int
spa_history_log_nvl(spa_t *spa, nvlist_t *nvl)
{
    int err = 0;
    dmu_tx_t *tx;
    nvlist_t *nvarg;

    if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY || !spa_writeable(spa))
        return (SET_ERROR(EINVAL));

    tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
    err = dmu_tx_assign(tx, TXG_WAIT);
    if (err) {
        dmu_tx_abort(tx);
        return (err);
    }

    VERIFY0(nvlist_dup(nvl, &nvarg, KM_SLEEP));
    if (spa_history_zone() != NULL) {
        fnvlist_add_string(nvarg, ZPOOL_HIST_ZONE,
                           spa_history_zone());
    }
    fnvlist_add_uint64(nvarg, ZPOOL_HIST_WHO, crgetruid(CRED()));

    /* Kick this off asynchronously; errors are ignored. */
    dsl_sync_task_nowait(spa_get_dsl(spa), spa_history_log_sync,
                         nvarg, 0, ZFS_SPACE_CHECK_NONE, tx);
    dmu_tx_commit(tx);

    /* spa_history_log_sync will free nvl */
    return (err);

}

/*
 * The nvlist will be consumed by this call.
 */
static void
log_internal(nvlist_t *nvl, const char *operation, spa_t *spa,
             dmu_tx_t *tx, const char *fmt, va_list adx)
{
    char *msg;

    /*
     * If this is part of creating a pool, not everything is
     * initialized yet, so don't bother logging the internal events.
     * Likewise if the pool is not writeable.
     */
    if (tx->tx_txg == TXG_INITIAL || !spa_writeable(spa)) {
        fnvlist_free(nvl);
        return;
    }

    msg = kmem_vasprintf(fmt, adx);
    fnvlist_add_string(nvl, ZPOOL_HIST_INT_STR, msg);
    strfree(msg);

    fnvlist_add_string(nvl, ZPOOL_HIST_INT_NAME, operation);
    fnvlist_add_uint64(nvl, ZPOOL_HIST_TXG, tx->tx_txg);

    if (dmu_tx_is_syncing(tx)) {
        spa_history_log_sync(nvl, tx);
    } else {
        dsl_sync_task_nowait(spa_get_dsl(spa),
                             spa_history_log_sync, nvl, 0, ZFS_SPACE_CHECK_NONE, tx);
    }
    /* spa_history_log_sync() will free nvl */
}

void
dsl_sync_task_do_nowait(dsl_pool_t *dp,
    dsl_checkfunc_t *checkfunc, dsl_syncfunc_t *syncfunc,
    void *arg1, void *arg2, int blocks_modified, dmu_tx_t *tx)
{
	dsl_sync_task_group_t *dstg;

	if (!spa_writeable(dp->dp_spa))
		return;

	dstg = dsl_sync_task_group_create(dp);
	dsl_sync_task_create(dstg, checkfunc, syncfunc,
	    arg1, arg2, blocks_modified);
	dsl_sync_task_group_nowait(dstg, tx);
}

static int
zvol_first_open(zvol_state_t *zv)
{
	objset_t *os;
	uint64_t volsize;
	int error;
	uint64_t ro;

	/* lie and say we're read-only */
	error = dmu_objset_own(zv->zv_name, DMU_OST_ZVOL, 1, zvol_tag, &os);
	if (error)
		return (SET_ERROR(-error));

	zv->zv_objset = os;

	error = dsl_prop_get_integer(zv->zv_name, "readonly", &ro, NULL);
	if (error)
		goto out_owned;

	error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize);
	if (error)
		goto out_owned;

	error = dmu_bonus_hold(os, ZVOL_OBJ, zvol_tag, &zv->zv_dbuf);
	if (error)
		goto out_owned;

	set_capacity(zv->zv_disk, volsize >> 9);
	zv->zv_volsize = volsize;
	zv->zv_zilog = zil_open(os, zvol_get_data);

	if (ro || dmu_objset_is_snapshot(os) ||
	    !spa_writeable(dmu_objset_spa(os))) {
		set_disk_ro(zv->zv_disk, 1);
		zv->zv_flags |= ZVOL_RDONLY;
	} else {
		set_disk_ro(zv->zv_disk, 0);
		zv->zv_flags &= ~ZVOL_RDONLY;
	}

out_owned:
	if (error) {
		dmu_objset_disown(os, zvol_tag);
		zv->zv_objset = NULL;
	}

	return (SET_ERROR(-error));
}

void
mmp_thread_start(spa_t *spa)
{
	mmp_thread_t *mmp = &spa->spa_mmp;

	if (spa_writeable(spa)) {
		mutex_enter(&mmp->mmp_thread_lock);
		if (!mmp->mmp_thread) {
			dprintf("mmp_thread_start pool %s\n",
			    spa->spa_name);
			mmp->mmp_thread = thread_create(NULL, 0, mmp_thread,
			    spa, 0, &p0, TS_RUN, defclsyspri);
		}
		mutex_exit(&mmp->mmp_thread_lock);
	}
}

int
dsl_sync_task_do(dsl_pool_t *dp,
    dsl_checkfunc_t *checkfunc, dsl_syncfunc_t *syncfunc,
    void *arg1, void *arg2, int blocks_modified)
{
	dsl_sync_task_group_t *dstg;
	int err;

