1
0
mirror of git://sourceware.org/git/lvm2.git synced 2024-12-22 17:35:59 +03:00
lvm2/libdm/libdm-deptree.c

3989 lines
102 KiB
C
Raw Normal View History

/*
* Copyright (C) 2005-2017 Red Hat, Inc. All rights reserved.
*
* This file is part of the device-mapper userspace tools.
*
* This copyrighted material is made available to anyone wishing to use,
* modify, copy, or redistribute it subject to the terms and conditions
* of the GNU Lesser General Public License v.2.1.
*
* You should have received a copy of the GNU Lesser General Public License
* along with this program; if not, write to the Free Software Foundation,
* Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "libdm/misc/dmlib.h"
#include "libdm-targets.h"
#include "libdm-common.h"
#include "libdm/misc/kdev_t.h"
#include "libdm/misc/dm-ioctl.h"
#include <stdarg.h>
#include <sys/utsname.h>
#define MAX_TARGET_PARAMSIZE 500000
/* Supported segment types */
enum {
SEG_CACHE,
SEG_CRYPT,
SEG_ERROR,
SEG_LINEAR,
SEG_MIRRORED,
SEG_SNAPSHOT,
SEG_SNAPSHOT_ORIGIN,
SEG_SNAPSHOT_MERGE,
SEG_STRIPED,
SEG_ZERO,
SEG_THIN_POOL,
SEG_THIN,
SEG_RAID0,
2016-07-02 00:20:54 +03:00
SEG_RAID0_META,
SEG_RAID1,
SEG_RAID10,
SEG_RAID4,
SEG_RAID5_N,
SEG_RAID5_LA,
SEG_RAID5_RA,
SEG_RAID5_LS,
SEG_RAID5_RS,
SEG_RAID6_N_6,
SEG_RAID6_ZR,
SEG_RAID6_NR,
SEG_RAID6_NC,
SEG_RAID6_LS_6,
SEG_RAID6_RS_6,
SEG_RAID6_LA_6,
SEG_RAID6_RA_6,
};
2005-11-09 17:10:50 +03:00
/* FIXME Add crypt and multipath support */
static const struct {
unsigned type;
const char target[16];
} _dm_segtypes[] = {
{ SEG_CACHE, "cache" },
{ SEG_CRYPT, "crypt" },
{ SEG_ERROR, "error" },
{ SEG_LINEAR, "linear" },
{ SEG_MIRRORED, "mirror" },
{ SEG_SNAPSHOT, "snapshot" },
{ SEG_SNAPSHOT_ORIGIN, "snapshot-origin" },
{ SEG_SNAPSHOT_MERGE, "snapshot-merge" },
{ SEG_STRIPED, "striped" },
{ SEG_ZERO, "zero"},
{ SEG_THIN_POOL, "thin-pool"},
{ SEG_THIN, "thin"},
{ SEG_RAID0, "raid0"},
2016-07-02 00:20:54 +03:00
{ SEG_RAID0_META, "raid0_meta"},
{ SEG_RAID1, "raid1"},
{ SEG_RAID10, "raid10"},
{ SEG_RAID4, "raid4"},
{ SEG_RAID5_N, "raid5_n"},
{ SEG_RAID5_LA, "raid5_la"},
{ SEG_RAID5_RA, "raid5_ra"},
{ SEG_RAID5_LS, "raid5_ls"},
{ SEG_RAID5_RS, "raid5_rs"},
{ SEG_RAID6_N_6,"raid6_n_6"},
{ SEG_RAID6_ZR, "raid6_zr"},
{ SEG_RAID6_NR, "raid6_nr"},
{ SEG_RAID6_NC, "raid6_nc"},
{ SEG_RAID6_LS_6, "raid6_ls_6"},
{ SEG_RAID6_RS_6, "raid6_rs_6"},
{ SEG_RAID6_LA_6, "raid6_la_6"},
{ SEG_RAID6_RA_6, "raid6_ra_6"},
/*
* WARNING: Since 'raid' target overloads this 1:1 mapping table
* for search do not add new enum elements past them!
*/
{ SEG_RAID5_LS, "raid5"}, /* same as "raid5_ls" (default for MD also) */
{ SEG_RAID6_ZR, "raid6"}, /* same as "raid6_zr" */
{ SEG_RAID10, "raid10_near"}, /* same as "raid10" */
};
/* Some segment types have a list of areas of other devices attached */
struct seg_area {
struct dm_list list;
2005-11-09 17:10:50 +03:00
struct dm_tree_node *dev_node;
uint64_t offset;
};
struct dm_thin_message {
dm_thin_message_t type;
union {
struct {
uint32_t device_id;
uint32_t origin_id;
} m_create_snap;
struct {
uint32_t device_id;
} m_create_thin;
struct {
uint32_t device_id;
} m_delete;
struct {
uint64_t current_id;
uint64_t new_id;
} m_set_transaction_id;
} u;
};
struct thin_message {
struct dm_list list;
struct dm_thin_message message;
int expected_errno;
};
/* Per-segment properties */
struct load_segment {
struct dm_list list;
unsigned type;
uint64_t size;
unsigned area_count; /* Linear + Striped + Mirrored + Crypt */
struct dm_list areas; /* Linear + Striped + Mirrored + Crypt */
uint32_t stripe_size; /* Striped + raid */
int persistent; /* Snapshot */
uint32_t chunk_size; /* Snapshot */
2005-11-09 17:10:50 +03:00
struct dm_tree_node *cow; /* Snapshot */
struct dm_tree_node *origin; /* Snapshot + Snapshot origin + Cache */
struct dm_tree_node *merge; /* Snapshot */
struct dm_tree_node *log; /* Mirror */
uint32_t region_size; /* Mirror + raid */
unsigned clustered; /* Mirror */
unsigned mirror_area_count; /* Mirror */
uint32_t flags; /* Mirror + raid + Cache */
2006-02-06 23:18:10 +03:00
char *uuid; /* Clustered mirror log */
const char *policy_name; /* Cache */
unsigned policy_argc; /* Cache */
struct dm_config_node *policy_settings; /* Cache */
const char *cipher; /* Crypt */
const char *chainmode; /* Crypt */
const char *iv; /* Crypt */
uint64_t iv_offset; /* Crypt */
const char *key; /* Crypt */
lvconvert: add infrastructure for RaidLV reshaping support In order to support striped raid5/6/10 LV reshaping (change of LV type, stripesize or number of legs), this patch introduces infrastructure prerequisites to be used by raid_manip.c extensions in followup patches. This base is needed for allocation of out-of-place reshape space required by the MD raid personalities to avoid writing over data in-place when reading off the current RAID layout or number of legs and writing out the new layout or to a different number of legs (i.e. restripe) Changes: - add members reshape_len to 'struct lv_segment' to store out-of-place reshape length per component rimage - add member data_copies to struct lv_segment to support more than 2 raid10 data copies - make alloc_lv_segment() aware of both reshape_len and data_copies - adjust all alloc_lv_segment() callers to the new API - add functions to retrieve the current data offset (needed for out-of-place reshaping space allocation) and the devices count from the kernel - make libdm deptree code aware of reshape_len - add LV flags for disk add/remove reshaping - support import/export of the new 'struct lv_segment' members - enhance lv_extend/_lv_reduce to cope with reshape_len - add seg_is_*/segtype_is_* macros related to reshaping - add target version check for reshaping - grow rebuilds/writemostly bitmaps to 246 bit to support kernel maximal - enhance libdm deptree code to support data_offset (out-of-place reshaping) and delta_disk (legs add/remove reshaping) target arguments Related: rhbz834579 Related: rhbz1191935 Related: rhbz1191978
2017-02-24 02:50:00 +03:00
int delta_disks; /* raid reshape number of disks */
int data_offset; /* raid reshape data offset on disk to set */
uint64_t rebuilds[RAID_BITMAP_SIZE]; /* raid */
uint64_t writemostly[RAID_BITMAP_SIZE]; /* raid */
uint32_t writebehind; /* raid */
uint32_t max_recovery_rate; /* raid kB/sec/disk */
uint32_t min_recovery_rate; /* raid kB/sec/disk */
lvconvert: add infrastructure for RaidLV reshaping support In order to support striped raid5/6/10 LV reshaping (change of LV type, stripesize or number of legs), this patch introduces infrastructure prerequisites to be used by raid_manip.c extensions in followup patches. This base is needed for allocation of out-of-place reshape space required by the MD raid personalities to avoid writing over data in-place when reading off the current RAID layout or number of legs and writing out the new layout or to a different number of legs (i.e. restripe) Changes: - add members reshape_len to 'struct lv_segment' to store out-of-place reshape length per component rimage - add member data_copies to struct lv_segment to support more than 2 raid10 data copies - make alloc_lv_segment() aware of both reshape_len and data_copies - adjust all alloc_lv_segment() callers to the new API - add functions to retrieve the current data offset (needed for out-of-place reshaping space allocation) and the devices count from the kernel - make libdm deptree code aware of reshape_len - add LV flags for disk add/remove reshaping - support import/export of the new 'struct lv_segment' members - enhance lv_extend/_lv_reduce to cope with reshape_len - add seg_is_*/segtype_is_* macros related to reshaping - add target version check for reshaping - grow rebuilds/writemostly bitmaps to 246 bit to support kernel maximal - enhance libdm deptree code to support data_offset (out-of-place reshaping) and delta_disk (legs add/remove reshaping) target arguments Related: rhbz834579 Related: rhbz1191935 Related: rhbz1191978
2017-02-24 02:50:00 +03:00
uint32_t data_copies; /* raid10 data_copies */
struct dm_tree_node *metadata; /* Thin_pool + Cache */
struct dm_tree_node *pool; /* Thin_pool, Thin */
struct dm_tree_node *external; /* Thin */
struct dm_list thin_messages; /* Thin_pool */
uint64_t transaction_id; /* Thin_pool */
uint64_t low_water_mark; /* Thin_pool */
uint32_t data_block_size; /* Thin_pool + cache */
uint32_t migration_threshold; /* Cache */
unsigned skip_block_zeroing; /* Thin_pool */
unsigned ignore_discard; /* Thin_pool target vsn 1.1 */
unsigned no_discard_passdown; /* Thin_pool target vsn 1.1 */
unsigned error_if_no_space; /* Thin pool target vsn 1.10 */
unsigned read_only; /* Thin pool target vsn 1.3 */
uint32_t device_id; /* Thin */
};
/* Per-device properties */
struct load_properties {
int read_only;
uint32_t major;
uint32_t minor;
uint32_t read_ahead;
uint32_t read_ahead_flags;
unsigned segment_count;
int size_changed;
struct dm_list segs;
const char *new_name;
/* If immediate_dev_node is set to 1, try to create the dev node
* as soon as possible (e.g. in preload stage even during traversal
* and processing of dm tree). This will also flush all stacked dev
* node operations, synchronizing with udev.
*/
unsigned immediate_dev_node;
/*
* If the device size changed from zero and this is set,
* don't resume the device immediately, even if the device
* has parents. This works provided the parents do not
* validate the device size and is required by pvmove to
* avoid starting the mirror resync operation too early.
*/
unsigned delay_resume_if_new;
/*
* Preload tree normally only loads and not resume, but there is
* automatic resume when target is extended, as it's believed
* there can be no i/o flying to this 'new' extedend space
* from any device above. Reason is that preloaded target above
* may actually need to see its bigger subdevice before it
* gets suspended. As long as devices are simple linears
* there is no problem to resume bigger device in preload (before commit).
* However complex targets like thin-pool (raid,cache...)
* they shall not be resumed before their commit.
*/
unsigned delay_resume_if_extended;
thin: move pool messaging from resume to suspend Existing messaging intarface for thin-pool has a few 'weak' points: * Message were posted with each 'resume' operation, thus not allowing activation of thin-pool with the existing state. * Acceleration skipped suspend step has not worked in cluster, since clvmd resumes only nodes which are suspended (have proper lock state). * Resume may fail and code is not really designed to 'fail' in this phase (generic rule here is resume DOES NOT fail unless something serious is wrong and lvm2 tool usually doesn't handle recovery path in this case.) * Full thin-pool suspend happened, when taken a thin-volume snapshot. With this patch the new method relocates message passing into suspend state. This has a few drawbacks with current API, but overal it performs better and gives are more posibilities to deal with errors. Patch introduces a new logic for 'origin-only' suspend of thin-pool and this also relates to thin-volume when taking snapshot. When suspend_origin_only operation is invoked on a pool with queued messages then only those messages are posted to thin-pool and actual suspend of thin pool and data and metadata volume is skipped. This makes taking a snapshot of thin-volume lighter operation and avoids blocking of other unrelated active thin volumes. Also fail now happens in 'suspend' state where the 'Fail' is more expected and it is better handled through error paths. Activation of thin-pool is now not sending any message and leaves upto a tool to decided later how to finish unfinished double-commit transaction. Problem which needs some API improvements relates to the lvm2 tree construction. For the suspend tree we do not add target table line into the tree, but only a device is inserted into a tree. Current mechanism to attach messages for thin-pool requires the libdm to know about thin-pool target, so lvm2 currently takes assumption, node is really a thin-pool and fills in the table line for this node (which should be ensured by the PRELOAD phase, but it's a misuse of internal API) we would possibly need to be able to attach message to 'any' node. Other thing to notice - current messaging interface in thin-pool target requires to suspend thin volume origin first and then send a create message, but this could not have any 'nice' solution on lvm2 side and IMHO we should introduce something like 'create_after_resume' message. Patch also changes the moment, where lvm2 transaction id is increased. Now it happens only after successful finish of kernel transaction id change. This change was needed to handle properly activation of pool, which is in the middle of unfinished transaction, and also this corrects usage of thin-pool by external apps like Docker.
2015-07-01 14:31:37 +03:00
/*
* Call node_send_messages(), set to 2 if there are messages
* When != 0, it validates matching transaction id, thus thin-pools
* where transation_id is passed as 0 are never validated, this
* allows external managment of thin-pool TID.
*/
unsigned send_messages;
/* Skip suspending node's children, used when sending messages to thin-pool */
int skip_suspend;
/* Suspend and Resume siblings after node activation with udev flags*/
unsigned reactivate_siblings;
uint16_t reactivate_udev_flags;
};
/* Two of these used to join two nodes with uses and used_by. */
2005-11-09 17:10:50 +03:00
struct dm_tree_link {
struct dm_list list;
2005-11-09 17:10:50 +03:00
struct dm_tree_node *node;
};
2005-11-09 17:10:50 +03:00
struct dm_tree_node {
struct dm_tree *dtree;
2011-08-19 21:02:48 +04:00
const char *name;
const char *uuid;
struct dm_info info;
2011-08-19 21:02:48 +04:00
struct dm_list uses; /* Nodes this node uses */
struct dm_list used_by; /* Nodes that use this node */
2005-11-22 23:00:35 +03:00
int activation_priority; /* 0 gets activated first */
int implicit_deps; /* 1 device only implicitly referenced */
2005-11-22 23:00:35 +03:00
uint16_t udev_flags; /* Udev control flags */
void *context; /* External supplied context */
struct load_properties props; /* For creation/table (re)load */
/*
* If presuspend of child node is needed
* Note: only direct child is allowed
*/
struct dm_tree_node *presuspend_node;
/* Callback */
dm_node_callback_fn callback;
void *callback_data;
int activated; /* tracks activation during preload */
};
2005-11-09 17:10:50 +03:00
struct dm_tree {
2005-10-17 02:57:20 +04:00
struct dm_pool *mem;
struct dm_hash_table *devs;
struct dm_hash_table *uuids;
2005-11-09 17:10:50 +03:00
struct dm_tree_node root;
2006-01-31 02:36:04 +03:00
int skip_lockfs; /* 1 skips lockfs (for non-snapshots) */
int no_flush; /* 1 sets noflush (mirrors/multipath) */
int retry_remove; /* 1 retries remove if not successful */
uint32_t cookie;
char buf[DM_NAME_LEN + 32]; /* print buffer for device_name (major:minor) */
const char **optional_uuid_suffixes; /* uuid suffixes ignored when matching */
};
/*
* Tree functions.
*/
2005-11-09 17:10:50 +03:00
struct dm_tree *dm_tree_create(void)
{
struct dm_pool *dmem;
2005-11-09 17:10:50 +03:00
struct dm_tree *dtree;
if (!(dmem = dm_pool_create("dtree", 1024)) ||
!(dtree = dm_pool_zalloc(dmem, sizeof(*dtree)))) {
log_error("Failed to allocate dtree.");
if (dmem)
dm_pool_destroy(dmem);
return NULL;
}
2005-11-09 17:10:50 +03:00
dtree->root.dtree = dtree;
dm_list_init(&dtree->root.uses);
dm_list_init(&dtree->root.used_by);
2006-01-31 02:36:04 +03:00
dtree->skip_lockfs = 0;
2007-01-09 22:44:07 +03:00
dtree->no_flush = 0;
dtree->mem = dmem;
dtree->optional_uuid_suffixes = NULL;
2005-11-09 17:10:50 +03:00
if (!(dtree->devs = dm_hash_create(8))) {
log_error("dtree hash creation failed");
dm_pool_destroy(dtree->mem);
return NULL;
}
2005-11-09 17:10:50 +03:00
if (!(dtree->uuids = dm_hash_create(32))) {
log_error("dtree uuid hash creation failed");
dm_hash_destroy(dtree->devs);
dm_pool_destroy(dtree->mem);
return NULL;
}
2005-11-09 17:10:50 +03:00
return dtree;
}
2005-11-09 17:10:50 +03:00
void dm_tree_free(struct dm_tree *dtree)
{
2005-11-09 17:10:50 +03:00
if (!dtree)
return;
2005-11-09 17:10:50 +03:00
dm_hash_destroy(dtree->uuids);
dm_hash_destroy(dtree->devs);
dm_pool_destroy(dtree->mem);
}
void dm_tree_set_cookie(struct dm_tree_node *node, uint32_t cookie)
{
node->dtree->cookie = cookie;
}
uint32_t dm_tree_get_cookie(struct dm_tree_node *node)
{
return node->dtree->cookie;
}
void dm_tree_skip_lockfs(struct dm_tree_node *dnode)
{
dnode->dtree->skip_lockfs = 1;
}
void dm_tree_use_no_flush_suspend(struct dm_tree_node *dnode)
{
dnode->dtree->no_flush = 1;
}
void dm_tree_retry_remove(struct dm_tree_node *dnode)
{
dnode->dtree->retry_remove = 1;
}
/*
* Node functions.
