2 * Copyright (C) 2005-2011 Red Hat, Inc. All rights reserved.
4 * This file is part of the device-mapper userspace tools.
6 * This copyrighted material is made available to anyone wishing to use,
7 * modify, copy, or redistribute it subject to the terms and conditions
8 * of the GNU Lesser General Public License v.2.1.
10 * You should have received a copy of the GNU Lesser General Public License
11 * along with this program; if not, write to the Free Software Foundation,
12 * Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
16 #include "libdm-targets.h"
17 #include "libdm-common.h"
22 #include <sys/param.h>
23 #include <sys/utsname.h>
25 #define MAX_TARGET_PARAMSIZE 500000
27 #define REPLICATOR_LOCAL_SITE 0
29 /* Supported segment types */
56 /* FIXME Add crypt and multipath support */
62 { SEG_CRYPT
, "crypt" },
63 { SEG_ERROR
, "error" },
64 { SEG_LINEAR
, "linear" },
65 { SEG_MIRRORED
, "mirror" },
66 { SEG_REPLICATOR
, "replicator" },
67 { SEG_REPLICATOR_DEV
, "replicator-dev" },
68 { SEG_SNAPSHOT
, "snapshot" },
69 { SEG_SNAPSHOT_ORIGIN
, "snapshot-origin" },
70 { SEG_SNAPSHOT_MERGE
, "snapshot-merge" },
71 { SEG_STRIPED
, "striped" },
73 { SEG_THIN_POOL
, "thin-pool"},
75 { SEG_RAID1
, "raid1"},
76 { SEG_RAID4
, "raid4"},
77 { SEG_RAID5_LA
, "raid5_la"},
78 { SEG_RAID5_RA
, "raid5_ra"},
79 { SEG_RAID5_LS
, "raid5_ls"},
80 { SEG_RAID5_RS
, "raid5_rs"},
81 { SEG_RAID6_ZR
, "raid6_zr"},
82 { SEG_RAID6_NR
, "raid6_nr"},
83 { SEG_RAID6_NC
, "raid6_nc"},
86 *WARNING: Since 'raid' target overloads this 1:1 mapping table
87 * for search do not add new enum elements past them!
89 { SEG_RAID5_LS
, "raid5"}, /* same as "raid5_ls" (default for MD also) */
90 { SEG_RAID6_ZR
, "raid6"}, /* same as "raid6_zr" */
94 /* Some segment types have a list of areas of other devices attached */
98 struct dm_tree_node
*dev_node
;
102 unsigned rsite_index
; /* Replicator site index */
103 struct dm_tree_node
*slog
; /* Replicator sync log node */
104 uint64_t region_size
; /* Replicator sync log size */
105 uint32_t flags
; /* Replicator sync log flags */
108 struct dm_thin_message
{
109 dm_thin_message_t type
;
124 } m_set_transaction_id
;
132 struct thin_message
{
134 struct dm_thin_message message
;
138 /* Replicator-log has a list of sites */
139 /* FIXME: maybe move to seg_area too? */
140 struct replicator_site
{
143 unsigned rsite_index
;
144 dm_replicator_mode_t mode
;
145 uint32_t async_timeout
;
146 uint32_t fall_behind_ios
;
147 uint64_t fall_behind_data
;
150 /* Per-segment properties */
151 struct load_segment
{
158 unsigned area_count
; /* Linear + Striped + Mirrored + Crypt + Replicator */
159 struct dm_list areas
; /* Linear + Striped + Mirrored + Crypt + Replicator */
161 uint32_t stripe_size
; /* Striped + raid */
163 int persistent
; /* Snapshot */
164 uint32_t chunk_size
; /* Snapshot */
165 struct dm_tree_node
*cow
; /* Snapshot */
166 struct dm_tree_node
*origin
; /* Snapshot + Snapshot origin */
167 struct dm_tree_node
*merge
; /* Snapshot */
169 struct dm_tree_node
*log
; /* Mirror + Replicator */
170 uint32_t region_size
; /* Mirror + raid */
171 unsigned clustered
; /* Mirror */
172 unsigned mirror_area_count
; /* Mirror */
173 uint32_t flags
; /* Mirror log */
174 char *uuid
; /* Clustered mirror log */
176 const char *cipher
; /* Crypt */
177 const char *chainmode
; /* Crypt */
178 const char *iv
; /* Crypt */
179 uint64_t iv_offset
; /* Crypt */
180 const char *key
; /* Crypt */
182 const char *rlog_type
; /* Replicator */
183 struct dm_list rsites
; /* Replicator */
184 unsigned rsite_count
; /* Replicator */
185 unsigned rdevice_count
; /* Replicator */
186 struct dm_tree_node
*replicator
;/* Replicator-dev */
187 uint64_t rdevice_index
; /* Replicator-dev */
189 uint64_t rebuilds
; /* raid */
191 struct dm_tree_node
*metadata
; /* Thin_pool */
192 struct dm_tree_node
*pool
; /* Thin_pool, Thin */
193 struct dm_list thin_messages
; /* Thin_pool */
194 uint64_t transaction_id
; /* Thin_pool */
195 uint64_t low_water_mark
; /* Thin_pool */
196 uint32_t data_block_size
; /* Thin_pool */
197 unsigned skip_block_zeroing
; /* Thin_pool */
198 uint32_t device_id
; /* Thin */
202 /* Per-device properties */
203 struct load_properties
{
209 uint32_t read_ahead_flags
;
211 unsigned segment_count
;
212 unsigned size_changed
;
215 const char *new_name
;
217 /* If immediate_dev_node is set to 1, try to create the dev node
218 * as soon as possible (e.g. in preload stage even during traversal
219 * and processing of dm tree). This will also flush all stacked dev
220 * node operations, synchronizing with udev.
222 unsigned immediate_dev_node
;
225 * If the device size changed from zero and this is set,
226 * don't resume the device immediately, even if the device
227 * has parents. This works provided the parents do not
228 * validate the device size and is required by pvmove to
229 * avoid starting the mirror resync operation too early.
231 unsigned delay_resume_if_new
;
233 /* Send messages for this node in preload */
234 unsigned send_messages
;
237 /* Two of these used to join two nodes with uses and used_by. */
238 struct dm_tree_link
{
240 struct dm_tree_node
*node
;
243 struct dm_tree_node
{
244 struct dm_tree
*dtree
;
250 struct dm_list uses
; /* Nodes this node uses */
251 struct dm_list used_by
; /* Nodes that use this node */
253 int activation_priority
; /* 0 gets activated first */
255 uint16_t udev_flags
; /* Udev control flags */
257 void *context
; /* External supplied context */
259 struct load_properties props
; /* For creation/table (re)load */
262 * If presuspend of child node is needed
263 * Note: only direct child is allowed
265 struct dm_tree_node
*presuspend_node
;
270 struct dm_hash_table
*devs
;
271 struct dm_hash_table
*uuids
;
272 struct dm_tree_node root
;
273 int skip_lockfs
; /* 1 skips lockfs (for non-snapshots) */
274 int no_flush
; /* 1 sets noflush (mirrors/multipath) */
275 int retry_remove
; /* 1 retries remove if not successful */
282 struct dm_tree
*dm_tree_create(void)
284 struct dm_pool
*dmem
;
285 struct dm_tree
*dtree
;
287 if (!(dmem
= dm_pool_create("dtree", 1024)) ||
288 !(dtree
= dm_pool_zalloc(dmem
, sizeof(*dtree
)))) {
289 log_error("Failed to allocate dtree.");
291 dm_pool_destroy(dmem
);
295 dtree
->root
.dtree
= dtree
;
296 dm_list_init(&dtree
->root
.uses
);
297 dm_list_init(&dtree
->root
.used_by
);
298 dtree
->skip_lockfs
= 0;
302 if (!(dtree
->devs
= dm_hash_create(8))) {
303 log_error("dtree hash creation failed");
304 dm_pool_destroy(dtree
->mem
);
308 if (!(dtree
->uuids
= dm_hash_create(32))) {
309 log_error("dtree uuid hash creation failed");
310 dm_hash_destroy(dtree
->devs
);
311 dm_pool_destroy(dtree
->mem
);
318 void dm_tree_free(struct dm_tree
*dtree
)
323 dm_hash_destroy(dtree
->uuids
);
324 dm_hash_destroy(dtree
->devs
);
325 dm_pool_destroy(dtree
->mem
);
328 void dm_tree_set_cookie(struct dm_tree_node
*node
, uint32_t cookie
)
330 node
->dtree
->cookie
= cookie
;
333 uint32_t dm_tree_get_cookie(struct dm_tree_node
*node
)
335 return node
->dtree
->cookie
;
338 void dm_tree_skip_lockfs(struct dm_tree_node
*dnode
)
340 dnode
->dtree
->skip_lockfs
= 1;
343 void dm_tree_use_no_flush_suspend(struct dm_tree_node
*dnode
)
345 dnode
->dtree
->no_flush
= 1;
348 void dm_tree_retry_remove(struct dm_tree_node
*dnode
)
350 dnode
->dtree
->retry_remove
= 1;
356 static int _nodes_are_linked(const struct dm_tree_node
*parent
,
357 const struct dm_tree_node
*child
)
359 struct dm_tree_link
*dlink
;
361 dm_list_iterate_items(dlink
, &parent
->uses
)
362 if (dlink
->node
== child
)
368 static int _link(struct dm_list
*list
, struct dm_tree_node
*node
)
370 struct dm_tree_link
*dlink
;
372 if (!(dlink
= dm_pool_alloc(node
->dtree
->mem
, sizeof(*dlink
)))) {
373 log_error("dtree link allocation failed");
378 dm_list_add(list
, &dlink
->list
);
383 static int _link_nodes(struct dm_tree_node
*parent
,
384 struct dm_tree_node
*child
)
386 if (_nodes_are_linked(parent
, child
))
389 if (!_link(&parent
->uses
, child
))
392 if (!_link(&child
->used_by
, parent
))
398 static void _unlink(struct dm_list
*list
, struct dm_tree_node
*node
)
400 struct dm_tree_link
*dlink
;
402 dm_list_iterate_items(dlink
, list
)
403 if (dlink
->node
== node
) {
404 dm_list_del(&dlink
->list
);
