/*
- * Copyright (C) 2005-2010 Red Hat, Inc. All rights reserved.
+ * Copyright (C) 2005-2011 Red Hat, Inc. All rights reserved.
*
* This file is part of the device-mapper userspace tools.
*
#define MAX_TARGET_PARAMSIZE 500000
-/* FIXME Fix interface so this is used only by LVM */
-#define UUID_PREFIX "LVM-"
-
#define REPLICATOR_LOCAL_SITE 0
/* Supported segment types */
SEG_SNAPSHOT_MERGE,
SEG_STRIPED,
SEG_ZERO,
+ SEG_THIN_POOL,
+ SEG_THIN,
SEG_RAID1,
SEG_RAID4,
SEG_RAID5_LA,
{ SEG_SNAPSHOT_MERGE, "snapshot-merge" },
{ SEG_STRIPED, "striped" },
{ SEG_ZERO, "zero"},
+ { SEG_THIN_POOL, "thin-pool"},
+ { SEG_THIN, "thin"},
{ SEG_RAID1, "raid1"},
{ SEG_RAID4, "raid4"},
{ SEG_RAID5_LA, "raid5_la"},
{ SEG_RAID6_ZR, "raid6_zr"},
{ SEG_RAID6_NR, "raid6_nr"},
{ SEG_RAID6_NC, "raid6_nc"},
+
+ /*
+ *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_LAST, NULL },
uint32_t flags; /* Replicator sync log flags */
};
+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;
+};
+
/* Replicator-log has a list of sites */
/* FIXME: maybe move to seg_area too? */
struct replicator_site {
uint64_t rdevice_index; /* Replicator-dev */
uint64_t rebuilds; /* raid */
+
+ struct dm_tree_node *metadata; /* Thin_pool */
+ struct dm_tree_node *pool; /* Thin_pool, 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 */
+ unsigned skip_block_zeroing; /* Thin_pool */
+ uint32_t device_id; /* Thin */
+
};
/* Per-device properties */
* avoid starting the mirror resync operation too early.
*/
unsigned delay_resume_if_new;
+
+ /* Send messages for this node in preload */
+ unsigned send_messages;
};
/* Two of these used to join two nodes with uses and used_by. */
* Note: only direct child is allowed
*/
struct dm_tree_node *presuspend_node;
+
+ /* Callback */
+ dm_node_callback_fn callback;
+ void *callback_data;
};
struct dm_tree {
struct dm_hash_table *uuids;
struct dm_tree_node root;
int skip_lockfs; /* 1 skips lockfs (for non-snapshots) */
- int no_flush; /* 1 sets noflush (mirrors/multipath) */
+ int no_flush; /* 1 sets noflush (mirrors/multipath) */
+ int retry_remove; /* 1 retries remove if not successful */
uint32_t cookie;
};
+/*
+ * Tree functions.
+ */
struct dm_tree *dm_tree_create(void)
{
+ struct dm_pool *dmem;
struct dm_tree *dtree;
- if (!(dtree = dm_zalloc(sizeof(*dtree)))) {
- log_error("dm_tree_create malloc failed");
+ 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;
}
dm_list_init(&dtree->root.used_by);
dtree->skip_lockfs = 0;
dtree->no_flush = 0;
-
- if (!(dtree->mem = dm_pool_create("dtree", 1024))) {
- log_error("dtree pool creation failed");
- dm_free(dtree);
- return NULL;
- }
+ dtree->mem = dmem;
if (!(dtree->devs = dm_hash_create(8))) {
log_error("dtree hash creation failed");
dm_pool_destroy(dtree->mem);
- dm_free(dtree);
return NULL;
}
log_error("dtree uuid hash creation failed");
dm_hash_destroy(dtree->devs);
dm_pool_destroy(dtree->mem);
- dm_free(dtree);
return NULL;
}
dm_hash_destroy(dtree->uuids);
dm_hash_destroy(dtree->devs);
dm_pool_destroy(dtree->mem);
- dm_free(dtree);
}
+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)
{
const char *uuid)
{
struct dm_tree_node *node;
+ const char *default_uuid_prefix;
+ size_t default_uuid_prefix_len;
if ((node = dm_hash_lookup(dtree->uuids, uuid)))
return node;
- if (strncmp(uuid, UUID_PREFIX, sizeof(UUID_PREFIX) - 1))
+ default_uuid_prefix = dm_uuid_prefix();
+ default_uuid_prefix_len = strlen(default_uuid_prefix);
+
+ if (strncmp(uuid, default_uuid_prefix, default_uuid_prefix_len))
return NULL;
- return dm_hash_lookup(dtree->uuids, uuid + sizeof(UUID_PREFIX) - 1);
+ return dm_hash_lookup(dtree->uuids, uuid + default_uuid_prefix_len);
+}
+
+void dm_tree_node_set_udev_flags(struct dm_tree_node *dnode, uint16_t udev_flags)
+
+{
+ struct dm_info *dinfo = &dnode->info;
+
+ if (udev_flags != dnode->udev_flags)
+ log_debug("Resetting %s (%" PRIu32 ":%" PRIu32
+ ") udev_flags from 0x%x to 0x%x",
+ dnode->name, dinfo->major, dinfo->minor,
+ 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))) {
+ stack; /* FIXME Is this normal? */