*/
static int _nodes_are_linked(const struct dm_tree_node *parent,
const struct dm_tree_node *child)
{
2005-11-09 17:10:50 +03:00
struct dm_tree_link *dlink;
dm_list_iterate_items(dlink, &parent->uses)
if (dlink->node == child)
return 1;
return 0;
}
static int _link(struct dm_list *list, struct dm_tree_node *node)
{
2005-11-09 17:10:50 +03:00
struct dm_tree_link *dlink;
2005-11-09 17:10:50 +03:00
if (!(dlink = dm_pool_alloc(node->dtree->mem, sizeof(*dlink)))) {
log_error("dtree link allocation failed");
return 0;
}
dlink->node = node;
dm_list_add(list, &dlink->list);
return 1;
}
2005-11-09 17:10:50 +03:00
static int _link_nodes(struct dm_tree_node *parent,
struct dm_tree_node *child)
{
if (_nodes_are_linked(parent, child))
return 1;
if (!_link(&parent->uses, child))
return 0;
if (!_link(&child->used_by, parent))
return 0;
return 1;
}
static void _unlink(struct dm_list *list, struct dm_tree_node *node)
{
2005-11-09 17:10:50 +03:00
struct dm_tree_link *dlink;
dm_list_iterate_items(dlink, list)
if (dlink->node == node) {
dm_list_del(&dlink->list);
break;
}
}
2005-11-09 17:10:50 +03:00
static void _unlink_nodes(struct dm_tree_node *parent,
struct dm_tree_node *child)
{
if (!_nodes_are_linked(parent, child))
return;
_unlink(&parent->uses, child);
_unlink(&child->used_by, parent);
}
2005-11-09 17:10:50 +03:00
static int _add_to_toplevel(struct dm_tree_node *node)
{
2005-11-09 17:10:50 +03:00
return _link_nodes(&node->dtree->root, node);
}
2005-11-09 17:10:50 +03:00
static void _remove_from_toplevel(struct dm_tree_node *node)
{
2009-12-11 16:16:37 +03:00
_unlink_nodes(&node->dtree->root, node);
}
2005-11-09 17:10:50 +03:00
static int _add_to_bottomlevel(struct dm_tree_node *node)
{
2005-11-09 17:10:50 +03:00
return _link_nodes(node, &node->dtree->root);
}
2005-11-09 17:10:50 +03:00
static void _remove_from_bottomlevel(struct dm_tree_node *node)
{
2009-12-11 16:16:37 +03:00
_unlink_nodes(node, &node->dtree->root);
}
2005-11-09 17:10:50 +03:00
static int _link_tree_nodes(struct dm_tree_node *parent, struct dm_tree_node *child)
{
/* Don't link to root node if child already has a parent */
if (parent == &parent->dtree->root) {
2005-11-09 17:10:50 +03:00
if (dm_tree_node_num_children(child, 1))
return 1;
} else
_remove_from_toplevel(child);
if (child == &child->dtree->root) {
2005-11-09 17:10:50 +03:00
if (dm_tree_node_num_children(parent, 0))
return 1;
} else
_remove_from_bottomlevel(parent);
return _link_nodes(parent, child);
}
2005-11-09 17:10:50 +03:00
static struct dm_tree_node *_create_dm_tree_node(struct dm_tree *dtree,
const char *name,
const char *uuid,
struct dm_info *info,
void *context,
uint16_t udev_flags)
{
2005-11-09 17:10:50 +03:00
struct dm_tree_node *node;
2012-06-21 14:55:30 +04:00
dev_t dev;
if (!(node = dm_pool_zalloc(dtree->mem, sizeof(*node))) ||
!(node->name = dm_pool_strdup(dtree->mem, name)) ||
!(node->uuid = dm_pool_strdup(dtree->mem, uuid))) {
log_error("_create_dm_tree_node alloc failed.");
return NULL;
}
2005-11-09 17:10:50 +03:00
node->dtree = dtree;
node->info = *info;
node->context = context;
node->udev_flags = udev_flags;
dm_list_init(&node->uses);
dm_list_init(&node->used_by);
dm_list_init(&node->props.segs);
dev = MKDEV(info->major, info->minor);
2005-11-09 17:10:50 +03:00
if (!dm_hash_insert_binary(dtree->devs, (const char *) &dev,
sizeof(dev), node)) {
2005-11-09 17:10:50 +03:00
log_error("dtree node hash insertion failed");
dm_pool_free(dtree->mem, node);
return NULL;
}
if (*uuid && !dm_hash_insert(dtree->uuids, uuid, node)) {
2005-11-09 17:10:50 +03:00
log_error("dtree uuid hash insertion failed");
dm_hash_remove_binary(dtree->devs, (const char *) &dev,
sizeof(dev));
2005-11-09 17:10:50 +03:00
dm_pool_free(dtree->mem, node);
return NULL;
}
return node;
}
2005-11-09 17:10:50 +03:00
static struct dm_tree_node *_find_dm_tree_node(struct dm_tree *dtree,
uint32_t major, uint32_t minor)
{
dev_t dev = MKDEV(major, minor);
2005-11-09 17:10:50 +03:00
return dm_hash_lookup_binary(dtree->devs, (const char *) &dev,
2014-02-14 23:46:55 +04:00
sizeof(dev));
}
void dm_tree_set_optional_uuid_suffixes(struct dm_tree *dtree, const char **optional_uuid_suffixes)
{
dtree->optional_uuid_suffixes = optional_uuid_suffixes;
}
2005-11-09 17:10:50 +03:00
static struct dm_tree_node *_find_dm_tree_node_by_uuid(struct dm_tree *dtree,
const char *uuid)
{
struct dm_tree_node *node;
const char *default_uuid_prefix;
size_t default_uuid_prefix_len;
const char *suffix, *suffix_position;
char uuid_without_suffix[DM_UUID_LEN];
unsigned i = 0;
const char **suffix_list = dtree->optional_uuid_suffixes;
if ((node = dm_hash_lookup(dtree->uuids, uuid))) {
log_debug("Matched uuid %s in deptree.", uuid);
return node;
}
default_uuid_prefix = dm_uuid_prefix();
default_uuid_prefix_len = strlen(default_uuid_prefix);
if (suffix_list && (suffix_position = strrchr(uuid, '-'))) {
while ((suffix = suffix_list[i++])) {
if (strcmp(suffix_position + 1, suffix))
continue;
(void) strncpy(uuid_without_suffix, uuid, sizeof(uuid_without_suffix));
uuid_without_suffix[suffix_position - uuid] = '\0';
if ((node = dm_hash_lookup(dtree->uuids, uuid_without_suffix))) {
log_debug("Matched uuid %s (missing suffix -%s) in deptree.", uuid_without_suffix, suffix);
return node;
}
break;
};
}
if (strncmp(uuid, default_uuid_prefix, default_uuid_prefix_len))
return NULL;
if ((node = dm_hash_lookup(dtree->uuids, uuid + default_uuid_prefix_len))) {
log_debug("Matched uuid %s (missing prefix) in deptree.", uuid + default_uuid_prefix_len);
return node;
}
log_debug("Not matched uuid %s in deptree.", uuid);
return NULL;
}
/* Return node's device_name (major:minor) for debug messages */
static const char *_node_name(struct dm_tree_node *dnode)
{
if (dm_snprintf(dnode->dtree->buf, sizeof(dnode->dtree->buf),
"%s (" FMTu32 ":" FMTu32 ")",
dnode->name ? dnode->name : "",
dnode->info.major, dnode->info.minor) < 0) {
stack;
return dnode->name;
}
return dnode->dtree->buf;
}
void dm_tree_node_set_udev_flags(struct dm_tree_node *dnode, uint16_t udev_flags)
{
if (udev_flags != dnode->udev_flags)
log_debug_activation("Resetting %s udev_flags from 0x%x to 0x%x.",
_node_name(dnode),
dnode->udev_flags, udev_flags);
dnode->udev_flags = udev_flags;
}
void dm_tree_node_set_read_ahead(struct dm_tree_node *dnode,
uint32_t read_ahead,
uint32_t read_ahead_flags)
{
dnode->props.read_ahead = read_ahead;
dnode->props.read_ahead_flags = read_ahead_flags;
}
void dm_tree_node_set_presuspend_node(struct dm_tree_node *node,
struct dm_tree_node *presuspend_node)
{
node->presuspend_node = presuspend_node;
}
const char *dm_tree_node_get_name(const struct dm_tree_node *node)
{
return node->info.exists ? node->name : "";
}
const char *dm_tree_node_get_uuid(const struct dm_tree_node *node)
{
return node->info.exists ? node->uuid : "";
}
const struct dm_info *dm_tree_node_get_info(const struct dm_tree_node *node)
{
return &node->info;
}
void *dm_tree_node_get_context(const struct dm_tree_node *node)
{
return node->context;
}
int dm_tree_node_size_changed(const struct dm_tree_node *dnode)
{
return dnode->props.size_changed;
}
int dm_tree_node_num_children(const struct dm_tree_node *node, uint32_t inverted)
{
if (inverted) {
if (_nodes_are_linked(&node->dtree->root, node))
return 0;
return dm_list_size(&node->used_by);
}
if (_nodes_are_linked(node, &node->dtree->root))
return 0;
return dm_list_size(&node->uses);
}
/*
* Returns 1 if no prefix supplied
*/
static int _uuid_prefix_matches(const char *uuid, const char *uuid_prefix, size_t uuid_prefix_len)
{
const char *default_uuid_prefix = dm_uuid_prefix();
size_t default_uuid_prefix_len = strlen(default_uuid_prefix);
if (!uuid_prefix)
return 1;
if (!strncmp(uuid, uuid_prefix, uuid_prefix_len))
return 1;
/* Handle transition: active device uuids might be missing the prefix */
if (uuid_prefix_len <= 4)
return 0;
if (!strncmp(uuid, default_uuid_prefix, default_uuid_prefix_len))
return 0;
if (strncmp(uuid_prefix, default_uuid_prefix, default_uuid_prefix_len))
return 0;
if (!strncmp(uuid, uuid_prefix + default_uuid_prefix_len, uuid_prefix_len - default_uuid_prefix_len))
return 1;
return 0;
}
/*
* Returns 1 if no children.
*/
static int _children_suspended(struct dm_tree_node *node,
uint32_t inverted,
const char *uuid_prefix,
size_t uuid_prefix_len)
{
struct dm_list *list;
struct dm_tree_link *dlink;
const struct dm_info *dinfo;
const char *uuid;
if (inverted) {
if (_nodes_are_linked(&node->dtree->root, node))
return 1;
list = &node->used_by;
} else {
if (_nodes_are_linked(node, &node->dtree->root))
return 1;
list = &node->uses;
}
dm_list_iterate_items(dlink, list) {
if (!(uuid = dm_tree_node_get_uuid(dlink->node))) {
stack;
continue;
}
/* Ignore if it doesn't belong to this VG */
if (!_uuid_prefix_matches(uuid, uuid_prefix, uuid_prefix_len))
continue;
/* Ignore if parent node wants to presuspend this node */
if (dlink->node->presuspend_node == node)
continue;
if (!(dinfo = dm_tree_node_get_info(dlink->node)))
return_0; /* FIXME Is this normal? */
if (!dinfo->suspended)
return 0;
}
return 1;
}
/*
* Set major and minor to zero for root of tree.
*/
struct dm_tree_node *dm_tree_find_node(struct dm_tree *dtree,
2014-02-14 23:46:55 +04:00
uint32_t major,
uint32_t minor)
{
if (!major && !minor)
return &dtree->root;
return _find_dm_tree_node(dtree, major, minor);
}
/*
* Set uuid to NULL for root of tree.
*/
struct dm_tree_node *dm_tree_find_node_by_uuid(struct dm_tree *dtree,
2014-02-14 23:46:55 +04:00
const char *uuid)
{
if (!uuid || !*uuid)
return &dtree->root;
return _find_dm_tree_node_by_uuid(dtree, uuid);
}
/*
* First time set *handle to NULL.
* Set inverted to invert the tree.
*/
struct dm_tree_node *dm_tree_next_child(void **handle,
const struct dm_tree_node *parent,
uint32_t inverted)
{
struct dm_list **dlink = (struct dm_list **) handle;
const struct dm_list *use_list;
if (inverted)
use_list = &parent->used_by;
else
use_list = &parent->uses;
if (!*dlink)
*dlink = dm_list_first(use_list);
else
*dlink = dm_list_next(use_list, *dlink);
return (*dlink) ? dm_list_item(*dlink, struct dm_tree_link)->node : NULL;
}
2005-10-17 02:57:20 +04:00
static int _deps(struct dm_task **dmt, struct dm_pool *mem, uint32_t major, uint32_t minor,
const char **name, const char **uuid, unsigned inactive_table,
struct dm_info *info, struct dm_deps **deps)
{
memset(info, 0, sizeof(*info));
*name = "";
*uuid = "";
*deps = NULL;
if (!dm_is_dm_major(major)) {
info->major = major;
info->minor = minor;
return 1;
}
2017-12-04 15:31:40 +03:00
if (!(*dmt = dm_task_create(DM_DEVICE_DEPS)))
return_0;
if (!dm_task_set_major(*dmt, major) || !dm_task_set_minor(*dmt, minor)) {
log_error("_deps: failed to set major:minor for (" FMTu32 ":" FMTu32 ").",
2005-11-09 17:10:50 +03:00
major, minor);
goto failed;
2005-11-09 17:10:50 +03:00
}
if (inactive_table && !dm_task_query_inactive_table(*dmt)) {
log_error("_deps: failed to set inactive table for (%" PRIu32 ":%" PRIu32 ")",
major, minor);
goto failed;
}
2005-11-09 17:10:50 +03:00
if (!dm_task_run(*dmt)) {
log_error("_deps: task run failed for (%" PRIu32 ":%" PRIu32 ")",
major, minor);
goto failed;
2005-11-09 17:10:50 +03:00
}
2005-11-09 17:10:50 +03:00
if (!dm_task_get_info(*dmt, info)) {
log_error("_deps: failed to get info for (%" PRIu32 ":%" PRIu32 ")",
major, minor);
goto failed;
2005-11-09 17:10:50 +03:00
}
if (info->exists) {
if (info->major != major) {
2005-11-09 17:10:50 +03:00
log_error("Inconsistent dtree major number: %u != %u",
major, info->major);
goto failed;
}
if (info->minor != minor) {
2005-11-09 17:10:50 +03:00
log_error("Inconsistent dtree minor number: %u != %u",
minor, info->minor);
goto failed;
}
*name = dm_task_get_name(*dmt);
*uuid = dm_task_get_uuid(*dmt);
*deps = dm_task_get_deps(*dmt);
}
return 1;
failed:
dm_task_destroy(*dmt);
*dmt = NULL;
return 0;
}
/*
* Deactivate a device with its dependencies if the uuid prefix matches.
*/
static int _info_by_dev(uint32_t major, uint32_t minor, int with_open_count,
struct dm_info *info, struct dm_pool *mem,
const char **name, const char **uuid)
{
struct dm_task *dmt;
int r = 0;
2017-12-04 15:31:40 +03:00
if (!(dmt = dm_task_create(DM_DEVICE_INFO)))
return_0;
if (!dm_task_set_major(dmt, major) || !dm_task_set_minor(dmt, minor)) {
log_error("_info_by_dev: Failed to set device number.");
goto out;
}
if (!with_open_count && !dm_task_no_open_count(dmt))
log_warn("WARNING: Failed to disable open_count.");
if (!dm_task_run(dmt))
goto_out;
if (!dm_task_get_info(dmt, info))
goto_out;
if (name && !(*name = dm_pool_strdup(mem, dm_task_get_name(dmt)))) {
log_error("name pool_strdup failed");
goto out;
}
if (uuid && !(*uuid = dm_pool_strdup(mem, dm_task_get_uuid(dmt)))) {
log_error("uuid pool_strdup failed");
goto out;
}
r = 1;
out:
dm_task_destroy(dmt);
return r;
}
static int _check_device_not_in_use(const char *name, struct dm_info *info)
{
const char *reason;
if (!info->exists)
return 1;
/* If sysfs is not used, use open_count information only. */
if (!*dm_sysfs_dir()) {
if (!info->open_count)
return 1;
reason = "in use";
} else if (dm_device_has_holders(info->major, info->minor))
reason = "is used by another device";
else if (dm_device_has_mounted_fs(info->major, info->minor))
reason = "constains a filesystem in use";
else
return 1;
log_error("Device %s (" FMTu32 ":" FMTu32 ") %s.",
name, info->major, info->minor, reason);
return 0;
}
/* Check if all parent nodes of given node have open_count == 0 */
static int _node_has_closed_parents(struct dm_tree_node *node,
const char *uuid_prefix,
size_t uuid_prefix_len)
{
struct dm_tree_link *dlink;
const struct dm_info *dinfo;
struct dm_info info;
const char *uuid;
/* Iterate through parents of this node */
dm_list_iterate_items(dlink, &node->used_by) {
if (!(uuid = dm_tree_node_get_uuid(dlink->node))) {
2005-11-09 17:10:50 +03:00
stack;
continue;
2005-11-09 17:10:50 +03:00
}
/* Ignore if it doesn't belong to this VG */
if (!_uuid_prefix_matches(uuid, uuid_prefix, uuid_prefix_len))
continue;
if (!(dinfo = dm_tree_node_get_info(dlink->node)))
return_0; /* FIXME Is this normal? */
/* Refresh open_count */
if (!_info_by_dev(dinfo->major, dinfo->minor, 1, &info, NULL, NULL, NULL))
return_0;
if (!info.exists)
continue;
if (info.open_count) {
log_debug_activation("Node %s %d:%d has open_count %d", uuid_prefix,
dinfo->major, dinfo->minor, info.open_count);
return 0;
}
}
return 1;
}
static int _deactivate_node(const char *name, uint32_t major, uint32_t minor,
uint32_t *cookie, uint16_t udev_flags, int retry)
{
struct dm_task *dmt;
int r = 0;
log_verbose("Removing %s (%" PRIu32 ":%" PRIu32 ")", name, major, minor);
if (!(dmt = dm_task_create(DM_DEVICE_REMOVE))) {
log_error("Deactivation dm_task creation failed for %s", name);
return 0;
}
if (!dm_task_set_major(dmt, major) || !dm_task_set_minor(dmt, minor)) {
log_error("Failed to set device number for %s deactivation", name);
goto out;
}
if (!dm_task_no_open_count(dmt))
log_warn("WARNING: Failed to disable open_count.");
if (cookie)
if (!dm_task_set_cookie(dmt, cookie, udev_flags))
goto out;
if (retry)
dm_task_retry_remove(dmt);
r = dm_task_run(dmt);
/* FIXME Until kernel returns actual name so dm-iface.c can handle it */
rm_dev_node(name, dmt->cookie_set && !(udev_flags & DM_UDEV_DISABLE_DM_RULES_FLAG),
dmt->cookie_set && (udev_flags & DM_UDEV_DISABLE_LIBRARY_FALLBACK));
/* FIXME Remove node from tree or mark invalid? */
out:
dm_task_destroy(dmt);
return r;
}
static int _node_clear_table(struct dm_tree_node *dnode, uint16_t udev_flags)
{
struct dm_task *dmt = NULL, *deps_dmt = NULL;
struct dm_info *info = &dnode->info, deps_info;
struct dm_deps *deps = NULL;
const char *name, *uuid, *depname, *depuuid;
const char *default_uuid_prefix;
size_t default_uuid_prefix_len;
uint32_t i;
int r = 0;
2005-11-09 17:10:50 +03:00
if (!(name = dm_tree_node_get_name(dnode))) {
log_error("_node_clear_table failed: missing name");
return 0;
}
/* Is there a table? */
if (!info->exists || !info->inactive_table)
return 1;
/* Get devices used by inactive table that's about to be deleted. */
if (!_deps(&deps_dmt, dnode->dtree->mem, info->major, info->minor, &depname, &depuuid, 1, info, &deps)) {
log_error("Failed to obtain dependencies for %s before clearing table.", name);
return 0;
}
log_verbose("Clearing inactive table %s (%" PRIu32 ":%" PRIu32 ")",
name, info->major, info->minor);
if (!(dmt = dm_task_create(DM_DEVICE_CLEAR))) {
log_error("Table clear dm_task creation failed for %s", name);
goto out;
}
if (!dm_task_set_major(dmt, info->major) ||
!dm_task_set_minor(dmt, info->minor)) {
log_error("Failed to set device number for %s table clear", name);
goto out;
}
r = dm_task_run(dmt);
if (!dm_task_get_info(dmt, info)) {
2005-11-09 17:10:50 +03:00
log_error("_node_clear_table failed: info missing after running task for %s", name);
r = 0;
}
if (!r || !deps)
goto_out;
/*
* Remove (incomplete) devices that the inactive table referred to but
* which are not in the tree, no longer referenced and don't have a live
* table.
*/
default_uuid_prefix = dm_uuid_prefix();
default_uuid_prefix_len = strlen(default_uuid_prefix);
for (i = 0; i < deps->count; i++) {
/* If already in tree, assume it's under control */
if (_find_dm_tree_node(dnode->dtree, MAJOR(deps->device[i]), MINOR(deps->device[i])))
continue;
2005-10-25 23:09:41 +04:00
if (!_info_by_dev(MAJOR(deps->device[i]), MINOR(deps->device[i]), 1,
&deps_info, dnode->dtree->mem, &name, &uuid))
goto_out;
/* Proceed if device is an 'orphan' - unreferenced and without a live table. */
if (!deps_info.exists || deps_info.live_table || deps_info.open_count)
continue;
2005-10-18 16:37:53 +04:00
if (strncmp(uuid, default_uuid_prefix, default_uuid_prefix_len))
continue;
/* Remove device. */
if (!_deactivate_node(name, deps_info.major, deps_info.minor, &dnode->dtree->cookie, udev_flags, 0)) {
log_error("Failed to deactivate no-longer-used device %s (%"
PRIu32 ":%" PRIu32 ")", name, deps_info.major, deps_info.minor);
} else if (deps_info.suspended)
dec_suspended();
}
out:
if (dmt)
dm_task_destroy(dmt);
if (deps_dmt)
dm_task_destroy(deps_dmt);
2005-10-18 16:37:53 +04:00
return r;
}
struct dm_tree_node *dm_tree_add_new_dev_with_udev_flags(struct dm_tree *dtree,
const char *name,
const char *uuid,
uint32_t major,
uint32_t minor,
int read_only,
int clear_inactive,
void *context,
uint16_t udev_flags)
{
struct dm_tree_node *dnode;
struct dm_info info = { 0 };
if (!name || !uuid) {
log_error("Cannot add device without name and uuid.");
return NULL;
}
/* Do we need to add node to tree? */
if (!(dnode = dm_tree_find_node_by_uuid(dtree, uuid))) {
if (!(dnode = _create_dm_tree_node(dtree, name, uuid, &info,
context, 0)))
return_NULL;
/* Attach to root node until a table is supplied */
if (!_add_to_toplevel(dnode) || !_add_to_bottomlevel(dnode))
return_NULL;
dnode->props.major = major;
dnode->props.minor = minor;
} else if (strcmp(name, dnode->name)) {
/* Do we need to rename node? */
if (!(dnode->props.new_name = dm_pool_strdup(dtree->mem, name))) {
log_error("name pool_strdup failed");
return NULL;
}
}
dnode->props.read_only = read_only ? 1 : 0;
dnode->props.read_ahead = DM_READ_AHEAD_AUTO;
dnode->props.read_ahead_flags = 0;
if (clear_inactive && !_node_clear_table(dnode, udev_flags))
return_NULL;
dnode->context = context;
dnode->udev_flags = udev_flags;
return dnode;
}
struct dm_tree_node *dm_tree_add_new_dev(struct dm_tree *dtree, const char *name,
const char *uuid, uint32_t major, uint32_t minor,
int read_only, int clear_inactive, void *context)
{
return dm_tree_add_new_dev_with_udev_flags(dtree, name, uuid, major, minor,
read_only, clear_inactive, context, 0);
}
static struct dm_tree_node *_add_dev(struct dm_tree *dtree,
struct dm_tree_node *parent,
uint32_t major, uint32_t minor,
uint16_t udev_flags,
int implicit_deps)
2005-10-18 16:37:53 +04:00
{
struct dm_task *dmt = NULL;
struct dm_info info;
struct dm_deps *deps = NULL;
const char *name = NULL;
const char *uuid = NULL;
struct dm_tree_node *node = NULL;
uint32_t i;
int new = 0;
2005-10-18 16:37:53 +04:00
/* Already in tree? */
if (!(node = _find_dm_tree_node(dtree, major, minor))) {
if (!_deps(&dmt, dtree->mem, major, minor, &name, &uuid, 0, &info, &deps))
return_NULL;
2005-10-18 16:37:53 +04:00
if (!(node = _create_dm_tree_node(dtree, name, uuid, &info,
NULL, udev_flags)))
goto_out;
new = 1;
node->implicit_deps = implicit_deps;
} else if (!implicit_deps && node->implicit_deps) {
node->udev_flags = udev_flags;
node->implicit_deps = 0;
2005-10-18 16:37:53 +04:00
}
if (!_link_tree_nodes(parent, node)) {
node = NULL;
goto_out;
2005-10-18 16:37:53 +04:00
}
/* If node was already in tree, no need to recurse. */
if (!new)
goto out;
/* Can't recurse if not a mapped device or there are no dependencies */
if (!node->info.exists || !deps || !deps->count) {
if (!_add_to_bottomlevel(node)) {
stack;
node = NULL;
}
goto out;
}
/* Add dependencies to tree */
for (i = 0; i < deps->count; i++)
/* Implicit devices are by default temporary */
if (!_add_dev(dtree, node, MAJOR(deps->device[i]),
MINOR(deps->device[i]), udev_flags |
DM_UDEV_DISABLE_SUBSYSTEM_RULES_FLAG |
DM_UDEV_DISABLE_DISK_RULES_FLAG |
DM_UDEV_DISABLE_OTHER_RULES_FLAG, 1)) {
node = NULL;
goto_out;
}
2005-10-18 16:37:53 +04:00
out:
if (dmt)
dm_task_destroy(dmt);
return node;
}
2005-10-25 23:09:41 +04:00
int dm_tree_add_dev(struct dm_tree *dtree, uint32_t major, uint32_t minor)
{
return _add_dev(dtree, &dtree->root, major, minor, 0, 0) ? 1 : 0;
}
2005-10-25 23:09:41 +04:00