409 static void _unlink_nodes(struct dm_tree_node
*parent
,
410 struct dm_tree_node
*child
)
412 if (!_nodes_are_linked(parent
, child
))
415 _unlink(&parent
->uses
, child
);
416 _unlink(&child
->used_by
, parent
);
419 static int _add_to_toplevel(struct dm_tree_node
*node
)
421 return _link_nodes(&node
->dtree
->root
, node
);
424 static void _remove_from_toplevel(struct dm_tree_node
*node
)
426 _unlink_nodes(&node
->dtree
->root
, node
);
429 static int _add_to_bottomlevel(struct dm_tree_node
*node
)
431 return _link_nodes(node
, &node
->dtree
->root
);
434 static void _remove_from_bottomlevel(struct dm_tree_node
*node
)
436 _unlink_nodes(node
, &node
->dtree
->root
);
439 static int _link_tree_nodes(struct dm_tree_node
*parent
, struct dm_tree_node
*child
)
441 /* Don't link to root node if child already has a parent */
442 if (parent
== &parent
->dtree
->root
) {
443 if (dm_tree_node_num_children(child
, 1))
446 _remove_from_toplevel(child
);
448 if (child
== &child
->dtree
->root
) {
449 if (dm_tree_node_num_children(parent
, 0))
452 _remove_from_bottomlevel(parent
);
454 return _link_nodes(parent
, child
);
457 static struct dm_tree_node
*_create_dm_tree_node(struct dm_tree
*dtree
,
460 struct dm_info
*info
,
464 struct dm_tree_node
*node
;
467 if (!(node
= dm_pool_zalloc(dtree
->mem
, sizeof(*node
)))) {
468 log_error("_create_dm_tree_node alloc failed");
477 node
->context
= context
;
478 node
->udev_flags
= udev_flags
;
479 node
->activation_priority
= 0;
481 dm_list_init(&node
->uses
);
482 dm_list_init(&node
->used_by
);
483 dm_list_init(&node
->props
.segs
);
485 dev
= MKDEV(info
->major
, info
->minor
);
487 if (!dm_hash_insert_binary(dtree
->devs
, (const char *) &dev
,
488 sizeof(dev
), node
)) {
489 log_error("dtree node hash insertion failed");
490 dm_pool_free(dtree
->mem
, node
);
495 !dm_hash_insert(dtree
->uuids
, uuid
, node
)) {
496 log_error("dtree uuid hash insertion failed");
497 dm_hash_remove_binary(dtree
->devs
, (const char *) &dev
,
499 dm_pool_free(dtree
->mem
, node
);
506 static struct dm_tree_node
*_find_dm_tree_node(struct dm_tree
*dtree
,
507 uint32_t major
, uint32_t minor
)
509 uint64_t dev
= MKDEV(major
, minor
);
511 return dm_hash_lookup_binary(dtree
->devs
, (const char *) &dev
,
515 static struct dm_tree_node
*_find_dm_tree_node_by_uuid(struct dm_tree
*dtree
,
518 struct dm_tree_node
*node
;
519 const char *default_uuid_prefix
;
520 size_t default_uuid_prefix_len
;
522 if ((node
= dm_hash_lookup(dtree
->uuids
, uuid
)))
525 default_uuid_prefix
= dm_uuid_prefix();
526 default_uuid_prefix_len
= strlen(default_uuid_prefix
);
528 if (strncmp(uuid
, default_uuid_prefix
, default_uuid_prefix_len
))
531 return dm_hash_lookup(dtree
->uuids
, uuid
+ default_uuid_prefix_len
);
534 void dm_tree_node_set_udev_flags(struct dm_tree_node
*dnode
, uint16_t udev_flags
)
537 struct dm_info
*dinfo
= &dnode
->info
;
539 if (udev_flags
!= dnode
->udev_flags
)
540 log_debug("Resetting %s (%" PRIu32
":%" PRIu32
541 ") udev_flags from 0x%x to 0x%x",
542 dnode
->name
, dinfo
->major
, dinfo
->minor
,
543 dnode
->udev_flags
, udev_flags
);
544 dnode
->udev_flags
= udev_flags
;
547 void dm_tree_node_set_read_ahead(struct dm_tree_node
*dnode
,
549 uint32_t read_ahead_flags
)
551 dnode
->props
.read_ahead
= read_ahead
;
552 dnode
->props
.read_ahead_flags
= read_ahead_flags
;
555 void dm_tree_node_set_presuspend_node(struct dm_tree_node
*node
,
556 struct dm_tree_node
*presuspend_node
)
558 node
->presuspend_node
= presuspend_node
;
561 const char *dm_tree_node_get_name(const struct dm_tree_node
*node
)
563 return node
->info
.exists
? node
->name
: "";
566 const char *dm_tree_node_get_uuid(const struct dm_tree_node
*node
)
568 return node
->info
.exists
? node
->uuid
: "";
571 const struct dm_info
*dm_tree_node_get_info(const struct dm_tree_node
*node
)
576 void *dm_tree_node_get_context(const struct dm_tree_node
*node
)
578 return node
->context
;
581 int dm_tree_node_size_changed(const struct dm_tree_node
*dnode
)
583 return dnode
->props
.size_changed
;
586 int dm_tree_node_num_children(const struct dm_tree_node
*node
, uint32_t inverted
)
589 if (_nodes_are_linked(&node
->dtree
->root
, node
))
591 return dm_list_size(&node
->used_by
);
594 if (_nodes_are_linked(node
, &node
->dtree
->root
))
597 return dm_list_size(&node
->uses
);
601 * Returns 1 if no prefix supplied
603 static int _uuid_prefix_matches(const char *uuid
, const char *uuid_prefix
, size_t uuid_prefix_len
)
605 const char *default_uuid_prefix
= dm_uuid_prefix();
606 size_t default_uuid_prefix_len
= strlen(default_uuid_prefix
);
611 if (!strncmp(uuid
, uuid_prefix
, uuid_prefix_len
))
614 /* Handle transition: active device uuids might be missing the prefix */
615 if (uuid_prefix_len
<= 4)
618 if (!strncmp(uuid
, default_uuid_prefix
, default_uuid_prefix_len
))
621 if (strncmp(uuid_prefix
, default_uuid_prefix
, default_uuid_prefix_len
))
624 if (!strncmp(uuid
, uuid_prefix
+ default_uuid_prefix_len
, uuid_prefix_len
- default_uuid_prefix_len
))
631 * Returns 1 if no children.
633 static int _children_suspended(struct dm_tree_node
*node
,
635 const char *uuid_prefix
,
636 size_t uuid_prefix_len
)
638 struct dm_list
*list
;
639 struct dm_tree_link
*dlink
;
640 const struct dm_info
*dinfo
;
644 if (_nodes_are_linked(&node
->dtree
->root
, node
))
646 list
= &node
->used_by
;
648 if (_nodes_are_linked(node
, &node
->dtree
->root
))
653 dm_list_iterate_items(dlink
, list
) {
654 if (!(uuid
= dm_tree_node_get_uuid(dlink
->node
))) {
659 /* Ignore if it doesn't belong to this VG */
660 if (!_uuid_prefix_matches(uuid
, uuid_prefix
, uuid_prefix_len
))
663 /* Ignore if parent node wants to presuspend this node */
664 if (dlink
->node
->presuspend_node
== node
)
667 if (!(dinfo
= dm_tree_node_get_info(dlink
->node
))) {
668 stack
; /* FIXME Is this normal? */
672 if (!dinfo
->suspended
)
680 * Set major and minor to zero for root of tree.
682 struct dm_tree_node
*dm_tree_find_node(struct dm_tree
*dtree
,
686 if (!major
&& !minor
)
689 return _find_dm_tree_node(dtree
, major
, minor
);
693 * Set uuid to NULL for root of tree.
695 struct dm_tree_node
*dm_tree_find_node_by_uuid(struct dm_tree
*dtree
,
701 return _find_dm_tree_node_by_uuid(dtree
, uuid
);
705 * First time set *handle to NULL.
706 * Set inverted to invert the tree.
708 struct dm_tree_node
*dm_tree_next_child(void **handle
,
709 const struct dm_tree_node
*parent
,
712 struct dm_list
**dlink
= (struct dm_list
**) handle
;
713 const struct dm_list
*use_list
;
716 use_list
= &parent
->used_by
;
718 use_list
= &parent
->uses
;
721 *dlink
= dm_list_first(use_list
);
723 *dlink
= dm_list_next(use_list
, *dlink
);
725 return (*dlink
) ? dm_list_item(*dlink
, struct dm_tree_link
)->node
: NULL
;
728 static int _deps(struct dm_task
**dmt
, struct dm_pool
*mem
, uint32_t major
, uint32_t minor
,
729 const char **name
, const char **uuid
, unsigned inactive_table
,
730 struct dm_info
*info
, struct dm_deps
**deps
)
732 memset(info
, 0, sizeof(*info
));
734 if (!dm_is_dm_major(major
)) {
745 if (!(*dmt
= dm_task_create(DM_DEVICE_DEPS
))) {
746 log_error("deps dm_task creation failed");
750 if (!dm_task_set_major(*dmt
, major
)) {
751 log_error("_deps: failed to set major for (%" PRIu32
":%" PRIu32
")",
756 if (!dm_task_set_minor(*dmt
, minor
)) {
757 log_error("_deps: failed to set minor for (%" PRIu32
":%" PRIu32
")",
762 if (inactive_table
&& !dm_task_query_inactive_table(*dmt
)) {
763 log_error("_deps: failed to set inactive table for (%" PRIu32
":%" PRIu32
")",
768 if (!dm_task_run(*dmt
)) {
769 log_error("_deps: task run failed for (%" PRIu32
":%" PRIu32
")",
774 if (!dm_task_get_info(*dmt
, info
)) {
775 log_error("_deps: failed to get info for (%" PRIu32
":%" PRIu32
")",
787 if (info
->major
!= major
) {
788 log_error("Inconsistent dtree major number: %u != %u",
792 if (info
->minor
!= minor
) {
793 log_error("Inconsistent dtree minor number: %u != %u",
797 if (name
&& !(*name
= dm_pool_strdup(mem
, dm_task_get_name(*dmt
)))) {
798 log_error("name pool_strdup failed");
801 if (uuid
&& !(*uuid
= dm_pool_strdup(mem
, dm_task_get_uuid(*dmt
)))) {
802 log_error("uuid pool_strdup failed");
805 *deps
= dm_task_get_deps(*dmt
);
811 dm_task_destroy(*dmt
);
816 * Deactivate a device with its dependencies if the uuid prefix matches.