+ return 0;
+ }
+
+ 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,
+ 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,
+ 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;
}
static int _deps(struct dm_task **dmt, struct dm_pool *mem, uint32_t major, uint32_t minor,
- const char **name, const char **uuid,
+ const char **name, const char **uuid, unsigned inactive_table,
struct dm_info *info, struct dm_deps **deps)
{
memset(info, 0, sizeof(*info));
if (!dm_is_dm_major(major)) {
- *name = "";
- *uuid = "";
+ if (name)
+ *name = "";
+ if (uuid)
+ *uuid = "";
*deps = NULL;
info->major = major;
info->minor = minor;
- info->exists = 0;
- info->live_table = 0;
- info->inactive_table = 0;
- info->read_only = 0;
return 1;
}
goto failed;
}
+ if (inactive_table && !dm_task_query_inactive_table(*dmt)) {
+ log_error("_deps: failed to set inactive table for (%" PRIu32 ":%" PRIu32 ")",
+ major, minor);
+ goto failed;
+ }
+
if (!dm_task_run(*dmt)) {
log_error("_deps: task run failed for (%" PRIu32 ":%" PRIu32 ")",
major, minor);
}
if (!info->exists) {
- *name = "";
- *uuid = "";
+ if (name)
+ *name = "";
+ if (uuid)
+ *uuid = "";
*deps = NULL;
} else {
if (info->major != major) {
minor, info->minor);
goto failed;
}
- if (!(*name = dm_pool_strdup(mem, dm_task_get_name(*dmt)))) {
+ if (name && !(*name = dm_pool_strdup(mem, dm_task_get_name(*dmt)))) {
log_error("name pool_strdup failed");
goto failed;
}
- if (!(*uuid = dm_pool_strdup(mem, dm_task_get_uuid(*dmt)))) {
+ if (uuid && !(*uuid = dm_pool_strdup(mem, dm_task_get_uuid(*dmt)))) {
log_error("uuid pool_strdup failed");
goto failed;
}
return 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)
+/*
+ * 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 = 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;
+ struct dm_task *dmt;
+ int r;
- /* Already in tree? */
- if (!(node = _find_dm_tree_node(dtree, major, minor))) {
- if (!_deps(&dmt, dtree->mem, major, minor, &name, &uuid, &info, &deps))
- return_NULL;
+ if (!(dmt = dm_task_create(DM_DEVICE_INFO))) {
+ log_error("_info_by_dev: dm_task creation failed");
+ return 0;
+ }
- if (!(node = _create_dm_tree_node(dtree, name, uuid, &info,
- NULL, udev_flags)))
- goto_out;
- new = 1;
+ if (!dm_task_set_major(dmt, major) || !dm_task_set_minor(dmt, minor)) {
+ log_error("_info_by_dev: Failed to set device number");
+ dm_task_destroy(dmt);
+ return 0;
}
- if (!_link_tree_nodes(parent, node)) {
- node = NULL;
+ if (!with_open_count && !dm_task_no_open_count(dmt))
+ log_error("Failed to disable open_count");
+
+ if (!(r = dm_task_run(dmt)))
+ goto_out;
+
+ if (!(r = 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");
+ r = 0;
goto_out;
}
- /* If node was already in tree, no need to recurse. */
- if (!new)
- goto out;
+ if (uuid && !(*uuid = dm_pool_strdup(mem, dm_task_get_uuid(dmt)))) {
+ log_error("uuid pool_strdup failed");
+ r = 0;
+ goto_out;
+ }
- /* Can't recurse if not a mapped device or there are no dependencies */
- if (!node->info.exists || !deps->count) {
- if (!_add_to_bottomlevel(node)) {
+out:
+ dm_task_destroy(dmt);
+
+ return r;
+}
+
+static int _check_device_not_in_use(const char *name, struct dm_info *info)
+{
+ if (!info->exists)
+ return 1;
+
+ /* If sysfs is not used, use open_count information only. */
+ if (!*dm_sysfs_dir()) {
+ if (info->open_count) {
+ log_error("Device %s (%" PRIu32 ":%" PRIu32 ") in use",
+ name, info->major, info->minor);
+ return 0;
+ }
+
+ return 1;
+ }
+
+ if (dm_device_has_holders(info->major, info->minor)) {
+ log_error("Device %s (%" PRIu32 ":%" PRIu32 ") is used "
+ "by another device.", name, info->major, info->minor);
+ return 0;
+ }
+
+ if (dm_device_has_mounted_fs(info->major, info->minor)) {
+ log_error("Device %s (%" PRIu32 ":%" PRIu32 ") contains "
+ "a filesystem in use.", name, info->major, info->minor);
+ return 0;
+ }
+
+ return 1;
+}
+
+/* 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))) {
stack;
- node = NULL;
+ continue;
+ }
+
+ /* 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))) {
+ stack; /* FIXME Is this normal? */