int dm_tree_add_dev_with_udev_flags(struct dm_tree *dtree, uint32_t major,
uint32_t minor, uint16_t udev_flags)
{
return _add_dev(dtree, &dtree->root, major, minor, udev_flags, 0) ? 1 : 0;
2005-10-25 23:09:41 +04:00
}
static int _rename_node(const char *old_name, const char *new_name, uint32_t major,
uint32_t minor, uint32_t *cookie, uint16_t udev_flags)
{
struct dm_task *dmt;
int r = 0;
log_verbose("Renaming %s (%" PRIu32 ":%" PRIu32 ") to %s", old_name, major, minor, new_name);
if (!(dmt = dm_task_create(DM_DEVICE_RENAME))) {
log_error("Rename dm_task creation failed for %s", old_name);
return 0;
}
if (!dm_task_set_name(dmt, old_name)) {
log_error("Failed to set name for %s rename.", old_name);
goto out;
}
2005-11-09 17:10:50 +03:00
if (!dm_task_set_newname(dmt, new_name))
2011-08-19 21:02:48 +04:00
goto_out;
if (!dm_task_no_open_count(dmt))
log_warn("WARNING: Failed to disable open_count.");
if (!dm_task_set_cookie(dmt, cookie, udev_flags))
goto out;
r = dm_task_run(dmt);
out:
dm_task_destroy(dmt);
return r;
}
/* FIXME Merge with _suspend_node? */
static int _resume_node(const char *name, uint32_t major, uint32_t minor,
uint32_t read_ahead, uint32_t read_ahead_flags,
struct dm_info *newinfo, uint32_t *cookie,
uint16_t udev_flags, int already_suspended)
{
struct dm_task *dmt;
int r = 0;
2017-12-04 16:37:01 +03:00
log_verbose("Resuming %s (" FMTu32 ":" FMTu32 ").", name, major, minor);
if (!(dmt = dm_task_create(DM_DEVICE_RESUME))) {
log_debug_activation("Suspend dm_task creation failed for %s.", name);
return 0;
}
2007-12-04 01:48:36 +03:00
/* FIXME Kernel should fill in name on return instead */
if (!dm_task_set_name(dmt, name)) {
log_debug_activation("Failed to set device name for %s resumption.", name);
goto out;
2007-12-04 01:48:36 +03:00
}
if (!dm_task_set_major(dmt, major) || !dm_task_set_minor(dmt, minor)) {
log_error("Failed to set device number for %s resumption.", name);
goto out;
}
if (!dm_task_no_open_count(dmt))
log_warn("WARNING: Failed to disable open_count.");
if (!dm_task_set_read_ahead(dmt, read_ahead, read_ahead_flags))
log_warn("WARNING: Failed to set read ahead.");
if (!dm_task_set_cookie(dmt, cookie, udev_flags))
goto_out;
if (!(r = dm_task_run(dmt)))
goto_out;
if (already_suspended)
dec_suspended();
if (!(r = dm_task_get_info(dmt, newinfo)))
stack;
out:
dm_task_destroy(dmt);
return r;
}
2005-10-25 23:09:41 +04:00
static int _suspend_node(const char *name, uint32_t major, uint32_t minor,
2007-01-09 22:44:07 +03:00
int skip_lockfs, int no_flush, struct dm_info *newinfo)
2005-10-25 23:09:41 +04:00
{
struct dm_task *dmt;
int r = 0;
2005-10-25 23:09:41 +04:00
2007-01-09 22:44:07 +03:00
log_verbose("Suspending %s (%" PRIu32 ":%" PRIu32 ")%s%s",
name, major, minor,
skip_lockfs ? "" : " with filesystem sync",
2008-06-25 18:24:17 +04:00
no_flush ? "" : " with device flush");
2005-10-25 23:09:41 +04:00
if (!(dmt = dm_task_create(DM_DEVICE_SUSPEND))) {
log_error("Suspend dm_task creation failed for %s", name);
return 0;
}
if (!dm_task_set_major(dmt, major) || !dm_task_set_minor(dmt, minor)) {
log_error("Failed to set device number for %s suspension.", name);
goto out;
2005-10-25 23:09:41 +04:00
}
if (!dm_task_no_open_count(dmt))
log_warn("WARNING: Failed to disable open_count.");
2005-10-25 23:09:41 +04:00
2006-01-31 02:36:04 +03:00
if (skip_lockfs && !dm_task_skip_lockfs(dmt))
log_warn("WARNING: Failed to set skip_lockfs flag.");
2006-01-31 02:36:04 +03:00
2007-01-09 22:44:07 +03:00
if (no_flush && !dm_task_no_flush(dmt))
log_warn("WARNING: Failed to set no_flush flag.");
2007-01-09 22:44:07 +03:00
if ((r = dm_task_run(dmt))) {
inc_suspended();
2005-10-25 23:09:41 +04:00
r = dm_task_get_info(dmt, newinfo);
}
out:
2005-10-18 16:37:53 +04:00
dm_task_destroy(dmt);
return r;
}
static int _thin_pool_get_status(struct dm_tree_node *dnode,
struct dm_status_thin_pool *s)
{
struct dm_task *dmt;
int r = 0;
uint64_t start, length;
char *type = NULL;
char *params = NULL;
if (!(dmt = dm_task_create(DM_DEVICE_STATUS)))
return_0;
if (!dm_task_set_major(dmt, dnode->info.major) ||
!dm_task_set_minor(dmt, dnode->info.minor)) {
log_error("Failed to set major minor.");
goto out;
}
if (!dm_task_no_flush(dmt))
log_warn("WARNING: Can't set no_flush flag."); /* Non fatal */
if (!dm_task_run(dmt))
goto_out;
dm_get_next_target(dmt, NULL, &start, &length, &type, &params);
2017-12-06 15:21:01 +03:00
if (!type || (strcmp(type, "thin-pool") != 0)) {
log_error("Expected thin-pool target for %s and got %s.",
2017-12-06 15:21:01 +03:00
_node_name(dnode), type ? : "no target");
goto out;
}
if (!parse_thin_pool_status(params, s))
goto_out;
log_debug_activation("Found transaction id %" PRIu64 " for thin pool %s "
"with status line: %s.",
s->transaction_id, _node_name(dnode), params);
r = 1;
out:
dm_task_destroy(dmt);
return r;
}
static int _node_message(uint32_t major, uint32_t minor,
int expected_errno, const char *message)
{
struct dm_task *dmt;
int r = 0;
if (!(dmt = dm_task_create(DM_DEVICE_TARGET_MSG)))
return_0;
if (!dm_task_set_major(dmt, major) ||
!dm_task_set_minor(dmt, minor)) {
log_error("Failed to set message major minor.");
goto out;
}
if (!dm_task_set_message(dmt, message))
goto_out;
/* Internal functionality of dm_task */
dmt->expected_errno = expected_errno;
if (!dm_task_run(dmt)) {
log_error("Failed to process message \"%s\".", message);
goto out;
}
r = 1;
out:
dm_task_destroy(dmt);
return r;
}
static int _thin_pool_node_message(struct dm_tree_node *dnode, struct thin_message *tm)
{
struct dm_thin_message *m = &tm->message;
char buf[64];
int r;
switch (m->type) {
case DM_THIN_MESSAGE_CREATE_SNAP:
r = dm_snprintf(buf, sizeof(buf), "create_snap %u %u",
m->u.m_create_snap.device_id,
m->u.m_create_snap.origin_id);
break;
case DM_THIN_MESSAGE_CREATE_THIN:
r = dm_snprintf(buf, sizeof(buf), "create_thin %u",
m->u.m_create_thin.device_id);
break;
case DM_THIN_MESSAGE_DELETE:
r = dm_snprintf(buf, sizeof(buf), "delete %u",
m->u.m_delete.device_id);
break;
case DM_THIN_MESSAGE_SET_TRANSACTION_ID:
r = dm_snprintf(buf, sizeof(buf),
"set_transaction_id %" PRIu64 " %" PRIu64,
m->u.m_set_transaction_id.current_id,
m->u.m_set_transaction_id.new_id);
break;
case DM_THIN_MESSAGE_RESERVE_METADATA_SNAP: /* target vsn 1.1 */
r = dm_snprintf(buf, sizeof(buf), "reserve_metadata_snap");
break;
case DM_THIN_MESSAGE_RELEASE_METADATA_SNAP: /* target vsn 1.1 */
r = dm_snprintf(buf, sizeof(buf), "release_metadata_snap");
break;
default:
r = -1;
}
if (r < 0) {
log_error("Failed to prepare message.");
return 0;
}
if (!_node_message(dnode->info.major, dnode->info.minor,
2020-09-25 20:10:30 +03:00
tm->expected_errno, buf)) {
switch (m->type) {
case DM_THIN_MESSAGE_CREATE_SNAP:
case DM_THIN_MESSAGE_CREATE_THIN:
if (errno == EEXIST) {
/*
* ATM errno from ioctl() is preserved through code error path chain
* If this would ever change, another way need to be used to
* obtain result from failed DM message
*/
log_error("Thin pool %s already contain thin device with device_id %u.",
_node_name(dnode), m->u.m_create_snap.device_id);
/*
* TODO:
*
* Give some useful advice how to solve this problem,
* until lvconvert --repair can handle this automatically
*/
log_error("Manual intervention may be required to remove device dev_id=%u in thin pool metadata.",
m->u.m_create_snap.device_id);
log_error("Optionally new thin volume with device_id=%u can be manually added into a volume group.",
m->u.m_create_snap.device_id);
log_warn("WARNING: When uncertain how to do this, contact support!");
return 0;
}
/* fall through */
default:
return_0;
}
}
return 1;
}
static struct load_segment *_get_last_load_segment(struct dm_tree_node *node)
{
if (dm_list_empty(&node->props.segs)) {
log_error("Node %s is missing a segment.", _node_name(node));
return NULL;
}
return dm_list_item(dm_list_last(&node->props.segs), struct load_segment);
}
/* For preload pass only validate pool's transaction_id */
static int _node_send_messages(struct dm_tree_node *dnode,
const char *uuid_prefix,
size_t uuid_prefix_len,
int send)
{
struct load_segment *seg;
struct thin_message *tmsg;
struct dm_status_thin_pool stp;
const char *uuid;
int have_messages;
if (!dnode->info.exists)
return 1;
if (!(seg = _get_last_load_segment(dnode)))
return_0;
if (seg->type != SEG_THIN_POOL)
return 1;
if (!(uuid = dm_tree_node_get_uuid(dnode)))
return_0;
if (!_uuid_prefix_matches(uuid, uuid_prefix, uuid_prefix_len)) {
log_debug_activation("UUID \"%s\" does not match.", uuid);
return 1;
}
if (!_thin_pool_get_status(dnode, &stp))
return_0;
have_messages = !dm_list_empty(&seg->thin_messages) ? 1 : 0;
if (stp.transaction_id == seg->transaction_id) {
dnode->props.send_messages = 0; /* messages already committed */
if (have_messages)
log_debug_activation("Thin pool %s transaction_id matches %"
PRIu64 ", skipping messages.",
_node_name(dnode), stp.transaction_id);
return 1;
}
/* Error if there are no stacked messages or id mismatches */
if ((stp.transaction_id + 1) != seg->transaction_id) {
log_error("Thin pool %s transaction_id is %" PRIu64 ", while expected %" PRIu64 ".",
_node_name(dnode), stp.transaction_id, seg->transaction_id - have_messages);
return 0;
}
if (!have_messages || !send)
return 1; /* transaction_id is matching */
if (stp.fail || stp.read_only || stp.needs_check) {
log_error("Cannot send messages to thin pool %s%s%s%s.",
_node_name(dnode),
stp.fail ? " in failed state" : "",
stp.read_only ? " with read only metadata" : "",
stp.needs_check ? " which needs check first" : "");
return 0;
}
dm_list_iterate_items(tmsg, &seg->thin_messages) {
if (!(_thin_pool_node_message(dnode, tmsg)))
return_0;
if (tmsg->message.type == DM_THIN_MESSAGE_SET_TRANSACTION_ID) {
if (!_thin_pool_get_status(dnode, &stp))
return_0;
if (stp.transaction_id != tmsg->message.u.m_set_transaction_id.new_id) {
log_error("Thin pool %s transaction_id is %" PRIu64
" and does not match expected %" PRIu64 ".",
_node_name(dnode), stp.transaction_id,
tmsg->message.u.m_set_transaction_id.new_id);
return 0;
}
}
}
dnode->props.send_messages = 0; /* messages posted */
return 1;
}
/*
* FIXME Don't attempt to deactivate known internal dependencies.
*/
static int _dm_tree_deactivate_children(struct dm_tree_node *dnode,
const char *uuid_prefix,
size_t uuid_prefix_len,
unsigned level)
2005-10-18 16:37:53 +04:00
{
int r = 1;
2005-10-18 16:37:53 +04:00
void *handle = NULL;
2005-11-09 17:10:50 +03:00
struct dm_tree_node *child = dnode;
2005-10-18 16:37:53 +04:00
struct dm_info info;
const struct dm_info *dinfo;
const char *name;
const char *uuid;
2005-11-09 17:10:50 +03:00
while ((child = dm_tree_next_child(&handle, dnode, 0))) {
if (!(dinfo = dm_tree_node_get_info(child))) {
2005-10-18 16:37:53 +04:00
stack;
continue;
}
2005-11-09 17:10:50 +03:00
if (!(name = dm_tree_node_get_name(child))) {
2005-10-18 16:37:53 +04:00
stack;
continue;
}
2005-11-09 17:10:50 +03:00
if (!(uuid = dm_tree_node_get_uuid(child))) {
2005-10-18 16:37:53 +04:00
stack;
continue;
}
/* Ignore if it doesn't belong to this VG */
2005-10-26 19:21:13 +04:00
if (!_uuid_prefix_matches(uuid, uuid_prefix, uuid_prefix_len))
2005-10-18 16:37:53 +04:00
continue;
/* Refresh open_count */
if (!_info_by_dev(dinfo->major, dinfo->minor, 1, &info, NULL, NULL, NULL))
return_0;
if (!info.exists)
2005-10-18 16:37:53 +04:00
continue;
if (info.open_count) {
/* Skip internal non-toplevel opened nodes */
if (level)
continue;
/* When retry is not allowed, error */
if (!child->dtree->retry_remove) {
2017-12-04 16:37:01 +03:00
log_error("Unable to deactivate open %s (" FMTu32 ":"
FMTu32 ").", name, info.major, info.minor);
r = 0;
continue;
}
/* Check toplevel node for holders/mounted fs */
if (!_check_device_not_in_use(name, &info)) {
stack;
r = 0;
continue;
}
/* Go on with retry */
}
/* Also checking open_count in parent nodes of presuspend_node */
if ((child->presuspend_node &&
!_node_has_closed_parents(child->presuspend_node,
uuid_prefix, uuid_prefix_len))) {
/* Only report error from (likely non-internal) dependency at top level */
if (!level) {
2017-12-04 16:37:01 +03:00
log_error("Unable to deactivate open %s (" FMTu32 ":"
FMTu32 ").", name, info.major, info.minor);
r = 0;
}
continue;
}
/* Suspend child node first if requested */
if (child->presuspend_node &&
!dm_tree_suspend_children(child, uuid_prefix, uuid_prefix_len))
continue;
if (!_deactivate_node(name, info.major, info.minor,
&child->dtree->cookie, child->udev_flags,
(level == 0) ? child->dtree->retry_remove : 0)) {
2017-12-04 16:37:01 +03:00
log_error("Unable to deactivate %s (" FMTu32 ":"
FMTu32 ").", name, info.major, info.minor);
r = 0;
2005-10-18 16:37:53 +04:00
continue;
2014-02-23 02:13:43 +04:00
}
if (info.suspended && info.live_table)
dec_suspended();
2005-10-18 16:37:53 +04:00
if (child->callback &&
!child->callback(child, DM_NODE_CALLBACK_DEACTIVATED,
child->callback_data))
stack;
/* FIXME Deactivation must currently ignore failure
* here so that lvremove can continue: we need an
* alternative way to handle this state without
* setting r=0. Or better, skip calling thin_check
* entirely if the device is about to be removed. */
if (dm_tree_node_num_children(child, 0) &&
!_dm_tree_deactivate_children(child, uuid_prefix, uuid_prefix_len, level + 1))
return_0;
2005-10-18 16:37:53 +04:00
}
return r;
2005-10-18 16:37:53 +04:00
}
2005-10-25 23:09:41 +04:00
int dm_tree_deactivate_children(struct dm_tree_node *dnode,
2014-01-08 19:51:11 +04:00
const char *uuid_prefix,
size_t uuid_prefix_len)
{
return _dm_tree_deactivate_children(dnode, uuid_prefix, uuid_prefix_len, 0);
}
2005-11-09 17:10:50 +03:00
int dm_tree_suspend_children(struct dm_tree_node *dnode,
2009-12-03 12:58:30 +03:00
const char *uuid_prefix,
size_t uuid_prefix_len)
2005-10-25 23:09:41 +04:00
{
int r = 1;
2005-10-25 23:09:41 +04:00
void *handle = NULL;
2005-11-09 17:10:50 +03:00
struct dm_tree_node *child = dnode;
2005-10-25 23:09:41 +04:00
struct dm_info info, newinfo;
const struct dm_info *dinfo;
const char *name;
const char *uuid;
2005-10-26 18:08:24 +04:00
/* Suspend nodes at this level of the tree */
2005-11-09 17:10:50 +03:00
while ((child = dm_tree_next_child(&handle, dnode, 0))) {
if (!(dinfo = dm_tree_node_get_info(child))) {
2005-10-25 23:09:41 +04:00
stack;
continue;
}
2005-11-09 17:10:50 +03:00
if (!(name = dm_tree_node_get_name(child))) {
2005-10-25 23:09:41 +04:00
stack;
continue;
}
2005-11-09 17:10:50 +03:00
if (!(uuid = dm_tree_node_get_uuid(child))) {
2005-10-25 23:09:41 +04:00
stack;
continue;
}
/* Ignore if it doesn't belong to this VG */
2005-10-26 19:21:13 +04:00
if (!_uuid_prefix_matches(uuid, uuid_prefix, uuid_prefix_len))
2005-10-25 23:09:41 +04:00
continue;
2005-10-26 18:08:24 +04:00
/* Ensure immediate parents are already suspended */
if (!_children_suspended(child, 1, uuid_prefix, uuid_prefix_len))
continue;
if (!_info_by_dev(dinfo->major, dinfo->minor, 0, &info, NULL, NULL, NULL))
return_0;
if (!info.exists || info.suspended)
2005-10-25 23:09:41 +04:00
continue;
/* If child has some real messages send them */
if ((child->props.send_messages > 1) && r) {
if (!(r = _node_send_messages(child, uuid_prefix, uuid_prefix_len, 1)))
stack;
else {
2016-09-13 10:24:38 +03:00
log_debug_activation("Sent messages to thin-pool %s and "
"skipping suspend of its children.",
_node_name(child));
child->props.skip_suspend++;
}
continue;
}
2006-01-31 02:36:04 +03:00
if (!_suspend_node(name, info.major, info.minor,
2007-01-09 22:44:07 +03:00
child->dtree->skip_lockfs,
child->dtree->no_flush, &newinfo)) {
2017-12-04 16:37:01 +03:00
log_error("Unable to suspend %s (" FMTu32 ":"
FMTu32 ")", name, info.major, info.minor);
r = 0;
2005-10-25 23:09:41 +04:00
continue;
}
/* Update cached info */
child->info = newinfo;
2005-10-26 18:08:24 +04:00
}
/* Then suspend any child nodes */
handle = NULL;
2005-11-09 17:10:50 +03:00
while ((child = dm_tree_next_child(&handle, dnode, 0))) {
if (child->props.skip_suspend)
continue;
2005-11-09 17:10:50 +03:00
if (!(uuid = dm_tree_node_get_uuid(child))) {
2005-10-26 18:08:24 +04:00
stack;
continue;
}
/* Ignore if it doesn't belong to this VG */
if (!_uuid_prefix_matches(uuid, uuid_prefix, uuid_prefix_len))
2005-10-26 18:08:24 +04:00
continue;
2005-10-25 23:09:41 +04:00
2005-11-09 17:10:50 +03:00
if (dm_tree_node_num_children(child, 0))
if (!dm_tree_suspend_children(child, uuid_prefix, uuid_prefix_len))
return_0;
2005-10-25 23:09:41 +04:00
}
return r;
2005-10-25 23:09:41 +04:00
}
/*
* _rename_conflict_exists
* @dnode
* @node
* @resolvable
*
* Check if there is a rename conflict with existing peers in
* this tree. 'resolvable' is set if the conflicting node will
* also be undergoing a rename. (Allowing that node to rename
* first would clear the conflict.)
*
* Returns: 1 if conflict, 0 otherwise
*/
static int _rename_conflict_exists(struct dm_tree_node *parent,
struct dm_tree_node *node,
int *resolvable)
{
void *handle = NULL;
const char *name = dm_tree_node_get_name(node);
const char *sibling_name;
struct dm_tree_node *sibling;
*resolvable = 0;
if (!name)
return_0;
while ((sibling = dm_tree_next_child(&handle, parent, 0))) {
if (sibling == node)
continue;
if (!(sibling_name = dm_tree_node_get_name(sibling))) {
stack;
continue;
}
if (!strcmp(node->props.new_name, sibling_name)) {
if (sibling->props.new_name)
*resolvable = 1;
return 1;
}
}
return 0;
}
/*
* Reactivation of sibling nodes
*
* Function is used when activating origin and its thick snapshots
* to ensure udev is processing first the origin LV and all the
* snapshot LVs are processed afterwards.
*/
static int _reactivate_siblings(struct dm_tree_node *dnode,
const char *uuid_prefix,
size_t uuid_prefix_len)
{
struct dm_tree_node *child;
const char *uuid;
void *handle = NULL;
int r = 1;
/* Wait for udev before reactivating siblings */
if (!dm_udev_wait(dm_tree_get_cookie(dnode)))
stack;
dm_tree_set_cookie(dnode, 0);
while ((child = dm_tree_next_child(&handle, dnode, 0))) {
if (child->props.reactivate_siblings) {
/* Skip 'leading' device in this group, marked with flag */
child->props.reactivate_siblings = 0;
continue;
}
if (!(uuid = dm_tree_node_get_uuid(child))) {
stack;
continue;
}
if (!_uuid_prefix_matches(uuid, uuid_prefix, uuid_prefix_len))
continue;
if (!_suspend_node(child->name, child->info.major, child->info.minor,
child->dtree->skip_lockfs,
child->dtree->no_flush, &child->info)) {
log_error("Unable to suspend %s (" FMTu32
":" FMTu32 ")", child->name,
child->info.major, child->info.minor);
r = 0;
continue;
}
if (!_resume_node(child->name, child->info.major, child->info.minor,
child->props.read_ahead, child->props.read_ahead_flags,
&child->info, &child->dtree->cookie,
child->props.reactivate_udev_flags, // use these flags
child->info.suspended)) {
log_error("Failed to suspend %s (" FMTu32
":" FMTu32 ")", child->name,
child->info.major, child->info.minor);
r = 0;
continue;
}
}
return r;
}
2005-11-09 17:10:50 +03:00
int dm_tree_activate_children(struct dm_tree_node *dnode,
2005-10-25 23:09:41 +04:00
const char *uuid_prefix,
size_t uuid_prefix_len)
{
int r = 1;
int resolvable_name_conflict, awaiting_peer_rename = 0;
2005-10-25 23:09:41 +04:00
void *handle = NULL;
2005-11-09 17:10:50 +03:00
struct dm_tree_node *child = dnode;
const char *name;
2005-10-25 23:09:41 +04:00
const char *uuid;
2005-11-22 23:00:35 +03:00
int priority;
2005-10-25 23:09:41 +04:00
/* Activate children first */
2005-11-09 17:10:50 +03:00
while ((child = dm_tree_next_child(&handle, dnode, 0))) {
if (!(uuid = dm_tree_node_get_uuid(child))) {
stack;
continue;
2005-10-25 23:09:41 +04:00
}
2005-11-09 18:41:42 +03:00
if (!_uuid_prefix_matches(uuid, uuid_prefix, uuid_prefix_len))
continue;
2005-10-25 23:09:41 +04:00
2005-11-09 17:10:50 +03:00
if (dm_tree_node_num_children(child, 0))
if (!dm_tree_activate_children(child, uuid_prefix, uuid_prefix_len))
return_0;
2005-11-22 23:00:35 +03:00
}
2005-11-22 23:00:35 +03:00
handle = NULL;
for (priority = 0; priority < 3; priority++) {
awaiting_peer_rename = 0;
2005-11-22 23:00:35 +03:00
while ((child = dm_tree_next_child(&handle, dnode, 0))) {
if (priority != child->activation_priority)
continue;
2005-11-22 23:00:35 +03:00
if (!(uuid = dm_tree_node_get_uuid(child))) {
stack;
continue;
}
2005-11-22 23:00:35 +03:00
if (!_uuid_prefix_matches(uuid, uuid_prefix, uuid_prefix_len))
continue;
2005-11-22 23:00:35 +03:00
if (!(name = dm_tree_node_get_name(child))) {
stack;
continue;
}
/* Rename? */
if (child->props.new_name) {
if (_rename_conflict_exists(dnode, child, &resolvable_name_conflict) &&
resolvable_name_conflict) {
awaiting_peer_rename++;
continue;
}
if (!_rename_node(name, child->props.new_name, child->info.major,
child->info.minor, &child->dtree->cookie,
child->udev_flags)) {
2005-11-22 23:00:35 +03:00
log_error("Failed to rename %s (%" PRIu32
":%" PRIu32 ") to %s", name, child->info.major,
child->info.minor, child->props.new_name);
return 0;
}
child->name = child->props.new_name;
child->props.new_name = NULL;
}
if (!child->info.inactive_table && !child->info.suspended)
continue;
if (!_resume_node(child->name, child->info.major, child->info.minor,
child->props.read_ahead, child->props.read_ahead_flags,
&child->info, &child->dtree->cookie, child->udev_flags, child->info.suspended)) {
log_error("Unable to resume %s.", _node_name(child));
r = 0;
2005-11-22 23:00:35 +03:00
continue;
}
/*
* FIXME: Implement delayed error reporting
* activation should be stopped only in the case,
* the submission of transation_id message fails,
* resume should continue further, just whole command
* has to report failure.
