1
0
mirror of git://sourceware.org/git/lvm2.git synced 2024-12-21 13:34:40 +03:00
lvm2/lib/activate/activate.c

2551 lines
64 KiB
C
Raw Normal View History

/*
2008-01-30 17:00:02 +03:00
* Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.
2016-04-27 01:01:08 +03:00
* Copyright (C) 2004-2016 Red Hat, Inc. All rights reserved.
*
2004-03-30 23:35:44 +04:00
* This file is part of LVM2.
*
* 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.
2004-03-30 23:35:44 +04:00
*
* You should have received a copy of the GNU Lesser General Public License
2004-03-30 23:35:44 +04:00
* along with this program; if not, write to the Free Software Foundation,
* Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
2002-11-18 17:01:16 +03:00
#include "lib.h"
2001-10-16 20:25:28 +04:00
#include "metadata.h"
#include "activate.h"
#include "memlock.h"
2001-11-15 17:27:51 +03:00
#include "display.h"
2001-11-12 15:20:58 +03:00
#include "fs.h"
#include "lvm-exec.h"
#include "lvm-file.h"
#include "lvm-string.h"
#include "toolcontext.h"
#include "dev_manager.h"
2004-03-08 21:13:22 +03:00
#include "str_list.h"
2004-05-05 01:25:57 +04:00
#include "config.h"
2005-12-02 23:35:07 +03:00
#include "segtype.h"
#include "sharedlib.h"
#include <limits.h>
#include <fcntl.h>
#include <unistd.h>
#define _skip(fmt, args...) log_very_verbose("Skipping: " fmt , ## args)
int lvm1_present(struct cmd_context *cmd)
{
static char path[PATH_MAX];
2006-08-21 16:54:53 +04:00
if (dm_snprintf(path, sizeof(path), "%s/lvm/global", cmd->proc_dir)
< 0) {
log_error("LVM1 proc global snprintf failed");
return 0;
}
2014-03-24 13:01:44 +04:00
return (path_exists(path)) ? 1 : 0;
}
int list_segment_modules(struct dm_pool *mem, const struct lv_segment *seg,
struct dm_list *modules)
{
unsigned int s;
struct lv_segment *seg2, *snap_seg;
struct dm_list *snh;
if (seg->segtype->ops->modules_needed &&
!seg->segtype->ops->modules_needed(mem, seg, modules)) {
log_error("module string allocation failed");
return 0;
}
if (lv_is_origin(seg->lv))
dm_list_iterate(snh, &seg->lv->snapshot_segs)
if (!list_lv_modules(mem,
dm_list_struct_base(snh,
struct lv_segment,
origin_list)->cow,
modules))
return_0;
if (lv_is_cow(seg->lv)) {
snap_seg = find_snapshot(seg->lv);
if (snap_seg->segtype->ops->modules_needed &&
!snap_seg->segtype->ops->modules_needed(mem, snap_seg,
modules)) {
log_error("snap_seg module string allocation failed");
return 0;
}
}
for (s = 0; s < seg->area_count; s++) {
switch (seg_type(seg, s)) {
case AREA_LV:
seg2 = find_seg_by_le(seg_lv(seg, s), seg_le(seg, s));
if (seg2 && !list_segment_modules(mem, seg2, modules))
return_0;
break;
case AREA_PV:
case AREA_UNASSIGNED:
;
}
}
return 1;
}
int list_lv_modules(struct dm_pool *mem, const struct logical_volume *lv,
struct dm_list *modules)
{
struct lv_segment *seg;
dm_list_iterate_items(seg, &lv->segments)
if (!list_segment_modules(mem, seg, modules))
return_0;
return 1;
}
static int _lv_passes_volumes_filter(struct cmd_context *cmd, const struct logical_volume *lv,
const struct dm_config_node *cn, const int cfg_id)
{
const struct dm_config_value *cv;
const char *str;
static char config_path[PATH_MAX];
size_t len = strlen(lv->vg->name);
config_def_get_path(config_path, sizeof(config_path), cfg_id);
log_verbose("%s configuration setting defined: "
"Checking the list to match %s.",
config_path, display_lvname(lv));
for (cv = cn->v; cv; cv = cv->next) {
if (cv->type == DM_CFG_EMPTY_ARRAY)
goto out;
if (cv->type != DM_CFG_STRING) {
log_print_unless_silent("Ignoring invalid string in config file %s.",
config_path);
continue;
}
str = cv->v.str;
if (!*str) {
log_print_unless_silent("Ignoring empty string in config file %s.",
config_path);
continue;
}
/* Tag? */
if (*str == '@') {
str++;
if (!*str) {
log_print_unless_silent("Ignoring empty tag in config file %s",
config_path);
continue;
}
/* If any host tag matches any LV or VG tag, activate */
if (!strcmp(str, "*")) {
if (str_list_match_list(&cmd->tags, &lv->tags, NULL)
|| str_list_match_list(&cmd->tags,
&lv->vg->tags, NULL))
return 1;
else
continue;
}
/* If supplied tag matches LV or VG tag, activate */
if (str_list_match_item(&lv->tags, str) ||
str_list_match_item(&lv->vg->tags, str))
return 1;
else
continue;
}
/* If supplied name is vgname[/lvname] */
if ((strncmp(str, lv->vg->name, len) == 0) &&
(!str[len] ||
((str[len] == '/') &&
!strcmp(str + len + 1, lv->name))))
return 1;
}
out:
log_verbose("No item supplied in %s configuration setting matches %s.",
config_path, display_lvname(lv));
return 0;
}
int lv_passes_auto_activation_filter(struct cmd_context *cmd, struct logical_volume *lv)
{
const struct dm_config_node *cn;
if (!(cn = find_config_tree_array(cmd, activation_auto_activation_volume_list_CFG, NULL))) {
log_verbose("activation/auto_activation_volume_list configuration setting "
"not defined: All logical volumes will be auto-activated.");
return 1;
}
return _lv_passes_volumes_filter(cmd, lv, cn, activation_auto_activation_volume_list_CFG);
}
#ifndef DEVMAPPER_SUPPORT
void set_activation(int act, int silent)
{
static int warned = 0;
if (warned || !act)
return;
log_error("Compiled without libdevmapper support. "
"Can't enable activation.");
warned = 1;
}
int activation(void)
{
return 0;
}
int library_version(char *version, size_t size)
{
return 0;
}
int driver_version(char *version, size_t size)
{
return 0;
}
int target_version(const char *target_name, uint32_t *maj,
uint32_t *min, uint32_t *patchlevel)
{
return 0;
}
int target_present(struct cmd_context *cmd, const char *target_name,
int use_modprobe)
2004-03-26 22:52:09 +03:00
{
return 0;
}
int lvm_dm_prefix_check(int major, int minor, const char *prefix)
{
return 0;
}
int lv_info(struct cmd_context *cmd, const struct logical_volume *lv, int use_layer,
struct lvinfo *info, int with_open_count, int with_read_ahead)
{
return 0;
}
int lv_info_by_lvid(struct cmd_context *cmd, const char *lvid_s, int use_layer,
struct lvinfo *info, int with_open_count, int with_read_ahead)
2004-03-26 18:35:01 +03:00
{
return 0;
}
int lv_info_with_seg_status(struct cmd_context *cmd, const struct logical_volume *lv,
const struct lv_segment *lv_seg, int use_layer,
struct lv_with_info_and_seg_status *status,
int with_open_count, int with_read_ahead)
{
return 0;
}
int lv_status(struct cmd_context *cmd, const struct lv_segment *lv_seg,
int use_layer, struct lv_seg_status *lv_seg_status)
{
return 0;
}
int lv_cache_status(const struct logical_volume *cache_lv,
struct lv_status_cache **status)
{
return 0;
}
int lv_check_not_in_use(const struct logical_volume *lv, int error_if_used)
{
return 0;
}
int lv_snapshot_percent(const struct logical_volume *lv, dm_percent_t *percent)
{
return 0;
}
int lv_mirror_percent(struct cmd_context *cmd, const struct logical_volume *lv,
int wait, dm_percent_t *percent, uint32_t *event_nr)
2003-04-30 19:26:25 +04:00
{
return 0;
}
int lv_raid_percent(const struct logical_volume *lv, dm_percent_t *percent)
{
return 0;
}
int lv_raid_dev_health(const struct logical_volume *lv, char **dev_health)
{
return 0;
}
RAID: Add scrubbing support for RAID LVs New options to 'lvchange' allow users to scrub their RAID LVs. Synopsis: lvchange --syncaction {check|repair} vg/raid_lv RAID scrubbing is the process of reading all the data and parity blocks in an array and checking to see whether they are coherent. 'lvchange' can now initaite the two scrubbing operations: "check" and "repair". "check" will go over the array and recored the number of discrepancies but not repair them. "repair" will correct the discrepancies as it finds them. 'lvchange --syncaction repair vg/raid_lv' is not to be confused with 'lvconvert --repair vg/raid_lv'. The former initiates a background synchronization operation on the array, while the latter is designed to repair/replace failed devices in a mirror or RAID logical volume. Additional reporting has been added for 'lvs' to support the new operations. Two new printable fields (which are not printed by default) have been added: "syncaction" and "mismatches". These can be accessed using the '-o' option to 'lvs', like: lvs -o +syncaction,mismatches vg/lv "syncaction" will print the current synchronization operation that the RAID volume is performing. It can be one of the following: - idle: All sync operations complete (doing nothing) - resync: Initializing an array or recovering after a machine failure - recover: Replacing a device in the array - check: Looking for array inconsistencies - repair: Looking for and repairing inconsistencies The "mismatches" field with print the number of descrepancies found during a check or repair operation. The 'Cpy%Sync' field already available to 'lvs' will print the progress of any of the above syncactions, including check and repair. Finally, the lv_attr field has changed to accomadate the scrubbing operations as well. The role of the 'p'artial character in the lv_attr report field as expanded. "Partial" is really an indicator for the health of a logical volume and it makes sense to extend this include other health indicators as well, specifically: 'm'ismatches: Indicates that there are discrepancies in a RAID LV. This character is shown after a scrubbing operation has detected that portions of the RAID are not coherent. 'r'efresh : Indicates that a device in a RAID array has suffered a failure and the kernel regards it as failed - even though LVM can read the device label and considers the device to be ok. The LV should be 'r'efreshed to notify the kernel that the device is now available, or the device should be 'r'eplaced if it is suspected of failing.
