]>
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. |
34dc7c2f BB |
25 | */ |
26 | ||
34dc7c2f BB |
27 | /* |
28 | * The objective of this program is to provide a DMU/ZAP/SPA stress test | |
29 | * that runs entirely in userland, is easy to use, and easy to extend. | |
30 | * | |
31 | * The overall design of the ztest program is as follows: | |
32 | * | |
33 | * (1) For each major functional area (e.g. adding vdevs to a pool, | |
34 | * creating and destroying datasets, reading and writing objects, etc) | |
35 | * we have a simple routine to test that functionality. These | |
36 | * individual routines do not have to do anything "stressful". | |
37 | * | |
38 | * (2) We turn these simple functionality tests into a stress test by | |
39 | * running them all in parallel, with as many threads as desired, | |
40 | * and spread across as many datasets, objects, and vdevs as desired. | |
41 | * | |
42 | * (3) While all this is happening, we inject faults into the pool to | |
43 | * verify that self-healing data really works. | |
44 | * | |
45 | * (4) Every time we open a dataset, we change its checksum and compression | |
46 | * functions. Thus even individual objects vary from block to block | |
47 | * in which checksum they use and whether they're compressed. | |
48 | * | |
49 | * (5) To verify that we never lose on-disk consistency after a crash, | |
50 | * we run the entire test in a child of the main process. | |
51 | * At random times, the child self-immolates with a SIGKILL. | |
52 | * This is the software equivalent of pulling the power cord. | |
53 | * The parent then runs the test again, using the existing | |
c242c188 CS |
54 | * storage pool, as many times as desired. If backwards compatability |
55 | * testing is enabled ztest will sometimes run the "older" version | |
56 | * of ztest after a SIGKILL. | |
34dc7c2f BB |
57 | * |
58 | * (6) To verify that we don't have future leaks or temporal incursions, | |
59 | * many of the functional tests record the transaction group number | |
60 | * as part of their data. When reading old data, they verify that | |
61 | * the transaction group number is less than the current, open txg. | |
62 | * If you add a new test, please do this if applicable. | |
63 | * | |
1e33ac1e BB |
64 | * (7) Threads are created with a reduced stack size, for sanity checking. |
65 | * Therefore, it's important not to allocate huge buffers on the stack. | |
66 | * | |
34dc7c2f BB |
67 | * When run with no arguments, ztest runs for about five minutes and |
68 | * produces no output if successful. To get a little bit of information, | |
69 | * specify -V. To get more information, specify -VV, and so on. | |
70 | * | |
71 | * To turn this into an overnight stress test, use -T to specify run time. | |
72 | * | |
73 | * You can ask more more vdevs [-v], datasets [-d], or threads [-t] | |
74 | * to increase the pool capacity, fanout, and overall stress level. | |
75 | * | |
c242c188 CS |
76 | * Use the -k option to set the desired frequency of kills. |
77 | * | |
78 | * When ztest invokes itself it passes all relevant information through a | |
79 | * temporary file which is mmap-ed in the child process. This allows shared | |
80 | * memory to survive the exec syscall. The ztest_shared_hdr_t struct is always | |
81 | * stored at offset 0 of this file and contains information on the size and | |
82 | * number of shared structures in the file. The information stored in this file | |
83 | * must remain backwards compatible with older versions of ztest so that | |
84 | * ztest can invoke them during backwards compatibility testing (-B). | |
34dc7c2f BB |
85 | */ |
86 | ||
87 | #include <sys/zfs_context.h> | |
88 | #include <sys/spa.h> | |
89 | #include <sys/dmu.h> | |
90 | #include <sys/txg.h> | |
9babb374 | 91 | #include <sys/dbuf.h> |
34dc7c2f | 92 | #include <sys/zap.h> |
34dc7c2f BB |
93 | #include <sys/dmu_objset.h> |
94 | #include <sys/poll.h> | |
95 | #include <sys/stat.h> | |
96 | #include <sys/time.h> | |
97 | #include <sys/wait.h> | |
98 | #include <sys/mman.h> | |
99 | #include <sys/resource.h> | |
100 | #include <sys/zio.h> | |
34dc7c2f | 101 | #include <sys/zil.h> |
428870ff | 102 | #include <sys/zil_impl.h> |
34dc7c2f | 103 | #include <sys/vdev_impl.h> |
b128c09f | 104 | #include <sys/vdev_file.h> |
34dc7c2f | 105 | #include <sys/spa_impl.h> |
428870ff | 106 | #include <sys/metaslab_impl.h> |
34dc7c2f | 107 | #include <sys/dsl_prop.h> |
9babb374 | 108 | #include <sys/dsl_dataset.h> |
428870ff BB |
109 | #include <sys/dsl_scan.h> |
110 | #include <sys/zio_checksum.h> | |
34dc7c2f BB |
111 | #include <sys/refcount.h> |
112 | #include <stdio.h> | |
113 | #include <stdio_ext.h> | |
114 | #include <stdlib.h> | |
115 | #include <unistd.h> | |
116 | #include <signal.h> | |
117 | #include <umem.h> | |
118 | #include <dlfcn.h> | |
119 | #include <ctype.h> | |
120 | #include <math.h> | |
121 | #include <sys/fs/zfs.h> | |
428870ff | 122 | #include <libnvpair.h> |
34dc7c2f | 123 | |
c242c188 CS |
124 | #define ZTEST_FD_DATA 3 |
125 | #define ZTEST_FD_RAND 4 | |
126 | ||
127 | typedef struct ztest_shared_hdr { | |
128 | uint64_t zh_hdr_size; | |
129 | uint64_t zh_opts_size; | |
130 | uint64_t zh_size; | |
131 | uint64_t zh_stats_size; | |
132 | uint64_t zh_stats_count; | |
133 | uint64_t zh_ds_size; | |
134 | uint64_t zh_ds_count; | |
135 | } ztest_shared_hdr_t; | |
136 | ||
137 | static ztest_shared_hdr_t *ztest_shared_hdr; | |
138 | ||
139 | typedef struct ztest_shared_opts { | |
140 | char zo_pool[MAXNAMELEN]; | |
141 | char zo_dir[MAXNAMELEN]; | |
142 | char zo_alt_ztest[MAXNAMELEN]; | |
143 | char zo_alt_libpath[MAXNAMELEN]; | |
144 | uint64_t zo_vdevs; | |
145 | uint64_t zo_vdevtime; | |
146 | size_t zo_vdev_size; | |
147 | int zo_ashift; | |
148 | int zo_mirrors; | |
149 | int zo_raidz; | |
150 | int zo_raidz_parity; | |
151 | int zo_datasets; | |
152 | int zo_threads; | |
153 | uint64_t zo_passtime; | |
154 | uint64_t zo_killrate; | |
155 | int zo_verbose; | |
156 | int zo_init; | |
157 | uint64_t zo_time; | |
158 | uint64_t zo_maxloops; | |
159 | uint64_t zo_metaslab_gang_bang; | |
160 | } ztest_shared_opts_t; | |
161 | ||
162 | static const ztest_shared_opts_t ztest_opts_defaults = { | |
163 | .zo_pool = { 'z', 't', 'e', 's', 't', '\0' }, | |
164 | .zo_dir = { '/', 't', 'm', 'p', '\0' }, | |
165 | .zo_alt_ztest = { '\0' }, | |
166 | .zo_alt_libpath = { '\0' }, | |
167 | .zo_vdevs = 5, | |
168 | .zo_ashift = SPA_MINBLOCKSHIFT, | |
169 | .zo_mirrors = 2, | |
170 | .zo_raidz = 4, | |
171 | .zo_raidz_parity = 1, | |
172 | .zo_vdev_size = SPA_MINDEVSIZE, | |
173 | .zo_datasets = 7, | |
174 | .zo_threads = 23, | |
175 | .zo_passtime = 60, /* 60 seconds */ | |
176 | .zo_killrate = 70, /* 70% kill rate */ | |
177 | .zo_verbose = 0, | |
178 | .zo_init = 1, | |
179 | .zo_time = 300, /* 5 minutes */ | |
180 | .zo_maxloops = 50, /* max loops during spa_freeze() */ | |
181 | .zo_metaslab_gang_bang = 32 << 10 | |
182 | }; | |
183 | ||
184 | extern uint64_t metaslab_gang_bang; | |
185 | extern uint64_t metaslab_df_alloc_threshold; | |
186 | ||
187 | static ztest_shared_opts_t *ztest_shared_opts; | |
188 | static ztest_shared_opts_t ztest_opts; | |
189 | ||
190 | typedef struct ztest_shared_ds { | |
191 | uint64_t zd_seq; | |
192 | } ztest_shared_ds_t; | |
193 | ||
194 | static ztest_shared_ds_t *ztest_shared_ds; | |
195 | #define ZTEST_GET_SHARED_DS(d) (&ztest_shared_ds[d]) | |
428870ff BB |
196 | |
197 | #define BT_MAGIC 0x123456789abcdefULL | |
c242c188 CS |
198 | #define MAXFAULTS() \ |
199 | (MAX(zs->zs_mirrors, 1) * (ztest_opts.zo_raidz_parity + 1) - 1) | |
428870ff BB |
200 | |
201 | enum ztest_io_type { | |
202 | ZTEST_IO_WRITE_TAG, | |
203 | ZTEST_IO_WRITE_PATTERN, | |
204 | ZTEST_IO_WRITE_ZEROES, | |
205 | ZTEST_IO_TRUNCATE, | |
206 | ZTEST_IO_SETATTR, | |
207 | ZTEST_IO_TYPES | |
208 | }; | |
34dc7c2f BB |
209 | |
210 | typedef struct ztest_block_tag { | |
428870ff | 211 | uint64_t bt_magic; |
34dc7c2f BB |
212 | uint64_t bt_objset; |
213 | uint64_t bt_object; | |
214 | uint64_t bt_offset; | |
428870ff | 215 | uint64_t bt_gen; |
34dc7c2f | 216 | uint64_t bt_txg; |
428870ff | 217 | uint64_t bt_crtxg; |
34dc7c2f BB |
218 | } ztest_block_tag_t; |
219 | ||
428870ff BB |
220 | typedef struct bufwad { |
221 | uint64_t bw_index; | |
222 | uint64_t bw_txg; | |
223 | uint64_t bw_data; | |
224 | } bufwad_t; | |
225 | ||
226 | /* | |
227 | * XXX -- fix zfs range locks to be generic so we can use them here. | |
228 | */ | |
229 | typedef enum { | |
230 | RL_READER, | |
231 | RL_WRITER, | |
232 | RL_APPEND | |
233 | } rl_type_t; | |
234 | ||
235 | typedef struct rll { | |
236 | void *rll_writer; | |
237 | int rll_readers; | |
1e33ac1e BB |
238 | kmutex_t rll_lock; |
239 | kcondvar_t rll_cv; | |
428870ff BB |
240 | } rll_t; |
241 | ||
242 | typedef struct rl { | |
243 | uint64_t rl_object; | |
244 | uint64_t rl_offset; | |
245 | uint64_t rl_size; | |
246 | rll_t *rl_lock; | |
247 | } rl_t; | |
248 | ||
249 | #define ZTEST_RANGE_LOCKS 64 | |
250 | #define ZTEST_OBJECT_LOCKS 64 | |
251 | ||
252 | /* | |
253 | * Object descriptor. Used as a template for object lookup/create/remove. | |
254 | */ | |
255 | typedef struct ztest_od { | |
256 | uint64_t od_dir; | |
257 | uint64_t od_object; | |
258 | dmu_object_type_t od_type; | |
259 | dmu_object_type_t od_crtype; | |
260 | uint64_t od_blocksize; | |
261 | uint64_t od_crblocksize; | |
262 | uint64_t od_gen; | |
263 | uint64_t od_crgen; | |
264 | char od_name[MAXNAMELEN]; | |
265 | } ztest_od_t; | |
34dc7c2f | 266 | |
428870ff BB |
267 | /* |
268 | * Per-dataset state. | |
269 | */ | |
270 | typedef struct ztest_ds { | |
c242c188 | 271 | ztest_shared_ds_t *zd_shared; |
428870ff | 272 | objset_t *zd_os; |
3e31d2b0 | 273 | krwlock_t zd_zilog_lock; |
428870ff | 274 | zilog_t *zd_zilog; |
428870ff BB |
275 | ztest_od_t *zd_od; /* debugging aid */ |
276 | char zd_name[MAXNAMELEN]; | |
1e33ac1e | 277 | kmutex_t zd_dirobj_lock; |
428870ff BB |
278 | rll_t zd_object_lock[ZTEST_OBJECT_LOCKS]; |
279 | rll_t zd_range_lock[ZTEST_RANGE_LOCKS]; | |
280 | } ztest_ds_t; | |
281 | ||
282 | /* | |
283 | * Per-iteration state. | |
284 | */ | |
285 | typedef void ztest_func_t(ztest_ds_t *zd, uint64_t id); | |
286 | ||
287 | typedef struct ztest_info { | |
288 | ztest_func_t *zi_func; /* test function */ | |
289 | uint64_t zi_iters; /* iterations per execution */ | |
290 | uint64_t *zi_interval; /* execute every <interval> seconds */ | |
428870ff | 291 | } ztest_info_t; |
34dc7c2f | 292 | |
c242c188 CS |
293 | typedef struct ztest_shared_callstate { |
294 | uint64_t zc_count; /* per-pass count */ | |
295 | uint64_t zc_time; /* per-pass time */ | |
296 | uint64_t zc_next; /* next time to call this function */ | |
297 | } ztest_shared_callstate_t; | |
298 | ||
299 | static ztest_shared_callstate_t *ztest_shared_callstate; | |
300 | #define ZTEST_GET_SHARED_CALLSTATE(c) (&ztest_shared_callstate[c]) | |
301 | ||
34dc7c2f BB |
302 | /* |
303 | * Note: these aren't static because we want dladdr() to work. | |
304 | */ | |
305 | ztest_func_t ztest_dmu_read_write; | |
306 | ztest_func_t ztest_dmu_write_parallel; | |
307 | ztest_func_t ztest_dmu_object_alloc_free; | |
428870ff | 308 | ztest_func_t ztest_dmu_commit_callbacks; |
34dc7c2f BB |
309 | ztest_func_t ztest_zap; |
310 | ztest_func_t ztest_zap_parallel; | |
428870ff | 311 | ztest_func_t ztest_zil_commit; |
3e31d2b0 | 312 | ztest_func_t ztest_zil_remount; |
428870ff | 313 | ztest_func_t ztest_dmu_read_write_zcopy; |
34dc7c2f | 314 | ztest_func_t ztest_dmu_objset_create_destroy; |
428870ff BB |
315 | ztest_func_t ztest_dmu_prealloc; |
316 | ztest_func_t ztest_fzap; | |
34dc7c2f | 317 | ztest_func_t ztest_dmu_snapshot_create_destroy; |
428870ff BB |
318 | ztest_func_t ztest_dsl_prop_get_set; |
319 | ztest_func_t ztest_spa_prop_get_set; | |
34dc7c2f BB |
320 | ztest_func_t ztest_spa_create_destroy; |
321 | ztest_func_t ztest_fault_inject; | |
428870ff BB |
322 | ztest_func_t ztest_ddt_repair; |
323 | ztest_func_t ztest_dmu_snapshot_hold; | |
b128c09f | 324 | ztest_func_t ztest_spa_rename; |
428870ff BB |
325 | ztest_func_t ztest_scrub; |
326 | ztest_func_t ztest_dsl_dataset_promote_busy; | |
34dc7c2f BB |
327 | ztest_func_t ztest_vdev_attach_detach; |
328 | ztest_func_t ztest_vdev_LUN_growth; | |
329 | ztest_func_t ztest_vdev_add_remove; | |
b128c09f | 330 | ztest_func_t ztest_vdev_aux_add_remove; |
428870ff | 331 | ztest_func_t ztest_split_pool; |
3541dc6d | 332 | ztest_func_t ztest_reguid; |
34dc7c2f | 333 | |
428870ff BB |
334 | uint64_t zopt_always = 0ULL * NANOSEC; /* all the time */ |
335 | uint64_t zopt_incessant = 1ULL * NANOSEC / 10; /* every 1/10 second */ | |
336 | uint64_t zopt_often = 1ULL * NANOSEC; /* every second */ | |
337 | uint64_t zopt_sometimes = 10ULL * NANOSEC; /* every 10 seconds */ | |
338 | uint64_t zopt_rarely = 60ULL * NANOSEC; /* every 60 seconds */ | |
34dc7c2f BB |
339 | |
340 | ztest_info_t ztest_info[] = { | |
341 | { ztest_dmu_read_write, 1, &zopt_always }, | |
428870ff | 342 | { ztest_dmu_write_parallel, 10, &zopt_always }, |
34dc7c2f | 343 | { ztest_dmu_object_alloc_free, 1, &zopt_always }, |
428870ff | 344 | { ztest_dmu_commit_callbacks, 1, &zopt_always }, |
34dc7c2f BB |
345 | { ztest_zap, 30, &zopt_always }, |
346 | { ztest_zap_parallel, 100, &zopt_always }, | |
428870ff BB |
347 | { ztest_split_pool, 1, &zopt_always }, |
348 | { ztest_zil_commit, 1, &zopt_incessant }, | |
3e31d2b0 | 349 | { ztest_zil_remount, 1, &zopt_sometimes }, |
428870ff BB |
350 | { ztest_dmu_read_write_zcopy, 1, &zopt_often }, |
351 | { ztest_dmu_objset_create_destroy, 1, &zopt_often }, | |
352 | { ztest_dsl_prop_get_set, 1, &zopt_often }, | |
353 | { ztest_spa_prop_get_set, 1, &zopt_sometimes }, | |
354 | #if 0 | |
355 | { ztest_dmu_prealloc, 1, &zopt_sometimes }, | |
356 | #endif | |
357 | { ztest_fzap, 1, &zopt_sometimes }, | |
358 | { ztest_dmu_snapshot_create_destroy, 1, &zopt_sometimes }, | |
359 | { ztest_spa_create_destroy, 1, &zopt_sometimes }, | |
34dc7c2f | 360 | { ztest_fault_inject, 1, &zopt_sometimes }, |
428870ff BB |
361 | { ztest_ddt_repair, 1, &zopt_sometimes }, |
362 | { ztest_dmu_snapshot_hold, 1, &zopt_sometimes }, | |
d1352457 ED |
363 | /* |
364 | * The reguid test is currently broken. Disable it until | |
365 | * we get around to fixing it. | |
366 | */ | |
367 | #if 0 | |
3541dc6d | 368 | { ztest_reguid, 1, &zopt_sometimes }, |
d1352457 | 369 | #endif |
34dc7c2f | 370 | { ztest_spa_rename, 1, &zopt_rarely }, |
428870ff | 371 | { ztest_scrub, 1, &zopt_rarely }, |
9babb374 | 372 | { ztest_dsl_dataset_promote_busy, 1, &zopt_rarely }, |
428870ff BB |
373 | { ztest_vdev_attach_detach, 1, &zopt_rarely }, |
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; | |
424 | static krwlock_t ztest_name_lock; | |
34dc7c2f | 425 | |
c242c188 | 426 | static boolean_t ztest_dump_core = B_TRUE; |
b128c09f | 427 | static boolean_t ztest_exiting; |
34dc7c2f | 428 | |
428870ff BB |
429 | /* Global commit callback list */ |
430 | static ztest_cb_list_t zcl; | |
090ff092 RC |
431 | /* Commit cb delay */ |
432 | static uint64_t zc_min_txg_delay = UINT64_MAX; | |
433 | static int zc_cb_counter = 0; | |
434 | ||
435 | /* | |
436 | * Minimum number of commit callbacks that need to be registered for us to check | |
437 | * whether the minimum txg delay is acceptable. | |
438 | */ | |
439 | #define ZTEST_COMMIT_CB_MIN_REG 100 | |
440 | ||
441 | /* | |
442 | * If a number of txgs equal to this threshold have been created after a commit | |
443 | * callback has been registered but not called, then we assume there is an | |
444 | * implementation bug. | |
445 | */ | |
446 | #define ZTEST_COMMIT_CB_THRESH (TXG_CONCURRENT_STATES + 1000) | |
428870ff | 447 | |
34dc7c2f | 448 | extern uint64_t metaslab_gang_bang; |
9babb374 | 449 | extern uint64_t metaslab_df_alloc_threshold; |
34dc7c2f | 450 | |
428870ff BB |
451 | enum ztest_object { |
452 | ZTEST_META_DNODE = 0, | |
453 | ZTEST_DIROBJ, | |
454 | ZTEST_OBJECTS | |
455 | }; | |
34dc7c2f BB |
456 | |
457 | static void usage(boolean_t) __NORETURN; | |
458 | ||
459 | /* | |
460 | * These libumem hooks provide a reasonable set of defaults for the allocator's | |
461 | * debugging facilities. | |
462 | */ | |
463 | const char * | |
0bc8fd78 | 464 | _umem_debug_init(void) |
34dc7c2f BB |
465 | { |
466 | return ("default,verbose"); /* $UMEM_DEBUG setting */ | |
467 | } | |
468 | ||
469 | const char * | |
470 | _umem_logging_init(void) | |
471 | { | |
472 | return ("fail,contents"); /* $UMEM_LOGGING setting */ | |
473 | } | |
474 | ||
475 | #define FATAL_MSG_SZ 1024 | |
476 | ||
477 | char *fatal_msg; | |
478 | ||
479 | static void | |
480 | fatal(int do_perror, char *message, ...) | |
481 | { | |
482 | va_list args; | |
483 | int save_errno = errno; | |
40b84e7a | 484 | char *buf; |
34dc7c2f BB |
485 | |
486 | (void) fflush(stdout); | |
40b84e7a | 487 | buf = umem_alloc(FATAL_MSG_SZ, UMEM_NOFAIL); |
34dc7c2f BB |
488 | |
489 | va_start(args, message); | |
490 | (void) sprintf(buf, "ztest: "); | |
491 | /* LINTED */ | |
492 | (void) vsprintf(buf + strlen(buf), message, args); | |
493 | va_end(args); | |
494 | if (do_perror) { | |
495 | (void) snprintf(buf + strlen(buf), FATAL_MSG_SZ - strlen(buf), | |
496 | ": %s", strerror(save_errno)); | |
497 | } | |
498 | (void) fprintf(stderr, "%s\n", buf); | |
499 | fatal_msg = buf; /* to ease debugging */ | |
500 | if (ztest_dump_core) | |
501 | abort(); | |
502 | exit(3); | |
503 | } | |
504 | ||
505 | static int | |
506 | str2shift(const char *buf) | |
507 | { | |
508 | const char *ends = "BKMGTPEZ"; | |
509 | int i; | |
510 | ||
511 | if (buf[0] == '\0') | |
512 | return (0); | |
513 | for (i = 0; i < strlen(ends); i++) { | |
514 | if (toupper(buf[0]) == ends[i]) | |
515 | break; | |
516 | } | |
517 | if (i == strlen(ends)) { | |
518 | (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", | |
519 | buf); | |
520 | usage(B_FALSE); | |
521 | } | |
522 | if (buf[1] == '\0' || (toupper(buf[1]) == 'B' && buf[2] == '\0')) { | |
523 | return (10*i); | |
524 | } | |
525 | (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", buf); | |
526 | usage(B_FALSE); | |
527 | /* NOTREACHED */ | |
528 | } | |
529 | ||
530 | static uint64_t | |
531 | nicenumtoull(const char *buf) | |
532 | { | |
533 | char *end; | |
534 | uint64_t val; | |
535 | ||
536 | val = strtoull(buf, &end, 0); | |
537 | if (end == buf) { | |
538 | (void) fprintf(stderr, "ztest: bad numeric value: %s\n", buf); | |
539 | usage(B_FALSE); | |
540 | } else if (end[0] == '.') { | |
541 | double fval = strtod(buf, &end); | |
542 | fval *= pow(2, str2shift(end)); | |
543 | if (fval > UINT64_MAX) { | |
544 | (void) fprintf(stderr, "ztest: value too large: %s\n", | |
545 | buf); | |
546 | usage(B_FALSE); | |
547 | } | |
548 | val = (uint64_t)fval; | |
549 | } else { | |
550 | int shift = str2shift(end); | |
551 | if (shift >= 64 || (val << shift) >> shift != val) { | |
552 | (void) fprintf(stderr, "ztest: value too large: %s\n", | |
553 | buf); | |
554 | usage(B_FALSE); | |
555 | } | |
556 | val <<= shift; | |
557 | } | |
558 | return (val); | |
559 | } | |
560 | ||
561 | static void | |
562 | usage(boolean_t requested) | |
563 | { | |
c242c188 CS |
564 | const ztest_shared_opts_t *zo = &ztest_opts_defaults; |
565 | ||
34dc7c2f BB |
566 | char nice_vdev_size[10]; |
567 | char nice_gang_bang[10]; | |
568 | FILE *fp = requested ? stdout : stderr; | |
569 | ||
c242c188 CS |
570 | nicenum(zo->zo_vdev_size, nice_vdev_size); |
571 | nicenum(zo->zo_metaslab_gang_bang, nice_gang_bang); | |
34dc7c2f BB |
572 | |
573 | (void) fprintf(fp, "Usage: %s\n" | |
574 | "\t[-v vdevs (default: %llu)]\n" | |
575 | "\t[-s size_of_each_vdev (default: %s)]\n" | |
428870ff | 576 | "\t[-a alignment_shift (default: %d)] use 0 for random\n" |
34dc7c2f BB |
577 | "\t[-m mirror_copies (default: %d)]\n" |
578 | "\t[-r raidz_disks (default: %d)]\n" | |
579 | "\t[-R raidz_parity (default: %d)]\n" | |
580 | "\t[-d datasets (default: %d)]\n" | |
581 | "\t[-t threads (default: %d)]\n" | |
582 | "\t[-g gang_block_threshold (default: %s)]\n" | |
428870ff BB |
583 | "\t[-i init_count (default: %d)] initialize pool i times\n" |
584 | "\t[-k kill_percentage (default: %llu%%)]\n" | |
34dc7c2f | 585 | "\t[-p pool_name (default: %s)]\n" |
428870ff BB |
586 | "\t[-f dir (default: %s)] file directory for vdev files\n" |
587 | "\t[-V] verbose (use multiple times for ever more blather)\n" | |
588 | "\t[-E] use existing pool instead of creating new one\n" | |
589 | "\t[-T time (default: %llu sec)] total run time\n" | |
590 | "\t[-F freezeloops (default: %llu)] max loops in spa_freeze()\n" | |
591 | "\t[-P passtime (default: %llu sec)] time per pass\n" | |
c242c188 | 592 | "\t[-B alt_ztest (default: <none>)] alternate ztest path\n" |
34dc7c2f BB |
593 | "\t[-h] (print help)\n" |
594 | "", | |
c242c188 CS |
595 | zo->zo_pool, |
596 | (u_longlong_t)zo->zo_vdevs, /* -v */ | |
34dc7c2f | 597 | nice_vdev_size, /* -s */ |
c242c188 CS |
598 | zo->zo_ashift, /* -a */ |
599 | zo->zo_mirrors, /* -m */ | |
600 | zo->zo_raidz, /* -r */ | |
601 | zo->zo_raidz_parity, /* -R */ | |
602 | zo->zo_datasets, /* -d */ | |
603 | zo->zo_threads, /* -t */ | |
34dc7c2f | 604 | nice_gang_bang, /* -g */ |
c242c188 CS |
605 | zo->zo_init, /* -i */ |
606 | (u_longlong_t)zo->zo_killrate, /* -k */ | |
607 | zo->zo_pool, /* -p */ | |
608 | zo->zo_dir, /* -f */ | |
609 | (u_longlong_t)zo->zo_time, /* -T */ | |
610 | (u_longlong_t)zo->zo_maxloops, /* -F */ | |
611 | (u_longlong_t)zo->zo_passtime); | |
34dc7c2f BB |
612 | exit(requested ? 0 : 1); |
613 | } | |
614 | ||
34dc7c2f BB |
615 | static void |
616 | process_options(int argc, char **argv) | |
617 | { | |
c242c188 CS |
618 | char *path; |
619 | ztest_shared_opts_t *zo = &ztest_opts; | |
620 | ||
34dc7c2f BB |
621 | int opt; |
622 | uint64_t value; | |
c242c188 | 623 | char altdir[MAXNAMELEN] = { 0 }; |
34dc7c2f | 624 | |
c242c188 | 625 | bcopy(&ztest_opts_defaults, zo, sizeof (*zo)); |
34dc7c2f | 626 | |
34dc7c2f | 627 | while ((opt = getopt(argc, argv, |
c242c188 | 628 | "v:s:a:m:r:R:d:t:g:i:k:p:f:VET:P:hF:B:")) != EOF) { |
34dc7c2f BB |
629 | value = 0; |
630 | switch (opt) { | |
631 | case 'v': | |
632 | case 's': | |
633 | case 'a': | |
634 | case 'm': | |
635 | case 'r': | |
636 | case 'R': | |
637 | case 'd': | |
638 | case 't': | |
639 | case 'g': | |
640 | case 'i': | |
641 | case 'k': | |
642 | case 'T': | |
643 | case 'P': | |
428870ff | 644 | case 'F': |
34dc7c2f BB |
645 | value = nicenumtoull(optarg); |
646 | } | |
647 | switch (opt) { | |
648 | case 'v': | |
c242c188 | 649 | zo->zo_vdevs = value; |
34dc7c2f BB |
650 | break; |
651 | case 's': | |
c242c188 | 652 | zo->zo_vdev_size = MAX(SPA_MINDEVSIZE, value); |
34dc7c2f BB |
653 | break; |
654 | case 'a': | |
c242c188 | 655 | zo->zo_ashift = value; |
34dc7c2f BB |
656 | break; |
657 | case 'm': | |
c242c188 | 658 | zo->zo_mirrors = value; |
34dc7c2f BB |
659 | break; |
660 | case 'r': | |
c242c188 | 661 | zo->zo_raidz = MAX(1, value); |
34dc7c2f BB |
662 | break; |
663 | case 'R': | |
c242c188 | 664 | zo->zo_raidz_parity = MIN(MAX(value, 1), 3); |
34dc7c2f BB |
665 | break; |
666 | case 'd': | |
c242c188 | 667 | zo->zo_datasets = MAX(1, value); |
34dc7c2f BB |
668 | break; |
669 | case 't': | |
c242c188 | 670 | zo->zo_threads = MAX(1, value); |
34dc7c2f BB |
671 | break; |
672 | case 'g': | |
c242c188 CS |
673 | zo->zo_metaslab_gang_bang = MAX(SPA_MINBLOCKSIZE << 1, |
674 | value); | |
34dc7c2f BB |
675 | break; |
676 | case 'i': | |
c242c188 | 677 | zo->zo_init = value; |
34dc7c2f BB |
678 | break; |
679 | case 'k': | |
c242c188 | 680 | zo->zo_killrate = value; |
34dc7c2f BB |
681 | break; |
682 | case 'p': | |
c242c188 CS |
683 | (void) strlcpy(zo->zo_pool, optarg, |
684 | sizeof (zo->zo_pool)); | |
34dc7c2f BB |
685 | break; |
686 | case 'f': | |
c242c188 CS |
687 | path = realpath(optarg, NULL); |
688 | if (path == NULL) { | |
689 | (void) fprintf(stderr, "error: %s: %s\n", | |
690 | optarg, strerror(errno)); | |
691 | usage(B_FALSE); | |
692 | } else { | |
693 | (void) strlcpy(zo->zo_dir, path, | |
694 | sizeof (zo->zo_dir)); | |
695 | } | |
34dc7c2f BB |
696 | break; |
697 | case 'V': | |
c242c188 | 698 | zo->zo_verbose++; |
34dc7c2f BB |
699 | break; |
700 | case 'E': | |
c242c188 | 701 | zo->zo_init = 0; |
34dc7c2f BB |
702 | break; |
703 | case 'T': | |
c242c188 | 704 | zo->zo_time = value; |
34dc7c2f BB |
705 | break; |
706 | case 'P': | |
c242c188 | 707 | zo->zo_passtime = MAX(1, value); |
34dc7c2f | 708 | break; |
428870ff | 709 | case 'F': |
c242c188 CS |
710 | zo->zo_maxloops = MAX(1, value); |
711 | break; | |
712 | case 'B': | |
713 | (void) strlcpy(altdir, optarg, sizeof (altdir)); | |
428870ff | 714 | break; |
34dc7c2f BB |
715 | case 'h': |
716 | usage(B_TRUE); | |
717 | break; | |
718 | case '?': | |
719 | default: | |
720 | usage(B_FALSE); | |
721 | break; | |
722 | } | |
723 | } | |
724 | ||
c242c188 | 725 | zo->zo_raidz_parity = MIN(zo->zo_raidz_parity, zo->zo_raidz - 1); |
34dc7c2f | 726 | |
c242c188 CS |
727 | zo->zo_vdevtime = |
728 | (zo->zo_vdevs > 0 ? zo->zo_time * NANOSEC / zo->zo_vdevs : | |
428870ff | 729 | UINT64_MAX >> 2); |
c242c188 CS |
730 | |
731 | if (strlen(altdir) > 0) { | |
732 | char cmd[MAXNAMELEN]; | |
733 | char realaltdir[MAXNAMELEN]; | |
734 | char *bin; | |
735 | char *ztest; | |
736 | char *isa; | |
737 | int isalen; | |
738 | ||
739 | (void) realpath(getexecname(), cmd); | |
740 | if (0 != access(altdir, F_OK)) { | |
741 | ztest_dump_core = B_FALSE; | |
742 | fatal(B_TRUE, "invalid alternate ztest path: %s", | |
743 | altdir); | |
744 | } | |
745 | VERIFY(NULL != realpath(altdir, realaltdir)); | |
746 | ||
747 | /* | |
748 | * 'cmd' should be of the form "<anything>/usr/bin/<isa>/ztest". | |
749 | * We want to extract <isa> to determine if we should use | |
750 | * 32 or 64 bit binaries. | |
751 | */ | |
752 | bin = strstr(cmd, "/usr/bin/"); | |
753 | ztest = strstr(bin, "/ztest"); | |
754 | isa = bin + 9; | |
755 | isalen = ztest - isa; | |
756 | (void) snprintf(zo->zo_alt_ztest, sizeof (zo->zo_alt_ztest), | |
757 | "%s/usr/bin/%.*s/ztest", realaltdir, isalen, isa); | |
758 | (void) snprintf(zo->zo_alt_libpath, sizeof (zo->zo_alt_libpath), | |
759 | "%s/usr/lib/%.*s", realaltdir, isalen, isa); | |
760 | ||
761 | if (0 != access(zo->zo_alt_ztest, X_OK)) { | |
762 | ztest_dump_core = B_FALSE; | |
763 | fatal(B_TRUE, "invalid alternate ztest: %s", | |
764 | zo->zo_alt_ztest); | |
765 | } else if (0 != access(zo->zo_alt_libpath, X_OK)) { | |
766 | ztest_dump_core = B_FALSE; | |
767 | fatal(B_TRUE, "invalid alternate lib directory %s", | |
768 | zo->zo_alt_libpath); | |
769 | } | |
770 | } | |
428870ff BB |
771 | } |
772 | ||
773 | static void | |
774 | ztest_kill(ztest_shared_t *zs) | |
775 | { | |
c242c188 CS |
776 | zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(ztest_spa)); |
777 | zs->zs_space = metaslab_class_get_space(spa_normal_class(ztest_spa)); | |
428870ff BB |
778 | (void) kill(getpid(), SIGKILL); |
779 | } | |
780 | ||
781 | static uint64_t | |
782 | ztest_random(uint64_t range) | |
783 | { | |
784 | uint64_t r; | |
785 | ||
786 | if (range == 0) | |
787 | return (0); | |
788 | ||
c242c188 | 789 | if (read(ZTEST_FD_RAND, &r, sizeof (r)) != sizeof (r)) |
428870ff BB |
790 | fatal(1, "short read from /dev/urandom"); |
791 | ||
792 | return (r % range); | |
793 | } | |
794 | ||
795 | /* ARGSUSED */ | |
796 | static void | |
797 | ztest_record_enospc(const char *s) | |
798 | { | |
799 | ztest_shared->zs_enospc_count++; | |
34dc7c2f BB |
800 | } |
801 | ||
802 | static uint64_t | |
803 | ztest_get_ashift(void) | |
804 | { | |
c242c188 | 805 | if (ztest_opts.zo_ashift == 0) |
34dc7c2f | 806 | return (SPA_MINBLOCKSHIFT + ztest_random(3)); |
c242c188 | 807 | return (ztest_opts.zo_ashift); |
34dc7c2f BB |
808 | } |
809 | ||
810 | static nvlist_t * | |
b128c09f | 811 | make_vdev_file(char *path, char *aux, size_t size, uint64_t ashift) |
34dc7c2f | 812 | { |
40b84e7a | 813 | char *pathbuf; |
34dc7c2f | 814 | uint64_t vdev; |
34dc7c2f BB |
815 | nvlist_t *file; |
816 | ||
40b84e7a BB |
817 | pathbuf = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
818 | ||
b128c09f BB |
819 | if (ashift == 0) |
820 | ashift = ztest_get_ashift(); | |
821 | ||
822 | if (path == NULL) { | |
823 | path = pathbuf; | |
824 | ||
825 | if (aux != NULL) { | |
826 | vdev = ztest_shared->zs_vdev_aux; | |
c242c188 CS |
827 | (void) snprintf(path, MAXPATHLEN, |
828 | ztest_aux_template, ztest_opts.zo_dir, | |
829 | ztest_opts.zo_pool, aux, vdev); | |
b128c09f | 830 | } else { |
428870ff | 831 | vdev = ztest_shared->zs_vdev_next_leaf++; |
c242c188 CS |
832 | (void) snprintf(path, MAXPATHLEN, |
833 | ztest_dev_template, ztest_opts.zo_dir, | |
834 | ztest_opts.zo_pool, vdev); | |
b128c09f BB |
835 | } |
836 | } | |
