/* * Copyright (C) 2014-2015 Red Hat, Inc. * * 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. */ #include "lib.h" #include "toolcontext.h" #include "metadata.h" #include "segtype.h" #include "activate.h" #include "lvmetad.h" #include "lvmlockd.h" #include "lvmcache.h" #include "lvmlockd-client.h" static daemon_handle _lvmlockd; static const char *_lvmlockd_socket = NULL; static struct cmd_context *_lvmlockd_cmd = NULL; static int _use_lvmlockd = 0; /* is 1 if command is configured to use lvmlockd */ static int _lvmlockd_connected = 0; /* is 1 if command is connected to lvmlockd */ static int _lvmlockd_init_failed = 0; /* used to suppress further warnings */ void lvmlockd_set_socket(const char *sock) { _lvmlockd_socket = sock; } /* * Set directly from global/use_lvmlockd */ void lvmlockd_set_use(int use) { _use_lvmlockd = use; } /* * Returns the value of global/use_lvmlockd being used by the command. */ int lvmlockd_use(void) { return _use_lvmlockd; } /* * The command continues even if init and/or connect fail, * because the command is allowed to use local VGs without lvmlockd, * and is allowed to read lockd VGs without locks from lvmlockd. */ void lvmlockd_init(struct cmd_context *cmd) { if (!_use_lvmlockd) { /* Should never happen, don't call init when not using lvmlockd. */ log_error("Should not initialize lvmlockd with use_lvmlockd=0."); } if (!_lvmlockd_socket) { log_warn("WARNING: lvmlockd socket location is not configured."); _lvmlockd_init_failed = 1; } if (!!access(LVMLOCKD_PIDFILE, F_OK)) { log_warn("WARNING: lvmlockd process is not running."); _lvmlockd_init_failed = 1; } else { _lvmlockd_init_failed = 0; } _lvmlockd_cmd = cmd; } void lvmlockd_connect(void) { if (!_use_lvmlockd) { /* Should never happen, don't call connect when not using lvmlockd. */ log_error("Should not connect to lvmlockd with use_lvmlockd=0."); } if (_lvmlockd_connected) { /* Should never happen, only call connect once. */ log_error("lvmlockd is already connected."); } if (_lvmlockd_init_failed) return; _lvmlockd = lvmlockd_open(_lvmlockd_socket); if (_lvmlockd.socket_fd >= 0 && !_lvmlockd.error) { log_debug("Successfully connected to lvmlockd on fd %d.", _lvmlockd.socket_fd); _lvmlockd_connected = 1; } else { log_warn("WARNING: lvmlockd connect failed."); } } void lvmlockd_disconnect(void) { if (_lvmlockd_connected) daemon_close(_lvmlockd); _lvmlockd_connected = 0; _lvmlockd_cmd = NULL; } /* Translate the result strings from lvmlockd to bit flags. */ static void _flags_str_to_lockd_flags(const char *flags_str, uint32_t *lockd_flags) { if (strstr(flags_str, "NO_LOCKSPACES")) *lockd_flags |= LD_RF_NO_LOCKSPACES; if (strstr(flags_str, "NO_GL_LS")) *lockd_flags |= LD_RF_NO_GL_LS; if (strstr(flags_str, "LOCAL_LS")) *lockd_flags |= LD_RF_LOCAL_LS; if (strstr(flags_str, "DUP_GL_LS")) *lockd_flags |= LD_RF_DUP_GL_LS; if (strstr(flags_str, "INACTIVE_LS")) *lockd_flags |= LD_RF_INACTIVE_LS; if (strstr(flags_str, "ADD_LS_ERROR")) *lockd_flags |= LD_RF_ADD_LS_ERROR; } /* * evaluate the reply from lvmlockd, check for errors, extract * the result and lockd_flags returned by lvmlockd. * 0 failure (no result/lockd_flags set) * 1 success (result/lockd_flags set) */ /* * This is an arbitrary number that we know lvmlockd * will not return. daemon_reply_int reverts to this * value if it finds no result value. */ #define NO_LOCKD_RESULT -1000 static int _lockd_result(daemon_reply reply, int *result, uint32_t *lockd_flags) { int reply_result; const char *flags_str = NULL; const char *lock_type = NULL; if (reply.error) { log_error("lockd_result reply error %d", reply.error); return 0; } if (strcmp(daemon_reply_str(reply, "response", ""), "OK")) { log_error("lockd_result bad response"); return 0; } reply_result = daemon_reply_int(reply, "op_result", NO_LOCKD_RESULT); if (reply_result == NO_LOCKD_RESULT) { log_error("lockd_result no op_result"); return 0; } /* The lock_type that lvmlockd used for locking. */ lock_type = daemon_reply_str(reply, "lock_type", "none"); *result = reply_result; if (lockd_flags) { if ((flags_str = daemon_reply_str(reply, "result_flags", NULL))) _flags_str_to_lockd_flags(flags_str, lockd_flags); } log_debug("lockd_result %d flags %s lm %s", reply_result, flags_str ? flags_str : "none", lock_type); return 1; } static daemon_reply _lockd_send(const char *req_name, ...) { va_list ap; daemon_reply repl; daemon_request req; req = daemon_request_make(req_name); va_start(ap, req_name); daemon_request_extend_v(req, ap); va_end(ap); repl = daemon_send(_lvmlockd, req); daemon_request_destroy(req); return repl; } /* * result/lockd_flags are values returned from lvmlockd. * * return 0 (failure) * return 1 (result/lockd_flags indicate success/failure) * * return 1 result 0 (success) * return 1 result < 0 (failure) * * caller may ignore result < 0 failure depending on * lockd_flags and the specific command/mode. * * When this function returns 0 (failure), no result/lockd_flags * were obtained from lvmlockd. * * When this function returns 1 (success), result/lockd_flags may * have been obtained from lvmlockd. This lvmlockd result may * indicate a locking failure. */ static int _lockd_request(struct cmd_context *cmd, const char *req_name, const char *vg_name, const char *vg_lock_type, const char *vg_lock_args, const char *lv_name, const char *lv_uuid, const char *lv_lock_args, const char *mode, const char *opts, int *result, uint32_t *lockd_flags) { const char *cmd_name = get_cmd_name(); daemon_reply reply; int pid = getpid(); *result = 0; *lockd_flags = 0; if (!strcmp(mode, "na")) return 1; if (!_use_lvmlockd) return 0; if (!_lvmlockd_connected) return 0; /* cmd and pid are passed for informational and debugging purposes */ if (!cmd_name || !cmd_name[0]) cmd_name = "none"; if (vg_name && lv_name) { reply = _lockd_send(req_name, "cmd = %s", cmd_name, "pid = %d", pid, "mode = %s", mode, "opts = %s", opts ?: "none", "vg_name = %s", vg_name, "lv_name = %s", lv_name, "lv_uuid = %s", lv_uuid, "vg_lock_type = %s", vg_lock_type ?: "none", "vg_lock_args = %s", vg_lock_args ?: "none", "lv_lock_args = %s", lv_lock_args ?: "none", NULL); if (!_lockd_result(reply, result, lockd_flags)) goto fail; log_debug("lvmlockd %s %s vg %s lv %s result %d %x", req_name, mode, vg_name, lv_name, *result, *lockd_flags); } else if (vg_name) { reply = _lockd_send(req_name, "cmd = %s", cmd_name, "pid = %d", pid, "mode = %s", mode, "opts = %s", opts ?: "none", "vg_name = %s", vg_name, "vg_lock_type = %s", vg_lock_type ?: "none", "vg_lock_args = %s", vg_lock_args ?: "none", NULL); if (!_lockd_result(reply, result, lockd_flags)) goto fail; log_debug("lvmlockd %s %s vg %s result %d %x", req_name, mode, vg_name, *result, *lockd_flags); } else { reply = _lockd_send(req_name, "cmd = %s", cmd_name, "pid = %d", pid, "mode = %s", mode, "opts = %s", opts ?: "none", "vg_lock_type = %s", vg_lock_type ?: "none", NULL); if (!_lockd_result(reply, result, lockd_flags)) goto fail; log_debug("lvmlockd %s %s result %d %x", req_name, mode, *result, *lockd_flags); } daemon_reply_destroy(reply); /* result/lockd_flags have lvmlockd result */ return 1; fail: /* no result was obtained from lvmlockd */ log_error("lvmlockd %s %s failed no result", req_name, mode); daemon_reply_destroy(reply); return 0; } /* * Eventually add an option to specify which pv the lvmlock lv should be placed on. * FIXME: when converting a VG from lock_type none to sanlock, we need to count * the number of existing LVs to ensure that the new sanlock_lv is large enough * for all of them that need locks. */ static int _create_sanlock_lv(struct cmd_context *cmd, struct volume_group *vg, const char *lock_lv_name, int extend_mb) { struct logical_volume *lv; struct lvcreate_params lp = { .activate = CHANGE_ALY, .alloc = ALLOC_INHERIT, .extents = (extend_mb * 1024 * 1024) / (vg->extent_size * SECTOR_SIZE), .major = -1, .minor = -1, .permission = LVM_READ | LVM_WRITE, .pvh = &vg->pvs, .read_ahead = DM_READ_AHEAD_NONE, .stripes = 1, .vg_name = vg->name, .lv_name = dm_pool_strdup(cmd->mem, lock_lv_name), .zero = 1, }; dm_list_init(&lp.tags); if (!