ztest.c revision a7a845e4bf22fd1b2a284729ccd95c7370a0438c
fa9e4066f08beec538e775443c5be79dd423fcabahrens * CDDL HEADER START
fa9e4066f08beec538e775443c5be79dd423fcabahrens * The contents of this file are subject to the terms of the
441d80aa4f613b6298fc8bd3151f4be02dbf84fclling * Common Development and Distribution License (the "License").
441d80aa4f613b6298fc8bd3151f4be02dbf84fclling * You may not use this file except in compliance with the License.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
fa9e4066f08beec538e775443c5be79dd423fcabahrens * See the License for the specific language governing permissions
fa9e4066f08beec538e775443c5be79dd423fcabahrens * and limitations under the License.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * When distributing Covered Code, include this CDDL HEADER in each
fa9e4066f08beec538e775443c5be79dd423fcabahrens * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * If applicable, add the following below this CDDL HEADER, with the
fa9e4066f08beec538e775443c5be79dd423fcabahrens * fields enclosed by brackets "[]" replaced with your own identifying
fa9e4066f08beec538e775443c5be79dd423fcabahrens * information: Portions Copyright [yyyy] [name of copyright owner]
fa9e4066f08beec538e775443c5be79dd423fcabahrens * CDDL HEADER END
47cb52daa729f19e298c85a84e8df069365c5232Jeff Bonwick * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Copyright (c) 2012 by Delphix. All rights reserved.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Copyright (c) 2013 Steven Hartland. All rights reserved.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * The objective of this program is to provide a DMU/ZAP/SPA stress test
fa9e4066f08beec538e775443c5be79dd423fcabahrens * that runs entirely in userland, is easy to use, and easy to extend.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * The overall design of the ztest program is as follows:
fa9e4066f08beec538e775443c5be79dd423fcabahrens * (1) For each major functional area (e.g. adding vdevs to a pool,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * creating and destroying datasets, reading and writing objects, etc)
fa9e4066f08beec538e775443c5be79dd423fcabahrens * we have a simple routine to test that functionality. These
fa9e4066f08beec538e775443c5be79dd423fcabahrens * individual routines do not have to do anything "stressful".
fa9e4066f08beec538e775443c5be79dd423fcabahrens * (2) We turn these simple functionality tests into a stress test by
fa9e4066f08beec538e775443c5be79dd423fcabahrens * running them all in parallel, with as many threads as desired,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * and spread across as many datasets, objects, and vdevs as desired.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * (3) While all this is happening, we inject faults into the pool to
fa9e4066f08beec538e775443c5be79dd423fcabahrens * verify that self-healing data really works.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * (4) Every time we open a dataset, we change its checksum and compression
fa9e4066f08beec538e775443c5be79dd423fcabahrens * functions. Thus even individual objects vary from block to block
fa9e4066f08beec538e775443c5be79dd423fcabahrens * in which checksum they use and whether they're compressed.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * (5) To verify that we never lose on-disk consistency after a crash,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * we run the entire test in a child of the main process.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * At random times, the child self-immolates with a SIGKILL.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * This is the software equivalent of pulling the power cord.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * The parent then runs the test again, using the existing
fa9e4066f08beec538e775443c5be79dd423fcabahrens * storage pool, as many times as desired. If backwards compatability
fa9e4066f08beec538e775443c5be79dd423fcabahrens * testing is enabled ztest will sometimes run the "older" version
fa9e4066f08beec538e775443c5be79dd423fcabahrens * of ztest after a SIGKILL.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * (6) To verify that we don't have future leaks or temporal incursions,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * many of the functional tests record the transaction group number
fa9e4066f08beec538e775443c5be79dd423fcabahrens * as part of their data. When reading old data, they verify that
fa9e4066f08beec538e775443c5be79dd423fcabahrens * the transaction group number is less than the current, open txg.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * If you add a new test, please do this if applicable.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * When run with no arguments, ztest runs for about five minutes and
fa9e4066f08beec538e775443c5be79dd423fcabahrens * produces no output if successful. To get a little bit of information,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * specify -V. To get more information, specify -VV, and so on.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * To turn this into an overnight stress test, use -T to specify run time.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * You can ask more more vdevs [-v], datasets [-d], or threads [-t]
fa9e4066f08beec538e775443c5be79dd423fcabahrens * to increase the pool capacity, fanout, and overall stress level.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Use the -k option to set the desired frequency of kills.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * When ztest invokes itself it passes all relevant information through a
fa9e4066f08beec538e775443c5be79dd423fcabahrens * temporary file which is mmap-ed in the child process. This allows shared
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * memory to survive the exec syscall. The ztest_shared_hdr_t struct is always
fa9e4066f08beec538e775443c5be79dd423fcabahrens * stored at offset 0 of this file and contains information on the size and
fa9e4066f08beec538e775443c5be79dd423fcabahrens * number of shared structures in the file. The information stored in this file
fa9e4066f08beec538e775443c5be79dd423fcabahrens * must remain backwards compatible with older versions of ztest so that
fa9e4066f08beec538e775443c5be79dd423fcabahrens * ztest can invoke them during backwards compatibility testing (-B).
fa9e4066f08beec538e775443c5be79dd423fcabahrenstypedef struct ztest_shared_hdr {
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickstatic const ztest_shared_opts_t ztest_opts_defaults = {
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick .zo_maxloops = 50, /* max loops during spa_freeze() */
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwicktypedef struct ztest_shared_ds {
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick#define ZTEST_GET_SHARED_DS(d) (&ztest_shared_ds[d])
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick (MAX(zs->zs_mirrors, 1) * (ztest_opts.zo_raidz_parity + 1) - 1)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwicktypedef struct ztest_block_tag {
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwicktypedef struct bufwad {
fa9e4066f08beec538e775443c5be79dd423fcabahrens * XXX -- fix zfs range locks to be generic so we can use them here.
fa9e4066f08beec538e775443c5be79dd423fcabahrenstypedef enum {
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwicktypedef struct rll {
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwicktypedef struct rl {
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick * Object descriptor. Used as a template for object lookup/create/remove.
fa9e4066f08beec538e775443c5be79dd423fcabahrenstypedef struct ztest_od {
e05725b117836db173257fae43fb0746eb857fb5bonwick * Per-dataset state.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwicktypedef struct ztest_ds {
5c987a37e6561d090231fb3c10ba9f6ea995f34dChris Kirby * Per-iteration state.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwicktypedef void ztest_func_t(ztest_ds_t *zd, uint64_t id);
3f9d6ad73e45c6823b409f93b0c8d4f62861d2d5Lin Lingtypedef struct ztest_info {
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick uint64_t *zi_interval; /* execute every <interval> seconds */
fa9e4066f08beec538e775443c5be79dd423fcabahrenstypedef struct ztest_shared_callstate {
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia uint64_t zc_next; /* next time to call this function */
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correiastatic ztest_shared_callstate_t *ztest_shared_callstate;
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia#define ZTEST_GET_SHARED_CALLSTATE(c) (&ztest_shared_callstate[c])
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia * Note: these aren't static because we want dladdr() to work.
fa9e4066f08beec538e775443c5be79dd423fcabahrensuint64_t zopt_always = 0ULL * NANOSEC; /* all the time */
fa9e4066f08beec538e775443c5be79dd423fcabahrensuint64_t zopt_incessant = 1ULL * NANOSEC / 10; /* every 1/10 second */
fa9e4066f08beec538e775443c5be79dd423fcabahrensuint64_t zopt_often = 1ULL * NANOSEC; /* every second */
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwickuint64_t zopt_sometimes = 10ULL * NANOSEC; /* every 10 seconds */
0a4e9518a44f226be6d39383330b5b1792d2f184gwuint64_t zopt_rarely = 60ULL * NANOSEC; /* every 60 seconds */
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia { ztest_dmu_read_write, 1, &zopt_always },
d6e555bdd793b8bc8fe57d5f12c3d69c813d0661George Wilson { ztest_dmu_object_alloc_free, 1, &zopt_always },
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick { ztest_dmu_commit_callbacks, 1, &zopt_always },
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick { ztest_dmu_read_write_zcopy, 1, &zopt_often },
fa9e4066f08beec538e775443c5be79dd423fcabahrens { ztest_dmu_snapshot_create_destroy, 1, &zopt_sometimes },
fa9e4066f08beec538e775443c5be79dd423fcabahrens#define ZTEST_FUNCS (sizeof (ztest_info) / sizeof (ztest_info_t))
fa9e4066f08beec538e775443c5be79dd423fcabahrens * The following struct is used to hold a list of uncalled commit callbacks.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * The callbacks are ordered by txg number.
fa9e4066f08beec538e775443c5be79dd423fcabahrenstypedef struct ztest_cb_list {
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Stuff we need to share writably between parent and child.
fa9e4066f08beec538e775443c5be79dd423fcabahrenstypedef struct ztest_shared {
f1b4288b3b466065edd5eb9868399db6a4ff283evb * The ztest_name_lock protects the pool and dataset namespace used by
fa9e4066f08beec538e775443c5be79dd423fcabahrens * the individual tests. To modify the namespace, consumers must grab
fa9e4066f08beec538e775443c5be79dd423fcabahrens * this lock as writer. Grabbing the lock as reader will ensure that the
fa9e4066f08beec538e775443c5be79dd423fcabahrens * namespace does not change while the lock is held.
fa9e4066f08beec538e775443c5be79dd423fcabahrens/* Global commit callback list */
fa9e4066f08beec538e775443c5be79dd423fcabahrens * These libumem hooks provide a reasonable set of defaults for the allocator's
fa9e4066f08beec538e775443c5be79dd423fcabahrens * debugging facilities.
f1b4288b3b466065edd5eb9868399db6a4ff283evbconst char *
fa9e4066f08beec538e775443c5be79dd423fcabahrensconst char *
fa9e4066f08beec538e775443c5be79dd423fcabahrens return ("fail,contents"); /* $UMEM_LOGGING setting */
fa9e4066f08beec538e775443c5be79dd423fcabahrens /* LINTED */
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) snprintf(buf + strlen(buf), FATAL_MSG_SZ - strlen(buf),
0a4e9518a44f226be6d39383330b5b1792d2f184gw return (0);
0a4e9518a44f226be6d39383330b5b1792d2f184gw (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n",
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (buf[1] == '\0' || (toupper(buf[1]) == 'B' && buf[2] == '\0')) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (10*i);
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) fprintf(stderr, "ztest: invalid bytes suffix: %s\n", buf);
fa9e4066f08beec538e775443c5be79dd423fcabahrens /* NOTREACHED */
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock (void) fprintf(stderr, "ztest: bad numeric value: %s\n", buf);
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock (void) fprintf(stderr, "ztest: value too large: %s\n",
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) fprintf(stderr, "ztest: value too large: %s\n",
fa9e4066f08beec538e775443c5be79dd423fcabahrens const ztest_shared_opts_t *zo = &ztest_opts_defaults;
99653d4ee642c6528e88224f12409a5f23060994eschrock nicenum(zo->zo_metaslab_gang_bang, nice_gang_bang);
fa9e4066f08beec538e775443c5be79dd423fcabahrens "\t[-v vdevs (default: %llu)]\n"
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock "\t[-s size_of_each_vdev (default: %s)]\n"
fa9e4066f08beec538e775443c5be79dd423fcabahrens "\t[-a alignment_shift (default: %d)] use 0 for random\n"
fa9e4066f08beec538e775443c5be79dd423fcabahrens "\t[-m mirror_copies (default: %d)]\n"
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock "\t[-r raidz_disks (default: %d)]\n"
e05725b117836db173257fae43fb0746eb857fb5bonwick "\t[-R raidz_parity (default: %d)]\n"
fa9e4066f08beec538e775443c5be79dd423fcabahrens "\t[-d datasets (default: %d)]\n"
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock "\t[-t threads (default: %d)]\n"
fa9e4066f08beec538e775443c5be79dd423fcabahrens "\t[-g gang_block_threshold (default: %s)]\n"
fa9e4066f08beec538e775443c5be79dd423fcabahrens "\t[-i init_count (default: %d)] initialize pool i times\n"
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock "\t[-k kill_percentage (default: %llu%%)]\n"
fa9e4066f08beec538e775443c5be79dd423fcabahrens "\t[-p pool_name (default: %s)]\n"
fa9e4066f08beec538e775443c5be79dd423fcabahrens "\t[-f dir (default: %s)] file directory for vdev files\n"
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock "\t[-V] verbose (use multiple times for ever more blather)\n"
fa9e4066f08beec538e775443c5be79dd423fcabahrens "\t[-E] use existing pool instead of creating new one\n"
fa9e4066f08beec538e775443c5be79dd423fcabahrens "\t[-T time (default: %llu sec)] total run time\n"
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock "\t[-F freezeloops (default: %llu)] max loops in spa_freeze()\n"
fa9e4066f08beec538e775443c5be79dd423fcabahrens "\t[-P passtime (default: %llu sec)] time per pass\n"
fa9e4066f08beec538e775443c5be79dd423fcabahrens "\t[-B alt_ztest (default: <none>)] alternate ztest path\n"
3d7072f8bd27709dba14f6fe336f149d25d9e207eschrock "\t[-h] (print help)\n"
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick "v:s:a:m:r:R:d:t:g:i:k:p:f:VET:P:hF:B:")) != EOF) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens zo->zo_metaslab_gang_bang = MAX(SPA_MINBLOCKSIZE << 1,
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick (void) strlcpy(altdir, optarg, sizeof (altdir));
fa9e4066f08beec538e775443c5be79dd423fcabahrens zo->zo_raidz_parity = MIN(zo->zo_raidz_parity, zo->zo_raidz - 1);
fa9e4066f08beec538e775443c5be79dd423fcabahrens (zo->zo_vdevs > 0 ? zo->zo_time * NANOSEC / zo->zo_vdevs :
fa9e4066f08beec538e775443c5be79dd423fcabahrens * 'cmd' should be of the form "<anything>/usr/bin/<isa>/ztest".
fa9e4066f08beec538e775443c5be79dd423fcabahrens * We want to extract <isa> to determine if we should use
fa9e4066f08beec538e775443c5be79dd423fcabahrens * 32 or 64 bit binaries.
