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

661 lines
15 KiB
C
Raw Normal View History

/*
2008-01-30 17:00:02 +03:00
* Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.
* Copyright (C) 2004-2007 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
*/
#include "lib.h"
#include "archiver.h"
#include "format-text.h"
#include "lvm-string.h"
#include "lvmcache.h"
#include "lvmetad.h"
#include "memlock.h"
#include "toolcontext.h"
#include "locking.h"
#include <unistd.h>
struct archive_params {
int enabled;
char *dir;
unsigned int keep_days;
unsigned int keep_number;
};
struct backup_params {
2003-09-15 19:03:22 +04:00
int enabled;
char *dir;
};
2003-09-15 19:03:22 +04:00
int archive_init(struct cmd_context *cmd, const char *dir,
unsigned int keep_days, unsigned int keep_min,
int enabled)
{
archive_exit(cmd);
if (!(cmd->archive_params = dm_pool_zalloc(cmd->libmem,
sizeof(*cmd->archive_params)))) {
log_error("archive_params alloc failed");
return 0;
}
cmd->archive_params->dir = NULL;
2002-01-09 16:17:14 +03:00
if (!*dir)
return 1;
if (!(cmd->archive_params->dir = dm_strdup(dir))) {
2002-01-09 16:17:14 +03:00
log_error("Couldn't copy archive directory name.");
return 0;
}
cmd->archive_params->keep_days = keep_days;
cmd->archive_params->keep_number = keep_min;
archive_enable(cmd, enabled);
return 1;
}
void archive_exit(struct cmd_context *cmd)
{
if (!cmd->archive_params)
return;
dm_free(cmd->archive_params->dir);
memset(cmd->archive_params, 0, sizeof(*cmd->archive_params));
}
void archive_enable(struct cmd_context *cmd, int flag)
{
cmd->archive_params->enabled = flag;
}
static char *_build_desc(struct dm_pool *mem, const char *line, int before)
{
size_t len = strlen(line) + 32;
char *buffer;
if (!(buffer = dm_pool_alloc(mem, len))) {
log_error("Failed to allocate desc.");
return NULL;
}
if (dm_snprintf(buffer, len, "Created %s executing '%s'",
before ? "*before*" : "*after*", line) < 0) {
log_error("Failed to build desc.");
return NULL;
}
return buffer;
}
static int _archive(struct volume_group *vg, int compulsory)
{
char *desc;
/* Don't archive orphan VGs. */
if (is_orphan_vg(vg->name))
return 1;
if (vg_is_archived(vg))
return 1; /* VG has been already archived */
if (!vg->cmd->archive_params->enabled || !vg->cmd->archive_params->dir) {
vg->status |= ARCHIVED_VG;
return 1;
}
if (test_mode()) {
vg->status |= ARCHIVED_VG;
2002-11-18 17:04:08 +03:00
log_verbose("Test mode: Skipping archiving of volume group.");
return 1;
}
if (!dm_create_dir(vg->cmd->archive_params->dir)) {
if (compulsory)
return_0;
return 1;
}
/* Trap a read-only file system */
2008-01-30 17:00:02 +03:00
if ((access(vg->cmd->archive_params->dir, R_OK | W_OK | X_OK) == -1) &&
(errno == EROFS)) {
if (compulsory) {
log_error("Cannot archive volume group metadata for %s to read-only filesystem.",
vg->name);
return 0;
}
return 1;
}
log_verbose("Archiving volume group \"%s\" metadata (seqno %u).", vg->name,
vg->seqno);
if (!(desc = _build_desc(vg->cmd->mem, vg->cmd->cmd_line, 1)))
return_0;
if (!archive_vg(vg, vg->cmd->archive_params->dir, desc,
vg->cmd->archive_params->keep_days,
vg->cmd->archive_params->keep_number))
return_0;
vg->status |= ARCHIVED_VG;
return 1;
}
int archive(struct volume_group *vg)
{
return _archive(vg, 1);
}
int archive_display(struct cmd_context *cmd, const char *vg_name)
{
2003-09-15 19:03:22 +04:00
int r1, r2;
r1 = archive_list(cmd, cmd->archive_params->dir, vg_name);
r2 = backup_list(cmd, cmd->backup_params->dir, vg_name);
2003-09-15 19:03:22 +04:00
return r1 && r2;
}
int archive_display_file(struct cmd_context *cmd, const char *file)
{
int r;
r = archive_list_file(cmd, file);
return r;
}
int backup_init(struct cmd_context *cmd, const char *dir,
int enabled)
{
backup_exit(cmd);
if (!(cmd->backup_params = dm_pool_zalloc(cmd->libmem,
sizeof(*cmd->backup_params)))) {
log_error("backup_params alloc failed");
return 0;
}
cmd->backup_params->dir = NULL;
2002-01-09 16:17:14 +03:00
if (!*dir)
return 1;
if (!(cmd->backup_params->dir = dm_strdup(dir))) {
2002-01-09 16:17:14 +03:00
log_error("Couldn't copy backup directory name.");
return 0;
}
backup_enable(cmd, enabled);
return 1;
}
void backup_exit(struct cmd_context *cmd)
{
if (!cmd->backup_params)
return;
dm_free(cmd->backup_params->dir);
memset(cmd->backup_params, 0, sizeof(*cmd->backup_params));
}
void backup_enable(struct cmd_context *cmd, int flag)
2002-01-07 12:05:31 +03:00
{
cmd->backup_params->enabled = flag;
2002-01-07 12:05:31 +03:00
}
static int _backup(struct volume_group *vg)
{
char name[PATH_MAX];
char *desc;
2008-01-30 16:19:47 +03:00
if (!(desc = _build_desc(vg->cmd->mem, vg->cmd->cmd_line, 0)))
return_0;
2006-08-21 16:54:53 +04:00
if (dm_snprintf(name, sizeof(name), "%s/%s",
vg->cmd->backup_params->dir, vg->name) < 0) {
2002-01-09 16:17:14 +03:00
log_error("Failed to generate volume group metadata backup "
"filename.");
return 0;
}
2002-11-18 17:04:08 +03:00
return backup_to_file(name, desc, vg);
}
int backup_locally(struct volume_group *vg)
{
if (!vg->cmd->backup_params->enabled || !vg->cmd->backup_params->dir) {
log_warn("WARNING: This metadata update is NOT backed up");
return 1;
2002-01-09 16:17:14 +03:00
}
if (test_mode()) {
2011-02-28 23:50:01 +03:00
log_verbose("Test mode: Skipping backup of volume group.");
return 1;
}
if (!dm_create_dir(vg->cmd->backup_params->dir))
return 0;