+ return 0;
+ }
+
+ /* Refresh open_count */
+ if (!_info_by_dev(dinfo->major, dinfo->minor, 1, &info, NULL, NULL, NULL) ||
+ !info.exists)
+ continue;
+
+ if (info.open_count) {
+ log_debug("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;
}
- /* Add dependencies to tree */
- for (i = 0; i < deps->count; i++)
- if (!_add_dev(dtree, node, MAJOR(deps->device[i]),
- MINOR(deps->device[i]), udev_flags)) {
- node = NULL;
- goto_out;
- }
+ if (!dm_task_no_open_count(dmt))
+ log_error("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:
- if (dmt)
- dm_task_destroy(dmt);
+ dm_task_destroy(dmt);
- return node;
+ return r;
}
-static int _node_clear_table(struct dm_tree_node *dnode)
+static int _node_clear_table(struct dm_tree_node *dnode, uint16_t udev_flags)
{
- struct dm_task *dmt;
- struct dm_info *info;
- const char *name;
- int r;
+ struct dm_task *dmt = NULL, *deps_dmt = NULL;
+ struct dm_info *info, deps_info;
+ struct dm_deps *deps = NULL;
+ const char *name, *uuid;
+ const char *default_uuid_prefix;
+ size_t default_uuid_prefix_len;
+ uint32_t i;
+ int r = 0;
if (!(info = &dnode->info)) {
log_error("_node_clear_table failed: missing info");
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, NULL, NULL, 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);
- return 0;
+ 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);
- dm_task_destroy(dmt);
- return 0;
+ goto_out;
}
r = dm_task_run(dmt);
r = 0;
}
- dm_task_destroy(dmt);
+ 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;
+
+ if (!_info_by_dev(MAJOR(deps->device[i]), MINOR(deps->device[i]), 1,
+ &deps_info, dnode->dtree->mem, &name, &uuid))
+ continue;
+
+ /* 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;
+
+ 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);
return r;
}
-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)
+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;
const char *name2;
const char *uuid2;
+ 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 (!(name2 = dm_pool_strdup(dtree->mem, name))) {
return NULL;
}
- info.major = 0;
- info.minor = 0;
- info.exists = 0;
- info.live_table = 0;
- info.inactive_table = 0;
- info.read_only = 0;
+ memset(&info, 0, sizeof(info));
if (!(dnode = _create_dm_tree_node(dtree, name2, uuid2, &info,
context, 0)))
/* Do we need to rename node? */
if (!(dnode->props.new_name = dm_pool_strdup(dtree->mem, name))) {
log_error("name pool_strdup failed");
- return 0;
+ return NULL;
}
}
dnode->props.read_ahead = DM_READ_AHEAD_AUTO;
dnode->props.read_ahead_flags = 0;
- if (clear_inactive && !_node_clear_table(dnode))
+ if (clear_inactive && !_node_clear_table(dnode, udev_flags))
return_NULL;
dnode->context = context;
- dnode->udev_flags = 0;
+ dnode->udev_flags = udev_flags;
return dnode;
}
-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 *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)
{
- struct dm_tree_node *node;
+ 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)
+{
+ 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;
- if ((node = dm_tree_add_new_dev(dtree, name, uuid, major, minor, read_only,
- clear_inactive, context)))
- node->udev_flags = udev_flags;
+ /* 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;
- return node;
-}
+ if (!(node = _create_dm_tree_node(dtree, name, uuid, &info,
+ NULL, udev_flags)))
+ goto_out;
+ new = 1;
+ }
+ if (!_link_tree_nodes(parent, node)) {
+ node = NULL;
+ goto_out;
+ }
-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;
-}
+ /* If node was already in tree, no need to recurse. */
+ if (!new)
+ goto out;
-void dm_tree_node_set_presuspend_node(struct dm_tree_node *node,
- struct dm_tree_node *presuspend_node)
-{
- node->presuspend_node = presuspend_node;
+ /* 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++)
+ if (!_add_dev(dtree, node, MAJOR(deps->device[i]),
+ MINOR(deps->device[i]), udev_flags)) {
+ node = NULL;
+ goto_out;
+ }
+
+out:
+ if (dmt)
+ dm_task_destroy(dmt);
+
+ return node;
}
int dm_tree_add_dev(struct dm_tree *dtree, uint32_t major, uint32_t minor)
