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

231 lines
6.5 KiB
C

/*
* Copyright (C) 2001-2004 Sistina Software, Inc. All rights reserved.
* Copyright (C) 2004-2009 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"
int vgcreate(struct cmd_context *cmd, int argc, char **argv)
{
struct processing_handle *handle;
struct pvcreate_params pp;
struct vgcreate_params vp_new;
struct vgcreate_params vp_def;
struct volume_group *vg;
const char *tag;
const char *clustered_message = "";
char *vg_name;
struct arg_value_group_list *current_group;
uint32_t rc;
if (!argc) {
log_error("Please provide volume group name and "
"physical volumes");
return EINVALID_CMD_LINE;
}
vg_name = argv[0];
argc--;
argv++;
pvcreate_params_set_defaults(&pp);
if (!pvcreate_params_from_args(cmd, &pp))
return EINVALID_CMD_LINE;
pp.pv_count = argc;
pp.pv_names = argv;
/* Don't create a new PV on top of an existing PV like pvcreate does. */
pp.preserve_existing = 1;
/* pvcreate within vgcreate cannot be forced. */
pp.force = 0;
if (!vgcreate_params_set_defaults(cmd, &vp_def, NULL))
return EINVALID_CMD_LINE;
vp_def.vg_name = vg_name;
if (!vgcreate_params_set_from_args(cmd, &vp_new, &vp_def))
return EINVALID_CMD_LINE;
if (!vgcreate_params_validate(cmd, &vp_new))
return EINVALID_CMD_LINE;
/*
* Needed to change the global VG namespace,
* and to change the set of orphan PVs.
*/
if (!lockd_gl_create(cmd, "ex", vp_new.lock_type))
return_ECMD_FAILED;
cmd->lockd_gl_disable = 1;
lvmcache_seed_infos_from_lvmetad(cmd);
/*
* Check if the VG name already exists. This should be done before
* creating PVs on any of the devices.
*/
if ((rc = vg_lock_newname(cmd, vp_new.vg_name)) != SUCCESS) {
if (rc == FAILED_EXIST)
log_error("A volume group called %s already exists.", vp_new.vg_name);
else
log_error("Can't get lock for %s.", vp_new.vg_name);
return ECMD_FAILED;
}
/*
* FIXME: we have to unlock/relock the new VG name around the pvcreate
* step because pvcreate needs to destroy lvmcache, which doesn't allow
* any locks to be held. There shouldn't be any reason to require this
* VG lock to be released, so the lvmcache destroy rule about locks
* seems to be unwarranted here.
*/
unlock_vg(cmd, NULL, vp_new.vg_name);
if (!(handle = init_processing_handle(cmd, NULL))) {
log_error("Failed to initialize processing handle.");
return ECMD_FAILED;
}
if (!pvcreate_each_device(cmd, handle, &pp)) {
destroy_processing_handle(cmd, handle);
return_ECMD_FAILED;
}
/* Relock the new VG name, see comment above. */
if (!lock_vol(cmd, vp_new.vg_name, LCK_VG_WRITE, NULL)) {
destroy_processing_handle(cmd, handle);
return_ECMD_FAILED;
}
/*
* pvcreate_each_device returns with the VG_ORPHANS write lock held,
* which was used to do pvcreate. Now to create the VG using those
* PVs, the VG lock will be taken (with the orphan lock already held.)
*/
if (!(vg = vg_create(cmd, vp_new.vg_name)))
goto_bad;
if (vg->fid->fmt->features & FMT_CONFIG_PROFILE)
vg->profile = vg->cmd->profile_params->global_metadata_profile;
if (!vg_set_extent_size(vg, vp_new.extent_size) ||
!vg_set_max_lv(vg, vp_new.max_lv) ||
!vg_set_max_pv(vg, vp_new.max_pv) ||
!vg_set_alloc_policy(vg, vp_new.alloc) ||
!vg_set_clustered(vg, vp_new.clustered) ||
!vg_set_system_id(vg, vp_new.system_id) ||
!vg_set_mda_copies(vg, vp_new.vgmetadatacopies))
goto_bad;
/* attach the pv's */
if (!vg_extend_each_pv(vg, &pp))
goto_bad;
if (vp_new.max_lv != vg->max_lv)
log_warn("WARNING: Setting maxlogicalvolumes to %d "
"(0 means unlimited)", vg->max_lv);
if (vp_new.max_pv != vg->max_pv)
log_warn("WARNING: Setting maxphysicalvolumes to %d "
"(0 means unlimited)", vg->max_pv);
if (arg_is_set(cmd, addtag_ARG)) {
dm_list_iterate_items(current_group, &cmd->arg_value_groups) {
if (!grouped_arg_is_set(current_group->arg_values, addtag_ARG))
continue;
if (!(tag = grouped_arg_str_value(current_group->arg_values, addtag_ARG, NULL))) {
log_error("Failed to get tag");
goto bad;
}
if (!vg_change_tag(vg, tag, 1))
goto_bad;
}
}
if (vg_is_clustered(vg))
clustered_message = "Clustered ";
else if (locking_is_clustered())
clustered_message = "Non-clustered ";
if (!archive(vg))
goto_bad;
/* Store VG on disk(s) */
if (!vg_write(vg) || !vg_commit(vg))
goto_bad;
/*
* The VG is initially written without lock_type set, i.e. it starts as
* a local VG. lockd_init_vg() then writes the VG a second time with
* both lock_type and lock_args set.
*/
if (!lockd_init_vg(cmd, vg, vp_new.lock_type, 0)) {
log_error("Failed to initialize lock args for lock type %s",
vp_new.lock_type);
vg_remove_pvs(vg);
vg_remove_direct(vg);
goto_bad;
}
unlock_vg(cmd, NULL, VG_ORPHANS);
unlock_vg(cmd, vg, vp_new.vg_name);
backup(vg);
log_print_unless_silent("%s%colume group \"%s\" successfully created%s%s",
clustered_message, *clustered_message ? 'v' : 'V', vg->name,
vg->system_id ? " with system ID " : "", vg->system_id ? : "");
/*
* Start the VG lockspace because it will likely be used right away.
* Optionally wait for the start to complete so the VG can be fully
* used after this command completes (otherwise, the VG can only be
* read without locks until the lockspace is done starting.)
*/
if (is_lockd_type(vg->lock_type)) {
const char *start_opt = arg_str_value(cmd, lockopt_ARG, NULL);
if (!lockd_start_vg(cmd, vg, 1)) {
log_error("Failed to start locking");
goto out;
}
lockd_gl(cmd, "un", 0);
if (!start_opt || !strcmp(start_opt, "wait")) {
/* It is OK if the user does Ctrl-C to cancel the wait. */
log_print_unless_silent("Starting locking. Waiting until locks are ready...");
lockd_start_wait(cmd);
} else if (!strcmp(start_opt, "nowait")) {
log_print_unless_silent("Starting locking. VG is read-only until locks are ready.");
}
}
out:
release_vg(vg);
destroy_processing_handle(cmd, handle);
return ECMD_PROCESSED;
bad:
unlock_vg(cmd, vg, vp_new.vg_name);
unlock_vg(cmd, NULL, VG_ORPHANS);
release_vg(vg);
destroy_processing_handle(cmd, handle);
return ECMD_FAILED;
}