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.
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.
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]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
24 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
25 * Copyright (c) 2013 Steven Hartland. All rights reserved.
29 * The objective of this program is to provide a DMU/ZAP/SPA stress test
30 * that runs entirely in userland, is easy to use, and easy to extend.
32 * The overall design of the ztest program is as follows:
34 * (1) For each major functional area (e.g. adding vdevs to a pool,
35 * creating and destroying datasets, reading and writing objects, etc)
36 * we have a simple routine to test that functionality. These
37 * individual routines do not have to do anything "stressful".
39 * (2) We turn these simple functionality tests into a stress test by
40 * running them all in parallel, with as many threads as desired,
41 * and spread across as many datasets, objects, and vdevs as desired.
43 * (3) While all this is happening, we inject faults into the pool to
44 * verify that self-healing data really works.
46 * (4) Every time we open a dataset, we change its checksum and compression
47 * functions. Thus even individual objects vary from block to block
48 * in which checksum they use and whether they're compressed.
50 * (5) To verify that we never lose on-disk consistency after a crash,
51 * we run the entire test in a child of the main process.
52 * At random times, the child self-immolates with a SIGKILL.
53 * This is the software equivalent of pulling the power cord.
54 * The parent then runs the test again, using the existing
55 * storage pool, as many times as desired. If backwards compatibility
56 * testing is enabled ztest will sometimes run the "older" version
57 * of ztest after a SIGKILL.
59 * (6) To verify that we don't have future leaks or temporal incursions,
60 * many of the functional tests record the transaction group number
61 * as part of their data. When reading old data, they verify that
62 * the transaction group number is less than the current, open txg.
63 * If you add a new test, please do this if applicable.
65 * (7) Threads are created with a reduced stack size, for sanity checking.
66 * Therefore, it's important not to allocate huge buffers on the stack.
68 * When run with no arguments, ztest runs for about five minutes and
69 * produces no output if successful. To get a little bit of information,
70 * specify -V. To get more information, specify -VV, and so on.
72 * To turn this into an overnight stress test, use -T to specify run time.
74 * You can ask more more vdevs [-v], datasets [-d], or threads [-t]
75 * to increase the pool capacity, fanout, and overall stress level.
77 * Use the -k option to set the desired frequency of kills.
79 * When ztest invokes itself it passes all relevant information through a
80 * temporary file which is mmap-ed in the child process. This allows shared
81 * memory to survive the exec syscall. The ztest_shared_hdr_t struct is always
82 * stored at offset 0 of this file and contains information on the size and
83 * number of shared structures in the file. The information stored in this file
84 * must remain backwards compatible with older versions of ztest so that
85 * ztest can invoke them during backwards compatibility testing (-B).
88 #include <sys/zfs_context.h>
94 #include <sys/dmu_objset.h>
100 #include <sys/resource.h>
103 #include <sys/zil_impl.h>
104 #include <sys/vdev_impl.h>
105 #include <sys/vdev_file.h>
106 #include <sys/spa_impl.h>
107 #include <sys/metaslab_impl.h>
108 #include <sys/dsl_prop.h>
109 #include <sys/dsl_dataset.h>
110 #include <sys/dsl_destroy.h>
111 #include <sys/dsl_scan.h>
112 #include <sys/zio_checksum.h>
113 #include <sys/refcount.h>
114 #include <sys/zfeature.h>
115 #include <sys/dsl_userhold.h>
117 #include <stdio_ext.h>
124 #include <sys/fs/zfs.h>
125 #include <libnvpair.h>
127 #include <execinfo.h> /* for backtrace() */
130 static int ztest_fd_data
= -1;
131 static int ztest_fd_rand
= -1;
133 typedef struct ztest_shared_hdr
{
134 uint64_t zh_hdr_size
;
135 uint64_t zh_opts_size
;
137 uint64_t zh_stats_size
;
138 uint64_t zh_stats_count
;
140 uint64_t zh_ds_count
;
141 } ztest_shared_hdr_t
;
143 static ztest_shared_hdr_t
*ztest_shared_hdr
;
145 typedef struct ztest_shared_opts
{
146 char zo_pool
[MAXNAMELEN
];
147 char zo_dir
[MAXNAMELEN
];
148 char zo_alt_ztest
[MAXNAMELEN
];
149 char zo_alt_libpath
[MAXNAMELEN
];
151 uint64_t zo_vdevtime
;
159 uint64_t zo_passtime
;
160 uint64_t zo_killrate
;
164 uint64_t zo_maxloops
;
165 uint64_t zo_metaslab_gang_bang
;
166 } ztest_shared_opts_t
;
168 static const ztest_shared_opts_t ztest_opts_defaults
= {
169 .zo_pool
= { 'z', 't', 'e', 's', 't', '\0' },
170 .zo_dir
= { '/', 't', 'm', 'p', '\0' },
171 .zo_alt_ztest
= { '\0' },
172 .zo_alt_libpath
= { '\0' },
174 .zo_ashift
= SPA_MINBLOCKSHIFT
,
177 .zo_raidz_parity
= 1,
178 .zo_vdev_size
= SPA_MINDEVSIZE
,
181 .zo_passtime
= 60, /* 60 seconds */
182 .zo_killrate
= 70, /* 70% kill rate */
185 .zo_time
= 300, /* 5 minutes */
186 .zo_maxloops
= 50, /* max loops during spa_freeze() */
187 .zo_metaslab_gang_bang
= 32 << 10
190 extern uint64_t metaslab_gang_bang
;
191 extern uint64_t metaslab_df_alloc_threshold
;
192 extern int metaslab_preload_limit
;
194 static ztest_shared_opts_t
*ztest_shared_opts
;
195 static ztest_shared_opts_t ztest_opts
;
197 typedef struct ztest_shared_ds
{
201 static ztest_shared_ds_t
*ztest_shared_ds
;
202 #define ZTEST_GET_SHARED_DS(d) (&ztest_shared_ds[d])
204 #define BT_MAGIC 0x123456789abcdefULL
205 #define MAXFAULTS() \
206 (MAX(zs->zs_mirrors, 1) * (ztest_opts.zo_raidz_parity + 1) - 1)
210 ZTEST_IO_WRITE_PATTERN
,
211 ZTEST_IO_WRITE_ZEROES
,
218 typedef struct ztest_block_tag
{
228 typedef struct bufwad
{
235 * XXX -- fix zfs range locks to be generic so we can use them here.
257 #define ZTEST_RANGE_LOCKS 64
258 #define ZTEST_OBJECT_LOCKS 64
261 * Object descriptor. Used as a template for object lookup/create/remove.
263 typedef struct ztest_od
{
266 dmu_object_type_t od_type
;
267 dmu_object_type_t od_crtype
;
268 uint64_t od_blocksize
;
269 uint64_t od_crblocksize
;
272 char od_name
[MAXNAMELEN
];
278 typedef struct ztest_ds
{
279 ztest_shared_ds_t
*zd_shared
;
281 rwlock_t zd_zilog_lock
;
283 ztest_od_t
*zd_od
; /* debugging aid */
284 char zd_name
[MAXNAMELEN
];
285 kmutex_t zd_dirobj_lock
;
286 rll_t zd_object_lock
[ZTEST_OBJECT_LOCKS
];
287 rll_t zd_range_lock
[ZTEST_RANGE_LOCKS
];
291 * Per-iteration state.
293 typedef void ztest_func_t(ztest_ds_t
*zd
, uint64_t id
);
295 typedef struct ztest_info
{
296 ztest_func_t
*zi_func
; /* test function */
297 uint64_t zi_iters
; /* iterations per execution */
298 uint64_t *zi_interval
; /* execute every <interval> seconds */
299 const char *zi_funcname
; /* name of test function */
302 typedef struct ztest_shared_callstate
{
303 uint64_t zc_count
; /* per-pass count */
304 uint64_t zc_time
; /* per-pass time */
305 uint64_t zc_next
; /* next time to call this function */
306 } ztest_shared_callstate_t
;
308 static ztest_shared_callstate_t
*ztest_shared_callstate
;
309 #define ZTEST_GET_SHARED_CALLSTATE(c) (&ztest_shared_callstate[c])
311 ztest_func_t ztest_dmu_read_write
;
312 ztest_func_t ztest_dmu_write_parallel
;
313 ztest_func_t ztest_dmu_object_alloc_free
;
314 ztest_func_t ztest_dmu_commit_callbacks
;
315 ztest_func_t ztest_zap
;
316 ztest_func_t ztest_zap_parallel
;
317 ztest_func_t ztest_zil_commit
;
318 ztest_func_t ztest_zil_remount
;
319 ztest_func_t ztest_dmu_read_write_zcopy
;
320 ztest_func_t ztest_dmu_objset_create_destroy
;
321 ztest_func_t ztest_dmu_prealloc
;
322 ztest_func_t ztest_fzap
;
323 ztest_func_t ztest_dmu_snapshot_create_destroy
;
324 ztest_func_t ztest_dsl_prop_get_set
;
325 ztest_func_t ztest_spa_prop_get_set
;
326 ztest_func_t ztest_spa_create_destroy
;
327 ztest_func_t ztest_fault_inject
;
328 ztest_func_t ztest_ddt_repair
;
329 ztest_func_t ztest_dmu_snapshot_hold
;
330 ztest_func_t ztest_spa_rename
;
331 ztest_func_t ztest_scrub
;
332 ztest_func_t ztest_dsl_dataset_promote_busy
;
333 ztest_func_t ztest_vdev_attach_detach
;
334 ztest_func_t ztest_vdev_LUN_growth
;
335 ztest_func_t ztest_vdev_add_remove
;
336 ztest_func_t ztest_vdev_aux_add_remove
;
337 ztest_func_t ztest_split_pool
;
338 ztest_func_t ztest_reguid
;
339 ztest_func_t ztest_spa_upgrade
;
341 uint64_t zopt_always
= 0ULL * NANOSEC
; /* all the time */
342 uint64_t zopt_incessant
= 1ULL * NANOSEC
/ 10; /* every 1/10 second */
343 uint64_t zopt_often
= 1ULL * NANOSEC
; /* every second */
344 uint64_t zopt_sometimes
= 10ULL * NANOSEC
; /* every 10 seconds */
345 uint64_t zopt_rarely
= 60ULL * NANOSEC
; /* every 60 seconds */
347 #define ZTI_INIT(func, iters, interval) \
348 { .zi_func = (func), \
349 .zi_iters = (iters), \
350 .zi_interval = (interval), \
351 .zi_funcname = # func }
353 ztest_info_t ztest_info
[] = {
354 ZTI_INIT(ztest_dmu_read_write
, 1, &zopt_always
),
355 ZTI_INIT(ztest_dmu_write_parallel
, 10, &zopt_always
),
356 ZTI_INIT(ztest_dmu_object_alloc_free
, 1, &zopt_always
),
357 ZTI_INIT(ztest_dmu_commit_callbacks
, 1, &zopt_always
),
358 ZTI_INIT(ztest_zap
, 30, &zopt_always
),
359 ZTI_INIT(ztest_zap_parallel
, 100, &zopt_always
),
360 ZTI_INIT(ztest_split_pool
, 1, &zopt_always
),
361 ZTI_INIT(ztest_zil_commit
, 1, &zopt_incessant
),
362 ZTI_INIT(ztest_zil_remount
, 1, &zopt_sometimes
),
363 ZTI_INIT(ztest_dmu_read_write_zcopy
, 1, &zopt_often
),
364 ZTI_INIT(ztest_dmu_objset_create_destroy
, 1, &zopt_often
),
365 ZTI_INIT(ztest_dsl_prop_get_set
, 1, &zopt_often
),
366 ZTI_INIT(ztest_spa_prop_get_set
, 1, &zopt_sometimes
),
368 ZTI_INIT(ztest_dmu_prealloc
, 1, &zopt_sometimes
),
370 ZTI_INIT(ztest_fzap
, 1, &zopt_sometimes
),
371 ZTI_INIT(ztest_dmu_snapshot_create_destroy
, 1, &zopt_sometimes
),
372 ZTI_INIT(ztest_spa_create_destroy
, 1, &zopt_sometimes
),
373 ZTI_INIT(ztest_fault_inject
, 1, &zopt_sometimes
),
374 ZTI_INIT(ztest_ddt_repair
, 1, &zopt_sometimes
),
375 ZTI_INIT(ztest_dmu_snapshot_hold
, 1, &zopt_sometimes
),
376 ZTI_INIT(ztest_reguid
, 1, &zopt_rarely
),
377 ZTI_INIT(ztest_spa_rename
, 1, &zopt_rarely
),
378 ZTI_INIT(ztest_scrub
, 1, &zopt_rarely
),
379 ZTI_INIT(ztest_spa_upgrade
, 1, &zopt_rarely
),
380 ZTI_INIT(ztest_dsl_dataset_promote_busy
, 1, &zopt_rarely
),
381 ZTI_INIT(ztest_vdev_attach_detach
, 1, &zopt_sometimes
),
382 ZTI_INIT(ztest_vdev_LUN_growth
, 1, &zopt_rarely
),
383 ZTI_INIT(ztest_vdev_add_remove
, 1, &ztest_opts
.zo_vdevtime
),
384 ZTI_INIT(ztest_vdev_aux_add_remove
, 1, &ztest_opts
.zo_vdevtime
),
387 #define ZTEST_FUNCS (sizeof (ztest_info) / sizeof (ztest_info_t))
390 * The following struct is used to hold a list of uncalled commit callbacks.
391 * The callbacks are ordered by txg number.
393 typedef struct ztest_cb_list
{
394 kmutex_t zcl_callbacks_lock
;
395 list_t zcl_callbacks
;
399 * Stuff we need to share writably between parent and child.
401 typedef struct ztest_shared
{
402 boolean_t zs_do_init
;
403 hrtime_t zs_proc_start
;
404 hrtime_t zs_proc_stop
;
405 hrtime_t zs_thread_start
;
406 hrtime_t zs_thread_stop
;
407 hrtime_t zs_thread_kill
;
408 uint64_t zs_enospc_count
;
409 uint64_t zs_vdev_next_leaf
;
410 uint64_t zs_vdev_aux
;
415 uint64_t zs_metaslab_sz
;
416 uint64_t zs_metaslab_df_alloc_threshold
;
420 #define ID_PARALLEL -1ULL
422 static char ztest_dev_template
[] = "%s/%s.%llua";
423 static char ztest_aux_template
[] = "%s/%s.%s.%llu";
424 ztest_shared_t
*ztest_shared
;
426 static spa_t
*ztest_spa
= NULL
;
427 static ztest_ds_t
*ztest_ds
;
429 static kmutex_t ztest_vdev_lock
;
432 * The ztest_name_lock protects the pool and dataset namespace used by
433 * the individual tests. To modify the namespace, consumers must grab
434 * this lock as writer. Grabbing the lock as reader will ensure that the
435 * namespace does not change while the lock is held.
437 static rwlock_t ztest_name_lock
;
439 static boolean_t ztest_dump_core
= B_TRUE
;
440 static boolean_t ztest_exiting
;
442 /* Global commit callback list */
443 static ztest_cb_list_t zcl
;
444 /* Commit cb delay */
445 static uint64_t zc_min_txg_delay
= UINT64_MAX
;
446 static int zc_cb_counter
= 0;
449 * Minimum number of commit callbacks that need to be registered for us to check
450 * whether the minimum txg delay is acceptable.
452 #define ZTEST_COMMIT_CB_MIN_REG 100
455 * If a number of txgs equal to this threshold have been created after a commit
456 * callback has been registered but not called, then we assume there is an
457 * implementation bug.
459 #define ZTEST_COMMIT_CB_THRESH (TXG_CONCURRENT_STATES + 1000)
461 extern uint64_t metaslab_gang_bang
;
462 extern uint64_t metaslab_df_alloc_threshold
;
465 ZTEST_META_DNODE
= 0,
470 static void usage(boolean_t
) __NORETURN
;
473 * These libumem hooks provide a reasonable set of defaults for the allocator's
474 * debugging facilities.
477 _umem_debug_init(void)
479 return ("default,verbose"); /* $UMEM_DEBUG setting */
483 _umem_logging_init(void)
485 return ("fail,contents"); /* $UMEM_LOGGING setting */
488 #define BACKTRACE_SZ 100
490 static void sig_handler(int signo
)
492 struct sigaction action
;
493 #ifdef __GNUC__ /* backtrace() is a GNU extension */
495 void *buffer
[BACKTRACE_SZ
];
497 nptrs
= backtrace(buffer
, BACKTRACE_SZ
);
498 backtrace_symbols_fd(buffer
, nptrs
, STDERR_FILENO
);
502 * Restore default action and re-raise signal so SIGSEGV and
503 * SIGABRT can trigger a core dump.
505 action
.sa_handler
= SIG_DFL
;
506 sigemptyset(&action
.sa_mask
);
508 (void) sigaction(signo
, &action
, NULL
);
512 #define FATAL_MSG_SZ 1024
517 fatal(int do_perror
, char *message
, ...)
520 int save_errno
= errno
;
523 (void) fflush(stdout
);
524 buf
= umem_alloc(FATAL_MSG_SZ
, UMEM_NOFAIL
);
526 va_start(args
, message
);
527 (void) sprintf(buf
, "ztest: ");
529 (void) vsprintf(buf
+ strlen(buf
), message
, args
);
532 (void) snprintf(buf
+ strlen(buf
), FATAL_MSG_SZ
- strlen(buf
),
533 ": %s", strerror(save_errno
));
535 (void) fprintf(stderr
, "%s\n", buf
);
536 fatal_msg
= buf
; /* to ease debugging */
543 str2shift(const char *buf
)
545 const char *ends
= "BKMGTPEZ";
550 for (i
= 0; i
< strlen(ends
); i
++) {
551 if (toupper(buf
[0]) == ends
[i
])
554 if (i
== strlen(ends
)) {
555 (void) fprintf(stderr
, "ztest: invalid bytes suffix: %s\n",
559 if (buf
[1] == '\0' || (toupper(buf
[1]) == 'B' && buf
[2] == '\0')) {
562 (void) fprintf(stderr
, "ztest: invalid bytes suffix: %s\n", buf
);
568 nicenumtoull(const char *buf
)
573 val
= strtoull(buf
, &end
, 0);
575 (void) fprintf(stderr
, "ztest: bad numeric value: %s\n", buf
);
577 } else if (end
[0] == '.') {
578 double fval
= strtod(buf
, &end
);
579 fval
*= pow(2, str2shift(end
));
580 if (fval
> UINT64_MAX
) {
581 (void) fprintf(stderr
, "ztest: value too large: %s\n",
585 val
= (uint64_t)fval
;
587 int shift
= str2shift(end
);
588 if (shift
>= 64 || (val
<< shift
) >> shift
!= val
) {
589 (void) fprintf(stderr
, "ztest: value too large: %s\n",
599 usage(boolean_t requested
)
601 const ztest_shared_opts_t
*zo
= &ztest_opts_defaults
;
603 char nice_vdev_size
[10];
604 char nice_gang_bang
[10];
605 FILE *fp
= requested
? stdout
: stderr
;
607 nicenum(zo
->zo_vdev_size
, nice_vdev_size
);
608 nicenum(zo
->zo_metaslab_gang_bang
, nice_gang_bang
);
610 (void) fprintf(fp
, "Usage: %s\n"
611 "\t[-v vdevs (default: %llu)]\n"
612 "\t[-s size_of_each_vdev (default: %s)]\n"
613 "\t[-a alignment_shift (default: %d)] use 0 for random\n"
614 "\t[-m mirror_copies (default: %d)]\n"
615 "\t[-r raidz_disks (default: %d)]\n"
616 "\t[-R raidz_parity (default: %d)]\n"
617 "\t[-d datasets (default: %d)]\n"
618 "\t[-t threads (default: %d)]\n"
619 "\t[-g gang_block_threshold (default: %s)]\n"
620 "\t[-i init_count (default: %d)] initialize pool i times\n"
621 "\t[-k kill_percentage (default: %llu%%)]\n"
622 "\t[-p pool_name (default: %s)]\n"
623 "\t[-f dir (default: %s)] file directory for vdev files\n"
624 "\t[-V] verbose (use multiple times for ever more blather)\n"
625 "\t[-E] use existing pool instead of creating new one\n"
626 "\t[-T time (default: %llu sec)] total run time\n"
627 "\t[-F freezeloops (default: %llu)] max loops in spa_freeze()\n"
628 "\t[-P passtime (default: %llu sec)] time per pass\n"
629 "\t[-B alt_ztest (default: <none>)] alternate ztest path\n"
630 "\t[-h] (print help)\n"
633 (u_longlong_t
)zo
->zo_vdevs
, /* -v */
634 nice_vdev_size
, /* -s */
635 zo
->zo_ashift
, /* -a */
636 zo
->zo_mirrors
, /* -m */
637 zo
->zo_raidz
, /* -r */
638 zo
->zo_raidz_parity
, /* -R */
639 zo
->zo_datasets
, /* -d */
640 zo
->zo_threads
, /* -t */
641 nice_gang_bang
, /* -g */
642 zo
->zo_init
, /* -i */
643 (u_longlong_t
)zo
->zo_killrate
, /* -k */
644 zo
->zo_pool
, /* -p */
646 (u_longlong_t
)zo
->zo_time
, /* -T */
647 (u_longlong_t
)zo
->zo_maxloops
, /* -F */
648 (u_longlong_t
)zo
->zo_passtime
);
649 exit(requested
? 0 : 1);
653 process_options(int argc
, char **argv
)
656 ztest_shared_opts_t
*zo
= &ztest_opts
;
660 char altdir
[MAXNAMELEN
] = { 0 };
662 bcopy(&ztest_opts_defaults
, zo
, sizeof (*zo
));
664 while ((opt
= getopt(argc
, argv
,
665 "v:s:a:m:r:R:d:t:g:i:k:p:f:VET:P:hF:B:")) != EOF
) {
682 value
= nicenumtoull(optarg
);
686 zo
->zo_vdevs
= value
;
689 zo
->zo_vdev_size
= MAX(SPA_MINDEVSIZE
, value
);
692 zo
->zo_ashift
= value
;
695 zo
->zo_mirrors
= value
;
698 zo
->zo_raidz
= MAX(1, value
);
701 zo
->zo_raidz_parity
= MIN(MAX(value
, 1), 3);
704 zo
->zo_datasets
= MAX(1, value
);
707 zo
->zo_threads
= MAX(1, value
);
710 zo
->zo_metaslab_gang_bang
= MAX(SPA_MINBLOCKSIZE
<< 1,
717 zo
->zo_killrate
= value
;
720 (void) strlcpy(zo
->zo_pool
, optarg
,
721 sizeof (zo
->zo_pool
));
724 path
= realpath(optarg
, NULL
);
726 (void) fprintf(stderr
, "error: %s: %s\n",
727 optarg
, strerror(errno
));
730 (void) strlcpy(zo
->zo_dir
, path
,
731 sizeof (zo
->zo_dir
));
744 zo
->zo_passtime
= MAX(1, value
);
747 zo
->zo_maxloops
= MAX(1, value
);
750 (void) strlcpy(altdir
, optarg
, sizeof (altdir
));
762 zo
->zo_raidz_parity
= MIN(zo
->zo_raidz_parity
, zo
->zo_raidz
- 1);
765 (zo
->zo_vdevs
> 0 ? zo
->zo_time
* NANOSEC
/ zo
->zo_vdevs
:
768 if (strlen(altdir
) > 0) {
776 cmd
= umem_alloc(MAXPATHLEN
, UMEM_NOFAIL
);
777 realaltdir
= umem_alloc(MAXPATHLEN
, UMEM_NOFAIL
);
779 VERIFY(NULL
!= realpath(getexecname(), cmd
));
780 if (0 != access(altdir
, F_OK
)) {
781 ztest_dump_core
= B_FALSE
;
782 fatal(B_TRUE
, "invalid alternate ztest path: %s",
785 VERIFY(NULL
!= realpath(altdir
, realaltdir
));
788 * 'cmd' should be of the form "<anything>/usr/bin/<isa>/ztest".
789 * We want to extract <isa> to determine if we should use
790 * 32 or 64 bit binaries.
792 bin
= strstr(cmd
, "/usr/bin/");
793 ztest
= strstr(bin
, "/ztest");
795 isalen
= ztest
- isa
;
796 (void) snprintf(zo
->zo_alt_ztest
, sizeof (zo
->zo_alt_ztest
),
797 "%s/usr/bin/%.*s/ztest", realaltdir
, isalen
, isa
);
798 (void) snprintf(zo
->zo_alt_libpath
, sizeof (zo
->zo_alt_libpath
),
799 "%s/usr/lib/%.*s", realaltdir
, isalen
, isa
);
801 if (0 != access(zo
->zo_alt_ztest
, X_OK
)) {
802 ztest_dump_core
= B_FALSE
;
803 fatal(B_TRUE
, "invalid alternate ztest: %s",
805 } else if (0 != access(zo
->zo_alt_libpath
, X_OK
)) {
806 ztest_dump_core
= B_FALSE
;
807 fatal(B_TRUE
, "invalid alternate lib directory %s",
811 umem_free(cmd
, MAXPATHLEN
);
812 umem_free(realaltdir
, MAXPATHLEN
);
817 ztest_kill(ztest_shared_t
*zs
)
819 zs
->zs_alloc
= metaslab_class_get_alloc(spa_normal_class(ztest_spa
));
820 zs
->zs_space
= metaslab_class_get_space(spa_normal_class(ztest_spa
));
823 * Before we kill off ztest, make sure that the config is updated.
824 * See comment above spa_config_sync().
