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 /* FIXME Fix interface so this is used only by LVM */
28 #define UUID_PREFIX "LVM-"
30 #define REPLICATOR_LOCAL_SITE 0
32 /* Supported segment types */
59 /* FIXME Add crypt and multipath support */
65 { SEG_CRYPT
, "crypt" },
66 { SEG_ERROR
, "error" },
67 { SEG_LINEAR
, "linear" },
68 { SEG_MIRRORED
, "mirror" },
69 { SEG_REPLICATOR
, "replicator" },
70 { SEG_REPLICATOR_DEV
, "replicator-dev" },
71 { SEG_SNAPSHOT
, "snapshot" },
72 { SEG_SNAPSHOT_ORIGIN
, "snapshot-origin" },
73 { SEG_SNAPSHOT_MERGE
, "snapshot-merge" },
74 { SEG_STRIPED
, "striped" },
76 { SEG_THIN_POOL
, "thin-pool"},
78 { SEG_RAID1
, "raid1"},
79 { SEG_RAID4
, "raid4"},
80 { SEG_RAID5_LA
, "raid5_la"},
81 { SEG_RAID5_RA
, "raid5_ra"},
82 { SEG_RAID5_LS
, "raid5_ls"},
83 { SEG_RAID5_RS
, "raid5_rs"},
84 { SEG_RAID6_ZR
, "raid6_zr"},
85 { SEG_RAID6_NR
, "raid6_nr"},
86 { SEG_RAID6_NC
, "raid6_nc"},
89 *WARNING: Since 'raid' target overloads this 1:1 mapping table
90 * for search do not add new enum elements past them!
92 { SEG_RAID5_LS
, "raid5"}, /* same as "raid5_ls" (default for MD also) */
93 { SEG_RAID6_ZR
, "raid6"}, /* same as "raid6_zr" */
97 /* Some segment types have a list of areas of other devices attached */
101 struct dm_tree_node
*dev_node
;
105 unsigned rsite_index
; /* Replicator site index */
106 struct dm_tree_node
*slog
; /* Replicator sync log node */
107 uint64_t region_size
; /* Replicator sync log size */
108 uint32_t flags
; /* Replicator sync log flags */
111 struct dm_thin_message
{
112 dm_thin_message_t type
;
127 } m_set_transaction_id
;
135 struct thin_message
{
137 struct dm_thin_message message
;
141 /* Replicator-log has a list of sites */
142 /* FIXME: maybe move to seg_area too? */
143 struct replicator_site
{
146 unsigned rsite_index
;
147 dm_replicator_mode_t mode
;
148 uint32_t async_timeout
;
149 uint32_t fall_behind_ios
;
150 uint64_t fall_behind_data
;
153 /* Per-segment properties */
154 struct load_segment
{
161 unsigned area_count
; /* Linear + Striped + Mirrored + Crypt + Replicator */
162 struct dm_list areas
; /* Linear + Striped + Mirrored + Crypt + Replicator */
164 uint32_t stripe_size
; /* Striped + raid */
166 int persistent
; /* Snapshot */
167 uint32_t chunk_size
; /* Snapshot */
168 struct dm_tree_node
*cow
; /* Snapshot */
169 struct dm_tree_node
*origin
; /* Snapshot + Snapshot origin */
170 struct dm_tree_node
*merge
; /* Snapshot */
172 struct dm_tree_node
*log
; /* Mirror + Replicator */
173 uint32_t region_size
; /* Mirror + raid */
174 unsigned clustered
; /* Mirror */
175 unsigned mirror_area_count
; /* Mirror */
176 uint32_t flags
; /* Mirror log */
177 char *uuid
; /* Clustered mirror log */
179 const char *cipher
; /* Crypt */
180 const char *chainmode
; /* Crypt */
181 const char *iv
; /* Crypt */
182 uint64_t iv_offset
; /* Crypt */
183 const char *key
; /* Crypt */
185 const char *rlog_type
; /* Replicator */
186 struct dm_list rsites
; /* Replicator */
187 unsigned rsite_count
; /* Replicator */
188 unsigned rdevice_count
; /* Replicator */
189 struct dm_tree_node
*replicator
;/* Replicator-dev */
190 uint64_t rdevice_index
; /* Replicator-dev */
192 uint64_t rebuilds
; /* raid */
194 struct dm_tree_node
*metadata
; /* Thin_pool */
195 struct dm_tree_node
*pool
; /* Thin_pool, Thin */
196 struct dm_list thin_messages
; /* Thin_pool */
197 uint64_t transaction_id
; /* Thin_pool */
198 uint64_t low_water_mark
; /* Thin_pool */
199 uint32_t data_block_size
; /* Thin_pool */
200 unsigned skip_block_zeroing
; /* Thin_pool */
201 uint32_t device_id
; /* Thin */
205 /* Per-device properties */
206 struct load_properties
{
212 uint32_t read_ahead_flags
;
214 unsigned segment_count
;
215 unsigned size_changed
;
218 const char *new_name
;
220 /* If immediate_dev_node is set to 1, try to create the dev node
221 * as soon as possible (e.g. in preload stage even during traversal
222 * and processing of dm tree). This will also flush all stacked dev
223 * node operations, synchronizing with udev.
225 unsigned immediate_dev_node
;
228 * If the device size changed from zero and this is set,
229 * don't resume the device immediately, even if the device
230 * has parents. This works provided the parents do not
231 * validate the device size and is required by pvmove to
232 * avoid starting the mirror resync operation too early.
234 unsigned delay_resume_if_new
;
236 /* Send messages for this node in preload */
237 unsigned send_messages
;
240 /* Two of these used to join two nodes with uses and used_by. */
241 struct dm_tree_link
{
243 struct dm_tree_node
*node
;
246 struct dm_tree_node
{
247 struct dm_tree
*dtree
;
253 struct dm_list uses
; /* Nodes this node uses */
254 struct dm_list used_by
; /* Nodes that use this node */
256 int activation_priority
; /* 0 gets activated first */
258 uint16_t udev_flags
; /* Udev control flags */
260 void *context
; /* External supplied context */
262 struct load_properties props
; /* For creation/table (re)load */
265 * If presuspend of child node is needed
266 * Note: only direct child is allowed
268 struct dm_tree_node
*presuspend_node
;
273 struct dm_hash_table
*devs
;
274 struct dm_hash_table
*uuids
;
275 struct dm_tree_node root
;
276 int skip_lockfs
; /* 1 skips lockfs (for non-snapshots) */
277 int no_flush
; /* 1 sets noflush (mirrors/multipath) */
278 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 static int _nodes_are_linked(const struct dm_tree_node
*parent
,
329 const struct dm_tree_node
*child
)
331 struct dm_tree_link
*dlink
;
333 dm_list_iterate_items(dlink
, &parent
->uses
)
334 if (dlink
->node
== child
)
340 static int _link(struct dm_list
*list
, struct dm_tree_node
*node
)
342 struct dm_tree_link
*dlink
;
344 if (!(dlink
= dm_pool_alloc(node
->dtree
->mem
, sizeof(*dlink
)))) {
345 log_error("dtree link allocation failed");
350 dm_list_add(list
, &dlink
->list
);
355 static int _link_nodes(struct dm_tree_node
*parent
,
356 struct dm_tree_node
*child
)
358 if (_nodes_are_linked(parent
, child
))
361 if (!_link(&parent
->uses
, child
))
364 if (!_link(&child
->used_by
, parent
))
370 static void _unlink(struct dm_list
*list
, struct dm_tree_node
*node
)
372 struct dm_tree_link
*dlink
;
374 dm_list_iterate_items(dlink
, list
)
375 if (dlink
->node
== node
) {
376 dm_list_del(&dlink
->list
);
381 static void _unlink_nodes(struct dm_tree_node
*parent
,
382 struct dm_tree_node
*child
)
384 if (!_nodes_are_linked(parent
, child
))
387 _unlink(&parent
->uses
, child
);
388 _unlink(&child
->used_by
, parent
);
391 static int _add_to_toplevel(struct dm_tree_node
*node
)
393 return _link_nodes(&node
->dtree
->root
, node
);
396 static void _remove_from_toplevel(struct dm_tree_node
*node
)
398 _unlink_nodes(&node
->dtree
->root
, node
);
401 static int _add_to_bottomlevel(struct dm_tree_node
*node
)
403 return _link_nodes(node
, &node
->dtree
->root
);
406 static void _remove_from_bottomlevel(struct dm_tree_node
*node
)
408 _unlink_nodes(node
, &node
->dtree
->root
);
411 static int _link_tree_nodes(struct dm_tree_node
*parent
, struct dm_tree_node
*child
)
413 /* Don't link to root node if child already has a parent */
414 if (parent
== &parent
->dtree
->root
) {
415 if (dm_tree_node_num_children(child
, 1))
418 _remove_from_toplevel(child
);
420 if (child
== &child
->dtree
->root
) {
421 if (dm_tree_node_num_children(parent
, 0))
424 _remove_from_bottomlevel(parent
);
426 return _link_nodes(parent
, child
);
429 static struct dm_tree_node
*_create_dm_tree_node(struct dm_tree
*dtree
,
432 struct dm_info
*info
,
436 struct dm_tree_node
*node
;
439 if (!(node
= dm_pool_zalloc(dtree
->mem
, sizeof(*node
)))) {
440 log_error("_create_dm_tree_node alloc failed");
449 node
->context
= context
;
450 node
->udev_flags
= udev_flags
;
451 node
->activation_priority
= 0;
453 dm_list_init(&node
->uses
);
454 dm_list_init(&node
->used_by
);
455 dm_list_init(&node
->props
.segs
);
457 dev
= MKDEV(info
->major
, info
->minor
);
459 if (!dm_hash_insert_binary(dtree
->devs
, (const char *) &dev
,
460 sizeof(dev
), node
)) {
461 log_error("dtree node hash insertion failed");
462 dm_pool_free(dtree
->mem
, node
);
467 !dm_hash_insert(dtree
->uuids
, uuid
, node
)) {
468 log_error("dtree uuid hash insertion failed");
469 dm_hash_remove_binary(dtree
->devs
, (const char *) &dev
,
471 dm_pool_free(dtree
->mem
, node
);
478 static struct dm_tree_node
*_find_dm_tree_node(struct dm_tree
*dtree
,
479 uint32_t major
, uint32_t minor
)
481 uint64_t dev
= MKDEV(major
, minor
);
483 return dm_hash_lookup_binary(dtree
->devs
, (const char *) &dev
,
487 static struct dm_tree_node
*_find_dm_tree_node_by_uuid(struct dm_tree
*dtree
,
490 struct dm_tree_node
*node
;
492 if ((node
= dm_hash_lookup(dtree
