#include <sys/zio.h>
#include <sys/zil.h>
#include <sys/zil_impl.h>
-#include <sys/zfs_rlock.h>
#include <sys/vdev_impl.h>
#include <sys/vdev_file.h>
#include <sys/spa_impl.h>
#include <sys/fs/zfs.h>
#include <zfs_fletcher.h>
#include <libnvpair.h>
-#include <libzfs.h>
+#include <libzutil.h>
#include <sys/crypto/icp.h>
#ifdef __GLIBC__
#include <execinfo.h> /* for backtrace() */
uint64_t zo_metaslab_force_ganging;
int zo_mmp_test;
int zo_special_vdevs;
+ int zo_dump_dbgmsg;
} ztest_shared_opts_t;
static const ztest_shared_opts_t ztest_opts_defaults = {
.zo_init = 1,
.zo_time = 300, /* 5 minutes */
.zo_maxloops = 50, /* max loops during spa_freeze() */
- .zo_metaslab_force_ganging = 32 << 10,
+ .zo_metaslab_force_ganging = 64 << 10,
.zo_special_vdevs = ZTEST_VDEV_CLASS_RND,
};
uint64_t bw_data;
} bufwad_t;
+/*
+ * It would be better to use a rangelock_t per object. Unfortunately
+ * the rangelock_t is not a drop-in replacement for rl_t, because we
+ * still need to map from object ID to rangelock_t.
+ */
+typedef enum {
+ RL_READER,
+ RL_WRITER,
+ RL_APPEND
+} rl_type_t;
+
typedef struct rll {
void *rll_writer;
int rll_readers;
kcondvar_t rll_cv;
} rll_t;
-typedef struct zll {
- list_t z_list;
- kmutex_t z_lock;
-} zll_t;
+typedef struct rl {
+ uint64_t rl_object;
+ uint64_t rl_offset;
+ uint64_t rl_size;
+ rll_t *rl_lock;
+} rl_t;
#define ZTEST_RANGE_LOCKS 64
#define ZTEST_OBJECT_LOCKS 64
char zd_name[ZFS_MAX_DATASET_NAME_LEN];
kmutex_t zd_dirobj_lock;
rll_t zd_object_lock[ZTEST_OBJECT_LOCKS];
- zll_t zd_range_lock[ZTEST_RANGE_LOCKS];
+ rll_t zd_range_lock[ZTEST_RANGE_LOCKS];
} ztest_ds_t;
/*
ztest_func_t ztest_ddt_repair;
ztest_func_t ztest_dmu_snapshot_hold;
ztest_func_t ztest_mmp_enable_disable;
-ztest_func_t ztest_spa_rename;
ztest_func_t ztest_scrub;
ztest_func_t ztest_dsl_dataset_promote_busy;
ztest_func_t ztest_vdev_attach_detach;
ZTI_INIT(ztest_dmu_snapshot_hold, 1, &zopt_sometimes),
ZTI_INIT(ztest_mmp_enable_disable, 1, &zopt_sometimes),
ZTI_INIT(ztest_reguid, 1, &zopt_rarely),
- ZTI_INIT(ztest_spa_rename, 1, &zopt_rarely),
ZTI_INIT(ztest_scrub, 1, &zopt_rarely),
ZTI_INIT(ztest_spa_upgrade, 1, &zopt_rarely),
ZTI_INIT(ztest_dsl_dataset_promote_busy, 1, &zopt_rarely),
static pthread_rwlock_t ztest_name_lock;
static boolean_t ztest_dump_core = B_TRUE;
-static boolean_t ztest_dump_debug_buffer = B_FALSE;
static boolean_t ztest_exiting;
/* Global commit callback list */
static void
dump_debug_buffer(void)
{
- if (!ztest_dump_debug_buffer)
+ ssize_t ret __attribute__((unused));
+
+ if (!ztest_opts.zo_dump_dbgmsg)
return;
- (void) printf("\n");
+ /*
+ * We use write() instead of printf() so that this function
+ * is safe to call from a signal handler.
