAX >> 2); |
c242c188 CS |
1008 | |
1009 | if (strlen(altdir) > 0) { | |
ae380cfa BB |
1010 | char *cmd; |
1011 | char *realaltdir; | |
c242c188 CS |
1012 | char *bin; |
1013 | char *ztest; | |
1014 | char *isa; | |
1015 | int isalen; | |
1016 | ||
ae380cfa BB |
1017 | cmd = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
1018 | realaltdir = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
1019 | ||
a9eaae06 | 1020 | VERIFY3P(NULL, !=, realpath(getexecname(), cmd)); |
c242c188 CS |
1021 | if (0 != access(altdir, F_OK)) { |
1022 | ztest_dump_core = B_FALSE; | |
1023 | fatal(B_TRUE, "invalid alternate ztest path: %s", | |
1024 | altdir); | |
1025 | } | |
a9eaae06 | 1026 | VERIFY3P(NULL, !=, realpath(altdir, realaltdir)); |
c242c188 CS |
1027 | |
1028 | /* | |
1029 | * 'cmd' should be of the form "<anything>/usr/bin/<isa>/ztest". | |
1030 | * We want to extract <isa> to determine if we should use | |
1031 | * 32 or 64 bit binaries. | |
1032 | */ | |
1033 | bin = strstr(cmd, "/usr/bin/"); | |
1034 | ztest = strstr(bin, "/ztest"); | |
1035 | isa = bin + 9; | |
1036 | isalen = ztest - isa; | |
1037 | (void) snprintf(zo->zo_alt_ztest, sizeof (zo->zo_alt_ztest), | |
1038 | "%s/usr/bin/%.*s/ztest", realaltdir, isalen, isa); | |
1039 | (void) snprintf(zo->zo_alt_libpath, sizeof (zo->zo_alt_libpath), | |
1040 | "%s/usr/lib/%.*s", realaltdir, isalen, isa); | |
1041 | ||
1042 | if (0 != access(zo->zo_alt_ztest, X_OK)) { | |
1043 | ztest_dump_core = B_FALSE; | |
1044 | fatal(B_TRUE, "invalid alternate ztest: %s", | |
1045 | zo->zo_alt_ztest); | |
1046 | } else if (0 != access(zo->zo_alt_libpath, X_OK)) { | |
1047 | ztest_dump_core = B_FALSE; | |
1048 | fatal(B_TRUE, "invalid alternate lib directory %s", | |
1049 | zo->zo_alt_libpath); | |
1050 | } | |
ae380cfa BB |
1051 | |
1052 | umem_free(cmd, MAXPATHLEN); | |
1053 | umem_free(realaltdir, MAXPATHLEN); | |
c242c188 | 1054 | } |
428870ff BB |
1055 | } |
1056 | ||
1057 | static void | |
1058 | ztest_kill(ztest_shared_t *zs) | |
1059 | { | |
c242c188 CS |
1060 | zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(ztest_spa)); |
1061 | zs->zs_space = metaslab_class_get_space(spa_normal_class(ztest_spa)); | |
5d1f7fb6 GW |
1062 | |
1063 | /* | |
1064 | * Before we kill off ztest, make sure that the config is updated. | |
a1d477c2 | 1065 | * See comment above spa_write_cachefile(). |
5d1f7fb6 GW |
1066 | */ |
1067 | mutex_enter(&spa_namespace_lock); | |
a1d477c2 | 1068 | spa_write_cachefile(ztest_spa, B_FALSE, B_FALSE); |
5d1f7fb6 GW |
1069 | mutex_exit(&spa_namespace_lock); |
1070 | ||
428870ff BB |
1071 | (void) kill(getpid(), SIGKILL); |
1072 | } | |
1073 | ||
428870ff BB |
1074 | /* ARGSUSED */ |
1075 | static void | |
1076 | ztest_record_enospc(const char *s) | |
1077 | { | |
1078 | ztest_shared->zs_enospc_count++; | |
34dc7c2f BB |
1079 | } |
1080 | ||
1081 | static uint64_t | |
1082 | ztest_get_ashift(void) | |
1083 | { | |
c242c188 | 1084 | if (ztest_opts.zo_ashift == 0) |
b02fe35d | 1085 | return (SPA_MINBLOCKSHIFT + ztest_random(5)); |
c242c188 | 1086 | return (ztest_opts.zo_ashift); |
34dc7c2f BB |
1087 | } |
1088 | ||
b2255edc BB |
1089 | static boolean_t |
1090 | ztest_is_draid_spare(const char *name) | |
1091 | { | |
1092 | uint64_t spare_id = 0, parity = 0, vdev_id = 0; | |
1093 | ||
1094 | if (sscanf(name, VDEV_TYPE_DRAID "%llu-%llu-%llu", | |
1095 | (u_longlong_t *)&parity, (u_longlong_t *)&vdev_id, | |
1096 | (u_longlong_t *)&spare_id) == 3) { | |
1097 | return (B_TRUE); | |
1098 | } | |
1099 | ||
1100 | return (B_FALSE); | |
1101 | } | |
1102 | ||
34dc7c2f | 1103 | static nvlist_t * |
ea0b2538 | 1104 | make_vdev_file(char *path, char *aux, char *pool, size_t size, uint64_t ashift) |
34dc7c2f | 1105 | { |
40b84e7a | 1106 | char *pathbuf; |
34dc7c2f | 1107 | uint64_t vdev; |
34dc7c2f | 1108 | nvlist_t *file; |
b2255edc | 1109 | boolean_t draid_spare = B_FALSE; |
34dc7c2f | 1110 | |
40b84e7a BB |
1111 | pathbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
1112 | ||
b128c09f BB |
1113 | if (ashift == 0) |
1114 | ashift = ztest_get_ashift(); | |
1115 | ||
1116 | if (path == NULL) { | |
1117 | path = pathbuf; | |
1118 | ||
1119 | if (aux != NULL) { | |
1120 | vdev = ztest_shared->zs_vdev_aux; | |
c242c188 CS |
1121 | (void) snprintf(path, MAXPATHLEN, |
1122 | ztest_aux_template, ztest_opts.zo_dir, | |
ea0b2538 GW |
1123 | pool == NULL ? ztest_opts.zo_pool : pool, |
1124 | aux, vdev); | |
b128c09f | 1125 | } else { |
428870ff | 1126 | vdev = ztest_shared->zs_vdev_next_leaf++; |
c242c188 CS |
1127 | (void) snprintf(path, MAXPATHLEN, |
1128 | ztest_dev_template, ztest_opts.zo_dir, | |
ea0b2538 | 1129 | pool == NULL ? ztest_opts.zo_pool : pool, vdev); |
b128c09f | 1130 | } |
b2255edc BB |
1131 | } else { |
1132 | draid_spare = ztest_is_draid_spare(path); | |
b128c09f | 1133 | } |
34dc7c2f | 1134 | |
b2255edc | 1135 | if (size != 0 && !draid_spare) { |
b128c09f | 1136 | int fd = open(path, O_RDWR | O_CREAT | O_TRUNC, 0666); |
34dc7c2f | 1137 | if (fd == -1) |
b128c09f | 1138 | fatal(1, "can't open %s", path); |
34dc7c2f | 1139 | if (ftruncate(fd, size) != 0) |
b128c09f | 1140 | fatal(1, "can't ftruncate %s", path); |
34dc7c2f BB |
1141 | (void) close(fd); |
1142 | } | |
1143 | ||
bea3fc71 RM |
1144 | file = fnvlist_alloc(); |
1145 | fnvlist_add_string(file, ZPOOL_CONFIG_TYPE, | |
1146 | draid_spare ? VDEV_TYPE_DRAID_SPARE : VDEV_TYPE_FILE); | |
1147 | fnvlist_add_string(file, ZPOOL_CONFIG_PATH, path); | |
1148 | fnvlist_add_uint64(file, ZPOOL_CONFIG_ASHIFT, ashift); | |
40b84e7a | 1149 | umem_free(pathbuf, MAXPATHLEN); |
34dc7c2f BB |
1150 | |
1151 | return (file); | |
1152 | } | |
1153 | ||
1154 | static nvlist_t * | |
b2255edc | 1155 | make_vdev_raid(char *path, char *aux, char *pool, size_t size, |
ea0b2538 | 1156 | uint64_t ashift, int r) |
34dc7c2f | 1157 | { |
b2255edc | 1158 | nvlist_t *raid, **child; |
34dc7c2f BB |
1159 | int c; |
1160 | ||
1161 | if (r < 2) | |
ea0b2538 | 1162 | return (make_vdev_file(path, aux, pool, size, ashift)); |
34dc7c2f BB |
1163 | child = umem_alloc(r * sizeof (nvlist_t *), UMEM_NOFAIL); |
1164 | ||
1165 | for (c = 0; c < r; c++) | |
ea0b2538 | 1166 | child[c] = make_vdev_file(path, aux, pool, size, ashift); |
34dc7c2f | 1167 | |
bea3fc71 RM |
1168 | raid = fnvlist_alloc(); |
1169 | fnvlist_add_string(raid, ZPOOL_CONFIG_TYPE, | |
1170 | ztest_opts.zo_raid_type); | |
1171 | fnvlist_add_uint64(raid, ZPOOL_CONFIG_NPARITY, | |
1172 | ztest_opts.zo_raid_parity); | |
1173 | fnvlist_add_nvlist_array(raid, ZPOOL_CONFIG_CHILDREN, child, r); | |
b2255edc BB |
1174 | |
1175 | if (strcmp(ztest_opts.zo_raid_type, VDEV_TYPE_DRAID) == 0) { | |
1176 | uint64_t ndata = ztest_opts.zo_draid_data; | |
1177 | uint64_t nparity = ztest_opts.zo_raid_parity; | |
1178 | uint64_t nspares = ztest_opts.zo_draid_spares; | |
1179 | uint64_t children = ztest_opts.zo_raid_children; | |
1180 | uint64_t ngroups = 1; | |
1181 | ||
1182 | /* | |
1183 | * Calculate the minimum number of groups required to fill a | |
1184 | * slice. This is the LCM of the stripe width (data + parity) | |
1185 | * and the number of data drives (children - spares). | |
1186 | */ | |
1187 | while (ngroups * (ndata + nparity) % (children - nspares) != 0) | |
1188 | ngroups++; | |
1189 | ||
1190 | /* Store the basic dRAID configuration. */ | |
1191 | fnvlist_add_uint64(raid, ZPOOL_CONFIG_DRAID_NDATA, ndata); | |
1192 | fnvlist_add_uint64(raid, ZPOOL_CONFIG_DRAID_NSPARES, nspares); | |
1193 | fnvlist_add_uint64(raid, ZPOOL_CONFIG_DRAID_NGROUPS, ngroups); | |
1194 | } | |
34dc7c2f BB |
1195 | |
1196 | for (c = 0; c < r; c++) | |
bea3fc71 | 1197 | fnvlist_free(child[c]); |
34dc7c2f BB |
1198 | |
1199 | umem_free(child, r * sizeof (nvlist_t *)); | |
1200 | ||
b2255edc | 1201 | return (raid); |
34dc7c2f BB |
1202 | } |
1203 | ||
1204 | static nvlist_t * | |
ea0b2538 GW |
1205 | make_vdev_mirror(char *path, char *aux, char *pool, size_t size, |
1206 | uint64_t ashift, int r, int m) | |
34dc7c2f BB |
1207 | { |
1208 | nvlist_t *mirror, **child; | |
1209 | int c; | |
1210 | ||
1211 | if (m < 1) | |
b2255edc | 1212 | return (make_vdev_raid(path, aux, pool, size, ashift, r)); |
34dc7c2f BB |
1213 | |
1214 | child = umem_alloc(m * sizeof (nvlist_t *), UMEM_NOFAIL); | |
1215 | ||
1216 | for (c = 0; c < m; c++) | |
b2255edc | 1217 | child[c] = make_vdev_raid(path, aux, pool, size, ashift, r); |
34dc7c2f | 1218 | |
bea3fc71 RM |
1219 | mirror = fnvlist_alloc(); |
1220 | fnvlist_add_string(mirror, ZPOOL_CONFIG_TYPE, VDEV_TYPE_MIRROR); | |
1221 | fnvlist_add_nvlist_array(mirror, ZPOOL_CONFIG_CHILDREN, child, m); | |
34dc7c2f BB |
1222 | |
1223 | for (c = 0; c < m; c++) | |
bea3fc71 | 1224 | fnvlist_free(child[c]); |
34dc7c2f BB |
1225 | |
1226 | umem_free(child, m * sizeof (nvlist_t *)); | |
1227 | ||
1228 | return (mirror); | |
1229 | } | |
1230 | ||
1231 | static nvlist_t * | |
ea0b2538 | 1232 | make_vdev_root(char *path, char *aux, char *pool, size_t size, uint64_t ashift, |
cc99f275 | 1233 | const char *class, int r, int m, int t) |
34dc7c2f BB |
1234 | { |
1235 | nvlist_t *root, **child; | |
1236 | int c; | |
cc99f275 | 1237 | boolean_t log; |
34dc7c2f | 1238 | |
a9eaae06 | 1239 | ASSERT3S(t, >, 0); |
34dc7c2f | 1240 | |
cc99f275 DB |
1241 | log = (class != NULL && strcmp(class, "log") == 0); |
1242 | ||
34dc7c2f BB |
1243 | child = umem_alloc(t * sizeof (nvlist_t *), UMEM_NOFAIL); |
1244 | ||
b128c09f | 1245 | for (c = 0; c < t; c++) { |
ea0b2538 GW |
1246 | child[c] = make_vdev_mirror(path, aux, pool, size, ashift, |
1247 | r, m); | |
bea3fc71 | 1248 | fnvlist_add_uint64(child[c], ZPOOL_CONFIG_IS_LOG, log); |
cc99f275 DB |
1249 | |
1250 | if (class != NULL && class[0] != '\0') { | |
1251 | ASSERT(m > 1 || log); /* expecting a mirror */ | |
bea3fc71 RM |
1252 | fnvlist_add_string(child[c], |
1253 | ZPOOL_CONFIG_ALLOCATION_BIAS, class); | |
cc99f275 | 1254 | } |
b128c09f | 1255 | } |
34dc7c2f | 1256 | |
bea3fc71 RM |
1257 | root = fnvlist_alloc(); |
1258 | fnvlist_add_string(root, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT); | |
1259 | fnvlist_add_nvlist_array(root, aux ? aux : ZPOOL_CONFIG_CHILDREN, | |
1260 | child, t); | |
34dc7c2f BB |
1261 | |
1262 | for (c = 0; c < t; c++) | |
bea3fc71 | 1263 | fnvlist_free(child[c]); |
34dc7c2f BB |
1264 | |
1265 | umem_free(child, t * sizeof (nvlist_t *)); | |
1266 | ||
1267 | return (root); | |
1268 | } | |
1269 | ||
ea0b2538 GW |
1270 | /* |
1271 | * Find a random spa version. Returns back a random spa version in the | |
1272 | * range [initial_version, SPA_VERSION_FEATURES]. | |
1273 | */ | |
1274 | static uint64_t | |
1275 | ztest_random_spa_version(uint64_t initial_version) | |
1276 | { | |
1277 | uint64_t version = initial_version; | |
1278 | ||
1279 | if (version <= SPA_VERSION_BEFORE_FEATURES) { | |
1280 | version = version + | |
1281 | ztest_random(SPA_VERSION_BEFORE_FEATURES - version + 1); | |
1282 | } | |
1283 | ||
1284 | if (version > SPA_VERSION_BEFORE_FEATURES) | |
1285 | version = SPA_VERSION_FEATURES; | |
1286 | ||
1287 | ASSERT(SPA_VERSION_IS_SUPPORTED(version)); | |
1288 | return (version); | |
1289 | } | |
1290 | ||
428870ff BB |
1291 | static int |
1292 | ztest_random_blocksize(void) | |
34dc7c2f | 1293 | { |
a9eaae06 | 1294 | ASSERT3U(ztest_spa->spa_max_ashift, !=, 0); |
cc99f275 | 1295 | |
f1512ee6 MA |
1296 | /* |
1297 | * Choose a block size >= the ashift. | |
1298 | * If the SPA supports new MAXBLOCKSIZE, test up to 1MB blocks. | |
1299 | */ | |
1300 | int maxbs = SPA_OLD_MAXBLOCKSHIFT; | |
1301 | if (spa_maxblocksize(ztest_spa) == SPA_MAXBLOCKSIZE) | |
1302 | maxbs = 20; | |
c3520e7f MA |
1303 | uint64_t block_shift = |
1304 | ztest_random(maxbs - ztest_spa->spa_max_ashift + 1); | |
b02fe35d | 1305 | return (1 << (SPA_MINBLOCKSHIFT + block_shift)); |
428870ff | 1306 | } |
34dc7c2f | 1307 | |
50c957f7 NB |
1308 | static int |
1309 | ztest_random_dnodesize(void) | |
1310 | { | |
1311 | int slots; | |
1312 | int max_slots = spa_maxdnodesize(ztest_spa) >> DNODE_SHIFT; | |
1313 | ||
1314 | if (max_slots == DNODE_MIN_SLOTS) | |
1315 | return (DNODE_MIN_SIZE); | |
1316 | ||
1317 | /* | |
1318 | * Weight the random distribution more heavily toward smaller | |
1319 | * dnode sizes since that is more likely to reflect real-world | |
1320 | * usage. | |
1321 | */ | |
1322 | ASSERT3U(max_slots, >, 4); | |
1323 | switch (ztest_random(10)) { | |
1324 | case 0: | |
1325 | slots = 5 + ztest_random(max_slots - 4); | |
1326 | break; | |
1327 | case 1 ... 4: | |
1328 | slots = 2 + ztest_random(3); | |
1329 | break; | |
1330 | default: | |
1331 | slots = 1; | |
1332 | break; | |
1333 | } | |
1334 | ||
1335 | return (slots << DNODE_SHIFT); | |
1336 | } | |
1337 | ||
428870ff BB |
1338 | static int |
1339 | ztest_random_ibshift(void) | |
1340 | { | |
1341 | return (DN_MIN_INDBLKSHIFT + | |
1342 | ztest_random(DN_MAX_INDBLKSHIFT - DN_MIN_INDBLKSHIFT + 1)); | |
34dc7c2f BB |
1343 | } |
1344 | ||
428870ff BB |
1345 | static uint64_t |
1346 | ztest_random_vdev_top(spa_t *spa, boolean_t log_ok) | |
34dc7c2f | 1347 | { |
428870ff BB |
1348 | uint64_t top; |
1349 | vdev_t *rvd = spa->spa_root_vdev; | |
1350 | vdev_t *tvd; | |
34dc7c2f | 1351 | |
a9eaae06 | 1352 | ASSERT3U(spa_config_held(spa, SCL_ALL, RW_READER), !=, 0); |
34dc7c2f | 1353 | |
428870ff BB |
1354 | do { |
1355 | top = ztest_random(rvd->vdev_children); | |
1356 | tvd = rvd->vdev_child[top]; | |
a1d477c2 | 1357 | } while (!vdev_is_concrete(tvd) || (tvd->vdev_islog && !log_ok) || |
428870ff | 1358 | tvd->vdev_mg == NULL || tvd->vdev_mg->mg_class == NULL); |
34dc7c2f | 1359 | |
428870ff | 1360 | return (top); |
34dc7c2f BB |
1361 | } |
1362 | ||
428870ff BB |
1363 | static uint64_t |
1364 | ztest_random_dsl_prop(zfs_prop_t prop) | |
34dc7c2f | 1365 | { |
428870ff BB |
1366 | uint64_t value; |
1367 | ||
1368 | do { | |
1369 | value = zfs_prop_random_value(prop, ztest_random(-1ULL)); | |
1370 | } while (prop == ZFS_PROP_CHECKSUM && value == ZIO_CHECKSUM_OFF); | |
1371 | ||
1372 | return (value); | |
34dc7c2f BB |
1373 | } |
1374 | ||
34dc7c2f | 1375 | static int |
428870ff BB |
1376 | ztest_dsl_prop_set_uint64(char *osname, zfs_prop_t prop, uint64_t value, |
1377 | boolean_t inherit) | |
34dc7c2f | 1378 | { |
428870ff BB |
1379 | const char *propname = zfs_prop_to_name(prop); |
1380 | const char *valname; | |
40b84e7a | 1381 | char *setpoint; |
428870ff | 1382 | uint64_t curval; |
34dc7c2f BB |
1383 | int error; |
1384 | ||
13fe0198 MA |
1385 | error = dsl_prop_set_int(osname, propname, |
1386 | (inherit ? ZPROP_SRC_NONE : ZPROP_SRC_LOCAL), value); | |
34dc7c2f | 1387 | |
428870ff BB |
1388 | if (error == ENOSPC) { |
1389 | ztest_record_enospc(FTAG); | |
34dc7c2f BB |
1390 | return (error); |
1391 | } | |
c99c9001 | 1392 | ASSERT0(error); |
34dc7c2f | 1393 | |
40b84e7a | 1394 | setpoint = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
13fe0198 | 1395 | VERIFY0(dsl_prop_get_integer(osname, propname, &curval, setpoint)); |
428870ff | 1396 | |
c242c188 | 1397 | if (ztest_opts.zo_verbose >= 6) { |
6bec4351 TC |
1398 | int err; |
1399 | ||
1400 | err = zfs_prop_index_to_string(prop, curval, &valname); | |
1401 | if (err) | |
02730c33 BB |
1402 | (void) printf("%s %s = %llu at '%s'\n", osname, |
1403 | propname, (unsigned long long)curval, setpoint); | |
6bec4351 TC |
1404 | else |
1405 | (void) printf("%s %s = %s at '%s'\n", | |
1406 | osname, propname, valname, setpoint); | |
34dc7c2f | 1407 | } |
40b84e7a | 1408 | umem_free(setpoint, MAXPATHLEN); |
34dc7c2f BB |
1409 | |
1410 | return (error); | |
1411 | } | |
1412 | ||
1413 | static int | |
c242c188 | 1414 | ztest_spa_prop_set_uint64(zpool_prop_t prop, uint64_t value) |
34dc7c2f | 1415 | { |
c242c188 | 1416 | spa_t *spa = ztest_spa; |
428870ff | 1417 | nvlist_t *props = NULL; |
34dc7c2f BB |
1418 | int error; |
1419 | ||
bea3fc71 RM |
1420 | props = fnvlist_alloc(); |
1421 | fnvlist_add_uint64(props, zpool_prop_to_name(prop), value); | |
34dc7c2f | 1422 | |
428870ff BB |
1423 | error = spa_prop_set(spa, props); |
1424 | ||
bea3fc71 | 1425 | fnvlist_free(props); |
428870ff BB |
1426 | |
1427 | if (error == ENOSPC) { | |
1428 | ztest_record_enospc(FTAG); | |
34dc7c2f BB |
1429 | return (error); |
1430 | } | |
c99c9001 | 1431 | ASSERT0(error); |
34dc7c2f BB |
1432 | |
1433 | return (error); | |
1434 | } | |
1435 | ||
4807c0ba TC |
1436 | static int |
1437 | ztest_dmu_objset_own(const char *name, dmu_objset_type_t type, | |
1438 | boolean_t readonly, boolean_t decrypt, void *tag, objset_t **osp) | |
1439 | { | |
1440 | int err; | |
f44ad929 TC |
1441 | char *cp = NULL; |
1442 | char ddname[ZFS_MAX_DATASET_NAME_LEN]; | |
1443 | ||
1444 | strcpy(ddname, name); | |
1445 | cp = strchr(ddname, '@'); | |
1446 | if (cp != NULL) | |
1447 | *cp = '\0'; | |
4807c0ba TC |
1448 | |
1449 | err = dmu_objset_own(name, type, readonly, decrypt, tag, osp); | |
f44ad929 | 1450 | while (decrypt && err == EACCES) { |
4807c0ba TC |
1451 | dsl_crypto_params_t *dcp; |
1452 | nvlist_t *crypto_args = fnvlist_alloc(); | |
4807c0ba TC |
1453 | |
1454 | fnvlist_add_uint8_array(crypto_args, "wkeydata", | |
1455 | (uint8_t *)ztest_wkeydata, WRAPPING_KEY_LEN); | |
1456 | VERIFY0(dsl_crypto_params_create_nvlist(DCP_CMD_NONE, NULL, | |
1457 | crypto_args, &dcp)); | |
1458 | err = spa_keystore_load_wkey(ddname, dcp, B_FALSE); | |
f014700a MA |
1459 | /* |
1460 | * Note: if there was an error loading, the wkey was not | |
1461 | * consumed, and needs to be freed. | |
1462 | */ | |
1463 | dsl_crypto_params_free(dcp, (err != 0)); | |
4807c0ba TC |
1464 | fnvlist_free(crypto_args); |
1465 | ||
f44ad929 TC |
1466 | if (err == EINVAL) { |
1467 | /* | |
1468 | * We couldn't load a key for this dataset so try | |
1469 | * the parent. This loop will eventually hit the | |
1470 | * encryption root since ztest only makes clones | |
1471 | * as children of their origin datasets. | |
1472 | */ | |
1473 | cp = strrchr(ddname, '/'); | |
1474 | if (cp == NULL) | |
1475 | return (err); | |
1476 | ||
1477 | *cp = '\0'; | |
1478 | err = EACCES; | |
1479 | continue; | |
1480 | } else if (err != 0) { | |
1481 | break; | |
1482 | } | |
4807c0ba TC |
1483 | |
1484 | err = dmu_objset_own(name, type, readonly, decrypt, tag, osp); | |
f44ad929 | 1485 | break; |
4807c0ba TC |
1486 | } |
1487 | ||
1488 | return (err); | |
1489 | } | |
1490 | ||
428870ff BB |
1491 | static void |
1492 | ztest_rll_init(rll_t *rll) | |
1493 | { | |
1494 | rll->rll_writer = NULL; | |
1495 | rll->rll_readers = 0; | |
1e33ac1e BB |
1496 | mutex_init(&rll->rll_lock, NULL, MUTEX_DEFAULT, NULL); |
1497 | cv_init(&rll->rll_cv, NULL, CV_DEFAULT, NULL); | |
428870ff | 1498 | } |
34dc7c2f | 1499 | |
428870ff BB |
1500 | static void |
1501 | ztest_rll_destroy(rll_t *rll) | |
34dc7c2f | 1502 | { |
a9eaae06 RM |
1503 | ASSERT3P(rll->rll_writer, ==, NULL); |
1504 | ASSERT0(rll->rll_readers); | |
1e33ac1e BB |
1505 | mutex_destroy(&rll->rll_lock); |
1506 | cv_destroy(&rll->rll_cv); | |
428870ff | 1507 | } |
34dc7c2f | 1508 | |
428870ff BB |
1509 | static void |
1510 | ztest_rll_lock(rll_t *rll, rl_type_t type) | |
1511 | { | |
1e33ac1e | 1512 | mutex_enter(&rll->rll_lock); |
34dc7c2f | 1513 | |
428870ff BB |
1514 | if (type == RL_READER) { |
1515 | while (rll->rll_writer != NULL) | |
1e33ac1e | 1516 | (void) cv_wait(&rll->rll_cv, &rll->rll_lock); |
428870ff BB |
1517 | rll->rll_readers++; |
1518 | } else { | |
1519 | while (rll->rll_writer != NULL || rll->rll_readers) | |
1e33ac1e | 1520 | (void) cv_wait(&rll->rll_cv, &rll->rll_lock); |
428870ff BB |
1521 | rll->rll_writer = curthread; |
1522 | } | |
34dc7c2f | 1523 | |
1e33ac1e | 1524 | mutex_exit(&rll->rll_lock); |
428870ff | 1525 | } |
34dc7c2f | 1526 | |
428870ff BB |
1527 | static void |
1528 | ztest_rll_unlock(rll_t *rll) | |
1529 | { | |
1e33ac1e | 1530 | mutex_enter(&rll->rll_lock); |
34dc7c2f | 1531 | |
428870ff | 1532 | if (rll->rll_writer) { |
a9eaae06 | 1533 | ASSERT0(rll->rll_readers); |
428870ff BB |
1534 | rll->rll_writer = NULL; |
1535 | } else { | |
a9eaae06 RM |
1536 | ASSERT3S(rll->rll_readers, >, 0); |
1537 | ASSERT3P(rll->rll_writer, ==, NULL); | |
428870ff BB |
1538 | rll->rll_readers--; |
1539 | } | |
34dc7c2f | 1540 | |
428870ff | 1541 | if (rll->rll_writer == NULL && rll->rll_readers == 0) |
1e33ac1e | 1542 | cv_broadcast(&rll->rll_cv); |
428870ff | 1543 | |
1e33ac1e | 1544 | mutex_exit(&rll->rll_lock); |
34dc7c2f BB |
1545 | } |
1546 | ||
428870ff BB |
1547 | static void |
1548 | ztest_object_lock(ztest_ds_t *zd, uint64_t object, rl_type_t type) | |
b128c09f | 1549 | { |
428870ff | 1550 | rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; |
b128c09f | 1551 | |
428870ff BB |
1552 | ztest_rll_lock(rll, type); |
1553 | } | |
b128c09f | 1554 | |
428870ff BB |
1555 | static void |
1556 | ztest_object_unlock(ztest_ds_t *zd, uint64_t object) | |
1557 | { | |
1558 | rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; | |
b128c09f | 1559 | |
428870ff | 1560 | ztest_rll_unlock(rll); |
b128c09f BB |
1561 | } |
1562 | ||
5d43cc9a | 1563 | static rl_t * |
e3a07cd0 BP |
1564 | ztest_range_lock(ztest_ds_t *zd, uint64_t object, uint64_t offset, |
1565 | uint64_t size, rl_type_t type) | |
1566 | { | |
5d43cc9a MA |
1567 | uint64_t hash = object ^ (offset % (ZTEST_RANGE_LOCKS + 1)); |
1568 | rll_t *rll = &zd->zd_range_lock[hash & (ZTEST_RANGE_LOCKS - 1)]; | |
1569 | rl_t *rl; | |
1570 | ||
1571 | rl = umem_alloc(sizeof (*rl), UMEM_NOFAIL); | |
1572 | rl->rl_object = object; | |
1573 | rl->rl_offset = offset; | |
1574 | rl->rl_size = size; | |
1575 | rl->rl_lock = rll; | |
1576 | ||
1577 | ztest_rll_lock(rll, type); | |
1578 | ||
1579 | return (rl); | |
e3a07cd0 | 1580 | } |
34dc7c2f | 1581 | |
e3a07cd0 | 1582 | static void |
5d43cc9a | 1583 | ztest_range_unlock(rl_t *rl) |
e3a07cd0 | 1584 | { |
5d43cc9a MA |
1585 | rll_t *rll = rl->rl_lock; |
1586 | ||
1587 | ztest_rll_unlock(rll); | |
1588 | ||
1589 | umem_free(rl, sizeof (*rl)); | |
428870ff | 1590 | } |
34dc7c2f | 1591 | |
428870ff | 1592 | static void |
c242c188 | 1593 | ztest_zd_init(ztest_ds_t *zd, ztest_shared_ds_t *szd, objset_t *os) |
428870ff BB |
1594 | { |
1595 | zd->zd_os = os; | |
1596 | zd->zd_zilog = dmu_objset_zil(os); | |
c242c188 | 1597 | zd->zd_shared = szd; |
428870ff | 1598 | dmu_objset_name(os, zd->zd_name); |
d6320ddb | 1599 | int l; |
428870ff | 1600 | |
c242c188 CS |
1601 | if (zd->zd_shared != NULL) |
1602 | zd->zd_shared->zd_seq = 0; | |
1603 | ||
2d9142c9 | 1604 | VERIFY0(pthread_rwlock_init(&zd->zd_zilog_lock, NULL)); |
1e33ac1e | 1605 | mutex_init(&zd->zd_dirobj_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 1606 | |
d6320ddb | 1607 | for (l = 0; l < ZTEST_OBJECT_LOCKS; l++) |
428870ff | 1608 | ztest_rll_init(&zd->zd_object_lock[l]); |
34dc7c2f | 1609 | |
d6320ddb | 1610 | for (l = 0; l < ZTEST_RANGE_LOCKS; l++) |
5d43cc9a | 1611 | ztest_rll_init(&zd->zd_range_lock[l]); |
34dc7c2f BB |
1612 | } |
1613 | ||
428870ff BB |
1614 | static void |
1615 | ztest_zd_fini(ztest_ds_t *zd) | |
34dc7c2f | 1616 | { |
d6320ddb BB |
1617 | int l; |
1618 | ||
1e33ac1e | 1619 | mutex_destroy(&zd->zd_dirobj_lock); |
2d9142c9 | 1620 | (void) pthread_rwlock_destroy(&zd->zd_zilog_lock); |
34dc7c2f | 1621 | |
d6320ddb | 1622 | for (l = 0; l < ZTEST_OBJECT_LOCKS; l++) |
428870ff | 1623 | ztest_rll_destroy(&zd->zd_object_lock[l]); |
b128c09f | 1624 | |
d6320ddb | 1625 | for (l = 0; l < ZTEST_RANGE_LOCKS; l++) |
5d43cc9a | 1626 | ztest_rll_destroy(&zd->zd_range_lock[l]); |
428870ff | 1627 | } |
b128c09f | 1628 | |
428870ff | 1629 | #define TXG_MIGHTWAIT (ztest_random(10) == 0 ? TXG_NOWAIT : TXG_WAIT) |
b128c09f | 1630 | |
428870ff BB |
1631 | static uint64_t |
1632 | ztest_tx_assign(dmu_tx_t *tx, uint64_t txg_how, const char *tag) | |
1633 | { | |
1634 | uint64_t txg; | |
1635 | int error; | |
1636 | ||
1637 | /* | |
1638 | * Attempt to assign tx to some transaction group. | |
1639 | */ | |
1640 | error = dmu_tx_assign(tx, txg_how); | |
1641 | if (error) { | |
1642 | if (error == ERESTART) { | |
a9eaae06 | 1643 | ASSERT3U(txg_how, ==, TXG_NOWAIT); |
428870ff BB |
1644 | dmu_tx_wait(tx); |
1645 | } else { | |
1646 | ASSERT3U(error, ==, ENOSPC); | |
1647 | ztest_record_enospc(tag); | |
1648 | } | |
1649 | dmu_tx_abort(tx); | |
1650 | return (0); | |
1651 | } | |
1652 | txg = dmu_tx_get_txg(tx); | |
a9eaae06 | 1653 | ASSERT3U(txg, !=, 0); |
428870ff BB |
1654 | return (txg); |
1655 | } | |
1656 | ||
428870ff BB |
1657 | static void |
1658 | ztest_bt_generate(ztest_block_tag_t *bt, objset_t *os, uint64_t object, | |
50c957f7 NB |
1659 | uint64_t dnodesize, uint64_t offset, uint64_t gen, uint64_t txg, |
1660 | uint64_t crtxg) | |
428870ff BB |
1661 | { |
1662 | bt->bt_magic = BT_MAGIC; | |
1663 | bt->bt_objset = dmu_objset_id(os); | |
1664 | bt->bt_object = object; | |
50c957f7 | 1665 | bt->bt_dnodesize = dnodesize; |
428870ff BB |
1666 | bt->bt_offset = offset; |
1667 | bt->bt_gen = gen; | |
1668 | bt->bt_txg = txg; | |
1669 | bt->bt_crtxg = crtxg; | |
1670 | } | |
1671 | ||
1672 | static void | |
1673 | ztest_bt_verify(ztest_block_tag_t *bt, objset_t *os, uint64_t object, | |
50c957f7 NB |
1674 | uint64_t dnodesize, uint64_t offset, uint64_t gen, uint64_t txg, |
1675 | uint64_t crtxg) | |
428870ff | 1676 | { |
9b67f605 MA |
1677 | ASSERT3U(bt->bt_magic, ==, BT_MAGIC); |
1678 | ASSERT3U(bt->bt_objset, ==, dmu_objset_id(os)); | |
1679 | ASSERT3U(bt->bt_object, ==, object); | |
50c957f7 | 1680 | ASSERT3U(bt->bt_dnodesize, ==, dnodesize); |
9b67f605 MA |
1681 | ASSERT3U(bt->bt_offset, ==, offset); |
1682 | ASSERT3U(bt->bt_gen, <=, gen); | |
1683 | ASSERT3U(bt->bt_txg, <=, txg); | |
1684 | ASSERT3U(bt->bt_crtxg, ==, crtxg); | |
428870ff BB |
1685 | } |
1686 | ||
1687 | static ztest_block_tag_t * | |
1688 | ztest_bt_bonus(dmu_buf_t *db) | |
1689 | { | |
1690 | dmu_object_info_t doi; | |
1691 | ztest_block_tag_t *bt; | |
1692 | ||
1693 | dmu_object_info_from_db(db, &doi); | |
1694 | ASSERT3U(doi.doi_bonus_size, <=, db->db_size); | |
1695 | ASSERT3U(doi.doi_bonus_size, >=, sizeof (*bt)); | |
1696 | bt = (void *)((char *)db->db_data + doi.doi_bonus_size - sizeof (*bt)); | |
1697 | ||
1698 | return (bt); | |
1699 | } | |
1700 | ||
50c957f7 NB |
1701 | /* |
1702 | * Generate a token to fill up unused bonus buffer space. Try to make | |
1703 | * it unique to the object, generation, and offset to verify that data | |
1704 | * is not getting overwritten by data from other dnodes. | |
1705 | */ | |
1706 | #define ZTEST_BONUS_FILL_TOKEN(obj, ds, gen, offset) \ | |
1707 | (((ds) << 48) | ((gen) << 32) | ((obj) << 8) | (offset)) | |
1708 | ||
1709 | /* | |
1710 | * Fill up the unused bonus buffer region before the block tag with a | |
1711 | * verifiable pattern. Filling the whole bonus area with non-zero data | |
1712 | * helps ensure that all dnode traversal code properly skips the | |
1713 | * interior regions of large dnodes. | |
1714 | */ | |
65c7cc49 | 1715 | static void |
50c957f7 NB |
1716 | ztest_fill_unused_bonus(dmu_buf_t *db, void *end, uint64_t obj, |
1717 | objset_t *os, uint64_t gen) | |
1718 | { | |
1719 | uint64_t *bonusp; | |
1720 | ||
1721 | ASSERT(IS_P2ALIGNED((char *)end - (char *)db->db_data, 8)); | |
1722 | ||
1723 | for (bonusp = db->db_data; bonusp < (uint64_t *)end; bonusp++) { | |
1724 | uint64_t token = ZTEST_BONUS_FILL_TOKEN(obj, dmu_objset_id(os), | |
1725 | gen, bonusp - (uint64_t *)db->db_data); | |
1726 | *bonusp = token; | |
1727 | } | |
1728 | } | |
1729 | ||
1730 | /* | |
1731 | * Verify that the unused area of a bonus buffer is filled with the | |
1732 | * expected tokens. | |
1733 | */ | |
65c7cc49 | 1734 | static void |
50c957f7 NB |
1735 | ztest_verify_unused_bonus(dmu_buf_t *db, void *end, uint64_t obj, |
1736 | objset_t *os, uint64_t gen) | |
1737 | { | |
1738 | uint64_t *bonusp; | |
1739 | ||
1740 | for (bonusp = db->db_data; bonusp < (uint64_t *)end; bonusp++) { | |
1741 | uint64_t token = ZTEST_BONUS_FILL_TOKEN(obj, dmu_objset_id(os), | |
1742 | gen, bonusp - (uint64_t *)db->db_data); | |
1743 | VERIFY3U(*bonusp, ==, token); | |
1744 | } | |
1745 | } | |
1746 | ||
428870ff BB |
1747 | /* |
1748 | * ZIL logging ops | |
1749 | */ | |
1750 | ||
1751 | #define lrz_type lr_mode | |
1752 | #define lrz_blocksize lr_uid | |
1753 | #define lrz_ibshift lr_gid | |
1754 | #define lrz_bonustype lr_rdev | |
50c957f7 | 1755 | #define lrz_dnodesize lr_crtime[1] |
428870ff | 1756 | |
572e2857 | 1757 | static void |
428870ff BB |
1758 | ztest_log_create(ztest_ds_t *zd, dmu_tx_t *tx, lr_create_t *lr) |
1759 | { | |
1760 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1761 | size_t namesize = strlen(name) + 1; | |
1762 | itx_t *itx; | |
1763 | ||
1764 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1765 | return; |
428870ff BB |
1766 | |
1767 | itx = zil_itx_create(TX_CREATE, sizeof (*lr) + namesize); | |
1768 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1769 | sizeof (*lr) + namesize - sizeof (lr_t)); | |
1770 | ||
572e2857 | 1771 | zil_itx_assign(zd->zd_zilog, itx, tx); |
428870ff BB |
1772 | } |
1773 | ||
572e2857 BB |
1774 | static void |
1775 | ztest_log_remove(ztest_ds_t *zd, dmu_tx_t *tx, lr_remove_t *lr, uint64_t object) | |
428870ff BB |
1776 | { |
1777 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1778 | size_t namesize = strlen(name) + 1; | |
1779 | itx_t *itx; | |
1780 | ||
1781 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1782 | return; |
428870ff BB |
1783 | |
1784 | itx = zil_itx_create(TX_REMOVE, sizeof (*lr) + namesize); | |
1785 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1786 | sizeof (*lr) + namesize - sizeof (lr_t)); | |
1787 | ||
572e2857 BB |
1788 | itx->itx_oid = object; |
1789 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1790 | } |
1791 | ||
572e2857 | 1792 | static void |
428870ff BB |
1793 | ztest_log_write(ztest_ds_t *zd, dmu_tx_t *tx, lr_write_t *lr) |
1794 | { | |
1795 | itx_t *itx; | |
1796 | itx_wr_state_t write_state = ztest_random(WR_NUM_STATES); | |
1797 | ||
1798 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1799 | return; |
428870ff | 1800 | |
b8738257 | 1801 | if (lr->lr_length > zil_max_log_data(zd->zd_zilog)) |
428870ff BB |
1802 | write_state = WR_INDIRECT; |
1803 | ||
1804 | itx = zil_itx_create(TX_WRITE, | |
1805 | sizeof (*lr) + (write_state == WR_COPIED ? lr->lr_length : 0)); | |
1806 | ||
1807 | if (write_state == WR_COPIED && | |
1808 | dmu_read(zd->zd_os, lr->lr_foid, lr->lr_offset, lr->lr_length, | |
1809 | ((lr_write_t *)&itx->itx_lr) + 1, DMU_READ_NO_PREFETCH) != 0) { | |
1810 | zil_itx_destroy(itx); | |
1811 | itx = zil_itx_create(TX_WRITE, sizeof (*lr)); | |
1812 | write_state = WR_NEED_COPY; | |
1813 | } | |
1814 | itx->itx_private = zd; | |
1815 | itx->itx_wr_state = write_state; | |
1816 | itx->itx_sync = (ztest_random(8) == 0); | |
428870ff BB |
1817 | |
1818 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1819 | sizeof (*lr) - sizeof (lr_t)); | |
1820 | ||
572e2857 | 1821 | zil_itx_assign(zd->zd_zilog, itx, tx); |
428870ff BB |
1822 | } |
1823 | ||
572e2857 | 1824 | static void |
428870ff BB |
1825 | ztest_log_truncate(ztest_ds_t *zd, dmu_tx_t *tx, lr_truncate_t *lr) |
1826 | { | |
1827 | itx_t *itx; | |
1828 | ||
1829 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1830 | return; |
428870ff BB |
1831 | |
1832 | itx = zil_itx_create(TX_TRUNCATE, sizeof (*lr)); | |
1833 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1834 | sizeof (*lr) - sizeof (lr_t)); | |
1835 | ||
572e2857 BB |
1836 | itx->itx_sync = B_FALSE; |
1837 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1838 | } |
1839 | ||
572e2857 | 1840 | static void |
428870ff BB |
1841 | ztest_log_setattr(ztest_ds_t *zd, dmu_tx_t *tx, lr_setattr_t *lr) |
1842 | { | |
1843 | itx_t *itx; | |
1844 | ||
1845 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1846 | return; |
428870ff BB |
1847 | |
1848 | itx = zil_itx_create(TX_SETATTR, sizeof (*lr)); | |
1849 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1850 | sizeof (*lr) - sizeof (lr_t)); | |
1851 | ||
572e2857 BB |
1852 | itx->itx_sync = B_FALSE; |
1853 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1854 | } |
1855 | ||
1856 | /* | |
1857 | * ZIL replay ops | |
1858 | */ | |
1859 | static int | |
867959b5 | 1860 | ztest_replay_create(void *arg1, void *arg2, boolean_t byteswap) |
428870ff | 1861 | { |
867959b5 BB |
1862 | ztest_ds_t *zd = arg1; |
1863 | lr_create_t *lr = arg2; | |
428870ff BB |
1864 | char *name = (void *)(lr + 1); /* name follows lr */ |
1865 | objset_t *os = zd->zd_os; | |
1866 | ztest_block_tag_t *bbt; | |
1867 | dmu_buf_t *db; | |
1868 | dmu_tx_t *tx; | |
1869 | uint64_t txg; | |
1870 | int error = 0; | |
50c957f7 | 1871 | int bonuslen; |
428870ff BB |
1872 | |
1873 | if (byteswap) | |
1874 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1875 | ||
a9eaae06 RM |
1876 | ASSERT3U(lr->lr_doid, ==, ZTEST_DIROBJ); |
1877 | ASSERT3S(name[0], !=, '\0'); | |
428870ff BB |
1878 | |
1879 | tx = dmu_tx_create(os); | |
1880 | ||
1881 | dmu_tx_hold_zap(tx, lr->lr_doid, B_TRUE, name); | |
1882 | ||
1883 | if (lr->lrz_type == DMU_OT_ZAP_OTHER) { | |
1884 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, B_TRUE, NULL); | |
1885 | } else { | |
1886 | dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT); | |
1887 | } | |
1888 | ||
1889 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1890 | if (txg == 0) | |
1891 | return (ENOSPC); | |
1892 | ||
a9eaae06 | 1893 | ASSERT3U(dmu_objset_zil(os)->zl_replay, ==, !!lr->lr_foid); |
50c957f7 | 1894 | bonuslen = DN_BONUS_SIZE(lr->lrz_dnodesize); |
428870ff BB |
1895 | |
1896 | if (lr->lrz_type == DMU_OT_ZAP_OTHER) { | |
1897 | if (lr->lr_foid == 0) { | |
50c957f7 | 1898 | lr->lr_foid = zap_create_dnsize(os, |
428870ff | 1899 | lr->lrz_type, lr->lrz_bonustype, |
50c957f7 | 1900 | bonuslen, lr->lrz_dnodesize, tx); |
428870ff | 1901 | } else { |
50c957f7 | 1902 | error = zap_create_claim_dnsize(os, lr->lr_foid, |
428870ff | 1903 | lr->lrz_type, lr->lrz_bonustype, |
50c957f7 | 1904 | bonuslen, lr->lrz_dnodesize, tx); |
428870ff BB |
1905 | } |
1906 | } else { | |
1907 | if (lr->lr_foid == 0) { | |