2013-04-12 00:33:59 +04:00
int lv_raid_mismatch_count(const struct logical_volume *lv, uint64_t *cnt)
{
return 0;
}
int lv_raid_sync_action(const struct logical_volume *lv, char **sync_action)
{
return 0;
}
int lv_raid_message(const struct logical_volume *lv, const char *msg)
{
return 0;
}
int lv_thin_pool_percent(const struct logical_volume *lv, int metadata,
dm_percent_t *percent)
{
return 0;
}
int lv_thin_percent(const struct logical_volume *lv, int mapped,
dm_percent_t *percent)
{
return 0;
}
int lv_thin_pool_transaction_id(const struct logical_volume *lv,
uint64_t *transaction_id)
{
return 0;
}
int lv_thin_device_id(const struct logical_volume *lv, uint32_t *device_id)
{
return 0;
}
int lvs_in_vg_activated(const struct volume_group *vg)
{
return 0;
}
int lvs_in_vg_opened(const struct volume_group *vg)
{
return 0;
}
/******
2004-03-08 21:54:13 +03:00
int lv_suspend(struct cmd_context *cmd, const char *lvid_s)
{
return 1;
}
*******/
int lv_suspend_if_active(struct cmd_context *cmd, const char *lvid_s, unsigned origin_only, unsigned exclusive,
const struct logical_volume *lv, const struct logical_volume *lv_pre)
{
return 1;
}
int lv_resume(struct cmd_context *cmd, const char *lvid_s, unsigned origin_only, const struct logical_volume *lv)
2004-03-08 21:54:13 +03:00
{
return 1;
}
int lv_resume_if_active(struct cmd_context *cmd, const char *lvid_s, unsigned origin_only,
unsigned exclusive, unsigned revert, const struct logical_volume *lv)
{
return 1;
}
int lv_deactivate(struct cmd_context *cmd, const char *lvid_s, const struct logical_volume *lv)
{
return 1;
}
2004-03-08 21:54:13 +03:00
int lv_activation_filter(struct cmd_context *cmd, const char *lvid_s,
int *activate_lv, const struct logical_volume *lv)
2004-03-08 21:54:13 +03:00
{
return 1;
}
int lv_activate(struct cmd_context *cmd, const char *lvid_s, int exclusive, int noscan,
int temporary, const struct logical_volume *lv)
{
return 1;
}
int lv_activate_with_filter(struct cmd_context *cmd, const char *lvid_s, int exclusive,
int noscan, int temporary, const struct logical_volume *lv)
2004-03-08 21:54:13 +03:00
{
return 1;
}
int lv_mknodes(struct cmd_context *cmd, const struct logical_volume *lv)
{
return 1;
}
int pv_uses_vg(struct physical_volume *pv,
struct volume_group *vg)
{
return 0;
}
void activation_release(void)
{
}
void activation_exit(void)
{
}
int lv_is_active(const struct logical_volume *lv)
{
return 0;
}
int lv_is_active_locally(const struct logical_volume *lv)
{
return 0;
}
2016-01-20 01:01:59 +03:00
int lv_is_active_remotely(const struct logical_volume *lv)
{
return 0;
}
int lv_is_active_but_not_locally(const struct logical_volume *lv)
{
return 0;
}
int lv_is_active_exclusive(const struct logical_volume *lv)
{
return 0;
}
int lv_is_active_exclusive_locally(const struct logical_volume *lv)
{
return 0;
}
int lv_is_active_exclusive_remotely(const struct logical_volume *lv)
{
return 0;
}
int lv_check_transient(struct logical_volume *lv)
{
return 1;
}
int monitor_dev_for_events(struct cmd_context *cmd, const struct logical_volume *lv,
const struct lv_activate_opts *laopts, int monitor)
{
return 1;
}
/* fs.c */
void fs_unlock(void)
{
}
/* dev_manager.c */
#include "targets.h"
int add_areas_line(struct dev_manager *dm, struct lv_segment *seg,
struct dm_tree_node *node, uint32_t start_area,
uint32_t areas)
{
return 0;
}
int device_is_usable(struct device *dev, struct dev_usable_check_params check)
{
return 0;
}
int lv_has_target_type(struct dm_pool *mem, const struct logical_volume *lv,
const char *layer, const char *target_type)
{
return 0;
}
#else /* DEVMAPPER_SUPPORT */
2002-11-18 17:01:16 +03:00
static int _activation = 1;
void set_activation(int act, int silent)
2002-11-18 17:01:16 +03:00
{
if (act == _activation)
2002-11-18 17:01:16 +03:00
return;
_activation = act;
2002-11-18 17:01:16 +03:00
if (_activation)
log_verbose("Activation enabled. Device-mapper kernel "
"driver will be used.");
else if (!silent)
log_warn("WARNING: Activation disabled. No device-mapper "
"interaction will be attempted.");
else
log_verbose("Activation disabled. No device-mapper "
"interaction will be attempted.");
2002-11-18 17:01:16 +03:00
}
int activation(void)
2002-11-18 17:01:16 +03:00
{
return _activation;
}
static int _passes_activation_filter(struct cmd_context *cmd,
const struct logical_volume *lv)
{
const struct dm_config_node *cn;
if (!(cn = find_config_tree_array(cmd, activation_volume_list_CFG, NULL))) {
log_verbose("activation/volume_list configuration setting "
"not defined: Checking only host tags for %s.",
display_lvname(lv));
/* If no host tags defined, activate */
if (dm_list_empty(&cmd->tags))
return 1;
/* If any host tag matches any LV or VG tag, activate */
if (str_list_match_list(&cmd->tags, &lv->tags, NULL) ||
str_list_match_list(&cmd->tags, &lv->vg->tags, NULL))
return 1;
log_verbose("No host tag matches %s", display_lvname(lv));
/* Don't activate */
return 0;
}
return _lv_passes_volumes_filter(cmd, lv, cn, activation_volume_list_CFG);
}
static int _passes_readonly_filter(struct cmd_context *cmd,
const struct logical_volume *lv)
{
const struct dm_config_node *cn;
if (!(cn = find_config_tree_array(cmd, activation_read_only_volume_list_CFG, NULL)))
return 0;
return _lv_passes_volumes_filter(cmd, lv, cn, activation_read_only_volume_list_CFG);
}
int library_version(char *version, size_t size)
{
2002-11-18 17:01:16 +03:00
if (!activation())
return 0;
return dm_get_library_version(version, size);
}
int driver_version(char *version, size_t size)
{
2002-11-18 17:01:16 +03:00
if (!activation())
return 0;
log_very_verbose("Getting driver version");
return dm_driver_version(version, size);
}
int target_version(const char *target_name, uint32_t *maj,
uint32_t *min, uint32_t *patchlevel)
2004-03-26 22:52:09 +03:00
{
int r = 0;
struct dm_task *dmt;
struct dm_versions *target, *last_target;
log_very_verbose("Getting target version for %s", target_name);
if (!(dmt = dm_task_create(DM_DEVICE_LIST_VERSIONS)))
return_0;
2004-03-26 22:52:09 +03:00
if (activation_checks() && !dm_task_enable_checks(dmt))
goto_out;
2004-03-26 22:52:09 +03:00
if (!dm_task_run(dmt)) {
log_debug_activation("Failed to get %s target version", target_name);
2004-03-26 22:52:09 +03:00
/* Assume this was because LIST_VERSIONS isn't supported */
*maj = 0;
*min = 0;
*patchlevel = 0;
r = 1;
goto out;
2004-03-26 22:52:09 +03:00
}
target = dm_task_get_versions(dmt);
do {
last_target = target;
if (!strcmp(target_name, target->name)) {
r = 1;
*maj = target->version[0];
*min = target->version[1];
*patchlevel = target->version[2];
2004-03-26 22:52:09 +03:00
goto out;
}
2010-12-20 16:37:26 +03:00
target = (struct dm_versions *)((char *) target + target->next);
2004-03-26 22:52:09 +03:00
} while (last_target != target);
out:
if (r)
log_very_verbose("Found %s target "
"v%" PRIu32 ".%" PRIu32 ".%" PRIu32 ".",
target_name, *maj, *min, *patchlevel);
2004-03-26 22:52:09 +03:00
dm_task_destroy(dmt);
return r;
}
2011-11-11 20:41:37 +04:00
int lvm_dm_prefix_check(int major, int minor, const char *prefix)
{
struct dm_task *dmt;
const char *uuid;
int r;
if (!(dmt = dm_task_create(DM_DEVICE_STATUS)))
2011-11-11 19:14:05 +04:00
return_0;
if (!dm_task_set_minor(dmt, minor) ||
!dm_task_set_major(dmt, major) ||
!dm_task_run(dmt) ||
!(uuid = dm_task_get_uuid(dmt))) {
dm_task_destroy(dmt);
return 0;
}
r = strncasecmp(uuid, prefix, strlen(prefix));
dm_task_destroy(dmt);
2011-11-11 19:14:05 +04:00
return r ? 0 : 1;
}
int module_present(struct cmd_context *cmd, const char *target_name)
{
int ret = 0;
2005-10-19 17:59:18 +04:00
#ifdef MODPROBE_CMD
char module[128];
const char *argv[] = { MODPROBE_CMD, module, NULL };
#endif
struct stat st;
char path[PATH_MAX];
int i = dm_snprintf(path, (sizeof(path) - 1), "%smodule/dm_%s",
dm_sysfs_dir(), target_name);
if (i > 0) {
while (path[--i] != '/') /* stop on dm_ */
if (path[i] == '-')
path[i] = '_'; /* replace '-' with '_' */
if ((lstat(path, &st) == 0) && S_ISDIR(st.st_mode)) {
2016-04-27 01:01:08 +03:00
log_debug_activation("Module directory %s exists.", path);
return 1;
}
}
#ifdef MODPROBE_CMD
if (dm_snprintf(module, sizeof(module), "dm-%s", target_name) < 0) {
log_error("module_present module name too long: %s",
target_name);
return 0;
}
ret = exec_cmd(cmd, argv, NULL, 0);
2005-10-19 17:59:18 +04:00
#endif
return ret;
}
int target_present_version(struct cmd_context *cmd, const char *target_name,
int use_modprobe,
uint32_t *maj, uint32_t *min, uint32_t *patchlevel)
{
if (!activation()) {
log_error(INTERNAL_ERROR "Target present version called when activation is disabled.");
return 0;
}
#ifdef MODPROBE_CMD
if (use_modprobe) {
if (target_version(target_name, maj, min, patchlevel))
return 1;
if (!module_present(cmd, target_name))
return_0;
}
#endif
return target_version(target_name, maj, min, patchlevel);
}
int target_present(struct cmd_context *cmd, const char *target_name,
int use_modprobe)
{
uint32_t maj, min, patchlevel;
return target_present_version(cmd, target_name, use_modprobe,
&maj, &min, &patchlevel);
}
static int _lv_info(struct cmd_context *cmd, const struct logical_volume *lv,
int use_layer, struct lvinfo *info,
const struct lv_segment *seg,
struct lv_seg_status *seg_status,
int with_open_count, int with_read_ahead)
2001-11-07 14:51:42 +03:00
{
struct dm_info dminfo;
/*
* If open_count info is requested and we have to be sure our own udev
* transactions are finished
* For non-clustered locking type we are only interested for non-delete operation
* in progress - as only those could lead to opened files
*/
if (with_open_count) {
if (locking_is_clustered() && !sync_local_dev_names(cmd)) /* Wait to have udev in sync */
return_0;
else if (fs_has_non_delete_ops())
fs_unlock(); /* For non clustered - wait if there are non-delete ops */
}
2002-11-18 17:01:16 +03:00
/* New thin-pool has no layer, but -tpool suffix needs to be queried */
if (!use_layer && lv_is_new_thin_pool(lv)) {
/* Check if there isn't existing old thin pool mapping in the table */
if (!dev_manager_info(cmd, lv, NULL, 0, 0, &dminfo, NULL, NULL))
return_0;
if (!dminfo.exists)
use_layer = 1;
}
if (seg_status) {
/* TODO: for now it's mess with seg_status */
seg_status->seg = seg;
if (lv_is_merging_cow(lv)) {
if (lv_has_target_type(cmd->mem, origin_from_cow(lv), NULL, TARGET_NAME_SNAPSHOT_MERGE)) {
/*
* When the snapshot-merge has not yet started, query COW LVs as is.
* When merge is in progress, query merging origin LV instead.
* COW volume is already mapped as error target in this case.
*/
lv = origin_from_cow(lv);
seg_status->seg = first_seg(lv);
log_debug_activation("Snapshot merge is in progress, querying status of %s instead.",
display_lvname(lv));
}
} else if (!use_layer && lv_is_origin(lv) && !lv_is_external_origin(lv)) {
/*
* Query status for 'layered' (-real) device most of the time,
* only when snapshot merge started, query its progress.
* TODO: single LV may need couple status to be exposed at once....
* but this needs more logical background
*/
if (!lv_is_merging_origin(lv) ||
!lv_has_target_type(cmd->mem, origin_from_cow(lv), NULL, TARGET_NAME_SNAPSHOT_MERGE))
use_layer = 1;
} else if (lv_is_cow(lv)) {
/* Hadle fictional lvm2 snapshot and query snapshotX volume */
seg_status->seg = find_snapshot(lv);
}
}
if (!dev_manager_info(cmd, lv,
(use_layer) ? lv_layer(lv) : NULL,
with_open_count, with_read_ahead,
&dminfo, (info) ? &info->read_ahead : NULL,
seg_status))
return_0;
if (!info)
return dminfo.exists;
info->exists = dminfo.exists;
info->suspended = dminfo.suspended;
info->open_count = dminfo.open_count;
info->major = dminfo.major;
info->minor = dminfo.minor;
info->read_only = dminfo.read_only;
info->live_table = dminfo.live_table;
info->inactive_table = dminfo.inactive_table;
return 1;
}
2002-02-11 20:42:02 +03:00
/*
* Returns 1 if info structure populated, else 0 on failure.
* When lvinfo* is NULL, it returns 1 if the device is locally active, 0 otherwise.
*/
int lv_info(struct cmd_context *cmd, const struct logical_volume *lv, int use_layer,
struct lvinfo *info, int with_open_count, int with_read_ahead)
{
if (!activation())
return 0;
return _lv_info(cmd, lv, use_layer, info, NULL, NULL, with_open_count, with_read_ahead);
}
int lv_info_by_lvid(struct cmd_context *cmd, const char *lvid_s, int use_layer,
struct lvinfo *info, int with_open_count, int with_read_ahead)
2004-03-26 18:35:01 +03:00
{
int r;
2004-03-26 18:35:01 +03:00
struct logical_volume *lv;
if (!(lv = lv_from_lvid(cmd, lvid_s, 0)))
2004-03-26 18:35:01 +03:00
return 0;
r = lv_info(cmd, lv, use_layer, info, with_open_count, with_read_ahead);
release_vg(lv->vg);
return r;
2004-03-26 18:35:01 +03:00
}
/*
* Returns 1 if lv_seg_status structure populated,
* else 0 on failure or if device not active locally.
*/
int lv_status(struct cmd_context *cmd, const struct lv_segment *lv_seg,
int use_layer, struct lv_seg_status *lv_seg_status)
{
if (!activation())
return 0;
return _lv_info(cmd, lv_seg->lv, use_layer, NULL, lv_seg, lv_seg_status, 0, 0);
}
/*
* Returns 1 if lv_with_info_and_seg_status structure populated,
* else 0 on failure or if device not active locally.
*
* This is the same as calling lv_info and lv_status,
* but* it's done in one go with one ioctl if possible! ]
*/
int lv_info_with_seg_status(struct cmd_context *cmd, const struct logical_volume *lv,
const struct lv_segment *lv_seg, int use_layer,
struct lv_with_info_and_seg_status *status,
int with_open_count, int with_read_ahead)
{
if (!activation())
return 0;
if (lv == lv_seg->lv)
return _lv_info(cmd, lv, use_layer, &status->info, lv_seg, &status->seg_status,
with_open_count, with_read_ahead);
/*
* If the info is requested for an LV and segment
* status for segment that belong to another LV,
* we need to acquire info and status separately!
*/
return _lv_info(cmd, lv, use_layer, &status->info, NULL, NULL, with_open_count, with_read_ahead) &&
2015-01-20 14:43:16 +03:00
_lv_info(cmd, lv_seg->lv, use_layer, NULL, lv_seg, &status->seg_status, 0, 0);
}
#define OPEN_COUNT_CHECK_RETRIES 25
#define OPEN_COUNT_CHECK_USLEEP_DELAY 200000
/* Only report error if error_if_used is set */
int lv_check_not_in_use(const struct logical_volume *lv, int error_if_used)
{
struct lvinfo info;
unsigned int open_count_check_retries;
if (!lv_info(lv->vg->cmd, lv, 0, &info, 1, 0) || !info.exists || !info.open_count)
return 1;
/* If sysfs is not used, use open_count information only. */
if (dm_sysfs_dir()) {
if (dm_device_has_holders(info.major, info.minor)) {
if (error_if_used)
log_error("Logical volume %s is used by another device.",
display_lvname(lv));
else
log_debug_activation("Logical volume %s is used by another device.",
display_lvname(lv));
return 0;
}
if (dm_device_has_mounted_fs(info.major, info.minor)) {
if (error_if_used)
log_error("Logical volume %s contains a filesystem in use.",
display_lvname(lv));
else
log_debug_activation("Logical volume %s contains a filesystem in use.",
display_lvname(lv));
return 0;
}
}
open_count_check_retries = retry_deactivation() ? OPEN_COUNT_CHECK_RETRIES : 1;
while (info.open_count > 0 && open_count_check_retries--) {
if (!open_count_check_retries) {
if (error_if_used)
log_error("Logical volume %s in use.", display_lvname(lv));
else
log_debug_activation("Logical volume %s in use.", display_lvname(lv));
return 0;
}
usleep(OPEN_COUNT_CHECK_USLEEP_DELAY);
log_debug_activation("Retrying open_count check for %s.",
display_lvname(lv));
if (!lv_info(lv->vg->cmd, lv, 0, &info, 1, 0)) {
stack; /* device dissappeared? */
break;
}
}
return 1;
}
/*
* Returns 1 if percent set, else 0 on failure.
