]>
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. |
c242c188 | 23 | * Copyright (c) 2012 by Delphix. All rights reserved. |
3541dc6d | 24 | * Copyright 2011 Nexenta Systems, Inc. All rights reserved. |
95fd54a1 | 25 | * Copyright (c) 2013 Steven Hartland. All rights reserved. |
34dc7c2f BB |
26 | */ |
27 | ||
34dc7c2f BB |
28 | /* |
29 | * The objective of this program is to provide a DMU/ZAP/SPA stress test | |
30 | * that runs entirely in userland, is easy to use, and easy to extend. | |
31 | * | |
32 | * The overall design of the ztest program is as follows: | |
33 | * | |
34 | * (1) For each major functional area (e.g. adding vdevs to a pool, | |
35 | * creating and destroying datasets, reading and writing objects, etc) | |
36 | * we have a simple routine to test that functionality. These | |
37 | * individual routines do not have to do anything "stressful". | |
38 | * | |
39 | * (2) We turn these simple functionality tests into a stress test by | |
40 | * running them all in parallel, with as many threads as desired, | |
41 | * and spread across as many datasets, objects, and vdevs as desired. | |
42 | * | |
43 | * (3) While all this is happening, we inject faults into the pool to | |
44 | * verify that self-healing data really works. | |
45 | * | |
46 | * (4) Every time we open a dataset, we change its checksum and compression | |
47 | * functions. Thus even individual objects vary from block to block | |
48 | * in which checksum they use and whether they're compressed. | |
49 | * | |
50 | * (5) To verify that we never lose on-disk consistency after a crash, | |
51 | * we run the entire test in a child of the main process. | |
52 | * At random times, the child self-immolates with a SIGKILL. | |
53 | * This is the software equivalent of pulling the power cord. | |
54 | * The parent then runs the test again, using the existing | |
c242c188 CS |
55 | * storage pool, as many times as desired. If backwards compatability |
56 | * testing is enabled ztest will sometimes run the "older" version | |
57 | * of ztest after a SIGKILL. | |
34dc7c2f BB |
58 | * |
59 | * (6) To verify that we don't have future leaks or temporal incursions, | |
60 | * many of the functional tests record the transaction group number | |
61 | * as part of their data. When reading old data, they verify that | |
62 | * the transaction group number is less than the current, open txg. | |
63 | * If you add a new test, please do this if applicable. | |
64 | * | |
1e33ac1e BB |
65 | * (7) Threads are created with a reduced stack size, for sanity checking. |
66 | * Therefore, it's important not to allocate huge buffers on the stack. | |
67 | * | |
34dc7c2f BB |
68 | * When run with no arguments, ztest runs for about five minutes and |
69 | * produces no output if successful. To get a little bit of information, | |
70 | * specify -V. To get more information, specify -VV, and so on. | |
71 | * | |
72 | * To turn this into an overnight stress test, use -T to specify run time. | |
73 | * | |
74 | * You can ask more more vdevs [-v], datasets [-d], or threads [-t] | |
75 | * to increase the pool capacity, fanout, and overall stress level. | |
76 | * | |
c242c188 CS |
77 | * Use the -k option to set the desired frequency of kills. |
78 | * | |
79 | * When ztest invokes itself it passes all relevant information through a | |
80 | * temporary file which is mmap-ed in the child process. This allows shared | |
81 | * memory to survive the exec syscall. The ztest_shared_hdr_t struct is always | |
82 | * stored at offset 0 of this file and contains information on the size and | |
83 | * number of shared structures in the file. The information stored in this file | |
84 | * must remain backwards compatible with older versions of ztest so that | |
85 | * ztest can invoke them during backwards compatibility testing (-B). | |
34dc7c2f BB |
86 | */ |
87 | ||
88 | #include <sys/zfs_context.h> | |
89 | #include <sys/spa.h> | |
90 | #include <sys/dmu.h> | |
91 | #include <sys/txg.h> | |
9babb374 | 92 | #include <sys/dbuf.h> |
34dc7c2f | 93 | #include <sys/zap.h> |
34dc7c2f BB |
94 | #include <sys/dmu_objset.h> |
95 | #include <sys/poll.h> | |
96 | #include <sys/stat.h> | |
97 | #include <sys/time.h> | |
98 | #include <sys/wait.h> | |
99 | #include <sys/mman.h> | |
100 | #include <sys/resource.h> | |
101 | #include <sys/zio.h> | |
34dc7c2f | 102 | #include <sys/zil.h> |
428870ff | 103 | #include <sys/zil_impl.h> |
34dc7c2f | 104 | #include <sys/vdev_impl.h> |
b128c09f | 105 | #include <sys/vdev_file.h> |
34dc7c2f | 106 | #include <sys/spa_impl.h> |
428870ff | 107 | #include <sys/metaslab_impl.h> |
34dc7c2f | 108 | #include <sys/dsl_prop.h> |
9babb374 | 109 | #include <sys/dsl_dataset.h> |
13fe0198 | 110 | #include <sys/dsl_destroy.h> |
428870ff BB |
111 | #include <sys/dsl_scan.h> |
112 | #include <sys/zio_checksum.h> | |
34dc7c2f | 113 | #include <sys/refcount.h> |
9ae529ec | 114 | #include <sys/zfeature.h> |
13fe0198 | 115 | #include <sys/dsl_userhold.h> |
34dc7c2f BB |
116 | #include <stdio.h> |
117 | #include <stdio_ext.h> | |
118 | #include <stdlib.h> | |
119 | #include <unistd.h> | |
120 | #include <signal.h> | |
121 | #include <umem.h> | |
122 | #include <dlfcn.h> | |
123 | #include <ctype.h> | |
124 | #include <math.h> | |
125 | #include <sys/fs/zfs.h> | |
428870ff | 126 | #include <libnvpair.h> |
34dc7c2f | 127 | |
9d81146b ED |
128 | static int ztest_fd_data = -1; |
129 | static int ztest_fd_rand = -1; | |
c242c188 CS |
130 | |
131 | typedef struct ztest_shared_hdr { | |
132 | uint64_t zh_hdr_size; | |
133 | uint64_t zh_opts_size; | |
134 | uint64_t zh_size; | |
135 | uint64_t zh_stats_size; | |
136 | uint64_t zh_stats_count; | |
137 | uint64_t zh_ds_size; | |
138 | uint64_t zh_ds_count; | |
139 | } ztest_shared_hdr_t; | |
140 | ||
141 | static ztest_shared_hdr_t *ztest_shared_hdr; | |
142 | ||
143 | typedef struct ztest_shared_opts { | |
144 | char zo_pool[MAXNAMELEN]; | |
145 | char zo_dir[MAXNAMELEN]; | |
146 | char zo_alt_ztest[MAXNAMELEN]; | |
147 | char zo_alt_libpath[MAXNAMELEN]; | |
148 | uint64_t zo_vdevs; | |
149 | uint64_t zo_vdevtime; | |
150 | size_t zo_vdev_size; | |
151 | int zo_ashift; | |
152 | int zo_mirrors; | |
153 | int zo_raidz; | |
154 | int zo_raidz_parity; | |
155 | int zo_datasets; | |
156 | int zo_threads; | |
157 | uint64_t zo_passtime; | |
158 | uint64_t zo_killrate; | |
159 | int zo_verbose; | |
160 | int zo_init; | |
161 | uint64_t zo_time; | |
162 | uint64_t zo_maxloops; | |
163 | uint64_t zo_metaslab_gang_bang; | |
164 | } ztest_shared_opts_t; | |
165 | ||
166 | static const ztest_shared_opts_t ztest_opts_defaults = { | |
167 | .zo_pool = { 'z', 't', 'e', 's', 't', '\0' }, | |
168 | .zo_dir = { '/', 't', 'm', 'p', '\0' }, | |
169 | .zo_alt_ztest = { '\0' }, | |
170 | .zo_alt_libpath = { '\0' }, | |
171 | .zo_vdevs = 5, | |
172 | .zo_ashift = SPA_MINBLOCKSHIFT, | |
173 | .zo_mirrors = 2, | |
174 | .zo_raidz = 4, | |
175 | .zo_raidz_parity = 1, | |
176 | .zo_vdev_size = SPA_MINDEVSIZE, | |
177 | .zo_datasets = 7, | |
178 | .zo_threads = 23, | |
179 | .zo_passtime = 60, /* 60 seconds */ | |
180 | .zo_killrate = 70, /* 70% kill rate */ | |
181 | .zo_verbose = 0, | |
182 | .zo_init = 1, | |
183 | .zo_time = 300, /* 5 minutes */ | |
184 | .zo_maxloops = 50, /* max loops during spa_freeze() */ | |
185 | .zo_metaslab_gang_bang = 32 << 10 | |
186 | }; | |
187 | ||
188 | extern uint64_t metaslab_gang_bang; | |
189 | extern uint64_t metaslab_df_alloc_threshold; | |
190 | ||
191 | static ztest_shared_opts_t *ztest_shared_opts; | |
192 | static ztest_shared_opts_t ztest_opts; | |
193 | ||
194 | typedef struct ztest_shared_ds { | |
195 | uint64_t zd_seq; | |
196 | } ztest_shared_ds_t; | |
197 | ||
198 | static ztest_shared_ds_t *ztest_shared_ds; | |
199 | #define ZTEST_GET_SHARED_DS(d) (&ztest_shared_ds[d]) | |
428870ff BB |
200 | |
201 | #define BT_MAGIC 0x123456789abcdefULL | |
c242c188 CS |
202 | #define MAXFAULTS() \ |
203 | (MAX(zs->zs_mirrors, 1) * (ztest_opts.zo_raidz_parity + 1) - 1) | |
428870ff BB |
204 | |
205 | enum ztest_io_type { | |
206 | ZTEST_IO_WRITE_TAG, | |
207 | ZTEST_IO_WRITE_PATTERN, | |
208 | ZTEST_IO_WRITE_ZEROES, | |
209 | ZTEST_IO_TRUNCATE, | |
210 | ZTEST_IO_SETATTR, | |
211 | ZTEST_IO_TYPES | |
212 | }; | |
34dc7c2f BB |
213 | |
214 | typedef struct ztest_block_tag { | |
428870ff | 215 | uint64_t bt_magic; |
34dc7c2f BB |
216 | uint64_t bt_objset; |
217 | uint64_t bt_object; | |
218 | uint64_t bt_offset; | |
428870ff | 219 | uint64_t bt_gen; |
34dc7c2f | 220 | uint64_t bt_txg; |
428870ff | 221 | uint64_t bt_crtxg; |
34dc7c2f BB |
222 | } ztest_block_tag_t; |
223 | ||
428870ff BB |
224 | typedef struct bufwad { |
225 | uint64_t bw_index; | |
226 | uint64_t bw_txg; | |
227 | uint64_t bw_data; | |
228 | } bufwad_t; | |
229 | ||
230 | /* | |
231 | * XXX -- fix zfs range locks to be generic so we can use them here. | |
232 | */ | |
233 | typedef enum { | |
234 | RL_READER, | |
235 | RL_WRITER, | |
236 | RL_APPEND | |
237 | } rl_type_t; | |
238 | ||
239 | typedef struct rll { | |
240 | void *rll_writer; | |
241 | int rll_readers; | |
1e33ac1e BB |
242 | kmutex_t rll_lock; |
243 | kcondvar_t rll_cv; | |
428870ff BB |
244 | } rll_t; |
245 | ||
246 | typedef struct rl { | |
247 | uint64_t rl_object; | |
248 | uint64_t rl_offset; | |
249 | uint64_t rl_size; | |
250 | rll_t *rl_lock; | |
251 | } rl_t; | |
252 | ||
253 | #define ZTEST_RANGE_LOCKS 64 | |
254 | #define ZTEST_OBJECT_LOCKS 64 | |
255 | ||
256 | /* | |
257 | * Object descriptor. Used as a template for object lookup/create/remove. | |
258 | */ | |
259 | typedef struct ztest_od { | |
260 | uint64_t od_dir; | |
261 | uint64_t od_object; | |
262 | dmu_object_type_t od_type; | |
263 | dmu_object_type_t od_crtype; | |
264 | uint64_t od_blocksize; | |
265 | uint64_t od_crblocksize; | |
266 | uint64_t od_gen; | |
267 | uint64_t od_crgen; | |
268 | char od_name[MAXNAMELEN]; | |
269 | } ztest_od_t; | |
34dc7c2f | 270 | |
428870ff BB |
271 | /* |
272 | * Per-dataset state. | |
273 | */ | |
274 | typedef struct ztest_ds { | |
c242c188 | 275 | ztest_shared_ds_t *zd_shared; |
428870ff | 276 | objset_t *zd_os; |
3e31d2b0 | 277 | krwlock_t zd_zilog_lock; |
428870ff | 278 | zilog_t *zd_zilog; |
428870ff BB |
279 | ztest_od_t *zd_od; /* debugging aid */ |
280 | char zd_name[MAXNAMELEN]; | |
1e33ac1e | 281 | kmutex_t zd_dirobj_lock; |
428870ff BB |
282 | rll_t zd_object_lock[ZTEST_OBJECT_LOCKS]; |
283 | rll_t zd_range_lock[ZTEST_RANGE_LOCKS]; | |
284 | } ztest_ds_t; | |
285 | ||
286 | /* | |
287 | * Per-iteration state. | |
288 | */ | |
289 | typedef void ztest_func_t(ztest_ds_t *zd, uint64_t id); | |
290 | ||
291 | typedef struct ztest_info { | |
292 | ztest_func_t *zi_func; /* test function */ | |
293 | uint64_t zi_iters; /* iterations per execution */ | |
294 | uint64_t *zi_interval; /* execute every <interval> seconds */ | |
428870ff | 295 | } ztest_info_t; |
34dc7c2f | 296 | |
c242c188 CS |
297 | typedef struct ztest_shared_callstate { |
298 | uint64_t zc_count; /* per-pass count */ | |
299 | uint64_t zc_time; /* per-pass time */ | |
300 | uint64_t zc_next; /* next time to call this function */ | |
301 | } ztest_shared_callstate_t; | |
302 | ||
303 | static ztest_shared_callstate_t *ztest_shared_callstate; | |
304 | #define ZTEST_GET_SHARED_CALLSTATE(c) (&ztest_shared_callstate[c]) | |
305 | ||
34dc7c2f BB |
306 | /* |
307 | * Note: these aren't static because we want dladdr() to work. | |
308 | */ | |
309 | ztest_func_t ztest_dmu_read_write; | |
310 | ztest_func_t ztest_dmu_write_parallel; | |
311 | ztest_func_t ztest_dmu_object_alloc_free; | |
428870ff | 312 | ztest_func_t ztest_dmu_commit_callbacks; |
34dc7c2f BB |
313 | ztest_func_t ztest_zap; |
314 | ztest_func_t ztest_zap_parallel; | |
428870ff | 315 | ztest_func_t ztest_zil_commit; |
3e31d2b0 | 316 | ztest_func_t ztest_zil_remount; |
428870ff | 317 | ztest_func_t ztest_dmu_read_write_zcopy; |
34dc7c2f | 318 | ztest_func_t ztest_dmu_objset_create_destroy; |
428870ff BB |
319 | ztest_func_t ztest_dmu_prealloc; |
320 | ztest_func_t ztest_fzap; | |
34dc7c2f | 321 | ztest_func_t ztest_dmu_snapshot_create_destroy; |
428870ff BB |
322 | ztest_func_t ztest_dsl_prop_get_set; |
323 | ztest_func_t ztest_spa_prop_get_set; | |
34dc7c2f BB |
324 | ztest_func_t ztest_spa_create_destroy; |
325 | ztest_func_t ztest_fault_inject; | |
428870ff BB |
326 | ztest_func_t ztest_ddt_repair; |
327 | ztest_func_t ztest_dmu_snapshot_hold; | |
b128c09f | 328 | ztest_func_t ztest_spa_rename; |
428870ff BB |
329 | ztest_func_t ztest_scrub; |
330 | ztest_func_t ztest_dsl_dataset_promote_busy; | |
34dc7c2f BB |
331 | ztest_func_t ztest_vdev_attach_detach; |
332 | ztest_func_t ztest_vdev_LUN_growth; | |
333 | ztest_func_t ztest_vdev_add_remove; | |
b128c09f | 334 | ztest_func_t ztest_vdev_aux_add_remove; |
428870ff | 335 | ztest_func_t ztest_split_pool; |
3541dc6d | 336 | ztest_func_t ztest_reguid; |
ea0b2538 | 337 | ztest_func_t ztest_spa_upgrade; |
34dc7c2f | 338 | |
428870ff BB |
339 | uint64_t zopt_always = 0ULL * NANOSEC; /* all the time */ |
340 | uint64_t zopt_incessant = 1ULL * NANOSEC / 10; /* every 1/10 second */ | |
341 | uint64_t zopt_often = 1ULL * NANOSEC; /* every second */ | |
342 | uint64_t zopt_sometimes = 10ULL * NANOSEC; /* every 10 seconds */ | |
343 | uint64_t zopt_rarely = 60ULL * NANOSEC; /* every 60 seconds */ | |
34dc7c2f BB |
344 | |
345 | ztest_info_t ztest_info[] = { | |
346 | { ztest_dmu_read_write, 1, &zopt_always }, | |
428870ff | 347 | { ztest_dmu_write_parallel, 10, &zopt_always }, |
34dc7c2f | 348 | { ztest_dmu_object_alloc_free, 1, &zopt_always }, |
428870ff | 349 | { ztest_dmu_commit_callbacks, 1, &zopt_always }, |
34dc7c2f BB |
350 | { ztest_zap, 30, &zopt_always }, |
351 | { ztest_zap_parallel, 100, &zopt_always }, | |
428870ff BB |
352 | { ztest_split_pool, 1, &zopt_always }, |
353 | { ztest_zil_commit, 1, &zopt_incessant }, | |
3e31d2b0 | 354 | { ztest_zil_remount, 1, &zopt_sometimes }, |
428870ff BB |
355 | { ztest_dmu_read_write_zcopy, 1, &zopt_often }, |
356 | { ztest_dmu_objset_create_destroy, 1, &zopt_often }, | |
357 | { ztest_dsl_prop_get_set, 1, &zopt_often }, | |
358 | { ztest_spa_prop_get_set, 1, &zopt_sometimes }, | |
359 | #if 0 | |
360 | { ztest_dmu_prealloc, 1, &zopt_sometimes }, | |
361 | #endif | |
362 | { ztest_fzap, 1, &zopt_sometimes }, | |
363 | { ztest_dmu_snapshot_create_destroy, 1, &zopt_sometimes }, | |
364 | { ztest_spa_create_destroy, 1, &zopt_sometimes }, | |
34dc7c2f | 365 | { ztest_fault_inject, 1, &zopt_sometimes }, |
428870ff BB |
366 | { ztest_ddt_repair, 1, &zopt_sometimes }, |
367 | { ztest_dmu_snapshot_hold, 1, &zopt_sometimes }, | |
3541dc6d | 368 | { ztest_reguid, 1, &zopt_sometimes }, |
34dc7c2f | 369 | { ztest_spa_rename, 1, &zopt_rarely }, |
428870ff | 370 | { ztest_scrub, 1, &zopt_rarely }, |
ea0b2538 | 371 | { ztest_spa_upgrade, 1, &zopt_rarely }, |
9babb374 | 372 | { ztest_dsl_dataset_promote_busy, 1, &zopt_rarely }, |
13fe0198 | 373 | { ztest_vdev_attach_detach, 1, &zopt_sometimes }, |
428870ff | 374 | { ztest_vdev_LUN_growth, 1, &zopt_rarely }, |
c242c188 CS |
375 | { ztest_vdev_add_remove, 1, |
376 | &ztest_opts.zo_vdevtime }, | |
377 | { ztest_vdev_aux_add_remove, 1, | |
378 | &ztest_opts.zo_vdevtime }, | |
34dc7c2f BB |
379 | }; |
380 | ||
381 | #define ZTEST_FUNCS (sizeof (ztest_info) / sizeof (ztest_info_t)) | |
382 | ||
428870ff BB |
383 | /* |
384 | * The following struct is used to hold a list of uncalled commit callbacks. | |
385 | * The callbacks are ordered by txg number. | |
386 | */ | |
387 | typedef struct ztest_cb_list { | |
1e33ac1e BB |
388 | kmutex_t zcl_callbacks_lock; |
389 | list_t zcl_callbacks; | |
428870ff | 390 | } ztest_cb_list_t; |
34dc7c2f BB |
391 | |
392 | /* | |
393 | * Stuff we need to share writably between parent and child. | |
394 | */ | |
395 | typedef struct ztest_shared { | |
c242c188 | 396 | boolean_t zs_do_init; |
428870ff BB |
397 | hrtime_t zs_proc_start; |
398 | hrtime_t zs_proc_stop; | |
399 | hrtime_t zs_thread_start; | |
400 | hrtime_t zs_thread_stop; | |
401 | hrtime_t zs_thread_kill; | |
34dc7c2f | 402 | uint64_t zs_enospc_count; |
428870ff BB |
403 | uint64_t zs_vdev_next_leaf; |
404 | uint64_t zs_vdev_aux; | |
34dc7c2f BB |
405 | uint64_t zs_alloc; |
406 | uint64_t zs_space; | |
428870ff BB |
407 | uint64_t zs_splits; |
408 | uint64_t zs_mirrors; | |
c242c188 CS |
409 | uint64_t zs_metaslab_sz; |
410 | uint64_t zs_metaslab_df_alloc_threshold; | |
411 | uint64_t zs_guid; | |
34dc7c2f BB |
412 | } ztest_shared_t; |
413 | ||
428870ff BB |
414 | #define ID_PARALLEL -1ULL |
415 | ||
34dc7c2f | 416 | static char ztest_dev_template[] = "%s/%s.%llua"; |
b128c09f | 417 | static char ztest_aux_template[] = "%s/%s.%s.%llu"; |
428870ff | 418 | ztest_shared_t *ztest_shared; |
34dc7c2f | 419 | |
c242c188 CS |
420 | static spa_t *ztest_spa = NULL; |
421 | static ztest_ds_t *ztest_ds; | |
422 | ||
423 | static kmutex_t ztest_vdev_lock; | |
3bc7e0fb GW |
424 | |
425 | /* | |
426 | * The ztest_name_lock protects the pool and dataset namespace used by | |
427 | * the individual tests. To modify the namespace, consumers must grab | |
428 | * this lock as writer. Grabbing the lock as reader will ensure that the | |
429 | * namespace does not change while the lock is held. | |
430 | */ | |
c242c188 | 431 | static krwlock_t ztest_name_lock; |
34dc7c2f | 432 | |
c242c188 | 433 | static boolean_t ztest_dump_core = B_TRUE; |
b128c09f | 434 | static boolean_t ztest_exiting; |
34dc7c2f | 435 | |
428870ff BB |
436 | /* Global commit callback list */ |
437 | static ztest_cb_list_t zcl; | |
090ff092 RC |
438 | /* Commit cb delay */ |
439 | static uint64_t zc_min_txg_delay = UINT64_MAX; | |
440 | static int zc_cb_counter = 0; | |
441 | ||
442 | /* | |
443 | * Minimum number of commit callbacks that need to be registered for us to check | |
444 | * whether the minimum txg delay is acceptable. | |
445 | */ | |
446 | #define ZTEST_COMMIT_CB_MIN_REG 100 | |
447 | ||
448 | /* | |
449 | * If a number of txgs equal to this threshold have been created after a commit | |
450 | * callback has been registered but not called, then we assume there is an | |
451 | * implementation bug. | |
452 | */ | |
453 | #define ZTEST_COMMIT_CB_THRESH (TXG_CONCURRENT_STATES + 1000) | |
428870ff | 454 | |
34dc7c2f | 455 | extern uint64_t metaslab_gang_bang; |
9babb374 | 456 | extern uint64_t metaslab_df_alloc_threshold; |
34dc7c2f | 457 | |
428870ff BB |
458 | enum ztest_object { |
459 | ZTEST_META_DNODE = 0, | |
460 | ZTEST_DIROBJ, | |
461 | ZTEST_OBJECTS | |
462 | }; | |
34dc7c2f BB |
463 | |
464 | static void usage(boolean_t) __NORETURN; | |
465 | ||
466 | /* | |
467 | * These libumem hooks provide a reasonable set of defaults for the allocator's | |
468 | * debugging facilities. | |
469 | */ | |
470 | const char * | |
0bc8fd78 | 471 | _umem_debug_init(void) |
34dc7c2f BB |
472 | { |
473 | return ("default,verbose"); /* $UMEM_DEBUG setting */ | |
474 | } | |
475 | ||
476 | const char * | |
477 | _umem_logging_init(void) | |
478 | { | |
479 | return ("fail,contents"); /* $UMEM_LOGGING setting */ | |
480 | } | |
481 | ||
482 | #define FATAL_MSG_SZ 1024 | |
483 | ||
484 | char *fatal_msg; | |
485 | ||
486 | static void | |
487 | fatal(int do_perror, char *message, ...) | |
488 | { | |
489 | va_list args; | |
490 | int save_errno = errno; | |
40b84e7a | 491 | char *buf; |
34dc7c2f BB |
492 | |
493 | (void) fflush(stdout); | |
40b84e7a | 494 | buf = umem_alloc(FATAL_MSG_SZ, UMEM_NOFAIL); |
34dc7c2f BB |
495 | |
496 | va_start(args, message); | |
497 | (void) sprintf(buf, "ztest: "); | |
498 | /* LINTED */ | |
499 | (void) vsprintf(buf + strlen(buf), message, args); | |
500 | va_end(args); | |
501 | if (do_perror) { | |
502 | (void) snprintf(buf + strlen(buf), FATAL_MSG_SZ - strlen(buf), | |
503 | ": %s", strerror(save_errno)); | |
504 | } | |
505 | (void) fprintf(stderr, "%s\n", buf); | |
506 | fatal_msg = buf; /* to ease debugging */ | |
507 | if (ztest_dump_core) | |
508 | abort(); | |
509 | exit(3); | |
510 | } | |
511 | ||
512 | static int | |
513 | str2shift(const char *buf) | |
514 | { | |
515 | const char *ends = "BKMGTPEZ"; | |
516 | int i; | |
517 | ||
518 | if (buf[0] == '\0') | |
519 | return (0); | |
520 | for (i = 0; i < strlen(ends); i++) { | |
521 | if (toupper(buf[0]) == ends[i]) | |
522 | break; | |
523 | } | |
524 | if (i == strlen(ends)) { | |
525 | (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", | |
526 | buf); | |
527 | usage(B_FALSE); | |
528 | } | |
529 | if (buf[1] == '\0' || (toupper(buf[1]) == 'B' && buf[2] == '\0')) { | |
530 | return (10*i); | |
531 | } | |
532 | (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", buf); | |
533 | usage(B_FALSE); | |
534 | /* NOTREACHED */ | |
535 | } | |
536 | ||
537 | static uint64_t | |
538 | nicenumtoull(const char *buf) | |
539 | { | |
540 | char *end; | |
541 | uint64_t val; | |
542 | ||
543 | val = strtoull(buf, &end, 0); | |
544 | if (end == buf) { | |
545 | (void) fprintf(stderr, "ztest: bad numeric value: %s\n", buf); | |
546 | usage(B_FALSE); | |
547 | } else if (end[0] == '.') { | |
548 | double fval = strtod(buf, &end); | |
549 | fval *= pow(2, str2shift(end)); | |
550 | if (fval > UINT64_MAX) { | |
551 | (void) fprintf(stderr, "ztest: value too large: %s\n", | |
552 | buf); | |
553 | usage(B_FALSE); | |
554 | } | |
555 | val = (uint64_t)fval; | |
556 | } else { | |
557 | int shift = str2shift(end); | |
558 | if (shift >= 64 || (val << shift) >> shift != val) { | |
559 | (void) fprintf(stderr, "ztest: value too large: %s\n", | |
560 | buf); | |
561 | usage(B_FALSE); | |
562 | } | |
563 | val <<= shift; | |
564 | } | |
565 | return (val); | |
566 | } | |
567 | ||
568 | static void | |
569 | usage(boolean_t requested) | |
570 | { | |
c242c188 CS |
571 | const ztest_shared_opts_t *zo = &ztest_opts_defaults; |
572 | ||
34dc7c2f BB |
573 | char nice_vdev_size[10]; |
574 | char nice_gang_bang[10]; | |
575 | FILE *fp = requested ? stdout : stderr; | |
576 | ||
c242c188 CS |
577 | nicenum(zo->zo_vdev_size, nice_vdev_size); |
578 | nicenum(zo->zo_metaslab_gang_bang, nice_gang_bang); | |
34dc7c2f BB |
579 | |
580 | (void) fprintf(fp, "Usage: %s\n" | |
581 | "\t[-v vdevs (default: %llu)]\n" | |
582 | "\t[-s size_of_each_vdev (default: %s)]\n" | |
428870ff | 583 | "\t[-a alignment_shift (default: %d)] use 0 for random\n" |
34dc7c2f BB |
584 | "\t[-m mirror_copies (default: %d)]\n" |
585 | "\t[-r raidz_disks (default: %d)]\n" | |
586 | "\t[-R raidz_parity (default: %d)]\n" | |
587 | "\t[-d datasets (default: %d)]\n" | |
588 | "\t[-t threads (default: %d)]\n" | |
589 | "\t[-g gang_block_threshold (default: %s)]\n" | |
428870ff BB |
590 | "\t[-i init_count (default: %d)] initialize pool i times\n" |
591 | "\t[-k kill_percentage (default: %llu%%)]\n" | |
34dc7c2f | 592 | "\t[-p pool_name (default: %s)]\n" |
428870ff BB |
593 | "\t[-f dir (default: %s)] file directory for vdev files\n" |
594 | "\t[-V] verbose (use multiple times for ever more blather)\n" | |
595 | "\t[-E] use existing pool instead of creating new one\n" | |
596 | "\t[-T time (default: %llu sec)] total run time\n" | |
597 | "\t[-F freezeloops (default: %llu)] max loops in spa_freeze()\n" | |
598 | "\t[-P passtime (default: %llu sec)] time per pass\n" | |
c242c188 | 599 | "\t[-B alt_ztest (default: <none>)] alternate ztest path\n" |
34dc7c2f BB |
600 | "\t[-h] (print help)\n" |
601 | "", | |
c242c188 CS |
602 | zo->zo_pool, |
603 | (u_longlong_t)zo->zo_vdevs, /* -v */ | |
34dc7c2f | 604 | nice_vdev_size, /* -s */ |
c242c188 CS |
605 | zo->zo_ashift, /* -a */ |
606 | zo->zo_mirrors, /* -m */ | |
607 | zo->zo_raidz, /* -r */ | |
608 | zo->zo_raidz_parity, /* -R */ | |
609 | zo->zo_datasets, /* -d */ | |
610 | zo->zo_threads, /* -t */ | |
34dc7c2f | 611 | nice_gang_bang, /* -g */ |
c242c188 CS |
612 | zo->zo_init, /* -i */ |
613 | (u_longlong_t)zo->zo_killrate, /* -k */ | |
614 | zo->zo_pool, /* -p */ | |
615 | zo->zo_dir, /* -f */ | |
616 | (u_longlong_t)zo->zo_time, /* -T */ | |
617 | (u_longlong_t)zo->zo_maxloops, /* -F */ | |
618 | (u_longlong_t)zo->zo_passtime); | |
34dc7c2f BB |
619 | exit(requested ? 0 : 1); |
620 | } | |
621 | ||
34dc7c2f BB |
622 | static void |
623 | process_options(int argc, char **argv) | |
624 | { | |
c242c188 CS |
625 | char *path; |
626 | ztest_shared_opts_t *zo = &ztest_opts; | |
627 | ||
34dc7c2f BB |
628 | int opt; |
629 | uint64_t value; | |
c242c188 | 630 | char altdir[MAXNAMELEN] = { 0 }; |
34dc7c2f | 631 | |
c242c188 | 632 | bcopy(&ztest_opts_defaults, zo, sizeof (*zo)); |
34dc7c2f | 633 | |
34dc7c2f | 634 | while ((opt = getopt(argc, argv, |
c242c188 | 635 | "v:s:a:m:r:R:d:t:g:i:k:p:f:VET:P:hF:B:")) != EOF) { |
34dc7c2f BB |
636 | value = 0; |
637 | switch (opt) { | |
638 | case 'v': | |
639 | case 's': | |
640 | case 'a': | |
641 | case 'm': | |
642 | case 'r': | |
643 | case 'R': | |
644 | case 'd': | |
645 | case 't': | |
646 | case 'g': | |
647 | case 'i': | |
648 | case 'k': | |
649 | case 'T': | |
650 | case 'P': | |
428870ff | 651 | case 'F': |
34dc7c2f BB |
652 | value = nicenumtoull(optarg); |
653 | } | |
654 | switch (opt) { | |
655 | case 'v': | |
c242c188 | 656 | zo->zo_vdevs = value; |
34dc7c2f BB |
657 | break; |
658 | case 's': | |
c242c188 | 659 | zo->zo_vdev_size = MAX(SPA_MINDEVSIZE, value); |
34dc7c2f BB |
660 | break; |
661 | case 'a': | |
c242c188 | 662 | zo->zo_ashift = value; |
34dc7c2f BB |
663 | break; |
664 | case 'm': | |
c242c188 | 665 | zo->zo_mirrors = value; |
34dc7c2f BB |
666 | break; |
667 | case 'r': | |
c242c188 | 668 | zo->zo_raidz = MAX(1, value); |
34dc7c2f BB |
669 | break; |
670 | case 'R': | |
c242c188 | 671 | zo->zo_raidz_parity = MIN(MAX(value, 1), 3); |
34dc7c2f BB |
672 | break; |
673 | case 'd': | |
c242c188 | 674 | zo->zo_datasets = MAX(1, value); |
34dc7c2f BB |
675 | break; |
676 | case 't': | |
c242c188 | 677 | zo->zo_threads = MAX(1, value); |
34dc7c2f BB |
678 | break; |
679 | case 'g': | |
c242c188 CS |
680 | zo->zo_metaslab_gang_bang = MAX(SPA_MINBLOCKSIZE << 1, |
681 | value); | |
34dc7c2f BB |
682 | break; |
683 | case 'i': | |
c242c188 | 684 | zo->zo_init = value; |
34dc7c2f BB |
685 | break; |
686 | case 'k': | |
c242c188 | 687 | zo->zo_killrate = value; |
34dc7c2f BB |
688 | break; |
689 | case 'p': | |
c242c188 CS |
690 | (void) strlcpy(zo->zo_pool, optarg, |
691 | sizeof (zo->zo_pool)); | |
34dc7c2f BB |
692 | break; |
693 | case 'f': | |
c242c188 CS |
694 | path = realpath(optarg, NULL); |
695 | if (path == NULL) { | |
696 | (void) fprintf(stderr, "error: %s: %s\n", | |
697 | optarg, strerror(errno)); | |
698 | usage(B_FALSE); | |
699 | } else { | |
700 | (void) strlcpy(zo->zo_dir, path, | |
701 | sizeof (zo->zo_dir)); | |
702 | } | |
34dc7c2f BB |
703 | break; |
704 | case 'V': | |
c242c188 | 705 | zo->zo_verbose++; |
34dc7c2f BB |
706 | break; |
707 | case 'E': | |
c242c188 | 708 | zo->zo_init = 0; |
34dc7c2f BB |
709 | break; |
710 | case 'T': | |
c242c188 | 711 | zo->zo_time = value; |
34dc7c2f BB |
712 | break; |
713 | case 'P': | |
c242c188 | 714 | zo->zo_passtime = MAX(1, value); |
34dc7c2f | 715 | break; |
428870ff | 716 | case 'F': |
c242c188 CS |
717 | zo->zo_maxloops = MAX(1, value); |
718 | break; | |
719 | case 'B': | |
720 | (void) strlcpy(altdir, optarg, sizeof (altdir)); | |
428870ff | 721 | break; |
34dc7c2f BB |
722 | case 'h': |
723 | usage(B_TRUE); | |
724 | break; | |
725 | case '?': | |
726 | default: | |
727 | usage(B_FALSE); | |
728 | break; | |
729 | } | |
730 | } | |
731 | ||
c242c188 | 732 | zo->zo_raidz_parity = MIN(zo->zo_raidz_parity, zo->zo_raidz - 1); |
34dc7c2f | 733 | |
c242c188 CS |
734 | zo->zo_vdevtime = |
735 | (zo->zo_vdevs > 0 ? zo->zo_time * NANOSEC / zo->zo_vdevs : | |
428870ff | 736 | UINT64_MAX >> 2); |
c242c188 CS |
737 | |
738 | if (strlen(altdir) > 0) { | |
ae380cfa BB |
739 | char *cmd; |
740 | char *realaltdir; | |
c242c188 CS |
741 | char *bin; |
742 | char *ztest; | |
743 | char *isa; | |
744 | int isalen; | |
745 | ||
ae380cfa BB |
746 | cmd = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
747 | realaltdir = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
748 | ||
5be98cfe | 749 | VERIFY(NULL != realpath(getexecname(), cmd)); |
c242c188 CS |
750 | if (0 != access(altdir, F_OK)) { |
751 | ztest_dump_core = B_FALSE; | |
752 | fatal(B_TRUE, "invalid alternate ztest path: %s", | |
753 | altdir); | |
754 | } | |
755 | VERIFY(NULL != realpath(altdir, realaltdir)); | |
756 | ||
757 | /* | |
758 | * 'cmd' should be of the form "<anything>/usr/bin/<isa>/ztest". | |
759 | * We want to extract <isa> to determine if we should use | |
760 | * 32 or 64 bit binaries. | |
761 | */ | |
762 | bin = strstr(cmd, "/usr/bin/"); | |
763 | ztest = strstr(bin, "/ztest"); | |
764 | isa = bin + 9; | |
765 | isalen = ztest - isa; | |
766 | (void) snprintf(zo->zo_alt_ztest, sizeof (zo->zo_alt_ztest), | |
767 | "%s/usr/bin/%.*s/ztest", realaltdir, isalen, isa); | |
768 | (void) snprintf(zo->zo_alt_libpath, sizeof (zo->zo_alt_libpath), | |
769 | "%s/usr/lib/%.*s", realaltdir, isalen, isa); | |
770 | ||
771 | if (0 != access(zo->zo_alt_ztest, X_OK)) { | |
772 | ztest_dump_core = B_FALSE; | |
773 | fatal(B_TRUE, "invalid alternate ztest: %s", | |
774 | zo->zo_alt_ztest); | |
775 | } else if (0 != access(zo->zo_alt_libpath, X_OK)) { | |
776 | ztest_dump_core = B_FALSE; | |
777 | fatal(B_TRUE, "invalid alternate lib directory %s", | |
778 | zo->zo_alt_libpath); | |
779 | } | |
ae380cfa BB |
780 | |
781 | umem_free(cmd, MAXPATHLEN); | |
782 | umem_free(realaltdir, MAXPATHLEN); | |
c242c188 | 783 | } |
428870ff BB |
784 | } |
785 | ||
786 | static void | |
787 | ztest_kill(ztest_shared_t *zs) | |
788 | { | |
c242c188 CS |
789 | zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(ztest_spa)); |
790 | zs->zs_space = metaslab_class_get_space(spa_normal_class(ztest_spa)); | |
428870ff BB |
791 | (void) kill(getpid(), SIGKILL); |
792 | } | |
793 | ||
794 | static uint64_t | |
795 | ztest_random(uint64_t range) | |
796 | { | |
797 | uint64_t r; | |
798 | ||
9d81146b ED |
799 | ASSERT3S(ztest_fd_rand, >=, 0); |
800 | ||