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) snprintf(zo->zo_alt_ztest, sizeof (zo->zo_alt_ztest),
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick (void) snprintf(zo->zo_alt_libpath, sizeof (zo->zo_alt_libpath),
fa9e4066f08beec538e775443c5be79dd423fcabahrens zs->zs_alloc = metaslab_class_get_alloc(spa_normal_class(ztest_spa));
fa9e4066f08beec538e775443c5be79dd423fcabahrens zs->zs_space = metaslab_class_get_space(spa_normal_class(ztest_spa));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick if (read(ztest_fd_rand, &r, sizeof (r)) != sizeof (r))
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick return (r % range);
fa9e4066f08beec538e775443c5be79dd423fcabahrens/* ARGSUSED */
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickmake_vdev_file(char *path, char *aux, char *pool, size_t size, uint64_t ashift)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick pool == NULL ? ztest_opts.zo_pool : pool, vdev);
92241e0b80813d0b83c08e730a29b9d1831794fcTom Erickson int fd = open(path, O_RDWR | O_CREAT | O_TRUNC, 0666);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(nvlist_alloc(&file, NV_UNIQUE_NAME, 0) == 0);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(nvlist_add_string(file, ZPOOL_CONFIG_TYPE, VDEV_TYPE_FILE) == 0);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(nvlist_add_string(file, ZPOOL_CONFIG_PATH, path) == 0);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(nvlist_add_uint64(file, ZPOOL_CONFIG_ASHIFT, ashift) == 0);
fa9e4066f08beec538e775443c5be79dd423fcabahrensmake_vdev_raidz(char *path, char *aux, char *pool, size_t size,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick return (make_vdev_file(path, aux, pool, size, ashift));
fa9e4066f08beec538e775443c5be79dd423fcabahrens child = umem_alloc(r * sizeof (nvlist_t *), UMEM_NOFAIL);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick for (c = 0; c < r; c++)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick child[c] = make_vdev_file(path, aux, pool, size, ashift);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(nvlist_alloc(&raidz, NV_UNIQUE_NAME, 0) == 0);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(nvlist_add_string(raidz, ZPOOL_CONFIG_TYPE,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(nvlist_add_uint64(raidz, ZPOOL_CONFIG_NPARITY,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(nvlist_add_nvlist_array(raidz, ZPOOL_CONFIG_CHILDREN,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick for (c = 0; c < r; c++)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickmake_vdev_mirror(char *path, char *aux, char *pool, size_t size,
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (make_vdev_raidz(path, aux, pool, size, ashift, r));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick child = umem_alloc(m * sizeof (nvlist_t *), UMEM_NOFAIL);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick for (c = 0; c < m; c++)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick child[c] = make_vdev_raidz(path, aux, pool, size, ashift, r);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(nvlist_alloc(&mirror, NV_UNIQUE_NAME, 0) == 0);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(nvlist_add_string(mirror, ZPOOL_CONFIG_TYPE,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(nvlist_add_nvlist_array(mirror, ZPOOL_CONFIG_CHILDREN,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick for (c = 0; c < m; c++)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickmake_vdev_root(char *path, char *aux, char *pool, size_t size, uint64_t ashift,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick int log, int r, int m, int t)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick child = umem_alloc(t * sizeof (nvlist_t *), UMEM_NOFAIL);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick for (c = 0; c < t; c++) {
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick child[c] = make_vdev_mirror(path, aux, pool, size, ashift,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(nvlist_add_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(nvlist_alloc(&root, NV_UNIQUE_NAME, 0) == 0);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(nvlist_add_string(root, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT) == 0);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(nvlist_add_nvlist_array(root, aux ? aux : ZPOOL_CONFIG_CHILDREN,
fa9e4066f08beec538e775443c5be79dd423fcabahrens for (c = 0; c < t; c++)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Find a random spa version. Returns back a random spa version in the
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * range [initial_version, SPA_VERSION_FEATURES].
31157203d871bc0b4bc67d925b0a8424570774a1Jeff Bonwickztest_random_spa_version(uint64_t initial_version)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ztest_random(SPA_VERSION_BEFORE_FEATURES - version + 1);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ztest_random(SPA_MAXBLOCKSHIFT - SPA_MINBLOCKSHIFT + 1)));
fa9e4066f08beec538e775443c5be79dd423fcabahrens ztest_random(DN_MAX_INDBLKSHIFT - DN_MIN_INDBLKSHIFT + 1));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_random_vdev_top(spa_t *spa, boolean_t log_ok)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ASSERT(spa_config_held(spa, SCL_ALL, RW_READER) != 0);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick } while (tvd->vdev_ishole || (tvd->vdev_islog && !log_ok) ||
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick tvd->vdev_mg == NULL || tvd->vdev_mg->mg_class == NULL);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick value = zfs_prop_random_value(prop, ztest_random(-1ULL));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick } while (prop == ZFS_PROP_CHECKSUM && value == ZIO_CHECKSUM_OFF);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_dsl_prop_set_uint64(char *osname, zfs_prop_t prop, uint64_t value,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick (inherit ? ZPROP_SRC_NONE : ZPROP_SRC_LOCAL), value);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY0(dsl_prop_get_integer(osname, propname, &curval, setpoint));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(zfs_prop_index_to_string(prop, curval, &valname) == 0);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_spa_prop_set_uint64(zpool_prop_t prop, uint64_t value)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(nvlist_alloc(&props, NV_UNIQUE_NAME, 0) == 0);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(nvlist_add_uint64(props, zpool_prop_to_name(prop), value) == 0);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(_mutex_init(&rll->rll_lock, USYNC_THREAD, NULL) == 0);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(cond_init(&rll->rll_cv, USYNC_THREAD, NULL) == 0);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick (void) cond_wait(&rll->rll_cv, &rll->rll_lock);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick while (rll->rll_writer != NULL || rll->rll_readers)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick (void) cond_wait(&rll->rll_cv, &rll->rll_lock);
5002558f6bfef3915c7f3b4ecb7c19c7f044bf5bNeil Perrin if (rll->rll_writer == NULL && rll->rll_readers == 0)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_object_lock(ztest_ds_t *zd, uint64_t object, rl_type_t type)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)];
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_object_unlock(ztest_ds_t *zd, uint64_t object)
5002558f6bfef3915c7f3b4ecb7c19c7f044bf5bNeil Perrin rll_t *rll = &zd->zd_object_lock[object & (ZTEST_OBJECT_LOCKS - 1)];
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_range_lock(ztest_ds_t *zd, uint64_t object, uint64_t offset,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick uint64_t hash = object ^ (offset % (ZTEST_RANGE_LOCKS + 1));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick rll_t *rll = &zd->zd_range_lock[hash & (ZTEST_RANGE_LOCKS - 1)];
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_zd_init(ztest_ds_t *zd, ztest_shared_ds_t *szd, objset_t *os)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(rwlock_init(&zd->zd_zilog_lock, USYNC_THREAD, NULL) == 0);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(_mutex_init(&zd->zd_dirobj_lock, USYNC_THREAD, NULL) == 0);
5002558f6bfef3915c7f3b4ecb7c19c7f044bf5bNeil Perrin for (int l = 0; l < ZTEST_OBJECT_LOCKS; l++)
5002558f6bfef3915c7f3b4ecb7c19c7f044bf5bNeil Perrin for (int l = 0; l < ZTEST_RANGE_LOCKS; l++)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(_mutex_destroy(&zd->zd_dirobj_lock) == 0);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick for (int l = 0; l < ZTEST_OBJECT_LOCKS; l++)
5002558f6bfef3915c7f3b4ecb7c19c7f044bf5bNeil Perrin for (int l = 0; l < ZTEST_RANGE_LOCKS; l++)
5002558f6bfef3915c7f3b4ecb7c19c7f044bf5bNeil Perrin#define TXG_MIGHTWAIT (ztest_random(10) == 0 ? TXG_NOWAIT : TXG_WAIT)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_tx_assign(dmu_tx_t *tx, uint64_t txg_how, const char *tag)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Attempt to assign tx to some transaction group.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_pattern_set(void *buf, uint64_t size, uint64_t value)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick uint64_t *ip_end = (uint64_t *)((uintptr_t)buf + (uintptr_t)size);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_pattern_match(void *buf, uint64_t size, uint64_t value)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick uint64_t *ip_end = (uint64_t *)((uintptr_t)buf + (uintptr_t)size);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick return (diff == 0);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_bt_generate(ztest_block_tag_t *bt, objset_t *os, uint64_t object,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick uint64_t offset, uint64_t gen, uint64_t txg, uint64_t crtxg)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_bt_verify(ztest_block_tag_t *bt, objset_t *os, uint64_t object,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick uint64_t offset, uint64_t gen, uint64_t txg, uint64_t crtxg)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ASSERT3U(doi.doi_bonus_size, >=, sizeof (*bt));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick bt = (void *)((char *)db->db_data + doi.doi_bonus_size - sizeof (*bt));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * ZIL logging ops
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_log_create(ztest_ds_t *zd, dmu_tx_t *tx, lr_create_t *lr)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick char *name = (void *)(lr + 1); /* name follows lr */
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick itx = zil_itx_create(TX_CREATE, sizeof (*lr) + namesize);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_log_remove(ztest_ds_t *zd, dmu_tx_t *tx, lr_remove_t *lr, uint64_t object)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick char *name = (void *)(lr + 1); /* name follows lr */
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick itx = zil_itx_create(TX_REMOVE, sizeof (*lr) + namesize);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_log_write(ztest_ds_t *zd, dmu_tx_t *tx, lr_write_t *lr)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick itx_wr_state_t write_state = ztest_random(WR_NUM_STATES);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick sizeof (*lr) + (write_state == WR_COPIED ? lr->lr_length : 0));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick dmu_read(zd->zd_os, lr->lr_foid, lr->lr_offset, lr->lr_length,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ((lr_write_t *)&itx->itx_lr) + 1, DMU_READ_NO_PREFETCH) != 0) {
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick itx->itx_sod += (write_state == WR_NEED_COPY ? lr->lr_length : 0);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_log_truncate(ztest_ds_t *zd, dmu_tx_t *tx, lr_truncate_t *lr)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick itx = zil_itx_create(TX_TRUNCATE, sizeof (*lr));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_log_setattr(ztest_ds_t *zd, dmu_tx_t *tx, lr_setattr_t *lr)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick itx = zil_itx_create(TX_SETATTR, sizeof (*lr));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * ZIL replay ops
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_replay_create(ztest_ds_t *zd, lr_create_t *lr, boolean_t byteswap)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick char *name = (void *)(lr + 1); /* name follows lr */
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick dmu_tx_hold_zap(tx, lr->lr_doid, B_TRUE, name);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, B_TRUE, NULL);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ASSERT(dmu_objset_zil(os)->zl_replay == !!lr->lr_foid);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY3U(0, ==, dmu_object_set_blocksize(os, lr->lr_foid,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ztest_bt_generate(bbt, os, lr->lr_foid, -1ULL, lr->lr_gen, txg, txg);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY3U(0, ==, zap_add(os, lr->lr_doid, name, sizeof (uint64_t), 1,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_replay_remove(ztest_ds_t *zd, lr_remove_t *lr, boolean_t byteswap)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick char *name = (void *)(lr + 1); /* name follows lr */
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick zap_lookup(os, lr->lr_doid, name, sizeof (object), 1, &object));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY3U(0, ==, dmu_object_info(os, object, &doi));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick dmu_tx_hold_zap(tx, lr->lr_doid, B_FALSE, name);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick dmu_tx_hold_free(tx, object, 0, DMU_OBJECT_END);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY3U(0, ==, dmu_object_free(os, object, tx));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY3U(0, ==, zap_remove(os, lr->lr_doid, name, tx));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_replay_write(ztest_ds_t *zd, lr_write_t *lr, boolean_t byteswap)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick /* If it's a dmu_sync() block, write the whole block */
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) {
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick rl = ztest_range_lock(zd, lr->lr_foid, offset, length, RL_WRITER);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick dmu_tx_hold_write(tx, lr->lr_foid, offset, length);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick if (ztest_random(8) == 0 && length == doi.doi_data_block_size &&
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Usually, verify the old data before writing new data --
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * but not always, because we also want to verify correct
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * behavior when the data was not recently read into cache.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Writes can appear to be newer than the bonus buffer because
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * the ztest_get_data() callback does a dmu_read() of the
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * open-context data, which may be different than the data