818 static int _info_by_dev(uint32_t major
, uint32_t minor
, int with_open_count
,
819 struct dm_info
*info
, struct dm_pool
*mem
,
820 const char **name
, const char **uuid
)
825 if (!(dmt
= dm_task_create(DM_DEVICE_INFO
))) {
826 log_error("_info_by_dev: dm_task creation failed");
830 if (!dm_task_set_major(dmt
, major
) || !dm_task_set_minor(dmt
, minor
)) {
831 log_error("_info_by_dev: Failed to set device number");
832 dm_task_destroy(dmt
);
836 if (!with_open_count
&& !dm_task_no_open_count(dmt
))
837 log_error("Failed to disable open_count");
839 if (!(r
= dm_task_run(dmt
)))
842 if (!(r
= dm_task_get_info(dmt
, info
)))
845 if (name
&& !(*name
= dm_pool_strdup(mem
, dm_task_get_name(dmt
)))) {
846 log_error("name pool_strdup failed");
851 if (uuid
&& !(*uuid
= dm_pool_strdup(mem
, dm_task_get_uuid(dmt
)))) {
852 log_error("uuid pool_strdup failed");
858 dm_task_destroy(dmt
);
863 static int _check_device_not_in_use(const char *name
, struct dm_info
*info
)
868 /* If sysfs is not used, use open_count information only. */
869 if (!*dm_sysfs_dir()) {
870 if (info
->open_count
) {
871 log_error("Device %s (%" PRIu32
":%" PRIu32
") in use",
872 name
, info
->major
, info
->minor
);
879 if (dm_device_has_holders(info
->major
, info
->minor
)) {
880 log_error("Device %s (%" PRIu32
":%" PRIu32
") is used "
881 "by another device.", name
, info
->major
, info
->minor
);
885 if (dm_device_has_mounted_fs(info
->major
, info
->minor
)) {
886 log_error("Device %s (%" PRIu32
":%" PRIu32
") contains "
887 "a filesystem in use.", name
, info
->major
, info
->minor
);
894 /* Check if all parent nodes of given node have open_count == 0 */
895 static int _node_has_closed_parents(struct dm_tree_node
*node
,
896 const char *uuid_prefix
,
897 size_t uuid_prefix_len
)
899 struct dm_tree_link
*dlink
;
900 const struct dm_info
*dinfo
;
904 /* Iterate through parents of this node */
905 dm_list_iterate_items(dlink
, &node
->used_by
) {
906 if (!(uuid
= dm_tree_node_get_uuid(dlink
->node
))) {
911 /* Ignore if it doesn't belong to this VG */
912 if (!_uuid_prefix_matches(uuid
, uuid_prefix
, uuid_prefix_len
))
915 if (!(dinfo
= dm_tree_node_get_info(dlink
->node
))) {
916 stack
; /* FIXME Is this normal? */
920 /* Refresh open_count */
921 if (!_info_by_dev(dinfo
->major
, dinfo
->minor
, 1, &info
, NULL
, NULL
, NULL
) ||
925 if (info
.open_count
) {
926 log_debug("Node %s %d:%d has open_count %d", uuid_prefix
,
927 dinfo
->major
, dinfo
->minor
, info
.open_count
);
935 static int _deactivate_node(const char *name
, uint32_t major
, uint32_t minor
,
936 uint32_t *cookie
, uint16_t udev_flags
, int retry
)
941 log_verbose("Removing %s (%" PRIu32
":%" PRIu32
")", name
, major
, minor
);
943 if (!(dmt
= dm_task_create(DM_DEVICE_REMOVE
))) {
944 log_error("Deactivation dm_task creation failed for %s", name
);
948 if (!dm_task_set_major(dmt
, major
) || !dm_task_set_minor(dmt
, minor
)) {
949 log_error("Failed to set device number for %s deactivation", name
);
953 if (!dm_task_no_open_count(dmt
))
954 log_error("Failed to disable open_count");
957 if (!dm_task_set_cookie(dmt
, cookie
, udev_flags
))
961 dm_task_retry_remove(dmt
);
963 r
= dm_task_run(dmt
);
965 /* FIXME Until kernel returns actual name so dm-iface.c can handle it */
966 rm_dev_node(name
, dmt
->cookie_set
&& !(udev_flags
& DM_UDEV_DISABLE_DM_RULES_FLAG
),
967 dmt
->cookie_set
&& (udev_flags
& DM_UDEV_DISABLE_LIBRARY_FALLBACK
));
969 /* FIXME Remove node from tree or mark invalid? */
972 dm_task_destroy(dmt
);
977 static int _node_clear_table(struct dm_tree_node
*dnode
, uint16_t udev_flags
)
979 struct dm_task
*dmt
= NULL
, *deps_dmt
= NULL
;
980 struct dm_info
*info
, deps_info
;
981 struct dm_deps
*deps
= NULL
;
982 const char *name
, *uuid
;
983 const char *default_uuid_prefix
;
984 size_t default_uuid_prefix_len
;
988 if (!(info
= &dnode
->info
)) {
989 log_error("_node_clear_table failed: missing info");
993 if (!(name
= dm_tree_node_get_name(dnode
))) {
994 log_error("_node_clear_table failed: missing name");
998 /* Is there a table? */
999 if (!info
->exists
|| !info
->inactive_table
)
1002 /* Get devices used by inactive table that's about to be deleted. */
1003 if (!_deps(&deps_dmt
, dnode
->dtree
->mem
, info
->major
, info
->minor
, NULL
, NULL
, 1, info
, &deps
)) {
1004 log_error("Failed to obtain dependencies for %s before clearing table.", name
);
1008 log_verbose("Clearing inactive table %s (%" PRIu32
":%" PRIu32
")",
1009 name
, info
->major
, info
->minor
);
1011 if (!(dmt
= dm_task_create(DM_DEVICE_CLEAR
))) {
1012 log_error("Table clear dm_task creation failed for %s", name
);
1016 if (!dm_task_set_major(dmt
, info
->major
) ||
1017 !dm_task_set_minor(dmt
, info
->minor
)) {
1018 log_error("Failed to set device number for %s table clear", name
);
1022 r
= dm_task_run(dmt
);
1024 if (!dm_task_get_info(dmt
, info
)) {
1025 log_error("_node_clear_table failed: info missing after running task for %s", name
);
1033 * Remove (incomplete) devices that the inactive table referred to but
1034 * which are not in the tree, no longer referenced and don't have a live
1037 default_uuid_prefix
= dm_uuid_prefix();
1038 default_uuid_prefix_len
= strlen(default_uuid_prefix
);
1040 for (i
= 0; i
< deps
->count
; i
++) {
1041 /* If already in tree, assume it's under control */
1042 if (_find_dm_tree_node(dnode
->dtree
, MAJOR(deps
->device
[i
]), MINOR(deps
->device
[i
])))
1045 if (!_info_by_dev(MAJOR(deps
->device
[i
]), MINOR(deps
->device
[i
]), 1,
1046 &deps_info
, dnode
->dtree
->mem
, &name
, &uuid
))
1049 /* Proceed if device is an 'orphan' - unreferenced and without a live table. */
1050 if (!deps_info
.exists
|| deps_info
.live_table
|| deps_info
.open_count
)
1053 if (strncmp(uuid
, default_uuid_prefix
, default_uuid_prefix_len
))
1056 /* Remove device. */
1057 if (!_deactivate_node(name
, deps_info
.major
, deps_info
.minor
, &dnode
->dtree
->cookie
, udev_flags
, 0)) {
1058 log_error("Failed to deactivate no-longer-used device %s (%"
1059 PRIu32
":%" PRIu32
")", name
, deps_info
.major
, deps_info
.minor
);
1060 } else if (deps_info
.suspended
)
1066 dm_task_destroy(dmt
);
1069 dm_task_destroy(deps_dmt
);
1074 struct dm_tree_node
*dm_tree_add_new_dev_with_udev_flags(struct dm_tree
*dtree
,
1082 uint16_t udev_flags
)
1084 struct dm_tree_node
*dnode
;
1085 struct dm_info info
;
1089 if (!name
|| !uuid
) {
1090 log_error("Cannot add device without name and uuid.");
1094 /* Do we need to add node to tree? */
1095 if (!(dnode
= dm_tree_find_node_by_uuid(dtree
, uuid
))) {
1096 if (!(name2
= dm_pool_strdup(dtree
->mem
, name
))) {
1097 log_error("name pool_strdup failed");
1100 if (!(uuid2
= dm_pool_strdup(dtree
->mem
, uuid
))) {
1101 log_error("uuid pool_strdup failed");
1105 memset(&info
, 0, sizeof(info
));
1107 if (!(dnode
= _create_dm_tree_node(dtree
, name2
, uuid2
, &info
,
1111 /* Attach to root node until a table is supplied */
1112 if (!_add_to_toplevel(dnode
) || !_add_to_bottomlevel(dnode
))
1115 dnode
->props
.major
= major
;
1116 dnode
->props
.minor
= minor
;
1117 dnode
->props
.new_name
= NULL
;
1118 dnode
->props
.size_changed
= 0;
1119 } else if (strcmp(name
, dnode
->name
)) {
1120 /* Do we need to rename node? */
1121 if (!(dnode
->props
.new_name
= dm_pool_strdup(dtree
->mem
, name
))) {
1122 log_error("name pool_strdup failed");
1127 dnode
->props
.read_only
= read_only
? 1 : 0;
1128 dnode
->props
.read_ahead
= DM_READ_AHEAD_AUTO
;
1129 dnode
->props
.read_ahead_flags
= 0;
1131 if (clear_inactive
&& !_node_clear_table(dnode
, udev_flags
))
1134 dnode
->context
= context
;
1135 dnode
->udev_flags
= udev_flags
;
1140 struct dm_tree_node
*dm_tree_add_new_dev(struct dm_tree
*dtree
, const char *name
,
1141 const char *uuid
, uint32_t major
, uint32_t minor
,
1142 int read_only
, int clear_inactive
, void *context
)
1144 return dm_tree_add_new_dev_with_udev_flags(dtree
, name
, uuid
, major
, minor
,
1145 read_only
, clear_inactive
, context
, 0);
1148 static struct dm_tree_node
*_add_dev(struct dm_tree
*dtree
,
1149 struct dm_tree_node
*parent
,
1150 uint32_t major
, uint32_t minor
,
1151 uint16_t udev_flags
)
1153 struct dm_task
*dmt
= NULL
;
1154 struct dm_info info
;
1155 struct dm_deps
*deps
= NULL
;
1156 const char *name
= NULL
;
1157 const char *uuid
= NULL
;
1158 struct dm_tree_node
*node
= NULL
;
1162 /* Already in tree? */
1163 if (!(node
= _find_dm_tree_node(dtree
, major
, minor
))) {
1164 if (!_deps(&dmt
, dtree
->mem
, major
, minor
, &name
, &uuid
, 0, &info
, &deps
))
1167 if (!(node
= _create_dm_tree_node(dtree
, name
, uuid
, &info
,
1173 if (!_link_tree_nodes(parent
, node
)) {
1178 /* If node was already in tree, no need to recurse. */
1182 /* Can't recurse if not a mapped device or there are no dependencies */
1183 if (!node
->info
.exists
|| !deps
->count
) {
1184 if (!_add_to_bottomlevel(node
)) {
1191 /* Add dependencies to tree */
1192 for (i
= 0; i
< deps
->count
; i
++)
1193 if (!_add_dev(dtree
, node
, MAJOR(deps
->device
[i
]),
1194 MINOR(deps
->device
[i
]), udev_flags
)) {
1201 dm_task_destroy(dmt
);
1206 int dm_tree_add_dev(struct dm_tree
*dtree
, uint32_t major
, uint32_t minor
)
1208 return _add_dev(dtree
, &dtree
->root
, major
, minor
, 0) ? 1 : 0;
1211 int dm_tree_add_dev_with_udev_flags(struct dm_tree
*dtree
, uint32_t major
,
1212 uint32_t minor
, uint16_t udev_flags
)
1214 return _add_dev(dtree
, &dtree
->root
, major
, minor
, udev_flags
) ? 1 : 0;
1217 static int _rename_node(const char *old_name
, const char *new_name
, uint32_t major
,
1218 uint32_t minor
, uint32_t *cookie
, uint16_t udev_flags
)
1220 struct dm_task
*dmt
;
1223 log_verbose("Renaming %s (%" PRIu32
":%" PRIu32
") to %s", old_name
, major
, minor
, new_name
);
1225 if (!(dmt
= dm_task_create(DM_DEVICE_RENAME
))) {
1226 log_error("Rename dm_task creation failed for %s", old_name
);
1230 if (!dm_task_set_name(dmt
, old_name
)) {
1231 log_error("Failed to set name for %s rename.", old_name
);
1235 if (!dm_task_set_newname(dmt
, new_name
))
1238 if (!dm_task_no_open_count(dmt
))
1239 log_error("Failed to disable open_count");
1241 if (!dm_task_set_cookie(dmt
, cookie
, udev_flags
))
1244 r
= dm_task_run(dmt
);
1247 dm_task_destroy(dmt
);
1252 /* FIXME Merge with _suspend_node? */
1253 static int _resume_node(const char *name
, uint32_t major
, uint32_t minor
,
1254 uint32_t read_ahead
, uint32_t read_ahead_flags
,
1255 struct dm_info
*newinfo
, uint32_t *cookie
,
1256 uint16_t udev_flags
, int already_suspended
)
1258 struct dm_task
*dmt
;
1261 log_verbose("Resuming %s (%" PRIu32
":%" PRIu32
")", name
, major
, minor
);
1263 if (!(dmt
= dm_task_create(DM_DEVICE_RESUME
))) {
1264 log_debug("Suspend dm_task creation failed for %s.", name
);
1268 /* FIXME Kernel should fill in name on return instead */
1269 if (!dm_task_set_name(dmt
, name
)) {
1270 log_debug("Failed to set device name for %s resumption.", name
);
1274 if (!dm_task_set_major(dmt
, major
) || !dm_task_set_minor(dmt
, minor
)) {
1275 log_error("Failed to set device number for %s resumption.", name
);
1279 if (!dm_task_no_open_count(dmt
))
