1
0
mirror of git://sourceware.org/git/lvm2.git synced 2024-10-27 18:55:19 +03:00
lvm2/tools/pvchange.c
David Teigland a7c45ddc59 lvmetad: two phase vg_update
Previously, a command sent lvmetad new VG metadata in vg_commit().
In vg_commit(), devices are suspended, so any memory allocation
done by the command while sending to lvmetad, or by lvmetad while
updating its cache could deadlock if memory reclaim was triggered.

Now lvmetad is updated in unlock_vg(), after devices are resumed.
The new method for updating VG metadata in lvmetad is in two phases:

1. In vg_write(), before devices are suspended, the command sends
   lvmetad a short message ("set_vg_info") telling it what the new
   VG seqno will be.  lvmetad sees that the seqno is newer than
   the seqno of its cached VG, so it sets the INVALID flag for the
   cached VG.  If sending the message to lvmetad fails, the command
   fails before the metadata is committed and the change is not made.
   If sending the message succeeds, vg_commit() is called.

2. In unlock_vg(), after devices are resumed, the command sends
   lvmetad the standard vg_update message with the new metadata.
   lvmetad sees that the seqno in the new metadata matches the
   seqno it saved from set_vg_info, and knows it has the latest
   copy, so it clears the INVALID flag for the cached VG.

If a command fails between 1 and 2 (after committing the VG on disk,
but before sending lvmetad the new metadata), the cached VG retains
the INVALID flag in lvmetad.  A subsequent command will read the
cached VG from lvmetad, see the INVALID flag, ignore the cached
copy, read the VG from disk instead, update the lvmetad copy
with the latest copy from disk, (this clears the INVALID flag
in lvmetad), and use the correct VG metadata for the command.

(This INVALID mechanism already existed for use by lvmlockd.)
2016-06-28 02:30:31 +01:00