*/
int lv_check_transient(struct logical_volume *lv)
{
int r;
struct dev_manager *dm;
if (!activation())
return 0;
log_debug_activation("Checking transient status for LV %s.",
display_lvname(lv));
if (!(dm = dev_manager_create(lv->vg->cmd, lv->vg->name, 1)))
return_0;
if (!(r = dev_manager_transient(dm, lv)))
stack;
dev_manager_destroy(dm);
return r;
}
/*
* Returns 1 if percent set, else 0 on failure.
*/
int lv_snapshot_percent(const struct logical_volume *lv, dm_percent_t *percent)
{
int r;
struct dev_manager *dm;
if (!lv_info(lv->vg->cmd, lv, 0, NULL, 0, 0))
2002-11-18 17:01:16 +03:00
return 0;
log_debug_activation("Checking snapshot percent for LV %s.",
display_lvname(lv));
if (!(dm = dev_manager_create(lv->vg->cmd, lv->vg->name, 1)))
return_0;
if (!(r = dev_manager_snapshot_percent(dm, lv, percent)))
stack;
2002-05-22 18:03:45 +04:00
dev_manager_destroy(dm);
return r;
}
2003-04-30 19:26:25 +04:00
/* FIXME Merge with snapshot_percent */
int lv_mirror_percent(struct cmd_context *cmd, const struct logical_volume *lv,
int wait, dm_percent_t *percent, uint32_t *event_nr)
2003-04-30 19:26:25 +04:00
{
int r;
struct dev_manager *dm;
/* If mirrored LV is temporarily shrinked to 1 area (= linear),
* it should be considered in-sync. */
if (dm_list_size(&lv->segments) == 1 && first_seg(lv)->area_count == 1) {
*percent = DM_PERCENT_100;
return 1;
}
if (!lv_info(cmd, lv, 0, NULL, 0, 0))
2003-04-30 19:26:25 +04:00
return 0;
log_debug_activation("Checking mirror percent for LV %s.",
display_lvname(lv));
if (!(dm = dev_manager_create(lv->vg->cmd, lv->vg->name, 1)))
return_0;
2003-04-30 19:26:25 +04:00
if (!(r = dev_manager_mirror_percent(dm, lv, wait, percent, event_nr)))
2003-04-30 19:26:25 +04:00
stack;
dev_manager_destroy(dm);
return r;
}
int lv_raid_percent(const struct logical_volume *lv, dm_percent_t *percent)
{
return lv_mirror_percent(lv->vg->cmd, lv, 0, percent, NULL);
}
int lv_raid_dev_health(const struct logical_volume *lv, char **dev_health)
{
int r;
struct dev_manager *dm;
struct dm_status_raid *status;
*dev_health = NULL;
if (!lv_info(lv->vg->cmd, lv, 0, NULL, 0, 0))
return 0;
log_debug_activation("Checking raid device health for LV %s.",
display_lvname(lv));
if (!(dm = dev_manager_create(lv->vg->cmd, lv->vg->name, 1)))
return_0;
if (!(r = dev_manager_raid_status(dm, lv, &status)) ||
!(*dev_health = dm_pool_strdup(lv->vg->cmd->mem,
status->dev_health))) {
dev_manager_destroy(dm);
return_0;
}
dev_manager_destroy(dm);
return r;
}
RAID: Add scrubbing support for RAID LVs New options to 'lvchange' allow users to scrub their RAID LVs. Synopsis: lvchange --syncaction {check|repair} vg/raid_lv RAID scrubbing is the process of reading all the data and parity blocks in an array and checking to see whether they are coherent. 'lvchange' can now initaite the two scrubbing operations: "check" and "repair". "check" will go over the array and recored the number of discrepancies but not repair them. "repair" will correct the discrepancies as it finds them. 'lvchange --syncaction repair vg/raid_lv' is not to be confused with 'lvconvert --repair vg/raid_lv'. The former initiates a background synchronization operation on the array, while the latter is designed to repair/replace failed devices in a mirror or RAID logical volume. Additional reporting has been added for 'lvs' to support the new operations. Two new printable fields (which are not printed by default) have been added: "syncaction" and "mismatches". These can be accessed using the '-o' option to 'lvs', like: lvs -o +syncaction,mismatches vg/lv "syncaction" will print the current synchronization operation that the RAID volume is performing. It can be one of the following: - idle: All sync operations complete (doing nothing) - resync: Initializing an array or recovering after a machine failure - recover: Replacing a device in the array - check: Looking for array inconsistencies - repair: Looking for and repairing inconsistencies The "mismatches" field with print the number of descrepancies found during a check or repair operation. The 'Cpy%Sync' field already available to 'lvs' will print the progress of any of the above syncactions, including check and repair. Finally, the lv_attr field has changed to accomadate the scrubbing operations as well. The role of the 'p'artial character in the lv_attr report field as expanded. "Partial" is really an indicator for the health of a logical volume and it makes sense to extend this include other health indicators as well, specifically: 'm'ismatches: Indicates that there are discrepancies in a RAID LV. This character is shown after a scrubbing operation has detected that portions of the RAID are not coherent. 'r'efresh : Indicates that a device in a RAID array has suffered a failure and the kernel regards it as failed - even though LVM can read the device label and considers the device to be ok. The LV should be 'r'efreshed to notify the kernel that the device is now available, or the device should be 'r'eplaced if it is suspected of failing.
2013-04-12 00:33:59 +04:00
int lv_raid_mismatch_count(const struct logical_volume *lv, uint64_t *cnt)
{
struct dev_manager *dm;
struct dm_status_raid *status;
*cnt = 0;
if (!lv_info(lv->vg->cmd, lv, 0, NULL, 0, 0))
RAID: Add scrubbing support for RAID LVs New options to 'lvchange' allow users to scrub their RAID LVs. Synopsis: lvchange --syncaction {check|repair} vg/raid_lv RAID scrubbing is the process of reading all the data and parity blocks in an array and checking to see whether they are coherent. 'lvchange' can now initaite the two scrubbing operations: "check" and "repair". "check" will go over the array and recored the number of discrepancies but not repair them. "repair" will correct the discrepancies as it finds them. 'lvchange --syncaction repair vg/raid_lv' is not to be confused with 'lvconvert --repair vg/raid_lv'. The former initiates a background synchronization operation on the array, while the latter is designed to repair/replace failed devices in a mirror or RAID logical volume. Additional reporting has been added for 'lvs' to support the new operations. Two new printable fields (which are not printed by default) have been added: "syncaction" and "mismatches". These can be accessed using the '-o' option to 'lvs', like: lvs -o +syncaction,mismatches vg/lv "syncaction" will print the current synchronization operation that the RAID volume is performing. It can be one of the following: - idle: All sync operations complete (doing nothing) - resync: Initializing an array or recovering after a machine failure - recover: Replacing a device in the array - check: Looking for array inconsistencies - repair: Looking for and repairing inconsistencies The "mismatches" field with print the number of descrepancies found during a check or repair operation. The 'Cpy%Sync' field already available to 'lvs' will print the progress of any of the above syncactions, including check and repair. Finally, the lv_attr field has changed to accomadate the scrubbing operations as well. The role of the 'p'artial character in the lv_attr report field as expanded. "Partial" is really an indicator for the health of a logical volume and it makes sense to extend this include other health indicators as well, specifically: 'm'ismatches: Indicates that there are discrepancies in a RAID LV. This character is shown after a scrubbing operation has detected that portions of the RAID are not coherent. 'r'efresh : Indicates that a device in a RAID array has suffered a failure and the kernel regards it as failed - even though LVM can read the device label and considers the device to be ok. The LV should be 'r'efreshed to notify the kernel that the device is now available, or the device should be 'r'eplaced if it is suspected of failing.
2013-04-12 00:33:59 +04:00
return 0;
log_debug_activation("Checking raid mismatch count for LV %s.",
display_lvname(lv));
RAID: Add scrubbing support for RAID LVs New options to 'lvchange' allow users to scrub their RAID LVs. Synopsis: lvchange --syncaction {check|repair} vg/raid_lv RAID scrubbing is the process of reading all the data and parity blocks in an array and checking to see whether they are coherent. 'lvchange' can now initaite the two scrubbing operations: "check" and "repair". "check" will go over the array and recored the number of discrepancies but not repair them. "repair" will correct the discrepancies as it finds them. 'lvchange --syncaction repair vg/raid_lv' is not to be confused with 'lvconvert --repair vg/raid_lv'. The former initiates a background synchronization operation on the array, while the latter is designed to repair/replace failed devices in a mirror or RAID logical volume. Additional reporting has been added for 'lvs' to support the new operations. Two new printable fields (which are not printed by default) have been added: "syncaction" and "mismatches". These can be accessed using the '-o' option to 'lvs', like: lvs -o +syncaction,mismatches vg/lv "syncaction" will print the current synchronization operation that the RAID volume is performing. It can be one of the following: - idle: All sync operations complete (doing nothing) - resync: Initializing an array or recovering after a machine failure - recover: Replacing a device in the array - check: Looking for array inconsistencies - repair: Looking for and repairing inconsistencies The "mismatches" field with print the number of descrepancies found during a check or repair operation. The 'Cpy%Sync' field already available to 'lvs' will print the progress of any of the above syncactions, including check and repair. Finally, the lv_attr field has changed to accomadate the scrubbing operations as well. The role of the 'p'artial character in the lv_attr report field as expanded. "Partial" is really an indicator for the health of a logical volume and it makes sense to extend this include other health indicators as well, specifically: 'm'ismatches: Indicates that there are discrepancies in a RAID LV. This character is shown after a scrubbing operation has detected that portions of the RAID are not coherent. 'r'efresh : Indicates that a device in a RAID array has suffered a failure and the kernel regards it as failed - even though LVM can read the device label and considers the device to be ok. The LV should be 'r'efreshed to notify the kernel that the device is now available, or the device should be 'r'eplaced if it is suspected of failing.
2013-04-12 00:33:59 +04:00
if (!(dm = dev_manager_create(lv->vg->cmd, lv->vg->name, 1)))
return_0;
if (!dev_manager_raid_status(dm, lv, &status)) {
dev_manager_destroy(dm);
return_0;
}
*cnt = status->mismatch_count;
dev_manager_destroy(dm);
return 1;
}
int lv_raid_sync_action(const struct logical_volume *lv, char **sync_action)
{
struct dev_manager *dm;
struct dm_status_raid *status;
char *action;
*sync_action = NULL;
if (!lv_info(lv->vg->cmd, lv, 0, NULL, 0, 0))
RAID: Add scrubbing support for RAID LVs New options to 'lvchange' allow users to scrub their RAID LVs. Synopsis: lvchange --syncaction {check|repair} vg/raid_lv RAID scrubbing is the process of reading all the data and parity blocks in an array and checking to see whether they are coherent. 'lvchange' can now initaite the two scrubbing operations: "check" and "repair". "check" will go over the array and recored the number of discrepancies but not repair them. "repair" will correct the discrepancies as it finds them. 'lvchange --syncaction repair vg/raid_lv' is not to be confused with 'lvconvert --repair vg/raid_lv'. The former initiates a background synchronization operation on the array, while the latter is designed to repair/replace failed devices in a mirror or RAID logical volume. Additional reporting has been added for 'lvs' to support the new operations. Two new printable fields (which are not printed by default) have been added: "syncaction" and "mismatches". These can be accessed using the '-o' option to 'lvs', like: lvs -o +syncaction,mismatches vg/lv "syncaction" will print the current synchronization operation that the RAID volume is performing. It can be one of the following: - idle: All sync operations complete (doing nothing) - resync: Initializing an array or recovering after a machine failure - recover: Replacing a device in the array - check: Looking for array inconsistencies - repair: Looking for and repairing inconsistencies The "mismatches" field with print the number of descrepancies found during a check or repair operation. The 'Cpy%Sync' field already available to 'lvs' will print the progress of any of the above syncactions, including check and repair. Finally, the lv_attr field has changed to accomadate the scrubbing operations as well. The role of the 'p'artial character in the lv_attr report field as expanded. "Partial" is really an indicator for the health of a logical volume and it makes sense to extend this include other health indicators as well, specifically: 'm'ismatches: Indicates that there are discrepancies in a RAID LV. This character is shown after a scrubbing operation has detected that portions of the RAID are not coherent. 'r'efresh : Indicates that a device in a RAID array has suffered a failure and the kernel regards it as failed - even though LVM can read the device label and considers the device to be ok. The LV should be 'r'efreshed to notify the kernel that the device is now available, or the device should be 'r'eplaced if it is suspected of failing.
2013-04-12 00:33:59 +04:00
return 0;
log_debug_activation("Checking raid sync_action for LV %s.",
display_lvname(lv));
RAID: Add scrubbing support for RAID LVs New options to 'lvchange' allow users to scrub their RAID LVs. Synopsis: lvchange --syncaction {check|repair} vg/raid_lv RAID scrubbing is the process of reading all the data and parity blocks in an array and checking to see whether they are coherent. 'lvchange' can now initaite the two scrubbing operations: "check" and "repair". "check" will go over the array and recored the number of discrepancies but not repair them. "repair" will correct the discrepancies as it finds them. 'lvchange --syncaction repair vg/raid_lv' is not to be confused with 'lvconvert --repair vg/raid_lv'. The former initiates a background synchronization operation on the array, while the latter is designed to repair/replace failed devices in a mirror or RAID logical volume. Additional reporting has been added for 'lvs' to support the new operations. Two new printable fields (which are not printed by default) have been added: "syncaction" and "mismatches". These can be accessed using the '-o' option to 'lvs', like: lvs -o +syncaction,mismatches vg/lv "syncaction" will print the current synchronization operation that the RAID volume is performing. It can be one of the following: - idle: All sync operations complete (doing nothing) - resync: Initializing an array or recovering after a machine failure - recover: Replacing a device in the array - check: Looking for array inconsistencies - repair: Looking for and repairing inconsistencies The "mismatches" field with print the number of descrepancies found during a check or repair operation. The 'Cpy%Sync' field already available to 'lvs' will print the progress of any of the above syncactions, including check and repair. Finally, the lv_attr field has changed to accomadate the scrubbing operations as well. The role of the 'p'artial character in the lv_attr report field as expanded. "Partial" is really an indicator for the health of a logical volume and it makes sense to extend this include other health indicators as well, specifically: 'm'ismatches: Indicates that there are discrepancies in a RAID LV. This character is shown after a scrubbing operation has detected that portions of the RAID are not coherent. 'r'efresh : Indicates that a device in a RAID array has suffered a failure and the kernel regards it as failed - even though LVM can read the device label and considers the device to be ok. The LV should be 'r'efreshed to notify the kernel that the device is now available, or the device should be 'r'eplaced if it is suspected of failing.