1280 log_error("Failed to disable open_count");
1282 if (!dm_task_set_read_ahead(dmt
, read_ahead
, read_ahead_flags
))
1283 log_error("Failed to set read ahead");
1285 if (!dm_task_set_cookie(dmt
, cookie
, udev_flags
))
1288 if (!(r
= dm_task_run(dmt
)))
1291 if (already_suspended
)
1294 if (!(r
= dm_task_get_info(dmt
, newinfo
)))
1298 dm_task_destroy(dmt
);
1303 static int _suspend_node(const char *name
, uint32_t major
, uint32_t minor
,
1304 int skip_lockfs
, int no_flush
, struct dm_info
*newinfo
)
1306 struct dm_task
*dmt
;
1309 log_verbose("Suspending %s (%" PRIu32
":%" PRIu32
")%s%s",
1311 skip_lockfs
? "" : " with filesystem sync",
1312 no_flush
? "" : " with device flush");
1314 if (!(dmt
= dm_task_create(DM_DEVICE_SUSPEND
))) {
1315 log_error("Suspend dm_task creation failed for %s", name
);
1319 if (!dm_task_set_major(dmt
, major
) || !dm_task_set_minor(dmt
, minor
)) {
1320 log_error("Failed to set device number for %s suspension.", name
);
1321 dm_task_destroy(dmt
);
1325 if (!dm_task_no_open_count(dmt
))
1326 log_error("Failed to disable open_count");
1328 if (skip_lockfs
&& !dm_task_skip_lockfs(dmt
))
1329 log_error("Failed to set skip_lockfs flag.");
1331 if (no_flush
&& !dm_task_no_flush(dmt
))
1332 log_error("Failed to set no_flush flag.");
1334 if ((r
= dm_task_run(dmt
))) {
1336 r
= dm_task_get_info(dmt
, newinfo
);
1339 dm_task_destroy(dmt
);
1344 static int _thin_pool_status_transaction_id(struct dm_tree_node
*dnode
, uint64_t *transaction_id
)
1346 struct dm_task
*dmt
;
1348 uint64_t start
, length
;
1350 char *params
= NULL
;
1352 if (!(dmt
= dm_task_create(DM_DEVICE_STATUS
)))
1355 if (!dm_task_set_major(dmt
, dnode
->info
.major
) ||
1356 !dm_task_set_minor(dmt
, dnode
->info
.minor
)) {
1357 log_error("Failed to set major minor.");
1361 if (!dm_task_run(dmt
))
1364 dm_get_next_target(dmt
, NULL
, &start
, &length
, &type
, ¶ms
);
1366 if (type
&& (strcmp(type
, "thin-pool") != 0)) {
1367 log_error("Expected thin-pool target for %d:%d and got %s.",
1368 dnode
->info
.major
, dnode
->info
.minor
, type
);
1372 if (!params
|| (sscanf(params
, "%" PRIu64
, transaction_id
) != 1)) {
1373 log_error("Failed to parse transaction_id from %s.", params
);
1377 log_debug("Thin pool transaction id: %" PRIu64
" status: %s.", *transaction_id
, params
);
1381 dm_task_destroy(dmt
);
1386 static int _thin_pool_node_message(struct dm_tree_node
*dnode
, struct thin_message
*tm
)
1388 struct dm_task
*dmt
;
1389 struct dm_thin_message
*m
= &tm
->message
;
1394 case DM_THIN_MESSAGE_CREATE_SNAP
:
1395 r
= dm_snprintf(buf
, sizeof(buf
), "create_snap %u %u",
1396 m
->u
.m_create_snap
.device_id
,
1397 m
->u
.m_create_snap
.origin_id
);
1399 case DM_THIN_MESSAGE_CREATE_THIN
:
1400 r
= dm_snprintf(buf
, sizeof(buf
), "create_thin %u",
1401 m
->u
.m_create_thin
.device_id
);
1403 case DM_THIN_MESSAGE_DELETE
:
1404 r
= dm_snprintf(buf
, sizeof(buf
), "delete %u",
1405 m
->u
.m_delete
.device_id
);
1407 case DM_THIN_MESSAGE_TRIM
:
1408 r
= dm_snprintf(buf
, sizeof(buf
), "trim %u %" PRIu64
,
1409 m
->u
.m_trim
.device_id
,
1410 m
->u
.m_trim
.new_size
);
1412 case DM_THIN_MESSAGE_SET_TRANSACTION_ID
:
1413 r
= dm_snprintf(buf
, sizeof(buf
),
1414 "set_transaction_id %" PRIu64
" %" PRIu64
,
1415 m
->u
.m_set_transaction_id
.current_id
,
1416 m
->u
.m_set_transaction_id
.new_id
);
1423 log_error("Failed to prepare message.");
1429 if (!(dmt
= dm_task_create(DM_DEVICE_TARGET_MSG
)))
1432 if (!dm_task_set_major(dmt
, dnode
->info
.major
) ||
1433 !dm_task_set_minor(dmt
, dnode
->info
.minor
)) {
1434 log_error("Failed to set message major minor.");
1438 if (!dm_task_set_message(dmt
, buf
))
1441 /* Internal functionality of dm_task */
1442 dmt
->expected_errno
= tm
->expected_errno
;
1444 if (!dm_task_run(dmt
))
1449 dm_task_destroy(dmt
);
1454 static int _node_send_messages(struct dm_tree_node
*dnode
,
1455 const char *uuid_prefix
,
1456 size_t uuid_prefix_len
)
1458 struct load_segment
*seg
;
1459 struct thin_message
*tmsg
;
1463 if (!dnode
->info
.exists
|| (dm_list_size(&dnode
->props
.segs
) != 1))
1466 seg
= dm_list_item(dm_list_last(&dnode
->props
.segs
), struct load_segment
);
1467 if (seg
->type
!= SEG_THIN_POOL
)
1470 if (!(uuid
= dm_tree_node_get_uuid(dnode
)))
1473 if (!_uuid_prefix_matches(uuid
, uuid_prefix
, uuid_prefix_len
)) {
1474 log_debug("UUID \"%s\" does not match.", uuid
);
1478 if (!_thin_pool_status_transaction_id(dnode
, &trans_id
))
1481 if (trans_id
== seg
->transaction_id
)
1482 return 1; /* In sync - skip messages */
1484 if (trans_id
!= (seg
->transaction_id
- 1)) {
1485 log_error("Thin pool transaction_id=%" PRIu64
", while expected: %" PRIu64
".",
1486 trans_id
, seg
->transaction_id
- 1);
1487 goto bad
; /* Nothing to send */
1490 dm_list_iterate_items(tmsg
, &seg
->thin_messages
)
1491 if (!(_thin_pool_node_message(dnode
, tmsg
)))
1496 /* Try to deactivate */
1497 if (!(dm_tree_deactivate_children(dnode
, uuid_prefix
, uuid_prefix_len
)))
1498 log_error("Failed to deactivate %s", dnode
->name
);
1504 * FIXME Don't attempt to deactivate known internal dependencies.
1506 static int _dm_tree_deactivate_children(struct dm_tree_node
*dnode
,
1507 const char *uuid_prefix
,
1508 size_t uuid_prefix_len
,
1512 void *handle
= NULL
;
1513 struct dm_tree_node
*child
= dnode
;
1514 struct dm_info info
;
1515 const struct dm_info
*dinfo
;
1519 while ((child
= dm_tree_next_child(&handle
, dnode
, 0))) {
1520 if (!(dinfo
= dm_tree_node_get_info(child
))) {
1525 if (!(name
= dm_tree_node_get_name(child
))) {
1530 if (!(uuid
= dm_tree_node_get_uuid(child
))) {
1535 /* Ignore if it doesn't belong to this VG */
1536 if (!_uuid_prefix_matches(uuid
, uuid_prefix
, uuid_prefix_len
))
1539 /* Refresh open_count */
1540 if (!_info_by_dev(dinfo
->major
, dinfo
->minor
, 1, &info
, NULL
, NULL
, NULL
) ||
1544 if (info
.open_count
) {
1545 /* Skip internal non-toplevel opened nodes */
1549 /* When retry is not allowed, error */
1550 if (!child
->dtree
->retry_remove
) {
1551 log_error("Unable to deactivate open %s (%" PRIu32
1552 ":%" PRIu32
")", name
, info
.major
, info
.minor
);
1557 /* Check toplevel node for holders/mounted fs */
1558 if (!_check_device_not_in_use(name
, &info
)) {
1563 /* Go on with retry */
1566 /* Also checking open_count in parent nodes of presuspend_node */
1567 if ((child
->presuspend_node
&&
1568 !_node_has_closed_parents(child
->presuspend_node
,
1569 uuid_prefix
, uuid_prefix_len
))) {
1570 /* Only report error from (likely non-internal) dependency at top level */
1572 log_error("Unable to deactivate open %s (%" PRIu32
1573 ":%" PRIu32
")", name
, info
.major
,
1580 /* Suspend child node first if requested */
1581 if (child
->presuspend_node
&&
1582 !dm_tree_suspend_children(child
, uuid_prefix
, uuid_prefix_len
))
1585 if (!_deactivate_node(name
, info
.major
, info
.minor
,
1586 &child
->dtree
->cookie
, child
->udev_flags
,
1587 (level
== 0) ? child
->dtree
->retry_remove
: 0)) {
1588 log_error("Unable to deactivate %s (%" PRIu32
1589 ":%" PRIu32
")", name
, info
.major
,
1593 } else if (info
.suspended
)
1596 if (dm_tree_node_num_children(child
, 0)) {
1597 if (!_dm_tree_deactivate_children(child
, uuid_prefix
, uuid_prefix_len
, level
+ 1))
1605 int dm_tree_deactivate_children(struct dm_tree_node
*dnode
,
1606 const char *uuid_prefix
,
1607 size_t uuid_prefix_len
)
1609 return _dm_tree_deactivate_children(dnode
, uuid_prefix
, uuid_prefix_len
, 0);
1612 int dm_tree_suspend_children(struct dm_tree_node
*dnode
,
1613 const char *uuid_prefix
,
1614 size_t uuid_prefix_len
)
1617 void *handle
= NULL
;
1618 struct dm_tree_node
*child
= dnode
;
1619 struct dm_info info
, newinfo
;
1620 const struct dm_info
*dinfo
;
1624 /* Suspend nodes at this level of the tree */
1625 while ((child
= dm_tree_next_child(&handle
, dnode
, 0))) {
1626 if (!(dinfo
= dm_tree_node_get_info(child
))) {
1631 if (!(name
= dm_tree_node_get_name(child
))) {
1636 if (!(uuid
= dm_tree_node_get_uuid(child
))) {
1641 /* Ignore if it doesn't belong to this VG */
1642 if (!_uuid_prefix_matches(uuid
, uuid_prefix
, uuid_prefix_len
))
1645 /* Ensure immediate parents are already suspended */
1646 if (!_children_suspended(child
, 1, uuid_prefix
, uuid_prefix_len
))
1649 if (!_info_by_dev(dinfo
->major
, dinfo
->minor
, 0, &info
, NULL
, NULL
, NULL
) ||
1650 !info
.exists
|| info
.suspended
)
1653 if (!_suspend_node(name
, info
.major
, info
.minor
,
1654 child
->dtree
->skip_lockfs
,
1655 child
->dtree
->no_flush
, &newinfo
)) {
1656 log_error("Unable to suspend %s (%" PRIu32
1657 ":%" PRIu32
")", name
, info
.major
,
1663 /* Update cached info */
1664 child
->info
= newinfo
;
1667 /* Then suspend any child nodes */
1670 while ((child
= dm_tree_next_child(&handle
, dnode
, 0))) {
1671 if (!(uuid
= dm_tree_node_get_uuid(child
))) {
1676 /* Ignore if it doesn't belong to this VG */
1677 if (!_uuid_prefix_matches(uuid
, uuid_prefix
, uuid_prefix_len
))
1680 if (dm_tree_node_num_children(child
, 0))
1681 if (!dm_tree_suspend_children(child
, uuid_prefix
, uuid_prefix_len
))
1688 int dm_tree_activate_children(struct dm_tree_node
*dnode
,
1689 const char *uuid_prefix
,
1690 size_t uuid_prefix_len
)
1693 void *handle
= NULL
;
1694 struct dm_tree_node
*child
= dnode
;
1695 struct dm_info newinfo
;
1700 /* Activate children first */
1701 while ((child
= dm_tree_next_child(&handle
, dnode
, 0))) {
1702 if (!(uuid
= dm_tree_node_get_uuid(child
))) {
1707 if (!_uuid_prefix_matches(uuid
, uuid_prefix
, uuid_prefix_len
))
1710 if (dm_tree_node_num_children(child
, 0))
1711 if (!dm_tree_activate_children(child
, uuid_prefix
, uuid_prefix_len
))
1717 for (priority
= 0; priority
< 3; priority
++) {
1718 while ((child
= dm_tree_next_child(&handle
, dnode
, 0))) {
1719 if (priority
!= child
->activation_priority
)
1722 if (!(uuid
= dm_tree_node_get_uuid(child
))) {
1727 if (!_uuid_prefix_matches(uuid
, uuid_prefix
, uuid_prefix_len
))
1730 if (!(name
= dm_tree_node_get_name(child
))) {
1736 if (child
->props
.new_name
) {
1737 if (!_rename_node(name
, child
->props
.new_name
, child
->info
.major
,
1738 child
->info
.minor
, &child
->dtree
->cookie
,
1739 child
->udev_flags
)) {
1740 log_error("Failed to rename %s (%" PRIu32
1741 ":%" PRIu32
") to %s", name
, child
->info
.major
,
1742 child
->info
.minor
, child
->props
.new_name
);
1745 child
->name
= child
->props
.new_name
;
1746 child
->props
.new_name
= NULL
;
1749 if (!child
->info
.inactive_table
&& !child
->info
.suspended
)
1752 if (!_resume_node(child
->name
, child
->info
.major
, child
->info
.minor
,
1753 child
->props
.read_ahead
, child
->props
.read_ahead_flags
,
1754 &newinfo
, &child
->dtree
->cookie
, child
->udev_flags
, child
->info
.suspended
)) {
1755 log_error("Unable to resume %s (%" PRIu32
1756 ":%" PRIu32
")", child
->name
, child
->info
.major
,
1762 /* Update cached info */
1763 child
->info
= newinfo
;
1768 * FIXME: Implement delayed error reporting
1769 * activation should be stopped only in the case,
1770 * the submission of transation_id message fails,
1771 * resume should continue further, just whole command
1772 * has to report failure.