return _add_dev(dtree, &dtree->root, major, minor, udev_flags) ? 1 : 0;
}
-const char *dm_tree_node_get_name(const struct dm_tree_node *node)
+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)
{
- return node->info.exists ? node->name : "";
-}
+ struct dm_task *dmt;
+ int r = 0;
-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)
-{
- 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, UUID_PREFIX, sizeof(UUID_PREFIX) - 1))
- return 0;
+ log_verbose("Renaming %s (%" PRIu32 ":%" PRIu32 ") to %s", old_name, major, minor, new_name);
- if (strncmp(uuid_prefix, UUID_PREFIX, sizeof(UUID_PREFIX) - 1))
+ if (!(dmt = dm_task_create(DM_DEVICE_RENAME))) {
+ log_error("Rename dm_task creation failed for %s", old_name);
return 0;
-
- if (!strncmp(uuid, uuid_prefix + sizeof(UUID_PREFIX) - 1, uuid_prefix_len - (sizeof(UUID_PREFIX) - 1)))
- 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))) {
- stack; /* FIXME Is this normal? */
- return 0;
- }
-
- if (!dinfo->suspended)
- return 0;
+ if (!dm_task_set_name(dmt, old_name)) {
+ log_error("Failed to set name for %s rename.", old_name);
+ goto out;
}
- return 1;
-}
-
-/*
- * Set major and minor to zero for root of tree.
- */
-struct dm_tree_node *dm_tree_find_node(struct dm_tree *dtree,
- 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,
- const char *uuid)
-{
- if (!uuid || !*uuid)
- return &dtree->root;
+ if (!dm_task_set_newname(dmt, new_name))
+ goto_out;
- return _find_dm_tree_node_by_uuid(dtree, uuid);
-}
+ if (!dm_task_no_open_count(dmt))
+ log_error("Failed to disable open_count");
-/*
- * 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 (!dm_task_set_cookie(dmt, cookie, udev_flags))
+ goto out;
- if (inverted)
- use_list = &parent->used_by;
- else
- use_list = &parent->uses;
+ r = dm_task_run(dmt);
- if (!*dlink)
- *dlink = dm_list_first(use_list);
- else
- *dlink = dm_list_next(use_list, *dlink);
+out:
+ dm_task_destroy(dmt);
- return (*dlink) ? dm_list_item(*dlink, struct dm_tree_link)->node : NULL;
+ return r;
}
-/*
- * 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)
+/* 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;
+ int r = 0;
- if (!(dmt = dm_task_create(DM_DEVICE_INFO))) {
- log_error("_info_by_dev: dm_task creation failed");
+ log_verbose("Resuming %s (%" PRIu32 ":%" PRIu32 ")", name, major, minor);
+
+ if (!(dmt = dm_task_create(DM_DEVICE_RESUME))) {
+ log_debug("Suspend dm_task creation failed for %s.", name);
return 0;
}
+ /* FIXME Kernel should fill in name on return instead */
+ if (!dm_task_set_name(dmt, name)) {
+ log_debug("Failed to set device name for %s resumption.", name);
+ goto out;
+ }
+
if (!dm_task_set_major(dmt, major) || !dm_task_set_minor(dmt, minor)) {
- log_error("_info_by_dev: Failed to set device number");
- dm_task_destroy(dmt);
- return 0;
+ log_error("Failed to set device number for %s resumption.", name);
+ goto out;
}
- if (!with_open_count && !dm_task_no_open_count(dmt))
+ if (!dm_task_no_open_count(dmt))
log_error("Failed to disable open_count");
- if ((r = dm_task_run(dmt)))
- r = dm_task_get_info(dmt, info);
-
- dm_task_destroy(dmt);
-
- return r;
-}
+ if (!dm_task_set_read_ahead(dmt, read_ahead, read_ahead_flags))
+ log_error("Failed to set read ahead");
-/* 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;
+ if (!dm_task_set_cookie(dmt, cookie, udev_flags))
+ goto_out;
- /* Iterate through parents of this node */
- dm_list_iterate_items(dlink, &node->used_by) {
- if (!(uuid = dm_tree_node_get_uuid(dlink->node))) {
- stack;
- continue;
- }
+ if (!(r = dm_task_run(dmt)))
+ goto_out;
- /* Ignore if it doesn't belong to this VG */
- if (!_uuid_prefix_matches(uuid, uuid_prefix, uuid_prefix_len))
- continue;
+ if (already_suspended)
+ dec_suspended();
- if (!(dinfo = dm_tree_node_get_info(dlink->node))) {
- stack; /* FIXME Is this normal? */
- return 0;
- }
+ if (!(r = dm_task_get_info(dmt, newinfo)))
+ stack;
- /* Refresh open_count */
- if (!_info_by_dev(dinfo->major, dinfo->minor, 1, &info) ||
- !info.exists)
- continue;
-
- if (info.open_count) {