34dc7c2f | 837 | |
b128c09f BB |
838 | if (size != 0) { |
839 | int fd = open(path, O_RDWR | O_CREAT | O_TRUNC, 0666); | |
34dc7c2f | 840 | if (fd == -1) |
b128c09f | 841 | fatal(1, "can't open %s", path); |
34dc7c2f | 842 | if (ftruncate(fd, size) != 0) |
b128c09f | 843 | fatal(1, "can't ftruncate %s", path); |
34dc7c2f BB |
844 | (void) close(fd); |
845 | } | |
846 | ||
847 | VERIFY(nvlist_alloc(&file, NV_UNIQUE_NAME, 0) == 0); | |
848 | VERIFY(nvlist_add_string(file, ZPOOL_CONFIG_TYPE, VDEV_TYPE_FILE) == 0); | |
b128c09f | 849 | VERIFY(nvlist_add_string(file, ZPOOL_CONFIG_PATH, path) == 0); |
34dc7c2f | 850 | VERIFY(nvlist_add_uint64(file, ZPOOL_CONFIG_ASHIFT, ashift) == 0); |
40b84e7a | 851 | umem_free(pathbuf, MAXPATHLEN); |
34dc7c2f BB |
852 | |
853 | return (file); | |
854 | } | |
855 | ||
856 | static nvlist_t * | |
b128c09f | 857 | make_vdev_raidz(char *path, char *aux, size_t size, uint64_t ashift, int r) |
34dc7c2f BB |
858 | { |
859 | nvlist_t *raidz, **child; | |
860 | int c; | |
861 | ||
862 | if (r < 2) | |
b128c09f | 863 | return (make_vdev_file(path, aux, size, ashift)); |
34dc7c2f BB |
864 | child = umem_alloc(r * sizeof (nvlist_t *), UMEM_NOFAIL); |
865 | ||
866 | for (c = 0; c < r; c++) | |
b128c09f | 867 | child[c] = make_vdev_file(path, aux, size, ashift); |
34dc7c2f BB |
868 | |
869 | VERIFY(nvlist_alloc(&raidz, NV_UNIQUE_NAME, 0) == 0); | |
870 | VERIFY(nvlist_add_string(raidz, ZPOOL_CONFIG_TYPE, | |
871 | VDEV_TYPE_RAIDZ) == 0); | |
872 | VERIFY(nvlist_add_uint64(raidz, ZPOOL_CONFIG_NPARITY, | |
c242c188 | 873 | ztest_opts.zo_raidz_parity) == 0); |
34dc7c2f BB |
874 | VERIFY(nvlist_add_nvlist_array(raidz, ZPOOL_CONFIG_CHILDREN, |
875 | child, r) == 0); | |
876 | ||
877 | for (c = 0; c < r; c++) | |
878 | nvlist_free(child[c]); | |
879 | ||
880 | umem_free(child, r * sizeof (nvlist_t *)); | |
881 | ||
882 | return (raidz); | |
883 | } | |
884 | ||
885 | static nvlist_t * | |
b128c09f BB |
886 | make_vdev_mirror(char *path, char *aux, size_t size, uint64_t ashift, |
887 | int r, int m) | |
34dc7c2f BB |
888 | { |
889 | nvlist_t *mirror, **child; | |
890 | int c; | |
891 | ||
892 | if (m < 1) | |
b128c09f | 893 | return (make_vdev_raidz(path, aux, size, ashift, r)); |
34dc7c2f BB |
894 | |
895 | child = umem_alloc(m * sizeof (nvlist_t *), UMEM_NOFAIL); | |
896 | ||
897 | for (c = 0; c < m; c++) | |
b128c09f | 898 | child[c] = make_vdev_raidz(path, aux, size, ashift, r); |
34dc7c2f BB |
899 | |
900 | VERIFY(nvlist_alloc(&mirror, NV_UNIQUE_NAME, 0) == 0); | |
901 | VERIFY(nvlist_add_string(mirror, ZPOOL_CONFIG_TYPE, | |
902 | VDEV_TYPE_MIRROR) == 0); | |
903 | VERIFY(nvlist_add_nvlist_array(mirror, ZPOOL_CONFIG_CHILDREN, | |
904 | child, m) == 0); | |
34dc7c2f BB |
905 | |
906 | for (c = 0; c < m; c++) | |
907 | nvlist_free(child[c]); | |
908 | ||
909 | umem_free(child, m * sizeof (nvlist_t *)); | |
910 | ||
911 | return (mirror); | |
912 | } | |
913 | ||
914 | static nvlist_t * | |
b128c09f BB |
915 | make_vdev_root(char *path, char *aux, size_t size, uint64_t ashift, |
916 | int log, int r, int m, int t) | |
34dc7c2f BB |
917 | { |
918 | nvlist_t *root, **child; | |
919 | int c; | |
920 | ||
921 | ASSERT(t > 0); | |
922 | ||
923 | child = umem_alloc(t * sizeof (nvlist_t *), UMEM_NOFAIL); | |
924 | ||
b128c09f BB |
925 | for (c = 0; c < t; c++) { |
926 | child[c] = make_vdev_mirror(path, aux, size, ashift, r, m); | |
927 | VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_IS_LOG, | |
928 | log) == 0); | |
929 | } | |
34dc7c2f BB |
930 | |
931 | VERIFY(nvlist_alloc(&root, NV_UNIQUE_NAME, 0) == 0); | |
932 | VERIFY(nvlist_add_string(root, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT) == 0); | |
b128c09f | 933 | VERIFY(nvlist_add_nvlist_array(root, aux ? aux : ZPOOL_CONFIG_CHILDREN, |
34dc7c2f BB |
934 | child, t) == 0); |
935 | ||
936 | for (c = 0; c < t; c++) | |
937 | nvlist_free(child[c]); | |
938 | ||
939 | umem_free(child, t * sizeof (nvlist_t *)); | |
940 | ||
941 | return (root); | |
942 | } | |
943 | ||
428870ff BB |
944 | static int |
945 | ztest_random_blocksize(void) | |
34dc7c2f | 946 | { |
428870ff BB |
947 | return (1 << (SPA_MINBLOCKSHIFT + |
948 | ztest_random(SPA_MAXBLOCKSHIFT - SPA_MINBLOCKSHIFT + 1))); | |
949 | } | |
34dc7c2f | 950 | |
428870ff BB |
951 | static int |
952 | ztest_random_ibshift(void) | |
953 | { | |
954 | return (DN_MIN_INDBLKSHIFT + | |
955 | ztest_random(DN_MAX_INDBLKSHIFT - DN_MIN_INDBLKSHIFT + 1)); | |
34dc7c2f BB |
956 | } |
957 | ||
428870ff BB |
958 | static uint64_t |
959 | ztest_random_vdev_top(spa_t *spa, boolean_t log_ok) | |
34dc7c2f | 960 | { |
428870ff BB |
961 | uint64_t top; |
962 | vdev_t *rvd = spa->spa_root_vdev; | |
963 | vdev_t *tvd; | |
34dc7c2f | 964 | |
428870ff | 965 | ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0); |
34dc7c2f | 966 | |
428870ff BB |
967 | do { |
968 | top = ztest_random(rvd->vdev_children); | |
969 | tvd = rvd->vdev_child[top]; | |
970 | } while (tvd->vdev_ishole || (tvd->vdev_islog && !log_ok) || | |
971 | tvd->vdev_mg == NULL || tvd->vdev_mg->mg_class == NULL); | |
34dc7c2f | 972 | |
428870ff | 973 | return (top); |
34dc7c2f BB |
974 | } |
975 | ||
428870ff BB |
976 | static uint64_t |
977 | ztest_random_dsl_prop(zfs_prop_t prop) | |
34dc7c2f | 978 | { |
428870ff BB |
979 | uint64_t value; |
980 | ||
981 | do { | |
982 | value = zfs_prop_random_value(prop, ztest_random(-1ULL)); | |
983 | } while (prop == ZFS_PROP_CHECKSUM && value == ZIO_CHECKSUM_OFF); | |
984 | ||
985 | return (value); | |
34dc7c2f BB |
986 | } |
987 | ||
34dc7c2f | 988 | static int |
428870ff BB |
989 | ztest_dsl_prop_set_uint64(char *osname, zfs_prop_t prop, uint64_t value, |
990 | boolean_t inherit) | |
34dc7c2f | 991 | { |
428870ff BB |
992 | const char *propname = zfs_prop_to_name(prop); |
993 | const char *valname; | |
40b84e7a | 994 | char *setpoint; |
428870ff | 995 | uint64_t curval; |
34dc7c2f BB |
996 | int error; |
997 | ||
428870ff BB |
998 | error = dsl_prop_set(osname, propname, |
999 | (inherit ? ZPROP_SRC_NONE : ZPROP_SRC_LOCAL), | |
1000 | sizeof (value), 1, &value); | |
34dc7c2f | 1001 | |
428870ff BB |
1002 | if (error == ENOSPC) { |
1003 | ztest_record_enospc(FTAG); | |
34dc7c2f BB |
1004 | return (error); |
1005 | } | |
34dc7c2f | 1006 | ASSERT3U(error, ==, 0); |
34dc7c2f | 1007 | |
40b84e7a | 1008 | setpoint = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
428870ff BB |
1009 | VERIFY3U(dsl_prop_get(osname, propname, sizeof (curval), |
1010 | 1, &curval, setpoint), ==, 0); | |
1011 | ||
c242c188 | 1012 | if (ztest_opts.zo_verbose >= 6) { |
428870ff BB |
1013 | VERIFY(zfs_prop_index_to_string(prop, curval, &valname) == 0); |
1014 | (void) printf("%s %s = %s at '%s'\n", | |
1015 | osname, propname, valname, setpoint); | |
34dc7c2f | 1016 | } |
40b84e7a | 1017 | umem_free(setpoint, MAXPATHLEN); |
34dc7c2f BB |
1018 | |
1019 | return (error); | |
1020 | } | |
1021 | ||
1022 | static int | |
c242c188 | 1023 | ztest_spa_prop_set_uint64(zpool_prop_t prop, uint64_t value) |
34dc7c2f | 1024 | { |
c242c188 | 1025 | spa_t *spa = ztest_spa; |
428870ff | 1026 | nvlist_t *props = NULL; |
34dc7c2f BB |
1027 | int error; |
1028 | ||
428870ff BB |
1029 | VERIFY(nvlist_alloc(&props, NV_UNIQUE_NAME, 0) == 0); |
1030 | VERIFY(nvlist_add_uint64(props, zpool_prop_to_name(prop), value) == 0); | |
34dc7c2f | 1031 | |
428870ff BB |
1032 | error = spa_prop_set(spa, props); |
1033 | ||
1034 | nvlist_free(props); | |
1035 | ||
1036 | if (error == ENOSPC) { | |
1037 | ztest_record_enospc(FTAG); | |
34dc7c2f BB |
1038 | return (error); |
1039 | } | |
428870ff | 1040 | ASSERT3U(error, ==, 0); |
34dc7c2f BB |
1041 | |
1042 | return (error); | |
1043 | } | |
1044 | ||
428870ff BB |
1045 | static void |
1046 | ztest_rll_init(rll_t *rll) | |
1047 | { | |
1048 | rll->rll_writer = NULL; | |
1049 | rll->rll_readers = 0; | |
1e33ac1e BB |
1050 | mutex_init(&rll->rll_lock, NULL, MUTEX_DEFAULT, NULL); |
1051 | cv_init(&rll->rll_cv, NULL, CV_DEFAULT, NULL); | |
428870ff | 1052 | } |
34dc7c2f | 1053 | |
428870ff BB |
1054 | static void |
1055 | ztest_rll_destroy(rll_t *rll) | |
34dc7c2f | 1056 | { |
428870ff BB |
1057 | ASSERT(rll->rll_writer == NULL); |
1058 | ASSERT(rll->rll_readers == 0); | |
1e33ac1e BB |
1059 | mutex_destroy(&rll->rll_lock); |
1060 | cv_destroy(&rll->rll_cv); | |
428870ff | 1061 | } |
34dc7c2f | 1062 | |
428870ff BB |
1063 | static void |
1064 | ztest_rll_lock(rll_t *rll, rl_type_t type) | |
1065 | { | |
1e33ac1e | 1066 | mutex_enter(&rll->rll_lock); |
34dc7c2f | 1067 | |
428870ff BB |
1068 | if (type == RL_READER) { |
1069 | while (rll->rll_writer != NULL) | |
1e33ac1e | 1070 | (void) cv_wait(&rll->rll_cv, &rll->rll_lock); |
428870ff BB |
1071 | rll->rll_readers++; |
1072 | } else { | |
1073 | while (rll->rll_writer != NULL || rll->rll_readers) | |
1e33ac1e | 1074 | (void) cv_wait(&rll->rll_cv, &rll->rll_lock); |
428870ff BB |
1075 | rll->rll_writer = curthread; |
1076 | } | |
34dc7c2f | 1077 | |
1e33ac1e | 1078 | mutex_exit(&rll->rll_lock); |
428870ff | 1079 | } |
34dc7c2f | 1080 | |
428870ff BB |
1081 | static void |
1082 | ztest_rll_unlock(rll_t *rll) | |
1083 | { | |
1e33ac1e | 1084 | mutex_enter(&rll->rll_lock); |
34dc7c2f | 1085 | |
428870ff BB |
1086 | if (rll->rll_writer) { |
1087 | ASSERT(rll->rll_readers == 0); | |
1088 | rll->rll_writer = NULL; | |
1089 | } else { | |
1090 | ASSERT(rll->rll_readers != 0); | |
1091 | ASSERT(rll->rll_writer == NULL); | |
1092 | rll->rll_readers--; | |
1093 | } | |
34dc7c2f | 1094 | |
428870ff | 1095 | if (rll->rll_writer == NULL && rll->rll_readers == 0) |
1e33ac1e | 1096 | cv_broadcast(&rll->rll_cv); |
428870ff | 1097 | |
1e33ac1e | 1098 | mutex_exit(&rll->rll_lock); |
34dc7c2f BB |
1099 | } |
1100 | ||
428870ff BB |
1101 | static void |
1102 | ztest_object_lock(ztest_ds_t *zd, uint64_t object, rl_type_t type) | |
b128c09f | 1103 | { |
428870ff | 1104 | rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; |
b128c09f | 1105 | |
428870ff BB |
1106 | ztest_rll_lock(rll, type); |
1107 | } | |
b128c09f | 1108 | |
428870ff BB |
1109 | static void |
1110 | ztest_object_unlock(ztest_ds_t *zd, uint64_t object) | |
1111 | { | |
1112 | rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)]; | |
b128c09f | 1113 | |
428870ff | 1114 | ztest_rll_unlock(rll); |
b128c09f BB |
1115 | } |
1116 | ||
428870ff BB |
1117 | static rl_t * |
1118 | ztest_range_lock(ztest_ds_t *zd, uint64_t object, uint64_t offset, | |
1119 | uint64_t size, rl_type_t type) | |
34dc7c2f | 1120 | { |
428870ff BB |
1121 | uint64_t hash = object ^ (offset % (ZTEST_RANGE_LOCKS + 1)); |
1122 | rll_t *rll = &zd->zd_range_lock[hash & (ZTEST_RANGE_LOCKS - 1)]; | |
1123 | rl_t *rl; | |
34dc7c2f | 1124 | |
428870ff BB |
1125 | rl = umem_alloc(sizeof (*rl), UMEM_NOFAIL); |
1126 | rl->rl_object = object; | |
1127 | rl->rl_offset = offset; | |
1128 | rl->rl_size = size; | |
1129 | rl->rl_lock = rll; | |
34dc7c2f | 1130 | |
428870ff BB |
1131 | ztest_rll_lock(rll, type); |
1132 | ||
1133 | return (rl); | |
1134 | } | |
34dc7c2f | 1135 | |
428870ff BB |
1136 | static void |
1137 | ztest_range_unlock(rl_t *rl) | |
1138 | { | |
1139 | rll_t *rll = rl->rl_lock; | |
34dc7c2f | 1140 | |
428870ff | 1141 | ztest_rll_unlock(rll); |
34dc7c2f | 1142 | |
428870ff BB |
1143 | umem_free(rl, sizeof (*rl)); |
1144 | } | |
34dc7c2f | 1145 | |
428870ff | 1146 | static void |
c242c188 | 1147 | ztest_zd_init(ztest_ds_t *zd, ztest_shared_ds_t *szd, objset_t *os) |
428870ff BB |
1148 | { |
1149 | zd->zd_os = os; | |
1150 | zd->zd_zilog = dmu_objset_zil(os); | |
c242c188 | 1151 | zd->zd_shared = szd; |
428870ff | 1152 | dmu_objset_name(os, zd->zd_name); |
d6320ddb | 1153 | int l; |
428870ff | 1154 | |
c242c188 CS |
1155 | if (zd->zd_shared != NULL) |
1156 | zd->zd_shared->zd_seq = 0; | |
1157 | ||
3e31d2b0 | 1158 | rw_init(&zd->zd_zilog_lock, NULL, RW_DEFAULT, NULL); |
1e33ac1e | 1159 | mutex_init(&zd->zd_dirobj_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 1160 | |
d6320ddb | 1161 | for (l = 0; l < ZTEST_OBJECT_LOCKS; l++) |
428870ff | 1162 | ztest_rll_init(&zd->zd_object_lock[l]); |
34dc7c2f | 1163 | |
d6320ddb | 1164 | for (l = 0; l < ZTEST_RANGE_LOCKS; l++) |
428870ff | 1165 | ztest_rll_init(&zd->zd_range_lock[l]); |
34dc7c2f BB |
1166 | } |
1167 | ||
428870ff BB |
1168 | static void |
1169 | ztest_zd_fini(ztest_ds_t *zd) | |
34dc7c2f | 1170 | { |
d6320ddb BB |
1171 | int l; |
1172 | ||
1e33ac1e | 1173 | mutex_destroy(&zd->zd_dirobj_lock); |
3e31d2b0 | 1174 | rw_destroy(&zd->zd_zilog_lock); |
34dc7c2f | 1175 | |
d6320ddb | 1176 | for (l = 0; l < ZTEST_OBJECT_LOCKS; l++) |
428870ff | 1177 | ztest_rll_destroy(&zd->zd_object_lock[l]); |
b128c09f | 1178 | |
d6320ddb | 1179 | for (l = 0; l < ZTEST_RANGE_LOCKS; l++) |
428870ff BB |
1180 | ztest_rll_destroy(&zd->zd_range_lock[l]); |
1181 | } | |
b128c09f | 1182 | |
428870ff | 1183 | #define TXG_MIGHTWAIT (ztest_random(10) == 0 ? TXG_NOWAIT : TXG_WAIT) |
b128c09f | 1184 | |
428870ff BB |
1185 | static uint64_t |
1186 | ztest_tx_assign(dmu_tx_t *tx, uint64_t txg_how, const char *tag) | |
1187 | { | |
1188 | uint64_t txg; | |
1189 | int error; | |
1190 | ||
1191 | /* | |
1192 | * Attempt to assign tx to some transaction group. | |
1193 | */ | |
1194 | error = dmu_tx_assign(tx, txg_how); | |
1195 | if (error) { | |
1196 | if (error == ERESTART) { | |
1197 | ASSERT(txg_how == TXG_NOWAIT); | |
1198 | dmu_tx_wait(tx); | |
1199 | } else { | |
1200 | ASSERT3U(error, ==, ENOSPC); | |
1201 | ztest_record_enospc(tag); | |
1202 | } | |
1203 | dmu_tx_abort(tx); | |
1204 | return (0); | |
1205 | } | |
1206 | txg = dmu_tx_get_txg(tx); | |
1207 | ASSERT(txg != 0); | |
1208 | return (txg); | |
1209 | } | |
1210 | ||
1211 | static void | |
1212 | ztest_pattern_set(void *buf, uint64_t size, uint64_t value) | |
1213 | { | |
1214 | uint64_t *ip = buf; | |
1215 | uint64_t *ip_end = (uint64_t *)((uintptr_t)buf + (uintptr_t)size); | |
1216 | ||
1217 | while (ip < ip_end) | |
1218 | *ip++ = value; | |
1219 | } | |
1220 | ||
1fde1e37 | 1221 | #ifndef NDEBUG |
428870ff BB |
1222 | static boolean_t |
1223 | ztest_pattern_match(void *buf, uint64_t size, uint64_t value) | |
1224 | { | |
1225 | uint64_t *ip = buf; | |
1226 | uint64_t *ip_end = (uint64_t *)((uintptr_t)buf + (uintptr_t)size); | |
1227 | uint64_t diff = 0; | |
1228 | ||
1229 | while (ip < ip_end) | |
1230 | diff |= (value - *ip++); | |
1231 | ||
1232 | return (diff == 0); | |
1233 | } | |
1fde1e37 | 1234 | #endif |
428870ff BB |
1235 | |
1236 | static void | |
1237 | ztest_bt_generate(ztest_block_tag_t *bt, objset_t *os, uint64_t object, | |
1238 | uint64_t offset, uint64_t gen, uint64_t txg, uint64_t crtxg) | |
1239 | { | |
1240 | bt->bt_magic = BT_MAGIC; | |
1241 | bt->bt_objset = dmu_objset_id(os); | |
1242 | bt->bt_object = object; | |
1243 | bt->bt_offset = offset; | |
1244 | bt->bt_gen = gen; | |
1245 | bt->bt_txg = txg; | |
1246 | bt->bt_crtxg = crtxg; | |
1247 | } | |
1248 | ||
1249 | static void | |
1250 | ztest_bt_verify(ztest_block_tag_t *bt, objset_t *os, uint64_t object, | |
1251 | uint64_t offset, uint64_t gen, uint64_t txg, uint64_t crtxg) | |
1252 | { | |
1253 | ASSERT(bt->bt_magic == BT_MAGIC); | |
1254 | ASSERT(bt->bt_objset == dmu_objset_id(os)); | |
1255 | ASSERT(bt->bt_object == object); | |
1256 | ASSERT(bt->bt_offset == offset); | |
1257 | ASSERT(bt->bt_gen <= gen); | |
1258 | ASSERT(bt->bt_txg <= txg); | |
1259 | ASSERT(bt->bt_crtxg == crtxg); | |
1260 | } | |
1261 | ||
1262 | static ztest_block_tag_t * | |
1263 | ztest_bt_bonus(dmu_buf_t *db) | |
1264 | { | |
1265 | dmu_object_info_t doi; | |
1266 | ztest_block_tag_t *bt; | |
1267 | ||
1268 | dmu_object_info_from_db(db, &doi); | |
1269 | ASSERT3U(doi.doi_bonus_size, <=, db->db_size); | |
1270 | ASSERT3U(doi.doi_bonus_size, >=, sizeof (*bt)); | |
1271 | bt = (void *)((char *)db->db_data + doi.doi_bonus_size - sizeof (*bt)); | |
1272 | ||
1273 | return (bt); | |
1274 | } | |
1275 | ||
1276 | /* | |
1277 | * ZIL logging ops | |
1278 | */ | |
1279 | ||
1280 | #define lrz_type lr_mode | |
1281 | #define lrz_blocksize lr_uid | |
1282 | #define lrz_ibshift lr_gid | |
1283 | #define lrz_bonustype lr_rdev | |
1284 | #define lrz_bonuslen lr_crtime[1] | |
1285 | ||
572e2857 | 1286 | static void |
428870ff BB |
1287 | ztest_log_create(ztest_ds_t *zd, dmu_tx_t *tx, lr_create_t *lr) |
1288 | { | |
1289 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1290 | size_t namesize = strlen(name) + 1; | |
1291 | itx_t *itx; | |
1292 | ||
1293 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1294 | return; |
428870ff BB |
1295 | |
1296 | itx = zil_itx_create(TX_CREATE, sizeof (*lr) + namesize); | |
1297 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1298 | sizeof (*lr) + namesize - sizeof (lr_t)); | |
1299 | ||
572e2857 | 1300 | zil_itx_assign(zd->zd_zilog, itx, tx); |
428870ff BB |
1301 | } |
1302 | ||
572e2857 BB |
1303 | static void |
1304 | ztest_log_remove(ztest_ds_t *zd, dmu_tx_t *tx, lr_remove_t *lr, uint64_t object) | |
428870ff BB |
1305 | { |
1306 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1307 | size_t namesize = strlen(name) + 1; | |
1308 | itx_t *itx; | |
1309 | ||
1310 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1311 | return; |
428870ff BB |
1312 | |
1313 | itx = zil_itx_create(TX_REMOVE, sizeof (*lr) + namesize); | |
1314 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1315 | sizeof (*lr) + namesize - sizeof (lr_t)); | |
1316 | ||
572e2857 BB |
1317 | itx->itx_oid = object; |
1318 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1319 | } |
1320 | ||
572e2857 | 1321 | static void |
428870ff BB |
1322 | ztest_log_write(ztest_ds_t *zd, dmu_tx_t *tx, lr_write_t *lr) |
1323 | { | |
1324 | itx_t *itx; | |
1325 | itx_wr_state_t write_state = ztest_random(WR_NUM_STATES); | |
1326 | ||
1327 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1328 | return; |
428870ff BB |
1329 | |
1330 | if (lr->lr_length > ZIL_MAX_LOG_DATA) | |
1331 | write_state = WR_INDIRECT; | |
1332 | ||
1333 | itx = zil_itx_create(TX_WRITE, | |
1334 | sizeof (*lr) + (write_state == WR_COPIED ? lr->lr_length : 0)); | |
1335 | ||
1336 | if (write_state == WR_COPIED && | |
1337 | dmu_read(zd->zd_os, lr->lr_foid, lr->lr_offset, lr->lr_length, | |
1338 | ((lr_write_t *)&itx->itx_lr) + 1, DMU_READ_NO_PREFETCH) != 0) { | |
1339 | zil_itx_destroy(itx); | |
1340 | itx = zil_itx_create(TX_WRITE, sizeof (*lr)); | |
1341 | write_state = WR_NEED_COPY; | |
1342 | } | |
1343 | itx->itx_private = zd; | |
1344 | itx->itx_wr_state = write_state; | |
1345 | itx->itx_sync = (ztest_random(8) == 0); | |
1346 | itx->itx_sod += (write_state == WR_NEED_COPY ? lr->lr_length : 0); | |
1347 | ||
1348 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1349 | sizeof (*lr) - sizeof (lr_t)); | |
1350 | ||
572e2857 | 1351 | zil_itx_assign(zd->zd_zilog, itx, tx); |
428870ff BB |
1352 | } |
1353 | ||
572e2857 | 1354 | static void |
428870ff BB |
1355 | ztest_log_truncate(ztest_ds_t *zd, dmu_tx_t *tx, lr_truncate_t *lr) |
1356 | { | |
1357 | itx_t *itx; | |
1358 | ||
1359 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1360 | return; |
428870ff BB |
1361 | |
1362 | itx = zil_itx_create(TX_TRUNCATE, sizeof (*lr)); | |
1363 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1364 | sizeof (*lr) - sizeof (lr_t)); | |
1365 | ||
572e2857 BB |
1366 | itx->itx_sync = B_FALSE; |
1367 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1368 | } |
1369 | ||
572e2857 | 1370 | static void |
428870ff BB |
1371 | ztest_log_setattr(ztest_ds_t *zd, dmu_tx_t *tx, lr_setattr_t *lr) |
1372 | { | |
1373 | itx_t *itx; | |
1374 | ||
1375 | if (zil_replaying(zd->zd_zilog, tx)) | |
572e2857 | 1376 | return; |
428870ff BB |
1377 | |
1378 | itx = zil_itx_create(TX_SETATTR, sizeof (*lr)); | |
1379 | bcopy(&lr->lr_common + 1, &itx->itx_lr + 1, | |
1380 | sizeof (*lr) - sizeof (lr_t)); | |
1381 | ||
572e2857 BB |
1382 | itx->itx_sync = B_FALSE; |
1383 | zil_itx_assign(zd->zd_zilog, itx, tx); | |
428870ff BB |
1384 | } |
1385 | ||
1386 | /* | |
1387 | * ZIL replay ops | |
1388 | */ | |
1389 | static int | |
1390 | ztest_replay_create(ztest_ds_t *zd, lr_create_t *lr, boolean_t byteswap) | |
1391 | { | |
1392 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1393 | objset_t *os = zd->zd_os; | |
1394 | ztest_block_tag_t *bbt; | |
1395 | dmu_buf_t *db; | |
1396 | dmu_tx_t *tx; | |
1397 | uint64_t txg; | |
1398 | int error = 0; | |
1399 | ||
1400 | if (byteswap) | |
1401 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1402 | ||
1403 | ASSERT(lr->lr_doid == ZTEST_DIROBJ); | |
1404 | ASSERT(name[0] != '\0'); | |
1405 | ||
1406 | tx = dmu_tx_create(os); | |
1407 | ||
1408 | dmu_tx_hold_zap(tx, lr->lr_doid, B_TRUE, name); | |
1409 | ||
1410 | if (lr->lrz_type == DMU_OT_ZAP_OTHER) { | |
1411 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, B_TRUE, NULL); | |
1412 | } else { | |
1413 | dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT); | |
1414 | } | |
1415 | ||
1416 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1417 | if (txg == 0) | |
1418 | return (ENOSPC); | |
1419 | ||
1420 | ASSERT(dmu_objset_zil(os)->zl_replay == !!lr->lr_foid); | |
1421 | ||
1422 | if (lr->lrz_type == DMU_OT_ZAP_OTHER) { | |
1423 | if (lr->lr_foid == 0) { | |
1424 | lr->lr_foid = zap_create(os, | |
1425 | lr->lrz_type, lr->lrz_bonustype, | |
1426 | lr->lrz_bonuslen, tx); | |
1427 | } else { | |
1428 | error = zap_create_claim(os, lr->lr_foid, | |
1429 | lr->lrz_type, lr->lrz_bonustype, | |
1430 | lr->lrz_bonuslen, tx); | |
1431 | } | |
1432 | } else { | |
1433 | if (lr->lr_foid == 0) { | |
1434 | lr->lr_foid = dmu_object_alloc(os, | |
1435 | lr->lrz_type, 0, lr->lrz_bonustype, | |
1436 | lr->lrz_bonuslen, tx); | |
1437 | } else { | |
1438 | error = dmu_object_claim(os, lr->lr_foid, | |
1439 | lr->lrz_type, 0, lr->lrz_bonustype, | |
1440 | lr->lrz_bonuslen, tx); | |
1441 | } | |
1442 | } | |
1443 | ||
1444 | if (error) { | |
1445 | ASSERT3U(error, ==, EEXIST); | |
1446 | ASSERT(zd->zd_zilog->zl_replay); | |
1447 | dmu_tx_commit(tx); | |
1448 | return (error); | |
1449 | } | |
1450 | ||
1451 | ASSERT(lr->lr_foid != 0); | |
1452 | ||
1453 | if (lr->lrz_type != DMU_OT_ZAP_OTHER) | |
1454 | VERIFY3U(0, ==, dmu_object_set_blocksize(os, lr->lr_foid, | |
1455 | lr->lrz_blocksize, lr->lrz_ibshift, tx)); | |
1456 | ||
1457 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
1458 | bbt = ztest_bt_bonus(db); | |
1459 | dmu_buf_will_dirty(db, tx); | |
1460 | ztest_bt_generate(bbt, os, lr->lr_foid, -1ULL, lr->lr_gen, txg, txg); | |
1461 | dmu_buf_rele(db, FTAG); | |
1462 | ||
1463 | VERIFY3U(0, ==, zap_add(os, lr->lr_doid, name, sizeof (uint64_t), 1, | |
1464 | &lr->lr_foid, tx)); | |
1465 | ||
1466 | (void) ztest_log_create(zd, tx, lr); | |
1467 | ||
1468 | dmu_tx_commit(tx); | |
1469 | ||
1470 | return (0); | |
1471 | } | |
1472 | ||
1473 | static int | |
1474 | ztest_replay_remove(ztest_ds_t *zd, lr_remove_t *lr, boolean_t byteswap) | |
1475 | { | |
1476 | char *name = (void *)(lr + 1); /* name follows lr */ | |
1477 | objset_t *os = zd->zd_os; | |
1478 | dmu_object_info_t doi; | |
1479 | dmu_tx_t *tx; | |
1480 | uint64_t object, txg; | |
1481 | ||
1482 | if (byteswap) | |
1483 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1484 | ||
1485 | ASSERT(lr->lr_doid == ZTEST_DIROBJ); | |
1486 | ASSERT(name[0] != '\0'); | |
1487 | ||
1488 | VERIFY3U(0, ==, | |
1489 | zap_lookup(os, lr->lr_doid, name, sizeof (object), 1, &object)); | |
1490 | ASSERT(object != 0); | |
1491 | ||
1492 | ztest_object_lock(zd, object, RL_WRITER); | |
1493 | ||
1494 | VERIFY3U(0, ==, dmu_object_info(os, object, &doi)); | |
1495 | ||
1496 | tx = dmu_tx_create(os); | |
1497 | ||
1498 | dmu_tx_hold_zap(tx, lr->lr_doid, B_FALSE, name); | |
1499 | dmu_tx_hold_free(tx, object, 0, DMU_OBJECT_END); | |
1500 | ||
1501 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1502 | if (txg == 0) { | |
1503 | ztest_object_unlock(zd, object); | |
1504 | return (ENOSPC); | |
1505 | } | |
1506 | ||
1507 | if (doi.doi_type == DMU_OT_ZAP_OTHER) { | |
1508 | VERIFY3U(0, ==, zap_destroy(os, object, tx)); | |
1509 | } else { | |
1510 | VERIFY3U(0, ==, dmu_object_free(os, object, tx)); | |
1511 | } | |
1512 | ||
1513 | VERIFY3U(0, ==, zap_remove(os, lr->lr_doid, name, tx)); | |
1514 | ||
572e2857 | 1515 | (void) ztest_log_remove(zd, tx, lr, object); |
428870ff BB |
1516 | |
1517 | dmu_tx_commit(tx); | |
1518 | ||
1519 | ztest_object_unlock(zd, object); | |
1520 | ||
1521 | return (0); | |
1522 | } | |
1523 | ||
1524 | static int | |
1525 | ztest_replay_write(ztest_ds_t *zd, lr_write_t *lr, boolean_t byteswap) | |
1526 | { | |
1527 | objset_t *os = zd->zd_os; | |
1528 | void *data = lr + 1; /* data follows lr */ | |
1529 | uint64_t offset, length; | |
1530 | ztest_block_tag_t *bt = data; | |
1531 | ztest_block_tag_t *bbt; | |
1532 | uint64_t gen, txg, lrtxg, crtxg; | |
1533 | dmu_object_info_t doi; | |
1534 | dmu_tx_t *tx; | |
1535 | dmu_buf_t *db; | |
1536 | arc_buf_t *abuf = NULL; | |
1537 | rl_t *rl; | |
1538 | ||
1539 | if (byteswap) | |
1540 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1541 | ||
1542 | offset = lr->lr_offset; | |
1543 | length = lr->lr_length; | |
1544 | ||
1545 | /* If it's a dmu_sync() block, write the whole block */ | |
1546 | if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) { | |
1547 | uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr); | |
1548 | if (length < blocksize) { | |
1549 | offset -= offset % blocksize; | |
1550 | length = blocksize; | |
1551 | } | |
1552 | } | |
1553 | ||
1554 | if (bt->bt_magic == BSWAP_64(BT_MAGIC)) | |
1555 | byteswap_uint64_array(bt, sizeof (*bt)); | |
1556 | ||
1557 | if (bt->bt_magic != BT_MAGIC) | |
1558 | bt = NULL; | |
1559 | ||
1560 | ztest_object_lock(zd, lr->lr_foid, RL_READER); | |
1561 | rl = ztest_range_lock(zd, lr->lr_foid, offset, length, RL_WRITER); | |
1562 | ||
1563 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
1564 | ||
1565 | dmu_object_info_from_db(db, &doi); | |
1566 | ||
1567 | bbt = ztest_bt_bonus(db); | |
1568 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
1569 | gen = bbt->bt_gen; | |
1570 | crtxg = bbt->bt_crtxg; | |
1571 | lrtxg = lr->lr_common.lrc_txg; | |
1572 | ||
1573 | tx = dmu_tx_create(os); | |
1574 | ||
1575 | dmu_tx_hold_write(tx, lr->lr_foid, offset, length); | |
1576 | ||
1577 | if (ztest_random(8) == 0 && length == doi.doi_data_block_size && | |
1578 | P2PHASE(offset, length) == 0) | |
1579 | abuf = dmu_request_arcbuf(db, length); | |
1580 | ||
1581 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1582 | if (txg == 0) { | |
1583 | if (abuf != NULL) | |
1584 | dmu_return_arcbuf(abuf); | |
1585 | dmu_buf_rele(db, FTAG); | |
1586 | ztest_range_unlock(rl); | |
1587 | ztest_object_unlock(zd, lr->lr_foid); | |
1588 | return (ENOSPC); | |
1589 | } | |
1590 | ||
1591 | if (bt != NULL) { | |
1592 | /* | |
1593 | * Usually, verify the old data before writing new data -- | |
1594 | * but not always, because we also want to verify correct | |
1595 | * behavior when the data was not recently read into cache. | |
1596 | */ | |
1597 | ASSERT(offset % doi.doi_data_block_size == 0); | |
1598 | if (ztest_random(4) != 0) { | |
1599 | int prefetch = ztest_random(2) ? | |
1600 | DMU_READ_PREFETCH : DMU_READ_NO_PREFETCH; | |
1601 | ztest_block_tag_t rbt; | |
1602 | ||
1603 | VERIFY(dmu_read(os, lr->lr_foid, offset, | |
1604 | sizeof (rbt), &rbt, prefetch) == 0); | |
1605 | if (rbt.bt_magic == BT_MAGIC) { | |
1606 | ztest_bt_verify(&rbt, os, lr->lr_foid, | |
1607 | offset, gen, txg, crtxg); | |
1608 | } | |
1609 | } | |
1610 | ||
1611 | /* | |
1612 | * Writes can appear to be newer than the bonus buffer because | |
1613 | * the ztest_get_data() callback does a dmu_read() of the | |
1614 | * open-context data, which may be different than the data | |
1615 | * as it was when the write was generated. | |
1616 | */ | |
1617 | if (zd->zd_zilog->zl_replay) { | |
1618 | ztest_bt_verify(bt, os, lr->lr_foid, offset, | |
1619 | MAX(gen, bt->bt_gen), MAX(txg, lrtxg), | |
1620 | bt->bt_crtxg); | |
1621 | } | |
1622 | ||
1623 | /* | |
1624 | * Set the bt's gen/txg to the bonus buffer's gen/txg | |
1625 | * so that all of the usual ASSERTs will work. | |
1626 | */ | |
1627 | ztest_bt_generate(bt, os, lr->lr_foid, offset, gen, txg, crtxg); | |
1628 | } | |
1629 | ||
1630 | if (abuf == NULL) { | |
1631 | dmu_write(os, lr->lr_foid, offset, length, data, tx); | |
1632 | } else { | |
1633 | bcopy(data, abuf->b_data, length); | |
1634 | dmu_assign_arcbuf(db, offset, abuf, tx); | |
1635 | } | |
1636 | ||
1637 | (void) ztest_log_write(zd, tx, lr); | |
1638 | ||
1639 | dmu_buf_rele(db, FTAG); | |
1640 | ||
1641 | dmu_tx_commit(tx); | |
1642 | ||
1643 | ztest_range_unlock(rl); | |
1644 | ztest_object_unlock(zd, lr->lr_foid); | |
1645 | ||
1646 | return (0); | |
1647 | } | |
1648 | ||
1649 | static int | |