50c957f7 | 1908 | lr->lr_foid = dmu_object_alloc_dnsize(os, |
428870ff | 1909 | lr->lrz_type, 0, lr->lrz_bonustype, |
50c957f7 | 1910 | bonuslen, lr->lrz_dnodesize, tx); |
428870ff | 1911 | } else { |
50c957f7 | 1912 | error = dmu_object_claim_dnsize(os, lr->lr_foid, |
428870ff | 1913 | lr->lrz_type, 0, lr->lrz_bonustype, |
50c957f7 | 1914 | bonuslen, lr->lrz_dnodesize, tx); |
428870ff BB |
1915 | } |
1916 | } | |
1917 | ||
1918 | if (error) { | |
1919 | ASSERT3U(error, ==, EEXIST); | |
1920 | ASSERT(zd->zd_zilog->zl_replay); | |
1921 | dmu_tx_commit(tx); | |
1922 | return (error); | |
1923 | } | |
1924 | ||
a9eaae06 | 1925 | ASSERT3U(lr->lr_foid, !=, 0); |
428870ff BB |
1926 | |
1927 | if (lr->lrz_type != DMU_OT_ZAP_OTHER) | |
a9eaae06 | 1928 | VERIFY0(dmu_object_set_blocksize(os, lr->lr_foid, |
428870ff BB |
1929 | lr->lrz_blocksize, lr->lrz_ibshift, tx)); |
1930 | ||
a9eaae06 | 1931 | VERIFY0(dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); |
428870ff BB |
1932 | bbt = ztest_bt_bonus(db); |
1933 | dmu_buf_will_dirty(db, tx); | |
50c957f7 NB |
1934 | ztest_bt_generate(bbt, os, lr->lr_foid, lr->lrz_dnodesize, -1ULL, |
1935 | lr->lr_gen, txg, txg); | |
1936 | ztest_fill_unused_bonus(db, bbt, lr->lr_foid, os, lr->lr_gen); | |
428870ff BB |
1937 | dmu_buf_rele(db, FTAG); |
1938 | ||
a9eaae06 | 1939 | VERIFY0(zap_add(os, lr->lr_doid, name, sizeof (uint64_t), 1, |
428870ff BB |
1940 | &lr->lr_foid, tx)); |
1941 | ||
1942 | (void) ztest_log_create(zd, tx, lr); | |
1943 | ||
1944 | dmu_tx_commit(tx); | |
1945 | ||
1946 | return (0); | |
1947 | } | |
1948 | ||
1949 | static int | |
867959b5 | 1950 | ztest_replay_remove(void *arg1, void *arg2, boolean_t byteswap) |
428870ff | 1951 | { |
867959b5 BB |
1952 | ztest_ds_t *zd = arg1; |
1953 | lr_remove_t *lr = arg2; | |
428870ff BB |
1954 | char *name = (void *)(lr + 1); /* name follows lr */ |
1955 | objset_t *os = zd->zd_os; | |
1956 | dmu_object_info_t doi; | |
1957 | dmu_tx_t *tx; | |
1958 | uint64_t object, txg; | |
1959 | ||
1960 | if (byteswap) | |
1961 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1962 | ||
a9eaae06 RM |
1963 | ASSERT3U(lr->lr_doid, ==, ZTEST_DIROBJ); |
1964 | ASSERT3S(name[0], !=, '\0'); | |
428870ff | 1965 | |
a9eaae06 | 1966 | VERIFY0( |
428870ff | 1967 | zap_lookup(os, lr->lr_doid, name, sizeof (object), 1, &object)); |
a9eaae06 | 1968 | ASSERT3U(object, !=, 0); |
428870ff BB |
1969 | |
1970 | ztest_object_lock(zd, object, RL_WRITER); | |
1971 | ||
a9eaae06 | 1972 | VERIFY0(dmu_object_info(os, object, &doi)); |
428870ff BB |
1973 | |
1974 | tx = dmu_tx_create(os); | |
1975 | ||
1976 | dmu_tx_hold_zap(tx, lr->lr_doid, B_FALSE, name); | |
1977 | dmu_tx_hold_free(tx, object, 0, DMU_OBJECT_END); | |
1978 | ||
1979 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1980 | if (txg == 0) { | |
1981 | ztest_object_unlock(zd, object); | |
1982 | return (ENOSPC); | |
1983 | } | |
1984 | ||
1985 | if (doi.doi_type == DMU_OT_ZAP_OTHER) { | |
a9eaae06 | 1986 | VERIFY0(zap_destroy(os, object, tx)); |
428870ff | 1987 | } else { |
a9eaae06 | 1988 | VERIFY0(dmu_object_free(os, object, tx)); |
428870ff BB |
1989 | } |
1990 | ||
a9eaae06 | 1991 | VERIFY0(zap_remove(os, lr->lr_doid, name, tx)); |
428870ff | 1992 | |
572e2857 | 1993 | (void) ztest_log_remove(zd, tx, lr, object); |
428870ff BB |
1994 | |
1995 | dmu_tx_commit(tx); | |
1996 | ||
1997 | ztest_object_unlock(zd, object); | |
1998 | ||
1999 | return (0); | |
2000 | } | |
2001 | ||
2002 | static int | |
867959b5 | 2003 | ztest_replay_write(void *arg1, void *arg2, boolean_t byteswap) |
428870ff | 2004 | { |
867959b5 BB |
2005 | ztest_ds_t *zd = arg1; |
2006 | lr_write_t *lr = arg2; | |
428870ff BB |
2007 | objset_t *os = zd->zd_os; |
2008 | void *data = lr + 1; /* data follows lr */ | |
2009 | uint64_t offset, length; | |
2010 | ztest_block_tag_t *bt = data; | |
2011 | ztest_block_tag_t *bbt; | |
2012 | uint64_t gen, txg, lrtxg, crtxg; | |
2013 | dmu_object_info_t doi; | |
2014 | dmu_tx_t *tx; | |
2015 | dmu_buf_t *db; | |
2016 | arc_buf_t *abuf = NULL; | |
5d43cc9a | 2017 | rl_t *rl; |
428870ff BB |
2018 | |
2019 | if (byteswap) | |
2020 | byteswap_uint64_array(lr, sizeof (*lr)); | |
2021 | ||
2022 | offset = lr->lr_offset; | |
2023 | length = lr->lr_length; | |
2024 | ||
2025 | /* If it's a dmu_sync() block, write the whole block */ | |
2026 | if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) { | |
2027 | uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr); | |
2028 | if (length < blocksize) { | |
2029 | offset -= offset % blocksize; | |
2030 | length = blocksize; | |
2031 | } | |
2032 | } | |
2033 | ||
2034 | if (bt->bt_magic == BSWAP_64(BT_MAGIC)) | |
2035 | byteswap_uint64_array(bt, sizeof (*bt)); | |
2036 | ||
2037 | if (bt->bt_magic != BT_MAGIC) | |
2038 | bt = NULL; | |
2039 | ||
2040 | ztest_object_lock(zd, lr->lr_foid, RL_READER); | |
2041 | rl = ztest_range_lock(zd, lr->lr_foid, offset, length, RL_WRITER); | |
2042 | ||
a9eaae06 | 2043 | VERIFY0(dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); |
428870ff BB |
2044 | |
2045 | dmu_object_info_from_db(db, &doi); | |
2046 | ||
2047 | bbt = ztest_bt_bonus(db); | |
2048 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
2049 | gen = bbt->bt_gen; | |
2050 | crtxg = bbt->bt_crtxg; | |
2051 | lrtxg = lr->lr_common.lrc_txg; | |
2052 | ||
2053 | tx = dmu_tx_create(os); | |
2054 | ||
2055 | dmu_tx_hold_write(tx, lr->lr_foid, offset, length); | |
2056 | ||
2057 | if (ztest_random(8) == 0 && length == doi.doi_data_block_size && | |
2058 | P2PHASE(offset, length) == 0) | |
2059 | abuf = dmu_request_arcbuf(db, length); | |
2060 | ||
2061 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
2062 | if (txg == 0) { | |
2063 | if (abuf != NULL) | |
2064 | dmu_return_arcbuf(abuf); | |
2065 | dmu_buf_rele(db, FTAG); | |
5d43cc9a | 2066 | ztest_range_unlock(rl); |
428870ff BB |
2067 | ztest_object_unlock(zd, lr->lr_foid); |
2068 | return (ENOSPC); | |
2069 | } | |
2070 | ||
2071 | if (bt != NULL) { | |
2072 | /* | |
2073 | * Usually, verify the old data before writing new data -- | |
2074 | * but not always, because we also want to verify correct | |
2075 | * behavior when the data was not recently read into cache. | |
2076 | */ | |
a9eaae06 | 2077 | ASSERT0(offset % doi.doi_data_block_size); |
428870ff BB |
2078 | if (ztest_random(4) != 0) { |
2079 | int prefetch = ztest_random(2) ? | |
2080 | DMU_READ_PREFETCH : DMU_READ_NO_PREFETCH; | |
2081 | ztest_block_tag_t rbt; | |
2082 | ||
2083 | VERIFY(dmu_read(os, lr->lr_foid, offset, | |
2084 | sizeof (rbt), &rbt, prefetch) == 0); | |
2085 | if (rbt.bt_magic == BT_MAGIC) { | |
50c957f7 | 2086 | ztest_bt_verify(&rbt, os, lr->lr_foid, 0, |
428870ff BB |
2087 | offset, gen, txg, crtxg); |
2088 | } | |
2089 | } | |
2090 | ||
2091 | /* | |
2092 | * Writes can appear to be newer than the bonus buffer because | |
2093 | * the ztest_get_data() callback does a dmu_read() of the | |
2094 | * open-context data, which may be different than the data | |
2095 | * as it was when the write was generated. | |
2096 | */ | |
2097 | if (zd->zd_zilog->zl_replay) { | |
50c957f7 | 2098 | ztest_bt_verify(bt, os, lr->lr_foid, 0, offset, |
428870ff BB |
2099 | MAX(gen, bt->bt_gen), MAX(txg, lrtxg), |
2100 | bt->bt_crtxg); | |
2101 | } | |
2102 | ||
2103 | /* | |
2104 | * Set the bt's gen/txg to the bonus buffer's gen/txg | |
2105 | * so that all of the usual ASSERTs will work. | |
2106 | */ | |
50c957f7 NB |
2107 | ztest_bt_generate(bt, os, lr->lr_foid, 0, offset, gen, txg, |
2108 | crtxg); | |
428870ff BB |
2109 | } |
2110 | ||
2111 | if (abuf == NULL) { | |
2112 | dmu_write(os, lr->lr_foid, offset, length, data, tx); | |
2113 | } else { | |
2114 | bcopy(data, abuf->b_data, length); | |
440a3eb9 | 2115 | dmu_assign_arcbuf_by_dbuf(db, offset, abuf, tx); |
428870ff BB |
2116 | } |
2117 | ||
2118 | (void) ztest_log_write(zd, tx, lr); | |
2119 | ||
2120 | dmu_buf_rele(db, FTAG); | |
2121 | ||
2122 | dmu_tx_commit(tx); | |
2123 | ||
5d43cc9a | 2124 | ztest_range_unlock(rl); |
428870ff BB |
2125 | ztest_object_unlock(zd, lr->lr_foid); |
2126 | ||
2127 | return (0); | |
2128 | } | |
2129 | ||
2130 | static int | |
867959b5 | 2131 | ztest_replay_truncate(void *arg1, void *arg2, boolean_t byteswap) |
428870ff | 2132 | { |
867959b5 BB |
2133 | ztest_ds_t *zd = arg1; |
2134 | lr_truncate_t *lr = arg2; | |
428870ff BB |
2135 | objset_t *os = zd->zd_os; |
2136 | dmu_tx_t *tx; | |
2137 | uint64_t txg; | |
5d43cc9a | 2138 | rl_t *rl; |
428870ff BB |
2139 | |
2140 | if (byteswap) | |
2141 | byteswap_uint64_array(lr, sizeof (*lr)); | |
2142 | ||
2143 | ztest_object_lock(zd, lr->lr_foid, RL_READER); | |
2144 | rl = ztest_range_lock(zd, lr->lr_foid, lr->lr_offset, lr->lr_length, | |
2145 | RL_WRITER); | |
2146 | ||
2147 | tx = dmu_tx_create(os); | |
2148 | ||
2149 | dmu_tx_hold_free(tx, lr->lr_foid, lr->lr_offset, lr->lr_length); | |
2150 | ||
2151 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
2152 | if (txg == 0) { | |
5d43cc9a | 2153 | ztest_range_unlock(rl); |
428870ff BB |
2154 | ztest_object_unlock(zd, lr->lr_foid); |
2155 | return (ENOSPC); | |
2156 | } | |
2157 | ||
a9eaae06 RM |
2158 | VERIFY0(dmu_free_range(os, lr->lr_foid, lr->lr_offset, |
2159 | lr->lr_length, tx)); | |
428870ff BB |
2160 | |
2161 | (void) ztest_log_truncate(zd, tx, lr); | |
2162 | ||
2163 | dmu_tx_commit(tx); | |
2164 | ||
5d43cc9a | 2165 | ztest_range_unlock(rl); |
428870ff BB |
2166 | ztest_object_unlock(zd, lr->lr_foid); |
2167 | ||
2168 | return (0); | |
2169 | } | |
2170 | ||
2171 | static int | |
867959b5 | 2172 | ztest_replay_setattr(void *arg1, void *arg2, boolean_t byteswap) |
428870ff | 2173 | { |
867959b5 BB |
2174 | ztest_ds_t *zd = arg1; |
2175 | lr_setattr_t *lr = arg2; | |
428870ff BB |
2176 | objset_t *os = zd->zd_os; |
2177 | dmu_tx_t *tx; | |
2178 | dmu_buf_t *db; | |
2179 | ztest_block_tag_t *bbt; | |
50c957f7 | 2180 | uint64_t txg, lrtxg, crtxg, dnodesize; |
428870ff BB |
2181 | |
2182 | if (byteswap) | |
2183 | byteswap_uint64_array(lr, sizeof (*lr)); | |
2184 | ||
2185 | ztest_object_lock(zd, lr->lr_foid, RL_WRITER); | |
2186 | ||
a9eaae06 | 2187 | VERIFY0(dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); |
428870ff BB |
2188 | |
2189 | tx = dmu_tx_create(os); | |
2190 | dmu_tx_hold_bonus(tx, lr->lr_foid); | |
2191 | ||
2192 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
2193 | if (txg == 0) { | |
2194 | dmu_buf_rele(db, FTAG); | |
2195 | ztest_object_unlock(zd, lr->lr_foid); | |
2196 | return (ENOSPC); | |
2197 | } | |
2198 | ||
2199 | bbt = ztest_bt_bonus(db); | |
2200 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
2201 | crtxg = bbt->bt_crtxg; | |
2202 | lrtxg = lr->lr_common.lrc_txg; | |
50c957f7 | 2203 | dnodesize = bbt->bt_dnodesize; |
428870ff BB |
2204 | |
2205 | if (zd->zd_zilog->zl_replay) { | |
a9eaae06 RM |
2206 | ASSERT3U(lr->lr_size, !=, 0); |
2207 | ASSERT3U(lr->lr_mode, !=, 0); | |
2208 | ASSERT3U(lrtxg, !=, 0); | |
428870ff BB |
2209 | } else { |
2210 | /* | |
2211 | * Randomly change the size and increment the generation. | |
2212 | */ | |
2213 | lr->lr_size = (ztest_random(db->db_size / sizeof (*bbt)) + 1) * | |
2214 | sizeof (*bbt); | |
2215 | lr->lr_mode = bbt->bt_gen + 1; | |
a9eaae06 | 2216 | ASSERT0(lrtxg); |
428870ff BB |
2217 | } |
2218 | ||
2219 | /* | |
2220 | * Verify that the current bonus buffer is not newer than our txg. | |
2221 | */ | |
50c957f7 | 2222 | ztest_bt_verify(bbt, os, lr->lr_foid, dnodesize, -1ULL, lr->lr_mode, |
428870ff BB |
2223 | MAX(txg, lrtxg), crtxg); |
2224 | ||
2225 | dmu_buf_will_dirty(db, tx); | |
2226 | ||
2227 | ASSERT3U(lr->lr_size, >=, sizeof (*bbt)); | |
2228 | ASSERT3U(lr->lr_size, <=, db->db_size); | |
c99c9001 | 2229 | VERIFY0(dmu_set_bonus(db, lr->lr_size, tx)); |
428870ff BB |
2230 | bbt = ztest_bt_bonus(db); |
2231 | ||
50c957f7 NB |
2232 | ztest_bt_generate(bbt, os, lr->lr_foid, dnodesize, -1ULL, lr->lr_mode, |
2233 | txg, crtxg); | |
2234 | ztest_fill_unused_bonus(db, bbt, lr->lr_foid, os, bbt->bt_gen); | |
428870ff BB |
2235 | dmu_buf_rele(db, FTAG); |
2236 | ||
2237 | (void) ztest_log_setattr(zd, tx, lr); | |
2238 | ||
2239 | dmu_tx_commit(tx); | |
2240 | ||
2241 | ztest_object_unlock(zd, lr->lr_foid); | |
2242 | ||
2243 | return (0); | |
2244 | } | |
2245 | ||
867959b5 BB |
2246 | zil_replay_func_t *ztest_replay_vector[TX_MAX_TYPE] = { |
2247 | NULL, /* 0 no such transaction type */ | |
2248 | ztest_replay_create, /* TX_CREATE */ | |
2249 | NULL, /* TX_MKDIR */ | |
2250 | NULL, /* TX_MKXATTR */ | |
2251 | NULL, /* TX_SYMLINK */ | |
2252 | ztest_replay_remove, /* TX_REMOVE */ | |
2253 | NULL, /* TX_RMDIR */ | |
2254 | NULL, /* TX_LINK */ | |
2255 | NULL, /* TX_RENAME */ | |
2256 | ztest_replay_write, /* TX_WRITE */ | |
2257 | ztest_replay_truncate, /* TX_TRUNCATE */ | |
2258 | ztest_replay_setattr, /* TX_SETATTR */ | |
2259 | NULL, /* TX_ACL */ | |
2260 | NULL, /* TX_CREATE_ACL */ | |
2261 | NULL, /* TX_CREATE_ATTR */ | |
2262 | NULL, /* TX_CREATE_ACL_ATTR */ | |
2263 | NULL, /* TX_MKDIR_ACL */ | |
2264 | NULL, /* TX_MKDIR_ATTR */ | |
2265 | NULL, /* TX_MKDIR_ACL_ATTR */ | |
2266 | NULL, /* TX_WRITE2 */ | |
428870ff BB |
2267 | }; |
2268 | ||
2269 | /* | |
2270 | * ZIL get_data callbacks | |
2271 | */ | |
2272 | ||
900d09b2 | 2273 | /* ARGSUSED */ |
428870ff BB |
2274 | static void |
2275 | ztest_get_done(zgd_t *zgd, int error) | |
2276 | { | |
5d43cc9a MA |
2277 | ztest_ds_t *zd = zgd->zgd_private; |
2278 | uint64_t object = ((rl_t *)zgd->zgd_lr)->rl_object; | |
428870ff BB |
2279 | |
2280 | if (zgd->zgd_db) | |
2281 | dmu_buf_rele(zgd->zgd_db, zgd); | |
2282 | ||
5d43cc9a | 2283 | ztest_range_unlock((rl_t *)zgd->zgd_lr); |
428870ff BB |
2284 | ztest_object_unlock(zd, object); |
2285 | ||
428870ff BB |
2286 | umem_free(zgd, sizeof (*zgd)); |
2287 | } | |
2288 | ||
2289 | static int | |
1ce23dca PS |
2290 | ztest_get_data(void *arg, lr_write_t *lr, char *buf, struct lwb *lwb, |
2291 | zio_t *zio) | |
428870ff BB |
2292 | { |
2293 | ztest_ds_t *zd = arg; | |
2294 | objset_t *os = zd->zd_os; | |
2295 | uint64_t object = lr->lr_foid; | |
2296 | uint64_t offset = lr->lr_offset; | |
2297 | uint64_t size = lr->lr_length; | |
428870ff BB |
2298 | uint64_t txg = lr->lr_common.lrc_txg; |
2299 | uint64_t crtxg; | |
2300 | dmu_object_info_t doi; | |
2301 | dmu_buf_t *db; | |
2302 | zgd_t *zgd; | |
2303 | int error; | |
2304 | ||
1ce23dca PS |
2305 | ASSERT3P(lwb, !=, NULL); |
2306 | ASSERT3P(zio, !=, NULL); | |
2307 | ASSERT3U(size, !=, 0); | |
2308 | ||
428870ff BB |
2309 | ztest_object_lock(zd, object, RL_READER); |
2310 | error = dmu_bonus_hold(os, object, FTAG, &db); | |
2311 | if (error) { | |
2312 | ztest_object_unlock(zd, object); | |
2313 | return (error); | |
2314 | } | |
2315 | ||
2316 | crtxg = ztest_bt_bonus(db)->bt_crtxg; | |
2317 | ||
2318 | if (crtxg == 0 || crtxg > txg) { | |
2319 | dmu_buf_rele(db, FTAG); | |
2320 | ztest_object_unlock(zd, object); | |
2321 | return (ENOENT); | |
2322 | } | |
2323 | ||
2324 | dmu_object_info_from_db(db, &doi); | |
2325 | dmu_buf_rele(db, FTAG); | |
2326 | db = NULL; | |
2327 | ||
2328 | zgd = umem_zalloc(sizeof (*zgd), UMEM_NOFAIL); | |
1ce23dca | 2329 | zgd->zgd_lwb = lwb; |
5d43cc9a | 2330 | zgd->zgd_private = zd; |
428870ff BB |
2331 | |
2332 | if (buf != NULL) { /* immediate write */ | |
bd4dde8e | 2333 | zgd->zgd_lr = (struct zfs_locked_range *)ztest_range_lock(zd, |
5d43cc9a | 2334 | object, offset, size, RL_READER); |
428870ff BB |
2335 | |
2336 | error = dmu_read(os, object, offset, size, buf, | |
2337 | DMU_READ_NO_PREFETCH); | |
a9eaae06 | 2338 | ASSERT0(error); |
428870ff BB |
2339 | } else { |
2340 | size = doi.doi_data_block_size; | |
2341 | if (ISP2(size)) { | |
2342 | offset = P2ALIGN(offset, size); | |
2343 | } else { | |
a9eaae06 | 2344 | ASSERT3U(offset, <, size); |
428870ff BB |
2345 | offset = 0; |
2346 | } | |
2347 | ||
bd4dde8e | 2348 | zgd->zgd_lr = (struct zfs_locked_range *)ztest_range_lock(zd, |
5d43cc9a | 2349 | object, offset, size, RL_READER); |
428870ff BB |
2350 | |
2351 | error = dmu_buf_hold(os, object, offset, zgd, &db, | |
2352 | DMU_READ_NO_PREFETCH); | |
2353 | ||
2354 | if (error == 0) { | |
02dc43bc | 2355 | blkptr_t *bp = &lr->lr_blkptr; |
03c6040b | 2356 | |
428870ff BB |
2357 | zgd->zgd_db = db; |
2358 | zgd->zgd_bp = bp; | |
2359 | ||
a9eaae06 RM |
2360 | ASSERT3U(db->db_offset, ==, offset); |
2361 | ASSERT3U(db->db_size, ==, size); | |
428870ff BB |
2362 | |
2363 | error = dmu_sync(zio, lr->lr_common.lrc_txg, | |
2364 | ztest_get_done, zgd); | |
2365 | ||
2366 | if (error == 0) | |
2367 | return (0); | |
2368 | } | |
2369 | } | |
2370 | ||
2371 | ztest_get_done(zgd, error); | |
2372 | ||
2373 | return (error); | |
2374 | } | |
2375 | ||
2376 | static void * | |
2377 | ztest_lr_alloc(size_t lrsize, char *name) | |
2378 | { | |
2379 | char *lr; | |
2380 | size_t namesize = name ? strlen(name) + 1 : 0; | |
2381 | ||
2382 | lr = umem_zalloc(lrsize + namesize, UMEM_NOFAIL); | |
2383 | ||
2384 | if (name) | |
2385 | bcopy(name, lr + lrsize, namesize); | |
2386 | ||
2387 | return (lr); | |
2388 | } | |
2389 | ||
65c7cc49 | 2390 | static void |
428870ff BB |
2391 | ztest_lr_free(void *lr, size_t lrsize, char *name) |
2392 | { | |
2393 | size_t namesize = name ? strlen(name) + 1 : 0; | |
2394 | ||
2395 | umem_free(lr, lrsize + namesize); | |
2396 | } | |
2397 | ||
2398 | /* | |
2399 | * Lookup a bunch of objects. Returns the number of objects not found. | |
2400 | */ | |
2401 | static int | |
2402 | ztest_lookup(ztest_ds_t *zd, ztest_od_t *od, int count) | |
2403 | { | |
2404 | int missing = 0; | |
2405 | int error; | |
d6320ddb | 2406 | int i; |
428870ff | 2407 | |
c25b8f99 | 2408 | ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock)); |
428870ff | 2409 | |
d6320ddb | 2410 | for (i = 0; i < count; i++, od++) { |
428870ff BB |
2411 | od->od_object = 0; |
2412 | error = zap_lookup(zd->zd_os, od->od_dir, od->od_name, | |
2413 | sizeof (uint64_t), 1, &od->od_object); | |
2414 | if (error) { | |
a9eaae06 RM |
2415 | ASSERT3S(error, ==, ENOENT); |
2416 | ASSERT0(od->od_object); | |
428870ff BB |
2417 | missing++; |
2418 | } else { | |
2419 | dmu_buf_t *db; | |
2420 | ztest_block_tag_t *bbt; | |
2421 | dmu_object_info_t doi; | |
2422 | ||
a9eaae06 RM |
2423 | ASSERT3U(od->od_object, !=, 0); |
2424 | ASSERT0(missing); /* there should be no gaps */ | |
428870ff BB |
2425 | |
2426 | ztest_object_lock(zd, od->od_object, RL_READER); | |
a9eaae06 RM |
2427 | VERIFY0(dmu_bonus_hold(zd->zd_os, od->od_object, |
2428 | FTAG, &db)); | |
428870ff BB |
2429 | dmu_object_info_from_db(db, &doi); |
2430 | bbt = ztest_bt_bonus(db); | |
2431 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
2432 | od->od_type = doi.doi_type; | |
2433 | od->od_blocksize = doi.doi_data_block_size; | |
2434 | od->od_gen = bbt->bt_gen; | |
2435 | dmu_buf_rele(db, FTAG); | |
2436 | ztest_object_unlock(zd, od->od_object); | |
2437 | } | |
2438 | } | |
2439 | ||
2440 | return (missing); | |
2441 | } | |
2442 | ||
2443 | static int | |
2444 | ztest_create(ztest_ds_t *zd, ztest_od_t *od, int count) | |
2445 | { | |
2446 | int missing = 0; | |
d6320ddb | 2447 | int i; |
428870ff | 2448 | |
c25b8f99 | 2449 | ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock)); |
428870ff | 2450 | |
d6320ddb | 2451 | for (i = 0; i < count; i++, od++) { |
428870ff BB |
2452 | if (missing) { |
2453 | od->od_object = 0; | |
2454 | missing++; | |
2455 | continue; | |
2456 | } | |
2457 | ||
2458 | lr_create_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name); | |
2459 | ||
2460 | lr->lr_doid = od->od_dir; | |
2461 | lr->lr_foid = 0; /* 0 to allocate, > 0 to claim */ | |
2462 | lr->lrz_type = od->od_crtype; | |
2463 | lr->lrz_blocksize = od->od_crblocksize; | |
2464 | lr->lrz_ibshift = ztest_random_ibshift(); | |
2465 | lr->lrz_bonustype = DMU_OT_UINT64_OTHER; | |
50c957f7 | 2466 | lr->lrz_dnodesize = od->od_crdnodesize; |
428870ff BB |
2467 | lr->lr_gen = od->od_crgen; |
2468 | lr->lr_crtime[0] = time(NULL); | |
2469 | ||
2470 | if (ztest_replay_create(zd, lr, B_FALSE) != 0) { | |
a9eaae06 | 2471 | ASSERT0(missing); |
428870ff BB |
2472 | od->od_object = 0; |
2473 | missing++; | |
2474 | } else { | |
2475 | od->od_object = lr->lr_foid; | |
2476 | od->od_type = od->od_crtype; | |
2477 | od->od_blocksize = od->od_crblocksize; | |
2478 | od->od_gen = od->od_crgen; | |
a9eaae06 | 2479 | ASSERT3U(od->od_object, !=, 0); |
428870ff BB |
2480 | } |
2481 | ||
2482 | ztest_lr_free(lr, sizeof (*lr), od->od_name); | |
2483 | } | |
2484 | ||
2485 | return (missing); | |
2486 | } | |
2487 | ||
2488 | static int | |
2489 | ztest_remove(ztest_ds_t *zd, ztest_od_t *od, int count) | |
2490 | { | |
2491 | int missing = 0; | |
2492 | int error; | |
d6320ddb | 2493 | int i; |
428870ff | 2494 | |
c25b8f99 | 2495 | ASSERT(MUTEX_HELD(&zd->zd_dirobj_lock)); |
428870ff BB |
2496 | |
2497 | od += count - 1; | |
2498 | ||
d6320ddb | 2499 | for (i = count - 1; i >= 0; i--, od--) { |
428870ff BB |
2500 | if (missing) { |
2501 | missing++; | |
2502 | continue; | |
2503 | } | |
2504 | ||
03c6040b GW |
2505 | /* |
2506 | * No object was found. | |
2507 | */ | |
428870ff BB |
2508 | if (od->od_object == 0) |
2509 | continue; | |
2510 | ||
2511 | lr_remove_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name); | |
2512 | ||
2513 | lr->lr_doid = od->od_dir; | |
2514 | ||
2515 | if ((error = ztest_replay_remove(zd, lr, B_FALSE)) != 0) { | |
2516 | ASSERT3U(error, ==, ENOSPC); | |
2517 | missing++; | |
2518 | } else { | |
2519 | od->od_object = 0; | |
2520 | } | |
2521 | ztest_lr_free(lr, sizeof (*lr), od->od_name); | |
2522 | } | |
2523 | ||
2524 | return (missing); | |
2525 | } | |
2526 | ||
2527 | static int | |
2528 | ztest_write(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size, | |
2529 | void *data) | |
2530 | { | |
2531 | lr_write_t *lr; | |
2532 | int error; | |
2533 | ||
2534 | lr = ztest_lr_alloc(sizeof (*lr) + size, NULL); | |
2535 | ||
2536 | lr->lr_foid = object; | |
2537 | lr->lr_offset = offset; | |
2538 | lr->lr_length = size; | |
2539 | lr->lr_blkoff = 0; | |
2540 | BP_ZERO(&lr->lr_blkptr); | |
2541 | ||
2542 | bcopy(data, lr + 1, size); | |
2543 | ||
2544 | error = ztest_replay_write(zd, lr, B_FALSE); | |
2545 | ||
2546 | ztest_lr_free(lr, sizeof (*lr) + size, NULL); | |
2547 | ||
2548 | return (error); | |
2549 | } | |
2550 | ||
2551 | static int | |
2552 | ztest_truncate(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size) | |
2553 | { | |
2554 | lr_truncate_t *lr; | |
2555 | int error; | |
2556 | ||
2557 | lr = ztest_lr_alloc(sizeof (*lr), NULL); | |
2558 | ||
2559 | lr->lr_foid = object; | |
2560 | lr->lr_offset = offset; | |
2561 | lr->lr_length = size; | |
2562 | ||
2563 | error = ztest_replay_truncate(zd, lr, B_FALSE); | |
2564 | ||
2565 | ztest_lr_free(lr, sizeof (*lr), NULL); | |
2566 | ||
2567 | return (error); | |
2568 | } | |
2569 | ||
2570 | static int | |
2571 | ztest_setattr(ztest_ds_t *zd, uint64_t object) | |
2572 | { | |
2573 | lr_setattr_t *lr; | |
2574 | int error; | |
2575 | ||
2576 | lr = ztest_lr_alloc(sizeof (*lr), NULL); | |
2577 | ||
2578 | lr->lr_foid = object; | |
2579 | lr->lr_size = 0; | |
2580 | lr->lr_mode = 0; | |
2581 | ||
2582 | error = ztest_replay_setattr(zd, lr, B_FALSE); | |
2583 | ||
2584 | ztest_lr_free(lr, sizeof (*lr), NULL); | |
2585 | ||
2586 | return (error); | |
2587 | } | |
2588 | ||
2589 | static void | |
2590 | ztest_prealloc(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size) | |
2591 | { | |
2592 | objset_t *os = zd->zd_os; | |
2593 | dmu_tx_t *tx; | |
2594 | uint64_t txg; | |
5d43cc9a | 2595 | rl_t *rl; |
428870ff BB |
2596 | |
2597 | txg_wait_synced(dmu_objset_pool(os), 0); | |
2598 | ||
2599 | ztest_object_lock(zd, object, RL_READER); | |
2600 | rl = ztest_range_lock(zd, object, offset, size, RL_WRITER); | |
2601 | ||
2602 | tx = dmu_tx_create(os); | |
2603 | ||
2604 | dmu_tx_hold_write(tx, object, offset, size); | |
2605 | ||
2606 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
2607 | ||
2608 | if (txg != 0) { | |
2609 | dmu_prealloc(os, object, offset, size, tx); | |
2610 | dmu_tx_commit(tx); | |
2611 | txg_wait_synced(dmu_objset_pool(os), txg); | |
2612 | } else { | |
2613 | (void) dmu_free_long_range(os, object, offset, size); | |
2614 | } | |
2615 | ||
5d43cc9a | 2616 | ztest_range_unlock(rl); |
428870ff BB |
2617 | ztest_object_unlock(zd, object); |
2618 | } | |
2619 | ||
2620 | static void | |
2621 | ztest_io(ztest_ds_t *zd, uint64_t object, uint64_t offset) | |
2622 | { | |
03c6040b | 2623 | int err; |
428870ff BB |
2624 | ztest_block_tag_t wbt; |
2625 | dmu_object_info_t doi; | |
2626 | enum ztest_io_type io_type; | |
2627 | uint64_t blocksize; | |
2628 | void *data; | |
2629 | ||
a9eaae06 | 2630 | VERIFY0(dmu_object_info(zd->zd_os, object, &doi)); |
428870ff BB |
2631 | blocksize = doi.doi_data_block_size; |
2632 | data = umem_alloc(blocksize, UMEM_NOFAIL); | |
2633 | ||
2634 | /* | |
2635 | * Pick an i/o type at random, biased toward writing block tags. | |
2636 | */ | |
2637 | io_type = ztest_random(ZTEST_IO_TYPES); | |
2638 | if (ztest_random(2) == 0) | |
2639 | io_type = ZTEST_IO_WRITE_TAG; | |
2640 | ||
2d9142c9 | 2641 | (void) pthread_rwlock_rdlock(&zd->zd_zilog_lock); |
3e31d2b0 | 2642 | |
428870ff BB |
2643 | switch (io_type) { |
2644 | ||
2645 | case ZTEST_IO_WRITE_TAG: | |
50c957f7 NB |
2646 | ztest_bt_generate(&wbt, zd->zd_os, object, doi.doi_dnodesize, |
2647 | offset, 0, 0, 0); | |
428870ff BB |
2648 | (void) ztest_write(zd, object, offset, sizeof (wbt), &wbt); |
2649 | break; | |
2650 | ||
2651 | case ZTEST_IO_WRITE_PATTERN: | |
2652 | (void) memset(data, 'a' + (object + offset) % 5, blocksize); | |
2653 | if (ztest_random(2) == 0) { | |
2654 | /* | |
2655 | * Induce fletcher2 collisions to ensure that | |
2656 | * zio_ddt_collision() detects and resolves them | |
2657 | * when using fletcher2-verify for deduplication. | |
2658 | */ | |
2659 | ((uint64_t *)data)[0] ^= 1ULL << 63; | |
2660 | ((uint64_t *)data)[4] ^= 1ULL << 63; | |
2661 | } | |
2662 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2663 | break; | |
2664 | ||
2665 | case ZTEST_IO_WRITE_ZEROES: | |
2666 | bzero(data, blocksize); | |
2667 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2668 | break; | |
2669 | ||
2670 | case ZTEST_IO_TRUNCATE: | |
2671 | (void) ztest_truncate(zd, object, offset, blocksize); | |
2672 | break; | |
2673 | ||
2674 | case ZTEST_IO_SETATTR: | |
2675 | (void) ztest_setattr(zd, object); | |
2676 | break; | |
e75c13c3 BB |
2677 | default: |
2678 | break; | |
03c6040b GW |
2679 | |
2680 | case ZTEST_IO_REWRITE: | |
2d9142c9 | 2681 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
03c6040b GW |
2682 | err = ztest_dsl_prop_set_uint64(zd->zd_name, |
2683 | ZFS_PROP_CHECKSUM, spa_dedup_checksum(ztest_spa), | |
2684 | B_FALSE); | |
2685 | VERIFY(err == 0 || err == ENOSPC); | |
2686 | err = ztest_dsl_prop_set_uint64(zd->zd_name, | |
2687 | ZFS_PROP_COMPRESSION, | |
2688 | ztest_random_dsl_prop(ZFS_PROP_COMPRESSION), | |
2689 | B_FALSE); | |
2690 | VERIFY(err == 0 || err == ENOSPC); | |
2d9142c9 | 2691 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
03c6040b GW |
2692 | |
2693 | VERIFY0(dmu_read(zd->zd_os, object, offset, blocksize, data, | |
2694 | DMU_READ_NO_PREFETCH)); | |
2695 | ||
2696 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2697 | break; | |
428870ff BB |
2698 | } |
2699 | ||
2d9142c9 | 2700 | (void) pthread_rwlock_unlock(&zd->zd_zilog_lock); |
3e31d2b0 | 2701 | |
428870ff BB |
2702 | umem_free(data, blocksize); |
2703 | } | |
2704 | ||
2705 | /* | |
2706 | * Initialize an object description template. | |
2707 | */ | |
2708 | static void | |
2709 | ztest_od_init(ztest_od_t *od, uint64_t id, char *tag, uint64_t index, | |
50c957f7 NB |
2710 | dmu_object_type_t type, uint64_t blocksize, uint64_t dnodesize, |
2711 | uint64_t gen) | |
428870ff BB |
2712 | { |
2713 | od->od_dir = ZTEST_DIROBJ; | |
2714 | od->od_object = 0; | |
2715 | ||
2716 | od->od_crtype = type; | |
2717 | od->od_crblocksize = blocksize ? blocksize : ztest_random_blocksize(); | |
50c957f7 | 2718 | od->od_crdnodesize = dnodesize ? dnodesize : ztest_random_dnodesize(); |
428870ff BB |
2719 | od->od_crgen = gen; |
2720 | ||
2721 | od->od_type = DMU_OT_NONE; | |
2722 | od->od_blocksize = 0; | |
2723 | od->od_gen = 0; | |
2724 | ||
2725 | (void) snprintf(od->od_name, sizeof (od->od_name), "%s(%lld)[%llu]", | |
b8864a23 | 2726 | tag, (longlong_t)id, (u_longlong_t)index); |
428870ff BB |
2727 | } |
2728 | ||
2729 | /* | |
2730 | * Lookup or create the objects for a test using the od template. | |
2731 | * If the objects do not all exist, or if 'remove' is specified, | |
2732 | * remove any existing objects and create new ones. Otherwise, | |
2733 | * use the existing objects. | |
2734 | */ | |
2735 | static int | |
2736 | ztest_object_init(ztest_ds_t *zd, ztest_od_t *od, size_t size, boolean_t remove) | |
2737 | { | |
2738 | int count = size / sizeof (*od); | |
2739 | int rv = 0; | |
2740 | ||
1e33ac1e | 2741 | mutex_enter(&zd->zd_dirobj_lock); |
428870ff BB |
2742 | if ((ztest_lookup(zd, od, count) != 0 || remove) && |
2743 | (ztest_remove(zd, od, count) != 0 || | |
2744 | ztest_create(zd, od, count) != 0)) | |
2745 | rv = -1; | |
2746 | zd->zd_od = od; | |
1e33ac1e | 2747 | mutex_exit(&zd->zd_dirobj_lock); |
428870ff BB |
2748 | |
2749 | return (rv); | |
2750 | } | |
2751 | ||
2752 | /* ARGSUSED */ | |
2753 | void | |
2754 | ztest_zil_commit(ztest_ds_t *zd, uint64_t id) | |
2755 | { | |
2756 | zilog_t *zilog = zd->zd_zilog; | |
2757 | ||
2d9142c9 | 2758 | (void) pthread_rwlock_rdlock(&zd->zd_zilog_lock); |
3e31d2b0 | 2759 | |
572e2857 | 2760 | zil_commit(zilog, ztest_random(ZTEST_OBJECTS)); |
428870ff BB |
2761 | |
2762 | /* | |
2763 | * Remember the committed values in zd, which is in parent/child | |
2764 | * shared memory. If we die, the next iteration of ztest_run() | |
2765 | * will verify that the log really does contain this record. | |
2766 | */ | |
2767 | mutex_enter(&zilog->zl_lock); | |
a9eaae06 | 2768 | ASSERT3P(zd->zd_shared, !=, NULL); |
c242c188 CS |
2769 | ASSERT3U(zd->zd_shared->zd_seq, <=, zilog->zl_commit_lr_seq); |
2770 | zd->zd_shared->zd_seq = zilog->zl_commit_lr_seq; | |
428870ff | 2771 | mutex_exit(&zilog->zl_lock); |
3e31d2b0 | 2772 | |
2d9142c9 | 2773 | (void) pthread_rwlock_unlock(&zd->zd_zilog_lock); |
3e31d2b0 ES |
2774 | } |
2775 | ||
2776 | /* | |
2777 | * This function is designed to simulate the operations that occur during a | |
2778 | * mount/unmount operation. We hold the dataset across these operations in an | |
2779 | * attempt to expose any implicit assumptions about ZIL management. | |
2780 | */ | |
2781 | /* ARGSUSED */ | |
2782 | void | |
2783 | ztest_zil_remount(ztest_ds_t *zd, uint64_t id) | |
2784 | { | |
2785 | objset_t *os = zd->zd_os; | |
2786 | ||
0b606cb3 TC |
2787 | /* |
2788 | * We hold the ztest_vdev_lock so we don't cause problems with | |
2789 | * other threads that wish to remove a log device, such as | |
2790 | * ztest_device_removal(). | |
2791 | */ | |
2792 | mutex_enter(&ztest_vdev_lock); | |
2793 | ||
03c6040b GW |
2794 | /* |
2795 | * We grab the zd_dirobj_lock to ensure that no other thread is | |
2796 | * updating the zil (i.e. adding in-memory log records) and the | |
2797 | * zd_zilog_lock to block any I/O. | |
2798 | */ | |
29809a6c | 2799 | mutex_enter(&zd->zd_dirobj_lock); |
2d9142c9 | 2800 | (void) pthread_rwlock_wrlock(&zd->zd_zilog_lock); |
3e31d2b0 | 2801 | |
f298b24d | 2802 | /* zfsvfs_teardown() */ |
3e31d2b0 ES |
2803 | zil_close(zd->zd_zilog); |
2804 | ||
2805 | /* zfsvfs_setup() */ | |
a9eaae06 | 2806 | VERIFY3P(zil_open(os, ztest_get_data), ==, zd->zd_zilog); |
3e31d2b0 ES |
2807 | zil_replay(os, zd, ztest_replay_vector); |
2808 | ||
2d9142c9 | 2809 | (void) pthread_rwlock_unlock(&zd->zd_zilog_lock); |
29809a6c | 2810 | mutex_exit(&zd->zd_dirobj_lock); |
0b606cb3 | 2811 | mutex_exit(&ztest_vdev_lock); |
428870ff BB |
2812 | } |
2813 | ||
2814 | /* | |
2815 | * Verify that we can't destroy an active pool, create an existing pool, | |
2816 | * or create a pool with a bad vdev spec. | |
2817 | */ | |
2818 | /* ARGSUSED */ | |
2819 | void | |
2820 | ztest_spa_create_destroy(ztest_ds_t *zd, uint64_t id) | |
2821 | { | |
c242c188 | 2822 | ztest_shared_opts_t *zo = &ztest_opts; |
428870ff BB |
2823 | spa_t *spa; |
2824 | nvlist_t *nvroot; | |
2825 | ||
379ca9cf OF |
2826 | if (zo->zo_mmp_test) |
2827 | return; | |
2828 | ||
428870ff BB |
2829 | /* |
2830 | * Attempt to create using a bad file. | |
2831 | */ | |
cc99f275 | 2832 | nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, NULL, 0, 0, 1); |
428870ff | 2833 | VERIFY3U(ENOENT, ==, |
b5256303 | 2834 | spa_create("ztest_bad_file", nvroot, NULL, NULL, NULL)); |
bea3fc71 | 2835 | fnvlist_free(nvroot); |
428870ff BB |
2836 | |
2837 | /* | |
2838 | * Attempt to create using a bad mirror. | |
2839 | */ | |
cc99f275 | 2840 | nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, NULL, 0, 2, 1); |
428870ff | 2841 | VERIFY3U(ENOENT, ==, |
b5256303 | 2842 | spa_create("ztest_bad_mirror", nvroot, NULL, NULL, NULL)); |
bea3fc71 | 2843 | fnvlist_free(nvroot); |
428870ff BB |
2844 | |
2845 | /* | |
2846 | * Attempt to create an existing pool. It shouldn't matter | |
2847 | * what's in the nvroot; we should fail with EEXIST. | |
2848 | */ | |
2d9142c9 | 2849 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
cc99f275 | 2850 | nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, NULL, 0, 0, 1); |
b5256303 TC |
2851 | VERIFY3U(EEXIST, ==, |
2852 | spa_create(zo->zo_pool, nvroot, NULL, NULL, NULL)); | |
bea3fc71 | 2853 | fnvlist_free(nvroot); |
43a85362 SD |
2854 | |
2855 | /* | |
2856 | * We open a reference to the spa and then we try to export it | |
2857 | * expecting one of the following errors: | |
2858 | * | |
2859 | * EBUSY | |
2860 | * Because of the reference we just opened. | |
2861 | * | |
2862 | * ZFS_ERR_EXPORT_IN_PROGRESS | |
2863 | * For the case that there is another ztest thread doing | |