/* Trap a read-only file system */
2008-01-30 17:00:02 +03:00
if ((access(vg->cmd->backup_params->dir, R_OK | W_OK | X_OK) == -1) &&
(errno == EROFS)) {
/* Will take a backup next time when FS is writable */
log_debug("Skipping backup of volume group on read-only filesystem.");
2008-01-30 17:00:02 +03:00
return 0;
}
if (!_backup(vg)) {
2002-01-09 16:17:14 +03:00
log_error("Backup of volume group %s metadata failed.",
vg->name);
return 0;
}
return 1;
}
int backup(struct volume_group *vg)
{
/* Unlock memory if possible */
memlock_unlock(vg->cmd);
/* Don't back up orphan VGs. */
if (is_orphan_vg(vg->name))
return 1;
if (vg_is_clustered(vg))
if (!remote_backup_metadata(vg))
stack;
return backup_locally(vg);
}
int backup_remove(struct cmd_context *cmd, const char *vg_name)
{
char path[PATH_MAX];
2006-08-21 16:54:53 +04:00
if (dm_snprintf(path, sizeof(path), "%s/%s",
cmd->backup_params->dir, vg_name) < 0) {
log_error("Failed to generate backup filename (for removal).");
return 0;
}
/*
* Let this fail silently.
*/
if (unlink(path))
log_sys_debug("unlink", path);
return 1;
}
2002-11-18 17:04:08 +03:00
struct volume_group *backup_read_vg(struct cmd_context *cmd,
const char *vg_name, const char *file)
2002-01-10 17:27:47 +03:00
{
struct volume_group *vg = NULL;
2002-01-10 17:27:47 +03:00
struct format_instance *tf;
struct format_instance_ctx fic;
struct text_context tc = {.path_live = file,
.path_edit = NULL,
.desc = cmd->cmd_line};
2002-11-18 17:04:08 +03:00
struct metadata_area *mda;
2002-01-10 17:27:47 +03:00
2012-02-13 03:01:19 +04:00
fic.type = FMT_INSTANCE_PRIVATE_MDAS;
fic.context.private = &tc;
if (!(tf = cmd->fmt_backup->ops->create_instance(cmd->fmt_backup, &fic))) {
2002-01-10 17:27:47 +03:00
log_error("Couldn't create text format object.");
return NULL;
2002-01-10 17:27:47 +03:00
}
dm_list_iterate_items(mda, &tf->metadata_areas_in_use) {
if (!(vg = mda->ops->vg_read(tf, vg_name, mda, NULL, NULL, 0)))
2002-11-18 17:04:08 +03:00
stack;
break;
}
2002-01-10 17:27:47 +03:00
if (!vg)
tf->fmt->ops->destroy_instance(tf);
return vg;
2002-01-10 17:27:47 +03:00
}
static int _restore_vg_should_write_pv(struct physical_volume *pv, int do_pvcreate)
backup: backup_restore_vg: register PVs that need writing via vg->pvs_to_write list The backup_restore_vg is used directly for restoring the VG from backup. It's also used to do the VG conversions from one metadata format to another which means vgconvert calls backup_restore_vg too. When restoring VG from backup, we need to rewrite/write PV headers as PVs may have been orphans before and now they're becoming part of some VG - we need to write the PV_EXT_USED flag at least. When using the backup_restore_vg for vgconvert, we need to write completely new PV header in different format. Avoid the special "pv_write" call and handling that was used before this patch in vgconvert (vgconvert_single function to be more precise) and reuse existing internal interface to register PV header for writing (or rewriting) via vg->pvs_to_write list instead like we do it elsewhere in the code. This patch also resolves a problem in which PV headers with target format were written in the vgconvert_single fn as orphans and VG metadata were added later on - this was a tiny hack actually. We can't do this now - we need to write the PV as belonging to a VG because otherwise the PV_EXT_USED flag won't be written properly (if the PV header is written as orphan, the PV_EXT_USED is set to 0, of course, even though metadata are attached later). So this patch removes this tiny inconsistency which was passing just fine before because we didn't have any relation to the VG in PV header before. Now we have the PV_EXT_USED flag which says the "PV is used in some VG".
2015-03-26 16:20:46 +03:00
{
struct lvmcache_info *info;
if (do_pvcreate)
backup: backup_restore_vg: register PVs that need writing via vg->pvs_to_write list The backup_restore_vg is used directly for restoring the VG from backup. It's also used to do the VG conversions from one metadata format to another which means vgconvert calls backup_restore_vg too. When restoring VG from backup, we need to rewrite/write PV headers as PVs may have been orphans before and now they're becoming part of some VG - we need to write the PV_EXT_USED flag at least. When using the backup_restore_vg for vgconvert, we need to write completely new PV header in different format. Avoid the special "pv_write" call and handling that was used before this patch in vgconvert (vgconvert_single function to be more precise) and reuse existing internal interface to register PV header for writing (or rewriting) via vg->pvs_to_write list instead like we do it elsewhere in the code. This patch also resolves a problem in which PV headers with target format were written in the vgconvert_single fn as orphans and VG metadata were added later on - this was a tiny hack actually. We can't do this now - we need to write the PV as belonging to a VG because otherwise the PV_EXT_USED flag won't be written properly (if the PV header is written as orphan, the PV_EXT_USED is set to 0, of course, even though metadata are attached later). So this patch removes this tiny inconsistency which was passing just fine before because we didn't have any relation to the VG in PV header before. Now we have the PV_EXT_USED flag which says the "PV is used in some VG".
