Beispiel #1
0
/* ARGSUSED */
static int
object_from_path(const char *dataset, const char *path, struct stat64 *statbuf,
    zinject_record_t *record)
{
	objset_t *os;
	int err;

	/*
	 * Before doing any libzpool operations, call sync() to ensure that the
	 * on-disk state is consistent with the in-core state.
	 */
	sync();

	err = dmu_objset_own(dataset, DMU_OST_ZFS, B_TRUE, FTAG, &os);
	if (err != 0) {
		(void) fprintf(stderr, "cannot open dataset '%s': %s\n",
		    dataset, strerror(err));
		return (-1);
	}

	record->zi_objset = dmu_objset_id(os);
	record->zi_object = statbuf->st_ino;

	dmu_objset_disown(os, FTAG);

	return (0);
}
Beispiel #2
0
Datei: zvol.c Projekt: alek-p/zfs
/*
 * Set ZFS_PROP_VOLSIZE set entry point.
 */
int
zvol_set_volsize(const char *name, uint64_t volsize)
{
	zvol_state_t *zv = NULL;
	objset_t *os = NULL;
	int error;
	dmu_object_info_t *doi;
	uint64_t readonly;
	boolean_t owned = B_FALSE;

	error = dsl_prop_get_integer(name,
	    zfs_prop_to_name(ZFS_PROP_READONLY), &readonly, NULL);
	if (error != 0)
		return (SET_ERROR(error));
	if (readonly)
		return (SET_ERROR(EROFS));

	mutex_enter(&zvol_state_lock);
	zv = zvol_find_by_name(name);

	if (zv == NULL || zv->zv_objset == NULL) {
		if ((error = dmu_objset_own(name, DMU_OST_ZVOL, B_FALSE,
		    FTAG, &os)) != 0) {
			mutex_exit(&zvol_state_lock);
			return (SET_ERROR(error));
		}
		owned = B_TRUE;
		if (zv != NULL)
			zv->zv_objset = os;
	} else {
		os = zv->zv_objset;
	}

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

	if ((error = dmu_object_info(os, ZVOL_OBJ, doi)) ||
	    (error = zvol_check_volsize(volsize, doi->doi_data_block_size)))
		goto out;

	error = zvol_update_volsize(volsize, os);
	kmem_free(doi, sizeof (dmu_object_info_t));

	if (error == 0 && zv != NULL)
		error = zvol_update_live_volsize(zv, volsize);
out:
	if (owned) {
		dmu_objset_disown(os, FTAG);
		if (zv != NULL)
			zv->zv_objset = NULL;
	}
	mutex_exit(&zvol_state_lock);
	return (error);
}
Beispiel #3
0
Datei: zvol.c Projekt: alek-p/zfs
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));
}
Beispiel #4
0
Datei: zvol.c Projekt: alek-p/zfs
/*
 * 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:

	if (error == 0) {
		zvol_insert(zv);
		/*
		 * Drop the lock to prevent deadlock with sys_open() ->
		 * zvol_open(), which first takes bd_disk->bd_mutex and then
		 * takes zvol_state_lock, whereas this code path first takes
		 * zvol_state_lock, and then takes bd_disk->bd_mutex.
		 */
		mutex_exit(&zvol_state_lock);
		add_disk(zv->zv_disk);
	} else {
		mutex_exit(&zvol_state_lock);
	}

	return (SET_ERROR(error));
}
Beispiel #5
0
static int
zpios_dmu_setup(run_args_t *run_args)
{
	zpios_time_t *t = &(run_args->stats.cr_time);
	objset_t *os;
	char name[32];
	uint64_t obj = 0ULL;
	int i, rc = 0, rc2;

	(void) zpios_upcall(run_args->pre, PHASE_PRE_CREATE, run_args, 0);
	t->start = zpios_timespec_now();

	(void) snprintf(name, 32, "%s/id_%d", run_args->pool, run_args->id);
	rc = dmu_objset_create(name, DMU_OST_OTHER, 0, NULL, NULL);
	if (rc) {
		zpios_print(run_args->file, "Error dmu_objset_create(%s, ...) "
			    "failed: %d\n", name, rc);
		goto out;
	}

	rc = dmu_objset_own(name, DMU_OST_OTHER, 0, zpios_tag, &os);
	if (rc) {
		zpios_print(run_args->file, "Error dmu_objset_own(%s, ...) "
			    "failed: %d\n", name, rc);
		goto out_destroy;
	}