428870ff BB |
801 | if (range == 0) |
802 | return (0); | |
803 | ||
9d81146b | 804 | if (read(ztest_fd_rand, &r, sizeof (r)) != sizeof (r)) |
428870ff BB |
805 | fatal(1, "short read from /dev/urandom"); |
806 | ||
807 | return (r % range); | |
808 | } | |
809 | ||
810 | /* ARGSUSED */ | |
811 | static void | |
812 | ztest_record_enospc(const char *s) | |
813 | { | |
814 | ztest_shared->zs_enospc_count++; | |
34dc7c2f BB |
815 | } |
816 | ||
817 | static uint64_t | |
818 | ztest_get_ashift(void) | |
819 | { | |
c242c188 | 820 | if (ztest_opts.zo_ashift == 0) |
34dc7c2f | 821 | return (SPA_MINBLOCKSHIFT + ztest_random(3)); |
c242c188 | 822 | return (ztest_opts.zo_ashift); |
34dc7c2f BB |
823 | } |
824 | ||
825 | static nvlist_t * | |
ea0b2538 | 826 | make_vdev_file(char *path, char *aux, char *pool, size_t size, uint64_t ashift) |
34dc7c2f | 827 | { |
40b84e7a | 828 | char *pathbuf; |
34dc7c2f | 829 | uint64_t vdev; |
34dc7c2f BB |
830 | nvlist_t *file; |
831 | ||
40b84e7a BB |
832 | pathbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
833 | ||
b128c09f BB |
834 | if (ashift == 0) |
835 | ashift = ztest_get_ashift(); | |
836 | ||
837 | if (path == NULL) { | |
838 | path = pathbuf; | |
839 | ||
840 | if (aux != NULL) { | |
841 | vdev = ztest_shared->zs_vdev_aux; | |
c242c188 CS |
842 | (void) snprintf(path, MAXPATHLEN, |
843 | ztest_aux_template, ztest_opts.zo_dir, | |
ea0b2538 GW |
844 | pool == NULL ? ztest_opts.zo_pool : pool, |
845 | aux, vdev); | |
b128c09f | 846 | } else { |
428870ff | 847 | vdev = ztest_shared->zs_vdev_next_leaf++; |
c242c188 CS |
848 | (void) snprintf(path, MAXPATHLEN, |
849 | ztest_dev_template, ztest_opts.zo_dir, | |
ea0b2538 | 850 | pool == NULL ? ztest_opts.zo_pool : pool, vdev); |
b128c09f BB |
851 | } |
852 | } | |
34dc7c2f | 853 | |
b128c09f BB |
854 | if (size != 0) { |
855 | int fd = open(path, O_RDWR | O_CREAT | O_TRUNC, 0666); | |
34dc7c2f | 856 | if (fd == -1) |
b128c09f | 857 | fatal(1, "can't open %s", path); |
34dc7c2f | 858 | if (ftruncate(fd, size) != 0) |
b128c09f | 859 | fatal(1, "can't ftruncate %s", path); |
34dc7c2f BB |
860 | (void) close(fd); |
861 | } | |
862 | ||
863 | VERIFY(nvlist_alloc(&file, NV_UNIQUE_NAME, 0) == 0); | |
864 | VERIFY(nvlist_add_string(file, ZPOOL_CONFIG_TYPE, VDEV_TYPE_FILE) == 0); | |
b128c09f | 865 | VERIFY(nvlist_add_string(file, ZPOOL_CONFIG_PATH, path) == 0); |
34dc7c2f | 866 | VERIFY(nvlist_add_uint64(file, ZPOOL_CONFIG_ASHIFT, ashift) == 0); |
40b84e7a | 867 | umem_free(pathbuf, MAXPATHLEN); |
34dc7c2f BB |
868 | |
869 | return (file); | |
870 | } | |
871 | ||
872 | static nvlist_t * | |
ea0b2538 GW |
873 | make_vdev_raidz(char *path, char *aux, char *pool, size_t size, |
874 | uint64_t ashift, int r) | |
34dc7c2f BB |
875 | { |
876 | nvlist_t *raidz, **child; | |
877 | int c; | |
878 | ||
879 | if (r < 2) | |
ea0b2538 | 880 | return (make_vdev_file(path, aux, pool, size, ashift)); |
34dc7c2f BB |
881 | child = umem_alloc(r * sizeof (nvlist_t *), UMEM_NOFAIL); |
882 | ||
883 | for (c = 0; c < r; c++) | |
ea0b2538 | 884 | child[c] = make_vdev_file(path, aux, pool, size, ashift); |
34dc7c2f BB |
885 | |
886 | VERIFY(nvlist_alloc(&raidz, NV_UNIQUE_NAME, 0) == 0); | |
887 | VERIFY(nvlist_add_string(raidz, ZPOOL_CONFIG_TYPE, | |
888 | VDEV_TYPE_RAIDZ) == 0); | |
889 | VERIFY(nvlist_add_uint64(raidz, ZPOOL_CONFIG_NPARITY, | |
c242c188 | 890 | ztest_opts.zo_raidz_parity) == 0); |
34dc7c2f BB |
891 | VERIFY(nvlist_add_nvlist_array(raidz, ZPOOL_CONFIG_CHILDREN, |
892 | child, r) == 0); | |
893 | ||
894 | for (c = 0; c < r; c++) | |
895 | nvlist_free(child[c]); | |
896 | ||
897 | umem_free(child, r * sizeof (nvlist_t *)); | |
898 | ||
899 | return (raidz); | |
900 | } | |
901 | ||
902 | static nvlist_t * | |
ea0b2538 GW |
903 | make_vdev_mirror(char *path, char *aux, char *pool, size_t size, |
904 | uint64_t ashift, int r, int m) | |
34dc7c2f BB |
905 | { |
906 | nvlist_t *mirror, **child; | |
907 | int c; | |
908 | ||
909 | if (m < 1) | |
ea0b2538 | 910 | return (make_vdev_raidz(path, aux, pool, size, ashift, r)); |
34dc7c2f BB |
911 | |
912 | child = umem_alloc(m * sizeof (nvlist_t *), UMEM_NOFAIL); | |
913 | ||
914 | for (c = 0; c < m; c++) | |
ea0b2538 | 915 | child[c] = make_vdev_raidz(path, aux, pool, size, ashift, r); |
34dc7c2f BB |
916 | |
917 | VERIFY(nvlist_alloc(&mirror, NV_UNIQUE_NAME, 0) == 0); | |
918 | VERIFY(nvlist_add_string(mirror, ZPOOL_CONFIG_TYPE, | |
919 | VDEV_TYPE_MIRROR) == 0); | |
920 | VERIFY(nvlist_add_nvlist_array(mirror, ZPOOL_CONFIG_CHILDREN, | |
921 | child, m) == 0); | |
34dc7c2f BB |
922 | |
923 | for (c = 0; c < m; c++) | |
924 | nvlist_free(child[c]); | |
925 | ||
926 | umem_free(child, m * sizeof (nvlist_t *)); | |
927 | ||
928 | return (mirror); | |
929 | } | |
930 | ||
931 | static nvlist_t * | |
ea0b2538 GW |
932 | make_vdev_root(char *path, char *aux, char *pool, size_t size, uint64_t ashift, |
933 | int log, int r, int m, int t) | |
34dc7c2f BB |
934 | { |
935 | nvlist_t *root, **child; | |
936 | int c; | |
937 | ||
938 | ASSERT(t > 0); | |
939 | ||
940 | child = umem_alloc(t * sizeof (nvlist_t *), UMEM_NOFAIL); | |
941 | ||
b128c09f | 942 | for (c = 0; c < t; c++) { |
ea0b2538 GW |
943 | child[c] = make_vdev_mirror(path, aux, pool, size, ashift, |
944 | r, m); | |
b128c09f BB |
945 | VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_IS_LOG, |
946 | log) == 0); | |
947 | } | |
34dc7c2f BB |
948 | |
949 | VERIFY(nvlist_alloc(&root, NV_UNIQUE_NAME, 0) == 0); | |
950 | VERIFY(nvlist_add_string(root, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT) == 0); | |
b128c09f | 951 | VERIFY(nvlist_add_nvlist_array(root, aux ? aux : ZPOOL_CONFIG_CHILDREN, |
34dc7c2f BB |
952 | child, t) == 0); |
953 | ||
954 | for (c = 0; c < t; c++) | |
955 | nvlist_free(child[c]); | |
956 | ||
957 | umem_free(child, t * sizeof (nvlist_t *)); | |
958 | ||
959 | return (root); | |
960 | } | |
961 | ||
ea0b2538 GW |
962 | /* |
963 | * Find a random spa version. Returns back a random spa version in the | |
964 | * range [initial_version, SPA_VERSION_FEATURES]. | |
965 | */ | |
966 | static uint64_t | |
967 | ztest_random_spa_version(uint64_t initial_version) | |
968 | { | |
969 | uint64_t version = initial_version; | |
970 | ||
971 | if (version <= SPA_VERSION_BEFORE_FEATURES) { | |
972 | version = version + | |
973 | ztest_random(SPA_VERSION_BEFORE_FEATURES - version + 1); | |
974 | } | |
975 | ||
976 | if (version > SPA_VERSION_BEFORE_FEATURES) | |
977 | version = SPA_VERSION_FEATURES; | |
978 | ||
979 | ASSERT(SPA_VERSION_IS_SUPPORTED(version)); | |
980 | return (version); | |
981 | } | |
982 | ||
428870ff BB |
983 | static int |
984 | ztest_random_blocksize(void) | |
34dc7c2f | 985 | { |
428870ff BB |
986 | return (1 << (SPA_MINBLOCKSHIFT + |
987 | ztest_random(SPA_MAXBLOCKSHIFT - SPA_MINBLOCKSHIFT + 1))); | |
988 | } | |
34dc7c2f | 989 | |
428870ff BB |
990 | static int |
991 | ztest_random_ibshift(void) | |
992 | { | |
993 | return (DN_MIN_INDBLKSHIFT + | |
994 | ztest_random(DN_MAX_INDBLKSHIFT - DN_MIN_INDBLKSHIFT + 1)); | |
34dc7c2f BB |
995 | } |
996 | ||
428870ff BB |
997 | static uint64_t |
998 | ztest_random_vdev_top(spa_t *spa, boolean_t log_ok) | |
34dc7c2f | 999 | { |
428870ff BB |
1000 | uint64_t top; |
1001 | vdev_t *rvd = spa->spa_root_vdev; | |
1002 | vdev_t *tvd; | |
34dc7c2f | 1003 | |
428870ff | 1004 | ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); |
34dc7c2f | 1005 | |
428870ff BB |
1006 | do { |
1007 | top = ztest_random(rvd->vdev_children); | |
1008 | tvd = rvd->vdev_child[top]; | |
1009 | } while (tvd->vdev_ishole || (tvd->vdev_islog && !log_ok) || | |
1010 | tvd->vdev_mg == NULL || tvd->vdev_mg->mg_class == NULL); | |
34dc7c2f | 1011 | |
428870ff | 1012 | return (top); |
34dc7c2f BB |
1013 | } |
1014 | ||
428870ff BB |
1015 | static uint64_t |
1016 | ztest_random_dsl_prop(zfs_prop_t prop) | |
34dc7c2f | 1017 | { |
428870ff BB |
1018 | uint64_t value; |
1019 | ||
1020 | do { | |
1021 | value = zfs_prop_random_value(prop, ztest_random(-1ULL)); | |
1022 | } while (prop == ZFS_PROP_CHECKSUM && value == ZIO_CHECKSUM_OFF); | |
1023 | ||
1024 | return (value); | |
34dc7c2f BB |
1025 | } |
1026 | ||
34dc7c2f | 1027 | static int |
428870ff BB |
1028 | ztest_dsl_prop_set_uint64(char *osname, zfs_prop_t prop, uint64_t value, |
1029 | boolean_t inherit) | |
34dc7c2f | 1030 | { |
428870ff BB |
1031 | const char *propname = zfs_prop_to_name(prop); |
1032 | const char *valname; | |
40b84e7a | 1033 | char *setpoint; |
428870ff | 1034 | uint64_t curval; |
34dc7c2f BB |
1035 | int error; |
1036 | ||
13fe0198 MA |
1037 | error = dsl_prop_set_int(osname, propname, |
1038 | (inherit ? ZPROP_SRC_NONE : ZPROP_SRC_LOCAL), value); | |
34dc7c2f | 1039 | |
428870ff BB |
1040 | if (error == ENOSPC) { |
1041 | ztest_record_enospc(FTAG); | |
34dc7c2f BB |
1042 | return (error); |
1043 | } | |
c99c9001 | 1044 | ASSERT0(error); |
34dc7c2f | 1045 | |
40b84e7a | 1046 | setpoint = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
13fe0198 | 1047 | VERIFY0(dsl_prop_get_integer(osname, propname, &curval, setpoint)); |
428870ff | 1048 | |
c242c188 | 1049 | if (ztest_opts.zo_verbose >= 6) { |
428870ff BB |
1050 | VERIFY(zfs_prop_index_to_string(prop, curval, &valname) == 0); |
1051 | (void) printf("%s %s = %s at '%s'\n", | |
1052 | osname, propname, valname, setpoint); | |
34dc7c2f | 1053 | } |
40b84e7a | 1054 | umem_free(setpoint, MAXPATHLEN); |
34dc7c2f BB |
1055 | |
1056 | return (error); | |
1057 | } | |
1058 | ||
1059 | static int | |
c242c188 | 1060 | ztest_spa_prop_set_uint64(zpool_prop_t prop, uint64_t value) |
34dc7c2f | 1061 | { |
c242c188 | 1062 | spa_t *spa = ztest_spa; |
428870ff | 1063 | nvlist_t *props = NULL; |
34dc7c2f BB |
1064 | int error; |
1065 | ||
428870ff BB |
1066 | VERIFY(nvlist_alloc(&props, NV_UNIQUE_NAME, 0) == 0); |
1067 | VERIFY(nvlist_add_uint64(props, zpool_prop_to_name(prop), value) == 0); | |
34dc7c2f | 1068 | |
428870ff BB |
1069 | error = spa_prop_set(spa, props); |
1070 | ||
1071 | nvlist_free(props); | |
1072 | ||
1073 | if (error == ENOSPC) { | |
1074 | ztest_record_enospc(FTAG); | |
34dc7c2f BB |
1075 | return (error); |
1076 | } | |
c99c9001 | 1077 | ASSERT0(error); |
34dc7c2f BB |
1078 | |
1079 | return (error); | |
1080 | } | |
1081 | ||
428870ff BB |
1082 | static void |
1083 | ztest_rll_init(rll_t *rll) | |
1084 | { | |
1085 | rll->rll_writer = NULL; | |
1086 | rll->rll_readers = 0; | |
1e33ac1e BB |
1087 | mutex_init(&rll->rll_lock, NULL, MUTEX_DEFAULT, NULL); |
1088 | cv_init(&rll->rll_cv, NULL, CV_DEFAULT, NULL); | |
428870ff | 1089 | } |
34dc7c2f | 1090 | |
428870ff BB |
1091 | static void |
1092 | ztest_rll_destroy(rll_t *rll) | |
34dc7c2f | 1093 | { |
428870ff BB |
1094 | ASSERT(rll->rll_writer == NULL); |
1095 | ASSERT(rll->rll_readers == 0); | |
1e33ac1e BB |
1096 | mutex_destroy(&rll->rll_lock); |
1097 | cv_destroy(&rll->rll_cv); | |
428870ff | 1098 | } |
34dc7c2f | 1099 | |
428870ff BB |
1100 | static void |
1101 | ztest_rll_lock(rll_t *rll, rl_type_t type) | |
1102 | { | |
1e33ac1e | 1103 | mutex_enter(&rll->rll_lock); |
34dc7c2f | 1104 | |
428870ff BB |
1105 | if (type == RL_READER) { |
1106 | while (rll->rll_writer != NULL) | |
1e33ac1e | 1107 | (void) cv_wait(&rll->rll_cv, &rll->rll_lock); |
428870ff BB |
1108 | rll->rll_readers++; |
1109 | } else { | |
1110 | while (rll->rll_writer != NULL || rll->rll_readers) | |
1e33ac1e | 1111 | (void) cv_wait(&rll->rll_cv, &rll->rll_lock); |
428870ff BB |
1112 | rll->rll_writer = curthread; |
1113 | } | |
34dc7c2f | 1114 | |
1e33ac1e | 1115 | mutex_exit(&rll->rll_lock); |
428870ff | 1116 | } |
34dc7c2f | 1117 | |
428870ff BB |
1118 | static void |
1119 | ztest_rll_unlock(rll_t *rll) | |
1120 | { | |
1e33ac1e | 1121 | mutex_enter(&rll->rll_lock); |
34dc7c2f | 1122 | |
428870ff BB |
1123 | if (rll->rll_writer) { |
1124 | ASSERT(rll->rll_readers == 0); | |
1125 | rll->rll_writer = NULL; | |
1126 | } else { | |
1127 | ASSERT(rll->rll_readers != 0); | |
1128 | ASSERT(rll->rll_writer == NULL); | |
1129 | rll->rll_readers--; | |
1130 | } | |
34dc7c2f | 1131 | |
428870ff | 1132 | if (rll->rll_writer == NULL && rll->rll_readers == 0) |
1e33ac1e | 1133 | cv_broadcast(&rll->rll_cv); |
428870ff | 1134 | |
1e33ac1e | 1135 | mutex_exit(&rll->rll_lock); |
34dc7c2f BB |
1136 | } |
1137 | ||
428870ff BB |
1138 | static void |
1139 | ztest_object_lock(ztest_ds_t *zd, uint64_t object, rl_type_t type) | |
b128c09f | 1140 | { |
428870ff | 1141 | rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; |
b128c09f | 1142 | |
428870ff BB |
1143 | ztest_rll_lock(rll, type); |
1144 | } | |
b128c09f | 1145 | |
428870ff BB |
1146 | static void |
1147 | ztest_object_unlock(ztest_ds_t *zd, uint64_t object) | |
1148 | { | |
1149 | rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; | |
b128c09f | 1150 | |
428870ff | 1151 | ztest_rll_unlock(rll); |
b128c09f BB |
1152 | } |
1153 | ||
428870ff BB |
1154 | static rl_t * |
1155 | ztest_range_lock(ztest_ds_t *zd, uint64_t object, uint64_t offset, | |
1156 | uint64_t size, rl_type_t type) | |
34dc7c2f | 1157 | { |
428870ff BB |
1158 | uint64_t hash = object ^ (offset % (ZTEST_RANGE_LOCKS + 1)); |
1159 | rll_t *rll = &zd->zd_range_lock[hash & (ZTEST_RANGE_LOCKS - 1)]; | |
1160 | rl_t *rl; | |
34dc7c2f | 1161 | |
428870ff BB |
1162 | rl = umem_alloc(sizeof (*rl), UMEM_NOFAIL); |
1163 | rl->rl_object = object; | |
1164 | rl->rl_offset = offset; | |
1165 | rl->rl_size = size; | |
1166 | rl->rl_lock = rll; | |
34dc7c2f | 1167 | |
428870ff BB |
1168 | ztest_rll_lock(rll, type); |
1169 | ||
1170 | return (rl); | |
1171 | } | |
34dc7c2f | 1172 | |
428870ff BB |
1173 | static void |
1174 | ztest_range_unlock(rl_t *rl) | |
1175 | { | |
1176 | rll_t *rll = rl->rl_lock; | |
34dc7c2f | 1177 | |
428870ff | 1178 | ztest_rll_unlock(rll); |
34dc7c2f | 1179 | |
428870ff BB |
1180 | umem_free(rl, sizeof (*rl)); |
1181 | } | |
34dc7c2f | 1182 | |
428870ff | 1183 | static void |
c242c188 | 1184 | ztest_zd_init(ztest_ds_t *zd, ztest_shared_ds_t *szd, objset_t *os) |
428870ff BB |
1185 | { |
1186 | zd->zd_os = os; | |
1187 | zd->zd_zilog = dmu_objset_zil(os); | |
c242c188 | 1188 | zd->zd_shared = szd; |
428870ff | 1189 | dmu_objset_name(os, zd->zd_name); |
d6320ddb | 1190 | int l; |
428870ff | 1191 | |
c242c188 CS |
1192 | if (zd->zd_shared != NULL) |
1193 | zd->zd_shared->zd_seq = 0; | |
1194 | ||
3e31d2b0 | 1195 | rw_init(&zd->zd_zilog_lock, NULL, RW_DEFAULT, NULL); |
1e33ac1e | 1196 | mutex_init(&zd->zd_dirobj_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 1197 | |
d6320ddb | 1198 | for (l = 0; l < ZTEST_OBJECT_LOCKS; l++) |
428870ff | 1199 | ztest_rll_init(&zd->zd_object_lock[l]); |
34dc7c2f | 1200 | |
d6320ddb | 1201 | for (l = 0; l < ZTEST_RANGE_LOCKS; l++) |
428870ff | 1202 | ztest_rll_init(&zd->zd_range_lock[l]); |
34dc7c2f BB |
1203 | } |
1204 | ||
428870ff BB |
1205 | static void |
1206 | ztest_zd_fini(ztest_ds_t *zd) | |
34dc7c2f | 1207 | { |
d6320ddb BB |
1208 | int l; |
1209 | ||
1e33ac1e | 1210 | mutex_destroy(&zd->zd_dirobj_lock); |
3e31d2b0 | 1211 | rw_destroy(&zd->zd_zilog_lock); |
34dc7c2f | 1212 | |
d6320ddb | 1213 | for (l = 0; l < ZTEST_OBJECT_LOCKS; l++) |
428870ff | 1214 | ztest_rll_destroy(&zd->zd_object_lock[l]); |
b128c09f | 1215 | |
d6320ddb | 1216 | for (l = 0; l < ZTEST_RANGE_LOCKS; l++) |
428870ff BB |
1217 | ztest_rll_destroy(&zd->zd_range_lock[l]); |
1218 | } | |
b128c09f | 1219 | |
428870ff | 1220 | #define TXG_MIGHTWAIT (ztest_random(10) == 0 ? TXG_NOWAIT : TXG_WAIT) |
b128c09f | 1221 | |
428870ff BB |
1222 | static uint64_t |
1223 | ztest_tx_assign(dmu_tx_t *tx, uint64_t txg_how, const char *tag) | |
1224 | { | |
1225 | uint64_t txg; | |
1226 | int error; | |
1227 | ||
1228 | /* | |
1229 | * Attempt to assign tx to some transaction group. | |
1230 | */ | |
1231 | error = dmu_tx_assign(tx, txg_how); | |
1232 | if (error) { | |
1233 | if (error == ERESTART) { | |
1234 | ASSERT(txg_how == TXG_NOWAIT); | |
1235 | dmu_tx_wait(tx); | |
1236 | } else { | |
1237 | ASSERT3U(error, ==, ENOSPC); | |
1238 | ztest_record_enospc(tag); | |
1239 | } | |
1240 | dmu_tx_abort(tx); | |
1241 | return (0); | |
1242 | } | |
1243 | txg = dmu_tx_get_txg(tx); | |
1244 | ASSERT(txg != 0); | |
1245 | return (txg); | |
1246 | } | |
1247 | ||
1248 | static void | |
1249 | ztest_pattern_set(void *buf, uint64_t size, uint64_t value) | |
1250 | { | |
1251 | uint64_t *ip = buf; | |
1252 | uint64_t *ip_end = (uint64_t *)((uintptr_t)buf + (uintptr_t)size); | |
1253 | ||
1254 | while (ip < ip_end) | |
1255 | *ip++ = value; | |
1256 | } | |
1257 | ||
1fde1e37 | 1258 | #ifndef NDEBUG |
428870ff BB |
1259 | static boolean_t |
1260 | ztest_pattern_match(void *buf, uint64_t size, uint64_t value) | |
1261 | { | |
1262 | uint64_t *ip = buf; | |
1263 | uint64_t *ip_end = (uint64_t *)((uintptr_t)buf + (uintptr_t)size); | |
1264 | uint64_t diff = 0; | |
1265 | ||
1266 | while (ip < ip_end) | |
1267 | diff |= (value - *ip++); | |
1268 | ||
1269 | return (diff == 0); | |
1270 | } | |
1fde1e37 | 1271 | #endif |
428870ff BB |
1272 | |
1273 | static void | |
1274 | ztest_bt_generate(ztest_block_tag_t *bt, objset_t *os, uint64_t object, | |
1275 | uint64_t offset, uint64_t gen, uint64_t txg, uint64_t crtxg) | |
1276 | { | |
1277 | bt->bt_magic = BT_MAGIC; | |
1278 | bt->bt_objset = dmu_objset_id(os); | |
1279 | bt->bt_object = object; | |
1280 | bt->bt_offset = offset; | |
1281 | bt->bt_gen = gen; | |
1282 | bt->bt_txg = txg; | |
1283 | bt->bt_crtxg = crtxg; | |
1284 | } | |
1285 | ||
1286 | static void | |
1287 | ztest_bt_verify(ztest_block_tag_t *bt, objset_t *os, uint64_t object, | |
1288 | uint64_t offset, uint64_t gen, uint64_t txg, uint64_t crtxg) | |
1289 | { | |
1290 | ASSERT(bt->bt_magic == BT_MAGIC); | |
1291 | ASSERT(bt->bt_objset == dmu_objset_id(os)); | |
1292 | ASSERT(bt->bt_object == object); | |
1293 | ASSERT(bt->bt_offset == offset); | |
1294 | ASSERT(bt->bt_gen <= gen); | |
1295 | ASSERT(bt->bt_txg <= txg); | |
1296 | ASSERT(bt->bt_crtxg == crtxg); | |
1297 | } | |
1298 | ||
1299 | static ztest_block_tag_t * | |
1300 | ztest_bt_bonus(dmu_buf_t *db) | |
1301 | { | |
1302 | dmu_object_info_t doi; | |
1303 | ztest_block_tag_t *bt; | |
1304 | ||
1305 | dmu_object_info_from_db(db, &doi); | |
1306 | ASSERT3U(doi.doi_bonus_size, <=, db->db_size); | |
1307 | ASSERT3U(doi.doi_bonus_size, >=, sizeof (*bt)); | |
1308 | bt = (void *)((char *)db->db_data + doi.doi_bonus_size - sizeof (*bt)); | |
1309 | ||
1310 | return (bt); | |
1311 | } | |
1312 | ||
1313 | /* | |
1314 | * ZIL logging ops | |
1315 | */ | |
1316 | ||
1317 | #define lrz_type lr_mode | |
1318 | #define lrz_blocksize lr_uid | |
1319 | #define lrz_ibshift lr_gid | |
1320 | #define lrz_bonustype lr_rdev | |
1321 | #define lrz_bonuslen lr_crtime[1] | |
1322 | ||
572e2857 | 1323 | static void |
428870ff BB |
1324 | ztest_log_create(ztest_ds_t *zd, dmu_tx_t *tx, lr_create_t *lr) |
1325 | { | |
1326 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1327 | size_t namesize = strlen(name) + 1; | |
1328 | itx_t *itx; | |
1329 | ||
1330 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1331 | return; |
428870ff BB |
1332 | |
1333 | itx = zil_itx_create(TX_CREATE, sizeof (*lr) + namesize); | |
1334 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1335 | sizeof (*lr) + namesize - sizeof (lr_t)); | |
1336 | ||
572e2857 | 1337 | zil_itx_assign(zd->zd_zilog, itx, tx); |
428870ff BB |
1338 | } |
1339 | ||
572e2857 BB |
1340 | static void |
1341 | ztest_log_remove(ztest_ds_t *zd, dmu_tx_t *tx, lr_remove_t *lr, uint64_t object) | |
428870ff BB |
1342 | { |
1343 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1344 | size_t namesize = strlen(name) + 1; | |
1345 | itx_t *itx; | |
1346 | ||
1347 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1348 | return; |
428870ff BB |
1349 | |
1350 | itx = zil_itx_create(TX_REMOVE, sizeof (*lr) + namesize); | |
1351 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1352 | sizeof (*lr) + namesize - sizeof (lr_t)); | |
1353 | ||
572e2857 BB |
1354 | itx->itx_oid = object; |
1355 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1356 | } |
1357 | ||
572e2857 | 1358 | static void |
428870ff BB |
1359 | ztest_log_write(ztest_ds_t *zd, dmu_tx_t *tx, lr_write_t *lr) |
1360 | { | |
1361 | itx_t *itx; | |
1362 | itx_wr_state_t write_state = ztest_random(WR_NUM_STATES); | |
1363 | ||
1364 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1365 | return; |
428870ff BB |
1366 | |
1367 | if (lr->lr_length > ZIL_MAX_LOG_DATA) | |
1368 | write_state = WR_INDIRECT; | |
1369 | ||
1370 | itx = zil_itx_create(TX_WRITE, | |
1371 | sizeof (*lr) + (write_state == WR_COPIED ? lr->lr_length : 0)); | |
1372 | ||
1373 | if (write_state == WR_COPIED && | |
1374 | dmu_read(zd->zd_os, lr->lr_foid, lr->lr_offset, lr->lr_length, | |
1375 | ((lr_write_t *)&itx->itx_lr) + 1, DMU_READ_NO_PREFETCH) != 0) { | |
1376 | zil_itx_destroy(itx); | |
1377 | itx = zil_itx_create(TX_WRITE, sizeof (*lr)); | |
1378 | write_state = WR_NEED_COPY; | |
1379 | } | |
1380 | itx->itx_private = zd; | |
1381 | itx->itx_wr_state = write_state; | |
1382 | itx->itx_sync = (ztest_random(8) == 0); | |
1383 | itx->itx_sod += (write_state == WR_NEED_COPY ? lr->lr_length : 0); | |
1384 | ||
1385 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1386 | sizeof (*lr) - sizeof (lr_t)); | |
1387 | ||
572e2857 | 1388 | zil_itx_assign(zd->zd_zilog, itx, tx); |
428870ff BB |
1389 | } |
1390 | ||
572e2857 | 1391 | static void |
428870ff BB |
1392 | ztest_log_truncate(ztest_ds_t *zd, dmu_tx_t *tx, lr_truncate_t *lr) |
1393 | { | |
1394 | itx_t *itx; | |
1395 | ||
1396 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1397 | return; |
428870ff BB |
1398 | |
1399 | itx = zil_itx_create(TX_TRUNCATE, sizeof (*lr)); | |
1400 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1401 | sizeof (*lr) - sizeof (lr_t)); | |
1402 | ||
572e2857 BB |
1403 | itx->itx_sync = B_FALSE; |
1404 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1405 | } |
1406 | ||
572e2857 | 1407 | static void |
428870ff BB |
1408 | ztest_log_setattr(ztest_ds_t *zd, dmu_tx_t *tx, lr_setattr_t *lr) |
1409 | { | |
1410 | itx_t *itx; | |
1411 | ||
1412 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1413 | return; |
428870ff BB |
1414 | |
1415 | itx = zil_itx_create(TX_SETATTR, sizeof (*lr)); | |
1416 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1417 | sizeof (*lr) - sizeof (lr_t)); | |
1418 | ||
572e2857 BB |
1419 | itx->itx_sync = B_FALSE; |
1420 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1421 | } |
1422 | ||
1423 | /* | |
1424 | * ZIL replay ops | |
1425 | */ | |
1426 | static int | |
1427 | ztest_replay_create(ztest_ds_t *zd, lr_create_t *lr, boolean_t byteswap) | |
1428 | { | |
1429 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1430 | objset_t *os = zd->zd_os; | |
1431 | ztest_block_tag_t *bbt; | |
1432 | dmu_buf_t *db; | |
1433 | dmu_tx_t *tx; | |
1434 | uint64_t txg; | |
1435 | int error = 0; | |
1436 | ||
1437 | if (byteswap) | |
1438 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1439 | ||
1440 | ASSERT(lr->lr_doid == ZTEST_DIROBJ); | |
1441 | ASSERT(name[0] != '\0'); | |
1442 | ||
1443 | tx = dmu_tx_create(os); | |
1444 | ||
1445 | dmu_tx_hold_zap(tx, lr->lr_doid, B_TRUE, name); | |
1446 | ||
1447 | if (lr->lrz_type == DMU_OT_ZAP_OTHER) { | |
1448 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, B_TRUE, NULL); | |
1449 | } else { | |
1450 | dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT); | |
1451 | } | |
1452 | ||
1453 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1454 | if (txg == 0) | |
1455 | return (ENOSPC); | |
1456 | ||
1457 | ASSERT(dmu_objset_zil(os)->zl_replay == !!lr->lr_foid); | |
1458 | ||
1459 | if (lr->lrz_type == DMU_OT_ZAP_OTHER) { | |
1460 | if (lr->lr_foid == 0) { | |
1461 | lr->lr_foid = zap_create(os, | |
1462 | lr->lrz_type, lr->lrz_bonustype, | |
1463 | lr->lrz_bonuslen, tx); | |
1464 | } else { | |
1465 | error = zap_create_claim(os, lr->lr_foid, | |
1466 | lr->lrz_type, lr->lrz_bonustype, | |
1467 | lr->lrz_bonuslen, tx); | |
1468 | } | |
1469 | } else { | |
1470 | if (lr->lr_foid == 0) { | |
1471 | lr->lr_foid = dmu_object_alloc(os, | |
1472 | lr->lrz_type, 0, lr->lrz_bonustype, | |
1473 | lr->lrz_bonuslen, tx); | |
1474 | } else { | |
1475 | error = dmu_object_claim(os, lr->lr_foid, | |
1476 | lr->lrz_type, 0, lr->lrz_bonustype, | |
1477 | lr->lrz_bonuslen, tx); | |
1478 | } | |
1479 | } | |
1480 | ||
1481 | if (error) { | |
1482 | ASSERT3U(error, ==, EEXIST); | |
1483 | ASSERT(zd->zd_zilog->zl_replay); | |
1484 | dmu_tx_commit(tx); | |
1485 | return (error); | |
1486 | } | |
1487 | ||
1488 | ASSERT(lr->lr_foid != 0); | |
1489 | ||
1490 | if (lr->lrz_type != DMU_OT_ZAP_OTHER) | |
1491 | VERIFY3U(0, ==, dmu_object_set_blocksize(os, lr->lr_foid, | |
1492 | lr->lrz_blocksize, lr->lrz_ibshift, tx)); | |
1493 | ||
1494 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
1495 | bbt = ztest_bt_bonus(db); | |
1496 | dmu_buf_will_dirty(db, tx); | |
1497 | ztest_bt_generate(bbt, os, lr->lr_foid, -1ULL, lr->lr_gen, txg, txg); | |
1498 | dmu_buf_rele(db, FTAG); | |
1499 | ||
1500 | VERIFY3U(0, ==, zap_add(os, lr->lr_doid, name, sizeof (uint64_t), 1, | |
1501 | &lr->lr_foid, tx)); | |
1502 | ||
1503 | (void) ztest_log_create(zd, tx, lr); | |
1504 | ||
1505 | dmu_tx_commit(tx); | |
1506 | ||
1507 | return (0); | |
1508 | } | |
1509 | ||
1510 | static int | |
1511 | ztest_replay_remove(ztest_ds_t *zd, lr_remove_t *lr, boolean_t byteswap) | |
1512 | { | |
1513 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1514 | objset_t *os = zd->zd_os; | |
1515 | dmu_object_info_t doi; | |
1516 | dmu_tx_t *tx; | |
1517 | uint64_t object, txg; | |
1518 | ||
1519 | if (byteswap) | |
1520 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1521 | ||
1522 | ASSERT(lr->lr_doid == ZTEST_DIROBJ); | |
1523 | ASSERT(name[0] != '\0'); | |
1524 | ||
1525 | VERIFY3U(0, ==, | |
1526 | zap_lookup(os, lr->lr_doid, name, sizeof (object), 1, &object)); | |
1527 | ASSERT(object != 0); | |
1528 | ||
1529 | ztest_object_lock(zd, object, RL_WRITER); | |
1530 | ||
1531 | VERIFY3U(0, ==, dmu_object_info(os, object, &doi)); | |
1532 | ||
1533 | tx = dmu_tx_create(os); | |
1534 | ||
1535 | dmu_tx_hold_zap(tx, lr->lr_doid, B_FALSE, name); | |
1536 | dmu_tx_hold_free(tx, object, 0, DMU_OBJECT_END); | |
1537 | ||
1538 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1539 | if (txg == 0) { | |
1540 | ztest_object_unlock(zd, object); | |
1541 | return (ENOSPC); | |
1542 | } | |
1543 | ||
1544 | if (doi.doi_type == DMU_OT_ZAP_OTHER) { | |
1545 | VERIFY3U(0, ==, zap_destroy(os, object, tx)); | |
1546 | } else { | |
1547 | VERIFY3U(0, ==, dmu_object_free(os, object, tx)); | |
1548 | } | |
1549 | ||
1550 | VERIFY3U(0, ==, zap_remove(os, lr->lr_doid, name, tx)); | |
1551 | ||
572e2857 | 1552 | (void) ztest_log_remove(zd, tx, lr, object); |
428870ff BB |
1553 | |
1554 | dmu_tx_commit(tx); | |
1555 | ||
1556 | ztest_object_unlock(zd, object); | |
1557 | ||
1558 | return (0); | |
1559 | } | |
1560 | ||
1561 | static int | |
1562 | ztest_replay_write(ztest_ds_t *zd, lr_write_t *lr, boolean_t byteswap) | |
1563 | { | |
1564 | objset_t *os = zd->zd_os; | |
1565 | void *data = lr + 1; /* data follows lr */ | |
1566 | uint64_t offset, length; | |
1567 | ztest_block_tag_t *bt = data; | |
1568 | ztest_block_tag_t *bbt; | |
1569 | uint64_t gen, txg, lrtxg, crtxg; | |
1570 | dmu_object_info_t doi; | |
1571 | dmu_tx_t *tx; | |
1572 | dmu_buf_t *db; | |
1573 | arc_buf_t *abuf = NULL; | |
1574 | rl_t *rl; | |
1575 | ||
1576 | if (byteswap) | |
1577 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1578 | ||
1579 | offset = lr->lr_offset; | |
1580 | length = lr->lr_length; | |
1581 | ||
1582 | /* If it's a dmu_sync() block, write the whole block */ | |
1583 | if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) { | |
1584 | uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr); | |
1585 | if (length < blocksize) { | |
1586 | offset -= offset % blocksize; | |
1587 | length = blocksize; | |
1588 | } | |
1589 | } | |
1590 | ||
1591 | if (bt->bt_magic == BSWAP_64(BT_MAGIC)) | |
1592 | byteswap_uint64_array(bt, sizeof (*bt)); | |
1593 | ||
1594 | if (bt->bt_magic != BT_MAGIC) | |
1595 | bt = NULL; | |
1596 | ||
1597 | ztest_object_lock(zd, lr->lr_foid, RL_READER); | |
1598 | rl = ztest_range_lock(zd, lr->lr_foid, offset, length, RL_WRITER); | |
1599 | ||
1600 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
1601 | ||
1602 | dmu_object_info_from_db(db, &doi); | |
1603 | ||
1604 | bbt = ztest_bt_bonus(db); | |
1605 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
1606 | gen = bbt->bt_gen; | |
1607 | crtxg = bbt->bt_crtxg; | |
1608 | lrtxg = lr->lr_common.lrc_txg; | |
1609 | ||
1610 | tx = dmu_tx_create(os); | |
1611 | ||
1612 | dmu_tx_hold_write(tx, lr->lr_foid, offset, length); | |
1613 | ||
1614 | if (ztest_random(8) == 0 && length == doi.doi_data_block_size && | |
1615 | P2PHASE(offset, length) == 0) | |
1616 | abuf = dmu_request_arcbuf(db, length); | |
1617 | ||
1618 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1619 | if (txg == 0) { | |
1620 | if (abuf != NULL) | |
1621 | dmu_return_arcbuf(abuf); | |
1622 | dmu_buf_rele(db, FTAG); | |
1623 | ztest_range_unlock(rl); | |
1624 | ztest_object_unlock(zd, lr->lr_foid); | |
1625 | return (ENOSPC); | |
1626 | } | |
1627 | ||
1628 | if (bt != NULL) { | |
1629 | /* | |
1630 | * Usually, verify the old data before writing new data -- | |
1631 | * but not always, because we also want to verify correct | |
1632 | * behavior when the data was not recently read into cache. | |
1633 | */ | |
1634 | ASSERT(offset % doi.doi_data_block_size == 0); | |
1635 | if (ztest_random(4) != 0) { | |
1636 | int prefetch = ztest_random(2) ? | |
1637 | DMU_READ_PREFETCH : DMU_READ_NO_PREFETCH; | |
1638 | ztest_block_tag_t rbt; | |
1639 | ||
1640 | VERIFY(dmu_read(os, lr->lr_foid, offset, | |
1641 | sizeof (rbt), &rbt, prefetch) == 0); | |
1642 | if (rbt.bt_magic == BT_MAGIC) { | |
1643 | ztest_bt_verify(&rbt, os, lr->lr_foid, | |
1644 | offset, gen, txg, crtxg); | |