2013-04-12 00:33:59 +04:00
if (!(dm = dev_manager_create(lv->vg->cmd, lv->vg->name, 1)))
return_0;
/* status->sync_action can be NULL if dm-raid version < 1.5.0 */
RAID: Add scrubbing support for RAID LVs New options to 'lvchange' allow users to scrub their RAID LVs. Synopsis: lvchange --syncaction {check|repair} vg/raid_lv RAID scrubbing is the process of reading all the data and parity blocks in an array and checking to see whether they are coherent. 'lvchange' can now initaite the two scrubbing operations: "check" and "repair". "check" will go over the array and recored the number of discrepancies but not repair them. "repair" will correct the discrepancies as it finds them. 'lvchange --syncaction repair vg/raid_lv' is not to be confused with 'lvconvert --repair vg/raid_lv'. The former initiates a background synchronization operation on the array, while the latter is designed to repair/replace failed devices in a mirror or RAID logical volume. Additional reporting has been added for 'lvs' to support the new operations. Two new printable fields (which are not printed by default) have been added: "syncaction" and "mismatches". These can be accessed using the '-o' option to 'lvs', like: lvs -o +syncaction,mismatches vg/lv "syncaction" will print the current synchronization operation that the RAID volume is performing. It can be one of the following: - idle: All sync operations complete (doing nothing) - resync: Initializing an array or recovering after a machine failure - recover: Replacing a device in the array - check: Looking for array inconsistencies - repair: Looking for and repairing inconsistencies The "mismatches" field with print the number of descrepancies found during a check or repair operation. The 'Cpy%Sync' field already available to 'lvs' will print the progress of any of the above syncactions, including check and repair. Finally, the lv_attr field has changed to accomadate the scrubbing operations as well. The role of the 'p'artial character in the lv_attr report field as expanded. "Partial" is really an indicator for the health of a logical volume and it makes sense to extend this include other health indicators as well, specifically: 'm'ismatches: Indicates that there are discrepancies in a RAID LV. This character is shown after a scrubbing operation has detected that portions of the RAID are not coherent. 'r'efresh : Indicates that a device in a RAID array has suffered a failure and the kernel regards it as failed - even though LVM can read the device label and considers the device to be ok. The LV should be 'r'efreshed to notify the kernel that the device is now available, or the device should be 'r'eplaced if it is suspected of failing.
2013-04-12 00:33:59 +04:00
if (!dev_manager_raid_status(dm, lv, &status) ||
!status->sync_action ||
RAID: Add scrubbing support for RAID LVs New options to 'lvchange' allow users to scrub their RAID LVs. Synopsis: lvchange --syncaction {check|repair} vg/raid_lv RAID scrubbing is the process of reading all the data and parity blocks in an array and checking to see whether they are coherent. 'lvchange' can now initaite the two scrubbing operations: "check" and "repair". "check" will go over the array and recored the number of discrepancies but not repair them. "repair" will correct the discrepancies as it finds them. 'lvchange --syncaction repair vg/raid_lv' is not to be confused with 'lvconvert --repair vg/raid_lv'. The former initiates a background synchronization operation on the array, while the latter is designed to repair/replace failed devices in a mirror or RAID logical volume. Additional reporting has been added for 'lvs' to support the new operations. Two new printable fields (which are not printed by default) have been added: "syncaction" and "mismatches". These can be accessed using the '-o' option to 'lvs', like: lvs -o +syncaction,mismatches vg/lv "syncaction" will print the current synchronization operation that the RAID volume is performing. It can be one of the following: - idle: All sync operations complete (doing nothing) - resync: Initializing an array or recovering after a machine failure - recover: Replacing a device in the array - check: Looking for array inconsistencies - repair: Looking for and repairing inconsistencies The "mismatches" field with print the number of descrepancies found during a check or repair operation. The 'Cpy%Sync' field already available to 'lvs' will print the progress of any of the above syncactions, including check and repair. Finally, the lv_attr field has changed to accomadate the scrubbing operations as well. The role of the 'p'artial character in the lv_attr report field as expanded. "Partial" is really an indicator for the health of a logical volume and it makes sense to extend this include other health indicators as well, specifically: 'm'ismatches: Indicates that there are discrepancies in a RAID LV. This character is shown after a scrubbing operation has detected that portions of the RAID are not coherent. 'r'efresh : Indicates that a device in a RAID array has suffered a failure and the kernel regards it as failed - even though LVM can read the device label and considers the device to be ok. The LV should be 'r'efreshed to notify the kernel that the device is now available, or the device should be 'r'eplaced if it is suspected of failing.
2013-04-12 00:33:59 +04:00
!(action = dm_pool_strdup(lv->vg->cmd->mem,
status->sync_action))) {
dev_manager_destroy(dm);
return_0;
}
*sync_action = action;
dev_manager_destroy(dm);
return 1;
}
int lv_raid_message(const struct logical_volume *lv, const char *msg)
{
int r = 0;
struct dev_manager *dm;
struct dm_status_raid *status;
if (!seg_is_raid(first_seg(lv))) {
/*
* Make it easier for user to know what to do when
* they are using thinpool.
*/
if (lv_is_thin_pool(lv) &&
(lv_is_raid(seg_lv(first_seg(lv), 0)) ||
lv_is_raid(first_seg(lv)->metadata_lv))) {
log_error("Thin pool data or metadata volume "
"must be specified. (E.g. \"%s_tdata\")",
display_lvname(lv));
return 0;
}
log_error("%s must be a RAID logical volume to perform this action.",
display_lvname(lv));
return 0;
}
if (!lv_info(lv->vg->cmd, lv, 0, NULL, 0, 0)) {
RAID: Add scrubbing support for RAID LVs New options to 'lvchange' allow users to scrub their RAID LVs. Synopsis: lvchange --syncaction {check|repair} vg/raid_lv RAID scrubbing is the process of reading all the data and parity blocks in an array and checking to see whether they are coherent. 'lvchange' can now initaite the two scrubbing operations: "check" and "repair". "check" will go over the array and recored the number of discrepancies but not repair them. "repair" will correct the discrepancies as it finds them. 'lvchange --syncaction repair vg/raid_lv' is not to be confused with 'lvconvert --repair vg/raid_lv'. The former initiates a background synchronization operation on the array, while the latter is designed to repair/replace failed devices in a mirror or RAID logical volume. Additional reporting has been added for 'lvs' to support the new operations. Two new printable fields (which are not printed by default) have been added: "syncaction" and "mismatches". These can be accessed using the '-o' option to 'lvs', like: lvs -o +syncaction,mismatches vg/lv "syncaction" will print the current synchronization operation that the RAID volume is performing. It can be one of the following: - idle: All sync operations complete (doing nothing) - resync: Initializing an array or recovering after a machine failure - recover: Replacing a device in the array - check: Looking for array inconsistencies - repair: Looking for and repairing inconsistencies The "mismatches" field with print the number of descrepancies found during a check or repair operation. The 'Cpy%Sync' field already available to 'lvs' will print the progress of any of the above syncactions, including check and repair. Finally, the lv_attr field has changed to accomadate the scrubbing operations as well. The role of the 'p'artial character in the lv_attr report field as expanded. "Partial" is really an indicator for the health of a logical volume and it makes sense to extend this include other health indicators as well, specifically: 'm'ismatches: Indicates that there are discrepancies in a RAID LV. This character is shown after a scrubbing operation has detected that portions of the RAID are not coherent. 'r'efresh : Indicates that a device in a RAID array has suffered a failure and the kernel regards it as failed - even though LVM can read the device label and considers the device to be ok. The LV should be 'r'efreshed to notify the kernel that the device is now available, or the device should be 'r'eplaced if it is suspected of failing.
2013-04-12 00:33:59 +04:00
log_error("Unable to send message to an inactive logical volume.");
return 0;
}
if (!(dm = dev_manager_create(lv->vg->cmd, lv->vg->name, 1)))
return_0;
if (!(r = dev_manager_raid_status(dm, lv, &status))) {
log_error("Failed to retrieve status of %s.",
display_lvname(lv));
RAID: Add scrubbing support for RAID LVs New options to 'lvchange' allow users to scrub their RAID LVs. Synopsis: lvchange --syncaction {check|repair} vg/raid_lv RAID scrubbing is the process of reading all the data and parity blocks in an array and checking to see whether they are coherent. 'lvchange' can now initaite the two scrubbing operations: "check" and "repair". "check" will go over the array and recored the number of discrepancies but not repair them. "repair" will correct the discrepancies as it finds them. 'lvchange --syncaction repair vg/raid_lv' is not to be confused with 'lvconvert --repair vg/raid_lv'. The former initiates a background synchronization operation on the array, while the latter is designed to repair/replace failed devices in a mirror or RAID logical volume. Additional reporting has been added for 'lvs' to support the new operations. Two new printable fields (which are not printed by default) have been added: "syncaction" and "mismatches". These can be accessed using the '-o' option to 'lvs', like: lvs -o +syncaction,mismatches vg/lv "syncaction" will print the current synchronization operation that the RAID volume is performing. It can be one of the following: - idle: All sync operations complete (doing nothing) - resync: Initializing an array or recovering after a machine failure - recover: Replacing a device in the array - check: Looking for array inconsistencies - repair: Looking for and repairing inconsistencies The "mismatches" field with print the number of descrepancies found during a check or repair operation. The 'Cpy%Sync' field already available to 'lvs' will print the progress of any of the above syncactions, including check and repair. Finally, the lv_attr field has changed to accomadate the scrubbing operations as well. The role of the 'p'artial character in the lv_attr report field as expanded. "Partial" is really an indicator for the health of a logical volume and it makes sense to extend this include other health indicators as well, specifically: 'm'ismatches: Indicates that there are discrepancies in a RAID LV. This character is shown after a scrubbing operation has detected that portions of the RAID are not coherent. 'r'efresh : Indicates that a device in a RAID array has suffered a failure and the kernel regards it as failed - even though LVM can read the device label and considers the device to be ok. The LV should be 'r'efreshed to notify the kernel that the device is now available, or the device should be 'r'eplaced if it is suspected of failing.
2013-04-12 00:33:59 +04:00
goto out;
}
if (!status->sync_action) {
log_error("Kernel driver does not support this action: %s", msg);
goto out;
}
/*
* Note that 'dev_manager_raid_message' allows us to pass down any
* currently valid message. However, this function restricts the
* number of user available combinations to a minimum. Specifically,
* "idle" -> "check"
* "idle" -> "repair"
* (The state automatically switches to "idle" when a sync process is
* complete.)
*/
if (strcmp(msg, "check") && strcmp(msg, "repair")) {
/*
* MD allows "frozen" to operate in a toggling fashion.
* We could allow this if we like...
*/
log_error("\"%s\" is not a supported sync operation.", msg);
goto out;
}
if (strcmp(status->sync_action, "idle")) {
log_error("%s state is currently \"%s\". Unable to switch to \"%s\".",
display_lvname(lv), status->sync_action, msg);
RAID: Add scrubbing support for RAID LVs New options to 'lvchange' allow users to scrub their RAID LVs. Synopsis: lvchange --syncaction {check|repair} vg/raid_lv RAID scrubbing is the process of reading all the data and parity blocks in an array and checking to see whether they are coherent. 'lvchange' can now initaite the two scrubbing operations: "check" and "repair". "check" will go over the array and recored the number of discrepancies but not repair them. "repair" will correct the discrepancies as it finds them. 'lvchange --syncaction repair vg/raid_lv' is not to be confused with 'lvconvert --repair vg/raid_lv'. The former initiates a background synchronization operation on the array, while the latter is designed to repair/replace failed devices in a mirror or RAID logical volume. Additional reporting has been added for 'lvs' to support the new operations. Two new printable fields (which are not printed by default) have been added: "syncaction" and "mismatches". These can be accessed using the '-o' option to 'lvs', like: lvs -o +syncaction,mismatches vg/lv "syncaction" will print the current synchronization operation that the RAID volume is performing. It can be one of the following: - idle: All sync operations complete (doing nothing) - resync: Initializing an array or recovering after a machine failure - recover: Replacing a device in the array - check: Looking for array inconsistencies - repair: Looking for and repairing inconsistencies The "mismatches" field with print the number of descrepancies found during a check or repair operation. The 'Cpy%Sync' field already available to 'lvs' will print the progress of any of the above syncactions, including check and repair. Finally, the lv_attr field has changed to accomadate the scrubbing operations as well. The role of the 'p'artial character in the lv_attr report field as expanded. "Partial" is really an indicator for the health of a logical volume and it makes sense to extend this include other health indicators as well, specifically: 'm'ismatches: Indicates that there are discrepancies in a RAID LV. This character is shown after a scrubbing operation has detected that portions of the RAID are not coherent. 'r'efresh : Indicates that a device in a RAID array has suffered a failure and the kernel regards it as failed - even though LVM can read the device label and considers the device to be ok. The LV should be 'r'efreshed to notify the kernel that the device is now available, or the device should be 'r'eplaced if it is suspected of failing.
2013-04-12 00:33:59 +04:00
goto out;
}
r = dev_manager_raid_message(dm, lv, msg);
out:
dev_manager_destroy(dm);
return r;
}
/*
* Return dm_status_cache for cache volume, accept also cache pool
*
* As there are too many variable for cache volumes, and it hard
* to make good API - so let's obtain dm_status_cache and return
* all info we have - user just has to release struct after its use.