->uuids
, uuid
)))
495 if (strncmp(uuid
, UUID_PREFIX
, sizeof(UUID_PREFIX
) - 1))
498 return dm_hash_lookup(dtree
->uuids
, uuid
+ sizeof(UUID_PREFIX
) - 1);
501 static int _deps(struct dm_task
**dmt
, struct dm_pool
*mem
, uint32_t major
, uint32_t minor
,
502 const char **name
, const char **uuid
,
503 struct dm_info
*info
, struct dm_deps
**deps
)
505 memset(info
, 0, sizeof(*info
));
507 if (!dm_is_dm_major(major
)) {
514 info
->live_table
= 0;
515 info
->inactive_table
= 0;
520 if (!(*dmt
= dm_task_create(DM_DEVICE_DEPS
))) {
521 log_error("deps dm_task creation failed");
525 if (!dm_task_set_major(*dmt
, major
)) {
526 log_error("_deps: failed to set major for (%" PRIu32
":%" PRIu32
")",
531 if (!dm_task_set_minor(*dmt
, minor
)) {
532 log_error("_deps: failed to set minor for (%" PRIu32
":%" PRIu32
")",
537 if (!dm_task_run(*dmt
)) {
538 log_error("_deps: task run failed for (%" PRIu32
":%" PRIu32
")",
543 if (!dm_task_get_info(*dmt
, info
)) {
544 log_error("_deps: failed to get info for (%" PRIu32
":%" PRIu32
")",
554 if (info
->major
!= major
) {
555 log_error("Inconsistent dtree major number: %u != %u",
559 if (info
->minor
!= minor
) {
560 log_error("Inconsistent dtree minor number: %u != %u",
564 if (!(*name
= dm_pool_strdup(mem
, dm_task_get_name(*dmt
)))) {
565 log_error("name pool_strdup failed");
568 if (!(*uuid
= dm_pool_strdup(mem
, dm_task_get_uuid(*dmt
)))) {
569 log_error("uuid pool_strdup failed");
572 *deps
= dm_task_get_deps(*dmt
);
578 dm_task_destroy(*dmt
);
582 static struct dm_tree_node
*_add_dev(struct dm_tree
*dtree
,
583 struct dm_tree_node
*parent
,
584 uint32_t major
, uint32_t minor
,
587 struct dm_task
*dmt
= NULL
;
589 struct dm_deps
*deps
= NULL
;
590 const char *name
= NULL
;
591 const char *uuid
= NULL
;
592 struct dm_tree_node
*node
= NULL
;
596 /* Already in tree? */
597 if (!(node
= _find_dm_tree_node(dtree
, major
, minor
))) {
598 if (!_deps(&dmt
, dtree
->mem
, major
, minor
, &name
, &uuid
, &info
, &deps
))
601 if (!(node
= _create_dm_tree_node(dtree
, name
, uuid
, &info
,
607 if (!_link_tree_nodes(parent
, node
)) {
612 /* If node was already in tree, no need to recurse. */
616 /* Can't recurse if not a mapped device or there are no dependencies */
617 if (!node
->info
.exists
|| !deps
->count
) {
618 if (!_add_to_bottomlevel(node
)) {
625 /* Add dependencies to tree */
626 for (i
= 0; i
< deps
->count
; i
++)
627 if (!_add_dev(dtree
, node
, MAJOR(deps
->device
[i
]),
628 MINOR(deps
->device
[i
]), udev_flags
)) {
635 dm_task_destroy(dmt
);
640 static int _node_clear_table(struct dm_tree_node
*dnode
)
643 struct dm_info
*info
;
647 if (!(info
= &dnode
->info
)) {
648 log_error("_node_clear_table failed: missing info");
652 if (!(name
= dm_tree_node_get_name(dnode
))) {
653 log_error("_node_clear_table failed: missing name");
657 /* Is there a table? */
658 if (!info
->exists
|| !info
->inactive_table
)
661 // FIXME Get inactive deps. If any dev referenced has 1 opener and no live table, remove it after the clear.
663 log_verbose("Clearing inactive table %s (%" PRIu32
":%" PRIu32
")",
664 name
, info
->major
, info
->minor
);
666 if (!(dmt
= dm_task_create(DM_DEVICE_CLEAR
))) {
667 log_error("Table clear dm_task creation failed for %s", name
);
671 if (!dm_task_set_major(dmt
, info
->major
) ||
672 !dm_task_set_minor(dmt
, info
->minor
)) {
673 log_error("Failed to set device number for %s table clear", name
);
674 dm_task_destroy(dmt
);
678 r
= dm_task_run(dmt
);
680 if (!dm_task_get_info(dmt
, info
)) {
681 log_error("_node_clear_table failed: info missing after running task for %s", name
);
685 dm_task_destroy(dmt
);
690 struct dm_tree_node
*dm_tree_add_new_dev(struct dm_tree
*dtree
,
693 uint32_t major
, uint32_t minor
,
698 struct dm_tree_node
*dnode
;
703 /* Do we need to add node to tree? */
704 if (!(dnode
= dm_tree_find_node_by_uuid(dtree
, uuid
))) {
705 if (!(name2
= dm_pool_strdup(dtree
->mem
, name
))) {
706 log_error("name pool_strdup failed");
709 if (!(uuid2
= dm_pool_strdup(dtree
->mem
, uuid
))) {
710 log_error("uuid pool_strdup failed");
718 info
.inactive_table
= 0;
721 if (!(dnode
= _create_dm_tree_node(dtree
, name2
, uuid2
, &info
,
725 /* Attach to root node until a table is supplied */
726 if (!_add_to_toplevel(dnode
) || !_add_to_bottomlevel(dnode
))
729 dnode
->props
.major
= major
;
730 dnode
->props
.minor
= minor
;
731 dnode
->props
.new_name
= NULL
;
732 dnode
->props
.size_changed
= 0;
733 } else if (strcmp(name
, dnode
->name
)) {
734 /* Do we need to rename node? */
735 if (!(dnode
->props
.new_name
= dm_pool_strdup(dtree
->mem
, name
))) {
736 log_error("name pool_strdup failed");
741 dnode
->props
.read_only
= read_only
? 1 : 0;
742 dnode
->props
.read_ahead
= DM_READ_AHEAD_AUTO
;
743 dnode
->props
.read_ahead_flags
= 0;
745 if (clear_inactive
&& !_node_clear_table(dnode
))
748 dnode
->context
= context
;
749 dnode
->udev_flags
= 0;
754 struct dm_tree_node
*dm_tree_add_new_dev_with_udev_flags(struct dm_tree
*dtree
,
764 struct dm_tree_node
*node
;
766 if ((node
= dm_tree_add_new_dev(dtree
, name
, uuid
, major
, minor
, read_only
,
767 clear_inactive
, context
)))
768 node
->udev_flags
= udev_flags
;
773 void dm_tree_node_set_udev_flags(struct dm_tree_node
*dnode
, uint16_t udev_flags
)
776 struct dm_info
*dinfo
= &dnode
->info
;
778 if (udev_flags
!= dnode
->udev_flags
)
779 log_debug("Resetting %s (%" PRIu32
":%" PRIu32
780 ") udev_flags from 0x%x to 0x%x",
781 dnode
->name
, dinfo
->major
, dinfo
->minor
,
782 dnode
->udev_flags
, udev_flags
);
783 dnode
->udev_flags
= udev_flags
;
786 void dm_tree_node_set_read_ahead(struct dm_tree_node
*dnode
,
788 uint32_t read_ahead_flags
)
790 dnode
->props
.read_ahead
= read_ahead
;
791 dnode
->props
.read_ahead_flags
= read_ahead_flags
;
794 void dm_tree_node_set_presuspend_node(struct dm_tree_node
*node
,
795 struct dm_tree_node
*presuspend_node
)
797 node
->presuspend_node
= presuspend_node
;
800 int dm_tree_add_dev(struct dm_tree
*dtree
, uint32_t major
, uint32_t minor
)
802 return _add_dev(dtree
, &dtree
->root
, major
, minor
, 0) ? 1 : 0;
805 int dm_tree_add_dev_with_udev_flags(struct dm_tree
*dtree
, uint32_t major
,
806 uint32_t minor
, uint16_t udev_flags
)
808 return _add_dev(dtree
, &dtree
->root
, major
, minor
, udev_flags
) ? 1 : 0;
811 const char *dm_tree_node_get_name(const struct dm_tree_node
*node
)
813 return node
->info
.exists
? node
->name
: "";
816 const char *dm_tree_node_get_uuid(const struct dm_tree_node
*node
)
818 return node
->info
.exists
? node
->uuid
: "";
821 const struct dm_info
*dm_tree_node_get_info(const struct dm_tree_node
*node
)
826 void *dm_tree_node_get_context(const struct dm_tree_node
*node
)
828 return node
->context
;
831 int dm_tree_node_size_changed(const struct dm_tree_node
*dnode
)
833 return dnode
->props
.size_changed
;
836 int dm_tree_node_num_children(const struct dm_tree_node
*node
, uint32_t inverted
)
839 if (_nodes_are_linked(&node
->dtree
->root
, node
))
841 return dm_list_size(&node
->used_by
);
844 if (_nodes_are_linked(node
, &node
->dtree
->root
))
847 return dm_list_size(&node
->uses
);
851 * Returns 1 if no prefix supplied
853 static int _uuid_prefix_matches(const char *uuid
, const char *uuid_prefix
, size_t uuid_prefix_len
)
858 if (!strncmp(uuid
, uuid_prefix
, uuid_prefix_len
))
861 /* Handle transition: active device uuids might be missing the prefix */
862 if (uuid_prefix_len
<= 4)
865 if (!strncmp(uuid
, UUID_PREFIX
, sizeof(UUID_PREFIX
) - 1))
868 if (strncmp(uuid_prefix
, UUID_PREFIX
, sizeof(UUID_PREFIX
) - 1))
871 if (!strncmp(uuid
, uuid_prefix
+ sizeof(UUID_PREFIX
) - 1, uuid_prefix_len
- (sizeof(UUID_PREFIX
) - 1)))
878 * Returns 1 if no children.
880 static int _children_suspended(struct dm_tree_node
*node
,
882 const char *uuid_prefix
,
883 size_t uuid_prefix_len
)
885 struct dm_list
*list
;
886 struct dm_tree_link
*dlink
;
887 const struct dm_info
*dinfo
;
891 if (_nodes_are_linked(&node
->dtree
->root
, node
))
893 list
= &node
->used_by
;
895 if (_nodes_are_linked(node
, &node
->dtree
->root
))
900 dm_list_iterate_items(dlink
, list
) {
901 if (!(uuid
= dm_tree_node_get_uuid(dlink
->node
))) {
906 /* Ignore if it doesn't belong to this VG */
907 if (!_uuid_prefix_matches(uuid
, uuid_prefix
, uuid_prefix_len
))
910 /* Ignore if parent node wants to presuspend this node */
911 if (dlink
->node
->presuspend_node
== node
)
914 if (!(dinfo
= dm_tree_node_get_info(dlink
->node
))) {
915 stack
; /* FIXME Is this normal? */
919 if (!dinfo
->suspended
)
927 * Set major and minor to zero for root of tree.
929 struct dm_tree_node
*dm_tree_find_node(struct dm_tree
*dtree
,
933 if (!major
&& !minor
)
936 return _find_dm_tree_node(dtree
, major
, minor
);
940 * Set uuid to NULL for root of tree.