1645 | } | |
1646 | } | |
1647 | ||
1648 | /* | |
1649 | * Writes can appear to be newer than the bonus buffer because | |
1650 | * the ztest_get_data() callback does a dmu_read() of the | |
1651 | * open-context data, which may be different than the data | |
1652 | * as it was when the write was generated. | |
1653 | */ | |
1654 | if (zd->zd_zilog->zl_replay) { | |
1655 | ztest_bt_verify(bt, os, lr->lr_foid, offset, | |
1656 | MAX(gen, bt->bt_gen), MAX(txg, lrtxg), | |
1657 | bt->bt_crtxg); | |
1658 | } | |
1659 | ||
1660 | /* | |
1661 | * Set the bt's gen/txg to the bonus buffer's gen/txg | |
1662 | * so that all of the usual ASSERTs will work. | |
1663 | */ | |
1664 | ztest_bt_generate(bt, os, lr->lr_foid, offset, gen, txg, crtxg); | |
1665 | } | |
1666 | ||
1667 | if (abuf == NULL) { | |
1668 | dmu_write(os, lr->lr_foid, offset, length, data, tx); | |
1669 | } else { | |
1670 | bcopy(data, abuf->b_data, length); | |
1671 | dmu_assign_arcbuf(db, offset, abuf, tx); | |
1672 | } | |
1673 | ||
1674 | (void) ztest_log_write(zd, tx, lr); | |
1675 | ||
1676 | dmu_buf_rele(db, FTAG); | |
1677 | ||
1678 | dmu_tx_commit(tx); | |
1679 | ||
1680 | ztest_range_unlock(rl); | |
1681 | ztest_object_unlock(zd, lr->lr_foid); | |
1682 | ||
1683 | return (0); | |
1684 | } | |
1685 | ||
1686 | static int | |
1687 | ztest_replay_truncate(ztest_ds_t *zd, lr_truncate_t *lr, boolean_t byteswap) | |
1688 | { | |
1689 | objset_t *os = zd->zd_os; | |
1690 | dmu_tx_t *tx; | |
1691 | uint64_t txg; | |
1692 | rl_t *rl; | |
1693 | ||
1694 | if (byteswap) | |
1695 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1696 | ||
1697 | ztest_object_lock(zd, lr->lr_foid, RL_READER); | |
1698 | rl = ztest_range_lock(zd, lr->lr_foid, lr->lr_offset, lr->lr_length, | |
1699 | RL_WRITER); | |
1700 | ||
1701 | tx = dmu_tx_create(os); | |
1702 | ||
1703 | dmu_tx_hold_free(tx, lr->lr_foid, lr->lr_offset, lr->lr_length); | |
1704 | ||
1705 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1706 | if (txg == 0) { | |
1707 | ztest_range_unlock(rl); | |
1708 | ztest_object_unlock(zd, lr->lr_foid); | |
1709 | return (ENOSPC); | |
1710 | } | |
1711 | ||
1712 | VERIFY(dmu_free_range(os, lr->lr_foid, lr->lr_offset, | |
1713 | lr->lr_length, tx) == 0); | |
1714 | ||
1715 | (void) ztest_log_truncate(zd, tx, lr); | |
1716 | ||
1717 | dmu_tx_commit(tx); | |
1718 | ||
1719 | ztest_range_unlock(rl); | |
1720 | ztest_object_unlock(zd, lr->lr_foid); | |
1721 | ||
1722 | return (0); | |
1723 | } | |
1724 | ||
1725 | static int | |
1726 | ztest_replay_setattr(ztest_ds_t *zd, lr_setattr_t *lr, boolean_t byteswap) | |
1727 | { | |
1728 | objset_t *os = zd->zd_os; | |
1729 | dmu_tx_t *tx; | |
1730 | dmu_buf_t *db; | |
1731 | ztest_block_tag_t *bbt; | |
1732 | uint64_t txg, lrtxg, crtxg; | |
1733 | ||
1734 | if (byteswap) | |
1735 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1736 | ||
1737 | ztest_object_lock(zd, lr->lr_foid, RL_WRITER); | |
1738 | ||
1739 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
1740 | ||
1741 | tx = dmu_tx_create(os); | |
1742 | dmu_tx_hold_bonus(tx, lr->lr_foid); | |
1743 | ||
1744 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1745 | if (txg == 0) { | |
1746 | dmu_buf_rele(db, FTAG); | |
1747 | ztest_object_unlock(zd, lr->lr_foid); | |
1748 | return (ENOSPC); | |
1749 | } | |
1750 | ||
1751 | bbt = ztest_bt_bonus(db); | |
1752 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
1753 | crtxg = bbt->bt_crtxg; | |
1754 | lrtxg = lr->lr_common.lrc_txg; | |
1755 | ||
1756 | if (zd->zd_zilog->zl_replay) { | |
1757 | ASSERT(lr->lr_size != 0); | |
1758 | ASSERT(lr->lr_mode != 0); | |
1759 | ASSERT(lrtxg != 0); | |
1760 | } else { | |
1761 | /* | |
1762 | * Randomly change the size and increment the generation. | |
1763 | */ | |
1764 | lr->lr_size = (ztest_random(db->db_size / sizeof (*bbt)) + 1) * | |
1765 | sizeof (*bbt); | |
1766 | lr->lr_mode = bbt->bt_gen + 1; | |
1767 | ASSERT(lrtxg == 0); | |
1768 | } | |
1769 | ||
1770 | /* | |
1771 | * Verify that the current bonus buffer is not newer than our txg. | |
1772 | */ | |
1773 | ztest_bt_verify(bbt, os, lr->lr_foid, -1ULL, lr->lr_mode, | |
1774 | MAX(txg, lrtxg), crtxg); | |
1775 | ||
1776 | dmu_buf_will_dirty(db, tx); | |
1777 | ||
1778 | ASSERT3U(lr->lr_size, >=, sizeof (*bbt)); | |
1779 | ASSERT3U(lr->lr_size, <=, db->db_size); | |
c99c9001 | 1780 | VERIFY0(dmu_set_bonus(db, lr->lr_size, tx)); |
428870ff BB |
1781 | bbt = ztest_bt_bonus(db); |
1782 | ||
1783 | ztest_bt_generate(bbt, os, lr->lr_foid, -1ULL, lr->lr_mode, txg, crtxg); | |
1784 | ||
1785 | dmu_buf_rele(db, FTAG); | |
1786 | ||
1787 | (void) ztest_log_setattr(zd, tx, lr); | |
1788 | ||
1789 | dmu_tx_commit(tx); | |
1790 | ||
1791 | ztest_object_unlock(zd, lr->lr_foid); | |
1792 | ||
1793 | return (0); | |
1794 | } | |
1795 | ||
b01615d5 | 1796 | zil_replay_func_t ztest_replay_vector[TX_MAX_TYPE] = { |
0bc8fd78 | 1797 | NULL, /* 0 no such transaction type */ |
b01615d5 | 1798 | (zil_replay_func_t)ztest_replay_create, /* TX_CREATE */ |
0bc8fd78 BB |
1799 | NULL, /* TX_MKDIR */ |
1800 | NULL, /* TX_MKXATTR */ | |
1801 | NULL, /* TX_SYMLINK */ | |
b01615d5 | 1802 | (zil_replay_func_t)ztest_replay_remove, /* TX_REMOVE */ |
0bc8fd78 BB |
1803 | NULL, /* TX_RMDIR */ |
1804 | NULL, /* TX_LINK */ | |
1805 | NULL, /* TX_RENAME */ | |
b01615d5 RY |
1806 | (zil_replay_func_t)ztest_replay_write, /* TX_WRITE */ |
1807 | (zil_replay_func_t)ztest_replay_truncate, /* TX_TRUNCATE */ | |
1808 | (zil_replay_func_t)ztest_replay_setattr, /* TX_SETATTR */ | |
0bc8fd78 BB |
1809 | NULL, /* TX_ACL */ |
1810 | NULL, /* TX_CREATE_ACL */ | |
1811 | NULL, /* TX_CREATE_ATTR */ | |
1812 | NULL, /* TX_CREATE_ACL_ATTR */ | |
1813 | NULL, /* TX_MKDIR_ACL */ | |
1814 | NULL, /* TX_MKDIR_ATTR */ | |
1815 | NULL, /* TX_MKDIR_ACL_ATTR */ | |
1816 | NULL, /* TX_WRITE2 */ | |
428870ff BB |
1817 | }; |
1818 | ||
1819 | /* | |
1820 | * ZIL get_data callbacks | |
1821 | */ | |
1822 | ||
1823 | static void | |
1824 | ztest_get_done(zgd_t *zgd, int error) | |
1825 | { | |
1826 | ztest_ds_t *zd = zgd->zgd_private; | |
1827 | uint64_t object = zgd->zgd_rl->rl_object; | |
1828 | ||
1829 | if (zgd->zgd_db) | |
1830 | dmu_buf_rele(zgd->zgd_db, zgd); | |
1831 | ||
1832 | ztest_range_unlock(zgd->zgd_rl); | |
1833 | ztest_object_unlock(zd, object); | |
1834 | ||
1835 | if (error == 0 && zgd->zgd_bp) | |
1836 | zil_add_block(zgd->zgd_zilog, zgd->zgd_bp); | |
1837 | ||
1838 | umem_free(zgd, sizeof (*zgd)); | |
1839 | } | |
1840 | ||
1841 | static int | |
1842 | ztest_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio) | |
1843 | { | |
1844 | ztest_ds_t *zd = arg; | |
1845 | objset_t *os = zd->zd_os; | |
1846 | uint64_t object = lr->lr_foid; | |
1847 | uint64_t offset = lr->lr_offset; | |
1848 | uint64_t size = lr->lr_length; | |
1849 | blkptr_t *bp = &lr->lr_blkptr; | |
1850 | uint64_t txg = lr->lr_common.lrc_txg; | |
1851 | uint64_t crtxg; | |
1852 | dmu_object_info_t doi; | |
1853 | dmu_buf_t *db; | |
1854 | zgd_t *zgd; | |
1855 | int error; | |
1856 | ||
1857 | ztest_object_lock(zd, object, RL_READER); | |
1858 | error = dmu_bonus_hold(os, object, FTAG, &db); | |
1859 | if (error) { | |
1860 | ztest_object_unlock(zd, object); | |
1861 | return (error); | |
1862 | } | |
1863 | ||
1864 | crtxg = ztest_bt_bonus(db)->bt_crtxg; | |
1865 | ||
1866 | if (crtxg == 0 || crtxg > txg) { | |
1867 | dmu_buf_rele(db, FTAG); | |
1868 | ztest_object_unlock(zd, object); | |
1869 | return (ENOENT); | |
1870 | } | |
1871 | ||
1872 | dmu_object_info_from_db(db, &doi); | |
1873 | dmu_buf_rele(db, FTAG); | |
1874 | db = NULL; | |
1875 | ||
1876 | zgd = umem_zalloc(sizeof (*zgd), UMEM_NOFAIL); | |
1877 | zgd->zgd_zilog = zd->zd_zilog; | |
1878 | zgd->zgd_private = zd; | |
1879 | ||
1880 | if (buf != NULL) { /* immediate write */ | |
1881 | zgd->zgd_rl = ztest_range_lock(zd, object, offset, size, | |
1882 | RL_READER); | |
1883 | ||
1884 | error = dmu_read(os, object, offset, size, buf, | |
1885 | DMU_READ_NO_PREFETCH); | |
1886 | ASSERT(error == 0); | |
1887 | } else { | |
1888 | size = doi.doi_data_block_size; | |
1889 | if (ISP2(size)) { | |
1890 | offset = P2ALIGN(offset, size); | |
1891 | } else { | |
1892 | ASSERT(offset < size); | |
1893 | offset = 0; | |
1894 | } | |
1895 | ||
1896 | zgd->zgd_rl = ztest_range_lock(zd, object, offset, size, | |
1897 | RL_READER); | |
1898 | ||
1899 | error = dmu_buf_hold(os, object, offset, zgd, &db, | |
1900 | DMU_READ_NO_PREFETCH); | |
1901 | ||
1902 | if (error == 0) { | |
1903 | zgd->zgd_db = db; | |
1904 | zgd->zgd_bp = bp; | |
1905 | ||
1906 | ASSERT(db->db_offset == offset); | |
1907 | ASSERT(db->db_size == size); | |
1908 | ||
1909 | error = dmu_sync(zio, lr->lr_common.lrc_txg, | |
1910 | ztest_get_done, zgd); | |
1911 | ||
1912 | if (error == 0) | |
1913 | return (0); | |
1914 | } | |
1915 | } | |
1916 | ||
1917 | ztest_get_done(zgd, error); | |
1918 | ||
1919 | return (error); | |
1920 | } | |
1921 | ||
1922 | static void * | |
1923 | ztest_lr_alloc(size_t lrsize, char *name) | |
1924 | { | |
1925 | char *lr; | |
1926 | size_t namesize = name ? strlen(name) + 1 : 0; | |
1927 | ||
1928 | lr = umem_zalloc(lrsize + namesize, UMEM_NOFAIL); | |
1929 | ||
1930 | if (name) | |
1931 | bcopy(name, lr + lrsize, namesize); | |
1932 | ||
1933 | return (lr); | |
1934 | } | |
1935 | ||
1936 | void | |
1937 | ztest_lr_free(void *lr, size_t lrsize, char *name) | |
1938 | { | |
1939 | size_t namesize = name ? strlen(name) + 1 : 0; | |
1940 | ||
1941 | umem_free(lr, lrsize + namesize); | |
1942 | } | |
1943 | ||
1944 | /* | |
1945 | * Lookup a bunch of objects. Returns the number of objects not found. | |
1946 | */ | |
1947 | static int | |
1948 | ztest_lookup(ztest_ds_t *zd, ztest_od_t *od, int count) | |
1949 | { | |
1950 | int missing = 0; | |
1951 | int error; | |
d6320ddb | 1952 | int i; |
428870ff | 1953 | |
1e33ac1e | 1954 | ASSERT(mutex_held(&zd->zd_dirobj_lock)); |
428870ff | 1955 | |
d6320ddb | 1956 | for (i = 0; i < count; i++, od++) { |
428870ff BB |
1957 | od->od_object = 0; |
1958 | error = zap_lookup(zd->zd_os, od->od_dir, od->od_name, | |
1959 | sizeof (uint64_t), 1, &od->od_object); | |
1960 | if (error) { | |
1961 | ASSERT(error == ENOENT); | |
1962 | ASSERT(od->od_object == 0); | |
1963 | missing++; | |
1964 | } else { | |
1965 | dmu_buf_t *db; | |
1966 | ztest_block_tag_t *bbt; | |
1967 | dmu_object_info_t doi; | |
1968 | ||
1969 | ASSERT(od->od_object != 0); | |
1970 | ASSERT(missing == 0); /* there should be no gaps */ | |
1971 | ||
1972 | ztest_object_lock(zd, od->od_object, RL_READER); | |
1973 | VERIFY3U(0, ==, dmu_bonus_hold(zd->zd_os, | |
1974 | od->od_object, FTAG, &db)); | |
1975 | dmu_object_info_from_db(db, &doi); | |
1976 | bbt = ztest_bt_bonus(db); | |
1977 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
1978 | od->od_type = doi.doi_type; | |
1979 | od->od_blocksize = doi.doi_data_block_size; | |
1980 | od->od_gen = bbt->bt_gen; | |
1981 | dmu_buf_rele(db, FTAG); | |
1982 | ztest_object_unlock(zd, od->od_object); | |
1983 | } | |
1984 | } | |
1985 | ||
1986 | return (missing); | |
1987 | } | |
1988 | ||
1989 | static int | |
1990 | ztest_create(ztest_ds_t *zd, ztest_od_t *od, int count) | |
1991 | { | |
1992 | int missing = 0; | |
d6320ddb | 1993 | int i; |
428870ff | 1994 | |
1e33ac1e | 1995 | ASSERT(mutex_held(&zd->zd_dirobj_lock)); |
428870ff | 1996 | |
d6320ddb | 1997 | for (i = 0; i < count; i++, od++) { |
428870ff BB |
1998 | if (missing) { |
1999 | od->od_object = 0; | |
2000 | missing++; | |
2001 | continue; | |
2002 | } | |
2003 | ||
2004 | lr_create_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name); | |
2005 | ||
2006 | lr->lr_doid = od->od_dir; | |
2007 | lr->lr_foid = 0; /* 0 to allocate, > 0 to claim */ | |
2008 | lr->lrz_type = od->od_crtype; | |
2009 | lr->lrz_blocksize = od->od_crblocksize; | |
2010 | lr->lrz_ibshift = ztest_random_ibshift(); | |
2011 | lr->lrz_bonustype = DMU_OT_UINT64_OTHER; | |
2012 | lr->lrz_bonuslen = dmu_bonus_max(); | |
2013 | lr->lr_gen = od->od_crgen; | |
2014 | lr->lr_crtime[0] = time(NULL); | |
2015 | ||
2016 | if (ztest_replay_create(zd, lr, B_FALSE) != 0) { | |
2017 | ASSERT(missing == 0); | |
2018 | od->od_object = 0; | |
2019 | missing++; | |
2020 | } else { | |
2021 | od->od_object = lr->lr_foid; | |
2022 | od->od_type = od->od_crtype; | |
2023 | od->od_blocksize = od->od_crblocksize; | |
2024 | od->od_gen = od->od_crgen; | |
2025 | ASSERT(od->od_object != 0); | |
2026 | } | |
2027 | ||
2028 | ztest_lr_free(lr, sizeof (*lr), od->od_name); | |
2029 | } | |
2030 | ||
2031 | return (missing); | |
2032 | } | |
2033 | ||
2034 | static int | |
2035 | ztest_remove(ztest_ds_t *zd, ztest_od_t *od, int count) | |
2036 | { | |
2037 | int missing = 0; | |
2038 | int error; | |
d6320ddb | 2039 | int i; |
428870ff | 2040 | |
1e33ac1e | 2041 | ASSERT(mutex_held(&zd->zd_dirobj_lock)); |
428870ff BB |
2042 | |
2043 | od += count - 1; | |
2044 | ||
d6320ddb | 2045 | for (i = count - 1; i >= 0; i--, od--) { |
428870ff BB |
2046 | if (missing) { |
2047 | missing++; | |
2048 | continue; | |
2049 | } | |
2050 | ||
2051 | if (od->od_object == 0) | |
2052 | continue; | |
2053 | ||
2054 | lr_remove_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name); | |
2055 | ||
2056 | lr->lr_doid = od->od_dir; | |
2057 | ||
2058 | if ((error = ztest_replay_remove(zd, lr, B_FALSE)) != 0) { | |
2059 | ASSERT3U(error, ==, ENOSPC); | |
2060 | missing++; | |
2061 | } else { | |
2062 | od->od_object = 0; | |
2063 | } | |
2064 | ztest_lr_free(lr, sizeof (*lr), od->od_name); | |
2065 | } | |
2066 | ||
2067 | return (missing); | |
2068 | } | |
2069 | ||
2070 | static int | |
2071 | ztest_write(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size, | |
2072 | void *data) | |
2073 | { | |
2074 | lr_write_t *lr; | |
2075 | int error; | |
2076 | ||
2077 | lr = ztest_lr_alloc(sizeof (*lr) + size, NULL); | |
2078 | ||
2079 | lr->lr_foid = object; | |
2080 | lr->lr_offset = offset; | |
2081 | lr->lr_length = size; | |
2082 | lr->lr_blkoff = 0; | |
2083 | BP_ZERO(&lr->lr_blkptr); | |
2084 | ||
2085 | bcopy(data, lr + 1, size); | |
2086 | ||
2087 | error = ztest_replay_write(zd, lr, B_FALSE); | |
2088 | ||
2089 | ztest_lr_free(lr, sizeof (*lr) + size, NULL); | |
2090 | ||
2091 | return (error); | |
2092 | } | |
2093 | ||
2094 | static int | |
2095 | ztest_truncate(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size) | |
2096 | { | |
2097 | lr_truncate_t *lr; | |
2098 | int error; | |
2099 | ||
2100 | lr = ztest_lr_alloc(sizeof (*lr), NULL); | |
2101 | ||
2102 | lr->lr_foid = object; | |
2103 | lr->lr_offset = offset; | |
2104 | lr->lr_length = size; | |
2105 | ||
2106 | error = ztest_replay_truncate(zd, lr, B_FALSE); | |
2107 | ||
2108 | ztest_lr_free(lr, sizeof (*lr), NULL); | |
2109 | ||
2110 | return (error); | |
2111 | } | |
2112 | ||
2113 | static int | |
2114 | ztest_setattr(ztest_ds_t *zd, uint64_t object) | |
2115 | { | |
2116 | lr_setattr_t *lr; | |
2117 | int error; | |
2118 | ||
2119 | lr = ztest_lr_alloc(sizeof (*lr), NULL); | |
2120 | ||
2121 | lr->lr_foid = object; | |
2122 | lr->lr_size = 0; | |
2123 | lr->lr_mode = 0; | |
2124 | ||
2125 | error = ztest_replay_setattr(zd, lr, B_FALSE); | |
2126 | ||
2127 | ztest_lr_free(lr, sizeof (*lr), NULL); | |
2128 | ||
2129 | return (error); | |
2130 | } | |
2131 | ||
2132 | static void | |
2133 | ztest_prealloc(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size) | |
2134 | { | |
2135 | objset_t *os = zd->zd_os; | |
2136 | dmu_tx_t *tx; | |
2137 | uint64_t txg; | |
2138 | rl_t *rl; | |
2139 | ||
2140 | txg_wait_synced(dmu_objset_pool(os), 0); | |
2141 | ||
2142 | ztest_object_lock(zd, object, RL_READER); | |
2143 | rl = ztest_range_lock(zd, object, offset, size, RL_WRITER); | |
2144 | ||
2145 | tx = dmu_tx_create(os); | |
2146 | ||
2147 | dmu_tx_hold_write(tx, object, offset, size); | |
2148 | ||
2149 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
2150 | ||
2151 | if (txg != 0) { | |
2152 | dmu_prealloc(os, object, offset, size, tx); | |
2153 | dmu_tx_commit(tx); | |
2154 | txg_wait_synced(dmu_objset_pool(os), txg); | |
2155 | } else { | |
2156 | (void) dmu_free_long_range(os, object, offset, size); | |
2157 | } | |
2158 | ||
2159 | ztest_range_unlock(rl); | |
2160 | ztest_object_unlock(zd, object); | |
2161 | } | |
2162 | ||
2163 | static void | |
2164 | ztest_io(ztest_ds_t *zd, uint64_t object, uint64_t offset) | |
2165 | { | |
2166 | ztest_block_tag_t wbt; | |
2167 | dmu_object_info_t doi; | |
2168 | enum ztest_io_type io_type; | |
2169 | uint64_t blocksize; | |
2170 | void *data; | |
2171 | ||
2172 | VERIFY(dmu_object_info(zd->zd_os, object, &doi) == 0); | |
2173 | blocksize = doi.doi_data_block_size; | |
2174 | data = umem_alloc(blocksize, UMEM_NOFAIL); | |
2175 | ||
2176 | /* | |
2177 | * Pick an i/o type at random, biased toward writing block tags. | |
2178 | */ | |
2179 | io_type = ztest_random(ZTEST_IO_TYPES); | |
2180 | if (ztest_random(2) == 0) | |
2181 | io_type = ZTEST_IO_WRITE_TAG; | |
2182 | ||
3e31d2b0 ES |
2183 | (void) rw_enter(&zd->zd_zilog_lock, RW_READER); |
2184 | ||
428870ff BB |
2185 | switch (io_type) { |
2186 | ||
2187 | case ZTEST_IO_WRITE_TAG: | |
2188 | ztest_bt_generate(&wbt, zd->zd_os, object, offset, 0, 0, 0); | |
2189 | (void) ztest_write(zd, object, offset, sizeof (wbt), &wbt); | |
2190 | break; | |
2191 | ||
2192 | case ZTEST_IO_WRITE_PATTERN: | |
2193 | (void) memset(data, 'a' + (object + offset) % 5, blocksize); | |
2194 | if (ztest_random(2) == 0) { | |
2195 | /* | |
2196 | * Induce fletcher2 collisions to ensure that | |
2197 | * zio_ddt_collision() detects and resolves them | |
2198 | * when using fletcher2-verify for deduplication. | |
2199 | */ | |
2200 | ((uint64_t *)data)[0] ^= 1ULL << 63; | |
2201 | ((uint64_t *)data)[4] ^= 1ULL << 63; | |
2202 | } | |
2203 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2204 | break; | |
2205 | ||
2206 | case ZTEST_IO_WRITE_ZEROES: | |
2207 | bzero(data, blocksize); | |
2208 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2209 | break; | |
2210 | ||
2211 | case ZTEST_IO_TRUNCATE: | |
2212 | (void) ztest_truncate(zd, object, offset, blocksize); | |
2213 | break; | |
2214 | ||
2215 | case ZTEST_IO_SETATTR: | |
2216 | (void) ztest_setattr(zd, object); | |
2217 | break; | |
e75c13c3 BB |
2218 | default: |
2219 | break; | |
428870ff BB |
2220 | } |
2221 | ||
3e31d2b0 ES |
2222 | (void) rw_exit(&zd->zd_zilog_lock); |
2223 | ||
428870ff BB |
2224 | umem_free(data, blocksize); |
2225 | } | |
2226 | ||
2227 | /* | |
2228 | * Initialize an object description template. | |
2229 | */ | |
2230 | static void | |
2231 | ztest_od_init(ztest_od_t *od, uint64_t id, char *tag, uint64_t index, | |
2232 | dmu_object_type_t type, uint64_t blocksize, uint64_t gen) | |
2233 | { | |
2234 | od->od_dir = ZTEST_DIROBJ; | |
2235 | od->od_object = 0; | |
2236 | ||
2237 | od->od_crtype = type; | |
2238 | od->od_crblocksize = blocksize ? blocksize : ztest_random_blocksize(); | |
2239 | od->od_crgen = gen; | |
2240 | ||
2241 | od->od_type = DMU_OT_NONE; | |
2242 | od->od_blocksize = 0; | |
2243 | od->od_gen = 0; | |
2244 | ||
2245 | (void) snprintf(od->od_name, sizeof (od->od_name), "%s(%lld)[%llu]", | |
b8864a23 | 2246 | tag, (longlong_t)id, (u_longlong_t)index); |
428870ff BB |
2247 | } |
2248 | ||
2249 | /* | |
2250 | * Lookup or create the objects for a test using the od template. | |
2251 | * If the objects do not all exist, or if 'remove' is specified, | |
2252 | * remove any existing objects and create new ones. Otherwise, | |
2253 | * use the existing objects. | |
2254 | */ | |
2255 | static int | |
2256 | ztest_object_init(ztest_ds_t *zd, ztest_od_t *od, size_t size, boolean_t remove) | |
2257 | { | |
2258 | int count = size / sizeof (*od); | |
2259 | int rv = 0; | |
2260 | ||
1e33ac1e | 2261 | mutex_enter(&zd->zd_dirobj_lock); |
428870ff BB |
2262 | if ((ztest_lookup(zd, od, count) != 0 || remove) && |
2263 | (ztest_remove(zd, od, count) != 0 || | |
2264 | ztest_create(zd, od, count) != 0)) | |
2265 | rv = -1; | |
2266 | zd->zd_od = od; | |
1e33ac1e | 2267 | mutex_exit(&zd->zd_dirobj_lock); |
428870ff BB |
2268 | |
2269 | return (rv); | |
2270 | } | |
2271 | ||
2272 | /* ARGSUSED */ | |
2273 | void | |
2274 | ztest_zil_commit(ztest_ds_t *zd, uint64_t id) | |
2275 | { | |
2276 | zilog_t *zilog = zd->zd_zilog; | |
2277 | ||
3e31d2b0 ES |
2278 | (void) rw_enter(&zd->zd_zilog_lock, RW_READER); |
2279 | ||
572e2857 | 2280 | zil_commit(zilog, ztest_random(ZTEST_OBJECTS)); |
428870ff BB |
2281 | |
2282 | /* | |
2283 | * Remember the committed values in zd, which is in parent/child | |
2284 | * shared memory. If we die, the next iteration of ztest_run() | |
2285 | * will verify that the log really does contain this record. | |
2286 | */ | |
2287 | mutex_enter(&zilog->zl_lock); | |
c242c188 CS |
2288 | ASSERT(zd->zd_shared != NULL); |
2289 | ASSERT3U(zd->zd_shared->zd_seq, <=, zilog->zl_commit_lr_seq); | |
2290 | zd->zd_shared->zd_seq = zilog->zl_commit_lr_seq; | |
428870ff | 2291 | mutex_exit(&zilog->zl_lock); |
3e31d2b0 ES |
2292 | |
2293 | (void) rw_exit(&zd->zd_zilog_lock); | |
2294 | } | |
2295 | ||
2296 | /* | |
2297 | * This function is designed to simulate the operations that occur during a | |
2298 | * mount/unmount operation. We hold the dataset across these operations in an | |
2299 | * attempt to expose any implicit assumptions about ZIL management. | |
2300 | */ | |
2301 | /* ARGSUSED */ | |
2302 | void | |
2303 | ztest_zil_remount(ztest_ds_t *zd, uint64_t id) | |
2304 | { | |
2305 | objset_t *os = zd->zd_os; | |
2306 | ||
29809a6c | 2307 | mutex_enter(&zd->zd_dirobj_lock); |
3e31d2b0 ES |
2308 | (void) rw_enter(&zd->zd_zilog_lock, RW_WRITER); |
2309 | ||
86f35f34 | 2310 | /* zfs_sb_teardown() */ |
3e31d2b0 ES |
2311 | zil_close(zd->zd_zilog); |
2312 | ||
2313 | /* zfsvfs_setup() */ | |
2314 | VERIFY(zil_open(os, ztest_get_data) == zd->zd_zilog); | |
2315 | zil_replay(os, zd, ztest_replay_vector); | |
2316 | ||
2317 | (void) rw_exit(&zd->zd_zilog_lock); | |
29809a6c | 2318 | mutex_exit(&zd->zd_dirobj_lock); |
428870ff BB |
2319 | } |
2320 | ||
2321 | /* | |
2322 | * Verify that we can't destroy an active pool, create an existing pool, | |
2323 | * or create a pool with a bad vdev spec. | |
2324 | */ | |
2325 | /* ARGSUSED */ | |
2326 | void | |
2327 | ztest_spa_create_destroy(ztest_ds_t *zd, uint64_t id) | |
2328 | { | |
c242c188 | 2329 | ztest_shared_opts_t *zo = &ztest_opts; |
428870ff BB |
2330 | spa_t *spa; |
2331 | nvlist_t *nvroot; | |
2332 | ||
2333 | /* | |
2334 | * Attempt to create using a bad file. | |
2335 | */ | |
ea0b2538 | 2336 | nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, 0, 0, 0, 1); |
428870ff | 2337 | VERIFY3U(ENOENT, ==, |
6f1ffb06 | 2338 | spa_create("ztest_bad_file", nvroot, NULL, NULL)); |
428870ff BB |
2339 | nvlist_free(nvroot); |
2340 | ||
2341 | /* | |
2342 | * Attempt to create using a bad mirror. | |
2343 | */ | |
ea0b2538 | 2344 | nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, 0, 0, 2, 1); |
428870ff | 2345 | VERIFY3U(ENOENT, ==, |
6f1ffb06 | 2346 | spa_create("ztest_bad_mirror", nvroot, NULL, NULL)); |
428870ff BB |
2347 | nvlist_free(nvroot); |
2348 | ||
2349 | /* | |
2350 | * Attempt to create an existing pool. It shouldn't matter | |
2351 | * what's in the nvroot; we should fail with EEXIST. | |
2352 | */ | |
c242c188 | 2353 | (void) rw_enter(&ztest_name_lock, RW_READER); |
ea0b2538 | 2354 | nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, 0, 0, 0, 1); |
6f1ffb06 | 2355 | VERIFY3U(EEXIST, ==, spa_create(zo->zo_pool, nvroot, NULL, NULL)); |
428870ff | 2356 | nvlist_free(nvroot); |
c242c188 CS |
2357 | VERIFY3U(0, ==, spa_open(zo->zo_pool, &spa, FTAG)); |
2358 | VERIFY3U(EBUSY, ==, spa_destroy(zo->zo_pool)); | |
428870ff BB |
2359 | spa_close(spa, FTAG); |
2360 | ||
c242c188 | 2361 | (void) rw_exit(&ztest_name_lock); |
428870ff BB |
2362 | } |
2363 | ||
ea0b2538 GW |
2364 | /* ARGSUSED */ |
2365 | void | |
2366 | ztest_spa_upgrade(ztest_ds_t *zd, uint64_t id) | |
2367 | { | |
2368 | spa_t *spa; | |
2369 | uint64_t initial_version = SPA_VERSION_INITIAL; | |
2370 | uint64_t version, newversion; | |
2371 | nvlist_t *nvroot, *props; | |
2372 | char *name; | |
2373 | ||
2374 | mutex_enter(&ztest_vdev_lock); | |
2375 | name = kmem_asprintf("%s_upgrade", ztest_opts.zo_pool); | |
2376 | ||
2377 | /* | |
2378 | * Clean up from previous runs. | |
2379 | */ | |
2380 | (void) spa_destroy(name); | |
2381 | ||
2382 | nvroot = make_vdev_root(NULL, NULL, name, ztest_opts.zo_vdev_size, 0, | |
2383 | 0, ztest_opts.zo_raidz, ztest_opts.zo_mirrors, 1); | |
2384 | ||
2385 | /* | |
2386 | * If we're configuring a RAIDZ device then make sure that the | |
2387 | * the initial version is capable of supporting that feature. | |
2388 | */ | |
2389 | switch (ztest_opts.zo_raidz_parity) { | |
2390 | case 0: | |
2391 | case 1: | |
2392 | initial_version = SPA_VERSION_INITIAL; | |
2393 | break; | |
2394 | case 2: | |
2395 | initial_version = SPA_VERSION_RAIDZ2; | |
2396 | break; | |
2397 | case 3: | |
2398 | initial_version = SPA_VERSION_RAIDZ3; | |
2399 | break; | |
2400 | } | |
2401 | ||
2402 | /* | |
2403 | * Create a pool with a spa version that can be upgraded. Pick | |
2404 | * a value between initial_version and SPA_VERSION_BEFORE_FEATURES. | |
2405 | */ | |
2406 | do { | |
2407 | version = ztest_random_spa_version(initial_version); | |
2408 | } while (version > SPA_VERSION_BEFORE_FEATURES); | |
2409 | ||
2410 | props = fnvlist_alloc(); | |
2411 | fnvlist_add_uint64(props, | |
2412 | zpool_prop_to_name(ZPOOL_PROP_VERSION), version); | |
6f1ffb06 | 2413 | VERIFY3S(spa_create(name, nvroot, props, NULL), ==, 0); |
ea0b2538 GW |
2414 | fnvlist_free(nvroot); |
2415 | fnvlist_free(props); | |
2416 | ||
2417 | VERIFY3S(spa_open(name, &spa, FTAG), ==, 0); | |
2418 | VERIFY3U(spa_version(spa), ==, version); | |
2419 | newversion = ztest_random_spa_version(version + 1); | |
2420 | ||
2421 | if (ztest_opts.zo_verbose >= 4) { | |
2422 | (void) printf("upgrading spa version from %llu to %llu\n", | |
2423 | (u_longlong_t)version, (u_longlong_t)newversion); | |
2424 | } | |
2425 | ||
2426 | spa_upgrade(spa, newversion); | |
2427 | VERIFY3U(spa_version(spa), >, version); | |
2428 | VERIFY3U(spa_version(spa), ==, fnvlist_lookup_uint64(spa->spa_config, | |
2429 | zpool_prop_to_name(ZPOOL_PROP_VERSION))); | |
2430 | spa_close(spa, FTAG); | |
2431 | ||
2432 | strfree(name); | |
2433 | mutex_exit(&ztest_vdev_lock); | |
2434 | } | |
2435 | ||
428870ff BB |
2436 | static vdev_t * |
2437 | vdev_lookup_by_path(vdev_t *vd, const char *path) | |
2438 | { | |
2439 | vdev_t *mvd; | |
d6320ddb | 2440 | int c; |
428870ff BB |
2441 | |
2442 | if (vd->vdev_path != NULL && strcmp(path, vd->vdev_path) == 0) | |
2443 | return (vd); | |
2444 | ||
d6320ddb | 2445 | for (c = 0; c < vd->vdev_children; c++) |
428870ff BB |
2446 | if ((mvd = vdev_lookup_by_path(vd->vdev_child[c], path)) != |
2447 | NULL) | |
2448 | return (mvd); | |
2449 | ||
2450 | return (NULL); | |
2451 | } | |
2452 | ||
2453 | /* | |
2454 | * Find the first available hole which can be used as a top-level. | |
2455 | */ | |
2456 | int | |
2457 | find_vdev_hole(spa_t *spa) | |
2458 | { | |
2459 | vdev_t *rvd = spa->spa_root_vdev; | |
2460 | int c; | |
2461 | ||
2462 | ASSERT(spa_config_held(spa, SCL_VDEV, RW_READER) == SCL_VDEV); | |
2463 | ||
2464 | for (c = 0; c < rvd->vdev_children; c++) { | |
2465 | vdev_t *cvd = rvd->vdev_child[c]; | |
2466 | ||
2467 | if (cvd->vdev_ishole) | |
2468 | break; | |
2469 | } | |
2470 | return (c); | |
2471 | } | |
2472 | ||
2473 | /* | |
2474 | * Verify that vdev_add() works as expected. | |
2475 | */ | |
2476 | /* ARGSUSED */ | |
2477 | void | |
2478 | ztest_vdev_add_remove(ztest_ds_t *zd, uint64_t id) | |
2479 | { | |
2480 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 2481 | spa_t *spa = ztest_spa; |
428870ff BB |
2482 | uint64_t leaves; |
2483 | uint64_t guid; | |
2484 | nvlist_t *nvroot; | |
2485 | int error; | |
2486 | ||
c242c188 | 2487 | mutex_enter(&ztest_vdev_lock); |
13fe0198 | 2488 | leaves = MAX(zs->zs_mirrors + zs->zs_splits, 1) * ztest_opts.zo_raidz; |
428870ff BB |
2489 | |
2490 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
2491 | ||
2492 | ztest_shared->zs_vdev_next_leaf = find_vdev_hole(spa) * leaves; | |
2493 | ||
2494 | /* | |
2495 | * If we have slogs then remove them 1/4 of the time. | |
2496 | */ | |
2497 | if (spa_has_slogs(spa) && ztest_random(4) == 0) { | |
2498 | /* | |
2499 | * Grab the guid from the head of the log class rotor. | |
2500 | */ | |
2501 | guid = spa_log_class(spa)->mc_rotor->mg_vd->vdev_guid; | |
2502 | ||
2503 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
2504 | ||
2505 | /* | |
2506 | * We have to grab the zs_name_lock as writer to | |
2507 | * prevent a race between removing a slog (dmu_objset_find) | |
2508 | * and destroying a dataset. Removing the slog will | |
2509 | * grab a reference on the dataset which may cause | |
13fe0198 | 2510 | * dsl_destroy_head() to fail with EBUSY thus |
428870ff BB |
2511 | * leaving the dataset in an inconsistent state. |
2512 | */ | |
c242c188 | 2513 | rw_enter(&ztest_name_lock, RW_WRITER); |
428870ff | 2514 | error = spa_vdev_remove(spa, guid, B_FALSE); |
c242c188 | 2515 | rw_exit(&ztest_name_lock); |
428870ff BB |
2516 | |
2517 | if (error && error != EEXIST) | |
2518 | fatal(0, "spa_vdev_remove() = %d", error); | |
2519 | } else { | |
2520 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
2521 | ||
2522 | /* | |
2523 | * Make 1/4 of the devices be log devices. | |
2524 | */ | |
ea0b2538 | 2525 | nvroot = make_vdev_root(NULL, NULL, NULL, |
c242c188 CS |
2526 | ztest_opts.zo_vdev_size, 0, |
2527 | ztest_random(4) == 0, ztest_opts.zo_raidz, | |
2528 | zs->zs_mirrors, 1); | |
428870ff BB |
2529 | |
2530 | error = spa_vdev_add(spa, nvroot); | |
2531 | nvlist_free(nvroot); | |
2532 | ||
2533 | if (error == ENOSPC) | |
2534 | ztest_record_enospc("spa_vdev_add"); | |
2535 | else if (error != 0) | |
2536 | fatal(0, "spa_vdev_add() = %d", error); | |