- log_debug("Node %s %d:%d has open_count %d", uuid_prefix,
- dinfo->major, dinfo->minor, info.open_count);
- return 0;
- }
- }
+out:
+ dm_task_destroy(dmt);
- return 1;
+ return r;
}
-static int _deactivate_node(const char *name, uint32_t major, uint32_t minor,
- uint32_t *cookie, uint16_t udev_flags)
+static int _suspend_node(const char *name, uint32_t major, uint32_t minor,
+ int skip_lockfs, int no_flush, struct dm_info *newinfo)
{
struct dm_task *dmt;
- int r = 0;
+ int r;
- log_verbose("Removing %s (%" PRIu32 ":%" PRIu32 ")", name, major, minor);
+ log_verbose("Suspending %s (%" PRIu32 ":%" PRIu32 ")%s%s",
+ name, major, minor,
+ skip_lockfs ? "" : " with filesystem sync",
+ no_flush ? "" : " with device flush");
- if (!(dmt = dm_task_create(DM_DEVICE_REMOVE))) {
- log_error("Deactivation dm_task creation failed for %s", name);
+ 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 deactivation", name);
- goto out;
+ log_error("Failed to set device number for %s suspension.", name);
+ dm_task_destroy(dmt);
+ return 0;
}
if (!dm_task_no_open_count(dmt))
log_error("Failed to disable open_count");
- if (!dm_task_set_cookie(dmt, cookie, udev_flags))
- goto out;
-
- r = dm_task_run(dmt);
+ if (skip_lockfs && !dm_task_skip_lockfs(dmt))
+ log_error("Failed to set skip_lockfs flag.");
- /* 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));
+ if (no_flush && !dm_task_no_flush(dmt))
+ log_error("Failed to set no_flush flag.");
- /* FIXME Remove node from tree or mark invalid? */
+ if ((r = dm_task_run(dmt))) {
+ inc_suspended();
+ r = dm_task_get_info(dmt, newinfo);
+ }
-out:
dm_task_destroy(dmt);
return r;
}
-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)
+static int _thin_pool_status_transaction_id(struct dm_tree_node *dnode, uint64_t *transaction_id)
{
struct dm_task *dmt;
int r = 0;
+ uint64_t start, length;
+ char *type = NULL;
+ char *params = NULL;
- 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 (!(dmt = dm_task_create(DM_DEVICE_STATUS)))
+ return_0;
- if (!dm_task_set_name(dmt, old_name)) {
- log_error("Failed to set name for %s rename.", old_name);
+ 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_set_newname(dmt, new_name))
+ if (!dm_task_run(dmt))
goto_out;
- if (!dm_task_no_open_count(dmt))
- log_error("Failed to disable open_count");
+ dm_get_next_target(dmt, NULL, &start, &length, &type, ¶ms);
- if (!dm_task_set_cookie(dmt, cookie, udev_flags))
+ if (type && (strcmp(type, "thin-pool") != 0)) {
+ log_error("Expected thin-pool target for %d:%d and got %s.",
+ dnode->info.major, dnode->info.minor, type);
+ goto out;
+ }
+
+ if (!params || (sscanf(params, "%" PRIu64, transaction_id) != 1)) {
+ log_error("Failed to parse transaction_id from %s.", params);
goto out;
+ }
- r = dm_task_run(dmt);
+ log_debug("Thin pool transaction id: %" PRIu64 " status: %s.", *transaction_id, params);
+ r = 1;
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)
+static int _thin_pool_node_message(struct dm_tree_node *dnode, struct thin_message *tm)
{
struct dm_task *dmt;
- int r = 0;
+ struct dm_thin_message *m = &tm->message;
+ char buf[64];
+ int r;
- log_verbose("Resuming %s (%" PRIu32 ":%" PRIu32 ")", name, major, minor);
+ 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;
+ default:
+ r = -1;
+ }
- if (!(dmt = dm_task_create(DM_DEVICE_RESUME))) {
- log_error("Suspend dm_task creation failed for %s", name);
+ if (r < 0) {
+ log_error("Failed to prepare message.");
return 0;
}
- /* FIXME Kernel should fill in name on return instead */
- if (!dm_task_set_name(dmt, name)) {
- log_error("Failed to set readahead device name for %s", name);
- goto out;
- }
+ r = 0;
- if (!dm_task_set_major(dmt, major) || !dm_task_set_minor(dmt, minor)) {
- log_error("Failed to set device number for %s resumption.", name);
+ if (!(dmt = dm_task_create(DM_DEVICE_TARGET_MSG)))
+ return_0;
+
+ if (!dm_task_set_major(dmt, dnode->info.major) ||
+ !dm_task_set_minor(dmt, dnode->info.minor)) {
+ log_error("Failed to set message major minor.");
goto out;
}
- if (!dm_task_no_open_count(dmt))