1650 | ztest_replay_truncate(ztest_ds_t *zd, lr_truncate_t *lr, boolean_t byteswap) | |
1651 | { | |
1652 | objset_t *os = zd->zd_os; | |
1653 | dmu_tx_t *tx; | |
1654 | uint64_t txg; | |
1655 | rl_t *rl; | |
1656 | ||
1657 | if (byteswap) | |
1658 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1659 | ||
1660 | ztest_object_lock(zd, lr->lr_foid, RL_READER); | |
1661 | rl = ztest_range_lock(zd, lr->lr_foid, lr->lr_offset, lr->lr_length, | |
1662 | RL_WRITER); | |
1663 | ||
1664 | tx = dmu_tx_create(os); | |
1665 | ||
1666 | dmu_tx_hold_free(tx, lr->lr_foid, lr->lr_offset, lr->lr_length); | |
1667 | ||
1668 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1669 | if (txg == 0) { | |
1670 | ztest_range_unlock(rl); | |
1671 | ztest_object_unlock(zd, lr->lr_foid); | |
1672 | return (ENOSPC); | |
1673 | } | |
1674 | ||
1675 | VERIFY(dmu_free_range(os, lr->lr_foid, lr->lr_offset, | |
1676 | lr->lr_length, tx) == 0); | |
1677 | ||
1678 | (void) ztest_log_truncate(zd, tx, lr); | |
1679 | ||
1680 | dmu_tx_commit(tx); | |
1681 | ||
1682 | ztest_range_unlock(rl); | |
1683 | ztest_object_unlock(zd, lr->lr_foid); | |
1684 | ||
1685 | return (0); | |
1686 | } | |
1687 | ||
1688 | static int | |
1689 | ztest_replay_setattr(ztest_ds_t *zd, lr_setattr_t *lr, boolean_t byteswap) | |
1690 | { | |
1691 | objset_t *os = zd->zd_os; | |
1692 | dmu_tx_t *tx; | |
1693 | dmu_buf_t *db; | |
1694 | ztest_block_tag_t *bbt; | |
1695 | uint64_t txg, lrtxg, crtxg; | |
1696 | ||
1697 | if (byteswap) | |
1698 | byteswap_uint64_array(lr, sizeof (*lr)); | |
1699 | ||
1700 | ztest_object_lock(zd, lr->lr_foid, RL_WRITER); | |
1701 | ||
1702 | VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db)); | |
1703 | ||
1704 | tx = dmu_tx_create(os); | |
1705 | dmu_tx_hold_bonus(tx, lr->lr_foid); | |
1706 | ||
1707 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
1708 | if (txg == 0) { | |
1709 | dmu_buf_rele(db, FTAG); | |
1710 | ztest_object_unlock(zd, lr->lr_foid); | |
1711 | return (ENOSPC); | |
1712 | } | |
1713 | ||
1714 | bbt = ztest_bt_bonus(db); | |
1715 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
1716 | crtxg = bbt->bt_crtxg; | |
1717 | lrtxg = lr->lr_common.lrc_txg; | |
1718 | ||
1719 | if (zd->zd_zilog->zl_replay) { | |
1720 | ASSERT(lr->lr_size != 0); | |
1721 | ASSERT(lr->lr_mode != 0); | |
1722 | ASSERT(lrtxg != 0); | |
1723 | } else { | |
1724 | /* | |
1725 | * Randomly change the size and increment the generation. | |
1726 | */ | |
1727 | lr->lr_size = (ztest_random(db->db_size / sizeof (*bbt)) + 1) * | |
1728 | sizeof (*bbt); | |
1729 | lr->lr_mode = bbt->bt_gen + 1; | |
1730 | ASSERT(lrtxg == 0); | |
1731 | } | |
1732 | ||
1733 | /* | |
1734 | * Verify that the current bonus buffer is not newer than our txg. | |
1735 | */ | |
1736 | ztest_bt_verify(bbt, os, lr->lr_foid, -1ULL, lr->lr_mode, | |
1737 | MAX(txg, lrtxg), crtxg); | |
1738 | ||
1739 | dmu_buf_will_dirty(db, tx); | |
1740 | ||
1741 | ASSERT3U(lr->lr_size, >=, sizeof (*bbt)); | |
1742 | ASSERT3U(lr->lr_size, <=, db->db_size); | |
1743 | VERIFY3U(dmu_set_bonus(db, lr->lr_size, tx), ==, 0); | |
1744 | bbt = ztest_bt_bonus(db); | |
1745 | ||
1746 | ztest_bt_generate(bbt, os, lr->lr_foid, -1ULL, lr->lr_mode, txg, crtxg); | |
1747 | ||
1748 | dmu_buf_rele(db, FTAG); | |
1749 | ||
1750 | (void) ztest_log_setattr(zd, tx, lr); | |
1751 | ||
1752 | dmu_tx_commit(tx); | |
1753 | ||
1754 | ztest_object_unlock(zd, lr->lr_foid); | |
1755 | ||
1756 | return (0); | |
1757 | } | |
1758 | ||
1759 | zil_replay_func_t *ztest_replay_vector[TX_MAX_TYPE] = { | |
0bc8fd78 BB |
1760 | NULL, /* 0 no such transaction type */ |
1761 | (zil_replay_func_t *)ztest_replay_create, /* TX_CREATE */ | |
1762 | NULL, /* TX_MKDIR */ | |
1763 | NULL, /* TX_MKXATTR */ | |
1764 | NULL, /* TX_SYMLINK */ | |
1765 | (zil_replay_func_t *)ztest_replay_remove, /* TX_REMOVE */ | |
1766 | NULL, /* TX_RMDIR */ | |
1767 | NULL, /* TX_LINK */ | |
1768 | NULL, /* TX_RENAME */ | |
1769 | (zil_replay_func_t *)ztest_replay_write, /* TX_WRITE */ | |
1770 | (zil_replay_func_t *)ztest_replay_truncate, /* TX_TRUNCATE */ | |
1771 | (zil_replay_func_t *)ztest_replay_setattr, /* TX_SETATTR */ | |
1772 | NULL, /* TX_ACL */ | |
1773 | NULL, /* TX_CREATE_ACL */ | |
1774 | NULL, /* TX_CREATE_ATTR */ | |
1775 | NULL, /* TX_CREATE_ACL_ATTR */ | |
1776 | NULL, /* TX_MKDIR_ACL */ | |
1777 | NULL, /* TX_MKDIR_ATTR */ | |
1778 | NULL, /* TX_MKDIR_ACL_ATTR */ | |
1779 | NULL, /* TX_WRITE2 */ | |
428870ff BB |
1780 | }; |
1781 | ||
1782 | /* | |
1783 | * ZIL get_data callbacks | |
1784 | */ | |
1785 | ||
1786 | static void | |
1787 | ztest_get_done(zgd_t *zgd, int error) | |
1788 | { | |
1789 | ztest_ds_t *zd = zgd->zgd_private; | |
1790 | uint64_t object = zgd->zgd_rl->rl_object; | |
1791 | ||
1792 | if (zgd->zgd_db) | |
1793 | dmu_buf_rele(zgd->zgd_db, zgd); | |
1794 | ||
1795 | ztest_range_unlock(zgd->zgd_rl); | |
1796 | ztest_object_unlock(zd, object); | |
1797 | ||
1798 | if (error == 0 && zgd->zgd_bp) | |
1799 | zil_add_block(zgd->zgd_zilog, zgd->zgd_bp); | |
1800 | ||
1801 | umem_free(zgd, sizeof (*zgd)); | |
1802 | } | |
1803 | ||
1804 | static int | |
1805 | ztest_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio) | |
1806 | { | |
1807 | ztest_ds_t *zd = arg; | |
1808 | objset_t *os = zd->zd_os; | |
1809 | uint64_t object = lr->lr_foid; | |
1810 | uint64_t offset = lr->lr_offset; | |
1811 | uint64_t size = lr->lr_length; | |
1812 | blkptr_t *bp = &lr->lr_blkptr; | |
1813 | uint64_t txg = lr->lr_common.lrc_txg; | |
1814 | uint64_t crtxg; | |
1815 | dmu_object_info_t doi; | |
1816 | dmu_buf_t *db; | |
1817 | zgd_t *zgd; | |
1818 | int error; | |
1819 | ||
1820 | ztest_object_lock(zd, object, RL_READER); | |
1821 | error = dmu_bonus_hold(os, object, FTAG, &db); | |
1822 | if (error) { | |
1823 | ztest_object_unlock(zd, object); | |
1824 | return (error); | |
1825 | } | |
1826 | ||
1827 | crtxg = ztest_bt_bonus(db)->bt_crtxg; | |
1828 | ||
1829 | if (crtxg == 0 || crtxg > txg) { | |
1830 | dmu_buf_rele(db, FTAG); | |
1831 | ztest_object_unlock(zd, object); | |
1832 | return (ENOENT); | |
1833 | } | |
1834 | ||
1835 | dmu_object_info_from_db(db, &doi); | |
1836 | dmu_buf_rele(db, FTAG); | |
1837 | db = NULL; | |
1838 | ||
1839 | zgd = umem_zalloc(sizeof (*zgd), UMEM_NOFAIL); | |
1840 | zgd->zgd_zilog = zd->zd_zilog; | |
1841 | zgd->zgd_private = zd; | |
1842 | ||
1843 | if (buf != NULL) { /* immediate write */ | |
1844 | zgd->zgd_rl = ztest_range_lock(zd, object, offset, size, | |
1845 | RL_READER); | |
1846 | ||
1847 | error = dmu_read(os, object, offset, size, buf, | |
1848 | DMU_READ_NO_PREFETCH); | |
1849 | ASSERT(error == 0); | |
1850 | } else { | |
1851 | size = doi.doi_data_block_size; | |
1852 | if (ISP2(size)) { | |
1853 | offset = P2ALIGN(offset, size); | |
1854 | } else { | |
1855 | ASSERT(offset < size); | |
1856 | offset = 0; | |
1857 | } | |
1858 | ||
1859 | zgd->zgd_rl = ztest_range_lock(zd, object, offset, size, | |
1860 | RL_READER); | |
1861 | ||
1862 | error = dmu_buf_hold(os, object, offset, zgd, &db, | |
1863 | DMU_READ_NO_PREFETCH); | |
1864 | ||
1865 | if (error == 0) { | |
1866 | zgd->zgd_db = db; | |
1867 | zgd->zgd_bp = bp; | |
1868 | ||
1869 | ASSERT(db->db_offset == offset); | |
1870 | ASSERT(db->db_size == size); | |
1871 | ||
1872 | error = dmu_sync(zio, lr->lr_common.lrc_txg, | |
1873 | ztest_get_done, zgd); | |
1874 | ||
1875 | if (error == 0) | |
1876 | return (0); | |
1877 | } | |
1878 | } | |
1879 | ||
1880 | ztest_get_done(zgd, error); | |
1881 | ||
1882 | return (error); | |
1883 | } | |
1884 | ||
1885 | static void * | |
1886 | ztest_lr_alloc(size_t lrsize, char *name) | |
1887 | { | |
1888 | char *lr; | |
1889 | size_t namesize = name ? strlen(name) + 1 : 0; | |
1890 | ||
1891 | lr = umem_zalloc(lrsize + namesize, UMEM_NOFAIL); | |
1892 | ||
1893 | if (name) | |
1894 | bcopy(name, lr + lrsize, namesize); | |
1895 | ||
1896 | return (lr); | |
1897 | } | |
1898 | ||
1899 | void | |
1900 | ztest_lr_free(void *lr, size_t lrsize, char *name) | |
1901 | { | |
1902 | size_t namesize = name ? strlen(name) + 1 : 0; | |
1903 | ||
1904 | umem_free(lr, lrsize + namesize); | |
1905 | } | |
1906 | ||
1907 | /* | |
1908 | * Lookup a bunch of objects. Returns the number of objects not found. | |
1909 | */ | |
1910 | static int | |
1911 | ztest_lookup(ztest_ds_t *zd, ztest_od_t *od, int count) | |
1912 | { | |
1913 | int missing = 0; | |
1914 | int error; | |
d6320ddb | 1915 | int i; |
428870ff | 1916 | |
1e33ac1e | 1917 | ASSERT(mutex_held(&zd->zd_dirobj_lock)); |
428870ff | 1918 | |
d6320ddb | 1919 | for (i = 0; i < count; i++, od++) { |
428870ff BB |
1920 | od->od_object = 0; |
1921 | error = zap_lookup(zd->zd_os, od->od_dir, od->od_name, | |
1922 | sizeof (uint64_t), 1, &od->od_object); | |
1923 | if (error) { | |
1924 | ASSERT(error == ENOENT); | |
1925 | ASSERT(od->od_object == 0); | |
1926 | missing++; | |
1927 | } else { | |
1928 | dmu_buf_t *db; | |
1929 | ztest_block_tag_t *bbt; | |
1930 | dmu_object_info_t doi; | |
1931 | ||
1932 | ASSERT(od->od_object != 0); | |
1933 | ASSERT(missing == 0); /* there should be no gaps */ | |
1934 | ||
1935 | ztest_object_lock(zd, od->od_object, RL_READER); | |
1936 | VERIFY3U(0, ==, dmu_bonus_hold(zd->zd_os, | |
1937 | od->od_object, FTAG, &db)); | |
1938 | dmu_object_info_from_db(db, &doi); | |
1939 | bbt = ztest_bt_bonus(db); | |
1940 | ASSERT3U(bbt->bt_magic, ==, BT_MAGIC); | |
1941 | od->od_type = doi.doi_type; | |
1942 | od->od_blocksize = doi.doi_data_block_size; | |
1943 | od->od_gen = bbt->bt_gen; | |
1944 | dmu_buf_rele(db, FTAG); | |
1945 | ztest_object_unlock(zd, od->od_object); | |
1946 | } | |
1947 | } | |
1948 | ||
1949 | return (missing); | |
1950 | } | |
1951 | ||
1952 | static int | |
1953 | ztest_create(ztest_ds_t *zd, ztest_od_t *od, int count) | |
1954 | { | |
1955 | int missing = 0; | |
d6320ddb | 1956 | int i; |
428870ff | 1957 | |
1e33ac1e | 1958 | ASSERT(mutex_held(&zd->zd_dirobj_lock)); |
428870ff | 1959 | |
d6320ddb | 1960 | for (i = 0; i < count; i++, od++) { |
428870ff BB |
1961 | if (missing) { |
1962 | od->od_object = 0; | |
1963 | missing++; | |
1964 | continue; | |
1965 | } | |
1966 | ||
1967 | lr_create_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name); | |
1968 | ||
1969 | lr->lr_doid = od->od_dir; | |
1970 | lr->lr_foid = 0; /* 0 to allocate, > 0 to claim */ | |
1971 | lr->lrz_type = od->od_crtype; | |
1972 | lr->lrz_blocksize = od->od_crblocksize; | |
1973 | lr->lrz_ibshift = ztest_random_ibshift(); | |
1974 | lr->lrz_bonustype = DMU_OT_UINT64_OTHER; | |
1975 | lr->lrz_bonuslen = dmu_bonus_max(); | |
1976 | lr->lr_gen = od->od_crgen; | |
1977 | lr->lr_crtime[0] = time(NULL); | |
1978 | ||
1979 | if (ztest_replay_create(zd, lr, B_FALSE) != 0) { | |
1980 | ASSERT(missing == 0); | |
1981 | od->od_object = 0; | |
1982 | missing++; | |
1983 | } else { | |
1984 | od->od_object = lr->lr_foid; | |
1985 | od->od_type = od->od_crtype; | |
1986 | od->od_blocksize = od->od_crblocksize; | |
1987 | od->od_gen = od->od_crgen; | |
1988 | ASSERT(od->od_object != 0); | |
1989 | } | |
1990 | ||
1991 | ztest_lr_free(lr, sizeof (*lr), od->od_name); | |
1992 | } | |
1993 | ||
1994 | return (missing); | |
1995 | } | |
1996 | ||
1997 | static int | |
1998 | ztest_remove(ztest_ds_t *zd, ztest_od_t *od, int count) | |
1999 | { | |
2000 | int missing = 0; | |
2001 | int error; | |
d6320ddb | 2002 | int i; |
428870ff | 2003 | |
1e33ac1e | 2004 | ASSERT(mutex_held(&zd->zd_dirobj_lock)); |
428870ff BB |
2005 | |
2006 | od += count - 1; | |
2007 | ||
d6320ddb | 2008 | for (i = count - 1; i >= 0; i--, od--) { |
428870ff BB |
2009 | if (missing) { |
2010 | missing++; | |
2011 | continue; | |
2012 | } | |
2013 | ||
2014 | if (od->od_object == 0) | |
2015 | continue; | |
2016 | ||
2017 | lr_remove_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name); | |
2018 | ||
2019 | lr->lr_doid = od->od_dir; | |
2020 | ||
2021 | if ((error = ztest_replay_remove(zd, lr, B_FALSE)) != 0) { | |
2022 | ASSERT3U(error, ==, ENOSPC); | |
2023 | missing++; | |
2024 | } else { | |
2025 | od->od_object = 0; | |
2026 | } | |
2027 | ztest_lr_free(lr, sizeof (*lr), od->od_name); | |
2028 | } | |
2029 | ||
2030 | return (missing); | |
2031 | } | |
2032 | ||
2033 | static int | |
2034 | ztest_write(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size, | |
2035 | void *data) | |
2036 | { | |
2037 | lr_write_t *lr; | |
2038 | int error; | |
2039 | ||
2040 | lr = ztest_lr_alloc(sizeof (*lr) + size, NULL); | |
2041 | ||
2042 | lr->lr_foid = object; | |
2043 | lr->lr_offset = offset; | |
2044 | lr->lr_length = size; | |
2045 | lr->lr_blkoff = 0; | |
2046 | BP_ZERO(&lr->lr_blkptr); | |
2047 | ||
2048 | bcopy(data, lr + 1, size); | |
2049 | ||
2050 | error = ztest_replay_write(zd, lr, B_FALSE); | |
2051 | ||
2052 | ztest_lr_free(lr, sizeof (*lr) + size, NULL); | |
2053 | ||
2054 | return (error); | |
2055 | } | |
2056 | ||
2057 | static int | |
2058 | ztest_truncate(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size) | |
2059 | { | |
2060 | lr_truncate_t *lr; | |
2061 | int error; | |
2062 | ||
2063 | lr = ztest_lr_alloc(sizeof (*lr), NULL); | |
2064 | ||
2065 | lr->lr_foid = object; | |
2066 | lr->lr_offset = offset; | |
2067 | lr->lr_length = size; | |
2068 | ||
2069 | error = ztest_replay_truncate(zd, lr, B_FALSE); | |
2070 | ||
2071 | ztest_lr_free(lr, sizeof (*lr), NULL); | |
2072 | ||
2073 | return (error); | |
2074 | } | |
2075 | ||
2076 | static int | |
2077 | ztest_setattr(ztest_ds_t *zd, uint64_t object) | |
2078 | { | |
2079 | lr_setattr_t *lr; | |
2080 | int error; | |
2081 | ||
2082 | lr = ztest_lr_alloc(sizeof (*lr), NULL); | |
2083 | ||
2084 | lr->lr_foid = object; | |
2085 | lr->lr_size = 0; | |
2086 | lr->lr_mode = 0; | |
2087 | ||
2088 | error = ztest_replay_setattr(zd, lr, B_FALSE); | |
2089 | ||
2090 | ztest_lr_free(lr, sizeof (*lr), NULL); | |
2091 | ||
2092 | return (error); | |
2093 | } | |
2094 | ||
2095 | static void | |
2096 | ztest_prealloc(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size) | |
2097 | { | |
2098 | objset_t *os = zd->zd_os; | |
2099 | dmu_tx_t *tx; | |
2100 | uint64_t txg; | |
2101 | rl_t *rl; | |
2102 | ||
2103 | txg_wait_synced(dmu_objset_pool(os), 0); | |
2104 | ||
2105 | ztest_object_lock(zd, object, RL_READER); | |
2106 | rl = ztest_range_lock(zd, object, offset, size, RL_WRITER); | |
2107 | ||
2108 | tx = dmu_tx_create(os); | |
2109 | ||
2110 | dmu_tx_hold_write(tx, object, offset, size); | |
2111 | ||
2112 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
2113 | ||
2114 | if (txg != 0) { | |
2115 | dmu_prealloc(os, object, offset, size, tx); | |
2116 | dmu_tx_commit(tx); | |
2117 | txg_wait_synced(dmu_objset_pool(os), txg); | |
2118 | } else { | |
2119 | (void) dmu_free_long_range(os, object, offset, size); | |
2120 | } | |
2121 | ||
2122 | ztest_range_unlock(rl); | |
2123 | ztest_object_unlock(zd, object); | |
2124 | } | |
2125 | ||
2126 | static void | |
2127 | ztest_io(ztest_ds_t *zd, uint64_t object, uint64_t offset) | |
2128 | { | |
2129 | ztest_block_tag_t wbt; | |
2130 | dmu_object_info_t doi; | |
2131 | enum ztest_io_type io_type; | |
2132 | uint64_t blocksize; | |
2133 | void *data; | |
2134 | ||
2135 | VERIFY(dmu_object_info(zd->zd_os, object, &doi) == 0); | |
2136 | blocksize = doi.doi_data_block_size; | |
2137 | data = umem_alloc(blocksize, UMEM_NOFAIL); | |
2138 | ||
2139 | /* | |
2140 | * Pick an i/o type at random, biased toward writing block tags. | |
2141 | */ | |
2142 | io_type = ztest_random(ZTEST_IO_TYPES); | |
2143 | if (ztest_random(2) == 0) | |
2144 | io_type = ZTEST_IO_WRITE_TAG; | |
2145 | ||
3e31d2b0 ES |
2146 | (void) rw_enter(&zd->zd_zilog_lock, RW_READER); |
2147 | ||
428870ff BB |
2148 | switch (io_type) { |
2149 | ||
2150 | case ZTEST_IO_WRITE_TAG: | |
2151 | ztest_bt_generate(&wbt, zd->zd_os, object, offset, 0, 0, 0); | |
2152 | (void) ztest_write(zd, object, offset, sizeof (wbt), &wbt); | |
2153 | break; | |
2154 | ||
2155 | case ZTEST_IO_WRITE_PATTERN: | |
2156 | (void) memset(data, 'a' + (object + offset) % 5, blocksize); | |
2157 | if (ztest_random(2) == 0) { | |
2158 | /* | |
2159 | * Induce fletcher2 collisions to ensure that | |
2160 | * zio_ddt_collision() detects and resolves them | |
2161 | * when using fletcher2-verify for deduplication. | |
2162 | */ | |
2163 | ((uint64_t *)data)[0] ^= 1ULL << 63; | |
2164 | ((uint64_t *)data)[4] ^= 1ULL << 63; | |
2165 | } | |
2166 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2167 | break; | |
2168 | ||
2169 | case ZTEST_IO_WRITE_ZEROES: | |
2170 | bzero(data, blocksize); | |
2171 | (void) ztest_write(zd, object, offset, blocksize, data); | |
2172 | break; | |
2173 | ||
2174 | case ZTEST_IO_TRUNCATE: | |
2175 | (void) ztest_truncate(zd, object, offset, blocksize); | |
2176 | break; | |
2177 | ||
2178 | case ZTEST_IO_SETATTR: | |
2179 | (void) ztest_setattr(zd, object); | |
2180 | break; | |
e75c13c3 BB |
2181 | default: |
2182 | break; | |
428870ff BB |
2183 | } |
2184 | ||
3e31d2b0 ES |
2185 | (void) rw_exit(&zd->zd_zilog_lock); |
2186 | ||
428870ff BB |
2187 | umem_free(data, blocksize); |
2188 | } | |
2189 | ||
2190 | /* | |
2191 | * Initialize an object description template. | |
2192 | */ | |
2193 | static void | |
2194 | ztest_od_init(ztest_od_t *od, uint64_t id, char *tag, uint64_t index, | |
2195 | dmu_object_type_t type, uint64_t blocksize, uint64_t gen) | |
2196 | { | |
2197 | od->od_dir = ZTEST_DIROBJ; | |
2198 | od->od_object = 0; | |
2199 | ||
2200 | od->od_crtype = type; | |
2201 | od->od_crblocksize = blocksize ? blocksize : ztest_random_blocksize(); | |
2202 | od->od_crgen = gen; | |
2203 | ||
2204 | od->od_type = DMU_OT_NONE; | |
2205 | od->od_blocksize = 0; | |
2206 | od->od_gen = 0; | |
2207 | ||
2208 | (void) snprintf(od->od_name, sizeof (od->od_name), "%s(%lld)[%llu]", | |
b8864a23 | 2209 | tag, (longlong_t)id, (u_longlong_t)index); |
428870ff BB |
2210 | } |
2211 | ||
2212 | /* | |
2213 | * Lookup or create the objects for a test using the od template. | |
2214 | * If the objects do not all exist, or if 'remove' is specified, | |
2215 | * remove any existing objects and create new ones. Otherwise, | |
2216 | * use the existing objects. | |
2217 | */ | |
2218 | static int | |
2219 | ztest_object_init(ztest_ds_t *zd, ztest_od_t *od, size_t size, boolean_t remove) | |
2220 | { | |
2221 | int count = size / sizeof (*od); | |
2222 | int rv = 0; | |
2223 | ||
1e33ac1e | 2224 | mutex_enter(&zd->zd_dirobj_lock); |
428870ff BB |
2225 | if ((ztest_lookup(zd, od, count) != 0 || remove) && |
2226 | (ztest_remove(zd, od, count) != 0 || | |
2227 | ztest_create(zd, od, count) != 0)) | |
2228 | rv = -1; | |
2229 | zd->zd_od = od; | |
1e33ac1e | 2230 | mutex_exit(&zd->zd_dirobj_lock); |
428870ff BB |
2231 | |
2232 | return (rv); | |
2233 | } | |
2234 | ||
2235 | /* ARGSUSED */ | |
2236 | void | |
2237 | ztest_zil_commit(ztest_ds_t *zd, uint64_t id) | |
2238 | { | |
2239 | zilog_t *zilog = zd->zd_zilog; | |
2240 | ||
3e31d2b0 ES |
2241 | (void) rw_enter(&zd->zd_zilog_lock, RW_READER); |
2242 | ||
572e2857 | 2243 | zil_commit(zilog, ztest_random(ZTEST_OBJECTS)); |
428870ff BB |
2244 | |
2245 | /* | |
2246 | * Remember the committed values in zd, which is in parent/child | |
2247 | * shared memory. If we die, the next iteration of ztest_run() | |
2248 | * will verify that the log really does contain this record. | |
2249 | */ | |
2250 | mutex_enter(&zilog->zl_lock); | |
c242c188 CS |
2251 | ASSERT(zd->zd_shared != NULL); |
2252 | ASSERT3U(zd->zd_shared->zd_seq, <=, zilog->zl_commit_lr_seq); | |
2253 | zd->zd_shared->zd_seq = zilog->zl_commit_lr_seq; | |
428870ff | 2254 | mutex_exit(&zilog->zl_lock); |
3e31d2b0 ES |
2255 | |
2256 | (void) rw_exit(&zd->zd_zilog_lock); | |
2257 | } | |
2258 | ||
2259 | /* | |
2260 | * This function is designed to simulate the operations that occur during a | |
2261 | * mount/unmount operation. We hold the dataset across these operations in an | |
2262 | * attempt to expose any implicit assumptions about ZIL management. | |
2263 | */ | |
2264 | /* ARGSUSED */ | |
2265 | void | |
2266 | ztest_zil_remount(ztest_ds_t *zd, uint64_t id) | |
2267 | { | |
2268 | objset_t *os = zd->zd_os; | |
2269 | ||
2270 | (void) rw_enter(&zd->zd_zilog_lock, RW_WRITER); | |
2271 | ||
86f35f34 | 2272 | /* zfs_sb_teardown() */ |
3e31d2b0 ES |
2273 | zil_close(zd->zd_zilog); |
2274 | ||
2275 | /* zfsvfs_setup() */ | |
2276 | VERIFY(zil_open(os, ztest_get_data) == zd->zd_zilog); | |
2277 | zil_replay(os, zd, ztest_replay_vector); | |
2278 | ||
2279 | (void) rw_exit(&zd->zd_zilog_lock); | |
428870ff BB |
2280 | } |
2281 | ||
2282 | /* | |
2283 | * Verify that we can't destroy an active pool, create an existing pool, | |
2284 | * or create a pool with a bad vdev spec. | |
2285 | */ | |
2286 | /* ARGSUSED */ | |
2287 | void | |
2288 | ztest_spa_create_destroy(ztest_ds_t *zd, uint64_t id) | |
2289 | { | |
c242c188 | 2290 | ztest_shared_opts_t *zo = &ztest_opts; |
428870ff BB |
2291 | spa_t *spa; |
2292 | nvlist_t *nvroot; | |
2293 | ||
2294 | /* | |
2295 | * Attempt to create using a bad file. | |
2296 | */ | |
2297 | nvroot = make_vdev_root("/dev/bogus", NULL, 0, 0, 0, 0, 0, 1); | |
2298 | VERIFY3U(ENOENT, ==, | |
2299 | spa_create("ztest_bad_file", nvroot, NULL, NULL, NULL)); | |
2300 | nvlist_free(nvroot); | |
2301 | ||
2302 | /* | |
2303 | * Attempt to create using a bad mirror. | |
2304 | */ | |
2305 | nvroot = make_vdev_root("/dev/bogus", NULL, 0, 0, 0, 0, 2, 1); | |
2306 | VERIFY3U(ENOENT, ==, | |
2307 | spa_create("ztest_bad_mirror", nvroot, NULL, NULL, NULL)); | |
2308 | nvlist_free(nvroot); | |
2309 | ||
2310 | /* | |
2311 | * Attempt to create an existing pool. It shouldn't matter | |
2312 | * what's in the nvroot; we should fail with EEXIST. | |
2313 | */ | |
c242c188 | 2314 | (void) rw_enter(&ztest_name_lock, RW_READER); |
428870ff | 2315 | nvroot = make_vdev_root("/dev/bogus", NULL, 0, 0, 0, 0, 0, 1); |
c242c188 | 2316 | VERIFY3U(EEXIST, ==, spa_create(zo->zo_pool, nvroot, NULL, NULL, NULL)); |
428870ff | 2317 | nvlist_free(nvroot); |
c242c188 CS |
2318 | VERIFY3U(0, ==, spa_open(zo->zo_pool, &spa, FTAG)); |
2319 | VERIFY3U(EBUSY, ==, spa_destroy(zo->zo_pool)); | |
428870ff BB |
2320 | spa_close(spa, FTAG); |
2321 | ||
c242c188 | 2322 | (void) rw_exit(&ztest_name_lock); |
428870ff BB |
2323 | } |
2324 | ||
2325 | static vdev_t * | |
2326 | vdev_lookup_by_path(vdev_t *vd, const char *path) | |
2327 | { | |
2328 | vdev_t *mvd; | |
d6320ddb | 2329 | int c; |
428870ff BB |
2330 | |
2331 | if (vd->vdev_path != NULL && strcmp(path, vd->vdev_path) == 0) | |
2332 | return (vd); | |
2333 | ||
d6320ddb | 2334 | for (c = 0; c < vd->vdev_children; c++) |
428870ff BB |
2335 | if ((mvd = vdev_lookup_by_path(vd->vdev_child[c], path)) != |
2336 | NULL) | |
2337 | return (mvd); | |
2338 | ||
2339 | return (NULL); | |
2340 | } | |
2341 | ||
2342 | /* | |
2343 | * Find the first available hole which can be used as a top-level. | |
2344 | */ | |
2345 | int | |
2346 | find_vdev_hole(spa_t *spa) | |
2347 | { | |
2348 | vdev_t *rvd = spa->spa_root_vdev; | |
2349 | int c; | |
2350 | ||
2351 | ASSERT(spa_config_held(spa, SCL_VDEV, RW_READER) == SCL_VDEV); | |
2352 | ||
2353 | for (c = 0; c < rvd->vdev_children; c++) { | |
2354 | vdev_t *cvd = rvd->vdev_child[c]; | |
2355 | ||
2356 | if (cvd->vdev_ishole) | |
2357 | break; | |
2358 | } | |
2359 | return (c); | |
2360 | } | |
2361 | ||
2362 | /* | |
2363 | * Verify that vdev_add() works as expected. | |
2364 | */ | |
2365 | /* ARGSUSED */ | |
2366 | void | |
2367 | ztest_vdev_add_remove(ztest_ds_t *zd, uint64_t id) | |
2368 | { | |
2369 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 2370 | spa_t *spa = ztest_spa; |
428870ff BB |
2371 | uint64_t leaves; |
2372 | uint64_t guid; | |
2373 | nvlist_t *nvroot; | |
2374 | int error; | |
2375 | ||
c242c188 CS |
2376 | mutex_enter(&ztest_vdev_lock); |
2377 | leaves = | |
2378 | MAX(zs->zs_mirrors + zs->zs_splits, 1) * ztest_opts.zo_raidz; | |
428870ff BB |
2379 | |
2380 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
2381 | ||
2382 | ztest_shared->zs_vdev_next_leaf = find_vdev_hole(spa) * leaves; | |
2383 | ||
2384 | /* | |
2385 | * If we have slogs then remove them 1/4 of the time. | |
2386 | */ | |
2387 | if (spa_has_slogs(spa) && ztest_random(4) == 0) { | |
2388 | /* | |
2389 | * Grab the guid from the head of the log class rotor. | |
2390 | */ | |
2391 | guid = spa_log_class(spa)->mc_rotor->mg_vd->vdev_guid; | |
2392 | ||
2393 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
2394 | ||
2395 | /* | |
2396 | * We have to grab the zs_name_lock as writer to | |
2397 | * prevent a race between removing a slog (dmu_objset_find) | |
2398 | * and destroying a dataset. Removing the slog will | |
2399 | * grab a reference on the dataset which may cause | |
2400 | * dmu_objset_destroy() to fail with EBUSY thus | |
2401 | * leaving the dataset in an inconsistent state. | |
2402 | */ | |
c242c188 | 2403 | rw_enter(&ztest_name_lock, RW_WRITER); |
428870ff | 2404 | error = spa_vdev_remove(spa, guid, B_FALSE); |
c242c188 | 2405 | rw_exit(&ztest_name_lock); |
428870ff BB |
2406 | |
2407 | if (error && error != EEXIST) | |
2408 | fatal(0, "spa_vdev_remove() = %d", error); | |
2409 | } else { | |
2410 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
2411 | ||
2412 | /* | |
2413 | * Make 1/4 of the devices be log devices. | |
2414 | */ | |
c242c188 CS |
2415 | nvroot = make_vdev_root(NULL, NULL, |
2416 | ztest_opts.zo_vdev_size, 0, | |
2417 | ztest_random(4) == 0, ztest_opts.zo_raidz, | |
2418 | zs->zs_mirrors, 1); | |
428870ff BB |
2419 | |
2420 | error = spa_vdev_add(spa, nvroot); | |
2421 | nvlist_free(nvroot); | |
2422 | ||
2423 | if (error == ENOSPC) | |
2424 | ztest_record_enospc("spa_vdev_add"); | |
2425 | else if (error != 0) | |
2426 | fatal(0, "spa_vdev_add() = %d", error); | |
2427 | } | |
2428 | ||
c242c188 | 2429 | mutex_exit(&ztest_vdev_lock); |
428870ff BB |
2430 | } |
2431 | ||
2432 | /* | |
2433 | * Verify that adding/removing aux devices (l2arc, hot spare) works as expected. | |
2434 | */ | |
2435 | /* ARGSUSED */ | |
2436 | void | |
2437 | ztest_vdev_aux_add_remove(ztest_ds_t *zd, uint64_t id) | |
2438 | { | |
2439 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 2440 | spa_t *spa = ztest_spa; |
428870ff BB |
2441 | vdev_t *rvd = spa->spa_root_vdev; |
2442 | spa_aux_vdev_t *sav; | |
2443 | char *aux; | |
40b84e7a | 2444 | char *path; |
428870ff BB |
2445 | uint64_t guid = 0; |
2446 | int error; | |
2447 | ||
40b84e7a BB |
2448 | path = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
2449 | ||
428870ff BB |
2450 | if (ztest_random(2) == 0) { |
2451 | sav = &spa->spa_spares; | |
2452 | aux = ZPOOL_CONFIG_SPARES; | |
2453 | } else { | |
2454 | sav = &spa->spa_l2cache; | |
2455 | aux = ZPOOL_CONFIG_L2CACHE; | |
2456 | } | |
2457 | ||
c242c188 | 2458 | mutex_enter(&ztest_vdev_lock); |
428870ff BB |
2459 | |
2460 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
2461 | ||
2462 | if (sav->sav_count != 0 && ztest_random(4) == 0) { | |
b128c09f BB |
2463 | /* |
2464 | * Pick a random device to remove. | |
2465 | */ | |
2466 | guid = sav->sav_vdevs[ztest_random(sav->sav_count)]->vdev_guid; | |
2467 | } else { | |
2468 | /* | |
2469 | * Find an unused device we can add. | |
2470 | */ | |
428870ff | 2471 | zs->zs_vdev_aux = 0; |
b128c09f | 2472 | for (;;) { |
b128c09f | 2473 | int c; |
c242c188 CS |
2474 | (void) snprintf(path, sizeof (path), ztest_aux_template, |
2475 | ztest_opts.zo_dir, ztest_opts.zo_pool, aux, | |
2476 | zs->zs_vdev_aux); | |
b128c09f BB |
2477 | for (c = 0; c < sav->sav_count; c++) |
2478 | if (strcmp(sav->sav_vdevs[c]->vdev_path, | |
2479 | path) == 0) | |
2480 | break; | |
2481 | if (c == sav->sav_count && | |
2482 | vdev_lookup_by_path(rvd, path) == NULL) | |
2483 | break; | |
428870ff | 2484 | zs->zs_vdev_aux++; |
34dc7c2f BB |
2485 | } |
2486 | } | |
2487 | ||
b128c09f | 2488 | spa_config_exit(spa, SCL_VDEV, FTAG); |
34dc7c2f | 2489 | |
b128c09f BB |
2490 | if (guid == 0) { |
2491 | /* | |
2492 | * Add a new device. | |
2493 | */ | |
2494 | nvlist_t *nvroot = make_vdev_root(NULL, aux, | |
c242c188 | 2495 | (ztest_opts.zo_vdev_size * 5) / 4, 0, 0, 0, 0, 1); |
b128c09f BB |
2496 | error = spa_vdev_add(spa, nvroot); |
2497 | if (error != 0) | |
2498 | fatal(0, "spa_vdev_add(%p) = %d", nvroot, error); | |
2499 | nvlist_free(nvroot); | |
2500 | } else { | |
2501 | /* | |
2502 | * Remove an existing device. Sometimes, dirty its | |
2503 | * vdev state first to make sure we handle removal | |
2504 | * of devices that have pending state changes. | |
2505 | */ | |
2506 | if (ztest_random(2) == 0) | |
9babb374 | 2507 | (void) vdev_online(spa, guid, 0, NULL); |
b128c09f BB |
2508 | |
2509 | error = spa_vdev_remove(spa, guid, B_FALSE); | |
2510 | if (error != 0 && error != EBUSY) | |