+ */
+ ret = write(STDOUT_FILENO, "\n", 1);
zfs_dbgmsg_print("ztest");
}
(void) fprintf(stderr, "%s\n", buf);
fatal_msg = buf; /* to ease debugging */
- dump_debug_buffer();
-
if (ztest_dump_core)
abort();
+ else
+ dump_debug_buffer();
+
exit(3);
}
usage(B_FALSE);
break;
case 'G':
- ztest_dump_debug_buffer = B_TRUE;
+ zo->zo_dump_dbgmsg = 1;
break;
case 'h':
usage(B_TRUE);
return (err);
}
-
-/*
- * Object and range lock mechanics
- */
-typedef struct {
- list_node_t z_lnode;
- zfs_refcount_t z_refcnt;
- uint64_t z_object;
- zfs_rlock_t z_range_lock;
-} ztest_znode_t;
-
-typedef struct {
- rl_t *z_rl;
- ztest_znode_t *z_ztznode;
-} ztest_zrl_t;
-
-static ztest_znode_t *
-ztest_znode_init(uint64_t object)
-{
- ztest_znode_t *zp = umem_alloc(sizeof (*zp), UMEM_NOFAIL);
-
- list_link_init(&zp->z_lnode);
- zfs_refcount_create(&zp->z_refcnt);
- zp->z_object = object;
- zfs_rlock_init(&zp->z_range_lock);
-
- return (zp);
-}
-
-static void
-ztest_znode_fini(ztest_znode_t *zp)
-{
- ASSERT(zfs_refcount_is_zero(&zp->z_refcnt));
- zfs_rlock_destroy(&zp->z_range_lock);
- zp->z_object = 0;
- zfs_refcount_destroy(&zp->z_refcnt);
- list_link_init(&zp->z_lnode);
- umem_free(zp, sizeof (*zp));
-}
-
-static void
-ztest_zll_init(zll_t *zll)
-{
- mutex_init(&zll->z_lock, NULL, MUTEX_DEFAULT, NULL);
- list_create(&zll->z_list, sizeof (ztest_znode_t),
- offsetof(ztest_znode_t, z_lnode));
-}
-
-static void
-ztest_zll_destroy(zll_t *zll)
-{
- list_destroy(&zll->z_list);
- mutex_destroy(&zll->z_lock);
-}
-
-#define RL_TAG "range_lock"
-static ztest_znode_t *
-ztest_znode_get(ztest_ds_t *zd, uint64_t object)
-{
- zll_t *zll = &zd->zd_range_lock[object & (ZTEST_OBJECT_LOCKS - 1)];
- ztest_znode_t *zp = NULL;
- mutex_enter(&zll->z_lock);
- for (zp = list_head(&zll->z_list); (zp);
- zp = list_next(&zll->z_list, zp)) {
- if (zp->z_object == object) {
- zfs_refcount_add(&zp->z_refcnt, RL_TAG);
- break;
- }
- }
- if (zp == NULL) {
- zp = ztest_znode_init(object);
- zfs_refcount_add(&zp->z_refcnt, RL_TAG);
- list_insert_head(&zll->z_list, zp);
- }
- mutex_exit(&zll->z_lock);
- return (zp);
-}
-
-static void
-ztest_znode_put(ztest_ds_t *zd, ztest_znode_t *zp)
-{
- zll_t *zll = NULL;
- ASSERT3U(zp->z_object, !=, 0);
- zll = &zd->zd_range_lock[zp->z_object & (ZTEST_OBJECT_LOCKS - 1)];
- mutex_enter(&zll->z_lock);
- zfs_refcount_remove(&zp->z_refcnt, RL_TAG);
- if (zfs_refcount_is_zero(&zp->z_refcnt)) {
- list_remove(&zll->z_list, zp);
- ztest_znode_fini(zp);
- }
- mutex_exit(&zll->z_lock);
-}
-
-
static void
ztest_rll_init(rll_t *rll)
{
ztest_rll_unlock(rll);
}
-static ztest_zrl_t *
-ztest_zrl_init(rl_t *rl, ztest_znode_t *zp)
-{
- ztest_zrl_t *zrl = umem_alloc(sizeof (*zrl), UMEM_NOFAIL);
- zrl->z_rl = rl;
- zrl->z_ztznode = zp;
- return (zrl);
-}
-
-static void
-ztest_zrl_fini(ztest_zrl_t *zrl)
-{
- umem_free(zrl, sizeof (*zrl));
-}
-
-static ztest_zrl_t *
+static rl_t *
ztest_range_lock(ztest_ds_t *zd, uint64_t object, uint64_t offset,
uint64_t size, rl_type_t type)
{
- ztest_znode_t *zp = ztest_znode_get(zd, object);
- rl_t *rl = zfs_range_lock(&zp->z_range_lock, offset,
- size, type);
- return (ztest_zrl_init(rl, zp));