826 mutex_enter(&spa_namespace_lock
);
827 spa_config_sync(ztest_spa
, B_FALSE
, B_FALSE
);
828 mutex_exit(&spa_namespace_lock
);
830 if (ztest_opts
.zo_verbose
>= 3)
831 zfs_dbgmsg_print(FTAG
);
833 (void) kill(getpid(), SIGKILL
);
837 ztest_random(uint64_t range
)
841 ASSERT3S(ztest_fd_rand
, >=, 0);
846 if (read(ztest_fd_rand
, &r
, sizeof (r
)) != sizeof (r
))
847 fatal(1, "short read from /dev/urandom");
854 ztest_record_enospc(const char *s
)
856 ztest_shared
->zs_enospc_count
++;
860 ztest_get_ashift(void)
862 if (ztest_opts
.zo_ashift
== 0)
863 return (SPA_MINBLOCKSHIFT
+ ztest_random(5));
864 return (ztest_opts
.zo_ashift
);
868 make_vdev_file(char *path
, char *aux
, char *pool
, size_t size
, uint64_t ashift
)
874 pathbuf
= umem_alloc(MAXPATHLEN
, UMEM_NOFAIL
);
877 ashift
= ztest_get_ashift();
883 vdev
= ztest_shared
->zs_vdev_aux
;
884 (void) snprintf(path
, MAXPATHLEN
,
885 ztest_aux_template
, ztest_opts
.zo_dir
,
886 pool
== NULL
? ztest_opts
.zo_pool
: pool
,
889 vdev
= ztest_shared
->zs_vdev_next_leaf
++;
890 (void) snprintf(path
, MAXPATHLEN
,
891 ztest_dev_template
, ztest_opts
.zo_dir
,
892 pool
== NULL
? ztest_opts
.zo_pool
: pool
, vdev
);
897 int fd
= open(path
, O_RDWR
| O_CREAT
| O_TRUNC
, 0666);
899 fatal(1, "can't open %s", path
);
900 if (ftruncate(fd
, size
) != 0)
901 fatal(1, "can't ftruncate %s", path
);
905 VERIFY(nvlist_alloc(&file
, NV_UNIQUE_NAME
, 0) == 0);
906 VERIFY(nvlist_add_string(file
, ZPOOL_CONFIG_TYPE
, VDEV_TYPE_FILE
) == 0);
907 VERIFY(nvlist_add_string(file
, ZPOOL_CONFIG_PATH
, path
) == 0);
908 VERIFY(nvlist_add_uint64(file
, ZPOOL_CONFIG_ASHIFT
, ashift
) == 0);
909 umem_free(pathbuf
, MAXPATHLEN
);
915 make_vdev_raidz(char *path
, char *aux
, char *pool
, size_t size
,
916 uint64_t ashift
, int r
)
918 nvlist_t
*raidz
, **child
;
922 return (make_vdev_file(path
, aux
, pool
, size
, ashift
));
923 child
= umem_alloc(r
* sizeof (nvlist_t
*), UMEM_NOFAIL
);
925 for (c
= 0; c
< r
; c
++)
926 child
[c
] = make_vdev_file(path
, aux
, pool
, size
, ashift
);
928 VERIFY(nvlist_alloc(&raidz
, NV_UNIQUE_NAME
, 0) == 0);
929 VERIFY(nvlist_add_string(raidz
, ZPOOL_CONFIG_TYPE
,
930 VDEV_TYPE_RAIDZ
) == 0);
931 VERIFY(nvlist_add_uint64(raidz
, ZPOOL_CONFIG_NPARITY
,
932 ztest_opts
.zo_raidz_parity
) == 0);
933 VERIFY(nvlist_add_nvlist_array(raidz
, ZPOOL_CONFIG_CHILDREN
,
936 for (c
= 0; c
< r
; c
++)
937 nvlist_free(child
[c
]);
939 umem_free(child
, r
* sizeof (nvlist_t
*));
945 make_vdev_mirror(char *path
, char *aux
, char *pool
, size_t size
,
946 uint64_t ashift
, int r
, int m
)
948 nvlist_t
*mirror
, **child
;
952 return (make_vdev_raidz(path
, aux
, pool
, size
, ashift
, r
));
954 child
= umem_alloc(m
* sizeof (nvlist_t
*), UMEM_NOFAIL
);
956 for (c
= 0; c
< m
; c
++)
957 child
[c
] = make_vdev_raidz(path
, aux
, pool
, size
, ashift
, r
);
959 VERIFY(nvlist_alloc(&mirror
, NV_UNIQUE_NAME
, 0) == 0);
960 VERIFY(nvlist_add_string(mirror
, ZPOOL_CONFIG_TYPE
,
961 VDEV_TYPE_MIRROR
) == 0);
962 VERIFY(nvlist_add_nvlist_array(mirror
, ZPOOL_CONFIG_CHILDREN
,
965 for (c
= 0; c
< m
; c
++)
966 nvlist_free(child
[c
]);
968 umem_free(child
, m
* sizeof (nvlist_t
*));
974 make_vdev_root(char *path
, char *aux
, char *pool
, size_t size
, uint64_t ashift
,
975 int log
, int r
, int m
, int t
)
977 nvlist_t
*root
, **child
;
982 child
= umem_alloc(t
* sizeof (nvlist_t
*), UMEM_NOFAIL
);
984 for (c
= 0; c
< t
; c
++) {
985 child
[c
] = make_vdev_mirror(path
, aux
, pool
, size
, ashift
,
987 VERIFY(nvlist_add_uint64(child
[c
], ZPOOL_CONFIG_IS_LOG
,
991 VERIFY(nvlist_alloc(&root
, NV_UNIQUE_NAME
, 0) == 0);
992 VERIFY(nvlist_add_string(root
, ZPOOL_CONFIG_TYPE
, VDEV_TYPE_ROOT
) == 0);
993 VERIFY(nvlist_add_nvlist_array(root
, aux
? aux
: ZPOOL_CONFIG_CHILDREN
,
996 for (c
= 0; c
< t
; c
++)
997 nvlist_free(child
[c
]);
999 umem_free(child
, t
* sizeof (nvlist_t
*));
1005 * Find a random spa version. Returns back a random spa version in the
1006 * range [initial_version, SPA_VERSION_FEATURES].
1009 ztest_random_spa_version(uint64_t initial_version
)
1011 uint64_t version
= initial_version
;
1013 if (version
<= SPA_VERSION_BEFORE_FEATURES
) {
1015 ztest_random(SPA_VERSION_BEFORE_FEATURES
- version
+ 1);
1018 if (version
> SPA_VERSION_BEFORE_FEATURES
)
1019 version
= SPA_VERSION_FEATURES
;
1021 ASSERT(SPA_VERSION_IS_SUPPORTED(version
));
1026 * Find the largest ashift used
1029 ztest_spa_get_ashift(void) {
1031 uint64_t ashift
= SPA_MINBLOCKSHIFT
;
1032 vdev_t
*rvd
= ztest_spa
->spa_root_vdev
;
1034 for (i
= 0; i
< rvd
->vdev_children
; i
++) {
1035 ashift
= MAX(ashift
, rvd
->vdev_child
[i
]->vdev_ashift
);
1041 ztest_random_blocksize(void)
1043 // Choose a block size >= the ashift.
1044 uint64_t block_shift
=
1045 ztest_random(SPA_MAXBLOCKSHIFT
- ztest_spa_get_ashift() + 1);
1046 return (1 << (SPA_MINBLOCKSHIFT
+ block_shift
));
1050 ztest_random_ibshift(void)
1052 return (DN_MIN_INDBLKSHIFT
+
1053 ztest_random(DN_MAX_INDBLKSHIFT
- DN_MIN_INDBLKSHIFT
+ 1));
1057 ztest_random_vdev_top(spa_t
*spa
, boolean_t log_ok
)
1060 vdev_t
*rvd
= spa
->spa_root_vdev
;
1063 ASSERT(spa_config_held(spa
, SCL_ALL
, RW_READER
) != 0);
1066 top
= ztest_random(rvd
->vdev_children
);
1067 tvd
= rvd
->vdev_child
[top
];
1068 } while (tvd
->vdev_ishole
|| (tvd
->vdev_islog
&& !log_ok
) ||
1069 tvd
->vdev_mg
== NULL
|| tvd
->vdev_mg
->mg_class
== NULL
);
1075 ztest_random_dsl_prop(zfs_prop_t prop
)
1080 value
= zfs_prop_random_value(prop
, ztest_random(-1ULL));
1081 } while (prop
== ZFS_PROP_CHECKSUM
&& value
== ZIO_CHECKSUM_OFF
);
1087 ztest_dsl_prop_set_uint64(char *osname
, zfs_prop_t prop
, uint64_t value
,
1090 const char *propname
= zfs_prop_to_name(prop
);
1091 const char *valname
;
1096 error
= dsl_prop_set_int(osname
, propname
,
1097 (inherit
? ZPROP_SRC_NONE
: ZPROP_SRC_LOCAL
), value
);
1099 if (error
== ENOSPC
) {
1100 ztest_record_enospc(FTAG
);
1105 setpoint
= umem_alloc(MAXPATHLEN
, UMEM_NOFAIL
);
1106 VERIFY0(dsl_prop_get_integer(osname
, propname
, &curval
, setpoint
));
1108 if (ztest_opts
.zo_verbose
>= 6) {
1109 VERIFY(zfs_prop_index_to_string(prop
, curval
, &valname
) == 0);
1110 (void) printf("%s %s = %s at '%s'\n",
1111 osname
, propname
, valname
, setpoint
);
1113 umem_free(setpoint
, MAXPATHLEN
);
1119 ztest_spa_prop_set_uint64(zpool_prop_t prop
, uint64_t value
)
1121 spa_t
*spa
= ztest_spa
;
1122 nvlist_t
*props
= NULL
;
1125 VERIFY(nvlist_alloc(&props
, NV_UNIQUE_NAME
, 0) == 0);
1126 VERIFY(nvlist_add_uint64(props
, zpool_prop_to_name(prop
), value
) == 0);
1128 error
= spa_prop_set(spa
, props
);
1132 if (error
== ENOSPC
) {
1133 ztest_record_enospc(FTAG
);
1142 ztest_rll_init(rll_t
*rll
)
1144 rll
->rll_writer
= NULL
;
1145 rll
->rll_readers
= 0;
1146 mutex_init(&rll
->rll_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
1147 cv_init(&rll
->rll_cv
, NULL
, CV_DEFAULT
, NULL
);
1151 ztest_rll_destroy(rll_t
*rll
)
1153 ASSERT(rll
->rll_writer
== NULL
);
1154 ASSERT(rll
->rll_readers
== 0);
1155 mutex_destroy(&rll
->rll_lock
);
1156 cv_destroy(&rll
->rll_cv
);
1160 ztest_rll_lock(rll_t
*rll
, rl_type_t type
)
1162 mutex_enter(&rll
->rll_lock
);
1164 if (type
== RL_READER
) {
1165 while (rll
->rll_writer
!= NULL
)
1166 (void) cv_wait(&rll
->rll_cv
, &rll
->rll_lock
);
1169 while (rll
->rll_writer
!= NULL
|| rll
->rll_readers
)
1170 (void) cv_wait(&rll
->rll_cv
, &rll
->rll_lock
);
1171 rll
->rll_writer
= curthread
;
1174 mutex_exit(&rll
->rll_lock
);
1178 ztest_rll_unlock(rll_t
*rll
)
1180 mutex_enter(&rll
->rll_lock
);
1182 if (rll
->rll_writer
) {
1183 ASSERT(rll
->rll_readers
== 0);
1184 rll
->rll_writer
= NULL
;
1186 ASSERT(rll
->rll_readers
!= 0);
1187 ASSERT(rll
->rll_writer
== NULL
);
1191 if (rll
->rll_writer
== NULL
&& rll
->rll_readers
== 0)
1192 cv_broadcast(&rll
->rll_cv
);
1194 mutex_exit(&rll
->rll_lock
);
1198 ztest_object_lock(ztest_ds_t
*zd
, uint64_t object
, rl_type_t type
)
1200 rll_t
*rll
= &zd
->zd_object_lock
[object
& (ZTEST_OBJECT_LOCKS
- 1)];
1202 ztest_rll_lock(rll
, type
);
1206 ztest_object_unlock(ztest_ds_t
*zd
, uint64_t object
)
1208 rll_t
*rll
= &zd
->zd_object_lock
[object
& (ZTEST_OBJECT_LOCKS
- 1)];
1210 ztest_rll_unlock(rll
);
1214 ztest_range_lock(ztest_ds_t
*zd
, uint64_t object
, uint64_t offset
,
1215 uint64_t size
, rl_type_t type
)
1217 uint64_t hash
= object
^ (offset
% (ZTEST_RANGE_LOCKS
+ 1));
1218 rll_t
*rll
= &zd
->zd_range_lock
[hash
& (ZTEST_RANGE_LOCKS
- 1)];
1221 rl
= umem_alloc(sizeof (*rl
), UMEM_NOFAIL
);
1222 rl
->rl_object
= object
;
1223 rl
->rl_offset
= offset
;
1227 ztest_rll_lock(rll
, type
);
1233 ztest_range_unlock(rl_t
*rl
)
1235 rll_t
*rll
= rl
->rl_lock
;
1237 ztest_rll_unlock(rll
);
1239 umem_free(rl
, sizeof (*rl
));
1243 ztest_zd_init(ztest_ds_t
*zd
, ztest_shared_ds_t
*szd
, objset_t
*os
)
1246 zd
->zd_zilog
= dmu_objset_zil(os
);
1247 zd
->zd_shared
= szd
;
1248 dmu_objset_name(os
, zd
->zd_name
);
1251 if (zd
->zd_shared
!= NULL
)
1252 zd
->zd_shared
->zd_seq
= 0;
1254 VERIFY(rwlock_init(&zd
->zd_zilog_lock
, USYNC_THREAD
, NULL
) == 0);
1255 mutex_init(&zd
->zd_dirobj_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
1257 for (l
= 0; l
< ZTEST_OBJECT_LOCKS
; l
++)
1258 ztest_rll_init(&zd
->zd_object_lock
[l
]);
1260 for (l
= 0; l
< ZTEST_RANGE_LOCKS
; l
++)
1261 ztest_rll_init(&zd
->zd_range_lock
[l
]);
1265 ztest_zd_fini(ztest_ds_t
*zd
)
1269 mutex_destroy(&zd
->zd_dirobj_lock
);
1270 (void) rwlock_destroy(&zd
->zd_zilog_lock
);
1272 for (l
= 0; l
< ZTEST_OBJECT_LOCKS
; l
++)
1273 ztest_rll_destroy(&zd
->zd_object_lock
[l
]);
1275 for (l
= 0; l
< ZTEST_RANGE_LOCKS
; l
++)
1276 ztest_rll_destroy(&zd
->zd_range_lock
[l
]);
1279 #define TXG_MIGHTWAIT (ztest_random(10) == 0 ? TXG_NOWAIT : TXG_WAIT)
1282 ztest_tx_assign(dmu_tx_t
*tx
, uint64_t txg_how
, const char *tag
)
1288 * Attempt to assign tx to some transaction group.
1290 error
= dmu_tx_assign(tx
, txg_how
);
1292 if (error
== ERESTART
) {
1293 ASSERT(txg_how
== TXG_NOWAIT
);
1296 ASSERT3U(error
, ==, ENOSPC
);
1297 ztest_record_enospc(tag
);
1302 txg
= dmu_tx_get_txg(tx
);
1308 ztest_pattern_set(void *buf
, uint64_t size
, uint64_t value
)
1311 uint64_t *ip_end
= (uint64_t *)((uintptr_t)buf
+ (uintptr_t)size
);
1319 ztest_pattern_match(void *buf
, uint64_t size
, uint64_t value
)
1322 uint64_t *ip_end
= (uint64_t *)((uintptr_t)buf
+ (uintptr_t)size
);
1326 diff
|= (value
- *ip
++);
1333 ztest_bt_generate(ztest_block_tag_t
*bt
, objset_t
*os
, uint64_t object
,
1334 uint64_t offset
, uint64_t gen
, uint64_t txg
, uint64_t crtxg
)
1336 bt
->bt_magic
= BT_MAGIC
;
1337 bt
->bt_objset
= dmu_objset_id(os
);
1338 bt
->bt_object
= object
;
1339 bt
->bt_offset
= offset
;
1342 bt
->bt_crtxg
= crtxg
;
1346 ztest_bt_verify(ztest_block_tag_t
*bt
, objset_t
*os
, uint64_t object
,
1347 uint64_t offset
, uint64_t gen
, uint64_t txg
, uint64_t crtxg
)
1349 ASSERT3U(bt
->bt_magic
, ==, BT_MAGIC
);
1350 ASSERT3U(bt
->bt_objset
, ==, dmu_objset_id(os
));
1351 ASSERT3U(bt
->bt_object
, ==, object
);
1352 ASSERT3U(bt
->bt_offset
, ==, offset
);
1353 ASSERT3U(bt
->bt_gen
, <=, gen
);
1354 ASSERT3U(bt
->bt_txg
, <=, txg
);
1355 ASSERT3U(bt
->bt_crtxg
, ==, crtxg
);
1358 static ztest_block_tag_t
*
1359 ztest_bt_bonus(dmu_buf_t
*db
)
1361 dmu_object_info_t doi
;
1362 ztest_block_tag_t
*bt
;
1364 dmu_object_info_from_db(db
, &doi
);
1365 ASSERT3U(doi
.doi_bonus_size
, <=, db
->db_size
);
1366 ASSERT3U(doi
.doi_bonus_size
, >=, sizeof (*bt
));
1367 bt
= (void *)((char *)db
->db_data
+ doi
.doi_bonus_size
- sizeof (*bt
));
1376 #define lrz_type lr_mode
1377 #define lrz_blocksize lr_uid
1378 #define lrz_ibshift lr_gid
1379 #define lrz_bonustype lr_rdev
1380 #define lrz_bonuslen lr_crtime[1]
1383 ztest_log_create(ztest_ds_t
*zd
, dmu_tx_t
*tx
, lr_create_t
*lr
)
1385 char *name
= (void *)(lr
+ 1); /* name follows lr */
1386 size_t namesize
= strlen(name
) + 1;
1389 if (zil_replaying(zd
->zd_zilog
, tx
))
1392 itx
= zil_itx_create(TX_CREATE
, sizeof (*lr
) + namesize
);
1393 bcopy(&lr
->lr_common
+ 1, &itx
->itx_lr
+ 1,
1394 sizeof (*lr
) + namesize
- sizeof (lr_t
));
1396 zil_itx_assign(zd
->zd_zilog
, itx
, tx
);
1400 ztest_log_remove(ztest_ds_t
*zd
, dmu_tx_t
*tx
, lr_remove_t
*lr
, uint64_t object
)
1402 char *name
= (void *)(lr
+ 1); /* name follows lr */
1403 size_t namesize
= strlen(name
) + 1;
1406 if (zil_replaying(zd
->zd_zilog
, tx
))
1409 itx
= zil_itx_create(TX_REMOVE
, sizeof (*lr
) + namesize
);
1410 bcopy(&lr
->lr_common
+ 1, &itx
->itx_lr
+ 1,
1411 sizeof (*lr
) + namesize
- sizeof (lr_t
));
1413 itx
->itx_oid
= object
;
1414 zil_itx_assign(zd
->zd_zilog
, itx
, tx
);
1418 ztest_log_write(ztest_ds_t
*zd
, dmu_tx_t
*tx
, lr_write_t
*lr
)
1421 itx_wr_state_t write_state
= ztest_random(WR_NUM_STATES
);
1423 if (zil_replaying(zd
->zd_zilog
, tx
))
1426 if (lr
->lr_length
> ZIL_MAX_LOG_DATA
)
1427 write_state
= WR_INDIRECT
;
1429 itx
= zil_itx_create(TX_WRITE
,
1430 sizeof (*lr
) + (write_state
== WR_COPIED
? lr
->lr_length
: 0));
1432 if (write_state
== WR_COPIED
&&
1433 dmu_read(zd
->zd_os
, lr
->lr_foid
, lr
->lr_offset
, lr
->lr_length
,
1434 ((lr_write_t
*)&itx
->itx_lr
) + 1, DMU_READ_NO_PREFETCH
) != 0) {
1435 zil_itx_destroy(itx
);
1436 itx
= zil_itx_create(TX_WRITE
, sizeof (*lr
));
1437 write_state
= WR_NEED_COPY
;
1439 itx
->itx_private
= zd
;
1440 itx
->itx_wr_state
= write_state
;
1441 itx
->itx_sync
= (ztest_random(8) == 0);
1442 itx
->itx_sod
+= (write_state
== WR_NEED_COPY
? lr
->lr_length
: 0);
1444 bcopy(&lr
->lr_common
+ 1, &itx
->itx_lr
+ 1,
1445 sizeof (*lr
) - sizeof (lr_t
));
1447 zil_itx_assign(zd
->zd_zilog
, itx
, tx
);
1451 ztest_log_truncate(ztest_ds_t
*zd
, dmu_tx_t
*tx
, lr_truncate_t
*lr
)
1455 if (zil_replaying(zd
->zd_zilog
, tx
))
1458 itx
= zil_itx_create(TX_TRUNCATE
, sizeof (*lr
));
1459 bcopy(&lr
->lr_common
+ 1, &itx
->itx_lr
+ 1,
1460 sizeof (*lr
) - sizeof (lr_t
));
1462 itx
->itx_sync
= B_FALSE
;
1463 zil_itx_assign(zd
->zd_zilog
, itx
, tx
);
1467 ztest_log_setattr(ztest_ds_t
*zd
, dmu_tx_t
*tx
, lr_setattr_t
*lr
)
1471 if (zil_replaying(zd
->zd_zilog
, tx
))
1474 itx
= zil_itx_create(TX_SETATTR
, sizeof (*lr
));
1475 bcopy(&lr
->lr_common
+ 1, &itx
->itx_lr
+ 1,
1476 sizeof (*lr
) - sizeof (lr_t
));
1478 itx
->itx_sync
= B_FALSE
;
1479 zil_itx_assign(zd
->zd_zilog
, itx
, tx
);
1486 ztest_replay_create(ztest_ds_t
*zd
, lr_create_t
*lr
, boolean_t byteswap
)
1488 char *name
= (void *)(lr
+ 1); /* name follows lr */
1489 objset_t
*os
= zd
->zd_os
;
1490 ztest_block_tag_t
*bbt
;
1497 byteswap_uint64_array(lr
, sizeof (*lr
));
1499 ASSERT(lr
->lr_doid
== ZTEST_DIROBJ
);
1500 ASSERT(name
[0] != '\0');
1502 tx
= dmu_tx_create(os
);
1504 dmu_tx_hold_zap(tx
, lr
->lr_doid
, B_TRUE
, name
);
1506 if (lr
->lrz_type
== DMU_OT_ZAP_OTHER
) {
1507 dmu_tx_hold_zap(tx
, DMU_NEW_OBJECT
, B_TRUE
, NULL
);
1509 dmu_tx_hold_bonus(tx
, DMU_NEW_OBJECT
);
1512 txg
= ztest_tx_assign(tx
, TXG_WAIT
, FTAG
);
1516 ASSERT(dmu_objset_zil(os
)->zl_replay
== !!lr
->lr_foid
);
1518 if (lr
->lrz_type
== DMU_OT_ZAP_OTHER
) {
1519 if (lr
->lr_foid
== 0) {
1520 lr
->lr_foid
= zap_create(os
,
1521 lr
->lrz_type
, lr
->lrz_bonustype
,
1522 lr
->lrz_bonuslen
, tx
);
1524 error
= zap_create_claim(os
, lr
->lr_foid
,
1525 lr
->lrz_type
, lr
->lrz_bonustype
,
1526 lr
->lrz_bonuslen
, tx
);
1529 if (lr
->lr_foid
== 0) {
1530 lr
->lr_foid
= dmu_object_alloc(os
,
1531 lr
->lrz_type
, 0, lr
->lrz_bonustype
,
1532 lr
->lrz_bonuslen
, tx
);
1534 error
= dmu_object_claim(os
, lr
->lr_foid
,
1535 lr
->lrz_type
, 0, lr
->lrz_bonustype
,
1536 lr
->lrz_bonuslen
, tx
);
1541 ASSERT3U(error
, ==, EEXIST
);
1542 ASSERT(zd
->zd_zilog
->zl_replay
);
1547 ASSERT(lr
->lr_foid
!= 0);
1549 if (lr
->lrz_type
!= DMU_OT_ZAP_OTHER
)
1550 VERIFY3U(0, ==, dmu_object_set_blocksize(os
, lr
->lr_foid
,
1551 lr
->lrz_blocksize
, lr
->lrz_ibshift
, tx
));
1553 VERIFY3U(0, ==, dmu_bonus_hold(os
, lr
->lr_foid
, FTAG
, &db
));
1554 bbt
= ztest_bt_bonus(db
);
1555 dmu_buf_will_dirty(db
, tx
);
1556 ztest_bt_generate(bbt
, os
, lr
->lr_foid
, -1ULL, lr
->lr_gen
, txg
, txg
);
1557 dmu_buf_rele(db
, FTAG
);
1559 VERIFY3U(0, ==, zap_add(os
, lr
->lr_doid
, name
, sizeof (uint64_t), 1,
1562 (void) ztest_log_create(zd
, tx
, lr
);
1570 ztest_replay_remove(ztest_ds_t
*zd
, lr_remove_t
*lr
, boolean_t byteswap
)
1572 char *name
= (void *)(lr
+ 1); /* name follows lr */
1573 objset_t
*os
= zd
->zd_os
;
1574 dmu_object_info_t doi
;
1576 uint64_t object
, txg
;
1579 byteswap_uint64_array(lr
, sizeof (*lr
));
1581 ASSERT(lr
->lr_doid
== ZTEST_DIROBJ
);
1582 ASSERT(name
[0] != '\0');
1585 zap_lookup(os
, lr
->lr_doid
, name
, sizeof (object
), 1, &object
));
1586 ASSERT(object
!= 0);
1588 ztest_object_lock(zd
, object
, RL_WRITER
);
1590 VERIFY3U(0, ==, dmu_object_info(os
, object
, &doi
));
1592 tx
= dmu_tx_create(os
);
1594 dmu_tx_hold_zap(tx
, lr
->lr_doid
, B_FALSE
, name
);
1595 dmu_tx_hold_free(tx
, object
, 0, DMU_OBJECT_END
);
1597 txg
= ztest_tx_assign(tx
, TXG_WAIT
, FTAG
);
1599 ztest_object_unlock(zd
, object
);
1603 if (doi
.doi_type
== DMU_OT_ZAP_OTHER
) {
1604 VERIFY3U(0, ==, zap_destroy(os
, object
, tx
));
1606 VERIFY3U(0, ==, dmu_object_free(os
, object
, tx
));
1609 VERIFY3U(0, ==, zap_remove(os
, lr
->lr_doid
, name
, tx
));
1611 (void) ztest_log_remove(zd
, tx
, lr
, object
);
1615 ztest_object_unlock(zd
, object
);
1621 ztest_replay_write(ztest_ds_t
*zd
, lr_write_t
*lr
, boolean_t byteswap
)
1623 objset_t
*os
= zd
->zd_os
;
1624 void *data
= lr
+ 1; /* data follows lr */
1625 uint64_t offset
, length
;
1626 ztest_block_tag_t
*bt
= data
;
1627 ztest_block_tag_t
*bbt
;
1628 uint64_t gen
, txg
, lrtxg
, crtxg
;
1629 dmu_object_info_t doi
;
1632 arc_buf_t
*abuf
= NULL
;
1636 byteswap_uint64_array(lr
, sizeof (*lr
));
1638 offset
= lr
->lr_offset
;
1639 length
= lr
->lr_length
;
1641 /* If it's a dmu_sync() block, write the whole block */
1642 if (lr
->lr_common
.lrc_reclen
== sizeof (lr_write_t
)) {
1643 uint64_t blocksize
= BP_GET_LSIZE(&lr
->lr_blkptr
);
1644 if (length
< blocksize
) {
1645 offset
-= offset
% blocksize
;
1650 if (bt
->bt_magic
== BSWAP_64(BT_MAGIC
))
1651 byteswap_uint64_array(bt
, sizeof (*bt
));
1653 if (bt
->bt_magic
!= BT_MAGIC
)
1656 ztest_object_lock(zd
, lr
->lr_foid
, RL_READER
);
1657 rl
= ztest_range_lock(zd
, lr
->lr_foid
, offset
, length
, RL_WRITER
);
1659 VERIFY3U(0, ==, dmu_bonus_hold(os
, lr
->lr_foid
, FTAG
, &db
));
1661 dmu_object_info_from_db(db
, &doi
);
1663 bbt
= ztest_bt_bonus(db
);
1664 ASSERT3U(bbt
->bt_magic
, ==, BT_MAGIC
);
1666 crtxg
= bbt
->bt_crtxg
;
1667 lrtxg
= lr
->lr_common
.lrc_txg
;
1669 tx
= dmu_tx_create(os
);
1671 dmu_tx_hold_write(tx
, lr
->lr_foid
, offset
, length
);
1673 if (ztest_random(8) == 0 && length
== doi
.doi_data_block_size
&&
1674 P2PHASE(offset
, length
) == 0)
1675 abuf
= dmu_request_arcbuf(db
, length
);
1677 txg
= ztest_tx_assign(tx
, TXG_WAIT
, FTAG
);
1680 dmu_return_arcbuf(abuf
);
1681 dmu_buf_rele(db
, FTAG
);
1682 ztest_range_unlock(rl
);
1683 ztest_object_unlock(zd
, lr
->lr_foid
);
1689 * Usually, verify the old data before writing new data --
1690 * but not always, because we also want to verify correct
1691 * behavior when the data was not recently read into cache.
1693 ASSERT(offset
% doi
.doi_data_block_size
== 0);
1694 if (ztest_random(4) != 0) {
1695 int prefetch
= ztest_random(2) ?
1696 DMU_READ_PREFETCH
: DMU_READ_NO_PREFETCH
;
1697 ztest_block_tag_t rbt
;
1699 VERIFY(dmu_read(os
, lr
->lr_foid
, offset
,
1700 sizeof (rbt
), &rbt
, prefetch
) == 0);
1701 if (rbt
.bt_magic
== BT_MAGIC
) {
1702 ztest_bt_verify(&rbt
, os
, lr
->lr_foid
,
1703 offset
, gen
, txg
, crtxg
);
1708 * Writes can appear to be newer than the bonus buffer because
1709 * the ztest_get_data() callback does a dmu_read() of the
1710 * open-context data, which may be different than the data
1711 * as it was when the write was generated.