942 struct dm_tree_node
*dm_tree_find_node_by_uuid(struct dm_tree
*dtree
,
948 return _find_dm_tree_node_by_uuid(dtree
, uuid
);
952 * First time set *handle to NULL.
953 * Set inverted to invert the tree.
955 struct dm_tree_node
*dm_tree_next_child(void **handle
,
956 const struct dm_tree_node
*parent
,
959 struct dm_list
**dlink
= (struct dm_list
**) handle
;
960 const struct dm_list
*use_list
;
963 use_list
= &parent
->used_by
;
965 use_list
= &parent
->uses
;
968 *dlink
= dm_list_first(use_list
);
970 *dlink
= dm_list_next(use_list
, *dlink
);
972 return (*dlink
) ? dm_list_item(*dlink
, struct dm_tree_link
)->node
: NULL
;
976 * Deactivate a device with its dependencies if the uuid prefix matches.
978 static int _info_by_dev(uint32_t major
, uint32_t minor
, int with_open_count
,
979 struct dm_info
*info
)
984 if (!(dmt
= dm_task_create(DM_DEVICE_INFO
))) {
985 log_error("_info_by_dev: dm_task creation failed");
989 if (!dm_task_set_major(dmt
, major
) || !dm_task_set_minor(dmt
, minor
)) {
990 log_error("_info_by_dev: Failed to set device number");
991 dm_task_destroy(dmt
);
995 if (!with_open_count
&& !dm_task_no_open_count(dmt
))
996 log_error("Failed to disable open_count");
998 if ((r
= dm_task_run(dmt
)))
999 r
= dm_task_get_info(dmt
, info
);
1001 dm_task_destroy(dmt
);
1006 static int _check_device_not_in_use(const char *name
, struct dm_info
*info
)
1011 /* If sysfs is not used, use open_count information only. */
1012 if (!*dm_sysfs_dir()) {
1013 if (info
->open_count
) {
1014 log_error("Device %s (%" PRIu32
":%" PRIu32
") in use",
1015 name
, info
->major
, info
->minor
);
1022 if (dm_device_has_holders(info
->major
, info
->minor
)) {
1023 log_error("Device %s (%" PRIu32
":%" PRIu32
") is used "
1024 "by another device.", name
, info
->major
, info
->minor
);
1028 if (dm_device_has_mounted_fs(info
->major
, info
->minor
)) {
1029 log_error("Device %s (%" PRIu32
":%" PRIu32
") contains "
1030 "a filesystem in use.", name
, info
->major
, info
->minor
);
1037 /* Check if all parent nodes of given node have open_count == 0 */
1038 static int _node_has_closed_parents(struct dm_tree_node
*node
,
1039 const char *uuid_prefix
,
1040 size_t uuid_prefix_len
)
1042 struct dm_tree_link
*dlink
;
1043 const struct dm_info
*dinfo
;
1044 struct dm_info info
;
1047 /* Iterate through parents of this node */
1048 dm_list_iterate_items(dlink
, &node
->used_by
) {
1049 if (!(uuid
= dm_tree_node_get_uuid(dlink
->node
))) {
1054 /* Ignore if it doesn't belong to this VG */
1055 if (!_uuid_prefix_matches(uuid
, uuid_prefix
, uuid_prefix_len
))
1058 if (!(dinfo
= dm_tree_node_get_info(dlink
->node
))) {
1059 stack
; /* FIXME Is this normal? */
1063 /* Refresh open_count */
1064 if (!_info_by_dev(dinfo
->major
, dinfo
->minor
, 1, &info
) ||
1068 if (info
.open_count
) {
1069 log_debug("Node %s %d:%d has open_count %d", uuid_prefix
,
1070 dinfo
->major
, dinfo
->minor
, info
.open_count
);
1078 static int _deactivate_node(const char *name
, uint32_t major
, uint32_t minor
,
1079 uint32_t *cookie
, uint16_t udev_flags
, int retry
)
1081 struct dm_task
*dmt
;
1084 log_verbose("Removing %s (%" PRIu32
":%" PRIu32
")", name
, major
, minor
);
1086 if (!(dmt
= dm_task_create(DM_DEVICE_REMOVE
))) {
1087 log_error("Deactivation dm_task creation failed for %s", name
);
1091 if (!dm_task_set_major(dmt
, major
) || !dm_task_set_minor(dmt
, minor
)) {
1092 log_error("Failed to set device number for %s deactivation", name
);
1096 if (!dm_task_no_open_count(dmt
))
1097 log_error("Failed to disable open_count");
1099 if (!dm_task_set_cookie(dmt
, cookie
, udev_flags
))
1104 dm_task_retry_remove(dmt
);
1106 r
= dm_task_run(dmt
);
1108 /* FIXME Until kernel returns actual name so dm-iface.c can handle it */
1109 rm_dev_node(name
, dmt
->cookie_set
&& !(udev_flags
& DM_UDEV_DISABLE_DM_RULES_FLAG
),
1110 dmt
->cookie_set
&& (udev_flags
& DM_UDEV_DISABLE_LIBRARY_FALLBACK
));
1112 /* FIXME Remove node from tree or mark invalid? */
1115 dm_task_destroy(dmt
);
1120 static int _rename_node(const char *old_name
, const char *new_name
, uint32_t major
,
1121 uint32_t minor
, uint32_t *cookie
, uint16_t udev_flags
)
1123 struct dm_task
*dmt
;
1126 log_verbose("Renaming %s (%" PRIu32
":%" PRIu32
") to %s", old_name
, major
, minor
, new_name
);
1128 if (!(dmt
= dm_task_create(DM_DEVICE_RENAME
))) {
1129 log_error("Rename dm_task creation failed for %s", old_name
);
1133 if (!dm_task_set_name(dmt
, old_name
)) {
1134 log_error("Failed to set name for %s rename.", old_name
);
1138 if (!dm_task_set_newname(dmt
, new_name
))
1141 if (!dm_task_no_open_count(dmt
))
1142 log_error("Failed to disable open_count");
1144 if (!dm_task_set_cookie(dmt
, cookie
, udev_flags
))
1147 r
= dm_task_run(dmt
);
1150 dm_task_destroy(dmt
);
1155 /* FIXME Merge with _suspend_node? */
1156 static int _resume_node(const char *name
, uint32_t major
, uint32_t minor
,
1157 uint32_t read_ahead
, uint32_t read_ahead_flags
,
1158 struct dm_info
*newinfo
, uint32_t *cookie
,
1159 uint16_t udev_flags
, int already_suspended
)
1161 struct dm_task
*dmt
;
1164 log_verbose("Resuming %s (%" PRIu32
":%" PRIu32
")", name
, major
, minor
);
1166 if (!(dmt
= dm_task_create(DM_DEVICE_RESUME
))) {
1167 log_debug("Suspend dm_task creation failed for %s.", name
);
1171 /* FIXME Kernel should fill in name on return instead */
1172 if (!dm_task_set_name(dmt
, name
)) {
1173 log_debug("Failed to set device name for %s resumption.", name
);
1177 if (!dm_task_set_major(dmt
, major
) || !dm_task_set_minor(dmt
, minor
)) {
1178 log_error("Failed to set device number for %s resumption.", name
);
1182 if (!dm_task_no_open_count(dmt
))
1183 log_error("Failed to disable open_count");
1185 if (!dm_task_set_read_ahead(dmt
, read_ahead
, read_ahead_flags
))
1186 log_error("Failed to set read ahead");
1188 if (!dm_task_set_cookie(dmt
, cookie
, udev_flags
))
1191 if (!(r
= dm_task_run(dmt
)))
1194 if (already_suspended
)
1197 if (!(r
= dm_task_get_info(dmt
, newinfo
)))
1201 dm_task_destroy(dmt
);
1206 static int _suspend_node(const char *name
, uint32_t major
, uint32_t minor
,
1207 int skip_lockfs
, int no_flush
, struct dm_info
*newinfo
)
1209 struct dm_task
*dmt
;
1212 log_verbose("Suspending %s (%" PRIu32
":%" PRIu32
")%s%s",
1214 skip_lockfs
? "" : " with filesystem sync",
1215 no_flush
? "" : " with device flush");
1217 if (!(dmt
= dm_task_create(DM_DEVICE_SUSPEND
))) {
1218 log_error("Suspend dm_task creation failed for %s", name
);
1222 if (!dm_task_set_major(dmt
, major
) || !dm_task_set_minor(dmt
, minor
)) {
1223 log_error("Failed to set device number for %s suspension.", name
);
1224 dm_task_destroy(dmt
);
1228 if (!dm_task_no_open_count(dmt
))
1229 log_error("Failed to disable open_count");
1231 if (skip_lockfs
&& !dm_task_skip_lockfs(dmt
))
1232 log_error("Failed to set skip_lockfs flag.");
1234 if (no_flush
&& !dm_task_no_flush(dmt
))
1235 log_error("Failed to set no_flush flag.");
1237 if ((r
= dm_task_run(dmt
))) {
1239 r
= dm_task_get_info(dmt
, newinfo
);
1242 dm_task_destroy(dmt
);
1247 static int _thin_pool_status_transaction_id(struct dm_tree_node
*dnode
, uint64_t *transaction_id
)
1249 struct dm_task
*dmt
;
1251 uint64_t start
, length
;
1253 char *params
= NULL
;
1255 if (!(dmt
= dm_task_create(DM_DEVICE_STATUS
)))
1258 if (!dm_task_set_major(dmt
, dnode
->info
.major
) ||
1259 !dm_task_set_minor(dmt
, dnode
->info
.minor
)) {
1260 log_error("Failed to set major minor.");
1264 if (!dm_task_run(dmt
))
1267 dm_get_next_target(dmt
, NULL
, &start
, &length
, &type
, ¶ms
);
1269 if (type
&& (strcmp(type
, "thin-pool") != 0)) {
1270 log_error("Expected thin-pool target for %d:%d and got %s.",
1271 dnode
->info
.major
, dnode
->info
.minor
, type
);
1275 if (!params
|| (sscanf(params
, "%" PRIu64
, transaction_id
) != 1)) {
1276 log_error("Failed to parse transaction_id from %s.", params
);
1280 log_debug("Thin pool transaction id: %" PRIu64
" status: %s.", *transaction_id
, params
);
1284 dm_task_destroy(dmt
);
1289 static int _thin_pool_node_message(struct dm_tree_node
*dnode
, struct thin_message
*tm
)
1291 struct dm_task
*dmt
;
1292 struct dm_thin_message
*m
= &tm
->message
;
1297 case DM_THIN_MESSAGE_CREATE_SNAP
:
1298 r
= dm_snprintf(buf
, sizeof(buf
), "create_snap %u %u",
1299 m
->u
.m_create_snap
.device_id
,
1300 m
->u
.m_create_snap
.origin_id
);
1302 case DM_THIN_MESSAGE_CREATE_THIN
:
1303 r
= dm_snprintf(buf
, sizeof(buf
), "create_thin %u",
1304 m
->u
.m_create_thin
.device_id
);
1306 case DM_THIN_MESSAGE_DELETE
:
1307 r
= dm_snprintf(buf
, sizeof(buf
), "delete %u",
1308 m
->u
.m_delete
.device_id
);
1310 case DM_THIN_MESSAGE_TRIM
:
1311 r
= dm_snprintf(buf
, sizeof(buf
), "trim %u %" PRIu64
,
1312 m
->u
.m_trim
.device_id
,
1313 m
->u
.m_trim
.new_size
);
1315 case DM_THIN_MESSAGE_SET_TRANSACTION_ID
:
1316 r
= dm_snprintf(buf
, sizeof(buf
),
1317 "set_transaction_id %" PRIu64
" %" PRIu64
,
1318 m
->u
.m_set_transaction_id
.current_id
,
1319 m
->u
.m_set_transaction_id
.new_id
);
1326 log_error("Failed to prepare message.");
1332 if (!(dmt
= dm_task_create(DM_DEVICE_TARGET_MSG
)))
1335 if (!dm_task_set_major(dmt
, dnode
->info
.major
) ||
1336 !dm_task_set_minor(dmt
, dnode
->info
.minor
)) {
1337 log_error("Failed to set message major minor.");
1341 if (!dm_task_set_message(dmt
, buf
))
1344 /* Internal functionality of dm_task */
1345 dmt
->expected_errno
= tm
->expected_errno
;
1347 if (!dm_task_run(dmt
))
1352 dm_task_destroy(dmt
);
1357 static int _node_send_messages(struct dm_tree_node
*dnode
,
1358 const char *uuid_prefix
,
1359 size_t uuid_prefix_len
)
1361 struct load_segment
*seg
;
1362 struct thin_message
*tmsg
;
1366 if (!dnode
->info
.exists
|| (dm_list_size(&dnode
->props
.segs
) != 1))
1369 seg
= dm_list_item(dm_list_last(&dnode
->props
.segs
), struct load_segment
);
1370 if (seg
->type
!= SEG_THIN_POOL
)
1373 if (!(uuid
= dm_tree_node_get_uuid(dnode
)))
1376 if (!_uuid_prefix_matches(uuid
, uuid_prefix
, uuid_prefix_len
)) {
1377 log_debug("UUID \"%s\" does not match.", uuid
);
1381 if (!_thin_pool_status_transaction_id(dnode
, &trans_id
))
1384 if (trans_id
== seg
->transaction_id
)
1385 return 1; /* In sync - skip messages */
1387 if (trans_id
!= (seg
->transaction_id
- 1)) {
1388 log_error("Thin pool transaction_id=%" PRIu64
", while expected: %" PRIu64
".",
1389 trans_id
, seg
->transaction_id
- 1);
1390 goto bad
; /* Nothing to send */
1393 dm_list_iterate_items(tmsg
, &seg
->thin_messages
)
1394 if (!(_thin_pool_node_message(dnode
, tmsg
)))
1399 /* Try to deactivate */
1400 if (!(dm_tree_deactivate_children(dnode
, uuid_prefix
, uuid_prefix_len
)))
1401 log_error("Failed to deactivate %s", dnode
->name
);
1407 * FIXME Don't attempt to deactivate known internal dependencies.