	ASSERT(spa_writeable(dp->dp_spa));

	dstg = dsl_sync_task_group_create(dp);
	dsl_sync_task_create(dstg, checkfunc, syncfunc,
	    arg1, arg2, blocks_modified);
	err = dsl_sync_task_group_wait(dstg);
	dsl_sync_task_group_destroy(dstg);
	return (err);
}

static void
traverse_zil(traverse_data_t *td, zil_header_t *zh)
{
	uint64_t claim_txg = zh->zh_claim_txg;
	zilog_t *zilog;

	/*
	 * We only want to visit blocks that have been claimed but not yet
	 * replayed; plus, in read-only mode, blocks that are already stable.
	 */
	if (claim_txg == 0 && spa_writeable(td->td_spa))
		return;

	zilog = zil_alloc(spa_get_dsl(td->td_spa)->dp_meta_objset, zh);

	(void) zil_parse(zilog, traverse_zil_block, traverse_zil_record, td,
	    claim_txg);

	zil_free(zilog);
}

static void
dsl_scan_zil(dsl_pool_t *dp, zil_header_t *zh)
{
	uint64_t claim_txg = zh->zh_claim_txg;
	zil_scan_arg_t zsa = { dp, zh };
	zilog_t *zilog;

	/*
	 * We only want to visit blocks that have been claimed but not yet
	 * replayed (or, in read-only mode, blocks that *would* be claimed).
	 */
	if (claim_txg == 0 && spa_writeable(dp->dp_spa))
		return;

	zilog = zil_alloc(dp->dp_meta_objset, zh);

	(void) zil_parse(zilog, dsl_scan_zil_block, dsl_scan_zil_record, &zsa,
	    claim_txg);

	zil_free(zilog);
}

/*
 * The nvlist will be consumed by this call.
 */
static void
log_internal(nvlist_t *nvl, const char *operation, spa_t *spa,
    dmu_tx_t *tx, const char *fmt, va_list adx)
{
	char *msg;
	va_list adx1;
	int size;

	/*
	 * If this is part of creating a pool, not everything is
	 * initialized yet, so don't bother logging the internal events.
	 * Likewise if the pool is not writeable.
	 */
	if (tx->tx_txg == TXG_INITIAL || !spa_writeable(spa)) {
		fnvlist_free(nvl);
		return;
	}

	va_copy(adx1, adx);
	size = vsnprintf(NULL, 0, fmt, adx1) + 1;
	msg = kmem_alloc(size, KM_PUSHPAGE);
	va_end(adx1);
	va_copy(adx1, adx);
	(void) vsprintf(msg, fmt, adx1);
	va_end(adx1);
	fnvlist_add_string(nvl, ZPOOL_HIST_INT_STR, msg);
	kmem_free(msg, size);

	fnvlist_add_string(nvl, ZPOOL_HIST_INT_NAME, operation);
	fnvlist_add_uint64(nvl, ZPOOL_HIST_TXG, tx->tx_txg);

	if (dmu_tx_is_syncing(tx)) {
		spa_history_log_sync(nvl, tx);
	} else {
		dsl_sync_task_nowait(spa_get_dsl(spa),
		    spa_history_log_sync, nvl, 0, tx);
	}
	/* spa_history_log_sync() will free nvl */
}

int
zfs_sb_setup(zfs_sb_t *zsb, boolean_t mounting)
{
	int error;

	error = zfs_register_callbacks(zsb);
	if (error)
		return (error);

	/*
	 * Set the objset user_ptr to track its zsb.
	 */
	mutex_enter(&zsb->z_os->os_user_ptr_lock);
	dmu_objset_set_user(zsb->z_os, zsb);
	mutex_exit(&zsb->z_os->os_user_ptr_lock);

	zsb->z_log = zil_open(zsb->z_os, zfs_get_data);

	/*
	 * If we are not mounting (ie: online recv), then we don't
	 * have to worry about replaying the log as we blocked all
	 * operations out since we closed the ZIL.
	 */
	if (mounting) {
		boolean_t readonly;

		/*
		 * During replay we remove the read only flag to
		 * allow replays to succeed.
		 */
		readonly = zfs_is_readonly(zsb);
		if (readonly != 0)
			readonly_changed_cb(zsb, B_FALSE);
		else
			zfs_unlinked_drain(zsb);

		/*
		 * Parse and replay the intent log.
		 *
		 * Because of ziltest, this must be done after
		 * zfs_unlinked_drain().  (Further note: ziltest
		 * doesn't use readonly mounts, where
		 * zfs_unlinked_drain() isn't called.)  This is because
		 * ziltest causes spa_sync() to think it's committed,
		 * but actually it is not, so the intent log contains
		 * many txg's worth of changes.
		 *
		 * In particular, if object N is in the unlinked set in
		 * the last txg to actually sync, then it could be
		 * actually freed in a later txg and then reallocated
		 * in a yet later txg.  This would write a "create
		 * object N" record to the intent log.  Normally, this
		 * would be fine because the spa_sync() would have
		 * written out the fact that object N is free, before
		 * we could write the "create object N" intent log
		 * record.
		 *
		 * But when we are in ziltest mode, we advance the "open
		 * txg" without actually spa_sync()-ing the changes to
		 * disk.  So we would see that object N is still
		 * allocated and in the unlinked set, and there is an
		 * intent log record saying to allocate it.
		 */
		if (spa_writeable(dmu_objset_spa(zsb->z_os))) {
			if (zil_replay_disable) {
				zil_destroy(zsb->z_log, B_FALSE);
			} else {
				zsb->z_replay = B_TRUE;
				zil_replay(zsb->z_os, zsb,
				    zfs_replay_vector);
				zsb->z_replay = B_FALSE;
			}
		}