2864 | * an export concurrently. | |
2865 | */ | |
a9eaae06 | 2866 | VERIFY0(spa_open(zo->zo_pool, &spa, FTAG)); |
43a85362 SD |
2867 | int error = spa_destroy(zo->zo_pool); |
2868 | if (error != EBUSY && error != ZFS_ERR_EXPORT_IN_PROGRESS) { | |
2869 | fatal(0, "spa_destroy(%s) returned unexpected value %d", | |
2870 | spa->spa_name, error); | |
2871 | } | |
428870ff BB |
2872 | spa_close(spa, FTAG); |
2873 | ||
2d9142c9 | 2874 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
428870ff BB |
2875 | } |
2876 | ||
0582e403 OF |
2877 | /* |
2878 | * Start and then stop the MMP threads to ensure the startup and shutdown code | |
2879 | * works properly. Actual protection and property-related code tested via ZTS. | |
2880 | */ | |
2881 | /* ARGSUSED */ | |
2882 | void | |
2883 | ztest_mmp_enable_disable(ztest_ds_t *zd, uint64_t id) | |
2884 | { | |
2885 | ztest_shared_opts_t *zo = &ztest_opts; | |
2886 | spa_t *spa = ztest_spa; | |
2887 | ||
2888 | if (zo->zo_mmp_test) | |
2889 | return; | |
2890 | ||
8133679f OF |
2891 | /* |
2892 | * Since enabling MMP involves setting a property, it could not be done | |
2893 | * while the pool is suspended. | |
2894 | */ | |
2895 | if (spa_suspended(spa)) | |
2896 | return; | |
2897 | ||
0582e403 OF |
2898 | spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); |
2899 | mutex_enter(&spa->spa_props_lock); | |
2900 | ||
52002377 OF |
2901 | zfs_multihost_fail_intervals = 0; |
2902 | ||
0582e403 OF |
2903 | if (!spa_multihost(spa)) { |
2904 | spa->spa_multihost = B_TRUE; | |
2905 | mmp_thread_start(spa); | |
2906 | } | |
2907 | ||
2908 | mutex_exit(&spa->spa_props_lock); | |
2909 | spa_config_exit(spa, SCL_CONFIG, FTAG); | |
2910 | ||
2911 | txg_wait_synced(spa_get_dsl(spa), 0); | |
2912 | mmp_signal_all_threads(); | |
2913 | txg_wait_synced(spa_get_dsl(spa), 0); | |
2914 | ||
2915 | spa_config_enter(spa, SCL_CONFIG, FTAG, RW_READER); | |
2916 | mutex_enter(&spa->spa_props_lock); | |
2917 | ||
2918 | if (spa_multihost(spa)) { | |
2919 | mmp_thread_stop(spa); | |
2920 | spa->spa_multihost = B_FALSE; | |
2921 | } | |
2922 | ||
2923 | mutex_exit(&spa->spa_props_lock); | |
2924 | spa_config_exit(spa, SCL_CONFIG, FTAG); | |
2925 | } | |
2926 | ||
ea0b2538 GW |
2927 | /* ARGSUSED */ |
2928 | void | |
2929 | ztest_spa_upgrade(ztest_ds_t *zd, uint64_t id) | |
2930 | { | |
2931 | spa_t *spa; | |
2932 | uint64_t initial_version = SPA_VERSION_INITIAL; | |
2933 | uint64_t version, newversion; | |
2934 | nvlist_t *nvroot, *props; | |
2935 | char *name; | |
2936 | ||
379ca9cf OF |
2937 | if (ztest_opts.zo_mmp_test) |
2938 | return; | |
2939 | ||
b2255edc BB |
2940 | /* dRAID added after feature flags, skip upgrade test. */ |
2941 | if (strcmp(ztest_opts.zo_raid_type, VDEV_TYPE_DRAID) == 0) | |
2942 | return; | |
2943 | ||
ea0b2538 GW |
2944 | mutex_enter(&ztest_vdev_lock); |
2945 | name = kmem_asprintf("%s_upgrade", ztest_opts.zo_pool); | |
2946 | ||
2947 | /* | |
2948 | * Clean up from previous runs. | |
2949 | */ | |
2950 | (void) spa_destroy(name); | |
2951 | ||
2952 | nvroot = make_vdev_root(NULL, NULL, name, ztest_opts.zo_vdev_size, 0, | |
b2255edc | 2953 | NULL, ztest_opts.zo_raid_children, ztest_opts.zo_mirrors, 1); |
ea0b2538 GW |
2954 | |
2955 | /* | |
2956 | * If we're configuring a RAIDZ device then make sure that the | |
d611989f | 2957 | * initial version is capable of supporting that feature. |
ea0b2538 | 2958 | */ |
b2255edc | 2959 | switch (ztest_opts.zo_raid_parity) { |
ea0b2538 GW |
2960 | case 0: |
2961 | case 1: | |
2962 | initial_version = SPA_VERSION_INITIAL; | |
2963 | break; | |
2964 | case 2: | |
2965 | initial_version = SPA_VERSION_RAIDZ2; | |
2966 | break; | |
2967 | case 3: | |
2968 | initial_version = SPA_VERSION_RAIDZ3; | |
2969 | break; | |
2970 | } | |
2971 | ||
2972 | /* | |
2973 | * Create a pool with a spa version that can be upgraded. Pick | |
2974 | * a value between initial_version and SPA_VERSION_BEFORE_FEATURES. | |
2975 | */ | |
2976 | do { | |
2977 | version = ztest_random_spa_version(initial_version); | |
2978 | } while (version > SPA_VERSION_BEFORE_FEATURES); | |
2979 | ||
2980 | props = fnvlist_alloc(); | |
2981 | fnvlist_add_uint64(props, | |
2982 | zpool_prop_to_name(ZPOOL_PROP_VERSION), version); | |
a9eaae06 | 2983 | VERIFY0(spa_create(name, nvroot, props, NULL, NULL)); |
ea0b2538 GW |
2984 | fnvlist_free(nvroot); |
2985 | fnvlist_free(props); | |
2986 | ||
a9eaae06 | 2987 | VERIFY0(spa_open(name, &spa, FTAG)); |
ea0b2538 GW |
2988 | VERIFY3U(spa_version(spa), ==, version); |
2989 | newversion = ztest_random_spa_version(version + 1); | |
2990 | ||
2991 | if (ztest_opts.zo_verbose >= 4) { | |
2992 | (void) printf("upgrading spa version from %llu to %llu\n", | |
2993 | (u_longlong_t)version, (u_longlong_t)newversion); | |
2994 | } | |
2995 | ||
2996 | spa_upgrade(spa, newversion); | |
2997 | VERIFY3U(spa_version(spa), >, version); | |
2998 | VERIFY3U(spa_version(spa), ==, fnvlist_lookup_uint64(spa->spa_config, | |
2999 | zpool_prop_to_name(ZPOOL_PROP_VERSION))); | |
3000 | spa_close(spa, FTAG); | |
3001 | ||
e4f5fa12 | 3002 | kmem_strfree(name); |
ea0b2538 GW |
3003 | mutex_exit(&ztest_vdev_lock); |
3004 | } | |
3005 | ||
d2734cce SD |
3006 | static void |
3007 | ztest_spa_checkpoint(spa_t *spa) | |
3008 | { | |
3009 | ASSERT(MUTEX_HELD(&ztest_checkpoint_lock)); | |
3010 | ||
3011 | int error = spa_checkpoint(spa->spa_name); | |
3012 | ||
3013 | switch (error) { | |
3014 | case 0: | |
3015 | case ZFS_ERR_DEVRM_IN_PROGRESS: | |
3016 | case ZFS_ERR_DISCARDING_CHECKPOINT: | |
3017 | case ZFS_ERR_CHECKPOINT_EXISTS: | |
3018 | break; | |
3019 | case ENOSPC: | |
3020 | ztest_record_enospc(FTAG); | |
3021 | break; | |
3022 | default: | |
3023 | fatal(0, "spa_checkpoint(%s) = %d", spa->spa_name, error); | |
3024 | } | |
3025 | } | |
3026 | ||
3027 | static void | |
3028 | ztest_spa_discard_checkpoint(spa_t *spa) | |
3029 | { | |
3030 | ASSERT(MUTEX_HELD(&ztest_checkpoint_lock)); | |
3031 | ||
3032 | int error = spa_checkpoint_discard(spa->spa_name); | |
3033 | ||
3034 | switch (error) { | |
3035 | case 0: | |
3036 | case ZFS_ERR_DISCARDING_CHECKPOINT: | |
3037 | case ZFS_ERR_NO_CHECKPOINT: | |
3038 | break; | |
3039 | default: | |
3040 | fatal(0, "spa_discard_checkpoint(%s) = %d", | |
3041 | spa->spa_name, error); | |
3042 | } | |
3043 | ||
3044 | } | |
3045 | ||
3046 | /* ARGSUSED */ | |
3047 | void | |
3048 | ztest_spa_checkpoint_create_discard(ztest_ds_t *zd, uint64_t id) | |
3049 | { | |
3050 | spa_t *spa = ztest_spa; | |
3051 | ||
3052 | mutex_enter(&ztest_checkpoint_lock); | |
3053 | if (ztest_random(2) == 0) { | |
3054 | ztest_spa_checkpoint(spa); | |
3055 | } else { | |
3056 | ztest_spa_discard_checkpoint(spa); | |
3057 | } | |
3058 | mutex_exit(&ztest_checkpoint_lock); | |
3059 | } | |
3060 | ||
3061 | ||
428870ff BB |
3062 | static vdev_t * |
3063 | vdev_lookup_by_path(vdev_t *vd, const char *path) | |
3064 | { | |
3065 | vdev_t *mvd; | |
d6320ddb | 3066 | int c; |
428870ff BB |
3067 | |
3068 | if (vd->vdev_path != NULL && strcmp(path, vd->vdev_path) == 0) | |
3069 | return (vd); | |
3070 | ||
d6320ddb | 3071 | for (c = 0; c < vd->vdev_children; c++) |
428870ff BB |
3072 | if ((mvd = vdev_lookup_by_path(vd->vdev_child[c], path)) != |
3073 | NULL) | |
3074 | return (mvd); | |
3075 | ||
3076 | return (NULL); | |
3077 | } | |
3078 | ||
93e28d66 SD |
3079 | static int |
3080 | spa_num_top_vdevs(spa_t *spa) | |
428870ff BB |
3081 | { |
3082 | vdev_t *rvd = spa->spa_root_vdev; | |
93e28d66 SD |
3083 | ASSERT3U(spa_config_held(spa, SCL_VDEV, RW_READER), ==, SCL_VDEV); |
3084 | return (rvd->vdev_children); | |
428870ff BB |
3085 | } |
3086 | ||
3087 | /* | |
3088 | * Verify that vdev_add() works as expected. | |
3089 | */ | |
3090 | /* ARGSUSED */ | |
3091 | void | |
3092 | ztest_vdev_add_remove(ztest_ds_t *zd, uint64_t id) | |
3093 | { | |
3094 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 3095 | spa_t *spa = ztest_spa; |
428870ff BB |
3096 | uint64_t leaves; |
3097 | uint64_t guid; | |
3098 | nvlist_t *nvroot; | |
3099 | int error; | |
3100 | ||
379ca9cf OF |
3101 | if (ztest_opts.zo_mmp_test) |
3102 | return; | |
3103 | ||
c242c188 | 3104 | mutex_enter(&ztest_vdev_lock); |
b2255edc BB |
3105 | leaves = MAX(zs->zs_mirrors + zs->zs_splits, 1) * |
3106 | ztest_opts.zo_raid_children; | |
428870ff BB |
3107 | |
3108 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
3109 | ||
93e28d66 | 3110 | ztest_shared->zs_vdev_next_leaf = spa_num_top_vdevs(spa) * leaves; |
428870ff BB |
3111 | |
3112 | /* | |
3113 | * If we have slogs then remove them 1/4 of the time. | |
3114 | */ | |
3115 | if (spa_has_slogs(spa) && ztest_random(4) == 0) { | |
cc99f275 DB |
3116 | metaslab_group_t *mg; |
3117 | ||
428870ff | 3118 | /* |
cc99f275 | 3119 | * find the first real slog in log allocation class |
428870ff | 3120 | */ |
f8020c93 | 3121 | mg = spa_log_class(spa)->mc_allocator[0].mca_rotor; |
cc99f275 DB |
3122 | while (!mg->mg_vd->vdev_islog) |
3123 | mg = mg->mg_next; | |
3124 | ||
3125 | guid = mg->mg_vd->vdev_guid; | |
428870ff BB |
3126 | |
3127 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
3128 | ||
3129 | /* | |
3130 | * We have to grab the zs_name_lock as writer to | |
3131 | * prevent a race between removing a slog (dmu_objset_find) | |
3132 | * and destroying a dataset. Removing the slog will | |
3133 | * grab a reference on the dataset which may cause | |
13fe0198 | 3134 | * dsl_destroy_head() to fail with EBUSY thus |
428870ff BB |
3135 | * leaving the dataset in an inconsistent state. |
3136 | */ | |
2d9142c9 | 3137 | pthread_rwlock_wrlock(&ztest_name_lock); |
428870ff | 3138 | error = spa_vdev_remove(spa, guid, B_FALSE); |
2d9142c9 | 3139 | pthread_rwlock_unlock(&ztest_name_lock); |
428870ff | 3140 | |
c91cf36f BB |
3141 | switch (error) { |
3142 | case 0: | |
3143 | case EEXIST: /* Generic zil_reset() error */ | |
3144 | case EBUSY: /* Replay required */ | |
3145 | case EACCES: /* Crypto key not loaded */ | |
d2734cce SD |
3146 | case ZFS_ERR_CHECKPOINT_EXISTS: |
3147 | case ZFS_ERR_DISCARDING_CHECKPOINT: | |
c91cf36f BB |
3148 | break; |
3149 | default: | |
428870ff | 3150 | fatal(0, "spa_vdev_remove() = %d", error); |
c91cf36f | 3151 | } |
428870ff BB |
3152 | } else { |
3153 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
3154 | ||
3155 | /* | |
cc99f275 | 3156 | * Make 1/4 of the devices be log devices |
428870ff | 3157 | */ |
ea0b2538 | 3158 | nvroot = make_vdev_root(NULL, NULL, NULL, |
cc99f275 | 3159 | ztest_opts.zo_vdev_size, 0, (ztest_random(4) == 0) ? |
b2255edc BB |
3160 | "log" : NULL, ztest_opts.zo_raid_children, zs->zs_mirrors, |
3161 | 1); | |
428870ff BB |
3162 | |
3163 | error = spa_vdev_add(spa, nvroot); | |
bea3fc71 | 3164 | fnvlist_free(nvroot); |
428870ff | 3165 | |
d2734cce SD |
3166 | switch (error) { |
3167 | case 0: | |
3168 | break; | |
3169 | case ENOSPC: | |
428870ff | 3170 | ztest_record_enospc("spa_vdev_add"); |
d2734cce SD |
3171 | break; |
3172 | default: | |
428870ff | 3173 | fatal(0, "spa_vdev_add() = %d", error); |
d2734cce | 3174 | } |
428870ff BB |
3175 | } |
3176 | ||
c242c188 | 3177 | mutex_exit(&ztest_vdev_lock); |
428870ff BB |
3178 | } |
3179 | ||
cc99f275 DB |
3180 | /* ARGSUSED */ |
3181 | void | |
3182 | ztest_vdev_class_add(ztest_ds_t *zd, uint64_t id) | |
3183 | { | |
3184 | ztest_shared_t *zs = ztest_shared; | |
3185 | spa_t *spa = ztest_spa; | |
3186 | uint64_t leaves; | |
3187 | nvlist_t *nvroot; | |
3188 | const char *class = (ztest_random(2) == 0) ? | |
3189 | VDEV_ALLOC_BIAS_SPECIAL : VDEV_ALLOC_BIAS_DEDUP; | |
3190 | int error; | |
3191 | ||
3192 | /* | |
3193 | * By default add a special vdev 50% of the time | |
3194 | */ | |
3195 | if ((ztest_opts.zo_special_vdevs == ZTEST_VDEV_CLASS_OFF) || | |
3196 | (ztest_opts.zo_special_vdevs == ZTEST_VDEV_CLASS_RND && | |
3197 | ztest_random(2) == 0)) { | |
3198 | return; | |
3199 | } | |
3200 | ||
3201 | mutex_enter(&ztest_vdev_lock); | |
3202 | ||
3203 | /* Only test with mirrors */ | |
3204 | if (zs->zs_mirrors < 2) { | |
3205 | mutex_exit(&ztest_vdev_lock); | |
3206 | return; | |
3207 | } | |
3208 | ||
3209 | /* requires feature@allocation_classes */ | |
3210 | if (!spa_feature_is_enabled(spa, SPA_FEATURE_ALLOCATION_CLASSES)) { | |
3211 | mutex_exit(&ztest_vdev_lock); | |
3212 | return; | |
3213 | } | |
3214 | ||
b2255edc BB |
3215 | leaves = MAX(zs->zs_mirrors + zs->zs_splits, 1) * |
3216 | ztest_opts.zo_raid_children; | |
cc99f275 DB |
3217 | |
3218 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
93e28d66 | 3219 | ztest_shared->zs_vdev_next_leaf = spa_num_top_vdevs(spa) * leaves; |
cc99f275 DB |
3220 | spa_config_exit(spa, SCL_VDEV, FTAG); |
3221 | ||
3222 | nvroot = make_vdev_root(NULL, NULL, NULL, ztest_opts.zo_vdev_size, 0, | |
b2255edc | 3223 | class, ztest_opts.zo_raid_children, zs->zs_mirrors, 1); |
cc99f275 DB |
3224 | |
3225 | error = spa_vdev_add(spa, nvroot); | |
bea3fc71 | 3226 | fnvlist_free(nvroot); |
cc99f275 DB |
3227 | |
3228 | if (error == ENOSPC) | |
3229 | ztest_record_enospc("spa_vdev_add"); | |
3230 | else if (error != 0) | |
3231 | fatal(0, "spa_vdev_add() = %d", error); | |
3232 | ||
3233 | /* | |
3234 | * 50% of the time allow small blocks in the special class | |
3235 | */ | |
3236 | if (error == 0 && | |
3237 | spa_special_class(spa)->mc_groups == 1 && ztest_random(2) == 0) { | |
3238 | if (ztest_opts.zo_verbose >= 3) | |
3239 | (void) printf("Enabling special VDEV small blocks\n"); | |
3240 | (void) ztest_dsl_prop_set_uint64(zd->zd_name, | |
3241 | ZFS_PROP_SPECIAL_SMALL_BLOCKS, 32768, B_FALSE); | |
3242 | } | |
3243 | ||
3244 | mutex_exit(&ztest_vdev_lock); | |
3245 | ||
3246 | if (ztest_opts.zo_verbose >= 3) { | |
3247 | metaslab_class_t *mc; | |
3248 | ||
3249 | if (strcmp(class, VDEV_ALLOC_BIAS_SPECIAL) == 0) | |
3250 | mc = spa_special_class(spa); | |
3251 | else | |
3252 | mc = spa_dedup_class(spa); | |
3253 | (void) printf("Added a %s mirrored vdev (of %d)\n", | |
3254 | class, (int)mc->mc_groups); | |
3255 | } | |
3256 | } | |
3257 | ||
428870ff BB |
3258 | /* |
3259 | * Verify that adding/removing aux devices (l2arc, hot spare) works as expected. | |
3260 | */ | |
3261 | /* ARGSUSED */ | |
3262 | void | |
3263 | ztest_vdev_aux_add_remove(ztest_ds_t *zd, uint64_t id) | |
3264 | { | |
3265 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 3266 | spa_t *spa = ztest_spa; |
428870ff BB |
3267 | vdev_t *rvd = spa->spa_root_vdev; |
3268 | spa_aux_vdev_t *sav; | |
3269 | char *aux; | |
40b84e7a | 3270 | char *path; |
428870ff | 3271 | uint64_t guid = 0; |
b2255edc | 3272 | int error, ignore_err = 0; |
428870ff | 3273 | |
379ca9cf OF |
3274 | if (ztest_opts.zo_mmp_test) |
3275 | return; | |
3276 | ||
40b84e7a BB |
3277 | path = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
3278 | ||
428870ff BB |
3279 | if (ztest_random(2) == 0) { |
3280 | sav = &spa->spa_spares; | |
3281 | aux = ZPOOL_CONFIG_SPARES; | |
3282 | } else { | |
3283 | sav = &spa->spa_l2cache; | |
3284 | aux = ZPOOL_CONFIG_L2CACHE; | |
3285 | } | |
3286 | ||
c242c188 | 3287 | mutex_enter(&ztest_vdev_lock); |
428870ff BB |
3288 | |
3289 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
3290 | ||
3291 | if (sav->sav_count != 0 && ztest_random(4) == 0) { | |
b128c09f BB |
3292 | /* |
3293 | * Pick a random device to remove. | |
3294 | */ | |
b2255edc BB |
3295 | vdev_t *svd = sav->sav_vdevs[ztest_random(sav->sav_count)]; |
3296 | ||
3297 | /* dRAID spares cannot be removed; try anyways to see ENOTSUP */ | |
3298 | if (strstr(svd->vdev_path, VDEV_TYPE_DRAID) != NULL) | |
3299 | ignore_err = ENOTSUP; | |
3300 | ||
3301 | guid = svd->vdev_guid; | |
b128c09f BB |
3302 | } else { |
3303 | /* | |
3304 | * Find an unused device we can add. | |
3305 | */ | |
428870ff | 3306 | zs->zs_vdev_aux = 0; |
b128c09f | 3307 | for (;;) { |
b128c09f | 3308 | int c; |
3db3ff4a | 3309 | (void) snprintf(path, MAXPATHLEN, ztest_aux_template, |
c242c188 CS |
3310 | ztest_opts.zo_dir, ztest_opts.zo_pool, aux, |
3311 | zs->zs_vdev_aux); | |
b128c09f BB |
3312 | for (c = 0; c < sav->sav_count; c++) |
3313 | if (strcmp(sav->sav_vdevs[c]->vdev_path, | |
3314 | path) == 0) | |
3315 | break; | |
3316 | if (c == sav->sav_count && | |
3317 | vdev_lookup_by_path(rvd, path) == NULL) | |
3318 | break; | |
428870ff | 3319 | zs->zs_vdev_aux++; |
34dc7c2f BB |
3320 | } |
3321 | } | |
3322 | ||
b128c09f | 3323 | spa_config_exit(spa, SCL_VDEV, FTAG); |
34dc7c2f | 3324 | |
b128c09f BB |
3325 | if (guid == 0) { |
3326 | /* | |
3327 | * Add a new device. | |
3328 | */ | |
ea0b2538 | 3329 | nvlist_t *nvroot = make_vdev_root(NULL, aux, NULL, |
cc99f275 | 3330 | (ztest_opts.zo_vdev_size * 5) / 4, 0, NULL, 0, 0, 1); |
b128c09f | 3331 | error = spa_vdev_add(spa, nvroot); |
d2734cce SD |
3332 | |
3333 | switch (error) { | |
3334 | case 0: | |
3335 | break; | |
3336 | default: | |
b128c09f | 3337 | fatal(0, "spa_vdev_add(%p) = %d", nvroot, error); |
d2734cce | 3338 | } |
bea3fc71 | 3339 | fnvlist_free(nvroot); |
b128c09f BB |
3340 | } else { |
3341 | /* | |
3342 | * Remove an existing device. Sometimes, dirty its | |
3343 | * vdev state first to make sure we handle removal | |
3344 | * of devices that have pending state changes. | |
3345 | */ | |
3346 | if (ztest_random(2) == 0) | |
9babb374 | 3347 | (void) vdev_online(spa, guid, 0, NULL); |
b128c09f BB |
3348 | |
3349 | error = spa_vdev_remove(spa, guid, B_FALSE); | |
d2734cce SD |
3350 | |
3351 | switch (error) { | |
3352 | case 0: | |
3353 | case EBUSY: | |
3354 | case ZFS_ERR_CHECKPOINT_EXISTS: | |
3355 | case ZFS_ERR_DISCARDING_CHECKPOINT: | |
3356 | break; | |
3357 | default: | |
b2255edc BB |
3358 | if (error != ignore_err) |
3359 | fatal(0, "spa_vdev_remove(%llu) = %d", guid, | |
3360 | error); | |
d2734cce | 3361 | } |
b128c09f BB |
3362 | } |
3363 | ||
c242c188 | 3364 | mutex_exit(&ztest_vdev_lock); |
40b84e7a BB |
3365 | |
3366 | umem_free(path, MAXPATHLEN); | |
428870ff BB |
3367 | } |
3368 | ||
3369 | /* | |
3370 | * split a pool if it has mirror tlvdevs | |
3371 | */ | |
3372 | /* ARGSUSED */ | |
3373 | void | |
3374 | ztest_split_pool(ztest_ds_t *zd, uint64_t id) | |
3375 | { | |
3376 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 3377 | spa_t *spa = ztest_spa; |
428870ff BB |
3378 | vdev_t *rvd = spa->spa_root_vdev; |
3379 | nvlist_t *tree, **child, *config, *split, **schild; | |
3380 | uint_t c, children, schildren = 0, lastlogid = 0; | |
3381 | int error = 0; | |
3382 | ||
379ca9cf OF |
3383 | if (ztest_opts.zo_mmp_test) |
3384 | return; | |
3385 | ||
c242c188 | 3386 | mutex_enter(&ztest_vdev_lock); |
428870ff | 3387 | |
d611989f | 3388 | /* ensure we have a usable config; mirrors of raidz aren't supported */ |
b2255edc | 3389 | if (zs->zs_mirrors < 3 || ztest_opts.zo_raid_children > 1) { |
c242c188 | 3390 | mutex_exit(&ztest_vdev_lock); |
428870ff BB |
3391 | return; |
3392 | } | |
3393 | ||
3394 | /* clean up the old pool, if any */ | |
3395 | (void) spa_destroy("splitp"); | |
3396 | ||
3397 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
3398 | ||
3399 | /* generate a config from the existing config */ | |
3400 | mutex_enter(&spa->spa_props_lock); | |
bea3fc71 | 3401 | tree = fnvlist_lookup_nvlist(spa->spa_config, ZPOOL_CONFIG_VDEV_TREE); |
428870ff BB |
3402 | mutex_exit(&spa->spa_props_lock); |
3403 | ||
bea3fc71 RM |
3404 | VERIFY0(nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, |
3405 | &child, &children)); | |
428870ff BB |
3406 | |
3407 | schild = malloc(rvd->vdev_children * sizeof (nvlist_t *)); | |
3408 | for (c = 0; c < children; c++) { | |
3409 | vdev_t *tvd = rvd->vdev_child[c]; | |
3410 | nvlist_t **mchild; | |
3411 | uint_t mchildren; | |
3412 | ||
3413 | if (tvd->vdev_islog || tvd->vdev_ops == &vdev_hole_ops) { | |
bea3fc71 RM |
3414 | schild[schildren] = fnvlist_alloc(); |
3415 | fnvlist_add_string(schild[schildren], | |
3416 | ZPOOL_CONFIG_TYPE, VDEV_TYPE_HOLE); | |
3417 | fnvlist_add_uint64(schild[schildren], | |
3418 | ZPOOL_CONFIG_IS_HOLE, 1); | |
428870ff BB |
3419 | if (lastlogid == 0) |
3420 | lastlogid = schildren; | |
3421 | ++schildren; | |
3422 | continue; | |
3423 | } | |
3424 | lastlogid = 0; | |
bea3fc71 RM |
3425 | VERIFY0(nvlist_lookup_nvlist_array(child[c], |
3426 | ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren)); | |
3427 | schild[schildren++] = fnvlist_dup(mchild[0]); | |
428870ff BB |
3428 | } |
3429 | ||
3430 | /* OK, create a config that can be used to split */ | |
bea3fc71 RM |
3431 | split = fnvlist_alloc(); |
3432 | fnvlist_add_string(split, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT); | |
3433 | fnvlist_add_nvlist_array(split, ZPOOL_CONFIG_CHILDREN, schild, | |
3434 | lastlogid != 0 ? lastlogid : schildren); | |
428870ff | 3435 | |
bea3fc71 RM |
3436 | config = fnvlist_alloc(); |
3437 | fnvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, split); | |
428870ff BB |
3438 | |
3439 | for (c = 0; c < schildren; c++) | |
bea3fc71 | 3440 | fnvlist_free(schild[c]); |
428870ff | 3441 | free(schild); |
bea3fc71 | 3442 | fnvlist_free(split); |
428870ff BB |
3443 | |
3444 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
3445 | ||
2d9142c9 | 3446 | (void) pthread_rwlock_wrlock(&ztest_name_lock); |
428870ff | 3447 | error = spa_vdev_split_mirror(spa, "splitp", config, NULL, B_FALSE); |
2d9142c9 | 3448 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
428870ff | 3449 | |
bea3fc71 | 3450 | fnvlist_free(config); |
428870ff BB |
3451 | |
3452 | if (error == 0) { | |
3453 | (void) printf("successful split - results:\n"); | |
3454 | mutex_enter(&spa_namespace_lock); | |
3455 | show_pool_stats(spa); | |
3456 | show_pool_stats(spa_lookup("splitp")); | |
3457 | mutex_exit(&spa_namespace_lock); | |
3458 | ++zs->zs_splits; | |
3459 | --zs->zs_mirrors; | |
3460 | } | |
c242c188 | 3461 | mutex_exit(&ztest_vdev_lock); |
34dc7c2f BB |
3462 | } |
3463 | ||
3464 | /* | |
3465 | * Verify that we can attach and detach devices. | |
3466 | */ | |
428870ff | 3467 | /* ARGSUSED */ |
34dc7c2f | 3468 | void |
428870ff | 3469 | ztest_vdev_attach_detach(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3470 | { |
428870ff | 3471 | ztest_shared_t *zs = ztest_shared; |
c242c188 | 3472 | spa_t *spa = ztest_spa; |
b128c09f | 3473 | spa_aux_vdev_t *sav = &spa->spa_spares; |
34dc7c2f BB |
3474 | vdev_t *rvd = spa->spa_root_vdev; |
3475 | vdev_t *oldvd, *newvd, *pvd; | |
b128c09f | 3476 | nvlist_t *root; |
428870ff | 3477 | uint64_t leaves; |
34dc7c2f BB |
3478 | uint64_t leaf, top; |
3479 | uint64_t ashift = ztest_get_ashift(); | |
fb5f0bc8 | 3480 | uint64_t oldguid, pguid; |
5d1f7fb6 | 3481 | uint64_t oldsize, newsize; |
40b84e7a | 3482 | char *oldpath, *newpath; |
34dc7c2f | 3483 | int replacing; |
b128c09f BB |
3484 | int oldvd_has_siblings = B_FALSE; |
3485 | int newvd_is_spare = B_FALSE; | |
b2255edc | 3486 | int newvd_is_dspare = B_FALSE; |
b128c09f | 3487 | int oldvd_is_log; |
34dc7c2f | 3488 | int error, expected_error; |
34dc7c2f | 3489 | |
379ca9cf OF |
3490 | if (ztest_opts.zo_mmp_test) |
3491 | return; | |
3492 | ||
40b84e7a BB |
3493 | oldpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
3494 | newpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
3495 | ||
c242c188 | 3496 | mutex_enter(&ztest_vdev_lock); |
b2255edc | 3497 | leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raid_children; |
34dc7c2f | 3498 | |
a1d477c2 MA |
3499 | spa_config_enter(spa, SCL_ALL, FTAG, RW_WRITER); |
3500 | ||
3501 | /* | |
3502 | * If a vdev is in the process of being removed, its removal may | |
3503 | * finish while we are in progress, leading to an unexpected error | |
3504 | * value. Don't bother trying to attach while we are in the middle | |
3505 | * of removal. | |
3506 | */ | |
9e052db4 | 3507 | if (ztest_device_removal_active) { |
a1d477c2 | 3508 | spa_config_exit(spa, SCL_ALL, FTAG); |
a0316ad2 | 3509 | goto out; |
a1d477c2 | 3510 | } |
34dc7c2f BB |
3511 | |
3512 | /* | |
3513 | * Decide whether to do an attach or a replace. | |
3514 | */ | |
3515 | replacing = ztest_random(2); | |
3516 | ||
3517 | /* | |
3518 | * Pick a random top-level vdev. | |
3519 | */ | |
428870ff | 3520 | top = ztest_random_vdev_top(spa, B_TRUE); |
34dc7c2f BB |
3521 | |
3522 | /* | |
3523 | * Pick a random leaf within it. | |
3524 | */ | |
3525 | leaf = ztest_random(leaves); | |
3526 | ||
3527 | /* | |
b128c09f | 3528 | * Locate this vdev. |
34dc7c2f | 3529 | */ |
b128c09f | 3530 | oldvd = rvd->vdev_child[top]; |
cc99f275 DB |
3531 | |
3532 | /* pick a child from the mirror */ | |
428870ff | 3533 | if (zs->zs_mirrors >= 1) { |
a9eaae06 RM |
3534 | ASSERT3P(oldvd->vdev_ops, ==, &vdev_mirror_ops); |
3535 | ASSERT3U(oldvd->vdev_children, >=, zs->zs_mirrors); | |
b2255edc | 3536 | oldvd = oldvd->vdev_child[leaf / ztest_opts.zo_raid_children]; |
fb5f0bc8 | 3537 | } |
cc99f275 DB |
3538 | |
3539 | /* pick a child out of the raidz group */ | |
b2255edc BB |
3540 | if (ztest_opts.zo_raid_children > 1) { |
3541 | if (strcmp(oldvd->vdev_ops->vdev_op_type, "raidz") == 0) | |
a9eaae06 | 3542 | ASSERT3P(oldvd->vdev_ops, ==, &vdev_raidz_ops); |
b2255edc | 3543 | else |
a9eaae06 RM |
3544 | ASSERT3P(oldvd->vdev_ops, ==, &vdev_draid_ops); |
3545 | ASSERT3U(oldvd->vdev_children, ==, ztest_opts.zo_raid_children); | |
b2255edc | 3546 | oldvd = oldvd->vdev_child[leaf % ztest_opts.zo_raid_children]; |
fb5f0bc8 | 3547 | } |
34dc7c2f BB |
3548 | |
3549 | /* | |
b128c09f BB |
3550 | * If we're already doing an attach or replace, oldvd may be a |
3551 | * mirror vdev -- in which case, pick a random child. | |
34dc7c2f | 3552 | */ |
b128c09f BB |
3553 | while (oldvd->vdev_children != 0) { |
3554 | oldvd_has_siblings = B_TRUE; | |
a9eaae06 | 3555 | ASSERT3U(oldvd->vdev_children, >=, 2); |
fb5f0bc8 | 3556 | oldvd = oldvd->vdev_child[ztest_random(oldvd->vdev_children)]; |
b128c09f BB |
3557 | } |
3558 | ||
3559 | oldguid = oldvd->vdev_guid; | |
9babb374 | 3560 | oldsize = vdev_get_min_asize(oldvd); |
b128c09f BB |
3561 | oldvd_is_log = oldvd->vdev_top->vdev_islog; |
3562 | (void) strcpy(oldpath, oldvd->vdev_path); | |
3563 | pvd = oldvd->vdev_parent; | |
fb5f0bc8 | 3564 | pguid = pvd->vdev_guid; |
34dc7c2f BB |
3565 | |
3566 | /* | |
ce5fb2a7 BB |
3567 | * If oldvd has siblings, then half of the time, detach it. Prior |
3568 | * to the detach the pool is scrubbed in order to prevent creating | |
3569 | * unrepairable blocks as a result of the data corruption injection. | |
34dc7c2f | 3570 | */ |
b128c09f | 3571 | if (oldvd_has_siblings && ztest_random(2) == 0) { |
a1d477c2 | 3572 | spa_config_exit(spa, SCL_ALL, FTAG); |
ce5fb2a7 BB |
3573 | |
3574 | error = ztest_scrub_impl(spa); | |
3575 | if (error) | |
3576 | goto out; | |
3577 | ||
fb5f0bc8 BB |
3578 | error = spa_vdev_detach(spa, oldguid, pguid, B_FALSE); |
3579 | if (error != 0 && error != ENODEV && error != EBUSY && | |
d2734cce SD |
3580 | error != ENOTSUP && error != ZFS_ERR_CHECKPOINT_EXISTS && |
3581 | error != ZFS_ERR_DISCARDING_CHECKPOINT) | |
fb5f0bc8 | 3582 | fatal(0, "detach (%s) returned %d", oldpath, error); |
40b84e7a | 3583 | goto out; |
b128c09f | 3584 | } |
34dc7c2f BB |
3585 | |
3586 | /* | |
b128c09f BB |
3587 | * For the new vdev, choose with equal probability between the two |
3588 | * standard paths (ending in either 'a' or 'b') or a random hot spare. | |
34dc7c2f | 3589 | */ |
b128c09f BB |
3590 | if (sav->sav_count != 0 && ztest_random(3) == 0) { |
3591 | newvd = sav->sav_vdevs[ztest_random(sav->sav_count)]; | |
3592 | newvd_is_spare = B_TRUE; | |
b2255edc BB |
3593 | |
3594 | if (newvd->vdev_ops == &vdev_draid_spare_ops) | |
3595 | newvd_is_dspare = B_TRUE; | |
3596 | ||
b128c09f BB |
3597 | (void) strcpy(newpath, newvd->vdev_path); |
3598 | } else { | |
6aec1cd5 | 3599 | (void) snprintf(newpath, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
3600 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
3601 | top * leaves + leaf); | |
b128c09f BB |
3602 | if (ztest_random(2) == 0) |
3603 | newpath[strlen(newpath) - 1] = 'b'; | |
3604 | newvd = vdev_lookup_by_path(rvd, newpath); | |
3605 | } | |
3606 | ||
3607 | if (newvd) { | |
a1d477c2 MA |
3608 | /* |
3609 | * Reopen to ensure the vdev's asize field isn't stale. | |
3610 | */ | |
3611 | vdev_reopen(newvd); | |
9babb374 | 3612 | newsize = vdev_get_min_asize(newvd); |
b128c09f BB |
3613 | } else { |
3614 | /* | |
3615 | * Make newsize a little bigger or smaller than oldsize. | |
3616 | * If it's smaller, the attach should fail. | |
3617 | * If it's larger, and we're doing a replace, | |
3618 | * we should get dynamic LUN growth when we're done. | |
3619 | */ | |
3620 | newsize = 10 * oldsize / (9 + ztest_random(3)); | |
3621 | } | |
34dc7c2f BB |
3622 | |
3623 | /* | |
3624 | * If pvd is not a mirror or root, the attach should fail with ENOTSUP, | |
3625 | * unless it's a replace; in that case any non-replacing parent is OK. | |
3626 | * | |
3627 | * If newvd is already part of the pool, it should fail with EBUSY. | |
3628 | * | |
3629 | * If newvd is too small, it should fail with EOVERFLOW. | |
b2255edc BB |
3630 | * |
3631 | * If newvd is a distributed spare and it's being attached to a | |
3632 | * dRAID which is not its parent it should fail with EINVAL. | |
34dc7c2f | 3633 | */ |
b128c09f BB |
3634 | if (pvd->vdev_ops != &vdev_mirror_ops && |
3635 | pvd->vdev_ops != &vdev_root_ops && (!replacing || | |
3636 | pvd->vdev_ops == &vdev_replacing_ops || | |
3637 | pvd->vdev_ops == &vdev_spare_ops)) | |
34dc7c2f | 3638 | expected_error = ENOTSUP; |
b128c09f BB |
3639 | else if (newvd_is_spare && (!replacing || oldvd_is_log)) |
3640 | expected_error = ENOTSUP; | |
3641 | else if (newvd == oldvd) | |
3642 | expected_error = replacing ? 0 : EBUSY; | |
3643 | else if (vdev_lookup_by_path(rvd, newpath) != NULL) | |
3644 | expected_error = EBUSY; | |
b2255edc | 3645 | else if (!newvd_is_dspare && newsize < oldsize) |
34dc7c2f BB |
3646 | expected_error = EOVERFLOW; |
3647 | else if (ashift > oldvd->vdev_top->vdev_ashift) | |
3648 | expected_error = EDOM; | |
b2255edc BB |
3649 | else if (newvd_is_dspare && pvd != vdev_draid_spare_get_parent(newvd)) |
3650 | expected_error = ENOTSUP; | |
34dc7c2f BB |
3651 | else |
3652 | expected_error = 0; | |
3653 | ||
a1d477c2 | 3654 | spa_config_exit(spa, SCL_ALL, FTAG); |
34dc7c2f BB |
3655 | |
3656 | /* | |
3657 | * Build the nvlist describing newpath. | |
3658 | */ | |
ea0b2538 | 3659 | root = make_vdev_root(newpath, NULL, NULL, newvd == NULL ? newsize : 0, |
cc99f275 | 3660 | ashift, NULL, 0, 0, 1); |
34dc7c2f | 3661 | |
9a49d3f3 BB |
3662 | /* |
3663 | * When supported select either a healing or sequential resilver. | |
3664 | */ | |
3665 | boolean_t rebuilding = B_FALSE; | |
3666 | if (pvd->vdev_ops == &vdev_mirror_ops || | |
3667 | pvd->vdev_ops == &vdev_root_ops) { | |
3668 | rebuilding = !!ztest_random(2); | |
3669 | } | |
3670 | ||
3671 | error = spa_vdev_attach(spa, oldguid, root, replacing, rebuilding); | |
34dc7c2f | 3672 | |
bea3fc71 | 3673 | fnvlist_free(root); |
34dc7c2f BB |
3674 | |
3675 | /* | |
3676 | * If our parent was the replacing vdev, but the replace completed, | |
3677 | * then instead of failing with ENOTSUP we may either succeed, | |
3678 | * fail with ENODEV, or fail with EOVERFLOW. | |
3679 | */ | |
3680 | if (expected_error == ENOTSUP && | |
3681 | (error == 0 || error == ENODEV || error == EOVERFLOW)) | |
3682 | expected_error = error; | |
3683 | ||
3684 | /* | |
3685 | * If someone grew the LUN, the replacement may be too small. | |
3686 | */ | |
b128c09f | 3687 | if (error == EOVERFLOW || error == EBUSY) |
34dc7c2f BB |
3688 | expected_error = error; |
3689 | ||
d2734cce | 3690 | if (error == ZFS_ERR_CHECKPOINT_EXISTS || |
9a49d3f3 BB |
3691 | error == ZFS_ERR_DISCARDING_CHECKPOINT || |
3692 | error == ZFS_ERR_RESILVER_IN_PROGRESS || | |
3693 | error == ZFS_ERR_REBUILD_IN_PROGRESS) | |
d2734cce SD |
3694 | expected_error = error; |
3695 | ||
b128c09f BB |
3696 | if (error != expected_error && expected_error != EBUSY) { |
3697 | fatal(0, "attach (%s %llu, %s %llu, %d) " | |
3698 | "returned %d, expected %d", | |
5d1f7fb6 GW |
3699 | oldpath, oldsize, newpath, |
3700 | newsize, replacing, error, expected_error); | |
34dc7c2f | 3701 | } |
40b84e7a | 3702 | out: |
c242c188 | 3703 | mutex_exit(&ztest_vdev_lock); |
40b84e7a BB |
3704 | |
3705 | umem_free(oldpath, MAXPATHLEN); | |
3706 | umem_free(newpath, MAXPATHLEN); | |
34dc7c2f BB |
3707 | } |
3708 | ||
a1d477c2 MA |
3709 | /* ARGSUSED */ |
3710 | void | |
3711 | ztest_device_removal(ztest_ds_t *zd, uint64_t id) | |
3712 | { | |
3713 | spa_t *spa = ztest_spa; | |
3714 | vdev_t *vd; | |
3715 | uint64_t guid; | |
9e052db4 | 3716 | int error; |
a1d477c2 MA |
3717 | |
3718 | mutex_enter(&ztest_vdev_lock); | |
3719 | ||
9e052db4 MA |
3720 | if (ztest_device_removal_active) { |
3721 | mutex_exit(&ztest_vdev_lock); | |
3722 | return; | |
3723 | } | |
3724 | ||
3725 | /* | |
3726 | * Remove a random top-level vdev and wait for removal to finish. | |
3727 | */ | |
a1d477c2 MA |
3728 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); |
3729 | vd = vdev_lookup_top(spa, ztest_random_vdev_top(spa, B_FALSE)); | |
3730 | guid = vd->vdev_guid; | |
3731 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
3732 | ||
9e052db4 MA |
3733 | error = spa_vdev_remove(spa, guid, B_FALSE); |
3734 | if (error == 0) { | |
3735 | ztest_device_removal_active = B_TRUE; | |
3736 | mutex_exit(&ztest_vdev_lock); | |
3737 | ||
7d658d29 TC |
3738 | /* |
3739 | * spa->spa_vdev_removal is created in a sync task that | |
3740 | * is initiated via dsl_sync_task_nowait(). Since the | |
3741 | * task may not run before spa_vdev_remove() returns, we | |
3742 | * must wait at least 1 txg to ensure that the removal | |
3743 | * struct has been created. | |
3744 | */ | |
3745 | txg_wait_synced(spa_get_dsl(spa), 0); | |
3746 | ||
c40a1124 | 3747 | while (spa->spa_removing_phys.sr_state == DSS_SCANNING) |
9e052db4 MA |
3748 | txg_wait_synced(spa_get_dsl(spa), 0); |
3749 | } else { | |
3750 | mutex_exit(&ztest_vdev_lock); | |
3751 | return; | |
3752 | } | |