2015-03-26 16:20:46 +03:00
return 1;
if (!(pv->fmt->features & FMT_PV_FLAGS))
return 0;
if (!pv->dev) {
log_error("Failed to find device for PV.");
return -1;
}
if (!(info = lvmcache_info_from_pvid(pv->dev->pvid, pv->dev, 0))) {
backup: backup_restore_vg: register PVs that need writing via vg->pvs_to_write list The backup_restore_vg is used directly for restoring the VG from backup. It's also used to do the VG conversions from one metadata format to another which means vgconvert calls backup_restore_vg too. When restoring VG from backup, we need to rewrite/write PV headers as PVs may have been orphans before and now they're becoming part of some VG - we need to write the PV_EXT_USED flag at least. When using the backup_restore_vg for vgconvert, we need to write completely new PV header in different format. Avoid the special "pv_write" call and handling that was used before this patch in vgconvert (vgconvert_single function to be more precise) and reuse existing internal interface to register PV header for writing (or rewriting) via vg->pvs_to_write list instead like we do it elsewhere in the code. This patch also resolves a problem in which PV headers with target format were written in the vgconvert_single fn as orphans and VG metadata were added later on - this was a tiny hack actually. We can't do this now - we need to write the PV as belonging to a VG because otherwise the PV_EXT_USED flag won't be written properly (if the PV header is written as orphan, the PV_EXT_USED is set to 0, of course, even though metadata are attached later). So this patch removes this tiny inconsistency which was passing just fine before because we didn't have any relation to the VG in PV header before. Now we have the PV_EXT_USED flag which says the "PV is used in some VG".
2015-03-26 16:20:46 +03:00
log_error("Failed to find cached info for PV %s.", pv_dev_name(pv));
return -1;
}
/*
* We're restoring a VG and if the PV_EXT_USED
* flag is not set yet in PV, we need to set it now!
* This may happen if we have plain PVs without a VG
* and we're restoring former VG from backup on top
* of these PVs.
*/
if (!(lvmcache_ext_flags(info) & PV_EXT_USED))
return 1;
return 0;
}
2002-11-18 17:04:08 +03:00
/* ORPHAN and VG locks held before calling this */
backup: backup_restore_vg: register PVs that need writing via vg->pvs_to_write list The backup_restore_vg is used directly for restoring the VG from backup. It's also used to do the VG conversions from one metadata format to another which means vgconvert calls backup_restore_vg too. When restoring VG from backup, we need to rewrite/write PV headers as PVs may have been orphans before and now they're becoming part of some VG - we need to write the PV_EXT_USED flag at least. When using the backup_restore_vg for vgconvert, we need to write completely new PV header in different format. Avoid the special "pv_write" call and handling that was used before this patch in vgconvert (vgconvert_single function to be more precise) and reuse existing internal interface to register PV header for writing (or rewriting) via vg->pvs_to_write list instead like we do it elsewhere in the code. This patch also resolves a problem in which PV headers with target format were written in the vgconvert_single fn as orphans and VG metadata were added later on - this was a tiny hack actually. We can't do this now - we need to write the PV as belonging to a VG because otherwise the PV_EXT_USED flag won't be written properly (if the PV header is written as orphan, the PV_EXT_USED is set to 0, of course, even though metadata are attached later). So this patch removes this tiny inconsistency which was passing just fine before because we didn't have any relation to the VG in PV header before. Now we have the PV_EXT_USED flag which says the "PV is used in some VG".
2015-03-26 16:20:46 +03:00
int backup_restore_vg(struct cmd_context *cmd, struct volume_group *vg,
int do_pvcreate, struct pv_create_args *pva)
2002-01-10 17:27:47 +03:00
{
backup: backup_restore_vg: register PVs that need writing via vg->pvs_to_write list The backup_restore_vg is used directly for restoring the VG from backup. It's also used to do the VG conversions from one metadata format to another which means vgconvert calls backup_restore_vg too. When restoring VG from backup, we need to rewrite/write PV headers as PVs may have been orphans before and now they're becoming part of some VG - we need to write the PV_EXT_USED flag at least. When using the backup_restore_vg for vgconvert, we need to write completely new PV header in different format. Avoid the special "pv_write" call and handling that was used before this patch in vgconvert (vgconvert_single function to be more precise) and reuse existing internal interface to register PV header for writing (or rewriting) via vg->pvs_to_write list instead like we do it elsewhere in the code. This patch also resolves a problem in which PV headers with target format were written in the vgconvert_single fn as orphans and VG metadata were added later on - this was a tiny hack actually. We can't do this now - we need to write the PV as belonging to a VG because otherwise the PV_EXT_USED flag won't be written properly (if the PV header is written as orphan, the PV_EXT_USED is set to 0, of course, even though metadata are attached later). So this patch removes this tiny inconsistency which was passing just fine before because we didn't have any relation to the VG in PV header before. Now we have the PV_EXT_USED flag which says the "PV is used in some VG".
2015-03-26 16:20:46 +03:00
struct dm_list new_pvs;
struct pv_list *pvl, *new_pvl;
struct physical_volume *existing_pv, *pv;
2016-02-23 14:12:29 +03:00
struct dm_list *pvs = &vg->pvs;
struct format_instance *fid;
struct format_instance_ctx fic;
backup: backup_restore_vg: register PVs that need writing via vg->pvs_to_write list The backup_restore_vg is used directly for restoring the VG from backup. It's also used to do the VG conversions from one metadata format to another which means vgconvert calls backup_restore_vg too. When restoring VG from backup, we need to rewrite/write PV headers as PVs may have been orphans before and now they're becoming part of some VG - we need to write the PV_EXT_USED flag at least. When using the backup_restore_vg for vgconvert, we need to write completely new PV header in different format. Avoid the special "pv_write" call and handling that was used before this patch in vgconvert (vgconvert_single function to be more precise) and reuse existing internal interface to register PV header for writing (or rewriting) via vg->pvs_to_write list instead like we do it elsewhere in the code. This patch also resolves a problem in which PV headers with target format were written in the vgconvert_single fn as orphans and VG metadata were added later on - this was a tiny hack actually. We can't do this now - we need to write the PV as belonging to a VG because otherwise the PV_EXT_USED flag won't be written properly (if the PV header is written as orphan, the PV_EXT_USED is set to 0, of course, even though metadata are attached later). So this patch removes this tiny inconsistency which was passing just fine before because we didn't have any relation to the VG in PV header before. Now we have the PV_EXT_USED flag which says the "PV is used in some VG".