+ uint64_t hash = object ^ (offset % (ZTEST_RANGE_LOCKS + 1));
+ rll_t *rll = &zd->zd_range_lock[hash & (ZTEST_RANGE_LOCKS - 1)];
+ rl_t *rl;
+
+ rl = umem_alloc(sizeof (*rl), UMEM_NOFAIL);
+ rl->rl_object = object;
+ rl->rl_offset = offset;
+ rl->rl_size = size;
+ rl->rl_lock = rll;
+
+ ztest_rll_lock(rll, type);
+
+ return (rl);
}
static void
-ztest_range_unlock(ztest_ds_t *zd, ztest_zrl_t *zrl)
+ztest_range_unlock(rl_t *rl)
{
- zfs_range_unlock(zrl->z_rl);
- ztest_znode_put(zd, zrl->z_ztznode);
- ztest_zrl_fini(zrl);
+ rll_t *rll = rl->rl_lock;
+
+ ztest_rll_unlock(rll);
+
+ umem_free(rl, sizeof (*rl));
}
static void
ztest_rll_init(&zd->zd_object_lock[l]);
for (l = 0; l < ZTEST_RANGE_LOCKS; l++)
- ztest_zll_init(&zd->zd_range_lock[l]);
+ ztest_rll_init(&zd->zd_range_lock[l]);
}
static void
ztest_rll_destroy(&zd->zd_object_lock[l]);
for (l = 0; l < ZTEST_RANGE_LOCKS; l++)
- ztest_zll_destroy(&zd->zd_range_lock[l]);
+ ztest_rll_destroy(&zd->zd_range_lock[l]);
}
#define TXG_MIGHTWAIT (ztest_random(10) == 0 ? TXG_NOWAIT : TXG_WAIT)
dmu_tx_t *tx;
dmu_buf_t *db;
arc_buf_t *abuf = NULL;
- ztest_zrl_t *rl;
+ rl_t *rl;
if (byteswap)
byteswap_uint64_array(lr, sizeof (*lr));
if (abuf != NULL)
dmu_return_arcbuf(abuf);
dmu_buf_rele(db, FTAG);
- ztest_range_unlock(zd, rl);
+ ztest_range_unlock(rl);
ztest_object_unlock(zd, lr->lr_foid);
return (ENOSPC);
}
dmu_tx_commit(tx);
- ztest_range_unlock(zd, rl);
+ ztest_range_unlock(rl);
ztest_object_unlock(zd, lr->lr_foid);
return (0);
objset_t *os = zd->zd_os;
dmu_tx_t *tx;
uint64_t txg;
- ztest_zrl_t *rl;
+ rl_t *rl;
if (byteswap)
byteswap_uint64_array(lr, sizeof (*lr));
txg = ztest_tx_assign(tx, TXG_WAIT, FTAG);
if (txg == 0) {
- ztest_range_unlock(zd, rl);
+ ztest_range_unlock(rl);
ztest_object_unlock(zd, lr->lr_foid);
return (ENOSPC);
}
dmu_tx_commit(tx);
- ztest_range_unlock(zd, rl);
+ ztest_range_unlock(rl);
ztest_object_unlock(zd, lr->lr_foid);
return (0);
/*
* ZIL get_data callbacks
*/
-typedef struct ztest_zgd_private {
- ztest_ds_t *z_zd;
- ztest_zrl_t *z_rl;
- uint64_t z_object;
-} ztest_zgd_private_t;
static void
ztest_get_done(zgd_t *zgd, int error)
{
- ztest_zgd_private_t *zzp = zgd->zgd_private;
- ztest_ds_t *zd = zzp->z_zd;
- uint64_t object = zzp->z_object;
+ ztest_ds_t *zd = zgd->zgd_private;
+ uint64_t object = ((rl_t *)zgd->zgd_lr)->rl_object;
if (zgd->zgd_db)
dmu_buf_rele(zgd->zgd_db, zgd);
- ztest_range_unlock(zd, zzp->z_rl);
+ ztest_range_unlock((rl_t *)zgd->zgd_lr);
ztest_object_unlock(zd, object);
if (error == 0 && zgd->zgd_bp)
zil_lwb_add_block(zgd->zgd_lwb, zgd->zgd_bp);
umem_free(zgd, sizeof (*zgd));
- umem_free(zzp, sizeof (*zzp));
}
static int
dmu_buf_t *db;
zgd_t *zgd;
int error;
- ztest_zgd_private_t *zgd_private;
ASSERT3P(lwb, !=, NULL);
ASSERT3P(zio, !=, NULL);
zgd = umem_zalloc(sizeof (*zgd), UMEM_NOFAIL);
zgd->zgd_lwb = lwb;
- zgd_private = umem_zalloc(sizeof (ztest_zgd_private_t), UMEM_NOFAIL);
- zgd_private->z_zd = zd;
- zgd_private->z_object = object;
- zgd->zgd_private = zgd_private;
+ zgd->zgd_private = zd;
if (buf != NULL) { /* immediate write */
- zgd_private->z_rl = ztest_range_lock(zd, object, offset, size,
- RL_READER);
- zgd->zgd_rl = zgd_private->z_rl->z_rl;