1409 static int _dm_tree_deactivate_children(struct dm_tree_node
*dnode
,
1410 const char *uuid_prefix
,
1411 size_t uuid_prefix_len
,
1415 void *handle
= NULL
;
1416 struct dm_tree_node
*child
= dnode
;
1417 struct dm_info info
;
1418 const struct dm_info
*dinfo
;
1422 while ((child
= dm_tree_next_child(&handle
, dnode
, 0))) {
1423 if (!(dinfo
= dm_tree_node_get_info(child
))) {
1428 if (!(name
= dm_tree_node_get_name(child
))) {
1433 if (!(uuid
= dm_tree_node_get_uuid(child
))) {
1438 /* Ignore if it doesn't belong to this VG */
1439 if (!_uuid_prefix_matches(uuid
, uuid_prefix
, uuid_prefix_len
))
1442 /* Refresh open_count */
1443 if (!_info_by_dev(dinfo
->major
, dinfo
->minor
, 1, &info
) ||
1447 if (info
.open_count
) {
1448 /* Skip internal non-toplevel opened nodes */
1452 /* When retry is not allowed, error */
1453 if (!child
->dtree
->retry_remove
) {
1454 log_error("Unable to deactivate open %s (%" PRIu32
1455 ":%" PRIu32
")", name
, info
.major
, info
.minor
);
1460 /* Check toplevel node for holders/mounted fs */
1461 if (!_check_device_not_in_use(name
, &info
)) {
1466 /* Go on with retry */
1469 /* Also checking open_count in parent nodes of presuspend_node */
1470 if ((child
->presuspend_node
&&
1471 !_node_has_closed_parents(child
->presuspend_node
,
1472 uuid_prefix
, uuid_prefix_len
))) {
1473 /* Only report error from (likely non-internal) dependency at top level */
1475 log_error("Unable to deactivate open %s (%" PRIu32
1476 ":%" PRIu32
")", name
, info
.major
,
1483 /* Suspend child node first if requested */
1484 if (child
->presuspend_node
&&
1485 !dm_tree_suspend_children(child
, uuid_prefix
, uuid_prefix_len
))
1488 if (!_deactivate_node(name
, info
.major
, info
.minor
,
1489 &child
->dtree
->cookie
, child
->udev_flags
,
1490 (level
== 0) ? child
->dtree
->retry_remove
: 0)) {
1491 log_error("Unable to deactivate %s (%" PRIu32
1492 ":%" PRIu32
")", name
, info
.major
,
1496 } else if (info
.suspended
)
1499 if (dm_tree_node_num_children(child
, 0)) {
1500 if (!_dm_tree_deactivate_children(child
, uuid_prefix
, uuid_prefix_len
, level
+ 1))
1508 int dm_tree_deactivate_children(struct dm_tree_node
*dnode
,
1509 const char *uuid_prefix
,
1510 size_t uuid_prefix_len
)
1512 return _dm_tree_deactivate_children(dnode
, uuid_prefix
, uuid_prefix_len
, 0);
1515 void dm_tree_skip_lockfs(struct dm_tree_node
*dnode
)
1517 dnode
->dtree
->skip_lockfs
= 1;
1520 void dm_tree_use_no_flush_suspend(struct dm_tree_node
*dnode
)
1522 dnode
->dtree
->no_flush
= 1;
1525 void dm_tree_retry_remove(struct dm_tree_node
*dnode
)
1527 dnode
->dtree
->retry_remove
= 1;
1530 int dm_tree_suspend_children(struct dm_tree_node
*dnode
,
1531 const char *uuid_prefix
,
1532 size_t uuid_prefix_len
)
1535 void *handle
= NULL
;
1536 struct dm_tree_node
*child
= dnode
;
1537 struct dm_info info
, newinfo
;
1538 const struct dm_info
*dinfo
;
1542 /* Suspend nodes at this level of the tree */
1543 while ((child
= dm_tree_next_child(&handle
, dnode
, 0))) {
1544 if (!(dinfo
= dm_tree_node_get_info(child
))) {
1549 if (!(name
= dm_tree_node_get_name(child
))) {
1554 if (!(uuid
= dm_tree_node_get_uuid(child
))) {
1559 /* Ignore if it doesn't belong to this VG */
1560 if (!_uuid_prefix_matches(uuid
, uuid_prefix
, uuid_prefix_len
))
1563 /* Ensure immediate parents are already suspended */
1564 if (!_children_suspended(child
, 1, uuid_prefix
, uuid_prefix_len
))
1567 if (!_info_by_dev(dinfo
->major
, dinfo
->minor
, 0, &info
) ||
1568 !info
.exists
|| info
.suspended
)
1571 if (!_suspend_node(name
, info
.major
, info
.minor
,
1572 child
->dtree
->skip_lockfs
,
1573 child
->dtree
->no_flush
, &newinfo
)) {
1574 log_error("Unable to suspend %s (%" PRIu32
1575 ":%" PRIu32
")", name
, info
.major
,
1581 /* Update cached info */
1582 child
->info
= newinfo
;
1585 /* Then suspend any child nodes */
1588 while ((child
= dm_tree_next_child(&handle
, dnode
, 0))) {
1589 if (!(uuid
= dm_tree_node_get_uuid(child
))) {
1594 /* Ignore if it doesn't belong to this VG */
1595 if (!_uuid_prefix_matches(uuid
, uuid_prefix
, uuid_prefix_len
))
1598 if (dm_tree_node_num_children(child
, 0))
1599 if (!dm_tree_suspend_children(child
, uuid_prefix
, uuid_prefix_len
))
1606 int dm_tree_activate_children(struct dm_tree_node
*dnode
,
1607 const char *uuid_prefix
,
1608 size_t uuid_prefix_len
)
1611 void *handle
= NULL
;
1612 struct dm_tree_node
*child
= dnode
;
1613 struct dm_info newinfo
;
1618 /* Activate children first */
1619 while ((child
= dm_tree_next_child(&handle
, dnode
, 0))) {
1620 if (!(uuid
= dm_tree_node_get_uuid(child
))) {
1625 if (!_uuid_prefix_matches(uuid
, uuid_prefix
, uuid_prefix_len
))
1628 if (dm_tree_node_num_children(child
, 0))
1629 if (!dm_tree_activate_children(child
, uuid_prefix
, uuid_prefix_len
))
1635 for (priority
= 0; priority
< 3; priority
++) {
1636 while ((child
= dm_tree_next_child(&handle
, dnode
, 0))) {
1637 if (priority
!= child
->activation_priority
)
1640 if (!(uuid
= dm_tree_node_get_uuid(child
))) {
1645 if (!_uuid_prefix_matches(uuid
, uuid_prefix
, uuid_prefix_len
))
1648 if (!(name
= dm_tree_node_get_name(child
))) {
1654 if (child
->props
.new_name
) {
1655 if (!_rename_node(name
, child
->props
.new_name
, child
->info
.major
,
1656 child
->info
.minor
, &child
->dtree
->cookie
,
1657 child
->udev_flags
)) {
1658 log_error("Failed to rename %s (%" PRIu32
1659 ":%" PRIu32
") to %s", name
, child
->info
.major
,
1660 child
->info
.minor
, child
->props
.new_name
);
1663 child
->name
= child
->props
.new_name
;
1664 child
->props
.new_name
= NULL
;
1667 if (!child
->info
.inactive_table
&& !child
->info
.suspended
)
1670 if (!_resume_node(child
->name
, child
->info
.major
, child
->info
.minor
,
1671 child
->props
.read_ahead
, child
->props
.read_ahead_flags
,
1672 &newinfo
, &child
->dtree
->cookie
, child
->udev_flags
, child
->info
.suspended
)) {
1673 log_error("Unable to resume %s (%" PRIu32
1674 ":%" PRIu32
")", child
->name
, child
->info
.major
,
1680 /* Update cached info */
1681 child
->info
= newinfo
;
1690 static int _create_node(struct dm_tree_node
*dnode
)
1693 struct dm_task
*dmt
;
1695 log_verbose("Creating %s", dnode
->name
);
1697 if (!(dmt
= dm_task_create(DM_DEVICE_CREATE
))) {
1698 log_error("Create dm_task creation failed for %s", dnode
->name
);
1702 if (!dm_task_set_name(dmt
, dnode
->name
)) {
1703 log_error("Failed to set device name for %s", dnode
->name
);
1707 if (!dm_task_set_uuid(dmt
, dnode
->uuid
)) {
1708 log_error("Failed to set uuid for %s", dnode
->name
);
1712 if (dnode
->props
.major
&&
1713 (!dm_task_set_major(dmt
, dnode
->props
.major
) ||
1714 !dm_task_set_minor(dmt
, dnode
->props
.minor
))) {
1715 log_error("Failed to set device number for %s creation.", dnode
->name
);
1719 if (dnode
->props
.read_only
&& !dm_task_set_ro(dmt
)) {
1720 log_error("Failed to set read only flag for %s", dnode
->name
);
1724 if (!dm_task_no_open_count(dmt
))
1725 log_error("Failed to disable open_count");
1727 if ((r
= dm_task_run(dmt
)))
1728 r
= dm_task_get_info(dmt
, &dnode
->info
);
1731 dm_task_destroy(dmt
);
1737 static int _build_dev_string(char *devbuf
, size_t bufsize
, struct dm_tree_node
*node
)
1739 if (!dm_format_dev(devbuf
, bufsize
, node
->info
.major
, node
->info
.minor
)) {
1740 log_error("Failed to format %s device number for %s as dm "
1742 node
->name
, node
->uuid
, node
->info
.major
, node
->info
.minor
);
1749 /* simplify string emiting code */
1750 #define EMIT_PARAMS(p, str...)\
1753 if ((w = dm_snprintf(params + p, paramsize - (size_t) p, str)) < 0) {\
1754 stack; /* Out of space */\
1763 * Returns: 1 on success, 0 on failure
1765 static int _emit_areas_line(struct dm_task
*dmt
__attribute__((unused
)),
1766 struct load_segment
*seg
, char *params
,
1767 size_t paramsize
, int *pos
)
1769 struct seg_area
*area
;
1770 char devbuf
[DM_FORMAT_DEV_BUFSIZE
];
1771 unsigned first_time
= 1;
1772 const char *logtype
, *synctype
;
1773 unsigned log_parm_count
;
1775 dm_list_iterate_items(area
, &seg
->areas
) {
1776 switch (seg
->type
) {
1777 case SEG_REPLICATOR_DEV
:
1778 if (!_build_dev_string(devbuf
, sizeof(devbuf
), area
->dev_node
))
1781 EMIT_PARAMS(*pos
, " %d 1 %s", area
->rsite_index
, devbuf
);
1783 EMIT_PARAMS(*pos
, " nolog 0");
1785 /* Remote devices */
1786 log_parm_count
= (area
->flags
&
1787 (DM_NOSYNC
| DM_FORCESYNC
)) ? 2 : 1;
1790 devbuf
[0] = 0; /* Only core log parameters */
1793 devbuf
[0] = ' '; /* Extra space before device name */
1794 if (!_build_dev_string(devbuf
+ 1,
1799 log_parm_count
++; /* Extra sync log device name parameter */
1802 EMIT_PARAMS(*pos
, " %s %u%s %" PRIu64
, logtype
,
1803 log_parm_count
, devbuf
, area
->region_size
);
1805 synctype
= (area
->flags
& DM_NOSYNC
) ?