1774 if (r
&& dnode
->props
.send_messages
&&
1775 !(r
= _node_send_messages(dnode
, uuid_prefix
, uuid_prefix_len
)))
1783 static int _create_node(struct dm_tree_node
*dnode
)
1786 struct dm_task
*dmt
;
1788 log_verbose("Creating %s", dnode
->name
);
1790 if (!(dmt
= dm_task_create(DM_DEVICE_CREATE
))) {
1791 log_error("Create dm_task creation failed for %s", dnode
->name
);
1795 if (!dm_task_set_name(dmt
, dnode
->name
)) {
1796 log_error("Failed to set device name for %s", dnode
->name
);
1800 if (!dm_task_set_uuid(dmt
, dnode
->uuid
)) {
1801 log_error("Failed to set uuid for %s", dnode
->name
);
1805 if (dnode
->props
.major
&&
1806 (!dm_task_set_major(dmt
, dnode
->props
.major
) ||
1807 !dm_task_set_minor(dmt
, dnode
->props
.minor
))) {
1808 log_error("Failed to set device number for %s creation.", dnode
->name
);
1812 if (dnode
->props
.read_only
&& !dm_task_set_ro(dmt
)) {
1813 log_error("Failed to set read only flag for %s", dnode
->name
);
1817 if (!dm_task_no_open_count(dmt
))
1818 log_error("Failed to disable open_count");
1820 if ((r
= dm_task_run(dmt
)))
1821 r
= dm_task_get_info(dmt
, &dnode
->info
);
1824 dm_task_destroy(dmt
);
1830 static int _build_dev_string(char *devbuf
, size_t bufsize
, struct dm_tree_node
*node
)
1832 if (!dm_format_dev(devbuf
, bufsize
, node
->info
.major
, node
->info
.minor
)) {
1833 log_error("Failed to format %s device number for %s as dm "
1835 node
->name
, node
->uuid
, node
->info
.major
, node
->info
.minor
);
1842 /* simplify string emiting code */
1843 #define EMIT_PARAMS(p, str...)\
1846 if ((w = dm_snprintf(params + p, paramsize - (size_t) p, str)) < 0) {\
1847 stack; /* Out of space */\
1856 * Returns: 1 on success, 0 on failure
1858 static int _emit_areas_line(struct dm_task
*dmt
__attribute__((unused
)),
1859 struct load_segment
*seg
, char *params
,
1860 size_t paramsize
, int *pos
)
1862 struct seg_area
*area
;
1863 char devbuf
[DM_FORMAT_DEV_BUFSIZE
];
1864 unsigned first_time
= 1;
1865 const char *logtype
, *synctype
;
1866 unsigned log_parm_count
;
1868 dm_list_iterate_items(area
, &seg
->areas
) {
1869 switch (seg
->type
) {
1870 case SEG_REPLICATOR_DEV
:
1871 if (!_build_dev_string(devbuf
, sizeof(devbuf
), area
->dev_node
))
1874 EMIT_PARAMS(*pos
, " %d 1 %s", area
->rsite_index
, devbuf
);
1876 EMIT_PARAMS(*pos
, " nolog 0");
1878 /* Remote devices */
1879 log_parm_count
= (area
->flags
&
1880 (DM_NOSYNC
| DM_FORCESYNC
)) ? 2 : 1;
1883 devbuf
[0] = 0; /* Only core log parameters */
1886 devbuf
[0] = ' '; /* Extra space before device name */
1887 if (!_build_dev_string(devbuf
+ 1,
1892 log_parm_count
++; /* Extra sync log device name parameter */
1895 EMIT_PARAMS(*pos
, " %s %u%s %" PRIu64
, logtype
,
1896 log_parm_count
, devbuf
, area
->region_size
);
1898 synctype
= (area
->flags
& DM_NOSYNC
) ?
1899 " nosync" : (area
->flags
& DM_FORCESYNC
) ?
1903 EMIT_PARAMS(*pos
, "%s", synctype
);
1915 if (!area
->dev_node
) {
1916 EMIT_PARAMS(*pos
, " -");
1919 if (!_build_dev_string(devbuf
, sizeof(devbuf
), area
->dev_node
))
1922 EMIT_PARAMS(*pos
, " %s", devbuf
);
1925 if (!_build_dev_string(devbuf
, sizeof(devbuf
), area
->dev_node
))
1928 EMIT_PARAMS(*pos
, "%s%s %" PRIu64
, first_time
? "" : " ",
1929 devbuf
, area
->offset
);
1938 static int _replicator_emit_segment_line(const struct load_segment
*seg
, char *params
,
1939 size_t paramsize
, int *pos
)
1941 const struct load_segment
*rlog_seg
;
1942 struct replicator_site
*rsite
;
1943 char rlogbuf
[DM_FORMAT_DEV_BUFSIZE
];
1944 unsigned parm_count
;
1946 if (!seg
->log
|| !_build_dev_string(rlogbuf
, sizeof(rlogbuf
), seg
->log
))
1949 rlog_seg
= dm_list_item(dm_list_last(&seg
->log
->props
.segs
),
1950 struct load_segment
);
1952 EMIT_PARAMS(*pos
, "%s 4 %s 0 auto %" PRIu64
,
1953 seg
->rlog_type
, rlogbuf
, rlog_seg
->size
);
1955 dm_list_iterate_items(rsite
, &seg
->rsites
) {
1956 parm_count
= (rsite
->fall_behind_data
1957 || rsite
->fall_behind_ios
1958 || rsite
->async_timeout
) ? 4 : 2;
1960 EMIT_PARAMS(*pos
, " blockdev %u %u %s", parm_count
, rsite
->rsite_index
,
1961 (rsite
->mode
== DM_REPLICATOR_SYNC
) ? "synchronous" : "asynchronous");
1963 if (rsite
->fall_behind_data
)
1964 EMIT_PARAMS(*pos
, " data %" PRIu64
, rsite
->fall_behind_data
);
1965 else if (rsite
->fall_behind_ios
)
1966 EMIT_PARAMS(*pos
, " ios %" PRIu32
, rsite
->fall_behind_ios
);
1967 else if (rsite
->async_timeout
)
1968 EMIT_PARAMS(*pos
, " timeout %" PRIu32
, rsite
->async_timeout
);
1975 * Returns: 1 on success, 0 on failure
1977 static int _mirror_emit_segment_line(struct dm_task
*dmt
, struct load_segment
*seg
,
1978 char *params
, size_t paramsize
)
1980 int block_on_error
= 0;
1981 int handle_errors
= 0;
1982 int dm_log_userspace
= 0;
1984 unsigned log_parm_count
;
1986 char logbuf
[DM_FORMAT_DEV_BUFSIZE
];
1987 const char *logtype
;
1988 unsigned kmaj
= 0, kmin
= 0, krel
= 0;
1990 if (uname(&uts
) == -1) {
1991 log_error("Cannot read kernel release version.");
1995 /* Kernels with a major number of 2 always had 3 parts. */
1996 parts
= sscanf(uts
.release
, "%u.%u.%u", &kmaj
, &kmin
, &krel
);
1997 if (parts
< 1 || (kmaj
< 3 && parts
< 3)) {
1998 log_error("Wrong kernel release version %s.", uts
.release
);
2002 if ((seg
->flags
& DM_BLOCK_ON_ERROR
)) {
2004 * Originally, block_on_error was an argument to the log
2005 * portion of the mirror CTR table. It was renamed to
2006 * "handle_errors" and now resides in the 'features'
2007 * section of the mirror CTR table (i.e. at the end).
2009 * We can identify whether to use "block_on_error" or
2010 * "handle_errors" by the dm-mirror module's version
2011 * number (>= 1.12) or by the kernel version (>= 2.6.22).
2013 if (KERNEL_VERSION(kmaj
, kmin
, krel
) >= KERNEL_VERSION(2, 6, 22))
2019 if (seg
->clustered
) {
2020 /* Cluster mirrors require a UUID */
2025 * Cluster mirrors used to have their own log
2026 * types. Now they are accessed through the
2027 * userspace log type.