*/
if (r && (child->props.send_messages > 1) &&
!(r = _node_send_messages(child, uuid_prefix, uuid_prefix_len, 1)))
stack;
/* Reactivate only for fresh activated origin */
if (r && child->props.reactivate_siblings &&
(!(r = _reactivate_siblings(dnode, uuid_prefix, uuid_prefix_len))))
stack;
2005-11-22 23:00:35 +03:00
}
if (awaiting_peer_rename)
priority--; /* redo priority level */
2005-10-25 23:09:41 +04:00
}
return r;
}
static int _create_node(struct dm_tree_node *dnode, struct dm_tree_node *parent)
{
int r = 0;
struct dm_task *dmt;
log_verbose("Creating %s", dnode->name);
if (!(dmt = dm_task_create(DM_DEVICE_CREATE))) {
log_error("Create dm_task creation failed for %s", dnode->name);
return 0;
}
if (!dm_task_set_name(dmt, dnode->name)) {
log_error("Failed to set device name for %s", dnode->name);
goto out;
}
if (!dm_task_set_uuid(dmt, dnode->uuid)) {
log_error("Failed to set uuid for %s", dnode->name);
goto out;
}
if (dnode->props.major &&
(!dm_task_set_major(dmt, dnode->props.major) ||
!dm_task_set_minor(dmt, dnode->props.minor))) {
log_error("Failed to set device number for %s creation.", dnode->name);
goto out;
}
if (dnode->props.read_only && !dm_task_set_ro(dmt)) {
log_error("Failed to set read only flag for %s", dnode->name);
goto out;
}
if (!dm_task_no_open_count(dmt))
log_warn("WARNING: Failed to disable open_count.");
if ((r = dm_task_run(dmt))) {
if (!(r = dm_task_get_info(dmt, &dnode->info)))
/*
* This should not be possible to occur. However,
* we print an error message anyway for the more
* absurd cases (e.g. memory corruption) so there
* is never any question as to which one failed.
*/
log_error(INTERNAL_ERROR
"Unable to get DM task info for %s.",
dnode->name);
}
if (r)
dnode->activated = 1;
out:
dm_task_destroy(dmt);
return r;
}
2005-11-09 17:10:50 +03:00
static int _build_dev_string(char *devbuf, size_t bufsize, struct dm_tree_node *node)
{
if (!dm_format_dev(devbuf, bufsize, node->info.major, node->info.minor)) {
2011-08-19 21:02:48 +04:00
log_error("Failed to format %s device number for %s as dm "
"target (%u,%u)",
node->name, node->uuid, node->info.major, node->info.minor);
return 0;
}
return 1;
}
/* simplify string emiting code */
#define EMIT_PARAMS(p, str...)\
do {\
int w;\
if ((w = dm_snprintf(params + p, paramsize - (size_t) p, str)) < 0) {\
stack; /* Out of space */\
return -1;\
}\
p += w;\
} while (0)
/*
* _emit_areas_line
*
* Returns: 1 on success, 0 on failure
*/
static int _emit_areas_line(struct dm_task *dmt __attribute__((unused)),
struct load_segment *seg, char *params,
size_t paramsize, int *pos)
{
struct seg_area *area;
char devbuf[DM_FORMAT_DEV_BUFSIZE];
unsigned first_time = 1;
dm_list_iterate_items(area, &seg->areas) {
switch (seg->type) {
case SEG_RAID0:
2016-07-02 00:20:54 +03:00
case SEG_RAID0_META:
case SEG_RAID1:
case SEG_RAID10:
case SEG_RAID4:
case SEG_RAID5_N:
case SEG_RAID5_LA:
case SEG_RAID5_RA:
case SEG_RAID5_LS:
case SEG_RAID5_RS:
case SEG_RAID6_N_6:
case SEG_RAID6_ZR:
case SEG_RAID6_NR:
case SEG_RAID6_NC:
case SEG_RAID6_LS_6:
case SEG_RAID6_RS_6:
case SEG_RAID6_LA_6:
case SEG_RAID6_RA_6:
Add the ability to split an image from the mirror and track changes. ~> lvconvert --splitmirrors 1 --trackchanges vg/lv The '--trackchanges' option allows a user the ability to use an image of a RAID1 array for the purposes of temporary read-only access. The image can be merged back into the array at a later time and only the blocks that have changed in the array since the split will be resync'ed. This operation can be thought of as a partial split. The image is never completely extracted from the array, in that the array reserves the position the device occupied and tracks the differences between the array and the split image via a bitmap. The image itself is rendered read-only and the name (<LV>_rimage_*) cannot be changed. The user can complete the split (permanently splitting the image from the array) by re-issuing the 'lvconvert' command without the '--trackchanges' argument and specifying the '--name' argument. ~> lvconvert --splitmirrors 1 --name my_split vg/lv Merging the tracked image back into the array is done with the '--merge' option (included in a follow-on patch). ~> lvconvert --merge vg/lv_rimage_<n> The internal mechanics of this are relatively simple. The 'raid' device- mapper target allows for the specification of an empty slot in an array via '- -'. This is what will be used if a partial activation of an array is ever required. (It would also be possible to use 'error' targets in place of the '- -'.) If a RAID image is found to be both read-only and visible, then it is considered separate from the array and '- -' is used to hold it's position in the array. So, all that needs to be done to temporarily split an image from the array /and/ cause the kernel target's bitmap to track (aka "mark") changes made is to make the specified image visible and read-only. To merge the device back into the array, the image needs to be returned to the read/write state of the top-level LV and made invisible.
2011-08-18 23:38:26 +04:00
if (!area->dev_node) {
EMIT_PARAMS(*pos, " -");
break;
}
if (!_build_dev_string(devbuf, sizeof(devbuf), area->dev_node))
return_0;
EMIT_PARAMS(*pos, " %s", devbuf);
break;
default:
Add the ability to split an image from the mirror and track changes. ~> lvconvert --splitmirrors 1 --trackchanges vg/lv The '--trackchanges' option allows a user the ability to use an image of a RAID1 array for the purposes of temporary read-only access. The image can be merged back into the array at a later time and only the blocks that have changed in the array since the split will be resync'ed. This operation can be thought of as a partial split. The image is never completely extracted from the array, in that the array reserves the position the device occupied and tracks the differences between the array and the split image via a bitmap. The image itself is rendered read-only and the name (<LV>_rimage_*) cannot be changed. The user can complete the split (permanently splitting the image from the array) by re-issuing the 'lvconvert' command without the '--trackchanges' argument and specifying the '--name' argument. ~> lvconvert --splitmirrors 1 --name my_split vg/lv Merging the tracked image back into the array is done with the '--merge' option (included in a follow-on patch). ~> lvconvert --merge vg/lv_rimage_<n> The internal mechanics of this are relatively simple. The 'raid' device- mapper target allows for the specification of an empty slot in an array via '- -'. This is what will be used if a partial activation of an array is ever required. (It would also be possible to use 'error' targets in place of the '- -'.) If a RAID image is found to be both read-only and visible, then it is considered separate from the array and '- -' is used to hold it's position in the array. So, all that needs to be done to temporarily split an image from the array /and/ cause the kernel target's bitmap to track (aka "mark") changes made is to make the specified image visible and read-only. To merge the device back into the array, the image needs to be returned to the read/write state of the top-level LV and made invisible.
2011-08-18 23:38:26 +04:00
if (!_build_dev_string(devbuf, sizeof(devbuf), area->dev_node))
return_0;
EMIT_PARAMS(*pos, "%s%s %" PRIu64, first_time ? "" : " ",
devbuf, area->offset);
}
first_time = 0;
}
return 1;
}
/*
* Returns: 1 on success, 0 on failure
*/
2011-09-07 12:37:48 +04:00
static int _mirror_emit_segment_line(struct dm_task *dmt, struct load_segment *seg,
char *params, size_t paramsize)
{
int block_on_error = 0;
int handle_errors = 0;
int dm_log_userspace = 0;
unsigned log_parm_count;
int pos = 0;
char logbuf[DM_FORMAT_DEV_BUFSIZE];
const char *logtype;
unsigned kmaj = 0, kmin = 0, krel = 0;
if (!get_uname_version(&kmaj, &kmin, &krel))
return_0;
2006-02-06 23:18:10 +03:00
if ((seg->flags & DM_BLOCK_ON_ERROR)) {
/*
* Originally, block_on_error was an argument to the log
* portion of the mirror CTR table. It was renamed to
* "handle_errors" and now resides in the 'features'
* section of the mirror CTR table (i.e. at the end).
*
* We can identify whether to use "block_on_error" or
* "handle_errors" by the dm-mirror module's version
* number (>= 1.12) or by the kernel version (>= 2.6.22).
*/
if (KERNEL_VERSION(kmaj, kmin, krel) >= KERNEL_VERSION(2, 6, 22))
handle_errors = 1;
else
block_on_error = 1;
}
if (seg->clustered) {
/* Cluster mirrors require a UUID */
if (!seg->uuid)
return_0;
/*
* Cluster mirrors used to have their own log
* types. Now they are accessed through the
* userspace log type.
*
* The dm-log-userspace module was added to the
* 2.6.31 kernel.
*/
if (KERNEL_VERSION(kmaj, kmin, krel) >= KERNEL_VERSION(2, 6, 31))
dm_log_userspace = 1;
}
/* Region size */
log_parm_count = 1;
/* [no]sync, block_on_error etc. */
log_parm_count += hweight32(seg->flags);
/* "handle_errors" is a feature arg now */
if (handle_errors)
log_parm_count--;
/* DM_CORELOG does not count in the param list */
if (seg->flags & DM_CORELOG)
log_parm_count--;
if (seg->clustered) {
log_parm_count++; /* For UUID */
if (!dm_log_userspace)
EMIT_PARAMS(pos, "clustered-");
else
/* For clustered-* type field inserted later */
log_parm_count++;
}
if (!seg->log)
logtype = "core";
else {
logtype = "disk";
log_parm_count++;
if (!_build_dev_string(logbuf, sizeof(logbuf), seg->log))
return_0;
}
if (dm_log_userspace)
EMIT_PARAMS(pos, "userspace %u %s clustered-%s",
log_parm_count, seg->uuid, logtype);
else
EMIT_PARAMS(pos, "%s %u", logtype, log_parm_count);
if (seg->log)
EMIT_PARAMS(pos, " %s", logbuf);
EMIT_PARAMS(pos, " %u", seg->region_size);
if (seg->clustered && !dm_log_userspace)
EMIT_PARAMS(pos, " %s", seg->uuid);
2006-02-06 23:18:10 +03:00
if ((seg->flags & DM_NOSYNC))
EMIT_PARAMS(pos, " nosync");
else if ((seg->flags & DM_FORCESYNC))
EMIT_PARAMS(pos, " sync");
if (block_on_error)
EMIT_PARAMS(pos, " block_on_error");
EMIT_PARAMS(pos, " %u ", seg->mirror_area_count);
2021-02-10 17:10:28 +03:00
if (!_emit_areas_line(dmt, seg, params, paramsize, &pos))
return_0;
if (handle_errors)
EMIT_PARAMS(pos, " 1 handle_errors");
return 1;
}
lvconvert: add infrastructure for RaidLV reshaping support In order to support striped raid5/6/10 LV reshaping (change of LV type, stripesize or number of legs), this patch introduces infrastructure prerequisites to be used by raid_manip.c extensions in followup patches. This base is needed for allocation of out-of-place reshape space required by the MD raid personalities to avoid writing over data in-place when reading off the current RAID layout or number of legs and writing out the new layout or to a different number of legs (i.e. restripe) Changes: - add members reshape_len to 'struct lv_segment' to store out-of-place reshape length per component rimage - add member data_copies to struct lv_segment to support more than 2 raid10 data copies - make alloc_lv_segment() aware of both reshape_len and data_copies - adjust all alloc_lv_segment() callers to the new API - add functions to retrieve the current data offset (needed for out-of-place reshaping space allocation) and the devices count from the kernel - make libdm deptree code aware of reshape_len - add LV flags for disk add/remove reshaping - support import/export of the new 'struct lv_segment' members - enhance lv_extend/_lv_reduce to cope with reshape_len - add seg_is_*/segtype_is_* macros related to reshaping - add target version check for reshaping - grow rebuilds/writemostly bitmaps to 246 bit to support kernel maximal - enhance libdm deptree code to support data_offset (out-of-place reshaping) and delta_disk (legs add/remove reshaping) target arguments Related: rhbz834579 Related: rhbz1191935 Related: rhbz1191978
2017-02-24 02:50:00 +03:00
static int _2_if_value(unsigned p)
{
return p ? 2 : 0;
}
lvconvert: add infrastructure for RaidLV reshaping support In order to support striped raid5/6/10 LV reshaping (change of LV type, stripesize or number of legs), this patch introduces infrastructure prerequisites to be used by raid_manip.c extensions in followup patches. This base is needed for allocation of out-of-place reshape space required by the MD raid personalities to avoid writing over data in-place when reading off the current RAID layout or number of legs and writing out the new layout or to a different number of legs (i.e. restripe) Changes: - add members reshape_len to 'struct lv_segment' to store out-of-place reshape length per component rimage - add member data_copies to struct lv_segment to support more than 2 raid10 data copies - make alloc_lv_segment() aware of both reshape_len and data_copies - adjust all alloc_lv_segment() callers to the new API - add functions to retrieve the current data offset (needed for out-of-place reshaping space allocation) and the devices count from the kernel - make libdm deptree code aware of reshape_len - add LV flags for disk add/remove reshaping - support import/export of the new 'struct lv_segment' members - enhance lv_extend/_lv_reduce to cope with reshape_len - add seg_is_*/segtype_is_* macros related to reshaping - add target version check for reshaping - grow rebuilds/writemostly bitmaps to 246 bit to support kernel maximal - enhance libdm deptree code to support data_offset (out-of-place reshaping) and delta_disk (legs add/remove reshaping) target arguments Related: rhbz834579 Related: rhbz1191935 Related: rhbz1191978
2017-02-24 02:50:00 +03:00
/* Return number of bits passed in @bits assuming 2 * 64 bit size */
2017-07-19 17:15:15 +03:00
static int _get_params_count(const uint64_t *bits)
{
int r = 0;
lvconvert: add infrastructure for RaidLV reshaping support In order to support striped raid5/6/10 LV reshaping (change of LV type, stripesize or number of legs), this patch introduces infrastructure prerequisites to be used by raid_manip.c extensions in followup patches. This base is needed for allocation of out-of-place reshape space required by the MD raid personalities to avoid writing over data in-place when reading off the current RAID layout or number of legs and writing out the new layout or to a different number of legs (i.e. restripe) Changes: - add members reshape_len to 'struct lv_segment' to store out-of-place reshape length per component rimage - add member data_copies to struct lv_segment to support more than 2 raid10 data copies - make alloc_lv_segment() aware of both reshape_len and data_copies - adjust all alloc_lv_segment() callers to the new API - add functions to retrieve the current data offset (needed for out-of-place reshaping space allocation) and the devices count from the kernel - make libdm deptree code aware of reshape_len - add LV flags for disk add/remove reshaping - support import/export of the new 'struct lv_segment' members - enhance lv_extend/_lv_reduce to cope with reshape_len - add seg_is_*/segtype_is_* macros related to reshaping - add target version check for reshaping - grow rebuilds/writemostly bitmaps to 246 bit to support kernel maximal - enhance libdm deptree code to support data_offset (out-of-place reshaping) and delta_disk (legs add/remove reshaping) target arguments Related: rhbz834579 Related: rhbz1191935 Related: rhbz1191978
2017-02-24 02:50:00 +03:00
int i = RAID_BITMAP_SIZE;
lvconvert: add infrastructure for RaidLV reshaping support In order to support striped raid5/6/10 LV reshaping (change of LV type, stripesize or number of legs), this patch introduces infrastructure prerequisites to be used by raid_manip.c extensions in followup patches. This base is needed for allocation of out-of-place reshape space required by the MD raid personalities to avoid writing over data in-place when reading off the current RAID layout or number of legs and writing out the new layout or to a different number of legs (i.e. restripe) Changes: - add members reshape_len to 'struct lv_segment' to store out-of-place reshape length per component rimage - add member data_copies to struct lv_segment to support more than 2 raid10 data copies - make alloc_lv_segment() aware of both reshape_len and data_copies - adjust all alloc_lv_segment() callers to the new API - add functions to retrieve the current data offset (needed for out-of-place reshaping space allocation) and the devices count from the kernel - make libdm deptree code aware of reshape_len - add LV flags for disk add/remove reshaping - support import/export of the new 'struct lv_segment' members - enhance lv_extend/_lv_reduce to cope with reshape_len - add seg_is_*/segtype_is_* macros related to reshaping - add target version check for reshaping - grow rebuilds/writemostly bitmaps to 246 bit to support kernel maximal - enhance libdm deptree code to support data_offset (out-of-place reshaping) and delta_disk (legs add/remove reshaping) target arguments Related: rhbz834579 Related: rhbz1191935 Related: rhbz1191978
2017-02-24 02:50:00 +03:00
while (i--) {
r += 2 * hweight32(bits[i] & 0xFFFFFFFF);
r += 2 * hweight32(bits[i] >> 32);
}
return r;
}
/*
* Get target version (major, minor and patchlevel) for @target_name
*
2017-03-24 02:12:41 +03:00
* FIXME: this function is derived from liblvm.
* Integrate with move of liblvm functions
* to libdm in future library layer purge
* (e.g. expose as API dm_target_version()?)