1713 if (zd
->zd_zilog
->zl_replay
) {
1714 ztest_bt_verify(bt
, os
, lr
->lr_foid
, offset
,
1715 MAX(gen
, bt
->bt_gen
), MAX(txg
, lrtxg
),
1720 * Set the bt's gen/txg to the bonus buffer's gen/txg
1721 * so that all of the usual ASSERTs will work.
1723 ztest_bt_generate(bt
, os
, lr
->lr_foid
, offset
, gen
, txg
, crtxg
);
1727 dmu_write(os
, lr
->lr_foid
, offset
, length
, data
, tx
);
1729 bcopy(data
, abuf
->b_data
, length
);
1730 dmu_assign_arcbuf(db
, offset
, abuf
, tx
);
1733 (void) ztest_log_write(zd
, tx
, lr
);
1735 dmu_buf_rele(db
, FTAG
);
1739 ztest_range_unlock(rl
);
1740 ztest_object_unlock(zd
, lr
->lr_foid
);
1746 ztest_replay_truncate(ztest_ds_t
*zd
, lr_truncate_t
*lr
, boolean_t byteswap
)
1748 objset_t
*os
= zd
->zd_os
;
1754 byteswap_uint64_array(lr
, sizeof (*lr
));
1756 ztest_object_lock(zd
, lr
->lr_foid
, RL_READER
);
1757 rl
= ztest_range_lock(zd
, lr
->lr_foid
, lr
->lr_offset
, lr
->lr_length
,
1760 tx
= dmu_tx_create(os
);
1762 dmu_tx_hold_free(tx
, lr
->lr_foid
, lr
->lr_offset
, lr
->lr_length
);
1764 txg
= ztest_tx_assign(tx
, TXG_WAIT
, FTAG
);
1766 ztest_range_unlock(rl
);
1767 ztest_object_unlock(zd
, lr
->lr_foid
);
1771 VERIFY(dmu_free_range(os
, lr
->lr_foid
, lr
->lr_offset
,
1772 lr
->lr_length
, tx
) == 0);
1774 (void) ztest_log_truncate(zd
, tx
, lr
);
1778 ztest_range_unlock(rl
);
1779 ztest_object_unlock(zd
, lr
->lr_foid
);
1785 ztest_replay_setattr(ztest_ds_t
*zd
, lr_setattr_t
*lr
, boolean_t byteswap
)
1787 objset_t
*os
= zd
->zd_os
;
1790 ztest_block_tag_t
*bbt
;
1791 uint64_t txg
, lrtxg
, crtxg
;
1794 byteswap_uint64_array(lr
, sizeof (*lr
));
1796 ztest_object_lock(zd
, lr
->lr_foid
, RL_WRITER
);
1798 VERIFY3U(0, ==, dmu_bonus_hold(os
, lr
->lr_foid
, FTAG
, &db
));
1800 tx
= dmu_tx_create(os
);
1801 dmu_tx_hold_bonus(tx
, lr
->lr_foid
);
1803 txg
= ztest_tx_assign(tx
, TXG_WAIT
, FTAG
);
1805 dmu_buf_rele(db
, FTAG
);
1806 ztest_object_unlock(zd
, lr
->lr_foid
);
1810 bbt
= ztest_bt_bonus(db
);
1811 ASSERT3U(bbt
->bt_magic
, ==, BT_MAGIC
);
1812 crtxg
= bbt
->bt_crtxg
;
1813 lrtxg
= lr
->lr_common
.lrc_txg
;
1815 if (zd
->zd_zilog
->zl_replay
) {
1816 ASSERT(lr
->lr_size
!= 0);
1817 ASSERT(lr
->lr_mode
!= 0);
1821 * Randomly change the size and increment the generation.
1823 lr
->lr_size
= (ztest_random(db
->db_size
/ sizeof (*bbt
)) + 1) *
1825 lr
->lr_mode
= bbt
->bt_gen
+ 1;
1830 * Verify that the current bonus buffer is not newer than our txg.
1832 ztest_bt_verify(bbt
, os
, lr
->lr_foid
, -1ULL, lr
->lr_mode
,
1833 MAX(txg
, lrtxg
), crtxg
);
1835 dmu_buf_will_dirty(db
, tx
);
1837 ASSERT3U(lr
->lr_size
, >=, sizeof (*bbt
));
1838 ASSERT3U(lr
->lr_size
, <=, db
->db_size
);
1839 VERIFY0(dmu_set_bonus(db
, lr
->lr_size
, tx
));
1840 bbt
= ztest_bt_bonus(db
);
1842 ztest_bt_generate(bbt
, os
, lr
->lr_foid
, -1ULL, lr
->lr_mode
, txg
, crtxg
);
1844 dmu_buf_rele(db
, FTAG
);
1846 (void) ztest_log_setattr(zd
, tx
, lr
);
1850 ztest_object_unlock(zd
, lr
->lr_foid
);
1855 zil_replay_func_t ztest_replay_vector
[TX_MAX_TYPE
] = {
1856 NULL
, /* 0 no such transaction type */
1857 (zil_replay_func_t
)ztest_replay_create
, /* TX_CREATE */
1858 NULL
, /* TX_MKDIR */
1859 NULL
, /* TX_MKXATTR */
1860 NULL
, /* TX_SYMLINK */
1861 (zil_replay_func_t
)ztest_replay_remove
, /* TX_REMOVE */
1862 NULL
, /* TX_RMDIR */
1864 NULL
, /* TX_RENAME */
1865 (zil_replay_func_t
)ztest_replay_write
, /* TX_WRITE */
1866 (zil_replay_func_t
)ztest_replay_truncate
, /* TX_TRUNCATE */
1867 (zil_replay_func_t
)ztest_replay_setattr
, /* TX_SETATTR */
1869 NULL
, /* TX_CREATE_ACL */
1870 NULL
, /* TX_CREATE_ATTR */
1871 NULL
, /* TX_CREATE_ACL_ATTR */
1872 NULL
, /* TX_MKDIR_ACL */
1873 NULL
, /* TX_MKDIR_ATTR */
1874 NULL
, /* TX_MKDIR_ACL_ATTR */
1875 NULL
, /* TX_WRITE2 */
1879 * ZIL get_data callbacks
1883 ztest_get_done(zgd_t
*zgd
, int error
)
1885 ztest_ds_t
*zd
= zgd
->zgd_private
;
1886 uint64_t object
= zgd
->zgd_rl
->rl_object
;
1889 dmu_buf_rele(zgd
->zgd_db
, zgd
);
1891 ztest_range_unlock(zgd
->zgd_rl
);
1892 ztest_object_unlock(zd
, object
);
1894 if (error
== 0 && zgd
->zgd_bp
)
1895 zil_add_block(zgd
->zgd_zilog
, zgd
->zgd_bp
);
1897 umem_free(zgd
, sizeof (*zgd
));
1901 ztest_get_data(void *arg
, lr_write_t
*lr
, char *buf
, zio_t
*zio
)
1903 ztest_ds_t
*zd
= arg
;
1904 objset_t
*os
= zd
->zd_os
;
1905 uint64_t object
= lr
->lr_foid
;
1906 uint64_t offset
= lr
->lr_offset
;
1907 uint64_t size
= lr
->lr_length
;
1908 blkptr_t
*bp
= &lr
->lr_blkptr
;
1909 uint64_t txg
= lr
->lr_common
.lrc_txg
;
1911 dmu_object_info_t doi
;
1916 ztest_object_lock(zd
, object
, RL_READER
);
1917 error
= dmu_bonus_hold(os
, object
, FTAG
, &db
);
1919 ztest_object_unlock(zd
, object
);
1923 crtxg
= ztest_bt_bonus(db
)->bt_crtxg
;
1925 if (crtxg
== 0 || crtxg
> txg
) {
1926 dmu_buf_rele(db
, FTAG
);
1927 ztest_object_unlock(zd
, object
);
1931 dmu_object_info_from_db(db
, &doi
);
1932 dmu_buf_rele(db
, FTAG
);
1935 zgd
= umem_zalloc(sizeof (*zgd
), UMEM_NOFAIL
);
1936 zgd
->zgd_zilog
= zd
->zd_zilog
;
1937 zgd
->zgd_private
= zd
;
1939 if (buf
!= NULL
) { /* immediate write */
1940 zgd
->zgd_rl
= ztest_range_lock(zd
, object
, offset
, size
,
1943 error
= dmu_read(os
, object
, offset
, size
, buf
,
1944 DMU_READ_NO_PREFETCH
);
1947 size
= doi
.doi_data_block_size
;
1949 offset
= P2ALIGN(offset
, size
);
1951 ASSERT(offset
< size
);
1955 zgd
->zgd_rl
= ztest_range_lock(zd
, object
, offset
, size
,
1958 error
= dmu_buf_hold(os
, object
, offset
, zgd
, &db
,
1959 DMU_READ_NO_PREFETCH
);
1962 blkptr_t
*obp
= dmu_buf_get_blkptr(db
);
1964 ASSERT(BP_IS_HOLE(bp
));
1971 ASSERT(db
->db_offset
== offset
);
1972 ASSERT(db
->db_size
== size
);
1974 error
= dmu_sync(zio
, lr
->lr_common
.lrc_txg
,
1975 ztest_get_done
, zgd
);
1982 ztest_get_done(zgd
, error
);
1988 ztest_lr_alloc(size_t lrsize
, char *name
)
1991 size_t namesize
= name
? strlen(name
) + 1 : 0;
1993 lr
= umem_zalloc(lrsize
+ namesize
, UMEM_NOFAIL
);
1996 bcopy(name
, lr
+ lrsize
, namesize
);
2002 ztest_lr_free(void *lr
, size_t lrsize
, char *name
)
2004 size_t namesize
= name
? strlen(name
) + 1 : 0;
2006 umem_free(lr
, lrsize
+ namesize
);
2010 * Lookup a bunch of objects. Returns the number of objects not found.
2013 ztest_lookup(ztest_ds_t
*zd
, ztest_od_t
*od
, int count
)
2019 ASSERT(mutex_held(&zd
->zd_dirobj_lock
));
2021 for (i
= 0; i
< count
; i
++, od
++) {
2023 error
= zap_lookup(zd
->zd_os
, od
->od_dir
, od
->od_name
,
2024 sizeof (uint64_t), 1, &od
->od_object
);
2026 ASSERT(error
== ENOENT
);
2027 ASSERT(od
->od_object
== 0);
2031 ztest_block_tag_t
*bbt
;
2032 dmu_object_info_t doi
;
2034 ASSERT(od
->od_object
!= 0);
2035 ASSERT(missing
== 0); /* there should be no gaps */
2037 ztest_object_lock(zd
, od
->od_object
, RL_READER
);
2038 VERIFY3U(0, ==, dmu_bonus_hold(zd
->zd_os
,
2039 od
->od_object
, FTAG
, &db
));
2040 dmu_object_info_from_db(db
, &doi
);
2041 bbt
= ztest_bt_bonus(db
);
2042 ASSERT3U(bbt
->bt_magic
, ==, BT_MAGIC
);
2043 od
->od_type
= doi
.doi_type
;
2044 od
->od_blocksize
= doi
.doi_data_block_size
;
2045 od
->od_gen
= bbt
->bt_gen
;
2046 dmu_buf_rele(db
, FTAG
);
2047 ztest_object_unlock(zd
, od
->od_object
);
2055 ztest_create(ztest_ds_t
*zd
, ztest_od_t
*od
, int count
)
2060 ASSERT(mutex_held(&zd
->zd_dirobj_lock
));
2062 for (i
= 0; i
< count
; i
++, od
++) {
2069 lr_create_t
*lr
= ztest_lr_alloc(sizeof (*lr
), od
->od_name
);
2071 lr
->lr_doid
= od
->od_dir
;
2072 lr
->lr_foid
= 0; /* 0 to allocate, > 0 to claim */
2073 lr
->lrz_type
= od
->od_crtype
;
2074 lr
->lrz_blocksize
= od
->od_crblocksize
;
2075 lr
->lrz_ibshift
= ztest_random_ibshift();
2076 lr
->lrz_bonustype
= DMU_OT_UINT64_OTHER
;
2077 lr
->lrz_bonuslen
= dmu_bonus_max();
2078 lr
->lr_gen
= od
->od_crgen
;
2079 lr
->lr_crtime
[0] = time(NULL
);
2081 if (ztest_replay_create(zd
, lr
, B_FALSE
) != 0) {
2082 ASSERT(missing
== 0);
2086 od
->od_object
= lr
->lr_foid
;
2087 od
->od_type
= od
->od_crtype
;
2088 od
->od_blocksize
= od
->od_crblocksize
;
2089 od
->od_gen
= od
->od_crgen
;
2090 ASSERT(od
->od_object
!= 0);
2093 ztest_lr_free(lr
, sizeof (*lr
), od
->od_name
);
2100 ztest_remove(ztest_ds_t
*zd
, ztest_od_t
*od
, int count
)
2106 ASSERT(mutex_held(&zd
->zd_dirobj_lock
));
2110 for (i
= count
- 1; i
>= 0; i
--, od
--) {
2117 * No object was found.
2119 if (od
->od_object
== 0)
2122 lr_remove_t
*lr
= ztest_lr_alloc(sizeof (*lr
), od
->od_name
);
2124 lr
->lr_doid
= od
->od_dir
;
2126 if ((error
= ztest_replay_remove(zd
, lr
, B_FALSE
)) != 0) {
2127 ASSERT3U(error
, ==, ENOSPC
);
2132 ztest_lr_free(lr
, sizeof (*lr
), od
->od_name
);
2139 ztest_write(ztest_ds_t
*zd
, uint64_t object
, uint64_t offset
, uint64_t size
,
2145 lr
= ztest_lr_alloc(sizeof (*lr
) + size
, NULL
);
2147 lr
->lr_foid
= object
;
2148 lr
->lr_offset
= offset
;
2149 lr
->lr_length
= size
;
2151 BP_ZERO(&lr
->lr_blkptr
);
2153 bcopy(data
, lr
+ 1, size
);
2155 error
= ztest_replay_write(zd
, lr
, B_FALSE
);
2157 ztest_lr_free(lr
, sizeof (*lr
) + size
, NULL
);
2163 ztest_truncate(ztest_ds_t
*zd
, uint64_t object
, uint64_t offset
, uint64_t size
)
2168 lr
= ztest_lr_alloc(sizeof (*lr
), NULL
);
2170 lr
->lr_foid
= object
;
2171 lr
->lr_offset
= offset
;
2172 lr
->lr_length
= size
;
2174 error
= ztest_replay_truncate(zd
, lr
, B_FALSE
);
2176 ztest_lr_free(lr
, sizeof (*lr
), NULL
);
2182 ztest_setattr(ztest_ds_t
*zd
, uint64_t object
)
2187 lr
= ztest_lr_alloc(sizeof (*lr
), NULL
);
2189 lr
->lr_foid
= object
;
2193 error
= ztest_replay_setattr(zd
, lr
, B_FALSE
);
2195 ztest_lr_free(lr
, sizeof (*lr
), NULL
);
2201 ztest_prealloc(ztest_ds_t
*zd
, uint64_t object
, uint64_t offset
, uint64_t size
)
2203 objset_t
*os
= zd
->zd_os
;
2208 txg_wait_synced(dmu_objset_pool(os
), 0);
2210 ztest_object_lock(zd
, object
, RL_READER
);
2211 rl
= ztest_range_lock(zd
, object
, offset
, size
, RL_WRITER
);
2213 tx
= dmu_tx_create(os
);
2215 dmu_tx_hold_write(tx
, object
, offset
, size
);
2217 txg
= ztest_tx_assign(tx
, TXG_WAIT
, FTAG
);
2220 dmu_prealloc(os
, object
, offset
, size
, tx
);
2222 txg_wait_synced(dmu_objset_pool(os
), txg
);
2224 (void) dmu_free_long_range(os
, object
, offset
, size
);
2227 ztest_range_unlock(rl
);
2228 ztest_object_unlock(zd
, object
);
2232 ztest_io(ztest_ds_t
*zd
, uint64_t object
, uint64_t offset
)
2235 ztest_block_tag_t wbt
;
2236 dmu_object_info_t doi
;
2237 enum ztest_io_type io_type
;
2241 VERIFY(dmu_object_info(zd
->zd_os
, object
, &doi
) == 0);
2242 blocksize
= doi
.doi_data_block_size
;
2243 data
= umem_alloc(blocksize
, UMEM_NOFAIL
);
2246 * Pick an i/o type at random, biased toward writing block tags.
2248 io_type
= ztest_random(ZTEST_IO_TYPES
);
2249 if (ztest_random(2) == 0)
2250 io_type
= ZTEST_IO_WRITE_TAG
;
2252 (void) rw_rdlock(&zd
->zd_zilog_lock
);
2256 case ZTEST_IO_WRITE_TAG
:
2257 ztest_bt_generate(&wbt
, zd
->zd_os
, object
, offset
, 0, 0, 0);
2258 (void) ztest_write(zd
, object
, offset
, sizeof (wbt
), &wbt
);
2261 case ZTEST_IO_WRITE_PATTERN
:
2262 (void) memset(data
, 'a' + (object
+ offset
) % 5, blocksize
);
2263 if (ztest_random(2) == 0) {
2265 * Induce fletcher2 collisions to ensure that
2266 * zio_ddt_collision() detects and resolves them
2267 * when using fletcher2-verify for deduplication.
2269 ((uint64_t *)data
)[0] ^= 1ULL << 63;
2270 ((uint64_t *)data
)[4] ^= 1ULL << 63;
2272 (void) ztest_write(zd
, object
, offset
, blocksize
, data
);
2275 case ZTEST_IO_WRITE_ZEROES
:
2276 bzero(data
, blocksize
);
2277 (void) ztest_write(zd
, object
, offset
, blocksize
, data
);
2280 case ZTEST_IO_TRUNCATE
:
2281 (void) ztest_truncate(zd
, object
, offset
, blocksize
);
2284 case ZTEST_IO_SETATTR
:
2285 (void) ztest_setattr(zd
, object
);
2290 case ZTEST_IO_REWRITE
:
2291 (void) rw_rdlock(&ztest_name_lock
);
2292 err
= ztest_dsl_prop_set_uint64(zd
->zd_name
,
2293 ZFS_PROP_CHECKSUM
, spa_dedup_checksum(ztest_spa
),
2295 VERIFY(err
== 0 || err
== ENOSPC
);
2296 err
= ztest_dsl_prop_set_uint64(zd
->zd_name
,
2297 ZFS_PROP_COMPRESSION
,
2298 ztest_random_dsl_prop(ZFS_PROP_COMPRESSION
),
2300 VERIFY(err
== 0 || err
== ENOSPC
);
2301 (void) rw_unlock(&ztest_name_lock
);
2303 VERIFY0(dmu_read(zd
->zd_os
, object
, offset
, blocksize
, data
,
2304 DMU_READ_NO_PREFETCH
));
2306 (void) ztest_write(zd
, object
, offset
, blocksize
, data
);
2310 (void) rw_unlock(&zd
->zd_zilog_lock
);
2312 umem_free(data
, blocksize
);
2316 * Initialize an object description template.
2319 ztest_od_init(ztest_od_t
*od
, uint64_t id
, char *tag
, uint64_t index
,
2320 dmu_object_type_t type
, uint64_t blocksize
, uint64_t gen
)
2322 od
->od_dir
= ZTEST_DIROBJ
;
2325 od
->od_crtype
= type
;
2326 od
->od_crblocksize
= blocksize
? blocksize
: ztest_random_blocksize();
2329 od
->od_type
= DMU_OT_NONE
;
2330 od
->od_blocksize
= 0;
2333 (void) snprintf(od
->od_name
, sizeof (od
->od_name
), "%s(%lld)[%llu]",
2334 tag
, (longlong_t
)id
, (u_longlong_t
)index
);
2338 * Lookup or create the objects for a test using the od template.
2339 * If the objects do not all exist, or if 'remove' is specified,
2340 * remove any existing objects and create new ones. Otherwise,
2341 * use the existing objects.
2344 ztest_object_init(ztest_ds_t
*zd
, ztest_od_t
*od
, size_t size
, boolean_t remove
)
2346 int count
= size
/ sizeof (*od
);
2349 mutex_enter(&zd
->zd_dirobj_lock
);
2350 if ((ztest_lookup(zd
, od
, count
) != 0 || remove
) &&
2351 (ztest_remove(zd
, od
, count
) != 0 ||
2352 ztest_create(zd
, od
, count
) != 0))
2355 mutex_exit(&zd
->zd_dirobj_lock
);
2362 ztest_zil_commit(ztest_ds_t
*zd
, uint64_t id
)
2364 zilog_t
*zilog
= zd
->zd_zilog
;
2366 (void) rw_rdlock(&zd
->zd_zilog_lock
);
2368 zil_commit(zilog
, ztest_random(ZTEST_OBJECTS
));
2371 * Remember the committed values in zd, which is in parent/child
2372 * shared memory. If we die, the next iteration of ztest_run()
2373 * will verify that the log really does contain this record.
2375 mutex_enter(&zilog
->zl_lock
);
2376 ASSERT(zd
->zd_shared
!= NULL
);
2377 ASSERT3U(zd
->zd_shared
->zd_seq
, <=, zilog
->zl_commit_lr_seq
);
2378 zd
->zd_shared
->zd_seq
= zilog
->zl_commit_lr_seq
;
2379 mutex_exit(&zilog
->zl_lock
);
2381 (void) rw_unlock(&zd
->zd_zilog_lock
);
2385 * This function is designed to simulate the operations that occur during a
2386 * mount/unmount operation. We hold the dataset across these operations in an
2387 * attempt to expose any implicit assumptions about ZIL management.
2391 ztest_zil_remount(ztest_ds_t
*zd
, uint64_t id
)
2393 objset_t
*os
= zd
->zd_os
;
2396 * We grab the zd_dirobj_lock to ensure that no other thread is
2397 * updating the zil (i.e. adding in-memory log records) and the
2398 * zd_zilog_lock to block any I/O.
2400 mutex_enter(&zd
->zd_dirobj_lock
);
2401 (void) rw_wrlock(&zd
->zd_zilog_lock
);
2403 /* zfs_sb_teardown() */
2404 zil_close(zd
->zd_zilog
);
2406 /* zfsvfs_setup() */
2407 VERIFY(zil_open(os
, ztest_get_data
) == zd
->zd_zilog
);
2408 zil_replay(os
, zd
, ztest_replay_vector
);
2410 (void) rw_unlock(&zd
->zd_zilog_lock
);
2411 mutex_exit(&zd
->zd_dirobj_lock
);
2415 * Verify that we can't destroy an active pool, create an existing pool,
2416 * or create a pool with a bad vdev spec.
2420 ztest_spa_create_destroy(ztest_ds_t
*zd
, uint64_t id
)
2422 ztest_shared_opts_t
*zo
= &ztest_opts
;
2427 * Attempt to create using a bad file.
2429 nvroot
= make_vdev_root("/dev/bogus", NULL
, NULL
, 0, 0, 0, 0, 0, 1);
2430 VERIFY3U(ENOENT
, ==,
2431 spa_create("ztest_bad_file", nvroot
, NULL
, NULL
));
2432 nvlist_free(nvroot
);
2435 * Attempt to create using a bad mirror.
2437 nvroot
= make_vdev_root("/dev/bogus", NULL
, NULL
, 0, 0, 0, 0, 2, 1);
2438 VERIFY3U(ENOENT
, ==,
2439 spa_create("ztest_bad_mirror", nvroot
, NULL
, NULL
));
2440 nvlist_free(nvroot
);
2443 * Attempt to create an existing pool. It shouldn't matter
2444 * what's in the nvroot; we should fail with EEXIST.
2446 (void) rw_rdlock(&ztest_name_lock
);
2447 nvroot
= make_vdev_root("/dev/bogus", NULL
, NULL
, 0, 0, 0, 0, 0, 1);
2448 VERIFY3U(EEXIST
, ==, spa_create(zo
->zo_pool
, nvroot
, NULL
, NULL
));
2449 nvlist_free(nvroot
);
2450 VERIFY3U(0, ==, spa_open(zo
->zo_pool
, &spa
, FTAG
));
2451 VERIFY3U(EBUSY
, ==, spa_destroy(zo
->zo_pool
));
2452 spa_close(spa
, FTAG
);
2454 (void) rw_unlock(&ztest_name_lock
);
2459 ztest_spa_upgrade(ztest_ds_t
*zd
, uint64_t id
)
2462 uint64_t initial_version
= SPA_VERSION_INITIAL
;
2463 uint64_t version
, newversion
;
2464 nvlist_t
*nvroot
, *props
;
2467 mutex_enter(&ztest_vdev_lock
);
2468 name
= kmem_asprintf("%s_upgrade", ztest_opts
.zo_pool
);
2471 * Clean up from previous runs.
2473 (void) spa_destroy(name
);
2475 nvroot
= make_vdev_root(NULL
, NULL
, name
, ztest_opts
.zo_vdev_size
, 0,
2476 0, ztest_opts
.zo_raidz
, ztest_opts
.zo_mirrors
, 1);
2479 * If we're configuring a RAIDZ device then make sure that the
2480 * the initial version is capable of supporting that feature.
2482 switch (ztest_opts
.zo_raidz_parity
) {
2485 initial_version
= SPA_VERSION_INITIAL
;
2488 initial_version
= SPA_VERSION_RAIDZ2
;
2491 initial_version
= SPA_VERSION_RAIDZ3
;
2496 * Create a pool with a spa version that can be upgraded. Pick
2497 * a value between initial_version and SPA_VERSION_BEFORE_FEATURES.
2500 version
= ztest_random_spa_version(initial_version
);
2501 } while (version
> SPA_VERSION_BEFORE_FEATURES
);
2503 props
= fnvlist_alloc();
2504 fnvlist_add_uint64(props
,
2505 zpool_prop_to_name(ZPOOL_PROP_VERSION
), version
);
2506 VERIFY3S(spa_create(name
, nvroot
, props
, NULL
), ==, 0);
2507 fnvlist_free(nvroot
);
2508 fnvlist_free(props
);
2510 VERIFY3S(spa_open(name
, &spa
, FTAG
), ==, 0);
2511 VERIFY3U(spa_version(spa
), ==, version
);
2512 newversion
= ztest_random_spa_version(version
+ 1);
2514 if (ztest_opts
.zo_verbose
>= 4) {
2515 (void) printf("upgrading spa version from %llu to %llu\n",
2516 (u_longlong_t
)version
, (u_longlong_t
)newversion
);
2519 spa_upgrade(spa
, newversion
);
2520 VERIFY3U(spa_version(spa
), >, version
);
2521 VERIFY3U(spa_version(spa
), ==, fnvlist_lookup_uint64(spa
->spa_config
,
2522 zpool_prop_to_name(ZPOOL_PROP_VERSION
)));
2523 spa_close(spa
, FTAG
);
2526 mutex_exit(&ztest_vdev_lock
);
2530 vdev_lookup_by_path(vdev_t
*vd
, const char *path
)
2535 if (vd
->vdev_path
!= NULL
&& strcmp(path
, vd
->vdev_path
) == 0)
2538 for (c
= 0; c
< vd
->vdev_children
; c
++)
2539 if ((mvd
= vdev_lookup_by_path(vd
->vdev_child
[c
], path
)) !=
2547 * Find the first available hole which can be used as a top-level.
2550 find_vdev_hole(spa_t
*spa
)
2552 vdev_t
*rvd
= spa
->spa_root_vdev
;
2555 ASSERT(spa_config_held(spa
, SCL_VDEV
, RW_READER
) == SCL_VDEV
);
2557 for (c
= 0; c
< rvd
->vdev_children
; c
++) {
2558 vdev_t
*cvd
= rvd
->vdev_child
[c
];
2560 if (cvd
->vdev_ishole
)
2567 * Verify that vdev_add() works as expected.