+ zgd->zgd_lr = (struct locked_range *)ztest_range_lock(zd,
+ object, offset, size, RL_READER);
error = dmu_read(os, object, offset, size, buf,
DMU_READ_NO_PREFETCH);
offset = 0;
}
- zgd_private->z_rl = ztest_range_lock(zd, object, offset, size,
- RL_READER);
- zgd->zgd_rl = zgd_private->z_rl->z_rl;
+ zgd->zgd_lr = (struct locked_range *)ztest_range_lock(zd,
+ object, offset, size, RL_READER);
error = dmu_buf_hold(os, object, offset, zgd, &db,
DMU_READ_NO_PREFETCH);
objset_t *os = zd->zd_os;
dmu_tx_t *tx;
uint64_t txg;
- ztest_zrl_t *rl;
+ rl_t *rl;
txg_wait_synced(dmu_objset_pool(os), 0);
(void) dmu_free_long_range(os, object, offset, size);
}
- ztest_range_unlock(zd, rl);
+ ztest_range_unlock(rl);
ztest_object_unlock(zd, object);
}
ztest_device_removal_active = B_TRUE;
mutex_exit(&ztest_vdev_lock);
+ /*
+ * spa->spa_vdev_removal is created in a sync task that
+ * is initiated via dsl_sync_task_nowait(). Since the
+ * task may not run before spa_vdev_remove() returns, we
+ * must wait at least 1 txg to ensure that the removal
+ * struct has been created.
+ */
+ txg_wait_synced(spa_get_dsl(spa), 0);
+
while (spa->spa_vdev_removal != NULL)
txg_wait_synced(spa_get_dsl(spa), 0);
} else {
VERIFY3U(load, ==, spa_load_guid(spa));
}
-/*
- * Rename the pool to a different name and then rename it back.
- */
-/* ARGSUSED */
-void
-ztest_spa_rename(ztest_ds_t *zd, uint64_t id)
-{
- char *oldname, *newname;
- spa_t *spa;
-
- if (ztest_opts.zo_mmp_test)
- return;
-
- (void) pthread_rwlock_wrlock(&ztest_name_lock);
-
- oldname = ztest_opts.zo_pool;
- newname = umem_alloc(strlen(oldname) + 5, UMEM_NOFAIL);
- (void) strcpy(newname, oldname);
- (void) strcat(newname, "_tmp");
-
- /*
- * Do the rename
- */
- VERIFY3U(0, ==, spa_rename(oldname, newname));
-
- /*
- * Try to open it under the old name, which shouldn't exist
- */
- VERIFY3U(ENOENT, ==, spa_open(oldname, &spa, FTAG));
-
- /*
- * Open it under the new name and make sure it's still the same spa_t.
- */
- VERIFY3U(0, ==, spa_open(newname, &spa, FTAG));
-
- ASSERT(spa == ztest_spa);
- spa_close(spa, FTAG);
-
- /*
- * Rename it back to the original
- */
- VERIFY3U(0, ==, spa_rename(newname, oldname));
-
- /*
- * Make sure it can still be opened
- */
- VERIFY3U(0, ==, spa_open(oldname, &spa, FTAG));
-
- ASSERT(spa == ztest_spa);
- spa_close(spa, FTAG);
-
- umem_free(newname, strlen(newname) + 1);
-
- (void) pthread_rwlock_unlock(&ztest_name_lock);
-}
-
void
ztest_fletcher(ztest_ds_t *zd, uint64_t id)
{
{
ztest_shared_t *zs = arg;
spa_t *spa = ztest_spa;
- hrtime_t delta, overdue, total = 0;
+ hrtime_t delay, overdue, last_run = gethrtime();
- for (;;) {
- delta = zs->zs_thread_stop - zs->zs_thread_start +
- MSEC2NSEC(zfs_deadman_synctime_ms);
+ delay = (zs->zs_thread_stop - zs->zs_thread_start) +
+ MSEC2NSEC(zfs_deadman_synctime_ms);
- (void) poll(NULL, 0, (int)NSEC2MSEC(delta));
+ while (!ztest_exiting) {
+ /*
+ * Wait for the delay timer while checking occasionally
+ * if we should stop.
+ */
+ if (gethrtime() < last_run + delay) {
+ (void) poll(NULL, 0, 1000);
+ continue;
+ }
/*
* If the pool is suspended then fail immediately. Otherwise,
* then it may be hung and is terminated.