2015-03-26 16:20:46 +03:00
int should_write_pv;
uint32_t tmp_extent_size;
2002-01-10 17:27:47 +03:00
/*
2002-11-18 17:04:08 +03:00
* FIXME: Check that the PVs referenced in the backup are
* not members of other existing VGs.
2002-01-10 17:27:47 +03:00
*/
2002-11-18 17:04:08 +03:00
backup: backup_restore_vg: register PVs that need writing via vg->pvs_to_write list The backup_restore_vg is used directly for restoring the VG from backup. It's also used to do the VG conversions from one metadata format to another which means vgconvert calls backup_restore_vg too. When restoring VG from backup, we need to rewrite/write PV headers as PVs may have been orphans before and now they're becoming part of some VG - we need to write the PV_EXT_USED flag at least. When using the backup_restore_vg for vgconvert, we need to write completely new PV header in different format. Avoid the special "pv_write" call and handling that was used before this patch in vgconvert (vgconvert_single function to be more precise) and reuse existing internal interface to register PV header for writing (or rewriting) via vg->pvs_to_write list instead like we do it elsewhere in the code. This patch also resolves a problem in which PV headers with target format were written in the vgconvert_single fn as orphans and VG metadata were added later on - this was a tiny hack actually. We can't do this now - we need to write the PV as belonging to a VG because otherwise the PV_EXT_USED flag won't be written properly (if the PV header is written as orphan, the PV_EXT_USED is set to 0, of course, even though metadata are attached later). So this patch removes this tiny inconsistency which was passing just fine before because we didn't have any relation to the VG in PV header before. Now we have the PV_EXT_USED flag which says the "PV is used in some VG".
2015-03-26 16:20:46 +03:00
/* Prepare new PVs if needed. */
if (do_pvcreate) {
backup: backup_restore_vg: register PVs that need writing via vg->pvs_to_write list The backup_restore_vg is used directly for restoring the VG from backup. It's also used to do the VG conversions from one metadata format to another which means vgconvert calls backup_restore_vg too. When restoring VG from backup, we need to rewrite/write PV headers as PVs may have been orphans before and now they're becoming part of some VG - we need to write the PV_EXT_USED flag at least. When using the backup_restore_vg for vgconvert, we need to write completely new PV header in different format. Avoid the special "pv_write" call and handling that was used before this patch in vgconvert (vgconvert_single function to be more precise) and reuse existing internal interface to register PV header for writing (or rewriting) via vg->pvs_to_write list instead like we do it elsewhere in the code. This patch also resolves a problem in which PV headers with target format were written in the vgconvert_single fn as orphans and VG metadata were added later on - this was a tiny hack actually. We can't do this now - we need to write the PV as belonging to a VG because otherwise the PV_EXT_USED flag won't be written properly (if the PV header is written as orphan, the PV_EXT_USED is set to 0, of course, even though metadata are attached later). So this patch removes this tiny inconsistency which was passing just fine before because we didn't have any relation to the VG in PV header before. Now we have the PV_EXT_USED flag which says the "PV is used in some VG".
2015-03-26 16:20:46 +03:00
dm_list_init(&new_pvs);
dm_list_iterate_items(pvl, &vg->pvs) {
existing_pv = pvl->pv;
pva->id = existing_pv->id;
pva->idp = &pva->id;
pva->pe_start = pv_pe_start(existing_pv);
pva->extent_count = pv_pe_count(existing_pv);
pva->extent_size = pv_pe_size(existing_pv);
backup: backup_restore_vg: register PVs that need writing via vg->pvs_to_write list The backup_restore_vg is used directly for restoring the VG from backup. It's also used to do the VG conversions from one metadata format to another which means vgconvert calls backup_restore_vg too. When restoring VG from backup, we need to rewrite/write PV headers as PVs may have been orphans before and now they're becoming part of some VG - we need to write the PV_EXT_USED flag at least. When using the backup_restore_vg for vgconvert, we need to write completely new PV header in different format. Avoid the special "pv_write" call and handling that was used before this patch in vgconvert (vgconvert_single function to be more precise) and reuse existing internal interface to register PV header for writing (or rewriting) via vg->pvs_to_write list instead like we do it elsewhere in the code. This patch also resolves a problem in which PV headers with target format were written in the vgconvert_single fn as orphans and VG metadata were added later on - this was a tiny hack actually. We can't do this now - we need to write the PV as belonging to a VG because otherwise the PV_EXT_USED flag won't be written properly (if the PV header is written as orphan, the PV_EXT_USED is set to 0, of course, even though metadata are attached later). So this patch removes this tiny inconsistency which was passing just fine before because we didn't have any relation to the VG in PV header before. Now we have the PV_EXT_USED flag which says the "PV is used in some VG".