2029 * The dm-log-userspace module was added to the
2032 if (KERNEL_VERSION(kmaj
, kmin
, krel
) >= KERNEL_VERSION(2, 6, 31))
2033 dm_log_userspace
= 1;
2039 /* [no]sync, block_on_error etc. */
2040 log_parm_count
+= hweight32(seg
->flags
);
2042 /* "handle_errors" is a feature arg now */
2046 /* DM_CORELOG does not count in the param list */
2047 if (seg
->flags
& DM_CORELOG
)
2050 if (seg
->clustered
) {
2051 log_parm_count
++; /* For UUID */
2053 if (!dm_log_userspace
)
2054 EMIT_PARAMS(pos
, "clustered-");
2056 /* For clustered-* type field inserted later */
2065 if (!_build_dev_string(logbuf
, sizeof(logbuf
), seg
->log
))
2069 if (dm_log_userspace
)
2070 EMIT_PARAMS(pos
, "userspace %u %s clustered-%s",
2071 log_parm_count
, seg
->uuid
, logtype
);
2073 EMIT_PARAMS(pos
, "%s %u", logtype
, log_parm_count
);
2076 EMIT_PARAMS(pos
, " %s", logbuf
);
2078 EMIT_PARAMS(pos
, " %u", seg
->region_size
);
2080 if (seg
->clustered
&& !dm_log_userspace
)
2081 EMIT_PARAMS(pos
, " %s", seg
->uuid
);
2083 if ((seg
->flags
& DM_NOSYNC
))
2084 EMIT_PARAMS(pos
, " nosync");
2085 else if ((seg
->flags
& DM_FORCESYNC
))
2086 EMIT_PARAMS(pos
, " sync");
2089 EMIT_PARAMS(pos
, " block_on_error");
2091 EMIT_PARAMS(pos
, " %u ", seg
->mirror_area_count
);
2093 if (_emit_areas_line(dmt
, seg
, params
, paramsize
, &pos
) <= 0)
2097 EMIT_PARAMS(pos
, " 1 handle_errors");
2102 static int _raid_emit_segment_line(struct dm_task
*dmt
, uint32_t major
,
2103 uint32_t minor
, struct load_segment
*seg
,
2104 uint64_t *seg_start
, char *params
,
2108 int param_count
= 1; /* mandatory 'chunk size'/'stripe size' arg */
2111 if ((seg
->flags
& DM_NOSYNC
) || (seg
->flags
& DM_FORCESYNC
))
2114 if (seg
->region_size
)
2117 /* rebuilds is 64-bit */
2118 param_count
+= 2 * hweight32(seg
->rebuilds
& 0xFFFFFFFF);
2119 param_count
+= 2 * hweight32(seg
->rebuilds
>> 32);
2121 if ((seg
->type
== SEG_RAID1
) && seg
->stripe_size
)
2122 log_error("WARNING: Ignoring RAID1 stripe size");
2124 EMIT_PARAMS(pos
, "%s %d %u", dm_segtypes
[seg
->type
].target
,
2125 param_count
, seg
->stripe_size
);
2127 if (seg
->flags
& DM_NOSYNC
)
2128 EMIT_PARAMS(pos
, " nosync");
2129 else if (seg
->flags
& DM_FORCESYNC
)
2130 EMIT_PARAMS(pos
, " sync");
2132 if (seg
->region_size
)
2133 EMIT_PARAMS(pos
, " region_size %u", seg
->region_size
);
2135 for (i
= 0; i
< (seg
->area_count
/ 2); i
++)
2136 if (seg
->rebuilds
& (1 << i
))
2137 EMIT_PARAMS(pos
, " rebuild %u", i
);
2139 /* Print number of metadata/data device pairs */
2140 EMIT_PARAMS(pos
, " %u", seg
->area_count
/2);
2142 if (_emit_areas_line(dmt
, seg
, params
, paramsize
, &pos
) <= 0)
2148 static int _emit_segment_line(struct dm_task
*dmt
, uint32_t major
,
2149 uint32_t minor
, struct load_segment
*seg
,
2150 uint64_t *seg_start
, char *params
,
2155 int target_type_is_raid
= 0;
2156 char originbuf
[DM_FORMAT_DEV_BUFSIZE
], cowbuf
[DM_FORMAT_DEV_BUFSIZE
];
2157 char pool
[DM_FORMAT_DEV_BUFSIZE
], metadata
[DM_FORMAT_DEV_BUFSIZE
];
2165 /* Mirrors are pretty complicated - now in separate function */
2166 r
= _mirror_emit_segment_line(dmt
, seg
, params
, paramsize
);
2170 case SEG_REPLICATOR
:
2171 if ((r
= _replicator_emit_segment_line(seg
, params
, paramsize
,
2177 case SEG_REPLICATOR_DEV
:
2178 if (!seg
->replicator
|| !_build_dev_string(originbuf
,
2183 EMIT_PARAMS(pos
, "%s %" PRIu64
, originbuf
, seg
->rdevice_index
);
2186 case SEG_SNAPSHOT_MERGE
:
2187 if (!_build_dev_string(originbuf
, sizeof(originbuf
), seg
->origin
))
2189 if (!_build_dev_string(cowbuf
, sizeof(cowbuf
), seg
->cow
))
2191 EMIT_PARAMS(pos
, "%s %s %c %d", originbuf
, cowbuf
,
2192 seg
->persistent
? 'P' : 'N', seg
->chunk_size
);
2194 case SEG_SNAPSHOT_ORIGIN
:
2195 if (!_build_dev_string(originbuf
, sizeof(originbuf
), seg
->origin
))
2197 EMIT_PARAMS(pos
, "%s", originbuf
);
2200 EMIT_PARAMS(pos
, "%u %u ", seg
->area_count
, seg
->stripe_size
);
2203 EMIT_PARAMS(pos
, "%s%s%s%s%s %s %" PRIu64
" ", seg
->cipher
,
2204 seg
->chainmode
? "-" : "", seg
->chainmode
?: "",
2205 seg
->iv
? "-" : "", seg
->iv
?: "", seg
->key
,
2206 seg
->iv_offset
!= DM_CRYPT_IV_DEFAULT
?
2207 seg
->iv_offset
: *seg_start
);
2218 target_type_is_raid
= 1;
2219 r
= _raid_emit_segment_line(dmt
, major
, minor
, seg
, seg_start
,
2226 if (!_build_dev_string(metadata
, sizeof(metadata
), seg
->metadata
))
2228 if (!_build_dev_string(pool
, sizeof(pool
), seg
->pool
))
2230 EMIT_PARAMS(pos
, "%s %s %d %" PRIu64
" %s", metadata
, pool
,
2231 seg
->data_block_size
, seg
->low_water_mark
,
2232 seg
->skip_block_zeroing
? "1 skip_block_zeroing" : "0");
2235 if (!_build_dev_string(pool
, sizeof(pool
), seg
->pool
))
2237 EMIT_PARAMS(pos
, "%s %d", pool
, seg
->device_id
);
2243 case SEG_REPLICATOR
:
2245 case SEG_SNAPSHOT_ORIGIN
:
2246 case SEG_SNAPSHOT_MERGE
:
2253 case SEG_REPLICATOR_DEV
:
2255 if ((r
= _emit_areas_line(dmt
, seg
, params
, paramsize
, &pos
)) <= 0) {
2260 log_error("No parameters supplied for %s target "
2261 "%u:%u.", dm_segtypes
[seg
->type
].target
,
2268 log_debug("Adding target to (%" PRIu32
":%" PRIu32
"): %" PRIu64
2269 " %" PRIu64
" %s %s", major
, minor
,
2270 *seg_start
, seg
->size
, target_type_is_raid
? "raid" :
2271 dm_segtypes
[seg
->type
].target
, params
);
2273 if (!dm_task_add_target(dmt
, *seg_start
, seg
->size
,
2274 target_type_is_raid
? "raid" :
2275 dm_segtypes
[seg
->type
].target
, params
))
2278 *seg_start
+= seg
->size
;
2285 static int _emit_segment(struct dm_task
*dmt
, uint32_t major
, uint32_t minor
,
2286 struct load_segment
*seg
, uint64_t *seg_start
)
2289 size_t paramsize
= 4096;
2293 if (!(params
= dm_malloc(paramsize
))) {
2294 log_error("Insufficient space for target parameters.");
2299 ret
= _emit_segment_line(dmt
, major
, minor
, seg
, seg_start
,
2309 log_debug("Insufficient space in params[%" PRIsize_t
2310 "] for target parameters.", paramsize
);
2313 } while (paramsize
< MAX_TARGET_PARAMSIZE
);
2315 log_error("Target parameter size too big. Aborting.");
2319 static int _load_node(struct dm_tree_node
*dnode
)
2322 struct dm_task
*dmt
;
2323 struct load_segment
*seg
;
2324 uint64_t seg_start
= 0, existing_table_size
;
2326 log_verbose("Loading %s table (%" PRIu32
":%" PRIu32
")", dnode
->name
,
2327 dnode
->info
.major
, dnode
->info
.minor
);
2329 if (!(dmt
= dm_task_create(DM_DEVICE_RELOAD
))) {
2330 log_error("Reload dm_task creation failed for %s", dnode
->name
);
2334 if (!dm_task_set_major(dmt
, dnode
->info
.major
) ||
2335 !dm_task_set_minor(dmt
, dnode
->info
.minor
)) {
2336 log_error("Failed to set device number for %s reload.", dnode
->name
);
2340 if (dnode
->props
.read_only
&& !dm_task_set_ro(dmt
)) {
2341 log_error("Failed to set read only flag for %s", dnode
->name
);
2345 if (!dm_task_no_open_count(dmt
))
2346 log_error("Failed to disable open_count");
2348 dm_list_iterate_items(seg
, &dnode
->props
.segs
)
2349 if (!_emit_segment(dmt
, dnode
->info
.major
, dnode
->info
.minor
,
2353 if (!dm_task_suppress_identical_reload(dmt
))
2354 log_error("Failed to suppress reload of identical tables.");
2356 if ((r
= dm_task_run(dmt
))) {
2357 r
= dm_task_get_info(dmt
, &dnode
->info
);
2358 if (r
&& !dnode
->info
.inactive_table
)
2359 log_verbose("Suppressed %s identical table reload.",
2362 existing_table_size
= dm_task_get_existing_table_size(dmt
);
2363 if ((dnode
->props
.size_changed
=
2364 (existing_table_size
== seg_start
) ? 0 : 1)) {
2365 log_debug("Table size changed from %" PRIu64
" to %"
2366 PRIu64
" for %s", existing_table_size
,
2367 seg_start
, dnode
->name
);
2369 * Kernel usually skips size validation on zero-length devices
2370 * now so no need to preload them.
2372 /* FIXME In which kernel version did this begin? */
2373 if (!existing_table_size
&& dnode
->props
.delay_resume_if_new
)
2374 dnode
->props
.size_changed
= 0;
2378 dnode
->props
.segment_count
= 0;
2381 dm_task_destroy(dmt
);
2386 int dm_tree_preload_children(struct dm_tree_node
*dnode
,
2387 const char *uuid_prefix
,
2388 size_t uuid_prefix_len
)
2391 void *handle
= NULL
;
2392 struct dm_tree_node
*child
;
2393 struct dm_info newinfo
;
2394 int update_devs_flag
= 0;
2396 /* Preload children first */
2397 while ((child
= dm_tree_next_child(&handle
, dnode
, 0))) {
2398 /* Skip existing non-device-mapper devices */
2399 if (!child
->info
.exists
&& child
->info
.major
)
2402 /* Ignore if it doesn't belong to this VG */
2403 if (child
->info
.exists
&&
2404 !_uuid_prefix_matches(child
->uuid
, uuid_prefix
, uuid_prefix_len
))
2407 if (dm_tree_node_num_children(child
, 0))
2408 if (!dm_tree_preload_children(child
, uuid_prefix
, uuid_prefix_len
))
2411 /* FIXME Cope if name exists with no uuid? */
2412 if (!child
->info
.exists
&& !_create_node(child
))
2415 if (!child
->info
.inactive_table
&&
2416 child
->props
.segment_count
&&
2420 /* Propagate device size change change */
2421 if (child
->props
.size_changed
)
2422 dnode
->props
.size_changed
= 1;
2424 /* Resume device immediately if it has parents and its size changed */
2425 if (!dm_tree_node_num_children(child
, 1) || !child
->props
.size_changed
)
2428 if (!child
->info
.inactive_table
&& !child
->info
.suspended
)
2431 if (!_resume_node(child
->name
, child
->info
.major
, child
->info
.minor
,
2432 child
->props
.read_ahead
, child
->props
.read_ahead_flags
,
2433 &newinfo
, &child
->dtree
->cookie
, child
->udev_flags
,
2434 child
->info
.suspended
)) {
2435 log_error("Unable to resume %s (%" PRIu32
2436 ":%" PRIu32
")", child
->name
, child
->info
.major
,
2442 /* Update cached info */
2443 child
->info
= newinfo
;
2445 * Prepare for immediate synchronization with udev and flush all stacked
2446 * dev node operations if requested by immediate_dev_node property. But
2447 * finish processing current level in the tree first.
2449 if (child
->props
.immediate_dev_node
)
2450 update_devs_flag
= 1;
2453 if (update_devs_flag
) {
2454 if (!dm_udev_wait(dm_tree_get_cookie(dnode
)))
2456 dm_tree_set_cookie(dnode
, 0);
2463 * Returns 1 if unsure.