2571 ztest_vdev_add_remove(ztest_ds_t
*zd
, uint64_t id
)
2573 ztest_shared_t
*zs
= ztest_shared
;
2574 spa_t
*spa
= ztest_spa
;
2580 mutex_enter(&ztest_vdev_lock
);
2581 leaves
= MAX(zs
->zs_mirrors
+ zs
->zs_splits
, 1) * ztest_opts
.zo_raidz
;
2583 spa_config_enter(spa
, SCL_VDEV
, FTAG
, RW_READER
);
2585 ztest_shared
->zs_vdev_next_leaf
= find_vdev_hole(spa
) * leaves
;
2588 * If we have slogs then remove them 1/4 of the time.
2590 if (spa_has_slogs(spa
) && ztest_random(4) == 0) {
2592 * Grab the guid from the head of the log class rotor.
2594 guid
= spa_log_class(spa
)->mc_rotor
->mg_vd
->vdev_guid
;
2596 spa_config_exit(spa
, SCL_VDEV
, FTAG
);
2599 * We have to grab the zs_name_lock as writer to
2600 * prevent a race between removing a slog (dmu_objset_find)
2601 * and destroying a dataset. Removing the slog will
2602 * grab a reference on the dataset which may cause
2603 * dsl_destroy_head() to fail with EBUSY thus
2604 * leaving the dataset in an inconsistent state.
2606 rw_wrlock(&ztest_name_lock
);
2607 error
= spa_vdev_remove(spa
, guid
, B_FALSE
);
2608 rw_unlock(&ztest_name_lock
);
2610 if (error
&& error
!= EEXIST
)
2611 fatal(0, "spa_vdev_remove() = %d", error
);
2613 spa_config_exit(spa
, SCL_VDEV
, FTAG
);
2616 * Make 1/4 of the devices be log devices.
2618 nvroot
= make_vdev_root(NULL
, NULL
, NULL
,
2619 ztest_opts
.zo_vdev_size
, 0,
2620 ztest_random(4) == 0, ztest_opts
.zo_raidz
,
2623 error
= spa_vdev_add(spa
, nvroot
);
2624 nvlist_free(nvroot
);
2626 if (error
== ENOSPC
)
2627 ztest_record_enospc("spa_vdev_add");
2628 else if (error
!= 0)
2629 fatal(0, "spa_vdev_add() = %d", error
);
2632 mutex_exit(&ztest_vdev_lock
);
2636 * Verify that adding/removing aux devices (l2arc, hot spare) works as expected.
2640 ztest_vdev_aux_add_remove(ztest_ds_t
*zd
, uint64_t id
)
2642 ztest_shared_t
*zs
= ztest_shared
;
2643 spa_t
*spa
= ztest_spa
;
2644 vdev_t
*rvd
= spa
->spa_root_vdev
;
2645 spa_aux_vdev_t
*sav
;
2651 path
= umem_alloc(MAXPATHLEN
, UMEM_NOFAIL
);
2653 if (ztest_random(2) == 0) {
2654 sav
= &spa
->spa_spares
;
2655 aux
= ZPOOL_CONFIG_SPARES
;
2657 sav
= &spa
->spa_l2cache
;
2658 aux
= ZPOOL_CONFIG_L2CACHE
;
2661 mutex_enter(&ztest_vdev_lock
);
2663 spa_config_enter(spa
, SCL_VDEV
, FTAG
, RW_READER
);
2665 if (sav
->sav_count
!= 0 && ztest_random(4) == 0) {
2667 * Pick a random device to remove.
2669 guid
= sav
->sav_vdevs
[ztest_random(sav
->sav_count
)]->vdev_guid
;
2672 * Find an unused device we can add.
2674 zs
->zs_vdev_aux
= 0;
2677 (void) snprintf(path
, MAXPATHLEN
, ztest_aux_template
,
2678 ztest_opts
.zo_dir
, ztest_opts
.zo_pool
, aux
,
2680 for (c
= 0; c
< sav
->sav_count
; c
++)
2681 if (strcmp(sav
->sav_vdevs
[c
]->vdev_path
,
2684 if (c
== sav
->sav_count
&&
2685 vdev_lookup_by_path(rvd
, path
) == NULL
)
2691 spa_config_exit(spa
, SCL_VDEV
, FTAG
);
2697 nvlist_t
*nvroot
= make_vdev_root(NULL
, aux
, NULL
,
2698 (ztest_opts
.zo_vdev_size
* 5) / 4, 0, 0, 0, 0, 1);
2699 error
= spa_vdev_add(spa
, nvroot
);
2701 fatal(0, "spa_vdev_add(%p) = %d", nvroot
, error
);
2702 nvlist_free(nvroot
);
2705 * Remove an existing device. Sometimes, dirty its
2706 * vdev state first to make sure we handle removal
2707 * of devices that have pending state changes.
2709 if (ztest_random(2) == 0)
2710 (void) vdev_online(spa
, guid
, 0, NULL
);
2712 error
= spa_vdev_remove(spa
, guid
, B_FALSE
);
2713 if (error
!= 0 && error
!= EBUSY
)
2714 fatal(0, "spa_vdev_remove(%llu) = %d", guid
, error
);
2717 mutex_exit(&ztest_vdev_lock
);
2719 umem_free(path
, MAXPATHLEN
);
2723 * split a pool if it has mirror tlvdevs
2727 ztest_split_pool(ztest_ds_t
*zd
, uint64_t id
)
2729 ztest_shared_t
*zs
= ztest_shared
;
2730 spa_t
*spa
= ztest_spa
;
2731 vdev_t
*rvd
= spa
->spa_root_vdev
;
2732 nvlist_t
*tree
, **child
, *config
, *split
, **schild
;
2733 uint_t c
, children
, schildren
= 0, lastlogid
= 0;
2736 mutex_enter(&ztest_vdev_lock
);
2738 /* ensure we have a useable config; mirrors of raidz aren't supported */
2739 if (zs
->zs_mirrors
< 3 || ztest_opts
.zo_raidz
> 1) {
2740 mutex_exit(&ztest_vdev_lock
);
2744 /* clean up the old pool, if any */
2745 (void) spa_destroy("splitp");
2747 spa_config_enter(spa
, SCL_VDEV
, FTAG
, RW_READER
);
2749 /* generate a config from the existing config */
2750 mutex_enter(&spa
->spa_props_lock
);
2751 VERIFY(nvlist_lookup_nvlist(spa
->spa_config
, ZPOOL_CONFIG_VDEV_TREE
,
2753 mutex_exit(&spa
->spa_props_lock
);
2755 VERIFY(nvlist_lookup_nvlist_array(tree
, ZPOOL_CONFIG_CHILDREN
, &child
,
2758 schild
= malloc(rvd
->vdev_children
* sizeof (nvlist_t
*));
2759 for (c
= 0; c
< children
; c
++) {
2760 vdev_t
*tvd
= rvd
->vdev_child
[c
];
2764 if (tvd
->vdev_islog
|| tvd
->vdev_ops
== &vdev_hole_ops
) {
2765 VERIFY(nvlist_alloc(&schild
[schildren
], NV_UNIQUE_NAME
,
2767 VERIFY(nvlist_add_string(schild
[schildren
],
2768 ZPOOL_CONFIG_TYPE
, VDEV_TYPE_HOLE
) == 0);
2769 VERIFY(nvlist_add_uint64(schild
[schildren
],
2770 ZPOOL_CONFIG_IS_HOLE
, 1) == 0);
2772 lastlogid
= schildren
;
2777 VERIFY(nvlist_lookup_nvlist_array(child
[c
],
2778 ZPOOL_CONFIG_CHILDREN
, &mchild
, &mchildren
) == 0);
2779 VERIFY(nvlist_dup(mchild
[0], &schild
[schildren
++], 0) == 0);
2782 /* OK, create a config that can be used to split */
2783 VERIFY(nvlist_alloc(&split
, NV_UNIQUE_NAME
, 0) == 0);
2784 VERIFY(nvlist_add_string(split
, ZPOOL_CONFIG_TYPE
,
2785 VDEV_TYPE_ROOT
) == 0);
2786 VERIFY(nvlist_add_nvlist_array(split
, ZPOOL_CONFIG_CHILDREN
, schild
,
2787 lastlogid
!= 0 ? lastlogid
: schildren
) == 0);
2789 VERIFY(nvlist_alloc(&config
, NV_UNIQUE_NAME
, 0) == 0);
2790 VERIFY(nvlist_add_nvlist(config
, ZPOOL_CONFIG_VDEV_TREE
, split
) == 0);
2792 for (c
= 0; c
< schildren
; c
++)
2793 nvlist_free(schild
[c
]);
2797 spa_config_exit(spa
, SCL_VDEV
, FTAG
);
2799 (void) rw_wrlock(&ztest_name_lock
);
2800 error
= spa_vdev_split_mirror(spa
, "splitp", config
, NULL
, B_FALSE
);
2801 (void) rw_unlock(&ztest_name_lock
);
2803 nvlist_free(config
);
2806 (void) printf("successful split - results:\n");
2807 mutex_enter(&spa_namespace_lock
);
2808 show_pool_stats(spa
);
2809 show_pool_stats(spa_lookup("splitp"));
2810 mutex_exit(&spa_namespace_lock
);
2814 mutex_exit(&ztest_vdev_lock
);
2819 * Verify that we can attach and detach devices.
2823 ztest_vdev_attach_detach(ztest_ds_t
*zd
, uint64_t id
)
2825 ztest_shared_t
*zs
= ztest_shared
;
2826 spa_t
*spa
= ztest_spa
;
2827 spa_aux_vdev_t
*sav
= &spa
->spa_spares
;
2828 vdev_t
*rvd
= spa
->spa_root_vdev
;
2829 vdev_t
*oldvd
, *newvd
, *pvd
;
2833 uint64_t ashift
= ztest_get_ashift();
2834 uint64_t oldguid
, pguid
;
2835 uint64_t oldsize
, newsize
;
2836 char *oldpath
, *newpath
;
2838 int oldvd_has_siblings
= B_FALSE
;
2839 int newvd_is_spare
= B_FALSE
;
2841 int error
, expected_error
;
2843 oldpath
= umem_alloc(MAXPATHLEN
, UMEM_NOFAIL
);
2844 newpath
= umem_alloc(MAXPATHLEN
, UMEM_NOFAIL
);
2846 mutex_enter(&ztest_vdev_lock
);
2847 leaves
= MAX(zs
->zs_mirrors
, 1) * ztest_opts
.zo_raidz
;
2849 spa_config_enter(spa
, SCL_VDEV
, FTAG
, RW_READER
);
2852 * Decide whether to do an attach or a replace.
2854 replacing
= ztest_random(2);
2857 * Pick a random top-level vdev.
2859 top
= ztest_random_vdev_top(spa
, B_TRUE
);
2862 * Pick a random leaf within it.
2864 leaf
= ztest_random(leaves
);
2869 oldvd
= rvd
->vdev_child
[top
];
2870 if (zs
->zs_mirrors
>= 1) {
2871 ASSERT(oldvd
->vdev_ops
== &vdev_mirror_ops
);
2872 ASSERT(oldvd
->vdev_children
>= zs
->zs_mirrors
);
2873 oldvd
= oldvd
->vdev_child
[leaf
/ ztest_opts
.zo_raidz
];
2875 if (ztest_opts
.zo_raidz
> 1) {
2876 ASSERT(oldvd
->vdev_ops
== &vdev_raidz_ops
);
2877 ASSERT(oldvd
->vdev_children
== ztest_opts
.zo_raidz
);
2878 oldvd
= oldvd
->vdev_child
[leaf
% ztest_opts
.zo_raidz
];
2882 * If we're already doing an attach or replace, oldvd may be a
2883 * mirror vdev -- in which case, pick a random child.
2885 while (oldvd
->vdev_children
!= 0) {
2886 oldvd_has_siblings
= B_TRUE
;
2887 ASSERT(oldvd
->vdev_children
>= 2);
2888 oldvd
= oldvd
->vdev_child
[ztest_random(oldvd
->vdev_children
)];
2891 oldguid
= oldvd
->vdev_guid
;
2892 oldsize
= vdev_get_min_asize(oldvd
);
2893 oldvd_is_log
= oldvd
->vdev_top
->vdev_islog
;
2894 (void) strcpy(oldpath
, oldvd
->vdev_path
);
2895 pvd
= oldvd
->vdev_parent
;
2896 pguid
= pvd
->vdev_guid
;
2899 * If oldvd has siblings, then half of the time, detach it.
2901 if (oldvd_has_siblings
&& ztest_random(2) == 0) {
2902 spa_config_exit(spa
, SCL_VDEV
, FTAG
);
2903 error
= spa_vdev_detach(spa
, oldguid
, pguid
, B_FALSE
);
2904 if (error
!= 0 && error
!= ENODEV
&& error
!= EBUSY
&&
2906 fatal(0, "detach (%s) returned %d", oldpath
, error
);
2911 * For the new vdev, choose with equal probability between the two
2912 * standard paths (ending in either 'a' or 'b') or a random hot spare.
2914 if (sav
->sav_count
!= 0 && ztest_random(3) == 0) {
2915 newvd
= sav
->sav_vdevs
[ztest_random(sav
->sav_count
)];
2916 newvd_is_spare
= B_TRUE
;
2917 (void) strcpy(newpath
, newvd
->vdev_path
);
2919 (void) snprintf(newpath
, MAXPATHLEN
, ztest_dev_template
,
2920 ztest_opts
.zo_dir
, ztest_opts
.zo_pool
,
2921 top
* leaves
+ leaf
);
2922 if (ztest_random(2) == 0)
2923 newpath
[strlen(newpath
) - 1] = 'b';
2924 newvd
= vdev_lookup_by_path(rvd
, newpath
);
2928 newsize
= vdev_get_min_asize(newvd
);
2931 * Make newsize a little bigger or smaller than oldsize.
2932 * If it's smaller, the attach should fail.
2933 * If it's larger, and we're doing a replace,
2934 * we should get dynamic LUN growth when we're done.
2936 newsize
= 10 * oldsize
/ (9 + ztest_random(3));
2940 * If pvd is not a mirror or root, the attach should fail with ENOTSUP,
2941 * unless it's a replace; in that case any non-replacing parent is OK.
2943 * If newvd is already part of the pool, it should fail with EBUSY.
2945 * If newvd is too small, it should fail with EOVERFLOW.
2947 if (pvd
->vdev_ops
!= &vdev_mirror_ops
&&
2948 pvd
->vdev_ops
!= &vdev_root_ops
&& (!replacing
||
2949 pvd
->vdev_ops
== &vdev_replacing_ops
||
2950 pvd
->vdev_ops
== &vdev_spare_ops
))
2951 expected_error
= ENOTSUP
;
2952 else if (newvd_is_spare
&& (!replacing
|| oldvd_is_log
))
2953 expected_error
= ENOTSUP
;
2954 else if (newvd
== oldvd
)
2955 expected_error
= replacing
? 0 : EBUSY
;
2956 else if (vdev_lookup_by_path(rvd
, newpath
) != NULL
)
2957 expected_error
= EBUSY
;
2958 else if (newsize
< oldsize
)
2959 expected_error
= EOVERFLOW
;
2960 else if (ashift
> oldvd
->vdev_top
->vdev_ashift
)
2961 expected_error
= EDOM
;
2965 spa_config_exit(spa
, SCL_VDEV
, FTAG
);
2968 * Build the nvlist describing newpath.
2970 root
= make_vdev_root(newpath
, NULL
, NULL
, newvd
== NULL
? newsize
: 0,
2971 ashift
, 0, 0, 0, 1);
2973 error
= spa_vdev_attach(spa
, oldguid
, root
, replacing
);
2978 * If our parent was the replacing vdev, but the replace completed,
2979 * then instead of failing with ENOTSUP we may either succeed,
2980 * fail with ENODEV, or fail with EOVERFLOW.
2982 if (expected_error
== ENOTSUP
&&
2983 (error
== 0 || error
== ENODEV
|| error
== EOVERFLOW
))
2984 expected_error
= error
;
2987 * If someone grew the LUN, the replacement may be too small.
2989 if (error
== EOVERFLOW
|| error
== EBUSY
)
2990 expected_error
= error
;
2992 /* XXX workaround 6690467 */
2993 if (error
!= expected_error
&& expected_error
!= EBUSY
) {
2994 fatal(0, "attach (%s %llu, %s %llu, %d) "
2995 "returned %d, expected %d",
2996 oldpath
, oldsize
, newpath
,
2997 newsize
, replacing
, error
, expected_error
);
3000 mutex_exit(&ztest_vdev_lock
);
3002 umem_free(oldpath
, MAXPATHLEN
);
3003 umem_free(newpath
, MAXPATHLEN
);
3007 * Callback function which expands the physical size of the vdev.
3010 grow_vdev(vdev_t
*vd
, void *arg
)
3012 ASSERTV(spa_t
*spa
= vd
->vdev_spa
);
3013 size_t *newsize
= arg
;
3017 ASSERT(spa_config_held(spa
, SCL_STATE
, RW_READER
) == SCL_STATE
);
3018 ASSERT(vd
->vdev_ops
->vdev_op_leaf
);
3020 if ((fd
= open(vd
->vdev_path
, O_RDWR
)) == -1)
3023 fsize
= lseek(fd
, 0, SEEK_END
);
3024 VERIFY(ftruncate(fd
, *newsize
) == 0);
3026 if (ztest_opts
.zo_verbose
>= 6) {
3027 (void) printf("%s grew from %lu to %lu bytes\n",
3028 vd
->vdev_path
, (ulong_t
)fsize
, (ulong_t
)*newsize
);
3035 * Callback function which expands a given vdev by calling vdev_online().
3039 online_vdev(vdev_t
*vd
, void *arg
)
3041 spa_t
*spa
= vd
->vdev_spa
;
3042 vdev_t
*tvd
= vd
->vdev_top
;
3043 uint64_t guid
= vd
->vdev_guid
;
3044 uint64_t generation
= spa
->spa_config_generation
+ 1;
3045 vdev_state_t newstate
= VDEV_STATE_UNKNOWN
;
3048 ASSERT(spa_config_held(spa
, SCL_STATE
, RW_READER
) == SCL_STATE
);
3049 ASSERT(vd
->vdev_ops
->vdev_op_leaf
);
3051 /* Calling vdev_online will initialize the new metaslabs */
3052 spa_config_exit(spa
, SCL_STATE
, spa
);
3053 error
= vdev_online(spa
, guid
, ZFS_ONLINE_EXPAND
, &newstate
);
3054 spa_config_enter(spa
, SCL_STATE
, spa
, RW_READER
);
3057 * If vdev_online returned an error or the underlying vdev_open
3058 * failed then we abort the expand. The only way to know that
3059 * vdev_open fails is by checking the returned newstate.
3061 if (error
|| newstate
!= VDEV_STATE_HEALTHY
) {
3062 if (ztest_opts
.zo_verbose
>= 5) {
3063 (void) printf("Unable to expand vdev, state %llu, "
3064 "error %d\n", (u_longlong_t
)newstate
, error
);
3068 ASSERT3U(newstate
, ==, VDEV_STATE_HEALTHY
);
3071 * Since we dropped the lock we need to ensure that we're
3072 * still talking to the original vdev. It's possible this
3073 * vdev may have been detached/replaced while we were
3074 * trying to online it.
3076 if (generation
!= spa
->spa_config_generation
) {
3077 if (ztest_opts
.zo_verbose
>= 5) {
3078 (void) printf("vdev configuration has changed, "
3079 "guid %llu, state %llu, expected gen %llu, "
3082 (u_longlong_t
)tvd
->vdev_state
,
3083 (u_longlong_t
)generation
,
3084 (u_longlong_t
)spa
->spa_config_generation
);
3092 * Traverse the vdev tree calling the supplied function.
3093 * We continue to walk the tree until we either have walked all
3094 * children or we receive a non-NULL return from the callback.
3095 * If a NULL callback is passed, then we just return back the first
3096 * leaf vdev we encounter.
3099 vdev_walk_tree(vdev_t
*vd
, vdev_t
*(*func
)(vdev_t
*, void *), void *arg
)
3103 if (vd
->vdev_ops
->vdev_op_leaf
) {
3107 return (func(vd
, arg
));
3110 for (c
= 0; c
< vd
->vdev_children
; c
++) {
3111 vdev_t
*cvd
= vd
->vdev_child
[c
];
3112 if ((cvd
= vdev_walk_tree(cvd
, func
, arg
)) != NULL
)
3119 * Verify that dynamic LUN growth works as expected.
3123 ztest_vdev_LUN_growth(ztest_ds_t
*zd
, uint64_t id
)
3125 spa_t
*spa
= ztest_spa
;
3127 metaslab_class_t
*mc
;
3128 metaslab_group_t
*mg
;
3129 size_t psize
, newsize
;
3131 uint64_t old_class_space
, new_class_space
, old_ms_count
, new_ms_count
;
3133 mutex_enter(&ztest_vdev_lock
);
3134 spa_config_enter(spa
, SCL_STATE
, spa
, RW_READER
);
3136 top
= ztest_random_vdev_top(spa
, B_TRUE
);
3138 tvd
= spa
->spa_root_vdev
->vdev_child
[top
];
3141 old_ms_count
= tvd
->vdev_ms_count
;
3142 old_class_space
= metaslab_class_get_space(mc
);
3145 * Determine the size of the first leaf vdev associated with
3146 * our top-level device.
3148 vd
= vdev_walk_tree(tvd
, NULL
, NULL
);
3149 ASSERT3P(vd
, !=, NULL
);
3150 ASSERT(vd
->vdev_ops
->vdev_op_leaf
);
3152 psize
= vd
->vdev_psize
;
3155 * We only try to expand the vdev if it's healthy, less than 4x its
3156 * original size, and it has a valid psize.
3158 if (tvd
->vdev_state
!= VDEV_STATE_HEALTHY
||
3159 psize
== 0 || psize
>= 4 * ztest_opts
.zo_vdev_size
) {
3160 spa_config_exit(spa
, SCL_STATE
, spa
);
3161 mutex_exit(&ztest_vdev_lock
);
3165 newsize
= psize
+ psize
/ 8;
3166 ASSERT3U(newsize
, >, psize
);
3168 if (ztest_opts
.zo_verbose
>= 6) {
3169 (void) printf("Expanding LUN %s from %lu to %lu\n",
3170 vd
->vdev_path
, (ulong_t
)psize
, (ulong_t
)newsize
);
3174 * Growing the vdev is a two step process:
3175 * 1). expand the physical size (i.e. relabel)
3176 * 2). online the vdev to create the new metaslabs
3178 if (vdev_walk_tree(tvd
, grow_vdev
, &newsize
) != NULL
||
3179 vdev_walk_tree(tvd
, online_vdev
, NULL
) != NULL
||
3180 tvd
->vdev_state
!= VDEV_STATE_HEALTHY
) {
3181 if (ztest_opts
.zo_verbose
>= 5) {
3182 (void) printf("Could not expand LUN because "
3183 "the vdev configuration changed.\n");
3185 spa_config_exit(spa
, SCL_STATE
, spa
);
3186 mutex_exit(&ztest_vdev_lock
);
3190 spa_config_exit(spa
, SCL_STATE
, spa
);
3193 * Expanding the LUN will update the config asynchronously,
3194 * thus we must wait for the async thread to complete any
3195 * pending tasks before proceeding.
3199 mutex_enter(&spa
->spa_async_lock
);
3200 done
= (spa
->spa_async_thread
== NULL
&& !spa
->spa_async_tasks
);
3201 mutex_exit(&spa
->spa_async_lock
);
3204 txg_wait_synced(spa_get_dsl(spa
), 0);
3205 (void) poll(NULL
, 0, 100);
3208 spa_config_enter(spa
, SCL_STATE
, spa
, RW_READER
);
3210 tvd
= spa
->spa_root_vdev
->vdev_child
[top
];
3211 new_ms_count
= tvd
->vdev_ms_count
;
3212 new_class_space
= metaslab_class_get_space(mc
);
3214 if (tvd
->vdev_mg
!= mg
|| mg
->mg_class
!= mc
) {
3215 if (ztest_opts
.zo_verbose
>= 5) {
3216 (void) printf("Could not verify LUN expansion due to "
3217 "intervening vdev offline or remove.\n");
3219 spa_config_exit(spa
, SCL_STATE
, spa
);
3220 mutex_exit(&ztest_vdev_lock
);
3225 * Make sure we were able to grow the vdev.
3227 if (new_ms_count
<= old_ms_count
)
3228 fatal(0, "LUN expansion failed: ms_count %llu <= %llu\n",
3229 old_ms_count
, new_ms_count
);
3232 * Make sure we were able to grow the pool.
3234 if (new_class_space
<= old_class_space
)
3235 fatal(0, "LUN expansion failed: class_space %llu <= %llu\n",
3236 old_class_space
, new_class_space
);
3238 if (ztest_opts
.zo_verbose
>= 5) {
3239 char oldnumbuf
[6], newnumbuf
[6];
3241 nicenum(old_class_space
, oldnumbuf
);
3242 nicenum(new_class_space
, newnumbuf
);
3243 (void) printf("%s grew from %s to %s\n",
3244 spa
->spa_name
, oldnumbuf
, newnumbuf
);
3247 spa_config_exit(spa
, SCL_STATE
, spa
);
3248 mutex_exit(&ztest_vdev_lock
);
3252 * Verify that dmu_objset_{create,destroy,open,close} work as expected.
3256 ztest_objset_create_cb(objset_t
*os
, void *arg
, cred_t
*cr
, dmu_tx_t
*tx
)
3259 * Create the objects common to all ztest datasets.
3261 VERIFY(zap_create_claim(os
, ZTEST_DIROBJ
,
3262 DMU_OT_ZAP_OTHER
, DMU_OT_NONE
, 0, tx
) == 0);
3266 ztest_dataset_create(char *dsname
)
3268 uint64_t zilset
= ztest_random(100);
3269 int err
= dmu_objset_create(dsname
, DMU_OST_OTHER
, 0,
3270 ztest_objset_create_cb
, NULL
);
3272 if (err
|| zilset
< 80)
3275 if (ztest_opts
.zo_verbose
>= 5)
3276 (void) printf("Setting dataset %s to sync always\n", dsname
);
3277 return (ztest_dsl_prop_set_uint64(dsname
, ZFS_PROP_SYNC
,
3278 ZFS_SYNC_ALWAYS
, B_FALSE
));
3283 ztest_objset_destroy_cb(const char *name
, void *arg
)
3286 dmu_object_info_t doi
;
3290 * Verify that the dataset contains a directory object.
3292 VERIFY0(dmu_objset_own(name
, DMU_OST_OTHER
, B_TRUE
, FTAG
, &os
));
3293 error
= dmu_object_info(os
, ZTEST_DIROBJ
, &doi
);
3294 if (error
!= ENOENT
) {
3295 /* We could have crashed in the middle of destroying it */
3297 ASSERT3U(doi
.doi_type
, ==, DMU_OT_ZAP_OTHER
);
3298 ASSERT3S(doi
.doi_physical_blocks_512
, >=, 0);
3300 dmu_objset_disown(os
, FTAG
);
3303 * Destroy the dataset.
3305 if (strchr(name
, '@') != NULL
) {
3306 VERIFY0(dsl_destroy_snapshot(name
, B_FALSE
));
3308 VERIFY0(dsl_destroy_head(name
));
3314 ztest_snapshot_create(char *osname
, uint64_t id
)
3316 char snapname
[MAXNAMELEN
];
3319 (void) snprintf(snapname
, sizeof (snapname
), "%llu", (u_longlong_t
)id
);
3321 error
= dmu_objset_snapshot_one(osname
, snapname
);
3322 if (error
== ENOSPC
) {
3323 ztest_record_enospc(FTAG
);
3326 if (error
!= 0 && error
!= EEXIST
) {
3327 fatal(0, "ztest_snapshot_create(%s@%s) = %d", osname
,
3334 ztest_snapshot_destroy(char *osname
, uint64_t id
)
3336 char snapname
[MAXNAMELEN
];
3339 (void) snprintf(snapname
, MAXNAMELEN
, "%s@%llu", osname
,
3342 error
= dsl_destroy_snapshot(snapname
, B_FALSE
);
3343 if (error
!= 0 && error
!= ENOENT
)
3344 fatal(0, "ztest_snapshot_destroy(%s) = %d", snapname
, error
);
3350 ztest_dmu_objset_create_destroy(ztest_ds_t
*zd
, uint64_t id
)
3360 zdtmp
= umem_alloc(sizeof (ztest_ds_t
), UMEM_NOFAIL
);
3361 name
= umem_alloc(MAXNAMELEN
, UMEM_NOFAIL
);
3363 (void) rw_rdlock(&ztest_name_lock
);
3365 (void) snprintf(name
, MAXNAMELEN
, "%s/temp_%llu",
3366 ztest_opts
.zo_pool
, (u_longlong_t
)id
);
3369 * If this dataset exists from a previous run, process its replay log
3370 * half of the time. If we don't replay it, then dsl_destroy_head()
3371 * (invoked from ztest_objset_destroy_cb()) should just throw it away.