1806 " nosync" : (area
->flags
& DM_FORCESYNC
) ?
1810 EMIT_PARAMS(*pos
, "%s", synctype
);
1822 if (!area
->dev_node
) {
1823 EMIT_PARAMS(*pos
, " -");
1826 if (!_build_dev_string(devbuf
, sizeof(devbuf
), area
->dev_node
))
1829 EMIT_PARAMS(*pos
, " %s", devbuf
);
1832 if (!_build_dev_string(devbuf
, sizeof(devbuf
), area
->dev_node
))
1835 EMIT_PARAMS(*pos
, "%s%s %" PRIu64
, first_time
? "" : " ",
1836 devbuf
, area
->offset
);
1845 static int _replicator_emit_segment_line(const struct load_segment
*seg
, char *params
,
1846 size_t paramsize
, int *pos
)
1848 const struct load_segment
*rlog_seg
;
1849 struct replicator_site
*rsite
;
1850 char rlogbuf
[DM_FORMAT_DEV_BUFSIZE
];
1851 unsigned parm_count
;
1853 if (!seg
->log
|| !_build_dev_string(rlogbuf
, sizeof(rlogbuf
), seg
->log
))
1856 rlog_seg
= dm_list_item(dm_list_last(&seg
->log
->props
.segs
),
1857 struct load_segment
);
1859 EMIT_PARAMS(*pos
, "%s 4 %s 0 auto %" PRIu64
,
1860 seg
->rlog_type
, rlogbuf
, rlog_seg
->size
);
1862 dm_list_iterate_items(rsite
, &seg
->rsites
) {
1863 parm_count
= (rsite
->fall_behind_data
1864 || rsite
->fall_behind_ios
1865 || rsite
->async_timeout
) ? 4 : 2;
1867 EMIT_PARAMS(*pos
, " blockdev %u %u %s", parm_count
, rsite
->rsite_index
,
1868 (rsite
->mode
== DM_REPLICATOR_SYNC
) ? "synchronous" : "asynchronous");
1870 if (rsite
->fall_behind_data
)
1871 EMIT_PARAMS(*pos
, " data %" PRIu64
, rsite
->fall_behind_data
);
1872 else if (rsite
->fall_behind_ios
)
1873 EMIT_PARAMS(*pos
, " ios %" PRIu32
, rsite
->fall_behind_ios
);
1874 else if (rsite
->async_timeout
)
1875 EMIT_PARAMS(*pos
, " timeout %" PRIu32
, rsite
->async_timeout
);
1882 * Returns: 1 on success, 0 on failure
1884 static int _mirror_emit_segment_line(struct dm_task
*dmt
, struct load_segment
*seg
,
1885 char *params
, size_t paramsize
)
1887 int block_on_error
= 0;
1888 int handle_errors
= 0;
1889 int dm_log_userspace
= 0;
1891 unsigned log_parm_count
;
1893 char logbuf
[DM_FORMAT_DEV_BUFSIZE
];
1894 const char *logtype
;
1895 unsigned kmaj
= 0, kmin
= 0, krel
= 0;
1897 if (uname(&uts
) == -1) {
1898 log_error("Cannot read kernel release version.");
1902 /* Kernels with a major number of 2 always had 3 parts. */
1903 parts
= sscanf(uts
.release
, "%u.%u.%u", &kmaj
, &kmin
, &krel
);
1904 if (parts
< 1 || (kmaj
< 3 && parts
< 3)) {
1905 log_error("Wrong kernel release version %s.", uts
.release
);
1909 if ((seg
->flags
& DM_BLOCK_ON_ERROR
)) {
1911 * Originally, block_on_error was an argument to the log
1912 * portion of the mirror CTR table. It was renamed to
1913 * "handle_errors" and now resides in the 'features'
1914 * section of the mirror CTR table (i.e. at the end).
1916 * We can identify whether to use "block_on_error" or
1917 * "handle_errors" by the dm-mirror module's version
1918 * number (>= 1.12) or by the kernel version (>= 2.6.22).
1920 if (KERNEL_VERSION(kmaj
, kmin
, krel
) >= KERNEL_VERSION(2, 6, 22))
1926 if (seg
->clustered
) {
1927 /* Cluster mirrors require a UUID */
1932 * Cluster mirrors used to have their own log
1933 * types. Now they are accessed through the
1934 * userspace log type.
1936 * The dm-log-userspace module was added to the
1939 if (KERNEL_VERSION(kmaj
, kmin
, krel
) >= KERNEL_VERSION(2, 6, 31))
1940 dm_log_userspace
= 1;
1946 /* [no]sync, block_on_error etc. */
1947 log_parm_count
+= hweight32(seg
->flags
);
1949 /* "handle_errors" is a feature arg now */
1953 /* DM_CORELOG does not count in the param list */
1954 if (seg
->flags
& DM_CORELOG
)
1957 if (seg
->clustered
) {
1958 log_parm_count
++; /* For UUID */
1960 if (!dm_log_userspace
)
1961 EMIT_PARAMS(pos
, "clustered-");
1963 /* For clustered-* type field inserted later */
1972 if (!_build_dev_string(logbuf
, sizeof(logbuf
), seg
->log
))
1976 if (dm_log_userspace
)
1977 EMIT_PARAMS(pos
, "userspace %u %s clustered-%s",
1978 log_parm_count
, seg
->uuid
, logtype
);
1980 EMIT_PARAMS(pos
, "%s %u", logtype
, log_parm_count
);
1983 EMIT_PARAMS(pos
, " %s", logbuf
);
1985 EMIT_PARAMS(pos
, " %u", seg
->region_size
);
1987 if (seg
->clustered
&& !dm_log_userspace
)
1988 EMIT_PARAMS(pos
, " %s", seg
->uuid
);
1990 if ((seg
->flags
& DM_NOSYNC
))
1991 EMIT_PARAMS(pos
, " nosync");
1992 else if ((seg
->flags
& DM_FORCESYNC
))
1993 EMIT_PARAMS(pos
, " sync");
1996 EMIT_PARAMS(pos
, " block_on_error");
1998 EMIT_PARAMS(pos
, " %u ", seg
->mirror_area_count
);
2000 if (_emit_areas_line(dmt
, seg
, params
, paramsize
, &pos
) <= 0)
2004 EMIT_PARAMS(pos
, " 1 handle_errors");
2009 static int _raid_emit_segment_line(struct dm_task
*dmt
, uint32_t major
,
2010 uint32_t minor
, struct load_segment
*seg
,
2011 uint64_t *seg_start
, char *params
,
2015 int param_count
= 1; /* mandatory 'chunk size'/'stripe size' arg */
2018 if ((seg
->flags
& DM_NOSYNC
) || (seg
->flags
& DM_FORCESYNC
))
2021 if (seg
->region_size
)
2024 /* rebuilds is 64-bit */
2025 param_count
+= 2 * hweight32(seg
->rebuilds
& 0xFFFFFFFF);
2026 param_count
+= 2 * hweight32(seg
->rebuilds
>> 32);
2028 if ((seg
->type
== SEG_RAID1
) && seg
->stripe_size
)
2029 log_error("WARNING: Ignoring RAID1 stripe size");
2031 EMIT_PARAMS(pos
, "%s %d %u", dm_segtypes
[seg
->type
].target
,
2032 param_count
, seg
->stripe_size
);
2034 if (seg
->flags
& DM_NOSYNC
)
2035 EMIT_PARAMS(pos
, " nosync");
2036 else if (seg
->flags
& DM_FORCESYNC
)
2037 EMIT_PARAMS(pos
, " sync");
2039 if (seg
->region_size
)
2040 EMIT_PARAMS(pos
, " region_size %u", seg
->region_size
);
2042 for (i
= 0; i
< (seg
->area_count
/ 2); i
++)
2043 if (seg
->rebuilds
& (1 << i
))
2044 EMIT_PARAMS(pos
, " rebuild %u", i
);
2046 /* Print number of metadata/data device pairs */
2047 EMIT_PARAMS(pos
, " %u", seg
->area_count
/2);
2049 if (_emit_areas_line(dmt
, seg
, params
, paramsize
, &pos
) <= 0)
2055 static int _emit_segment_line(struct dm_task
*dmt
, uint32_t major
,
2056 uint32_t minor
, struct load_segment
*seg
,
2057 uint64_t *seg_start
, char *params
,
2062 int target_type_is_raid
= 0;
2063 char originbuf
[DM_FORMAT_DEV_BUFSIZE
], cowbuf
[DM_FORMAT_DEV_BUFSIZE
];
2064 char pool
[DM_FORMAT_DEV_BUFSIZE
], metadata
[DM_FORMAT_DEV_BUFSIZE
];
2072 /* Mirrors are pretty complicated - now in separate function */
2073 r
= _mirror_emit_segment_line(dmt
, seg
, params
, paramsize
);
2077 case SEG_REPLICATOR
:
2078 if ((r
= _replicator_emit_segment_line(seg
, params
, paramsize
,
2084 case SEG_REPLICATOR_DEV
:
2085 if (!seg
->replicator
|| !_build_dev_string(originbuf
,
2090 EMIT_PARAMS(pos
, "%s %" PRIu64
, originbuf
, seg
->rdevice_index
);
2093 case SEG_SNAPSHOT_MERGE
:
2094 if (!_build_dev_string(originbuf
, sizeof(originbuf
), seg
->origin
))
2096 if (!_build_dev_string(cowbuf
, sizeof(cowbuf
), seg
->cow
))
2098 EMIT_PARAMS(pos
, "%s %s %c %d", originbuf
, cowbuf
,
2099 seg
->persistent
? 'P' : 'N', seg
->chunk_size
);
2101 case SEG_SNAPSHOT_ORIGIN
:
2102 if (!_build_dev_string(originbuf
, sizeof(originbuf
), seg
->origin
))
2104 EMIT_PARAMS(pos
, "%s", originbuf
);
2107 EMIT_PARAMS(pos
, "%u %u ", seg
->area_count
, seg
->stripe_size
);
2110 EMIT_PARAMS(pos
, "%s%s%s%s%s %s %" PRIu64
" ", seg
->cipher
,
2111 seg
->chainmode
? "-" : "", seg
->chainmode
?: "",
2112 seg
->iv
? "-" : "", seg
->iv
?: "", seg
->key
,
2113 seg
->iv_offset
!= DM_CRYPT_IV_DEFAULT
?