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * as it was when the write was generated.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Set the bt's gen/txg to the bonus buffer's gen/txg
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * so that all of the usual ASSERTs will work.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ztest_bt_generate(bt, os, lr->lr_foid, offset, gen, txg, crtxg);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick dmu_write(os, lr->lr_foid, offset, length, data, tx);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_replay_truncate(ztest_ds_t *zd, lr_truncate_t *lr, boolean_t byteswap)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick rl = ztest_range_lock(zd, lr->lr_foid, lr->lr_offset, lr->lr_length,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick dmu_tx_hold_free(tx, lr->lr_foid, lr->lr_offset, lr->lr_length);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(dmu_free_range(os, lr->lr_foid, lr->lr_offset,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_replay_setattr(ztest_ds_t *zd, lr_setattr_t *lr, boolean_t byteswap)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY3U(0, ==, dmu_bonus_hold(os, lr->lr_foid, FTAG, &db));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Randomly change the size and increment the generation.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick lr->lr_size = (ztest_random(db->db_size / sizeof (*bbt)) + 1) *
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Verify that the current bonus buffer is not newer than our txg.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ztest_bt_verify(bbt, os, lr->lr_foid, -1ULL, lr->lr_mode,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ztest_bt_generate(bbt, os, lr->lr_foid, -1ULL, lr->lr_mode, txg, crtxg);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickzil_replay_func_t *ztest_replay_vector[TX_MAX_TYPE] = {
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * ZIL get_data callbacks
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick zgd->zgd_rl = ztest_range_lock(zd, object, offset, size,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick error = dmu_read(os, object, offset, size, buf,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick zgd->zgd_rl = ztest_range_lock(zd, object, offset, size,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick error = dmu_buf_hold(os, object, offset, zgd, &db,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickstatic void *
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick lr = umem_zalloc(lrsize + namesize, UMEM_NOFAIL);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_lr_free(void *lr, size_t lrsize, char *name)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Lookup a bunch of objects. Returns the number of objects not found.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_lookup(ztest_ds_t *zd, ztest_od_t *od, int count)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick error = zap_lookup(zd->zd_os, od->od_dir, od->od_name,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ASSERT(missing == 0); /* there should be no gaps */
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ztest_object_lock(zd, od->od_object, RL_READER);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_create(ztest_ds_t *zd, ztest_od_t *od, int count)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick lr_create_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick lr->lr_foid = 0; /* 0 to allocate, > 0 to claim */
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick if (ztest_replay_create(zd, lr, B_FALSE) != 0) {
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_remove(ztest_ds_t *zd, ztest_od_t *od, int count)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * No object was found.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick lr_remove_t *lr = ztest_lr_alloc(sizeof (*lr), od->od_name);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick if ((error = ztest_replay_remove(zd, lr, B_FALSE)) != 0) {
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_write(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick lr = ztest_lr_alloc(sizeof (*lr) + size, NULL);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_truncate(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick error = ztest_replay_truncate(zd, lr, B_FALSE);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_prealloc(ztest_ds_t *zd, uint64_t object, uint64_t offset, uint64_t size)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick rl = ztest_range_lock(zd, object, offset, size, RL_WRITER);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick (void) dmu_free_long_range(os, object, offset, size);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_io(ztest_ds_t *zd, uint64_t object, uint64_t offset)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(dmu_object_info(zd->zd_os, object, &doi) == 0);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Pick an i/o type at random, biased toward writing block tags.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ztest_bt_generate(&wbt, zd->zd_os, object, offset, 0, 0, 0);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick (void) ztest_write(zd, object, offset, sizeof (wbt), &wbt);
88ecc943b4eb72f7c4fbbd8435997b85ef171fc3George Wilson (void) memset(data, 'a' + (object + offset) % 5, blocksize);
88ecc943b4eb72f7c4fbbd8435997b85ef171fc3George Wilson * Induce fletcher2 collisions to ensure that
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * zio_ddt_collision() detects and resolves them
8654d0253136055bd4cc2423d87378e8a37f2eb5perrin * when using fletcher2-verify for deduplication.
88ecc943b4eb72f7c4fbbd8435997b85ef171fc3George Wilson (void) ztest_write(zd, object, offset, blocksize, data);
88ecc943b4eb72f7c4fbbd8435997b85ef171fc3George Wilson (void) ztest_write(zd, object, offset, blocksize, data);
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) ztest_truncate(zd, object, offset, blocksize);
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante ZFS_PROP_CHECKSUM, spa_dedup_checksum(ztest_spa),
88ecc943b4eb72f7c4fbbd8435997b85ef171fc3George Wilson VERIFY0(dmu_read(zd->zd_os, object, offset, blocksize, data,
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) ztest_write(zd, object, offset, blocksize, data);
31157203d871bc0b4bc67d925b0a8424570774a1Jeff Bonwick * Initialize an object description template.
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwickztest_od_init(ztest_od_t *od, uint64_t id, char *tag, uint64_t index,
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick dmu_object_type_t type, uint64_t blocksize, uint64_t gen)
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick od->od_crblocksize = blocksize ? blocksize : ztest_random_blocksize();
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick (void) snprintf(od->od_name, sizeof (od->od_name), "%s(%lld)[%llu]",
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick * Lookup or create the objects for a test using the od template.
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick * If the objects do not all exist, or if 'remove' is specified,
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick * remove any existing objects and create new ones. Otherwise,
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick * use the existing objects.
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwickztest_object_init(ztest_ds_t *zd, ztest_od_t *od, size_t size, boolean_t remove)
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick if ((ztest_lookup(zd, od, count) != 0 || remove) &&
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick VERIFY(mutex_unlock(&zd->zd_dirobj_lock) == 0);
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick/* ARGSUSED */
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick zil_commit(zilog, ztest_random(ZTEST_OBJECTS));
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick * Remember the committed values in zd, which is in parent/child
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick * shared memory. If we die, the next iteration of ztest_run()
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick * will verify that the log really does contain this record.
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick ASSERT3U(zd->zd_shared->zd_seq, <=, zilog->zl_commit_lr_seq);
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick zd->zd_shared->zd_seq = zilog->zl_commit_lr_seq;
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick * This function is designed to simulate the operations that occur during a
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * mount/unmount operation. We hold the dataset across these operations in an
fa9e4066f08beec538e775443c5be79dd423fcabahrens * attempt to expose any implicit assumptions about ZIL management.
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante/* ARGSUSED */
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musanteztest_zil_remount(ztest_ds_t *zd, uint64_t id)
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante * We grab the zd_dirobj_lock to ensure that no other thread is
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante * updating the zil (i.e. adding in-memory log records) and the
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante * zd_zilog_lock to block any I/O.
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante /* zfsvfs_teardown() */
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante /* zfsvfs_setup() */
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante VERIFY(zil_open(os, ztest_get_data) == zd->zd_zilog);
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante VERIFY(mutex_unlock(&zd->zd_dirobj_lock) == 0);
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante * Verify that we can't destroy an active pool, create an existing pool,
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante * or create a pool with a bad vdev spec.
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante/* ARGSUSED */
98295d61142c413b360ac29a3d3957bdb90886e1Mark J Musanteztest_spa_create_destroy(ztest_ds_t *zd, uint64_t id)
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante * Attempt to create using a bad file.
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, 0, 0, 0, 1);
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante spa_create("ztest_bad_file", nvroot, NULL, NULL));
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante * Attempt to create using a bad mirror.
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, 0, 0, 2, 1);
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante spa_create("ztest_bad_mirror", nvroot, NULL, NULL));
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante * Attempt to create an existing pool. It shouldn't matter
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante * what's in the nvroot; we should fail with EEXIST.
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante nvroot = make_vdev_root("/dev/bogus", NULL, NULL, 0, 0, 0, 0, 0, 1);
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante VERIFY3U(EEXIST, ==, spa_create(zo->zo_pool, nvroot, NULL, NULL));
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante VERIFY3U(0, ==, spa_open(zo->zo_pool, &spa, FTAG));
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante VERIFY3U(EBUSY, ==, spa_destroy(zo->zo_pool));
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante/* ARGSUSED */
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musanteztest_spa_upgrade(ztest_ds_t *zd, uint64_t id)
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante uint64_t initial_version = SPA_VERSION_INITIAL;
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante name = kmem_asprintf("%s_upgrade", ztest_opts.zo_pool);
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante * Clean up from previous runs.
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante nvroot = make_vdev_root(NULL, NULL, name, ztest_opts.zo_vdev_size, 0,
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante 0, ztest_opts.zo_raidz, ztest_opts.zo_mirrors, 1);
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante * If we're configuring a RAIDZ device then make sure that the
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante * the initial version is capable of supporting that feature.
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock * Create a pool with a spa version that can be upgraded. Pick
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick * a value between initial_version and SPA_VERSION_BEFORE_FEATURES.
ecc2d604e885a75cc75e647b5641af99d5a6f4a6bonwick version = ztest_random_spa_version(initial_version);
8ad4d6dd86f5bc65fb3afa566c8133f3bac21648Jeff Bonwick } while (version > SPA_VERSION_BEFORE_FEATURES);
9af0a4dffb12fece9e4ca0a97721e6e05ca2bca9Jeff Bonwick zpool_prop_to_name(ZPOOL_PROP_VERSION), version);
31157203d871bc0b4bc67d925b0a8424570774a1Jeff Bonwick VERIFY0(spa_create(name, nvroot, props, NULL));
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) printf("upgrading spa version from %llu to %llu\n",
fa9e4066f08beec538e775443c5be79dd423fcabahrens VERIFY3U(spa_version(spa), ==, fnvlist_lookup_uint64(spa->spa_config,
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante if (vd->vdev_path != NULL && strcmp(path, vd->vdev_path) == 0)
8ad4d6dd86f5bc65fb3afa566c8133f3bac21648Jeff Bonwick if ((mvd = vdev_lookup_by_path(vd->vdev_child[c], path)) !=
31157203d871bc0b4bc67d925b0a8424570774a1Jeff Bonwick * Find the first available hole which can be used as a top-level.
fa9e4066f08beec538e775443c5be79dd423fcabahrens ASSERT(spa_config_held(spa, SCL_VDEV, RW_READER) == SCL_VDEV);
9af0a4dffb12fece9e4ca0a97721e6e05ca2bca9Jeff Bonwick * Verify that vdev_add() works as expected.
8ad4d6dd86f5bc65fb3afa566c8133f3bac21648Jeff Bonwick/* ARGSUSED */
8ad4d6dd86f5bc65fb3afa566c8133f3bac21648Jeff Bonwickztest_vdev_add_remove(ztest_ds_t *zd, uint64_t id)
31157203d871bc0b4bc67d925b0a8424570774a1Jeff Bonwick leaves = MAX(zs->zs_mirrors + zs->zs_splits, 1) * ztest_opts.zo_raidz;
31157203d871bc0b4bc67d925b0a8424570774a1Jeff Bonwick spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
31157203d871bc0b4bc67d925b0a8424570774a1Jeff Bonwick ztest_shared->zs_vdev_next_leaf = find_vdev_hole(spa) * leaves;
31157203d871bc0b4bc67d925b0a8424570774a1Jeff Bonwick * If we have slogs then remove them 1/4 of the time.
31157203d871bc0b4bc67d925b0a8424570774a1Jeff Bonwick if (spa_has_slogs(spa) && ztest_random(4) == 0) {
31157203d871bc0b4bc67d925b0a8424570774a1Jeff Bonwick * Grab the guid from the head of the log class rotor.
31157203d871bc0b4bc67d925b0a8424570774a1Jeff Bonwick guid = spa_log_class(spa)->mc_rotor->mg_vd->vdev_guid;
31157203d871bc0b4bc67d925b0a8424570774a1Jeff Bonwick * We have to grab the zs_name_lock as writer to
31157203d871bc0b4bc67d925b0a8424570774a1Jeff Bonwick * prevent a race between removing a slog (dmu_objset_find)
31157203d871bc0b4bc67d925b0a8424570774a1Jeff Bonwick * and destroying a dataset. Removing the slog will
31157203d871bc0b4bc67d925b0a8424570774a1Jeff Bonwick * grab a reference on the dataset which may cause
fa9e4066f08beec538e775443c5be79dd423fcabahrens * dmu_objset_destroy() to fail with EBUSY thus
fa9e4066f08beec538e775443c5be79dd423fcabahrens * leaving the dataset in an inconsistent state.