3373 if (ztest_random(2) == 0 &&
3374 dmu_objset_own(name
, DMU_OST_OTHER
, B_FALSE
, FTAG
, &os
) == 0) {
3375 ztest_zd_init(zdtmp
, NULL
, os
);
3376 zil_replay(os
, zdtmp
, ztest_replay_vector
);
3377 ztest_zd_fini(zdtmp
);
3378 dmu_objset_disown(os
, FTAG
);
3382 * There may be an old instance of the dataset we're about to
3383 * create lying around from a previous run. If so, destroy it
3384 * and all of its snapshots.
3386 (void) dmu_objset_find(name
, ztest_objset_destroy_cb
, NULL
,
3387 DS_FIND_CHILDREN
| DS_FIND_SNAPSHOTS
);
3390 * Verify that the destroyed dataset is no longer in the namespace.
3392 VERIFY3U(ENOENT
, ==, dmu_objset_own(name
, DMU_OST_OTHER
, B_TRUE
,
3396 * Verify that we can create a new dataset.
3398 error
= ztest_dataset_create(name
);
3400 if (error
== ENOSPC
) {
3401 ztest_record_enospc(FTAG
);
3404 fatal(0, "dmu_objset_create(%s) = %d", name
, error
);
3407 VERIFY0(dmu_objset_own(name
, DMU_OST_OTHER
, B_FALSE
, FTAG
, &os
));
3409 ztest_zd_init(zdtmp
, NULL
, os
);
3412 * Open the intent log for it.
3414 zilog
= zil_open(os
, ztest_get_data
);
3417 * Put some objects in there, do a little I/O to them,
3418 * and randomly take a couple of snapshots along the way.
3420 iters
= ztest_random(5);
3421 for (i
= 0; i
< iters
; i
++) {
3422 ztest_dmu_object_alloc_free(zdtmp
, id
);
3423 if (ztest_random(iters
) == 0)
3424 (void) ztest_snapshot_create(name
, i
);
3428 * Verify that we cannot create an existing dataset.
3430 VERIFY3U(EEXIST
, ==,
3431 dmu_objset_create(name
, DMU_OST_OTHER
, 0, NULL
, NULL
));
3434 * Verify that we can hold an objset that is also owned.
3436 VERIFY3U(0, ==, dmu_objset_hold(name
, FTAG
, &os2
));
3437 dmu_objset_rele(os2
, FTAG
);
3440 * Verify that we cannot own an objset that is already owned.
3443 dmu_objset_own(name
, DMU_OST_OTHER
, B_FALSE
, FTAG
, &os2
));
3446 dmu_objset_disown(os
, FTAG
);
3447 ztest_zd_fini(zdtmp
);
3449 (void) rw_unlock(&ztest_name_lock
);
3451 umem_free(name
, MAXNAMELEN
);
3452 umem_free(zdtmp
, sizeof (ztest_ds_t
));
3456 * Verify that dmu_snapshot_{create,destroy,open,close} work as expected.
3459 ztest_dmu_snapshot_create_destroy(ztest_ds_t
*zd
, uint64_t id
)
3461 (void) rw_rdlock(&ztest_name_lock
);
3462 (void) ztest_snapshot_destroy(zd
->zd_name
, id
);
3463 (void) ztest_snapshot_create(zd
->zd_name
, id
);
3464 (void) rw_unlock(&ztest_name_lock
);
3468 * Cleanup non-standard snapshots and clones.
3471 ztest_dsl_dataset_cleanup(char *osname
, uint64_t id
)
3480 snap1name
= umem_alloc(MAXNAMELEN
, UMEM_NOFAIL
);
3481 clone1name
= umem_alloc(MAXNAMELEN
, UMEM_NOFAIL
);
3482 snap2name
= umem_alloc(MAXNAMELEN
, UMEM_NOFAIL
);
3483 clone2name
= umem_alloc(MAXNAMELEN
, UMEM_NOFAIL
);
3484 snap3name
= umem_alloc(MAXNAMELEN
, UMEM_NOFAIL
);
3486 (void) snprintf(snap1name
, MAXNAMELEN
, "%s@s1_%llu",
3487 osname
, (u_longlong_t
)id
);
3488 (void) snprintf(clone1name
, MAXNAMELEN
, "%s/c1_%llu",
3489 osname
, (u_longlong_t
)id
);
3490 (void) snprintf(snap2name
, MAXNAMELEN
, "%s@s2_%llu",
3491 clone1name
, (u_longlong_t
)id
);
3492 (void) snprintf(clone2name
, MAXNAMELEN
, "%s/c2_%llu",
3493 osname
, (u_longlong_t
)id
);
3494 (void) snprintf(snap3name
, MAXNAMELEN
, "%s@s3_%llu",
3495 clone1name
, (u_longlong_t
)id
);
3497 error
= dsl_destroy_head(clone2name
);
3498 if (error
&& error
!= ENOENT
)
3499 fatal(0, "dsl_destroy_head(%s) = %d", clone2name
, error
);
3500 error
= dsl_destroy_snapshot(snap3name
, B_FALSE
);
3501 if (error
&& error
!= ENOENT
)
3502 fatal(0, "dsl_destroy_snapshot(%s) = %d", snap3name
, error
);
3503 error
= dsl_destroy_snapshot(snap2name
, B_FALSE
);
3504 if (error
&& error
!= ENOENT
)
3505 fatal(0, "dsl_destroy_snapshot(%s) = %d", snap2name
, error
);
3506 error
= dsl_destroy_head(clone1name
);
3507 if (error
&& error
!= ENOENT
)
3508 fatal(0, "dsl_destroy_head(%s) = %d", clone1name
, error
);
3509 error
= dsl_destroy_snapshot(snap1name
, B_FALSE
);
3510 if (error
&& error
!= ENOENT
)
3511 fatal(0, "dsl_destroy_snapshot(%s) = %d", snap1name
, error
);
3513 umem_free(snap1name
, MAXNAMELEN
);
3514 umem_free(clone1name
, MAXNAMELEN
);
3515 umem_free(snap2name
, MAXNAMELEN
);
3516 umem_free(clone2name
, MAXNAMELEN
);
3517 umem_free(snap3name
, MAXNAMELEN
);
3521 * Verify dsl_dataset_promote handles EBUSY
3524 ztest_dsl_dataset_promote_busy(ztest_ds_t
*zd
, uint64_t id
)
3532 char *osname
= zd
->zd_name
;
3535 snap1name
= umem_alloc(MAXNAMELEN
, UMEM_NOFAIL
);
3536 clone1name
= umem_alloc(MAXNAMELEN
, UMEM_NOFAIL
);
3537 snap2name
= umem_alloc(MAXNAMELEN
, UMEM_NOFAIL
);
3538 clone2name
= umem_alloc(MAXNAMELEN
, UMEM_NOFAIL
);
3539 snap3name
= umem_alloc(MAXNAMELEN
, UMEM_NOFAIL
);
3541 (void) rw_rdlock(&ztest_name_lock
);
3543 ztest_dsl_dataset_cleanup(osname
, id
);
3545 (void) snprintf(snap1name
, MAXNAMELEN
, "%s@s1_%llu",
3546 osname
, (u_longlong_t
)id
);
3547 (void) snprintf(clone1name
, MAXNAMELEN
, "%s/c1_%llu",
3548 osname
, (u_longlong_t
)id
);
3549 (void) snprintf(snap2name
, MAXNAMELEN
, "%s@s2_%llu",
3550 clone1name
, (u_longlong_t
)id
);
3551 (void) snprintf(clone2name
, MAXNAMELEN
, "%s/c2_%llu",
3552 osname
, (u_longlong_t
)id
);
3553 (void) snprintf(snap3name
, MAXNAMELEN
, "%s@s3_%llu",
3554 clone1name
, (u_longlong_t
)id
);
3556 error
= dmu_objset_snapshot_one(osname
, strchr(snap1name
, '@') + 1);
3557 if (error
&& error
!= EEXIST
) {
3558 if (error
== ENOSPC
) {
3559 ztest_record_enospc(FTAG
);
3562 fatal(0, "dmu_take_snapshot(%s) = %d", snap1name
, error
);
3565 error
= dmu_objset_clone(clone1name
, snap1name
);
3567 if (error
== ENOSPC
) {
3568 ztest_record_enospc(FTAG
);
3571 fatal(0, "dmu_objset_create(%s) = %d", clone1name
, error
);
3574 error
= dmu_objset_snapshot_one(clone1name
, strchr(snap2name
, '@') + 1);
3575 if (error
&& error
!= EEXIST
) {
3576 if (error
== ENOSPC
) {
3577 ztest_record_enospc(FTAG
);
3580 fatal(0, "dmu_open_snapshot(%s) = %d", snap2name
, error
);
3583 error
= dmu_objset_snapshot_one(clone1name
, strchr(snap3name
, '@') + 1);
3584 if (error
&& error
!= EEXIST
) {
3585 if (error
== ENOSPC
) {
3586 ztest_record_enospc(FTAG
);
3589 fatal(0, "dmu_open_snapshot(%s) = %d", snap3name
, error
);
3592 error
= dmu_objset_clone(clone2name
, snap3name
);
3594 if (error
== ENOSPC
) {
3595 ztest_record_enospc(FTAG
);
3598 fatal(0, "dmu_objset_create(%s) = %d", clone2name
, error
);
3601 error
= dmu_objset_own(snap2name
, DMU_OST_ANY
, B_TRUE
, FTAG
, &os
);
3603 fatal(0, "dmu_objset_own(%s) = %d", snap2name
, error
);
3604 error
= dsl_dataset_promote(clone2name
, NULL
);
3605 if (error
== ENOSPC
) {
3606 dmu_objset_disown(os
, FTAG
);
3607 ztest_record_enospc(FTAG
);
3611 fatal(0, "dsl_dataset_promote(%s), %d, not EBUSY", clone2name
,
3613 dmu_objset_disown(os
, FTAG
);
3616 ztest_dsl_dataset_cleanup(osname
, id
);
3618 (void) rw_unlock(&ztest_name_lock
);
3620 umem_free(snap1name
, MAXNAMELEN
);
3621 umem_free(clone1name
, MAXNAMELEN
);
3622 umem_free(snap2name
, MAXNAMELEN
);
3623 umem_free(clone2name
, MAXNAMELEN
);
3624 umem_free(snap3name
, MAXNAMELEN
);
3627 #undef OD_ARRAY_SIZE
3628 #define OD_ARRAY_SIZE 4
3631 * Verify that dmu_object_{alloc,free} work as expected.
3634 ztest_dmu_object_alloc_free(ztest_ds_t
*zd
, uint64_t id
)
3641 size
= sizeof (ztest_od_t
) * OD_ARRAY_SIZE
;
3642 od
= umem_alloc(size
, UMEM_NOFAIL
);
3643 batchsize
= OD_ARRAY_SIZE
;
3645 for (b
= 0; b
< batchsize
; b
++)
3646 ztest_od_init(od
+ b
, id
, FTAG
, b
, DMU_OT_UINT64_OTHER
, 0, 0);
3649 * Destroy the previous batch of objects, create a new batch,
3650 * and do some I/O on the new objects.
3652 if (ztest_object_init(zd
, od
, size
, B_TRUE
) != 0)
3655 while (ztest_random(4 * batchsize
) != 0)
3656 ztest_io(zd
, od
[ztest_random(batchsize
)].od_object
,
3657 ztest_random(ZTEST_RANGE_LOCKS
) << SPA_MAXBLOCKSHIFT
);
3659 umem_free(od
, size
);
3662 #undef OD_ARRAY_SIZE
3663 #define OD_ARRAY_SIZE 2
3666 * Verify that dmu_{read,write} work as expected.
3669 ztest_dmu_read_write(ztest_ds_t
*zd
, uint64_t id
)
3674 objset_t
*os
= zd
->zd_os
;
3675 size
= sizeof (ztest_od_t
) * OD_ARRAY_SIZE
;
3676 od
= umem_alloc(size
, UMEM_NOFAIL
);
3678 int i
, freeit
, error
;
3680 bufwad_t
*packbuf
, *bigbuf
, *pack
, *bigH
, *bigT
;
3681 uint64_t packobj
, packoff
, packsize
, bigobj
, bigoff
, bigsize
;
3682 uint64_t chunksize
= (1000 + ztest_random(1000)) * sizeof (uint64_t);
3683 uint64_t regions
= 997;
3684 uint64_t stride
= 123456789ULL;
3685 uint64_t width
= 40;
3686 int free_percent
= 5;
3689 * This test uses two objects, packobj and bigobj, that are always
3690 * updated together (i.e. in the same tx) so that their contents are
3691 * in sync and can be compared. Their contents relate to each other
3692 * in a simple way: packobj is a dense array of 'bufwad' structures,
3693 * while bigobj is a sparse array of the same bufwads. Specifically,
3694 * for any index n, there are three bufwads that should be identical:
3696 * packobj, at offset n * sizeof (bufwad_t)
3697 * bigobj, at the head of the nth chunk
3698 * bigobj, at the tail of the nth chunk
3700 * The chunk size is arbitrary. It doesn't have to be a power of two,
3701 * and it doesn't have any relation to the object blocksize.
3702 * The only requirement is that it can hold at least two bufwads.
3704 * Normally, we write the bufwad to each of these locations.
3705 * However, free_percent of the time we instead write zeroes to
3706 * packobj and perform a dmu_free_range() on bigobj. By comparing
3707 * bigobj to packobj, we can verify that the DMU is correctly
3708 * tracking which parts of an object are allocated and free,
3709 * and that the contents of the allocated blocks are correct.
3713 * Read the directory info. If it's the first time, set things up.
3715 ztest_od_init(od
, id
, FTAG
, 0, DMU_OT_UINT64_OTHER
, 0, chunksize
);
3716 ztest_od_init(od
+ 1, id
, FTAG
, 1, DMU_OT_UINT64_OTHER
, 0, chunksize
);
3718 if (ztest_object_init(zd
, od
, size
, B_FALSE
) != 0) {
3719 umem_free(od
, size
);
3723 bigobj
= od
[0].od_object
;
3724 packobj
= od
[1].od_object
;
3725 chunksize
= od
[0].od_gen
;
3726 ASSERT(chunksize
== od
[1].od_gen
);
3729 * Prefetch a random chunk of the big object.
3730 * Our aim here is to get some async reads in flight
3731 * for blocks that we may free below; the DMU should
3732 * handle this race correctly.
3734 n
= ztest_random(regions
) * stride
+ ztest_random(width
);
3735 s
= 1 + ztest_random(2 * width
- 1);
3736 dmu_prefetch(os
, bigobj
, n
* chunksize
, s
* chunksize
);
3739 * Pick a random index and compute the offsets into packobj and bigobj.
3741 n
= ztest_random(regions
) * stride
+ ztest_random(width
);
3742 s
= 1 + ztest_random(width
- 1);
3744 packoff
= n
* sizeof (bufwad_t
);
3745 packsize
= s
* sizeof (bufwad_t
);
3747 bigoff
= n
* chunksize
;
3748 bigsize
= s
* chunksize
;
3750 packbuf
= umem_alloc(packsize
, UMEM_NOFAIL
);
3751 bigbuf
= umem_alloc(bigsize
, UMEM_NOFAIL
);
3754 * free_percent of the time, free a range of bigobj rather than
3757 freeit
= (ztest_random(100) < free_percent
);
3760 * Read the current contents of our objects.
3762 error
= dmu_read(os
, packobj
, packoff
, packsize
, packbuf
,
3765 error
= dmu_read(os
, bigobj
, bigoff
, bigsize
, bigbuf
,
3770 * Get a tx for the mods to both packobj and bigobj.
3772 tx
= dmu_tx_create(os
);
3774 dmu_tx_hold_write(tx
, packobj
, packoff
, packsize
);
3777 dmu_tx_hold_free(tx
, bigobj
, bigoff
, bigsize
);
3779 dmu_tx_hold_write(tx
, bigobj
, bigoff
, bigsize
);
3781 /* This accounts for setting the checksum/compression. */
3782 dmu_tx_hold_bonus(tx
, bigobj
);
3784 txg
= ztest_tx_assign(tx
, TXG_MIGHTWAIT
, FTAG
);
3786 umem_free(packbuf
, packsize
);
3787 umem_free(bigbuf
, bigsize
);
3788 umem_free(od
, size
);
3792 enum zio_checksum cksum
;
3794 cksum
= (enum zio_checksum
)
3795 ztest_random_dsl_prop(ZFS_PROP_CHECKSUM
);
3796 } while (cksum
>= ZIO_CHECKSUM_LEGACY_FUNCTIONS
);
3797 dmu_object_set_checksum(os
, bigobj
, cksum
, tx
);
3799 enum zio_compress comp
;
3801 comp
= (enum zio_compress
)
3802 ztest_random_dsl_prop(ZFS_PROP_COMPRESSION
);
3803 } while (comp
>= ZIO_COMPRESS_LEGACY_FUNCTIONS
);
3804 dmu_object_set_compress(os
, bigobj
, comp
, tx
);
3807 * For each index from n to n + s, verify that the existing bufwad
3808 * in packobj matches the bufwads at the head and tail of the
3809 * corresponding chunk in bigobj. Then update all three bufwads
3810 * with the new values we want to write out.
3812 for (i
= 0; i
< s
; i
++) {
3814 pack
= (bufwad_t
*)((char *)packbuf
+ i
* sizeof (bufwad_t
));
3816 bigH
= (bufwad_t
*)((char *)bigbuf
+ i
* chunksize
);
3818 bigT
= (bufwad_t
*)((char *)bigH
+ chunksize
) - 1;
3820 ASSERT((uintptr_t)bigH
- (uintptr_t)bigbuf
< bigsize
);
3821 ASSERT((uintptr_t)bigT
- (uintptr_t)bigbuf
< bigsize
);
3823 if (pack
->bw_txg
> txg
)
3824 fatal(0, "future leak: got %llx, open txg is %llx",
3827 if (pack
->bw_data
!= 0 && pack
->bw_index
!= n
+ i
)
3828 fatal(0, "wrong index: got %llx, wanted %llx+%llx",
3829 pack
->bw_index
, n
, i
);
3831 if (bcmp(pack
, bigH
, sizeof (bufwad_t
)) != 0)
3832 fatal(0, "pack/bigH mismatch in %p/%p", pack
, bigH
);
3834 if (bcmp(pack
, bigT
, sizeof (bufwad_t
)) != 0)
3835 fatal(0, "pack/bigT mismatch in %p/%p", pack
, bigT
);
3838 bzero(pack
, sizeof (bufwad_t
));
3840 pack
->bw_index
= n
+ i
;
3842 pack
->bw_data
= 1 + ztest_random(-2ULL);
3849 * We've verified all the old bufwads, and made new ones.
3850 * Now write them out.
3852 dmu_write(os
, packobj
, packoff
, packsize
, packbuf
, tx
);
3855 if (ztest_opts
.zo_verbose
>= 7) {
3856 (void) printf("freeing offset %llx size %llx"
3858 (u_longlong_t
)bigoff
,
3859 (u_longlong_t
)bigsize
,
3862 VERIFY(0 == dmu_free_range(os
, bigobj
, bigoff
, bigsize
, tx
));
3864 if (ztest_opts
.zo_verbose
>= 7) {
3865 (void) printf("writing offset %llx size %llx"
3867 (u_longlong_t
)bigoff
,
3868 (u_longlong_t
)bigsize
,
3871 dmu_write(os
, bigobj
, bigoff
, bigsize
, bigbuf
, tx
);
3877 * Sanity check the stuff we just wrote.
3880 void *packcheck
= umem_alloc(packsize
, UMEM_NOFAIL
);
3881 void *bigcheck
= umem_alloc(bigsize
, UMEM_NOFAIL
);
3883 VERIFY(0 == dmu_read(os
, packobj
, packoff
,
3884 packsize
, packcheck
, DMU_READ_PREFETCH
));
3885 VERIFY(0 == dmu_read(os
, bigobj
, bigoff
,
3886 bigsize
, bigcheck
, DMU_READ_PREFETCH
));
3888 ASSERT(bcmp(packbuf
, packcheck
, packsize
) == 0);
3889 ASSERT(bcmp(bigbuf
, bigcheck
, bigsize
) == 0);
3891 umem_free(packcheck
, packsize
);
3892 umem_free(bigcheck
, bigsize
);
3895 umem_free(packbuf
, packsize
);
3896 umem_free(bigbuf
, bigsize
);
3897 umem_free(od
, size
);
3901 compare_and_update_pbbufs(uint64_t s
, bufwad_t
*packbuf
, bufwad_t
*bigbuf
,
3902 uint64_t bigsize
, uint64_t n
, uint64_t chunksize
, uint64_t txg
)
3910 * For each index from n to n + s, verify that the existing bufwad
3911 * in packobj matches the bufwads at the head and tail of the
3912 * corresponding chunk in bigobj. Then update all three bufwads
3913 * with the new values we want to write out.
3915 for (i
= 0; i
< s
; i
++) {
3917 pack
= (bufwad_t
*)((char *)packbuf
+ i
* sizeof (bufwad_t
));
3919 bigH
= (bufwad_t
*)((char *)bigbuf
+ i
* chunksize
);
3921 bigT
= (bufwad_t
*)((char *)bigH
+ chunksize
) - 1;
3923 ASSERT((uintptr_t)bigH
- (uintptr_t)bigbuf
< bigsize
);
3924 ASSERT((uintptr_t)bigT
- (uintptr_t)bigbuf
< bigsize
);
3926 if (pack
->bw_txg
> txg
)
3927 fatal(0, "future leak: got %llx, open txg is %llx",
3930 if (pack
->bw_data
!= 0 && pack
->bw_index
!= n
+ i
)
3931 fatal(0, "wrong index: got %llx, wanted %llx+%llx",
3932 pack
->bw_index
, n
, i
);
3934 if (bcmp(pack
, bigH
, sizeof (bufwad_t
)) != 0)
3935 fatal(0, "pack/bigH mismatch in %p/%p", pack
, bigH
);
3937 if (bcmp(pack
, bigT
, sizeof (bufwad_t
)) != 0)
3938 fatal(0, "pack/bigT mismatch in %p/%p", pack
, bigT
);
3940 pack
->bw_index
= n
+ i
;
3942 pack
->bw_data
= 1 + ztest_random(-2ULL);
3949 #undef OD_ARRAY_SIZE
3950 #define OD_ARRAY_SIZE 2
3953 ztest_dmu_read_write_zcopy(ztest_ds_t
*zd
, uint64_t id
)
3955 objset_t
*os
= zd
->zd_os
;
3962 bufwad_t
*packbuf
, *bigbuf
;
3963 uint64_t packobj
, packoff
, packsize
, bigobj
, bigoff
, bigsize
;
3964 uint64_t blocksize
= ztest_random_blocksize();
3965 uint64_t chunksize
= blocksize
;
3966 uint64_t regions
= 997;
3967 uint64_t stride
= 123456789ULL;
3969 dmu_buf_t
*bonus_db
;
3970 arc_buf_t
**bigbuf_arcbufs
;
3971 dmu_object_info_t doi
;
3973 size
= sizeof (ztest_od_t
) * OD_ARRAY_SIZE
;
3974 od
= umem_alloc(size
, UMEM_NOFAIL
);
3977 * This test uses two objects, packobj and bigobj, that are always
3978 * updated together (i.e. in the same tx) so that their contents are
3979 * in sync and can be compared. Their contents relate to each other
3980 * in a simple way: packobj is a dense array of 'bufwad' structures,
3981 * while bigobj is a sparse array of the same bufwads. Specifically,
3982 * for any index n, there are three bufwads that should be identical:
3984 * packobj, at offset n * sizeof (bufwad_t)
3985 * bigobj, at the head of the nth chunk
3986 * bigobj, at the tail of the nth chunk
3988 * The chunk size is set equal to bigobj block size so that
3989 * dmu_assign_arcbuf() can be tested for object updates.
3993 * Read the directory info. If it's the first time, set things up.
3995 ztest_od_init(od
, id
, FTAG
, 0, DMU_OT_UINT64_OTHER
, blocksize
, 0);
3996 ztest_od_init(od
+ 1, id
, FTAG
, 1, DMU_OT_UINT64_OTHER
, 0, chunksize
);
3999 if (ztest_object_init(zd
, od
, size
, B_FALSE
) != 0) {
4000 umem_free(od
, size
);
4004 bigobj
= od
[0].od_object
;
4005 packobj
= od
[1].od_object
;
4006 blocksize
= od
[0].od_blocksize
;
4007 chunksize
= blocksize
;
4008 ASSERT(chunksize
== od
[1].od_gen
);
4010 VERIFY(dmu_object_info(os
, bigobj
, &doi
) == 0);
4011 VERIFY(ISP2(doi
.doi_data_block_size
));
4012 VERIFY(chunksize
== doi
.doi_data_block_size
);
4013 VERIFY(chunksize
>= 2 * sizeof (bufwad_t
));
4016 * Pick a random index and compute the offsets into packobj and bigobj.
4018 n
= ztest_random(regions
) * stride
+ ztest_random(width
);
4019 s
= 1 + ztest_random(width
- 1);
4021 packoff
= n
* sizeof (bufwad_t
);
4022 packsize
= s
* sizeof (bufwad_t
);
4024 bigoff
= n
* chunksize
;
4025 bigsize
= s
* chunksize
;
4027 packbuf
= umem_zalloc(packsize
, UMEM_NOFAIL
);
4028 bigbuf
= umem_zalloc(bigsize
, UMEM_NOFAIL
);
4030 VERIFY3U(0, ==, dmu_bonus_hold(os
, bigobj
, FTAG
, &bonus_db
));
4032 bigbuf_arcbufs
= umem_zalloc(2 * s
* sizeof (arc_buf_t
*), UMEM_NOFAIL
);
4035 * Iteration 0 test zcopy for DB_UNCACHED dbufs.
4036 * Iteration 1 test zcopy to already referenced dbufs.
4037 * Iteration 2 test zcopy to dirty dbuf in the same txg.
4038 * Iteration 3 test zcopy to dbuf dirty in previous txg.
4039 * Iteration 4 test zcopy when dbuf is no longer dirty.
4040 * Iteration 5 test zcopy when it can't be done.
4041 * Iteration 6 one more zcopy write.
4043 for (i
= 0; i
< 7; i
++) {
4048 * In iteration 5 (i == 5) use arcbufs
4049 * that don't match bigobj blksz to test
4050 * dmu_assign_arcbuf() when it can't directly
4051 * assign an arcbuf to a dbuf.
4053 for (j
= 0; j
< s
; j
++) {
4056 dmu_request_arcbuf(bonus_db
, chunksize
);
4058 bigbuf_arcbufs
[2 * j
] =
4059 dmu_request_arcbuf(bonus_db
, chunksize
/ 2);
4060 bigbuf_arcbufs
[2 * j
+ 1] =
4061 dmu_request_arcbuf(bonus_db
, chunksize
/ 2);
4066 * Get a tx for the mods to both packobj and bigobj.