		/* restore readonly bit */
		if (readonly != 0)
			readonly_changed_cb(zsb, B_TRUE);
	}

	return (0);
}

int
zfs_register_callbacks(zfs_sb_t *zsb)
{
	struct dsl_dataset *ds = NULL;
	objset_t *os = zsb->z_os;
	zfs_mntopts_t *zmo = zsb->z_mntopts;
	int error = 0;

	ASSERT(zsb);
	ASSERT(zmo);

	/*
	 * The act of registering our callbacks will destroy any mount
	 * options we may have.  In order to enable temporary overrides
	 * of mount options, we stash away the current values and
	 * restore them after we register the callbacks.
	 */
	if (zfs_is_readonly(zsb) || !spa_writeable(dmu_objset_spa(os))) {
		zmo->z_do_readonly = B_TRUE;
		zmo->z_readonly = B_TRUE;
	}

	/*
	 * Register property callbacks.
	 *
	 * It would probably be fine to just check for i/o error from
	 * the first prop_register(), but I guess I like to go
	 * overboard...
	 */
	ds = dmu_objset_ds(os);
	dsl_pool_config_enter(dmu_objset_pool(os), FTAG);
	error = dsl_prop_register(ds,
	    zfs_prop_to_name(ZFS_PROP_ATIME), atime_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    zfs_prop_to_name(ZFS_PROP_RELATIME), relatime_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    zfs_prop_to_name(ZFS_PROP_XATTR), xattr_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    zfs_prop_to_name(ZFS_PROP_RECORDSIZE), blksz_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    zfs_prop_to_name(ZFS_PROP_READONLY), readonly_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    zfs_prop_to_name(ZFS_PROP_DEVICES), devices_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    zfs_prop_to_name(ZFS_PROP_SETUID), setuid_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    zfs_prop_to_name(ZFS_PROP_EXEC), exec_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    zfs_prop_to_name(ZFS_PROP_SNAPDIR), snapdir_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    zfs_prop_to_name(ZFS_PROP_ACLTYPE), acltype_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    zfs_prop_to_name(ZFS_PROP_ACLINHERIT), acl_inherit_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    zfs_prop_to_name(ZFS_PROP_VSCAN), vscan_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    zfs_prop_to_name(ZFS_PROP_NBMAND), nbmand_changed_cb, zsb);
	dsl_pool_config_exit(dmu_objset_pool(os), FTAG);
	if (error)
		goto unregister;

	/*
	 * Invoke our callbacks to restore temporary mount options.
	 */
	if (zmo->z_do_readonly)
		readonly_changed_cb(zsb, zmo->z_readonly);
	if (zmo->z_do_setuid)
		setuid_changed_cb(zsb, zmo->z_setuid);
	if (zmo->z_do_exec)
		exec_changed_cb(zsb, zmo->z_exec);
	if (zmo->z_do_devices)
		devices_changed_cb(zsb, zmo->z_devices);
	if (zmo->z_do_xattr)
		xattr_changed_cb(zsb, zmo->z_xattr);
	if (zmo->z_do_atime)
		atime_changed_cb(zsb, zmo->z_atime);
	if (zmo->z_do_relatime)
		relatime_changed_cb(zsb, zmo->z_relatime);
	if (zmo->z_do_nbmand)
		nbmand_changed_cb(zsb, zmo->z_nbmand);

	return (0);

unregister:
	/*
	 * We may attempt to unregister some callbacks that are not
	 * registered, but this is OK; it will simply return ENOMSG,
	 * which we will ignore.
	 */
	(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ATIME),
	    atime_changed_cb, zsb);
	(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_RELATIME),
	    relatime_changed_cb, zsb);
	(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_XATTR),
	    xattr_changed_cb, zsb);
	(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_RECORDSIZE),
	    blksz_changed_cb, zsb);
	(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_READONLY),
	    readonly_changed_cb, zsb);
	(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_DEVICES),
//.........这里部分代码省略.........

static int
__zvol_create_minor(const char *name, boolean_t ignore_snapdev)
{
	zvol_state_t *zv;
	objset_t *os;
	dmu_object_info_t *doi;
	uint64_t volsize;
	unsigned minor = 0;
	int error = 0;