2114 seg
->iv_offset
: *seg_start
);
2125 target_type_is_raid
= 1;
2126 r
= _raid_emit_segment_line(dmt
, major
, minor
, seg
, seg_start
,
2133 if (!_build_dev_string(metadata
, sizeof(metadata
), seg
->metadata
))
2135 if (!_build_dev_string(pool
, sizeof(pool
), seg
->pool
))
2137 EMIT_PARAMS(pos
, "%s %s %d %" PRIu64
" %s", metadata
, pool
,
2138 seg
->data_block_size
, seg
->low_water_mark
,
2139 seg
->skip_block_zeroing
? "1 skip_block_zeroing" : "0");
2142 if (!_build_dev_string(pool
, sizeof(pool
), seg
->pool
))
2144 EMIT_PARAMS(pos
, "%s %d", pool
, seg
->device_id
);
2150 case SEG_REPLICATOR
:
2152 case SEG_SNAPSHOT_ORIGIN
:
2153 case SEG_SNAPSHOT_MERGE
:
2160 case SEG_REPLICATOR_DEV
:
2162 if ((r
= _emit_areas_line(dmt
, seg
, params
, paramsize
, &pos
)) <= 0) {
2167 log_error("No parameters supplied for %s target "
2168 "%u:%u.", dm_segtypes
[seg
->type
].target
,
2175 log_debug("Adding target to (%" PRIu32
":%" PRIu32
"): %" PRIu64
2176 " %" PRIu64
" %s %s", major
, minor
,
2177 *seg_start
, seg
->size
, target_type_is_raid
? "raid" :
2178 dm_segtypes
[seg
->type
].target
, params
);
2180 if (!dm_task_add_target(dmt
, *seg_start
, seg
->size
,
2181 target_type_is_raid
? "raid" :
2182 dm_segtypes
[seg
->type
].target
, params
))
2185 *seg_start
+= seg
->size
;
2192 static int _emit_segment(struct dm_task
*dmt
, uint32_t major
, uint32_t minor
,
2193 struct load_segment
*seg
, uint64_t *seg_start
)
2196 size_t paramsize
= 4096;
2200 if (!(params
= dm_malloc(paramsize
))) {
2201 log_error("Insufficient space for target parameters.");
2206 ret
= _emit_segment_line(dmt
, major
, minor
, seg
, seg_start
,
2216 log_debug("Insufficient space in params[%" PRIsize_t
2217 "] for target parameters.", paramsize
);
2220 } while (paramsize
< MAX_TARGET_PARAMSIZE
);
2222 log_error("Target parameter size too big. Aborting.");
2226 static int _load_node(struct dm_tree_node
*dnode
)
2229 struct dm_task
*dmt
;
2230 struct load_segment
*seg
;
2231 uint64_t seg_start
= 0, existing_table_size
;
2233 log_verbose("Loading %s table (%" PRIu32
":%" PRIu32
")", dnode
->name
,
2234 dnode
->info
.major
, dnode
->info
.minor
);
2236 if (!(dmt
= dm_task_create(DM_DEVICE_RELOAD
))) {
2237 log_error("Reload dm_task creation failed for %s", dnode
->name
);
2241 if (!dm_task_set_major(dmt
, dnode
->info
.major
) ||
2242 !dm_task_set_minor(dmt
, dnode
->info
.minor
)) {
2243 log_error("Failed to set device number for %s reload.", dnode
->name
);
2247 if (dnode
->props
.read_only
&& !dm_task_set_ro(dmt
)) {
2248 log_error("Failed to set read only flag for %s", dnode
->name
);
2252 if (!dm_task_no_open_count(dmt
))
2253 log_error("Failed to disable open_count");
2255 dm_list_iterate_items(seg
, &dnode
->props
.segs
)
2256 if (!_emit_segment(dmt
, dnode
->info
.major
, dnode
->info
.minor
,
2260 if (!dm_task_suppress_identical_reload(dmt
))
2261 log_error("Failed to suppress reload of identical tables.");
2263 if ((r
= dm_task_run(dmt
))) {
2264 r
= dm_task_get_info(dmt
, &dnode
->info
);
2265 if (r
&& !dnode
->info
.inactive_table
)
2266 log_verbose("Suppressed %s identical table reload.",
2269 existing_table_size
= dm_task_get_existing_table_size(dmt
);
2270 if ((dnode
->props
.size_changed
=
2271 (existing_table_size
== seg_start
) ? 0 : 1)) {
2272 log_debug("Table size changed from %" PRIu64
" to %"
2273 PRIu64
" for %s", existing_table_size
,
2274 seg_start
, dnode
->name
);
2276 * Kernel usually skips size validation on zero-length devices
2277 * now so no need to preload them.
2279 /* FIXME In which kernel version did this begin? */
2280 if (!existing_table_size
&& dnode
->props
.delay_resume_if_new
)
2281 dnode
->props
.size_changed
= 0;
2285 dnode
->props
.segment_count
= 0;
2288 dm_task_destroy(dmt
);
2293 int dm_tree_preload_children(struct dm_tree_node
*dnode
,
2294 const char *uuid_prefix
,
2295 size_t uuid_prefix_len
)
2298 void *handle
= NULL
;
2299 struct dm_tree_node
*child
;
2300 struct dm_info newinfo
;
2301 int update_devs_flag
= 0;
2303 /* Preload children first */
2304 while ((child
= dm_tree_next_child(&handle
, dnode
, 0))) {
2305 /* Skip existing non-device-mapper devices */
2306 if (!child
->info
.exists
&& child
->info
.major
)
2309 /* Ignore if it doesn't belong to this VG */
2310 if (child
->info
.exists
&&
2311 !_uuid_prefix_matches(child
->uuid
, uuid_prefix
, uuid_prefix_len
))
2314 if (dm_tree_node_num_children(child
, 0))
2315 if (!dm_tree_preload_children(child
, uuid_prefix
, uuid_prefix_len
))
2318 /* FIXME Cope if name exists with no uuid? */
2319 if (!child
->info
.exists
&& !_create_node(child
))
2322 if (!child
->info
.inactive_table
&&
2323 child
->props
.segment_count
&&
2327 /* Propagate device size change change */
2328 if (child
->props
.size_changed
)
2329 dnode
->props
.size_changed
= 1;
2331 /* Resume device immediately if it has parents and its size changed */
2332 if (!dm_tree_node_num_children(child
, 1) || !child
->props
.size_changed
)
2335 if (!child
->info
.inactive_table
&& !child
->info
.suspended
)
2338 if (!_resume_node(child
->name
, child
->info
.major
, child
->info
.minor
,
2339 child
->props
.read_ahead
, child
->props
.read_ahead_flags
,
2340 &newinfo
, &child
->dtree
->cookie
, child
->udev_flags
,
2341 child
->info
.suspended
)) {
2342 log_error("Unable to resume %s (%" PRIu32
2343 ":%" PRIu32
")", child
->name
, child
->info
.major
,
2349 /* Update cached info */
2350 child
->info
= newinfo
;
2351 if (child
->props
.send_messages
&&
2352 !(r
= _node_send_messages(child
, uuid_prefix
, uuid_prefix_len
))) {
2357 * Prepare for immediate synchronization with udev and flush all stacked
2358 * dev node operations if requested by immediate_dev_node property. But
2359 * finish processing current level in the tree first.
2361 if (child
->props
.immediate_dev_node
)
2362 update_devs_flag
= 1;
2365 if (r
&& dnode
->props
.send_messages
&&
2366 !(r
= _node_send_messages(dnode
, uuid_prefix
, uuid_prefix_len
)))
2369 if (update_devs_flag
) {
2370 if (!dm_udev_wait(dm_tree_get_cookie(dnode
)))
2372 dm_tree_set_cookie(dnode
, 0);
2375 if (r
&& !_node_send_messages(dnode
, uuid_prefix
, uuid_prefix_len
)) {
2377 if (!(dm_tree_deactivate_children(dnode
, uuid_prefix
, uuid_prefix_len
)))
2378 log_error("Failed to deactivate %s", dnode
->name
);
2386 * Returns 1 if unsure.