(lp.segtype = get_segtype_from_string(vg->cmd, "striped"))) return_0; lv = lv_create_single(vg, &lp); if (!lv) { log_error("Failed to create sanlock lv %s in vg %s", lock_lv_name, vg->name); return 0; } vg->sanlock_lv = lv; return 1; } static int _remove_sanlock_lv(struct cmd_context *cmd, struct volume_group *vg) { if (!lv_remove(vg->sanlock_lv)) { log_error("Failed to remove sanlock LV %s/%s", vg->name, vg->sanlock_lv->name); return 0; } return 1; } static int _extend_sanlock_lv(struct cmd_context *cmd, struct volume_group *vg, int extend_mb) { struct logical_volume *lv = vg->sanlock_lv; struct lvresize_params lp = { .lv_name = vg->sanlock_lv->name, .sign = SIGN_NONE, .percent = PERCENT_NONE, .resize = LV_EXTEND, .ac_force = 1, .sizeargs = 1, }; lp.size = lv->size + ((extend_mb * 1024 * 1024) / SECTOR_SIZE); if (!lv_resize_prepare(cmd, lv, &lp, &vg->pvs) || !lv_resize(cmd, lv, &lp, &vg->pvs)) { log_error("Extend LV %s/%s to size %llu failed.", vg->name, lv->name, (unsigned long long)lp.size); return 0; } return 1; } /* When one host does _extend_sanlock_lv, the others need to refresh the size. */ static int _refresh_sanlock_lv(struct cmd_context *cmd, struct volume_group *vg) { if (!lv_refresh_suspend_resume(cmd, vg->sanlock_lv)) { log_error("Failed to refresh %s.", vg->sanlock_lv->name); return 0; } return 1; } /* * Called at the beginning of lvcreate in a sanlock VG to ensure * that there is space in the sanlock LV for a new lock. If it's * full, then this extends it. */ int handle_sanlock_lv(struct cmd_context *cmd, struct volume_group *vg) { daemon_reply reply; int extend_mb; int result; int ret; if (!_use_lvmlockd) return 1; if (!_lvmlockd_connected) return 0; extend_mb = find_config_tree_int(cmd, global_sanlock_lv_extend_CFG, NULL); /* * User can choose to not automatically extend the lvmlock LV * so they can manually extend it. */ if (!extend_mb) return 1; /* * Another host may have extended the lvmlock LV already. * Refresh so that we'll find the new space they added * when we search for new space. */ if (!_refresh_sanlock_lv(cmd, vg)) return 0; /* * Ask lvmlockd/sanlock to look for an unused lock. */ reply = _lockd_send("find_free_lock", "pid = %d", getpid(), "vg_name = %s", vg->name, NULL); if (!_lockd_result(reply, &result, NULL)) { ret = 0; } else { ret = (result < 0) ? 0 : 1; } /* No space on the lvmlock lv for a new lease. */ if (result == -EMSGSIZE) ret = _extend_sanlock_lv(cmd, vg, extend_mb); daemon_reply_destroy(reply); return ret; } static int _activate_sanlock_lv(struct cmd_context *cmd, struct volume_group *vg) { if (!activate_lv(cmd, vg->sanlock_lv)) { log_error("Failed to activate sanlock lv %s/%s", vg->name, vg->sanlock_lv->name); return 0; } return 1; } static int _deactivate_sanlock_lv(struct cmd_context *cmd, struct volume_group *vg) { if (!deactivate_lv(cmd, vg->sanlock_lv)) { log_error("Failed to deactivate sanlock lv %s/%s", vg->name, vg->sanlock_lv->name); return 0; } return 1; } static int _init_vg_dlm(struct cmd_context *cmd, struct volume_group *vg) { daemon_reply reply; const char *reply_str; const char *vg_lock_args = NULL; int result; int ret; if (!_use_lvmlockd) return 0; if (!_lvmlockd_connected) return 0; reply = _lockd_send("init_vg", "pid = %d", getpid(), "vg_name = %s", vg->name, "vg_lock_type = %s", "dlm", NULL); if (!_lockd_result(reply, &result, NULL)) { ret = 0; result = -ELOCKD; } else { ret = (result < 0) ? 0 : 1; } switch (result) { case 0: break; case -ELOCKD: log_error("VG %s init failed: lvmlockd not available", vg->name); break; case -EARGS: log_error("VG %s init failed: invalid parameters for dlm", vg->name); break; case -EMANAGER: log_error("VG %s init failed: lock manager dlm is not running", vg->name); break; default: log_error("VG %s init failed: %d", vg->name, result); } if (!ret) goto out; if (!(reply_str = daemon_reply_str(reply, "vg_lock_args", NULL))) { log_error("VG %s init failed: lock_args not returned", vg->name); ret = 0; goto out; } if (!(vg_lock_args = dm_pool_strdup(cmd->mem, reply_str))) { log_error("VG %s init failed: lock_args alloc failed", vg->name); ret = 0; goto out; } vg->lock_type = "dlm"; vg->lock_args = vg_lock_args; if (!vg_write(vg) || !vg_commit(vg)) { log_error("VG %s init failed: vg_write vg_commit", vg->name); ret = 0; goto out; } ret = 1; out: daemon_reply_destroy(reply); return ret; } static int _init_vg_sanlock(struct cmd_context *cmd, struct volume_group *vg) { daemon_reply reply; const char *reply_str; const char *vg_lock_args = NULL; const char *opts = NULL; int extend_mb; int result; int ret; if (!_use_lvmlockd) return 0; if (!_lvmlockd_connected) return 0; /* * Automatic extension of the sanlock lv is disabled by * setting sanlock_lv_extend to 0. Zero won't work as * an initial size, so in this case, use the default as * the initial size. */ if (!(extend_mb = find_config_tree_int(cmd, global_sanlock_lv_extend_CFG, NULL))) extend_mb = DEFAULT_SANLOCK_LV_EXTEND_MB; /* * Creating the sanlock LV writes the VG containing the new lvmlock * LV, then activates the lvmlock LV. The lvmlock LV must be active * before we ask lvmlockd to initialize the VG because sanlock needs * to initialize leases on the lvmlock LV. */ if (!_create_sanlock_lv(cmd, vg, LOCKD_SANLOCK_LV_NAME, extend_mb)) { log_error("Failed to create internal lv."); return 0; } /* * N.B. this passes the sanlock lv name as vg_lock_args * even though it is only part of the final args string * which will be returned from lvmlockd. */ reply = _lockd_send("init_vg", "pid = %d", getpid(), "vg_name = %s", vg->name, "vg_lock_type = %s", "sanlock", "vg_lock_args = %s", vg->sanlock_lv->name, "opts = %s", opts ?: "none", NULL); if (!_lockd_result(reply, &result, NULL)) { ret = 0; result = -ELOCKD; } else { ret = (result < 0) ? 0 : 1; } switch (result) { case 0: break; case -ELOCKD: log_error("VG %s init failed: lvmlockd not available", vg->name); break; case -EARGS: log_error("VG %s init failed: invalid parameters for sanlock", vg->name); break; case -EMANAGER: log_error("VG %s init failed: lock manager sanlock is not running", vg->name); break; case -EMSGSIZE: log_error("VG %s init failed: no disk space for leases", vg->name); break; default: log_error("VG %s init failed: %d", vg->name, result); } if (!ret) goto out; if (!(reply_str = daemon_reply_str(reply, "vg_lock_args", NULL))) { log_error("VG %s init failed: lock_args not returned", vg->name); ret = 0; goto out; } if (!(vg_lock_args = dm_pool_strdup(cmd->mem, reply_str))) { log_error("VG %s init failed: lock_args alloc failed", vg->name); ret = 0; goto out; } lv_set_hidden(vg->sanlock_lv); vg->sanlock_lv->status |= LOCKD_SANLOCK_LV; vg->lock_type = "sanlock"; vg->lock_args = vg_lock_args; if (!vg_write(vg) || !vg_commit(vg)) { log_error("VG %s init failed: vg_write vg_commit", vg->name); ret = 0; goto out; } ret = 1; out: if (!ret) { /* * The usleep delay gives sanlock time to close the lock lv, * and usually avoids having an annoying error printed. */ usleep(1000000); _deactivate_sanlock_lv(cmd, vg); _remove_sanlock_lv(cmd, vg); if (!vg_write(vg) || !vg_commit(vg)) stack; } daemon_reply_destroy(reply); return ret; } /* called after vg_remove on disk */ static int _free_vg_dlm(struct cmd_context *cmd, struct volume_group *vg) { uint32_t lockd_flags; int result; int ret; if (!