2015-03-26 16:20:46 +03:00
/* pe_end = pv_pe_count(existing_pv) * pv_pe_size(existing_pv) + pe_start - 1 */
if (!(pv = pv_create(cmd, pv_dev(existing_pv), pva))) {
backup: backup_restore_vg: register PVs that need writing via vg->pvs_to_write list The backup_restore_vg is used directly for restoring the VG from backup. It's also used to do the VG conversions from one metadata format to another which means vgconvert calls backup_restore_vg too. When restoring VG from backup, we need to rewrite/write PV headers as PVs may have been orphans before and now they're becoming part of some VG - we need to write the PV_EXT_USED flag at least. When using the backup_restore_vg for vgconvert, we need to write completely new PV header in different format. Avoid the special "pv_write" call and handling that was used before this patch in vgconvert (vgconvert_single function to be more precise) and reuse existing internal interface to register PV header for writing (or rewriting) via vg->pvs_to_write list instead like we do it elsewhere in the code. This patch also resolves a problem in which PV headers with target format were written in the vgconvert_single fn as orphans and VG metadata were added later on - this was a tiny hack actually. We can't do this now - we need to write the PV as belonging to a VG because otherwise the PV_EXT_USED flag won't be written properly (if the PV header is written as orphan, the PV_EXT_USED is set to 0, of course, even though metadata are attached later). So this patch removes this tiny inconsistency which was passing just fine before because we didn't have any relation to the VG in PV header before. Now we have the PV_EXT_USED flag which says the "PV is used in some VG".
2015-03-26 16:20:46 +03:00
log_error("Failed to setup physical volume \"%s\".",
pv_dev_name(existing_pv));
return 0;
}
pv->vg_name = vg->name;
pv->vgid = vg->id;
if (!(new_pvl = dm_pool_zalloc(vg->vgmem, sizeof(*new_pvl)))) {
log_error("Failed to allocate PV list item for \"%s\".",
pv_dev_name(pvl->pv));
return 0;
}
backup: backup_restore_vg: register PVs that need writing via vg->pvs_to_write list The backup_restore_vg is used directly for restoring the VG from backup. It's also used to do the VG conversions from one metadata format to another which means vgconvert calls backup_restore_vg too. When restoring VG from backup, we need to rewrite/write PV headers as PVs may have been orphans before and now they're becoming part of some VG - we need to write the PV_EXT_USED flag at least. When using the backup_restore_vg for vgconvert, we need to write completely new PV header in different format. Avoid the special "pv_write" call and handling that was used before this patch in vgconvert (vgconvert_single function to be more precise) and reuse existing internal interface to register PV header for writing (or rewriting) via vg->pvs_to_write list instead like we do it elsewhere in the code. This patch also resolves a problem in which PV headers with target format were written in the vgconvert_single fn as orphans and VG metadata were added later on - this was a tiny hack actually. We can't do this now - we need to write the PV as belonging to a VG because otherwise the PV_EXT_USED flag won't be written properly (if the PV header is written as orphan, the PV_EXT_USED is set to 0, of course, even though metadata are attached later). So this patch removes this tiny inconsistency which was passing just fine before because we didn't have any relation to the VG in PV header before. Now we have the PV_EXT_USED flag which says the "PV is used in some VG".
2015-03-26 16:20:46 +03:00
new_pvl->pv = pv;
dm_list_add(&new_pvs, &new_pvl->list);
log_verbose("Set up physical volume for \"%s\" with %" PRIu64
" available sectors.", pv_dev_name(pv), pv_size(pv));
}
pvs = &new_pvs;
}
2002-11-18 17:04:08 +03:00
/* Attempt to write out using currently active format */
2012-02-13 03:01:19 +04:00
fic.type = FMT_INSTANCE_AUX_MDAS;
fic.context.vg_ref.vg_name = vg->name;
fic.context.vg_ref.vg_id = NULL;
if (!(fid = cmd->fmt->ops->create_instance(cmd->fmt, &fic))) {
backup: backup_restore_vg: register PVs that need writing via vg->pvs_to_write list The backup_restore_vg is used directly for restoring the VG from backup. It's also used to do the VG conversions from one metadata format to another which means vgconvert calls backup_restore_vg too. When restoring VG from backup, we need to rewrite/write PV headers as PVs may have been orphans before and now they're becoming part of some VG - we need to write the PV_EXT_USED flag at least. When using the backup_restore_vg for vgconvert, we need to write completely new PV header in different format. Avoid the special "pv_write" call and handling that was used before this patch in vgconvert (vgconvert_single function to be more precise) and reuse existing internal interface to register PV header for writing (or rewriting) via vg->pvs_to_write list instead like we do it elsewhere in the code. This patch also resolves a problem in which PV headers with target format were written in the vgconvert_single fn as orphans and VG metadata were added later on - this was a tiny hack actually. We can't do this now - we need to write the PV as belonging to a VG because otherwise the PV_EXT_USED flag won't be written properly (if the PV header is written as orphan, the PV_EXT_USED is set to 0, of course, even though metadata are attached later). So this patch removes this tiny inconsistency which was passing just fine before because we didn't have any relation to the VG in PV header before. Now we have the PV_EXT_USED flag which says the "PV is used in some VG".
2015-03-26 16:20:46 +03:00
log_error("Failed to allocate format instance.");
2002-01-10 17:27:47 +03:00
return 0;
}
backup: backup_restore_vg: register PVs that need writing via vg->pvs_to_write list The backup_restore_vg is used directly for restoring the VG from backup. It's also used to do the VG conversions from one metadata format to another which means vgconvert calls backup_restore_vg too. When restoring VG from backup, we need to rewrite/write PV headers as PVs may have been orphans before and now they're becoming part of some VG - we need to write the PV_EXT_USED flag at least. When using the backup_restore_vg for vgconvert, we need to write completely new PV header in different format. Avoid the special "pv_write" call and handling that was used before this patch in vgconvert (vgconvert_single function to be more precise) and reuse existing internal interface to register PV header for writing (or rewriting) via vg->pvs_to_write list instead like we do it elsewhere in the code. This patch also resolves a problem in which PV headers with target format were written in the vgconvert_single fn as orphans and VG metadata were added later on - this was a tiny hack actually. We can't do this now - we need to write the PV as belonging to a VG because otherwise the PV_EXT_USED flag won't be written properly (if the PV header is written as orphan, the PV_EXT_USED is set to 0, of course, even though metadata are attached later). So this patch removes this tiny inconsistency which was passing just fine before because we didn't have any relation to the VG in PV header before. Now we have the PV_EXT_USED flag which says the "PV is used in some VG".