268 lines
7.4 KiB
C

/*
* Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.
* Copyright (C) 2004-2015 Red Hat, Inc. All rights reserved.
*
* 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.
*
* You should have received a copy of the GNU Lesser General Public License
* along with this program; if not, write to the Free Software Foundation,
* Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "tools.h"
struct pvchange_params {
unsigned done;
unsigned total;
};
static int _pvchange_single(struct cmd_context *cmd, struct volume_group *vg,
struct physical_volume *pv, struct processing_handle *handle)
{
struct pvchange_params *params = (struct pvchange_params *) handle->custom_handle;
const char *pv_name = pv_dev_name(pv);
char uuid[64] __attribute__((aligned(8)));
unsigned done = 0;
int used;
int allocatable = arg_int_value(cmd, allocatable_ARG, 0);
int mda_ignore = arg_int_value(cmd, metadataignore_ARG, 0);
int tagargs = arg_is_set(cmd, addtag_ARG) + arg_is_set(cmd, deltag_ARG);
params->total++;
if (vg && vg_is_exported(vg)) {
log_error("Volume group %s is exported", vg->name);
goto bad;
}
/*
* The primary location of this check is in vg_write(), but it needs
* to be copied here to prevent the pv_write() which is called before
* the vg_write().
*/
if (vg && lvmcache_found_duplicate_pvs() && vg_has_duplicate_pvs(vg)) {
if (!find_config_tree_bool(vg->cmd, devices_allow_changes_with_duplicate_pvs_CFG, NULL)) {
log_error("Cannot update volume group %s with duplicate PV devices.",
vg->name);
goto bad;
}
if (arg_is_set(cmd, uuid_ARG)) {
log_error("Resolve duplicate PV UUIDs with vgimportclone (or filters).");
goto bad;
}
}
/* If in a VG, must change using volume group. */
if (!is_orphan(pv)) {
if (tagargs && !(vg->fid->fmt->features & FMT_TAGS)) {
log_error("Volume group containing %s does not "
"support tags", pv_name);
goto bad;
}
if (arg_is_set(cmd, uuid_ARG) && lvs_in_vg_activated(vg)) {
log_error("Volume group containing %s has active "
"logical volumes", pv_name);
goto bad;
}
if (!archive(vg))
goto_bad;
} else {
if (tagargs) {
log_error("Can't change tag on Physical Volume %s not "
"in volume group", pv_name);
goto bad;
}
if ((used = is_used_pv(pv)) < 0)
goto_bad;
if (used && (arg_count(cmd, force_ARG) != DONT_PROMPT_OVERRIDE)) {
log_error("PV %s is used by a VG but its metadata is missing.", pv_name);
log_error("Can't change PV '%s' without -ff.", pv_name);
goto bad;
}
}
if (arg_is_set(cmd, allocatable_ARG)) {
if (is_orphan(pv) &&
!(pv->fmt->features & FMT_ORPHAN_ALLOCATABLE)) {
log_error("Allocatability not supported by orphan "
"%s format PV %s", pv->fmt->name, pv_name);
goto bad;
}
/* change allocatability for a PV */
if (allocatable && (pv_status(pv) & ALLOCATABLE_PV)) {
log_warn("Physical volume \"%s\" is already "
"allocatable.", pv_name);
} else if (!allocatable && !(pv_status(pv) & ALLOCATABLE_PV)) {
log_warn("Physical volume \"%s\" is already "
"unallocatable.", pv_name);
} else if (allocatable) {
log_verbose("Setting physical volume \"%s\" "
"allocatable", pv_name);
pv->status |= ALLOCATABLE_PV;
done = 1;
} else {
log_verbose("Setting physical volume \"%s\" NOT "
"allocatable", pv_name);
pv->status &= ~ALLOCATABLE_PV;
done = 1;
}
}
/*
* Needed to change a property on an orphan PV.
* i.e. the global lock is only needed for orphans.
* Convert sh to ex.
*/
if (is_orphan(pv)) {
if (!lockd_gl(cmd, "ex", 0))
return_ECMD_FAILED;
cmd->lockd_gl_disable = 1;
}
if (tagargs) {
/* tag or deltag */
if (arg_is_set(cmd, addtag_ARG) && !change_tag(cmd, NULL, NULL, pv, addtag_ARG))
goto_bad;
if (arg_is_set(cmd, deltag_ARG) && !change_tag(cmd, NULL, NULL, pv, deltag_ARG))
goto_bad;
done = 1;
}
if (arg_is_set(cmd, metadataignore_ARG)) {
if ((vg_mda_copies(vg) != VGMETADATACOPIES_UNMANAGED) &&
(arg_count(cmd, force_ARG) == PROMPT) &&
yes_no_prompt("Override preferred number of copies "
"of VG %s metadata? [y/n]: ",
pv_vg_name(pv)) == 'n')
goto_bad;
if (!pv_change_metadataignore(pv, mda_ignore))
goto_bad;
done = 1;
}
if (arg_is_set(cmd, uuid_ARG)) {
/* --uuid: Change PV ID randomly */
memcpy(&pv->old_id, &pv->id, sizeof(pv->id));
if (!id_create(&pv->id)) {
log_error("Failed to generate new random UUID for %s.",
pv_name);
goto bad;
}
if (!id_write_format(&pv->id, uuid, sizeof(uuid)))
goto_bad;
log_verbose("Changing uuid of %s to %s.", pv_name, uuid);
if (!is_orphan(pv) && (!pv_write(cmd, pv, 1))) {
log_error("pv_write with new uuid failed "
"for %s.", pv_name);
goto bad;
}
done = 1;
}
if (!done) {
log_print_unless_silent("Physical volume %s not changed", pv_name);
return ECMD_PROCESSED;
}
log_verbose("Updating physical volume \"%s\"", pv_name);
if (!is_orphan(pv)) {
if (!vg_write(vg) || !vg_commit(vg)) {
log_error("Failed to store physical volume \"%s\" in "
"volume group \"%s\"", pv_name, vg->name);
goto bad;
}
backup(vg);
} else if (!(pv_write(cmd, pv, 0))) {
log_error("Failed to store physical volume \"%s\"",
pv_name);
goto bad;
}
log_print_unless_silent("Physical volume \"%s\" changed", pv_name);
params->done++;
return ECMD_PROCESSED;
bad:
log_error("Physical volume %s not changed", pv_name);
return ECMD_FAILED;
}
int pvchange(struct cmd_context *cmd, int argc, char **argv)
{
struct pvchange_params params = { 0 };
struct processing_handle *handle = NULL;
int ret;
if (!(arg_is_set(cmd, allocatable_ARG) + arg_is_set(cmd, addtag_ARG) +
arg_is_set(cmd, deltag_ARG) + arg_is_set(cmd, uuid_ARG) +
arg_is_set(cmd, metadataignore_ARG))) {
log_error("Please give one or more of -x, -uuid, "
"--addtag, --deltag or --metadataignore");
ret = EINVALID_CMD_LINE;
goto out;
}
if (!(handle = init_processing_handle(cmd, NULL))) {
log_error("Failed to initialize processing handle.");
ret = ECMD_FAILED;
goto out;
}
handle->custom_handle = &params;
if (!(arg_is_set(cmd, all_ARG)) && !argc && !handle->internal_report_for_select) {
log_error("Please give a physical volume path or use --select for selection.");
ret = EINVALID_CMD_LINE;
goto out;
}
if (arg_is_set(cmd, all_ARG) && argc) {
log_error("Option --all and PhysicalVolumePath are exclusive.");
ret = EINVALID_CMD_LINE;
goto out;
}
if (!argc) {
/*
* Take the global lock here so the lvmcache remains
* consistent across orphan/non-orphan vg locks. If we don't
* take the lock here, pvs with 0 mdas in a non-orphan VG will
* be processed twice.
*/
if (!lock_vol(cmd, VG_GLOBAL, LCK_VG_WRITE, NULL)) {
log_error("Unable to obtain global lock.");
ret = ECMD_FAILED;
goto out;
}
}
set_pv_notify(cmd);
ret = process_each_pv(cmd, argc, argv, NULL, 0, READ_FOR_UPDATE | READ_ALLOW_EXPORTED, handle, _pvchange_single);
if (!argc)
unlock_vg(cmd, NULL, VG_GLOBAL);
log_print_unless_silent("%d physical volume%s changed / %d physical volume%s not changed",
params.done, params.done == 1 ? "" : "s",
params.total - params.done, (params.total - params.done) == 1 ? "" : "s");
out:
destroy_processing_handle(cmd, handle);
return ret;
}