31157203d871bc0b4bc67d925b0a8424570774a1Jeff Bonwick * Make 1/4 of the devices be log devices.
fa9e4066f08beec538e775443c5be79dd423fcabahrens else if (error != 0)
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick * Verify that adding/removing aux devices (l2arc, hot spare) works as expected.
31157203d871bc0b4bc67d925b0a8424570774a1Jeff Bonwick/* ARGSUSED */
fa9e4066f08beec538e775443c5be79dd423fcabahrensztest_vdev_aux_add_remove(ztest_ds_t *zd, uint64_t id)
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Pick a random device to remove.
fa9e4066f08beec538e775443c5be79dd423fcabahrens guid = sav->sav_vdevs[ztest_random(sav->sav_count)]->vdev_guid;
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson * Find an unused device we can add.
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson (void) snprintf(path, sizeof (path), ztest_aux_template,
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson * Add a new device.
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson nvlist_t *nvroot = make_vdev_root(NULL, aux, NULL,
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson (ztest_opts.zo_vdev_size * 5) / 4, 0, 0, 0, 0, 1);
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson fatal(0, "spa_vdev_add(%p) = %d", nvroot, error);
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson * Remove an existing device. Sometimes, dirty its
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson * vdev state first to make sure we handle removal
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson * of devices that have pending state changes.
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson fatal(0, "spa_vdev_remove(%llu) = %d", guid, error);
095bcd6622e3b3520eb3b71039a3be5cfab25b74George Wilson * split a pool if it has mirror tlvdevs
095bcd6622e3b3520eb3b71039a3be5cfab25b74George Wilson/* ARGSUSED */
095bcd6622e3b3520eb3b71039a3be5cfab25b74George Wilson nvlist_t *tree, **child, *config, *split, **schild;
095bcd6622e3b3520eb3b71039a3be5cfab25b74George Wilson uint_t c, children, schildren = 0, lastlogid = 0;
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson /* ensure we have a useable config; mirrors of raidz aren't supported */
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson if (zs->zs_mirrors < 3 || ztest_opts.zo_raidz > 1) {
8f18d1fadf6a0c20fac9ff7259a5368faa3c3bfbGeorge Wilson /* clean up the old pool, if any */
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
8f18d1fadf6a0c20fac9ff7259a5368faa3c3bfbGeorge Wilson /* generate a config from the existing config */
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson VERIFY(nvlist_lookup_nvlist(spa->spa_config, ZPOOL_CONFIG_VDEV_TREE,
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson VERIFY(nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child,
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson schild = malloc(rvd->vdev_children * sizeof (nvlist_t *));
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson for (c = 0; c < children; c++) {
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson if (tvd->vdev_islog || tvd->vdev_ops == &vdev_hole_ops) {
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson VERIFY(nvlist_alloc(&schild[schildren], NV_UNIQUE_NAME,
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0);
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson VERIFY(nvlist_dup(mchild[0], &schild[schildren++], 0) == 0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens /* OK, create a config that can be used to split */
fa9e4066f08beec538e775443c5be79dd423fcabahrens VERIFY(nvlist_alloc(&split, NV_UNIQUE_NAME, 0) == 0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens VERIFY(nvlist_add_nvlist_array(split, ZPOOL_CONFIG_CHILDREN, schild,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(nvlist_alloc(&config, NV_UNIQUE_NAME, 0) == 0);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, split) == 0);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick for (c = 0; c < schildren; c++)
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson error = spa_vdev_split_mirror(spa, "splitp", config, NULL, B_FALSE);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick (void) printf("successful split - results:\n");
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson * Verify that we can attach and detach devices.
8f18d1fadf6a0c20fac9ff7259a5368faa3c3bfbGeorge Wilson/* ARGSUSED */
8f18d1fadf6a0c20fac9ff7259a5368faa3c3bfbGeorge Wilsonztest_vdev_attach_detach(ztest_ds_t *zd, uint64_t id)
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson char oldpath[MAXPATHLEN], newpath[MAXPATHLEN];
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raidz;
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Decide whether to do an attach or a replace.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Pick a random top-level vdev.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Pick a random leaf within it.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Locate this vdev.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ASSERT(oldvd->vdev_children >= zs->zs_mirrors);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick oldvd = oldvd->vdev_child[leaf / ztest_opts.zo_raidz];
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ASSERT(oldvd->vdev_children == ztest_opts.zo_raidz);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick oldvd = oldvd->vdev_child[leaf % ztest_opts.zo_raidz];
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * If we're already doing an attach or replace, oldvd may be a
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * mirror vdev -- in which case, pick a random child.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick oldvd = oldvd->vdev_child[ztest_random(oldvd->vdev_children)];
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson * If oldvd has siblings, then half of the time, detach it.
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson if (oldvd_has_siblings && ztest_random(2) == 0) {
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick error = spa_vdev_detach(spa, oldguid, pguid, B_FALSE);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick if (error != 0 && error != ENODEV && error != EBUSY &&
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick fatal(0, "detach (%s) returned %d", oldpath, error);
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson * For the new vdev, choose with equal probability between the two
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson * standard paths (ending in either 'a' or 'b') or a random hot spare.
573ca77e53dd31dcaebef023e7eb41969e6896c1George Wilson if (sav->sav_count != 0 && ztest_random(3) == 0) {
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick newvd = sav->sav_vdevs[ztest_random(sav->sav_count)];
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick (void) snprintf(newpath, sizeof (newpath), ztest_dev_template,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Make newsize a little bigger or smaller than oldsize.
55da60b91d96984f12de050ce428373ea25c7f35Mark J Musante * If it's smaller, the attach should fail.
55da60b91d96984f12de050ce428373ea25c7f35Mark J Musante * If it's larger, and we're doing a replace,
55da60b91d96984f12de050ce428373ea25c7f35Mark J Musante * we should get dynamic LUN growth when we're done.
55da60b91d96984f12de050ce428373ea25c7f35Mark J Musante newsize = 10 * oldsize / (9 + ztest_random(3));
55da60b91d96984f12de050ce428373ea25c7f35Mark J Musante * If pvd is not a mirror or root, the attach should fail with ENOTSUP,
55da60b91d96984f12de050ce428373ea25c7f35Mark J Musante * unless it's a replace; in that case any non-replacing parent is OK.
55da60b91d96984f12de050ce428373ea25c7f35Mark J Musante * If newvd is already part of the pool, it should fail with EBUSY.
55da60b91d96984f12de050ce428373ea25c7f35Mark J Musante * If newvd is too small, it should fail with EOVERFLOW.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick else if (newvd_is_spare && (!replacing || oldvd_is_log))
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Build the nvlist describing newpath.
fa9e4066f08beec538e775443c5be79dd423fcabahrens root = make_vdev_root(newpath, NULL, NULL, newvd == NULL ? newsize : 0,
fa9e4066f08beec538e775443c5be79dd423fcabahrens error = spa_vdev_attach(spa, oldguid, root, replacing);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * If our parent was the replacing vdev, but the replace completed,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * then instead of failing with ENOTSUP we may either succeed,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * fail with ENODEV, or fail with EOVERFLOW.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick (error == 0 || error == ENODEV || error == EOVERFLOW))
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * If someone grew the LUN, the replacement may be too small.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick /* XXX workaround 6690467 */
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick if (error != expected_error && expected_error != EBUSY) {
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick "returned %d, expected %d",
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick (longlong_t)newsize, replacing, error, expected_error);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Callback function which expands the physical size of the vdev.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick (void) printf("%s grew from %lu to %lu bytes\n",
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick vd->vdev_path, (ulong_t)fsize, (ulong_t)*newsize);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Callback function which expands a given vdev by calling vdev_online().
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick/* ARGSUSED */
fa9e4066f08beec538e775443c5be79dd423fcabahrens uint64_t generation = spa->spa_config_generation + 1;
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ASSERT(spa_config_held(spa, SCL_STATE, RW_READER) == SCL_STATE);
fa9e4066f08beec538e775443c5be79dd423fcabahrens /* Calling vdev_online will initialize the new metaslabs */
fa9e4066f08beec538e775443c5be79dd423fcabahrens error = vdev_online(spa, guid, ZFS_ONLINE_EXPAND, &newstate);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick spa_config_enter(spa, SCL_STATE, spa, RW_READER);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * If vdev_online returned an error or the underlying vdev_open
fa9e4066f08beec538e775443c5be79dd423fcabahrens * failed then we abort the expand. The only way to know that
55da60b91d96984f12de050ce428373ea25c7f35Mark J Musante * vdev_open fails is by checking the returned newstate.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick (void) printf("Unable to expand vdev, state %llu, "
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Since we dropped the lock we need to ensure that we're
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * still talking to the original vdev. It's possible this
fa9e4066f08beec538e775443c5be79dd423fcabahrens * vdev may have been detached/replaced while we were
fa9e4066f08beec538e775443c5be79dd423fcabahrens * trying to online it.
fa9e4066f08beec538e775443c5be79dd423fcabahrens "guid %llu, state %llu, expected gen %llu, "
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick "got gen %llu\n",
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Traverse the vdev tree calling the supplied function.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * We continue to walk the tree until we either have walked all
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * children or we receive a non-NULL return from the callback.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * If a NULL callback is passed, then we just return back the first
fa9e4066f08beec538e775443c5be79dd423fcabahrens * leaf vdev we encounter.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickvdev_walk_tree(vdev_t *vd, vdev_t *(*func)(vdev_t *, void *), void *arg)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick if ((cvd = vdev_walk_tree(cvd, func, arg)) != NULL)
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Verify that dynamic LUN growth works as expected.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick/* ARGSUSED */
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_vdev_LUN_growth(ztest_ds_t *zd, uint64_t id)
718d718a734e79bad0b9f8a264219b238fabaf44George Wilson uint64_t old_class_space, new_class_space, old_ms_count, new_ms_count;
718d718a734e79bad0b9f8a264219b238fabaf44George Wilson spa_config_enter(spa, SCL_STATE, spa, RW_READER);
718d718a734e79bad0b9f8a264219b238fabaf44George Wilson old_class_space = metaslab_class_get_space(mc);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Determine the size of the first leaf vdev associated with
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * our top-level device.
718d718a734e79bad0b9f8a264219b238fabaf44George Wilson * We only try to expand the vdev if it's healthy, less than 4x its
842727c2f41f01b380de4f5e787d905702870f23Chris Kirby * original size, and it has a valid psize.
842727c2f41f01b380de4f5e787d905702870f23Chris Kirby psize == 0 || psize >= 4 * ztest_opts.zo_vdev_size) {
4f5064b73b1cc9de1d0f1a2ae700d519d4d565dfMark J Musante (void) printf("Expanding LUN %s from %lu to %lu\n",
4f5064b73b1cc9de1d0f1a2ae700d519d4d565dfMark J Musante vd->vdev_path, (ulong_t)psize, (ulong_t)newsize);
4f5064b73b1cc9de1d0f1a2ae700d519d4d565dfMark J Musante * Growing the vdev is a two step process:
4f5064b73b1cc9de1d0f1a2ae700d519d4d565dfMark J Musante * 1). expand the physical size (i.e. relabel)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * 2). online the vdev to create the new metaslabs
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick if (vdev_walk_tree(tvd, grow_vdev, &newsize) != NULL ||
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick vdev_walk_tree(tvd, online_vdev, NULL) != NULL ||
4f5064b73b1cc9de1d0f1a2ae700d519d4d565dfMark J Musante "the vdev configuration changed.\n");
05312e2ca4a1e68606ae0db2c449aee7e7009323Mark J Musante * Expanding the LUN will update the config asynchronously,
ea2f5b9e5bf4966630882d6d681a94768aea1d75Matthew Ahrens * thus we must wait for the async thread to complete any
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * pending tasks before proceeding.
b2d22b66bc47a107891381986ae05e32ff517d10George Wilson done = (spa->spa_async_thread == NULL && !spa->spa_async_tasks);
ae46e4c775f2becc5343ff90b60a95acb79735f9Matthew Ahrens spa_config_enter(spa, SCL_STATE, spa, RW_READER);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick new_class_space = metaslab_class_get_space(mc);
b2d22b66bc47a107891381986ae05e32ff517d10George Wilson if (tvd->vdev_mg != mg || mg->mg_class != mc) {
b2d22b66bc47a107891381986ae05e32ff517d10George Wilson (void) printf("Could not verify LUN expansion due to "
4f5064b73b1cc9de1d0f1a2ae700d519d4d565dfMark J Musante "intervening vdev offline or remove.\n");
b2d22b66bc47a107891381986ae05e32ff517d10George Wilson * Make sure we were able to grow the vdev.