2537 | } | |
2538 | ||
c242c188 | 2539 | mutex_exit(&ztest_vdev_lock); |
428870ff BB |
2540 | } |
2541 | ||
2542 | /* | |
2543 | * Verify that adding/removing aux devices (l2arc, hot spare) works as expected. | |
2544 | */ | |
2545 | /* ARGSUSED */ | |
2546 | void | |
2547 | ztest_vdev_aux_add_remove(ztest_ds_t *zd, uint64_t id) | |
2548 | { | |
2549 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 2550 | spa_t *spa = ztest_spa; |
428870ff BB |
2551 | vdev_t *rvd = spa->spa_root_vdev; |
2552 | spa_aux_vdev_t *sav; | |
2553 | char *aux; | |
40b84e7a | 2554 | char *path; |
428870ff BB |
2555 | uint64_t guid = 0; |
2556 | int error; | |
2557 | ||
40b84e7a BB |
2558 | path = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
2559 | ||
428870ff BB |
2560 | if (ztest_random(2) == 0) { |
2561 | sav = &spa->spa_spares; | |
2562 | aux = ZPOOL_CONFIG_SPARES; | |
2563 | } else { | |
2564 | sav = &spa->spa_l2cache; | |
2565 | aux = ZPOOL_CONFIG_L2CACHE; | |
2566 | } | |
2567 | ||
c242c188 | 2568 | mutex_enter(&ztest_vdev_lock); |
428870ff BB |
2569 | |
2570 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
2571 | ||
2572 | if (sav->sav_count != 0 && ztest_random(4) == 0) { | |
b128c09f BB |
2573 | /* |
2574 | * Pick a random device to remove. | |
2575 | */ | |
2576 | guid = sav->sav_vdevs[ztest_random(sav->sav_count)]->vdev_guid; | |
2577 | } else { | |
2578 | /* | |
2579 | * Find an unused device we can add. | |
2580 | */ | |
428870ff | 2581 | zs->zs_vdev_aux = 0; |
b128c09f | 2582 | for (;;) { |
b128c09f | 2583 | int c; |
3db3ff4a | 2584 | (void) snprintf(path, MAXPATHLEN, ztest_aux_template, |
c242c188 CS |
2585 | ztest_opts.zo_dir, ztest_opts.zo_pool, aux, |
2586 | zs->zs_vdev_aux); | |
b128c09f BB |
2587 | for (c = 0; c < sav->sav_count; c++) |
2588 | if (strcmp(sav->sav_vdevs[c]->vdev_path, | |
2589 | path) == 0) | |
2590 | break; | |
2591 | if (c == sav->sav_count && | |
2592 | vdev_lookup_by_path(rvd, path) == NULL) | |
2593 | break; | |
428870ff | 2594 | zs->zs_vdev_aux++; |
34dc7c2f BB |
2595 | } |
2596 | } | |
2597 | ||
b128c09f | 2598 | spa_config_exit(spa, SCL_VDEV, FTAG); |
34dc7c2f | 2599 | |
b128c09f BB |
2600 | if (guid == 0) { |
2601 | /* | |
2602 | * Add a new device. | |
2603 | */ | |
ea0b2538 | 2604 | nvlist_t *nvroot = make_vdev_root(NULL, aux, NULL, |
c242c188 | 2605 | (ztest_opts.zo_vdev_size * 5) / 4, 0, 0, 0, 0, 1); |
b128c09f BB |
2606 | error = spa_vdev_add(spa, nvroot); |
2607 | if (error != 0) | |
2608 | fatal(0, "spa_vdev_add(%p) = %d", nvroot, error); | |
2609 | nvlist_free(nvroot); | |
2610 | } else { | |
2611 | /* | |
2612 | * Remove an existing device. Sometimes, dirty its | |
2613 | * vdev state first to make sure we handle removal | |
2614 | * of devices that have pending state changes. | |
2615 | */ | |
2616 | if (ztest_random(2) == 0) | |
9babb374 | 2617 | (void) vdev_online(spa, guid, 0, NULL); |
b128c09f BB |
2618 | |
2619 | error = spa_vdev_remove(spa, guid, B_FALSE); | |
2620 | if (error != 0 && error != EBUSY) | |
2621 | fatal(0, "spa_vdev_remove(%llu) = %d", guid, error); | |
2622 | } | |
2623 | ||
c242c188 | 2624 | mutex_exit(&ztest_vdev_lock); |
40b84e7a BB |
2625 | |
2626 | umem_free(path, MAXPATHLEN); | |
428870ff BB |
2627 | } |
2628 | ||
2629 | /* | |
2630 | * split a pool if it has mirror tlvdevs | |
2631 | */ | |
2632 | /* ARGSUSED */ | |
2633 | void | |
2634 | ztest_split_pool(ztest_ds_t *zd, uint64_t id) | |
2635 | { | |
2636 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 2637 | spa_t *spa = ztest_spa; |
428870ff BB |
2638 | vdev_t *rvd = spa->spa_root_vdev; |
2639 | nvlist_t *tree, **child, *config, *split, **schild; | |
2640 | uint_t c, children, schildren = 0, lastlogid = 0; | |
2641 | int error = 0; | |
2642 | ||
c242c188 | 2643 | mutex_enter(&ztest_vdev_lock); |
428870ff BB |
2644 | |
2645 | /* ensure we have a useable config; mirrors of raidz aren't supported */ | |
c242c188 CS |
2646 | if (zs->zs_mirrors < 3 || ztest_opts.zo_raidz > 1) { |
2647 | mutex_exit(&ztest_vdev_lock); | |
428870ff BB |
2648 | return; |
2649 | } | |
2650 | ||
2651 | /* clean up the old pool, if any */ | |
2652 | (void) spa_destroy("splitp"); | |
2653 | ||
2654 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
2655 | ||
2656 | /* generate a config from the existing config */ | |
2657 | mutex_enter(&spa->spa_props_lock); | |
2658 | VERIFY(nvlist_lookup_nvlist(spa->spa_config, ZPOOL_CONFIG_VDEV_TREE, | |
2659 | &tree) == 0); | |
2660 | mutex_exit(&spa->spa_props_lock); | |
2661 | ||
2662 | VERIFY(nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child, | |
2663 | &children) == 0); | |
2664 | ||
2665 | schild = malloc(rvd->vdev_children * sizeof (nvlist_t *)); | |
2666 | for (c = 0; c < children; c++) { | |
2667 | vdev_t *tvd = rvd->vdev_child[c]; | |
2668 | nvlist_t **mchild; | |
2669 | uint_t mchildren; | |
2670 | ||
2671 | if (tvd->vdev_islog || tvd->vdev_ops == &vdev_hole_ops) { | |
2672 | VERIFY(nvlist_alloc(&schild[schildren], NV_UNIQUE_NAME, | |
2673 | 0) == 0); | |
2674 | VERIFY(nvlist_add_string(schild[schildren], | |
2675 | ZPOOL_CONFIG_TYPE, VDEV_TYPE_HOLE) == 0); | |
2676 | VERIFY(nvlist_add_uint64(schild[schildren], | |
2677 | ZPOOL_CONFIG_IS_HOLE, 1) == 0); | |
2678 | if (lastlogid == 0) | |
2679 | lastlogid = schildren; | |
2680 | ++schildren; | |
2681 | continue; | |
2682 | } | |
2683 | lastlogid = 0; | |
2684 | VERIFY(nvlist_lookup_nvlist_array(child[c], | |
2685 | ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0); | |
2686 | VERIFY(nvlist_dup(mchild[0], &schild[schildren++], 0) == 0); | |
2687 | } | |
2688 | ||
2689 | /* OK, create a config that can be used to split */ | |
2690 | VERIFY(nvlist_alloc(&split, NV_UNIQUE_NAME, 0) == 0); | |
2691 | VERIFY(nvlist_add_string(split, ZPOOL_CONFIG_TYPE, | |
2692 | VDEV_TYPE_ROOT) == 0); | |
2693 | VERIFY(nvlist_add_nvlist_array(split, ZPOOL_CONFIG_CHILDREN, schild, | |
2694 | lastlogid != 0 ? lastlogid : schildren) == 0); | |
2695 | ||
2696 | VERIFY(nvlist_alloc(&config, NV_UNIQUE_NAME, 0) == 0); | |
2697 | VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, split) == 0); | |
2698 | ||
2699 | for (c = 0; c < schildren; c++) | |
2700 | nvlist_free(schild[c]); | |
2701 | free(schild); | |
2702 | nvlist_free(split); | |
2703 | ||
2704 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
2705 | ||
c242c188 | 2706 | (void) rw_enter(&ztest_name_lock, RW_WRITER); |
428870ff | 2707 | error = spa_vdev_split_mirror(spa, "splitp", config, NULL, B_FALSE); |
c242c188 | 2708 | (void) rw_exit(&ztest_name_lock); |
428870ff BB |
2709 | |
2710 | nvlist_free(config); | |
2711 | ||
2712 | if (error == 0) { | |
2713 | (void) printf("successful split - results:\n"); | |
2714 | mutex_enter(&spa_namespace_lock); | |
2715 | show_pool_stats(spa); | |
2716 | show_pool_stats(spa_lookup("splitp")); | |
2717 | mutex_exit(&spa_namespace_lock); | |
2718 | ++zs->zs_splits; | |
2719 | --zs->zs_mirrors; | |
2720 | } | |
c242c188 | 2721 | mutex_exit(&ztest_vdev_lock); |
428870ff | 2722 | |
34dc7c2f BB |
2723 | } |
2724 | ||
2725 | /* | |
2726 | * Verify that we can attach and detach devices. | |
2727 | */ | |
428870ff | 2728 | /* ARGSUSED */ |
34dc7c2f | 2729 | void |
428870ff | 2730 | ztest_vdev_attach_detach(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 2731 | { |
428870ff | 2732 | ztest_shared_t *zs = ztest_shared; |
c242c188 | 2733 | spa_t *spa = ztest_spa; |
b128c09f | 2734 | spa_aux_vdev_t *sav = &spa->spa_spares; |
34dc7c2f BB |
2735 | vdev_t *rvd = spa->spa_root_vdev; |
2736 | vdev_t *oldvd, *newvd, *pvd; | |
b128c09f | 2737 | nvlist_t *root; |
428870ff | 2738 | uint64_t leaves; |
34dc7c2f BB |
2739 | uint64_t leaf, top; |
2740 | uint64_t ashift = ztest_get_ashift(); | |
fb5f0bc8 | 2741 | uint64_t oldguid, pguid; |
34dc7c2f | 2742 | size_t oldsize, newsize; |
40b84e7a | 2743 | char *oldpath, *newpath; |
34dc7c2f | 2744 | int replacing; |
b128c09f BB |
2745 | int oldvd_has_siblings = B_FALSE; |
2746 | int newvd_is_spare = B_FALSE; | |
2747 | int oldvd_is_log; | |
34dc7c2f | 2748 | int error, expected_error; |
34dc7c2f | 2749 | |
40b84e7a BB |
2750 | oldpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
2751 | newpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
2752 | ||
c242c188 CS |
2753 | mutex_enter(&ztest_vdev_lock); |
2754 | leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raidz; | |
34dc7c2f | 2755 | |
b128c09f | 2756 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); |
34dc7c2f BB |
2757 | |
2758 | /* | |
2759 | * Decide whether to do an attach or a replace. | |
2760 | */ | |
2761 | replacing = ztest_random(2); | |
2762 | ||
2763 | /* | |
2764 | * Pick a random top-level vdev. | |
2765 | */ | |
428870ff | 2766 | top = ztest_random_vdev_top(spa, B_TRUE); |
34dc7c2f BB |
2767 | |
2768 | /* | |
2769 | * Pick a random leaf within it. | |
2770 | */ | |
2771 | leaf = ztest_random(leaves); | |
2772 | ||
2773 | /* | |
b128c09f | 2774 | * Locate this vdev. |
34dc7c2f | 2775 | */ |
b128c09f | 2776 | oldvd = rvd->vdev_child[top]; |
428870ff | 2777 | if (zs->zs_mirrors >= 1) { |
fb5f0bc8 | 2778 | ASSERT(oldvd->vdev_ops == &vdev_mirror_ops); |
428870ff | 2779 | ASSERT(oldvd->vdev_children >= zs->zs_mirrors); |
c242c188 | 2780 | oldvd = oldvd->vdev_child[leaf / ztest_opts.zo_raidz]; |
fb5f0bc8 | 2781 | } |
c242c188 | 2782 | if (ztest_opts.zo_raidz > 1) { |
fb5f0bc8 | 2783 | ASSERT(oldvd->vdev_ops == &vdev_raidz_ops); |
c242c188 CS |
2784 | ASSERT(oldvd->vdev_children == ztest_opts.zo_raidz); |
2785 | oldvd = oldvd->vdev_child[leaf % ztest_opts.zo_raidz]; | |
fb5f0bc8 | 2786 | } |
34dc7c2f BB |
2787 | |
2788 | /* | |
b128c09f BB |
2789 | * If we're already doing an attach or replace, oldvd may be a |
2790 | * mirror vdev -- in which case, pick a random child. | |
34dc7c2f | 2791 | */ |
b128c09f BB |
2792 | while (oldvd->vdev_children != 0) { |
2793 | oldvd_has_siblings = B_TRUE; | |
fb5f0bc8 BB |
2794 | ASSERT(oldvd->vdev_children >= 2); |
2795 | oldvd = oldvd->vdev_child[ztest_random(oldvd->vdev_children)]; | |
b128c09f BB |
2796 | } |
2797 | ||
2798 | oldguid = oldvd->vdev_guid; | |
9babb374 | 2799 | oldsize = vdev_get_min_asize(oldvd); |
b128c09f BB |
2800 | oldvd_is_log = oldvd->vdev_top->vdev_islog; |
2801 | (void) strcpy(oldpath, oldvd->vdev_path); | |
2802 | pvd = oldvd->vdev_parent; | |
fb5f0bc8 | 2803 | pguid = pvd->vdev_guid; |
34dc7c2f BB |
2804 | |
2805 | /* | |
b128c09f | 2806 | * If oldvd has siblings, then half of the time, detach it. |
34dc7c2f | 2807 | */ |
b128c09f BB |
2808 | if (oldvd_has_siblings && ztest_random(2) == 0) { |
2809 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
fb5f0bc8 BB |
2810 | error = spa_vdev_detach(spa, oldguid, pguid, B_FALSE); |
2811 | if (error != 0 && error != ENODEV && error != EBUSY && | |
2812 | error != ENOTSUP) | |
2813 | fatal(0, "detach (%s) returned %d", oldpath, error); | |
40b84e7a | 2814 | goto out; |
b128c09f | 2815 | } |
34dc7c2f BB |
2816 | |
2817 | /* | |
b128c09f BB |
2818 | * For the new vdev, choose with equal probability between the two |
2819 | * standard paths (ending in either 'a' or 'b') or a random hot spare. | |
34dc7c2f | 2820 | */ |
b128c09f BB |
2821 | if (sav->sav_count != 0 && ztest_random(3) == 0) { |
2822 | newvd = sav->sav_vdevs[ztest_random(sav->sav_count)]; | |
2823 | newvd_is_spare = B_TRUE; | |
2824 | (void) strcpy(newpath, newvd->vdev_path); | |
2825 | } else { | |
6aec1cd5 | 2826 | (void) snprintf(newpath, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
2827 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
2828 | top * leaves + leaf); | |
b128c09f BB |
2829 | if (ztest_random(2) == 0) |
2830 | newpath[strlen(newpath) - 1] = 'b'; | |
2831 | newvd = vdev_lookup_by_path(rvd, newpath); | |
2832 | } | |
2833 | ||
2834 | if (newvd) { | |
9babb374 | 2835 | newsize = vdev_get_min_asize(newvd); |
b128c09f BB |
2836 | } else { |
2837 | /* | |
2838 | * Make newsize a little bigger or smaller than oldsize. | |
2839 | * If it's smaller, the attach should fail. | |
2840 | * If it's larger, and we're doing a replace, | |
2841 | * we should get dynamic LUN growth when we're done. | |
2842 | */ | |
2843 | newsize = 10 * oldsize / (9 + ztest_random(3)); | |
2844 | } | |
34dc7c2f BB |
2845 | |
2846 | /* | |
2847 | * If pvd is not a mirror or root, the attach should fail with ENOTSUP, | |
2848 | * unless it's a replace; in that case any non-replacing parent is OK. | |
2849 | * | |
2850 | * If newvd is already part of the pool, it should fail with EBUSY. | |
2851 | * | |
2852 | * If newvd is too small, it should fail with EOVERFLOW. | |
2853 | */ | |
b128c09f BB |
2854 | if (pvd->vdev_ops != &vdev_mirror_ops && |
2855 | pvd->vdev_ops != &vdev_root_ops && (!replacing || | |
2856 | pvd->vdev_ops == &vdev_replacing_ops || | |
2857 | pvd->vdev_ops == &vdev_spare_ops)) | |
34dc7c2f | 2858 | expected_error = ENOTSUP; |
b128c09f BB |
2859 | else if (newvd_is_spare && (!replacing || oldvd_is_log)) |
2860 | expected_error = ENOTSUP; | |
2861 | else if (newvd == oldvd) | |
2862 | expected_error = replacing ? 0 : EBUSY; | |
2863 | else if (vdev_lookup_by_path(rvd, newpath) != NULL) | |
2864 | expected_error = EBUSY; | |
34dc7c2f BB |
2865 | else if (newsize < oldsize) |
2866 | expected_error = EOVERFLOW; | |
2867 | else if (ashift > oldvd->vdev_top->vdev_ashift) | |
2868 | expected_error = EDOM; | |
2869 | else | |
2870 | expected_error = 0; | |
2871 | ||
b128c09f | 2872 | spa_config_exit(spa, SCL_VDEV, FTAG); |
34dc7c2f BB |
2873 | |
2874 | /* | |
2875 | * Build the nvlist describing newpath. | |
2876 | */ | |
ea0b2538 | 2877 | root = make_vdev_root(newpath, NULL, NULL, newvd == NULL ? newsize : 0, |
b128c09f | 2878 | ashift, 0, 0, 0, 1); |
34dc7c2f | 2879 | |
b128c09f | 2880 | error = spa_vdev_attach(spa, oldguid, root, replacing); |
34dc7c2f | 2881 | |
34dc7c2f BB |
2882 | nvlist_free(root); |
2883 | ||
2884 | /* | |
2885 | * If our parent was the replacing vdev, but the replace completed, | |
2886 | * then instead of failing with ENOTSUP we may either succeed, | |
2887 | * fail with ENODEV, or fail with EOVERFLOW. | |
2888 | */ | |
2889 | if (expected_error == ENOTSUP && | |
2890 | (error == 0 || error == ENODEV || error == EOVERFLOW)) | |
2891 | expected_error = error; | |
2892 | ||
2893 | /* | |
2894 | * If someone grew the LUN, the replacement may be too small. | |
2895 | */ | |
b128c09f | 2896 | if (error == EOVERFLOW || error == EBUSY) |
34dc7c2f BB |
2897 | expected_error = error; |
2898 | ||
b128c09f BB |
2899 | /* XXX workaround 6690467 */ |
2900 | if (error != expected_error && expected_error != EBUSY) { | |
2901 | fatal(0, "attach (%s %llu, %s %llu, %d) " | |
2902 | "returned %d, expected %d", | |
2903 | oldpath, (longlong_t)oldsize, newpath, | |
2904 | (longlong_t)newsize, replacing, error, expected_error); | |
34dc7c2f | 2905 | } |
40b84e7a | 2906 | out: |
c242c188 | 2907 | mutex_exit(&ztest_vdev_lock); |
40b84e7a BB |
2908 | |
2909 | umem_free(oldpath, MAXPATHLEN); | |
2910 | umem_free(newpath, MAXPATHLEN); | |
34dc7c2f BB |
2911 | } |
2912 | ||
9babb374 BB |
2913 | /* |
2914 | * Callback function which expands the physical size of the vdev. | |
2915 | */ | |
2916 | vdev_t * | |
2917 | grow_vdev(vdev_t *vd, void *arg) | |
2918 | { | |
1fde1e37 | 2919 | ASSERTV(spa_t *spa = vd->vdev_spa); |
9babb374 BB |
2920 | size_t *newsize = arg; |
2921 | size_t fsize; | |
2922 | int fd; | |
2923 | ||
2924 | ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE); | |
2925 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
2926 | ||
2927 | if ((fd = open(vd->vdev_path, O_RDWR)) == -1) | |
2928 | return (vd); | |
2929 | ||
2930 | fsize = lseek(fd, 0, SEEK_END); | |
0e5b68e0 | 2931 | VERIFY(ftruncate(fd, *newsize) == 0); |
9babb374 | 2932 | |
c242c188 | 2933 | if (ztest_opts.zo_verbose >= 6) { |
9babb374 BB |
2934 | (void) printf("%s grew from %lu to %lu bytes\n", |
2935 | vd->vdev_path, (ulong_t)fsize, (ulong_t)*newsize); | |
2936 | } | |
2937 | (void) close(fd); | |
2938 | return (NULL); | |
2939 | } | |
2940 | ||
2941 | /* | |
2942 | * Callback function which expands a given vdev by calling vdev_online(). | |
2943 | */ | |
2944 | /* ARGSUSED */ | |
2945 | vdev_t * | |
2946 | online_vdev(vdev_t *vd, void *arg) | |
2947 | { | |
2948 | spa_t *spa = vd->vdev_spa; | |
2949 | vdev_t *tvd = vd->vdev_top; | |
9babb374 | 2950 | uint64_t guid = vd->vdev_guid; |
428870ff BB |
2951 | uint64_t generation = spa->spa_config_generation + 1; |
2952 | vdev_state_t newstate = VDEV_STATE_UNKNOWN; | |
2953 | int error; | |
9babb374 BB |
2954 | |
2955 | ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE); | |
2956 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
2957 | ||
2958 | /* Calling vdev_online will initialize the new metaslabs */ | |
2959 | spa_config_exit(spa, SCL_STATE, spa); | |
428870ff | 2960 | error = vdev_online(spa, guid, ZFS_ONLINE_EXPAND, &newstate); |
9babb374 BB |
2961 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); |
2962 | ||
428870ff BB |
2963 | /* |
2964 | * If vdev_online returned an error or the underlying vdev_open | |
2965 | * failed then we abort the expand. The only way to know that | |
2966 | * vdev_open fails is by checking the returned newstate. | |
2967 | */ | |
2968 | if (error || newstate != VDEV_STATE_HEALTHY) { | |
c242c188 | 2969 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
2970 | (void) printf("Unable to expand vdev, state %llu, " |
2971 | "error %d\n", (u_longlong_t)newstate, error); | |
2972 | } | |
2973 | return (vd); | |
2974 | } | |
2975 | ASSERT3U(newstate, ==, VDEV_STATE_HEALTHY); | |
2976 | ||
9babb374 BB |
2977 | /* |
2978 | * Since we dropped the lock we need to ensure that we're | |
2979 | * still talking to the original vdev. It's possible this | |
2980 | * vdev may have been detached/replaced while we were | |
2981 | * trying to online it. | |
2982 | */ | |
428870ff | 2983 | if (generation != spa->spa_config_generation) { |
c242c188 | 2984 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
2985 | (void) printf("vdev configuration has changed, " |
2986 | "guid %llu, state %llu, expected gen %llu, " | |
2987 | "got gen %llu\n", | |
2988 | (u_longlong_t)guid, | |
2989 | (u_longlong_t)tvd->vdev_state, | |
2990 | (u_longlong_t)generation, | |
2991 | (u_longlong_t)spa->spa_config_generation); | |
9babb374 BB |
2992 | } |
2993 | return (vd); | |
2994 | } | |
2995 | return (NULL); | |
2996 | } | |
2997 | ||
2998 | /* | |
2999 | * Traverse the vdev tree calling the supplied function. | |
3000 | * We continue to walk the tree until we either have walked all | |
3001 | * children or we receive a non-NULL return from the callback. | |
3002 | * If a NULL callback is passed, then we just return back the first | |
3003 | * leaf vdev we encounter. | |
3004 | */ | |
3005 | vdev_t * | |
3006 | vdev_walk_tree(vdev_t *vd, vdev_t *(*func)(vdev_t *, void *), void *arg) | |
3007 | { | |
d6320ddb BB |
3008 | uint_t c; |
3009 | ||
9babb374 BB |
3010 | if (vd->vdev_ops->vdev_op_leaf) { |
3011 | if (func == NULL) | |
3012 | return (vd); | |
3013 | else | |
3014 | return (func(vd, arg)); | |
3015 | } | |
3016 | ||
d6320ddb | 3017 | for (c = 0; c < vd->vdev_children; c++) { |
9babb374 BB |
3018 | vdev_t *cvd = vd->vdev_child[c]; |
3019 | if ((cvd = vdev_walk_tree(cvd, func, arg)) != NULL) | |
3020 | return (cvd); | |
3021 | } | |
3022 | return (NULL); | |
3023 | } | |
3024 | ||
34dc7c2f BB |
3025 | /* |
3026 | * Verify that dynamic LUN growth works as expected. | |
3027 | */ | |
428870ff | 3028 | /* ARGSUSED */ |
34dc7c2f | 3029 | void |
428870ff | 3030 | ztest_vdev_LUN_growth(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3031 | { |
c242c188 | 3032 | spa_t *spa = ztest_spa; |
428870ff BB |
3033 | vdev_t *vd, *tvd; |
3034 | metaslab_class_t *mc; | |
3035 | metaslab_group_t *mg; | |
9babb374 | 3036 | size_t psize, newsize; |
428870ff BB |
3037 | uint64_t top; |
3038 | uint64_t old_class_space, new_class_space, old_ms_count, new_ms_count; | |
34dc7c2f | 3039 | |
c242c188 | 3040 | mutex_enter(&ztest_vdev_lock); |
9babb374 BB |
3041 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); |
3042 | ||
428870ff | 3043 | top = ztest_random_vdev_top(spa, B_TRUE); |
9babb374 | 3044 | |
428870ff BB |
3045 | tvd = spa->spa_root_vdev->vdev_child[top]; |
3046 | mg = tvd->vdev_mg; | |
3047 | mc = mg->mg_class; | |
3048 | old_ms_count = tvd->vdev_ms_count; | |
3049 | old_class_space = metaslab_class_get_space(mc); | |
34dc7c2f BB |
3050 | |
3051 | /* | |
9babb374 BB |
3052 | * Determine the size of the first leaf vdev associated with |
3053 | * our top-level device. | |
34dc7c2f | 3054 | */ |
9babb374 BB |
3055 | vd = vdev_walk_tree(tvd, NULL, NULL); |
3056 | ASSERT3P(vd, !=, NULL); | |
3057 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
34dc7c2f | 3058 | |
9babb374 | 3059 | psize = vd->vdev_psize; |
34dc7c2f | 3060 | |
9babb374 | 3061 | /* |
428870ff BB |
3062 | * We only try to expand the vdev if it's healthy, less than 4x its |
3063 | * original size, and it has a valid psize. | |
9babb374 | 3064 | */ |
428870ff | 3065 | if (tvd->vdev_state != VDEV_STATE_HEALTHY || |
c242c188 | 3066 | psize == 0 || psize >= 4 * ztest_opts.zo_vdev_size) { |
9babb374 | 3067 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 3068 | mutex_exit(&ztest_vdev_lock); |
9babb374 BB |
3069 | return; |
3070 | } | |
3071 | ASSERT(psize > 0); | |
3072 | newsize = psize + psize / 8; | |
3073 | ASSERT3U(newsize, >, psize); | |
34dc7c2f | 3074 | |
c242c188 | 3075 | if (ztest_opts.zo_verbose >= 6) { |
428870ff | 3076 | (void) printf("Expanding LUN %s from %lu to %lu\n", |
9babb374 BB |
3077 | vd->vdev_path, (ulong_t)psize, (ulong_t)newsize); |
3078 | } | |
3079 | ||
9babb374 BB |
3080 | /* |
3081 | * Growing the vdev is a two step process: | |
3082 | * 1). expand the physical size (i.e. relabel) | |
3083 | * 2). online the vdev to create the new metaslabs | |
3084 | */ | |
3085 | if (vdev_walk_tree(tvd, grow_vdev, &newsize) != NULL || | |
3086 | vdev_walk_tree(tvd, online_vdev, NULL) != NULL || | |
3087 | tvd->vdev_state != VDEV_STATE_HEALTHY) { | |
c242c188 | 3088 | if (ztest_opts.zo_verbose >= 5) { |
9babb374 | 3089 | (void) printf("Could not expand LUN because " |
428870ff | 3090 | "the vdev configuration changed.\n"); |
34dc7c2f | 3091 | } |
428870ff | 3092 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 3093 | mutex_exit(&ztest_vdev_lock); |
9babb374 | 3094 | return; |
34dc7c2f BB |
3095 | } |
3096 | ||
428870ff | 3097 | spa_config_exit(spa, SCL_STATE, spa); |
9babb374 BB |
3098 | |
3099 | /* | |
3100 | * Expanding the LUN will update the config asynchronously, | |
3101 | * thus we must wait for the async thread to complete any | |
3102 | * pending tasks before proceeding. | |
3103 | */ | |
428870ff BB |
3104 | for (;;) { |
3105 | boolean_t done; | |
3106 | mutex_enter(&spa->spa_async_lock); | |
3107 | done = (spa->spa_async_thread == NULL && !spa->spa_async_tasks); | |
3108 | mutex_exit(&spa->spa_async_lock); | |
3109 | if (done) | |
3110 | break; | |
3111 | txg_wait_synced(spa_get_dsl(spa), 0); | |
3112 | (void) poll(NULL, 0, 100); | |
3113 | } | |
9babb374 BB |
3114 | |
3115 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); | |
428870ff BB |
3116 | |
3117 | tvd = spa->spa_root_vdev->vdev_child[top]; | |
3118 | new_ms_count = tvd->vdev_ms_count; | |
3119 | new_class_space = metaslab_class_get_space(mc); | |
3120 | ||
3121 | if (tvd->vdev_mg != mg || mg->mg_class != mc) { | |
c242c188 | 3122 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
3123 | (void) printf("Could not verify LUN expansion due to " |
3124 | "intervening vdev offline or remove.\n"); | |
3125 | } | |
3126 | spa_config_exit(spa, SCL_STATE, spa); | |
c242c188 | 3127 | mutex_exit(&ztest_vdev_lock); |
428870ff BB |
3128 | return; |
3129 | } | |
3130 | ||
3131 | /* | |
3132 | * Make sure we were able to grow the vdev. | |
3133 | */ | |
3134 | if (new_ms_count <= old_ms_count) | |
3135 | fatal(0, "LUN expansion failed: ms_count %llu <= %llu\n", | |
3136 | old_ms_count, new_ms_count); | |
9babb374 BB |
3137 | |
3138 | /* | |
3139 | * Make sure we were able to grow the pool. | |
3140 | */ | |
428870ff BB |
3141 | if (new_class_space <= old_class_space) |
3142 | fatal(0, "LUN expansion failed: class_space %llu <= %llu\n", | |
3143 | old_class_space, new_class_space); | |
3144 | ||
c242c188 | 3145 | if (ztest_opts.zo_verbose >= 5) { |
9babb374 BB |
3146 | char oldnumbuf[6], newnumbuf[6]; |
3147 | ||
428870ff BB |
3148 | nicenum(old_class_space, oldnumbuf); |
3149 | nicenum(new_class_space, newnumbuf); | |
9babb374 BB |
3150 | (void) printf("%s grew from %s to %s\n", |
3151 | spa->spa_name, oldnumbuf, newnumbuf); | |
3152 | } | |
428870ff | 3153 | |
9babb374 | 3154 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 3155 | mutex_exit(&ztest_vdev_lock); |
34dc7c2f BB |
3156 | } |
3157 | ||
428870ff BB |
3158 | /* |
3159 | * Verify that dmu_objset_{create,destroy,open,close} work as expected. | |
3160 | */ | |
34dc7c2f BB |
3161 | /* ARGSUSED */ |
3162 | static void | |
428870ff | 3163 | ztest_objset_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) |
34dc7c2f BB |
3164 | { |
3165 | /* | |
428870ff | 3166 | * Create the objects common to all ztest datasets. |
34dc7c2f | 3167 | */ |
428870ff | 3168 | VERIFY(zap_create_claim(os, ZTEST_DIROBJ, |
34dc7c2f | 3169 | DMU_OT_ZAP_OTHER, DMU_OT_NONE, 0, tx) == 0); |
428870ff | 3170 | } |
34dc7c2f | 3171 | |
428870ff BB |
3172 | static int |
3173 | ztest_dataset_create(char *dsname) | |
3174 | { | |
3175 | uint64_t zilset = ztest_random(100); | |
3176 | int err = dmu_objset_create(dsname, DMU_OST_OTHER, 0, | |
3177 | ztest_objset_create_cb, NULL); | |
3178 | ||
3179 | if (err || zilset < 80) | |
3180 | return (err); | |
3181 | ||
c242c188 | 3182 | if (ztest_opts.zo_verbose >= 5) |
b815ff9a | 3183 | (void) printf("Setting dataset %s to sync always\n", dsname); |
428870ff BB |
3184 | return (ztest_dsl_prop_set_uint64(dsname, ZFS_PROP_SYNC, |
3185 | ZFS_SYNC_ALWAYS, B_FALSE)); | |
34dc7c2f BB |
3186 | } |
3187 | ||
428870ff | 3188 | /* ARGSUSED */ |
34dc7c2f | 3189 | static int |
428870ff | 3190 | ztest_objset_destroy_cb(const char *name, void *arg) |
34dc7c2f | 3191 | { |
34dc7c2f | 3192 | objset_t *os; |
428870ff | 3193 | dmu_object_info_t doi; |
34dc7c2f BB |
3194 | int error; |
3195 | ||
3196 | /* | |
3197 | * Verify that the dataset contains a directory object. | |
3198 | */ | |
13fe0198 | 3199 | VERIFY0(dmu_objset_own(name, DMU_OST_OTHER, B_TRUE, FTAG, &os)); |
428870ff | 3200 | error = dmu_object_info(os, ZTEST_DIROBJ, &doi); |
34dc7c2f BB |
3201 | if (error != ENOENT) { |
3202 | /* We could have crashed in the middle of destroying it */ | |
c99c9001 | 3203 | ASSERT0(error); |
428870ff BB |
3204 | ASSERT3U(doi.doi_type, ==, DMU_OT_ZAP_OTHER); |
3205 | ASSERT3S(doi.doi_physical_blocks_512, >=, 0); | |
34dc7c2f | 3206 | } |
13fe0198 | 3207 | dmu_objset_disown(os, FTAG); |
34dc7c2f BB |
3208 | |
3209 | /* | |
3210 | * Destroy the dataset. | |
3211 | */ | |
13fe0198 MA |
3212 | if (strchr(name, '@') != NULL) { |
3213 | VERIFY0(dsl_destroy_snapshot(name, B_FALSE)); | |
3214 | } else { | |
3215 | VERIFY0(dsl_destroy_head(name)); | |
3216 | } | |
34dc7c2f BB |
3217 | return (0); |
3218 | } | |
3219 | ||
428870ff BB |
3220 | static boolean_t |
3221 | ztest_snapshot_create(char *osname, uint64_t id) | |
34dc7c2f | 3222 | { |
428870ff BB |
3223 | char snapname[MAXNAMELEN]; |
3224 | int error; | |
3225 | ||
13fe0198 | 3226 | (void) snprintf(snapname, sizeof (snapname), "%llu", (u_longlong_t)id); |
428870ff | 3227 | |
13fe0198 | 3228 | error = dmu_objset_snapshot_one(osname, snapname); |
428870ff BB |
3229 | if (error == ENOSPC) { |
3230 | ztest_record_enospc(FTAG); | |
3231 | return (B_FALSE); | |
3232 | } | |
13fe0198 MA |
3233 | if (error != 0 && error != EEXIST) { |
3234 | fatal(0, "ztest_snapshot_create(%s@%s) = %d", osname, | |
3235 | snapname, error); | |
3236 | } | |
428870ff BB |
3237 | return (B_TRUE); |
3238 | } | |
3239 | ||
3240 | static boolean_t | |
3241 | ztest_snapshot_destroy(char *osname, uint64_t id) | |
3242 | { | |
3243 | char snapname[MAXNAMELEN]; | |
3244 | int error; | |
3245 | ||
3246 | (void) snprintf(snapname, MAXNAMELEN, "%s@%llu", osname, | |
3247 | (u_longlong_t)id); | |
3248 | ||
13fe0198 | 3249 | error = dsl_destroy_snapshot(snapname, B_FALSE); |
428870ff BB |
3250 | if (error != 0 && error != ENOENT) |
3251 | fatal(0, "ztest_snapshot_destroy(%s) = %d", snapname, error); | |
3252 | return (B_TRUE); | |
34dc7c2f BB |
3253 | } |
3254 | ||
428870ff | 3255 | /* ARGSUSED */ |
34dc7c2f | 3256 | void |
428870ff | 3257 | ztest_dmu_objset_create_destroy(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3258 | { |
40b84e7a | 3259 | ztest_ds_t *zdtmp; |
428870ff | 3260 | int iters; |
34dc7c2f | 3261 | int error; |
b128c09f | 3262 | objset_t *os, *os2; |
40b84e7a | 3263 | char *name; |
34dc7c2f | 3264 | zilog_t *zilog; |
d6320ddb | 3265 | int i; |
34dc7c2f | 3266 | |
40b84e7a BB |
3267 | zdtmp = umem_alloc(sizeof (ztest_ds_t), UMEM_NOFAIL); |
3268 | name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3269 | ||
c242c188 | 3270 | (void) rw_enter(&ztest_name_lock, RW_READER); |
34dc7c2f | 3271 | |
428870ff | 3272 | (void) snprintf(name, MAXNAMELEN, "%s/temp_%llu", |
c242c188 | 3273 | ztest_opts.zo_pool, (u_longlong_t)id); |
34dc7c2f BB |
3274 | |
3275 | /* | |
3276 | * If this dataset exists from a previous run, process its replay log | |
13fe0198 | 3277 | * half of the time. If we don't replay it, then dsl_destroy_head() |
428870ff | 3278 | * (invoked from ztest_objset_destroy_cb()) should just throw it away. |
34dc7c2f BB |
3279 | */ |
3280 | if (ztest_random(2) == 0 && | |
428870ff | 3281 | dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, FTAG, &os) == 0) { |
c242c188 | 3282 | ztest_zd_init(zdtmp, NULL, os); |
40b84e7a BB |
3283 | zil_replay(os, zdtmp, ztest_replay_vector); |
3284 | ztest_zd_fini(zdtmp); | |
428870ff | 3285 | dmu_objset_disown(os, FTAG); |
34dc7c2f BB |
3286 | } |
3287 | ||
3288 | /* | |
3289 | * There may be an old instance of the dataset we're about to | |