a1d477c2 | 3753 | |
9e052db4 MA |
3754 | /* |
3755 | * The pool needs to be scrubbed after completing device removal. | |
3756 | * Failure to do so may result in checksum errors due to the | |
3757 | * strategy employed by ztest_fault_inject() when selecting which | |
3758 | * offset are redundant and can be damaged. | |
3759 | */ | |
3760 | error = spa_scan(spa, POOL_SCAN_SCRUB); | |
3761 | if (error == 0) { | |
3762 | while (dsl_scan_scrubbing(spa_get_dsl(spa))) | |
3763 | txg_wait_synced(spa_get_dsl(spa), 0); | |
3764 | } | |
3765 | ||
3766 | mutex_enter(&ztest_vdev_lock); | |
3767 | ztest_device_removal_active = B_FALSE; | |
a1d477c2 MA |
3768 | mutex_exit(&ztest_vdev_lock); |
3769 | } | |
3770 | ||
9babb374 BB |
3771 | /* |
3772 | * Callback function which expands the physical size of the vdev. | |
3773 | */ | |
65c7cc49 | 3774 | static vdev_t * |
9babb374 BB |
3775 | grow_vdev(vdev_t *vd, void *arg) |
3776 | { | |
2a8ba608 | 3777 | spa_t *spa __maybe_unused = vd->vdev_spa; |
9babb374 BB |
3778 | size_t *newsize = arg; |
3779 | size_t fsize; | |
3780 | int fd; | |
3781 | ||
a9eaae06 | 3782 | ASSERT3S(spa_config_held(spa, SCL_STATE, RW_READER), ==, SCL_STATE); |
9babb374 BB |
3783 | ASSERT(vd->vdev_ops->vdev_op_leaf); |
3784 | ||
3785 | if ((fd = open(vd->vdev_path, O_RDWR)) == -1) | |
3786 | return (vd); | |
3787 | ||
3788 | fsize = lseek(fd, 0, SEEK_END); | |
a9eaae06 | 3789 | VERIFY0(ftruncate(fd, *newsize)); |
9babb374 | 3790 | |
c242c188 | 3791 | if (ztest_opts.zo_verbose >= 6) { |
9babb374 BB |
3792 | (void) printf("%s grew from %lu to %lu bytes\n", |
3793 | vd->vdev_path, (ulong_t)fsize, (ulong_t)*newsize); | |
3794 | } | |
3795 | (void) close(fd); | |
3796 | return (NULL); | |
3797 | } | |
3798 | ||
3799 | /* | |
3800 | * Callback function which expands a given vdev by calling vdev_online(). | |
3801 | */ | |
3802 | /* ARGSUSED */ | |
65c7cc49 | 3803 | static vdev_t * |
9babb374 BB |
3804 | online_vdev(vdev_t *vd, void *arg) |
3805 | { | |
3806 | spa_t *spa = vd->vdev_spa; | |
3807 | vdev_t *tvd = vd->vdev_top; | |
9babb374 | 3808 | uint64_t guid = vd->vdev_guid; |
428870ff BB |
3809 | uint64_t generation = spa->spa_config_generation + 1; |
3810 | vdev_state_t newstate = VDEV_STATE_UNKNOWN; | |
3811 | int error; | |
9babb374 | 3812 | |
a9eaae06 | 3813 | ASSERT3S(spa_config_held(spa, SCL_STATE, RW_READER), ==, SCL_STATE); |
9babb374 BB |
3814 | ASSERT(vd->vdev_ops->vdev_op_leaf); |
3815 | ||
3816 | /* Calling vdev_online will initialize the new metaslabs */ | |
3817 | spa_config_exit(spa, SCL_STATE, spa); | |
428870ff | 3818 | error = vdev_online(spa, guid, ZFS_ONLINE_EXPAND, &newstate); |
9babb374 BB |
3819 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); |
3820 | ||
428870ff BB |
3821 | /* |
3822 | * If vdev_online returned an error or the underlying vdev_open | |
3823 | * failed then we abort the expand. The only way to know that | |
3824 | * vdev_open fails is by checking the returned newstate. | |
3825 | */ | |
3826 | if (error || newstate != VDEV_STATE_HEALTHY) { | |
c242c188 | 3827 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
3828 | (void) printf("Unable to expand vdev, state %llu, " |
3829 | "error %d\n", (u_longlong_t)newstate, error); | |
3830 | } | |
3831 | return (vd); | |
3832 | } | |
3833 | ASSERT3U(newstate, ==, VDEV_STATE_HEALTHY); | |
3834 | ||
9babb374 BB |
3835 | /* |
3836 | * Since we dropped the lock we need to ensure that we're | |
3837 | * still talking to the original vdev. It's possible this | |
3838 | * vdev may have been detached/replaced while we were | |
3839 | * trying to online it. | |
3840 | */ | |
428870ff | 3841 | if (generation != spa->spa_config_generation) { |
c242c188 | 3842 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
3843 | (void) printf("vdev configuration has changed, " |
3844 | "guid %llu, state %llu, expected gen %llu, " | |
3845 | "got gen %llu\n", | |
3846 | (u_longlong_t)guid, | |
3847 | (u_longlong_t)tvd->vdev_state, | |
3848 | (u_longlong_t)generation, | |
3849 | (u_longlong_t)spa->spa_config_generation); | |
9babb374 BB |
3850 | } |
3851 | return (vd); | |
3852 | } | |
3853 | return (NULL); | |
3854 | } | |
3855 | ||
3856 | /* | |
3857 | * Traverse the vdev tree calling the supplied function. | |
3858 | * We continue to walk the tree until we either have walked all | |
3859 | * children or we receive a non-NULL return from the callback. | |
3860 | * If a NULL callback is passed, then we just return back the first | |
3861 | * leaf vdev we encounter. | |
3862 | */ | |
65c7cc49 | 3863 | static vdev_t * |
9babb374 BB |
3864 | vdev_walk_tree(vdev_t *vd, vdev_t *(*func)(vdev_t *, void *), void *arg) |
3865 | { | |
d6320ddb BB |
3866 | uint_t c; |
3867 | ||
9babb374 BB |
3868 | if (vd->vdev_ops->vdev_op_leaf) { |
3869 | if (func == NULL) | |
3870 | return (vd); | |
3871 | else | |
3872 | return (func(vd, arg)); | |
3873 | } | |
3874 | ||
d6320ddb | 3875 | for (c = 0; c < vd->vdev_children; c++) { |
9babb374 BB |
3876 | vdev_t *cvd = vd->vdev_child[c]; |
3877 | if ((cvd = vdev_walk_tree(cvd, func, arg)) != NULL) | |
3878 | return (cvd); | |
3879 | } | |
3880 | return (NULL); | |
3881 | } | |
3882 | ||
34dc7c2f BB |
3883 | /* |
3884 | * Verify that dynamic LUN growth works as expected. | |
3885 | */ | |
428870ff | 3886 | /* ARGSUSED */ |
34dc7c2f | 3887 | void |
428870ff | 3888 | ztest_vdev_LUN_growth(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3889 | { |
c242c188 | 3890 | spa_t *spa = ztest_spa; |
428870ff BB |
3891 | vdev_t *vd, *tvd; |
3892 | metaslab_class_t *mc; | |
3893 | metaslab_group_t *mg; | |
9babb374 | 3894 | size_t psize, newsize; |
428870ff BB |
3895 | uint64_t top; |
3896 | uint64_t old_class_space, new_class_space, old_ms_count, new_ms_count; | |
34dc7c2f | 3897 | |
d2734cce | 3898 | mutex_enter(&ztest_checkpoint_lock); |
c242c188 | 3899 | mutex_enter(&ztest_vdev_lock); |
9babb374 BB |
3900 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); |
3901 | ||
a1d477c2 MA |
3902 | /* |
3903 | * If there is a vdev removal in progress, it could complete while | |
3904 | * we are running, in which case we would not be able to verify | |
3905 | * that the metaslab_class space increased (because it decreases | |
3906 | * when the device removal completes). | |
3907 | */ | |
9e052db4 | 3908 | if (ztest_device_removal_active) { |
d2734cce | 3909 | spa_config_exit(spa, SCL_STATE, spa); |
a1d477c2 | 3910 | mutex_exit(&ztest_vdev_lock); |
d2734cce | 3911 | mutex_exit(&ztest_checkpoint_lock); |
a1d477c2 MA |
3912 | return; |
3913 | } | |
3914 | ||
428870ff | 3915 | top = ztest_random_vdev_top(spa, B_TRUE); |
9babb374 | 3916 | |
428870ff BB |
3917 | tvd = spa->spa_root_vdev->vdev_child[top]; |
3918 | mg = tvd->vdev_mg; | |
3919 | mc = mg->mg_class; | |
3920 | old_ms_count = tvd->vdev_ms_count; | |
3921 | old_class_space = metaslab_class_get_space(mc); | |
34dc7c2f BB |
3922 | |
3923 | /* | |
9babb374 BB |
3924 | * Determine the size of the first leaf vdev associated with |
3925 | * our top-level device. | |
34dc7c2f | 3926 | */ |
9babb374 BB |
3927 | vd = vdev_walk_tree(tvd, NULL, NULL); |
3928 | ASSERT3P(vd, !=, NULL); | |
3929 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
34dc7c2f | 3930 | |
9babb374 | 3931 | psize = vd->vdev_psize; |
34dc7c2f | 3932 | |
9babb374 | 3933 | /* |
428870ff BB |
3934 | * We only try to expand the vdev if it's healthy, less than 4x its |
3935 | * original size, and it has a valid psize. | |
9babb374 | 3936 | */ |
428870ff | 3937 | if (tvd->vdev_state != VDEV_STATE_HEALTHY || |
c242c188 | 3938 | psize == 0 || psize >= 4 * ztest_opts.zo_vdev_size) { |
9babb374 | 3939 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 3940 | mutex_exit(&ztest_vdev_lock); |
d2734cce | 3941 | mutex_exit(&ztest_checkpoint_lock); |
9babb374 BB |
3942 | return; |
3943 | } | |
a9eaae06 | 3944 | ASSERT3U(psize, >, 0); |
cc99f275 | 3945 | newsize = psize + MAX(psize / 8, SPA_MAXBLOCKSIZE); |
9babb374 | 3946 | ASSERT3U(newsize, >, psize); |
34dc7c2f | 3947 | |
c242c188 | 3948 | if (ztest_opts.zo_verbose >= 6) { |
428870ff | 3949 | (void) printf("Expanding LUN %s from %lu to %lu\n", |
9babb374 BB |
3950 | vd->vdev_path, (ulong_t)psize, (ulong_t)newsize); |
3951 | } | |
3952 | ||
9babb374 BB |
3953 | /* |
3954 | * Growing the vdev is a two step process: | |
3955 | * 1). expand the physical size (i.e. relabel) | |
3956 | * 2). online the vdev to create the new metaslabs | |
3957 | */ | |
3958 | if (vdev_walk_tree(tvd, grow_vdev, &newsize) != NULL || | |
3959 | vdev_walk_tree(tvd, online_vdev, NULL) != NULL || | |
3960 | tvd->vdev_state != VDEV_STATE_HEALTHY) { | |
c242c188 | 3961 | if (ztest_opts.zo_verbose >= 5) { |
9babb374 | 3962 | (void) printf("Could not expand LUN because " |
428870ff | 3963 | "the vdev configuration changed.\n"); |
34dc7c2f | 3964 | } |
428870ff | 3965 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 3966 | mutex_exit(&ztest_vdev_lock); |
d2734cce | 3967 | mutex_exit(&ztest_checkpoint_lock); |
9babb374 | 3968 | return; |
34dc7c2f BB |
3969 | } |
3970 | ||
428870ff | 3971 | spa_config_exit(spa, SCL_STATE, spa); |
9babb374 BB |
3972 | |
3973 | /* | |
3974 | * Expanding the LUN will update the config asynchronously, | |
3975 | * thus we must wait for the async thread to complete any | |
3976 | * pending tasks before proceeding. | |
3977 | */ | |
428870ff BB |
3978 | for (;;) { |
3979 | boolean_t done; | |
3980 | mutex_enter(&spa->spa_async_lock); | |
3981 | done = (spa->spa_async_thread == NULL && !spa->spa_async_tasks); | |
3982 | mutex_exit(&spa->spa_async_lock); | |
3983 | if (done) | |
3984 | break; | |
3985 | txg_wait_synced(spa_get_dsl(spa), 0); | |
3986 | (void) poll(NULL, 0, 100); | |
3987 | } | |
9babb374 BB |
3988 | |
3989 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); | |
428870ff BB |
3990 | |
3991 | tvd = spa->spa_root_vdev->vdev_child[top]; | |
3992 | new_ms_count = tvd->vdev_ms_count; | |
3993 | new_class_space = metaslab_class_get_space(mc); | |
3994 | ||
3995 | if (tvd->vdev_mg != mg || mg->mg_class != mc) { | |
c242c188 | 3996 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
3997 | (void) printf("Could not verify LUN expansion due to " |
3998 | "intervening vdev offline or remove.\n"); | |
3999 | } | |
4000 | spa_config_exit(spa, SCL_STATE, spa); | |
c242c188 | 4001 | mutex_exit(&ztest_vdev_lock); |
d2734cce | 4002 | mutex_exit(&ztest_checkpoint_lock); |
428870ff BB |
4003 | return; |
4004 | } | |
4005 | ||
4006 | /* | |
4007 | * Make sure we were able to grow the vdev. | |
4008 | */ | |
a1d477c2 MA |
4009 | if (new_ms_count <= old_ms_count) { |
4010 | fatal(0, "LUN expansion failed: ms_count %llu < %llu\n", | |
428870ff | 4011 | old_ms_count, new_ms_count); |
a1d477c2 | 4012 | } |
9babb374 BB |
4013 | |
4014 | /* | |
4015 | * Make sure we were able to grow the pool. | |
4016 | */ | |
a1d477c2 MA |
4017 | if (new_class_space <= old_class_space) { |
4018 | fatal(0, "LUN expansion failed: class_space %llu < %llu\n", | |
428870ff | 4019 | old_class_space, new_class_space); |
a1d477c2 | 4020 | } |
428870ff | 4021 | |
c242c188 | 4022 | if (ztest_opts.zo_verbose >= 5) { |
f3c8c9e6 | 4023 | char oldnumbuf[NN_NUMBUF_SZ], newnumbuf[NN_NUMBUF_SZ]; |
9babb374 | 4024 | |
f3c8c9e6 JK |
4025 | nicenum(old_class_space, oldnumbuf, sizeof (oldnumbuf)); |
4026 | nicenum(new_class_space, newnumbuf, sizeof (newnumbuf)); | |
9babb374 BB |
4027 | (void) printf("%s grew from %s to %s\n", |
4028 | spa->spa_name, oldnumbuf, newnumbuf); | |
4029 | } | |
428870ff | 4030 | |
9babb374 | 4031 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 4032 | mutex_exit(&ztest_vdev_lock); |
d2734cce | 4033 | mutex_exit(&ztest_checkpoint_lock); |
34dc7c2f BB |
4034 | } |
4035 | ||
428870ff BB |
4036 | /* |
4037 | * Verify that dmu_objset_{create,destroy,open,close} work as expected. | |
4038 | */ | |
34dc7c2f BB |
4039 | /* ARGSUSED */ |
4040 | static void | |
428870ff | 4041 | ztest_objset_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) |
34dc7c2f BB |
4042 | { |
4043 | /* | |
428870ff | 4044 | * Create the objects common to all ztest datasets. |
34dc7c2f | 4045 | */ |
a9eaae06 RM |
4046 | VERIFY0(zap_create_claim(os, ZTEST_DIROBJ, |
4047 | DMU_OT_ZAP_OTHER, DMU_OT_NONE, 0, tx)); | |
428870ff | 4048 | } |
34dc7c2f | 4049 | |
428870ff BB |
4050 | static int |
4051 | ztest_dataset_create(char *dsname) | |
4052 | { | |
4807c0ba TC |
4053 | int err; |
4054 | uint64_t rand; | |
4055 | dsl_crypto_params_t *dcp = NULL; | |
4056 | ||
4057 | /* | |
4058 | * 50% of the time, we create encrypted datasets | |
4059 | * using a random cipher suite and a hard-coded | |
4060 | * wrapping key. | |
4061 | */ | |
4062 | rand = ztest_random(2); | |
4063 | if (rand != 0) { | |
4064 | nvlist_t *crypto_args = fnvlist_alloc(); | |
4065 | nvlist_t *props = fnvlist_alloc(); | |
4066 | ||
4067 | /* slight bias towards the default cipher suite */ | |
4068 | rand = ztest_random(ZIO_CRYPT_FUNCTIONS); | |
4069 | if (rand < ZIO_CRYPT_AES_128_CCM) | |
4070 | rand = ZIO_CRYPT_ON; | |
4071 | ||
4072 | fnvlist_add_uint64(props, | |
4073 | zfs_prop_to_name(ZFS_PROP_ENCRYPTION), rand); | |
4074 | fnvlist_add_uint8_array(crypto_args, "wkeydata", | |
4075 | (uint8_t *)ztest_wkeydata, WRAPPING_KEY_LEN); | |
4076 | ||
4077 | /* | |
4078 | * These parameters aren't really used by the kernel. They | |
4079 | * are simply stored so that userspace knows how to load | |
4080 | * the wrapping key. | |
4081 | */ | |
4082 | fnvlist_add_uint64(props, | |
4083 | zfs_prop_to_name(ZFS_PROP_KEYFORMAT), ZFS_KEYFORMAT_RAW); | |
4084 | fnvlist_add_string(props, | |
4085 | zfs_prop_to_name(ZFS_PROP_KEYLOCATION), "prompt"); | |
4086 | fnvlist_add_uint64(props, | |
4087 | zfs_prop_to_name(ZFS_PROP_PBKDF2_SALT), 0ULL); | |
4088 | fnvlist_add_uint64(props, | |
4089 | zfs_prop_to_name(ZFS_PROP_PBKDF2_ITERS), 0ULL); | |
4090 | ||
4091 | VERIFY0(dsl_crypto_params_create_nvlist(DCP_CMD_NONE, props, | |
4092 | crypto_args, &dcp)); | |
4093 | ||
010d1247 NL |
4094 | /* |
4095 | * Cycle through all available encryption implementations | |
4096 | * to verify interoperability. | |
4097 | */ | |
4098 | VERIFY0(gcm_impl_set("cycle")); | |
4099 | VERIFY0(aes_impl_set("cycle")); | |
4100 | ||
4807c0ba TC |
4101 | fnvlist_free(crypto_args); |
4102 | fnvlist_free(props); | |
4103 | } | |
4104 | ||
4105 | err = dmu_objset_create(dsname, DMU_OST_OTHER, 0, dcp, | |
428870ff | 4106 | ztest_objset_create_cb, NULL); |
4807c0ba | 4107 | dsl_crypto_params_free(dcp, !!err); |
428870ff | 4108 | |
4807c0ba TC |
4109 | rand = ztest_random(100); |
4110 | if (err || rand < 80) | |
428870ff BB |
4111 | return (err); |
4112 | ||
c242c188 | 4113 | if (ztest_opts.zo_verbose >= 5) |
b815ff9a | 4114 | (void) printf("Setting dataset %s to sync always\n", dsname); |
428870ff BB |
4115 | return (ztest_dsl_prop_set_uint64(dsname, ZFS_PROP_SYNC, |
4116 | ZFS_SYNC_ALWAYS, B_FALSE)); | |
34dc7c2f BB |
4117 | } |
4118 | ||
428870ff | 4119 | /* ARGSUSED */ |
34dc7c2f | 4120 | static int |
428870ff | 4121 | ztest_objset_destroy_cb(const char *name, void *arg) |
34dc7c2f | 4122 | { |
34dc7c2f | 4123 | objset_t *os; |
428870ff | 4124 | dmu_object_info_t doi; |
34dc7c2f BB |
4125 | int error; |
4126 | ||
4127 | /* | |
4128 | * Verify that the dataset contains a directory object. | |
4129 | */ | |
4807c0ba TC |
4130 | VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_TRUE, |
4131 | B_TRUE, FTAG, &os)); | |
428870ff | 4132 | error = dmu_object_info(os, ZTEST_DIROBJ, &doi); |
34dc7c2f BB |
4133 | if (error != ENOENT) { |
4134 | /* We could have crashed in the middle of destroying it */ | |
c99c9001 | 4135 | ASSERT0(error); |
428870ff BB |
4136 | ASSERT3U(doi.doi_type, ==, DMU_OT_ZAP_OTHER); |
4137 | ASSERT3S(doi.doi_physical_blocks_512, >=, 0); | |
34dc7c2f | 4138 | } |
b5256303 | 4139 | dmu_objset_disown(os, B_TRUE, FTAG); |
34dc7c2f BB |
4140 | |
4141 | /* | |
4142 | * Destroy the dataset. | |
4143 | */ | |
13fe0198 | 4144 | if (strchr(name, '@') != NULL) { |
dc1fbc43 | 4145 | VERIFY0(dsl_destroy_snapshot(name, B_TRUE)); |
13fe0198 | 4146 | } else { |
dc1fbc43 | 4147 | error = dsl_destroy_head(name); |
0dd6b6bf BB |
4148 | if (error == ENOSPC) { |
4149 | /* There could be checkpoint or insufficient slop */ | |
4150 | ztest_record_enospc(FTAG); | |
4151 | } else if (error != EBUSY) { | |
4152 | /* There could be a hold on this dataset */ | |
dc1fbc43 | 4153 | ASSERT0(error); |
0dd6b6bf | 4154 | } |
13fe0198 | 4155 | } |
34dc7c2f BB |
4156 | return (0); |
4157 | } | |
4158 | ||
428870ff BB |
4159 | static boolean_t |
4160 | ztest_snapshot_create(char *osname, uint64_t id) | |
34dc7c2f | 4161 | { |
eca7b760 | 4162 | char snapname[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
4163 | int error; |
4164 | ||
13fe0198 | 4165 | (void) snprintf(snapname, sizeof (snapname), "%llu", (u_longlong_t)id); |
428870ff | 4166 | |
13fe0198 | 4167 | error = dmu_objset_snapshot_one(osname, snapname); |
428870ff BB |
4168 | if (error == ENOSPC) { |
4169 | ztest_record_enospc(FTAG); | |
4170 | return (B_FALSE); | |
4171 | } | |
13fe0198 MA |
4172 | if (error != 0 && error != EEXIST) { |
4173 | fatal(0, "ztest_snapshot_create(%s@%s) = %d", osname, | |
4174 | snapname, error); | |
4175 | } | |
428870ff BB |
4176 | return (B_TRUE); |
4177 | } | |
4178 | ||
4179 | static boolean_t | |
4180 | ztest_snapshot_destroy(char *osname, uint64_t id) | |
4181 | { | |
eca7b760 | 4182 | char snapname[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
4183 | int error; |
4184 | ||
eca7b760 | 4185 | (void) snprintf(snapname, sizeof (snapname), "%s@%llu", osname, |
428870ff BB |
4186 | (u_longlong_t)id); |
4187 | ||
13fe0198 | 4188 | error = dsl_destroy_snapshot(snapname, B_FALSE); |
428870ff BB |
4189 | if (error != 0 && error != ENOENT) |
4190 | fatal(0, "ztest_snapshot_destroy(%s) = %d", snapname, error); | |
4191 | return (B_TRUE); | |
34dc7c2f BB |
4192 | } |
4193 | ||
428870ff | 4194 | /* ARGSUSED */ |
34dc7c2f | 4195 | void |
428870ff | 4196 | ztest_dmu_objset_create_destroy(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4197 | { |
40b84e7a | 4198 | ztest_ds_t *zdtmp; |
428870ff | 4199 | int iters; |
34dc7c2f | 4200 | int error; |
b128c09f | 4201 | objset_t *os, *os2; |
eca7b760 | 4202 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
34dc7c2f | 4203 | zilog_t *zilog; |
d6320ddb | 4204 | int i; |
34dc7c2f | 4205 | |
40b84e7a | 4206 | zdtmp = umem_alloc(sizeof (ztest_ds_t), UMEM_NOFAIL); |
40b84e7a | 4207 | |
2d9142c9 | 4208 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
34dc7c2f | 4209 | |
eca7b760 | 4210 | (void) snprintf(name, sizeof (name), "%s/temp_%llu", |
c242c188 | 4211 | ztest_opts.zo_pool, (u_longlong_t)id); |
34dc7c2f BB |
4212 | |
4213 | /* | |
4214 | * If this dataset exists from a previous run, process its replay log | |
13fe0198 | 4215 | * half of the time. If we don't replay it, then dsl_destroy_head() |
428870ff | 4216 | * (invoked from ztest_objset_destroy_cb()) should just throw it away. |
34dc7c2f BB |
4217 | */ |
4218 | if (ztest_random(2) == 0 && | |
4807c0ba | 4219 | ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, |
b5256303 | 4220 | B_TRUE, FTAG, &os) == 0) { |
c242c188 | 4221 | ztest_zd_init(zdtmp, NULL, os); |
40b84e7a BB |
4222 | zil_replay(os, zdtmp, ztest_replay_vector); |
4223 | ztest_zd_fini(zdtmp); | |
b5256303 | 4224 | dmu_objset_disown(os, B_TRUE, FTAG); |
34dc7c2f BB |
4225 | } |
4226 | ||
4227 | /* | |
4228 | * There may be an old instance of the dataset we're about to | |
4229 | * create lying around from a previous run. If so, destroy it | |
4230 | * and all of its snapshots. | |
4231 | */ | |
428870ff | 4232 | (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL, |
34dc7c2f BB |
4233 | DS_FIND_CHILDREN | DS_FIND_SNAPSHOTS); |
4234 | ||
4235 | /* | |
4236 | * Verify that the destroyed dataset is no longer in the namespace. | |
4237 | */ | |
4807c0ba TC |
4238 | VERIFY3U(ENOENT, ==, ztest_dmu_objset_own(name, DMU_OST_OTHER, B_TRUE, |
4239 | B_TRUE, FTAG, &os)); | |
34dc7c2f BB |
4240 | |
4241 | /* | |
4242 | * Verify that we can create a new dataset. | |
4243 | */ | |
428870ff | 4244 | error = ztest_dataset_create(name); |
34dc7c2f BB |
4245 | if (error) { |
4246 | if (error == ENOSPC) { | |
428870ff | 4247 | ztest_record_enospc(FTAG); |
40b84e7a | 4248 | goto out; |
34dc7c2f BB |
4249 | } |
4250 | fatal(0, "dmu_objset_create(%s) = %d", name, error); | |
4251 | } | |
4252 | ||
4807c0ba | 4253 | VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, B_TRUE, |
b5256303 | 4254 | FTAG, &os)); |
428870ff | 4255 | |
c242c188 | 4256 | ztest_zd_init(zdtmp, NULL, os); |
34dc7c2f BB |
4257 | |
4258 | /* | |
4259 | * Open the intent log for it. | |
4260 | */ | |
428870ff | 4261 | zilog = zil_open(os, ztest_get_data); |
34dc7c2f BB |
4262 | |
4263 | /* | |
428870ff BB |
4264 | * Put some objects in there, do a little I/O to them, |
4265 | * and randomly take a couple of snapshots along the way. | |
34dc7c2f | 4266 | */ |
428870ff | 4267 | iters = ztest_random(5); |
d6320ddb | 4268 | for (i = 0; i < iters; i++) { |
40b84e7a | 4269 | ztest_dmu_object_alloc_free(zdtmp, id); |
428870ff BB |
4270 | if (ztest_random(iters) == 0) |
4271 | (void) ztest_snapshot_create(name, i); | |
34dc7c2f BB |
4272 | } |
4273 | ||
4274 | /* | |
4275 | * Verify that we cannot create an existing dataset. | |
4276 | */ | |
428870ff | 4277 | VERIFY3U(EEXIST, ==, |
b5256303 | 4278 | dmu_objset_create(name, DMU_OST_OTHER, 0, NULL, NULL, NULL)); |
34dc7c2f BB |
4279 | |
4280 | /* | |
428870ff | 4281 | * Verify that we can hold an objset that is also owned. |
b128c09f | 4282 | */ |
a9eaae06 | 4283 | VERIFY0(dmu_objset_hold(name, FTAG, &os2)); |
428870ff | 4284 | dmu_objset_rele(os2, FTAG); |
34dc7c2f | 4285 | |
428870ff BB |
4286 | /* |
4287 | * Verify that we cannot own an objset that is already owned. | |
4288 | */ | |
4807c0ba TC |
4289 | VERIFY3U(EBUSY, ==, ztest_dmu_objset_own(name, DMU_OST_OTHER, |
4290 | B_FALSE, B_TRUE, FTAG, &os2)); | |
34dc7c2f | 4291 | |
428870ff | 4292 | zil_close(zilog); |
b5256303 | 4293 | dmu_objset_disown(os, B_TRUE, FTAG); |
40b84e7a BB |
4294 | ztest_zd_fini(zdtmp); |
4295 | out: | |
2d9142c9 | 4296 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
40b84e7a | 4297 | |
40b84e7a | 4298 | umem_free(zdtmp, sizeof (ztest_ds_t)); |
34dc7c2f BB |
4299 | } |
4300 | ||
4301 | /* | |
4302 | * Verify that dmu_snapshot_{create,destroy,open,close} work as expected. | |
4303 | */ | |
4304 | void | |
428870ff | 4305 | ztest_dmu_snapshot_create_destroy(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4306 | { |
2d9142c9 | 4307 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
428870ff BB |
4308 | (void) ztest_snapshot_destroy(zd->zd_name, id); |
4309 | (void) ztest_snapshot_create(zd->zd_name, id); | |
2d9142c9 | 4310 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
34dc7c2f BB |
4311 | } |
4312 | ||
9babb374 BB |
4313 | /* |
4314 | * Cleanup non-standard snapshots and clones. | |
4315 | */ | |
65c7cc49 | 4316 | static void |
428870ff | 4317 | ztest_dsl_dataset_cleanup(char *osname, uint64_t id) |
9babb374 | 4318 | { |
40b84e7a BB |
4319 | char *snap1name; |
4320 | char *clone1name; | |
4321 | char *snap2name; | |
4322 | char *clone2name; | |
4323 | char *snap3name; | |
9babb374 BB |
4324 | int error; |
4325 | ||
eca7b760 IK |
4326 | snap1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); |
4327 | clone1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4328 | snap2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4329 | clone2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4330 | snap3name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4331 | ||
4332 | (void) snprintf(snap1name, ZFS_MAX_DATASET_NAME_LEN, | |
4333 | "%s@s1_%llu", osname, (u_longlong_t)id); | |
4334 | (void) snprintf(clone1name, ZFS_MAX_DATASET_NAME_LEN, | |
4335 | "%s/c1_%llu", osname, (u_longlong_t)id); | |
4336 | (void) snprintf(snap2name, ZFS_MAX_DATASET_NAME_LEN, | |
4337 | "%s@s2_%llu", clone1name, (u_longlong_t)id); | |
4338 | (void) snprintf(clone2name, ZFS_MAX_DATASET_NAME_LEN, | |
4339 | "%s/c2_%llu", osname, (u_longlong_t)id); | |
4340 | (void) snprintf(snap3name, ZFS_MAX_DATASET_NAME_LEN, | |
4341 | "%s@s3_%llu", clone1name, (u_longlong_t)id); | |
9babb374 | 4342 | |
13fe0198 | 4343 | error = dsl_destroy_head(clone2name); |
9babb374 | 4344 | if (error && error != ENOENT) |
13fe0198 MA |
4345 | fatal(0, "dsl_destroy_head(%s) = %d", clone2name, error); |
4346 | error = dsl_destroy_snapshot(snap3name, B_FALSE); | |
9babb374 | 4347 | if (error && error != ENOENT) |
13fe0198 MA |
4348 | fatal(0, "dsl_destroy_snapshot(%s) = %d", snap3name, error); |
4349 | error = dsl_destroy_snapshot(snap2name, B_FALSE); | |
9babb374 | 4350 | if (error && error != ENOENT) |
13fe0198 MA |
4351 | fatal(0, "dsl_destroy_snapshot(%s) = %d", snap2name, error); |
4352 | error = dsl_destroy_head(clone1name); | |
9babb374 | 4353 | if (error && error != ENOENT) |
13fe0198 MA |
4354 | fatal(0, "dsl_destroy_head(%s) = %d", clone1name, error); |
4355 | error = dsl_destroy_snapshot(snap1name, B_FALSE); | |
9babb374 | 4356 | if (error && error != ENOENT) |
13fe0198 | 4357 | fatal(0, "dsl_destroy_snapshot(%s) = %d", snap1name, error); |
40b84e7a | 4358 | |
eca7b760 IK |
4359 | umem_free(snap1name, ZFS_MAX_DATASET_NAME_LEN); |
4360 | umem_free(clone1name, ZFS_MAX_DATASET_NAME_LEN); | |
4361 | umem_free(snap2name, ZFS_MAX_DATASET_NAME_LEN); | |
4362 | umem_free(clone2name, ZFS_MAX_DATASET_NAME_LEN); | |
4363 | umem_free(snap3name, ZFS_MAX_DATASET_NAME_LEN); | |
9babb374 BB |
4364 | } |
4365 | ||
4366 | /* | |
4367 | * Verify dsl_dataset_promote handles EBUSY | |
4368 | */ | |
4369 | void | |
428870ff | 4370 | ztest_dsl_dataset_promote_busy(ztest_ds_t *zd, uint64_t id) |
9babb374 | 4371 | { |
13fe0198 | 4372 | objset_t *os; |
40b84e7a BB |
4373 | char *snap1name; |
4374 | char *clone1name; | |
4375 | char *snap2name; | |
4376 | char *clone2name; | |
4377 | char *snap3name; | |
428870ff BB |
4378 | char *osname = zd->zd_name; |
4379 | int error; | |
9babb374 | 4380 | |
eca7b760 IK |
4381 | snap1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); |
4382 | clone1name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4383 | snap2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4384 | clone2name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
4385 | snap3name = umem_alloc(ZFS_MAX_DATASET_NAME_LEN, UMEM_NOFAIL); | |
40b84e7a | 4386 | |
2d9142c9 | 4387 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
9babb374 | 4388 | |
428870ff | 4389 | ztest_dsl_dataset_cleanup(osname, id); |
9babb374 | 4390 | |
eca7b760 IK |
4391 | (void) snprintf(snap1name, ZFS_MAX_DATASET_NAME_LEN, |
4392 | "%s@s1_%llu", osname, (u_longlong_t)id); | |
4393 | (void) snprintf(clone1name, ZFS_MAX_DATASET_NAME_LEN, | |
4394 | "%s/c1_%llu", osname, (u_longlong_t)id); | |
4395 | (void) snprintf(snap2name, ZFS_MAX_DATASET_NAME_LEN, | |
4396 | "%s@s2_%llu", clone1name, (u_longlong_t)id); | |
4397 | (void) snprintf(clone2name, ZFS_MAX_DATASET_NAME_LEN, | |
4398 | "%s/c2_%llu", osname, (u_longlong_t)id); | |
4399 | (void) snprintf(snap3name, ZFS_MAX_DATASET_NAME_LEN, | |
4400 | "%s@s3_%llu", clone1name, (u_longlong_t)id); | |
9babb374 | 4401 | |
6f1ffb06 | 4402 | error = dmu_objset_snapshot_one(osname, strchr(snap1name, '@') + 1); |
9babb374 BB |
4403 | if (error && error != EEXIST) { |
4404 | if (error == ENOSPC) { | |
428870ff | 4405 | ztest_record_enospc(FTAG); |
9babb374 BB |
4406 | goto out; |
4407 | } | |
4408 | fatal(0, "dmu_take_snapshot(%s) = %d", snap1name, error); | |
4409 | } | |
4410 | ||
13fe0198 | 4411 | error = dmu_objset_clone(clone1name, snap1name); |
9babb374 BB |
4412 | if (error) { |
4413 | if (error == ENOSPC) { | |
428870ff | 4414 | ztest_record_enospc(FTAG); |
9babb374 BB |
4415 | goto out; |
4416 | } | |
4417 | fatal(0, "dmu_objset_create(%s) = %d", clone1name, error); | |
4418 | } | |
4419 | ||
6f1ffb06 | 4420 | error = dmu_objset_snapshot_one(clone1name, strchr(snap2name, '@') + 1); |
428870ff BB |
4421 | if (error && error != EEXIST) { |
4422 | if (error == ENOSPC) { | |
4423 | ztest_record_enospc(FTAG); | |
4424 | goto out; | |
34dc7c2f | 4425 | } |
428870ff BB |
4426 | fatal(0, "dmu_open_snapshot(%s) = %d", snap2name, error); |
4427 | } | |
34dc7c2f | 4428 | |
6f1ffb06 | 4429 | error = dmu_objset_snapshot_one(clone1name, strchr(snap3name, '@') + 1); |
428870ff BB |
4430 | if (error && error != EEXIST) { |
4431 | if (error == ENOSPC) { | |
4432 | ztest_record_enospc(FTAG); | |
4433 | goto out; | |
4434 | } | |
4435 | fatal(0, "dmu_open_snapshot(%s) = %d", snap3name, error); | |
4436 | } | |
34dc7c2f | 4437 | |
13fe0198 | 4438 | error = dmu_objset_clone(clone2name, snap3name); |
428870ff BB |
4439 | if (error) { |
4440 | if (error == ENOSPC) { | |
4441 | ztest_record_enospc(FTAG); | |
4442 | goto out; | |
4443 | } | |
4444 | fatal(0, "dmu_objset_create(%s) = %d", clone2name, error); | |
4445 | } | |
34dc7c2f | 4446 | |
4807c0ba | 4447 | error = ztest_dmu_objset_own(snap2name, DMU_OST_ANY, B_TRUE, B_TRUE, |
b5256303 | 4448 | FTAG, &os); |
428870ff | 4449 | if (error) |
13fe0198 | 4450 | fatal(0, "dmu_objset_own(%s) = %d", snap2name, error); |
428870ff | 4451 | error = dsl_dataset_promote(clone2name, NULL); |
9b67f605 | 4452 | if (error == ENOSPC) { |
b5256303 | 4453 | dmu_objset_disown(os, B_TRUE, FTAG); |
9b67f605 MA |
4454 | ztest_record_enospc(FTAG); |
4455 | goto out; | |
4456 | } | |
428870ff BB |
4457 | if (error != EBUSY) |
4458 | fatal(0, "dsl_dataset_promote(%s), %d, not EBUSY", clone2name, | |
4459 | error); | |
b5256303 | 4460 | dmu_objset_disown(os, B_TRUE, FTAG); |
34dc7c2f | 4461 | |
428870ff BB |
4462 | out: |
4463 | ztest_dsl_dataset_cleanup(osname, id); | |
34dc7c2f | 4464 | |
2d9142c9 | 4465 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
40b84e7a | 4466 | |
eca7b760 IK |
4467 | umem_free(snap1name, ZFS_MAX_DATASET_NAME_LEN); |
4468 | umem_free(clone1name, ZFS_MAX_DATASET_NAME_LEN); | |
4469 | umem_free(snap2name, ZFS_MAX_DATASET_NAME_LEN); | |
4470 | umem_free(clone2name, ZFS_MAX_DATASET_NAME_LEN); | |
4471 | umem_free(snap3name, ZFS_MAX_DATASET_NAME_LEN); | |
428870ff | 4472 | } |
34dc7c2f | 4473 | |
40b84e7a | 4474 | #undef OD_ARRAY_SIZE |
d1d7e268 | 4475 | #define OD_ARRAY_SIZE 4 |
40b84e7a | 4476 | |
428870ff BB |
4477 | /* |
4478 | * Verify that dmu_object_{alloc,free} work as expected. | |
4479 | */ | |
4480 | void | |
4481 | ztest_dmu_object_alloc_free(ztest_ds_t *zd, uint64_t id) | |
4482 | { | |
40b84e7a BB |
4483 | ztest_od_t *od; |
4484 | int batchsize; | |
4485 | int size; | |
d6320ddb | 4486 | int b; |
34dc7c2f | 4487 | |
d1d7e268 | 4488 | size = sizeof (ztest_od_t) * OD_ARRAY_SIZE; |
40b84e7a BB |
4489 | od = umem_alloc(size, UMEM_NOFAIL); |
4490 | batchsize = OD_ARRAY_SIZE; | |
4491 | ||
d6320ddb | 4492 | for (b = 0; b < batchsize; b++) |
50c957f7 NB |
4493 | ztest_od_init(od + b, id, FTAG, b, DMU_OT_UINT64_OTHER, |
4494 | 0, 0, 0); | |
34dc7c2f | 4495 | |
428870ff BB |
4496 | /* |
4497 | * Destroy the previous batch of objects, create a new batch, | |
4498 | * and do some I/O on the new objects. | |
4499 | */ | |
40b84e7a | 4500 | if (ztest_object_init(zd, od, size, B_TRUE) != 0) |
428870ff | 4501 | return; |
34dc7c2f | 4502 | |
428870ff BB |
4503 | while (ztest_random(4 * batchsize) != 0) |
4504 | ztest_io(zd, od[ztest_random(batchsize)].od_object, | |
4505 | ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
40b84e7a BB |
4506 | |
4507 | umem_free(od, size); | |
34dc7c2f BB |
4508 | } |
4509 | ||
047116ac TC |
4510 | /* |
4511 | * Rewind the global allocator to verify object allocation backfilling. | |
4512 | */ | |
4513 | void | |
4514 | ztest_dmu_object_next_chunk(ztest_ds_t *zd, uint64_t id) | |
4515 | { | |
4516 | objset_t *os = zd->zd_os; | |
4517 | int dnodes_per_chunk = 1 << dmu_object_alloc_chunk_shift; | |
4518 | uint64_t object; | |
4519 | ||
4520 | /* | |
4521 | * Rewind the global allocator randomly back to a lower object number | |
4522 | * to force backfilling and reclamation of recently freed dnodes. | |
4523 | */ | |
4524 | mutex_enter(&os->os_obj_lock); | |
4525 | object = ztest_random(os->os_obj_next_chunk); | |
4526 | os->os_obj_next_chunk = P2ALIGN(object, dnodes_per_chunk); | |
4527 | mutex_exit(&os->os_obj_lock); | |
4528 | } | |
4529 | ||
40b84e7a | 4530 | #undef OD_ARRAY_SIZE |
d1d7e268 | 4531 | #define OD_ARRAY_SIZE 2 |
40b84e7a | 4532 | |
34dc7c2f BB |
4533 | /* |
4534 | * Verify that dmu_{read,write} work as expected. | |
4535 | */ | |
34dc7c2f | 4536 | void |
428870ff | 4537 | ztest_dmu_read_write(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4538 | { |
40b84e7a BB |
4539 | int size; |
4540 | ztest_od_t *od; | |
4541 | ||
428870ff | 4542 | objset_t *os = zd->zd_os; |
d1d7e268 | 4543 | size = sizeof (ztest_od_t) * OD_ARRAY_SIZE; |
40b84e7a | 4544 | od = umem_alloc(size, UMEM_NOFAIL); |
34dc7c2f BB |
4545 | dmu_tx_t *tx; |
4546 | int i, freeit, error; | |
4547 | uint64_t n, s, txg; | |
4548 | bufwad_t *packbuf, *bigbuf, *pack, *bigH, *bigT; | |
428870ff BB |
4549 | uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize; |
4550 | uint64_t chunksize = (1000 + ztest_random(1000)) * sizeof (uint64_t); | |
34dc7c2f BB |
4551 | uint64_t regions = 997; |
4552 | uint64_t stride = 123456789ULL; | |
4553 | uint64_t width = 40; | |
4554 | int free_percent = 5; | |
4555 | ||
4556 | /* | |
4557 | * This test uses two objects, packobj and bigobj, that are always | |
4558 | * updated together (i.e. in the same tx) so that their contents are | |
4559 | * in sync and can be compared. Their contents relate to each other | |
4560 | * in a simple way: packobj is a dense array of 'bufwad' structures, | |
4561 | * while bigobj is a sparse array of the same bufwads. Specifically, | |
4562 | * for any index n, there are three bufwads that should be identical: | |
4563 | * | |
4564 | * packobj, at offset n * sizeof (bufwad_t) | |
4565 | * bigobj, at the head of the nth chunk | |
4566 | * bigobj, at the tail of the nth chunk | |
4567 | * | |