4068 tx
= dmu_tx_create(os
);
4070 dmu_tx_hold_write(tx
, packobj
, packoff
, packsize
);
4071 dmu_tx_hold_write(tx
, bigobj
, bigoff
, bigsize
);
4073 txg
= ztest_tx_assign(tx
, TXG_MIGHTWAIT
, FTAG
);
4075 umem_free(packbuf
, packsize
);
4076 umem_free(bigbuf
, bigsize
);
4077 for (j
= 0; j
< s
; j
++) {
4079 dmu_return_arcbuf(bigbuf_arcbufs
[j
]);
4082 bigbuf_arcbufs
[2 * j
]);
4084 bigbuf_arcbufs
[2 * j
+ 1]);
4087 umem_free(bigbuf_arcbufs
, 2 * s
* sizeof (arc_buf_t
*));
4088 umem_free(od
, size
);
4089 dmu_buf_rele(bonus_db
, FTAG
);
4094 * 50% of the time don't read objects in the 1st iteration to
4095 * test dmu_assign_arcbuf() for the case when there're no
4096 * existing dbufs for the specified offsets.
4098 if (i
!= 0 || ztest_random(2) != 0) {
4099 error
= dmu_read(os
, packobj
, packoff
,
4100 packsize
, packbuf
, DMU_READ_PREFETCH
);
4102 error
= dmu_read(os
, bigobj
, bigoff
, bigsize
,
4103 bigbuf
, DMU_READ_PREFETCH
);
4106 compare_and_update_pbbufs(s
, packbuf
, bigbuf
, bigsize
,
4110 * We've verified all the old bufwads, and made new ones.
4111 * Now write them out.
4113 dmu_write(os
, packobj
, packoff
, packsize
, packbuf
, tx
);
4114 if (ztest_opts
.zo_verbose
>= 7) {
4115 (void) printf("writing offset %llx size %llx"
4117 (u_longlong_t
)bigoff
,
4118 (u_longlong_t
)bigsize
,
4121 for (off
= bigoff
, j
= 0; j
< s
; j
++, off
+= chunksize
) {
4124 bcopy((caddr_t
)bigbuf
+ (off
- bigoff
),
4125 bigbuf_arcbufs
[j
]->b_data
, chunksize
);
4127 bcopy((caddr_t
)bigbuf
+ (off
- bigoff
),
4128 bigbuf_arcbufs
[2 * j
]->b_data
,
4130 bcopy((caddr_t
)bigbuf
+ (off
- bigoff
) +
4132 bigbuf_arcbufs
[2 * j
+ 1]->b_data
,
4137 VERIFY(dmu_buf_hold(os
, bigobj
, off
,
4138 FTAG
, &dbt
, DMU_READ_NO_PREFETCH
) == 0);
4141 dmu_assign_arcbuf(bonus_db
, off
,
4142 bigbuf_arcbufs
[j
], tx
);
4144 dmu_assign_arcbuf(bonus_db
, off
,
4145 bigbuf_arcbufs
[2 * j
], tx
);
4146 dmu_assign_arcbuf(bonus_db
,
4147 off
+ chunksize
/ 2,
4148 bigbuf_arcbufs
[2 * j
+ 1], tx
);
4151 dmu_buf_rele(dbt
, FTAG
);
4157 * Sanity check the stuff we just wrote.
4160 void *packcheck
= umem_alloc(packsize
, UMEM_NOFAIL
);
4161 void *bigcheck
= umem_alloc(bigsize
, UMEM_NOFAIL
);
4163 VERIFY(0 == dmu_read(os
, packobj
, packoff
,
4164 packsize
, packcheck
, DMU_READ_PREFETCH
));
4165 VERIFY(0 == dmu_read(os
, bigobj
, bigoff
,
4166 bigsize
, bigcheck
, DMU_READ_PREFETCH
));
4168 ASSERT(bcmp(packbuf
, packcheck
, packsize
) == 0);
4169 ASSERT(bcmp(bigbuf
, bigcheck
, bigsize
) == 0);
4171 umem_free(packcheck
, packsize
);
4172 umem_free(bigcheck
, bigsize
);
4175 txg_wait_open(dmu_objset_pool(os
), 0);
4176 } else if (i
== 3) {
4177 txg_wait_synced(dmu_objset_pool(os
), 0);
4181 dmu_buf_rele(bonus_db
, FTAG
);
4182 umem_free(packbuf
, packsize
);
4183 umem_free(bigbuf
, bigsize
);
4184 umem_free(bigbuf_arcbufs
, 2 * s
* sizeof (arc_buf_t
*));
4185 umem_free(od
, size
);
4190 ztest_dmu_write_parallel(ztest_ds_t
*zd
, uint64_t id
)
4194 od
= umem_alloc(sizeof (ztest_od_t
), UMEM_NOFAIL
);
4195 uint64_t offset
= (1ULL << (ztest_random(20) + 43)) +
4196 (ztest_random(ZTEST_RANGE_LOCKS
) << SPA_MAXBLOCKSHIFT
);
4199 * Have multiple threads write to large offsets in an object
4200 * to verify that parallel writes to an object -- even to the
4201 * same blocks within the object -- doesn't cause any trouble.
4203 ztest_od_init(od
, ID_PARALLEL
, FTAG
, 0, DMU_OT_UINT64_OTHER
, 0, 0);
4205 if (ztest_object_init(zd
, od
, sizeof (ztest_od_t
), B_FALSE
) != 0)
4208 while (ztest_random(10) != 0)
4209 ztest_io(zd
, od
->od_object
, offset
);
4211 umem_free(od
, sizeof (ztest_od_t
));
4215 ztest_dmu_prealloc(ztest_ds_t
*zd
, uint64_t id
)
4218 uint64_t offset
= (1ULL << (ztest_random(4) + SPA_MAXBLOCKSHIFT
)) +
4219 (ztest_random(ZTEST_RANGE_LOCKS
) << SPA_MAXBLOCKSHIFT
);
4220 uint64_t count
= ztest_random(20) + 1;
4221 uint64_t blocksize
= ztest_random_blocksize();
4224 od
= umem_alloc(sizeof (ztest_od_t
), UMEM_NOFAIL
);
4226 ztest_od_init(od
, id
, FTAG
, 0, DMU_OT_UINT64_OTHER
, blocksize
, 0);
4228 if (ztest_object_init(zd
, od
, sizeof (ztest_od_t
),
4229 !ztest_random(2)) != 0) {
4230 umem_free(od
, sizeof (ztest_od_t
));
4234 if (ztest_truncate(zd
, od
->od_object
, offset
, count
* blocksize
) != 0) {
4235 umem_free(od
, sizeof (ztest_od_t
));
4239 ztest_prealloc(zd
, od
->od_object
, offset
, count
* blocksize
);
4241 data
= umem_zalloc(blocksize
, UMEM_NOFAIL
);
4243 while (ztest_random(count
) != 0) {
4244 uint64_t randoff
= offset
+ (ztest_random(count
) * blocksize
);
4245 if (ztest_write(zd
, od
->od_object
, randoff
, blocksize
,
4248 while (ztest_random(4) != 0)
4249 ztest_io(zd
, od
->od_object
, randoff
);
4252 umem_free(data
, blocksize
);
4253 umem_free(od
, sizeof (ztest_od_t
));
4257 * Verify that zap_{create,destroy,add,remove,update} work as expected.
4259 #define ZTEST_ZAP_MIN_INTS 1
4260 #define ZTEST_ZAP_MAX_INTS 4
4261 #define ZTEST_ZAP_MAX_PROPS 1000
4264 ztest_zap(ztest_ds_t
*zd
, uint64_t id
)
4266 objset_t
*os
= zd
->zd_os
;
4269 uint64_t txg
, last_txg
;
4270 uint64_t value
[ZTEST_ZAP_MAX_INTS
];
4271 uint64_t zl_ints
, zl_intsize
, prop
;
4274 char propname
[100], txgname
[100];
4276 char *hc
[2] = { "s.acl.h", ".s.open.h.hyLZlg" };
4278 od
= umem_alloc(sizeof (ztest_od_t
), UMEM_NOFAIL
);
4279 ztest_od_init(od
, id
, FTAG
, 0, DMU_OT_ZAP_OTHER
, 0, 0);
4281 if (ztest_object_init(zd
, od
, sizeof (ztest_od_t
),
4282 !ztest_random(2)) != 0)
4285 object
= od
->od_object
;
4288 * Generate a known hash collision, and verify that
4289 * we can lookup and remove both entries.
4291 tx
= dmu_tx_create(os
);
4292 dmu_tx_hold_zap(tx
, object
, B_TRUE
, NULL
);
4293 txg
= ztest_tx_assign(tx
, TXG_MIGHTWAIT
, FTAG
);
4296 for (i
= 0; i
< 2; i
++) {
4298 VERIFY3U(0, ==, zap_add(os
, object
, hc
[i
], sizeof (uint64_t),
4301 for (i
= 0; i
< 2; i
++) {
4302 VERIFY3U(EEXIST
, ==, zap_add(os
, object
, hc
[i
],
4303 sizeof (uint64_t), 1, &value
[i
], tx
));
4305 zap_length(os
, object
, hc
[i
], &zl_intsize
, &zl_ints
));
4306 ASSERT3U(zl_intsize
, ==, sizeof (uint64_t));
4307 ASSERT3U(zl_ints
, ==, 1);
4309 for (i
= 0; i
< 2; i
++) {
4310 VERIFY3U(0, ==, zap_remove(os
, object
, hc
[i
], tx
));
4315 * Generate a buch of random entries.
4317 ints
= MAX(ZTEST_ZAP_MIN_INTS
, object
% ZTEST_ZAP_MAX_INTS
);
4319 prop
= ztest_random(ZTEST_ZAP_MAX_PROPS
);
4320 (void) sprintf(propname
, "prop_%llu", (u_longlong_t
)prop
);
4321 (void) sprintf(txgname
, "txg_%llu", (u_longlong_t
)prop
);
4322 bzero(value
, sizeof (value
));
4326 * If these zap entries already exist, validate their contents.
4328 error
= zap_length(os
, object
, txgname
, &zl_intsize
, &zl_ints
);
4330 ASSERT3U(zl_intsize
, ==, sizeof (uint64_t));
4331 ASSERT3U(zl_ints
, ==, 1);
4333 VERIFY(zap_lookup(os
, object
, txgname
, zl_intsize
,
4334 zl_ints
, &last_txg
) == 0);
4336 VERIFY(zap_length(os
, object
, propname
, &zl_intsize
,
4339 ASSERT3U(zl_intsize
, ==, sizeof (uint64_t));
4340 ASSERT3U(zl_ints
, ==, ints
);
4342 VERIFY(zap_lookup(os
, object
, propname
, zl_intsize
,
4343 zl_ints
, value
) == 0);
4345 for (i
= 0; i
< ints
; i
++) {
4346 ASSERT3U(value
[i
], ==, last_txg
+ object
+ i
);
4349 ASSERT3U(error
, ==, ENOENT
);
4353 * Atomically update two entries in our zap object.
4354 * The first is named txg_%llu, and contains the txg
4355 * in which the property was last updated. The second
4356 * is named prop_%llu, and the nth element of its value
4357 * should be txg + object + n.
4359 tx
= dmu_tx_create(os
);
4360 dmu_tx_hold_zap(tx
, object
, B_TRUE
, NULL
);
4361 txg
= ztest_tx_assign(tx
, TXG_MIGHTWAIT
, FTAG
);
4366 fatal(0, "zap future leak: old %llu new %llu", last_txg
, txg
);
4368 for (i
= 0; i
< ints
; i
++)
4369 value
[i
] = txg
+ object
+ i
;
4371 VERIFY3U(0, ==, zap_update(os
, object
, txgname
, sizeof (uint64_t),
4373 VERIFY3U(0, ==, zap_update(os
, object
, propname
, sizeof (uint64_t),
4379 * Remove a random pair of entries.
4381 prop
= ztest_random(ZTEST_ZAP_MAX_PROPS
);
4382 (void) sprintf(propname
, "prop_%llu", (u_longlong_t
)prop
);
4383 (void) sprintf(txgname
, "txg_%llu", (u_longlong_t
)prop
);
4385 error
= zap_length(os
, object
, txgname
, &zl_intsize
, &zl_ints
);
4387 if (error
== ENOENT
)
4392 tx
= dmu_tx_create(os
);
4393 dmu_tx_hold_zap(tx
, object
, B_TRUE
, NULL
);
4394 txg
= ztest_tx_assign(tx
, TXG_MIGHTWAIT
, FTAG
);
4397 VERIFY3U(0, ==, zap_remove(os
, object
, txgname
, tx
));
4398 VERIFY3U(0, ==, zap_remove(os
, object
, propname
, tx
));
4401 umem_free(od
, sizeof (ztest_od_t
));
4405 * Testcase to test the upgrading of a microzap to fatzap.
4408 ztest_fzap(ztest_ds_t
*zd
, uint64_t id
)
4410 objset_t
*os
= zd
->zd_os
;
4412 uint64_t object
, txg
;
4415 od
= umem_alloc(sizeof (ztest_od_t
), UMEM_NOFAIL
);
4416 ztest_od_init(od
, id
, FTAG
, 0, DMU_OT_ZAP_OTHER
, 0, 0);
4418 if (ztest_object_init(zd
, od
, sizeof (ztest_od_t
),
4419 !ztest_random(2)) != 0)
4421 object
= od
->od_object
;
4424 * Add entries to this ZAP and make sure it spills over
4425 * and gets upgraded to a fatzap. Also, since we are adding
4426 * 2050 entries we should see ptrtbl growth and leaf-block split.
4428 for (i
= 0; i
< 2050; i
++) {
4429 char name
[MAXNAMELEN
];
4434 (void) snprintf(name
, sizeof (name
), "fzap-%llu-%llu",
4435 (u_longlong_t
)id
, (u_longlong_t
)value
);
4437 tx
= dmu_tx_create(os
);
4438 dmu_tx_hold_zap(tx
, object
, B_TRUE
, name
);
4439 txg
= ztest_tx_assign(tx
, TXG_MIGHTWAIT
, FTAG
);
4442 error
= zap_add(os
, object
, name
, sizeof (uint64_t), 1,
4444 ASSERT(error
== 0 || error
== EEXIST
);
4448 umem_free(od
, sizeof (ztest_od_t
));
4453 ztest_zap_parallel(ztest_ds_t
*zd
, uint64_t id
)
4455 objset_t
*os
= zd
->zd_os
;
4457 uint64_t txg
, object
, count
, wsize
, wc
, zl_wsize
, zl_wc
;
4459 int i
, namelen
, error
;
4460 int micro
= ztest_random(2);
4461 char name
[20], string_value
[20];
4464 od
= umem_alloc(sizeof (ztest_od_t
), UMEM_NOFAIL
);
4465 ztest_od_init(od
, ID_PARALLEL
, FTAG
, micro
, DMU_OT_ZAP_OTHER
, 0, 0);
4467 if (ztest_object_init(zd
, od
, sizeof (ztest_od_t
), B_FALSE
) != 0) {
4468 umem_free(od
, sizeof (ztest_od_t
));
4472 object
= od
->od_object
;
4475 * Generate a random name of the form 'xxx.....' where each
4476 * x is a random printable character and the dots are dots.
4477 * There are 94 such characters, and the name length goes from
4478 * 6 to 20, so there are 94^3 * 15 = 12,458,760 possible names.
4480 namelen
= ztest_random(sizeof (name
) - 5) + 5 + 1;
4482 for (i
= 0; i
< 3; i
++)
4483 name
[i
] = '!' + ztest_random('~' - '!' + 1);
4484 for (; i
< namelen
- 1; i
++)
4488 if ((namelen
& 1) || micro
) {
4489 wsize
= sizeof (txg
);
4495 data
= string_value
;
4499 VERIFY0(zap_count(os
, object
, &count
));
4500 ASSERT(count
!= -1ULL);
4503 * Select an operation: length, lookup, add, update, remove.
4505 i
= ztest_random(5);
4508 tx
= dmu_tx_create(os
);
4509 dmu_tx_hold_zap(tx
, object
, B_TRUE
, NULL
);
4510 txg
= ztest_tx_assign(tx
, TXG_MIGHTWAIT
, FTAG
);
4513 bcopy(name
, string_value
, namelen
);
4517 bzero(string_value
, namelen
);
4523 error
= zap_length(os
, object
, name
, &zl_wsize
, &zl_wc
);
4525 ASSERT3U(wsize
, ==, zl_wsize
);
4526 ASSERT3U(wc
, ==, zl_wc
);
4528 ASSERT3U(error
, ==, ENOENT
);
4533 error
= zap_lookup(os
, object
, name
, wsize
, wc
, data
);
4535 if (data
== string_value
&&
4536 bcmp(name
, data
, namelen
) != 0)
4537 fatal(0, "name '%s' != val '%s' len %d",
4538 name
, data
, namelen
);
4540 ASSERT3U(error
, ==, ENOENT
);
4545 error
= zap_add(os
, object
, name
, wsize
, wc
, data
, tx
);
4546 ASSERT(error
== 0 || error
== EEXIST
);
4550 VERIFY(zap_update(os
, object
, name
, wsize
, wc
, data
, tx
) == 0);
4554 error
= zap_remove(os
, object
, name
, tx
);
4555 ASSERT(error
== 0 || error
== ENOENT
);
4562 umem_free(od
, sizeof (ztest_od_t
));
4566 * Commit callback data.
4568 typedef struct ztest_cb_data
{
4569 list_node_t zcd_node
;
4571 int zcd_expected_err
;
4572 boolean_t zcd_added
;
4573 boolean_t zcd_called
;
4577 /* This is the actual commit callback function */
4579 ztest_commit_callback(void *arg
, int error
)
4581 ztest_cb_data_t
*data
= arg
;
4582 uint64_t synced_txg
;
4584 VERIFY(data
!= NULL
);
4585 VERIFY3S(data
->zcd_expected_err
, ==, error
);
4586 VERIFY(!data
->zcd_called
);
4588 synced_txg
= spa_last_synced_txg(data
->zcd_spa
);
4589 if (data
->zcd_txg
> synced_txg
)
4590 fatal(0, "commit callback of txg %" PRIu64
" called prematurely"
4591 ", last synced txg = %" PRIu64
"\n", data
->zcd_txg
,
4594 data
->zcd_called
= B_TRUE
;
4596 if (error
== ECANCELED
) {
4597 ASSERT0(data
->zcd_txg
);
4598 ASSERT(!data
->zcd_added
);
4601 * The private callback data should be destroyed here, but
4602 * since we are going to check the zcd_called field after
4603 * dmu_tx_abort(), we will destroy it there.
4608 ASSERT(data
->zcd_added
);
4609 ASSERT3U(data
->zcd_txg
, !=, 0);
4611 (void) mutex_enter(&zcl
.zcl_callbacks_lock
);
4613 /* See if this cb was called more quickly */
4614 if ((synced_txg
- data
->zcd_txg
) < zc_min_txg_delay
)
4615 zc_min_txg_delay
= synced_txg
- data
->zcd_txg
;
4617 /* Remove our callback from the list */
4618 list_remove(&zcl
.zcl_callbacks
, data
);
4620 (void) mutex_exit(&zcl
.zcl_callbacks_lock
);
4622 umem_free(data
, sizeof (ztest_cb_data_t
));
4625 /* Allocate and initialize callback data structure */
4626 static ztest_cb_data_t
*
4627 ztest_create_cb_data(objset_t
*os
, uint64_t txg
)
4629 ztest_cb_data_t
*cb_data
;
4631 cb_data
= umem_zalloc(sizeof (ztest_cb_data_t
), UMEM_NOFAIL
);
4633 cb_data
->zcd_txg
= txg
;
4634 cb_data
->zcd_spa
= dmu_objset_spa(os
);
4635 list_link_init(&cb_data
->zcd_node
);
4641 * Commit callback test.
4644 ztest_dmu_commit_callbacks(ztest_ds_t
*zd
, uint64_t id
)
4646 objset_t
*os
= zd
->zd_os
;
4649 ztest_cb_data_t
*cb_data
[3], *tmp_cb
;
4650 uint64_t old_txg
, txg
;
4653 od
= umem_alloc(sizeof (ztest_od_t
), UMEM_NOFAIL
);
4654 ztest_od_init(od
, id
, FTAG
, 0, DMU_OT_UINT64_OTHER
, 0, 0);
4656 if (ztest_object_init(zd
, od
, sizeof (ztest_od_t
), B_FALSE
) != 0) {
4657 umem_free(od
, sizeof (ztest_od_t
));
4661 tx
= dmu_tx_create(os
);
4663 cb_data
[0] = ztest_create_cb_data(os
, 0);
4664 dmu_tx_callback_register(tx
, ztest_commit_callback
, cb_data
[0]);
4666 dmu_tx_hold_write(tx
, od
->od_object
, 0, sizeof (uint64_t));
4668 /* Every once in a while, abort the transaction on purpose */
4669 if (ztest_random(100) == 0)
4673 error
= dmu_tx_assign(tx
, TXG_NOWAIT
);
4675 txg
= error
? 0 : dmu_tx_get_txg(tx
);
4677 cb_data
[0]->zcd_txg
= txg
;
4678 cb_data
[1] = ztest_create_cb_data(os
, txg
);
4679 dmu_tx_callback_register(tx
, ztest_commit_callback
, cb_data
[1]);
4683 * It's not a strict requirement to call the registered
4684 * callbacks from inside dmu_tx_abort(), but that's what
4685 * it's supposed to happen in the current implementation
4686 * so we will check for that.
4688 for (i
= 0; i
< 2; i
++) {
4689 cb_data
[i
]->zcd_expected_err
= ECANCELED
;
4690 VERIFY(!cb_data
[i
]->zcd_called
);
4695 for (i
= 0; i
< 2; i
++) {
4696 VERIFY(cb_data
[i
]->zcd_called
);
4697 umem_free(cb_data
[i
], sizeof (ztest_cb_data_t
));
4700 umem_free(od
, sizeof (ztest_od_t
));
4704 cb_data
[2] = ztest_create_cb_data(os
, txg
);
4705 dmu_tx_callback_register(tx
, ztest_commit_callback
, cb_data
[2]);
4708 * Read existing data to make sure there isn't a future leak.
4710 VERIFY(0 == dmu_read(os
, od
->od_object
, 0, sizeof (uint64_t),
4711 &old_txg
, DMU_READ_PREFETCH
));
4714 fatal(0, "future leak: got %" PRIu64
", open txg is %" PRIu64
,
4717 dmu_write(os
, od
->od_object
, 0, sizeof (uint64_t), &txg
, tx
);
4719 (void) mutex_enter(&zcl
.zcl_callbacks_lock
);
4722 * Since commit callbacks don't have any ordering requirement and since
4723 * it is theoretically possible for a commit callback to be called
4724 * after an arbitrary amount of time has elapsed since its txg has been
4725 * synced, it is difficult to reliably determine whether a commit
4726 * callback hasn't been called due to high load or due to a flawed
4729 * In practice, we will assume that if after a certain number of txgs a
4730 * commit callback hasn't been called, then most likely there's an
4731 * implementation bug..
4733 tmp_cb
= list_head(&zcl
.zcl_callbacks
);
4734 if (tmp_cb
!= NULL
&&
4735 tmp_cb
->zcd_txg
+ ZTEST_COMMIT_CB_THRESH
< txg
) {
4736 fatal(0, "Commit callback threshold exceeded, oldest txg: %"
4737 PRIu64
", open txg: %" PRIu64
"\n", tmp_cb
->zcd_txg
, txg
);
4741 * Let's find the place to insert our callbacks.
4743 * Even though the list is ordered by txg, it is possible for the
4744 * insertion point to not be the end because our txg may already be
4745 * quiescing at this point and other callbacks in the open txg
4746 * (from other objsets) may have sneaked in.
4748 tmp_cb
= list_tail(&zcl
.zcl_callbacks
);
4749 while (tmp_cb
!= NULL
&& tmp_cb
->zcd_txg
> txg
)
4750 tmp_cb
= list_prev(&zcl
.zcl_callbacks
, tmp_cb
);
4752 /* Add the 3 callbacks to the list */
4753 for (i
= 0; i
< 3; i
++) {
4755 list_insert_head(&zcl
.zcl_callbacks
, cb_data
[i
]);
4757 list_insert_after(&zcl
.zcl_callbacks
, tmp_cb
,
4760 cb_data
[i
]->zcd_added
= B_TRUE
;
4761 VERIFY(!cb_data
[i
]->zcd_called
);
4763 tmp_cb
= cb_data
[i
];
4768 (void) mutex_exit(&zcl
.zcl_callbacks_lock
);
4772 umem_free(od
, sizeof (ztest_od_t
));
4777 ztest_dsl_prop_get_set(ztest_ds_t
*zd
, uint64_t id
)
4779 zfs_prop_t proplist
[] = {
4781 ZFS_PROP_COMPRESSION
,
4787 (void) rw_rdlock(&ztest_name_lock
);
4789 for (p
= 0; p
< sizeof (proplist
) / sizeof (proplist
[0]); p
++)
4790 (void) ztest_dsl_prop_set_uint64(zd
->zd_name
, proplist
[p
],
4791 ztest_random_dsl_prop(proplist
[p
]), (int)ztest_random(2));
4793 (void) rw_unlock(&ztest_name_lock
);
4798 ztest_spa_prop_get_set(ztest_ds_t
*zd
, uint64_t id
)
4800 nvlist_t
*props
= NULL
;
4802 (void) rw_rdlock(&ztest_name_lock
);
4804 (void) ztest_spa_prop_set_uint64(ZPOOL_PROP_DEDUPDITTO
,
4805 ZIO_DEDUPDITTO_MIN
+ ztest_random(ZIO_DEDUPDITTO_MIN
));
4807 VERIFY0(spa_prop_get(ztest_spa
, &props
));
4809 if (ztest_opts
.zo_verbose
>= 6)
4810 dump_nvlist(props
, 4);
4814 (void) rw_unlock(&ztest_name_lock
);
4818 user_release_one(const char *snapname
, const char *holdname
)
4820 nvlist_t
*snaps
, *holds
;
4823 snaps
= fnvlist_alloc();
4824 holds
= fnvlist_alloc();
4825 fnvlist_add_boolean(holds
, holdname
);
4826 fnvlist_add_nvlist(snaps
, snapname
, holds
);
4827 fnvlist_free(holds
);
4828 error
= dsl_dataset_user_release(snaps
, NULL
);
4829 fnvlist_free(snaps
);
4834 * Test snapshot hold/release and deferred destroy.
4837 ztest_dmu_snapshot_hold(ztest_ds_t
*zd
, uint64_t id
)
4840 objset_t
*os
= zd
->zd_os
;
4844 char clonename
[100];
4846 char osname
[MAXNAMELEN
];
4849 (void) rw_rdlock(&ztest_name_lock
);
4851 dmu_objset_name(os
, osname
);
4853 (void) snprintf(snapname
, sizeof (snapname
), "sh1_%llu",
4855 (void) snprintf(fullname
, sizeof (fullname
), "%s@%s", osname
, snapname
);
4856 (void) snprintf(clonename
, sizeof (clonename
),
4857 "%s/ch1_%llu", osname
, (u_longlong_t
)id
);
4858 (void) snprintf(tag
, sizeof (tag
), "tag_%llu", (u_longlong_t
)id
);
4861 * Clean up from any previous run.
4863 error
= dsl_destroy_head(clonename
);
4864 if (error
!= ENOENT
)
4866 error
= user_release_one(fullname
, tag
);
4867 if (error
!= ESRCH
&& error
!= ENOENT
)
4869 error
= dsl_destroy_snapshot(fullname
, B_FALSE
);
4870 if (error
!= ENOENT
)
4874 * Create snapshot, clone it, mark snap for deferred destroy,
4875 * destroy clone, verify snap was also destroyed.