_use_lvmlockd) return 0; if (!_lvmlockd_connected) return 0; /* * Unlocking the vg lock here preempts the lvmlockd unlock in * toollib.c which happens too late since the lockspace is * left here. */ /* Equivalent to a standard unlock. */ ret = _lockd_request(cmd, "lock_vg", vg->name, NULL, NULL, NULL, NULL, NULL, "un", NULL, &result, &lockd_flags); if (!ret || result < 0) { log_error("_free_vg_dlm lvmlockd result %d", result); return 0; } /* Leave the dlm lockspace. */ lockd_stop_vg(cmd, vg); return 1; } /* called before vg_remove on disk */ static int _free_vg_sanlock(struct cmd_context *cmd, struct volume_group *vg) { daemon_reply reply; int result; int ret; if (!_use_lvmlockd) return 0; if (!_lvmlockd_connected) return 0; if (!vg->lock_args || !strlen(vg->lock_args)) { /* Shouldn't happen in general, but maybe in some error cases? */ log_debug("_free_vg_sanlock %s no lock_args", vg->name); return 1; } reply = _lockd_send("free_vg", "pid = %d", getpid(), "vg_name = %s", vg->name, "vg_lock_type = %s", vg->lock_type, "vg_lock_args = %s", vg->lock_args, NULL); if (!_lockd_result(reply, &result, NULL)) { ret = 0; } else { ret = (result < 0) ? 0 : 1; } /* * Other hosts could still be joined to the lockspace, which means they * are using the internal sanlock LV, which means we cannot remove the * VG. Once other hosts stop using the VG it can be removed. */ if (result == -EBUSY) { log_error("Lockspace for \"%s\" not stopped on other hosts", vg->name); goto out; } if (!ret) { log_error("_free_vg_sanlock lvmlockd result %d", result); goto out; } /* * The usleep delay gives sanlock time to close the lock lv, * and usually avoids having an annoying error printed. */ usleep(1000000); _deactivate_sanlock_lv(cmd, vg); _remove_sanlock_lv(cmd, vg); out: daemon_reply_destroy(reply); return ret; } /* * Tell lvmlockd to forget about an old VG name. * lvmlockd remembers previous lockd VGs so that it can provide more * informative error messages (see INACTIVE_LS, ADD_LS_ERROR). * * If a new local VG is created with the same name as a previous lockd VG, * lvmlockd's memory of the previous lockd VG interferes (causes incorrect * lockd_vg failures). * * We could also remove the list of inactive (old) VG names from lvmlockd, * and then this function would not be needed, but this would also reduce * the ability to have helpful error messages. */ static void _forget_vg_name(struct cmd_context *cmd, struct volume_group *vg) { daemon_reply reply; if (!_use_lvmlockd) return; if (!_lvmlockd_connected) return; reply = _lockd_send("forget_vg_name", "pid = %d", getpid(), "vg_name = %s", vg->name, NULL); daemon_reply_destroy(reply); } /* vgcreate */ int lockd_init_vg(struct cmd_context *cmd, struct volume_group *vg, const char *lock_type) { switch (get_lock_type_from_string(lock_type)) { case LOCK_TYPE_NONE: _forget_vg_name(cmd, vg); return 1; case LOCK_TYPE_CLVM: return 1; case LOCK_TYPE_DLM: return _init_vg_dlm(cmd, vg); case LOCK_TYPE_SANLOCK: return _init_vg_sanlock(cmd, vg); default: log_error("Unknown lock_type."); return 0; } } /* vgremove before the vg is removed */ int lockd_free_vg_before(struct cmd_context *cmd, struct volume_group *vg) { switch (get_lock_type_from_string(vg->lock_type)) { case LOCK_TYPE_NONE: case LOCK_TYPE_CLVM: case LOCK_TYPE_DLM: return 1; case LOCK_TYPE_SANLOCK: /* returning an error will prevent vg_remove() */ return _free_vg_sanlock(cmd, vg); default: log_error("Unknown lock_type."); return 0; } } /* vgremove after the vg is removed */ void lockd_free_vg_final(struct cmd_context *cmd, struct volume_group *vg) { switch (get_lock_type_from_string(vg->lock_type)) { case LOCK_TYPE_NONE: case LOCK_TYPE_CLVM: case LOCK_TYPE_SANLOCK: break; case LOCK_TYPE_DLM: _free_vg_dlm(cmd, vg); break; default: log_error("Unknown lock_type."); } } /* * Starting a vg involves: * 1. reading the vg without a lock * 2. getting the lock_type/lock_args from the vg metadata * 3. doing start_vg in lvmlockd for the lock_type; * this means joining the lockspace * * The vg read in step 1 should not be used for anything * other than getting the lock_type/lock_args/uuid necessary * for starting the lockspace. To use the vg after starting * the lockspace, follow the standard method which is: * lock the vg, read/use/write the vg, unlock the vg. */ int lockd_start_vg(struct cmd_context *cmd, struct volume_group *vg) { char uuid[64] __attribute__((aligned(8))); daemon_reply reply; int host_id = 0; int result; int ret; memset(uuid, 0, sizeof(uuid)); if (!is_lockd_type(vg->lock_type)) return 1; if (!_use_lvmlockd) { log_error("VG %s start failed: lvmlockd is not enabled", vg->name); return 0; } if (!_lvmlockd_connected) { log_error("VG %s start failed: lvmlockd is not running", vg->name); return 0; } log_debug("lockd start VG %s lock_type %s", vg->name, vg->lock_type ? vg->lock_type : "empty"); if (!id_write_format(&vg->id, uuid, sizeof(uuid))) return_0; if (vg->lock_type && !strcmp(vg->lock_type, "sanlock")) { /* * This is the big difference between starting * sanlock vgs vs starting dlm vgs: the internal * sanlock lv needs to be activated before lvmlockd * does the start because sanlock needs to use the lv * to access locks. */ if (!_activate_sanlock_lv(cmd, vg)) return 0; host_id = find_config_tree_int(cmd, local_host_id_CFG, NULL); } reply = _lockd_send("start_vg", "pid = %d", getpid(), "vg_name = %s", vg->name, "vg_lock_type = %s", vg->lock_type, "vg_lock_args = %s", vg->lock_args ?: "none", "vg_uuid = %s", uuid[0] ? uuid : "none", "version = %d", (int64_t)vg->seqno, "host_id = %d", host_id, NULL); if (!_lockd_result(reply, &result, NULL)) { ret = 0; result = -ELOCKD; } else { ret = (result < 0) ? 0 : 1; } switch (result) { case 0: log_print_unless_silent("VG %s starting %s lockspace", vg->name, vg->lock_type); break; case -ELOCKD: log_error("VG %s start failed: lvmlockd not available", vg->name); break; case -EEXIST: log_debug("VG %s start error: already started", vg->name); ret = 1; break; case -EARGS: log_error("VG %s start failed: invalid parameters for %s", vg->name, vg->lock_type); break; case -EHOSTID: log_error("VG %s start failed: invalid sanlock host_id, set in lvmlocal.conf", vg->name); break; case -EMANAGER: log_error("VG %s start failed: lock manager %s is not running", vg->name, vg->lock_type); break; default: log_error("VG %s start failed: %d", vg->name, result); } daemon_reply_destroy(reply); return ret; } int lockd_stop_vg(struct cmd_context *cmd, struct volume_group *vg) { daemon_reply reply; int result; int ret; if (!is_lockd_type(vg->lock_type)) return 1; if (!_use_lvmlockd) return 0; if (!_lvmlockd_connected) return 0; log_debug("lockd stop VG %s lock_type %s", vg->name, vg->lock_type ? vg->lock_type : "empty"); reply = _lockd_send("stop_vg", "pid = %d", getpid(), "vg_name = %s", vg->name, NULL); if (!_lockd_result(reply, &result, NULL)) { ret = 0; } else { ret = (result < 0) ? 0 : 1; } if (result == -ENOLS) { ret = 1; goto out; } if (result == -EBUSY) { log_error("VG %s stop failed: LVs must first be deactivated", vg->name); goto out; } if (!ret) { log_error("VG %s stop failed: %d", vg->name, result); goto out; } if (!strcmp(vg->lock_type, "sanlock")) { log_debug("lockd_stop_vg deactivate sanlock lv"); _deactivate_sanlock_lv(cmd, vg); } out: daemon_reply_destroy(reply); return ret; } int lockd_start_wait(struct cmd_context *cmd) { daemon_reply reply; int result; int ret; if (!_use_lvmlockd) return 0; if (!_lvmlockd_connected) return 0; reply = _lockd_send("start_wait", "pid = %d", getpid(), NULL); if (!