4568 | * The chunk size is arbitrary. It doesn't have to be a power of two, | |
4569 | * and it doesn't have any relation to the object blocksize. | |
4570 | * The only requirement is that it can hold at least two bufwads. | |
4571 | * | |
4572 | * Normally, we write the bufwad to each of these locations. | |
4573 | * However, free_percent of the time we instead write zeroes to | |
4574 | * packobj and perform a dmu_free_range() on bigobj. By comparing | |
4575 | * bigobj to packobj, we can verify that the DMU is correctly | |
4576 | * tracking which parts of an object are allocated and free, | |
4577 | * and that the contents of the allocated blocks are correct. | |
4578 | */ | |
4579 | ||
4580 | /* | |
4581 | * Read the directory info. If it's the first time, set things up. | |
4582 | */ | |
50c957f7 NB |
4583 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, chunksize); |
4584 | ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, 0, | |
4585 | chunksize); | |
34dc7c2f | 4586 | |
40b84e7a BB |
4587 | if (ztest_object_init(zd, od, size, B_FALSE) != 0) { |
4588 | umem_free(od, size); | |
428870ff | 4589 | return; |
40b84e7a | 4590 | } |
34dc7c2f | 4591 | |
428870ff BB |
4592 | bigobj = od[0].od_object; |
4593 | packobj = od[1].od_object; | |
4594 | chunksize = od[0].od_gen; | |
a9eaae06 | 4595 | ASSERT3U(chunksize, ==, od[1].od_gen); |
34dc7c2f BB |
4596 | |
4597 | /* | |
4598 | * Prefetch a random chunk of the big object. | |
4599 | * Our aim here is to get some async reads in flight | |
4600 | * for blocks that we may free below; the DMU should | |
4601 | * handle this race correctly. | |
4602 | */ | |
4603 | n = ztest_random(regions) * stride + ztest_random(width); | |
4604 | s = 1 + ztest_random(2 * width - 1); | |
fcff0f35 PD |
4605 | dmu_prefetch(os, bigobj, 0, n * chunksize, s * chunksize, |
4606 | ZIO_PRIORITY_SYNC_READ); | |
34dc7c2f BB |
4607 | |
4608 | /* | |
4609 | * Pick a random index and compute the offsets into packobj and bigobj. | |
4610 | */ | |
4611 | n = ztest_random(regions) * stride + ztest_random(width); | |
4612 | s = 1 + ztest_random(width - 1); | |
4613 | ||
4614 | packoff = n * sizeof (bufwad_t); | |
4615 | packsize = s * sizeof (bufwad_t); | |
4616 | ||
428870ff BB |
4617 | bigoff = n * chunksize; |
4618 | bigsize = s * chunksize; | |
34dc7c2f BB |
4619 | |
4620 | packbuf = umem_alloc(packsize, UMEM_NOFAIL); | |
4621 | bigbuf = umem_alloc(bigsize, UMEM_NOFAIL); | |
4622 | ||
4623 | /* | |
4624 | * free_percent of the time, free a range of bigobj rather than | |
4625 | * overwriting it. | |
4626 | */ | |
4627 | freeit = (ztest_random(100) < free_percent); | |
4628 | ||
4629 | /* | |
4630 | * Read the current contents of our objects. | |
4631 | */ | |
428870ff | 4632 | error = dmu_read(os, packobj, packoff, packsize, packbuf, |
9babb374 | 4633 | DMU_READ_PREFETCH); |
c99c9001 | 4634 | ASSERT0(error); |
428870ff | 4635 | error = dmu_read(os, bigobj, bigoff, bigsize, bigbuf, |
9babb374 | 4636 | DMU_READ_PREFETCH); |
c99c9001 | 4637 | ASSERT0(error); |
34dc7c2f BB |
4638 | |
4639 | /* | |
4640 | * Get a tx for the mods to both packobj and bigobj. | |
4641 | */ | |
4642 | tx = dmu_tx_create(os); | |
4643 | ||
428870ff | 4644 | dmu_tx_hold_write(tx, packobj, packoff, packsize); |
34dc7c2f BB |
4645 | |
4646 | if (freeit) | |
428870ff | 4647 | dmu_tx_hold_free(tx, bigobj, bigoff, bigsize); |
34dc7c2f | 4648 | else |
428870ff | 4649 | dmu_tx_hold_write(tx, bigobj, bigoff, bigsize); |
34dc7c2f | 4650 | |
383fc4a9 MA |
4651 | /* This accounts for setting the checksum/compression. */ |
4652 | dmu_tx_hold_bonus(tx, bigobj); | |
4653 | ||
428870ff BB |
4654 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); |
4655 | if (txg == 0) { | |
34dc7c2f BB |
4656 | umem_free(packbuf, packsize); |
4657 | umem_free(bigbuf, bigsize); | |
40b84e7a | 4658 | umem_free(od, size); |
34dc7c2f BB |
4659 | return; |
4660 | } | |
4661 | ||
9b67f605 MA |
4662 | enum zio_checksum cksum; |
4663 | do { | |
4664 | cksum = (enum zio_checksum) | |
4665 | ztest_random_dsl_prop(ZFS_PROP_CHECKSUM); | |
4666 | } while (cksum >= ZIO_CHECKSUM_LEGACY_FUNCTIONS); | |
4667 | dmu_object_set_checksum(os, bigobj, cksum, tx); | |
428870ff | 4668 | |
9b67f605 MA |
4669 | enum zio_compress comp; |
4670 | do { | |
4671 | comp = (enum zio_compress) | |
4672 | ztest_random_dsl_prop(ZFS_PROP_COMPRESSION); | |
4673 | } while (comp >= ZIO_COMPRESS_LEGACY_FUNCTIONS); | |
4674 | dmu_object_set_compress(os, bigobj, comp, tx); | |
34dc7c2f BB |
4675 | |
4676 | /* | |
4677 | * For each index from n to n + s, verify that the existing bufwad | |
4678 | * in packobj matches the bufwads at the head and tail of the | |
4679 | * corresponding chunk in bigobj. Then update all three bufwads | |
4680 | * with the new values we want to write out. | |
4681 | */ | |
4682 | for (i = 0; i < s; i++) { | |
4683 | /* LINTED */ | |
4684 | pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t)); | |
4685 | /* LINTED */ | |
428870ff | 4686 | bigH = (bufwad_t *)((char *)bigbuf + i * chunksize); |
34dc7c2f | 4687 | /* LINTED */ |
428870ff | 4688 | bigT = (bufwad_t *)((char *)bigH + chunksize) - 1; |
34dc7c2f | 4689 | |
a9eaae06 RM |
4690 | ASSERT3U((uintptr_t)bigH - (uintptr_t)bigbuf, <, bigsize); |
4691 | ASSERT3U((uintptr_t)bigT - (uintptr_t)bigbuf, <, bigsize); | |
34dc7c2f BB |
4692 | |
4693 | if (pack->bw_txg > txg) | |
4694 | fatal(0, "future leak: got %llx, open txg is %llx", | |
4695 | pack->bw_txg, txg); | |
4696 | ||
4697 | if (pack->bw_data != 0 && pack->bw_index != n + i) | |
4698 | fatal(0, "wrong index: got %llx, wanted %llx+%llx", | |
4699 | pack->bw_index, n, i); | |
4700 | ||
4701 | if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0) | |
4702 | fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH); | |
4703 | ||
4704 | if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0) | |
4705 | fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT); | |
4706 | ||
4707 | if (freeit) { | |
4708 | bzero(pack, sizeof (bufwad_t)); | |
4709 | } else { | |
4710 | pack->bw_index = n + i; | |
4711 | pack->bw_txg = txg; | |
4712 | pack->bw_data = 1 + ztest_random(-2ULL); | |
4713 | } | |
4714 | *bigH = *pack; | |
4715 | *bigT = *pack; | |
4716 | } | |
4717 | ||
4718 | /* | |
4719 | * We've verified all the old bufwads, and made new ones. | |
4720 | * Now write them out. | |
4721 | */ | |
428870ff | 4722 | dmu_write(os, packobj, packoff, packsize, packbuf, tx); |
34dc7c2f BB |
4723 | |
4724 | if (freeit) { | |
c242c188 | 4725 | if (ztest_opts.zo_verbose >= 7) { |
34dc7c2f BB |
4726 | (void) printf("freeing offset %llx size %llx" |
4727 | " txg %llx\n", | |
4728 | (u_longlong_t)bigoff, | |
4729 | (u_longlong_t)bigsize, | |
4730 | (u_longlong_t)txg); | |
4731 | } | |
a9eaae06 | 4732 | VERIFY0(dmu_free_range(os, bigobj, bigoff, bigsize, tx)); |
34dc7c2f | 4733 | } else { |
c242c188 | 4734 | if (ztest_opts.zo_verbose >= 7) { |
34dc7c2f BB |
4735 | (void) printf("writing offset %llx size %llx" |
4736 | " txg %llx\n", | |
4737 | (u_longlong_t)bigoff, | |
4738 | (u_longlong_t)bigsize, | |
4739 | (u_longlong_t)txg); | |
4740 | } | |
428870ff | 4741 | dmu_write(os, bigobj, bigoff, bigsize, bigbuf, tx); |
34dc7c2f BB |
4742 | } |
4743 | ||
4744 | dmu_tx_commit(tx); | |
4745 | ||
4746 | /* | |
4747 | * Sanity check the stuff we just wrote. | |
4748 | */ | |
4749 | { | |
4750 | void *packcheck = umem_alloc(packsize, UMEM_NOFAIL); | |
4751 | void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL); | |
4752 | ||
a9eaae06 | 4753 | VERIFY0(dmu_read(os, packobj, packoff, |
9babb374 | 4754 | packsize, packcheck, DMU_READ_PREFETCH)); |
a9eaae06 | 4755 | VERIFY0(dmu_read(os, bigobj, bigoff, |
9babb374 | 4756 | bigsize, bigcheck, DMU_READ_PREFETCH)); |
34dc7c2f | 4757 | |
a9eaae06 RM |
4758 | ASSERT0(bcmp(packbuf, packcheck, packsize)); |
4759 | ASSERT0(bcmp(bigbuf, bigcheck, bigsize)); | |
34dc7c2f BB |
4760 | |
4761 | umem_free(packcheck, packsize); | |
4762 | umem_free(bigcheck, bigsize); | |
4763 | } | |
4764 | ||
4765 | umem_free(packbuf, packsize); | |
4766 | umem_free(bigbuf, bigsize); | |
40b84e7a | 4767 | umem_free(od, size); |
34dc7c2f BB |
4768 | } |
4769 | ||
65c7cc49 | 4770 | static void |
9babb374 | 4771 | compare_and_update_pbbufs(uint64_t s, bufwad_t *packbuf, bufwad_t *bigbuf, |
428870ff | 4772 | uint64_t bigsize, uint64_t n, uint64_t chunksize, uint64_t txg) |
9babb374 BB |
4773 | { |
4774 | uint64_t i; | |
4775 | bufwad_t *pack; | |
4776 | bufwad_t *bigH; | |
4777 | bufwad_t *bigT; | |
4778 | ||
4779 | /* | |
4780 | * For each index from n to n + s, verify that the existing bufwad | |
4781 | * in packobj matches the bufwads at the head and tail of the | |
4782 | * corresponding chunk in bigobj. Then update all three bufwads | |
4783 | * with the new values we want to write out. | |
4784 | */ | |
4785 | for (i = 0; i < s; i++) { | |
4786 | /* LINTED */ | |
4787 | pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t)); | |
4788 | /* LINTED */ | |
428870ff | 4789 | bigH = (bufwad_t *)((char *)bigbuf + i * chunksize); |
9babb374 | 4790 | /* LINTED */ |
428870ff | 4791 | bigT = (bufwad_t *)((char *)bigH + chunksize) - 1; |
9babb374 | 4792 | |
a9eaae06 RM |
4793 | ASSERT3U((uintptr_t)bigH - (uintptr_t)bigbuf, <, bigsize); |
4794 | ASSERT3U((uintptr_t)bigT - (uintptr_t)bigbuf, <, bigsize); | |
9babb374 BB |
4795 | |
4796 | if (pack->bw_txg > txg) | |
4797 | fatal(0, "future leak: got %llx, open txg is %llx", | |
4798 | pack->bw_txg, txg); | |
4799 | ||
4800 | if (pack->bw_data != 0 && pack->bw_index != n + i) | |
4801 | fatal(0, "wrong index: got %llx, wanted %llx+%llx", | |
4802 | pack->bw_index, n, i); | |
4803 | ||
4804 | if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0) | |
4805 | fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH); | |
4806 | ||
4807 | if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0) | |
4808 | fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT); | |
4809 | ||
4810 | pack->bw_index = n + i; | |
4811 | pack->bw_txg = txg; | |
4812 | pack->bw_data = 1 + ztest_random(-2ULL); | |
4813 | ||
4814 | *bigH = *pack; | |
4815 | *bigT = *pack; | |
4816 | } | |
4817 | } | |
4818 | ||
40b84e7a | 4819 | #undef OD_ARRAY_SIZE |
d1d7e268 | 4820 | #define OD_ARRAY_SIZE 2 |
40b84e7a | 4821 | |
9babb374 | 4822 | void |
428870ff | 4823 | ztest_dmu_read_write_zcopy(ztest_ds_t *zd, uint64_t id) |
9babb374 | 4824 | { |
428870ff | 4825 | objset_t *os = zd->zd_os; |
40b84e7a | 4826 | ztest_od_t *od; |
9babb374 BB |
4827 | dmu_tx_t *tx; |
4828 | uint64_t i; | |
4829 | int error; | |
40b84e7a | 4830 | int size; |
9babb374 BB |
4831 | uint64_t n, s, txg; |
4832 | bufwad_t *packbuf, *bigbuf; | |
428870ff BB |
4833 | uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize; |
4834 | uint64_t blocksize = ztest_random_blocksize(); | |
4835 | uint64_t chunksize = blocksize; | |
9babb374 BB |
4836 | uint64_t regions = 997; |
4837 | uint64_t stride = 123456789ULL; | |
4838 | uint64_t width = 9; | |
4839 | dmu_buf_t *bonus_db; | |
4840 | arc_buf_t **bigbuf_arcbufs; | |
428870ff | 4841 | dmu_object_info_t doi; |
9babb374 | 4842 | |
d1d7e268 | 4843 | size = sizeof (ztest_od_t) * OD_ARRAY_SIZE; |
40b84e7a BB |
4844 | od = umem_alloc(size, UMEM_NOFAIL); |
4845 | ||
9babb374 BB |
4846 | /* |
4847 | * This test uses two objects, packobj and bigobj, that are always | |
4848 | * updated together (i.e. in the same tx) so that their contents are | |
4849 | * in sync and can be compared. Their contents relate to each other | |
4850 | * in a simple way: packobj is a dense array of 'bufwad' structures, | |
4851 | * while bigobj is a sparse array of the same bufwads. Specifically, | |
4852 | * for any index n, there are three bufwads that should be identical: | |
4853 | * | |
4854 | * packobj, at offset n * sizeof (bufwad_t) | |
4855 | * bigobj, at the head of the nth chunk | |
4856 | * bigobj, at the tail of the nth chunk | |
4857 | * | |
4858 | * The chunk size is set equal to bigobj block size so that | |
440a3eb9 | 4859 | * dmu_assign_arcbuf_by_dbuf() can be tested for object updates. |
9babb374 BB |
4860 | */ |
4861 | ||
4862 | /* | |
4863 | * Read the directory info. If it's the first time, set things up. | |
4864 | */ | |
50c957f7 NB |
4865 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0, 0); |
4866 | ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, 0, | |
4867 | chunksize); | |
40b84e7a | 4868 | |
9babb374 | 4869 | |
40b84e7a BB |
4870 | if (ztest_object_init(zd, od, size, B_FALSE) != 0) { |
4871 | umem_free(od, size); | |
428870ff | 4872 | return; |
40b84e7a | 4873 | } |
9babb374 | 4874 | |
428870ff BB |
4875 | bigobj = od[0].od_object; |
4876 | packobj = od[1].od_object; | |
4877 | blocksize = od[0].od_blocksize; | |
4878 | chunksize = blocksize; | |
a9eaae06 | 4879 | ASSERT3U(chunksize, ==, od[1].od_gen); |
9babb374 | 4880 | |
a9eaae06 | 4881 | VERIFY0(dmu_object_info(os, bigobj, &doi)); |
428870ff | 4882 | VERIFY(ISP2(doi.doi_data_block_size)); |
a9eaae06 RM |
4883 | VERIFY3U(chunksize, ==, doi.doi_data_block_size); |
4884 | VERIFY3U(chunksize, >=, 2 * sizeof (bufwad_t)); | |
9babb374 BB |
4885 | |
4886 | /* | |
4887 | * Pick a random index and compute the offsets into packobj and bigobj. | |
4888 | */ | |
4889 | n = ztest_random(regions) * stride + ztest_random(width); | |
4890 | s = 1 + ztest_random(width - 1); | |
4891 | ||
4892 | packoff = n * sizeof (bufwad_t); | |
4893 | packsize = s * sizeof (bufwad_t); | |
4894 | ||
428870ff BB |
4895 | bigoff = n * chunksize; |
4896 | bigsize = s * chunksize; | |
9babb374 BB |
4897 | |
4898 | packbuf = umem_zalloc(packsize, UMEM_NOFAIL); | |
4899 | bigbuf = umem_zalloc(bigsize, UMEM_NOFAIL); | |
4900 | ||
a9eaae06 | 4901 | VERIFY0(dmu_bonus_hold(os, bigobj, FTAG, &bonus_db)); |
9babb374 BB |
4902 | |
4903 | bigbuf_arcbufs = umem_zalloc(2 * s * sizeof (arc_buf_t *), UMEM_NOFAIL); | |
4904 | ||
4905 | /* | |
4906 | * Iteration 0 test zcopy for DB_UNCACHED dbufs. | |
4907 | * Iteration 1 test zcopy to already referenced dbufs. | |
4908 | * Iteration 2 test zcopy to dirty dbuf in the same txg. | |
4909 | * Iteration 3 test zcopy to dbuf dirty in previous txg. | |
4910 | * Iteration 4 test zcopy when dbuf is no longer dirty. | |
4911 | * Iteration 5 test zcopy when it can't be done. | |
4912 | * Iteration 6 one more zcopy write. | |
4913 | */ | |
4914 | for (i = 0; i < 7; i++) { | |
4915 | uint64_t j; | |
4916 | uint64_t off; | |
4917 | ||
4918 | /* | |
4919 | * In iteration 5 (i == 5) use arcbufs | |
4920 | * that don't match bigobj blksz to test | |
440a3eb9 | 4921 | * dmu_assign_arcbuf_by_dbuf() when it can't directly |
9babb374 BB |
4922 | * assign an arcbuf to a dbuf. |
4923 | */ | |
4924 | for (j = 0; j < s; j++) { | |
b9541d6b | 4925 | if (i != 5 || chunksize < (SPA_MINBLOCKSIZE * 2)) { |
9babb374 | 4926 | bigbuf_arcbufs[j] = |
428870ff | 4927 | dmu_request_arcbuf(bonus_db, chunksize); |
9babb374 BB |
4928 | } else { |
4929 | bigbuf_arcbufs[2 * j] = | |
428870ff | 4930 | dmu_request_arcbuf(bonus_db, chunksize / 2); |
9babb374 | 4931 | bigbuf_arcbufs[2 * j + 1] = |
428870ff | 4932 | dmu_request_arcbuf(bonus_db, chunksize / 2); |
9babb374 BB |
4933 | } |
4934 | } | |
4935 | ||
4936 | /* | |
4937 | * Get a tx for the mods to both packobj and bigobj. | |
4938 | */ | |
4939 | tx = dmu_tx_create(os); | |
4940 | ||
428870ff BB |
4941 | dmu_tx_hold_write(tx, packobj, packoff, packsize); |
4942 | dmu_tx_hold_write(tx, bigobj, bigoff, bigsize); | |
9babb374 | 4943 | |
428870ff BB |
4944 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); |
4945 | if (txg == 0) { | |
9babb374 BB |
4946 | umem_free(packbuf, packsize); |
4947 | umem_free(bigbuf, bigsize); | |
4948 | for (j = 0; j < s; j++) { | |
b9541d6b CW |
4949 | if (i != 5 || |
4950 | chunksize < (SPA_MINBLOCKSIZE * 2)) { | |
9babb374 BB |
4951 | dmu_return_arcbuf(bigbuf_arcbufs[j]); |
4952 | } else { | |
4953 | dmu_return_arcbuf( | |
4954 | bigbuf_arcbufs[2 * j]); | |
4955 | dmu_return_arcbuf( | |
4956 | bigbuf_arcbufs[2 * j + 1]); | |
4957 | } | |
4958 | } | |
4959 | umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *)); | |
40b84e7a | 4960 | umem_free(od, size); |
9babb374 BB |
4961 | dmu_buf_rele(bonus_db, FTAG); |
4962 | return; | |
4963 | } | |
4964 | ||
9babb374 BB |
4965 | /* |
4966 | * 50% of the time don't read objects in the 1st iteration to | |
440a3eb9 TC |
4967 | * test dmu_assign_arcbuf_by_dbuf() for the case when there are |
4968 | * no existing dbufs for the specified offsets. | |
9babb374 BB |
4969 | */ |
4970 | if (i != 0 || ztest_random(2) != 0) { | |
428870ff | 4971 | error = dmu_read(os, packobj, packoff, |
9babb374 | 4972 | packsize, packbuf, DMU_READ_PREFETCH); |
c99c9001 | 4973 | ASSERT0(error); |
428870ff | 4974 | error = dmu_read(os, bigobj, bigoff, bigsize, |
9babb374 | 4975 | bigbuf, DMU_READ_PREFETCH); |
c99c9001 | 4976 | ASSERT0(error); |
9babb374 BB |
4977 | } |
4978 | compare_and_update_pbbufs(s, packbuf, bigbuf, bigsize, | |
428870ff | 4979 | n, chunksize, txg); |
9babb374 BB |
4980 | |
4981 | /* | |
4982 | * We've verified all the old bufwads, and made new ones. | |
4983 | * Now write them out. | |
4984 | */ | |
428870ff | 4985 | dmu_write(os, packobj, packoff, packsize, packbuf, tx); |
c242c188 | 4986 | if (ztest_opts.zo_verbose >= 7) { |
9babb374 BB |
4987 | (void) printf("writing offset %llx size %llx" |
4988 | " txg %llx\n", | |
4989 | (u_longlong_t)bigoff, | |
4990 | (u_longlong_t)bigsize, | |
4991 | (u_longlong_t)txg); | |
4992 | } | |
428870ff | 4993 | for (off = bigoff, j = 0; j < s; j++, off += chunksize) { |
9babb374 | 4994 | dmu_buf_t *dbt; |
b9541d6b | 4995 | if (i != 5 || chunksize < (SPA_MINBLOCKSIZE * 2)) { |
9babb374 | 4996 | bcopy((caddr_t)bigbuf + (off - bigoff), |
428870ff | 4997 | bigbuf_arcbufs[j]->b_data, chunksize); |
9babb374 BB |
4998 | } else { |
4999 | bcopy((caddr_t)bigbuf + (off - bigoff), | |
5000 | bigbuf_arcbufs[2 * j]->b_data, | |
428870ff | 5001 | chunksize / 2); |
9babb374 | 5002 | bcopy((caddr_t)bigbuf + (off - bigoff) + |
428870ff | 5003 | chunksize / 2, |
9babb374 | 5004 | bigbuf_arcbufs[2 * j + 1]->b_data, |
428870ff | 5005 | chunksize / 2); |
9babb374 BB |
5006 | } |
5007 | ||
5008 | if (i == 1) { | |
428870ff BB |
5009 | VERIFY(dmu_buf_hold(os, bigobj, off, |
5010 | FTAG, &dbt, DMU_READ_NO_PREFETCH) == 0); | |
9babb374 | 5011 | } |
b9541d6b | 5012 | if (i != 5 || chunksize < (SPA_MINBLOCKSIZE * 2)) { |
305781da TC |
5013 | VERIFY0(dmu_assign_arcbuf_by_dbuf(bonus_db, |
5014 | off, bigbuf_arcbufs[j], tx)); | |
9babb374 | 5015 | } else { |
305781da TC |
5016 | VERIFY0(dmu_assign_arcbuf_by_dbuf(bonus_db, |
5017 | off, bigbuf_arcbufs[2 * j], tx)); | |
5018 | VERIFY0(dmu_assign_arcbuf_by_dbuf(bonus_db, | |
428870ff | 5019 | off + chunksize / 2, |
305781da | 5020 | bigbuf_arcbufs[2 * j + 1], tx)); |
9babb374 BB |
5021 | } |
5022 | if (i == 1) { | |
5023 | dmu_buf_rele(dbt, FTAG); | |
5024 | } | |
5025 | } | |
5026 | dmu_tx_commit(tx); | |
5027 | ||
5028 | /* | |
5029 | * Sanity check the stuff we just wrote. | |
5030 | */ | |
5031 | { | |
5032 | void *packcheck = umem_alloc(packsize, UMEM_NOFAIL); | |
5033 | void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL); | |
5034 | ||
b2255edc | 5035 | VERIFY0(dmu_read(os, packobj, packoff, |
9babb374 | 5036 | packsize, packcheck, DMU_READ_PREFETCH)); |
b2255edc | 5037 | VERIFY0(dmu_read(os, bigobj, bigoff, |
9babb374 BB |
5038 | bigsize, bigcheck, DMU_READ_PREFETCH)); |
5039 | ||
b2255edc BB |
5040 | ASSERT0(bcmp(packbuf, packcheck, packsize)); |
5041 | ASSERT0(bcmp(bigbuf, bigcheck, bigsize)); | |
9babb374 BB |
5042 | |
5043 | umem_free(packcheck, packsize); | |
5044 | umem_free(bigcheck, bigsize); | |
5045 | } | |
5046 | if (i == 2) { | |
1b939560 | 5047 | txg_wait_open(dmu_objset_pool(os), 0, B_TRUE); |
9babb374 BB |
5048 | } else if (i == 3) { |
5049 | txg_wait_synced(dmu_objset_pool(os), 0); | |
5050 | } | |
5051 | } | |
5052 | ||
5053 | dmu_buf_rele(bonus_db, FTAG); | |
5054 | umem_free(packbuf, packsize); | |
5055 | umem_free(bigbuf, bigsize); | |
5056 | umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *)); | |
40b84e7a | 5057 | umem_free(od, size); |
9babb374 BB |
5058 | } |
5059 | ||
428870ff | 5060 | /* ARGSUSED */ |
34dc7c2f | 5061 | void |
428870ff | 5062 | ztest_dmu_write_parallel(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 5063 | { |
40b84e7a BB |
5064 | ztest_od_t *od; |
5065 | ||
d1d7e268 | 5066 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
428870ff BB |
5067 | uint64_t offset = (1ULL << (ztest_random(20) + 43)) + |
5068 | (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
34dc7c2f BB |
5069 | |
5070 | /* | |
428870ff BB |
5071 | * Have multiple threads write to large offsets in an object |
5072 | * to verify that parallel writes to an object -- even to the | |
5073 | * same blocks within the object -- doesn't cause any trouble. | |
34dc7c2f | 5074 | */ |
50c957f7 | 5075 | ztest_od_init(od, ID_PARALLEL, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0); |
9babb374 | 5076 | |
40b84e7a | 5077 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) |
34dc7c2f | 5078 | return; |
34dc7c2f | 5079 | |
428870ff | 5080 | while (ztest_random(10) != 0) |
40b84e7a BB |
5081 | ztest_io(zd, od->od_object, offset); |
5082 | ||
d1d7e268 | 5083 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 5084 | } |
34dc7c2f | 5085 | |
428870ff BB |
5086 | void |
5087 | ztest_dmu_prealloc(ztest_ds_t *zd, uint64_t id) | |
5088 | { | |
40b84e7a | 5089 | ztest_od_t *od; |
428870ff BB |
5090 | uint64_t offset = (1ULL << (ztest_random(4) + SPA_MAXBLOCKSHIFT)) + |
5091 | (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
5092 | uint64_t count = ztest_random(20) + 1; | |
5093 | uint64_t blocksize = ztest_random_blocksize(); | |
5094 | void *data; | |
34dc7c2f | 5095 | |
d1d7e268 | 5096 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
34dc7c2f | 5097 | |
50c957f7 | 5098 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0, 0); |
40b84e7a | 5099 | |
d1d7e268 MK |
5100 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
5101 | !ztest_random(2)) != 0) { | |
5102 | umem_free(od, sizeof (ztest_od_t)); | |
34dc7c2f | 5103 | return; |
40b84e7a | 5104 | } |
34dc7c2f | 5105 | |
40b84e7a | 5106 | if (ztest_truncate(zd, od->od_object, offset, count * blocksize) != 0) { |
d1d7e268 | 5107 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f | 5108 | return; |
40b84e7a | 5109 | } |
34dc7c2f | 5110 | |
40b84e7a | 5111 | ztest_prealloc(zd, od->od_object, offset, count * blocksize); |
34dc7c2f | 5112 | |
428870ff | 5113 | data = umem_zalloc(blocksize, UMEM_NOFAIL); |
34dc7c2f | 5114 | |
428870ff BB |
5115 | while (ztest_random(count) != 0) { |
5116 | uint64_t randoff = offset + (ztest_random(count) * blocksize); | |
40b84e7a | 5117 | if (ztest_write(zd, od->od_object, randoff, blocksize, |
428870ff BB |
5118 | data) != 0) |
5119 | break; | |
5120 | while (ztest_random(4) != 0) | |
40b84e7a | 5121 | ztest_io(zd, od->od_object, randoff); |
9babb374 | 5122 | } |
34dc7c2f | 5123 | |
428870ff | 5124 | umem_free(data, blocksize); |
d1d7e268 | 5125 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f BB |
5126 | } |
5127 | ||
5128 | /* | |
5129 | * Verify that zap_{create,destroy,add,remove,update} work as expected. | |
5130 | */ | |
5131 | #define ZTEST_ZAP_MIN_INTS 1 | |
5132 | #define ZTEST_ZAP_MAX_INTS 4 | |
5133 | #define ZTEST_ZAP_MAX_PROPS 1000 | |
5134 | ||
5135 | void | |
428870ff | 5136 | ztest_zap(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 5137 | { |
428870ff | 5138 | objset_t *os = zd->zd_os; |
40b84e7a | 5139 | ztest_od_t *od; |
34dc7c2f BB |
5140 | uint64_t object; |
5141 | uint64_t txg, last_txg; | |
5142 | uint64_t value[ZTEST_ZAP_MAX_INTS]; | |
5143 | uint64_t zl_ints, zl_intsize, prop; | |
5144 | int i, ints; | |
5145 | dmu_tx_t *tx; | |
5146 | char propname[100], txgname[100]; | |
5147 | int error; | |
34dc7c2f BB |
5148 | char *hc[2] = { "s.acl.h", ".s.open.h.hyLZlg" }; |
5149 | ||
d1d7e268 | 5150 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 5151 | ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0, 0); |
34dc7c2f | 5152 | |
40b84e7a | 5153 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
02730c33 | 5154 | !ztest_random(2)) != 0) |
40b84e7a | 5155 | goto out; |
34dc7c2f | 5156 | |
40b84e7a | 5157 | object = od->od_object; |
34dc7c2f | 5158 | |
428870ff BB |
5159 | /* |
5160 | * Generate a known hash collision, and verify that | |
5161 | * we can lookup and remove both entries. | |
5162 | */ | |
5163 | tx = dmu_tx_create(os); | |
5164 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); | |
5165 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
5166 | if (txg == 0) | |
40b84e7a | 5167 | goto out; |
428870ff BB |
5168 | for (i = 0; i < 2; i++) { |
5169 | value[i] = i; | |
a9eaae06 | 5170 | VERIFY0(zap_add(os, object, hc[i], sizeof (uint64_t), |
428870ff BB |
5171 | 1, &value[i], tx)); |
5172 | } | |
5173 | for (i = 0; i < 2; i++) { | |
5174 | VERIFY3U(EEXIST, ==, zap_add(os, object, hc[i], | |
5175 | sizeof (uint64_t), 1, &value[i], tx)); | |
a9eaae06 | 5176 | VERIFY0( |
428870ff BB |
5177 | zap_length(os, object, hc[i], &zl_intsize, &zl_ints)); |
5178 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
5179 | ASSERT3U(zl_ints, ==, 1); | |
5180 | } | |
5181 | for (i = 0; i < 2; i++) { | |
a9eaae06 | 5182 | VERIFY0(zap_remove(os, object, hc[i], tx)); |
34dc7c2f | 5183 | } |
428870ff | 5184 | dmu_tx_commit(tx); |
34dc7c2f | 5185 | |
428870ff | 5186 | /* |
d611989f | 5187 | * Generate a bunch of random entries. |
428870ff | 5188 | */ |
34dc7c2f BB |
5189 | ints = MAX(ZTEST_ZAP_MIN_INTS, object % ZTEST_ZAP_MAX_INTS); |
5190 | ||
5191 | prop = ztest_random(ZTEST_ZAP_MAX_PROPS); | |
5192 | (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop); | |
5193 | (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop); | |
5194 | bzero(value, sizeof (value)); | |
5195 | last_txg = 0; | |
5196 | ||
5197 | /* | |
5198 | * If these zap entries already exist, validate their contents. | |
5199 | */ | |
5200 | error = zap_length(os, object, txgname, &zl_intsize, &zl_ints); | |
5201 | if (error == 0) { | |
5202 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
5203 | ASSERT3U(zl_ints, ==, 1); | |
5204 | ||
a9eaae06 RM |
5205 | VERIFY0(zap_lookup(os, object, txgname, zl_intsize, |
5206 | zl_ints, &last_txg)); | |
34dc7c2f | 5207 | |
a9eaae06 RM |
5208 | VERIFY0(zap_length(os, object, propname, &zl_intsize, |
5209 | &zl_ints)); | |
34dc7c2f BB |
5210 | |
5211 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
5212 | ASSERT3U(zl_ints, ==, ints); | |
5213 | ||
a9eaae06 RM |
5214 | VERIFY0(zap_lookup(os, object, propname, zl_intsize, |
5215 | zl_ints, value)); | |
34dc7c2f BB |
5216 | |
5217 | for (i = 0; i < ints; i++) { | |
5218 | ASSERT3U(value[i], ==, last_txg + object + i); | |
5219 | } | |
5220 | } else { | |
5221 | ASSERT3U(error, ==, ENOENT); | |
5222 | } | |
5223 | ||
5224 | /* | |
5225 | * Atomically update two entries in our zap object. | |
5226 | * The first is named txg_%llu, and contains the txg | |
5227 | * in which the property was last updated. The second | |
5228 | * is named prop_%llu, and the nth element of its value | |
5229 | * should be txg + object + n. | |
5230 | */ | |
5231 | tx = dmu_tx_create(os); | |
428870ff BB |
5232 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
5233 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
5234 | if (txg == 0) | |
40b84e7a | 5235 | goto out; |
34dc7c2f BB |
5236 | |
5237 | if (last_txg > txg) | |
5238 | fatal(0, "zap future leak: old %llu new %llu", last_txg, txg); | |
5239 | ||
5240 | for (i = 0; i < ints; i++) | |
5241 | value[i] = txg + object + i; | |
5242 | ||
a9eaae06 | 5243 | VERIFY0(zap_update(os, object, txgname, sizeof (uint64_t), |
428870ff | 5244 | 1, &txg, tx)); |
a9eaae06 | 5245 | VERIFY0(zap_update(os, object, propname, sizeof (uint64_t), |
428870ff | 5246 | ints, value, tx)); |
34dc7c2f BB |
5247 | |
5248 | dmu_tx_commit(tx); | |
5249 | ||
5250 | /* | |
5251 | * Remove a random pair of entries. | |
5252 | */ | |
5253 | prop = ztest_random(ZTEST_ZAP_MAX_PROPS); | |
5254 | (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop); | |
5255 | (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop); | |
5256 | ||
5257 | error = zap_length(os, object, txgname, &zl_intsize, &zl_ints); | |
5258 | ||
5259 | if (error == ENOENT) | |
40b84e7a | 5260 | goto out; |
34dc7c2f | 5261 | |
c99c9001 | 5262 | ASSERT0(error); |
34dc7c2f BB |
5263 | |
5264 | tx = dmu_tx_create(os); | |
428870ff BB |
5265 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
5266 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
5267 | if (txg == 0) | |
40b84e7a | 5268 | goto out; |
a9eaae06 RM |
5269 | VERIFY0(zap_remove(os, object, txgname, tx)); |
5270 | VERIFY0(zap_remove(os, object, propname, tx)); | |
428870ff | 5271 | dmu_tx_commit(tx); |
40b84e7a | 5272 | out: |
d1d7e268 | 5273 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 5274 | } |
34dc7c2f | 5275 | |
428870ff | 5276 | /* |
d611989f | 5277 | * Test case to test the upgrading of a microzap to fatzap. |
428870ff BB |
5278 | */ |
5279 | void | |
5280 | ztest_fzap(ztest_ds_t *zd, uint64_t id) | |
5281 | { | |
5282 | objset_t *os = zd->zd_os; | |
40b84e7a | 5283 | ztest_od_t *od; |
428870ff | 5284 | uint64_t object, txg; |
d6320ddb | 5285 | int i; |
34dc7c2f | 5286 | |
d1d7e268 | 5287 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 5288 | ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0, 0); |
428870ff | 5289 | |
40b84e7a | 5290 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
02730c33 | 5291 | !ztest_random(2)) != 0) |
40b84e7a BB |
5292 | goto out; |
5293 | object = od->od_object; | |
34dc7c2f BB |
5294 | |
5295 | /* | |
428870ff BB |
5296 | * Add entries to this ZAP and make sure it spills over |
5297 | * and gets upgraded to a fatzap. Also, since we are adding | |
5298 | * 2050 entries we should see ptrtbl growth and leaf-block split. | |
34dc7c2f | 5299 | */ |
d6320ddb | 5300 | for (i = 0; i < 2050; i++) { |
eca7b760 | 5301 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
5302 | uint64_t value = i; |
5303 | dmu_tx_t *tx; | |
5304 | int error; | |
34dc7c2f | 5305 | |
428870ff | 5306 | (void) snprintf(name, sizeof (name), "fzap-%llu-%llu", |
b8864a23 | 5307 | (u_longlong_t)id, (u_longlong_t)value); |
428870ff BB |
5308 | |
5309 | tx = dmu_tx_create(os); | |
5310 | dmu_tx_hold_zap(tx, object, B_TRUE, name); | |
5311 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
5312 | if (txg == 0) | |
40b84e7a | 5313 | goto out; |
428870ff BB |
5314 | error = zap_add(os, object, name, sizeof (uint64_t), 1, |
5315 | &value, tx); | |
5316 | ASSERT(error == 0 || error == EEXIST); | |
5317 | dmu_tx_commit(tx); | |
34dc7c2f | 5318 | } |
40b84e7a | 5319 | out: |
d1d7e268 | 5320 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f BB |
5321 | } |
5322 | ||
428870ff | 5323 | /* ARGSUSED */ |
34dc7c2f | 5324 | void |
428870ff | 5325 | ztest_zap_parallel(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 5326 | { |
428870ff | 5327 | objset_t *os = zd->zd_os; |
40b84e7a | 5328 | ztest_od_t *od; |
34dc7c2f BB |
5329 | uint64_t txg, object, count, wsize, wc, zl_wsize, zl_wc; |
5330 | dmu_tx_t *tx; | |
5331 | int i, namelen, error; | |
428870ff | 5332 | int micro = ztest_random(2); |
34dc7c2f BB |
5333 | char name[20], string_value[20]; |
5334 | void *data; | |
5335 | ||
d1d7e268 | 5336 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 5337 | ztest_od_init(od, ID_PARALLEL, FTAG, micro, DMU_OT_ZAP_OTHER, 0, 0, 0); |
428870ff | 5338 | |
40b84e7a | 5339 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
d1d7e268 | 5340 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 5341 | return; |
40b84e7a | 5342 | } |
428870ff | 5343 | |
40b84e7a | 5344 | object = od->od_object; |
428870ff | 5345 | |
34dc7c2f BB |
5346 | /* |
5347 | * Generate a random name of the form 'xxx.....' where each | |
5348 | * x is a random printable character and the dots are dots. | |
5349 | * There are 94 such characters, and the name length goes from | |
5350 | * 6 to 20, so there are 94^3 * 15 = 12,458,760 possible names. | |
5351 | */ | |
5352 | namelen = ztest_random(sizeof (name) - 5) + 5 + 1; | |
5353 | ||
5354 | for (i = 0; i < 3; i++) | |
5355 | name[i] = '!' + ztest_random('~' - '!' + 1); | |
5356 | for (; i < namelen - 1; i++) | |
5357 | name[i] = '.'; | |
5358 | name[i] = '\0'; | |
5359 | ||
428870ff | 5360 | if ((namelen & 1) || micro) { |
34dc7c2f BB |
5361 | wsize = sizeof (txg); |
5362 | wc = 1; | |
5363 | data = &txg; | |
5364 | } else { | |
5365 | wsize = 1; | |
5366 | wc = namelen; | |
5367 | data = string_value; | |
5368 | } | |
5369 | ||
5370 | count = -1ULL; | |
13fe0198 | 5371 | VERIFY0(zap_count(os, object, &count)); |
a9eaae06 | 5372 | ASSERT3S(count, !=, -1ULL); |
34dc7c2f BB |
5373 | |
5374 | /* | |
5375 | * Select an operation: length, lookup, add, update, remove. | |
5376 | */ | |
5377 | i = ztest_random(5); | |
5378 | ||
5379 | if (i >= 2) { | |
5380 | tx = dmu_tx_create(os); | |
428870ff BB |
5381 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
5382 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
df053d67 GN |
5383 | if (txg == 0) { |
5384 | umem_free(od, sizeof (ztest_od_t)); | |
34dc7c2f | 5385 | return; |
df053d67 | 5386 | } |
34dc7c2f BB |
5387 | bcopy(name, string_value, namelen); |
5388 | } else { | |
5389 | tx = NULL; | |
5390 | txg = 0; | |
5391 | bzero(string_value, namelen); | |
5392 | } | |
5393 | ||
5394 | switch (i) { | |
5395 | ||
5396 | case 0: | |
5397 | error = zap_length(os, object, name, &zl_wsize, &zl_wc); | |
5398 | if (error == 0) { | |
5399 | ASSERT3U(wsize, ==, zl_wsize); | |
5400 | ASSERT3U(wc, ==, zl_wc); | |
5401 | } else { | |
5402 | ASSERT3U(error, ==, ENOENT); | |
5403 | } | |
5404 | break; | |
5405 | ||
5406 | case 1: | |
5407 | error = zap_lookup(os, object, name, wsize, wc, data); | |
5408 | if (error == 0) { | |
5409 | if (data == string_value && | |
5410 | bcmp(name, data, namelen) != 0) | |