4f5064b73b1cc9de1d0f1a2ae700d519d4d565dfMark J Musante fatal(0, "LUN expansion failed: ms_count %llu <= %llu\n",
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Make sure we were able to grow the pool.
b2d22b66bc47a107891381986ae05e32ff517d10George Wilson fatal(0, "LUN expansion failed: class_space %llu <= %llu\n",
cde58dbc6a23d4d38db7c8866312be83221c765fMatthew Ahrens * Verify that dmu_objset_{create,destroy,open,close} work as expected.
cde58dbc6a23d4d38db7c8866312be83221c765fMatthew Ahrens/* ARGSUSED */
4f5064b73b1cc9de1d0f1a2ae700d519d4d565dfMark J Musanteztest_objset_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
4f5064b73b1cc9de1d0f1a2ae700d519d4d565dfMark J Musante * Create the objects common to all ztest datasets.
fa9e4066f08beec538e775443c5be79dd423fcabahrens int err = dmu_objset_create(dsname, DMU_OST_OTHER, 0,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick (void) printf("Setting dataset %s to sync always\n", dsname);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick return (ztest_dsl_prop_set_uint64(dsname, ZFS_PROP_SYNC,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick/* ARGSUSED */
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_objset_destroy_cb(const char *name, void *arg)
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Verify that the dataset contains a directory object.
fa9e4066f08beec538e775443c5be79dd423fcabahrens VERIFY0(dmu_objset_own(name, DMU_OST_OTHER, B_TRUE, FTAG, &os));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick /* We could have crashed in the middle of destroying it */
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Destroy the dataset.
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) snprintf(snapname, sizeof (snapname), "%llu", (u_longlong_t)id);
fa9e4066f08beec538e775443c5be79dd423fcabahrens fatal(0, "ztest_snapshot_create(%s@%s) = %d", osname,
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) snprintf(snapname, MAXNAMELEN, "%s@%llu", osname,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick fatal(0, "ztest_snapshot_destroy(%s) = %d", snapname, error);
fa9e4066f08beec538e775443c5be79dd423fcabahrens/* ARGSUSED */
fa9e4066f08beec538e775443c5be79dd423fcabahrensztest_dmu_objset_create_destroy(ztest_ds_t *zd, uint64_t id)
fa9e4066f08beec538e775443c5be79dd423fcabahrens * If this dataset exists from a previous run, process its replay log
fa9e4066f08beec538e775443c5be79dd423fcabahrens * half of the time. If we don't replay it, then dmu_objset_destroy()
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * (invoked from ztest_objset_destroy_cb()) should just throw it away.
fa9e4066f08beec538e775443c5be79dd423fcabahrens dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, FTAG, &os) == 0) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens * There may be an old instance of the dataset we're about to
fa9e4066f08beec538e775443c5be79dd423fcabahrens * create lying around from a previous run. If so, destroy it
fa9e4066f08beec538e775443c5be79dd423fcabahrens * and all of its snapshots.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL,
7bfdf011e081684f853a3242d0296695110d9d84Neil Perrin * Verify that the destroyed dataset is no longer in the namespace.
fa9e4066f08beec538e775443c5be79dd423fcabahrens VERIFY3U(ENOENT, ==, dmu_objset_own(name, DMU_OST_OTHER, B_TRUE,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Verify that we can create a new dataset.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick fatal(0, "dmu_objset_create(%s) = %d", name, error);
fa9e4066f08beec538e775443c5be79dd423fcabahrens VERIFY0(dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, FTAG, &os));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Open the intent log for it.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Put some objects in there, do a little I/O to them,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * and randomly take a couple of snapshots along the way.
fa9e4066f08beec538e775443c5be79dd423fcabahrens for (int i = 0; i < iters; i++) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Verify that we cannot create an existing dataset.
fa9e4066f08beec538e775443c5be79dd423fcabahrens dmu_objset_create(name, DMU_OST_OTHER, 0, NULL, NULL));
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Verify that we can hold an objset that is also owned.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Verify that we cannot own an objset that is already owned.
fa9e4066f08beec538e775443c5be79dd423fcabahrens dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, FTAG, &os2));
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Verify that dmu_snapshot_{create,destroy,open,close} work as expected.
fa9e4066f08beec538e775443c5be79dd423fcabahrensztest_dmu_snapshot_create_destroy(ztest_ds_t *zd, uint64_t id)
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Cleanup non-standard snapshots and clones.
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) snprintf(snap1name, MAXNAMELEN, "%s@s1_%llu", osname, id);
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) snprintf(clone1name, MAXNAMELEN, "%s/c1_%llu", osname, id);
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) snprintf(snap2name, MAXNAMELEN, "%s@s2_%llu", clone1name, id);
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) snprintf(clone2name, MAXNAMELEN, "%s/c2_%llu", osname, id);
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) snprintf(snap3name, MAXNAMELEN, "%s@s3_%llu", clone1name, id);
fa9e4066f08beec538e775443c5be79dd423fcabahrens fatal(0, "dsl_destroy_head(%s) = %d", clone2name, error);
fa9e4066f08beec538e775443c5be79dd423fcabahrens fatal(0, "dsl_destroy_snapshot(%s) = %d", snap3name, error);
fa9e4066f08beec538e775443c5be79dd423fcabahrens fatal(0, "dsl_destroy_snapshot(%s) = %d", snap2name, error);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick fatal(0, "dsl_destroy_head(%s) = %d", clone1name, error);
7bfdf011e081684f853a3242d0296695110d9d84Neil Perrin error = dsl_destroy_snapshot(snap1name, B_FALSE);
7bfdf011e081684f853a3242d0296695110d9d84Neil Perrin fatal(0, "dsl_destroy_snapshot(%s) = %d", snap1name, error);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Verify dsl_dataset_promote handles EBUSY
fa9e4066f08beec538e775443c5be79dd423fcabahrensztest_dsl_dataset_promote_busy(ztest_ds_t *zd, uint64_t id)
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy (void) snprintf(snap1name, MAXNAMELEN, "%s@s1_%llu", osname, id);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy (void) snprintf(clone1name, MAXNAMELEN, "%s/c1_%llu", osname, id);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy (void) snprintf(snap2name, MAXNAMELEN, "%s@s2_%llu", clone1name, id);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy (void) snprintf(clone2name, MAXNAMELEN, "%s/c2_%llu", osname, id);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy (void) snprintf(snap3name, MAXNAMELEN, "%s@s3_%llu", clone1name, id);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy error = dmu_objset_snapshot_one(osname, strchr(snap1name, '@') + 1);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick fatal(0, "dmu_take_snapshot(%s) = %d", snap1name, error);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy error = dmu_objset_clone(clone1name, snap1name);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy fatal(0, "dmu_objset_create(%s) = %d", clone1name, error);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy error = dmu_objset_snapshot_one(clone1name, strchr(snap2name, '@') + 1);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy fatal(0, "dmu_open_snapshot(%s) = %d", snap2name, error);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy error = dmu_objset_snapshot_one(clone1name, strchr(snap3name, '@') + 1);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy fatal(0, "dmu_open_snapshot(%s) = %d", snap3name, error);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy error = dmu_objset_clone(clone2name, snap3name);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy fatal(0, "dmu_objset_create(%s) = %d", clone2name, error);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick error = dmu_objset_own(snap2name, DMU_OST_ANY, B_TRUE, FTAG, &os);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy fatal(0, "dmu_objset_own(%s) = %d", snap2name, error);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy error = dsl_dataset_promote(clone2name, NULL);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy fatal(0, "dsl_dataset_promote(%s), %d, not EBUSY", clone2name,
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * Verify that dmu_object_{alloc,free} work as expected.
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiyztest_dmu_object_alloc_free(ztest_ds_t *zd, uint64_t id)
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy int batchsize = sizeof (od) / sizeof (od[0]);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy for (int b = 0; b < batchsize; b++)
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy ztest_od_init(&od[b], id, FTAG, b, DMU_OT_UINT64_OTHER, 0, 0);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Destroy the previous batch of objects, create a new batch,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * and do some I/O on the new objects.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick if (ztest_object_init(zd, od, sizeof (od), B_TRUE) != 0)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ztest_io(zd, od[ztest_random(batchsize)].od_object,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Verify that dmu_{read,write} work as expected.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_dmu_read_write(ztest_ds_t *zd, uint64_t id)
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy bufwad_t *packbuf, *bigbuf, *pack, *bigH, *bigT;
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize;
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy uint64_t chunksize = (1000 + ztest_random(1000)) * sizeof (uint64_t);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * This test uses two objects, packobj and bigobj, that are always
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * updated together (i.e. in the same tx) so that their contents are
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * in sync and can be compared. Their contents relate to each other
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * in a simple way: packobj is a dense array of 'bufwad' structures,
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * while bigobj is a sparse array of the same bufwads. Specifically,
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * for any index n, there are three bufwads that should be identical:
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * packobj, at offset n * sizeof (bufwad_t)
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * bigobj, at the head of the nth chunk
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * bigobj, at the tail of the nth chunk
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * The chunk size is arbitrary. It doesn't have to be a power of two,
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * and it doesn't have any relation to the object blocksize.
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * The only requirement is that it can hold at least two bufwads.
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * Normally, we write the bufwad to each of these locations.
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * However, free_percent of the time we instead write zeroes to
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * packobj and perform a dmu_free_range() on bigobj. By comparing
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * bigobj to packobj, we can verify that the DMU is correctly
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * tracking which parts of an object are allocated and free,
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * and that the contents of the allocated blocks are correct.
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * Read the directory info. If it's the first time, set things up.
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy ztest_od_init(&od[0], id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, chunksize);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ztest_od_init(&od[1], id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, chunksize);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy if (ztest_object_init(zd, od, sizeof (od), B_FALSE) != 0)
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * Prefetch a random chunk of the big object.
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * Our aim here is to get some async reads in flight
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * for blocks that we may free below; the DMU should
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * handle this race correctly.
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy n = ztest_random(regions) * stride + ztest_random(width);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick dmu_prefetch(os, bigobj, n * chunksize, s * chunksize);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * Pick a random index and compute the offsets into packobj and bigobj.
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy n = ztest_random(regions) * stride + ztest_random(width);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy packbuf = umem_alloc(packsize, UMEM_NOFAIL);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy bigbuf = umem_alloc(bigsize, UMEM_NOFAIL);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * free_percent of the time, free a range of bigobj rather than
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * overwriting it.
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy freeit = (ztest_random(100) < free_percent);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Read the current contents of our objects.
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy error = dmu_read(os, packobj, packoff, packsize, packbuf,
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy error = dmu_read(os, bigobj, bigoff, bigsize, bigbuf,
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * Get a tx for the mods to both packobj and bigobj.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick dmu_tx_hold_write(tx, packobj, packoff, packsize);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy dmu_tx_hold_free(tx, bigobj, bigoff, bigsize);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy dmu_tx_hold_write(tx, bigobj, bigoff, bigsize);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy (enum zio_checksum)ztest_random_dsl_prop(ZFS_PROP_CHECKSUM), tx);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick (enum zio_compress)ztest_random_dsl_prop(ZFS_PROP_COMPRESSION), tx);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * For each index from n to n + s, verify that the existing bufwad
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * in packobj matches the bufwads at the head and tail of the
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy * corresponding chunk in bigobj. Then update all three bufwads
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * with the new values we want to write out.
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy for (i = 0; i < s; i++) {
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t));
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy bigH = (bufwad_t *)((char *)bigbuf + i * chunksize);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy bigT = (bufwad_t *)((char *)bigH + chunksize) - 1;
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ASSERT((uintptr_t)bigH - (uintptr_t)bigbuf < bigsize);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy ASSERT((uintptr_t)bigT - (uintptr_t)bigbuf < bigsize);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy fatal(0, "future leak: got %llx, open txg is %llx",
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy if (pack->bw_data != 0 && pack->bw_index != n + i)
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy fatal(0, "wrong index: got %llx, wanted %llx+%llx",
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy if (bcmp(pack, bigH, sizeof (bufwad_t)) != 0)
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy fatal(0, "pack/bigH mismatch in %p/%p", pack, bigH);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy if (bcmp(pack, bigT, sizeof (bufwad_t)) != 0)
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy fatal(0, "pack/bigT mismatch in %p/%p", pack, bigT);
2fdbea25c2ba89186b8a6b7c6840ebc9f4dff245Aleksandr Guzovskiy pack->bw_data = 1 + ztest_random(-2ULL);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * We've verified all the old bufwads, and made new ones.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Now write them out.
e05725b117836db173257fae43fb0746eb857fb5bonwick dmu_write(os, packobj, packoff, packsize, packbuf, tx);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick " txg %llx\n",
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(0 == dmu_free_range(os, bigobj, bigoff, bigsize, tx));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick " txg %llx\n",
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick dmu_write(os, bigobj, bigoff, bigsize, bigbuf, tx);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Sanity check the stuff we just wrote.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick void *packcheck = umem_alloc(packsize, UMEM_NOFAIL);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ASSERT(bcmp(packbuf, packcheck, packsize) == 0);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickcompare_and_update_pbbufs(uint64_t s, bufwad_t *packbuf, bufwad_t *bigbuf,
fa9e4066f08beec538e775443c5be79dd423fcabahrens uint64_t bigsize, uint64_t n, uint64_t chunksize, uint64_t txg)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * For each index from n to n + s, verify that the existing bufwad
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * in packobj matches the bufwads at the head and tail of the
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * corresponding chunk in bigobj. Then update all three bufwads
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * with the new values we want to write out.