_lockd_result(reply, &result, NULL)) { ret = 0; } else { ret = (result < 0) ? 0 : 1; } if (!ret) log_error("Lock start failed"); /* * FIXME: get a list of vgs that started so we can * better report what worked and what didn't? */ daemon_reply_destroy(reply); return ret; } /* * lockd_gl_create() is a variation of lockd_gl() used only by vgcreate. * It handles the case that when using sanlock, the global lock does * not exist until after the first vgcreate is complete, since the global * lock exists on storage within an actual VG. So, the first vgcreate * needs special logic to detect this bootstrap case. * * When the vgcreate is not creating the first VG, then lockd_gl_create() * behaves the same as lockd_gl(). * * vgcreate will have a lock_type for the new VG which lockd_gl_create() * can provide in the lock-gl call. * * lockd_gl() and lockd_gl_create() differ in the specific cases where * ENOLS (no lockspace found) is overriden. In the vgcreate case, the * override cases are related to sanlock bootstrap, and the lock_type of * the vg being created is needed. * * 1. vgcreate of the first lockd-type vg calls lockd_gl_create() * to acquire the global lock. * * 2. vgcreate/lockd_gl_create passes gl lock request to lvmlockd, * along with lock_type of the new vg. * * 3. lvmlockd finds no global lockspace/lock. * * 4. dlm: * If the lock_type from vgcreate is dlm, lvmlockd creates the * dlm global lockspace, and queues the global lock request * for vgcreate. lockd_gl_create returns sucess with the gl held. * * sanlock: * If the lock_type from vgcreate is sanlock, lvmlockd returns -ENOLS * with the NO_GL_LS flag. lvmlockd cannot create or acquire a sanlock * global lock until the VG exists on disk (the locks live within the VG). * * lockd_gl_create sees sanlock/ENOLS/NO_GL_LS (and optionally the * "enable" lock-gl arg), determines that this is the sanlock * bootstrap special case, and returns success without the global lock. * * vgcreate creates the VG on disk, and calls lockd_init_vg() which * initializes/enables a global lock on the new VG's internal sanlock lv. * Future lockd_gl/lockd_gl_create calls will acquire the existing gl. */ int lockd_gl_create(struct cmd_context *cmd, const char *def_mode, const char *vg_lock_type) { const char *mode = NULL; uint32_t lockd_flags; int retries = 0; int result; /* * There are four variations of creating a local/lockd VG * with/without use_lvmlockd set. * * use_lvmlockd=1, lockd VG: * This function should acquire or create the global lock. * * use_lvmlockd=0, local VG: * This function is a no-op, just returns 1. * * use_lvmlockd=0, lockd VG * An error is returned in vgcreate_params_set_from_args (before this is called). * * use_lvmlockd=1, local VG * This function should acquire the global lock. */ if (!_use_lvmlockd) { if (!is_lockd_type(vg_lock_type)) return 1; log_error("Cannot create VG with lock_type %s without lvmlockd.", vg_lock_type); return 0; } log_debug("lockd global lock_type %s", vg_lock_type); if (!mode) mode = def_mode; if (!mode) { log_error("Unknown lock-gl mode"); return 0; } req: if (!_lockd_request(cmd, "lock_gl", NULL, vg_lock_type, NULL, NULL, NULL, NULL, mode, NULL, &result, &lockd_flags)) { /* No result from lvmlockd, it is probably not running. */ log_error("Global lock failed: check that lvmlockd is running."); return 0; } if (result == -EAGAIN) { if (retries < find_config_tree_int(cmd, global_lvmlockd_lock_retries_CFG, NULL)) { log_warn("Retrying %s global lock", mode); sleep(1); retries++; goto req; } } /* * ENOLS: no lockspace was found with a global lock. * It may not exist (perhaps this command is creating the first), * or it may not be visible or started on the system yet. */ if (result == -ENOLS) { if (!strcmp(mode, "un")) return 1; /* * This is the explicit sanlock bootstrap condition for * proceding without the global lock: a chicken/egg case * for the first sanlock VG that is created. * * When creating the first sanlock VG, there is no global * lock to acquire because the gl will exist in the VG * being created. The "enable" option makes explicit that * this is expected: * * vgcreate --lock-type sanlock --lock-gl enable * * There are three indications that this is the unique * first-sanlock-vg bootstrap case: * * - result from lvmlockd is -ENOLS because lvmlockd found * no lockspace for this VG; expected because it's being * created here. * * - result flag LD_RF_NO_GL_LS from lvmlockd means that * lvmlockd has seen no other lockspace with a global lock. * This implies that this is probably the first sanlock vg * to be created. If other sanlock vgs exist, the global * lock should be available from one of them. * * - command line lock-gl arg is "enable" which means the * user expects this to be the first sanlock vg, and the * global lock should be enabled in it. */ if ((lockd_flags & LD_RF_NO_GL_LS) && !strcmp(vg_lock_type, "sanlock") && !strcmp(mode, "enable")) { log_print_unless_silent("Enabling sanlock global lock"); lvmetad_validate_global_cache(cmd, 1); return 1; } /* * This is an implicit sanlock bootstrap condition for * proceeding without the global lock. The command line does * not indicate explicitly that this is a bootstrap situation * (via "enable"), but it seems likely to be because lvmlockd * has seen no lockd-type vgs. It is possible that a global * lock does exist in a vg that has not yet been seen. If that * vg appears after this creates a new vg with a new enabled * gl, then there will be two enabled global locks, and one * will need to be disabled. (We could instead return an error * here and insist with an error message that the --lock-gl * enable option be used to exercise the explicit case above.) */ if ((lockd_flags & LD_RF_NO_GL_LS) && (lockd_flags & LD_RF_NO_LOCKSPACES) && !strcmp(vg_lock_type, "sanlock")) { log_print_unless_silent("Enabling sanlock global lock"); lvmetad_validate_global_cache(cmd, 1); return 1; } if (!strcmp(vg_lock_type, "sanlock")) log_error("Global lock failed: check that VG holding global lock exists and is started."); else log_error("Global lock failed: check that global lockspace is started."); return 0; } /* * Check for each specific error that can be returned so a helpful * message can be printed for it. */ if (result < 0) { if (result == -ESTARTING) log_error("Global lock failed: lockspace is starting."); else if (result -EAGAIN) log_error("Global lock failed: held by other host."); else log_error("Global lock failed: error %d", result); return 0; } lvmetad_validate_global_cache(cmd, 1); return 1; } /* * The global lock protects: * * - The global VG namespace. Two VGs cannot have the same name. * Used by any command that creates or removes a VG name, * e.g. vgcreate, vgremove, vgrename, vgsplit, vgmerge. * * - The set of orphan PVs. * Used by any command that changes a non-PV device into an orphan PV, * an orphan PV into a device, a non-orphan PV (in a VG) into an orphan PV * (not in a VG), or an orphan PV into a non-orphan PV, * e.g. pvcreate, pvremove, vgcreate, vgremove, vgextend, vgreduce. * * - The properties of orphan PVs. It is possible to make changes to the * properties of an orphan PV, e.g. pvresize, pvchange. * * These are things that cannot be protected by a VG lock alone, since * orphan PVs do not belong to a real VG (an artificial VG does not * apply since a sanlock lock only exists on real storage.) * * If a command will change any of the things above, it must first acquire * the global lock in exclusive mode. * * If command is reading any of the things above, it must acquire the global * lock in shared mode. A number