2015-03-26 16:20:46 +03:00
if (do_pvcreate) {
backup: backup_restore_vg: register PVs that need writing via vg->pvs_to_write list The backup_restore_vg is used directly for restoring the VG from backup. It's also used to do the VG conversions from one metadata format to another which means vgconvert calls backup_restore_vg too. When restoring VG from backup, we need to rewrite/write PV headers as PVs may have been orphans before and now they're becoming part of some VG - we need to write the PV_EXT_USED flag at least. When using the backup_restore_vg for vgconvert, we need to write completely new PV header in different format. Avoid the special "pv_write" call and handling that was used before this patch in vgconvert (vgconvert_single function to be more precise) and reuse existing internal interface to register PV header for writing (or rewriting) via vg->pvs_to_write list instead like we do it elsewhere in the code. This patch also resolves a problem in which PV headers with target format were written in the vgconvert_single fn as orphans and VG metadata were added later on - this was a tiny hack actually. We can't do this now - we need to write the PV as belonging to a VG because otherwise the PV_EXT_USED flag won't be written properly (if the PV header is written as orphan, the PV_EXT_USED is set to 0, of course, even though metadata are attached later). So this patch removes this tiny inconsistency which was passing just fine before because we didn't have any relation to the VG in PV header before. Now we have the PV_EXT_USED flag which says the "PV is used in some VG".
2015-03-26 16:20:46 +03:00
log_verbose("Deleting existing metadata for VG %s.", vg->name);
if (!vg_remove_mdas(vg)) {
cmd->fmt->ops->destroy_instance(fid);
log_error("Removal of existing metadata for VG %s failed.", vg->name);
return 0;
}
}
vg_set_fid(vg, fid);
2002-01-10 17:27:47 +03:00
/*
* Setting vg->old_name to a blank value will explicitly
* disable any attempt to check VG name in existing metadata.
*/
if (!(vg->old_name = dm_pool_strdup(vg->vgmem, ""))) {
log_error("Failed to duplicate empty name.");
return 0;
}
2002-11-18 17:04:08 +03:00
/* Add any metadata areas on the PVs */
backup: backup_restore_vg: register PVs that need writing via vg->pvs_to_write list The backup_restore_vg is used directly for restoring the VG from backup. It's also used to do the VG conversions from one metadata format to another which means vgconvert calls backup_restore_vg too. When restoring VG from backup, we need to rewrite/write PV headers as PVs may have been orphans before and now they're becoming part of some VG - we need to write the PV_EXT_USED flag at least. When using the backup_restore_vg for vgconvert, we need to write completely new PV header in different format. Avoid the special "pv_write" call and handling that was used before this patch in vgconvert (vgconvert_single function to be more precise) and reuse existing internal interface to register PV header for writing (or rewriting) via vg->pvs_to_write list instead like we do it elsewhere in the code. This patch also resolves a problem in which PV headers with target format were written in the vgconvert_single fn as orphans and VG metadata were added later on - this was a tiny hack actually. We can't do this now - we need to write the PV as belonging to a VG because otherwise the PV_EXT_USED flag won't be written properly (if the PV header is written as orphan, the PV_EXT_USED is set to 0, of course, even though metadata are attached later). So this patch removes this tiny inconsistency which was passing just fine before because we didn't have any relation to the VG in PV header before. Now we have the PV_EXT_USED flag which says the "PV is used in some VG".
2015-03-26 16:20:46 +03:00
dm_list_iterate_items(pvl, pvs) {
if ((should_write_pv = _restore_vg_should_write_pv(pvl->pv, do_pvcreate)) < 0)
backup: backup_restore_vg: register PVs that need writing via vg->pvs_to_write list The backup_restore_vg is used directly for restoring the VG from backup. It's also used to do the VG conversions from one metadata format to another which means vgconvert calls backup_restore_vg too. When restoring VG from backup, we need to rewrite/write PV headers as PVs may have been orphans before and now they're becoming part of some VG - we need to write the PV_EXT_USED flag at least. When using the backup_restore_vg for vgconvert, we need to write completely new PV header in different format. Avoid the special "pv_write" call and handling that was used before this patch in vgconvert (vgconvert_single function to be more precise) and reuse existing internal interface to register PV header for writing (or rewriting) via vg->pvs_to_write list instead like we do it elsewhere in the code. This patch also resolves a problem in which PV headers with target format were written in the vgconvert_single fn as orphans and VG metadata were added later on - this was a tiny hack actually. We can't do this now - we need to write the PV as belonging to a VG because otherwise the PV_EXT_USED flag won't be written properly (if the PV header is written as orphan, the PV_EXT_USED is set to 0, of course, even though metadata are attached later). So this patch removes this tiny inconsistency which was passing just fine before because we didn't have any relation to the VG in PV header before. Now we have the PV_EXT_USED flag which says the "PV is used in some VG".
2015-03-26 16:20:46 +03:00
return_0;
if (should_write_pv) {
if (!(new_pvl = dm_pool_zalloc(vg->vgmem, sizeof(*new_pvl)))) {
backup: backup_restore_vg: register PVs that need writing via vg->pvs_to_write list The backup_restore_vg is used directly for restoring the VG from backup. It's also used to do the VG conversions from one metadata format to another which means vgconvert calls backup_restore_vg too. When restoring VG from backup, we need to rewrite/write PV headers as PVs may have been orphans before and now they're becoming part of some VG - we need to write the PV_EXT_USED flag at least. When using the backup_restore_vg for vgconvert, we need to write completely new PV header in different format. Avoid the special "pv_write" call and handling that was used before this patch in vgconvert (vgconvert_single function to be more precise) and reuse existing internal interface to register PV header for writing (or rewriting) via vg->pvs_to_write list instead like we do it elsewhere in the code. This patch also resolves a problem in which PV headers with target format were written in the vgconvert_single fn as orphans and VG metadata were added later on - this was a tiny hack actually. We can't do this now - we need to write the PV as belonging to a VG because otherwise the PV_EXT_USED flag won't be written properly (if the PV header is written as orphan, the PV_EXT_USED is set to 0, of course, even though metadata are attached later). So this patch removes this tiny inconsistency which was passing just fine before because we didn't have any relation to the VG in PV header before. Now we have the PV_EXT_USED flag which says the "PV is used in some VG".