	ASSERT(MUTEX_HELD(&zvol_state_lock));

	zv = zvol_find_by_name(name);
	if (zv) {
		error = EEXIST;
		goto out;
	}

	if (ignore_snapdev == B_FALSE) {
		error = __zvol_snapdev_hidden(name);
		if (error)
			goto out;
	}

	doi = kmem_alloc(sizeof(dmu_object_info_t), KM_SLEEP);

	error = dmu_objset_own(name, DMU_OST_ZVOL, B_TRUE, zvol_tag, &os);
	if (error)
		goto out_doi;

	error = dmu_object_info(os, ZVOL_OBJ, doi);
	if (error)
		goto out_dmu_objset_disown;

	error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize);
	if (error)
		goto out_dmu_objset_disown;

	error = zvol_find_minor(&minor);
	if (error)
		goto out_dmu_objset_disown;

	zv = zvol_alloc(MKDEV(zvol_major, minor), name);
	if (zv == NULL) {
		error = EAGAIN;
		goto out_dmu_objset_disown;
	}

	if (dmu_objset_is_snapshot(os))
		zv->zv_flags |= ZVOL_RDONLY;

	zv->zv_volblocksize = doi->doi_data_block_size;
	zv->zv_volsize = volsize;
	zv->zv_objset = os;

	set_capacity(zv->zv_disk, zv->zv_volsize >> 9);

	blk_queue_max_hw_sectors(zv->zv_queue, UINT_MAX);
	blk_queue_max_segments(zv->zv_queue, UINT16_MAX);
	blk_queue_max_segment_size(zv->zv_queue, UINT_MAX);
	blk_queue_physical_block_size(zv->zv_queue, zv->zv_volblocksize);
	blk_queue_io_opt(zv->zv_queue, zv->zv_volblocksize);
#ifdef HAVE_BLK_QUEUE_DISCARD
	blk_queue_max_discard_sectors(zv->zv_queue,
	    (zvol_max_discard_blocks * zv->zv_volblocksize) >> 9);
	blk_queue_discard_granularity(zv->zv_queue, zv->zv_volblocksize);
	queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, zv->zv_queue);
#endif
#ifdef HAVE_BLK_QUEUE_NONROT
	queue_flag_set_unlocked(QUEUE_FLAG_NONROT, zv->zv_queue);
#endif

	if (spa_writeable(dmu_objset_spa(os))) {
		if (zil_replay_disable)
			zil_destroy(dmu_objset_zil(os), B_FALSE);
		else
			zil_replay(os, zv, zvol_replay_vector);
	}

	zv->zv_objset = NULL;
out_dmu_objset_disown:
	dmu_objset_disown(os, zvol_tag);
out_doi:
	kmem_free(doi, sizeof(dmu_object_info_t));
out:

	if (error == 0) {
		zvol_insert(zv);
		add_disk(zv->zv_disk);
	}

	return (error);
}

//.........这里部分代码省略.........
#ifdef illumos
		if (vd->vdev_physpath != NULL &&
		    (dev = ddi_pathname_to_dev_t(vd->vdev_physpath)) != NODEV)
			error = ldi_open_by_dev(&dev, OTYP_BLK, spa_mode(spa),
			    kcred, &dvd->vd_lh, zfs_li);
#endif

		/*
		 * Note that we don't support the legacy auto-wholedisk support
		 * as above.  This hasn't been used in a very long time and we
		 * don't need to propagate its oddities to this edge condition.
		 */
		if (error && vd->vdev_path != NULL)
			error = ldi_open_by_name(vd->vdev_path, spa_mode(spa),
			    kcred, &dvd->vd_lh, zfs_li);
	}

	if (error) {
		vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
		vdev_dbgmsg(vd, "vdev_disk_open: failed to open [error=%d]",
		    error);
		return (error);
	}

/*
 * XXX Apple - We must not set or modify the devid. Import on Solaris/illumos
 * expects a valid devid and fails if it cannot be decoded.
 */
#ifdef illumos
	/*
	 * Now that the device has been successfully opened, update the devid
	 * if necessary.
	 */
	if (validate_devid && spa_writeable(spa) &&
	    ldi_get_devid(dvd->vd_lh, &devid) == 0) {
		if (ddi_devid_compare(devid, dvd->vd_devid) != 0) {
			char *vd_devid;

			vd_devid = ddi_devid_str_encode(devid, dvd->vd_minor);
			vdev_dbgmsg(vd, "vdev_disk_open: update devid from "
			    "'%s' to '%s'", vd->vdev_devid, vd_devid);
			spa_strfree(vd->vdev_devid);
			vd->vdev_devid = spa_strdup(vd_devid);
			ddi_devid_str_free(vd_devid);
		}
		ddi_devid_free(devid);
	}
#endif

/* XXX Apple to do, needs IORegistry physpath interface */
#ifdef illumos
	/*
	 * Once a device is opened, verify that the physical device path (if
	 * available) is up to date.
	 */
	if (ldi_get_dev(dvd->vd_lh, &dev) == 0 &&
	    ldi_get_otyp(dvd->vd_lh, &otyp) == 0) {
		char *physpath, *minorname;

		physpath = kmem_alloc(MAXPATHLEN, KM_SLEEP);
		minorname = NULL;
		if (ddi_dev_pathname(dev, otyp, physpath) == 0 &&
		    ldi_get_minor_name(dvd->vd_lh, &minorname) == 0 &&
		    (vd->vdev_physpath == NULL ||
		    strcmp(vd->vdev_physpath, physpath) != 0)) {
			if (vd->vdev_physpath)

/*
 * Synchronize pool configuration to disk.  This must be called with the
 * namespace lock held. Synchronizing the pool cache is typically done after
 * the configuration has been synced to the MOS. This exposes a window where
 * the MOS config will have been updated but the cache file has not. If
 * the system were to crash at that instant then the cached config may not
 * contain the correct information to open the pool and an explicity import
 * would be required.
 */
void
spa_config_sync(spa_t *target, boolean_t removing, boolean_t postsysevent)
{
	spa_config_dirent_t *dp, *tdp;
	nvlist_t *nvl;
	char *pool_name;