2511 | fatal(0, "spa_vdev_remove(%llu) = %d", guid, error); | |
2512 | } | |
2513 | ||
c242c188 | 2514 | mutex_exit(&ztest_vdev_lock); |
40b84e7a BB |
2515 | |
2516 | umem_free(path, MAXPATHLEN); | |
428870ff BB |
2517 | } |
2518 | ||
2519 | /* | |
2520 | * split a pool if it has mirror tlvdevs | |
2521 | */ | |
2522 | /* ARGSUSED */ | |
2523 | void | |
2524 | ztest_split_pool(ztest_ds_t *zd, uint64_t id) | |
2525 | { | |
2526 | ztest_shared_t *zs = ztest_shared; | |
c242c188 | 2527 | spa_t *spa = ztest_spa; |
428870ff BB |
2528 | vdev_t *rvd = spa->spa_root_vdev; |
2529 | nvlist_t *tree, **child, *config, *split, **schild; | |
2530 | uint_t c, children, schildren = 0, lastlogid = 0; | |
2531 | int error = 0; | |
2532 | ||
c242c188 | 2533 | mutex_enter(&ztest_vdev_lock); |
428870ff BB |
2534 | |
2535 | /* ensure we have a useable config; mirrors of raidz aren't supported */ | |
c242c188 CS |
2536 | if (zs->zs_mirrors < 3 || ztest_opts.zo_raidz > 1) { |
2537 | mutex_exit(&ztest_vdev_lock); | |
428870ff BB |
2538 | return; |
2539 | } | |
2540 | ||
2541 | /* clean up the old pool, if any */ | |
2542 | (void) spa_destroy("splitp"); | |
2543 | ||
2544 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); | |
2545 | ||
2546 | /* generate a config from the existing config */ | |
2547 | mutex_enter(&spa->spa_props_lock); | |
2548 | VERIFY(nvlist_lookup_nvlist(spa->spa_config, ZPOOL_CONFIG_VDEV_TREE, | |
2549 | &tree) == 0); | |
2550 | mutex_exit(&spa->spa_props_lock); | |
2551 | ||
2552 | VERIFY(nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child, | |
2553 | &children) == 0); | |
2554 | ||
2555 | schild = malloc(rvd->vdev_children * sizeof (nvlist_t *)); | |
2556 | for (c = 0; c < children; c++) { | |
2557 | vdev_t *tvd = rvd->vdev_child[c]; | |
2558 | nvlist_t **mchild; | |
2559 | uint_t mchildren; | |
2560 | ||
2561 | if (tvd->vdev_islog || tvd->vdev_ops == &vdev_hole_ops) { | |
2562 | VERIFY(nvlist_alloc(&schild[schildren], NV_UNIQUE_NAME, | |
2563 | 0) == 0); | |
2564 | VERIFY(nvlist_add_string(schild[schildren], | |
2565 | ZPOOL_CONFIG_TYPE, VDEV_TYPE_HOLE) == 0); | |
2566 | VERIFY(nvlist_add_uint64(schild[schildren], | |
2567 | ZPOOL_CONFIG_IS_HOLE, 1) == 0); | |
2568 | if (lastlogid == 0) | |
2569 | lastlogid = schildren; | |
2570 | ++schildren; | |
2571 | continue; | |
2572 | } | |
2573 | lastlogid = 0; | |
2574 | VERIFY(nvlist_lookup_nvlist_array(child[c], | |
2575 | ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0); | |
2576 | VERIFY(nvlist_dup(mchild[0], &schild[schildren++], 0) == 0); | |
2577 | } | |
2578 | ||
2579 | /* OK, create a config that can be used to split */ | |
2580 | VERIFY(nvlist_alloc(&split, NV_UNIQUE_NAME, 0) == 0); | |
2581 | VERIFY(nvlist_add_string(split, ZPOOL_CONFIG_TYPE, | |
2582 | VDEV_TYPE_ROOT) == 0); | |
2583 | VERIFY(nvlist_add_nvlist_array(split, ZPOOL_CONFIG_CHILDREN, schild, | |
2584 | lastlogid != 0 ? lastlogid : schildren) == 0); | |
2585 | ||
2586 | VERIFY(nvlist_alloc(&config, NV_UNIQUE_NAME, 0) == 0); | |
2587 | VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, split) == 0); | |
2588 | ||
2589 | for (c = 0; c < schildren; c++) | |
2590 | nvlist_free(schild[c]); | |
2591 | free(schild); | |
2592 | nvlist_free(split); | |
2593 | ||
2594 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
2595 | ||
c242c188 | 2596 | (void) rw_enter(&ztest_name_lock, RW_WRITER); |
428870ff | 2597 | error = spa_vdev_split_mirror(spa, "splitp", config, NULL, B_FALSE); |
c242c188 | 2598 | (void) rw_exit(&ztest_name_lock); |
428870ff BB |
2599 | |
2600 | nvlist_free(config); | |
2601 | ||
2602 | if (error == 0) { | |
2603 | (void) printf("successful split - results:\n"); | |
2604 | mutex_enter(&spa_namespace_lock); | |
2605 | show_pool_stats(spa); | |
2606 | show_pool_stats(spa_lookup("splitp")); | |
2607 | mutex_exit(&spa_namespace_lock); | |
2608 | ++zs->zs_splits; | |
2609 | --zs->zs_mirrors; | |
2610 | } | |
c242c188 | 2611 | mutex_exit(&ztest_vdev_lock); |
428870ff | 2612 | |
34dc7c2f BB |
2613 | } |
2614 | ||
2615 | /* | |
2616 | * Verify that we can attach and detach devices. | |
2617 | */ | |
428870ff | 2618 | /* ARGSUSED */ |
34dc7c2f | 2619 | void |
428870ff | 2620 | ztest_vdev_attach_detach(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 2621 | { |
428870ff | 2622 | ztest_shared_t *zs = ztest_shared; |
c242c188 | 2623 | spa_t *spa = ztest_spa; |
b128c09f | 2624 | spa_aux_vdev_t *sav = &spa->spa_spares; |
34dc7c2f BB |
2625 | vdev_t *rvd = spa->spa_root_vdev; |
2626 | vdev_t *oldvd, *newvd, *pvd; | |
b128c09f | 2627 | nvlist_t *root; |
428870ff | 2628 | uint64_t leaves; |
34dc7c2f BB |
2629 | uint64_t leaf, top; |
2630 | uint64_t ashift = ztest_get_ashift(); | |
fb5f0bc8 | 2631 | uint64_t oldguid, pguid; |
34dc7c2f | 2632 | size_t oldsize, newsize; |
40b84e7a | 2633 | char *oldpath, *newpath; |
34dc7c2f | 2634 | int replacing; |
b128c09f BB |
2635 | int oldvd_has_siblings = B_FALSE; |
2636 | int newvd_is_spare = B_FALSE; | |
2637 | int oldvd_is_log; | |
34dc7c2f | 2638 | int error, expected_error; |
34dc7c2f | 2639 | |
40b84e7a BB |
2640 | oldpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
2641 | newpath = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
2642 | ||
c242c188 CS |
2643 | mutex_enter(&ztest_vdev_lock); |
2644 | leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raidz; | |
34dc7c2f | 2645 | |
b128c09f | 2646 | spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); |
34dc7c2f BB |
2647 | |
2648 | /* | |
2649 | * Decide whether to do an attach or a replace. | |
2650 | */ | |
2651 | replacing = ztest_random(2); | |
2652 | ||
2653 | /* | |
2654 | * Pick a random top-level vdev. | |
2655 | */ | |
428870ff | 2656 | top = ztest_random_vdev_top(spa, B_TRUE); |
34dc7c2f BB |
2657 | |
2658 | /* | |
2659 | * Pick a random leaf within it. | |
2660 | */ | |
2661 | leaf = ztest_random(leaves); | |
2662 | ||
2663 | /* | |
b128c09f | 2664 | * Locate this vdev. |
34dc7c2f | 2665 | */ |
b128c09f | 2666 | oldvd = rvd->vdev_child[top]; |
428870ff | 2667 | if (zs->zs_mirrors >= 1) { |
fb5f0bc8 | 2668 | ASSERT(oldvd->vdev_ops == &vdev_mirror_ops); |
428870ff | 2669 | ASSERT(oldvd->vdev_children >= zs->zs_mirrors); |
c242c188 | 2670 | oldvd = oldvd->vdev_child[leaf / ztest_opts.zo_raidz]; |
fb5f0bc8 | 2671 | } |
c242c188 | 2672 | if (ztest_opts.zo_raidz > 1) { |
fb5f0bc8 | 2673 | ASSERT(oldvd->vdev_ops == &vdev_raidz_ops); |
c242c188 CS |
2674 | ASSERT(oldvd->vdev_children == ztest_opts.zo_raidz); |
2675 | oldvd = oldvd->vdev_child[leaf % ztest_opts.zo_raidz]; | |
fb5f0bc8 | 2676 | } |
34dc7c2f BB |
2677 | |
2678 | /* | |
b128c09f BB |
2679 | * If we're already doing an attach or replace, oldvd may be a |
2680 | * mirror vdev -- in which case, pick a random child. | |
34dc7c2f | 2681 | */ |
b128c09f BB |
2682 | while (oldvd->vdev_children != 0) { |
2683 | oldvd_has_siblings = B_TRUE; | |
fb5f0bc8 BB |
2684 | ASSERT(oldvd->vdev_children >= 2); |
2685 | oldvd = oldvd->vdev_child[ztest_random(oldvd->vdev_children)]; | |
b128c09f BB |
2686 | } |
2687 | ||
2688 | oldguid = oldvd->vdev_guid; | |
9babb374 | 2689 | oldsize = vdev_get_min_asize(oldvd); |
b128c09f BB |
2690 | oldvd_is_log = oldvd->vdev_top->vdev_islog; |
2691 | (void) strcpy(oldpath, oldvd->vdev_path); | |
2692 | pvd = oldvd->vdev_parent; | |
fb5f0bc8 | 2693 | pguid = pvd->vdev_guid; |
34dc7c2f BB |
2694 | |
2695 | /* | |
b128c09f | 2696 | * If oldvd has siblings, then half of the time, detach it. |
34dc7c2f | 2697 | */ |
b128c09f BB |
2698 | if (oldvd_has_siblings && ztest_random(2) == 0) { |
2699 | spa_config_exit(spa, SCL_VDEV, FTAG); | |
fb5f0bc8 BB |
2700 | error = spa_vdev_detach(spa, oldguid, pguid, B_FALSE); |
2701 | if (error != 0 && error != ENODEV && error != EBUSY && | |
2702 | error != ENOTSUP) | |
2703 | fatal(0, "detach (%s) returned %d", oldpath, error); | |
40b84e7a | 2704 | goto out; |
b128c09f | 2705 | } |
34dc7c2f BB |
2706 | |
2707 | /* | |
b128c09f BB |
2708 | * For the new vdev, choose with equal probability between the two |
2709 | * standard paths (ending in either 'a' or 'b') or a random hot spare. | |
34dc7c2f | 2710 | */ |
b128c09f BB |
2711 | if (sav->sav_count != 0 && ztest_random(3) == 0) { |
2712 | newvd = sav->sav_vdevs[ztest_random(sav->sav_count)]; | |
2713 | newvd_is_spare = B_TRUE; | |
2714 | (void) strcpy(newpath, newvd->vdev_path); | |
2715 | } else { | |
6aec1cd5 | 2716 | (void) snprintf(newpath, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
2717 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
2718 | top * leaves + leaf); | |
b128c09f BB |
2719 | if (ztest_random(2) == 0) |
2720 | newpath[strlen(newpath) - 1] = 'b'; | |
2721 | newvd = vdev_lookup_by_path(rvd, newpath); | |
2722 | } | |
2723 | ||
2724 | if (newvd) { | |
9babb374 | 2725 | newsize = vdev_get_min_asize(newvd); |
b128c09f BB |
2726 | } else { |
2727 | /* | |
2728 | * Make newsize a little bigger or smaller than oldsize. | |
2729 | * If it's smaller, the attach should fail. | |
2730 | * If it's larger, and we're doing a replace, | |
2731 | * we should get dynamic LUN growth when we're done. | |
2732 | */ | |
2733 | newsize = 10 * oldsize / (9 + ztest_random(3)); | |
2734 | } | |
34dc7c2f BB |
2735 | |
2736 | /* | |
2737 | * If pvd is not a mirror or root, the attach should fail with ENOTSUP, | |
2738 | * unless it's a replace; in that case any non-replacing parent is OK. | |
2739 | * | |
2740 | * If newvd is already part of the pool, it should fail with EBUSY. | |
2741 | * | |
2742 | * If newvd is too small, it should fail with EOVERFLOW. | |
2743 | */ | |
b128c09f BB |
2744 | if (pvd->vdev_ops != &vdev_mirror_ops && |
2745 | pvd->vdev_ops != &vdev_root_ops && (!replacing || | |
2746 | pvd->vdev_ops == &vdev_replacing_ops || | |
2747 | pvd->vdev_ops == &vdev_spare_ops)) | |
34dc7c2f | 2748 | expected_error = ENOTSUP; |
b128c09f BB |
2749 | else if (newvd_is_spare && (!replacing || oldvd_is_log)) |
2750 | expected_error = ENOTSUP; | |
2751 | else if (newvd == oldvd) | |
2752 | expected_error = replacing ? 0 : EBUSY; | |
2753 | else if (vdev_lookup_by_path(rvd, newpath) != NULL) | |
2754 | expected_error = EBUSY; | |
34dc7c2f BB |
2755 | else if (newsize < oldsize) |
2756 | expected_error = EOVERFLOW; | |
2757 | else if (ashift > oldvd->vdev_top->vdev_ashift) | |
2758 | expected_error = EDOM; | |
2759 | else | |
2760 | expected_error = 0; | |
2761 | ||
b128c09f | 2762 | spa_config_exit(spa, SCL_VDEV, FTAG); |
34dc7c2f BB |
2763 | |
2764 | /* | |
2765 | * Build the nvlist describing newpath. | |
2766 | */ | |
b128c09f BB |
2767 | root = make_vdev_root(newpath, NULL, newvd == NULL ? newsize : 0, |
2768 | ashift, 0, 0, 0, 1); | |
34dc7c2f | 2769 | |
b128c09f | 2770 | error = spa_vdev_attach(spa, oldguid, root, replacing); |
34dc7c2f | 2771 | |
34dc7c2f BB |
2772 | nvlist_free(root); |
2773 | ||
2774 | /* | |
2775 | * If our parent was the replacing vdev, but the replace completed, | |
2776 | * then instead of failing with ENOTSUP we may either succeed, | |
2777 | * fail with ENODEV, or fail with EOVERFLOW. | |
2778 | */ | |
2779 | if (expected_error == ENOTSUP && | |
2780 | (error == 0 || error == ENODEV || error == EOVERFLOW)) | |
2781 | expected_error = error; | |
2782 | ||
2783 | /* | |
2784 | * If someone grew the LUN, the replacement may be too small. | |
2785 | */ | |
b128c09f | 2786 | if (error == EOVERFLOW || error == EBUSY) |
34dc7c2f BB |
2787 | expected_error = error; |
2788 | ||
b128c09f BB |
2789 | /* XXX workaround 6690467 */ |
2790 | if (error != expected_error && expected_error != EBUSY) { | |
2791 | fatal(0, "attach (%s %llu, %s %llu, %d) " | |
2792 | "returned %d, expected %d", | |
2793 | oldpath, (longlong_t)oldsize, newpath, | |
2794 | (longlong_t)newsize, replacing, error, expected_error); | |
34dc7c2f | 2795 | } |
40b84e7a | 2796 | out: |
c242c188 | 2797 | mutex_exit(&ztest_vdev_lock); |
40b84e7a BB |
2798 | |
2799 | umem_free(oldpath, MAXPATHLEN); | |
2800 | umem_free(newpath, MAXPATHLEN); | |
34dc7c2f BB |
2801 | } |
2802 | ||
9babb374 BB |
2803 | /* |
2804 | * Callback function which expands the physical size of the vdev. | |
2805 | */ | |
2806 | vdev_t * | |
2807 | grow_vdev(vdev_t *vd, void *arg) | |
2808 | { | |
1fde1e37 | 2809 | ASSERTV(spa_t *spa = vd->vdev_spa); |
9babb374 BB |
2810 | size_t *newsize = arg; |
2811 | size_t fsize; | |
2812 | int fd; | |
2813 | ||
2814 | ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE); | |
2815 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
2816 | ||
2817 | if ((fd = open(vd->vdev_path, O_RDWR)) == -1) | |
2818 | return (vd); | |
2819 | ||
2820 | fsize = lseek(fd, 0, SEEK_END); | |
0e5b68e0 | 2821 | VERIFY(ftruncate(fd, *newsize) == 0); |
9babb374 | 2822 | |
c242c188 | 2823 | if (ztest_opts.zo_verbose >= 6) { |
9babb374 BB |
2824 | (void) printf("%s grew from %lu to %lu bytes\n", |
2825 | vd->vdev_path, (ulong_t)fsize, (ulong_t)*newsize); | |
2826 | } | |
2827 | (void) close(fd); | |
2828 | return (NULL); | |
2829 | } | |
2830 | ||
2831 | /* | |
2832 | * Callback function which expands a given vdev by calling vdev_online(). | |
2833 | */ | |
2834 | /* ARGSUSED */ | |
2835 | vdev_t * | |
2836 | online_vdev(vdev_t *vd, void *arg) | |
2837 | { | |
2838 | spa_t *spa = vd->vdev_spa; | |
2839 | vdev_t *tvd = vd->vdev_top; | |
9babb374 | 2840 | uint64_t guid = vd->vdev_guid; |
428870ff BB |
2841 | uint64_t generation = spa->spa_config_generation + 1; |
2842 | vdev_state_t newstate = VDEV_STATE_UNKNOWN; | |
2843 | int error; | |
9babb374 BB |
2844 | |
2845 | ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE); | |
2846 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
2847 | ||
2848 | /* Calling vdev_online will initialize the new metaslabs */ | |
2849 | spa_config_exit(spa, SCL_STATE, spa); | |
428870ff | 2850 | error = vdev_online(spa, guid, ZFS_ONLINE_EXPAND, &newstate); |
9babb374 BB |
2851 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); |
2852 | ||
428870ff BB |
2853 | /* |
2854 | * If vdev_online returned an error or the underlying vdev_open | |
2855 | * failed then we abort the expand. The only way to know that | |
2856 | * vdev_open fails is by checking the returned newstate. | |
2857 | */ | |
2858 | if (error || newstate != VDEV_STATE_HEALTHY) { | |
c242c188 | 2859 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
2860 | (void) printf("Unable to expand vdev, state %llu, " |
2861 | "error %d\n", (u_longlong_t)newstate, error); | |
2862 | } | |
2863 | return (vd); | |
2864 | } | |
2865 | ASSERT3U(newstate, ==, VDEV_STATE_HEALTHY); | |
2866 | ||
9babb374 BB |
2867 | /* |
2868 | * Since we dropped the lock we need to ensure that we're | |
2869 | * still talking to the original vdev. It's possible this | |
2870 | * vdev may have been detached/replaced while we were | |
2871 | * trying to online it. | |
2872 | */ | |
428870ff | 2873 | if (generation != spa->spa_config_generation) { |
c242c188 | 2874 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
2875 | (void) printf("vdev configuration has changed, " |
2876 | "guid %llu, state %llu, expected gen %llu, " | |
2877 | "got gen %llu\n", | |
2878 | (u_longlong_t)guid, | |
2879 | (u_longlong_t)tvd->vdev_state, | |
2880 | (u_longlong_t)generation, | |
2881 | (u_longlong_t)spa->spa_config_generation); | |
9babb374 BB |
2882 | } |
2883 | return (vd); | |
2884 | } | |
2885 | return (NULL); | |
2886 | } | |
2887 | ||
2888 | /* | |
2889 | * Traverse the vdev tree calling the supplied function. | |
2890 | * We continue to walk the tree until we either have walked all | |
2891 | * children or we receive a non-NULL return from the callback. | |
2892 | * If a NULL callback is passed, then we just return back the first | |
2893 | * leaf vdev we encounter. | |
2894 | */ | |
2895 | vdev_t * | |
2896 | vdev_walk_tree(vdev_t *vd, vdev_t *(*func)(vdev_t *, void *), void *arg) | |
2897 | { | |
d6320ddb BB |
2898 | uint_t c; |
2899 | ||
9babb374 BB |
2900 | if (vd->vdev_ops->vdev_op_leaf) { |
2901 | if (func == NULL) | |
2902 | return (vd); | |
2903 | else | |
2904 | return (func(vd, arg)); | |
2905 | } | |
2906 | ||
d6320ddb | 2907 | for (c = 0; c < vd->vdev_children; c++) { |
9babb374 BB |
2908 | vdev_t *cvd = vd->vdev_child[c]; |
2909 | if ((cvd = vdev_walk_tree(cvd, func, arg)) != NULL) | |
2910 | return (cvd); | |
2911 | } | |
2912 | return (NULL); | |
2913 | } | |
2914 | ||
34dc7c2f BB |
2915 | /* |
2916 | * Verify that dynamic LUN growth works as expected. | |
2917 | */ | |
428870ff | 2918 | /* ARGSUSED */ |
34dc7c2f | 2919 | void |
428870ff | 2920 | ztest_vdev_LUN_growth(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 2921 | { |
c242c188 | 2922 | spa_t *spa = ztest_spa; |
428870ff BB |
2923 | vdev_t *vd, *tvd; |
2924 | metaslab_class_t *mc; | |
2925 | metaslab_group_t *mg; | |
9babb374 | 2926 | size_t psize, newsize; |
428870ff BB |
2927 | uint64_t top; |
2928 | uint64_t old_class_space, new_class_space, old_ms_count, new_ms_count; | |
34dc7c2f | 2929 | |
c242c188 | 2930 | mutex_enter(&ztest_vdev_lock); |
9babb374 BB |
2931 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); |
2932 | ||
428870ff | 2933 | top = ztest_random_vdev_top(spa, B_TRUE); |
9babb374 | 2934 | |
428870ff BB |
2935 | tvd = spa->spa_root_vdev->vdev_child[top]; |
2936 | mg = tvd->vdev_mg; | |
2937 | mc = mg->mg_class; | |
2938 | old_ms_count = tvd->vdev_ms_count; | |
2939 | old_class_space = metaslab_class_get_space(mc); | |
34dc7c2f BB |
2940 | |
2941 | /* | |
9babb374 BB |
2942 | * Determine the size of the first leaf vdev associated with |
2943 | * our top-level device. | |
34dc7c2f | 2944 | */ |
9babb374 BB |
2945 | vd = vdev_walk_tree(tvd, NULL, NULL); |
2946 | ASSERT3P(vd, !=, NULL); | |
2947 | ASSERT(vd->vdev_ops->vdev_op_leaf); | |
34dc7c2f | 2948 | |
9babb374 | 2949 | psize = vd->vdev_psize; |
34dc7c2f | 2950 | |
9babb374 | 2951 | /* |
428870ff BB |
2952 | * We only try to expand the vdev if it's healthy, less than 4x its |
2953 | * original size, and it has a valid psize. | |
9babb374 | 2954 | */ |
428870ff | 2955 | if (tvd->vdev_state != VDEV_STATE_HEALTHY || |
c242c188 | 2956 | psize == 0 || psize >= 4 * ztest_opts.zo_vdev_size) { |
9babb374 | 2957 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 2958 | mutex_exit(&ztest_vdev_lock); |
9babb374 BB |
2959 | return; |
2960 | } | |
2961 | ASSERT(psize > 0); | |
2962 | newsize = psize + psize / 8; | |
2963 | ASSERT3U(newsize, >, psize); | |
34dc7c2f | 2964 | |
c242c188 | 2965 | if (ztest_opts.zo_verbose >= 6) { |
428870ff | 2966 | (void) printf("Expanding LUN %s from %lu to %lu\n", |
9babb374 BB |
2967 | vd->vdev_path, (ulong_t)psize, (ulong_t)newsize); |
2968 | } | |
2969 | ||
9babb374 BB |
2970 | /* |
2971 | * Growing the vdev is a two step process: | |
2972 | * 1). expand the physical size (i.e. relabel) | |
2973 | * 2). online the vdev to create the new metaslabs | |
2974 | */ | |
2975 | if (vdev_walk_tree(tvd, grow_vdev, &newsize) != NULL || | |
2976 | vdev_walk_tree(tvd, online_vdev, NULL) != NULL || | |
2977 | tvd->vdev_state != VDEV_STATE_HEALTHY) { | |
c242c188 | 2978 | if (ztest_opts.zo_verbose >= 5) { |
9babb374 | 2979 | (void) printf("Could not expand LUN because " |
428870ff | 2980 | "the vdev configuration changed.\n"); |
34dc7c2f | 2981 | } |
428870ff | 2982 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 2983 | mutex_exit(&ztest_vdev_lock); |
9babb374 | 2984 | return; |
34dc7c2f BB |
2985 | } |
2986 | ||
428870ff | 2987 | spa_config_exit(spa, SCL_STATE, spa); |
9babb374 BB |
2988 | |
2989 | /* | |
2990 | * Expanding the LUN will update the config asynchronously, | |
2991 | * thus we must wait for the async thread to complete any | |
2992 | * pending tasks before proceeding. | |
2993 | */ | |
428870ff BB |
2994 | for (;;) { |
2995 | boolean_t done; | |
2996 | mutex_enter(&spa->spa_async_lock); | |
2997 | done = (spa->spa_async_thread == NULL && !spa->spa_async_tasks); | |
2998 | mutex_exit(&spa->spa_async_lock); | |
2999 | if (done) | |
3000 | break; | |
3001 | txg_wait_synced(spa_get_dsl(spa), 0); | |
3002 | (void) poll(NULL, 0, 100); | |
3003 | } | |
9babb374 BB |
3004 | |
3005 | spa_config_enter(spa, SCL_STATE, spa, RW_READER); | |
428870ff BB |
3006 | |
3007 | tvd = spa->spa_root_vdev->vdev_child[top]; | |
3008 | new_ms_count = tvd->vdev_ms_count; | |
3009 | new_class_space = metaslab_class_get_space(mc); | |
3010 | ||
3011 | if (tvd->vdev_mg != mg || mg->mg_class != mc) { | |
c242c188 | 3012 | if (ztest_opts.zo_verbose >= 5) { |
428870ff BB |
3013 | (void) printf("Could not verify LUN expansion due to " |
3014 | "intervening vdev offline or remove.\n"); | |
3015 | } | |
3016 | spa_config_exit(spa, SCL_STATE, spa); | |
c242c188 | 3017 | mutex_exit(&ztest_vdev_lock); |
428870ff BB |
3018 | return; |
3019 | } | |
3020 | ||
3021 | /* | |
3022 | * Make sure we were able to grow the vdev. | |
3023 | */ | |
3024 | if (new_ms_count <= old_ms_count) | |
3025 | fatal(0, "LUN expansion failed: ms_count %llu <= %llu\n", | |
3026 | old_ms_count, new_ms_count); | |
9babb374 BB |
3027 | |
3028 | /* | |
3029 | * Make sure we were able to grow the pool. | |
3030 | */ | |
428870ff BB |
3031 | if (new_class_space <= old_class_space) |
3032 | fatal(0, "LUN expansion failed: class_space %llu <= %llu\n", | |
3033 | old_class_space, new_class_space); | |
3034 | ||
c242c188 | 3035 | if (ztest_opts.zo_verbose >= 5) { |
9babb374 BB |
3036 | char oldnumbuf[6], newnumbuf[6]; |
3037 | ||
428870ff BB |
3038 | nicenum(old_class_space, oldnumbuf); |
3039 | nicenum(new_class_space, newnumbuf); | |
9babb374 BB |
3040 | (void) printf("%s grew from %s to %s\n", |
3041 | spa->spa_name, oldnumbuf, newnumbuf); | |
3042 | } | |
428870ff | 3043 | |
9babb374 | 3044 | spa_config_exit(spa, SCL_STATE, spa); |
c242c188 | 3045 | mutex_exit(&ztest_vdev_lock); |
34dc7c2f BB |
3046 | } |
3047 | ||
428870ff BB |
3048 | /* |
3049 | * Verify that dmu_objset_{create,destroy,open,close} work as expected. | |
3050 | */ | |
34dc7c2f BB |
3051 | /* ARGSUSED */ |
3052 | static void | |
428870ff | 3053 | ztest_objset_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) |
34dc7c2f BB |
3054 | { |
3055 | /* | |
428870ff | 3056 | * Create the objects common to all ztest datasets. |
34dc7c2f | 3057 | */ |
428870ff | 3058 | VERIFY(zap_create_claim(os, ZTEST_DIROBJ, |
34dc7c2f | 3059 | DMU_OT_ZAP_OTHER, DMU_OT_NONE, 0, tx) == 0); |
428870ff | 3060 | } |
34dc7c2f | 3061 | |
428870ff BB |
3062 | static int |
3063 | ztest_dataset_create(char *dsname) | |
3064 | { | |
3065 | uint64_t zilset = ztest_random(100); | |
3066 | int err = dmu_objset_create(dsname, DMU_OST_OTHER, 0, | |
3067 | ztest_objset_create_cb, NULL); | |
3068 | ||
3069 | if (err || zilset < 80) | |
3070 | return (err); | |
3071 | ||
c242c188 | 3072 | if (ztest_opts.zo_verbose >= 5) |
b815ff9a | 3073 | (void) printf("Setting dataset %s to sync always\n", dsname); |
428870ff BB |
3074 | return (ztest_dsl_prop_set_uint64(dsname, ZFS_PROP_SYNC, |
3075 | ZFS_SYNC_ALWAYS, B_FALSE)); | |
34dc7c2f BB |
3076 | } |
3077 | ||
428870ff | 3078 | /* ARGSUSED */ |
34dc7c2f | 3079 | static int |
428870ff | 3080 | ztest_objset_destroy_cb(const char *name, void *arg) |
34dc7c2f | 3081 | { |
34dc7c2f | 3082 | objset_t *os; |
428870ff | 3083 | dmu_object_info_t doi; |
34dc7c2f BB |
3084 | int error; |
3085 | ||
3086 | /* | |
3087 | * Verify that the dataset contains a directory object. | |
3088 | */ | |
428870ff BB |
3089 | VERIFY3U(0, ==, dmu_objset_hold(name, FTAG, &os)); |
3090 | error = dmu_object_info(os, ZTEST_DIROBJ, &doi); | |
34dc7c2f BB |
3091 | if (error != ENOENT) { |
3092 | /* We could have crashed in the middle of destroying it */ | |
3093 | ASSERT3U(error, ==, 0); | |
428870ff BB |
3094 | ASSERT3U(doi.doi_type, ==, DMU_OT_ZAP_OTHER); |
3095 | ASSERT3S(doi.doi_physical_blocks_512, >=, 0); | |
34dc7c2f | 3096 | } |
428870ff | 3097 | dmu_objset_rele(os, FTAG); |
34dc7c2f BB |
3098 | |
3099 | /* | |
3100 | * Destroy the dataset. | |
3101 | */ | |
428870ff | 3102 | VERIFY3U(0, ==, dmu_objset_destroy(name, B_FALSE)); |
34dc7c2f BB |
3103 | return (0); |
3104 | } | |
3105 | ||
428870ff BB |
3106 | static boolean_t |
3107 | ztest_snapshot_create(char *osname, uint64_t id) | |
34dc7c2f | 3108 | { |
428870ff BB |
3109 | char snapname[MAXNAMELEN]; |
3110 | int error; | |
3111 | ||
3112 | (void) snprintf(snapname, MAXNAMELEN, "%s@%llu", osname, | |
3113 | (u_longlong_t)id); | |
3114 | ||
3115 | error = dmu_objset_snapshot(osname, strchr(snapname, '@') + 1, | |
572e2857 | 3116 | NULL, NULL, B_FALSE, B_FALSE, -1); |
428870ff BB |
3117 | if (error == ENOSPC) { |
3118 | ztest_record_enospc(FTAG); | |
3119 | return (B_FALSE); | |
3120 | } | |
3121 | if (error != 0 && error != EEXIST) | |
3122 | fatal(0, "ztest_snapshot_create(%s) = %d", snapname, error); | |
3123 | return (B_TRUE); | |
3124 | } | |
3125 | ||
3126 | static boolean_t | |
3127 | ztest_snapshot_destroy(char *osname, uint64_t id) | |
3128 | { | |
3129 | char snapname[MAXNAMELEN]; | |
3130 | int error; | |
3131 | ||
3132 | (void) snprintf(snapname, MAXNAMELEN, "%s@%llu", osname, | |
3133 | (u_longlong_t)id); | |
3134 | ||
3135 | error = dmu_objset_destroy(snapname, B_FALSE); | |
3136 | if (error != 0 && error != ENOENT) | |
3137 | fatal(0, "ztest_snapshot_destroy(%s) = %d", snapname, error); | |
3138 | return (B_TRUE); | |
34dc7c2f BB |
3139 | } |
3140 | ||
428870ff | 3141 | /* ARGSUSED */ |
34dc7c2f | 3142 | void |
428870ff | 3143 | ztest_dmu_objset_create_destroy(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3144 | { |
40b84e7a | 3145 | ztest_ds_t *zdtmp; |
428870ff | 3146 | int iters; |
34dc7c2f | 3147 | int error; |
b128c09f | 3148 | objset_t *os, *os2; |
40b84e7a | 3149 | char *name; |
34dc7c2f | 3150 | zilog_t *zilog; |
d6320ddb | 3151 | int i; |
34dc7c2f | 3152 | |
40b84e7a BB |
3153 | zdtmp = umem_alloc(sizeof (ztest_ds_t), UMEM_NOFAIL); |
3154 | name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3155 | ||
c242c188 | 3156 | (void) rw_enter(&ztest_name_lock, RW_READER); |
34dc7c2f | 3157 | |
428870ff | 3158 | (void) snprintf(name, MAXNAMELEN, "%s/temp_%llu", |
c242c188 | 3159 | ztest_opts.zo_pool, (u_longlong_t)id); |
34dc7c2f BB |
3160 | |
3161 | /* | |
3162 | * If this dataset exists from a previous run, process its replay log | |
3163 | * half of the time. If we don't replay it, then dmu_objset_destroy() | |
428870ff | 3164 | * (invoked from ztest_objset_destroy_cb()) should just throw it away. |
34dc7c2f BB |
3165 | */ |
3166 | if (ztest_random(2) == 0 && | |
428870ff | 3167 | dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, FTAG, &os) == 0) { |
c242c188 | 3168 | ztest_zd_init(zdtmp, NULL, os); |
40b84e7a BB |
3169 | zil_replay(os, zdtmp, ztest_replay_vector); |
3170 | ztest_zd_fini(zdtmp); | |
428870ff | 3171 | dmu_objset_disown(os, FTAG); |
34dc7c2f BB |
3172 | } |
3173 | ||
3174 | /* | |
3175 | * There may be an old instance of the dataset we're about to | |
3176 | * create lying around from a previous run. If so, destroy it | |
3177 | * and all of its snapshots. | |
3178 | */ | |
428870ff | 3179 | (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL, |
34dc7c2f BB |
3180 | DS_FIND_CHILDREN | DS_FIND_SNAPSHOTS); |
3181 | ||
3182 | /* | |
3183 | * Verify that the destroyed dataset is no longer in the namespace. | |
3184 | */ | |
428870ff | 3185 | VERIFY3U(ENOENT, ==, dmu_objset_hold(name, FTAG, &os)); |
34dc7c2f BB |
3186 | |
3187 | /* | |
3188 | * Verify that we can create a new dataset. | |
3189 | */ | |
428870ff | 3190 | error = ztest_dataset_create(name); |
34dc7c2f BB |
3191 | if (error) { |
3192 | if (error == ENOSPC) { | |
428870ff | 3193 | ztest_record_enospc(FTAG); |
40b84e7a | 3194 | goto out; |
34dc7c2f BB |
3195 | } |
3196 | fatal(0, "dmu_objset_create(%s) = %d", name, error); | |
3197 | } | |
3198 | ||
428870ff BB |
3199 | VERIFY3U(0, ==, |
3200 | dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, FTAG, &os)); | |
3201 | ||
c242c188 | 3202 | ztest_zd_init(zdtmp, NULL, os); |
34dc7c2f BB |
3203 | |
3204 | /* | |
3205 | * Open the intent log for it. | |
3206 | */ | |
428870ff | 3207 | zilog = zil_open(os, ztest_get_data); |
34dc7c2f BB |
3208 | |
3209 | /* | |
428870ff BB |
3210 | * Put some objects in there, do a little I/O to them, |
3211 | * and randomly take a couple of snapshots along the way. | |
34dc7c2f | 3212 | */ |
428870ff | 3213 | iters = ztest_random(5); |
d6320ddb | 3214 | for (i = 0; i < iters; i++) { |
40b84e7a | 3215 | ztest_dmu_object_alloc_free(zdtmp, id); |
428870ff BB |
3216 | if (ztest_random(iters) == 0) |
3217 | (void) ztest_snapshot_create(name, i); | |
34dc7c2f BB |
3218 | } |
3219 | ||
3220 | /* | |
3221 | * Verify that we cannot create an existing dataset. | |
3222 | */ | |
428870ff BB |
3223 | VERIFY3U(EEXIST, ==, |
3224 | dmu_objset_create(name, DMU_OST_OTHER, 0, NULL, NULL)); | |
34dc7c2f BB |
3225 | |
3226 | /* | |
428870ff | 3227 | * Verify that we can hold an objset that is also owned. |