2015-03-26 16:20:46 +03:00
log_error("Failed to allocate structure for scheduled "
"writing of PV '%s'.", pv_dev_name(pvl->pv));
return 0;
}
backup: backup_restore_vg: register PVs that need writing via vg->pvs_to_write list The backup_restore_vg is used directly for restoring the VG from backup. It's also used to do the VG conversions from one metadata format to another which means vgconvert calls backup_restore_vg too. When restoring VG from backup, we need to rewrite/write PV headers as PVs may have been orphans before and now they're becoming part of some VG - we need to write the PV_EXT_USED flag at least. When using the backup_restore_vg for vgconvert, we need to write completely new PV header in different format. Avoid the special "pv_write" call and handling that was used before this patch in vgconvert (vgconvert_single function to be more precise) and reuse existing internal interface to register PV header for writing (or rewriting) via vg->pvs_to_write list instead like we do it elsewhere in the code. This patch also resolves a problem in which PV headers with target format were written in the vgconvert_single fn as orphans and VG metadata were added later on - this was a tiny hack actually. We can't do this now - we need to write the PV as belonging to a VG because otherwise the PV_EXT_USED flag won't be written properly (if the PV header is written as orphan, the PV_EXT_USED is set to 0, of course, even though metadata are attached later). So this patch removes this tiny inconsistency which was passing just fine before because we didn't have any relation to the VG in PV header before. Now we have the PV_EXT_USED flag which says the "PV is used in some VG".
2015-03-26 16:20:46 +03:00
new_pvl->pv = pvl->pv;
dm_list_add(&vg->pv_write_list, &new_pvl->list);
}
backup: backup_restore_vg: register PVs that need writing via vg->pvs_to_write list The backup_restore_vg is used directly for restoring the VG from backup. It's also used to do the VG conversions from one metadata format to another which means vgconvert calls backup_restore_vg too. When restoring VG from backup, we need to rewrite/write PV headers as PVs may have been orphans before and now they're becoming part of some VG - we need to write the PV_EXT_USED flag at least. When using the backup_restore_vg for vgconvert, we need to write completely new PV header in different format. Avoid the special "pv_write" call and handling that was used before this patch in vgconvert (vgconvert_single function to be more precise) and reuse existing internal interface to register PV header for writing (or rewriting) via vg->pvs_to_write list instead like we do it elsewhere in the code. This patch also resolves a problem in which PV headers with target format were written in the vgconvert_single fn as orphans and VG metadata were added later on - this was a tiny hack actually. We can't do this now - we need to write the PV as belonging to a VG because otherwise the PV_EXT_USED flag won't be written properly (if the PV header is written as orphan, the PV_EXT_USED is set to 0, of course, even though metadata are attached later). So this patch removes this tiny inconsistency which was passing just fine before because we didn't have any relation to the VG in PV header before. Now we have the PV_EXT_USED flag which says the "PV is used in some VG".
2015-03-26 16:20:46 +03:00
/* Add any metadata areas on the PV. */
tmp_extent_size = vg->extent_size;
vg->extent_size = 0;
if (!vg->fid->fmt->ops->pv_setup(vg->fid->fmt, pvl->pv, vg)) {
backup: backup_restore_vg: register PVs that need writing via vg->pvs_to_write list The backup_restore_vg is used directly for restoring the VG from backup. It's also used to do the VG conversions from one metadata format to another which means vgconvert calls backup_restore_vg too. When restoring VG from backup, we need to rewrite/write PV headers as PVs may have been orphans before and now they're becoming part of some VG - we need to write the PV_EXT_USED flag at least. When using the backup_restore_vg for vgconvert, we need to write completely new PV header in different format. Avoid the special "pv_write" call and handling that was used before this patch in vgconvert (vgconvert_single function to be more precise) and reuse existing internal interface to register PV header for writing (or rewriting) via vg->pvs_to_write list instead like we do it elsewhere in the code. This patch also resolves a problem in which PV headers with target format were written in the vgconvert_single fn as orphans and VG metadata were added later on - this was a tiny hack actually. We can't do this now - we need to write the PV as belonging to a VG because otherwise the PV_EXT_USED flag won't be written properly (if the PV header is written as orphan, the PV_EXT_USED is set to 0, of course, even though metadata are attached later). So this patch removes this tiny inconsistency which was passing just fine before because we didn't have any relation to the VG in PV header before. Now we have the PV_EXT_USED flag which says the "PV is used in some VG".
2015-03-26 16:20:46 +03:00
vg->extent_size = tmp_extent_size;
log_error("Format-specific setup for %s failed.",
pv_dev_name(pvl->pv));
2002-11-18 17:04:08 +03:00
return 0;
}
backup: backup_restore_vg: register PVs that need writing via vg->pvs_to_write list The backup_restore_vg is used directly for restoring the VG from backup. It's also used to do the VG conversions from one metadata format to another which means vgconvert calls backup_restore_vg too. When restoring VG from backup, we need to rewrite/write PV headers as PVs may have been orphans before and now they're becoming part of some VG - we need to write the PV_EXT_USED flag at least. When using the backup_restore_vg for vgconvert, we need to write completely new PV header in different format. Avoid the special "pv_write" call and handling that was used before this patch in vgconvert (vgconvert_single function to be more precise) and reuse existing internal interface to register PV header for writing (or rewriting) via vg->pvs_to_write list instead like we do it elsewhere in the code. This patch also resolves a problem in which PV headers with target format were written in the vgconvert_single fn as orphans and VG metadata were added later on - this was a tiny hack actually. We can't do this now - we need to write the PV as belonging to a VG because otherwise the PV_EXT_USED flag won't be written properly (if the PV header is written as orphan, the PV_EXT_USED is set to 0, of course, even though metadata are attached later). So this patch removes this tiny inconsistency which was passing just fine before because we didn't have any relation to the VG in PV header before. Now we have the PV_EXT_USED flag which says the "PV is used in some VG".