	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	if (rootdir == NULL || !(spa_mode_global & FWRITE))
		return;

	/*
	 * Iterate over all cachefiles for the pool, past or present.  When the
	 * cachefile is changed, the new one is pushed onto this list, allowing
	 * us to update previous cachefiles that no longer contain this pool.
	 */
	for (dp = list_head(&target->spa_config_list); dp != NULL;
	    dp = list_next(&target->spa_config_list, dp)) {
		spa_t *spa = NULL;
		if (dp->scd_path == NULL)
			continue;

		/*
		 * Iterate over all pools, adding any matching pools to 'nvl'.
		 */
		nvl = NULL;
		while ((spa = spa_next(spa)) != NULL) {
			/*
			 * Skip over our own pool if we're about to remove
			 * ourselves from the spa namespace or any pool that
			 * is readonly. Since we cannot guarantee that a
			 * readonly pool would successfully import upon reboot,
			 * we don't allow them to be written to the cache file.
			 */
			if ((spa == target && removing) ||
			    !spa_writeable(spa))
				continue;

			mutex_enter(&spa->spa_props_lock);
			tdp = list_head(&spa->spa_config_list);
			if (spa->spa_config == NULL ||
			    tdp->scd_path == NULL ||
			    strcmp(tdp->scd_path, dp->scd_path) != 0) {
				mutex_exit(&spa->spa_props_lock);
				continue;
			}

			if (nvl == NULL)
				VERIFY(nvlist_alloc(&nvl, NV_UNIQUE_NAME,
				    KM_SLEEP) == 0);

			if (spa->spa_import_flags & ZFS_IMPORT_TEMP_NAME) {
				VERIFY0(nvlist_lookup_string(spa->spa_config,
					ZPOOL_CONFIG_POOL_NAME, &pool_name));
			} else
				pool_name = spa_name(spa);

			VERIFY(nvlist_add_nvlist(nvl, pool_name,
			    spa->spa_config) == 0);
			mutex_exit(&spa->spa_props_lock);
		}

		spa_config_write(dp, nvl);
		nvlist_free(nvl);
	}

	/*
	 * Remove any config entries older than the current one.
	 */
	dp = list_head(&target->spa_config_list);
	while ((tdp = list_next(&target->spa_config_list, dp)) != NULL) {
		list_remove(&target->spa_config_list, tdp);
		if (tdp->scd_path != NULL)
			spa_strfree(tdp->scd_path);
		kmem_free(tdp, sizeof (spa_config_dirent_t));
	}

	spa_config_generation++;

	if (postsysevent)
		spa_event_notify(target, NULL, FM_EREPORT_ZFS_CONFIG_SYNC);
}

/*
 * Synchronize pool configuration to disk.  This must be called with the
 * namespace lock held. Synchronizing the pool cache is typically done after
 * the configuration has been synced to the MOS. This exposes a window where
 * the MOS config will have been updated but the cache file has not. If
 * the system were to crash at that instant then the cached config may not
 * contain the correct information to open the pool and an explicity import
 * would be required.
 */
void
spa_config_sync(spa_t *target, boolean_t removing, boolean_t postsysevent)
{
	spa_config_dirent_t *dp, *tdp;
	nvlist_t *nvl;
	boolean_t ccw_failure;
	int error;

	ASSERT(MUTEX_HELD(&spa_namespace_lock));

	if (rootdir == NULL || !(spa_mode_global & FWRITE))
		return;

	/*
	 * Iterate over all cachefiles for the pool, past or present.  When the
	 * cachefile is changed, the new one is pushed onto this list, allowing
	 * us to update previous cachefiles that no longer contain this pool.
	 */
	ccw_failure = B_FALSE;
	for (dp = list_head(&target->spa_config_list); dp != NULL;
	    dp = list_next(&target->spa_config_list, dp)) {
		spa_t *spa = NULL;
		if (dp->scd_path == NULL)
			continue;

		/*
		 * Iterate over all pools, adding any matching pools to 'nvl'.
		 */
		nvl = NULL;
		while ((spa = spa_next(spa)) != NULL) {
			nvlist_t *nvroot = NULL;
			/*
			 * Skip over our own pool if we're about to remove
			 * ourselves from the spa namespace or any pool that
			 * is readonly. Since we cannot guarantee that a
			 * readonly pool would successfully import upon reboot,
			 * we don't allow them to be written to the cache file.
			 */
			if ((spa == target && removing) ||
			    (spa_state(spa) == POOL_STATE_ACTIVE &&
			    !spa_writeable(spa)))
				continue;

			mutex_enter(&spa->spa_props_lock);
			tdp = list_head(&spa->spa_config_list);
			if (spa->spa_config == NULL ||
			    tdp->scd_path == NULL ||
			    strcmp(tdp->scd_path, dp->scd_path) != 0) {
				mutex_exit(&spa->spa_props_lock);
				continue;
			}

			if (nvl == NULL)
				VERIFY(nvlist_alloc(&nvl, NV_UNIQUE_NAME,
				    KM_SLEEP) == 0);