b128c09f | 3228 | */ |
428870ff BB |
3229 | VERIFY3U(0, ==, dmu_objset_hold(name, FTAG, &os2)); |
3230 | dmu_objset_rele(os2, FTAG); | |
34dc7c2f | 3231 | |
428870ff BB |
3232 | /* |
3233 | * Verify that we cannot own an objset that is already owned. | |
3234 | */ | |
3235 | VERIFY3U(EBUSY, ==, | |
3236 | dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, FTAG, &os2)); | |
34dc7c2f | 3237 | |
428870ff BB |
3238 | zil_close(zilog); |
3239 | dmu_objset_disown(os, FTAG); | |
40b84e7a BB |
3240 | ztest_zd_fini(zdtmp); |
3241 | out: | |
c242c188 | 3242 | (void) rw_exit(&ztest_name_lock); |
40b84e7a BB |
3243 | |
3244 | umem_free(name, MAXNAMELEN); | |
3245 | umem_free(zdtmp, sizeof (ztest_ds_t)); | |
34dc7c2f BB |
3246 | } |
3247 | ||
3248 | /* | |
3249 | * Verify that dmu_snapshot_{create,destroy,open,close} work as expected. | |
3250 | */ | |
3251 | void | |
428870ff | 3252 | ztest_dmu_snapshot_create_destroy(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3253 | { |
c242c188 | 3254 | (void) rw_enter(&ztest_name_lock, RW_READER); |
428870ff BB |
3255 | (void) ztest_snapshot_destroy(zd->zd_name, id); |
3256 | (void) ztest_snapshot_create(zd->zd_name, id); | |
c242c188 | 3257 | (void) rw_exit(&ztest_name_lock); |
34dc7c2f BB |
3258 | } |
3259 | ||
9babb374 BB |
3260 | /* |
3261 | * Cleanup non-standard snapshots and clones. | |
3262 | */ | |
3263 | void | |
428870ff | 3264 | ztest_dsl_dataset_cleanup(char *osname, uint64_t id) |
9babb374 | 3265 | { |
40b84e7a BB |
3266 | char *snap1name; |
3267 | char *clone1name; | |
3268 | char *snap2name; | |
3269 | char *clone2name; | |
3270 | char *snap3name; | |
9babb374 BB |
3271 | int error; |
3272 | ||
40b84e7a BB |
3273 | snap1name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); |
3274 | clone1name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3275 | snap2name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3276 | clone2name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3277 | snap3name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3278 | ||
b8864a23 BB |
3279 | (void) snprintf(snap1name, MAXNAMELEN, "%s@s1_%llu", |
3280 | osname, (u_longlong_t)id); | |
3281 | (void) snprintf(clone1name, MAXNAMELEN, "%s/c1_%llu", | |
3282 | osname, (u_longlong_t)id); | |
3283 | (void) snprintf(snap2name, MAXNAMELEN, "%s@s2_%llu", | |
3284 | clone1name, (u_longlong_t)id); | |
3285 | (void) snprintf(clone2name, MAXNAMELEN, "%s/c2_%llu", | |
3286 | osname, (u_longlong_t)id); | |
3287 | (void) snprintf(snap3name, MAXNAMELEN, "%s@s3_%llu", | |
3288 | clone1name, (u_longlong_t)id); | |
9babb374 | 3289 | |
45d1cae3 | 3290 | error = dmu_objset_destroy(clone2name, B_FALSE); |
9babb374 BB |
3291 | if (error && error != ENOENT) |
3292 | fatal(0, "dmu_objset_destroy(%s) = %d", clone2name, error); | |
45d1cae3 | 3293 | error = dmu_objset_destroy(snap3name, B_FALSE); |
9babb374 BB |
3294 | if (error && error != ENOENT) |
3295 | fatal(0, "dmu_objset_destroy(%s) = %d", snap3name, error); | |
45d1cae3 | 3296 | error = dmu_objset_destroy(snap2name, B_FALSE); |
9babb374 BB |
3297 | if (error && error != ENOENT) |
3298 | fatal(0, "dmu_objset_destroy(%s) = %d", snap2name, error); | |
45d1cae3 | 3299 | error = dmu_objset_destroy(clone1name, B_FALSE); |
9babb374 BB |
3300 | if (error && error != ENOENT) |
3301 | fatal(0, "dmu_objset_destroy(%s) = %d", clone1name, error); | |
45d1cae3 | 3302 | error = dmu_objset_destroy(snap1name, B_FALSE); |
9babb374 BB |
3303 | if (error && error != ENOENT) |
3304 | fatal(0, "dmu_objset_destroy(%s) = %d", snap1name, error); | |
40b84e7a BB |
3305 | |
3306 | umem_free(snap1name, MAXNAMELEN); | |
3307 | umem_free(clone1name, MAXNAMELEN); | |
3308 | umem_free(snap2name, MAXNAMELEN); | |
3309 | umem_free(clone2name, MAXNAMELEN); | |
3310 | umem_free(snap3name, MAXNAMELEN); | |
9babb374 BB |
3311 | } |
3312 | ||
3313 | /* | |
3314 | * Verify dsl_dataset_promote handles EBUSY | |
3315 | */ | |
3316 | void | |
428870ff | 3317 | ztest_dsl_dataset_promote_busy(ztest_ds_t *zd, uint64_t id) |
9babb374 | 3318 | { |
9babb374 BB |
3319 | objset_t *clone; |
3320 | dsl_dataset_t *ds; | |
40b84e7a BB |
3321 | char *snap1name; |
3322 | char *clone1name; | |
3323 | char *snap2name; | |
3324 | char *clone2name; | |
3325 | char *snap3name; | |
428870ff BB |
3326 | char *osname = zd->zd_name; |
3327 | int error; | |
9babb374 | 3328 | |
40b84e7a BB |
3329 | snap1name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); |
3330 | clone1name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3331 | snap2name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3332 | clone2name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3333 | snap3name = umem_alloc(MAXNAMELEN, UMEM_NOFAIL); | |
3334 | ||
c242c188 | 3335 | (void) rw_enter(&ztest_name_lock, RW_READER); |
9babb374 | 3336 | |
428870ff | 3337 | ztest_dsl_dataset_cleanup(osname, id); |
9babb374 | 3338 | |
b8864a23 BB |
3339 | (void) snprintf(snap1name, MAXNAMELEN, "%s@s1_%llu", |
3340 | osname, (u_longlong_t)id); | |
3341 | (void) snprintf(clone1name, MAXNAMELEN, "%s/c1_%llu", | |
3342 | osname, (u_longlong_t)id); | |
3343 | (void) snprintf(snap2name, MAXNAMELEN, "%s@s2_%llu", | |
3344 | clone1name, (u_longlong_t)id); | |
3345 | (void) snprintf(clone2name, MAXNAMELEN, "%s/c2_%llu", | |
3346 | osname, (u_longlong_t)id); | |
3347 | (void) snprintf(snap3name, MAXNAMELEN, "%s@s3_%llu", | |
3348 | clone1name, (u_longlong_t)id); | |
9babb374 BB |
3349 | |
3350 | error = dmu_objset_snapshot(osname, strchr(snap1name, '@')+1, | |
572e2857 | 3351 | NULL, NULL, B_FALSE, B_FALSE, -1); |
9babb374 BB |
3352 | if (error && error != EEXIST) { |
3353 | if (error == ENOSPC) { | |
428870ff | 3354 | ztest_record_enospc(FTAG); |
9babb374 BB |
3355 | goto out; |
3356 | } | |
3357 | fatal(0, "dmu_take_snapshot(%s) = %d", snap1name, error); | |
3358 | } | |
3359 | ||
428870ff | 3360 | error = dmu_objset_hold(snap1name, FTAG, &clone); |
9babb374 BB |
3361 | if (error) |
3362 | fatal(0, "dmu_open_snapshot(%s) = %d", snap1name, error); | |
3363 | ||
428870ff BB |
3364 | error = dmu_objset_clone(clone1name, dmu_objset_ds(clone), 0); |
3365 | dmu_objset_rele(clone, FTAG); | |
9babb374 BB |
3366 | if (error) { |
3367 | if (error == ENOSPC) { | |
428870ff | 3368 | ztest_record_enospc(FTAG); |
9babb374 BB |
3369 | goto out; |
3370 | } | |
3371 | fatal(0, "dmu_objset_create(%s) = %d", clone1name, error); | |
3372 | } | |
3373 | ||
428870ff | 3374 | error = dmu_objset_snapshot(clone1name, strchr(snap2name, '@')+1, |
572e2857 | 3375 | NULL, NULL, B_FALSE, B_FALSE, -1); |
428870ff BB |
3376 | if (error && error != EEXIST) { |
3377 | if (error == ENOSPC) { | |
3378 | ztest_record_enospc(FTAG); | |
3379 | goto out; | |
34dc7c2f | 3380 | } |
428870ff BB |
3381 | fatal(0, "dmu_open_snapshot(%s) = %d", snap2name, error); |
3382 | } | |
34dc7c2f | 3383 | |
428870ff | 3384 | error = dmu_objset_snapshot(clone1name, strchr(snap3name, '@')+1, |
572e2857 | 3385 | NULL, NULL, B_FALSE, B_FALSE, -1); |
428870ff BB |
3386 | if (error && error != EEXIST) { |
3387 | if (error == ENOSPC) { | |
3388 | ztest_record_enospc(FTAG); | |
3389 | goto out; | |
3390 | } | |
3391 | fatal(0, "dmu_open_snapshot(%s) = %d", snap3name, error); | |
3392 | } | |
34dc7c2f | 3393 | |
428870ff BB |
3394 | error = dmu_objset_hold(snap3name, FTAG, &clone); |
3395 | if (error) | |
3396 | fatal(0, "dmu_open_snapshot(%s) = %d", snap3name, error); | |
34dc7c2f | 3397 | |
428870ff BB |
3398 | error = dmu_objset_clone(clone2name, dmu_objset_ds(clone), 0); |
3399 | dmu_objset_rele(clone, FTAG); | |
3400 | if (error) { | |
3401 | if (error == ENOSPC) { | |
3402 | ztest_record_enospc(FTAG); | |
3403 | goto out; | |
3404 | } | |
3405 | fatal(0, "dmu_objset_create(%s) = %d", clone2name, error); | |
3406 | } | |
34dc7c2f | 3407 | |
428870ff BB |
3408 | error = dsl_dataset_own(snap2name, B_FALSE, FTAG, &ds); |
3409 | if (error) | |
3410 | fatal(0, "dsl_dataset_own(%s) = %d", snap2name, error); | |
3411 | error = dsl_dataset_promote(clone2name, NULL); | |
3412 | if (error != EBUSY) | |
3413 | fatal(0, "dsl_dataset_promote(%s), %d, not EBUSY", clone2name, | |
3414 | error); | |
3415 | dsl_dataset_disown(ds, FTAG); | |
34dc7c2f | 3416 | |
428870ff BB |
3417 | out: |
3418 | ztest_dsl_dataset_cleanup(osname, id); | |
34dc7c2f | 3419 | |
c242c188 | 3420 | (void) rw_exit(&ztest_name_lock); |
40b84e7a BB |
3421 | |
3422 | umem_free(snap1name, MAXNAMELEN); | |
3423 | umem_free(clone1name, MAXNAMELEN); | |
3424 | umem_free(snap2name, MAXNAMELEN); | |
3425 | umem_free(clone2name, MAXNAMELEN); | |
3426 | umem_free(snap3name, MAXNAMELEN); | |
428870ff | 3427 | } |
34dc7c2f | 3428 | |
40b84e7a BB |
3429 | #undef OD_ARRAY_SIZE |
3430 | #define OD_ARRAY_SIZE 4 | |
3431 | ||
428870ff BB |
3432 | /* |
3433 | * Verify that dmu_object_{alloc,free} work as expected. | |
3434 | */ | |
3435 | void | |
3436 | ztest_dmu_object_alloc_free(ztest_ds_t *zd, uint64_t id) | |
3437 | { | |
40b84e7a BB |
3438 | ztest_od_t *od; |
3439 | int batchsize; | |
3440 | int size; | |
d6320ddb | 3441 | int b; |
34dc7c2f | 3442 | |
40b84e7a BB |
3443 | size = sizeof(ztest_od_t) * OD_ARRAY_SIZE; |
3444 | od = umem_alloc(size, UMEM_NOFAIL); | |
3445 | batchsize = OD_ARRAY_SIZE; | |
3446 | ||
d6320ddb | 3447 | for (b = 0; b < batchsize; b++) |
40b84e7a | 3448 | ztest_od_init(od + b, id, FTAG, b, DMU_OT_UINT64_OTHER, 0, 0); |
34dc7c2f | 3449 | |
428870ff BB |
3450 | /* |
3451 | * Destroy the previous batch of objects, create a new batch, | |
3452 | * and do some I/O on the new objects. | |
3453 | */ | |
40b84e7a | 3454 | if (ztest_object_init(zd, od, size, B_TRUE) != 0) |
428870ff | 3455 | return; |
34dc7c2f | 3456 | |
428870ff BB |
3457 | while (ztest_random(4 * batchsize) != 0) |
3458 | ztest_io(zd, od[ztest_random(batchsize)].od_object, | |
3459 | ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
40b84e7a BB |
3460 | |
3461 | umem_free(od, size); | |
34dc7c2f BB |
3462 | } |
3463 | ||
40b84e7a BB |
3464 | #undef OD_ARRAY_SIZE |
3465 | #define OD_ARRAY_SIZE 2 | |
3466 | ||
34dc7c2f BB |
3467 | /* |
3468 | * Verify that dmu_{read,write} work as expected. | |
3469 | */ | |
34dc7c2f | 3470 | void |
428870ff | 3471 | ztest_dmu_read_write(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3472 | { |
40b84e7a BB |
3473 | int size; |
3474 | ztest_od_t *od; | |
3475 | ||
428870ff | 3476 | objset_t *os = zd->zd_os; |
40b84e7a BB |
3477 | size = sizeof(ztest_od_t) * OD_ARRAY_SIZE; |
3478 | od = umem_alloc(size, UMEM_NOFAIL); | |
34dc7c2f BB |
3479 | dmu_tx_t *tx; |
3480 | int i, freeit, error; | |
3481 | uint64_t n, s, txg; | |
3482 | bufwad_t *packbuf, *bigbuf, *pack, *bigH, *bigT; | |
428870ff BB |
3483 | uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize; |
3484 | uint64_t chunksize = (1000 + ztest_random(1000)) * sizeof (uint64_t); | |
34dc7c2f BB |
3485 | uint64_t regions = 997; |
3486 | uint64_t stride = 123456789ULL; | |
3487 | uint64_t width = 40; | |
3488 | int free_percent = 5; | |
3489 | ||
3490 | /* | |
3491 | * This test uses two objects, packobj and bigobj, that are always | |
3492 | * updated together (i.e. in the same tx) so that their contents are | |
3493 | * in sync and can be compared. Their contents relate to each other | |
3494 | * in a simple way: packobj is a dense array of 'bufwad' structures, | |
3495 | * while bigobj is a sparse array of the same bufwads. Specifically, | |
3496 | * for any index n, there are three bufwads that should be identical: | |
3497 | * | |
3498 | * packobj, at offset n * sizeof (bufwad_t) | |
3499 | * bigobj, at the head of the nth chunk | |
3500 | * bigobj, at the tail of the nth chunk | |
3501 | * | |
3502 | * The chunk size is arbitrary. It doesn't have to be a power of two, | |
3503 | * and it doesn't have any relation to the object blocksize. | |
3504 | * The only requirement is that it can hold at least two bufwads. | |
3505 | * | |
3506 | * Normally, we write the bufwad to each of these locations. | |
3507 | * However, free_percent of the time we instead write zeroes to | |
3508 | * packobj and perform a dmu_free_range() on bigobj. By comparing | |
3509 | * bigobj to packobj, we can verify that the DMU is correctly | |
3510 | * tracking which parts of an object are allocated and free, | |
3511 | * and that the contents of the allocated blocks are correct. | |
3512 | */ | |
3513 | ||
3514 | /* | |
3515 | * Read the directory info. If it's the first time, set things up. | |
3516 | */ | |
40b84e7a BB |
3517 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, chunksize); |
3518 | ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, chunksize); | |
34dc7c2f | 3519 | |
40b84e7a BB |
3520 | if (ztest_object_init(zd, od, size, B_FALSE) != 0) { |
3521 | umem_free(od, size); | |
428870ff | 3522 | return; |
40b84e7a | 3523 | } |
34dc7c2f | 3524 | |
428870ff BB |
3525 | bigobj = od[0].od_object; |
3526 | packobj = od[1].od_object; | |
3527 | chunksize = od[0].od_gen; | |
3528 | ASSERT(chunksize == od[1].od_gen); | |
34dc7c2f BB |
3529 | |
3530 | /* | |
3531 | * Prefetch a random chunk of the big object. | |
3532 | * Our aim here is to get some async reads in flight | |
3533 | * for blocks that we may free below; the DMU should | |
3534 | * handle this race correctly. | |
3535 | */ | |
3536 | n = ztest_random(regions) * stride + ztest_random(width); | |
3537 | s = 1 + ztest_random(2 * width - 1); | |
428870ff | 3538 | dmu_prefetch(os, bigobj, n * chunksize, s * chunksize); |
34dc7c2f BB |
3539 | |
3540 | /* | |
3541 | * Pick a random index and compute the offsets into packobj and bigobj. | |
3542 | */ | |
3543 | n = ztest_random(regions) * stride + ztest_random(width); | |
3544 | s = 1 + ztest_random(width - 1); | |
3545 | ||
3546 | packoff = n * sizeof (bufwad_t); | |
3547 | packsize = s * sizeof (bufwad_t); | |
3548 | ||
428870ff BB |
3549 | bigoff = n * chunksize; |
3550 | bigsize = s * chunksize; | |
34dc7c2f BB |
3551 | |
3552 | packbuf = umem_alloc(packsize, UMEM_NOFAIL); | |
3553 | bigbuf = umem_alloc(bigsize, UMEM_NOFAIL); | |
3554 | ||
3555 | /* | |
3556 | * free_percent of the time, free a range of bigobj rather than | |
3557 | * overwriting it. | |
3558 | */ | |
3559 | freeit = (ztest_random(100) < free_percent); | |
3560 | ||
3561 | /* | |
3562 | * Read the current contents of our objects. | |
3563 | */ | |
428870ff | 3564 | error = dmu_read(os, packobj, packoff, packsize, packbuf, |
9babb374 | 3565 | DMU_READ_PREFETCH); |
34dc7c2f | 3566 | ASSERT3U(error, ==, 0); |
428870ff | 3567 | error = dmu_read(os, bigobj, bigoff, bigsize, bigbuf, |
9babb374 | 3568 | DMU_READ_PREFETCH); |
34dc7c2f BB |
3569 | ASSERT3U(error, ==, 0); |
3570 | ||
3571 | /* | |
3572 | * Get a tx for the mods to both packobj and bigobj. | |
3573 | */ | |
3574 | tx = dmu_tx_create(os); | |
3575 | ||
428870ff | 3576 | dmu_tx_hold_write(tx, packobj, packoff, packsize); |
34dc7c2f BB |
3577 | |
3578 | if (freeit) | |
428870ff | 3579 | dmu_tx_hold_free(tx, bigobj, bigoff, bigsize); |
34dc7c2f | 3580 | else |
428870ff | 3581 | dmu_tx_hold_write(tx, bigobj, bigoff, bigsize); |
34dc7c2f | 3582 | |
428870ff BB |
3583 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); |
3584 | if (txg == 0) { | |
34dc7c2f BB |
3585 | umem_free(packbuf, packsize); |
3586 | umem_free(bigbuf, bigsize); | |
40b84e7a | 3587 | umem_free(od, size); |
34dc7c2f BB |
3588 | return; |
3589 | } | |
3590 | ||
428870ff BB |
3591 | dmu_object_set_checksum(os, bigobj, |
3592 | (enum zio_checksum)ztest_random_dsl_prop(ZFS_PROP_CHECKSUM), tx); | |
3593 | ||
3594 | dmu_object_set_compress(os, bigobj, | |
3595 | (enum zio_compress)ztest_random_dsl_prop(ZFS_PROP_COMPRESSION), tx); | |
34dc7c2f BB |
3596 | |
3597 | /* | |
3598 | * For each index from n to n + s, verify that the existing bufwad | |
3599 | * in packobj matches the bufwads at the head and tail of the | |
3600 | * corresponding chunk in bigobj. Then update all three bufwads | |
3601 | * with the new values we want to write out. | |
3602 | */ | |
3603 | for (i = 0; i < s; i++) { | |
3604 | /* LINTED */ | |
3605 | pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t)); | |
3606 | /* LINTED */ | |
428870ff | 3607 | bigH = (bufwad_t *)((char *)bigbuf + i * chunksize); |
34dc7c2f | 3608 | /* LINTED */ |
428870ff | 3609 | bigT = (bufwad_t *)((char *)bigH + chunksize) - 1; |
34dc7c2f BB |
3610 | |
3611 | ASSERT((uintptr_t)bigH - (uintptr_t)bigbuf < bigsize); | |
3612 | ASSERT((uintptr_t)bigT - (uintptr_t)bigbuf < bigsize); | |
3613 | ||
3614 | if (pack->bw_txg > txg) | |
3615 | fatal(0, "future leak: got %llx, open txg is %llx", | |
3616 | pack->bw_txg, txg); | |
3617 | ||
3618 | if (pack->bw_data != 0 && pack->bw_index != n + i) | |
3619 | fatal(0, "wrong index: got %llx, wanted %llx+%llx", | |
3620 | pack->bw_index, n, i); | |
3621 | ||
3622 | if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0) | |
3623 | fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH); | |
3624 | ||
3625 | if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0) | |
3626 | fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT); | |
3627 | ||
3628 | if (freeit) { | |
3629 | bzero(pack, sizeof (bufwad_t)); | |
3630 | } else { | |
3631 | pack->bw_index = n + i; | |
3632 | pack->bw_txg = txg; | |
3633 | pack->bw_data = 1 + ztest_random(-2ULL); | |
3634 | } | |
3635 | *bigH = *pack; | |
3636 | *bigT = *pack; | |
3637 | } | |
3638 | ||
3639 | /* | |
3640 | * We've verified all the old bufwads, and made new ones. | |
3641 | * Now write them out. | |
3642 | */ | |
428870ff | 3643 | dmu_write(os, packobj, packoff, packsize, packbuf, tx); |
34dc7c2f BB |
3644 | |
3645 | if (freeit) { | |
c242c188 | 3646 | if (ztest_opts.zo_verbose >= 7) { |
34dc7c2f BB |
3647 | (void) printf("freeing offset %llx size %llx" |
3648 | " txg %llx\n", | |
3649 | (u_longlong_t)bigoff, | |
3650 | (u_longlong_t)bigsize, | |
3651 | (u_longlong_t)txg); | |
3652 | } | |
428870ff | 3653 | VERIFY(0 == dmu_free_range(os, bigobj, bigoff, bigsize, tx)); |
34dc7c2f | 3654 | } else { |
c242c188 | 3655 | if (ztest_opts.zo_verbose >= 7) { |
34dc7c2f BB |
3656 | (void) printf("writing offset %llx size %llx" |
3657 | " txg %llx\n", | |
3658 | (u_longlong_t)bigoff, | |
3659 | (u_longlong_t)bigsize, | |
3660 | (u_longlong_t)txg); | |
3661 | } | |
428870ff | 3662 | dmu_write(os, bigobj, bigoff, bigsize, bigbuf, tx); |
34dc7c2f BB |
3663 | } |
3664 | ||
3665 | dmu_tx_commit(tx); | |
3666 | ||
3667 | /* | |
3668 | * Sanity check the stuff we just wrote. | |
3669 | */ | |
3670 | { | |
3671 | void *packcheck = umem_alloc(packsize, UMEM_NOFAIL); | |
3672 | void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL); | |
3673 | ||
428870ff | 3674 | VERIFY(0 == dmu_read(os, packobj, packoff, |
9babb374 | 3675 | packsize, packcheck, DMU_READ_PREFETCH)); |
428870ff | 3676 | VERIFY(0 == dmu_read(os, bigobj, bigoff, |
9babb374 | 3677 | bigsize, bigcheck, DMU_READ_PREFETCH)); |
34dc7c2f BB |
3678 | |
3679 | ASSERT(bcmp(packbuf, packcheck, packsize) == 0); | |
3680 | ASSERT(bcmp(bigbuf, bigcheck, bigsize) == 0); | |
3681 | ||
3682 | umem_free(packcheck, packsize); | |
3683 | umem_free(bigcheck, bigsize); | |
3684 | } | |
3685 | ||
3686 | umem_free(packbuf, packsize); | |
3687 | umem_free(bigbuf, bigsize); | |
40b84e7a | 3688 | umem_free(od, size); |
34dc7c2f BB |
3689 | } |
3690 | ||
9babb374 BB |
3691 | void |
3692 | compare_and_update_pbbufs(uint64_t s, bufwad_t *packbuf, bufwad_t *bigbuf, | |
428870ff | 3693 | uint64_t bigsize, uint64_t n, uint64_t chunksize, uint64_t txg) |
9babb374 BB |
3694 | { |
3695 | uint64_t i; | |
3696 | bufwad_t *pack; | |
3697 | bufwad_t *bigH; | |
3698 | bufwad_t *bigT; | |
3699 | ||
3700 | /* | |
3701 | * For each index from n to n + s, verify that the existing bufwad | |
3702 | * in packobj matches the bufwads at the head and tail of the | |
3703 | * corresponding chunk in bigobj. Then update all three bufwads | |
3704 | * with the new values we want to write out. | |
3705 | */ | |
3706 | for (i = 0; i < s; i++) { | |
3707 | /* LINTED */ | |
3708 | pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t)); | |
3709 | /* LINTED */ | |
428870ff | 3710 | bigH = (bufwad_t *)((char *)bigbuf + i * chunksize); |
9babb374 | 3711 | /* LINTED */ |
428870ff | 3712 | bigT = (bufwad_t *)((char *)bigH + chunksize) - 1; |
9babb374 BB |
3713 | |
3714 | ASSERT((uintptr_t)bigH - (uintptr_t)bigbuf < bigsize); | |
3715 | ASSERT((uintptr_t)bigT - (uintptr_t)bigbuf < bigsize); | |
3716 | ||
3717 | if (pack->bw_txg > txg) | |
3718 | fatal(0, "future leak: got %llx, open txg is %llx", | |
3719 | pack->bw_txg, txg); | |
3720 | ||
3721 | if (pack->bw_data != 0 && pack->bw_index != n + i) | |
3722 | fatal(0, "wrong index: got %llx, wanted %llx+%llx", | |
3723 | pack->bw_index, n, i); | |
3724 | ||
3725 | if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0) | |
3726 | fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH); | |
3727 | ||
3728 | if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0) | |
3729 | fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT); | |
3730 | ||
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 | ||
40b84e7a BB |
3740 | #undef OD_ARRAY_SIZE |
3741 | #define OD_ARRAY_SIZE 2 | |
3742 | ||
9babb374 | 3743 | void |
428870ff | 3744 | ztest_dmu_read_write_zcopy(ztest_ds_t *zd, uint64_t id) |
9babb374 | 3745 | { |
428870ff | 3746 | objset_t *os = zd->zd_os; |
40b84e7a | 3747 | ztest_od_t *od; |
9babb374 BB |
3748 | dmu_tx_t *tx; |
3749 | uint64_t i; | |
3750 | int error; | |
40b84e7a | 3751 | int size; |
9babb374 BB |
3752 | uint64_t n, s, txg; |
3753 | bufwad_t *packbuf, *bigbuf; | |
428870ff BB |
3754 | uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize; |
3755 | uint64_t blocksize = ztest_random_blocksize(); | |
3756 | uint64_t chunksize = blocksize; | |
9babb374 BB |
3757 | uint64_t regions = 997; |
3758 | uint64_t stride = 123456789ULL; | |
3759 | uint64_t width = 9; | |
3760 | dmu_buf_t *bonus_db; | |
3761 | arc_buf_t **bigbuf_arcbufs; | |
428870ff | 3762 | dmu_object_info_t doi; |
9babb374 | 3763 | |
40b84e7a BB |
3764 | size = sizeof(ztest_od_t) * OD_ARRAY_SIZE; |
3765 | od = umem_alloc(size, UMEM_NOFAIL); | |
3766 | ||
9babb374 BB |
3767 | /* |
3768 | * This test uses two objects, packobj and bigobj, that are always | |
3769 | * updated together (i.e. in the same tx) so that their contents are | |
3770 | * in sync and can be compared. Their contents relate to each other | |
3771 | * in a simple way: packobj is a dense array of 'bufwad' structures, | |
3772 | * while bigobj is a sparse array of the same bufwads. Specifically, | |
3773 | * for any index n, there are three bufwads that should be identical: | |
3774 | * | |
3775 | * packobj, at offset n * sizeof (bufwad_t) | |
3776 | * bigobj, at the head of the nth chunk | |
3777 | * bigobj, at the tail of the nth chunk | |
3778 | * | |
3779 | * The chunk size is set equal to bigobj block size so that | |
3780 | * dmu_assign_arcbuf() can be tested for object updates. | |
3781 | */ | |
3782 | ||
3783 | /* | |
3784 | * Read the directory info. If it's the first time, set things up. | |
3785 | */ | |
40b84e7a BB |
3786 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0); |
3787 | ztest_od_init(od + 1, id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, chunksize); | |
3788 | ||
9babb374 | 3789 | |
40b84e7a BB |
3790 | if (ztest_object_init(zd, od, size, B_FALSE) != 0) { |
3791 | umem_free(od, size); | |
428870ff | 3792 | return; |
40b84e7a | 3793 | } |
9babb374 | 3794 | |
428870ff BB |
3795 | bigobj = od[0].od_object; |
3796 | packobj = od[1].od_object; | |
3797 | blocksize = od[0].od_blocksize; | |
3798 | chunksize = blocksize; | |
3799 | ASSERT(chunksize == od[1].od_gen); | |
9babb374 | 3800 | |
428870ff BB |
3801 | VERIFY(dmu_object_info(os, bigobj, &doi) == 0); |
3802 | VERIFY(ISP2(doi.doi_data_block_size)); | |
3803 | VERIFY(chunksize == doi.doi_data_block_size); | |
3804 | VERIFY(chunksize >= 2 * sizeof (bufwad_t)); | |
9babb374 BB |
3805 | |
3806 | /* | |
3807 | * Pick a random index and compute the offsets into packobj and bigobj. | |
3808 | */ | |
3809 | n = ztest_random(regions) * stride + ztest_random(width); | |
3810 | s = 1 + ztest_random(width - 1); | |
3811 | ||
3812 | packoff = n * sizeof (bufwad_t); | |
3813 | packsize = s * sizeof (bufwad_t); | |
3814 | ||
428870ff BB |
3815 | bigoff = n * chunksize; |
3816 | bigsize = s * chunksize; | |
9babb374 BB |
3817 | |
3818 | packbuf = umem_zalloc(packsize, UMEM_NOFAIL); | |
3819 | bigbuf = umem_zalloc(bigsize, UMEM_NOFAIL); | |
3820 | ||
428870ff | 3821 | VERIFY3U(0, ==, dmu_bonus_hold(os, bigobj, FTAG, &bonus_db)); |
9babb374 BB |
3822 | |
3823 | bigbuf_arcbufs = umem_zalloc(2 * s * sizeof (arc_buf_t *), UMEM_NOFAIL); | |
3824 | ||
3825 | /* | |
3826 | * Iteration 0 test zcopy for DB_UNCACHED dbufs. | |
3827 | * Iteration 1 test zcopy to already referenced dbufs. | |
3828 | * Iteration 2 test zcopy to dirty dbuf in the same txg. | |
3829 | * Iteration 3 test zcopy to dbuf dirty in previous txg. | |
3830 | * Iteration 4 test zcopy when dbuf is no longer dirty. | |
3831 | * Iteration 5 test zcopy when it can't be done. | |
3832 | * Iteration 6 one more zcopy write. | |
3833 | */ | |
3834 | for (i = 0; i < 7; i++) { | |
3835 | uint64_t j; | |
3836 | uint64_t off; | |
3837 | ||
3838 | /* | |
3839 | * In iteration 5 (i == 5) use arcbufs | |
3840 | * that don't match bigobj blksz to test | |
3841 | * dmu_assign_arcbuf() when it can't directly | |
3842 | * assign an arcbuf to a dbuf. | |
3843 | */ | |
3844 | for (j = 0; j < s; j++) { | |
3845 | if (i != 5) { | |
3846 | bigbuf_arcbufs[j] = | |
428870ff | 3847 | dmu_request_arcbuf(bonus_db, chunksize); |
9babb374 BB |
3848 | } else { |
3849 | bigbuf_arcbufs[2 * j] = | |
428870ff | 3850 | dmu_request_arcbuf(bonus_db, chunksize / 2); |
9babb374 | 3851 | bigbuf_arcbufs[2 * j + 1] = |
428870ff | 3852 | dmu_request_arcbuf(bonus_db, chunksize / 2); |
9babb374 BB |
3853 | } |
3854 | } | |
3855 | ||
3856 | /* | |
3857 | * Get a tx for the mods to both packobj and bigobj. | |
3858 | */ | |
3859 | tx = dmu_tx_create(os); | |
3860 | ||
428870ff BB |
3861 | dmu_tx_hold_write(tx, packobj, packoff, packsize); |
3862 | dmu_tx_hold_write(tx, bigobj, bigoff, bigsize); | |
9babb374 | 3863 | |
428870ff BB |
3864 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); |
3865 | if (txg == 0) { | |
9babb374 BB |
3866 | umem_free(packbuf, packsize); |
3867 | umem_free(bigbuf, bigsize); | |
3868 | for (j = 0; j < s; j++) { | |
3869 | if (i != 5) { | |
3870 | dmu_return_arcbuf(bigbuf_arcbufs[j]); | |
3871 | } else { | |
3872 | dmu_return_arcbuf( | |
3873 | bigbuf_arcbufs[2 * j]); | |
3874 | dmu_return_arcbuf( | |
3875 | bigbuf_arcbufs[2 * j + 1]); | |
3876 | } | |
3877 | } | |
3878 | umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *)); | |
40b84e7a | 3879 | umem_free(od, size); |
9babb374 BB |
3880 | dmu_buf_rele(bonus_db, FTAG); |
3881 | return; | |
3882 | } | |
3883 | ||
9babb374 BB |
3884 | /* |
3885 | * 50% of the time don't read objects in the 1st iteration to | |
3886 | * test dmu_assign_arcbuf() for the case when there're no | |
3887 | * existing dbufs for the specified offsets. | |
3888 | */ | |
3889 | if (i != 0 || ztest_random(2) != 0) { | |
428870ff | 3890 | error = dmu_read(os, packobj, packoff, |
9babb374 BB |
3891 | packsize, packbuf, DMU_READ_PREFETCH); |
3892 | ASSERT3U(error, ==, 0); | |
428870ff | 3893 | error = dmu_read(os, bigobj, bigoff, bigsize, |
9babb374 BB |
3894 | bigbuf, DMU_READ_PREFETCH); |
3895 | ASSERT3U(error, ==, 0); | |
3896 | } | |
3897 | compare_and_update_pbbufs(s, packbuf, bigbuf, bigsize, | |
428870ff | 3898 | n, chunksize, txg); |
9babb374 BB |
3899 | |
3900 | /* | |
3901 | * We've verified all the old bufwads, and made new ones. | |
3902 | * Now write them out. | |
3903 | */ | |
428870ff | 3904 | dmu_write(os, packobj, packoff, packsize, packbuf, tx); |
c242c188 | 3905 | if (ztest_opts.zo_verbose >= 7) { |
9babb374 BB |
3906 | (void) printf("writing offset %llx size %llx" |
3907 | " txg %llx\n", | |
3908 | (u_longlong_t)bigoff, | |
3909 | (u_longlong_t)bigsize, | |
3910 | (u_longlong_t)txg); | |
3911 | } | |
428870ff | 3912 | for (off = bigoff, j = 0; j < s; j++, off += chunksize) { |
9babb374 BB |
3913 | dmu_buf_t *dbt; |
3914 | if (i != 5) { | |
3915 | bcopy((caddr_t)bigbuf + (off - bigoff), | |
428870ff | 3916 | bigbuf_arcbufs[j]->b_data, chunksize); |
9babb374 BB |
3917 | } else { |
3918 | bcopy((caddr_t)bigbuf + (off - bigoff), | |
3919 | bigbuf_arcbufs[2 * j]->b_data, | |
428870ff | 3920 | chunksize / 2); |
9babb374 | 3921 | bcopy((caddr_t)bigbuf + (off - bigoff) + |
428870ff | 3922 | chunksize / 2, |
9babb374 | 3923 | bigbuf_arcbufs[2 * j + 1]->b_data, |
428870ff | 3924 | chunksize / 2); |
9babb374 BB |
3925 | } |
3926 | ||
3927 | if (i == 1) { | |
428870ff BB |
3928 | VERIFY(dmu_buf_hold(os, bigobj, off, |