*/
static int _target_version(const char *target_name, uint32_t *maj,
uint32_t *min, uint32_t *patchlevel)
{
int r = 0;
struct dm_task *dmt;
struct dm_versions *target, *last_target = NULL;
log_very_verbose("Getting target version for %s", target_name);
if (!(dmt = dm_task_create(DM_DEVICE_LIST_VERSIONS)))
return_0;
if (!dm_task_run(dmt)) {
log_debug_activation("Failed to get %s target versions", target_name);
/* Assume this was because LIST_VERSIONS isn't supported */
*maj = *min = *patchlevel = 0;
r = 1;
} else
for (target = dm_task_get_versions(dmt);
target != last_target;
last_target = target, target = (struct dm_versions *)((char *) target + target->next))
if (!strcmp(target_name, target->name)) {
*maj = target->version[0];
*min = target->version[1];
*patchlevel = target->version[2];
log_very_verbose("Found %s target "
"v%" PRIu32 ".%" PRIu32 ".%" PRIu32 ".",
target_name, *maj, *min, *patchlevel);
r = 1;
break;
}
dm_task_destroy(dmt);
return r;
}
static int _raid_emit_segment_line(struct dm_task *dmt, uint32_t major,
uint32_t minor, struct load_segment *seg,
uint64_t *seg_start, char *params,
size_t paramsize)
{
uint32_t i;
lvconvert: add infrastructure for RaidLV reshaping support In order to support striped raid5/6/10 LV reshaping (change of LV type, stripesize or number of legs), this patch introduces infrastructure prerequisites to be used by raid_manip.c extensions in followup patches. This base is needed for allocation of out-of-place reshape space required by the MD raid personalities to avoid writing over data in-place when reading off the current RAID layout or number of legs and writing out the new layout or to a different number of legs (i.e. restripe) Changes: - add members reshape_len to 'struct lv_segment' to store out-of-place reshape length per component rimage - add member data_copies to struct lv_segment to support more than 2 raid10 data copies - make alloc_lv_segment() aware of both reshape_len and data_copies - adjust all alloc_lv_segment() callers to the new API - add functions to retrieve the current data offset (needed for out-of-place reshaping space allocation) and the devices count from the kernel - make libdm deptree code aware of reshape_len - add LV flags for disk add/remove reshaping - support import/export of the new 'struct lv_segment' members - enhance lv_extend/_lv_reduce to cope with reshape_len - add seg_is_*/segtype_is_* macros related to reshaping - add target version check for reshaping - grow rebuilds/writemostly bitmaps to 246 bit to support kernel maximal - enhance libdm deptree code to support data_offset (out-of-place reshaping) and delta_disk (legs add/remove reshaping) target arguments Related: rhbz834579 Related: rhbz1191935 Related: rhbz1191978
2017-02-24 02:50:00 +03:00
uint32_t area_count = seg->area_count / 2;
uint32_t maj, min, patchlevel;
int param_count = 1; /* mandatory 'chunk size'/'stripe size' arg */
int pos = 0;
lvconvert: add infrastructure for RaidLV reshaping support In order to support striped raid5/6/10 LV reshaping (change of LV type, stripesize or number of legs), this patch introduces infrastructure prerequisites to be used by raid_manip.c extensions in followup patches. This base is needed for allocation of out-of-place reshape space required by the MD raid personalities to avoid writing over data in-place when reading off the current RAID layout or number of legs and writing out the new layout or to a different number of legs (i.e. restripe) Changes: - add members reshape_len to 'struct lv_segment' to store out-of-place reshape length per component rimage - add member data_copies to struct lv_segment to support more than 2 raid10 data copies - make alloc_lv_segment() aware of both reshape_len and data_copies - adjust all alloc_lv_segment() callers to the new API - add functions to retrieve the current data offset (needed for out-of-place reshaping space allocation) and the devices count from the kernel - make libdm deptree code aware of reshape_len - add LV flags for disk add/remove reshaping - support import/export of the new 'struct lv_segment' members - enhance lv_extend/_lv_reduce to cope with reshape_len - add seg_is_*/segtype_is_* macros related to reshaping - add target version check for reshaping - grow rebuilds/writemostly bitmaps to 246 bit to support kernel maximal - enhance libdm deptree code to support data_offset (out-of-place reshaping) and delta_disk (legs add/remove reshaping) target arguments Related: rhbz834579 Related: rhbz1191935 Related: rhbz1191978
2017-02-24 02:50:00 +03:00
unsigned type;
if (seg->area_count % 2)
return 0;
if ((seg->flags & DM_NOSYNC) || (seg->flags & DM_FORCESYNC))
param_count++;
lvconvert: add infrastructure for RaidLV reshaping support In order to support striped raid5/6/10 LV reshaping (change of LV type, stripesize or number of legs), this patch introduces infrastructure prerequisites to be used by raid_manip.c extensions in followup patches. This base is needed for allocation of out-of-place reshape space required by the MD raid personalities to avoid writing over data in-place when reading off the current RAID layout or number of legs and writing out the new layout or to a different number of legs (i.e. restripe) Changes: - add members reshape_len to 'struct lv_segment' to store out-of-place reshape length per component rimage - add member data_copies to struct lv_segment to support more than 2 raid10 data copies - make alloc_lv_segment() aware of both reshape_len and data_copies - adjust all alloc_lv_segment() callers to the new API - add functions to retrieve the current data offset (needed for out-of-place reshaping space allocation) and the devices count from the kernel - make libdm deptree code aware of reshape_len - add LV flags for disk add/remove reshaping - support import/export of the new 'struct lv_segment' members - enhance lv_extend/_lv_reduce to cope with reshape_len - add seg_is_*/segtype_is_* macros related to reshaping - add target version check for reshaping - grow rebuilds/writemostly bitmaps to 246 bit to support kernel maximal - enhance libdm deptree code to support data_offset (out-of-place reshaping) and delta_disk (legs add/remove reshaping) target arguments Related: rhbz834579 Related: rhbz1191935 Related: rhbz1191978
2017-02-24 02:50:00 +03:00
param_count += _2_if_value(seg->data_offset) +
_2_if_value(seg->delta_disks) +
_2_if_value(seg->region_size) +
_2_if_value(seg->writebehind) +
_2_if_value(seg->min_recovery_rate) +
_2_if_value(seg->max_recovery_rate) +
_2_if_value(seg->data_copies > 1);
lvconvert: add infrastructure for RaidLV reshaping support In order to support striped raid5/6/10 LV reshaping (change of LV type, stripesize or number of legs), this patch introduces infrastructure prerequisites to be used by raid_manip.c extensions in followup patches. This base is needed for allocation of out-of-place reshape space required by the MD raid personalities to avoid writing over data in-place when reading off the current RAID layout or number of legs and writing out the new layout or to a different number of legs (i.e. restripe) Changes: - add members reshape_len to 'struct lv_segment' to store out-of-place reshape length per component rimage - add member data_copies to struct lv_segment to support more than 2 raid10 data copies - make alloc_lv_segment() aware of both reshape_len and data_copies - adjust all alloc_lv_segment() callers to the new API - add functions to retrieve the current data offset (needed for out-of-place reshaping space allocation) and the devices count from the kernel - make libdm deptree code aware of reshape_len - add LV flags for disk add/remove reshaping - support import/export of the new 'struct lv_segment' members - enhance lv_extend/_lv_reduce to cope with reshape_len - add seg_is_*/segtype_is_* macros related to reshaping - add target version check for reshaping - grow rebuilds/writemostly bitmaps to 246 bit to support kernel maximal - enhance libdm deptree code to support data_offset (out-of-place reshaping) and delta_disk (legs add/remove reshaping) target arguments Related: rhbz834579 Related: rhbz1191935 Related: rhbz1191978
2017-02-24 02:50:00 +03:00
/* rebuilds and writemostly are BITMAP_SIZE * 64 bits */
param_count += _get_params_count(seg->rebuilds);
param_count += _get_params_count(seg->writemostly);
RAID: Add writemostly/writebehind support for RAID1 'lvchange' is used to alter a RAID 1 logical volume's write-mostly and write-behind characteristics. The '--writemostly' parameter takes a PV as an argument with an optional trailing character to specify whether to set ('y'), unset ('n'), or toggle ('t') the value. If no trailing character is given, it will set the flag. Synopsis: lvchange [--writemostly <PV>:{t|y|n}] [--writebehind <count>] vg/lv Example: lvchange --writemostly /dev/sdb1:y --writebehind 512 vg/raid1_lv The last character in the 'lv_attr' field is used to show whether a device has the WriteMostly flag set. It is signified with a 'w'. If the device has failed, the 'p'artial flag has priority. Example ("nosync" raid1 with mismatch_cnt and writemostly): [~]# lvs -a --segment vg LV VG Attr #Str Type SSize raid1 vg Rwi---r-m 2 raid1 500.00m [raid1_rimage_0] vg Iwi---r-- 1 linear 500.00m [raid1_rimage_1] vg Iwi---r-w 1 linear 500.00m [raid1_rmeta_0] vg ewi---r-- 1 linear 4.00m [raid1_rmeta_1] vg ewi---r-- 1 linear 4.00m Example (raid1 with mismatch_cnt, writemostly - but failed drive): [~]# lvs -a --segment vg LV VG Attr #Str Type SSize raid1 vg rwi---r-p 2 raid1 500.00m [raid1_rimage_0] vg Iwi---r-- 1 linear 500.00m [raid1_rimage_1] vg Iwi---r-p 1 linear 500.00m [raid1_rmeta_0] vg ewi---r-- 1 linear 4.00m [raid1_rmeta_1] vg ewi---r-p 1 linear 4.00m A new reportable field has been added for writebehind as well. If write-behind has not been set or the LV is not RAID1, the field will be blank. Example (writebehind is set): [~]# lvs -a -o name,attr,writebehind vg LV Attr WBehind lv rwi-a-r-- 512 [lv_rimage_0] iwi-aor-w [lv_rimage_1] iwi-aor-- [lv_rmeta_0] ewi-aor-- [lv_rmeta_1] ewi-aor-- Example (writebehind is not set): [~]# lvs -a -o name,attr,writebehind vg LV Attr WBehind lv rwi-a-r-- [lv_rimage_0] iwi-aor-w [lv_rimage_1] iwi-aor-- [lv_rmeta_0] ewi-aor-- [lv_rmeta_1] ewi-aor--
2013-04-15 22:59:46 +04:00
lvconvert: add infrastructure for RaidLV reshaping support In order to support striped raid5/6/10 LV reshaping (change of LV type, stripesize or number of legs), this patch introduces infrastructure prerequisites to be used by raid_manip.c extensions in followup patches. This base is needed for allocation of out-of-place reshape space required by the MD raid personalities to avoid writing over data in-place when reading off the current RAID layout or number of legs and writing out the new layout or to a different number of legs (i.e. restripe) Changes: - add members reshape_len to 'struct lv_segment' to store out-of-place reshape length per component rimage - add member data_copies to struct lv_segment to support more than 2 raid10 data copies - make alloc_lv_segment() aware of both reshape_len and data_copies - adjust all alloc_lv_segment() callers to the new API - add functions to retrieve the current data offset (needed for out-of-place reshaping space allocation) and the devices count from the kernel - make libdm deptree code aware of reshape_len - add LV flags for disk add/remove reshaping - support import/export of the new 'struct lv_segment' members - enhance lv_extend/_lv_reduce to cope with reshape_len - add seg_is_*/segtype_is_* macros related to reshaping - add target version check for reshaping - grow rebuilds/writemostly bitmaps to 246 bit to support kernel maximal - enhance libdm deptree code to support data_offset (out-of-place reshaping) and delta_disk (legs add/remove reshaping) target arguments Related: rhbz834579 Related: rhbz1191935 Related: rhbz1191978
2017-02-24 02:50:00 +03:00
if ((seg->type == SEG_RAID1) && seg->stripe_size)
log_info("WARNING: Ignoring RAID1 stripe size");
2016-07-02 00:20:54 +03:00
/* Kernel only expects "raid0", not "raid0_meta" */
lvconvert: add infrastructure for RaidLV reshaping support In order to support striped raid5/6/10 LV reshaping (change of LV type, stripesize or number of legs), this patch introduces infrastructure prerequisites to be used by raid_manip.c extensions in followup patches. This base is needed for allocation of out-of-place reshape space required by the MD raid personalities to avoid writing over data in-place when reading off the current RAID layout or number of legs and writing out the new layout or to a different number of legs (i.e. restripe) Changes: - add members reshape_len to 'struct lv_segment' to store out-of-place reshape length per component rimage - add member data_copies to struct lv_segment to support more than 2 raid10 data copies - make alloc_lv_segment() aware of both reshape_len and data_copies - adjust all alloc_lv_segment() callers to the new API - add functions to retrieve the current data offset (needed for out-of-place reshaping space allocation) and the devices count from the kernel - make libdm deptree code aware of reshape_len - add LV flags for disk add/remove reshaping - support import/export of the new 'struct lv_segment' members - enhance lv_extend/_lv_reduce to cope with reshape_len - add seg_is_*/segtype_is_* macros related to reshaping - add target version check for reshaping - grow rebuilds/writemostly bitmaps to 246 bit to support kernel maximal - enhance libdm deptree code to support data_offset (out-of-place reshaping) and delta_disk (legs add/remove reshaping) target arguments Related: rhbz834579 Related: rhbz1191935 Related: rhbz1191978
2017-02-24 02:50:00 +03:00
type = seg->type;
2016-07-02 00:20:54 +03:00
if (type == SEG_RAID0_META)
type = SEG_RAID0;
lvconvert: add infrastructure for RaidLV reshaping support In order to support striped raid5/6/10 LV reshaping (change of LV type, stripesize or number of legs), this patch introduces infrastructure prerequisites to be used by raid_manip.c extensions in followup patches. This base is needed for allocation of out-of-place reshape space required by the MD raid personalities to avoid writing over data in-place when reading off the current RAID layout or number of legs and writing out the new layout or to a different number of legs (i.e. restripe) Changes: - add members reshape_len to 'struct lv_segment' to store out-of-place reshape length per component rimage - add member data_copies to struct lv_segment to support more than 2 raid10 data copies - make alloc_lv_segment() aware of both reshape_len and data_copies - adjust all alloc_lv_segment() callers to the new API - add functions to retrieve the current data offset (needed for out-of-place reshaping space allocation) and the devices count from the kernel - make libdm deptree code aware of reshape_len - add LV flags for disk add/remove reshaping - support import/export of the new 'struct lv_segment' members - enhance lv_extend/_lv_reduce to cope with reshape_len - add seg_is_*/segtype_is_* macros related to reshaping - add target version check for reshaping - grow rebuilds/writemostly bitmaps to 246 bit to support kernel maximal - enhance libdm deptree code to support data_offset (out-of-place reshaping) and delta_disk (legs add/remove reshaping) target arguments Related: rhbz834579 Related: rhbz1191935 Related: rhbz1191978
2017-02-24 02:50:00 +03:00
EMIT_PARAMS(pos, "%s %d %u",
type == SEG_RAID10 ? "raid10" : _dm_segtypes[type].target,
param_count, seg->stripe_size);
if (!_target_version("raid", &maj, &min, &patchlevel))
return_0;
/*
* Target version prior to 1.9.0 and >= 1.11.0 emit
* order of parameters as of kernel target documentation
*/
if (maj > 1 || (maj == 1 && (min < 9 || min >= 11))) {
if (seg->flags & DM_NOSYNC)
EMIT_PARAMS(pos, " nosync");
else if (seg->flags & DM_FORCESYNC)
EMIT_PARAMS(pos, " sync");
for (i = 0; i < area_count; i++)
if (seg->rebuilds[i/64] & (1ULL << (i%64)))
EMIT_PARAMS(pos, " rebuild %u", i);
if (seg->min_recovery_rate)
EMIT_PARAMS(pos, " min_recovery_rate %u",
seg->min_recovery_rate);
if (seg->max_recovery_rate)
EMIT_PARAMS(pos, " max_recovery_rate %u",
seg->max_recovery_rate);
for (i = 0; i < area_count; i++)
if (seg->writemostly[i/64] & (1ULL << (i%64)))
EMIT_PARAMS(pos, " write_mostly %u", i);
if (seg->writebehind)
EMIT_PARAMS(pos, " max_write_behind %u", seg->writebehind);
if (seg->region_size)
EMIT_PARAMS(pos, " region_size %u", seg->region_size);
if (seg->data_copies > 1 && type == SEG_RAID10)
EMIT_PARAMS(pos, " raid10_copies %u", seg->data_copies);
if (seg->delta_disks)
EMIT_PARAMS(pos, " delta_disks %d", seg->delta_disks);
lvconvert: add infrastructure for RaidLV reshaping support In order to support striped raid5/6/10 LV reshaping (change of LV type, stripesize or number of legs), this patch introduces infrastructure prerequisites to be used by raid_manip.c extensions in followup patches. This base is needed for allocation of out-of-place reshape space required by the MD raid personalities to avoid writing over data in-place when reading off the current RAID layout or number of legs and writing out the new layout or to a different number of legs (i.e. restripe) Changes: - add members reshape_len to 'struct lv_segment' to store out-of-place reshape length per component rimage - add member data_copies to struct lv_segment to support more than 2 raid10 data copies - make alloc_lv_segment() aware of both reshape_len and data_copies - adjust all alloc_lv_segment() callers to the new API - add functions to retrieve the current data offset (needed for out-of-place reshaping space allocation) and the devices count from the kernel - make libdm deptree code aware of reshape_len - add LV flags for disk add/remove reshaping - support import/export of the new 'struct lv_segment' members - enhance lv_extend/_lv_reduce to cope with reshape_len - add seg_is_*/segtype_is_* macros related to reshaping - add target version check for reshaping - grow rebuilds/writemostly bitmaps to 246 bit to support kernel maximal - enhance libdm deptree code to support data_offset (out-of-place reshaping) and delta_disk (legs add/remove reshaping) target arguments Related: rhbz834579 Related: rhbz1191935 Related: rhbz1191978
2017-02-24 02:50:00 +03:00
/* If seg-data_offset == 1, kernel needs a zero offset to adjust to it */
if (seg->data_offset)
EMIT_PARAMS(pos, " data_offset %d", seg->data_offset == 1 ? 0 : seg->data_offset);
lvconvert: add infrastructure for RaidLV reshaping support In order to support striped raid5/6/10 LV reshaping (change of LV type, stripesize or number of legs), this patch introduces infrastructure prerequisites to be used by raid_manip.c extensions in followup patches. This base is needed for allocation of out-of-place reshape space required by the MD raid personalities to avoid writing over data in-place when reading off the current RAID layout or number of legs and writing out the new layout or to a different number of legs (i.e. restripe) Changes: - add members reshape_len to 'struct lv_segment' to store out-of-place reshape length per component rimage - add member data_copies to struct lv_segment to support more than 2 raid10 data copies - make alloc_lv_segment() aware of both reshape_len and data_copies - adjust all alloc_lv_segment() callers to the new API - add functions to retrieve the current data offset (needed for out-of-place reshaping space allocation) and the devices count from the kernel - make libdm deptree code aware of reshape_len - add LV flags for disk add/remove reshaping - support import/export of the new 'struct lv_segment' members - enhance lv_extend/_lv_reduce to cope with reshape_len - add seg_is_*/segtype_is_* macros related to reshaping - add target version check for reshaping - grow rebuilds/writemostly bitmaps to 246 bit to support kernel maximal - enhance libdm deptree code to support data_offset (out-of-place reshaping) and delta_disk (legs add/remove reshaping) target arguments Related: rhbz834579 Related: rhbz1191935 Related: rhbz1191978
2017-02-24 02:50:00 +03:00
/* Target version >= 1.9.0 && < 1.11.0 had a table line parameter ordering flaw */
} else {
if (seg->data_copies > 1 && type == SEG_RAID10)
EMIT_PARAMS(pos, " raid10_copies %u", seg->data_copies);
lvconvert: add infrastructure for RaidLV reshaping support In order to support striped raid5/6/10 LV reshaping (change of LV type, stripesize or number of legs), this patch introduces infrastructure prerequisites to be used by raid_manip.c extensions in followup patches. This base is needed for allocation of out-of-place reshape space required by the MD raid personalities to avoid writing over data in-place when reading off the current RAID layout or number of legs and writing out the new layout or to a different number of legs (i.e. restripe) Changes: - add members reshape_len to 'struct lv_segment' to store out-of-place reshape length per component rimage - add member data_copies to struct lv_segment to support more than 2 raid10 data copies - make alloc_lv_segment() aware of both reshape_len and data_copies - adjust all alloc_lv_segment() callers to the new API - add functions to retrieve the current data offset (needed for out-of-place reshaping space allocation) and the devices count from the kernel - make libdm deptree code aware of reshape_len - add LV flags for disk add/remove reshaping - support import/export of the new 'struct lv_segment' members - enhance lv_extend/_lv_reduce to cope with reshape_len - add seg_is_*/segtype_is_* macros related to reshaping - add target version check for reshaping - grow rebuilds/writemostly bitmaps to 246 bit to support kernel maximal - enhance libdm deptree code to support data_offset (out-of-place reshaping) and delta_disk (legs add/remove reshaping) target arguments Related: rhbz834579 Related: rhbz1191935 Related: rhbz1191978
2017-02-24 02:50:00 +03:00
if (seg->flags & DM_NOSYNC)
EMIT_PARAMS(pos, " nosync");
else if (seg->flags & DM_FORCESYNC)
EMIT_PARAMS(pos, " sync");
if (seg->region_size)
EMIT_PARAMS(pos, " region_size %u", seg->region_size);
/* If seg-data_offset == 1, kernel needs a zero offset to adjust to it */
if (seg->data_offset)
EMIT_PARAMS(pos, " data_offset %d", seg->data_offset == 1 ? 0 : seg->data_offset);
if (seg->delta_disks)
EMIT_PARAMS(pos, " delta_disks %d", seg->delta_disks);
for (i = 0; i < area_count; i++)
if (seg->rebuilds[i/64] & (1ULL << (i%64)))
EMIT_PARAMS(pos, " rebuild %u", i);
for (i = 0; i < area_count; i++)
if (seg->writemostly[i/64] & (1ULL << (i%64)))
EMIT_PARAMS(pos, " write_mostly %u", i);
if (seg->writebehind)
EMIT_PARAMS(pos, " max_write_behind %u", seg->writebehind);
if (seg->max_recovery_rate)
EMIT_PARAMS(pos, " max_recovery_rate %u",
seg->max_recovery_rate);
if (seg->min_recovery_rate)
EMIT_PARAMS(pos, " min_recovery_rate %u",
seg->min_recovery_rate);
}
lvconvert: add infrastructure for RaidLV reshaping support In order to support striped raid5/6/10 LV reshaping (change of LV type, stripesize or number of legs), this patch introduces infrastructure prerequisites to be used by raid_manip.c extensions in followup patches. This base is needed for allocation of out-of-place reshape space required by the MD raid personalities to avoid writing over data in-place when reading off the current RAID layout or number of legs and writing out the new layout or to a different number of legs (i.e. restripe) Changes: - add members reshape_len to 'struct lv_segment' to store out-of-place reshape length per component rimage - add member data_copies to struct lv_segment to support more than 2 raid10 data copies - make alloc_lv_segment() aware of both reshape_len and data_copies - adjust all alloc_lv_segment() callers to the new API - add functions to retrieve the current data offset (needed for out-of-place reshaping space allocation) and the devices count from the kernel - make libdm deptree code aware of reshape_len - add LV flags for disk add/remove reshaping - support import/export of the new 'struct lv_segment' members - enhance lv_extend/_lv_reduce to cope with reshape_len - add seg_is_*/segtype_is_* macros related to reshaping - add target version check for reshaping - grow rebuilds/writemostly bitmaps to 246 bit to support kernel maximal - enhance libdm deptree code to support data_offset (out-of-place reshaping) and delta_disk (legs add/remove reshaping) target arguments Related: rhbz834579 Related: rhbz1191935 Related: rhbz1191978
2017-02-24 02:50:00 +03:00
/* Print number of metadata/data device pairs */
lvconvert: add infrastructure for RaidLV reshaping support In order to support striped raid5/6/10 LV reshaping (change of LV type, stripesize or number of legs), this patch introduces infrastructure prerequisites to be used by raid_manip.c extensions in followup patches. This base is needed for allocation of out-of-place reshape space required by the MD raid personalities to avoid writing over data in-place when reading off the current RAID layout or number of legs and writing out the new layout or to a different number of legs (i.e. restripe) Changes: - add members reshape_len to 'struct lv_segment' to store out-of-place reshape length per component rimage - add member data_copies to struct lv_segment to support more than 2 raid10 data copies - make alloc_lv_segment() aware of both reshape_len and data_copies - adjust all alloc_lv_segment() callers to the new API - add functions to retrieve the current data offset (needed for out-of-place reshaping space allocation) and the devices count from the kernel - make libdm deptree code aware of reshape_len - add LV flags for disk add/remove reshaping - support import/export of the new 'struct lv_segment' members - enhance lv_extend/_lv_reduce to cope with reshape_len - add seg_is_*/segtype_is_* macros related to reshaping - add target version check for reshaping - grow rebuilds/writemostly bitmaps to 246 bit to support kernel maximal - enhance libdm deptree code to support data_offset (out-of-place reshaping) and delta_disk (legs add/remove reshaping) target arguments Related: rhbz834579 Related: rhbz1191935 Related: rhbz1191978
2017-02-24 02:50:00 +03:00
EMIT_PARAMS(pos, " %u", area_count);
2021-02-10 17:10:28 +03:00
if (!_emit_areas_line(dmt, seg, params, paramsize, &pos))
return_0;
return 1;
}
static int _cache_emit_segment_line(struct dm_task *dmt,
struct load_segment *seg,
char *params, size_t paramsize)
{
int pos = 0;
/* unsigned feature_count; */
char data[DM_FORMAT_DEV_BUFSIZE];
char metadata[DM_FORMAT_DEV_BUFSIZE];
char origin[DM_FORMAT_DEV_BUFSIZE];
const char *name;
struct dm_config_node *cn;
/* Cache Dev */
if (!_build_dev_string(data, sizeof(data), seg->pool))
return_0;
/* Metadata Dev */
if (!_build_dev_string(metadata, sizeof(metadata), seg->metadata))
return_0;
/* Origin Dev */
if (!_build_dev_string(origin, sizeof(origin), seg->origin))
return_0;
EMIT_PARAMS(pos, "%s %s %s", metadata, data, origin);
/* Data block size */
EMIT_PARAMS(pos, " %u", seg->data_block_size);
/* Features */
/* feature_count = hweight32(seg->flags); */
/* EMIT_PARAMS(pos, " %u", feature_count); */
if (seg->flags & DM_CACHE_FEATURE_METADATA2)
EMIT_PARAMS(pos, " 2 metadata2 ");
else
EMIT_PARAMS(pos, " 1 ");
if (seg->flags & DM_CACHE_FEATURE_PASSTHROUGH)
EMIT_PARAMS(pos, "passthrough");
else if (seg->flags & DM_CACHE_FEATURE_WRITEBACK)
EMIT_PARAMS(pos, "writeback");
else
EMIT_PARAMS(pos, "writethrough");
/* Cache Policy */
name = seg->policy_name ? : "default";
EMIT_PARAMS(pos, " %s", name);
/* Do not pass migration_threshold 2048 which is default */
EMIT_PARAMS(pos, " %u", (seg->policy_argc + ((seg->migration_threshold != 2048) ? 1 : 0)) * 2);
if (seg->migration_threshold != 2048)
EMIT_PARAMS(pos, " migration_threshold %u", seg->migration_threshold);
if (seg->policy_settings)
for (cn = seg->policy_settings->child; cn; cn = cn->sib)
if (cn->v) /* Skip deleted entry */
EMIT_PARAMS(pos, " %s %" PRIu64, cn->key, cn->v->v.i);
return 1;
}
static int _thin_pool_emit_segment_line(struct dm_task *dmt,
struct load_segment *seg,
char *params, size_t paramsize)
{
int pos = 0;
char pool[DM_FORMAT_DEV_BUFSIZE], metadata[DM_FORMAT_DEV_BUFSIZE];
int features = (seg->error_if_no_space ? 1 : 0) +
(seg->read_only ? 1 : 0) +
(seg->ignore_discard ? 1 : 0) +
(seg->no_discard_passdown ? 1 : 0) +
(seg->skip_block_zeroing ? 1 : 0);
if (!_build_dev_string(metadata, sizeof(metadata), seg->metadata))
return_0;
if (!_build_dev_string(pool, sizeof(pool), seg->pool))
return_0;
EMIT_PARAMS(pos, "%s %s %d %" PRIu64 " %d%s%s%s%s%s", metadata, pool,
seg->data_block_size, seg->low_water_mark, features,
seg->skip_block_zeroing ? " skip_block_zeroing" : "",
seg->ignore_discard ? " ignore_discard" : "",
seg->no_discard_passdown ? " no_discard_passdown" : "",
seg->error_if_no_space ? " error_if_no_space" : "",
seg->read_only ? " read_only" : ""
);
return 1;
}
static int _thin_emit_segment_line(struct dm_task *dmt,
struct load_segment *seg,
char *params, size_t paramsize)
{
int pos = 0;
char pool[DM_FORMAT_DEV_BUFSIZE];
char external[DM_FORMAT_DEV_BUFSIZE + 1];
if (!_build_dev_string(pool, sizeof(pool), seg->pool))
return_0;
if (!seg->external)
*external = 0;
else {
*external = ' ';
if (!_build_dev_string(external + 1, sizeof(external) - 1,
seg->external))
return_0;
}
EMIT_PARAMS(pos, "%s %d%s", pool, seg->device_id, external);
return 1;
}
static int _emit_segment_line(struct dm_task *dmt, uint32_t major,
uint32_t minor, struct load_segment *seg,
uint64_t *seg_start, char *params,
size_t paramsize)
{
int pos = 0;
int target_type_is_raid = 0;
char originbuf[DM_FORMAT_DEV_BUFSIZE], cowbuf[DM_FORMAT_DEV_BUFSIZE];
switch(seg->type) {
case SEG_ERROR:
case SEG_ZERO:
case SEG_LINEAR:
break;
case SEG_MIRRORED:
/* Mirrors are pretty complicated - now in separate function */
2021-02-10 17:10:28 +03:00
if (!_mirror_emit_segment_line(dmt, seg, params, paramsize))
return_0;
break;
case SEG_SNAPSHOT:
case SEG_SNAPSHOT_MERGE:
2005-11-09 17:10:50 +03:00
if (!_build_dev_string(originbuf, sizeof(originbuf), seg->origin))
return_0;
if (!_build_dev_string(cowbuf, sizeof(cowbuf), seg->cow))
return_0;
EMIT_PARAMS(pos, "%s %s %c %d", originbuf, cowbuf,
seg->persistent ? 'P' : 'N', seg->chunk_size);
break;
case SEG_SNAPSHOT_ORIGIN:
2005-11-09 17:10:50 +03:00
if (!_build_dev_string(originbuf, sizeof(originbuf), seg->origin))
return_0;
EMIT_PARAMS(pos, "%s", originbuf);
break;
case SEG_STRIPED:
EMIT_PARAMS(pos, "%u %u ", seg->area_count, seg->stripe_size);
break;
case SEG_CRYPT:
EMIT_PARAMS(pos, "%s%s%s%s%s %s %" PRIu64 " ", seg->cipher,
seg->chainmode ? "-" : "", seg->chainmode ?: "",
seg->iv ? "-" : "", seg->iv ?: "", seg->key,
seg->iv_offset != DM_CRYPT_IV_DEFAULT ?