			VERIFY(nvlist_add_nvlist(nvl, spa->spa_name,
			    spa->spa_config) == 0);
			mutex_exit(&spa->spa_props_lock);

			if (nvlist_lookup_nvlist(nvl, spa->spa_name, &nvroot) == 0)
				spa_config_clean(nvroot);
		}

		error = spa_config_write(dp, nvl);
		if (error != 0)
			ccw_failure = B_TRUE;
		nvlist_free(nvl);
	}

	if (ccw_failure) {
		/*
		 * Keep trying so that configuration data is
		 * written if/when any temporary filesystem
		 * resource issues are resolved.
		 */
		if (target->spa_ccw_fail_time == 0) {
			zfs_ereport_post(FM_EREPORT_ZFS_CONFIG_CACHE_WRITE,
			    target, NULL, NULL, 0, 0);
		}
		target->spa_ccw_fail_time = gethrtime();
		spa_async_request(target, SPA_ASYNC_CONFIG_UPDATE);
	} else {
		/*
		 * Do not rate limit future attempts to update
		 * the config cache.
		 */
		target->spa_ccw_fail_time = 0;
	}

	/*
//.........这里部分代码省略.........

static int
zvol_first_open(zvol_state_t *zv)
{
	objset_t *os;
	uint64_t volsize;
	int locked = 0;
	int error;
	uint64_t ro;

	/*
	 * In all other cases the spa_namespace_lock is taken before the
	 * bdev->bd_mutex lock.  But in this case the Linux __blkdev_get()
	 * function calls fops->open() with the bdev->bd_mutex lock held.
	 *
	 * To avoid a potential lock inversion deadlock we preemptively
	 * try to take the spa_namespace_lock().  Normally it will not
	 * be contended and this is safe because spa_open_common() handles
	 * the case where the caller already holds the spa_namespace_lock.
	 *
	 * When it is contended we risk a lock inversion if we were to
	 * block waiting for the lock.  Luckily, the __blkdev_get()
	 * function allows us to return -ERESTARTSYS which will result in
	 * bdev->bd_mutex being dropped, reacquired, and fops->open() being
	 * called again.  This process can be repeated safely until both
	 * locks are acquired.
	 */
	if (!mutex_owned(&spa_namespace_lock)) {
		locked = mutex_tryenter(&spa_namespace_lock);
		if (!locked)
			return (-ERESTARTSYS);
	}

	/* lie and say we're read-only */
	error = dmu_objset_own(zv->zv_name, DMU_OST_ZVOL, 1, zvol_tag, &os);
	if (error)
		goto out_mutex;

	error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize);
	if (error) {
		dmu_objset_disown(os, zvol_tag);
		goto out_mutex;
	}

	zv->zv_objset = os;
	error = dmu_bonus_hold(os, ZVOL_OBJ, zvol_tag, &zv->zv_dbuf);
	if (error) {
		dmu_objset_disown(os, zvol_tag);
		goto out_mutex;
	}

	set_capacity(zv->zv_disk, volsize >> 9);
	zv->zv_volsize = volsize;
	zv->zv_zilog = zil_open(os, zvol_get_data);

	VERIFY(dsl_prop_get_integer(zv->zv_name, "readonly", &ro, NULL) == 0);
	if (ro || dmu_objset_is_snapshot(os) ||
	    !spa_writeable(dmu_objset_spa(os))) {
		set_disk_ro(zv->zv_disk, 1);
		zv->zv_flags |= ZVOL_RDONLY;
	} else {
		set_disk_ro(zv->zv_disk, 0);
		zv->zv_flags &= ~ZVOL_RDONLY;
	}

out_mutex:
	if (locked)
		mutex_exit(&spa_namespace_lock);

	return (-error);
}

/*
 * Create a block device minor node and setup the linkage between it
 * and the specified volume.  Once this function returns the block
 * device is live and ready for use.
 */
static int
zvol_create_minor_impl(const char *name)
{
	zvol_state_t *zv;
	objset_t *os;
	dmu_object_info_t *doi;
	uint64_t volsize;
	uint64_t len;
	unsigned minor = 0;
	int error = 0;

	mutex_enter(&zvol_state_lock);

	zv = zvol_find_by_name(name);
	if (zv) {
		error = SET_ERROR(EEXIST);
		goto out;
	}

	doi = kmem_alloc(sizeof (dmu_object_info_t), KM_SLEEP);

	error = dmu_objset_own(name, DMU_OST_ZVOL, B_TRUE, zvol_tag, &os);
	if (error)
		goto out_doi;

	error = dmu_object_info(os, ZVOL_OBJ, doi);
	if (error)
		goto out_dmu_objset_disown;

	error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize);
	if (error)
		goto out_dmu_objset_disown;

	error = zvol_find_minor(&minor);
	if (error)
		goto out_dmu_objset_disown;

	zv = zvol_alloc(MKDEV(zvol_major, minor), name);
	if (zv == NULL) {
		error = SET_ERROR(EAGAIN);
		goto out_dmu_objset_disown;
	}

	if (dmu_objset_is_snapshot(os))
		zv->zv_flags |= ZVOL_RDONLY;

	zv->zv_volblocksize = doi->doi_data_block_size;
	zv->zv_volsize = volsize;
	zv->zv_objset = os;

	set_capacity(zv->zv_disk, zv->zv_volsize >> 9);