of commands read the things above, including: * * - Reporting/display commands which show all VGs. Any command that * will iterate through the entire VG namespace must first acquire the * global lock shared so that it has an accurate view of the namespace. * * - A command where a tag name is used to identify what to process. * A tag requires reading all VGs to check if they match the tag. * * In these cases, the global lock must be acquired before the list of * all VGs is created. * * The global lock is not generally unlocked explicitly in the code. * When the command disconnects from lvmlockd, lvmlockd automatically * releases the locks held by the command. The exception is if a command * will continue running for a long time while not needing the global lock, * e.g. commands that poll to report progress. * * Acquiring the global lock also updates the local lvmetad cache if * necessary. lockd_gl() first acquires the lock via lvmlockd, then * before returning to the caller, it checks that the global information * (e.g. VG namespace, set of orphans) is up to date in lvmetad. If * not, it scans disks and updates the lvmetad cache before returning * to the caller. It does this checking using a version number associated * with the global lock. The version number is incremented each time * a change is made to the state associated with the global lock, and * if the local version number is lower than the version number in the * lock, then the local lvmetad state must be updated. * * There are two cases where the global lock can be taken in shared mode, * and then later converted to ex. pvchange and pvresize use process_each_pv * which does lockd_gl("sh") to get the list of VGs. Later, in the "_single" * function called within process_each_pv, the PV may be an orphan, in which * case the ex global lock is needed, so it's converted to ex at that point. * * Effects of misconfiguring use_lvmlockd. * * - Setting use_lvmlockd=1 tells lvm commands to use the global lock. * This should not be set unless a lock manager and lockd VGs will * be used. Setting use_lvmlockd=1 without setting up a lock manager * or using lockd VGs will cause lvm commands to fail when they attempt * to change any global state (requiring the ex global lock), and will * cause warnings when the commands read global state (requiring the sh * global lock). In this condition, lvm is nominally useful, and existing * local VGs can continue to be used mostly as usual. But, the * warnings/errors should lead a user to either set up a lock manager * and lockd VGs, or set use_lvmlockd to 0. * * - Setting use_lvmlockd=0 tells lvm commands to not use the global lock. * If use_lvmlockd=0 when lockd VGs exist which require lvmlockd, the * lockd_gl() calls become no-ops, but the lockd_vg() calls for the lockd * VGs will fail. The warnings/errors from accessing the lockd VGs * should lead the user to set use_lvmlockd to 1 and run the necessary * lock manager. In this condition, lvm reverts to the behavior of * the following case, in which system ID largely protects shared * devices, but has limitations. * * - Setting use_lvmlockd=0 with shared devices, no lockd VGs and * no lock manager is a recognized mode of operation that is * described in the lvmsystemid man page. Using lvm on shared * devices this way is made safe by using system IDs to assign * ownership of VGs to single hosts. The main limitation of this * mode (among others outlined in the man page), is that orphan PVs * are unprotected. */ int lockd_gl(struct cmd_context *cmd, const char *def_mode, uint32_t flags) { const char *mode = NULL; const char *opts = NULL; uint32_t lockd_flags; int retries = 0; int result; if (!_use_lvmlockd) return 1; if (cmd->lockd_gl_disable) return 1; if (def_mode && !strcmp(def_mode, "un")) { mode = "un"; goto req; } if (!mode) mode = def_mode; if (!mode) { log_error("Unknown lock-gl mode"); return 0; } req: log_debug("lockd global mode %s", mode); if (!_lockd_request(cmd, "lock_gl", NULL, NULL, NULL, NULL, NULL, NULL, mode, opts, &result, &lockd_flags)) { /* No result from lvmlockd, it is probably not running. */ /* We don't care if an unlock fails. */ if (!strcmp(mode, "un")) return 1; /* We can continue reading if a shared lock fails. */ if (!strcmp(mode, "sh")) { log_warn("Reading without shared global lock."); lvmetad_validate_global_cache(cmd, 1); return 1; } log_error("Global lock failed: check that lvmlockd is running."); return 0; } if (result == -EAGAIN) { if (retries < find_config_tree_int(cmd, global_lvmlockd_lock_retries_CFG, NULL)) { log_warn("Retrying %s global lock", mode); sleep(1); retries++; goto req; } } /* * ENOLS: no lockspace was found with a global lock. * The VG with the global lock may not be visible or started yet, * this should be a temporary condition. * * ESTARTING: the lockspace with the gl is starting. * The VG with the global lock is starting and should finish shortly. */ if (result == -ENOLS || result == -ESTARTING) { if (!strcmp(mode, "un")) return 1; /* * If an ex global lock fails, then the command fails. */ if (strcmp(mode, "sh")) { if (result == -ESTARTING) log_error("Global lock failed: lockspace is starting."); else if (result == -ENOLS) log_error("Global lock failed: check that global lockspace is started."); else log_error("Global lock failed: error %d", result); return 0; } /* * If a sh global lock fails, then the command can continue * reading without it, but force a global cache validation, * and print a warning. */ if (result == -ESTARTING) { log_warn("Skipping global lock: lockspace is starting"); lvmetad_validate_global_cache(cmd, 1); return 1; } if ((lockd_flags & LD_RF_NO_GL_LS) || (lockd_flags & LD_RF_NO_LOCKSPACES)) { log_warn("Skipping global lock: lockspace not found or started"); lvmetad_validate_global_cache(cmd, 1); return 1; } /* * This is for completeness. If we reach here, then * a specific check for the error should be added above * with a more helpful message. */ log_error("Global lock failed: error %d", result); return 0; } if ((lockd_flags & LD_RF_DUP_GL_LS) && strcmp(mode, "un")) log_warn("Duplicate sanlock global locks should be corrected"); if (result < 0) { if (ignorelockingfailure()) { log_debug("Ignore failed locking for global lock"); lvmetad_validate_global_cache(cmd, 1); return 1; } else if (result == -EAGAIN) { /* * Most of the time, retries should avoid this case. */ log_error("Global lock failed: held by other host."); return 0; } else { /* * We don't intend to reach this. We should check * any known/possible error specifically and print * a more helpful message. This is for completeness. */ log_error("Global lock failed: error %d.", result); return 0; } } if (!