	if (!(run_args->flags & DMU_FPP)) {
		obj = zpios_dmu_object_create(run_args, os);
		if (obj == 0) {
			rc = -EBADF;
			zpios_print(run_args->file, "Error zpios_dmu_"
				    "object_create() failed, %d\n", rc);
			goto out_destroy;
		}
	}

	for (i = 0; i < run_args->region_count; i++) {
		zpios_region_t *region;

		region = &run_args->regions[i];
		mutex_init(&region->lock, NULL, MUTEX_DEFAULT, NULL);

		if (run_args->flags & DMU_FPP) {
			/* File per process */
			region->obj.os  = os;
			region->obj.obj = zpios_dmu_object_create(run_args, os);
			ASSERT(region->obj.obj > 0); /* XXX - Handle this */
			region->wr_offset   = run_args->offset;
			region->rd_offset   = run_args->offset;
			region->init_offset = run_args->offset;
			region->max_offset  = run_args->offset +
			    run_args->region_size;
		} else {
			/* Single shared file */
			region->obj.os  = os;
			region->obj.obj = obj;
			region->wr_offset   = run_args->offset * i;
			region->rd_offset   = run_args->offset * i;
			region->init_offset = run_args->offset * i;
			region->max_offset  = run_args->offset *
			    i + run_args->region_size;
		}
	}

	run_args->os = os;
out_destroy:
	if (rc) {
		rc2 = dsl_destroy_head(name);
		if (rc2)
			zpios_print(run_args->file, "Error dsl_destroy_head"
				    "(%s, ...) failed: %d\n", name, rc2);
	}
out:
	t->stop  = zpios_timespec_now();
	t->delta = zpios_timespec_sub(t->stop, t->start);
	(void) zpios_upcall(run_args->post, PHASE_POST_CREATE, run_args, rc);

	return (rc);
}
Beispiel #6
0
static int
__zvol_create_minor(const char *name)
{
	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;
	}

	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;

    /* Make sure we have the key loaded if we need one. */
    error = dsl_crypto_key_inherit(name);
    if (error != 0 && error != EEXIST)
		goto out_dmu_objset_disown;

	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 (zil_replay_disable)
		zil_destroy(dmu_objset_zil(os), B_FALSE);
	else
		zil_replay(os, zv, zvol_replay_vector);

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

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

	return (error);
}
Beispiel #7
0
int
zfs_sb_create(const char *osname, zfs_mntopts_t *zmo, zfs_sb_t **zsbp)
{
	objset_t *os;
	zfs_sb_t *zsb;
	uint64_t zval;
	int i, error;
	uint64_t sa_obj;

	zsb = kmem_zalloc(sizeof (zfs_sb_t), KM_SLEEP);

	/*
	 * We claim to always be readonly so we can open snapshots;
	 * other ZPL code will prevent us from writing to snapshots.
	 */
	error = dmu_objset_own(osname, DMU_OST_ZFS, B_TRUE, zsb, &os);
	if (error) {
		kmem_free(zsb, sizeof (zfs_sb_t));
		return (error);
	}

	/*
	 * Optional temporary mount options, free'd in zfs_sb_free().
	 */
	zsb->z_mntopts = (zmo ? zmo : zfs_mntopts_alloc());

	/*
	 * Initialize the zfs-specific filesystem structure.
	 * Should probably make this a kmem cache, shuffle fields,
	 * and just bzero up to z_hold_mtx[].
	 */
	zsb->z_sb = NULL;
	zsb->z_parent = zsb;
	zsb->z_max_blksz = SPA_OLD_MAXBLOCKSIZE;
	zsb->z_show_ctldir = ZFS_SNAPDIR_VISIBLE;
	zsb->z_os = os;

	error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zsb->z_version);
	if (error) {
		goto out;
	} else if (zsb->z_version > ZPL_VERSION) {
		error = SET_ERROR(ENOTSUP);
		goto out;
	}
	if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0)
		goto out;
	zsb->z_norm = (int)zval;

	if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0)
		goto out;
	zsb->z_utf8 = (zval != 0);

	if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0)
		goto out;
	zsb->z_case = (uint_t)zval;

	if ((error = zfs_get_zplprop(os, ZFS_PROP_ACLTYPE, &zval)) != 0)
		goto out;
	zsb->z_acl_type = (uint_t)zval;