*/
int lv_cache_status(const struct logical_volume *cache_lv,
struct lv_status_cache **status)
{
struct dev_manager *dm;
struct lv_segment *cache_seg;
if (lv_is_cache_pool(cache_lv)) {
if (dm_list_empty(&cache_lv->segs_using_this_lv) ||
!(cache_seg = get_only_segment_using_this_lv(cache_lv))) {
log_error(INTERNAL_ERROR "Cannot check status for unused cache pool %s.",
display_lvname(cache_lv));
return 0;
}
cache_lv = cache_seg->lv;
}
if (lv_is_pending_delete(cache_lv)) {
log_error("Cannot check status for deleted cache volume %s.",
display_lvname(cache_lv));
return 0;
}
if (!lv_info(cache_lv->vg->cmd, cache_lv, 0, NULL, 0, 0)) {
log_error("Cannot check status for locally inactive cache volume %s.",
display_lvname(cache_lv));
return 0;
}
log_debug_activation("Checking status for cache volume %s.",
display_lvname(cache_lv));
if (!(dm = dev_manager_create(cache_lv->vg->cmd, cache_lv->vg->name, 1)))
return_0;
if (!dev_manager_cache_status(dm, cache_lv, status)) {
dev_manager_destroy(dm);
return_0;
}
/* User has to call dm_pool_destroy(status->mem)! */
return 1;
}
2011-12-21 17:10:05 +04:00
/*
* Returns data or metadata percent usage, depends on metadata 0/1.
2011-12-21 17:10:05 +04:00
* Returns 1 if percent set, else 0 on failure.
*/
int lv_thin_pool_percent(const struct logical_volume *lv, int metadata,
dm_percent_t *percent)
2011-12-21 17:10:05 +04:00
{
int r;
struct dev_manager *dm;
if (!lv_info(lv->vg->cmd, lv, 1, NULL, 0, 0))
2011-12-21 17:10:05 +04:00
return 0;
log_debug_activation("Checking thin %sdata percent for LV %s.",
(metadata) ? "meta" : "", display_lvname(lv));
2011-12-21 17:10:05 +04:00
if (!(dm = dev_manager_create(lv->vg->cmd, lv->vg->name, 1)))
return_0;
if (!(r = dev_manager_thin_pool_percent(dm, lv, metadata, percent)))
2011-12-21 17:10:05 +04:00
stack;
dev_manager_destroy(dm);
return r;
}
/*
* Returns 1 if percent set, else 0 on failure.
*/
int lv_thin_percent(const struct logical_volume *lv,
int mapped, dm_percent_t *percent)
{
int r;
struct dev_manager *dm;
if (!lv_info(lv->vg->cmd, lv, 0, NULL, 0, 0))
return 0;
log_debug_activation("Checking thin percent for LV %s.",
display_lvname(lv));
if (!(dm = dev_manager_create(lv->vg->cmd, lv->vg->name, 1)))
return_0;
if (!(r = dev_manager_thin_percent(dm, lv, mapped, percent)))
stack;
dev_manager_destroy(dm);
return r;
}
/*
* Returns 1 if transaction_id set, else 0 on failure.
*/
int lv_thin_pool_transaction_id(const struct logical_volume *lv,
uint64_t *transaction_id)
{
int r;
struct dev_manager *dm;
struct dm_status_thin_pool *status;
if (!lv_info(lv->vg->cmd, lv, 1, NULL, 0, 0))
return 0;
log_debug_activation("Checking thin-pool transaction id for LV %s.",
display_lvname(lv));
if (!(dm = dev_manager_create(lv->vg->cmd, lv->vg->name, 1)))
return_0;
if (!(r = dev_manager_thin_pool_status(dm, lv, &status, 0)))
stack;
else
*transaction_id = status->transaction_id;
dev_manager_destroy(dm);
return r;
}
int lv_thin_device_id(const struct logical_volume *lv, uint32_t *device_id)
{
int r;
struct dev_manager *dm;
if (!lv_info(lv->vg->cmd, lv, 0, NULL, 0, 0))
return 0;
log_debug_activation("Checking device id for LV %s.",
display_lvname(lv));
if (!(dm = dev_manager_create(lv->vg->cmd, lv->vg->name, 1)))
return_0;
if (!(r = dev_manager_thin_device_id(dm, lv, device_id)))
stack;
dev_manager_destroy(dm);
return r;
}
static int _lv_active(struct cmd_context *cmd, const struct logical_volume *lv)
2001-11-07 18:02:07 +03:00
{
struct lvinfo info;
2001-11-07 18:02:07 +03:00
if (!lv_info(cmd, lv, 0, &info, 0, 0)) {
log_debug("Cannot determine activation status of %s%s.",
display_lvname(lv),
activation() ? "" : " (no device driver)");
return 0;
2001-11-07 18:02:07 +03:00
}
return info.exists;
2001-11-07 18:02:07 +03:00
}
static int _lv_open_count(struct cmd_context *cmd, const struct logical_volume *lv)
{
struct lvinfo info;
if (!lv_info(cmd, lv, 0, &info, 1, 0)) {
stack;
return -1;
}
return info.open_count;
}
static int _lv_activate_lv(const struct logical_volume *lv, struct lv_activate_opts *laopts)
{
int r;
struct dev_manager *dm;
if (!(dm = dev_manager_create(lv->vg->cmd, lv->vg->name, !lv_is_pvmove(lv))))
return_0;
2001-11-02 16:45:05 +03:00
if (!(r = dev_manager_activate(dm, lv, laopts)))
stack;
2001-11-02 16:45:05 +03:00
dev_manager_destroy(dm);
2001-11-02 16:45:05 +03:00
return r;
}
2001-10-16 20:25:28 +04:00
static int _lv_preload(const struct logical_volume *lv, struct lv_activate_opts *laopts,
int *flush_required)
{
int r = 0;
struct dev_manager *dm;
int old_readonly = laopts->read_only;
if (!(dm = dev_manager_create(lv->vg->cmd, lv->vg->name, !lv_is_pvmove(lv))))
goto_out;
laopts->read_only = _passes_readonly_filter(lv->vg->cmd, lv);
if (!(r = dev_manager_preload(dm, lv, laopts, flush_required)))
stack;
dev_manager_destroy(dm);
laopts->read_only = old_readonly;
out:
return r;
}
static int _lv_deactivate(const struct logical_volume *lv)
{
int r;
struct dev_manager *dm;
if (!(dm = dev_manager_create(lv->vg->cmd, lv->vg->name, 1)))
return_0;
2001-11-07 14:51:42 +03:00
if (!(r = dev_manager_deactivate(dm, lv)))
2001-11-07 14:51:42 +03:00
stack;
dev_manager_destroy(dm);
return r;
2001-11-07 14:51:42 +03:00
}
static int _lv_suspend_lv(const struct logical_volume *lv, struct lv_activate_opts *laopts,
int lockfs, int flush_required)
{
2002-03-14 18:36:07 +03:00
int r;
struct dev_manager *dm;
laopts->read_only = _passes_readonly_filter(lv->vg->cmd, lv);
/*
* When we are asked to manipulate (normally suspend/resume) the PVMOVE
* device directly, we don't want to touch the devices that use it.
*/
if (!(dm = dev_manager_create(lv->vg->cmd, lv->vg->name, !lv_is_pvmove(lv))))
return_0;
if (!(r = dev_manager_suspend(dm, lv, laopts, lockfs, flush_required)))
2001-11-07 14:51:42 +03:00
stack;
2002-03-14 18:36:07 +03:00
dev_manager_destroy(dm);
return r;
}
2002-03-01 22:08:11 +03:00
/*
2004-05-11 22:18:14 +04:00
* These two functions return the number of visible LVs in the state,
* or -1 on error. FIXME Check this.
2002-03-01 22:08:11 +03:00
*/
int lvs_in_vg_activated(const struct volume_group *vg)
2001-11-02 19:28:04 +03:00
{
2005-06-01 20:51:55 +04:00
struct lv_list *lvl;
2001-11-08 19:15:58 +03:00
int count = 0;
2001-11-07 14:51:42 +03:00
2002-11-18 17:01:16 +03:00
if (!activation())
return 0;
dm_list_iterate_items(lvl, &vg->lvs)
if (lv_is_visible(lvl->lv))
count += (_lv_active(vg->cmd, lvl->lv) == 1);
log_debug_activation("Counted %d active LVs in VG %s", count, vg->name);
2001-11-07 14:51:42 +03:00
return count;
2001-11-02 19:28:04 +03:00
}
2001-11-07 18:02:07 +03:00
int lvs_in_vg_opened(const struct volume_group *vg)
2001-11-07 18:02:07 +03:00
{
const struct lv_list *lvl;
2001-11-08 19:15:58 +03:00
int count = 0;
2001-11-07 18:02:07 +03:00
2002-11-18 17:01:16 +03:00
if (!activation())
return 0;
dm_list_iterate_items(lvl, &vg->lvs)
if (lv_is_visible(lvl->lv))
count += (_lv_open_count(vg->cmd, lvl->lv) > 0);
log_debug_activation("Counted %d open LVs in VG %s.", count, vg->name);
2001-11-07 18:02:07 +03:00
return count;
}
/*
* _lv_is_active
* @lv: logical volume being queried
* @locally: set if active locally (when provided)
2016-01-20 01:01:59 +03:00
* @remotely: set if active remotely (when provided)
* @exclusive: set if active exclusively (when provided)
*
* Determine whether an LV is active locally or in a cluster.
* In addition to the return code which indicates whether or
* not the LV is active somewhere, two other values are set
* to yield more information about the status of the activation:
2016-01-20 01:01:59 +03:00
*
* return locally exclusively status
* ====== ======= =========== ======
* 0 0 0 not active
* 1 0 0 active remotely
* 1 0 1 exclusive remotely
* 1 1 0 active locally and possibly remotely
* 1 1 1 exclusive locally (or local && !cluster)
* The VG lock must be held to call this function.
*
* Returns: 0 or 1
*/
static int _lv_is_active(const struct logical_volume *lv,
2016-01-20 01:01:59 +03:00
int *locally, int *remotely, int *exclusive)
{
int r, l, e; /* remote, local, and exclusive */
2016-01-20 01:01:59 +03:00
int skip_cluster_query = 0;
r = l = e = 0;
if (_lv_active(lv->vg->cmd, lv))
l = 1;
if (!vg_is_clustered(lv->vg)) {
if (l)
e = 1; /* exclusive by definition */
goto out;
}
2016-01-20 01:01:59 +03:00
/* Active locally, and the caller doesn't care about exclusive or remotely */
if (l && !exclusive && !remotely)
skip_cluster_query = 1;
if (skip_cluster_query)
goto out;
if ((r = cluster_lock_held(lv->lvid.s, "", &e)) >= 0)
goto out;
/*
* If lock query is not supported (due to interfacing with old
* code), then we cannot evaluate exclusivity properly.
*
* Old users of this function will never be affected by this,
* since they are only concerned about active vs. not active.
* New users of this function who specifically ask for 'exclusive'
* will be given a warning message.
*/
log_warn("WARNING: Unable to determine exclusivity of %s.", display_lvname(lv));
e = 0;
2016-01-20 01:01:59 +03:00
/* Also set remotely as a precaution, as we don't know */
r = 1;
/*
* We used to attempt activate_lv_excl_local(lv->vg->cmd, lv) here,
* but it's unreliable.
*/
out:
if (locally)
*locally = l;
if (exclusive)
*exclusive = e;
2016-01-20 01:01:59 +03:00
if (remotely)
*remotely = r;
2016-01-20 01:01:59 +03:00
log_very_verbose("%s is %sactive%s%s%s%s",
display_lvname(lv),
(r || l) ? "" : "not ",
(exclusive && e) ? " exclusive" : "",
2016-01-20 01:01:59 +03:00
l ? " locally" : "",
(!skip_cluster_query && l && r) ? " and" : "",
(!skip_cluster_query && r) ? " remotely" : "");
return r || l;
}
int lv_is_active(const struct logical_volume *lv)
{
2016-01-20 01:01:59 +03:00
return _lv_is_active(lv, NULL, NULL, NULL);
}
int lv_is_active_locally(const struct logical_volume *lv)
{
int l;
2016-01-20 01:01:59 +03:00
return _lv_is_active(lv, &l, NULL, NULL) && l;
}
int lv_is_active_remotely(const struct logical_volume *lv)
{
int r;
return _lv_is_active(lv, NULL, &r, NULL) && r;
}
int lv_is_active_but_not_locally(const struct logical_volume *lv)
{
int l;
2016-01-20 01:01:59 +03:00
return _lv_is_active(lv, &l, NULL, NULL) && !l;
}
int lv_is_active_exclusive(const struct logical_volume *lv)
{
int e;
2016-01-20 01:01:59 +03:00
return _lv_is_active(lv, NULL, NULL, &e) && e;
}
int lv_is_active_exclusive_locally(const struct logical_volume *lv)
{
int l, e;
2016-01-20 01:01:59 +03:00
return _lv_is_active(lv, &l, NULL, &e) && l && e;
}
int lv_is_active_exclusive_remotely(const struct logical_volume *lv)
{
int l, e;
2016-01-20 01:01:59 +03:00
return _lv_is_active(lv, &l, NULL, &e) && !l && e;
}
#ifdef DMEVENTD
static struct dm_event_handler *_create_dm_event_handler(struct cmd_context *cmd, const char *dmuuid, const char *dso,
const int timeout, enum dm_event_mask mask)
{
struct dm_event_handler *dmevh;
if (!(dmevh = dm_event_handler_create()))
return_NULL;
if (dm_event_handler_set_dmeventd_path(dmevh, find_config_tree_str(cmd, dmeventd_executable_CFG, NULL)))
goto_bad;
if (dso && dm_event_handler_set_dso(dmevh, dso))
goto_bad;
if (dm_event_handler_set_uuid(dmevh, dmuuid))
goto_bad;
dm_event_handler_set_timeout(dmevh, timeout);
dm_event_handler_set_event_mask(dmevh, mask);
return dmevh;
bad:
dm_event_handler_destroy(dmevh);
return NULL;
}
char *get_monitor_dso_path(struct cmd_context *cmd, const char *libpath)
{
char *path;
if (!(path = dm_pool_alloc(cmd->mem, PATH_MAX))) {
log_error("Failed to allocate dmeventd library path.");
return NULL;
}
get_shared_library_path(cmd, libpath, path, PATH_MAX);
return path;
}
static char *_build_target_uuid(struct cmd_context *cmd, const struct logical_volume *lv)
{
const char *layer;
if (lv_is_thin_pool(lv))
layer = "tpool"; /* Monitor "tpool" for the "thin pool". */
else if (lv_is_origin(lv))
layer = "real"; /* Monitor "real" for "snapshot-origin". */
else
layer = NULL;
return build_dm_uuid(cmd->mem, lv, layer);
}
static int _device_registered_with_dmeventd(struct cmd_context *cmd, const struct logical_volume *lv, int *pending, const char **dso)
{
char *uuid;
enum dm_event_mask evmask = 0;
struct dm_event_handler *dmevh;
*pending = 0;
if (!(uuid = _build_target_uuid(cmd, lv)))
return_0;
if (!(dmevh = _create_dm_event_handler(cmd, uuid, NULL, 0, DM_EVENT_ALL_ERRORS)))
return_0;
if (dm_event_get_registered_device(dmevh, 0)) {
dm_event_handler_destroy(dmevh);
return 0;
}
evmask = dm_event_handler_get_event_mask(dmevh);
if (evmask & DM_EVENT_REGISTRATION_PENDING) {
*pending = 1;
evmask &= ~DM_EVENT_REGISTRATION_PENDING;
}
if (dso && (*dso = dm_event_handler_get_dso(dmevh)) && !(*dso = dm_pool_strdup(cmd->mem, *dso)))
log_error("Failed to duplicate dso name.");
dm_event_handler_destroy(dmevh);
return evmask;
}
int target_registered_with_dmeventd(struct cmd_context *cmd, const char *dso,
const struct logical_volume *lv, int *pending)
{
char *uuid;
enum dm_event_mask evmask = 0;
struct dm_event_handler *dmevh;
*pending = 0;
if (!dso)
return_0;
if (!(uuid = _build_target_uuid(cmd, lv)))
return_0;
if (!(dmevh = _create_dm_event_handler(cmd, uuid, dso, 0, DM_EVENT_ALL_ERRORS)))
return_0;
if (dm_event_get_registered_device(dmevh, 0)) {
dm_event_handler_destroy(dmevh);
return 0;
}
evmask = dm_event_handler_get_event_mask(dmevh);
if (evmask & DM_EVENT_REGISTRATION_PENDING) {
*pending = 1;
evmask &= ~DM_EVENT_REGISTRATION_PENDING;
}
dm_event_handler_destroy(dmevh);
return evmask;
}
int target_register_events(struct cmd_context *cmd, const char *dso, const struct logical_volume *lv,
int evmask __attribute__((unused)), int set, int timeout)
{
char *uuid;
struct dm_event_handler *dmevh;
int r;
if (!dso)
return_0;
/* We always monitor the "real" device, never the "snapshot-origin" itself. */
if (!(uuid = _build_target_uuid(cmd, lv)))
return_0;
if (!(dmevh = _create_dm_event_handler(cmd, uuid, dso, timeout,
DM_EVENT_ALL_ERRORS | (timeout ? DM_EVENT_TIMEOUT : 0))))
return_0;
r = set ? dm_event_register_handler(dmevh) : dm_event_unregister_handler(dmevh);
dm_event_handler_destroy(dmevh);
if (!r)
return_0;
log_info("%s %s for events", set ? "Monitored" : "Unmonitored", uuid);
return 1;
}
#endif
/*
2007-01-20 01:21:45 +03:00
* Returns 0 if an attempt to (un)monitor the device failed.