	blk_queue_max_hw_sectors(zv->zv_queue, (DMU_MAX_ACCESS / 4) >> 9);
	blk_queue_max_segments(zv->zv_queue, UINT16_MAX);
	blk_queue_max_segment_size(zv->zv_queue, UINT_MAX);
	blk_queue_physical_block_size(zv->zv_queue, zv->zv_volblocksize);
	blk_queue_io_opt(zv->zv_queue, zv->zv_volblocksize);
	blk_queue_max_discard_sectors(zv->zv_queue,
	    (zvol_max_discard_blocks * zv->zv_volblocksize) >> 9);
	blk_queue_discard_granularity(zv->zv_queue, zv->zv_volblocksize);
	queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, zv->zv_queue);
#ifdef QUEUE_FLAG_NONROT
	queue_flag_set_unlocked(QUEUE_FLAG_NONROT, zv->zv_queue);
#endif
#ifdef QUEUE_FLAG_ADD_RANDOM
	queue_flag_clear_unlocked(QUEUE_FLAG_ADD_RANDOM, zv->zv_queue);
#endif

	if (spa_writeable(dmu_objset_spa(os))) {
		if (zil_replay_disable)
			zil_destroy(dmu_objset_zil(os), B_FALSE);
		else
			zil_replay(os, zv, zvol_replay_vector);
	}

	/*
	 * When udev detects the addition of the device it will immediately
	 * invoke blkid(8) to determine the type of content on the device.
	 * Prefetching the blocks commonly scanned by blkid(8) will speed
	 * up this process.
	 */
	len = MIN(MAX(zvol_prefetch_bytes, 0), SPA_MAXBLOCKSIZE);
	if (len > 0) {
		dmu_prefetch(os, ZVOL_OBJ, 0, 0, len, ZIO_PRIORITY_SYNC_READ);
		dmu_prefetch(os, ZVOL_OBJ, 0, volsize - len, len,
			ZIO_PRIORITY_SYNC_READ);
	}

	zv->zv_objset = NULL;
out_dmu_objset_disown:
	dmu_objset_disown(os, zvol_tag);
out_doi:
	kmem_free(doi, sizeof (dmu_object_info_t));
out:

//.........这里部分代码省略.........

int
zfs_register_callbacks(zfs_sb_t *zsb)
{
	struct dsl_dataset *ds = NULL;
	objset_t *os = zsb->z_os;
	boolean_t do_readonly = B_FALSE;
	int error = 0;

	if (zfs_is_readonly(zsb) || !spa_writeable(dmu_objset_spa(os)))
		do_readonly = B_TRUE;

	/*
	 * Register property callbacks.
	 *
	 * It would probably be fine to just check for i/o error from
	 * the first prop_register(), but I guess I like to go
	 * overboard...
	 */
	ds = dmu_objset_ds(os);
	dsl_pool_config_enter(dmu_objset_pool(os), FTAG);
	error = dsl_prop_register(ds,
	    zfs_prop_to_name(ZFS_PROP_ATIME), atime_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    zfs_prop_to_name(ZFS_PROP_XATTR), xattr_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    zfs_prop_to_name(ZFS_PROP_RECORDSIZE), blksz_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    zfs_prop_to_name(ZFS_PROP_READONLY), readonly_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    zfs_prop_to_name(ZFS_PROP_DEVICES), devices_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    zfs_prop_to_name(ZFS_PROP_SETUID), setuid_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    zfs_prop_to_name(ZFS_PROP_EXEC), exec_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    zfs_prop_to_name(ZFS_PROP_SNAPDIR), snapdir_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    zfs_prop_to_name(ZFS_PROP_ACLINHERIT), acl_inherit_changed_cb,
	    zsb);
	error = error ? error : dsl_prop_register(ds,
	    zfs_prop_to_name(ZFS_PROP_VSCAN), vscan_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    zfs_prop_to_name(ZFS_PROP_NBMAND), nbmand_changed_cb, zsb);
	dsl_pool_config_exit(dmu_objset_pool(os), FTAG);
	if (error)
		goto unregister;

	if (do_readonly)
		readonly_changed_cb(zsb, B_TRUE);

	return (0);

unregister:
	/*
	 * We may attempt to unregister some callbacks that are not
	 * registered, but this is OK; it will simply return ENOMSG,
	 * which we will ignore.
	 */
	(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ATIME),
	    atime_changed_cb, zsb);
	(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_XATTR),
	    xattr_changed_cb, zsb);
	(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_RECORDSIZE),
	    blksz_changed_cb, zsb);
	(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_READONLY),
	    readonly_changed_cb, zsb);
	(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_DEVICES),
	    devices_changed_cb, zsb);
	(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_SETUID),
	    setuid_changed_cb, zsb);
	(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_EXEC),
	    exec_changed_cb, zsb);
	(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_SNAPDIR),
	    snapdir_changed_cb, zsb);
	(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ACLINHERIT),
	    acl_inherit_changed_cb, zsb);
	(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_VSCAN),
	    vscan_changed_cb, zsb);
	(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_NBMAND),
	    nbmand_changed_cb, zsb);

	return (error);
}

/*
 * Read out the command history.
 */
int
spa_history_get(spa_t *spa, uint64_t *offp, uint64_t *len, char *buf)
{
	objset_t *mos = spa->spa_meta_objset;
	dmu_buf_t *dbp;
	uint64_t read_len, phys_read_off, phys_eof;
	uint64_t leftover = 0;
	spa_history_phys_t *shpp;
	int err;