5411 | fatal(0, "name '%s' != val '%s' len %d", | |
5412 | name, data, namelen); | |
5413 | } else { | |
5414 | ASSERT3U(error, ==, ENOENT); | |
5415 | } | |
5416 | break; | |
5417 | ||
5418 | case 2: | |
5419 | error = zap_add(os, object, name, wsize, wc, data, tx); | |
5420 | ASSERT(error == 0 || error == EEXIST); | |
5421 | break; | |
5422 | ||
5423 | case 3: | |
a9eaae06 | 5424 | VERIFY0(zap_update(os, object, name, wsize, wc, data, tx)); |
34dc7c2f BB |
5425 | break; |
5426 | ||
5427 | case 4: | |
5428 | error = zap_remove(os, object, name, tx); | |
5429 | ASSERT(error == 0 || error == ENOENT); | |
5430 | break; | |
5431 | } | |
5432 | ||
5433 | if (tx != NULL) | |
5434 | dmu_tx_commit(tx); | |
40b84e7a | 5435 | |
d1d7e268 | 5436 | umem_free(od, sizeof (ztest_od_t)); |
34dc7c2f BB |
5437 | } |
5438 | ||
428870ff BB |
5439 | /* |
5440 | * Commit callback data. | |
5441 | */ | |
5442 | typedef struct ztest_cb_data { | |
5443 | list_node_t zcd_node; | |
5444 | uint64_t zcd_txg; | |
5445 | int zcd_expected_err; | |
5446 | boolean_t zcd_added; | |
5447 | boolean_t zcd_called; | |
5448 | spa_t *zcd_spa; | |
5449 | } ztest_cb_data_t; | |
5450 | ||
5451 | /* This is the actual commit callback function */ | |
5452 | static void | |
5453 | ztest_commit_callback(void *arg, int error) | |
5454 | { | |
5455 | ztest_cb_data_t *data = arg; | |
5456 | uint64_t synced_txg; | |
5457 | ||
a9eaae06 | 5458 | VERIFY3P(data, !=, NULL); |
428870ff BB |
5459 | VERIFY3S(data->zcd_expected_err, ==, error); |
5460 | VERIFY(!data->zcd_called); | |
5461 | ||
5462 | synced_txg = spa_last_synced_txg(data->zcd_spa); | |
5463 | if (data->zcd_txg > synced_txg) | |
5464 | fatal(0, "commit callback of txg %" PRIu64 " called prematurely" | |
5465 | ", last synced txg = %" PRIu64 "\n", data->zcd_txg, | |
5466 | synced_txg); | |
5467 | ||
5468 | data->zcd_called = B_TRUE; | |
5469 | ||
5470 | if (error == ECANCELED) { | |
c99c9001 | 5471 | ASSERT0(data->zcd_txg); |
428870ff BB |
5472 | ASSERT(!data->zcd_added); |
5473 | ||
5474 | /* | |
5475 | * The private callback data should be destroyed here, but | |
5476 | * since we are going to check the zcd_called field after | |
5477 | * dmu_tx_abort(), we will destroy it there. | |
5478 | */ | |
5479 | return; | |
5480 | } | |
5481 | ||
090ff092 | 5482 | ASSERT(data->zcd_added); |
428870ff BB |
5483 | ASSERT3U(data->zcd_txg, !=, 0); |
5484 | ||
1e33ac1e | 5485 | (void) mutex_enter(&zcl.zcl_callbacks_lock); |
090ff092 RC |
5486 | |
5487 | /* See if this cb was called more quickly */ | |
5488 | if ((synced_txg - data->zcd_txg) < zc_min_txg_delay) | |
5489 | zc_min_txg_delay = synced_txg - data->zcd_txg; | |
5490 | ||
5491 | /* Remove our callback from the list */ | |
428870ff | 5492 | list_remove(&zcl.zcl_callbacks, data); |
090ff092 | 5493 | |
1e33ac1e | 5494 | (void) mutex_exit(&zcl.zcl_callbacks_lock); |
428870ff | 5495 | |
428870ff BB |
5496 | umem_free(data, sizeof (ztest_cb_data_t)); |
5497 | } | |
5498 | ||
5499 | /* Allocate and initialize callback data structure */ | |
5500 | static ztest_cb_data_t * | |
5501 | ztest_create_cb_data(objset_t *os, uint64_t txg) | |
5502 | { | |
5503 | ztest_cb_data_t *cb_data; | |
5504 | ||
5505 | cb_data = umem_zalloc(sizeof (ztest_cb_data_t), UMEM_NOFAIL); | |
5506 | ||
5507 | cb_data->zcd_txg = txg; | |
5508 | cb_data->zcd_spa = dmu_objset_spa(os); | |
1e33ac1e | 5509 | list_link_init(&cb_data->zcd_node); |
428870ff BB |
5510 | |
5511 | return (cb_data); | |
5512 | } | |
5513 | ||
428870ff BB |
5514 | /* |
5515 | * Commit callback test. | |
5516 | */ | |
34dc7c2f | 5517 | void |
428870ff BB |
5518 | ztest_dmu_commit_callbacks(ztest_ds_t *zd, uint64_t id) |
5519 | { | |
5520 | objset_t *os = zd->zd_os; | |
40b84e7a | 5521 | ztest_od_t *od; |
428870ff BB |
5522 | dmu_tx_t *tx; |
5523 | ztest_cb_data_t *cb_data[3], *tmp_cb; | |
5524 | uint64_t old_txg, txg; | |
090ff092 | 5525 | int i, error = 0; |
428870ff | 5526 | |
d1d7e268 | 5527 | od = umem_alloc(sizeof (ztest_od_t), UMEM_NOFAIL); |
50c957f7 | 5528 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0); |
428870ff | 5529 | |
40b84e7a | 5530 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
d1d7e268 | 5531 | umem_free(od, sizeof (ztest_od_t)); |
428870ff | 5532 | return; |
40b84e7a | 5533 | } |
428870ff BB |
5534 | |
5535 | tx = dmu_tx_create(os); | |
5536 | ||
5537 | cb_data[0] = ztest_create_cb_data(os, 0); | |
5538 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[0]); | |
5539 | ||
40b84e7a | 5540 | dmu_tx_hold_write(tx, od->od_object, 0, sizeof (uint64_t)); |
428870ff BB |
5541 | |
5542 | /* Every once in a while, abort the transaction on purpose */ | |
5543 | if (ztest_random(100) == 0) | |
5544 | error = -1; | |
5545 | ||
5546 | if (!error) | |
5547 | error = dmu_tx_assign(tx, TXG_NOWAIT); | |
5548 | ||
5549 | txg = error ? 0 : dmu_tx_get_txg(tx); | |
5550 | ||
5551 | cb_data[0]->zcd_txg = txg; | |
5552 | cb_data[1] = ztest_create_cb_data(os, txg); | |
5553 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[1]); | |
5554 | ||
5555 | if (error) { | |
5556 | /* | |
5557 | * It's not a strict requirement to call the registered | |
5558 | * callbacks from inside dmu_tx_abort(), but that's what | |
5559 | * it's supposed to happen in the current implementation | |
5560 | * so we will check for that. | |
5561 | */ | |
5562 | for (i = 0; i < 2; i++) { | |
5563 | cb_data[i]->zcd_expected_err = ECANCELED; | |
5564 | VERIFY(!cb_data[i]->zcd_called); | |
5565 | } | |
5566 | ||
5567 | dmu_tx_abort(tx); | |
5568 | ||
5569 | for (i = 0; i < 2; i++) { | |
5570 | VERIFY(cb_data[i]->zcd_called); | |
5571 | umem_free(cb_data[i], sizeof (ztest_cb_data_t)); | |
5572 | } | |
5573 | ||
d1d7e268 | 5574 | umem_free(od, sizeof (ztest_od_t)); |
428870ff BB |
5575 | return; |
5576 | } | |
5577 | ||
5578 | cb_data[2] = ztest_create_cb_data(os, txg); | |
5579 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[2]); | |
5580 | ||
5581 | /* | |
5582 | * Read existing data to make sure there isn't a future leak. | |
5583 | */ | |
a9eaae06 | 5584 | VERIFY0(dmu_read(os, od->od_object, 0, sizeof (uint64_t), |
428870ff BB |
5585 | &old_txg, DMU_READ_PREFETCH)); |
5586 | ||
5587 | if (old_txg > txg) | |
5588 | fatal(0, "future leak: got %" PRIu64 ", open txg is %" PRIu64, | |
5589 | old_txg, txg); | |
5590 | ||
40b84e7a | 5591 | dmu_write(os, od->od_object, 0, sizeof (uint64_t), &txg, tx); |
428870ff | 5592 | |
1e33ac1e | 5593 | (void) mutex_enter(&zcl.zcl_callbacks_lock); |
428870ff BB |
5594 | |
5595 | /* | |
5596 | * Since commit callbacks don't have any ordering requirement and since | |
5597 | * it is theoretically possible for a commit callback to be called | |
5598 | * after an arbitrary amount of time has elapsed since its txg has been | |
5599 | * synced, it is difficult to reliably determine whether a commit | |
5600 | * callback hasn't been called due to high load or due to a flawed | |
5601 | * implementation. | |
5602 | * | |
5603 | * In practice, we will assume that if after a certain number of txgs a | |
5604 | * commit callback hasn't been called, then most likely there's an | |
5605 | * implementation bug.. | |
5606 | */ | |
5607 | tmp_cb = list_head(&zcl.zcl_callbacks); | |
5608 | if (tmp_cb != NULL && | |
090ff092 | 5609 | tmp_cb->zcd_txg + ZTEST_COMMIT_CB_THRESH < txg) { |
428870ff BB |
5610 | fatal(0, "Commit callback threshold exceeded, oldest txg: %" |
5611 | PRIu64 ", open txg: %" PRIu64 "\n", tmp_cb->zcd_txg, txg); | |
5612 | } | |
5613 | ||
5614 | /* | |
5615 | * Let's find the place to insert our callbacks. | |
5616 | * | |
5617 | * Even though the list is ordered by txg, it is possible for the | |
5618 | * insertion point to not be the end because our txg may already be | |
5619 | * quiescing at this point and other callbacks in the open txg | |
5620 | * (from other objsets) may have sneaked in. | |
5621 | */ | |
5622 | tmp_cb = list_tail(&zcl.zcl_callbacks); | |
5623 | while (tmp_cb != NULL && tmp_cb->zcd_txg > txg) | |
5624 | tmp_cb = list_prev(&zcl.zcl_callbacks, tmp_cb); | |
5625 | ||
5626 | /* Add the 3 callbacks to the list */ | |
5627 | for (i = 0; i < 3; i++) { | |
5628 | if (tmp_cb == NULL) | |
5629 | list_insert_head(&zcl.zcl_callbacks, cb_data[i]); | |
5630 | else | |
5631 | list_insert_after(&zcl.zcl_callbacks, tmp_cb, | |
5632 | cb_data[i]); | |
5633 | ||
5634 | cb_data[i]->zcd_added = B_TRUE; | |
5635 | VERIFY(!cb_data[i]->zcd_called); | |
5636 | ||
5637 | tmp_cb = cb_data[i]; | |
5638 | } | |
5639 | ||
090ff092 RC |
5640 | zc_cb_counter += 3; |
5641 | ||
1e33ac1e | 5642 | (void) mutex_exit(&zcl.zcl_callbacks_lock); |
428870ff BB |
5643 | |
5644 | dmu_tx_commit(tx); | |
40b84e7a | 5645 | |
d1d7e268 | 5646 | umem_free(od, sizeof (ztest_od_t)); |
428870ff BB |
5647 | } |
5648 | ||
50c957f7 NB |
5649 | /* |
5650 | * Visit each object in the dataset. Verify that its properties | |
5651 | * are consistent what was stored in the block tag when it was created, | |
5652 | * and that its unused bonus buffer space has not been overwritten. | |
5653 | */ | |
817b1b6e | 5654 | /* ARGSUSED */ |
50c957f7 NB |
5655 | void |
5656 | ztest_verify_dnode_bt(ztest_ds_t *zd, uint64_t id) | |
5657 | { | |
5658 | objset_t *os = zd->zd_os; | |
5659 | uint64_t obj; | |
5660 | int err = 0; | |
5661 | ||
5662 | for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) { | |
5663 | ztest_block_tag_t *bt = NULL; | |
5664 | dmu_object_info_t doi; | |
5665 | dmu_buf_t *db; | |
5666 | ||
06401e42 BB |
5667 | ztest_object_lock(zd, obj, RL_READER); |
5668 | if (dmu_bonus_hold(os, obj, FTAG, &db) != 0) { | |
5669 | ztest_object_unlock(zd, obj); | |
50c957f7 | 5670 | continue; |
06401e42 | 5671 | } |
50c957f7 NB |
5672 | |
5673 | dmu_object_info_from_db(db, &doi); | |
5674 | if (doi.doi_bonus_size >= sizeof (*bt)) | |
5675 | bt = ztest_bt_bonus(db); | |
5676 | ||
5677 | if (bt && bt->bt_magic == BT_MAGIC) { | |
5678 | ztest_bt_verify(bt, os, obj, doi.doi_dnodesize, | |
5679 | bt->bt_offset, bt->bt_gen, bt->bt_txg, | |
5680 | bt->bt_crtxg); | |
5681 | ztest_verify_unused_bonus(db, bt, obj, os, bt->bt_gen); | |
5682 | } | |
5683 | ||
5684 | dmu_buf_rele(db, FTAG); | |
06401e42 | 5685 | ztest_object_unlock(zd, obj); |
50c957f7 NB |
5686 | } |
5687 | } | |
5688 | ||
428870ff BB |
5689 | /* ARGSUSED */ |
5690 | void | |
5691 | ztest_dsl_prop_get_set(ztest_ds_t *zd, uint64_t id) | |
5692 | { | |
5693 | zfs_prop_t proplist[] = { | |
5694 | ZFS_PROP_CHECKSUM, | |
5695 | ZFS_PROP_COMPRESSION, | |
5696 | ZFS_PROP_COPIES, | |
5697 | ZFS_PROP_DEDUP | |
5698 | }; | |
d6320ddb | 5699 | int p; |
428870ff | 5700 | |
2d9142c9 | 5701 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
428870ff | 5702 | |
d6320ddb | 5703 | for (p = 0; p < sizeof (proplist) / sizeof (proplist[0]); p++) |
428870ff BB |
5704 | (void) ztest_dsl_prop_set_uint64(zd->zd_name, proplist[p], |
5705 | ztest_random_dsl_prop(proplist[p]), (int)ztest_random(2)); | |
5706 | ||
76d52067 BB |
5707 | VERIFY0(ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_RECORDSIZE, |
5708 | ztest_random_blocksize(), (int)ztest_random(2))); | |
5709 | ||
2d9142c9 | 5710 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
428870ff BB |
5711 | } |
5712 | ||
5713 | /* ARGSUSED */ | |
5714 | void | |
5715 | ztest_spa_prop_get_set(ztest_ds_t *zd, uint64_t id) | |
5716 | { | |
428870ff BB |
5717 | nvlist_t *props = NULL; |
5718 | ||
2d9142c9 | 5719 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
428870ff | 5720 | |
1b939560 BB |
5721 | (void) ztest_spa_prop_set_uint64(ZPOOL_PROP_AUTOTRIM, ztest_random(2)); |
5722 | ||
c99c9001 | 5723 | VERIFY0(spa_prop_get(ztest_spa, &props)); |
428870ff | 5724 | |
c242c188 | 5725 | if (ztest_opts.zo_verbose >= 6) |
428870ff BB |
5726 | dump_nvlist(props, 4); |
5727 | ||
bea3fc71 | 5728 | fnvlist_free(props); |
428870ff | 5729 | |
2d9142c9 | 5730 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
428870ff BB |
5731 | } |
5732 | ||
13fe0198 MA |
5733 | static int |
5734 | user_release_one(const char *snapname, const char *holdname) | |
5735 | { | |
5736 | nvlist_t *snaps, *holds; | |
5737 | int error; | |
5738 | ||
5739 | snaps = fnvlist_alloc(); | |
5740 | holds = fnvlist_alloc(); | |
5741 | fnvlist_add_boolean(holds, holdname); | |
5742 | fnvlist_add_nvlist(snaps, snapname, holds); | |
5743 | fnvlist_free(holds); | |
5744 | error = dsl_dataset_user_release(snaps, NULL); | |
5745 | fnvlist_free(snaps); | |
5746 | return (error); | |
5747 | } | |
5748 | ||
428870ff BB |
5749 | /* |
5750 | * Test snapshot hold/release and deferred destroy. | |
5751 | */ | |
5752 | void | |
5753 | ztest_dmu_snapshot_hold(ztest_ds_t *zd, uint64_t id) | |
34dc7c2f | 5754 | { |
34dc7c2f | 5755 | int error; |
428870ff BB |
5756 | objset_t *os = zd->zd_os; |
5757 | objset_t *origin; | |
5758 | char snapname[100]; | |
5759 | char fullname[100]; | |
5760 | char clonename[100]; | |
5761 | char tag[100]; | |
eca7b760 | 5762 | char osname[ZFS_MAX_DATASET_NAME_LEN]; |
13fe0198 | 5763 | nvlist_t *holds; |
34dc7c2f | 5764 | |
2d9142c9 | 5765 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
34dc7c2f BB |
5766 | |
5767 | dmu_objset_name(os, osname); | |
5768 | ||
d1d7e268 MK |
5769 | (void) snprintf(snapname, sizeof (snapname), "sh1_%llu", |
5770 | (u_longlong_t)id); | |
13fe0198 MA |
5771 | (void) snprintf(fullname, sizeof (fullname), "%s@%s", osname, snapname); |
5772 | (void) snprintf(clonename, sizeof (clonename), | |
d1d7e268 MK |
5773 | "%s/ch1_%llu", osname, (u_longlong_t)id); |
5774 | (void) snprintf(tag, sizeof (tag), "tag_%llu", (u_longlong_t)id); | |
428870ff BB |
5775 | |
5776 | /* | |
5777 | * Clean up from any previous run. | |
5778 | */ | |
13fe0198 MA |
5779 | error = dsl_destroy_head(clonename); |
5780 | if (error != ENOENT) | |
5781 | ASSERT0(error); | |
5782 | error = user_release_one(fullname, tag); | |
5783 | if (error != ESRCH && error != ENOENT) | |
5784 | ASSERT0(error); | |
5785 | error = dsl_destroy_snapshot(fullname, B_FALSE); | |
5786 | if (error != ENOENT) | |
5787 | ASSERT0(error); | |
428870ff BB |
5788 | |
5789 | /* | |
5790 | * Create snapshot, clone it, mark snap for deferred destroy, | |
5791 | * destroy clone, verify snap was also destroyed. | |
5792 | */ | |
6f1ffb06 | 5793 | error = dmu_objset_snapshot_one(osname, snapname); |
428870ff BB |
5794 | if (error) { |
5795 | if (error == ENOSPC) { | |
5796 | ztest_record_enospc("dmu_objset_snapshot"); | |
5797 | goto out; | |
34dc7c2f | 5798 | } |
428870ff BB |
5799 | fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error); |
5800 | } | |
34dc7c2f | 5801 | |
13fe0198 | 5802 | error = dmu_objset_clone(clonename, fullname); |
428870ff | 5803 | if (error) { |
34dc7c2f | 5804 | if (error == ENOSPC) { |
428870ff BB |
5805 | ztest_record_enospc("dmu_objset_clone"); |
5806 | goto out; | |
34dc7c2f | 5807 | } |
428870ff BB |
5808 | fatal(0, "dmu_objset_clone(%s) = %d", clonename, error); |
5809 | } | |
34dc7c2f | 5810 | |
13fe0198 | 5811 | error = dsl_destroy_snapshot(fullname, B_TRUE); |
428870ff | 5812 | if (error) { |
13fe0198 | 5813 | fatal(0, "dsl_destroy_snapshot(%s, B_TRUE) = %d", |
428870ff BB |
5814 | fullname, error); |
5815 | } | |
34dc7c2f | 5816 | |
13fe0198 | 5817 | error = dsl_destroy_head(clonename); |
428870ff | 5818 | if (error) |
13fe0198 | 5819 | fatal(0, "dsl_destroy_head(%s) = %d", clonename, error); |
34dc7c2f | 5820 | |
428870ff BB |
5821 | error = dmu_objset_hold(fullname, FTAG, &origin); |
5822 | if (error != ENOENT) | |
5823 | fatal(0, "dmu_objset_hold(%s) = %d", fullname, error); | |
34dc7c2f | 5824 | |
428870ff BB |
5825 | /* |
5826 | * Create snapshot, add temporary hold, verify that we can't | |
5827 | * destroy a held snapshot, mark for deferred destroy, | |
5828 | * release hold, verify snapshot was destroyed. | |
5829 | */ | |
6f1ffb06 | 5830 | error = dmu_objset_snapshot_one(osname, snapname); |
428870ff BB |
5831 | if (error) { |
5832 | if (error == ENOSPC) { | |
5833 | ztest_record_enospc("dmu_objset_snapshot"); | |
5834 | goto out; | |
34dc7c2f | 5835 | } |
428870ff BB |
5836 | fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error); |
5837 | } | |
5838 | ||
13fe0198 MA |
5839 | holds = fnvlist_alloc(); |
5840 | fnvlist_add_string(holds, fullname, tag); | |
5841 | error = dsl_dataset_user_hold(holds, 0, NULL); | |
5842 | fnvlist_free(holds); | |
5843 | ||
9b67f605 MA |
5844 | if (error == ENOSPC) { |
5845 | ztest_record_enospc("dsl_dataset_user_hold"); | |
5846 | goto out; | |
5847 | } else if (error) { | |
5848 | fatal(0, "dsl_dataset_user_hold(%s, %s) = %u", | |
5849 | fullname, tag, error); | |
5850 | } | |
428870ff | 5851 | |
13fe0198 | 5852 | error = dsl_destroy_snapshot(fullname, B_FALSE); |
428870ff | 5853 | if (error != EBUSY) { |
13fe0198 | 5854 | fatal(0, "dsl_destroy_snapshot(%s, B_FALSE) = %d", |
428870ff BB |
5855 | fullname, error); |
5856 | } | |
5857 | ||
13fe0198 | 5858 | error = dsl_destroy_snapshot(fullname, B_TRUE); |
428870ff | 5859 | if (error) { |
13fe0198 | 5860 | fatal(0, "dsl_destroy_snapshot(%s, B_TRUE) = %d", |
428870ff | 5861 | fullname, error); |
34dc7c2f BB |
5862 | } |
5863 | ||
13fe0198 | 5864 | error = user_release_one(fullname, tag); |
428870ff | 5865 | if (error) |
95fd54a1 | 5866 | fatal(0, "user_release_one(%s, %s) = %d", fullname, tag, error); |
428870ff | 5867 | |
13fe0198 | 5868 | VERIFY3U(dmu_objset_hold(fullname, FTAG, &origin), ==, ENOENT); |
428870ff BB |
5869 | |
5870 | out: | |
2d9142c9 | 5871 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
34dc7c2f BB |
5872 | } |
5873 | ||
34dc7c2f BB |
5874 | /* |
5875 | * Inject random faults into the on-disk data. | |
5876 | */ | |
428870ff | 5877 | /* ARGSUSED */ |
34dc7c2f | 5878 | void |
428870ff | 5879 | ztest_fault_inject(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 5880 | { |
428870ff | 5881 | ztest_shared_t *zs = ztest_shared; |
c242c188 | 5882 | spa_t *spa = ztest_spa; |
34dc7c2f BB |
5883 | int fd; |
5884 | uint64_t offset; | |
428870ff | 5885 | uint64_t leaves; |
c5b3a7bb | 5886 | uint64_t bad = 0x1990c0ffeedecadeull; |
34dc7c2f | 5887 | uint64_t top, leaf; |
40b84e7a BB |
5888 | char *path0; |
5889 | char *pathrand; | |
34dc7c2f | 5890 | size_t fsize; |
2014c09f | 5891 | int bshift = SPA_MAXBLOCKSHIFT + 2; |
34dc7c2f | 5892 | int iters = 1000; |
428870ff BB |
5893 | int maxfaults; |
5894 | int mirror_save; | |
b128c09f | 5895 | vdev_t *vd0 = NULL; |
34dc7c2f | 5896 | uint64_t guid0 = 0; |
428870ff BB |
5897 | boolean_t islog = B_FALSE; |
5898 | ||
40b84e7a BB |
5899 | path0 = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
5900 | pathrand = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
5901 | ||
c242c188 | 5902 | mutex_enter(&ztest_vdev_lock); |
9e052db4 MA |
5903 | |
5904 | /* | |
5905 | * Device removal is in progress, fault injection must be disabled | |
5906 | * until it completes and the pool is scrubbed. The fault injection | |
5907 | * strategy for damaging blocks does not take in to account evacuated | |
5908 | * blocks which may have already been damaged. | |
5909 | */ | |
5910 | if (ztest_device_removal_active) { | |
5911 | mutex_exit(&ztest_vdev_lock); | |
5912 | goto out; | |
5913 | } | |
5914 | ||
60b82074 | 5915 | maxfaults = MAXFAULTS(zs); |
b2255edc | 5916 | leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raid_children; |
428870ff | 5917 | mirror_save = zs->zs_mirrors; |
c242c188 | 5918 | mutex_exit(&ztest_vdev_lock); |
34dc7c2f | 5919 | |
a9eaae06 | 5920 | ASSERT3U(leaves, >=, 1); |
34dc7c2f | 5921 | |
ce5fb2a7 BB |
5922 | /* |
5923 | * While ztest is running the number of leaves will not change. This | |
5924 | * is critical for the fault injection logic as it determines where | |
5925 | * errors can be safely injected such that they are always repairable. | |
5926 | * | |
5927 | * When restarting ztest a different number of leaves may be requested | |
5928 | * which will shift the regions to be damaged. This is fine as long | |
5929 | * as the pool has been scrubbed prior to using the new mapping. | |
5930 | * Failure to do can result in non-repairable damage being injected. | |
5931 | */ | |
5932 | if (ztest_pool_scrubbed == B_FALSE) | |
5933 | goto out; | |
5934 | ||
621dd7bb GW |
5935 | /* |
5936 | * Grab the name lock as reader. There are some operations | |
5937 | * which don't like to have their vdevs changed while | |
5938 | * they are in progress (i.e. spa_change_guid). Those | |
5939 | * operations will have grabbed the name lock as writer. | |
5940 | */ | |
2d9142c9 | 5941 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
621dd7bb | 5942 | |
34dc7c2f | 5943 | /* |
b128c09f | 5944 | * We need SCL_STATE here because we're going to look at vd0->vdev_tsd. |
34dc7c2f | 5945 | */ |
b128c09f | 5946 | spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); |
34dc7c2f | 5947 | |
b128c09f BB |
5948 | if (ztest_random(2) == 0) { |
5949 | /* | |
428870ff | 5950 | * Inject errors on a normal data device or slog device. |
b128c09f | 5951 | */ |
428870ff BB |
5952 | top = ztest_random_vdev_top(spa, B_TRUE); |
5953 | leaf = ztest_random(leaves) + zs->zs_splits; | |
34dc7c2f | 5954 | |
b128c09f BB |
5955 | /* |
5956 | * Generate paths to the first leaf in this top-level vdev, | |
5957 | * and to the random leaf we selected. We'll induce transient | |
5958 | * write failures and random online/offline activity on leaf 0, | |
5959 | * and we'll write random garbage to the randomly chosen leaf. | |
5960 | */ | |
6aec1cd5 | 5961 | (void) snprintf(path0, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
5962 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
5963 | top * leaves + zs->zs_splits); | |
6aec1cd5 | 5964 | (void) snprintf(pathrand, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
5965 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
5966 | top * leaves + leaf); | |
34dc7c2f | 5967 | |
b128c09f | 5968 | vd0 = vdev_lookup_by_path(spa->spa_root_vdev, path0); |
428870ff BB |
5969 | if (vd0 != NULL && vd0->vdev_top->vdev_islog) |
5970 | islog = B_TRUE; | |
5971 | ||
621dd7bb GW |
5972 | /* |
5973 | * If the top-level vdev needs to be resilvered | |
5974 | * then we only allow faults on the device that is | |
5975 | * resilvering. | |
5976 | */ | |
5977 | if (vd0 != NULL && maxfaults != 1 && | |
5978 | (!vdev_resilver_needed(vd0->vdev_top, NULL, NULL) || | |
5d1f7fb6 | 5979 | vd0->vdev_resilver_txg != 0)) { |
b128c09f BB |
5980 | /* |
5981 | * Make vd0 explicitly claim to be unreadable, | |
5982 | * or unwriteable, or reach behind its back | |
5983 | * and close the underlying fd. We can do this if | |
5984 | * maxfaults == 0 because we'll fail and reexecute, | |
5985 | * and we can do it if maxfaults >= 2 because we'll | |
5986 | * have enough redundancy. If maxfaults == 1, the | |
5987 | * combination of this with injection of random data | |
5988 | * corruption below exceeds the pool's fault tolerance. | |
5989 | */ | |
5990 | vdev_file_t *vf = vd0->vdev_tsd; | |
5991 | ||
a1d477c2 MA |
5992 | zfs_dbgmsg("injecting fault to vdev %llu; maxfaults=%d", |
5993 | (long long)vd0->vdev_id, (int)maxfaults); | |
5994 | ||
b128c09f | 5995 | if (vf != NULL && ztest_random(3) == 0) { |
da92d5cb MM |
5996 | (void) close(vf->vf_file->f_fd); |
5997 | vf->vf_file->f_fd = -1; | |
b128c09f BB |
5998 | } else if (ztest_random(2) == 0) { |
5999 | vd0->vdev_cant_read = B_TRUE; | |
6000 | } else { | |
6001 | vd0->vdev_cant_write = B_TRUE; | |
6002 | } | |
6003 | guid0 = vd0->vdev_guid; | |
6004 | } | |
6005 | } else { | |
6006 | /* | |
6007 | * Inject errors on an l2cache device. | |
6008 | */ | |
6009 | spa_aux_vdev_t *sav = &spa->spa_l2cache; | |
34dc7c2f | 6010 | |
b128c09f BB |
6011 | if (sav->sav_count == 0) { |
6012 | spa_config_exit(spa, SCL_STATE, FTAG); | |
2d9142c9 | 6013 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
40b84e7a | 6014 | goto out; |
b128c09f BB |
6015 | } |
6016 | vd0 = sav->sav_vdevs[ztest_random(sav->sav_count)]; | |
34dc7c2f | 6017 | guid0 = vd0->vdev_guid; |
b128c09f BB |
6018 | (void) strcpy(path0, vd0->vdev_path); |
6019 | (void) strcpy(pathrand, vd0->vdev_path); | |
6020 | ||
6021 | leaf = 0; | |
6022 | leaves = 1; | |
6023 | maxfaults = INT_MAX; /* no limit on cache devices */ | |
34dc7c2f BB |
6024 | } |
6025 | ||
b128c09f | 6026 | spa_config_exit(spa, SCL_STATE, FTAG); |
2d9142c9 | 6027 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
b128c09f | 6028 | |
34dc7c2f | 6029 | /* |
428870ff BB |
6030 | * If we can tolerate two or more faults, or we're dealing |
6031 | * with a slog, randomly online/offline vd0. | |
34dc7c2f | 6032 | */ |
428870ff | 6033 | if ((maxfaults >= 2 || islog) && guid0 != 0) { |
fb5f0bc8 BB |
6034 | if (ztest_random(10) < 6) { |
6035 | int flags = (ztest_random(2) == 0 ? | |
6036 | ZFS_OFFLINE_TEMPORARY : 0); | |
428870ff BB |
6037 | |
6038 | /* | |
6039 | * We have to grab the zs_name_lock as writer to | |
6040 | * prevent a race between offlining a slog and | |
6041 | * destroying a dataset. Offlining the slog will | |
6042 | * grab a reference on the dataset which may cause | |
13fe0198 | 6043 | * dsl_destroy_head() to fail with EBUSY thus |
428870ff BB |
6044 | * leaving the dataset in an inconsistent state. |
6045 | */ | |
6046 | if (islog) | |
2d9142c9 | 6047 | (void) pthread_rwlock_wrlock(&ztest_name_lock); |
428870ff | 6048 | |
a9eaae06 | 6049 | VERIFY3U(vdev_offline(spa, guid0, flags), !=, EBUSY); |
428870ff BB |
6050 | |
6051 | if (islog) | |
2d9142c9 | 6052 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
fb5f0bc8 | 6053 | } else { |
1eb5bfa3 GW |
6054 | /* |
6055 | * Ideally we would like to be able to randomly | |
6056 | * call vdev_[on|off]line without holding locks | |
6057 | * to force unpredictable failures but the side | |
6058 | * effects of vdev_[on|off]line prevent us from | |
6059 | * doing so. We grab the ztest_vdev_lock here to | |
6060 | * prevent a race between injection testing and | |
6061 | * aux_vdev removal. | |
6062 | */ | |
6063 | mutex_enter(&ztest_vdev_lock); | |
fb5f0bc8 | 6064 | (void) vdev_online(spa, guid0, 0, NULL); |
1eb5bfa3 | 6065 | mutex_exit(&ztest_vdev_lock); |
fb5f0bc8 | 6066 | } |
34dc7c2f BB |
6067 | } |
6068 | ||
428870ff | 6069 | if (maxfaults == 0) |
40b84e7a | 6070 | goto out; |
428870ff | 6071 | |
34dc7c2f BB |
6072 | /* |
6073 | * We have at least single-fault tolerance, so inject data corruption. | |
6074 | */ | |
6075 | fd = open(pathrand, O_RDWR); | |
6076 | ||
d2734cce | 6077 | if (fd == -1) /* we hit a gap in the device namespace */ |
40b84e7a | 6078 | goto out; |
34dc7c2f BB |
6079 | |
6080 | fsize = lseek(fd, 0, SEEK_END); | |
6081 | ||
6082 | while (--iters != 0) { | |
91d88843 MA |
6083 | /* |
6084 | * The offset must be chosen carefully to ensure that | |
6085 | * we do not inject a given logical block with errors | |
6086 | * on two different leaf devices, because ZFS can not | |
6087 | * tolerate that (if maxfaults==1). | |
6088 | * | |
42d8d1d6 SD |
6089 | * To achieve this we divide each leaf device into |
6090 | * chunks of size (# leaves * SPA_MAXBLOCKSIZE * 4). | |
6091 | * Each chunk is further divided into error-injection | |
6092 | * ranges (can accept errors) and clear ranges (we do | |
6093 | * not inject errors in those). Each error-injection | |
6094 | * range can accept errors only for a single leaf vdev. | |
6095 | * Error-injection ranges are separated by clear ranges. | |
91d88843 MA |
6096 | * |
6097 | * For example, with 3 leaves, each chunk looks like: | |
6098 | * 0 to 32M: injection range for leaf 0 | |
42d8d1d6 | 6099 | * 32M to 64M: clear range - no injection allowed |
91d88843 | 6100 | * 64M to 96M: injection range for leaf 1 |
42d8d1d6 | 6101 | * 96M to 128M: clear range - no injection allowed |
91d88843 | 6102 | * 128M to 160M: injection range for leaf 2 |
42d8d1d6 SD |
6103 | * 160M to 192M: clear range - no injection allowed |
6104 | * | |
6105 | * Each clear range must be large enough such that a | |
6106 | * single block cannot straddle it. This way a block | |
6107 | * can't be a target in two different injection ranges | |
6108 | * (on different leaf vdevs). | |
91d88843 | 6109 | */ |
34dc7c2f BB |
6110 | offset = ztest_random(fsize / (leaves << bshift)) * |
6111 | (leaves << bshift) + (leaf << bshift) + | |
6112 | (ztest_random(1ULL << (bshift - 1)) & -8ULL); | |
6113 | ||
2014c09f GM |
6114 | /* |
6115 | * Only allow damage to the labels at one end of the vdev. | |
6116 | * | |
6117 | * If all labels are damaged, the device will be totally | |
6118 | * inaccessible, which will result in loss of data, | |
6119 | * because we also damage (parts of) the other side of | |
6120 | * the mirror/raidz. | |
6121 | * | |
6122 | * Additionally, we will always have both an even and an | |
6123 | * odd label, so that we can handle crashes in the | |
6124 | * middle of vdev_config_sync(). | |
6125 | */ | |
6126 | if ((leaf & 1) == 0 && offset < VDEV_LABEL_START_SIZE) | |
6127 | continue; | |
6128 | ||
6129 | /* | |
6130 | * The two end labels are stored at the "end" of the disk, but | |
6131 | * the end of the disk (vdev_psize) is aligned to | |
6132 | * sizeof (vdev_label_t). | |
6133 | */ | |
6134 | uint64_t psize = P2ALIGN(fsize, sizeof (vdev_label_t)); | |
6135 | if ((leaf & 1) == 1 && | |
6136 | offset + sizeof (bad) > psize - VDEV_LABEL_END_SIZE) | |
34dc7c2f BB |
6137 | continue; |
6138 | ||
c242c188 | 6139 | mutex_enter(&ztest_vdev_lock); |
428870ff | 6140 | if (mirror_save != zs->zs_mirrors) { |
c242c188 | 6141 | mutex_exit(&ztest_vdev_lock); |
428870ff | 6142 | (void) close(fd); |
40b84e7a | 6143 | goto out; |
428870ff | 6144 | } |
34dc7c2f BB |
6145 | |
6146 | if (pwrite(fd, &bad, sizeof (bad), offset) != sizeof (bad)) | |
6147 | fatal(1, "can't inject bad word at 0x%llx in %s", | |
6148 | offset, pathrand); | |
428870ff | 6149 | |
c242c188 | 6150 | mutex_exit(&ztest_vdev_lock); |
428870ff | 6151 | |
c242c188 | 6152 | if (ztest_opts.zo_verbose >= 7) |
428870ff BB |
6153 | (void) printf("injected bad word into %s," |
6154 | " offset 0x%llx\n", pathrand, (u_longlong_t)offset); | |
34dc7c2f BB |
6155 | } |
6156 | ||
6157 | (void) close(fd); | |
40b84e7a BB |
6158 | out: |
6159 | umem_free(path0, MAXPATHLEN); | |
6160 | umem_free(pathrand, MAXPATHLEN); | |
34dc7c2f BB |
6161 | } |
6162 | ||
ce5fb2a7 BB |
6163 | /* |
6164 | * By design ztest will never inject uncorrectable damage in to the pool. | |
6165 | * Issue a scrub, wait for it to complete, and verify there is never any | |
b2255edc | 6166 | * persistent damage. |
ce5fb2a7 BB |
6167 | * |
6168 | * Only after a full scrub has been completed is it safe to start injecting | |
6169 | * data corruption. See the comment in zfs_fault_inject(). | |
6170 | */ | |
6171 | static int | |
6172 | ztest_scrub_impl(spa_t *spa) | |
6173 | { | |
6174 | int error = spa_scan(spa, POOL_SCAN_SCRUB); | |
6175 | if (error) | |
6176 | return (error); | |
6177 | ||
6178 | while (dsl_scan_scrubbing(spa_get_dsl(spa))) | |
6179 | txg_wait_synced(spa_get_dsl(spa), 0); | |
6180 | ||
6181 | if (spa_get_errlog_size(spa) > 0) | |
6182 | return (ECKSUM); | |
6183 | ||
6184 | ztest_pool_scrubbed = B_TRUE; | |
6185 | ||
6186 | return (0); | |
6187 | } | |
6188 | ||
34dc7c2f | 6189 | /* |
428870ff | 6190 | * Scrub the pool. |
34dc7c2f | 6191 | */ |
428870ff BB |
6192 | /* ARGSUSED */ |
6193 | void | |
6194 | ztest_scrub(ztest_ds_t *zd, uint64_t id) | |
34dc7c2f | 6195 | { |
c242c188 | 6196 | spa_t *spa = ztest_spa; |
ce5fb2a7 | 6197 | int error; |
34dc7c2f | 6198 | |
9e052db4 MA |
6199 | /* |
6200 | * Scrub in progress by device removal. | |
6201 | */ | |
6202 | if (ztest_device_removal_active) | |
6203 | return; | |
6204 | ||
ce5fb2a7 BB |
6205 | /* |
6206 | * Start a scrub, wait a moment, then force a restart. | |
6207 | */ | |
428870ff | 6208 | (void) spa_scan(spa, POOL_SCAN_SCRUB); |
ce5fb2a7 BB |
6209 | (void) poll(NULL, 0, 100); |
6210 | ||
6211 | error = ztest_scrub_impl(spa); | |
6212 | if (error == EBUSY) | |
6213 | error = 0; | |
6214 | ASSERT0(error); | |
428870ff | 6215 | } |
34dc7c2f | 6216 | |
3541dc6d GA |
6217 | /* |
6218 | * Change the guid for the pool. | |
6219 | */ | |
6220 | /* ARGSUSED */ | |
6221 | void | |
6222 | ztest_reguid(ztest_ds_t *zd, uint64_t id) | |
6223 | { | |
c242c188 | 6224 | spa_t *spa = ztest_spa; |
3541dc6d | 6225 | uint64_t orig, load; |
3bc7e0fb | 6226 | int error; |
3541dc6d | 6227 | |
b33d668d OF |
6228 | if (ztest_opts.zo_mmp_test) |
6229 | return; | |
6230 | ||
3541dc6d GA |
6231 | orig = spa_guid(spa); |
6232 | load = spa_load_guid(spa); | |
3bc7e0fb | 6233 | |
2d9142c9 | 6234 | (void) pthread_rwlock_wrlock(&ztest_name_lock); |
3bc7e0fb | 6235 | error = spa_change_guid(spa); |
2d9142c9 | 6236 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
3bc7e0fb GW |
6237 | |
6238 | if (error != 0) | |
3541dc6d GA |
6239 | return; |
6240 | ||
ea0b2538 | 6241 | if (ztest_opts.zo_verbose >= 4) { |
3541dc6d GA |
6242 | (void) printf("Changed guid old %llu -> %llu\n", |
6243 | (u_longlong_t)orig, (u_longlong_t)spa_guid(spa)); | |
6244 | } | |
6245 | ||
6246 | VERIFY3U(orig, !=, spa_guid(spa)); | |
6247 | VERIFY3U(load, ==, spa_load_guid(spa)); | |
6248 | } | |
6249 | ||
1eeb4562 JX |
6250 | void |
6251 | ztest_fletcher(ztest_ds_t *zd, uint64_t id) | |
6252 | { | |
6253 | hrtime_t end = gethrtime() + NANOSEC; | |
6254 | ||
6255 | while (gethrtime() <= end) { | |
6256 | int run_count = 100; | |
6257 | void *buf; | |
2fe36b0b | 6258 | struct abd *abd_data, *abd_meta; |
1eeb4562 JX |
6259 | uint32_t size; |
6260 | int *ptr; | |
6261 | int i; | |
6262 | zio_cksum_t zc_ref; | |
6263 | zio_cksum_t zc_ref_byteswap; | |
6264 | ||
6265 | size = ztest_random_blocksize(); | |
2fe36b0b | 6266 | |
1eeb4562 | 6267 | buf = umem_alloc(size, UMEM_NOFAIL); |
2fe36b0b DQ |
6268 | abd_data = abd_alloc(size, B_FALSE); |
6269 | abd_meta = abd_alloc(size, B_TRUE); | |
1eeb4562 JX |
6270 | |
6271 | for (i = 0, ptr = buf; i < size / sizeof (*ptr); i++, ptr++) | |
6272 | *ptr = ztest_random(UINT_MAX); | |
6273 | ||
2fe36b0b DQ |
6274 | abd_copy_from_buf_off(abd_data, buf, 0, size); |
6275 | abd_copy_from_buf_off(abd_meta, buf, 0, size); | |