3290 | * create lying around from a previous run. If so, destroy it | |
3291 | * and all of its snapshots. | |
3292 | */ | |
428870ff | 3293 | (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL, |
34dc7c2f BB |
3294 | DS_FIND_CHILDREN | DS_FIND_SNAPSHOTS); |
3295 | ||
3296 | /* | |
3297 | * Verify that the destroyed dataset is no longer in the namespace. | |
3298 | */ | |
13fe0198 MA |
3299 | VERIFY3U(ENOENT, ==, dmu_objset_own(name, DMU_OST_OTHER, B_TRUE, |
3300 | FTAG, &os)); | |
34dc7c2f BB |
3301 | |
3302 | /* | |
3303 | * Verify that we can create a new dataset. | |
3304 | */ | |
428870ff | 3305 | error = ztest_dataset_create(name); |
34dc7c2f BB |
3306 | if (error) { |
3307 | if (error == ENOSPC) { | |
428870ff | 3308 | ztest_record_enospc(FTAG); |
40b84e7a | 3309 | goto out; |
34dc7c2f BB |
3310 | } |
3311 | fatal(0, "dmu_objset_create(%s) = %d", name, error); | |
3312 | } | |
3313 | ||
13fe0198 | 3314 | VERIFY0(dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, FTAG, &os)); |
428870ff | 3315 | |
c242c188 | 3316 | ztest_zd_init(zdtmp, NULL, os); |
34dc7c2f BB |
3317 | |
3318 | /* | |
3319 | * Open the intent log for it. | |
3320 | */ | |
428870ff | 3321 | zilog = zil_open(os, ztest_get_data); |
34dc7c2f BB |
3322 | |
3323 | /* | |
428870ff BB |
3324 | * Put some objects in there, do a little I/O to them, |
3325 | * and randomly take a couple of snapshots along the way. | |
34dc7c2f | 3326 | */ |
428870ff | 3327 | iters = ztest_random(5); |
d6320ddb | 3328 | for (i = 0; i < iters; i++) { |
40b84e7a | 3329 | ztest_dmu_object_alloc_free(zdtmp, id); |
428870ff BB |
3330 | if (ztest_random(iters) == 0) |
3331 | (void) ztest_snapshot_create(name, i); | |
34dc7c2f BB |
3332 | } |
3333 | ||
3334 | /* | |
3335 | * Verify that we cannot create an existing dataset. | |
3336 | */ | |
428870ff BB |
3337 | VERIFY3U(EEXIST, ==, |
3338 | dmu_objset_create(name, DMU_OST_OTHER, 0, NULL, NULL)); | |
34dc7c2f BB |
3339 | |
3340 | /* | |
428870ff | 3341 | * Verify that we can hold an objset that is also owned. |
b128c09f | 3342 | */ |
428870ff BB |
3343 | VERIFY3U(0, ==, dmu_objset_hold(name, FTAG, &os2)); |
3344 | dmu_objset_rele(os2, FTAG); | |
34dc7c2f | 3345 | |
428870ff BB |
3346 | /* |
3347 | * Verify that we cannot own an objset that is already owned. | |
3348 | */ | |
3349 | VERIFY3U(EBUSY, ==, | |
3350 | dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, FTAG, &os2)); | |
34dc7c2f | 3351 | |
428870ff BB |
3352 | zil_close(zilog); |
3353 | dmu_objset_disown(os, FTAG); | |
40b84e7a BB |
3354 | ztest_zd_fini(zdtmp); |
3355 | out: | |
c242c188 | 3356 | (void) rw_exit(&ztest_name_lock); |
40b84e7a BB |
3357 | |
3358 | umem_free(name, MAXNAMELEN); | |
3359 | umem_free(zdtmp, sizeof (ztest_ds_t)); | |
34dc7c2f BB |
3360 | } |
3361 | ||
3362 | /* | |
3363 | * Verify that dmu_snapshot_{create,destroy,open,close} work as expected. | |
3364 | */ | |
3365 | void | |
428870ff | 3366 | ztest_dmu_snapshot_create_destroy(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3367 | { |
c242c188 | 3368 | (void) rw_enter(&ztest_name_lock, RW_READER); |
428870ff BB |
3369 | (void) ztest_snapshot_destroy(zd->zd_name, id); |
3370 | (void) ztest_snapshot_create(zd->zd_name, id); | |
c242c188 | 3371 | (void) rw_exit(&ztest_name_lock); |
34dc7c2f BB |
3372 | } |
3373 | ||
9babb374 BB |
3374 | /* |
3375 | * Cleanup non-standard snapshots and clones. | |
3376 | */ | |
3377 | void | |
428870ff | 3378 | ztest_dsl_dataset_cleanup(char *osname, uint64_t id) |
9babb374 | 3379 | { |
40b84e7a BB |
3380 | char *snap1name; |
3381 | char *clone1name; | |
3382 | char *snap2name; | |
3383 | char *clone2name; | |
3384 | char *snap3name; | |
9babb374 BB |
3385 | int error; |
3386 | ||
40b84e7a BB |
3387 | snap1name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); |
3388 | clone1name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3389 | snap2name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3390 | clone2name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3391 | snap3name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3392 | ||
b8864a23 BB |
3393 | (void) snprintf(snap1name, MAXNAMELEN, "%s@s1_%llu", |
3394 | osname, (u_longlong_t)id); | |
3395 | (void) snprintf(clone1name, MAXNAMELEN, "%s/c1_%llu", | |
3396 | osname, (u_longlong_t)id); | |
3397 | (void) snprintf(snap2name, MAXNAMELEN, "%s@s2_%llu", | |
3398 | clone1name, (u_longlong_t)id); | |
3399 | (void) snprintf(clone2name, MAXNAMELEN, "%s/c2_%llu", | |
3400 | osname, (u_longlong_t)id); | |
3401 | (void) snprintf(snap3name, MAXNAMELEN, "%s@s3_%llu", | |
3402 | clone1name, (u_longlong_t)id); | |
9babb374 | 3403 | |
13fe0198 | 3404 | error = dsl_destroy_head(clone2name); |
9babb374 | 3405 | if (error && error != ENOENT) |
13fe0198 MA |
3406 | fatal(0, "dsl_destroy_head(%s) = %d", clone2name, error); |
3407 | error = dsl_destroy_snapshot(snap3name, B_FALSE); | |
9babb374 | 3408 | if (error && error != ENOENT) |
13fe0198 MA |
3409 | fatal(0, "dsl_destroy_snapshot(%s) = %d", snap3name, error); |
3410 | error = dsl_destroy_snapshot(snap2name, B_FALSE); | |
9babb374 | 3411 | if (error && error != ENOENT) |
13fe0198 MA |
3412 | fatal(0, "dsl_destroy_snapshot(%s) = %d", snap2name, error); |
3413 | error = dsl_destroy_head(clone1name); | |
9babb374 | 3414 | if (error && error != ENOENT) |
13fe0198 MA |
3415 | fatal(0, "dsl_destroy_head(%s) = %d", clone1name, error); |
3416 | error = dsl_destroy_snapshot(snap1name, B_FALSE); | |
9babb374 | 3417 | if (error && error != ENOENT) |
13fe0198 | 3418 | fatal(0, "dsl_destroy_snapshot(%s) = %d", snap1name, error); |
40b84e7a BB |
3419 | |
3420 | umem_free(snap1name, MAXNAMELEN); | |
3421 | umem_free(clone1name, MAXNAMELEN); | |
3422 | umem_free(snap2name, MAXNAMELEN); | |
3423 | umem_free(clone2name, MAXNAMELEN); | |
3424 | umem_free(snap3name, MAXNAMELEN); | |
9babb374 BB |
3425 | } |
3426 | ||
3427 | /* | |
3428 | * Verify dsl_dataset_promote handles EBUSY | |
3429 | */ | |
3430 | void | |
428870ff | 3431 | ztest_dsl_dataset_promote_busy(ztest_ds_t *zd, uint64_t id) |
9babb374 | 3432 | { |
13fe0198 | 3433 | objset_t *os; |
40b84e7a BB |
3434 | char *snap1name; |
3435 | char *clone1name; | |
3436 | char *snap2name; | |
3437 | char *clone2name; | |
3438 | char *snap3name; | |
428870ff BB |
3439 | char *osname = zd->zd_name; |
3440 | int error; | |
9babb374 | 3441 | |
40b84e7a BB |
3442 | snap1name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); |
3443 | clone1name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3444 | snap2name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3445 | clone2name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3446 | snap3name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3447 | ||
c242c188 | 3448 | (void) rw_enter(&ztest_name_lock, RW_READER); |
9babb374 | 3449 | |
428870ff | 3450 | ztest_dsl_dataset_cleanup(osname, id); |
9babb374 | 3451 | |
b8864a23 BB |
3452 | (void) snprintf(snap1name, MAXNAMELEN, "%s@s1_%llu", |
3453 | osname, (u_longlong_t)id); | |
3454 | (void) snprintf(clone1name, MAXNAMELEN, "%s/c1_%llu", | |
3455 | osname, (u_longlong_t)id); | |
3456 | (void) snprintf(snap2name, MAXNAMELEN, "%s@s2_%llu", | |
3457 | clone1name, (u_longlong_t)id); | |
3458 | (void) snprintf(clone2name, MAXNAMELEN, "%s/c2_%llu", | |
3459 | osname, (u_longlong_t)id); | |
3460 | (void) snprintf(snap3name, MAXNAMELEN, "%s@s3_%llu", | |
3461 | clone1name, (u_longlong_t)id); | |
9babb374 | 3462 | |
6f1ffb06 | 3463 | error = dmu_objset_snapshot_one(osname, strchr(snap1name, '@') + 1); |
9babb374 BB |
3464 | if (error && error != EEXIST) { |
3465 | if (error == ENOSPC) { | |
428870ff | 3466 | ztest_record_enospc(FTAG); |
9babb374 BB |
3467 | goto out; |
3468 | } | |
3469 | fatal(0, "dmu_take_snapshot(%s) = %d", snap1name, error); | |
3470 | } | |
3471 | ||
13fe0198 | 3472 | error = dmu_objset_clone(clone1name, snap1name); |
9babb374 BB |
3473 | if (error) { |
3474 | if (error == ENOSPC) { | |
428870ff | 3475 | ztest_record_enospc(FTAG); |
9babb374 BB |
3476 | goto out; |
3477 | } | |
3478 | fatal(0, "dmu_objset_create(%s) = %d", clone1name, error); | |
3479 | } | |
3480 | ||
6f1ffb06 | 3481 | error = dmu_objset_snapshot_one(clone1name, strchr(snap2name, '@') + 1); |
428870ff BB |
3482 | if (error && error != EEXIST) { |
3483 | if (error == ENOSPC) { | |
3484 | ztest_record_enospc(FTAG); | |
3485 | goto out; | |
34dc7c2f | 3486 | } |
428870ff BB |
3487 | fatal(0, "dmu_open_snapshot(%s) = %d", snap2name, error); |
3488 | } | |
34dc7c2f | 3489 | |
6f1ffb06 | 3490 | error = dmu_objset_snapshot_one(clone1name, strchr(snap3name, '@') + 1); |
428870ff BB |
3491 | if (error && error != EEXIST) { |
3492 | if (error == ENOSPC) { | |
3493 | ztest_record_enospc(FTAG); | |
3494 | goto out; | |
3495 | } | |
3496 | fatal(0, "dmu_open_snapshot(%s) = %d", snap3name, error); | |
3497 | } | |
34dc7c2f | 3498 | |
13fe0198 | 3499 | error = dmu_objset_clone(clone2name, snap3name); |
428870ff BB |
3500 | if (error) { |
3501 | if (error == ENOSPC) { | |
3502 | ztest_record_enospc(FTAG); | |
3503 | goto out; | |
3504 | } | |
3505 | fatal(0, "dmu_objset_create(%s) = %d", clone2name, error); | |
3506 | } | |
34dc7c2f | 3507 | |
13fe0198 | 3508 | error = dmu_objset_own(snap2name, DMU_OST_ANY, B_TRUE, FTAG, &os); |
428870ff | 3509 | if (error) |
13fe0198 | 3510 | fatal(0, "dmu_objset_own(%s) = %d", snap2name, error); |
428870ff BB |
3511 | error = dsl_dataset_promote(clone2name, NULL); |
3512 | if (error != EBUSY) | |
3513 | fatal(0, "dsl_dataset_promote(%s), %d, not EBUSY", clone2name, | |
3514 | error); | |
13fe0198 | 3515 | dmu_objset_disown(os, FTAG); |
34dc7c2f | 3516 | |
428870ff BB |
3517 | out: |
3518 | ztest_dsl_dataset_cleanup(osname, id); | |
34dc7c2f | 3519 | |
c242c188 | 3520 | (void) rw_exit(&ztest_name_lock); |
40b84e7a BB |
3521 | |
3522 | umem_free(snap1name, MAXNAMELEN); | |
3523 | umem_free(clone1name, MAXNAMELEN); | |
3524 | umem_free(snap2name, MAXNAMELEN); | |
3525 | umem_free(clone2name, MAXNAMELEN); | |
3526 | umem_free(snap3name, MAXNAMELEN); | |
428870ff | 3527 | } |
34dc7c2f | 3528 | |
40b84e7a BB |
3529 | #undef OD_ARRAY_SIZE |
3530 | #define OD_ARRAY_SIZE 4 | |
3531 | ||
428870ff BB |
3532 | /* |
3533 | * Verify that dmu_object_{alloc,free} work as expected. | |
3534 | */ | |
3535 | void | |
3536 | ztest_dmu_object_alloc_free(ztest_ds_t *zd, uint64_t id) | |
3537 | { | |
40b84e7a BB |
3538 | ztest_od_t *od; |
3539 | int batchsize; | |
3540 | int size; | |
d6320ddb | 3541 | int b; |
34dc7c2f | 3542 | |
40b84e7a BB |
3543 | size = sizeof(ztest_od_t) * OD_ARRAY_SIZE; |
3544 | od = umem_alloc(size, UMEM_NOFAIL); | |
3545 | batchsize = OD_ARRAY_SIZE; | |
3546 | ||
d6320ddb | 3547 | for (b = 0; b < batchsize; b++) |
40b84e7a | 3548 | ztest_od_init(od + b, id, FTAG, b, DMU_OT_UINT64_OTHER, 0, 0); |
34dc7c2f | 3549 | |
428870ff BB |
3550 | /* |
3551 | * Destroy the previous batch of objects, create a new batch, | |
3552 | * and do some I/O on the new objects. | |
3553 | */ | |
40b84e7a | 3554 | if (ztest_object_init(zd, od, size, B_TRUE) != 0) |
428870ff | 3555 | return; |
34dc7c2f | 3556 | |
428870ff BB |
3557 | while (ztest_random(4 * batchsize) != 0) |
3558 | ztest_io(zd, od[ztest_random(batchsize)].od_object, | |
3559 | ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
40b84e7a BB |
3560 | |
3561 | umem_free(od, size); | |
34dc7c2f BB |
3562 | } |
3563 | ||
40b84e7a BB |
3564 | #undef OD_ARRAY_SIZE |
3565 | #define OD_ARRAY_SIZE 2 | |
3566 | ||
34dc7c2f BB |
3567 | /* |
3568 | * Verify that dmu_{read,write} work as expected. | |
3569 | */ | |
34dc7c2f | 3570 | void |
428870ff | 3571 | ztest_dmu_read_write(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3572 | { |
40b84e7a BB |
3573 | int size; |
3574 | ztest_od_t *od; | |
3575 | ||
428870ff | 3576 | objset_t *os = zd->zd_os; |
40b84e7a BB |
3577 | size = sizeof(ztest_od_t) * OD_ARRAY_SIZE; |
3578 | od = umem_alloc(size, UMEM_NOFAIL); | |
34dc7c2f BB |
3579 | dmu_tx_t *tx; |
3580 | int i, freeit, error; | |
3581 | uint64_t n, s, txg; | |
3582 | bufwad_t *packbuf, *bigbuf, *pack, *bigH, *bigT; | |
428870ff BB |
3583 | uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize; |
3584 | uint64_t chunksize = (1000 + ztest_random(1000)) * sizeof (uint64_t); | |
34dc7c2f BB |
3585 | uint64_t regions = 997; |
3586 | uint64_t stride = 123456789ULL; | |
3587 | uint64_t width = 40; | |
3588 | int free_percent = 5; | |
3589 | ||
3590 | /* | |
3591 | * This test uses two objects, packobj and bigobj, that are always | |
3592 | * updated together (i.e. in the same tx) so that their contents are | |
3593 | * in sync and can be compared. Their contents relate to each other | |
3594 | * in a simple way: packobj is a dense array of 'bufwad' structures, | |
3595 | * while bigobj is a sparse array of the same bufwads. Specifically, | |
3596 | * for any index n, there are three bufwads that should be identical: | |
3597 | * | |
3598 | * packobj, at offset n * sizeof (bufwad_t) | |
3599 | * bigobj, at the head of the nth chunk | |
3600 | * bigobj, at the tail of the nth chunk | |
3601 | * | |
3602 | * The chunk size is arbitrary. It doesn't have to be a power of two, | |
3603 | * and it doesn't have any relation to the object blocksize. | |
3604 | * The only requirement is that it can hold at least two bufwads. | |
3605 | * | |
3606 | * Normally, we write the bufwad to each of these locations. | |
3607 | * However, free_percent of the time we instead write zeroes to | |
3608 | * packobj and perform a dmu_free_range() on bigobj. By comparing | |
3609 | * bigobj to packobj, we can verify that the DMU is correctly | |
3610 | * tracking which parts of an object are allocated and free, | |
3611 | * and that the contents of the allocated blocks are correct. | |
3612 | */ | |
3613 | ||
3614 | /* | |
3615 | * Read the directory info. If it's the first time, set things up. | |
3616 | */ | |
40b84e7a BB |
3617 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, chunksize); |
3618 | ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, chunksize); | |
34dc7c2f | 3619 | |
40b84e7a BB |
3620 | if (ztest_object_init(zd, od, size, B_FALSE) != 0) { |
3621 | umem_free(od, size); | |
428870ff | 3622 | return; |
40b84e7a | 3623 | } |
34dc7c2f | 3624 | |
428870ff BB |
3625 | bigobj = od[0].od_object; |
3626 | packobj = od[1].od_object; | |
3627 | chunksize = od[0].od_gen; | |
3628 | ASSERT(chunksize == od[1].od_gen); | |
34dc7c2f BB |
3629 | |
3630 | /* | |
3631 | * Prefetch a random chunk of the big object. | |
3632 | * Our aim here is to get some async reads in flight | |
3633 | * for blocks that we may free below; the DMU should | |
3634 | * handle this race correctly. | |
3635 | */ | |
3636 | n = ztest_random(regions) * stride + ztest_random(width); | |
3637 | s = 1 + ztest_random(2 * width - 1); | |
428870ff | 3638 | dmu_prefetch(os, bigobj, n * chunksize, s * chunksize); |
34dc7c2f BB |
3639 | |
3640 | /* | |
3641 | * Pick a random index and compute the offsets into packobj and bigobj. | |
3642 | */ | |
3643 | n = ztest_random(regions) * stride + ztest_random(width); | |
3644 | s = 1 + ztest_random(width - 1); | |
3645 | ||
3646 | packoff = n * sizeof (bufwad_t); | |
3647 | packsize = s * sizeof (bufwad_t); | |
3648 | ||
428870ff BB |
3649 | bigoff = n * chunksize; |
3650 | bigsize = s * chunksize; | |
34dc7c2f BB |
3651 | |
3652 | packbuf = umem_alloc(packsize, UMEM_NOFAIL); | |
3653 | bigbuf = umem_alloc(bigsize, UMEM_NOFAIL); | |
3654 | ||
3655 | /* | |
3656 | * free_percent of the time, free a range of bigobj rather than | |
3657 | * overwriting it. | |
3658 | */ | |
3659 | freeit = (ztest_random(100) < free_percent); | |
3660 | ||
3661 | /* | |
3662 | * Read the current contents of our objects. | |
3663 | */ | |
428870ff | 3664 | error = dmu_read(os, packobj, packoff, packsize, packbuf, |
9babb374 | 3665 | DMU_READ_PREFETCH); |
c99c9001 | 3666 | ASSERT0(error); |
428870ff | 3667 | error = dmu_read(os, bigobj, bigoff, bigsize, bigbuf, |
9babb374 | 3668 | DMU_READ_PREFETCH); |
c99c9001 | 3669 | ASSERT0(error); |
34dc7c2f BB |
3670 | |
3671 | /* | |
3672 | * Get a tx for the mods to both packobj and bigobj. | |
3673 | */ | |
3674 | tx = dmu_tx_create(os); | |
3675 | ||
428870ff | 3676 | dmu_tx_hold_write(tx, packobj, packoff, packsize); |
34dc7c2f BB |
3677 | |
3678 | if (freeit) | |
428870ff | 3679 | dmu_tx_hold_free(tx, bigobj, bigoff, bigsize); |
34dc7c2f | 3680 | else |
428870ff | 3681 | dmu_tx_hold_write(tx, bigobj, bigoff, bigsize); |
34dc7c2f | 3682 | |
428870ff BB |
3683 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); |
3684 | if (txg == 0) { | |
34dc7c2f BB |
3685 | umem_free(packbuf, packsize); |
3686 | umem_free(bigbuf, bigsize); | |
40b84e7a | 3687 | umem_free(od, size); |
34dc7c2f BB |
3688 | return; |
3689 | } | |
3690 | ||
428870ff BB |
3691 | dmu_object_set_checksum(os, bigobj, |
3692 | (enum zio_checksum)ztest_random_dsl_prop(ZFS_PROP_CHECKSUM), tx); | |
3693 | ||
3694 | dmu_object_set_compress(os, bigobj, | |
3695 | (enum zio_compress)ztest_random_dsl_prop(ZFS_PROP_COMPRESSION), tx); | |
34dc7c2f BB |
3696 | |
3697 | /* | |
3698 | * For each index from n to n + s, verify that the existing bufwad | |
3699 | * in packobj matches the bufwads at the head and tail of the | |
3700 | * corresponding chunk in bigobj. Then update all three bufwads | |
3701 | * with the new values we want to write out. | |
3702 | */ | |
3703 | for (i = 0; i < s; i++) { | |
3704 | /* LINTED */ | |
3705 | pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t)); | |
3706 | /* LINTED */ | |
428870ff | 3707 | bigH = (bufwad_t *)((char *)bigbuf + i * chunksize); |
34dc7c2f | 3708 | /* LINTED */ |
428870ff | 3709 | bigT = (bufwad_t *)((char *)bigH + chunksize) - 1; |
34dc7c2f BB |
3710 | |
3711 | ASSERT((uintptr_t)bigH - (uintptr_t)bigbuf < bigsize); | |
3712 | ASSERT((uintptr_t)bigT - (uintptr_t)bigbuf < bigsize); | |
3713 | ||
3714 | if (pack->bw_txg > txg) | |
3715 | fatal(0, "future leak: got %llx, open txg is %llx", | |
3716 | pack->bw_txg, txg); | |
3717 | ||
3718 | if (pack->bw_data != 0 && pack->bw_index != n + i) | |
3719 | fatal(0, "wrong index: got %llx, wanted %llx+%llx", | |
3720 | pack->bw_index, n, i); | |
3721 | ||
3722 | if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0) | |
3723 | fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH); | |
3724 | ||
3725 | if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0) | |
3726 | fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT); | |
3727 | ||
3728 | if (freeit) { | |
3729 | bzero(pack, sizeof (bufwad_t)); | |
3730 | } else { | |
3731 | pack->bw_index = n + i; | |
3732 | pack->bw_txg = txg; | |
3733 | pack->bw_data = 1 + ztest_random(-2ULL); | |
3734 | } | |
3735 | *bigH = *pack; | |
3736 | *bigT = *pack; | |
3737 | } | |
3738 | ||
3739 | /* | |
3740 | * We've verified all the old bufwads, and made new ones. | |
3741 | * Now write them out. | |
3742 | */ | |
428870ff | 3743 | dmu_write(os, packobj, packoff, packsize, packbuf, tx); |
34dc7c2f BB |
3744 | |
3745 | if (freeit) { | |
c242c188 | 3746 | if (ztest_opts.zo_verbose >= 7) { |
34dc7c2f BB |
3747 | (void) printf("freeing offset %llx size %llx" |
3748 | " txg %llx\n", | |
3749 | (u_longlong_t)bigoff, | |
3750 | (u_longlong_t)bigsize, | |
3751 | (u_longlong_t)txg); | |
3752 | } | |
428870ff | 3753 | VERIFY(0 == dmu_free_range(os, bigobj, bigoff, bigsize, tx)); |
34dc7c2f | 3754 | } else { |
c242c188 | 3755 | if (ztest_opts.zo_verbose >= 7) { |
34dc7c2f BB |
3756 | (void) printf("writing offset %llx size %llx" |
3757 | " txg %llx\n", | |
3758 | (u_longlong_t)bigoff, | |
3759 | (u_longlong_t)bigsize, | |
3760 | (u_longlong_t)txg); | |
3761 | } | |
428870ff | 3762 | dmu_write(os, bigobj, bigoff, bigsize, bigbuf, tx); |
34dc7c2f BB |
3763 | } |
3764 | ||
3765 | dmu_tx_commit(tx); | |
3766 | ||
3767 | /* | |
3768 | * Sanity check the stuff we just wrote. | |
3769 | */ | |
3770 | { | |
3771 | void *packcheck = umem_alloc(packsize, UMEM_NOFAIL); | |
3772 | void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL); | |
3773 | ||
428870ff | 3774 | VERIFY(0 == dmu_read(os, packobj, packoff, |
9babb374 | 3775 | packsize, packcheck, DMU_READ_PREFETCH)); |
428870ff | 3776 | VERIFY(0 == dmu_read(os, bigobj, bigoff, |
9babb374 | 3777 | bigsize, bigcheck, DMU_READ_PREFETCH)); |
34dc7c2f BB |
3778 | |
3779 | ASSERT(bcmp(packbuf, packcheck, packsize) == 0); | |
3780 | ASSERT(bcmp(bigbuf, bigcheck, bigsize) == 0); | |
3781 | ||
3782 | umem_free(packcheck, packsize); | |
3783 | umem_free(bigcheck, bigsize); | |
3784 | } | |
3785 | ||
3786 | umem_free(packbuf, packsize); | |
3787 | umem_free(bigbuf, bigsize); | |
40b84e7a | 3788 | umem_free(od, size); |
34dc7c2f BB |
3789 | } |
3790 | ||
9babb374 BB |
3791 | void |
3792 | compare_and_update_pbbufs(uint64_t s, bufwad_t *packbuf, bufwad_t *bigbuf, | |
428870ff | 3793 | uint64_t bigsize, uint64_t n, uint64_t chunksize, uint64_t txg) |
9babb374 BB |
3794 | { |
3795 | uint64_t i; | |
3796 | bufwad_t *pack; | |
3797 | bufwad_t *bigH; | |
3798 | bufwad_t *bigT; | |
3799 | ||
3800 | /* | |
3801 | * For each index from n to n + s, verify that the existing bufwad | |
3802 | * in packobj matches the bufwads at the head and tail of the | |
3803 | * corresponding chunk in bigobj. Then update all three bufwads | |
3804 | * with the new values we want to write out. | |
3805 | */ | |
3806 | for (i = 0; i < s; i++) { | |
3807 | /* LINTED */ | |
3808 | pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t)); | |
3809 | /* LINTED */ | |
428870ff | 3810 | bigH = (bufwad_t *)((char *)bigbuf + i * chunksize); |
9babb374 | 3811 | /* LINTED */ |
428870ff | 3812 | bigT = (bufwad_t *)((char *)bigH + chunksize) - 1; |
9babb374 BB |
3813 | |
3814 | ASSERT((uintptr_t)bigH - (uintptr_t)bigbuf < bigsize); | |
3815 | ASSERT((uintptr_t)bigT - (uintptr_t)bigbuf < bigsize); | |
3816 | ||
3817 | if (pack->bw_txg > txg) | |
3818 | fatal(0, "future leak: got %llx, open txg is %llx", | |
3819 | pack->bw_txg, txg); | |
3820 | ||
3821 | if (pack->bw_data != 0 && pack->bw_index != n + i) | |
3822 | fatal(0, "wrong index: got %llx, wanted %llx+%llx", | |
3823 | pack->bw_index, n, i); | |
3824 | ||
3825 | if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0) | |
3826 | fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH); | |
3827 | ||
3828 | if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0) | |
3829 | fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT); | |
3830 | ||
3831 | pack->bw_index = n + i; | |
3832 | pack->bw_txg = txg; | |
3833 | pack->bw_data = 1 + ztest_random(-2ULL); | |
3834 | ||
3835 | *bigH = *pack; | |
3836 | *bigT = *pack; | |
3837 | } | |
3838 | } | |
3839 | ||
40b84e7a BB |
3840 | #undef OD_ARRAY_SIZE |
3841 | #define OD_ARRAY_SIZE 2 | |
3842 | ||
9babb374 | 3843 | void |
428870ff | 3844 | ztest_dmu_read_write_zcopy(ztest_ds_t *zd, uint64_t id) |
9babb374 | 3845 | { |
428870ff | 3846 | objset_t *os = zd->zd_os; |
40b84e7a | 3847 | ztest_od_t *od; |
9babb374 BB |
3848 | dmu_tx_t *tx; |
3849 | uint64_t i; | |
3850 | int error; | |
40b84e7a | 3851 | int size; |
9babb374 BB |
3852 | uint64_t n, s, txg; |
3853 | bufwad_t *packbuf, *bigbuf; | |
428870ff BB |
3854 | uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize; |
3855 | uint64_t blocksize = ztest_random_blocksize(); | |
3856 | uint64_t chunksize = blocksize; | |
9babb374 BB |
3857 | uint64_t regions = 997; |
3858 | uint64_t stride = 123456789ULL; | |
3859 | uint64_t width = 9; | |
3860 | dmu_buf_t *bonus_db; | |
3861 | arc_buf_t **bigbuf_arcbufs; | |
428870ff | 3862 | dmu_object_info_t doi; |
9babb374 | 3863 | |
40b84e7a BB |
3864 | size = sizeof(ztest_od_t) * OD_ARRAY_SIZE; |
3865 | od = umem_alloc(size, UMEM_NOFAIL); | |
3866 | ||
9babb374 BB |
3867 | /* |
3868 | * This test uses two objects, packobj and bigobj, that are always | |
3869 | * updated together (i.e. in the same tx) so that their contents are | |
3870 | * in sync and can be compared. Their contents relate to each other | |
3871 | * in a simple way: packobj is a dense array of 'bufwad' structures, | |
3872 | * while bigobj is a sparse array of the same bufwads. Specifically, | |
3873 | * for any index n, there are three bufwads that should be identical: | |
3874 | * | |
3875 | * packobj, at offset n * sizeof (bufwad_t) | |
3876 | * bigobj, at the head of the nth chunk | |
3877 | * bigobj, at the tail of the nth chunk | |
3878 | * | |
3879 | * The chunk size is set equal to bigobj block size so that | |
3880 | * dmu_assign_arcbuf() can be tested for object updates. | |
3881 | */ | |
3882 | ||
3883 | /* | |
3884 | * Read the directory info. If it's the first time, set things up. | |
3885 | */ | |
40b84e7a BB |
3886 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0); |
3887 | ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, chunksize); | |
3888 | ||
9babb374 | 3889 | |
40b84e7a BB |
3890 | if (ztest_object_init(zd, od, size, B_FALSE) != 0) { |
3891 | umem_free(od, size); | |
428870ff | 3892 | return; |
40b84e7a | 3893 | } |
9babb374 | 3894 | |
428870ff BB |
3895 | bigobj = od[0].od_object; |
3896 | packobj = od[1].od_object; | |
3897 | blocksize = od[0].od_blocksize; | |
3898 | chunksize = blocksize; | |
3899 | ASSERT(chunksize == od[1].od_gen); | |
9babb374 | 3900 | |
428870ff BB |
3901 | VERIFY(dmu_object_info(os, bigobj, &doi) == 0); |
3902 | VERIFY(ISP2(doi.doi_data_block_size)); | |
3903 | VERIFY(chunksize == doi.doi_data_block_size); | |
3904 | VERIFY(chunksize >= 2 * sizeof (bufwad_t)); | |
9babb374 BB |
3905 | |
3906 | /* | |
3907 | * Pick a random index and compute the offsets into packobj and bigobj. | |
3908 | */ | |
3909 | n = ztest_random(regions) * stride + ztest_random(width); | |
3910 | s = 1 + ztest_random(width - 1); | |
3911 | ||
3912 | packoff = n * sizeof (bufwad_t); | |
3913 | packsize = s * sizeof (bufwad_t); | |
3914 | ||
428870ff BB |
3915 | bigoff = n * chunksize; |
3916 | bigsize = s * chunksize; | |
9babb374 BB |
3917 | |
3918 | packbuf = umem_zalloc(packsize, UMEM_NOFAIL); | |
3919 | bigbuf = umem_zalloc(bigsize, UMEM_NOFAIL); | |
3920 | ||
428870ff | 3921 | VERIFY3U(0, ==, dmu_bonus_hold(os, bigobj, FTAG, &bonus_db)); |
9babb374 BB |
3922 | |
3923 | bigbuf_arcbufs = umem_zalloc(2 * s * sizeof (arc_buf_t *), UMEM_NOFAIL); | |
3924 | ||
3925 | /* | |
3926 | * Iteration 0 test zcopy for DB_UNCACHED dbufs. | |
3927 | * Iteration 1 test zcopy to already referenced dbufs. | |
3928 | * Iteration 2 test zcopy to dirty dbuf in the same txg. | |
3929 | * Iteration 3 test zcopy to dbuf dirty in previous txg. | |
3930 | * Iteration 4 test zcopy when dbuf is no longer dirty. | |
3931 | * Iteration 5 test zcopy when it can't be done. | |
3932 | * Iteration 6 one more zcopy write. | |
3933 | */ | |
3934 | for (i = 0; i < 7; i++) { | |
3935 | uint64_t j; | |
3936 | uint64_t off; | |
3937 | ||
3938 | /* | |
3939 | * In iteration 5 (i == 5) use arcbufs | |
3940 | * that don't match bigobj blksz to test | |
3941 | * dmu_assign_arcbuf() when it can't directly | |
3942 | * assign an arcbuf to a dbuf. | |
3943 | */ | |
3944 | for (j = 0; j < s; j++) { | |
3945 | if (i != 5) { | |
3946 | bigbuf_arcbufs[j] = | |
428870ff | 3947 | dmu_request_arcbuf(bonus_db, chunksize); |
9babb374 BB |
3948 | } else { |
3949 | bigbuf_arcbufs[2 * j] = | |
428870ff | 3950 | dmu_request_arcbuf(bonus_db, chunksize / 2); |
9babb374 | 3951 | bigbuf_arcbufs[2 * j + 1] = |
428870ff | 3952 | dmu_request_arcbuf(bonus_db, chunksize / 2); |
9babb374 BB |
3953 | } |
3954 | } | |
3955 | ||
3956 | /* | |
3957 | * Get a tx for the mods to both packobj and bigobj. | |
3958 | */ | |
3959 | tx = dmu_tx_create(os); | |
3960 | ||
428870ff BB |
3961 | dmu_tx_hold_write(tx, packobj, packoff, packsize); |
3962 | dmu_tx_hold_write(tx, bigobj, bigoff, bigsize); | |
9babb374 | 3963 | |
428870ff BB |
3964 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); |
3965 | if (txg == 0) { | |
9babb374 BB |
3966 | umem_free(packbuf, packsize); |
3967 | umem_free(bigbuf, bigsize); | |
3968 | for (j = 0; j < s; j++) { | |
3969 | if (i != 5) { | |
3970 | dmu_return_arcbuf(bigbuf_arcbufs[j]); | |
3971 | } else { | |
3972 | dmu_return_arcbuf( | |
3973 | bigbuf_arcbufs[2 * j]); | |
3974 | dmu_return_arcbuf( | |
3975 | bigbuf_arcbufs[2 * j + 1]); | |
3976 | } | |
3977 | } | |
3978 | umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *)); | |
40b84e7a | 3979 | umem_free(od, size); |
9babb374 BB |
3980 | dmu_buf_rele(bonus_db, FTAG); |
3981 | return; | |
3982 | } | |
3983 | ||
9babb374 BB |
3984 | /* |
3985 | * 50% of the time don't read objects in the 1st iteration to | |
3986 | * test dmu_assign_arcbuf() for the case when there're no | |
3987 | * existing dbufs for the specified offsets. | |
3988 | */ | |
3989 | if (i != 0 || ztest_random(2) != 0) { | |
428870ff | 3990 | error = dmu_read(os, packobj, packoff, |
9babb374 | 3991 | packsize, packbuf, DMU_READ_PREFETCH); |
c99c9001 | 3992 | ASSERT0(error); |
428870ff | 3993 | error = dmu_read(os, bigobj, bigoff, bigsize, |
9babb374 | 3994 | bigbuf, DMU_READ_PREFETCH); |
c99c9001 | 3995 | ASSERT0(error); |
9babb374 BB |
3996 | } |
3997 | compare_and_update_pbbufs(s, packbuf, bigbuf, bigsize, | |
428870ff | 3998 | n, chunksize, txg); |
9babb374 BB |
3999 | |
4000 | /* | |
4001 | * We've verified all the old bufwads, and made new ones. | |
4002 | * Now write them out. | |
4003 | */ | |
428870ff | 4004 | dmu_write(os, packobj, packoff, packsize, packbuf, tx); |
c242c188 | 4005 | if (ztest_opts.zo_verbose >= 7) { |
9babb374 BB |
4006 | (void) printf("writing offset %llx size %llx" |
4007 | " txg %llx\n", | |
4008 | (u_longlong_t)bigoff, | |
4009 | (u_longlong_t)bigsize, | |
4010 | (u_longlong_t)txg); | |
4011 | } | |
428870ff | 4012 | for (off = bigoff, j = 0; j < s; j++, off += chunksize) { |
9babb374 BB |
4013 | dmu_buf_t *dbt; |
4014 | if (i != 5) { | |
4015 | bcopy((caddr_t)bigbuf + (off - bigoff), | |
428870ff | 4016 | bigbuf_arcbufs[j]->b_data, chunksize); |
9babb374 BB |
4017 | } else { |
4018 | bcopy((caddr_t)bigbuf + (off - bigoff), | |
4019 | bigbuf_arcbufs[2 * j]->b_data, | |