seg->iv_offset : *seg_start);
break;
case SEG_RAID0:
2016-07-02 00:20:54 +03:00
case SEG_RAID0_META:
case SEG_RAID1:
case SEG_RAID10:
case SEG_RAID4:
case SEG_RAID5_N:
case SEG_RAID5_LA:
case SEG_RAID5_RA:
case SEG_RAID5_LS:
case SEG_RAID5_RS:
case SEG_RAID6_N_6:
case SEG_RAID6_ZR:
case SEG_RAID6_NR:
case SEG_RAID6_NC:
case SEG_RAID6_LS_6:
case SEG_RAID6_RS_6:
case SEG_RAID6_LA_6:
case SEG_RAID6_RA_6:
target_type_is_raid = 1;
2021-02-10 17:10:28 +03:00
if (!_raid_emit_segment_line(dmt, major, minor, seg, seg_start,
params, paramsize))
return_0;
break;
case SEG_THIN_POOL:
if (!_thin_pool_emit_segment_line(dmt, seg, params, paramsize))
return_0;
break;
case SEG_THIN:
if (!_thin_emit_segment_line(dmt, seg, params, paramsize))
return_0;
break;
case SEG_CACHE:
if (!_cache_emit_segment_line(dmt, seg, params, paramsize))
return_0;
break;
}
switch(seg->type) {
case SEG_ERROR:
case SEG_SNAPSHOT:
case SEG_SNAPSHOT_ORIGIN:
case SEG_SNAPSHOT_MERGE:
case SEG_ZERO:
case SEG_THIN_POOL:
case SEG_THIN:
case SEG_CACHE:
break;
case SEG_CRYPT:
case SEG_LINEAR:
case SEG_STRIPED:
2021-02-10 17:10:28 +03:00
if (!_emit_areas_line(dmt, seg, params, paramsize, &pos))
return_0;
if (!params[0]) {
log_error("No parameters supplied for %s target "
"%u:%u.", _dm_segtypes[seg->type].target,
major, minor);
return 0;
}
break;
}
log_debug_activation("Adding target to (%" PRIu32 ":%" PRIu32 "): %" PRIu64
" %" PRIu64 " %s %s", major, minor,
*seg_start, seg->size, target_type_is_raid ? "raid" :
_dm_segtypes[seg->type].target, params);
if (!dm_task_add_target(dmt, *seg_start, seg->size,
target_type_is_raid ? "raid" :
_dm_segtypes[seg->type].target, params))
2005-11-09 17:10:50 +03:00
return_0;
*seg_start += seg->size;
return 1;
}
#undef EMIT_PARAMS
static int _emit_segment(struct dm_task *dmt, uint32_t major, uint32_t minor,
struct load_segment *seg, uint64_t *seg_start)
{
char *params;
size_t paramsize = 4096; /* FIXME: too small for long RAID lines when > 64 devices supported */
int ret;
do {
if (!(params = dm_malloc(paramsize))) {
log_error("Insufficient space for target parameters.");
return 0;
}
params[0] = '\0';
ret = _emit_segment_line(dmt, major, minor, seg, seg_start,
params, paramsize);
dm_free(params);
if (!ret)
stack;
if (ret >= 0)
return ret;
log_debug_activation("Insufficient space in params[%" PRIsize_t
"] for target parameters.", paramsize);
paramsize *= 2;
} while (paramsize < MAX_TARGET_PARAMSIZE);
log_error("Target parameter size too big. Aborting.");
return 0;
}
2005-11-09 17:10:50 +03:00
static int _load_node(struct dm_tree_node *dnode)
{
int r = 0;
struct dm_task *dmt;
struct load_segment *seg;
uint64_t seg_start = 0, existing_table_size;
log_verbose("Loading table for %s.", _node_name(dnode));
if (!(dmt = dm_task_create(DM_DEVICE_RELOAD))) {
log_error("Reload dm_task creation failed for %s.", _node_name(dnode));
return 0;
}
if (!dm_task_set_major(dmt, dnode->info.major) ||
!dm_task_set_minor(dmt, dnode->info.minor)) {
log_error("Failed to set device number for %s reload.", _node_name(dnode));
goto out;
}
if (dnode->props.read_only && !dm_task_set_ro(dmt)) {
log_error("Failed to set read only flag for %s.", _node_name(dnode));
goto out;
}
if (!dm_task_no_open_count(dmt))
log_warn("WARNING: Failed to disable open_count.");
dm_list_iterate_items(seg, &dnode->props.segs)
if (!_emit_segment(dmt, dnode->info.major, dnode->info.minor,
seg, &seg_start))
2005-11-09 17:10:50 +03:00
goto_out;
if (!dm_task_suppress_identical_reload(dmt))
log_warn("WARNING: Failed to suppress reload of identical tables.");
if ((r = dm_task_run(dmt))) {
r = dm_task_get_info(dmt, &dnode->info);
if (r && !dnode->info.inactive_table)
log_verbose("Suppressed %s identical table reload.",
_node_name(dnode));
existing_table_size = dm_task_get_existing_table_size(dmt);
if ((dnode->props.size_changed =
(existing_table_size == seg_start) ? 0 :
(existing_table_size > seg_start) ? -1 : 1)) {
/*
* Kernel usually skips size validation on zero-length devices
* now so no need to preload them.
*/
/* FIXME In which kernel version did this begin? */
if (!existing_table_size && dnode->props.delay_resume_if_new)
dnode->props.size_changed = 0;
log_debug_activation("Table size changed from %" PRIu64 " to %"
PRIu64 " for %s.%s", existing_table_size,
seg_start, _node_name(dnode),
dnode->props.size_changed ? "" : " (Ignoring.)");
/*
* FIXME: code here has known design problem.
* LVM2 does NOT resize thin-pool on top of other LV in 2 steps -
* where raid would be resized with 1st. transaction
* followed by 2nd. thin-pool resize - RHBZ #1285063
*/
if (existing_table_size && dnode->props.delay_resume_if_extended) {
log_debug_activation("Resume of table of extended device %s delayed.",
_node_name(dnode));
dnode->props.size_changed = 0;
}
}
}
dnode->props.segment_count = 0;
out:
dm_task_destroy(dmt);
return r;
}
/* Try to deactivate only nodes created during preload. */
static int _dm_tree_revert_activated(struct dm_tree_node *dnode)
{
void *handle = NULL;
struct dm_tree_node *child;
while ((child = dm_tree_next_child(&handle, dnode, 0))) {
if (child->activated) {
if (child->callback) {
log_debug_activation("Dropping callback for %s.", _node_name(child));
child->callback = NULL;
}
log_debug_activation("Reverting %s.", _node_name(child));
if (!_deactivate_node(child->name, child->info.major, child->info.minor,
&child->dtree->cookie, child->udev_flags, 0)) {
log_debug_activation("Unable to deactivate %s.", _node_name(child));
return 0;
}
}
if (dm_tree_node_num_children(child, 0) &&
!_dm_tree_revert_activated(child))
return_0;
}
return 1;
}
static int _dm_tree_wait_and_revert_activated(struct dm_tree_node *dnode)
{
if (!dm_udev_wait(dm_tree_get_cookie(dnode)))
stack;
dm_tree_set_cookie(dnode, 0);
return _dm_tree_revert_activated(dnode);
}
2005-11-09 17:10:50 +03:00
int dm_tree_preload_children(struct dm_tree_node *dnode,
const char *uuid_prefix,
size_t uuid_prefix_len)
{
int r = 1, node_created = 0;
void *handle = NULL;
2005-11-09 17:10:50 +03:00
struct dm_tree_node *child;
int update_devs_flag = 0;
/* Preload children first */
2005-11-09 17:10:50 +03:00
while ((child = dm_tree_next_child(&handle, dnode, 0))) {
/* Propagate delay of resume from parent node */
if (dnode->props.delay_resume_if_new > 1)
child->props.delay_resume_if_new = dnode->props.delay_resume_if_new;
/* Skip existing non-device-mapper devices */
if (!child->info.exists && child->info.major)
continue;
/* Ignore if it doesn't belong to this VG */
if (child->info.exists &&
!_uuid_prefix_matches(child->uuid, uuid_prefix, uuid_prefix_len))
continue;
2005-11-09 17:10:50 +03:00
if (dm_tree_node_num_children(child, 0))
if (!dm_tree_preload_children(child, uuid_prefix, uuid_prefix_len))
return_0;
/* FIXME Cope if name exists with no uuid? */
if (!child->info.exists && !(node_created = _create_node(child, dnode)))
return_0;
/* Propagate delayed resume from exteded child node */
if (child->props.delay_resume_if_extended)
dnode->props.delay_resume_if_extended = 1;
if (!child->info.inactive_table &&
child->props.segment_count &&
!_load_node(child)) {
stack;
/*
* If the table load fails, try to device in the kernel
* together with other created and preloaded devices.
*/
if (!_dm_tree_wait_and_revert_activated(dnode))
stack;
r = 0;
continue;
}
/* No resume for a device without parents or with unchanged or smaller size */
if (!dm_tree_node_num_children(child, 1) || (child->props.size_changed <= 0))
continue;
2005-11-22 22:31:20 +03:00
if (!child->info.inactive_table && !child->info.suspended)
continue;
if (!_resume_node(child->name, child->info.major, child->info.minor,
child->props.read_ahead, child->props.read_ahead_flags,
&child->info, &child->dtree->cookie, child->udev_flags,
child->info.suspended)) {
log_error("Unable to resume %s.", _node_name(child));
if (!_dm_tree_wait_and_revert_activated(dnode))
stack;
r = 0;
continue;
}
if (node_created) {
/* When creating new node also check transaction_id. */
if (child->props.send_messages &&
!_node_send_messages(child, uuid_prefix, uuid_prefix_len, 0)) {
stack;
if (!_dm_tree_wait_and_revert_activated(dnode))
stack;
r = 0;
continue;
}
}
/*
* Prepare for immediate synchronization with udev and flush all stacked
* dev node operations if requested by immediate_dev_node property. But
* finish processing current level in the tree first.
*/
if (child->props.immediate_dev_node)
update_devs_flag = 1;
}
if (update_devs_flag ||
(r && !dnode->info.exists && dnode->callback)) {
if (!dm_udev_wait(dm_tree_get_cookie(dnode)))
stack;
dm_tree_set_cookie(dnode, 0);
if (r && !dnode->info.exists && dnode->callback &&
!dnode->callback(dnode, DM_NODE_CALLBACK_PRELOADED,
dnode->callback_data))
{
/* Try to deactivate what has been activated in preload phase */
(void) _dm_tree_revert_activated(dnode);
return_0;
}
}
return r;
}
/*
* Returns 1 if unsure.
*/
2005-11-09 17:10:50 +03:00
int dm_tree_children_use_uuid(struct dm_tree_node *dnode,
const char *uuid_prefix,
size_t uuid_prefix_len)
{
void *handle = NULL;
2005-11-09 17:10:50 +03:00
struct dm_tree_node *child = dnode;
const char *uuid;
2005-11-09 17:10:50 +03:00
while ((child = dm_tree_next_child(&handle, dnode, 0))) {
if (!(uuid = dm_tree_node_get_uuid(child))) {
log_warn("WARNING: Failed to get uuid for dtree node %s.",
_node_name(child));
return 1;
}
if (_uuid_prefix_matches(uuid, uuid_prefix, uuid_prefix_len))
return 1;
2005-11-09 17:10:50 +03:00
if (dm_tree_node_num_children(child, 0))
dm_tree_children_use_uuid(child, uuid_prefix, uuid_prefix_len);
}
return 0;
}
/*
* Target functions
*/
2005-11-09 17:10:50 +03:00
static struct load_segment *_add_segment(struct dm_tree_node *dnode, unsigned type, uint64_t size)
{
struct load_segment *seg;
2005-11-09 17:10:50 +03:00
if (!(seg = dm_pool_zalloc(dnode->dtree->mem, sizeof(*seg)))) {
log_error("dtree node segment allocation failed");
return NULL;
}
seg->type = type;
seg->size = size;
dm_list_init(&seg->areas);
dm_list_add(&dnode->props.segs, &seg->list);
dnode->props.segment_count++;
return seg;
}
2005-11-09 17:10:50 +03:00
int dm_tree_node_add_snapshot_origin_target(struct dm_tree_node *dnode,
2014-02-14 23:46:55 +04:00
uint64_t size,
const char *origin_uuid)
{
struct load_segment *seg;
2005-11-09 17:10:50 +03:00
struct dm_tree_node *origin_node;
2005-11-09 17:10:50 +03:00
if (!(seg = _add_segment(dnode, SEG_SNAPSHOT_ORIGIN, size)))
return_0;
2005-11-09 17:10:50 +03:00
if (!(origin_node = dm_tree_find_node_by_uuid(dnode->dtree, origin_uuid))) {
log_error("Couldn't find snapshot origin uuid %s.", origin_uuid);
return 0;
}
seg->origin = origin_node;
2005-11-09 17:10:50 +03:00
if (!_link_tree_nodes(dnode, origin_node))
return_0;
2005-11-22 23:00:35 +03:00
/* Resume snapshot origins after new snapshots */
dnode->activation_priority = 1;
if (!dnode->info.exists)
/* Reactivate siblings for this origin after being resumed */
dnode->props.reactivate_siblings = 1;
/*
* Don't resume the origin immediately in case it is a non-trivial
* target that must not be active more than once concurrently!
*/
origin_node->props.delay_resume_if_new = 1;
return 1;
}
static int _add_snapshot_target(struct dm_tree_node *node,
2014-02-14 23:46:55 +04:00
uint64_t size,
const char *origin_uuid,
const char *cow_uuid,
const char *merge_uuid,
int persistent,
uint32_t chunk_size)
{
struct load_segment *seg;
struct dm_tree_node *origin_node, *cow_node, *merge_node;
unsigned seg_type;
seg_type = !merge_uuid ? SEG_SNAPSHOT : SEG_SNAPSHOT_MERGE;
if (!(seg = _add_segment(node, seg_type, size)))
2005-11-09 17:10:50 +03:00
return_0;
2005-11-09 17:10:50 +03:00
if (!(origin_node = dm_tree_find_node_by_uuid(node->dtree, origin_uuid))) {
log_error("Couldn't find snapshot origin uuid %s.", origin_uuid);
return 0;
}
seg->origin = origin_node;
2005-11-09 17:10:50 +03:00
if (!_link_tree_nodes(node, origin_node))
return_0;
2005-11-09 17:10:50 +03:00
if (!(cow_node = dm_tree_find_node_by_uuid(node->dtree, cow_uuid))) {
log_error("Couldn't find snapshot COW device uuid %s.", cow_uuid);
return 0;
}
seg->cow = cow_node;
2005-11-09 17:10:50 +03:00
if (!_link_tree_nodes(node, cow_node))
return_0;
seg->persistent = persistent ? 1 : 0;
seg->chunk_size = chunk_size;
if (merge_uuid) {
if (!(merge_node = dm_tree_find_node_by_uuid(node->dtree, merge_uuid))) {
/* not a pure error, merging snapshot may have been deactivated */
log_verbose("Couldn't find merging snapshot uuid %s.", merge_uuid);
} else {
seg->merge = merge_node;
/* must not link merging snapshot, would undermine activation_priority below */
}
/* Resume snapshot-merge (acting origin) after other snapshots */
node->activation_priority = 1;
if (seg->merge) {
/* Resume merging snapshot after snapshot-merge */
seg->merge->activation_priority = 2;
}
} else if (!origin_node->info.exists) {
/* Keep original udev_flags for reactivation. */
node->props.reactivate_udev_flags = node->udev_flags;
/* Reactivation is needed if the origin's -real device is not in DM table.
* For this case after the resume of its origin LV we resume its snapshots
* with updated udev_flags to completely avoid udev scanning for the first resume.
* Reactivation then resumes snapshots with original udev_flags.