2015-03-26 16:20:46 +03:00
vg->extent_size = tmp_extent_size;
}
if (do_pvcreate) {
dm_list_iterate_items(pvl, &vg->pv_write_list) {
struct device *dev = pv_dev(pvl->pv);
const char *pv_name = dev_name(dev);
if (!label_remove(dev)) {
log_error("Failed to wipe existing label on %s", pv_name);
return 0;
}
log_verbose("Zeroing start of device %s", pv_name);
if (!dev_open_quiet(dev)) {
log_error("%s not opened: device not zeroed", pv_name);
return 0;
}
if (!dev_set(dev, UINT64_C(0), (size_t) 2048, 0)) {
log_error("%s not wiped: aborting", pv_name);
if (!dev_close(dev))
stack;
return 0;
}
if (!dev_close(dev))
stack;
}
}
if (!vg_write(vg))
return_0;
if (!vg_commit(vg))
2008-01-30 16:19:47 +03:00
return_0;
2002-01-10 17:27:47 +03:00
return 1;
}
2002-11-18 17:04:08 +03:00
/* ORPHAN and VG locks held before calling this */
int backup_restore_from_file(struct cmd_context *cmd, const char *vg_name,
const char *file, int force)
2002-11-18 17:04:08 +03:00
{
struct volume_group *vg;
int missing_pvs, r = 0;
const struct lv_list *lvl;
2002-11-18 17:04:08 +03:00
/*
* Read in the volume group from the text file.
*/
2008-01-30 16:19:47 +03:00
if (!(vg = backup_read_vg(cmd, vg_name, file)))
return_0;
2002-11-18 17:04:08 +03:00
/* FIXME: Restore support is missing for now */
dm_list_iterate_items(lvl, &vg->lvs) {
if (lv_is_thin_type(lvl->lv)) {
if (!force) {
log_error("Consider using option --force to restore "
"Volume Group %s with thin volumes.",
vg->name);
goto out;
} else {
log_warn("WARNING: Forced restore of Volume Group "
"%s with thin volumes.", vg->name);
break;
}
}
}
missing_pvs = vg_missing_pv_count(vg);
if (missing_pvs == 0)
r = backup_restore_vg(cmd, vg, 0, NULL);
else
log_error("Cannot restore Volume Group %s with %i PVs "
"marked as missing.", vg->name, missing_pvs);
out:
release_vg(vg);
return r;
2002-11-18 17:04:08 +03:00
}
int backup_restore(struct cmd_context *cmd, const char *vg_name, int force)
2002-01-10 17:27:47 +03:00
{
char path[PATH_MAX];
2006-08-21 16:54:53 +04:00
if (dm_snprintf(path, sizeof(path), "%s/%s",
cmd->backup_params->dir, vg_name) < 0) {
log_error("Failed to generate backup filename (for restore).");
2002-01-10 17:27:47 +03:00
return 0;
}
return backup_restore_from_file(cmd, vg_name, path, force);
2002-01-10 17:27:47 +03:00
}
2002-11-18 17:04:08 +03:00
int backup_to_file(const char *file, const char *desc, struct volume_group *vg)
{
int r = 0;
2002-11-18 17:04:08 +03:00
struct format_instance *tf;
struct format_instance_ctx fic;
struct text_context tc = {.path_live = file,
.path_edit = NULL,
.desc = desc};
2002-11-18 17:04:08 +03:00
struct metadata_area *mda;
struct cmd_context *cmd;
cmd = vg->cmd;
log_verbose("Creating volume group backup \"%s\" (seqno %u).", file, vg->seqno);
2012-02-13 03:01:19 +04:00
fic.type = FMT_INSTANCE_PRIVATE_MDAS;
fic.context.private = &tc;
if (!(tf = cmd->fmt_backup->ops->create_instance(cmd->fmt_backup, &fic))) {
2002-11-18 17:04:08 +03:00
log_error("Couldn't create backup object.");
return 0;
}
if (dm_list_empty(&tf->metadata_areas_in_use)) {
log_error(INTERNAL_ERROR "No in use metadata areas to write.");
tf->fmt->ops->destroy_instance(tf);
return 0;
}
2002-11-18 17:04:08 +03:00
/* Write and commit the metadata area */
dm_list_iterate_items(mda, &tf->metadata_areas_in_use) {
2002-11-18 17:04:08 +03:00
if (!(r = mda->ops->vg_write(tf, vg, mda))) {
stack;
continue;
}
if (mda->ops->vg_commit &&
!(r = mda->ops->vg_commit(tf, vg, mda))) {
stack;
}
}
tf->fmt->ops->destroy_instance(tf);
return r;
}
/*
* Update backup (and archive) if they're out-of-date or don't exist.
*
* This function is not supposed to log_error
* when the filesystem with archive/backup dir is read-only.
*/
void check_current_backup(struct volume_group *vg)
{
char path[PATH_MAX];
struct volume_group *vg_backup;
int old_suppress;
if (!vg->cmd->backup_params->enabled || !vg->cmd->backup_params->dir) {
log_debug("Skipping check for current backup, since backup is disabled.");
return;
}
if (vg_is_exported(vg))
return;
2006-08-21 16:54:53 +04:00
if (dm_snprintf(path, sizeof(path), "%s/%s",
vg->cmd->backup_params->dir, vg->name) < 0) {
log_warn("WARNING: Failed to generate backup pathname %s/%s.",
vg->cmd->backup_params->dir, vg->name);
return;
}
old_suppress = log_suppress(1);
/* Up-to-date backup exists? */
if ((vg_backup = backup_read_vg(vg->cmd, vg->name, path)) &&
(vg->seqno == vg_backup->seqno) &&
(id_equal(&vg->id, &vg_backup->id))) {
log_suppress(old_suppress);
release_vg(vg_backup);
return;
}
log_suppress(old_suppress);
if (vg_backup) {
if (!_archive(vg_backup, 0))
stack;
release_vg(vg_backup);
}
if (!_archive(vg, 0))
stack;
if (!backup_locally(vg))
stack;
}