6276 | ||
1eeb4562 | 6277 | VERIFY0(fletcher_4_impl_set("scalar")); |
3c67d83a TH |
6278 | fletcher_4_native(buf, size, NULL, &zc_ref); |
6279 | fletcher_4_byteswap(buf, size, NULL, &zc_ref_byteswap); | |
1eeb4562 JX |
6280 | |
6281 | VERIFY0(fletcher_4_impl_set("cycle")); | |
6282 | while (run_count-- > 0) { | |
6283 | zio_cksum_t zc; | |
6284 | zio_cksum_t zc_byteswap; | |
6285 | ||
3c67d83a TH |
6286 | fletcher_4_byteswap(buf, size, NULL, &zc_byteswap); |
6287 | fletcher_4_native(buf, size, NULL, &zc); | |
1eeb4562 JX |
6288 | |
6289 | VERIFY0(bcmp(&zc, &zc_ref, sizeof (zc))); | |
6290 | VERIFY0(bcmp(&zc_byteswap, &zc_ref_byteswap, | |
6291 | sizeof (zc_byteswap))); | |
2fe36b0b DQ |
6292 | |
6293 | /* Test ABD - data */ | |
6294 | abd_fletcher_4_byteswap(abd_data, size, NULL, | |
6295 | &zc_byteswap); | |
6296 | abd_fletcher_4_native(abd_data, size, NULL, &zc); | |
6297 | ||
6298 | VERIFY0(bcmp(&zc, &zc_ref, sizeof (zc))); | |
6299 | VERIFY0(bcmp(&zc_byteswap, &zc_ref_byteswap, | |
6300 | sizeof (zc_byteswap))); | |
6301 | ||
6302 | /* Test ABD - metadata */ | |
6303 | abd_fletcher_4_byteswap(abd_meta, size, NULL, | |
6304 | &zc_byteswap); | |
6305 | abd_fletcher_4_native(abd_meta, size, NULL, &zc); | |
6306 | ||
6307 | VERIFY0(bcmp(&zc, &zc_ref, sizeof (zc))); | |
6308 | VERIFY0(bcmp(&zc_byteswap, &zc_ref_byteswap, | |
6309 | sizeof (zc_byteswap))); | |
6310 | ||
1eeb4562 JX |
6311 | } |
6312 | ||
6313 | umem_free(buf, size); | |
2fe36b0b DQ |
6314 | abd_free(abd_data); |
6315 | abd_free(abd_meta); | |
1eeb4562 JX |
6316 | } |
6317 | } | |
6318 | ||
37f520db GN |
6319 | void |
6320 | ztest_fletcher_incr(ztest_ds_t *zd, uint64_t id) | |
6321 | { | |
6322 | void *buf; | |
6323 | size_t size; | |
6324 | int *ptr; | |
6325 | int i; | |
6326 | zio_cksum_t zc_ref; | |
6327 | zio_cksum_t zc_ref_bswap; | |
6328 | ||
6329 | hrtime_t end = gethrtime() + NANOSEC; | |
6330 | ||
6331 | while (gethrtime() <= end) { | |
6332 | int run_count = 100; | |
6333 | ||
6334 | size = ztest_random_blocksize(); | |
6335 | buf = umem_alloc(size, UMEM_NOFAIL); | |
6336 | ||
6337 | for (i = 0, ptr = buf; i < size / sizeof (*ptr); i++, ptr++) | |
6338 | *ptr = ztest_random(UINT_MAX); | |
6339 | ||
6340 | VERIFY0(fletcher_4_impl_set("scalar")); | |
6341 | fletcher_4_native(buf, size, NULL, &zc_ref); | |
6342 | fletcher_4_byteswap(buf, size, NULL, &zc_ref_bswap); | |
6343 | ||
6344 | VERIFY0(fletcher_4_impl_set("cycle")); | |
6345 | ||
6346 | while (run_count-- > 0) { | |
6347 | zio_cksum_t zc; | |
6348 | zio_cksum_t zc_bswap; | |
6349 | size_t pos = 0; | |
6350 | ||
6351 | ZIO_SET_CHECKSUM(&zc, 0, 0, 0, 0); | |
6352 | ZIO_SET_CHECKSUM(&zc_bswap, 0, 0, 0, 0); | |
6353 | ||
6354 | while (pos < size) { | |
6355 | size_t inc = 64 * ztest_random(size / 67); | |
6356 | /* sometimes add few bytes to test non-simd */ | |
6357 | if (ztest_random(100) < 10) | |
6358 | inc += P2ALIGN(ztest_random(64), | |
6359 | sizeof (uint32_t)); | |
6360 | ||
6361 | if (inc > (size - pos)) | |
6362 | inc = size - pos; | |
6363 | ||
6364 | fletcher_4_incremental_native(buf + pos, inc, | |
6365 | &zc); | |
6366 | fletcher_4_incremental_byteswap(buf + pos, inc, | |
6367 | &zc_bswap); | |
6368 | ||
6369 | pos += inc; | |
6370 | } | |
6371 | ||
6372 | VERIFY3U(pos, ==, size); | |
6373 | ||
6374 | VERIFY(ZIO_CHECKSUM_EQUAL(zc, zc_ref)); | |
6375 | VERIFY(ZIO_CHECKSUM_EQUAL(zc_bswap, zc_ref_bswap)); | |
6376 | ||
6377 | /* | |
6378 | * verify if incremental on the whole buffer is | |
6379 | * equivalent to non-incremental version | |
6380 | */ | |
6381 | ZIO_SET_CHECKSUM(&zc, 0, 0, 0, 0); | |
6382 | ZIO_SET_CHECKSUM(&zc_bswap, 0, 0, 0, 0); | |
6383 | ||
6384 | fletcher_4_incremental_native(buf, size, &zc); | |
6385 | fletcher_4_incremental_byteswap(buf, size, &zc_bswap); | |
6386 | ||
6387 | VERIFY(ZIO_CHECKSUM_EQUAL(zc, zc_ref)); | |
6388 | VERIFY(ZIO_CHECKSUM_EQUAL(zc_bswap, zc_ref_bswap)); | |
6389 | } | |
6390 | ||
6391 | umem_free(buf, size); | |
6392 | } | |
6393 | } | |
6394 | ||
52cb284f CS |
6395 | static int |
6396 | ztest_set_global_vars(void) | |
6397 | { | |
6398 | for (size_t i = 0; i < ztest_opts.zo_gvars_count; i++) { | |
6399 | char *kv = ztest_opts.zo_gvars[i]; | |
6400 | VERIFY3U(strlen(kv), <=, ZO_GVARS_MAX_ARGLEN); | |
6401 | VERIFY3U(strlen(kv), >, 0); | |
6402 | int err = set_global_var(kv); | |
6403 | if (ztest_opts.zo_verbose > 0) { | |
6404 | (void) printf("setting global var %s ... %s\n", kv, | |
6405 | err ? "failed" : "ok"); | |
6406 | } | |
6407 | if (err != 0) { | |
6408 | (void) fprintf(stderr, | |
6409 | "failed to set global var '%s'\n", kv); | |
6410 | return (err); | |
6411 | } | |
6412 | } | |
6413 | return (0); | |
6414 | } | |
6415 | ||
6416 | static char ** | |
6417 | ztest_global_vars_to_zdb_args(void) | |
6418 | { | |
6419 | char **args = calloc(2*ztest_opts.zo_gvars_count + 1, sizeof (char *)); | |
6420 | char **cur = args; | |
6421 | for (size_t i = 0; i < ztest_opts.zo_gvars_count; i++) { | |
6422 | char *kv = ztest_opts.zo_gvars[i]; | |
6423 | *cur = "-o"; | |
6424 | cur++; | |
6425 | *cur = strdup(kv); | |
6426 | cur++; | |
6427 | } | |
6428 | ASSERT3P(cur, ==, &args[2*ztest_opts.zo_gvars_count]); | |
6429 | *cur = NULL; | |
6430 | return (args); | |
6431 | } | |
6432 | ||
6433 | /* The end of strings is indicated by a NULL element */ | |
6434 | static char * | |
6435 | join_strings(char **strings, const char *sep) | |
6436 | { | |
6437 | size_t totallen = 0; | |
6438 | for (char **sp = strings; *sp != NULL; sp++) { | |
6439 | totallen += strlen(*sp); | |
6440 | totallen += strlen(sep); | |
6441 | } | |
6442 | if (totallen > 0) { | |
6443 | ASSERT(totallen >= strlen(sep)); | |
6444 | totallen -= strlen(sep); | |
6445 | } | |
6446 | ||
6447 | size_t buflen = totallen + 1; | |
6448 | char *o = malloc(buflen); /* trailing 0 byte */ | |
6449 | o[0] = '\0'; | |
6450 | for (char **sp = strings; *sp != NULL; sp++) { | |
6451 | size_t would; | |
6452 | would = strlcat(o, *sp, buflen); | |
6453 | VERIFY3U(would, <, buflen); | |
6454 | if (*(sp+1) == NULL) { | |
6455 | break; | |
6456 | } | |
6457 | would = strlcat(o, sep, buflen); | |
6458 | VERIFY3U(would, <, buflen); | |
6459 | } | |
6460 | ASSERT3S(strlen(o), ==, totallen); | |
6461 | return (o); | |
6462 | } | |
6463 | ||
90aa094d CC |
6464 | static int |
6465 | ztest_check_path(char *path) | |
6466 | { | |
6467 | struct stat s; | |
6468 | /* return true on success */ | |
6469 | return (!stat(path, &s)); | |
6470 | } | |
6471 | ||
6472 | static void | |
6473 | ztest_get_zdb_bin(char *bin, int len) | |
6474 | { | |
6475 | char *zdb_path; | |
6476 | /* | |
6477 | * Try to use ZDB_PATH and in-tree zdb path. If not successful, just | |
6478 | * let popen to search through PATH. | |
6479 | */ | |
6480 | if ((zdb_path = getenv("ZDB_PATH"))) { | |
6481 | strlcpy(bin, zdb_path, len); /* In env */ | |
6482 | if (!ztest_check_path(bin)) { | |
6483 | ztest_dump_core = 0; | |
6484 | fatal(1, "invalid ZDB_PATH '%s'", bin); | |
6485 | } | |
6486 | return; | |
6487 | } | |
6488 | ||
a9eaae06 | 6489 | VERIFY3P(realpath(getexecname(), bin), !=, NULL); |
90aa094d CC |
6490 | if (strstr(bin, "/ztest/")) { |
6491 | strstr(bin, "/ztest/")[0] = '\0'; /* In-tree */ | |
6492 | strcat(bin, "/zdb/zdb"); | |
6493 | if (ztest_check_path(bin)) | |
6494 | return; | |
6495 | } | |
6496 | strcpy(bin, "zdb"); | |
6497 | } | |
6498 | ||
619f0976 GW |
6499 | static vdev_t * |
6500 | ztest_random_concrete_vdev_leaf(vdev_t *vd) | |
6501 | { | |
6502 | if (vd == NULL) | |
6503 | return (NULL); | |
6504 | ||
6505 | if (vd->vdev_children == 0) | |
6506 | return (vd); | |
6507 | ||
6508 | vdev_t *eligible[vd->vdev_children]; | |
6509 | int eligible_idx = 0, i; | |
6510 | for (i = 0; i < vd->vdev_children; i++) { | |
6511 | vdev_t *cvd = vd->vdev_child[i]; | |
6512 | if (cvd->vdev_top->vdev_removing) | |
6513 | continue; | |
6514 | if (cvd->vdev_children > 0 || | |
6515 | (vdev_is_concrete(cvd) && !cvd->vdev_detached)) { | |
6516 | eligible[eligible_idx++] = cvd; | |
6517 | } | |
6518 | } | |
a9eaae06 | 6519 | VERIFY3S(eligible_idx, >, 0); |
619f0976 GW |
6520 | |
6521 | uint64_t child_no = ztest_random(eligible_idx); | |
6522 | return (ztest_random_concrete_vdev_leaf(eligible[child_no])); | |
6523 | } | |
6524 | ||
6525 | /* ARGSUSED */ | |
6526 | void | |
6527 | ztest_initialize(ztest_ds_t *zd, uint64_t id) | |
6528 | { | |
6529 | spa_t *spa = ztest_spa; | |
6530 | int error = 0; | |
6531 | ||
6532 | mutex_enter(&ztest_vdev_lock); | |
6533 | ||
6534 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
6535 | ||
6536 | /* Random leaf vdev */ | |
6537 | vdev_t *rand_vd = ztest_random_concrete_vdev_leaf(spa->spa_root_vdev); | |
6538 | if (rand_vd == NULL) { | |
6539 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
6540 | mutex_exit(&ztest_vdev_lock); | |
6541 | return; | |
6542 | } | |
6543 | ||
6544 | /* | |
6545 | * The random vdev we've selected may change as soon as we | |
6546 | * drop the spa_config_lock. We create local copies of things | |
6547 | * we're interested in. | |
6548 | */ | |
6549 | uint64_t guid = rand_vd->vdev_guid; | |
6550 | char *path = strdup(rand_vd->vdev_path); | |
6551 | boolean_t active = rand_vd->vdev_initialize_thread != NULL; | |
6552 | ||
a887d653 | 6553 | zfs_dbgmsg("vd %px, guid %llu", rand_vd, guid); |
619f0976 GW |
6554 | spa_config_exit(spa, SCL_VDEV, FTAG); |
6555 | ||
6556 | uint64_t cmd = ztest_random(POOL_INITIALIZE_FUNCS); | |
c10d37dd GW |
6557 | |
6558 | nvlist_t *vdev_guids = fnvlist_alloc(); | |
6559 | nvlist_t *vdev_errlist = fnvlist_alloc(); | |
6560 | fnvlist_add_uint64(vdev_guids, path, guid); | |
6561 | error = spa_vdev_initialize(spa, vdev_guids, cmd, vdev_errlist); | |
6562 | fnvlist_free(vdev_guids); | |
6563 | fnvlist_free(vdev_errlist); | |
6564 | ||
619f0976 GW |
6565 | switch (cmd) { |
6566 | case POOL_INITIALIZE_CANCEL: | |
6567 | if (ztest_opts.zo_verbose >= 4) { | |
6568 | (void) printf("Cancel initialize %s", path); | |
6569 | if (!active) | |
6570 | (void) printf(" failed (no initialize active)"); | |
6571 | (void) printf("\n"); | |
6572 | } | |
6573 | break; | |
1b939560 | 6574 | case POOL_INITIALIZE_START: |
619f0976 GW |
6575 | if (ztest_opts.zo_verbose >= 4) { |
6576 | (void) printf("Start initialize %s", path); | |
6577 | if (active && error == 0) | |
6578 | (void) printf(" failed (already active)"); | |
6579 | else if (error != 0) | |
6580 | (void) printf(" failed (error %d)", error); | |
6581 | (void) printf("\n"); | |
6582 | } | |
6583 | break; | |
6584 | case POOL_INITIALIZE_SUSPEND: | |
6585 | if (ztest_opts.zo_verbose >= 4) { | |
6586 | (void) printf("Suspend initialize %s", path); | |
6587 | if (!active) | |
6588 | (void) printf(" failed (no initialize active)"); | |
6589 | (void) printf("\n"); | |
6590 | } | |
6591 | break; | |
6592 | } | |
6593 | free(path); | |
6594 | mutex_exit(&ztest_vdev_lock); | |
6595 | } | |
6596 | ||
1b939560 BB |
6597 | /* ARGSUSED */ |
6598 | void | |
6599 | ztest_trim(ztest_ds_t *zd, uint64_t id) | |
6600 | { | |
6601 | spa_t *spa = ztest_spa; | |
6602 | int error = 0; | |
6603 | ||
6604 | mutex_enter(&ztest_vdev_lock); | |
6605 | ||
6606 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
6607 | ||
6608 | /* Random leaf vdev */ | |
6609 | vdev_t *rand_vd = ztest_random_concrete_vdev_leaf(spa->spa_root_vdev); | |
6610 | if (rand_vd == NULL) { | |
6611 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
6612 | mutex_exit(&ztest_vdev_lock); | |
6613 | return; | |
6614 | } | |
6615 | ||
6616 | /* | |
6617 | * The random vdev we've selected may change as soon as we | |
6618 | * drop the spa_config_lock. We create local copies of things | |
6619 | * we're interested in. | |
6620 | */ | |
6621 | uint64_t guid = rand_vd->vdev_guid; | |
6622 | char *path = strdup(rand_vd->vdev_path); | |
6623 | boolean_t active = rand_vd->vdev_trim_thread != NULL; | |
6624 | ||
6625 | zfs_dbgmsg("vd %p, guid %llu", rand_vd, guid); | |
6626 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
6627 | ||
6628 | uint64_t cmd = ztest_random(POOL_TRIM_FUNCS); | |
6629 | uint64_t rate = 1 << ztest_random(30); | |
6630 | boolean_t partial = (ztest_random(5) > 0); | |
6631 | boolean_t secure = (ztest_random(5) > 0); | |
6632 | ||
6633 | nvlist_t *vdev_guids = fnvlist_alloc(); | |
6634 | nvlist_t *vdev_errlist = fnvlist_alloc(); | |
6635 | fnvlist_add_uint64(vdev_guids, path, guid); | |
6636 | error = spa_vdev_trim(spa, vdev_guids, cmd, rate, partial, | |
6637 | secure, vdev_errlist); | |
6638 | fnvlist_free(vdev_guids); | |
6639 | fnvlist_free(vdev_errlist); | |
6640 | ||
6641 | switch (cmd) { | |
6642 | case POOL_TRIM_CANCEL: | |
6643 | if (ztest_opts.zo_verbose >= 4) { | |
6644 | (void) printf("Cancel TRIM %s", path); | |
6645 | if (!active) | |
6646 | (void) printf(" failed (no TRIM active)"); | |
6647 | (void) printf("\n"); | |
6648 | } | |
6649 | break; | |
6650 | case POOL_TRIM_START: | |
6651 | if (ztest_opts.zo_verbose >= 4) { | |
6652 | (void) printf("Start TRIM %s", path); | |
6653 | if (active && error == 0) | |
6654 | (void) printf(" failed (already active)"); | |
6655 | else if (error != 0) | |
6656 | (void) printf(" failed (error %d)", error); | |
6657 | (void) printf("\n"); | |
6658 | } | |
6659 | break; | |
6660 | case POOL_TRIM_SUSPEND: | |
6661 | if (ztest_opts.zo_verbose >= 4) { | |
6662 | (void) printf("Suspend TRIM %s", path); | |
6663 | if (!active) | |
6664 | (void) printf(" failed (no TRIM active)"); | |
6665 | (void) printf("\n"); | |
6666 | } | |
6667 | break; | |
6668 | } | |
6669 | free(path); | |
6670 | mutex_exit(&ztest_vdev_lock); | |
6671 | } | |
6672 | ||
428870ff BB |
6673 | /* |
6674 | * Verify pool integrity by running zdb. | |
6675 | */ | |
34dc7c2f | 6676 | static void |
428870ff | 6677 | ztest_run_zdb(char *pool) |
34dc7c2f BB |
6678 | { |
6679 | int status; | |
34dc7c2f | 6680 | char *bin; |
0e8d1b2d BB |
6681 | char *zdb; |
6682 | char *zbuf; | |
90aa094d | 6683 | const int len = MAXPATHLEN + MAXNAMELEN + 20; |
34dc7c2f BB |
6684 | FILE *fp; |
6685 | ||
90aa094d CC |
6686 | bin = umem_alloc(len, UMEM_NOFAIL); |
6687 | zdb = umem_alloc(len, UMEM_NOFAIL); | |
0e8d1b2d | 6688 | zbuf = umem_alloc(1024, UMEM_NOFAIL); |
34dc7c2f | 6689 | |
90aa094d | 6690 | ztest_get_zdb_bin(bin, len); |
0e8d1b2d | 6691 | |
52cb284f CS |
6692 | char **set_gvars_args = ztest_global_vars_to_zdb_args(); |
6693 | char *set_gvars_args_joined = join_strings(set_gvars_args, " "); | |
6694 | free(set_gvars_args); | |
6695 | ||
6696 | size_t would = snprintf(zdb, len, | |
6697 | "%s -bcc%s%s -G -d -Y -e -y %s -p %s %s", | |
0e8d1b2d | 6698 | bin, |
c242c188 CS |
6699 | ztest_opts.zo_verbose >= 3 ? "s" : "", |
6700 | ztest_opts.zo_verbose >= 4 ? "v" : "", | |
52cb284f | 6701 | set_gvars_args_joined, |
3d93161b | 6702 | ztest_opts.zo_dir, |
b128c09f | 6703 | pool); |
52cb284f CS |
6704 | ASSERT3U(would, <, len); |
6705 | ||
6706 | free(set_gvars_args_joined); | |
34dc7c2f | 6707 | |
c242c188 | 6708 | if (ztest_opts.zo_verbose >= 5) |
34dc7c2f BB |
6709 | (void) printf("Executing %s\n", strstr(zdb, "zdb ")); |
6710 | ||
6711 | fp = popen(zdb, "r"); | |
6712 | ||
6aec1cd5 | 6713 | while (fgets(zbuf, 1024, fp) != NULL) |
c242c188 | 6714 | if (ztest_opts.zo_verbose >= 3) |
34dc7c2f BB |
6715 | (void) printf("%s", zbuf); |
6716 | ||
6717 | status = pclose(fp); | |
6718 | ||
6719 | if (status == 0) | |
0e8d1b2d | 6720 | goto out; |
34dc7c2f BB |
6721 | |
6722 | ztest_dump_core = 0; | |
6723 | if (WIFEXITED(status)) | |
6724 | fatal(0, "'%s' exit code %d", zdb, WEXITSTATUS(status)); | |
6725 | else | |
6726 | fatal(0, "'%s' died with signal %d", zdb, WTERMSIG(status)); | |
0e8d1b2d | 6727 | out: |
90aa094d CC |
6728 | umem_free(bin, len); |
6729 | umem_free(zdb, len); | |
0e8d1b2d | 6730 | umem_free(zbuf, 1024); |
34dc7c2f BB |
6731 | } |
6732 | ||
6733 | static void | |
6734 | ztest_walk_pool_directory(char *header) | |
6735 | { | |
6736 | spa_t *spa = NULL; | |
6737 | ||
c242c188 | 6738 | if (ztest_opts.zo_verbose >= 6) |
34dc7c2f BB |
6739 | (void) printf("%s\n", header); |
6740 | ||
6741 | mutex_enter(&spa_namespace_lock); | |
6742 | while ((spa = spa_next(spa)) != NULL) | |
c242c188 | 6743 | if (ztest_opts.zo_verbose >= 6) |
34dc7c2f BB |
6744 | (void) printf("\t%s\n", spa_name(spa)); |
6745 | mutex_exit(&spa_namespace_lock); | |
6746 | } | |
6747 | ||
6748 | static void | |
6749 | ztest_spa_import_export(char *oldname, char *newname) | |
6750 | { | |
fb5f0bc8 | 6751 | nvlist_t *config, *newconfig; |
34dc7c2f BB |
6752 | uint64_t pool_guid; |
6753 | spa_t *spa; | |
13fe0198 | 6754 | int error; |
34dc7c2f | 6755 | |
c242c188 | 6756 | if (ztest_opts.zo_verbose >= 4) { |
34dc7c2f BB |
6757 | (void) printf("import/export: old = %s, new = %s\n", |
6758 | oldname, newname); | |
6759 | } | |
6760 | ||
6761 | /* | |
6762 | * Clean up from previous runs. | |
6763 | */ | |
6764 | (void) spa_destroy(newname); | |
6765 | ||
6766 | /* | |
6767 | * Get the pool's configuration and guid. | |
6768 | */ | |
a9eaae06 | 6769 | VERIFY0(spa_open(oldname, &spa, FTAG)); |
34dc7c2f | 6770 | |
fb5f0bc8 BB |
6771 | /* |
6772 | * Kick off a scrub to tickle scrub/export races. | |
6773 | */ | |
6774 | if (ztest_random(2) == 0) | |
428870ff | 6775 | (void) spa_scan(spa, POOL_SCAN_SCRUB); |
fb5f0bc8 | 6776 | |
34dc7c2f BB |
6777 | pool_guid = spa_guid(spa); |
6778 | spa_close(spa, FTAG); | |
6779 | ||
6780 | ztest_walk_pool_directory("pools before export"); | |
6781 | ||
6782 | /* | |
6783 | * Export it. | |
6784 | */ | |
a9eaae06 | 6785 | VERIFY0(spa_export(oldname, &config, B_FALSE, B_FALSE)); |
34dc7c2f BB |
6786 | |
6787 | ztest_walk_pool_directory("pools after export"); | |
6788 | ||
fb5f0bc8 BB |
6789 | /* |
6790 | * Try to import it. | |
6791 | */ | |
6792 | newconfig = spa_tryimport(config); | |
a9eaae06 | 6793 | ASSERT3P(newconfig, !=, NULL); |
bea3fc71 | 6794 | fnvlist_free(newconfig); |
fb5f0bc8 | 6795 | |
34dc7c2f BB |
6796 | /* |
6797 | * Import it under the new name. | |
6798 | */ | |
13fe0198 MA |
6799 | error = spa_import(newname, config, NULL, 0); |
6800 | if (error != 0) { | |
6801 | dump_nvlist(config, 0); | |
6802 | fatal(B_FALSE, "couldn't import pool %s as %s: error %u", | |
6803 | oldname, newname, error); | |
6804 | } | |
34dc7c2f BB |
6805 | |
6806 | ztest_walk_pool_directory("pools after import"); | |
6807 | ||
6808 | /* | |
6809 | * Try to import it again -- should fail with EEXIST. | |
6810 | */ | |
572e2857 | 6811 | VERIFY3U(EEXIST, ==, spa_import(newname, config, NULL, 0)); |
34dc7c2f BB |
6812 | |
6813 | /* | |
6814 | * Try to import it under a different name -- should fail with EEXIST. | |
6815 | */ | |
572e2857 | 6816 | VERIFY3U(EEXIST, ==, spa_import(oldname, config, NULL, 0)); |
34dc7c2f BB |
6817 | |
6818 | /* | |
6819 | * Verify that the pool is no longer visible under the old name. | |
6820 | */ | |
428870ff | 6821 | VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f BB |
6822 | |
6823 | /* | |
6824 | * Verify that we can open and close the pool using the new name. | |
6825 | */ | |
a9eaae06 RM |
6826 | VERIFY0(spa_open(newname, &spa, FTAG)); |
6827 | ASSERT3U(pool_guid, ==, spa_guid(spa)); | |
34dc7c2f BB |
6828 | spa_close(spa, FTAG); |
6829 | ||
bea3fc71 | 6830 | fnvlist_free(config); |
34dc7c2f BB |
6831 | } |
6832 | ||
fb5f0bc8 BB |
6833 | static void |
6834 | ztest_resume(spa_t *spa) | |
6835 | { | |
c242c188 | 6836 | if (spa_suspended(spa) && ztest_opts.zo_verbose >= 6) |
428870ff BB |
6837 | (void) printf("resuming from suspended state\n"); |
6838 | spa_vdev_state_enter(spa, SCL_NONE); | |
6839 | vdev_clear(spa, NULL); | |
6840 | (void) spa_vdev_state_exit(spa, NULL, 0); | |
6841 | (void) zio_resume(spa); | |
fb5f0bc8 BB |
6842 | } |
6843 | ||
c25b8f99 | 6844 | static void |
fb5f0bc8 | 6845 | ztest_resume_thread(void *arg) |
34dc7c2f | 6846 | { |
b128c09f | 6847 | spa_t *spa = arg; |
34dc7c2f BB |
6848 | |
6849 | while (!ztest_exiting) { | |
428870ff BB |
6850 | if (spa_suspended(spa)) |
6851 | ztest_resume(spa); | |
6852 | (void) poll(NULL, 0, 100); | |
d3c2ae1c GW |
6853 | |
6854 | /* | |
6855 | * Periodically change the zfs_compressed_arc_enabled setting. | |
6856 | */ | |
6857 | if (ztest_random(10) == 0) | |
6858 | zfs_compressed_arc_enabled = ztest_random(2); | |
a6255b7f DQ |
6859 | |
6860 | /* | |
6861 | * Periodically change the zfs_abd_scatter_enabled setting. | |
6862 | */ | |
6863 | if (ztest_random(10) == 0) | |
6864 | zfs_abd_scatter_enabled = ztest_random(2); | |
34dc7c2f | 6865 | } |
34dc7c2f | 6866 | |
1e33ac1e | 6867 | thread_exit(); |
1e33ac1e | 6868 | } |
428870ff | 6869 | |
1e33ac1e | 6870 | static void |
8fb1ede1 | 6871 | ztest_deadman_thread(void *arg) |
1e33ac1e | 6872 | { |
8fb1ede1 BB |
6873 | ztest_shared_t *zs = arg; |
6874 | spa_t *spa = ztest_spa; | |
94102578 | 6875 | hrtime_t delay, overdue, last_run = gethrtime(); |
8fb1ede1 | 6876 | |
94102578 TC |
6877 | delay = (zs->zs_thread_stop - zs->zs_thread_start) + |
6878 | MSEC2NSEC(zfs_deadman_synctime_ms); | |
8fb1ede1 | 6879 | |
94102578 TC |
6880 | while (!ztest_exiting) { |
6881 | /* | |
6882 | * Wait for the delay timer while checking occasionally | |
6883 | * if we should stop. | |
6884 | */ | |
6885 | if (gethrtime() < last_run + delay) { | |
6886 | (void) poll(NULL, 0, 1000); | |
6887 | continue; | |
6888 | } | |
8fb1ede1 BB |
6889 | |
6890 | /* | |
6891 | * If the pool is suspended then fail immediately. Otherwise, | |
6892 | * check to see if the pool is making any progress. If | |
6893 | * vdev_deadman() discovers that there hasn't been any recent | |
6894 | * I/Os then it will end up aborting the tests. | |
6895 | */ | |
6896 | if (spa_suspended(spa) || spa->spa_root_vdev == NULL) { | |
6897 | fatal(0, "aborting test after %llu seconds because " | |
6898 | "pool has transitioned to a suspended state.", | |
6899 | zfs_deadman_synctime_ms / 1000); | |
6900 | } | |
6901 | vdev_deadman(spa->spa_root_vdev, FTAG); | |
6902 | ||
6903 | /* | |
6904 | * If the process doesn't complete within a grace period of | |
6905 | * zfs_deadman_synctime_ms over the expected finish time, | |
6906 | * then it may be hung and is terminated. | |
6907 | */ | |
6908 | overdue = zs->zs_proc_stop + MSEC2NSEC(zfs_deadman_synctime_ms); | |
8fb1ede1 BB |
6909 | if (gethrtime() > overdue) { |
6910 | fatal(0, "aborting test after %llu seconds because " | |
94102578 TC |
6911 | "the process is overdue for termination.", |
6912 | (gethrtime() - zs->zs_proc_start) / NANOSEC); | |
8fb1ede1 BB |
6913 | } |
6914 | ||
6915 | (void) printf("ztest has been running for %lld seconds\n", | |
94102578 TC |
6916 | (gethrtime() - zs->zs_proc_start) / NANOSEC); |
6917 | ||
6918 | last_run = gethrtime(); | |
6919 | delay = MSEC2NSEC(zfs_deadman_checktime_ms); | |
8fb1ede1 | 6920 | } |
94102578 TC |
6921 | |
6922 | thread_exit(); | |
428870ff BB |
6923 | } |
6924 | ||
6925 | static void | |
c242c188 | 6926 | ztest_execute(int test, ztest_info_t *zi, uint64_t id) |
428870ff | 6927 | { |
c242c188 CS |
6928 | ztest_ds_t *zd = &ztest_ds[id % ztest_opts.zo_datasets]; |
6929 | ztest_shared_callstate_t *zc = ZTEST_GET_SHARED_CALLSTATE(test); | |
428870ff | 6930 | hrtime_t functime = gethrtime(); |
d6320ddb | 6931 | int i; |
428870ff | 6932 | |
d6320ddb | 6933 | for (i = 0; i < zi->zi_iters; i++) |
428870ff BB |
6934 | zi->zi_func(zd, id); |
6935 | ||
6936 | functime = gethrtime() - functime; | |
6937 | ||
c242c188 CS |
6938 | atomic_add_64(&zc->zc_count, 1); |
6939 | atomic_add_64(&zc->zc_time, functime); | |
428870ff | 6940 | |
fdc5d982 | 6941 | if (ztest_opts.zo_verbose >= 4) |
428870ff | 6942 | (void) printf("%6.2f sec in %s\n", |
fdc5d982 | 6943 | (double)functime / NANOSEC, zi->zi_funcname); |
428870ff BB |
6944 | } |
6945 | ||
c25b8f99 | 6946 | static void |
34dc7c2f BB |
6947 | ztest_thread(void *arg) |
6948 | { | |
c242c188 | 6949 | int rand; |
428870ff | 6950 | uint64_t id = (uintptr_t)arg; |
34dc7c2f | 6951 | ztest_shared_t *zs = ztest_shared; |
428870ff BB |
6952 | uint64_t call_next; |
6953 | hrtime_t now; | |
34dc7c2f | 6954 | ztest_info_t *zi; |
c242c188 | 6955 | ztest_shared_callstate_t *zc; |
34dc7c2f | 6956 | |
428870ff | 6957 | while ((now = gethrtime()) < zs->zs_thread_stop) { |
34dc7c2f BB |
6958 | /* |
6959 | * See if it's time to force a crash. | |
6960 | */ | |
428870ff BB |
6961 | if (now > zs->zs_thread_kill) |
6962 | ztest_kill(zs); | |
34dc7c2f BB |
6963 | |
6964 | /* | |
428870ff | 6965 | * If we're getting ENOSPC with some regularity, stop. |
34dc7c2f | 6966 | */ |
428870ff BB |
6967 | if (zs->zs_enospc_count > 10) |
6968 | break; | |
34dc7c2f BB |
6969 | |
6970 | /* | |
428870ff | 6971 | * Pick a random function to execute. |
34dc7c2f | 6972 | */ |
c242c188 CS |
6973 | rand = ztest_random(ZTEST_FUNCS); |
6974 | zi = &ztest_info[rand]; | |
6975 | zc = ZTEST_GET_SHARED_CALLSTATE(rand); | |
6976 | call_next = zc->zc_next; | |
34dc7c2f | 6977 | |
428870ff | 6978 | if (now >= call_next && |
c242c188 CS |
6979 | atomic_cas_64(&zc->zc_next, call_next, call_next + |
6980 | ztest_random(2 * zi->zi_interval[0] + 1)) == call_next) { | |
6981 | ztest_execute(rand, zi, id); | |
6982 | } | |
428870ff | 6983 | } |
34dc7c2f | 6984 | |
1e33ac1e | 6985 | thread_exit(); |
428870ff | 6986 | } |
34dc7c2f | 6987 | |
428870ff BB |
6988 | static void |
6989 | ztest_dataset_name(char *dsname, char *pool, int d) | |
6990 | { | |
eca7b760 | 6991 | (void) snprintf(dsname, ZFS_MAX_DATASET_NAME_LEN, "%s/ds_%d", pool, d); |
428870ff | 6992 | } |
34dc7c2f | 6993 | |
428870ff | 6994 | static void |
c242c188 | 6995 | ztest_dataset_destroy(int d) |
428870ff | 6996 | { |
eca7b760 | 6997 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
d6320ddb | 6998 | int t; |
34dc7c2f | 6999 | |
c242c188 | 7000 | ztest_dataset_name(name, ztest_opts.zo_pool, d); |
34dc7c2f | 7001 | |
c242c188 | 7002 | if (ztest_opts.zo_verbose >= 3) |
428870ff | 7003 | (void) printf("Destroying %s to free up space\n", name); |
34dc7c2f | 7004 | |
428870ff BB |
7005 | /* |
7006 | * Cleanup any non-standard clones and snapshots. In general, | |
7007 | * ztest thread t operates on dataset (t % zopt_datasets), | |
7008 | * so there may be more than one thing to clean up. | |
7009 | */ | |
c242c188 CS |
7010 | for (t = d; t < ztest_opts.zo_threads; |
7011 | t += ztest_opts.zo_datasets) | |
428870ff BB |
7012 | ztest_dsl_dataset_cleanup(name, t); |
7013 | ||
7014 | (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL, | |
7015 | DS_FIND_SNAPSHOTS | DS_FIND_CHILDREN); | |
7016 | } | |
7017 | ||
7018 | static void | |
7019 | ztest_dataset_dirobj_verify(ztest_ds_t *zd) | |
7020 | { | |
7021 | uint64_t usedobjs, dirobjs, scratch; | |
7022 | ||
7023 | /* | |
7024 | * ZTEST_DIROBJ is the object directory for the entire dataset. | |
7025 | * Therefore, the number of objects in use should equal the | |
7026 | * number of ZTEST_DIROBJ entries, +1 for ZTEST_DIROBJ itself. | |
7027 | * If not, we have an object leak. | |
7028 | * | |
7029 | * Note that we can only check this in ztest_dataset_open(), | |
7030 | * when the open-context and syncing-context values agree. | |
7031 | * That's because zap_count() returns the open-context value, | |
7032 | * while dmu_objset_space() returns the rootbp fill count. | |
7033 | */ | |
a9eaae06 | 7034 | VERIFY0(zap_count(zd->zd_os, ZTEST_DIROBJ, &dirobjs)); |
428870ff BB |
7035 | dmu_objset_space(zd->zd_os, &scratch, &scratch, &usedobjs, &scratch); |
7036 | ASSERT3U(dirobjs + 1, ==, usedobjs); | |
7037 | } | |
7038 | ||
7039 | static int | |
c242c188 | 7040 | ztest_dataset_open(int d) |
428870ff | 7041 | { |
c242c188 CS |
7042 | ztest_ds_t *zd = &ztest_ds[d]; |
7043 | uint64_t committed_seq = ZTEST_GET_SHARED_DS(d)->zd_seq; | |
428870ff BB |
7044 | objset_t *os; |
7045 | zilog_t *zilog; | |
eca7b760 | 7046 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
428870ff BB |
7047 | int error; |
7048 | ||
c242c188 | 7049 | ztest_dataset_name(name, ztest_opts.zo_pool, d); |
428870ff | 7050 | |
2d9142c9 | 7051 | (void) pthread_rwlock_rdlock(&ztest_name_lock); |
428870ff BB |
7052 | |
7053 | error = ztest_dataset_create(name); | |
7054 | if (error == ENOSPC) { | |
2d9142c9 | 7055 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
428870ff BB |
7056 | ztest_record_enospc(FTAG); |
7057 | return (error); | |
34dc7c2f | 7058 | } |
428870ff | 7059 | ASSERT(error == 0 || error == EEXIST); |
34dc7c2f | 7060 | |
4807c0ba TC |
7061 | VERIFY0(ztest_dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, |
7062 | B_TRUE, zd, &os)); | |
2d9142c9 | 7063 | (void) pthread_rwlock_unlock(&ztest_name_lock); |
428870ff | 7064 | |
c242c188 | 7065 | ztest_zd_init(zd, ZTEST_GET_SHARED_DS(d), os); |
428870ff BB |
7066 | |
7067 | zilog = zd->zd_zilog; | |
7068 | ||
7069 | if (zilog->zl_header->zh_claim_lr_seq != 0 && | |
7070 | zilog->zl_header->zh_claim_lr_seq < committed_seq) | |
7071 | fatal(0, "missing log records: claimed %llu < committed %llu", | |
7072 | zilog->zl_header->zh_claim_lr_seq, committed_seq); | |
7073 | ||
7074 | ztest_dataset_dirobj_verify(zd); | |
7075 | ||
7076 | zil_replay(os, zd, ztest_replay_vector); | |
7077 | ||
7078 | ztest_dataset_dirobj_verify(zd); | |
7079 | ||
c242c188 | 7080 | if (ztest_opts.zo_verbose >= 6) |
428870ff BB |
7081 | (void) printf("%s replay %llu blocks, %llu records, seq %llu\n", |
7082 | zd->zd_name, | |
7083 | (u_longlong_t)zilog->zl_parse_blk_count, | |
7084 | (u_longlong_t)zilog->zl_parse_lr_count, | |
7085 | (u_longlong_t)zilog->zl_replaying_seq); | |
7086 | ||
7087 | zilog = zil_open(os, ztest_get_data); | |
7088 | ||
7089 | if (zilog->zl_replaying_seq != 0 && | |
7090 | zilog->zl_replaying_seq < committed_seq) | |
7091 | fatal(0, "missing log records: replayed %llu < committed %llu", | |
7092 | zilog->zl_replaying_seq, committed_seq); | |
7093 | ||
7094 | return (0); | |
7095 | } | |
7096 | ||
7097 | static void | |
c242c188 | 7098 | ztest_dataset_close(int d) |
428870ff | 7099 | { |
c242c188 | 7100 | ztest_ds_t *zd = &ztest_ds[d]; |
428870ff BB |
7101 | |
7102 | zil_close(zd->zd_zilog); | |
b5256303 | 7103 | dmu_objset_disown(zd->zd_os, B_TRUE, zd); |
428870ff BB |
7104 | |
7105 | ztest_zd_fini(zd); | |
34dc7c2f BB |
7106 | } |
7107 | ||
f44ad929 TC |
7108 | /* ARGSUSED */ |
7109 | static int | |
7110 | ztest_replay_zil_cb(const char *name, void *arg) | |
7111 | { | |
7112 | objset_t *os; | |
7113 | ztest_ds_t *zdtmp; | |
7114 | ||
7115 | VERIFY0(ztest_dmu_objset_own(name, DMU_OST_ANY, B_TRUE, | |
7116 | B_TRUE, FTAG, &os)); | |
7117 | ||
7118 | zdtmp = umem_alloc(sizeof (ztest_ds_t), UMEM_NOFAIL); | |
7119 | ||
7120 | ztest_zd_init(zdtmp, NULL, os); | |
7121 | zil_replay(os, zdtmp, ztest_replay_vector); | |
7122 | ztest_zd_fini(zdtmp); | |
7123 | ||
7124 | if (dmu_objset_zil(os)->zl_parse_lr_count != 0 && | |
7125 | ztest_opts.zo_verbose >= 6) { | |
7126 | zilog_t *zilog = dmu_objset_zil(os); | |
7127 | ||
7128 | (void) printf("%s replay %llu blocks, %llu records, seq %llu\n", | |
7129 | name, | |
7130 | (u_longlong_t)zilog->zl_parse_blk_count, | |
7131 | (u_longlong_t)zilog->zl_parse_lr_count, | |
7132 | (u_longlong_t)zilog->zl_replaying_seq); | |
7133 | } | |
7134 | ||
7135 | umem_free(zdtmp, sizeof (ztest_ds_t)); | |
7136 | ||
7137 | dmu_objset_disown(os, B_TRUE, FTAG); | |
7138 | return (0); | |
7139 | } | |
7140 | ||
c1f3de18 BB |
7141 | static void |
7142 | ztest_freeze(void) | |
7143 | { | |
7144 | ztest_ds_t *zd = &ztest_ds[0]; | |
7145 | spa_t *spa; | |
7146 | int numloops = 0; | |
7147 | ||
7148 | if (ztest_opts.zo_verbose >= 3) | |
7149 | (void) printf("testing spa_freeze()...\n"); | |
7150 | ||
7151 | kernel_init(SPA_MODE_READ | SPA_MODE_WRITE); | |
a9eaae06 RM |
7152 | VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
7153 | VERIFY0(ztest_dataset_open(0)); | |
c1f3de18 BB |
7154 | ztest_spa = spa; |
7155 | ||
7156 | /* | |
7157 | * Force the first log block to be transactionally allocated. | |
7158 | * We have to do this before we freeze the pool -- otherwise | |
7159 | * the log chain won't be anchored. | |
7160 | */ | |
7161 | while (BP_IS_HOLE(&zd->zd_zilog->zl_header->zh_log)) { | |
7162 | ztest_dmu_object_alloc_free(zd, 0); | |
7163 | zil_commit(zd->zd_zilog, 0); | |
7164 | } | |
7165 | ||
7166 | txg_wait_synced(spa_get_dsl(spa), 0); | |
7167 | ||
7168 | /* | |
7169 | * Freeze the pool. This stops spa_sync() from doing anything, | |
7170 | * so that the only way to record changes from now on is the ZIL. | |
7171 | */ | |
7172 | spa_freeze(spa); | |
7173 | ||
7174 | /* | |
7175 | * Because it is hard to predict how much space a write will actually | |
7176 | * require beforehand, we leave ourselves some fudge space to write over | |
7177 | * capacity. | |
7178 | */ | |
7179 | uint64_t capacity = metaslab_class_get_space(spa_normal_class(spa)) / 2; | |
7180 | ||
7181 | /* | |
7182 | * Run tests that generate log records but don't alter the pool config | |
7183 | * or depend on DSL sync tasks (snapshots, objset create/destroy, etc). | |
7184 | * We do a txg_wait_synced() after each iteration to force the txg | |
7185 | * to increase well beyond the last synced value in the uberblock. | |
7186 | * The ZIL should be OK with that. | |
7187 | * | |
7188 | * Run a random number of times less than zo_maxloops and ensure we do | |
7189 | * not run out of space on the pool. | |
7190 | */ | |
7191 | while (ztest_random(10) != 0 && | |
7192 | numloops++ < ztest_opts.zo_maxloops && | |
7193 | metaslab_class_get_alloc(spa_normal_class(spa)) < capacity) { | |
7194 | ztest_od_t od; | |
7195 | ztest_od_init(&od, 0, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0, 0); | |
7196 | VERIFY0(ztest_object_init(zd, &od, sizeof (od), B_FALSE)); | |
7197 | ztest_io(zd, od.od_object, | |
7198 | ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
7199 | txg_wait_synced(spa_get_dsl(spa), 0); | |
7200 | } | |
7201 | ||
7202 | /* | |
7203 | * Commit all of the changes we just generated. | |
7204 | */ | |
7205 | zil_commit(zd->zd_zilog, 0); | |
7206 | txg_wait_synced(spa_get_dsl(spa), 0); | |
7207 | ||
7208 | /* | |