- log_error("Failed to disable open_count");
-
- if (!dm_task_set_read_ahead(dmt, read_ahead, read_ahead_flags))
- log_error("Failed to set read ahead");
+ if (!dm_task_set_message(dmt, buf))
+ goto_out;
- if (!dm_task_set_cookie(dmt, cookie, udev_flags))
- goto out;
+ /* Internal functionality of dm_task */
+ dmt->expected_errno = tm->expected_errno;
- if ((r = dm_task_run(dmt))) {
- if (already_suspended)
- dec_suspended();
- r = dm_task_get_info(dmt, newinfo);
- }
+ if (!dm_task_run(dmt))
+ goto_out;
+ r = 1;
out:
dm_task_destroy(dmt);
return r;
}
-static int _suspend_node(const char *name, uint32_t major, uint32_t minor,
- int skip_lockfs, int no_flush, struct dm_info *newinfo)
+static int _node_send_messages(struct dm_tree_node *dnode,
+ const char *uuid_prefix,
+ size_t uuid_prefix_len)
{
- struct dm_task *dmt;
- int r;
+ struct load_segment *seg;
+ struct thin_message *tmsg;
+ uint64_t trans_id;
+ const char *uuid;
- log_verbose("Suspending %s (%" PRIu32 ":%" PRIu32 ")%s%s",
- name, major, minor,
- skip_lockfs ? "" : " with filesystem sync",
- no_flush ? "" : " with device flush");
+ if (!dnode->info.exists || (dm_list_size(&dnode->props.segs) != 1))
+ return 1;
- if (!(dmt = dm_task_create(DM_DEVICE_SUSPEND))) {
- log_error("Suspend dm_task creation failed for %s", name);
- return 0;
- }
+ seg = dm_list_item(dm_list_last(&dnode->props.segs), struct load_segment);
+ if (seg->type != SEG_THIN_POOL)
+ return 1;
- if (!dm_task_set_major(dmt, major) || !dm_task_set_minor(dmt, minor)) {
- log_error("Failed to set device number for %s suspension.", name);
- dm_task_destroy(dmt);
- return 0;
- }
+ if (!(uuid = dm_tree_node_get_uuid(dnode)))
+ return_0;
- if (!dm_task_no_open_count(dmt))
- log_error("Failed to disable open_count");
+ if (!_uuid_prefix_matches(uuid, uuid_prefix, uuid_prefix_len)) {
+ log_debug("UUID \"%s\" does not match.", uuid);
+ return 1;
+ }
- if (skip_lockfs && !dm_task_skip_lockfs(dmt))
- log_error("Failed to set skip_lockfs flag.");
+ if (!_thin_pool_status_transaction_id(dnode, &trans_id))
+ goto_bad;
- if (no_flush && !dm_task_no_flush(dmt))
- log_error("Failed to set no_flush flag.");
+ if (trans_id == seg->transaction_id)
+ return 1; /* In sync - skip messages */
- if ((r = dm_task_run(dmt))) {
- inc_suspended();
- r = dm_task_get_info(dmt, newinfo);
+ if (trans_id != (seg->transaction_id - 1)) {
+ log_error("Thin pool transaction_id=%" PRIu64 ", while expected: %" PRIu64 ".",
+ trans_id, seg->transaction_id - 1);
+ goto bad; /* Nothing to send */
}
- dm_task_destroy(dmt);
+ dm_list_iterate_items(tmsg, &seg->thin_messages)
+ if (!(_thin_pool_node_message(dnode, tmsg)))
+ goto_bad;
- return r;
+ return 1;
+bad:
+ /* Try to deactivate */
+ if (!(dm_tree_deactivate_children(dnode, uuid_prefix, uuid_prefix_len)))
+ log_error("Failed to deactivate %s", dnode->name);
+
+ return 0;
}
/*
continue;
/* Refresh open_count */
- if (!_info_by_dev(dinfo->major, dinfo->minor, 1, &info) ||
+ if (!_info_by_dev(dinfo->major, dinfo->minor, 1, &info, NULL, NULL, NULL) ||
!info.exists)
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) {
+ log_error("Unable to deactivate open %s (%" PRIu32
+ ":%" PRIu32 ")", 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 (info.open_count ||
- (child->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 */
continue;
if (!_deactivate_node(name, info.major, info.minor,
- &child->dtree->cookie, child->udev_flags)) {
+ &child->dtree->cookie, child->udev_flags,
+ (level == 0) ? child->dtree->retry_remove : 0)) {
log_error("Unable to deactivate %s (%" PRIu32
":%" PRIu32 ")", name, info.major,
info.minor);
} else if (info.suspended)
dec_suspended();
- if (dm_tree_node_num_children(child, 0)) {
- if (!_dm_tree_deactivate_children(child, uuid_prefix, uuid_prefix_len, level + 1))
- return_0;
- }
+ if (child->callback &&
+ !child->callback(child, DM_NODE_CALLBACK_DEACTIVATED,
+ child->callback_data))
+ stack;
+ // FIXME: We need to let lvremove pass,
+ // so for now deactivation ignores check result
+ //r = 0; // FIXME: _node_clear_table() without callback ?