2388 int dm_tree_children_use_uuid(struct dm_tree_node
*dnode
,
2389 const char *uuid_prefix
,
2390 size_t uuid_prefix_len
)
2392 void *handle
= NULL
;
2393 struct dm_tree_node
*child
= dnode
;
2396 while ((child
= dm_tree_next_child(&handle
, dnode
, 0))) {
2397 if (!(uuid
= dm_tree_node_get_uuid(child
))) {
2398 log_error("Failed to get uuid for dtree node.");
2402 if (_uuid_prefix_matches(uuid
, uuid_prefix
, uuid_prefix_len
))
2405 if (dm_tree_node_num_children(child
, 0))
2406 dm_tree_children_use_uuid(child
, uuid_prefix
, uuid_prefix_len
);
2415 static struct load_segment
*_add_segment(struct dm_tree_node
*dnode
, unsigned type
, uint64_t size
)
2417 struct load_segment
*seg
;
2419 if (!(seg
= dm_pool_zalloc(dnode
->dtree
->mem
, sizeof(*seg
)))) {
2420 log_error("dtree node segment allocation failed");
2426 seg
->area_count
= 0;
2427 dm_list_init(&seg
->areas
);
2428 seg
->stripe_size
= 0;
2429 seg
->persistent
= 0;
2430 seg
->chunk_size
= 0;
2435 dm_list_add(&dnode
->props
.segs
, &seg
->list
);
2436 dnode
->props
.segment_count
++;
2441 int dm_tree_node_add_snapshot_origin_target(struct dm_tree_node
*dnode
,
2443 const char *origin_uuid
)
2445 struct load_segment
*seg
;
2446 struct dm_tree_node
*origin_node
;
2448 if (!(seg
= _add_segment(dnode
, SEG_SNAPSHOT_ORIGIN
, size
)))
2451 if (!(origin_node
= dm_tree_find_node_by_uuid(dnode
->dtree
, origin_uuid
))) {
2452 log_error("Couldn't find snapshot origin uuid %s.", origin_uuid
);
2456 seg
->origin
= origin_node
;
2457 if (!_link_tree_nodes(dnode
, origin_node
))
2460 /* Resume snapshot origins after new snapshots */
2461 dnode
->activation_priority
= 1;
2466 static int _add_snapshot_target(struct dm_tree_node
*node
,
2468 const char *origin_uuid
,
2469 const char *cow_uuid
,
2470 const char *merge_uuid
,
2472 uint32_t chunk_size
)
2474 struct load_segment
*seg
;
2475 struct dm_tree_node
*origin_node
, *cow_node
, *merge_node
;
2478 seg_type
= !merge_uuid
? SEG_SNAPSHOT
: SEG_SNAPSHOT_MERGE
;
2480 if (!(seg
= _add_segment(node
, seg_type
, size
)))
2483 if (!(origin_node
= dm_tree_find_node_by_uuid(node
->dtree
, origin_uuid
))) {
2484 log_error("Couldn't find snapshot origin uuid %s.", origin_uuid
);
2488 seg
->origin
= origin_node
;
2489 if (!_link_tree_nodes(node
, origin_node
))
2492 if (!(cow_node
= dm_tree_find_node_by_uuid(node
->dtree
, cow_uuid
))) {
2493 log_error("Couldn't find snapshot COW device uuid %s.", cow_uuid
);
2497 seg
->cow
= cow_node
;
2498 if (!_link_tree_nodes(node
, cow_node
))
2501 seg
->persistent
= persistent
? 1 : 0;
2502 seg
->chunk_size
= chunk_size
;
2505 if (!(merge_node
= dm_tree_find_node_by_uuid(node
->dtree
, merge_uuid
))) {
2506 /* not a pure error, merging snapshot may have been deactivated */
2507 log_verbose("Couldn't find merging snapshot uuid %s.", merge_uuid
);
2509 seg
->merge
= merge_node
;
2510 /* must not link merging snapshot, would undermine activation_priority below */
2513 /* Resume snapshot-merge (acting origin) after other snapshots */
2514 node
->activation_priority
= 1;
2516 /* Resume merging snapshot after snapshot-merge */
2517 seg
->merge
->activation_priority
= 2;
2525 int dm_tree_node_add_snapshot_target(struct dm_tree_node
*node
,
2527 const char *origin_uuid
,
2528 const char *cow_uuid
,
2530 uint32_t chunk_size
)
2532 return _add_snapshot_target(node
, size
, origin_uuid
, cow_uuid
,
2533 NULL
, persistent
, chunk_size
);
2536 int dm_tree_node_add_snapshot_merge_target(struct dm_tree_node
*node
,
2538 const char *origin_uuid
,
2539 const char *cow_uuid
,
2540 const char *merge_uuid
,
2541 uint32_t chunk_size
)
2543 return _add_snapshot_target(node
, size
, origin_uuid
, cow_uuid
,
2544 merge_uuid
, 1, chunk_size
);
2547 int dm_tree_node_add_error_target(struct dm_tree_node
*node
,
2550 if (!_add_segment(node
, SEG_ERROR
, size
))
2556 int dm_tree_node_add_zero_target(struct dm_tree_node
*node
,
2559 if (!_add_segment(node
, SEG_ZERO
, size
))
2565 int dm_tree_node_add_linear_target(struct dm_tree_node
*node
,
2568 if (!_add_segment(node
, SEG_LINEAR
, size
))
2574 int dm_tree_node_add_striped_target(struct dm_tree_node
*node
,
2576 uint32_t stripe_size
)
2578 struct load_segment
*seg
;
2580 if (!(seg
= _add_segment(node
, SEG_STRIPED
, size
)))
2583 seg
->stripe_size
= stripe_size
;
2588 int dm_tree_node_add_crypt_target(struct dm_tree_node
*node
,
2591 const char *chainmode
,
2596 struct load_segment
*seg
;
2598 if (!(seg
= _add_segment(node
, SEG_CRYPT
, size
)))
2601 seg
->cipher
= cipher
;
2602 seg
->chainmode
= chainmode
;
2604 seg
->iv_offset
= iv_offset
;
2610 int dm_tree_node_add_mirror_target_log(struct dm_tree_node
*node
,
2611 uint32_t region_size
,
2613 const char *log_uuid
,
2614 unsigned area_count
,
2617 struct dm_tree_node
*log_node
= NULL
;
2618 struct load_segment
*seg
;
2620 if (!node
->props
.segment_count
) {
2621 log_error(INTERNAL_ERROR
"Attempt to add target area to missing segment.");
2625 seg
= dm_list_item(dm_list_last(&node
->props
.segs
), struct load_segment
);
2628 if (!(seg
->uuid
= dm_pool_strdup(node
->dtree
->mem
, log_uuid
))) {
2629 log_error("log uuid pool_strdup failed");
2632 if ((flags
& DM_CORELOG
))
2633 /* For pvmove: immediate resume (for size validation) isn't needed. */
2634 node
->props
.delay_resume_if_new
= 1;
2636 if (!(log_node
= dm_tree_find_node_by_uuid(node
->dtree
, log_uuid
))) {
2637 log_error("Couldn't find mirror log uuid %s.", log_uuid
);
2642 log_node
->props
.immediate_dev_node
= 1;
2644 /* The kernel validates the size of disk logs. */
2645 /* FIXME Propagate to any devices below */
2646 log_node
->props
.delay_resume_if_new
= 0;
2648 if (!_link_tree_nodes(node
, log_node
))
2653 seg
->log
= log_node
;
2654 seg
->region_size
= region_size
;
2655 seg
->clustered
= clustered
;
2656 seg
->mirror_area_count
= area_count
;
2662 int dm_tree_node_add_mirror_target(struct dm_tree_node
*node
,
2665 if (!_add_segment(node
, SEG_MIRRORED
, size
))
2671 int dm_tree_node_add_raid_target(struct dm_tree_node
*node
,
2673 const char *raid_type
,
2674 uint32_t region_size
,
2675 uint32_t stripe_size
,
2680 struct load_segment
*seg
= NULL
;
2682 for (i
= 0; dm_segtypes
[i
].target
&& !seg
; i
++)
2683 if (!strcmp(raid_type
, dm_segtypes
[i
].target
))
2684 if (!(seg
= _add_segment(node
,
2685 dm_segtypes
[i
].type
, size
)))
2691 seg
->region_size
= region_size
;
2692 seg
->stripe_size
= stripe_size
;
2693 seg
->area_count
= 0;
2694 seg
->rebuilds
= rebuilds
;
2699 int dm_tree_node_add_replicator_target(struct dm_tree_node
*node
,
2701 const char *rlog_uuid
,
2702 const char *rlog_type
,
2703 unsigned rsite_index
,
2704 dm_replicator_mode_t mode
,
2705 uint32_t async_timeout
,
2706 uint64_t fall_behind_data
,
2707 uint32_t fall_behind_ios
)
2709 struct load_segment
*rseg
;
2710 struct replicator_site
*rsite
;
2712 /* Local site0 - adds replicator segment and links rlog device */
2713 if (rsite_index
== REPLICATOR_LOCAL_SITE
) {
2714 if (node
->props
.segment_count
) {
2715 log_error(INTERNAL_ERROR
"Attempt to add replicator segment to already used node.");
2719 if (!(rseg
= _add_segment(node
, SEG_REPLICATOR
, size
)))
2722 if (!(rseg
->log
= dm_tree_find_node_by_uuid(node
->dtree
, rlog_uuid
))) {
2723 log_error("Missing replicator log uuid %s.", rlog_uuid
);
2727 if (!_link_tree_nodes(node
, rseg
->log
))
2730 if (strcmp(rlog_type
, "ringbuffer") != 0) {
2731 log_error("Unsupported replicator log type %s.", rlog_type
);
2735 if (!(rseg
->rlog_type
= dm_pool_strdup(node
->dtree
->mem
, rlog_type
)))
2738 dm_list_init(&rseg
->rsites
);
2739 rseg
->rdevice_count
= 0;
2740 node
->activation_priority
= 1;
2743 /* Add site to segment */
2744 if (mode
== DM_REPLICATOR_SYNC
2745 && (async_timeout
|| fall_behind_ios
|| fall_behind_data
)) {
2746 log_error("Async parameters passed for synchronnous replicator.");
2750 if (node
->props
.segment_count
!= 1) {
2751 log_error(INTERNAL_ERROR
"Attempt to add remote site area before setting replicator log.");
2755 rseg
= dm_list_item(dm_list_last(&node
->props
.segs
), struct load_segment
);
2756 if (rseg
->type
!= SEG_REPLICATOR
) {
2757 log_error(INTERNAL_ERROR
"Attempt to use non replicator segment %s.",
2758 dm_segtypes
[rseg
->type
].target
);
2762 if (!(rsite
= dm_pool_zalloc(node
->dtree
->mem
, sizeof(*rsite
)))) {
2763 log_error("Failed to allocate remote site segment.");
2767 dm_list_add(&rseg
->rsites
, &rsite
->list
);
2768 rseg
->rsite_count
++;
2771 rsite
->async_timeout
= async_timeout
;
2772 rsite
->fall_behind_data
= fall_behind_data
;
2773 rsite
->fall_behind_ios
= fall_behind_ios
;
2774 rsite
->rsite_index
= rsite_index
;
2779 /* Appends device node to Replicator */