*/
node->udev_flags |= DM_SUBSYSTEM_UDEV_FLAG0 |
DM_UDEV_DISABLE_DISK_RULES_FLAG |
DM_UDEV_DISABLE_OTHER_RULES_FLAG;
log_debug_activation("Using udev_flags 0x%x for activation of %s.",
node->udev_flags, node->name);
}
return 1;
}
int dm_tree_node_add_snapshot_target(struct dm_tree_node *node,
uint64_t size,
const char *origin_uuid,
const char *cow_uuid,
int persistent,
uint32_t chunk_size)
{
return _add_snapshot_target(node, size, origin_uuid, cow_uuid,
NULL, persistent, chunk_size);
}
int dm_tree_node_add_snapshot_merge_target(struct dm_tree_node *node,
uint64_t size,
const char *origin_uuid,
const char *cow_uuid,
const char *merge_uuid,
uint32_t chunk_size)
{
return _add_snapshot_target(node, size, origin_uuid, cow_uuid,
merge_uuid, 1, chunk_size);
}
2005-11-09 17:10:50 +03:00
int dm_tree_node_add_error_target(struct dm_tree_node *node,
2014-02-14 23:46:55 +04:00
uint64_t size)
{
2005-11-09 17:10:50 +03:00
if (!_add_segment(node, SEG_ERROR, size))
return_0;
return 1;
}
2005-11-09 17:10:50 +03:00
int dm_tree_node_add_zero_target(struct dm_tree_node *node,
2014-02-14 23:46:55 +04:00
uint64_t size)
{
2005-11-09 17:10:50 +03:00
if (!_add_segment(node, SEG_ZERO, size))
return_0;
return 1;
}
2005-11-09 17:10:50 +03:00
int dm_tree_node_add_linear_target(struct dm_tree_node *node,
2014-02-14 23:46:55 +04:00
uint64_t size)
{
2005-11-09 17:10:50 +03:00
if (!_add_segment(node, SEG_LINEAR, size))
return_0;
return 1;
}
2005-11-09 17:10:50 +03:00
int dm_tree_node_add_striped_target(struct dm_tree_node *node,
2014-02-14 23:46:55 +04:00
uint64_t size,
uint32_t stripe_size)
{
struct load_segment *seg;
2005-11-09 17:10:50 +03:00
if (!(seg = _add_segment(node, SEG_STRIPED, size)))
return_0;
seg->stripe_size = stripe_size;
return 1;
}
int dm_tree_node_add_crypt_target(struct dm_tree_node *node,
uint64_t size,
const char *cipher,
const char *chainmode,
const char *iv,
uint64_t iv_offset,
const char *key)
{
struct load_segment *seg;
if (!(seg = _add_segment(node, SEG_CRYPT, size)))
return_0;
seg->cipher = cipher;
seg->chainmode = chainmode;
seg->iv = iv;
seg->iv_offset = iv_offset;
seg->key = key;
return 1;
}
2005-11-09 17:10:50 +03:00
int dm_tree_node_add_mirror_target_log(struct dm_tree_node *node,
2014-02-14 23:46:55 +04:00
uint32_t region_size,
unsigned clustered,
const char *log_uuid,
unsigned area_count,
uint32_t flags)
{
2005-11-09 18:41:42 +03:00
struct dm_tree_node *log_node = NULL;
struct load_segment *seg;
if (!(seg = _get_last_load_segment(node)))
return_0;
2005-11-09 20:32:31 +03:00
if (log_uuid) {
2006-02-06 23:18:10 +03:00
if (!(seg->uuid = dm_pool_strdup(node->dtree->mem, log_uuid))) {
log_error("log uuid pool_strdup failed");
return 0;
}
if ((flags & DM_CORELOG))
/* For pvmove: immediate resume (for size validation) isn't needed. */
/* pvmove flag passed via unused UUID and its suffix */
node->props.delay_resume_if_new = strstr(log_uuid, "pvmove") ? 2 : 1;
else {
if (!(log_node = dm_tree_find_node_by_uuid(node->dtree, log_uuid))) {
log_error("Couldn't find mirror log uuid %s.", log_uuid);
return 0;
}
if (clustered)
log_node->props.immediate_dev_node = 1;
2011-06-11 16:55:31 +04:00
/* The kernel validates the size of disk logs. */
/* FIXME Propagate to any devices below */
log_node->props.delay_resume_if_new = 0;
if (!_link_tree_nodes(node, log_node))
return_0;
}
}
seg->log = log_node;
seg->region_size = region_size;
seg->clustered = clustered;
seg->mirror_area_count = area_count;
seg->flags = flags;
return 1;
}
2005-11-09 17:10:50 +03:00
int dm_tree_node_add_mirror_target(struct dm_tree_node *node,
2014-02-14 23:46:55 +04:00
uint64_t size)
{
if (!_add_segment(node, SEG_MIRRORED, size))
2005-11-09 17:10:50 +03:00
return_0;
return 1;
}
RAID: Add writemostly/writebehind support for RAID1 'lvchange' is used to alter a RAID 1 logical volume's write-mostly and write-behind characteristics. The '--writemostly' parameter takes a PV as an argument with an optional trailing character to specify whether to set ('y'), unset ('n'), or toggle ('t') the value. If no trailing character is given, it will set the flag. Synopsis: lvchange [--writemostly <PV>:{t|y|n}] [--writebehind <count>] vg/lv Example: lvchange --writemostly /dev/sdb1:y --writebehind 512 vg/raid1_lv The last character in the 'lv_attr' field is used to show whether a device has the WriteMostly flag set. It is signified with a 'w'. If the device has failed, the 'p'artial flag has priority. Example ("nosync" raid1 with mismatch_cnt and writemostly): [~]# lvs -a --segment vg LV VG Attr #Str Type SSize raid1 vg Rwi---r-m 2 raid1 500.00m [raid1_rimage_0] vg Iwi---r-- 1 linear 500.00m [raid1_rimage_1] vg Iwi---r-w 1 linear 500.00m [raid1_rmeta_0] vg ewi---r-- 1 linear 4.00m [raid1_rmeta_1] vg ewi---r-- 1 linear 4.00m Example (raid1 with mismatch_cnt, writemostly - but failed drive): [~]# lvs -a --segment vg LV VG Attr #Str Type SSize raid1 vg rwi---r-p 2 raid1 500.00m [raid1_rimage_0] vg Iwi---r-- 1 linear 500.00m [raid1_rimage_1] vg Iwi---r-p 1 linear 500.00m [raid1_rmeta_0] vg ewi---r-- 1 linear 4.00m [raid1_rmeta_1] vg ewi---r-p 1 linear 4.00m A new reportable field has been added for writebehind as well. If write-behind has not been set or the LV is not RAID1, the field will be blank. Example (writebehind is set): [~]# lvs -a -o name,attr,writebehind vg LV Attr WBehind lv rwi-a-r-- 512 [lv_rimage_0] iwi-aor-w [lv_rimage_1] iwi-aor-- [lv_rmeta_0] ewi-aor-- [lv_rmeta_1] ewi-aor-- Example (writebehind is not set): [~]# lvs -a -o name,attr,writebehind vg LV Attr WBehind lv rwi-a-r-- [lv_rimage_0] iwi-aor-w [lv_rimage_1] iwi-aor-- [lv_rmeta_0] ewi-aor-- [lv_rmeta_1] ewi-aor--
2013-04-15 22:59:46 +04:00
int dm_tree_node_add_raid_target_with_params(struct dm_tree_node *node,
uint64_t size,
const struct dm_tree_node_raid_params *p)
{
2014-11-08 17:33:17 +03:00
unsigned i;
struct load_segment *seg = NULL;
for (i = 0; i < DM_ARRAY_SIZE(_dm_segtypes) && !seg; ++i)
if (!strcmp(p->raid_type, _dm_segtypes[i].target))
if (!(seg = _add_segment(node,
_dm_segtypes[i].type, size)))
return_0;
if (!seg) {
log_error("Unsupported raid type %s.", p->raid_type);
return 0;
}
RAID: Add writemostly/writebehind support for RAID1 'lvchange' is used to alter a RAID 1 logical volume's write-mostly and write-behind characteristics. The '--writemostly' parameter takes a PV as an argument with an optional trailing character to specify whether to set ('y'), unset ('n'), or toggle ('t') the value. If no trailing character is given, it will set the flag. Synopsis: lvchange [--writemostly <PV>:{t|y|n}] [--writebehind <count>] vg/lv Example: lvchange --writemostly /dev/sdb1:y --writebehind 512 vg/raid1_lv The last character in the 'lv_attr' field is used to show whether a device has the WriteMostly flag set. It is signified with a 'w'. If the device has failed, the 'p'artial flag has priority. Example ("nosync" raid1 with mismatch_cnt and writemostly): [~]# lvs -a --segment vg LV VG Attr #Str Type SSize raid1 vg Rwi---r-m 2 raid1 500.00m [raid1_rimage_0] vg Iwi---r-- 1 linear 500.00m [raid1_rimage_1] vg Iwi---r-w 1 linear 500.00m [raid1_rmeta_0] vg ewi---r-- 1 linear 4.00m [raid1_rmeta_1] vg ewi---r-- 1 linear 4.00m Example (raid1 with mismatch_cnt, writemostly - but failed drive): [~]# lvs -a --segment vg LV VG Attr #Str Type SSize raid1 vg rwi---r-p 2 raid1 500.00m [raid1_rimage_0] vg Iwi---r-- 1 linear 500.00m [raid1_rimage_1] vg Iwi---r-p 1 linear 500.00m [raid1_rmeta_0] vg ewi---r-- 1 linear 4.00m [raid1_rmeta_1] vg ewi---r-p 1 linear 4.00m A new reportable field has been added for writebehind as well. If write-behind has not been set or the LV is not RAID1, the field will be blank. Example (writebehind is set): [~]# lvs -a -o name,attr,writebehind vg LV Attr WBehind lv rwi-a-r-- 512 [lv_rimage_0] iwi-aor-w [lv_rimage_1] iwi-aor-- [lv_rmeta_0] ewi-aor-- [lv_rmeta_1] ewi-aor-- Example (writebehind is not set): [~]# lvs -a -o name,attr,writebehind vg LV Attr WBehind lv rwi-a-r-- [lv_rimage_0] iwi-aor-w [lv_rimage_1] iwi-aor-- [lv_rmeta_0] ewi-aor-- [lv_rmeta_1] ewi-aor--
2013-04-15 22:59:46 +04:00
seg->region_size = p->region_size;
seg->stripe_size = p->stripe_size;
seg->area_count = 0;
memset(seg->rebuilds, 0, sizeof(seg->rebuilds));
seg->rebuilds[0] = p->rebuilds;
memset(seg->writemostly, 0, sizeof(seg->writemostly));
seg->writemostly[0] = p->writemostly;
RAID: Add writemostly/writebehind support for RAID1 'lvchange' is used to alter a RAID 1 logical volume's write-mostly and write-behind characteristics. The '--writemostly' parameter takes a PV as an argument with an optional trailing character to specify whether to set ('y'), unset ('n'), or toggle ('t') the value. If no trailing character is given, it will set the flag. Synopsis: lvchange [--writemostly <PV>:{t|y|n}] [--writebehind <count>] vg/lv Example: lvchange --writemostly /dev/sdb1:y --writebehind 512 vg/raid1_lv The last character in the 'lv_attr' field is used to show whether a device has the WriteMostly flag set. It is signified with a 'w'. If the device has failed, the 'p'artial flag has priority. Example ("nosync" raid1 with mismatch_cnt and writemostly): [~]# lvs -a --segment vg LV VG Attr #Str Type SSize raid1 vg Rwi---r-m 2 raid1 500.00m [raid1_rimage_0] vg Iwi---r-- 1 linear 500.00m [raid1_rimage_1] vg Iwi---r-w 1 linear 500.00m [raid1_rmeta_0] vg ewi---r-- 1 linear 4.00m [raid1_rmeta_1] vg ewi---r-- 1 linear 4.00m Example (raid1 with mismatch_cnt, writemostly - but failed drive): [~]# lvs -a --segment vg LV VG Attr #Str Type SSize raid1 vg rwi---r-p 2 raid1 500.00m [raid1_rimage_0] vg Iwi---r-- 1 linear 500.00m [raid1_rimage_1] vg Iwi---r-p 1 linear 500.00m [raid1_rmeta_0] vg ewi---r-- 1 linear 4.00m [raid1_rmeta_1] vg ewi---r-p 1 linear 4.00m A new reportable field has been added for writebehind as well. If write-behind has not been set or the LV is not RAID1, the field will be blank. Example (writebehind is set): [~]# lvs -a -o name,attr,writebehind vg LV Attr WBehind lv rwi-a-r-- 512 [lv_rimage_0] iwi-aor-w [lv_rimage_1] iwi-aor-- [lv_rmeta_0] ewi-aor-- [lv_rmeta_1] ewi-aor-- Example (writebehind is not set): [~]# lvs -a -o name,attr,writebehind vg LV Attr WBehind lv rwi-a-r-- [lv_rimage_0] iwi-aor-w [lv_rimage_1] iwi-aor-- [lv_rmeta_0] ewi-aor-- [lv_rmeta_1] ewi-aor--
2013-04-15 22:59:46 +04:00
seg->writebehind = p->writebehind;
seg->min_recovery_rate = p->min_recovery_rate;
seg->max_recovery_rate = p->max_recovery_rate;
RAID: Add writemostly/writebehind support for RAID1 'lvchange' is used to alter a RAID 1 logical volume's write-mostly and write-behind characteristics. The '--writemostly' parameter takes a PV as an argument with an optional trailing character to specify whether to set ('y'), unset ('n'), or toggle ('t') the value. If no trailing character is given, it will set the flag. Synopsis: lvchange [--writemostly <PV>:{t|y|n}] [--writebehind <count>] vg/lv Example: lvchange --writemostly /dev/sdb1:y --writebehind 512 vg/raid1_lv The last character in the 'lv_attr' field is used to show whether a device has the WriteMostly flag set. It is signified with a 'w'. If the device has failed, the 'p'artial flag has priority. Example ("nosync" raid1 with mismatch_cnt and writemostly): [~]# lvs -a --segment vg LV VG Attr #Str Type SSize raid1 vg Rwi---r-m 2 raid1 500.00m [raid1_rimage_0] vg Iwi---r-- 1 linear 500.00m [raid1_rimage_1] vg Iwi---r-w 1 linear 500.00m [raid1_rmeta_0] vg ewi---r-- 1 linear 4.00m [raid1_rmeta_1] vg ewi---r-- 1 linear 4.00m Example (raid1 with mismatch_cnt, writemostly - but failed drive): [~]# lvs -a --segment vg LV VG Attr #Str Type SSize raid1 vg rwi---r-p 2 raid1 500.00m [raid1_rimage_0] vg Iwi---r-- 1 linear 500.00m [raid1_rimage_1] vg Iwi---r-p 1 linear 500.00m [raid1_rmeta_0] vg ewi---r-- 1 linear 4.00m [raid1_rmeta_1] vg ewi---r-p 1 linear 4.00m A new reportable field has been added for writebehind as well. If write-behind has not been set or the LV is not RAID1, the field will be blank. Example (writebehind is set): [~]# lvs -a -o name,attr,writebehind vg LV Attr WBehind lv rwi-a-r-- 512 [lv_rimage_0] iwi-aor-w [lv_rimage_1] iwi-aor-- [lv_rmeta_0] ewi-aor-- [lv_rmeta_1] ewi-aor-- Example (writebehind is not set): [~]# lvs -a -o name,attr,writebehind vg LV Attr WBehind lv rwi-a-r-- [lv_rimage_0] iwi-aor-w [lv_rimage_1] iwi-aor-- [lv_rmeta_0] ewi-aor-- [lv_rmeta_1] ewi-aor--
2013-04-15 22:59:46 +04:00
seg->flags = p->flags;
return 1;
}
RAID: Add writemostly/writebehind support for RAID1 'lvchange' is used to alter a RAID 1 logical volume's write-mostly and write-behind characteristics. The '--writemostly' parameter takes a PV as an argument with an optional trailing character to specify whether to set ('y'), unset ('n'), or toggle ('t') the value. If no trailing character is given, it will set the flag. Synopsis: lvchange [--writemostly <PV>:{t|y|n}] [--writebehind <count>] vg/lv Example: lvchange --writemostly /dev/sdb1:y --writebehind 512 vg/raid1_lv The last character in the 'lv_attr' field is used to show whether a device has the WriteMostly flag set. It is signified with a 'w'. If the device has failed, the 'p'artial flag has priority. Example ("nosync" raid1 with mismatch_cnt and writemostly): [~]# lvs -a --segment vg LV VG Attr #Str Type SSize raid1 vg Rwi---r-m 2 raid1 500.00m [raid1_rimage_0] vg Iwi---r-- 1 linear 500.00m [raid1_rimage_1] vg Iwi---r-w 1 linear 500.00m [raid1_rmeta_0] vg ewi---r-- 1 linear 4.00m [raid1_rmeta_1] vg ewi---r-- 1 linear 4.00m Example (raid1 with mismatch_cnt, writemostly - but failed drive): [~]# lvs -a --segment vg LV VG Attr #Str Type SSize raid1 vg rwi---r-p 2 raid1 500.00m [raid1_rimage_0] vg Iwi---r-- 1 linear 500.00m [raid1_rimage_1] vg Iwi---r-p 1 linear 500.00m [raid1_rmeta_0] vg ewi---r-- 1 linear 4.00m [raid1_rmeta_1] vg ewi---r-p 1 linear 4.00m A new reportable field has been added for writebehind as well. If write-behind has not been set or the LV is not RAID1, the field will be blank. Example (writebehind is set): [~]# lvs -a -o name,attr,writebehind vg LV Attr WBehind lv rwi-a-r-- 512 [lv_rimage_0] iwi-aor-w [lv_rimage_1] iwi-aor-- [lv_rmeta_0] ewi-aor-- [lv_rmeta_1] ewi-aor-- Example (writebehind is not set): [~]# lvs -a -o name,attr,writebehind vg LV Attr WBehind lv rwi-a-r-- [lv_rimage_0] iwi-aor-w [lv_rimage_1] iwi-aor-- [lv_rmeta_0] ewi-aor-- [lv_rmeta_1] ewi-aor--
2013-04-15 22:59:46 +04:00
int dm_tree_node_add_raid_target(struct dm_tree_node *node,
uint64_t size,
const char *raid_type,
uint32_t region_size,
uint32_t stripe_size,
uint64_t rebuilds,
RAID: Add writemostly/writebehind support for RAID1 'lvchange' is used to alter a RAID 1 logical volume's write-mostly and write-behind characteristics. The '--writemostly' parameter takes a PV as an argument with an optional trailing character to specify whether to set ('y'), unset ('n'), or toggle ('t') the value. If no trailing character is given, it will set the flag. Synopsis: lvchange [--writemostly <PV>:{t|y|n}] [--writebehind <count>] vg/lv Example: lvchange --writemostly /dev/sdb1:y --writebehind 512 vg/raid1_lv The last character in the 'lv_attr' field is used to show whether a device has the WriteMostly flag set. It is signified with a 'w'. If the device has failed, the 'p'artial flag has priority. Example ("nosync" raid1 with mismatch_cnt and writemostly): [~]# lvs -a --segment vg LV VG Attr #Str Type SSize raid1 vg Rwi---r-m 2 raid1 500.00m [raid1_rimage_0] vg Iwi---r-- 1 linear 500.00m [raid1_rimage_1] vg Iwi---r-w 1 linear 500.00m [raid1_rmeta_0] vg ewi---r-- 1 linear 4.00m [raid1_rmeta_1] vg ewi---r-- 1 linear 4.00m Example (raid1 with mismatch_cnt, writemostly - but failed drive): [~]# lvs -a --segment vg LV VG Attr #Str Type SSize raid1 vg rwi---r-p 2 raid1 500.00m [raid1_rimage_0] vg Iwi---r-- 1 linear 500.00m [raid1_rimage_1] vg Iwi---r-p 1 linear 500.00m [raid1_rmeta_0] vg ewi---r-- 1 linear 4.00m [raid1_rmeta_1] vg ewi---r-p 1 linear 4.00m A new reportable field has been added for writebehind as well. If write-behind has not been set or the LV is not RAID1, the field will be blank. Example (writebehind is set): [~]# lvs -a -o name,attr,writebehind vg LV Attr WBehind lv rwi-a-r-- 512 [lv_rimage_0] iwi-aor-w [lv_rimage_1] iwi-aor-- [lv_rmeta_0] ewi-aor-- [lv_rmeta_1] ewi-aor-- Example (writebehind is not set): [~]# lvs -a -o name,attr,writebehind vg LV Attr WBehind lv rwi-a-r-- [lv_rimage_0] iwi-aor-w [lv_rimage_1] iwi-aor-- [lv_rmeta_0] ewi-aor-- [lv_rmeta_1] ewi-aor--
2013-04-15 22:59:46 +04:00
uint64_t flags)
{
2014-02-15 00:36:56 +04:00
struct dm_tree_node_raid_params params = {
.raid_type = raid_type,
.region_size = region_size,
.stripe_size = stripe_size,
.rebuilds = rebuilds,
.flags = flags
};
return dm_tree_node_add_raid_target_with_params(node, size, &params);
}
/*
* Version 2 of dm_tree_node_add_raid_target() allowing for:
*
* - maximum 253 legs in a raid set (MD kernel limitation)
* - delta_disks for disk add/remove reshaping
* - data_offset for out-of-place reshaping
* - data_copies to cope witth odd numbers of raid10 disks
*/
int dm_tree_node_add_raid_target_with_params_v2(struct dm_tree_node *node,
uint64_t size,
const struct dm_tree_node_raid_params_v2 *p)
{
unsigned i;
struct load_segment *seg = NULL;
for (i = 0; i < DM_ARRAY_SIZE(_dm_segtypes) && !seg; ++i)
if (!strcmp(p->raid_type, _dm_segtypes[i].target))
if (!(seg = _add_segment(node,
_dm_segtypes[i].type, size)))
return_0;
if (!seg) {
log_error("Unsupported raid type %s.", p->raid_type);
return 0;
}
seg->region_size = p->region_size;
seg->stripe_size = p->stripe_size;
seg->area_count = 0;
seg->delta_disks = p->delta_disks;
seg->data_offset = p->data_offset;
memcpy(seg->rebuilds, p->rebuilds, sizeof(seg->rebuilds));
memcpy(seg->writemostly, p->writemostly, sizeof(seg->writemostly));
seg->writebehind = p->writebehind;
seg->data_copies = p->data_copies;
seg->min_recovery_rate = p->min_recovery_rate;
seg->max_recovery_rate = p->max_recovery_rate;
seg->flags = p->flags;
return 1;
}
DM_EXPORT_NEW_SYMBOL(int, dm_tree_node_add_cache_target, 1_02_138)
(struct dm_tree_node *node,
uint64_t size,
uint64_t feature_flags, /* DM_CACHE_FEATURE_* */
const char *metadata_uuid,
const char *data_uuid,
const char *origin_uuid,
const char *policy_name,
const struct dm_config_node *policy_settings,
uint32_t data_block_size)
{
struct dm_config_node *cn;
struct load_segment *seg;
static const uint64_t _modemask =
DM_CACHE_FEATURE_PASSTHROUGH |
DM_CACHE_FEATURE_WRITETHROUGH |
DM_CACHE_FEATURE_WRITEBACK;
/* Detect unknown (bigger) feature bit */
if (feature_flags >= (DM_CACHE_FEATURE_METADATA2 * 2)) {
log_error("Unsupported cache's feature flags set " FMTu64 ".",
feature_flags);
return 0;
}
switch (feature_flags & _modemask) {
case DM_CACHE_FEATURE_PASSTHROUGH:
case DM_CACHE_FEATURE_WRITEBACK:
if (strcmp(policy_name, "cleaner") == 0) {
/* Enforce writethrough mode for cleaner policy */
feature_flags = ~_modemask;
feature_flags |= DM_CACHE_FEATURE_WRITETHROUGH;
}
/* Fall through */
case DM_CACHE_FEATURE_WRITETHROUGH:
break;
default:
log_error("Invalid cache's feature flag " FMTu64 ".",
feature_flags);
return 0;
}
if (data_block_size < DM_CACHE_MIN_DATA_BLOCK_SIZE) {
log_error("Data block size %u is lower then %u sectors.",
data_block_size, DM_CACHE_MIN_DATA_BLOCK_SIZE);
return 0;
}
if (data_block_size > DM_CACHE_MAX_DATA_BLOCK_SIZE) {
log_error("Data block size %u is higher then %u sectors.",
data_block_size, DM_CACHE_MAX_DATA_BLOCK_SIZE);
return 0;
}
if (!(seg = _add_segment(node, SEG_CACHE, size)))
return_0;
if (!(seg->pool = dm_tree_find_node_by_uuid(node->dtree,
data_uuid))) {
log_error("Missing cache's data uuid %s.",
data_uuid);
return 0;
}
if (!_link_tree_nodes(node, seg->pool))
return_0;
if (!(seg->metadata = dm_tree_find_node_by_uuid(node->dtree,
metadata_uuid))) {
log_error("Missing cache's metadata uuid %s.",
metadata_uuid);
return 0;
}
if (!_link_tree_nodes(node, seg->metadata))
return_0;
if (!(seg->origin = dm_tree_find_node_by_uuid(node->dtree,
origin_uuid))) {
log_error("Missing cache's origin uuid %s.",
metadata_uuid);
return 0;
}
if (!_link_tree_nodes(node, seg->origin))
return_0;
seg->data_block_size = data_block_size;
seg->flags = feature_flags;
seg->policy_name = policy_name;
seg->migration_threshold = 2048; /* Default migration threshold 1MiB */
/* FIXME: better validation missing */
if (policy_settings) {
if (!(seg->policy_settings = dm_config_clone_node_with_mem(node->dtree->mem, policy_settings, 0)))
return_0;
for (cn = seg->policy_settings->child; cn; cn = cn->sib) {
if (!cn->v || (cn->v->type != DM_CFG_INT)) {
/* For now only <key> = <int> pairs are supported */
log_error("Cache policy parameter %s is without integer value.", cn->key);
return 0;
}
if (strcmp(cn->key, "migration_threshold") == 0) {
seg->migration_threshold = cn->v->v.i;
cn->v = NULL; /* skip this entry */
} else
seg->policy_argc++;
}
}
/* Always some throughput available for cache to proceed */
if (seg->migration_threshold < data_block_size * 8)
seg->migration_threshold = data_block_size * 8;
return 1;
}
int dm_tree_node_add_replicator_target(struct dm_tree_node *node,
uint64_t size,
const char *rlog_uuid,
const char *rlog_type,
unsigned rsite_index,
dm_replicator_mode_t mode,
uint32_t async_timeout,
uint64_t fall_behind_data,
uint32_t fall_behind_ios)
{
log_error("Replicator segment is unsupported.");
return 0;
}
/* Appends device node to Replicator */
int dm_tree_node_add_replicator_dev_target(struct dm_tree_node *node,
uint64_t size,
const char *replicator_uuid,
uint64_t rdevice_index,
const char *rdev_uuid,
unsigned rsite_index,
const char *slog_uuid,
uint32_t slog_flags,
uint32_t slog_region_size)
{
log_error("Replicator targer is unsupported.");
return 0;
}
static struct load_segment *_get_single_load_segment(struct dm_tree_node *node,
unsigned type)
{
struct load_segment *seg;