2465 int dm_tree_children_use_uuid(struct dm_tree_node
*dnode
,
2466 const char *uuid_prefix
,
2467 size_t uuid_prefix_len
)
2469 void *handle
= NULL
;
2470 struct dm_tree_node
*child
= dnode
;
2473 while ((child
= dm_tree_next_child(&handle
, dnode
, 0))) {
2474 if (!(uuid
= dm_tree_node_get_uuid(child
))) {
2475 log_error("Failed to get uuid for dtree node.");
2479 if (_uuid_prefix_matches(uuid
, uuid_prefix
, uuid_prefix_len
))
2482 if (dm_tree_node_num_children(child
, 0))
2483 dm_tree_children_use_uuid(child
, uuid_prefix
, uuid_prefix_len
);
2492 static struct load_segment
*_add_segment(struct dm_tree_node
*dnode
, unsigned type
, uint64_t size
)
2494 struct load_segment
*seg
;
2496 if (!(seg
= dm_pool_zalloc(dnode
->dtree
->mem
, sizeof(*seg
)))) {
2497 log_error("dtree node segment allocation failed");
2503 seg
->area_count
= 0;
2504 dm_list_init(&seg
->areas
);
2505 seg
->stripe_size
= 0;
2506 seg
->persistent
= 0;
2507 seg
->chunk_size
= 0;
2512 dm_list_add(&dnode
->props
.segs
, &seg
->list
);
2513 dnode
->props
.segment_count
++;
2518 int dm_tree_node_add_snapshot_origin_target(struct dm_tree_node
*dnode
,
2520 const char *origin_uuid
)
2522 struct load_segment
*seg
;
2523 struct dm_tree_node
*origin_node
;
2525 if (!(seg
= _add_segment(dnode
, SEG_SNAPSHOT_ORIGIN
, size
)))
2528 if (!(origin_node
= dm_tree_find_node_by_uuid(dnode
->dtree
, origin_uuid
))) {
2529 log_error("Couldn't find snapshot origin uuid %s.", origin_uuid
);
2533 seg
->origin
= origin_node
;
2534 if (!_link_tree_nodes(dnode
, origin_node
))
2537 /* Resume snapshot origins after new snapshots */
2538 dnode
->activation_priority
= 1;
2543 static int _add_snapshot_target(struct dm_tree_node
*node
,
2545 const char *origin_uuid
,
2546 const char *cow_uuid
,
2547 const char *merge_uuid
,
2549 uint32_t chunk_size
)
2551 struct load_segment
*seg
;
2552 struct dm_tree_node
*origin_node
, *cow_node
, *merge_node
;
2555 seg_type
= !merge_uuid
? SEG_SNAPSHOT
: SEG_SNAPSHOT_MERGE
;
2557 if (!(seg
= _add_segment(node
, seg_type
, size
)))
2560 if (!(origin_node
= dm_tree_find_node_by_uuid(node
->dtree
, origin_uuid
))) {
2561 log_error("Couldn't find snapshot origin uuid %s.", origin_uuid
);
2565 seg
->origin
= origin_node
;
2566 if (!_link_tree_nodes(node
, origin_node
))
2569 if (!(cow_node
= dm_tree_find_node_by_uuid(node
->dtree
, cow_uuid
))) {
2570 log_error("Couldn't find snapshot COW device uuid %s.", cow_uuid
);
2574 seg
->cow
= cow_node
;
2575 if (!_link_tree_nodes(node
, cow_node
))
2578 seg
->persistent
= persistent
? 1 : 0;
2579 seg
->chunk_size
= chunk_size
;
2582 if (!(merge_node
= dm_tree_find_node_by_uuid(node
->dtree
, merge_uuid
))) {
2583 /* not a pure error, merging snapshot may have been deactivated */
2584 log_verbose("Couldn't find merging snapshot uuid %s.", merge_uuid
);
2586 seg
->merge
= merge_node
;
2587 /* must not link merging snapshot, would undermine activation_priority below */
2590 /* Resume snapshot-merge (acting origin) after other snapshots */
2591 node
->activation_priority
= 1;
2593 /* Resume merging snapshot after snapshot-merge */
2594 seg
->merge
->activation_priority
= 2;
2602 int dm_tree_node_add_snapshot_target(struct dm_tree_node
*node
,
2604 const char *origin_uuid
,
2605 const char *cow_uuid
,
2607 uint32_t chunk_size
)
2609 return _add_snapshot_target(node
, size
, origin_uuid
, cow_uuid
,
2610 NULL
, persistent
, chunk_size
);
2613 int dm_tree_node_add_snapshot_merge_target(struct dm_tree_node
*node
,
2615 const char *origin_uuid
,
2616 const char *cow_uuid
,
2617 const char *merge_uuid
,
2618 uint32_t chunk_size
)
2620 return _add_snapshot_target(node
, size
, origin_uuid
, cow_uuid
,
2621 merge_uuid
, 1, chunk_size
);
2624 int dm_tree_node_add_error_target(struct dm_tree_node
*node
,
2627 if (!_add_segment(node
, SEG_ERROR
, size
))
2633 int dm_tree_node_add_zero_target(struct dm_tree_node
*node
,
2636 if (!_add_segment(node
, SEG_ZERO
, size
))
2642 int dm_tree_node_add_linear_target(struct dm_tree_node
*node
,
2645 if (!_add_segment(node
, SEG_LINEAR
, size
))
2651 int dm_tree_node_add_striped_target(struct dm_tree_node
*node
,
2653 uint32_t stripe_size
)
2655 struct load_segment
*seg
;
2657 if (!(seg
= _add_segment(node
, SEG_STRIPED
, size
)))
2660 seg
->stripe_size
= stripe_size
;
2665 int dm_tree_node_add_crypt_target(struct dm_tree_node
*node
,
2668 const char *chainmode
,
2673 struct load_segment
*seg
;
2675 if (!(seg
= _add_segment(node
, SEG_CRYPT
, size
)))
2678 seg
->cipher
= cipher
;
2679 seg
->chainmode
= chainmode
;
2681 seg
->iv_offset
= iv_offset
;
2687 int dm_tree_node_add_mirror_target_log(struct dm_tree_node
*node
,
2688 uint32_t region_size
,
2690 const char *log_uuid
,
2691 unsigned area_count
,
2694 struct dm_tree_node
*log_node
= NULL
;
2695 struct load_segment
*seg
;
2697 if (!node
->props
.segment_count
) {
2698 log_error(INTERNAL_ERROR
"Attempt to add target area to missing segment.");
2702 seg
= dm_list_item(dm_list_last(&node
->props
.segs
), struct load_segment
);
2705 if (!(seg
->uuid
= dm_pool_strdup(node
->dtree
->mem
, log_uuid
))) {
2706 log_error("log uuid pool_strdup failed");
2709 if ((flags
& DM_CORELOG
))
2710 /* For pvmove: immediate resume (for size validation) isn't needed. */
2711 node
->props
.delay_resume_if_new
= 1;
2713 if (!(log_node
= dm_tree_find_node_by_uuid(node
->dtree
, log_uuid
))) {
2714 log_error("Couldn't find mirror log uuid %s.", log_uuid
);
2719 log_node
->props
.immediate_dev_node
= 1;
2721 /* The kernel validates the size of disk logs. */
2722 /* FIXME Propagate to any devices below */
2723 log_node
->props
.delay_resume_if_new
= 0;
2725 if (!_link_tree_nodes(node
, log_node
))
2730 seg
->log
= log_node
;
2731 seg
->region_size
= region_size
;
2732 seg
->clustered
= clustered
;
2733 seg
->mirror_area_count
= area_count
;
2739 int dm_tree_node_add_mirror_target(struct dm_tree_node
*node
,
2742 if (!_add_segment(node
, SEG_MIRRORED
, size
))
2748 int dm_tree_node_add_raid_target(struct dm_tree_node
*node
,
2750 const char *raid_type
,
2751 uint32_t region_size
,
2752 uint32_t stripe_size
,
2757 struct load_segment
*seg
= NULL
;
2759 for (i
= 0; dm_segtypes
[i
].target
&& !seg
; i
++)
2760 if (!strcmp(raid_type
, dm_segtypes
[i
].target
))
2761 if (!(seg
= _add_segment(node
,
2762 dm_segtypes
[i
].type
, size
)))
2768 seg
->region_size
= region_size
;
2769 seg
->stripe_size
= stripe_size
;
2770 seg
->area_count
= 0;
2771 seg
->rebuilds
= rebuilds
;
2776 int dm_tree_node_add_replicator_target(struct dm_tree_node
*node
,
2778 const char *rlog_uuid
,
2779 const char *rlog_type
,
2780 unsigned rsite_index
,
2781 dm_replicator_mode_t mode
,
2782 uint32_t async_timeout
,
2783 uint64_t fall_behind_data
,
2784 uint32_t fall_behind_ios
)
2786 struct load_segment
*rseg
;
2787 struct replicator_site
*rsite
;
2789 /* Local site0 - adds replicator segment and links rlog device */
2790 if (rsite_index
== REPLICATOR_LOCAL_SITE
) {
2791 if (node
->props
.segment_count
) {
2792 log_error(INTERNAL_ERROR
"Attempt to add replicator segment to already used node.");
2796 if (!(rseg
= _add_segment(node
, SEG_REPLICATOR
, size
)))
2799 if (!(rseg
->log
= dm_tree_find_node_by_uuid(node
->dtree
, rlog_uuid
))) {
2800 log_error("Missing replicator log uuid %s.", rlog_uuid
);
2804 if (!_link_tree_nodes(node
, rseg
->log
))
2807 if (strcmp(rlog_type
, "ringbuffer") != 0) {
2808 log_error("Unsupported replicator log type %s.", rlog_type
);
2812 if (!(rseg
->rlog_type
= dm_pool_strdup(node
->dtree
->mem
, rlog_type
)))
2815 dm_list_init(&rseg
->rsites
);
2816 rseg
->rdevice_count
= 0;
2817 node
->activation_priority
= 1;
2820 /* Add site to segment */
2821 if (mode
== DM_REPLICATOR_SYNC
2822 && (async_timeout
|| fall_behind_ios
|| fall_behind_data
)) {
2823 log_error("Async parameters passed for synchronnous replicator.");
2827 if (node
->props
.segment_count
!= 1) {
2828 log_error(INTERNAL_ERROR
"Attempt to add remote site area before setting replicator log.");
2832 rseg
= dm_list_item(dm_list_last(&node
->props
.segs
), struct load_segment
);
2833 if (rseg
->type
!= SEG_REPLICATOR
) {
2834 log_error(INTERNAL_ERROR
"Attempt to use non replicator segment %s.",
2835 dm_segtypes
[rseg
->type
].target
);
2839 if (!(rsite
= dm_pool_zalloc(node
->dtree
->mem
, sizeof(*rsite
)))) {
2840 log_error("Failed to allocate remote site segment.");
2844 dm_list_add(&rseg
->rsites
, &rsite
->list
);
2845 rseg
->rsite_count
++;
2848 rsite
->async_timeout
= async_timeout
;
2849 rsite
->fall_behind_data
= fall_behind_data
;
2850 rsite
->fall_behind_ios
= fall_behind_ios
;
2851 rsite
->rsite_index
= rsite_index
;
2856 /* Appends device node to Replicator */
2857 int dm_tree_node_add_replicator_dev_target(struct dm_tree_node
*node
,
2859 const char *replicator_uuid
,
2860 uint64_t rdevice_index
,
2861 const char *rdev_uuid
,
2862 unsigned rsite_index
,
2863 const char *slog_uuid
,
2864 uint32_t slog_flags
,
2865 uint32_t slog_region_size
)
2867 struct seg_area
*area
;
2868 struct load_segment
*rseg
;
2869 struct load_segment
*rep_seg
;
2871 if (rsite_index
== REPLICATOR_LOCAL_SITE
) {