428870ff | 4020 | chunksize / 2); |
9babb374 | 4021 | bcopy((caddr_t)bigbuf + (off - bigoff) + |
428870ff | 4022 | chunksize / 2, |
9babb374 | 4023 | bigbuf_arcbufs[2 * j + 1]->b_data, |
428870ff | 4024 | chunksize / 2); |
9babb374 BB |
4025 | } |
4026 | ||
4027 | if (i == 1) { | |
428870ff BB |
4028 | VERIFY(dmu_buf_hold(os, bigobj, off, |
4029 | FTAG, &dbt, DMU_READ_NO_PREFETCH) == 0); | |
9babb374 BB |
4030 | } |
4031 | if (i != 5) { | |
4032 | dmu_assign_arcbuf(bonus_db, off, | |
4033 | bigbuf_arcbufs[j], tx); | |
4034 | } else { | |
4035 | dmu_assign_arcbuf(bonus_db, off, | |
4036 | bigbuf_arcbufs[2 * j], tx); | |
4037 | dmu_assign_arcbuf(bonus_db, | |
428870ff | 4038 | off + chunksize / 2, |
9babb374 BB |
4039 | bigbuf_arcbufs[2 * j + 1], tx); |
4040 | } | |
4041 | if (i == 1) { | |
4042 | dmu_buf_rele(dbt, FTAG); | |
4043 | } | |
4044 | } | |
4045 | dmu_tx_commit(tx); | |
4046 | ||
4047 | /* | |
4048 | * Sanity check the stuff we just wrote. | |
4049 | */ | |
4050 | { | |
4051 | void *packcheck = umem_alloc(packsize, UMEM_NOFAIL); | |
4052 | void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL); | |
4053 | ||
428870ff | 4054 | VERIFY(0 == dmu_read(os, packobj, packoff, |
9babb374 | 4055 | packsize, packcheck, DMU_READ_PREFETCH)); |
428870ff | 4056 | VERIFY(0 == dmu_read(os, bigobj, bigoff, |
9babb374 BB |
4057 | bigsize, bigcheck, DMU_READ_PREFETCH)); |
4058 | ||
4059 | ASSERT(bcmp(packbuf, packcheck, packsize) == 0); | |
4060 | ASSERT(bcmp(bigbuf, bigcheck, bigsize) == 0); | |
4061 | ||
4062 | umem_free(packcheck, packsize); | |
4063 | umem_free(bigcheck, bigsize); | |
4064 | } | |
4065 | if (i == 2) { | |
4066 | txg_wait_open(dmu_objset_pool(os), 0); | |
4067 | } else if (i == 3) { | |
4068 | txg_wait_synced(dmu_objset_pool(os), 0); | |
4069 | } | |
4070 | } | |
4071 | ||
4072 | dmu_buf_rele(bonus_db, FTAG); | |
4073 | umem_free(packbuf, packsize); | |
4074 | umem_free(bigbuf, bigsize); | |
4075 | umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *)); | |
40b84e7a | 4076 | umem_free(od, size); |
9babb374 BB |
4077 | } |
4078 | ||
428870ff | 4079 | /* ARGSUSED */ |
34dc7c2f | 4080 | void |
428870ff | 4081 | ztest_dmu_write_parallel(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4082 | { |
40b84e7a BB |
4083 | ztest_od_t *od; |
4084 | ||
4085 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); | |
428870ff BB |
4086 | uint64_t offset = (1ULL << (ztest_random(20) + 43)) + |
4087 | (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
34dc7c2f BB |
4088 | |
4089 | /* | |
428870ff BB |
4090 | * Have multiple threads write to large offsets in an object |
4091 | * to verify that parallel writes to an object -- even to the | |
4092 | * same blocks within the object -- doesn't cause any trouble. | |
34dc7c2f | 4093 | */ |
40b84e7a | 4094 | ztest_od_init(od, ID_PARALLEL, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0); |
9babb374 | 4095 | |
40b84e7a | 4096 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) |
34dc7c2f | 4097 | return; |
34dc7c2f | 4098 | |
428870ff | 4099 | while (ztest_random(10) != 0) |
40b84e7a BB |
4100 | ztest_io(zd, od->od_object, offset); |
4101 | ||
4102 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff | 4103 | } |
34dc7c2f | 4104 | |
428870ff BB |
4105 | void |
4106 | ztest_dmu_prealloc(ztest_ds_t *zd, uint64_t id) | |
4107 | { | |
40b84e7a | 4108 | ztest_od_t *od; |
428870ff BB |
4109 | uint64_t offset = (1ULL << (ztest_random(4) + SPA_MAXBLOCKSHIFT)) + |
4110 | (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
4111 | uint64_t count = ztest_random(20) + 1; | |
4112 | uint64_t blocksize = ztest_random_blocksize(); | |
4113 | void *data; | |
34dc7c2f | 4114 | |
40b84e7a | 4115 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
34dc7c2f | 4116 | |
40b84e7a BB |
4117 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0); |
4118 | ||
4119 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), !ztest_random(2)) != 0) { | |
4120 | umem_free(od, sizeof(ztest_od_t)); | |
34dc7c2f | 4121 | return; |
40b84e7a | 4122 | } |
34dc7c2f | 4123 | |
40b84e7a BB |
4124 | if (ztest_truncate(zd, od->od_object, offset, count * blocksize) != 0) { |
4125 | umem_free(od, sizeof(ztest_od_t)); | |
34dc7c2f | 4126 | return; |
40b84e7a | 4127 | } |
34dc7c2f | 4128 | |
40b84e7a | 4129 | ztest_prealloc(zd, od->od_object, offset, count * blocksize); |
34dc7c2f | 4130 | |
428870ff | 4131 | data = umem_zalloc(blocksize, UMEM_NOFAIL); |
34dc7c2f | 4132 | |
428870ff BB |
4133 | while (ztest_random(count) != 0) { |
4134 | uint64_t randoff = offset + (ztest_random(count) * blocksize); | |
40b84e7a | 4135 | if (ztest_write(zd, od->od_object, randoff, blocksize, |
428870ff BB |
4136 | data) != 0) |
4137 | break; | |
4138 | while (ztest_random(4) != 0) | |
40b84e7a | 4139 | ztest_io(zd, od->od_object, randoff); |
9babb374 | 4140 | } |
34dc7c2f | 4141 | |
428870ff | 4142 | umem_free(data, blocksize); |
40b84e7a | 4143 | umem_free(od, sizeof(ztest_od_t)); |
34dc7c2f BB |
4144 | } |
4145 | ||
4146 | /* | |
4147 | * Verify that zap_{create,destroy,add,remove,update} work as expected. | |
4148 | */ | |
4149 | #define ZTEST_ZAP_MIN_INTS 1 | |
4150 | #define ZTEST_ZAP_MAX_INTS 4 | |
4151 | #define ZTEST_ZAP_MAX_PROPS 1000 | |
4152 | ||
4153 | void | |
428870ff | 4154 | ztest_zap(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4155 | { |
428870ff | 4156 | objset_t *os = zd->zd_os; |
40b84e7a | 4157 | ztest_od_t *od; |
34dc7c2f BB |
4158 | uint64_t object; |
4159 | uint64_t txg, last_txg; | |
4160 | uint64_t value[ZTEST_ZAP_MAX_INTS]; | |
4161 | uint64_t zl_ints, zl_intsize, prop; | |
4162 | int i, ints; | |
4163 | dmu_tx_t *tx; | |
4164 | char propname[100], txgname[100]; | |
4165 | int error; | |
34dc7c2f BB |
4166 | char *hc[2] = { "s.acl.h", ".s.open.h.hyLZlg" }; |
4167 | ||
40b84e7a BB |
4168 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
4169 | ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0); | |
34dc7c2f | 4170 | |
40b84e7a BB |
4171 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
4172 | !ztest_random(2)) != 0) | |
4173 | goto out; | |
34dc7c2f | 4174 | |
40b84e7a | 4175 | object = od->od_object; |
34dc7c2f | 4176 | |
428870ff BB |
4177 | /* |
4178 | * Generate a known hash collision, and verify that | |
4179 | * we can lookup and remove both entries. | |
4180 | */ | |
4181 | tx = dmu_tx_create(os); | |
4182 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); | |
4183 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4184 | if (txg == 0) | |
40b84e7a | 4185 | goto out; |
428870ff BB |
4186 | for (i = 0; i < 2; i++) { |
4187 | value[i] = i; | |
4188 | VERIFY3U(0, ==, zap_add(os, object, hc[i], sizeof (uint64_t), | |
4189 | 1, &value[i], tx)); | |
4190 | } | |
4191 | for (i = 0; i < 2; i++) { | |
4192 | VERIFY3U(EEXIST, ==, zap_add(os, object, hc[i], | |
4193 | sizeof (uint64_t), 1, &value[i], tx)); | |
4194 | VERIFY3U(0, ==, | |
4195 | zap_length(os, object, hc[i], &zl_intsize, &zl_ints)); | |
4196 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
4197 | ASSERT3U(zl_ints, ==, 1); | |
4198 | } | |
4199 | for (i = 0; i < 2; i++) { | |
4200 | VERIFY3U(0, ==, zap_remove(os, object, hc[i], tx)); | |
34dc7c2f | 4201 | } |
428870ff | 4202 | dmu_tx_commit(tx); |
34dc7c2f | 4203 | |
428870ff BB |
4204 | /* |
4205 | * Generate a buch of random entries. | |
4206 | */ | |
34dc7c2f BB |
4207 | ints = MAX(ZTEST_ZAP_MIN_INTS, object % ZTEST_ZAP_MAX_INTS); |
4208 | ||
4209 | prop = ztest_random(ZTEST_ZAP_MAX_PROPS); | |
4210 | (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop); | |
4211 | (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop); | |
4212 | bzero(value, sizeof (value)); | |
4213 | last_txg = 0; | |
4214 | ||
4215 | /* | |
4216 | * If these zap entries already exist, validate their contents. | |
4217 | */ | |
4218 | error = zap_length(os, object, txgname, &zl_intsize, &zl_ints); | |
4219 | if (error == 0) { | |
4220 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
4221 | ASSERT3U(zl_ints, ==, 1); | |
4222 | ||
4223 | VERIFY(zap_lookup(os, object, txgname, zl_intsize, | |
4224 | zl_ints, &last_txg) == 0); | |
4225 | ||
4226 | VERIFY(zap_length(os, object, propname, &zl_intsize, | |
4227 | &zl_ints) == 0); | |
4228 | ||
4229 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
4230 | ASSERT3U(zl_ints, ==, ints); | |
4231 | ||
4232 | VERIFY(zap_lookup(os, object, propname, zl_intsize, | |
4233 | zl_ints, value) == 0); | |
4234 | ||
4235 | for (i = 0; i < ints; i++) { | |
4236 | ASSERT3U(value[i], ==, last_txg + object + i); | |
4237 | } | |
4238 | } else { | |
4239 | ASSERT3U(error, ==, ENOENT); | |
4240 | } | |
4241 | ||
4242 | /* | |
4243 | * Atomically update two entries in our zap object. | |
4244 | * The first is named txg_%llu, and contains the txg | |
4245 | * in which the property was last updated. The second | |
4246 | * is named prop_%llu, and the nth element of its value | |
4247 | * should be txg + object + n. | |
4248 | */ | |
4249 | tx = dmu_tx_create(os); | |
428870ff BB |
4250 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
4251 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4252 | if (txg == 0) | |
40b84e7a | 4253 | goto out; |
34dc7c2f BB |
4254 | |
4255 | if (last_txg > txg) | |
4256 | fatal(0, "zap future leak: old %llu new %llu", last_txg, txg); | |
4257 | ||
4258 | for (i = 0; i < ints; i++) | |
4259 | value[i] = txg + object + i; | |
4260 | ||
428870ff BB |
4261 | VERIFY3U(0, ==, zap_update(os, object, txgname, sizeof (uint64_t), |
4262 | 1, &txg, tx)); | |
4263 | VERIFY3U(0, ==, zap_update(os, object, propname, sizeof (uint64_t), | |
4264 | ints, value, tx)); | |
34dc7c2f BB |
4265 | |
4266 | dmu_tx_commit(tx); | |
4267 | ||
4268 | /* | |
4269 | * Remove a random pair of entries. | |
4270 | */ | |
4271 | prop = ztest_random(ZTEST_ZAP_MAX_PROPS); | |
4272 | (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop); | |
4273 | (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop); | |
4274 | ||
4275 | error = zap_length(os, object, txgname, &zl_intsize, &zl_ints); | |
4276 | ||
4277 | if (error == ENOENT) | |
40b84e7a | 4278 | goto out; |
34dc7c2f | 4279 | |
c99c9001 | 4280 | ASSERT0(error); |
34dc7c2f BB |
4281 | |
4282 | tx = dmu_tx_create(os); | |
428870ff BB |
4283 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
4284 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4285 | if (txg == 0) | |
40b84e7a | 4286 | goto out; |
428870ff BB |
4287 | VERIFY3U(0, ==, zap_remove(os, object, txgname, tx)); |
4288 | VERIFY3U(0, ==, zap_remove(os, object, propname, tx)); | |
4289 | dmu_tx_commit(tx); | |
40b84e7a BB |
4290 | out: |
4291 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff | 4292 | } |
34dc7c2f | 4293 | |
428870ff BB |
4294 | /* |
4295 | * Testcase to test the upgrading of a microzap to fatzap. | |
4296 | */ | |
4297 | void | |
4298 | ztest_fzap(ztest_ds_t *zd, uint64_t id) | |
4299 | { | |
4300 | objset_t *os = zd->zd_os; | |
40b84e7a | 4301 | ztest_od_t *od; |
428870ff | 4302 | uint64_t object, txg; |
d6320ddb | 4303 | int i; |
34dc7c2f | 4304 | |
40b84e7a BB |
4305 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
4306 | ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0); | |
428870ff | 4307 | |
40b84e7a BB |
4308 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
4309 | !ztest_random(2)) != 0) | |
4310 | goto out; | |
4311 | object = od->od_object; | |
34dc7c2f BB |
4312 | |
4313 | /* | |
428870ff BB |
4314 | * Add entries to this ZAP and make sure it spills over |
4315 | * and gets upgraded to a fatzap. Also, since we are adding | |
4316 | * 2050 entries we should see ptrtbl growth and leaf-block split. | |
34dc7c2f | 4317 | */ |
d6320ddb | 4318 | for (i = 0; i < 2050; i++) { |
428870ff BB |
4319 | char name[MAXNAMELEN]; |
4320 | uint64_t value = i; | |
4321 | dmu_tx_t *tx; | |
4322 | int error; | |
34dc7c2f | 4323 | |
428870ff | 4324 | (void) snprintf(name, sizeof (name), "fzap-%llu-%llu", |
b8864a23 | 4325 | (u_longlong_t)id, (u_longlong_t)value); |
428870ff BB |
4326 | |
4327 | tx = dmu_tx_create(os); | |
4328 | dmu_tx_hold_zap(tx, object, B_TRUE, name); | |
4329 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4330 | if (txg == 0) | |
40b84e7a | 4331 | goto out; |
428870ff BB |
4332 | error = zap_add(os, object, name, sizeof (uint64_t), 1, |
4333 | &value, tx); | |
4334 | ASSERT(error == 0 || error == EEXIST); | |
4335 | dmu_tx_commit(tx); | |
34dc7c2f | 4336 | } |
40b84e7a BB |
4337 | out: |
4338 | umem_free(od, sizeof(ztest_od_t)); | |
34dc7c2f BB |
4339 | } |
4340 | ||
428870ff | 4341 | /* ARGSUSED */ |
34dc7c2f | 4342 | void |
428870ff | 4343 | ztest_zap_parallel(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4344 | { |
428870ff | 4345 | objset_t *os = zd->zd_os; |
40b84e7a | 4346 | ztest_od_t *od; |
34dc7c2f BB |
4347 | uint64_t txg, object, count, wsize, wc, zl_wsize, zl_wc; |
4348 | dmu_tx_t *tx; | |
4349 | int i, namelen, error; | |
428870ff | 4350 | int micro = ztest_random(2); |
34dc7c2f BB |
4351 | char name[20], string_value[20]; |
4352 | void *data; | |
4353 | ||
40b84e7a BB |
4354 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
4355 | ztest_od_init(od, ID_PARALLEL, FTAG, micro, DMU_OT_ZAP_OTHER, 0, 0); | |
428870ff | 4356 | |
40b84e7a BB |
4357 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
4358 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff | 4359 | return; |
40b84e7a | 4360 | } |
428870ff | 4361 | |
40b84e7a | 4362 | object = od->od_object; |
428870ff | 4363 | |
34dc7c2f BB |
4364 | /* |
4365 | * Generate a random name of the form 'xxx.....' where each | |
4366 | * x is a random printable character and the dots are dots. | |
4367 | * There are 94 such characters, and the name length goes from | |
4368 | * 6 to 20, so there are 94^3 * 15 = 12,458,760 possible names. | |
4369 | */ | |
4370 | namelen = ztest_random(sizeof (name) - 5) + 5 + 1; | |
4371 | ||
4372 | for (i = 0; i < 3; i++) | |
4373 | name[i] = '!' + ztest_random('~' - '!' + 1); | |
4374 | for (; i < namelen - 1; i++) | |
4375 | name[i] = '.'; | |
4376 | name[i] = '\0'; | |
4377 | ||
428870ff | 4378 | if ((namelen & 1) || micro) { |
34dc7c2f BB |
4379 | wsize = sizeof (txg); |
4380 | wc = 1; | |
4381 | data = &txg; | |
4382 | } else { | |
4383 | wsize = 1; | |
4384 | wc = namelen; | |
4385 | data = string_value; | |
4386 | } | |
4387 | ||
4388 | count = -1ULL; | |
13fe0198 | 4389 | VERIFY0(zap_count(os, object, &count)); |
34dc7c2f BB |
4390 | ASSERT(count != -1ULL); |
4391 | ||
4392 | /* | |
4393 | * Select an operation: length, lookup, add, update, remove. | |
4394 | */ | |
4395 | i = ztest_random(5); | |
4396 | ||
4397 | if (i >= 2) { | |
4398 | tx = dmu_tx_create(os); | |
428870ff BB |
4399 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
4400 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4401 | if (txg == 0) | |
34dc7c2f | 4402 | return; |
34dc7c2f BB |
4403 | bcopy(name, string_value, namelen); |
4404 | } else { | |
4405 | tx = NULL; | |
4406 | txg = 0; | |
4407 | bzero(string_value, namelen); | |
4408 | } | |
4409 | ||
4410 | switch (i) { | |
4411 | ||
4412 | case 0: | |
4413 | error = zap_length(os, object, name, &zl_wsize, &zl_wc); | |
4414 | if (error == 0) { | |
4415 | ASSERT3U(wsize, ==, zl_wsize); | |
4416 | ASSERT3U(wc, ==, zl_wc); | |
4417 | } else { | |
4418 | ASSERT3U(error, ==, ENOENT); | |
4419 | } | |
4420 | break; | |
4421 | ||
4422 | case 1: | |
4423 | error = zap_lookup(os, object, name, wsize, wc, data); | |
4424 | if (error == 0) { | |
4425 | if (data == string_value && | |
4426 | bcmp(name, data, namelen) != 0) | |
4427 | fatal(0, "name '%s' != val '%s' len %d", | |
4428 | name, data, namelen); | |
4429 | } else { | |
4430 | ASSERT3U(error, ==, ENOENT); | |
4431 | } | |
4432 | break; | |
4433 | ||
4434 | case 2: | |
4435 | error = zap_add(os, object, name, wsize, wc, data, tx); | |
4436 | ASSERT(error == 0 || error == EEXIST); | |
4437 | break; | |
4438 | ||
4439 | case 3: | |
4440 | VERIFY(zap_update(os, object, name, wsize, wc, data, tx) == 0); | |
4441 | break; | |
4442 | ||
4443 | case 4: | |
4444 | error = zap_remove(os, object, name, tx); | |
4445 | ASSERT(error == 0 || error == ENOENT); | |
4446 | break; | |
4447 | } | |
4448 | ||
4449 | if (tx != NULL) | |
4450 | dmu_tx_commit(tx); | |
40b84e7a BB |
4451 | |
4452 | umem_free(od, sizeof(ztest_od_t)); | |
34dc7c2f BB |
4453 | } |
4454 | ||
428870ff BB |
4455 | /* |
4456 | * Commit callback data. | |
4457 | */ | |
4458 | typedef struct ztest_cb_data { | |
4459 | list_node_t zcd_node; | |
4460 | uint64_t zcd_txg; | |
4461 | int zcd_expected_err; | |
4462 | boolean_t zcd_added; | |
4463 | boolean_t zcd_called; | |
4464 | spa_t *zcd_spa; | |
4465 | } ztest_cb_data_t; | |
4466 | ||
4467 | /* This is the actual commit callback function */ | |
4468 | static void | |
4469 | ztest_commit_callback(void *arg, int error) | |
4470 | { | |
4471 | ztest_cb_data_t *data = arg; | |
4472 | uint64_t synced_txg; | |
4473 | ||
4474 | VERIFY(data != NULL); | |
4475 | VERIFY3S(data->zcd_expected_err, ==, error); | |
4476 | VERIFY(!data->zcd_called); | |
4477 | ||
4478 | synced_txg = spa_last_synced_txg(data->zcd_spa); | |
4479 | if (data->zcd_txg > synced_txg) | |
4480 | fatal(0, "commit callback of txg %" PRIu64 " called prematurely" | |
4481 | ", last synced txg = %" PRIu64 "\n", data->zcd_txg, | |
4482 | synced_txg); | |
4483 | ||
4484 | data->zcd_called = B_TRUE; | |
4485 | ||
4486 | if (error == ECANCELED) { | |
c99c9001 | 4487 | ASSERT0(data->zcd_txg); |
428870ff BB |
4488 | ASSERT(!data->zcd_added); |
4489 | ||
4490 | /* | |
4491 | * The private callback data should be destroyed here, but | |
4492 | * since we are going to check the zcd_called field after | |
4493 | * dmu_tx_abort(), we will destroy it there. | |
4494 | */ | |
4495 | return; | |
4496 | } | |
4497 | ||
090ff092 | 4498 | ASSERT(data->zcd_added); |
428870ff BB |
4499 | ASSERT3U(data->zcd_txg, !=, 0); |
4500 | ||
1e33ac1e | 4501 | (void) mutex_enter(&zcl.zcl_callbacks_lock); |
090ff092 RC |
4502 | |
4503 | /* See if this cb was called more quickly */ | |
4504 | if ((synced_txg - data->zcd_txg) < zc_min_txg_delay) | |
4505 | zc_min_txg_delay = synced_txg - data->zcd_txg; | |
4506 | ||
4507 | /* Remove our callback from the list */ | |
428870ff | 4508 | list_remove(&zcl.zcl_callbacks, data); |
090ff092 | 4509 | |
1e33ac1e | 4510 | (void) mutex_exit(&zcl.zcl_callbacks_lock); |
428870ff | 4511 | |
428870ff BB |
4512 | umem_free(data, sizeof (ztest_cb_data_t)); |
4513 | } | |
4514 | ||
4515 | /* Allocate and initialize callback data structure */ | |
4516 | static ztest_cb_data_t * | |
4517 | ztest_create_cb_data(objset_t *os, uint64_t txg) | |
4518 | { | |
4519 | ztest_cb_data_t *cb_data; | |
4520 | ||
4521 | cb_data = umem_zalloc(sizeof (ztest_cb_data_t), UMEM_NOFAIL); | |
4522 | ||
4523 | cb_data->zcd_txg = txg; | |
4524 | cb_data->zcd_spa = dmu_objset_spa(os); | |
1e33ac1e | 4525 | list_link_init(&cb_data->zcd_node); |
428870ff BB |
4526 | |
4527 | return (cb_data); | |
4528 | } | |
4529 | ||
428870ff BB |
4530 | /* |
4531 | * Commit callback test. | |
4532 | */ | |
34dc7c2f | 4533 | void |
428870ff BB |
4534 | ztest_dmu_commit_callbacks(ztest_ds_t *zd, uint64_t id) |
4535 | { | |
4536 | objset_t *os = zd->zd_os; | |
40b84e7a | 4537 | ztest_od_t *od; |
428870ff BB |
4538 | dmu_tx_t *tx; |
4539 | ztest_cb_data_t *cb_data[3], *tmp_cb; | |
4540 | uint64_t old_txg, txg; | |
090ff092 | 4541 | int i, error = 0; |
428870ff | 4542 | |
40b84e7a BB |
4543 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
4544 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0); | |
428870ff | 4545 | |
40b84e7a BB |
4546 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
4547 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff | 4548 | return; |
40b84e7a | 4549 | } |
428870ff BB |
4550 | |
4551 | tx = dmu_tx_create(os); | |
4552 | ||
4553 | cb_data[0] = ztest_create_cb_data(os, 0); | |
4554 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[0]); | |
4555 | ||
40b84e7a | 4556 | dmu_tx_hold_write(tx, od->od_object, 0, sizeof (uint64_t)); |
428870ff BB |
4557 | |
4558 | /* Every once in a while, abort the transaction on purpose */ | |
4559 | if (ztest_random(100) == 0) | |
4560 | error = -1; | |
4561 | ||
4562 | if (!error) | |
4563 | error = dmu_tx_assign(tx, TXG_NOWAIT); | |
4564 | ||
4565 | txg = error ? 0 : dmu_tx_get_txg(tx); | |
4566 | ||
4567 | cb_data[0]->zcd_txg = txg; | |
4568 | cb_data[1] = ztest_create_cb_data(os, txg); | |
4569 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[1]); | |
4570 | ||
4571 | if (error) { | |
4572 | /* | |
4573 | * It's not a strict requirement to call the registered | |
4574 | * callbacks from inside dmu_tx_abort(), but that's what | |
4575 | * it's supposed to happen in the current implementation | |
4576 | * so we will check for that. | |
4577 | */ | |
4578 | for (i = 0; i < 2; i++) { | |
4579 | cb_data[i]->zcd_expected_err = ECANCELED; | |
4580 | VERIFY(!cb_data[i]->zcd_called); | |
4581 | } | |
4582 | ||
4583 | dmu_tx_abort(tx); | |
4584 | ||
4585 | for (i = 0; i < 2; i++) { | |
4586 | VERIFY(cb_data[i]->zcd_called); | |
4587 | umem_free(cb_data[i], sizeof (ztest_cb_data_t)); | |
4588 | } | |
4589 | ||
40b84e7a | 4590 | umem_free(od, sizeof(ztest_od_t)); |
428870ff BB |
4591 | return; |
4592 | } | |
4593 | ||
4594 | cb_data[2] = ztest_create_cb_data(os, txg); | |
4595 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[2]); | |
4596 | ||
4597 | /* | |
4598 | * Read existing data to make sure there isn't a future leak. | |
4599 | */ | |
40b84e7a | 4600 | VERIFY(0 == dmu_read(os, od->od_object, 0, sizeof (uint64_t), |
428870ff BB |
4601 | &old_txg, DMU_READ_PREFETCH)); |
4602 | ||
4603 | if (old_txg > txg) | |
4604 | fatal(0, "future leak: got %" PRIu64 ", open txg is %" PRIu64, | |
4605 | old_txg, txg); | |
4606 | ||
40b84e7a | 4607 | dmu_write(os, od->od_object, 0, sizeof (uint64_t), &txg, tx); |
428870ff | 4608 | |
1e33ac1e | 4609 | (void) mutex_enter(&zcl.zcl_callbacks_lock); |
428870ff BB |
4610 | |
4611 | /* | |
4612 | * Since commit callbacks don't have any ordering requirement and since | |
4613 | * it is theoretically possible for a commit callback to be called | |
4614 | * after an arbitrary amount of time has elapsed since its txg has been | |
4615 | * synced, it is difficult to reliably determine whether a commit | |
4616 | * callback hasn't been called due to high load or due to a flawed | |
4617 | * implementation. | |
4618 | * | |
4619 | * In practice, we will assume that if after a certain number of txgs a | |
4620 | * commit callback hasn't been called, then most likely there's an | |
4621 | * implementation bug.. | |
4622 | */ | |
4623 | tmp_cb = list_head(&zcl.zcl_callbacks); | |
4624 | if (tmp_cb != NULL && | |
090ff092 | 4625 | tmp_cb->zcd_txg + ZTEST_COMMIT_CB_THRESH < txg) { |
428870ff BB |
4626 | fatal(0, "Commit callback threshold exceeded, oldest txg: %" |
4627 | PRIu64 ", open txg: %" PRIu64 "\n", tmp_cb->zcd_txg, txg); | |
4628 | } | |
4629 | ||
4630 | /* | |
4631 | * Let's find the place to insert our callbacks. | |
4632 | * | |
4633 | * Even though the list is ordered by txg, it is possible for the | |
4634 | * insertion point to not be the end because our txg may already be | |
4635 | * quiescing at this point and other callbacks in the open txg | |
4636 | * (from other objsets) may have sneaked in. | |
4637 | */ | |
4638 | tmp_cb = list_tail(&zcl.zcl_callbacks); | |
4639 | while (tmp_cb != NULL && tmp_cb->zcd_txg > txg) | |
4640 | tmp_cb = list_prev(&zcl.zcl_callbacks, tmp_cb); | |
4641 | ||
4642 | /* Add the 3 callbacks to the list */ | |
4643 | for (i = 0; i < 3; i++) { | |
4644 | if (tmp_cb == NULL) | |
4645 | list_insert_head(&zcl.zcl_callbacks, cb_data[i]); | |
4646 | else | |
4647 | list_insert_after(&zcl.zcl_callbacks, tmp_cb, | |
4648 | cb_data[i]); | |
4649 | ||
4650 | cb_data[i]->zcd_added = B_TRUE; | |
4651 | VERIFY(!cb_data[i]->zcd_called); | |
4652 | ||
4653 | tmp_cb = cb_data[i]; | |
4654 | } | |
4655 | ||
090ff092 RC |
4656 | zc_cb_counter += 3; |
4657 | ||
1e33ac1e | 4658 | (void) mutex_exit(&zcl.zcl_callbacks_lock); |
428870ff BB |
4659 | |
4660 | dmu_tx_commit(tx); | |
40b84e7a BB |
4661 | |
4662 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff BB |
4663 | } |
4664 | ||
4665 | /* ARGSUSED */ | |
4666 | void | |
4667 | ztest_dsl_prop_get_set(ztest_ds_t *zd, uint64_t id) | |
4668 | { | |
4669 | zfs_prop_t proplist[] = { | |
4670 | ZFS_PROP_CHECKSUM, | |
4671 | ZFS_PROP_COMPRESSION, | |
4672 | ZFS_PROP_COPIES, | |
4673 | ZFS_PROP_DEDUP | |
4674 | }; | |
d6320ddb | 4675 | int p; |
428870ff | 4676 | |
c242c188 | 4677 | (void) rw_enter(&ztest_name_lock, RW_READER); |
428870ff | 4678 | |
d6320ddb | 4679 | for (p = 0; p < sizeof (proplist) / sizeof (proplist[0]); p++) |
428870ff BB |
4680 | (void) ztest_dsl_prop_set_uint64(zd->zd_name, proplist[p], |
4681 | ztest_random_dsl_prop(proplist[p]), (int)ztest_random(2)); | |
4682 | ||
c242c188 | 4683 | (void) rw_exit(&ztest_name_lock); |
428870ff BB |
4684 | } |
4685 | ||
4686 | /* ARGSUSED */ | |
4687 | void | |
4688 | ztest_spa_prop_get_set(ztest_ds_t *zd, uint64_t id) | |
4689 | { | |
428870ff BB |
4690 | nvlist_t *props = NULL; |
4691 | ||
c242c188 | 4692 | (void) rw_enter(&ztest_name_lock, RW_READER); |
428870ff | 4693 | |
c242c188 | 4694 | (void) ztest_spa_prop_set_uint64(ZPOOL_PROP_DEDUPDITTO, |
428870ff BB |
4695 | ZIO_DEDUPDITTO_MIN + ztest_random(ZIO_DEDUPDITTO_MIN)); |
4696 | ||
c99c9001 | 4697 | VERIFY0(spa_prop_get(ztest_spa, &props)); |
428870ff | 4698 | |
c242c188 | 4699 | if (ztest_opts.zo_verbose >= 6) |
428870ff BB |
4700 | dump_nvlist(props, 4); |
4701 | ||
4702 | nvlist_free(props); | |
4703 | ||
c242c188 | 4704 | (void) rw_exit(&ztest_name_lock); |
428870ff BB |
4705 | } |
4706 | ||
13fe0198 MA |
4707 | static int |
4708 | user_release_one(const char *snapname, const char *holdname) | |
4709 | { | |
4710 | nvlist_t *snaps, *holds; | |
4711 | int error; | |
4712 | ||
4713 | snaps = fnvlist_alloc(); | |
4714 | holds = fnvlist_alloc(); | |
4715 | fnvlist_add_boolean(holds, holdname); | |
4716 | fnvlist_add_nvlist(snaps, snapname, holds); | |
4717 | fnvlist_free(holds); | |
4718 | error = dsl_dataset_user_release(snaps, NULL); | |
4719 | fnvlist_free(snaps); | |
4720 | return (error); | |
4721 | } | |
4722 | ||
428870ff BB |
4723 | /* |
4724 | * Test snapshot hold/release and deferred destroy. | |
4725 | */ | |
4726 | void | |
4727 | ztest_dmu_snapshot_hold(ztest_ds_t *zd, uint64_t id) | |
34dc7c2f | 4728 | { |
34dc7c2f | 4729 | int error; |
428870ff BB |
4730 | objset_t *os = zd->zd_os; |
4731 | objset_t *origin; | |
4732 | char snapname[100]; | |
4733 | char fullname[100]; | |
4734 | char clonename[100]; | |
4735 | char tag[100]; | |
4736 | char osname[MAXNAMELEN]; | |
13fe0198 | 4737 | nvlist_t *holds; |
34dc7c2f | 4738 | |
c242c188 | 4739 | (void) rw_enter(&ztest_name_lock, RW_READER); |
34dc7c2f BB |
4740 | |
4741 | dmu_objset_name(os, osname); | |
4742 | ||
13fe0198 MA |
4743 | (void) snprintf(snapname, sizeof (snapname), "sh1_%llu", (long long unsigned int)id); |
4744 | (void) snprintf(fullname, sizeof (fullname), "%s@%s", osname, snapname); | |
4745 | (void) snprintf(clonename, sizeof (clonename), | |
4746 | "%s/ch1_%llu", osname, (long long unsigned int)id); | |
4747 | (void) snprintf(tag, sizeof (tag), "tag_%llu", (long long unsigned int)id); | |
428870ff BB |
4748 | |
4749 | /* | |
4750 | * Clean up from any previous run. | |
4751 | */ | |
13fe0198 MA |
4752 | error = dsl_destroy_head(clonename); |
4753 | if (error != ENOENT) | |
4754 | ASSERT0(error); | |
4755 | error = user_release_one(fullname, tag); | |
4756 | if (error != ESRCH && error != ENOENT) | |
4757 | ASSERT0(error); | |
4758 | error = dsl_destroy_snapshot(fullname, B_FALSE); | |
4759 | if (error != ENOENT) | |
4760 | ASSERT0(error); | |
428870ff BB |
4761 | |
4762 | /* | |
4763 | * Create snapshot, clone it, mark snap for deferred destroy, | |
4764 | * destroy clone, verify snap was also destroyed. | |
4765 | */ | |
6f1ffb06 | 4766 | error = dmu_objset_snapshot_one(osname, snapname); |
428870ff BB |
4767 | if (error) { |
4768 | if (error == ENOSPC) { | |
4769 | ztest_record_enospc("dmu_objset_snapshot"); | |
4770 | goto out; | |
34dc7c2f | 4771 | } |
428870ff BB |
4772 | fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error); |
4773 | } | |
34dc7c2f | 4774 | |
13fe0198 | 4775 | error = dmu_objset_clone(clonename, fullname); |
428870ff | 4776 | if (error) { |
34dc7c2f | 4777 | if (error == ENOSPC) { |
428870ff BB |
4778 | ztest_record_enospc("dmu_objset_clone"); |
4779 | goto out; | |
34dc7c2f | 4780 | } |
428870ff BB |
4781 | fatal(0, "dmu_objset_clone(%s) = %d", clonename, error); |
4782 | } | |
34dc7c2f | 4783 | |
13fe0198 | 4784 | error = dsl_destroy_snapshot(fullname, B_TRUE); |
428870ff | 4785 | if (error) { |
13fe0198 | 4786 | fatal(0, "dsl_destroy_snapshot(%s, B_TRUE) = %d", |
428870ff BB |
4787 | fullname, error); |
4788 | } | |
34dc7c2f | 4789 | |
13fe0198 | 4790 | error = dsl_destroy_head(clonename); |
428870ff | 4791 | if (error) |
13fe0198 | 4792 | fatal(0, "dsl_destroy_head(%s) = %d", clonename, error); |
34dc7c2f | 4793 | |
428870ff BB |
4794 | error = dmu_objset_hold(fullname, FTAG, &origin); |
4795 | if (error != ENOENT) | |
4796 | fatal(0, "dmu_objset_hold(%s) = %d", fullname, error); | |
34dc7c2f | 4797 | |
428870ff BB |
4798 | /* |
4799 | * Create snapshot, add temporary hold, verify that we can't | |
4800 | * destroy a held snapshot, mark for deferred destroy, | |
4801 | * release hold, verify snapshot was destroyed. | |
4802 | */ | |
6f1ffb06 | 4803 | error = dmu_objset_snapshot_one(osname, snapname); |
428870ff BB |
4804 | if (error) { |
4805 | if (error == ENOSPC) { | |
4806 | ztest_record_enospc("dmu_objset_snapshot"); | |
4807 | goto out; | |
34dc7c2f | 4808 | } |
428870ff BB |
4809 | fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error); |
4810 | } | |
4811 | ||
13fe0198 MA |
4812 | holds = fnvlist_alloc(); |
4813 | fnvlist_add_string(holds, fullname, tag); | |