*/
overdue = zs->zs_proc_stop + MSEC2NSEC(zfs_deadman_synctime_ms);
- total += zfs_deadman_synctime_ms / 1000;
if (gethrtime() > overdue) {
fatal(0, "aborting test after %llu seconds because "
- "the process is overdue for termination.", total);
+ "the process is overdue for termination.",
+ (gethrtime() - zs->zs_proc_start) / NANOSEC);
}
(void) printf("ztest has been running for %lld seconds\n",
- total);
+ (gethrtime() - zs->zs_proc_start) / NANOSEC);
+
+ last_run = gethrtime();
+ delay = MSEC2NSEC(zfs_deadman_checktime_ms);
}
+
+ thread_exit();
}
static void
{
spa_t *spa;
objset_t *os;
- kthread_t *resume_thread;
+ kthread_t *resume_thread, *deadman_thread;
kthread_t **run_threads;
uint64_t object;
int error;
/*
* Create a deadman thread and set to panic if we hang.
*/
- (void) thread_create(NULL, 0, ztest_deadman_thread,
+ deadman_thread = thread_create(NULL, 0, ztest_deadman_thread,
zs, 0, NULL, TS_RUN | TS_JOINABLE, defclsyspri);
spa->spa_deadman_failmode = ZIO_FAILURE_MODE_PANIC;
umem_free(run_threads, ztest_opts.zo_threads * sizeof (kthread_t *));
- /* Kill the resume thread */
+ /* Kill the resume and deadman threads */
ztest_exiting = B_TRUE;
VERIFY0(thread_join(resume_thread));
+ VERIFY0(thread_join(deadman_thread));
ztest_resume(spa);
/*
static void
ztest_import(ztest_shared_t *zs)
{
- libzfs_handle_t *hdl;
importargs_t args = { 0 };
spa_t *spa;
nvlist_t *cfg = NULL;
VERIFY0(pthread_rwlock_init(&ztest_name_lock, NULL));
kernel_init(FREAD | FWRITE);
- hdl = libzfs_init();
searchdirs[0] = ztest_opts.zo_dir;
args.paths = nsearch;
args.path = searchdirs;
args.can_be_active = B_FALSE;
- error = zpool_tryimport(hdl, name, &cfg, &args);
+ error = zpool_find_config(NULL, name, &cfg, &args,
+ &libzpool_config_ops);
if (error)
(void) fatal(0, "No pools found\n");
1ULL << spa->spa_root_vdev->vdev_child[0]->vdev_ms_shift;
spa_close(spa, FTAG);
- libzfs_fini(hdl);
kernel_fini();
if (!ztest_opts.zo_mmp_test) {
ztest_shared_callstate_t *zc;
char timebuf[100];
char numbuf[NN_NUMBUF_SZ];
- spa_t *spa;
char *cmd;
boolean_t hasalt;
int f;
dprintf_setup(&argc, argv);
zfs_deadman_synctime_ms = 300000;
+ zfs_deadman_checktime_ms = 30000;
/*
* As two-word space map entries may not come up often (especially
* if pool and vdev sizes are small) we want to force at least some
* Verify that even extensively damaged split blocks with many
* segments can be reconstructed in a reasonable amount of time
* when reconstruction is known to be possible.
+ *
+ * Note: the lower this value is, the more damage we inflict, and
+ * the more time ztest spends in recovering that damage. We chose
+ * to induce damage 1/100th of the time so recovery is tested but
+ * not so frequently that ztest doesn't get to test other code paths.
*/
- zfs_reconstruct_indirect_damage_fraction = 4;
+ zfs_reconstruct_indirect_damage_fraction = 100;
action.sa_handler = sig_handler;
sigemptyset(&action.sa_mask);
(void) printf("\n");
}
- /*
- * It's possible that we killed a child during a rename test,
- * in which case we'll have a 'ztest_tmp' pool lying around
- * instead of 'ztest'. Do a blind rename in case this happened.
- */
- kernel_init(FREAD);
- if (spa_open(ztest_opts.zo_pool, &spa, FTAG) == 0) {
- spa_close(spa, FTAG);
- } else {
- char tmpname[ZFS_MAX_DATASET_NAME_LEN];
- kernel_fini();
- kernel_init(FREAD | FWRITE);
- (void) snprintf(tmpname, sizeof (tmpname), "%s_tmp",
- ztest_opts.zo_pool);
- (void) spa_rename(tmpname, ztest_opts.zo_pool);
- }
- kernel_fini();
-
if (!ztest_opts.zo_mmp_test)
ztest_run_zdb(ztest_opts.zo_pool);
}