4877 error
= dmu_objset_snapshot_one(osname
, snapname
);
4879 if (error
== ENOSPC
) {
4880 ztest_record_enospc("dmu_objset_snapshot");
4883 fatal(0, "dmu_objset_snapshot(%s) = %d", fullname
, error
);
4886 error
= dmu_objset_clone(clonename
, fullname
);
4888 if (error
== ENOSPC
) {
4889 ztest_record_enospc("dmu_objset_clone");
4892 fatal(0, "dmu_objset_clone(%s) = %d", clonename
, error
);
4895 error
= dsl_destroy_snapshot(fullname
, B_TRUE
);
4897 fatal(0, "dsl_destroy_snapshot(%s, B_TRUE) = %d",
4901 error
= dsl_destroy_head(clonename
);
4903 fatal(0, "dsl_destroy_head(%s) = %d", clonename
, error
);
4905 error
= dmu_objset_hold(fullname
, FTAG
, &origin
);
4906 if (error
!= ENOENT
)
4907 fatal(0, "dmu_objset_hold(%s) = %d", fullname
, error
);
4910 * Create snapshot, add temporary hold, verify that we can't
4911 * destroy a held snapshot, mark for deferred destroy,
4912 * release hold, verify snapshot was destroyed.
4914 error
= dmu_objset_snapshot_one(osname
, snapname
);
4916 if (error
== ENOSPC
) {
4917 ztest_record_enospc("dmu_objset_snapshot");
4920 fatal(0, "dmu_objset_snapshot(%s) = %d", fullname
, error
);
4923 holds
= fnvlist_alloc();
4924 fnvlist_add_string(holds
, fullname
, tag
);
4925 error
= dsl_dataset_user_hold(holds
, 0, NULL
);
4926 fnvlist_free(holds
);
4928 if (error
== ENOSPC
) {
4929 ztest_record_enospc("dsl_dataset_user_hold");
4932 fatal(0, "dsl_dataset_user_hold(%s, %s) = %u",
4933 fullname
, tag
, error
);
4936 error
= dsl_destroy_snapshot(fullname
, B_FALSE
);
4937 if (error
!= EBUSY
) {
4938 fatal(0, "dsl_destroy_snapshot(%s, B_FALSE) = %d",
4942 error
= dsl_destroy_snapshot(fullname
, B_TRUE
);
4944 fatal(0, "dsl_destroy_snapshot(%s, B_TRUE) = %d",
4948 error
= user_release_one(fullname
, tag
);
4950 fatal(0, "user_release_one(%s, %s) = %d", fullname
, tag
, error
);
4952 VERIFY3U(dmu_objset_hold(fullname
, FTAG
, &origin
), ==, ENOENT
);
4955 (void) rw_unlock(&ztest_name_lock
);
4959 * Inject random faults into the on-disk data.
4963 ztest_fault_inject(ztest_ds_t
*zd
, uint64_t id
)
4965 ztest_shared_t
*zs
= ztest_shared
;
4966 spa_t
*spa
= ztest_spa
;
4970 uint64_t bad
= 0x1990c0ffeedecadeull
;
4975 int bshift
= SPA_MAXBLOCKSHIFT
+ 2; /* don't scrog all labels */
4981 boolean_t islog
= B_FALSE
;
4983 path0
= umem_alloc(MAXPATHLEN
, UMEM_NOFAIL
);
4984 pathrand
= umem_alloc(MAXPATHLEN
, UMEM_NOFAIL
);
4986 mutex_enter(&ztest_vdev_lock
);
4987 maxfaults
= MAXFAULTS();
4988 leaves
= MAX(zs
->zs_mirrors
, 1) * ztest_opts
.zo_raidz
;
4989 mirror_save
= zs
->zs_mirrors
;
4990 mutex_exit(&ztest_vdev_lock
);
4992 ASSERT(leaves
>= 1);
4995 * Grab the name lock as reader. There are some operations
4996 * which don't like to have their vdevs changed while
4997 * they are in progress (i.e. spa_change_guid). Those
4998 * operations will have grabbed the name lock as writer.
5000 (void) rw_rdlock(&ztest_name_lock
);
5003 * We need SCL_STATE here because we're going to look at vd0->vdev_tsd.
5005 spa_config_enter(spa
, SCL_STATE
, FTAG
, RW_READER
);
5007 if (ztest_random(2) == 0) {
5009 * Inject errors on a normal data device or slog device.
5011 top
= ztest_random_vdev_top(spa
, B_TRUE
);
5012 leaf
= ztest_random(leaves
) + zs
->zs_splits
;
5015 * Generate paths to the first leaf in this top-level vdev,
5016 * and to the random leaf we selected. We'll induce transient
5017 * write failures and random online/offline activity on leaf 0,
5018 * and we'll write random garbage to the randomly chosen leaf.
5020 (void) snprintf(path0
, MAXPATHLEN
, ztest_dev_template
,
5021 ztest_opts
.zo_dir
, ztest_opts
.zo_pool
,
5022 top
* leaves
+ zs
->zs_splits
);
5023 (void) snprintf(pathrand
, MAXPATHLEN
, ztest_dev_template
,
5024 ztest_opts
.zo_dir
, ztest_opts
.zo_pool
,
5025 top
* leaves
+ leaf
);
5027 vd0
= vdev_lookup_by_path(spa
->spa_root_vdev
, path0
);
5028 if (vd0
!= NULL
&& vd0
->vdev_top
->vdev_islog
)
5032 * If the top-level vdev needs to be resilvered
5033 * then we only allow faults on the device that is
5036 if (vd0
!= NULL
&& maxfaults
!= 1 &&
5037 (!vdev_resilver_needed(vd0
->vdev_top
, NULL
, NULL
) ||
5038 vd0
->vdev_resilver_txg
!= 0)) {
5040 * Make vd0 explicitly claim to be unreadable,
5041 * or unwriteable, or reach behind its back
5042 * and close the underlying fd. We can do this if
5043 * maxfaults == 0 because we'll fail and reexecute,
5044 * and we can do it if maxfaults >= 2 because we'll
5045 * have enough redundancy. If maxfaults == 1, the
5046 * combination of this with injection of random data
5047 * corruption below exceeds the pool's fault tolerance.
5049 vdev_file_t
*vf
= vd0
->vdev_tsd
;
5051 if (vf
!= NULL
&& ztest_random(3) == 0) {
5052 (void) close(vf
->vf_vnode
->v_fd
);
5053 vf
->vf_vnode
->v_fd
= -1;
5054 } else if (ztest_random(2) == 0) {
5055 vd0
->vdev_cant_read
= B_TRUE
;
5057 vd0
->vdev_cant_write
= B_TRUE
;
5059 guid0
= vd0
->vdev_guid
;
5063 * Inject errors on an l2cache device.
5065 spa_aux_vdev_t
*sav
= &spa
->spa_l2cache
;
5067 if (sav
->sav_count
== 0) {
5068 spa_config_exit(spa
, SCL_STATE
, FTAG
);
5069 (void) rw_unlock(&ztest_name_lock
);
5072 vd0
= sav
->sav_vdevs
[ztest_random(sav
->sav_count
)];
5073 guid0
= vd0
->vdev_guid
;
5074 (void) strcpy(path0
, vd0
->vdev_path
);
5075 (void) strcpy(pathrand
, vd0
->vdev_path
);
5079 maxfaults
= INT_MAX
; /* no limit on cache devices */
5082 spa_config_exit(spa
, SCL_STATE
, FTAG
);
5083 (void) rw_unlock(&ztest_name_lock
);
5086 * If we can tolerate two or more faults, or we're dealing
5087 * with a slog, randomly online/offline vd0.
5089 if ((maxfaults
>= 2 || islog
) && guid0
!= 0) {
5090 if (ztest_random(10) < 6) {
5091 int flags
= (ztest_random(2) == 0 ?
5092 ZFS_OFFLINE_TEMPORARY
: 0);
5095 * We have to grab the zs_name_lock as writer to
5096 * prevent a race between offlining a slog and
5097 * destroying a dataset. Offlining the slog will
5098 * grab a reference on the dataset which may cause
5099 * dsl_destroy_head() to fail with EBUSY thus
5100 * leaving the dataset in an inconsistent state.
5103 (void) rw_wrlock(&ztest_name_lock
);
5105 VERIFY(vdev_offline(spa
, guid0
, flags
) != EBUSY
);
5108 (void) rw_unlock(&ztest_name_lock
);
5111 * Ideally we would like to be able to randomly
5112 * call vdev_[on|off]line without holding locks
5113 * to force unpredictable failures but the side
5114 * effects of vdev_[on|off]line prevent us from
5115 * doing so. We grab the ztest_vdev_lock here to
5116 * prevent a race between injection testing and
5119 mutex_enter(&ztest_vdev_lock
);
5120 (void) vdev_online(spa
, guid0
, 0, NULL
);
5121 mutex_exit(&ztest_vdev_lock
);
5129 * We have at least single-fault tolerance, so inject data corruption.
5131 fd
= open(pathrand
, O_RDWR
);
5133 if (fd
== -1) /* we hit a gap in the device namespace */
5136 fsize
= lseek(fd
, 0, SEEK_END
);
5138 while (--iters
!= 0) {
5139 offset
= ztest_random(fsize
/ (leaves
<< bshift
)) *
5140 (leaves
<< bshift
) + (leaf
<< bshift
) +
5141 (ztest_random(1ULL << (bshift
- 1)) & -8ULL);
5143 if (offset
>= fsize
)
5146 mutex_enter(&ztest_vdev_lock
);
5147 if (mirror_save
!= zs
->zs_mirrors
) {
5148 mutex_exit(&ztest_vdev_lock
);
5153 if (pwrite(fd
, &bad
, sizeof (bad
), offset
) != sizeof (bad
))
5154 fatal(1, "can't inject bad word at 0x%llx in %s",
5157 mutex_exit(&ztest_vdev_lock
);
5159 if (ztest_opts
.zo_verbose
>= 7)
5160 (void) printf("injected bad word into %s,"
5161 " offset 0x%llx\n", pathrand
, (u_longlong_t
)offset
);
5166 umem_free(path0
, MAXPATHLEN
);
5167 umem_free(pathrand
, MAXPATHLEN
);
5171 * Verify that DDT repair works as expected.
5174 ztest_ddt_repair(ztest_ds_t
*zd
, uint64_t id
)
5176 ztest_shared_t
*zs
= ztest_shared
;
5177 spa_t
*spa
= ztest_spa
;
5178 objset_t
*os
= zd
->zd_os
;
5180 uint64_t object
, blocksize
, txg
, pattern
, psize
;
5181 enum zio_checksum checksum
= spa_dedup_checksum(spa
);
5186 int copies
= 2 * ZIO_DEDUPDITTO_MIN
;
5189 blocksize
= ztest_random_blocksize();
5190 blocksize
= MIN(blocksize
, 2048); /* because we write so many */
5192 od
= umem_alloc(sizeof (ztest_od_t
), UMEM_NOFAIL
);
5193 ztest_od_init(od
, id
, FTAG
, 0, DMU_OT_UINT64_OTHER
, blocksize
, 0);
5195 if (ztest_object_init(zd
, od
, sizeof (ztest_od_t
), B_FALSE
) != 0) {
5196 umem_free(od
, sizeof (ztest_od_t
));
5201 * Take the name lock as writer to prevent anyone else from changing
5202 * the pool and dataset properies we need to maintain during this test.
5204 (void) rw_wrlock(&ztest_name_lock
);
5206 if (ztest_dsl_prop_set_uint64(zd
->zd_name
, ZFS_PROP_DEDUP
, checksum
,
5208 ztest_dsl_prop_set_uint64(zd
->zd_name
, ZFS_PROP_COPIES
, 1,
5210 (void) rw_unlock(&ztest_name_lock
);
5211 umem_free(od
, sizeof (ztest_od_t
));
5215 object
= od
[0].od_object
;
5216 blocksize
= od
[0].od_blocksize
;
5217 pattern
= zs
->zs_guid
^ dmu_objset_fsid_guid(os
);
5219 ASSERT(object
!= 0);
5221 tx
= dmu_tx_create(os
);
5222 dmu_tx_hold_write(tx
, object
, 0, copies
* blocksize
);
5223 txg
= ztest_tx_assign(tx
, TXG_WAIT
, FTAG
);
5225 (void) rw_unlock(&ztest_name_lock
);
5226 umem_free(od
, sizeof (ztest_od_t
));
5231 * Write all the copies of our block.
5233 for (i
= 0; i
< copies
; i
++) {
5234 uint64_t offset
= i
* blocksize
;
5235 int error
= dmu_buf_hold(os
, object
, offset
, FTAG
, &db
,
5236 DMU_READ_NO_PREFETCH
);
5238 fatal(B_FALSE
, "dmu_buf_hold(%p, %llu, %llu) = %u",
5239 os
, (long long)object
, (long long) offset
, error
);
5241 ASSERT(db
->db_offset
== offset
);
5242 ASSERT(db
->db_size
== blocksize
);
5243 ASSERT(ztest_pattern_match(db
->db_data
, db
->db_size
, pattern
) ||
5244 ztest_pattern_match(db
->db_data
, db
->db_size
, 0ULL));
5245 dmu_buf_will_fill(db
, tx
);
5246 ztest_pattern_set(db
->db_data
, db
->db_size
, pattern
);
5247 dmu_buf_rele(db
, FTAG
);
5251 txg_wait_synced(spa_get_dsl(spa
), txg
);
5254 * Find out what block we got.
5256 VERIFY0(dmu_buf_hold(os
, object
, 0, FTAG
, &db
,
5257 DMU_READ_NO_PREFETCH
));
5258 blk
= *((dmu_buf_impl_t
*)db
)->db_blkptr
;
5259 dmu_buf_rele(db
, FTAG
);
5262 * Damage the block. Dedup-ditto will save us when we read it later.
5264 psize
= BP_GET_PSIZE(&blk
);
5265 buf
= zio_buf_alloc(psize
);
5266 ztest_pattern_set(buf
, psize
, ~pattern
);
5268 (void) zio_wait(zio_rewrite(NULL
, spa
, 0, &blk
,
5269 buf
, psize
, NULL
, NULL
, ZIO_PRIORITY_SYNC_WRITE
,
5270 ZIO_FLAG_CANFAIL
| ZIO_FLAG_INDUCE_DAMAGE
, NULL
));
5272 zio_buf_free(buf
, psize
);
5274 (void) rw_unlock(&ztest_name_lock
);
5275 umem_free(od
, sizeof (ztest_od_t
));
5283 ztest_scrub(ztest_ds_t
*zd
, uint64_t id
)
5285 spa_t
*spa
= ztest_spa
;
5287 (void) spa_scan(spa
, POOL_SCAN_SCRUB
);
5288 (void) poll(NULL
, 0, 100); /* wait a moment, then force a restart */
5289 (void) spa_scan(spa
, POOL_SCAN_SCRUB
);
5293 * Change the guid for the pool.
5297 ztest_reguid(ztest_ds_t
*zd
, uint64_t id
)
5299 spa_t
*spa
= ztest_spa
;
5300 uint64_t orig
, load
;
5303 orig
= spa_guid(spa
);
5304 load
= spa_load_guid(spa
);
5306 (void) rw_wrlock(&ztest_name_lock
);
5307 error
= spa_change_guid(spa
);
5308 (void) rw_unlock(&ztest_name_lock
);
5313 if (ztest_opts
.zo_verbose
>= 4) {
5314 (void) printf("Changed guid old %llu -> %llu\n",
5315 (u_longlong_t
)orig
, (u_longlong_t
)spa_guid(spa
));
5318 VERIFY3U(orig
, !=, spa_guid(spa
));
5319 VERIFY3U(load
, ==, spa_load_guid(spa
));
5323 * Rename the pool to a different name and then rename it back.
5327 ztest_spa_rename(ztest_ds_t
*zd
, uint64_t id
)
5329 char *oldname
, *newname
;
5332 (void) rw_wrlock(&ztest_name_lock
);
5334 oldname
= ztest_opts
.zo_pool
;
5335 newname
= umem_alloc(strlen(oldname
) + 5, UMEM_NOFAIL
);
5336 (void) strcpy(newname
, oldname
);
5337 (void) strcat(newname
, "_tmp");
5342 VERIFY3U(0, ==, spa_rename(oldname
, newname
));
5345 * Try to open it under the old name, which shouldn't exist
5347 VERIFY3U(ENOENT
, ==, spa_open(oldname
, &spa
, FTAG
));
5350 * Open it under the new name and make sure it's still the same spa_t.
5352 VERIFY3U(0, ==, spa_open(newname
, &spa
, FTAG
));
5354 ASSERT(spa
== ztest_spa
);
5355 spa_close(spa
, FTAG
);
5358 * Rename it back to the original
5360 VERIFY3U(0, ==, spa_rename(newname
, oldname
));
5363 * Make sure it can still be opened
5365 VERIFY3U(0, ==, spa_open(oldname
, &spa
, FTAG
));
5367 ASSERT(spa
== ztest_spa
);
5368 spa_close(spa
, FTAG
);
5370 umem_free(newname
, strlen(newname
) + 1);
5372 (void) rw_unlock(&ztest_name_lock
);
5376 * Verify pool integrity by running zdb.
5379 ztest_run_zdb(char *pool
)
5387 bin
= umem_alloc(MAXPATHLEN
+ MAXNAMELEN
+ 20, UMEM_NOFAIL
);
5388 zdb
= umem_alloc(MAXPATHLEN
+ MAXNAMELEN
+ 20, UMEM_NOFAIL
);
5389 zbuf
= umem_alloc(1024, UMEM_NOFAIL
);
5391 VERIFY(realpath(getexecname(), bin
) != NULL
);
5392 if (strncmp(bin
, "/usr/sbin/ztest", 15) == 0) {
5393 strcpy(bin
, "/usr/sbin/zdb"); /* Installed */
5394 } else if (strncmp(bin
, "/sbin/ztest", 11) == 0) {
5395 strcpy(bin
, "/sbin/zdb"); /* Installed */
5397 strstr(bin
, "/ztest/")[0] = '\0'; /* In-tree */
5398 strcat(bin
, "/zdb/zdb");
5402 "%s -bcc%s%s -d -U %s %s",
5404 ztest_opts
.zo_verbose
>= 3 ? "s" : "",
5405 ztest_opts
.zo_verbose
>= 4 ? "v" : "",
5409 if (ztest_opts
.zo_verbose
>= 5)
5410 (void) printf("Executing %s\n", strstr(zdb
, "zdb "));
5412 fp
= popen(zdb
, "r");
5414 while (fgets(zbuf
, 1024, fp
) != NULL
)
5415 if (ztest_opts
.zo_verbose
>= 3)
5416 (void) printf("%s", zbuf
);
5418 status
= pclose(fp
);
5423 ztest_dump_core
= 0;
5424 if (WIFEXITED(status
))
5425 fatal(0, "'%s' exit code %d", zdb
, WEXITSTATUS(status
));
5427 fatal(0, "'%s' died with signal %d", zdb
, WTERMSIG(status
));
5429 umem_free(bin
, MAXPATHLEN
+ MAXNAMELEN
+ 20);
5430 umem_free(zdb
, MAXPATHLEN
+ MAXNAMELEN
+ 20);
5431 umem_free(zbuf
, 1024);
5435 ztest_walk_pool_directory(char *header
)
5439 if (ztest_opts
.zo_verbose
>= 6)
5440 (void) printf("%s\n", header
);
5442 mutex_enter(&spa_namespace_lock
);
5443 while ((spa
= spa_next(spa
)) != NULL
)
5444 if (ztest_opts
.zo_verbose
>= 6)
5445 (void) printf("\t%s\n", spa_name(spa
));
5446 mutex_exit(&spa_namespace_lock
);
5450 ztest_spa_import_export(char *oldname
, char *newname
)
5452 nvlist_t
*config
, *newconfig
;
5457 if (ztest_opts
.zo_verbose
>= 4) {
5458 (void) printf("import/export: old = %s, new = %s\n",
5463 * Clean up from previous runs.
5465 (void) spa_destroy(newname
);
5468 * Get the pool's configuration and guid.
5470 VERIFY3U(0, ==, spa_open(oldname
, &spa
, FTAG
));
5473 * Kick off a scrub to tickle scrub/export races.
5475 if (ztest_random(2) == 0)
5476 (void) spa_scan(spa
, POOL_SCAN_SCRUB
);
5478 pool_guid
= spa_guid(spa
);
5479 spa_close(spa
, FTAG
);
5481 ztest_walk_pool_directory("pools before export");
5486 VERIFY3U(0, ==, spa_export(oldname
, &config
, B_FALSE
, B_FALSE
));
5488 ztest_walk_pool_directory("pools after export");
5493 newconfig
= spa_tryimport(config
);
5494 ASSERT(newconfig
!= NULL
);
5495 nvlist_free(newconfig
);
5498 * Import it under the new name.
5500 error
= spa_import(newname
, config
, NULL
, 0);
5502 dump_nvlist(config
, 0);
5503 fatal(B_FALSE
, "couldn't import pool %s as %s: error %u",
5504 oldname
, newname
, error
);
5507 ztest_walk_pool_directory("pools after import");
5510 * Try to import it again -- should fail with EEXIST.
5512 VERIFY3U(EEXIST
, ==, spa_import(newname
, config
, NULL
, 0));
5515 * Try to import it under a different name -- should fail with EEXIST.
5517 VERIFY3U(EEXIST
, ==, spa_import(oldname
, config
, NULL
, 0));
5520 * Verify that the pool is no longer visible under the old name.
5522 VERIFY3U(ENOENT
, ==, spa_open(oldname
, &spa
, FTAG
));
5525 * Verify that we can open and close the pool using the new name.
5527 VERIFY3U(0, ==, spa_open(newname
, &spa
, FTAG
));
5528 ASSERT(pool_guid
== spa_guid(spa
));
5529 spa_close(spa
, FTAG
);
5531 nvlist_free(config
);
5535 ztest_resume(spa_t
*spa
)
5537 if (spa_suspended(spa
) && ztest_opts
.zo_verbose
>= 6)
5538 (void) printf("resuming from suspended state\n");
5539 spa_vdev_state_enter(spa
, SCL_NONE
);
5540 vdev_clear(spa
, NULL
);
5541 (void) spa_vdev_state_exit(spa
, NULL
, 0);
5542 (void) zio_resume(spa
);
5546 ztest_resume_thread(void *arg
)
5550 while (!ztest_exiting
) {
5551 if (spa_suspended(spa
))
5553 (void) poll(NULL
, 0, 100);
5565 ztest_deadman_alarm(int sig
)
5567 fatal(0, "failed to complete within %d seconds of deadline", GRACE
);
5572 ztest_execute(int test
, ztest_info_t
*zi
, uint64_t id
)
5574 ztest_ds_t
*zd
= &ztest_ds
[id
% ztest_opts
.zo_datasets
];
5575 ztest_shared_callstate_t
*zc
= ZTEST_GET_SHARED_CALLSTATE(test
);
5576 hrtime_t functime
= gethrtime();
5579 for (i
= 0; i
< zi
->zi_iters
; i
++)
5580 zi
->zi_func(zd
, id
);
5582 functime
= gethrtime() - functime
;
5584 atomic_add_64(&zc
->zc_count
, 1);
5585 atomic_add_64(&zc
->zc_time
, functime
);
5587 if (ztest_opts
.zo_verbose
>= 4)
5588 (void) printf("%6.2f sec in %s\n",
5589 (double)functime
/ NANOSEC
, zi
->zi_funcname
);
5593 ztest_thread(void *arg
)
5596 uint64_t id
= (uintptr_t)arg
;
5597 ztest_shared_t
*zs
= ztest_shared
;
5601 ztest_shared_callstate_t
*zc
;
5603 while ((now
= gethrtime()) < zs
->zs_thread_stop
) {
5605 * See if it's time to force a crash.
5607 if (now
> zs
->zs_thread_kill
)
5611 * If we're getting ENOSPC with some regularity, stop.
5613 if (zs
->zs_enospc_count
> 10)
5617 * Pick a random function to execute.
5619 rand
= ztest_random(ZTEST_FUNCS
);
5620 zi
= &ztest_info
[rand
];
5621 zc
= ZTEST_GET_SHARED_CALLSTATE(rand
);
5622 call_next
= zc
->zc_next
;
5624 if (now
>= call_next
&&
5625 atomic_cas_64(&zc
->zc_next
, call_next
, call_next
+
5626 ztest_random(2 * zi
->zi_interval
[0] + 1)) == call_next
) {
5627 ztest_execute(rand
, zi
, id
);
5637 ztest_dataset_name(char *dsname
, char *pool
, int d
)
5639 (void) snprintf(dsname
, MAXNAMELEN
, "%s/ds_%d", pool
, d
);
5643 ztest_dataset_destroy(int d
)
5645 char name
[MAXNAMELEN
];
5648 ztest_dataset_name(name
, ztest_opts
.zo_pool
, d
);
5650 if (ztest_opts
.zo_verbose
>= 3)
5651 (void) printf("Destroying %s to free up space\n", name
);
5654 * Cleanup any non-standard clones and snapshots. In general,
5655 * ztest thread t operates on dataset (t % zopt_datasets),
5656 * so there may be more than one thing to clean up.
5658 for (t
= d
; t
< ztest_opts
.zo_threads
;
5659 t
+= ztest_opts
.zo_datasets
)
5660 ztest_dsl_dataset_cleanup(name
, t
);
5662 (void) dmu_objset_find(name
, ztest_objset_destroy_cb
, NULL
,
5663 DS_FIND_SNAPSHOTS
| DS_FIND_CHILDREN
);
5667 ztest_dataset_dirobj_verify(ztest_ds_t
*zd
)
5669 uint64_t usedobjs
, dirobjs
, scratch
;
5672 * ZTEST_DIROBJ is the object directory for the entire dataset.
5673 * Therefore, the number of objects in use should equal the
5674 * number of ZTEST_DIROBJ entries, +1 for ZTEST_DIROBJ itself.
5675 * If not, we have an object leak.
5677 * Note that we can only check this in ztest_dataset_open(),
5678 * when the open-context and syncing-context values agree.
5679 * That's because zap_count() returns the open-context value,
5680 * while dmu_objset_space() returns the rootbp fill count.
5682 VERIFY3U(0, ==, zap_count(zd
->zd_os
, ZTEST_DIROBJ
, &dirobjs
));
5683 dmu_objset_space(zd
->zd_os
, &scratch
, &scratch
, &usedobjs
, &scratch
);
5684 ASSERT3U(dirobjs
+ 1, ==, usedobjs
);
5688 ztest_dataset_open(int d
)
5690 ztest_ds_t
*zd
= &ztest_ds
[d
];
5691 uint64_t committed_seq
= ZTEST_GET_SHARED_DS(d
)->zd_seq
;
5694 char name
[MAXNAMELEN
];
5697 ztest_dataset_name(name
, ztest_opts
.zo_pool
, d
);
5699 (void) rw_rdlock(&ztest_name_lock
);
5701 error
= ztest_dataset_create(name
);
5702 if (error
== ENOSPC
) {
5703 (void) rw_unlock(&ztest_name_lock
);
5704 ztest_record_enospc(FTAG
);
5707 ASSERT(error
== 0 || error
== EEXIST
);
5709 VERIFY0(dmu_objset_own(name
, DMU_OST_OTHER
, B_FALSE
, zd
, &os
));
5710 (void) rw_unlock(&ztest_name_lock
);
5712 ztest_zd_init(zd
, ZTEST_GET_SHARED_DS(d
), os
);
5714 zilog
= zd
->zd_zilog
;
5716 if (zilog
->zl_header
->zh_claim_lr_seq
!= 0 &&
5717 zilog
->zl_header
->zh_claim_lr_seq
< committed_seq
)
5718 fatal(0, "missing log records: claimed %llu < committed %llu",
5719 zilog
->zl_header
->zh_claim_lr_seq
, committed_seq
);
5721 ztest_dataset_dirobj_verify(zd
);
5723 zil_replay(os
, zd
, ztest_replay_vector
);
5725 ztest_dataset_dirobj_verify(zd
);
5727 if (ztest_opts
.zo_verbose
>= 6)
5728 (void) printf("%s replay %llu blocks, %llu records, seq %llu\n",
5730 (u_longlong_t
)zilog
->zl_parse_blk_count
,
5731 (u_longlong_t
)zilog
->zl_parse_lr_count
,
5732 (u_longlong_t
)zilog
->zl_replaying_seq
);
5734 zilog
= zil_open(os
, ztest_get_data
);
5736 if (zilog
->zl_replaying_seq
!= 0 &&
5737 zilog
->zl_replaying_seq
< committed_seq
)
5738 fatal(0, "missing log records: replayed %llu < committed %llu",
5739 zilog
->zl_replaying_seq
, committed_seq
);
5745 ztest_dataset_close(int d
)
5747 ztest_ds_t
*zd
= &ztest_ds
[d
];
5749 zil_close(zd
->zd_zilog
);
5750 dmu_objset_disown(zd
->zd_os
, zd
);
5756 * Kick off threads to run tests on all datasets in parallel.