fa9e4066f08beec538e775443c5be79dd423fcabahrens for (i = 0; i < s; i++) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens /* LINTED */
fa9e4066f08beec538e775443c5be79dd423fcabahrens pack = (bufwad_t *)((char *)packbuf + i * sizeof (bufwad_t));
fa9e4066f08beec538e775443c5be79dd423fcabahrens /* LINTED */
fa9e4066f08beec538e775443c5be79dd423fcabahrens /* LINTED */
fa9e4066f08beec538e775443c5be79dd423fcabahrens ASSERT((uintptr_t)bigH - (uintptr_t)bigbuf < bigsize);
fa9e4066f08beec538e775443c5be79dd423fcabahrens ASSERT((uintptr_t)bigT - (uintptr_t)bigbuf < bigsize);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick fatal(0, "future leak: got %llx, open txg is %llx",
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_dmu_read_write_zcopy(ztest_ds_t *zd, uint64_t id)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick uint64_t packobj, packoff, packsize, bigobj, bigoff, bigsize;
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * This test uses two objects, packobj and bigobj, that are always
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * updated together (i.e. in the same tx) so that their contents are
fa9e4066f08beec538e775443c5be79dd423fcabahrens * in sync and can be compared. Their contents relate to each other
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * in a simple way: packobj is a dense array of 'bufwad' structures,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * while bigobj is a sparse array of the same bufwads. Specifically,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * for any index n, there are three bufwads that should be identical:
fa9e4066f08beec538e775443c5be79dd423fcabahrens * packobj, at offset n * sizeof (bufwad_t)
e05725b117836db173257fae43fb0746eb857fb5bonwick * bigobj, at the head of the nth chunk
e05725b117836db173257fae43fb0746eb857fb5bonwick * bigobj, at the tail of the nth chunk
e05725b117836db173257fae43fb0746eb857fb5bonwick * The chunk size is set equal to bigobj block size so that
e05725b117836db173257fae43fb0746eb857fb5bonwick * dmu_assign_arcbuf() can be tested for object updates.
e05725b117836db173257fae43fb0746eb857fb5bonwick * Read the directory info. If it's the first time, set things up.
e05725b117836db173257fae43fb0746eb857fb5bonwick ztest_od_init(&od[0], id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0);
e05725b117836db173257fae43fb0746eb857fb5bonwick ztest_od_init(&od[1], id, FTAG, 1, DMU_OT_UINT64_OTHER, 0, chunksize);
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (ztest_object_init(zd, od, sizeof (od), B_FALSE) != 0)
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Pick a random index and compute the offsets into packobj and bigobj.
e05725b117836db173257fae43fb0746eb857fb5bonwick n = ztest_random(regions) * stride + ztest_random(width);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY3U(0, ==, dmu_bonus_hold(os, bigobj, FTAG, &bonus_db));
e05725b117836db173257fae43fb0746eb857fb5bonwick bigbuf_arcbufs = umem_zalloc(2 * s * sizeof (arc_buf_t *), UMEM_NOFAIL);
e05725b117836db173257fae43fb0746eb857fb5bonwick * Iteration 0 test zcopy for DB_UNCACHED dbufs.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Iteration 1 test zcopy to already referenced dbufs.
e05725b117836db173257fae43fb0746eb857fb5bonwick * Iteration 2 test zcopy to dirty dbuf in the same txg.
e05725b117836db173257fae43fb0746eb857fb5bonwick * Iteration 3 test zcopy to dbuf dirty in previous txg.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Iteration 4 test zcopy when dbuf is no longer dirty.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Iteration 5 test zcopy when it can't be done.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Iteration 6 one more zcopy write.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick for (i = 0; i < 7; i++) {
e05725b117836db173257fae43fb0746eb857fb5bonwick * In iteration 5 (i == 5) use arcbufs
e05725b117836db173257fae43fb0746eb857fb5bonwick * that don't match bigobj blksz to test
e05725b117836db173257fae43fb0746eb857fb5bonwick * dmu_assign_arcbuf() when it can't directly
e05725b117836db173257fae43fb0746eb857fb5bonwick * assign an arcbuf to a dbuf.
fa9e4066f08beec538e775443c5be79dd423fcabahrens for (j = 0; j < s; j++) {
e05725b117836db173257fae43fb0746eb857fb5bonwick if (i != 5) {
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Get a tx for the mods to both packobj and bigobj.
12a2833a0260ba374a4e6d8225376193e908784aSanjeev Bagewadi dmu_tx_hold_write(tx, packobj, packoff, packsize);
12a2833a0260ba374a4e6d8225376193e908784aSanjeev Bagewadi dmu_tx_hold_write(tx, bigobj, bigoff, bigsize);
12a2833a0260ba374a4e6d8225376193e908784aSanjeev Bagewadi txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick for (j = 0; j < s; j++) {
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick if (i != 5) {
12a2833a0260ba374a4e6d8225376193e908784aSanjeev Bagewadi umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * 50% of the time don't read objects in the 1st iteration to
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * test dmu_assign_arcbuf() for the case when there're no
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * existing dbufs for the specified offsets.
12a2833a0260ba374a4e6d8225376193e908784aSanjeev Bagewadi error = dmu_read(os, bigobj, bigoff, bigsize,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick compare_and_update_pbbufs(s, packbuf, bigbuf, bigsize,
12a2833a0260ba374a4e6d8225376193e908784aSanjeev Bagewadi * We've verified all the old bufwads, and made new ones.
12a2833a0260ba374a4e6d8225376193e908784aSanjeev Bagewadi * Now write them out.
fa9e4066f08beec538e775443c5be79dd423fcabahrens dmu_write(os, packobj, packoff, packsize, packbuf, tx);
fa9e4066f08beec538e775443c5be79dd423fcabahrens " txg %llx\n",
fa9e4066f08beec538e775443c5be79dd423fcabahrens for (off = bigoff, j = 0; j < s; j++, off += chunksize) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (i != 5) {
e05725b117836db173257fae43fb0746eb857fb5bonwick if (i == 1) {
e05725b117836db173257fae43fb0746eb857fb5bonwick if (i != 5) {
e05725b117836db173257fae43fb0746eb857fb5bonwick if (i == 1) {
e05725b117836db173257fae43fb0746eb857fb5bonwick * Sanity check the stuff we just wrote.
e05725b117836db173257fae43fb0746eb857fb5bonwick void *packcheck = umem_alloc(packsize, UMEM_NOFAIL);
e05725b117836db173257fae43fb0746eb857fb5bonwick if (i == 2) {
e05725b117836db173257fae43fb0746eb857fb5bonwick } else if (i == 3) {
e05725b117836db173257fae43fb0746eb857fb5bonwick umem_free(bigbuf_arcbufs, 2 * s * sizeof (arc_buf_t *));
e05725b117836db173257fae43fb0746eb857fb5bonwick/* ARGSUSED */
e05725b117836db173257fae43fb0746eb857fb5bonwickztest_dmu_write_parallel(ztest_ds_t *zd, uint64_t id)
fa9e4066f08beec538e775443c5be79dd423fcabahrens uint64_t offset = (1ULL << (ztest_random(20) + 43)) +
e05725b117836db173257fae43fb0746eb857fb5bonwick (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT);
e05725b117836db173257fae43fb0746eb857fb5bonwick * Have multiple threads write to large offsets in an object
e05725b117836db173257fae43fb0746eb857fb5bonwick * to verify that parallel writes to an object -- even to the
e05725b117836db173257fae43fb0746eb857fb5bonwick * same blocks within the object -- doesn't cause any trouble.
e05725b117836db173257fae43fb0746eb857fb5bonwick ztest_od_init(&od[0], ID_PARALLEL, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0);
e05725b117836db173257fae43fb0746eb857fb5bonwick if (ztest_object_init(zd, od, sizeof (od), B_FALSE) != 0)
fa9e4066f08beec538e775443c5be79dd423fcabahrens uint64_t offset = (1ULL << (ztest_random(4) + SPA_MAXBLOCKSHIFT)) +
e05725b117836db173257fae43fb0746eb857fb5bonwick (ztest_random(ZTEST_RANGE_LOCKS) << SPA_MAXBLOCKSHIFT);
e05725b117836db173257fae43fb0746eb857fb5bonwick ztest_od_init(&od[0], id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0);
e05725b117836db173257fae43fb0746eb857fb5bonwick if (ztest_object_init(zd, od, sizeof (od), !ztest_random(2)) != 0)
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia if (ztest_truncate(zd, od[0].od_object, offset, count * blocksize) != 0)
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia ztest_prealloc(zd, od[0].od_object, offset, count * blocksize);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia data = umem_zalloc(blocksize, UMEM_NOFAIL);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia uint64_t randoff = offset + (ztest_random(count) * blocksize);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia if (ztest_write(zd, od[0].od_object, randoff, blocksize,
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia * Verify that zap_{create,destroy,add,remove,update} work as expected.
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia char *hc[2] = { "s.acl.h", ".s.open.h.hyLZlg" };
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia ztest_od_init(&od[0], id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia if (ztest_object_init(zd, od, sizeof (od), !ztest_random(2)) != 0)
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia * Generate a known hash collision, and verify that
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia * we can lookup and remove both entries.
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia dmu_tx_hold_zap(tx, object, B_TRUE, NULL);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia for (i = 0; i < 2; i++) {
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia VERIFY3U(0, ==, zap_add(os, object, hc[i], sizeof (uint64_t),
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia for (i = 0; i < 2; i++) {
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia VERIFY3U(EEXIST, ==, zap_add(os, object, hc[i],
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia zap_length(os, object, hc[i], &zl_intsize, &zl_ints));
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia ASSERT3U(zl_intsize, ==, sizeof (uint64_t));
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia for (i = 0; i < 2; i++) {
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia VERIFY3U(0, ==, zap_remove(os, object, hc[i], tx));
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia * Generate a buch of random entries.
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia ints = MAX(ZTEST_ZAP_MIN_INTS, object % ZTEST_ZAP_MAX_INTS);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia prop = ztest_random(ZTEST_ZAP_MAX_PROPS);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia * If these zap entries already exist, validate their contents.
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia error = zap_length(os, object, txgname, &zl_intsize, &zl_ints);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia ASSERT3U(zl_intsize, ==, sizeof (uint64_t));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(zap_lookup(os, object, txgname, zl_intsize,
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia VERIFY(zap_length(os, object, propname, &zl_intsize,
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia ASSERT3U(zl_intsize, ==, sizeof (uint64_t));
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia VERIFY(zap_lookup(os, object, propname, zl_intsize,
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia for (i = 0; i < ints; i++) {
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia ASSERT3U(value[i], ==, last_txg + object + i);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia * Atomically update two entries in our zap object.
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia * The first is named txg_%llu, and contains the txg
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia * in which the property was last updated. The second
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia * is named prop_%llu, and the nth element of its value
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia * should be txg + object + n.
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia dmu_tx_hold_zap(tx, object, B_TRUE, NULL);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia fatal(0, "zap future leak: old %llu new %llu", last_txg, txg);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia for (i = 0; i < ints; i++)
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia VERIFY3U(0, ==, zap_update(os, object, txgname, sizeof (uint64_t),
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia VERIFY3U(0, ==, zap_update(os, object, propname, sizeof (uint64_t),
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia * Remove a random pair of entries.
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia prop = ztest_random(ZTEST_ZAP_MAX_PROPS);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia (void) sprintf(propname, "prop_%llu", (u_longlong_t)prop);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia (void) sprintf(txgname, "txg_%llu", (u_longlong_t)prop);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia error = zap_length(os, object, txgname, &zl_intsize, &zl_ints);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia dmu_tx_hold_zap(tx, object, B_TRUE, NULL);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia VERIFY3U(0, ==, zap_remove(os, object, txgname, tx));
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia VERIFY3U(0, ==, zap_remove(os, object, propname, tx));
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia * Testcase to test the upgrading of a microzap to fatzap.
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia ztest_od_init(&od[0], id, FTAG, 0, DMU_OT_ZAP_OTHER, 0, 0);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia if (ztest_object_init(zd, od, sizeof (od), !ztest_random(2)) != 0)
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia * Add entries to this ZAP and make sure it spills over
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia * and gets upgraded to a fatzap. Also, since we are adding
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia * 2050 entries we should see ptrtbl growth and leaf-block split.