3929 | FTAG, &dbt, DMU_READ_NO_PREFETCH) == 0); | |
9babb374 BB |
3930 | } |
3931 | if (i != 5) { | |
3932 | dmu_assign_arcbuf(bonus_db, off, | |
3933 | bigbuf_arcbufs[j], tx); | |
3934 | } else { | |
3935 | dmu_assign_arcbuf(bonus_db, off, | |
3936 | bigbuf_arcbufs[2 * j], tx); | |
3937 | dmu_assign_arcbuf(bonus_db, | |
428870ff | 3938 | off + chunksize / 2, |
9babb374 BB |
3939 | bigbuf_arcbufs[2 * j + 1], tx); |
3940 | } | |
3941 | if (i == 1) { | |
3942 | dmu_buf_rele(dbt, FTAG); | |
3943 | } | |
3944 | } | |
3945 | dmu_tx_commit(tx); | |
3946 | ||
3947 | /* | |
3948 | * Sanity check the stuff we just wrote. | |
3949 | */ | |
3950 | { | |
3951 | void *packcheck = umem_alloc(packsize, UMEM_NOFAIL); | |
3952 | void *bigcheck = umem_alloc(bigsize, UMEM_NOFAIL); | |
3953 | ||
428870ff | 3954 | VERIFY(0 == dmu_read(os, packobj, packoff, |
9babb374 | 3955 | packsize, packcheck, DMU_READ_PREFETCH)); |
428870ff | 3956 | VERIFY(0 == dmu_read(os, bigobj, bigoff, |
9babb374 BB |
3957 | bigsize, bigcheck, DMU_READ_PREFETCH)); |
3958 | ||
3959 | ASSERT(bcmp(packbuf, packcheck, packsize) == 0); | |
3960 | ASSERT(bcmp(bigbuf, bigcheck, bigsize) == 0); | |
3961 | ||
3962 | umem_free(packcheck, packsize); | |
3963 | umem_free(bigcheck, bigsize); | |
3964 | } | |
3965 | if (i == 2) { | |
3966 | txg_wait_open(dmu_objset_pool(os), 0); | |
3967 | } else if (i == 3) { | |
3968 | txg_wait_synced(dmu_objset_pool(os), 0); | |
3969 | } | |
3970 | } | |
3971 | ||
3972 | dmu_buf_rele(bonus_db, FTAG); | |
3973 | umem_free(packbuf, packsize); | |
3974 | umem_free(bigbuf, bigsize); | |
3975 | umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *)); | |
40b84e7a | 3976 | umem_free(od, size); |
9babb374 BB |
3977 | } |
3978 | ||
428870ff | 3979 | /* ARGSUSED */ |
34dc7c2f | 3980 | void |
428870ff | 3981 | ztest_dmu_write_parallel(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 3982 | { |
40b84e7a BB |
3983 | ztest_od_t *od; |
3984 | ||
3985 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); | |
428870ff BB |
3986 | uint64_t offset = (1ULL << (ztest_random(20) + 43)) + |
3987 | (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
34dc7c2f BB |
3988 | |
3989 | /* | |
428870ff BB |
3990 | * Have multiple threads write to large offsets in an object |
3991 | * to verify that parallel writes to an object -- even to the | |
3992 | * same blocks within the object -- doesn't cause any trouble. | |
34dc7c2f | 3993 | */ |
40b84e7a | 3994 | ztest_od_init(od, ID_PARALLEL, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0); |
9babb374 | 3995 | |
40b84e7a | 3996 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) |
34dc7c2f | 3997 | return; |
34dc7c2f | 3998 | |
428870ff | 3999 | while (ztest_random(10) != 0) |
40b84e7a BB |
4000 | ztest_io(zd, od->od_object, offset); |
4001 | ||
4002 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff | 4003 | } |
34dc7c2f | 4004 | |
428870ff BB |
4005 | void |
4006 | ztest_dmu_prealloc(ztest_ds_t *zd, uint64_t id) | |
4007 | { | |
40b84e7a | 4008 | ztest_od_t *od; |
428870ff BB |
4009 | uint64_t offset = (1ULL << (ztest_random(4) + SPA_MAXBLOCKSHIFT)) + |
4010 | (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT); | |
4011 | uint64_t count = ztest_random(20) + 1; | |
4012 | uint64_t blocksize = ztest_random_blocksize(); | |
4013 | void *data; | |
34dc7c2f | 4014 | |
40b84e7a | 4015 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
34dc7c2f | 4016 | |
40b84e7a BB |
4017 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0); |
4018 | ||
4019 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), !ztest_random(2)) != 0) { | |
4020 | umem_free(od, sizeof(ztest_od_t)); | |
34dc7c2f | 4021 | return; |
40b84e7a | 4022 | } |
34dc7c2f | 4023 | |
40b84e7a BB |
4024 | if (ztest_truncate(zd, od->od_object, offset, count * blocksize) != 0) { |
4025 | umem_free(od, sizeof(ztest_od_t)); | |
34dc7c2f | 4026 | return; |
40b84e7a | 4027 | } |
34dc7c2f | 4028 | |
40b84e7a | 4029 | ztest_prealloc(zd, od->od_object, offset, count * blocksize); |
34dc7c2f | 4030 | |
428870ff | 4031 | data = umem_zalloc(blocksize, UMEM_NOFAIL); |
34dc7c2f | 4032 | |
428870ff BB |
4033 | while (ztest_random(count) != 0) { |
4034 | uint64_t randoff = offset + (ztest_random(count) * blocksize); | |
40b84e7a | 4035 | if (ztest_write(zd, od->od_object, randoff, blocksize, |
428870ff BB |
4036 | data) != 0) |
4037 | break; | |
4038 | while (ztest_random(4) != 0) | |
40b84e7a | 4039 | ztest_io(zd, od->od_object, randoff); |
9babb374 | 4040 | } |
34dc7c2f | 4041 | |
428870ff | 4042 | umem_free(data, blocksize); |
40b84e7a | 4043 | umem_free(od, sizeof(ztest_od_t)); |
34dc7c2f BB |
4044 | } |
4045 | ||
4046 | /* | |
4047 | * Verify that zap_{create,destroy,add,remove,update} work as expected. | |
4048 | */ | |
4049 | #define ZTEST_ZAP_MIN_INTS 1 | |
4050 | #define ZTEST_ZAP_MAX_INTS 4 | |
4051 | #define ZTEST_ZAP_MAX_PROPS 1000 | |
4052 | ||
4053 | void | |
428870ff | 4054 | ztest_zap(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4055 | { |
428870ff | 4056 | objset_t *os = zd->zd_os; |
40b84e7a | 4057 | ztest_od_t *od; |
34dc7c2f BB |
4058 | uint64_t object; |
4059 | uint64_t txg, last_txg; | |
4060 | uint64_t value[ZTEST_ZAP_MAX_INTS]; | |
4061 | uint64_t zl_ints, zl_intsize, prop; | |
4062 | int i, ints; | |
4063 | dmu_tx_t *tx; | |
4064 | char propname[100], txgname[100]; | |
4065 | int error; | |
34dc7c2f BB |
4066 | char *hc[2] = { "s.acl.h", ".s.open.h.hyLZlg" }; |
4067 | ||
40b84e7a BB |
4068 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
4069 | ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0); | |
34dc7c2f | 4070 | |
40b84e7a BB |
4071 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
4072 | !ztest_random(2)) != 0) | |
4073 | goto out; | |
34dc7c2f | 4074 | |
40b84e7a | 4075 | object = od->od_object; |
34dc7c2f | 4076 | |
428870ff BB |
4077 | /* |
4078 | * Generate a known hash collision, and verify that | |
4079 | * we can lookup and remove both entries. | |
4080 | */ | |
4081 | tx = dmu_tx_create(os); | |
4082 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); | |
4083 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4084 | if (txg == 0) | |
40b84e7a | 4085 | goto out; |
428870ff BB |
4086 | for (i = 0; i < 2; i++) { |
4087 | value[i] = i; | |
4088 | VERIFY3U(0, ==, zap_add(os, object, hc[i], sizeof (uint64_t), | |
4089 | 1, &value[i], tx)); | |
4090 | } | |
4091 | for (i = 0; i < 2; i++) { | |
4092 | VERIFY3U(EEXIST, ==, zap_add(os, object, hc[i], | |
4093 | sizeof (uint64_t), 1, &value[i], tx)); | |
4094 | VERIFY3U(0, ==, | |
4095 | zap_length(os, object, hc[i], &zl_intsize, &zl_ints)); | |
4096 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
4097 | ASSERT3U(zl_ints, ==, 1); | |
4098 | } | |
4099 | for (i = 0; i < 2; i++) { | |
4100 | VERIFY3U(0, ==, zap_remove(os, object, hc[i], tx)); | |
34dc7c2f | 4101 | } |
428870ff | 4102 | dmu_tx_commit(tx); |
34dc7c2f | 4103 | |
428870ff BB |
4104 | /* |
4105 | * Generate a buch of random entries. | |
4106 | */ | |
34dc7c2f BB |
4107 | ints = MAX(ZTEST_ZAP_MIN_INTS, object % ZTEST_ZAP_MAX_INTS); |
4108 | ||
4109 | prop = ztest_random(ZTEST_ZAP_MAX_PROPS); | |
4110 | (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop); | |
4111 | (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop); | |
4112 | bzero(value, sizeof (value)); | |
4113 | last_txg = 0; | |
4114 | ||
4115 | /* | |
4116 | * If these zap entries already exist, validate their contents. | |
4117 | */ | |
4118 | error = zap_length(os, object, txgname, &zl_intsize, &zl_ints); | |
4119 | if (error == 0) { | |
4120 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
4121 | ASSERT3U(zl_ints, ==, 1); | |
4122 | ||
4123 | VERIFY(zap_lookup(os, object, txgname, zl_intsize, | |
4124 | zl_ints, &last_txg) == 0); | |
4125 | ||
4126 | VERIFY(zap_length(os, object, propname, &zl_intsize, | |
4127 | &zl_ints) == 0); | |
4128 | ||
4129 | ASSERT3U(zl_intsize, ==, sizeof (uint64_t)); | |
4130 | ASSERT3U(zl_ints, ==, ints); | |
4131 | ||
4132 | VERIFY(zap_lookup(os, object, propname, zl_intsize, | |
4133 | zl_ints, value) == 0); | |
4134 | ||
4135 | for (i = 0; i < ints; i++) { | |
4136 | ASSERT3U(value[i], ==, last_txg + object + i); | |
4137 | } | |
4138 | } else { | |
4139 | ASSERT3U(error, ==, ENOENT); | |
4140 | } | |
4141 | ||
4142 | /* | |
4143 | * Atomically update two entries in our zap object. | |
4144 | * The first is named txg_%llu, and contains the txg | |
4145 | * in which the property was last updated. The second | |
4146 | * is named prop_%llu, and the nth element of its value | |
4147 | * should be txg + object + n. | |
4148 | */ | |
4149 | tx = dmu_tx_create(os); | |
428870ff BB |
4150 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
4151 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4152 | if (txg == 0) | |
40b84e7a | 4153 | goto out; |
34dc7c2f BB |
4154 | |
4155 | if (last_txg > txg) | |
4156 | fatal(0, "zap future leak: old %llu new %llu", last_txg, txg); | |
4157 | ||
4158 | for (i = 0; i < ints; i++) | |
4159 | value[i] = txg + object + i; | |
4160 | ||
428870ff BB |
4161 | VERIFY3U(0, ==, zap_update(os, object, txgname, sizeof (uint64_t), |
4162 | 1, &txg, tx)); | |
4163 | VERIFY3U(0, ==, zap_update(os, object, propname, sizeof (uint64_t), | |
4164 | ints, value, tx)); | |
34dc7c2f BB |
4165 | |
4166 | dmu_tx_commit(tx); | |
4167 | ||
4168 | /* | |
4169 | * Remove a random pair of entries. | |
4170 | */ | |
4171 | prop = ztest_random(ZTEST_ZAP_MAX_PROPS); | |
4172 | (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop); | |
4173 | (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop); | |
4174 | ||
4175 | error = zap_length(os, object, txgname, &zl_intsize, &zl_ints); | |
4176 | ||
4177 | if (error == ENOENT) | |
40b84e7a | 4178 | goto out; |
34dc7c2f BB |
4179 | |
4180 | ASSERT3U(error, ==, 0); | |
4181 | ||
4182 | tx = dmu_tx_create(os); | |
428870ff BB |
4183 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
4184 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4185 | if (txg == 0) | |
40b84e7a | 4186 | goto out; |
428870ff BB |
4187 | VERIFY3U(0, ==, zap_remove(os, object, txgname, tx)); |
4188 | VERIFY3U(0, ==, zap_remove(os, object, propname, tx)); | |
4189 | dmu_tx_commit(tx); | |
40b84e7a BB |
4190 | out: |
4191 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff | 4192 | } |
34dc7c2f | 4193 | |
428870ff BB |
4194 | /* |
4195 | * Testcase to test the upgrading of a microzap to fatzap. | |
4196 | */ | |
4197 | void | |
4198 | ztest_fzap(ztest_ds_t *zd, uint64_t id) | |
4199 | { | |
4200 | objset_t *os = zd->zd_os; | |
40b84e7a | 4201 | ztest_od_t *od; |
428870ff | 4202 | uint64_t object, txg; |
d6320ddb | 4203 | int i; |
34dc7c2f | 4204 | |
40b84e7a BB |
4205 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
4206 | ztest_od_init(od, id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0); | |
428870ff | 4207 | |
40b84e7a BB |
4208 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), |
4209 | !ztest_random(2)) != 0) | |
4210 | goto out; | |
4211 | object = od->od_object; | |
34dc7c2f BB |
4212 | |
4213 | /* | |
428870ff BB |
4214 | * Add entries to this ZAP and make sure it spills over |
4215 | * and gets upgraded to a fatzap. Also, since we are adding | |
4216 | * 2050 entries we should see ptrtbl growth and leaf-block split. | |
34dc7c2f | 4217 | */ |
d6320ddb | 4218 | for (i = 0; i < 2050; i++) { |
428870ff BB |
4219 | char name[MAXNAMELEN]; |
4220 | uint64_t value = i; | |
4221 | dmu_tx_t *tx; | |
4222 | int error; | |
34dc7c2f | 4223 | |
428870ff | 4224 | (void) snprintf(name, sizeof (name), "fzap-%llu-%llu", |
b8864a23 | 4225 | (u_longlong_t)id, (u_longlong_t)value); |
428870ff BB |
4226 | |
4227 | tx = dmu_tx_create(os); | |
4228 | dmu_tx_hold_zap(tx, object, B_TRUE, name); | |
4229 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4230 | if (txg == 0) | |
40b84e7a | 4231 | goto out; |
428870ff BB |
4232 | error = zap_add(os, object, name, sizeof (uint64_t), 1, |
4233 | &value, tx); | |
4234 | ASSERT(error == 0 || error == EEXIST); | |
4235 | dmu_tx_commit(tx); | |
34dc7c2f | 4236 | } |
40b84e7a BB |
4237 | out: |
4238 | umem_free(od, sizeof(ztest_od_t)); | |
34dc7c2f BB |
4239 | } |
4240 | ||
428870ff | 4241 | /* ARGSUSED */ |
34dc7c2f | 4242 | void |
428870ff | 4243 | ztest_zap_parallel(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4244 | { |
428870ff | 4245 | objset_t *os = zd->zd_os; |
40b84e7a | 4246 | ztest_od_t *od; |
34dc7c2f BB |
4247 | uint64_t txg, object, count, wsize, wc, zl_wsize, zl_wc; |
4248 | dmu_tx_t *tx; | |
4249 | int i, namelen, error; | |
428870ff | 4250 | int micro = ztest_random(2); |
34dc7c2f BB |
4251 | char name[20], string_value[20]; |
4252 | void *data; | |
4253 | ||
40b84e7a BB |
4254 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
4255 | ztest_od_init(od, ID_PARALLEL, FTAG, micro, DMU_OT_ZAP_OTHER, 0, 0); | |
428870ff | 4256 | |
40b84e7a BB |
4257 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
4258 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff | 4259 | return; |
40b84e7a | 4260 | } |
428870ff | 4261 | |
40b84e7a | 4262 | object = od->od_object; |
428870ff | 4263 | |
34dc7c2f BB |
4264 | /* |
4265 | * Generate a random name of the form 'xxx.....' where each | |
4266 | * x is a random printable character and the dots are dots. | |
4267 | * There are 94 such characters, and the name length goes from | |
4268 | * 6 to 20, so there are 94^3 * 15 = 12,458,760 possible names. | |
4269 | */ | |
4270 | namelen = ztest_random(sizeof (name) - 5) + 5 + 1; | |
4271 | ||
4272 | for (i = 0; i < 3; i++) | |
4273 | name[i] = '!' + ztest_random('~' - '!' + 1); | |
4274 | for (; i < namelen - 1; i++) | |
4275 | name[i] = '.'; | |
4276 | name[i] = '\0'; | |
4277 | ||
428870ff | 4278 | if ((namelen & 1) || micro) { |
34dc7c2f BB |
4279 | wsize = sizeof (txg); |
4280 | wc = 1; | |
4281 | data = &txg; | |
4282 | } else { | |
4283 | wsize = 1; | |
4284 | wc = namelen; | |
4285 | data = string_value; | |
4286 | } | |
4287 | ||
4288 | count = -1ULL; | |
4289 | VERIFY(zap_count(os, object, &count) == 0); | |
4290 | ASSERT(count != -1ULL); | |
4291 | ||
4292 | /* | |
4293 | * Select an operation: length, lookup, add, update, remove. | |
4294 | */ | |
4295 | i = ztest_random(5); | |
4296 | ||
4297 | if (i >= 2) { | |
4298 | tx = dmu_tx_create(os); | |
428870ff BB |
4299 | dmu_tx_hold_zap(tx, object, B_TRUE, NULL); |
4300 | txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG); | |
4301 | if (txg == 0) | |
34dc7c2f | 4302 | return; |
34dc7c2f BB |
4303 | bcopy(name, string_value, namelen); |
4304 | } else { | |
4305 | tx = NULL; | |
4306 | txg = 0; | |
4307 | bzero(string_value, namelen); | |
4308 | } | |
4309 | ||
4310 | switch (i) { | |
4311 | ||
4312 | case 0: | |
4313 | error = zap_length(os, object, name, &zl_wsize, &zl_wc); | |
4314 | if (error == 0) { | |
4315 | ASSERT3U(wsize, ==, zl_wsize); | |
4316 | ASSERT3U(wc, ==, zl_wc); | |
4317 | } else { | |
4318 | ASSERT3U(error, ==, ENOENT); | |
4319 | } | |
4320 | break; | |
4321 | ||
4322 | case 1: | |
4323 | error = zap_lookup(os, object, name, wsize, wc, data); | |
4324 | if (error == 0) { | |
4325 | if (data == string_value && | |
4326 | bcmp(name, data, namelen) != 0) | |
4327 | fatal(0, "name '%s' != val '%s' len %d", | |
4328 | name, data, namelen); | |
4329 | } else { | |
4330 | ASSERT3U(error, ==, ENOENT); | |
4331 | } | |
4332 | break; | |
4333 | ||
4334 | case 2: | |
4335 | error = zap_add(os, object, name, wsize, wc, data, tx); | |
4336 | ASSERT(error == 0 || error == EEXIST); | |
4337 | break; | |
4338 | ||
4339 | case 3: | |
4340 | VERIFY(zap_update(os, object, name, wsize, wc, data, tx) == 0); | |
4341 | break; | |
4342 | ||
4343 | case 4: | |
4344 | error = zap_remove(os, object, name, tx); | |
4345 | ASSERT(error == 0 || error == ENOENT); | |
4346 | break; | |
4347 | } | |
4348 | ||
4349 | if (tx != NULL) | |
4350 | dmu_tx_commit(tx); | |
40b84e7a BB |
4351 | |
4352 | umem_free(od, sizeof(ztest_od_t)); | |
34dc7c2f BB |
4353 | } |
4354 | ||
428870ff BB |
4355 | /* |
4356 | * Commit callback data. | |
4357 | */ | |
4358 | typedef struct ztest_cb_data { | |
4359 | list_node_t zcd_node; | |
4360 | uint64_t zcd_txg; | |
4361 | int zcd_expected_err; | |
4362 | boolean_t zcd_added; | |
4363 | boolean_t zcd_called; | |
4364 | spa_t *zcd_spa; | |
4365 | } ztest_cb_data_t; | |
4366 | ||
4367 | /* This is the actual commit callback function */ | |
4368 | static void | |
4369 | ztest_commit_callback(void *arg, int error) | |
4370 | { | |
4371 | ztest_cb_data_t *data = arg; | |
4372 | uint64_t synced_txg; | |
4373 | ||
4374 | VERIFY(data != NULL); | |
4375 | VERIFY3S(data->zcd_expected_err, ==, error); | |
4376 | VERIFY(!data->zcd_called); | |
4377 | ||
4378 | synced_txg = spa_last_synced_txg(data->zcd_spa); | |
4379 | if (data->zcd_txg > synced_txg) | |
4380 | fatal(0, "commit callback of txg %" PRIu64 " called prematurely" | |
4381 | ", last synced txg = %" PRIu64 "\n", data->zcd_txg, | |
4382 | synced_txg); | |
4383 | ||
4384 | data->zcd_called = B_TRUE; | |
4385 | ||
4386 | if (error == ECANCELED) { | |
4387 | ASSERT3U(data->zcd_txg, ==, 0); | |
4388 | ASSERT(!data->zcd_added); | |
4389 | ||
4390 | /* | |
4391 | * The private callback data should be destroyed here, but | |
4392 | * since we are going to check the zcd_called field after | |
4393 | * dmu_tx_abort(), we will destroy it there. | |
4394 | */ | |
4395 | return; | |
4396 | } | |
4397 | ||
090ff092 | 4398 | ASSERT(data->zcd_added); |
428870ff BB |
4399 | ASSERT3U(data->zcd_txg, !=, 0); |
4400 | ||
1e33ac1e | 4401 | (void) mutex_enter(&zcl.zcl_callbacks_lock); |
090ff092 RC |
4402 | |
4403 | /* See if this cb was called more quickly */ | |
4404 | if ((synced_txg - data->zcd_txg) < zc_min_txg_delay) | |
4405 | zc_min_txg_delay = synced_txg - data->zcd_txg; | |
4406 | ||
4407 | /* Remove our callback from the list */ | |
428870ff | 4408 | list_remove(&zcl.zcl_callbacks, data); |
090ff092 | 4409 | |
1e33ac1e | 4410 | (void) mutex_exit(&zcl.zcl_callbacks_lock); |
428870ff | 4411 | |
428870ff BB |
4412 | umem_free(data, sizeof (ztest_cb_data_t)); |
4413 | } | |
4414 | ||
4415 | /* Allocate and initialize callback data structure */ | |
4416 | static ztest_cb_data_t * | |
4417 | ztest_create_cb_data(objset_t *os, uint64_t txg) | |
4418 | { | |
4419 | ztest_cb_data_t *cb_data; | |
4420 | ||
4421 | cb_data = umem_zalloc(sizeof (ztest_cb_data_t), UMEM_NOFAIL); | |
4422 | ||
4423 | cb_data->zcd_txg = txg; | |
4424 | cb_data->zcd_spa = dmu_objset_spa(os); | |
1e33ac1e | 4425 | list_link_init(&cb_data->zcd_node); |
428870ff BB |
4426 | |
4427 | return (cb_data); | |
4428 | } | |
4429 | ||
428870ff BB |
4430 | /* |
4431 | * Commit callback test. | |
4432 | */ | |
34dc7c2f | 4433 | void |
428870ff BB |
4434 | ztest_dmu_commit_callbacks(ztest_ds_t *zd, uint64_t id) |
4435 | { | |
4436 | objset_t *os = zd->zd_os; | |
40b84e7a | 4437 | ztest_od_t *od; |
428870ff BB |
4438 | dmu_tx_t *tx; |
4439 | ztest_cb_data_t *cb_data[3], *tmp_cb; | |
4440 | uint64_t old_txg, txg; | |
090ff092 | 4441 | int i, error = 0; |
428870ff | 4442 | |
40b84e7a BB |
4443 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
4444 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0); | |
428870ff | 4445 | |
40b84e7a BB |
4446 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
4447 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff | 4448 | return; |
40b84e7a | 4449 | } |
428870ff BB |
4450 | |
4451 | tx = dmu_tx_create(os); | |
4452 | ||
4453 | cb_data[0] = ztest_create_cb_data(os, 0); | |
4454 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[0]); | |
4455 | ||
40b84e7a | 4456 | dmu_tx_hold_write(tx, od->od_object, 0, sizeof (uint64_t)); |
428870ff BB |
4457 | |
4458 | /* Every once in a while, abort the transaction on purpose */ | |
4459 | if (ztest_random(100) == 0) | |
4460 | error = -1; | |
4461 | ||
4462 | if (!error) | |
4463 | error = dmu_tx_assign(tx, TXG_NOWAIT); | |
4464 | ||
4465 | txg = error ? 0 : dmu_tx_get_txg(tx); | |
4466 | ||
4467 | cb_data[0]->zcd_txg = txg; | |
4468 | cb_data[1] = ztest_create_cb_data(os, txg); | |
4469 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[1]); | |
4470 | ||
4471 | if (error) { | |
4472 | /* | |
4473 | * It's not a strict requirement to call the registered | |
4474 | * callbacks from inside dmu_tx_abort(), but that's what | |
4475 | * it's supposed to happen in the current implementation | |
4476 | * so we will check for that. | |
4477 | */ | |
4478 | for (i = 0; i < 2; i++) { | |
4479 | cb_data[i]->zcd_expected_err = ECANCELED; | |
4480 | VERIFY(!cb_data[i]->zcd_called); | |
4481 | } | |
4482 | ||
4483 | dmu_tx_abort(tx); | |
4484 | ||
4485 | for (i = 0; i < 2; i++) { | |
4486 | VERIFY(cb_data[i]->zcd_called); | |
4487 | umem_free(cb_data[i], sizeof (ztest_cb_data_t)); | |
4488 | } | |
4489 | ||
40b84e7a | 4490 | umem_free(od, sizeof(ztest_od_t)); |
428870ff BB |
4491 | return; |
4492 | } | |
4493 | ||
4494 | cb_data[2] = ztest_create_cb_data(os, txg); | |
4495 | dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[2]); | |
4496 | ||
4497 | /* | |
4498 | * Read existing data to make sure there isn't a future leak. | |
4499 | */ | |
40b84e7a | 4500 | VERIFY(0 == dmu_read(os, od->od_object, 0, sizeof (uint64_t), |
428870ff BB |
4501 | &old_txg, DMU_READ_PREFETCH)); |
4502 | ||
4503 | if (old_txg > txg) | |
4504 | fatal(0, "future leak: got %" PRIu64 ", open txg is %" PRIu64, | |
4505 | old_txg, txg); | |
4506 | ||
40b84e7a | 4507 | dmu_write(os, od->od_object, 0, sizeof (uint64_t), &txg, tx); |
428870ff | 4508 | |
1e33ac1e | 4509 | (void) mutex_enter(&zcl.zcl_callbacks_lock); |
428870ff BB |
4510 | |
4511 | /* | |
4512 | * Since commit callbacks don't have any ordering requirement and since | |
4513 | * it is theoretically possible for a commit callback to be called | |
4514 | * after an arbitrary amount of time has elapsed since its txg has been | |
4515 | * synced, it is difficult to reliably determine whether a commit | |
4516 | * callback hasn't been called due to high load or due to a flawed | |
4517 | * implementation. | |
4518 | * | |
4519 | * In practice, we will assume that if after a certain number of txgs a | |
4520 | * commit callback hasn't been called, then most likely there's an | |
4521 | * implementation bug.. | |
4522 | */ | |
4523 | tmp_cb = list_head(&zcl.zcl_callbacks); | |
4524 | if (tmp_cb != NULL && | |
090ff092 | 4525 | tmp_cb->zcd_txg + ZTEST_COMMIT_CB_THRESH < txg) { |
428870ff BB |
4526 | fatal(0, "Commit callback threshold exceeded, oldest txg: %" |
4527 | PRIu64 ", open txg: %" PRIu64 "\n", tmp_cb->zcd_txg, txg); | |
4528 | } | |
4529 | ||
4530 | /* | |
4531 | * Let's find the place to insert our callbacks. | |
4532 | * | |
4533 | * Even though the list is ordered by txg, it is possible for the | |
4534 | * insertion point to not be the end because our txg may already be | |
4535 | * quiescing at this point and other callbacks in the open txg | |
4536 | * (from other objsets) may have sneaked in. | |
4537 | */ | |
4538 | tmp_cb = list_tail(&zcl.zcl_callbacks); | |
4539 | while (tmp_cb != NULL && tmp_cb->zcd_txg > txg) | |
4540 | tmp_cb = list_prev(&zcl.zcl_callbacks, tmp_cb); | |
4541 | ||
4542 | /* Add the 3 callbacks to the list */ | |
4543 | for (i = 0; i < 3; i++) { | |
4544 | if (tmp_cb == NULL) | |
4545 | list_insert_head(&zcl.zcl_callbacks, cb_data[i]); | |
4546 | else | |
4547 | list_insert_after(&zcl.zcl_callbacks, tmp_cb, | |
4548 | cb_data[i]); | |
4549 | ||
4550 | cb_data[i]->zcd_added = B_TRUE; | |
4551 | VERIFY(!cb_data[i]->zcd_called); | |
4552 | ||
4553 | tmp_cb = cb_data[i]; | |
4554 | } | |
4555 | ||
090ff092 RC |
4556 | zc_cb_counter += 3; |
4557 | ||
1e33ac1e | 4558 | (void) mutex_exit(&zcl.zcl_callbacks_lock); |
428870ff BB |
4559 | |
4560 | dmu_tx_commit(tx); | |
40b84e7a BB |
4561 | |
4562 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff BB |
4563 | } |
4564 | ||
4565 | /* ARGSUSED */ | |
4566 | void | |
4567 | ztest_dsl_prop_get_set(ztest_ds_t *zd, uint64_t id) | |
4568 | { | |
4569 | zfs_prop_t proplist[] = { | |
4570 | ZFS_PROP_CHECKSUM, | |
4571 | ZFS_PROP_COMPRESSION, | |
4572 | ZFS_PROP_COPIES, | |
4573 | ZFS_PROP_DEDUP | |
4574 | }; | |
d6320ddb | 4575 | int p; |
428870ff | 4576 | |
c242c188 | 4577 | (void) rw_enter(&ztest_name_lock, RW_READER); |
428870ff | 4578 | |
d6320ddb | 4579 | for (p = 0; p < sizeof (proplist) / sizeof (proplist[0]); p++) |
428870ff BB |
4580 | (void) ztest_dsl_prop_set_uint64(zd->zd_name, proplist[p], |
4581 | ztest_random_dsl_prop(proplist[p]), (int)ztest_random(2)); | |
4582 | ||
c242c188 | 4583 | (void) rw_exit(&ztest_name_lock); |
428870ff BB |
4584 | } |
4585 | ||
4586 | /* ARGSUSED */ | |
4587 | void | |
4588 | ztest_spa_prop_get_set(ztest_ds_t *zd, uint64_t id) | |
4589 | { | |
428870ff BB |
4590 | nvlist_t *props = NULL; |
4591 | ||
c242c188 | 4592 | (void) rw_enter(&ztest_name_lock, RW_READER); |
428870ff | 4593 | |
c242c188 | 4594 | (void) ztest_spa_prop_set_uint64(ZPOOL_PROP_DEDUPDITTO, |
428870ff BB |
4595 | ZIO_DEDUPDITTO_MIN + ztest_random(ZIO_DEDUPDITTO_MIN)); |
4596 | ||
c242c188 | 4597 | VERIFY3U(spa_prop_get(ztest_spa, &props), ==, 0); |
428870ff | 4598 | |
c242c188 | 4599 | if (ztest_opts.zo_verbose >= 6) |
428870ff BB |
4600 | dump_nvlist(props, 4); |
4601 | ||
4602 | nvlist_free(props); | |
4603 | ||
c242c188 | 4604 | (void) rw_exit(&ztest_name_lock); |
428870ff BB |
4605 | } |
4606 | ||
4607 | /* | |
4608 | * Test snapshot hold/release and deferred destroy. | |
4609 | */ | |
4610 | void | |
4611 | ztest_dmu_snapshot_hold(ztest_ds_t *zd, uint64_t id) | |
34dc7c2f | 4612 | { |
34dc7c2f | 4613 | int error; |
428870ff BB |
4614 | objset_t *os = zd->zd_os; |
4615 | objset_t *origin; | |
4616 | char snapname[100]; | |
4617 | char fullname[100]; | |
4618 | char clonename[100]; | |
4619 | char tag[100]; | |
4620 | char osname[MAXNAMELEN]; | |
34dc7c2f | 4621 | |
c242c188 | 4622 | (void) rw_enter(&ztest_name_lock, RW_READER); |
34dc7c2f BB |
4623 | |
4624 | dmu_objset_name(os, osname); | |
4625 | ||
1e33ac1e | 4626 | (void) snprintf(snapname, 100, "sh1_%llu", (u_longlong_t)id); |
428870ff | 4627 | (void) snprintf(fullname, 100, "%s@%s", osname, snapname); |
1e33ac1e BB |
4628 | (void) snprintf(clonename, 100, "%s/ch1_%llu",osname,(u_longlong_t)id); |
4629 | (void) snprintf(tag, 100, "tag_%llu", (u_longlong_t)id); | |
428870ff BB |
4630 | |
4631 | /* | |
4632 | * Clean up from any previous run. | |
4633 | */ | |
4634 | (void) dmu_objset_destroy(clonename, B_FALSE); | |
4635 | (void) dsl_dataset_user_release(osname, snapname, tag, B_FALSE); | |
4636 | (void) dmu_objset_destroy(fullname, B_FALSE); | |
4637 | ||
4638 | /* | |
4639 | * Create snapshot, clone it, mark snap for deferred destroy, | |
4640 | * destroy clone, verify snap was also destroyed. | |
4641 | */ | |
572e2857 BB |
4642 | error = dmu_objset_snapshot(osname, snapname, NULL, NULL, FALSE, |
4643 | FALSE, -1); | |
428870ff BB |
4644 | if (error) { |
4645 | if (error == ENOSPC) { | |
4646 | ztest_record_enospc("dmu_objset_snapshot"); | |
4647 | goto out; | |
34dc7c2f | 4648 | } |
428870ff BB |
4649 | fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error); |
4650 | } | |
34dc7c2f | 4651 | |
428870ff BB |
4652 | error = dmu_objset_hold(fullname, FTAG, &origin); |
4653 | if (error) | |
4654 | fatal(0, "dmu_objset_hold(%s) = %d", fullname, error); | |
34dc7c2f | 4655 | |
428870ff BB |
4656 | error = dmu_objset_clone(clonename, dmu_objset_ds(origin), 0); |
4657 | dmu_objset_rele(origin, FTAG); | |
4658 | if (error) { | |
34dc7c2f | 4659 | if (error == ENOSPC) { |
428870ff BB |
4660 | ztest_record_enospc("dmu_objset_clone"); |
4661 | goto out; | |
34dc7c2f | 4662 | } |
428870ff BB |
4663 | fatal(0, "dmu_objset_clone(%s) = %d", clonename, error); |
4664 | } | |
34dc7c2f | 4665 | |
428870ff BB |
4666 | error = dmu_objset_destroy(fullname, B_TRUE); |
4667 | if (error) { | |
4668 | fatal(0, "dmu_objset_destroy(%s, B_TRUE) = %d", | |
4669 | fullname, error); | |
4670 | } | |
34dc7c2f | 4671 | |
428870ff BB |
4672 | error = dmu_objset_destroy(clonename, B_FALSE); |
4673 | if (error) | |
4674 | fatal(0, "dmu_objset_destroy(%s) = %d", clonename, error); | |
34dc7c2f | 4675 | |
428870ff BB |
4676 | error = dmu_objset_hold(fullname, FTAG, &origin); |
4677 | if (error != ENOENT) | |
4678 | fatal(0, "dmu_objset_hold(%s) = %d", fullname, error); | |
34dc7c2f | 4679 | |
428870ff BB |
4680 | /* |
4681 | * Create snapshot, add temporary hold, verify that we can't | |
4682 | * destroy a held snapshot, mark for deferred destroy, | |
4683 | * release hold, verify snapshot was destroyed. | |
4684 | */ | |
572e2857 BB |
4685 | error = dmu_objset_snapshot(osname, snapname, NULL, NULL, FALSE, |
4686 | FALSE, -1); | |
428870ff BB |
4687 | if (error) { |