	/*
	 * Fold case on file systems that are always or sometimes case
	 * insensitive.
	 */
	if (zsb->z_case == ZFS_CASE_INSENSITIVE ||
	    zsb->z_case == ZFS_CASE_MIXED)
		zsb->z_norm |= U8_TEXTPREP_TOUPPER;

	zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os);
	zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os);

	if (zsb->z_use_sa) {
		/* should either have both of these objects or none */
		error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1,
		    &sa_obj);
		if (error)
			goto out;

		error = zfs_get_zplprop(os, ZFS_PROP_XATTR, &zval);
		if ((error == 0) && (zval == ZFS_XATTR_SA))
			zsb->z_xattr_sa = B_TRUE;
	} else {
		/*
		 * Pre SA versions file systems should never touch
		 * either the attribute registration or layout objects.
		 */
		sa_obj = 0;
	}

	error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
	    &zsb->z_attr_table);
	if (error)
		goto out;

	if (zsb->z_version >= ZPL_VERSION_SA)
		sa_register_update_callback(os, zfs_sa_upgrade);

	error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1,
	    &zsb->z_root);
	if (error)
		goto out;
	ASSERT(zsb->z_root != 0);

	error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1,
	    &zsb->z_unlinkedobj);
	if (error)
		goto out;

	error = zap_lookup(os, MASTER_NODE_OBJ,
	    zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA],
	    8, 1, &zsb->z_userquota_obj);
	if (error && error != ENOENT)
		goto out;

	error = zap_lookup(os, MASTER_NODE_OBJ,
	    zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA],
	    8, 1, &zsb->z_groupquota_obj);
	if (error && error != ENOENT)
		goto out;

	error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1,
	    &zsb->z_fuid_obj);
	if (error && error != ENOENT)
		goto out;

	error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1,
	    &zsb->z_shares_dir);
	if (error && error != ENOENT)
		goto out;

	mutex_init(&zsb->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
	mutex_init(&zsb->z_lock, NULL, MUTEX_DEFAULT, NULL);
	list_create(&zsb->z_all_znodes, sizeof (znode_t),
	    offsetof(znode_t, z_link_node));
	rrm_init(&zsb->z_teardown_lock, B_FALSE);
	rw_init(&zsb->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL);
	rw_init(&zsb->z_fuid_lock, NULL, RW_DEFAULT, NULL);

	zsb->z_hold_mtx = vmem_zalloc(sizeof (kmutex_t) * ZFS_OBJ_MTX_SZ,
	    KM_SLEEP);
	for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
		mutex_init(&zsb->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);

	*zsbp = zsb;
	return (0);

out:
	dmu_objset_disown(os, zsb);
	*zsbp = NULL;

	vmem_free(zsb->z_hold_mtx, sizeof (kmutex_t) * ZFS_OBJ_MTX_SZ);
	kmem_free(zsb, sizeof (zfs_sb_t));
	return (error);
}
Beispiel #8
0
/*
 * Calculate the real range based on the type, level, and range given.
 */
static int
calculate_range(const char *dataset, err_type_t type, int level, char *range,
    zinject_record_t *record)
{
	objset_t *os = NULL;
	dnode_t *dn = NULL;
	int err;
	int ret = -1;

	/*
	 * Determine the numeric range from the string.
	 */
	if (range == NULL) {
		/*
		 * If range is unspecified, set the range to [0,-1], which
		 * indicates that the whole object should be treated as an
		 * error.
		 */
		record->zi_start = 0;
		record->zi_end = -1ULL;
	} else {
		char *end;

		/* XXX add support for suffixes */
		record->zi_start = strtoull(range, &end, 10);


		if (*end == '\0')
			record->zi_end = record->zi_start + 1;
		else if (*end == ',')
			record->zi_end = strtoull(end + 1, &end, 10);

		if (*end != '\0') {
			(void) fprintf(stderr, "invalid range '%s': must be "
			    "a numeric range of the form 'start[,end]'\n",
			    range);
			goto out;
		}
	}

	switch (type) {
	case TYPE_DATA:
		break;

	case TYPE_DNODE:
		/*
		 * If this is a request to inject faults into the dnode, then we
		 * must translate the current (objset,object) pair into an
		 * offset within the metadnode for the objset.  Specifying any
		 * kind of range with type 'dnode' is illegal.
		 */
		if (range != NULL) {
			(void) fprintf(stderr, "range cannot be specified when "
			    "type is 'dnode'\n");
			goto out;
		}

		record->zi_start = record->zi_object * sizeof (dnode_phys_t);
		record->zi_end = record->zi_start + sizeof (dnode_phys_t);
		record->zi_object = 0;
		break;
	}