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia for (int i = 0; i < 2050; i++) {
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia (void) snprintf(name, sizeof (name), "fzap-%llu-%llu",
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia dmu_tx_hold_zap(tx, object, B_TRUE, name);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia txg = ztest_tx_assign(tx, TXG_MIGHTWAIT, FTAG);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia error = zap_add(os, object, name, sizeof (uint64_t), 1,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick/* ARGSUSED */
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick uint64_t txg, object, count, wsize, wc, zl_wsize, zl_wc;
fa9e4066f08beec538e775443c5be79dd423fcabahrens ztest_od_init(&od[0], ID_PARALLEL, FTAG, micro, DMU_OT_ZAP_OTHER, 0, 0);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick if (ztest_object_init(zd, od, sizeof (od), B_FALSE) != 0)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Generate a random name of the form 'xxx.....' where each
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * x is a random printable character and the dots are dots.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * There are 94 such characters, and the name length goes from
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * 6 to 20, so there are 94^3 * 15 = 12,458,760 possible names.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick namelen = ztest_random(sizeof (name) - 5) + 5 + 1;
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick for (i = 0; i < 3; i++)
5c987a37e6561d090231fb3c10ba9f6ea995f34dChris Kirby * Select an operation: length, lookup, add, update, remove.
5c987a37e6561d090231fb3c10ba9f6ea995f34dChris Kirby if (i >= 2) {
5c987a37e6561d090231fb3c10ba9f6ea995f34dChris Kirby switch (i) {
5c987a37e6561d090231fb3c10ba9f6ea995f34dChris Kirby error = zap_length(os, object, name, &zl_wsize, &zl_wc);
5c987a37e6561d090231fb3c10ba9f6ea995f34dChris Kirby error = zap_lookup(os, object, name, wsize, wc, data);
5c987a37e6561d090231fb3c10ba9f6ea995f34dChris Kirby error = zap_add(os, object, name, wsize, wc, data, tx);
5c987a37e6561d090231fb3c10ba9f6ea995f34dChris Kirby VERIFY(zap_update(os, object, name, wsize, wc, data, tx) == 0);
5c987a37e6561d090231fb3c10ba9f6ea995f34dChris Kirby * Commit callback data.
5c987a37e6561d090231fb3c10ba9f6ea995f34dChris Kirbytypedef struct ztest_cb_data {
5c987a37e6561d090231fb3c10ba9f6ea995f34dChris Kirby/* This is the actual commit callback function */
5c987a37e6561d090231fb3c10ba9f6ea995f34dChris Kirby synced_txg = spa_last_synced_txg(data->zcd_spa);
5c987a37e6561d090231fb3c10ba9f6ea995f34dChris Kirby fatal(0, "commit callback of txg %" PRIu64 " called prematurely"
5c987a37e6561d090231fb3c10ba9f6ea995f34dChris Kirby ", last synced txg = %" PRIu64 "\n", data->zcd_txg,
5c987a37e6561d090231fb3c10ba9f6ea995f34dChris Kirby * The private callback data should be destroyed here, but
5c987a37e6561d090231fb3c10ba9f6ea995f34dChris Kirby * since we are going to check the zcd_called field after
5c987a37e6561d090231fb3c10ba9f6ea995f34dChris Kirby * dmu_tx_abort(), we will destroy it there.
5c987a37e6561d090231fb3c10ba9f6ea995f34dChris Kirby /* Was this callback added to the global callback list? */
fa9e4066f08beec538e775443c5be79dd423fcabahrens /* Remove our callback from the list */
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante/* Allocate and initialize callback data structure */
fa9e4066f08beec538e775443c5be79dd423fcabahrens cb_data = umem_zalloc(sizeof (ztest_cb_data_t), UMEM_NOFAIL);
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante * If a number of txgs equal to this threshold have been created after a commit
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante * callback has been registered but not called, then we assume there is an
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante * implementation bug.
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante#define ZTEST_COMMIT_CALLBACK_THRESH (TXG_CONCURRENT_STATES + 2)
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Commit callback test.
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwickztest_dmu_commit_callbacks(ztest_ds_t *zd, uint64_t id)
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick ztest_od_init(&od[0], id, FTAG, 0, DMU_OT_UINT64_OTHER, 0, 0);
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick if (ztest_object_init(zd, od, sizeof (od), B_FALSE) != 0)
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[0]);
fa9e4066f08beec538e775443c5be79dd423fcabahrens dmu_tx_hold_write(tx, od[0].od_object, 0, sizeof (uint64_t));
8f18d1fadf6a0c20fac9ff7259a5368faa3c3bfbGeorge Wilson /* Every once in a while, abort the transaction on purpose */
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[1]);
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick * It's not a strict requirement to call the registered
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick * callbacks from inside dmu_tx_abort(), but that's what
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick * it's supposed to happen in the current implementation
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick * so we will check for that.
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick for (i = 0; i < 2; i++) {
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick for (i = 0; i < 2; i++) {
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick umem_free(cb_data[i], sizeof (ztest_cb_data_t));
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick dmu_tx_callback_register(tx, ztest_commit_callback, cb_data[2]);
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick * Read existing data to make sure there isn't a future leak.
e14bb3258d05c1b1077e2db7cf77088924e56919Jeff Bonwick VERIFY(0 == dmu_read(os, od[0].od_object, 0, sizeof (uint64_t),
fa9e4066f08beec538e775443c5be79dd423fcabahrens fatal(0, "future leak: got %" PRIu64 ", open txg is %" PRIu64,
fa9e4066f08beec538e775443c5be79dd423fcabahrens dmu_write(os, od[0].od_object, 0, sizeof (uint64_t), &txg, tx);
8ad4d6dd86f5bc65fb3afa566c8133f3bac21648Jeff Bonwick * Since commit callbacks don't have any ordering requirement and since
8ad4d6dd86f5bc65fb3afa566c8133f3bac21648Jeff Bonwick * it is theoretically possible for a commit callback to be called
8ad4d6dd86f5bc65fb3afa566c8133f3bac21648Jeff Bonwick * after an arbitrary amount of time has elapsed since its txg has been
8f18d1fadf6a0c20fac9ff7259a5368faa3c3bfbGeorge Wilson * synced, it is difficult to reliably determine whether a commit
8f18d1fadf6a0c20fac9ff7259a5368faa3c3bfbGeorge Wilson * callback hasn't been called due to high load or due to a flawed
8f18d1fadf6a0c20fac9ff7259a5368faa3c3bfbGeorge Wilson * implementation.
8f18d1fadf6a0c20fac9ff7259a5368faa3c3bfbGeorge Wilson * In practice, we will assume that if after a certain number of txgs a
8f18d1fadf6a0c20fac9ff7259a5368faa3c3bfbGeorge Wilson * commit callback hasn't been called, then most likely there's an
8f18d1fadf6a0c20fac9ff7259a5368faa3c3bfbGeorge Wilson * implementation bug..
8f18d1fadf6a0c20fac9ff7259a5368faa3c3bfbGeorge Wilson (txg - ZTEST_COMMIT_CALLBACK_THRESH) > tmp_cb->zcd_txg) {
8f18d1fadf6a0c20fac9ff7259a5368faa3c3bfbGeorge Wilson fatal(0, "Commit callback threshold exceeded, oldest txg: %"
8ad4d6dd86f5bc65fb3afa566c8133f3bac21648Jeff Bonwick PRIu64 ", open txg: %" PRIu64 "\n", tmp_cb->zcd_txg, txg);
8ad4d6dd86f5bc65fb3afa566c8133f3bac21648Jeff Bonwick * Let's find the place to insert our callbacks.
8ad4d6dd86f5bc65fb3afa566c8133f3bac21648Jeff Bonwick * Even though the list is ordered by txg, it is possible for the
fa9e4066f08beec538e775443c5be79dd423fcabahrens * insertion point to not be the end because our txg may already be
fa9e4066f08beec538e775443c5be79dd423fcabahrens * quiescing at this point and other callbacks in the open txg
8f18d1fadf6a0c20fac9ff7259a5368faa3c3bfbGeorge Wilson * (from other objsets) may have sneaked in.
fa9e4066f08beec538e775443c5be79dd423fcabahrens /* Add the 3 callbacks to the list */
fa9e4066f08beec538e775443c5be79dd423fcabahrens for (i = 0; i < 3; i++) {
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante/* ARGSUSED */
fa9e4066f08beec538e775443c5be79dd423fcabahrens for (int p = 0; p < sizeof (proplist) / sizeof (proplist[0]); p++)
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) ztest_dsl_prop_set_uint64(zd->zd_name, proplist[p],
fa9e4066f08beec538e775443c5be79dd423fcabahrens ztest_random_dsl_prop(proplist[p]), (int)ztest_random(2));
fa9e4066f08beec538e775443c5be79dd423fcabahrens/* ARGSUSED */
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick (void) ztest_spa_prop_set_uint64(ZPOOL_PROP_DEDUPDITTO,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ZIO_DEDUPDITTO_MIN + ztest_random(ZIO_DEDUPDITTO_MIN));
fa9e4066f08beec538e775443c5be79dd423fcabahrensuser_release_one(const char *snapname, const char *holdname)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Test snapshot hold/release and deferred destroy.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_dmu_snapshot_hold(ztest_ds_t *zd, uint64_t id)
47cb52daa729f19e298c85a84e8df069365c5232Jeff Bonwick (void) snprintf(snapname, sizeof (snapname), "sh1_%llu", id);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick (void) snprintf(fullname, sizeof (fullname), "%s@%s", osname, snapname);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick (void) snprintf(tag, sizeof (tag), "tag_%llu", id);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Clean up from any previous run.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Create snapshot, clone it, mark snap for deferred destroy,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * destroy clone, verify snap was also destroyed.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick fatal(0, "dmu_objset_clone(%s) = %d", clonename, error);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick error = dsl_destroy_snapshot(fullname, B_TRUE);
3f9d6ad73e45c6823b409f93b0c8d4f62861d2d5Lin Ling fatal(0, "dsl_destroy_head(%s) = %d", clonename, error);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick fatal(0, "dmu_objset_hold(%s) = %d", fullname, error);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Create snapshot, add temporary hold, verify that we can't
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * destroy a held snapshot, mark for deferred destroy,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * release hold, verify snapshot was destroyed.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick error = dmu_objset_snapshot_one(osname, snapname);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick fatal(0, "dmu_objset_snapshot(%s) = %d", fullname, error);
fa9e4066f08beec538e775443c5be79dd423fcabahrens fatal(0, "dsl_dataset_user_hold(%s)", fullname, tag);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick error = dsl_destroy_snapshot(fullname, B_TRUE);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick fatal(0, "dsl_destroy_snapshot(%s, B_TRUE) = %d",
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick fatal(0, "user_release_one(%s, %s) = %d", fullname, tag, error);
fa9e4066f08beec538e775443c5be79dd423fcabahrens VERIFY3U(dmu_objset_hold(fullname, FTAG, &origin), ==, ENOENT);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Inject random faults into the on-disk data.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick/* ARGSUSED */
8654d0253136055bd4cc2423d87378e8a37f2eb5perrin int bshift = SPA_MAXBLOCKSHIFT + 2; /* don't scrog all labels */
8654d0253136055bd4cc2423d87378e8a37f2eb5perrin leaves = MAX(zs->zs_mirrors, 1) * ztest_opts.zo_raidz;
8654d0253136055bd4cc2423d87378e8a37f2eb5perrin * We need SCL_STATE here because we're going to look at vd0->vdev_tsd.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Inject errors on a normal data device or slog device.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Generate paths to the first leaf in this top-level vdev,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * and to the random leaf we selected. We'll induce transient
fa9e4066f08beec538e775443c5be79dd423fcabahrens * write failures and random online/offline activity on leaf 0,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * and we'll write random garbage to the randomly chosen leaf.
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) snprintf(path0, sizeof (path0), ztest_dev_template,
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) snprintf(pathrand, sizeof (pathrand), ztest_dev_template,
fa9e4066f08beec538e775443c5be79dd423fcabahrens vd0 = vdev_lookup_by_path(spa->spa_root_vdev, path0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Make vd0 explicitly claim to be unreadable,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * or unwriteable, or reach behind its back
fa9e4066f08beec538e775443c5be79dd423fcabahrens * and close the underlying fd. We can do this if
fa9e4066f08beec538e775443c5be79dd423fcabahrens * maxfaults == 0 because we'll fail and reexecute,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * and we can do it if maxfaults >= 2 because we'll
fa9e4066f08beec538e775443c5be79dd423fcabahrens * have enough redundancy. If maxfaults == 1, the
fa9e4066f08beec538e775443c5be79dd423fcabahrens * combination of this with injection of random data
fa9e4066f08beec538e775443c5be79dd423fcabahrens * corruption below exceeds the pool's fault tolerance.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Inject errors on an l2cache device.
8ad4d6dd86f5bc65fb3afa566c8133f3bac21648Jeff Bonwick maxfaults = INT_MAX; /* no limit on cache devices */
fa9e4066f08beec538e775443c5be79dd423fcabahrens * If we can tolerate two or more faults, or we're dealing
fa9e4066f08beec538e775443c5be79dd423fcabahrens * with a slog, randomly online/offline vd0.