4688 | if (error == ENOSPC) { | |
4689 | ztest_record_enospc("dmu_objset_snapshot"); | |
4690 | goto out; | |
34dc7c2f | 4691 | } |
428870ff BB |
4692 | fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error); |
4693 | } | |
4694 | ||
572e2857 BB |
4695 | error = dsl_dataset_user_hold(osname, snapname, tag, B_FALSE, |
4696 | B_TRUE, -1); | |
428870ff BB |
4697 | if (error) |
4698 | fatal(0, "dsl_dataset_user_hold(%s)", fullname, tag); | |
4699 | ||
4700 | error = dmu_objset_destroy(fullname, B_FALSE); | |
4701 | if (error != EBUSY) { | |
4702 | fatal(0, "dmu_objset_destroy(%s, B_FALSE) = %d", | |
4703 | fullname, error); | |
4704 | } | |
4705 | ||
4706 | error = dmu_objset_destroy(fullname, B_TRUE); | |
4707 | if (error) { | |
4708 | fatal(0, "dmu_objset_destroy(%s, B_TRUE) = %d", | |
4709 | fullname, error); | |
34dc7c2f BB |
4710 | } |
4711 | ||
428870ff BB |
4712 | error = dsl_dataset_user_release(osname, snapname, tag, B_FALSE); |
4713 | if (error) | |
4714 | fatal(0, "dsl_dataset_user_release(%s)", fullname, tag); | |
4715 | ||
4716 | VERIFY(dmu_objset_hold(fullname, FTAG, &origin) == ENOENT); | |
4717 | ||
4718 | out: | |
c242c188 | 4719 | (void) rw_exit(&ztest_name_lock); |
34dc7c2f BB |
4720 | } |
4721 | ||
34dc7c2f BB |
4722 | /* |
4723 | * Inject random faults into the on-disk data. | |
4724 | */ | |
428870ff | 4725 | /* ARGSUSED */ |
34dc7c2f | 4726 | void |
428870ff | 4727 | ztest_fault_inject(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4728 | { |
428870ff | 4729 | ztest_shared_t *zs = ztest_shared; |
c242c188 | 4730 | spa_t *spa = ztest_spa; |
34dc7c2f BB |
4731 | int fd; |
4732 | uint64_t offset; | |
428870ff | 4733 | uint64_t leaves; |
c5b3a7bb | 4734 | uint64_t bad = 0x1990c0ffeedecadeull; |
34dc7c2f | 4735 | uint64_t top, leaf; |
40b84e7a BB |
4736 | char *path0; |
4737 | char *pathrand; | |
34dc7c2f | 4738 | size_t fsize; |
34dc7c2f BB |
4739 | int bshift = SPA_MAXBLOCKSHIFT + 2; /* don't scrog all labels */ |
4740 | int iters = 1000; | |
428870ff BB |
4741 | int maxfaults; |
4742 | int mirror_save; | |
b128c09f | 4743 | vdev_t *vd0 = NULL; |
34dc7c2f | 4744 | uint64_t guid0 = 0; |
428870ff BB |
4745 | boolean_t islog = B_FALSE; |
4746 | ||
40b84e7a BB |
4747 | path0 = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); |
4748 | pathrand = umem_alloc(MAXPATHLEN, UMEM_NOFAIL); | |
4749 | ||
c242c188 | 4750 | mutex_enter(&ztest_vdev_lock); |
428870ff | 4751 | maxfaults = MAXFAULTS(); |
c242c188 | 4752 | leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raidz; |
428870ff | 4753 | mirror_save = zs->zs_mirrors; |
c242c188 | 4754 | mutex_exit(&ztest_vdev_lock); |
34dc7c2f | 4755 | |
b128c09f | 4756 | ASSERT(leaves >= 1); |
34dc7c2f BB |
4757 | |
4758 | /* | |
b128c09f | 4759 | * We need SCL_STATE here because we're going to look at vd0->vdev_tsd. |
34dc7c2f | 4760 | */ |
b128c09f | 4761 | spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); |
34dc7c2f | 4762 | |
b128c09f BB |
4763 | if (ztest_random(2) == 0) { |
4764 | /* | |
428870ff | 4765 | * Inject errors on a normal data device or slog device. |
b128c09f | 4766 | */ |
428870ff BB |
4767 | top = ztest_random_vdev_top(spa, B_TRUE); |
4768 | leaf = ztest_random(leaves) + zs->zs_splits; | |
34dc7c2f | 4769 | |
b128c09f BB |
4770 | /* |
4771 | * Generate paths to the first leaf in this top-level vdev, | |
4772 | * and to the random leaf we selected. We'll induce transient | |
4773 | * write failures and random online/offline activity on leaf 0, | |
4774 | * and we'll write random garbage to the randomly chosen leaf. | |
4775 | */ | |
6aec1cd5 | 4776 | (void) snprintf(path0, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
4777 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
4778 | top * leaves + zs->zs_splits); | |
6aec1cd5 | 4779 | (void) snprintf(pathrand, MAXPATHLEN, ztest_dev_template, |
c242c188 CS |
4780 | ztest_opts.zo_dir, ztest_opts.zo_pool, |
4781 | top * leaves + leaf); | |
34dc7c2f | 4782 | |
b128c09f | 4783 | vd0 = vdev_lookup_by_path(spa->spa_root_vdev, path0); |
428870ff BB |
4784 | if (vd0 != NULL && vd0->vdev_top->vdev_islog) |
4785 | islog = B_TRUE; | |
4786 | ||
b128c09f BB |
4787 | if (vd0 != NULL && maxfaults != 1) { |
4788 | /* | |
4789 | * Make vd0 explicitly claim to be unreadable, | |
4790 | * or unwriteable, or reach behind its back | |
4791 | * and close the underlying fd. We can do this if | |
4792 | * maxfaults == 0 because we'll fail and reexecute, | |
4793 | * and we can do it if maxfaults >= 2 because we'll | |
4794 | * have enough redundancy. If maxfaults == 1, the | |
4795 | * combination of this with injection of random data | |
4796 | * corruption below exceeds the pool's fault tolerance. | |
4797 | */ | |
4798 | vdev_file_t *vf = vd0->vdev_tsd; | |
4799 | ||
4800 | if (vf != NULL && ztest_random(3) == 0) { | |
4801 | (void) close(vf->vf_vnode->v_fd); | |
4802 | vf->vf_vnode->v_fd = -1; | |
4803 | } else if (ztest_random(2) == 0) { | |
4804 | vd0->vdev_cant_read = B_TRUE; | |
4805 | } else { | |
4806 | vd0->vdev_cant_write = B_TRUE; | |
4807 | } | |
4808 | guid0 = vd0->vdev_guid; | |
4809 | } | |
4810 | } else { | |
4811 | /* | |
4812 | * Inject errors on an l2cache device. | |
4813 | */ | |
4814 | spa_aux_vdev_t *sav = &spa->spa_l2cache; | |
34dc7c2f | 4815 | |
b128c09f BB |
4816 | if (sav->sav_count == 0) { |
4817 | spa_config_exit(spa, SCL_STATE, FTAG); | |
40b84e7a | 4818 | goto out; |
b128c09f BB |
4819 | } |
4820 | vd0 = sav->sav_vdevs[ztest_random(sav->sav_count)]; | |
34dc7c2f | 4821 | guid0 = vd0->vdev_guid; |
b128c09f BB |
4822 | (void) strcpy(path0, vd0->vdev_path); |
4823 | (void) strcpy(pathrand, vd0->vdev_path); | |
4824 | ||
4825 | leaf = 0; | |
4826 | leaves = 1; | |
4827 | maxfaults = INT_MAX; /* no limit on cache devices */ | |
34dc7c2f BB |
4828 | } |
4829 | ||
b128c09f BB |
4830 | spa_config_exit(spa, SCL_STATE, FTAG); |
4831 | ||
34dc7c2f | 4832 | /* |
428870ff BB |
4833 | * If we can tolerate two or more faults, or we're dealing |
4834 | * with a slog, randomly online/offline vd0. | |
34dc7c2f | 4835 | */ |
428870ff | 4836 | if ((maxfaults >= 2 || islog) && guid0 != 0) { |
fb5f0bc8 BB |
4837 | if (ztest_random(10) < 6) { |
4838 | int flags = (ztest_random(2) == 0 ? | |
4839 | ZFS_OFFLINE_TEMPORARY : 0); | |
428870ff BB |
4840 | |
4841 | /* | |
4842 | * We have to grab the zs_name_lock as writer to | |
4843 | * prevent a race between offlining a slog and | |
4844 | * destroying a dataset. Offlining the slog will | |
4845 | * grab a reference on the dataset which may cause | |
4846 | * dmu_objset_destroy() to fail with EBUSY thus | |
4847 | * leaving the dataset in an inconsistent state. | |
4848 | */ | |
4849 | if (islog) | |
c242c188 | 4850 | (void) rw_enter(&ztest_name_lock, |
1e33ac1e | 4851 | RW_WRITER); |
428870ff | 4852 | |
fb5f0bc8 | 4853 | VERIFY(vdev_offline(spa, guid0, flags) != EBUSY); |
428870ff BB |
4854 | |
4855 | if (islog) | |
c242c188 | 4856 | (void) rw_exit(&ztest_name_lock); |
fb5f0bc8 BB |
4857 | } else { |
4858 | (void) vdev_online(spa, guid0, 0, NULL); | |
4859 | } | |
34dc7c2f BB |
4860 | } |
4861 | ||
428870ff | 4862 | if (maxfaults == 0) |
40b84e7a | 4863 | goto out; |
428870ff | 4864 | |
34dc7c2f BB |
4865 | /* |
4866 | * We have at least single-fault tolerance, so inject data corruption. | |
4867 | */ | |
4868 | fd = open(pathrand, O_RDWR); | |
4869 | ||
4870 | if (fd == -1) /* we hit a gap in the device namespace */ | |
40b84e7a | 4871 | goto out; |
34dc7c2f BB |
4872 | |
4873 | fsize = lseek(fd, 0, SEEK_END); | |
4874 | ||
4875 | while (--iters != 0) { | |
4876 | offset = ztest_random(fsize / (leaves << bshift)) * | |
4877 | (leaves << bshift) + (leaf << bshift) + | |
4878 | (ztest_random(1ULL << (bshift - 1)) & -8ULL); | |
4879 | ||
4880 | if (offset >= fsize) | |
4881 | continue; | |
4882 | ||
c242c188 | 4883 | mutex_enter(&ztest_vdev_lock); |
428870ff | 4884 | if (mirror_save != zs->zs_mirrors) { |
c242c188 | 4885 | mutex_exit(&ztest_vdev_lock); |
428870ff | 4886 | (void) close(fd); |
40b84e7a | 4887 | goto out; |
428870ff | 4888 | } |
34dc7c2f BB |
4889 | |
4890 | if (pwrite(fd, &bad, sizeof (bad), offset) != sizeof (bad)) | |
4891 | fatal(1, "can't inject bad word at 0x%llx in %s", | |
4892 | offset, pathrand); | |
428870ff | 4893 | |
c242c188 | 4894 | mutex_exit(&ztest_vdev_lock); |
428870ff | 4895 | |
c242c188 | 4896 | if (ztest_opts.zo_verbose >= 7) |
428870ff BB |
4897 | (void) printf("injected bad word into %s," |
4898 | " offset 0x%llx\n", pathrand, (u_longlong_t)offset); | |
34dc7c2f BB |
4899 | } |
4900 | ||
4901 | (void) close(fd); | |
40b84e7a BB |
4902 | out: |
4903 | umem_free(path0, MAXPATHLEN); | |
4904 | umem_free(pathrand, MAXPATHLEN); | |
34dc7c2f BB |
4905 | } |
4906 | ||
4907 | /* | |
428870ff | 4908 | * Verify that DDT repair works as expected. |
34dc7c2f BB |
4909 | */ |
4910 | void | |
428870ff | 4911 | ztest_ddt_repair(ztest_ds_t *zd, uint64_t id) |
34dc7c2f | 4912 | { |
428870ff | 4913 | ztest_shared_t *zs = ztest_shared; |
c242c188 | 4914 | spa_t *spa = ztest_spa; |
428870ff | 4915 | objset_t *os = zd->zd_os; |
40b84e7a | 4916 | ztest_od_t *od; |
428870ff BB |
4917 | uint64_t object, blocksize, txg, pattern, psize; |
4918 | enum zio_checksum checksum = spa_dedup_checksum(spa); | |
4919 | dmu_buf_t *db; | |
4920 | dmu_tx_t *tx; | |
4921 | void *buf; | |
4922 | blkptr_t blk; | |
4923 | int copies = 2 * ZIO_DEDUPDITTO_MIN; | |
d6320ddb | 4924 | int i; |
34dc7c2f | 4925 | |
428870ff BB |
4926 | blocksize = ztest_random_blocksize(); |
4927 | blocksize = MIN(blocksize, 2048); /* because we write so many */ | |
34dc7c2f | 4928 | |
40b84e7a BB |
4929 | od = umem_alloc(sizeof(ztest_od_t), UMEM_NOFAIL); |
4930 | ztest_od_init(od, id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0); | |
34dc7c2f | 4931 | |
40b84e7a BB |
4932 | if (ztest_object_init(zd, od, sizeof (ztest_od_t), B_FALSE) != 0) { |
4933 | umem_free(od, sizeof(ztest_od_t)); | |
428870ff | 4934 | return; |
40b84e7a | 4935 | } |
34dc7c2f BB |
4936 | |
4937 | /* | |
428870ff BB |
4938 | * Take the name lock as writer to prevent anyone else from changing |
4939 | * the pool and dataset properies we need to maintain during this test. | |
34dc7c2f | 4940 | */ |
c242c188 | 4941 | (void) rw_enter(&ztest_name_lock, RW_WRITER); |
34dc7c2f | 4942 | |
428870ff BB |
4943 | if (ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_DEDUP, checksum, |
4944 | B_FALSE) != 0 || | |
4945 | ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_COPIES, 1, | |
4946 | B_FALSE) != 0) { | |
c242c188 | 4947 | (void) rw_exit(&ztest_name_lock); |
40b84e7a | 4948 | umem_free(od, sizeof(ztest_od_t)); |
428870ff BB |
4949 | return; |
4950 | } | |
4951 | ||
4952 | object = od[0].od_object; | |
4953 | blocksize = od[0].od_blocksize; | |
3541dc6d | 4954 | pattern = zs->zs_guid ^ dmu_objset_fsid_guid(os); |
428870ff BB |
4955 | |
4956 | ASSERT(object != 0); | |
4957 | ||
4958 | tx = dmu_tx_create(os); | |
4959 | dmu_tx_hold_write(tx, object, 0, copies * blocksize); | |
4960 | txg = ztest_tx_assign(tx, TXG_WAIT, FTAG); | |
4961 | if (txg == 0) { | |
c242c188 | 4962 | (void) rw_exit(&ztest_name_lock); |
40b84e7a | 4963 | umem_free(od, sizeof(ztest_od_t)); |
428870ff BB |
4964 | return; |
4965 | } | |
34dc7c2f BB |
4966 | |
4967 | /* | |
428870ff | 4968 | * Write all the copies of our block. |
34dc7c2f | 4969 | */ |
d6320ddb | 4970 | for (i = 0; i < copies; i++) { |
428870ff BB |
4971 | uint64_t offset = i * blocksize; |
4972 | VERIFY(dmu_buf_hold(os, object, offset, FTAG, &db, | |
4973 | DMU_READ_NO_PREFETCH) == 0); | |
4974 | ASSERT(db->db_offset == offset); | |
4975 | ASSERT(db->db_size == blocksize); | |
4976 | ASSERT(ztest_pattern_match(db->db_data, db->db_size, pattern) || | |
4977 | ztest_pattern_match(db->db_data, db->db_size, 0ULL)); | |
4978 | dmu_buf_will_fill(db, tx); | |
4979 | ztest_pattern_set(db->db_data, db->db_size, pattern); | |
4980 | dmu_buf_rele(db, FTAG); | |
4981 | } | |
34dc7c2f | 4982 | |
428870ff BB |
4983 | dmu_tx_commit(tx); |
4984 | txg_wait_synced(spa_get_dsl(spa), txg); | |
34dc7c2f BB |
4985 | |
4986 | /* | |
428870ff | 4987 | * Find out what block we got. |
34dc7c2f | 4988 | */ |
428870ff BB |
4989 | VERIFY(dmu_buf_hold(os, object, 0, FTAG, &db, |
4990 | DMU_READ_NO_PREFETCH) == 0); | |
4991 | blk = *((dmu_buf_impl_t *)db)->db_blkptr; | |
4992 | dmu_buf_rele(db, FTAG); | |
34dc7c2f BB |
4993 | |
4994 | /* | |
428870ff | 4995 | * Damage the block. Dedup-ditto will save us when we read it later. |
34dc7c2f | 4996 | */ |
428870ff BB |
4997 | psize = BP_GET_PSIZE(&blk); |
4998 | buf = zio_buf_alloc(psize); | |
4999 | ztest_pattern_set(buf, psize, ~pattern); | |
34dc7c2f | 5000 | |
428870ff BB |
5001 | (void) zio_wait(zio_rewrite(NULL, spa, 0, &blk, |
5002 | buf, psize, NULL, NULL, ZIO_PRIORITY_SYNC_WRITE, | |
5003 | ZIO_FLAG_CANFAIL | ZIO_FLAG_INDUCE_DAMAGE, NULL)); | |
34dc7c2f | 5004 | |
428870ff | 5005 | zio_buf_free(buf, psize); |
34dc7c2f | 5006 | |
c242c188 | 5007 | (void) rw_exit(&ztest_name_lock); |
40b84e7a | 5008 | umem_free(od, sizeof(ztest_od_t)); |
34dc7c2f BB |
5009 | } |
5010 | ||
34dc7c2f | 5011 | /* |
428870ff | 5012 | * Scrub the pool. |
34dc7c2f | 5013 | */ |
428870ff BB |
5014 | /* ARGSUSED */ |
5015 | void | |
5016 | ztest_scrub(ztest_ds_t *zd, uint64_t id) | |
34dc7c2f | 5017 | { |
c242c188 | 5018 | spa_t *spa = ztest_spa; |
34dc7c2f | 5019 | |
428870ff BB |
5020 | (void) spa_scan(spa, POOL_SCAN_SCRUB); |
5021 | (void) poll(NULL, 0, 100); /* wait a moment, then force a restart */ | |
5022 | (void) spa_scan(spa, POOL_SCAN_SCRUB); | |
5023 | } | |
34dc7c2f | 5024 | |
3541dc6d GA |
5025 | /* |
5026 | * Change the guid for the pool. | |
5027 | */ | |
5028 | /* ARGSUSED */ | |
5029 | void | |
5030 | ztest_reguid(ztest_ds_t *zd, uint64_t id) | |
5031 | { | |
c242c188 | 5032 | spa_t *spa = ztest_spa; |
3541dc6d GA |
5033 | uint64_t orig, load; |
5034 | ||
5035 | orig = spa_guid(spa); | |
5036 | load = spa_load_guid(spa); | |
5037 | if (spa_change_guid(spa) != 0) | |
5038 | return; | |
5039 | ||
c242c188 | 5040 | if (ztest_opts.zo_verbose >= 3) { |
3541dc6d GA |
5041 | (void) printf("Changed guid old %llu -> %llu\n", |
5042 | (u_longlong_t)orig, (u_longlong_t)spa_guid(spa)); | |
5043 | } | |
5044 | ||
5045 | VERIFY3U(orig, !=, spa_guid(spa)); | |
5046 | VERIFY3U(load, ==, spa_load_guid(spa)); | |
5047 | } | |
5048 | ||
428870ff BB |
5049 | /* |
5050 | * Rename the pool to a different name and then rename it back. | |
5051 | */ | |
5052 | /* ARGSUSED */ | |
5053 | void | |
5054 | ztest_spa_rename(ztest_ds_t *zd, uint64_t id) | |
5055 | { | |
428870ff BB |
5056 | char *oldname, *newname; |
5057 | spa_t *spa; | |
34dc7c2f | 5058 | |
c242c188 | 5059 | (void) rw_enter(&ztest_name_lock, RW_WRITER); |
34dc7c2f | 5060 | |
c242c188 | 5061 | oldname = ztest_opts.zo_pool; |
428870ff BB |
5062 | newname = umem_alloc(strlen(oldname) + 5, UMEM_NOFAIL); |
5063 | (void) strcpy(newname, oldname); | |
5064 | (void) strcat(newname, "_tmp"); | |
34dc7c2f BB |
5065 | |
5066 | /* | |
428870ff | 5067 | * Do the rename |
34dc7c2f | 5068 | */ |
428870ff | 5069 | VERIFY3U(0, ==, spa_rename(oldname, newname)); |
34dc7c2f BB |
5070 | |
5071 | /* | |
428870ff | 5072 | * Try to open it under the old name, which shouldn't exist |
34dc7c2f | 5073 | */ |
428870ff | 5074 | VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f BB |
5075 | |
5076 | /* | |
428870ff BB |
5077 | * Open it under the new name and make sure it's still the same spa_t. |
5078 | */ | |
5079 | VERIFY3U(0, ==, spa_open(newname, &spa, FTAG)); | |
34dc7c2f | 5080 | |
c242c188 | 5081 | ASSERT(spa == ztest_spa); |
428870ff | 5082 | spa_close(spa, FTAG); |
34dc7c2f BB |
5083 | |
5084 | /* | |
428870ff | 5085 | * Rename it back to the original |
34dc7c2f | 5086 | */ |
428870ff | 5087 | VERIFY3U(0, ==, spa_rename(newname, oldname)); |
34dc7c2f | 5088 | |
428870ff BB |
5089 | /* |
5090 | * Make sure it can still be opened | |
5091 | */ | |
5092 | VERIFY3U(0, ==, spa_open(oldname, &spa, FTAG)); | |
34dc7c2f | 5093 | |
c242c188 | 5094 | ASSERT(spa == ztest_spa); |
428870ff BB |
5095 | spa_close(spa, FTAG); |
5096 | ||
5097 | umem_free(newname, strlen(newname) + 1); | |
5098 | ||
c242c188 | 5099 | (void) rw_exit(&ztest_name_lock); |
34dc7c2f BB |
5100 | } |
5101 | ||
428870ff BB |
5102 | /* |
5103 | * Verify pool integrity by running zdb. | |
5104 | */ | |
34dc7c2f | 5105 | static void |
428870ff | 5106 | ztest_run_zdb(char *pool) |
34dc7c2f BB |
5107 | { |
5108 | int status; | |
34dc7c2f | 5109 | char *bin; |
0e8d1b2d BB |
5110 | char *zdb; |
5111 | char *zbuf; | |
34dc7c2f BB |
5112 | FILE *fp; |
5113 | ||
0e8d1b2d BB |
5114 | bin = umem_alloc(MAXPATHLEN + MAXNAMELEN + 20, UMEM_NOFAIL); |
5115 | zdb = umem_alloc(MAXPATHLEN + MAXNAMELEN + 20, UMEM_NOFAIL); | |
5116 | zbuf = umem_alloc(1024, UMEM_NOFAIL); | |
34dc7c2f | 5117 | |
0e8d1b2d | 5118 | VERIFY(realpath(getexecname(), bin) != NULL); |
341b5f1d | 5119 | if (strncmp(bin, "/usr/sbin/ztest", 15) == 0) { |
0e8d1b2d | 5120 | strcpy(bin, "/usr/sbin/zdb"); /* Installed */ |
341b5f1d BB |
5121 | } else if (strncmp(bin, "/sbin/ztest", 11) == 0) { |
5122 | strcpy(bin, "/sbin/zdb"); /* Installed */ | |
0e8d1b2d BB |
5123 | } else { |
5124 | strstr(bin, "/ztest/")[0] = '\0'; /* In-tree */ | |
5125 | strcat(bin, "/zdb/zdb"); | |
5126 | } | |
5127 | ||
5128 | (void) sprintf(zdb, | |
5129 | "%s -bcc%s%s -U %s %s", | |
5130 | bin, | |
c242c188 CS |
5131 | ztest_opts.zo_verbose >= 3 ? "s" : "", |
5132 | ztest_opts.zo_verbose >= 4 ? "v" : "", | |
428870ff | 5133 | spa_config_path, |
b128c09f | 5134 | pool); |
34dc7c2f | 5135 | |
c242c188 | 5136 | if (ztest_opts.zo_verbose >= 5) |
34dc7c2f BB |
5137 | (void) printf("Executing %s\n", strstr(zdb, "zdb ")); |
5138 | ||
5139 | fp = popen(zdb, "r"); | |
5140 | ||
6aec1cd5 | 5141 | while (fgets(zbuf, 1024, fp) != NULL) |
c242c188 | 5142 | if (ztest_opts.zo_verbose >= 3) |
34dc7c2f BB |
5143 | (void) printf("%s", zbuf); |
5144 | ||
5145 | status = pclose(fp); | |
5146 | ||
5147 | if (status == 0) | |
0e8d1b2d | 5148 | goto out; |
34dc7c2f BB |
5149 | |
5150 | ztest_dump_core = 0; | |
5151 | if (WIFEXITED(status)) | |
5152 | fatal(0, "'%s' exit code %d", zdb, WEXITSTATUS(status)); | |
5153 | else | |
5154 | fatal(0, "'%s' died with signal %d", zdb, WTERMSIG(status)); | |
0e8d1b2d BB |
5155 | out: |
5156 | umem_free(bin, MAXPATHLEN + MAXNAMELEN + 20); | |
5157 | umem_free(zdb, MAXPATHLEN + MAXNAMELEN + 20); | |
5158 | umem_free(zbuf, 1024); | |
34dc7c2f BB |
5159 | } |
5160 | ||
5161 | static void | |
5162 | ztest_walk_pool_directory(char *header) | |
5163 | { | |
5164 | spa_t *spa = NULL; | |
5165 | ||
c242c188 | 5166 | if (ztest_opts.zo_verbose >= 6) |
34dc7c2f BB |
5167 | (void) printf("%s\n", header); |
5168 | ||
5169 | mutex_enter(&spa_namespace_lock); | |
5170 | while ((spa = spa_next(spa)) != NULL) | |
c242c188 | 5171 | if (ztest_opts.zo_verbose >= 6) |
34dc7c2f BB |
5172 | (void) printf("\t%s\n", spa_name(spa)); |
5173 | mutex_exit(&spa_namespace_lock); | |
5174 | } | |
5175 | ||
5176 | static void | |
5177 | ztest_spa_import_export(char *oldname, char *newname) | |
5178 | { | |
fb5f0bc8 | 5179 | nvlist_t *config, *newconfig; |
34dc7c2f BB |
5180 | uint64_t pool_guid; |
5181 | spa_t *spa; | |
34dc7c2f | 5182 | |
c242c188 | 5183 | if (ztest_opts.zo_verbose >= 4) { |
34dc7c2f BB |
5184 | (void) printf("import/export: old = %s, new = %s\n", |
5185 | oldname, newname); | |
5186 | } | |
5187 | ||
5188 | /* | |
5189 | * Clean up from previous runs. | |
5190 | */ | |
5191 | (void) spa_destroy(newname); | |
5192 | ||
5193 | /* | |
5194 | * Get the pool's configuration and guid. | |
5195 | */ | |
428870ff | 5196 | VERIFY3U(0, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f | 5197 | |
fb5f0bc8 BB |
5198 | /* |
5199 | * Kick off a scrub to tickle scrub/export races. | |
5200 | */ | |
5201 | if (ztest_random(2) == 0) | |
428870ff | 5202 | (void) spa_scan(spa, POOL_SCAN_SCRUB); |
fb5f0bc8 | 5203 | |
34dc7c2f BB |
5204 | pool_guid = spa_guid(spa); |
5205 | spa_close(spa, FTAG); | |
5206 | ||
5207 | ztest_walk_pool_directory("pools before export"); | |
5208 | ||
5209 | /* | |
5210 | * Export it. | |
5211 | */ | |
428870ff | 5212 | VERIFY3U(0, ==, spa_export(oldname, &config, B_FALSE, B_FALSE)); |
34dc7c2f BB |
5213 | |
5214 | ztest_walk_pool_directory("pools after export"); | |
5215 | ||
fb5f0bc8 BB |
5216 | /* |
5217 | * Try to import it. | |
5218 | */ | |
5219 | newconfig = spa_tryimport(config); | |
5220 | ASSERT(newconfig != NULL); | |
5221 | nvlist_free(newconfig); | |
5222 | ||
34dc7c2f BB |
5223 | /* |
5224 | * Import it under the new name. | |
5225 | */ | |
572e2857 | 5226 | VERIFY3U(0, ==, spa_import(newname, config, NULL, 0)); |
34dc7c2f BB |
5227 | |
5228 | ztest_walk_pool_directory("pools after import"); | |
5229 | ||
5230 | /* | |
5231 | * Try to import it again -- should fail with EEXIST. | |
5232 | */ | |
572e2857 | 5233 | VERIFY3U(EEXIST, ==, spa_import(newname, config, NULL, 0)); |
34dc7c2f BB |
5234 | |
5235 | /* | |
5236 | * Try to import it under a different name -- should fail with EEXIST. | |
5237 | */ | |
572e2857 | 5238 | VERIFY3U(EEXIST, ==, spa_import(oldname, config, NULL, 0)); |
34dc7c2f BB |
5239 | |
5240 | /* | |
5241 | * Verify that the pool is no longer visible under the old name. | |
5242 | */ | |
428870ff | 5243 | VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG)); |
34dc7c2f BB |
5244 | |
5245 | /* | |
5246 | * Verify that we can open and close the pool using the new name. | |
5247 | */ | |
428870ff | 5248 | VERIFY3U(0, ==, spa_open(newname, &spa, FTAG)); |
34dc7c2f BB |
5249 | ASSERT(pool_guid == spa_guid(spa)); |
5250 | spa_close(spa, FTAG); | |
5251 | ||
5252 | nvlist_free(config); | |
5253 | } | |
5254 | ||
fb5f0bc8 BB |
5255 | static void |
5256 | ztest_resume(spa_t *spa) | |
5257 | { | |
c242c188 | 5258 | if (spa_suspended(spa) && ztest_opts.zo_verbose >= 6) |
428870ff BB |
5259 | (void) printf("resuming from suspended state\n"); |
5260 | spa_vdev_state_enter(spa, SCL_NONE); | |
5261 | vdev_clear(spa, NULL); | |
5262 | (void) spa_vdev_state_exit(spa, NULL, 0); | |
5263 | (void) zio_resume(spa); | |
fb5f0bc8 BB |
5264 | } |
5265 | ||
34dc7c2f | 5266 | static void * |
fb5f0bc8 | 5267 | ztest_resume_thread(void *arg) |
34dc7c2f | 5268 | { |
b128c09f | 5269 | spa_t *spa = arg; |
34dc7c2f BB |
5270 | |
5271 | while (!ztest_exiting) { | |
428870ff BB |
5272 | if (spa_suspended(spa)) |
5273 | ztest_resume(spa); | |
5274 | (void) poll(NULL, 0, 100); | |
34dc7c2f | 5275 | } |
34dc7c2f | 5276 | |
1e33ac1e | 5277 | thread_exit(); |
428870ff | 5278 | |
1e33ac1e BB |
5279 | return (NULL); |
5280 | } | |
428870ff | 5281 | |
1e33ac1e | 5282 | #define GRACE 300 |
428870ff | 5283 | |
1e33ac1e BB |
5284 | static void |
5285 | ztest_deadman_alarm(int sig) | |
5286 | { | |
5287 | fatal(0, "failed to complete within %d seconds of deadline", GRACE); | |
428870ff BB |
5288 | } |
5289 | ||
5290 | static void | |
c242c188 | 5291 | ztest_execute(int test, ztest_info_t *zi, uint64_t id) |
428870ff | 5292 | { |
c242c188 CS |
5293 | ztest_ds_t *zd = &ztest_ds[id % ztest_opts.zo_datasets]; |
5294 | ztest_shared_callstate_t *zc = ZTEST_GET_SHARED_CALLSTATE(test); | |
428870ff | 5295 | hrtime_t functime = gethrtime(); |
d6320ddb | 5296 | int i; |
428870ff | 5297 | |
d6320ddb | 5298 | for (i = 0; i < zi->zi_iters; i++) |
428870ff BB |
5299 | zi->zi_func(zd, id); |
5300 | ||
5301 | functime = gethrtime() - functime; | |
5302 | ||
c242c188 CS |
5303 | atomic_add_64(&zc->zc_count, 1); |
5304 | atomic_add_64(&zc->zc_time, functime); | |
428870ff | 5305 | |
c242c188 | 5306 | if (ztest_opts.zo_verbose >= 4) { |
428870ff BB |
5307 | Dl_info dli; |
5308 | (void) dladdr((void *)zi->zi_func, &dli); | |
5309 | (void) printf("%6.2f sec in %s\n", | |
5310 | (double)functime / NANOSEC, dli.dli_sname); | |
5311 | } | |
5312 | } | |
5313 | ||
34dc7c2f BB |
5314 | static void * |
5315 | ztest_thread(void *arg) | |
5316 | { | |
c242c188 | 5317 | int rand; |
428870ff | 5318 | uint64_t id = (uintptr_t)arg; |
34dc7c2f | 5319 | ztest_shared_t *zs = ztest_shared; |
428870ff BB |
5320 | uint64_t call_next; |
5321 | hrtime_t now; | |
34dc7c2f | 5322 | ztest_info_t *zi; |
c242c188 | 5323 | ztest_shared_callstate_t *zc; |
34dc7c2f | 5324 | |
428870ff | 5325 | while ((now = gethrtime()) < zs->zs_thread_stop) { |
34dc7c2f BB |
5326 | /* |
5327 | * See if it's time to force a crash. | |
5328 | */ | |
428870ff BB |
5329 | if (now > zs->zs_thread_kill) |
5330 | ztest_kill(zs); | |
34dc7c2f BB |
5331 | |
5332 | /* | |
428870ff | 5333 | * If we're getting ENOSPC with some regularity, stop. |
34dc7c2f | 5334 | */ |
428870ff BB |
5335 | if (zs->zs_enospc_count > 10) |
5336 | break; | |
34dc7c2f BB |
5337 | |
5338 | /* | |
428870ff | 5339 | * Pick a random function to execute. |
34dc7c2f | 5340 | */ |
c242c188 CS |
5341 | rand = ztest_random(ZTEST_FUNCS); |
5342 | zi = &ztest_info[rand]; | |
5343 | zc = ZTEST_GET_SHARED_CALLSTATE(rand); | |
5344 | call_next = zc->zc_next; | |
34dc7c2f | 5345 | |
428870ff | 5346 | if (now >= call_next && |
c242c188 CS |
5347 | atomic_cas_64(&zc->zc_next, call_next, call_next + |
5348 | ztest_random(2 * zi->zi_interval[0] + 1)) == call_next) { | |
5349 | ztest_execute(rand, zi, id); | |
5350 | } | |
428870ff | 5351 | } |
34dc7c2f | 5352 | |
1e33ac1e BB |
5353 | thread_exit(); |
5354 | ||
428870ff BB |
5355 | return (NULL); |
5356 | } | |
34dc7c2f | 5357 | |
428870ff BB |
5358 | static void |
5359 | ztest_dataset_name(char *dsname, char *pool, int d) | |
5360 | { | |
5361 | (void) snprintf(dsname, MAXNAMELEN, "%s/ds_%d", pool, d); | |
5362 | } | |
34dc7c2f | 5363 | |
428870ff | 5364 | static void |
c242c188 | 5365 | ztest_dataset_destroy(int d) |
428870ff BB |
5366 | { |
5367 | char name[MAXNAMELEN]; | |
d6320ddb | 5368 | int t; |
34dc7c2f | 5369 | |
c242c188 | 5370 | ztest_dataset_name(name, ztest_opts.zo_pool, d); |
34dc7c2f | 5371 | |
c242c188 | 5372 | if (ztest_opts.zo_verbose >= 3) |
428870ff | 5373 | (void) printf("Destroying %s to free up space\n", name); |
34dc7c2f | 5374 | |
428870ff BB |
5375 | /* |
5376 | * Cleanup any non-standard clones and snapshots. In general, | |
5377 | * ztest thread t operates on dataset (t % zopt_datasets), | |
5378 | * so there may be more than one thing to clean up. | |
5379 | */ | |
c242c188 CS |
5380 | for (t = d; t < ztest_opts.zo_threads; |
5381 | t += ztest_opts.zo_datasets) | |
428870ff BB |
5382 | ztest_dsl_dataset_cleanup(name, t); |
5383 | ||
5384 | (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL, | |
5385 | DS_FIND_SNAPSHOTS | DS_FIND_CHILDREN); | |
5386 | } | |
5387 | ||
5388 | static void | |
5389 | ztest_dataset_dirobj_verify(ztest_ds_t *zd) | |
5390 | { | |
5391 | uint64_t usedobjs, dirobjs, scratch; | |
5392 | ||
5393 | /* | |
5394 | * ZTEST_DIROBJ is the object directory for the entire dataset. | |
5395 | * Therefore, the number of objects in use should equal the | |
5396 | * number of ZTEST_DIROBJ entries, +1 for ZTEST_DIROBJ itself. | |
5397 | * If not, we have an object leak. | |
5398 | * | |
5399 | * Note that we can only check this in ztest_dataset_open(), | |
5400 | * when the open-context and syncing-context values agree. | |
5401 | * That's because zap_count() returns the open-context value, | |
5402 | * while dmu_objset_space() returns the rootbp fill count. | |
5403 | */ | |
5404 | VERIFY3U(0, ==, zap_count(zd->zd_os, ZTEST_DIROBJ, &dirobjs)); | |
5405 | dmu_objset_space(zd->zd_os, &scratch, &scratch, &usedobjs, &scratch); | |
5406 | ASSERT3U(dirobjs + 1, ==, usedobjs); | |
5407 | } | |
5408 | ||
5409 | static int | |
c242c188 | 5410 | ztest_dataset_open(int d) |
428870ff | 5411 | { |
c242c188 CS |
5412 | ztest_ds_t *zd = &ztest_ds[d]; |
5413 | uint64_t committed_seq = ZTEST_GET_SHARED_DS(d)->zd_seq; | |
428870ff BB |
5414 | objset_t *os; |
5415 | zilog_t *zilog; | |
5416 | char name[MAXNAMELEN]; | |
5417 | int error; | |
5418 | ||
c242c188 | 5419 | ztest_dataset_name(name, ztest_opts.zo_pool, d); |
428870ff | 5420 | |
c242c188 | 5421 | (void) rw_enter(&ztest_name_lock, RW_READER); |
428870ff BB |
5422 | |
5423 | error = ztest_dataset_create(name); | |
5424 | if (error == ENOSPC) { | |
c242c188 | 5425 | (void) rw_exit(&ztest_name_lock); |
428870ff BB |
5426 | ztest_record_enospc(FTAG); |
5427 | return (error); | |
34dc7c2f | 5428 | } |
428870ff | 5429 | ASSERT(error == 0 || error == EEXIST); |
34dc7c2f | 5430 | |
428870ff | 5431 | VERIFY3U(dmu_objset_hold(name, zd, &os), ==, 0); |
c242c188 | 5432 | (void) rw_exit(&ztest_name_lock); |
428870ff | 5433 | |
c242c188 | 5434 | ztest_zd_init(zd, ZTEST_GET_SHARED_DS(d), os); |
428870ff BB |
5435 | |
5436 | zilog = zd->zd_zilog; | |
5437 | ||
5438 | if (zilog->zl_header->zh_claim_lr_seq != 0 && | |
5439 | zilog->zl_header->zh_claim_lr_seq < committed_seq) | |
5440 | fatal(0, "missing log records: claimed %llu < committed %llu", | |