(flags & LDGL_SKIP_CACHE_VALIDATE)) lvmetad_validate_global_cache(cmd, 0); return 1; } /* * VG lock * * Return 1: continue, lockd_state may still indicate an error * Return 0: failure, do not continue * * lvmlockd could also return the lock_type that it used for the VG, * and we could encode that in lockd_state, and verify later that it * matches vg->lock_type. * * The result of the VG lock operation needs to be saved in lockd_state * because the result needs to be passed into vg_read so it can be * assessed in combination with vg->lock_state. * * The VG lock protects the VG metadata on disk from concurrent access * among hosts. The VG lock also ensures that the local lvmetad cache * contains the latest version of the VG metadata from disk. (Since * another host may have changed the VG since it was last read.) * * The VG lock must be acquired before the VG is read, i.e. before vg_read(). * The result from lockd_vg() is saved in the "lockd_state" variable, and * this result is passed into vg_read(). After vg_read() reads the VG, * it checks if the VG lock_type (sanlock or dlm) requires a lock to be * held, and if so, it verifies that the lock was correctly acquired by * looking at lockd_state. If vg_read() sees that the VG is a local VG, * i.e. lock_type is not sanlock or dlm, then no lock is required, and it * ignores lockd_state (which would indicate no lock was found.) * * When acquiring the VG lock, lvmlockd checks if the local cached copy * of the VG metadata in lvmetad is up to date. If not, it invalidates * the VG cached in lvmetad. This would happen if another host changed * the VG since it was last read. When lvm commands read the VG from * lvmetad, they will check if the metadata is invalid, and if so they * will reread it from disk, and update the copy in lvmetad. */ int lockd_vg(struct cmd_context *cmd, const char *vg_name, const char *def_mode, uint32_t flags, uint32_t *lockd_state) { const char *mode = NULL; uint32_t lockd_flags; uint32_t prev_state = *lockd_state; int retries = 0; int result; int ret; *lockd_state = 0; if (!is_real_vg(vg_name)) return 1; /* * Some special cases need to disable the vg lock. */ if (cmd->lockd_vg_disable) return 1; /* * An unlock is simply sent or skipped without any need * for the mode checking for sh/ex. * * Look at lockd_state from the sh/ex lock, and if it failed, * don't bother sending the unlock to lvmlockd. The main * purpose of this is to avoid sending an unnecessary unlock * for local VGs (the lockd_state from sh/ex on the local VG * will be failed.) This implies that the lockd_state value * should be preserved from the sh/ex lockd_vg() call and * passed back to lockd_vg() for the corresponding unlock. */ if (def_mode && !strcmp(def_mode, "un")) { if (prev_state & LDST_FAIL) { log_debug("VG %s unlock skipped: lockd_state is failed", vg_name); return 1; } mode = "un"; goto req; } /* * The default mode may not have been provided in the * function args. This happens when lockd_vg is called * from a process_each function that handles different * commands. Commands that only read/check/report/display * the vg have LOCKD_VG_SH set in commands.h, which is * copied to lockd_vg_default_sh. Commands without this * set modify the vg and need ex. */ if (!mode) mode = def_mode; if (!mode) mode = cmd->lockd_vg_default_sh ? "sh" : "ex"; if (!strcmp(mode, "ex")) *lockd_state |= LDST_EX; req: /* * This check is not at the top of the function so that * we can first set LDST_EX which will be used later to * decide whether a failure can be ignored or not. * * We do not know if this is a local VG or lockd VG yet, * so we must return success, go ahead and read the VG, * then check if the lock_type required lvmlockd or not. */ if (!_use_lvmlockd) { *lockd_state |= LDST_FAIL_REQUEST; return 1; } log_debug("lockd VG %s mode %s", vg_name, mode); if (!_lockd_request(cmd, "lock_vg", vg_name, NULL, NULL, NULL, NULL, NULL, mode, NULL, &result, &lockd_flags)) { /* * No result from lvmlockd, it is probably not running. * Decide if it is ok to continue without a lock in * access_vg_lock_type() after the VG has been read and * the lock_type can be checked. We don't care about * this error for local VGs, but we do care for lockd VGs. */ *lockd_state |= LDST_FAIL_REQUEST; return 1; } if (result == -EAGAIN) { if (retries < find_config_tree_int(cmd, global_lvmlockd_lock_retries_CFG, NULL)) { log_warn("Retrying %s lock on VG %s", mode, vg_name); sleep(1); retries++; goto req; } } switch (result) { case 0: /* success */ break; case -ENOLS: *lockd_state |= LDST_FAIL_NOLS; break; case -ESTARTING: *lockd_state |= LDST_FAIL_STARTING; break; default: *lockd_state |= LDST_FAIL_OTHER; } /* * Normal success. */ if (!result) { ret = 1; goto out; } /* * The lockspace for the VG is starting (the VG must not * be local), and is not yet ready to do locking. Allow * reading without a sh lock during this period. */ if (result == -ESTARTING) { if (!strcmp(mode, "un")) { ret = 1; goto out; } else if (!strcmp(mode, "sh")) { log_warn("VG %s lock skipped: lock start in progress", vg_name); ret = 1; goto out; } else { log_error("VG %s lock failed: lock start in progress", vg_name); ret = 0; goto out; } } /* * An unused/previous lockspace for the VG was found. * This means it must be a lockd VG, not local. The * lockspace needs to be started to be used. */ if ((result == -ENOLS) && (lockd_flags & LD_RF_INACTIVE_LS)) { if (!strcmp(mode, "un")) { ret = 1; goto out; } else if (!strcmp(mode, "sh")) { log_warn("VG %s lock skipped: lockspace is inactive", vg_name); ret = 1; goto out; } else { log_error("VG %s lock failed: lockspace is inactive", vg_name); ret = 0; goto out; } } /* * An unused lockspace for the VG was found. The previous * start of the lockspace failed, so we can print a more useful * error message. */ if ((result == -ENOLS) && (lockd_flags & LD_RF_ADD_LS_ERROR)) { if (!strcmp(mode, "un")) { ret = 1; goto out; } else if (!strcmp(mode, "sh")) { log_warn("VG %s lock skipped: lockspace start error", vg_name); ret = 1; goto out; } else { log_error("VG %s lock failed: lockspace start error", vg_name); ret = 0; goto out; } } /* * No lockspace for the VG was found. It may be a local * VG that lvmlockd doesn't keep track of, or it may be * a lockd VG that lvmlockd doesn't yet know about (it hasn't * been started yet.) Decide what to do after the VG is * read and we can see the lock_type. */ if (result == -ENOLS) { ret = 1; goto out; } /* * Another error. We don't intend to reach here, but * want to check for each specific error above so that * a helpful message can be printed. */ if (result) { if (!strcmp(mode, "un")) { ret = 1; goto out; } else if (!strcmp(mode, "sh")) { log_warn("VG %s lock skipped: error %d", vg_name, result); ret = 1; goto out; } else { log_error("VG %s lock failed: error %d", vg_name, result); ret = 0; goto out; } } out: /* * A notice from lvmlockd that duplicate gl locks have been found. * It would be good for the user to disable one of them. */ if ((lockd_flags & LD_RF_DUP_GL_LS) && strcmp(mode, "un")) log_warn("Duplicate sanlock global lock in VG %s", vg_name); if (!ret && ignorelockingfailure()) { log_debug("Ignore failed locking for VG %s", vg_name); return 1; } return ret; } /* * This must be called before a new version of the VG metadata is * written to disk. For local VGs, this is a no-op, but for lockd * VGs, this notifies lvmlockd of the new VG seqno. lvmlockd must * know the latest VG seqno so that it can save it within the lock's * LVB. The VG seqno in the VG lock's LVB is used by other hosts to * detect when their cached copy of the VG metadata is stale, i.e. * the cached VG metadata has a lower seqno than the seqno seen in * the VG lock. */ int lockd_vg_update(struct volume_group *vg) { daemon_reply reply; int result; int ret; if (!is_lockd_type(vg->lock_type)) return 1; if (!_use_lvmlockd) return 0; if (!_lvmlockd_connected) return 0; reply = _lockd_send("vg_update", "pid = %d", getpid(), "vg_name = %s", vg->name, "version = %d", (int64_t)vg->seqno, NULL); if (!_lockd_result(reply, &result, NULL)) { ret = 0; } else { ret = (result < 0) ? 0 : 1; } daemon_reply_destroy(reply); return ret; } /* * When this is called directly (as opposed to being called from * lockd_lv), the caller knows that the LV has a lock. */ int lockd_lv_name(struct cmd_context *cmd, struct volume_group *vg, const char *lv_name, struct id *lv_id, const char *lock_args, const char *def_mode, uint32_t flags) { char lv_uuid[64] __attribute__((aligned(8))); const char *mode = NULL; const char *opts = NULL; uint32_t lockd_flags; int refreshed = 0; int result; if (cmd->lockd_lv_disable) return 1; if (!