4814 | error = dsl_dataset_user_hold(holds, 0, NULL); | |
4815 | fnvlist_free(holds); | |
4816 | ||
428870ff BB |
4817 | if (error) |
4818 | fatal(0, "dsl_dataset_user_hold(%s)", fullname, tag); | |
4819 | ||
13fe0198 | 4820 | error = dsl_destroy_snapshot(fullname, B_FALSE); |
428870ff | 4821 | if (error != EBUSY) { |
13fe0198 | 4822 | fatal(0, "dsl_destroy_snapshot(%s, B_FALSE) = %d", |
428870ff BB |
4823 | fullname, error); |
4824 | } | |
4825 | ||
13fe0198 | 4826 | error = dsl_destroy_snapshot(fullname, B_TRUE); |
428870ff | 4827 | if (error) { |
13fe0198 | 4828 | fatal(0, "dsl_destroy_snapshot(%s, B_TRUE) = %d", |
428870ff | 4829 | fullname, error); |
34dc7c2f BB |
4830 | } |
4831 | ||
13fe0198 | 4832 | error = user_release_one(fullname, tag); |
428870ff | 4833 | if (error) |
95fd54a1 | 4834 | fatal(0, "user_release_one(%s, %s) = %d", fullname, tag, error); |
428870ff | 4835 | |
13fe0198 | 4836 | VERIFY3U(dmu_objset_hold(fullname, FTAG, &origin), ==, ENOENT); |
428870ff BB |
4837 | |
4838 | out: | |
c242c188 | 4839 | (void) rw_exit(&ztest_name_lock); |
34dc7c2f BB |
4840 | } |
4841 | ||
34dc7c2f BB |
4842 | /* |
4843 | * Inject random faults into the on-disk data. | |
4844 | */ | |
428870ff | 4845 | /* ARGSUSED */ |
34dc7c2f | 4846 | void |
428870ff | 4847 | ztest_fault_inject(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4848 | { |
428870ff | 4849 | ztest_shared_t *zs = ztest_shared; |
c242c188 | 4850 | spa_t *spa = ztest_spa; |
34dc7c2f BB |
4851 | int fd; |
4852 | uint64_t offset; | |
428870ff | 4853 | uint64_t leaves; |
c5b3a7bb | 4854 | uint64_t bad = 0x1990c0ffeedecadeull; |
34dc7c2f | 4855 | uint64_t top, leaf; |
40b84e7a BB |
4856 | char *path0; |
4857 | char *pathrand; | |
34dc7c2f | 4858 | size_t fsize; |
34dc7c2f BB |
4859 | int bshift = SPA_MAXBLOCKSHIFT + 2; /* don't scrog all labels */ |
4860 | int iters = 1000; | |
428870ff BB |
4861 | int maxfaults; |
4862 | int mirror_save; | |
b128c09f | 4863 | vdev_t *vd0 = NULL; |
34dc7c2f | 4864 | uint64_t guid0 = 0; |
428870ff BB |
4865 | boolean_t islog = B_FALSE; |
4866 | ||
40b84e7a BB |
4867 | path0 = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
4868 | pathrand = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
4869 | ||
c242c188 | 4870 | mutex_enter(&ztest_vdev_lock); |
428870ff | 4871 | maxfaults = MAXFAULTS(); |
c242c188 | 4872 | leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raidz; |
428870ff | 4873 | mirror_save = zs->zs_mirrors; |
c242c188 | 4874 | mutex_exit(&ztest_vdev_lock); |
34dc7c2f | 4875 | |
b128c09f | 4876 | ASSERT(leaves >= 1); |
34dc7c2f BB |
4877 | |
4878 | /* | |
b128c09f | 4879 | * We need SCL_STATE here because we're going to look at vd0->vdev_tsd. |
34dc7c2f | 4880 | */ |
b128c09f | 4881 | spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); |
34dc7c2f | 4882 | |
b128c09f BB |
4883 | if (ztest_random(2) == 0) { |
4884 | /* | |
428870ff | 4885 | * Inject errors on a normal data device or slog device. |
b128c09f | 4886 | */ |
428870ff BB |
4887 | top = ztest_random_vdev_top(spa, B_TRUE); |
4888 | leaf = ztest_random(leaves) + zs->zs_splits; | |
34dc7c2f | 4889 | |
b128c09f BB |
4890 | /* |
4891 | * Generate paths to the first leaf in this top-level vdev, | |
4892 | * and to the random leaf we selected. We'll induce transient | |
4893 | * write failures and random online/offline activity on leaf 0, | |
4894 | * and we'll write random garbage to the randomly chosen leaf. | |
4895 | */ | |
6aec1cd5 | 4896 | (void) snprintf(path0, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
4897 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
4898 | top * leaves + zs->zs_splits); | |
6aec1cd5 | 4899 | (void) snprintf(pathrand, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
4900 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
4901 | top * leaves + leaf); | |
34dc7c2f | 4902 | |
b128c09f | 4903 | vd0 = vdev_lookup_by_path(spa->spa_root_vdev, path0); |
428870ff BB |
4904 | if (vd0 != NULL && vd0->vdev_top->vdev_islog) |
4905 | islog = B_TRUE; | |
4906 | ||
b128c09f BB |
4907 | if (vd0 != NULL && maxfaults != 1) { |
4908 | /* | |
4909 | * Make vd0 explicitly claim to be unreadable, | |
4910 | * or unwriteable, or reach behind its back | |
4911 | * and close the underlying fd. We can do this if | |
4912 | * maxfaults == 0 because we'll fail and reexecute, | |
4913 | * and we can do it if maxfaults >= 2 because we'll | |
4914 | * have enough redundancy. If maxfaults == 1, the | |
4915 | * combination of this with injection of random data | |
4916 | * corruption below exceeds the pool's fault tolerance. | |
4917 | */ | |
4918 | vdev_file_t *vf = vd0->vdev_tsd; | |
4919 | ||
4920 | if (vf != NULL && ztest_random(3) == 0) { | |
4921 | (void) close(vf->vf_vnode->v_fd); | |
4922 | vf->vf_vnode->v_fd = -1; | |
4923 | } else if (ztest_random(2) == 0) { | |
4924 | vd0->vdev_cant_read = B_TRUE; | |
4925 | } else { | |
4926 | vd0->vdev_cant_write = B_TRUE; | |
4927 | } | |
4928 | guid0 = vd0->vdev_guid; | |
4929 | } | |
4930 | } else { | |
4931 | /* | |
4932 | * Inject errors on an l2cache device. | |
4933 | */ | |
4934 | spa_aux_vdev_t *sav = &spa->spa_l2cache; | |
34dc7c2f | 4935 | |
b128c09f BB |
4936 | if (sav->sav_count == 0) { |
4937 | spa_config_exit(spa, SCL_STATE, FTAG); | |
40b84e7a | 4938 | goto out; |
b128c09f BB |
4939 | } |
4940 | vd0 = sav->sav_vdevs[ztest_random(sav->sav_count)]; | |
34dc7c2f | 4941 | guid0 = vd0->vdev_guid; |
b128c09f BB |
4942 | (void) strcpy(path0, vd0->vdev_path); |
4943 | (void) strcpy(pathrand, vd0->vdev_path); | |
4944 | ||
4945 | leaf = 0; | |
4946 | leaves = 1; | |
4947 | maxfaults = INT_MAX; /* no limit on cache devices */ | |
34dc7c2f BB |
4948 | } |
4949 | ||
b128c09f BB |
4950 | spa_config_exit(spa, SCL_STATE, FTAG); |
4951 | ||
34dc7c2f | 4952 | /* |
428870ff BB |
4953 | * If we can tolerate two or more faults, or we're dealing |
4954 | * with a slog, randomly online/offline vd0. | |
34dc7c2f | 4955 | */ |
428870ff | 4956 | if ((maxfaults >= 2 || islog) && guid0 != 0) { |
fb5f0bc8 BB |
4957 | if (ztest_random(10) < 6) { |
4958 | int flags = (ztest_random(2) == 0 ? | |
4959 | ZFS_OFFLINE_TEMPORARY : 0); | |
428870ff BB |
4960 | |
4961 | /* | |
4962 | * We have to grab the zs_name_lock as writer to | |
4963 | * prevent a race between offlining a slog and | |
4964 | * destroying a dataset. Offlining the slog will | |
4965 | * grab a reference on the dataset which may cause | |
13fe0198 | 4966 | * dsl_destroy_head() to fail with EBUSY thus |
428870ff BB |
4967 | * leaving the dataset in an inconsistent state. |
4968 | */ | |
4969 | if (islog) | |
c242c188 | 4970 | (void) rw_enter(&ztest_name_lock, |
1e33ac1e | 4971 | RW_WRITER); |
428870ff | 4972 | |
fb5f0bc8 | 4973 | VERIFY(vdev_offline(spa, guid0, flags) != EBUSY); |
428870ff BB |
4974 | |
4975 | if (islog) | |
c242c188 | 4976 | (void) rw_exit(&ztest_name_lock); |
fb5f0bc8 | 4977 | } else { |
1eb5bfa3 GW |
4978 | /* |
4979 | * Ideally we would like to be able to randomly | |
4980 | * call vdev_[on|off]line without holding locks | |
4981 | * to force unpredictable failures but the side | |
4982 | * effects of vdev_[on|off]line prevent us from | |
4983 | * doing so. We grab the ztest_vdev_lock here to | |
4984 | * prevent a race between injection testing and | |
4985 | * aux_vdev removal. | |
4986 | */ | |
4987 | mutex_enter(&ztest_vdev_lock); | |
fb5f0bc8 | 4988 | (void) vdev_online(spa, guid0, 0, NULL); |
1eb5bfa3 | 4989 | mutex_exit(&ztest_vdev_lock); |
fb5f0bc8 | 4990 | } |
34dc7c2f BB |
4991 | } |
4992 | ||
428870ff | 4993 | if (maxfaults == 0) |
40b84e7a | 4994 | goto out; |
428870ff | 4995 | |
34dc7c2f BB |
4996 | /* |
4997 | * We have at least single-fault tolerance, so inject data corruption. | |
4998 | */ | |
4999 | fd = open(pathrand, O_RDWR); | |
5000 | ||
5001 | if (fd == -1) /* we hit a gap in the device namespace */ | |
40b84e7a | 5002 | goto out; |
34dc7c2f BB |
5003 | |
5004 | fsize = lseek(fd, 0, SEEK_END); | |
5005 | ||
5006 | while (--iters != 0) { | |
5007 | offset = ztest_random(fsize / (leaves << bshift)) * | |
5008 | (leaves << bshift) + (leaf << bshift) + | |
5009 | (ztest_random(1ULL << (bshift - 1)) & -8ULL); | |
5010 | ||
5011 | if (offset >= fsize) | |
5012 | continue; | |
5013 | ||
c242c188 | 5014 | mutex_enter(&ztest_vdev_lock); |
428870ff | 5015 | if (mirror_save != zs->zs_mirrors) { |
c242c188 | 5016 | mutex_exit(&ztest_vdev_lock); |
428870ff | 5017 | (void) close(fd); |
40b84e7a | 5018 | goto out; |
428870ff | 5019 | } |
34dc7c2f BB |
5020 | |
5021 | if (pwrite(fd, &bad, sizeof (bad), offset) != sizeof (bad)) | |
5022 | fatal(1, "can't inject bad word at 0x%llx in %s", | |
5023 | offset, pathrand); | |
428870ff | 5024 | |
c242c188 | 5025 | mutex_exit(&ztest_vdev_lock); |
428870ff | 5026 | |
c242c188 | 5027 | if (ztest_opts.zo_verbose >= 7) |
428870ff BB |
5028 | (void) printf("injected bad word into %s," |
5029 | " offset 0x%llx\n", pathrand, (u_longlong_t)offset); | |
34dc7c2f BB |
5030 | } |
5031 | ||
5032 | (void) close(fd); | |
40b84e7a BB |
5033 | out: |
5034 | umem_free(path0, MAXPATHLEN); | |
5035 | umem_free(pathrand, MAXPATHLEN); | |
34dc7c2f BB |
5036 | } |
5037 | ||
5038 | /* | |
428870ff | 5039 | * Verify that DDT repair works as expected. |
34dc7c2f BB |
5040 | */ |
5041 | void | |
428870ff | 5042 | ztest_ddt_repair(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 5043 | { |
428870ff | 5044 | ztest_shared_t *zs = ztest_shared; |
c242c188 | 5045 | spa_t *spa = ztest_spa; |
428870ff | 5046 | objset_t *os = zd->zd_os; |
40b84e7a | 5047 | ztest_od_t *od; |
428870ff BB |
5048 | uint64_t object, blocksize, txg, pattern, psize; |
5049 | enum zio_checksum checksum = spa_dedup_checksum(spa); | |
5050 | dmu_buf_t *db; | |
5051 | dmu_tx_t *tx; | |
5052 | void *buf; | |
5053 | blkptr_t blk; | |
5054 | int copies = 2 * ZIO_DEDUPDITTO_MIN; | |
d6320ddb | 5055 | int i; |
34dc7c2f | 5056 | |
428870ff BB |
5057 | blocksize = ztest_random_blocksize(); |
5058 | blocksize = MIN(blocksize, 2048); /* because we write so many */ | |
34dc7c2f | 5059 | |
40b84e7a BB |
5060 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
5061 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0); | |
34dc7c2f | 5062 | |
40b84e7a BB |
5063 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
5064 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff | 5065 | return; |
40b84e7a | 5066 | } |
34dc7c2f BB |
5067 | |
5068 | /* | |
428870ff BB |
5069 | * Take the name lock as writer to prevent anyone else from changing |
5070 | * the pool and dataset properies we need to maintain during this test. | |
34dc7c2f | 5071 | */ |
c242c188 | 5072 | (void) rw_enter(&ztest_name_lock, RW_WRITER); |
34dc7c2f | 5073 | |
428870ff BB |
5074 | if (ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_DEDUP, checksum, |
5075 | B_FALSE) != 0 || | |
5076 | ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_COPIES, 1, | |
5077 | B_FALSE) != 0) { | |
c242c188 | 5078 | (void) rw_exit(&ztest_name_lock); |
40b84e7a | 5079 | umem_free(od, sizeof(ztest_od_t)); |
428870ff BB |
5080 | return; |
5081 | } | |
5082 | ||
5083 | object = od[0].od_object; | |
5084 | blocksize = od[0].od_blocksize; | |
3541dc6d | 5085 | pattern = zs->zs_guid ^ dmu_objset_fsid_guid(os); |
428870ff BB |
5086 | |
5087 | ASSERT(object != 0); | |
5088 | ||
5089 | tx = dmu_tx_create(os); | |
5090 | dmu_tx_hold_write(tx, object, 0, copies * blocksize); | |
5091 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
5092 | if (txg == 0) { | |
c242c188 | 5093 | (void) rw_exit(&ztest_name_lock); |
40b84e7a | 5094 | umem_free(od, sizeof(ztest_od_t)); |
428870ff BB |
5095 | return; |
5096 | } | |
34dc7c2f BB |
5097 | |
5098 | /* | |
428870ff | 5099 | * Write all the copies of our block. |
34dc7c2f | 5100 | */ |
d6320ddb | 5101 | for (i = 0; i < copies; i++) { |
428870ff | 5102 | uint64_t offset = i * blocksize; |
13fe0198 MA |
5103 | int error = dmu_buf_hold(os, object, offset, FTAG, &db, |
5104 | DMU_READ_NO_PREFETCH); | |
5105 | if (error != 0) { | |
5106 | fatal(B_FALSE, "dmu_buf_hold(%p, %llu, %llu) = %u", | |
5107 | os, (long long)object, (long long) offset, error); | |
5108 | } | |
428870ff BB |
5109 | ASSERT(db->db_offset == offset); |
5110 | ASSERT(db->db_size == blocksize); | |
5111 | ASSERT(ztest_pattern_match(db->db_data, db->db_size, pattern) || | |
5112 | ztest_pattern_match(db->db_data, db->db_size, 0ULL)); | |
5113 | dmu_buf_will_fill(db, tx); | |
5114 | ztest_pattern_set(db->db_data, db->db_size, pattern); | |
5115 | dmu_buf_rele(db, FTAG); | |
5116 | } | |
34dc7c2f | 5117 | |
428870ff BB |
5118 | dmu_tx_commit(tx); |
5119 | txg_wait_synced(spa_get_dsl(spa), txg); | |
34dc7c2f BB |
5120 | |
5121 | /* | |
428870ff | 5122 | * Find out what block we got. |
34dc7c2f | 5123 | */ |
428870ff BB |
5124 | VERIFY(dmu_buf_hold(os, object, 0, FTAG, &db, |
5125 | DMU_READ_NO_PREFETCH) == 0); | |
5126 | blk = *((dmu_buf_impl_t *)db)->db_blkptr; | |
5127 | dmu_buf_rele(db, FTAG); | |
34dc7c2f BB |
5128 | |
5129 | /* | |
428870ff | 5130 | * Damage the block. Dedup-ditto will save us when we read it later. |
34dc7c2f | 5131 | */ |
428870ff BB |
5132 | psize = BP_GET_PSIZE(&blk); |
5133 | buf = zio_buf_alloc(psize); | |
5134 | ztest_pattern_set(buf, psize, ~pattern); | |
34dc7c2f | 5135 | |
428870ff BB |
5136 | (void) zio_wait(zio_rewrite(NULL, spa, 0, &blk, |
5137 | buf, psize, NULL, NULL, ZIO_PRIORITY_SYNC_WRITE, | |
5138 | ZIO_FLAG_CANFAIL | ZIO_FLAG_INDUCE_DAMAGE, NULL)); | |
34dc7c2f | 5139 | |
428870ff | 5140 | zio_buf_free(buf, psize); |
34dc7c2f | 5141 | |
c242c188 | 5142 | (void) rw_exit(&ztest_name_lock); |
40b84e7a | 5143 | umem_free(od, sizeof(ztest_od_t)); |
34dc7c2f BB |
5144 | } |
5145 | ||
34dc7c2f | 5146 | /* |
428870ff | 5147 | * Scrub the pool. |
34dc7c2f | 5148 | */ |
428870ff BB |
5149 | /* ARGSUSED */ |
5150 | void | |
5151 | ztest_scrub(ztest_ds_t *zd, uint64_t id) | |
34dc7c2f | 5152 | { |
c242c188 | 5153 | spa_t *spa = ztest_spa; |
34dc7c2f | 5154 | |
428870ff BB |
5155 | (void) spa_scan(spa, POOL_SCAN_SCRUB); |
5156 | (void) poll(NULL, 0, 100); /* wait a moment, then force a restart */ | |
5157 | (void) spa_scan(spa, POOL_SCAN_SCRUB); | |
5158 | } | |
34dc7c2f | 5159 | |
3541dc6d GA |
5160 | /* |
5161 | * Change the guid for the pool. | |
5162 | */ | |
5163 | /* ARGSUSED */ | |
5164 | void | |
5165 | ztest_reguid(ztest_ds_t *zd, uint64_t id) | |
5166 | { | |
c242c188 | 5167 | spa_t *spa = ztest_spa; |
3541dc6d | 5168 | uint64_t orig, load; |
3bc7e0fb | 5169 | int error; |
3541dc6d GA |
5170 | |
5171 | orig = spa_guid(spa); | |
5172 | load = spa_load_guid(spa); | |
3bc7e0fb GW |
5173 | |
5174 | (void) rw_enter(&ztest_name_lock, RW_WRITER); | |
5175 | error = spa_change_guid(spa); | |
5176 | (void) rw_exit(&ztest_name_lock); | |
5177 | ||
5178 | if (error != 0) | |
3541dc6d GA |
5179 | return; |
5180 | ||
ea0b2538 | 5181 | if (ztest_opts.zo_verbose >= 4) { |
3541dc6d GA |
5182 | (void) printf("Changed guid old %llu -> %llu\n", |
5183 | (u_longlong_t)orig, (u_longlong_t)spa_guid(spa)); | |
5184 | } | |
5185 | ||
5186 | VERIFY3U(orig, !=, spa_guid(spa)); | |
5187 | VERIFY3U(load, ==, spa_load_guid(spa)); | |
5188 | } | |
5189 | ||
428870ff BB |
5190 | /* |
5191 | * Rename the pool to a different name and then rename it back. | |
5192 | */ | |
5193 | /* ARGSUSED */ | |
5194 | void | |
5195 | ztest_spa_rename(ztest_ds_t *zd, uint64_t id) | |
5196 | { | |
428870ff BB |
5197 | char *oldname, *newname; |
5198 | spa_t *spa; | |
34dc7c2f | 5199 | |
c242c188 | 5200 | (void) rw_enter(&ztest_name_lock, RW_WRITER); |
34dc7c2f | 5201 | |
c242c188 | 5202 | oldname = ztest_opts.zo_pool; |
428870ff BB |
5203 | newname = umem_alloc(strlen(oldname) + 5, UMEM_NOFAIL); |
5204 | (void) strcpy(newname, oldname); | |
5205 | (void) strcat(newname, "_tmp"); | |
34dc7c2f BB |
5206 | |
5207 | /* | |
428870ff | 5208 | * Do the rename |
34dc7c2f | 5209 | */ |
428870ff | 5210 | VERIFY3U(0, ==, spa_rename(oldname, newname)); |
34dc7c2f BB |
5211 | |
5212 | /* | |
428870ff | 5213 | * Try to open it under the old name, which shouldn't exist |
34dc7c2f | 5214 | */ |
428870ff | 5215 | VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f BB |
5216 | |
5217 | /* | |
428870ff BB |
5218 | * Open it under the new name and make sure it's still the same spa_t. |
5219 | */ | |
5220 | VERIFY3U(0, ==, spa_open(newname, &spa, FTAG)); | |
34dc7c2f | 5221 | |
c242c188 | 5222 | ASSERT(spa == ztest_spa); |
428870ff | 5223 | spa_close(spa, FTAG); |
34dc7c2f BB |
5224 | |
5225 | /* | |
428870ff | 5226 | * Rename it back to the original |
34dc7c2f | 5227 | */ |
428870ff | 5228 | VERIFY3U(0, ==, spa_rename(newname, oldname)); |
34dc7c2f | 5229 | |
428870ff BB |
5230 | /* |
5231 | * Make sure it can still be opened | |
5232 | */ | |
5233 | VERIFY3U(0, ==, spa_open(oldname, &spa, FTAG)); | |
34dc7c2f | 5234 | |
c242c188 | 5235 | ASSERT(spa == ztest_spa); |
428870ff BB |
5236 | spa_close(spa, FTAG); |
5237 | ||
5238 | umem_free(newname, strlen(newname) + 1); | |
5239 | ||
c242c188 | 5240 | (void) rw_exit(&ztest_name_lock); |
34dc7c2f BB |
5241 | } |
5242 | ||
428870ff BB |
5243 | /* |
5244 | * Verify pool integrity by running zdb. | |
5245 | */ | |
34dc7c2f | 5246 | static void |
428870ff | 5247 | ztest_run_zdb(char *pool) |
34dc7c2f BB |
5248 | { |
5249 | int status; | |
34dc7c2f | 5250 | char *bin; |
0e8d1b2d BB |
5251 | char *zdb; |
5252 | char *zbuf; | |
34dc7c2f BB |
5253 | FILE *fp; |
5254 | ||
0e8d1b2d BB |
5255 | bin = umem_alloc(MAXPATHLEN + MAXNAMELEN + 20, UMEM_NOFAIL); |
5256 | zdb = umem_alloc(MAXPATHLEN + MAXNAMELEN + 20, UMEM_NOFAIL); | |
5257 | zbuf = umem_alloc(1024, UMEM_NOFAIL); | |
34dc7c2f | 5258 | |
0e8d1b2d | 5259 | VERIFY(realpath(getexecname(), bin) != NULL); |
341b5f1d | 5260 | if (strncmp(bin, "/usr/sbin/ztest", 15) == 0) { |
0e8d1b2d | 5261 | strcpy(bin, "/usr/sbin/zdb"); /* Installed */ |
341b5f1d BB |
5262 | } else if (strncmp(bin, "/sbin/ztest", 11) == 0) { |
5263 | strcpy(bin, "/sbin/zdb"); /* Installed */ | |
0e8d1b2d BB |
5264 | } else { |
5265 | strstr(bin, "/ztest/")[0] = '\0'; /* In-tree */ | |
5266 | strcat(bin, "/zdb/zdb"); | |
5267 | } | |
5268 | ||
5269 | (void) sprintf(zdb, | |
5270 | "%s -bcc%s%s -U %s %s", | |
5271 | bin, | |
c242c188 CS |
5272 | ztest_opts.zo_verbose >= 3 ? "s" : "", |
5273 | ztest_opts.zo_verbose >= 4 ? "v" : "", | |
428870ff | 5274 | spa_config_path, |
b128c09f | 5275 | pool); |
34dc7c2f | 5276 | |
c242c188 | 5277 | if (ztest_opts.zo_verbose >= 5) |
34dc7c2f BB |
5278 | (void) printf("Executing %s\n", strstr(zdb, "zdb ")); |
5279 | ||
5280 | fp = popen(zdb, "r"); | |
5281 | ||
6aec1cd5 | 5282 | while (fgets(zbuf, 1024, fp) != NULL) |
c242c188 | 5283 | if (ztest_opts.zo_verbose >= 3) |
34dc7c2f BB |
5284 | (void) printf("%s", zbuf); |
5285 | ||
5286 | status = pclose(fp); | |
5287 | ||
5288 | if (status == 0) | |
0e8d1b2d | 5289 | goto out; |
34dc7c2f BB |
5290 | |
5291 | ztest_dump_core = 0; | |
5292 | if (WIFEXITED(status)) | |
5293 | fatal(0, "'%s' exit code %d", zdb, WEXITSTATUS(status)); | |
5294 | else | |
5295 | fatal(0, "'%s' died with signal %d", zdb, WTERMSIG(status)); | |
0e8d1b2d BB |
5296 | out: |
5297 | umem_free(bin, MAXPATHLEN + MAXNAMELEN + 20); | |
5298 | umem_free(zdb, MAXPATHLEN + MAXNAMELEN + 20); | |
5299 | umem_free(zbuf, 1024); | |
34dc7c2f BB |
5300 | } |
5301 | ||
5302 | static void | |
5303 | ztest_walk_pool_directory(char *header) | |
5304 | { | |
5305 | spa_t *spa = NULL; | |
5306 | ||
c242c188 | 5307 | if (ztest_opts.zo_verbose >= 6) |
34dc7c2f BB |
5308 | (void) printf("%s\n", header); |
5309 | ||
5310 | mutex_enter(&spa_namespace_lock); | |
5311 | while ((spa = spa_next(spa)) != NULL) | |
c242c188 | 5312 | if (ztest_opts.zo_verbose >= 6) |
34dc7c2f BB |
5313 | (void) printf("\t%s\n", spa_name(spa)); |
5314 | mutex_exit(&spa_namespace_lock); | |
5315 | } | |
5316 | ||
5317 | static void | |
5318 | ztest_spa_import_export(char *oldname, char *newname) | |
5319 | { | |
fb5f0bc8 | 5320 | nvlist_t *config, *newconfig; |
34dc7c2f BB |
5321 | uint64_t pool_guid; |
5322 | spa_t *spa; | |
13fe0198 | 5323 | int error; |
34dc7c2f | 5324 | |
c242c188 | 5325 | if (ztest_opts.zo_verbose >= 4) { |
34dc7c2f BB |
5326 | (void) printf("import/export: old = %s, new = %s\n", |
5327 | oldname, newname); | |
5328 | } | |
5329 | ||
5330 | /* | |
5331 | * Clean up from previous runs. | |
5332 | */ | |
5333 | (void) spa_destroy(newname); | |
5334 | ||
5335 | /* | |
5336 | * Get the pool's configuration and guid. | |
5337 | */ | |
428870ff | 5338 | VERIFY3U(0, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f | 5339 | |
fb5f0bc8 BB |
5340 | /* |
5341 | * Kick off a scrub to tickle scrub/export races. | |
5342 | */ | |
5343 | if (ztest_random(2) == 0) | |
428870ff | 5344 | (void) spa_scan(spa, POOL_SCAN_SCRUB); |
fb5f0bc8 | 5345 | |
34dc7c2f BB |
5346 | pool_guid = spa_guid(spa); |
5347 | spa_close(spa, FTAG); | |
5348 | ||
5349 | ztest_walk_pool_directory("pools before export"); | |
5350 | ||
5351 | /* | |
5352 | * Export it. | |
5353 | */ | |
428870ff | 5354 | VERIFY3U(0, ==, spa_export(oldname, &config, B_FALSE, B_FALSE)); |
34dc7c2f BB |
5355 | |
5356 | ztest_walk_pool_directory("pools after export"); | |
5357 | ||
fb5f0bc8 BB |
5358 | /* |
5359 | * Try to import it. | |
5360 | */ | |
5361 | newconfig = spa_tryimport(config); | |
5362 | ASSERT(newconfig != NULL); | |
5363 | nvlist_free(newconfig); | |
5364 | ||
34dc7c2f BB |
5365 | /* |
5366 | * Import it under the new name. | |
5367 | */ | |
13fe0198 MA |
5368 | error = spa_import(newname, config, NULL, 0); |
5369 | if (error != 0) { | |
5370 | dump_nvlist(config, 0); | |
5371 | fatal(B_FALSE, "couldn't import pool %s as %s: error %u", | |
5372 | oldname, newname, error); | |
5373 | } | |
34dc7c2f BB |
5374 | |
5375 | ztest_walk_pool_directory("pools after import"); | |
5376 | ||
5377 | /* | |
5378 | * Try to import it again -- should fail with EEXIST. | |
5379 | */ | |
572e2857 | 5380 | VERIFY3U(EEXIST, ==, spa_import(newname, config, NULL, 0)); |
34dc7c2f BB |
5381 | |
5382 | /* | |
5383 | * Try to import it under a different name -- should fail with EEXIST. | |
5384 | */ | |
572e2857 | 5385 | VERIFY3U(EEXIST, ==, spa_import(oldname, config, NULL, 0)); |
34dc7c2f BB |
5386 | |
5387 | /* | |
5388 | * Verify that the pool is no longer visible under the old name. | |
5389 | */ | |
428870ff | 5390 | VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f BB |
5391 | |
5392 | /* | |
5393 | * Verify that we can open and close the pool using the new name. | |
5394 | */ | |
428870ff | 5395 | VERIFY3U(0, ==, spa_open(newname, &spa, FTAG)); |
34dc7c2f BB |
5396 | ASSERT(pool_guid == spa_guid(spa)); |
5397 | spa_close(spa, FTAG); | |
5398 | ||
5399 | nvlist_free(config); | |
5400 | } | |
5401 | ||
fb5f0bc8 BB |
5402 | static void |
5403 | ztest_resume(spa_t *spa) | |
5404 | { | |
c242c188 | 5405 | if (spa_suspended(spa) && ztest_opts.zo_verbose >= 6) |
428870ff BB |
5406 | (void) printf("resuming from suspended state\n"); |
5407 | spa_vdev_state_enter(spa, SCL_NONE); | |
5408 | vdev_clear(spa, NULL); | |
5409 | (void) spa_vdev_state_exit(spa, NULL, 0); | |
5410 | (void) zio_resume(spa); | |
fb5f0bc8 BB |
5411 | } |
5412 | ||
34dc7c2f | 5413 | static void * |
fb5f0bc8 | 5414 | ztest_resume_thread(void *arg) |
34dc7c2f | 5415 | { |
b128c09f | 5416 | spa_t *spa = arg; |
34dc7c2f BB |
5417 | |
5418 | while (!ztest_exiting) { | |
428870ff BB |
5419 | if (spa_suspended(spa)) |
5420 | ztest_resume(spa); | |
5421 | (void) poll(NULL, 0, 100); | |
34dc7c2f | 5422 | } |
34dc7c2f | 5423 | |
1e33ac1e | 5424 | thread_exit(); |
428870ff | 5425 | |
1e33ac1e BB |
5426 | return (NULL); |
5427 | } | |
428870ff | 5428 | |
1e33ac1e | 5429 | #define GRACE 300 |
428870ff | 5430 | |
26099167 | 5431 | #if 0 |
1e33ac1e BB |
5432 | static void |
5433 | ztest_deadman_alarm(int sig) | |
5434 | { | |
5435 | fatal(0, "failed to complete within %d seconds of deadline", GRACE); | |
428870ff | 5436 | } |
26099167 | 5437 | #endif |
428870ff BB |
5438 | |
5439 | static void | |
c242c188 | 5440 | ztest_execute(int test, ztest_info_t *zi, uint64_t id) |
428870ff | 5441 | { |
c242c188 CS |
5442 | ztest_ds_t *zd = &ztest_ds[id % ztest_opts.zo_datasets]; |
5443 | ztest_shared_callstate_t *zc = ZTEST_GET_SHARED_CALLSTATE(test); | |
428870ff | 5444 | hrtime_t functime = gethrtime(); |
d6320ddb | 5445 | int i; |
428870ff | 5446 | |
d6320ddb | 5447 | for (i = 0; i < zi->zi_iters; i++) |
428870ff BB |
5448 | zi->zi_func(zd, id); |
5449 | ||
5450 | functime = gethrtime() - functime; | |
5451 | ||
c242c188 CS |
5452 | atomic_add_64(&zc->zc_count, 1); |
5453 | atomic_add_64(&zc->zc_time, functime); | |
428870ff | 5454 | |
c242c188 | 5455 | if (ztest_opts.zo_verbose >= 4) { |
428870ff BB |
5456 | Dl_info dli; |
5457 | (void) dladdr((void *)zi->zi_func, &dli); | |
5458 | (void) printf("%6.2f sec in %s\n", | |
5459 | (double)functime / NANOSEC, dli.dli_sname); | |
5460 | } | |
5461 | } | |
5462 | ||
34dc7c2f BB |
5463 | static void * |
5464 | ztest_thread(void *arg) | |
5465 | { | |
c242c188 | 5466 | int rand; |
428870ff | 5467 | uint64_t id = (uintptr_t)arg; |
34dc7c2f | 5468 | ztest_shared_t *zs = ztest_shared; |
428870ff BB |
5469 | uint64_t call_next; |
5470 | hrtime_t now; | |
34dc7c2f | 5471 | ztest_info_t *zi; |
c242c188 | 5472 | ztest_shared_callstate_t *zc; |
34dc7c2f | 5473 | |
428870ff | 5474 | while ((now = gethrtime()) < zs->zs_thread_stop) { |
34dc7c2f BB |
5475 | /* |
5476 | * See if it's time to force a crash. | |
5477 | */ | |
428870ff BB |
5478 | if (now > zs->zs_thread_kill) |
5479 | ztest_kill(zs); | |
34dc7c2f BB |
5480 | |
5481 | /* | |
428870ff | 5482 | * If we're getting ENOSPC with some regularity, stop. |
34dc7c2f | 5483 | */ |
428870ff BB |
5484 | if (zs->zs_enospc_count > 10) |
5485 | break; | |
34dc7c2f BB |
5486 | |
5487 | /* | |
428870ff | 5488 | * Pick a random function to execute. |
34dc7c2f | 5489 | */ |
c242c188 CS |
5490 | rand = ztest_random(ZTEST_FUNCS); |
5491 | zi = &ztest_info[rand]; | |
5492 | zc = ZTEST_GET_SHARED_CALLSTATE(rand); | |
5493 | call_next = zc->zc_next; | |
34dc7c2f | 5494 | |
428870ff | 5495 | if (now >= call_next && |
c242c188 CS |
5496 | atomic_cas_64(&zc->zc_next, call_next, call_next + |
5497 | ztest_random(2 * zi->zi_interval[0] + 1)) == call_next) { | |
5498 | ztest_execute(rand, zi, id); | |
5499 | } | |
428870ff | 5500 | } |
34dc7c2f | 5501 | |
1e33ac1e BB |
5502 | thread_exit(); |
5503 | ||
428870ff BB |
5504 | return (NULL); |
5505 | } | |
34dc7c2f | 5506 | |
428870ff BB |
5507 | static void |
5508 | ztest_dataset_name(char *dsname, char *pool, int d) | |
5509 | { | |
5510 | (void) snprintf(dsname, MAXNAMELEN, "%s/ds_%d", pool, d); | |
5511 | } | |
34dc7c2f | 5512 | |
428870ff | 5513 | static void |
c242c188 | 5514 | ztest_dataset_destroy(int d) |
428870ff BB |
5515 | { |
5516 | char name[MAXNAMELEN]; | |
d6320ddb | 5517 | int t; |
34dc7c2f | 5518 | |
c242c188 | 5519 | ztest_dataset_name(name, ztest_opts.zo_pool, d); |
34dc7c2f | 5520 | |
c242c188 | 5521 | if (ztest_opts.zo_verbose >= 3) |
428870ff | 5522 | (void) printf("Destroying %s to free up space\n", name); |
34dc7c2f | 5523 | |
428870ff BB |
5524 | /* |
5525 | * Cleanup any non-standard clones and snapshots. In general, | |
5526 | * ztest thread t operates on dataset (t % zopt_datasets), | |
5527 | * so there may be more than one thing to clean up. | |
5528 | */ | |
c242c188 CS |
5529 | for (t = d; t < ztest_opts.zo_threads; |
5530 | t += ztest_opts.zo_datasets) | |
428870ff BB |
5531 | ztest_dsl_dataset_cleanup(name, t); |
5532 | ||
5533 | (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL, | |
5534 | DS_FIND_SNAPSHOTS | DS_FIND_CHILDREN); | |
5535 | } | |
5536 | ||
5537 | static void | |
5538 | ztest_dataset_dirobj_verify(ztest_ds_t *zd) | |
5539 | { | |
5540 | uint64_t usedobjs, dirobjs, scratch; | |
5541 | ||
5542 | /* | |
5543 | * ZTEST_DIROBJ is the object directory for the entire dataset. | |
5544 | * Therefore, the number of objects in use should equal the | |
5545 | * number of ZTEST_DIROBJ entries, +1 for ZTEST_DIROBJ itself. | |
5546 | * If not, we have an object leak. | |
5547 | * | |
5548 | * Note that we can only check this in ztest_dataset_open(), | |
5549 | * when the open-context and syncing-context values agree. | |
5550 | * That's because zap_count() returns the open-context value, | |
5551 | * while dmu_objset_space() returns the rootbp fill count. | |
5552 | */ | |
5553 | VERIFY3U(0, ==, zap_count(zd->zd_os, ZTEST_DIROBJ, &dirobjs)); | |
5554 | dmu_objset_space(zd->zd_os, &scratch, &scratch, &usedobjs, &scratch); | |
5555 | ASSERT3U(dirobjs + 1, ==, usedobjs); | |
5556 | } | |
5557 | ||
5558 | static int | |
c242c188 | 5559 | ztest_dataset_open(int d) |
428870ff | 5560 | { |
c242c188 CS |
5561 | ztest_ds_t *zd = &ztest_ds[d]; |
5562 | uint64_t committed_seq = ZTEST_GET_SHARED_DS(d)->zd_seq; | |
428870ff BB |
5563 | objset_t *os; |
5564 | zilog_t *zilog; | |
5565 | char name[MAXNAMELEN]; | |
5566 | int error; | |
5567 | ||
c242c188 | 5568 | ztest_dataset_name(name, ztest_opts.zo_pool, d); |
428870ff | 5569 | |
c242c188 | 5570 | (void) rw_enter(&ztest_name_lock, RW_READER); |
428870ff BB |
5571 | |
5572 | error = ztest_dataset_create(name); | |
5573 | if (error == ENOSPC) { | |
c242c188 | 5574 | (void) rw_exit(&ztest_name_lock); |
428870ff BB |
5575 | ztest_record_enospc(FTAG); |
5576 | return (error); | |
34dc7c2f | 5577 | } |
428870ff | 5578 | ASSERT(error == 0 || error == EEXIST); |
34dc7c2f | 5579 | |
13fe0198 | 5580 | VERIFY0(dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, zd, &os)); |
c242c188 | 5581 | (void) rw_exit(&ztest_name_lock); |
428870ff | 5582 | |
c242c188 | 5583 | ztest_zd_init(zd, ZTEST_GET_SHARED_DS(d), os); |
428870ff BB |
5584 | |
5585 | zilog = zd->zd_zilog; | |
5586 | ||
5587 | if (zilog->zl_header->zh_claim_lr_seq != 0 && | |
5588 | zilog->zl_header->zh_claim_lr_seq < committed_seq) | |
5589 | fatal(0, "missing log records: claimed %llu < committed %llu", | |
5590 | zilog->zl_header->zh_claim_lr_seq, committed_seq); | |
5591 | ||
5592 | ztest_dataset_dirobj_verify(zd); | |