8ad4d6dd86f5bc65fb3afa566c8133f3bac21648Jeff Bonwick * We have to grab the zs_name_lock as writer to
8ad4d6dd86f5bc65fb3afa566c8133f3bac21648Jeff Bonwick * prevent a race between offlining a slog and
8ad4d6dd86f5bc65fb3afa566c8133f3bac21648Jeff Bonwick * destroying a dataset. Offlining the slog will
8ad4d6dd86f5bc65fb3afa566c8133f3bac21648Jeff Bonwick * grab a reference on the dataset which may cause
8ad4d6dd86f5bc65fb3afa566c8133f3bac21648Jeff Bonwick * dmu_objset_destroy() to fail with EBUSY thus
8ad4d6dd86f5bc65fb3afa566c8133f3bac21648Jeff Bonwick * leaving the dataset in an inconsistent state.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY(vdev_offline(spa, guid0, flags) != EBUSY);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Ideally we would like to be able to randomly
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * call vdev_[on|off]line without holding locks
fa9e4066f08beec538e775443c5be79dd423fcabahrens * to force unpredictable failures but the side
fa9e4066f08beec538e775443c5be79dd423fcabahrens * effects of vdev_[on|off]line prevent us from
fa9e4066f08beec538e775443c5be79dd423fcabahrens * doing so. We grab the ztest_vdev_lock here to
fa9e4066f08beec538e775443c5be79dd423fcabahrens * prevent a race between injection testing and
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * aux_vdev removal.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * We have at least single-fault tolerance, so inject data corruption.
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (fd == -1) /* we hit a gap in the device namespace */
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick while (--iters != 0) {
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick offset = ztest_random(fsize / (leaves << bshift)) *
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick if (pwrite(fd, &bad, sizeof (bad), offset) != sizeof (bad))
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick " offset 0x%llx\n", pathrand, (u_longlong_t)offset);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Verify that DDT repair works as expected.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick uint64_t object, blocksize, txg, pattern, psize;
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick enum zio_checksum checksum = spa_dedup_checksum(spa);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick blocksize = MIN(blocksize, 2048); /* because we write so many */
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ztest_od_init(&od[0], id, FTAG, 0, DMU_OT_UINT64_OTHER, blocksize, 0);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick if (ztest_object_init(zd, od, sizeof (od), B_FALSE) != 0)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Take the name lock as writer to prevent anyone else from changing
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * the pool and dataset properies we need to maintain during this test.
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_DEDUP, checksum,
fa9e4066f08beec538e775443c5be79dd423fcabahrens ztest_dsl_prop_set_uint64(zd->zd_name, ZFS_PROP_COPIES, 1,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick dmu_tx_hold_write(tx, object, 0, copies * blocksize);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Write all the copies of our block.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick for (int i = 0; i < copies; i++) {
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick int error = dmu_buf_hold(os, object, offset, FTAG, &db,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick fatal(B_FALSE, "dmu_buf_hold(%p, %llu, %llu) = %u",
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ASSERT(ztest_pattern_match(db->db_data, db->db_size, pattern) ||
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ztest_pattern_match(db->db_data, db->db_size, 0ULL));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ztest_pattern_set(db->db_data, db->db_size, pattern);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Find out what block we got.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Damage the block. Dedup-ditto will save us when we read it later.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick (void) zio_wait(zio_rewrite(NULL, spa, 0, &blk,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick buf, psize, NULL, NULL, ZIO_PRIORITY_SYNC_WRITE,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ZIO_FLAG_CANFAIL | ZIO_FLAG_INDUCE_DAMAGE, NULL));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Scrub the pool.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick/* ARGSUSED */
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick (void) poll(NULL, 0, 100); /* wait a moment, then force a restart */
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Change the guid for the pool.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick/* ARGSUSED */
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick (u_longlong_t)orig, (u_longlong_t)spa_guid(spa));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Rename the pool to a different name and then rename it back.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick/* ARGSUSED */
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick newname = umem_alloc(strlen(oldname) + 5, UMEM_NOFAIL);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Do the rename
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Try to open it under the old name, which shouldn't exist
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Open it under the new name and make sure it's still the same spa_t.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY3U(0, ==, spa_open(newname, &spa, FTAG));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Rename it back to the original
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Make sure it can still be opened
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY3U(0, ==, spa_open(oldname, &spa, FTAG));
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Verify pool integrity by running zdb.
0a4e9518a44f226be6d39383330b5b1792d2f184gw /* zdb lives in /usr/sbin, while ztest lives in /usr/bin */
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick /* LINTED */
8ad4d6dd86f5bc65fb3afa566c8133f3bac21648Jeff Bonwick (void) printf("Executing %s\n", strstr(zdb, "zdb "));
fa9e4066f08beec538e775443c5be79dd423fcabahrens fatal(0, "'%s' exit code %d", zdb, WEXITSTATUS(status));
fa9e4066f08beec538e775443c5be79dd423fcabahrens fatal(0, "'%s' died with signal %d", zdb, WTERMSIG(status));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick (void) printf("import/export: old = %s, new = %s\n",
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Clean up from previous runs.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Get the pool's configuration and guid.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY3U(0, ==, spa_open(oldname, &spa, FTAG));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Kick off a scrub to tickle scrub/export races.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ztest_walk_pool_directory("pools before export");
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Export it.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY3U(0, ==, spa_export(oldname, &config, B_FALSE, B_FALSE));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ztest_walk_pool_directory("pools after export");
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Try to import it.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Import it under the new name.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick fatal(B_FALSE, "couldn't import pool %s as %s: error %u",
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ztest_walk_pool_directory("pools after import");
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Try to import it again -- should fail with EEXIST.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY3U(EEXIST, ==, spa_import(newname, config, NULL, 0));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Try to import it under a different name -- should fail with EEXIST.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY3U(EEXIST, ==, spa_import(oldname, config, NULL, 0));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Verify that the pool is no longer visible under the old name.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG));
2215e990accbb747e6562e1cf14b772c9341dd35Mark J Musante * Verify that we can open and close the pool using the new name.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY3U(0, ==, spa_open(newname, &spa, FTAG));
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick if (spa_suspended(spa) && ztest_opts.zo_verbose >= 6)
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick (void) printf("resuming from suspended state\n");
fa9e4066f08beec538e775443c5be79dd423fcabahrensstatic void *
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickstatic void *
fa9e4066f08beec538e775443c5be79dd423fcabahrens delta = (zs->zs_thread_stop - zs->zs_thread_start) / NANOSEC + grace;
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick fatal(0, "failed to complete within %d seconds of deadline", grace);
e05725b117836db173257fae43fb0746eb857fb5bonwickztest_execute(int test, ztest_info_t *zi, uint64_t id)
e05725b117836db173257fae43fb0746eb857fb5bonwick ztest_ds_t *zd = &ztest_ds[id % ztest_opts.zo_datasets];
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ztest_shared_callstate_t *zc = ZTEST_GET_SHARED_CALLSTATE(test);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia (void) dladdr((void *)zi->zi_func, &dli);
d20e665c84abf083a9e8b62cca93383ecb55afdfRicardo M. Correia (double)functime / NANOSEC, dli.dli_sname);
fa9e4066f08beec538e775443c5be79dd423fcabahrensstatic void *
fa9e4066f08beec538e775443c5be79dd423fcabahrens * See if it's time to force a crash.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * If we're getting ENOSPC with some regularity, stop.
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante * Pick a random function to execute.
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante atomic_cas_64(&zc->zc_next, call_next, call_next +
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante ztest_random(2 * zi->zi_interval[0] + 1)) == call_next) {
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickztest_dataset_name(char *dsname, char *pool, int d)
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) snprintf(dsname, MAXNAMELEN, "%s/ds_%d", pool, d);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick ztest_dataset_name(name, ztest_opts.zo_pool, d);
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) printf("Destroying %s to free up space\n", name);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Cleanup any non-standard clones and snapshots. In general,
88ecc943b4eb72f7c4fbbd8435997b85ef171fc3George Wilson * ztest thread t operates on dataset (t % zopt_datasets),
1195e687f1c03c8d57417b5999578922e20a3554Mark J Musante * so there may be more than one thing to clean up.
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) dmu_objset_find(name, ztest_objset_destroy_cb, NULL,
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * ZTEST_DIROBJ is the object directory for the entire dataset.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * Therefore, the number of objects in use should equal the
fa9e4066f08beec538e775443c5be79dd423fcabahrens * number of ZTEST_DIROBJ entries, +1 for ZTEST_DIROBJ itself.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * If not, we have an object leak.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Note that we can only check this in ztest_dataset_open(),
fa9e4066f08beec538e775443c5be79dd423fcabahrens * when the open-context and syncing-context values agree.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * That's because zap_count() returns the open-context value,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * while dmu_objset_space() returns the rootbp fill count.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick VERIFY3U(0, ==, zap_count(zd->zd_os, ZTEST_DIROBJ, &dirobjs));
fa9e4066f08beec538e775443c5be79dd423fcabahrens dmu_objset_space(zd->zd_os, &scratch, &scratch, &usedobjs, &scratch);
fa9e4066f08beec538e775443c5be79dd423fcabahrens uint64_t committed_seq = ZTEST_GET_SHARED_DS(d)->zd_seq;
fa9e4066f08beec538e775443c5be79dd423fcabahrens VERIFY0(dmu_objset_own(name, DMU_OST_OTHER, B_FALSE, zd, &os));
fa9e4066f08beec538e775443c5be79dd423fcabahrens fatal(0, "missing log records: claimed %llu < committed %llu",
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick zilog->zl_header->zh_claim_lr_seq, committed_seq);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick (void) printf("%s replay %llu blocks, %llu records, seq %llu\n",
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick fatal(0, "missing log records: replayed %llu < committed %llu",
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Kick off threads to run tests on all datasets in parallel.
004388ebfdfe2ed7dfd2d153a876dfcc22d2c006casper * Initialize parent/child shared state.
fa9e4066f08beec538e775443c5be79dd423fcabahrens VERIFY(_mutex_init(&ztest_vdev_lock, USYNC_THREAD, NULL) == 0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens VERIFY(rwlock_init(&ztest_name_lock, USYNC_THREAD, NULL) == 0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens zs->zs_thread_start + ztest_opts.zo_passtime * NANOSEC;
fa9e4066f08beec538e775443c5be79dd423fcabahrens zs->zs_thread_stop = MIN(zs->zs_thread_stop, zs->zs_proc_stop);
5ad820458efd0fdb914baff9c1447c22b819fa23nd (void) _mutex_init(&zcl.zcl_callbacks_lock, USYNC_THREAD, NULL);
fa9e4066f08beec538e775443c5be79dd423fcabahrens list_create(&zcl.zcl_callbacks, sizeof (ztest_cb_data_t),
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Open our pool.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * We don't expect the pool to suspend unless maxfaults == 0,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * in which case ztest_fault_inject() temporarily takes away
fa9e4066f08beec538e775443c5be79dd423fcabahrens * the only valid replica.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Create a thread to periodically resume suspended I/O.
fa9e4066f08beec538e775443c5be79dd423fcabahrens VERIFY(thr_create(0, 0, ztest_resume_thread, spa, THR_BOUND,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Create a deadman thread to abort() if we hang.
fa9e4066f08beec538e775443c5be79dd423fcabahrens VERIFY(thr_create(0, 0, ztest_deadman_thread, zs, THR_BOUND,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Verify that we can safely inquire about about any object,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * whether it's allocated or not. To make it interesting,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * we probe a 5-wide window around each power of two.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * This hits all edge cases, including zero and the max.
fa9e4066f08beec538e775443c5be79dd423fcabahrens for (int t = 0; t < 64; t++) {
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * If we got any ENOSPC errors on the previous run, destroy something.
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick tid = umem_zalloc(ztest_opts.zo_threads * sizeof (thread_t),
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Kick off all the tests that run in parallel.
fa9e4066f08beec538e775443c5be79dd423fcabahrens VERIFY(thr_create(0, 0, ztest_thread, (void *)(uintptr_t)t,
kernel_fini();
ztest_freeze(void)
int numloops = 0;
kernel_fini();
kernel_fini();
static nvlist_t *
return (props);
return (props);
for (int i = 0; i < SPA_FEATURES; i++) {
kernel_fini();
ztest_freeze();
setup_data_fd(void)
int size;
return (size);
setup_hdr(void)
int size;
setup_data(void)
static boolean_t
int status;
return (B_FALSE);
return (B_TRUE);
ztest_run_init(void)
* Blow away any existing copy of zpool.cache
int kills = 0;
int iters = 0;
int older = 0;
int newer = 0;
char *cmd;
if (!fd_data_str) {
setup_hdr();
setup_data();
sizeof (*ztest_shared_opts));
setup_data();
/* Override location of zpool.cache */
if (fd_data_str) {
exit(0);
for (int f = 0; f < ZTEST_FUNCS; f++) {
int status;
for (int f = 0; f < ZTEST_FUNCS; f++) {
cmd);
newer++;
older++;
if (killed)
kills++;
iters++;
for (int f = 0; f < ZTEST_FUNCS; f++) {
kernel_fini();
kernel_fini();
if (hasalt) {
cmd);