_use_lvmlockd) return 0; if (!_lvmlockd_connected) return 0; if (!id_write_format(lv_id, lv_uuid, sizeof(lv_uuid))) return_0; /* * For lvchange/vgchange activation, def_mode is "sh" or "ex" * according to the specific -a{e,s}y mode designation. * No e,s designation gives NULL def_mode. */ if (def_mode) mode = def_mode; if (mode && !strcmp(mode, "sh") && (flags & LDLV_MODE_NO_SH)) { log_error("Shared activation not compatible with LV type: %s/%s", vg->name, lv_name); return 0; } if (!mode) mode = "ex"; if (flags & LDLV_PERSISTENT) opts = "persistent"; retry: log_debug("lockd LV %s/%s mode %s uuid %s", vg->name, lv_name, mode, lv_uuid); if (!_lockd_request(cmd, "lock_lv", vg->name, vg->lock_type, vg->lock_args, lv_name, lv_uuid, lock_args, mode, opts, &result, &lockd_flags)) { /* No result from lvmlockd, it is probably not running. */ log_error("Locking failed for LV %s/%s", vg->name, lv_name); return 0; } /* The lv was not active/locked. */ if (result == -ENOENT && !strcmp(mode, "un")) return 1; if (result == -EALREADY) return 1; if (result == -EAGAIN) { log_error("LV locked by other host: %s/%s", vg->name, lv_name); return 0; } if (result == -EMSGSIZE) { /* Another host probably extended lvmlock. */ if (!refreshed++) { log_debug("Refresh lvmlock"); _refresh_sanlock_lv(cmd, vg); goto retry; } } if (result == -ENOLS) { log_error("LV %s/%s lock failed: lockspace is inactive", vg->name, lv_name); return 0; } if (result < 0) { log_error("LV %s/%s lock failed: error %d", vg->name, lv_name, result); return 0; } return 1; } /* * Direct the lock request to the pool LV. * For a thin pool and all its thin volumes, one ex lock is used. * It is the one specified in metadata of the pool data lv. */ static int _lockd_lv_thin(struct cmd_context *cmd, struct logical_volume *lv, const char *def_mode, uint32_t flags) { struct logical_volume *pool_lv; if (lv_is_thin_volume(lv)) { struct lv_segment *pool_seg = first_seg(lv); pool_lv = pool_seg ? pool_seg->pool_lv : NULL; } else if (lv_is_thin_pool(lv)) { pool_lv = lv; } else { /* This should not happen AFAIK. */ log_error("Lock on incorrect thin lv type %s/%s", lv->vg->name, lv->name); return 0; } if (!pool_lv) { /* This should not happen. */ log_error("Cannot find thin pool for %s/%s", lv->vg->name, lv->name); return 0; } /* * Locking a locked lv (pool in this case) is a no-op. * Unlock when the pool is no longer active. */ if (def_mode && !strcmp(def_mode, "un") && pool_is_active(pool_lv)) return 1; flags |= LDLV_MODE_NO_SH; return lockd_lv_name(cmd, pool_lv->vg, pool_lv->name, &pool_lv->lvid.id[1], pool_lv->lock_args, def_mode, flags); } /* * If the VG has no lock_type, then this function can return immediately. * The LV itself may have no lock (NULL lv->lock_args), but the lock request * may be directed to another lock, e.g. the pool LV lock in _lockd_lv_thin. * If the lock request is not directed to another LV, and the LV has no * lock_type set, it means that the LV has no lock, and no locking is done * for it. * * An LV lock is acquired before the LV is activated, and released * after the LV is deactivated. If the LV lock cannot be acquired, * it means that the LV is active on another host and the activation * fails. Commands that modify an inactive LV also acquire the LV lock. * * In non-lockd VGs, this is a no-op. * * In lockd VGs, normal LVs each have their own lock, but other * LVs do not have their own lock, e.g. the lock for a thin LV is * acquired on the thin pool LV, and a thin LV does not have a lock * of its own. A cache pool LV does not have a lock of its own. * When the cache pool LV is linked to an origin LV, the lock of * the orgin LV protects the combined origin + cache pool. */ int lockd_lv(struct cmd_context *cmd, struct logical_volume *lv, const char *def_mode, uint32_t flags) { if (!is_lockd_type(lv->vg->lock_type)) return 1; if (!_use_lvmlockd) { log_error("LV in VG %s with lock_type %s requires lvmlockd.", lv->vg->name, lv->vg->lock_type); return 0; } if (!_lvmlockd_connected) return 0; if (lv_is_thin_type(lv)) return _lockd_lv_thin(cmd, lv, def_mode, flags); /* * An LV with NULL lock_args does not have a lock of its own. */ if (!lv->lock_args) return 1; /* * LV type cannot be active concurrently on multiple hosts, * so shared mode activation is not allowed. */ if (lv_is_external_origin(lv) || lv_is_thin_type(lv) || lv_is_mirror_type(lv) || lv_is_raid_type(lv) || lv_is_cache_type(lv)) { flags |= LDLV_MODE_NO_SH; } return lockd_lv_name(cmd, lv->vg, lv->name, &lv->lvid.id[1], lv->lock_args, def_mode, flags); } static int _init_lv_sanlock(struct cmd_context *cmd, struct volume_group *vg, const char *lv_name, struct id *lv_id, const char **lock_args_ret) { char lv_uuid[64] __attribute__((aligned(8))); daemon_reply reply; const char *reply_str; const char *lv_lock_args = NULL; int result; int ret; if (!_use_lvmlockd) return 0; if (!_lvmlockd_connected) return 0; if (!id_write_format(lv_id, lv_uuid, sizeof(lv_uuid))) return_0; reply = _lockd_send("init_lv", "pid = %d", getpid(), "vg_name = %s", vg->name, "lv_name = %s", lv_name, "lv_uuid = %s", lv_uuid, "vg_lock_type = %s", "sanlock", "vg_lock_args = %s", vg->lock_args, NULL); if (!_lockd_result(reply, &result, NULL)) { ret = 0; } else { ret = (result < 0) ? 0 : 1; } if (result == -EEXIST) { log_error("Lock already exists for LV %s/%s", vg->name, lv_name); goto out; } if (result == -EMSGSIZE) { /* * No space on the lvmlock lv for a new lease, this should be * detected by handle_sanlock_lv() called before. */ log_error("No sanlock space for lock for LV %s/%s", vg->name, lv_name); goto out; } if (!ret) { log_error("_init_lv_sanlock lvmlockd result %d", result); goto out; } if (!(reply_str = daemon_reply_str(reply, "lv_lock_args", NULL))) { log_error("lv_lock_args not returned"); ret = 0; goto out; } if (!(lv_lock_args = dm_pool_strdup(cmd->mem, reply_str))) { log_error("lv_lock_args allocation failed"); ret = 0; } out: daemon_reply_destroy(reply); *lock_args_ret = lv_lock_args; return ret; } static int _free_lv(struct cmd_context *cmd, struct volume_group *vg, const char *lv_name, struct id *lv_id, const char *lock_args) { char lv_uuid[64] __attribute__((aligned(8))); daemon_reply reply; int result; int ret; if (!_use_lvmlockd) return 0; if (!_lvmlockd_connected) return 0; if (!id_write_format(lv_id, lv_uuid, sizeof(lv_uuid))) return_0; reply = _lockd_send("free_lv", "pid = %d", getpid(), "vg_name = %s", vg->name, "lv_name = %s", lv_name, "lv_uuid = %s", lv_uuid, "vg_lock_type = %s", vg->lock_type, "vg_lock_args = %s", vg->lock_args, "lv_lock_args = %s", lock_args ?: "none", NULL); if (!_lockd_result(reply, &result, NULL)) { ret = 0; } else { ret = (result < 0) ? 0 : 1; } if (!ret) log_error("_free_lv lvmlockd result %d", result); daemon_reply_destroy(reply); return ret; } int lockd_init_lv_args(struct cmd_context *cmd, struct volume_group *vg, struct logical_volume *lv, const char *lock_type, const char **lock_args) { /* sanlock is the only lock type that sets per-LV lock_args. */ if (!strcmp(lock_type, "sanlock")) return _init_lv_sanlock(cmd, vg, lv->name, &lv->lvid.id[1], lock_args); return 1; } /* * lvcreate * * An LV created in a lockd VG inherits the lock_type of the VG. In some * cases, e.g. thin LVs, this function may decide that the LV should not be * given a lock, in which case it sets lp lock_args to NULL, which will cause * the LV to not have lock_args set in its metadata. A lockd_lv() request on * an LV with no lock_args will do nothing (unless the LV type causes the lock * request to be directed to another LV with a lock, e.g. to the thin pool LV * for thin LVs.) * * Current limitations: * - cache-type LV's in a lockd VG must be created with lvconvert. * - creating a thin pool and thin lv in one command is not allowed. */ int lockd_init_lv(struct cmd_context *cmd, struct volume_group *vg, struct logical_volume *lv, struct lvcreate_params *lp) { int lock_type_num = get_lock_type_from_string(vg->lock_type); switch (lock_type_num) { case LOCK_TYPE_NONE: case LOCK_TYPE_CLVM: return 1; case LOCK_TYPE_SANLOCK: case LOCK_TYPE_DLM: break; default: log_error("lockd_init_lv: unknown lock_type."); return 0; } if (!