	/*
	 * Get the dnode associated with object, so we can calculate the block
	 * size.
	 */
	if ((err = dmu_objset_own(dataset, DMU_OST_ANY,
	    B_TRUE, FTAG, &os)) != 0) {
		(void) fprintf(stderr, "cannot open dataset '%s': %s\n",
		    dataset, strerror(err));
		goto out;
	}

	if (record->zi_object == 0) {
		dn = DMU_META_DNODE(os);
	} else {
		err = dnode_hold(os, record->zi_object, FTAG, &dn);
		if (err != 0) {
			(void) fprintf(stderr, "failed to hold dnode "
			    "for object %llu\n",
			    (u_longlong_t)record->zi_object);
			goto out;
		}
	}


	ziprintf("data shift: %d\n", (int)dn->dn_datablkshift);
	ziprintf(" ind shift: %d\n", (int)dn->dn_indblkshift);

	/*
	 * Translate range into block IDs.
	 */
	if (record->zi_start != 0 || record->zi_end != -1ULL) {
		record->zi_start >>= dn->dn_datablkshift;
		record->zi_end >>= dn->dn_datablkshift;
	}
Beispiel #9
0
/*
 * Reopen zfs_sb_t and release VFS ops.
 */
int
zfs_resume_fs(zfs_sb_t *zsb, const char *osname)
{
	int err, err2;

	ASSERT(RRW_WRITE_HELD(&zsb->z_teardown_lock));
	ASSERT(RW_WRITE_HELD(&zsb->z_teardown_inactive_lock));

	err = dmu_objset_own(osname, DMU_OST_ZFS, B_FALSE, zsb, &zsb->z_os);
	if (err) {
		zsb->z_os = NULL;
	} else {
		znode_t *zp;
		uint64_t sa_obj = 0;

		err2 = zap_lookup(zsb->z_os, MASTER_NODE_OBJ,
		    ZFS_SA_ATTRS, 8, 1, &sa_obj);

		if ((err || err2) && zsb->z_version >= ZPL_VERSION_SA)
			goto bail;


		if ((err = sa_setup(zsb->z_os, sa_obj,
		    zfs_attr_table,  ZPL_END, &zsb->z_attr_table)) != 0)
			goto bail;

		VERIFY(zfs_sb_setup(zsb, B_FALSE) == 0);
		zsb->z_rollback_time = jiffies;

		/*
		 * Attempt to re-establish all the active inodes with their
		 * dbufs.  If a zfs_rezget() fails, then we unhash the inode
		 * and mark it stale.  This prevents a collision if a new
		 * inode/object is created which must use the same inode
		 * number.  The stale inode will be be released when the
		 * VFS prunes the dentry holding the remaining references
		 * on the stale inode.
		 */
		mutex_enter(&zsb->z_znodes_lock);
		for (zp = list_head(&zsb->z_all_znodes); zp;
		    zp = list_next(&zsb->z_all_znodes, zp)) {
			err2 = zfs_rezget(zp);
			if (err2) {
				remove_inode_hash(ZTOI(zp));
				zp->z_is_stale = B_TRUE;
			}
		}
		mutex_exit(&zsb->z_znodes_lock);
	}

bail:
	/* release the VFS ops */
	rw_exit(&zsb->z_teardown_inactive_lock);
	rrw_exit(&zsb->z_teardown_lock, FTAG);

	if (err) {
		/*
		 * Since we couldn't reopen zfs_sb_t or, setup the
		 * sa framework, force unmount this file system.
		 */
		if (zsb->z_os)
			(void) zfs_umount(zsb->z_sb);
	}
	return (err);
}
Beispiel #10
0
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);
}
Beispiel #11
0
static int
__zvol_create_minor(const char *name)
{
	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;
	}

	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);

	if (zil_replay_disable)
		zil_destroy(dmu_objset_zil(os), B_FALSE);
	else
		zil_replay(os, zv, zvol_replay_vector);

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

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

	return (error);
}
Beispiel #12
0
/*
 * Given a full path to a file, translate into a dataset name and a relative
 * path within the dataset.  'dataset' must be at least MAXNAMELEN characters,
 * and 'relpath' must be at least MAXPATHLEN characters.  We also pass a stat
 * buffer, which we need later to get the object ID.
 */
static int
parse_pathname(const char *inpath, char *dataset, char *relpath,
    struct stat *statbuf)
{
	struct extmnttab mp;
	FILE *fp;
	int match;
	const char *rel;
	char fullpath[MAXPATHLEN];

	compress_slashes(inpath, fullpath);

	if (fullpath[0] != '/') {
		(void) fprintf(stderr, "invalid object '%s': must be full "
		    "path\n", fullpath);
		usage();
		return (-1);
	}

	if (strlen(fullpath) >= MAXPATHLEN) {
		(void) fprintf(stderr, "invalid object; pathname too long\n");
		return (-1);
	}

	if (stat(fullpath, statbuf) != 0) {
		(void) fprintf(stderr, "cannot open '%s': %s\n",
		    fullpath, strerror(errno));
		return (-1);
	}