2780 int dm_tree_node_add_replicator_dev_target(struct dm_tree_node
*node
,
2782 const char *replicator_uuid
,
2783 uint64_t rdevice_index
,
2784 const char *rdev_uuid
,
2785 unsigned rsite_index
,
2786 const char *slog_uuid
,
2787 uint32_t slog_flags
,
2788 uint32_t slog_region_size
)
2790 struct seg_area
*area
;
2791 struct load_segment
*rseg
;
2792 struct load_segment
*rep_seg
;
2794 if (rsite_index
== REPLICATOR_LOCAL_SITE
) {
2795 /* Site index for local target */
2796 if (!(rseg
= _add_segment(node
, SEG_REPLICATOR_DEV
, size
)))
2799 if (!(rseg
->replicator
= dm_tree_find_node_by_uuid(node
->dtree
, replicator_uuid
))) {
2800 log_error("Missing replicator uuid %s.", replicator_uuid
);
2804 /* Local slink0 for replicator must be always initialized first */
2805 if (rseg
->replicator
->props
.segment_count
!= 1) {
2806 log_error(INTERNAL_ERROR
"Attempt to use non replicator segment.");
2810 rep_seg
= dm_list_item(dm_list_last(&rseg
->replicator
->props
.segs
), struct load_segment
);
2811 if (rep_seg
->type
!= SEG_REPLICATOR
) {
2812 log_error(INTERNAL_ERROR
"Attempt to use non replicator segment %s.",
2813 dm_segtypes
[rep_seg
->type
].target
);
2816 rep_seg
->rdevice_count
++;
2818 if (!_link_tree_nodes(node
, rseg
->replicator
))
2821 rseg
->rdevice_index
= rdevice_index
;
2823 /* Local slink0 for replicator must be always initialized first */
2824 if (node
->props
.segment_count
!= 1) {
2825 log_error(INTERNAL_ERROR
"Attempt to use non replicator-dev segment.");
2829 rseg
= dm_list_item(dm_list_last(&node
->props
.segs
), struct load_segment
);
2830 if (rseg
->type
!= SEG_REPLICATOR_DEV
) {
2831 log_error(INTERNAL_ERROR
"Attempt to use non replicator-dev segment %s.",
2832 dm_segtypes
[rseg
->type
].target
);
2837 if (!(slog_flags
& DM_CORELOG
) && !slog_uuid
) {
2838 log_error("Unspecified sync log uuid.");
2842 if (!dm_tree_node_add_target_area(node
, NULL
, rdev_uuid
, 0))
2845 area
= dm_list_item(dm_list_last(&rseg
->areas
), struct seg_area
);
2847 if (!(slog_flags
& DM_CORELOG
)) {
2848 if (!(area
->slog
= dm_tree_find_node_by_uuid(node
->dtree
, slog_uuid
))) {
2849 log_error("Couldn't find sync log uuid %s.", slog_uuid
);
2853 if (!_link_tree_nodes(node
, area
->slog
))
2857 area
->flags
= slog_flags
;
2858 area
->region_size
= slog_region_size
;
2859 area
->rsite_index
= rsite_index
;
2864 static int _thin_validate_device_id(uint32_t device_id
)
2866 if (device_id
> DM_THIN_MAX_DEVICE_ID
) {
2867 log_error("Device id %u is higher then %u.",
2868 device_id
, DM_THIN_MAX_DEVICE_ID
);
2875 int dm_tree_node_add_thin_pool_target(struct dm_tree_node
*node
,
2877 uint64_t transaction_id
,
2878 const char *metadata_uuid
,
2879 const char *pool_uuid
,
2880 uint32_t data_block_size
,
2881 uint64_t low_water_mark
,
2882 unsigned skip_block_zeroing
)
2884 struct load_segment
*seg
;
2886 if (data_block_size
< DM_THIN_MIN_DATA_BLOCK_SIZE
) {
2887 log_error("Data block size %u is lower then %u sectors.",
2888 data_block_size
, DM_THIN_MIN_DATA_BLOCK_SIZE
);
2892 if (data_block_size
> DM_THIN_MAX_DATA_BLOCK_SIZE
) {
2893 log_error("Data block size %u is higher then %u sectors.",
2894 data_block_size
, DM_THIN_MAX_DATA_BLOCK_SIZE
);
2898 if (!(seg
= _add_segment(node
, SEG_THIN_POOL
, size
)))
2901 if (!(seg
->metadata
= dm_tree_find_node_by_uuid(node
->dtree
, metadata_uuid
))) {
2902 log_error("Missing metadata uuid %s.", metadata_uuid
);
2906 if (!_link_tree_nodes(node
, seg
->metadata
))
2909 if (!(seg
->pool
= dm_tree_find_node_by_uuid(node
->dtree
, pool_uuid
))) {
2910 log_error("Missing pool uuid %s.", pool_uuid
);
2914 if (!_link_tree_nodes(node
, seg
->pool
))
2917 node
->props
.send_messages
= 1;
2918 seg
->transaction_id
= transaction_id
;
2919 seg
->low_water_mark
= low_water_mark
;
2920 seg
->data_block_size
= data_block_size
;
2921 seg
->skip_block_zeroing
= skip_block_zeroing
;
2922 dm_list_init(&seg
->thin_messages
);
2927 int dm_tree_node_add_thin_pool_message(struct dm_tree_node
*node
,
2928 dm_thin_message_t type
,
2929 uint64_t id1
, uint64_t id2
)
2931 struct load_segment
*seg
;
2932 struct thin_message
*tm
;
2934 if (node
->props
.segment_count
!= 1) {
2935 log_error("Thin pool node must have only one segment.");
2939 seg
= dm_list_item(dm_list_last(&node
->props
.segs
), struct load_segment
);
2940 if (seg
->type
!= SEG_THIN_POOL
) {
2941 log_error("Thin pool node has segment type %s.",
2942 dm_segtypes
[seg
->type
].target
);
2946 if (!(tm
= dm_pool_zalloc(node
->dtree
->mem
, sizeof (*tm
)))) {
2947 log_error("Failed to allocate thin message.");
2952 case DM_THIN_MESSAGE_CREATE_SNAP
:
2953 /* If the thin origin is active, it must be suspend first! */
2955 log_error("Cannot use same device id for origin and its snapshot.");
2958 if (!_thin_validate_device_id(id1
) ||
2959 !_thin_validate_device_id(id2
))
2961 tm
->message
.u
.m_create_snap
.device_id
= id1
;
2962 tm
->message
.u
.m_create_snap
.origin_id
= id2
;
2964 case DM_THIN_MESSAGE_CREATE_THIN
:
2965 if (!_thin_validate_device_id(id1
))
2967 tm
->message
.u
.m_create_thin
.device_id
= id1
;
2968 tm
->expected_errno
= EEXIST
;
2970 case DM_THIN_MESSAGE_DELETE
:
2971 if (!_thin_validate_device_id(id1
))
2973 tm
->message
.u
.m_delete
.device_id
= id1
;
2974 tm
->expected_errno
= ENODATA
;
2976 case DM_THIN_MESSAGE_TRIM
:
2977 if (!_thin_validate_device_id(id1
))
2979 tm
->message
.u
.m_trim
.device_id
= id1
;
2980 tm
->message
.u
.m_trim
.new_size
= id2
;
2982 case DM_THIN_MESSAGE_SET_TRANSACTION_ID
:
2983 if ((id1
+ 1) != id2
) {
2984 log_error("New transaction id must be sequential.");
2985 return 0; /* FIXME: Maybe too strict here? */
2987 if (id2
!= seg
->transaction_id
) {
2988 log_error("Current transaction id is different from thin pool.");
2989 return 0; /* FIXME: Maybe too strict here? */
2991 tm
->message
.u
.m_set_transaction_id
.current_id
= id1
;
2992 tm
->message
.u
.m_set_transaction_id
.new_id
= id2
;
2995 log_error("Unsupported message type %d.", (int) type
);
2999 tm
->message
.type
= type
;
3000 dm_list_add(&seg
->thin_messages
, &tm
->list
);
3005 int dm_tree_node_add_thin_target(struct dm_tree_node
*node
,
3007 const char *pool_uuid
,
3010 struct dm_tree_node
*pool
;
3011 struct load_segment
*seg
;
3013 if (!(pool
= dm_tree_find_node_by_uuid(node
->dtree
, pool_uuid
))) {
3014 log_error("Missing thin pool uuid %s.", pool_uuid
);
3018 if (!_link_tree_nodes(node
, pool
))
3021 if (device_id
== DM_THIN_ERROR_DEVICE_ID
) {
3022 if (!dm_tree_node_add_error_target(node
, size
))
3025 if (!_thin_validate_device_id(device_id
))
3028 if (!(seg
= _add_segment(node
, SEG_THIN
, size
)))
3032 seg
->device_id
= device_id
;
3038 static int _add_area(struct dm_tree_node
*node
, struct load_segment
*seg
, struct dm_tree_node
*dev_node
, uint64_t offset
)
3040 struct seg_area
*area
;
3042 if (!(area
= dm_pool_zalloc(node
->dtree
->mem
, sizeof (*area
)))) {
3043 log_error("Failed to allocate target segment area.");
3047 area
->dev_node
= dev_node
;
3048 area
->offset
= offset
;
3050 dm_list_add(&seg
->areas
, &area
->list
);
3056 int dm_tree_node_add_target_area(struct dm_tree_node
*node
,
3057 const char *dev_name
,
3061 struct load_segment
*seg
;
3063 struct dm_tree_node
*dev_node
;
3065 if ((!dev_name
|| !*dev_name
) && (!uuid
|| !*uuid
)) {
3066 log_error("dm_tree_node_add_target_area called without device");
3071 if (!(dev_node
= dm_tree_find_node_by_uuid(node
->dtree
, uuid
))) {
3072 log_error("Couldn't find area uuid %s.", uuid
);
3075 if (!_link_tree_nodes(node
, dev_node
))
3078 if (stat(dev_name
, &info
) < 0) {
3079 log_error("Device %s not found.", dev_name
);
3083 if (!S_ISBLK(info
.st_mode
)) {
3084 log_error("Device %s is not a block device.", dev_name
);
3088 /* FIXME Check correct macro use */
3089 if (!(dev_node
= _add_dev(node
->dtree
, node
, MAJOR(info
.st_rdev
),
3090 MINOR(info
.st_rdev
), 0)))
3094 if (!node
->props
.segment_count
) {
3095 log_error(INTERNAL_ERROR
"Attempt to add target area to missing segment.");
3099 seg
= dm_list_item(dm_list_last(&node
->props
.segs
), struct load_segment
);
3101 if (!_add_area(node
, seg
, dev_node
, offset
))
3107 int dm_tree_node_add_null_area(struct dm_tree_node
*node
, uint64_t offset
)
3109 struct load_segment
*seg
;
3111 seg
= dm_list_item(dm_list_last(&node
->props
.segs
), struct load_segment
);
3113 switch (seg
->type
) {
3125 log_error("dm_tree_node_add_null_area() called on an unsupported segment type");
3129 if (!_add_area(node
, seg
, NULL
, offset
))
3135 void dm_tree_set_cookie(struct dm_tree_node
*node
, uint32_t cookie
)
3137 node
->dtree
->cookie
= cookie
;
3140 uint32_t dm_tree_get_cookie(struct dm_tree_node
*node
)
3142 return node
->dtree
->cookie
;