* Returns 1 otherwise.
*/
int monitor_dev_for_events(struct cmd_context *cmd, const struct logical_volume *lv,
const struct lv_activate_opts *laopts, int monitor)
2005-12-02 23:35:07 +03:00
{
#ifdef DMEVENTD
2007-01-20 01:21:45 +03:00
int i, pending = 0, monitored;
int r = 1;
2013-04-25 13:46:17 +04:00
struct dm_list *snh, *snht;
2005-12-02 23:35:07 +03:00
struct lv_segment *seg;
struct lv_segment *log_seg;
2007-01-25 01:06:11 +03:00
int (*monitor_fn) (struct lv_segment *s, int e);
uint32_t s;
static const struct lv_activate_opts zlaopts = { 0 };
struct lvinfo info;
const char *dso = NULL;
int new_unmonitor;
if (!laopts)
laopts = &zlaopts;
2005-12-02 23:35:07 +03:00
/* skip dmeventd code altogether */
if (dmeventd_monitor_mode() == DMEVENTD_MONITOR_IGNORE)
return 1;
2007-01-20 01:21:45 +03:00
/*
* Nothing to do if dmeventd configured not to be used.
*/
if (monitor && !dmeventd_monitor_mode())
return 1;
/*
* Activation of unused cache-pool activates metadata device as
* a public LV for clearing purpose.
* FIXME:
* As VG lock is held across whole operation unmonitored volume
* is usually OK since dmeventd couldn't do anything.
* However in case command would have crashed, such LV is
* left unmonitored and may potentially require dmeventd.
*/
if ((lv_is_cache_pool_data(lv) || lv_is_cache_pool_metadata(lv)) &&
!lv_is_used_cache_pool((find_pool_seg(first_seg(lv))->lv))) {
log_debug_activation("Skipping %smonitor of %s.%s",
(monitor) ? "" : "un", display_lvname(lv),
(monitor) ? " Cache pool activation for clearing only." : "");
return 1;
}
/*
* Allow to unmonitor thin pool via explicit pool unmonitor
* or unmonitor before the last thin pool user deactivation
* Skip unmonitor, if invoked via deactivation of thin volume
* and there is another thin pool user (open_count > 1)
* FIXME think about watch ruler influence.
*/
if (laopts->skip_in_use && lv_is_thin_pool(lv) &&
lv_info(lv->vg->cmd, lv, 1, &info, 1, 0) && (info.open_count > 1)) {
log_debug_activation("Skipping unmonitor of opened %s (open:%d)",
display_lvname(lv), info.open_count);
return 1;
}
/* Do not monitor snapshot that already covers origin */
if (monitor && lv_is_cow_covering_origin(lv)) {
log_debug_activation("Skipping monitor of snapshot larger "
"then origin %s.", display_lvname(lv));
return 1;
}
/*
* In case of a snapshot device, we monitor lv->snapshot->lv,
* not the actual LV itself.
*/
if (lv_is_cow(lv) && (laopts->no_merging || !lv_is_merging_cow(lv))) {
if (!(r = monitor_dev_for_events(cmd, lv->snapshot->lv, NULL, monitor)))
stack;
return r;
}
/*
* In case this LV is a snapshot origin, we instead monitor
* each of its respective snapshots. The origin itself may
* also need to be monitored if it is a mirror, for example.
*/
if (!laopts->origin_only && lv_is_origin(lv))
dm_list_iterate_safe(snh, snht, &lv->snapshot_segs)
if (!monitor_dev_for_events(cmd, dm_list_struct_base(snh,
struct lv_segment, origin_list)->cow, NULL, monitor)) {
stack;
2008-01-17 20:17:09 +03:00
r = 0;
}
/*
* If the volume is mirrored and its log is also mirrored, monitor
* the log volume as well.
*/
if ((seg = first_seg(lv)) != NULL && seg->log_lv != NULL &&
(log_seg = first_seg(seg->log_lv)) != NULL &&
seg_is_mirrored(log_seg))
if (!monitor_dev_for_events(cmd, seg->log_lv, NULL, monitor)) {
stack;
r = 0;
}
2013-04-25 13:46:17 +04:00
dm_list_iterate_items(seg, &lv->segments) {
/* Recurse for AREA_LV */
for (s = 0; s < seg->area_count; s++) {
if (seg_type(seg, s) != AREA_LV)
continue;
if (!monitor_dev_for_events(cmd, seg_lv(seg, s), NULL,
monitor)) {
log_error("Failed to %smonitor %s",
monitor ? "" : "un",
display_lvname(seg_lv(seg, s)));
r = 0;
}
}
/*
* If requested unmonitoring of thin volume, preserve skip_in_use flag.
*
* FIXME: code here looks like _lv_postorder()
*/
if (seg->pool_lv &&
!monitor_dev_for_events(cmd, seg->pool_lv,
(!monitor) ? laopts : NULL, monitor)) {
stack;
r = 0;
}
if (seg->metadata_lv &&
!monitor_dev_for_events(cmd, seg->metadata_lv, NULL, monitor)) {
stack;
r = 0;
}
if (!seg_monitored(seg) ||
(seg->status & PVMOVE) ||
!seg->segtype->ops->target_monitored) /* doesn't support registration */
continue;
2007-01-20 01:21:45 +03:00
if (!monitor)
/* When unmonitoring, obtain existing dso being used. */
monitored = _device_registered_with_dmeventd(cmd, seg->lv, &pending, &dso);
else
monitored = seg->segtype->ops->target_monitored(seg, &pending);
/* FIXME: We should really try again if pending */
2007-01-20 01:21:45 +03:00
monitored = (pending) ? 0 : monitored;
monitor_fn = NULL;
new_unmonitor = 0;
2007-01-20 01:21:45 +03:00
if (monitor) {
if (monitored)
log_verbose("%s already monitored.", display_lvname(lv));
else if (seg->segtype->ops->target_monitor_events) {
log_verbose("Monitoring %s%s", display_lvname(lv), test_mode() ? " [Test mode: skipping this]" : "");
2007-01-20 01:21:45 +03:00
monitor_fn = seg->segtype->ops->target_monitor_events;
}
} else {
2007-01-20 01:21:45 +03:00
if (!monitored)
log_verbose("%s already not monitored.", display_lvname(lv));
else if (dso && *dso) {
/*
* Divert unmonitor away from code that depends on the new segment
* type instead of the existing one if it's changing.
*/
log_verbose("Not monitoring %s with %s%s", display_lvname(lv), dso, test_mode() ? " [Test mode: skipping this]" : "");
new_unmonitor = 1;
}
}
2005-12-02 23:35:07 +03:00
/* FIXME Test mode should really continue a bit further. */
if (test_mode())
continue;
if (new_unmonitor) {
if (!target_register_events(cmd, dso, lv, 0, 0, 10)) {
log_error("%s: segment unmonitoring failed.",
display_lvname(lv));
return 0;
}
} else if (monitor_fn) {
/* FIXME specify events */
if (!monitor_fn(seg, 0)) {
log_error("%s: %s segment monitoring function failed.",
display_lvname(lv), seg->segtype->name);
return 0;
}
} else
continue;
/* Check [un]monitor results */
/* Try a couple times if pending, but not forever... */
for (i = 0; i < 40; i++) {
pending = 0;
2007-01-20 01:21:45 +03:00
monitored = seg->segtype->ops->target_monitored(seg, &pending);
if (pending ||
2007-01-20 01:21:45 +03:00
(!monitored && monitor) ||
(monitored && !monitor))
log_very_verbose("%s %smonitoring still pending: waiting...",
display_lvname(lv), monitor ? "" : "un");
else
break;
usleep(10000 * i);
}
if (r)
r = (monitored && monitor) || (!monitored && !monitor);
2005-12-02 23:35:07 +03:00
}
if (!r && !error_message_produced())
log_error("%sonitoring %s failed.", monitor ? "M" : "Not m",
display_lvname(lv));
return r;
#else
2005-12-02 23:35:07 +03:00
return 1;
#endif
2005-12-02 23:35:07 +03:00
}
struct detached_lv_data {
const struct logical_volume *lv_pre;
struct lv_activate_opts *laopts;
int *flush_required;
};
static int _preload_detached_lv(struct logical_volume *lv, void *data)
{
struct detached_lv_data *detached = data;
struct logical_volume *lv_pre;
/* Check and preload removed raid image leg or metadata */
if (lv_is_raid_image(lv)) {
if ((lv_pre = find_lv_in_vg_by_lvid(detached->lv_pre->vg, &lv->lvid)) &&
!lv_is_raid_image(lv_pre) && lv_is_active(lv) &&
!_lv_preload(lv_pre, detached->laopts, detached->flush_required))
return_0;
} else if (lv_is_raid_metadata(lv)) {
if ((lv_pre = find_lv_in_vg_by_lvid(detached->lv_pre->vg, &lv->lvid)) &&
!lv_is_raid_metadata(lv_pre) && lv_is_active(lv) &&
!_lv_preload(lv_pre, detached->laopts, detached->flush_required))
return_0;
}
/* FIXME: condition here should be far more limiting to really
* detect detached LVs */
if ((lv_pre = find_lv(detached->lv_pre->vg, lv->name))) {
if (lv_is_visible(lv_pre) && lv_is_active(lv) &&
!lv_is_pool(lv) &&
(!lv_is_cow(lv) || !lv_is_cow(lv_pre)) &&
!_lv_preload(lv_pre, detached->laopts, detached->flush_required))
return_0;
}
return 1;
}
2004-03-08 21:54:13 +03:00
static int _lv_suspend(struct cmd_context *cmd, const char *lvid_s,
struct lv_activate_opts *laopts, int error_if_not_suspended,
const struct logical_volume *lv, const struct logical_volume *lv_pre)
{
const struct logical_volume *pvmove_lv = NULL;
const struct logical_volume *lv_to_free = NULL;
const struct logical_volume *lv_pre_to_free = NULL;
struct logical_volume *lv_pre_tmp;
struct seg_list *sl;
2016-04-06 11:28:02 +03:00
struct lv_segment *snap_seg;
struct lvinfo info;
int r = 0, lockfs = 0, flush_required = 0;
struct detached_lv_data detached;
2002-11-18 17:01:16 +03:00
if (!activation())
return 1;
/* lv comes from committed metadata */
if (!lv && !(lv_to_free = lv = lv_from_lvid(cmd, lvid_s, 0)))
goto_out;
/* Use precommitted metadata if present */
if (!lv_pre && !(lv_pre_to_free = lv_pre = lv_from_lvid(cmd, lvid_s, 1)))
goto_out;
/* Ignore origin_only unless LV is origin in both old and new metadata */
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
/* or LV is thin or thin pool volume */
if (!lv_is_thin_volume(lv) && !lv_is_thin_pool(lv) &&
!(lv_is_origin(lv) && lv_is_origin(lv_pre)))
laopts->origin_only = 0;
2002-03-14 18:36:07 +03:00
if (test_mode()) {
_skip("Suspending %s%s.", display_lvname(lv),
laopts->origin_only ? " origin without snapshots" : "");
r = 1;
goto out;
2002-03-14 18:36:07 +03:00
}
if (!lv_info(cmd, lv, laopts->origin_only, &info, 0, 0))
goto_out;
2002-03-11 22:02:28 +03:00
if (!info.exists || info.suspended) {
if (!error_if_not_suspended) {
r = 1;
if (info.suspended)
critical_section_inc(cmd, "already suspended");
}
goto out;
}
if (!lv_read_replicator_vgs(lv))
goto_out;
lv_calculate_readahead(lv, NULL);
/*
* Preload devices for the LV.