_use_lvmlockd) return 0; if (!_lvmlockd_connected) return 0; if (!lp->needs_lockd_init) { /* needs_lock_init is set for LVs that need a lockd lock. */ return 1; } else if (seg_is_cache(lp) || seg_is_cache_pool(lp)) { log_error("Use lvconvert for cache with lock type %s", vg->lock_type); return 0; } else if (!seg_is_thin_volume(lp) && lp->snapshot) { struct logical_volume *origin_lv; /* * COW snapshots are associated with their origin LV, * and only the origin LV needs its own lock, which * represents itself and all associated cow snapshots. */ if (!(origin_lv = find_lv(vg, lp->origin_name))) { log_error("Failed to find origin LV %s/%s", vg->name, lp->origin_name); return 0; } if (!lockd_lv(cmd, origin_lv, "ex", LDLV_PERSISTENT)) { log_error("Failed to lock origin LV %s/%s", vg->name, lp->origin_name); return 0; } lv->lock_args = NULL; return 1; } else if (seg_is_thin(lp)) { if ((seg_is_thin_volume(lp) && !lp->create_pool) || (!seg_is_thin_volume(lp) && lp->snapshot)) { struct lv_list *lvl; /* * Creating a new thin lv or snapshot. These lvs do not get * their own lock but use the pool lock. If an lv does not * use its own lock, its lock_args is set to NULL. */ if (!(lvl = find_lv_in_vg(vg, lp->pool_name))) { log_error("Failed to find thin pool %s/%s", vg->name, lp->pool_name); return 0; } if (!lockd_lv(cmd, lvl->lv, "ex", LDLV_PERSISTENT)) { log_error("Failed to lock thin pool %s/%s", vg->name, lp->pool_name); return 0; } lv->lock_args = NULL; return 1; } else if (seg_is_thin_volume(lp) && lp->create_pool) { /* * Creating a thin pool and a thin lv in it. We could * probably make this work. */ log_error("Create thin pool and thin LV separately with lock type %s", vg->lock_type); return 0; } else if (!seg_is_thin_volume(lp) && lp->create_pool) { /* Creating a thin pool only. */ /* lv_name_lock = lp->pool_name; */ } else { log_error("Unknown thin options for lock init."); return 0; } } else { /* Creating a normal lv. */ /* lv_name_lock = lv_name; */ } /* * The LV gets its own lock, so set lock_args to non-NULL. * * lockd_init_lv_args() will be called during vg_write() * to complete the sanlock LV lock initialization, where * actual space on disk is allocated. Waiting to do this * last step until vg_write() avoids the need to revert * the sanlock allocation if the lvcreate function isn't * completed. * * This works, but would leave the sanlock lease allocated * unless the lease was freed on each early exit path from * lvcreate: * * return lockd_init_lv_args(cmd, vg, lv_name_lock, lv_id, * vg->lock_type, &lv->lock_args); */ if (!strcmp(vg->lock_type, "sanlock")) lv->lock_args = "pending"; else if (!strcmp(vg->lock_type, "dlm")) lv->lock_args = "dlm"; return 1; } /* lvremove */ int lockd_free_lv(struct cmd_context *cmd, struct volume_group *vg, const char *lv_name, struct id *lv_id, const char *lock_args) { switch (get_lock_type_from_string(vg->lock_type)) { case LOCK_TYPE_NONE: case LOCK_TYPE_CLVM: return 1; case LOCK_TYPE_DLM: case LOCK_TYPE_SANLOCK: if (!lock_args) return 1; return _free_lv(cmd, vg, lv_name, lv_id, lock_args); default: log_error("lockd_free_lv: unknown lock_type."); return 0; } } int lockd_rename_vg_before(struct cmd_context *cmd, struct volume_group *vg) { struct lv_list *lvl; daemon_reply reply; int result; int ret; if (!is_lockd_type(vg->lock_type)) return 1; if (!_use_lvmlockd) return 0; if (!_lvmlockd_connected) return 0; if (lvs_in_vg_activated(vg)) { log_error("LVs must be inactive before vgrename."); return 0; } /* Check that no LVs are active on other hosts. */ dm_list_iterate_items(lvl, &vg->lvs) { if (!lockd_lv(cmd, lvl->lv, "ex", 0)) { log_error("LV %s/%s must be inactive on all hosts before vgrename.", vg->name, lvl->lv->name); return 0; } if (!lockd_lv(cmd, lvl->lv, "un", 0)) { log_error("Failed to unlock LV %s/%s.", vg->name, lvl->lv->name); return 0; } } /* * lvmlockd: * checks for other hosts in lockspace * leaves the lockspace */ reply = _lockd_send("rename_vg_before", "pid = %d", getpid(), "vg_name = %s", vg->name, "vg_lock_type = %s", vg->lock_type, "vg_lock_args = %s", vg->lock_args, NULL); if (!_lockd_result(reply, &result, NULL)) { ret = 0; } else { ret = (result < 0) ? 0 : 1; } daemon_reply_destroy(reply); if (!ret) { log_error("lockd_rename_vg_before lvmlockd result %d", result); return 0; } if (!strcmp(vg->lock_type, "sanlock")) { log_debug("lockd_rename_vg_before deactivate sanlock lv"); _deactivate_sanlock_lv(cmd, vg); } return 1; } int lockd_rename_vg_final(struct cmd_context *cmd, struct volume_group *vg, int success) { daemon_reply reply; int result; int ret; if (!is_lockd_type(vg->lock_type)) return 1; if (!_use_lvmlockd) return 0; if (!_lvmlockd_connected) return 0; if (!success) { /* * Depending on the problem that caused the rename to * fail, it may make sense to not restart the VG here. */ if (!lockd_start_vg(cmd, vg)) log_error("Failed to restart VG %s lockspace.", vg->name); return 1; } if (!strcmp(vg->lock_type, "sanlock")) { if (!_activate_sanlock_lv(cmd, vg)) return 0; /* * lvmlockd needs to rewrite the leases on disk * with the new VG (lockspace) name. */ reply = _lockd_send("rename_vg_final", "pid = %d", getpid(), "vg_name = %s", vg->name, "vg_lock_type = %s", vg->lock_type, "vg_lock_args = %s", vg->lock_args, NULL); if (!_lockd_result(reply, &result, NULL)) { ret = 0; } else { ret = (result < 0) ? 0 : 1; } daemon_reply_destroy(reply); if (!ret) { /* * The VG has been renamed on disk, but renaming the * sanlock leases failed. Cleaning this up can * probably be done by converting the VG to lock_type * none, then converting back to sanlock. */ log_error("lockd_rename_vg_final lvmlockd result %d", result); return 0; } } if (!lockd_start_vg(cmd, vg)) log_error("Failed to start VG %s lockspace.", vg->name); return 1; } const char *lockd_running_lock_type(struct cmd_context *cmd) { daemon_reply reply; const char *lock_type = NULL; int result; if (!_use_lvmlockd) return NULL; if (!_lvmlockd_connected) return NULL; reply = _lockd_send("running_lm", "pid = %d", getpid(), NULL); if (!_lockd_result(reply, &result, NULL)) { log_error("Failed to get result from lvmlockd"); goto out; } switch (result) { case -EXFULL: log_error("lvmlockd found multiple lock managers, use --lock-type to select one."); break; case -ENOLCK: log_error("lvmlockd found no lock manager running."); break; case LOCK_TYPE_SANLOCK: log_debug("lvmlockd found sanlock"); lock_type = "sanlock"; break; case LOCK_TYPE_DLM: log_debug("lvmlockd found dlm"); lock_type = "dlm"; break; default: log_error("Failed to find a running lock manager."); break; } out: daemon_reply_destroy(reply); return lock_type; } /* Some LV types have no lock. */ int lockd_lv_uses_lock(struct logical_volume *lv) { if (!lv_is_visible(lv) || lv_is_thin_volume(lv) || lv_is_thin_pool_data(lv) || lv_is_thin_pool_metadata(lv) || lv_is_pool_metadata_spare(lv) || lv_is_cache_pool(lv) || lv_is_cache_pool_data(lv) || lv_is_cache_pool_metadata(lv) || lv_is_lockd_sanlock_lv(lv)) return 0; return 1; }