2872 /* Site index for local target */
2873 if (!(rseg
= _add_segment(node
, SEG_REPLICATOR_DEV
, size
)))
2876 if (!(rseg
->replicator
= dm_tree_find_node_by_uuid(node
->dtree
, replicator_uuid
))) {
2877 log_error("Missing replicator uuid %s.", replicator_uuid
);
2881 /* Local slink0 for replicator must be always initialized first */
2882 if (rseg
->replicator
->props
.segment_count
!= 1) {
2883 log_error(INTERNAL_ERROR
"Attempt to use non replicator segment.");
2887 rep_seg
= dm_list_item(dm_list_last(&rseg
->replicator
->props
.segs
), struct load_segment
);
2888 if (rep_seg
->type
!= SEG_REPLICATOR
) {
2889 log_error(INTERNAL_ERROR
"Attempt to use non replicator segment %s.",
2890 dm_segtypes
[rep_seg
->type
].target
);
2893 rep_seg
->rdevice_count
++;
2895 if (!_link_tree_nodes(node
, rseg
->replicator
))
2898 rseg
->rdevice_index
= rdevice_index
;
2900 /* Local slink0 for replicator must be always initialized first */
2901 if (node
->props
.segment_count
!= 1) {
2902 log_error(INTERNAL_ERROR
"Attempt to use non replicator-dev segment.");
2906 rseg
= dm_list_item(dm_list_last(&node
->props
.segs
), struct load_segment
);
2907 if (rseg
->type
!= SEG_REPLICATOR_DEV
) {
2908 log_error(INTERNAL_ERROR
"Attempt to use non replicator-dev segment %s.",
2909 dm_segtypes
[rseg
->type
].target
);
2914 if (!(slog_flags
& DM_CORELOG
) && !slog_uuid
) {
2915 log_error("Unspecified sync log uuid.");
2919 if (!dm_tree_node_add_target_area(node
, NULL
, rdev_uuid
, 0))
2922 area
= dm_list_item(dm_list_last(&rseg
->areas
), struct seg_area
);
2924 if (!(slog_flags
& DM_CORELOG
)) {
2925 if (!(area
->slog
= dm_tree_find_node_by_uuid(node
->dtree
, slog_uuid
))) {
2926 log_error("Couldn't find sync log uuid %s.", slog_uuid
);
2930 if (!_link_tree_nodes(node
, area
->slog
))
2934 area
->flags
= slog_flags
;
2935 area
->region_size
= slog_region_size
;
2936 area
->rsite_index
= rsite_index
;
2941 static int _thin_validate_device_id(uint32_t device_id
)
2943 if (device_id
> DM_THIN_MAX_DEVICE_ID
) {
2944 log_error("Device id %u is higher then %u.",
2945 device_id
, DM_THIN_MAX_DEVICE_ID
);
2952 int dm_tree_node_add_thin_pool_target(struct dm_tree_node
*node
,
2954 uint64_t transaction_id
,
2955 const char *metadata_uuid
,
2956 const char *pool_uuid
,
2957 uint32_t data_block_size
,
2958 uint64_t low_water_mark
,
2959 unsigned skip_block_zeroing
)
2961 struct load_segment
*seg
;
2963 if (data_block_size
< DM_THIN_MIN_DATA_BLOCK_SIZE
) {
2964 log_error("Data block size %u is lower then %u sectors.",
2965 data_block_size
, DM_THIN_MIN_DATA_BLOCK_SIZE
);
2969 if (data_block_size
> DM_THIN_MAX_DATA_BLOCK_SIZE
) {
2970 log_error("Data block size %u is higher then %u sectors.",
2971 data_block_size
, DM_THIN_MAX_DATA_BLOCK_SIZE
);
2975 if (!(seg
= _add_segment(node
, SEG_THIN_POOL
, size
)))
2978 if (!(seg
->metadata
= dm_tree_find_node_by_uuid(node
->dtree
, metadata_uuid
))) {
2979 log_error("Missing metadata uuid %s.", metadata_uuid
);
2983 if (!_link_tree_nodes(node
, seg
->metadata
))
2986 if (!(seg
->pool
= dm_tree_find_node_by_uuid(node
->dtree
, pool_uuid
))) {
2987 log_error("Missing pool uuid %s.", pool_uuid
);
2991 if (!_link_tree_nodes(node
, seg
->pool
))
2994 node
->props
.send_messages
= 1;
2995 seg
->transaction_id
= transaction_id
;
2996 seg
->low_water_mark
= low_water_mark
;
2997 seg
->data_block_size
= data_block_size
;
2998 seg
->skip_block_zeroing
= skip_block_zeroing
;
2999 dm_list_init(&seg
->thin_messages
);
3004 int dm_tree_node_add_thin_pool_message(struct dm_tree_node
*node
,
3005 dm_thin_message_t type
,
3006 uint64_t id1
, uint64_t id2
)
3008 struct load_segment
*seg
;
3009 struct thin_message
*tm
;
3011 if (node
->props
.segment_count
!= 1) {
3012 log_error("Thin pool node must have only one segment.");
3016 seg
= dm_list_item(dm_list_last(&node
->props
.segs
), struct load_segment
);
3017 if (seg
->type
!= SEG_THIN_POOL
) {
3018 log_error("Thin pool node has segment type %s.",
3019 dm_segtypes
[seg
->type
].target
);
3023 if (!(tm
= dm_pool_zalloc(node
->dtree
->mem
, sizeof (*tm
)))) {
3024 log_error("Failed to allocate thin message.");
3029 case DM_THIN_MESSAGE_CREATE_SNAP
:
3030 /* If the thin origin is active, it must be suspend first! */
3032 log_error("Cannot use same device id for origin and its snapshot.");
3035 if (!_thin_validate_device_id(id1
) ||
3036 !_thin_validate_device_id(id2
))
3038 tm
->message
.u
.m_create_snap
.device_id
= id1
;
3039 tm
->message
.u
.m_create_snap
.origin_id
= id2
;
3041 case DM_THIN_MESSAGE_CREATE_THIN
:
3042 if (!_thin_validate_device_id(id1
))
3044 tm
->message
.u
.m_create_thin
.device_id
= id1
;
3045 tm
->expected_errno
= EEXIST
;
3047 case DM_THIN_MESSAGE_DELETE
:
3048 if (!_thin_validate_device_id(id1
))
3050 tm
->message
.u
.m_delete
.device_id
= id1
;
3051 tm
->expected_errno
= ENODATA
;
3053 case DM_THIN_MESSAGE_TRIM
:
3054 if (!_thin_validate_device_id(id1
))
3056 tm
->message
.u
.m_trim
.device_id
= id1
;
3057 tm
->message
.u
.m_trim
.new_size
= id2
;
3059 case DM_THIN_MESSAGE_SET_TRANSACTION_ID
:
3060 if ((id1
+ 1) != id2
) {
3061 log_error("New transaction id must be sequential.");
3062 return 0; /* FIXME: Maybe too strict here? */
3064 if (id2
!= seg
->transaction_id
) {
3065 log_error("Current transaction id is different from thin pool.");
3066 return 0; /* FIXME: Maybe too strict here? */
3068 tm
->message
.u
.m_set_transaction_id
.current_id
= id1
;
3069 tm
->message
.u
.m_set_transaction_id
.new_id
= id2
;
3072 log_error("Unsupported message type %d.", (int) type
);
3076 tm
->message
.type
= type
;
3077 dm_list_add(&seg
->thin_messages
, &tm
->list
);
3082 int dm_tree_node_add_thin_target(struct dm_tree_node
*node
,
3084 const char *pool_uuid
,
3087 struct dm_tree_node
*pool
;
3088 struct load_segment
*seg
;
3090 if (!(pool
= dm_tree_find_node_by_uuid(node
->dtree
, pool_uuid
))) {
3091 log_error("Missing thin pool uuid %s.", pool_uuid
);
3095 if (!_link_tree_nodes(node
, pool
))
3098 if (!_thin_validate_device_id(device_id
))
3101 if (!(seg
= _add_segment(node
, SEG_THIN
, size
)))
3105 seg
->device_id
= device_id
;
3111 int dm_get_status_thin_pool(struct dm_pool
*mem
, const char *params
,
3112 struct dm_status_thin_pool
**status
)
3114 struct dm_status_thin_pool
*s
;
3116 if (!(s
= dm_pool_zalloc(mem
, sizeof(struct dm_status_thin_pool
)))) {
3117 log_error("Failed to allocate thin_pool status structure.");
3121 /* FIXME: add support for held metadata root */
3122 if (sscanf(params
, "%" PRIu64
" %" PRIu64
"/%" PRIu64
" %" PRIu64
"/%" PRIu64
,
3124 &s
->used_metadata_blocks
,
3125 &s
->total_metadata_blocks
,
3126 &s
->used_data_blocks
,
3127 &s
->total_data_blocks
) != 5) {
3128 log_error("Failed to parse thin pool params: %s.", params
);
3137 int dm_get_status_thin(struct dm_pool
*mem
, const char *params
,
3138 struct dm_status_thin
**status
)
3140 struct dm_status_thin
*s
;
3142 if (!(s
= dm_pool_zalloc(mem
, sizeof(struct dm_status_thin
)))) {
3143 log_error("Failed to allocate thin status structure.");
3147 if (strchr(params
, '-')) {
3148 s
->mapped_sectors
= 0;
3149 s
->highest_mapped_sector
= 0;
3150 } else if (sscanf(params
, "%" PRIu64
" %" PRIu64
,
3152 &s
->highest_mapped_sector
) != 2) {
3153 log_error("Failed to parse thin params: %s.", params
);
3162 static int _add_area(struct dm_tree_node
*node
, struct load_segment
*seg
, struct dm_tree_node
*dev_node
, uint64_t offset
)
3164 struct seg_area
*area
;
3166 if (!(area
= dm_pool_zalloc(node
->dtree
->mem
, sizeof (*area
)))) {
3167 log_error("Failed to allocate target segment area.");
3171 area
->dev_node
= dev_node
;
3172 area
->offset
= offset
;
3174 dm_list_add(&seg
->areas
, &area
->list
);
3180 int dm_tree_node_add_target_area(struct dm_tree_node
*node
,
3181 const char *dev_name
,
3185 struct load_segment
*seg
;
3187 struct dm_tree_node
*dev_node
;
3189 if ((!dev_name
|| !*dev_name
) && (!uuid
|| !*uuid
)) {
3190 log_error("dm_tree_node_add_target_area called without device");
3195 if (!(dev_node
= dm_tree_find_node_by_uuid(node
->dtree
, uuid
))) {
3196 log_error("Couldn't find area uuid %s.", uuid
);
3199 if (!_link_tree_nodes(node
, dev_node
))
3202 if (stat(dev_name
, &info
) < 0) {
3203 log_error("Device %s not found.", dev_name
);
3207 if (!S_ISBLK(info
.st_mode
)) {
3208 log_error("Device %s is not a block device.", dev_name
);
3212 /* FIXME Check correct macro use */
3213 if (!(dev_node
= _add_dev(node
->dtree
, node
, MAJOR(info
.st_rdev
),
3214 MINOR(info
.st_rdev
), 0)))
3218 if (!node
->props
.segment_count
) {
3219 log_error(INTERNAL_ERROR
"Attempt to add target area to missing segment.");
3223 seg
= dm_list_item(dm_list_last(&node
->props
.segs
), struct load_segment
);
3225 if (!_add_area(node
, seg
, dev_node
, offset
))
3231 int dm_tree_node_add_null_area(struct dm_tree_node
*node
, uint64_t offset
)
3233 struct load_segment
*seg
;
3235 seg
= dm_list_item(dm_list_last(&node
->props
.segs
), struct load_segment
);
3237 switch (seg
->type
) {
3249 log_error("dm_tree_node_add_null_area() called on an unsupported segment type");
3253 if (!_add_area(node
, seg
, NULL
, offset
))