* If the PVMOVE LV is being removed, it's only present in the old
* metadata and not the new, so we must explicitly add the new
* tables for all the changed LVs here, as the relationships
* are not found by walking the new metadata.
*/
if (lv_is_locked(lv) && !lv_is_locked(lv_pre) &&
(pvmove_lv = find_pvmove_lv_in_lv(lv))) {
/* Preload all the LVs above the PVMOVE LV */
dm_list_iterate_items(sl, &pvmove_lv->segs_using_this_lv) {
if (!(lv_pre_tmp = find_lv(lv_pre->vg, sl->seg->lv->name))) {
log_error(INTERNAL_ERROR "LV %s missing from preload metadata.",
display_lvname(sl->seg->lv));
goto out;
}
if (!_lv_preload(lv_pre_tmp, laopts, &flush_required))
goto_out;
}
/* Now preload the PVMOVE LV itself */
if (!(lv_pre_tmp = find_lv(lv_pre->vg, pvmove_lv->name))) {
log_error(INTERNAL_ERROR "LV %s missing from preload metadata.",
display_lvname(pvmove_lv));
goto out;
}
if (!_lv_preload(lv_pre_tmp, laopts, &flush_required))
goto_out;
} else {
if (!_lv_preload(lv_pre, laopts, &flush_required))
/* FIXME Revert preloading */
goto_out;
/*
* Search for existing LVs that have become detached and preload them.
*/
detached.lv_pre = lv_pre;
detached.laopts = laopts;
detached.flush_required = &flush_required;
if (!for_each_sub_lv((struct logical_volume *)lv, &_preload_detached_lv, &detached))
goto_out;
/*
* Preload any snapshots that are being removed.
*/
if (!laopts->origin_only && lv_is_origin(lv)) {
dm_list_iterate_items_gen(snap_seg, &lv->snapshot_segs, origin_list) {
if (!(lv_pre_tmp = find_lv_in_vg_by_lvid(lv_pre->vg, &snap_seg->cow->lvid))) {
log_error(INTERNAL_ERROR "LV %s (%s) missing from preload metadata.",
display_lvname(snap_seg->cow),
snap_seg->cow->lvid.id[1].uuid);
goto out;
}
if (!lv_is_cow(lv_pre_tmp) &&
!_lv_preload(lv_pre_tmp, laopts, &flush_required))
goto_out;
}
}
}
/* Flush is ATM required for the tested cases
* NOTE: Mirror repair requires noflush for proper repair!
* TODO: Relax this limiting condition further */
if (!flush_required &&
(lv_is_pvmove(lv) ||
(!lv_is_mirror(lv) && !lv_is_thin_pool(lv) && !lv_is_thin_volume(lv)))) {
log_debug("Requiring flush for LV %s.", display_lvname(lv));
flush_required = 1;
}
if (!monitor_dev_for_events(cmd, lv, laopts, 0))
/* FIXME Consider aborting here */
stack;
critical_section_inc(cmd, "suspending");
if (pvmove_lv)
critical_section_inc(cmd, "suspending pvmove LV");
if (!laopts->origin_only &&
(lv_is_origin(lv_pre) || lv_is_cow(lv_pre)))
lockfs = 1;
/* Converting non-thin LV to thin external origin ? */
if (!lv_is_thin_volume(lv) && lv_is_thin_volume(lv_pre))
lockfs = 1; /* Sync before conversion */
if (laopts->origin_only && lv_is_thin_volume(lv) && lv_is_thin_volume(lv_pre))
lockfs = 1;
/*
* Suspending an LV directly above a PVMOVE LV also
* suspends other LVs using that same PVMOVE LV.
* FIXME Remove this and delay the 'clear node' until
* after the code knows whether there's a different
* inactive table to load or not instead so lv_suspend
* can be called separately for each LV safely.
*/
if ((lv_pre->vg->status & PRECOMMITTED) &&
lv_is_locked(lv_pre) && find_pvmove_lv_in_lv(lv_pre)) {
if (!_lv_suspend_lv(lv_pre, laopts, lockfs, flush_required)) {
critical_section_dec(cmd, "failed precommitted suspend");
if (pvmove_lv)
critical_section_dec(cmd, "failed precommitted suspend (pvmove)");
goto_out;
}
} else {
/* Normal suspend */
if (!_lv_suspend_lv(lv, laopts, lockfs, flush_required)) {
critical_section_dec(cmd, "failed suspend");
if (pvmove_lv)
critical_section_dec(cmd, "failed suspend (pvmove)");
goto_out;
}
}
2002-03-01 22:08:11 +03:00
r = 1;
out:
if (lv_pre_to_free)
release_vg(lv_pre_to_free->vg);
if (lv_to_free) {
lv_release_replicator_vgs(lv_to_free);
release_vg(lv_to_free->vg);
}
return r;
}
2012-01-20 07:46:52 +04:00
/*
* In a cluster, set exclusive to indicate that only one node is using the
* device. Any preloaded tables may then use non-clustered targets.
*
* Returns success if the device is not active
*/
int lv_suspend_if_active(struct cmd_context *cmd, const char *lvid_s, unsigned origin_only, unsigned exclusive,
const struct logical_volume *lv, const struct logical_volume *lv_pre)
2004-03-08 21:54:13 +03:00
{
struct lv_activate_opts laopts = {
.origin_only = origin_only,
.exclusive = exclusive
};
return _lv_suspend(cmd, lvid_s, &laopts, 0, lv, lv_pre);
2004-03-08 21:54:13 +03:00
}
static int _lv_resume(struct cmd_context *cmd, const char *lvid_s,
struct lv_activate_opts *laopts, int error_if_not_active,
const struct logical_volume *lv)
{
const struct logical_volume *lv_to_free = NULL;
struct lvinfo info;
int r = 0;
2002-11-18 17:01:16 +03:00
if (!activation())
return 1;
if (!lv && !(lv_to_free = lv = lv_from_lvid(cmd, lvid_s, 0)))
goto_out;
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
if (!lv_is_origin(lv) && !lv_is_thin_volume(lv) && !lv_is_thin_pool(lv))
laopts->origin_only = 0;
2002-03-14 18:36:07 +03:00
if (test_mode()) {
_skip("Resuming %s%s%s.", display_lvname(lv),
laopts->origin_only ? " without snapshots" : "",
laopts->revert ? " (reverting)" : "");
r = 1;
goto out;
2002-03-14 18:36:07 +03:00
}
log_debug_activation("Resuming LV %s%s%s%s.", display_lvname(lv),
error_if_not_active ? "" : " if active",
laopts->origin_only ?
(lv_is_thin_pool(lv) ? " pool only" :
lv_is_thin_volume(lv) ? " thin only" : " without snapshots") : "",
laopts->revert ? " (reverting)" : "");
if (!lv_info(cmd, lv, laopts->origin_only, &info, 0, 0))
goto_out;
2002-03-11 22:02:28 +03:00
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
if (!info.exists || !info.suspended) {
if (error_if_not_active)
goto_out;
r = 1;
if (!info.suspended)
critical_section_dec(cmd, "already resumed");
goto out;
}
laopts->read_only = _passes_readonly_filter(cmd, lv);
laopts->resuming = 1;
if (!_lv_activate_lv(lv, laopts))
goto_out;
critical_section_dec(cmd, "resumed");
if (!monitor_dev_for_events(cmd, lv, laopts, 1))
stack;
2005-12-02 23:35:07 +03:00
r = 1;
out:
if (lv_to_free)
release_vg(lv_to_free->vg);
return r;
}
2012-01-20 07:46:52 +04:00
/*
* In a cluster, set exclusive to indicate that only one node is using the
* device. Any tables loaded may then use non-clustered targets.
*
* @origin_only
* @exclusive This parameter only has an affect in cluster-context.
* It forces local target type to be used (instead of
* cluster-aware type).
2012-01-20 07:46:52 +04:00
* Returns success if the device is not active
*/
int lv_resume_if_active(struct cmd_context *cmd, const char *lvid_s,
2012-01-20 07:46:52 +04:00
unsigned origin_only, unsigned exclusive,
unsigned revert, const struct logical_volume *lv)
2004-03-08 21:54:13 +03:00
{
struct lv_activate_opts laopts = {
.origin_only = origin_only,
.exclusive = exclusive,
.revert = revert
};
return _lv_resume(cmd, lvid_s, &laopts, 0, lv);
2004-03-08 21:54:13 +03:00
}
int lv_resume(struct cmd_context *cmd, const char *lvid_s, unsigned origin_only,
const struct logical_volume *lv)
2004-03-08 21:54:13 +03:00
{
struct lv_activate_opts laopts = { .origin_only = origin_only, };
return _lv_resume(cmd, lvid_s, &laopts, 1, lv);
2004-03-08 21:54:13 +03:00
}
static int _lv_has_open_snapshots(const struct logical_volume *lv)
{
struct lv_segment *snap_seg;
int r = 0;
dm_list_iterate_items_gen(snap_seg, &lv->snapshot_segs, origin_list)
if (!lv_check_not_in_use(snap_seg->cow, 1))
r++;
if (r)
log_error("LV %s has open %d snapshot(s), not deactivating.",
display_lvname(lv), r);
return r;
}
int lv_deactivate(struct cmd_context *cmd, const char *lvid_s, const struct logical_volume *lv)
{
const struct logical_volume *lv_to_free = NULL;
struct lvinfo info;
static const struct lv_activate_opts laopts = { .skip_in_use = 1 };
struct dm_list *snh;
int r = 0;
2002-11-18 17:01:16 +03:00
if (!activation())
return 1;
if (!lv && !(lv_to_free = lv = lv_from_lvid(cmd, lvid_s, 0)))
goto out;
2002-03-14 18:36:07 +03:00
if (test_mode()) {
_skip("Deactivating %s.", display_lvname(lv));
r = 1;
goto out;
2002-03-14 18:36:07 +03:00
}
log_debug_activation("Deactivating %s.", display_lvname(lv));
if (!lv_info(cmd, lv, 0, &info, 0, 0))
goto_out;
2002-03-11 22:02:28 +03:00
if (!info.exists) {
r = 1;
/* Check attached snapshot segments are also inactive */
dm_list_iterate(snh, &lv->snapshot_segs) {
if (!lv_info(cmd, dm_list_struct_base(snh, struct lv_segment, origin_list)->cow,
0, &info, 0, 0))
goto_out;
if (info.exists) {
r = 0; /* Snapshot left in table? */
break;
}
}
if (r)
goto out;
}
if (lv_is_visible(lv) || lv_is_virtual_origin(lv) ||
lv_is_merging_thin_snapshot(lv)) {
if (!lv_check_not_in_use(lv, 1))
goto_out;
if (lv_is_origin(lv) && _lv_has_open_snapshots(lv))
goto_out;
}
if (!lv_read_replicator_vgs(lv))
goto_out;
if (!monitor_dev_for_events(cmd, lv, &laopts, 0))
stack;
2005-12-02 23:35:07 +03:00
critical_section_inc(cmd, "deactivating");
r = _lv_deactivate(lv);
critical_section_dec(cmd, "deactivated");
if (!lv_info(cmd, lv, 0, &info, 0, 0) || info.exists) {
/* Turn into log_error, but we do not log error */
log_debug_activation("Deactivated volume is still %s present.",
display_lvname(lv));
r = 0;
}
out:
if (lv_to_free) {
lv_release_replicator_vgs(lv_to_free);
release_vg(lv_to_free->vg);
}
return r;
}
2004-03-08 21:54:13 +03:00
/* Test if LV passes filter */
int lv_activation_filter(struct cmd_context *cmd, const char *lvid_s,
int *activate_lv, const struct logical_volume *lv)
2004-03-08 21:54:13 +03:00
{
const struct logical_volume *lv_to_free = NULL;
int r = 0;
2004-03-08 21:54:13 +03:00
if (!activation()) {
*activate_lv = 1;
return 1;
}
2004-03-08 21:54:13 +03:00
if (!lv && !(lv_to_free = lv = lv_from_lvid(cmd, lvid_s, 0)))
goto_out;
2004-03-08 21:54:13 +03:00
if (!_passes_activation_filter(cmd, lv)) {
log_verbose("Not activating %s since it does not pass "
"activation filter.", display_lvname(lv));
2004-03-08 21:54:13 +03:00
*activate_lv = 0;
} else
*activate_lv = 1;
r = 1;
out:
if (lv_to_free)
release_vg(lv_to_free->vg);
2004-03-08 21:54:13 +03:00
return r;
2004-03-08 21:54:13 +03:00
}
static int _lv_activate(struct cmd_context *cmd, const char *lvid_s,
struct lv_activate_opts *laopts, int filter,
const struct logical_volume *lv)
{
const struct logical_volume *lv_to_free = NULL;
struct lvinfo info;
int r = 0;
2002-11-18 17:01:16 +03:00
if (!activation())
return 1;
if (!lv && !(lv_to_free = lv = lv_from_lvid(cmd, lvid_s, 0)))
goto out;
2004-03-08 21:54:13 +03:00
if (filter && !_passes_activation_filter(cmd, lv)) {
log_verbose("Not activating %s since it does not pass "
"activation filter.", display_lvname(lv));
r = 1;
goto out;
2004-03-08 21:13:22 +03:00
}
if ((!lv->vg->cmd->partial_activation) && lv_is_partial(lv)) {
if (!lv_is_raid_type(lv) || !partial_raid_lv_supports_degraded_activation(lv)) {
activation: Add "degraded" activation mode Currently, we have two modes of activation, an unnamed nominal mode (which I will refer to as "complete") and "partial" mode. The "complete" mode requires that a volume group be 'complete' - that is, no missing PVs. If there are any missing PVs, no affected LVs are allowed to activate - even RAID LVs which might be able to tolerate a failure. The "partial" mode allows anything to be activated (or at least attempted). If a non-redundant LV is missing a portion of its addressable space due to a device failure, it will be replaced with an error target. RAID LVs will either activate or fail to activate depending on how badly their redundancy is compromised. This patch adds a third option, "degraded" mode. This mode can be selected via the '--activationmode {complete|degraded|partial}' option to lvchange/vgchange. It can also be set in lvm.conf. The "degraded" activation mode allows RAID LVs with a sufficient level of redundancy to activate (e.g. a RAID5 LV with one device failure, a RAID6 with two device failures, or RAID1 with n-1 failures). RAID LVs with too many device failures are not allowed to activate - nor are any non-redundant LVs that may have been affected. This patch also makes the "degraded" mode the default activation mode. The degraded activation mode does not yet work in a cluster. A new cluster lock flag (LCK_DEGRADED_MODE) will need to be created to make that work. Currently, there is limited space for this extra flag and I am looking for possible solutions. One possible solution is to usurp LCK_CONVERT, as it is not used. When the locking_type is 3, the degraded mode flag simply gets dropped and the old ("complete") behavior is exhibited.