#ifdef HAVE_SETMNTENT
	if ((fp = setmntent(MNTTAB, "r")) == NULL) {
#else
	if ((fp = fopen(MNTTAB, "r")) == NULL) {
#endif
		(void) fprintf(stderr, "cannot open /etc/mtab\n");
		return (-1);
	}

	match = 0;
	while (getextmntent(fp, &mp, sizeof (mp)) == 0) {
		if (makedev(mp.mnt_major, mp.mnt_minor) == statbuf->st_dev) {
			match = 1;
			break;
		}
	}

	if (!match) {
		(void) fprintf(stderr, "cannot find mountpoint for '%s'\n",
		    fullpath);
		return (-1);
	}

	if (strcmp(mp.mnt_fstype, MNTTYPE_ZFS) != 0) {
		(void) fprintf(stderr, "invalid path '%s': not a ZFS "
		    "filesystem\n", fullpath);
		return (-1);
	}

	if (strncmp(fullpath, mp.mnt_mountp, strlen(mp.mnt_mountp)) != 0) {
		(void) fprintf(stderr, "invalid path '%s': mountpoint "
		    "doesn't match path\n", fullpath);
		return (-1);
	}

	(void) strcpy(dataset, mp.mnt_special);

	rel = fullpath + strlen(mp.mnt_mountp);
	if (rel[0] == '/')
		rel++;
	(void) strcpy(relpath, rel);

	return (0);
}
#endif

//From FreeBSD
static int
parse_pathname(const char *inpath, char *dataset, char *relpath,
    struct stat *statbuf)
{
        struct statfs sfs;
        const char *rel;
        char fullpath[MAXPATHLEN];

        compress_slashes(inpath, fullpath);

        if (fullpath[0] != '/') {
                (void) fprintf(stderr, "invalid object '%s': must be full "
                    "path\n", fullpath);
                usage();
                return (-1);
        }

        if (strlen(fullpath) >= MAXPATHLEN) {
                (void) fprintf(stderr, "invalid object; pathname too long\n");
                return (-1);
        }

        if (stat(fullpath, statbuf) != 0) {
                (void) fprintf(stderr, "cannot open '%s': %s\n",
                    fullpath, strerror(errno));
                return (-1);
        }

        if (statfs(fullpath, &sfs) == -1) {
                (void) fprintf(stderr, "cannot find mountpoint for '%s': %s\n",
                    fullpath, strerror(errno));
                return (-1);
        }

        if (strcmp(sfs.f_fstypename, MNTTYPE_ZFS) != 0) {
                (void) fprintf(stderr, "invalid path '%s': not a ZFS "
                    "filesystem\n", fullpath);
                return (-1);
        }

        if (strncmp(fullpath, sfs.f_mntonname, strlen(sfs.f_mntonname)) != 0) {
                (void) fprintf(stderr, "invalid path '%s': mountpoint "
                    "doesn't match path\n", fullpath);
                return (-1);
        }

        (void) strcpy(dataset, sfs.f_mntfromname);

        rel = fullpath + strlen(sfs.f_mntonname);
        if (rel[0] == '/')
                rel++;
        (void) strcpy(relpath, rel);

        return (0);
}

/*
 * Convert from a (dataset, path) pair into a (objset, object) pair.  Note that
 * we grab the object number from the inode number, since looking this up via
 * libzpool is a real pain.
 */
/* ARGSUSED */
static int
object_from_path(const char *dataset, const char *path, struct stat *statbuf,
    zinject_record_t *record)
{
	objset_t *os;
	int err;

	/*
	 * Before doing any libzpool operations, call sync() to ensure that the
	 * on-disk state is consistent with the in-core state.
	 */
	sync();

	err = dmu_objset_own(dataset, DMU_OST_ZFS, B_TRUE, B_FALSE, FTAG, &os);
	if (err != 0) {
		(void) fprintf(stderr, "cannot open dataset '%s': %s\n",
		    dataset, strerror(err));
		return (-1);
	}

	record->zi_objset = dmu_objset_id(os);
	record->zi_object = statbuf->st_ino;

	dmu_objset_disown(os, B_FALSE, FTAG);

	return (0);
}