+
+ if (dm_tree_node_num_children(child, 0) &&
+ !_dm_tree_deactivate_children(child, uuid_prefix, uuid_prefix_len, level + 1))
+ return_0;
}
return r;
return _dm_tree_deactivate_children(dnode, uuid_prefix, uuid_prefix_len, 0);
}
-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;
-}
-
int dm_tree_suspend_children(struct dm_tree_node *dnode,
const char *uuid_prefix,
size_t uuid_prefix_len)
if (!_children_suspended(child, 1, uuid_prefix, uuid_prefix_len))
continue;
- if (!_info_by_dev(dinfo->major, dinfo->minor, 0, &info) ||
+ if (!_info_by_dev(dinfo->major, dinfo->minor, 0, &info, NULL, NULL, NULL) ||
!info.exists || info.suspended)
continue;
for (priority = 0; priority < 3; priority++) {
while ((child = dm_tree_next_child(&handle, dnode, 0))) {
+ if (priority != child->activation_priority)
+ continue;
+
if (!(uuid = dm_tree_node_get_uuid(child))) {
stack;
continue;
if (!_uuid_prefix_matches(uuid, uuid_prefix, uuid_prefix_len))
continue;
- if (priority != child->activation_priority)
- continue;
-
if (!(name = dm_tree_node_get_name(child))) {
stack;
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 && dnode->props.send_messages &&
+ !(r = _node_send_messages(dnode, uuid_prefix, uuid_prefix_len)))
+ stack;
+
handle = NULL;
return r;
uint64_t *seg_start, char *params,
size_t paramsize)
{
- uint32_t i, *tmp;
+ uint32_t i;
int param_count = 1; /* mandatory 'chunk size'/'stripe size' arg */
int pos = 0;
if (seg->region_size)
param_count += 2;
- tmp = (uint32_t *)(&seg->rebuilds); /* rebuilds is 64-bit */
- param_count += 2 * hweight32(tmp[0]);
- param_count += 2 * hweight32(tmp[1]);
+ /* rebuilds is 64-bit */
+ param_count += 2 * hweight32(seg->rebuilds & 0xFFFFFFFF);
+ param_count += 2 * hweight32(seg->rebuilds >> 32);
if ((seg->type == SEG_RAID1) && seg->stripe_size)
log_error("WARNING: Ignoring RAID1 stripe size");
int r;
int target_type_is_raid = 0;
char originbuf[DM_FORMAT_DEV_BUFSIZE], cowbuf[DM_FORMAT_DEV_BUFSIZE];
+ char pool[DM_FORMAT_DEV_BUFSIZE], metadata[DM_FORMAT_DEV_BUFSIZE];
switch(seg->type) {
case SEG_ERROR:
return_0;
break;
+ case SEG_THIN_POOL:
+ 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 " %s", metadata, pool,
+ seg->data_block_size, seg->low_water_mark,
+ seg->skip_block_zeroing ? "1 skip_block_zeroing" : "0");
+ break;
+ case SEG_THIN:
+ if (!_build_dev_string(pool, sizeof(pool), seg->pool))
+ return_0;
+ EMIT_PARAMS(pos, "%s %d", pool, seg->device_id);
+ break;
}
switch(seg->type) {
case SEG_SNAPSHOT_ORIGIN:
case SEG_SNAPSHOT_MERGE:
case SEG_ZERO:
+ case SEG_THIN_POOL:
+ case SEG_THIN:
break;
case SEG_CRYPT:
case SEG_LINEAR:
existing_table_size = dm_task_get_existing_table_size(dmt);
if ((dnode->props.size_changed =
(existing_table_size == seg_start) ? 0 : 1)) {
- log_debug("Table size changed from %" PRIu64 " to %"
- PRIu64 " for %s", existing_table_size,
- seg_start, dnode->name);
/*
* 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("Table size changed from %" PRIu64 " to %"
+ PRIu64 " for %s.%s", existing_table_size,
+ seg_start, dnode->name,
+ dnode->props.size_changed ? "" : " (Ignoring.)");
}
}
return_0;
/* FIXME Cope if name exists with no uuid? */
- if (!child->info.exists) {
- if (!_create_node(child)) {
- stack;
- return 0;
- }
- }
+ if (!child->info.exists && !_create_node(child))
+ return_0;
- if (!child->info.inactive_table && child->props.segment_count) {
- if (!_load_node(child)) {
- stack;
- return 0;
- }
- }
+ if (!child->info.inactive_table &&
+ child->props.segment_count &&
+ !_load_node(child))
+ return_0;
/* Propagate device size change change */
if (child->props.size_changed)
/* Update cached info */
child->info = newinfo;
-
/*
* Prepare for immediate synchronization with udev and flush all stacked