if (!(seg = _get_last_load_segment(node)))
return_NULL;
/* Never used past _load_node(), so can test segment_count */
if (node->props.segment_count != 1) {
log_error("Node %s must have only one segment.",
_dm_segtypes[type].target);
return NULL;
}
if (seg->type != type) {
log_error("Node %s has segment type %s.",
_dm_segtypes[type].target,
_dm_segtypes[seg->type].target);
return NULL;
}
return seg;
}
2011-10-17 18:15:26 +04:00
static int _thin_validate_device_id(uint32_t device_id)
{
if (device_id > DM_THIN_MAX_DEVICE_ID) {
log_error("Device id %u is higher then %u.",
device_id, DM_THIN_MAX_DEVICE_ID);
return 0;
}
return 1;
}
int dm_tree_node_add_thin_pool_target(struct dm_tree_node *node,
uint64_t size,
uint64_t transaction_id,
const char *metadata_uuid,
const char *pool_uuid,
uint32_t data_block_size,
uint64_t low_water_mark,
unsigned skip_block_zeroing)
{
return dm_tree_node_add_thin_pool_target_v1(node, size, transaction_id,
metadata_uuid, pool_uuid,
data_block_size,
low_water_mark,
skip_block_zeroing,
1);
}
int dm_tree_node_add_thin_pool_target_v1(struct dm_tree_node *node,
uint64_t size,
uint64_t transaction_id,
const char *metadata_uuid,
const char *pool_uuid,
uint32_t data_block_size,
uint64_t low_water_mark,
unsigned skip_block_zeroing,
unsigned crop_metadata)
{
struct load_segment *seg, *mseg;
uint64_t devsize = 0;
if (data_block_size < DM_THIN_MIN_DATA_BLOCK_SIZE) {
log_error("Data block size %u is lower then %u sectors.",
data_block_size, DM_THIN_MIN_DATA_BLOCK_SIZE);
return 0;
}
if (data_block_size > DM_THIN_MAX_DATA_BLOCK_SIZE) {
log_error("Data block size %u is higher then %u sectors.",
data_block_size, DM_THIN_MAX_DATA_BLOCK_SIZE);
return 0;
}
if (!(seg = _add_segment(node, SEG_THIN_POOL, size)))
return_0;
if (!(seg->metadata = dm_tree_find_node_by_uuid(node->dtree, metadata_uuid))) {
log_error("Missing metadata uuid %s.", metadata_uuid);
return 0;
}
if (!_link_tree_nodes(node, seg->metadata))
return_0;
if (crop_metadata)
/* FIXME: more complex target may need more tweaks */
dm_list_iterate_items(mseg, &seg->metadata->props.segs) {
devsize += mseg->size;
if (devsize > DM_THIN_MAX_METADATA_SIZE) {
log_debug_activation("Ignoring %" PRIu64 " of device.",
devsize - DM_THIN_MAX_METADATA_SIZE);
mseg->size -= (devsize - DM_THIN_MAX_METADATA_SIZE);
devsize = DM_THIN_MAX_METADATA_SIZE;
/* FIXME: drop remaining segs */
}
}
if (!(seg->pool = dm_tree_find_node_by_uuid(node->dtree, pool_uuid))) {
log_error("Missing pool uuid %s.", pool_uuid);
return 0;
}
if (!_link_tree_nodes(node, seg->pool))
return_0;
/* Clean flag delay_resume_if_new - so corelog gets resumed */
seg->metadata->props.delay_resume_if_new = 0;
seg->pool->props.delay_resume_if_new = 0;
/* Preload must not resume extended running thin-pool before it's committed */
node->props.delay_resume_if_extended = 1;
/* Validate only transaction_id > 0 when activating thin-pool */
node->props.send_messages = transaction_id ? 1 : 0;
seg->transaction_id = transaction_id;
seg->low_water_mark = low_water_mark;
seg->data_block_size = data_block_size;
seg->skip_block_zeroing = skip_block_zeroing;
dm_list_init(&seg->thin_messages);
return 1;
}
int dm_tree_node_add_thin_pool_message(struct dm_tree_node *node,
dm_thin_message_t type,
uint64_t id1, uint64_t id2)
{
struct thin_message *tm;
struct load_segment *seg;
if (!(seg = _get_single_load_segment(node, SEG_THIN_POOL)))
return_0;
if (!(tm = dm_pool_zalloc(node->dtree->mem, sizeof (*tm)))) {
log_error("Failed to allocate thin message.");
return 0;
}
switch (type) {
case DM_THIN_MESSAGE_CREATE_SNAP:
/* If the thin origin is active, it must be suspend first! */
if (id1 == id2) {
log_error("Cannot use same device id for origin and its snapshot.");
return 0;
}
if (!_thin_validate_device_id(id1) ||
!_thin_validate_device_id(id2))
return_0;
tm->message.u.m_create_snap.device_id = id1;
tm->message.u.m_create_snap.origin_id = id2;
break;
case DM_THIN_MESSAGE_CREATE_THIN:
if (!_thin_validate_device_id(id1))
return_0;
tm->message.u.m_create_thin.device_id = id1;
tm->expected_errno = EEXIST;
break;
case DM_THIN_MESSAGE_DELETE:
if (!_thin_validate_device_id(id1))
return_0;
tm->message.u.m_delete.device_id = id1;
tm->expected_errno = ENODATA;
break;
case DM_THIN_MESSAGE_SET_TRANSACTION_ID:
if ((id1 + 1) != id2) {
log_error("New transaction id must be sequential.");
return 0; /* FIXME: Maybe too strict here? */
}
if (id2 != seg->transaction_id) {
log_error("Current transaction id is different from thin pool.");
return 0; /* FIXME: Maybe too strict here? */
}
tm->message.u.m_set_transaction_id.current_id = id1;
tm->message.u.m_set_transaction_id.new_id = id2;
break;
default:
log_error("Unsupported message type %d.", (int) type);
return 0;
}
tm->message.type = type;
dm_list_add(&seg->thin_messages, &tm->list);
thin: move pool messaging from resume to suspend Existing messaging intarface for thin-pool has a few 'weak' points: * Message were posted with each 'resume' operation, thus not allowing activation of thin-pool with the existing state. * Acceleration skipped suspend step has not worked in cluster, since clvmd resumes only nodes which are suspended (have proper lock state). * Resume may fail and code is not really designed to 'fail' in this phase (generic rule here is resume DOES NOT fail unless something serious is wrong and lvm2 tool usually doesn't handle recovery path in this case.) * Full thin-pool suspend happened, when taken a thin-volume snapshot. With this patch the new method relocates message passing into suspend state. This has a few drawbacks with current API, but overal it performs better and gives are more posibilities to deal with errors. Patch introduces a new logic for 'origin-only' suspend of thin-pool and this also relates to thin-volume when taking snapshot. When suspend_origin_only operation is invoked on a pool with queued messages then only those messages are posted to thin-pool and actual suspend of thin pool and data and metadata volume is skipped. This makes taking a snapshot of thin-volume lighter operation and avoids blocking of other unrelated active thin volumes. Also fail now happens in 'suspend' state where the 'Fail' is more expected and it is better handled through error paths. Activation of thin-pool is now not sending any message and leaves upto a tool to decided later how to finish unfinished double-commit transaction. Problem which needs some API improvements relates to the lvm2 tree construction. For the suspend tree we do not add target table line into the tree, but only a device is inserted into a tree. Current mechanism to attach messages for thin-pool requires the libdm to know about thin-pool target, so lvm2 currently takes assumption, node is really a thin-pool and fills in the table line for this node (which should be ensured by the PRELOAD phase, but it's a misuse of internal API) we would possibly need to be able to attach message to 'any' node. Other thing to notice - current messaging interface in thin-pool target requires to suspend thin volume origin first and then send a create message, but this could not have any 'nice' solution on lvm2 side and IMHO we should introduce something like 'create_after_resume' message. Patch also changes the moment, where lvm2 transaction id is increased. Now it happens only after successful finish of kernel transaction id change. This change was needed to handle properly activation of pool, which is in the middle of unfinished transaction, and also this corrects usage of thin-pool by external apps like Docker.
2015-07-01 14:31:37 +03:00
/* Higher value >1 identifies there are really some messages */
node->props.send_messages = 2;
return 1;
}
int dm_tree_node_set_thin_pool_discard(struct dm_tree_node *node,
unsigned ignore,
unsigned no_passdown)
{
struct load_segment *seg;
if (!(seg = _get_single_load_segment(node, SEG_THIN_POOL)))
return_0;
seg->ignore_discard = ignore;
seg->no_discard_passdown = no_passdown;
return 1;
}
int dm_tree_node_set_thin_pool_error_if_no_space(struct dm_tree_node *node,
unsigned error_if_no_space)
{
struct load_segment *seg;
if (!(seg = _get_single_load_segment(node, SEG_THIN_POOL)))
return_0;
seg->error_if_no_space = error_if_no_space;
return 1;
}
int dm_tree_node_set_thin_pool_read_only(struct dm_tree_node *node,
unsigned read_only)
{
struct load_segment *seg;
if (!(seg = _get_single_load_segment(node, SEG_THIN_POOL)))
return_0;
seg->read_only = read_only;
return 1;
}
int dm_tree_node_add_thin_target(struct dm_tree_node *node,
uint64_t size,
const char *pool_uuid,
uint32_t device_id)
{
struct dm_tree_node *pool;
struct load_segment *seg;
if (!(pool = dm_tree_find_node_by_uuid(node->dtree, pool_uuid))) {
log_error("Missing thin pool uuid %s.", pool_uuid);
return 0;
}
if (!_link_tree_nodes(node, pool))
return_0;
if (!_thin_validate_device_id(device_id))
return_0;
if (!(seg = _add_segment(node, SEG_THIN, size)))
return_0;
seg->pool = pool;
seg->device_id = device_id;
return 1;
}
int dm_tree_node_set_thin_external_origin(struct dm_tree_node *node,
const char *external_uuid)
{
struct dm_tree_node *external;
struct load_segment *seg;
if (!(seg = _get_single_load_segment(node, SEG_THIN)))
return_0;
if (!(external = dm_tree_find_node_by_uuid(node->dtree,
external_uuid))) {
log_error("Missing thin external origin uuid %s.",
external_uuid);
return 0;
}
if (!_link_tree_nodes(node, external))
return_0;
seg->external = external;
return 1;
}
2005-11-09 17:10:50 +03:00
static int _add_area(struct dm_tree_node *node, struct load_segment *seg, struct dm_tree_node *dev_node, uint64_t offset)
{
struct seg_area *area;
2005-11-09 17:10:50 +03:00
if (!(area = dm_pool_zalloc(node->dtree->mem, sizeof (*area)))) {
log_error("Failed to allocate target segment area.");
return 0;
}
area->dev_node = dev_node;
area->offset = offset;
dm_list_add(&seg->areas, &area->list);
seg->area_count++;
return 1;
}
2005-11-09 17:10:50 +03:00
int dm_tree_node_add_target_area(struct dm_tree_node *node,
2014-02-14 23:46:55 +04:00
const char *dev_name,
const char *uuid,
uint64_t offset)
{
struct load_segment *seg;
struct stat info;
2005-11-09 17:10:50 +03:00
struct dm_tree_node *dev_node;
if ((!dev_name || !*dev_name) && (!uuid || !*uuid)) {
2005-11-09 17:10:50 +03:00
log_error("dm_tree_node_add_target_area called without device");
return 0;
}
if (uuid) {
2005-11-09 17:10:50 +03:00
if (!(dev_node = dm_tree_find_node_by_uuid(node->dtree, uuid))) {
log_error("Couldn't find area uuid %s.", uuid);
return 0;
}
2005-11-09 17:10:50 +03:00
if (!_link_tree_nodes(node, dev_node))
return_0;
} else {
Add the ability to split an image from the mirror and track changes. ~> lvconvert --splitmirrors 1 --trackchanges vg/lv The '--trackchanges' option allows a user the ability to use an image of a RAID1 array for the purposes of temporary read-only access. The image can be merged back into the array at a later time and only the blocks that have changed in the array since the split will be resync'ed. This operation can be thought of as a partial split. The image is never completely extracted from the array, in that the array reserves the position the device occupied and tracks the differences between the array and the split image via a bitmap. The image itself is rendered read-only and the name (<LV>_rimage_*) cannot be changed. The user can complete the split (permanently splitting the image from the array) by re-issuing the 'lvconvert' command without the '--trackchanges' argument and specifying the '--name' argument. ~> lvconvert --splitmirrors 1 --name my_split vg/lv Merging the tracked image back into the array is done with the '--merge' option (included in a follow-on patch). ~> lvconvert --merge vg/lv_rimage_<n> The internal mechanics of this are relatively simple. The 'raid' device- mapper target allows for the specification of an empty slot in an array via '- -'. This is what will be used if a partial activation of an array is ever required. (It would also be possible to use 'error' targets in place of the '- -'.) If a RAID image is found to be both read-only and visible, then it is considered separate from the array and '- -' is used to hold it's position in the array. So, all that needs to be done to temporarily split an image from the array /and/ cause the kernel target's bitmap to track (aka "mark") changes made is to make the specified image visible and read-only. To merge the device back into the array, the image needs to be returned to the read/write state of the top-level LV and made invisible.
2011-08-18 23:38:26 +04:00
if (stat(dev_name, &info) < 0) {
log_error("Device %s not found.", dev_name);
return 0;
}
2011-08-19 21:02:48 +04:00
if (!S_ISBLK(info.st_mode)) {
log_error("Device %s is not a block device.", dev_name);
return 0;
}
/* FIXME Check correct macro use */
if (!(dev_node = _add_dev(node->dtree, node, MAJOR(info.st_rdev),
MINOR(info.st_rdev), 0, 0)))
2005-11-09 17:10:50 +03:00
return_0;
}
if (!(seg = _get_last_load_segment(node)))
return_0;
2005-11-09 17:10:50 +03:00
if (!_add_area(node, seg, dev_node, offset))
return_0;
return 1;
2005-10-25 23:09:41 +04:00
}
Add the ability to split an image from the mirror and track changes. ~> lvconvert --splitmirrors 1 --trackchanges vg/lv The '--trackchanges' option allows a user the ability to use an image of a RAID1 array for the purposes of temporary read-only access. The image can be merged back into the array at a later time and only the blocks that have changed in the array since the split will be resync'ed. This operation can be thought of as a partial split. The image is never completely extracted from the array, in that the array reserves the position the device occupied and tracks the differences between the array and the split image via a bitmap. The image itself is rendered read-only and the name (<LV>_rimage_*) cannot be changed. The user can complete the split (permanently splitting the image from the array) by re-issuing the 'lvconvert' command without the '--trackchanges' argument and specifying the '--name' argument. ~> lvconvert --splitmirrors 1 --name my_split vg/lv Merging the tracked image back into the array is done with the '--merge' option (included in a follow-on patch). ~> lvconvert --merge vg/lv_rimage_<n> The internal mechanics of this are relatively simple. The 'raid' device- mapper target allows for the specification of an empty slot in an array via '- -'. This is what will be used if a partial activation of an array is ever required. (It would also be possible to use 'error' targets in place of the '- -'.) If a RAID image is found to be both read-only and visible, then it is considered separate from the array and '- -' is used to hold it's position in the array. So, all that needs to be done to temporarily split an image from the array /and/ cause the kernel target's bitmap to track (aka "mark") changes made is to make the specified image visible and read-only. To merge the device back into the array, the image needs to be returned to the read/write state of the top-level LV and made invisible.
2011-08-18 23:38:26 +04:00
int dm_tree_node_add_null_area(struct dm_tree_node *node, uint64_t offset)
{
struct load_segment *seg;
if (!(seg = _get_last_load_segment(node)))
return_0;
Add the ability to split an image from the mirror and track changes. ~> lvconvert --splitmirrors 1 --trackchanges vg/lv The '--trackchanges' option allows a user the ability to use an image of a RAID1 array for the purposes of temporary read-only access. The image can be merged back into the array at a later time and only the blocks that have changed in the array since the split will be resync'ed. This operation can be thought of as a partial split. The image is never completely extracted from the array, in that the array reserves the position the device occupied and tracks the differences between the array and the split image via a bitmap. The image itself is rendered read-only and the name (<LV>_rimage_*) cannot be changed. The user can complete the split (permanently splitting the image from the array) by re-issuing the 'lvconvert' command without the '--trackchanges' argument and specifying the '--name' argument. ~> lvconvert --splitmirrors 1 --name my_split vg/lv Merging the tracked image back into the array is done with the '--merge' option (included in a follow-on patch). ~> lvconvert --merge vg/lv_rimage_<n> The internal mechanics of this are relatively simple. The 'raid' device- mapper target allows for the specification of an empty slot in an array via '- -'. This is what will be used if a partial activation of an array is ever required. (It would also be possible to use 'error' targets in place of the '- -'.) If a RAID image is found to be both read-only and visible, then it is considered separate from the array and '- -' is used to hold it's position in the array. So, all that needs to be done to temporarily split an image from the array /and/ cause the kernel target's bitmap to track (aka "mark") changes made is to make the specified image visible and read-only. To merge the device back into the array, the image needs to be returned to the read/write state of the top-level LV and made invisible.
2011-08-18 23:38:26 +04:00
2011-08-19 20:26:02 +04:00
switch (seg->type) {
case SEG_RAID0:
2016-07-02 00:20:54 +03:00
case SEG_RAID0_META:
2011-08-19 20:26:02 +04:00
case SEG_RAID1:
case SEG_RAID4:
case SEG_RAID5_N:
2011-08-19 20:26:02 +04:00
case SEG_RAID5_LA:
case SEG_RAID5_RA:
case SEG_RAID5_LS:
case SEG_RAID5_RS:
case SEG_RAID6_N_6:
2011-08-19 20:26:02 +04:00
case SEG_RAID6_ZR:
case SEG_RAID6_NR:
case SEG_RAID6_NC:
case SEG_RAID6_LS_6:
case SEG_RAID6_RS_6:
case SEG_RAID6_LA_6:
case SEG_RAID6_RA_6:
2011-08-19 20:26:02 +04:00
break;
default:
log_error("dm_tree_node_add_null_area() called on an unsupported segment type");
return 0;
}
Add the ability to split an image from the mirror and track changes. ~> lvconvert --splitmirrors 1 --trackchanges vg/lv The '--trackchanges' option allows a user the ability to use an image of a RAID1 array for the purposes of temporary read-only access. The image can be merged back into the array at a later time and only the blocks that have changed in the array since the split will be resync'ed. This operation can be thought of as a partial split. The image is never completely extracted from the array, in that the array reserves the position the device occupied and tracks the differences between the array and the split image via a bitmap. The image itself is rendered read-only and the name (<LV>_rimage_*) cannot be changed. The user can complete the split (permanently splitting the image from the array) by re-issuing the 'lvconvert' command without the '--trackchanges' argument and specifying the '--name' argument. ~> lvconvert --splitmirrors 1 --name my_split vg/lv Merging the tracked image back into the array is done with the '--merge' option (included in a follow-on patch). ~> lvconvert --merge vg/lv_rimage_<n> The internal mechanics of this are relatively simple. The 'raid' device- mapper target allows for the specification of an empty slot in an array via '- -'. This is what will be used if a partial activation of an array is ever required. (It would also be possible to use 'error' targets in place of the '- -'.) If a RAID image is found to be both read-only and visible, then it is considered separate from the array and '- -' is used to hold it's position in the array. So, all that needs to be done to temporarily split an image from the array /and/ cause the kernel target's bitmap to track (aka "mark") changes made is to make the specified image visible and read-only. To merge the device back into the array, the image needs to be returned to the read/write state of the top-level LV and made invisible.
2011-08-18 23:38:26 +04:00
if (!_add_area(node, seg, NULL, offset))
return_0;
return 1;
}
void dm_tree_node_set_callback(struct dm_tree_node *dnode,
dm_node_callback_fn cb, void *data)
{
dnode->callback = cb;
dnode->callback_data = data;
}
#if defined(GNU_SYMVER)
/*
* Backward compatible implementations.
*
* Keep these at the end of the file to make sure that
* no code in this file accidentally calls it.
*/
/* Backward compatible dm_tree_node_size_changed() implementations. */
DM_EXPORT_SYMBOL_BASE(dm_tree_node_size_changed)
int dm_tree_node_size_changed_base(const struct dm_tree_node *dnode);
int dm_tree_node_size_changed_base(const struct dm_tree_node *dnode)
{
/* Base does not make difference between smaller and bigger */
return dm_tree_node_size_changed(dnode) ? 1 : 0;
}
/*
* Retain ABI compatibility after adding the DM_CACHE_FEATURE_METADATA2
* in version 1.02.138.
*
* Binaries compiled against version 1.02.138 onwards will use
* the new function dm_tree_node_add_cache_target which detects unknown
* feature flags and returns error for them.
*/
DM_EXPORT_SYMBOL_BASE(dm_tree_node_add_cache_target)
int dm_tree_node_add_cache_target_base(struct dm_tree_node *node,
uint64_t size,
uint64_t feature_flags, /* DM_CACHE_FEATURE_* */
const char *metadata_uuid,
const char *data_uuid,
const char *origin_uuid,
const char *policy_name,
const struct dm_config_node *policy_settings,
uint32_t data_block_size);
int dm_tree_node_add_cache_target_base(struct dm_tree_node *node,
uint64_t size,
uint64_t feature_flags,
const char *metadata_uuid,
const char *data_uuid,
const char *origin_uuid,
const char *policy_name,
const struct dm_config_node *policy_settings,
uint32_t data_block_size)
{
/* Old version supported only these FEATURE bits, others were ignored so masked them */
static const uint64_t _mask =
DM_CACHE_FEATURE_WRITEBACK |
DM_CACHE_FEATURE_WRITETHROUGH |
DM_CACHE_FEATURE_PASSTHROUGH;
return dm_tree_node_add_cache_target(node, size, feature_flags & _mask,
metadata_uuid, data_uuid, origin_uuid,
policy_name, policy_settings, data_block_size);
}
#endif