5593 | ||
5594 | zil_replay(os, zd, ztest_replay_vector); | |
5595 | ||
5596 | ztest_dataset_dirobj_verify(zd); | |
5597 | ||
c242c188 | 5598 | if (ztest_opts.zo_verbose >= 6) |
428870ff BB |
5599 | (void) printf("%s replay %llu blocks, %llu records, seq %llu\n", |
5600 | zd->zd_name, | |
5601 | (u_longlong_t)zilog->zl_parse_blk_count, | |
5602 | (u_longlong_t)zilog->zl_parse_lr_count, | |
5603 | (u_longlong_t)zilog->zl_replaying_seq); | |
5604 | ||
5605 | zilog = zil_open(os, ztest_get_data); | |
5606 | ||
5607 | if (zilog->zl_replaying_seq != 0 && | |
5608 | zilog->zl_replaying_seq < committed_seq) | |
5609 | fatal(0, "missing log records: replayed %llu < committed %llu", | |
5610 | zilog->zl_replaying_seq, committed_seq); | |
5611 | ||
5612 | return (0); | |
5613 | } | |
5614 | ||
5615 | static void | |
c242c188 | 5616 | ztest_dataset_close(int d) |
428870ff | 5617 | { |
c242c188 | 5618 | ztest_ds_t *zd = &ztest_ds[d]; |
428870ff BB |
5619 | |
5620 | zil_close(zd->zd_zilog); | |
13fe0198 | 5621 | dmu_objset_disown(zd->zd_os, zd); |
428870ff BB |
5622 | |
5623 | ztest_zd_fini(zd); | |
34dc7c2f BB |
5624 | } |
5625 | ||
5626 | /* | |
5627 | * Kick off threads to run tests on all datasets in parallel. | |
5628 | */ | |
5629 | static void | |
428870ff | 5630 | ztest_run(ztest_shared_t *zs) |
34dc7c2f | 5631 | { |
1e33ac1e | 5632 | kt_did_t *tid; |
34dc7c2f | 5633 | spa_t *spa; |
3541dc6d | 5634 | objset_t *os; |
1e33ac1e BB |
5635 | kthread_t *resume_thread; |
5636 | uint64_t object; | |
428870ff | 5637 | int error; |
d6320ddb | 5638 | int t, d; |
b128c09f BB |
5639 | |
5640 | ztest_exiting = B_FALSE; | |
34dc7c2f | 5641 | |
34dc7c2f | 5642 | /* |
428870ff | 5643 | * Initialize parent/child shared state. |
34dc7c2f | 5644 | */ |
c242c188 CS |
5645 | mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); |
5646 | rw_init(&ztest_name_lock, NULL, RW_DEFAULT, NULL); | |
34dc7c2f | 5647 | |
428870ff | 5648 | zs->zs_thread_start = gethrtime(); |
c242c188 CS |
5649 | zs->zs_thread_stop = |
5650 | zs->zs_thread_start + ztest_opts.zo_passtime * NANOSEC; | |
428870ff BB |
5651 | zs->zs_thread_stop = MIN(zs->zs_thread_stop, zs->zs_proc_stop); |
5652 | zs->zs_thread_kill = zs->zs_thread_stop; | |
c242c188 CS |
5653 | if (ztest_random(100) < ztest_opts.zo_killrate) { |
5654 | zs->zs_thread_kill -= | |
5655 | ztest_random(ztest_opts.zo_passtime * NANOSEC); | |
5656 | } | |
34dc7c2f | 5657 | |
1e33ac1e | 5658 | mutex_init(&zcl.zcl_callbacks_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 5659 | |
428870ff BB |
5660 | list_create(&zcl.zcl_callbacks, sizeof (ztest_cb_data_t), |
5661 | offsetof(ztest_cb_data_t, zcd_node)); | |
34dc7c2f BB |
5662 | |
5663 | /* | |
b128c09f | 5664 | * Open our pool. |
34dc7c2f | 5665 | */ |
428870ff | 5666 | kernel_init(FREAD | FWRITE); |
13fe0198 | 5667 | VERIFY0(spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
6d974228 | 5668 | spa->spa_debug = B_TRUE; |
c242c188 | 5669 | ztest_spa = spa; |
428870ff | 5670 | |
13fe0198 MA |
5671 | VERIFY0(dmu_objset_own(ztest_opts.zo_pool, |
5672 | DMU_OST_ANY, B_TRUE, FTAG, &os)); | |
3541dc6d | 5673 | zs->zs_guid = dmu_objset_fsid_guid(os); |
13fe0198 | 5674 | dmu_objset_disown(os, FTAG); |
3541dc6d | 5675 | |
428870ff | 5676 | spa->spa_dedup_ditto = 2 * ZIO_DEDUPDITTO_MIN; |
34dc7c2f | 5677 | |
fb5f0bc8 BB |
5678 | /* |
5679 | * We don't expect the pool to suspend unless maxfaults == 0, | |
5680 | * in which case ztest_fault_inject() temporarily takes away | |
5681 | * the only valid replica. | |
5682 | */ | |
428870ff | 5683 | if (MAXFAULTS() == 0) |
fb5f0bc8 BB |
5684 | spa->spa_failmode = ZIO_FAILURE_MODE_WAIT; |
5685 | else | |
5686 | spa->spa_failmode = ZIO_FAILURE_MODE_PANIC; | |
5687 | ||
34dc7c2f | 5688 | /* |
b128c09f | 5689 | * Create a thread to periodically resume suspended I/O. |
34dc7c2f | 5690 | */ |
0aebd4f9 ED |
5691 | VERIFY3P((resume_thread = zk_thread_create(NULL, 0, |
5692 | (thread_func_t)ztest_resume_thread, spa, TS_RUN, NULL, 0, 0, | |
5693 | PTHREAD_CREATE_JOINABLE)), !=, NULL); | |
34dc7c2f | 5694 | |
26099167 | 5695 | #if 0 |
428870ff | 5696 | /* |
1e33ac1e | 5697 | * Set a deadman alarm to abort() if we hang. |
428870ff | 5698 | */ |
1e33ac1e BB |
5699 | signal(SIGALRM, ztest_deadman_alarm); |
5700 | alarm((zs->zs_thread_stop - zs->zs_thread_start) / NANOSEC + GRACE); | |
26099167 | 5701 | #endif |
428870ff | 5702 | |
34dc7c2f BB |
5703 | /* |
5704 | * Verify that we can safely inquire about about any object, | |
5705 | * whether it's allocated or not. To make it interesting, | |
5706 | * we probe a 5-wide window around each power of two. | |
5707 | * This hits all edge cases, including zero and the max. | |
5708 | */ | |
d6320ddb BB |
5709 | for (t = 0; t < 64; t++) { |
5710 | for (d = -5; d <= 5; d++) { | |
34dc7c2f BB |
5711 | error = dmu_object_info(spa->spa_meta_objset, |
5712 | (1ULL << t) + d, NULL); | |
5713 | ASSERT(error == 0 || error == ENOENT || | |
5714 | error == EINVAL); | |
5715 | } | |
5716 | } | |
5717 | ||
5718 | /* | |
428870ff | 5719 | * If we got any ENOSPC errors on the previous run, destroy something. |
34dc7c2f | 5720 | */ |
428870ff | 5721 | if (zs->zs_enospc_count != 0) { |
c242c188 CS |
5722 | int d = ztest_random(ztest_opts.zo_datasets); |
5723 | ztest_dataset_destroy(d); | |
428870ff | 5724 | } |
34dc7c2f BB |
5725 | zs->zs_enospc_count = 0; |
5726 | ||
c242c188 CS |
5727 | tid = umem_zalloc(ztest_opts.zo_threads * sizeof (kt_did_t), |
5728 | UMEM_NOFAIL); | |
34dc7c2f | 5729 | |
c242c188 | 5730 | if (ztest_opts.zo_verbose >= 4) |
34dc7c2f BB |
5731 | (void) printf("starting main threads...\n"); |
5732 | ||
428870ff BB |
5733 | /* |
5734 | * Kick off all the tests that run in parallel. | |
5735 | */ | |
c242c188 | 5736 | for (t = 0; t < ztest_opts.zo_threads; t++) { |
1e33ac1e BB |
5737 | kthread_t *thread; |
5738 | ||
c242c188 CS |
5739 | if (t < ztest_opts.zo_datasets && |
5740 | ztest_dataset_open(t) != 0) | |
428870ff | 5741 | return; |
1e33ac1e | 5742 | |
0aebd4f9 ED |
5743 | VERIFY3P(thread = zk_thread_create(NULL, 0, |
5744 | (thread_func_t)ztest_thread, | |
5745 | (void *)(uintptr_t)t, TS_RUN, NULL, 0, 0, | |
5746 | PTHREAD_CREATE_JOINABLE), !=, NULL); | |
1e33ac1e | 5747 | tid[t] = thread->t_tid; |
34dc7c2f BB |
5748 | } |
5749 | ||
428870ff BB |
5750 | /* |
5751 | * Wait for all of the tests to complete. We go in reverse order | |
5752 | * so we don't close datasets while threads are still using them. | |
5753 | */ | |
c242c188 | 5754 | for (t = ztest_opts.zo_threads - 1; t >= 0; t--) { |
1e33ac1e | 5755 | thread_join(tid[t]); |
c242c188 CS |
5756 | if (t < ztest_opts.zo_datasets) |
5757 | ztest_dataset_close(t); | |
34dc7c2f BB |
5758 | } |
5759 | ||
34dc7c2f BB |
5760 | txg_wait_synced(spa_get_dsl(spa), 0); |
5761 | ||
428870ff BB |
5762 | zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(spa)); |
5763 | zs->zs_space = metaslab_class_get_space(spa_normal_class(spa)); | |
5764 | ||
c242c188 | 5765 | umem_free(tid, ztest_opts.zo_threads * sizeof (kt_did_t)); |
428870ff BB |
5766 | |
5767 | /* Kill the resume thread */ | |
5768 | ztest_exiting = B_TRUE; | |
1e33ac1e | 5769 | thread_join(resume_thread->t_tid); |
428870ff | 5770 | ztest_resume(spa); |
34dc7c2f BB |
5771 | |
5772 | /* | |
428870ff BB |
5773 | * Right before closing the pool, kick off a bunch of async I/O; |
5774 | * spa_close() should wait for it to complete. | |
34dc7c2f | 5775 | */ |
1e33ac1e | 5776 | for (object = 1; object < 50; object++) |
428870ff BB |
5777 | dmu_prefetch(spa->spa_meta_objset, object, 0, 1ULL << 20); |
5778 | ||
090ff092 RC |
5779 | /* Verify that at least one commit cb was called in a timely fashion */ |
5780 | if (zc_cb_counter >= ZTEST_COMMIT_CB_MIN_REG) | |
c99c9001 | 5781 | VERIFY0(zc_min_txg_delay); |
090ff092 | 5782 | |
428870ff BB |
5783 | spa_close(spa, FTAG); |
5784 | ||
5785 | /* | |
5786 | * Verify that we can loop over all pools. | |
5787 | */ | |
5788 | mutex_enter(&spa_namespace_lock); | |
5789 | for (spa = spa_next(NULL); spa != NULL; spa = spa_next(spa)) | |
c242c188 | 5790 | if (ztest_opts.zo_verbose > 3) |
428870ff BB |
5791 | (void) printf("spa_next: found %s\n", spa_name(spa)); |
5792 | mutex_exit(&spa_namespace_lock); | |
5793 | ||
5794 | /* | |
5795 | * Verify that we can export the pool and reimport it under a | |
5796 | * different name. | |
5797 | */ | |
5798 | if (ztest_random(2) == 0) { | |
5799 | char name[MAXNAMELEN]; | |
c242c188 CS |
5800 | (void) snprintf(name, MAXNAMELEN, "%s_import", |
5801 | ztest_opts.zo_pool); | |
5802 | ztest_spa_import_export(ztest_opts.zo_pool, name); | |
5803 | ztest_spa_import_export(name, ztest_opts.zo_pool); | |
428870ff BB |
5804 | } |
5805 | ||
5806 | kernel_fini(); | |
572e2857 BB |
5807 | |
5808 | list_destroy(&zcl.zcl_callbacks); | |
0e8d1b2d | 5809 | mutex_destroy(&zcl.zcl_callbacks_lock); |
c242c188 CS |
5810 | rw_destroy(&ztest_name_lock); |
5811 | mutex_destroy(&ztest_vdev_lock); | |
428870ff BB |
5812 | } |
5813 | ||
5814 | static void | |
c242c188 | 5815 | ztest_freeze(void) |
428870ff | 5816 | { |
c242c188 | 5817 | ztest_ds_t *zd = &ztest_ds[0]; |
428870ff BB |
5818 | spa_t *spa; |
5819 | int numloops = 0; | |
5820 | ||
c242c188 | 5821 | if (ztest_opts.zo_verbose >= 3) |
428870ff | 5822 | (void) printf("testing spa_freeze()...\n"); |
9babb374 | 5823 | |
428870ff | 5824 | kernel_init(FREAD | FWRITE); |
c242c188 CS |
5825 | VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
5826 | VERIFY3U(0, ==, ztest_dataset_open(0)); | |
9babb374 | 5827 | |
428870ff BB |
5828 | /* |
5829 | * Force the first log block to be transactionally allocated. | |
5830 | * We have to do this before we freeze the pool -- otherwise | |
5831 | * the log chain won't be anchored. | |
5832 | */ | |
5833 | while (BP_IS_HOLE(&zd->zd_zilog->zl_header->zh_log)) { | |
5834 | ztest_dmu_object_alloc_free(zd, 0); | |
572e2857 | 5835 | zil_commit(zd->zd_zilog, 0); |
34dc7c2f BB |
5836 | } |
5837 | ||
5838 | txg_wait_synced(spa_get_dsl(spa), 0); | |
5839 | ||
428870ff BB |
5840 | /* |
5841 | * Freeze the pool. This stops spa_sync() from doing anything, | |
5842 | * so that the only way to record changes from now on is the ZIL. | |
5843 | */ | |
5844 | spa_freeze(spa); | |
b128c09f | 5845 | |
428870ff BB |
5846 | /* |
5847 | * Run tests that generate log records but don't alter the pool config | |
5848 | * or depend on DSL sync tasks (snapshots, objset create/destroy, etc). | |
5849 | * We do a txg_wait_synced() after each iteration to force the txg | |
5850 | * to increase well beyond the last synced value in the uberblock. | |
5851 | * The ZIL should be OK with that. | |
5852 | */ | |
c242c188 CS |
5853 | while (ztest_random(10) != 0 && |
5854 | numloops++ < ztest_opts.zo_maxloops) { | |
428870ff BB |
5855 | ztest_dmu_write_parallel(zd, 0); |
5856 | ztest_dmu_object_alloc_free(zd, 0); | |
5857 | txg_wait_synced(spa_get_dsl(spa), 0); | |
5858 | } | |
b128c09f | 5859 | |
34dc7c2f | 5860 | /* |
428870ff | 5861 | * Commit all of the changes we just generated. |
34dc7c2f | 5862 | */ |
572e2857 | 5863 | zil_commit(zd->zd_zilog, 0); |
428870ff | 5864 | txg_wait_synced(spa_get_dsl(spa), 0); |
34dc7c2f | 5865 | |
428870ff BB |
5866 | /* |
5867 | * Close our dataset and close the pool. | |
5868 | */ | |
c242c188 | 5869 | ztest_dataset_close(0); |
34dc7c2f | 5870 | spa_close(spa, FTAG); |
428870ff | 5871 | kernel_fini(); |
34dc7c2f | 5872 | |
428870ff BB |
5873 | /* |
5874 | * Open and close the pool and dataset to induce log replay. | |
5875 | */ | |
5876 | kernel_init(FREAD | FWRITE); | |
c242c188 | 5877 | VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
29809a6c | 5878 | ASSERT(spa_freeze_txg(spa) == UINT64_MAX); |
c242c188 CS |
5879 | VERIFY3U(0, ==, ztest_dataset_open(0)); |
5880 | ztest_dataset_close(0); | |
3bc7e0fb GW |
5881 | |
5882 | spa->spa_debug = B_TRUE; | |
5883 | ztest_spa = spa; | |
5884 | txg_wait_synced(spa_get_dsl(spa), 0); | |
5885 | ztest_reguid(NULL, 0); | |
5886 | ||
428870ff | 5887 | spa_close(spa, FTAG); |
34dc7c2f BB |
5888 | kernel_fini(); |
5889 | } | |
5890 | ||
5891 | void | |
5892 | print_time(hrtime_t t, char *timebuf) | |
5893 | { | |
5894 | hrtime_t s = t / NANOSEC; | |
5895 | hrtime_t m = s / 60; | |
5896 | hrtime_t h = m / 60; | |
5897 | hrtime_t d = h / 24; | |
5898 | ||
5899 | s -= m * 60; | |
5900 | m -= h * 60; | |
5901 | h -= d * 24; | |
5902 | ||
5903 | timebuf[0] = '\0'; | |
5904 | ||
5905 | if (d) | |
5906 | (void) sprintf(timebuf, | |
5907 | "%llud%02lluh%02llum%02llus", d, h, m, s); | |
5908 | else if (h) | |
5909 | (void) sprintf(timebuf, "%lluh%02llum%02llus", h, m, s); | |
5910 | else if (m) | |
5911 | (void) sprintf(timebuf, "%llum%02llus", m, s); | |
5912 | else | |
5913 | (void) sprintf(timebuf, "%llus", s); | |
5914 | } | |
5915 | ||
428870ff | 5916 | static nvlist_t * |
0bc8fd78 | 5917 | make_random_props(void) |
428870ff BB |
5918 | { |
5919 | nvlist_t *props; | |
5920 | ||
428870ff | 5921 | VERIFY(nvlist_alloc(&props, NV_UNIQUE_NAME, 0) == 0); |
9ae529ec CS |
5922 | if (ztest_random(2) == 0) |
5923 | return (props); | |
428870ff BB |
5924 | VERIFY(nvlist_add_uint64(props, "autoreplace", 1) == 0); |
5925 | ||
428870ff BB |
5926 | return (props); |
5927 | } | |
5928 | ||
34dc7c2f BB |
5929 | /* |
5930 | * Create a storage pool with the given name and initial vdev size. | |
428870ff | 5931 | * Then test spa_freeze() functionality. |
34dc7c2f BB |
5932 | */ |
5933 | static void | |
428870ff | 5934 | ztest_init(ztest_shared_t *zs) |
34dc7c2f BB |
5935 | { |
5936 | spa_t *spa; | |
428870ff | 5937 | nvlist_t *nvroot, *props; |
9ae529ec | 5938 | int i; |
428870ff | 5939 | |
c242c188 CS |
5940 | mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); |
5941 | rw_init(&ztest_name_lock, NULL, RW_DEFAULT, NULL); | |
34dc7c2f BB |
5942 | |
5943 | kernel_init(FREAD | FWRITE); | |
5944 | ||
5945 | /* | |
5946 | * Create the storage pool. | |
5947 | */ | |
c242c188 | 5948 | (void) spa_destroy(ztest_opts.zo_pool); |
428870ff BB |
5949 | ztest_shared->zs_vdev_next_leaf = 0; |
5950 | zs->zs_splits = 0; | |
c242c188 | 5951 | zs->zs_mirrors = ztest_opts.zo_mirrors; |
ea0b2538 | 5952 | nvroot = make_vdev_root(NULL, NULL, NULL, ztest_opts.zo_vdev_size, 0, |
c242c188 | 5953 | 0, ztest_opts.zo_raidz, zs->zs_mirrors, 1); |
428870ff | 5954 | props = make_random_props(); |
9ae529ec CS |
5955 | for (i = 0; i < SPA_FEATURES; i++) { |
5956 | char *buf; | |
5957 | VERIFY3S(-1, !=, asprintf(&buf, "feature@%s", | |
5958 | spa_feature_table[i].fi_uname)); | |
5959 | VERIFY3U(0, ==, nvlist_add_uint64(props, buf, 0)); | |
5960 | free(buf); | |
5961 | } | |
6f1ffb06 | 5962 | VERIFY3U(0, ==, spa_create(ztest_opts.zo_pool, nvroot, props, NULL)); |
34dc7c2f BB |
5963 | nvlist_free(nvroot); |
5964 | ||
c242c188 CS |
5965 | VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
5966 | zs->zs_metaslab_sz = | |
5967 | 1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift; | |
34dc7c2f BB |
5968 | spa_close(spa, FTAG); |
5969 | ||
5970 | kernel_fini(); | |
428870ff | 5971 | |
c242c188 CS |
5972 | ztest_run_zdb(ztest_opts.zo_pool); |
5973 | ||
5974 | ztest_freeze(); | |
5975 | ||
5976 | ztest_run_zdb(ztest_opts.zo_pool); | |
5977 | ||
5978 | rw_destroy(&ztest_name_lock); | |
5979 | mutex_destroy(&ztest_vdev_lock); | |
5980 | } | |
5981 | ||
5982 | static void | |
9d81146b | 5983 | setup_data_fd(void) |
c242c188 | 5984 | { |
facbbe43 BB |
5985 | static char ztest_name_data[] = "/tmp/ztest.data.XXXXXX"; |
5986 | ||
5987 | ztest_fd_data = mkstemp(ztest_name_data); | |
9d81146b | 5988 | ASSERT3S(ztest_fd_data, >=, 0); |
facbbe43 | 5989 | (void) unlink(ztest_name_data); |
c242c188 CS |
5990 | } |
5991 | ||
22257dc0 CS |
5992 | static int |
5993 | shared_data_size(ztest_shared_hdr_t *hdr) | |
5994 | { | |
5995 | int size; | |
5996 | ||
5997 | size = hdr->zh_hdr_size; | |
5998 | size += hdr->zh_opts_size; | |
5999 | size += hdr->zh_size; | |
6000 | size += hdr->zh_stats_size * hdr->zh_stats_count; | |
6001 | size += hdr->zh_ds_size * hdr->zh_ds_count; | |
6002 | ||
6003 | return (size); | |
6004 | } | |
6005 | ||
c242c188 CS |
6006 | static void |
6007 | setup_hdr(void) | |
6008 | { | |
22257dc0 | 6009 | int size; |
c242c188 CS |
6010 | ztest_shared_hdr_t *hdr; |
6011 | ||
6012 | hdr = (void *)mmap(0, P2ROUNDUP(sizeof (*hdr), getpagesize()), | |
9d81146b | 6013 | PROT_READ | PROT_WRITE, MAP_SHARED, ztest_fd_data, 0); |
c242c188 CS |
6014 | ASSERT(hdr != MAP_FAILED); |
6015 | ||
9d81146b | 6016 | VERIFY3U(0, ==, ftruncate(ztest_fd_data, sizeof (ztest_shared_hdr_t))); |
22257dc0 | 6017 | |
c242c188 CS |
6018 | hdr->zh_hdr_size = sizeof (ztest_shared_hdr_t); |
6019 | hdr->zh_opts_size = sizeof (ztest_shared_opts_t); | |
6020 | hdr->zh_size = sizeof (ztest_shared_t); | |
6021 | hdr->zh_stats_size = sizeof (ztest_shared_callstate_t); | |
6022 | hdr->zh_stats_count = ZTEST_FUNCS; | |
6023 | hdr->zh_ds_size = sizeof (ztest_shared_ds_t); | |
6024 | hdr->zh_ds_count = ztest_opts.zo_datasets; | |
6025 | ||
22257dc0 | 6026 | size = shared_data_size(hdr); |
9d81146b | 6027 | VERIFY3U(0, ==, ftruncate(ztest_fd_data, size)); |
22257dc0 | 6028 | |
c242c188 CS |
6029 | (void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize())); |
6030 | } | |
6031 | ||
6032 | static void | |
6033 | setup_data(void) | |
6034 | { | |
6035 | int size, offset; | |
6036 | ztest_shared_hdr_t *hdr; | |
6037 | uint8_t *buf; | |
6038 | ||
6039 | hdr = (void *)mmap(0, P2ROUNDUP(sizeof (*hdr), getpagesize()), | |
9d81146b | 6040 | PROT_READ, MAP_SHARED, ztest_fd_data, 0); |
c242c188 CS |
6041 | ASSERT(hdr != MAP_FAILED); |
6042 | ||
22257dc0 | 6043 | size = shared_data_size(hdr); |
c242c188 CS |
6044 | |
6045 | (void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize())); | |
6046 | hdr = ztest_shared_hdr = (void *)mmap(0, P2ROUNDUP(size, getpagesize()), | |
9d81146b | 6047 | PROT_READ | PROT_WRITE, MAP_SHARED, ztest_fd_data, 0); |
c242c188 CS |
6048 | ASSERT(hdr != MAP_FAILED); |
6049 | buf = (uint8_t *)hdr; | |
6050 | ||
6051 | offset = hdr->zh_hdr_size; | |
6052 | ztest_shared_opts = (void *)&buf[offset]; | |
6053 | offset += hdr->zh_opts_size; | |
6054 | ztest_shared = (void *)&buf[offset]; | |
6055 | offset += hdr->zh_size; | |
6056 | ztest_shared_callstate = (void *)&buf[offset]; | |
6057 | offset += hdr->zh_stats_size * hdr->zh_stats_count; | |
6058 | ztest_shared_ds = (void *)&buf[offset]; | |
6059 | } | |
6060 | ||
6061 | static boolean_t | |
6062 | exec_child(char *cmd, char *libpath, boolean_t ignorekill, int *statusp) | |
6063 | { | |
6064 | pid_t pid; | |
6065 | int status; | |
483106eb | 6066 | char *cmdbuf = NULL; |
c242c188 CS |
6067 | |
6068 | pid = fork(); | |
6069 | ||
6070 | if (cmd == NULL) { | |
483106eb BB |
6071 | cmdbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
6072 | (void) strlcpy(cmdbuf, getexecname(), MAXPATHLEN); | |
c242c188 CS |
6073 | cmd = cmdbuf; |
6074 | } | |
6075 | ||
6076 | if (pid == -1) | |
6077 | fatal(1, "fork failed"); | |
6078 | ||
6079 | if (pid == 0) { /* child */ | |
6080 | char *emptyargv[2] = { cmd, NULL }; | |
9d81146b | 6081 | char fd_data_str[12]; |
c242c188 CS |
6082 | |
6083 | struct rlimit rl = { 1024, 1024 }; | |
6084 | (void) setrlimit(RLIMIT_NOFILE, &rl); | |
9d81146b ED |
6085 | |
6086 | (void) close(ztest_fd_rand); | |
6087 | VERIFY(11 >= snprintf(fd_data_str, 12, "%d", ztest_fd_data)); | |
6088 | VERIFY(0 == setenv("ZTEST_FD_DATA", fd_data_str, 1)); | |
6089 | ||
c242c188 CS |
6090 | (void) enable_extended_FILE_stdio(-1, -1); |
6091 | if (libpath != NULL) | |
6092 | VERIFY(0 == setenv("LD_LIBRARY_PATH", libpath, 1)); | |
6093 | (void) execv(cmd, emptyargv); | |
6094 | ztest_dump_core = B_FALSE; | |
6095 | fatal(B_TRUE, "exec failed: %s", cmd); | |
6096 | } | |
6097 | ||
483106eb BB |
6098 | if (cmdbuf != NULL) { |
6099 | umem_free(cmdbuf, MAXPATHLEN); | |
6100 | cmd = NULL; | |
6101 | } | |
6102 | ||
c242c188 CS |
6103 | while (waitpid(pid, &status, 0) != pid) |
6104 | continue; | |
6105 | if (statusp != NULL) | |
6106 | *statusp = status; | |
6107 | ||
6108 | if (WIFEXITED(status)) { | |
6109 | if (WEXITSTATUS(status) != 0) { | |
6110 | (void) fprintf(stderr, "child exited with code %d\n", | |
6111 | WEXITSTATUS(status)); | |
6112 | exit(2); | |
6113 | } | |
6114 | return (B_FALSE); | |
6115 | } else if (WIFSIGNALED(status)) { | |
6116 | if (!ignorekill || WTERMSIG(status) != SIGKILL) { | |
6117 | (void) fprintf(stderr, "child died with signal %d\n", | |
6118 | WTERMSIG(status)); | |
6119 | exit(3); | |
6120 | } | |
6121 | return (B_TRUE); | |
6122 | } else { | |
6123 | (void) fprintf(stderr, "something strange happened to child\n"); | |
6124 | exit(4); | |
6125 | /* NOTREACHED */ | |
6126 | } | |
6127 | } | |
6128 | ||
6129 | static void | |
6130 | ztest_run_init(void) | |
6131 | { | |
6132 | int i; | |
428870ff | 6133 | |
c242c188 | 6134 | ztest_shared_t *zs = ztest_shared; |
428870ff | 6135 | |
c242c188 | 6136 | ASSERT(ztest_opts.zo_init != 0); |
572e2857 | 6137 | |
c242c188 CS |
6138 | /* |
6139 | * Blow away any existing copy of zpool.cache | |
6140 | */ | |
6141 | (void) remove(spa_config_path); | |
6142 | ||
6143 | /* | |
6144 | * Create and initialize our storage pool. | |
6145 | */ | |
6146 | for (i = 1; i <= ztest_opts.zo_init; i++) { | |
6147 | bzero(zs, sizeof (ztest_shared_t)); | |
6148 | if (ztest_opts.zo_verbose >= 3 && | |
6149 | ztest_opts.zo_init != 1) { | |
6150 | (void) printf("ztest_init(), pass %d\n", i); | |
6151 | } | |
6152 | ztest_init(zs); | |
6153 | } | |
34dc7c2f BB |
6154 | } |
6155 | ||
6156 | int | |
6157 | main(int argc, char **argv) | |
6158 | { | |
6159 | int kills = 0; | |
6160 | int iters = 0; | |
c242c188 CS |
6161 | int older = 0; |
6162 | int newer = 0; | |
34dc7c2f BB |
6163 | ztest_shared_t *zs; |
6164 | ztest_info_t *zi; | |
c242c188 | 6165 | ztest_shared_callstate_t *zc; |
34dc7c2f BB |
6166 | char timebuf[100]; |
6167 | char numbuf[6]; | |
428870ff | 6168 | spa_t *spa; |
483106eb | 6169 | char *cmd; |
c242c188 CS |
6170 | boolean_t hasalt; |
6171 | int f; | |
9d81146b | 6172 | char *fd_data_str = getenv("ZTEST_FD_DATA"); |
34dc7c2f BB |
6173 | |
6174 | (void) setvbuf(stdout, NULL, _IOLBF, 0); | |
6175 | ||
9d81146b ED |
6176 | ztest_fd_rand = open("/dev/urandom", O_RDONLY); |
6177 | ASSERT3S(ztest_fd_rand, >=, 0); | |
6178 | ||
6179 | if (!fd_data_str) { | |
c242c188 CS |
6180 | dprintf_setup(&argc, argv); |
6181 | process_options(argc, argv); | |
34dc7c2f | 6182 | |
9d81146b | 6183 | setup_data_fd(); |
c242c188 CS |
6184 | setup_hdr(); |
6185 | setup_data(); | |
6186 | bcopy(&ztest_opts, ztest_shared_opts, | |
6187 | sizeof (*ztest_shared_opts)); | |
6188 | } else { | |
9d81146b | 6189 | ztest_fd_data = atoi(fd_data_str); |
c242c188 CS |
6190 | setup_data(); |
6191 | bcopy(ztest_shared_opts, &ztest_opts, sizeof (ztest_opts)); | |
6192 | } | |
6193 | ASSERT3U(ztest_opts.zo_datasets, ==, ztest_shared_hdr->zh_ds_count); | |
34dc7c2f | 6194 | |
428870ff | 6195 | /* Override location of zpool.cache */ |
5be98cfe BB |
6196 | VERIFY(asprintf((char **)&spa_config_path, "%s/zpool.cache", |
6197 | ztest_opts.zo_dir) != -1); | |
9d81146b | 6198 | |
c242c188 CS |
6199 | ztest_ds = umem_alloc(ztest_opts.zo_datasets * sizeof (ztest_ds_t), |
6200 | UMEM_NOFAIL); | |
6201 | zs = ztest_shared; | |
6202 | ||
9d81146b | 6203 | if (fd_data_str) { |
c242c188 CS |
6204 | metaslab_gang_bang = ztest_opts.zo_metaslab_gang_bang; |
6205 | metaslab_df_alloc_threshold = | |
6206 | zs->zs_metaslab_df_alloc_threshold; | |
6207 | ||
6208 | if (zs->zs_do_init) | |
6209 | ztest_run_init(); | |
6210 | else | |
6211 | ztest_run(zs); | |
6212 | exit(0); | |
6213 | } | |
34dc7c2f | 6214 | |
c242c188 | 6215 | hasalt = (strlen(ztest_opts.zo_alt_ztest) != 0); |
34dc7c2f | 6216 | |
c242c188 | 6217 | if (ztest_opts.zo_verbose >= 1) { |
34dc7c2f BB |
6218 | (void) printf("%llu vdevs, %d datasets, %d threads," |
6219 | " %llu seconds...\n", | |
c242c188 CS |
6220 | (u_longlong_t)ztest_opts.zo_vdevs, |
6221 | ztest_opts.zo_datasets, | |
6222 | ztest_opts.zo_threads, | |
6223 | (u_longlong_t)ztest_opts.zo_time); | |
34dc7c2f BB |
6224 | } |
6225 | ||
483106eb BB |
6226 | cmd = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); |
6227 | (void) strlcpy(cmd, getexecname(), MAXNAMELEN); | |
c242c188 CS |
6228 | |
6229 | zs->zs_do_init = B_TRUE; | |
6230 | if (strlen(ztest_opts.zo_alt_ztest) != 0) { | |
6231 | if (ztest_opts.zo_verbose >= 1) { | |
6232 | (void) printf("Executing older ztest for " | |
6233 | "initialization: %s\n", ztest_opts.zo_alt_ztest); | |
6234 | } | |
6235 | VERIFY(!exec_child(ztest_opts.zo_alt_ztest, | |
6236 | ztest_opts.zo_alt_libpath, B_FALSE, NULL)); | |
6237 | } else { | |
6238 | VERIFY(!exec_child(NULL, NULL, B_FALSE, NULL)); | |
34dc7c2f | 6239 | } |
c242c188 | 6240 | zs->zs_do_init = B_FALSE; |
34dc7c2f | 6241 | |
428870ff | 6242 | zs->zs_proc_start = gethrtime(); |
c242c188 | 6243 | zs->zs_proc_stop = zs->zs_proc_start + ztest_opts.zo_time * NANOSEC; |
34dc7c2f | 6244 | |
d6320ddb | 6245 | for (f = 0; f < ZTEST_FUNCS; f++) { |
c242c188 CS |
6246 | zi = &ztest_info[f]; |
6247 | zc = ZTEST_GET_SHARED_CALLSTATE(f); | |
428870ff | 6248 | if (zs->zs_proc_start + zi->zi_interval[0] > zs->zs_proc_stop) |
c242c188 | 6249 | zc->zc_next = UINT64_MAX; |
34dc7c2f | 6250 | else |
c242c188 | 6251 | zc->zc_next = zs->zs_proc_start + |
428870ff | 6252 | ztest_random(2 * zi->zi_interval[0] + 1); |
34dc7c2f BB |
6253 | } |
6254 | ||
34dc7c2f BB |
6255 | /* |
6256 | * Run the tests in a loop. These tests include fault injection | |
6257 | * to verify that self-healing data works, and forced crashes | |
6258 | * to verify that we never lose on-disk consistency. | |
6259 | */ | |
428870ff | 6260 | while (gethrtime() < zs->zs_proc_stop) { |
34dc7c2f | 6261 | int status; |
c242c188 | 6262 | boolean_t killed; |
34dc7c2f BB |
6263 | |
6264 | /* | |
6265 | * Initialize the workload counters for each function. | |
6266 | */ | |
d6320ddb | 6267 | for (f = 0; f < ZTEST_FUNCS; f++) { |
c242c188 CS |
6268 | zc = ZTEST_GET_SHARED_CALLSTATE(f); |
6269 | zc->zc_count = 0; | |
6270 | zc->zc_time = 0; | |
34dc7c2f BB |
6271 | } |
6272 | ||
9babb374 | 6273 | /* Set the allocation switch size */ |
c242c188 CS |
6274 | zs->zs_metaslab_df_alloc_threshold = |
6275 | ztest_random(zs->zs_metaslab_sz / 4) + 1; | |
9babb374 | 6276 | |
c242c188 CS |
6277 | if (!hasalt || ztest_random(2) == 0) { |
6278 | if (hasalt && ztest_opts.zo_verbose >= 1) { | |
6279 | (void) printf("Executing newer ztest: %s\n", | |
6280 | cmd); | |
34dc7c2f | 6281 | } |
c242c188 CS |
6282 | newer++; |
6283 | killed = exec_child(cmd, NULL, B_TRUE, &status); | |
34dc7c2f | 6284 | } else { |
c242c188 CS |
6285 | if (hasalt && ztest_opts.zo_verbose >= 1) { |
6286 | (void) printf("Executing older ztest: %s\n", | |
6287 | ztest_opts.zo_alt_ztest); | |
6288 | } | |
6289 | older++; | |
6290 | killed = exec_child(ztest_opts.zo_alt_ztest, | |
6291 | ztest_opts.zo_alt_libpath, B_TRUE, &status); | |
34dc7c2f BB |
6292 | } |
6293 | ||
c242c188 CS |
6294 | if (killed) |
6295 | kills++; | |
34dc7c2f BB |
6296 | iters++; |
6297 | ||
c242c188 | 6298 | if (ztest_opts.zo_verbose >= 1) { |
34dc7c2f BB |
6299 | hrtime_t now = gethrtime(); |
6300 | ||
428870ff BB |
6301 | now = MIN(now, zs->zs_proc_stop); |
6302 | print_time(zs->zs_proc_stop - now, timebuf); | |
34dc7c2f BB |
6303 | nicenum(zs->zs_space, numbuf); |
6304 | ||
6305 | (void) printf("Pass %3d, %8s, %3llu ENOSPC, " | |
6306 | "%4.1f%% of %5s used, %3.0f%% done, %8s to go\n", | |
6307 | iters, | |
6308 | WIFEXITED(status) ? "Complete" : "SIGKILL", | |
6309 | (u_longlong_t)zs->zs_enospc_count, | |
6310 | 100.0 * zs->zs_alloc / zs->zs_space, | |
6311 | numbuf, | |
428870ff | 6312 | 100.0 * (now - zs->zs_proc_start) / |
c242c188 | 6313 | (ztest_opts.zo_time * NANOSEC), timebuf); |
34dc7c2f BB |
6314 | } |
6315 | ||
c242c188 | 6316 | if (ztest_opts.zo_verbose >= 2) { |
34dc7c2f BB |
6317 | (void) printf("\nWorkload summary:\n\n"); |
6318 | (void) printf("%7s %9s %s\n", | |
6319 | "Calls", "Time", "Function"); | |
6320 | (void) printf("%7s %9s %s\n", | |
6321 | "-----", "----", "--------"); | |
d6320ddb | 6322 | for (f = 0; f < ZTEST_FUNCS; f++) { |
34dc7c2f BB |
6323 | Dl_info dli; |
6324 | ||
c242c188 CS |
6325 | zi = &ztest_info[f]; |
6326 | zc = ZTEST_GET_SHARED_CALLSTATE(f); | |
6327 | print_time(zc->zc_time, timebuf); | |
34dc7c2f BB |
6328 | (void) dladdr((void *)zi->zi_func, &dli); |
6329 | (void) printf("%7llu %9s %s\n", | |
c242c188 | 6330 | (u_longlong_t)zc->zc_count, timebuf, |
34dc7c2f BB |
6331 | dli.dli_sname); |
6332 | } | |
6333 | (void) printf("\n"); | |
6334 | } | |
6335 | ||
6336 | /* | |
428870ff BB |
6337 | * It's possible that we killed a child during a rename test, |
6338 | * in which case we'll have a 'ztest_tmp' pool lying around | |
6339 | * instead of 'ztest'. Do a blind rename in case this happened. | |
34dc7c2f | 6340 | */ |
428870ff | 6341 | kernel_init(FREAD); |
c242c188 | 6342 | if (spa_open(ztest_opts.zo_pool, &spa, FTAG) == 0) { |
428870ff BB |
6343 | spa_close(spa, FTAG); |
6344 | } else { | |
6345 | char tmpname[MAXNAMELEN]; | |
6346 | kernel_fini(); | |
6347 | kernel_init(FREAD | FWRITE); | |
6348 | (void) snprintf(tmpname, sizeof (tmpname), "%s_tmp", | |
c242c188 CS |
6349 | ztest_opts.zo_pool); |
6350 | (void) spa_rename(tmpname, ztest_opts.zo_pool); | |
428870ff | 6351 | } |
34dc7c2f | 6352 | kernel_fini(); |
34dc7c2f | 6353 | |
c242c188 | 6354 | ztest_run_zdb(ztest_opts.zo_pool); |
428870ff | 6355 | } |
34dc7c2f | 6356 | |
c242c188 CS |
6357 | if (ztest_opts.zo_verbose >= 1) { |
6358 | if (hasalt) { | |
6359 | (void) printf("%d runs of older ztest: %s\n", older, | |
6360 | ztest_opts.zo_alt_ztest); | |
6361 | (void) printf("%d runs of newer ztest: %s\n", newer, | |
6362 | cmd); | |
6363 | } | |
34dc7c2f BB |
6364 | (void) printf("%d killed, %d completed, %.0f%% kill rate\n", |
6365 | kills, iters - kills, (100.0 * kills) / MAX(1, iters)); | |
6366 | } | |
6367 | ||
483106eb BB |
6368 | umem_free(cmd, MAXNAMELEN); |
6369 | ||
34dc7c2f BB |
6370 | return (0); |
6371 | } |