5759 ztest_run(ztest_shared_t
*zs
)
5764 kthread_t
*resume_thread
;
5769 ztest_exiting
= B_FALSE
;
5772 * Initialize parent/child shared state.
5774 mutex_init(&ztest_vdev_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
5775 VERIFY(rwlock_init(&ztest_name_lock
, USYNC_THREAD
, NULL
) == 0);
5777 zs
->zs_thread_start
= gethrtime();
5778 zs
->zs_thread_stop
=
5779 zs
->zs_thread_start
+ ztest_opts
.zo_passtime
* NANOSEC
;
5780 zs
->zs_thread_stop
= MIN(zs
->zs_thread_stop
, zs
->zs_proc_stop
);
5781 zs
->zs_thread_kill
= zs
->zs_thread_stop
;
5782 if (ztest_random(100) < ztest_opts
.zo_killrate
) {
5783 zs
->zs_thread_kill
-=
5784 ztest_random(ztest_opts
.zo_passtime
* NANOSEC
);
5787 mutex_init(&zcl
.zcl_callbacks_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
5789 list_create(&zcl
.zcl_callbacks
, sizeof (ztest_cb_data_t
),
5790 offsetof(ztest_cb_data_t
, zcd_node
));
5795 kernel_init(FREAD
| FWRITE
);
5796 VERIFY0(spa_open(ztest_opts
.zo_pool
, &spa
, FTAG
));
5797 spa
->spa_debug
= B_TRUE
;
5798 metaslab_preload_limit
= ztest_random(20) + 1;
5801 VERIFY0(dmu_objset_own(ztest_opts
.zo_pool
,
5802 DMU_OST_ANY
, B_TRUE
, FTAG
, &os
));
5803 zs
->zs_guid
= dmu_objset_fsid_guid(os
);
5804 dmu_objset_disown(os
, FTAG
);
5806 spa
->spa_dedup_ditto
= 2 * ZIO_DEDUPDITTO_MIN
;
5809 * We don't expect the pool to suspend unless maxfaults == 0,
5810 * in which case ztest_fault_inject() temporarily takes away
5811 * the only valid replica.
5813 if (MAXFAULTS() == 0)
5814 spa
->spa_failmode
= ZIO_FAILURE_MODE_WAIT
;
5816 spa
->spa_failmode
= ZIO_FAILURE_MODE_PANIC
;
5819 * Create a thread to periodically resume suspended I/O.
5821 VERIFY3P((resume_thread
= zk_thread_create(NULL
, 0,
5822 (thread_func_t
)ztest_resume_thread
, spa
, TS_RUN
, NULL
, 0, 0,
5823 PTHREAD_CREATE_JOINABLE
)), !=, NULL
);
5827 * Set a deadman alarm to abort() if we hang.
5829 signal(SIGALRM
, ztest_deadman_alarm
);
5830 alarm((zs
->zs_thread_stop
- zs
->zs_thread_start
) / NANOSEC
+ GRACE
);
5834 * Verify that we can safely inquire about about any object,
5835 * whether it's allocated or not. To make it interesting,
5836 * we probe a 5-wide window around each power of two.
5837 * This hits all edge cases, including zero and the max.
5839 for (t
= 0; t
< 64; t
++) {
5840 for (d
= -5; d
<= 5; d
++) {
5841 error
= dmu_object_info(spa
->spa_meta_objset
,
5842 (1ULL << t
) + d
, NULL
);
5843 ASSERT(error
== 0 || error
== ENOENT
||
5849 * If we got any ENOSPC errors on the previous run, destroy something.
5851 if (zs
->zs_enospc_count
!= 0) {
5852 int d
= ztest_random(ztest_opts
.zo_datasets
);
5853 ztest_dataset_destroy(d
);
5855 zs
->zs_enospc_count
= 0;
5857 tid
= umem_zalloc(ztest_opts
.zo_threads
* sizeof (kt_did_t
),
5860 if (ztest_opts
.zo_verbose
>= 4)
5861 (void) printf("starting main threads...\n");
5864 * Kick off all the tests that run in parallel.
5866 for (t
= 0; t
< ztest_opts
.zo_threads
; t
++) {
5869 if (t
< ztest_opts
.zo_datasets
&&
5870 ztest_dataset_open(t
) != 0)
5873 VERIFY3P(thread
= zk_thread_create(NULL
, 0,
5874 (thread_func_t
)ztest_thread
,
5875 (void *)(uintptr_t)t
, TS_RUN
, NULL
, 0, 0,
5876 PTHREAD_CREATE_JOINABLE
), !=, NULL
);
5877 tid
[t
] = thread
->t_tid
;
5881 * Wait for all of the tests to complete. We go in reverse order
5882 * so we don't close datasets while threads are still using them.
5884 for (t
= ztest_opts
.zo_threads
- 1; t
>= 0; t
--) {
5885 thread_join(tid
[t
]);
5886 if (t
< ztest_opts
.zo_datasets
)
5887 ztest_dataset_close(t
);
5890 txg_wait_synced(spa_get_dsl(spa
), 0);
5892 zs
->zs_alloc
= metaslab_class_get_alloc(spa_normal_class(spa
));
5893 zs
->zs_space
= metaslab_class_get_space(spa_normal_class(spa
));
5895 if (ztest_opts
.zo_verbose
>= 3)
5896 zfs_dbgmsg_print(FTAG
);
5898 umem_free(tid
, ztest_opts
.zo_threads
* sizeof (kt_did_t
));
5900 /* Kill the resume thread */
5901 ztest_exiting
= B_TRUE
;
5902 thread_join(resume_thread
->t_tid
);
5906 * Right before closing the pool, kick off a bunch of async I/O;
5907 * spa_close() should wait for it to complete.
5909 for (object
= 1; object
< 50; object
++)
5910 dmu_prefetch(spa
->spa_meta_objset
, object
, 0, 1ULL << 20);
5912 /* Verify that at least one commit cb was called in a timely fashion */
5913 if (zc_cb_counter
>= ZTEST_COMMIT_CB_MIN_REG
)
5914 VERIFY0(zc_min_txg_delay
);
5916 spa_close(spa
, FTAG
);
5919 * Verify that we can loop over all pools.
5921 mutex_enter(&spa_namespace_lock
);
5922 for (spa
= spa_next(NULL
); spa
!= NULL
; spa
= spa_next(spa
))
5923 if (ztest_opts
.zo_verbose
> 3)
5924 (void) printf("spa_next: found %s\n", spa_name(spa
));
5925 mutex_exit(&spa_namespace_lock
);
5928 * Verify that we can export the pool and reimport it under a
5931 if (ztest_random(2) == 0) {
5932 char name
[MAXNAMELEN
];
5933 (void) snprintf(name
, MAXNAMELEN
, "%s_import",
5934 ztest_opts
.zo_pool
);
5935 ztest_spa_import_export(ztest_opts
.zo_pool
, name
);
5936 ztest_spa_import_export(name
, ztest_opts
.zo_pool
);
5941 list_destroy(&zcl
.zcl_callbacks
);
5942 mutex_destroy(&zcl
.zcl_callbacks_lock
);
5943 (void) rwlock_destroy(&ztest_name_lock
);
5944 mutex_destroy(&ztest_vdev_lock
);
5950 ztest_ds_t
*zd
= &ztest_ds
[0];
5954 if (ztest_opts
.zo_verbose
>= 3)
5955 (void) printf("testing spa_freeze()...\n");
5957 kernel_init(FREAD
| FWRITE
);
5958 VERIFY3U(0, ==, spa_open(ztest_opts
.zo_pool
, &spa
, FTAG
));
5959 VERIFY3U(0, ==, ztest_dataset_open(0));
5960 spa
->spa_debug
= B_TRUE
;
5964 * Force the first log block to be transactionally allocated.
5965 * We have to do this before we freeze the pool -- otherwise
5966 * the log chain won't be anchored.
5968 while (BP_IS_HOLE(&zd
->zd_zilog
->zl_header
->zh_log
)) {
5969 ztest_dmu_object_alloc_free(zd
, 0);
5970 zil_commit(zd
->zd_zilog
, 0);
5973 txg_wait_synced(spa_get_dsl(spa
), 0);
5976 * Freeze the pool. This stops spa_sync() from doing anything,
5977 * so that the only way to record changes from now on is the ZIL.
5982 * Because it is hard to predict how much space a write will actually
5983 * require beforehand, we leave ourselves some fudge space to write over
5986 uint64_t capacity
= metaslab_class_get_space(spa_normal_class(spa
)) / 2;
5989 * Run tests that generate log records but don't alter the pool config
5990 * or depend on DSL sync tasks (snapshots, objset create/destroy, etc).
5991 * We do a txg_wait_synced() after each iteration to force the txg
5992 * to increase well beyond the last synced value in the uberblock.
5993 * The ZIL should be OK with that.
5995 * Run a random number of times less than zo_maxloops and ensure we do
5996 * not run out of space on the pool.
5998 while (ztest_random(10) != 0 &&
5999 numloops
++ < ztest_opts
.zo_maxloops
&&
6000 metaslab_class_get_alloc(spa_normal_class(spa
)) < capacity
) {
6002 ztest_od_init(&od
, 0, FTAG
, 0, DMU_OT_UINT64_OTHER
, 0, 0);
6003 VERIFY0(ztest_object_init(zd
, &od
, sizeof (od
), B_FALSE
));
6004 ztest_io(zd
, od
.od_object
,
6005 ztest_random(ZTEST_RANGE_LOCKS
) << SPA_MAXBLOCKSHIFT
);
6006 txg_wait_synced(spa_get_dsl(spa
), 0);
6010 * Commit all of the changes we just generated.
6012 zil_commit(zd
->zd_zilog
, 0);
6013 txg_wait_synced(spa_get_dsl(spa
), 0);
6016 * Close our dataset and close the pool.
6018 ztest_dataset_close(0);
6019 spa_close(spa
, FTAG
);
6023 * Open and close the pool and dataset to induce log replay.
6025 kernel_init(FREAD
| FWRITE
);
6026 VERIFY3U(0, ==, spa_open(ztest_opts
.zo_pool
, &spa
, FTAG
));
6027 ASSERT(spa_freeze_txg(spa
) == UINT64_MAX
);
6028 VERIFY3U(0, ==, ztest_dataset_open(0));
6029 ztest_dataset_close(0);
6031 spa
->spa_debug
= B_TRUE
;
6033 txg_wait_synced(spa_get_dsl(spa
), 0);
6034 ztest_reguid(NULL
, 0);
6036 spa_close(spa
, FTAG
);
6041 print_time(hrtime_t t
, char *timebuf
)
6043 hrtime_t s
= t
/ NANOSEC
;
6044 hrtime_t m
= s
/ 60;
6045 hrtime_t h
= m
/ 60;
6046 hrtime_t d
= h
/ 24;
6055 (void) sprintf(timebuf
,
6056 "%llud%02lluh%02llum%02llus", d
, h
, m
, s
);
6058 (void) sprintf(timebuf
, "%lluh%02llum%02llus", h
, m
, s
);
6060 (void) sprintf(timebuf
, "%llum%02llus", m
, s
);
6062 (void) sprintf(timebuf
, "%llus", s
);
6066 make_random_props(void)
6070 VERIFY(nvlist_alloc(&props
, NV_UNIQUE_NAME
, 0) == 0);
6071 if (ztest_random(2) == 0)
6073 VERIFY(nvlist_add_uint64(props
, "autoreplace", 1) == 0);
6079 * Create a storage pool with the given name and initial vdev size.
6080 * Then test spa_freeze() functionality.
6083 ztest_init(ztest_shared_t
*zs
)
6086 nvlist_t
*nvroot
, *props
;
6089 mutex_init(&ztest_vdev_lock
, NULL
, MUTEX_DEFAULT
, NULL
);
6090 VERIFY(rwlock_init(&ztest_name_lock
, USYNC_THREAD
, NULL
) == 0);
6092 kernel_init(FREAD
| FWRITE
);
6095 * Create the storage pool.
6097 (void) spa_destroy(ztest_opts
.zo_pool
);
6098 ztest_shared
->zs_vdev_next_leaf
= 0;
6100 zs
->zs_mirrors
= ztest_opts
.zo_mirrors
;
6101 nvroot
= make_vdev_root(NULL
, NULL
, NULL
, ztest_opts
.zo_vdev_size
, 0,
6102 0, ztest_opts
.zo_raidz
, zs
->zs_mirrors
, 1);
6103 props
= make_random_props();
6104 for (i
= 0; i
< SPA_FEATURES
; i
++) {
6106 VERIFY3S(-1, !=, asprintf(&buf
, "feature@%s",
6107 spa_feature_table
[i
].fi_uname
));
6108 VERIFY3U(0, ==, nvlist_add_uint64(props
, buf
, 0));
6111 VERIFY3U(0, ==, spa_create(ztest_opts
.zo_pool
, nvroot
, props
, NULL
));
6112 nvlist_free(nvroot
);
6115 VERIFY3U(0, ==, spa_open(ztest_opts
.zo_pool
, &spa
, FTAG
));
6116 zs
->zs_metaslab_sz
=
6117 1ULL << spa
->spa_root_vdev
->vdev_child
[0]->vdev_ms_shift
;
6118 spa_close(spa
, FTAG
);
6122 ztest_run_zdb(ztest_opts
.zo_pool
);
6126 ztest_run_zdb(ztest_opts
.zo_pool
);
6128 (void) rwlock_destroy(&ztest_name_lock
);
6129 mutex_destroy(&ztest_vdev_lock
);
6135 static char ztest_name_data
[] = "/tmp/ztest.data.XXXXXX";
6137 ztest_fd_data
= mkstemp(ztest_name_data
);
6138 ASSERT3S(ztest_fd_data
, >=, 0);
6139 (void) unlink(ztest_name_data
);
6143 shared_data_size(ztest_shared_hdr_t
*hdr
)
6147 size
= hdr
->zh_hdr_size
;
6148 size
+= hdr
->zh_opts_size
;
6149 size
+= hdr
->zh_size
;
6150 size
+= hdr
->zh_stats_size
* hdr
->zh_stats_count
;
6151 size
+= hdr
->zh_ds_size
* hdr
->zh_ds_count
;
6160 ztest_shared_hdr_t
*hdr
;
6162 hdr
= (void *)mmap(0, P2ROUNDUP(sizeof (*hdr
), getpagesize()),
6163 PROT_READ
| PROT_WRITE
, MAP_SHARED
, ztest_fd_data
, 0);
6164 VERIFY3P(hdr
, !=, MAP_FAILED
);
6166 VERIFY3U(0, ==, ftruncate(ztest_fd_data
, sizeof (ztest_shared_hdr_t
)));
6168 hdr
->zh_hdr_size
= sizeof (ztest_shared_hdr_t
);
6169 hdr
->zh_opts_size
= sizeof (ztest_shared_opts_t
);
6170 hdr
->zh_size
= sizeof (ztest_shared_t
);
6171 hdr
->zh_stats_size
= sizeof (ztest_shared_callstate_t
);
6172 hdr
->zh_stats_count
= ZTEST_FUNCS
;
6173 hdr
->zh_ds_size
= sizeof (ztest_shared_ds_t
);
6174 hdr
->zh_ds_count
= ztest_opts
.zo_datasets
;
6176 size
= shared_data_size(hdr
);
6177 VERIFY3U(0, ==, ftruncate(ztest_fd_data
, size
));
6179 (void) munmap((caddr_t
)hdr
, P2ROUNDUP(sizeof (*hdr
), getpagesize()));
6186 ztest_shared_hdr_t
*hdr
;
6189 hdr
= (void *)mmap(0, P2ROUNDUP(sizeof (*hdr
), getpagesize()),
6190 PROT_READ
, MAP_SHARED
, ztest_fd_data
, 0);
6191 VERIFY3P(hdr
, !=, MAP_FAILED
);
6193 size
= shared_data_size(hdr
);
6195 (void) munmap((caddr_t
)hdr
, P2ROUNDUP(sizeof (*hdr
), getpagesize()));
6196 hdr
= ztest_shared_hdr
= (void *)mmap(0, P2ROUNDUP(size
, getpagesize()),
6197 PROT_READ
| PROT_WRITE
, MAP_SHARED
, ztest_fd_data
, 0);
6198 VERIFY3P(hdr
, !=, MAP_FAILED
);
6199 buf
= (uint8_t *)hdr
;
6201 offset
= hdr
->zh_hdr_size
;
6202 ztest_shared_opts
= (void *)&buf
[offset
];
6203 offset
+= hdr
->zh_opts_size
;
6204 ztest_shared
= (void *)&buf
[offset
];
6205 offset
+= hdr
->zh_size
;
6206 ztest_shared_callstate
= (void *)&buf
[offset
];
6207 offset
+= hdr
->zh_stats_size
* hdr
->zh_stats_count
;
6208 ztest_shared_ds
= (void *)&buf
[offset
];
6212 exec_child(char *cmd
, char *libpath
, boolean_t ignorekill
, int *statusp
)
6216 char *cmdbuf
= NULL
;
6221 cmdbuf
= umem_alloc(MAXPATHLEN
, UMEM_NOFAIL
);
6222 (void) strlcpy(cmdbuf
, getexecname(), MAXPATHLEN
);
6227 fatal(1, "fork failed");
6229 if (pid
== 0) { /* child */
6230 char *emptyargv
[2] = { cmd
, NULL
};
6231 char fd_data_str
[12];
6233 struct rlimit rl
= { 1024, 1024 };
6234 (void) setrlimit(RLIMIT_NOFILE
, &rl
);
6236 (void) close(ztest_fd_rand
);
6237 VERIFY(11 >= snprintf(fd_data_str
, 12, "%d", ztest_fd_data
));
6238 VERIFY(0 == setenv("ZTEST_FD_DATA", fd_data_str
, 1));
6240 (void) enable_extended_FILE_stdio(-1, -1);
6241 if (libpath
!= NULL
)
6242 VERIFY(0 == setenv("LD_LIBRARY_PATH", libpath
, 1));
6243 (void) execv(cmd
, emptyargv
);
6244 ztest_dump_core
= B_FALSE
;
6245 fatal(B_TRUE
, "exec failed: %s", cmd
);
6248 if (cmdbuf
!= NULL
) {
6249 umem_free(cmdbuf
, MAXPATHLEN
);
6253 while (waitpid(pid
, &status
, 0) != pid
)
6255 if (statusp
!= NULL
)
6258 if (WIFEXITED(status
)) {
6259 if (WEXITSTATUS(status
) != 0) {
6260 (void) fprintf(stderr
, "child exited with code %d\n",
6261 WEXITSTATUS(status
));
6265 } else if (WIFSIGNALED(status
)) {
6266 if (!ignorekill
|| WTERMSIG(status
) != SIGKILL
) {
6267 (void) fprintf(stderr
, "child died with signal %d\n",
6273 (void) fprintf(stderr
, "something strange happened to child\n");
6280 ztest_run_init(void)
6284 ztest_shared_t
*zs
= ztest_shared
;
6286 ASSERT(ztest_opts
.zo_init
!= 0);
6289 * Blow away any existing copy of zpool.cache
6291 (void) remove(spa_config_path
);
6294 * Create and initialize our storage pool.
6296 for (i
= 1; i
<= ztest_opts
.zo_init
; i
++) {
6297 bzero(zs
, sizeof (ztest_shared_t
));
6298 if (ztest_opts
.zo_verbose
>= 3 &&
6299 ztest_opts
.zo_init
!= 1) {
6300 (void) printf("ztest_init(), pass %d\n", i
);
6307 main(int argc
, char **argv
)
6315 ztest_shared_callstate_t
*zc
;
6322 char *fd_data_str
= getenv("ZTEST_FD_DATA");
6323 struct sigaction action
;
6325 (void) setvbuf(stdout
, NULL
, _IOLBF
, 0);
6327 dprintf_setup(&argc
, argv
);
6329 action
.sa_handler
= sig_handler
;
6330 sigemptyset(&action
.sa_mask
);
6331 action
.sa_flags
= 0;
6333 if (sigaction(SIGSEGV
, &action
, NULL
) < 0) {
6334 (void) fprintf(stderr
, "ztest: cannot catch SIGSEGV: %s.\n",
6339 if (sigaction(SIGABRT
, &action
, NULL
) < 0) {
6340 (void) fprintf(stderr
, "ztest: cannot catch SIGABRT: %s.\n",
6345 ztest_fd_rand
= open("/dev/urandom", O_RDONLY
);
6346 ASSERT3S(ztest_fd_rand
, >=, 0);
6349 process_options(argc
, argv
);
6354 bcopy(&ztest_opts
, ztest_shared_opts
,
6355 sizeof (*ztest_shared_opts
));
6357 ztest_fd_data
= atoi(fd_data_str
);
6359 bcopy(ztest_shared_opts
, &ztest_opts
, sizeof (ztest_opts
));
6361 ASSERT3U(ztest_opts
.zo_datasets
, ==, ztest_shared_hdr
->zh_ds_count
);
6363 /* Override location of zpool.cache */
6364 VERIFY(asprintf((char **)&spa_config_path
, "%s/zpool.cache",
6365 ztest_opts
.zo_dir
) != -1);
6367 ztest_ds
= umem_alloc(ztest_opts
.zo_datasets
* sizeof (ztest_ds_t
),
6372 metaslab_gang_bang
= ztest_opts
.zo_metaslab_gang_bang
;
6373 metaslab_df_alloc_threshold
=
6374 zs
->zs_metaslab_df_alloc_threshold
;
6383 hasalt
= (strlen(ztest_opts
.zo_alt_ztest
) != 0);
6385 if (ztest_opts
.zo_verbose
>= 1) {
6386 (void) printf("%llu vdevs, %d datasets, %d threads,"
6387 " %llu seconds...\n",
6388 (u_longlong_t
)ztest_opts
.zo_vdevs
,
6389 ztest_opts
.zo_datasets
,
6390 ztest_opts
.zo_threads
,
6391 (u_longlong_t
)ztest_opts
.zo_time
);
6394 cmd
= umem_alloc(MAXNAMELEN
, UMEM_NOFAIL
);
6395 (void) strlcpy(cmd
, getexecname(), MAXNAMELEN
);
6397 zs
->zs_do_init
= B_TRUE
;
6398 if (strlen(ztest_opts
.zo_alt_ztest
) != 0) {
6399 if (ztest_opts
.zo_verbose
>= 1) {
6400 (void) printf("Executing older ztest for "
6401 "initialization: %s\n", ztest_opts
.zo_alt_ztest
);
6403 VERIFY(!exec_child(ztest_opts
.zo_alt_ztest
,
6404 ztest_opts
.zo_alt_libpath
, B_FALSE
, NULL
));
6406 VERIFY(!exec_child(NULL
, NULL
, B_FALSE
, NULL
));
6408 zs
->zs_do_init
= B_FALSE
;
6410 zs
->zs_proc_start
= gethrtime();
6411 zs
->zs_proc_stop
= zs
->zs_proc_start
+ ztest_opts
.zo_time
* NANOSEC
;
6413 for (f
= 0; f
< ZTEST_FUNCS
; f
++) {
6414 zi
= &ztest_info
[f
];
6415 zc
= ZTEST_GET_SHARED_CALLSTATE(f
);
6416 if (zs
->zs_proc_start
+ zi
->zi_interval
[0] > zs
->zs_proc_stop
)
6417 zc
->zc_next
= UINT64_MAX
;
6419 zc
->zc_next
= zs
->zs_proc_start
+
6420 ztest_random(2 * zi
->zi_interval
[0] + 1);
6424 * Run the tests in a loop. These tests include fault injection
6425 * to verify that self-healing data works, and forced crashes
6426 * to verify that we never lose on-disk consistency.
6428 while (gethrtime() < zs
->zs_proc_stop
) {
6433 * Initialize the workload counters for each function.
6435 for (f
= 0; f
< ZTEST_FUNCS
; f
++) {
6436 zc
= ZTEST_GET_SHARED_CALLSTATE(f
);
6441 /* Set the allocation switch size */
6442 zs
->zs_metaslab_df_alloc_threshold
=
6443 ztest_random(zs
->zs_metaslab_sz
/ 4) + 1;
6445 if (!hasalt
|| ztest_random(2) == 0) {
6446 if (hasalt
&& ztest_opts
.zo_verbose
>= 1) {
6447 (void) printf("Executing newer ztest: %s\n",
6451 killed
= exec_child(cmd
, NULL
, B_TRUE
, &status
);
6453 if (hasalt
&& ztest_opts
.zo_verbose
>= 1) {
6454 (void) printf("Executing older ztest: %s\n",
6455 ztest_opts
.zo_alt_ztest
);
6458 killed
= exec_child(ztest_opts
.zo_alt_ztest
,
6459 ztest_opts
.zo_alt_libpath
, B_TRUE
, &status
);
6466 if (ztest_opts
.zo_verbose
>= 1) {
6467 hrtime_t now
= gethrtime();
6469 now
= MIN(now
, zs
->zs_proc_stop
);
6470 print_time(zs
->zs_proc_stop
- now
, timebuf
);
6471 nicenum(zs
->zs_space
, numbuf
);
6473 (void) printf("Pass %3d, %8s, %3llu ENOSPC, "
6474 "%4.1f%% of %5s used, %3.0f%% done, %8s to go\n",
6476 WIFEXITED(status
) ? "Complete" : "SIGKILL",
6477 (u_longlong_t
)zs
->zs_enospc_count
,
6478 100.0 * zs
->zs_alloc
/ zs
->zs_space
,
6480 100.0 * (now
- zs
->zs_proc_start
) /
6481 (ztest_opts
.zo_time
* NANOSEC
), timebuf
);
6484 if (ztest_opts
.zo_verbose
>= 2) {
6485 (void) printf("\nWorkload summary:\n\n");
6486 (void) printf("%7s %9s %s\n",
6487 "Calls", "Time", "Function");
6488 (void) printf("%7s %9s %s\n",
6489 "-----", "----", "--------");
6490 for (f
= 0; f
< ZTEST_FUNCS
; f
++) {
6491 zi
= &ztest_info
[f
];
6492 zc
= ZTEST_GET_SHARED_CALLSTATE(f
);
6493 print_time(zc
->zc_time
, timebuf
);
6494 (void) printf("%7llu %9s %s\n",
6495 (u_longlong_t
)zc
->zc_count
, timebuf
,
6498 (void) printf("\n");
6502 * It's possible that we killed a child during a rename test,
6503 * in which case we'll have a 'ztest_tmp' pool lying around
6504 * instead of 'ztest'. Do a blind rename in case this happened.
6507 if (spa_open(ztest_opts
.zo_pool
, &spa
, FTAG
) == 0) {
6508 spa_close(spa
, FTAG
);
6510 char tmpname
[MAXNAMELEN
];
6512 kernel_init(FREAD
| FWRITE
);
6513 (void) snprintf(tmpname
, sizeof (tmpname
), "%s_tmp",
6514 ztest_opts
.zo_pool
);
6515 (void) spa_rename(tmpname
, ztest_opts
.zo_pool
);
6519 ztest_run_zdb(ztest_opts
.zo_pool
);
6522 if (ztest_opts
.zo_verbose
>= 1) {
6524 (void) printf("%d runs of older ztest: %s\n", older
,
6525 ztest_opts
.zo_alt_ztest
);
6526 (void) printf("%d runs of newer ztest: %s\n", newer
,
6529 (void) printf("%d killed, %d completed, %.0f%% kill rate\n",
6530 kills
, iters
- kills
, (100.0 * kills
) / MAX(1, iters
));
6533 umem_free(cmd
, MAXNAMELEN
);