2014-07-10 07:56:11 +04:00
log_error("Refusing activation of partial LV %s. "
"Use '--activationmode partial' to override.",
display_lvname(lv));
activation: Add "degraded" activation mode Currently, we have two modes of activation, an unnamed nominal mode (which I will refer to as "complete") and "partial" mode. The "complete" mode requires that a volume group be 'complete' - that is, no missing PVs. If there are any missing PVs, no affected LVs are allowed to activate - even RAID LVs which might be able to tolerate a failure. The "partial" mode allows anything to be activated (or at least attempted). If a non-redundant LV is missing a portion of its addressable space due to a device failure, it will be replaced with an error target. RAID LVs will either activate or fail to activate depending on how badly their redundancy is compromised. This patch adds a third option, "degraded" mode. This mode can be selected via the '--activationmode {complete|degraded|partial}' option to lvchange/vgchange. It can also be set in lvm.conf. The "degraded" activation mode allows RAID LVs with a sufficient level of redundancy to activate (e.g. a RAID5 LV with one device failure, a RAID6 with two device failures, or RAID1 with n-1 failures). RAID LVs with too many device failures are not allowed to activate - nor are any non-redundant LVs that may have been affected. This patch also makes the "degraded" mode the default activation mode. The degraded activation mode does not yet work in a cluster. A new cluster lock flag (LCK_DEGRADED_MODE) will need to be created to make that work. Currently, there is limited space for this extra flag and I am looking for possible solutions. One possible solution is to usurp LCK_CONVERT, as it is not used. When the locking_type is 3, the degraded mode flag simply gets dropped and the old ("complete") behavior is exhibited.
2014-07-10 07:56:11 +04:00
goto out;
}
if (!lv->vg->cmd->degraded_activation) {
activation: Add "degraded" activation mode Currently, we have two modes of activation, an unnamed nominal mode (which I will refer to as "complete") and "partial" mode. The "complete" mode requires that a volume group be 'complete' - that is, no missing PVs. If there are any missing PVs, no affected LVs are allowed to activate - even RAID LVs which might be able to tolerate a failure. The "partial" mode allows anything to be activated (or at least attempted). If a non-redundant LV is missing a portion of its addressable space due to a device failure, it will be replaced with an error target. RAID LVs will either activate or fail to activate depending on how badly their redundancy is compromised. This patch adds a third option, "degraded" mode. This mode can be selected via the '--activationmode {complete|degraded|partial}' option to lvchange/vgchange. It can also be set in lvm.conf. The "degraded" activation mode allows RAID LVs with a sufficient level of redundancy to activate (e.g. a RAID5 LV with one device failure, a RAID6 with two device failures, or RAID1 with n-1 failures). RAID LVs with too many device failures are not allowed to activate - nor are any non-redundant LVs that may have been affected. This patch also makes the "degraded" mode the default activation mode. The degraded activation mode does not yet work in a cluster. A new cluster lock flag (LCK_DEGRADED_MODE) will need to be created to make that work. Currently, there is limited space for this extra flag and I am looking for possible solutions. One possible solution is to usurp LCK_CONVERT, as it is not used. When the locking_type is 3, the degraded mode flag simply gets dropped and the old ("complete") behavior is exhibited.
2014-07-10 07:56:11 +04:00
log_error("Refusing activation of partial LV %s. "
"Try '--activationmode degraded'.",
display_lvname(lv));
activation: Add "degraded" activation mode Currently, we have two modes of activation, an unnamed nominal mode (which I will refer to as "complete") and "partial" mode. The "complete" mode requires that a volume group be 'complete' - that is, no missing PVs. If there are any missing PVs, no affected LVs are allowed to activate - even RAID LVs which might be able to tolerate a failure. The "partial" mode allows anything to be activated (or at least attempted). If a non-redundant LV is missing a portion of its addressable space due to a device failure, it will be replaced with an error target. RAID LVs will either activate or fail to activate depending on how badly their redundancy is compromised. This patch adds a third option, "degraded" mode. This mode can be selected via the '--activationmode {complete|degraded|partial}' option to lvchange/vgchange. It can also be set in lvm.conf. The "degraded" activation mode allows RAID LVs with a sufficient level of redundancy to activate (e.g. a RAID5 LV with one device failure, a RAID6 with two device failures, or RAID1 with n-1 failures). RAID LVs with too many device failures are not allowed to activate - nor are any non-redundant LVs that may have been affected. This patch also makes the "degraded" mode the default activation mode. The degraded activation mode does not yet work in a cluster. A new cluster lock flag (LCK_DEGRADED_MODE) will need to be created to make that work. Currently, there is limited space for this extra flag and I am looking for possible solutions. One possible solution is to usurp LCK_CONVERT, as it is not used. When the locking_type is 3, the degraded mode flag simply gets dropped and the old ("complete") behavior is exhibited.
2014-07-10 07:56:11 +04:00
goto out;
}
}
if (lv_has_unknown_segments(lv)) {
log_error("Refusing activation of LV %s containing "
"an unrecognised segment.", display_lvname(lv));
goto out;
}
/*
2015-04-13 12:00:07 +03:00
* Check if cmirrord is running for clustered mirrors.
*/
if (!laopts->exclusive && vg_is_clustered(lv->vg) &&
lv_is_mirror(lv) && !lv_is_raid(lv) &&
!cluster_mirror_is_available(lv->vg->cmd)) {
log_error("Shared cluster mirrors are not available.");
goto out;
}
2002-03-14 18:36:07 +03:00
if (test_mode()) {
_skip("Activating %s.", display_lvname(lv));
r = 1;
goto out;
2002-03-14 18:36:07 +03:00
}
if (filter)
laopts->read_only = _passes_readonly_filter(cmd, lv);
log_debug_activation("Activating %s%s%s%s%s.", display_lvname(lv),
laopts->exclusive ? " exclusively" : "",
laopts->read_only ? " read-only" : "",
activation: flag temporary LVs internally Add LV_TEMPORARY flag for LVs with limited existence during command execution. Such LVs are temporary in way that they need to be activated, some action done and then removed immediately. Such LVs are just like any normal LV - the only difference is that they are removed during LVM command execution. This is also the case for LVs representing future pool metadata spare LVs which we need to initialize by using the usual LV before they are declared as pool metadata spare. We can optimize some other parts like udev to do a better job if it knows that the LV is temporary and any processing on it is just useless. This flag is orthogonal to LV_NOSCAN flag introduced recently as LV_NOSCAN flag is primarily used to mark an LV for the scanning to be avoided before the zeroing of the device happens. The LV_TEMPORARY flag makes a difference between a full-fledged LV visible in the system and the LV just used as a temporary overlay for some action that needs to be done on underlying PVs. For example: lvcreate --thinpool POOL --zero n -L 1G vg - first, the usual LV is created to do a clean up for pool metadata spare. The LV is activated, zeroed, deactivated. - between "activated" and "zeroed" stage, the LV_NOSCAN flag is used to avoid any scanning in udev - betwen "zeroed" and "deactivated" stage, we need to avoid the WATCH udev rule, but since the LV is just a usual LV, we can't make a difference. The LV_TEMPORARY internal LV flag helps here. If we create the LV with this flag, the DM_UDEV_DISABLE_DISK_RULES and DM_UDEV_DISABLE_OTHER_RULES flag are set (just like as it is with "invisible" and non-top-level LVs) - udev is directed to skip WATCH rule use. - if the LV_TEMPORARY flag was not used, there would normally be a WATCH event generated once the LV is closed after "zeroed" stage. This will make problems with immediated deactivation that follows.
2013-10-23 16:06:39 +04:00
laopts->noscan ? " noscan" : "",
laopts->temporary ? " temporary" : "");
if (!lv_info(cmd, lv, 0, &info, 0, 0))
goto_out;
2002-03-01 22:08:11 +03:00
/*
* Nothing to do?
*/
if (info.exists && !info.suspended && info.live_table &&
(info.read_only == read_only_lv(lv, laopts))) {
r = 1;
log_debug_activation("LV %s is already active.", display_lvname(lv));
goto out;
}
if (!lv_read_replicator_vgs(lv))
goto_out;
lv_calculate_readahead(lv, NULL);
critical_section_inc(cmd, "activating");
if (!(r = _lv_activate_lv(lv, laopts)))
stack;
critical_section_dec(cmd, "activated");
if (r && !monitor_dev_for_events(cmd, lv, laopts, 1))
stack;
2005-12-02 23:35:07 +03:00
out:
if (lv_to_free) {
lv_release_replicator_vgs(lv_to_free);
release_vg(lv_to_free->vg);
}
return r;
}
2004-03-08 21:54:13 +03:00
/* Activate LV */
int lv_activate(struct cmd_context *cmd, const char *lvid_s, int exclusive,
int noscan, int temporary, const struct logical_volume *lv)
2004-03-08 21:54:13 +03:00
{
activation: flag temporary LVs internally Add LV_TEMPORARY flag for LVs with limited existence during command execution. Such LVs are temporary in way that they need to be activated, some action done and then removed immediately. Such LVs are just like any normal LV - the only difference is that they are removed during LVM command execution. This is also the case for LVs representing future pool metadata spare LVs which we need to initialize by using the usual LV before they are declared as pool metadata spare. We can optimize some other parts like udev to do a better job if it knows that the LV is temporary and any processing on it is just useless. This flag is orthogonal to LV_NOSCAN flag introduced recently as LV_NOSCAN flag is primarily used to mark an LV for the scanning to be avoided before the zeroing of the device happens. The LV_TEMPORARY flag makes a difference between a full-fledged LV visible in the system and the LV just used as a temporary overlay for some action that needs to be done on underlying PVs. For example: lvcreate --thinpool POOL --zero n -L 1G vg - first, the usual LV is created to do a clean up for pool metadata spare. The LV is activated, zeroed, deactivated. - between "activated" and "zeroed" stage, the LV_NOSCAN flag is used to avoid any scanning in udev - betwen "zeroed" and "deactivated" stage, we need to avoid the WATCH udev rule, but since the LV is just a usual LV, we can't make a difference. The LV_TEMPORARY internal LV flag helps here. If we create the LV with this flag, the DM_UDEV_DISABLE_DISK_RULES and DM_UDEV_DISABLE_OTHER_RULES flag are set (just like as it is with "invisible" and non-top-level LVs) - udev is directed to skip WATCH rule use. - if the LV_TEMPORARY flag was not used, there would normally be a WATCH event generated once the LV is closed after "zeroed" stage. This will make problems with immediated deactivation that follows.
2013-10-23 16:06:39 +04:00
struct lv_activate_opts laopts = { .exclusive = exclusive,
.noscan = noscan,
.temporary = temporary };
if (!_lv_activate(cmd, lvid_s, &laopts, 0, lv))
return_0;
return 1;
2004-03-08 21:54:13 +03:00
}
/* Activate LV only if it passes filter */
int lv_activate_with_filter(struct cmd_context *cmd, const char *lvid_s, int exclusive,
int noscan, int temporary, const struct logical_volume *lv)
2004-03-08 21:54:13 +03:00
{
activation: flag temporary LVs internally Add LV_TEMPORARY flag for LVs with limited existence during command execution. Such LVs are temporary in way that they need to be activated, some action done and then removed immediately. Such LVs are just like any normal LV - the only difference is that they are removed during LVM command execution. This is also the case for LVs representing future pool metadata spare LVs which we need to initialize by using the usual LV before they are declared as pool metadata spare. We can optimize some other parts like udev to do a better job if it knows that the LV is temporary and any processing on it is just useless. This flag is orthogonal to LV_NOSCAN flag introduced recently as LV_NOSCAN flag is primarily used to mark an LV for the scanning to be avoided before the zeroing of the device happens. The LV_TEMPORARY flag makes a difference between a full-fledged LV visible in the system and the LV just used as a temporary overlay for some action that needs to be done on underlying PVs. For example: lvcreate --thinpool POOL --zero n -L 1G vg - first, the usual LV is created to do a clean up for pool metadata spare. The LV is activated, zeroed, deactivated. - between "activated" and "zeroed" stage, the LV_NOSCAN flag is used to avoid any scanning in udev - betwen "zeroed" and "deactivated" stage, we need to avoid the WATCH udev rule, but since the LV is just a usual LV, we can't make a difference. The LV_TEMPORARY internal LV flag helps here. If we create the LV with this flag, the DM_UDEV_DISABLE_DISK_RULES and DM_UDEV_DISABLE_OTHER_RULES flag are set (just like as it is with "invisible" and non-top-level LVs) - udev is directed to skip WATCH rule use. - if the LV_TEMPORARY flag was not used, there would normally be a WATCH event generated once the LV is closed after "zeroed" stage. This will make problems with immediated deactivation that follows.
2013-10-23 16:06:39 +04:00
struct lv_activate_opts laopts = { .exclusive = exclusive,
.noscan = noscan,
.temporary = temporary };
if (!_lv_activate(cmd, lvid_s, &laopts, 1, lv))
return_0;
return 1;
2004-03-08 21:54:13 +03:00
}
int lv_mknodes(struct cmd_context *cmd, const struct logical_volume *lv)
{
int r;
if (!lv) {
r = dm_mknodes(NULL);
fs_unlock();
return r;
}
if (!activation())
return 1;
r = dev_manager_mknodes(lv);
fs_unlock();
return r;
}
/*
* Does PV use VG somewhere in its construction?
* Returns 1 on failure.
*/
int pv_uses_vg(struct physical_volume *pv,
struct volume_group *vg)
{
if (!activation() || !pv->dev)
return 0;
if (!dm_is_dm_major(MAJOR(pv->dev->dev)))
return 0;
return dev_manager_device_uses_vg(pv->dev, vg);
}
void activation_release(void)
{
if (critical_section())
/* May leak stacked operation */
log_error("Releasing activation in critical section.");
fs_unlock(); /* Implicit dev_manager_release(); */
}
void activation_exit(void)
{
activation_release();
dev_manager_exit();
}
#endif