5441 | zilog->zl_header->zh_claim_lr_seq, committed_seq); | |
5442 | ||
5443 | ztest_dataset_dirobj_verify(zd); | |
5444 | ||
5445 | zil_replay(os, zd, ztest_replay_vector); | |
5446 | ||
5447 | ztest_dataset_dirobj_verify(zd); | |
5448 | ||
c242c188 | 5449 | if (ztest_opts.zo_verbose >= 6) |
428870ff BB |
5450 | (void) printf("%s replay %llu blocks, %llu records, seq %llu\n", |
5451 | zd->zd_name, | |
5452 | (u_longlong_t)zilog->zl_parse_blk_count, | |
5453 | (u_longlong_t)zilog->zl_parse_lr_count, | |
5454 | (u_longlong_t)zilog->zl_replaying_seq); | |
5455 | ||
5456 | zilog = zil_open(os, ztest_get_data); | |
5457 | ||
5458 | if (zilog->zl_replaying_seq != 0 && | |
5459 | zilog->zl_replaying_seq < committed_seq) | |
5460 | fatal(0, "missing log records: replayed %llu < committed %llu", | |
5461 | zilog->zl_replaying_seq, committed_seq); | |
5462 | ||
5463 | return (0); | |
5464 | } | |
5465 | ||
5466 | static void | |
c242c188 | 5467 | ztest_dataset_close(int d) |
428870ff | 5468 | { |
c242c188 | 5469 | ztest_ds_t *zd = &ztest_ds[d]; |
428870ff BB |
5470 | |
5471 | zil_close(zd->zd_zilog); | |
5472 | dmu_objset_rele(zd->zd_os, zd); | |
5473 | ||
5474 | ztest_zd_fini(zd); | |
34dc7c2f BB |
5475 | } |
5476 | ||
5477 | /* | |
5478 | * Kick off threads to run tests on all datasets in parallel. | |
5479 | */ | |
5480 | static void | |
428870ff | 5481 | ztest_run(ztest_shared_t *zs) |
34dc7c2f | 5482 | { |
1e33ac1e | 5483 | kt_did_t *tid; |
34dc7c2f | 5484 | spa_t *spa; |
3541dc6d | 5485 | objset_t *os; |
1e33ac1e BB |
5486 | kthread_t *resume_thread; |
5487 | uint64_t object; | |
428870ff | 5488 | int error; |
d6320ddb | 5489 | int t, d; |
b128c09f BB |
5490 | |
5491 | ztest_exiting = B_FALSE; | |
34dc7c2f | 5492 | |
34dc7c2f | 5493 | /* |
428870ff | 5494 | * Initialize parent/child shared state. |
34dc7c2f | 5495 | */ |
c242c188 CS |
5496 | mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); |
5497 | rw_init(&ztest_name_lock, NULL, RW_DEFAULT, NULL); | |
34dc7c2f | 5498 | |
428870ff | 5499 | zs->zs_thread_start = gethrtime(); |
c242c188 CS |
5500 | zs->zs_thread_stop = |
5501 | zs->zs_thread_start + ztest_opts.zo_passtime * NANOSEC; | |
428870ff BB |
5502 | zs->zs_thread_stop = MIN(zs->zs_thread_stop, zs->zs_proc_stop); |
5503 | zs->zs_thread_kill = zs->zs_thread_stop; | |
c242c188 CS |
5504 | if (ztest_random(100) < ztest_opts.zo_killrate) { |
5505 | zs->zs_thread_kill -= | |
5506 | ztest_random(ztest_opts.zo_passtime * NANOSEC); | |
5507 | } | |
34dc7c2f | 5508 | |
1e33ac1e | 5509 | mutex_init(&zcl.zcl_callbacks_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 5510 | |
428870ff BB |
5511 | list_create(&zcl.zcl_callbacks, sizeof (ztest_cb_data_t), |
5512 | offsetof(ztest_cb_data_t, zcd_node)); | |
34dc7c2f BB |
5513 | |
5514 | /* | |
b128c09f | 5515 | * Open our pool. |
34dc7c2f | 5516 | */ |
428870ff | 5517 | kernel_init(FREAD | FWRITE); |
c242c188 | 5518 | VERIFY(spa_open(ztest_opts.zo_pool, &spa, FTAG) == 0); |
6d974228 | 5519 | spa->spa_debug = B_TRUE; |
c242c188 | 5520 | ztest_spa = spa; |
428870ff | 5521 | |
c242c188 | 5522 | VERIFY3U(0, ==, dmu_objset_hold(ztest_opts.zo_pool, FTAG, &os)); |
3541dc6d GA |
5523 | zs->zs_guid = dmu_objset_fsid_guid(os); |
5524 | dmu_objset_rele(os, FTAG); | |
5525 | ||
428870ff | 5526 | spa->spa_dedup_ditto = 2 * ZIO_DEDUPDITTO_MIN; |
34dc7c2f | 5527 | |
fb5f0bc8 BB |
5528 | /* |
5529 | * We don't expect the pool to suspend unless maxfaults == 0, | |
5530 | * in which case ztest_fault_inject() temporarily takes away | |
5531 | * the only valid replica. | |
5532 | */ | |
428870ff | 5533 | if (MAXFAULTS() == 0) |
fb5f0bc8 BB |
5534 | spa->spa_failmode = ZIO_FAILURE_MODE_WAIT; |
5535 | else | |
5536 | spa->spa_failmode = ZIO_FAILURE_MODE_PANIC; | |
5537 | ||
34dc7c2f | 5538 | /* |
b128c09f | 5539 | * Create a thread to periodically resume suspended I/O. |
34dc7c2f | 5540 | */ |
0aebd4f9 ED |
5541 | VERIFY3P((resume_thread = zk_thread_create(NULL, 0, |
5542 | (thread_func_t)ztest_resume_thread, spa, TS_RUN, NULL, 0, 0, | |
5543 | PTHREAD_CREATE_JOINABLE)), !=, NULL); | |
34dc7c2f | 5544 | |
428870ff | 5545 | /* |
1e33ac1e | 5546 | * Set a deadman alarm to abort() if we hang. |
428870ff | 5547 | */ |
1e33ac1e BB |
5548 | signal(SIGALRM, ztest_deadman_alarm); |
5549 | alarm((zs->zs_thread_stop - zs->zs_thread_start) / NANOSEC + GRACE); | |
428870ff | 5550 | |
34dc7c2f BB |
5551 | /* |
5552 | * Verify that we can safely inquire about about any object, | |
5553 | * whether it's allocated or not. To make it interesting, | |
5554 | * we probe a 5-wide window around each power of two. | |
5555 | * This hits all edge cases, including zero and the max. | |
5556 | */ | |
d6320ddb BB |
5557 | for (t = 0; t < 64; t++) { |
5558 | for (d = -5; d <= 5; d++) { | |
34dc7c2f BB |
5559 | error = dmu_object_info(spa->spa_meta_objset, |
5560 | (1ULL << t) + d, NULL); | |
5561 | ASSERT(error == 0 || error == ENOENT || | |
5562 | error == EINVAL); | |
5563 | } | |
5564 | } | |
5565 | ||
5566 | /* | |
428870ff | 5567 | * If we got any ENOSPC errors on the previous run, destroy something. |
34dc7c2f | 5568 | */ |
428870ff | 5569 | if (zs->zs_enospc_count != 0) { |
c242c188 CS |
5570 | int d = ztest_random(ztest_opts.zo_datasets); |
5571 | ztest_dataset_destroy(d); | |
428870ff | 5572 | } |
34dc7c2f BB |
5573 | zs->zs_enospc_count = 0; |
5574 | ||
c242c188 CS |
5575 | tid = umem_zalloc(ztest_opts.zo_threads * sizeof (kt_did_t), |
5576 | UMEM_NOFAIL); | |
34dc7c2f | 5577 | |
c242c188 | 5578 | if (ztest_opts.zo_verbose >= 4) |
34dc7c2f BB |
5579 | (void) printf("starting main threads...\n"); |
5580 | ||
428870ff BB |
5581 | /* |
5582 | * Kick off all the tests that run in parallel. | |
5583 | */ | |
c242c188 | 5584 | for (t = 0; t < ztest_opts.zo_threads; t++) { |
1e33ac1e BB |
5585 | kthread_t *thread; |
5586 | ||
c242c188 CS |
5587 | if (t < ztest_opts.zo_datasets && |
5588 | ztest_dataset_open(t) != 0) | |
428870ff | 5589 | return; |
1e33ac1e | 5590 | |
0aebd4f9 ED |
5591 | VERIFY3P(thread = zk_thread_create(NULL, 0, |
5592 | (thread_func_t)ztest_thread, | |
5593 | (void *)(uintptr_t)t, TS_RUN, NULL, 0, 0, | |
5594 | PTHREAD_CREATE_JOINABLE), !=, NULL); | |
1e33ac1e | 5595 | tid[t] = thread->t_tid; |
34dc7c2f BB |
5596 | } |
5597 | ||
428870ff BB |
5598 | /* |
5599 | * Wait for all of the tests to complete. We go in reverse order | |
5600 | * so we don't close datasets while threads are still using them. | |
5601 | */ | |
c242c188 | 5602 | for (t = ztest_opts.zo_threads - 1; t >= 0; t--) { |
1e33ac1e | 5603 | thread_join(tid[t]); |
c242c188 CS |
5604 | if (t < ztest_opts.zo_datasets) |
5605 | ztest_dataset_close(t); | |
34dc7c2f BB |
5606 | } |
5607 | ||
34dc7c2f BB |
5608 | txg_wait_synced(spa_get_dsl(spa), 0); |
5609 | ||
428870ff BB |
5610 | zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(spa)); |
5611 | zs->zs_space = metaslab_class_get_space(spa_normal_class(spa)); | |
5612 | ||
c242c188 | 5613 | umem_free(tid, ztest_opts.zo_threads * sizeof (kt_did_t)); |
428870ff BB |
5614 | |
5615 | /* Kill the resume thread */ | |
5616 | ztest_exiting = B_TRUE; | |
1e33ac1e | 5617 | thread_join(resume_thread->t_tid); |
428870ff | 5618 | ztest_resume(spa); |
34dc7c2f BB |
5619 | |
5620 | /* | |
428870ff BB |
5621 | * Right before closing the pool, kick off a bunch of async I/O; |
5622 | * spa_close() should wait for it to complete. | |
34dc7c2f | 5623 | */ |
1e33ac1e | 5624 | for (object = 1; object < 50; object++) |
428870ff BB |
5625 | dmu_prefetch(spa->spa_meta_objset, object, 0, 1ULL << 20); |
5626 | ||
090ff092 RC |
5627 | /* Verify that at least one commit cb was called in a timely fashion */ |
5628 | if (zc_cb_counter >= ZTEST_COMMIT_CB_MIN_REG) | |
5629 | VERIFY3U(zc_min_txg_delay, ==, 0); | |
5630 | ||
428870ff BB |
5631 | spa_close(spa, FTAG); |
5632 | ||
5633 | /* | |
5634 | * Verify that we can loop over all pools. | |
5635 | */ | |
5636 | mutex_enter(&spa_namespace_lock); | |
5637 | for (spa = spa_next(NULL); spa != NULL; spa = spa_next(spa)) | |
c242c188 | 5638 | if (ztest_opts.zo_verbose > 3) |
428870ff BB |
5639 | (void) printf("spa_next: found %s\n", spa_name(spa)); |
5640 | mutex_exit(&spa_namespace_lock); | |
5641 | ||
5642 | /* | |
5643 | * Verify that we can export the pool and reimport it under a | |
5644 | * different name. | |
5645 | */ | |
5646 | if (ztest_random(2) == 0) { | |
5647 | char name[MAXNAMELEN]; | |
c242c188 CS |
5648 | (void) snprintf(name, MAXNAMELEN, "%s_import", |
5649 | ztest_opts.zo_pool); | |
5650 | ztest_spa_import_export(ztest_opts.zo_pool, name); | |
5651 | ztest_spa_import_export(name, ztest_opts.zo_pool); | |
428870ff BB |
5652 | } |
5653 | ||
5654 | kernel_fini(); | |
572e2857 BB |
5655 | |
5656 | list_destroy(&zcl.zcl_callbacks); | |
0e8d1b2d | 5657 | mutex_destroy(&zcl.zcl_callbacks_lock); |
c242c188 CS |
5658 | rw_destroy(&ztest_name_lock); |
5659 | mutex_destroy(&ztest_vdev_lock); | |
428870ff BB |
5660 | } |
5661 | ||
5662 | static void | |
c242c188 | 5663 | ztest_freeze(void) |
428870ff | 5664 | { |
c242c188 | 5665 | ztest_ds_t *zd = &ztest_ds[0]; |
428870ff BB |
5666 | spa_t *spa; |
5667 | int numloops = 0; | |
5668 | ||
c242c188 | 5669 | if (ztest_opts.zo_verbose >= 3) |
428870ff | 5670 | (void) printf("testing spa_freeze()...\n"); |
9babb374 | 5671 | |
428870ff | 5672 | kernel_init(FREAD | FWRITE); |
c242c188 CS |
5673 | VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
5674 | VERIFY3U(0, ==, ztest_dataset_open(0)); | |
9babb374 | 5675 | |
428870ff BB |
5676 | /* |
5677 | * Force the first log block to be transactionally allocated. | |
5678 | * We have to do this before we freeze the pool -- otherwise | |
5679 | * the log chain won't be anchored. | |
5680 | */ | |
5681 | while (BP_IS_HOLE(&zd->zd_zilog->zl_header->zh_log)) { | |
5682 | ztest_dmu_object_alloc_free(zd, 0); | |
572e2857 | 5683 | zil_commit(zd->zd_zilog, 0); |
34dc7c2f BB |
5684 | } |
5685 | ||
5686 | txg_wait_synced(spa_get_dsl(spa), 0); | |
5687 | ||
428870ff BB |
5688 | /* |
5689 | * Freeze the pool. This stops spa_sync() from doing anything, | |
5690 | * so that the only way to record changes from now on is the ZIL. | |
5691 | */ | |
5692 | spa_freeze(spa); | |
b128c09f | 5693 | |
428870ff BB |
5694 | /* |
5695 | * Run tests that generate log records but don't alter the pool config | |
5696 | * or depend on DSL sync tasks (snapshots, objset create/destroy, etc). | |
5697 | * We do a txg_wait_synced() after each iteration to force the txg | |
5698 | * to increase well beyond the last synced value in the uberblock. | |
5699 | * The ZIL should be OK with that. | |
5700 | */ | |
c242c188 CS |
5701 | while (ztest_random(10) != 0 && |
5702 | numloops++ < ztest_opts.zo_maxloops) { | |
428870ff BB |
5703 | ztest_dmu_write_parallel(zd, 0); |
5704 | ztest_dmu_object_alloc_free(zd, 0); | |
5705 | txg_wait_synced(spa_get_dsl(spa), 0); | |
5706 | } | |
b128c09f | 5707 | |
34dc7c2f | 5708 | /* |
428870ff | 5709 | * Commit all of the changes we just generated. |
34dc7c2f | 5710 | */ |
572e2857 | 5711 | zil_commit(zd->zd_zilog, 0); |
428870ff | 5712 | txg_wait_synced(spa_get_dsl(spa), 0); |
34dc7c2f | 5713 | |
428870ff BB |
5714 | /* |
5715 | * Close our dataset and close the pool. | |
5716 | */ | |
c242c188 | 5717 | ztest_dataset_close(0); |
34dc7c2f | 5718 | spa_close(spa, FTAG); |
428870ff | 5719 | kernel_fini(); |
34dc7c2f | 5720 | |
428870ff BB |
5721 | /* |
5722 | * Open and close the pool and dataset to induce log replay. | |
5723 | */ | |
5724 | kernel_init(FREAD | FWRITE); | |
c242c188 CS |
5725 | VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
5726 | VERIFY3U(0, ==, ztest_dataset_open(0)); | |
5727 | ztest_dataset_close(0); | |
428870ff | 5728 | spa_close(spa, FTAG); |
34dc7c2f BB |
5729 | kernel_fini(); |
5730 | } | |
5731 | ||
5732 | void | |
5733 | print_time(hrtime_t t, char *timebuf) | |
5734 | { | |
5735 | hrtime_t s = t / NANOSEC; | |
5736 | hrtime_t m = s / 60; | |
5737 | hrtime_t h = m / 60; | |
5738 | hrtime_t d = h / 24; | |
5739 | ||
5740 | s -= m * 60; | |
5741 | m -= h * 60; | |
5742 | h -= d * 24; | |
5743 | ||
5744 | timebuf[0] = '\0'; | |
5745 | ||
5746 | if (d) | |
5747 | (void) sprintf(timebuf, | |
5748 | "%llud%02lluh%02llum%02llus", d, h, m, s); | |
5749 | else if (h) | |
5750 | (void) sprintf(timebuf, "%lluh%02llum%02llus", h, m, s); | |
5751 | else if (m) | |
5752 | (void) sprintf(timebuf, "%llum%02llus", m, s); | |
5753 | else | |
5754 | (void) sprintf(timebuf, "%llus", s); | |
5755 | } | |
5756 | ||
428870ff | 5757 | static nvlist_t * |
0bc8fd78 | 5758 | make_random_props(void) |
428870ff BB |
5759 | { |
5760 | nvlist_t *props; | |
5761 | ||
5762 | if (ztest_random(2) == 0) | |
5763 | return (NULL); | |
5764 | ||
5765 | VERIFY(nvlist_alloc(&props, NV_UNIQUE_NAME, 0) == 0); | |
5766 | VERIFY(nvlist_add_uint64(props, "autoreplace", 1) == 0); | |
5767 | ||
428870ff BB |
5768 | return (props); |
5769 | } | |
5770 | ||
34dc7c2f BB |
5771 | /* |
5772 | * Create a storage pool with the given name and initial vdev size. | |
428870ff | 5773 | * Then test spa_freeze() functionality. |
34dc7c2f BB |
5774 | */ |
5775 | static void | |
428870ff | 5776 | ztest_init(ztest_shared_t *zs) |
34dc7c2f BB |
5777 | { |
5778 | spa_t *spa; | |
428870ff BB |
5779 | nvlist_t *nvroot, *props; |
5780 | ||
c242c188 CS |
5781 | mutex_init(&ztest_vdev_lock, NULL, MUTEX_DEFAULT, NULL); |
5782 | rw_init(&ztest_name_lock, NULL, RW_DEFAULT, NULL); | |
34dc7c2f BB |
5783 | |
5784 | kernel_init(FREAD | FWRITE); | |
5785 | ||
5786 | /* | |
5787 | * Create the storage pool. | |
5788 | */ | |
c242c188 | 5789 | (void) spa_destroy(ztest_opts.zo_pool); |
428870ff BB |
5790 | ztest_shared->zs_vdev_next_leaf = 0; |
5791 | zs->zs_splits = 0; | |
c242c188 CS |
5792 | zs->zs_mirrors = ztest_opts.zo_mirrors; |
5793 | nvroot = make_vdev_root(NULL, NULL, ztest_opts.zo_vdev_size, 0, | |
5794 | 0, ztest_opts.zo_raidz, zs->zs_mirrors, 1); | |
428870ff | 5795 | props = make_random_props(); |
c242c188 CS |
5796 | VERIFY3U(0, ==, spa_create(ztest_opts.zo_pool, nvroot, props, |
5797 | NULL, NULL)); | |
34dc7c2f BB |
5798 | nvlist_free(nvroot); |
5799 | ||
c242c188 CS |
5800 | VERIFY3U(0, ==, spa_open(ztest_opts.zo_pool, &spa, FTAG)); |
5801 | zs->zs_metaslab_sz = | |
5802 | 1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift; | |
34dc7c2f BB |
5803 | spa_close(spa, FTAG); |
5804 | ||
5805 | kernel_fini(); | |
428870ff | 5806 | |
c242c188 CS |
5807 | ztest_run_zdb(ztest_opts.zo_pool); |
5808 | ||
5809 | ztest_freeze(); | |
5810 | ||
5811 | ztest_run_zdb(ztest_opts.zo_pool); | |
5812 | ||
5813 | rw_destroy(&ztest_name_lock); | |
5814 | mutex_destroy(&ztest_vdev_lock); | |
5815 | } | |
5816 | ||
5817 | static void | |
5818 | setup_fds(void) | |
5819 | { | |
5820 | int fd; | |
5821 | ||
5822 | char *tmp = tempnam(NULL, NULL); | |
5823 | fd = open(tmp, O_RDWR | O_CREAT, 0700); | |
5824 | ASSERT3S(fd, >=, 0); | |
c242c188 CS |
5825 | if (fd != ZTEST_FD_DATA) { |
5826 | VERIFY3S(dup2(fd, ZTEST_FD_DATA), ==, ZTEST_FD_DATA); | |
5827 | close(fd); | |
5828 | } | |
5829 | (void) unlink(tmp); | |
5830 | free(tmp); | |
5831 | ||
5832 | fd = open("/dev/urandom", O_RDONLY); | |
5833 | ASSERT3S(fd, >=, 0); | |
5834 | if (fd != ZTEST_FD_RAND) { | |
5835 | VERIFY3S(dup2(fd, ZTEST_FD_RAND), ==, ZTEST_FD_RAND); | |
5836 | close(fd); | |
5837 | } | |
5838 | } | |
5839 | ||
22257dc0 CS |
5840 | static int |
5841 | shared_data_size(ztest_shared_hdr_t *hdr) | |
5842 | { | |
5843 | int size; | |
5844 | ||
5845 | size = hdr->zh_hdr_size; | |
5846 | size += hdr->zh_opts_size; | |
5847 | size += hdr->zh_size; | |
5848 | size += hdr->zh_stats_size * hdr->zh_stats_count; | |
5849 | size += hdr->zh_ds_size * hdr->zh_ds_count; | |
5850 | ||
5851 | return (size); | |
5852 | } | |
5853 | ||
c242c188 CS |
5854 | static void |
5855 | setup_hdr(void) | |
5856 | { | |
22257dc0 | 5857 | int size; |
c242c188 CS |
5858 | ztest_shared_hdr_t *hdr; |
5859 | ||
5860 | hdr = (void *)mmap(0, P2ROUNDUP(sizeof (*hdr), getpagesize()), | |
5861 | PROT_READ | PROT_WRITE, MAP_SHARED, ZTEST_FD_DATA, 0); | |
5862 | ASSERT(hdr != MAP_FAILED); | |
5863 | ||
22257dc0 CS |
5864 | VERIFY3U(0, ==, ftruncate(ZTEST_FD_DATA, sizeof (ztest_shared_hdr_t))); |
5865 | ||
c242c188 CS |
5866 | hdr->zh_hdr_size = sizeof (ztest_shared_hdr_t); |
5867 | hdr->zh_opts_size = sizeof (ztest_shared_opts_t); | |
5868 | hdr->zh_size = sizeof (ztest_shared_t); | |
5869 | hdr->zh_stats_size = sizeof (ztest_shared_callstate_t); | |
5870 | hdr->zh_stats_count = ZTEST_FUNCS; | |
5871 | hdr->zh_ds_size = sizeof (ztest_shared_ds_t); | |
5872 | hdr->zh_ds_count = ztest_opts.zo_datasets; | |
5873 | ||
22257dc0 CS |
5874 | size = shared_data_size(hdr); |
5875 | VERIFY3U(0, ==, ftruncate(ZTEST_FD_DATA, size)); | |
5876 | ||
c242c188 CS |
5877 | (void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize())); |
5878 | } | |
5879 | ||
5880 | static void | |
5881 | setup_data(void) | |
5882 | { | |
5883 | int size, offset; | |
5884 | ztest_shared_hdr_t *hdr; | |
5885 | uint8_t *buf; | |
5886 | ||
5887 | hdr = (void *)mmap(0, P2ROUNDUP(sizeof (*hdr), getpagesize()), | |
5888 | PROT_READ, MAP_SHARED, ZTEST_FD_DATA, 0); | |
5889 | ASSERT(hdr != MAP_FAILED); | |
5890 | ||
22257dc0 | 5891 | size = shared_data_size(hdr); |
c242c188 CS |
5892 | |
5893 | (void) munmap((caddr_t)hdr, P2ROUNDUP(sizeof (*hdr), getpagesize())); | |
5894 | hdr = ztest_shared_hdr = (void *)mmap(0, P2ROUNDUP(size, getpagesize()), | |
5895 | PROT_READ | PROT_WRITE, MAP_SHARED, ZTEST_FD_DATA, 0); | |
5896 | ASSERT(hdr != MAP_FAILED); | |
5897 | buf = (uint8_t *)hdr; | |
5898 | ||
5899 | offset = hdr->zh_hdr_size; | |
5900 | ztest_shared_opts = (void *)&buf[offset]; | |
5901 | offset += hdr->zh_opts_size; | |
5902 | ztest_shared = (void *)&buf[offset]; | |
5903 | offset += hdr->zh_size; | |
5904 | ztest_shared_callstate = (void *)&buf[offset]; | |
5905 | offset += hdr->zh_stats_size * hdr->zh_stats_count; | |
5906 | ztest_shared_ds = (void *)&buf[offset]; | |
5907 | } | |
5908 | ||
5909 | static boolean_t | |
5910 | exec_child(char *cmd, char *libpath, boolean_t ignorekill, int *statusp) | |
5911 | { | |
5912 | pid_t pid; | |
5913 | int status; | |
5914 | char cmdbuf[MAXPATHLEN]; | |
5915 | ||
5916 | pid = fork(); | |
5917 | ||
5918 | if (cmd == NULL) { | |
5919 | (void) strlcpy(cmdbuf, getexecname(), sizeof (cmdbuf)); | |
5920 | cmd = cmdbuf; | |
5921 | } | |
5922 | ||
5923 | if (pid == -1) | |
5924 | fatal(1, "fork failed"); | |
5925 | ||
5926 | if (pid == 0) { /* child */ | |
5927 | char *emptyargv[2] = { cmd, NULL }; | |
5928 | ||
5929 | struct rlimit rl = { 1024, 1024 }; | |
5930 | (void) setrlimit(RLIMIT_NOFILE, &rl); | |
5931 | (void) enable_extended_FILE_stdio(-1, -1); | |
5932 | if (libpath != NULL) | |
5933 | VERIFY(0 == setenv("LD_LIBRARY_PATH", libpath, 1)); | |
5934 | (void) execv(cmd, emptyargv); | |
5935 | ztest_dump_core = B_FALSE; | |
5936 | fatal(B_TRUE, "exec failed: %s", cmd); | |
5937 | } | |
5938 | ||
5939 | while (waitpid(pid, &status, 0) != pid) | |
5940 | continue; | |
5941 | if (statusp != NULL) | |
5942 | *statusp = status; | |
5943 | ||
5944 | if (WIFEXITED(status)) { | |
5945 | if (WEXITSTATUS(status) != 0) { | |
5946 | (void) fprintf(stderr, "child exited with code %d\n", | |
5947 | WEXITSTATUS(status)); | |
5948 | exit(2); | |
5949 | } | |
5950 | return (B_FALSE); | |
5951 | } else if (WIFSIGNALED(status)) { | |
5952 | if (!ignorekill || WTERMSIG(status) != SIGKILL) { | |
5953 | (void) fprintf(stderr, "child died with signal %d\n", | |
5954 | WTERMSIG(status)); | |
5955 | exit(3); | |
5956 | } | |
5957 | return (B_TRUE); | |
5958 | } else { | |
5959 | (void) fprintf(stderr, "something strange happened to child\n"); | |
5960 | exit(4); | |
5961 | /* NOTREACHED */ | |
5962 | } | |
5963 | } | |
5964 | ||
5965 | static void | |
5966 | ztest_run_init(void) | |
5967 | { | |
5968 | int i; | |
428870ff | 5969 | |
c242c188 | 5970 | ztest_shared_t *zs = ztest_shared; |
428870ff | 5971 | |
c242c188 | 5972 | ASSERT(ztest_opts.zo_init != 0); |
572e2857 | 5973 | |
c242c188 CS |
5974 | /* |
5975 | * Blow away any existing copy of zpool.cache | |
5976 | */ | |
5977 | (void) remove(spa_config_path); | |
5978 | ||
5979 | /* | |
5980 | * Create and initialize our storage pool. | |
5981 | */ | |
5982 | for (i = 1; i <= ztest_opts.zo_init; i++) { | |
5983 | bzero(zs, sizeof (ztest_shared_t)); | |
5984 | if (ztest_opts.zo_verbose >= 3 && | |
5985 | ztest_opts.zo_init != 1) { | |
5986 | (void) printf("ztest_init(), pass %d\n", i); | |
5987 | } | |
5988 | ztest_init(zs); | |
5989 | } | |
34dc7c2f BB |
5990 | } |
5991 | ||
5992 | int | |
5993 | main(int argc, char **argv) | |
5994 | { | |
5995 | int kills = 0; | |
5996 | int iters = 0; | |
c242c188 CS |
5997 | int older = 0; |
5998 | int newer = 0; | |
34dc7c2f BB |
5999 | ztest_shared_t *zs; |
6000 | ztest_info_t *zi; | |
c242c188 | 6001 | ztest_shared_callstate_t *zc; |
34dc7c2f BB |
6002 | char timebuf[100]; |
6003 | char numbuf[6]; | |
428870ff | 6004 | spa_t *spa; |
c242c188 CS |
6005 | char cmd[MAXNAMELEN]; |
6006 | boolean_t hasalt; | |
6007 | int f; | |
6008 | boolean_t ischild = (0 == lseek(ZTEST_FD_DATA, 0, SEEK_CUR)); | |
6009 | ||
6010 | ASSERT(ischild || errno == EBADF || errno == ESPIPE); | |
34dc7c2f BB |
6011 | |
6012 | (void) setvbuf(stdout, NULL, _IOLBF, 0); | |
6013 | ||
c242c188 CS |
6014 | if (!ischild) { |
6015 | dprintf_setup(&argc, argv); | |
6016 | process_options(argc, argv); | |
34dc7c2f | 6017 | |
c242c188 CS |
6018 | setup_fds(); |
6019 | setup_hdr(); | |
6020 | setup_data(); | |
6021 | bcopy(&ztest_opts, ztest_shared_opts, | |
6022 | sizeof (*ztest_shared_opts)); | |
6023 | } else { | |
6024 | setup_data(); | |
6025 | bcopy(ztest_shared_opts, &ztest_opts, sizeof (ztest_opts)); | |
6026 | } | |
6027 | ASSERT3U(ztest_opts.zo_datasets, ==, ztest_shared_hdr->zh_ds_count); | |
34dc7c2f | 6028 | |
428870ff | 6029 | /* Override location of zpool.cache */ |
c242c188 CS |
6030 | (void) asprintf((char **)&spa_config_path, "%s/zpool.cache", |
6031 | ztest_opts.zo_dir); | |
6032 | ||
6033 | ztest_ds = umem_alloc(ztest_opts.zo_datasets * sizeof (ztest_ds_t), | |
6034 | UMEM_NOFAIL); | |
6035 | zs = ztest_shared; | |
6036 | ||
6037 | if (ischild) { | |
6038 | metaslab_gang_bang = ztest_opts.zo_metaslab_gang_bang; | |
6039 | metaslab_df_alloc_threshold = | |
6040 | zs->zs_metaslab_df_alloc_threshold; | |
6041 | ||
6042 | if (zs->zs_do_init) | |
6043 | ztest_run_init(); | |
6044 | else | |
6045 | ztest_run(zs); | |
6046 | exit(0); | |
6047 | } | |
34dc7c2f | 6048 | |
c242c188 | 6049 | hasalt = (strlen(ztest_opts.zo_alt_ztest) != 0); |
34dc7c2f | 6050 | |
c242c188 | 6051 | if (ztest_opts.zo_verbose >= 1) { |
34dc7c2f BB |
6052 | (void) printf("%llu vdevs, %d datasets, %d threads," |
6053 | " %llu seconds...\n", | |
c242c188 CS |
6054 | (u_longlong_t)ztest_opts.zo_vdevs, |
6055 | ztest_opts.zo_datasets, | |
6056 | ztest_opts.zo_threads, | |
6057 | (u_longlong_t)ztest_opts.zo_time); | |
34dc7c2f BB |
6058 | } |
6059 | ||
c242c188 CS |
6060 | (void) strlcpy(cmd, getexecname(), sizeof (cmd)); |
6061 | ||
6062 | zs->zs_do_init = B_TRUE; | |
6063 | if (strlen(ztest_opts.zo_alt_ztest) != 0) { | |
6064 | if (ztest_opts.zo_verbose >= 1) { | |
6065 | (void) printf("Executing older ztest for " | |
6066 | "initialization: %s\n", ztest_opts.zo_alt_ztest); | |
6067 | } | |
6068 | VERIFY(!exec_child(ztest_opts.zo_alt_ztest, | |
6069 | ztest_opts.zo_alt_libpath, B_FALSE, NULL)); | |
6070 | } else { | |
6071 | VERIFY(!exec_child(NULL, NULL, B_FALSE, NULL)); | |
34dc7c2f | 6072 | } |
c242c188 | 6073 | zs->zs_do_init = B_FALSE; |
34dc7c2f | 6074 | |
428870ff | 6075 | zs->zs_proc_start = gethrtime(); |
c242c188 | 6076 | zs->zs_proc_stop = zs->zs_proc_start + ztest_opts.zo_time * NANOSEC; |
34dc7c2f | 6077 | |
d6320ddb | 6078 | for (f = 0; f < ZTEST_FUNCS; f++) { |
c242c188 CS |
6079 | zi = &ztest_info[f]; |
6080 | zc = ZTEST_GET_SHARED_CALLSTATE(f); | |
428870ff | 6081 | if (zs->zs_proc_start + zi->zi_interval[0] > zs->zs_proc_stop) |
c242c188 | 6082 | zc->zc_next = UINT64_MAX; |
34dc7c2f | 6083 | else |
c242c188 | 6084 | zc->zc_next = zs->zs_proc_start + |
428870ff | 6085 | ztest_random(2 * zi->zi_interval[0] + 1); |
34dc7c2f BB |
6086 | } |
6087 | ||
34dc7c2f BB |
6088 | /* |
6089 | * Run the tests in a loop. These tests include fault injection | |
6090 | * to verify that self-healing data works, and forced crashes | |
6091 | * to verify that we never lose on-disk consistency. | |
6092 | */ | |
428870ff | 6093 | while (gethrtime() < zs->zs_proc_stop) { |
34dc7c2f | 6094 | int status; |
c242c188 | 6095 | boolean_t killed; |
34dc7c2f BB |
6096 | |
6097 | /* | |
6098 | * Initialize the workload counters for each function. | |
6099 | */ | |
d6320ddb | 6100 | for (f = 0; f < ZTEST_FUNCS; f++) { |
c242c188 CS |
6101 | zc = ZTEST_GET_SHARED_CALLSTATE(f); |
6102 | zc->zc_count = 0; | |
6103 | zc->zc_time = 0; | |
34dc7c2f BB |
6104 | } |
6105 | ||
9babb374 | 6106 | /* Set the allocation switch size */ |
c242c188 CS |
6107 | zs->zs_metaslab_df_alloc_threshold = |
6108 | ztest_random(zs->zs_metaslab_sz / 4) + 1; | |
9babb374 | 6109 | |
c242c188 CS |
6110 | if (!hasalt || ztest_random(2) == 0) { |
6111 | if (hasalt && ztest_opts.zo_verbose >= 1) { | |
6112 | (void) printf("Executing newer ztest: %s\n", | |
6113 | cmd); | |
34dc7c2f | 6114 | } |
c242c188 CS |
6115 | newer++; |
6116 | killed = exec_child(cmd, NULL, B_TRUE, &status); | |
34dc7c2f | 6117 | } else { |
c242c188 CS |
6118 | if (hasalt && ztest_opts.zo_verbose >= 1) { |
6119 | (void) printf("Executing older ztest: %s\n", | |
6120 | ztest_opts.zo_alt_ztest); | |
6121 | } | |
6122 | older++; | |
6123 | killed = exec_child(ztest_opts.zo_alt_ztest, | |
6124 | ztest_opts.zo_alt_libpath, B_TRUE, &status); | |
34dc7c2f BB |
6125 | } |
6126 | ||
c242c188 CS |
6127 | if (killed) |
6128 | kills++; | |
34dc7c2f BB |
6129 | iters++; |
6130 | ||
c242c188 | 6131 | if (ztest_opts.zo_verbose >= 1) { |
34dc7c2f BB |
6132 | hrtime_t now = gethrtime(); |
6133 | ||
428870ff BB |
6134 | now = MIN(now, zs->zs_proc_stop); |
6135 | print_time(zs->zs_proc_stop - now, timebuf); | |
34dc7c2f BB |
6136 | nicenum(zs->zs_space, numbuf); |
6137 | ||
6138 | (void) printf("Pass %3d, %8s, %3llu ENOSPC, " | |
6139 | "%4.1f%% of %5s used, %3.0f%% done, %8s to go\n", | |
6140 | iters, | |
6141 | WIFEXITED(status) ? "Complete" : "SIGKILL", | |
6142 | (u_longlong_t)zs->zs_enospc_count, | |
6143 | 100.0 * zs->zs_alloc / zs->zs_space, | |
6144 | numbuf, | |
428870ff | 6145 | 100.0 * (now - zs->zs_proc_start) / |
c242c188 | 6146 | (ztest_opts.zo_time * NANOSEC), timebuf); |
34dc7c2f BB |
6147 | } |
6148 | ||
c242c188 | 6149 | if (ztest_opts.zo_verbose >= 2) { |
34dc7c2f BB |
6150 | (void) printf("\nWorkload summary:\n\n"); |
6151 | (void) printf("%7s %9s %s\n", | |
6152 | "Calls", "Time", "Function"); | |
6153 | (void) printf("%7s %9s %s\n", | |
6154 | "-----", "----", "--------"); | |
d6320ddb | 6155 | for (f = 0; f < ZTEST_FUNCS; f++) { |
34dc7c2f BB |
6156 | Dl_info dli; |
6157 | ||
c242c188 CS |
6158 | zi = &ztest_info[f]; |
6159 | zc = ZTEST_GET_SHARED_CALLSTATE(f); | |
6160 | print_time(zc->zc_time, timebuf); | |
34dc7c2f BB |
6161 | (void) dladdr((void *)zi->zi_func, &dli); |
6162 | (void) printf("%7llu %9s %s\n", | |
c242c188 | 6163 | (u_longlong_t)zc->zc_count, timebuf, |
34dc7c2f BB |
6164 | dli.dli_sname); |
6165 | } | |
6166 | (void) printf("\n"); | |
6167 | } | |
6168 | ||
6169 | /* | |
428870ff BB |
6170 | * It's possible that we killed a child during a rename test, |
6171 | * in which case we'll have a 'ztest_tmp' pool lying around | |
6172 | * instead of 'ztest'. Do a blind rename in case this happened. | |
34dc7c2f | 6173 | */ |
428870ff | 6174 | kernel_init(FREAD); |
c242c188 | 6175 | if (spa_open(ztest_opts.zo_pool, &spa, FTAG) == 0) { |
428870ff BB |
6176 | spa_close(spa, FTAG); |
6177 | } else { | |
6178 | char tmpname[MAXNAMELEN]; | |
6179 | kernel_fini(); | |
6180 | kernel_init(FREAD | FWRITE); | |
6181 | (void) snprintf(tmpname, sizeof (tmpname), "%s_tmp", | |
c242c188 CS |
6182 | ztest_opts.zo_pool); |
6183 | (void) spa_rename(tmpname, ztest_opts.zo_pool); | |
428870ff | 6184 | } |
34dc7c2f | 6185 | kernel_fini(); |
34dc7c2f | 6186 | |
c242c188 | 6187 | ztest_run_zdb(ztest_opts.zo_pool); |
428870ff | 6188 | } |
34dc7c2f | 6189 | |
c242c188 CS |
6190 | if (ztest_opts.zo_verbose >= 1) { |
6191 | if (hasalt) { | |
6192 | (void) printf("%d runs of older ztest: %s\n", older, | |
6193 | ztest_opts.zo_alt_ztest); | |
6194 | (void) printf("%d runs of newer ztest: %s\n", newer, | |
6195 | cmd); | |
6196 | } | |
34dc7c2f BB |
6197 | (void) printf("%d killed, %d completed, %.0f%% kill rate\n", |
6198 | kills, iters - kills, (100.0 * kills) / MAX(1, iters)); | |
6199 | } | |
6200 | ||
6201 | return (0); | |
6202 | } |