	/*
	 * If the command history  doesn't exist (older pool),
	 * that's ok, just return ENOENT.
	 */
	if (!spa->spa_history)
		return (ENOENT);

	/*
	 * The history is logged asynchronously, so when they request
	 * the first chunk of history, make sure everything has been
	 * synced to disk so that we get it.
	 */
	if (*offp == 0 && spa_writeable(spa))
		txg_wait_synced(spa_get_dsl(spa), 0);

	if ((err = dmu_bonus_hold(mos, spa->spa_history, FTAG, &dbp)) != 0)
		return (err);
	shpp = dbp->db_data;

#ifdef ZFS_DEBUG
	{
		dmu_object_info_t doi;
		dmu_object_info_from_db(dbp, &doi);
		ASSERT3U(doi.doi_bonus_type, ==, DMU_OT_SPA_HISTORY_OFFSETS);
	}
#endif

	mutex_enter(&spa->spa_history_lock);
	phys_eof = spa_history_log_to_phys(shpp->sh_eof, shpp);

	if (*offp < shpp->sh_pool_create_len) {
		/* read in just the zpool create history */
		phys_read_off = *offp;
		read_len = MIN(*len, shpp->sh_pool_create_len -
		    phys_read_off);
	} else {
		/*
		 * Need to reset passed in offset to BOF if the passed in
		 * offset has since been overwritten.
		 */
		*offp = MAX(*offp, shpp->sh_bof);
		phys_read_off = spa_history_log_to_phys(*offp, shpp);

		/*
		 * Read up to the minimum of what the user passed down or
		 * the EOF (physical or logical).  If we hit physical EOF,
		 * use 'leftover' to read from the physical BOF.
		 */
		if (phys_read_off <= phys_eof) {
			read_len = MIN(*len, phys_eof - phys_read_off);
		} else {
			read_len = MIN(*len,
			    shpp->sh_phys_max_off - phys_read_off);
			if (phys_read_off + *len > shpp->sh_phys_max_off) {
				leftover = MIN(*len - read_len,
				    phys_eof - shpp->sh_pool_create_len);
			}
		}
	}

	/* offset for consumer to use next */
	*offp += read_len + leftover;

	/* tell the consumer how much you actually read */
	*len = read_len + leftover;

	if (read_len == 0) {
		mutex_exit(&spa->spa_history_lock);
		dmu_buf_rele(dbp, FTAG);
		return (0);
	}

	err = dmu_read(mos, spa->spa_history, phys_read_off, read_len, buf,
	    DMU_READ_PREFETCH);
	if (leftover && err == 0) {
		err = dmu_read(mos, spa->spa_history, shpp->sh_pool_create_len,
		    leftover, buf + read_len, DMU_READ_PREFETCH);
	}
	mutex_exit(&spa->spa_history_lock);

	dmu_buf_rele(dbp, FTAG);
	return (err);
}

int
zfs_register_callbacks(zfs_sb_t *zsb)
{
	struct dsl_dataset *ds = NULL;
	objset_t *os = zsb->z_os;
	int error = 0;

	if (zfs_is_readonly(zsb) || !spa_writeable(dmu_objset_spa(os)))
		readonly_changed_cb(zsb, B_TRUE);

	/*
	 * Register property callbacks.
	 *
	 * It would probably be fine to just check for i/o error from
	 * the first prop_register(), but I guess I like to go
	 * overboard...
	 */
	ds = dmu_objset_ds(os);
	error = dsl_prop_register(ds,
	    "atime", atime_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    "xattr", xattr_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    "recordsize", blksz_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    "readonly", readonly_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    "devices", devices_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    "setuid", setuid_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    "exec", exec_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    "snapdir", snapdir_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    "aclinherit", acl_inherit_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    "vscan", vscan_changed_cb, zsb);
	error = error ? error : dsl_prop_register(ds,
	    "nbmand", nbmand_changed_cb, zsb);
	if (error)
		goto unregister;

	return (0);

unregister:
	/*
	 * We may attempt to unregister some callbacks that are not
	 * registered, but this is OK; it will simply return ENOMSG,
	 * which we will ignore.
	 */
	(void) dsl_prop_unregister(ds, "atime", atime_changed_cb, zsb);
	(void) dsl_prop_unregister(ds, "xattr", xattr_changed_cb, zsb);
	(void) dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, zsb);
	(void) dsl_prop_unregister(ds, "readonly", readonly_changed_cb, zsb);
	(void) dsl_prop_unregister(ds, "devices", devices_changed_cb, zsb);
	(void) dsl_prop_unregister(ds, "setuid", setuid_changed_cb, zsb);
	(void) dsl_prop_unregister(ds, "exec", exec_changed_cb, zsb);
	(void) dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, zsb);
	(void) dsl_prop_unregister(ds, "aclinherit", acl_inherit_changed_cb,
	    zsb);
	(void) dsl_prop_unregister(ds, "vscan", vscan_changed_cb, zsb);
	(void) dsl_prop_unregister(ds, "nbmand", nbmand_changed_cb, zsb);

	return (error);
}