* dev node operations if requested by immediate_dev_node property. But
*/
if (child->props.immediate_dev_node)
update_devs_flag = 1;
-
}
- handle = NULL;
-
- if (update_devs_flag) {
+ if (update_devs_flag ||
+ (!dnode->info.exists && dnode->callback)) {
if (!dm_udev_wait(dm_tree_get_cookie(dnode)))
stack;
dm_tree_set_cookie(dnode, 0);
+
+ if (!dnode->info.exists && dnode->callback &&
+ !dnode->callback(child, DM_NODE_CALLBACK_PRELOADED,
+ dnode->callback_data))
+ return_0;
}
return r;
uint32_t region_size,
uint32_t stripe_size,
uint64_t rebuilds,
- uint64_t reserved2)
+ uint64_t flags)
{
int i;
struct load_segment *seg = NULL;
seg->stripe_size = stripe_size;
seg->area_count = 0;
seg->rebuilds = rebuilds;
+ seg->flags = flags;
return 1;
}
return 1;
}
+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)
+{
+ struct load_segment *seg, *mseg;
+ uint64_t devsize = 0;
+ /*
+ * Max supported size for thin pool metadata device
+ * Limitation is hardcoded into kernel and bigger
+ * device size is not accepted. (16978542592)
+ */
+ const uint64_t max_metadata_size =
+ 255ULL * (1 << 14) * (4096 / (1 << 9)) - 256 * 1024;
+
+ 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;
+
+ /* FIXME: more complex target may need more tweaks */
+ dm_list_iterate_items(mseg, &seg->metadata->props.segs) {
+ devsize += mseg->size;
+ if (devsize > max_metadata_size) {
+ log_debug("Ignoring %" PRIu64 " of device.",
+ devsize - max_metadata_size);
+ mseg->size -= (devsize - max_metadata_size);
+ devsize = 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;
+
+ node->props.send_messages = 1;
+ 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 load_segment *seg;
+ struct thin_message *tm;
+
+ if (node->props.segment_count != 1) {
+ log_error("Thin pool node must have only one segment.");
+ return 0;
+ }
+
+ seg = dm_list_item(dm_list_last(&node->props.segs), struct load_segment);
+ if (seg->type != SEG_THIN_POOL) {
+ log_error("Thin pool node has segment type %s.",
+ dm_segtypes[seg->type].target);
+ 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);
+
+ 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_get_status_thin_pool(struct dm_pool *mem, const char *params,
+ struct dm_status_thin_pool **status)
+{
+ struct dm_status_thin_pool *s;
+
+ if (!(s = dm_pool_zalloc(mem, sizeof(struct dm_status_thin_pool)))) {
+ log_error("Failed to allocate thin_pool status structure.");
+ return 0;
+ }
+
+ /* FIXME: add support for held metadata root */
+ if (sscanf(params, "%" PRIu64 " %" PRIu64 "/%" PRIu64 " %" PRIu64 "/%" PRIu64,
+ &s->transaction_id,
+ &s->used_metadata_blocks,
+ &s->total_metadata_blocks,
+ &s->used_data_blocks,
+ &s->total_data_blocks) != 5) {
+ log_error("Failed to parse thin pool params: %s.", params);
+ return 0;
+ }
+
+ *status = s;
+
+ return 1;
+}
+
+int dm_get_status_thin(struct dm_pool *mem, const char *params,
+ struct dm_status_thin **status)
+{
+ struct dm_status_thin *s;
+
+ if (!(s = dm_pool_zalloc(mem, sizeof(struct dm_status_thin)))) {
+ log_error("Failed to allocate thin status structure.");
+ return 0;
+ }
+
+ if (strchr(params, '-')) {
+ s->mapped_sectors = 0;
+ s->highest_mapped_sector = 0;
+ } else if (sscanf(params, "%" PRIu64 " %" PRIu64,
+ &s->mapped_sectors,
+ &s->highest_mapped_sector) != 2) {
+ log_error("Failed to parse thin params: %s.", params);
+ return 0;
+ }
+
+ *status = s;
+
+ return 1;
+}
+
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;
return 1;
}
-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)
+void dm_tree_node_set_callback(struct dm_tree_node *dnode,
+ dm_node_callback_fn cb, void *data)
{
- return node->dtree->cookie;
+ dnode->callback = cb;
+ dnode->callback_data = data;
}