7209 | * Close our dataset and close the pool. | |
7210 | */ | |
7211 | ztest_dataset_close(0); | |
7212 | spa_close(spa, FTAG); | |
7213 | kernel_fini(); | |
7214 | ||
7215 | /* | |
7216 | * Open and close the pool and dataset to induce log replay. | |
7217 | */ | |
7218 | kernel_init(SPA_MODE_READ | SPA_MODE_WRITE); | |
a9eaae06 RM |
7219 | VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
7220 | ASSERT3U(spa_freeze_txg(spa), ==, UINT64_MAX); | |
7221 | VERIFY0(ztest_dataset_open(0)); | |
c1f3de18 BB |
7222 | ztest_spa = spa; |
7223 | txg_wait_synced(spa_get_dsl(spa), 0); | |
7224 | ztest_dataset_close(0); | |
7225 | ztest_reguid(NULL, 0); | |
7226 | ||
7227 | spa_close(spa, FTAG); | |
7228 | kernel_fini(); | |
7229 | } | |
7230 | ||
7231 | static void | |
7232 | ztest_import_impl(ztest_shared_t *zs) | |
7233 | { | |
7234 | importargs_t args = { 0 }; | |
7235 | nvlist_t *cfg = NULL; | |
7236 | int nsearch = 1; | |
7237 | char *searchdirs[nsearch]; | |
7238 | int flags = ZFS_IMPORT_MISSING_LOG; | |
7239 | ||
7240 | searchdirs[0] = ztest_opts.zo_dir; | |
7241 | args.paths = nsearch; | |
7242 | args.path = searchdirs; | |
7243 | args.can_be_active = B_FALSE; | |
7244 | ||
7245 | VERIFY0(zpool_find_config(NULL, ztest_opts.zo_pool, &cfg, &args, | |
7246 | &libzpool_config_ops)); | |
7247 | VERIFY0(spa_import(ztest_opts.zo_pool, cfg, NULL, flags)); | |
b6722b87 | 7248 | fnvlist_free(cfg); |
c1f3de18 BB |
7249 | } |
7250 | ||
7251 | /* | |
7252 | * Import a storage pool with the given name. | |
7253 | */ | |
7254 | static void | |
7255 | ztest_import(ztest_shared_t *zs) | |
7256 | { | |
7257 | spa_t *spa; | |
7258 | ||
7259 | mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); | |
7260 | mutex_init(&ztest_checkpoint_lock, NULL, MUTEX_DEFAULT, NULL); | |
7261 | VERIFY0(pthread_rwlock_init(&ztest_name_lock, NULL)); | |
7262 | ||
7263 | kernel_init(SPA_MODE_READ | SPA_MODE_WRITE); | |
7264 | ||
7265 | ztest_import_impl(zs); | |
7266 | ||
7267 | VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG)); | |
7268 | zs->zs_metaslab_sz = | |
7269 | 1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift; | |
7270 | spa_close(spa, FTAG); | |
7271 | ||
7272 | kernel_fini(); | |
7273 | ||
7274 | if (!ztest_opts.zo_mmp_test) { | |
7275 | ztest_run_zdb(ztest_opts.zo_pool); | |
7276 | ztest_freeze(); | |
7277 | ztest_run_zdb(ztest_opts.zo_pool); | |
7278 | } | |
7279 | ||
7280 | (void) pthread_rwlock_destroy(&ztest_name_lock); | |
7281 | mutex_destroy(&ztest_vdev_lock); | |
7282 | mutex_destroy(&ztest_checkpoint_lock); | |
7283 | } | |
7284 | ||
34dc7c2f BB |
7285 | /* |
7286 | * Kick off threads to run tests on all datasets in parallel. | |
7287 | */ | |
7288 | static void | |
428870ff | 7289 | ztest_run(ztest_shared_t *zs) |
34dc7c2f | 7290 | { |
34dc7c2f | 7291 | spa_t *spa; |
3541dc6d | 7292 | objset_t *os; |
94102578 | 7293 | kthread_t *resume_thread, *deadman_thread; |
c25b8f99 | 7294 | kthread_t **run_threads; |
1e33ac1e | 7295 | uint64_t object; |
428870ff | 7296 | int error; |
d6320ddb | 7297 | int t, d; |
b128c09f BB |
7298 | |
7299 | ztest_exiting = B_FALSE; | |
34dc7c2f | 7300 | |
34dc7c2f | 7301 | /* |
428870ff | 7302 | * Initialize parent/child shared state. |
34dc7c2f | 7303 | */ |
c242c188 | 7304 | mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); |
d2734cce | 7305 | mutex_init(&ztest_checkpoint_lock, NULL, MUTEX_DEFAULT, NULL); |
2d9142c9 | 7306 | VERIFY0(pthread_rwlock_init(&ztest_name_lock, NULL)); |
34dc7c2f | 7307 | |
428870ff | 7308 | zs->zs_thread_start = gethrtime(); |
c242c188 CS |
7309 | zs->zs_thread_stop = |
7310 | zs->zs_thread_start + ztest_opts.zo_passtime * NANOSEC; | |
428870ff BB |
7311 | zs->zs_thread_stop = MIN(zs->zs_thread_stop, zs->zs_proc_stop); |
7312 | zs->zs_thread_kill = zs->zs_thread_stop; | |
c242c188 CS |
7313 | if (ztest_random(100) < ztest_opts.zo_killrate) { |
7314 | zs->zs_thread_kill -= | |
7315 | ztest_random(ztest_opts.zo_passtime * NANOSEC); | |
7316 | } | |
34dc7c2f | 7317 | |
1e33ac1e | 7318 | mutex_init(&zcl.zcl_callbacks_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 7319 | |
428870ff BB |
7320 | list_create(&zcl.zcl_callbacks, sizeof (ztest_cb_data_t), |
7321 | offsetof(ztest_cb_data_t, zcd_node)); | |
34dc7c2f BB |
7322 | |
7323 | /* | |
c1f3de18 BB |
7324 | * Open our pool. It may need to be imported first depending on |
7325 | * what tests were running when the previous pass was terminated. | |
34dc7c2f | 7326 | */ |
da92d5cb | 7327 | kernel_init(SPA_MODE_READ | SPA_MODE_WRITE); |
c1f3de18 BB |
7328 | error = spa_open(ztest_opts.zo_pool, &spa, FTAG); |
7329 | if (error) { | |
7330 | VERIFY3S(error, ==, ENOENT); | |
7331 | ztest_import_impl(zs); | |
7332 | VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG)); | |
7333 | zs->zs_metaslab_sz = | |
7334 | 1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift; | |
7335 | } | |
7336 | ||
080b3100 | 7337 | metaslab_preload_limit = ztest_random(20) + 1; |
c242c188 | 7338 | ztest_spa = spa; |
428870ff | 7339 | |
e5db3134 BB |
7340 | VERIFY0(vdev_raidz_impl_set("cycle")); |
7341 | ||
546d32ca | 7342 | dmu_objset_stats_t dds; |
4807c0ba | 7343 | VERIFY0(ztest_dmu_objset_own(ztest_opts.zo_pool, |
b5256303 | 7344 | DMU_OST_ANY, B_TRUE, B_TRUE, FTAG, &os)); |
546d32ca BB |
7345 | dsl_pool_config_enter(dmu_objset_pool(os), FTAG); |
7346 | dmu_objset_fast_stat(os, &dds); | |
7347 | dsl_pool_config_exit(dmu_objset_pool(os), FTAG); | |
7348 | zs->zs_guid = dds.dds_guid; | |
b5256303 | 7349 | dmu_objset_disown(os, B_TRUE, FTAG); |
3541dc6d | 7350 | |
34dc7c2f | 7351 | /* |
b128c09f | 7352 | * Create a thread to periodically resume suspended I/O. |
34dc7c2f | 7353 | */ |
c25b8f99 BB |
7354 | resume_thread = thread_create(NULL, 0, ztest_resume_thread, |
7355 | spa, 0, NULL, TS_RUN | TS_JOINABLE, defclsyspri); | |
34dc7c2f | 7356 | |
428870ff | 7357 | /* |
8fb1ede1 | 7358 | * Create a deadman thread and set to panic if we hang. |
428870ff | 7359 | */ |
94102578 | 7360 | deadman_thread = thread_create(NULL, 0, ztest_deadman_thread, |
8fb1ede1 BB |
7361 | zs, 0, NULL, TS_RUN | TS_JOINABLE, defclsyspri); |
7362 | ||
7363 | spa->spa_deadman_failmode = ZIO_FAILURE_MODE_PANIC; | |
428870ff | 7364 | |
34dc7c2f | 7365 | /* |
d2734cce | 7366 | * Verify that we can safely inquire about any object, |
34dc7c2f BB |
7367 | * whether it's allocated or not. To make it interesting, |
7368 | * we probe a 5-wide window around each power of two. | |
7369 | * This hits all edge cases, including zero and the max. | |
7370 | */ | |
d6320ddb BB |
7371 | for (t = 0; t < 64; t++) { |
7372 | for (d = -5; d <= 5; d++) { | |
34dc7c2f BB |
7373 | error = dmu_object_info(spa->spa_meta_objset, |
7374 | (1ULL << t) + d, NULL); | |
7375 | ASSERT(error == 0 || error == ENOENT || | |
7376 | error == EINVAL); | |
7377 | } | |
7378 | } | |
7379 | ||
7380 | /* | |
428870ff | 7381 | * If we got any ENOSPC errors on the previous run, destroy something. |
34dc7c2f | 7382 | */ |
428870ff | 7383 | if (zs->zs_enospc_count != 0) { |
c242c188 CS |
7384 | int d = ztest_random(ztest_opts.zo_datasets); |
7385 | ztest_dataset_destroy(d); | |
428870ff | 7386 | } |
34dc7c2f BB |
7387 | zs->zs_enospc_count = 0; |
7388 | ||
c40a1124 TC |
7389 | /* |
7390 | * If we were in the middle of ztest_device_removal() and were killed | |
7391 | * we need to ensure the removal and scrub complete before running | |
7392 | * any tests that check ztest_device_removal_active. The removal will | |
7393 | * be restarted automatically when the spa is opened, but we need to | |
d611989f | 7394 | * initiate the scrub manually if it is not already in progress. Note |
c40a1124 TC |
7395 | * that we always run the scrub whenever an indirect vdev exists |
7396 | * because we have no way of knowing for sure if ztest_device_removal() | |
7397 | * fully completed its scrub before the pool was reimported. | |
7398 | */ | |
7399 | if (spa->spa_removing_phys.sr_state == DSS_SCANNING || | |
7400 | spa->spa_removing_phys.sr_prev_indirect_vdev != -1) { | |
7401 | while (spa->spa_removing_phys.sr_state == DSS_SCANNING) | |
7402 | txg_wait_synced(spa_get_dsl(spa), 0); | |
7403 | ||
ce5fb2a7 BB |
7404 | error = ztest_scrub_impl(spa); |
7405 | if (error == EBUSY) | |
7406 | error = 0; | |
7407 | ASSERT0(error); | |
c40a1124 TC |
7408 | } |
7409 | ||
c25b8f99 | 7410 | run_threads = umem_zalloc(ztest_opts.zo_threads * sizeof (kthread_t *), |
c242c188 | 7411 | UMEM_NOFAIL); |
34dc7c2f | 7412 | |
c242c188 | 7413 | if (ztest_opts.zo_verbose >= 4) |
34dc7c2f BB |
7414 | (void) printf("starting main threads...\n"); |
7415 | ||
f44ad929 TC |
7416 | /* |
7417 | * Replay all logs of all datasets in the pool. This is primarily for | |
7418 | * temporary datasets which wouldn't otherwise get replayed, which | |
7419 | * can trigger failures when attempting to offline a SLOG in | |
7420 | * ztest_fault_inject(). | |
7421 | */ | |
7422 | (void) dmu_objset_find(ztest_opts.zo_pool, ztest_replay_zil_cb, | |
7423 | NULL, DS_FIND_CHILDREN); | |
7424 | ||
428870ff BB |
7425 | /* |
7426 | * Kick off all the tests that run in parallel. | |
7427 | */ | |
c242c188 | 7428 | for (t = 0; t < ztest_opts.zo_threads; t++) { |
c25b8f99 BB |
7429 | if (t < ztest_opts.zo_datasets && ztest_dataset_open(t) != 0) { |
7430 | umem_free(run_threads, ztest_opts.zo_threads * | |
7431 | sizeof (kthread_t *)); | |
428870ff | 7432 | return; |
06cf4d98 | 7433 | } |
1e33ac1e | 7434 | |
c25b8f99 BB |
7435 | run_threads[t] = thread_create(NULL, 0, ztest_thread, |
7436 | (void *)(uintptr_t)t, 0, NULL, TS_RUN | TS_JOINABLE, | |
7437 | defclsyspri); | |
34dc7c2f BB |
7438 | } |
7439 | ||
428870ff | 7440 | /* |
a2d88f77 | 7441 | * Wait for all of the tests to complete. |
428870ff | 7442 | */ |
a2d88f77 | 7443 | for (t = 0; t < ztest_opts.zo_threads; t++) |
c25b8f99 | 7444 | VERIFY0(thread_join(run_threads[t])); |
a2d88f77 TC |
7445 | |
7446 | /* | |
7447 | * Close all datasets. This must be done after all the threads | |
7448 | * are joined so we can be sure none of the datasets are in-use | |
7449 | * by any of the threads. | |
7450 | */ | |
7451 | for (t = 0; t < ztest_opts.zo_threads; t++) { | |
c242c188 CS |
7452 | if (t < ztest_opts.zo_datasets) |
7453 | ztest_dataset_close(t); | |
34dc7c2f BB |
7454 | } |
7455 | ||
34dc7c2f BB |
7456 | txg_wait_synced(spa_get_dsl(spa), 0); |
7457 | ||
428870ff BB |
7458 | zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(spa)); |
7459 | zs->zs_space = metaslab_class_get_space(spa_normal_class(spa)); | |
7460 | ||
c25b8f99 | 7461 | umem_free(run_threads, ztest_opts.zo_threads * sizeof (kthread_t *)); |
428870ff | 7462 | |
94102578 | 7463 | /* Kill the resume and deadman threads */ |
428870ff | 7464 | ztest_exiting = B_TRUE; |
c25b8f99 | 7465 | VERIFY0(thread_join(resume_thread)); |
94102578 | 7466 | VERIFY0(thread_join(deadman_thread)); |
428870ff | 7467 | ztest_resume(spa); |
34dc7c2f BB |
7468 | |
7469 | /* | |
428870ff BB |
7470 | * Right before closing the pool, kick off a bunch of async I/O; |
7471 | * spa_close() should wait for it to complete. | |
34dc7c2f | 7472 | */ |
fcff0f35 PD |
7473 | for (object = 1; object < 50; object++) { |
7474 | dmu_prefetch(spa->spa_meta_objset, object, 0, 0, 1ULL << 20, | |
7475 | ZIO_PRIORITY_SYNC_READ); | |
7476 | } | |
428870ff | 7477 | |
090ff092 RC |
7478 | /* Verify that at least one commit cb was called in a timely fashion */ |
7479 | if (zc_cb_counter >= ZTEST_COMMIT_CB_MIN_REG) | |
c99c9001 | 7480 | VERIFY0(zc_min_txg_delay); |
090ff092 | 7481 | |
428870ff BB |
7482 | spa_close(spa, FTAG); |
7483 | ||
7484 | /* | |
7485 | * Verify that we can loop over all pools. | |
7486 | */ | |
7487 | mutex_enter(&spa_namespace_lock); | |
7488 | for (spa = spa_next(NULL); spa != NULL; spa = spa_next(spa)) | |
c242c188 | 7489 | if (ztest_opts.zo_verbose > 3) |
428870ff BB |
7490 | (void) printf("spa_next: found %s\n", spa_name(spa)); |
7491 | mutex_exit(&spa_namespace_lock); | |
7492 | ||
7493 | /* | |
7494 | * Verify that we can export the pool and reimport it under a | |
7495 | * different name. | |
7496 | */ | |
379ca9cf | 7497 | if ((ztest_random(2) == 0) && !ztest_opts.zo_mmp_test) { |
eca7b760 IK |
7498 | char name[ZFS_MAX_DATASET_NAME_LEN]; |
7499 | (void) snprintf(name, sizeof (name), "%s_import", | |
c242c188 CS |
7500 | ztest_opts.zo_pool); |
7501 | ztest_spa_import_export(ztest_opts.zo_pool, name); | |
7502 | ztest_spa_import_export(name, ztest_opts.zo_pool); | |
428870ff BB |
7503 | } |
7504 | ||
7505 | kernel_fini(); | |
572e2857 BB |
7506 | |
7507 | list_destroy(&zcl.zcl_callbacks); | |
0e8d1b2d | 7508 | mutex_destroy(&zcl.zcl_callbacks_lock); |
2d9142c9 | 7509 | (void) pthread_rwlock_destroy(&ztest_name_lock); |
c242c188 | 7510 | mutex_destroy(&ztest_vdev_lock); |
d2734cce | 7511 | mutex_destroy(&ztest_checkpoint_lock); |
428870ff BB |
7512 | } |
7513 | ||
65c7cc49 | 7514 | static void |
34dc7c2f BB |
7515 | print_time(hrtime_t t, char *timebuf) |
7516 | { | |
7517 | hrtime_t s = t / NANOSEC; | |
7518 | hrtime_t m = s / 60; | |
7519 | hrtime_t h = m / 60; | |
7520 | hrtime_t d = h / 24; | |
7521 | ||
7522 | s -= m * 60; | |
7523 | m -= h * 60; | |
7524 | h -= d * 24; | |
7525 | ||
7526 | timebuf[0] = '\0'; | |
7527 | ||
7528 | if (d) | |
7529 | (void) sprintf(timebuf, | |
7530 | "%llud%02lluh%02llum%02llus", d, h, m, s); | |
7531 | else if (h) | |
7532 | (void) sprintf(timebuf, "%lluh%02llum%02llus", h, m, s); | |
7533 | else if (m) | |
7534 | (void) sprintf(timebuf, "%llum%02llus", m, s); | |
7535 | else | |
7536 | (void) sprintf(timebuf, "%llus", s); | |
7537 | } | |
7538 | ||
428870ff | 7539 | static nvlist_t * |
0bc8fd78 | 7540 | make_random_props(void) |
428870ff BB |
7541 | { |
7542 | nvlist_t *props; | |
7543 | ||
bea3fc71 | 7544 | props = fnvlist_alloc(); |
cc99f275 | 7545 | |
9ae529ec CS |
7546 | if (ztest_random(2) == 0) |
7547 | return (props); | |
60b82074 | 7548 | |
bea3fc71 RM |
7549 | fnvlist_add_uint64(props, |
7550 | zpool_prop_to_name(ZPOOL_PROP_AUTOREPLACE), 1); | |
428870ff | 7551 | |
428870ff BB |
7552 | return (props); |
7553 | } | |
7554 | ||
34dc7c2f BB |
7555 | /* |
7556 | * Create a storage pool with the given name and initial vdev size. | |
428870ff | 7557 | * Then test spa_freeze() functionality. |
34dc7c2f BB |
7558 | */ |
7559 | static void | |
428870ff | 7560 | ztest_init(ztest_shared_t *zs) |
34dc7c2f BB |
7561 | { |
7562 | spa_t *spa; | |
428870ff | 7563 | nvlist_t *nvroot, *props; |
9ae529ec | 7564 | int i; |
428870ff | 7565 | |
c242c188 | 7566 | mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); |
d2734cce | 7567 | mutex_init(&ztest_checkpoint_lock, NULL, MUTEX_DEFAULT, NULL); |
2d9142c9 | 7568 | VERIFY0(pthread_rwlock_init(&ztest_name_lock, NULL)); |
34dc7c2f | 7569 | |
da92d5cb | 7570 | kernel_init(SPA_MODE_READ | SPA_MODE_WRITE); |
34dc7c2f BB |
7571 | |
7572 | /* | |
7573 | * Create the storage pool. | |
7574 | */ | |
c242c188 | 7575 | (void) spa_destroy(ztest_opts.zo_pool); |
428870ff BB |
7576 | ztest_shared->zs_vdev_next_leaf = 0; |
7577 | zs->zs_splits = 0; | |
c242c188 | 7578 | zs->zs_mirrors = ztest_opts.zo_mirrors; |
ea0b2538 | 7579 | nvroot = make_vdev_root(NULL, NULL, NULL, ztest_opts.zo_vdev_size, 0, |
b2255edc | 7580 | NULL, ztest_opts.zo_raid_children, zs->zs_mirrors, 1); |
428870ff | 7581 | props = make_random_props(); |
60b82074 BB |
7582 | |
7583 | /* | |
7584 | * We don't expect the pool to suspend unless maxfaults == 0, | |
7585 | * in which case ztest_fault_inject() temporarily takes away | |
7586 | * the only valid replica. | |
7587 | */ | |
bea3fc71 | 7588 | fnvlist_add_uint64(props, |
60b82074 | 7589 | zpool_prop_to_name(ZPOOL_PROP_FAILUREMODE), |
bea3fc71 | 7590 | MAXFAULTS(zs) ? ZIO_FAILURE_MODE_PANIC : ZIO_FAILURE_MODE_WAIT); |
60b82074 | 7591 | |
9ae529ec CS |
7592 | for (i = 0; i < SPA_FEATURES; i++) { |
7593 | char *buf; | |
93e28d66 | 7594 | |
03ef8f09 MM |
7595 | if (!spa_feature_table[i].fi_zfs_mod_supported) |
7596 | continue; | |
7597 | ||
93e28d66 SD |
7598 | /* |
7599 | * 75% chance of using the log space map feature. We want ztest | |
7600 | * to exercise both the code paths that use the log space map | |
7601 | * feature and the ones that don't. | |
7602 | */ | |
7603 | if (i == SPA_FEATURE_LOG_SPACEMAP && ztest_random(4) == 0) | |
7604 | continue; | |
7605 | ||
9ae529ec CS |
7606 | VERIFY3S(-1, !=, asprintf(&buf, "feature@%s", |
7607 | spa_feature_table[i].fi_uname)); | |
bea3fc71 | 7608 | fnvlist_add_uint64(props, buf, 0); |
9ae529ec CS |
7609 | free(buf); |
7610 | } | |
60b82074 BB |
7611 | |
7612 | VERIFY0(spa_create(ztest_opts.zo_pool, nvroot, props, NULL, NULL)); | |
bea3fc71 RM |
7613 | fnvlist_free(nvroot); |
7614 | fnvlist_free(props); | |
34dc7c2f | 7615 | |
a9eaae06 | 7616 | VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
c242c188 CS |
7617 | zs->zs_metaslab_sz = |
7618 | 1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift; | |
34dc7c2f BB |
7619 | spa_close(spa, FTAG); |
7620 | ||
7621 | kernel_fini(); | |
428870ff | 7622 | |
379ca9cf OF |
7623 | if (!ztest_opts.zo_mmp_test) { |
7624 | ztest_run_zdb(ztest_opts.zo_pool); | |
7625 | ztest_freeze(); | |
7626 | ztest_run_zdb(ztest_opts.zo_pool); | |
7627 | } | |
c242c188 | 7628 | |
2d9142c9 | 7629 | (void) pthread_rwlock_destroy(&ztest_name_lock); |
c242c188 | 7630 | mutex_destroy(&ztest_vdev_lock); |
d2734cce | 7631 | mutex_destroy(&ztest_checkpoint_lock); |
c242c188 CS |
7632 | } |
7633 | ||
7634 | static void | |
9d81146b | 7635 | setup_data_fd(void) |
c242c188 | 7636 | { |
facbbe43 BB |
7637 | static char ztest_name_data[] = "/tmp/ztest.data.XXXXXX"; |
7638 | ||
7639 | ztest_fd_data = mkstemp(ztest_name_data); | |
9d81146b | 7640 | ASSERT3S(ztest_fd_data, >=, 0); |
facbbe43 | 7641 | (void) unlink(ztest_name_data); |
c242c188 CS |
7642 | } |
7643 | ||
22257dc0 CS |
7644 | static int |
7645 | shared_data_size(ztest_shared_hdr_t *hdr) | |
7646 | { | |
7647 | int size; | |
7648 | ||
7649 | size = hdr->zh_hdr_size; | |
7650 | size += hdr->zh_opts_size; | |
7651 | size += hdr->zh_size; | |
7652 | size += hdr->zh_stats_size * hdr->zh_stats_count; | |
7653 | size += hdr->zh_ds_size * hdr->zh_ds_count; | |
7654 | ||
7655 | return (size); | |
7656 | } | |
7657 | ||
c242c188 CS |
7658 | static void |
7659 | setup_hdr(void) | |
7660 | { | |
22257dc0 | 7661 | int size; |
c242c188 CS |
7662 | ztest_shared_hdr_t *hdr; |
7663 | ||
7664 | hdr = (void *)mmap(0, P2ROUNDUP(sizeof (*hdr), getpagesize()), | |
9d81146b | 7665 | PROT_READ | PROT_WRITE, MAP_SHARED, ztest_fd_data, 0); |
a9eaae06 | 7666 | ASSERT3P(hdr, !=, MAP_FAILED); |
c242c188 | 7667 | |
a9eaae06 | 7668 | VERIFY0(ftruncate(ztest_fd_data, sizeof (ztest_shared_hdr_t))); |
22257dc0 | 7669 | |
c242c188 CS |
7670 | hdr->zh_hdr_size = sizeof (ztest_shared_hdr_t); |
7671 | hdr->zh_opts_size = sizeof (ztest_shared_opts_t); | |
7672 | hdr->zh_size = sizeof (ztest_shared_t); | |
7673 | hdr->zh_stats_size = sizeof (ztest_shared_callstate_t); | |
7674 | hdr->zh_stats_count = ZTEST_FUNCS; | |
7675 | hdr->zh_ds_size = sizeof (ztest_shared_ds_t); | |
7676 | hdr->zh_ds_count = ztest_opts.zo_datasets; | |
7677 | ||
22257dc0 | 7678 | size = shared_data_size(hdr); |
a9eaae06 | 7679 | VERIFY0(ftruncate(ztest_fd_data, size)); |
22257dc0 | 7680 | |
c242c188 CS |
7681 | (void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize())); |
7682 | } | |
7683 | ||
7684 | static void | |
7685 | setup_data(void) | |
7686 | { | |
7687 | int size, offset; | |
7688 | ztest_shared_hdr_t *hdr; | |
7689 | uint8_t *buf; | |
7690 | ||
7691 | hdr = (void *)mmap(0, P2ROUNDUP(sizeof (*hdr), getpagesize()), | |
9d81146b | 7692 | PROT_READ, MAP_SHARED, ztest_fd_data, 0); |
a9eaae06 | 7693 | ASSERT3P(hdr, !=, MAP_FAILED); |
c242c188 | 7694 | |
22257dc0 | 7695 | size = shared_data_size(hdr); |
c242c188 CS |
7696 | |
7697 | (void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize())); | |
7698 | hdr = ztest_shared_hdr = (void *)mmap(0, P2ROUNDUP(size, getpagesize()), | |
9d81146b | 7699 | PROT_READ | PROT_WRITE, MAP_SHARED, ztest_fd_data, 0); |
a9eaae06 | 7700 | ASSERT3P(hdr, !=, MAP_FAILED); |
c242c188 CS |
7701 | buf = (uint8_t *)hdr; |
7702 | ||
7703 | offset = hdr->zh_hdr_size; | |
7704 | ztest_shared_opts = (void *)&buf[offset]; | |
7705 | offset += hdr->zh_opts_size; | |
7706 | ztest_shared = (void *)&buf[offset]; | |
7707 | offset += hdr->zh_size; | |
7708 | ztest_shared_callstate = (void *)&buf[offset]; | |
7709 | offset += hdr->zh_stats_size * hdr->zh_stats_count; | |
7710 | ztest_shared_ds = (void *)&buf[offset]; | |
7711 | } | |
7712 | ||
7713 | static boolean_t | |
7714 | exec_child(char *cmd, char *libpath, boolean_t ignorekill, int *statusp) | |
7715 | { | |
7716 | pid_t pid; | |
7717 | int status; | |
483106eb | 7718 | char *cmdbuf = NULL; |
c242c188 CS |
7719 | |
7720 | pid = fork(); | |
7721 | ||
7722 | if (cmd == NULL) { | |
483106eb BB |
7723 | cmdbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
7724 | (void) strlcpy(cmdbuf, getexecname(), MAXPATHLEN); | |
c242c188 CS |
7725 | cmd = cmdbuf; |
7726 | } | |
7727 | ||
7728 | if (pid == -1) | |
7729 | fatal(1, "fork failed"); | |
7730 | ||
7731 | if (pid == 0) { /* child */ | |
7732 | char *emptyargv[2] = { cmd, NULL }; | |
9d81146b | 7733 | char fd_data_str[12]; |
c242c188 CS |
7734 | |
7735 | struct rlimit rl = { 1024, 1024 }; | |
7736 | (void) setrlimit(RLIMIT_NOFILE, &rl); | |
9d81146b ED |
7737 | |
7738 | (void) close(ztest_fd_rand); | |
a9eaae06 RM |
7739 | VERIFY3S(11, >=, |
7740 | snprintf(fd_data_str, 12, "%d", ztest_fd_data)); | |
7741 | VERIFY0(setenv("ZTEST_FD_DATA", fd_data_str, 1)); | |
9d81146b | 7742 | |
c242c188 CS |
7743 | (void) enable_extended_FILE_stdio(-1, -1); |
7744 | if (libpath != NULL) | |
a9eaae06 | 7745 | VERIFY0(setenv("LD_LIBRARY_PATH", libpath, 1)); |
c242c188 CS |
7746 | (void) execv(cmd, emptyargv); |
7747 | ztest_dump_core = B_FALSE; | |
7748 | fatal(B_TRUE, "exec failed: %s", cmd); | |
7749 | } | |
7750 | ||
483106eb BB |
7751 | if (cmdbuf != NULL) { |
7752 | umem_free(cmdbuf, MAXPATHLEN); | |
7753 | cmd = NULL; | |
7754 | } | |
7755 | ||
c242c188 CS |
7756 | while (waitpid(pid, &status, 0) != pid) |
7757 | continue; | |
7758 | if (statusp != NULL) | |
7759 | *statusp = status; | |
7760 | ||
7761 | if (WIFEXITED(status)) { | |
7762 | if (WEXITSTATUS(status) != 0) { | |
7763 | (void) fprintf(stderr, "child exited with code %d\n", | |
7764 | WEXITSTATUS(status)); | |
7765 | exit(2); | |
7766 | } | |
7767 | return (B_FALSE); | |
7768 | } else if (WIFSIGNALED(status)) { | |
7769 | if (!ignorekill || WTERMSIG(status) != SIGKILL) { | |
7770 | (void) fprintf(stderr, "child died with signal %d\n", | |
7771 | WTERMSIG(status)); | |
7772 | exit(3); | |
7773 | } | |
7774 | return (B_TRUE); | |
7775 | } else { | |
7776 | (void) fprintf(stderr, "something strange happened to child\n"); | |
7777 | exit(4); | |
7778 | /* NOTREACHED */ | |
7779 | } | |
7780 | } | |
7781 | ||
7782 | static void | |
7783 | ztest_run_init(void) | |
7784 | { | |
7785 | int i; | |
428870ff | 7786 | |
c242c188 | 7787 | ztest_shared_t *zs = ztest_shared; |
428870ff | 7788 | |
c242c188 CS |
7789 | /* |
7790 | * Blow away any existing copy of zpool.cache | |
7791 | */ | |
7792 | (void) remove(spa_config_path); | |
7793 | ||
379ca9cf OF |
7794 | if (ztest_opts.zo_init == 0) { |
7795 | if (ztest_opts.zo_verbose >= 1) | |
7796 | (void) printf("Importing pool %s\n", | |
7797 | ztest_opts.zo_pool); | |
7798 | ztest_import(zs); | |
7799 | return; | |
7800 | } | |
7801 | ||
c242c188 CS |
7802 | /* |
7803 | * Create and initialize our storage pool. | |
7804 | */ | |
7805 | for (i = 1; i <= ztest_opts.zo_init; i++) { | |
7806 | bzero(zs, sizeof (ztest_shared_t)); | |
7807 | if (ztest_opts.zo_verbose >= 3 && | |
7808 | ztest_opts.zo_init != 1) { | |
7809 | (void) printf("ztest_init(), pass %d\n", i); | |
7810 | } | |
7811 | ztest_init(zs); | |
7812 | } | |
34dc7c2f BB |
7813 | } |
7814 | ||
7815 | int | |
7816 | main(int argc, char **argv) | |
7817 | { | |
7818 | int kills = 0; | |
7819 | int iters = 0; | |
c242c188 CS |
7820 | int older = 0; |
7821 | int newer = 0; | |
34dc7c2f BB |
7822 | ztest_shared_t *zs; |
7823 | ztest_info_t *zi; | |
c242c188 | 7824 | ztest_shared_callstate_t *zc; |
34dc7c2f | 7825 | char timebuf[100]; |
f3c8c9e6 | 7826 | char numbuf[NN_NUMBUF_SZ]; |
483106eb | 7827 | char *cmd; |
c242c188 | 7828 | boolean_t hasalt; |
2801e68d | 7829 | int f, err; |
9d81146b | 7830 | char *fd_data_str = getenv("ZTEST_FD_DATA"); |
e82cdc3a | 7831 | struct sigaction action; |
34dc7c2f BB |
7832 | |
7833 | (void) setvbuf(stdout, NULL, _IOLBF, 0); | |
7834 | ||
498877ba | 7835 | dprintf_setup(&argc, argv); |
8fb1ede1 | 7836 | zfs_deadman_synctime_ms = 300000; |
94102578 | 7837 | zfs_deadman_checktime_ms = 30000; |
4d044c4c SD |
7838 | /* |
7839 | * As two-word space map entries may not come up often (especially | |
7840 | * if pool and vdev sizes are small) we want to force at least some | |
7841 | * of them so the feature get tested. | |
7842 | */ | |
7843 | zfs_force_some_double_word_sm_entries = B_TRUE; | |
498877ba | 7844 | |
1258bd77 BB |
7845 | /* |
7846 | * Verify that even extensively damaged split blocks with many | |
7847 | * segments can be reconstructed in a reasonable amount of time | |
7848 | * when reconstruction is known to be possible. | |
4a7eb69a TC |
7849 | * |
7850 | * Note: the lower this value is, the more damage we inflict, and | |
7851 | * the more time ztest spends in recovering that damage. We chose | |
7852 | * to induce damage 1/100th of the time so recovery is tested but | |
7853 | * not so frequently that ztest doesn't get to test other code paths. | |
1258bd77 | 7854 | */ |
4a7eb69a | 7855 | zfs_reconstruct_indirect_damage_fraction = 100; |
1258bd77 | 7856 | |
e82cdc3a NB |
7857 | action.sa_handler = sig_handler; |
7858 | sigemptyset(&action.sa_mask); | |
7859 | action.sa_flags = 0; | |
7860 | ||
7861 | if (sigaction(SIGSEGV, &action, NULL) < 0) { | |
7862 | (void) fprintf(stderr, "ztest: cannot catch SIGSEGV: %s.\n", | |
7863 | strerror(errno)); | |
7864 | exit(EXIT_FAILURE); | |
7865 | } | |
7866 | ||
7867 | if (sigaction(SIGABRT, &action, NULL) < 0) { | |
7868 | (void) fprintf(stderr, "ztest: cannot catch SIGABRT: %s.\n", | |
7869 | strerror(errno)); | |
7870 | exit(EXIT_FAILURE); | |
7871 | } | |
7872 | ||
e1a0850c BB |
7873 | /* |
7874 | * Force random_get_bytes() to use /dev/urandom in order to prevent | |
7875 | * ztest from needlessly depleting the system entropy pool. | |
7876 | */ | |
7877 | random_path = "/dev/urandom"; | |
7878 | ztest_fd_rand = open(random_path, O_RDONLY); | |
9d81146b ED |
7879 | ASSERT3S(ztest_fd_rand, >=, 0); |
7880 | ||
7881 | if (!fd_data_str) { | |
c242c188 | 7882 | process_options(argc, argv); |
34dc7c2f | 7883 | |
9d81146b | 7884 | setup_data_fd(); |
c242c188 CS |
7885 | setup_hdr(); |
7886 | setup_data(); | |
7887 | bcopy(&ztest_opts, ztest_shared_opts, | |
7888 | sizeof (*ztest_shared_opts)); | |
7889 | } else { | |
9d81146b | 7890 | ztest_fd_data = atoi(fd_data_str); |
c242c188 CS |
7891 | setup_data(); |
7892 | bcopy(ztest_shared_opts, &ztest_opts, sizeof (ztest_opts)); | |
7893 | } | |
7894 | ASSERT3U(ztest_opts.zo_datasets, ==, ztest_shared_hdr->zh_ds_count); | |
34dc7c2f | 7895 | |
2801e68d CS |
7896 | err = ztest_set_global_vars(); |
7897 | if (err != 0 && !fd_data_str) { | |
7898 | /* error message done by ztest_set_global_vars */ | |
7899 | exit(EXIT_FAILURE); | |
7900 | } else { | |
7901 | /* children should not be spawned if setting gvars fails */ | |
7902 | VERIFY3S(err, ==, 0); | |
7903 | } | |
7904 | ||
428870ff | 7905 | /* Override location of zpool.cache */ |
a9eaae06 RM |
7906 | VERIFY3S(asprintf((char **)&spa_config_path, "%s/zpool.cache", |
7907 | ztest_opts.zo_dir), !=, -1); | |
9d81146b | 7908 | |
c242c188 CS |
7909 | ztest_ds = umem_alloc(ztest_opts.zo_datasets * sizeof (ztest_ds_t), |
7910 | UMEM_NOFAIL); | |
7911 | zs = ztest_shared; | |
7912 | ||
9d81146b | 7913 | if (fd_data_str) { |
d830d479 | 7914 | metaslab_force_ganging = ztest_opts.zo_metaslab_force_ganging; |
c242c188 CS |
7915 | metaslab_df_alloc_threshold = |
7916 | zs->zs_metaslab_df_alloc_threshold; | |
7917 | ||
7918 | if (zs->zs_do_init) | |
7919 | ztest_run_init(); | |
7920 | else | |
7921 | ztest_run(zs); | |
7922 | exit(0); | |
7923 | } | |
34dc7c2f | 7924 | |
c242c188 | 7925 | hasalt = (strlen(ztest_opts.zo_alt_ztest) != 0); |
34dc7c2f | 7926 | |
c242c188 | 7927 | if (ztest_opts.zo_verbose >= 1) { |
34dc7c2f | 7928 | (void) printf("%llu vdevs, %d datasets, %d threads," |
b2255edc | 7929 | "%d %s disks, %llu seconds...\n\n", |
c242c188 CS |
7930 | (u_longlong_t)ztest_opts.zo_vdevs, |
7931 | ztest_opts.zo_datasets, | |
7932 | ztest_opts.zo_threads, | |
b2255edc BB |
7933 | ztest_opts.zo_raid_children, |
7934 | ztest_opts.zo_raid_type, | |
c242c188 | 7935 | (u_longlong_t)ztest_opts.zo_time); |
34dc7c2f BB |
7936 | } |
7937 | ||
483106eb BB |
7938 | cmd = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); |
7939 | (void) strlcpy(cmd, getexecname(), MAXNAMELEN); | |
c242c188 CS |
7940 | |
7941 | zs->zs_do_init = B_TRUE; | |
7942 | if (strlen(ztest_opts.zo_alt_ztest) != 0) { | |
7943 | if (ztest_opts.zo_verbose >= 1) { | |
7944 | (void) printf("Executing older ztest for " | |
7945 | "initialization: %s\n", ztest_opts.zo_alt_ztest); | |
7946 | } | |
7947 | VERIFY(!exec_child(ztest_opts.zo_alt_ztest, | |
7948 | ztest_opts.zo_alt_libpath, B_FALSE, NULL)); | |
7949 | } else { | |
7950 | VERIFY(!exec_child(NULL, NULL, B_FALSE, NULL)); | |
34dc7c2f | 7951 | } |
c242c188 | 7952 | zs->zs_do_init = B_FALSE; |
34dc7c2f | 7953 | |
428870ff | 7954 | zs->zs_proc_start = gethrtime(); |
c242c188 | 7955 | zs->zs_proc_stop = zs->zs_proc_start + ztest_opts.zo_time * NANOSEC; |
34dc7c2f | 7956 | |
d6320ddb | 7957 | for (f = 0; f < ZTEST_FUNCS; f++) { |
c242c188 CS |
7958 | zi = &ztest_info[f]; |
7959 | zc = ZTEST_GET_SHARED_CALLSTATE(f); | |
428870ff | 7960 | if (zs->zs_proc_start + zi->zi_interval[0] > zs->zs_proc_stop) |
c242c188 | 7961 | zc->zc_next = UINT64_MAX; |
34dc7c2f | 7962 | else |
c242c188 | 7963 | zc->zc_next = zs->zs_proc_start + |
428870ff | 7964 | ztest_random(2 * zi->zi_interval[0] + 1); |
34dc7c2f BB |
7965 | } |
7966 | ||
34dc7c2f BB |
7967 | /* |
7968 | * Run the tests in a loop. These tests include fault injection | |
7969 | * to verify that self-healing data works, and forced crashes | |
7970 | * to verify that we never lose on-disk consistency. | |
7971 | */ | |
428870ff | 7972 | while (gethrtime() < zs->zs_proc_stop) { |
34dc7c2f | 7973 | int status; |
c242c188 | 7974 | boolean_t killed; |
34dc7c2f BB |
7975 | |
7976 | /* | |
7977 | * Initialize the workload counters for each function. | |
7978 | */ | |
d6320ddb | 7979 | for (f = 0; f < ZTEST_FUNCS; f++) { |
c242c188 CS |
7980 | zc = ZTEST_GET_SHARED_CALLSTATE(f); |
7981 | zc->zc_count = 0; | |
7982 | zc->zc_time = 0; | |
34dc7c2f BB |
7983 | } |
7984 | ||
9babb374 | 7985 | /* Set the allocation switch size */ |
c242c188 CS |
7986 | zs->zs_metaslab_df_alloc_threshold = |
7987 | ztest_random(zs->zs_metaslab_sz / 4) + 1; | |
9babb374 | 7988 | |
c242c188 CS |
7989 | if (!hasalt || ztest_random(2) == 0) { |
7990 | if (hasalt && ztest_opts.zo_verbose >= 1) { | |
7991 | (void) printf("Executing newer ztest: %s\n", | |
7992 | cmd); | |
34dc7c2f | 7993 | } |
c242c188 CS |
7994 | newer++; |
7995 | killed = exec_child(cmd, NULL, B_TRUE, &status); | |
34dc7c2f | 7996 | } else { |
c242c188 CS |
7997 | if (hasalt && ztest_opts.zo_verbose >= 1) { |
7998 | (void) printf("Executing older ztest: %s\n", | |
7999 | ztest_opts.zo_alt_ztest); | |
8000 | } | |
8001 | older++; | |
8002 | killed = exec_child(ztest_opts.zo_alt_ztest, | |
8003 | ztest_opts.zo_alt_libpath, B_TRUE, &status); | |
34dc7c2f BB |
8004 | } |
8005 | ||
c242c188 CS |
8006 | if (killed) |
8007 | kills++; | |
34dc7c2f BB |
8008 | iters++; |
8009 | ||
c242c188 | 8010 | if (ztest_opts.zo_verbose >= 1) { |
34dc7c2f BB |
8011 | hrtime_t now = gethrtime(); |
8012 | ||
428870ff BB |
8013 | now = MIN(now, zs->zs_proc_stop); |
8014 | print_time(zs->zs_proc_stop - now, timebuf); | |
f3c8c9e6 | 8015 | nicenum(zs->zs_space, numbuf, sizeof (numbuf)); |
34dc7c2f BB |
8016 | |
8017 | (void) printf("Pass %3d, %8s, %3llu ENOSPC, " | |
8018 | "%4.1f%% of %5s used, %3.0f%% done, %8s to go\n", | |
8019 | iters, | |
8020 | WIFEXITED(status) ? "Complete" : "SIGKILL", | |
8021 | (u_longlong_t)zs->zs_enospc_count, | |
8022 | 100.0 * zs->zs_alloc / zs->zs_space, | |
8023 | numbuf, | |
428870ff | 8024 | 100.0 * (now - zs->zs_proc_start) / |
c242c188 | 8025 | (ztest_opts.zo_time * NANOSEC), timebuf); |
34dc7c2f BB |
8026 | } |
8027 | ||
c242c188 | 8028 | if (ztest_opts.zo_verbose >= 2) { |
34dc7c2f BB |
8029 | (void) printf("\nWorkload summary:\n\n"); |
8030 | (void) printf("%7s %9s %s\n", | |
8031 | "Calls", "Time", "Function"); | |
8032 | (void) printf("%7s %9s %s\n", | |
8033 | "-----", "----", "--------"); | |
d6320ddb | 8034 | for (f = 0; f < ZTEST_FUNCS; f++) { |
c242c188 CS |
8035 | zi = &ztest_info[f]; |
8036 | zc = ZTEST_GET_SHARED_CALLSTATE(f); | |
8037 | print_time(zc->zc_time, timebuf); | |
34dc7c2f | 8038 | (void) printf("%7llu %9s %s\n", |
c242c188 | 8039 | (u_longlong_t)zc->zc_count, timebuf, |
fdc5d982 | 8040 | zi->zi_funcname); |
34dc7c2f BB |
8041 | } |
8042 | (void) printf("\n"); | |
8043 | } | |
8044 | ||
379ca9cf OF |
8045 | if (!ztest_opts.zo_mmp_test) |
8046 | ztest_run_zdb(ztest_opts.zo_pool); | |
428870ff | 8047 | } |
34dc7c2f | 8048 | |
c242c188 CS |
8049 | if (ztest_opts.zo_verbose >= 1) { |
8050 | if (hasalt) { | |
8051 | (void) printf("%d runs of older ztest: %s\n", older, | |
8052 | ztest_opts.zo_alt_ztest); | |
8053 | (void) printf("%d runs of newer ztest: %s\n", newer, | |
8054 | cmd); | |
8055 | } | |
34dc7c2f BB |
8056 | (void) printf("%d killed, %d completed, %.0f%% kill rate\n", |
8057 | kills, iters - kills, (100.0 * kills) / MAX(1, iters)); | |
8058 | } | |
8059 | ||
483106eb BB |
8060 | umem_free(cmd, MAXNAMELEN); |
8061 | ||
34dc7c2f BB |
8062 | return (0); |
8063 | } |